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/if_tunnel.h>
33 #include <linux/net.h>
34 #include <linux/in6.h>
35 #include <linux/netdevice.h>
36 #include <linux/if_arp.h>
37 #include <linux/icmpv6.h>
38 #include <linux/init.h>
39 #include <linux/route.h>
40 #include <linux/rtnetlink.h>
41 #include <linux/netfilter_ipv6.h>
42 #include <linux/slab.h>
43 #include <linux/hash.h>
45 #include <asm/uaccess.h>
46 #include <linux/atomic.h>
51 #include <net/ip6_route.h>
52 #include <net/addrconf.h>
53 #include <net/ip6_tunnel.h>
55 #include <net/dsfield.h>
56 #include <net/inet_ecn.h>
57 #include <net/net_namespace.h>
58 #include <net/netns/generic.h>
60 MODULE_AUTHOR("Ville Nuorvala");
61 MODULE_DESCRIPTION("IPv6 tunneling device");
62 MODULE_LICENSE("GPL");
63 MODULE_ALIAS_NETDEV("ip6tnl0");
66 #define IP6_TNL_TRACE(x...) pr_debug("%s:" x "\n", __func__)
68 #define IP6_TNL_TRACE(x...) do {;} while(0)
71 #define IPV6_TCLASS_MASK (IPV6_FLOWINFO_MASK & ~IPV6_FLOWLABEL_MASK)
72 #define IPV6_TCLASS_SHIFT 20
74 #define HASH_SIZE_SHIFT 5
75 #define HASH_SIZE (1 << HASH_SIZE_SHIFT)
77 static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2)
79 u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2);
81 return hash_32(hash, HASH_SIZE_SHIFT);
84 static int ip6_tnl_dev_init(struct net_device *dev);
85 static void ip6_tnl_dev_setup(struct net_device *dev);
86 static struct rtnl_link_ops ip6_link_ops __read_mostly;
88 static int ip6_tnl_net_id __read_mostly;
90 /* the IPv6 tunnel fallback device */
91 struct net_device *fb_tnl_dev;
92 /* lists for storing tunnels in use */
93 struct ip6_tnl __rcu *tnls_r_l[HASH_SIZE];
94 struct ip6_tnl __rcu *tnls_wc[1];
95 struct ip6_tnl __rcu **tnls[2];
98 static struct net_device_stats *ip6_get_stats(struct net_device *dev)
100 struct pcpu_tstats sum = { 0 };
103 for_each_possible_cpu(i) {
104 const struct pcpu_tstats *tstats = per_cpu_ptr(dev->tstats, i);
106 sum.rx_packets += tstats->rx_packets;
107 sum.rx_bytes += tstats->rx_bytes;
108 sum.tx_packets += tstats->tx_packets;
109 sum.tx_bytes += tstats->tx_bytes;
111 dev->stats.rx_packets = sum.rx_packets;
112 dev->stats.rx_bytes = sum.rx_bytes;
113 dev->stats.tx_packets = sum.tx_packets;
114 dev->stats.tx_bytes = sum.tx_bytes;
119 * Locking : hash tables are protected by RCU and RTNL
122 struct dst_entry *ip6_tnl_dst_check(struct ip6_tnl *t)
124 struct dst_entry *dst = t->dst_cache;
126 if (dst && dst->obsolete &&
127 dst->ops->check(dst, t->dst_cookie) == NULL) {
135 EXPORT_SYMBOL_GPL(ip6_tnl_dst_check);
137 void ip6_tnl_dst_reset(struct ip6_tnl *t)
139 dst_release(t->dst_cache);
142 EXPORT_SYMBOL_GPL(ip6_tnl_dst_reset);
144 void ip6_tnl_dst_store(struct ip6_tnl *t, struct dst_entry *dst)
146 struct rt6_info *rt = (struct rt6_info *) dst;
147 t->dst_cookie = rt->rt6i_node ? rt->rt6i_node->fn_sernum : 0;
148 dst_release(t->dst_cache);
151 EXPORT_SYMBOL_GPL(ip6_tnl_dst_store);
154 * ip6_tnl_lookup - fetch tunnel matching the end-point addresses
155 * @remote: the address of the tunnel exit-point
156 * @local: the address of the tunnel entry-point
159 * tunnel matching given end-points if found,
160 * else fallback tunnel if its device is up,
164 #define for_each_ip6_tunnel_rcu(start) \
165 for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
167 static struct ip6_tnl *
168 ip6_tnl_lookup(struct net *net, const struct in6_addr *remote, const struct in6_addr *local)
170 unsigned int hash = HASH(remote, local);
172 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
174 for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
175 if (ipv6_addr_equal(local, &t->parms.laddr) &&
176 ipv6_addr_equal(remote, &t->parms.raddr) &&
177 (t->dev->flags & IFF_UP))
180 t = rcu_dereference(ip6n->tnls_wc[0]);
181 if (t && (t->dev->flags & IFF_UP))
188 * ip6_tnl_bucket - get head of list matching given tunnel parameters
189 * @p: parameters containing tunnel end-points
192 * ip6_tnl_bucket() returns the head of the list matching the
193 * &struct in6_addr entries laddr and raddr in @p.
