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_RTNL_LINK("ip6tnl");
61 MODULE_ALIAS_NETDEV("ip6tnl0");
64 #define IP6_TNL_TRACE(x...) printk(KERN_DEBUG "%s:" x "\n", __func__)
66 #define IP6_TNL_TRACE(x...) do {;} while(0)
69 #define IPV6_TCLASS_MASK (IPV6_FLOWINFO_MASK & ~IPV6_FLOWLABEL_MASK)
70 #define IPV6_TCLASS_SHIFT 20
74 #define HASH(addr) ((__force u32)((addr)->s6_addr32[0] ^ (addr)->s6_addr32[1] ^ \
75 (addr)->s6_addr32[2] ^ (addr)->s6_addr32[3]) & \
78 static int ip6_tnl_dev_init(struct net_device *dev);
79 static void ip6_tnl_dev_setup(struct net_device *dev);
81 static int ip6_tnl_net_id __read_mostly;
83 /* the IPv6 tunnel fallback device */
84 struct net_device *fb_tnl_dev;
85 /* lists for storing tunnels in use */
86 struct ip6_tnl __rcu *tnls_r_l[HASH_SIZE];
87 struct ip6_tnl __rcu *tnls_wc[1];
88 struct ip6_tnl __rcu **tnls[2];
91 /* often modified stats are per cpu, other are shared (netdev->stats) */
93 unsigned long rx_packets;
94 unsigned long rx_bytes;
95 unsigned long tx_packets;
96 unsigned long tx_bytes;
99 static struct net_device_stats *ip6_get_stats(struct net_device *dev)
101 struct pcpu_tstats sum = { 0 };
104 for_each_possible_cpu(i) {
105 const struct pcpu_tstats *tstats = per_cpu_ptr(dev->tstats, i);
107 sum.rx_packets += tstats->rx_packets;
108 sum.rx_bytes += tstats->rx_bytes;
109 sum.tx_packets += tstats->tx_packets;
110 sum.tx_bytes += tstats->tx_bytes;
112 dev->stats.rx_packets = sum.rx_packets;
113 dev->stats.rx_bytes = sum.rx_bytes;
114 dev->stats.tx_packets = sum.tx_packets;
115 dev->stats.tx_bytes = sum.tx_bytes;
120 * Locking : hash tables are protected by RCU and RTNL
123 static inline struct dst_entry *ip6_tnl_dst_check(struct ip6_tnl *t)
125 struct dst_entry *dst = t->dst_cache;
127 if (dst && dst->obsolete &&
128 dst->ops->check(dst, t->dst_cookie) == NULL) {
137 static inline void ip6_tnl_dst_reset(struct ip6_tnl *t)
139 dst_release(t->dst_cache);
143 static inline void ip6_tnl_dst_store(struct ip6_tnl *t, struct dst_entry *dst)
145 struct rt6_info *rt = (struct rt6_info *) dst;
146 t->dst_cookie = rt->rt6i_node ? rt->rt6i_node->fn_sernum : 0;
147 dst_release(t->dst_cache);
152 * ip6_tnl_lookup - fetch tunnel matching the end-point addresses
153 * @remote: the address of the tunnel exit-point
154 * @local: the address of the tunnel entry-point
157 * tunnel matching given end-points if found,
158 * else fallback tunnel if its device is up,
162 #define for_each_ip6_tunnel_rcu(start) \
163 for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
165 static struct ip6_tnl *
166 ip6_tnl_lookup(struct net *net, const struct in6_addr *remote, const struct in6_addr *local)
168 unsigned int h0 = HASH(remote);
169 unsigned int h1 = HASH(local);
171 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
173 for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[h0 ^ h1]) {
174 if (ipv6_addr_equal(local, &t->parms.laddr) &&
175 ipv6_addr_equal(remote, &t->parms.raddr) &&
176 (t->dev->flags & IFF_UP))
179 t = rcu_dereference(ip6n->tnls_wc[0]);
180 if (t && (t->dev->flags & IFF_UP))
187 * ip6_tnl_bucket - get head of list matching given tunnel parameters
188 * @p: parameters containing tunnel end-points
191 * ip6_tnl_bucket() returns the head of the list matching the
192 * &struct in6_addr entries laddr and raddr in @p.
