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>
65 #include <net/ioam6.h>
66 #include <net/rawv6.h>
68 #include <linux/uaccess.h>
69 #include <linux/mroute6.h>
71 #include "ip6_offload.h"
73 MODULE_AUTHOR("Cast of dozens");
74 MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
75 MODULE_LICENSE("GPL");
77 /* The inetsw6 table contains everything that inet6_create needs to
80 static struct list_head inetsw6[SOCK_MAX];
81 static DEFINE_SPINLOCK(inetsw6_lock);
83 struct ipv6_params ipv6_defaults = {
88 static int disable_ipv6_mod;
90 module_param_named(disable, disable_ipv6_mod, int, 0444);
91 MODULE_PARM_DESC(disable, "Disable IPv6 module such that it is non-functional");
93 module_param_named(disable_ipv6, ipv6_defaults.disable_ipv6, int, 0444);
94 MODULE_PARM_DESC(disable_ipv6, "Disable IPv6 on all interfaces");
96 module_param_named(autoconf, ipv6_defaults.autoconf, int, 0444);
97 MODULE_PARM_DESC(autoconf, "Enable IPv6 address autoconfiguration on all interfaces");
99 bool ipv6_mod_enabled(void)
101 return disable_ipv6_mod == 0;
103 EXPORT_SYMBOL_GPL(ipv6_mod_enabled);
105 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
107 const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
109 return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
112 static int inet6_create(struct net *net, struct socket *sock, int protocol,
115 struct inet_sock *inet;
116 struct ipv6_pinfo *np;
118 struct inet_protosw *answer;
119 struct proto *answer_prot;
120 unsigned char answer_flags;
121 int try_loading_module = 0;
124 if (protocol < 0 || protocol >= IPPROTO_MAX)
127 /* Look for the requested type/protocol pair. */
129 err = -ESOCKTNOSUPPORT;
131 list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
134 /* Check the non-wild match. */
135 if (protocol == answer->protocol) {
136 if (protocol != IPPROTO_IP)
139 /* Check for the two wild cases. */
140 if (IPPROTO_IP == protocol) {
141 protocol = answer->protocol;
144 if (IPPROTO_IP == answer->protocol)
147 err = -EPROTONOSUPPORT;
151 if (try_loading_module < 2) {
154 * Be more specific, e.g. net-pf-10-proto-132-type-1
155 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
157 if (++try_loading_module == 1)
158 request_module("net-pf-%d-proto-%d-type-%d",
159 PF_INET6, protocol, sock->type);
161 * Fall back to generic, e.g. net-pf-10-proto-132
162 * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
165 request_module("net-pf-%d-proto-%d",
167 goto lookup_protocol;
173 if (sock->type == SOCK_RAW && !kern &&
174 !ns_capable(net->user_ns, CAP_NET_RAW))
177 sock->ops = answer->ops;
178 answer_prot = answer->prot;
179 answer_flags = answer->flags;
182 WARN_ON(!answer_prot->slab);
185 sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot, kern);
189 sock_init_data(sock, sk);
192 if (INET_PROTOSW_REUSE & answer_flags)
193 sk->sk_reuse = SK_CAN_REUSE;
196 inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
198 if (SOCK_RAW == sock->type) {
199 inet->inet_num = protocol;
200 if (IPPROTO_RAW == protocol)
204 sk->sk_destruct = inet_sock_destruct;
205 sk->sk_family = PF_INET6;
206 sk->sk_protocol = protocol;
208 sk->sk_backlog_rcv = answer->prot->backlog_rcv;
210 inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
212 np->mcast_hops = IPV6_DEFAULT_MCASTHOPS;
215 np->pmtudisc = IPV6_PMTUDISC_WANT;
216 np->repflow = net->ipv6.sysctl.flowlabel_reflect & FLOWLABEL_REFLECT_ESTABLISHED;
217 sk->sk_ipv6only = net->ipv6.sysctl.bindv6only;
219 /* Init the ipv4 part of the socket since we can have sockets
220 * using v6 API for ipv4.
