Add linux-next specific files for 20110504
[linux-2.6/next.git] / net / ipv6 / sit.c
blob01ab1ac5bc6ab6a18ba29b976f11ba6c47fc226d
1 /*
2 * IPv6 over IPv4 tunnel device - Simple Internet Transition (SIT)
3 * Linux INET6 implementation
5 * Authors:
6 * Pedro Roque <roque@di.fc.ul.pt>
7 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
9 * This program is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU General Public License
11 * as published by the Free Software Foundation; either version
12 * 2 of the License, or (at your option) any later version.
14 * Changes:
15 * Roger Venning <r.venning@telstra.com>: 6to4 support
16 * Nate Thompson <nate@thebog.net>: 6to4 support
17 * Fred Templin <fred.l.templin@boeing.com>: isatap support
20 #include <linux/module.h>
21 #include <linux/capability.h>
22 #include <linux/errno.h>
23 #include <linux/types.h>
24 #include <linux/socket.h>
25 #include <linux/sockios.h>
26 #include <linux/net.h>
27 #include <linux/in6.h>
28 #include <linux/netdevice.h>
29 #include <linux/if_arp.h>
30 #include <linux/icmp.h>
31 #include <linux/slab.h>
32 #include <asm/uaccess.h>
33 #include <linux/init.h>
34 #include <linux/netfilter_ipv4.h>
35 #include <linux/if_ether.h>
37 #include <net/sock.h>
38 #include <net/snmp.h>
40 #include <net/ipv6.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>
47 #include <net/ip.h>
48 #include <net/udp.h>
49 #include <net/icmp.h>
50 #include <net/ipip.h>
51 #include <net/inet_ecn.h>
52 #include <net/xfrm.h>
53 #include <net/dsfield.h>
54 #include <net/net_namespace.h>
55 #include <net/netns/generic.h>
58 This version of net/ipv6/sit.c is cloned of net/ipv4/ip_gre.c
60 For comments look at net/ipv4/ip_gre.c --ANK
63 #define HASH_SIZE 16
64 #define HASH(addr) (((__force u32)addr^((__force u32)addr>>4))&0xF)
66 static int ipip6_tunnel_init(struct net_device *dev);
67 static void ipip6_tunnel_setup(struct net_device *dev);
68 static void ipip6_dev_free(struct net_device *dev);
70 static int sit_net_id __read_mostly;
71 struct sit_net {
72 struct ip_tunnel __rcu *tunnels_r_l[HASH_SIZE];
73 struct ip_tunnel __rcu *tunnels_r[HASH_SIZE];
74 struct ip_tunnel __rcu *tunnels_l[HASH_SIZE];
75 struct ip_tunnel __rcu *tunnels_wc[1];
76 struct ip_tunnel __rcu **tunnels[4];
78 struct net_device *fb_tunnel_dev;
82 * Locking : hash tables are protected by RCU and RTNL
85 #define for_each_ip_tunnel_rcu(start) \
86 for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
88 /* often modified stats are per cpu, other are shared (netdev->stats) */
89 struct pcpu_tstats {
90 unsigned long rx_packets;
91 unsigned long rx_bytes;
92 unsigned long tx_packets;
93 unsigned long tx_bytes;
96 static struct net_device_stats *ipip6_get_stats(struct net_device *dev)
98 struct pcpu_tstats sum = { 0 };
99 int i;
101 for_each_possible_cpu(i) {
102 const struct pcpu_tstats *tstats = per_cpu_ptr(dev->tstats, i);
104 sum.rx_packets += tstats->rx_packets;
105 sum.rx_bytes += tstats->rx_bytes;
106 sum.tx_packets += tstats->tx_packets;
107 sum.tx_bytes += tstats->tx_bytes;
109 dev->stats.rx_packets = sum.rx_packets;
110 dev->stats.rx_bytes = sum.rx_bytes;
111 dev->stats.tx_packets = sum.tx_packets;
112 dev->stats.tx_bytes = sum.tx_bytes;
113 return &dev->stats;
116 * Must be invoked with rcu_read_lock
118 static struct ip_tunnel * ipip6_tunnel_lookup(struct net *net,
119 struct net_device *dev, __be32 remote, __be32 local)
121 unsigned int h0 = HASH(remote);
122 unsigned int h1 = HASH(local);
123 struct ip_tunnel *t;
124 struct sit_net *sitn = net_generic(net, sit_net_id);
126 for_each_ip_tunnel_rcu(sitn->tunnels_r_l[h0 ^ h1]) {
127 if (local == t->parms.iph.saddr &&
128 remote == t->parms.iph.daddr &&
129 (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
130 (t->dev->flags & IFF_UP))
131 return t;
133 for_each_ip_tunnel_rcu(sitn->tunnels_r[h0]) {
134 if (remote == t->parms.iph.daddr &&
135 (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
136 (t->dev->flags & IFF_UP))
137 return t;
139 for_each_ip_tunnel_rcu(sitn->tunnels_l[h1]) {
