net: return operator cleanup
[linux-3.10.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 (!ipv6_chk_addr(net, &addr->sin6_addr,
347                                                    dev, 0)) {
348                                         err = -EADDRNOTAVAIL;
349                                         goto out_unlock;
350                                 }
351                         }
352                         rcu_read_unlock();
353                 }
354         }
355
356         inet->inet_rcv_saddr = v4addr;
357         inet->inet_saddr = v4addr;
358
359         ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
360
361         if (!(addr_type & IPV6_ADDR_MULTICAST))
362                 ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
363
364         /* Make sure we are allowed to bind here. */
365         if (sk->sk_prot->get_port(sk, snum)) {
366                 inet_reset_saddr(sk);
367                 err = -EADDRINUSE;
368                 goto out;
369         }
370
371         if (addr_type != IPV6_ADDR_ANY) {
372                 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
373                 if (addr_type != IPV6_ADDR_MAPPED)
374                         np->ipv6only = 1;
375         }
376         if (snum)
377                 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
378         inet->inet_sport = htons(inet->inet_num);
379         inet->inet_dport = 0;
380         inet->inet_daddr = 0;
381 out:
382         release_sock(sk);
383         return err;
384 out_unlock:
385         rcu_read_unlock();
386         goto out;
387 }
388
389 EXPORT_SYMBOL(inet6_bind);
390
391 int inet6_release(struct socket *sock)
392 {
393         struct sock *sk = sock->sk;
394
395         if (sk == NULL)
396                 return -EINVAL;
397
398         /* Free mc lists */
399         ipv6_sock_mc_close(sk);
400
401         /* Free ac lists */
402         ipv6_sock_ac_close(sk);
403
404         return inet_release(sock);
405 }
406
407 EXPORT_SYMBOL(inet6_release);
408
409 void inet6_destroy_sock(struct sock *sk)
410 {
411         struct ipv6_pinfo *np = inet6_sk(sk);
412         struct sk_buff *skb;
413         struct ipv6_txoptions *opt;
414
415         /* Release rx options */
416
417         if ((skb = xchg(&np->pktoptions, NULL)) != NULL)
418                 kfree_skb(skb);
419
420         if ((skb = xchg(&np->rxpmtu, NULL)) != NULL)
421                 kfree_skb(skb);
422
423         /* Free flowlabels */
424         fl6_free_socklist(sk);
425
426         /* Free tx options */
427
428         if ((opt = xchg(&np->opt, NULL)) != NULL)
429                 sock_kfree_s(sk, opt, opt->tot_len);
430 }
431
432 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
433
434 /*
435  *      This does both peername and sockname.
436  */
437
438 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
439                  int *uaddr_len, int peer)
440 {
441         struct sockaddr_in6 *sin=(struct sockaddr_in6 *)uaddr;
442         struct sock *sk = sock->sk;
443         struct inet_sock *inet = inet_sk(sk);
444         struct ipv6_pinfo *np = inet6_sk(sk);
445
446         sin->sin6_family = AF_INET6;
447         sin->sin6_flowinfo = 0;
448         sin->sin6_scope_id = 0;
449         if (peer) {
450                 if (!inet->inet_dport)
451                         return -ENOTCONN;
452                 if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
453                     peer == 1)
454                         return -ENOTCONN;
455                 sin->sin6_port = inet->inet_dport;
456                 ipv6_addr_copy(&sin->sin6_addr, &np->daddr);
457                 if (np->sndflow)
458                         sin->sin6_flowinfo = np->flow_label;
459         } else {
460                 if (ipv6_addr_any(&np->rcv_saddr))
461                         ipv6_addr_copy(&sin->sin6_addr, &np->saddr);
462                 else
463                         ipv6_addr_copy(&sin->sin6_addr, &np->rcv_saddr);
464
465                 sin->sin6_port = inet->inet_sport;
466         }
467         if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL)
468                 sin->sin6_scope_id = sk->sk_bound_dev_if;
469         *uaddr_len = sizeof(*sin);
470         return 0;
471 }
472
473 EXPORT_SYMBOL(inet6_getname);
474
475 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
476 {
477         struct sock *sk = sock->sk;
478         struct net *net = sock_net(sk);
479
480         switch(cmd)
481         {
482         case SIOCGSTAMP:
483                 return sock_get_timestamp(sk, (struct timeval __user *)arg);
484
485         case SIOCGSTAMPNS:
486                 return sock_get_timestampns(sk, (struct timespec __user *)arg);
487
488         case SIOCADDRT:
489         case SIOCDELRT:
490
491                 return ipv6_route_ioctl(net, cmd, (void __user *)arg);
492
493         case SIOCSIFADDR:
494                 return addrconf_add_ifaddr(net, (void __user *) arg);
495         case SIOCDIFADDR:
496                 return addrconf_del_ifaddr(net, (void __user *) arg);
497         case SIOCSIFDSTADDR:
498                 return addrconf_set_dstaddr(net, (void __user *) arg);
499         default:
500                 if (!sk->sk_prot->ioctl)
501                         return -ENOIOCTLCMD;
502                 return sk->sk_prot->ioctl(sk, cmd, arg);
503         }
504         /*NOTREACHED*/
505         return 0;
506 }
507
508 EXPORT_SYMBOL(inet6_ioctl);
509
510 const struct proto_ops inet6_stream_ops = {
511         .family            = PF_INET6,
512         .owner             = THIS_MODULE,
513         .release           = inet6_release,
514         .bind              = inet6_bind,
515         .connect           = inet_stream_connect,       /* ok           */
516         .socketpair        = sock_no_socketpair,        /* a do nothing */
517         .accept            = inet_accept,               /* ok           */
518         .getname           = inet6_getname,
519         .poll              = tcp_poll,                  /* ok           */
520         .ioctl             = inet6_ioctl,               /* must change  */
521         .listen            = inet_listen,               /* ok           */
522         .shutdown          = inet_shutdown,             /* ok           */
523         .setsockopt        = sock_common_setsockopt,    /* ok           */
524         .getsockopt        = sock_common_getsockopt,    /* ok           */
525         .sendmsg           = inet_sendmsg,              /* ok           */
526         .recvmsg           = inet_recvmsg,              /* ok           */
527         .mmap              = sock_no_mmap,
528         .sendpage          = inet_sendpage,
529         .splice_read       = tcp_splice_read,
530 #ifdef CONFIG_COMPAT
531         .compat_setsockopt = compat_sock_common_setsockopt,
532         .compat_getsockopt = compat_sock_common_getsockopt,
533 #endif
534 };
535
536 const struct proto_ops inet6_dgram_ops = {
537         .family            = PF_INET6,
538         .owner             = THIS_MODULE,
539         .release           = inet6_release,
540         .bind              = inet6_bind,
541         .connect           = inet_dgram_connect,        /* ok           */
542         .socketpair        = sock_no_socketpair,        /* a do nothing */
543         .accept            = sock_no_accept,            /* a do nothing */
544         .getname           = inet6_getname,
545         .poll              = udp_poll,                  /* ok           */
546         .ioctl             = inet6_ioctl,               /* must change  */
547         .listen            = sock_no_listen,            /* ok           */
548         .shutdown          = inet_shutdown,             /* ok           */
549         .setsockopt        = sock_common_setsockopt,    /* ok           */
550         .getsockopt        = sock_common_getsockopt,    /* ok           */
551         .sendmsg           = inet_sendmsg,              /* ok           */
552         .recvmsg           = inet_recvmsg,              /* ok           */
553         .mmap              = sock_no_mmap,
554         .sendpage          = sock_no_sendpage,
555 #ifdef CONFIG_COMPAT
556         .compat_setsockopt = compat_sock_common_setsockopt,
557         .compat_getsockopt = compat_sock_common_getsockopt,
558 #endif
559 };
560
561 static const struct net_proto_family inet6_family_ops = {
562         .family = PF_INET6,
563         .create = inet6_create,
564         .owner  = THIS_MODULE,
565 };
566
567 int inet6_register_protosw(struct inet_protosw *p)
568 {
569         struct list_head *lh;
570         struct inet_protosw *answer;
571         struct list_head *last_perm;
572         int protocol = p->protocol;
573         int ret;
574
575         spin_lock_bh(&inetsw6_lock);
576
577         ret = -EINVAL;
578         if (p->type >= SOCK_MAX)
579                 goto out_illegal;
580
581         /* If we are trying to override a permanent protocol, bail. */
582         answer = NULL;
583         ret = -EPERM;
584         last_perm = &inetsw6[p->type];
585         list_for_each(lh, &inetsw6[p->type]) {
586                 answer = list_entry(lh, struct inet_protosw, list);
587
588                 /* Check only the non-wild match. */
589                 if (INET_PROTOSW_PERMANENT & answer->flags) {
590                         if (protocol == answer->protocol)
591                                 break;
592                         last_perm = lh;
593                 }
594
595                 answer = NULL;
596         }
597         if (answer)
598                 goto out_permanent;
599
600         /* Add the new entry after the last permanent entry if any, so that
601          * the new entry does not override a permanent entry when matched with
602          * a wild-card protocol. But it is allowed to override any existing
603          * non-permanent entry.  This means that when we remove this entry, the
604          * system automatically returns to the old behavior.
