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 net *net, 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_net *ip6n, 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_net *ip6n, struct ip6_tnl *t)
191 struct ip6_tnl **tp = ip6_tnl_bucket(ip6n, &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_net *ip6n, struct ip6_tnl *t)
209 for (tp = ip6_tnl_bucket(ip6n, &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 net *net, struct ip6_tnl_parm *p)
233 struct net_device *dev;
237 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
240 strlcpy(name, p->name, IFNAMSIZ);
242 sprintf(name, "ip6tnl%%d");
244 dev = alloc_netdev(sizeof (*t), name, ip6_tnl_dev_setup);
248 if (strchr(name, '%')) {
249 if (dev_alloc_name(dev, name) < 0)
253 t = netdev_priv(dev);
254 dev->init = ip6_tnl_dev_init;
257 if ((err = register_netdevice(dev)) < 0)
261 ip6_tnl_link(ip6n, t);
271 * ip6_tnl_locate - find or create tunnel matching given parameters
272 * @p: tunnel parameters
273 * @create: != 0 if allowed to create new tunnel if no match found
276 * ip6_tnl_locate() first tries to locate an existing tunnel
277 * based on @parms. If this is unsuccessful, but @create is set a new
278 * tunnel device is created and registered for use.
281 * matching tunnel or NULL
284 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
285 struct ip6_tnl_parm *p, int create)
287 struct in6_addr *remote = &p->raddr;
288 struct in6_addr *local = &p->laddr;
290 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
292 for (t = *ip6_tnl_bucket(ip6n, p); t; t = t->next) {
293 if (ipv6_addr_equal(local, &t->parms.laddr) &&
294 ipv6_addr_equal(remote, &t->parms.raddr))
299 return ip6_tnl_create(net, p);
303 * ip6_tnl_dev_uninit - tunnel device uninitializer
304 * @dev: the device to be destroyed
307 * ip6_tnl_dev_uninit() removes tunnel from its list
311 ip6_tnl_dev_uninit(struct net_device *dev)
313 struct ip6_tnl *t = netdev_priv(dev);
314 struct net *net = dev_net(dev);
315 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
317 if (dev == ip6_fb_tnl_dev) {
318 write_lock_bh(&ip6_tnl_lock);
320 write_unlock_bh(&ip6_tnl_lock);
322 ip6_tnl_unlink(ip6n, t);
324 ip6_tnl_dst_reset(t);
329 * parse_tvl_tnl_enc_lim - handle encapsulation limit option
330 * @skb: received socket buffer
333 * 0 if none was found,
334 * else index to encapsulation limit
338 parse_tlv_tnl_enc_lim(struct sk_buff *skb, __u8 * raw)
340 struct ipv6hdr *ipv6h = (struct ipv6hdr *) raw;
341 __u8 nexthdr = ipv6h->nexthdr;
342 __u16 off = sizeof (*ipv6h);
344 while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
346 struct ipv6_opt_hdr *hdr;
347 if (raw + off + sizeof (*hdr) > skb->data &&
348 !pskb_may_pull(skb, raw - skb->data + off + sizeof (*hdr)))
351 hdr = (struct ipv6_opt_hdr *) (raw + off);
352 if (nexthdr == NEXTHDR_FRAGMENT) {
353 struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
354 if (frag_hdr->frag_off)
357 } else if (nexthdr == NEXTHDR_AUTH) {
358 optlen = (hdr->hdrlen + 2) << 2;
360 optlen = ipv6_optlen(hdr);
362 if (nexthdr == NEXTHDR_DEST) {
365 struct ipv6_tlv_tnl_enc_lim *tel;
367 /* No more room for encapsulation limit */
368 if (i + sizeof (*tel) > off + optlen)
371 tel = (struct ipv6_tlv_tnl_enc_lim *) &raw[i];
372 /* return index of option if found and valid */
373 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
376 /* else jump to next option */
378 i += tel->length + 2;
383 nexthdr = hdr->nexthdr;
390 * ip6_tnl_err - tunnel error handler
393 * ip6_tnl_err() should handle errors in the tunnel according
394 * to the specifications in RFC 2473.
