]> asedeno.scripts.mit.edu Git - linux.git/blob - net/ipv6/af_inet6.c
2c694912df2e77b414de5cc2aa43e2ec59286836
[linux.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->pmtudisc    = IPV6_PMTUDISC_WANT;
213         np->repflow     = net->ipv6.sysctl.flowlabel_reflect;
214         sk->sk_ipv6only = net->ipv6.sysctl.bindv6only;
215
216         /* Init the ipv4 part of the socket since we can have sockets
217          * using v6 API for ipv4.
218          */
219         inet->uc_ttl    = -1;
220
221         inet->mc_loop   = 1;
222         inet->mc_ttl    = 1;
223         inet->mc_index  = 0;
224         inet->mc_list   = NULL;
225         inet->rcv_tos   = 0;
226
227         if (net->ipv4.sysctl_ip_no_pmtu_disc)
228                 inet->pmtudisc = IP_PMTUDISC_DONT;
229         else
230                 inet->pmtudisc = IP_PMTUDISC_WANT;
231         /*
232          * Increment only the relevant sk_prot->socks debug field, this changes
233          * the previous behaviour of incrementing both the equivalent to
234          * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
235          *
236          * This allows better debug granularity as we'll know exactly how many
237          * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
238          * transport protocol socks. -acme
239          */
240         sk_refcnt_debug_inc(sk);
241
242         if (inet->inet_num) {
243                 /* It assumes that any protocol which allows
244                  * the user to assign a number at socket
245                  * creation time automatically shares.
246                  */
247                 inet->inet_sport = htons(inet->inet_num);
248                 err = sk->sk_prot->hash(sk);
249                 if (err) {
250                         sk_common_release(sk);
251                         goto out;
252                 }
253         }
254         if (sk->sk_prot->init) {
255                 err = sk->sk_prot->init(sk);
256                 if (err) {
257                         sk_common_release(sk);
258                         goto out;
259                 }
260         }
261
262         if (!kern) {
263                 err = BPF_CGROUP_RUN_PROG_INET_SOCK(sk);
264                 if (err) {
265                         sk_common_release(sk);
266                         goto out;
267                 }
268         }
269 out:
270         return err;
271 out_rcu_unlock:
272         rcu_read_unlock();
273         goto out;
274 }
275
276
277 /* bind for INET6 API */
278 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
279 {
280         struct sock *sk = sock->sk;
281         int err = 0;
282
283         /* If the socket has its own bind function then use it. */
284         if (sk->sk_prot->bind)
285                 return sk->sk_prot->bind(sk, uaddr, addr_len);
286
287         if (addr_len < SIN6_LEN_RFC2133)
288                 return -EINVAL;
289
290         /* BPF prog is run before any checks are done so that if the prog
291          * changes context in a wrong way it will be caught.
292          */
293         err = BPF_CGROUP_RUN_PROG_INET6_BIND(sk, uaddr);
294         if (err)
295                 return err;
296
297         return __inet6_bind(sk, uaddr, addr_len, false, true);
298 }
299 EXPORT_SYMBOL(inet6_bind);
300
301 int __inet6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len,
302                  bool force_bind_address_no_port, bool with_lock)
303 {
304         struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr;
305         struct inet_sock *inet = inet_sk(sk);
306         struct ipv6_pinfo *np = inet6_sk(sk);
307         struct net *net = sock_net(sk);
308         __be32 v4addr = 0;
309         unsigned short snum;
310         bool saved_ipv6only;
311         int addr_type = 0;
312         int err = 0;
313
314         if (addr->sin6_family != AF_INET6)
315                 return -EAFNOSUPPORT;
316
317         addr_type = ipv6_addr_type(&addr->sin6_addr);
318         if ((addr_type & IPV6_ADDR_MULTICAST) && sk->sk_type == SOCK_STREAM)
319                 return -EINVAL;
320
321         snum = ntohs(addr->sin6_port);
322         if (snum && snum < inet_prot_sock(net) &&
323             !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
324                 return -EACCES;
325
326         if (with_lock)
327                 lock_sock(sk);
328
329         /* Check these errors (active socket, double bind). */
330         if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
331                 err = -EINVAL;
332                 goto out;
333         }
334