227 RCU_INIT_POINTER(inet->mc_list, NULL);
230 if (READ_ONCE(net->ipv4.sysctl_ip_no_pmtu_disc))
231 inet->pmtudisc = IP_PMTUDISC_DONT;
233 inet->pmtudisc = IP_PMTUDISC_WANT;
235 * Increment only the relevant sk_prot->socks debug field, this changes
236 * the previous behaviour of incrementing both the equivalent to
237 * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
239 * This allows better debug granularity as we'll know exactly how many
240 * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
241 * transport protocol socks. -acme
243 sk_refcnt_debug_inc(sk);
245 if (inet->inet_num) {
246 /* It assumes that any protocol which allows
247 * the user to assign a number at socket
248 * creation time automatically shares.
250 inet->inet_sport = htons(inet->inet_num);
251 err = sk->sk_prot->hash(sk);
253 sk_common_release(sk);
257 if (sk->sk_prot->init) {
258 err = sk->sk_prot->init(sk);
260 sk_common_release(sk);
266 err = BPF_CGROUP_RUN_PROG_INET_SOCK(sk);
268 sk_common_release(sk);
279 static int __inet6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len,
282 struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr;
283 struct inet_sock *inet = inet_sk(sk);
284 struct ipv6_pinfo *np = inet6_sk(sk);
285 struct net *net = sock_net(sk);
292 if (addr->sin6_family != AF_INET6)
293 return -EAFNOSUPPORT;
295 addr_type = ipv6_addr_type(&addr->sin6_addr);
296 if ((addr_type & IPV6_ADDR_MULTICAST) && sk->sk_type == SOCK_STREAM)
299 snum = ntohs(addr->sin6_port);
300 if (!(flags & BIND_NO_CAP_NET_BIND_SERVICE) &&
301 snum && inet_port_requires_bind_service(net, snum) &&
302 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
305 if (flags & BIND_WITH_LOCK)
308 /* Check these errors (active socket, double bind). */
309 if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
314 /* Check if the address belongs to the host. */
315 if (addr_type == IPV6_ADDR_MAPPED) {
316 struct net_device *dev = NULL;
319 /* Binding to v4-mapped address on a v6-only socket
322 if (ipv6_only_sock(sk)) {
328 if (sk->sk_bound_dev_if) {
329 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
336 /* Reproduce AF_INET checks to make the bindings consistent */
337 v4addr = addr->sin6_addr.s6_addr32[3];
338 chk_addr_ret = inet_addr_type_dev_table(net, dev, v4addr);
341 if (!inet_addr_valid_or_nonlocal(net, inet, v4addr,
343 err = -EADDRNOTAVAIL;
347 if (addr_type != IPV6_ADDR_ANY) {
348 struct net_device *dev = NULL;
351 if (__ipv6_addr_needs_scope_id(addr_type)) {
352 if (addr_len >= sizeof(struct sockaddr_in6) &&
353 addr->sin6_scope_id) {
354 /* Override any existing binding, if another one
355 * is supplied by user.
357 sk->sk_bound_dev_if = addr->sin6_scope_id;
360 /* Binding to link-local address requires an interface */
361 if (!sk->sk_bound_dev_if) {
367 if (sk->sk_bound_dev_if) {
368 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
375 /* ipv4 addr of the socket is invalid. Only the
376 * unspecified and mapped address have a v4 equivalent.
