[NET] NETNS: Omit sock->sk_net without CONFIG_NET_NS.
[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  *      $Id: af_inet6.c,v 1.66 2002/02/01 22:01:04 davem Exp $
11  *
12  *      Fixes:
13  *      piggy, Karl Knutson     :       Socket protocol table
14  *      Hideaki YOSHIFUJI       :       sin6_scope_id support
15  *      Arnaldo Melo            :       check proc_net_create return, cleanups
16  *
17  *      This program is free software; you can redistribute it and/or
18  *      modify it under the terms of the GNU General Public License
19  *      as published by the Free Software Foundation; either version
20  *      2 of the License, or (at your option) any later version.
21  */
22
23
24 #include <linux/module.h>
25 #include <linux/capability.h>
26 #include <linux/errno.h>
27 #include <linux/types.h>
28 #include <linux/socket.h>
29 #include <linux/in.h>
30 #include <linux/kernel.h>
31 #include <linux/timer.h>
32 #include <linux/string.h>
33 #include <linux/sockios.h>
34 #include <linux/net.h>
35 #include <linux/fcntl.h>
36 #include <linux/mm.h>
37 #include <linux/interrupt.h>
38 #include <linux/proc_fs.h>
39 #include <linux/stat.h>
40 #include <linux/init.h>
41
42 #include <linux/inet.h>
43 #include <linux/netdevice.h>
44 #include <linux/icmpv6.h>
45 #include <linux/netfilter_ipv6.h>
46
47 #include <net/ip.h>
48 #include <net/ipv6.h>
49 #include <net/udp.h>
50 #include <net/udplite.h>
51 #include <net/tcp.h>
52 #include <net/ipip.h>
53 #include <net/protocol.h>
54 #include <net/inet_common.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
65 MODULE_AUTHOR("Cast of dozens");
66 MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
67 MODULE_LICENSE("GPL");
68
69 /* The inetsw6 table contains everything that inet6_create needs to
70  * build a new socket.
71  */
72 static struct list_head inetsw6[SOCK_MAX];
73 static DEFINE_SPINLOCK(inetsw6_lock);
74
75 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
76 {
77         const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
78
79         return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
80 }
81
82 static int inet6_create(struct net *net, struct socket *sock, int protocol)
83 {
84         struct inet_sock *inet;
85         struct ipv6_pinfo *np;
86         struct sock *sk;
87         struct list_head *p;
88         struct inet_protosw *answer;
89         struct proto *answer_prot;
90         unsigned char answer_flags;
91         char answer_no_check;
92         int try_loading_module = 0;
93         int err;
94
95         if (sock->type != SOCK_RAW &&
96             sock->type != SOCK_DGRAM &&
97             !inet_ehash_secret)
98                 build_ehash_secret();
99
100         /* Look for the requested type/protocol pair. */
101         answer = NULL;
102 lookup_protocol:
103         err = -ESOCKTNOSUPPORT;
104         rcu_read_lock();
105         list_for_each_rcu(p, &inetsw6[sock->type]) {
106                 answer = list_entry(p, struct inet_protosw, list);
107
108                 /* Check the non-wild match. */
109                 if (protocol == answer->protocol) {
110                         if (protocol != IPPROTO_IP)
111                                 break;
112                 } else {
113                         /* Check for the two wild cases. */
114                         if (IPPROTO_IP == protocol) {
115                                 protocol = answer->protocol;
116                                 break;
117                         }
118                         if (IPPROTO_IP == answer->protocol)
119                                 break;
120                 }
121                 err = -EPROTONOSUPPORT;
122                 answer = NULL;
123         }
124
125         if (!answer) {
126                 if (try_loading_module < 2) {
127                         rcu_read_unlock();
128                         /*
129                          * Be more specific, e.g. net-pf-10-proto-132-type-1
130                          * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
131                          */
132                         if (++try_loading_module == 1)
133                                 request_module("net-pf-%d-proto-%d-type-%d",
134                                                 PF_INET6, protocol, sock->type);
135                         /*
136                          * Fall back to generic, e.g. net-pf-10-proto-132
137                          * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
138                          */
139                         else
140                                 request_module("net-pf-%d-proto-%d",
141                                                 PF_INET6, protocol);
142                         goto lookup_protocol;
143                 } else
144                         goto out_rcu_unlock;
145         }
146
147         err = -EPERM;
148         if (answer->capability > 0 && !capable(answer->capability))
149                 goto out_rcu_unlock;
150
151         sock->ops = answer->ops;
152         answer_prot = answer->prot;
153         answer_no_check = answer->no_check;
154         answer_flags = answer->flags;
155         rcu_read_unlock();
156
157         BUG_TRAP(answer_prot->slab != NULL);
158
159         err = -ENOBUFS;
160         sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot);
161         if (sk == NULL)
162                 goto out;
163
164         sock_init_data(sock, sk);
165
166         err = 0;
167         sk->sk_no_check = answer_no_check;
168         if (INET_PROTOSW_REUSE & answer_flags)
169                 sk->sk_reuse = 1;
170
171         inet = inet_sk(sk);
172         inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
173
174         if (SOCK_RAW == sock->type) {
175                 inet->num = protocol;
176                 if (IPPROTO_RAW == protocol)
177                         inet->hdrincl = 1;
178         }
179
180         sk->sk_destruct         = inet_sock_destruct;
181         sk->sk_family           = PF_INET6;
182         sk->sk_protocol         = protocol;
183
184         sk->sk_backlog_rcv      = answer->prot->backlog_rcv;
185
186         inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
187         np->hop_limit   = -1;
188         np->mcast_hops  = -1;
189         np->mc_loop     = 1;
190         np->pmtudisc    = IPV6_PMTUDISC_WANT;
191         np->ipv6only    = init_net.ipv6.sysctl.bindv6only;
192
193         /* Init the ipv4 part of the socket since we can have sockets
194          * using v6 API for ipv4.
