x86/tracing: Disentangle pagefault and resched IPI tracing key
[linux-stable.git] / net / ipv6 / af_inet6.c
bloba88b5b5b79558948dcbad84cbdda99c1d574102e
1 /*
2 * PF_INET6 socket protocol family
3 * Linux INET6 implementation
5 * Authors:
6 * Pedro Roque <roque@di.fc.ul.pt>
8 * Adapted from linux/net/ipv4/af_inet.c
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
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.
21 #define pr_fmt(fmt) "IPv6: " fmt
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>
42 #include <linux/inet.h>
43 #include <linux/netdevice.h>
44 #include <linux/icmpv6.h>
45 #include <linux/netfilter_ipv6.h>
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>
66 #include <linux/uaccess.h>
67 #include <linux/mroute6.h>
69 #include "ip6_offload.h"
71 MODULE_AUTHOR("Cast of dozens");
72 MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
73 MODULE_LICENSE("GPL");
75 /* The inetsw6 table contains everything that inet6_create needs to
76 * build a new socket.
78 static struct list_head inetsw6[SOCK_MAX];
79 static DEFINE_SPINLOCK(inetsw6_lock);
81 struct ipv6_params ipv6_defaults = {
82 .disable_ipv6 = 0,
83 .autoconf = 1,
86 static int disable_ipv6_mod;
88 module_param_named(disable, disable_ipv6_mod, int, 0444);
89 MODULE_PARM_DESC(disable, "Disable IPv6 module such that it is non-functional");
91 module_param_named(disable_ipv6, ipv6_defaults.disable_ipv6, int, 0444);
92 MODULE_PARM_DESC(disable_ipv6, "Disable IPv6 on all interfaces");
94 module_param_named(autoconf, ipv6_defaults.autoconf, int, 0444);
95 MODULE_PARM_DESC(autoconf, "Enable IPv6 address autoconfiguration on all interfaces");
97 bool ipv6_mod_enabled(void)
99 return disable_ipv6_mod == 0;
101 EXPORT_SYMBOL_GPL(ipv6_mod_enabled);
103 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
105 const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
107 return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
110 static int inet6_create(struct net *net, struct socket *sock, int protocol,
111 int kern)
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;
122 if (protocol < 0 || protocol >= IPPROTO_MAX)
123 return -EINVAL;
125 /* Look for the requested type/protocol pair. */
126 lookup_protocol:
127 err = -ESOCKTNOSUPPORT;
128 rcu_read_lock();
129 list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
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;
142 if (IPPROTO_IP == answer->protocol)
143 break;
145 err = -EPROTONOSUPPORT;
148 if (err) {
149 if (try_loading_module < 2) {
150 rcu_read_unlock();
152 * Be more specific, e.g. net-pf-10-proto-132-type-1
153 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
155 if (++try_loading_module == 1)
156 request_module("net-pf-%d-proto-%d-type-%d",
157 PF_INET6, protocol, sock->type);
159 * Fall back to generic, e.g. net-pf-10-proto-132
160 * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
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;
170 err = -EPERM;
171 if (sock->type == SOCK_RAW && !kern &&
172 !ns_capable(net->user_ns, CAP_NET_RAW))
173 goto out_rcu_unlock;
175 sock->ops = answer->ops;
176 answer_prot = answer->prot;
177 answer_flags = answer->flags;
178 rcu_read_unlock();
180 WARN_ON(!answer_prot->slab);
182 err = -ENOBUFS;
183 sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot, kern);
184 if (!sk)
185 goto out;
187 sock_init_data(sock, sk);
189 err = 0;
190 if (INET_PROTOSW_REUSE & answer_flags)
191 sk->sk_reuse = SK_CAN_REUSE;
193 inet = inet_sk(sk);
194 inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
196 if (SOCK_RAW == sock->type) {
197 inet->inet_num = protocol;
198 if (IPPROTO_RAW == protocol)
199 inet->hdrincl = 1;
202 sk->sk_destruct = inet_sock_destruct;
203 sk->sk_family = PF_INET6;
204 sk->sk_protocol = protocol;
206 sk->sk_backlog_rcv = answer->prot->backlog_rcv;
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;
216 /* Init the ipv4 part of the socket since we can have sockets
217 * using v6 API for ipv4.
