2 * IPv6 tunneling device
3 * Linux INET6 implementation
6 * Ville Nuorvala <vnuorval@tcs.hut.fi>
7 * Yasuyuki Kozakai <kozakai@linux-ipv6.org>
12 * linux/net/ipv6/sit.c and linux/net/ipv4/ipip.c
16 * This program is free software; you can redistribute it and/or
17 * modify it under the terms of the GNU General Public License
18 * as published by the Free Software Foundation; either version
19 * 2 of the License, or (at your option) any later version.
23 #include <linux/module.h>
24 #include <linux/capability.h>
25 #include <linux/errno.h>
26 #include <linux/types.h>
27 #include <linux/sockios.h>
28 #include <linux/icmp.h>
32 #include <linux/if_tunnel.h>
33 #include <linux/net.h>
34 #include <linux/in6.h>
35 #include <linux/netdevice.h>
36 #include <linux/if_arp.h>
37 #include <linux/icmpv6.h>
38 #include <linux/init.h>
39 #include <linux/route.h>
40 #include <linux/rtnetlink.h>
41 #include <linux/netfilter_ipv6.h>
43 #include <asm/uaccess.h>
44 #include <asm/atomic.h>
49 #include <net/ip6_route.h>
50 #include <net/addrconf.h>
51 #include <net/ip6_tunnel.h>
53 #include <net/dsfield.h>
54 #include <net/inet_ecn.h>
55 #include <net/net_namespace.h>
56 #include <net/netns/generic.h>
58 MODULE_AUTHOR("Ville Nuorvala");
59 MODULE_DESCRIPTION("IPv6 tunneling device");
60 MODULE_LICENSE("GPL");
62 #define IPV6_TLV_TEL_DST_SIZE 8
65 #define IP6_TNL_TRACE(x...) printk(KERN_DEBUG "%s:" x "\n", __func__)
67 #define IP6_TNL_TRACE(x...) do {;} while(0)
70 #define IPV6_TCLASS_MASK (IPV6_FLOWINFO_MASK & ~IPV6_FLOWLABEL_MASK)
71 #define IPV6_TCLASS_SHIFT 20
75 #define HASH(addr) ((__force u32)((addr)->s6_addr32[0] ^ (addr)->s6_addr32[1] ^ \
76 (addr)->s6_addr32[2] ^ (addr)->s6_addr32[3]) & \
79 static int ip6_fb_tnl_dev_init(struct net_device *dev);
80 static int ip6_tnl_dev_init(struct net_device *dev);
81 static void ip6_tnl_dev_setup(struct net_device *dev);
83 static int ip6_tnl_net_id;
87 /* the IPv6 tunnel fallback device */
88 static struct net_device *ip6_fb_tnl_dev;
91 /* lists for storing tunnels in use */
92 static struct ip6_tnl *tnls_r_l[HASH_SIZE];
93 static struct ip6_tnl *tnls_wc[1];
94 static struct ip6_tnl **tnls[2] = { tnls_wc, tnls_r_l };
96 /* lock for the tunnel lists */
97 static DEFINE_RWLOCK(ip6_tnl_lock);
99 static inline struct dst_entry *ip6_tnl_dst_check(struct ip6_tnl *t)
101 struct dst_entry *dst = t->dst_cache;
103 if (dst && dst->obsolete &&
104 dst->ops->check(dst, t->dst_cookie) == NULL) {
113 static inline void ip6_tnl_dst_reset(struct ip6_tnl *t)
115 dst_release(t->dst_cache);
119 static inline void ip6_tnl_dst_store(struct ip6_tnl *t, struct dst_entry *dst)
121 struct rt6_info *rt = (struct rt6_info *) dst;
122 t->dst_cookie = rt->rt6i_node ? rt->rt6i_node->fn_sernum : 0;
123 dst_release(t->dst_cache);
128 * ip6_tnl_lookup - fetch tunnel matching the end-point addresses
129 * @remote: the address of the tunnel exit-point
130 * @local: the address of the tunnel entry-point
133 * tunnel matching given end-points if found,
134 * else fallback tunnel if its device is up,
138 static struct ip6_tnl *
139 ip6_tnl_lookup(struct in6_addr *remote, struct in6_addr *local)
141 unsigned h0 = HASH(remote);
142 unsigned h1 = HASH(local);
145 for (t = tnls_r_l[h0 ^ h1]; t; t = t->next) {
146 if (ipv6_addr_equal(local, &t->parms.laddr) &&
147 ipv6_addr_equal(remote, &t->parms.raddr) &&
148 (t->dev->flags & IFF_UP))
151 if ((t = tnls_wc[0]) != NULL && (t->dev->flags & IFF_UP))
158 * ip6_tnl_bucket - get head of list matching given tunnel parameters
159 * @p: parameters containing tunnel end-points
162 * ip6_tnl_bucket() returns the head of the list matching the
163 * &struct in6_addr entries laddr and raddr in @p.