194 * Return: head of IPv6 tunnel list
197 static struct ip6_tnl __rcu **
198 ip6_tnl_bucket(struct ip6_tnl_net *ip6n, const struct ip6_tnl_parm *p)
200 const struct in6_addr *remote = &p->raddr;
201 const struct in6_addr *local = &p->laddr;
205 if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
207 h = HASH(remote) ^ HASH(local);
209 return &ip6n->tnls[prio][h];
213 * ip6_tnl_link - add tunnel to hash table
214 * @t: tunnel to be added
218 ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
220 struct ip6_tnl __rcu **tp = ip6_tnl_bucket(ip6n, &t->parms);
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 for (tp = ip6_tnl_bucket(ip6n, &t->parms);
238 (iter = rtnl_dereference(*tp)) != NULL;
241 rcu_assign_pointer(*tp, t->next);
247 static void ip6_dev_free(struct net_device *dev)
249 free_percpu(dev->tstats);
254 * ip6_tnl_create() - create a new tunnel
255 * @p: tunnel parameters
256 * @pt: pointer to new tunnel
259 * Create tunnel matching given parameters.
262 * created tunnel or NULL
265 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct ip6_tnl_parm *p)
267 struct net_device *dev;
271 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
274 strlcpy(name, p->name, IFNAMSIZ);
276 sprintf(name, "ip6tnl%%d");
278 dev = alloc_netdev(sizeof (*t), name, ip6_tnl_dev_setup);
282 dev_net_set(dev, net);
284 t = netdev_priv(dev);
286 err = ip6_tnl_dev_init(dev);
290 if ((err = register_netdevice(dev)) < 0)
293 strcpy(t->parms.name, dev->name);
296 ip6_tnl_link(ip6n, t);
306 * ip6_tnl_locate - find or create tunnel matching given parameters
307 * @p: tunnel parameters
308 * @create: != 0 if allowed to create new tunnel if no match found
311 * ip6_tnl_locate() first tries to locate an existing tunnel
312 * based on @parms. If this is unsuccessful, but @create is set a new
313 * tunnel device is created and registered for use.
316 * matching tunnel or NULL
319 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
320 struct ip6_tnl_parm *p, int create)
322 const struct in6_addr *remote = &p->raddr;
323 const struct in6_addr *local = &p->laddr;
324 struct ip6_tnl __rcu **tp;
326 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
328 for (tp = ip6_tnl_bucket(ip6n, p);
329 (t = rtnl_dereference(*tp)) != NULL;
331 if (ipv6_addr_equal(local, &t->parms.laddr) &&
332 ipv6_addr_equal(remote, &t->parms.raddr))
337 return ip6_tnl_create(net, p);
341 * ip6_tnl_dev_uninit - tunnel device uninitializer
342 * @dev: the device to be destroyed
345 * ip6_tnl_dev_uninit() removes tunnel from its list
349 ip6_tnl_dev_uninit(struct net_device *dev)
351 struct ip6_tnl *t = netdev_priv(dev);
352 struct net *net = dev_net(dev);
353 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
355 if (dev == ip6n->fb_tnl_dev)
356 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
358 ip6_tnl_unlink(ip6n, t);
359 ip6_tnl_dst_reset(t);
364 * parse_tvl_tnl_enc_lim - handle encapsulation limit option
365 * @skb: received socket buffer
368 * 0 if none was found,
369 * else index to encapsulation limit
373 parse_tlv_tnl_enc_lim(struct sk_buff *skb, __u8 * raw)
375 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) raw;
376 __u8 nexthdr = ipv6h->nexthdr;
377 __u16 off = sizeof (*ipv6h);
379 while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
381 struct ipv6_opt_hdr *hdr;
382 if (raw + off + sizeof (*hdr) > skb->data &&
383 !pskb_may_pull(skb, raw - skb->data + off + sizeof (*hdr)))
386 hdr = (struct ipv6_opt_hdr *) (raw + off);
387 if (nexthdr == NEXTHDR_FRAGMENT) {
388 struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
389 if (frag_hdr->frag_off)
392 } else if (nexthdr == NEXTHDR_AUTH) {
393 optlen = (hdr->hdrlen + 2) << 2;
395 optlen = ipv6_optlen(hdr);
397 if (nexthdr == NEXTHDR_DEST) {
400 struct ipv6_tlv_tnl_enc_lim *tel;
402 /* No more room for encapsulation limit */
403 if (i + sizeof (*tel) > off + optlen)
406 tel = (struct ipv6_tlv_tnl_enc_lim *) &raw[i];
407 /* return index of option if found and valid */
408 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
411 /* else jump to next option */
413 i += tel->length + 2;
418 nexthdr = hdr->nexthdr;
425 * ip6_tnl_err - tunnel error handler
428 * ip6_tnl_err() should handle errors in the tunnel according
429 * to the specifications in RFC 2473.