219 inet->uc_ttl = -1;
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;
227 if (net->ipv4.sysctl_ip_no_pmtu_disc)
228 inet->pmtudisc = IP_PMTUDISC_DONT;
229 else
230 inet->pmtudisc = IP_PMTUDISC_WANT;
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.
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
240 sk_refcnt_debug_inc(sk);
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.
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;
254 if (sk->sk_prot->init) {
255 err = sk->sk_prot->init(sk);
256 if (err) {
257 sk_common_release(sk);
258 goto out;
262 if (!kern) {
263 err = BPF_CGROUP_RUN_PROG_INET_SOCK(sk);
264 if (err) {
265 sk_common_release(sk);
266 goto out;
269 out:
270 return err;
271 out_rcu_unlock:
272 rcu_read_unlock();
273 goto out;
277 /* bind for INET6 API */
278 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
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;
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);
294 if (addr_len < SIN6_LEN_RFC2133)
295 return -EINVAL;
297 if (addr->sin6_family != AF_INET6)
298 return -EAFNOSUPPORT;
300 addr_type = ipv6_addr_type(&addr->sin6_addr);
301 if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
302 return -EINVAL;
304 snum = ntohs(addr->sin6_port);
305 if (snum && snum < inet_prot_sock(net) &&
306 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
307 return -EACCES;
309 lock_sock(sk);
311 /* Check these errors (active socket, double bind). */
312 if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
313 err = -EINVAL;
314 goto out;
317 /* Check if the address belongs to the host. */
318 if (addr_type == IPV6_ADDR_MAPPED) {
319 int chk_addr_ret;
321 /* Binding to v4-mapped address on a v6-only socket
322 * makes no sense
324 if (sk->sk_ipv6only) {
325 err = -EINVAL;
326 goto out;
329 /* Reproduce AF_INET checks to make the bindings consistent */
330 v4addr = addr->sin6_addr.s6_addr32[3];
331 chk_addr_ret = inet_addr_type(net, v4addr);
332 if (!net->ipv4.sysctl_ip_nonlocal_bind &&
333 !(inet->freebind || inet->transparent) &&
334 v4addr != htonl(INADDR_ANY) &&
335 chk_addr_ret != RTN_LOCAL &&
336 chk_addr_ret != RTN_MULTICAST &&
337 chk_addr_ret != RTN_BROADCAST) {
338 err = -EADDRNOTAVAIL;
339 goto out;
341 } else {
342 if (addr_type != IPV6_ADDR_ANY) {
343 struct net_device *dev = NULL;
345 rcu_read_lock();
346 if (__ipv6_addr_needs_scope_id(addr_type)) {
347 if (addr_len >= sizeof(struct sockaddr_in6) &&
348 addr->sin6_scope_id) {
349 /* Override any existing binding, if another one
350 * is supplied by user.
352 sk->sk_bound_dev_if = addr->sin6_scope_id;
355 /* Binding to link-local address requires an interface */
356 if (!sk->sk_bound_dev_if) {
357 err = -EINVAL;
358 goto out_unlock;
360 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
361 if (!dev) {
362 err = -ENODEV;
363 goto out_unlock;
367 /* ipv4 addr of the socket is invalid. Only the
368 * unspecified and mapped address have a v4 equivalent.
