ipv4: Create and use route lookup helpers.
[linux-3.10.git] / net / ipv6 / sit.c
1 /*
2  *      IPv6 over IPv4 tunnel device - Simple Internet Transition (SIT)
3  *      Linux INET6 implementation
4  *
5  *      Authors:
6  *      Pedro Roque             <roque@di.fc.ul.pt>
7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
8  *
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.
13  *
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
18  */
19
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>
36
37 #include <net/sock.h>
38 #include <net/snmp.h>
39
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>
56
57 /*
58    This version of net/ipv6/sit.c is cloned of net/ipv4/ip_gre.c
59
60    For comments look at net/ipv4/ip_gre.c --ANK
61  */
62
63 #define HASH_SIZE  16
64 #define HASH(addr) (((__force u32)addr^((__force u32)addr>>4))&0xF)
65
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);
69
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];
77
78         struct net_device *fb_tunnel_dev;
79 };
80
81 /*
82  * Locking : hash tables are protected by RCU and RTNL
83  */
84
85 #define for_each_ip_tunnel_rcu(start) \
86         for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
87
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;
94 };
95
96 static struct net_device_stats *ipip6_get_stats(struct net_device *dev)
97 {
98         struct pcpu_tstats sum = { 0 };
99         int i;
100
101         for_each_possible_cpu(i) {
102                 const struct pcpu_tstats *tstats = per_cpu_ptr(dev->tstats, i);
103
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;
108         }
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;
114 }
115 /*
116  * Must be invoked with rcu_read_lock
117  */
118 static struct ip_tunnel * ipip6_tunnel_lookup(struct net *net,
119                 struct net_device *dev, __be32 remote, __be32 local)
120 {
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);
125
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;
132         }
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;
138         }
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;
144         }
145         t = rcu_dereference(sitn->tunnels_wc[0]);
146         if ((t != NULL) && (t->dev->flags & IFF_UP))
147                 return t;
148         return NULL;
149 }
150
151 static struct ip_tunnel __rcu **__ipip6_bucket(struct sit_net *sitn,
152                 struct ip_tunnel_parm *parms)
153 {
154         __be32 remote = parms->iph.daddr;
155         __be32 local = parms->iph.saddr;
156         unsigned int h = 0;
157         int prio = 0;
158
159         if (remote) {
160                 prio |= 2;
161                 h ^= HASH(remote);
162         }
163         if (local) {
164                 prio |= 1;
165                 h ^= HASH(local);
166         }
167         return &sitn->tunnels[prio][h];
168 }
169
170 static inline struct ip_tunnel __rcu **ipip6_bucket(struct sit_net *sitn,
171                 struct ip_tunnel *t)
172 {
173         return __ipip6_bucket(sitn, &t->parms);
174 }
175
176 static void ipip6_tunnel_unlink(struct sit_net *sitn, struct ip_tunnel *t)
177 {
178         struct ip_tunnel __rcu **tp;
179         struct ip_tunnel *iter;
180
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;
187                 }
188         }
189 }
190
191 static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t)
192 {
193         struct ip_tunnel __rcu **tp = ipip6_bucket(sitn, t);
194
195         rcu_assign_pointer(t->next, rtnl_dereference(*tp));
196         rcu_assign_pointer(*tp, t);
197 }
198
199 static void ipip6_tunnel_clone_6rd(struct net_device *dev, struct sit_net *sitn)
200 {
201 #ifdef CONFIG_IPV6_SIT_6RD
202         struct ip_tunnel *t = netdev_priv(dev);
203
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));
212         }
213 #endif
214 }
215
216 static struct ip_tunnel *ipip6_tunnel_locate(struct net *net,
217                 struct ip_tunnel_parm *parms, int create)
218 {
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);
226
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;
237                 }
238         }
239         if (!create)
240                 goto failed;
241
242         if (parms->name[0])
243                 strlcpy(name, parms->name, IFNAMSIZ);
244         else
245                 strcpy(name, "sit%d");
246
247         dev = alloc_netdev(sizeof(*t), name, ipip6_tunnel_setup);
248         if (dev == NULL)
249                 return NULL;
250
251         dev_net_set(dev, net);
252
253         if (strchr(name, '%')) {
254                 if (dev_alloc_name(dev, name) < 0)
255                         goto failed_free;
256         }
257
258         nt = netdev_priv(dev);
259
260         nt->parms = *parms;
261         if (ipip6_tunnel_init(dev) < 0)
262                 goto failed_free;
263         ipip6_tunnel_clone_6rd(dev, sitn);
264
265         if (parms->i_flags & SIT_ISATAP)
266                 dev->priv_flags |= IFF_ISATAP;
267
268         if (register_netdevice(dev) < 0)
269                 goto failed_free;
270
271         dev_hold(dev);
272
273         ipip6_tunnel_link(sitn, nt);
274         return nt;
275
276 failed_free:
277         ipip6_dev_free(dev);
278 failed:
279         return NULL;
280 }
281
282 #define for_each_prl_rcu(start)                 \
283         for (prl = rcu_dereference(start);      \
284              prl;                               \
285              prl = rcu_dereference(prl->next))
286
287 static struct ip_tunnel_prl_entry *
288 __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
289 {
290         struct ip_tunnel_prl_entry *prl;
291
292         for_each_prl_rcu(t->prl)
293                 if (prl->addr == addr)
294                         break;
295         return prl;
296
297 }
298
299 static int ipip6_tunnel_get_prl(struct ip_tunnel *t,
300                                 struct ip_tunnel_prl __user *a)
301 {
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;
306
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;
312
313         /* For simple GET or for root users,
314          * we try harder to allocate.