140 if (local == t->parms.iph.saddr &&
141 (!dev || !t->parms.link || dev->iflink == t->parms.link) &&
142 (t->dev->flags & IFF_UP))
143 return t;
145 t = rcu_dereference(sitn->tunnels_wc[0]);
146 if ((t != NULL) && (t->dev->flags & IFF_UP))
147 return t;
148 return NULL;
151 static struct ip_tunnel __rcu **__ipip6_bucket(struct sit_net *sitn,
152 struct ip_tunnel_parm *parms)
154 __be32 remote = parms->iph.daddr;
155 __be32 local = parms->iph.saddr;
156 unsigned int h = 0;
157 int prio = 0;
159 if (remote) {
160 prio |= 2;
161 h ^= HASH(remote);
163 if (local) {
164 prio |= 1;
165 h ^= HASH(local);
167 return &sitn->tunnels[prio][h];
170 static inline struct ip_tunnel __rcu **ipip6_bucket(struct sit_net *sitn,
171 struct ip_tunnel *t)
173 return __ipip6_bucket(sitn, &t->parms);
176 static void ipip6_tunnel_unlink(struct sit_net *sitn, struct ip_tunnel *t)
178 struct ip_tunnel __rcu **tp;
179 struct ip_tunnel *iter;
181 for (tp = ipip6_bucket(sitn, t);
182 (iter = rtnl_dereference(*tp)) != NULL;
183 tp = &iter->next) {
184 if (t == iter) {
185 rcu_assign_pointer(*tp, t->next);
186 break;
191 static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t)
193 struct ip_tunnel __rcu **tp = ipip6_bucket(sitn, t);
195 rcu_assign_pointer(t->next, rtnl_dereference(*tp));
196 rcu_assign_pointer(*tp, t);
199 static void ipip6_tunnel_clone_6rd(struct net_device *dev, struct sit_net *sitn)
201 #ifdef CONFIG_IPV6_SIT_6RD
202 struct ip_tunnel *t = netdev_priv(dev);
204 if (t->dev == sitn->fb_tunnel_dev) {
205 ipv6_addr_set(&t->ip6rd.prefix, htonl(0x20020000), 0, 0, 0);
206 t->ip6rd.relay_prefix = 0;
207 t->ip6rd.prefixlen = 16;
208 t->ip6rd.relay_prefixlen = 0;
209 } else {
210 struct ip_tunnel *t0 = netdev_priv(sitn->fb_tunnel_dev);
211 memcpy(&t->ip6rd, &t0->ip6rd, sizeof(t->ip6rd));
213 #endif
216 static struct ip_tunnel *ipip6_tunnel_locate(struct net *net,
217 struct ip_tunnel_parm *parms, int create)
219 __be32 remote = parms->iph.daddr;
220 __be32 local = parms->iph.saddr;
221 struct ip_tunnel *t, *nt;
222 struct ip_tunnel __rcu **tp;
223 struct net_device *dev;
224 char name[IFNAMSIZ];
225 struct sit_net *sitn = net_generic(net, sit_net_id);
227 for (tp = __ipip6_bucket(sitn, parms);
228 (t = rtnl_dereference(*tp)) != NULL;
229 tp = &t->next) {
230 if (local == t->parms.iph.saddr &&
231 remote == t->parms.iph.daddr &&
232 parms->link == t->parms.link) {
233 if (create)
234 return NULL;
235 else
236 return t;
239 if (!create)
240 goto failed;
242 if (parms->name[0])
243 strlcpy(name, parms->name, IFNAMSIZ);
244 else
245 strcpy(name, "sit%d");
247 dev = alloc_netdev(sizeof(*t), name, ipip6_tunnel_setup);
248 if (dev == NULL)
249 return NULL;
251 dev_net_set(dev, net);
253 if (strchr(name, '%')) {
254 if (dev_alloc_name(dev, name) < 0)
255 goto failed_free;
258 nt = netdev_priv(dev);
260 nt->parms = *parms;
261 if (ipip6_tunnel_init(dev) < 0)
262 goto failed_free;
263 ipip6_tunnel_clone_6rd(dev, sitn);
265 if (parms->i_flags & SIT_ISATAP)
266 dev->priv_flags |= IFF_ISATAP;
268 if (register_netdevice(dev) < 0)
269 goto failed_free;
271 dev_hold(dev);
273 ipip6_tunnel_link(sitn, nt);
274 return nt;
276 failed_free:
277 ipip6_dev_free(dev);
278 failed:
279 return NULL;
282 #define for_each_prl_rcu(start) \
283 for (prl = rcu_dereference(start); \
284 prl; \
285 prl = rcu_dereference(prl->next))
287 static struct ip_tunnel_prl_entry *
288 __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
290 struct ip_tunnel_prl_entry *prl;
292 for_each_prl_rcu(t->prl)
293 if (prl->addr == addr)
294 break;
295 return prl;
299 static int ipip6_tunnel_get_prl(struct ip_tunnel *t,
300 struct ip_tunnel_prl __user *a)
302 struct ip_tunnel_prl kprl, *kp;
303 struct ip_tunnel_prl_entry *prl;
304 unsigned int cmax, c = 0, ca, len;
305 int ret = 0;
307 if (copy_from_user(&kprl, a, sizeof(kprl)))
308 return -EFAULT;
309 cmax = kprl.datalen / sizeof(kprl);
310 if (cmax > 1 && kprl.addr != htonl(INADDR_ANY))
311 cmax = 1;
313 /* For simple GET or for root users,
314 * we try harder to allocate.