195 * Return: head of IPv6 tunnel list
198 static struct ip6_tnl __rcu **
199 ip6_tnl_bucket(struct ip6_tnl_net *ip6n, const struct __ip6_tnl_parm *p)
201 const struct in6_addr *remote = &p->raddr;
202 const struct in6_addr *local = &p->laddr;
206 if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
208 h = HASH(remote, local);
210 return &ip6n->tnls[prio][h];
214 * ip6_tnl_link - add tunnel to hash table
215 * @t: tunnel to be added
219 ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
221 struct ip6_tnl __rcu **tp = ip6_tnl_bucket(ip6n, &t->parms);
223 rcu_assign_pointer(t->next , rtnl_dereference(*tp));
224 rcu_assign_pointer(*tp, t);
228 * ip6_tnl_unlink - remove tunnel from hash table
229 * @t: tunnel to be removed
233 ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
235 struct ip6_tnl __rcu **tp;
236 struct ip6_tnl *iter;
238 for (tp = ip6_tnl_bucket(ip6n, &t->parms);
239 (iter = rtnl_dereference(*tp)) != NULL;
242 rcu_assign_pointer(*tp, t->next);
248 static void ip6_dev_free(struct net_device *dev)
250 free_percpu(dev->tstats);
255 * ip6_tnl_create - create a new tunnel
256 * @p: tunnel parameters
257 * @pt: pointer to new tunnel
260 * Create tunnel matching given parameters.
263 * created tunnel or NULL
266 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
268 struct net_device *dev;
272 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
275 strlcpy(name, p->name, IFNAMSIZ);
277 sprintf(name, "ip6tnl%%d");
279 dev = alloc_netdev(sizeof (*t), name, ip6_tnl_dev_setup);
283 dev_net_set(dev, net);
285 t = netdev_priv(dev);
287 err = ip6_tnl_dev_init(dev);
291 if ((err = register_netdevice(dev)) < 0)
294 strcpy(t->parms.name, dev->name);
295 dev->rtnl_link_ops = &ip6_link_ops;
298 ip6_tnl_link(ip6n, t);
308 * ip6_tnl_locate - find or create tunnel matching given parameters
309 * @p: tunnel parameters
310 * @create: != 0 if allowed to create new tunnel if no match found
313 * ip6_tnl_locate() first tries to locate an existing tunnel
314 * based on @parms. If this is unsuccessful, but @create is set a new
315 * tunnel device is created and registered for use.
318 * matching tunnel or NULL
321 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
322 struct __ip6_tnl_parm *p, int create)
324 const struct in6_addr *remote = &p->raddr;
325 const struct in6_addr *local = &p->laddr;
326 struct ip6_tnl __rcu **tp;
328 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
330 for (tp = ip6_tnl_bucket(ip6n, p);
331 (t = rtnl_dereference(*tp)) != NULL;
333 if (ipv6_addr_equal(local, &t->parms.laddr) &&
334 ipv6_addr_equal(remote, &t->parms.raddr))
339 return ip6_tnl_create(net, p);
343 * ip6_tnl_dev_uninit - tunnel device uninitializer
344 * @dev: the device to be destroyed
347 * ip6_tnl_dev_uninit() removes tunnel from its list
351 ip6_tnl_dev_uninit(struct net_device *dev)
353 struct ip6_tnl *t = netdev_priv(dev);
354 struct net *net = dev_net(dev);
355 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
357 if (dev == ip6n->fb_tnl_dev)
358 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
360 ip6_tnl_unlink(ip6n, t);
361 ip6_tnl_dst_reset(t);
366 * parse_tvl_tnl_enc_lim - handle encapsulation limit option
367 * @skb: received socket buffer
370 * 0 if none was found,
371 * else index to encapsulation limit
374 __u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw)
376 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) raw;
377 __u8 nexthdr = ipv6h->nexthdr;
378 __u16 off = sizeof (*ipv6h);
380 while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
382 struct ipv6_opt_hdr *hdr;
383 if (raw + off + sizeof (*hdr) > skb->data &&
384 !pskb_may_pull(skb, raw - skb->data + off + sizeof (*hdr)))
387 hdr = (struct ipv6_opt_hdr *) (raw + off);
388 if (nexthdr == NEXTHDR_FRAGMENT) {
389 struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
390 if (frag_hdr->frag_off)
393 } else if (nexthdr == NEXTHDR_AUTH) {
394 optlen = (hdr->hdrlen + 2) << 2;
396 optlen = ipv6_optlen(hdr);
398 if (nexthdr == NEXTHDR_DEST) {
401 struct ipv6_tlv_tnl_enc_lim *tel;
403 /* No more room for encapsulation limit */
404 if (i + sizeof (*tel) > off + optlen)
407 tel = (struct ipv6_tlv_tnl_enc_lim *) &raw[i];
408 /* return index of option if found and valid */
409 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
412 /* else jump to next option */
414 i += tel->length + 2;
419 nexthdr = hdr->nexthdr;
424 EXPORT_SYMBOL(ip6_tnl_parse_tlv_enc_lim);
427 * ip6_tnl_err - tunnel error handler
430 * ip6_tnl_err() should handle errors in the tunnel according
431 * to the specifications in RFC 2473.
