Merge branch 'for-3.2-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tj...
[pandora-kernel.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_INIT_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_INIT_POINTER(t->next, rtnl_dereference(*tp));
196         RCU_INIT_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         nt = netdev_priv(dev);
254
255         nt->parms = *parms;
256         if (ipip6_tunnel_init(dev) < 0)
257                 goto failed_free;
258         ipip6_tunnel_clone_6rd(dev, sitn);
259
260         if (parms->i_flags & SIT_ISATAP)
261                 dev->priv_flags |= IFF_ISATAP;
262
263         if (register_netdevice(dev) < 0)
264                 goto failed_free;
265
266         strcpy(nt->parms.name, dev->name);
267
268         dev_hold(dev);
269
270         ipip6_tunnel_link(sitn, nt);
271         return nt;
272
273 failed_free:
274         ipip6_dev_free(dev);
275 failed:
276         return NULL;
277 }
278
279 #define for_each_prl_rcu(start)                 \
280         for (prl = rcu_dereference(start);      \
281              prl;                               \
282              prl = rcu_dereference(prl->next))
283
284 static struct ip_tunnel_prl_entry *
285 __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
286 {
287         struct ip_tunnel_prl_entry *prl;
288
289         for_each_prl_rcu(t->prl)
290                 if (prl->addr == addr)
291                         break;
292         return prl;
293
294 }
295
296 static int ipip6_tunnel_get_prl(struct ip_tunnel *t,
297                                 struct ip_tunnel_prl __user *a)
298 {
299         struct ip_tunnel_prl kprl, *kp;
300         struct ip_tunnel_prl_entry *prl;
301         unsigned int cmax, c = 0, ca, len;
302         int ret = 0;
303
304         if (copy_from_user(&kprl, a, sizeof(kprl)))
305                 return -EFAULT;
306         cmax = kprl.datalen / sizeof(kprl);
307         if (cmax > 1 && kprl.addr != htonl(INADDR_ANY))
308                 cmax = 1;
309
310         /* For simple GET or for root users,
311          * we try harder to allocate.
312          */
313         kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
314                 kcalloc(cmax, sizeof(*kp), GFP_KERNEL) :
315                 NULL;
316
317         rcu_read_lock();
318
319         ca = t->prl_count < cmax ? t->prl_count : cmax;
320
321         if (!kp) {
322                 /* We don't try hard to allocate much memory for
323                  * non-root users.
324                  * For root users, retry allocating enough memory for
325                  * the answer.
326                  */
327                 kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC);
328                 if (!kp) {
329                         ret = -ENOMEM;
330                         goto out;
331                 }
332         }
333
334         c = 0;
335         for_each_prl_rcu(t->prl) {
336                 if (c >= cmax)
337                         break;
338                 if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr)
339                         continue;
340                 kp[c].addr = prl->addr;
341                 kp[c].flags = prl->flags;
342                 c++;
343                 if (kprl.addr != htonl(INADDR_ANY))
344                         break;
345         }
346 out:
347         rcu_read_unlock();
348
349         len = sizeof(*kp) * c;
350         ret = 0;
351         if ((len && copy_to_user(a + 1, kp, len)) || put_user(len, &a->datalen))
352                 ret = -EFAULT;
353
354         kfree(kp);
355
356         return ret;
357 }
358
359 static int
360 ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
361 {
362         struct ip_tunnel_prl_entry *p;
363         int err = 0;
364
365         if (a->addr == htonl(INADDR_ANY))
366                 return -EINVAL;
367
368         ASSERT_RTNL();
369
370         for (p = rtnl_dereference(t->prl); p; p = rtnl_dereference(p->next)) {
371                 if (p->addr == a->addr) {
372                         if (chg) {
373                                 p->flags = a->flags;
374                                 goto out;
375                         }
376                         err = -EEXIST;
377                         goto out;
378                 }
379         }
380
381         if (chg) {
382                 err = -ENXIO;
383                 goto out;
384         }
385
386         p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
