2 * IPv6 over IPv4 tunnel device - Simple Internet Transition (SIT)
3 * Linux INET6 implementation
6 * Pedro Roque <roque@di.fc.ul.pt>
7 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
9 * $Id: sit.c,v 1.53 2001/09/25 05:09:53 davem Exp $
11 * This program is free software; you can redistribute it and/or
12 * modify it under the terms of the GNU General Public License
13 * as published by the Free Software Foundation; either version
14 * 2 of the License, or (at your option) any later version.
17 * Roger Venning <r.venning@telstra.com>: 6to4 support
18 * Nate Thompson <nate@thebog.net>: 6to4 support
21 #include <linux/config.h>
22 #include <linux/module.h>
23 #include <linux/errno.h>
24 #include <linux/types.h>
25 #include <linux/socket.h>
26 #include <linux/sockios.h>
27 #include <linux/sched.h>
28 #include <linux/net.h>
29 #include <linux/in6.h>
30 #include <linux/netdevice.h>
31 #include <linux/if_arp.h>
32 #include <linux/icmp.h>
33 #include <asm/uaccess.h>
34 #include <linux/init.h>
35 #include <linux/netfilter_ipv4.h>
41 #include <net/protocol.h>
42 #include <net/transp_v6.h>
43 #include <net/ip6_fib.h>
44 #include <net/ip6_route.h>
45 #include <net/ndisc.h>
46 #include <net/addrconf.h>
51 #include <net/inet_ecn.h>
55 This version of net/ipv6/sit.c is cloned of net/ipv4/ip_gre.c
57 For comments look at net/ipv4/ip_gre.c --ANK
61 #define HASH(addr) ((addr^(addr>>4))&0xF)
63 static int ipip6_fb_tunnel_init(struct net_device *dev);
64 static int ipip6_tunnel_init(struct net_device *dev);
65 static void ipip6_tunnel_setup(struct net_device *dev);
67 static struct net_device *ipip6_fb_tunnel_dev;
69 static struct ip_tunnel *tunnels_r_l[HASH_SIZE];
70 static struct ip_tunnel *tunnels_r[HASH_SIZE];
71 static struct ip_tunnel *tunnels_l[HASH_SIZE];
72 static struct ip_tunnel *tunnels_wc[1];
73 static struct ip_tunnel **tunnels[4] = { tunnels_wc, tunnels_l, tunnels_r, tunnels_r_l };
75 static rwlock_t ipip6_lock = RW_LOCK_UNLOCKED;
77 static struct ip_tunnel * ipip6_tunnel_lookup(u32 remote, u32 local)
79 unsigned h0 = HASH(remote);
80 unsigned h1 = HASH(local);
83 for (t = tunnels_r_l[h0^h1]; t; t = t->next) {
84 if (local == t->parms.iph.saddr &&
85 remote == t->parms.iph.daddr && (t->dev->flags&IFF_UP))
88 for (t = tunnels_r[h0]; t; t = t->next) {
89 if (remote == t->parms.iph.daddr && (t->dev->flags&IFF_UP))
92 for (t = tunnels_l[h1]; t; t = t->next) {
93 if (local == t->parms.iph.saddr && (t->dev->flags&IFF_UP))
96 if ((t = tunnels_wc[0]) != NULL && (t->dev->flags&IFF_UP))
101 static struct ip_tunnel ** ipip6_bucket(struct ip_tunnel *t)
103 u32 remote = t->parms.iph.daddr;
104 u32 local = t->parms.iph.saddr;
116 return &tunnels[prio][h];
119 static void ipip6_tunnel_unlink(struct ip_tunnel *t)
121 struct ip_tunnel **tp;
123 for (tp = ipip6_bucket(t); *tp; tp = &(*tp)->next) {
125 write_lock_bh(&ipip6_lock);
127 write_unlock_bh(&ipip6_lock);
133 static void ipip6_tunnel_link(struct ip_tunnel *t)