165 * Return: head of IPv6 tunnel list
168 static struct ip6_tnl **
169 ip6_tnl_bucket(struct ip6_tnl_parm *p)
171 struct in6_addr *remote = &p->raddr;
172 struct in6_addr *local = &p->laddr;
176 if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
178 h = HASH(remote) ^ HASH(local);
180 return &tnls[prio][h];
184 * ip6_tnl_link - add tunnel to hash table
185 * @t: tunnel to be added
189 ip6_tnl_link(struct ip6_tnl *t)
191 struct ip6_tnl **tp = ip6_tnl_bucket(&t->parms);
194 write_lock_bh(&ip6_tnl_lock);
196 write_unlock_bh(&ip6_tnl_lock);
200 * ip6_tnl_unlink - remove tunnel from hash table
201 * @t: tunnel to be removed
205 ip6_tnl_unlink(struct ip6_tnl *t)
209 for (tp = ip6_tnl_bucket(&t->parms); *tp; tp = &(*tp)->next) {
211 write_lock_bh(&ip6_tnl_lock);
213 write_unlock_bh(&ip6_tnl_lock);
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 ip6_tnl_parm *p)
233 struct net_device *dev;
239 strlcpy(name, p->name, IFNAMSIZ);
241 sprintf(name, "ip6tnl%%d");
243 dev = alloc_netdev(sizeof (*t), name, ip6_tnl_dev_setup);
247 if (strchr(name, '%')) {
248 if (dev_alloc_name(dev, name) < 0)
252 t = netdev_priv(dev);
253 dev->init = ip6_tnl_dev_init;
256 if ((err = register_netdevice(dev)) < 0)
270 * ip6_tnl_locate - find or create tunnel matching given parameters
271 * @p: tunnel parameters
272 * @create: != 0 if allowed to create new tunnel if no match found
275 * ip6_tnl_locate() first tries to locate an existing tunnel
276 * based on @parms. If this is unsuccessful, but @create is set a new
277 * tunnel device is created and registered for use.
280 * matching tunnel or NULL
283 static struct ip6_tnl *ip6_tnl_locate(struct ip6_tnl_parm *p, int create)
285 struct in6_addr *remote = &p->raddr;
286 struct in6_addr *local = &p->laddr;
289 for (t = *ip6_tnl_bucket(p); t; t = t->next) {
290 if (ipv6_addr_equal(local, &t->parms.laddr) &&
291 ipv6_addr_equal(remote, &t->parms.raddr))
296 return ip6_tnl_create(p);
300 * ip6_tnl_dev_uninit - tunnel device uninitializer
301 * @dev: the device to be destroyed
304 * ip6_tnl_dev_uninit() removes tunnel from its list
308 ip6_tnl_dev_uninit(struct net_device *dev)
310 struct ip6_tnl *t = netdev_priv(dev);
312 if (dev == ip6_fb_tnl_dev) {
313 write_lock_bh(&ip6_tnl_lock);
315 write_unlock_bh(&ip6_tnl_lock);
319 ip6_tnl_dst_reset(t);
324 * parse_tvl_tnl_enc_lim - handle encapsulation limit option
325 * @skb: received socket buffer
328 * 0 if none was found,
329 * else index to encapsulation limit
333 parse_tlv_tnl_enc_lim(struct sk_buff *skb, __u8 * raw)
335 struct ipv6hdr *ipv6h = (struct ipv6hdr *) raw;
336 __u8 nexthdr = ipv6h->nexthdr;
337 __u16 off = sizeof (*ipv6h);
339 while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
341 struct ipv6_opt_hdr *hdr;
342 if (raw + off + sizeof (*hdr) > skb->data &&
343 !pskb_may_pull(skb, raw - skb->data + off + sizeof (*hdr)))
346 hdr = (struct ipv6_opt_hdr *) (raw + off);
347 if (nexthdr == NEXTHDR_FRAGMENT) {
348 struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
349 if (frag_hdr->frag_off)
352 } else if (nexthdr == NEXTHDR_AUTH) {
353 optlen = (hdr->hdrlen + 2) << 2;
355 optlen = ipv6_optlen(hdr);
357 if (nexthdr == NEXTHDR_DEST) {
360 struct ipv6_tlv_tnl_enc_lim *tel;
362 /* No more room for encapsulation limit */
363 if (i + sizeof (*tel) > off + optlen)
366 tel = (struct ipv6_tlv_tnl_enc_lim *) &raw[i];
367 /* return index of option if found and valid */
368 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
371 /* else jump to next option */
373 i += tel->length + 2;
378 nexthdr = hdr->nexthdr;
385 * ip6_tnl_err - tunnel error handler
388 * ip6_tnl_err() should handle errors in the tunnel according
389 * to the specifications in RFC 2473.
