Merge tag 'ext4_for_linus_stable' of git://git.kernel.org/pub/scm/linux/kernel/git...
[sfrench/cifs-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 #define pr_fmt(fmt) "IPv6: " fmt
22
23 #include <linux/module.h>
24 #include <linux/capability.h>
25 #include <linux/errno.h>
26 #include <linux/types.h>
27 #include <linux/socket.h>
28 #include <linux/in.h>
29 #include <linux/kernel.h>
30 #include <linux/timer.h>
31 #include <linux/string.h>
32 #include <linux/sockios.h>
33 #include <linux/net.h>
34 #include <linux/fcntl.h>
35 #include <linux/mm.h>
36 #include <linux/interrupt.h>
37 #include <linux/proc_fs.h>
38 #include <linux/stat.h>
39 #include <linux/init.h>
40 #include <linux/slab.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/ping.h>
53 #include <net/protocol.h>
54 #include <net/inet_common.h>
55 #include <net/route.h>
56 #include <net/transp_v6.h>
57 #include <net/ip6_route.h>
58 #include <net/addrconf.h>
59 #include <net/ndisc.h>
60 #ifdef CONFIG_IPV6_TUNNEL
61 #include <net/ip6_tunnel.h>
62 #endif
63 #include <net/calipso.h>
64 #include <net/seg6.h>
65
66 #include <linux/uaccess.h>
67 #include <linux/mroute6.h>
68
69 #include "ip6_offload.h"
70
71 MODULE_AUTHOR("Cast of dozens");
72 MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
73 MODULE_LICENSE("GPL");
74
75 /* The inetsw6 table contains everything that inet6_create needs to
76  * build a new socket.
77  */
78 static struct list_head inetsw6[SOCK_MAX];
79 static DEFINE_SPINLOCK(inetsw6_lock);
80
81 struct ipv6_params ipv6_defaults = {
82         .disable_ipv6 = 0,
83         .autoconf = 1,
84 };
85
86 static int disable_ipv6_mod;
87
88 module_param_named(disable, disable_ipv6_mod, int, 0444);
89 MODULE_PARM_DESC(disable, "Disable IPv6 module such that it is non-functional");
90
91 module_param_named(disable_ipv6, ipv6_defaults.disable_ipv6, int, 0444);
92 MODULE_PARM_DESC(disable_ipv6, "Disable IPv6 on all interfaces");
93
94 module_param_named(autoconf, ipv6_defaults.autoconf, int, 0444);
95 MODULE_PARM_DESC(autoconf, "Enable IPv6 address autoconfiguration on all interfaces");
96
97 bool ipv6_mod_enabled(void)
98 {
99         return disable_ipv6_mod == 0;
100 }
101 EXPORT_SYMBOL_GPL(ipv6_mod_enabled);
102
103 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
104 {
105         const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
106
107         return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
108 }
109
110 static int inet6_create(struct net *net, struct socket *sock, int protocol,
111                         int kern)
112 {
113         struct inet_sock *inet;
114         struct ipv6_pinfo *np;
115         struct sock *sk;
116         struct inet_protosw *answer;
117         struct proto *answer_prot;
118         unsigned char answer_flags;
119         int try_loading_module = 0;
120         int err;
121
122         if (protocol < 0 || protocol >= IPPROTO_MAX)
123                 return -EINVAL;
124
125         /* Look for the requested type/protocol pair. */
126 lookup_protocol:
127         err = -ESOCKTNOSUPPORT;
128         rcu_read_lock();
129         list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
130
131                 err = 0;
132                 /* Check the non-wild match. */
133                 if (protocol == answer->protocol) {
134                         if (protocol != IPPROTO_IP)
135                                 break;
136                 } else {
137                         /* Check for the two wild cases. */
138                         if (IPPROTO_IP == protocol) {
139                                 protocol = answer->protocol;
140                                 break;
141                         }
142                         if (IPPROTO_IP == answer->protocol)
143                                 break;
144                 }
145                 err = -EPROTONOSUPPORT;
146         }
147
148         if (err) {
149                 if (try_loading_module < 2) {
150                         rcu_read_unlock();
151                         /*
152                          * Be more specific, e.g. net-pf-10-proto-132-type-1
153                          * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
154                          */
155                         if (++try_loading_module == 1)
156                                 request_module("net-pf-%d-proto-%d-type-%d",
157                                                 PF_INET6, protocol, sock->type);
158                         /*
159                          * Fall back to generic, e.g. net-pf-10-proto-132
160                          * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
161                          */
162                         else
163                                 request_module("net-pf-%d-proto-%d",
164                                                 PF_INET6, protocol);
165                         goto lookup_protocol;
166                 } else
167                         goto out_rcu_unlock;
168         }
169
170         err = -EPERM;
171         if (sock->type == SOCK_RAW && !kern &&
172             !ns_capable(net->user_ns, CAP_NET_RAW))
173                 goto out_rcu_unlock;
174
175         sock->ops = answer->ops;
176         answer_prot = answer->prot;
177         answer_flags = answer->flags;
178         rcu_read_unlock();
179
180         WARN_ON(!answer_prot->slab);
181
182         err = -ENOBUFS;
183         sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot, kern);
184         if (!sk)
185                 goto out;
186
187         sock_init_data(sock, sk);
188
189         err = 0;
190         if (INET_PROTOSW_REUSE & answer_flags)
191                 sk->sk_reuse = SK_CAN_REUSE;
192
193         inet = inet_sk(sk);
194         inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
195
196         if (SOCK_RAW == sock->type) {
