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 <asm/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");
61 #define IPV6_TLV_TEL_DST_SIZE 8
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 void 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];
92 * Locking : hash tables are protected by RCU and RTNL
95 static inline struct dst_entry *ip6_tnl_dst_check(struct ip6_tnl *t)
97 struct dst_entry *dst = t->dst_cache;
99 if (dst && dst->obsolete &&
100 dst->ops->check(dst, t->dst_cookie) == NULL) {
109 static inline void ip6_tnl_dst_reset(struct ip6_tnl *t)
111 dst_release(t->dst_cache);
115 static inline void ip6_tnl_dst_store(struct ip6_tnl *t, struct dst_entry *dst)
117 struct rt6_info *rt = (struct rt6_info *) dst;
118 t->dst_cookie = rt->rt6i_node ? rt->rt6i_node->fn_sernum : 0;
119 dst_release(t->dst_cache);
124 * ip6_tnl_lookup - fetch tunnel matching the end-point addresses
125 * @remote: the address of the tunnel exit-point
126 * @local: the address of the tunnel entry-point
129 * tunnel matching given end-points if found,
130 * else fallback tunnel if its device is up,
134 #define for_each_ip6_tunnel_rcu(start) \
135 for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
137 static struct ip6_tnl *
138 ip6_tnl_lookup(struct net *net, struct in6_addr *remote, struct in6_addr *local)
140 unsigned int h0 = HASH(remote);
141 unsigned int h1 = HASH(local);
143 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
145 for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[h0 ^ h1]) {
146 if (ipv6_addr_equal(local, &t->parms.laddr) &&
147 ipv6_addr_equal(remote, &t->parms.raddr) &&
148 (t->dev->flags & IFF_UP))
151 t = rcu_dereference(ip6n->tnls_wc[0]);
152 if (t && (t->dev->flags & IFF_UP))
159 * ip6_tnl_bucket - get head of list matching given tunnel parameters
160 * @p: parameters containing tunnel end-points
163 * ip6_tnl_bucket() returns the head of the list matching the
164 * &struct in6_addr entries laddr and raddr in @p.
166 * Return: head of IPv6 tunnel list
169 static struct ip6_tnl __rcu **
170 ip6_tnl_bucket(struct ip6_tnl_net *ip6n, struct ip6_tnl_parm *p)
172 struct in6_addr *remote = &p->raddr;
173 struct in6_addr *local = &p->laddr;
177 if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
179 h = HASH(remote) ^ HASH(local);
181 return &ip6n->tnls[prio][h];
185 * ip6_tnl_link - add tunnel to hash table
186 * @t: tunnel to be added
190 ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
192 struct ip6_tnl __rcu **tp = ip6_tnl_bucket(ip6n, &t->parms);
194 rcu_assign_pointer(t->next , rtnl_dereference(*tp));
195 rcu_assign_pointer(*tp, t);
199 * ip6_tnl_unlink - remove tunnel from hash table
200 * @t: tunnel to be removed
204 ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
206 struct ip6_tnl __rcu **tp;
207 struct ip6_tnl *iter;
209 for (tp = ip6_tnl_bucket(ip6n, &t->parms);
210 (iter = rtnl_dereference(*tp)) != NULL;
213 rcu_assign_pointer(*tp, t->next);
220 * ip6_tnl_create() - create a new tunnel
221 * @p: tunnel parameters
222 * @pt: pointer to new tunnel
225 * Create tunnel matching given parameters.
228 * created tunnel or NULL
231 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct ip6_tnl_parm *p)
233 struct net_device *dev;
237 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
240 strlcpy(name, p->name, IFNAMSIZ);
242 sprintf(name, "ip6tnl%%d");
244 dev = alloc_netdev(sizeof (*t), name, ip6_tnl_dev_setup);
248 dev_net_set(dev, net);
250 if (strchr(name, '%')) {
251 if (dev_alloc_name(dev, name) < 0)
255 t = netdev_priv(dev);
257 ip6_tnl_dev_init(dev);
259 if ((err = register_netdevice(dev)) < 0)
263 ip6_tnl_link(ip6n, t);
273 * ip6_tnl_locate - find or create tunnel matching given parameters
274 * @p: tunnel parameters
275 * @create: != 0 if allowed to create new tunnel if no match found
278 * ip6_tnl_locate() first tries to locate an existing tunnel
279 * based on @parms. If this is unsuccessful, but @create is set a new
280 * tunnel device is created and registered for use.