315          */
316         kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
317                 kcalloc(cmax, sizeof(*kp), GFP_KERNEL) :
318                 NULL;
319
320         rcu_read_lock();
321
322         ca = t->prl_count < cmax ? t->prl_count : cmax;
323
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.
329                  */
330                 kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC);
331                 if (!kp) {
332                         ret = -ENOMEM;
333                         goto out;
334                 }
335         }
336
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;
348         }
349 out:
350         rcu_read_unlock();
351
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;
356
357         kfree(kp);
358
359         return ret;
360 }
361
362 static int
363 ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
364 {
365         struct ip_tunnel_prl_entry *p;
366         int err = 0;
367
368         if (a->addr == htonl(INADDR_ANY))
369                 return -EINVAL;
370
371         ASSERT_RTNL();
372
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;
378                         }
379                         err = -EEXIST;
380                         goto out;
381                 }
382         }
383
384         if (chg) {
385                 err = -ENXIO;
386                 goto out;
387         }
388
389         p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
390         if (!p) {
391                 err = -ENOBUFS;
392                 goto out;
393         }
394
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;
402 }
403
404 static void prl_entry_destroy_rcu(struct rcu_head *head)
405 {
406         kfree(container_of(head, struct ip_tunnel_prl_entry, rcu_head));
407 }
408
409 static void prl_list_destroy_rcu(struct rcu_head *head)
410 {
411         struct ip_tunnel_prl_entry *p, *n;
412
413         p = container_of(head, struct ip_tunnel_prl_entry, rcu_head);
414         do {
415                 n = rcu_dereference_protected(p->next, 1);
416                 kfree(p);
417                 p = n;
418         } while (p);
419 }
420
421 static int
422 ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
423 {
424         struct ip_tunnel_prl_entry *x;
425         struct ip_tunnel_prl_entry __rcu **p;
426         int err = 0;
427
428         ASSERT_RTNL();
429
430         if (a && a->addr != htonl(INADDR_ANY)) {
431                 for (p = &t->prl;
432                      (x = rtnl_dereference(*p)) != NULL;
433                      p = &x->next) {
434                         if (x->addr == a->addr) {
435                                 *p = x->next;
436                                 call_rcu(&x->rcu_head, prl_entry_destroy_rcu);
437                                 t->prl_count--;
438                                 goto out;
439                         }
440                 }
441                 err = -ENXIO;
442         } else {
443                 x = rtnl_dereference(t->prl);
444                 if (x) {
445                         t->prl_count = 0;
446                         call_rcu(&x->rcu_head, prl_list_destroy_rcu);
447                         t->prl = NULL;
448                 }
449         }
450 out:
451         return err;
452 }
453
454 static int
455 isatap_chksrc(struct sk_buff *skb, struct iphdr *iph, struct ip_tunnel *t)
456 {
457         struct ip_tunnel_prl_entry *p;
458         int ok = 1;
459
460         rcu_read_lock();
461         p = __ipip6_tunnel_locate_prl(t, iph->saddr);
462         if (p) {
463                 if (p->flags & PRL_DEFAULT)
464                         skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
465                 else
466                         skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
467         } else {
468                 struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
469                 if (ipv6_addr_is_isatap(addr6) &&
470                     (addr6->s6_addr32[3] == iph->saddr) &&
471                     ipv6_chk_prefix(addr6, t->dev))
472                         skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
473                 else
474                         ok = 0;
475         }
476         rcu_read_unlock();
477         return ok;
478 }
479
480 static void ipip6_tunnel_uninit(struct net_device *dev)
481 {
482         struct net *net = dev_net(dev);
483         struct sit_net *sitn = net_generic(net, sit_net_id);
484
485         if (dev == sitn->fb_tunnel_dev) {
486                 rcu_assign_pointer(sitn->tunnels_wc[0], NULL);
487         } else {
488                 ipip6_tunnel_unlink(sitn, netdev_priv(dev));
489                 ipip6_tunnel_del_prl(netdev_priv(dev), NULL);
490         }
491         dev_put(dev);
492 }
493
494
495 static int ipip6_err(struct sk_buff *skb, u32 info)
