Loading...
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
64#include <asm/uaccess.h>
65#include <linux/mroute6.h>
66
67MODULE_AUTHOR("Cast of dozens");
68MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
69MODULE_LICENSE("GPL");
70
71/* The inetsw6 table contains everything that inet6_create needs to
72 * build a new socket.
73 */
74static struct list_head inetsw6[SOCK_MAX];
75static DEFINE_SPINLOCK(inetsw6_lock);
76
77struct ipv6_params ipv6_defaults = {
78 .disable_ipv6 = 0,
79 .autoconf = 1,
80};
81
82static int disable_ipv6_mod;
83
84module_param_named(disable, disable_ipv6_mod, int, 0444);
85MODULE_PARM_DESC(disable, "Disable IPv6 module such that it is non-functional");
86
87module_param_named(disable_ipv6, ipv6_defaults.disable_ipv6, int, 0444);
88MODULE_PARM_DESC(disable_ipv6, "Disable IPv6 on all interfaces");
89
90module_param_named(autoconf, ipv6_defaults.autoconf, int, 0444);
91MODULE_PARM_DESC(autoconf, "Enable IPv6 address autoconfiguration on all interfaces");
92
93static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
94{
95 const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
96
97 return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
98}
99
100static int inet6_create(struct net *net, struct socket *sock, int protocol,
101 int kern)
102{
103 struct inet_sock *inet;
104 struct ipv6_pinfo *np;
105 struct sock *sk;
106 struct inet_protosw *answer;
107 struct proto *answer_prot;
108 unsigned char answer_flags;
109 char answer_no_check;
110 int try_loading_module = 0;
111 int err;
112
113 /* Look for the requested type/protocol pair. */
114lookup_protocol:
115 err = -ESOCKTNOSUPPORT;
116 rcu_read_lock();
117 list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
118
119 err = 0;
120 /* Check the non-wild match. */
121 if (protocol == answer->protocol) {
122 if (protocol != IPPROTO_IP)
123 break;
124 } else {
125 /* Check for the two wild cases. */
126 if (IPPROTO_IP == protocol) {
127 protocol = answer->protocol;
128 break;
129 }
130 if (IPPROTO_IP == answer->protocol)
131 break;
132 }
133 err = -EPROTONOSUPPORT;
134 }
135
136 if (err) {
137 if (try_loading_module < 2) {
138 rcu_read_unlock();
139 /*
140 * Be more specific, e.g. net-pf-10-proto-132-type-1
141 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
142 */
143 if (++try_loading_module == 1)
144 request_module("net-pf-%d-proto-%d-type-%d",
145 PF_INET6, protocol, sock->type);
146 /*
147 * Fall back to generic, e.g. net-pf-10-proto-132
148 * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
149 */
150 else
151 request_module("net-pf-%d-proto-%d",
152 PF_INET6, protocol);
153 goto lookup_protocol;
154 } else
155 goto out_rcu_unlock;
156 }
157
158 err = -EPERM;
159 if (sock->type == SOCK_RAW && !kern &&
160 !ns_capable(net->user_ns, CAP_NET_RAW))
161 goto out_rcu_unlock;
162
163 sock->ops = answer->ops;
164 answer_prot = answer->prot;
165 answer_no_check = answer->no_check;
166 answer_flags = answer->flags;
167 rcu_read_unlock();
168
169 WARN_ON(answer_prot->slab == NULL);
170
171 err = -ENOBUFS;
172 sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot);
173 if (sk == NULL)
174 goto out;
175
176 sock_init_data(sock, sk);
177
178 err = 0;
179 sk->sk_no_check = answer_no_check;
180 if (INET_PROTOSW_REUSE & answer_flags)
181 sk->sk_reuse = SK_CAN_REUSE;
182
183 inet = inet_sk(sk);
184 inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
185
186 if (SOCK_RAW == sock->type) {
187 inet->inet_num = protocol;
188 if (IPPROTO_RAW == protocol)
189 inet->hdrincl = 1;
190 }
191
192 sk->sk_destruct = inet_sock_destruct;
193 sk->sk_family = PF_INET6;
194 sk->sk_protocol = protocol;
195
196 sk->sk_backlog_rcv = answer->prot->backlog_rcv;
197
198 inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
199 np->hop_limit = -1;
200 np->mcast_hops = IPV6_DEFAULT_MCASTHOPS;
201 np->mc_loop = 1;
202 np->pmtudisc = IPV6_PMTUDISC_WANT;
203 np->ipv6only = net->ipv6.sysctl.bindv6only;
204
205 /* Init the ipv4 part of the socket since we can have sockets
206 * using v6 API for ipv4.
207 */
208 inet->uc_ttl = -1;
209
210 inet->mc_loop = 1;
211 inet->mc_ttl = 1;
212 inet->mc_index = 0;
213 inet->mc_list = NULL;
214 inet->rcv_tos = 0;
215
216 if (net->ipv4.sysctl_ip_no_pmtu_disc)
217 inet->pmtudisc = IP_PMTUDISC_DONT;
218 else
219 inet->pmtudisc = IP_PMTUDISC_WANT;
220 /*
221 * Increment only the relevant sk_prot->socks debug field, this changes
222 * the previous behaviour of incrementing both the equivalent to
223 * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
224 *
225 * This allows better debug granularity as we'll know exactly how many
226 * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
227 * transport protocol socks. -acme
228 */
229 sk_refcnt_debug_inc(sk);
230
231 if (inet->inet_num) {
232 /* It assumes that any protocol which allows
233 * the user to assign a number at socket
234 * creation time automatically shares.
235 */
236 inet->inet_sport = htons(inet->inet_num);
237 sk->sk_prot->hash(sk);
238 }
239 if (sk->sk_prot->init) {
240 err = sk->sk_prot->init(sk);
241 if (err) {
242 sk_common_release(sk);
243 goto out;
244 }
245 }
246out:
247 return err;
248out_rcu_unlock:
249 rcu_read_unlock();
250 goto out;
251}
252
253
254/* bind for INET6 API */
255int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
256{
257 struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr;
258 struct sock *sk = sock->sk;
259 struct inet_sock *inet = inet_sk(sk);
260 struct ipv6_pinfo *np = inet6_sk(sk);
261 struct net *net = sock_net(sk);
262 __be32 v4addr = 0;
263 unsigned short snum;
264 int addr_type = 0;
265 int err = 0;
266
267 /* If the socket has its own bind function then use it. */
268 if (sk->sk_prot->bind)
269 return sk->sk_prot->bind(sk, uaddr, addr_len);
270
271 if (addr_len < SIN6_LEN_RFC2133)
272 return -EINVAL;
273
274 if (addr->sin6_family != AF_INET6)
275 return -EAFNOSUPPORT;
276
277 addr_type = ipv6_addr_type(&addr->sin6_addr);
278 if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
279 return -EINVAL;
280
281 snum = ntohs(addr->sin6_port);
282 if (snum && snum < PROT_SOCK && !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
283 return -EACCES;
284
285 lock_sock(sk);
286
287 /* Check these errors (active socket, double bind). */
288 if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
289 err = -EINVAL;
290 goto out;
291 }
292
293 /* Check if the address belongs to the host. */
294 if (addr_type == IPV6_ADDR_MAPPED) {
295 int chk_addr_ret;
296
297 /* Binding to v4-mapped address on a v6-only socket
298 * makes no sense
299 */
300 if (np->ipv6only) {
301 err = -EINVAL;
302 goto out;
303 }
304
305 /* Reproduce AF_INET checks to make the bindings consistent */
306 v4addr = addr->sin6_addr.s6_addr32[3];
307 chk_addr_ret = inet_addr_type(net, v4addr);
308 if (!sysctl_ip_nonlocal_bind &&
309 !(inet->freebind || inet->transparent) &&
310 v4addr != htonl(INADDR_ANY) &&
311 chk_addr_ret != RTN_LOCAL &&
312 chk_addr_ret != RTN_MULTICAST &&
313 chk_addr_ret != RTN_BROADCAST) {
314 err = -EADDRNOTAVAIL;
315 goto out;
316 }
317 } else {
318 if (addr_type != IPV6_ADDR_ANY) {
319 struct net_device *dev = NULL;
320
321 rcu_read_lock();
322 if (__ipv6_addr_needs_scope_id(addr_type)) {
323 if (addr_len >= sizeof(struct sockaddr_in6) &&
324 addr->sin6_scope_id) {
325 /* Override any existing binding, if another one
326 * is supplied by user.
327 */
328 sk->sk_bound_dev_if = addr->sin6_scope_id;
329 }
330
331 /* Binding to link-local address requires an interface */
332 if (!sk->sk_bound_dev_if) {
333 err = -EINVAL;
334 goto out_unlock;
335 }
336 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
337 if (!dev) {
338 err = -ENODEV;
339 goto out_unlock;
340 }
341 }
342
343 /* ipv4 addr of the socket is invalid. Only the
344 * unspecified and mapped address have a v4 equivalent.
