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/*
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
71MODULE_AUTHOR("Cast of dozens");
72MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
73MODULE_LICENSE("GPL");
74
75/* The inetsw6 table contains everything that inet6_create needs to
76 * build a new socket.
77 */
78static struct list_head inetsw6[SOCK_MAX];
79static DEFINE_SPINLOCK(inetsw6_lock);
80
81struct ipv6_params ipv6_defaults = {
82 .disable_ipv6 = 0,
83 .autoconf = 1,
84};
85
86static int disable_ipv6_mod;
87
88module_param_named(disable, disable_ipv6_mod, int, 0444);
89MODULE_PARM_DESC(disable, "Disable IPv6 module such that it is non-functional");
90
91module_param_named(disable_ipv6, ipv6_defaults.disable_ipv6, int, 0444);
92MODULE_PARM_DESC(disable_ipv6, "Disable IPv6 on all interfaces");
93
94module_param_named(autoconf, ipv6_defaults.autoconf, int, 0444);
95MODULE_PARM_DESC(autoconf, "Enable IPv6 address autoconfiguration on all interfaces");
96
97bool ipv6_mod_enabled(void)
98{
99 return disable_ipv6_mod == 0;
100}
101EXPORT_SYMBOL_GPL(ipv6_mod_enabled);
102
103static __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
110static 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. */
126lookup_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->autoflowlabel = ip6_default_np_autolabel(sock_net(sk));
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 }
269out:
270 return err;
271out_rcu_unlock:
272 rcu_read_unlock();
273 goto out;
274}
275
276
277/* bind for INET6 API */
278int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
279{
280 struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr;
281 struct sock *sk = sock->sk;
282 struct inet_sock *inet = inet_sk(sk);
283 struct ipv6_pinfo *np = inet6_sk(sk);
284 struct net *net = sock_net(sk);
285 __be32 v4addr = 0;
286 unsigned short snum;
287 int addr_type = 0;
288 int err = 0;
289
290 /* If the socket has its own bind function then use it. */
291 if (sk->sk_prot->bind)
292 return sk->sk_prot->bind(sk, uaddr, addr_len);
293
294 if (addr_len < SIN6_LEN_RFC2133)
295 return -EINVAL;
296
297 if (addr->sin6_family != AF_INET6)
298 return -EAFNOSUPPORT;
299
300 addr_type = ipv6_addr_type(&addr->sin6_addr);
301 if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
302 return -EINVAL;
303
304 snum = ntohs(addr->sin6_port);
305 if (snum && snum < PROT_SOCK && !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
306 return -EACCES;
307
308 lock_sock(sk);
309
310 /* Check these errors (active socket, double bind). */
311 if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
312 err = -EINVAL;
313 goto out;
314 }
315
316 /* Check if the address belongs to the host. */
317 if (addr_type == IPV6_ADDR_MAPPED) {
318 int chk_addr_ret;
319
320 /* Binding to v4-mapped address on a v6-only socket
321 * makes no sense
322 */
323 if (sk->sk_ipv6only) {
324 err = -EINVAL;
325 goto out;
326 }
327
328 /* Reproduce AF_INET checks to make the bindings consistent */
329 v4addr = addr->sin6_addr.s6_addr32[3];
330 chk_addr_ret = inet_addr_type(net, v4addr);
331 if (!net->ipv4.sysctl_ip_nonlocal_bind &&
332 !(inet->freebind || inet->transparent) &&
333 v4addr != htonl(INADDR_ANY) &&
334 chk_addr_ret != RTN_LOCAL &&
335 chk_addr_ret != RTN_MULTICAST &&
336 chk_addr_ret != RTN_BROADCAST) {
337 err = -EADDRNOTAVAIL;
338 goto out;
339 }
340 } else {
341 if (addr_type != IPV6_ADDR_ANY) {
342 struct net_device *dev = NULL;
343
344 rcu_read_lock();
345 if (__ipv6_addr_needs_scope_id(addr_type)) {
346 if (addr_len >= sizeof(struct sockaddr_in6) &&
347 addr->sin6_scope_id) {
348 /* Override any existing binding, if another one
349 * is supplied by user.
