missing bits of net-namespace / sysctl
[linux-2.6.git] / net / ipv6 / af_inet6.c
1 /*
2  *      PF_INET6 socket protocol family
3  *      Linux INET6 implementation
4  *
5  *      Authors:
6  *      Pedro Roque             <roque@di.fc.ul.pt>
7  *
8  *      Adapted from linux/net/ipv4/af_inet.c
9  *
10  *      Fixes:
11  *      piggy, Karl Knutson     :       Socket protocol table
12  *      Hideaki YOSHIFUJI       :       sin6_scope_id support
13  *      Arnaldo Melo            :       check proc_net_create return, cleanups
14  *
15  *      This program is free software; you can redistribute it and/or
16  *      modify it under the terms of the GNU General Public License
17  *      as published by the Free Software Foundation; either version
18  *      2 of the License, or (at your option) any later version.
19  */
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/in.h>
28 #include <linux/kernel.h>
29 #include <linux/timer.h>
30 #include <linux/string.h>
31 #include <linux/sockios.h>
32 #include <linux/net.h>
33 #include <linux/fcntl.h>
34 #include <linux/mm.h>
35 #include <linux/interrupt.h>
36 #include <linux/proc_fs.h>
37 #include <linux/stat.h>
38 #include <linux/init.h>
39
40 #include <linux/inet.h>
41 #include <linux/netdevice.h>
42 #include <linux/icmpv6.h>
43 #include <linux/netfilter_ipv6.h>
44
45 #include <net/ip.h>
46 #include <net/ipv6.h>
47 #include <net/udp.h>
48 #include <net/udplite.h>
49 #include <net/tcp.h>
50 #include <net/ipip.h>
51 #include <net/protocol.h>
52 #include <net/inet_common.h>
53 #include <net/transp_v6.h>
54 #include <net/ip6_route.h>
55 #include <net/addrconf.h>
56 #ifdef CONFIG_IPV6_TUNNEL
57 #include <net/ip6_tunnel.h>
58 #endif
59
60 #include <asm/uaccess.h>
61 #include <asm/system.h>
62 #include <linux/mroute6.h>
63
64 MODULE_AUTHOR("Cast of dozens");
65 MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
66 MODULE_LICENSE("GPL");
67
68 /* The inetsw6 table contains everything that inet6_create needs to
69  * build a new socket.
70  */
71 static struct list_head inetsw6[SOCK_MAX];
72 static DEFINE_SPINLOCK(inetsw6_lock);
73
74 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
75 {
76         const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
77
78         return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
79 }
80
81 static int inet6_create(struct net *net, struct socket *sock, int protocol)
82 {
83         struct inet_sock *inet;
84         struct ipv6_pinfo *np;
85         struct sock *sk;
86         struct inet_protosw *answer;
87         struct proto *answer_prot;
88         unsigned char answer_flags;
89         char answer_no_check;
90         int try_loading_module = 0;
91         int err;
92
93         if (sock->type != SOCK_RAW &&
94             sock->type != SOCK_DGRAM &&
95             !inet_ehash_secret)
96                 build_ehash_secret();
97
98         /* Look for the requested type/protocol pair. */
99 lookup_protocol:
100         err = -ESOCKTNOSUPPORT;
101         rcu_read_lock();
102         list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
103
104                 err = 0;
105                 /* Check the non-wild match. */
106                 if (protocol == answer->protocol) {
107                         if (protocol != IPPROTO_IP)
108                                 break;
109                 } else {
110                         /* Check for the two wild cases. */
111                         if (IPPROTO_IP == protocol) {
112                                 protocol = answer->protocol;
113                                 break;
114                         }
115                         if (IPPROTO_IP == answer->protocol)
116                                 break;
117                 }
118                 err = -EPROTONOSUPPORT;
119         }
120
121         if (err) {
122                 if (try_loading_module < 2) {
123                         rcu_read_unlock();
124                         /*
125                          * Be more specific, e.g. net-pf-10-proto-132-type-1
126                          * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
127                          */
128                         if (++try_loading_module == 1)
129                                 request_module("net-pf-%d-proto-%d-type-%d",
130                                                 PF_INET6, protocol, sock->type);
131                         /*
132                          * Fall back to generic, e.g. net-pf-10-proto-132
133                          * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
134                          */
135                         else
136                                 request_module("net-pf-%d-proto-%d",
137                                                 PF_INET6, protocol);
138                         goto lookup_protocol;
139                 } else
140                         goto out_rcu_unlock;
141         }
142
143         err = -EPERM;
144         if (answer->capability > 0 && !capable(answer->capability))
145                 goto out_rcu_unlock;
146
147         sock->ops = answer->ops;
148         answer_prot = answer->prot;
149         answer_no_check = answer->no_check;
150         answer_flags = answer->flags;
151         rcu_read_unlock();
152
153         WARN_ON(answer_prot->slab == NULL);
154
155         err = -ENOBUFS;
156         sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot);
157         if (sk == NULL)
158                 goto out;
159
160         sock_init_data(sock, sk);
161
162         err = 0;
163         sk->sk_no_check = answer_no_check;
164         if (INET_PROTOSW_REUSE & answer_flags)
165                 sk->sk_reuse = 1;
166
167         inet = inet_sk(sk);
168         inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
169
170         if (SOCK_RAW == sock->type) {
171                 inet->num = protocol;
172                 if (IPPROTO_RAW == protocol)
173                         inet->hdrincl = 1;
174         }
175
176         sk->sk_destruct         = inet_sock_destruct;
177         sk->sk_family           = PF_INET6;
178         sk->sk_protocol         = protocol;
179
180         sk->sk_backlog_rcv      = answer->prot->backlog_rcv;
181
182         inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
183         np->hop_limit   = -1;
184         np->mcast_hops  = -1;
185         np->mc_loop     = 1;
186         np->pmtudisc    = IPV6_PMTUDISC_WANT;
187         np->ipv6only    = net->ipv6.sysctl.bindv6only;
188
189         /* Init the ipv4 part of the socket since we can have sockets
190          * using v6 API for ipv4.
