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>
41 #include <asm/uaccess.h>
42 #include <asm/atomic.h>
47 #include <net/ip6_route.h>
48 #include <net/addrconf.h>
49 #include <net/ip6_tunnel.h>
51 #include <net/dsfield.h>
52 #include <net/inet_ecn.h>
53 #include <net/net_namespace.h>
54 #include <net/netns/generic.h>
56 MODULE_AUTHOR("Ville Nuorvala");
57 MODULE_DESCRIPTION("IPv6 tunneling device");
58 MODULE_LICENSE("GPL");
60 #define IPV6_TLV_TEL_DST_SIZE 8
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 void ip6_fb_tnl_dev_init(struct net_device *dev);
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 *tnls_r_l[HASH_SIZE];
87 struct ip6_tnl *tnls_wc[1];
88 struct ip6_tnl **tnls[2];
92 * Locking : hash tables are protected by RCU and a spinlock
94 static DEFINE_SPINLOCK(ip6_tnl_lock);
96 static inline struct dst_entry *ip6_tnl_dst_check(struct ip6_tnl *t)
98 struct dst_entry *dst = t->dst_cache;
100 if (dst && dst->obsolete &&
101 dst->ops->check(dst, t->dst_cookie) == NULL) {
110 static inline void ip6_tnl_dst_reset(struct ip6_tnl *t)
112 dst_release(t->dst_cache);
116 static inline void ip6_tnl_dst_store(struct ip6_tnl *t, struct dst_entry *dst)
118 struct rt6_info *rt = (struct rt6_info *) dst;
119 t->dst_cookie = rt->rt6i_node ? rt->rt6i_node->fn_sernum : 0;
120 dst_release(t->dst_cache);
125 * ip6_tnl_lookup - fetch tunnel matching the end-point addresses
126 * @remote: the address of the tunnel exit-point
127 * @local: the address of the tunnel entry-point
130 * tunnel matching given end-points if found,
131 * else fallback tunnel if its device is up,
135 #define for_each_ip6_tunnel_rcu(start) \
136 for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
138 static struct ip6_tnl *
139 ip6_tnl_lookup(struct net *net, struct in6_addr *remote, struct in6_addr *local)
141 unsigned h0 = HASH(remote);
142 unsigned h1 = HASH(local);
144 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
146 for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[h0 ^ h1]) {
147 if (ipv6_addr_equal(local, &t->parms.laddr) &&
148 ipv6_addr_equal(remote, &t->parms.raddr) &&
149 (t->dev->flags & IFF_UP))
152 t = rcu_dereference(ip6n->tnls_wc[0]);
153 if (t && (t->dev->flags & IFF_UP))
160 * ip6_tnl_bucket - get head of list matching given tunnel parameters
161 * @p: parameters containing tunnel end-points
164 * ip6_tnl_bucket() returns the head of the list matching the
165 * &struct in6_addr entries laddr and raddr in @p.
167 * Return: head of IPv6 tunnel list
170 static struct ip6_tnl **
171 ip6_tnl_bucket(struct ip6_tnl_net *ip6n, struct ip6_tnl_parm *p)
173 struct in6_addr *remote = &p->raddr;
174 struct in6_addr *local = &p->laddr;
178 if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
180 h = HASH(remote) ^ HASH(local);
182 return &ip6n->tnls[prio][h];
186 * ip6_tnl_link - add tunnel to hash table
187 * @t: tunnel to be added
191 ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
193 struct ip6_tnl **tp = ip6_tnl_bucket(ip6n, &t->parms);
195 spin_lock_bh(&ip6_tnl_lock);
197 rcu_assign_pointer(*tp, t);
198 spin_unlock_bh(&ip6_tnl_lock);
202 * ip6_tnl_unlink - remove tunnel from hash table
203 * @t: tunnel to be removed
207 ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
211 for (tp = ip6_tnl_bucket(ip6n, &t->parms); *tp; tp = &(*tp)->next) {
213 spin_lock_bh(&ip6_tnl_lock);
215 spin_unlock_bh(&ip6_tnl_lock);
222 * ip6_tnl_create() - create a new tunnel
223 * @p: tunnel parameters
224 * @pt: pointer to new tunnel
227 * Create tunnel matching given parameters.