335         /* Check if the address belongs to the host. */
336         if (addr_type == IPV6_ADDR_MAPPED) {
337                 int chk_addr_ret;
338
339                 /* Binding to v4-mapped address on a v6-only socket
340                  * makes no sense
341                  */
342                 if (sk->sk_ipv6only) {
343                         err = -EINVAL;
344                         goto out;
345                 }
346
347                 /* Reproduce AF_INET checks to make the bindings consistent */
348                 v4addr = addr->sin6_addr.s6_addr32[3];
349                 chk_addr_ret = inet_addr_type(net, v4addr);
350                 if (!net->ipv4.sysctl_ip_nonlocal_bind &&
351                     !(inet->freebind || inet->transparent) &&
352                     v4addr != htonl(INADDR_ANY) &&
353                     chk_addr_ret != RTN_LOCAL &&
354                     chk_addr_ret != RTN_MULTICAST &&
355                     chk_addr_ret != RTN_BROADCAST) {
356                         err = -EADDRNOTAVAIL;
357                         goto out;
358                 }
359         } else {
360                 if (addr_type != IPV6_ADDR_ANY) {
361                         struct net_device *dev = NULL;
362
363                         rcu_read_lock();
364                         if (__ipv6_addr_needs_scope_id(addr_type)) {
365                                 if (addr_len >= sizeof(struct sockaddr_in6) &&
366                                     addr->sin6_scope_id) {
367                                         /* Override any existing binding, if another one
368                                          * is supplied by user.
369                                          */
370                                         sk->sk_bound_dev_if = addr->sin6_scope_id;
371                                 }
372
373                                 /* Binding to link-local address requires an interface */
374                                 if (!sk->sk_bound_dev_if) {
375                                         err = -EINVAL;
376                                         goto out_unlock;
377                                 }
378                                 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
379                                 if (!dev) {
380                                         err = -ENODEV;
381                                         goto out_unlock;
382                                 }
383                         }
384
385                         /* ipv4 addr of the socket is invalid.  Only the
386                          * unspecified and mapped address have a v4 equivalent.
387                          */
388                         v4addr = LOOPBACK4_IPV6;
389                         if (!(addr_type & IPV6_ADDR_MULTICAST)) {
390                                 if (!net->ipv6.sysctl.ip_nonlocal_bind &&
391                                     !(inet->freebind || inet->transparent) &&
392                                     !ipv6_chk_addr(net, &addr->sin6_addr,
393                                                    dev, 0)) {
394                                         err = -EADDRNOTAVAIL;
395                                         goto out_unlock;
396                                 }
397                         }
398                         rcu_read_unlock();
399                 }
400         }
401
402         inet->inet_rcv_saddr = v4addr;
403         inet->inet_saddr = v4addr;
404
405         sk->sk_v6_rcv_saddr = addr->sin6_addr;
406
407         if (!(addr_type & IPV6_ADDR_MULTICAST))
408                 np->saddr = addr->sin6_addr;
409
410         saved_ipv6only = sk->sk_ipv6only;
411         if (addr_type != IPV6_ADDR_ANY && addr_type != IPV6_ADDR_MAPPED)
412                 sk->sk_ipv6only = 1;
413
414         /* Make sure we are allowed to bind here. */
415         if (snum || !(inet->bind_address_no_port ||
416                       force_bind_address_no_port)) {
417                 if (sk->sk_prot->get_port(sk, snum)) {
418                         sk->sk_ipv6only = saved_ipv6only;
419                         inet_reset_saddr(sk);
420                         err = -EADDRINUSE;
421                         goto out;
422                 }
423                 err = BPF_CGROUP_RUN_PROG_INET6_POST_BIND(sk);
424                 if (err) {
425                         sk->sk_ipv6only = saved_ipv6only;
426                         inet_reset_saddr(sk);
427                         goto out;
428                 }
429         }
430
431         if (addr_type != IPV6_ADDR_ANY)
