1 // SPDX-License-Identifier: GPL-2.0-or-later
3 * PF_INET6 socket protocol family
4 * Linux INET6 implementation
7 * Pedro Roque <roque@di.fc.ul.pt>
9 * Adapted from linux/net/ipv4/af_inet.c
12 * piggy, Karl Knutson : Socket protocol table
13 * Hideaki YOSHIFUJI : sin6_scope_id support
14 * Arnaldo Melo : check proc_net_create return, cleanups
17 #define pr_fmt(fmt) "IPv6: " fmt
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>
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>
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>
38 #include <linux/inet.h>
39 #include <linux/netdevice.h>
40 #include <linux/icmpv6.h>
41 #include <linux/netfilter_ipv6.h>
46 #include <net/udplite.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>
60 #include <net/calipso.h>
63 #include <net/compat.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
78 static struct list_head inetsw6[SOCK_MAX];
79 static DEFINE_SPINLOCK(inetsw6_lock);
81 struct ipv6_params ipv6_defaults = {
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,
113 struct inet_sock *inet;
114 struct ipv6_pinfo *np;
116 struct inet_protosw *answer;
117 struct proto *answer_prot;
118 unsigned char answer_flags;
119 int try_loading_module = 0;
122 if (protocol < 0 || protocol >= IPPROTO_MAX)
125 /* Look for the requested type/protocol pair. */
127 err = -ESOCKTNOSUPPORT;
129 list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
132 /* Check the non-wild match. */
133 if (protocol == answer->protocol) {
134 if (protocol != IPPROTO_IP)
137 /* Check for the two wild cases. */
138 if (IPPROTO_IP == protocol) {
139 protocol = answer->protocol;
142 if (IPPROTO_IP == answer->protocol)
145 err = -EPROTONOSUPPORT;
149 if (try_loading_module < 2) {
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)
163 request_module("net-pf-%d-proto-%d",
165 goto lookup_protocol;
171 if (sock->type == SOCK_RAW && !kern &&
172 !ns_capable(net->user_ns, CAP_NET_RAW))
175 sock->ops = answer->ops;
176 answer_prot = answer->prot;
177 answer_flags = answer->flags;
180 WARN_ON(!answer_prot->slab);
183 sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot, kern);
187 sock_init_data(sock, sk);
190 if (INET_PROTOSW_REUSE & answer_flags)
191 sk->sk_reuse = SK_CAN_REUSE;
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)
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);
210 np->mcast_hops = IPV6_DEFAULT_MCASTHOPS;
213 np->pmtudisc = IPV6_PMTUDISC_WANT;
214 np->repflow = net->ipv6.sysctl.flowlabel_reflect & FLOWLABEL_REFLECT_ESTABLISHED;
215 sk->sk_ipv6only = net->ipv6.sysctl.bindv6only;
217 /* Init the ipv4 part of the socket since we can have sockets
218 * using v6 API for ipv4.
225 inet->mc_list = NULL;
228 if (net->ipv4.sysctl_ip_no_pmtu_disc)
229 inet->pmtudisc = IP_PMTUDISC_DONT;
231 inet->pmtudisc = IP_PMTUDISC_WANT;
233 * Increment only the relevant sk_prot->socks debug field, this changes
234 * the previous behaviour of incrementing both the equivalent to
235 * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
237 * This allows better debug granularity as we'll know exactly how many
238 * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
239 * transport protocol socks. -acme
241 sk_refcnt_debug_inc(sk);
243 if (inet->inet_num) {
244 /* It assumes that any protocol which allows
245 * the user to assign a number at socket
246 * creation time automatically shares.