283 * matching tunnel or NULL
286 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
287 struct ip6_tnl_parm *p, int create)
289 struct in6_addr *remote = &p->raddr;
290 struct in6_addr *local = &p->laddr;
291 struct ip6_tnl __rcu **tp;
293 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
295 for (tp = ip6_tnl_bucket(ip6n, p);
296 (t = rtnl_dereference(*tp)) != NULL;
298 if (ipv6_addr_equal(local, &t->parms.laddr) &&
299 ipv6_addr_equal(remote, &t->parms.raddr))
304 return ip6_tnl_create(net, p);
308 * ip6_tnl_dev_uninit - tunnel device uninitializer
309 * @dev: the device to be destroyed
312 * ip6_tnl_dev_uninit() removes tunnel from its list
316 ip6_tnl_dev_uninit(struct net_device *dev)
318 struct ip6_tnl *t = netdev_priv(dev);
319 struct net *net = dev_net(dev);
320 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
322 if (dev == ip6n->fb_tnl_dev)
323 rcu_assign_pointer(ip6n->tnls_wc[0], NULL);
325 ip6_tnl_unlink(ip6n, t);
326 ip6_tnl_dst_reset(t);
331 * parse_tvl_tnl_enc_lim - handle encapsulation limit option
332 * @skb: received socket buffer
335 * 0 if none was found,
336 * else index to encapsulation limit
340 parse_tlv_tnl_enc_lim(struct sk_buff *skb, __u8 * raw)
342 struct ipv6hdr *ipv6h = (struct ipv6hdr *) raw;
343 __u8 nexthdr = ipv6h->nexthdr;
344 __u16 off = sizeof (*ipv6h);
346 while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
348 struct ipv6_opt_hdr *hdr;
349 if (raw + off + sizeof (*hdr) > skb->data &&
350 !pskb_may_pull(skb, raw - skb->data + off + sizeof (*hdr)))
353 hdr = (struct ipv6_opt_hdr *) (raw + off);
354 if (nexthdr == NEXTHDR_FRAGMENT) {
355 struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
356 if (frag_hdr->frag_off)
359 } else if (nexthdr == NEXTHDR_AUTH) {
360 optlen = (hdr->hdrlen + 2) << 2;
362 optlen = ipv6_optlen(hdr);
364 if (nexthdr == NEXTHDR_DEST) {
367 struct ipv6_tlv_tnl_enc_lim *tel;
369 /* No more room for encapsulation limit */
370 if (i + sizeof (*tel) > off + optlen)
373 tel = (struct ipv6_tlv_tnl_enc_lim *) &raw[i];
374 /* return index of option if found and valid */
375 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
378 /* else jump to next option */
380 i += tel->length + 2;
385 nexthdr = hdr->nexthdr;
392 * ip6_tnl_err - tunnel error handler
395 * ip6_tnl_err() should handle errors in the tunnel according
396 * to the specifications in RFC 2473.