195          */
196         inet->uc_ttl    = -1;
197
198         inet->mc_loop   = 1;
199         inet->mc_ttl    = 1;
200         inet->mc_index  = 0;
201         inet->mc_list   = NULL;
202
203         if (ipv4_config.no_pmtu_disc)
204                 inet->pmtudisc = IP_PMTUDISC_DONT;
205         else
206                 inet->pmtudisc = IP_PMTUDISC_WANT;
207         /*
208          * Increment only the relevant sk_prot->socks debug field, this changes
209          * the previous behaviour of incrementing both the equivalent to
210          * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
211          *
212          * This allows better debug granularity as we'll know exactly how many
213          * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
214          * transport protocol socks. -acme
215          */
216         sk_refcnt_debug_inc(sk);
217
218         if (inet->num) {
219                 /* It assumes that any protocol which allows
220                  * the user to assign a number at socket
221                  * creation time automatically shares.
222                  */
223                 inet->sport = htons(inet->num);
224                 sk->sk_prot->hash(sk);
225         }
226         if (sk->sk_prot->init) {
227                 err = sk->sk_prot->init(sk);
228                 if (err) {
229                         sk_common_release(sk);
230                         goto out;
231                 }
232         }
233 out:
234         return err;
235 out_rcu_unlock:
236         rcu_read_unlock();
237         goto out;
238 }
239
240
241 /* bind for INET6 API */
242 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
243 {
244         struct sockaddr_in6 *addr=(struct sockaddr_in6 *)uaddr;
245         struct sock *sk = sock->sk;
246         struct inet_sock *inet = inet_sk(sk);
247         struct ipv6_pinfo *np = inet6_sk(sk);
248         struct net *net = sock_net(sk);
249         __be32 v4addr = 0;
250         unsigned short snum;
251         int addr_type = 0;
252         int err = 0;
253
254         /* If the socket has its own bind function then use it. */
255         if (sk->sk_prot->bind)
256                 return sk->sk_prot->bind(sk, uaddr, addr_len);
257
258         if (addr_len < SIN6_LEN_RFC2133)
259                 return -EINVAL;
260         addr_type = ipv6_addr_type(&addr->sin6_addr);
261         if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
262                 return -EINVAL;
263
264         snum = ntohs(addr->sin6_port);
265         if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
266                 return -EACCES;
267
268         lock_sock(sk);
269
270         /* Check these errors (active socket, double bind). */
271         if (sk->sk_state != TCP_CLOSE || inet->num) {
272                 err = -EINVAL;
273                 goto out;
274         }
275
276         /* Check if the address belongs to the host. */
277         if (addr_type == IPV6_ADDR_MAPPED) {
278                 v4addr = addr->sin6_addr.s6_addr32[3];
279                 if (inet_addr_type(net, v4addr) != RTN_LOCAL) {
280                         err = -EADDRNOTAVAIL;
281                         goto out;
282                 }
283         } else {
284                 if (addr_type != IPV6_ADDR_ANY) {
285                         struct net_device *dev = NULL;
286
287                         if (addr_type & IPV6_ADDR_LINKLOCAL) {
288                                 if (addr_len >= sizeof(struct sockaddr_in6) &&
289                                     addr->sin6_scope_id) {
290                                         /* Override any existing binding, if another one
291                                          * is supplied by user.