230 * created tunnel or NULL
233 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct ip6_tnl_parm *p)
235 struct net_device *dev;
239 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
242 strlcpy(name, p->name, IFNAMSIZ);
244 sprintf(name, "ip6tnl%%d");
246 dev = alloc_netdev(sizeof (*t), name, ip6_tnl_dev_setup);
250 dev_net_set(dev, net);
252 if (strchr(name, '%')) {
253 if (dev_alloc_name(dev, name) < 0)
257 t = netdev_priv(dev);
259 ip6_tnl_dev_init(dev);
261 if ((err = register_netdevice(dev)) < 0)
265 ip6_tnl_link(ip6n, t);
275 * ip6_tnl_locate - find or create tunnel matching given parameters
276 * @p: tunnel parameters
277 * @create: != 0 if allowed to create new tunnel if no match found
280 * ip6_tnl_locate() first tries to locate an existing tunnel
281 * based on @parms. If this is unsuccessful, but @create is set a new
282 * tunnel device is created and registered for use.
285 * matching tunnel or NULL
288 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
289 struct ip6_tnl_parm *p, int create)
291 struct in6_addr *remote = &p->raddr;
292 struct in6_addr *local = &p->laddr;
294 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
296 for (t = *ip6_tnl_bucket(ip6n, p); t; t = t->next) {
297 if (ipv6_addr_equal(local, &t->parms.laddr) &&
298 ipv6_addr_equal(remote, &t->parms.raddr))
303 return ip6_tnl_create(net, p);
307 * ip6_tnl_dev_uninit - tunnel device uninitializer
308 * @dev: the device to be destroyed
311 * ip6_tnl_dev_uninit() removes tunnel from its list
315 ip6_tnl_dev_uninit(struct net_device *dev)
317 struct ip6_tnl *t = netdev_priv(dev);
318 struct net *net = dev_net(dev);
319 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
321 if (dev == ip6n->fb_tnl_dev) {
322 spin_lock_bh(&ip6_tnl_lock);
323 ip6n->tnls_wc[0] = NULL;
324 spin_unlock_bh(&ip6_tnl_lock);
326 ip6_tnl_unlink(ip6n, t);
328 ip6_tnl_dst_reset(t);
333 * parse_tvl_tnl_enc_lim - handle encapsulation limit option
334 * @skb: received socket buffer
337 * 0 if none was found,
338 * else index to encapsulation limit
342 parse_tlv_tnl_enc_lim(struct sk_buff *skb, __u8 * raw)
344 struct ipv6hdr *ipv6h = (struct ipv6hdr *) raw;
345 __u8 nexthdr = ipv6h->nexthdr;
346 __u16 off = sizeof (*ipv6h);
348 while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
350 struct ipv6_opt_hdr *hdr;
351 if (raw + off + sizeof (*hdr) > skb->data &&
352 !pskb_may_pull(skb, raw - skb->data + off + sizeof (*hdr)))
355 hdr = (struct ipv6_opt_hdr *) (raw + off);
356 if (nexthdr == NEXTHDR_FRAGMENT) {
357 struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
358 if (frag_hdr->frag_off)
361 } else if (nexthdr == NEXTHDR_AUTH) {
362 optlen = (hdr->hdrlen + 2) << 2;
364 optlen = ipv6_optlen(hdr);
366 if (nexthdr == NEXTHDR_DEST) {
369 struct ipv6_tlv_tnl_enc_lim *tel;
371 /* No more room for encapsulation limit */
372 if (i + sizeof (*tel) > off + optlen)
375 tel = (struct ipv6_tlv_tnl_enc_lim *) &raw[i];
376 /* return index of option if found and valid */
377 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
380 /* else jump to next option */
382 i += tel->length + 2;
387 nexthdr = hdr->nexthdr;
394 * ip6_tnl_err - tunnel error handler
397 * ip6_tnl_err() should handle errors in the tunnel according
398 * to the specifications in RFC 2473.