400 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
401 u8 *type, u8 *code, int *msg, __u32 *info, int offset)
403 struct ipv6hdr *ipv6h = (struct ipv6hdr *) skb->data;
406 u8 rel_type = ICMPV6_DEST_UNREACH;
407 u8 rel_code = ICMPV6_ADDR_UNREACH;
412 /* If the packet doesn't contain the original IPv6 header we are
413 in trouble since we might need the source address for further
414 processing of the error. */
417 if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr,
418 &ipv6h->saddr)) == NULL)
421 if (t->parms.proto != ipproto && t->parms.proto != 0)
428 struct ipv6_tlv_tnl_enc_lim *tel;
430 case ICMPV6_DEST_UNREACH:
433 "%s: Path to destination invalid "
434 "or inactive!\n", t->parms.name);
437 case ICMPV6_TIME_EXCEED:
438 if ((*code) == ICMPV6_EXC_HOPLIMIT) {
441 "%s: Too small hop limit or "
442 "routing loop in tunnel!\n",
447 case ICMPV6_PARAMPROB:
449 if ((*code) == ICMPV6_HDR_FIELD)
450 teli = parse_tlv_tnl_enc_lim(skb, skb->data);
452 if (teli && teli == *info - 2) {
453 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
454 if (tel->encap_limit == 0) {
457 "%s: Too small encapsulation "
458 "limit or routing loop in "
459 "tunnel!\n", t->parms.name);
462 } else if (net_ratelimit()) {
464 "%s: Recipient unable to parse tunneled "
465 "packet!\n ", t->parms.name);
468 case ICMPV6_PKT_TOOBIG:
469 mtu = *info - offset;
470 if (mtu < IPV6_MIN_MTU)
474 if ((len = sizeof (*ipv6h) + ntohs(ipv6h->payload_len)) > mtu) {
475 rel_type = ICMPV6_PKT_TOOBIG;
494 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
495 u8 type, u8 code, int offset, __be32 info)
500 __u32 rel_info = ntohl(info);
502 struct sk_buff *skb2;
507 err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
508 &rel_msg, &rel_info, offset);
516 case ICMPV6_DEST_UNREACH:
517 if (rel_code != ICMPV6_ADDR_UNREACH)
519 rel_type = ICMP_DEST_UNREACH;
520 rel_code = ICMP_HOST_UNREACH;
522 case ICMPV6_PKT_TOOBIG:
525 rel_type = ICMP_DEST_UNREACH;
526 rel_code = ICMP_FRAG_NEEDED;
532 if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
535 skb2 = skb_clone(skb, GFP_ATOMIC);
541 skb_pull(skb2, offset);
542 skb_reset_network_header(skb2);
545 /* Try to guess incoming interface */
546 memset(&fl, 0, sizeof(fl));
547 fl.fl4_dst = eiph->saddr;
548 fl.fl4_tos = RT_TOS(eiph->tos);
549 fl.proto = IPPROTO_IPIP;
550 if (ip_route_output_key(dev_net(skb->dev), &rt, &fl))
553 skb2->dev = rt->dst.dev;
555 /* route "incoming" packet */
556 if (rt->rt_flags & RTCF_LOCAL) {
559 fl.fl4_dst = eiph->daddr;
560 fl.fl4_src = eiph->saddr;
561 fl.fl4_tos = eiph->tos;
562 if (ip_route_output_key(dev_net(skb->dev), &rt, &fl) ||
563 rt->dst.dev->type != ARPHRD_TUNNEL) {
567 skb_dst_set(skb2, (struct dst_entry *)rt);
570 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
572 skb_dst(skb2)->dev->type != ARPHRD_TUNNEL)
576 /* change mtu on this route */
577 if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
578 if (rel_info > dst_mtu(skb_dst(skb2)))
581 skb_dst(skb2)->ops->update_pmtu(skb_dst(skb2), rel_info);
584 icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
592 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
593 u8 type, u8 code, int offset, __be32 info)
598 __u32 rel_info = ntohl(info);
601 err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
602 &rel_msg, &rel_info, offset);
606 if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
608 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
614 skb_pull(skb2, offset);
615 skb_reset_network_header(skb2);
617 /* Try to guess incoming interface */
618 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
621 if (rt && rt->rt6i_dev)
622 skb2->dev = rt->rt6i_dev;
624 icmpv6_send(skb2, rel_type, rel_code, rel_info);
627 dst_release(&rt->dst);
635 static void ip4ip6_dscp_ecn_decapsulate(struct ip6_tnl *t,
636 struct ipv6hdr *ipv6h,
639 __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
641 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