292                                          */
293                                         sk->sk_bound_dev_if = addr->sin6_scope_id;
294                                 }
295
296                                 /* Binding to link-local address requires an interface */
297                                 if (!sk->sk_bound_dev_if) {
298                                         err = -EINVAL;
299                                         goto out;
300                                 }
301                                 dev = dev_get_by_index(net, sk->sk_bound_dev_if);
302                                 if (!dev) {
303                                         err = -ENODEV;
304                                         goto out;
305                                 }
306                         }
307
308                         /* ipv4 addr of the socket is invalid.  Only the
309                          * unspecified and mapped address have a v4 equivalent.
310                          */
311                         v4addr = LOOPBACK4_IPV6;
312                         if (!(addr_type & IPV6_ADDR_MULTICAST)) {
313                                 if (!ipv6_chk_addr(net, &addr->sin6_addr,
314                                                    dev, 0)) {
315                                         if (dev)
316                                                 dev_put(dev);
317                                         err = -EADDRNOTAVAIL;
318                                         goto out;
319                                 }
320                         }
321                         if (dev)
322                                 dev_put(dev);
323                 }
324         }
325
326         inet->rcv_saddr = v4addr;
327         inet->saddr = v4addr;
328
329         ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
330
331         if (!(addr_type & IPV6_ADDR_MULTICAST))
332                 ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
333
334         /* Make sure we are allowed to bind here. */
335         if (sk->sk_prot->get_port(sk, snum)) {
336                 inet_reset_saddr(sk);
337                 err = -EADDRINUSE;
338                 goto out;
339         }
340
341         if (addr_type != IPV6_ADDR_ANY)
342                 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
343         if (snum)
344                 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
345         inet->sport = htons(inet->num);
346         inet->dport = 0;
347         inet->daddr = 0;
348 out:
349         release_sock(sk);
350         return err;
351 }
352
353 EXPORT_SYMBOL(inet6_bind);
354
355 int inet6_release(struct socket *sock)
356 {
357         struct sock *sk = sock->sk;
358
359         if (sk == NULL)
360                 return -EINVAL;
361
362         /* Free mc lists */
363         ipv6_sock_mc_close(sk);
364
365         /* Free ac lists */
366         ipv6_sock_ac_close(sk);
367
368         return inet_release(sock);
369 }
370
371 EXPORT_SYMBOL(inet6_release);
372
373 int inet6_destroy_sock(struct sock *sk)
374 {
375         struct ipv6_pinfo *np = inet6_sk(sk);
376         struct sk_buff *skb;
377         struct ipv6_txoptions *opt;
378
379         /* Release rx options */
380
381         if ((skb = xchg(&np->pktoptions, NULL)) != NULL)
382                 kfree_skb(skb);
383
384         /* Free flowlabels */
385         fl6_free_socklist(sk);
386
387         /* Free tx options */
388
389         if ((opt = xchg(&np->opt, NULL)) != NULL)
390                 sock_kfree_s(sk, opt, opt->tot_len);
391
392         return 0;
393 }
394
395 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
396
397 /*
398  *      This does both peername and sockname.
399  */
400
401 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
402                  int *uaddr_len, int peer)
403 {
404         struct sockaddr_in6 *sin=(struct sockaddr_in6 *)uaddr;
405         struct sock *sk = sock->sk;
406         struct inet_sock *inet = inet_sk(sk);
407         struct ipv6_pinfo *np = inet6_sk(sk);
408
409         sin->sin6_family = AF_INET6;
410         sin->sin6_flowinfo = 0;
411         sin->sin6_scope_id = 0;
412         if (peer) {
413                 if (!inet->dport)
414                         return -ENOTCONN;
415                 if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
416                     peer == 1)
417                         return -ENOTCONN;
418                 sin->sin6_port = inet->dport;
419                 ipv6_addr_copy(&sin->sin6_addr, &np->daddr);
420                 if (np->sndflow)
421                         sin->sin6_flowinfo = np->flow_label;
422         } else {
423                 if (ipv6_addr_any(&np->rcv_saddr))
424                         ipv6_addr_copy(&sin->sin6_addr, &np->saddr);
425                 else
426                         ipv6_addr_copy(&sin->sin6_addr, &np->rcv_saddr);
427
428                 sin->sin6_port = inet->sport;
429         }
430         if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL)
431                 sin->sin6_scope_id = sk->sk_bound_dev_if;
432         *uaddr_len = sizeof(*sin);
433         return(0);
434 }
435
436 EXPORT_SYMBOL(inet6_getname);
437
438 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
439 {
440         struct sock *sk = sock->sk;
441         struct net *net = sock_net(sk);
442
443         switch(cmd)
444         {
445         case SIOCGSTAMP:
446                 return sock_get_timestamp(sk, (struct timeval __user *)arg);
447
448         case SIOCGSTAMPNS:
449                 return sock_get_timestampns(sk, (struct timespec __user *)arg);
450
451         case SIOCADDRT:
452         case SIOCDELRT:
453
454                 return(ipv6_route_ioctl(net, cmd, (void __user *)arg));
455
456         case SIOCSIFADDR:
457                 return addrconf_add_ifaddr(net, (void __user *) arg);
458         case SIOCDIFADDR:
459                 return addrconf_del_ifaddr(net, (void __user *) arg);
460         case SIOCSIFDSTADDR:
461                 return addrconf_set_dstaddr(net, (void __user *) arg);
462         default:
463                 if (!sk->sk_prot->ioctl)
464                         return -ENOIOCTLCMD;
465                 return sk->sk_prot->ioctl(sk, cmd, arg);
466         }
467         /*NOTREACHED*/
468         return(0);
469 }
470
471 EXPORT_SYMBOL(inet6_ioctl);
472
473 const struct proto_ops inet6_stream_ops = {
474         .family            = PF_INET6,
475         .owner             = THIS_MODULE,
476         .release           = inet6_release,
477         .bind              = inet6_bind,
478         .connect           = inet_stream_connect,       /* ok           */
479         .socketpair        = sock_no_socketpair,        /* a do nothing */
480         .accept            = inet_accept,               /* ok           */
481         .getname           = inet6_getname,
482         .poll              = tcp_poll,                  /* ok           */
483         .ioctl             = inet6_ioctl,               /* must change  */
484         .listen            = inet_listen,               /* ok           */
485         .shutdown          = inet_shutdown,             /* ok           */
486         .setsockopt        = sock_common_setsockopt,    /* ok           */
487         .getsockopt        = sock_common_getsockopt,    /* ok           */
488         .sendmsg           = tcp_sendmsg,               /* ok           */
489         .recvmsg           = sock_common_recvmsg,       /* ok           */
490         .mmap              = sock_no_mmap,
491         .sendpage          = tcp_sendpage,
492         .splice_read       = tcp_splice_read,
493 #ifdef CONFIG_COMPAT
494         .compat_setsockopt = compat_sock_common_setsockopt,
495         .compat_getsockopt = compat_sock_common_getsockopt,
496 #endif
497 };
498
499 const struct proto_ops inet6_dgram_ops = {
500         .family            = PF_INET6,
501         .owner             = THIS_MODULE,
502         .release           = inet6_release,
503         .bind              = inet6_bind,
504         .connect           = inet_dgram_connect,        /* ok           */
505         .socketpair        = sock_no_socketpair,        /* a do nothing */
506         .accept            = sock_no_accept,            /* a do nothing */
507         .getname           = inet6_getname,
508         .poll              = udp_poll,                  /* ok           */
509         .ioctl             = inet6_ioctl,               /* must change  */
510         .listen            = sock_no_listen,            /* ok           */
511         .shutdown          = inet_shutdown,             /* ok           */
512         .setsockopt        = sock_common_setsockopt,    /* ok           */
513         .getsockopt        = sock_common_getsockopt,    /* ok           */
514         .sendmsg           = inet_sendmsg,              /* ok           */
515         .recvmsg           = sock_common_recvmsg,       /* ok           */
516         .mmap              = sock_no_mmap,
517         .sendpage          = sock_no_sendpage,
518 #ifdef CONFIG_COMPAT
519         .compat_setsockopt = compat_sock_common_setsockopt,
520         .compat_getsockopt = compat_sock_common_getsockopt,
521 #endif
522 };
523
524 static struct net_proto_family inet6_family_ops = {
525         .family = PF_INET6,
526         .create = inet6_create,
527         .owner  = THIS_MODULE,
528 };
529
530 int inet6_register_protosw(struct inet_protosw *p)
531 {
532         struct list_head *lh;
533         struct inet_protosw *answer;
534         struct list_head *last_perm;
535         int protocol = p->protocol;
536         int ret;
537
538         spin_lock_bh(&inetsw6_lock);
539
540         ret = -EINVAL;
541         if (p->type >= SOCK_MAX)
542                 goto out_illegal;
543
544         /* If we are trying to override a permanent protocol, bail. */
545         answer = NULL;
546         ret = -EPERM;
547         last_perm = &inetsw6[p->type];
548         list_for_each(lh, &inetsw6[p->type]) {
549                 answer = list_entry(lh, struct inet_protosw, list);
550
551                 /* Check only the non-wild match. */
552                 if (INET_PROTOSW_PERMANENT & answer->flags) {
553                         if (protocol == answer->protocol)
554                                 break;
555                         last_perm = lh;
556                 }
557
558                 answer = NULL;
559         }
560         if (answer)
561                 goto out_permanent;
562
563         /* Add the new entry after the last permanent entry if any, so that
564          * the new entry does not override a permanent entry when matched with
565          * a wild-card protocol. But it is allowed to override any existing
566          * non-permanent entry.  This means that when we remove this entry, the
567          * system automatically returns to the old behavior.