345 */
346 v4addr = LOOPBACK4_IPV6;
347 if (!(addr_type & IPV6_ADDR_MULTICAST)) {
348 if (!(inet->freebind || inet->transparent) &&
349 !ipv6_chk_addr(net, &addr->sin6_addr,
350 dev, 0)) {
351 err = -EADDRNOTAVAIL;
352 goto out_unlock;
353 }
354 }
355 rcu_read_unlock();
356 }
357 }
358
359 inet->inet_rcv_saddr = v4addr;
360 inet->inet_saddr = v4addr;
361
362 sk->sk_v6_rcv_saddr = addr->sin6_addr;
363
364 if (!(addr_type & IPV6_ADDR_MULTICAST))
365 np->saddr = addr->sin6_addr;
366
367 /* Make sure we are allowed to bind here. */
368 if (sk->sk_prot->get_port(sk, snum)) {
369 inet_reset_saddr(sk);
370 err = -EADDRINUSE;
371 goto out;
372 }
373
374 if (addr_type != IPV6_ADDR_ANY) {
375 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
376 if (addr_type != IPV6_ADDR_MAPPED)
377 np->ipv6only = 1;
378 }
379 if (snum)
380 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
381 inet->inet_sport = htons(inet->inet_num);
382 inet->inet_dport = 0;
383 inet->inet_daddr = 0;
384out:
385 release_sock(sk);
386 return err;
387out_unlock:
388 rcu_read_unlock();
389 goto out;
390}
391EXPORT_SYMBOL(inet6_bind);
392
393int inet6_release(struct socket *sock)
394{
395 struct sock *sk = sock->sk;
396
397 if (sk == NULL)
398 return -EINVAL;
399
400 /* Free mc lists */
401 ipv6_sock_mc_close(sk);
402
403 /* Free ac lists */
404 ipv6_sock_ac_close(sk);
405
406 return inet_release(sock);
407}
408EXPORT_SYMBOL(inet6_release);
409
410void inet6_destroy_sock(struct sock *sk)
411{
412 struct ipv6_pinfo *np = inet6_sk(sk);
413 struct sk_buff *skb;
414 struct ipv6_txoptions *opt;
415
416 /* Release rx options */
417
418 skb = xchg(&np->pktoptions, NULL);
419 if (skb != NULL)
420 kfree_skb(skb);
421
422 skb = xchg(&np->rxpmtu, NULL);
423 if (skb != NULL)
424 kfree_skb(skb);
425
426 /* Free flowlabels */
427 fl6_free_socklist(sk);
428
429 /* Free tx options */
430
431 opt = xchg(&np->opt, NULL);
432 if (opt != NULL)
433 sock_kfree_s(sk, opt, opt->tot_len);
434}
435EXPORT_SYMBOL_GPL(inet6_destroy_sock);
436
437/*
438 * This does both peername and sockname.
439 */
440
441int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
442 int *uaddr_len, int peer)
443{
444 struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
445 struct sock *sk = sock->sk;
446 struct inet_sock *inet = inet_sk(sk);
447 struct ipv6_pinfo *np = inet6_sk(sk);
448
449 sin->sin6_family = AF_INET6;
450 sin->sin6_flowinfo = 0;
451 sin->sin6_scope_id = 0;
452 if (peer) {
453 if (!inet->inet_dport)
454 return -ENOTCONN;
455 if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
456 peer == 1)
457 return -ENOTCONN;
458 sin->sin6_port = inet->inet_dport;
459 sin->sin6_addr = sk->sk_v6_daddr;
460 if (np->sndflow)
461 sin->sin6_flowinfo = np->flow_label;
462 } else {
463 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
464 sin->sin6_addr = np->saddr;
465 else
466 sin->sin6_addr = sk->sk_v6_rcv_saddr;
467
468 sin->sin6_port = inet->inet_sport;
469 }
470 sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
471 sk->sk_bound_dev_if);
472 *uaddr_len = sizeof(*sin);
473 return 0;
474}
475EXPORT_SYMBOL(inet6_getname);
476
477int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
478{
479 struct sock *sk = sock->sk;
480 struct net *net = sock_net(sk);
481
482 switch (cmd) {
483 case SIOCGSTAMP:
484 return sock_get_timestamp(sk, (struct timeval __user *)arg);
485
486 case SIOCGSTAMPNS:
487 return sock_get_timestampns(sk, (struct timespec __user *)arg);
488
489 case SIOCADDRT:
490 case SIOCDELRT:
491
492 return ipv6_route_ioctl(net, cmd, (void __user *)arg);
493
494 case SIOCSIFADDR:
495 return addrconf_add_ifaddr(net, (void __user *) arg);
496 case SIOCDIFADDR:
497 return addrconf_del_ifaddr(net, (void __user *) arg);
498 case SIOCSIFDSTADDR:
499 return addrconf_set_dstaddr(net, (void __user *) arg);
500 default:
501 if (!sk->sk_prot->ioctl)
502 return -ENOIOCTLCMD;
503 return sk->sk_prot->ioctl(sk, cmd, arg);
504 }
505 /*NOTREACHED*/
506 return 0;
507}
508EXPORT_SYMBOL(inet6_ioctl);
509
510const struct proto_ops inet6_stream_ops = {
511 .family = PF_INET6,
512 .owner = THIS_MODULE,
513 .release = inet6_release,
514 .bind = inet6_bind,
515 .connect = inet_stream_connect, /* ok */
516 .socketpair = sock_no_socketpair, /* a do nothing */
517 .accept = inet_accept, /* ok */
518 .getname = inet6_getname,
519 .poll = tcp_poll, /* ok */
520 .ioctl = inet6_ioctl, /* must change */
521 .listen = inet_listen, /* ok */
522 .shutdown = inet_shutdown, /* ok */
523 .setsockopt = sock_common_setsockopt, /* ok */
524 .getsockopt = sock_common_getsockopt, /* ok */
525 .sendmsg = inet_sendmsg, /* ok */
526 .recvmsg = inet_recvmsg, /* ok */
527 .mmap = sock_no_mmap,
528 .sendpage = inet_sendpage,
529 .splice_read = tcp_splice_read,
530#ifdef CONFIG_COMPAT
531 .compat_setsockopt = compat_sock_common_setsockopt,
532 .compat_getsockopt = compat_sock_common_getsockopt,
533#endif
534};
535
536const struct proto_ops inet6_dgram_ops = {
537 .family = PF_INET6,
538 .owner = THIS_MODULE,
539 .release = inet6_release,
540 .bind = inet6_bind,
541 .connect = inet_dgram_connect, /* ok */
542 .socketpair = sock_no_socketpair, /* a do nothing */
543 .accept = sock_no_accept, /* a do nothing */
544 .getname = inet6_getname,
545 .poll = udp_poll, /* ok */
546 .ioctl = inet6_ioctl, /* must change */
547 .listen = sock_no_listen, /* ok */
548 .shutdown = inet_shutdown, /* ok */
549 .setsockopt = sock_common_setsockopt, /* ok */
550 .getsockopt = sock_common_getsockopt, /* ok */
551 .sendmsg = inet_sendmsg, /* ok */
552 .recvmsg = inet_recvmsg, /* ok */
553 .mmap = sock_no_mmap,
554 .sendpage = sock_no_sendpage,
555#ifdef CONFIG_COMPAT
556 .compat_setsockopt = compat_sock_common_setsockopt,
557 .compat_getsockopt = compat_sock_common_getsockopt,
558#endif
559};
560
561static const struct net_proto_family inet6_family_ops = {
562 .family = PF_INET6,
563 .create = inet6_create,
564 .owner = THIS_MODULE,
565};
566
567int inet6_register_protosw(struct inet_protosw *p)
568{
569 struct list_head *lh;
570 struct inet_protosw *answer;
571 struct list_head *last_perm;
572 int protocol = p->protocol;
573 int ret;
574
575 spin_lock_bh(&inetsw6_lock);
576
577 ret = -EINVAL;
578 if (p->type >= SOCK_MAX)
579 goto out_illegal;
580
581 /* If we are trying to override a permanent protocol, bail. */
582 answer = NULL;
583 ret = -EPERM;
584 last_perm = &inetsw6[p->type];
585 list_for_each(lh, &inetsw6[p->type]) {
586 answer = list_entry(lh, struct inet_protosw, list);
587
588 /* Check only the non-wild match. */
589 if (INET_PROTOSW_PERMANENT & answer->flags) {
590 if (protocol == answer->protocol)
591 break;
592 last_perm = lh;
593 }
594
595 answer = NULL;
596 }
597 if (answer)
598 goto out_permanent;
599
600 /* Add the new entry after the last permanent entry if any, so that
601 * the new entry does not override a permanent entry when matched with
602 * a wild-card protocol. But it is allowed to override any existing
603 * non-permanent entry. This means that when we remove this entry, the
604 * system automatically returns to the old behavior.