248 inet->inet_sport = htons(inet->inet_num);
249 err = sk->sk_prot->hash(sk);
251 sk_common_release(sk);
255 if (sk->sk_prot->init) {
256 err = sk->sk_prot->init(sk);
258 sk_common_release(sk);
264 err = BPF_CGROUP_RUN_PROG_INET_SOCK(sk);
266 sk_common_release(sk);
277 static int __inet6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len,
280 struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr;
281 struct inet_sock *inet = inet_sk(sk);
282 struct ipv6_pinfo *np = inet6_sk(sk);
283 struct net *net = sock_net(sk);
290 if (addr->sin6_family != AF_INET6)
291 return -EAFNOSUPPORT;
293 addr_type = ipv6_addr_type(&addr->sin6_addr);
294 if ((addr_type & IPV6_ADDR_MULTICAST) && sk->sk_type == SOCK_STREAM)
297 snum = ntohs(addr->sin6_port);
298 if (!(flags & BIND_NO_CAP_NET_BIND_SERVICE) &&
299 snum && inet_port_requires_bind_service(net, snum) &&
300 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
303 if (flags & BIND_WITH_LOCK)
306 /* Check these errors (active socket, double bind). */
307 if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
312 /* Check if the address belongs to the host. */
313 if (addr_type == IPV6_ADDR_MAPPED) {
314 struct net_device *dev = NULL;
317 /* Binding to v4-mapped address on a v6-only socket
320 if (sk->sk_ipv6only) {
326 if (sk->sk_bound_dev_if) {
327 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
334 /* Reproduce AF_INET checks to make the bindings consistent */
335 v4addr = addr->sin6_addr.s6_addr32[3];
336 chk_addr_ret = inet_addr_type_dev_table(net, dev, v4addr);
339 if (!inet_can_nonlocal_bind(net, inet) &&
340 v4addr != htonl(INADDR_ANY) &&
341 chk_addr_ret != RTN_LOCAL &&
342 chk_addr_ret != RTN_MULTICAST &&
343 chk_addr_ret != RTN_BROADCAST) {
344 err = -EADDRNOTAVAIL;
348 if (addr_type != IPV6_ADDR_ANY) {
349 struct net_device *dev = NULL;
352 if (__ipv6_addr_needs_scope_id(addr_type)) {
353 if (addr_len >= sizeof(struct sockaddr_in6) &&
354 addr->sin6_scope_id) {
355 /* Override any existing binding, if another one
356 * is supplied by user.
358 sk->sk_bound_dev_if = addr->sin6_scope_id;
361 /* Binding to link-local address requires an interface */
362 if (!sk->sk_bound_dev_if) {
368 if (sk->sk_bound_dev_if) {
369 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
376 /* ipv4 addr of the socket is invalid. Only the
377 * unspecified and mapped address have a v4 equivalent.
379 v4addr = LOOPBACK4_IPV6;
380 if (!(addr_type & IPV6_ADDR_MULTICAST)) {
381 if (!ipv6_can_nonlocal_bind(net, inet) &&
382 !ipv6_chk_addr(net, &addr->sin6_addr,
384 err = -EADDRNOTAVAIL;
392 inet->inet_rcv_saddr = v4addr;
393 inet->inet_saddr = v4addr;
395 sk->sk_v6_rcv_saddr = addr->sin6_addr;
397 if (!(addr_type & IPV6_ADDR_MULTICAST))
398 np->saddr = addr->sin6_addr;
400 saved_ipv6only = sk->sk_ipv6only;
401 if (addr_type != IPV6_ADDR_ANY && addr_type != IPV6_ADDR_MAPPED)
404 /* Make sure we are allowed to bind here. */
405 if (snum || !(inet->bind_address_no_port ||
406 (flags & BIND_FORCE_ADDRESS_NO_PORT))) {
407 if (sk->sk_prot->get_port(sk, snum)) {
408 sk->sk_ipv6only = saved_ipv6only;
409 inet_reset_saddr(sk);
413 if (!(flags & BIND_FROM_BPF)) {
414 err = BPF_CGROUP_RUN_PROG_INET6_POST_BIND(sk);
416 sk->sk_ipv6only = saved_ipv6only;
417 inet_reset_saddr(sk);
423 if (addr_type != IPV6_ADDR_ANY)
424 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
426 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
427 inet->inet_sport = htons(inet->inet_num);
428 inet->inet_dport = 0;
429 inet->inet_daddr = 0;
431 if (flags & BIND_WITH_LOCK)
439 /* bind for INET6 API */
440 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
442 struct sock *sk = sock->sk;
443 u32 flags = BIND_WITH_LOCK;
446 /* If the socket has its own bind function then use it. */
447 if (sk->sk_prot->bind)
448 return sk->sk_prot->bind(sk, uaddr, addr_len);
450 if (addr_len < SIN6_LEN_RFC2133)
453 /* BPF prog is run before any checks are done so that if the prog
454 * changes context in a wrong way it will be caught.