433 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
434 u8 *type, u8 *code, int *msg, __u32 *info, int offset)
436 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) skb->data;
439 u8 rel_type = ICMPV6_DEST_UNREACH;
440 u8 rel_code = ICMPV6_ADDR_UNREACH;
445 /* If the packet doesn't contain the original IPv6 header we are
446 in trouble since we might need the source address for further
447 processing of the error. */
450 if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr,
451 &ipv6h->saddr)) == NULL)
454 if (t->parms.proto != ipproto && t->parms.proto != 0)
461 struct ipv6_tlv_tnl_enc_lim *tel;
463 case ICMPV6_DEST_UNREACH:
466 "%s: Path to destination invalid "
467 "or inactive!\n", t->parms.name);
470 case ICMPV6_TIME_EXCEED:
471 if ((*code) == ICMPV6_EXC_HOPLIMIT) {
474 "%s: Too small hop limit or "
475 "routing loop in tunnel!\n",
480 case ICMPV6_PARAMPROB:
482 if ((*code) == ICMPV6_HDR_FIELD)
483 teli = parse_tlv_tnl_enc_lim(skb, skb->data);
485 if (teli && teli == *info - 2) {
486 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
487 if (tel->encap_limit == 0) {
490 "%s: Too small encapsulation "
491 "limit or routing loop in "
492 "tunnel!\n", t->parms.name);
495 } else if (net_ratelimit()) {
497 "%s: Recipient unable to parse tunneled "
498 "packet!\n ", t->parms.name);
501 case ICMPV6_PKT_TOOBIG:
502 mtu = *info - offset;
503 if (mtu < IPV6_MIN_MTU)
507 if ((len = sizeof (*ipv6h) + ntohs(ipv6h->payload_len)) > mtu) {
508 rel_type = ICMPV6_PKT_TOOBIG;
527 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
528 u8 type, u8 code, int offset, __be32 info)
533 __u32 rel_info = ntohl(info);
535 struct sk_buff *skb2;
536 const struct iphdr *eiph;
540 err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
541 &rel_msg, &rel_info, offset);
549 case ICMPV6_DEST_UNREACH:
550 if (rel_code != ICMPV6_ADDR_UNREACH)
552 rel_type = ICMP_DEST_UNREACH;
553 rel_code = ICMP_HOST_UNREACH;
555 case ICMPV6_PKT_TOOBIG:
558 rel_type = ICMP_DEST_UNREACH;
559 rel_code = ICMP_FRAG_NEEDED;
565 if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
568 skb2 = skb_clone(skb, GFP_ATOMIC);
574 skb_pull(skb2, offset);
575 skb_reset_network_header(skb2);
578 /* Try to guess incoming interface */
579 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
582 IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
586 skb2->dev = rt->dst.dev;
588 /* route "incoming" packet */
589 if (rt->rt_flags & RTCF_LOCAL) {
592 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
593 eiph->daddr, eiph->saddr,
596 RT_TOS(eiph->tos), 0);
598 rt->dst.dev->type != ARPHRD_TUNNEL) {
603 skb_dst_set(skb2, &rt->dst);
606 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
608 skb_dst(skb2)->dev->type != ARPHRD_TUNNEL)
612 /* change mtu on this route */
613 if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
614 if (rel_info > dst_mtu(skb_dst(skb2)))
617 skb_dst(skb2)->ops->update_pmtu(skb_dst(skb2), rel_info);
620 icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
628 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
629 u8 type, u8 code, int offset, __be32 info)
634 __u32 rel_info = ntohl(info);
637 err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
638 &rel_msg, &rel_info, offset);
642 if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
644 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
650 skb_pull(skb2, offset);
651 skb_reset_network_header(skb2);
