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 int ip6_fb_tnl_dev_init(struct net_device *dev);
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;
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];
91 /* lock for the tunnel lists */
92 static DEFINE_RWLOCK(ip6_tnl_lock);
94 static inline struct dst_entry *ip6_tnl_dst_check(struct ip6_tnl *t)
96 struct dst_entry *dst = t->dst_cache;
98 if (dst && dst->obsolete &&
99 dst->ops->check(dst, t->dst_cookie) == NULL) {
108 static inline void ip6_tnl_dst_reset(struct ip6_tnl *t)
110 dst_release(t->dst_cache);
114 static inline void ip6_tnl_dst_store(struct ip6_tnl *t, struct dst_entry *dst)
116 struct rt6_info *rt = (struct rt6_info *) dst;
117 t->dst_cookie = rt->rt6i_node ? rt->rt6i_node->fn_sernum : 0;
118 dst_release(t->dst_cache);
123 * ip6_tnl_lookup - fetch tunnel matching the end-point addresses
124 * @remote: the address of the tunnel exit-point
125 * @local: the address of the tunnel entry-point
128 * tunnel matching given end-points if found,
129 * else fallback tunnel if its device is up,
133 static struct ip6_tnl *
134 ip6_tnl_lookup(struct net *net, struct in6_addr *remote, struct in6_addr *local)
136 unsigned h0 = HASH(remote);
137 unsigned h1 = HASH(local);
139 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
141 for (t = ip6n->tnls_r_l[h0 ^ h1]; t; t = t->next) {
142 if (ipv6_addr_equal(local, &t->parms.laddr) &&
143 ipv6_addr_equal(remote, &t->parms.raddr) &&
144 (t->dev->flags & IFF_UP))
147 if ((t = ip6n->tnls_wc[0]) != NULL && (t->dev->flags & IFF_UP))
154 * ip6_tnl_bucket - get head of list matching given tunnel parameters
155 * @p: parameters containing tunnel end-points
158 * ip6_tnl_bucket() returns the head of the list matching the
159 * &struct in6_addr entries laddr and raddr in @p.
161 * Return: head of IPv6 tunnel list
164 static struct ip6_tnl **
165 ip6_tnl_bucket(struct ip6_tnl_net *ip6n, struct ip6_tnl_parm *p)
167 struct in6_addr *remote = &p->raddr;
168 struct in6_addr *local = &p->laddr;
172 if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
174 h = HASH(remote) ^ HASH(local);
176 return &ip6n->tnls[prio][h];
180 * ip6_tnl_link - add tunnel to hash table
181 * @t: tunnel to be added
185 ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
187 struct ip6_tnl **tp = ip6_tnl_bucket(ip6n, &t->parms);
190 write_lock_bh(&ip6_tnl_lock);
192 write_unlock_bh(&ip6_tnl_lock);
196 * ip6_tnl_unlink - remove tunnel from hash table
197 * @t: tunnel to be removed
201 ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
205 for (tp = ip6_tnl_bucket(ip6n, &t->parms); *tp; tp = &(*tp)->next) {
207 write_lock_bh(&ip6_tnl_lock);
209 write_unlock_bh(&ip6_tnl_lock);
216 * ip6_tnl_create() - create a new tunnel
217 * @p: tunnel parameters
218 * @pt: pointer to new tunnel
221 * Create tunnel matching given parameters.
224 * created tunnel or NULL
227 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct ip6_tnl_parm *p)
229 struct net_device *dev;
233 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
236 strlcpy(name, p->name, IFNAMSIZ);
238 sprintf(name, "ip6tnl%%d");
240 dev = alloc_netdev(sizeof (*t), name, ip6_tnl_dev_setup);
244 dev_net_set(dev, net);
246 if (strchr(name, '%')) {
247 if (dev_alloc_name(dev, name) < 0)
251 t = netdev_priv(dev);
252 dev->init = ip6_tnl_dev_init;
255 if ((err = register_netdevice(dev)) < 0)
259 ip6_tnl_link(ip6n, t);
269 * ip6_tnl_locate - find or create tunnel matching given parameters
270 * @p: tunnel parameters
271 * @create: != 0 if allowed to create new tunnel if no match found
274 * ip6_tnl_locate() first tries to locate an existing tunnel
275 * based on @parms. If this is unsuccessful, but @create is set a new
276 * tunnel device is created and registered for use.