456 err = BPF_CGROUP_RUN_PROG_INET_BIND_LOCK(sk, uaddr,
457 BPF_CGROUP_INET6_BIND, &flags);
461 return __inet6_bind(sk, uaddr, addr_len, flags);
463 EXPORT_SYMBOL(inet6_bind);
465 int inet6_release(struct socket *sock)
467 struct sock *sk = sock->sk;
473 ipv6_sock_mc_close(sk);
476 ipv6_sock_ac_close(sk);
478 return inet_release(sock);
480 EXPORT_SYMBOL(inet6_release);
482 void inet6_destroy_sock(struct sock *sk)
484 struct ipv6_pinfo *np = inet6_sk(sk);
486 struct ipv6_txoptions *opt;
488 /* Release rx options */
490 skb = xchg(&np->pktoptions, NULL);
493 skb = xchg(&np->rxpmtu, NULL);
496 /* Free flowlabels */
497 fl6_free_socklist(sk);
499 /* Free tx options */
501 opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL);
503 atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
507 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
510 * This does both peername and sockname.
512 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
515 struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
516 struct sock *sk = sock->sk;
517 struct inet_sock *inet = inet_sk(sk);
518 struct ipv6_pinfo *np = inet6_sk(sk);
520 sin->sin6_family = AF_INET6;
521 sin->sin6_flowinfo = 0;
522 sin->sin6_scope_id = 0;
524 if (!inet->inet_dport)
526 if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
529 sin->sin6_port = inet->inet_dport;
530 sin->sin6_addr = sk->sk_v6_daddr;
532 sin->sin6_flowinfo = np->flow_label;
533 BPF_CGROUP_RUN_SA_PROG_LOCK(sk, (struct sockaddr *)sin,
534 BPF_CGROUP_INET6_GETPEERNAME,
537 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
538 sin->sin6_addr = np->saddr;
540 sin->sin6_addr = sk->sk_v6_rcv_saddr;
541 sin->sin6_port = inet->inet_sport;
542 BPF_CGROUP_RUN_SA_PROG_LOCK(sk, (struct sockaddr *)sin,
543 BPF_CGROUP_INET6_GETSOCKNAME,
546 sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
547 sk->sk_bound_dev_if);
550 EXPORT_SYMBOL(inet6_getname);
552 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
554 void __user *argp = (void __user *)arg;
555 struct sock *sk = sock->sk;
556 struct net *net = sock_net(sk);
561 struct in6_rtmsg rtmsg;
563 if (copy_from_user(&rtmsg, argp, sizeof(rtmsg)))
565 return ipv6_route_ioctl(net, cmd, &rtmsg);
568 return addrconf_add_ifaddr(net, argp);
570 return addrconf_del_ifaddr(net, argp);
572 return addrconf_set_dstaddr(net, argp);
574 if (!sk->sk_prot->ioctl)
576 return sk->sk_prot->ioctl(sk, cmd, arg);
581 EXPORT_SYMBOL(inet6_ioctl);
584 struct compat_in6_rtmsg {
585 struct in6_addr rtmsg_dst;
586 struct in6_addr rtmsg_src;
587 struct in6_addr rtmsg_gateway;
597 static int inet6_compat_routing_ioctl(struct sock *sk, unsigned int cmd,
598 struct compat_in6_rtmsg __user *ur)
602 if (copy_from_user(&rt.rtmsg_dst, &ur->rtmsg_dst,
603 3 * sizeof(struct in6_addr)) ||