568          */
569         list_add_rcu(&p->list, last_perm);
570         ret = 0;
571 out:
572         spin_unlock_bh(&inetsw6_lock);
573         return ret;
574
575 out_permanent:
576         printk(KERN_ERR "Attempt to override permanent protocol %d.\n",
577                protocol);
578         goto out;
579
580 out_illegal:
581         printk(KERN_ERR
582                "Ignoring attempt to register invalid socket type %d.\n",
583                p->type);
584         goto out;
585 }
586
587 EXPORT_SYMBOL(inet6_register_protosw);
588
589 void
590 inet6_unregister_protosw(struct inet_protosw *p)
591 {
592         if (INET_PROTOSW_PERMANENT & p->flags) {
593                 printk(KERN_ERR
594                        "Attempt to unregister permanent protocol %d.\n",
595                        p->protocol);
596         } else {
597                 spin_lock_bh(&inetsw6_lock);
598                 list_del_rcu(&p->list);
599                 spin_unlock_bh(&inetsw6_lock);
600
601                 synchronize_net();
602         }
603 }
604
605 EXPORT_SYMBOL(inet6_unregister_protosw);
606
607 int inet6_sk_rebuild_header(struct sock *sk)
608 {
609         int err;
610         struct dst_entry *dst;
611         struct ipv6_pinfo *np = inet6_sk(sk);
612
613         dst = __sk_dst_check(sk, np->dst_cookie);
614
615         if (dst == NULL) {
616                 struct inet_sock *inet = inet_sk(sk);
617                 struct in6_addr *final_p = NULL, final;
618                 struct flowi fl;
619
620                 memset(&fl, 0, sizeof(fl));
621                 fl.proto = sk->sk_protocol;
622                 ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
623                 ipv6_addr_copy(&fl.fl6_src, &np->saddr);
624                 fl.fl6_flowlabel = np->flow_label;
625                 fl.oif = sk->sk_bound_dev_if;
626                 fl.fl_ip_dport = inet->dport;
627                 fl.fl_ip_sport = inet->sport;
628                 security_sk_classify_flow(sk, &fl);
629
630                 if (np->opt && np->opt->srcrt) {
631                         struct rt0_hdr *rt0 = (struct rt0_hdr *) np->opt->srcrt;
632                         ipv6_addr_copy(&final, &fl.fl6_dst);
633                         ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
634                         final_p = &final;
635                 }
636
637                 err = ip6_dst_lookup(sk, &dst, &fl);
638                 if (err) {
639                         sk->sk_route_caps = 0;
640                         return err;
641                 }
642                 if (final_p)
643                         ipv6_addr_copy(&fl.fl6_dst, final_p);
644
645                 if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0) {
646                         sk->sk_err_soft = -err;
647                         return err;
648                 }
649
650                 __ip6_dst_store(sk, dst, NULL, NULL);
651         }
652
653         return 0;
654 }
655
656 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
657
658 int ipv6_opt_accepted(struct sock *sk, struct sk_buff *skb)
659 {
660         struct ipv6_pinfo *np = inet6_sk(sk);
661         struct inet6_skb_parm *opt = IP6CB(skb);
662
663         if (np->rxopt.all) {
664                 if ((opt->hop && (np->rxopt.bits.hopopts ||
665                                   np->rxopt.bits.ohopopts)) ||
666                     ((IPV6_FLOWINFO_MASK &
667                       *(__be32 *)skb_network_header(skb)) &&
668                      np->rxopt.bits.rxflow) ||
669                     (opt->srcrt && (np->rxopt.bits.srcrt ||
670                      np->rxopt.bits.osrcrt)) ||
671                     ((opt->dst1 || opt->dst0) &&
672                      (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
673                         return 1;
674         }
675         return 0;
676 }
677
678 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
679
680 static struct inet6_protocol *ipv6_gso_pull_exthdrs(struct sk_buff *skb,
681                                                     int proto)
682 {
683         struct inet6_protocol *ops = NULL;
684
685         for (;;) {
686                 struct ipv6_opt_hdr *opth;
687                 int len;
688
689                 if (proto != NEXTHDR_HOP) {
690                         ops = rcu_dereference(inet6_protos[proto]);
691
692                         if (unlikely(!ops))
693                                 break;
694
