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 #include <linux/module.h>
22 #include <linux/capability.h>
23 #include <linux/errno.h>
24 #include <linux/types.h>
25 #include <linux/sockios.h>
26 #include <linux/icmp.h>
30 #include <linux/if_tunnel.h>
31 #include <linux/net.h>
32 #include <linux/in6.h>
33 #include <linux/netdevice.h>
34 #include <linux/if_arp.h>
35 #include <linux/icmpv6.h>
36 #include <linux/init.h>
37 #include <linux/route.h>
38 #include <linux/rtnetlink.h>
39 #include <linux/netfilter_ipv6.h>
40 #include <linux/slab.h>
42 #include <asm/uaccess.h>
43 #include <linux/atomic.h>
48 #include <net/ip6_route.h>
49 #include <net/addrconf.h>
50 #include <net/ip6_tunnel.h>
52 #include <net/dsfield.h>
53 #include <net/inet_ecn.h>
54 #include <net/net_namespace.h>
55 #include <net/netns/generic.h>
57 MODULE_AUTHOR("Ville Nuorvala");
58 MODULE_DESCRIPTION("IPv6 tunneling device");
59 MODULE_LICENSE("GPL");
60 MODULE_ALIAS_NETDEV("ip6tnl0");
63 #define IP6_TNL_TRACE(x...) printk(KERN_DEBUG "%s:" x "\n", __func__)
65 #define IP6_TNL_TRACE(x...) do {;} while(0)
68 #define IPV6_TCLASS_MASK (IPV6_FLOWINFO_MASK & ~IPV6_FLOWLABEL_MASK)
69 #define IPV6_TCLASS_SHIFT 20
73 #define HASH(addr) ((__force u32)((addr)->s6_addr32[0] ^ (addr)->s6_addr32[1] ^ \
74 (addr)->s6_addr32[2] ^ (addr)->s6_addr32[3]) & \
77 static int ip6_tnl_dev_init(struct net_device *dev);
78 static void ip6_tnl_dev_setup(struct net_device *dev);
80 static int ip6_tnl_net_id __read_mostly;
82 /* the IPv6 tunnel fallback device */
83 struct net_device *fb_tnl_dev;
84 /* lists for storing tunnels in use */
85 struct ip6_tnl __rcu *tnls_r_l[HASH_SIZE];
86 struct ip6_tnl __rcu *tnls_wc[1];
87 struct ip6_tnl __rcu **tnls[2];
90 /* often modified stats are per cpu, other are shared (netdev->stats) */
92 unsigned long rx_packets;
93 unsigned long rx_bytes;
94 unsigned long tx_packets;
95 unsigned long tx_bytes;
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 static inline 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) {
136 static inline void ip6_tnl_dst_reset(struct ip6_tnl *t)
138 dst_release(t->dst_cache);
142 static inline void ip6_tnl_dst_store(struct ip6_tnl *t, struct dst_entry *dst)
144 struct rt6_info *rt = (struct rt6_info *) dst;
145 t->dst_cookie = rt->rt6i_node ? rt->rt6i_node->fn_sernum : 0;
146 dst_release(t->dst_cache);
151 * ip6_tnl_lookup - fetch tunnel matching the end-point addresses
152 * @remote: the address of the tunnel exit-point
153 * @local: the address of the tunnel entry-point
156 * tunnel matching given end-points if found,
157 * else fallback tunnel if its device is up,
161 #define for_each_ip6_tunnel_rcu(start) \
162 for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
164 static struct ip6_tnl *
165 ip6_tnl_lookup(struct net *net, const struct in6_addr *remote, const struct in6_addr *local)
167 unsigned int h0 = HASH(remote);
168 unsigned int h1 = HASH(local);
170 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
172 for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[h0 ^ h1]) {
173 if (ipv6_addr_equal(local, &t->parms.laddr) &&
174 ipv6_addr_equal(remote, &t->parms.raddr) &&
175 (t->dev->flags & IFF_UP))
178 t = rcu_dereference(ip6n->tnls_wc[0]);
179 if (t && (t->dev->flags & IFF_UP))
186 * ip6_tnl_bucket - get head of list matching given tunnel parameters
187 * @p: parameters containing tunnel end-points
190 * ip6_tnl_bucket() returns the head of the list matching the
191 * &struct in6_addr entries laddr and raddr in @p.