605 */
606 list_add_rcu(&p->list, last_perm);
607 ret = 0;
608out:
609 spin_unlock_bh(&inetsw6_lock);
610 return ret;
611
612out_permanent:
613 pr_err("Attempt to override permanent protocol %d\n", protocol);
614 goto out;
615
616out_illegal:
617 pr_err("Ignoring attempt to register invalid socket type %d\n",
618 p->type);
619 goto out;
620}
621EXPORT_SYMBOL(inet6_register_protosw);
622
623void
624inet6_unregister_protosw(struct inet_protosw *p)
625{
626 if (INET_PROTOSW_PERMANENT & p->flags) {
627 pr_err("Attempt to unregister permanent protocol %d\n",
628 p->protocol);
629 } else {
630 spin_lock_bh(&inetsw6_lock);
631 list_del_rcu(&p->list);
632 spin_unlock_bh(&inetsw6_lock);
633
634 synchronize_net();
635 }
636}
637EXPORT_SYMBOL(inet6_unregister_protosw);
638
639int inet6_sk_rebuild_header(struct sock *sk)
640{
641 struct ipv6_pinfo *np = inet6_sk(sk);
642 struct dst_entry *dst;
643
644 dst = __sk_dst_check(sk, np->dst_cookie);
645
646 if (dst == NULL) {
647 struct inet_sock *inet = inet_sk(sk);
648 struct in6_addr *final_p, final;
649 struct flowi6 fl6;
650
651 memset(&fl6, 0, sizeof(fl6));
652 fl6.flowi6_proto = sk->sk_protocol;
653 fl6.daddr = sk->sk_v6_daddr;
654 fl6.saddr = np->saddr;
655 fl6.flowlabel = np->flow_label;
656 fl6.flowi6_oif = sk->sk_bound_dev_if;
657 fl6.flowi6_mark = sk->sk_mark;
658 fl6.fl6_dport = inet->inet_dport;
659 fl6.fl6_sport = inet->inet_sport;
660 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
661
662 final_p = fl6_update_dst(&fl6, np->opt, &final);
663
664 dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
665 if (IS_ERR(dst)) {
666 sk->sk_route_caps = 0;
667 sk->sk_err_soft = -PTR_ERR(dst);
668 return PTR_ERR(dst);
669 }
670
671 __ip6_dst_store(sk, dst, NULL, NULL);
672 }
673
674 return 0;
675}
676EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
677
678bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb)
679{
680 const struct ipv6_pinfo *np = inet6_sk(sk);
681 const struct inet6_skb_parm *opt = IP6CB(skb);
682
683 if (np->rxopt.all) {
684 if ((opt->hop && (np->rxopt.bits.hopopts ||
685 np->rxopt.bits.ohopopts)) ||
686 (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) &&
687 np->rxopt.bits.rxflow) ||
688 (opt->srcrt && (np->rxopt.bits.srcrt ||
689 np->rxopt.bits.osrcrt)) ||
690 ((opt->dst1 || opt->dst0) &&
691 (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
692 return true;
693 }
694 return false;
695}
696EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
697
698static struct packet_type ipv6_packet_type __read_mostly = {
699 .type = cpu_to_be16(ETH_P_IPV6),
700 .func = ipv6_rcv,
701};
702
703static int __init ipv6_packet_init(void)
704{
705 dev_add_pack(&ipv6_packet_type);
706 return 0;
707}
708
709static void ipv6_packet_cleanup(void)
710{
711 dev_remove_pack(&ipv6_packet_type);
712}
713
714static int __net_init ipv6_init_mibs(struct net *net)
715{
716 int i;
717
718 if (snmp_mib_init((void __percpu **)net->mib.udp_stats_in6,
719 sizeof(struct udp_mib),
720 __alignof__(struct udp_mib)) < 0)
721 return -ENOMEM;
722 if (snmp_mib_init((void __percpu **)net->mib.udplite_stats_in6,
723 sizeof(struct udp_mib),
724 __alignof__(struct udp_mib)) < 0)
725 goto err_udplite_mib;
726 if (snmp_mib_init((void __percpu **)net->mib.ipv6_statistics,
727 sizeof(struct ipstats_mib),
728 __alignof__(struct ipstats_mib)) < 0)
729 goto err_ip_mib;
730
731 for_each_possible_cpu(i) {
732 struct ipstats_mib *af_inet6_stats;
733 af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics[0], i);
734 u64_stats_init(&af_inet6_stats->syncp);
735#if SNMP_ARRAY_SZ == 2
736 af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics[1], i);
737 u64_stats_init(&af_inet6_stats->syncp);
738#endif
739 }
740
741
742 if (snmp_mib_init((void __percpu **)net->mib.icmpv6_statistics,
743 sizeof(struct icmpv6_mib),
744 __alignof__(struct icmpv6_mib)) < 0)
745 goto err_icmp_mib;
746 net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
747 GFP_KERNEL);
748 if (!net->mib.icmpv6msg_statistics)
749 goto err_icmpmsg_mib;
750 return 0;
751
752err_icmpmsg_mib:
753 snmp_mib_free((void __percpu **)net->mib.icmpv6_statistics);
754err_icmp_mib:
755 snmp_mib_free((void __percpu **)net->mib.ipv6_statistics);
756err_ip_mib:
757 snmp_mib_free((void __percpu **)net->mib.udplite_stats_in6);
758err_udplite_mib:
759 snmp_mib_free((void __percpu **)net->mib.udp_stats_in6);
760 return -ENOMEM;
761}
762
763static void ipv6_cleanup_mibs(struct net *net)
764{
765 snmp_mib_free((void __percpu **)net->mib.udp_stats_in6);
766 snmp_mib_free((void __percpu **)net->mib.udplite_stats_in6);
767 snmp_mib_free((void __percpu **)net->mib.ipv6_statistics);
768 snmp_mib_free((void __percpu **)net->mib.icmpv6_statistics);
769 kfree(net->mib.icmpv6msg_statistics);
770}
771
772static int __net_init inet6_net_init(struct net *net)
773{
774 int err = 0;
775
776 net->ipv6.sysctl.bindv6only = 0;
777 net->ipv6.sysctl.icmpv6_time = 1*HZ;
778 net->ipv6.sysctl.flowlabel_consistency = 1;
779 atomic_set(&net->ipv6.rt_genid, 0);
780
781 err = ipv6_init_mibs(net);
782 if (err)
783 return err;
784#ifdef CONFIG_PROC_FS
785 err = udp6_proc_init(net);
786 if (err)
787 goto out;
788 err = tcp6_proc_init(net);
789 if (err)
790 goto proc_tcp6_fail;
791 err = ac6_proc_init(net);
792 if (err)
793 goto proc_ac6_fail;
794#endif
795 return err;
796
797#ifdef CONFIG_PROC_FS
798proc_ac6_fail:
799 tcp6_proc_exit(net);
800proc_tcp6_fail:
801 udp6_proc_exit(net);
802out:
803 ipv6_cleanup_mibs(net);
804 return err;
805#endif
806}
807
808static void __net_exit inet6_net_exit(struct net *net)
809{
810#ifdef CONFIG_PROC_FS
811 udp6_proc_exit(net);
812 tcp6_proc_exit(net);
813 ac6_proc_exit(net);
814#endif
815 ipv6_cleanup_mibs(net);
816}
817
818static struct pernet_operations inet6_net_ops = {
819 .init = inet6_net_init,
820 .exit = inet6_net_exit,
821};
822
823static const struct ipv6_stub ipv6_stub_impl = {
824 .ipv6_sock_mc_join = ipv6_sock_mc_join,
825 .ipv6_sock_mc_drop = ipv6_sock_mc_drop,
826 .ipv6_dst_lookup = ip6_dst_lookup,
827 .udpv6_encap_enable = udpv6_encap_enable,
828 .ndisc_send_na = ndisc_send_na,
829 .nd_tbl = &nd_tbl,
830};
831
832static int __init inet6_init(void)
833{
834 struct list_head *r;
835 int err = 0;
836
837 BUILD_BUG_ON(sizeof(struct inet6_skb_parm) > FIELD_SIZEOF(struct sk_buff, cb));
838
839 /* Register the socket-side information for inet6_create. */
840 for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
841 INIT_LIST_HEAD(r);
842
843 if (disable_ipv6_mod) {
844 pr_info("Loaded, but administratively disabled, reboot required to enable\n");
845 goto out;
846 }
847
848 err = proto_register(&tcpv6_prot, 1);
849 if (err)
850 goto out;
851
852 err = proto_register(&udpv6_prot, 1);
853 if (err)
854 goto out_unregister_tcp_proto;
855
856 err = proto_register(&udplitev6_prot, 1);
857 if (err)
858 goto out_unregister_udp_proto;
859
860 err = proto_register(&rawv6_prot, 1);
861 if (err)
862 goto out_unregister_udplite_proto;
863
864 err = proto_register(&pingv6_prot, 1);
865 if (err)
866 goto out_unregister_ping_proto;
867
868 /* We MUST register RAW sockets before we create the ICMP6,
869 * IGMP6, or NDISC control sockets.
870 */
871 err = rawv6_init();
872 if (err)
873 goto out_unregister_raw_proto;
874
875 /* Register the family here so that the init calls below will
876 * be able to create sockets. (?? is this dangerous ??)
877 */
878 err = sock_register(&inet6_family_ops);
879 if (err)
880 goto out_sock_register_fail;
881
882 /*
883 * ipngwg API draft makes clear that the correct semantics
884 * for TCP and UDP is to consider one TCP and UDP instance
885 * in a host available by both INET and INET6 APIs and
886 * able to communicate via both network protocols.