350 */
351 sk->sk_bound_dev_if = addr->sin6_scope_id;
352 }
353
354 /* Binding to link-local address requires an interface */
355 if (!sk->sk_bound_dev_if) {
356 err = -EINVAL;
357 goto out_unlock;
358 }
359 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
360 if (!dev) {
361 err = -ENODEV;
362 goto out_unlock;
363 }
364 }
365
366 /* ipv4 addr of the socket is invalid. Only the
367 * unspecified and mapped address have a v4 equivalent.
368 */
369 v4addr = LOOPBACK4_IPV6;
370 if (!(addr_type & IPV6_ADDR_MULTICAST)) {
371 if (!net->ipv6.sysctl.ip_nonlocal_bind &&
372 !(inet->freebind || inet->transparent) &&
373 !ipv6_chk_addr(net, &addr->sin6_addr,
374 dev, 0)) {
375 err = -EADDRNOTAVAIL;
376 goto out_unlock;
377 }
378 }
379 rcu_read_unlock();
380 }
381 }
382
383 inet->inet_rcv_saddr = v4addr;
384 inet->inet_saddr = v4addr;
385
386 sk->sk_v6_rcv_saddr = addr->sin6_addr;
387
388 if (!(addr_type & IPV6_ADDR_MULTICAST))
389 np->saddr = addr->sin6_addr;
390
391 /* Make sure we are allowed to bind here. */
392 if ((snum || !inet->bind_address_no_port) &&
393 sk->sk_prot->get_port(sk, snum)) {
394 inet_reset_saddr(sk);
395 err = -EADDRINUSE;
396 goto out;
397 }
398
399 if (addr_type != IPV6_ADDR_ANY) {
400 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
401 if (addr_type != IPV6_ADDR_MAPPED)
402 sk->sk_ipv6only = 1;
403 }
404 if (snum)
405 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
406 inet->inet_sport = htons(inet->inet_num);
407 inet->inet_dport = 0;
408 inet->inet_daddr = 0;
409out:
410 release_sock(sk);
411 return err;
412out_unlock:
413 rcu_read_unlock();
414 goto out;
415}
416EXPORT_SYMBOL(inet6_bind);
417
418int inet6_release(struct socket *sock)
419{
420 struct sock *sk = sock->sk;
421
422 if (!sk)
423 return -EINVAL;
424
425 /* Free mc lists */
426 ipv6_sock_mc_close(sk);
427
428 /* Free ac lists */
429 ipv6_sock_ac_close(sk);
430
431 return inet_release(sock);
432}
433EXPORT_SYMBOL(inet6_release);
434
435void inet6_destroy_sock(struct sock *sk)
436{
437 struct ipv6_pinfo *np = inet6_sk(sk);
438 struct sk_buff *skb;
439 struct ipv6_txoptions *opt;
440
441 /* Release rx options */
442
443 skb = xchg(&np->pktoptions, NULL);
444 if (skb)
445 kfree_skb(skb);
446
447 skb = xchg(&np->rxpmtu, NULL);
448 if (skb)
449 kfree_skb(skb);
450
451 /* Free flowlabels */
452 fl6_free_socklist(sk);
453
454 /* Free tx options */
455
456 opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL);
457 if (opt) {
458 atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
459 txopt_put(opt);
460 }
461}
462EXPORT_SYMBOL_GPL(inet6_destroy_sock);
463
464/*
465 * This does both peername and sockname.