197                 inet->inet_num = protocol;
198                 if (IPPROTO_RAW == protocol)
199                         inet->hdrincl = 1;
200         }
201
202         sk->sk_destruct         = inet_sock_destruct;
203         sk->sk_family           = PF_INET6;
204         sk->sk_protocol         = protocol;
205
206         sk->sk_backlog_rcv      = answer->prot->backlog_rcv;
207
208         inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
209         np->hop_limit   = -1;
210         np->mcast_hops  = IPV6_DEFAULT_MCASTHOPS;
211         np->mc_loop     = 1;
212         np->mc_all      = 1;
213         np->pmtudisc    = IPV6_PMTUDISC_WANT;
214         np->repflow     = net->ipv6.sysctl.flowlabel_reflect;
215         sk->sk_ipv6only = net->ipv6.sysctl.bindv6only;
216
217         /* Init the ipv4 part of the socket since we can have sockets
218          * using v6 API for ipv4.
219          */
220         inet->uc_ttl    = -1;
221
222         inet->mc_loop   = 1;
223         inet->mc_ttl    = 1;
224         inet->mc_index  = 0;
225         inet->mc_list   = NULL;
226         inet->rcv_tos   = 0;
227
228         if (net->ipv4.sysctl_ip_no_pmtu_disc)
229                 inet->pmtudisc = IP_PMTUDISC_DONT;
230         else
231                 inet->pmtudisc = IP_PMTUDISC_WANT;
232         /*
233          * Increment only the relevant sk_prot->socks debug field, this changes
234          * the previous behaviour of incrementing both the equivalent to
235          * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
236          *
237          * This allows better debug granularity as we'll know exactly how many
238          * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
239          * transport protocol socks. -acme
240          */
241         sk_refcnt_debug_inc(sk);
242
243         if (inet->inet_num) {
244                 /* It assumes that any protocol which allows
245                  * the user to assign a number at socket
246                  * creation time automatically shares.
247                  */
248                 inet->inet_sport = htons(inet->inet_num);
249                 err = sk->sk_prot->hash(sk);
250                 if (err) {
251                         sk_common_release(sk);
252                         goto out;
253                 }
254         }
255         if (sk->sk_prot->init) {
256                 err = sk->sk_prot->init(sk);
257                 if (err) {
258                         sk_common_release(sk);
259                         goto out;
260                 }
261         }
262
263         if (!kern) {
264                 err = BPF_CGROUP_RUN_PROG_INET_SOCK(sk);
265                 if (err) {
266                         sk_common_release(sk);
267                         goto out;
268                 }
269         }
270 out:
271         return err;
272 out_rcu_unlock:
273         rcu_read_unlock();
274         goto out;
275 }
276
277 static int __inet6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len,
278                         bool force_bind_address_no_port, bool with_lock)
279 {
280         struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr;
281         struct inet_sock *inet = inet_sk(sk);
282         struct ipv6_pinfo *np = inet6_sk(sk);
283         struct net *net = sock_net(sk);
284         __be32 v4addr = 0;
285         unsigned short snum;
286         bool saved_ipv6only;
287         int addr_type = 0;
288         int err = 0;
289
290         if (addr->sin6_family != AF_INET6)
291                 return -EAFNOSUPPORT;
292
293         addr_type = ipv6_addr_type(&addr->sin6_addr);
294         if ((addr_type & IPV6_ADDR_MULTICAST) && sk->sk_type == SOCK_STREAM)
295                 return -EINVAL;
296
297         snum = ntohs(addr->sin6_port);
298         if (snum && snum < inet_prot_sock(net) &&
299             !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
300                 return -EACCES;
301
302         if (with_lock)
303                 lock_sock(sk);
304
305         /* Check these errors (active socket, double bind). */
306         if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
307                 err = -EINVAL;
308                 goto out;
309         }
310
311         /* Check if the address belongs to the host. */
312         if (addr_type == IPV6_ADDR_MAPPED) {
313                 int chk_addr_ret;
314
315                 /* Binding to v4-mapped address on a v6-only socket
316                  * makes no sense
317                  */
318                 if (sk->sk_ipv6only) {
319                         err = -EINVAL;
320                         goto out;
321                 }
322
323                 /* Reproduce AF_INET checks to make the bindings consistent */
324                 v4addr = addr->sin6_addr.s6_addr32[3];
325                 chk_addr_ret = inet_addr_type(net, v4addr);
326                 if (!inet_can_nonlocal_bind(net, inet) &&
327                     v4addr != htonl(INADDR_ANY) &&
328                     chk_addr_ret != RTN_LOCAL &&
329                     chk_addr_ret != RTN_MULTICAST &&
330                     chk_addr_ret != RTN_BROADCAST) {
331                         err = -EADDRNOTAVAIL;
332                         goto out;
333                 }
334         } else {
335                 if (addr_type != IPV6_ADDR_ANY) {
336                         struct net_device *dev = NULL;
337
338                         rcu_read_lock();
339                         if (__ipv6_addr_needs_scope_id(addr_type)) {
340                                 if (addr_len >= sizeof(struct sockaddr_in6) &&
341                                     addr->sin6_scope_id) {
342                                         /* Override any existing binding, if another one
343                                          * is supplied by user.