432                 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
433         if (snum)
434                 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
435         inet->inet_sport = htons(inet->inet_num);
436         inet->inet_dport = 0;
437         inet->inet_daddr = 0;
438 out:
439         if (with_lock)
440                 release_sock(sk);
441         return err;
442 out_unlock:
443         rcu_read_unlock();
444         goto out;
445 }
446
447 int inet6_release(struct socket *sock)
448 {
449         struct sock *sk = sock->sk;
450
451         if (!sk)
452                 return -EINVAL;
453
454         /* Free mc lists */
455         ipv6_sock_mc_close(sk);
456
457         /* Free ac lists */
458         ipv6_sock_ac_close(sk);
459
460         return inet_release(sock);
461 }
462 EXPORT_SYMBOL(inet6_release);
463
464 void inet6_destroy_sock(struct sock *sk)
465 {
466         struct ipv6_pinfo *np = inet6_sk(sk);
467         struct sk_buff *skb;
468         struct ipv6_txoptions *opt;
469
470         /* Release rx options */
471
472         skb = xchg(&np->pktoptions, NULL);
473         if (skb)
474                 kfree_skb(skb);
475
476         skb = xchg(&np->rxpmtu, NULL);
477         if (skb)
478                 kfree_skb(skb);
479
480         /* Free flowlabels */
481         fl6_free_socklist(sk);
482
483         /* Free tx options */
484
485         opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL);
486         if (opt) {
487                 atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
488                 txopt_put(opt);
489         }
490 }
491 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
492
493 /*
494  *      This does both peername and sockname.
495  */
496
497 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
498                  int peer)
499 {
500         struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
501         struct sock *sk = sock->sk;
502         struct inet_sock *inet = inet_sk(sk);
503         struct ipv6_pinfo *np = inet6_sk(sk);
504
505         sin->sin6_family = AF_INET6;
506         sin->sin6_flowinfo = 0;
507         sin->sin6_scope_id = 0;
508         if (peer) {
509                 if (!inet->inet_dport)
510                         return -ENOTCONN;
511                 if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
512                     peer == 1)
513                         return -ENOTCONN;
514                 sin->sin6_port = inet->inet_dport;
515                 sin->sin6_addr = sk->sk_v6_daddr;
516                 if (np->sndflow)
517                         sin->sin6_flowinfo = np->flow_label;
518         } else {
519                 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
520                         sin->sin6_addr = np->saddr;
521                 else
522                         sin->sin6_addr = sk->sk_v6_rcv_saddr;
523
524                 sin->sin6_port = inet->inet_sport;
525         }
526         sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
527                                                  sk->sk_bound_dev_if);
528         return sizeof(*sin);
529 }
530 EXPORT_SYMBOL(inet6_getname);
531
532 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
533 {
534         struct sock *sk = sock->sk;
535         struct net *net = sock_net(sk);
536
537         switch (cmd) {
538         case SIOCGSTAMP:
539                 return sock_get_timestamp(sk, (struct timeval __user *)arg);
540
541         case SIOCGSTAMPNS:
542                 return sock_get_timestampns(sk, (struct timespec __user *)arg);
543
544         case SIOCADDRT:
545         case SIOCDELRT:
546
547                 return ipv6_route_ioctl(net, cmd, (void __user *)arg);
548
549         case SIOCSIFADDR:
550                 return addrconf_add_ifaddr(net, (void __user *) arg);
551         case SIOCDIFADDR:
552                 return addrconf_del_ifaddr(net, (void __user *) arg);
553         case SIOCSIFDSTADDR:
554                 return addrconf_set_dstaddr(net, (void __user *) arg);
555         default:
556                 if (!sk->sk_prot->ioctl)
557                         return -ENOIOCTLCMD;
558                 return sk->sk_prot->ioctl(sk, cmd, arg);
559         }
560         /*NOTREACHED*/
561         return 0;
562 }
563 EXPORT_SYMBOL(inet6_ioctl);
564
565 const struct proto_ops inet6_stream_ops = {
566         .family            = PF_INET6,
567         .owner             = THIS_MODULE,
568         .release           = inet6_release,
569         .bind              = inet6_bind,
570         .connect           = inet_stream_connect,       /* ok           */