316 kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
317 kcalloc(cmax, sizeof(*kp), GFP_KERNEL) :
318 NULL;
320 rcu_read_lock();
322 ca = t->prl_count < cmax ? t->prl_count : cmax;
324 if (!kp) {
325 /* We don't try hard to allocate much memory for
326 * non-root users.
327 * For root users, retry allocating enough memory for
328 * the answer.
330 kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC);
331 if (!kp) {
332 ret = -ENOMEM;
333 goto out;
337 c = 0;
338 for_each_prl_rcu(t->prl) {
339 if (c >= cmax)
340 break;
341 if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr)
342 continue;
343 kp[c].addr = prl->addr;
344 kp[c].flags = prl->flags;
345 c++;
346 if (kprl.addr != htonl(INADDR_ANY))
347 break;
349 out:
350 rcu_read_unlock();
352 len = sizeof(*kp) * c;
353 ret = 0;
354 if ((len && copy_to_user(a + 1, kp, len)) || put_user(len, &a->datalen))
355 ret = -EFAULT;
357 kfree(kp);
359 return ret;
362 static int
363 ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
365 struct ip_tunnel_prl_entry *p;
366 int err = 0;
368 if (a->addr == htonl(INADDR_ANY))
369 return -EINVAL;
371 ASSERT_RTNL();
373 for (p = rtnl_dereference(t->prl); p; p = rtnl_dereference(p->next)) {
374 if (p->addr == a->addr) {
375 if (chg) {
376 p->flags = a->flags;
377 goto out;
379 err = -EEXIST;
380 goto out;
384 if (chg) {
385 err = -ENXIO;
386 goto out;
389 p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
390 if (!p) {
391 err = -ENOBUFS;
392 goto out;
395 p->next = t->prl;
396 p->addr = a->addr;
397 p->flags = a->flags;
398 t->prl_count++;
399 rcu_assign_pointer(t->prl, p);
400 out:
401 return err;
404 static void prl_list_destroy_rcu(struct rcu_head *head)
406 struct ip_tunnel_prl_entry *p, *n;
408 p = container_of(head, struct ip_tunnel_prl_entry, rcu_head);
409 do {
410 n = rcu_dereference_protected(p->next, 1);
411 kfree(p);
412 p = n;
413 } while (p);
416 static int
417 ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
419 struct ip_tunnel_prl_entry *x;
420 struct ip_tunnel_prl_entry __rcu **p;
421 int err = 0;
423 ASSERT_RTNL();
425 if (a && a->addr != htonl(INADDR_ANY)) {
426 for (p = &t->prl;
427 (x = rtnl_dereference(*p)) != NULL;
428 p = &x->next) {
429 if (x->addr == a->addr) {
430 *p = x->next;
431 kfree_rcu(x, rcu_head);
432 t->prl_count--;
433 goto out;
436 err = -ENXIO;
437 } else {
438 x = rtnl_dereference(t->prl);
439 if (x) {
440 t->prl_count = 0;
441 call_rcu(&x->rcu_head, prl_list_destroy_rcu);
442 t->prl = NULL;
445 out:
446 return err;
449 static int
450 isatap_chksrc(struct sk_buff *skb, const struct iphdr *iph, struct ip_tunnel *t)
452 struct ip_tunnel_prl_entry *p;
453 int ok = 1;
455 rcu_read_lock();
456 p = __ipip6_tunnel_locate_prl(t, iph->saddr);
457 if (p) {
458 if (p->flags & PRL_DEFAULT)
459 skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
460 else
461 skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
462 } else {
463 const struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
465 if (ipv6_addr_is_isatap(addr6) &&
466 (addr6->s6_addr32[3] == iph->saddr) &&
467 ipv6_chk_prefix(addr6, t->dev))
468 skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
469 else
470 ok = 0;
472 rcu_read_unlock();
473 return ok;
476 static void ipip6_tunnel_uninit(struct net_device *dev)
478 struct net *net = dev_net(dev);
479 struct sit_net *sitn = net_generic(net, sit_net_id);
481 if (dev == sitn->fb_tunnel_dev) {
482 rcu_assign_pointer(sitn->tunnels_wc[0], NULL);
483 } else {
484 ipip6_tunnel_unlink(sitn, netdev_priv(dev));
485 ipip6_tunnel_del_prl(netdev_priv(dev), NULL);
487 dev_put(dev);
491 static int ipip6_err(struct sk_buff *skb, u32 info)