193 * Return: head of IPv6 tunnel list
196 static struct ip6_tnl __rcu **
197 ip6_tnl_bucket(struct ip6_tnl_net *ip6n, const struct ip6_tnl_parm *p)
199 const struct in6_addr *remote = &p->raddr;
200 const struct in6_addr *local = &p->laddr;
204 if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
206 h = HASH(remote) ^ HASH(local);
208 return &ip6n->tnls[prio][h];
212 * ip6_tnl_link - add tunnel to hash table
213 * @t: tunnel to be added
217 ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
219 struct ip6_tnl __rcu **tp = ip6_tnl_bucket(ip6n, &t->parms);
221 RCU_INIT_POINTER(t->next , rtnl_dereference(*tp));
222 RCU_INIT_POINTER(*tp, t);
226 * ip6_tnl_unlink - remove tunnel from hash table
227 * @t: tunnel to be removed
231 ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
233 struct ip6_tnl __rcu **tp;
234 struct ip6_tnl *iter;
236 for (tp = ip6_tnl_bucket(ip6n, &t->parms);
237 (iter = rtnl_dereference(*tp)) != NULL;
240 RCU_INIT_POINTER(*tp, t->next);
246 static void ip6_dev_free(struct net_device *dev)
248 free_percpu(dev->tstats);
253 * ip6_tnl_create() - create a new tunnel
254 * @p: tunnel parameters
255 * @pt: pointer to new tunnel
258 * Create tunnel matching given parameters.
261 * created tunnel or NULL
264 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct ip6_tnl_parm *p)
266 struct net_device *dev;
270 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
273 strlcpy(name, p->name, IFNAMSIZ);
275 sprintf(name, "ip6tnl%%d");
277 dev = alloc_netdev(sizeof (*t), name, ip6_tnl_dev_setup);
281 dev_net_set(dev, net);
283 t = netdev_priv(dev);
285 err = ip6_tnl_dev_init(dev);
289 if ((err = register_netdevice(dev)) < 0)
293 ip6_tnl_link(ip6n, t);
303 * ip6_tnl_locate - find or create tunnel matching given parameters
304 * @p: tunnel parameters
305 * @create: != 0 if allowed to create new tunnel if no match found
308 * ip6_tnl_locate() first tries to locate an existing tunnel
309 * based on @parms. If this is unsuccessful, but @create is set a new
310 * tunnel device is created and registered for use.
313 * matching tunnel or NULL
316 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
317 struct ip6_tnl_parm *p, int create)
319 const struct in6_addr *remote = &p->raddr;
320 const struct in6_addr *local = &p->laddr;
321 struct ip6_tnl __rcu **tp;
323 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
325 for (tp = ip6_tnl_bucket(ip6n, p);
326 (t = rtnl_dereference(*tp)) != NULL;
328 if (ipv6_addr_equal(local, &t->parms.laddr) &&
329 ipv6_addr_equal(remote, &t->parms.raddr))
334 return ip6_tnl_create(net, p);
338 * ip6_tnl_dev_uninit - tunnel device uninitializer
339 * @dev: the device to be destroyed
342 * ip6_tnl_dev_uninit() removes tunnel from its list
346 ip6_tnl_dev_uninit(struct net_device *dev)
348 struct ip6_tnl *t = netdev_priv(dev);
349 struct net *net = dev_net(dev);
350 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
352 if (dev == ip6n->fb_tnl_dev)
353 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
355 ip6_tnl_unlink(ip6n, t);
356 ip6_tnl_dst_reset(t);
361 * parse_tvl_tnl_enc_lim - handle encapsulation limit option
362 * @skb: received socket buffer
365 * 0 if none was found,
366 * else index to encapsulation limit
370 parse_tlv_tnl_enc_lim(struct sk_buff *skb, __u8 * raw)
372 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) raw;
373 __u8 nexthdr = ipv6h->nexthdr;
374 __u16 off = sizeof (*ipv6h);
376 while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
378 struct ipv6_opt_hdr *hdr;
379 if (raw + off + sizeof (*hdr) > skb->data &&
380 !pskb_may_pull(skb, raw - skb->data + off + sizeof (*hdr)))
383 hdr = (struct ipv6_opt_hdr *) (raw + off);
384 if (nexthdr == NEXTHDR_FRAGMENT) {
385 struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
386 if (frag_hdr->frag_off)
389 } else if (nexthdr == NEXTHDR_AUTH) {
390 optlen = (hdr->hdrlen + 2) << 2;
392 optlen = ipv6_optlen(hdr);
394 if (nexthdr == NEXTHDR_DEST) {
397 struct ipv6_tlv_tnl_enc_lim *tel;
399 /* No more room for encapsulation limit */
400 if (i + sizeof (*tel) > off + optlen)
403 tel = (struct ipv6_tlv_tnl_enc_lim *) &raw[i];
404 /* return index of option if found and valid */
405 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
408 /* else jump to next option */
410 i += tel->length + 2;
415 nexthdr = hdr->nexthdr;
422 * ip6_tnl_err - tunnel error handler
425 * ip6_tnl_err() should handle errors in the tunnel according
426 * to the specifications in RFC 2473.