605          */
606         list_add_rcu(&p->list, last_perm);
607         ret = 0;
608 out:
609         spin_unlock_bh(&inetsw6_lock);
610         return ret;
611
612 out_permanent:
613         printk(KERN_ERR "Attempt to override permanent protocol %d.\n",
614                protocol);
615         goto out;
616
617 out_illegal:
618         printk(KERN_ERR
619                "Ignoring attempt to register invalid socket type %d.\n",
620                p->type);
621         goto out;
622 }
623
624 EXPORT_SYMBOL(inet6_register_protosw);
625
626 void
627 inet6_unregister_protosw(struct inet_protosw *p)
628 {
629         if (INET_PROTOSW_PERMANENT & p->flags) {
630                 printk(KERN_ERR
631                        "Attempt to unregister permanent protocol %d.\n",
632                        p->protocol);
633         } else {
634                 spin_lock_bh(&inetsw6_lock);
635                 list_del_rcu(&p->list);
636                 spin_unlock_bh(&inetsw6_lock);
637
638                 synchronize_net();
639         }
640 }
641
642 EXPORT_SYMBOL(inet6_unregister_protosw);
643
644 int inet6_sk_rebuild_header(struct sock *sk)
645 {
646         int err;
647         struct dst_entry *dst;
648         struct ipv6_pinfo *np = inet6_sk(sk);
649
650         dst = __sk_dst_check(sk, np->dst_cookie);
651
652         if (dst == NULL) {
653                 struct inet_sock *inet = inet_sk(sk);
654                 struct in6_addr *final_p, final;
655                 struct flowi fl;
656
657                 memset(&fl, 0, sizeof(fl));
658                 fl.proto = sk->sk_protocol;
659                 ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
660                 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
661                 fl.fl6_flowlabel = np->flow_label;
662                 fl.oif = sk->sk_bound_dev_if;
663                 fl.mark = sk->sk_mark;
664                 fl.fl_ip_dport = inet->inet_dport;
665                 fl.fl_ip_sport = inet->inet_sport;
666                 security_sk_classify_flow(sk, &fl);
667
668                 final_p = fl6_update_dst(&fl, np->opt, &final);
669
670                 err = ip6_dst_lookup(sk, &dst, &fl);
671                 if (err) {
672                         sk->sk_route_caps = 0;
673                         return err;
674                 }
675                 if (final_p)
676                         ipv6_addr_copy(&fl.fl6_dst, final_p);
677
678                 if ((err = xfrm_lookup(sock_net(sk), &dst, &fl, sk, 0)) < 0) {
679                         sk->sk_err_soft = -err;
680                         return err;
681                 }
682
683                 __ip6_dst_store(sk, dst, NULL, NULL);
684         }
685
686         return 0;
687 }
688
689 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
690
691 int ipv6_opt_accepted(struct sock *sk, struct sk_buff *skb)
692 {
693         struct ipv6_pinfo *np = inet6_sk(sk);
694         struct inet6_skb_parm *opt = IP6CB(skb);
695
696         if (np->rxopt.all) {
697                 if ((opt->hop && (np->rxopt.bits.hopopts ||
698                                   np->rxopt.bits.ohopopts)) ||
699                     ((IPV6_FLOWINFO_MASK &
700                       *(__be32 *)skb_network_header(skb)) &&
701                      np->rxopt.bits.rxflow) ||
702                     (opt->srcrt && (np->rxopt.bits.srcrt ||
703                      np->rxopt.bits.osrcrt)) ||
704                     ((opt->dst1 || opt->dst0) &&
705                      (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
706                         return 1;
707         }
708         return 0;
709 }
710
711 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
712
713 static int ipv6_gso_pull_exthdrs(struct sk_buff *skb, int proto)
714 {
715         const struct inet6_protocol *ops = NULL;
716
717         for (;;) {
718                 struct ipv6_opt_hdr *opth;
719                 int len;
720
721                 if (proto != NEXTHDR_HOP) {
722                         ops = rcu_dereference(inet6_protos[proto]);
723
724                         if (unlikely(!ops))
725                                 break;
726
727                         if (!(ops->flags & INET6_PROTO_GSO_EXTHDR))