370 v4addr = LOOPBACK4_IPV6;
371 if (!(addr_type & IPV6_ADDR_MULTICAST)) {
372 if (!net->ipv6.sysctl.ip_nonlocal_bind &&
373 !(inet->freebind || inet->transparent) &&
374 !ipv6_chk_addr(net, &addr->sin6_addr,
375 dev, 0)) {
376 err = -EADDRNOTAVAIL;
377 goto out_unlock;
380 rcu_read_unlock();
384 inet->inet_rcv_saddr = v4addr;
385 inet->inet_saddr = v4addr;
387 sk->sk_v6_rcv_saddr = addr->sin6_addr;
389 if (!(addr_type & IPV6_ADDR_MULTICAST))
390 np->saddr = addr->sin6_addr;
392 /* Make sure we are allowed to bind here. */
393 if ((snum || !inet->bind_address_no_port) &&
394 sk->sk_prot->get_port(sk, snum)) {
395 inet_reset_saddr(sk);
396 err = -EADDRINUSE;
397 goto out;
400 if (addr_type != IPV6_ADDR_ANY) {
401 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
402 if (addr_type != IPV6_ADDR_MAPPED)
403 sk->sk_ipv6only = 1;
405 if (snum)
406 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
407 inet->inet_sport = htons(inet->inet_num);
408 inet->inet_dport = 0;
409 inet->inet_daddr = 0;
410 out:
411 release_sock(sk);
412 return err;
413 out_unlock:
414 rcu_read_unlock();
415 goto out;
417 EXPORT_SYMBOL(inet6_bind);
419 int inet6_release(struct socket *sock)
421 struct sock *sk = sock->sk;
423 if (!sk)
424 return -EINVAL;
426 /* Free mc lists */
427 ipv6_sock_mc_close(sk);
429 /* Free ac lists */
430 ipv6_sock_ac_close(sk);
432 return inet_release(sock);
434 EXPORT_SYMBOL(inet6_release);
436 void inet6_destroy_sock(struct sock *sk)
438 struct ipv6_pinfo *np = inet6_sk(sk);
439 struct sk_buff *skb;
440 struct ipv6_txoptions *opt;
442 /* Release rx options */
444 skb = xchg(&np->pktoptions, NULL);
445 if (skb)
446 kfree_skb(skb);
448 skb = xchg(&np->rxpmtu, NULL);
449 if (skb)
450 kfree_skb(skb);
452 /* Free flowlabels */
453 fl6_free_socklist(sk);
455 /* Free tx options */
457 opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL);
458 if (opt) {
459 atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
460 txopt_put(opt);
463 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
466 * This does both peername and sockname.
469 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
470 int *uaddr_len, int peer)
472 struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
473 struct sock *sk = sock->sk;
474 struct inet_sock *inet = inet_sk(sk);
475 struct ipv6_pinfo *np = inet6_sk(sk);
477 sin->sin6_family = AF_INET6;
478 sin->sin6_flowinfo = 0;
479 sin->sin6_scope_id = 0;
480 if (peer) {
481 if (!inet->inet_dport)
482 return -ENOTCONN;
483 if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
484 peer == 1)
485 return -ENOTCONN;
486 sin->sin6_port = inet->inet_dport;
487 sin->sin6_addr = sk->sk_v6_daddr;
488 if (np->sndflow)
489 sin->sin6_flowinfo = np->flow_label;
490 } else {
491 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
492 sin->sin6_addr = np->saddr;
493 else
494 sin->sin6_addr = sk->sk_v6_rcv_saddr;
496 sin->sin6_port = inet->inet_sport;
498 sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
499 sk->sk_bound_dev_if);
500 *uaddr_len = sizeof(*sin);
501 return 0;
503 EXPORT_SYMBOL(inet6_getname);
505 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
507 struct sock *sk = sock->sk;
508 struct net *net = sock_net(sk);
510 switch (cmd) {
511 case SIOCGSTAMP:
512 return sock_get_timestamp(sk, (struct timeval __user *)arg);
514 case SIOCGSTAMPNS:
515 return sock_get_timestampns(sk, (struct timespec __user *)arg);
517 case SIOCADDRT:
518 case SIOCDELRT:
520 return ipv6_route_ioctl(net, cmd, (void __user *)arg);