571         .socketpair        = sock_no_socketpair,        /* a do nothing */
572         .accept            = inet_accept,               /* ok           */
573         .getname           = inet6_getname,
574         .poll              = tcp_poll,                  /* ok           */
575         .ioctl             = inet6_ioctl,               /* must change  */
576         .listen            = inet_listen,               /* ok           */
577         .shutdown          = inet_shutdown,             /* ok           */
578         .setsockopt        = sock_common_setsockopt,    /* ok           */
579         .getsockopt        = sock_common_getsockopt,    /* ok           */
580         .sendmsg           = inet_sendmsg,              /* ok           */
581         .recvmsg           = inet_recvmsg,              /* ok           */
582         .mmap              = tcp_mmap,
583         .sendpage          = inet_sendpage,
584         .sendmsg_locked    = tcp_sendmsg_locked,
585         .sendpage_locked   = tcp_sendpage_locked,
586         .splice_read       = tcp_splice_read,
587         .read_sock         = tcp_read_sock,
588         .peek_len          = tcp_peek_len,
589 #ifdef CONFIG_COMPAT
590         .compat_setsockopt = compat_sock_common_setsockopt,
591         .compat_getsockopt = compat_sock_common_getsockopt,
592 #endif
593         .set_rcvlowat      = tcp_set_rcvlowat,
594 };
595
596 const struct proto_ops inet6_dgram_ops = {
597         .family            = PF_INET6,
598         .owner             = THIS_MODULE,
599         .release           = inet6_release,
600         .bind              = inet6_bind,
601         .connect           = inet_dgram_connect,        /* ok           */
602         .socketpair        = sock_no_socketpair,        /* a do nothing */
603         .accept            = sock_no_accept,            /* a do nothing */
604         .getname           = inet6_getname,
605         .poll              = udp_poll,                  /* ok           */
606         .ioctl             = inet6_ioctl,               /* must change  */
607         .listen            = sock_no_listen,            /* ok           */
608         .shutdown          = inet_shutdown,             /* ok           */
609         .setsockopt        = sock_common_setsockopt,    /* ok           */
610         .getsockopt        = sock_common_getsockopt,    /* ok           */
611         .sendmsg           = inet_sendmsg,              /* ok           */
612         .recvmsg           = inet_recvmsg,              /* ok           */
613         .mmap              = sock_no_mmap,
614         .sendpage          = sock_no_sendpage,
615         .set_peek_off      = sk_set_peek_off,
616 #ifdef CONFIG_COMPAT
617         .compat_setsockopt = compat_sock_common_setsockopt,
618         .compat_getsockopt = compat_sock_common_getsockopt,
619 #endif
620 };
621
622 static const struct net_proto_family inet6_family_ops = {
623         .family = PF_INET6,
624         .create = inet6_create,
625         .owner  = THIS_MODULE,
626 };
627
628 int inet6_register_protosw(struct inet_protosw *p)
629 {
630         struct list_head *lh;
631         struct inet_protosw *answer;
632         struct list_head *last_perm;
633         int protocol = p->protocol;
634         int ret;
635
636         spin_lock_bh(&inetsw6_lock);
637
638         ret = -EINVAL;
639         if (p->type >= SOCK_MAX)
640                 goto out_illegal;
641
642         /* If we are trying to override a permanent protocol, bail. */
643         answer = NULL;
644         ret = -EPERM;
645         last_perm = &inetsw6[p->type];
646         list_for_each(lh, &inetsw6[p->type]) {
647                 answer = list_entry(lh, struct inet_protosw, list);
648
649                 /* Check only the non-wild match. */
650                 if (INET_PROTOSW_PERMANENT & answer->flags) {
651                         if (protocol == answer->protocol)
652                                 break;
653                         last_perm = lh;
654                 }
655
656                 answer = NULL;
657         }
658         if (answer)
659                 goto out_permanent;
660
661         /* Add the new entry after the last permanent entry if any, so that
662          * the new entry does not override a permanent entry when matched with
663          * a wild-card protocol. But it is allowed to override any existing
664          * non-permanent entry.  This means that when we remove this entry, the
665          * system automatically returns to the old behavior.