135 struct ip_tunnel **tp = ipip6_bucket(t);
137 write_lock_bh(&ipip6_lock);
139 write_unlock_bh(&ipip6_lock);
143 static struct ip_tunnel * ipip6_tunnel_locate(struct ip_tunnel_parm *parms, int create)
145 u32 remote = parms->iph.daddr;
146 u32 local = parms->iph.saddr;
147 struct ip_tunnel *t, **tp, *nt;
148 struct net_device *dev;
161 for (tp = &tunnels[prio][h]; (t = *tp) != NULL; tp = &t->next) {
162 if (local == t->parms.iph.saddr && remote == t->parms.iph.daddr)
169 strlcpy(name, parms->name, IFNAMSIZ);
172 for (i=1; i<100; i++) {
173 sprintf(name, "sit%d", i);
174 if (__dev_get_by_name(name) == NULL)
181 dev = alloc_netdev(sizeof(*t), name, ipip6_tunnel_setup);
186 dev->init = ipip6_tunnel_init;
189 if (register_netdevice(dev) < 0) {
196 ipip6_tunnel_link(nt);
197 /* Do not decrement MOD_USE_COUNT here. */
204 static void ipip6_tunnel_uninit(struct net_device *dev)
206 if (dev == ipip6_fb_tunnel_dev) {
207 write_lock_bh(&ipip6_lock);
208 tunnels_wc[0] = NULL;
209 write_unlock_bh(&ipip6_lock);
212 ipip6_tunnel_unlink((struct ip_tunnel*)dev->priv);
218 static void ipip6_err(struct sk_buff *skb, u32 info)
220 #ifndef I_WISH_WORLD_WERE_PERFECT
222 /* It is not :-( All the routers (except for Linux) return only
223 8 bytes of packet payload. It means, that precise relaying of
224 ICMP in the real Internet is absolutely infeasible.
226 struct iphdr *iph = (struct iphdr*)skb->data;
227 int type = skb->h.icmph->type;
228 int code = skb->h.icmph->code;
233 case ICMP_PARAMETERPROB:
236 case ICMP_DEST_UNREACH:
239 case ICMP_PORT_UNREACH:
240 /* Impossible event. */
242 case ICMP_FRAG_NEEDED:
243 /* Soft state for pmtu is maintained by IP core. */
246 /* All others are translated to HOST_UNREACH.
247 rfc2003 contains "deep thoughts" about NET_UNREACH,
248 I believe they are just ether pollution. --ANK
253 case ICMP_TIME_EXCEEDED:
254 if (code != ICMP_EXC_TTL)
259 read_lock(&ipip6_lock);
260 t = ipip6_tunnel_lookup(iph->daddr, iph->saddr);
261 if (t == NULL || t->parms.iph.daddr == 0)
263 if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
266 if (jiffies - t->err_time < IPTUNNEL_ERR_TIMEO)
270 t->err_time = jiffies;
272 read_unlock(&ipip6_lock);
275 struct iphdr *iph = (struct iphdr*)dp;
276 int hlen = iph->ihl<<2;
277 struct ipv6hdr *iph6;
278 int type = skb->h.icmph->type;
279 int code = skb->h.icmph->code;
283 struct sk_buff *skb2;
284 struct rt6_info *rt6i;
286 if (len < hlen + sizeof(struct ipv6hdr))
288 iph6 = (struct ipv6hdr*)(dp + hlen);
293 case ICMP_PARAMETERPROB:
294 if (skb->h.icmph->un.gateway < hlen)
297 /* So... This guy found something strange INSIDE encapsulated
298 packet. Well, he is fool, but what can we do ?
300 rel_type = ICMPV6_PARAMPROB;
301 rel_info = skb->h.icmph->un.gateway - hlen;
304 case ICMP_DEST_UNREACH:
307 case ICMP_PORT_UNREACH:
308 /* Impossible event. */
310 case ICMP_FRAG_NEEDED:
311 /* Too complicated case ... */
314 /* All others are translated to HOST_UNREACH.