378 v4addr = LOOPBACK4_IPV6;
379 if (!(addr_type & IPV6_ADDR_MULTICAST)) {
380 if (!ipv6_can_nonlocal_bind(net, inet) &&
381 !ipv6_chk_addr(net, &addr->sin6_addr,
383 err = -EADDRNOTAVAIL;
391 inet->inet_rcv_saddr = v4addr;
392 inet->inet_saddr = v4addr;
394 sk->sk_v6_rcv_saddr = addr->sin6_addr;
396 if (!(addr_type & IPV6_ADDR_MULTICAST))
397 np->saddr = addr->sin6_addr;
399 saved_ipv6only = sk->sk_ipv6only;
400 if (addr_type != IPV6_ADDR_ANY && addr_type != IPV6_ADDR_MAPPED)
403 /* Make sure we are allowed to bind here. */
404 if (snum || !(inet->bind_address_no_port ||
405 (flags & BIND_FORCE_ADDRESS_NO_PORT))) {
406 if (sk->sk_prot->get_port(sk, snum)) {
407 sk->sk_ipv6only = saved_ipv6only;
408 inet_reset_saddr(sk);
412 if (!(flags & BIND_FROM_BPF)) {
413 err = BPF_CGROUP_RUN_PROG_INET6_POST_BIND(sk);
415 sk->sk_ipv6only = saved_ipv6only;
416 inet_reset_saddr(sk);
417 if (sk->sk_prot->put_port)
418 sk->sk_prot->put_port(sk);
424 if (addr_type != IPV6_ADDR_ANY)
425 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
427 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
428 inet->inet_sport = htons(inet->inet_num);
429 inet->inet_dport = 0;
430 inet->inet_daddr = 0;
432 if (flags & BIND_WITH_LOCK)
440 /* bind for INET6 API */
441 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
443 struct sock *sk = sock->sk;
444 u32 flags = BIND_WITH_LOCK;
445 const struct proto *prot;
448 /* IPV6_ADDRFORM can change sk->sk_prot under us. */
449 prot = READ_ONCE(sk->sk_prot);
450 /* If the socket has its own bind function then use it. */
452 return prot->bind(sk, uaddr, addr_len);
454 if (addr_len < SIN6_LEN_RFC2133)
457 /* BPF prog is run before any checks are done so that if the prog
458 * changes context in a wrong way it will be caught.
460 err = BPF_CGROUP_RUN_PROG_INET_BIND_LOCK(sk, uaddr,
461 CGROUP_INET6_BIND, &flags);
465 return __inet6_bind(sk, uaddr, addr_len, flags);
467 EXPORT_SYMBOL(inet6_bind);
469 int inet6_release(struct socket *sock)
471 struct sock *sk = sock->sk;
477 ipv6_sock_mc_close(sk);
480 ipv6_sock_ac_close(sk);
482 return inet_release(sock);
484 EXPORT_SYMBOL(inet6_release);
486 void inet6_destroy_sock(struct sock *sk)
488 struct ipv6_pinfo *np = inet6_sk(sk);
490 struct ipv6_txoptions *opt;
492 /* Release rx options */
494 skb = xchg(&np->pktoptions, NULL);
497 skb = xchg(&np->rxpmtu, NULL);
500 /* Free flowlabels */
501 fl6_free_socklist(sk);
503 /* Free tx options */
505 opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL);
507 atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
511 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
514 * This does both peername and sockname.
516 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
519 struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
520 struct sock *sk = sock->sk;
521 struct inet_sock *inet = inet_sk(sk);
522 struct ipv6_pinfo *np = inet6_sk(sk);
524 sin->sin6_family = AF_INET6;
525 sin->sin6_flowinfo = 0;
526 sin->sin6_scope_id = 0;
529 if (!inet->inet_dport ||
530 (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
535 sin->sin6_port = inet->inet_dport;
536 sin->sin6_addr = sk->sk_v6_daddr;
538 sin->sin6_flowinfo = np->flow_label;
539 BPF_CGROUP_RUN_SA_PROG(sk, (struct sockaddr *)sin,
540 CGROUP_INET6_GETPEERNAME);
542 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
543 sin->sin6_addr = np->saddr;
545 sin->sin6_addr = sk->sk_v6_rcv_saddr;
546 sin->sin6_port = inet->inet_sport;