402 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
403 u8 *type, u8 *code, int *msg, __u32 *info, int offset)
405 struct ipv6hdr *ipv6h = (struct ipv6hdr *) skb->data;
408 u8 rel_type = ICMPV6_DEST_UNREACH;
409 u8 rel_code = ICMPV6_ADDR_UNREACH;
414 /* If the packet doesn't contain the original IPv6 header we are
415 in trouble since we might need the source address for further
416 processing of the error. */
419 if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr,
420 &ipv6h->saddr)) == NULL)
423 if (t->parms.proto != ipproto && t->parms.proto != 0)
430 struct ipv6_tlv_tnl_enc_lim *tel;
432 case ICMPV6_DEST_UNREACH:
435 "%s: Path to destination invalid "
436 "or inactive!\n", t->parms.name);
439 case ICMPV6_TIME_EXCEED:
440 if ((*code) == ICMPV6_EXC_HOPLIMIT) {
443 "%s: Too small hop limit or "
444 "routing loop in tunnel!\n",
449 case ICMPV6_PARAMPROB:
451 if ((*code) == ICMPV6_HDR_FIELD)
452 teli = parse_tlv_tnl_enc_lim(skb, skb->data);
454 if (teli && teli == *info - 2) {
455 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
456 if (tel->encap_limit == 0) {
459 "%s: Too small encapsulation "
460 "limit or routing loop in "
461 "tunnel!\n", t->parms.name);
464 } else if (net_ratelimit()) {
466 "%s: Recipient unable to parse tunneled "
467 "packet!\n ", t->parms.name);
470 case ICMPV6_PKT_TOOBIG:
471 mtu = *info - offset;
472 if (mtu < IPV6_MIN_MTU)
476 if ((len = sizeof (*ipv6h) + ntohs(ipv6h->payload_len)) > mtu) {
477 rel_type = ICMPV6_PKT_TOOBIG;
496 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
497 u8 type, u8 code, int offset, __be32 info)
502 __u32 rel_info = ntohl(info);
504 struct sk_buff *skb2;
509 err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
510 &rel_msg, &rel_info, offset);
518 case ICMPV6_DEST_UNREACH:
519 if (rel_code != ICMPV6_ADDR_UNREACH)
521 rel_type = ICMP_DEST_UNREACH;
522 rel_code = ICMP_HOST_UNREACH;
524 case ICMPV6_PKT_TOOBIG:
527 rel_type = ICMP_DEST_UNREACH;
528 rel_code = ICMP_FRAG_NEEDED;
534 if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
537 skb2 = skb_clone(skb, GFP_ATOMIC);
543 skb_pull(skb2, offset);
544 skb_reset_network_header(skb2);
547 /* Try to guess incoming interface */
548 memset(&fl, 0, sizeof(fl));
549 fl.fl4_dst = eiph->saddr;
550 fl.fl4_tos = RT_TOS(eiph->tos);
551 fl.proto = IPPROTO_IPIP;
552 if (ip_route_output_key(dev_net(skb->dev), &rt, &fl))
555 skb2->dev = rt->u.dst.dev;
557 /* route "incoming" packet */
558 if (rt->rt_flags & RTCF_LOCAL) {
561 fl.fl4_dst = eiph->daddr;
562 fl.fl4_src = eiph->saddr;
563 fl.fl4_tos = eiph->tos;
564 if (ip_route_output_key(dev_net(skb->dev), &rt, &fl) ||
565 rt->u.dst.dev->type != ARPHRD_TUNNEL) {
569 skb_dst_set(skb2, (struct dst_entry *)rt);
572 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
574 skb_dst(skb2)->dev->type != ARPHRD_TUNNEL)
578 /* change mtu on this route */
579 if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
580 if (rel_info > dst_mtu(skb_dst(skb2)))
583 skb_dst(skb2)->ops->update_pmtu(skb_dst(skb2), rel_info);
586 icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
594 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
595 u8 type, u8 code, int offset, __be32 info)
600 __u32 rel_info = ntohl(info);
603 err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
604 &rel_msg, &rel_info, offset);
608 if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