604 get_user(rt.rtmsg_type, &ur->rtmsg_type) ||
605 get_user(rt.rtmsg_dst_len, &ur->rtmsg_dst_len) ||
606 get_user(rt.rtmsg_src_len, &ur->rtmsg_src_len) ||
607 get_user(rt.rtmsg_metric, &ur->rtmsg_metric) ||
608 get_user(rt.rtmsg_info, &ur->rtmsg_info) ||
609 get_user(rt.rtmsg_flags, &ur->rtmsg_flags) ||
610 get_user(rt.rtmsg_ifindex, &ur->rtmsg_ifindex))
614 return ipv6_route_ioctl(sock_net(sk), cmd, &rt);
617 int inet6_compat_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
619 void __user *argp = compat_ptr(arg);
620 struct sock *sk = sock->sk;
625 return inet6_compat_routing_ioctl(sk, cmd, argp);
630 EXPORT_SYMBOL_GPL(inet6_compat_ioctl);
631 #endif /* CONFIG_COMPAT */
633 INDIRECT_CALLABLE_DECLARE(int udpv6_sendmsg(struct sock *, struct msghdr *,
635 int inet6_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
637 struct sock *sk = sock->sk;
639 if (unlikely(inet_send_prepare(sk)))
642 return INDIRECT_CALL_2(sk->sk_prot->sendmsg, tcp_sendmsg, udpv6_sendmsg,
646 INDIRECT_CALLABLE_DECLARE(int udpv6_recvmsg(struct sock *, struct msghdr *,
647 size_t, int, int, int *));
648 int inet6_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
651 struct sock *sk = sock->sk;
655 if (likely(!(flags & MSG_ERRQUEUE)))
656 sock_rps_record_flow(sk);
658 err = INDIRECT_CALL_2(sk->sk_prot->recvmsg, tcp_recvmsg, udpv6_recvmsg,
659 sk, msg, size, flags & MSG_DONTWAIT,
660 flags & ~MSG_DONTWAIT, &addr_len);
662 msg->msg_namelen = addr_len;
666 const struct proto_ops inet6_stream_ops = {
668 .owner = THIS_MODULE,
669 .release = inet6_release,
671 .connect = inet_stream_connect, /* ok */
672 .socketpair = sock_no_socketpair, /* a do nothing */
673 .accept = inet_accept, /* ok */
674 .getname = inet6_getname,
675 .poll = tcp_poll, /* ok */
676 .ioctl = inet6_ioctl, /* must change */
677 .gettstamp = sock_gettstamp,
678 .listen = inet_listen, /* ok */
679 .shutdown = inet_shutdown, /* ok */
680 .setsockopt = sock_common_setsockopt, /* ok */
681 .getsockopt = sock_common_getsockopt, /* ok */
682 .sendmsg = inet6_sendmsg, /* retpoline's sake */
683 .recvmsg = inet6_recvmsg, /* retpoline's sake */
687 .sendpage = inet_sendpage,
688 .sendmsg_locked = tcp_sendmsg_locked,
689 .sendpage_locked = tcp_sendpage_locked,
690 .splice_read = tcp_splice_read,
691 .read_sock = tcp_read_sock,
692 .peek_len = tcp_peek_len,
694 .compat_ioctl = inet6_compat_ioctl,
696 .set_rcvlowat = tcp_set_rcvlowat,
699 const struct proto_ops inet6_dgram_ops = {
701 .owner = THIS_MODULE,
702 .release = inet6_release,
704 .connect = inet_dgram_connect, /* ok */
705 .socketpair = sock_no_socketpair, /* a do nothing */
706 .accept = sock_no_accept, /* a do nothing */
707 .getname = inet6_getname,
708 .poll = udp_poll, /* ok */
709 .ioctl = inet6_ioctl, /* must change */