522 case SIOCSIFADDR:
523 return addrconf_add_ifaddr(net, (void __user *) arg);
524 case SIOCDIFADDR:
525 return addrconf_del_ifaddr(net, (void __user *) arg);
526 case SIOCSIFDSTADDR:
527 return addrconf_set_dstaddr(net, (void __user *) arg);
528 default:
529 if (!sk->sk_prot->ioctl)
530 return -ENOIOCTLCMD;
531 return sk->sk_prot->ioctl(sk, cmd, arg);
533 /*NOTREACHED*/
534 return 0;
536 EXPORT_SYMBOL(inet6_ioctl);
538 const struct proto_ops inet6_stream_ops = {
539 .family = PF_INET6,
540 .owner = THIS_MODULE,
541 .release = inet6_release,
542 .bind = inet6_bind,
543 .connect = inet_stream_connect, /* ok */
544 .socketpair = sock_no_socketpair, /* a do nothing */
545 .accept = inet_accept, /* ok */
546 .getname = inet6_getname,
547 .poll = tcp_poll, /* ok */
548 .ioctl = inet6_ioctl, /* must change */
549 .listen = inet_listen, /* ok */
550 .shutdown = inet_shutdown, /* ok */
551 .setsockopt = sock_common_setsockopt, /* ok */
552 .getsockopt = sock_common_getsockopt, /* ok */
553 .sendmsg = inet_sendmsg, /* ok */
554 .recvmsg = inet_recvmsg, /* ok */
555 .mmap = sock_no_mmap,
556 .sendpage = inet_sendpage,
557 .splice_read = tcp_splice_read,
558 .read_sock = tcp_read_sock,
559 .peek_len = tcp_peek_len,
560 #ifdef CONFIG_COMPAT
561 .compat_setsockopt = compat_sock_common_setsockopt,
562 .compat_getsockopt = compat_sock_common_getsockopt,
563 #endif
566 const struct proto_ops inet6_dgram_ops = {
567 .family = PF_INET6,
568 .owner = THIS_MODULE,
569 .release = inet6_release,
570 .bind = inet6_bind,
571 .connect = inet_dgram_connect, /* ok */
572 .socketpair = sock_no_socketpair, /* a do nothing */
573 .accept = sock_no_accept, /* a do nothing */
574 .getname = inet6_getname,
575 .poll = udp_poll, /* ok */
576 .ioctl = inet6_ioctl, /* must change */
577 .listen = sock_no_listen, /* ok */
578 .shutdown = inet_shutdown, /* ok */
579 .setsockopt = sock_common_setsockopt, /* ok */
580 .getsockopt = sock_common_getsockopt, /* ok */
581 .sendmsg = inet_sendmsg, /* ok */
582 .recvmsg = inet_recvmsg, /* ok */
583 .mmap = sock_no_mmap,
584 .sendpage = sock_no_sendpage,
585 .set_peek_off = sk_set_peek_off,
586 #ifdef CONFIG_COMPAT
587 .compat_setsockopt = compat_sock_common_setsockopt,
588 .compat_getsockopt = compat_sock_common_getsockopt,
589 #endif
592 static const struct net_proto_family inet6_family_ops = {
593 .family = PF_INET6,
594 .create = inet6_create,
595 .owner = THIS_MODULE,
598 int inet6_register_protosw(struct inet_protosw *p)
600 struct list_head *lh;
601 struct inet_protosw *answer;
602 struct list_head *last_perm;
603 int protocol = p->protocol;
604 int ret;
606 spin_lock_bh(&inetsw6_lock);
608 ret = -EINVAL;
609 if (p->type >= SOCK_MAX)
610 goto out_illegal;
612 /* If we are trying to override a permanent protocol, bail. */
613 answer = NULL;
614 ret = -EPERM;
615 last_perm = &inetsw6[p->type];
616 list_for_each(lh, &inetsw6[p->type]) {
617 answer = list_entry(lh, struct inet_protosw, list);
619 /* Check only the non-wild match. */
620 if (INET_PROTOSW_PERMANENT & answer->flags) {
621 if (protocol == answer->protocol)
622 break;
623 last_perm = lh;
626 answer = NULL;
628 if (answer)
629 goto out_permanent;
631 /* Add the new entry after the last permanent entry if any, so that
632 * the new entry does not override a permanent entry when matched with
633 * a wild-card protocol. But it is allowed to override any existing
634 * non-permanent entry. This means that when we remove this entry, the
635 * system automatically returns to the old behavior.