393 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
394 int *type, int *code, int *msg, __u32 *info, int offset)
396 struct ipv6hdr *ipv6h = (struct ipv6hdr *) skb->data;
399 int rel_type = ICMPV6_DEST_UNREACH;
400 int rel_code = ICMPV6_ADDR_UNREACH;
405 /* If the packet doesn't contain the original IPv6 header we are
406 in trouble since we might need the source address for further
407 processing of the error. */
409 read_lock(&ip6_tnl_lock);
410 if ((t = ip6_tnl_lookup(&ipv6h->daddr, &ipv6h->saddr)) == NULL)
413 if (t->parms.proto != ipproto && t->parms.proto != 0)
420 struct ipv6_tlv_tnl_enc_lim *tel;
422 case ICMPV6_DEST_UNREACH:
425 "%s: Path to destination invalid "
426 "or inactive!\n", t->parms.name);
429 case ICMPV6_TIME_EXCEED:
430 if ((*code) == ICMPV6_EXC_HOPLIMIT) {
433 "%s: Too small hop limit or "
434 "routing loop in tunnel!\n",
439 case ICMPV6_PARAMPROB:
441 if ((*code) == ICMPV6_HDR_FIELD)
442 teli = parse_tlv_tnl_enc_lim(skb, skb->data);
444 if (teli && teli == *info - 2) {
445 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
446 if (tel->encap_limit == 0) {
449 "%s: Too small encapsulation "
450 "limit or routing loop in "
451 "tunnel!\n", t->parms.name);
454 } else if (net_ratelimit()) {
456 "%s: Recipient unable to parse tunneled "
457 "packet!\n ", t->parms.name);
460 case ICMPV6_PKT_TOOBIG:
461 mtu = *info - offset;
462 if (mtu < IPV6_MIN_MTU)
466 if ((len = sizeof (*ipv6h) + ntohs(ipv6h->payload_len)) > mtu) {
467 rel_type = ICMPV6_PKT_TOOBIG;
481 read_unlock(&ip6_tnl_lock);
486 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
487 int type, int code, int offset, __be32 info)
492 __u32 rel_info = ntohl(info);
494 struct sk_buff *skb2;
499 err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
500 &rel_msg, &rel_info, offset);
508 case ICMPV6_DEST_UNREACH:
509 if (rel_code != ICMPV6_ADDR_UNREACH)
511 rel_type = ICMP_DEST_UNREACH;
512 rel_code = ICMP_HOST_UNREACH;
514 case ICMPV6_PKT_TOOBIG:
517 rel_type = ICMP_DEST_UNREACH;
518 rel_code = ICMP_FRAG_NEEDED;
524 if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
527 skb2 = skb_clone(skb, GFP_ATOMIC);
531 dst_release(skb2->dst);
533 skb_pull(skb2, offset);
534 skb_reset_network_header(skb2);
537 /* Try to guess incoming interface */
538 memset(&fl, 0, sizeof(fl));
539 fl.fl4_dst = eiph->saddr;
540 fl.fl4_tos = RT_TOS(eiph->tos);
541 fl.proto = IPPROTO_IPIP;
542 if (ip_route_output_key(&init_net, &rt, &fl))
545 skb2->dev = rt->u.dst.dev;
547 /* route "incoming" packet */
548 if (rt->rt_flags & RTCF_LOCAL) {
551 fl.fl4_dst = eiph->daddr;
552 fl.fl4_src = eiph->saddr;
553 fl.fl4_tos = eiph->tos;
554 if (ip_route_output_key(&init_net, &rt, &fl) ||
555 rt->u.dst.dev->type != ARPHRD_TUNNEL) {
559 skb2->dst = (struct dst_entry *)rt;
562 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
564 skb2->dst->dev->type != ARPHRD_TUNNEL)
568 /* change mtu on this route */
569 if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
570 if (rel_info > dst_mtu(skb2->dst))
573 skb2->dst->ops->update_pmtu(skb2->dst, rel_info);
576 icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
584 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
585 int type, int code, int offset, __be32 info)
590 __u32 rel_info = ntohl(info);
593 err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
594 &rel_msg, &rel_info, offset);
598 if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
600 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
605 dst_release(skb2->dst);