710 .gettstamp = sock_gettstamp,
711 .listen = sock_no_listen, /* ok */
712 .shutdown = inet_shutdown, /* ok */
713 .setsockopt = sock_common_setsockopt, /* ok */
714 .getsockopt = sock_common_getsockopt, /* ok */
715 .sendmsg = inet6_sendmsg, /* retpoline's sake */
716 .recvmsg = inet6_recvmsg, /* retpoline's sake */
717 .mmap = sock_no_mmap,
718 .sendpage = sock_no_sendpage,
719 .set_peek_off = sk_set_peek_off,
721 .compat_ioctl = inet6_compat_ioctl,
725 static const struct net_proto_family inet6_family_ops = {
727 .create = inet6_create,
728 .owner = THIS_MODULE,
731 int inet6_register_protosw(struct inet_protosw *p)
733 struct list_head *lh;
734 struct inet_protosw *answer;
735 struct list_head *last_perm;
736 int protocol = p->protocol;
739 spin_lock_bh(&inetsw6_lock);
742 if (p->type >= SOCK_MAX)
745 /* If we are trying to override a permanent protocol, bail. */
748 last_perm = &inetsw6[p->type];
749 list_for_each(lh, &inetsw6[p->type]) {
750 answer = list_entry(lh, struct inet_protosw, list);
752 /* Check only the non-wild match. */
753 if (INET_PROTOSW_PERMANENT & answer->flags) {
754 if (protocol == answer->protocol)
764 /* Add the new entry after the last permanent entry if any, so that
765 * the new entry does not override a permanent entry when matched with
766 * a wild-card protocol. But it is allowed to override any existing
767 * non-permanent entry. This means that when we remove this entry, the
768 * system automatically returns to the old behavior.
770 list_add_rcu(&p->list, last_perm);
773 spin_unlock_bh(&inetsw6_lock);
777 pr_err("Attempt to override permanent protocol %d\n", protocol);
781 pr_err("Ignoring attempt to register invalid socket type %d\n",
785 EXPORT_SYMBOL(inet6_register_protosw);
788 inet6_unregister_protosw(struct inet_protosw *p)
790 if (INET_PROTOSW_PERMANENT & p->flags) {
791 pr_err("Attempt to unregister permanent protocol %d\n",
794 spin_lock_bh(&inetsw6_lock);
795 list_del_rcu(&p->list);
796 spin_unlock_bh(&inetsw6_lock);
801 EXPORT_SYMBOL(inet6_unregister_protosw);
803 int inet6_sk_rebuild_header(struct sock *sk)
805 struct ipv6_pinfo *np = inet6_sk(sk);
806 struct dst_entry *dst;
808 dst = __sk_dst_check(sk, np->dst_cookie);
811 struct inet_sock *inet = inet_sk(sk);
812 struct in6_addr *final_p, final;
815 memset(&fl6, 0, sizeof(fl6));
816 fl6.flowi6_proto = sk->sk_protocol;
817 fl6.daddr = sk->sk_v6_daddr;
818 fl6.saddr = np->saddr;
819 fl6.flowlabel = np->flow_label;
820 fl6.flowi6_oif = sk->sk_bound_dev_if;
821 fl6.flowi6_mark = sk->sk_mark;
822 fl6.fl6_dport = inet->inet_dport;
823 fl6.fl6_sport = inet->inet_sport;
824 fl6.flowi6_uid = sk->sk_uid;
825 security_sk_classify_flow(sk, flowi6_to_flowi_common(&fl6));
828 final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt),
