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>
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", __FUNCTION__)
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 ip6ip6_fb_tnl_dev_init(struct net_device *dev);
78 static int ip6ip6_tnl_dev_init(struct net_device *dev);
79 static void ip6ip6_tnl_dev_setup(struct net_device *dev);
81 /* the IPv6 tunnel fallback device */
82 static struct net_device *ip6ip6_fb_tnl_dev;
85 /* lists for storing tunnels in use */
86 static struct ip6_tnl *tnls_r_l[HASH_SIZE];
87 static struct ip6_tnl *tnls_wc[1];
88 static struct ip6_tnl **tnls[2] = { tnls_wc, tnls_r_l };
90 /* lock for the tunnel lists */
91 static DEFINE_RWLOCK(ip6ip6_lock);
93 static inline struct dst_entry *ip6_tnl_dst_check(struct ip6_tnl *t)
95 struct dst_entry *dst = t->dst_cache;
97 if (dst && dst->obsolete &&
98 dst->ops->check(dst, t->dst_cookie) == NULL) {
107 static inline void ip6_tnl_dst_reset(struct ip6_tnl *t)
109 dst_release(t->dst_cache);
113 static inline void ip6_tnl_dst_store(struct ip6_tnl *t, struct dst_entry *dst)
115 struct rt6_info *rt = (struct rt6_info *) dst;
116 t->dst_cookie = rt->rt6i_node ? rt->rt6i_node->fn_sernum : 0;
117 dst_release(t->dst_cache);
122 * ip6ip6_tnl_lookup - fetch tunnel matching the end-point addresses
123 * @remote: the address of the tunnel exit-point
124 * @local: the address of the tunnel entry-point
127 * tunnel matching given end-points if found,
128 * else fallback tunnel if its device is up,
132 static struct ip6_tnl *
133 ip6ip6_tnl_lookup(struct in6_addr *remote, struct in6_addr *local)
135 unsigned h0 = HASH(remote);
136 unsigned h1 = HASH(local);
139 for (t = tnls_r_l[h0 ^ h1]; t; t = t->next) {
140 if (ipv6_addr_equal(local, &t->parms.laddr) &&
141 ipv6_addr_equal(remote, &t->parms.raddr) &&
142 (t->dev->flags & IFF_UP))
145 if ((t = tnls_wc[0]) != NULL && (t->dev->flags & IFF_UP))
152 * ip6ip6_bucket - get head of list matching given tunnel parameters
153 * @p: parameters containing tunnel end-points
156 * ip6ip6_bucket() returns the head of the list matching the
157 * &struct in6_addr entries laddr and raddr in @p.
159 * Return: head of IPv6 tunnel list
162 static struct ip6_tnl **
163 ip6ip6_bucket(struct ip6_tnl_parm *p)
165 struct in6_addr *remote = &p->raddr;
166 struct in6_addr *local = &p->laddr;
170 if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
172 h = HASH(remote) ^ HASH(local);
174 return &tnls[prio][h];
178 * ip6ip6_tnl_link - add tunnel to hash table
179 * @t: tunnel to be added
183 ip6ip6_tnl_link(struct ip6_tnl *t)
185 struct ip6_tnl **tp = ip6ip6_bucket(&t->parms);
188 write_lock_bh(&ip6ip6_lock);
190 write_unlock_bh(&ip6ip6_lock);
194 * ip6ip6_tnl_unlink - remove tunnel from hash table
195 * @t: tunnel to be removed
199 ip6ip6_tnl_unlink(struct ip6_tnl *t)
203 for (tp = ip6ip6_bucket(&t->parms); *tp; tp = &(*tp)->next) {
205 write_lock_bh(&ip6ip6_lock);
207 write_unlock_bh(&ip6ip6_lock);
214 * ip6_tnl_create() - create a new tunnel
215 * @p: tunnel parameters
216 * @pt: pointer to new tunnel
219 * Create tunnel matching given parameters.