695                         if (!(ops->flags & INET6_PROTO_GSO_EXTHDR))
696                                 break;
697                 }
698
699                 if (unlikely(!pskb_may_pull(skb, 8)))
700                         break;
701
702                 opth = (void *)skb->data;
703                 len = ipv6_optlen(opth);
704
705                 if (unlikely(!pskb_may_pull(skb, len)))
706                         break;
707
708                 proto = opth->nexthdr;
709                 __skb_pull(skb, len);
710         }
711
712         return ops;
713 }
714
715 static int ipv6_gso_send_check(struct sk_buff *skb)
716 {
717         struct ipv6hdr *ipv6h;
718         struct inet6_protocol *ops;
719         int err = -EINVAL;
720
721         if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
722                 goto out;
723
724         ipv6h = ipv6_hdr(skb);
725         __skb_pull(skb, sizeof(*ipv6h));
726         err = -EPROTONOSUPPORT;
727
728         rcu_read_lock();
729         ops = ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr);
730         if (likely(ops && ops->gso_send_check)) {
731                 skb_reset_transport_header(skb);
732                 err = ops->gso_send_check(skb);
733         }
734         rcu_read_unlock();
735
736 out:
737         return err;
738 }
739
740 static struct sk_buff *ipv6_gso_segment(struct sk_buff *skb, int features)
741 {
742         struct sk_buff *segs = ERR_PTR(-EINVAL);
743         struct ipv6hdr *ipv6h;
744         struct inet6_protocol *ops;
745
746         if (!(features & NETIF_F_V6_CSUM))
747                 features &= ~NETIF_F_SG;
748
749         if (unlikely(skb_shinfo(skb)->gso_type &
750                      ~(SKB_GSO_UDP |
751                        SKB_GSO_DODGY |
752                        SKB_GSO_TCP_ECN |
753                        SKB_GSO_TCPV6 |
754                        0)))
755                 goto out;
756
757         if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
758                 goto out;
759
760         ipv6h = ipv6_hdr(skb);
761         __skb_pull(skb, sizeof(*ipv6h));
762         segs = ERR_PTR(-EPROTONOSUPPORT);
763
764         rcu_read_lock();
765         ops = ipv6_gso_pull_exthdrs(skb, ipv6h->nexthdr);
766         if (likely(ops && ops->gso_segment)) {
767                 skb_reset_transport_header(skb);
768                 segs = ops->gso_segment(skb, features);
769         }
770         rcu_read_unlock();
771
772         if (unlikely(IS_ERR(segs)))
773                 goto out;
774
775         for (skb = segs; skb; skb = skb->next) {
776                 ipv6h = ipv6_hdr(skb);
777                 ipv6h->payload_len = htons(skb->len - skb->mac_len -
778                                            sizeof(*ipv6h));
779         }
780
781 out:
782         return segs;
783 }
784
785 static struct packet_type ipv6_packet_type = {
786         .type = __constant_htons(ETH_P_IPV6),
787         .func = ipv6_rcv,
788         .gso_send_check = ipv6_gso_send_check,
789         .gso_segment = ipv6_gso_segment,
790 };
791
792 static int __init ipv6_packet_init(void)
793 {
794         dev_add_pack(&ipv6_packet_type);
795         return 0;
796 }
797
798 static void ipv6_packet_cleanup(void)
799 {
800         dev_remove_pack(&ipv6_packet_type);
801 }
802
803 static int __init init_ipv6_mibs(void)
804 {
805         if (snmp_mib_init((void **)ipv6_statistics,
806                           sizeof(struct ipstats_mib)) < 0)
807                 goto err_ip_mib;
808         if (snmp_mib_init((void **)icmpv6_statistics,
809                           sizeof(struct icmpv6_mib)) < 0)
810                 goto err_icmp_mib;
811         if (snmp_mib_init((void **)icmpv6msg_statistics,
812                           sizeof(struct icmpv6msg_mib)) < 0)
813                 goto err_icmpmsg_mib;
814         if (snmp_mib_init((void **)udp_stats_in6, sizeof (struct udp_mib)) < 0)
815                 goto err_udp_mib;
816         if (snmp_mib_init((void **)udplite_stats_in6,
817                           sizeof (struct udp_mib)) < 0)
818                 goto err_udplite_mib;
819         return 0;
820
821 err_udplite_mib:
822         snmp_mib_free((void **)udp_stats_in6);
823 err_udp_mib:
824         snmp_mib_free((void **)icmpv6msg_statistics);
825 err_icmpmsg_mib:
826         snmp_mib_free((void **)icmpv6_statistics);
827 err_icmp_mib:
828         snmp_mib_free((void **)ipv6_statistics);
829 err_ip_mib:
830         return -ENOMEM;
831
832 }