494 /* All the routers (except for Linux) return only
495 8 bytes of packet payload. It means, that precise relaying of
496 ICMP in the real Internet is absolutely infeasible.
498 const struct iphdr *iph = (const struct iphdr *)skb->data;
499 const int type = icmp_hdr(skb)->type;
500 const int code = icmp_hdr(skb)->code;
501 struct ip_tunnel *t;
502 int err;
504 switch (type) {
505 default:
506 case ICMP_PARAMETERPROB:
507 return 0;
509 case ICMP_DEST_UNREACH:
510 switch (code) {
511 case ICMP_SR_FAILED:
512 case ICMP_PORT_UNREACH:
513 /* Impossible event. */
514 return 0;
515 case ICMP_FRAG_NEEDED:
516 /* Soft state for pmtu is maintained by IP core. */
517 return 0;
518 default:
519 /* All others are translated to HOST_UNREACH.
520 rfc2003 contains "deep thoughts" about NET_UNREACH,
521 I believe they are just ether pollution. --ANK
523 break;
525 break;
526 case ICMP_TIME_EXCEEDED:
527 if (code != ICMP_EXC_TTL)
528 return 0;
529 break;
532 err = -ENOENT;
534 rcu_read_lock();
535 t = ipip6_tunnel_lookup(dev_net(skb->dev),
536 skb->dev,
537 iph->daddr,
538 iph->saddr);
539 if (t == NULL || t->parms.iph.daddr == 0)
540 goto out;
542 err = 0;
543 if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
544 goto out;
546 if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
547 t->err_count++;
548 else
549 t->err_count = 1;
550 t->err_time = jiffies;
551 out:
552 rcu_read_unlock();
553 return err;
556 static inline void ipip6_ecn_decapsulate(const struct iphdr *iph, struct sk_buff *skb)
558 if (INET_ECN_is_ce(iph->tos))
559 IP6_ECN_set_ce(ipv6_hdr(skb));
562 static int ipip6_rcv(struct sk_buff *skb)
564 const struct iphdr *iph;
565 struct ip_tunnel *tunnel;
567 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
568 goto out;
570 iph = ip_hdr(skb);
572 rcu_read_lock();
573 tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
574 iph->saddr, iph->daddr);
575 if (tunnel != NULL) {
576 struct pcpu_tstats *tstats;
578 secpath_reset(skb);
579 skb->mac_header = skb->network_header;
580 skb_reset_network_header(skb);
581 IPCB(skb)->flags = 0;
582 skb->protocol = htons(ETH_P_IPV6);
583 skb->pkt_type = PACKET_HOST;
585 if ((tunnel->dev->priv_flags & IFF_ISATAP) &&
586 !isatap_chksrc(skb, iph, tunnel)) {
587 tunnel->dev->stats.rx_errors++;
588 rcu_read_unlock();
589 kfree_skb(skb);
590 return 0;
593 tstats = this_cpu_ptr(tunnel->dev->tstats);
594 tstats->rx_packets++;
595 tstats->rx_bytes += skb->len;
597 __skb_tunnel_rx(skb, tunnel->dev);
599 ipip6_ecn_decapsulate(iph, skb);
601 netif_rx(skb);
603 rcu_read_unlock();
604 return 0;
607 /* no tunnel matched, let upstream know, ipsec may handle it */
608 rcu_read_unlock();
609 return 1;
610 out:
611 kfree_skb(skb);
612 return 0;
616 * Returns the embedded IPv4 address if the IPv6 address
617 * comes from 6rd / 6to4 (RFC 3056) addr space.
619 static inline
620 __be32 try_6rd(const struct in6_addr *v6dst, struct ip_tunnel *tunnel)
622 __be32 dst = 0;
624 #ifdef CONFIG_IPV6_SIT_6RD
625 if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
626 tunnel->ip6rd.prefixlen)) {
627 unsigned int pbw0, pbi0;
628 int pbi1;
629 u32 d;
631 pbw0 = tunnel->ip6rd.prefixlen >> 5;
632 pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
634 d = (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
635 tunnel->ip6rd.relay_prefixlen;
637 pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
638 if (pbi1 > 0)
639 d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
640 (32 - pbi1);
642 dst = tunnel->ip6rd.relay_prefix | htonl(d);
644 #else
645 if (v6dst->s6_addr16[0] == htons(0x2002)) {
646 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
647 memcpy(&dst, &v6dst->s6_addr16[1], 4);
649 #endif
650 return dst;
654 * This function assumes it is being called from dev_queue_xmit()
655 * and that skb is filled properly by that function.