642 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
644 if (INET_ECN_is_ce(dsfield))
645 IP_ECN_set_ce(ip_hdr(skb));
648 static void ip6ip6_dscp_ecn_decapsulate(struct ip6_tnl *t,
649 struct ipv6hdr *ipv6h,
652 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
653 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
655 if (INET_ECN_is_ce(ipv6_get_dsfield(ipv6h)))
656 IP6_ECN_set_ce(ipv6_hdr(skb));
659 /* called with rcu_read_lock() */
660 static inline int ip6_tnl_rcv_ctl(struct ip6_tnl *t)
662 struct ip6_tnl_parm *p = &t->parms;
664 struct net *net = dev_net(t->dev);
666 if (p->flags & IP6_TNL_F_CAP_RCV) {
667 struct net_device *ldev = NULL;
670 ldev = dev_get_by_index_rcu(net, p->link);
672 if ((ipv6_addr_is_multicast(&p->laddr) ||
673 likely(ipv6_chk_addr(net, &p->laddr, ldev, 0))) &&
674 likely(!ipv6_chk_addr(net, &p->raddr, NULL, 0)))
682 * ip6_tnl_rcv - decapsulate IPv6 packet and retransmit it locally
683 * @skb: received socket buffer
684 * @protocol: ethernet protocol ID
685 * @dscp_ecn_decapsulate: the function to decapsulate DSCP code and ECN
690 static int ip6_tnl_rcv(struct sk_buff *skb, __u16 protocol,
692 void (*dscp_ecn_decapsulate)(struct ip6_tnl *t,
693 struct ipv6hdr *ipv6h,
694 struct sk_buff *skb))
697 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
701 if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr,
702 &ipv6h->daddr)) != NULL) {
703 if (t->parms.proto != ipproto && t->parms.proto != 0) {
708 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
713 if (!ip6_tnl_rcv_ctl(t)) {
714 t->dev->stats.rx_dropped++;
719 skb->mac_header = skb->network_header;
720 skb_reset_network_header(skb);
721 skb->protocol = htons(protocol);
722 skb->pkt_type = PACKET_HOST;
723 memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
725 skb_tunnel_rx(skb, t->dev);
727 dscp_ecn_decapsulate(t, ipv6h, skb);
729 if (netif_rx(skb) == NET_RX_DROP)
730 t->dev->stats.rx_dropped++;
743 static int ip4ip6_rcv(struct sk_buff *skb)
745 return ip6_tnl_rcv(skb, ETH_P_IP, IPPROTO_IPIP,
746 ip4ip6_dscp_ecn_decapsulate);
749 static int ip6ip6_rcv(struct sk_buff *skb)
751 return ip6_tnl_rcv(skb, ETH_P_IPV6, IPPROTO_IPV6,
752 ip6ip6_dscp_ecn_decapsulate);
755 struct ipv6_tel_txoption {
756 struct ipv6_txoptions ops;
760 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
762 memset(opt, 0, sizeof(struct ipv6_tel_txoption));
764 opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
766 opt->dst_opt[4] = encap_limit;
767 opt->dst_opt[5] = IPV6_TLV_PADN;
770 opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
771 opt->ops.opt_nflen = 8;
775 * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
776 * @t: the outgoing tunnel device
777 * @hdr: IPv6 header from the incoming packet
780 * Avoid trivial tunneling loop by checking that tunnel exit-point
781 * doesn't match source of incoming packet.
789 ip6_tnl_addr_conflict(struct ip6_tnl *t, struct ipv6hdr *hdr)
791 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
794 static inline int ip6_tnl_xmit_ctl(struct ip6_tnl *t)
796 struct ip6_tnl_parm *p = &t->parms;
798 struct net *net = dev_net(t->dev);
800 if (p->flags & IP6_TNL_F_CAP_XMIT) {
801 struct net_device *ldev = NULL;
805 ldev = dev_get_by_index_rcu(net, p->link);
807 if (unlikely(!ipv6_chk_addr(net, &p->laddr, ldev, 0)))
809 "%s xmit: Local address not yet configured!\n",
811 else if (!ipv6_addr_is_multicast(&p->raddr) &&
812 unlikely(ipv6_chk_addr(net, &p->raddr, NULL, 0)))
814 "%s xmit: Routing loop! "
815 "Remote address found on this node!\n",
824 * ip6_tnl_xmit2 - encapsulate packet and send
825 * @skb: the outgoing socket buffer
826 * @dev: the outgoing tunnel device
827 * @dsfield: dscp code for outer header
828 * @fl: flow of tunneled packet
829 * @encap_limit: encapsulation limit
830 * @pmtu: Path MTU is stored if packet is too big
833 * Build new header and do some sanity checks on the packet before sending
839 * %-EMSGSIZE message too big. return mtu in this case.