435 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
436 u8 *type, u8 *code, int *msg, __u32 *info, int offset)
438 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) skb->data;
441 u8 rel_type = ICMPV6_DEST_UNREACH;
442 u8 rel_code = ICMPV6_ADDR_UNREACH;
447 /* If the packet doesn't contain the original IPv6 header we are
448 in trouble since we might need the source address for further
449 processing of the error. */
452 if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr,
453 &ipv6h->saddr)) == NULL)
456 if (t->parms.proto != ipproto && t->parms.proto != 0)
463 struct ipv6_tlv_tnl_enc_lim *tel;
465 case ICMPV6_DEST_UNREACH:
466 net_warn_ratelimited("%s: Path to destination invalid or inactive!\n",
470 case ICMPV6_TIME_EXCEED:
471 if ((*code) == ICMPV6_EXC_HOPLIMIT) {
472 net_warn_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
477 case ICMPV6_PARAMPROB:
479 if ((*code) == ICMPV6_HDR_FIELD)
480 teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
482 if (teli && teli == *info - 2) {
483 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
484 if (tel->encap_limit == 0) {
485 net_warn_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
490 net_warn_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
494 case ICMPV6_PKT_TOOBIG:
495 mtu = *info - offset;
496 if (mtu < IPV6_MIN_MTU)
500 if ((len = sizeof (*ipv6h) + ntohs(ipv6h->payload_len)) > mtu) {
501 rel_type = ICMPV6_PKT_TOOBIG;
520 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
521 u8 type, u8 code, int offset, __be32 info)
526 __u32 rel_info = ntohl(info);
528 struct sk_buff *skb2;
529 const struct iphdr *eiph;
533 err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
534 &rel_msg, &rel_info, offset);
542 case ICMPV6_DEST_UNREACH:
543 if (rel_code != ICMPV6_ADDR_UNREACH)
545 rel_type = ICMP_DEST_UNREACH;
546 rel_code = ICMP_HOST_UNREACH;
548 case ICMPV6_PKT_TOOBIG:
551 rel_type = ICMP_DEST_UNREACH;
552 rel_code = ICMP_FRAG_NEEDED;
555 rel_type = ICMP_REDIRECT;
556 rel_code = ICMP_REDIR_HOST;
561 if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
564 skb2 = skb_clone(skb, GFP_ATOMIC);
570 skb_pull(skb2, offset);
571 skb_reset_network_header(skb2);
574 /* Try to guess incoming interface */
575 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
578 IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
582 skb2->dev = rt->dst.dev;
584 /* route "incoming" packet */
585 if (rt->rt_flags & RTCF_LOCAL) {
588 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
589 eiph->daddr, eiph->saddr,
592 RT_TOS(eiph->tos), 0);
594 rt->dst.dev->type != ARPHRD_TUNNEL) {
599 skb_dst_set(skb2, &rt->dst);
602 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
604 skb_dst(skb2)->dev->type != ARPHRD_TUNNEL)
608 /* change mtu on this route */
609 if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
610 if (rel_info > dst_mtu(skb_dst(skb2)))
613 skb_dst(skb2)->ops->update_pmtu(skb_dst(skb2), NULL, skb2, rel_info);
615 if (rel_type == ICMP_REDIRECT)
616 skb_dst(skb2)->ops->redirect(skb_dst(skb2), NULL, skb2);
618 icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
626 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
627 u8 type, u8 code, int offset, __be32 info)
632 __u32 rel_info = ntohl(info);
635 err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
636 &rel_msg, &rel_info, offset);
640 if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
642 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
648 skb_pull(skb2, offset);
649 skb_reset_network_header(skb2);
651 /* Try to guess incoming interface */
652 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
655 if (rt && rt->dst.dev)
656 skb2->dev = rt->dst.dev;
658 icmpv6_send(skb2, rel_type, rel_code, rel_info);
668 static void ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
669 const struct ipv6hdr *ipv6h,
672 __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
674 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