430 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
431 u8 *type, u8 *code, int *msg, __u32 *info, int offset)
433 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) skb->data;
436 u8 rel_type = ICMPV6_DEST_UNREACH;
437 u8 rel_code = ICMPV6_ADDR_UNREACH;
442 /* If the packet doesn't contain the original IPv6 header we are
443 in trouble since we might need the source address for further
444 processing of the error. */
447 if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr,
448 &ipv6h->saddr)) == NULL)
451 if (t->parms.proto != ipproto && t->parms.proto != 0)
458 struct ipv6_tlv_tnl_enc_lim *tel;
460 case ICMPV6_DEST_UNREACH:
463 "%s: Path to destination invalid "
464 "or inactive!\n", t->parms.name);
467 case ICMPV6_TIME_EXCEED:
468 if ((*code) == ICMPV6_EXC_HOPLIMIT) {
471 "%s: Too small hop limit or "
472 "routing loop in tunnel!\n",
477 case ICMPV6_PARAMPROB:
479 if ((*code) == ICMPV6_HDR_FIELD)
480 teli = parse_tlv_tnl_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) {
487 "%s: Too small encapsulation "
488 "limit or routing loop in "
489 "tunnel!\n", t->parms.name);
492 } else if (net_ratelimit()) {
494 "%s: Recipient unable to parse tunneled "
495 "packet!\n ", t->parms.name);
498 case ICMPV6_PKT_TOOBIG:
499 mtu = *info - offset;
500 if (mtu < IPV6_MIN_MTU)
504 if ((len = sizeof (*ipv6h) + ntohs(ipv6h->payload_len)) > mtu) {
505 rel_type = ICMPV6_PKT_TOOBIG;
524 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
525 u8 type, u8 code, int offset, __be32 info)
530 __u32 rel_info = ntohl(info);
532 struct sk_buff *skb2;
533 const struct iphdr *eiph;
537 err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
538 &rel_msg, &rel_info, offset);
546 case ICMPV6_DEST_UNREACH:
547 if (rel_code != ICMPV6_ADDR_UNREACH)
549 rel_type = ICMP_DEST_UNREACH;
550 rel_code = ICMP_HOST_UNREACH;
552 case ICMPV6_PKT_TOOBIG:
555 rel_type = ICMP_DEST_UNREACH;
556 rel_code = ICMP_FRAG_NEEDED;
562 if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
565 skb2 = skb_clone(skb, GFP_ATOMIC);
571 skb_pull(skb2, offset);
572 skb_reset_network_header(skb2);
575 /* Try to guess incoming interface */
576 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
579 IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
583 skb2->dev = rt->dst.dev;
585 /* route "incoming" packet */
586 if (rt->rt_flags & RTCF_LOCAL) {
589 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
590 eiph->daddr, eiph->saddr,
593 RT_TOS(eiph->tos), 0);
595 rt->dst.dev->type != ARPHRD_TUNNEL) {
600 skb_dst_set(skb2, &rt->dst);
603 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
605 skb_dst(skb2)->dev->type != ARPHRD_TUNNEL)
609 /* change mtu on this route */
610 if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
611 if (rel_info > dst_mtu(skb_dst(skb2)))
614 skb_dst(skb2)->ops->update_pmtu(skb_dst(skb2), rel_info);
617 icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
625 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
626 u8 type, u8 code, int offset, __be32 info)
631 __u32 rel_info = ntohl(info);