666          */
667         list_add_rcu(&p->list, last_perm);
668         ret = 0;
669 out:
670         spin_unlock_bh(&inetsw6_lock);
671         return ret;
672
673 out_permanent:
674         pr_err("Attempt to override permanent protocol %d\n", protocol);
675         goto out;
676
677 out_illegal:
678         pr_err("Ignoring attempt to register invalid socket type %d\n",
679                p->type);
680         goto out;
681 }
682 EXPORT_SYMBOL(inet6_register_protosw);
683
684 void
685 inet6_unregister_protosw(struct inet_protosw *p)
686 {
687         if (INET_PROTOSW_PERMANENT & p->flags) {
688                 pr_err("Attempt to unregister permanent protocol %d\n",
689                        p->protocol);
690         } else {
691                 spin_lock_bh(&inetsw6_lock);
692                 list_del_rcu(&p->list);
693                 spin_unlock_bh(&inetsw6_lock);
694
695                 synchronize_net();
696         }
697 }
698 EXPORT_SYMBOL(inet6_unregister_protosw);
699
700 int inet6_sk_rebuild_header(struct sock *sk)
701 {
702         struct ipv6_pinfo *np = inet6_sk(sk);
703         struct dst_entry *dst;
704
705         dst = __sk_dst_check(sk, np->dst_cookie);
706
707         if (!dst) {
708                 struct inet_sock *inet = inet_sk(sk);
709                 struct in6_addr *final_p, final;
710                 struct flowi6 fl6;
711
712                 memset(&fl6, 0, sizeof(fl6));
713                 fl6.flowi6_proto = sk->sk_protocol;
714                 fl6.daddr = sk->sk_v6_daddr;
715                 fl6.saddr = np->saddr;
716                 fl6.flowlabel = np->flow_label;
717                 fl6.flowi6_oif = sk->sk_bound_dev_if;
718                 fl6.flowi6_mark = sk->sk_mark;
719                 fl6.fl6_dport = inet->inet_dport;
720                 fl6.fl6_sport = inet->inet_sport;
721                 fl6.flowi6_uid = sk->sk_uid;
722                 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
723
724                 rcu_read_lock();
725                 final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt),
726                                          &final);
727                 rcu_read_unlock();
728
729                 dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
730                 if (IS_ERR(dst)) {
731                         sk->sk_route_caps = 0;
732                         sk->sk_err_soft = -PTR_ERR(dst);
733                         return PTR_ERR(dst);
734                 }
735
736                 ip6_dst_store(sk, dst, NULL, NULL);
737         }
738
739         return 0;
740 }
741 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
742
743 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb,
744                        const struct inet6_skb_parm *opt)
745 {
746         const struct ipv6_pinfo *np = inet6_sk(sk);
747
748         if (np->rxopt.all) {
749                 if (((opt->flags & IP6SKB_HOPBYHOP) &&
750                      (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) ||
751                     (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) &&
752                      np->rxopt.bits.rxflow) ||
753                     (opt->srcrt && (np->rxopt.bits.srcrt ||
754                      np->rxopt.bits.osrcrt)) ||
755                     ((opt->dst1 || opt->dst0) &&
756                      (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
757                         return true;
758         }
759         return false;
760 }
761 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
762
763 static struct packet_type ipv6_packet_type __read_mostly = {
764         .type = cpu_to_be16(ETH_P_IPV6),
765         .func = ipv6_rcv,
766 };
767
768 static int __init ipv6_packet_init(void)
769 {
770         dev_add_pack(&ipv6_packet_type);
771         return 0;