887 */
888
889 err = register_pernet_subsys(&inet6_net_ops);
890 if (err)
891 goto register_pernet_fail;
892 err = icmpv6_init();
893 if (err)
894 goto icmp_fail;
895 err = ip6_mr_init();
896 if (err)
897 goto ipmr_fail;
898 err = ndisc_init();
899 if (err)
900 goto ndisc_fail;
901 err = igmp6_init();
902 if (err)
903 goto igmp_fail;
904
905 ipv6_stub = &ipv6_stub_impl;
906
907 err = ipv6_netfilter_init();
908 if (err)
909 goto netfilter_fail;
910 /* Create /proc/foo6 entries. */
911#ifdef CONFIG_PROC_FS
912 err = -ENOMEM;
913 if (raw6_proc_init())
914 goto proc_raw6_fail;
915 if (udplite6_proc_init())
916 goto proc_udplite6_fail;
917 if (ipv6_misc_proc_init())
918 goto proc_misc6_fail;
919 if (if6_proc_init())
920 goto proc_if6_fail;
921#endif
922 err = ip6_route_init();
923 if (err)
924 goto ip6_route_fail;
925 err = ndisc_late_init();
926 if (err)
927 goto ndisc_late_fail;
928 err = ip6_flowlabel_init();
929 if (err)
930 goto ip6_flowlabel_fail;
931 err = addrconf_init();
932 if (err)
933 goto addrconf_fail;
934
935 /* Init v6 extension headers. */
936 err = ipv6_exthdrs_init();
937 if (err)
938 goto ipv6_exthdrs_fail;
939
940 err = ipv6_frag_init();
941 if (err)
942 goto ipv6_frag_fail;
943
944 /* Init v6 transport protocols. */
945 err = udpv6_init();
946 if (err)
947 goto udpv6_fail;
948
949 err = udplitev6_init();
950 if (err)
951 goto udplitev6_fail;
952
953 err = tcpv6_init();
954 if (err)
955 goto tcpv6_fail;
956
957 err = ipv6_packet_init();
958 if (err)
959 goto ipv6_packet_fail;
960
961 err = pingv6_init();
962 if (err)
963 goto pingv6_fail;
964
965#ifdef CONFIG_SYSCTL
966 err = ipv6_sysctl_register();
967 if (err)
968 goto sysctl_fail;
969#endif
970out:
971 return err;
972
973#ifdef CONFIG_SYSCTL
974sysctl_fail:
975 pingv6_exit();
976#endif
977pingv6_fail:
978 ipv6_packet_cleanup();
979ipv6_packet_fail:
980 tcpv6_exit();
981tcpv6_fail:
982 udplitev6_exit();
983udplitev6_fail:
984 udpv6_exit();
985udpv6_fail:
986 ipv6_frag_exit();
987ipv6_frag_fail:
988 ipv6_exthdrs_exit();
989ipv6_exthdrs_fail:
990 addrconf_cleanup();
991addrconf_fail:
992 ip6_flowlabel_cleanup();
993ip6_flowlabel_fail:
994 ndisc_late_cleanup();
995ndisc_late_fail:
996 ip6_route_cleanup();
997ip6_route_fail:
998#ifdef CONFIG_PROC_FS
999 if6_proc_exit();
1000proc_if6_fail:
1001 ipv6_misc_proc_exit();
1002proc_misc6_fail:
1003 udplite6_proc_exit();
1004proc_udplite6_fail:
1005 raw6_proc_exit();
1006proc_raw6_fail:
1007#endif
1008 ipv6_netfilter_fini();
1009netfilter_fail:
1010 igmp6_cleanup();
1011igmp_fail:
1012 ndisc_cleanup();
1013ndisc_fail:
1014 ip6_mr_cleanup();
1015ipmr_fail:
1016 icmpv6_cleanup();
1017icmp_fail:
1018 unregister_pernet_subsys(&inet6_net_ops);
1019register_pernet_fail:
1020 sock_unregister(PF_INET6);
1021 rtnl_unregister_all(PF_INET6);
1022out_sock_register_fail:
1023 rawv6_exit();
1024out_unregister_ping_proto:
1025 proto_unregister(&pingv6_prot);
1026out_unregister_raw_proto:
1027 proto_unregister(&rawv6_prot);
1028out_unregister_udplite_proto:
1029 proto_unregister(&udplitev6_prot);
1030out_unregister_udp_proto:
1031 proto_unregister(&udpv6_prot);
1032out_unregister_tcp_proto:
1033 proto_unregister(&tcpv6_prot);
1034 goto out;
1035}
1036module_init(inet6_init);
1037
1038MODULE_ALIAS_NETPROTO(PF_INET6);
1// SPDX-License-Identifier: GPL-2.0-or-later
2/*
3 * PF_INET6 socket protocol family
4 * Linux INET6 implementation
5 *
6 * Authors:
7 * Pedro Roque <roque@di.fc.ul.pt>
8 *
9 * Adapted from linux/net/ipv4/af_inet.c
10 *
11 * Fixes:
12 * piggy, Karl Knutson : Socket protocol table
13 * Hideaki YOSHIFUJI : sin6_scope_id support
14 * Arnaldo Melo : check proc_net_create return, cleanups
15 */
16
17#define pr_fmt(fmt) "IPv6: " fmt
18
19#include <linux/module.h>
20#include <linux/capability.h>
21#include <linux/errno.h>
22#include <linux/types.h>
23#include <linux/socket.h>
24#include <linux/in.h>
25#include <linux/kernel.h>
26#include <linux/timer.h>
27#include <linux/string.h>
28#include <linux/sockios.h>
29#include <linux/net.h>
30#include <linux/fcntl.h>
31#include <linux/mm.h>
32#include <linux/interrupt.h>
33#include <linux/proc_fs.h>
34#include <linux/stat.h>
35#include <linux/init.h>
36#include <linux/slab.h>
37
38#include <linux/inet.h>
39#include <linux/netdevice.h>
40#include <linux/icmpv6.h>
41#include <linux/netfilter_ipv6.h>
42
43#include <net/ip.h>
44#include <net/ipv6.h>
45#include <net/udp.h>
46#include <net/udplite.h>
47#include <net/tcp.h>
48#include <net/ping.h>
49#include <net/protocol.h>
50#include <net/inet_common.h>
51#include <net/route.h>
52#include <net/transp_v6.h>
53#include <net/ip6_route.h>
54#include <net/addrconf.h>
55#include <net/ipv6_stubs.h>
56#include <net/ndisc.h>
57#ifdef CONFIG_IPV6_TUNNEL
58#include <net/ip6_tunnel.h>
59#endif
60#include <net/calipso.h>
61#include <net/seg6.h>
62#include <net/rpl.h>
63#include <net/compat.h>
64#include <net/xfrm.h>
65#include <net/ioam6.h>
66#include <net/rawv6.h>
67#include <net/rps.h>
68
69#include <linux/uaccess.h>
70#include <linux/mroute6.h>
71
72#include "ip6_offload.h"
73
74MODULE_AUTHOR("Cast of dozens");
75MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
76MODULE_LICENSE("GPL");
77
78/* The inetsw6 table contains everything that inet6_create needs to
79 * build a new socket.
80 */
81static struct list_head inetsw6[SOCK_MAX];
82static DEFINE_SPINLOCK(inetsw6_lock);
83
84struct ipv6_params ipv6_defaults = {
85 .disable_ipv6 = 0,
86 .autoconf = 1,
87};
88
89static int disable_ipv6_mod;
90
91module_param_named(disable, disable_ipv6_mod, int, 0444);
92MODULE_PARM_DESC(disable, "Disable IPv6 module such that it is non-functional");
93
94module_param_named(disable_ipv6, ipv6_defaults.disable_ipv6, int, 0444);
95MODULE_PARM_DESC(disable_ipv6, "Disable IPv6 on all interfaces");
96
97module_param_named(autoconf, ipv6_defaults.autoconf, int, 0444);
98MODULE_PARM_DESC(autoconf, "Enable IPv6 address autoconfiguration on all interfaces");
99
100bool ipv6_mod_enabled(void)
101{
102 return disable_ipv6_mod == 0;
103}
104EXPORT_SYMBOL_GPL(ipv6_mod_enabled);
105
106static struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
107{
108 const int offset = sk->sk_prot->ipv6_pinfo_offset;
109
110 return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
111}
112
113void inet6_sock_destruct(struct sock *sk)
114{
115 inet6_cleanup_sock(sk);
116 inet_sock_destruct(sk);
117}
118EXPORT_SYMBOL_GPL(inet6_sock_destruct);
119
120static int inet6_create(struct net *net, struct socket *sock, int protocol,
121 int kern)
122{
123 struct inet_sock *inet;
124 struct ipv6_pinfo *np;
125 struct sock *sk;
126 struct inet_protosw *answer;
127 struct proto *answer_prot;
128 unsigned char answer_flags;
129 int try_loading_module = 0;
130 int err;
131
132 if (protocol < 0 || protocol >= IPPROTO_MAX)
133 return -EINVAL;
134
135 /* Look for the requested type/protocol pair. */
136lookup_protocol:
137 err = -ESOCKTNOSUPPORT;
138 rcu_read_lock();
139 list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
140
141 err = 0;
142 /* Check the non-wild match. */
143 if (protocol == answer->protocol) {
144 if (protocol != IPPROTO_IP)
145 break;
146 } else {
147 /* Check for the two wild cases. */
148 if (IPPROTO_IP == protocol) {
149 protocol = answer->protocol;
150 break;
151 }
152 if (IPPROTO_IP == answer->protocol)
153 break;
154 }
155 err = -EPROTONOSUPPORT;
156 }
157
158 if (err) {
159 if (try_loading_module < 2) {
160 rcu_read_unlock();
161 /*
162 * Be more specific, e.g. net-pf-10-proto-132-type-1
163 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
164 */
165 if (++try_loading_module == 1)
166 request_module("net-pf-%d-proto-%d-type-%d",
167 PF_INET6, protocol, sock->type);
168 /*
169 * Fall back to generic, e.g. net-pf-10-proto-132
170 * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
171 */
172 else
173 request_module("net-pf-%d-proto-%d",
174 PF_INET6, protocol);
175 goto lookup_protocol;
176 } else
177 goto out_rcu_unlock;
178 }
179
180 err = -EPERM;
181 if (sock->type == SOCK_RAW && !kern &&
182 !ns_capable(net->user_ns, CAP_NET_RAW))
183 goto out_rcu_unlock;
184
185 sock->ops = answer->ops;
186 answer_prot = answer->prot;
187 answer_flags = answer->flags;
188 rcu_read_unlock();
189
190 WARN_ON(!answer_prot->slab);
191
192 err = -ENOBUFS;
193 sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot, kern);
194 if (!sk)
195 goto out;
196
197 sock_init_data(sock, sk);
198
199 err = 0;
200 if (INET_PROTOSW_REUSE & answer_flags)
201 sk->sk_reuse = SK_CAN_REUSE;
202
203 if (INET_PROTOSW_ICSK & answer_flags)
204 inet_init_csk_locks(sk);
205
206 inet = inet_sk(sk);
207 inet_assign_bit(IS_ICSK, sk, INET_PROTOSW_ICSK & answer_flags);
208
209 if (SOCK_RAW == sock->type) {
210 inet->inet_num = protocol;
211 if (IPPROTO_RAW == protocol)
212 inet_set_bit(HDRINCL, sk);
213 }
214
215 sk->sk_destruct = inet6_sock_destruct;
216 sk->sk_family = PF_INET6;
217 sk->sk_protocol = protocol;
218
219 sk->sk_backlog_rcv = answer->prot->backlog_rcv;
220
221 inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
222 np->hop_limit = -1;
223 np->mcast_hops = IPV6_DEFAULT_MCASTHOPS;
224 inet6_set_bit(MC6_LOOP, sk);
225 inet6_set_bit(MC6_ALL, sk);
226 np->pmtudisc = IPV6_PMTUDISC_WANT;
227 inet6_assign_bit(REPFLOW, sk, net->ipv6.sysctl.flowlabel_reflect &
228 FLOWLABEL_REFLECT_ESTABLISHED);
229 sk->sk_ipv6only = net->ipv6.sysctl.bindv6only;
230 sk->sk_txrehash = READ_ONCE(net->core.sysctl_txrehash);
231
232 /* Init the ipv4 part of the socket since we can have sockets
233 * using v6 API for ipv4.