398 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
399 int *type, int *code, int *msg, __u32 *info, int offset)
401 struct ipv6hdr *ipv6h = (struct ipv6hdr *) skb->data;
404 int rel_type = ICMPV6_DEST_UNREACH;
405 int rel_code = ICMPV6_ADDR_UNREACH;
410 /* If the packet doesn't contain the original IPv6 header we are
411 in trouble since we might need the source address for further
412 processing of the error. */
414 read_lock(&ip6_tnl_lock);
415 if ((t = ip6_tnl_lookup(&init_net, &ipv6h->daddr,
416 &ipv6h->saddr)) == NULL)
419 if (t->parms.proto != ipproto && t->parms.proto != 0)
426 struct ipv6_tlv_tnl_enc_lim *tel;
428 case ICMPV6_DEST_UNREACH:
431 "%s: Path to destination invalid "
432 "or inactive!\n", t->parms.name);
435 case ICMPV6_TIME_EXCEED:
436 if ((*code) == ICMPV6_EXC_HOPLIMIT) {
439 "%s: Too small hop limit or "
440 "routing loop in tunnel!\n",
445 case ICMPV6_PARAMPROB:
447 if ((*code) == ICMPV6_HDR_FIELD)
448 teli = parse_tlv_tnl_enc_lim(skb, skb->data);
450 if (teli && teli == *info - 2) {
451 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
452 if (tel->encap_limit == 0) {
455 "%s: Too small encapsulation "
456 "limit or routing loop in "
457 "tunnel!\n", t->parms.name);
460 } else if (net_ratelimit()) {
462 "%s: Recipient unable to parse tunneled "
463 "packet!\n ", t->parms.name);
466 case ICMPV6_PKT_TOOBIG:
467 mtu = *info - offset;
468 if (mtu < IPV6_MIN_MTU)
472 if ((len = sizeof (*ipv6h) + ntohs(ipv6h->payload_len)) > mtu) {
473 rel_type = ICMPV6_PKT_TOOBIG;
487 read_unlock(&ip6_tnl_lock);
492 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
493 int type, int code, int offset, __be32 info)
498 __u32 rel_info = ntohl(info);
500 struct sk_buff *skb2;
505 err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
506 &rel_msg, &rel_info, offset);
514 case ICMPV6_DEST_UNREACH:
515 if (rel_code != ICMPV6_ADDR_UNREACH)
517 rel_type = ICMP_DEST_UNREACH;
518 rel_code = ICMP_HOST_UNREACH;
520 case ICMPV6_PKT_TOOBIG:
523 rel_type = ICMP_DEST_UNREACH;
524 rel_code = ICMP_FRAG_NEEDED;
530 if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
533 skb2 = skb_clone(skb, GFP_ATOMIC);
537 dst_release(skb2->dst);
539 skb_pull(skb2, offset);
540 skb_reset_network_header(skb2);
543 /* Try to guess incoming interface */
544 memset(&fl, 0, sizeof(fl));
545 fl.fl4_dst = eiph->saddr;
546 fl.fl4_tos = RT_TOS(eiph->tos);
547 fl.proto = IPPROTO_IPIP;
548 if (ip_route_output_key(&init_net, &rt, &fl))
551 skb2->dev = rt->u.dst.dev;
553 /* route "incoming" packet */
554 if (rt->rt_flags & RTCF_LOCAL) {
557 fl.fl4_dst = eiph->daddr;
558 fl.fl4_src = eiph->saddr;
559 fl.fl4_tos = eiph->tos;
560 if (ip_route_output_key(&init_net, &rt, &fl) ||
561 rt->u.dst.dev->type != ARPHRD_TUNNEL) {
565 skb2->dst = (struct dst_entry *)rt;
568 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
570 skb2->dst->dev->type != ARPHRD_TUNNEL)
574 /* change mtu on this route */
575 if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
576 if (rel_info > dst_mtu(skb2->dst))
579 skb2->dst->ops->update_pmtu(skb2->dst, rel_info);
582 icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
590 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
591 int type, int code, int offset, __be32 info)
596 __u32 rel_info = ntohl(info);
599 err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
600 &rel_msg, &rel_info, offset);
604 if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