466 */
467
468int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
469 int *uaddr_len, int peer)
470{
471 struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
472 struct sock *sk = sock->sk;
473 struct inet_sock *inet = inet_sk(sk);
474 struct ipv6_pinfo *np = inet6_sk(sk);
475
476 sin->sin6_family = AF_INET6;
477 sin->sin6_flowinfo = 0;
478 sin->sin6_scope_id = 0;
479 if (peer) {
480 if (!inet->inet_dport)
481 return -ENOTCONN;
482 if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
483 peer == 1)
484 return -ENOTCONN;
485 sin->sin6_port = inet->inet_dport;
486 sin->sin6_addr = sk->sk_v6_daddr;
487 if (np->sndflow)
488 sin->sin6_flowinfo = np->flow_label;
489 } else {
490 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
491 sin->sin6_addr = np->saddr;
492 else
493 sin->sin6_addr = sk->sk_v6_rcv_saddr;
494
495 sin->sin6_port = inet->inet_sport;
496 }
497 sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
498 sk->sk_bound_dev_if);
499 *uaddr_len = sizeof(*sin);
500 return 0;
501}
502EXPORT_SYMBOL(inet6_getname);
503
504int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
505{
506 struct sock *sk = sock->sk;
507 struct net *net = sock_net(sk);
508
509 switch (cmd) {
510 case SIOCGSTAMP:
511 return sock_get_timestamp(sk, (struct timeval __user *)arg);
512
513 case SIOCGSTAMPNS:
514 return sock_get_timestampns(sk, (struct timespec __user *)arg);
515
516 case SIOCADDRT:
517 case SIOCDELRT:
518
519 return ipv6_route_ioctl(net, cmd, (void __user *)arg);
520
521 case SIOCSIFADDR:
522 return addrconf_add_ifaddr(net, (void __user *) arg);
523 case SIOCDIFADDR:
524 return addrconf_del_ifaddr(net, (void __user *) arg);
525 case SIOCSIFDSTADDR:
526 return addrconf_set_dstaddr(net, (void __user *) arg);
527 default:
528 if (!sk->sk_prot->ioctl)
529 return -ENOIOCTLCMD;
530 return sk->sk_prot->ioctl(sk, cmd, arg);
531 }
532 /*NOTREACHED*/
533 return 0;
534}
535EXPORT_SYMBOL(inet6_ioctl);
536
537const struct proto_ops inet6_stream_ops = {
538 .family = PF_INET6,
539 .owner = THIS_MODULE,
540 .release = inet6_release,
541 .bind = inet6_bind,
542 .connect = inet_stream_connect, /* ok */
543 .socketpair = sock_no_socketpair, /* a do nothing */
544 .accept = inet_accept, /* ok */
545 .getname = inet6_getname,
546 .poll = tcp_poll, /* ok */
547 .ioctl = inet6_ioctl, /* must change */
548 .listen = inet_listen, /* ok */
549 .shutdown = inet_shutdown, /* ok */
550 .setsockopt = sock_common_setsockopt, /* ok */
551 .getsockopt = sock_common_getsockopt, /* ok */
552 .sendmsg = inet_sendmsg, /* ok */
553 .recvmsg = inet_recvmsg, /* ok */
554 .mmap = sock_no_mmap,
555 .sendpage = inet_sendpage,
556 .splice_read = tcp_splice_read,
557 .read_sock = tcp_read_sock,
558 .peek_len = tcp_peek_len,
559#ifdef CONFIG_COMPAT
560 .compat_setsockopt = compat_sock_common_setsockopt,
561 .compat_getsockopt = compat_sock_common_getsockopt,
562#endif
563};
564
565const struct proto_ops inet6_dgram_ops = {
566 .family = PF_INET6,
567 .owner = THIS_MODULE,
568 .release = inet6_release,
569 .bind = inet6_bind,
570 .connect = inet_dgram_connect, /* ok */
571 .socketpair = sock_no_socketpair, /* a do nothing */
572 .accept = sock_no_accept, /* a do nothing */
573 .getname = inet6_getname,