832 dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p);
834 sk->sk_route_caps = 0;
835 sk->sk_err_soft = -PTR_ERR(dst);
839 ip6_dst_store(sk, dst, NULL, NULL);
844 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
846 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb,
847 const struct inet6_skb_parm *opt)
849 const struct ipv6_pinfo *np = inet6_sk(sk);
852 if (((opt->flags & IP6SKB_HOPBYHOP) &&
853 (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) ||
854 (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) &&
855 np->rxopt.bits.rxflow) ||
856 (opt->srcrt && (np->rxopt.bits.srcrt ||
857 np->rxopt.bits.osrcrt)) ||
858 ((opt->dst1 || opt->dst0) &&
859 (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
864 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
866 static struct packet_type ipv6_packet_type __read_mostly = {
867 .type = cpu_to_be16(ETH_P_IPV6),
869 .list_func = ipv6_list_rcv,
872 static int __init ipv6_packet_init(void)
874 dev_add_pack(&ipv6_packet_type);
878 static void ipv6_packet_cleanup(void)
880 dev_remove_pack(&ipv6_packet_type);
883 static int __net_init ipv6_init_mibs(struct net *net)
887 net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib);
888 if (!net->mib.udp_stats_in6)
890 net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib);
891 if (!net->mib.udplite_stats_in6)
892 goto err_udplite_mib;
893 net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib);
894 if (!net->mib.ipv6_statistics)
897 for_each_possible_cpu(i) {
898 struct ipstats_mib *af_inet6_stats;
899 af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i);
900 u64_stats_init(&af_inet6_stats->syncp);
904 net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib);
905 if (!net->mib.icmpv6_statistics)
907 net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
909 if (!net->mib.icmpv6msg_statistics)
910 goto err_icmpmsg_mib;
914 free_percpu(net->mib.icmpv6_statistics);
916 free_percpu(net->mib.ipv6_statistics);
918 free_percpu(net->mib.udplite_stats_in6);
920 free_percpu(net->mib.udp_stats_in6);
924 static void ipv6_cleanup_mibs(struct net *net)
926 free_percpu(net->mib.udp_stats_in6);
927 free_percpu(net->mib.udplite_stats_in6);
928 free_percpu(net->mib.ipv6_statistics);
929 free_percpu(net->mib.icmpv6_statistics);
930 kfree(net->mib.icmpv6msg_statistics);
933 static int __net_init inet6_net_init(struct net *net)
937 net->ipv6.sysctl.bindv6only = 0;
938 net->ipv6.sysctl.icmpv6_time = 1*HZ;
939 net->ipv6.sysctl.icmpv6_echo_ignore_all = 0;
940 net->ipv6.sysctl.icmpv6_echo_ignore_multicast = 0;
941 net->ipv6.sysctl.icmpv6_echo_ignore_anycast = 0;
943 /* By default, rate limit error messages.
944 * Except for pmtu discovery, it would break it.
945 * proc_do_large_bitmap needs pointer to the bitmap.