634 err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
635 &rel_msg, &rel_info, offset);
639 if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
641 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
647 skb_pull(skb2, offset);
648 skb_reset_network_header(skb2);
650 /* Try to guess incoming interface */
651 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
654 if (rt && rt->rt6i_dev)
655 skb2->dev = rt->rt6i_dev;
657 icmpv6_send(skb2, rel_type, rel_code, rel_info);
660 dst_release(&rt->dst);
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 /* called with rcu_read_lock() */
693 static inline int ip6_tnl_rcv_ctl(struct ip6_tnl *t)
695 struct ip6_tnl_parm *p = &t->parms;
697 struct net *net = dev_net(t->dev);
699 if (p->flags & IP6_TNL_F_CAP_RCV) {
700 struct net_device *ldev = NULL;
703 ldev = dev_get_by_index_rcu(net, p->link);
705 if ((ipv6_addr_is_multicast(&p->laddr) ||
706 likely(ipv6_chk_addr(net, &p->laddr, ldev, 0))) &&
707 likely(!ipv6_chk_addr(net, &p->raddr, NULL, 0)))
715 * ip6_tnl_rcv - decapsulate IPv6 packet and retransmit it locally
716 * @skb: received socket buffer
717 * @protocol: ethernet protocol ID
718 * @dscp_ecn_decapsulate: the function to decapsulate DSCP code and ECN
723 static int ip6_tnl_rcv(struct sk_buff *skb, __u16 protocol,
725 void (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
726 const struct ipv6hdr *ipv6h,
727 struct sk_buff *skb))
730 const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
734 if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr,
735 &ipv6h->daddr)) != NULL) {
736 struct pcpu_tstats *tstats;
738 if (t->parms.proto != ipproto && t->parms.proto != 0) {
743 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
748 if (!ip6_tnl_rcv_ctl(t)) {
749 t->dev->stats.rx_dropped++;
754 skb->mac_header = skb->network_header;
755 skb_reset_network_header(skb);
756 skb->protocol = htons(protocol);
757 skb->pkt_type = PACKET_HOST;
758 memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
760 tstats = this_cpu_ptr(t->dev->tstats);
761 tstats->rx_packets++;
762 tstats->rx_bytes += skb->len;
764 __skb_tunnel_rx(skb, t->dev);
766 dscp_ecn_decapsulate(t, ipv6h, skb);
781 static int ip4ip6_rcv(struct sk_buff *skb)
783 return ip6_tnl_rcv(skb, ETH_P_IP, IPPROTO_IPIP,
784 ip4ip6_dscp_ecn_decapsulate);
787 static int ip6ip6_rcv(struct sk_buff *skb)
789 return ip6_tnl_rcv(skb, ETH_P_IPV6, IPPROTO_IPV6,
790 ip6ip6_dscp_ecn_decapsulate);
793 struct ipv6_tel_txoption {
794 struct ipv6_txoptions ops;
798 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
800 memset(opt, 0, sizeof(struct ipv6_tel_txoption));
802 opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
804 opt->dst_opt[4] = encap_limit;
805 opt->dst_opt[5] = IPV6_TLV_PADN;
808 opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
809 opt->ops.opt_nflen = 8;
813 * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
814 * @t: the outgoing tunnel device
815 * @hdr: IPv6 header from the incoming packet
818 * Avoid trivial tunneling loop by checking that tunnel exit-point
819 * doesn't match source of incoming packet.