653 /* Try to guess incoming interface */
654 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
657 if (rt && rt->rt6i_dev)
658 skb2->dev = rt->rt6i_dev;
660 icmpv6_send(skb2, rel_type, rel_code, rel_info);
663 dst_release(&rt->dst);
671 static void ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
672 const struct ipv6hdr *ipv6h,
675 __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
677 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
678 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
680 if (INET_ECN_is_ce(dsfield))
681 IP_ECN_set_ce(ip_hdr(skb));
684 static void ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
685 const struct ipv6hdr *ipv6h,
688 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
689 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
691 if (INET_ECN_is_ce(ipv6_get_dsfield(ipv6h)))
692 IP6_ECN_set_ce(skb, ipv6_hdr(skb));
695 /* called with rcu_read_lock() */
696 static inline int ip6_tnl_rcv_ctl(struct ip6_tnl *t)
698 struct ip6_tnl_parm *p = &t->parms;
700 struct net *net = dev_net(t->dev);
702 if (p->flags & IP6_TNL_F_CAP_RCV) {
703 struct net_device *ldev = NULL;
706 ldev = dev_get_by_index_rcu(net, p->link);
708 if ((ipv6_addr_is_multicast(&p->laddr) ||
709 likely(ipv6_chk_addr(net, &p->laddr, ldev, 0))) &&
710 likely(!ipv6_chk_addr(net, &p->raddr, NULL, 0)))
718 * ip6_tnl_rcv - decapsulate IPv6 packet and retransmit it locally
719 * @skb: received socket buffer
720 * @protocol: ethernet protocol ID
721 * @dscp_ecn_decapsulate: the function to decapsulate DSCP code and ECN
726 static int ip6_tnl_rcv(struct sk_buff *skb, __u16 protocol,
728 void (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
729 const struct ipv6hdr *ipv6h,
730 struct sk_buff *skb))
733 const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
737 if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr,
738 &ipv6h->daddr)) != NULL) {
739 struct pcpu_tstats *tstats;
741 if (t->parms.proto != ipproto && t->parms.proto != 0) {
746 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
751 if (!ip6_tnl_rcv_ctl(t)) {
752 t->dev->stats.rx_dropped++;
757 skb->mac_header = skb->network_header;
758 skb_reset_network_header(skb);
759 skb->protocol = htons(protocol);
760 skb->pkt_type = PACKET_HOST;
761 memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
763 tstats = this_cpu_ptr(t->dev->tstats);
764 tstats->rx_packets++;
765 tstats->rx_bytes += skb->len;
767 __skb_tunnel_rx(skb, t->dev);
769 dscp_ecn_decapsulate(t, ipv6h, skb);
784 static int ip4ip6_rcv(struct sk_buff *skb)
786 return ip6_tnl_rcv(skb, ETH_P_IP, IPPROTO_IPIP,
787 ip4ip6_dscp_ecn_decapsulate);
790 static int ip6ip6_rcv(struct sk_buff *skb)
792 return ip6_tnl_rcv(skb, ETH_P_IPV6, IPPROTO_IPV6,
793 ip6ip6_dscp_ecn_decapsulate);
796 struct ipv6_tel_txoption {
797 struct ipv6_txoptions ops;
801 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
803 memset(opt, 0, sizeof(struct ipv6_tel_txoption));
805 opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
807 opt->dst_opt[4] = encap_limit;
808 opt->dst_opt[5] = IPV6_TLV_PADN;
811 opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
812 opt->ops.opt_nflen = 8;
816 * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
817 * @t: the outgoing tunnel device
818 * @hdr: IPv6 header from the incoming packet
821 * Avoid trivial tunneling loop by checking that tunnel exit-point
822 * doesn't match source of incoming packet.