547 BPF_CGROUP_RUN_SA_PROG(sk, (struct sockaddr *)sin,
548 CGROUP_INET6_GETSOCKNAME);
550 sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
551 sk->sk_bound_dev_if);
555 EXPORT_SYMBOL(inet6_getname);
557 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
559 void __user *argp = (void __user *)arg;
560 struct sock *sk = sock->sk;
561 struct net *net = sock_net(sk);
562 const struct proto *prot;
567 struct in6_rtmsg rtmsg;
569 if (copy_from_user(&rtmsg, argp, sizeof(rtmsg)))
571 return ipv6_route_ioctl(net, cmd, &rtmsg);
574 return addrconf_add_ifaddr(net, argp);
576 return addrconf_del_ifaddr(net, argp);
578 return addrconf_set_dstaddr(net, argp);
580 /* IPV6_ADDRFORM can change sk->sk_prot under us. */
581 prot = READ_ONCE(sk->sk_prot);
584 return prot->ioctl(sk, cmd, arg);
589 EXPORT_SYMBOL(inet6_ioctl);
592 struct compat_in6_rtmsg {
593 struct in6_addr rtmsg_dst;
594 struct in6_addr rtmsg_src;
595 struct in6_addr rtmsg_gateway;
605 static int inet6_compat_routing_ioctl(struct sock *sk, unsigned int cmd,
606 struct compat_in6_rtmsg __user *ur)
610 if (copy_from_user(&rt.rtmsg_dst, &ur->rtmsg_dst,
611 3 * sizeof(struct in6_addr)) ||
612 get_user(rt.rtmsg_type, &ur->rtmsg_type) ||
613 get_user(rt.rtmsg_dst_len, &ur->rtmsg_dst_len) ||
614 get_user(rt.rtmsg_src_len, &ur->rtmsg_src_len) ||
615 get_user(rt.rtmsg_metric, &ur->rtmsg_metric) ||
616 get_user(rt.rtmsg_info, &ur->rtmsg_info) ||
617 get_user(rt.rtmsg_flags, &ur->rtmsg_flags) ||
618 get_user(rt.rtmsg_ifindex, &ur->rtmsg_ifindex))
622 return ipv6_route_ioctl(sock_net(sk), cmd, &rt);
625 int inet6_compat_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
627 void __user *argp = compat_ptr(arg);
628 struct sock *sk = sock->sk;
633 return inet6_compat_routing_ioctl(sk, cmd, argp);
638 EXPORT_SYMBOL_GPL(inet6_compat_ioctl);
639 #endif /* CONFIG_COMPAT */
641 INDIRECT_CALLABLE_DECLARE(int udpv6_sendmsg(struct sock *, struct msghdr *,
643 int inet6_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
645 struct sock *sk = sock->sk;
646 const struct proto *prot;
648 if (unlikely(inet_send_prepare(sk)))
651 /* IPV6_ADDRFORM can change sk->sk_prot under us. */
652 prot = READ_ONCE(sk->sk_prot);
653 return INDIRECT_CALL_2(prot->sendmsg, tcp_sendmsg, udpv6_sendmsg,
657 INDIRECT_CALLABLE_DECLARE(int udpv6_recvmsg(struct sock *, struct msghdr *,
658 size_t, int, int *));
659 int inet6_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
662 struct sock *sk = sock->sk;
663 const struct proto *prot;
667 if (likely(!(flags & MSG_ERRQUEUE)))
668 sock_rps_record_flow(sk);
670 /* IPV6_ADDRFORM can change sk->sk_prot under us. */
671 prot = READ_ONCE(sk->sk_prot);
672 err = INDIRECT_CALL_2(prot->recvmsg, tcp_recvmsg, udpv6_recvmsg,
673 sk, msg, size, flags, &addr_len);
675 msg->msg_namelen = addr_len;
679 const struct proto_ops inet6_stream_ops = {
681 .owner = THIS_MODULE,
682 .release = inet6_release,
684 .connect = inet_stream_connect, /* ok */
685 .socketpair = sock_no_socketpair, /* a do nothing */
686 .accept = inet_accept, /* ok */
687 .getname = inet6_getname,
688 .poll = tcp_poll, /* ok */
689 .ioctl = inet6_ioctl, /* must change */
690 .gettstamp = sock_gettstamp,
691 .listen = inet_listen, /* ok */
692 .shutdown = inet_shutdown, /* ok */
693 .setsockopt = sock_common_setsockopt, /* ok */
694 .getsockopt = sock_common_getsockopt, /* ok */
695 .sendmsg = inet6_sendmsg, /* retpoline's sake */
696 .recvmsg = inet6_recvmsg, /* retpoline's sake */
700 .sendpage = inet_sendpage,