315 rfc2003 contains "deep thoughts" about NET_UNREACH,
316 I believe, it is just ether pollution. --ANK
318 rel_type = ICMPV6_DEST_UNREACH;
319 rel_code = ICMPV6_ADDR_UNREACH;
323 case ICMP_TIME_EXCEEDED:
324 if (code != ICMP_EXC_TTL)
326 rel_type = ICMPV6_TIME_EXCEED;
327 rel_code = ICMPV6_EXC_HOPLIMIT;
331 /* Prepare fake skb to feed it to icmpv6_send */
332 skb2 = skb_clone(skb, GFP_ATOMIC);
335 dst_release(skb2->dst);
337 skb_pull(skb2, skb->data - (u8*)iph6);
338 skb2->nh.raw = skb2->data;
340 /* Try to guess incoming interface */
341 rt6i = rt6_lookup(&iph6->saddr, NULL, NULL, 0);
342 if (rt6i && rt6i->rt6i_dev) {
343 skb2->dev = rt6i->rt6i_dev;
345 rt6i = rt6_lookup(&iph6->daddr, &iph6->saddr, NULL, 0);
347 if (rt6i && rt6i->rt6i_dev && rt6i->rt6i_dev->type == ARPHRD_SIT) {
348 struct ip_tunnel * t = (struct ip_tunnel*)rt6i->rt6i_dev->priv;
349 if (rel_type == ICMPV6_TIME_EXCEED && t->parms.iph.ttl) {
350 rel_type = ICMPV6_DEST_UNREACH;
351 rel_code = ICMPV6_ADDR_UNREACH;
353 icmpv6_send(skb2, rel_type, rel_code, rel_info, skb2->dev);
361 static inline void ipip6_ecn_decapsulate(struct iphdr *iph, struct sk_buff *skb)
363 if (INET_ECN_is_ce(iph->tos) &&
364 INET_ECN_is_not_ce(ip6_get_dsfield(skb->nh.ipv6h)))
365 IP6_ECN_set_ce(skb->nh.ipv6h);
368 static int ipip6_rcv(struct sk_buff *skb)
371 struct ip_tunnel *tunnel;
373 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
378 read_lock(&ipip6_lock);
379 if ((tunnel = ipip6_tunnel_lookup(iph->saddr, iph->daddr)) != NULL) {
381 skb->mac.raw = skb->nh.raw;
382 skb->nh.raw = skb->data;
383 memset(&(IPCB(skb)->opt), 0, sizeof(struct ip_options));
384 skb->protocol = htons(ETH_P_IPV6);
385 skb->pkt_type = PACKET_HOST;
386 tunnel->stat.rx_packets++;
387 tunnel->stat.rx_bytes += skb->len;
388 skb->dev = tunnel->dev;
389 dst_release(skb->dst);
391 #ifdef CONFIG_NETFILTER
392 nf_conntrack_put(skb->nfct);
394 #ifdef CONFIG_NETFILTER_DEBUG
398 ipip6_ecn_decapsulate(iph, skb);
400 read_unlock(&ipip6_lock);
404 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PROT_UNREACH, 0);
406 read_unlock(&ipip6_lock);
411 /* Returns the embedded IPv4 address if the IPv6 address
412 comes from 6to4 (RFC 3056) addr space */
414 static inline u32 try_6to4(struct in6_addr *v6dst)
418 if (v6dst->s6_addr16[0] == htons(0x2002)) {
419 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
420 memcpy(&dst, &v6dst->s6_addr16[1], 4);
426 * This function assumes it is being called from dev_queue_xmit()
427 * and that skb is filled properly by that function.