607 skb_pull(skb2, offset);
608 skb_reset_network_header(skb2);
610 /* Try to guess incoming interface */
611 rt = rt6_lookup(&init_net, &ipv6_hdr(skb2)->saddr, NULL, 0, 0);
613 if (rt && rt->rt6i_dev)
614 skb2->dev = rt->rt6i_dev;
616 icmpv6_send(skb2, rel_type, rel_code, rel_info, skb2->dev);
619 dst_release(&rt->u.dst);
627 static void ip4ip6_dscp_ecn_decapsulate(struct ip6_tnl *t,
628 struct ipv6hdr *ipv6h,
631 __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
633 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
634 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
636 if (INET_ECN_is_ce(dsfield))
637 IP_ECN_set_ce(ip_hdr(skb));
640 static void ip6ip6_dscp_ecn_decapsulate(struct ip6_tnl *t,
641 struct ipv6hdr *ipv6h,
644 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
645 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
647 if (INET_ECN_is_ce(ipv6_get_dsfield(ipv6h)))
648 IP6_ECN_set_ce(ipv6_hdr(skb));
651 static inline int ip6_tnl_rcv_ctl(struct ip6_tnl *t)
653 struct ip6_tnl_parm *p = &t->parms;
656 if (p->flags & IP6_TNL_F_CAP_RCV) {
657 struct net_device *ldev = NULL;
660 ldev = dev_get_by_index(&init_net, p->link);
662 if ((ipv6_addr_is_multicast(&p->laddr) ||
663 likely(ipv6_chk_addr(&init_net, &p->laddr, ldev, 0))) &&
664 likely(!ipv6_chk_addr(&init_net, &p->raddr, NULL, 0)))
674 * ip6_tnl_rcv - decapsulate IPv6 packet and retransmit it locally
675 * @skb: received socket buffer
676 * @protocol: ethernet protocol ID
677 * @dscp_ecn_decapsulate: the function to decapsulate DSCP code and ECN
682 static int ip6_tnl_rcv(struct sk_buff *skb, __u16 protocol,
684 void (*dscp_ecn_decapsulate)(struct ip6_tnl *t,
685 struct ipv6hdr *ipv6h,
686 struct sk_buff *skb))
689 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
691 read_lock(&ip6_tnl_lock);
693 if ((t = ip6_tnl_lookup(&ipv6h->saddr, &ipv6h->daddr)) != NULL) {
694 if (t->parms.proto != ipproto && t->parms.proto != 0) {
695 read_unlock(&ip6_tnl_lock);
699 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
700 read_unlock(&ip6_tnl_lock);
704 if (!ip6_tnl_rcv_ctl(t)) {
705 t->stat.rx_dropped++;
706 read_unlock(&ip6_tnl_lock);
710 skb->mac_header = skb->network_header;
711 skb_reset_network_header(skb);
712 skb->protocol = htons(protocol);
713 skb->pkt_type = PACKET_HOST;
714 memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
716 dst_release(skb->dst);
720 dscp_ecn_decapsulate(t, ipv6h, skb);
722 t->stat.rx_packets++;
723 t->stat.rx_bytes += skb->len;
725 read_unlock(&ip6_tnl_lock);
728 read_unlock(&ip6_tnl_lock);
736 static int ip4ip6_rcv(struct sk_buff *skb)
738 return ip6_tnl_rcv(skb, ETH_P_IP, IPPROTO_IPIP,
739 ip4ip6_dscp_ecn_decapsulate);
742 static int ip6ip6_rcv(struct sk_buff *skb)
744 return ip6_tnl_rcv(skb, ETH_P_IPV6, IPPROTO_IPV6,
745 ip6ip6_dscp_ecn_decapsulate);
748 struct ipv6_tel_txoption {
749 struct ipv6_txoptions ops;
753 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
755 memset(opt, 0, sizeof(struct ipv6_tel_txoption));
757 opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
759 opt->dst_opt[4] = encap_limit;
760 opt->dst_opt[5] = IPV6_TLV_PADN;
763 opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
764 opt->ops.opt_nflen = 8;
768 * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
769 * @t: the outgoing tunnel device
770 * @hdr: IPv6 header from the incoming packet
773 * Avoid trivial tunneling loop by checking that tunnel exit-point
774 * doesn't match source of incoming packet.