279 * matching tunnel or NULL
282 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
283 struct ip6_tnl_parm *p, int create)
285 struct in6_addr *remote = &p->raddr;
286 struct in6_addr *local = &p->laddr;
288 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
290 for (t = *ip6_tnl_bucket(ip6n, p); t; t = t->next) {
291 if (ipv6_addr_equal(local, &t->parms.laddr) &&
292 ipv6_addr_equal(remote, &t->parms.raddr))
297 return ip6_tnl_create(net, p);
301 * ip6_tnl_dev_uninit - tunnel device uninitializer
302 * @dev: the device to be destroyed
305 * ip6_tnl_dev_uninit() removes tunnel from its list
309 ip6_tnl_dev_uninit(struct net_device *dev)
311 struct ip6_tnl *t = netdev_priv(dev);
312 struct net *net = dev_net(dev);
313 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
315 if (dev == ip6n->fb_tnl_dev) {
316 write_lock_bh(&ip6_tnl_lock);
317 ip6n->tnls_wc[0] = NULL;
318 write_unlock_bh(&ip6_tnl_lock);
320 ip6_tnl_unlink(ip6n, t);
322 ip6_tnl_dst_reset(t);
327 * parse_tvl_tnl_enc_lim - handle encapsulation limit option
328 * @skb: received socket buffer
331 * 0 if none was found,
332 * else index to encapsulation limit
336 parse_tlv_tnl_enc_lim(struct sk_buff *skb, __u8 * raw)
338 struct ipv6hdr *ipv6h = (struct ipv6hdr *) raw;
339 __u8 nexthdr = ipv6h->nexthdr;
340 __u16 off = sizeof (*ipv6h);
342 while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
344 struct ipv6_opt_hdr *hdr;
345 if (raw + off + sizeof (*hdr) > skb->data &&
346 !pskb_may_pull(skb, raw - skb->data + off + sizeof (*hdr)))
349 hdr = (struct ipv6_opt_hdr *) (raw + off);
350 if (nexthdr == NEXTHDR_FRAGMENT) {
351 struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
352 if (frag_hdr->frag_off)
355 } else if (nexthdr == NEXTHDR_AUTH) {
356 optlen = (hdr->hdrlen + 2) << 2;
358 optlen = ipv6_optlen(hdr);
360 if (nexthdr == NEXTHDR_DEST) {
363 struct ipv6_tlv_tnl_enc_lim *tel;
365 /* No more room for encapsulation limit */
366 if (i + sizeof (*tel) > off + optlen)
369 tel = (struct ipv6_tlv_tnl_enc_lim *) &raw[i];
370 /* return index of option if found and valid */
371 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
374 /* else jump to next option */
376 i += tel->length + 2;
381 nexthdr = hdr->nexthdr;
388 * ip6_tnl_err - tunnel error handler
391 * ip6_tnl_err() should handle errors in the tunnel according
392 * to the specifications in RFC 2473.
396 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
397 int *type, int *code, int *msg, __u32 *info, int offset)
399 struct ipv6hdr *ipv6h = (struct ipv6hdr *) skb->data;
402 int rel_type = ICMPV6_DEST_UNREACH;
403 int rel_code = ICMPV6_ADDR_UNREACH;
408 /* If the packet doesn't contain the original IPv6 header we are
409 in trouble since we might need the source address for further
410 processing of the error. */
412 read_lock(&ip6_tnl_lock);
413 if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr,
414 &ipv6h->saddr)) == NULL)
417 if (t->parms.proto != ipproto && t->parms.proto != 0)
424 struct ipv6_tlv_tnl_enc_lim *tel;
426 case ICMPV6_DEST_UNREACH:
429 "%s: Path to destination invalid "
430 "or inactive!\n", t->parms.name);
433 case ICMPV6_TIME_EXCEED:
434 if ((*code) == ICMPV6_EXC_HOPLIMIT) {
437 "%s: Too small hop limit or "
438 "routing loop in tunnel!\n",
443 case ICMPV6_PARAMPROB:
445 if ((*code) == ICMPV6_HDR_FIELD)
446 teli = parse_tlv_tnl_enc_lim(skb, skb->data);
448 if (teli && teli == *info - 2) {
449 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
450 if (tel->encap_limit == 0) {
453 "%s: Too small encapsulation "
454 "limit or routing loop in "
455 "tunnel!\n", t->parms.name);
458 } else if (net_ratelimit()) {
460 "%s: Recipient unable to parse tunneled "
461 "packet!\n ", t->parms.name);
464 case ICMPV6_PKT_TOOBIG:
465 mtu = *info - offset;
466 if (mtu < IPV6_MIN_MTU)
470 if ((len = sizeof (*ipv6h) + ntohs(ipv6h->payload_len)) > mtu) {
471 rel_type = ICMPV6_PKT_TOOBIG;
485 read_unlock(&ip6_tnl_lock);
490 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
491 int type, int code, int offset, __be32 info)
496 __u32 rel_info = ntohl(info);
498 struct sk_buff *skb2;
503 err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
504 &rel_msg, &rel_info, offset);
512 case ICMPV6_DEST_UNREACH:
513 if (rel_code != ICMPV6_ADDR_UNREACH)
515 rel_type = ICMP_DEST_UNREACH;
516 rel_code = ICMP_HOST_UNREACH;
518 case ICMPV6_PKT_TOOBIG:
521 rel_type = ICMP_DEST_UNREACH;
522 rel_code = ICMP_FRAG_NEEDED;
528 if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
531 skb2 = skb_clone(skb, GFP_ATOMIC);
535 dst_release(skb2->dst);
537 skb_pull(skb2, offset);
538 skb_reset_network_header(skb2);
541 /* Try to guess incoming interface */
542 memset(&fl, 0, sizeof(fl));
543 fl.fl4_dst = eiph->saddr;
544 fl.fl4_tos = RT_TOS(eiph->tos);
545 fl.proto = IPPROTO_IPIP;
546 if (ip_route_output_key(dev_net(skb->dev), &rt, &fl))
549 skb2->dev = rt->u.dst.dev;
551 /* route "incoming" packet */
552 if (rt->rt_flags & RTCF_LOCAL) {
555 fl.fl4_dst = eiph->daddr;
556 fl.fl4_src = eiph->saddr;
557 fl.fl4_tos = eiph->tos;
558 if (ip_route_output_key(dev_net(skb->dev), &rt, &fl) ||
559 rt->u.dst.dev->type != ARPHRD_TUNNEL) {
563 skb2->dst = (struct dst_entry *)rt;
566 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
568 skb2->dst->dev->type != ARPHRD_TUNNEL)
572 /* change mtu on this route */
573 if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
574 if (rel_info > dst_mtu(skb2->dst))
577 skb2->dst->ops->update_pmtu(skb2->dst, rel_info);
580 icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
588 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
589 int type, int code, int offset, __be32 info)
594 __u32 rel_info = ntohl(info);
597 err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
598 &rel_msg, &rel_info, offset);
602 if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
604 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
609 dst_release(skb2->dst);
611 skb_pull(skb2, offset);
612 skb_reset_network_header(skb2);
614 /* Try to guess incoming interface */
615 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
618 if (rt && rt->rt6i_dev)
619 skb2->dev = rt->rt6i_dev;
621 icmpv6_send(skb2, rel_type, rel_code, rel_info, skb2->dev);
624 dst_release(&rt->u.dst);
632 static void ip4ip6_dscp_ecn_decapsulate(struct ip6_tnl *t,
633 struct ipv6hdr *ipv6h,
636 __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
638 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
639 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
641 if (INET_ECN_is_ce(dsfield))
642 IP_ECN_set_ce(ip_hdr(skb));
645 static void ip6ip6_dscp_ecn_decapsulate(struct ip6_tnl *t,
646 struct ipv6hdr *ipv6h,
649 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
650 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
652 if (INET_ECN_is_ce(ipv6_get_dsfield(ipv6h)))
653 IP6_ECN_set_ce(ipv6_hdr(skb));
656 static inline int ip6_tnl_rcv_ctl(struct ip6_tnl *t)
658 struct ip6_tnl_parm *p = &t->parms;
660 struct net *net = dev_net(t->dev);
662 if (p->flags & IP6_TNL_F_CAP_RCV) {
663 struct net_device *ldev = NULL;
666 ldev = dev_get_by_index(net, p->link);
668 if ((ipv6_addr_is_multicast(&p->laddr) ||
669 likely(ipv6_chk_addr(net, &p->laddr, ldev, 0))) &&
670 likely(!ipv6_chk_addr(net, &p->raddr, NULL, 0)))
680 * ip6_tnl_rcv - decapsulate IPv6 packet and retransmit it locally
681 * @skb: received socket buffer
682 * @protocol: ethernet protocol ID
683 * @dscp_ecn_decapsulate: the function to decapsulate DSCP code and ECN