574 .poll = udp_poll, /* ok */
575 .ioctl = inet6_ioctl, /* must change */
576 .listen = sock_no_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 = sock_no_mmap,
583 .sendpage = sock_no_sendpage,
584 .set_peek_off = sk_set_peek_off,
585#ifdef CONFIG_COMPAT
586 .compat_setsockopt = compat_sock_common_setsockopt,
587 .compat_getsockopt = compat_sock_common_getsockopt,
588#endif
589};
590
591static const struct net_proto_family inet6_family_ops = {
592 .family = PF_INET6,
593 .create = inet6_create,
594 .owner = THIS_MODULE,
595};
596
597int inet6_register_protosw(struct inet_protosw *p)
598{
599 struct list_head *lh;
600 struct inet_protosw *answer;
601 struct list_head *last_perm;
602 int protocol = p->protocol;
603 int ret;
604
605 spin_lock_bh(&inetsw6_lock);
606
607 ret = -EINVAL;
608 if (p->type >= SOCK_MAX)
609 goto out_illegal;
610
611 /* If we are trying to override a permanent protocol, bail. */
612 answer = NULL;
613 ret = -EPERM;
614 last_perm = &inetsw6[p->type];
615 list_for_each(lh, &inetsw6[p->type]) {
616 answer = list_entry(lh, struct inet_protosw, list);
617
618 /* Check only the non-wild match. */
619 if (INET_PROTOSW_PERMANENT & answer->flags) {
620 if (protocol == answer->protocol)
621 break;
622 last_perm = lh;
623 }
624
625 answer = NULL;
626 }
627 if (answer)
628 goto out_permanent;
629
630 /* Add the new entry after the last permanent entry if any, so that
631 * the new entry does not override a permanent entry when matched with
632 * a wild-card protocol. But it is allowed to override any existing
633 * non-permanent entry. This means that when we remove this entry, the
634 * system automatically returns to the old behavior.
635 */
636 list_add_rcu(&p->list, last_perm);
637 ret = 0;
638out:
639 spin_unlock_bh(&inetsw6_lock);
640 return ret;
641
642out_permanent:
643 pr_err("Attempt to override permanent protocol %d\n", protocol);
644 goto out;
645
646out_illegal:
647 pr_err("Ignoring attempt to register invalid socket type %d\n",
648 p->type);
649 goto out;
650}
651EXPORT_SYMBOL(inet6_register_protosw);
652
653void
654inet6_unregister_protosw(struct inet_protosw *p)
655{
656 if (INET_PROTOSW_PERMANENT & p->flags) {
657 pr_err("Attempt to unregister permanent protocol %d\n",
658 p->protocol);
659 } else {
660 spin_lock_bh(&inetsw6_lock);
661 list_del_rcu(&p->list);
662 spin_unlock_bh(&inetsw6_lock);
663
664 synchronize_net();
665 }
666}
667EXPORT_SYMBOL(inet6_unregister_protosw);
668
669int inet6_sk_rebuild_header(struct sock *sk)
670{
671 struct ipv6_pinfo *np = inet6_sk(sk);
672 struct dst_entry *dst;
673
674 dst = __sk_dst_check(sk, np->dst_cookie);
675
676 if (!dst) {
677 struct inet_sock *inet = inet_sk(sk);
678 struct in6_addr *final_p, final;
679 struct flowi6 fl6;
680
681 memset(&fl6, 0, sizeof(fl6));
682 fl6.flowi6_proto = sk->sk_protocol;
683 fl6.daddr = sk->sk_v6_daddr;
684 fl6.saddr = np->saddr;
685 fl6.flowlabel = np->flow_label;
686 fl6.flowi6_oif = sk->sk_bound_dev_if;
687 fl6.flowi6_mark = sk->sk_mark;
688 fl6.fl6_dport = inet->inet_dport;
689 fl6.fl6_sport = inet->inet_sport;
690 fl6.flowi6_uid = sk->sk_uid;
691 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
692
693 rcu_read_lock();