782 ip6_tnl_addr_conflict(struct ip6_tnl *t, struct ipv6hdr *hdr)
784 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
787 static inline int ip6_tnl_xmit_ctl(struct ip6_tnl *t)
789 struct ip6_tnl_parm *p = &t->parms;
792 if (p->flags & IP6_TNL_F_CAP_XMIT) {
793 struct net_device *ldev = NULL;
796 ldev = dev_get_by_index(&init_net, p->link);
798 if (unlikely(!ipv6_chk_addr(&init_net, &p->laddr, ldev, 0)))
800 "%s xmit: Local address not yet configured!\n",
802 else if (!ipv6_addr_is_multicast(&p->raddr) &&
803 unlikely(ipv6_chk_addr(&init_net, &p->raddr, NULL, 0)))
805 "%s xmit: Routing loop! "
806 "Remote address found on this node!\n",
816 * ip6_tnl_xmit2 - encapsulate packet and send
817 * @skb: the outgoing socket buffer
818 * @dev: the outgoing tunnel device
819 * @dsfield: dscp code for outer header
820 * @fl: flow of tunneled packet
821 * @encap_limit: encapsulation limit
822 * @pmtu: Path MTU is stored if packet is too big
825 * Build new header and do some sanity checks on the packet before sending
831 * %-EMSGSIZE message too big. return mtu in this case.
834 static int ip6_tnl_xmit2(struct sk_buff *skb,
835 struct net_device *dev,
841 struct ip6_tnl *t = netdev_priv(dev);
842 struct net_device_stats *stats = &t->stat;
843 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
844 struct ipv6_tel_txoption opt;
845 struct dst_entry *dst;
846 struct net_device *tdev;
848 unsigned int max_headroom = sizeof(struct ipv6hdr);
853 if ((dst = ip6_tnl_dst_check(t)) != NULL)
856 dst = ip6_route_output(&init_net, NULL, fl);
858 if (dst->error || xfrm_lookup(&dst, fl, NULL, 0) < 0)
859 goto tx_err_link_failure;
868 "%s: Local routing loop detected!\n",
870 goto tx_err_dst_release;
872 mtu = dst_mtu(dst) - sizeof (*ipv6h);
873 if (encap_limit >= 0) {
877 if (mtu < IPV6_MIN_MTU)
880 skb->dst->ops->update_pmtu(skb->dst, mtu);
881 if (skb->len > mtu) {
884 goto tx_err_dst_release;
888 * Okay, now see if we can stuff it in the buffer as-is.
890 max_headroom += LL_RESERVED_SPACE(tdev);
892 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
893 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
894 struct sk_buff *new_skb;
896 if (!(new_skb = skb_realloc_headroom(skb, max_headroom)))
897 goto tx_err_dst_release;
900 skb_set_owner_w(new_skb, skb->sk);
904 dst_release(skb->dst);
905 skb->dst = dst_clone(dst);
907 skb->transport_header = skb->network_header;
910 if (encap_limit >= 0) {
911 init_tel_txopt(&opt, encap_limit);
912 ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
914 skb_push(skb, sizeof(struct ipv6hdr));
915 skb_reset_network_header(skb);
916 ipv6h = ipv6_hdr(skb);
917 *(__be32*)ipv6h = fl->fl6_flowlabel | htonl(0x60000000);
918 dsfield = INET_ECN_encapsulate(0, dsfield);
919 ipv6_change_dsfield(ipv6h, ~INET_ECN_MASK, dsfield);
920 ipv6h->hop_limit = t->parms.hop_limit;
921 ipv6h->nexthdr = proto;
922 ipv6_addr_copy(&ipv6h->saddr, &fl->fl6_src);
923 ipv6_addr_copy(&ipv6h->daddr, &fl->fl6_dst);
926 err = ip6_local_out(skb);
928 if (net_xmit_eval(err) == 0) {
929 stats->tx_bytes += pkt_len;
933 stats->tx_aborted_errors++;
935 ip6_tnl_dst_store(t, dst);
938 stats->tx_carrier_errors++;
939 dst_link_failure(skb);
946 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
948 struct ip6_tnl *t = netdev_priv(dev);
949 struct iphdr *iph = ip_hdr(skb);
950 int encap_limit = -1;
956 if ((t->parms.proto != IPPROTO_IPIP && t->parms.proto != 0) ||
957 !ip6_tnl_xmit_ctl(t))
960 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
961 encap_limit = t->parms.encap_limit;
963 memcpy(&fl, &t->fl, sizeof (fl));
964 fl.proto = IPPROTO_IPIP;
966 dsfield = ipv4_get_dsfield(iph);