191          */
192         inet->uc_ttl    = -1;
193
194         inet->mc_loop   = 1;
195         inet->mc_ttl    = 1;
196         inet->mc_index  = 0;
197         inet->mc_list   = NULL;
198
199         if (ipv4_config.no_pmtu_disc)
200                 inet->pmtudisc = IP_PMTUDISC_DONT;
201         else
202                 inet->pmtudisc = IP_PMTUDISC_WANT;
203         /*
204          * Increment only the relevant sk_prot->socks debug field, this changes
205          * the previous behaviour of incrementing both the equivalent to
206          * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
207          *
208          * This allows better debug granularity as we'll know exactly how many
209          * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
210          * transport protocol socks. -acme
211          */
212         sk_refcnt_debug_inc(sk);
213
214         if (inet->num) {
215                 /* It assumes that any protocol which allows
216                  * the user to assign a number at socket
217                  * creation time automatically shares.
218                  */
219                 inet->sport = htons(inet->num);
220                 sk->sk_prot->hash(sk);
221         }
222         if (sk->sk_prot->init) {
223                 err = sk->sk_prot->init(sk);
224                 if (err) {
225                         sk_common_release(sk);
226                         goto out;
227                 }
228         }
229 out:
230         return err;
231 out_rcu_unlock:
232         rcu_read_unlock();
233         goto out;
234 }
235
236
237 /* bind for INET6 API */
238 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
239 {
240         struct sockaddr_in6 *addr=(struct sockaddr_in6 *)uaddr;
241         struct sock *sk = sock->sk;
242         struct inet_sock *inet = inet_sk(sk);
243         struct ipv6_pinfo *np = inet6_sk(sk);
244         struct net *net = sock_net(sk);
245         __be32 v4addr = 0;
246         unsigned short snum;
247         int addr_type = 0;
248         int err = 0;
249
250         /* If the socket has its own bind function then use it. */
251         if (sk->sk_prot->bind)
252                 return sk->sk_prot->bind(sk, uaddr, addr_len);
253
254         if (addr_len < SIN6_LEN_RFC2133)
255                 return -EINVAL;
256         addr_type = ipv6_addr_type(&addr->sin6_addr);
257         if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
258                 return -EINVAL;
259
260         snum = ntohs(addr->sin6_port);
261         if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
262                 return -EACCES;
263
264         lock_sock(sk);
265
266         /* Check these errors (active socket, double bind). */
267         if (sk->sk_state != TCP_CLOSE || inet->num) {
268                 err = -EINVAL;
269                 goto out;
270         }
271
272         /* Check if the address belongs to the host. */
273         if (addr_type == IPV6_ADDR_MAPPED) {
274                 v4addr = addr->sin6_addr.s6_addr32[3];
275                 if (inet_addr_type(net, v4addr) != RTN_LOCAL) {
276                         err = -EADDRNOTAVAIL;
277                         goto out;
278                 }
279         } else {
280                 if (addr_type != IPV6_ADDR_ANY) {
281                         struct net_device *dev = NULL;
282
283                         if (addr_type & IPV6_ADDR_LINKLOCAL) {
284                                 if (addr_len >= sizeof(struct sockaddr_in6) &&
285                                     addr->sin6_scope_id) {
286                                         /* Override any existing binding, if another one
287                                          * is supplied by user.