344                                          */
345                                         sk->sk_bound_dev_if = addr->sin6_scope_id;
346                                 }
347
348                                 /* Binding to link-local address requires an interface */
349                                 if (!sk->sk_bound_dev_if) {
350                                         err = -EINVAL;
351                                         goto out_unlock;
352                                 }
353                         }
354
355                         if (sk->sk_bound_dev_if) {
356                                 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
357                                 if (!dev) {
358                                         err = -ENODEV;
359                                         goto out_unlock;
360                                 }
361                         }
362
363                         /* ipv4 addr of the socket is invalid.  Only the
364                          * unspecified and mapped address have a v4 equivalent.
365                          */
366                         v4addr = LOOPBACK4_IPV6;
367                         if (!(addr_type & IPV6_ADDR_MULTICAST)) {
368                                 if (!ipv6_can_nonlocal_bind(net, inet) &&
369                                     !ipv6_chk_addr(net, &addr->sin6_addr,
370                                                    dev, 0)) {
371                                         err = -EADDRNOTAVAIL;
372                                         goto out_unlock;
373                                 }
374                         }
375                         rcu_read_unlock();
376                 }
377         }
378
379         inet->inet_rcv_saddr = v4addr;
380         inet->inet_saddr = v4addr;
381
382         sk->sk_v6_rcv_saddr = addr->sin6_addr;
383
384         if (!(addr_type & IPV6_ADDR_MULTICAST))
385                 np->saddr = addr->sin6_addr;
386
387         saved_ipv6only = sk->sk_ipv6only;
388         if (addr_type != IPV6_ADDR_ANY && addr_type != IPV6_ADDR_MAPPED)
389                 sk->sk_ipv6only = 1;
390
391         /* Make sure we are allowed to bind here. */
392         if (snum || !(inet->bind_address_no_port ||
393                       force_bind_address_no_port)) {
394                 if (sk->sk_prot->get_port(sk, snum)) {
395                         sk->sk_ipv6only = saved_ipv6only;
396                         inet_reset_saddr(sk);
397                         err = -EADDRINUSE;
398                         goto out;
399                 }
400                 err = BPF_CGROUP_RUN_PROG_INET6_POST_BIND(sk);
401                 if (err) {
402                         sk->sk_ipv6only = saved_ipv6only;
403                         inet_reset_saddr(sk);
404                         goto out;
405                 }
406         }
407
408         if (addr_type != IPV6_ADDR_ANY)
409                 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
410         if (snum)
411                 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
412         inet->inet_sport = htons(inet->inet_num);
413         inet->inet_dport = 0;
414         inet->inet_daddr = 0;
415 out:
416         if (with_lock)
417                 release_sock(sk);
418         return err;
419 out_unlock:
420         rcu_read_unlock();
421         goto out;
422 }
423
424 /* bind for INET6 API */
425 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
426 {
427         struct sock *sk = sock->sk;
428         int err = 0;
429
430         /* If the socket has its own bind function then use it. */
431         if (sk->sk_prot->bind)
432                 return sk->sk_prot->bind(sk, uaddr, addr_len);
433
434         if (addr_len < SIN6_LEN_RFC2133)
435                 return -EINVAL;
436
437         /* BPF prog is run before any checks are done so that if the prog
438          * changes context in a wrong way it will be caught.
