Merge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/kaber/nf-next-2.6
[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 #include <linux/slab.h>
40
41 #include <linux/inet.h>
42 #include <linux/netdevice.h>
43 #include <linux/icmpv6.h>
44 #include <linux/netfilter_ipv6.h>
45
46 #include <net/ip.h>
47 #include <net/ipv6.h>
48 #include <net/udp.h>
49 #include <net/udplite.h>
50 #include <net/tcp.h>
51 #include <net/ipip.h>
52 #include <net/protocol.h>
53 #include <net/inet_common.h>
54 #include <net/route.h>
55 #include <net/transp_v6.h>
56 #include <net/ip6_route.h>
57 #include <net/addrconf.h>
58 #ifdef CONFIG_IPV6_TUNNEL
59 #include <net/ip6_tunnel.h>
60 #endif
61
62 #include <asm/uaccess.h>
63 #include <asm/system.h>
64 #include <linux/mroute6.h>
65
66 MODULE_AUTHOR("Cast of dozens");
67 MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
68 MODULE_LICENSE("GPL");
69
70 /* The inetsw6 table contains everything that inet6_create needs to
71  * build a new socket.
72  */
73 static struct list_head inetsw6[SOCK_MAX];
74 static DEFINE_SPINLOCK(inetsw6_lock);
75
76 struct ipv6_params ipv6_defaults = {
77         .disable_ipv6 = 0,
78         .autoconf = 1,
79 };
80
81 static int disable_ipv6_mod = 0;
82
83 module_param_named(disable, disable_ipv6_mod, int, 0444);
84 MODULE_PARM_DESC(disable, "Disable IPv6 module such that it is non-functional");
85
86 module_param_named(disable_ipv6, ipv6_defaults.disable_ipv6, int, 0444);
87 MODULE_PARM_DESC(disable_ipv6, "Disable IPv6 on all interfaces");
88
89 module_param_named(autoconf, ipv6_defaults.autoconf, int, 0444);
90 MODULE_PARM_DESC(autoconf, "Enable IPv6 address autoconfiguration on all interfaces");
91
92 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
93 {
94         const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
95
96         return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
97 }
98
99 static int inet6_create(struct net *net, struct socket *sock, int protocol,
100                         int kern)
101 {
102         struct inet_sock *inet;
103         struct ipv6_pinfo *np;
104         struct sock *sk;
105         struct inet_protosw *answer;
106         struct proto *answer_prot;
107         unsigned char answer_flags;
108         char answer_no_check;
109         int try_loading_module = 0;
110         int err;
111
112         if (sock->type != SOCK_RAW &&
113             sock->type != SOCK_DGRAM &&
114             !inet_ehash_secret)
115                 build_ehash_secret();
116
117         /* Look for the requested type/protocol pair. */
118 lookup_protocol:
119         err = -ESOCKTNOSUPPORT;
120         rcu_read_lock();
121         list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
122
123                 err = 0;
124                 /* Check the non-wild match. */
125                 if (protocol == answer->protocol) {
126                         if (protocol != IPPROTO_IP)
127                                 break;
128                 } else {
129                         /* Check for the two wild cases. */
130                         if (IPPROTO_IP == protocol) {
131                                 protocol = answer->protocol;
132                                 break;
133                         }
134                         if (IPPROTO_IP == answer->protocol)
135                                 break;
136                 }
137                 err = -EPROTONOSUPPORT;
138         }
139
140         if (err) {
141                 if (try_loading_module < 2) {
142                         rcu_read_unlock();
143                         /*
144                          * Be more specific, e.g. net-pf-10-proto-132-type-1
145                          * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
146                          */
147                         if (++try_loading_module == 1)
148                                 request_module("net-pf-%d-proto-%d-type-%d",
149                                                 PF_INET6, protocol, sock->type);
150                         /*
151                          * Fall back to generic, e.g. net-pf-10-proto-132
152                          * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
153                          */
154                         else
155                                 request_module("net-pf-%d-proto-%d",
156                                                 PF_INET6, protocol);
157                         goto lookup_protocol;
158                 } else
159                         goto out_rcu_unlock;
160         }
161
162         err = -EPERM;
163         if (sock->type == SOCK_RAW && !kern && !capable(CAP_NET_RAW))
164                 goto out_rcu_unlock;
165
166         sock->ops = answer->ops;
167         answer_prot = answer->prot;
168         answer_no_check = answer->no_check;
169         answer_flags = answer->flags;
170         rcu_read_unlock();
171
172         WARN_ON(answer_prot->slab == NULL);
173
174         err = -ENOBUFS;
175         sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot);
176         if (sk == NULL)
177                 goto out;
178
179         sock_init_data(sock, sk);
180
181         err = 0;
182         sk->sk_no_check = answer_no_check;
183         if (INET_PROTOSW_REUSE & answer_flags)
184                 sk->sk_reuse = 1;
185
186         inet = inet_sk(sk);
187         inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
188
189         if (SOCK_RAW == sock->type) {
190                 inet->inet_num = protocol;
191                 if (IPPROTO_RAW == protocol)
192                         inet->hdrincl = 1;
193         }
194
195         sk->sk_destruct         = inet_sock_destruct;
196         sk->sk_family           = PF_INET6;
197         sk->sk_protocol         = protocol;
198
199         sk->sk_backlog_rcv      = answer->prot->backlog_rcv;
200
201         inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
202         np->hop_limit   = -1;
203         np->mcast_hops  = IPV6_DEFAULT_MCASTHOPS;
204         np->mc_loop     = 1;
205         np->pmtudisc    = IPV6_PMTUDISC_WANT;
206         np->ipv6only    = net->ipv6.sysctl.bindv6only;
207
208         /* Init the ipv4 part of the socket since we can have sockets
209          * using v6 API for ipv4.
