2 * IPv6 tunneling device
3 * Linux INET6 implementation
6 * Ville Nuorvala <vnuorval@tcs.hut.fi>
7 * Yasuyuki Kozakai <kozakai@linux-ipv6.org>
10 * linux/net/ipv6/sit.c and linux/net/ipv4/ipip.c
14 * This program is free software; you can redistribute it and/or
15 * modify it under the terms of the GNU General Public License
16 * as published by the Free Software Foundation; either version
17 * 2 of the License, or (at your option) any later version.
21 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
23 #include <linux/module.h>
24 #include <linux/capability.h>
25 #include <linux/errno.h>
26 #include <linux/types.h>
27 #include <linux/sockios.h>
28 #include <linux/icmp.h>
32 #include <linux/net.h>
33 #include <linux/in6.h>
34 #include <linux/netdevice.h>
35 #include <linux/if_arp.h>
36 #include <linux/icmpv6.h>
37 #include <linux/init.h>
38 #include <linux/route.h>
39 #include <linux/rtnetlink.h>
40 #include <linux/netfilter_ipv6.h>
41 #include <linux/slab.h>
42 #include <linux/hash.h>
43 #include <linux/etherdevice.h>
45 #include <asm/uaccess.h>
46 #include <linux/atomic.h>
50 #include <net/ip_tunnels.h>
52 #include <net/ip6_route.h>
53 #include <net/addrconf.h>
54 #include <net/ip6_tunnel.h>
56 #include <net/dsfield.h>
57 #include <net/inet_ecn.h>
58 #include <net/net_namespace.h>
59 #include <net/netns/generic.h>
60 #include <net/dst_metadata.h>
62 MODULE_AUTHOR("Ville Nuorvala");
63 MODULE_DESCRIPTION("IPv6 tunneling device");
64 MODULE_LICENSE("GPL");
65 MODULE_ALIAS_RTNL_LINK("ip6tnl");
66 MODULE_ALIAS_NETDEV("ip6tnl0");
68 #define IP6_TUNNEL_HASH_SIZE_SHIFT 5
69 #define IP6_TUNNEL_HASH_SIZE (1 << IP6_TUNNEL_HASH_SIZE_SHIFT)
71 static bool log_ecn_error = true;
72 module_param(log_ecn_error, bool, 0644);
73 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
75 static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2)
77 u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2);
79 return hash_32(hash, IP6_TUNNEL_HASH_SIZE_SHIFT);
82 static int ip6_tnl_dev_init(struct net_device *dev);
83 static void ip6_tnl_dev_setup(struct net_device *dev);
84 static struct rtnl_link_ops ip6_link_ops __read_mostly;
86 static int ip6_tnl_net_id __read_mostly;
88 /* the IPv6 tunnel fallback device */
89 struct net_device *fb_tnl_dev;
90 /* lists for storing tunnels in use */
91 struct ip6_tnl __rcu *tnls_r_l[IP6_TUNNEL_HASH_SIZE];
92 struct ip6_tnl __rcu *tnls_wc[1];
93 struct ip6_tnl __rcu **tnls[2];
94 struct ip6_tnl __rcu *collect_md_tun;
97 static struct net_device_stats *ip6_get_stats(struct net_device *dev)
99 struct pcpu_sw_netstats tmp, sum = { 0 };
102 for_each_possible_cpu(i) {
104 const struct pcpu_sw_netstats *tstats =
105 per_cpu_ptr(dev->tstats, i);
108 start = u64_stats_fetch_begin_irq(&tstats->syncp);
109 tmp.rx_packets = tstats->rx_packets;
110 tmp.rx_bytes = tstats->rx_bytes;
111 tmp.tx_packets = tstats->tx_packets;
112 tmp.tx_bytes = tstats->tx_bytes;
113 } while (u64_stats_fetch_retry_irq(&tstats->syncp, start));
115 sum.rx_packets += tmp.rx_packets;
116 sum.rx_bytes += tmp.rx_bytes;
117 sum.tx_packets += tmp.tx_packets;
118 sum.tx_bytes += tmp.tx_bytes;
120 dev->stats.rx_packets = sum.rx_packets;
121 dev->stats.rx_bytes = sum.rx_bytes;
122 dev->stats.tx_packets = sum.tx_packets;
123 dev->stats.tx_bytes = sum.tx_bytes;
128 * ip6_tnl_lookup - fetch tunnel matching the end-point addresses
129 * @remote: the address of the tunnel exit-point
130 * @local: the address of the tunnel entry-point
133 * tunnel matching given end-points if found,
134 * else fallback tunnel if its device is up,
138 #define for_each_ip6_tunnel_rcu(start) \
139 for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
141 static struct ip6_tnl *
142 ip6_tnl_lookup(struct net *net, const struct in6_addr *remote, const struct in6_addr *local)
144 unsigned int hash = HASH(remote, local);
146 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
149 for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
150 if (ipv6_addr_equal(local, &t->parms.laddr) &&
151 ipv6_addr_equal(remote, &t->parms.raddr) &&
152 (t->dev->flags & IFF_UP))
156 memset(&any, 0, sizeof(any));
157 hash = HASH(&any, local);
158 for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
159 if (ipv6_addr_equal(local, &t->parms.laddr) &&
160 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);
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);
260 static int ip6_tnl_create2(struct net_device *dev)
262 struct ip6_tnl *t = netdev_priv(dev);
263 struct net *net = dev_net(dev);
264 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
267 t = netdev_priv(dev);
269 dev->rtnl_link_ops = &ip6_link_ops;
270 err = register_netdevice(dev);
274 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;
480 u8 rel_type = ICMPV6_DEST_UNREACH;
481 u8 rel_code = ICMPV6_ADDR_UNREACH;
487 /* If the packet doesn't contain the original IPv6 header we are
488 in trouble since we might need the source address for further
489 processing of the error. */
492 t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr, &ipv6h->saddr);
496 tproto = ACCESS_ONCE(t->parms.proto);