387         if (!p) {
388                 err = -ENOBUFS;
389                 goto out;
390         }
391
392         p->next = t->prl;
393         p->addr = a->addr;
394         p->flags = a->flags;
395         t->prl_count++;
396         RCU_INIT_POINTER(t->prl, p);
397 out:
398         return err;
399 }
400
401 static void prl_list_destroy_rcu(struct rcu_head *head)
402 {
403         struct ip_tunnel_prl_entry *p, *n;
404
405         p = container_of(head, struct ip_tunnel_prl_entry, rcu_head);
406         do {
407                 n = rcu_dereference_protected(p->next, 1);
408                 kfree(p);
409                 p = n;
410         } while (p);
411 }
412
413 static int
414 ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
415 {
416         struct ip_tunnel_prl_entry *x;
417         struct ip_tunnel_prl_entry __rcu **p;
418         int err = 0;
419
420         ASSERT_RTNL();
421
422         if (a && a->addr != htonl(INADDR_ANY)) {
423                 for (p = &t->prl;
424                      (x = rtnl_dereference(*p)) != NULL;
425                      p = &x->next) {
426                         if (x->addr == a->addr) {
427                                 *p = x->next;
428                                 kfree_rcu(x, rcu_head);
429                                 t->prl_count--;
430                                 goto out;
431                         }
432                 }
433                 err = -ENXIO;
434         } else {
435                 x = rtnl_dereference(t->prl);
436                 if (x) {
437                         t->prl_count = 0;
438                         call_rcu(&x->rcu_head, prl_list_destroy_rcu);
439                         t->prl = NULL;
440                 }
441         }
442 out:
443         return err;
444 }
445
446 static int
447 isatap_chksrc(struct sk_buff *skb, const struct iphdr *iph, struct ip_tunnel *t)
448 {
449         struct ip_tunnel_prl_entry *p;
450         int ok = 1;
451
452         rcu_read_lock();
453         p = __ipip6_tunnel_locate_prl(t, iph->saddr);
454         if (p) {
455                 if (p->flags & PRL_DEFAULT)
456                         skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
457                 else
458                         skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
459         } else {
460                 const struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
461
462                 if (ipv6_addr_is_isatap(addr6) &&
463                     (addr6->s6_addr32[3] == iph->saddr) &&
464                     ipv6_chk_prefix(addr6, t->dev))
465                         skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
466                 else
467                         ok = 0;
468         }
469         rcu_read_unlock();
470         return ok;
471 }
472
473 static void ipip6_tunnel_uninit(struct net_device *dev)
474 {
475         struct net *net = dev_net(dev);
476         struct sit_net *sitn = net_generic(net, sit_net_id);
477
478         if (dev == sitn->fb_tunnel_dev) {
479                 RCU_INIT_POINTER(sitn->tunnels_wc[0], NULL);
480         } else {
481                 ipip6_tunnel_unlink(sitn, netdev_priv(dev));
482                 ipip6_tunnel_del_prl(netdev_priv(dev), NULL);
483         }
484         dev_put(dev);
485 }
486
487
488 static int ipip6_err(struct sk_buff *skb, u32 info)
489 {
490
491 /* All the routers (except for Linux) return only
492    8 bytes of packet payload. It means, that precise relaying of
493    ICMP in the real Internet is absolutely infeasible.
494  */
495         const struct iphdr *iph = (const struct iphdr *)skb->data;
496         const int type = icmp_hdr(skb)->type;
497         const int code = icmp_hdr(skb)->code;
498         struct ip_tunnel *t;
499         int err;
500
501         switch (type) {
502         default:
503         case ICMP_PARAMETERPROB:
504                 return 0;
505
506         case ICMP_DEST_UNREACH:
507                 switch (code) {
508                 case ICMP_SR_FAILED:
509                 case ICMP_PORT_UNREACH:
510                         /* Impossible event. */
511                         return 0;
512                 case ICMP_FRAG_NEEDED:
513                         /* Soft state for pmtu is maintained by IP core. */
514                         return 0;
515                 default:
516                         /* All others are translated to HOST_UNREACH.