968 if ((t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS))
969 fl.fl6_flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
972 err = ip6_tnl_xmit2(skb, dev, dsfield, &fl, encap_limit, &mtu);
974 /* XXX: send ICMP error even if DF is not set. */
975 if (err == -EMSGSIZE)
976 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
985 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
987 struct ip6_tnl *t = netdev_priv(dev);
988 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
989 int encap_limit = -1;
996 if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) ||
997 !ip6_tnl_xmit_ctl(t) || ip6_tnl_addr_conflict(t, ipv6h))
1000 offset = parse_tlv_tnl_enc_lim(skb, skb_network_header(skb));
1002 struct ipv6_tlv_tnl_enc_lim *tel;
1003 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
1004 if (tel->encap_limit == 0) {
1005 icmpv6_send(skb, ICMPV6_PARAMPROB,
1006 ICMPV6_HDR_FIELD, offset + 2, skb->dev);
1009 encap_limit = tel->encap_limit - 1;
1010 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1011 encap_limit = t->parms.encap_limit;
1013 memcpy(&fl, &t->fl, sizeof (fl));
1014 fl.proto = IPPROTO_IPV6;
1016 dsfield = ipv6_get_dsfield(ipv6h);
1017 if ((t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS))
1018 fl.fl6_flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
1019 if ((t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL))
1020 fl.fl6_flowlabel |= (*(__be32 *) ipv6h & IPV6_FLOWLABEL_MASK);
1022 err = ip6_tnl_xmit2(skb, dev, dsfield, &fl, encap_limit, &mtu);
1024 if (err == -EMSGSIZE)
1025 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, dev);
1033 ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1035 struct ip6_tnl *t = netdev_priv(dev);
1036 struct net_device_stats *stats = &t->stat;
1039 if (t->recursion++) {
1040 t->stat.collisions++;
1044 switch (skb->protocol) {
1045 case __constant_htons(ETH_P_IP):
1046 ret = ip4ip6_tnl_xmit(skb, dev);
1048 case __constant_htons(ETH_P_IPV6):
1049 ret = ip6ip6_tnl_xmit(skb, dev);
1063 stats->tx_dropped++;
1069 static void ip6_tnl_set_cap(struct ip6_tnl *t)
1071 struct ip6_tnl_parm *p = &t->parms;
1072 int ltype = ipv6_addr_type(&p->laddr);
1073 int rtype = ipv6_addr_type(&p->raddr);
1075 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV);
1077 if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
1078 rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
1079 !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
1080 (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
1081 if (ltype&IPV6_ADDR_UNICAST)
1082 p->flags |= IP6_TNL_F_CAP_XMIT;
1083 if (rtype&IPV6_ADDR_UNICAST)
1084 p->flags |= IP6_TNL_F_CAP_RCV;
1088 static void ip6_tnl_link_config(struct ip6_tnl *t)
1090 struct net_device *dev = t->dev;
1091 struct ip6_tnl_parm *p = &t->parms;
1092 struct flowi *fl = &t->fl;
1094 memcpy(&dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1095 memcpy(&dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1097 /* Set up flowi template */
1098 ipv6_addr_copy(&fl->fl6_src, &p->laddr);
1099 ipv6_addr_copy(&fl->fl6_dst, &p->raddr);
1101 fl->fl6_flowlabel = 0;
1103 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1104 fl->fl6_flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1105 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1106 fl->fl6_flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1110 if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1111 dev->flags |= IFF_POINTOPOINT;
1113 dev->flags &= ~IFF_POINTOPOINT;
1115 dev->iflink = p->link;
1117 if (p->flags & IP6_TNL_F_CAP_XMIT) {
1118 int strict = (ipv6_addr_type(&p->raddr) &
1119 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1121 struct rt6_info *rt = rt6_lookup(&init_net, &p->raddr, &p->laddr,