496 {
497
498 /* All the routers (except for Linux) return only
499    8 bytes of packet payload. It means, that precise relaying of
500    ICMP in the real Internet is absolutely infeasible.
501  */
502         struct iphdr *iph = (struct iphdr*)skb->data;
503         const int type = icmp_hdr(skb)->type;
504         const int code = icmp_hdr(skb)->code;
505         struct ip_tunnel *t;
506         int err;
507
508         switch (type) {
509         default:
510         case ICMP_PARAMETERPROB:
511                 return 0;
512
513         case ICMP_DEST_UNREACH:
514                 switch (code) {
515                 case ICMP_SR_FAILED:
516                 case ICMP_PORT_UNREACH:
517                         /* Impossible event. */
518                         return 0;
519                 case ICMP_FRAG_NEEDED:
520                         /* Soft state for pmtu is maintained by IP core. */
521                         return 0;
522                 default:
523                         /* All others are translated to HOST_UNREACH.
524                            rfc2003 contains "deep thoughts" about NET_UNREACH,
525                            I believe they are just ether pollution. --ANK
526                          */
527                         break;
528                 }
529                 break;
530         case ICMP_TIME_EXCEEDED:
531                 if (code != ICMP_EXC_TTL)
532                         return 0;
533                 break;
534         }
535
536         err = -ENOENT;
537
538         rcu_read_lock();
539         t = ipip6_tunnel_lookup(dev_net(skb->dev),
540                                 skb->dev,
541                                 iph->daddr,
542                                 iph->saddr);
543         if (t == NULL || t->parms.iph.daddr == 0)
544                 goto out;
545
546         err = 0;
547         if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
548                 goto out;
549
550         if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
551                 t->err_count++;
552         else
553                 t->err_count = 1;
554         t->err_time = jiffies;
555 out:
556         rcu_read_unlock();
557         return err;
558 }
559
560 static inline void ipip6_ecn_decapsulate(struct iphdr *iph, struct sk_buff *skb)
561 {
562         if (INET_ECN_is_ce(iph->tos))
563                 IP6_ECN_set_ce(ipv6_hdr(skb));
564 }
565
566 static int ipip6_rcv(struct sk_buff *skb)
567 {
568         struct iphdr *iph;
569         struct ip_tunnel *tunnel;
570
571         if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
572                 goto out;
573
574         iph = ip_hdr(skb);
575
576         rcu_read_lock();
577         tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
578                                      iph->saddr, iph->daddr);
579         if (tunnel != NULL) {
580                 struct pcpu_tstats *tstats;
581
582                 secpath_reset(skb);
583                 skb->mac_header = skb->network_header;
584                 skb_reset_network_header(skb);
585                 IPCB(skb)->flags = 0;
586                 skb->protocol = htons(ETH_P_IPV6);
587                 skb->pkt_type = PACKET_HOST;
588
589                 if ((tunnel->dev->priv_flags & IFF_ISATAP) &&
590                     !isatap_chksrc(skb, iph, tunnel)) {
591                         tunnel->dev->stats.rx_errors++;
592                         rcu_read_unlock();
593                         kfree_skb(skb);
594                         return 0;
595                 }
596
597                 tstats = this_cpu_ptr(tunnel->dev->tstats);
598                 tstats->rx_packets++;
599                 tstats->rx_bytes += skb->len;
600
601                 __skb_tunnel_rx(skb, tunnel->dev);
602
603                 ipip6_ecn_decapsulate(iph, skb);
604
605                 netif_rx(skb);
606
607                 rcu_read_unlock();
608                 return 0;
609         }
610
611         /* no tunnel matched,  let upstream know, ipsec may handle it */
612         rcu_read_unlock();
613         return 1;
614 out:
615         kfree_skb(skb);
616         return 0;
617 }
618
619 /*
620  * Returns the embedded IPv4 address if the IPv6 address
621  * comes from 6rd / 6to4 (RFC 3056) addr space.