439          */
440         err = BPF_CGROUP_RUN_PROG_INET6_BIND(sk, uaddr);
441         if (err)
442                 return err;
443
444         return __inet6_bind(sk, uaddr, addr_len, false, true);
445 }
446 EXPORT_SYMBOL(inet6_bind);
447
448 int inet6_release(struct socket *sock)
449 {
450         struct sock *sk = sock->sk;
451
452         if (!sk)
453                 return -EINVAL;
454
455         /* Free mc lists */
456         ipv6_sock_mc_close(sk);
457
458         /* Free ac lists */
459         ipv6_sock_ac_close(sk);
460
461         return inet_release(sock);
462 }
463 EXPORT_SYMBOL(inet6_release);
464
465 void inet6_destroy_sock(struct sock *sk)
466 {
467         struct ipv6_pinfo *np = inet6_sk(sk);
468         struct sk_buff *skb;
469         struct ipv6_txoptions *opt;
470
471         /* Release rx options */
472
473         skb = xchg(&np->pktoptions, NULL);
474         kfree_skb(skb);
475
476         skb = xchg(&np->rxpmtu, NULL);
477         kfree_skb(skb);
478
479         /* Free flowlabels */
480         fl6_free_socklist(sk);
481
482         /* Free tx options */
483
484         opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL);
485         if (opt) {
486                 atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
487                 txopt_put(opt);
488         }
489 }
490 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
491
492 /*
493  *      This does both peername and sockname.
494  */
495
496 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
497                  int peer)
498 {
499         struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
500         struct sock *sk = sock->sk;
501         struct inet_sock *inet = inet_sk(sk);
502         struct ipv6_pinfo *np = inet6_sk(sk);
503
504         sin->sin6_family = AF_INET6;
505         sin->sin6_flowinfo = 0;
506         sin->sin6_scope_id = 0;
507         if (peer) {
508                 if (!inet->inet_dport)
509                         return -ENOTCONN;
510                 if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
511                     peer == 1)
512                         return -ENOTCONN;
513                 sin->sin6_port = inet->inet_dport;
514                 sin->sin6_addr = sk->sk_v6_daddr;
515                 if (np->sndflow)
516                         sin->sin6_flowinfo = np->flow_label;
517         } else {
518                 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
519                         sin->sin6_addr = np->saddr;
520                 else
521                         sin->sin6_addr = sk->sk_v6_rcv_saddr;
522
523                 sin->sin6_port = inet->inet_sport;
524         }
525         sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
526                                                  sk->sk_bound_dev_if);
527         return sizeof(*sin);
528 }
529 EXPORT_SYMBOL(inet6_getname);
530
531 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
532 {
533         struct sock *sk = sock->sk;
534         struct net *net = sock_net(sk);
535
536         switch (cmd) {
537         case SIOCGSTAMP:
538                 return sock_get_timestamp(sk, (struct timeval __user *)arg);
539
540         case SIOCGSTAMPNS:
541                 return sock_get_timestampns(sk, (struct timespec __user *)arg);
542
543         case SIOCADDRT:
544         case SIOCDELRT:
545
546                 return ipv6_route_ioctl(net, cmd, (void __user *)arg);
547
548         case SIOCSIFADDR:
549                 return addrconf_add_ifaddr(net, (void __user *) arg);
550         case SIOCDIFADDR:
551                 return addrconf_del_ifaddr(net, (void __user *) arg);
552         case SIOCSIFDSTADDR:
553                 return addrconf_set_dstaddr(net, (void __user *) arg);
554         default:
555                 if (!sk->sk_prot->ioctl)
556                         return -ENOIOCTLCMD;
557                 return sk->sk_prot->ioctl(sk, cmd, arg);
558         }
559         /*NOTREACHED*/
560         return 0;
561 }
562 EXPORT_SYMBOL(inet6_ioctl);
563
564 const struct proto_ops inet6_stream_ops = {
565         .family            = PF_INET6,
566         .owner             = THIS_MODULE,
567         .release           = inet6_release,
568         .bind              = inet6_bind,
569         .connect           = inet_stream_connect,       /* ok           */
570         .socketpair        = sock_no_socketpair,        /* a do nothing */
571         .accept            = inet_accept,               /* ok           */
572         .getname           = inet6_getname,
573         .poll              = tcp_poll,                  /* ok           */
574         .ioctl             = inet6_ioctl,               /* must change  */
575         .listen            = inet_listen,               /* ok           */
576         .shutdown          = inet_shutdown,             /* ok           */