497 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 mtu = *info - offset;
537 if (mtu < IPV6_MIN_MTU)
541 len = sizeof(*ipv6h) + ntohs(ipv6h->payload_len);
543 rel_type = ICMPV6_PKT_TOOBIG;
562 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
563 u8 type, u8 code, int offset, __be32 info)
568 __u32 rel_info = ntohl(info);
570 struct sk_buff *skb2;
571 const struct iphdr *eiph;
575 err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
576 &rel_msg, &rel_info, offset);
584 case ICMPV6_DEST_UNREACH:
585 if (rel_code != ICMPV6_ADDR_UNREACH)
587 rel_type = ICMP_DEST_UNREACH;
588 rel_code = ICMP_HOST_UNREACH;
590 case ICMPV6_PKT_TOOBIG:
593 rel_type = ICMP_DEST_UNREACH;
594 rel_code = ICMP_FRAG_NEEDED;
597 rel_type = ICMP_REDIRECT;
598 rel_code = ICMP_REDIR_HOST;
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,
620 IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
624 skb2->dev = rt->dst.dev;
626 /* route "incoming" packet */
627 if (rt->rt_flags & RTCF_LOCAL) {
630 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
631 eiph->daddr, eiph->saddr,
634 RT_TOS(eiph->tos), 0);
636 rt->dst.dev->type != ARPHRD_TUNNEL6) {
641 skb_dst_set(skb2, &rt->dst);
644 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
646 skb_dst(skb2)->dev->type != ARPHRD_TUNNEL6)
650 /* change mtu on this route */
651 if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
652 if (rel_info > dst_mtu(skb_dst(skb2)))
655 skb_dst(skb2)->ops->update_pmtu(skb_dst(skb2), NULL, skb2, rel_info);
657 if (rel_type == ICMP_REDIRECT)
658 skb_dst(skb2)->ops->redirect(skb_dst(skb2), NULL, skb2);
660 icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
668 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
669 u8 type, u8 code, int offset, __be32 info)
674 __u32 rel_info = ntohl(info);
677 err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
678 &rel_msg, &rel_info, offset);
682 if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
684 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
690 skb_pull(skb2, offset);
691 skb_reset_network_header(skb2);
693 /* Try to guess incoming interface */
694 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
697 if (rt && rt->dst.dev)
698 skb2->dev = rt->dst.dev;
700 icmpv6_send(skb2, rel_type, rel_code, rel_info);
710 static int ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
711 const struct ipv6hdr *ipv6h,
714 __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
716 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
717 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
719 return IP6_ECN_decapsulate(ipv6h, skb);
722 static int ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
723 const struct ipv6hdr *ipv6h,
726 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
727 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
729 return IP6_ECN_decapsulate(ipv6h, skb);
732 __u32 ip6_tnl_get_cap(struct ip6_tnl *t,
733 const struct in6_addr *laddr,
734 const struct in6_addr *raddr)
736 struct __ip6_tnl_parm *p = &t->parms;
737 int ltype = ipv6_addr_type(laddr);
738 int rtype = ipv6_addr_type(raddr);
741 if (ltype == IPV6_ADDR_ANY || rtype == IPV6_ADDR_ANY) {
742 flags = IP6_TNL_F_CAP_PER_PACKET;
743 } else if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
744 rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
745 !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
746 (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
747 if (ltype&IPV6_ADDR_UNICAST)
748 flags |= IP6_TNL_F_CAP_XMIT;
749 if (rtype&IPV6_ADDR_UNICAST)
750 flags |= IP6_TNL_F_CAP_RCV;
754 EXPORT_SYMBOL(ip6_tnl_get_cap);
756 /* called with rcu_read_lock() */
757 int ip6_tnl_rcv_ctl(struct ip6_tnl *t,
758 const struct in6_addr *laddr,
759 const struct in6_addr *raddr)
761 struct __ip6_tnl_parm *p = &t->parms;
763 struct net *net = t->net;
765 if ((p->flags & IP6_TNL_F_CAP_RCV) ||
766 ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
767 (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_RCV))) {
768 struct net_device *ldev = NULL;
771 ldev = dev_get_by_index_rcu(net, p->link);
773 if ((ipv6_addr_is_multicast(laddr) ||
774 likely(ipv6_chk_addr(net, laddr, ldev, 0))) &&
775 likely(!ipv6_chk_addr(net, raddr, NULL, 0)))
780 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl);
782 static int __ip6_tnl_rcv(struct ip6_tnl *tunnel, struct sk_buff *skb,
783 const struct tnl_ptk_info *tpi,
784 struct metadata_dst *tun_dst,
785 int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
786 const struct ipv6hdr *ipv6h,
787 struct sk_buff *skb),
790 struct pcpu_sw_netstats *tstats;
791 const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
794 if ((!(tpi->flags & TUNNEL_CSUM) &&
795 (tunnel->parms.i_flags & TUNNEL_CSUM)) ||
796 ((tpi->flags & TUNNEL_CSUM) &&
797 !(tunnel->parms.i_flags & TUNNEL_CSUM))) {
798 tunnel->dev->stats.rx_crc_errors++;
799 tunnel->dev->stats.rx_errors++;
803 if (tunnel->parms.i_flags & TUNNEL_SEQ) {
804 if (!(tpi->flags & TUNNEL_SEQ) ||
806 (s32)(ntohl(tpi->seq) - tunnel->i_seqno) < 0)) {
807 tunnel->dev->stats.rx_fifo_errors++;
808 tunnel->dev->stats.rx_errors++;