610 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
616 skb_pull(skb2, offset);
617 skb_reset_network_header(skb2);
619 /* Try to guess incoming interface */
620 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
623 if (rt && rt->rt6i_dev)
624 skb2->dev = rt->rt6i_dev;
626 icmpv6_send(skb2, rel_type, rel_code, rel_info, skb2->dev);
629 dst_release(&rt->u.dst);
637 static void ip4ip6_dscp_ecn_decapsulate(struct ip6_tnl *t,
638 struct ipv6hdr *ipv6h,
641 __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
643 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
644 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
646 if (INET_ECN_is_ce(dsfield))
647 IP_ECN_set_ce(ip_hdr(skb));
650 static void ip6ip6_dscp_ecn_decapsulate(struct ip6_tnl *t,
651 struct ipv6hdr *ipv6h,
654 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
655 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
657 if (INET_ECN_is_ce(ipv6_get_dsfield(ipv6h)))
658 IP6_ECN_set_ce(ipv6_hdr(skb));
661 /* called with rcu_read_lock() */
662 static inline int ip6_tnl_rcv_ctl(struct ip6_tnl *t)
664 struct ip6_tnl_parm *p = &t->parms;
666 struct net *net = dev_net(t->dev);
668 if (p->flags & IP6_TNL_F_CAP_RCV) {
669 struct net_device *ldev = NULL;
672 ldev = dev_get_by_index_rcu(net, p->link);
674 if ((ipv6_addr_is_multicast(&p->laddr) ||
675 likely(ipv6_chk_addr(net, &p->laddr, ldev, 0))) &&
676 likely(!ipv6_chk_addr(net, &p->raddr, NULL, 0)))
684 * ip6_tnl_rcv - decapsulate IPv6 packet and retransmit it locally
685 * @skb: received socket buffer
686 * @protocol: ethernet protocol ID
687 * @dscp_ecn_decapsulate: the function to decapsulate DSCP code and ECN
692 static int ip6_tnl_rcv(struct sk_buff *skb, __u16 protocol,
694 void (*dscp_ecn_decapsulate)(struct ip6_tnl *t,
695 struct ipv6hdr *ipv6h,
696 struct sk_buff *skb))
699 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
703 if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr,
704 &ipv6h->daddr)) != NULL) {
705 if (t->parms.proto != ipproto && t->parms.proto != 0) {
710 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
715 if (!ip6_tnl_rcv_ctl(t)) {
716 t->dev->stats.rx_dropped++;
721 skb->mac_header = skb->network_header;
722 skb_reset_network_header(skb);
723 skb->protocol = htons(protocol);
724 skb->pkt_type = PACKET_HOST;
725 memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
730 dscp_ecn_decapsulate(t, ipv6h, skb);
732 t->dev->stats.rx_packets++;
733 t->dev->stats.rx_bytes += skb->len;
746 static int ip4ip6_rcv(struct sk_buff *skb)
748 return ip6_tnl_rcv(skb, ETH_P_IP, IPPROTO_IPIP,
749 ip4ip6_dscp_ecn_decapsulate);
752 static int ip6ip6_rcv(struct sk_buff *skb)
754 return ip6_tnl_rcv(skb, ETH_P_IPV6, IPPROTO_IPV6,
755 ip6ip6_dscp_ecn_decapsulate);
758 struct ipv6_tel_txoption {
759 struct ipv6_txoptions ops;
763 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
765 memset(opt, 0, sizeof(struct ipv6_tel_txoption));
767 opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
769 opt->dst_opt[4] = encap_limit;
770 opt->dst_opt[5] = IPV6_TLV_PADN;
773 opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
774 opt->ops.opt_nflen = 8;
778 * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
779 * @t: the outgoing tunnel device
780 * @hdr: IPv6 header from the incoming packet
783 * Avoid trivial tunneling loop by checking that tunnel exit-point
784 * doesn't match source of incoming packet.