577         .setsockopt        = sock_common_setsockopt,    /* ok           */
578         .getsockopt        = sock_common_getsockopt,    /* ok           */
579         .sendmsg           = inet_sendmsg,              /* ok           */
580         .recvmsg           = inet_recvmsg,              /* ok           */
581 #ifdef CONFIG_MMU
582         .mmap              = tcp_mmap,
583 #endif
584         .sendpage          = inet_sendpage,
585         .sendmsg_locked    = tcp_sendmsg_locked,
586         .sendpage_locked   = tcp_sendpage_locked,
587         .splice_read       = tcp_splice_read,
588         .read_sock         = tcp_read_sock,
589         .peek_len          = tcp_peek_len,
590 #ifdef CONFIG_COMPAT
591         .compat_setsockopt = compat_sock_common_setsockopt,
592         .compat_getsockopt = compat_sock_common_getsockopt,
593 #endif
594         .set_rcvlowat      = tcp_set_rcvlowat,
595 };
596
597 const struct proto_ops inet6_dgram_ops = {
598         .family            = PF_INET6,
599         .owner             = THIS_MODULE,
600         .release           = inet6_release,
601         .bind              = inet6_bind,
602         .connect           = inet_dgram_connect,        /* ok           */
603         .socketpair        = sock_no_socketpair,        /* a do nothing */
604         .accept            = sock_no_accept,            /* a do nothing */
605         .getname           = inet6_getname,
606         .poll              = udp_poll,                  /* ok           */
607         .ioctl             = inet6_ioctl,               /* must change  */
608         .listen            = sock_no_listen,            /* ok           */
609         .shutdown          = inet_shutdown,             /* ok           */
610         .setsockopt        = sock_common_setsockopt,    /* ok           */
611         .getsockopt        = sock_common_getsockopt,    /* ok           */
612         .sendmsg           = inet_sendmsg,              /* ok           */
613         .recvmsg           = inet_recvmsg,              /* ok           */
614         .mmap              = sock_no_mmap,
615         .sendpage          = sock_no_sendpage,
616         .set_peek_off      = sk_set_peek_off,
617 #ifdef CONFIG_COMPAT
618         .compat_setsockopt = compat_sock_common_setsockopt,
619         .compat_getsockopt = compat_sock_common_getsockopt,
620 #endif
621 };
622
623 static const struct net_proto_family inet6_family_ops = {
624         .family = PF_INET6,
625         .create = inet6_create,
626         .owner  = THIS_MODULE,
627 };
628
629 int inet6_register_protosw(struct inet_protosw *p)
630 {
631         struct list_head *lh;
632         struct inet_protosw *answer;
633         struct list_head *last_perm;
634         int protocol = p->protocol;
635         int ret;
636
637         spin_lock_bh(&inetsw6_lock);
638
639         ret = -EINVAL;
640         if (p->type >= SOCK_MAX)
641                 goto out_illegal;
642
643         /* If we are trying to override a permanent protocol, bail. */
644         answer = NULL;
645         ret = -EPERM;
646         last_perm = &inetsw6[p->type];
647         list_for_each(lh, &inetsw6[p->type]) {
648                 answer = list_entry(lh, struct inet_protosw, list);
649
650                 /* Check only the non-wild match. */
651                 if (INET_PROTOSW_PERMANENT & answer->flags) {
652                         if (protocol == answer->protocol)
653                                 break;
654                         last_perm = lh;
655                 }
656
657                 answer = NULL;
658         }
659         if (answer)
660                 goto out_permanent;
661
662         /* Add the new entry after the last permanent entry if any, so that
663          * the new entry does not override a permanent entry when matched with
664          * a wild-card protocol. But it is allowed to override any existing
665          * non-permanent entry.  This means that when we remove this entry, the
666          * system automatically returns to the old behavior.
667          */
668         list_add_rcu(&p->list, last_perm);
669         ret = 0;
670 out:
671         spin_unlock_bh(&inetsw6_lock);
672         return ret;
673
674 out_permanent:
675         pr_err("Attempt to override permanent protocol %d\n", protocol);
676         goto out;
677
678 out_illegal:
679         pr_err("Ignoring attempt to register invalid socket type %d\n",
680                p->type);
681         goto out;
682 }
683 EXPORT_SYMBOL(inet6_register_protosw);
684
685 void
686 inet6_unregister_protosw(struct inet_protosw *p)
687 {
688         if (INET_PROTOSW_PERMANENT & p->flags) {
689                 pr_err("Attempt to unregister permanent protocol %d\n",
690                        p->protocol);
691         } else {
692                 spin_lock_bh(&inetsw6_lock);
