ipv6 sit: Avoid extra need for compat layer in PRL management.
[linux-2.6] / 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  *      $Id: sit.c,v 1.53 2001/09/25 05:09:53 davem Exp $
10  *
11  *      This program is free software; you can redistribute it and/or
12  *      modify it under the terms of the GNU General Public License
13  *      as published by the Free Software Foundation; either version
14  *      2 of the License, or (at your option) any later version.
15  *
16  *      Changes:
17  * Roger Venning <r.venning@telstra.com>:       6to4 support
18  * Nate Thompson <nate@thebog.net>:             6to4 support
19  * Fred Templin <fred.l.templin@boeing.com>:    isatap support
20  */
21
22 #include <linux/module.h>
23 #include <linux/capability.h>
24 #include <linux/errno.h>
25 #include <linux/types.h>
26 #include <linux/socket.h>
27 #include <linux/sockios.h>
28 #include <linux/net.h>
29 #include <linux/in6.h>
30 #include <linux/netdevice.h>
31 #include <linux/if_arp.h>
32 #include <linux/icmp.h>
33 #include <asm/uaccess.h>
34 #include <linux/init.h>
35 #include <linux/netfilter_ipv4.h>
36 #include <linux/if_ether.h>
37
38 #include <net/sock.h>
39 #include <net/snmp.h>
40
41 #include <net/ipv6.h>
42 #include <net/protocol.h>
43 #include <net/transp_v6.h>
44 #include <net/ip6_fib.h>
45 #include <net/ip6_route.h>
46 #include <net/ndisc.h>
47 #include <net/addrconf.h>
48 #include <net/ip.h>
49 #include <net/udp.h>
50 #include <net/icmp.h>
51 #include <net/ipip.h>
52 #include <net/inet_ecn.h>
53 #include <net/xfrm.h>
54 #include <net/dsfield.h>
55 #include <net/net_namespace.h>
56 #include <net/netns/generic.h>
57
58 /*
59    This version of net/ipv6/sit.c is cloned of net/ipv4/ip_gre.c
60
61    For comments look at net/ipv4/ip_gre.c --ANK
62  */
63
64 #define HASH_SIZE  16
65 #define HASH(addr) (((__force u32)addr^((__force u32)addr>>4))&0xF)
66
67 static int ipip6_fb_tunnel_init(struct net_device *dev);
68 static int ipip6_tunnel_init(struct net_device *dev);
69 static void ipip6_tunnel_setup(struct net_device *dev);
70
71 static int sit_net_id;
72 struct sit_net {
73         struct ip_tunnel *tunnels_r_l[HASH_SIZE];
74         struct ip_tunnel *tunnels_r[HASH_SIZE];
75         struct ip_tunnel *tunnels_l[HASH_SIZE];
76         struct ip_tunnel *tunnels_wc[1];
77         struct ip_tunnel **tunnels[4];
78
79         struct net_device *fb_tunnel_dev;
80 };
81
82 static DEFINE_RWLOCK(ipip6_lock);
83
84 static struct ip_tunnel * ipip6_tunnel_lookup(struct net *net,
85                 __be32 remote, __be32 local)
86 {
87         unsigned h0 = HASH(remote);
88         unsigned h1 = HASH(local);
89         struct ip_tunnel *t;
90         struct sit_net *sitn = net_generic(net, sit_net_id);
91
92         for (t = sitn->tunnels_r_l[h0^h1]; t; t = t->next) {
93                 if (local == t->parms.iph.saddr &&
94                     remote == t->parms.iph.daddr && (t->dev->flags&IFF_UP))
95                         return t;
96         }
97         for (t = sitn->tunnels_r[h0]; t; t = t->next) {
98                 if (remote == t->parms.iph.daddr && (t->dev->flags&IFF_UP))
99                         return t;
100         }
101         for (t = sitn->tunnels_l[h1]; t; t = t->next) {
102                 if (local == t->parms.iph.saddr && (t->dev->flags&IFF_UP))
103                         return t;
104         }
105         if ((t = sitn->tunnels_wc[0]) != NULL && (t->dev->flags&IFF_UP))
106                 return t;
107         return NULL;
108 }
109
110 static struct ip_tunnel **__ipip6_bucket(struct sit_net *sitn,
111                 struct ip_tunnel_parm *parms)
112 {
113         __be32 remote = parms->iph.daddr;
114         __be32 local = parms->iph.saddr;
115         unsigned h = 0;
116         int prio = 0;
117
118         if (remote) {
119                 prio |= 2;
120                 h ^= HASH(remote);
121         }
122         if (local) {
123                 prio |= 1;
124                 h ^= HASH(local);
125         }
126         return &sitn->tunnels[prio][h];
127 }
128