222 * created tunnel or NULL
225 static struct ip6_tnl *ip6_tnl_create(struct ip6_tnl_parm *p)
227 struct net_device *dev;
233 strlcpy(name, p->name, IFNAMSIZ);
236 for (i = 1; i < IP6_TNL_MAX; i++) {
237 sprintf(name, "ip6tnl%d", i);
238 if (__dev_get_by_name(name) == NULL)
241 if (i == IP6_TNL_MAX)
244 dev = alloc_netdev(sizeof (*t), name, ip6ip6_tnl_dev_setup);
248 t = netdev_priv(dev);
249 dev->init = ip6ip6_tnl_dev_init;
252 if ((err = register_netdevice(dev)) < 0) {
264 * ip6ip6_tnl_locate - find or create tunnel matching given parameters
265 * @p: tunnel parameters
266 * @create: != 0 if allowed to create new tunnel if no match found
269 * ip6ip6_tnl_locate() first tries to locate an existing tunnel
270 * based on @parms. If this is unsuccessful, but @create is set a new
271 * tunnel device is created and registered for use.
274 * matching tunnel or NULL
277 static struct ip6_tnl *ip6ip6_tnl_locate(struct ip6_tnl_parm *p, int create)
279 struct in6_addr *remote = &p->raddr;
280 struct in6_addr *local = &p->laddr;
283 for (t = *ip6ip6_bucket(p); t; t = t->next) {
284 if (ipv6_addr_equal(local, &t->parms.laddr) &&
285 ipv6_addr_equal(remote, &t->parms.raddr))
290 return ip6_tnl_create(p);
294 * ip6ip6_tnl_dev_uninit - tunnel device uninitializer
295 * @dev: the device to be destroyed
298 * ip6ip6_tnl_dev_uninit() removes tunnel from its list
302 ip6ip6_tnl_dev_uninit(struct net_device *dev)
304 struct ip6_tnl *t = netdev_priv(dev);
306 if (dev == ip6ip6_fb_tnl_dev) {
307 write_lock_bh(&ip6ip6_lock);
309 write_unlock_bh(&ip6ip6_lock);
311 ip6ip6_tnl_unlink(t);
313 ip6_tnl_dst_reset(t);
318 * parse_tvl_tnl_enc_lim - handle encapsulation limit option
319 * @skb: received socket buffer
322 * 0 if none was found,
323 * else index to encapsulation limit
327 parse_tlv_tnl_enc_lim(struct sk_buff *skb, __u8 * raw)
329 struct ipv6hdr *ipv6h = (struct ipv6hdr *) raw;
330 __u8 nexthdr = ipv6h->nexthdr;
331 __u16 off = sizeof (*ipv6h);
333 while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
335 struct ipv6_opt_hdr *hdr;
336 if (raw + off + sizeof (*hdr) > skb->data &&
337 !pskb_may_pull(skb, raw - skb->data + off + sizeof (*hdr)))
340 hdr = (struct ipv6_opt_hdr *) (raw + off);
341 if (nexthdr == NEXTHDR_FRAGMENT) {
342 struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
343 if (frag_hdr->frag_off)
346 } else if (nexthdr == NEXTHDR_AUTH) {
347 optlen = (hdr->hdrlen + 2) << 2;
349 optlen = ipv6_optlen(hdr);
351 if (nexthdr == NEXTHDR_DEST) {
354 struct ipv6_tlv_tnl_enc_lim *tel;
356 /* No more room for encapsulation limit */
357 if (i + sizeof (*tel) > off + optlen)
360 tel = (struct ipv6_tlv_tnl_enc_lim *) &raw[i];
361 /* return index of option if found and valid */
362 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
365 /* else jump to next option */
367 i += tel->length + 2;
372 nexthdr = hdr->nexthdr;
379 * ip6_tnl_err - tunnel error handler
382 * ip6_tnl_err() should handle errors in the tunnel according
383 * to the specifications in RFC 2473.