830 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
832 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
835 static inline int ip6_tnl_xmit_ctl(struct ip6_tnl *t)
837 struct ip6_tnl_parm *p = &t->parms;
839 struct net *net = dev_net(t->dev);
841 if (p->flags & IP6_TNL_F_CAP_XMIT) {
842 struct net_device *ldev = NULL;
846 ldev = dev_get_by_index_rcu(net, p->link);
848 if (unlikely(!ipv6_chk_addr(net, &p->laddr, ldev, 0)))
850 "%s xmit: Local address not yet configured!\n",
852 else if (!ipv6_addr_is_multicast(&p->raddr) &&
853 unlikely(ipv6_chk_addr(net, &p->raddr, NULL, 0)))
855 "%s xmit: Routing loop! "
856 "Remote address found on this node!\n",
865 * ip6_tnl_xmit2 - encapsulate packet and send
866 * @skb: the outgoing socket buffer
867 * @dev: the outgoing tunnel device
868 * @dsfield: dscp code for outer header
869 * @fl: flow of tunneled packet
870 * @encap_limit: encapsulation limit
871 * @pmtu: Path MTU is stored if packet is too big
874 * Build new header and do some sanity checks on the packet before sending
880 * %-EMSGSIZE message too big. return mtu in this case.
883 static int ip6_tnl_xmit2(struct sk_buff *skb,
884 struct net_device *dev,
890 struct net *net = dev_net(dev);
891 struct ip6_tnl *t = netdev_priv(dev);
892 struct net_device_stats *stats = &t->dev->stats;
893 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
894 struct ipv6_tel_txoption opt;
895 struct dst_entry *dst = NULL, *ndst = NULL;
896 struct net_device *tdev;
898 unsigned int max_headroom = sizeof(struct ipv6hdr);
903 if (!fl6->flowi6_mark)
904 dst = ip6_tnl_dst_check(t);
906 ndst = ip6_route_output(net, NULL, fl6);
909 goto tx_err_link_failure;
910 ndst = xfrm_lookup(net, ndst, flowi6_to_flowi(fl6), NULL, 0);
914 goto tx_err_link_failure;
925 "%s: Local routing loop detected!\n",
927 goto tx_err_dst_release;
929 mtu = dst_mtu(dst) - sizeof (*ipv6h);
930 if (encap_limit >= 0) {
934 if (mtu < IPV6_MIN_MTU)
937 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), mtu);
938 if (skb->len > mtu) {
941 goto tx_err_dst_release;
945 * Okay, now see if we can stuff it in the buffer as-is.
947 max_headroom += LL_RESERVED_SPACE(tdev);
949 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
950 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
951 struct sk_buff *new_skb;
953 if (!(new_skb = skb_realloc_headroom(skb, max_headroom)))
954 goto tx_err_dst_release;
957 skb_set_owner_w(new_skb, skb->sk);
962 if (fl6->flowi6_mark) {
963 skb_dst_set(skb, dst);
966 skb_dst_set_noref(skb, dst);
968 skb->transport_header = skb->network_header;
970 proto = fl6->flowi6_proto;
971 if (encap_limit >= 0) {
972 init_tel_txopt(&opt, encap_limit);
973 ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
975 skb_push(skb, sizeof(struct ipv6hdr));
976 skb_reset_network_header(skb);
977 ipv6h = ipv6_hdr(skb);
978 *(__be32*)ipv6h = fl6->flowlabel | htonl(0x60000000);
979 dsfield = INET_ECN_encapsulate(0, dsfield);
980 ipv6_change_dsfield(ipv6h, ~INET_ECN_MASK, dsfield);
981 ipv6h->hop_limit = t->parms.hop_limit;
982 ipv6h->nexthdr = proto;
983 ipv6_addr_copy(&ipv6h->saddr, &fl6->saddr);
984 ipv6_addr_copy(&ipv6h->daddr, &fl6->daddr);
987 err = ip6_local_out(skb);
989 if (net_xmit_eval(err) == 0) {
990 struct pcpu_tstats *tstats = this_cpu_ptr(t->dev->tstats);
992 tstats->tx_bytes += pkt_len;
993 tstats->tx_packets++;
996 stats->tx_aborted_errors++;
999 ip6_tnl_dst_store(t, ndst);
1001 tx_err_link_failure:
1002 stats->tx_carrier_errors++;
1003 dst_link_failure(skb);
1010 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1012 struct ip6_tnl *t = netdev_priv(dev);
1013 const struct iphdr *iph = ip_hdr(skb);
1014 int encap_limit = -1;
1020 if ((t->parms.proto != IPPROTO_IPIP && t->parms.proto != 0) ||
1021 !ip6_tnl_xmit_ctl(t))
1024 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1025 encap_limit = t->parms.encap_limit;
1027 memcpy(&fl6, &t->fl.u.ip6, sizeof (fl6));
1028 fl6.flowi6_proto = IPPROTO_IPIP;