430 static int ipip6_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
432 struct ip_tunnel *tunnel = (struct ip_tunnel*)dev->priv;
433 struct net_device_stats *stats = &tunnel->stat;
434 struct iphdr *tiph = &tunnel->parms.iph;
435 struct ipv6hdr *iph6 = skb->nh.ipv6h;
436 u8 tos = tunnel->parms.iph.tos;
437 struct rtable *rt; /* Route to the other host */
438 struct net_device *tdev; /* Device to other host */
439 struct iphdr *iph; /* Our new IP header */
440 int max_headroom; /* The extra header space needed */
441 u32 dst = tiph->daddr;
443 struct in6_addr *addr6;
446 if (tunnel->recursion++) {
447 tunnel->stat.collisions++;
451 if (skb->protocol != htons(ETH_P_IPV6))
455 dst = try_6to4(&iph6->daddr);
458 struct neighbour *neigh = NULL;
461 neigh = skb->dst->neighbour;
465 printk(KERN_DEBUG "sit: nexthop == NULL\n");
469 addr6 = (struct in6_addr*)&neigh->primary_key;
470 addr_type = ipv6_addr_type(addr6);
472 if (addr_type == IPV6_ADDR_ANY) {
473 addr6 = &skb->nh.ipv6h->daddr;
474 addr_type = ipv6_addr_type(addr6);
477 if ((addr_type & IPV6_ADDR_COMPATv4) == 0)
480 dst = addr6->s6_addr32[3];
484 struct flowi fl = { .nl_u = { .ip4_u =
486 .saddr = tiph->saddr,
487 .tos = RT_TOS(tos) } },
488 .oif = tunnel->parms.link,
489 .proto = IPPROTO_IPV6 };
490 if (ip_route_output_key(&rt, &fl)) {
491 tunnel->stat.tx_carrier_errors++;
495 if (rt->rt_type != RTN_UNICAST) {
496 tunnel->stat.tx_carrier_errors++;
499 tdev = rt->u.dst.dev;
503 tunnel->stat.collisions++;
508 mtu = dst_pmtu(&rt->u.dst) - sizeof(struct iphdr);
510 mtu = skb->dst ? dst_pmtu(skb->dst) : dev->mtu;
513 tunnel->stat.collisions++;
517 if (mtu < IPV6_MIN_MTU)
519 if (tunnel->parms.iph.daddr && skb->dst)
520 skb->dst->ops->update_pmtu(skb->dst, mtu);
522 if (skb->len > mtu) {
523 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, dev);
528 if (tunnel->err_count > 0) {
529 if (jiffies - tunnel->err_time < IPTUNNEL_ERR_TIMEO) {
531 dst_link_failure(skb);
533 tunnel->err_count = 0;
537 * Okay, now see if we can stuff it in the buffer as-is.
539 max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr);
541 if (skb_headroom(skb) < max_headroom || skb_cloned(skb) || skb_shared(skb)) {
542 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
551 skb_set_owner_w(new_skb, skb->sk);
554 iph6 = skb->nh.ipv6h;
557 skb->h.raw = skb->nh.raw;
558 skb->nh.raw = skb_push(skb, sizeof(struct iphdr));
559 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
560 dst_release(skb->dst);
561 skb->dst = &rt->u.dst;
564 * Push down and install the IPIP header.
569 iph->ihl = sizeof(struct iphdr)>>2;
570 if (mtu > IPV6_MIN_MTU)
571 iph->frag_off = htons(IP_DF);
575 iph->protocol = IPPROTO_IPV6;
576 iph->tos = INET_ECN_encapsulate(tos, ip6_get_dsfield(iph6));
577 iph->daddr = rt->rt_dst;
578 iph->saddr = rt->rt_src;
580 if ((iph->ttl = tiph->ttl) == 0)
581 iph->ttl = iph6->hop_limit;
583 #ifdef CONFIG_NETFILTER
584 nf_conntrack_put(skb->nfct);
586 #ifdef CONFIG_NETFILTER_DEBUG
596 dst_link_failure(skb);
605 ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
608 struct ip_tunnel_parm p;
614 if (dev == ipip6_fb_tunnel_dev) {
615 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
619 t = ipip6_tunnel_locate(&p, 0);
622 t = (struct ip_tunnel*)dev->priv;
623 memcpy(&p, &t->parms, sizeof(p));
624 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
631 if (!capable(CAP_NET_ADMIN))
635 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
639 if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPV6 ||
640 p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
643 p.iph.frag_off |= htons(IP_DF);
645 t = ipip6_tunnel_locate(&p, cmd == SIOCADDTUNNEL);