387 ip6_tnl_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
388 int *type, int *code, int *msg, __be32 *info, int offset)
390 struct ipv6hdr *ipv6h = (struct ipv6hdr *) skb->data;
393 int rel_type = ICMPV6_DEST_UNREACH;
394 int rel_code = ICMPV6_ADDR_UNREACH;
399 /* If the packet doesn't contain the original IPv6 header we are
400 in trouble since we might need the source address for further
401 processing of the error. */
403 read_lock(&ip6ip6_lock);
404 if ((t = ip6ip6_tnl_lookup(&ipv6h->daddr, &ipv6h->saddr)) == NULL)
411 struct ipv6_tlv_tnl_enc_lim *tel;
413 case ICMPV6_DEST_UNREACH:
416 "%s: Path to destination invalid "
417 "or inactive!\n", t->parms.name);
420 case ICMPV6_TIME_EXCEED:
421 if ((*code) == ICMPV6_EXC_HOPLIMIT) {
424 "%s: Too small hop limit or "
425 "routing loop in tunnel!\n",
430 case ICMPV6_PARAMPROB:
432 if ((*code) == ICMPV6_HDR_FIELD)
433 teli = parse_tlv_tnl_enc_lim(skb, skb->data);
435 if (teli && teli == ntohl(*info) - 2) {
436 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
437 if (tel->encap_limit == 0) {
440 "%s: Too small encapsulation "
441 "limit or routing loop in "
442 "tunnel!\n", t->parms.name);
445 } else if (net_ratelimit()) {
447 "%s: Recipient unable to parse tunneled "
448 "packet!\n ", t->parms.name);
451 case ICMPV6_PKT_TOOBIG:
452 mtu = ntohl(*info) - offset;
453 if (mtu < IPV6_MIN_MTU)
457 if ((len = sizeof (*ipv6h) + ntohs(ipv6h->payload_len)) > mtu) {
458 rel_type = ICMPV6_PKT_TOOBIG;
472 read_unlock(&ip6ip6_lock);
477 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
478 int type, int code, int offset, __u32 info)
483 __u32 rel_info = info;
485 struct sk_buff *skb2;
490 err = ip6_tnl_err(skb, opt, &rel_type, &rel_code, &rel_msg, &rel_info,
499 case ICMPV6_DEST_UNREACH:
500 if (rel_code != ICMPV6_ADDR_UNREACH)
502 rel_type = ICMP_DEST_UNREACH;
503 rel_code = ICMP_HOST_UNREACH;
505 case ICMPV6_PKT_TOOBIG:
508 rel_type = ICMP_DEST_UNREACH;
509 rel_code = ICMP_FRAG_NEEDED;
515 if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
518 skb2 = skb_clone(skb, GFP_ATOMIC);
522 dst_release(skb2->dst);
524 skb_pull(skb2, offset);
525 skb2->nh.raw = skb2->data;
528 /* Try to guess incoming interface */
529 memset(&fl, 0, sizeof(fl));
530 fl.fl4_dst = eiph->saddr;
531 fl.fl4_tos = RT_TOS(eiph->tos);
532 fl.proto = IPPROTO_IPIP;
533 if (ip_route_output_key(&rt, &fl))
536 skb2->dev = rt->u.dst.dev;
538 /* route "incoming" packet */
539 if (rt->rt_flags & RTCF_LOCAL) {
542 fl.fl4_dst = eiph->daddr;
543 fl.fl4_src = eiph->saddr;
544 fl.fl4_tos = eiph->tos;
545 if (ip_route_output_key(&rt, &fl) ||
546 rt->u.dst.dev->type != ARPHRD_TUNNEL) {
552 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
554 skb2->dst->dev->type != ARPHRD_TUNNEL)
558 /* change mtu on this route */
559 if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
560 if (rel_info > dst_mtu(skb2->dst))
563 skb2->dst->ops->update_pmtu(skb2->dst, rel_info);
564 rel_info = htonl(rel_info);
567 icmp_send(skb2, rel_type, rel_code, rel_info);
575 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
576 int type, int code, int offset, __u32 info)
581 __u32 rel_info = info;
584 err = ip6_tnl_err(skb, opt, &rel_type, &rel_code, &rel_msg, &rel_info,
589 if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
591 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