129 static inline struct ip_tunnel **ipip6_bucket(struct sit_net *sitn,
130                 struct ip_tunnel *t)
131 {
132         return __ipip6_bucket(sitn, &t->parms);
133 }
134
135 static void ipip6_tunnel_unlink(struct sit_net *sitn, struct ip_tunnel *t)
136 {
137         struct ip_tunnel **tp;
138
139         for (tp = ipip6_bucket(sitn, t); *tp; tp = &(*tp)->next) {
140                 if (t == *tp) {
141                         write_lock_bh(&ipip6_lock);
142                         *tp = t->next;
143                         write_unlock_bh(&ipip6_lock);
144                         break;
145                 }
146         }
147 }
148
149 static void ipip6_tunnel_link(struct sit_net *sitn, struct ip_tunnel *t)
150 {
151         struct ip_tunnel **tp = ipip6_bucket(sitn, t);
152
153         t->next = *tp;
154         write_lock_bh(&ipip6_lock);
155         *tp = t;
156         write_unlock_bh(&ipip6_lock);
157 }
158
159 static struct ip_tunnel * ipip6_tunnel_locate(struct net *net,
160                 struct ip_tunnel_parm *parms, int create)
161 {
162         __be32 remote = parms->iph.daddr;
163         __be32 local = parms->iph.saddr;
164         struct ip_tunnel *t, **tp, *nt;
165         struct net_device *dev;
166         char name[IFNAMSIZ];
167         struct sit_net *sitn = net_generic(net, sit_net_id);
168
169         for (tp = __ipip6_bucket(sitn, parms); (t = *tp) != NULL; tp = &t->next) {
170                 if (local == t->parms.iph.saddr && remote == t->parms.iph.daddr)
171                         return t;
172         }
173         if (!create)
174                 goto failed;
175
176         if (parms->name[0])
177                 strlcpy(name, parms->name, IFNAMSIZ);
178         else
179                 sprintf(name, "sit%%d");
180
181         dev = alloc_netdev(sizeof(*t), name, ipip6_tunnel_setup);
182         if (dev == NULL)
183                 return NULL;
184
185         dev_net_set(dev, net);
186
187         if (strchr(name, '%')) {
188                 if (dev_alloc_name(dev, name) < 0)
189                         goto failed_free;
190         }
191
192         nt = netdev_priv(dev);
193         dev->init = ipip6_tunnel_init;
194         nt->parms = *parms;
195
196         if (parms->i_flags & SIT_ISATAP)
197                 dev->priv_flags |= IFF_ISATAP;
198
199         if (register_netdevice(dev) < 0)
200                 goto failed_free;
201
202         dev_hold(dev);
203
204         ipip6_tunnel_link(sitn, nt);
205         return nt;
206
207 failed_free:
208         free_netdev(dev);
209 failed:
210         return NULL;
211 }
212
213 static struct ip_tunnel_prl_entry *
214 __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
215 {
216         struct ip_tunnel_prl_entry *p = (struct ip_tunnel_prl_entry *)NULL;
217
218         for (p = t->prl; p; p = p->next)
219                 if (p->addr == addr)
220                         break;
221         return p;
222
223 }
224
225 static int ipip6_tunnel_get_prl(struct ip_tunnel *t,
226                                 struct ip_tunnel_prl __user *a)
227 {
228         struct ip_tunnel_prl kprl, *kp;
229         struct ip_tunnel_prl_entry *prl;
230         unsigned int cmax, c = 0, ca, len;
231         int ret = 0;
232
233         if (copy_from_user(&kprl, a, sizeof(kprl)))
234                 return -EFAULT;
235         cmax = kprl.datalen / sizeof(kprl);
236         if (cmax > 1 && kprl.addr != htonl(INADDR_ANY))
237                 cmax = 1;
238
239         /* For simple GET or for root users,
240          * we try harder to allocate.
241          */
242         kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
243                 kcalloc(cmax, sizeof(*kp), GFP_KERNEL) :
244                 NULL;
245
246         read_lock(&ipip6_lock);
247
248         ca = t->prl_count < cmax ? t->prl_count : cmax;
249
250         if (!kp) {
251                 /* We don't try hard to allocate much memory for
252                  * non-root users.
253                  * For root users, retry allocating enough memory for
254                  * the answer.