947 bitmap_set(net->ipv6.sysctl.icmpv6_ratemask, 0, ICMPV6_ERRMSG_MAX + 1);
948 bitmap_clear(net->ipv6.sysctl.icmpv6_ratemask, ICMPV6_PKT_TOOBIG, 1);
949 net->ipv6.sysctl.icmpv6_ratemask_ptr = net->ipv6.sysctl.icmpv6_ratemask;
951 net->ipv6.sysctl.flowlabel_consistency = 1;
952 net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS;
953 net->ipv6.sysctl.idgen_retries = 3;
954 net->ipv6.sysctl.idgen_delay = 1 * HZ;
955 net->ipv6.sysctl.flowlabel_state_ranges = 0;
956 net->ipv6.sysctl.max_dst_opts_cnt = IP6_DEFAULT_MAX_DST_OPTS_CNT;
957 net->ipv6.sysctl.max_hbh_opts_cnt = IP6_DEFAULT_MAX_HBH_OPTS_CNT;
958 net->ipv6.sysctl.max_dst_opts_len = IP6_DEFAULT_MAX_DST_OPTS_LEN;
959 net->ipv6.sysctl.max_hbh_opts_len = IP6_DEFAULT_MAX_HBH_OPTS_LEN;
960 net->ipv6.sysctl.fib_notify_on_flag_change = 0;
961 atomic_set(&net->ipv6.fib6_sernum, 1);
963 err = ipv6_init_mibs(net);
966 #ifdef CONFIG_PROC_FS
967 err = udp6_proc_init(net);
970 err = tcp6_proc_init(net);
973 err = ac6_proc_init(net);
979 #ifdef CONFIG_PROC_FS
985 ipv6_cleanup_mibs(net);
990 static void __net_exit inet6_net_exit(struct net *net)
992 #ifdef CONFIG_PROC_FS
997 ipv6_cleanup_mibs(net);
1000 static struct pernet_operations inet6_net_ops = {
1001 .init = inet6_net_init,
1002 .exit = inet6_net_exit,
1005 static int ipv6_route_input(struct sk_buff *skb)
1007 ip6_route_input(skb);
1008 return skb_dst(skb)->error;
1011 static const struct ipv6_stub ipv6_stub_impl = {
1012 .ipv6_sock_mc_join = ipv6_sock_mc_join,
1013 .ipv6_sock_mc_drop = ipv6_sock_mc_drop,
1014 .ipv6_dst_lookup_flow = ip6_dst_lookup_flow,
1015 .ipv6_route_input = ipv6_route_input,
1016 .fib6_get_table = fib6_get_table,
1017 .fib6_table_lookup = fib6_table_lookup,
1018 .fib6_lookup = fib6_lookup,
1019 .fib6_select_path = fib6_select_path,
1020 .ip6_mtu_from_fib6 = ip6_mtu_from_fib6,
1021 .fib6_nh_init = fib6_nh_init,
1022 .fib6_nh_release = fib6_nh_release,
1023 .fib6_update_sernum = fib6_update_sernum_stub,
1024 .fib6_rt_update = fib6_rt_update,
1025 .ip6_del_rt = ip6_del_rt,
1026 .udpv6_encap_enable = udpv6_encap_enable,
1027 .ndisc_send_na = ndisc_send_na,
1028 #if IS_ENABLED(CONFIG_XFRM)
1029 .xfrm6_local_rxpmtu = xfrm6_local_rxpmtu,
1030 .xfrm6_udp_encap_rcv = xfrm6_udp_encap_rcv,
1031 .xfrm6_rcv_encap = xfrm6_rcv_encap,
1034 .ipv6_fragment = ip6_fragment,
1037 static const struct ipv6_bpf_stub ipv6_bpf_stub_impl = {
1038 .inet6_bind = __inet6_bind,
1039 .udp6_lib_lookup = __udp6_lib_lookup,
1042 static int __init inet6_init(void)
1044 struct list_head *r;
1047 sock_skb_cb_check_size(sizeof(struct inet6_skb_parm));
1049 /* Register the socket-side information for inet6_create. */
1050 for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
1053 if (disable_ipv6_mod) {
1054 pr_info("Loaded, but administratively disabled, reboot required to enable\n");
1058 err = proto_register(&tcpv6_prot, 1);
1062 err = proto_register(&udpv6_prot, 1);
1064 goto out_unregister_tcp_proto;
1066 err = proto_register(&udplitev6_prot, 1);
1068 goto out_unregister_udp_proto;
1070 err = proto_register(&rawv6_prot, 1);
1072 goto out_unregister_udplite_proto;
1074 err = proto_register(&pingv6_prot, 1);
1076 goto out_unregister_raw_proto;
1078 /* We MUST register RAW sockets before we create the ICMP6,
1079 * IGMP6, or NDISC control sockets.