622  */
623 static inline
624 __be32 try_6rd(struct in6_addr *v6dst, struct ip_tunnel *tunnel)
625 {
626         __be32 dst = 0;
627
628 #ifdef CONFIG_IPV6_SIT_6RD
629         if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
630                               tunnel->ip6rd.prefixlen)) {
631                 unsigned int pbw0, pbi0;
632                 int pbi1;
633                 u32 d;
634
635                 pbw0 = tunnel->ip6rd.prefixlen >> 5;
636                 pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
637
638                 d = (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
639                     tunnel->ip6rd.relay_prefixlen;
640
641                 pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
642                 if (pbi1 > 0)
643                         d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
644                              (32 - pbi1);
645
646                 dst = tunnel->ip6rd.relay_prefix | htonl(d);
647         }
648 #else
649         if (v6dst->s6_addr16[0] == htons(0x2002)) {
650                 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
651                 memcpy(&dst, &v6dst->s6_addr16[1], 4);
652         }
653 #endif
654         return dst;
655 }
656
657 /*
658  *      This function assumes it is being called from dev_queue_xmit()
659  *      and that skb is filled properly by that function.
660  */
661
662 static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
663                                      struct net_device *dev)
664 {
665         struct ip_tunnel *tunnel = netdev_priv(dev);
666         struct pcpu_tstats *tstats;
667         struct iphdr  *tiph = &tunnel->parms.iph;
668         struct ipv6hdr *iph6 = ipv6_hdr(skb);
669         u8     tos = tunnel->parms.iph.tos;
670         __be16 df = tiph->frag_off;
671         struct rtable *rt;                      /* Route to the other host */
672         struct net_device *tdev;                /* Device to other host */
673         struct iphdr  *iph;                     /* Our new IP header */
674         unsigned int max_headroom;              /* The extra header space needed */
675         __be32 dst = tiph->daddr;
676         int    mtu;
677         struct in6_addr *addr6;
678         int addr_type;
679
680         if (skb->protocol != htons(ETH_P_IPV6))
681                 goto tx_error;
682
683         /* ISATAP (RFC4214) - must come before 6to4 */
684         if (dev->priv_flags & IFF_ISATAP) {
685                 struct neighbour *neigh = NULL;
686
687                 if (skb_dst(skb))
688                         neigh = skb_dst(skb)->neighbour;
689
690                 if (neigh == NULL) {
691                         if (net_ratelimit())
692                                 printk(KERN_DEBUG "sit: nexthop == NULL\n");
693                         goto tx_error;
694                 }
695
696                 addr6 = (struct in6_addr*)&neigh->primary_key;
697                 addr_type = ipv6_addr_type(addr6);
698
699                 if ((addr_type & IPV6_ADDR_UNICAST) &&
700                      ipv6_addr_is_isatap(addr6))
701                         dst = addr6->s6_addr32[3];
702                 else
703                         goto tx_error;
704         }
705
706         if (!dst)
707                 dst = try_6rd(&iph6->daddr, tunnel);
708
709         if (!dst) {
710                 struct neighbour *neigh = NULL;
711
712                 if (skb_dst(skb))
713                         neigh = skb_dst(skb)->neighbour;
714
715                 if (neigh == NULL) {
716                         if (net_ratelimit())
717                                 printk(KERN_DEBUG "sit: nexthop == NULL\n");
718                         goto tx_error;
719                 }
720
721                 addr6 = (struct in6_addr*)&neigh->primary_key;
722                 addr_type = ipv6_addr_type(addr6);
723
724                 if (addr_type == IPV6_ADDR_ANY) {
725                         addr6 = &ipv6_hdr(skb)->daddr;
726                         addr_type = ipv6_addr_type(addr6);
727                 }
728
729                 if ((addr_type & IPV6_ADDR_COMPATv4) == 0)
730                         goto tx_error_icmp;
731
732                 dst = addr6->s6_addr32[3];
733         }
734
735         rt = ip_route_output_ports(dev_net(dev), NULL,
736                                    dst, tiph->saddr,
737                                    0, 0,
738                                    IPPROTO_IPV6, RT_TOS(tos),
739                                    tunnel->parms.link);
740         if (IS_ERR(rt)) {
741                 dev->stats.tx_carrier_errors++;
742                 goto tx_error_icmp;
743         }
744         if (rt->rt_type != RTN_UNICAST) {
745                 ip_rt_put(rt);
746                 dev->stats.tx_carrier_errors++;
747                 goto tx_error_icmp;
748         }
749         tdev = rt->dst.dev;
750
751         if (tdev == dev) {
752                 ip_rt_put(rt);
753                 dev->stats.collisions++;
754                 goto tx_error;
755         }
756
757         if (df) {
758                 mtu = dst_mtu(&rt->dst) - sizeof(struct iphdr);
759
760                 if (mtu < 68) {
761                         dev->stats.collisions++;
762                         ip_rt_put(rt);
763                         goto tx_error;
764                 }
765
766                 if (mtu < IPV6_MIN_MTU) {
767                         mtu = IPV6_MIN_MTU;
768                         df = 0;
769                 }
770
771                 if (tunnel->parms.iph.daddr && skb_dst(skb))
772                         skb_dst(skb)->ops->update_pmtu(skb_dst(skb), mtu);
773
774                 if (skb->len > mtu) {
775                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
776                         ip_rt_put(rt);
777                         goto tx_error;
778                 }
779         }
780
781         if (tunnel->err_count > 0) {
782                 if (time_before(jiffies,
783                                 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
784                         tunnel->err_count--;
785                         dst_link_failure(skb);
786                 } else
787                         tunnel->err_count = 0;
788         }
789
790         /*
791          * Okay, now see if we can stuff it in the buffer as-is.