596 dst_release(skb2->dst);
598 skb_pull(skb2, offset);
599 skb2->nh.raw = skb2->data;
601 /* Try to guess incoming interface */
602 rt = rt6_lookup(&skb2->nh.ipv6h->saddr, NULL, 0, 0);
604 if (rt && rt->rt6i_dev)
605 skb2->dev = rt->rt6i_dev;
607 icmpv6_send(skb2, rel_type, rel_code, rel_info, skb2->dev);
610 dst_release(&rt->u.dst);
618 static void ip4ip6_dscp_ecn_decapsulate(struct ip6_tnl *t,
619 struct ipv6hdr *ipv6h,
622 __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
624 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
625 ipv4_change_dsfield(skb->nh.iph, INET_ECN_MASK, dsfield);
627 if (INET_ECN_is_ce(dsfield))
628 IP_ECN_set_ce(skb->nh.iph);
631 static void ip6ip6_dscp_ecn_decapsulate(struct ip6_tnl *t,
632 struct ipv6hdr *ipv6h,
635 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
636 ipv6_copy_dscp(ipv6h, skb->nh.ipv6h);
638 if (INET_ECN_is_ce(ipv6_get_dsfield(ipv6h)))
639 IP6_ECN_set_ce(skb->nh.ipv6h);
642 static inline int ip6_tnl_rcv_ctl(struct ip6_tnl *t)
644 struct ip6_tnl_parm *p = &t->parms;
647 if (p->flags & IP6_TNL_F_CAP_RCV) {
648 struct net_device *ldev = NULL;
651 ldev = dev_get_by_index(p->link);
653 if ((ipv6_addr_is_multicast(&p->laddr) ||
654 likely(ipv6_chk_addr(&p->laddr, ldev, 0))) &&
655 likely(!ipv6_chk_addr(&p->raddr, NULL, 0)))
665 * ip6ip6_rcv - decapsulate IPv6 packet and retransmit it locally
666 * @skb: received socket buffer
667 * @protocol: ethernet protocol ID
668 * @dscp_ecn_decapsulate: the function to decapsulate DSCP code and ECN
673 static int ip6_tnl_rcv(struct sk_buff *skb, __u16 protocol,
674 void (*dscp_ecn_decapsulate)(struct ip6_tnl *t,
675 struct ipv6hdr *ipv6h,
676 struct sk_buff *skb))
678 struct ipv6hdr *ipv6h;
681 ipv6h = skb->nh.ipv6h;
683 read_lock(&ip6ip6_lock);
685 if ((t = ip6ip6_tnl_lookup(&ipv6h->saddr, &ipv6h->daddr)) != NULL) {
686 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
687 read_unlock(&ip6ip6_lock);
691 if (!ip6_tnl_rcv_ctl(t)) {
692 t->stat.rx_dropped++;
693 read_unlock(&ip6ip6_lock);
697 skb->mac.raw = skb->nh.raw;
698 skb->nh.raw = skb->data;
699 skb->protocol = htons(protocol);
700 skb->pkt_type = PACKET_HOST;
701 memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
703 dst_release(skb->dst);
707 dscp_ecn_decapsulate(t, ipv6h, skb);
709 t->stat.rx_packets++;
710 t->stat.rx_bytes += skb->len;
712 read_unlock(&ip6ip6_lock);
715 read_unlock(&ip6ip6_lock);
723 static int ip4ip6_rcv(struct sk_buff *skb)
725 return ip6_tnl_rcv(skb, ETH_P_IP, ip4ip6_dscp_ecn_decapsulate);
728 static int ip6ip6_rcv(struct sk_buff *skb)
730 return ip6_tnl_rcv(skb, ETH_P_IPV6, ip6ip6_dscp_ecn_decapsulate);
733 struct ipv6_tel_txoption {
734 struct ipv6_txoptions ops;
738 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
740 memset(opt, 0, sizeof(struct ipv6_tel_txoption));
742 opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
744 opt->dst_opt[4] = encap_limit;
745 opt->dst_opt[5] = IPV6_TLV_PADN;
748 opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
749 opt->ops.opt_nflen = 8;
753 * ip6ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
754 * @t: the outgoing tunnel device
755 * @hdr: IPv6 header from the incoming packet
758 * Avoid trivial tunneling loop by checking that tunnel exit-point
759 * doesn't match source of incoming packet.
