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_tnl_dev_init(struct net_device *dev);
78 static void ip6_tnl_dev_setup(struct net_device *dev);
80 static int ip6_tnl_net_id __read_mostly;
82 /* the IPv6 tunnel fallback device */
83 struct net_device *fb_tnl_dev;
84 /* lists for storing tunnels in use */
85 struct ip6_tnl *tnls_r_l[HASH_SIZE];
86 struct ip6_tnl *tnls_wc[1];
87 struct ip6_tnl **tnls[2];
91 * Locking : hash tables are protected by RCU and a spinlock
93 static DEFINE_SPINLOCK(ip6_tnl_lock);
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 h0 = HASH(remote);
141 unsigned 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 **
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 **tp = ip6_tnl_bucket(ip6n, &t->parms);
194 spin_lock_bh(&ip6_tnl_lock);
196 rcu_assign_pointer(*tp, t);
197 spin_unlock_bh(&ip6_tnl_lock);
201 * ip6_tnl_unlink - remove tunnel from hash table
202 * @t: tunnel to be removed
206 ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
210 for (tp = ip6_tnl_bucket(ip6n, &t->parms); *tp; tp = &(*tp)->next) {
212 spin_lock_bh(&ip6_tnl_lock);
214 spin_unlock_bh(&ip6_tnl_lock);
221 * ip6_tnl_create() - create a new tunnel
222 * @p: tunnel parameters
223 * @pt: pointer to new tunnel
226 * Create tunnel matching given parameters.
229 * created tunnel or NULL
232 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct ip6_tnl_parm *p)
234 struct net_device *dev;
238 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
241 strlcpy(name, p->name, IFNAMSIZ);
243 sprintf(name, "ip6tnl%%d");
245 dev = alloc_netdev(sizeof (*t), name, ip6_tnl_dev_setup);
249 dev_net_set(dev, net);
251 if (strchr(name, '%')) {
252 if (dev_alloc_name(dev, name) < 0)
256 t = netdev_priv(dev);
258 ip6_tnl_dev_init(dev);
260 if ((err = register_netdevice(dev)) < 0)
264 ip6_tnl_link(ip6n, t);
274 * ip6_tnl_locate - find or create tunnel matching given parameters
275 * @p: tunnel parameters
276 * @create: != 0 if allowed to create new tunnel if no match found
279 * ip6_tnl_locate() first tries to locate an existing tunnel
280 * based on @parms. If this is unsuccessful, but @create is set a new
281 * tunnel device is created and registered for use.
284 * matching tunnel or NULL
287 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
288 struct ip6_tnl_parm *p, int create)
290 struct in6_addr *remote = &p->raddr;
291 struct in6_addr *local = &p->laddr;
293 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
295 for (t = *ip6_tnl_bucket(ip6n, p); t; t = t->next) {
296 if (ipv6_addr_equal(local, &t->parms.laddr) &&
297 ipv6_addr_equal(remote, &t->parms.raddr))
302 return ip6_tnl_create(net, p);
306 * ip6_tnl_dev_uninit - tunnel device uninitializer
307 * @dev: the device to be destroyed
310 * ip6_tnl_dev_uninit() removes tunnel from its list
314 ip6_tnl_dev_uninit(struct net_device *dev)
316 struct ip6_tnl *t = netdev_priv(dev);
317 struct net *net = dev_net(dev);
318 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
320 if (dev == ip6n->fb_tnl_dev) {
321 spin_lock_bh(&ip6_tnl_lock);
322 ip6n->tnls_wc[0] = NULL;
323 spin_unlock_bh(&ip6_tnl_lock);
325 ip6_tnl_unlink(ip6n, t);
327 ip6_tnl_dst_reset(t);
332 * parse_tvl_tnl_enc_lim - handle encapsulation limit option
333 * @skb: received socket buffer
336 * 0 if none was found,
337 * else index to encapsulation limit
341 parse_tlv_tnl_enc_lim(struct sk_buff *skb, __u8 * raw)
343 struct ipv6hdr *ipv6h = (struct ipv6hdr *) raw;
344 __u8 nexthdr = ipv6h->nexthdr;
345 __u16 off = sizeof (*ipv6h);
347 while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
349 struct ipv6_opt_hdr *hdr;
350 if (raw + off + sizeof (*hdr) > skb->data &&
351 !pskb_may_pull(skb, raw - skb->data + off + sizeof (*hdr)))
354 hdr = (struct ipv6_opt_hdr *) (raw + off);
355 if (nexthdr == NEXTHDR_FRAGMENT) {
356 struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
357 if (frag_hdr->frag_off)
360 } else if (nexthdr == NEXTHDR_AUTH) {
361 optlen = (hdr->hdrlen + 2) << 2;
363 optlen = ipv6_optlen(hdr);
365 if (nexthdr == NEXTHDR_DEST) {
368 struct ipv6_tlv_tnl_enc_lim *tel;
370 /* No more room for encapsulation limit */
371 if (i + sizeof (*tel) > off + optlen)
374 tel = (struct ipv6_tlv_tnl_enc_lim *) &raw[i];
375 /* return index of option if found and valid */
376 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
379 /* else jump to next option */
381 i += tel->length + 2;
386 nexthdr = hdr->nexthdr;
393 * ip6_tnl_err - tunnel error handler
396 * ip6_tnl_err() should handle errors in the tunnel according
397 * to the specifications in RFC 2473.