701 .sendmsg_locked = tcp_sendmsg_locked,
702 .sendpage_locked = tcp_sendpage_locked,
703 .splice_read = tcp_splice_read,
704 .read_sock = tcp_read_sock,
705 .peek_len = tcp_peek_len,
707 .compat_ioctl = inet6_compat_ioctl,
709 .set_rcvlowat = tcp_set_rcvlowat,
712 const struct proto_ops inet6_dgram_ops = {
714 .owner = THIS_MODULE,
715 .release = inet6_release,
717 .connect = inet_dgram_connect, /* ok */
718 .socketpair = sock_no_socketpair, /* a do nothing */
719 .accept = sock_no_accept, /* a do nothing */
720 .getname = inet6_getname,
721 .poll = udp_poll, /* ok */
722 .ioctl = inet6_ioctl, /* must change */
723 .gettstamp = sock_gettstamp,
724 .listen = sock_no_listen, /* ok */
725 .shutdown = inet_shutdown, /* ok */
726 .setsockopt = sock_common_setsockopt, /* ok */
727 .getsockopt = sock_common_getsockopt, /* ok */
728 .sendmsg = inet6_sendmsg, /* retpoline's sake */
729 .recvmsg = inet6_recvmsg, /* retpoline's sake */
730 .read_sock = udp_read_sock,
731 .mmap = sock_no_mmap,
732 .sendpage = sock_no_sendpage,
733 .set_peek_off = sk_set_peek_off,
735 .compat_ioctl = inet6_compat_ioctl,
739 static const struct net_proto_family inet6_family_ops = {
741 .create = inet6_create,
742 .owner = THIS_MODULE,
745 int inet6_register_protosw(struct inet_protosw *p)
747 struct list_head *lh;
748 struct inet_protosw *answer;
749 struct list_head *last_perm;
750 int protocol = p->protocol;
753 spin_lock_bh(&inetsw6_lock);
756 if (p->type >= SOCK_MAX)
759 /* If we are trying to override a permanent protocol, bail. */
762 last_perm = &inetsw6[p->type];
763 list_for_each(lh, &inetsw6[p->type]) {
764 answer = list_entry(lh, struct inet_protosw, list);
766 /* Check only the non-wild match. */
767 if (INET_PROTOSW_PERMANENT & answer->flags) {
768 if (protocol == answer->protocol)
778 /* Add the new entry after the last permanent entry if any, so that
779 * the new entry does not override a permanent entry when matched with
780 * a wild-card protocol. But it is allowed to override any existing
781 * non-permanent entry. This means that when we remove this entry, the
782 * system automatically returns to the old behavior.
784 list_add_rcu(&p->list, last_perm);
787 spin_unlock_bh(&inetsw6_lock);
791 pr_err("Attempt to override permanent protocol %d\n", protocol);
795 pr_err("Ignoring attempt to register invalid socket type %d\n",
799 EXPORT_SYMBOL(inet6_register_protosw);
802 inet6_unregister_protosw(struct inet_protosw *p)
804 if (INET_PROTOSW_PERMANENT & p->flags) {
805 pr_err("Attempt to unregister permanent protocol %d\n",
808 spin_lock_bh(&inetsw6_lock);
809 list_del_rcu(&p->list);
810 spin_unlock_bh(&inetsw6_lock);
815 EXPORT_SYMBOL(inet6_unregister_protosw);
817 int inet6_sk_rebuild_header(struct sock *sk)
819 struct ipv6_pinfo *np = inet6_sk(sk);
820 struct dst_entry *dst;
822 dst = __sk_dst_check(sk, np->dst_cookie);
825 struct inet_sock *inet = inet_sk(sk);
826 struct in6_addr *final_p, final;
829 memset(&fl6, 0, sizeof(fl6));
830 fl6.flowi6_proto = sk->sk_protocol;
831 fl6.daddr = sk->sk_v6_daddr;
832 fl6.saddr = np->saddr;
833 fl6.flowlabel = np->flow_label;
834 fl6.flowi6_oif = sk->sk_bound_dev_if;
835 fl6.flowi6_mark = sk->sk_mark;
836 fl6.fl6_dport = inet->inet_dport;
837 fl6.fl6_sport = inet->inet_sport;
838 fl6.flowi6_uid = sk->sk_uid;
839 security_sk_classify_flow(sk, flowi6_to_flowi_common(&fl6));
842 final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt),