842 static int ip6_tnl_xmit2(struct sk_buff *skb,
843 struct net_device *dev,
849 struct net *net = dev_net(dev);
850 struct ip6_tnl *t = netdev_priv(dev);
851 struct net_device_stats *stats = &t->dev->stats;
852 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
853 struct ipv6_tel_txoption opt;
854 struct dst_entry *dst;
855 struct net_device *tdev;
857 unsigned int max_headroom = sizeof(struct ipv6hdr);
862 if ((dst = ip6_tnl_dst_check(t)) != NULL)
865 dst = ip6_route_output(net, NULL, fl);
867 if (dst->error || xfrm_lookup(net, &dst, fl, NULL, 0) < 0)
868 goto tx_err_link_failure;
877 "%s: Local routing loop detected!\n",
879 goto tx_err_dst_release;
881 mtu = dst_mtu(dst) - sizeof (*ipv6h);
882 if (encap_limit >= 0) {
886 if (mtu < IPV6_MIN_MTU)
889 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), mtu);
890 if (skb->len > mtu) {
893 goto tx_err_dst_release;
897 * Okay, now see if we can stuff it in the buffer as-is.
899 max_headroom += LL_RESERVED_SPACE(tdev);
901 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
902 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
903 struct sk_buff *new_skb;
905 if (!(new_skb = skb_realloc_headroom(skb, max_headroom)))
906 goto tx_err_dst_release;
909 skb_set_owner_w(new_skb, skb->sk);
914 skb_dst_set(skb, dst_clone(dst));
916 skb->transport_header = skb->network_header;
919 if (encap_limit >= 0) {
920 init_tel_txopt(&opt, encap_limit);
921 ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
923 skb_push(skb, sizeof(struct ipv6hdr));
924 skb_reset_network_header(skb);
925 ipv6h = ipv6_hdr(skb);
926 *(__be32*)ipv6h = fl->fl6_flowlabel | htonl(0x60000000);
927 dsfield = INET_ECN_encapsulate(0, dsfield);
928 ipv6_change_dsfield(ipv6h, ~INET_ECN_MASK, dsfield);
929 ipv6h->hop_limit = t->parms.hop_limit;
930 ipv6h->nexthdr = proto;
931 ipv6_addr_copy(&ipv6h->saddr, &fl->fl6_src);
932 ipv6_addr_copy(&ipv6h->daddr, &fl->fl6_dst);
935 err = ip6_local_out(skb);
937 if (net_xmit_eval(err) == 0) {
938 stats->tx_bytes += pkt_len;
942 stats->tx_aborted_errors++;
944 ip6_tnl_dst_store(t, dst);
947 stats->tx_carrier_errors++;
948 dst_link_failure(skb);
955 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
957 struct ip6_tnl *t = netdev_priv(dev);
958 struct iphdr *iph = ip_hdr(skb);
959 int encap_limit = -1;
965 if ((t->parms.proto != IPPROTO_IPIP && t->parms.proto != 0) ||
966 !ip6_tnl_xmit_ctl(t))
969 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
970 encap_limit = t->parms.encap_limit;
972 memcpy(&fl, &t->fl, sizeof (fl));
973 fl.proto = IPPROTO_IPIP;
975 dsfield = ipv4_get_dsfield(iph);
977 if ((t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS))
978 fl.fl6_flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
981 err = ip6_tnl_xmit2(skb, dev, dsfield, &fl, encap_limit, &mtu);
983 /* XXX: send ICMP error even if DF is not set. */
984 if (err == -EMSGSIZE)
985 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
994 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
996 struct ip6_tnl *t = netdev_priv(dev);
997 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
998 int encap_limit = -1;
1005 if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) ||
1006 !ip6_tnl_xmit_ctl(t) || ip6_tnl_addr_conflict(t, ipv6h))
1009 offset = parse_tlv_tnl_enc_lim(skb, skb_network_header(skb));
1011 struct ipv6_tlv_tnl_enc_lim *tel;
1012 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
1013 if (tel->encap_limit == 0) {
1014 icmpv6_send(skb, ICMPV6_PARAMPROB,