772 }
773
774 static void ipv6_packet_cleanup(void)
775 {
776         dev_remove_pack(&ipv6_packet_type);
777 }
778
779 static int __net_init ipv6_init_mibs(struct net *net)
780 {
781         int i;
782
783         net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib);
784         if (!net->mib.udp_stats_in6)
785                 return -ENOMEM;
786         net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib);
787         if (!net->mib.udplite_stats_in6)
788                 goto err_udplite_mib;
789         net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib);
790         if (!net->mib.ipv6_statistics)
791                 goto err_ip_mib;
792
793         for_each_possible_cpu(i) {
794                 struct ipstats_mib *af_inet6_stats;
795                 af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i);
796                 u64_stats_init(&af_inet6_stats->syncp);
797         }
798
799
800         net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib);
801         if (!net->mib.icmpv6_statistics)
802                 goto err_icmp_mib;
803         net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
804                                                 GFP_KERNEL);
805         if (!net->mib.icmpv6msg_statistics)
806                 goto err_icmpmsg_mib;
807         return 0;
808
809 err_icmpmsg_mib:
810         free_percpu(net->mib.icmpv6_statistics);
811 err_icmp_mib:
812         free_percpu(net->mib.ipv6_statistics);
813 err_ip_mib:
814         free_percpu(net->mib.udplite_stats_in6);
815 err_udplite_mib:
816         free_percpu(net->mib.udp_stats_in6);
817         return -ENOMEM;
818 }
819
820 static void ipv6_cleanup_mibs(struct net *net)
821 {
822         free_percpu(net->mib.udp_stats_in6);
823         free_percpu(net->mib.udplite_stats_in6);
824         free_percpu(net->mib.ipv6_statistics);
825         free_percpu(net->mib.icmpv6_statistics);
826         kfree(net->mib.icmpv6msg_statistics);
827 }
828
829 static int __net_init inet6_net_init(struct net *net)
830 {
831         int err = 0;
832
833         net->ipv6.sysctl.bindv6only = 0;
834         net->ipv6.sysctl.icmpv6_time = 1*HZ;
835         net->ipv6.sysctl.flowlabel_consistency = 1;
836         net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS;
837         net->ipv6.sysctl.idgen_retries = 3;
838         net->ipv6.sysctl.idgen_delay = 1 * HZ;
839         net->ipv6.sysctl.flowlabel_state_ranges = 0;
840         net->ipv6.sysctl.max_dst_opts_cnt = IP6_DEFAULT_MAX_DST_OPTS_CNT;
841         net->ipv6.sysctl.max_hbh_opts_cnt = IP6_DEFAULT_MAX_HBH_OPTS_CNT;
842         net->ipv6.sysctl.max_dst_opts_len = IP6_DEFAULT_MAX_DST_OPTS_LEN;
843         net->ipv6.sysctl.max_hbh_opts_len = IP6_DEFAULT_MAX_HBH_OPTS_LEN;
844         atomic_set(&net->ipv6.fib6_sernum, 1);
845
846         err = ipv6_init_mibs(net);
847         if (err)
848                 return err;
849 #ifdef CONFIG_PROC_FS
850         err = udp6_proc_init(net);
851         if (err)
852                 goto out;
853         err = tcp6_proc_init(net);
854         if (err)
855                 goto proc_tcp6_fail;
856         err = ac6_proc_init(net);
857         if (err)
858                 goto proc_ac6_fail;
859 #endif
860         return err;
861
862 #ifdef CONFIG_PROC_FS
863 proc_ac6_fail:
864         tcp6_proc_exit(net);
865 proc_tcp6_fail:
866         udp6_proc_exit(net);
867 out:
868         ipv6_cleanup_mibs(net);
869         return err;
870 #endif
871 }
872
873 static void __net_exit inet6_net_exit(struct net *net)
874 {
875 #ifdef CONFIG_PROC_FS
876         udp6_proc_exit(net);
877         tcp6_proc_exit(net);
878         ac6_proc_exit(net);
879 #endif
880         ipv6_cleanup_mibs(net);
881 }
882
883 static struct pernet_operations inet6_net_ops = {
884         .init = inet6_net_init,