658 static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
659 struct net_device *dev)
661 struct ip_tunnel *tunnel = netdev_priv(dev);
662 struct pcpu_tstats *tstats;
663 const struct iphdr *tiph = &tunnel->parms.iph;
664 const struct ipv6hdr *iph6 = ipv6_hdr(skb);
665 u8 tos = tunnel->parms.iph.tos;
666 __be16 df = tiph->frag_off;
667 struct rtable *rt; /* Route to the other host */
668 struct net_device *tdev; /* Device to other host */
669 struct iphdr *iph; /* Our new IP header */
670 unsigned int max_headroom; /* The extra header space needed */
671 __be32 dst = tiph->daddr;
672 int mtu;
673 const struct in6_addr *addr6;
674 int addr_type;
676 if (skb->protocol != htons(ETH_P_IPV6))
677 goto tx_error;
679 /* ISATAP (RFC4214) - must come before 6to4 */
680 if (dev->priv_flags & IFF_ISATAP) {
681 struct neighbour *neigh = NULL;
683 if (skb_dst(skb))
684 neigh = skb_dst(skb)->neighbour;
686 if (neigh == NULL) {
687 if (net_ratelimit())
688 printk(KERN_DEBUG "sit: nexthop == NULL\n");
689 goto tx_error;
692 addr6 = (const struct in6_addr*)&neigh->primary_key;
693 addr_type = ipv6_addr_type(addr6);
695 if ((addr_type & IPV6_ADDR_UNICAST) &&
696 ipv6_addr_is_isatap(addr6))
697 dst = addr6->s6_addr32[3];
698 else
699 goto tx_error;
702 if (!dst)
703 dst = try_6rd(&iph6->daddr, tunnel);
705 if (!dst) {
706 struct neighbour *neigh = NULL;
708 if (skb_dst(skb))
709 neigh = skb_dst(skb)->neighbour;
711 if (neigh == NULL) {
712 if (net_ratelimit())
713 printk(KERN_DEBUG "sit: nexthop == NULL\n");
714 goto tx_error;
717 addr6 = (const struct in6_addr*)&neigh->primary_key;
718 addr_type = ipv6_addr_type(addr6);
720 if (addr_type == IPV6_ADDR_ANY) {
721 addr6 = &ipv6_hdr(skb)->daddr;
722 addr_type = ipv6_addr_type(addr6);
725 if ((addr_type & IPV6_ADDR_COMPATv4) == 0)
726 goto tx_error_icmp;
728 dst = addr6->s6_addr32[3];
731 rt = ip_route_output_ports(dev_net(dev), NULL,
732 dst, tiph->saddr,
733 0, 0,
734 IPPROTO_IPV6, RT_TOS(tos),
735 tunnel->parms.link);
736 if (IS_ERR(rt)) {
737 dev->stats.tx_carrier_errors++;
738 goto tx_error_icmp;
740 if (rt->rt_type != RTN_UNICAST) {
741 ip_rt_put(rt);
742 dev->stats.tx_carrier_errors++;
743 goto tx_error_icmp;
745 tdev = rt->dst.dev;
747 if (tdev == dev) {
748 ip_rt_put(rt);
749 dev->stats.collisions++;
750 goto tx_error;
753 if (df) {
754 mtu = dst_mtu(&rt->dst) - sizeof(struct iphdr);
756 if (mtu < 68) {
757 dev->stats.collisions++;
758 ip_rt_put(rt);
759 goto tx_error;
762 if (mtu < IPV6_MIN_MTU) {
763 mtu = IPV6_MIN_MTU;
764 df = 0;
767 if (tunnel->parms.iph.daddr && skb_dst(skb))
768 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), mtu);
770 if (skb->len > mtu) {
771 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
772 ip_rt_put(rt);
773 goto tx_error;
777 if (tunnel->err_count > 0) {
778 if (time_before(jiffies,
779 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
780 tunnel->err_count--;
781 dst_link_failure(skb);
782 } else
783 tunnel->err_count = 0;
787 * Okay, now see if we can stuff it in the buffer as-is.
789 max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr);
791 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
792 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
793 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
794 if (!new_skb) {
795 ip_rt_put(rt);
796 dev->stats.tx_dropped++;
797 dev_kfree_skb(skb);
798 return NETDEV_TX_OK;
800 if (skb->sk)
801 skb_set_owner_w(new_skb, skb->sk);
802 dev_kfree_skb(skb);
803 skb = new_skb;
804 iph6 = ipv6_hdr(skb);
807 skb->transport_header = skb->network_header;
808 skb_push(skb, sizeof(struct iphdr));
809 skb_reset_network_header(skb);
810 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
811 IPCB(skb)->flags = 0;
812 skb_dst_drop(skb);
813 skb_dst_set(skb, &rt->dst);
816 * Push down and install the IPIP header.