688 static int ip6_tnl_rcv(struct sk_buff *skb, __u16 protocol,
690 void (*dscp_ecn_decapsulate)(struct ip6_tnl *t,
691 struct ipv6hdr *ipv6h,
692 struct sk_buff *skb))
695 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
697 read_lock(&ip6_tnl_lock);
699 if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr,
700 &ipv6h->daddr)) != NULL) {
701 if (t->parms.proto != ipproto && t->parms.proto != 0) {
702 read_unlock(&ip6_tnl_lock);
706 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
707 read_unlock(&ip6_tnl_lock);
711 if (!ip6_tnl_rcv_ctl(t)) {
712 t->dev->stats.rx_dropped++;
713 read_unlock(&ip6_tnl_lock);
717 skb->mac_header = skb->network_header;
718 skb_reset_network_header(skb);
719 skb->protocol = htons(protocol);
720 skb->pkt_type = PACKET_HOST;
721 memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
723 dst_release(skb->dst);
727 dscp_ecn_decapsulate(t, ipv6h, skb);
729 t->dev->stats.rx_packets++;
730 t->dev->stats.rx_bytes += skb->len;
732 read_unlock(&ip6_tnl_lock);
735 read_unlock(&ip6_tnl_lock);
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;
804 ldev = dev_get_by_index(net, p->link);
806 if (unlikely(!ipv6_chk_addr(net, &p->laddr, ldev, 0)))
808 "%s xmit: Local address not yet configured!\n",
810 else if (!ipv6_addr_is_multicast(&p->raddr) &&
811 unlikely(ipv6_chk_addr(net, &p->raddr, NULL, 0)))
813 "%s xmit: Routing loop! "
814 "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 ip6_tnl *t = netdev_priv(dev);
850 struct net_device_stats *stats = &t->dev->stats;
851 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
852 struct ipv6_tel_txoption opt;
853 struct dst_entry *dst;
854 struct net_device *tdev;
856 unsigned int max_headroom = sizeof(struct ipv6hdr);
861 if ((dst = ip6_tnl_dst_check(t)) != NULL)
864 dst = ip6_route_output(dev_net(dev), NULL, fl);
866 if (dst->error || xfrm_lookup(&dst, fl, NULL, 0) < 0)
867 goto tx_err_link_failure;
876 "%s: Local routing loop detected!\n",
878 goto tx_err_dst_release;
880 mtu = dst_mtu(dst) - sizeof (*ipv6h);
881 if (encap_limit >= 0) {
885 if (mtu < IPV6_MIN_MTU)
888 skb->dst->ops->update_pmtu(skb->dst, mtu);
889 if (skb->len > mtu) {
892 goto tx_err_dst_release;
896 * Okay, now see if we can stuff it in the buffer as-is.
898 max_headroom += LL_RESERVED_SPACE(tdev);
900 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
901 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
902 struct sk_buff *new_skb;
904 if (!(new_skb = skb_realloc_headroom(skb, max_headroom)))
905 goto tx_err_dst_release;
908 skb_set_owner_w(new_skb, skb->sk);
912 dst_release(skb->dst);
913 skb->dst = dst_clone(dst);
915 skb->transport_header = skb->network_header;
918 if (encap_limit >= 0) {
919 init_tel_txopt(&opt, encap_limit);
920 ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
922 skb_push(skb, sizeof(struct ipv6hdr));
923 skb_reset_network_header(skb);
924 ipv6h = ipv6_hdr(skb);
925 *(__be32*)ipv6h = fl->fl6_flowlabel | htonl(0x60000000);
926 dsfield = INET_ECN_encapsulate(0, dsfield);
927 ipv6_change_dsfield(ipv6h, ~INET_ECN_MASK, dsfield);
928 ipv6h->hop_limit = t->parms.hop_limit;
929 ipv6h->nexthdr = proto;
930 ipv6_addr_copy(&ipv6h->saddr, &fl->fl6_src);
931 ipv6_addr_copy(&ipv6h->daddr, &fl->fl6_dst);
934 err = ip6_local_out(skb);
936 if (net_xmit_eval(err) == 0) {
937 stats->tx_bytes += pkt_len;
941 stats->tx_aborted_errors++;
943 ip6_tnl_dst_store(t, dst);
946 stats->tx_carrier_errors++;
947 dst_link_failure(skb);
954 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
956 struct ip6_tnl *t = netdev_priv(dev);
957 struct iphdr *iph = ip_hdr(skb);
958 int encap_limit = -1;
964 if ((t->parms.proto != IPPROTO_IPIP && t->parms.proto != 0) ||
965 !ip6_tnl_xmit_ctl(t))
968 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