792          */
793         max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr);
794
795         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
796             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
797                 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
798                 if (!new_skb) {
799                         ip_rt_put(rt);
800                         dev->stats.tx_dropped++;
801                         dev_kfree_skb(skb);
802                         return NETDEV_TX_OK;
803                 }
804                 if (skb->sk)
805                         skb_set_owner_w(new_skb, skb->sk);
806                 dev_kfree_skb(skb);
807                 skb = new_skb;
808                 iph6 = ipv6_hdr(skb);
809         }
810
811         skb->transport_header = skb->network_header;
812         skb_push(skb, sizeof(struct iphdr));
813         skb_reset_network_header(skb);
814         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
815         IPCB(skb)->flags = 0;
816         skb_dst_drop(skb);
817         skb_dst_set(skb, &rt->dst);
818
819         /*
820          *      Push down and install the IPIP header.
821          */
822
823         iph                     =       ip_hdr(skb);
824         iph->version            =       4;
825         iph->ihl                =       sizeof(struct iphdr)>>2;
826         iph->frag_off           =       df;
827         iph->protocol           =       IPPROTO_IPV6;
828         iph->tos                =       INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
829         iph->daddr              =       rt->rt_dst;
830         iph->saddr              =       rt->rt_src;
831
832         if ((iph->ttl = tiph->ttl) == 0)
833                 iph->ttl        =       iph6->hop_limit;
834
835         nf_reset(skb);
836         tstats = this_cpu_ptr(dev->tstats);
837         __IPTUNNEL_XMIT(tstats, &dev->stats);
838         return NETDEV_TX_OK;
839
840 tx_error_icmp:
841         dst_link_failure(skb);
842 tx_error:
843         dev->stats.tx_errors++;
844         dev_kfree_skb(skb);
845         return NETDEV_TX_OK;
846 }
847
848 static void ipip6_tunnel_bind_dev(struct net_device *dev)
849 {
850         struct net_device *tdev = NULL;
851         struct ip_tunnel *tunnel;
852         struct iphdr *iph;
853
854         tunnel = netdev_priv(dev);
855         iph = &tunnel->parms.iph;
856
857         if (iph->daddr) {
858                 struct rtable *rt = ip_route_output_ports(dev_net(dev), NULL,
859                                                           iph->daddr, iph->saddr,
860                                                           0, 0,
861                                                           IPPROTO_IPV6,
862                                                           RT_TOS(iph->tos),
863                                                           tunnel->parms.link);
864
865                 if (!IS_ERR(rt)) {
866                         tdev = rt->dst.dev;
867                         ip_rt_put(rt);
868                 }
869                 dev->flags |= IFF_POINTOPOINT;
870         }
871
872         if (!tdev && tunnel->parms.link)
873                 tdev = __dev_get_by_index(dev_net(dev), tunnel->parms.link);
874
875         if (tdev) {
876                 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
877                 dev->mtu = tdev->mtu - sizeof(struct iphdr);
878                 if (dev->mtu < IPV6_MIN_MTU)
879                         dev->mtu = IPV6_MIN_MTU;
880         }
881         dev->iflink = tunnel->parms.link;
882 }
883
884 static int
885 ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
886 {
887         int err = 0;
888         struct ip_tunnel_parm p;
889         struct ip_tunnel_prl prl;
890         struct ip_tunnel *t;
891         struct net *net = dev_net(dev);
892         struct sit_net *sitn = net_generic(net, sit_net_id);
893 #ifdef CONFIG_IPV6_SIT_6RD
894         struct ip_tunnel_6rd ip6rd;
895 #endif
896
897         switch (cmd) {
898         case SIOCGETTUNNEL:
899 #ifdef CONFIG_IPV6_SIT_6RD
900         case SIOCGET6RD:
901 #endif
902                 t = NULL;
903                 if (dev == sitn->fb_tunnel_dev) {
904                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
905                                 err = -EFAULT;
906                                 break;
907                         }
908                         t = ipip6_tunnel_locate(net, &p, 0);
909                 }
910                 if (t == NULL)
911                         t = netdev_priv(dev);
912
913                 err = -EFAULT;
914                 if (cmd == SIOCGETTUNNEL) {
915                         memcpy(&p, &t->parms, sizeof(p));
916                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p,
917                                          sizeof(p)))
918                                 goto done;