846 dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p);
848 sk->sk_route_caps = 0;
849 sk->sk_err_soft = -PTR_ERR(dst);
853 ip6_dst_store(sk, dst, NULL, NULL);
858 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
860 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb,
861 const struct inet6_skb_parm *opt)
863 const struct ipv6_pinfo *np = inet6_sk(sk);
866 if (((opt->flags & IP6SKB_HOPBYHOP) &&
867 (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) ||
868 (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) &&
869 np->rxopt.bits.rxflow) ||
870 (opt->srcrt && (np->rxopt.bits.srcrt ||
871 np->rxopt.bits.osrcrt)) ||
872 ((opt->dst1 || opt->dst0) &&
873 (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
878 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
880 static struct packet_type ipv6_packet_type __read_mostly = {
881 .type = cpu_to_be16(ETH_P_IPV6),
883 .list_func = ipv6_list_rcv,
886 static int __init ipv6_packet_init(void)
888 dev_add_pack(&ipv6_packet_type);
892 static void ipv6_packet_cleanup(void)
894 dev_remove_pack(&ipv6_packet_type);
897 static int __net_init ipv6_init_mibs(struct net *net)
901 net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib);
902 if (!net->mib.udp_stats_in6)
904 net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib);
905 if (!net->mib.udplite_stats_in6)
906 goto err_udplite_mib;
907 net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib);
908 if (!net->mib.ipv6_statistics)
911 for_each_possible_cpu(i) {
912 struct ipstats_mib *af_inet6_stats;
913 af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i);
914 u64_stats_init(&af_inet6_stats->syncp);
918 net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib);
919 if (!net->mib.icmpv6_statistics)
921 net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
923 if (!net->mib.icmpv6msg_statistics)
924 goto err_icmpmsg_mib;
928 free_percpu(net->mib.icmpv6_statistics);
930 free_percpu(net->mib.ipv6_statistics);
932 free_percpu(net->mib.udplite_stats_in6);
934 free_percpu(net->mib.udp_stats_in6);
938 static void ipv6_cleanup_mibs(struct net *net)
940 free_percpu(net->mib.udp_stats_in6);
941 free_percpu(net->mib.udplite_stats_in6);
942 free_percpu(net->mib.ipv6_statistics);
943 free_percpu(net->mib.icmpv6_statistics);
944 kfree(net->mib.icmpv6msg_statistics);
947 static int __net_init inet6_net_init(struct net *net)
951 net->ipv6.sysctl.bindv6only = 0;
952 net->ipv6.sysctl.icmpv6_time = 1*HZ;
953 net->ipv6.sysctl.icmpv6_echo_ignore_all = 0;
954 net->ipv6.sysctl.icmpv6_echo_ignore_multicast = 0;
955 net->ipv6.sysctl.icmpv6_echo_ignore_anycast = 0;
957 /* By default, rate limit error messages.
958 * Except for pmtu discovery, it would break it.
959 * proc_do_large_bitmap needs pointer to the bitmap.
961 bitmap_set(net->ipv6.sysctl.icmpv6_ratemask, 0, ICMPV6_ERRMSG_MAX + 1);
962 bitmap_clear(net->ipv6.sysctl.icmpv6_ratemask, ICMPV6_PKT_TOOBIG, 1);
963 net->ipv6.sysctl.icmpv6_ratemask_ptr = net->ipv6.sysctl.icmpv6_ratemask;
965 net->ipv6.sysctl.flowlabel_consistency = 1;
966 net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS;
967 net->ipv6.sysctl.idgen_retries = 3;
968 net->ipv6.sysctl.idgen_delay = 1 * HZ;
969 net->ipv6.sysctl.flowlabel_state_ranges = 0;
970 net->ipv6.sysctl.max_dst_opts_cnt = IP6_DEFAULT_MAX_DST_OPTS_CNT;
971 net->ipv6.sysctl.max_hbh_opts_cnt = IP6_DEFAULT_MAX_HBH_OPTS_CNT;
972 net->ipv6.sysctl.max_dst_opts_len = IP6_DEFAULT_MAX_DST_OPTS_LEN;
973 net->ipv6.sysctl.max_hbh_opts_len = IP6_DEFAULT_MAX_HBH_OPTS_LEN;