969 encap_limit = t->parms.encap_limit;
971 memcpy(&fl, &t->fl, sizeof (fl));
972 fl.proto = IPPROTO_IPIP;
974 dsfield = ipv4_get_dsfield(iph);
976 if ((t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS))
977 fl.fl6_flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
980 err = ip6_tnl_xmit2(skb, dev, dsfield, &fl, encap_limit, &mtu);
982 /* XXX: send ICMP error even if DF is not set. */
983 if (err == -EMSGSIZE)
984 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
993 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
995 struct ip6_tnl *t = netdev_priv(dev);
996 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
997 int encap_limit = -1;
1004 if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) ||
1005 !ip6_tnl_xmit_ctl(t) || ip6_tnl_addr_conflict(t, ipv6h))
1008 offset = parse_tlv_tnl_enc_lim(skb, skb_network_header(skb));
1010 struct ipv6_tlv_tnl_enc_lim *tel;
1011 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
1012 if (tel->encap_limit == 0) {
1013 icmpv6_send(skb, ICMPV6_PARAMPROB,
1014 ICMPV6_HDR_FIELD, offset + 2, skb->dev);
1017 encap_limit = tel->encap_limit - 1;
1018 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1019 encap_limit = t->parms.encap_limit;
1021 memcpy(&fl, &t->fl, sizeof (fl));
1022 fl.proto = IPPROTO_IPV6;
1024 dsfield = ipv6_get_dsfield(ipv6h);
1025 if ((t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS))
1026 fl.fl6_flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
1027 if ((t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL))
1028 fl.fl6_flowlabel |= (*(__be32 *) ipv6h & IPV6_FLOWLABEL_MASK);
1030 err = ip6_tnl_xmit2(skb, dev, dsfield, &fl, encap_limit, &mtu);
1032 if (err == -EMSGSIZE)
1033 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, dev);
1041 ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1043 struct ip6_tnl *t = netdev_priv(dev);
1044 struct net_device_stats *stats = &t->dev->stats;
1047 if (t->recursion++) {
1048 stats->collisions++;
1052 switch (skb->protocol) {
1053 case htons(ETH_P_IP):
1054 ret = ip4ip6_tnl_xmit(skb, dev);
1056 case htons(ETH_P_IPV6):
1057 ret = ip6ip6_tnl_xmit(skb, dev);
1071 stats->tx_dropped++;
1077 static void ip6_tnl_set_cap(struct ip6_tnl *t)
1079 struct ip6_tnl_parm *p = &t->parms;
1080 int ltype = ipv6_addr_type(&p->laddr);
1081 int rtype = ipv6_addr_type(&p->raddr);
1083 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV);
1085 if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
1086 rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
1087 !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
1088 (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
1089 if (ltype&IPV6_ADDR_UNICAST)
1090 p->flags |= IP6_TNL_F_CAP_XMIT;
1091 if (rtype&IPV6_ADDR_UNICAST)
1092 p->flags |= IP6_TNL_F_CAP_RCV;
1096 static void ip6_tnl_link_config(struct ip6_tnl *t)
1098 struct net_device *dev = t->dev;
1099 struct ip6_tnl_parm *p = &t->parms;
1100 struct flowi *fl = &t->fl;
1102 memcpy(&dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1103 memcpy(&dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1105 /* Set up flowi template */
1106 ipv6_addr_copy(&fl->fl6_src, &p->laddr);
1107 ipv6_addr_copy(&fl->fl6_dst, &p->raddr);
1109 fl->fl6_flowlabel = 0;
1111 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1112 fl->fl6_flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1113 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1114 fl->fl6_flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1118 if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1119 dev->flags |= IFF_POINTOPOINT;
1121 dev->flags &= ~IFF_POINTOPOINT;
1123 dev->iflink = p->link;
1125 if (p->flags & IP6_TNL_F_CAP_XMIT) {
1126 int strict = (ipv6_addr_type(&p->raddr) &