767 ip6ip6_tnl_addr_conflict(struct ip6_tnl *t, struct ipv6hdr *hdr)
769 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
772 static inline int ip6_tnl_xmit_ctl(struct ip6_tnl *t)
774 struct ip6_tnl_parm *p = &t->parms;
777 if (p->flags & IP6_TNL_F_CAP_XMIT) {
778 struct net_device *ldev = NULL;
781 ldev = dev_get_by_index(p->link);
783 if (unlikely(!ipv6_chk_addr(&p->laddr, ldev, 0)))
785 "%s xmit: Local address not yet configured!\n",
787 else if (!ipv6_addr_is_multicast(&p->raddr) &&
788 unlikely(ipv6_chk_addr(&p->raddr, NULL, 0)))
790 "%s xmit: Routing loop! "
791 "Remote address found on this node!\n",
801 * ip6_tnl_xmit2 - encapsulate packet and send
802 * @skb: the outgoing socket buffer
803 * @dev: the outgoing tunnel device
804 * @dsfield: dscp code for outer header
805 * @fl: flow of tunneled packet
806 * @encap_limit: encapsulation limit
807 * @pmtu: Path MTU is stored if packet is too big
810 * Build new header and do some sanity checks on the packet before sending
816 * %-EMSGSIZE message too big. return mtu in this case.
819 static int ip6_tnl_xmit2(struct sk_buff *skb,
820 struct net_device *dev,
826 struct ip6_tnl *t = netdev_priv(dev);
827 struct net_device_stats *stats = &t->stat;
828 struct ipv6hdr *ipv6h = skb->nh.ipv6h;
829 struct ipv6_tel_txoption opt;
830 struct dst_entry *dst;
831 struct net_device *tdev;
833 int max_headroom = sizeof(struct ipv6hdr);
838 if ((dst = ip6_tnl_dst_check(t)) != NULL)
841 dst = ip6_route_output(NULL, fl);
843 if (dst->error || xfrm_lookup(&dst, fl, NULL, 0) < 0)
844 goto tx_err_link_failure;
853 "%s: Local routing loop detected!\n",
855 goto tx_err_dst_release;
857 mtu = dst_mtu(dst) - sizeof (*ipv6h);
858 if (encap_limit >= 0) {
862 if (mtu < IPV6_MIN_MTU)
865 skb->dst->ops->update_pmtu(skb->dst, mtu);
866 if (skb->len > mtu) {
869 goto tx_err_dst_release;
873 * Okay, now see if we can stuff it in the buffer as-is.
875 max_headroom += LL_RESERVED_SPACE(tdev);
877 if (skb_headroom(skb) < max_headroom ||
878 skb_cloned(skb) || skb_shared(skb)) {
879 struct sk_buff *new_skb;
881 if (!(new_skb = skb_realloc_headroom(skb, max_headroom)))
882 goto tx_err_dst_release;
885 skb_set_owner_w(new_skb, skb->sk);
889 dst_release(skb->dst);
890 skb->dst = dst_clone(dst);
892 skb->h.raw = skb->nh.raw;
895 if (encap_limit >= 0) {
896 init_tel_txopt(&opt, encap_limit);
897 ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
899 skb->nh.raw = skb_push(skb, sizeof(struct ipv6hdr));
900 ipv6h = skb->nh.ipv6h;
901 *(__be32*)ipv6h = fl->fl6_flowlabel | htonl(0x60000000);
902 dsfield = INET_ECN_encapsulate(0, dsfield);
903 ipv6_change_dsfield(ipv6h, ~INET_ECN_MASK, dsfield);
904 ipv6h->payload_len = htons(skb->len - sizeof(struct ipv6hdr));
905 ipv6h->hop_limit = t->parms.hop_limit;
906 ipv6h->nexthdr = proto;
907 ipv6_addr_copy(&ipv6h->saddr, &fl->fl6_src);
908 ipv6_addr_copy(&ipv6h->daddr, &fl->fl6_dst);
911 err = NF_HOOK(PF_INET6, NF_IP6_LOCAL_OUT, skb, NULL,
912 skb->dst->dev, dst_output);
914 if (net_xmit_eval(err) == 0) {
915 stats->tx_bytes += pkt_len;
919 stats->tx_aborted_errors++;
921 ip6_tnl_dst_store(t, dst);
924 stats->tx_carrier_errors++;
925 dst_link_failure(skb);
932 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
934 struct ip6_tnl *t = netdev_priv(dev);
935 struct iphdr *iph = skb->nh.iph;
936 int encap_limit = -1;
942 if (!ip6_tnl_xmit_ctl(t))
945 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
946 encap_limit = t->parms.encap_limit;
948 memcpy(&fl, &t->fl, sizeof (fl));