401 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
402 u8 *type, u8 *code, int *msg, __u32 *info, int offset)
404 struct ipv6hdr *ipv6h = (struct ipv6hdr *) skb->data;
407 u8 rel_type = ICMPV6_DEST_UNREACH;
408 u8 rel_code = ICMPV6_ADDR_UNREACH;
413 /* If the packet doesn't contain the original IPv6 header we are
414 in trouble since we might need the source address for further
415 processing of the error. */
418 if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr,
419 &ipv6h->saddr)) == NULL)
422 if (t->parms.proto != ipproto && t->parms.proto != 0)
429 struct ipv6_tlv_tnl_enc_lim *tel;
431 case ICMPV6_DEST_UNREACH:
434 "%s: Path to destination invalid "
435 "or inactive!\n", t->parms.name);
438 case ICMPV6_TIME_EXCEED:
439 if ((*code) == ICMPV6_EXC_HOPLIMIT) {
442 "%s: Too small hop limit or "
443 "routing loop in tunnel!\n",
448 case ICMPV6_PARAMPROB:
450 if ((*code) == ICMPV6_HDR_FIELD)
451 teli = parse_tlv_tnl_enc_lim(skb, skb->data);
453 if (teli && teli == *info - 2) {
454 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
455 if (tel->encap_limit == 0) {
458 "%s: Too small encapsulation "
459 "limit or routing loop in "
460 "tunnel!\n", t->parms.name);
463 } else if (net_ratelimit()) {
465 "%s: Recipient unable to parse tunneled "
466 "packet!\n ", t->parms.name);
469 case ICMPV6_PKT_TOOBIG:
470 mtu = *info - offset;
471 if (mtu < IPV6_MIN_MTU)
475 if ((len = sizeof (*ipv6h) + ntohs(ipv6h->payload_len)) > mtu) {
476 rel_type = ICMPV6_PKT_TOOBIG;
495 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
496 u8 type, u8 code, int offset, __be32 info)
501 __u32 rel_info = ntohl(info);
503 struct sk_buff *skb2;
508 err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
509 &rel_msg, &rel_info, offset);
517 case ICMPV6_DEST_UNREACH:
518 if (rel_code != ICMPV6_ADDR_UNREACH)
520 rel_type = ICMP_DEST_UNREACH;
521 rel_code = ICMP_HOST_UNREACH;
523 case ICMPV6_PKT_TOOBIG:
526 rel_type = ICMP_DEST_UNREACH;
527 rel_code = ICMP_FRAG_NEEDED;
533 if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
536 skb2 = skb_clone(skb, GFP_ATOMIC);
542 skb_pull(skb2, offset);
543 skb_reset_network_header(skb2);
546 /* Try to guess incoming interface */
547 memset(&fl, 0, sizeof(fl));
548 fl.fl4_dst = eiph->saddr;
549 fl.fl4_tos = RT_TOS(eiph->tos);
550 fl.proto = IPPROTO_IPIP;
551 if (ip_route_output_key(dev_net(skb->dev), &rt, &fl))
554 skb2->dev = rt->u.dst.dev;
556 /* route "incoming" packet */
557 if (rt->rt_flags & RTCF_LOCAL) {
560 fl.fl4_dst = eiph->daddr;
561 fl.fl4_src = eiph->saddr;
562 fl.fl4_tos = eiph->tos;
563 if (ip_route_output_key(dev_net(skb->dev), &rt, &fl) ||
564 rt->u.dst.dev->type != ARPHRD_TUNNEL) {
568 skb_dst_set(skb2, (struct dst_entry *)rt);
571 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
573 skb_dst(skb2)->dev->type != ARPHRD_TUNNEL)
577 /* change mtu on this route */
578 if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
579 if (rel_info > dst_mtu(skb_dst(skb2)))
582 skb_dst(skb2)->ops->update_pmtu(skb_dst(skb2), rel_info);
585 icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
593 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
594 u8 type, u8 code, int offset, __be32 info)
599 __u32 rel_info = ntohl(info);
602 err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
603 &rel_msg, &rel_info, offset);
607 if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