288                                          */
289                                         sk->sk_bound_dev_if = addr->sin6_scope_id;
290                                 }
291
292                                 /* Binding to link-local address requires an interface */
293                                 if (!sk->sk_bound_dev_if) {
294                                         err = -EINVAL;
295                                         goto out;
296                                 }
297                                 dev = dev_get_by_index(net, sk->sk_bound_dev_if);
298                                 if (!dev) {
299                                         err = -ENODEV;
300                                         goto out;
301                                 }
302                         }
303
304                         /* ipv4 addr of the socket is invalid.  Only the
305                          * unspecified and mapped address have a v4 equivalent.
306                          */
307                         v4addr = LOOPBACK4_IPV6;
308                         if (!(addr_type & IPV6_ADDR_MULTICAST)) {
309                                 if (!ipv6_chk_addr(net, &addr->sin6_addr,
310                                                    dev, 0)) {
311                                         if (dev)
312                                                 dev_put(dev);
313                                         err = -EADDRNOTAVAIL;
314                                         goto out;
315                                 }
316                         }
317                         if (dev)
318                                 dev_put(dev);
319                 }
320         }
321
322         inet->rcv_saddr = v4addr;
323         inet->saddr = v4addr;
324
325         ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
326
327         if (!(addr_type & IPV6_ADDR_MULTICAST))
328                 ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
329
330         /* Make sure we are allowed to bind here. */
331         if (sk->sk_prot->get_port(sk, snum)) {
332                 inet_reset_saddr(sk);
333                 err = -EADDRINUSE;
334                 goto out;
335         }
336
337         if (addr_type != IPV6_ADDR_ANY)
338                 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
339         if (snum)
340                 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
341         inet->sport = htons(inet->num);
342         inet->dport = 0;
343         inet->daddr = 0;
344 out:
345         release_sock(sk);
346         return err;
347 }
348
349 EXPORT_SYMBOL(inet6_bind);
350
351 int inet6_release(struct socket *sock)
352 {
353         struct sock *sk = sock->sk;
354
355         if (sk == NULL)
356                 return -EINVAL;
357
358         /* Free mc lists */
359         ipv6_sock_mc_close(sk);
360
361         /* Free ac lists */
362         ipv6_sock_ac_close(sk);
363
364         return inet_release(sock);
365 }
366
367 EXPORT_SYMBOL(inet6_release);
368
369 void inet6_destroy_sock(struct sock *sk)
370 {
371         struct ipv6_pinfo *np = inet6_sk(sk);
372         struct sk_buff *skb;
373         struct ipv6_txoptions *opt;
374
375         /* Release rx options */
376
377         if ((skb = xchg(&np->pktoptions, NULL)) != NULL)
378                 kfree_skb(skb);
379
380         /* Free flowlabels */
381         fl6_free_socklist(sk);
382
383         /* Free tx options */
384
385         if ((opt = xchg(&np->opt, NULL)) != NULL)
386                 sock_kfree_s(sk, opt, opt->tot_len);
387 }
388
389 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
390
391 /*
392  *      This does both peername and sockname.
393  */
394
395 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
396                  int *uaddr_len, int peer)
397 {
398         struct sockaddr_in6 *sin=(struct sockaddr_in6 *)uaddr;
399         struct sock *sk = sock->sk;
400         struct inet_sock *inet = inet_sk(sk);
401         struct ipv6_pinfo *np = inet6_sk(sk);
402
403         sin->sin6_family = AF_INET6;
404         sin->sin6_flowinfo = 0;
405         sin->sin6_scope_id = 0;
406         if (peer) {
407                 if (!inet->dport)
408                         return -ENOTCONN;
409                 if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
410                     peer == 1)
411                         return -ENOTCONN;
412                 sin->sin6_port = inet->dport;
413                 ipv6_addr_copy(&sin->sin6_addr, &np->daddr);
414                 if (np->sndflow)
415                         sin->sin6_flowinfo = np->flow_label;
416         } else {
417                 if (ipv6_addr_any(&np->rcv_saddr))
418                         ipv6_addr_copy(&sin->sin6_addr, &np->saddr);
419                 else
420                         ipv6_addr_copy(&sin->sin6_addr, &np->rcv_saddr);
421
422                 sin->sin6_port = inet->sport;
423         }
424         if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL)
425                 sin->sin6_scope_id = sk->sk_bound_dev_if;
426         *uaddr_len = sizeof(*sin);
427         return(0);
428 }
429
430 EXPORT_SYMBOL(inet6_getname);
431
432 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
433 {
434         struct sock *sk = sock->sk;
435         struct net *net = sock_net(sk);
436
437         switch(cmd)
438         {
439         case SIOCGSTAMP:
440                 return sock_get_timestamp(sk, (struct timeval __user *)arg);