675 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
677 if (INET_ECN_is_ce(dsfield))
678 IP_ECN_set_ce(ip_hdr(skb));
681 static void ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
682 const struct ipv6hdr *ipv6h,
685 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
686 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
688 if (INET_ECN_is_ce(ipv6_get_dsfield(ipv6h)))
689 IP6_ECN_set_ce(ipv6_hdr(skb));
692 __u32 ip6_tnl_get_cap(struct ip6_tnl *t,
693 const struct in6_addr *laddr,
694 const struct in6_addr *raddr)
696 struct __ip6_tnl_parm *p = &t->parms;
697 int ltype = ipv6_addr_type(laddr);
698 int rtype = ipv6_addr_type(raddr);
701 if (ltype == IPV6_ADDR_ANY || rtype == IPV6_ADDR_ANY) {
702 flags = IP6_TNL_F_CAP_PER_PACKET;
703 } else if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
704 rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
705 !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
706 (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
707 if (ltype&IPV6_ADDR_UNICAST)
708 flags |= IP6_TNL_F_CAP_XMIT;
709 if (rtype&IPV6_ADDR_UNICAST)
710 flags |= IP6_TNL_F_CAP_RCV;
714 EXPORT_SYMBOL(ip6_tnl_get_cap);
716 /* called with rcu_read_lock() */
717 int ip6_tnl_rcv_ctl(struct ip6_tnl *t,
718 const struct in6_addr *laddr,
719 const struct in6_addr *raddr)
721 struct __ip6_tnl_parm *p = &t->parms;
723 struct net *net = dev_net(t->dev);
725 if ((p->flags & IP6_TNL_F_CAP_RCV) ||
726 ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
727 (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_RCV))) {
728 struct net_device *ldev = NULL;
731 ldev = dev_get_by_index_rcu(net, p->link);
733 if ((ipv6_addr_is_multicast(laddr) ||
734 likely(ipv6_chk_addr(net, laddr, ldev, 0))) &&
735 likely(!ipv6_chk_addr(net, raddr, NULL, 0)))
740 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl);
743 * ip6_tnl_rcv - decapsulate IPv6 packet and retransmit it locally
744 * @skb: received socket buffer
745 * @protocol: ethernet protocol ID
746 * @dscp_ecn_decapsulate: the function to decapsulate DSCP code and ECN
751 static int ip6_tnl_rcv(struct sk_buff *skb, __u16 protocol,
753 void (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
754 const struct ipv6hdr *ipv6h,
755 struct sk_buff *skb))
758 const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
762 if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr,
763 &ipv6h->daddr)) != NULL) {
764 struct pcpu_tstats *tstats;
766 if (t->parms.proto != ipproto && t->parms.proto != 0) {
771 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
776 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr)) {
777 t->dev->stats.rx_dropped++;
782 skb->mac_header = skb->network_header;
783 skb_reset_network_header(skb);
784 skb->protocol = htons(protocol);
785 skb->pkt_type = PACKET_HOST;
786 memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
788 tstats = this_cpu_ptr(t->dev->tstats);
789 tstats->rx_packets++;
790 tstats->rx_bytes += skb->len;
792 __skb_tunnel_rx(skb, t->dev);
794 dscp_ecn_decapsulate(t, ipv6h, skb);
809 static int ip4ip6_rcv(struct sk_buff *skb)
811 return ip6_tnl_rcv(skb, ETH_P_IP, IPPROTO_IPIP,
812 ip4ip6_dscp_ecn_decapsulate);
815 static int ip6ip6_rcv(struct sk_buff *skb)
817 return ip6_tnl_rcv(skb, ETH_P_IPV6, IPPROTO_IPV6,
818 ip6ip6_dscp_ecn_decapsulate);
821 struct ipv6_tel_txoption {
822 struct ipv6_txoptions ops;
826 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
828 memset(opt, 0, sizeof(struct ipv6_tel_txoption));
830 opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
832 opt->dst_opt[4] = encap_limit;
833 opt->dst_opt[5] = IPV6_TLV_PADN;
836 opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
837 opt->ops.opt_nflen = 8;
841 * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
842 * @t: the outgoing tunnel device
843 * @hdr: IPv6 header from the incoming packet
846 * Avoid trivial tunneling loop by checking that tunnel exit-point
847 * doesn't match source of incoming packet.