792 ip6_tnl_addr_conflict(struct ip6_tnl *t, struct ipv6hdr *hdr)
794 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
797 static inline int ip6_tnl_xmit_ctl(struct ip6_tnl *t)
799 struct ip6_tnl_parm *p = &t->parms;
801 struct net *net = dev_net(t->dev);
803 if (p->flags & IP6_TNL_F_CAP_XMIT) {
804 struct net_device *ldev = NULL;
808 ldev = dev_get_by_index_rcu(net, p->link);
810 if (unlikely(!ipv6_chk_addr(net, &p->laddr, ldev, 0)))
812 "%s xmit: Local address not yet configured!\n",
814 else if (!ipv6_addr_is_multicast(&p->raddr) &&
815 unlikely(ipv6_chk_addr(net, &p->raddr, NULL, 0)))
817 "%s xmit: Routing loop! "
818 "Remote address found on this node!\n",
827 * ip6_tnl_xmit2 - encapsulate packet and send
828 * @skb: the outgoing socket buffer
829 * @dev: the outgoing tunnel device
830 * @dsfield: dscp code for outer header
831 * @fl: flow of tunneled packet
832 * @encap_limit: encapsulation limit
833 * @pmtu: Path MTU is stored if packet is too big
836 * Build new header and do some sanity checks on the packet before sending
842 * %-EMSGSIZE message too big. return mtu in this case.
845 static int ip6_tnl_xmit2(struct sk_buff *skb,
846 struct net_device *dev,
852 struct net *net = dev_net(dev);
853 struct ip6_tnl *t = netdev_priv(dev);
854 struct net_device_stats *stats = &t->dev->stats;
855 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
856 struct ipv6_tel_txoption opt;
857 struct dst_entry *dst;
858 struct net_device *tdev;
860 unsigned int max_headroom = sizeof(struct ipv6hdr);
865 if ((dst = ip6_tnl_dst_check(t)) != NULL)
868 dst = ip6_route_output(net, NULL, fl);
870 if (dst->error || xfrm_lookup(net, &dst, fl, NULL, 0) < 0)
871 goto tx_err_link_failure;
880 "%s: Local routing loop detected!\n",
882 goto tx_err_dst_release;
884 mtu = dst_mtu(dst) - sizeof (*ipv6h);
885 if (encap_limit >= 0) {
889 if (mtu < IPV6_MIN_MTU)
892 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), mtu);
893 if (skb->len > mtu) {
896 goto tx_err_dst_release;
900 * Okay, now see if we can stuff it in the buffer as-is.
902 max_headroom += LL_RESERVED_SPACE(tdev);
904 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
905 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
906 struct sk_buff *new_skb;
908 if (!(new_skb = skb_realloc_headroom(skb, max_headroom)))
909 goto tx_err_dst_release;
912 skb_set_owner_w(new_skb, skb->sk);
917 skb_dst_set(skb, dst_clone(dst));
919 skb->transport_header = skb->network_header;
922 if (encap_limit >= 0) {
923 init_tel_txopt(&opt, encap_limit);
924 ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
926 skb_push(skb, sizeof(struct ipv6hdr));
927 skb_reset_network_header(skb);
928 ipv6h = ipv6_hdr(skb);
929 *(__be32*)ipv6h = fl->fl6_flowlabel | htonl(0x60000000);
930 dsfield = INET_ECN_encapsulate(0, dsfield);
931 ipv6_change_dsfield(ipv6h, ~INET_ECN_MASK, dsfield);
932 ipv6h->hop_limit = t->parms.hop_limit;
933 ipv6h->nexthdr = proto;
934 ipv6_addr_copy(&ipv6h->saddr, &fl->fl6_src);
935 ipv6_addr_copy(&ipv6h->daddr, &fl->fl6_dst);
938 err = ip6_local_out(skb);
940 if (net_xmit_eval(err) == 0) {
941 stats->tx_bytes += pkt_len;
945 stats->tx_aborted_errors++;
947 ip6_tnl_dst_store(t, dst);
950 stats->tx_carrier_errors++;
951 dst_link_failure(skb);
958 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
960 struct ip6_tnl *t = netdev_priv(dev);
961 struct iphdr *iph = ip_hdr(skb);
962 int encap_limit = -1;
968 if ((t->parms.proto != IPPROTO_IPIP && t->parms.proto != 0) ||
969 !ip6_tnl_xmit_ctl(t))
972 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
973 encap_limit = t->parms.encap_limit;
975 memcpy(&fl, &t->fl, sizeof (fl));
976 fl.proto = IPPROTO_IPIP;
978 dsfield = ipv4_get_dsfield(iph);