606 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
611 dst_release(skb2->dst);
613 skb_pull(skb2, offset);
614 skb_reset_network_header(skb2);
616 /* Try to guess incoming interface */
617 rt = rt6_lookup(&init_net, &ipv6_hdr(skb2)->saddr, NULL, 0, 0);
619 if (rt && rt->rt6i_dev)
620 skb2->dev = rt->rt6i_dev;
622 icmpv6_send(skb2, rel_type, rel_code, rel_info, skb2->dev);
625 dst_release(&rt->u.dst);
633 static void ip4ip6_dscp_ecn_decapsulate(struct ip6_tnl *t,
634 struct ipv6hdr *ipv6h,
637 __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
639 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
640 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
642 if (INET_ECN_is_ce(dsfield))
643 IP_ECN_set_ce(ip_hdr(skb));
646 static void ip6ip6_dscp_ecn_decapsulate(struct ip6_tnl *t,
647 struct ipv6hdr *ipv6h,
650 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
651 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
653 if (INET_ECN_is_ce(ipv6_get_dsfield(ipv6h)))
654 IP6_ECN_set_ce(ipv6_hdr(skb));
657 static inline int ip6_tnl_rcv_ctl(struct ip6_tnl *t)
659 struct ip6_tnl_parm *p = &t->parms;
662 if (p->flags & IP6_TNL_F_CAP_RCV) {
663 struct net_device *ldev = NULL;
666 ldev = dev_get_by_index(&init_net, p->link);
668 if ((ipv6_addr_is_multicast(&p->laddr) ||
669 likely(ipv6_chk_addr(&init_net, &p->laddr, ldev, 0))) &&
670 likely(!ipv6_chk_addr(&init_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(&init_net, &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->stat.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->stat.rx_packets++;
730 t->stat.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;
799 if (p->flags & IP6_TNL_F_CAP_XMIT) {
800 struct net_device *ldev = NULL;
803 ldev = dev_get_by_index(&init_net, p->link);
805 if (unlikely(!ipv6_chk_addr(&init_net, &p->laddr, ldev, 0)))
807 "%s xmit: Local address not yet configured!\n",
809 else if (!ipv6_addr_is_multicast(&p->raddr) &&
810 unlikely(ipv6_chk_addr(&init_net, &p->raddr, NULL, 0)))
812 "%s xmit: Routing loop! "
813 "Remote address found on this node!\n",
823 * ip6_tnl_xmit2 - encapsulate packet and send
824 * @skb: the outgoing socket buffer
825 * @dev: the outgoing tunnel device
826 * @dsfield: dscp code for outer header
827 * @fl: flow of tunneled packet
828 * @encap_limit: encapsulation limit
829 * @pmtu: Path MTU is stored if packet is too big
832 * Build new header and do some sanity checks on the packet before sending
838 * %-EMSGSIZE message too big. return mtu in this case.
841 static int ip6_tnl_xmit2(struct sk_buff *skb,
842 struct net_device *dev,
848 struct ip6_tnl *t = netdev_priv(dev);
849 struct net_device_stats *stats = &t->stat;
850 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
851 struct ipv6_tel_txoption opt;
852 struct dst_entry *dst;
853 struct net_device *tdev;
855 unsigned int max_headroom = sizeof(struct ipv6hdr);
860 if ((dst = ip6_tnl_dst_check(t)) != NULL)
863 dst = ip6_route_output(&init_net, NULL, fl);
865 if (dst->error || xfrm_lookup(&dst, fl, NULL, 0) < 0)
866 goto tx_err_link_failure;
875 "%s: Local routing loop detected!\n",
877 goto tx_err_dst_release;
879 mtu = dst_mtu(dst) - sizeof (*ipv6h);
880 if (encap_limit >= 0) {
884 if (mtu < IPV6_MIN_MTU)
887 skb->dst->ops->update_pmtu(skb->dst, mtu);
888 if (skb->len > mtu) {
891 goto tx_err_dst_release;
895 * Okay, now see if we can stuff it in the buffer as-is.