811 tunnel->i_seqno = ntohl(tpi->seq) + 1;
814 skb->protocol = tpi->proto;
816 /* Warning: All skb pointers will be invalidated! */
817 if (tunnel->dev->type == ARPHRD_ETHER) {
818 if (!pskb_may_pull(skb, ETH_HLEN)) {
819 tunnel->dev->stats.rx_length_errors++;
820 tunnel->dev->stats.rx_errors++;
824 ipv6h = ipv6_hdr(skb);
825 skb->protocol = eth_type_trans(skb, tunnel->dev);
826 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
828 skb->dev = tunnel->dev;
831 skb_reset_network_header(skb);
832 memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
834 __skb_tunnel_rx(skb, tunnel->dev, tunnel->net);
836 err = dscp_ecn_decapsulate(tunnel, ipv6h, skb);
839 net_info_ratelimited("non-ECT from %pI6 with DS=%#x\n",
841 ipv6_get_dsfield(ipv6h));
843 ++tunnel->dev->stats.rx_frame_errors;
844 ++tunnel->dev->stats.rx_errors;
849 tstats = this_cpu_ptr(tunnel->dev->tstats);
850 u64_stats_update_begin(&tstats->syncp);
851 tstats->rx_packets++;
852 tstats->rx_bytes += skb->len;
853 u64_stats_update_end(&tstats->syncp);
855 skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(tunnel->dev)));
858 skb_dst_set(skb, (struct dst_entry *)tun_dst);
860 gro_cells_receive(&tunnel->gro_cells, skb);
868 int ip6_tnl_rcv(struct ip6_tnl *t, struct sk_buff *skb,
869 const struct tnl_ptk_info *tpi,
870 struct metadata_dst *tun_dst,
873 int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
874 const struct ipv6hdr *ipv6h,
875 struct sk_buff *skb);
877 dscp_ecn_decapsulate = ip6ip6_dscp_ecn_decapsulate;
878 if (tpi->proto == htons(ETH_P_IP))
879 dscp_ecn_decapsulate = ip4ip6_dscp_ecn_decapsulate;
881 return __ip6_tnl_rcv(t, skb, tpi, NULL, dscp_ecn_decapsulate,
884 EXPORT_SYMBOL(ip6_tnl_rcv);
886 static const struct tnl_ptk_info tpi_v6 = {
887 /* no tunnel info required for ipxip6. */
888 .proto = htons(ETH_P_IPV6),
891 static const struct tnl_ptk_info tpi_v4 = {
892 /* no tunnel info required for ipxip6. */
893 .proto = htons(ETH_P_IP),
896 static int ipxip6_rcv(struct sk_buff *skb, u8 ipproto,
897 const struct tnl_ptk_info *tpi,
898 int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
899 const struct ipv6hdr *ipv6h,
900 struct sk_buff *skb))
903 const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
904 struct metadata_dst *tun_dst = NULL;
908 t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, &ipv6h->daddr);
911 u8 tproto = ACCESS_ONCE(t->parms.proto);
913 if (tproto != ipproto && tproto != 0)
915 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
917 ipv6h = ipv6_hdr(skb);
918 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr))
920 if (iptunnel_pull_header(skb, 0, tpi->proto, false))
922 if (t->parms.collect_md) {
923 tun_dst = ipv6_tun_rx_dst(skb, 0, 0, 0);
927 ret = __ip6_tnl_rcv(t, skb, tpi, tun_dst, dscp_ecn_decapsulate,
941 static int ip4ip6_rcv(struct sk_buff *skb)
943 return ipxip6_rcv(skb, IPPROTO_IPIP, &tpi_v4,
944 ip4ip6_dscp_ecn_decapsulate);
947 static int ip6ip6_rcv(struct sk_buff *skb)
949 return ipxip6_rcv(skb, IPPROTO_IPV6, &tpi_v6,
950 ip6ip6_dscp_ecn_decapsulate);
953 struct ipv6_tel_txoption {
954 struct ipv6_txoptions ops;
958 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
960 memset(opt, 0, sizeof(struct ipv6_tel_txoption));
962 opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
964 opt->dst_opt[4] = encap_limit;
965 opt->dst_opt[5] = IPV6_TLV_PADN;
968 opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
969 opt->ops.opt_nflen = 8;
973 * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
974 * @t: the outgoing tunnel device
975 * @hdr: IPv6 header from the incoming packet
978 * Avoid trivial tunneling loop by checking that tunnel exit-point
979 * doesn't match source of incoming packet.
987 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
989 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
992 int ip6_tnl_xmit_ctl(struct ip6_tnl *t,
993 const struct in6_addr *laddr,
994 const struct in6_addr *raddr)
996 struct __ip6_tnl_parm *p = &t->parms;
998 struct net *net = t->net;
1000 if ((p->flags & IP6_TNL_F_CAP_XMIT) ||
1001 ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
1002 (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_XMIT))) {
1003 struct net_device *ldev = NULL;
1007 ldev = dev_get_by_index_rcu(net, p->link);
1009 if (unlikely(!ipv6_chk_addr(net, laddr, ldev, 0)))
1010 pr_warn_ratelimited("%s xmit: Local address not yet configured!\n",
1012 else if (!ipv6_addr_is_multicast(raddr) &&
1013 unlikely(ipv6_chk_addr(net, raddr, NULL, 0)))
1014 pr_warn_ratelimited("%s xmit: Routing loop! Remote address found on this node!\n",
1022 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl);
1025 * ip6_tnl_xmit - encapsulate packet and send
1026 * @skb: the outgoing socket buffer
1027 * @dev: the outgoing tunnel device
1028 * @dsfield: dscp code for outer header
1029 * @fl6: flow of tunneled packet
1030 * @encap_limit: encapsulation limit
1031 * @pmtu: Path MTU is stored if packet is too big
1032 * @proto: next header value
1035 * Build new header and do some sanity checks on the packet before sending
1041 * %-EMSGSIZE message too big. return mtu in this case.