637 list_add_rcu(&p->list, last_perm);
638 ret = 0;
639 out:
640 spin_unlock_bh(&inetsw6_lock);
641 return ret;
643 out_permanent:
644 pr_err("Attempt to override permanent protocol %d\n", protocol);
645 goto out;
647 out_illegal:
648 pr_err("Ignoring attempt to register invalid socket type %d\n",
649 p->type);
650 goto out;
652 EXPORT_SYMBOL(inet6_register_protosw);
654 void
655 inet6_unregister_protosw(struct inet_protosw *p)
657 if (INET_PROTOSW_PERMANENT & p->flags) {
658 pr_err("Attempt to unregister permanent protocol %d\n",
659 p->protocol);
660 } else {
661 spin_lock_bh(&inetsw6_lock);
662 list_del_rcu(&p->list);
663 spin_unlock_bh(&inetsw6_lock);
665 synchronize_net();
668 EXPORT_SYMBOL(inet6_unregister_protosw);
670 int inet6_sk_rebuild_header(struct sock *sk)
672 struct ipv6_pinfo *np = inet6_sk(sk);
673 struct dst_entry *dst;
675 dst = __sk_dst_check(sk, np->dst_cookie);
677 if (!dst) {
678 struct inet_sock *inet = inet_sk(sk);
679 struct in6_addr *final_p, final;
680 struct flowi6 fl6;
682 memset(&fl6, 0, sizeof(fl6));
683 fl6.flowi6_proto = sk->sk_protocol;
684 fl6.daddr = sk->sk_v6_daddr;
685 fl6.saddr = np->saddr;
686 fl6.flowlabel = np->flow_label;
687 fl6.flowi6_oif = sk->sk_bound_dev_if;
688 fl6.flowi6_mark = sk->sk_mark;
689 fl6.fl6_dport = inet->inet_dport;
690 fl6.fl6_sport = inet->inet_sport;
691 fl6.flowi6_uid = sk->sk_uid;
692 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
694 rcu_read_lock();
695 final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt),
696 &final);
697 rcu_read_unlock();
699 dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
700 if (IS_ERR(dst)) {
701 sk->sk_route_caps = 0;
702 sk->sk_err_soft = -PTR_ERR(dst);
703 return PTR_ERR(dst);
706 ip6_dst_store(sk, dst, NULL, NULL);
709 return 0;
711 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
713 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb,
714 const struct inet6_skb_parm *opt)
716 const struct ipv6_pinfo *np = inet6_sk(sk);
718 if (np->rxopt.all) {
719 if (((opt->flags & IP6SKB_HOPBYHOP) &&
720 (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) ||
721 (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) &&
722 np->rxopt.bits.rxflow) ||
723 (opt->srcrt && (np->rxopt.bits.srcrt ||
724 np->rxopt.bits.osrcrt)) ||
725 ((opt->dst1 || opt->dst0) &&
726 (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
727 return true;
729 return false;
731 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
733 static struct packet_type ipv6_packet_type __read_mostly = {
734 .type = cpu_to_be16(ETH_P_IPV6),
735 .func = ipv6_rcv,
738 static int __init ipv6_packet_init(void)
740 dev_add_pack(&ipv6_packet_type);
741 return 0;
744 static void ipv6_packet_cleanup(void)
746 dev_remove_pack(&ipv6_packet_type);
749 static int __net_init ipv6_init_mibs(struct net *net)
751 int i;
753 net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib);
754 if (!net->mib.udp_stats_in6)
755 return -ENOMEM;
756 net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib);
757 if (!net->mib.udplite_stats_in6)
758 goto err_udplite_mib;
759 net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib);
760 if (!net->mib.ipv6_statistics)
761 goto err_ip_mib;
763 for_each_possible_cpu(i) {
764 struct ipstats_mib *af_inet6_stats;
765 af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i);
766 u64_stats_init(&af_inet6_stats->syncp);
770 net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib);
771 if (!net->mib.icmpv6_statistics)
772 goto err_icmp_mib;
773 net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
774 GFP_KERNEL);