517                            rfc2003 contains "deep thoughts" about NET_UNREACH,
518                            I believe they are just ether pollution. --ANK
519                          */
520                         break;
521                 }
522                 break;
523         case ICMP_TIME_EXCEEDED:
524                 if (code != ICMP_EXC_TTL)
525                         return 0;
526                 break;
527         }
528
529         err = -ENOENT;
530
531         rcu_read_lock();
532         t = ipip6_tunnel_lookup(dev_net(skb->dev),
533                                 skb->dev,
534                                 iph->daddr,
535                                 iph->saddr);
536         if (t == NULL || t->parms.iph.daddr == 0)
537                 goto out;
538
539         err = 0;
540         if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
541                 goto out;
542
543         if (time_before(jiffies, t->err_time + IPTUNNEL_ERR_TIMEO))
544                 t->err_count++;
545         else
546                 t->err_count = 1;
547         t->err_time = jiffies;
548 out:
549         rcu_read_unlock();
550         return err;
551 }
552
553 static inline void ipip6_ecn_decapsulate(const struct iphdr *iph, struct sk_buff *skb)
554 {
555         if (INET_ECN_is_ce(iph->tos))
556                 IP6_ECN_set_ce(ipv6_hdr(skb));
557 }
558
559 static int ipip6_rcv(struct sk_buff *skb)
560 {
561         const struct iphdr *iph;
562         struct ip_tunnel *tunnel;
563
564         if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
565                 goto out;
566
567         iph = ip_hdr(skb);
568
569         rcu_read_lock();
570         tunnel = ipip6_tunnel_lookup(dev_net(skb->dev), skb->dev,
571                                      iph->saddr, iph->daddr);
572         if (tunnel != NULL) {
573                 struct pcpu_tstats *tstats;
574
575                 secpath_reset(skb);
576                 skb->mac_header = skb->network_header;
577                 skb_reset_network_header(skb);
578                 IPCB(skb)->flags = 0;
579                 skb->protocol = htons(ETH_P_IPV6);
580                 skb->pkt_type = PACKET_HOST;
581
582                 if ((tunnel->dev->priv_flags & IFF_ISATAP) &&
583                     !isatap_chksrc(skb, iph, tunnel)) {
584                         tunnel->dev->stats.rx_errors++;
585                         rcu_read_unlock();
586                         kfree_skb(skb);
587                         return 0;
588                 }
589
590                 tstats = this_cpu_ptr(tunnel->dev->tstats);
591                 tstats->rx_packets++;
592                 tstats->rx_bytes += skb->len;
593
594                 __skb_tunnel_rx(skb, tunnel->dev);
595
596                 ipip6_ecn_decapsulate(iph, skb);
597
598                 netif_rx(skb);
599
600                 rcu_read_unlock();
601                 return 0;
602         }
603
604         /* no tunnel matched,  let upstream know, ipsec may handle it */
605         rcu_read_unlock();
606         return 1;
607 out:
608         kfree_skb(skb);
609         return 0;
610 }
611
612 /*
613  * Returns the embedded IPv4 address if the IPv6 address
614  * comes from 6rd / 6to4 (RFC 3056) addr space.
615  */
616 static inline
617 __be32 try_6rd(const struct in6_addr *v6dst, struct ip_tunnel *tunnel)
618 {
619         __be32 dst = 0;
620
621 #ifdef CONFIG_IPV6_SIT_6RD
622         if (ipv6_prefix_equal(v6dst, &tunnel->ip6rd.prefix,
623                               tunnel->ip6rd.prefixlen)) {
624                 unsigned int pbw0, pbi0;
625                 int pbi1;
626                 u32 d;
627
628                 pbw0 = tunnel->ip6rd.prefixlen >> 5;
629                 pbi0 = tunnel->ip6rd.prefixlen & 0x1f;
630
631                 d = (ntohl(v6dst->s6_addr32[pbw0]) << pbi0) >>
632                     tunnel->ip6rd.relay_prefixlen;
633
634                 pbi1 = pbi0 - tunnel->ip6rd.relay_prefixlen;
635                 if (pbi1 > 0)
636                         d |= ntohl(v6dst->s6_addr32[pbw0 + 1]) >>
637                              (32 - pbi1);
638
639                 dst = tunnel->ip6rd.relay_prefix | htonl(d);
640         }
641 #else
642         if (v6dst->s6_addr16[0] == htons(0x2002)) {
643                 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
644                 memcpy(&dst, &v6dst->s6_addr16[1], 4);
645         }
646 #endif
647         return dst;
648 }
649
650 /*
651  *      This function assumes it is being called from dev_queue_xmit()
652  *      and that skb is filled properly by that function.