441
442         case SIOCGSTAMPNS:
443                 return sock_get_timestampns(sk, (struct timespec __user *)arg);
444
445         case SIOCADDRT:
446         case SIOCDELRT:
447
448                 return(ipv6_route_ioctl(net, cmd, (void __user *)arg));
449
450         case SIOCSIFADDR:
451                 return addrconf_add_ifaddr(net, (void __user *) arg);
452         case SIOCDIFADDR:
453                 return addrconf_del_ifaddr(net, (void __user *) arg);
454         case SIOCSIFDSTADDR:
455                 return addrconf_set_dstaddr(net, (void __user *) arg);
456         default:
457                 if (!sk->sk_prot->ioctl)
458                         return -ENOIOCTLCMD;
459                 return sk->sk_prot->ioctl(sk, cmd, arg);
460         }
461         /*NOTREACHED*/
462         return(0);
463 }
464
465 EXPORT_SYMBOL(inet6_ioctl);
466
467 const struct proto_ops inet6_stream_ops = {
468         .family            = PF_INET6,
469         .owner             = THIS_MODULE,
470         .release           = inet6_release,
471         .bind              = inet6_bind,
472         .connect           = inet_stream_connect,       /* ok           */
473         .socketpair        = sock_no_socketpair,        /* a do nothing */
474         .accept            = inet_accept,               /* ok           */
475         .getname           = inet6_getname,
476         .poll              = tcp_poll,                  /* ok           */
477         .ioctl             = inet6_ioctl,               /* must change  */
478         .listen            = inet_listen,               /* ok           */
479         .shutdown          = inet_shutdown,             /* ok           */
480         .setsockopt        = sock_common_setsockopt,    /* ok           */
481         .getsockopt        = sock_common_getsockopt,    /* ok           */
482         .sendmsg           = tcp_sendmsg,               /* ok           */
483         .recvmsg           = sock_common_recvmsg,       /* ok           */
484         .mmap              = sock_no_mmap,
485         .sendpage          = tcp_sendpage,
486         .splice_read       = tcp_splice_read,
487 #ifdef CONFIG_COMPAT
488         .compat_setsockopt = compat_sock_common_setsockopt,
489         .compat_getsockopt = compat_sock_common_getsockopt,
490 #endif
491 };
492
493 const struct proto_ops inet6_dgram_ops = {
494         .family            = PF_INET6,
495         .owner             = THIS_MODULE,
496         .release           = inet6_release,
497         .bind              = inet6_bind,
498         .connect           = inet_dgram_connect,        /* ok           */
499         .socketpair        = sock_no_socketpair,        /* a do nothing */
500         .accept            = sock_no_accept,            /* a do nothing */
501         .getname           = inet6_getname,
502         .poll              = udp_poll,                  /* ok           */
503         .ioctl             = inet6_ioctl,               /* must change  */
504         .listen            = sock_no_listen,            /* ok           */
505         .shutdown          = inet_shutdown,             /* ok           */
506         .setsockopt        = sock_common_setsockopt,    /* ok           */
507         .getsockopt        = sock_common_getsockopt,    /* ok           */
508         .sendmsg           = inet_sendmsg,              /* ok           */
509         .recvmsg           = sock_common_recvmsg,       /* ok           */
510         .mmap              = sock_no_mmap,
511         .sendpage          = sock_no_sendpage,
512 #ifdef CONFIG_COMPAT
513         .compat_setsockopt = compat_sock_common_setsockopt,
514         .compat_getsockopt = compat_sock_common_getsockopt,
515 #endif
516 };
517
518 static struct net_proto_family inet6_family_ops = {
519         .family = PF_INET6,
520         .create = inet6_create,
521         .owner  = THIS_MODULE,
522 };
523
524 int inet6_register_protosw(struct inet_protosw *p)
525 {
526         struct list_head *lh;
527         struct inet_protosw *answer;
528         struct list_head *last_perm;
529         int protocol = p->protocol;
530         int ret;
531
532         spin_lock_bh(&inetsw6_lock);
533
534         ret = -EINVAL;
535         if (p->type >= SOCK_MAX)
536                 goto out_illegal;
537
538         /* If we are trying to override a permanent protocol, bail. */
539         answer = NULL;
540         ret = -EPERM;
541         last_perm = &inetsw6[p->type];
542         list_for_each(lh, &inetsw6[p->type]) {
543                 answer = list_entry(lh, struct inet_protosw, list);
544
545                 /* Check only the non-wild match. */
546                 if (INET_PROTOSW_PERMANENT & answer->flags) {
547                         if (protocol == answer->protocol)
548                                 break;
549                         last_perm = lh;
550                 }
551
552                 answer = NULL;
553         }
554         if (answer)
555                 goto out_permanent;
556
557         /* Add the new entry after the last permanent entry if any, so that
558          * the new entry does not override a permanent entry when matched with
559          * a wild-card protocol. But it is allowed to override any existing
560          * non-permanent entry.  This means that when we remove this entry, the
561          * system automatically returns to the old behavior.