897 max_headroom += LL_RESERVED_SPACE(tdev);
899 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
900 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
901 struct sk_buff *new_skb;
903 if (!(new_skb = skb_realloc_headroom(skb, max_headroom)))
904 goto tx_err_dst_release;
907 skb_set_owner_w(new_skb, skb->sk);
911 dst_release(skb->dst);
912 skb->dst = dst_clone(dst);
914 skb->transport_header = skb->network_header;
917 if (encap_limit >= 0) {
918 init_tel_txopt(&opt, encap_limit);
919 ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
921 skb_push(skb, sizeof(struct ipv6hdr));
922 skb_reset_network_header(skb);
923 ipv6h = ipv6_hdr(skb);
924 *(__be32*)ipv6h = fl->fl6_flowlabel | htonl(0x60000000);
925 dsfield = INET_ECN_encapsulate(0, dsfield);
926 ipv6_change_dsfield(ipv6h, ~INET_ECN_MASK, dsfield);
927 ipv6h->hop_limit = t->parms.hop_limit;
928 ipv6h->nexthdr = proto;
929 ipv6_addr_copy(&ipv6h->saddr, &fl->fl6_src);
930 ipv6_addr_copy(&ipv6h->daddr, &fl->fl6_dst);
933 err = ip6_local_out(skb);
935 if (net_xmit_eval(err) == 0) {
936 stats->tx_bytes += pkt_len;
940 stats->tx_aborted_errors++;
942 ip6_tnl_dst_store(t, dst);
945 stats->tx_carrier_errors++;
946 dst_link_failure(skb);
953 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
955 struct ip6_tnl *t = netdev_priv(dev);
956 struct iphdr *iph = ip_hdr(skb);
957 int encap_limit = -1;
963 if ((t->parms.proto != IPPROTO_IPIP && t->parms.proto != 0) ||
964 !ip6_tnl_xmit_ctl(t))
967 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
968 encap_limit = t->parms.encap_limit;
970 memcpy(&fl, &t->fl, sizeof (fl));
971 fl.proto = IPPROTO_IPIP;
973 dsfield = ipv4_get_dsfield(iph);
975 if ((t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS))
976 fl.fl6_flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
979 err = ip6_tnl_xmit2(skb, dev, dsfield, &fl, encap_limit, &mtu);
981 /* XXX: send ICMP error even if DF is not set. */
982 if (err == -EMSGSIZE)
983 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
992 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
994 struct ip6_tnl *t = netdev_priv(dev);
995 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
996 int encap_limit = -1;
1003 if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) ||
1004 !ip6_tnl_xmit_ctl(t) || ip6_tnl_addr_conflict(t, ipv6h))
1007 offset = parse_tlv_tnl_enc_lim(skb, skb_network_header(skb));
1009 struct ipv6_tlv_tnl_enc_lim *tel;
1010 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
1011 if (tel->encap_limit == 0) {
1012 icmpv6_send(skb, ICMPV6_PARAMPROB,
1013 ICMPV6_HDR_FIELD, offset + 2, skb->dev);
1016 encap_limit = tel->encap_limit - 1;
1017 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1018 encap_limit = t->parms.encap_limit;
1020 memcpy(&fl, &t->fl, sizeof (fl));
1021 fl.proto = IPPROTO_IPV6;
1023 dsfield = ipv6_get_dsfield(ipv6h);
1024 if ((t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS))
1025 fl.fl6_flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
1026 if ((t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL))
1027 fl.fl6_flowlabel |= (*(__be32 *) ipv6h & IPV6_FLOWLABEL_MASK);
1029 err = ip6_tnl_xmit2(skb, dev, dsfield, &fl, encap_limit, &mtu);
1031 if (err == -EMSGSIZE)
1032 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, dev);
1040 ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1042 struct ip6_tnl *t = netdev_priv(dev);
1043 struct net_device_stats *stats = &t->stat;
1046 if (t->recursion++) {
1047 t->stat.collisions++;
1051 switch (skb->protocol) {
1052 case __constant_htons(ETH_P_IP):
1053 ret = ip4ip6_tnl_xmit(skb, dev);
1055 case __constant_htons(ETH_P_IPV6):
1056 ret = ip6ip6_tnl_xmit(skb, dev);
1070 stats->tx_dropped++;
1076 static void ip6_tnl_set_cap(struct ip6_tnl *t)
1078 struct ip6_tnl_parm *p = &t->parms;
1079 int ltype = ipv6_addr_type(&p->laddr);
1080 int rtype = ipv6_addr_type(&p->raddr);
1082 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV);
1084 if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
1085 rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
1086 !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
1087 (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