653  */
654
655 static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
656                                      struct net_device *dev)
657 {
658         struct ip_tunnel *tunnel = netdev_priv(dev);
659         struct pcpu_tstats *tstats;
660         const struct iphdr  *tiph = &tunnel->parms.iph;
661         const struct ipv6hdr *iph6 = ipv6_hdr(skb);
662         u8     tos = tunnel->parms.iph.tos;
663         __be16 df = tiph->frag_off;
664         struct rtable *rt;                      /* Route to the other host */
665         struct net_device *tdev;                /* Device to other host */
666         struct iphdr  *iph;                     /* Our new IP header */
667         unsigned int max_headroom;              /* The extra header space needed */
668         __be32 dst = tiph->daddr;
669         struct flowi4 fl4;
670         int    mtu;
671         const struct in6_addr *addr6;
672         int addr_type;
673
674         if (skb->protocol != htons(ETH_P_IPV6))
675                 goto tx_error;
676
677         if (tos == 1)
678                 tos = ipv6_get_dsfield(iph6);
679
680         /* ISATAP (RFC4214) - must come before 6to4 */
681         if (dev->priv_flags & IFF_ISATAP) {
682                 struct neighbour *neigh = NULL;
683
684                 if (skb_dst(skb))
685                         neigh = dst_get_neighbour(skb_dst(skb));
686
687                 if (neigh == NULL) {
688                         if (net_ratelimit())
689                                 printk(KERN_DEBUG "sit: nexthop == NULL\n");
690                         goto tx_error;
691                 }
692
693                 addr6 = (const struct in6_addr*)&neigh->primary_key;
694                 addr_type = ipv6_addr_type(addr6);
695
696                 if ((addr_type & IPV6_ADDR_UNICAST) &&
697                      ipv6_addr_is_isatap(addr6))
698                         dst = addr6->s6_addr32[3];
699                 else
700                         goto tx_error;
701         }
702
703         if (!dst)
704                 dst = try_6rd(&iph6->daddr, tunnel);
705
706         if (!dst) {
707                 struct neighbour *neigh = NULL;
708
709                 if (skb_dst(skb))
710                         neigh = dst_get_neighbour(skb_dst(skb));
711
712                 if (neigh == NULL) {
713                         if (net_ratelimit())
714                                 printk(KERN_DEBUG "sit: nexthop == NULL\n");
715                         goto tx_error;
716                 }
717
718                 addr6 = (const struct in6_addr*)&neigh->primary_key;
719                 addr_type = ipv6_addr_type(addr6);
720
721                 if (addr_type == IPV6_ADDR_ANY) {
722                         addr6 = &ipv6_hdr(skb)->daddr;
723                         addr_type = ipv6_addr_type(addr6);
724                 }
725
726                 if ((addr_type & IPV6_ADDR_COMPATv4) == 0)
727                         goto tx_error_icmp;
728
729                 dst = addr6->s6_addr32[3];
730         }
731
732         rt = ip_route_output_ports(dev_net(dev), &fl4, NULL,
733                                    dst, tiph->saddr,
734                                    0, 0,
735                                    IPPROTO_IPV6, RT_TOS(tos),
736                                    tunnel->parms.link);
737         if (IS_ERR(rt)) {
738                 dev->stats.tx_carrier_errors++;
739                 goto tx_error_icmp;
740         }
741         if (rt->rt_type != RTN_UNICAST) {
742                 ip_rt_put(rt);
743                 dev->stats.tx_carrier_errors++;
744                 goto tx_error_icmp;
745         }
746         tdev = rt->dst.dev;
747
748         if (tdev == dev) {
749                 ip_rt_put(rt);
750                 dev->stats.collisions++;
751                 goto tx_error;
752         }
753
754         if (df) {
755                 mtu = dst_mtu(&rt->dst) - sizeof(struct iphdr);
756
757                 if (mtu < 68) {
758                         dev->stats.collisions++;
759                         ip_rt_put(rt);
760                         goto tx_error;
761                 }
762
763                 if (mtu < IPV6_MIN_MTU) {
764                         mtu = IPV6_MIN_MTU;
765                         df = 0;
766                 }
767
768                 if (tunnel->parms.iph.daddr && skb_dst(skb))
769                         skb_dst(skb)->ops->update_pmtu(skb_dst(skb), mtu);
770
771                 if (skb->len > mtu) {
772                         icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
773                         ip_rt_put(rt);
774                         goto tx_error;
775                 }
776         }
777
778         if (tunnel->err_count > 0) {
779                 if (time_before(jiffies,
780                                 tunnel->err_time + IPTUNNEL_ERR_TIMEO)) {
781                         tunnel->err_count--;
782                         dst_link_failure(skb);
783                 } else
784                         tunnel->err_count = 0;
785         }
786
787         /*
788          * Okay, now see if we can stuff it in the buffer as-is.