728                                 break;
729                 }
730
731                 if (unlikely(!pskb_may_pull(skb, 8)))
732                         break;
733
734                 opth = (void *)skb->data;
735                 len = ipv6_optlen(opth);
736
737                 if (unlikely(!pskb_may_pull(skb, len)))
738                         break;
739
740                 proto = opth->nexthdr;
741                 __skb_pull(skb, len);
742         }
743
744         return proto;
745 }
746
747 static int ipv6_gso_send_check(struct sk_buff *skb)
748 {
749         struct ipv6hdr *ipv6h;
750         const struct inet6_protocol *ops;
751         int err = -EINVAL;
752
753         if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
754                 goto out;
755
756         ipv6h = ipv6_hdr(skb);
757         __skb_pull(skb, sizeof(*ipv6h));
758         err = -EPROTONOSUPPORT;
759
760         rcu_read_lock();
761         ops = rcu_dereference(inet6_protos[
762                 ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr)]);
763
764         if (likely(ops && ops->gso_send_check)) {
765                 skb_reset_transport_header(skb);
766                 err = ops->gso_send_check(skb);
767         }
768         rcu_read_unlock();
769
770 out:
771         return err;
772 }
773
774 static struct sk_buff *ipv6_gso_segment(struct sk_buff *skb, int features)
775 {
776         struct sk_buff *segs = ERR_PTR(-EINVAL);
777         struct ipv6hdr *ipv6h;
778         const struct inet6_protocol *ops;
779         int proto;
780         struct frag_hdr *fptr;
781         unsigned int unfrag_ip6hlen;
782         u8 *prevhdr;
783         int offset = 0;
784
785         if (!(features & NETIF_F_V6_CSUM))
786                 features &= ~NETIF_F_SG;
787
788         if (unlikely(skb_shinfo(skb)->gso_type &
789                      ~(SKB_GSO_UDP |
790                        SKB_GSO_DODGY |
791                        SKB_GSO_TCP_ECN |
792                        SKB_GSO_TCPV6 |
793                        0)))
794                 goto out;
795
796         if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
797                 goto out;
798
799         ipv6h = ipv6_hdr(skb);
800         __skb_pull(skb, sizeof(*ipv6h));
801         segs = ERR_PTR(-EPROTONOSUPPORT);
802
803         proto = ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr);
804         rcu_read_lock();
805         ops = rcu_dereference(inet6_protos[proto]);
806         if (likely(ops && ops->gso_segment)) {
807                 skb_reset_transport_header(skb);
808                 segs = ops->gso_segment(skb, features);
809         }
810         rcu_read_unlock();
811
812         if (unlikely(IS_ERR(segs)))
813                 goto out;
814
815         for (skb = segs; skb; skb = skb->next) {
816                 ipv6h = ipv6_hdr(skb);
817                 ipv6h->payload_len = htons(skb->len - skb->mac_len -
818                                            sizeof(*ipv6h));
819                 if (proto == IPPROTO_UDP) {
820                         unfrag_ip6hlen = ip6_find_1stfragopt(skb, &prevhdr);
821                         fptr = (struct frag_hdr *)(skb_network_header(skb) +
822                                 unfrag_ip6hlen);
823                         fptr->frag_off = htons(offset);
824                         if (skb->next != NULL)
825                                 fptr->frag_off |= htons(IP6_MF);
826                         offset += (ntohs(ipv6h->payload_len) -
827                                    sizeof(struct frag_hdr));
828                 }
829         }
830
831 out:
832         return segs;
833 }
834
835 struct ipv6_gro_cb {
836         struct napi_gro_cb napi;
837         int proto;
838 };
839
840 #define IPV6_GRO_CB(skb) ((struct ipv6_gro_cb *)(skb)->cb)
841
842 static struct sk_buff **ipv6_gro_receive(struct sk_buff **head,
843                                          struct sk_buff *skb)
844 {
845         const struct inet6_protocol *ops;
846         struct sk_buff **pp = NULL;
847         struct sk_buff *p;
848         struct ipv6hdr *iph;
849         unsigned int nlen;
850         unsigned int hlen;
851         unsigned int off;
852         int flush = 1;