1088 if (ltype&IPV6_ADDR_UNICAST)
1089 p->flags |= IP6_TNL_F_CAP_XMIT;
1090 if (rtype&IPV6_ADDR_UNICAST)
1091 p->flags |= IP6_TNL_F_CAP_RCV;
1095 static void ip6_tnl_link_config(struct ip6_tnl *t)
1097 struct net_device *dev = t->dev;
1098 struct ip6_tnl_parm *p = &t->parms;
1099 struct flowi *fl = &t->fl;
1101 memcpy(&dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1102 memcpy(&dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1104 /* Set up flowi template */
1105 ipv6_addr_copy(&fl->fl6_src, &p->laddr);
1106 ipv6_addr_copy(&fl->fl6_dst, &p->raddr);
1108 fl->fl6_flowlabel = 0;
1110 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1111 fl->fl6_flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1112 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1113 fl->fl6_flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1117 if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1118 dev->flags |= IFF_POINTOPOINT;
1120 dev->flags &= ~IFF_POINTOPOINT;
1122 dev->iflink = p->link;
1124 if (p->flags & IP6_TNL_F_CAP_XMIT) {
1125 int strict = (ipv6_addr_type(&p->raddr) &
1126 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1128 struct rt6_info *rt = rt6_lookup(&init_net, &p->raddr, &p->laddr,
1135 dev->hard_header_len = rt->rt6i_dev->hard_header_len +
1136 sizeof (struct ipv6hdr);
1138 dev->mtu = rt->rt6i_dev->mtu - sizeof (struct ipv6hdr);
1140 if (dev->mtu < IPV6_MIN_MTU)
1141 dev->mtu = IPV6_MIN_MTU;
1143 dst_release(&rt->u.dst);
1148 * ip6_tnl_change - update the tunnel parameters
1149 * @t: tunnel to be changed
1150 * @p: tunnel configuration parameters
1151 * @active: != 0 if tunnel is ready for use
1154 * ip6_tnl_change() updates the tunnel parameters
1158 ip6_tnl_change(struct ip6_tnl *t, struct ip6_tnl_parm *p)
1160 ipv6_addr_copy(&t->parms.laddr, &p->laddr);
1161 ipv6_addr_copy(&t->parms.raddr, &p->raddr);
1162 t->parms.flags = p->flags;
1163 t->parms.hop_limit = p->hop_limit;
1164 t->parms.encap_limit = p->encap_limit;
1165 t->parms.flowinfo = p->flowinfo;
1166 t->parms.link = p->link;
1167 t->parms.proto = p->proto;
1168 ip6_tnl_dst_reset(t);
1169 ip6_tnl_link_config(t);
1174 * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1175 * @dev: virtual device associated with tunnel
1176 * @ifr: parameters passed from userspace
1177 * @cmd: command to be performed
1180 * ip6_tnl_ioctl() is used for managing IPv6 tunnels
1183 * The possible commands are the following:
1184 * %SIOCGETTUNNEL: get tunnel parameters for device
1185 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1186 * %SIOCCHGTUNNEL: change tunnel parameters to those given
1187 * %SIOCDELTUNNEL: delete tunnel
1189 * The fallback device "ip6tnl0", created during module
1190 * initialization, can be used for creating other tunnel devices.
1194 * %-EFAULT if unable to copy data to or from userspace,
1195 * %-EPERM if current process hasn't %CAP_NET_ADMIN set
1196 * %-EINVAL if passed tunnel parameters are invalid,
1197 * %-EEXIST if changing a tunnel's parameters would cause a conflict
1198 * %-ENODEV if attempting to change or delete a nonexisting device
1202 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1205 struct ip6_tnl_parm p;
1206 struct ip6_tnl *t = NULL;
1207 struct net *net = dev_net(dev);
1208 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1212 if (dev == ip6_fb_tnl_dev) {
1213 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p))) {
1217 t = ip6_tnl_locate(net, &p, 0);
1220 t = netdev_priv(dev);
1221 memcpy(&p, &t->parms, sizeof (p));
1222 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof (p))) {
1229 if (!capable(CAP_NET_ADMIN))
1232 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
1235 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1238 t = ip6_tnl_locate(net, &p, cmd == SIOCADDTUNNEL);
1239 if (dev != ip6_fb_tnl_dev && cmd == SIOCCHGTUNNEL) {
1241 if (t->dev != dev) {
1246 t = netdev_priv(dev);
1248 ip6_tnl_unlink(ip6n, t);
1249 err = ip6_tnl_change(t, &p);
1250 ip6_tnl_link(ip6n, t);
1251 netdev_state_change(dev);
1255 if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof (p)))
1259 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
1263 if (!capable(CAP_NET_ADMIN))