210          */
211         inet->uc_ttl    = -1;
212
213         inet->mc_loop   = 1;
214         inet->mc_ttl    = 1;
215         inet->mc_index  = 0;
216         inet->mc_list   = NULL;
217
218         if (ipv4_config.no_pmtu_disc)
219                 inet->pmtudisc = IP_PMTUDISC_DONT;
220         else
221                 inet->pmtudisc = IP_PMTUDISC_WANT;
222         /*
223          * Increment only the relevant sk_prot->socks debug field, this changes
224          * the previous behaviour of incrementing both the equivalent to
225          * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
226          *
227          * This allows better debug granularity as we'll know exactly how many
228          * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
229          * transport protocol socks. -acme
230          */
231         sk_refcnt_debug_inc(sk);
232
233         if (inet->inet_num) {
234                 /* It assumes that any protocol which allows
235                  * the user to assign a number at socket
236                  * creation time automatically shares.
237                  */
238                 inet->inet_sport = htons(inet->inet_num);
239                 sk->sk_prot->hash(sk);
240         }
241         if (sk->sk_prot->init) {
242                 err = sk->sk_prot->init(sk);
243                 if (err) {
244                         sk_common_release(sk);
245                         goto out;
246                 }
247         }
248 out:
249         return err;
250 out_rcu_unlock:
251         rcu_read_unlock();
252         goto out;
253 }
254
255
256 /* bind for INET6 API */
257 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
258 {
259         struct sockaddr_in6 *addr=(struct sockaddr_in6 *)uaddr;
260         struct sock *sk = sock->sk;
261         struct inet_sock *inet = inet_sk(sk);
262         struct ipv6_pinfo *np = inet6_sk(sk);
263         struct net *net = sock_net(sk);
264         __be32 v4addr = 0;
265         unsigned short snum;
266         int addr_type = 0;
267         int err = 0;
268
269         /* If the socket has its own bind function then use it. */
270         if (sk->sk_prot->bind)
271                 return sk->sk_prot->bind(sk, uaddr, addr_len);
272
273         if (addr_len < SIN6_LEN_RFC2133)
274                 return -EINVAL;
275         addr_type = ipv6_addr_type(&addr->sin6_addr);
276         if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
277                 return -EINVAL;
278
279         snum = ntohs(addr->sin6_port);
280         if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
281                 return -EACCES;
282
283         lock_sock(sk);
284
285         /* Check these errors (active socket, double bind). */
286         if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
287                 err = -EINVAL;
288                 goto out;
289         }
290
291         /* Check if the address belongs to the host. */
292         if (addr_type == IPV6_ADDR_MAPPED) {
293                 int chk_addr_ret;
294
295                 /* Binding to v4-mapped address on a v6-only socket
296                  * makes no sense
297                  */
298                 if (np->ipv6only) {
299                         err = -EINVAL;
300                         goto out;
301                 }
302
303                 /* Reproduce AF_INET checks to make the bindings consitant */
304                 v4addr = addr->sin6_addr.s6_addr32[3];
305                 chk_addr_ret = inet_addr_type(net, v4addr);
306                 if (!sysctl_ip_nonlocal_bind &&
307                     !(inet->freebind || inet->transparent) &&
308                     v4addr != htonl(INADDR_ANY) &&
309                     chk_addr_ret != RTN_LOCAL &&
310                     chk_addr_ret != RTN_MULTICAST &&
311                     chk_addr_ret != RTN_BROADCAST) {
312                         err = -EADDRNOTAVAIL;
313                         goto out;
314                 }
315         } else {
316                 if (addr_type != IPV6_ADDR_ANY) {
317                         struct net_device *dev = NULL;
318
319                         rcu_read_lock();
320                         if (addr_type & IPV6_ADDR_LINKLOCAL) {
321                                 if (addr_len >= sizeof(struct sockaddr_in6) &&
322                                     addr->sin6_scope_id) {
323                                         /* Override any existing binding, if another one
324                                          * is supplied by user.
325                                          */
326                                         sk->sk_bound_dev_if = addr->sin6_scope_id;
327                                 }
328
329                                 /* Binding to link-local address requires an interface */
330                                 if (!sk->sk_bound_dev_if) {
331                                         err = -EINVAL;
332                                         goto out_unlock;
333                                 }
334                                 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
335                                 if (!dev) {
336                                         err = -ENODEV;
337                                         goto out_unlock;
338                                 }
339                         }
340
341                         /* ipv4 addr of the socket is invalid.  Only the
342                          * unspecified and mapped address have a v4 equivalent.