855 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
857 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
860 int ip6_tnl_xmit_ctl(struct ip6_tnl *t)
862 struct __ip6_tnl_parm *p = &t->parms;
864 struct net *net = dev_net(t->dev);
866 if (p->flags & IP6_TNL_F_CAP_XMIT) {
867 struct net_device *ldev = NULL;
871 ldev = dev_get_by_index_rcu(net, p->link);
873 if (unlikely(!ipv6_chk_addr(net, &p->laddr, ldev, 0)))
874 pr_warn("%s xmit: Local address not yet configured!\n",
876 else if (!ipv6_addr_is_multicast(&p->raddr) &&
877 unlikely(ipv6_chk_addr(net, &p->raddr, NULL, 0)))
878 pr_warn("%s xmit: Routing loop! Remote address found on this node!\n",
886 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl);
889 * ip6_tnl_xmit2 - encapsulate packet and send
890 * @skb: the outgoing socket buffer
891 * @dev: the outgoing tunnel device
892 * @dsfield: dscp code for outer header
893 * @fl: flow of tunneled packet
894 * @encap_limit: encapsulation limit
895 * @pmtu: Path MTU is stored if packet is too big
898 * Build new header and do some sanity checks on the packet before sending
904 * %-EMSGSIZE message too big. return mtu in this case.
907 static int ip6_tnl_xmit2(struct sk_buff *skb,
908 struct net_device *dev,
914 struct net *net = dev_net(dev);
915 struct ip6_tnl *t = netdev_priv(dev);
916 struct net_device_stats *stats = &t->dev->stats;
917 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
918 struct ipv6_tel_txoption opt;
919 struct dst_entry *dst = NULL, *ndst = NULL;
920 struct net_device *tdev;
922 unsigned int max_headroom = sizeof(struct ipv6hdr);
927 if (!fl6->flowi6_mark)
928 dst = ip6_tnl_dst_check(t);
930 ndst = ip6_route_output(net, NULL, fl6);
933 goto tx_err_link_failure;
934 ndst = xfrm_lookup(net, ndst, flowi6_to_flowi(fl6), NULL, 0);
938 goto tx_err_link_failure;
947 net_warn_ratelimited("%s: Local routing loop detected!\n",
949 goto tx_err_dst_release;
951 mtu = dst_mtu(dst) - sizeof (*ipv6h);
952 if (encap_limit >= 0) {
956 if (mtu < IPV6_MIN_MTU)
959 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
960 if (skb->len > mtu) {
963 goto tx_err_dst_release;
967 * Okay, now see if we can stuff it in the buffer as-is.
969 max_headroom += LL_RESERVED_SPACE(tdev);
971 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
972 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
973 struct sk_buff *new_skb;
975 if (!(new_skb = skb_realloc_headroom(skb, max_headroom)))
976 goto tx_err_dst_release;
979 skb_set_owner_w(new_skb, skb->sk);
984 if (fl6->flowi6_mark) {
985 skb_dst_set(skb, dst);
988 skb_dst_set_noref(skb, dst);
990 skb->transport_header = skb->network_header;
992 proto = fl6->flowi6_proto;
993 if (encap_limit >= 0) {
994 init_tel_txopt(&opt, encap_limit);
995 ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
997 skb_push(skb, sizeof(struct ipv6hdr));
998 skb_reset_network_header(skb);
999 ipv6h = ipv6_hdr(skb);
1000 *(__be32*)ipv6h = fl6->flowlabel | htonl(0x60000000);
1001 dsfield = INET_ECN_encapsulate(0, dsfield);
1002 ipv6_change_dsfield(ipv6h, ~INET_ECN_MASK, dsfield);
1003 ipv6h->hop_limit = t->parms.hop_limit;
1004 ipv6h->nexthdr = proto;
1005 ipv6h->saddr = fl6->saddr;
1006 ipv6h->daddr = fl6->daddr;
1009 err = ip6_local_out(skb);
1011 if (net_xmit_eval(err) == 0) {
1012 struct pcpu_tstats *tstats = this_cpu_ptr(t->dev->tstats);
1014 tstats->tx_bytes += pkt_len;
1015 tstats->tx_packets++;
1018 stats->tx_aborted_errors++;
1021 ip6_tnl_dst_store(t, ndst);
1023 tx_err_link_failure:
1024 stats->tx_carrier_errors++;
1025 dst_link_failure(skb);
1032 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1034 struct ip6_tnl *t = netdev_priv(dev);
1035 const struct iphdr *iph = ip_hdr(skb);
1036 int encap_limit = -1;
1042 if ((t->parms.proto != IPPROTO_IPIP && t->parms.proto != 0) ||
1043 !ip6_tnl_xmit_ctl(t))
1046 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1047 encap_limit = t->parms.encap_limit;
1049 memcpy(&fl6, &t->fl.u.ip6, sizeof (fl6));
1050 fl6.flowi6_proto = IPPROTO_IPIP;
1052 dsfield = ipv4_get_dsfield(iph);
1054 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1055 fl6.flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
1057 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1058 fl6.flowi6_mark = skb->mark;
1060 err = ip6_tnl_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
1062 /* XXX: send ICMP error even if DF is not set. */
1063 if (err == -EMSGSIZE)
1064 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
1073 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1075 struct ip6_tnl *t = netdev_priv(dev);
1076 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
1077 int encap_limit = -1;
1084 if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) ||