1015 ICMPV6_HDR_FIELD, offset + 2);
1018 encap_limit = tel->encap_limit - 1;
1019 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1020 encap_limit = t->parms.encap_limit;
1022 memcpy(&fl, &t->fl, sizeof (fl));
1023 fl.proto = IPPROTO_IPV6;
1025 dsfield = ipv6_get_dsfield(ipv6h);
1026 if ((t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS))
1027 fl.fl6_flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
1028 if ((t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL))
1029 fl.fl6_flowlabel |= (*(__be32 *) ipv6h & IPV6_FLOWLABEL_MASK);
1031 err = ip6_tnl_xmit2(skb, dev, dsfield, &fl, encap_limit, &mtu);
1033 if (err == -EMSGSIZE)
1034 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1042 ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1044 struct ip6_tnl *t = netdev_priv(dev);
1045 struct net_device_stats *stats = &t->dev->stats;
1048 switch (skb->protocol) {
1049 case htons(ETH_P_IP):
1050 ret = ip4ip6_tnl_xmit(skb, dev);
1052 case htons(ETH_P_IPV6):
1053 ret = ip6ip6_tnl_xmit(skb, dev);
1062 return NETDEV_TX_OK;
1066 stats->tx_dropped++;
1068 return NETDEV_TX_OK;
1071 static void ip6_tnl_set_cap(struct ip6_tnl *t)
1073 struct ip6_tnl_parm *p = &t->parms;
1074 int ltype = ipv6_addr_type(&p->laddr);
1075 int rtype = ipv6_addr_type(&p->raddr);
1077 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV);
1079 if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
1080 rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
1081 !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
1082 (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
1083 if (ltype&IPV6_ADDR_UNICAST)
1084 p->flags |= IP6_TNL_F_CAP_XMIT;
1085 if (rtype&IPV6_ADDR_UNICAST)
1086 p->flags |= IP6_TNL_F_CAP_RCV;
1090 static void ip6_tnl_link_config(struct ip6_tnl *t)
1092 struct net_device *dev = t->dev;
1093 struct ip6_tnl_parm *p = &t->parms;
1094 struct flowi *fl = &t->fl;
1096 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1097 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1099 /* Set up flowi template */
1100 ipv6_addr_copy(&fl->fl6_src, &p->laddr);
1101 ipv6_addr_copy(&fl->fl6_dst, &p->raddr);
1103 fl->fl6_flowlabel = 0;
1105 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1106 fl->fl6_flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1107 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1108 fl->fl6_flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1112 if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1113 dev->flags |= IFF_POINTOPOINT;
1115 dev->flags &= ~IFF_POINTOPOINT;
1117 dev->iflink = p->link;
1119 if (p->flags & IP6_TNL_F_CAP_XMIT) {
1120 int strict = (ipv6_addr_type(&p->raddr) &
1121 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1123 struct rt6_info *rt = rt6_lookup(dev_net(dev),
1124 &p->raddr, &p->laddr,
1131 dev->hard_header_len = rt->rt6i_dev->hard_header_len +
1132 sizeof (struct ipv6hdr);
1134 dev->mtu = rt->rt6i_dev->mtu - sizeof (struct ipv6hdr);
1136 if (dev->mtu < IPV6_MIN_MTU)
1137 dev->mtu = IPV6_MIN_MTU;
1139 dst_release(&rt->dst);
1144 * ip6_tnl_change - update the tunnel parameters
1145 * @t: tunnel to be changed
1146 * @p: tunnel configuration parameters
1149 * ip6_tnl_change() updates the tunnel parameters
1153 ip6_tnl_change(struct ip6_tnl *t, struct ip6_tnl_parm *p)
1155 ipv6_addr_copy(&t->parms.laddr, &p->laddr);
1156 ipv6_addr_copy(&t->parms.raddr, &p->raddr);
1157 t->parms.flags = p->flags;
1158 t->parms.hop_limit = p->hop_limit;