853         int proto;
854         __wsum csum;
855
856         off = skb_gro_offset(skb);
857         hlen = off + sizeof(*iph);
858         iph = skb_gro_header_fast(skb, off);
859         if (skb_gro_header_hard(skb, hlen)) {
860                 iph = skb_gro_header_slow(skb, hlen, off);
861                 if (unlikely(!iph))
862                         goto out;
863         }
864
865         skb_gro_pull(skb, sizeof(*iph));
866         skb_set_transport_header(skb, skb_gro_offset(skb));
867
868         flush += ntohs(iph->payload_len) != skb_gro_len(skb);
869
870         rcu_read_lock();
871         proto = iph->nexthdr;
872         ops = rcu_dereference(inet6_protos[proto]);
873         if (!ops || !ops->gro_receive) {
874                 __pskb_pull(skb, skb_gro_offset(skb));
875                 proto = ipv6_gso_pull_exthdrs(skb, proto);
876                 skb_gro_pull(skb, -skb_transport_offset(skb));
877                 skb_reset_transport_header(skb);
878                 __skb_push(skb, skb_gro_offset(skb));
879
880                 if (!ops || !ops->gro_receive)
881                         goto out_unlock;
882
883                 iph = ipv6_hdr(skb);
884         }
885
886         IPV6_GRO_CB(skb)->proto = proto;
887
888         flush--;
889         nlen = skb_network_header_len(skb);
890
891         for (p = *head; p; p = p->next) {
892                 struct ipv6hdr *iph2;
893
894                 if (!NAPI_GRO_CB(p)->same_flow)
895                         continue;
896
897                 iph2 = ipv6_hdr(p);
898
899                 /* All fields must match except length. */
900                 if (nlen != skb_network_header_len(p) ||
901                     memcmp(iph, iph2, offsetof(struct ipv6hdr, payload_len)) ||
902                     memcmp(&iph->nexthdr, &iph2->nexthdr,
903                            nlen - offsetof(struct ipv6hdr, nexthdr))) {
904                         NAPI_GRO_CB(p)->same_flow = 0;
905                         continue;
906                 }
907
908                 NAPI_GRO_CB(p)->flush |= flush;
909         }
910
911         NAPI_GRO_CB(skb)->flush |= flush;
912
913         csum = skb->csum;
914         skb_postpull_rcsum(skb, iph, skb_network_header_len(skb));
915
916         pp = ops->gro_receive(head, skb);
917
918         skb->csum = csum;
919
920 out_unlock:
921         rcu_read_unlock();
922
923 out:
924         NAPI_GRO_CB(skb)->flush |= flush;
925
926         return pp;
927 }
928
929 static int ipv6_gro_complete(struct sk_buff *skb)
930 {
931         const struct inet6_protocol *ops;
932         struct ipv6hdr *iph = ipv6_hdr(skb);
933         int err = -ENOSYS;
934
935         iph->payload_len = htons(skb->len - skb_network_offset(skb) -
936                                  sizeof(*iph));
937
938         rcu_read_lock();
939         ops = rcu_dereference(inet6_protos[IPV6_GRO_CB(skb)->proto]);
940         if (WARN_ON(!ops || !ops->gro_complete))
941                 goto out_unlock;
942
943         err = ops->gro_complete(skb);
944
945 out_unlock:
946         rcu_read_unlock();
947
948         return err;
949 }
950
951 static struct packet_type ipv6_packet_type __read_mostly = {
952         .type = cpu_to_be16(ETH_P_IPV6),
953         .func = ipv6_rcv,
954         .gso_send_check = ipv6_gso_send_check,
955         .gso_segment = ipv6_gso_segment,
956         .gro_receive = ipv6_gro_receive,
957         .gro_complete = ipv6_gro_complete,
958 };
959
960 static int __init ipv6_packet_init(void)
961 {
962         dev_add_pack(&ipv6_packet_type);
963         return 0;
964 }
965
966 static void ipv6_packet_cleanup(void)
967 {
968         dev_remove_pack(&ipv6_packet_type);
969 }
970
971 static int __net_init ipv6_init_mibs(struct net *net)
972 {
973         if (snmp_mib_init((void __percpu **)net->mib.udp_stats_in6,
974                           sizeof(struct udp_mib),
975                           __alignof__(struct udp_mib)) < 0)
976                 return -ENOMEM;
977         if (snmp_mib_init((void __percpu **)net->mib.udplite_stats_in6,
978                           sizeof(struct udp_mib),
979                           __alignof__(struct udp_mib)) < 0)
980                 goto err_udplite_mib;