255                  */
256                 kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC);
257                 if (!kp) {
258                         ret = -ENOMEM;
259                         goto out;
260                 }
261         }
262
263         c = 0;
264         for (prl = t->prl; prl; prl = prl->next) {
265                 if (c > cmax)
266                         break;
267                 if (kprl.addr != htonl(INADDR_ANY) && prl->addr != kprl.addr)
268                         continue;
269                 kp[c].addr = prl->addr;
270                 kp[c].flags = prl->flags;
271                 c++;
272                 if (kprl.addr != htonl(INADDR_ANY))
273                         break;
274         }
275 out:
276         read_unlock(&ipip6_lock);
277
278         len = sizeof(*kp) * c;
279         ret = 0;
280         if ((len && copy_to_user(a + 1, kp, len)) || put_user(len, &a->datalen))
281                 ret = -EFAULT;
282
283         kfree(kp);
284
285         return ret;
286 }
287
288 static int
289 ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
290 {
291         struct ip_tunnel_prl_entry *p;
292         int err = 0;
293
294         if (a->addr == htonl(INADDR_ANY))
295                 return -EINVAL;
296
297         write_lock(&ipip6_lock);
298
299         for (p = t->prl; p; p = p->next) {
300                 if (p->addr == a->addr) {
301                         if (chg)
302                                 goto update;
303                         err = -EEXIST;
304                         goto out;
305                 }
306         }
307
308         if (chg) {
309                 err = -ENXIO;
310                 goto out;
311         }
312
313         p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
314         if (!p) {
315                 err = -ENOBUFS;
316                 goto out;
317         }
318
319         p->next = t->prl;
320         t->prl = p;
321         t->prl_count++;
322 update:
323         p->addr = a->addr;
324         p->flags = a->flags;
325 out:
326         write_unlock(&ipip6_lock);
327         return err;
328 }
329
330 static int
331 ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
332 {
333         struct ip_tunnel_prl_entry *x, **p;
334         int err = 0;
335
336         write_lock(&ipip6_lock);
337
338         if (a && a->addr != htonl(INADDR_ANY)) {
339                 for (p = &t->prl; *p; p = &(*p)->next) {
340                         if ((*p)->addr == a->addr) {
341                                 x = *p;
342                                 *p = x->next;
343                                 kfree(x);
344                                 t->prl_count--;
345                                 goto out;
346                         }
347                 }
348                 err = -ENXIO;
349         } else {
350                 while (t->prl) {
351                         x = t->prl;
352                         t->prl = t->prl->next;
353                         kfree(x);
354                         t->prl_count--;
355                 }
356         }
357 out:
358         write_unlock(&ipip6_lock);
359         return 0;
360 }
361
362 static int
363 isatap_chksrc(struct sk_buff *skb, struct iphdr *iph, struct ip_tunnel *t)
364 {
365         struct ip_tunnel_prl_entry *p;
366         int ok = 1;
367
368         read_lock(&ipip6_lock);
369         p = __ipip6_tunnel_locate_prl(t, iph->saddr);
370         if (p) {
371                 if (p->flags & PRL_DEFAULT)
372                         skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
373                 else
374                         skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
375         } else {
376                 struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
377                 if (ipv6_addr_is_isatap(addr6) &&
378                     (addr6->s6_addr32[3] == iph->saddr) &&
379                     ipv6_chk_prefix(addr6, t->dev))
380                         skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
381                 else
382                         ok = 0;
383         }
384         read_unlock(&ipip6_lock);
385         return ok;
386 }
387
388 static void ipip6_tunnel_uninit(struct net_device *dev)
389 {
390         struct net *net = dev_net(dev);
391         struct sit_net *sitn = net_generic(net, sit_net_id);
392
393         if (dev == sitn->fb_tunnel_dev) {
394                 write_lock_bh(&ipip6_lock);
395                 sitn->tunnels_wc[0] = NULL;
396                 write_unlock_bh(&ipip6_lock);
397                 dev_put(dev);
398         } else {
399                 ipip6_tunnel_unlink(sitn, netdev_priv(dev));
400                 ipip6_tunnel_del_prl(netdev_priv(dev), NULL);
401                 dev_put(dev);
402         }
403 }
404
405
406 static int ipip6_err(struct sk_buff *skb, u32 info)
407 {
408
409 /* All the routers (except for Linux) return only
410    8 bytes of packet payload. It means, that precise relaying of
411    ICMP in the real Internet is absolutely infeasible.
412  */
413         struct iphdr *iph = (struct iphdr*)skb->data;
414         const int type = icmp_hdr(skb)->type;
415         const int code = icmp_hdr(skb)->code;
416         struct ip_tunnel *t;
417         int err;
418
419         switch (type) {
420         default:
421         case ICMP_PARAMETERPROB:
422                 return 0;
423
424         case ICMP_DEST_UNREACH:
425                 switch (code) {
426                 case ICMP_SR_FAILED:
427                 case ICMP_PORT_UNREACH:
428                         /* Impossible event. */
429                         return 0;
430                 case ICMP_FRAG_NEEDED:
431                         /* Soft state for pmtu is maintained by IP core. */
432                         return 0;
433                 default:
434                         /* All others are translated to HOST_UNREACH.