827 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
829 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
832 static inline int ip6_tnl_xmit_ctl(struct ip6_tnl *t)
834 struct ip6_tnl_parm *p = &t->parms;
836 struct net *net = dev_net(t->dev);
838 if (p->flags & IP6_TNL_F_CAP_XMIT) {
839 struct net_device *ldev = NULL;
843 ldev = dev_get_by_index_rcu(net, p->link);
845 if (unlikely(!ipv6_chk_addr(net, &p->laddr, ldev, 0)))
847 "%s xmit: Local address not yet configured!\n",
849 else if (!ipv6_addr_is_multicast(&p->raddr) &&
850 unlikely(ipv6_chk_addr(net, &p->raddr, NULL, 0)))
852 "%s xmit: Routing loop! "
853 "Remote address found on this node!\n",
862 * ip6_tnl_xmit2 - encapsulate packet and send
863 * @skb: the outgoing socket buffer
864 * @dev: the outgoing tunnel device
865 * @dsfield: dscp code for outer header
866 * @fl: flow of tunneled packet
867 * @encap_limit: encapsulation limit
868 * @pmtu: Path MTU is stored if packet is too big
871 * Build new header and do some sanity checks on the packet before sending
877 * %-EMSGSIZE message too big. return mtu in this case.
880 static int ip6_tnl_xmit2(struct sk_buff *skb,
881 struct net_device *dev,
887 struct net *net = dev_net(dev);
888 struct ip6_tnl *t = netdev_priv(dev);
889 struct net_device_stats *stats = &t->dev->stats;
890 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
891 struct ipv6_tel_txoption opt;
892 struct dst_entry *dst, *ndst = NULL;
893 struct net_device *tdev;
895 unsigned int max_headroom = sizeof(struct ipv6hdr);
900 dst = ip6_tnl_dst_check(t);
902 ndst = ip6_route_output(net, NULL, fl6);
905 goto tx_err_link_failure;
906 ndst = xfrm_lookup(net, ndst, flowi6_to_flowi(fl6), NULL, 0);
910 goto tx_err_link_failure;
921 "%s: Local routing loop detected!\n",
923 goto tx_err_dst_release;
925 mtu = dst_mtu(dst) - sizeof (*ipv6h);
926 if (encap_limit >= 0) {
930 if (mtu < IPV6_MIN_MTU)
933 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), mtu);
934 if (skb->len > mtu) {
937 goto tx_err_dst_release;
941 * Okay, now see if we can stuff it in the buffer as-is.
943 max_headroom += LL_RESERVED_SPACE(tdev);
945 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
946 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
947 struct sk_buff *new_skb;
949 if (!(new_skb = skb_realloc_headroom(skb, max_headroom)))
950 goto tx_err_dst_release;
953 skb_set_owner_w(new_skb, skb->sk);
958 skb_dst_set_noref(skb, dst);
960 skb->transport_header = skb->network_header;
962 proto = fl6->flowi6_proto;
963 if (encap_limit >= 0) {
964 init_tel_txopt(&opt, encap_limit);
965 ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
967 skb_push(skb, sizeof(struct ipv6hdr));
968 skb_reset_network_header(skb);
969 ipv6h = ipv6_hdr(skb);
970 *(__be32*)ipv6h = fl6->flowlabel | htonl(0x60000000);
971 dsfield = INET_ECN_encapsulate(0, dsfield);
972 ipv6_change_dsfield(ipv6h, ~INET_ECN_MASK, dsfield);
973 ipv6h->hop_limit = t->parms.hop_limit;
974 ipv6h->nexthdr = proto;
975 ipv6_addr_copy(&ipv6h->saddr, &fl6->saddr);
976 ipv6_addr_copy(&ipv6h->daddr, &fl6->daddr);
979 err = ip6_local_out(skb);
981 if (net_xmit_eval(err) == 0) {
982 struct pcpu_tstats *tstats = this_cpu_ptr(t->dev->tstats);
984 tstats->tx_bytes += pkt_len;
985 tstats->tx_packets++;
988 stats->tx_aborted_errors++;
991 ip6_tnl_dst_store(t, ndst);
994 stats->tx_carrier_errors++;
995 dst_link_failure(skb);
1002 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1004 struct ip6_tnl *t = netdev_priv(dev);
1005 const struct iphdr *iph = ip_hdr(skb);
1006 int encap_limit = -1;
1012 if ((t->parms.proto != IPPROTO_IPIP && t->parms.proto != 0) ||
1013 !ip6_tnl_xmit_ctl(t))
1016 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1017 encap_limit = t->parms.encap_limit;
1019 memcpy(&fl6, &t->fl.u.ip6, sizeof (fl6));
1020 fl6.flowi6_proto = IPPROTO_IPIP;
1022 dsfield = ipv4_get_dsfield(iph);