981         if (snmp_mib_init((void __percpu **)net->mib.ipv6_statistics,
982                           sizeof(struct ipstats_mib),
983                           __alignof__(struct ipstats_mib)) < 0)
984                 goto err_ip_mib;
985         if (snmp_mib_init((void __percpu **)net->mib.icmpv6_statistics,
986                           sizeof(struct icmpv6_mib),
987                           __alignof__(struct icmpv6_mib)) < 0)
988                 goto err_icmp_mib;
989         if (snmp_mib_init((void __percpu **)net->mib.icmpv6msg_statistics,
990                           sizeof(struct icmpv6msg_mib),
991                           __alignof__(struct icmpv6msg_mib)) < 0)
992                 goto err_icmpmsg_mib;
993         return 0;
994
995 err_icmpmsg_mib:
996         snmp_mib_free((void __percpu **)net->mib.icmpv6_statistics);
997 err_icmp_mib:
998         snmp_mib_free((void __percpu **)net->mib.ipv6_statistics);
999 err_ip_mib:
1000         snmp_mib_free((void __percpu **)net->mib.udplite_stats_in6);
1001 err_udplite_mib:
1002         snmp_mib_free((void __percpu **)net->mib.udp_stats_in6);
1003         return -ENOMEM;
1004 }
1005
1006 static void ipv6_cleanup_mibs(struct net *net)
1007 {
1008         snmp_mib_free((void __percpu **)net->mib.udp_stats_in6);
1009         snmp_mib_free((void __percpu **)net->mib.udplite_stats_in6);
1010         snmp_mib_free((void __percpu **)net->mib.ipv6_statistics);
1011         snmp_mib_free((void __percpu **)net->mib.icmpv6_statistics);
1012         snmp_mib_free((void __percpu **)net->mib.icmpv6msg_statistics);
1013 }
1014
1015 static int __net_init inet6_net_init(struct net *net)
1016 {
1017         int err = 0;
1018
1019         net->ipv6.sysctl.bindv6only = 0;
1020         net->ipv6.sysctl.icmpv6_time = 1*HZ;
1021
1022         err = ipv6_init_mibs(net);
1023         if (err)
1024                 return err;
1025 #ifdef CONFIG_PROC_FS
1026         err = udp6_proc_init(net);
1027         if (err)
1028                 goto out;
1029         err = tcp6_proc_init(net);
1030         if (err)
1031                 goto proc_tcp6_fail;
1032         err = ac6_proc_init(net);
1033         if (err)
1034                 goto proc_ac6_fail;
1035 #endif
1036         return err;
1037
1038 #ifdef CONFIG_PROC_FS
1039 proc_ac6_fail:
1040         tcp6_proc_exit(net);
1041 proc_tcp6_fail:
1042         udp6_proc_exit(net);
1043 out:
1044         ipv6_cleanup_mibs(net);
1045         return err;
1046 #endif
1047 }
1048
1049 static void __net_exit inet6_net_exit(struct net *net)
1050 {
1051 #ifdef CONFIG_PROC_FS
1052         udp6_proc_exit(net);
1053         tcp6_proc_exit(net);
1054         ac6_proc_exit(net);
1055 #endif
1056         ipv6_cleanup_mibs(net);
1057 }
1058
1059 static struct pernet_operations inet6_net_ops = {
1060         .init = inet6_net_init,
1061         .exit = inet6_net_exit,
1062 };
1063
1064 static int __init inet6_init(void)
1065 {
1066         struct sk_buff *dummy_skb;
1067         struct list_head *r;
1068         int err = 0;
1069
1070         BUILD_BUG_ON(sizeof(struct inet6_skb_parm) > sizeof(dummy_skb->cb));
1071
1072         /* Register the socket-side information for inet6_create.  */
1073         for(r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
1074                 INIT_LIST_HEAD(r);
1075
1076         if (disable_ipv6_mod) {
1077                 printk(KERN_INFO
1078                        "IPv6: Loaded, but administratively disabled, "
1079                        "reboot required to enable\n");
1080                 goto out;
1081         }
1082
1083         err = proto_register(&tcpv6_prot, 1);
1084         if (err)
1085                 goto out;
1086
1087         err = proto_register(&udpv6_prot, 1);
1088         if (err)
1089                 goto out_unregister_tcp_proto;
1090
1091         err = proto_register(&udplitev6_prot, 1);
1092         if (err)
1093                 goto out_unregister_udp_proto;
1094
1095         err = proto_register(&rawv6_prot, 1);
1096         if (err)
1097                 goto out_unregister_udplite_proto;
1098
1099
1100         /* We MUST register RAW sockets before we create the ICMP6,
1101          * IGMP6, or NDISC control sockets.