609 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
615 skb_pull(skb2, offset);
616 skb_reset_network_header(skb2);
618 /* Try to guess incoming interface */
619 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
622 if (rt && rt->rt6i_dev)
623 skb2->dev = rt->rt6i_dev;
625 icmpv6_send(skb2, rel_type, rel_code, rel_info);
628 dst_release(&rt->u.dst);
636 static void ip4ip6_dscp_ecn_decapsulate(struct ip6_tnl *t,
637 struct ipv6hdr *ipv6h,
640 __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
642 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
643 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
645 if (INET_ECN_is_ce(dsfield))
646 IP_ECN_set_ce(ip_hdr(skb));
649 static void ip6ip6_dscp_ecn_decapsulate(struct ip6_tnl *t,
650 struct ipv6hdr *ipv6h,
653 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
654 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
656 if (INET_ECN_is_ce(ipv6_get_dsfield(ipv6h)))
657 IP6_ECN_set_ce(ipv6_hdr(skb));
660 /* called with rcu_read_lock() */
661 static inline int ip6_tnl_rcv_ctl(struct ip6_tnl *t)
663 struct ip6_tnl_parm *p = &t->parms;
665 struct net *net = dev_net(t->dev);
667 if (p->flags & IP6_TNL_F_CAP_RCV) {
668 struct net_device *ldev = NULL;
671 ldev = dev_get_by_index_rcu(net, p->link);
673 if ((ipv6_addr_is_multicast(&p->laddr) ||
674 likely(ipv6_chk_addr(net, &p->laddr, ldev, 0))) &&
675 likely(!ipv6_chk_addr(net, &p->raddr, NULL, 0)))
683 * ip6_tnl_rcv - decapsulate IPv6 packet and retransmit it locally
684 * @skb: received socket buffer
685 * @protocol: ethernet protocol ID
686 * @dscp_ecn_decapsulate: the function to decapsulate DSCP code and ECN
691 static int ip6_tnl_rcv(struct sk_buff *skb, __u16 protocol,
693 void (*dscp_ecn_decapsulate)(struct ip6_tnl *t,
694 struct ipv6hdr *ipv6h,
695 struct sk_buff *skb))
698 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
702 if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr,
703 &ipv6h->daddr)) != NULL) {
704 if (t->parms.proto != ipproto && t->parms.proto != 0) {
709 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
714 if (!ip6_tnl_rcv_ctl(t)) {
715 t->dev->stats.rx_dropped++;
720 skb->mac_header = skb->network_header;
721 skb_reset_network_header(skb);
722 skb->protocol = htons(protocol);
723 skb->pkt_type = PACKET_HOST;
724 memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
729 dscp_ecn_decapsulate(t, ipv6h, skb);
731 t->dev->stats.rx_packets++;
732 t->dev->stats.rx_bytes += skb->len;
745 static int ip4ip6_rcv(struct sk_buff *skb)
747 return ip6_tnl_rcv(skb, ETH_P_IP, IPPROTO_IPIP,
748 ip4ip6_dscp_ecn_decapsulate);
751 static int ip6ip6_rcv(struct sk_buff *skb)
753 return ip6_tnl_rcv(skb, ETH_P_IPV6, IPPROTO_IPV6,
754 ip6ip6_dscp_ecn_decapsulate);
757 struct ipv6_tel_txoption {
758 struct ipv6_txoptions ops;
762 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
764 memset(opt, 0, sizeof(struct ipv6_tel_txoption));
766 opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
768 opt->dst_opt[4] = encap_limit;
769 opt->dst_opt[5] = IPV6_TLV_PADN;
772 opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
773 opt->ops.opt_nflen = 8;
777 * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
778 * @t: the outgoing tunnel device
779 * @hdr: IPv6 header from the incoming packet
782 * Avoid trivial tunneling loop by checking that tunnel exit-point
783 * doesn't match source of incoming packet.