949 fl.proto = IPPROTO_IPIP;
951 dsfield = ipv4_get_dsfield(iph);
953 if ((t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS))
954 fl.fl6_flowlabel |= ntohl(((__u32)iph->tos << IPV6_TCLASS_SHIFT)
957 err = ip6_tnl_xmit2(skb, dev, dsfield, &fl, encap_limit, &mtu);
959 /* XXX: send ICMP error even if DF is not set. */
960 if (err == -EMSGSIZE)
961 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
970 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
972 struct ip6_tnl *t = netdev_priv(dev);
973 struct ipv6hdr *ipv6h = skb->nh.ipv6h;
974 int encap_limit = -1;
981 if (!ip6_tnl_xmit_ctl(t) || ip6ip6_tnl_addr_conflict(t, ipv6h))
984 if ((offset = parse_tlv_tnl_enc_lim(skb, skb->nh.raw)) > 0) {
985 struct ipv6_tlv_tnl_enc_lim *tel;
986 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->nh.raw[offset];
987 if (tel->encap_limit == 0) {
988 icmpv6_send(skb, ICMPV6_PARAMPROB,
989 ICMPV6_HDR_FIELD, offset + 2, skb->dev);
992 encap_limit = tel->encap_limit - 1;
993 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
994 encap_limit = t->parms.encap_limit;
996 memcpy(&fl, &t->fl, sizeof (fl));
997 fl.proto = IPPROTO_IPV6;
999 dsfield = ipv6_get_dsfield(ipv6h);
1000 if ((t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS))
1001 fl.fl6_flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
1002 if ((t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL))
1003 fl.fl6_flowlabel |= (*(__be32 *) ipv6h & IPV6_FLOWLABEL_MASK);
1005 err = ip6_tnl_xmit2(skb, dev, dsfield, &fl, encap_limit, &mtu);
1007 if (err == -EMSGSIZE)
1008 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, dev);
1016 ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1018 struct ip6_tnl *t = netdev_priv(dev);
1019 struct net_device_stats *stats = &t->stat;
1022 if (t->recursion++) {
1023 t->stat.collisions++;
1027 switch (skb->protocol) {
1028 case __constant_htons(ETH_P_IP):
1029 ret = ip4ip6_tnl_xmit(skb, dev);
1031 case __constant_htons(ETH_P_IPV6):
1032 ret = ip6ip6_tnl_xmit(skb, dev);
1046 stats->tx_dropped++;
1052 static void ip6_tnl_set_cap(struct ip6_tnl *t)
1054 struct ip6_tnl_parm *p = &t->parms;
1055 int ltype = ipv6_addr_type(&p->laddr);
1056 int rtype = ipv6_addr_type(&p->raddr);
1058 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV);
1060 if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
1061 rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
1062 !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
1063 (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
1064 if (ltype&IPV6_ADDR_UNICAST)
1065 p->flags |= IP6_TNL_F_CAP_XMIT;
1066 if (rtype&IPV6_ADDR_UNICAST)
1067 p->flags |= IP6_TNL_F_CAP_RCV;
1071 static void ip6ip6_tnl_link_config(struct ip6_tnl *t)
1073 struct net_device *dev = t->dev;
1074 struct ip6_tnl_parm *p = &t->parms;
1075 struct flowi *fl = &t->fl;
1077 memcpy(&dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1078 memcpy(&dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1080 /* Set up flowi template */
1081 ipv6_addr_copy(&fl->fl6_src, &p->laddr);
1082 ipv6_addr_copy(&fl->fl6_dst, &p->raddr);
1084 fl->fl6_flowlabel = 0;
1086 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1087 fl->fl6_flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1088 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1089 fl->fl6_flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1093 if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1094 dev->flags |= IFF_POINTOPOINT;
1096 dev->flags &= ~IFF_POINTOPOINT;
1098 dev->iflink = p->link;