819 iph = ip_hdr(skb);
820 iph->version = 4;
821 iph->ihl = sizeof(struct iphdr)>>2;
822 iph->frag_off = df;
823 iph->protocol = IPPROTO_IPV6;
824 iph->tos = INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
825 iph->daddr = rt->rt_dst;
826 iph->saddr = rt->rt_src;
828 if ((iph->ttl = tiph->ttl) == 0)
829 iph->ttl = iph6->hop_limit;
831 nf_reset(skb);
832 tstats = this_cpu_ptr(dev->tstats);
833 __IPTUNNEL_XMIT(tstats, &dev->stats);
834 return NETDEV_TX_OK;
836 tx_error_icmp:
837 dst_link_failure(skb);
838 tx_error:
839 dev->stats.tx_errors++;
840 dev_kfree_skb(skb);
841 return NETDEV_TX_OK;
844 static void ipip6_tunnel_bind_dev(struct net_device *dev)
846 struct net_device *tdev = NULL;
847 struct ip_tunnel *tunnel;
848 const struct iphdr *iph;
850 tunnel = netdev_priv(dev);
851 iph = &tunnel->parms.iph;
853 if (iph->daddr) {
854 struct rtable *rt = ip_route_output_ports(dev_net(dev), NULL,
855 iph->daddr, iph->saddr,
856 0, 0,
857 IPPROTO_IPV6,
858 RT_TOS(iph->tos),
859 tunnel->parms.link);
861 if (!IS_ERR(rt)) {
862 tdev = rt->dst.dev;
863 ip_rt_put(rt);
865 dev->flags |= IFF_POINTOPOINT;
868 if (!tdev && tunnel->parms.link)
869 tdev = __dev_get_by_index(dev_net(dev), tunnel->parms.link);
871 if (tdev) {
872 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
873 dev->mtu = tdev->mtu - sizeof(struct iphdr);
874 if (dev->mtu < IPV6_MIN_MTU)
875 dev->mtu = IPV6_MIN_MTU;
877 dev->iflink = tunnel->parms.link;
880 static int
881 ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
883 int err = 0;
884 struct ip_tunnel_parm p;
885 struct ip_tunnel_prl prl;
886 struct ip_tunnel *t;
887 struct net *net = dev_net(dev);
888 struct sit_net *sitn = net_generic(net, sit_net_id);
889 #ifdef CONFIG_IPV6_SIT_6RD
890 struct ip_tunnel_6rd ip6rd;
891 #endif
893 switch (cmd) {
894 case SIOCGETTUNNEL:
895 #ifdef CONFIG_IPV6_SIT_6RD
896 case SIOCGET6RD:
897 #endif
898 t = NULL;
899 if (dev == sitn->fb_tunnel_dev) {
900 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
901 err = -EFAULT;
902 break;
904 t = ipip6_tunnel_locate(net, &p, 0);
906 if (t == NULL)
907 t = netdev_priv(dev);
909 err = -EFAULT;
910 if (cmd == SIOCGETTUNNEL) {
911 memcpy(&p, &t->parms, sizeof(p));
912 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p,
913 sizeof(p)))
914 goto done;
915 #ifdef CONFIG_IPV6_SIT_6RD
916 } else {
917 ipv6_addr_copy(&ip6rd.prefix, &t->ip6rd.prefix);
918 ip6rd.relay_prefix = t->ip6rd.relay_prefix;
919 ip6rd.prefixlen = t->ip6rd.prefixlen;
920 ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
921 if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd,
922 sizeof(ip6rd)))
923 goto done;
924 #endif
926 err = 0;
927 break;
929 case SIOCADDTUNNEL:
930 case SIOCCHGTUNNEL:
931 err = -EPERM;
932 if (!capable(CAP_NET_ADMIN))
933 goto done;
935 err = -EFAULT;
936 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
937 goto done;
939 err = -EINVAL;
940 if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPV6 ||
941 p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
942 goto done;
943 if (p.iph.ttl)
944 p.iph.frag_off |= htons(IP_DF);
946 t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
948 if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
949 if (t != NULL) {
950 if (t->dev != dev) {
951 err = -EEXIST;
952 break;
954 } else {
955 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
956 (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
957 err = -EINVAL;
958 break;
960 t = netdev_priv(dev);
961 ipip6_tunnel_unlink(sitn, t);
962 synchronize_net();
963 t->parms.iph.saddr = p.iph.saddr;
964 t->parms.iph.daddr = p.iph.daddr;
965 memcpy(dev->dev_addr, &p.iph.saddr, 4);
966 memcpy(dev->broadcast, &p.iph.daddr, 4);
967 ipip6_tunnel_link(sitn, t);
968 netdev_state_change(dev);
972 if (t) {
973 err = 0;
974 if (cmd == SIOCCHGTUNNEL) {
975 t->parms.iph.ttl = p.iph.ttl;