435                            rfc2003 contains "deep thoughts" about NET_UNREACH,
436                            I believe they are just ether pollution. --ANK
437                          */
438                         break;
439                 }
440                 break;
441         case ICMP_TIME_EXCEEDED:
442                 if (code != ICMP_EXC_TTL)
443                         return 0;
444                 break;
445         }
446
447         err = -ENOENT;
448
449         read_lock(&ipip6_lock);
450         t = ipip6_tunnel_lookup(dev_net(skb->dev), iph->daddr, iph->saddr);
451         if (t == NULL || t->parms.iph.daddr == 0)
452                 goto out;
453
454         err = 0;
455         if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
456                 goto out;
457
458         if (jiffies - t->err_time < IPTUNNEL_ERR_TIMEO)
459                 t->err_count++;
460         else
461                 t->err_count = 1;
462         t->err_time = jiffies;
463 out:
464         read_unlock(&ipip6_lock);
465         return err;
466 }
467
468 static inline void ipip6_ecn_decapsulate(struct iphdr *iph, struct sk_buff *skb)
469 {
470         if (INET_ECN_is_ce(iph->tos))
471                 IP6_ECN_set_ce(ipv6_hdr(skb));
472 }
473
474 static int ipip6_rcv(struct sk_buff *skb)
475 {
476         struct iphdr *iph;
477         struct ip_tunnel *tunnel;
478
479         if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
480                 goto out;
481
482         iph = ip_hdr(skb);
483
484         read_lock(&ipip6_lock);
485         if ((tunnel = ipip6_tunnel_lookup(dev_net(skb->dev),
486                                         iph->saddr, iph->daddr)) != NULL) {
487                 secpath_reset(skb);
488                 skb->mac_header = skb->network_header;
489                 skb_reset_network_header(skb);
490                 IPCB(skb)->flags = 0;
491                 skb->protocol = htons(ETH_P_IPV6);
492                 skb->pkt_type = PACKET_HOST;
493
494                 if ((tunnel->dev->priv_flags & IFF_ISATAP) &&
495                     !isatap_chksrc(skb, iph, tunnel)) {
496                         tunnel->stat.rx_errors++;
497                         read_unlock(&ipip6_lock);
498                         kfree_skb(skb);
499                         return 0;
500                 }
501                 tunnel->stat.rx_packets++;
502                 tunnel->stat.rx_bytes += skb->len;
503                 skb->dev = tunnel->dev;
504                 dst_release(skb->dst);
505                 skb->dst = NULL;
506                 nf_reset(skb);
507                 ipip6_ecn_decapsulate(iph, skb);
508                 netif_rx(skb);
509                 read_unlock(&ipip6_lock);
510                 return 0;
511         }
512
513         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);
514         read_unlock(&ipip6_lock);
515 out:
516         kfree_skb(skb);
517         return 0;
518 }
519
520 /* Returns the embedded IPv4 address if the IPv6 address
521    comes from 6to4 (RFC 3056) addr space */
522
523 static inline __be32 try_6to4(struct in6_addr *v6dst)
524 {
525         __be32 dst = 0;
526
527         if (v6dst->s6_addr16[0] == htons(0x2002)) {
528                 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
529                 memcpy(&dst, &v6dst->s6_addr16[1], 4);
530         }
531         return dst;
532 }
533
534 /*
535  *      This function assumes it is being called from dev_queue_xmit()
536  *      and that skb is filled properly by that function.