234 */
235 inet->uc_ttl = -1;
236
237 inet_set_bit(MC_LOOP, sk);
238 inet->mc_ttl = 1;
239 inet->mc_index = 0;
240 RCU_INIT_POINTER(inet->mc_list, NULL);
241 inet->rcv_tos = 0;
242
243 if (READ_ONCE(net->ipv4.sysctl_ip_no_pmtu_disc))
244 inet->pmtudisc = IP_PMTUDISC_DONT;
245 else
246 inet->pmtudisc = IP_PMTUDISC_WANT;
247
248 if (inet->inet_num) {
249 /* It assumes that any protocol which allows
250 * the user to assign a number at socket
251 * creation time automatically shares.
252 */
253 inet->inet_sport = htons(inet->inet_num);
254 err = sk->sk_prot->hash(sk);
255 if (err) {
256 sk_common_release(sk);
257 goto out;
258 }
259 }
260 if (sk->sk_prot->init) {
261 err = sk->sk_prot->init(sk);
262 if (err) {
263 sk_common_release(sk);
264 goto out;
265 }
266 }
267
268 if (!kern) {
269 err = BPF_CGROUP_RUN_PROG_INET_SOCK(sk);
270 if (err) {
271 sk_common_release(sk);
272 goto out;
273 }
274 }
275out:
276 return err;
277out_rcu_unlock:
278 rcu_read_unlock();
279 goto out;
280}
281
282static int __inet6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len,
283 u32 flags)
284{
285 struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr;
286 struct inet_sock *inet = inet_sk(sk);
287 struct ipv6_pinfo *np = inet6_sk(sk);
288 struct net *net = sock_net(sk);
289 __be32 v4addr = 0;
290 unsigned short snum;
291 bool saved_ipv6only;
292 int addr_type = 0;
293 int err = 0;
294
295 if (addr->sin6_family != AF_INET6)
296 return -EAFNOSUPPORT;
297
298 addr_type = ipv6_addr_type(&addr->sin6_addr);
299 if ((addr_type & IPV6_ADDR_MULTICAST) && sk->sk_type == SOCK_STREAM)
300 return -EINVAL;
301
302 snum = ntohs(addr->sin6_port);
303 if (!(flags & BIND_NO_CAP_NET_BIND_SERVICE) &&
304 snum && inet_port_requires_bind_service(net, snum) &&
305 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
306 return -EACCES;
307
308 if (flags & BIND_WITH_LOCK)
309 lock_sock(sk);
310
311 /* Check these errors (active socket, double bind). */
312 if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
313 err = -EINVAL;
314 goto out;
315 }
316
317 /* Check if the address belongs to the host. */
318 if (addr_type == IPV6_ADDR_MAPPED) {
319 struct net_device *dev = NULL;
320 int chk_addr_ret;
321
322 /* Binding to v4-mapped address on a v6-only socket
323 * makes no sense
324 */
325 if (ipv6_only_sock(sk)) {
326 err = -EINVAL;
327 goto out;
328 }
329
330 rcu_read_lock();
331 if (sk->sk_bound_dev_if) {
332 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
333 if (!dev) {
334 err = -ENODEV;
335 goto out_unlock;
336 }
337 }
338
339 /* Reproduce AF_INET checks to make the bindings consistent */
340 v4addr = addr->sin6_addr.s6_addr32[3];
341 chk_addr_ret = inet_addr_type_dev_table(net, dev, v4addr);
342 rcu_read_unlock();
343
344 if (!inet_addr_valid_or_nonlocal(net, inet, v4addr,
345 chk_addr_ret)) {
346 err = -EADDRNOTAVAIL;
347 goto out;
348 }
349 } else {
350 if (addr_type != IPV6_ADDR_ANY) {
351 struct net_device *dev = NULL;
352
353 rcu_read_lock();
354 if (__ipv6_addr_needs_scope_id(addr_type)) {
355 if (addr_len >= sizeof(struct sockaddr_in6) &&
356 addr->sin6_scope_id) {
357 /* Override any existing binding, if another one
358 * is supplied by user.
359 */
360 sk->sk_bound_dev_if = addr->sin6_scope_id;
361 }
362
363 /* Binding to link-local address requires an interface */
364 if (!sk->sk_bound_dev_if) {
365 err = -EINVAL;
366 goto out_unlock;
367 }
368 }
369
370 if (sk->sk_bound_dev_if) {
371 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
372 if (!dev) {
373 err = -ENODEV;
374 goto out_unlock;
375 }
376 }
377
378 /* ipv4 addr of the socket is invalid. Only the
379 * unspecified and mapped address have a v4 equivalent.
380 */
381 v4addr = LOOPBACK4_IPV6;
382 if (!(addr_type & IPV6_ADDR_MULTICAST)) {
383 if (!ipv6_can_nonlocal_bind(net, inet) &&
384 !ipv6_chk_addr(net, &addr->sin6_addr,
385 dev, 0)) {
386 err = -EADDRNOTAVAIL;
387 goto out_unlock;
388 }
389 }
390 rcu_read_unlock();
391 }
392 }
393
394 inet->inet_rcv_saddr = v4addr;
395 inet->inet_saddr = v4addr;
396
397 sk->sk_v6_rcv_saddr = addr->sin6_addr;
398
399 if (!(addr_type & IPV6_ADDR_MULTICAST))
400 np->saddr = addr->sin6_addr;
401
402 saved_ipv6only = sk->sk_ipv6only;
403 if (addr_type != IPV6_ADDR_ANY && addr_type != IPV6_ADDR_MAPPED)
404 sk->sk_ipv6only = 1;
405
406 /* Make sure we are allowed to bind here. */
407 if (snum || !(inet_test_bit(BIND_ADDRESS_NO_PORT, sk) ||
408 (flags & BIND_FORCE_ADDRESS_NO_PORT))) {
409 err = sk->sk_prot->get_port(sk, snum);
410 if (err) {
411 sk->sk_ipv6only = saved_ipv6only;
412 inet_reset_saddr(sk);
413 goto out;
414 }
415 if (!(flags & BIND_FROM_BPF)) {
416 err = BPF_CGROUP_RUN_PROG_INET6_POST_BIND(sk);
417 if (err) {
418 sk->sk_ipv6only = saved_ipv6only;
419 inet_reset_saddr(sk);
420 if (sk->sk_prot->put_port)
421 sk->sk_prot->put_port(sk);
422 goto out;
423 }
424 }
425 }
426
427 if (addr_type != IPV6_ADDR_ANY)
428 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
429 if (snum)
430 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
431 inet->inet_sport = htons(inet->inet_num);
432 inet->inet_dport = 0;
433 inet->inet_daddr = 0;
434out:
435 if (flags & BIND_WITH_LOCK)
436 release_sock(sk);
437 return err;
438out_unlock:
439 rcu_read_unlock();
440 goto out;
441}
442
443int inet6_bind_sk(struct sock *sk, struct sockaddr *uaddr, int addr_len)
444{
445 u32 flags = BIND_WITH_LOCK;
446 const struct proto *prot;
447 int err = 0;
448
449 /* IPV6_ADDRFORM can change sk->sk_prot under us. */
450 prot = READ_ONCE(sk->sk_prot);
451 /* If the socket has its own bind function then use it. */
452 if (prot->bind)
453 return prot->bind(sk, uaddr, addr_len);
454
455 if (addr_len < SIN6_LEN_RFC2133)
456 return -EINVAL;
457
458 /* BPF prog is run before any checks are done so that if the prog
459 * changes context in a wrong way it will be caught.
460 */
461 err = BPF_CGROUP_RUN_PROG_INET_BIND_LOCK(sk, uaddr, &addr_len,
462 CGROUP_INET6_BIND, &flags);
463 if (err)
464 return err;
465
466 return __inet6_bind(sk, uaddr, addr_len, flags);
467}
468
469/* bind for INET6 API */
470int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
471{
472 return inet6_bind_sk(sock->sk, uaddr, addr_len);
473}
474EXPORT_SYMBOL(inet6_bind);
475
476int inet6_release(struct socket *sock)
477{
478 struct sock *sk = sock->sk;
479
480 if (!sk)
481 return -EINVAL;
482
483 /* Free mc lists */
484 ipv6_sock_mc_close(sk);
485
486 /* Free ac lists */
487 ipv6_sock_ac_close(sk);
488
489 return inet_release(sock);
490}
491EXPORT_SYMBOL(inet6_release);
492
493void inet6_cleanup_sock(struct sock *sk)
494{
495 struct ipv6_pinfo *np = inet6_sk(sk);
496 struct sk_buff *skb;
497 struct ipv6_txoptions *opt;
498
499 /* Release rx options */
500
501 skb = xchg(&np->pktoptions, NULL);
502 kfree_skb(skb);
503
504 skb = xchg(&np->rxpmtu, NULL);
505 kfree_skb(skb);
506
507 /* Free flowlabels */
508 fl6_free_socklist(sk);
509
510 /* Free tx options */
511
512 opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL);
513 if (opt) {
514 atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
515 txopt_put(opt);
516 }
517}
518EXPORT_SYMBOL_GPL(inet6_cleanup_sock);
519
520/*
521 * This does both peername and sockname.