1024 if ((t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS))
1025 fl6.flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
1028 err = ip6_tnl_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
1030 /* XXX: send ICMP error even if DF is not set. */
1031 if (err == -EMSGSIZE)
1032 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
1041 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1043 struct ip6_tnl *t = netdev_priv(dev);
1044 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
1045 int encap_limit = -1;
1052 if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) ||
1053 !ip6_tnl_xmit_ctl(t) || ip6_tnl_addr_conflict(t, ipv6h))
1056 offset = parse_tlv_tnl_enc_lim(skb, skb_network_header(skb));
1058 struct ipv6_tlv_tnl_enc_lim *tel;
1059 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
1060 if (tel->encap_limit == 0) {
1061 icmpv6_send(skb, ICMPV6_PARAMPROB,
1062 ICMPV6_HDR_FIELD, offset + 2);
1065 encap_limit = tel->encap_limit - 1;
1066 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1067 encap_limit = t->parms.encap_limit;
1069 memcpy(&fl6, &t->fl.u.ip6, sizeof (fl6));
1070 fl6.flowi6_proto = IPPROTO_IPV6;
1072 dsfield = ipv6_get_dsfield(ipv6h);
1073 if ((t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS))
1074 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
1075 if ((t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL))
1076 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_FLOWLABEL_MASK);
1078 err = ip6_tnl_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
1080 if (err == -EMSGSIZE)
1081 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1089 ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1091 struct ip6_tnl *t = netdev_priv(dev);
1092 struct net_device_stats *stats = &t->dev->stats;
1095 switch (skb->protocol) {
1096 case htons(ETH_P_IP):
1097 ret = ip4ip6_tnl_xmit(skb, dev);
1099 case htons(ETH_P_IPV6):
1100 ret = ip6ip6_tnl_xmit(skb, dev);
1109 return NETDEV_TX_OK;
1113 stats->tx_dropped++;
1115 return NETDEV_TX_OK;
1118 static void ip6_tnl_set_cap(struct ip6_tnl *t)
1120 struct ip6_tnl_parm *p = &t->parms;
1121 int ltype = ipv6_addr_type(&p->laddr);
1122 int rtype = ipv6_addr_type(&p->raddr);
1124 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV);
1126 if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
1127 rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
1128 !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
1129 (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
1130 if (ltype&IPV6_ADDR_UNICAST)
1131 p->flags |= IP6_TNL_F_CAP_XMIT;
1132 if (rtype&IPV6_ADDR_UNICAST)
1133 p->flags |= IP6_TNL_F_CAP_RCV;
1137 static void ip6_tnl_link_config(struct ip6_tnl *t)
1139 struct net_device *dev = t->dev;
1140 struct ip6_tnl_parm *p = &t->parms;
1141 struct flowi6 *fl6 = &t->fl.u.ip6;
1143 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1144 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1146 /* Set up flowi template */
1147 ipv6_addr_copy(&fl6->saddr, &p->laddr);
1148 ipv6_addr_copy(&fl6->daddr, &p->raddr);
1149 fl6->flowi6_oif = p->link;
1152 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1153 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1154 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1155 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1159 if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1160 dev->flags |= IFF_POINTOPOINT;
1162 dev->flags &= ~IFF_POINTOPOINT;
1164 dev->iflink = p->link;
1166 if (p->flags & IP6_TNL_F_CAP_XMIT) {
1167 int strict = (ipv6_addr_type(&p->raddr) &
1168 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1170 struct rt6_info *rt = rt6_lookup(dev_net(dev),
1171 &p->raddr, &p->laddr,
1178 dev->hard_header_len = rt->rt6i_dev->hard_header_len +
1179 sizeof (struct ipv6hdr);
1181 dev->mtu = rt->rt6i_dev->mtu - sizeof (struct ipv6hdr);