537  */
538
539 static int ipip6_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
540 {
541         struct ip_tunnel *tunnel = netdev_priv(dev);
542         struct net_device_stats *stats = &tunnel->stat;
543         struct iphdr  *tiph = &tunnel->parms.iph;
544         struct ipv6hdr *iph6 = ipv6_hdr(skb);
545         u8     tos = tunnel->parms.iph.tos;
546         struct rtable *rt;                      /* Route to the other host */
547         struct net_device *tdev;                        /* Device to other host */
548         struct iphdr  *iph;                     /* Our new IP header */
549         unsigned int max_headroom;              /* The extra header space needed */
550         __be32 dst = tiph->daddr;
551         int    mtu;
552         struct in6_addr *addr6;
553         int addr_type;
554
555         if (tunnel->recursion++) {
556                 tunnel->stat.collisions++;
557                 goto tx_error;
558         }
559
560         if (skb->protocol != htons(ETH_P_IPV6))
561                 goto tx_error;
562
563         /* ISATAP (RFC4214) - must come before 6to4 */
564         if (dev->priv_flags & IFF_ISATAP) {
565                 struct neighbour *neigh = NULL;
566
567                 if (skb->dst)
568                         neigh = skb->dst->neighbour;
569
570                 if (neigh == NULL) {
571                         if (net_ratelimit())
572                                 printk(KERN_DEBUG "sit: nexthop == NULL\n");
573                         goto tx_error;
574                 }
575
576                 addr6 = (struct in6_addr*)&neigh->primary_key;
577                 addr_type = ipv6_addr_type(addr6);
578
579                 if ((addr_type & IPV6_ADDR_UNICAST) &&
580                      ipv6_addr_is_isatap(addr6))
581                         dst = addr6->s6_addr32[3];
582                 else
583                         goto tx_error;
584         }
585
586         if (!dst)
587                 dst = try_6to4(&iph6->daddr);
588
589         if (!dst) {
590                 struct neighbour *neigh = NULL;
591
592                 if (skb->dst)
593                         neigh = skb->dst->neighbour;
594
595                 if (neigh == NULL) {
596                         if (net_ratelimit())
597                                 printk(KERN_DEBUG "sit: nexthop == NULL\n");
598                         goto tx_error;
599                 }
600
601                 addr6 = (struct in6_addr*)&neigh->primary_key;
602                 addr_type = ipv6_addr_type(addr6);
603
604                 if (addr_type == IPV6_ADDR_ANY) {
605                         addr6 = &ipv6_hdr(skb)->daddr;
606                         addr_type = ipv6_addr_type(addr6);
607                 }
608
609                 if ((addr_type & IPV6_ADDR_COMPATv4) == 0)
610                         goto tx_error_icmp;
611
612                 dst = addr6->s6_addr32[3];
613         }
614
615         {
616                 struct flowi fl = { .nl_u = { .ip4_u =
617                                               { .daddr = dst,
618                                                 .saddr = tiph->saddr,
619                                                 .tos = RT_TOS(tos) } },
620                                     .oif = tunnel->parms.link,
621                                     .proto = IPPROTO_IPV6 };
622                 if (ip_route_output_key(dev_net(dev), &rt, &fl)) {
623                         tunnel->stat.tx_carrier_errors++;
624                         goto tx_error_icmp;
625                 }
626         }
627         if (rt->rt_type != RTN_UNICAST) {
628                 ip_rt_put(rt);
629                 tunnel->stat.tx_carrier_errors++;
630                 goto tx_error_icmp;
631         }
632         tdev = rt->u.dst.dev;
633
634         if (tdev == dev) {
635                 ip_rt_put(rt);
636                 tunnel->stat.collisions++;
637                 goto tx_error;
638         }
639
640         if (tiph->frag_off)
641                 mtu = dst_mtu(&rt->u.dst) - sizeof(struct iphdr);
642         else
643                 mtu = skb->dst ? dst_mtu(skb->dst) : dev->mtu;
644
645         if (mtu < 68) {
646                 tunnel->stat.collisions++;
647                 ip_rt_put(rt);
648                 goto tx_error;
649         }
650         if (mtu < IPV6_MIN_MTU)
651                 mtu = IPV6_MIN_MTU;
652         if (tunnel->parms.iph.daddr && skb->dst)
653                 skb->dst->ops->update_pmtu(skb->dst, mtu);
654
655         if (skb->len > mtu) {
656                 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, dev);
657                 ip_rt_put(rt);
658                 goto tx_error;
659         }
660
661         if (tunnel->err_count > 0) {
662                 if (jiffies - tunnel->err_time < IPTUNNEL_ERR_TIMEO) {
663                         tunnel->err_count--;
664                         dst_link_failure(skb);
665                 } else
666                         tunnel->err_count = 0;
667         }
668
669         /*
670          * Okay, now see if we can stuff it in the buffer as-is.