343                          */
344                         v4addr = LOOPBACK4_IPV6;
345                         if (!(addr_type & IPV6_ADDR_MULTICAST)) {
346                                 if (!inet->transparent &&
347                                     !ipv6_chk_addr(net, &addr->sin6_addr,
348                                                    dev, 0)) {
349                                         err = -EADDRNOTAVAIL;
350                                         goto out_unlock;
351                                 }
352                         }
353                         rcu_read_unlock();
354                 }
355         }
356
357         inet->inet_rcv_saddr = v4addr;
358         inet->inet_saddr = v4addr;
359
360         ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
361
362         if (!(addr_type & IPV6_ADDR_MULTICAST))
363                 ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
364
365         /* Make sure we are allowed to bind here. */
366         if (sk->sk_prot->get_port(sk, snum)) {
367                 inet_reset_saddr(sk);
368                 err = -EADDRINUSE;
369                 goto out;
370         }
371
372         if (addr_type != IPV6_ADDR_ANY) {
373                 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
374                 if (addr_type != IPV6_ADDR_MAPPED)
375                         np->ipv6only = 1;
376         }
377         if (snum)
378                 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
379         inet->inet_sport = htons(inet->inet_num);
380         inet->inet_dport = 0;
381         inet->inet_daddr = 0;
382 out:
383         release_sock(sk);
384         return err;
385 out_unlock:
386         rcu_read_unlock();
387         goto out;
388 }
389
390 EXPORT_SYMBOL(inet6_bind);
391
392 int inet6_release(struct socket *sock)
393 {
394         struct sock *sk = sock->sk;
395
396         if (sk == NULL)
397                 return -EINVAL;
398
399         /* Free mc lists */
400         ipv6_sock_mc_close(sk);
401
402         /* Free ac lists */
403         ipv6_sock_ac_close(sk);
404
405         return inet_release(sock);
406 }
407
408 EXPORT_SYMBOL(inet6_release);
409
410 void inet6_destroy_sock(struct sock *sk)
411 {
412         struct ipv6_pinfo *np = inet6_sk(sk);
413         struct sk_buff *skb;
414         struct ipv6_txoptions *opt;
415
416         /* Release rx options */
417
418         if ((skb = xchg(&np->pktoptions, NULL)) != NULL)
419                 kfree_skb(skb);
420
421         if ((skb = xchg(&np->rxpmtu, NULL)) != NULL)
422                 kfree_skb(skb);
423
424         /* Free flowlabels */
425         fl6_free_socklist(sk);
426
427         /* Free tx options */
428
429         if ((opt = xchg(&np->opt, NULL)) != NULL)
430                 sock_kfree_s(sk, opt, opt->tot_len);
431 }
432
433 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
434
435 /*
436  *      This does both peername and sockname.
437  */
438
439 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
440                  int *uaddr_len, int peer)
441 {
442         struct sockaddr_in6 *sin=(struct sockaddr_in6 *)uaddr;
443         struct sock *sk = sock->sk;
444         struct inet_sock *inet = inet_sk(sk);
445         struct ipv6_pinfo *np = inet6_sk(sk);
446
447         sin->sin6_family = AF_INET6;
448         sin->sin6_flowinfo = 0;
449         sin->sin6_scope_id = 0;
450         if (peer) {
451                 if (!inet->inet_dport)
452                         return -ENOTCONN;
453                 if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
454                     peer == 1)
455                         return -ENOTCONN;
456                 sin->sin6_port = inet->inet_dport;
457                 ipv6_addr_copy(&sin->sin6_addr, &np->daddr);
458                 if (np->sndflow)
459                         sin->sin6_flowinfo = np->flow_label;
460         } else {
461                 if (ipv6_addr_any(&np->rcv_saddr))
462                         ipv6_addr_copy(&sin->sin6_addr, &np->saddr);
463                 else
464                         ipv6_addr_copy(&sin->sin6_addr, &np->rcv_saddr);
465
466                 sin->sin6_port = inet->inet_sport;
467         }
468         if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL)
469                 sin->sin6_scope_id = sk->sk_bound_dev_if;
470         *uaddr_len = sizeof(*sin);
471         return 0;
472 }
473
474 EXPORT_SYMBOL(inet6_getname);
475
476 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
477 {
478         struct sock *sk = sock->sk;
479         struct net *net = sock_net(sk);
480
481         switch(cmd)
482         {
483         case SIOCGSTAMP:
484                 return sock_get_timestamp(sk, (struct timeval __user *)arg);
485
486         case SIOCGSTAMPNS:
487                 return sock_get_timestampns(sk, (struct timespec __user *)arg);
488
489         case SIOCADDRT:
490         case SIOCDELRT:
491
492                 return ipv6_route_ioctl(net, cmd, (void __user *)arg);
493
494         case SIOCSIFADDR:
495                 return addrconf_add_ifaddr(net, (void __user *) arg);
496         case SIOCDIFADDR:
497                 return addrconf_del_ifaddr(net, (void __user *) arg);
498         case SIOCSIFDSTADDR:
499                 return addrconf_set_dstaddr(net, (void __user *) arg);
500         default:
501                 if (!sk->sk_prot->ioctl)
502                         return -ENOIOCTLCMD;
503                 return sk->sk_prot->ioctl(sk, cmd, arg);
504         }
505         /*NOTREACHED*/
506         return 0;
507 }
508