976 t->parms.iph.tos = p.iph.tos;
977 if (t->parms.link != p.link) {
978 t->parms.link = p.link;
979 ipip6_tunnel_bind_dev(dev);
980 netdev_state_change(dev);
983 if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
984 err = -EFAULT;
985 } else
986 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
987 break;
989 case SIOCDELTUNNEL:
990 err = -EPERM;
991 if (!capable(CAP_NET_ADMIN))
992 goto done;
994 if (dev == sitn->fb_tunnel_dev) {
995 err = -EFAULT;
996 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
997 goto done;
998 err = -ENOENT;
999 if ((t = ipip6_tunnel_locate(net, &p, 0)) == NULL)
1000 goto done;
1001 err = -EPERM;
1002 if (t == netdev_priv(sitn->fb_tunnel_dev))
1003 goto done;
1004 dev = t->dev;
1006 unregister_netdevice(dev);
1007 err = 0;
1008 break;
1010 case SIOCGETPRL:
1011 err = -EINVAL;
1012 if (dev == sitn->fb_tunnel_dev)
1013 goto done;
1014 err = -ENOENT;
1015 if (!(t = netdev_priv(dev)))
1016 goto done;
1017 err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
1018 break;
1020 case SIOCADDPRL:
1021 case SIOCDELPRL:
1022 case SIOCCHGPRL:
1023 err = -EPERM;
1024 if (!capable(CAP_NET_ADMIN))
1025 goto done;
1026 err = -EINVAL;
1027 if (dev == sitn->fb_tunnel_dev)
1028 goto done;
1029 err = -EFAULT;
1030 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
1031 goto done;
1032 err = -ENOENT;
1033 if (!(t = netdev_priv(dev)))
1034 goto done;
1036 switch (cmd) {
1037 case SIOCDELPRL:
1038 err = ipip6_tunnel_del_prl(t, &prl);
1039 break;
1040 case SIOCADDPRL:
1041 case SIOCCHGPRL:
1042 err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
1043 break;
1045 netdev_state_change(dev);
1046 break;
1048 #ifdef CONFIG_IPV6_SIT_6RD
1049 case SIOCADD6RD:
1050 case SIOCCHG6RD:
1051 case SIOCDEL6RD:
1052 err = -EPERM;
1053 if (!capable(CAP_NET_ADMIN))
1054 goto done;
1056 err = -EFAULT;
1057 if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data,
1058 sizeof(ip6rd)))
1059 goto done;
1061 t = netdev_priv(dev);
1063 if (cmd != SIOCDEL6RD) {
1064 struct in6_addr prefix;
1065 __be32 relay_prefix;
1067 err = -EINVAL;
1068 if (ip6rd.relay_prefixlen > 32 ||
1069 ip6rd.prefixlen + (32 - ip6rd.relay_prefixlen) > 64)
1070 goto done;
1072 ipv6_addr_prefix(&prefix, &ip6rd.prefix,
1073 ip6rd.prefixlen);
1074 if (!ipv6_addr_equal(&prefix, &ip6rd.prefix))
1075 goto done;
1076 if (ip6rd.relay_prefixlen)
1077 relay_prefix = ip6rd.relay_prefix &
1078 htonl(0xffffffffUL <<
1079 (32 - ip6rd.relay_prefixlen));
1080 else
1081 relay_prefix = 0;
1082 if (relay_prefix != ip6rd.relay_prefix)
1083 goto done;
1085 ipv6_addr_copy(&t->ip6rd.prefix, &prefix);
1086 t->ip6rd.relay_prefix = relay_prefix;
1087 t->ip6rd.prefixlen = ip6rd.prefixlen;
1088 t->ip6rd.relay_prefixlen = ip6rd.relay_prefixlen;
1089 } else
1090 ipip6_tunnel_clone_6rd(dev, sitn);
1092 err = 0;
1093 break;
1094 #endif
1096 default:
1097 err = -EINVAL;
1100 done:
1101 return err;
1104 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1106 if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
1107 return -EINVAL;
1108 dev->mtu = new_mtu;
1109 return 0;
1112 static const struct net_device_ops ipip6_netdev_ops = {
1113 .ndo_uninit = ipip6_tunnel_uninit,
1114 .ndo_start_xmit = ipip6_tunnel_xmit,
1115 .ndo_do_ioctl = ipip6_tunnel_ioctl,
1116 .ndo_change_mtu = ipip6_tunnel_change_mtu,
1117 .ndo_get_stats = ipip6_get_stats,
1120 static void ipip6_dev_free(struct net_device *dev)
1122 free_percpu(dev->tstats);
1123 free_netdev(dev);
1126 static void ipip6_tunnel_setup(struct net_device *dev)
1128 dev->netdev_ops = &ipip6_netdev_ops;
1129 dev->destructor = ipip6_dev_free;
1131 dev->type = ARPHRD_SIT;
1132 dev->hard_header_len = LL_MAX_HEADER + sizeof(struct iphdr);
1133 dev->mtu = ETH_DATA_LEN - sizeof(struct iphdr);
1134 dev->flags = IFF_NOARP;
1135 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