1030 dsfield = ipv4_get_dsfield(iph);
1032 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1033 fl6.flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
1035 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1036 fl6.flowi6_mark = skb->mark;
1038 err = ip6_tnl_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
1040 /* XXX: send ICMP error even if DF is not set. */
1041 if (err == -EMSGSIZE)
1042 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
1051 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1053 struct ip6_tnl *t = netdev_priv(dev);
1054 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
1055 int encap_limit = -1;
1062 if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) ||
1063 !ip6_tnl_xmit_ctl(t) || ip6_tnl_addr_conflict(t, ipv6h))
1066 offset = parse_tlv_tnl_enc_lim(skb, skb_network_header(skb));
1068 struct ipv6_tlv_tnl_enc_lim *tel;
1069 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
1070 if (tel->encap_limit == 0) {
1071 icmpv6_send(skb, ICMPV6_PARAMPROB,
1072 ICMPV6_HDR_FIELD, offset + 2);
1075 encap_limit = tel->encap_limit - 1;
1076 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1077 encap_limit = t->parms.encap_limit;
1079 memcpy(&fl6, &t->fl.u.ip6, sizeof (fl6));
1080 fl6.flowi6_proto = IPPROTO_IPV6;
1082 dsfield = ipv6_get_dsfield(ipv6h);
1083 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1084 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
1085 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
1086 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_FLOWLABEL_MASK);
1087 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1088 fl6.flowi6_mark = skb->mark;
1090 err = ip6_tnl_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
1092 if (err == -EMSGSIZE)
1093 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1101 ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1103 struct ip6_tnl *t = netdev_priv(dev);
1104 struct net_device_stats *stats = &t->dev->stats;
1107 switch (skb->protocol) {
1108 case htons(ETH_P_IP):
1109 ret = ip4ip6_tnl_xmit(skb, dev);
1111 case htons(ETH_P_IPV6):
1112 ret = ip6ip6_tnl_xmit(skb, dev);
1121 return NETDEV_TX_OK;
1125 stats->tx_dropped++;
1127 return NETDEV_TX_OK;
1130 static void ip6_tnl_set_cap(struct ip6_tnl *t)
1132 struct ip6_tnl_parm *p = &t->parms;
1133 int ltype = ipv6_addr_type(&p->laddr);
1134 int rtype = ipv6_addr_type(&p->raddr);
1136 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV);
1138 if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
1139 rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
1140 !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
1141 (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
1142 if (ltype&IPV6_ADDR_UNICAST)
1143 p->flags |= IP6_TNL_F_CAP_XMIT;
1144 if (rtype&IPV6_ADDR_UNICAST)
1145 p->flags |= IP6_TNL_F_CAP_RCV;
1149 static void ip6_tnl_link_config(struct ip6_tnl *t)
1151 struct net_device *dev = t->dev;
1152 struct ip6_tnl_parm *p = &t->parms;
1153 struct flowi6 *fl6 = &t->fl.u.ip6;
1155 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1156 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1158 /* Set up flowi template */
1159 ipv6_addr_copy(&fl6->saddr, &p->laddr);
1160 ipv6_addr_copy(&fl6->daddr, &p->raddr);
1161 fl6->flowi6_oif = p->link;
1164 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1165 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1166 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1167 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1171 if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1172 dev->flags |= IFF_POINTOPOINT;
1174 dev->flags &= ~IFF_POINTOPOINT;
1176 dev->iflink = p->link;
1178 if (p->flags & IP6_TNL_F_CAP_XMIT) {
1179 int strict = (ipv6_addr_type(&p->raddr) &
1180 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1182 struct rt6_info *rt = rt6_lookup(dev_net(dev),
1183 &p->raddr, &p->laddr,
1190 dev->hard_header_len = rt->rt6i_dev->hard_header_len +