694 final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt),
695 &final);
696 rcu_read_unlock();
697
698 dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
699 if (IS_ERR(dst)) {
700 sk->sk_route_caps = 0;
701 sk->sk_err_soft = -PTR_ERR(dst);
702 return PTR_ERR(dst);
703 }
704
705 ip6_dst_store(sk, dst, NULL, NULL);
706 }
707
708 return 0;
709}
710EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
711
712bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb,
713 const struct inet6_skb_parm *opt)
714{
715 const struct ipv6_pinfo *np = inet6_sk(sk);
716
717 if (np->rxopt.all) {
718 if (((opt->flags & IP6SKB_HOPBYHOP) &&
719 (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) ||
720 (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) &&
721 np->rxopt.bits.rxflow) ||
722 (opt->srcrt && (np->rxopt.bits.srcrt ||
723 np->rxopt.bits.osrcrt)) ||
724 ((opt->dst1 || opt->dst0) &&
725 (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
726 return true;
727 }
728 return false;
729}
730EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
731
732static struct packet_type ipv6_packet_type __read_mostly = {
733 .type = cpu_to_be16(ETH_P_IPV6),
734 .func = ipv6_rcv,
735};
736
737static int __init ipv6_packet_init(void)
738{
739 dev_add_pack(&ipv6_packet_type);
740 return 0;
741}
742
743static void ipv6_packet_cleanup(void)
744{
745 dev_remove_pack(&ipv6_packet_type);
746}
747
748static int __net_init ipv6_init_mibs(struct net *net)
749{
750 int i;
751
752 net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib);
753 if (!net->mib.udp_stats_in6)
754 return -ENOMEM;
755 net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib);
756 if (!net->mib.udplite_stats_in6)
757 goto err_udplite_mib;
758 net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib);
759 if (!net->mib.ipv6_statistics)
760 goto err_ip_mib;
761
762 for_each_possible_cpu(i) {
763 struct ipstats_mib *af_inet6_stats;
764 af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i);
765 u64_stats_init(&af_inet6_stats->syncp);
766 }
767
768
769 net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib);
770 if (!net->mib.icmpv6_statistics)
771 goto err_icmp_mib;
772 net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
773 GFP_KERNEL);
774 if (!net->mib.icmpv6msg_statistics)
775 goto err_icmpmsg_mib;
776 return 0;
777
778err_icmpmsg_mib:
779 free_percpu(net->mib.icmpv6_statistics);
780err_icmp_mib:
781 free_percpu(net->mib.ipv6_statistics);
782err_ip_mib:
783 free_percpu(net->mib.udplite_stats_in6);
784err_udplite_mib:
785 free_percpu(net->mib.udp_stats_in6);
786 return -ENOMEM;
787}
788
789static void ipv6_cleanup_mibs(struct net *net)
790{
791 free_percpu(net->mib.udp_stats_in6);
792 free_percpu(net->mib.udplite_stats_in6);
793 free_percpu(net->mib.ipv6_statistics);
794 free_percpu(net->mib.icmpv6_statistics);
795 kfree(net->mib.icmpv6msg_statistics);
796}
797
798static int __net_init inet6_net_init(struct net *net)
799{
800 int err = 0;
801
802 net->ipv6.sysctl.bindv6only = 0;
803 net->ipv6.sysctl.icmpv6_time = 1*HZ;
804 net->ipv6.sysctl.flowlabel_consistency = 1;
805 net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS;
806 net->ipv6.sysctl.idgen_retries = 3;