522 */
523int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
524 int peer)
525{
526 struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
527 int sin_addr_len = sizeof(*sin);
528 struct sock *sk = sock->sk;
529 struct inet_sock *inet = inet_sk(sk);
530 struct ipv6_pinfo *np = inet6_sk(sk);
531
532 sin->sin6_family = AF_INET6;
533 sin->sin6_flowinfo = 0;
534 sin->sin6_scope_id = 0;
535 lock_sock(sk);
536 if (peer) {
537 if (!inet->inet_dport ||
538 (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
539 peer == 1)) {
540 release_sock(sk);
541 return -ENOTCONN;
542 }
543 sin->sin6_port = inet->inet_dport;
544 sin->sin6_addr = sk->sk_v6_daddr;
545 if (inet6_test_bit(SNDFLOW, sk))
546 sin->sin6_flowinfo = np->flow_label;
547 BPF_CGROUP_RUN_SA_PROG(sk, (struct sockaddr *)sin, &sin_addr_len,
548 CGROUP_INET6_GETPEERNAME);
549 } else {
550 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
551 sin->sin6_addr = np->saddr;
552 else
553 sin->sin6_addr = sk->sk_v6_rcv_saddr;
554 sin->sin6_port = inet->inet_sport;
555 BPF_CGROUP_RUN_SA_PROG(sk, (struct sockaddr *)sin, &sin_addr_len,
556 CGROUP_INET6_GETSOCKNAME);
557 }
558 sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
559 sk->sk_bound_dev_if);
560 release_sock(sk);
561 return sin_addr_len;
562}
563EXPORT_SYMBOL(inet6_getname);
564
565int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
566{
567 void __user *argp = (void __user *)arg;
568 struct sock *sk = sock->sk;
569 struct net *net = sock_net(sk);
570 const struct proto *prot;
571
572 switch (cmd) {
573 case SIOCADDRT:
574 case SIOCDELRT: {
575 struct in6_rtmsg rtmsg;
576
577 if (copy_from_user(&rtmsg, argp, sizeof(rtmsg)))
578 return -EFAULT;
579 return ipv6_route_ioctl(net, cmd, &rtmsg);
580 }
581 case SIOCSIFADDR:
582 return addrconf_add_ifaddr(net, argp);
583 case SIOCDIFADDR:
584 return addrconf_del_ifaddr(net, argp);
585 case SIOCSIFDSTADDR:
586 return addrconf_set_dstaddr(net, argp);
587 default:
588 /* IPV6_ADDRFORM can change sk->sk_prot under us. */
589 prot = READ_ONCE(sk->sk_prot);
590 if (!prot->ioctl)
591 return -ENOIOCTLCMD;
592 return sk_ioctl(sk, cmd, (void __user *)arg);
593 }
594 /*NOTREACHED*/
595 return 0;
596}
597EXPORT_SYMBOL(inet6_ioctl);
598
599#ifdef CONFIG_COMPAT
600struct compat_in6_rtmsg {
601 struct in6_addr rtmsg_dst;
602 struct in6_addr rtmsg_src;
603 struct in6_addr rtmsg_gateway;
604 u32 rtmsg_type;
605 u16 rtmsg_dst_len;
606 u16 rtmsg_src_len;
607 u32 rtmsg_metric;
608 u32 rtmsg_info;
609 u32 rtmsg_flags;
610 s32 rtmsg_ifindex;
611};
612
613static int inet6_compat_routing_ioctl(struct sock *sk, unsigned int cmd,
614 struct compat_in6_rtmsg __user *ur)
615{
616 struct in6_rtmsg rt;
617
618 if (copy_from_user(&rt.rtmsg_dst, &ur->rtmsg_dst,
619 3 * sizeof(struct in6_addr)) ||
620 get_user(rt.rtmsg_type, &ur->rtmsg_type) ||
621 get_user(rt.rtmsg_dst_len, &ur->rtmsg_dst_len) ||
622 get_user(rt.rtmsg_src_len, &ur->rtmsg_src_len) ||
623 get_user(rt.rtmsg_metric, &ur->rtmsg_metric) ||
624 get_user(rt.rtmsg_info, &ur->rtmsg_info) ||
625 get_user(rt.rtmsg_flags, &ur->rtmsg_flags) ||
626 get_user(rt.rtmsg_ifindex, &ur->rtmsg_ifindex))
627 return -EFAULT;
628
629
630 return ipv6_route_ioctl(sock_net(sk), cmd, &rt);
631}
632
633int inet6_compat_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
634{
635 void __user *argp = compat_ptr(arg);
636 struct sock *sk = sock->sk;
637
638 switch (cmd) {
639 case SIOCADDRT:
640 case SIOCDELRT:
641 return inet6_compat_routing_ioctl(sk, cmd, argp);
642 default:
643 return -ENOIOCTLCMD;
644 }
645}
646EXPORT_SYMBOL_GPL(inet6_compat_ioctl);
647#endif /* CONFIG_COMPAT */
648
649INDIRECT_CALLABLE_DECLARE(int udpv6_sendmsg(struct sock *, struct msghdr *,
650 size_t));
651int inet6_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
652{
653 struct sock *sk = sock->sk;
654 const struct proto *prot;
655
656 if (unlikely(inet_send_prepare(sk)))
657 return -EAGAIN;
658
659 /* IPV6_ADDRFORM can change sk->sk_prot under us. */
660 prot = READ_ONCE(sk->sk_prot);
661 return INDIRECT_CALL_2(prot->sendmsg, tcp_sendmsg, udpv6_sendmsg,
662 sk, msg, size);
663}
664
665INDIRECT_CALLABLE_DECLARE(int udpv6_recvmsg(struct sock *, struct msghdr *,
666 size_t, int, int *));
667int inet6_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
668 int flags)
669{
670 struct sock *sk = sock->sk;
671 const struct proto *prot;
672 int addr_len = 0;
673 int err;
674
675 if (likely(!(flags & MSG_ERRQUEUE)))
676 sock_rps_record_flow(sk);
677
678 /* IPV6_ADDRFORM can change sk->sk_prot under us. */
679 prot = READ_ONCE(sk->sk_prot);
680 err = INDIRECT_CALL_2(prot->recvmsg, tcp_recvmsg, udpv6_recvmsg,
681 sk, msg, size, flags, &addr_len);
682 if (err >= 0)
683 msg->msg_namelen = addr_len;
684 return err;
685}
686
687const struct proto_ops inet6_stream_ops = {
688 .family = PF_INET6,
689 .owner = THIS_MODULE,
690 .release = inet6_release,
691 .bind = inet6_bind,
692 .connect = inet_stream_connect, /* ok */
693 .socketpair = sock_no_socketpair, /* a do nothing */
694 .accept = inet_accept, /* ok */
695 .getname = inet6_getname,
696 .poll = tcp_poll, /* ok */
697 .ioctl = inet6_ioctl, /* must change */
698 .gettstamp = sock_gettstamp,
699 .listen = inet_listen, /* ok */
700 .shutdown = inet_shutdown, /* ok */
701 .setsockopt = sock_common_setsockopt, /* ok */
702 .getsockopt = sock_common_getsockopt, /* ok */
703 .sendmsg = inet6_sendmsg, /* retpoline's sake */
704 .recvmsg = inet6_recvmsg, /* retpoline's sake */
705#ifdef CONFIG_MMU
706 .mmap = tcp_mmap,
707#endif
708 .splice_eof = inet_splice_eof,
709 .sendmsg_locked = tcp_sendmsg_locked,
710 .splice_read = tcp_splice_read,
711 .read_sock = tcp_read_sock,
712 .read_skb = tcp_read_skb,
713 .peek_len = tcp_peek_len,
714#ifdef CONFIG_COMPAT
715 .compat_ioctl = inet6_compat_ioctl,
716#endif
717 .set_rcvlowat = tcp_set_rcvlowat,
718};
719
720const struct proto_ops inet6_dgram_ops = {
721 .family = PF_INET6,
722 .owner = THIS_MODULE,
723 .release = inet6_release,
724 .bind = inet6_bind,
725 .connect = inet_dgram_connect, /* ok */
726 .socketpair = sock_no_socketpair, /* a do nothing */
727 .accept = sock_no_accept, /* a do nothing */
728 .getname = inet6_getname,
729 .poll = udp_poll, /* ok */
730 .ioctl = inet6_ioctl, /* must change */
731 .gettstamp = sock_gettstamp,
732 .listen = sock_no_listen, /* ok */
733 .shutdown = inet_shutdown, /* ok */
734 .setsockopt = sock_common_setsockopt, /* ok */
735 .getsockopt = sock_common_getsockopt, /* ok */
736 .sendmsg = inet6_sendmsg, /* retpoline's sake */
737 .recvmsg = inet6_recvmsg, /* retpoline's sake */
738 .read_skb = udp_read_skb,
739 .mmap = sock_no_mmap,
740 .set_peek_off = udp_set_peek_off,
741#ifdef CONFIG_COMPAT
742 .compat_ioctl = inet6_compat_ioctl,
743#endif
744};
745
746static const struct net_proto_family inet6_family_ops = {
747 .family = PF_INET6,
748 .create = inet6_create,
749 .owner = THIS_MODULE,
750};
751
752int inet6_register_protosw(struct inet_protosw *p)
753{
754 struct list_head *lh;
755 struct inet_protosw *answer;
756 struct list_head *last_perm;
757 int protocol = p->protocol;
758 int ret;
759
760 spin_lock_bh(&inetsw6_lock);
761
762 ret = -EINVAL;
763 if (p->type >= SOCK_MAX)
764 goto out_illegal;
765
766 /* If we are trying to override a permanent protocol, bail. */
767 answer = NULL;
768 ret = -EPERM;
769 last_perm = &inetsw6[p->type];
770 list_for_each(lh, &inetsw6[p->type]) {
771 answer = list_entry(lh, struct inet_protosw, list);
772
773 /* Check only the non-wild match. */
774 if (INET_PROTOSW_PERMANENT & answer->flags) {
775 if (protocol == answer->protocol)
776 break;
777 last_perm = lh;
778 }
779
780 answer = NULL;
781 }
782 if (answer)
783 goto out_permanent;
784
785 /* Add the new entry after the last permanent entry if any, so that
786 * the new entry does not override a permanent entry when matched with
787 * a wild-card protocol. But it is allowed to override any existing
788 * non-permanent entry. This means that when we remove this entry, the
789 * system automatically returns to the old behavior.