509 EXPORT_SYMBOL(inet6_ioctl);
510
511 const struct proto_ops inet6_stream_ops = {
512         .family            = PF_INET6,
513         .owner             = THIS_MODULE,
514         .release           = inet6_release,
515         .bind              = inet6_bind,
516         .connect           = inet_stream_connect,       /* ok           */
517         .socketpair        = sock_no_socketpair,        /* a do nothing */
518         .accept            = inet_accept,               /* ok           */
519         .getname           = inet6_getname,
520         .poll              = tcp_poll,                  /* ok           */
521         .ioctl             = inet6_ioctl,               /* must change  */
522         .listen            = inet_listen,               /* ok           */
523         .shutdown          = inet_shutdown,             /* ok           */
524         .setsockopt        = sock_common_setsockopt,    /* ok           */
525         .getsockopt        = sock_common_getsockopt,    /* ok           */
526         .sendmsg           = inet_sendmsg,              /* ok           */
527         .recvmsg           = inet_recvmsg,              /* ok           */
528         .mmap              = sock_no_mmap,
529         .sendpage          = inet_sendpage,
530         .splice_read       = tcp_splice_read,
531 #ifdef CONFIG_COMPAT
532         .compat_setsockopt = compat_sock_common_setsockopt,
533         .compat_getsockopt = compat_sock_common_getsockopt,
534 #endif
535 };
536
537 const struct proto_ops inet6_dgram_ops = {
538         .family            = PF_INET6,
539         .owner             = THIS_MODULE,
540         .release           = inet6_release,
541         .bind              = inet6_bind,
542         .connect           = inet_dgram_connect,        /* ok           */
543         .socketpair        = sock_no_socketpair,        /* a do nothing */
544         .accept            = sock_no_accept,            /* a do nothing */
545         .getname           = inet6_getname,
546         .poll              = udp_poll,                  /* ok           */
547         .ioctl             = inet6_ioctl,               /* must change  */
548         .listen            = sock_no_listen,            /* ok           */
549         .shutdown          = inet_shutdown,             /* ok           */
550         .setsockopt        = sock_common_setsockopt,    /* ok           */
551         .getsockopt        = sock_common_getsockopt,    /* ok           */
552         .sendmsg           = inet_sendmsg,              /* ok           */
553         .recvmsg           = inet_recvmsg,              /* ok           */
554         .mmap              = sock_no_mmap,
555         .sendpage          = sock_no_sendpage,
556 #ifdef CONFIG_COMPAT
557         .compat_setsockopt = compat_sock_common_setsockopt,
558         .compat_getsockopt = compat_sock_common_getsockopt,
559 #endif
560 };
561
562 static const struct net_proto_family inet6_family_ops = {
563         .family = PF_INET6,
564         .create = inet6_create,
565         .owner  = THIS_MODULE,
566 };
567
568 int inet6_register_protosw(struct inet_protosw *p)
569 {
570         struct list_head *lh;
571         struct inet_protosw *answer;
572         struct list_head *last_perm;
573         int protocol = p->protocol;
574         int ret;
575
576         spin_lock_bh(&inetsw6_lock);
577
578         ret = -EINVAL;
579         if (p->type >= SOCK_MAX)
580                 goto out_illegal;
581
582         /* If we are trying to override a permanent protocol, bail. */
583         answer = NULL;
584         ret = -EPERM;
585         last_perm = &inetsw6[p->type];
586         list_for_each(lh, &inetsw6[p->type]) {
587                 answer = list_entry(lh, struct inet_protosw, list);
588
589                 /* Check only the non-wild match. */
590                 if (INET_PROTOSW_PERMANENT & answer->flags) {
591                         if (protocol == answer->protocol)
592                                 break;
593                         last_perm = lh;
594                 }
595
596                 answer = NULL;
597         }
598         if (answer)
599                 goto out_permanent;
600
601         /* Add the new entry after the last permanent entry if any, so that
602          * the new entry does not override a permanent entry when matched with
603          * a wild-card protocol. But it is allowed to override any existing
604          * non-permanent entry.  This means that when we remove this entry, the
605          * system automatically returns to the old behavior.
606          */
607         list_add_rcu(&p->list, last_perm);
608         ret = 0;
609 out:
610         spin_unlock_bh(&inetsw6_lock);
611         return ret;
612
613 out_permanent:
614         printk(KERN_ERR "Attempt to override permanent protocol %d.\n",
615                protocol);
616         goto out;
617
618 out_illegal:
619         printk(KERN_ERR
620                "Ignoring attempt to register invalid socket type %d.\n",
621                p->type);
622         goto out;
623 }
624
625 EXPORT_SYMBOL(inet6_register_protosw);
626
627 void
628 inet6_unregister_protosw(struct inet_protosw *p)
629 {
630         if (INET_PROTOSW_PERMANENT & p->flags) {
631                 printk(KERN_ERR
632                        "Attempt to unregister permanent protocol %d.\n",
633                        p->protocol);
634         } else {
635                 spin_lock_bh(&inetsw6_lock);
636                 list_del_rcu(&p->list);
637                 spin_unlock_bh(&inetsw6_lock);
638