1044 int ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev, __u8 dsfield,
1045 struct flowi6 *fl6, int encap_limit, __u32 *pmtu,
1048 struct ip6_tnl *t = netdev_priv(dev);
1049 struct net *net = t->net;
1050 struct net_device_stats *stats = &t->dev->stats;
1051 struct ipv6hdr *ipv6h;
1052 struct ipv6_tel_txoption opt;
1053 struct dst_entry *dst = NULL, *ndst = NULL;
1054 struct net_device *tdev;
1056 unsigned int eth_hlen = t->dev->type == ARPHRD_ETHER ? ETH_HLEN : 0;
1057 unsigned int psh_hlen = sizeof(struct ipv6hdr) + t->encap_hlen;
1058 unsigned int max_headroom = psh_hlen;
1059 bool use_cache = false;
1063 if (t->parms.collect_md) {
1064 hop_limit = skb_tunnel_info(skb)->key.ttl;
1067 hop_limit = t->parms.hop_limit;
1071 if (ipv6_addr_any(&t->parms.raddr)) {
1072 if (skb->protocol == htons(ETH_P_IPV6)) {
1073 struct in6_addr *addr6;
1074 struct neighbour *neigh;
1078 goto tx_err_link_failure;
1080 neigh = dst_neigh_lookup(skb_dst(skb),
1081 &ipv6_hdr(skb)->daddr);
1083 goto tx_err_link_failure;
1085 addr6 = (struct in6_addr *)&neigh->primary_key;
1086 addr_type = ipv6_addr_type(addr6);
1088 if (addr_type == IPV6_ADDR_ANY)
1089 addr6 = &ipv6_hdr(skb)->daddr;
1091 memcpy(&fl6->daddr, addr6, sizeof(fl6->daddr));
1092 neigh_release(neigh);
1094 } else if (t->parms.proto != 0 && !(t->parms.flags &
1095 (IP6_TNL_F_USE_ORIG_TCLASS |
1096 IP6_TNL_F_USE_ORIG_FWMARK))) {
1097 /* enable the cache only if neither the outer protocol nor the
1098 * routing decision depends on the current inner header value
1104 dst = dst_cache_get(&t->dst_cache);
1106 if (!ip6_tnl_xmit_ctl(t, &fl6->saddr, &fl6->daddr))
1107 goto tx_err_link_failure;
1111 /* add dsfield to flowlabel for route lookup */
1112 fl6->flowlabel = ip6_make_flowinfo(dsfield, fl6->flowlabel);
1114 dst = ip6_route_output(net, NULL, fl6);
1117 goto tx_err_link_failure;
1118 dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), NULL, 0);
1122 goto tx_err_link_failure;
1124 if (t->parms.collect_md &&
1125 ipv6_dev_get_saddr(net, ip6_dst_idev(dst)->dev,
1126 &fl6->daddr, 0, &fl6->saddr))
1127 goto tx_err_link_failure;
1134 stats->collisions++;
1135 net_warn_ratelimited("%s: Local routing loop detected!\n",
1137 goto tx_err_dst_release;
1139 mtu = dst_mtu(dst) - eth_hlen - psh_hlen - t->tun_hlen;
1140 if (encap_limit >= 0) {
1144 mtu = max(mtu, skb->protocol == htons(ETH_P_IPV6) ?
1145 IPV6_MIN_MTU : IPV4_MIN_MTU);
1147 if (skb_dst(skb) && !t->parms.collect_md)
1148 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
1149 if (skb->len - t->tun_hlen - eth_hlen > mtu && !skb_is_gso(skb)) {
1152 goto tx_err_dst_release;
1155 if (t->err_count > 0) {
1156 if (time_before(jiffies,
1157 t->err_time + IP6TUNNEL_ERR_TIMEO)) {
1160 dst_link_failure(skb);
1166 skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
1169 * Okay, now see if we can stuff it in the buffer as-is.
1171 max_headroom += LL_RESERVED_SPACE(tdev);
1173 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1174 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1175 struct sk_buff *new_skb;
1177 new_skb = skb_realloc_headroom(skb, max_headroom);
1179 goto tx_err_dst_release;
1182 skb_set_owner_w(new_skb, skb->sk);
1187 if (t->parms.collect_md) {
1188 if (t->encap.type != TUNNEL_ENCAP_NONE)
1189 goto tx_err_dst_release;
1191 if (use_cache && ndst)
1192 dst_cache_set_ip6(&t->dst_cache, ndst, &fl6->saddr);
1194 skb_dst_set(skb, dst);
1196 /* Calculate max headroom for all the headers and adjust
1197 * needed_headroom if necessary.
1199 max_headroom = LL_RESERVED_SPACE(dst->dev) + sizeof(struct ipv6hdr)
1200 + dst->header_len + t->hlen;
1201 if (max_headroom > dev->needed_headroom)
1202 dev->needed_headroom = max_headroom;
1204 err = ip6_tnl_encap(skb, t, &proto, fl6);
1208 if (encap_limit >= 0) {
1209 init_tel_txopt(&opt, encap_limit);
1210 ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
1213 skb_push(skb, sizeof(struct ipv6hdr));
1214 skb_reset_network_header(skb);
1215 ipv6h = ipv6_hdr(skb);
1216 ip6_flow_hdr(ipv6h, dsfield,
1217 ip6_make_flowlabel(net, skb, fl6->flowlabel, true, fl6));
1218 ipv6h->hop_limit = hop_limit;
1219 ipv6h->nexthdr = proto;
1220 ipv6h->saddr = fl6->saddr;
1221 ipv6h->daddr = fl6->daddr;
1222 ip6tunnel_xmit(NULL, skb, dev);
1224 tx_err_link_failure:
1225 stats->tx_carrier_errors++;
1226 dst_link_failure(skb);
1231 EXPORT_SYMBOL(ip6_tnl_xmit);
1234 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1236 struct ip6_tnl *t = netdev_priv(dev);
1237 const struct iphdr *iph;
1238 int encap_limit = -1;
1245 /* ensure we can access the full inner ip header */
1246 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
1250 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1252 tproto = ACCESS_ONCE(t->parms.proto);
1253 if (tproto != IPPROTO_IPIP && tproto != 0)
1256 if (t->parms.collect_md) {
1257 struct ip_tunnel_info *tun_info;
1258 const struct ip_tunnel_key *key;
1260 tun_info = skb_tunnel_info(skb);
1261 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
1262 ip_tunnel_info_af(tun_info) != AF_INET6))
1264 key = &tun_info->key;
1265 memset(&fl6, 0, sizeof(fl6));
1266 fl6.flowi6_proto = IPPROTO_IPIP;
1267 fl6.daddr = key->u.ipv6.dst;
1268 fl6.flowlabel = key->label;
1271 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1272 encap_limit = t->parms.encap_limit;
1274 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1275 fl6.flowi6_proto = IPPROTO_IPIP;
1277 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1278 dsfield = ipv4_get_dsfield(iph);
1280 dsfield = ip6_tclass(t->parms.flowinfo);
1281 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1282 fl6.flowi6_mark = skb->mark;
1285 dsfield = INET_ECN_encapsulate(dsfield, ipv4_get_dsfield(iph));
1287 if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6))
1290 skb_set_inner_ipproto(skb, IPPROTO_IPIP);
1292 err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1295 /* XXX: send ICMP error even if DF is not set. */
1296 if (err == -EMSGSIZE)
1297 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
1306 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1308 struct ip6_tnl *t = netdev_priv(dev);