1127 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1129 struct rt6_info *rt = rt6_lookup(dev_net(dev),
1130 &p->raddr, &p->laddr,
1137 dev->hard_header_len = rt->rt6i_dev->hard_header_len +
1138 sizeof (struct ipv6hdr);
1140 dev->mtu = rt->rt6i_dev->mtu - sizeof (struct ipv6hdr);
1142 if (dev->mtu < IPV6_MIN_MTU)
1143 dev->mtu = IPV6_MIN_MTU;
1145 dst_release(&rt->u.dst);
1150 * ip6_tnl_change - update the tunnel parameters
1151 * @t: tunnel to be changed
1152 * @p: tunnel configuration parameters
1153 * @active: != 0 if tunnel is ready for use
1156 * ip6_tnl_change() updates the tunnel parameters
1160 ip6_tnl_change(struct ip6_tnl *t, struct ip6_tnl_parm *p)
1162 ipv6_addr_copy(&t->parms.laddr, &p->laddr);
1163 ipv6_addr_copy(&t->parms.raddr, &p->raddr);
1164 t->parms.flags = p->flags;
1165 t->parms.hop_limit = p->hop_limit;
1166 t->parms.encap_limit = p->encap_limit;
1167 t->parms.flowinfo = p->flowinfo;
1168 t->parms.link = p->link;
1169 t->parms.proto = p->proto;
1170 ip6_tnl_dst_reset(t);
1171 ip6_tnl_link_config(t);
1176 * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1177 * @dev: virtual device associated with tunnel
1178 * @ifr: parameters passed from userspace
1179 * @cmd: command to be performed
1182 * ip6_tnl_ioctl() is used for managing IPv6 tunnels
1185 * The possible commands are the following:
1186 * %SIOCGETTUNNEL: get tunnel parameters for device
1187 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1188 * %SIOCCHGTUNNEL: change tunnel parameters to those given
1189 * %SIOCDELTUNNEL: delete tunnel
1191 * The fallback device "ip6tnl0", created during module
1192 * initialization, can be used for creating other tunnel devices.
1196 * %-EFAULT if unable to copy data to or from userspace,
1197 * %-EPERM if current process hasn't %CAP_NET_ADMIN set
1198 * %-EINVAL if passed tunnel parameters are invalid,
1199 * %-EEXIST if changing a tunnel's parameters would cause a conflict
1200 * %-ENODEV if attempting to change or delete a nonexisting device
1204 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1207 struct ip6_tnl_parm p;
1208 struct ip6_tnl *t = NULL;
1209 struct net *net = dev_net(dev);
1210 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1214 if (dev == ip6n->fb_tnl_dev) {
1215 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p))) {
1219 t = ip6_tnl_locate(net, &p, 0);
1222 t = netdev_priv(dev);
1223 memcpy(&p, &t->parms, sizeof (p));
1224 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof (p))) {
1231 if (!capable(CAP_NET_ADMIN))
1234 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
1237 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1240 t = ip6_tnl_locate(net, &p, cmd == SIOCADDTUNNEL);
1241 if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) {
1243 if (t->dev != dev) {
1248 t = netdev_priv(dev);
1250 ip6_tnl_unlink(ip6n, t);
1251 err = ip6_tnl_change(t, &p);
1252 ip6_tnl_link(ip6n, t);
1253 netdev_state_change(dev);
1257 if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof (p)))
1261 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1265 if (!capable(CAP_NET_ADMIN))
1268 if (dev == ip6n->fb_tnl_dev) {
1270 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
1273 if ((t = ip6_tnl_locate(net, &p, 0)) == NULL)
1276 if (t->dev == ip6n->fb_tnl_dev)
1281 unregister_netdevice(dev);
1290 * ip6_tnl_change_mtu - change mtu manually for tunnel device
1291 * @dev: virtual device associated with tunnel
1292 * @new_mtu: the new mtu
1296 * %-EINVAL if mtu too small
1300 ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1302 if (new_mtu < IPV6_MIN_MTU) {
1310 * ip6_tnl_dev_setup - setup virtual tunnel device
1311 * @dev: virtual device associated with tunnel
1314 * Initialize function pointers and device parameters