1085 !ip6_tnl_xmit_ctl(t) || ip6_tnl_addr_conflict(t, ipv6h))
1088 offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
1090 struct ipv6_tlv_tnl_enc_lim *tel;
1091 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
1092 if (tel->encap_limit == 0) {
1093 icmpv6_send(skb, ICMPV6_PARAMPROB,
1094 ICMPV6_HDR_FIELD, offset + 2);
1097 encap_limit = tel->encap_limit - 1;
1098 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1099 encap_limit = t->parms.encap_limit;
1101 memcpy(&fl6, &t->fl.u.ip6, sizeof (fl6));
1102 fl6.flowi6_proto = IPPROTO_IPV6;
1104 dsfield = ipv6_get_dsfield(ipv6h);
1105 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1106 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
1107 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
1108 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_FLOWLABEL_MASK);
1109 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1110 fl6.flowi6_mark = skb->mark;
1112 err = ip6_tnl_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
1114 if (err == -EMSGSIZE)
1115 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1123 ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1125 struct ip6_tnl *t = netdev_priv(dev);
1126 struct net_device_stats *stats = &t->dev->stats;
1129 switch (skb->protocol) {
1130 case htons(ETH_P_IP):
1131 ret = ip4ip6_tnl_xmit(skb, dev);
1133 case htons(ETH_P_IPV6):
1134 ret = ip6ip6_tnl_xmit(skb, dev);
1143 return NETDEV_TX_OK;
1147 stats->tx_dropped++;
1149 return NETDEV_TX_OK;
1152 static void ip6_tnl_link_config(struct ip6_tnl *t)
1154 struct net_device *dev = t->dev;
1155 struct __ip6_tnl_parm *p = &t->parms;
1156 struct flowi6 *fl6 = &t->fl.u.ip6;
1158 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1159 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1161 /* Set up flowi template */
1162 fl6->saddr = p->laddr;
1163 fl6->daddr = p->raddr;
1164 fl6->flowi6_oif = p->link;
1167 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1168 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1169 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1170 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1172 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1173 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1175 if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1176 dev->flags |= IFF_POINTOPOINT;
1178 dev->flags &= ~IFF_POINTOPOINT;
1180 dev->iflink = p->link;
1182 if (p->flags & IP6_TNL_F_CAP_XMIT) {
1183 int strict = (ipv6_addr_type(&p->raddr) &
1184 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1186 struct rt6_info *rt = rt6_lookup(dev_net(dev),
1187 &p->raddr, &p->laddr,
1194 dev->hard_header_len = rt->dst.dev->hard_header_len +
1195 sizeof (struct ipv6hdr);
1197 dev->mtu = rt->dst.dev->mtu - sizeof (struct ipv6hdr);
1198 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1201 if (dev->mtu < IPV6_MIN_MTU)
1202 dev->mtu = IPV6_MIN_MTU;
1209 * ip6_tnl_change - update the tunnel parameters
1210 * @t: tunnel to be changed
1211 * @p: tunnel configuration parameters
1214 * ip6_tnl_change() updates the tunnel parameters
1218 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
1220 t->parms.laddr = p->laddr;
1221 t->parms.raddr = p->raddr;
1222 t->parms.flags = p->flags;
1223 t->parms.hop_limit = p->hop_limit;
1224 t->parms.encap_limit = p->encap_limit;
1225 t->parms.flowinfo = p->flowinfo;
1226 t->parms.link = p->link;
1227 t->parms.proto = p->proto;
1228 ip6_tnl_dst_reset(t);
1229 ip6_tnl_link_config(t);
1234 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u)
1236 p->laddr = u->laddr;
1237 p->raddr = u->raddr;
1238 p->flags = u->flags;
1239 p->hop_limit = u->hop_limit;
1240 p->encap_limit = u->encap_limit;
1241 p->flowinfo = u->flowinfo;
1243 p->proto = u->proto;
1244 memcpy(p->name, u->name, sizeof(u->name));
1248 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p)
1250 u->laddr = p->laddr;
1251 u->raddr = p->raddr;
1252 u->flags = p->flags;
1253 u->hop_limit = p->hop_limit;
1254 u->encap_limit = p->encap_limit;
1255 u->flowinfo = p->flowinfo;
1257 u->proto = p->proto;
1258 memcpy(u->name, p->name, sizeof(u->name));
1262 * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1263 * @dev: virtual device associated with tunnel
1264 * @ifr: parameters passed from userspace
1265 * @cmd: command to be performed
1268 * ip6_tnl_ioctl() is used for managing IPv6 tunnels
1271 * The possible commands are the following:
1272 * %SIOCGETTUNNEL: get tunnel parameters for device
1273 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1274 * %SIOCCHGTUNNEL: change tunnel parameters to those given
1275 * %SIOCDELTUNNEL: delete tunnel
1277 * The fallback device "ip6tnl0", created during module
1278 * initialization, can be used for creating other tunnel devices.