789          */
790         max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr);
791
792         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
793             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
794                 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
795                 if (!new_skb) {
796                         ip_rt_put(rt);
797                         dev->stats.tx_dropped++;
798                         dev_kfree_skb(skb);
799                         return NETDEV_TX_OK;
800                 }
801                 if (skb->sk)
802                         skb_set_owner_w(new_skb, skb->sk);
803                 dev_kfree_skb(skb);
804                 skb = new_skb;
805                 iph6 = ipv6_hdr(skb);
806         }
807
808         skb->transport_header = skb->network_header;
809         skb_push(skb, sizeof(struct iphdr));
810         skb_reset_network_header(skb);
811         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
812         IPCB(skb)->flags = 0;
813         skb_dst_drop(skb);
814         skb_dst_set(skb, &rt->dst);
815
816         /*
817          *      Push down and install the IPIP header.
818          */
819
820         iph                     =       ip_hdr(skb);
821         iph->version            =       4;
822         iph->ihl                =       sizeof(struct iphdr)>>2;
823         iph->frag_off           =       df;
824         iph->protocol           =       IPPROTO_IPV6;
825         iph->tos                =       INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
826         iph->daddr              =       fl4.daddr;
827         iph->saddr              =       fl4.saddr;
828
829         if ((iph->ttl = tiph->ttl) == 0)
830                 iph->ttl        =       iph6->hop_limit;
831
832         nf_reset(skb);
833         tstats = this_cpu_ptr(dev->tstats);
834         __IPTUNNEL_XMIT(tstats, &dev->stats);
835         return NETDEV_TX_OK;
836
837 tx_error_icmp:
838         dst_link_failure(skb);
839 tx_error:
840         dev->stats.tx_errors++;
841         dev_kfree_skb(skb);
842         return NETDEV_TX_OK;
843 }
844
845 static void ipip6_tunnel_bind_dev(struct net_device *dev)
846 {
847         struct net_device *tdev = NULL;
848         struct ip_tunnel *tunnel;
849         const struct iphdr *iph;
850         struct flowi4 fl4;
851
852         tunnel = netdev_priv(dev);
853         iph = &tunnel->parms.iph;
854
855         if (iph->daddr) {
856                 struct rtable *rt = ip_route_output_ports(dev_net(dev), &fl4, NULL,
857                                                           iph->daddr, iph->saddr,
858                                                           0, 0,
859                                                           IPPROTO_IPV6,
860                                                           RT_TOS(iph->tos),
861                                                           tunnel->parms.link);
862
863                 if (!IS_ERR(rt)) {
864                         tdev = rt->dst.dev;
865                         ip_rt_put(rt);
866                 }
867                 dev->flags |= IFF_POINTOPOINT;
868         }
869
870         if (!tdev && tunnel->parms.link)
871                 tdev = __dev_get_by_index(dev_net(dev), tunnel->parms.link);
872
873         if (tdev) {
874                 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
875                 dev->mtu = tdev->mtu - sizeof(struct iphdr);
876                 if (dev->mtu < IPV6_MIN_MTU)
877                         dev->mtu = IPV6_MIN_MTU;
878         }
879         dev->iflink = tunnel->parms.link;
880 }
881
882 static int
883 ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
884 {
885         int err = 0;
886         struct ip_tunnel_parm p;
887         struct ip_tunnel_prl prl;
888         struct ip_tunnel *t;
889         struct net *net = dev_net(dev);
890         struct sit_net *sitn = net_generic(net, sit_net_id);
891 #ifdef CONFIG_IPV6_SIT_6RD
892         struct ip_tunnel_6rd ip6rd;
893 #endif
894
895         switch (cmd) {
896         case SIOCGETTUNNEL:
897 #ifdef CONFIG_IPV6_SIT_6RD
898         case SIOCGET6RD:
899 #endif
900                 t = NULL;
901                 if (dev == sitn->fb_tunnel_dev) {
902                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
903                                 err = -EFAULT;
904                                 break;
905                         }
906                         t = ipip6_tunnel_locate(net, &p, 0);
907                 }
908                 if (t == NULL)
909                         t = netdev_priv(dev);
910
911                 err = -EFAULT;
912                 if (cmd == SIOCGETTUNNEL) {
913                         memcpy(&p, &t->parms, sizeof(p));
914                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &p,
915                                          sizeof(p)))