1309 struct ipv6hdr *ipv6h;
1310 int encap_limit = -1;
1318 if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
1321 ipv6h = ipv6_hdr(skb);
1322 tproto = ACCESS_ONCE(t->parms.proto);
1323 if ((tproto != IPPROTO_IPV6 && tproto != 0) ||
1324 ip6_tnl_addr_conflict(t, ipv6h))
1327 if (t->parms.collect_md) {
1328 struct ip_tunnel_info *tun_info;
1329 const struct ip_tunnel_key *key;
1331 tun_info = skb_tunnel_info(skb);
1332 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
1333 ip_tunnel_info_af(tun_info) != AF_INET6))
1335 key = &tun_info->key;
1336 memset(&fl6, 0, sizeof(fl6));
1337 fl6.flowi6_proto = IPPROTO_IPV6;
1338 fl6.daddr = key->u.ipv6.dst;
1339 fl6.flowlabel = key->label;
1342 offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
1343 /* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
1344 ipv6h = ipv6_hdr(skb);
1346 struct ipv6_tlv_tnl_enc_lim *tel;
1348 tel = (void *)&skb_network_header(skb)[offset];
1349 if (tel->encap_limit == 0) {
1350 icmpv6_send(skb, ICMPV6_PARAMPROB,
1351 ICMPV6_HDR_FIELD, offset + 2);
1354 encap_limit = tel->encap_limit - 1;
1355 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) {
1356 encap_limit = t->parms.encap_limit;
1359 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1360 fl6.flowi6_proto = IPPROTO_IPV6;
1362 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1363 dsfield = ipv6_get_dsfield(ipv6h);
1365 dsfield = ip6_tclass(t->parms.flowinfo);
1366 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
1367 fl6.flowlabel |= ip6_flowlabel(ipv6h);
1368 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1369 fl6.flowi6_mark = skb->mark;
1372 dsfield = INET_ECN_encapsulate(dsfield, ipv6_get_dsfield(ipv6h));
1374 if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6))
1377 skb_set_inner_ipproto(skb, IPPROTO_IPV6);
1379 err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1382 if (err == -EMSGSIZE)
1383 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1391 ip6_tnl_start_xmit(struct sk_buff *skb, struct net_device *dev)
1393 struct ip6_tnl *t = netdev_priv(dev);
1394 struct net_device_stats *stats = &t->dev->stats;
1397 switch (skb->protocol) {
1398 case htons(ETH_P_IP):
1399 ret = ip4ip6_tnl_xmit(skb, dev);
1401 case htons(ETH_P_IPV6):
1402 ret = ip6ip6_tnl_xmit(skb, dev);
1411 return NETDEV_TX_OK;
1415 stats->tx_dropped++;
1417 return NETDEV_TX_OK;
1420 static void ip6_tnl_link_config(struct ip6_tnl *t)
1422 struct net_device *dev = t->dev;
1423 struct __ip6_tnl_parm *p = &t->parms;
1424 struct flowi6 *fl6 = &t->fl.u.ip6;
1427 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1428 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1430 /* Set up flowi template */
1431 fl6->saddr = p->laddr;
1432 fl6->daddr = p->raddr;
1433 fl6->flowi6_oif = p->link;
1436 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1437 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1438 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1439 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1441 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1442 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1444 if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1445 dev->flags |= IFF_POINTOPOINT;
1447 dev->flags &= ~IFF_POINTOPOINT;
1450 t->hlen = t->encap_hlen + t->tun_hlen;
1451 t_hlen = t->hlen + sizeof(struct ipv6hdr);
1453 if (p->flags & IP6_TNL_F_CAP_XMIT) {
1454 int strict = (ipv6_addr_type(&p->raddr) &
1455 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1457 struct rt6_info *rt = rt6_lookup(t->net,
1458 &p->raddr, &p->laddr,
1465 dev->hard_header_len = rt->dst.dev->hard_header_len +
1468 dev->mtu = rt->dst.dev->mtu - t_hlen;
1469 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1472 if (dev->mtu < IPV6_MIN_MTU)
1473 dev->mtu = IPV6_MIN_MTU;
1480 * ip6_tnl_change - update the tunnel parameters
1481 * @t: tunnel to be changed
1482 * @p: tunnel configuration parameters
1485 * ip6_tnl_change() updates the tunnel parameters
1489 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
1491 t->parms.laddr = p->laddr;
1492 t->parms.raddr = p->raddr;
1493 t->parms.flags = p->flags;
1494 t->parms.hop_limit = p->hop_limit;
1495 t->parms.encap_limit = p->encap_limit;
1496 t->parms.flowinfo = p->flowinfo;
1497 t->parms.link = p->link;
1498 t->parms.proto = p->proto;
1499 dst_cache_reset(&t->dst_cache);
1500 ip6_tnl_link_config(t);
1504 static int ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1506 struct net *net = t->net;
1507 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1510 ip6_tnl_unlink(ip6n, t);
1512 err = ip6_tnl_change(t, p);
1513 ip6_tnl_link(ip6n, t);
1514 netdev_state_change(t->dev);
1518 static int ip6_tnl0_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1520 /* for default tnl0 device allow to change only the proto */
1521 t->parms.proto = p->proto;
1522 netdev_state_change(t->dev);
1527 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u)
1529 p->laddr = u->laddr;
1530 p->raddr = u->raddr;
1531 p->flags = u->flags;
1532 p->hop_limit = u->hop_limit;
1533 p->encap_limit = u->encap_limit;
1534 p->flowinfo = u->flowinfo;
1536 p->proto = u->proto;
1537 memcpy(p->name, u->name, sizeof(u->name));
1541 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p)
1543 u->laddr = p->laddr;
1544 u->raddr = p->raddr;
1545 u->flags = p->flags;
1546 u->hop_limit = p->hop_limit;
1547 u->encap_limit = p->encap_limit;
1548 u->flowinfo = p->flowinfo;
1550 u->proto = p->proto;
1551 memcpy(u->name, p->name, sizeof(u->name));
1555 * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1556 * @dev: virtual device associated with tunnel
1557 * @ifr: parameters passed from userspace
1558 * @cmd: command to be performed
1561 * ip6_tnl_ioctl() is used for managing IPv6 tunnels
1564 * The possible commands are the following:
1565 * %SIOCGETTUNNEL: get tunnel parameters for device
1566 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1567 * %SIOCCHGTUNNEL: change tunnel parameters to those given
1568 * %SIOCDELTUNNEL: delete tunnel
1570 * The fallback device "ip6tnl0", created during module
1571 * initialization, can be used for creating other tunnel devices.