885         .exit = inet6_net_exit,
886 };
887
888 static const struct ipv6_stub ipv6_stub_impl = {
889         .ipv6_sock_mc_join = ipv6_sock_mc_join,
890         .ipv6_sock_mc_drop = ipv6_sock_mc_drop,
891         .ipv6_dst_lookup = ip6_dst_lookup,
892         .udpv6_encap_enable = udpv6_encap_enable,
893         .ndisc_send_na = ndisc_send_na,
894         .nd_tbl = &nd_tbl,
895 };
896
897 static const struct ipv6_bpf_stub ipv6_bpf_stub_impl = {
898         .inet6_bind = __inet6_bind,
899 };
900
901 static int __init inet6_init(void)
902 {
903         struct list_head *r;
904         int err = 0;
905
906         sock_skb_cb_check_size(sizeof(struct inet6_skb_parm));
907
908         /* Register the socket-side information for inet6_create.  */
909         for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
910                 INIT_LIST_HEAD(r);
911
912         if (disable_ipv6_mod) {
913                 pr_info("Loaded, but administratively disabled, reboot required to enable\n");
914                 goto out;
915         }
916
917         err = proto_register(&tcpv6_prot, 1);
918         if (err)
919                 goto out;
920
921         err = proto_register(&udpv6_prot, 1);
922         if (err)
923                 goto out_unregister_tcp_proto;
924
925         err = proto_register(&udplitev6_prot, 1);
926         if (err)
927                 goto out_unregister_udp_proto;
928
929         err = proto_register(&rawv6_prot, 1);
930         if (err)
931                 goto out_unregister_udplite_proto;
932
933         err = proto_register(&pingv6_prot, 1);
934         if (err)
935                 goto out_unregister_ping_proto;
936
937         /* We MUST register RAW sockets before we create the ICMP6,
938          * IGMP6, or NDISC control sockets.
939          */
940         err = rawv6_init();
941         if (err)
942                 goto out_unregister_raw_proto;
943
944         /* Register the family here so that the init calls below will
945          * be able to create sockets. (?? is this dangerous ??)
946          */
947         err = sock_register(&inet6_family_ops);
948         if (err)
949                 goto out_sock_register_fail;
950
951         /*
952          *      ipngwg API draft makes clear that the correct semantics
953          *      for TCP and UDP is to consider one TCP and UDP instance
954          *      in a host available by both INET and INET6 APIs and
955          *      able to communicate via both network protocols.
956          */
957
958         err = register_pernet_subsys(&inet6_net_ops);
959         if (err)
960                 goto register_pernet_fail;
961         err = ip6_mr_init();
962         if (err)
963                 goto ipmr_fail;
964         err = icmpv6_init();
965         if (err)
966                 goto icmp_fail;
967         err = ndisc_init();
968         if (err)
969                 goto ndisc_fail;
970         err = igmp6_init();
971         if (err)
972                 goto igmp_fail;
973
974         err = ipv6_netfilter_init();
975         if (err)
976                 goto netfilter_fail;
977         /* Create /proc/foo6 entries. */
978 #ifdef CONFIG_PROC_FS
979         err = -ENOMEM;
980         if (raw6_proc_init())
981                 goto proc_raw6_fail;
982         if (udplite6_proc_init())
983                 goto proc_udplite6_fail;
984         if (ipv6_misc_proc_init())
985                 goto proc_misc6_fail;
986         if (if6_proc_init())
987                 goto proc_if6_fail;
988 #endif
989         err = ip6_route_init();
990         if (err)
991                 goto ip6_route_fail;
992         err = ndisc_late_init();
993         if (err)
994                 goto ndisc_late_fail;
995         err = ip6_flowlabel_init();
996         if (err)