693                 list_del_rcu(&p->list);
694                 spin_unlock_bh(&inetsw6_lock);
695
696                 synchronize_net();
697         }
698 }
699 EXPORT_SYMBOL(inet6_unregister_protosw);
700
701 int inet6_sk_rebuild_header(struct sock *sk)
702 {
703         struct ipv6_pinfo *np = inet6_sk(sk);
704         struct dst_entry *dst;
705
706         dst = __sk_dst_check(sk, np->dst_cookie);
707
708         if (!dst) {
709                 struct inet_sock *inet = inet_sk(sk);
710                 struct in6_addr *final_p, final;
711                 struct flowi6 fl6;
712
713                 memset(&fl6, 0, sizeof(fl6));
714                 fl6.flowi6_proto = sk->sk_protocol;
715                 fl6.daddr = sk->sk_v6_daddr;
716                 fl6.saddr = np->saddr;
717                 fl6.flowlabel = np->flow_label;
718                 fl6.flowi6_oif = sk->sk_bound_dev_if;
719                 fl6.flowi6_mark = sk->sk_mark;
720                 fl6.fl6_dport = inet->inet_dport;
721                 fl6.fl6_sport = inet->inet_sport;
722                 fl6.flowi6_uid = sk->sk_uid;
723                 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
724
725                 rcu_read_lock();
726                 final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt),
727                                          &final);
728                 rcu_read_unlock();
729
730                 dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
731                 if (IS_ERR(dst)) {
732                         sk->sk_route_caps = 0;
733                         sk->sk_err_soft = -PTR_ERR(dst);
734                         return PTR_ERR(dst);
735                 }
736
737                 ip6_dst_store(sk, dst, NULL, NULL);
738         }
739
740         return 0;
741 }
742 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
743
744 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb,
745                        const struct inet6_skb_parm *opt)
746 {
747         const struct ipv6_pinfo *np = inet6_sk(sk);
748
749         if (np->rxopt.all) {
750                 if (((opt->flags & IP6SKB_HOPBYHOP) &&
751                      (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) ||
752                     (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) &&
753                      np->rxopt.bits.rxflow) ||
754                     (opt->srcrt && (np->rxopt.bits.srcrt ||
755                      np->rxopt.bits.osrcrt)) ||
756                     ((opt->dst1 || opt->dst0) &&
757                      (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
758                         return true;
759         }
760         return false;
761 }
762 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
763
764 static struct packet_type ipv6_packet_type __read_mostly = {
765         .type = cpu_to_be16(ETH_P_IPV6),
766         .func = ipv6_rcv,
767         .list_func = ipv6_list_rcv,
768 };
769
770 static int __init ipv6_packet_init(void)
771 {
772         dev_add_pack(&ipv6_packet_type);
773         return 0;
774 }
775
776 static void ipv6_packet_cleanup(void)
777 {
778         dev_remove_pack(&ipv6_packet_type);
779 }
780
781 static int __net_init ipv6_init_mibs(struct net *net)
782 {
783         int i;
784
785         net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib);
786         if (!net->mib.udp_stats_in6)
787                 return -ENOMEM;
788         net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib);
789         if (!net->mib.udplite_stats_in6)
790                 goto err_udplite_mib;
791         net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib);
792         if (!net->mib.ipv6_statistics)
793                 goto err_ip_mib;
794
795         for_each_possible_cpu(i) {
796                 struct ipstats_mib *af_inet6_stats;
797                 af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i);
798                 u64_stats_init(&af_inet6_stats->syncp);
799         }
800
801
802         net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib);
803         if (!net->mib.icmpv6_statistics)
804                 goto err_icmp_mib;
805         net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
806                                                 GFP_KERNEL);
807         if (!net->mib.icmpv6msg_statistics)
808                 goto err_icmpmsg_mib;
809         return 0;
810
811 err_icmpmsg_mib:
812         free_percpu(net->mib.icmpv6_statistics);
813 err_icmp_mib:
814         free_percpu(net->mib.ipv6_statistics);
815 err_ip_mib:
816         free_percpu(net->mib.udplite_stats_in6);
817 err_udplite_mib:
818         free_percpu(net->mib.udp_stats_in6);
819         return -ENOMEM;
820 }
821
822 static void ipv6_cleanup_mibs(struct net *net)
823 {
824         free_percpu(net->mib.udp_stats_in6);
825         free_percpu(net->mib.udplite_stats_in6);