980 if ((t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS))
981 fl.fl6_flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
984 err = ip6_tnl_xmit2(skb, dev, dsfield, &fl, encap_limit, &mtu);
986 /* XXX: send ICMP error even if DF is not set. */
987 if (err == -EMSGSIZE)
988 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
997 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
999 struct ip6_tnl *t = netdev_priv(dev);
1000 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
1001 int encap_limit = -1;
1008 if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) ||
1009 !ip6_tnl_xmit_ctl(t) || ip6_tnl_addr_conflict(t, ipv6h))
1012 offset = parse_tlv_tnl_enc_lim(skb, skb_network_header(skb));
1014 struct ipv6_tlv_tnl_enc_lim *tel;
1015 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
1016 if (tel->encap_limit == 0) {
1017 icmpv6_send(skb, ICMPV6_PARAMPROB,
1018 ICMPV6_HDR_FIELD, offset + 2, skb->dev);
1021 encap_limit = tel->encap_limit - 1;
1022 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1023 encap_limit = t->parms.encap_limit;
1025 memcpy(&fl, &t->fl, sizeof (fl));
1026 fl.proto = IPPROTO_IPV6;
1028 dsfield = ipv6_get_dsfield(ipv6h);
1029 if ((t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS))
1030 fl.fl6_flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
1031 if ((t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL))
1032 fl.fl6_flowlabel |= (*(__be32 *) ipv6h & IPV6_FLOWLABEL_MASK);
1034 err = ip6_tnl_xmit2(skb, dev, dsfield, &fl, encap_limit, &mtu);
1036 if (err == -EMSGSIZE)
1037 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, dev);
1045 ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1047 struct ip6_tnl *t = netdev_priv(dev);
1048 struct net_device_stats *stats = &t->dev->stats;
1051 switch (skb->protocol) {
1052 case htons(ETH_P_IP):
1053 ret = ip4ip6_tnl_xmit(skb, dev);
1055 case htons(ETH_P_IPV6):
1056 ret = ip6ip6_tnl_xmit(skb, dev);
1065 return NETDEV_TX_OK;
1069 stats->tx_dropped++;
1071 return NETDEV_TX_OK;
1074 static void ip6_tnl_set_cap(struct ip6_tnl *t)
1076 struct ip6_tnl_parm *p = &t->parms;
1077 int ltype = ipv6_addr_type(&p->laddr);
1078 int rtype = ipv6_addr_type(&p->raddr);
1080 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV);
1082 if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
1083 rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
1084 !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
1085 (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
1086 if (ltype&IPV6_ADDR_UNICAST)
1087 p->flags |= IP6_TNL_F_CAP_XMIT;
1088 if (rtype&IPV6_ADDR_UNICAST)
1089 p->flags |= IP6_TNL_F_CAP_RCV;
1093 static void ip6_tnl_link_config(struct ip6_tnl *t)
1095 struct net_device *dev = t->dev;
1096 struct ip6_tnl_parm *p = &t->parms;
1097 struct flowi *fl = &t->fl;
1099 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1100 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1102 /* Set up flowi template */
1103 ipv6_addr_copy(&fl->fl6_src, &p->laddr);
1104 ipv6_addr_copy(&fl->fl6_dst, &p->raddr);
1106 fl->fl6_flowlabel = 0;
1108 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1109 fl->fl6_flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1110 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1111 fl->fl6_flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1115 if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1116 dev->flags |= IFF_POINTOPOINT;
1118 dev->flags &= ~IFF_POINTOPOINT;
1120 dev->iflink = p->link;
1122 if (p->flags & IP6_TNL_F_CAP_XMIT) {
1123 int strict = (ipv6_addr_type(&p->raddr) &
1124 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1126 struct rt6_info *rt = rt6_lookup(dev_net(dev),
1127 &p->raddr, &p->laddr,