791 ip6_tnl_addr_conflict(struct ip6_tnl *t, struct ipv6hdr *hdr)
793 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
796 static inline int ip6_tnl_xmit_ctl(struct ip6_tnl *t)
798 struct ip6_tnl_parm *p = &t->parms;
800 struct net *net = dev_net(t->dev);
802 if (p->flags & IP6_TNL_F_CAP_XMIT) {
803 struct net_device *ldev = NULL;
807 ldev = dev_get_by_index_rcu(net, p->link);
809 if (unlikely(!ipv6_chk_addr(net, &p->laddr, ldev, 0)))
811 "%s xmit: Local address not yet configured!\n",
813 else if (!ipv6_addr_is_multicast(&p->raddr) &&
814 unlikely(ipv6_chk_addr(net, &p->raddr, NULL, 0)))
816 "%s xmit: Routing loop! "
817 "Remote address found on this node!\n",
826 * ip6_tnl_xmit2 - encapsulate packet and send
827 * @skb: the outgoing socket buffer
828 * @dev: the outgoing tunnel device
829 * @dsfield: dscp code for outer header
830 * @fl: flow of tunneled packet
831 * @encap_limit: encapsulation limit
832 * @pmtu: Path MTU is stored if packet is too big
835 * Build new header and do some sanity checks on the packet before sending
841 * %-EMSGSIZE message too big. return mtu in this case.
844 static int ip6_tnl_xmit2(struct sk_buff *skb,
845 struct net_device *dev,
851 struct net *net = dev_net(dev);
852 struct ip6_tnl *t = netdev_priv(dev);
853 struct net_device_stats *stats = &t->dev->stats;
854 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
855 struct ipv6_tel_txoption opt;
856 struct dst_entry *dst;
857 struct net_device *tdev;
859 unsigned int max_headroom = sizeof(struct ipv6hdr);
864 if ((dst = ip6_tnl_dst_check(t)) != NULL)
867 dst = ip6_route_output(net, NULL, fl);
869 if (dst->error || xfrm_lookup(net, &dst, fl, NULL, 0) < 0)
870 goto tx_err_link_failure;
879 "%s: Local routing loop detected!\n",
881 goto tx_err_dst_release;
883 mtu = dst_mtu(dst) - sizeof (*ipv6h);
884 if (encap_limit >= 0) {
888 if (mtu < IPV6_MIN_MTU)
891 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), mtu);
892 if (skb->len > mtu) {
895 goto tx_err_dst_release;
899 * Okay, now see if we can stuff it in the buffer as-is.