916                                 goto done;
917 #ifdef CONFIG_IPV6_SIT_6RD
918                 } else {
919                         ipv6_addr_copy(&ip6rd.prefix, &t->ip6rd.prefix);
920                         ip6rd.relay_prefix = t->ip6rd.relay_prefix;
921                         ip6rd.prefixlen = t->ip6rd.prefixlen;
922                         ip6rd.relay_prefixlen = t->ip6rd.relay_prefixlen;
923                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &ip6rd,
924                                          sizeof(ip6rd)))
925                                 goto done;
926 #endif
927                 }
928                 err = 0;
929                 break;
930
931         case SIOCADDTUNNEL:
932         case SIOCCHGTUNNEL:
933                 err = -EPERM;
934                 if (!capable(CAP_NET_ADMIN))
935                         goto done;
936
937                 err = -EFAULT;
938                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
939                         goto done;
940
941                 err = -EINVAL;
942                 if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPV6 ||
943                     p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
944                         goto done;
945                 if (p.iph.ttl)
946                         p.iph.frag_off |= htons(IP_DF);
947
948                 t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
949
950                 if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
951                         if (t != NULL) {
952                                 if (t->dev != dev) {
953                                         err = -EEXIST;
954                                         break;
955                                 }
956                         } else {
957                                 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
958                                     (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
959                                         err = -EINVAL;
960                                         break;
961                                 }
962                                 t = netdev_priv(dev);
963                                 ipip6_tunnel_unlink(sitn, t);
964                                 synchronize_net();
965                                 t->parms.iph.saddr = p.iph.saddr;
966                                 t->parms.iph.daddr = p.iph.daddr;
967                                 memcpy(dev->dev_addr, &p.iph.saddr, 4);
968                                 memcpy(dev->broadcast, &p.iph.daddr, 4);
969                                 ipip6_tunnel_link(sitn, t);
970                                 netdev_state_change(dev);
971                         }
972                 }
973
974                 if (t) {
975                         err = 0;
976                         if (cmd == SIOCCHGTUNNEL) {
977                                 t->parms.iph.ttl = p.iph.ttl;
978                                 t->parms.iph.tos = p.iph.tos;
979                                 if (t->parms.link != p.link) {
980                                         t->parms.link = p.link;
981                                         ipip6_tunnel_bind_dev(dev);
982                                         netdev_state_change(dev);
983                                 }
984                         }
985                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
986                                 err = -EFAULT;
987                 } else
988                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
989                 break;
990
991         case SIOCDELTUNNEL:
992                 err = -EPERM;
993                 if (!capable(CAP_NET_ADMIN))
994                         goto done;
995
996                 if (dev == sitn->fb_tunnel_dev) {
997                         err = -EFAULT;
998                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
999                                 goto done;
1000                         err = -ENOENT;
1001                         if ((t = ipip6_tunnel_locate(net, &p, 0)) == NULL)
1002                                 goto done;
1003                         err = -EPERM;
1004                         if (t == netdev_priv(sitn->fb_tunnel_dev))
1005                                 goto done;
1006                         dev = t->dev;
1007                 }
1008                 unregister_netdevice(dev);
1009                 err = 0;
1010                 break;
1011
1012         case SIOCGETPRL:
1013                 err = -EINVAL;
1014                 if (dev == sitn->fb_tunnel_dev)
1015                         goto done;
1016                 err = -ENOENT;
1017                 if (!(t = netdev_priv(dev)))
1018                         goto done;
1019                 err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
1020                 break;
1021
1022         case SIOCADDPRL:
1023         case SIOCDELPRL:
1024         case SIOCCHGPRL:
1025                 err = -EPERM;
1026                 if (!capable(CAP_NET_ADMIN))
1027                         goto done;
1028                 err = -EINVAL;
1029                 if (dev == sitn->fb_tunnel_dev)
1030                         goto done;
1031                 err = -EFAULT;
1032                 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
1033                         goto done;
1034                 err = -ENOENT;
1035                 if (!(t = netdev_priv(dev)))
1036                         goto done;
1037
1038                 switch (cmd) {
1039                 case SIOCDELPRL:
1040                         err = ipip6_tunnel_del_prl(t, &prl);
1041                         break;
1042                 case SIOCADDPRL:
1043                 case SIOCCHGPRL:
1044                         err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
1045                         break;
1046                 }
1047                 netdev_state_change(dev);
1048                 break;
1049
1050 #ifdef CONFIG_IPV6_SIT_6RD
1051         case SIOCADD6RD:
1052         case SIOCCHG6RD:
1053         case SIOCDEL6RD:
1054                 err = -EPERM;
1055                 if (!capable(CAP_NET_ADMIN))
1056                         goto done;
1057
1058                 err = -EFAULT;
1059                 if (copy_from_user(&ip6rd, ifr->ifr_ifru.ifru_data,
1060                                    sizeof(ip6rd)))
1061                         goto done;
1062
1063                 t = netdev_priv(dev);
1064
1065                 if (cmd != SIOCDEL6RD) {
1066                         struct in6_addr prefix;
1067                         __be32 relay_prefix;
1068
1069                         err = -EINVAL;
1070                         if (ip6rd.relay_prefixlen > 32 ||
1071                             ip6rd.prefixlen + (32 - ip6rd.relay_prefixlen) > 64)
1072                                 goto done;
1073
1074                         ipv6_addr_prefix(&prefix, &ip6rd.prefix,
1075                                          ip6rd.prefixlen);
1076                         if (!ipv6_addr_equal(&prefix, &ip6rd.prefix))
1077                                 goto done;
1078                         if (ip6rd.relay_prefixlen)
1079                                 relay_prefix = ip6rd.relay_prefix &
1080                                                htonl(0xffffffffUL <<
1081                                                      (32 - ip6rd.relay_prefixlen));
1082                         else
1083                                 relay_prefix = 0;
1084                         if (relay_prefix != ip6rd.relay_prefix)
1085                                 goto done;
1086
1087                         ipv6_addr_copy(&t->ip6rd.prefix, &prefix);
1088                         t->ip6rd.relay_prefix = relay_prefix;
1089                         t->ip6rd.prefixlen = ip6rd.prefixlen;
1090                         t->ip6rd.relay_prefixlen = ip6rd.relay_prefixlen;
1091                 } else
1092                         ipip6_tunnel_clone_6rd(dev, sitn);
1093
1094                 err = 0;
1095                 break;
1096 #endif
1097
1098         default:
1099                 err = -EINVAL;
1100         }
1101
1102 done:
1103         return err;
1104 }
1105
1106 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
1107 {
1108         if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
1109                 return -EINVAL;
1110         dev->mtu = new_mtu;
1111         return 0;
1112 }
1113
1114 static const struct net_device_ops ipip6_netdev_ops = {
1115         .ndo_uninit     = ipip6_tunnel_uninit,
1116         .ndo_start_xmit = ipip6_tunnel_xmit,
1117         .ndo_do_ioctl   = ipip6_tunnel_ioctl,
1118         .ndo_change_mtu = ipip6_tunnel_change_mtu,
1119         .ndo_get_stats  = ipip6_get_stats,
1120 };
1121
1122 static void ipip6_dev_free(struct net_device *dev)
1123 {
1124         free_percpu(dev->tstats);
1125         free_netdev(dev);
1126 }
1127
1128 static void ipip6_tunnel_setup(struct net_device *dev)
1129 {
1130         dev->netdev_ops         = &ipip6_netdev_ops;
1131         dev->destructor         = ipip6_dev_free;
1132
1133         dev->type               = ARPHRD_SIT;
1134         dev->hard_header_len    = LL_MAX_HEADER + sizeof(struct iphdr);
1135         dev->mtu                = ETH_DATA_LEN - sizeof(struct iphdr);
1136         dev->flags              = IFF_NOARP;
1137         dev->priv_flags        &= ~IFF_XMIT_DST_RELEASE;
1138         dev->iflink             = 0;
1139         dev->addr_len           = 4;
1140         dev->features           |= NETIF_F_NETNS_LOCAL;
1141         dev->features           |= NETIF_F_LLTX;
1142 }
1143
1144 static int ipip6_tunnel_init(struct net_device *dev)
1145 {
1146         struct ip_tunnel *tunnel = netdev_priv(dev);
1147
1148         tunnel->dev = dev;
1149