1575 * %-EFAULT if unable to copy data to or from userspace,
1576 * %-EPERM if current process hasn't %CAP_NET_ADMIN set
1577 * %-EINVAL if passed tunnel parameters are invalid,
1578 * %-EEXIST if changing a tunnel's parameters would cause a conflict
1579 * %-ENODEV if attempting to change or delete a nonexisting device
1583 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1586 struct ip6_tnl_parm p;
1587 struct __ip6_tnl_parm p1;
1588 struct ip6_tnl *t = netdev_priv(dev);
1589 struct net *net = t->net;
1590 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1592 memset(&p1, 0, sizeof(p1));
1596 if (dev == ip6n->fb_tnl_dev) {
1597 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1601 ip6_tnl_parm_from_user(&p1, &p);
1602 t = ip6_tnl_locate(net, &p1, 0);
1604 t = netdev_priv(dev);
1606 memset(&p, 0, sizeof(p));
1608 ip6_tnl_parm_to_user(&p, &t->parms);
1609 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) {
1616 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1619 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1622 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1625 ip6_tnl_parm_from_user(&p1, &p);
1626 t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL);
1627 if (cmd == SIOCCHGTUNNEL) {
1629 if (t->dev != dev) {
1634 t = netdev_priv(dev);
1635 if (dev == ip6n->fb_tnl_dev)
1636 err = ip6_tnl0_update(t, &p1);
1638 err = ip6_tnl_update(t, &p1);
1642 ip6_tnl_parm_to_user(&p, &t->parms);
1643 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1652 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1655 if (dev == ip6n->fb_tnl_dev) {
1657 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1660 ip6_tnl_parm_from_user(&p1, &p);
1661 t = ip6_tnl_locate(net, &p1, 0);
1665 if (t->dev == ip6n->fb_tnl_dev)
1670 unregister_netdevice(dev);
1679 * ip6_tnl_change_mtu - change mtu manually for tunnel device
1680 * @dev: virtual device associated with tunnel
1681 * @new_mtu: the new mtu
1685 * %-EINVAL if mtu too small
1688 int ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1690 struct ip6_tnl *tnl = netdev_priv(dev);
1692 if (tnl->parms.proto == IPPROTO_IPIP) {
1696 if (new_mtu < IPV6_MIN_MTU)
1699 if (new_mtu > 0xFFF8 - dev->hard_header_len)
1704 EXPORT_SYMBOL(ip6_tnl_change_mtu);
1706 int ip6_tnl_get_iflink(const struct net_device *dev)
1708 struct ip6_tnl *t = netdev_priv(dev);
1710 return t->parms.link;
1712 EXPORT_SYMBOL(ip6_tnl_get_iflink);
1714 int ip6_tnl_encap_add_ops(const struct ip6_tnl_encap_ops *ops,
1717 if (num >= MAX_IPTUN_ENCAP_OPS)
1720 return !cmpxchg((const struct ip6_tnl_encap_ops **)
1721 &ip6tun_encaps[num],
1722 NULL, ops) ? 0 : -1;
1724 EXPORT_SYMBOL(ip6_tnl_encap_add_ops);
1726 int ip6_tnl_encap_del_ops(const struct ip6_tnl_encap_ops *ops,
1731 if (num >= MAX_IPTUN_ENCAP_OPS)
1734 ret = (cmpxchg((const struct ip6_tnl_encap_ops **)
1735 &ip6tun_encaps[num],
1736 ops, NULL) == ops) ? 0 : -1;
1742 EXPORT_SYMBOL(ip6_tnl_encap_del_ops);
1744 int ip6_tnl_encap_setup(struct ip6_tnl *t,
1745 struct ip_tunnel_encap *ipencap)
1749 memset(&t->encap, 0, sizeof(t->encap));
1751 hlen = ip6_encap_hlen(ipencap);
1755 t->encap.type = ipencap->type;
1756 t->encap.sport = ipencap->sport;
1757 t->encap.dport = ipencap->dport;
1758 t->encap.flags = ipencap->flags;
1760 t->encap_hlen = hlen;
1761 t->hlen = t->encap_hlen + t->tun_hlen;
1765 EXPORT_SYMBOL_GPL(ip6_tnl_encap_setup);
1767 static const struct net_device_ops ip6_tnl_netdev_ops = {
1768 .ndo_init = ip6_tnl_dev_init,
1769 .ndo_uninit = ip6_tnl_dev_uninit,
1770 .ndo_start_xmit = ip6_tnl_start_xmit,
1771 .ndo_do_ioctl = ip6_tnl_ioctl,
1772 .ndo_change_mtu = ip6_tnl_change_mtu,
1773 .ndo_get_stats = ip6_get_stats,
1774 .ndo_get_iflink = ip6_tnl_get_iflink,
1777 #define IPXIPX_FEATURES (NETIF_F_SG | \
1778 NETIF_F_FRAGLIST | \
1780 NETIF_F_GSO_SOFTWARE | \
1784 * ip6_tnl_dev_setup - setup virtual tunnel device
1785 * @dev: virtual device associated with tunnel
1788 * Initialize function pointers and device parameters
1791 static void ip6_tnl_dev_setup(struct net_device *dev)
1793 dev->netdev_ops = &ip6_tnl_netdev_ops;
1794 dev->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))
1851 dst_cache_destroy(&t->dst_cache);
1853 free_percpu(dev->tstats);
1860 * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1861 * @dev: virtual device associated with tunnel
1864 static int ip6_tnl_dev_init(struct net_device *dev)
1866 struct ip6_tnl *t = netdev_priv(dev);
1867 int err = ip6_tnl_dev_init_gen(dev);
1871 ip6_tnl_link_config(t);
1872 if (t->parms.collect_md)
1873 netif_keep_dst(dev);
1878 * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1879 * @dev: fallback device
1884 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1886 struct ip6_tnl *t = netdev_priv(dev);
1887 struct net *net = dev_net(dev);
1888 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1890 t->parms.proto = IPPROTO_IPV6;