997                 goto ip6_flowlabel_fail;
998         err = addrconf_init();
999         if (err)
1000                 goto addrconf_fail;
1001
1002         /* Init v6 extension headers. */
1003         err = ipv6_exthdrs_init();
1004         if (err)
1005                 goto ipv6_exthdrs_fail;
1006
1007         err = ipv6_frag_init();
1008         if (err)
1009                 goto ipv6_frag_fail;
1010
1011         /* Init v6 transport protocols. */
1012         err = udpv6_init();
1013         if (err)
1014                 goto udpv6_fail;
1015
1016         err = udplitev6_init();
1017         if (err)
1018                 goto udplitev6_fail;
1019
1020         err = udpv6_offload_init();
1021         if (err)
1022                 goto udpv6_offload_fail;
1023
1024         err = tcpv6_init();
1025         if (err)
1026                 goto tcpv6_fail;
1027
1028         err = ipv6_packet_init();
1029         if (err)
1030                 goto ipv6_packet_fail;
1031
1032         err = pingv6_init();
1033         if (err)
1034                 goto pingv6_fail;
1035
1036         err = calipso_init();
1037         if (err)
1038                 goto calipso_fail;
1039
1040         err = seg6_init();
1041         if (err)
1042                 goto seg6_fail;
1043
1044         err = igmp6_late_init();
1045         if (err)
1046                 goto igmp6_late_err;
1047
1048 #ifdef CONFIG_SYSCTL
1049         err = ipv6_sysctl_register();
1050         if (err)
1051                 goto sysctl_fail;
1052 #endif
1053
1054         /* ensure that ipv6 stubs are visible only after ipv6 is ready */
1055         wmb();
1056         ipv6_stub = &ipv6_stub_impl;
1057         ipv6_bpf_stub = &ipv6_bpf_stub_impl;
1058 out:
1059         return err;
1060
1061 #ifdef CONFIG_SYSCTL
1062 sysctl_fail:
1063         igmp6_late_cleanup();
1064 #endif
1065 igmp6_late_err:
1066         seg6_exit();
1067 seg6_fail:
1068         calipso_exit();
1069 calipso_fail:
1070         pingv6_exit();
1071 pingv6_fail:
1072         ipv6_packet_cleanup();
1073 ipv6_packet_fail:
1074         tcpv6_exit();
1075 tcpv6_fail:
1076         udpv6_offload_exit();
1077 udpv6_offload_fail:
1078         udplitev6_exit();
1079 udplitev6_fail:
1080         udpv6_exit();
1081 udpv6_fail:
1082         ipv6_frag_exit();
1083 ipv6_frag_fail:
1084         ipv6_exthdrs_exit();
1085 ipv6_exthdrs_fail:
1086         addrconf_cleanup();
1087 addrconf_fail:
1088         ip6_flowlabel_cleanup();
1089 ip6_flowlabel_fail:
1090         ndisc_late_cleanup();
1091 ndisc_late_fail:
1092         ip6_route_cleanup();
1093 ip6_route_fail:
1094 #ifdef CONFIG_PROC_FS
1095         if6_proc_exit();
1096 proc_if6_fail:
1097         ipv6_misc_proc_exit();
1098 proc_misc6_fail:
1099         udplite6_proc_exit();
1100 proc_udplite6_fail:
1101         raw6_proc_exit();
1102 proc_raw6_fail:
1103 #endif
1104         ipv6_netfilter_fini();
1105 netfilter_fail:
1106         igmp6_cleanup();
1107 igmp_fail:
1108         ndisc_cleanup();
1109 ndisc_fail:
1110         ip6_mr_cleanup();
1111 icmp_fail:
1112         unregister_pernet_subsys(&inet6_net_ops);
1113 ipmr_fail:
1114         icmpv6_cleanup();
1115 register_pernet_fail:
1116         sock_unregister(PF_INET6);
1117         rtnl_unregister_all(PF_INET6);
1118 out_sock_register_fail:
1119         rawv6_exit();
1120 out_unregister_ping_proto:
1121         proto_unregister(&pingv6_prot);
1122 out_unregister_raw_proto:
1123         proto_unregister(&rawv6_prot);
1124 out_unregister_udplite_proto:
1125         proto_unregister(&udplitev6_prot);
1126 out_unregister_udp_proto:
1127         proto_unregister(&udpv6_prot);
1128 out_unregister_tcp_proto:
1129         proto_unregister(&tcpv6_prot);
1130         goto out;
1131 }
1132 module_init(inet6_init);
1133
1134 MODULE_ALIAS_NETPROTO(PF_INET6);