901 max_headroom += LL_RESERVED_SPACE(tdev);
903 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
904 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
905 struct sk_buff *new_skb;
907 if (!(new_skb = skb_realloc_headroom(skb, max_headroom)))
908 goto tx_err_dst_release;
911 skb_set_owner_w(new_skb, skb->sk);
916 skb_dst_set(skb, dst_clone(dst));
918 skb->transport_header = skb->network_header;
921 if (encap_limit >= 0) {
922 init_tel_txopt(&opt, encap_limit);
923 ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
925 skb_push(skb, sizeof(struct ipv6hdr));
926 skb_reset_network_header(skb);
927 ipv6h = ipv6_hdr(skb);
928 *(__be32*)ipv6h = fl->fl6_flowlabel | htonl(0x60000000);
929 dsfield = INET_ECN_encapsulate(0, dsfield);
930 ipv6_change_dsfield(ipv6h, ~INET_ECN_MASK, dsfield);
931 ipv6h->hop_limit = t->parms.hop_limit;
932 ipv6h->nexthdr = proto;
933 ipv6_addr_copy(&ipv6h->saddr, &fl->fl6_src);
934 ipv6_addr_copy(&ipv6h->daddr, &fl->fl6_dst);
937 err = ip6_local_out(skb);
939 if (net_xmit_eval(err) == 0) {
940 stats->tx_bytes += pkt_len;
944 stats->tx_aborted_errors++;
946 ip6_tnl_dst_store(t, dst);
949 stats->tx_carrier_errors++;
950 dst_link_failure(skb);
957 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
959 struct ip6_tnl *t = netdev_priv(dev);
960 struct iphdr *iph = ip_hdr(skb);
961 int encap_limit = -1;
967 if ((t->parms.proto != IPPROTO_IPIP && t->parms.proto != 0) ||
968 !ip6_tnl_xmit_ctl(t))
971 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
972 encap_limit = t->parms.encap_limit;
974 memcpy(&fl, &t->fl, sizeof (fl));
975 fl.proto = IPPROTO_IPIP;
977 dsfield = ipv4_get_dsfield(iph);
979 if ((t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS))
980 fl.fl6_flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
983 err = ip6_tnl_xmit2(skb, dev, dsfield, &fl, encap_limit, &mtu);
985 /* XXX: send ICMP error even if DF is not set. */
986 if (err == -EMSGSIZE)
987 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
996 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
998 struct ip6_tnl *t = netdev_priv(dev);
999 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
1000 int encap_limit = -1;
1007 if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) ||
1008 !ip6_tnl_xmit_ctl(t) || ip6_tnl_addr_conflict(t, ipv6h))
1011 offset = parse_tlv_tnl_enc_lim(skb, skb_network_header(skb));
1013 struct ipv6_tlv_tnl_enc_lim *tel;
1014 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
1015 if (tel->encap_limit == 0) {
1016 icmpv6_send(skb, ICMPV6_PARAMPROB,
1017 ICMPV6_HDR_FIELD, offset + 2);
1020 encap_limit = tel->encap_limit - 1;
1021 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1022 encap_limit = t->parms.encap_limit;
1024 memcpy(&fl, &t->fl, sizeof (fl));
1025 fl.proto = IPPROTO_IPV6;
1027 dsfield = ipv6_get_dsfield(ipv6h);
1028 if ((t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS))
1029 fl.fl6_flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
1030 if ((t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL))
1031 fl.fl6_flowlabel |= (*(__be32 *) ipv6h & IPV6_FLOWLABEL_MASK);
1033 err = ip6_tnl_xmit2(skb, dev, dsfield, &fl, encap_limit, &mtu);
1035 if (err == -EMSGSIZE)
1036 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1044 ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1046 struct ip6_tnl *t = netdev_priv(dev);
1047 struct net_device_stats *stats = &t->dev->stats;
1050 switch (skb->protocol) {
1051 case htons(ETH_P_IP):
1052 ret = ip4ip6_tnl_xmit(skb, dev);
1054 case htons(ETH_P_IPV6):
1055 ret = ip6ip6_tnl_xmit(skb, dev);
1064 return NETDEV_TX_OK;
1068 stats->tx_dropped++;
1070 return NETDEV_TX_OK;
1073 static void ip6_tnl_set_cap(struct ip6_tnl *t)
1075 struct ip6_tnl_parm *p = &t->parms;
1076 int ltype = ipv6_addr_type(&p->laddr);
1077 int rtype = ipv6_addr_type(&p->raddr);
1079 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV);
1081 if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
1082 rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
1083 !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