826         free_percpu(net->mib.ipv6_statistics);
827         free_percpu(net->mib.icmpv6_statistics);
828         kfree(net->mib.icmpv6msg_statistics);
829 }
830
831 static int __net_init inet6_net_init(struct net *net)
832 {
833         int err = 0;
834
835         net->ipv6.sysctl.bindv6only = 0;
836         net->ipv6.sysctl.icmpv6_time = 1*HZ;
837         net->ipv6.sysctl.icmpv6_echo_ignore_all = 0;
838         net->ipv6.sysctl.flowlabel_consistency = 1;
839         net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS;
840         net->ipv6.sysctl.idgen_retries = 3;
841         net->ipv6.sysctl.idgen_delay = 1 * HZ;
842         net->ipv6.sysctl.flowlabel_state_ranges = 0;
843         net->ipv6.sysctl.max_dst_opts_cnt = IP6_DEFAULT_MAX_DST_OPTS_CNT;
844         net->ipv6.sysctl.max_hbh_opts_cnt = IP6_DEFAULT_MAX_HBH_OPTS_CNT;
845         net->ipv6.sysctl.max_dst_opts_len = IP6_DEFAULT_MAX_DST_OPTS_LEN;
846         net->ipv6.sysctl.max_hbh_opts_len = IP6_DEFAULT_MAX_HBH_OPTS_LEN;
847         atomic_set(&net->ipv6.fib6_sernum, 1);
848
849         err = ipv6_init_mibs(net);
850         if (err)
851                 return err;
852 #ifdef CONFIG_PROC_FS
853         err = udp6_proc_init(net);
854         if (err)
855                 goto out;
856         err = tcp6_proc_init(net);
857         if (err)
858                 goto proc_tcp6_fail;
859         err = ac6_proc_init(net);
860         if (err)
861                 goto proc_ac6_fail;
862 #endif
863         return err;
864
865 #ifdef CONFIG_PROC_FS
866 proc_ac6_fail:
867         tcp6_proc_exit(net);
868 proc_tcp6_fail:
869         udp6_proc_exit(net);
870 out:
871         ipv6_cleanup_mibs(net);
872         return err;
873 #endif
874 }
875
876 static void __net_exit inet6_net_exit(struct net *net)
877 {
878 #ifdef CONFIG_PROC_FS
879         udp6_proc_exit(net);
880         tcp6_proc_exit(net);
881         ac6_proc_exit(net);
882 #endif
883         ipv6_cleanup_mibs(net);
884 }
885
886 static struct pernet_operations inet6_net_ops = {
887         .init = inet6_net_init,
888         .exit = inet6_net_exit,
889 };
890
891 static const struct ipv6_stub ipv6_stub_impl = {
892         .ipv6_sock_mc_join = ipv6_sock_mc_join,
893         .ipv6_sock_mc_drop = ipv6_sock_mc_drop,
894         .ipv6_dst_lookup   = ip6_dst_lookup,
895         .fib6_get_table    = fib6_get_table,
896         .fib6_table_lookup = fib6_table_lookup,
897         .fib6_lookup       = fib6_lookup,
898         .fib6_multipath_select = fib6_multipath_select,
899         .ip6_mtu_from_fib6 = ip6_mtu_from_fib6,
900         .udpv6_encap_enable = udpv6_encap_enable,
901         .ndisc_send_na = ndisc_send_na,
902         .nd_tbl = &nd_tbl,
903 };
904
905 static const struct ipv6_bpf_stub ipv6_bpf_stub_impl = {
906         .inet6_bind = __inet6_bind,
907         .udp6_lib_lookup = __udp6_lib_lookup,
908 };
909
910 static int __init inet6_init(void)
911 {
912         struct list_head *r;
913         int err = 0;
914
915         sock_skb_cb_check_size(sizeof(struct inet6_skb_parm));
916
917         /* Register the socket-side information for inet6_create.  */
918         for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
919                 INIT_LIST_HEAD(r);
920
921         if (disable_ipv6_mod) {
922                 pr_info("Loaded, but administratively disabled, reboot required to enable\n");
923                 goto out;
924         }
925
926         err = proto_register(&tcpv6_prot, 1);
927         if (err)
928                 goto out;
929
930         err = proto_register(&udpv6_prot, 1);
931         if (err)
932                 goto out_unregister_tcp_proto;
933
934         err = proto_register(&udplitev6_prot, 1);
935         if (err)
936                 goto out_unregister_udp_proto;
937
938         err = proto_register(&rawv6_prot, 1);
939         if (err)
940                 goto out_unregister_udplite_proto;
941
942         err = proto_register(&pingv6_prot, 1);
943         if (err)
944                 goto out_unregister_raw_proto;
945
946         /* We MUST register RAW sockets before we create the ICMP6,
947          * IGMP6, or NDISC control sockets.
948          */
949         err = rawv6_init();
950         if (err)
951                 goto out_unregister_ping_proto;
952
953         /* Register the family here so that the init calls below will
954          * be able to create sockets. (?? is this dangerous ??)
955          */
956         err = sock_register(&inet6_family_ops);
957         if (err)
958                 goto out_sock_register_fail;
959
960         /*
961          *      ipngwg API draft makes clear that the correct semantics
962          *      for TCP and UDP is to consider one TCP and UDP instance
963          *      in a host available by both INET and INET6 APIs and
964          *      able to communicate via both network protocols.