671          */
672         max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr);
673
674         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
675             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
676                 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
677                 if (!new_skb) {
678                         ip_rt_put(rt);
679                         stats->tx_dropped++;
680                         dev_kfree_skb(skb);
681                         tunnel->recursion--;
682                         return 0;
683                 }
684                 if (skb->sk)
685                         skb_set_owner_w(new_skb, skb->sk);
686                 dev_kfree_skb(skb);
687                 skb = new_skb;
688                 iph6 = ipv6_hdr(skb);
689         }
690
691         skb->transport_header = skb->network_header;
692         skb_push(skb, sizeof(struct iphdr));
693         skb_reset_network_header(skb);
694         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
695         IPCB(skb)->flags = 0;
696         dst_release(skb->dst);
697         skb->dst = &rt->u.dst;
698
699         /*
700          *      Push down and install the IPIP header.
701          */
702
703         iph                     =       ip_hdr(skb);
704         iph->version            =       4;
705         iph->ihl                =       sizeof(struct iphdr)>>2;
706         if (mtu > IPV6_MIN_MTU)
707                 iph->frag_off   =       htons(IP_DF);
708         else
709                 iph->frag_off   =       0;
710
711         iph->protocol           =       IPPROTO_IPV6;
712         iph->tos                =       INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
713         iph->daddr              =       rt->rt_dst;
714         iph->saddr              =       rt->rt_src;
715
716         if ((iph->ttl = tiph->ttl) == 0)
717                 iph->ttl        =       iph6->hop_limit;
718
719         nf_reset(skb);
720
721         IPTUNNEL_XMIT();
722         tunnel->recursion--;
723         return 0;
724
725 tx_error_icmp:
726         dst_link_failure(skb);
727 tx_error:
728         stats->tx_errors++;
729         dev_kfree_skb(skb);
730         tunnel->recursion--;
731         return 0;
732 }
733
734 static void ipip6_tunnel_bind_dev(struct net_device *dev)
735 {
736         struct net_device *tdev = NULL;
737         struct ip_tunnel *tunnel;
738         struct iphdr *iph;
739
740         tunnel = netdev_priv(dev);
741         iph = &tunnel->parms.iph;
742
743         if (iph->daddr) {
744                 struct flowi fl = { .nl_u = { .ip4_u =
745                                               { .daddr = iph->daddr,
746                                                 .saddr = iph->saddr,
747                                                 .tos = RT_TOS(iph->tos) } },
748                                     .oif = tunnel->parms.link,
749                                     .proto = IPPROTO_IPV6 };
750                 struct rtable *rt;
751                 if (!ip_route_output_key(dev_net(dev), &rt, &fl)) {
752                         tdev = rt->u.dst.dev;
753                         ip_rt_put(rt);
754                 }
755                 dev->flags |= IFF_POINTOPOINT;
756         }
757
758         if (!tdev && tunnel->parms.link)
759                 tdev = __dev_get_by_index(dev_net(dev), tunnel->parms.link);
760
761         if (tdev) {
762                 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
763                 dev->mtu = tdev->mtu - sizeof(struct iphdr);
764                 if (dev->mtu < IPV6_MIN_MTU)
765                         dev->mtu = IPV6_MIN_MTU;
766         }
767         dev->iflink = tunnel->parms.link;
768 }
769
770 static int
771 ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
772 {
773         int err = 0;
774         struct ip_tunnel_parm p;
775         struct ip_tunnel_prl prl;
776         struct ip_tunnel *t;
777         struct net *net = dev_net(dev);
778         struct sit_net *sitn = net_generic(net, sit_net_id);
779
780         switch (cmd) {
781         case SIOCGETTUNNEL:
782                 t = NULL;
783                 if (dev == sitn->fb_tunnel_dev) {
784                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
785                                 err = -EFAULT;
786                                 break;
787                         }
788                         t = ipip6_tunnel_locate(net, &p, 0);
789                 }
790                 if (t == NULL)
791                         t = netdev_priv(dev);
792                 memcpy(&p, &t->parms, sizeof(p));
793                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
794                         err = -EFAULT;
795                 break;
796
797         case SIOCADDTUNNEL:
798         case SIOCCHGTUNNEL:
799                 err = -EPERM;
800                 if (!capable(CAP_NET_ADMIN))
801                         goto done;
802
803                 err = -EFAULT;
804                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
805                         goto done;
806
807                 err = -EINVAL;
808                 if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPV6 ||
809                     p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