833
834 static void cleanup_ipv6_mibs(void)
835 {
836         snmp_mib_free((void **)ipv6_statistics);
837         snmp_mib_free((void **)icmpv6_statistics);
838         snmp_mib_free((void **)icmpv6msg_statistics);
839         snmp_mib_free((void **)udp_stats_in6);
840         snmp_mib_free((void **)udplite_stats_in6);
841 }
842
843 static int inet6_net_init(struct net *net)
844 {
845         int err = 0;
846
847         net->ipv6.sysctl.bindv6only = 0;
848         net->ipv6.sysctl.flush_delay = 0;
849         net->ipv6.sysctl.ip6_rt_max_size = 4096;
850         net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
851         net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
852         net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
853         net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
854         net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
855         net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
856         net->ipv6.sysctl.icmpv6_time = 1*HZ;
857
858 #ifdef CONFIG_PROC_FS
859         err = udp6_proc_init(net);
860         if (err)
861                 goto out;
862         err = tcp6_proc_init(net);
863         if (err)
864                 goto proc_tcp6_fail;
865 out:
866 #endif
867         return err;
868
869 #ifdef CONFIG_PROC_FS
870 proc_tcp6_fail:
871         udp6_proc_exit(net);
872         goto out;
873 #endif
874 }
875
876 static void inet6_net_exit(struct net *net)
877 {
878 #ifdef CONFIG_PROC_FS
879         udp6_proc_exit(net);
880         tcp6_proc_exit(net);
881 #endif
882 }
883
884 static struct pernet_operations inet6_net_ops = {
885         .init = inet6_net_init,
886         .exit = inet6_net_exit,
887 };
888
889 static int __init inet6_init(void)
890 {
891         struct sk_buff *dummy_skb;
892         struct list_head *r;
893         int err;
894
895         BUILD_BUG_ON(sizeof(struct inet6_skb_parm) > sizeof(dummy_skb->cb));
896
897         err = proto_register(&tcpv6_prot, 1);
898         if (err)
899                 goto out;
900
901         err = proto_register(&udpv6_prot, 1);
902         if (err)
903                 goto out_unregister_tcp_proto;
904
905         err = proto_register(&udplitev6_prot, 1);
906         if (err)
907                 goto out_unregister_udp_proto;
908
909         err = proto_register(&rawv6_prot, 1);
910         if (err)
911                 goto out_unregister_udplite_proto;
912
913
914         /* Register the socket-side information for inet6_create.  */
915         for(r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
916                 INIT_LIST_HEAD(r);
917
918         /* We MUST register RAW sockets before we create the ICMP6,
919          * IGMP6, or NDISC control sockets.
920          */
921         err = rawv6_init();
922         if (err)
923                 goto out_unregister_raw_proto;
924
925         /* Register the family here so that the init calls below will
926          * be able to create sockets. (?? is this dangerous ??)
927          */
928         err = sock_register(&inet6_family_ops);
929         if (err)
930                 goto out_sock_register_fail;
931
932         /* Initialise ipv6 mibs */
933         err = init_ipv6_mibs();
934         if (err)
935                 goto out_unregister_sock;
936
937         /*
938          *      ipngwg API draft makes clear that the correct semantics
939          *      for TCP and UDP is to consider one TCP and UDP instance
940          *      in a host availiable by both INET and INET6 APIs and
941          *      able to communicate via both network protocols.
942          */
943
944         err = register_pernet_subsys(&inet6_net_ops);
945         if (err)
946                 goto register_pernet_fail;
947         err = icmpv6_init();
948         if (err)
949                 goto icmp_fail;
950         err = ndisc_init();
951         if (err)
952                 goto ndisc_fail;
953         err = igmp6_init();
954         if (err)
955                 goto igmp_fail;
956         err = ipv6_netfilter_init();
957         if (err)
958                 goto netfilter_fail;
959         /* Create /proc/foo6 entries. */
960 #ifdef CONFIG_PROC_FS
961         err = -ENOMEM;
962         if (raw6_proc_init())
963                 goto proc_raw6_fail;
964         if (udplite6_proc_init())
965                 goto proc_udplite6_fail;
966         if (ipv6_misc_proc_init())
967                 goto proc_misc6_fail;
968