1134 dev->hard_header_len = rt->rt6i_dev->hard_header_len +
1135 sizeof (struct ipv6hdr);
1137 dev->mtu = rt->rt6i_dev->mtu - sizeof (struct ipv6hdr);
1139 if (dev->mtu < IPV6_MIN_MTU)
1140 dev->mtu = IPV6_MIN_MTU;
1142 dst_release(&rt->u.dst);
1147 * ip6_tnl_change - update the tunnel parameters
1148 * @t: tunnel to be changed
1149 * @p: tunnel configuration parameters
1152 * ip6_tnl_change() updates the tunnel parameters
1156 ip6_tnl_change(struct ip6_tnl *t, struct ip6_tnl_parm *p)
1158 ipv6_addr_copy(&t->parms.laddr, &p->laddr);
1159 ipv6_addr_copy(&t->parms.raddr, &p->raddr);
1160 t->parms.flags = p->flags;
1161 t->parms.hop_limit = p->hop_limit;
1162 t->parms.encap_limit = p->encap_limit;
1163 t->parms.flowinfo = p->flowinfo;
1164 t->parms.link = p->link;
1165 t->parms.proto = p->proto;
1166 ip6_tnl_dst_reset(t);
1167 ip6_tnl_link_config(t);
1172 * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1173 * @dev: virtual device associated with tunnel
1174 * @ifr: parameters passed from userspace
1175 * @cmd: command to be performed
1178 * ip6_tnl_ioctl() is used for managing IPv6 tunnels
1181 * The possible commands are the following:
1182 * %SIOCGETTUNNEL: get tunnel parameters for device
1183 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1184 * %SIOCCHGTUNNEL: change tunnel parameters to those given
1185 * %SIOCDELTUNNEL: delete tunnel
1187 * The fallback device "ip6tnl0", created during module
1188 * initialization, can be used for creating other tunnel devices.
1192 * %-EFAULT if unable to copy data to or from userspace,
1193 * %-EPERM if current process hasn't %CAP_NET_ADMIN set
1194 * %-EINVAL if passed tunnel parameters are invalid,
1195 * %-EEXIST if changing a tunnel's parameters would cause a conflict
1196 * %-ENODEV if attempting to change or delete a nonexisting device
1200 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1203 struct ip6_tnl_parm p;
1204 struct ip6_tnl *t = NULL;
1205 struct net *net = dev_net(dev);
1206 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1210 if (dev == ip6n->fb_tnl_dev) {
1211 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p))) {
1215 t = ip6_tnl_locate(net, &p, 0);
1218 t = netdev_priv(dev);
1219 memcpy(&p, &t->parms, sizeof (p));
1220 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof (p))) {
1227 if (!capable(CAP_NET_ADMIN))
1230 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
1233 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1236 t = ip6_tnl_locate(net, &p, cmd == SIOCADDTUNNEL);
1237 if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) {
1239 if (t->dev != dev) {
1244 t = netdev_priv(dev);
1246 ip6_tnl_unlink(ip6n, t);
1247 err = ip6_tnl_change(t, &p);
1248 ip6_tnl_link(ip6n, t);
1249 netdev_state_change(dev);
1253 if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof (p)))
1257 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1261 if (!capable(CAP_NET_ADMIN))
1264 if (dev == ip6n->fb_tnl_dev) {
1266 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
1269 if ((t = ip6_tnl_locate(net, &p, 0)) == NULL)
1272 if (t->dev == ip6n->fb_tnl_dev)
1277 unregister_netdevice(dev);
1286 * ip6_tnl_change_mtu - change mtu manually for tunnel device
1287 * @dev: virtual device associated with tunnel
1288 * @new_mtu: the new mtu
1292 * %-EINVAL if mtu too small
1296 ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1298 if (new_mtu < IPV6_MIN_MTU) {
1306 static const struct net_device_ops ip6_tnl_netdev_ops = {
1307 .ndo_uninit = ip6_tnl_dev_uninit,
1308 .ndo_start_xmit = ip6_tnl_xmit,
1309 .ndo_do_ioctl = ip6_tnl_ioctl,
1310 .ndo_change_mtu = ip6_tnl_change_mtu,
1314 * ip6_tnl_dev_setup - setup virtual tunnel device
1315 * @dev: virtual device associated with tunnel