1102          */
1103         err = rawv6_init();
1104         if (err)
1105                 goto out_unregister_raw_proto;
1106
1107         /* Register the family here so that the init calls below will
1108          * be able to create sockets. (?? is this dangerous ??)
1109          */
1110         err = sock_register(&inet6_family_ops);
1111         if (err)
1112                 goto out_sock_register_fail;
1113
1114 #ifdef CONFIG_SYSCTL
1115         err = ipv6_static_sysctl_register();
1116         if (err)
1117                 goto static_sysctl_fail;
1118 #endif
1119         /*
1120          *      ipngwg API draft makes clear that the correct semantics
1121          *      for TCP and UDP is to consider one TCP and UDP instance
1122          *      in a host availiable by both INET and INET6 APIs and
1123          *      able to communicate via both network protocols.
1124          */
1125
1126         err = register_pernet_subsys(&inet6_net_ops);
1127         if (err)
1128                 goto register_pernet_fail;
1129         err = icmpv6_init();
1130         if (err)
1131                 goto icmp_fail;
1132         err = ip6_mr_init();
1133         if (err)
1134                 goto ipmr_fail;
1135         err = ndisc_init();
1136         if (err)
1137                 goto ndisc_fail;
1138         err = igmp6_init();
1139         if (err)
1140                 goto igmp_fail;
1141         err = ipv6_netfilter_init();
1142         if (err)
1143                 goto netfilter_fail;
1144         /* Create /proc/foo6 entries. */
1145 #ifdef CONFIG_PROC_FS
1146         err = -ENOMEM;
1147         if (raw6_proc_init())
1148                 goto proc_raw6_fail;
1149         if (udplite6_proc_init())
1150                 goto proc_udplite6_fail;
1151         if (ipv6_misc_proc_init())
1152                 goto proc_misc6_fail;
1153         if (if6_proc_init())
1154                 goto proc_if6_fail;
1155 #endif
1156         err = ip6_route_init();
1157         if (err)
1158                 goto ip6_route_fail;
1159         err = ip6_flowlabel_init();
1160         if (err)
1161                 goto ip6_flowlabel_fail;
1162         err = addrconf_init();
1163         if (err)
1164                 goto addrconf_fail;
1165
1166         /* Init v6 extension headers. */
1167         err = ipv6_exthdrs_init();
1168         if (err)
1169                 goto ipv6_exthdrs_fail;
1170
1171         err = ipv6_frag_init();
1172         if (err)
1173                 goto ipv6_frag_fail;
1174
1175         /* Init v6 transport protocols. */
1176         err = udpv6_init();
1177         if (err)
1178                 goto udpv6_fail;
1179
1180         err = udplitev6_init();
1181         if (err)
1182                 goto udplitev6_fail;
1183
1184         err = tcpv6_init();
1185         if (err)
1186                 goto tcpv6_fail;
1187
1188         err = ipv6_packet_init();
1189         if (err)
1190                 goto ipv6_packet_fail;
1191
1192 #ifdef CONFIG_SYSCTL
1193         err = ipv6_sysctl_register();
1194         if (err)
1195                 goto sysctl_fail;
1196 #endif
1197 out:
1198         return err;
1199
1200 #ifdef CONFIG_SYSCTL
1201 sysctl_fail:
1202         ipv6_packet_cleanup();
1203 #endif