1282 * %-EFAULT if unable to copy data to or from userspace,
1283 * %-EPERM if current process hasn't %CAP_NET_ADMIN set
1284 * %-EINVAL if passed tunnel parameters are invalid,
1285 * %-EEXIST if changing a tunnel's parameters would cause a conflict
1286 * %-ENODEV if attempting to change or delete a nonexisting device
1290 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1293 struct ip6_tnl_parm p;
1294 struct __ip6_tnl_parm p1;
1295 struct ip6_tnl *t = NULL;
1296 struct net *net = dev_net(dev);
1297 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1301 if (dev == ip6n->fb_tnl_dev) {
1302 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p))) {
1306 ip6_tnl_parm_from_user(&p1, &p);
1307 t = ip6_tnl_locate(net, &p1, 0);
1309 memset(&p, 0, sizeof(p));
1312 t = netdev_priv(dev);
1313 ip6_tnl_parm_to_user(&p, &t->parms);
1314 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof (p))) {
1321 if (!capable(CAP_NET_ADMIN))
1324 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
1327 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1330 ip6_tnl_parm_from_user(&p1, &p);
1331 t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL);
1332 if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) {
1334 if (t->dev != dev) {
1339 t = netdev_priv(dev);
1341 ip6_tnl_unlink(ip6n, t);
1343 err = ip6_tnl_change(t, &p1);
1344 ip6_tnl_link(ip6n, t);
1345 netdev_state_change(dev);
1349 ip6_tnl_parm_to_user(&p, &t->parms);
1350 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1354 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1358 if (!capable(CAP_NET_ADMIN))
1361 if (dev == ip6n->fb_tnl_dev) {
1363 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
1366 ip6_tnl_parm_from_user(&p1, &p);
1367 t = ip6_tnl_locate(net, &p1, 0);
1371 if (t->dev == ip6n->fb_tnl_dev)
1376 unregister_netdevice(dev);
1385 * ip6_tnl_change_mtu - change mtu manually for tunnel device
1386 * @dev: virtual device associated with tunnel
1387 * @new_mtu: the new mtu
1391 * %-EINVAL if mtu too small
1395 ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1397 if (new_mtu < IPV6_MIN_MTU) {
1405 static const struct net_device_ops ip6_tnl_netdev_ops = {
1406 .ndo_uninit = ip6_tnl_dev_uninit,
1407 .ndo_start_xmit = ip6_tnl_xmit,
1408 .ndo_do_ioctl = ip6_tnl_ioctl,
1409 .ndo_change_mtu = ip6_tnl_change_mtu,
1410 .ndo_get_stats = ip6_get_stats,
1415 * ip6_tnl_dev_setup - setup virtual tunnel device
1416 * @dev: virtual device associated with tunnel
1419 * Initialize function pointers and device parameters
1422 static void ip6_tnl_dev_setup(struct net_device *dev)
1426 dev->netdev_ops = &ip6_tnl_netdev_ops;
1427 dev->destructor = ip6_dev_free;
1429 dev->type = ARPHRD_TUNNEL6;
1430 dev->hard_header_len = LL_MAX_HEADER + sizeof (struct ipv6hdr);
1431 dev->mtu = ETH_DATA_LEN - sizeof (struct ipv6hdr);
1432 t = netdev_priv(dev);
1433 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1435 dev->flags |= IFF_NOARP;
1436 dev->addr_len = sizeof(struct in6_addr);
1437 dev->features |= NETIF_F_NETNS_LOCAL;
1438 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
1443 * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1444 * @dev: virtual device associated with tunnel
1448 ip6_tnl_dev_init_gen(struct net_device *dev)
1450 struct ip6_tnl *t = netdev_priv(dev);
1453 dev->tstats = alloc_percpu(struct pcpu_tstats);
1460 * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1461 * @dev: virtual device associated with tunnel
1464 static int ip6_tnl_dev_init(struct net_device *dev)
1466 struct ip6_tnl *t = netdev_priv(dev);
1467 int err = ip6_tnl_dev_init_gen(dev);
1471 ip6_tnl_link_config(t);
1476 * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1477 * @dev: fallback device
1482 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1484 struct ip6_tnl *t = netdev_priv(dev);
1485 struct net *net = dev_net(dev);
1486 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1487 int err = ip6_tnl_dev_init_gen(dev);
1492 t->parms.proto = IPPROTO_IPV6;
1495 ip6_tnl_link_config(t);