919 #ifdef CONFIG_IPV6_SIT_6RD
920                 } else {
921                         ipv6_addr_copy(&ip6rd.prefix, &t->ip6rd.prefix);
922                         ip6rd.relay_prefix = t->ip6rd.relay_prefix;
923                         ip6rd.prefixlen = t->ip6rd.prefixlen;
924                         ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
925                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd,
926                                          sizeof(ip6rd)))
927                                 goto done;
928 #endif
929                 }
930                 err = 0;
931                 break;
932
933         case SIOCADDTUNNEL:
934         case SIOCCHGTUNNEL:
935                 err = -EPERM;
936                 if (!capable(CAP_NET_ADMIN))
937                         goto done;
938
939                 err = -EFAULT;
940                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
941                         goto done;
942
943                 err = -EINVAL;
944                 if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPV6 ||
945                     p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
946                         goto done;
947                 if (p.iph.ttl)
948                         p.iph.frag_off |= htons(IP_DF);
949
950                 t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
951
952                 if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
953                         if (t != NULL) {
954                                 if (t->dev != dev) {
955                                         err = -EEXIST;
956                                         break;
957                                 }
958                         } else {
959                                 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
960                                     (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
961                                         err = -EINVAL;
962                                         break;
963                                 }
964                                 t = netdev_priv(dev);
965                                 ipip6_tunnel_unlink(sitn, t);
966                                 synchronize_net();
967                                 t->parms.iph.saddr = p.iph.saddr;
968                                 t->parms.iph.daddr = p.iph.daddr;
969                                 memcpy(dev->dev_addr, &p.iph.saddr, 4);
970                                 memcpy(dev->broadcast, &p.iph.daddr, 4);
971                                 ipip6_tunnel_link(sitn, t);
972                                 netdev_state_change(dev);
973                         }
974                 }
975
976                 if (t) {
977                         err = 0;
978                         if (cmd == SIOCCHGTUNNEL) {
979                                 t->parms.iph.ttl = p.iph.ttl;
980                                 t->parms.iph.tos = p.iph.tos;
981                                 if (t->parms.link != p.link) {
982                                         t->parms.link = p.link;
983                                         ipip6_tunnel_bind_dev(dev);
984                                         netdev_state_change(dev);
985                                 }
986                         }
987                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
988                                 err = -EFAULT;
989                 } else
990                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
991                 break;
992
993         case SIOCDELTUNNEL:
994                 err = -EPERM;
995                 if (!capable(CAP_NET_ADMIN))
996                         goto done;
997
998                 if (dev == sitn->fb_tunnel_dev) {
999                         err = -EFAULT;
1000                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1001                                 goto done;
1002                         err = -ENOENT;
1003                         if ((t = ipip6_tunnel_locate(net, &p, 0)) == NULL)
1004                                 goto done;
1005                         err = -EPERM;
1006                         if (t == netdev_priv(sitn->fb_tunnel_dev))
1007                                 goto done;
1008                         dev = t->dev;
1009                 }
1010                 unregister_netdevice(dev);
1011                 err = 0;
1012                 break;
1013
1014         case SIOCGETPRL:
1015                 err = -EINVAL;
1016                 if (dev == sitn->fb_tunnel_dev)
1017                         goto done;
1018                 err = -ENOENT;
1019                 if (!(t = netdev_priv(dev)))
1020                         goto done;
1021                 err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
1022                 break;
1023