1204 ipv6_packet_fail:
1205         tcpv6_exit();
1206 tcpv6_fail:
1207         udplitev6_exit();
1208 udplitev6_fail:
1209         udpv6_exit();
1210 udpv6_fail:
1211         ipv6_frag_exit();
1212 ipv6_frag_fail:
1213         ipv6_exthdrs_exit();
1214 ipv6_exthdrs_fail:
1215         addrconf_cleanup();
1216 addrconf_fail:
1217         ip6_flowlabel_cleanup();
1218 ip6_flowlabel_fail:
1219         ip6_route_cleanup();
1220 ip6_route_fail:
1221 #ifdef CONFIG_PROC_FS
1222         if6_proc_exit();
1223 proc_if6_fail:
1224         ipv6_misc_proc_exit();
1225 proc_misc6_fail:
1226         udplite6_proc_exit();
1227 proc_udplite6_fail:
1228         raw6_proc_exit();
1229 proc_raw6_fail:
1230 #endif
1231         ipv6_netfilter_fini();
1232 netfilter_fail:
1233         igmp6_cleanup();
1234 igmp_fail:
1235         ndisc_cleanup();
1236 ndisc_fail:
1237         ip6_mr_cleanup();
1238 ipmr_fail:
1239         icmpv6_cleanup();
1240 icmp_fail:
1241         unregister_pernet_subsys(&inet6_net_ops);
1242 register_pernet_fail:
1243 #ifdef CONFIG_SYSCTL
1244         ipv6_static_sysctl_unregister();
1245 static_sysctl_fail:
1246 #endif
1247         sock_unregister(PF_INET6);
1248         rtnl_unregister_all(PF_INET6);
1249 out_sock_register_fail:
1250         rawv6_exit();
1251 out_unregister_raw_proto:
1252         proto_unregister(&rawv6_prot);
1253 out_unregister_udplite_proto:
1254         proto_unregister(&udplitev6_prot);
1255 out_unregister_udp_proto:
1256         proto_unregister(&udpv6_prot);
1257 out_unregister_tcp_proto:
1258         proto_unregister(&tcpv6_prot);
1259         goto out;
1260 }
1261 module_init(inet6_init);
1262
1263 static void __exit inet6_exit(void)
1264 {
1265         if (disable_ipv6_mod)
1266                 return;
1267
1268         /* First of all disallow new sockets creation. */
1269         sock_unregister(PF_INET6);
1270         /* Disallow any further netlink messages */
1271         rtnl_unregister_all(PF_INET6);
1272
1273 #ifdef CONFIG_SYSCTL
1274         ipv6_sysctl_unregister();
1275 #endif
1276         udpv6_exit();
1277         udplitev6_exit();
1278         tcpv6_exit();
1279
1280         /* Cleanup code parts. */
1281         ipv6_packet_cleanup();
1282         ipv6_frag_exit();
1283         ipv6_exthdrs_exit();
1284         addrconf_cleanup();
1285         ip6_flowlabel_cleanup();
1286         ip6_route_cleanup();
1287 #ifdef CONFIG_PROC_FS
1288
1289         /* Cleanup code parts. */
1290         if6_proc_exit();
1291         ipv6_misc_proc_exit();
1292         udplite6_proc_exit();
1293         raw6_proc_exit();
1294 #endif
1295         ipv6_netfilter_fini();
1296         igmp6_cleanup();
1297         ndisc_cleanup();
1298         ip6_mr_cleanup();
1299         icmpv6_cleanup();
1300         rawv6_exit();
1301
1302         unregister_pernet_subsys(&inet6_net_ops);
1303 #ifdef CONFIG_SYSCTL
1304         ipv6_static_sysctl_unregister();
1305 #endif
1306         proto_unregister(&rawv6_prot);
1307         proto_unregister(&udplitev6_prot);
1308         proto_unregister(&udpv6_prot);
1309         proto_unregister(&tcpv6_prot);
1310
1311         rcu_barrier(); /* Wait for completion of call_rcu()'s */
1312 }
1313 module_exit(inet6_exit);
1314
1315 MODULE_ALIAS_NETPROTO(PF_INET6);