639                 synchronize_net();
640         }
641 }
642
643 EXPORT_SYMBOL(inet6_unregister_protosw);
644
645 int inet6_sk_rebuild_header(struct sock *sk)
646 {
647         int err;
648         struct dst_entry *dst;
649         struct ipv6_pinfo *np = inet6_sk(sk);
650
651         dst = __sk_dst_check(sk, np->dst_cookie);
652
653         if (dst == NULL) {
654                 struct inet_sock *inet = inet_sk(sk);
655                 struct in6_addr *final_p, final;
656                 struct flowi fl;
657
658                 memset(&fl, 0, sizeof(fl));
659                 fl.proto = sk->sk_protocol;
660                 ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
661                 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
662                 fl.fl6_flowlabel = np->flow_label;
663                 fl.oif = sk->sk_bound_dev_if;
664                 fl.mark = sk->sk_mark;
665                 fl.fl_ip_dport = inet->inet_dport;
666                 fl.fl_ip_sport = inet->inet_sport;
667                 security_sk_classify_flow(sk, &fl);
668
669                 final_p = fl6_update_dst(&fl, np->opt, &final);
670
671                 err = ip6_dst_lookup(sk, &dst, &fl);
672                 if (err) {
673                         sk->sk_route_caps = 0;
674                         return err;
675                 }
676                 if (final_p)
677                         ipv6_addr_copy(&fl.fl6_dst, final_p);
678
679                 if ((err = xfrm_lookup(sock_net(sk), &dst, &fl, sk, 0)) < 0) {
680                         sk->sk_err_soft = -err;
681                         return err;
682                 }
683
684                 __ip6_dst_store(sk, dst, NULL, NULL);
685         }
686
687         return 0;
688 }
689
690 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
691
692 int ipv6_opt_accepted(struct sock *sk, struct sk_buff *skb)
693 {
694         struct ipv6_pinfo *np = inet6_sk(sk);
695         struct inet6_skb_parm *opt = IP6CB(skb);
696
697         if (np->rxopt.all) {
698                 if ((opt->hop && (np->rxopt.bits.hopopts ||
699                                   np->rxopt.bits.ohopopts)) ||
700                     ((IPV6_FLOWINFO_MASK &
701                       *(__be32 *)skb_network_header(skb)) &&
702                      np->rxopt.bits.rxflow) ||
703                     (opt->srcrt && (np->rxopt.bits.srcrt ||
704                      np->rxopt.bits.osrcrt)) ||
705                     ((opt->dst1 || opt->dst0) &&
706                      (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
707                         return 1;
708         }
709         return 0;
710 }
711
712 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
713
714 static int ipv6_gso_pull_exthdrs(struct sk_buff *skb, int proto)
715 {
716         const struct inet6_protocol *ops = NULL;
717
718         for (;;) {
719                 struct ipv6_opt_hdr *opth;
720                 int len;
721
722                 if (proto != NEXTHDR_HOP) {
723                         ops = rcu_dereference(inet6_protos[proto]);
724
725                         if (unlikely(!ops))
726                                 break;
727
728                         if (!(ops->flags & INET6_PROTO_GSO_EXTHDR))
729                                 break;
730                 }
731
732                 if (unlikely(!pskb_may_pull(skb, 8)))
733                         break;
734
735                 opth = (void *)skb->data;
736                 len = ipv6_optlen(opth);
737
738                 if (unlikely(!pskb_may_pull(skb, len)))
739                         break;
740
741                 proto = opth->nexthdr;
742                 __skb_pull(skb, len);
743         }
744
745         return proto;
746 }
747
748 static int ipv6_gso_send_check(struct sk_buff *skb)
749 {
750         struct ipv6hdr *ipv6h;
751         const struct inet6_protocol *ops;
752         int err = -EINVAL;
753
754         if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
755                 goto out;
756
757         ipv6h = ipv6_hdr(skb);
758         __skb_pull(skb, sizeof(*ipv6h));
759         err = -EPROTONOSUPPORT;
760
761         rcu_read_lock();
762         ops = rcu_dereference(inet6_protos[
763                 ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr)]);
764
765         if (likely(ops && ops->gso_send_check)) {
766                 skb_reset_transport_header(skb);
767                 err = ops->gso_send_check(skb);
768         }
769         rcu_read_unlock();
770
771 out:
772         return err;
773 }
774
775 static struct sk_buff *ipv6_gso_segment(struct sk_buff *skb, int features)
776 {
777         struct sk_buff *segs = ERR_PTR(-EINVAL);
778         struct ipv6hdr *ipv6h;
779         const struct inet6_protocol *ops;
780         int proto;
781         struct frag_hdr *fptr;
782         unsigned int unfrag_ip6hlen;
783         u8 *prevhdr;
784         int offset = 0;
785
786         if (!(features & NETIF_F_V6_CSUM))
787                 features &= ~NETIF_F_SG;
788
789         if (unlikely(skb_shinfo(skb)->gso_type &
790                      ~(SKB_GSO_UDP |
791                        SKB_GSO_DODGY |
792                        SKB_GSO_TCP_ECN |
793                        SKB_GSO_TCPV6 |
794                        0)))
795                 goto out;
796
797         if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
798                 goto out;
799
800         ipv6h = ipv6_hdr(skb);
801         __skb_pull(skb, sizeof(*ipv6h));
802         segs = ERR_PTR(-EPROTONOSUPPORT);
803
804         proto = ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr);