807 net->ipv6.sysctl.idgen_delay = 1 * HZ;
808 net->ipv6.sysctl.flowlabel_state_ranges = 0;
809 atomic_set(&net->ipv6.fib6_sernum, 1);
810
811 err = ipv6_init_mibs(net);
812 if (err)
813 return err;
814#ifdef CONFIG_PROC_FS
815 err = udp6_proc_init(net);
816 if (err)
817 goto out;
818 err = tcp6_proc_init(net);
819 if (err)
820 goto proc_tcp6_fail;
821 err = ac6_proc_init(net);
822 if (err)
823 goto proc_ac6_fail;
824#endif
825 return err;
826
827#ifdef CONFIG_PROC_FS
828proc_ac6_fail:
829 tcp6_proc_exit(net);
830proc_tcp6_fail:
831 udp6_proc_exit(net);
832out:
833 ipv6_cleanup_mibs(net);
834 return err;
835#endif
836}
837
838static void __net_exit inet6_net_exit(struct net *net)
839{
840#ifdef CONFIG_PROC_FS
841 udp6_proc_exit(net);
842 tcp6_proc_exit(net);
843 ac6_proc_exit(net);
844#endif
845 ipv6_cleanup_mibs(net);
846}
847
848static struct pernet_operations inet6_net_ops = {
849 .init = inet6_net_init,
850 .exit = inet6_net_exit,
851};
852
853static const struct ipv6_stub ipv6_stub_impl = {
854 .ipv6_sock_mc_join = ipv6_sock_mc_join,
855 .ipv6_sock_mc_drop = ipv6_sock_mc_drop,
856 .ipv6_dst_lookup = ip6_dst_lookup,
857 .udpv6_encap_enable = udpv6_encap_enable,
858 .ndisc_send_na = ndisc_send_na,
859 .nd_tbl = &nd_tbl,
860};
861
862static int __init inet6_init(void)
863{
864 struct list_head *r;
865 int err = 0;
866
867 sock_skb_cb_check_size(sizeof(struct inet6_skb_parm));
868
869 /* Register the socket-side information for inet6_create. */
870 for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
871 INIT_LIST_HEAD(r);
872
873 if (disable_ipv6_mod) {
874 pr_info("Loaded, but administratively disabled, reboot required to enable\n");
875 goto out;
876 }
877
878 err = proto_register(&tcpv6_prot, 1);
879 if (err)
880 goto out;
881
882 err = proto_register(&udpv6_prot, 1);
883 if (err)
884 goto out_unregister_tcp_proto;
885
886 err = proto_register(&udplitev6_prot, 1);
887 if (err)
888 goto out_unregister_udp_proto;
889
890 err = proto_register(&rawv6_prot, 1);
891 if (err)
892 goto out_unregister_udplite_proto;
893
894 err = proto_register(&pingv6_prot, 1);
895 if (err)
896 goto out_unregister_ping_proto;
897
898 /* We MUST register RAW sockets before we create the ICMP6,
899 * IGMP6, or NDISC control sockets.
900 */
901 err = rawv6_init();
902 if (err)
903 goto out_unregister_raw_proto;
904
905 /* Register the family here so that the init calls below will
906 * be able to create sockets. (?? is this dangerous ??)
907 */
908 err = sock_register(&inet6_family_ops);
909 if (err)
910 goto out_sock_register_fail;
911
912 /*
913 * ipngwg API draft makes clear that the correct semantics
914 * for TCP and UDP is to consider one TCP and UDP instance
915 * in a host available by both INET and INET6 APIs and
916 * able to communicate via both network protocols.
917 */
918
919 err = register_pernet_subsys(&inet6_net_ops);
920 if (err)
921 goto register_pernet_fail;
922 err = icmpv6_init();
923 if (err)
924 goto icmp_fail;
925 err = ip6_mr_init();
926 if (err)
927 goto ipmr_fail;
928 err = ndisc_init();
929 if (err)
930 goto ndisc_fail;
931 err = igmp6_init();
932 if (err)
933 goto igmp_fail;
934
935 ipv6_stub = &ipv6_stub_impl;
936
937 err = ipv6_netfilter_init();
938 if (err)