1128 dev->hard_header_len = rt->rt6i_dev->hard_header_len +
1129 sizeof (struct ipv6hdr);
1131 dev->mtu = rt->rt6i_dev->mtu - sizeof (struct ipv6hdr);
1133 if (dev->mtu < IPV6_MIN_MTU)
1134 dev->mtu = IPV6_MIN_MTU;
1136 dst_release(&rt->u.dst);
1141 * ip6_tnl_change - update the tunnel parameters
1142 * @t: tunnel to be changed
1143 * @p: tunnel configuration parameters
1144 * @active: != 0 if tunnel is ready for use
1147 * ip6_tnl_change() updates the tunnel parameters
1151 ip6_tnl_change(struct ip6_tnl *t, struct ip6_tnl_parm *p)
1153 ipv6_addr_copy(&t->parms.laddr, &p->laddr);
1154 ipv6_addr_copy(&t->parms.raddr, &p->raddr);
1155 t->parms.flags = p->flags;
1156 t->parms.hop_limit = p->hop_limit;
1157 t->parms.encap_limit = p->encap_limit;
1158 t->parms.flowinfo = p->flowinfo;
1159 t->parms.link = p->link;
1160 t->parms.proto = p->proto;
1161 ip6_tnl_dst_reset(t);
1162 ip6_tnl_link_config(t);
1167 * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1168 * @dev: virtual device associated with tunnel
1169 * @ifr: parameters passed from userspace
1170 * @cmd: command to be performed
1173 * ip6_tnl_ioctl() is used for managing IPv6 tunnels
1176 * The possible commands are the following:
1177 * %SIOCGETTUNNEL: get tunnel parameters for device
1178 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1179 * %SIOCCHGTUNNEL: change tunnel parameters to those given
1180 * %SIOCDELTUNNEL: delete tunnel
1182 * The fallback device "ip6tnl0", created during module
1183 * initialization, can be used for creating other tunnel devices.
1187 * %-EFAULT if unable to copy data to or from userspace,
1188 * %-EPERM if current process hasn't %CAP_NET_ADMIN set
1189 * %-EINVAL if passed tunnel parameters are invalid,
1190 * %-EEXIST if changing a tunnel's parameters would cause a conflict
1191 * %-ENODEV if attempting to change or delete a nonexisting device
1195 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1198 struct ip6_tnl_parm p;
1199 struct ip6_tnl *t = NULL;
1203 if (dev == ip6_fb_tnl_dev) {
1204 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p))) {
1208 t = ip6_tnl_locate(&p, 0);
1211 t = netdev_priv(dev);
1212 memcpy(&p, &t->parms, sizeof (p));
1213 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof (p))) {
1220 if (!capable(CAP_NET_ADMIN))
1223 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
1226 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1229 t = ip6_tnl_locate(&p, cmd == SIOCADDTUNNEL);
1230 if (dev != ip6_fb_tnl_dev && cmd == SIOCCHGTUNNEL) {
1232 if (t->dev != dev) {
1237 t = netdev_priv(dev);
1240 err = ip6_tnl_change(t, &p);
1242 netdev_state_change(dev);
1246 if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof (p)))
1250 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1254 if (!capable(CAP_NET_ADMIN))
1257 if (dev == ip6_fb_tnl_dev) {
1259 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
1262 if ((t = ip6_tnl_locate(&p, 0)) == NULL)
1265 if (t->dev == ip6_fb_tnl_dev)
1270 unregister_netdevice(dev);
1279 * ip6_tnl_get_stats - return the stats for tunnel device
1280 * @dev: virtual device associated with tunnel
1282 * Return: stats for device
1285 static struct net_device_stats *
1286 ip6_tnl_get_stats(struct net_device *dev)
1288 return &(((struct ip6_tnl *)netdev_priv(dev))->stat);
1292 * ip6_tnl_change_mtu - change mtu manually for tunnel device
1293 * @dev: virtual device associated with tunnel
1294 * @new_mtu: the new mtu
1298 * %-EINVAL if mtu too small
1302 ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1304 if (new_mtu < IPV6_MIN_MTU) {
1312 * ip6_tnl_dev_setup - setup virtual tunnel device