1084 (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
1085 if (ltype&IPV6_ADDR_UNICAST)
1086 p->flags |= IP6_TNL_F_CAP_XMIT;
1087 if (rtype&IPV6_ADDR_UNICAST)
1088 p->flags |= IP6_TNL_F_CAP_RCV;
1092 static void ip6_tnl_link_config(struct ip6_tnl *t)
1094 struct net_device *dev = t->dev;
1095 struct ip6_tnl_parm *p = &t->parms;
1096 struct flowi *fl = &t->fl;
1098 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1099 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1101 /* Set up flowi template */
1102 ipv6_addr_copy(&fl->fl6_src, &p->laddr);
1103 ipv6_addr_copy(&fl->fl6_dst, &p->raddr);
1105 fl->fl6_flowlabel = 0;
1107 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1108 fl->fl6_flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1109 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1110 fl->fl6_flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1114 if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1115 dev->flags |= IFF_POINTOPOINT;
1117 dev->flags &= ~IFF_POINTOPOINT;
1119 dev->iflink = p->link;
1121 if (p->flags & IP6_TNL_F_CAP_XMIT) {
1122 int strict = (ipv6_addr_type(&p->raddr) &
1123 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1125 struct rt6_info *rt = rt6_lookup(dev_net(dev),
1126 &p->raddr, &p->laddr,
1133 dev->hard_header_len = rt->rt6i_dev->hard_header_len +
1134 sizeof (struct ipv6hdr);
1136 dev->mtu = rt->rt6i_dev->mtu - sizeof (struct ipv6hdr);
1138 if (dev->mtu < IPV6_MIN_MTU)
1139 dev->mtu = IPV6_MIN_MTU;
1141 dst_release(&rt->u.dst);
1146 * ip6_tnl_change - update the tunnel parameters
1147 * @t: tunnel to be changed
1148 * @p: tunnel configuration parameters
1151 * ip6_tnl_change() updates the tunnel parameters
1155 ip6_tnl_change(struct ip6_tnl *t, struct ip6_tnl_parm *p)
1157 ipv6_addr_copy(&t->parms.laddr, &p->laddr);
1158 ipv6_addr_copy(&t->parms.raddr, &p->raddr);
1159 t->parms.flags = p->flags;
1160 t->parms.hop_limit = p->hop_limit;
1161 t->parms.encap_limit = p->encap_limit;
1162 t->parms.flowinfo = p->flowinfo;
1163 t->parms.link = p->link;
1164 t->parms.proto = p->proto;
1165 ip6_tnl_dst_reset(t);
1166 ip6_tnl_link_config(t);
1171 * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1172 * @dev: virtual device associated with tunnel
1173 * @ifr: parameters passed from userspace
1174 * @cmd: command to be performed
1177 * ip6_tnl_ioctl() is used for managing IPv6 tunnels
1180 * The possible commands are the following:
1181 * %SIOCGETTUNNEL: get tunnel parameters for device
1182 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1183 * %SIOCCHGTUNNEL: change tunnel parameters to those given
1184 * %SIOCDELTUNNEL: delete tunnel
1186 * The fallback device "ip6tnl0", created during module
1187 * initialization, can be used for creating other tunnel devices.
1191 * %-EFAULT if unable to copy data to or from userspace,
1192 * %-EPERM if current process hasn't %CAP_NET_ADMIN set
1193 * %-EINVAL if passed tunnel parameters are invalid,
1194 * %-EEXIST if changing a tunnel's parameters would cause a conflict
1195 * %-ENODEV if attempting to change or delete a nonexisting device
1199 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1202 struct ip6_tnl_parm p;
1203 struct ip6_tnl *t = NULL;
1204 struct net *net = dev_net(dev);
1205 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1209 if (dev == ip6n->fb_tnl_dev) {
1210 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p))) {
1214 t = ip6_tnl_locate(net, &p, 0);
1217 t = netdev_priv(dev);
1218 memcpy(&p, &t->parms, sizeof (p));
1219 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof (p))) {
1226 if (!capable(CAP_NET_ADMIN))
1229 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
1232 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1235 t = ip6_tnl_locate(net, &p, cmd == SIOCADDTUNNEL);
1236 if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) {
1238 if (t->dev != dev) {
1243 t = netdev_priv(dev);
1245 ip6_tnl_unlink(ip6n, t);
1246 err = ip6_tnl_change(t, &p);
1247 ip6_tnl_link(ip6n, t);
1248 netdev_state_change(dev);
1252 if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof (p)))