1159 t->parms.encap_limit = p->encap_limit;
1160 t->parms.flowinfo = p->flowinfo;
1161 t->parms.link = p->link;
1162 t->parms.proto = p->proto;
1163 ip6_tnl_dst_reset(t);
1164 ip6_tnl_link_config(t);
1169 * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1170 * @dev: virtual device associated with tunnel
1171 * @ifr: parameters passed from userspace
1172 * @cmd: command to be performed
1175 * ip6_tnl_ioctl() is used for managing IPv6 tunnels
1178 * The possible commands are the following:
1179 * %SIOCGETTUNNEL: get tunnel parameters for device
1180 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1181 * %SIOCCHGTUNNEL: change tunnel parameters to those given
1182 * %SIOCDELTUNNEL: delete tunnel
1184 * The fallback device "ip6tnl0", created during module
1185 * initialization, can be used for creating other tunnel devices.
1189 * %-EFAULT if unable to copy data to or from userspace,
1190 * %-EPERM if current process hasn't %CAP_NET_ADMIN set
1191 * %-EINVAL if passed tunnel parameters are invalid,
1192 * %-EEXIST if changing a tunnel's parameters would cause a conflict
1193 * %-ENODEV if attempting to change or delete a nonexisting device
1197 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1200 struct ip6_tnl_parm p;
1201 struct ip6_tnl *t = NULL;
1202 struct net *net = dev_net(dev);
1203 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1207 if (dev == ip6n->fb_tnl_dev) {
1208 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p))) {
1212 t = ip6_tnl_locate(net, &p, 0);
1215 t = netdev_priv(dev);
1216 memcpy(&p, &t->parms, sizeof (p));
1217 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof (p))) {
1224 if (!capable(CAP_NET_ADMIN))
1227 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
1230 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1233 t = ip6_tnl_locate(net, &p, cmd == SIOCADDTUNNEL);
1234 if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) {
1236 if (t->dev != dev) {
1241 t = netdev_priv(dev);
1243 ip6_tnl_unlink(ip6n, t);
1244 err = ip6_tnl_change(t, &p);
1245 ip6_tnl_link(ip6n, t);
1246 netdev_state_change(dev);
1250 if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof (p)))
1254 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1258 if (!capable(CAP_NET_ADMIN))
1261 if (dev == ip6n->fb_tnl_dev) {
1263 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
1266 if ((t = ip6_tnl_locate(net, &p, 0)) == NULL)
1269 if (t->dev == ip6n->fb_tnl_dev)
1274 unregister_netdevice(dev);
1283 * ip6_tnl_change_mtu - change mtu manually for tunnel device
1284 * @dev: virtual device associated with tunnel
1285 * @new_mtu: the new mtu
1289 * %-EINVAL if mtu too small
1293 ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1295 if (new_mtu < IPV6_MIN_MTU) {
1303 static const struct net_device_ops ip6_tnl_netdev_ops = {
1304 .ndo_uninit = ip6_tnl_dev_uninit,
1305 .ndo_start_xmit = ip6_tnl_xmit,
1306 .ndo_do_ioctl = ip6_tnl_ioctl,
1307 .ndo_change_mtu = ip6_tnl_change_mtu,
1311 * ip6_tnl_dev_setup - setup virtual tunnel device
1312 * @dev: virtual device associated with tunnel
1315 * Initialize function pointers and device parameters
1318 static void ip6_tnl_dev_setup(struct net_device *dev)
1320 dev->netdev_ops = &ip6_tnl_netdev_ops;
1321 dev->destructor = free_netdev;
1323 dev->type = ARPHRD_TUNNEL6;
1324 dev->hard_header_len = LL_MAX_HEADER + sizeof (struct ipv6hdr);
1325 dev->mtu = ETH_DATA_LEN - sizeof (struct ipv6hdr);
1326 dev->flags |= IFF_NOARP;
1327 dev->addr_len = sizeof(struct in6_addr);