1083 goto out_unregister_ping_proto;
1085 /* Register the family here so that the init calls below will
1086 * be able to create sockets. (?? is this dangerous ??)
1088 err = sock_register(&inet6_family_ops);
1090 goto out_sock_register_fail;
1093 * ipngwg API draft makes clear that the correct semantics
1094 * for TCP and UDP is to consider one TCP and UDP instance
1095 * in a host available by both INET and INET6 APIs and
1096 * able to communicate via both network protocols.
1099 err = register_pernet_subsys(&inet6_net_ops);
1101 goto register_pernet_fail;
1102 err = ip6_mr_init();
1105 err = icmpv6_init();
1115 err = ipv6_netfilter_init();
1117 goto netfilter_fail;
1118 /* Create /proc/foo6 entries. */
1119 #ifdef CONFIG_PROC_FS
1121 if (raw6_proc_init())
1122 goto proc_raw6_fail;
1123 if (udplite6_proc_init())
1124 goto proc_udplite6_fail;
1125 if (ipv6_misc_proc_init())
1126 goto proc_misc6_fail;
1127 if (if6_proc_init())
1130 err = ip6_route_init();
1132 goto ip6_route_fail;
1133 err = ndisc_late_init();
1135 goto ndisc_late_fail;
1136 err = ip6_flowlabel_init();
1138 goto ip6_flowlabel_fail;
1139 err = ipv6_anycast_init();
1141 goto ipv6_anycast_fail;
1142 err = addrconf_init();
1146 /* Init v6 extension headers. */
1147 err = ipv6_exthdrs_init();
1149 goto ipv6_exthdrs_fail;
1151 err = ipv6_frag_init();
1153 goto ipv6_frag_fail;
1155 /* Init v6 transport protocols. */
1160 err = udplitev6_init();
1162 goto udplitev6_fail;
1164 err = udpv6_offload_init();
1166 goto udpv6_offload_fail;
1172 err = ipv6_packet_init();
1174 goto ipv6_packet_fail;
1176 err = pingv6_init();
1180 err = calipso_init();
1192 err = igmp6_late_init();
1194 goto igmp6_late_err;
1196 #ifdef CONFIG_SYSCTL
1197 err = ipv6_sysctl_register();
1202 /* ensure that ipv6 stubs are visible only after ipv6 is ready */
1204 ipv6_stub = &ipv6_stub_impl;
1205 ipv6_bpf_stub = &ipv6_bpf_stub_impl;
1209 #ifdef CONFIG_SYSCTL
1211 igmp6_late_cleanup();
1222 ipv6_packet_cleanup();
1226 udpv6_offload_exit();
1234 ipv6_exthdrs_exit();
1238 ipv6_anycast_cleanup();
1240 ip6_flowlabel_cleanup();
1242 ndisc_late_cleanup();
1244 ip6_route_cleanup();
1246 #ifdef CONFIG_PROC_FS
1249 ipv6_misc_proc_exit();
1251 udplite6_proc_exit();
1256 ipv6_netfilter_fini();
1266 unregister_pernet_subsys(&inet6_net_ops);
1267 register_pernet_fail:
1268 sock_unregister(PF_INET6);
1269 rtnl_unregister_all(PF_INET6);
1270 out_sock_register_fail:
1272 out_unregister_ping_proto:
1273 proto_unregister(&pingv6_prot);
1274 out_unregister_raw_proto:
1275 proto_unregister(&rawv6_prot);
1276 out_unregister_udplite_proto:
1277 proto_unregister(&udplitev6_prot);
1278 out_unregister_udp_proto:
1279 proto_unregister(&udpv6_prot);
1280 out_unregister_tcp_proto:
1281 proto_unregister(&tcpv6_prot);
1284 module_init(inet6_init);
1286 MODULE_ALIAS_NETPROTO(PF_INET6);