1256 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1260 if (!capable(CAP_NET_ADMIN))
1263 if (dev == ip6n->fb_tnl_dev) {
1265 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
1268 if ((t = ip6_tnl_locate(net, &p, 0)) == NULL)
1271 if (t->dev == ip6n->fb_tnl_dev)
1276 unregister_netdevice(dev);
1285 * ip6_tnl_change_mtu - change mtu manually for tunnel device
1286 * @dev: virtual device associated with tunnel
1287 * @new_mtu: the new mtu
1291 * %-EINVAL if mtu too small
1295 ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1297 if (new_mtu < IPV6_MIN_MTU) {
1305 static const struct net_device_ops ip6_tnl_netdev_ops = {
1306 .ndo_uninit = ip6_tnl_dev_uninit,
1307 .ndo_start_xmit = ip6_tnl_xmit,
1308 .ndo_do_ioctl = ip6_tnl_ioctl,
1309 .ndo_change_mtu = ip6_tnl_change_mtu,
1313 * ip6_tnl_dev_setup - setup virtual tunnel device
1314 * @dev: virtual device associated with tunnel
1317 * Initialize function pointers and device parameters
1320 static void ip6_tnl_dev_setup(struct net_device *dev)
1322 dev->netdev_ops = &ip6_tnl_netdev_ops;
1323 dev->destructor = free_netdev;
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
1352 static void ip6_tnl_dev_init(struct net_device *dev)
1354 struct ip6_tnl *t = netdev_priv(dev);
1355 ip6_tnl_dev_init_gen(dev);
1356 ip6_tnl_link_config(t);
1360 * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1361 * @dev: fallback device
1366 static void __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1368 struct ip6_tnl *t = netdev_priv(dev);
1369 struct net *net = dev_net(dev);
1370 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1372 ip6_tnl_dev_init_gen(dev);
1373 t->parms.proto = IPPROTO_IPV6;
1375 ip6n->tnls_wc[0] = t;
1378 static struct xfrm6_tunnel ip4ip6_handler = {
1379 .handler = ip4ip6_rcv,
1380 .err_handler = ip4ip6_err,
1384 static struct xfrm6_tunnel ip6ip6_handler = {
1385 .handler = ip6ip6_rcv,
1386 .err_handler = ip6ip6_err,
1390 static void __net_exit ip6_tnl_destroy_tunnels(struct ip6_tnl_net *ip6n)
1396 for (h = 0; h < HASH_SIZE; h++) {
1397 t = ip6n->tnls_r_l[h];
1399 unregister_netdevice_queue(t->dev, &list);
1404 t = ip6n->tnls_wc[0];
1405 unregister_netdevice_queue(t->dev, &list);
1406 unregister_netdevice_many(&list);
1409 static int __net_init ip6_tnl_init_net(struct net *net)
1411 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1414 ip6n->tnls[0] = ip6n->tnls_wc;
1415 ip6n->tnls[1] = ip6n->tnls_r_l;
1418 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
1421 if (!ip6n->fb_tnl_dev)
1423 dev_net_set(ip6n->fb_tnl_dev, net);
1425 ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
1427 err = register_netdev(ip6n->fb_tnl_dev);
1433 free_netdev(ip6n->fb_tnl_dev);
1438 static void __net_exit ip6_tnl_exit_net(struct net *net)
1440 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1443 ip6_tnl_destroy_tunnels(ip6n);
1447 static struct pernet_operations ip6_tnl_net_ops = {
1448 .init = ip6_tnl_init_net,
1449 .exit = ip6_tnl_exit_net,
1450 .id = &ip6_tnl_net_id,
1451 .size = sizeof(struct ip6_tnl_net),
1455 * ip6_tunnel_init - register protocol and reserve needed resources
1457 * Return: 0 on success
1460 static int __init ip6_tunnel_init(void)
1464 err = register_pernet_device(&ip6_tnl_net_ops);
1468 err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
1470 printk(KERN_ERR "ip6_tunnel init: can't register ip4ip6\n");
1474 err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
1476 printk(KERN_ERR "ip6_tunnel init: can't register ip6ip6\n");
1483 xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
1485 unregister_pernet_device(&ip6_tnl_net_ops);
1491 * ip6_tunnel_cleanup - free resources and unregister protocol
1494 static void __exit ip6_tunnel_cleanup(void)
1496 if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
1497 printk(KERN_INFO "ip6_tunnel close: can't deregister ip4ip6\n");
1499 if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
1500 printk(KERN_INFO "ip6_tunnel close: can't deregister ip6ip6\n");
1502 unregister_pernet_device(&ip6_tnl_net_ops);
1505 module_init(ip6_tunnel_init);
1506 module_exit(ip6_tunnel_cleanup);