1318 * Initialize function pointers and device parameters
1321 static void ip6_tnl_dev_setup(struct net_device *dev)
1323 dev->netdev_ops = &ip6_tnl_netdev_ops;
1324 dev->destructor = free_netdev;
1326 dev->type = ARPHRD_TUNNEL6;
1327 dev->hard_header_len = LL_MAX_HEADER + sizeof (struct ipv6hdr);
1328 dev->mtu = ETH_DATA_LEN - sizeof (struct ipv6hdr);
1329 dev->flags |= IFF_NOARP;
1330 dev->addr_len = sizeof(struct in6_addr);
1331 dev->features |= NETIF_F_NETNS_LOCAL;
1336 * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1337 * @dev: virtual device associated with tunnel
1341 ip6_tnl_dev_init_gen(struct net_device *dev)
1343 struct ip6_tnl *t = netdev_priv(dev);
1345 strcpy(t->parms.name, dev->name);
1349 * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1350 * @dev: virtual device associated with tunnel
1353 static void ip6_tnl_dev_init(struct net_device *dev)
1355 struct ip6_tnl *t = netdev_priv(dev);
1356 ip6_tnl_dev_init_gen(dev);
1357 ip6_tnl_link_config(t);
1361 * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1362 * @dev: fallback device
1367 static void ip6_fb_tnl_dev_init(struct net_device *dev)
1369 struct ip6_tnl *t = netdev_priv(dev);
1370 struct net *net = dev_net(dev);
1371 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1373 ip6_tnl_dev_init_gen(dev);
1374 t->parms.proto = IPPROTO_IPV6;
1376 ip6n->tnls_wc[0] = t;
1379 static struct xfrm6_tunnel ip4ip6_handler = {
1380 .handler = ip4ip6_rcv,
1381 .err_handler = ip4ip6_err,
1385 static struct xfrm6_tunnel ip6ip6_handler = {
1386 .handler = ip6ip6_rcv,
1387 .err_handler = ip6ip6_err,
1391 static void ip6_tnl_destroy_tunnels(struct ip6_tnl_net *ip6n)
1397 for (h = 0; h < HASH_SIZE; h++) {
1398 t = ip6n->tnls_r_l[h];
1400 unregister_netdevice_queue(t->dev, &list);
1405 t = ip6n->tnls_wc[0];
1406 unregister_netdevice_queue(t->dev, &list);
1407 unregister_netdevice_many(&list);
1410 static int ip6_tnl_init_net(struct net *net)
1412 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1415 ip6n->tnls[0] = ip6n->tnls_wc;
1416 ip6n->tnls[1] = ip6n->tnls_r_l;
1419 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
1422 if (!ip6n->fb_tnl_dev)
1424 dev_net_set(ip6n->fb_tnl_dev, net);
1426 ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
1428 err = register_netdev(ip6n->fb_tnl_dev);
1434 free_netdev(ip6n->fb_tnl_dev);
1439 static void ip6_tnl_exit_net(struct net *net)
1441 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1444 ip6_tnl_destroy_tunnels(ip6n);
1448 static struct pernet_operations ip6_tnl_net_ops = {
1449 .init = ip6_tnl_init_net,
1450 .exit = ip6_tnl_exit_net,
1451 .id = &ip6_tnl_net_id,
1452 .size = sizeof(struct ip6_tnl_net),
1456 * ip6_tunnel_init - register protocol and reserve needed resources
1458 * Return: 0 on success
1461 static int __init ip6_tunnel_init(void)
1465 if (xfrm6_tunnel_register(&ip4ip6_handler, AF_INET)) {
1466 printk(KERN_ERR "ip6_tunnel init: can't register ip4ip6\n");
1471 if (xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6)) {
1472 printk(KERN_ERR "ip6_tunnel init: can't register ip6ip6\n");
1477 err = register_pernet_device(&ip6_tnl_net_ops);
1482 xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
1484 xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
1490 * ip6_tunnel_cleanup - free resources and unregister protocol
1493 static void __exit ip6_tunnel_cleanup(void)
1495 if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
1496 printk(KERN_INFO "ip6_tunnel close: can't deregister ip4ip6\n");
1498 if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
1499 printk(KERN_INFO "ip6_tunnel close: can't deregister ip6ip6\n");
1501 unregister_pernet_device(&ip6_tnl_net_ops);
1504 module_init(ip6_tunnel_init);
1505 module_exit(ip6_tunnel_cleanup);