939 goto netfilter_fail;
940 /* Create /proc/foo6 entries. */
941#ifdef CONFIG_PROC_FS
942 err = -ENOMEM;
943 if (raw6_proc_init())
944 goto proc_raw6_fail;
945 if (udplite6_proc_init())
946 goto proc_udplite6_fail;
947 if (ipv6_misc_proc_init())
948 goto proc_misc6_fail;
949 if (if6_proc_init())
950 goto proc_if6_fail;
951#endif
952 err = ip6_route_init();
953 if (err)
954 goto ip6_route_fail;
955 err = ndisc_late_init();
956 if (err)
957 goto ndisc_late_fail;
958 err = ip6_flowlabel_init();
959 if (err)
960 goto ip6_flowlabel_fail;
961 err = addrconf_init();
962 if (err)
963 goto addrconf_fail;
964
965 /* Init v6 extension headers. */
966 err = ipv6_exthdrs_init();
967 if (err)
968 goto ipv6_exthdrs_fail;
969
970 err = ipv6_frag_init();
971 if (err)
972 goto ipv6_frag_fail;
973
974 /* Init v6 transport protocols. */
975 err = udpv6_init();
976 if (err)
977 goto udpv6_fail;
978
979 err = udplitev6_init();
980 if (err)
981 goto udplitev6_fail;
982
983 err = udpv6_offload_init();
984 if (err)
985 goto udpv6_offload_fail;
986
987 err = tcpv6_init();
988 if (err)
989 goto tcpv6_fail;
990
991 err = ipv6_packet_init();
992 if (err)
993 goto ipv6_packet_fail;
994
995 err = pingv6_init();
996 if (err)
997 goto pingv6_fail;
998
999 err = calipso_init();
1000 if (err)
1001 goto calipso_fail;
1002
1003 err = seg6_init();
1004 if (err)
1005 goto seg6_fail;
1006
1007#ifdef CONFIG_SYSCTL
1008 err = ipv6_sysctl_register();
1009 if (err)
1010 goto sysctl_fail;
1011#endif
1012out:
1013 return err;
1014
1015#ifdef CONFIG_SYSCTL
1016sysctl_fail:
1017 seg6_exit();
1018#endif
1019seg6_fail:
1020 calipso_exit();
1021calipso_fail:
1022 pingv6_exit();
1023pingv6_fail:
1024 ipv6_packet_cleanup();
1025ipv6_packet_fail:
1026 tcpv6_exit();
1027tcpv6_fail:
1028 udpv6_offload_exit();
1029udpv6_offload_fail:
1030 udplitev6_exit();
1031udplitev6_fail:
1032 udpv6_exit();
1033udpv6_fail:
1034 ipv6_frag_exit();
1035ipv6_frag_fail:
1036 ipv6_exthdrs_exit();
1037ipv6_exthdrs_fail:
1038 addrconf_cleanup();
1039addrconf_fail:
1040 ip6_flowlabel_cleanup();
1041ip6_flowlabel_fail:
1042 ndisc_late_cleanup();
1043ndisc_late_fail:
1044 ip6_route_cleanup();
1045ip6_route_fail:
1046#ifdef CONFIG_PROC_FS
1047 if6_proc_exit();
1048proc_if6_fail:
1049 ipv6_misc_proc_exit();
1050proc_misc6_fail:
1051 udplite6_proc_exit();
1052proc_udplite6_fail:
1053 raw6_proc_exit();
1054proc_raw6_fail:
1055#endif
1056 ipv6_netfilter_fini();
1057netfilter_fail:
1058 igmp6_cleanup();
1059igmp_fail:
1060 ndisc_cleanup();
1061ndisc_fail:
1062 ip6_mr_cleanup();
1063ipmr_fail:
1064 icmpv6_cleanup();
1065icmp_fail:
1066 unregister_pernet_subsys(&inet6_net_ops);
1067register_pernet_fail:
1068 sock_unregister(PF_INET6);
1069 rtnl_unregister_all(PF_INET6);
1070out_sock_register_fail:
1071 rawv6_exit();
1072out_unregister_ping_proto:
1073 proto_unregister(&pingv6_prot);
1074out_unregister_raw_proto:
1075 proto_unregister(&rawv6_prot);
1076out_unregister_udplite_proto:
1077 proto_unregister(&udplitev6_prot);
1078out_unregister_udp_proto:
1079 proto_unregister(&udpv6_prot);
1080out_unregister_tcp_proto:
1081 proto_unregister(&tcpv6_prot);
1082 goto out;
1083}
1084module_init(inet6_init);
1085
1086MODULE_ALIAS_NETPROTO(PF_INET6);