1136 dev->iflink = 0;
1137 dev->addr_len = 4;
1138 dev->features |= NETIF_F_NETNS_LOCAL;
1139 dev->features |= NETIF_F_LLTX;
1142 static int ipip6_tunnel_init(struct net_device *dev)
1144 struct ip_tunnel *tunnel = netdev_priv(dev);
1146 tunnel->dev = dev;
1147 strcpy(tunnel->parms.name, dev->name);
1149 memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
1150 memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
1152 ipip6_tunnel_bind_dev(dev);
1153 dev->tstats = alloc_percpu(struct pcpu_tstats);
1154 if (!dev->tstats)
1155 return -ENOMEM;
1157 return 0;
1160 static int __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1162 struct ip_tunnel *tunnel = netdev_priv(dev);
1163 struct iphdr *iph = &tunnel->parms.iph;
1164 struct net *net = dev_net(dev);
1165 struct sit_net *sitn = net_generic(net, sit_net_id);
1167 tunnel->dev = dev;
1168 strcpy(tunnel->parms.name, dev->name);
1170 iph->version = 4;
1171 iph->protocol = IPPROTO_IPV6;
1172 iph->ihl = 5;
1173 iph->ttl = 64;
1175 dev->tstats = alloc_percpu(struct pcpu_tstats);
1176 if (!dev->tstats)
1177 return -ENOMEM;
1178 dev_hold(dev);
1179 rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1180 return 0;
1183 static struct xfrm_tunnel sit_handler __read_mostly = {
1184 .handler = ipip6_rcv,
1185 .err_handler = ipip6_err,
1186 .priority = 1,
1189 static void __net_exit sit_destroy_tunnels(struct sit_net *sitn, struct list_head *head)
1191 int prio;
1193 for (prio = 1; prio < 4; prio++) {
1194 int h;
1195 for (h = 0; h < HASH_SIZE; h++) {
1196 struct ip_tunnel *t;
1198 t = rtnl_dereference(sitn->tunnels[prio][h]);
1199 while (t != NULL) {
1200 unregister_netdevice_queue(t->dev, head);
1201 t = rtnl_dereference(t->next);
1207 static int __net_init sit_init_net(struct net *net)
1209 struct sit_net *sitn = net_generic(net, sit_net_id);
1210 int err;
1212 sitn->tunnels[0] = sitn->tunnels_wc;
1213 sitn->tunnels[1] = sitn->tunnels_l;
1214 sitn->tunnels[2] = sitn->tunnels_r;
1215 sitn->tunnels[3] = sitn->tunnels_r_l;
1217 sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1218 ipip6_tunnel_setup);
1219 if (!sitn->fb_tunnel_dev) {
1220 err = -ENOMEM;
1221 goto err_alloc_dev;
1223 dev_net_set(sitn->fb_tunnel_dev, net);
1225 err = ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1226 if (err)
1227 goto err_dev_free;
1229 ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1231 if ((err = register_netdev(sitn->fb_tunnel_dev)))
1232 goto err_reg_dev;
1234 return 0;
1236 err_reg_dev:
1237 dev_put(sitn->fb_tunnel_dev);
1238 err_dev_free:
1239 ipip6_dev_free(sitn->fb_tunnel_dev);
1240 err_alloc_dev:
1241 return err;
1244 static void __net_exit sit_exit_net(struct net *net)
1246 struct sit_net *sitn = net_generic(net, sit_net_id);
1247 LIST_HEAD(list);
1249 rtnl_lock();
1250 sit_destroy_tunnels(sitn, &list);
1251 unregister_netdevice_queue(sitn->fb_tunnel_dev, &list);
1252 unregister_netdevice_many(&list);
1253 rtnl_unlock();
1256 static struct pernet_operations sit_net_ops = {
1257 .init = sit_init_net,
1258 .exit = sit_exit_net,
1259 .id = &sit_net_id,
1260 .size = sizeof(struct sit_net),
1263 static void __exit sit_cleanup(void)
1265 xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1267 unregister_pernet_device(&sit_net_ops);
1268 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1271 static int __init sit_init(void)
1273 int err;
1275 printk(KERN_INFO "IPv6 over IPv4 tunneling driver\n");
1277 err = register_pernet_device(&sit_net_ops);
1278 if (err < 0)
1279 return err;
1280 err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1281 if (err < 0) {
1282 unregister_pernet_device(&sit_net_ops);
1283 printk(KERN_INFO "sit init: Can't add protocol\n");
1285 return err;
1288 module_init(sit_init);
1289 module_exit(sit_cleanup);
1290 MODULE_LICENSE("GPL");
1291 MODULE_ALIAS_NETDEV("sit0");