1191 sizeof (struct ipv6hdr);
1193 dev->mtu = rt->rt6i_dev->mtu - sizeof (struct ipv6hdr);
1194 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1197 if (dev->mtu < IPV6_MIN_MTU)
1198 dev->mtu = IPV6_MIN_MTU;
1200 dst_release(&rt->dst);
1205 * ip6_tnl_change - update the tunnel parameters
1206 * @t: tunnel to be changed
1207 * @p: tunnel configuration parameters
1210 * ip6_tnl_change() updates the tunnel parameters
1214 ip6_tnl_change(struct ip6_tnl *t, struct ip6_tnl_parm *p)
1216 ipv6_addr_copy(&t->parms.laddr, &p->laddr);
1217 ipv6_addr_copy(&t->parms.raddr, &p->raddr);
1218 t->parms.flags = p->flags;
1219 t->parms.hop_limit = p->hop_limit;
1220 t->parms.encap_limit = p->encap_limit;
1221 t->parms.flowinfo = p->flowinfo;
1222 t->parms.link = p->link;
1223 t->parms.proto = p->proto;
1224 ip6_tnl_dst_reset(t);
1225 ip6_tnl_link_config(t);
1230 * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1231 * @dev: virtual device associated with tunnel
1232 * @ifr: parameters passed from userspace
1233 * @cmd: command to be performed
1236 * ip6_tnl_ioctl() is used for managing IPv6 tunnels
1239 * The possible commands are the following:
1240 * %SIOCGETTUNNEL: get tunnel parameters for device
1241 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1242 * %SIOCCHGTUNNEL: change tunnel parameters to those given
1243 * %SIOCDELTUNNEL: delete tunnel
1245 * The fallback device "ip6tnl0", created during module
1246 * initialization, can be used for creating other tunnel devices.
1250 * %-EFAULT if unable to copy data to or from userspace,
1251 * %-EPERM if current process hasn't %CAP_NET_ADMIN set
1252 * %-EINVAL if passed tunnel parameters are invalid,
1253 * %-EEXIST if changing a tunnel's parameters would cause a conflict
1254 * %-ENODEV if attempting to change or delete a nonexisting device
1258 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1261 struct ip6_tnl_parm p;
1262 struct ip6_tnl *t = NULL;
1263 struct net *net = dev_net(dev);
1264 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1268 if (dev == ip6n->fb_tnl_dev) {
1269 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p))) {
1273 t = ip6_tnl_locate(net, &p, 0);
1276 t = netdev_priv(dev);
1277 memcpy(&p, &t->parms, sizeof (p));
1278 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof (p))) {
1285 if (!capable(CAP_NET_ADMIN))
1288 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
1291 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1294 t = ip6_tnl_locate(net, &p, cmd == SIOCADDTUNNEL);
1295 if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) {
1297 if (t->dev != dev) {
1302 t = netdev_priv(dev);
1304 ip6_tnl_unlink(ip6n, t);
1306 err = ip6_tnl_change(t, &p);
1307 ip6_tnl_link(ip6n, t);
1308 netdev_state_change(dev);
1312 if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof (p)))
1316 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1320 if (!capable(CAP_NET_ADMIN))
1323 if (dev == ip6n->fb_tnl_dev) {
1325 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
1328 if ((t = ip6_tnl_locate(net, &p, 0)) == NULL)
1331 if (t->dev == ip6n->fb_tnl_dev)
1336 unregister_netdevice(dev);
1345 * ip6_tnl_change_mtu - change mtu manually for tunnel device
1346 * @dev: virtual device associated with tunnel
1347 * @new_mtu: the new mtu
1351 * %-EINVAL if mtu too small
1355 ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1357 if (new_mtu < IPV6_MIN_MTU) {
1365 static const struct net_device_ops ip6_tnl_netdev_ops = {
1366 .ndo_uninit = ip6_tnl_dev_uninit,
1367 .ndo_start_xmit = ip6_tnl_xmit,
1368 .ndo_do_ioctl = ip6_tnl_ioctl,
1369 .ndo_change_mtu = ip6_tnl_change_mtu,
1370 .ndo_get_stats = ip6_get_stats,
1375 * ip6_tnl_dev_setup - setup virtual tunnel device
1376 * @dev: virtual device associated with tunnel
1379 * Initialize function pointers and device parameters
1382 static void ip6_tnl_dev_setup(struct net_device *dev)
1386 dev->netdev_ops = &ip6_tnl_netdev_ops;
1387 dev->destructor = ip6_dev_free;
1389 dev->type = ARPHRD_TUNNEL6;