562          */
563         list_add_rcu(&p->list, last_perm);
564         ret = 0;
565 out:
566         spin_unlock_bh(&inetsw6_lock);
567         return ret;
568
569 out_permanent:
570         printk(KERN_ERR "Attempt to override permanent protocol %d.\n",
571                protocol);
572         goto out;
573
574 out_illegal:
575         printk(KERN_ERR
576                "Ignoring attempt to register invalid socket type %d.\n",
577                p->type);
578         goto out;
579 }
580
581 EXPORT_SYMBOL(inet6_register_protosw);
582
583 void
584 inet6_unregister_protosw(struct inet_protosw *p)
585 {
586         if (INET_PROTOSW_PERMANENT & p->flags) {
587                 printk(KERN_ERR
588                        "Attempt to unregister permanent protocol %d.\n",
589                        p->protocol);
590         } else {
591                 spin_lock_bh(&inetsw6_lock);
592                 list_del_rcu(&p->list);
593                 spin_unlock_bh(&inetsw6_lock);
594
595                 synchronize_net();
596         }
597 }
598
599 EXPORT_SYMBOL(inet6_unregister_protosw);
600
601 int inet6_sk_rebuild_header(struct sock *sk)
602 {
603         int err;
604         struct dst_entry *dst;
605         struct ipv6_pinfo *np = inet6_sk(sk);
606
607         dst = __sk_dst_check(sk, np->dst_cookie);
608
609         if (dst == NULL) {
610                 struct inet_sock *inet = inet_sk(sk);
611                 struct in6_addr *final_p = NULL, final;
612                 struct flowi fl;
613
614                 memset(&fl, 0, sizeof(fl));
615                 fl.proto = sk->sk_protocol;
616                 ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
617                 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
618                 fl.fl6_flowlabel = np->flow_label;
619                 fl.oif = sk->sk_bound_dev_if;
620                 fl.fl_ip_dport = inet->dport;
621                 fl.fl_ip_sport = inet->sport;
622                 security_sk_classify_flow(sk, &fl);
623
624                 if (np->opt && np->opt->srcrt) {
625                         struct rt0_hdr *rt0 = (struct rt0_hdr *) np->opt->srcrt;
626                         ipv6_addr_copy(&final, &fl.fl6_dst);
627                         ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
628                         final_p = &final;
629                 }
630
631                 err = ip6_dst_lookup(sk, &dst, &fl);
632                 if (err) {
633                         sk->sk_route_caps = 0;
634                         return err;
635                 }
636                 if (final_p)
637                         ipv6_addr_copy(&fl.fl6_dst, final_p);
638
639                 if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0) {
640                         sk->sk_err_soft = -err;
641                         return err;
642                 }
643
644                 __ip6_dst_store(sk, dst, NULL, NULL);
645         }
646
647         return 0;
648 }
649
650 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
651
652 int ipv6_opt_accepted(struct sock *sk, struct sk_buff *skb)
653 {
654         struct ipv6_pinfo *np = inet6_sk(sk);
655         struct inet6_skb_parm *opt = IP6CB(skb);
656
657         if (np->rxopt.all) {
658                 if ((opt->hop && (np->rxopt.bits.hopopts ||
659                                   np->rxopt.bits.ohopopts)) ||
660                     ((IPV6_FLOWINFO_MASK &
661                       *(__be32 *)skb_network_header(skb)) &&
662                      np->rxopt.bits.rxflow) ||
663                     (opt->srcrt && (np->rxopt.bits.srcrt ||
664                      np->rxopt.bits.osrcrt)) ||
665                     ((opt->dst1 || opt->dst0) &&
666                      (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
667                         return 1;
668         }
669         return 0;
670 }
671
672 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
673
674 static struct inet6_protocol *ipv6_gso_pull_exthdrs(struct sk_buff *skb,
675                                                     int proto)
676 {
677         struct inet6_protocol *ops = NULL;
678
679         for (;;) {
680                 struct ipv6_opt_hdr *opth;
681                 int len;
682
683                 if (proto != NEXTHDR_HOP) {
684                         ops = rcu_dereference(inet6_protos[proto]);
685
686                         if (unlikely(!ops))
687                                 break;
688
689                         if (!(ops->flags & INET6_PROTO_GSO_EXTHDR))
690                                 break;
691                 }
692