1182 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1185 if (dev->mtu < IPV6_MIN_MTU)
1186 dev->mtu = IPV6_MIN_MTU;
1188 dst_release(&rt->dst);
1193 * ip6_tnl_change - update the tunnel parameters
1194 * @t: tunnel to be changed
1195 * @p: tunnel configuration parameters
1198 * ip6_tnl_change() updates the tunnel parameters
1202 ip6_tnl_change(struct ip6_tnl *t, struct ip6_tnl_parm *p)
1204 ipv6_addr_copy(&t->parms.laddr, &p->laddr);
1205 ipv6_addr_copy(&t->parms.raddr, &p->raddr);
1206 t->parms.flags = p->flags;
1207 t->parms.hop_limit = p->hop_limit;
1208 t->parms.encap_limit = p->encap_limit;
1209 t->parms.flowinfo = p->flowinfo;
1210 t->parms.link = p->link;
1211 t->parms.proto = p->proto;
1212 ip6_tnl_dst_reset(t);
1213 ip6_tnl_link_config(t);
1218 * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1219 * @dev: virtual device associated with tunnel
1220 * @ifr: parameters passed from userspace
1221 * @cmd: command to be performed
1224 * ip6_tnl_ioctl() is used for managing IPv6 tunnels
1227 * The possible commands are the following:
1228 * %SIOCGETTUNNEL: get tunnel parameters for device
1229 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1230 * %SIOCCHGTUNNEL: change tunnel parameters to those given
1231 * %SIOCDELTUNNEL: delete tunnel
1233 * The fallback device "ip6tnl0", created during module
1234 * initialization, can be used for creating other tunnel devices.
1238 * %-EFAULT if unable to copy data to or from userspace,
1239 * %-EPERM if current process hasn't %CAP_NET_ADMIN set
1240 * %-EINVAL if passed tunnel parameters are invalid,
1241 * %-EEXIST if changing a tunnel's parameters would cause a conflict
1242 * %-ENODEV if attempting to change or delete a nonexisting device
1246 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1249 struct ip6_tnl_parm p;
1250 struct ip6_tnl *t = NULL;
1251 struct net *net = dev_net(dev);
1252 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1256 if (dev == ip6n->fb_tnl_dev) {
1257 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p))) {
1261 t = ip6_tnl_locate(net, &p, 0);
1264 t = netdev_priv(dev);
1265 memcpy(&p, &t->parms, sizeof (p));
1266 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof (p))) {
1273 if (!capable(CAP_NET_ADMIN))
1276 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
1279 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1282 t = ip6_tnl_locate(net, &p, cmd == SIOCADDTUNNEL);
1283 if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) {
1285 if (t->dev != dev) {
1290 t = netdev_priv(dev);
1292 ip6_tnl_unlink(ip6n, t);
1294 err = ip6_tnl_change(t, &p);
1295 ip6_tnl_link(ip6n, t);
1296 netdev_state_change(dev);
1300 if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof (p)))
1304 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1308 if (!capable(CAP_NET_ADMIN))
1311 if (dev == ip6n->fb_tnl_dev) {
1313 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
1316 if ((t = ip6_tnl_locate(net, &p, 0)) == NULL)
1319 if (t->dev == ip6n->fb_tnl_dev)
1324 unregister_netdevice(dev);
1333 * ip6_tnl_change_mtu - change mtu manually for tunnel device
1334 * @dev: virtual device associated with tunnel
1335 * @new_mtu: the new mtu
1339 * %-EINVAL if mtu too small
1343 ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1345 if (new_mtu < IPV6_MIN_MTU) {
1353 static const struct net_device_ops ip6_tnl_netdev_ops = {
1354 .ndo_uninit = ip6_tnl_dev_uninit,
1355 .ndo_start_xmit = ip6_tnl_xmit,
1356 .ndo_do_ioctl = ip6_tnl_ioctl,
1357 .ndo_change_mtu = ip6_tnl_change_mtu,
1358 .ndo_get_stats = ip6_get_stats,
1363 * ip6_tnl_dev_setup - setup virtual tunnel device
1364 * @dev: virtual device associated with tunnel
1367 * Initialize function pointers and device parameters
1370 static void ip6_tnl_dev_setup(struct net_device *dev)
1374 dev->netdev_ops = &ip6_tnl_netdev_ops;
1375 dev->destructor = ip6_dev_free;