1892 rcu_assign_pointer(ip6n->tnls_wc[0], t);
1896 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[])
1900 if (!data || !data[IFLA_IPTUN_PROTO])
1903 proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1904 if (proto != IPPROTO_IPV6 &&
1905 proto != IPPROTO_IPIP &&
1912 static void ip6_tnl_netlink_parms(struct nlattr *data[],
1913 struct __ip6_tnl_parm *parms)
1915 memset(parms, 0, sizeof(*parms));
1920 if (data[IFLA_IPTUN_LINK])
1921 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1923 if (data[IFLA_IPTUN_LOCAL])
1924 parms->laddr = nla_get_in6_addr(data[IFLA_IPTUN_LOCAL]);
1926 if (data[IFLA_IPTUN_REMOTE])
1927 parms->raddr = nla_get_in6_addr(data[IFLA_IPTUN_REMOTE]);
1929 if (data[IFLA_IPTUN_TTL])
1930 parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]);
1932 if (data[IFLA_IPTUN_ENCAP_LIMIT])
1933 parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]);
1935 if (data[IFLA_IPTUN_FLOWINFO])
1936 parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]);
1938 if (data[IFLA_IPTUN_FLAGS])
1939 parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]);
1941 if (data[IFLA_IPTUN_PROTO])
1942 parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1944 if (data[IFLA_IPTUN_COLLECT_METADATA])
1945 parms->collect_md = true;
1948 static bool ip6_tnl_netlink_encap_parms(struct nlattr *data[],
1949 struct ip_tunnel_encap *ipencap)
1953 memset(ipencap, 0, sizeof(*ipencap));
1958 if (data[IFLA_IPTUN_ENCAP_TYPE]) {
1960 ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
1963 if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
1965 ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
1968 if (data[IFLA_IPTUN_ENCAP_SPORT]) {
1970 ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
1973 if (data[IFLA_IPTUN_ENCAP_DPORT]) {
1975 ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
1981 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev,
1982 struct nlattr *tb[], struct nlattr *data[])
1984 struct net *net = dev_net(dev);
1985 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1986 struct ip_tunnel_encap ipencap;
1987 struct ip6_tnl *nt, *t;
1990 nt = netdev_priv(dev);
1992 if (ip6_tnl_netlink_encap_parms(data, &ipencap)) {
1993 err = ip6_tnl_encap_setup(nt, &ipencap);
1998 ip6_tnl_netlink_parms(data, &nt->parms);
2000 if (nt->parms.collect_md) {
2001 if (rtnl_dereference(ip6n->collect_md_tun))
2004 t = ip6_tnl_locate(net, &nt->parms, 0);
2009 err = ip6_tnl_create2(dev);
2010 if (!err && tb[IFLA_MTU])
2011 ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
2016 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[],
2017 struct nlattr *data[])
2019 struct ip6_tnl *t = netdev_priv(dev);
2020 struct __ip6_tnl_parm p;
2021 struct net *net = t->net;
2022 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2023 struct ip_tunnel_encap ipencap;
2025 if (dev == ip6n->fb_tnl_dev)
2028 if (ip6_tnl_netlink_encap_parms(data, &ipencap)) {
2029 int err = ip6_tnl_encap_setup(t, &ipencap);
2034 ip6_tnl_netlink_parms(data, &p);
2038 t = ip6_tnl_locate(net, &p, 0);
2043 t = netdev_priv(dev);
2045 return ip6_tnl_update(t, &p);
2048 static void ip6_tnl_dellink(struct net_device *dev, struct list_head *head)
2050 struct net *net = dev_net(dev);
2051 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2053 if (dev != ip6n->fb_tnl_dev)
2054 unregister_netdevice_queue(dev, head);
2057 static size_t ip6_tnl_get_size(const struct net_device *dev)
2060 /* IFLA_IPTUN_LINK */
2062 /* IFLA_IPTUN_LOCAL */
2063 nla_total_size(sizeof(struct in6_addr)) +
2064 /* IFLA_IPTUN_REMOTE */
2065 nla_total_size(sizeof(struct in6_addr)) +
2066 /* IFLA_IPTUN_TTL */
2068 /* IFLA_IPTUN_ENCAP_LIMIT */
2070 /* IFLA_IPTUN_FLOWINFO */
2072 /* IFLA_IPTUN_FLAGS */
2074 /* IFLA_IPTUN_PROTO */
2076 /* IFLA_IPTUN_ENCAP_TYPE */
2078 /* IFLA_IPTUN_ENCAP_FLAGS */
2080 /* IFLA_IPTUN_ENCAP_SPORT */
2082 /* IFLA_IPTUN_ENCAP_DPORT */
2084 /* IFLA_IPTUN_COLLECT_METADATA */
2089 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev)
2091 struct ip6_tnl *tunnel = netdev_priv(dev);
2092 struct __ip6_tnl_parm *parm = &tunnel->parms;
2094 if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
2095 nla_put_in6_addr(skb, IFLA_IPTUN_LOCAL, &parm->laddr) ||
2096 nla_put_in6_addr(skb, IFLA_IPTUN_REMOTE, &parm->raddr) ||
2097 nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) ||
2098 nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) ||
2099 nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) ||
2100 nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) ||
2101 nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto))
2102 goto nla_put_failure;
2104 if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE, tunnel->encap.type) ||
2105 nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT, tunnel->encap.sport) ||
2106 nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT, tunnel->encap.dport) ||