805         rcu_read_lock();
806         ops = rcu_dereference(inet6_protos[proto]);
807         if (likely(ops && ops->gso_segment)) {
808                 skb_reset_transport_header(skb);
809                 segs = ops->gso_segment(skb, features);
810         }
811         rcu_read_unlock();
812
813         if (unlikely(IS_ERR(segs)))
814                 goto out;
815
816         for (skb = segs; skb; skb = skb->next) {
817                 ipv6h = ipv6_hdr(skb);
818                 ipv6h->payload_len = htons(skb->len - skb->mac_len -
819                                            sizeof(*ipv6h));
820                 if (proto == IPPROTO_UDP) {
821                         unfrag_ip6hlen = ip6_find_1stfragopt(skb, &prevhdr);
822                         fptr = (struct frag_hdr *)(skb_network_header(skb) +
823                                 unfrag_ip6hlen);
824                         fptr->frag_off = htons(offset);
825                         if (skb->next != NULL)
826                                 fptr->frag_off |= htons(IP6_MF);
827                         offset += (ntohs(ipv6h->payload_len) -
828                                    sizeof(struct frag_hdr));
829                 }
830         }
831
832 out:
833         return segs;
834 }
835
836 struct ipv6_gro_cb {
837         struct napi_gro_cb napi;
838         int proto;
839 };
840
841 #define IPV6_GRO_CB(skb) ((struct ipv6_gro_cb *)(skb)->cb)
842
843 static struct sk_buff **ipv6_gro_receive(struct sk_buff **head,
844                                          struct sk_buff *skb)
845 {
846         const struct inet6_protocol *ops;
847         struct sk_buff **pp = NULL;
848         struct sk_buff *p;
849         struct ipv6hdr *iph;
850         unsigned int nlen;
851         unsigned int hlen;
852         unsigned int off;
853         int flush = 1;
854         int proto;
855         __wsum csum;
856
857         off = skb_gro_offset(skb);
858         hlen = off + sizeof(*iph);
859         iph = skb_gro_header_fast(skb, off);
860         if (skb_gro_header_hard(skb, hlen)) {
861                 iph = skb_gro_header_slow(skb, hlen, off);
862                 if (unlikely(!iph))
863                         goto out;
864         }
865
866         skb_gro_pull(skb, sizeof(*iph));
867         skb_set_transport_header(skb, skb_gro_offset(skb));
868
869         flush += ntohs(iph->payload_len) != skb_gro_len(skb);
870
871         rcu_read_lock();
872         proto = iph->nexthdr;
873         ops = rcu_dereference(inet6_protos[proto]);
874         if (!ops || !ops->gro_receive) {
875                 __pskb_pull(skb, skb_gro_offset(skb));
876                 proto = ipv6_gso_pull_exthdrs(skb, proto);
877                 skb_gro_pull(skb, -skb_transport_offset(skb));
878                 skb_reset_transport_header(skb);
879                 __skb_push(skb, skb_gro_offset(skb));
880
881                 if (!ops || !ops->gro_receive)
882                         goto out_unlock;
883
884                 iph = ipv6_hdr(skb);
885         }
886
887         IPV6_GRO_CB(skb)->proto = proto;
888
889         flush--;
890         nlen = skb_network_header_len(skb);
891
892         for (p = *head; p; p = p->next) {
893                 struct ipv6hdr *iph2;
894
895                 if (!NAPI_GRO_CB(p)->same_flow)
896                         continue;
897
898                 iph2 = ipv6_hdr(p);
899
900                 /* All fields must match except length. */
901                 if (nlen != skb_network_header_len(p) ||
902                     memcmp(iph, iph2, offsetof(struct ipv6hdr, payload_len)) ||
903                     memcmp(&iph->nexthdr, &iph2->nexthdr,
904                            nlen - offsetof(struct ipv6hdr, nexthdr))) {
905                         NAPI_GRO_CB(p)->same_flow = 0;
906                         continue;
907                 }
908
909                 NAPI_GRO_CB(p)->flush |= flush;
910         }
911
912         NAPI_GRO_CB(skb)->flush |= flush;
913
914         csum = skb->csum;
915         skb_postpull_rcsum(skb, iph, skb_network_header_len(skb));
916
917         pp = ops->gro_receive(head, skb);
918
919         skb->csum = csum;
920
921 out_unlock:
922         rcu_read_unlock();
923
924 out:
925         NAPI_GRO_CB(skb)->flush |= flush;
926
927         return pp;
928 }
929
930 static int ipv6_gro_complete(struct sk_buff *skb)
931 {
932         const struct inet6_protocol *ops;
933         struct ipv6hdr *iph = ipv6_hdr(skb);
934         int err = -ENOSYS;
935
936         iph->payload_len = htons(skb->len - skb_network_offset(skb) -
937                                  sizeof(*iph));
938
939         rcu_read_lock();
940         ops = rcu_dereference(inet6_protos[IPV6_GRO_CB(skb)->proto]);
941         if (WARN_ON(!ops || !ops->gro_complete))
942                 goto out_unlock;
943
944         err = ops->gro_complete(skb);
945
946 out_unlock:
947         rcu_read_unlock();
948
949         return err;
950 }
951
952 static struct packet_type ipv6_packet_type __read_mostly = {
953         .type = cpu_to_be16(ETH_P_IPV6),
954         .func = ipv6_rcv,
955         .gso_send_check = ipv6_gso_send_check,
956         .gso_segment = ipv6_gso_segment,
957         .gro_receive = ipv6_gro_receive,
958         .gro_complete = ipv6_gro_complete,
959 };
960
961 static int __init ipv6_packet_init(void)
962 {
963         dev_add_pack(&ipv6_packet_type);
964         return 0;
965 }
966
967 static void ipv6_packet_cleanup(void)
968 {
969         dev_remove_pack(&ipv6_packet_type);
970 }
971
972 static int __net_init ipv6_init_mibs(struct net *net)