790 */
791 list_add_rcu(&p->list, last_perm);
792 ret = 0;
793out:
794 spin_unlock_bh(&inetsw6_lock);
795 return ret;
796
797out_permanent:
798 pr_err("Attempt to override permanent protocol %d\n", protocol);
799 goto out;
800
801out_illegal:
802 pr_err("Ignoring attempt to register invalid socket type %d\n",
803 p->type);
804 goto out;
805}
806EXPORT_SYMBOL(inet6_register_protosw);
807
808void
809inet6_unregister_protosw(struct inet_protosw *p)
810{
811 if (INET_PROTOSW_PERMANENT & p->flags) {
812 pr_err("Attempt to unregister permanent protocol %d\n",
813 p->protocol);
814 } else {
815 spin_lock_bh(&inetsw6_lock);
816 list_del_rcu(&p->list);
817 spin_unlock_bh(&inetsw6_lock);
818
819 synchronize_net();
820 }
821}
822EXPORT_SYMBOL(inet6_unregister_protosw);
823
824int inet6_sk_rebuild_header(struct sock *sk)
825{
826 struct ipv6_pinfo *np = inet6_sk(sk);
827 struct dst_entry *dst;
828
829 dst = __sk_dst_check(sk, np->dst_cookie);
830
831 if (!dst) {
832 struct inet_sock *inet = inet_sk(sk);
833 struct in6_addr *final_p, final;
834 struct flowi6 fl6;
835
836 memset(&fl6, 0, sizeof(fl6));
837 fl6.flowi6_proto = sk->sk_protocol;
838 fl6.daddr = sk->sk_v6_daddr;
839 fl6.saddr = np->saddr;
840 fl6.flowlabel = np->flow_label;
841 fl6.flowi6_oif = sk->sk_bound_dev_if;
842 fl6.flowi6_mark = sk->sk_mark;
843 fl6.fl6_dport = inet->inet_dport;
844 fl6.fl6_sport = inet->inet_sport;
845 fl6.flowi6_uid = sk->sk_uid;
846 security_sk_classify_flow(sk, flowi6_to_flowi_common(&fl6));
847
848 rcu_read_lock();
849 final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt),
850 &final);
851 rcu_read_unlock();
852
853 dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p);
854 if (IS_ERR(dst)) {
855 sk->sk_route_caps = 0;
856 WRITE_ONCE(sk->sk_err_soft, -PTR_ERR(dst));
857 return PTR_ERR(dst);
858 }
859
860 ip6_dst_store(sk, dst, NULL, NULL);
861 }
862
863 return 0;
864}
865EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
866
867bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb,
868 const struct inet6_skb_parm *opt)
869{
870 const struct ipv6_pinfo *np = inet6_sk(sk);
871
872 if (np->rxopt.all) {
873 if (((opt->flags & IP6SKB_HOPBYHOP) &&
874 (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) ||
875 (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) &&
876 np->rxopt.bits.rxflow) ||
877 (opt->srcrt && (np->rxopt.bits.srcrt ||
878 np->rxopt.bits.osrcrt)) ||
879 ((opt->dst1 || opt->dst0) &&
880 (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
881 return true;
882 }
883 return false;
884}
885EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
886
887static struct packet_type ipv6_packet_type __read_mostly = {
888 .type = cpu_to_be16(ETH_P_IPV6),
889 .func = ipv6_rcv,
890 .list_func = ipv6_list_rcv,
891};
892
893static int __init ipv6_packet_init(void)
894{
895 dev_add_pack(&ipv6_packet_type);
896 return 0;
897}
898
899static void ipv6_packet_cleanup(void)
900{
901 dev_remove_pack(&ipv6_packet_type);
902}
903
904static int __net_init ipv6_init_mibs(struct net *net)
905{
906 int i;
907
908 net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib);
909 if (!net->mib.udp_stats_in6)
910 return -ENOMEM;
911 net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib);
912 if (!net->mib.udplite_stats_in6)
913 goto err_udplite_mib;
914 net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib);
915 if (!net->mib.ipv6_statistics)
916 goto err_ip_mib;
917
918 for_each_possible_cpu(i) {
919 struct ipstats_mib *af_inet6_stats;
920 af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i);
921 u64_stats_init(&af_inet6_stats->syncp);
922 }
923
924
925 net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib);
926 if (!net->mib.icmpv6_statistics)
927 goto err_icmp_mib;
928 net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
929 GFP_KERNEL);
930 if (!net->mib.icmpv6msg_statistics)
931 goto err_icmpmsg_mib;
932 return 0;
933
934err_icmpmsg_mib:
935 free_percpu(net->mib.icmpv6_statistics);
936err_icmp_mib:
937 free_percpu(net->mib.ipv6_statistics);
938err_ip_mib:
939 free_percpu(net->mib.udplite_stats_in6);
940err_udplite_mib:
941 free_percpu(net->mib.udp_stats_in6);
942 return -ENOMEM;
943}
944
945static void ipv6_cleanup_mibs(struct net *net)
946{
947 free_percpu(net->mib.udp_stats_in6);
948 free_percpu(net->mib.udplite_stats_in6);
949 free_percpu(net->mib.ipv6_statistics);
950 free_percpu(net->mib.icmpv6_statistics);
951 kfree(net->mib.icmpv6msg_statistics);
952}
953
954static int __net_init inet6_net_init(struct net *net)
955{
956 int err = 0;
957
958 net->ipv6.sysctl.bindv6only = 0;
959 net->ipv6.sysctl.icmpv6_time = 1*HZ;
960 net->ipv6.sysctl.icmpv6_echo_ignore_all = 0;
961 net->ipv6.sysctl.icmpv6_echo_ignore_multicast = 0;
962 net->ipv6.sysctl.icmpv6_echo_ignore_anycast = 0;
963 net->ipv6.sysctl.icmpv6_error_anycast_as_unicast = 0;
964
965 /* By default, rate limit error messages.
966 * Except for pmtu discovery, it would break it.
967 * proc_do_large_bitmap needs pointer to the bitmap.
968 */
969 bitmap_set(net->ipv6.sysctl.icmpv6_ratemask, 0, ICMPV6_ERRMSG_MAX + 1);
970 bitmap_clear(net->ipv6.sysctl.icmpv6_ratemask, ICMPV6_PKT_TOOBIG, 1);
971 net->ipv6.sysctl.icmpv6_ratemask_ptr = net->ipv6.sysctl.icmpv6_ratemask;
972
973 net->ipv6.sysctl.flowlabel_consistency = 1;
974 net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS;
975 net->ipv6.sysctl.idgen_retries = 3;
976 net->ipv6.sysctl.idgen_delay = 1 * HZ;
977 net->ipv6.sysctl.flowlabel_state_ranges = 0;
978 net->ipv6.sysctl.max_dst_opts_cnt = IP6_DEFAULT_MAX_DST_OPTS_CNT;
979 net->ipv6.sysctl.max_hbh_opts_cnt = IP6_DEFAULT_MAX_HBH_OPTS_CNT;
980 net->ipv6.sysctl.max_dst_opts_len = IP6_DEFAULT_MAX_DST_OPTS_LEN;
981 net->ipv6.sysctl.max_hbh_opts_len = IP6_DEFAULT_MAX_HBH_OPTS_LEN;
982 net->ipv6.sysctl.fib_notify_on_flag_change = 0;
983 atomic_set(&net->ipv6.fib6_sernum, 1);
984
985 net->ipv6.sysctl.ioam6_id = IOAM6_DEFAULT_ID;
986 net->ipv6.sysctl.ioam6_id_wide = IOAM6_DEFAULT_ID_WIDE;
987
988 err = ipv6_init_mibs(net);
989 if (err)
990 return err;
991#ifdef CONFIG_PROC_FS
992 err = udp6_proc_init(net);
993 if (err)
994 goto out;
995 err = tcp6_proc_init(net);
996 if (err)
997 goto proc_tcp6_fail;
998 err = ac6_proc_init(net);
999 if (err)
1000 goto proc_ac6_fail;
1001#endif
1002 return err;
1003
1004#ifdef CONFIG_PROC_FS
1005proc_ac6_fail:
1006 tcp6_proc_exit(net);
1007proc_tcp6_fail:
1008 udp6_proc_exit(net);
1009out:
1010 ipv6_cleanup_mibs(net);
1011 return err;
1012#endif
1013}
1014
1015static void __net_exit inet6_net_exit(struct net *net)
1016{
1017#ifdef CONFIG_PROC_FS
1018 udp6_proc_exit(net);
1019 tcp6_proc_exit(net);
1020 ac6_proc_exit(net);
1021#endif
1022 ipv6_cleanup_mibs(net);
1023}
1024
1025static struct pernet_operations inet6_net_ops = {
1026 .init = inet6_net_init,
1027 .exit = inet6_net_exit,
1028};
1029
1030static int ipv6_route_input(struct sk_buff *skb)
1031{
1032 ip6_route_input(skb);
1033 return skb_dst(skb)->error;
1034}
1035
1036static const struct ipv6_stub ipv6_stub_impl = {
1037 .ipv6_sock_mc_join = ipv6_sock_mc_join,
1038 .ipv6_sock_mc_drop = ipv6_sock_mc_drop,
1039 .ipv6_dst_lookup_flow = ip6_dst_lookup_flow,