647 if (dev != ipip6_fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
654 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
655 (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
659 t = (struct ip_tunnel*)dev->priv;
660 ipip6_tunnel_unlink(t);
661 t->parms.iph.saddr = p.iph.saddr;
662 t->parms.iph.daddr = p.iph.daddr;
663 memcpy(dev->dev_addr, &p.iph.saddr, 4);
664 memcpy(dev->broadcast, &p.iph.daddr, 4);
665 ipip6_tunnel_link(t);
666 netdev_state_change(dev);
672 if (cmd == SIOCCHGTUNNEL) {
673 t->parms.iph.ttl = p.iph.ttl;
674 t->parms.iph.tos = p.iph.tos;
676 if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
679 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
684 if (!capable(CAP_NET_ADMIN))
687 if (dev == ipip6_fb_tunnel_dev) {
689 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
692 if ((t = ipip6_tunnel_locate(&p, 0)) == NULL)
695 if (t == ipip6_fb_tunnel_dev->priv)
699 err = unregister_netdevice(dev);
710 static struct net_device_stats *ipip6_tunnel_get_stats(struct net_device *dev)
712 return &(((struct ip_tunnel*)dev->priv)->stat);
715 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
717 if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
723 static void ipip6_tunnel_setup(struct net_device *dev)
725 SET_MODULE_OWNER(dev);
726 dev->uninit = ipip6_tunnel_uninit;
727 dev->destructor = free_netdev;
728 dev->hard_start_xmit = ipip6_tunnel_xmit;
729 dev->get_stats = ipip6_tunnel_get_stats;
730 dev->do_ioctl = ipip6_tunnel_ioctl;
731 dev->change_mtu = ipip6_tunnel_change_mtu;
733 dev->type = ARPHRD_SIT;
734 dev->hard_header_len = LL_MAX_HEADER + sizeof(struct iphdr);
735 dev->mtu = 1500 - sizeof(struct iphdr);
736 dev->flags = IFF_NOARP;
741 static int ipip6_tunnel_init(struct net_device *dev)
743 struct net_device *tdev = NULL;
744 struct ip_tunnel *tunnel;
747 tunnel = (struct ip_tunnel*)dev->priv;
748 iph = &tunnel->parms.iph;
751 strcpy(tunnel->parms.name, dev->name);
753 memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
754 memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
757 struct flowi fl = { .nl_u = { .ip4_u =
758 { .daddr = iph->daddr,
760 .tos = RT_TOS(iph->tos) } },
761 .oif = tunnel->parms.link,
762 .proto = IPPROTO_IPV6 };
764 if (!ip_route_output_key(&rt, &fl)) {
765 tdev = rt->u.dst.dev;
768 dev->flags |= IFF_POINTOPOINT;
771 if (!tdev && tunnel->parms.link)
772 tdev = __dev_get_by_index(tunnel->parms.link);
775 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
776 dev->mtu = tdev->mtu - sizeof(struct iphdr);
777 if (dev->mtu < IPV6_MIN_MTU)
778 dev->mtu = IPV6_MIN_MTU;
780 dev->iflink = tunnel->parms.link;
785 int __init ipip6_fb_tunnel_init(struct net_device *dev)
787 struct ip_tunnel *tunnel = dev->priv;
788 struct iphdr *iph = &tunnel->parms.iph;
791 strcpy(tunnel->parms.name, dev->name);
794 iph->protocol = IPPROTO_IPV6;
799 tunnels_wc[0] = tunnel;
803 static struct inet_protocol sit_protocol = {
804 .handler = ipip6_rcv,
805 .err_handler = ipip6_err,
808 void __exit sit_cleanup(void)
810 inet_del_protocol(&sit_protocol, IPPROTO_IPV6);
811 unregister_netdev(ipip6_fb_tunnel_dev);
814 int __init sit_init(void)
818 printk(KERN_INFO "IPv6 over IPv4 tunneling driver\n");
820 if (inet_add_protocol(&sit_protocol, IPPROTO_IPV6) < 0) {
821 printk(KERN_INFO "sit init: Can't add protocol\n");
825 ipip6_fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
827 if (!ipip6_fb_tunnel_dev) {
832 ipip6_fb_tunnel_dev->init = ipip6_fb_tunnel_init;
834 if ((err = register_netdev(ipip6_fb_tunnel_dev)))
840 inet_del_protocol(&sit_protocol, IPPROTO_IPV6);
841 free_netdev(ipip6_fb_tunnel_dev);