775 if (!net->mib.icmpv6msg_statistics)
776 goto err_icmpmsg_mib;
777 return 0;
779 err_icmpmsg_mib:
780 free_percpu(net->mib.icmpv6_statistics);
781 err_icmp_mib:
782 free_percpu(net->mib.ipv6_statistics);
783 err_ip_mib:
784 free_percpu(net->mib.udplite_stats_in6);
785 err_udplite_mib:
786 free_percpu(net->mib.udp_stats_in6);
787 return -ENOMEM;
790 static void ipv6_cleanup_mibs(struct net *net)
792 free_percpu(net->mib.udp_stats_in6);
793 free_percpu(net->mib.udplite_stats_in6);
794 free_percpu(net->mib.ipv6_statistics);
795 free_percpu(net->mib.icmpv6_statistics);
796 kfree(net->mib.icmpv6msg_statistics);
799 static int __net_init inet6_net_init(struct net *net)
801 int err = 0;
803 net->ipv6.sysctl.bindv6only = 0;
804 net->ipv6.sysctl.icmpv6_time = 1*HZ;
805 net->ipv6.sysctl.flowlabel_consistency = 1;
806 net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS;
807 net->ipv6.sysctl.idgen_retries = 3;
808 net->ipv6.sysctl.idgen_delay = 1 * HZ;
809 net->ipv6.sysctl.flowlabel_state_ranges = 0;
810 atomic_set(&net->ipv6.fib6_sernum, 1);
812 err = ipv6_init_mibs(net);
813 if (err)
814 return err;
815 #ifdef CONFIG_PROC_FS
816 err = udp6_proc_init(net);
817 if (err)
818 goto out;
819 err = tcp6_proc_init(net);
820 if (err)
821 goto proc_tcp6_fail;
822 err = ac6_proc_init(net);
823 if (err)
824 goto proc_ac6_fail;
825 #endif
826 return err;
828 #ifdef CONFIG_PROC_FS
829 proc_ac6_fail:
830 tcp6_proc_exit(net);
831 proc_tcp6_fail:
832 udp6_proc_exit(net);
833 out:
834 ipv6_cleanup_mibs(net);
835 return err;
836 #endif
839 static void __net_exit inet6_net_exit(struct net *net)
841 #ifdef CONFIG_PROC_FS
842 udp6_proc_exit(net);
843 tcp6_proc_exit(net);
844 ac6_proc_exit(net);
845 #endif
846 ipv6_cleanup_mibs(net);
849 static struct pernet_operations inet6_net_ops = {
850 .init = inet6_net_init,
851 .exit = inet6_net_exit,
854 static const struct ipv6_stub ipv6_stub_impl = {
855 .ipv6_sock_mc_join = ipv6_sock_mc_join,
856 .ipv6_sock_mc_drop = ipv6_sock_mc_drop,
857 .ipv6_dst_lookup = ip6_dst_lookup,
858 .udpv6_encap_enable = udpv6_encap_enable,
859 .ndisc_send_na = ndisc_send_na,
860 .nd_tbl = &nd_tbl,
863 static int __init inet6_init(void)
865 struct list_head *r;
866 int err = 0;
868 sock_skb_cb_check_size(sizeof(struct inet6_skb_parm));
870 /* Register the socket-side information for inet6_create. */
871 for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
872 INIT_LIST_HEAD(r);
874 if (disable_ipv6_mod) {
875 pr_info("Loaded, but administratively disabled, reboot required to enable\n");
876 goto out;
879 err = proto_register(&tcpv6_prot, 1);
880 if (err)
881 goto out;
883 err = proto_register(&udpv6_prot, 1);
884 if (err)
885 goto out_unregister_tcp_proto;
887 err = proto_register(&udplitev6_prot, 1);
888 if (err)
889 goto out_unregister_udp_proto;
891 err = proto_register(&rawv6_prot, 1);
892 if (err)
893 goto out_unregister_udplite_proto;
895 err = proto_register(&pingv6_prot, 1);
896 if (err)
897 goto out_unregister_ping_proto;
899 /* We MUST register RAW sockets before we create the ICMP6,
900 * IGMP6, or NDISC control sockets.
902 err = rawv6_init();
903 if (err)
904 goto out_unregister_raw_proto;
906 /* Register the family here so that the init calls below will
907 * be able to create sockets. (?? is this dangerous ??)
909 err = sock_register(&inet6_family_ops);
910 if (err)
911 goto out_sock_register_fail;
914 * ipngwg API draft makes clear that the correct semantics
915 * for TCP and UDP is to consider one TCP and UDP instance
916 * in a host available by both INET and INET6 APIs and
917 * able to communicate via both network protocols.