2107 nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS, tunnel->encap.flags))
2108 goto nla_put_failure;
2110 if (parm->collect_md)
2111 if (nla_put_flag(skb, IFLA_IPTUN_COLLECT_METADATA))
2112 goto nla_put_failure;
2119 struct net *ip6_tnl_get_link_net(const struct net_device *dev)
2121 struct ip6_tnl *tunnel = netdev_priv(dev);
2125 EXPORT_SYMBOL(ip6_tnl_get_link_net);
2127 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = {
2128 [IFLA_IPTUN_LINK] = { .type = NLA_U32 },
2129 [IFLA_IPTUN_LOCAL] = { .len = sizeof(struct in6_addr) },
2130 [IFLA_IPTUN_REMOTE] = { .len = sizeof(struct in6_addr) },
2131 [IFLA_IPTUN_TTL] = { .type = NLA_U8 },
2132 [IFLA_IPTUN_ENCAP_LIMIT] = { .type = NLA_U8 },
2133 [IFLA_IPTUN_FLOWINFO] = { .type = NLA_U32 },
2134 [IFLA_IPTUN_FLAGS] = { .type = NLA_U32 },
2135 [IFLA_IPTUN_PROTO] = { .type = NLA_U8 },
2136 [IFLA_IPTUN_ENCAP_TYPE] = { .type = NLA_U16 },
2137 [IFLA_IPTUN_ENCAP_FLAGS] = { .type = NLA_U16 },
2138 [IFLA_IPTUN_ENCAP_SPORT] = { .type = NLA_U16 },
2139 [IFLA_IPTUN_ENCAP_DPORT] = { .type = NLA_U16 },
2140 [IFLA_IPTUN_COLLECT_METADATA] = { .type = NLA_FLAG },
2143 static struct rtnl_link_ops ip6_link_ops __read_mostly = {
2145 .maxtype = IFLA_IPTUN_MAX,
2146 .policy = ip6_tnl_policy,
2147 .priv_size = sizeof(struct ip6_tnl),
2148 .setup = ip6_tnl_dev_setup,
2149 .validate = ip6_tnl_validate,
2150 .newlink = ip6_tnl_newlink,
2151 .changelink = ip6_tnl_changelink,
2152 .dellink = ip6_tnl_dellink,
2153 .get_size = ip6_tnl_get_size,
2154 .fill_info = ip6_tnl_fill_info,
2155 .get_link_net = ip6_tnl_get_link_net,
2158 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
2159 .handler = ip4ip6_rcv,
2160 .err_handler = ip4ip6_err,
2164 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
2165 .handler = ip6ip6_rcv,
2166 .err_handler = ip6ip6_err,
2170 static void __net_exit ip6_tnl_destroy_tunnels(struct net *net)
2172 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2173 struct net_device *dev, *aux;
2178 for_each_netdev_safe(net, dev, aux)
2179 if (dev->rtnl_link_ops == &ip6_link_ops)
2180 unregister_netdevice_queue(dev, &list);
2182 for (h = 0; h < IP6_TUNNEL_HASH_SIZE; h++) {
2183 t = rtnl_dereference(ip6n->tnls_r_l[h]);
2185 /* If dev is in the same netns, it has already
2186 * been added to the list by the previous loop.
2188 if (!net_eq(dev_net(t->dev), net))
2189 unregister_netdevice_queue(t->dev, &list);
2190 t = rtnl_dereference(t->next);
2194 unregister_netdevice_many(&list);
2197 static int __net_init ip6_tnl_init_net(struct net *net)
2199 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2200 struct ip6_tnl *t = NULL;
2203 ip6n->tnls[0] = ip6n->tnls_wc;
2204 ip6n->tnls[1] = ip6n->tnls_r_l;
2207 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
2208 NET_NAME_UNKNOWN, ip6_tnl_dev_setup);
2210 if (!ip6n->fb_tnl_dev)
2212 dev_net_set(ip6n->fb_tnl_dev, net);
2213 ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops;
2214 /* FB netdevice is special: we have one, and only one per netns.
2215 * Allowing to move it to another netns is clearly unsafe.
2217 ip6n->fb_tnl_dev->features |= NETIF_F_NETNS_LOCAL;
2219 err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
2223 err = register_netdev(ip6n->fb_tnl_dev);
2227 t = netdev_priv(ip6n->fb_tnl_dev);
2229 strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
2233 ip6_dev_free(ip6n->fb_tnl_dev);
2238 static void __net_exit ip6_tnl_exit_net(struct net *net)
2241 ip6_tnl_destroy_tunnels(net);
2245 static struct pernet_operations ip6_tnl_net_ops = {
2246 .init = ip6_tnl_init_net,
2247 .exit = ip6_tnl_exit_net,
2248 .id = &ip6_tnl_net_id,
2249 .size = sizeof(struct ip6_tnl_net),
2253 * ip6_tunnel_init - register protocol and reserve needed resources
2255 * Return: 0 on success
2258 static int __init ip6_tunnel_init(void)
2262 if (!ipv6_mod_enabled())
2265 err = register_pernet_device(&ip6_tnl_net_ops);
2269 err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
2271 pr_err("%s: can't register ip4ip6\n", __func__);
2275 err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
2277 pr_err("%s: can't register ip6ip6\n", __func__);
2280 err = rtnl_link_register(&ip6_link_ops);
2282 goto rtnl_link_failed;
2287 xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
2289 xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
2291 unregister_pernet_device(&ip6_tnl_net_ops);
2297 * ip6_tunnel_cleanup - free resources and unregister protocol
2300 static void __exit ip6_tunnel_cleanup(void)
2302 rtnl_link_unregister(&ip6_link_ops);
2303 if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
2304 pr_info("%s: can't deregister ip4ip6\n", __func__);
2306 if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
2307 pr_info("%s: can't deregister ip6ip6\n", __func__);
2309 unregister_pernet_device(&ip6_tnl_net_ops);
2312 module_init(ip6_tunnel_init);
2313 module_exit(ip6_tunnel_cleanup);