969         if (ac6_proc_init())
970                 goto proc_anycast6_fail;
971         if (if6_proc_init())
972                 goto proc_if6_fail;
973 #endif
974         err = ip6_route_init();
975         if (err)
976                 goto ip6_route_fail;
977         err = ip6_flowlabel_init();
978         if (err)
979                 goto ip6_flowlabel_fail;
980         err = addrconf_init();
981         if (err)
982                 goto addrconf_fail;
983
984         /* Init v6 extension headers. */
985         err = ipv6_exthdrs_init();
986         if (err)
987                 goto ipv6_exthdrs_fail;
988
989         err = ipv6_frag_init();
990         if (err)
991                 goto ipv6_frag_fail;
992
993         /* Init v6 transport protocols. */
994         err = udpv6_init();
995         if (err)
996                 goto udpv6_fail;
997
998         err = udplitev6_init();
999         if (err)
1000                 goto udplitev6_fail;
1001
1002         err = tcpv6_init();
1003         if (err)
1004                 goto tcpv6_fail;
1005
1006         err = ipv6_packet_init();
1007         if (err)
1008                 goto ipv6_packet_fail;
1009
1010 #ifdef CONFIG_SYSCTL
1011         err = ipv6_sysctl_register();
1012         if (err)
1013                 goto sysctl_fail;
1014 #endif
1015 out:
1016         return err;
1017
1018 #ifdef CONFIG_SYSCTL
1019 sysctl_fail:
1020         ipv6_packet_cleanup();
1021 #endif
1022 ipv6_packet_fail:
1023         tcpv6_exit();
1024 tcpv6_fail:
1025         udplitev6_exit();
1026 udplitev6_fail:
1027         udpv6_exit();
1028 udpv6_fail:
1029         ipv6_frag_exit();
1030 ipv6_frag_fail:
1031         ipv6_exthdrs_exit();
1032 ipv6_exthdrs_fail:
1033         addrconf_cleanup();
1034 addrconf_fail:
1035         ip6_flowlabel_cleanup();
1036 ip6_flowlabel_fail:
1037         ip6_route_cleanup();
1038 ip6_route_fail:
1039 #ifdef CONFIG_PROC_FS
1040         if6_proc_exit();
1041 proc_if6_fail:
1042         ac6_proc_exit();
1043 proc_anycast6_fail:
1044         ipv6_misc_proc_exit();
1045 proc_misc6_fail:
1046         udplite6_proc_exit();
1047 proc_udplite6_fail:
1048         raw6_proc_exit();
1049 proc_raw6_fail:
1050 #endif
1051         ipv6_netfilter_fini();
1052 netfilter_fail:
1053         igmp6_cleanup();
1054 igmp_fail:
1055         ndisc_cleanup();
1056 ndisc_fail:
1057         icmpv6_cleanup();
1058 icmp_fail:
1059         unregister_pernet_subsys(&inet6_net_ops);
1060 register_pernet_fail:
1061         cleanup_ipv6_mibs();
1062 out_unregister_sock:
1063         sock_unregister(PF_INET6);
1064         rtnl_unregister_all(PF_INET6);
1065 out_sock_register_fail:
1066         rawv6_exit();
1067 out_unregister_raw_proto:
1068         proto_unregister(&rawv6_prot);
1069 out_unregister_udplite_proto:
1070         proto_unregister(&udplitev6_prot);
1071 out_unregister_udp_proto:
1072         proto_unregister(&udpv6_prot);
1073 out_unregister_tcp_proto:
1074         proto_unregister(&tcpv6_prot);
1075         goto out;
1076 }
1077 module_init(inet6_init);
1078
1079 static void __exit inet6_exit(void)
1080 {
1081         /* First of all disallow new sockets creation. */
1082         sock_unregister(PF_INET6);
1083         /* Disallow any further netlink messages */
1084         rtnl_unregister_all(PF_INET6);
1085
1086 #ifdef CONFIG_SYSCTL
1087         ipv6_sysctl_unregister();
1088 #endif
1089         udpv6_exit();
1090         udplitev6_exit();
1091         tcpv6_exit();
1092
1093         /* Cleanup code parts. */
1094         ipv6_packet_cleanup();
1095         ipv6_frag_exit();
1096         ipv6_exthdrs_exit();
1097         addrconf_cleanup();
1098         ip6_flowlabel_cleanup();
1099         ip6_route_cleanup();
1100 #ifdef CONFIG_PROC_FS
1101
1102         /* Cleanup code parts. */
1103         if6_proc_exit();
1104         ac6_proc_exit();
1105         ipv6_misc_proc_exit();
1106         udplite6_proc_exit();
1107         raw6_proc_exit();
1108 #endif
1109         ipv6_netfilter_fini();
1110         igmp6_cleanup();
1111         ndisc_cleanup();
1112         icmpv6_cleanup();
1113         rawv6_exit();
1114
1115         unregister_pernet_subsys(&inet6_net_ops);
1116         cleanup_ipv6_mibs();
1117         proto_unregister(&rawv6_prot);
1118         proto_unregister(&udplitev6_prot);
1119         proto_unregister(&udpv6_prot);
1120         proto_unregister(&tcpv6_prot);
1121 }
1122 module_exit(inet6_exit);
1123
1124 MODULE_ALIAS_NETPROTO(PF_INET6);