920 err = register_pernet_subsys(&inet6_net_ops);
921 if (err)
922 goto register_pernet_fail;
923 err = ip6_mr_init();
924 if (err)
925 goto ipmr_fail;
926 err = icmpv6_init();
927 if (err)
928 goto icmp_fail;
929 err = ndisc_init();
930 if (err)
931 goto ndisc_fail;
932 err = igmp6_init();
933 if (err)
934 goto igmp_fail;
936 err = ipv6_netfilter_init();
937 if (err)
938 goto netfilter_fail;
939 /* Create /proc/foo6 entries. */
940 #ifdef CONFIG_PROC_FS
941 err = -ENOMEM;
942 if (raw6_proc_init())
943 goto proc_raw6_fail;
944 if (udplite6_proc_init())
945 goto proc_udplite6_fail;
946 if (ipv6_misc_proc_init())
947 goto proc_misc6_fail;
948 if (if6_proc_init())
949 goto proc_if6_fail;
950 #endif
951 err = ip6_route_init();
952 if (err)
953 goto ip6_route_fail;
954 err = ndisc_late_init();
955 if (err)
956 goto ndisc_late_fail;
957 err = ip6_flowlabel_init();
958 if (err)
959 goto ip6_flowlabel_fail;
960 err = addrconf_init();
961 if (err)
962 goto addrconf_fail;
964 /* Init v6 extension headers. */
965 err = ipv6_exthdrs_init();
966 if (err)
967 goto ipv6_exthdrs_fail;
969 err = ipv6_frag_init();
970 if (err)
971 goto ipv6_frag_fail;
973 /* Init v6 transport protocols. */
974 err = udpv6_init();
975 if (err)
976 goto udpv6_fail;
978 err = udplitev6_init();
979 if (err)
980 goto udplitev6_fail;
982 err = udpv6_offload_init();
983 if (err)
984 goto udpv6_offload_fail;
986 err = tcpv6_init();
987 if (err)
988 goto tcpv6_fail;
990 err = ipv6_packet_init();
991 if (err)
992 goto ipv6_packet_fail;
994 err = pingv6_init();
995 if (err)
996 goto pingv6_fail;
998 err = calipso_init();
999 if (err)
1000 goto calipso_fail;
1002 err = seg6_init();
1003 if (err)
1004 goto seg6_fail;
1006 err = igmp6_late_init();
1007 if (err)
1008 goto igmp6_late_err;
1010 #ifdef CONFIG_SYSCTL
1011 err = ipv6_sysctl_register();
1012 if (err)
1013 goto sysctl_fail;
1014 #endif
1016 /* ensure that ipv6 stubs are visible only after ipv6 is ready */
1017 wmb();
1018 ipv6_stub = &ipv6_stub_impl;
1019 out:
1020 return err;
1022 #ifdef CONFIG_SYSCTL
1023 sysctl_fail:
1024 igmp6_late_cleanup();
1025 #endif
1026 igmp6_late_err:
1027 seg6_exit();
1028 seg6_fail:
1029 calipso_exit();
1030 calipso_fail:
1031 pingv6_exit();
1032 pingv6_fail:
1033 ipv6_packet_cleanup();
1034 ipv6_packet_fail:
1035 tcpv6_exit();
1036 tcpv6_fail:
1037 udpv6_offload_exit();
1038 udpv6_offload_fail:
1039 udplitev6_exit();
1040 udplitev6_fail:
1041 udpv6_exit();
1042 udpv6_fail:
1043 ipv6_frag_exit();
1044 ipv6_frag_fail:
1045 ipv6_exthdrs_exit();
1046 ipv6_exthdrs_fail:
1047 addrconf_cleanup();
1048 addrconf_fail:
1049 ip6_flowlabel_cleanup();
1050 ip6_flowlabel_fail:
1051 ndisc_late_cleanup();
1052 ndisc_late_fail:
1053 ip6_route_cleanup();
1054 ip6_route_fail:
1055 #ifdef CONFIG_PROC_FS
1056 if6_proc_exit();
1057 proc_if6_fail:
1058 ipv6_misc_proc_exit();
1059 proc_misc6_fail:
1060 udplite6_proc_exit();
1061 proc_udplite6_fail:
1062 raw6_proc_exit();
1063 proc_raw6_fail:
1064 #endif
1065 ipv6_netfilter_fini();
1066 netfilter_fail:
1067 igmp6_cleanup();
1068 igmp_fail:
1069 ndisc_cleanup();
1070 ndisc_fail:
1071 ip6_mr_cleanup();
1072 icmp_fail:
1073 unregister_pernet_subsys(&inet6_net_ops);
1074 ipmr_fail:
1075 icmpv6_cleanup();
1076 register_pernet_fail:
1077 sock_unregister(PF_INET6);
1078 rtnl_unregister_all(PF_INET6);
1079 out_sock_register_fail:
1080 rawv6_exit();
1081 out_unregister_ping_proto:
1082 proto_unregister(&pingv6_prot);
1083 out_unregister_raw_proto:
1084 proto_unregister(&rawv6_prot);
1085 out_unregister_udplite_proto:
1086 proto_unregister(&udplitev6_prot);
1087 out_unregister_udp_proto:
1088 proto_unregister(&udpv6_prot);
1089 out_unregister_tcp_proto:
1090 proto_unregister(&tcpv6_prot);
1091 goto out;
1093 module_init(inet6_init);
1095 MODULE_ALIAS_NETPROTO(PF_INET6);