1317 static void ip6_tnl_dev_setup(struct net_device *dev)
1319 dev->uninit = ip6_tnl_dev_uninit;
1320 dev->destructor = free_netdev;
1321 dev->hard_start_xmit = ip6_tnl_xmit;
1322 dev->do_ioctl = ip6_tnl_ioctl;
1323 dev->change_mtu = ip6_tnl_change_mtu;
1325 dev->type = ARPHRD_TUNNEL6;
1326 dev->hard_header_len = LL_MAX_HEADER + sizeof (struct ipv6hdr);
1327 dev->mtu = ETH_DATA_LEN - sizeof (struct ipv6hdr);
1328 dev->flags |= IFF_NOARP;
1329 dev->addr_len = sizeof(struct in6_addr);
1330 dev->features |= NETIF_F_NETNS_LOCAL;
1335 * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1336 * @dev: virtual device associated with tunnel
1340 ip6_tnl_dev_init_gen(struct net_device *dev)
1342 struct ip6_tnl *t = netdev_priv(dev);
1344 strcpy(t->parms.name, dev->name);
1348 * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1349 * @dev: virtual device associated with tunnel
1353 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);
1362 * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1363 * @dev: fallback device
1369 ip6_fb_tnl_dev_init(struct net_device *dev)
1371 struct ip6_tnl *t = netdev_priv(dev);
1372 struct net *net = dev_net(dev);
1373 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1375 ip6_tnl_dev_init_gen(dev);
1376 t->parms.proto = IPPROTO_IPV6;
1378 ip6n->tnls_wc[0] = t;
1382 static struct xfrm6_tunnel ip4ip6_handler = {
1383 .handler = ip4ip6_rcv,
1384 .err_handler = ip4ip6_err,
1388 static struct xfrm6_tunnel ip6ip6_handler = {
1389 .handler = ip6ip6_rcv,
1390 .err_handler = ip6ip6_err,
1394 static void ip6_tnl_destroy_tunnels(struct ip6_tnl_net *ip6n)
1399 for (h = 0; h < HASH_SIZE; h++) {
1400 while ((t = ip6n->tnls_r_l[h]) != NULL)
1401 unregister_netdevice(t->dev);
1404 t = ip6n->tnls_wc[0];
1405 unregister_netdevice(t->dev);
1408 static int ip6_tnl_init_net(struct net *net)
1411 struct ip6_tnl_net *ip6n;
1414 ip6n = kzalloc(sizeof(struct ip6_tnl_net), GFP_KERNEL);
1418 err = net_assign_generic(net, ip6_tnl_net_id, ip6n);
1422 ip6n->tnls[0] = ip6n->tnls_wc;
1423 ip6n->tnls[1] = ip6n->tnls_r_l;
1426 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
1429 if (!ip6n->fb_tnl_dev)
1432 ip6n->fb_tnl_dev->init = ip6_fb_tnl_dev_init;
1433 dev_net_set(ip6n->fb_tnl_dev, net);
1435 err = register_netdev(ip6n->fb_tnl_dev);
1441 free_netdev(ip6n->fb_tnl_dev);
1450 static void ip6_tnl_exit_net(struct net *net)
1452 struct ip6_tnl_net *ip6n;
1454 ip6n = net_generic(net, ip6_tnl_net_id);
1456 ip6_tnl_destroy_tunnels(ip6n);
1461 static struct pernet_operations ip6_tnl_net_ops = {
1462 .init = ip6_tnl_init_net,
1463 .exit = ip6_tnl_exit_net,
1467 * ip6_tunnel_init - register protocol and reserve needed resources
1469 * Return: 0 on success
1472 static int __init ip6_tunnel_init(void)
1476 if (xfrm6_tunnel_register(&ip4ip6_handler, AF_INET)) {
1477 printk(KERN_ERR "ip6_tunnel init: can't register ip4ip6\n");
1482 if (xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6)) {
1483 printk(KERN_ERR "ip6_tunnel init: can't register ip6ip6\n");
1488 err = register_pernet_gen_device(&ip6_tnl_net_id, &ip6_tnl_net_ops);
1493 xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
1495 xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
1501 * ip6_tunnel_cleanup - free resources and unregister protocol
1504 static void __exit ip6_tunnel_cleanup(void)
1506 if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
1507 printk(KERN_INFO "ip6_tunnel close: can't deregister ip4ip6\n");
1509 if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
1510 printk(KERN_INFO "ip6_tunnel close: can't deregister ip6ip6\n");
1512 unregister_pernet_gen_device(ip6_tnl_net_id, &ip6_tnl_net_ops);
1515 module_init(ip6_tunnel_init);
1516 module_exit(ip6_tunnel_cleanup);