1150         memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
1151         memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
1152
1153         ipip6_tunnel_bind_dev(dev);
1154         dev->tstats = alloc_percpu(struct pcpu_tstats);
1155         if (!dev->tstats)
1156                 return -ENOMEM;
1157
1158         return 0;
1159 }
1160
1161 static int __net_init ipip6_fb_tunnel_init(struct net_device *dev)
1162 {
1163         struct ip_tunnel *tunnel = netdev_priv(dev);
1164         struct iphdr *iph = &tunnel->parms.iph;
1165         struct net *net = dev_net(dev);
1166         struct sit_net *sitn = net_generic(net, sit_net_id);
1167
1168         tunnel->dev = dev;
1169         strcpy(tunnel->parms.name, dev->name);
1170
1171         iph->version            = 4;
1172         iph->protocol           = IPPROTO_IPV6;
1173         iph->ihl                = 5;
1174         iph->ttl                = 64;
1175
1176         dev->tstats = alloc_percpu(struct pcpu_tstats);
1177         if (!dev->tstats)
1178                 return -ENOMEM;
1179         dev_hold(dev);
1180         RCU_INIT_POINTER(sitn->tunnels_wc[0], tunnel);
1181         return 0;
1182 }
1183
1184 static struct xfrm_tunnel sit_handler __read_mostly = {
1185         .handler        =       ipip6_rcv,
1186         .err_handler    =       ipip6_err,
1187         .priority       =       1,
1188 };
1189
1190 static void __net_exit sit_destroy_tunnels(struct sit_net *sitn, struct list_head *head)
1191 {
1192         int prio;
1193
1194         for (prio = 1; prio < 4; prio++) {
1195                 int h;
1196                 for (h = 0; h < HASH_SIZE; h++) {
1197                         struct ip_tunnel *t;
1198
1199                         t = rtnl_dereference(sitn->tunnels[prio][h]);
1200                         while (t != NULL) {
1201                                 unregister_netdevice_queue(t->dev, head);
1202                                 t = rtnl_dereference(t->next);
1203                         }
1204                 }
1205         }
1206 }
1207
1208 static int __net_init sit_init_net(struct net *net)
1209 {
1210         struct sit_net *sitn = net_generic(net, sit_net_id);
1211         struct ip_tunnel *t;
1212         int err;
1213
1214         sitn->tunnels[0] = sitn->tunnels_wc;
1215         sitn->tunnels[1] = sitn->tunnels_l;
1216         sitn->tunnels[2] = sitn->tunnels_r;
1217         sitn->tunnels[3] = sitn->tunnels_r_l;
1218
1219         sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1220                                            ipip6_tunnel_setup);
1221         if (!sitn->fb_tunnel_dev) {
1222                 err = -ENOMEM;
1223                 goto err_alloc_dev;
1224         }
1225         dev_net_set(sitn->fb_tunnel_dev, net);
1226
1227         err = ipip6_fb_tunnel_init(sitn->fb_tunnel_dev);
1228         if (err)
1229                 goto err_dev_free;
1230
1231         ipip6_tunnel_clone_6rd(sitn->fb_tunnel_dev, sitn);
1232
1233         if ((err = register_netdev(sitn->fb_tunnel_dev)))
1234                 goto err_reg_dev;
1235
1236         t = netdev_priv(sitn->fb_tunnel_dev);
1237
1238         strcpy(t->parms.name, sitn->fb_tunnel_dev->name);
1239         return 0;
1240
1241 err_reg_dev:
1242         dev_put(sitn->fb_tunnel_dev);
1243 err_dev_free:
1244         ipip6_dev_free(sitn->fb_tunnel_dev);
1245 err_alloc_dev:
1246         return err;
1247 }
1248
1249 static void __net_exit sit_exit_net(struct net *net)
1250 {
1251         struct sit_net *sitn = net_generic(net, sit_net_id);
1252         LIST_HEAD(list);
1253
1254         rtnl_lock();
1255         sit_destroy_tunnels(sitn, &list);
1256         unregister_netdevice_queue(sitn->fb_tunnel_dev, &list);
1257         unregister_netdevice_many(&list);
1258         rtnl_unlock();
1259 }
1260
1261 static struct pernet_operations sit_net_ops = {
1262         .init = sit_init_net,
1263         .exit = sit_exit_net,
1264         .id   = &sit_net_id,
1265         .size = sizeof(struct sit_net),
1266 };
1267
1268 static void __exit sit_cleanup(void)
1269 {
1270         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1271
1272         unregister_pernet_device(&sit_net_ops);
1273         rcu_barrier(); /* Wait for completion of call_rcu()'s */
1274 }
1275
1276 static int __init sit_init(void)
1277 {
1278         int err;
1279
1280         printk(KERN_INFO "IPv6 over IPv4 tunneling driver\n");
1281
1282         err = register_pernet_device(&sit_net_ops);
1283         if (err < 0)
1284                 return err;
1285         err = xfrm4_tunnel_register(&sit_handler, AF_INET6);
1286         if (err < 0) {
1287                 unregister_pernet_device(&sit_net_ops);
1288                 printk(KERN_INFO "sit init: Can't add protocol\n");
1289         }
1290         return err;
1291 }
1292
1293 module_init(sit_init);
1294 module_exit(sit_cleanup);
1295 MODULE_LICENSE("GPL");
1296 MODULE_ALIAS_NETDEV("sit0");