1497 rcu_assign_pointer(ip6n->tnls_wc[0], t);
1501 static size_t ip6_get_size(const struct net_device *dev)
1504 /* IFLA_IPTUN_LINK */
1506 /* IFLA_IPTUN_LOCAL */
1507 nla_total_size(sizeof(struct in6_addr)) +
1508 /* IFLA_IPTUN_REMOTE */
1509 nla_total_size(sizeof(struct in6_addr)) +
1510 /* IFLA_IPTUN_TTL */
1512 /* IFLA_IPTUN_ENCAP_LIMIT */
1514 /* IFLA_IPTUN_FLOWINFO */
1516 /* IFLA_IPTUN_FLAGS */
1521 static int ip6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1523 struct ip6_tnl *tunnel = netdev_priv(dev);
1524 struct __ip6_tnl_parm *parm = &tunnel->parms;
1526 if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1527 nla_put(skb, IFLA_IPTUN_LOCAL, sizeof(struct in6_addr),
1529 nla_put(skb, IFLA_IPTUN_REMOTE, sizeof(struct in6_addr),
1531 nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) ||
1532 nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) ||
1533 nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) ||
1534 nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags))
1535 goto nla_put_failure;
1542 static struct rtnl_link_ops ip6_link_ops __read_mostly = {
1544 .maxtype = IFLA_IPTUN_MAX,
1545 .priv_size = sizeof(struct ip6_tnl),
1546 .get_size = ip6_get_size,
1547 .fill_info = ip6_fill_info,
1550 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
1551 .handler = ip4ip6_rcv,
1552 .err_handler = ip4ip6_err,
1556 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
1557 .handler = ip6ip6_rcv,
1558 .err_handler = ip6ip6_err,
1562 static void __net_exit ip6_tnl_destroy_tunnels(struct ip6_tnl_net *ip6n)
1568 for (h = 0; h < HASH_SIZE; h++) {
1569 t = rtnl_dereference(ip6n->tnls_r_l[h]);
1571 unregister_netdevice_queue(t->dev, &list);
1572 t = rtnl_dereference(t->next);
1576 t = rtnl_dereference(ip6n->tnls_wc[0]);
1577 unregister_netdevice_queue(t->dev, &list);
1578 unregister_netdevice_many(&list);
1581 static int __net_init ip6_tnl_init_net(struct net *net)
1583 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1584 struct ip6_tnl *t = NULL;
1587 ip6n->tnls[0] = ip6n->tnls_wc;
1588 ip6n->tnls[1] = ip6n->tnls_r_l;
1591 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
1594 if (!ip6n->fb_tnl_dev)
1596 dev_net_set(ip6n->fb_tnl_dev, net);
1598 err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
1602 err = register_netdev(ip6n->fb_tnl_dev);
1606 t = netdev_priv(ip6n->fb_tnl_dev);
1608 strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
1612 ip6_dev_free(ip6n->fb_tnl_dev);
1617 static void __net_exit ip6_tnl_exit_net(struct net *net)
1619 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1622 ip6_tnl_destroy_tunnels(ip6n);
1626 static struct pernet_operations ip6_tnl_net_ops = {
1627 .init = ip6_tnl_init_net,
1628 .exit = ip6_tnl_exit_net,
1629 .id = &ip6_tnl_net_id,
1630 .size = sizeof(struct ip6_tnl_net),
1634 * ip6_tunnel_init - register protocol and reserve needed resources
1636 * Return: 0 on success
1639 static int __init ip6_tunnel_init(void)
1643 err = register_pernet_device(&ip6_tnl_net_ops);
1647 err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
1649 pr_err("%s: can't register ip4ip6\n", __func__);
1653 err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
1655 pr_err("%s: can't register ip6ip6\n", __func__);
1658 err = rtnl_link_register(&ip6_link_ops);
1660 goto rtnl_link_failed;
1665 xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
1667 xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
1669 unregister_pernet_device(&ip6_tnl_net_ops);
1675 * ip6_tunnel_cleanup - free resources and unregister protocol
1678 static void __exit ip6_tunnel_cleanup(void)
1680 rtnl_link_unregister(&ip6_link_ops);
1681 if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
1682 pr_info("%s: can't deregister ip4ip6\n", __func__);
1684 if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
1685 pr_info("%s: can't deregister ip6ip6\n", __func__);
1687 unregister_pernet_device(&ip6_tnl_net_ops);
1690 module_init(ip6_tunnel_init);
1691 module_exit(ip6_tunnel_cleanup);