1390 dev->hard_header_len = LL_MAX_HEADER + sizeof (struct ipv6hdr);
1391 dev->mtu = ETH_DATA_LEN - sizeof (struct ipv6hdr);
1392 t = netdev_priv(dev);
1393 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1395 dev->flags |= IFF_NOARP;
1396 dev->addr_len = sizeof(struct in6_addr);
1397 dev->features |= NETIF_F_NETNS_LOCAL;
1398 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
1403 * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1404 * @dev: virtual device associated with tunnel
1408 ip6_tnl_dev_init_gen(struct net_device *dev)
1410 struct ip6_tnl *t = netdev_priv(dev);
1413 dev->tstats = alloc_percpu(struct pcpu_tstats);
1420 * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1421 * @dev: virtual device associated with tunnel
1424 static int ip6_tnl_dev_init(struct net_device *dev)
1426 struct ip6_tnl *t = netdev_priv(dev);
1427 int err = ip6_tnl_dev_init_gen(dev);
1431 ip6_tnl_link_config(t);
1436 * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1437 * @dev: fallback device
1442 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1444 struct ip6_tnl *t = netdev_priv(dev);
1445 struct net *net = dev_net(dev);
1446 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1447 int err = ip6_tnl_dev_init_gen(dev);
1452 t->parms.proto = IPPROTO_IPV6;
1454 rcu_assign_pointer(ip6n->tnls_wc[0], t);
1458 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
1459 .handler = ip4ip6_rcv,
1460 .err_handler = ip4ip6_err,
1464 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
1465 .handler = ip6ip6_rcv,
1466 .err_handler = ip6ip6_err,
1470 static void __net_exit ip6_tnl_destroy_tunnels(struct ip6_tnl_net *ip6n)
1476 for (h = 0; h < HASH_SIZE; h++) {
1477 t = rtnl_dereference(ip6n->tnls_r_l[h]);
1479 unregister_netdevice_queue(t->dev, &list);
1480 t = rtnl_dereference(t->next);
1484 t = rtnl_dereference(ip6n->tnls_wc[0]);
1485 unregister_netdevice_queue(t->dev, &list);
1486 unregister_netdevice_many(&list);
1489 static int __net_init ip6_tnl_init_net(struct net *net)
1491 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1492 struct ip6_tnl *t = NULL;
1495 ip6n->tnls[0] = ip6n->tnls_wc;
1496 ip6n->tnls[1] = ip6n->tnls_r_l;
1499 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
1502 if (!ip6n->fb_tnl_dev)
1504 dev_net_set(ip6n->fb_tnl_dev, net);
1506 err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
1510 err = register_netdev(ip6n->fb_tnl_dev);
1514 t = netdev_priv(ip6n->fb_tnl_dev);
1516 strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
1520 ip6_dev_free(ip6n->fb_tnl_dev);
1525 static void __net_exit ip6_tnl_exit_net(struct net *net)
1527 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1530 ip6_tnl_destroy_tunnels(ip6n);
1534 static struct pernet_operations ip6_tnl_net_ops = {
1535 .init = ip6_tnl_init_net,
1536 .exit = ip6_tnl_exit_net,
1537 .id = &ip6_tnl_net_id,
1538 .size = sizeof(struct ip6_tnl_net),
1542 * ip6_tunnel_init - register protocol and reserve needed resources
1544 * Return: 0 on success
1547 static int __init ip6_tunnel_init(void)
1551 err = register_pernet_device(&ip6_tnl_net_ops);
1555 err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
1557 printk(KERN_ERR "ip6_tunnel init: can't register ip4ip6\n");
1561 err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
1563 printk(KERN_ERR "ip6_tunnel init: can't register ip6ip6\n");
1570 xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
1572 unregister_pernet_device(&ip6_tnl_net_ops);
1578 * ip6_tunnel_cleanup - free resources and unregister protocol
1581 static void __exit ip6_tunnel_cleanup(void)
1583 if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
1584 printk(KERN_INFO "ip6_tunnel close: can't deregister ip4ip6\n");
1586 if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
1587 printk(KERN_INFO "ip6_tunnel close: can't deregister ip6ip6\n");
1589 unregister_pernet_device(&ip6_tnl_net_ops);
1592 module_init(ip6_tunnel_init);
1593 module_exit(ip6_tunnel_cleanup);