973 {
974         if (snmp_mib_init((void __percpu **)net->mib.udp_stats_in6,
975                           sizeof(struct udp_mib),
976                           __alignof__(struct udp_mib)) < 0)
977                 return -ENOMEM;
978         if (snmp_mib_init((void __percpu **)net->mib.udplite_stats_in6,
979                           sizeof(struct udp_mib),
980                           __alignof__(struct udp_mib)) < 0)
981                 goto err_udplite_mib;
982         if (snmp_mib_init((void __percpu **)net->mib.ipv6_statistics,
983                           sizeof(struct ipstats_mib),
984                           __alignof__(struct ipstats_mib)) < 0)
985                 goto err_ip_mib;
986         if (snmp_mib_init((void __percpu **)net->mib.icmpv6_statistics,
987                           sizeof(struct icmpv6_mib),
988                           __alignof__(struct icmpv6_mib)) < 0)
989                 goto err_icmp_mib;
990         if (snmp_mib_init((void __percpu **)net->mib.icmpv6msg_statistics,
991                           sizeof(struct icmpv6msg_mib),
992                           __alignof__(struct icmpv6msg_mib)) < 0)
993                 goto err_icmpmsg_mib;
994         return 0;
995
996 err_icmpmsg_mib:
997         snmp_mib_free((void __percpu **)net->mib.icmpv6_statistics);
998 err_icmp_mib:
999         snmp_mib_free((void __percpu **)net->mib.ipv6_statistics);
1000 err_ip_mib:
1001         snmp_mib_free((void __percpu **)net->mib.udplite_stats_in6);
1002 err_udplite_mib:
1003         snmp_mib_free((void __percpu **)net->mib.udp_stats_in6);
1004         return -ENOMEM;
1005 }
1006
1007 static void ipv6_cleanup_mibs(struct net *net)
1008 {
1009         snmp_mib_free((void __percpu **)net->mib.udp_stats_in6);
1010         snmp_mib_free((void __percpu **)net->mib.udplite_stats_in6);
1011         snmp_mib_free((void __percpu **)net->mib.ipv6_statistics);
1012         snmp_mib_free((void __percpu **)net->mib.icmpv6_statistics);
1013         snmp_mib_free((void __percpu **)net->mib.icmpv6msg_statistics);
1014 }
1015
1016 static int __net_init inet6_net_init(struct net *net)
1017 {
1018         int err = 0;
1019
1020         net->ipv6.sysctl.bindv6only = 0;
1021         net->ipv6.sysctl.icmpv6_time = 1*HZ;
1022
1023         err = ipv6_init_mibs(net);
1024         if (err)
1025                 return err;
1026 #ifdef CONFIG_PROC_FS
1027         err = udp6_proc_init(net);
1028         if (err)
1029                 goto out;
1030         err = tcp6_proc_init(net);
1031         if (err)
1032                 goto proc_tcp6_fail;
1033         err = ac6_proc_init(net);
1034         if (err)
1035                 goto proc_ac6_fail;
1036 #endif
1037         return err;
1038
1039 #ifdef CONFIG_PROC_FS
1040 proc_ac6_fail:
1041         tcp6_proc_exit(net);
1042 proc_tcp6_fail:
1043         udp6_proc_exit(net);
1044 out:
1045         ipv6_cleanup_mibs(net);
1046         return err;
1047 #endif
1048 }
1049
1050 static void __net_exit inet6_net_exit(struct net *net)
1051 {
1052 #ifdef CONFIG_PROC_FS
1053         udp6_proc_exit(net);
1054         tcp6_proc_exit(net);
1055         ac6_proc_exit(net);
1056 #endif
1057         ipv6_cleanup_mibs(net);
1058 }
1059
1060 static struct pernet_operations inet6_net_ops = {
1061         .init = inet6_net_init,
1062         .exit = inet6_net_exit,
1063 };
1064
1065 static int __init inet6_init(void)
1066 {
1067         struct sk_buff *dummy_skb;
1068         struct list_head *r;
1069         int err = 0;
1070
1071         BUILD_BUG_ON(sizeof(struct inet6_skb_parm) > sizeof(dummy_skb->cb));
1072
1073         /* Register the socket-side information for inet6_create.  */
1074         for(r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
1075                 INIT_LIST_HEAD(r);
1076
1077         if (disable_ipv6_mod) {
1078                 printk(KERN_INFO
1079                        "IPv6: Loaded, but administratively disabled, "
1080                        "reboot required to enable\n");
1081                 goto out;
1082         }
1083
1084         err = proto_register(&tcpv6_prot, 1);
1085         if (err)
1086                 goto out;
1087
1088         err = proto_register(&udpv6_prot, 1);
1089         if (err)
1090                 goto out_unregister_tcp_proto;
1091
1092         err = proto_register(&udplitev6_prot, 1);
1093         if (err)
1094                 goto out_unregister_udp_proto;
1095
1096         err = proto_register(&rawv6_prot, 1);
1097         if (err)
1098                 goto out_unregister_udplite_proto;
1099
1100
1101         /* We MUST register RAW sockets before we create the ICMP6,
1102          * IGMP6, or NDISC control sockets.
1103          */
1104         err = rawv6_init();
1105         if (err)
1106                 goto out_unregister_raw_proto;
1107
1108         /* Register the family here so that the init calls below will
1109          * be able to create sockets. (?? is this dangerous ??)
1110          */
1111         err = sock_register(&inet6_family_ops);
1112         if (err)
1113                 goto out_sock_register_fail;
1114
1115 #ifdef CONFIG_SYSCTL
1116         err = ipv6_static_sysctl_register();
1117         if (err)
1118                 goto static_sysctl_fail;
1119 #endif
1120         /*
1121          *      ipngwg API draft makes clear that the correct semantics
1122          *      for TCP and UDP is to consider one TCP and UDP instance
1123          *      in a host availiable by both INET and INET6 APIs and
1124          *      able to communicate via both network protocols.