693                 if (unlikely(!pskb_may_pull(skb, 8)))
694                         break;
695
696                 opth = (void *)skb->data;
697                 len = ipv6_optlen(opth);
698
699                 if (unlikely(!pskb_may_pull(skb, len)))
700                         break;
701
702                 proto = opth->nexthdr;
703                 __skb_pull(skb, len);
704         }
705
706         return ops;
707 }
708
709 static int ipv6_gso_send_check(struct sk_buff *skb)
710 {
711         struct ipv6hdr *ipv6h;
712         struct inet6_protocol *ops;
713         int err = -EINVAL;
714
715         if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
716                 goto out;
717
718         ipv6h = ipv6_hdr(skb);
719         __skb_pull(skb, sizeof(*ipv6h));
720         err = -EPROTONOSUPPORT;
721
722         rcu_read_lock();
723         ops = ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr);
724         if (likely(ops && ops->gso_send_check)) {
725                 skb_reset_transport_header(skb);
726                 err = ops->gso_send_check(skb);
727         }
728         rcu_read_unlock();
729
730 out:
731         return err;
732 }
733
734 static struct sk_buff *ipv6_gso_segment(struct sk_buff *skb, int features)
735 {
736         struct sk_buff *segs = ERR_PTR(-EINVAL);
737         struct ipv6hdr *ipv6h;
738         struct inet6_protocol *ops;
739
740         if (!(features & NETIF_F_V6_CSUM))
741                 features &= ~NETIF_F_SG;
742
743         if (unlikely(skb_shinfo(skb)->gso_type &
744                      ~(SKB_GSO_UDP |
745                        SKB_GSO_DODGY |
746                        SKB_GSO_TCP_ECN |
747                        SKB_GSO_TCPV6 |
748                        0)))
749                 goto out;
750
751         if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
752                 goto out;
753
754         ipv6h = ipv6_hdr(skb);
755         __skb_pull(skb, sizeof(*ipv6h));
756         segs = ERR_PTR(-EPROTONOSUPPORT);
757
758         rcu_read_lock();
759         ops = ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr);
760         if (likely(ops && ops->gso_segment)) {
761                 skb_reset_transport_header(skb);
762                 segs = ops->gso_segment(skb, features);
763         }
764         rcu_read_unlock();
765
766         if (unlikely(IS_ERR(segs)))
767                 goto out;
768
769         for (skb = segs; skb; skb = skb->next) {
770                 ipv6h = ipv6_hdr(skb);
771                 ipv6h->payload_len = htons(skb->len - skb->mac_len -
772                                            sizeof(*ipv6h));
773         }
774
775 out:
776         return segs;
777 }
778
779 static struct packet_type ipv6_packet_type = {
780         .type = __constant_htons(ETH_P_IPV6),
781         .func = ipv6_rcv,
782         .gso_send_check = ipv6_gso_send_check,
783         .gso_segment = ipv6_gso_segment,
784 };
785
786 static int __init ipv6_packet_init(void)
787 {
788         dev_add_pack(&ipv6_packet_type);
789         return 0;
790 }
791
792 static void ipv6_packet_cleanup(void)
793 {
794         dev_remove_pack(&ipv6_packet_type);
795 }
796
797 static int __init init_ipv6_mibs(void)
798 {
799         if (snmp_mib_init((void **)ipv6_statistics,
800                           sizeof(struct ipstats_mib)) < 0)
801                 goto err_ip_mib;
802         if (snmp_mib_init((void **)icmpv6_statistics,
803                           sizeof(struct icmpv6_mib)) < 0)
804                 goto err_icmp_mib;
805         if (snmp_mib_init((void **)icmpv6msg_statistics,
806                           sizeof(struct icmpv6msg_mib)) < 0)
807                 goto err_icmpmsg_mib;
808         if (snmp_mib_init((void **)udp_stats_in6, sizeof (struct udp_mib)) < 0)
809                 goto err_udp_mib;
810         if (snmp_mib_init((void **)udplite_stats_in6,
811                           sizeof (struct udp_mib)) < 0)
812                 goto err_udplite_mib;
813         return 0;
814
815 err_udplite_mib:
816         snmp_mib_free((void **)udp_stats_in6);
817 err_udp_mib:
818         snmp_mib_free((void **)icmpv6msg_statistics);
819 err_icmpmsg_mib:
820         snmp_mib_free((void **)icmpv6_statistics);
821 err_icmp_mib:
822         snmp_mib_free((void **)ipv6_statistics);
823 err_ip_mib:
824         return -ENOMEM;
825
826 }
827
828 static void cleanup_ipv6_mibs(void)
829 {
830         snmp_mib_free((void **)ipv6_statistics);
831         snmp_mib_free((void **)icmpv6_statistics);
832         snmp_mib_free((void **)icmpv6msg_statistics);
833         snmp_mib_free((void **)udp_stats_in6);