1100 if (p->flags & IP6_TNL_F_CAP_XMIT) {
1101 int strict = (ipv6_addr_type(&p->raddr) &
1102 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1104 struct rt6_info *rt = rt6_lookup(&p->raddr, &p->laddr,
1111 dev->hard_header_len = rt->rt6i_dev->hard_header_len +
1112 sizeof (struct ipv6hdr);
1114 dev->mtu = rt->rt6i_dev->mtu - sizeof (struct ipv6hdr);
1116 if (dev->mtu < IPV6_MIN_MTU)
1117 dev->mtu = IPV6_MIN_MTU;
1119 dst_release(&rt->u.dst);
1124 * ip6ip6_tnl_change - update the tunnel parameters
1125 * @t: tunnel to be changed
1126 * @p: tunnel configuration parameters
1127 * @active: != 0 if tunnel is ready for use
1130 * ip6ip6_tnl_change() updates the tunnel parameters
1134 ip6ip6_tnl_change(struct ip6_tnl *t, struct ip6_tnl_parm *p)
1136 ipv6_addr_copy(&t->parms.laddr, &p->laddr);
1137 ipv6_addr_copy(&t->parms.raddr, &p->raddr);
1138 t->parms.flags = p->flags;
1139 t->parms.hop_limit = p->hop_limit;
1140 t->parms.encap_limit = p->encap_limit;
1141 t->parms.flowinfo = p->flowinfo;
1142 t->parms.link = p->link;
1143 ip6_tnl_dst_reset(t);
1144 ip6ip6_tnl_link_config(t);
1149 * ip6ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1150 * @dev: virtual device associated with tunnel
1151 * @ifr: parameters passed from userspace
1152 * @cmd: command to be performed
1155 * ip6ip6_tnl_ioctl() is used for managing IPv6 tunnels
1158 * The possible commands are the following:
1159 * %SIOCGETTUNNEL: get tunnel parameters for device
1160 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1161 * %SIOCCHGTUNNEL: change tunnel parameters to those given
1162 * %SIOCDELTUNNEL: delete tunnel
1164 * The fallback device "ip6tnl0", created during module
1165 * initialization, can be used for creating other tunnel devices.
1169 * %-EFAULT if unable to copy data to or from userspace,
1170 * %-EPERM if current process hasn't %CAP_NET_ADMIN set
1171 * %-EINVAL if passed tunnel parameters are invalid,
1172 * %-EEXIST if changing a tunnel's parameters would cause a conflict
1173 * %-ENODEV if attempting to change or delete a nonexisting device
1177 ip6ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1180 struct ip6_tnl_parm p;
1181 struct ip6_tnl *t = NULL;
1185 if (dev == ip6ip6_fb_tnl_dev) {
1186 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p))) {
1190 t = ip6ip6_tnl_locate(&p, 0);
1193 t = netdev_priv(dev);
1194 memcpy(&p, &t->parms, sizeof (p));
1195 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof (p))) {
1202 if (!capable(CAP_NET_ADMIN))
1205 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
1208 if (p.proto != IPPROTO_IPV6)
1210 t = ip6ip6_tnl_locate(&p, cmd == SIOCADDTUNNEL);
1211 if (dev != ip6ip6_fb_tnl_dev && cmd == SIOCCHGTUNNEL) {
1213 if (t->dev != dev) {
1218 t = netdev_priv(dev);
1220 ip6ip6_tnl_unlink(t);
1221 err = ip6ip6_tnl_change(t, &p);
1223 netdev_state_change(dev);
1227 if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof (p)))
1231 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1235 if (!capable(CAP_NET_ADMIN))
1238 if (dev == ip6ip6_fb_tnl_dev) {
1240 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
1243 if ((t = ip6ip6_tnl_locate(&p, 0)) == NULL)
1246 if (t->dev == ip6ip6_fb_tnl_dev)
1251 unregister_netdevice(dev);
1260 * ip6ip6_tnl_get_stats - return the stats for tunnel device
1261 * @dev: virtual device associated with tunnel
1263 * Return: stats for device
1266 static struct net_device_stats *
1267 ip6ip6_tnl_get_stats(struct net_device *dev)
1269 return &(((struct ip6_tnl *)netdev_priv(dev))->stat);