1125          */
1126
1127         err = register_pernet_subsys(&inet6_net_ops);
1128         if (err)
1129                 goto register_pernet_fail;
1130         err = icmpv6_init();
1131         if (err)
1132                 goto icmp_fail;
1133         err = ip6_mr_init();
1134         if (err)
1135                 goto ipmr_fail;
1136         err = ndisc_init();
1137         if (err)
1138                 goto ndisc_fail;
1139         err = igmp6_init();
1140         if (err)
1141                 goto igmp_fail;
1142         err = ipv6_netfilter_init();
1143         if (err)
1144                 goto netfilter_fail;
1145         /* Create /proc/foo6 entries. */
1146 #ifdef CONFIG_PROC_FS
1147         err = -ENOMEM;
1148         if (raw6_proc_init())
1149                 goto proc_raw6_fail;
1150         if (udplite6_proc_init())
1151                 goto proc_udplite6_fail;
1152         if (ipv6_misc_proc_init())
1153                 goto proc_misc6_fail;
1154         if (if6_proc_init())
1155                 goto proc_if6_fail;
1156 #endif
1157         err = ip6_route_init();
1158         if (err)
1159                 goto ip6_route_fail;
1160         err = ip6_flowlabel_init();
1161         if (err)
1162                 goto ip6_flowlabel_fail;
1163         err = addrconf_init();
1164         if (err)
1165                 goto addrconf_fail;
1166
1167         /* Init v6 extension headers. */
1168         err = ipv6_exthdrs_init();
1169         if (err)
1170                 goto ipv6_exthdrs_fail;
1171
1172         err = ipv6_frag_init();
1173         if (err)
1174                 goto ipv6_frag_fail;
1175
1176         /* Init v6 transport protocols. */
1177         err = udpv6_init();
1178         if (err)
1179                 goto udpv6_fail;
1180
1181         err = udplitev6_init();
1182         if (err)
1183                 goto udplitev6_fail;
1184
1185         err = tcpv6_init();
1186         if (err)
1187                 goto tcpv6_fail;
1188
1189         err = ipv6_packet_init();
1190         if (err)
1191                 goto ipv6_packet_fail;
1192
1193 #ifdef CONFIG_SYSCTL
1194         err = ipv6_sysctl_register();
1195         if (err)
1196                 goto sysctl_fail;
1197 #endif
1198 out:
1199         return err;
1200
1201 #ifdef CONFIG_SYSCTL
1202 sysctl_fail:
1203         ipv6_packet_cleanup();
1204 #endif
1205 ipv6_packet_fail:
1206         tcpv6_exit();
1207 tcpv6_fail:
1208         udplitev6_exit();
1209 udplitev6_fail:
1210         udpv6_exit();
1211 udpv6_fail:
1212         ipv6_frag_exit();
1213 ipv6_frag_fail:
1214         ipv6_exthdrs_exit();
1215 ipv6_exthdrs_fail:
1216         addrconf_cleanup();
1217 addrconf_fail:
1218         ip6_flowlabel_cleanup();
1219 ip6_flowlabel_fail:
1220         ip6_route_cleanup();
1221 ip6_route_fail:
1222 #ifdef CONFIG_PROC_FS
1223         if6_proc_exit();
1224 proc_if6_fail:
1225         ipv6_misc_proc_exit();
1226 proc_misc6_fail:
1227         udplite6_proc_exit();
1228 proc_udplite6_fail:
1229         raw6_proc_exit();
1230 proc_raw6_fail:
1231 #endif
1232         ipv6_netfilter_fini();
1233 netfilter_fail:
1234         igmp6_cleanup();
1235 igmp_fail:
1236         ndisc_cleanup();
1237 ndisc_fail:
1238         ip6_mr_cleanup();
1239 ipmr_fail:
1240         icmpv6_cleanup();
1241 icmp_fail:
1242         unregister_pernet_subsys(&inet6_net_ops);
1243 register_pernet_fail:
1244 #ifdef CONFIG_SYSCTL
1245         ipv6_static_sysctl_unregister();
1246 static_sysctl_fail:
1247 #endif
1248         sock_unregister(PF_INET6);
1249         rtnl_unregister_all(PF_INET6);
1250 out_sock_register_fail:
1251         rawv6_exit();
1252 out_unregister_raw_proto:
1253         proto_unregister(&rawv6_prot);
1254 out_unregister_udplite_proto:
1255         proto_unregister(&udplitev6_prot);
1256 out_unregister_udp_proto:
1257         proto_unregister(&udpv6_prot);
1258 out_unregister_tcp_proto:
1259         proto_unregister(&tcpv6_prot);
1260         goto out;
1261 }
1262 module_init(inet6_init);
1263
1264 static void __exit inet6_exit(void)
1265 {
1266         if (disable_ipv6_mod)
1267                 return;
1268
1269         /* First of all disallow new sockets creation. */
1270         sock_unregister(PF_INET6);
1271         /* Disallow any further netlink messages */
1272         rtnl_unregister_all(PF_INET6);
1273
1274 #ifdef CONFIG_SYSCTL
1275         ipv6_sysctl_unregister();
1276 #endif
1277         udpv6_exit();
1278         udplitev6_exit();
1279         tcpv6_exit();
1280
1281         /* Cleanup code parts. */
1282         ipv6_packet_cleanup();
1283         ipv6_frag_exit();
1284         ipv6_exthdrs_exit();
1285         addrconf_cleanup();
1286         ip6_flowlabel_cleanup();
1287         ip6_route_cleanup();
1288 #ifdef CONFIG_PROC_FS
1289
1290         /* Cleanup code parts. */
1291         if6_proc_exit();
1292         ipv6_misc_proc_exit();
1293         udplite6_proc_exit();
1294         raw6_proc_exit();
1295 #endif
1296         ipv6_netfilter_fini();
1297         igmp6_cleanup();
1298         ndisc_cleanup();
1299         ip6_mr_cleanup();
1300         icmpv6_cleanup();
1301         rawv6_exit();
1302
1303         unregister_pernet_subsys(&inet6_net_ops);
1304 #ifdef CONFIG_SYSCTL
1305         ipv6_static_sysctl_unregister();
1306 #endif
1307         proto_unregister(&rawv6_prot);
1308         proto_unregister(&udplitev6_prot);
1309         proto_unregister(&udpv6_prot);
1310         proto_unregister(&tcpv6_prot);
1311
1312         rcu_barrier(); /* Wait for completion of call_rcu()'s */
1313 }
1314 module_exit(inet6_exit);
1315
1316 MODULE_ALIAS_NETPROTO(PF_INET6);