1377 dev->type = ARPHRD_TUNNEL6;
1378 dev->hard_header_len = LL_MAX_HEADER + sizeof (struct ipv6hdr);
1379 dev->mtu = ETH_DATA_LEN - sizeof (struct ipv6hdr);
1380 t = netdev_priv(dev);
1381 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1383 dev->flags |= IFF_NOARP;
1384 dev->addr_len = sizeof(struct in6_addr);
1385 dev->features |= NETIF_F_NETNS_LOCAL;
1386 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
1391 * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1392 * @dev: virtual device associated with tunnel
1396 ip6_tnl_dev_init_gen(struct net_device *dev)
1398 struct ip6_tnl *t = netdev_priv(dev);
1401 strcpy(t->parms.name, dev->name);
1402 dev->tstats = alloc_percpu(struct pcpu_tstats);
1409 * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1410 * @dev: virtual device associated with tunnel
1413 static int ip6_tnl_dev_init(struct net_device *dev)
1415 struct ip6_tnl *t = netdev_priv(dev);
1416 int err = ip6_tnl_dev_init_gen(dev);
1420 ip6_tnl_link_config(t);
1425 * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1426 * @dev: fallback device
1431 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1433 struct ip6_tnl *t = netdev_priv(dev);
1434 struct net *net = dev_net(dev);
1435 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1436 int err = ip6_tnl_dev_init_gen(dev);
1441 t->parms.proto = IPPROTO_IPV6;
1443 RCU_INIT_POINTER(ip6n->tnls_wc[0], t);
1447 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
1448 .handler = ip4ip6_rcv,
1449 .err_handler = ip4ip6_err,
1453 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
1454 .handler = ip6ip6_rcv,
1455 .err_handler = ip6ip6_err,
1459 static void __net_exit ip6_tnl_destroy_tunnels(struct ip6_tnl_net *ip6n)
1465 for (h = 0; h < HASH_SIZE; h++) {
1466 t = rtnl_dereference(ip6n->tnls_r_l[h]);
1468 unregister_netdevice_queue(t->dev, &list);
1469 t = rtnl_dereference(t->next);
1473 t = rtnl_dereference(ip6n->tnls_wc[0]);
1474 unregister_netdevice_queue(t->dev, &list);
1475 unregister_netdevice_many(&list);
1478 static int __net_init ip6_tnl_init_net(struct net *net)
1480 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1483 ip6n->tnls[0] = ip6n->tnls_wc;
1484 ip6n->tnls[1] = ip6n->tnls_r_l;
1487 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
1490 if (!ip6n->fb_tnl_dev)
1492 dev_net_set(ip6n->fb_tnl_dev, net);
1494 err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
1498 err = register_netdev(ip6n->fb_tnl_dev);
1504 ip6_dev_free(ip6n->fb_tnl_dev);
1509 static void __net_exit ip6_tnl_exit_net(struct net *net)
1511 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1514 ip6_tnl_destroy_tunnels(ip6n);
1518 static struct pernet_operations ip6_tnl_net_ops = {
1519 .init = ip6_tnl_init_net,
1520 .exit = ip6_tnl_exit_net,
1521 .id = &ip6_tnl_net_id,
1522 .size = sizeof(struct ip6_tnl_net),
1526 * ip6_tunnel_init - register protocol and reserve needed resources
1528 * Return: 0 on success
1531 static int __init ip6_tunnel_init(void)
1535 err = register_pernet_device(&ip6_tnl_net_ops);
1539 err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
1541 printk(KERN_ERR "ip6_tunnel init: can't register ip4ip6\n");
1545 err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
1547 printk(KERN_ERR "ip6_tunnel init: can't register ip6ip6\n");
1554 xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
1556 unregister_pernet_device(&ip6_tnl_net_ops);
1562 * ip6_tunnel_cleanup - free resources and unregister protocol
1565 static void __exit ip6_tunnel_cleanup(void)
1567 if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
1568 printk(KERN_INFO "ip6_tunnel close: can't deregister ip4ip6\n");
1570 if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
1571 printk(KERN_INFO "ip6_tunnel close: can't deregister ip6ip6\n");
1573 unregister_pernet_device(&ip6_tnl_net_ops);
1576 module_init(ip6_tunnel_init);
1577 module_exit(ip6_tunnel_cleanup);