810                         goto done;
811                 if (p.iph.ttl)
812                         p.iph.frag_off |= htons(IP_DF);
813
814                 t = ipip6_tunnel_locate(net, &p, cmd == SIOCADDTUNNEL);
815
816                 if (dev != sitn->fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
817                         if (t != NULL) {
818                                 if (t->dev != dev) {
819                                         err = -EEXIST;
820                                         break;
821                                 }
822                         } else {
823                                 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
824                                     (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
825                                         err = -EINVAL;
826                                         break;
827                                 }
828                                 t = netdev_priv(dev);
829                                 ipip6_tunnel_unlink(sitn, t);
830                                 t->parms.iph.saddr = p.iph.saddr;
831                                 t->parms.iph.daddr = p.iph.daddr;
832                                 memcpy(dev->dev_addr, &p.iph.saddr, 4);
833                                 memcpy(dev->broadcast, &p.iph.daddr, 4);
834                                 ipip6_tunnel_link(sitn, t);
835                                 netdev_state_change(dev);
836                         }
837                 }
838
839                 if (t) {
840                         err = 0;
841                         if (cmd == SIOCCHGTUNNEL) {
842                                 t->parms.iph.ttl = p.iph.ttl;
843                                 t->parms.iph.tos = p.iph.tos;
844                                 if (t->parms.link != p.link) {
845                                         t->parms.link = p.link;
846                                         ipip6_tunnel_bind_dev(dev);
847                                         netdev_state_change(dev);
848                                 }
849                         }
850                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
851                                 err = -EFAULT;
852                 } else
853                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
854                 break;
855
856         case SIOCDELTUNNEL:
857                 err = -EPERM;
858                 if (!capable(CAP_NET_ADMIN))
859                         goto done;
860
861                 if (dev == sitn->fb_tunnel_dev) {
862                         err = -EFAULT;
863                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
864                                 goto done;
865                         err = -ENOENT;
866                         if ((t = ipip6_tunnel_locate(net, &p, 0)) == NULL)
867                                 goto done;
868                         err = -EPERM;
869                         if (t == netdev_priv(sitn->fb_tunnel_dev))
870                                 goto done;
871                         dev = t->dev;
872                 }
873                 unregister_netdevice(dev);
874                 err = 0;
875                 break;
876
877         case SIOCGETPRL:
878                 err = -EINVAL;
879                 if (dev == sitn->fb_tunnel_dev)
880                         goto done;
881                 err = -ENOENT;
882                 if (!(t = netdev_priv(dev)))
883                         goto done;
884                 err = ipip6_tunnel_get_prl(t, ifr->ifr_ifru.ifru_data);
885                 break;
886
887         case SIOCADDPRL:
888         case SIOCDELPRL:
889         case SIOCCHGPRL:
890                 err = -EPERM;
891                 if (!capable(CAP_NET_ADMIN))
892                         goto done;
893                 err = -EINVAL;
894                 if (dev == sitn->fb_tunnel_dev)
895                         goto done;
896                 err = -EFAULT;
897                 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
898                         goto done;
899                 err = -ENOENT;
900                 if (!(t = netdev_priv(dev)))
901                         goto done;
902
903                 switch (cmd) {
904                 case SIOCDELPRL:
905                         err = ipip6_tunnel_del_prl(t, &prl);
906                         break;
907                 case SIOCADDPRL:
908                 case SIOCCHGPRL:
909                         err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
910                         break;
911                 }
912                 netdev_state_change(dev);
913                 break;
914
915         default:
916                 err = -EINVAL;
917         }
918
919 done:
920         return err;
921 }
922
923 static struct net_device_stats *ipip6_tunnel_get_stats(struct net_device *dev)
924 {
925         return &(((struct ip_tunnel*)netdev_priv(dev))->stat);
926 }
927
928 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
929 {
930         if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
931                 return -EINVAL;
932         dev->mtu = new_mtu;
933         return 0;
934 }
935