1040 .ipv6_route_input = ipv6_route_input,
1041 .fib6_get_table = fib6_get_table,
1042 .fib6_table_lookup = fib6_table_lookup,
1043 .fib6_lookup = fib6_lookup,
1044 .fib6_select_path = fib6_select_path,
1045 .ip6_mtu_from_fib6 = ip6_mtu_from_fib6,
1046 .fib6_nh_init = fib6_nh_init,
1047 .fib6_nh_release = fib6_nh_release,
1048 .fib6_nh_release_dsts = fib6_nh_release_dsts,
1049 .fib6_update_sernum = fib6_update_sernum_stub,
1050 .fib6_rt_update = fib6_rt_update,
1051 .ip6_del_rt = ip6_del_rt,
1052 .udpv6_encap_enable = udpv6_encap_enable,
1053 .ndisc_send_na = ndisc_send_na,
1054#if IS_ENABLED(CONFIG_XFRM)
1055 .xfrm6_local_rxpmtu = xfrm6_local_rxpmtu,
1056 .xfrm6_udp_encap_rcv = xfrm6_udp_encap_rcv,
1057 .xfrm6_gro_udp_encap_rcv = xfrm6_gro_udp_encap_rcv,
1058 .xfrm6_rcv_encap = xfrm6_rcv_encap,
1059#endif
1060 .nd_tbl = &nd_tbl,
1061 .ipv6_fragment = ip6_fragment,
1062 .ipv6_dev_find = ipv6_dev_find,
1063};
1064
1065static const struct ipv6_bpf_stub ipv6_bpf_stub_impl = {
1066 .inet6_bind = __inet6_bind,
1067 .udp6_lib_lookup = __udp6_lib_lookup,
1068 .ipv6_setsockopt = do_ipv6_setsockopt,
1069 .ipv6_getsockopt = do_ipv6_getsockopt,
1070 .ipv6_dev_get_saddr = ipv6_dev_get_saddr,
1071};
1072
1073static int __init inet6_init(void)
1074{
1075 struct list_head *r;
1076 int err = 0;
1077
1078 sock_skb_cb_check_size(sizeof(struct inet6_skb_parm));
1079
1080 /* Register the socket-side information for inet6_create. */
1081 for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
1082 INIT_LIST_HEAD(r);
1083
1084 raw_hashinfo_init(&raw_v6_hashinfo);
1085
1086 if (disable_ipv6_mod) {
1087 pr_info("Loaded, but administratively disabled, reboot required to enable\n");
1088 goto out;
1089 }
1090
1091 err = proto_register(&tcpv6_prot, 1);
1092 if (err)
1093 goto out;
1094
1095 err = proto_register(&udpv6_prot, 1);
1096 if (err)
1097 goto out_unregister_tcp_proto;
1098
1099 err = proto_register(&udplitev6_prot, 1);
1100 if (err)
1101 goto out_unregister_udp_proto;
1102
1103 err = proto_register(&rawv6_prot, 1);
1104 if (err)
1105 goto out_unregister_udplite_proto;
1106
1107 err = proto_register(&pingv6_prot, 1);
1108 if (err)
1109 goto out_unregister_raw_proto;
1110
1111 /* We MUST register RAW sockets before we create the ICMP6,
1112 * IGMP6, or NDISC control sockets.
1113 */
1114 err = rawv6_init();
1115 if (err)
1116 goto out_unregister_ping_proto;
1117
1118 /* Register the family here so that the init calls below will
1119 * be able to create sockets. (?? is this dangerous ??)
1120 */
1121 err = sock_register(&inet6_family_ops);
1122 if (err)
1123 goto out_sock_register_fail;
1124
1125 /*
1126 * ipngwg API draft makes clear that the correct semantics
1127 * for TCP and UDP is to consider one TCP and UDP instance
1128 * in a host available by both INET and INET6 APIs and
1129 * able to communicate via both network protocols.
1130 */
1131
1132 err = register_pernet_subsys(&inet6_net_ops);
1133 if (err)
1134 goto register_pernet_fail;
1135 err = ip6_mr_init();
1136 if (err)
1137 goto ipmr_fail;
1138 err = icmpv6_init();
1139 if (err)
1140 goto icmp_fail;
1141 err = ndisc_init();
1142 if (err)
1143 goto ndisc_fail;
1144 err = igmp6_init();
1145 if (err)
1146 goto igmp_fail;
1147
1148 err = ipv6_netfilter_init();
1149 if (err)
1150 goto netfilter_fail;
1151 /* Create /proc/foo6 entries. */
1152#ifdef CONFIG_PROC_FS
1153 err = -ENOMEM;
1154 if (raw6_proc_init())
1155 goto proc_raw6_fail;
1156 if (udplite6_proc_init())
1157 goto proc_udplite6_fail;
1158 if (ipv6_misc_proc_init())
1159 goto proc_misc6_fail;
1160 if (if6_proc_init())
1161 goto proc_if6_fail;
1162#endif
1163 err = ip6_route_init();
1164 if (err)
1165 goto ip6_route_fail;
1166 err = ndisc_late_init();
1167 if (err)
1168 goto ndisc_late_fail;
1169 err = ip6_flowlabel_init();
1170 if (err)
1171 goto ip6_flowlabel_fail;
1172 err = ipv6_anycast_init();
1173 if (err)
1174 goto ipv6_anycast_fail;
1175 err = addrconf_init();
1176 if (err)
1177 goto addrconf_fail;
1178
1179 /* Init v6 extension headers. */
1180 err = ipv6_exthdrs_init();
1181 if (err)
1182 goto ipv6_exthdrs_fail;
1183
1184 err = ipv6_frag_init();
1185 if (err)
1186 goto ipv6_frag_fail;
1187
1188 /* Init v6 transport protocols. */
1189 err = udpv6_init();
1190 if (err)
1191 goto udpv6_fail;
1192
1193 err = udplitev6_init();
1194 if (err)
1195 goto udplitev6_fail;
1196
1197 err = udpv6_offload_init();
1198 if (err)
1199 goto udpv6_offload_fail;
1200
1201 err = tcpv6_init();
1202 if (err)
1203 goto tcpv6_fail;
1204
1205 err = ipv6_packet_init();
1206 if (err)
1207 goto ipv6_packet_fail;
1208
1209 err = pingv6_init();
1210 if (err)
1211 goto pingv6_fail;
1212
1213 err = calipso_init();
1214 if (err)
1215 goto calipso_fail;
1216
1217 err = seg6_init();
1218 if (err)
1219 goto seg6_fail;
1220
1221 err = rpl_init();
1222 if (err)
1223 goto rpl_fail;
1224
1225 err = ioam6_init();
1226 if (err)
1227 goto ioam6_fail;
1228
1229 err = igmp6_late_init();
1230 if (err)
1231 goto igmp6_late_err;
1232
1233#ifdef CONFIG_SYSCTL
1234 err = ipv6_sysctl_register();
1235 if (err)
1236 goto sysctl_fail;
1237#endif
1238
1239 /* ensure that ipv6 stubs are visible only after ipv6 is ready */
1240 wmb();
1241 ipv6_stub = &ipv6_stub_impl;
1242 ipv6_bpf_stub = &ipv6_bpf_stub_impl;
1243out:
1244 return err;
1245
1246#ifdef CONFIG_SYSCTL
1247sysctl_fail:
1248 igmp6_late_cleanup();
1249#endif
1250igmp6_late_err:
1251 ioam6_exit();
1252ioam6_fail:
1253 rpl_exit();
1254rpl_fail:
1255 seg6_exit();
1256seg6_fail:
1257 calipso_exit();
1258calipso_fail:
1259 pingv6_exit();
1260pingv6_fail:
1261 ipv6_packet_cleanup();
1262ipv6_packet_fail:
1263 tcpv6_exit();
1264tcpv6_fail:
1265 udpv6_offload_exit();
1266udpv6_offload_fail:
1267 udplitev6_exit();
1268udplitev6_fail:
1269 udpv6_exit();
1270udpv6_fail:
1271 ipv6_frag_exit();
1272ipv6_frag_fail:
1273 ipv6_exthdrs_exit();
1274ipv6_exthdrs_fail:
1275 addrconf_cleanup();
1276addrconf_fail:
1277 ipv6_anycast_cleanup();
1278ipv6_anycast_fail:
1279 ip6_flowlabel_cleanup();
1280ip6_flowlabel_fail:
1281 ndisc_late_cleanup();
1282ndisc_late_fail:
1283 ip6_route_cleanup();
1284ip6_route_fail:
1285#ifdef CONFIG_PROC_FS
1286 if6_proc_exit();
1287proc_if6_fail:
1288 ipv6_misc_proc_exit();
1289proc_misc6_fail:
1290 udplite6_proc_exit();
1291proc_udplite6_fail:
1292 raw6_proc_exit();
1293proc_raw6_fail:
1294#endif
1295 ipv6_netfilter_fini();
1296netfilter_fail:
1297 igmp6_cleanup();
1298igmp_fail:
1299 ndisc_cleanup();
1300ndisc_fail:
1301 icmpv6_cleanup();
1302icmp_fail:
1303 ip6_mr_cleanup();
1304ipmr_fail:
1305 unregister_pernet_subsys(&inet6_net_ops);
1306register_pernet_fail:
1307 sock_unregister(PF_INET6);
1308 rtnl_unregister_all(PF_INET6);
1309out_sock_register_fail:
1310 rawv6_exit();
1311out_unregister_ping_proto:
1312 proto_unregister(&pingv6_prot);
1313out_unregister_raw_proto:
1314 proto_unregister(&rawv6_prot);
1315out_unregister_udplite_proto:
1316 proto_unregister(&udplitev6_prot);
1317out_unregister_udp_proto:
1318 proto_unregister(&udpv6_prot);
1319out_unregister_tcp_proto:
1320 proto_unregister(&tcpv6_prot);
1321 goto out;
1322}
1323module_init(inet6_init);
1324
1325MODULE_ALIAS_NETPROTO(PF_INET6);