834         snmp_mib_free((void **)udplite_stats_in6);
835 }
836
837 static int inet6_net_init(struct net *net)
838 {
839         int err = 0;
840
841         net->ipv6.sysctl.bindv6only = 0;
842         net->ipv6.sysctl.flush_delay = 0;
843         net->ipv6.sysctl.ip6_rt_max_size = 4096;
844         net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
845         net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
846         net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
847         net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
848         net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
849         net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
850         net->ipv6.sysctl.icmpv6_time = 1*HZ;
851
852 #ifdef CONFIG_PROC_FS
853         err = udp6_proc_init(net);
854         if (err)
855                 goto out;
856         err = tcp6_proc_init(net);
857         if (err)
858                 goto proc_tcp6_fail;
859         err = ac6_proc_init(net);
860         if (err)
861                 goto proc_ac6_fail;
862 out:
863 #endif
864         return err;
865
866 #ifdef CONFIG_PROC_FS
867 proc_ac6_fail:
868         tcp6_proc_exit(net);
869 proc_tcp6_fail:
870         udp6_proc_exit(net);
871         goto out;
872 #endif
873 }
874
875 static void inet6_net_exit(struct net *net)
876 {
877 #ifdef CONFIG_PROC_FS
878         udp6_proc_exit(net);
879         tcp6_proc_exit(net);
880         ac6_proc_exit(net);
881 #endif
882 }
883
884 static struct pernet_operations inet6_net_ops = {
885         .init = inet6_net_init,
886         .exit = inet6_net_exit,
887 };
888
889 static int __init inet6_init(void)
890 {
891         struct sk_buff *dummy_skb;
892         struct list_head *r;
893         int err;
894
895         BUILD_BUG_ON(sizeof(struct inet6_skb_parm) > sizeof(dummy_skb->cb));
896
897         err = proto_register(&tcpv6_prot, 1);
898         if (err)
899                 goto out;
900
901         err = proto_register(&udpv6_prot, 1);
902         if (err)
903                 goto out_unregister_tcp_proto;
904
905         err = proto_register(&udplitev6_prot, 1);
906         if (err)
907                 goto out_unregister_udp_proto;
908
909         err = proto_register(&rawv6_prot, 1);
910         if (err)
911                 goto out_unregister_udplite_proto;
912
913
914         /* Register the socket-side information for inet6_create.  */
915         for(r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
916                 INIT_LIST_HEAD(r);
917
918         /* We MUST register RAW sockets before we create the ICMP6,
919          * IGMP6, or NDISC control sockets.
920          */
921         err = rawv6_init();
922         if (err)
923                 goto out_unregister_raw_proto;
924
925         /* Register the family here so that the init calls below will
926          * be able to create sockets. (?? is this dangerous ??)
927          */
928         err = sock_register(&inet6_family_ops);
929         if (err)
930                 goto out_sock_register_fail;
931
932         /* Initialise ipv6 mibs */
933         err = init_ipv6_mibs();
934         if (err)
935                 goto out_unregister_sock;
936
937 #ifdef CONFIG_SYSCTL
938         err = ipv6_static_sysctl_register();
939         if (err)
940                 goto static_sysctl_fail;
941 #endif
942         /*
943          *      ipngwg API draft makes clear that the correct semantics
944          *      for TCP and UDP is to consider one TCP and UDP instance
945          *      in a host availiable by both INET and INET6 APIs and
946          *      able to communicate via both network protocols.
947          */
948
949         err = register_pernet_subsys(&inet6_net_ops);
950         if (err)
951                 goto register_pernet_fail;
952         err = icmpv6_init();
953         if (err)
954                 goto icmp_fail;
955         err = ip6_mr_init();
956         if (err)
957                 goto ipmr_fail;
958         err = ndisc_init();
959         if (err)
960                 goto ndisc_fail;
961         err = igmp6_init();
962         if (err)
963                 goto igmp_fail;
964         err = ipv6_netfilter_init();
965         if (err)
966                 goto netfilter_fail;
967         /* Create /proc/foo6 entries. */
968 #ifdef CONFIG_PROC_FS
969         err = -ENOMEM;
970         if (raw6_proc_init())
971                 goto proc_raw6_fail;
972         if (udplite6_proc_init())
973                 goto proc_udplite6_fail;
974         if (ipv6_misc_proc_init())