965          */
966
967         err = register_pernet_subsys(&inet6_net_ops);
968         if (err)
969                 goto register_pernet_fail;
970         err = ip6_mr_init();
971         if (err)
972                 goto ipmr_fail;
973         err = icmpv6_init();
974         if (err)
975                 goto icmp_fail;
976         err = ndisc_init();
977         if (err)
978                 goto ndisc_fail;
979         err = igmp6_init();
980         if (err)
981                 goto igmp_fail;
982
983         err = ipv6_netfilter_init();
984         if (err)
985                 goto netfilter_fail;
986         /* Create /proc/foo6 entries. */
987 #ifdef CONFIG_PROC_FS
988         err = -ENOMEM;
989         if (raw6_proc_init())
990                 goto proc_raw6_fail;
991         if (udplite6_proc_init())
992                 goto proc_udplite6_fail;
993         if (ipv6_misc_proc_init())
994                 goto proc_misc6_fail;
995         if (if6_proc_init())
996                 goto proc_if6_fail;
997 #endif
998         err = ip6_route_init();
999         if (err)
1000                 goto ip6_route_fail;
1001         err = ndisc_late_init();
1002         if (err)
1003                 goto ndisc_late_fail;
1004         err = ip6_flowlabel_init();
1005         if (err)
1006                 goto ip6_flowlabel_fail;
1007         err = ipv6_anycast_init();
1008         if (err)
1009                 goto ipv6_anycast_fail;
1010         err = addrconf_init();
1011         if (err)
1012                 goto addrconf_fail;
1013
1014         /* Init v6 extension headers. */
1015         err = ipv6_exthdrs_init();
1016         if (err)
1017                 goto ipv6_exthdrs_fail;
1018
1019         err = ipv6_frag_init();
1020         if (err)
1021                 goto ipv6_frag_fail;
1022
1023         /* Init v6 transport protocols. */
1024         err = udpv6_init();
1025         if (err)
1026                 goto udpv6_fail;
1027
1028         err = udplitev6_init();
1029         if (err)
1030                 goto udplitev6_fail;
1031
1032         err = udpv6_offload_init();
1033         if (err)
1034                 goto udpv6_offload_fail;
1035
1036         err = tcpv6_init();
1037         if (err)
1038                 goto tcpv6_fail;
1039
1040         err = ipv6_packet_init();
1041         if (err)
1042                 goto ipv6_packet_fail;
1043
1044         err = pingv6_init();
1045         if (err)
1046                 goto pingv6_fail;
1047
1048         err = calipso_init();
1049         if (err)
1050                 goto calipso_fail;
1051
1052         err = seg6_init();
1053         if (err)
1054                 goto seg6_fail;
1055
1056         err = igmp6_late_init();
1057         if (err)
1058                 goto igmp6_late_err;
1059
1060 #ifdef CONFIG_SYSCTL
1061         err = ipv6_sysctl_register();
1062         if (err)
1063                 goto sysctl_fail;
1064 #endif
1065
1066         /* ensure that ipv6 stubs are visible only after ipv6 is ready */
1067         wmb();
1068         ipv6_stub = &ipv6_stub_impl;
1069         ipv6_bpf_stub = &ipv6_bpf_stub_impl;
1070 out:
1071         return err;
1072
1073 #ifdef CONFIG_SYSCTL
1074 sysctl_fail:
1075         igmp6_late_cleanup();
1076 #endif
1077 igmp6_late_err:
1078         seg6_exit();
1079 seg6_fail:
1080         calipso_exit();
1081 calipso_fail:
1082         pingv6_exit();
1083 pingv6_fail:
1084         ipv6_packet_cleanup();
1085 ipv6_packet_fail:
1086         tcpv6_exit();
1087 tcpv6_fail:
1088         udpv6_offload_exit();
1089 udpv6_offload_fail:
1090         udplitev6_exit();
1091 udplitev6_fail:
1092         udpv6_exit();
1093 udpv6_fail:
1094         ipv6_frag_exit();
1095 ipv6_frag_fail:
1096         ipv6_exthdrs_exit();
1097 ipv6_exthdrs_fail:
1098         addrconf_cleanup();
1099 addrconf_fail:
1100         ipv6_anycast_cleanup();
1101 ipv6_anycast_fail:
1102         ip6_flowlabel_cleanup();
1103 ip6_flowlabel_fail:
1104         ndisc_late_cleanup();
1105 ndisc_late_fail:
1106         ip6_route_cleanup();
1107 ip6_route_fail:
1108 #ifdef CONFIG_PROC_FS
1109         if6_proc_exit();
1110 proc_if6_fail:
1111         ipv6_misc_proc_exit();
1112 proc_misc6_fail:
1113         udplite6_proc_exit();
1114 proc_udplite6_fail:
1115         raw6_proc_exit();
1116 proc_raw6_fail:
1117 #endif
1118         ipv6_netfilter_fini();
1119 netfilter_fail:
1120         igmp6_cleanup();
1121 igmp_fail:
1122         ndisc_cleanup();
1123 ndisc_fail:
1124         icmpv6_cleanup();
1125 icmp_fail:
1126         ip6_mr_cleanup();
1127 ipmr_fail:
1128         unregister_pernet_subsys(&inet6_net_ops);
1129 register_pernet_fail:
1130         sock_unregister(PF_INET6);
1131         rtnl_unregister_all(PF_INET6);
1132 out_sock_register_fail:
1133         rawv6_exit();
1134 out_unregister_ping_proto:
1135         proto_unregister(&pingv6_prot);
1136 out_unregister_raw_proto:
1137         proto_unregister(&rawv6_prot);
1138 out_unregister_udplite_proto:
1139         proto_unregister(&udplitev6_prot);
1140 out_unregister_udp_proto:
1141         proto_unregister(&udpv6_prot);
1142 out_unregister_tcp_proto:
1143         proto_unregister(&tcpv6_prot);
1144         goto out;
1145 }
1146 module_init(inet6_init);
1147
1148 MODULE_ALIAS_NETPROTO(PF_INET6);