1024         case SIOCADDPRL:
1025         case SIOCDELPRL:
1026         case SIOCCHGPRL:
1027                 err = -EPERM;
1028                 if (!capable(CAP_NET_ADMIN))
1029                         goto done;
1030                 err = -EINVAL;
1031                 if (dev == sitn->fb_tunnel_dev)
1032                         goto done;
1033                 err = -EFAULT;
1034                 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
1035                         goto done;
1036                 err = -ENOENT;
1037                 if (!(t = netdev_priv(dev)))
1038                         goto done;
1039
1040                 switch (cmd) {
1041                 case SIOCDELPRL:
1042                         err = ipip6_tunnel_del_prl(t, &prl);
1043                         break;
1044                 case SIOCADDPRL:
1045                 case SIOCCHGPRL:
1046                         err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
1047                         break;
1048                 }
1049                 netdev_state_change(dev);
1050                 break;
1051
1052 #ifdef CONFIG_IPV6_SIT_6RD
1053         case SIOCADD6RD:
1054         case SIOCCHG6RD:
1055         case SIOCDEL6RD:
1056                 err = -EPERM;
1057                 if (!capable(CAP_NET_ADMIN))
1058                         goto done;
1059
1060                 err = -EFAULT;
1061                 if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data,
1062                                    sizeof(ip6rd)))
1063                         goto done;
1064
1065                 t = netdev_priv(dev);
1066
1067                 if (cmd != SIOCDEL6RD) {
1068                         struct in6_addr prefix;
1069                         __be32 relay_prefix;
1070
1071                         err = -EINVAL;
1072                         if (ip6rd.relay_prefixlen > 32 ||
1073                             ip6rd.prefixlen + (32 - ip6rd.relay_prefixlen) > 64)
1074                                 goto done;
1075
1076                         ipv6_addr_prefix(&prefix, &ip6rd.prefix,
1077                                          ip6rd.prefixlen);
1078                         if (!ipv6_addr_equal(&prefix, &ip6rd.prefix))
1079                                 goto done;
1080                         if (ip6rd.relay_prefixlen)
1081                                 relay_prefix = ip6rd.relay_prefix &
1082                                                htonl(0xffffffffUL <<
1083                                                      (32 - ip6rd.relay_prefixlen));
1084                         else
1085                                 relay_prefix = 0;
1086                         if (relay_prefix != ip6rd.relay_prefix)
1087                                 goto done;
1088
1089                         ipv6_addr_copy(&t->ip6rd.prefix, &prefix);
1090                         t->ip6rd.relay_prefix = relay_prefix;
1091                         t->ip6rd.prefixlen = ip6rd.prefixlen;
1092                         t->ip6rd.relay_prefixlen = ip6rd.relay_prefixlen;
1093                 } else
1094                         ipip6_tunnel_clone_6rd(dev, sitn);
1095
1096                 err = 0;
1097                 break;
1098 #endif
1099
1100         default:
1101                 err = -EINVAL;
1102         }
1103
1104 done:
1105         return err;
1106 }
1107
1108 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1109 {
1110         if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
1111                 return -EINVAL;
1112         dev->mtu = new_mtu;
1113         return 0;
1114 }
1115
1116 static const struct net_device_ops ipip6_netdev_ops = {
1117         .ndo_uninit     = ipip6_tunnel_uninit,
1118         .ndo_start_xmit = ipip6_tunnel_xmit,
1119         .ndo_do_ioctl   = ipip6_tunnel_ioctl,
1120         .ndo_change_mtu = ipip6_tunnel_change_mtu,
1121         .ndo_get_stats  = ipip6_get_stats,
1122 };
1123
1124 static void ipip6_dev_free(struct net_device *dev)
1125 {
1126         free_percpu(dev->tstats);
1127         free_netdev(dev);
1128 }
1129
1130 static void ipip6_tunnel_setup(struct net_device *dev)
1131 {
1132         dev->netdev_ops         = &ipip6_netdev_ops;
1133         dev->destructor         = ipip6_dev_free;
1134
1135         dev->type               = ARPHRD_SIT;
1136         dev->hard_header_len    = LL_MAX_HEADER + sizeof(struct iphdr);
1137         dev->mtu                = ETH_DATA_LEN - sizeof(struct iphdr);
1138         dev->flags              = IFF_NOARP;
1139         dev->priv_flags        &= ~IFF_XMIT_DST_RELEASE;
1140         dev->iflink             = 0;
1141         dev->addr_len           = 4;
1142         dev->features           |= NETIF_F_NETNS_LOCAL;
1143         dev->features           |= NETIF_F_LLTX;
1144 }
1145
1146 static int ipip6_tunnel_init(struct net_device *dev)
1147 {
1148         struct ip_tunnel *tunnel = netdev_priv(dev);