936 static void ipip6_tunnel_setup(struct net_device *dev)
937 {
938         dev->uninit             = ipip6_tunnel_uninit;
939         dev->destructor         = free_netdev;
940         dev->hard_start_xmit    = ipip6_tunnel_xmit;
941         dev->get_stats          = ipip6_tunnel_get_stats;
942         dev->do_ioctl           = ipip6_tunnel_ioctl;
943         dev->change_mtu         = ipip6_tunnel_change_mtu;
944
945         dev->type               = ARPHRD_SIT;
946         dev->hard_header_len    = LL_MAX_HEADER + sizeof(struct iphdr);
947         dev->mtu                = ETH_DATA_LEN - sizeof(struct iphdr);
948         dev->flags              = IFF_NOARP;
949         dev->iflink             = 0;
950         dev->addr_len           = 4;
951         dev->features           |= NETIF_F_NETNS_LOCAL;
952 }
953
954 static int ipip6_tunnel_init(struct net_device *dev)
955 {
956         struct ip_tunnel *tunnel;
957
958         tunnel = netdev_priv(dev);
959
960         tunnel->dev = dev;
961         strcpy(tunnel->parms.name, dev->name);
962
963         memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
964         memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
965
966         ipip6_tunnel_bind_dev(dev);
967
968         return 0;
969 }
970
971 static int ipip6_fb_tunnel_init(struct net_device *dev)
972 {
973         struct ip_tunnel *tunnel = netdev_priv(dev);
974         struct iphdr *iph = &tunnel->parms.iph;
975         struct net *net = dev_net(dev);
976         struct sit_net *sitn = net_generic(net, sit_net_id);
977
978         tunnel->dev = dev;
979         strcpy(tunnel->parms.name, dev->name);
980
981         iph->version            = 4;
982         iph->protocol           = IPPROTO_IPV6;
983         iph->ihl                = 5;
984         iph->ttl                = 64;
985
986         dev_hold(dev);
987         sitn->tunnels_wc[0]     = tunnel;
988         return 0;
989 }
990
991 static struct xfrm_tunnel sit_handler = {
992         .handler        =       ipip6_rcv,
993         .err_handler    =       ipip6_err,
994         .priority       =       1,
995 };
996
997 static void sit_destroy_tunnels(struct sit_net *sitn)
998 {
999         int prio;
1000
1001         for (prio = 1; prio < 4; prio++) {
1002                 int h;
1003                 for (h = 0; h < HASH_SIZE; h++) {
1004                         struct ip_tunnel *t;
1005                         while ((t = sitn->tunnels[prio][h]) != NULL)
1006                                 unregister_netdevice(t->dev);
1007                 }
1008         }
1009 }
1010
1011 static int sit_init_net(struct net *net)
1012 {
1013         int err;
1014         struct sit_net *sitn;
1015
1016         err = -ENOMEM;
1017         sitn = kzalloc(sizeof(struct sit_net), GFP_KERNEL);
1018         if (sitn == NULL)
1019                 goto err_alloc;
1020
1021         err = net_assign_generic(net, sit_net_id, sitn);
1022         if (err < 0)
1023                 goto err_assign;
1024
1025         sitn->tunnels[0] = sitn->tunnels_wc;
1026         sitn->tunnels[1] = sitn->tunnels_l;
1027         sitn->tunnels[2] = sitn->tunnels_r;
1028         sitn->tunnels[3] = sitn->tunnels_r_l;
1029
1030         sitn->fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1031                                            ipip6_tunnel_setup);
1032         if (!sitn->fb_tunnel_dev) {
1033                 err = -ENOMEM;
1034                 goto err_alloc_dev;
1035         }
1036
1037         sitn->fb_tunnel_dev->init = ipip6_fb_tunnel_init;
1038         dev_net_set(sitn->fb_tunnel_dev, net);
1039
1040         if ((err = register_netdev(sitn->fb_tunnel_dev)))
1041                 goto err_reg_dev;
1042
1043         return 0;
1044
1045 err_reg_dev:
1046         free_netdev(sitn->fb_tunnel_dev);
1047 err_alloc_dev:
1048         /* nothing */
1049 err_assign:
1050         kfree(sitn);
1051 err_alloc:
1052         return err;
1053 }
1054
1055 static void sit_exit_net(struct net *net)
1056 {
1057         struct sit_net *sitn;
1058
1059         sitn = net_generic(net, sit_net_id);
1060         rtnl_lock();
1061         sit_destroy_tunnels(sitn);
1062         unregister_netdevice(sitn->fb_tunnel_dev);
1063         rtnl_unlock();
1064         kfree(sitn);
1065 }
1066
1067 static struct pernet_operations sit_net_ops = {
1068         .init = sit_init_net,
1069         .exit = sit_exit_net,
1070 };
1071
1072 static void __exit sit_cleanup(void)
1073 {
1074         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1075
1076         unregister_pernet_gen_device(sit_net_id, &sit_net_ops);
1077 }
1078
1079 static int __init sit_init(void)
1080 {
1081         int err;
1082
1083         printk(KERN_INFO "IPv6 over IPv4 tunneling driver\n");
1084
1085         if (xfrm4_tunnel_register(&sit_handler, AF_INET6) < 0) {
1086                 printk(KERN_INFO "sit init: Can't add protocol\n");
1087                 return -EAGAIN;
1088         }
1089
1090         err = register_pernet_gen_device(&sit_net_id, &sit_net_ops);
1091         if (err < 0)
1092                 xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1093
1094         return err;
1095 }
1096
1097 module_init(sit_init);
1098 module_exit(sit_cleanup);
1099 MODULE_LICENSE("GPL");
1100 MODULE_ALIAS("sit0");