1328 dev->features |= NETIF_F_NETNS_LOCAL;
1333 * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1334 * @dev: virtual device associated with tunnel
1338 ip6_tnl_dev_init_gen(struct net_device *dev)
1340 struct ip6_tnl *t = netdev_priv(dev);
1342 strcpy(t->parms.name, dev->name);
1346 * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1347 * @dev: virtual device associated with tunnel
1350 static void ip6_tnl_dev_init(struct net_device *dev)
1352 struct ip6_tnl *t = netdev_priv(dev);
1353 ip6_tnl_dev_init_gen(dev);
1354 ip6_tnl_link_config(t);
1358 * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1359 * @dev: fallback device
1364 static void __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1366 struct ip6_tnl *t = netdev_priv(dev);
1367 struct net *net = dev_net(dev);
1368 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1370 ip6_tnl_dev_init_gen(dev);
1371 t->parms.proto = IPPROTO_IPV6;
1373 rcu_assign_pointer(ip6n->tnls_wc[0], t);
1376 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
1377 .handler = ip4ip6_rcv,
1378 .err_handler = ip4ip6_err,
1382 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
1383 .handler = ip6ip6_rcv,
1384 .err_handler = ip6ip6_err,
1388 static void __net_exit ip6_tnl_destroy_tunnels(struct ip6_tnl_net *ip6n)
1394 for (h = 0; h < HASH_SIZE; h++) {
1395 t = rtnl_dereference(ip6n->tnls_r_l[h]);
1397 unregister_netdevice_queue(t->dev, &list);
1398 t = rtnl_dereference(t->next);
1402 t = rtnl_dereference(ip6n->tnls_wc[0]);
1403 unregister_netdevice_queue(t->dev, &list);
1404 unregister_netdevice_many(&list);
1407 static int __net_init ip6_tnl_init_net(struct net *net)
1409 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1412 ip6n->tnls[0] = ip6n->tnls_wc;
1413 ip6n->tnls[1] = ip6n->tnls_r_l;
1416 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
1419 if (!ip6n->fb_tnl_dev)
1421 dev_net_set(ip6n->fb_tnl_dev, net);
1423 ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
1425 err = register_netdev(ip6n->fb_tnl_dev);
1431 free_netdev(ip6n->fb_tnl_dev);
1436 static void __net_exit ip6_tnl_exit_net(struct net *net)
1438 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1441 ip6_tnl_destroy_tunnels(ip6n);
1445 static struct pernet_operations ip6_tnl_net_ops = {
1446 .init = ip6_tnl_init_net,
1447 .exit = ip6_tnl_exit_net,
1448 .id = &ip6_tnl_net_id,
1449 .size = sizeof(struct ip6_tnl_net),
1453 * ip6_tunnel_init - register protocol and reserve needed resources
1455 * Return: 0 on success
1458 static int __init ip6_tunnel_init(void)
1462 err = register_pernet_device(&ip6_tnl_net_ops);
1466 err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
1468 printk(KERN_ERR "ip6_tunnel init: can't register ip4ip6\n");
1472 err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
1474 printk(KERN_ERR "ip6_tunnel init: can't register ip6ip6\n");
1481 xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
1483 unregister_pernet_device(&ip6_tnl_net_ops);
1489 * ip6_tunnel_cleanup - free resources and unregister protocol
1492 static void __exit ip6_tunnel_cleanup(void)
1494 if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
1495 printk(KERN_INFO "ip6_tunnel close: can't deregister ip4ip6\n");
1497 if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
1498 printk(KERN_INFO "ip6_tunnel close: can't deregister ip6ip6\n");
1500 unregister_pernet_device(&ip6_tnl_net_ops);
1503 module_init(ip6_tunnel_init);
1504 module_exit(ip6_tunnel_cleanup);