1149
1150         tunnel->dev = dev;
1151         strcpy(tunnel->parms.name, dev->name);
1152
1153         memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
1154         memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
1155
1156         ipip6_tunnel_bind_dev(dev);
1157         dev->tstats = alloc_percpu(struct pcpu_tstats);
1158         if (!dev->tstats)
1159                 return -ENOMEM;
1160
1161         return 0;
1162 }
1163
1164 static int __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1165 {
1166         struct ip_tunnel *tunnel = netdev_priv(dev);
1167         struct iphdr *iph = &tunnel->parms.iph;
1168         struct net *net = dev_net(dev);
1169         struct sit_net *sitn = net_generic(net, sit_net_id);
1170
1171         tunnel->dev = dev;
1172         strcpy(tunnel->parms.name, dev->name);
1173
1174         iph->version            = 4;
1175         iph->protocol           = IPPROTO_IPV6;
1176         iph->ihl                = 5;
1177         iph->ttl                = 64;
1178
1179         dev->tstats = alloc_percpu(struct pcpu_tstats);
1180         if (!dev->tstats)
1181                 return -ENOMEM;
1182         dev_hold(dev);
1183         rcu_assign_pointer(sitn->tunnels_wc[0], tunnel);
1184         return 0;
1185 }
1186
1187 static struct xfrm_tunnel sit_handler __read_mostly = {
1188         .handler        =       ipip6_rcv,
1189         .err_handler    =       ipip6_err,
1190         .priority       =       1,
1191 };
1192
1193 static void __net_exit sit_destroy_tunnels(struct sit_net *sitn, struct list_head *head)
1194 {
1195         int prio;
1196
1197         for (prio = 1; prio < 4; prio++) {
1198                 int h;
1199                 for (h = 0; h < HASH_SIZE; h++) {
1200                         struct ip_tunnel *t;
1201
1202                         t = rtnl_dereference(sitn->tunnels[prio][h]);
1203                         while (t != NULL) {
1204                                 unregister_netdevice_queue(t->dev, head);
1205                                 t = rtnl_dereference(t->next);
1206                         }
1207                 }
1208         }
1209 }
1210
1211 static int __net_init sit_init_net(struct net *net)
1212 {
1213         struct sit_net *sitn = net_generic(net, sit_net_id);
1214         int err;
1215
1216         sitn->tunnels[0] = sitn->tunnels_wc;
1217         sitn->tunnels[1] = sitn->tunnels_l;
1218         sitn->tunnels[2] = sitn->tunnels_r;
1219         sitn->tunnels[3] = sitn->tunnels_r_l;
1220
1221         sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1222                                            ipip6_tunnel_setup);
1223         if (!sitn->fb_tunnel_dev) {
1224                 err = -ENOMEM;
1225                 goto err_alloc_dev;
1226         }
1227         dev_net_set(sitn->fb_tunnel_dev, net);
1228
1229         err = ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1230         if (err)
1231                 goto err_dev_free;
1232
1233         ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1234
1235         if ((err = register_netdev(sitn->fb_tunnel_dev)))
1236                 goto err_reg_dev;
1237
1238         return 0;
1239
1240 err_reg_dev:
1241         dev_put(sitn->fb_tunnel_dev);
1242 err_dev_free:
1243         ipip6_dev_free(sitn->fb_tunnel_dev);
1244 err_alloc_dev:
1245         return err;
1246 }
1247
1248 static void __net_exit sit_exit_net(struct net *net)
1249 {
1250         struct sit_net *sitn = net_generic(net, sit_net_id);
1251         LIST_HEAD(list);
1252
1253         rtnl_lock();
1254         sit_destroy_tunnels(sitn, &list);
1255         unregister_netdevice_queue(sitn->fb_tunnel_dev, &list);
1256         unregister_netdevice_many(&list);
1257         rtnl_unlock();
1258 }
1259
1260 static struct pernet_operations sit_net_ops = {
1261         .init = sit_init_net,
1262         .exit = sit_exit_net,
1263         .id   = &sit_net_id,
1264         .size = sizeof(struct sit_net),
1265 };
1266
1267 static void __exit sit_cleanup(void)
1268 {
1269         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1270
1271         unregister_pernet_device(&sit_net_ops);
1272         rcu_barrier(); /* Wait for completion of call_rcu()'s */
1273 }
1274
1275 static int __init sit_init(void)
1276 {
1277         int err;
1278
1279         printk(KERN_INFO "IPv6 over IPv4 tunneling driver\n");
1280
1281         err = register_pernet_device(&sit_net_ops);
1282         if (err < 0)
1283                 return err;
1284         err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1285         if (err < 0) {
1286                 unregister_pernet_device(&sit_net_ops);
1287                 printk(KERN_INFO "sit init: Can't add protocol\n");
1288         }
1289         return err;
1290 }
1291
1292 module_init(sit_init);
1293 module_exit(sit_cleanup);
1294 MODULE_LICENSE("GPL");
1295 MODULE_ALIAS_NETDEV("sit0");