1273 * ip6ip6_tnl_change_mtu - change mtu manually for tunnel device
1274 * @dev: virtual device associated with tunnel
1275 * @new_mtu: the new mtu
1279 * %-EINVAL if mtu too small
1283 ip6ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1285 if (new_mtu < IPV6_MIN_MTU) {
1293 * ip6ip6_tnl_dev_setup - setup virtual tunnel device
1294 * @dev: virtual device associated with tunnel
1297 * Initialize function pointers and device parameters
1300 static void ip6ip6_tnl_dev_setup(struct net_device *dev)
1302 SET_MODULE_OWNER(dev);
1303 dev->uninit = ip6ip6_tnl_dev_uninit;
1304 dev->destructor = free_netdev;
1305 dev->hard_start_xmit = ip6_tnl_xmit;
1306 dev->get_stats = ip6ip6_tnl_get_stats;
1307 dev->do_ioctl = ip6ip6_tnl_ioctl;
1308 dev->change_mtu = ip6ip6_tnl_change_mtu;
1310 dev->type = ARPHRD_TUNNEL6;
1311 dev->hard_header_len = LL_MAX_HEADER + sizeof (struct ipv6hdr);
1312 dev->mtu = ETH_DATA_LEN - sizeof (struct ipv6hdr);
1313 dev->flags |= IFF_NOARP;
1314 dev->addr_len = sizeof(struct in6_addr);
1319 * ip6ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1320 * @dev: virtual device associated with tunnel
1324 ip6ip6_tnl_dev_init_gen(struct net_device *dev)
1326 struct ip6_tnl *t = netdev_priv(dev);
1327 t->fl.proto = IPPROTO_IPV6;
1329 strcpy(t->parms.name, dev->name);
1333 * ip6ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1334 * @dev: virtual device associated with tunnel
1338 ip6ip6_tnl_dev_init(struct net_device *dev)
1340 struct ip6_tnl *t = netdev_priv(dev);
1341 ip6ip6_tnl_dev_init_gen(dev);
1342 ip6ip6_tnl_link_config(t);
1347 * ip6ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1348 * @dev: fallback device
1354 ip6ip6_fb_tnl_dev_init(struct net_device *dev)
1356 struct ip6_tnl *t = netdev_priv(dev);
1357 ip6ip6_tnl_dev_init_gen(dev);
1363 static struct xfrm6_tunnel ip4ip6_handler = {
1364 .handler = ip4ip6_rcv,
1365 .err_handler = ip4ip6_err,
1369 static struct xfrm6_tunnel ip6ip6_handler = {
1370 .handler = ip6ip6_rcv,
1371 .err_handler = ip6ip6_err,
1376 * ip6_tunnel_init - register protocol and reserve needed resources
1378 * Return: 0 on success
1381 static int __init ip6_tunnel_init(void)
1385 if (xfrm6_tunnel_register(&ip4ip6_handler, AF_INET)) {
1386 printk(KERN_ERR "ip4ip6 init: can't register tunnel\n");
1391 if (xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6)) {
1392 printk(KERN_ERR "ip6ip6 init: can't register tunnel\n");
1396 ip6ip6_fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
1397 ip6ip6_tnl_dev_setup);
1399 if (!ip6ip6_fb_tnl_dev) {
1403 ip6ip6_fb_tnl_dev->init = ip6ip6_fb_tnl_dev_init;
1405 if ((err = register_netdev(ip6ip6_fb_tnl_dev))) {
1406 free_netdev(ip6ip6_fb_tnl_dev);
1411 xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
1413 xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
1418 static void __exit ip6ip6_destroy_tunnels(void)
1423 for (h = 0; h < HASH_SIZE; h++) {
1424 while ((t = tnls_r_l[h]) != NULL)
1425 unregister_netdevice(t->dev);
1429 unregister_netdevice(t->dev);
1433 * ip6_tunnel_cleanup - free resources and unregister protocol
1436 static void __exit ip6_tunnel_cleanup(void)
1438 if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
1439 printk(KERN_INFO "ip4ip6 close: can't deregister tunnel\n");
1441 if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
1442 printk(KERN_INFO "ip6ip6 close: can't deregister tunnel\n");
1445 ip6ip6_destroy_tunnels();
1449 module_init(ip6_tunnel_init);
1450 module_exit(ip6_tunnel_cleanup);