1313 * @dev: virtual device associated with tunnel
1316 * Initialize function pointers and device parameters
1319 static void ip6_tnl_dev_setup(struct net_device *dev)
1321 dev->uninit = ip6_tnl_dev_uninit;
1322 dev->destructor = free_netdev;
1323 dev->hard_start_xmit = ip6_tnl_xmit;
1324 dev->get_stats = ip6_tnl_get_stats;
1325 dev->do_ioctl = ip6_tnl_ioctl;
1326 dev->change_mtu = ip6_tnl_change_mtu;
1328 dev->type = ARPHRD_TUNNEL6;
1329 dev->hard_header_len = LL_MAX_HEADER + sizeof (struct ipv6hdr);
1330 dev->mtu = ETH_DATA_LEN - sizeof (struct ipv6hdr);
1331 dev->flags |= IFF_NOARP;
1332 dev->addr_len = sizeof(struct in6_addr);
1337 * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1338 * @dev: virtual device associated with tunnel
1342 ip6_tnl_dev_init_gen(struct net_device *dev)
1344 struct ip6_tnl *t = netdev_priv(dev);
1346 strcpy(t->parms.name, dev->name);
1350 * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1351 * @dev: virtual device associated with tunnel
1355 ip6_tnl_dev_init(struct net_device *dev)
1357 struct ip6_tnl *t = netdev_priv(dev);
1358 ip6_tnl_dev_init_gen(dev);
1359 ip6_tnl_link_config(t);
1364 * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1365 * @dev: fallback device
1371 ip6_fb_tnl_dev_init(struct net_device *dev)
1373 struct ip6_tnl *t = netdev_priv(dev);
1374 ip6_tnl_dev_init_gen(dev);
1375 t->parms.proto = IPPROTO_IPV6;
1381 static struct xfrm6_tunnel ip4ip6_handler = {
1382 .handler = ip4ip6_rcv,
1383 .err_handler = ip4ip6_err,
1387 static struct xfrm6_tunnel ip6ip6_handler = {
1388 .handler = ip6ip6_rcv,
1389 .err_handler = ip6ip6_err,
1393 static int ip6_tnl_init_net(struct net *net)
1396 struct ip6_tnl_net *ip6n;
1399 ip6n = kmalloc(sizeof(struct ip6_tnl_net), GFP_KERNEL);
1403 err = net_assign_generic(net, ip6_tnl_net_id, ip6n);
1415 static void ip6_tnl_exit_net(struct net *net)
1417 struct ip6_tnl_net *ip6n;
1419 ip6n = net_generic(net, ip6_tnl_net_id);
1423 static struct pernet_operations ip6_tnl_net_ops = {
1424 .init = ip6_tnl_init_net,
1425 .exit = ip6_tnl_exit_net,
1429 * ip6_tunnel_init - register protocol and reserve needed resources
1431 * Return: 0 on success
1434 static int __init ip6_tunnel_init(void)
1438 if (xfrm6_tunnel_register(&ip4ip6_handler, AF_INET)) {
1439 printk(KERN_ERR "ip6_tunnel init: can't register ip4ip6\n");
1444 if (xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6)) {
1445 printk(KERN_ERR "ip6_tunnel init: can't register ip6ip6\n");
1449 ip6_fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
1452 if (!ip6_fb_tnl_dev) {
1456 ip6_fb_tnl_dev->init = ip6_fb_tnl_dev_init;
1458 if ((err = register_netdev(ip6_fb_tnl_dev))) {
1459 free_netdev(ip6_fb_tnl_dev);
1463 err = register_pernet_gen_device(&ip6_tnl_net_id, &ip6_tnl_net_ops);
1468 unregister_netdevice(ip6_fb_tnl_dev);
1470 xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
1472 xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
1477 static void __exit ip6_tnl_destroy_tunnels(void)
1482 for (h = 0; h < HASH_SIZE; h++) {
1483 while ((t = tnls_r_l[h]) != NULL)
1484 unregister_netdevice(t->dev);
1488 unregister_netdevice(t->dev);
1492 * ip6_tunnel_cleanup - free resources and unregister protocol
1495 static void __exit ip6_tunnel_cleanup(void)
1497 if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
1498 printk(KERN_INFO "ip6_tunnel close: can't deregister ip4ip6\n");
1500 if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
1501 printk(KERN_INFO "ip6_tunnel close: can't deregister ip6ip6\n");
1504 ip6_tnl_destroy_tunnels();
1507 unregister_pernet_gen_device(ip6_tnl_net_id, &ip6_tnl_net_ops);
1510 module_init(ip6_tunnel_init);
1511 module_exit(ip6_tunnel_cleanup);