1266 if (dev == ip6_fb_tnl_dev) {
1268 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
1271 if ((t = ip6_tnl_locate(net, &p, 0)) == NULL)
1274 if (t->dev == ip6_fb_tnl_dev)
1279 unregister_netdevice(dev);
1288 * ip6_tnl_get_stats - return the stats for tunnel device
1289 * @dev: virtual device associated with tunnel
1291 * Return: stats for device
1294 static struct net_device_stats *
1295 ip6_tnl_get_stats(struct net_device *dev)
1297 return &(((struct ip6_tnl *)netdev_priv(dev))->stat);
1301 * ip6_tnl_change_mtu - change mtu manually for tunnel device
1302 * @dev: virtual device associated with tunnel
1303 * @new_mtu: the new mtu
1307 * %-EINVAL if mtu too small
1311 ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1313 if (new_mtu < IPV6_MIN_MTU) {
1321 * ip6_tnl_dev_setup - setup virtual tunnel device
1322 * @dev: virtual device associated with tunnel
1325 * Initialize function pointers and device parameters
1328 static void ip6_tnl_dev_setup(struct net_device *dev)
1330 dev->uninit = ip6_tnl_dev_uninit;
1331 dev->destructor = free_netdev;
1332 dev->hard_start_xmit = ip6_tnl_xmit;
1333 dev->get_stats = ip6_tnl_get_stats;
1334 dev->do_ioctl = ip6_tnl_ioctl;
1335 dev->change_mtu = ip6_tnl_change_mtu;
1337 dev->type = ARPHRD_TUNNEL6;
1338 dev->hard_header_len = LL_MAX_HEADER + sizeof (struct ipv6hdr);
1339 dev->mtu = ETH_DATA_LEN - sizeof (struct ipv6hdr);
1340 dev->flags |= IFF_NOARP;
1341 dev->addr_len = sizeof(struct in6_addr);
1346 * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1347 * @dev: virtual device associated with tunnel
1351 ip6_tnl_dev_init_gen(struct net_device *dev)
1353 struct ip6_tnl *t = netdev_priv(dev);
1355 strcpy(t->parms.name, dev->name);
1359 * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1360 * @dev: virtual device associated with tunnel
1364 ip6_tnl_dev_init(struct net_device *dev)
1366 struct ip6_tnl *t = netdev_priv(dev);
1367 ip6_tnl_dev_init_gen(dev);
1368 ip6_tnl_link_config(t);
1373 * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1374 * @dev: fallback device
1380 ip6_fb_tnl_dev_init(struct net_device *dev)
1382 struct ip6_tnl *t = netdev_priv(dev);
1383 ip6_tnl_dev_init_gen(dev);
1384 t->parms.proto = IPPROTO_IPV6;
1390 static struct xfrm6_tunnel ip4ip6_handler = {
1391 .handler = ip4ip6_rcv,
1392 .err_handler = ip4ip6_err,
1396 static struct xfrm6_tunnel ip6ip6_handler = {
1397 .handler = ip6ip6_rcv,
1398 .err_handler = ip6ip6_err,
1402 static int ip6_tnl_init_net(struct net *net)
1405 struct ip6_tnl_net *ip6n;
1408 ip6n = kmalloc(sizeof(struct ip6_tnl_net), GFP_KERNEL);
1412 err = net_assign_generic(net, ip6_tnl_net_id, ip6n);
1424 static void ip6_tnl_exit_net(struct net *net)
1426 struct ip6_tnl_net *ip6n;
1428 ip6n = net_generic(net, ip6_tnl_net_id);
1432 static struct pernet_operations ip6_tnl_net_ops = {
1433 .init = ip6_tnl_init_net,
1434 .exit = ip6_tnl_exit_net,
1438 * ip6_tunnel_init - register protocol and reserve needed resources
1440 * Return: 0 on success
1443 static int __init ip6_tunnel_init(void)
1447 if (xfrm6_tunnel_register(&ip4ip6_handler, AF_INET)) {
1448 printk(KERN_ERR "ip6_tunnel init: can't register ip4ip6\n");
1453 if (xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6)) {
1454 printk(KERN_ERR "ip6_tunnel init: can't register ip6ip6\n");
1458 ip6_fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
1461 if (!ip6_fb_tnl_dev) {
1465 ip6_fb_tnl_dev->init = ip6_fb_tnl_dev_init;
1467 if ((err = register_netdev(ip6_fb_tnl_dev))) {
1468 free_netdev(ip6_fb_tnl_dev);
1472 err = register_pernet_gen_device(&ip6_tnl_net_id, &ip6_tnl_net_ops);
1477 unregister_netdevice(ip6_fb_tnl_dev);
1479 xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
1481 xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
1486 static void __exit ip6_tnl_destroy_tunnels(void)
1491 for (h = 0; h < HASH_SIZE; h++) {
1492 while ((t = tnls_r_l[h]) != NULL)
1493 unregister_netdevice(t->dev);
1497 unregister_netdevice(t->dev);
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");
1513 ip6_tnl_destroy_tunnels();
1516 unregister_pernet_gen_device(ip6_tnl_net_id, &ip6_tnl_net_ops);
1519 module_init(ip6_tunnel_init);
1520 module_exit(ip6_tunnel_cleanup);