975                 goto proc_misc6_fail;
976         if (if6_proc_init())
977                 goto proc_if6_fail;
978 #endif
979         err = ip6_route_init();
980         if (err)
981                 goto ip6_route_fail;
982         err = ip6_flowlabel_init();
983         if (err)
984                 goto ip6_flowlabel_fail;
985         err = addrconf_init();
986         if (err)
987                 goto addrconf_fail;
988
989         /* Init v6 extension headers. */
990         err = ipv6_exthdrs_init();
991         if (err)
992                 goto ipv6_exthdrs_fail;
993
994         err = ipv6_frag_init();
995         if (err)
996                 goto ipv6_frag_fail;
997
998         /* Init v6 transport protocols. */
999         err = udpv6_init();
1000         if (err)
1001                 goto udpv6_fail;
1002
1003         err = udplitev6_init();
1004         if (err)
1005                 goto udplitev6_fail;
1006
1007         err = tcpv6_init();
1008         if (err)
1009                 goto tcpv6_fail;
1010
1011         err = ipv6_packet_init();
1012         if (err)
1013                 goto ipv6_packet_fail;
1014
1015 #ifdef CONFIG_SYSCTL
1016         err = ipv6_sysctl_register();
1017         if (err)
1018                 goto sysctl_fail;
1019 #endif
1020 out:
1021         return err;
1022
1023 #ifdef CONFIG_SYSCTL
1024 sysctl_fail:
1025         ipv6_packet_cleanup();
1026 #endif
1027 ipv6_packet_fail:
1028         tcpv6_exit();
1029 tcpv6_fail:
1030         udplitev6_exit();
1031 udplitev6_fail:
1032         udpv6_exit();
1033 udpv6_fail:
1034         ipv6_frag_exit();
1035 ipv6_frag_fail:
1036         ipv6_exthdrs_exit();
1037 ipv6_exthdrs_fail:
1038         addrconf_cleanup();
1039 addrconf_fail:
1040         ip6_flowlabel_cleanup();
1041 ip6_flowlabel_fail:
1042         ip6_route_cleanup();
1043 ip6_route_fail:
1044 #ifdef CONFIG_PROC_FS
1045         if6_proc_exit();
1046 proc_if6_fail:
1047         ipv6_misc_proc_exit();
1048 proc_misc6_fail:
1049         udplite6_proc_exit();
1050 proc_udplite6_fail:
1051         raw6_proc_exit();
1052 proc_raw6_fail:
1053 #endif
1054         ipv6_netfilter_fini();
1055 netfilter_fail:
1056         igmp6_cleanup();
1057 igmp_fail:
1058         ndisc_cleanup();
1059 ndisc_fail:
1060         ip6_mr_cleanup();
1061 ipmr_fail:
1062         icmpv6_cleanup();
1063 icmp_fail:
1064         unregister_pernet_subsys(&inet6_net_ops);
1065 register_pernet_fail:
1066 #ifdef CONFIG_SYSCTL
1067         ipv6_static_sysctl_unregister();
1068 static_sysctl_fail:
1069 #endif
1070         cleanup_ipv6_mibs();
1071 out_unregister_sock:
1072         sock_unregister(PF_INET6);
1073         rtnl_unregister_all(PF_INET6);
1074 out_sock_register_fail:
1075         rawv6_exit();
1076 out_unregister_raw_proto:
1077         proto_unregister(&rawv6_prot);
1078 out_unregister_udplite_proto:
1079         proto_unregister(&udplitev6_prot);
1080 out_unregister_udp_proto:
1081         proto_unregister(&udpv6_prot);
1082 out_unregister_tcp_proto:
1083         proto_unregister(&tcpv6_prot);
1084         goto out;
1085 }
1086 module_init(inet6_init);
1087
1088 static void __exit inet6_exit(void)
1089 {
1090         /* First of all disallow new sockets creation. */
1091         sock_unregister(PF_INET6);
1092         /* Disallow any further netlink messages */
1093         rtnl_unregister_all(PF_INET6);
1094
1095 #ifdef CONFIG_SYSCTL
1096         ipv6_sysctl_unregister();
1097 #endif
1098         udpv6_exit();
1099         udplitev6_exit();
1100         tcpv6_exit();
1101
1102         /* Cleanup code parts. */
1103         ipv6_packet_cleanup();
1104         ipv6_frag_exit();
1105         ipv6_exthdrs_exit();
1106         addrconf_cleanup();
1107         ip6_flowlabel_cleanup();
1108         ip6_route_cleanup();
1109 #ifdef CONFIG_PROC_FS
1110
1111         /* Cleanup code parts. */
1112         if6_proc_exit();
1113         ipv6_misc_proc_exit();
1114         udplite6_proc_exit();
1115         raw6_proc_exit();
1116 #endif
1117         ipv6_netfilter_fini();
1118         igmp6_cleanup();
1119         ndisc_cleanup();
1120         ip6_mr_cleanup();
1121         icmpv6_cleanup();
1122         rawv6_exit();
1123
1124         unregister_pernet_subsys(&inet6_net_ops);
1125 #ifdef CONFIG_SYSCTL
1126         ipv6_static_sysctl_unregister();
1127 #endif
1128         cleanup_ipv6_mibs();
1129         proto_unregister(&rawv6_prot);
1130         proto_unregister(&udplitev6_prot);
1131         proto_unregister(&udpv6_prot);
1132         proto_unregister(&tcpv6_prot);
1133 }
1134 module_exit(inet6_exit);
1135
1136 MODULE_ALIAS_NETPROTO(PF_INET6);