974 net->ipv6.sysctl.fib_notify_on_flag_change = 0;
975 atomic_set(&net->ipv6.fib6_sernum, 1);
977 net->ipv6.sysctl.ioam6_id = IOAM6_DEFAULT_ID;
978 net->ipv6.sysctl.ioam6_id_wide = IOAM6_DEFAULT_ID_WIDE;
980 err = ipv6_init_mibs(net);
983 #ifdef CONFIG_PROC_FS
984 err = udp6_proc_init(net);
987 err = tcp6_proc_init(net);
990 err = ac6_proc_init(net);
996 #ifdef CONFIG_PROC_FS
1000 udp6_proc_exit(net);
1002 ipv6_cleanup_mibs(net);
1007 static void __net_exit inet6_net_exit(struct net *net)
1009 #ifdef CONFIG_PROC_FS
1010 udp6_proc_exit(net);
1011 tcp6_proc_exit(net);
1014 ipv6_cleanup_mibs(net);
1017 static struct pernet_operations inet6_net_ops = {
1018 .init = inet6_net_init,
1019 .exit = inet6_net_exit,
1022 static int ipv6_route_input(struct sk_buff *skb)
1024 ip6_route_input(skb);
1025 return skb_dst(skb)->error;
1028 static const struct ipv6_stub ipv6_stub_impl = {
1029 .ipv6_sock_mc_join = ipv6_sock_mc_join,
1030 .ipv6_sock_mc_drop = ipv6_sock_mc_drop,
1031 .ipv6_dst_lookup_flow = ip6_dst_lookup_flow,
1032 .ipv6_route_input = ipv6_route_input,
1033 .fib6_get_table = fib6_get_table,
1034 .fib6_table_lookup = fib6_table_lookup,
1035 .fib6_lookup = fib6_lookup,
1036 .fib6_select_path = fib6_select_path,
1037 .ip6_mtu_from_fib6 = ip6_mtu_from_fib6,
1038 .fib6_nh_init = fib6_nh_init,
1039 .fib6_nh_release = fib6_nh_release,
1040 .fib6_nh_release_dsts = fib6_nh_release_dsts,
1041 .fib6_update_sernum = fib6_update_sernum_stub,
1042 .fib6_rt_update = fib6_rt_update,
1043 .ip6_del_rt = ip6_del_rt,
1044 .udpv6_encap_enable = udpv6_encap_enable,
1045 .ndisc_send_na = ndisc_send_na,
1046 #if IS_ENABLED(CONFIG_XFRM)
1047 .xfrm6_local_rxpmtu = xfrm6_local_rxpmtu,
1048 .xfrm6_udp_encap_rcv = xfrm6_udp_encap_rcv,
1049 .xfrm6_rcv_encap = xfrm6_rcv_encap,
1052 .ipv6_fragment = ip6_fragment,
1053 .ipv6_dev_find = ipv6_dev_find,
1056 static const struct ipv6_bpf_stub ipv6_bpf_stub_impl = {
1057 .inet6_bind = __inet6_bind,
1058 .udp6_lib_lookup = __udp6_lib_lookup,
1061 static int __init inet6_init(void)
1063 struct list_head *r;
1066 sock_skb_cb_check_size(sizeof(struct inet6_skb_parm));
1068 /* Register the socket-side information for inet6_create. */
1069 for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
1072 raw_hashinfo_init(&raw_v6_hashinfo);
1074 if (disable_ipv6_mod) {
1075 pr_info("Loaded, but administratively disabled, reboot required to enable\n");
1079 err = proto_register(&tcpv6_prot, 1);
1083 err = proto_register(&udpv6_prot, 1);
1085 goto out_unregister_tcp_proto;
1087 err = proto_register(&udplitev6_prot, 1);
1089 goto out_unregister_udp_proto;
1091 err = proto_register(&rawv6_prot, 1);
1093 goto out_unregister_udplite_proto;
1095 err = proto_register(&pingv6_prot, 1);
1097 goto out_unregister_raw_proto;
1099 /* We MUST register RAW sockets before we create the ICMP6,
1100 * IGMP6, or NDISC control sockets.
1104 goto out_unregister_ping_proto;
1106 /* Register the family here so that the init calls below will
1107 * be able to create sockets. (?? is this dangerous ??)
1109 err = sock_register(&inet6_family_ops);
1111 goto out_sock_register_fail;
1114 * ipngwg API draft makes clear that the correct semantics
1115 * for TCP and UDP is to consider one TCP and UDP instance
1116 * in a host available by both INET and INET6 APIs and
1117 * able to communicate via both network protocols.
1120 err = register_pernet_subsys(&inet6_net_ops);
1122 goto register_pernet_fail;
1123 err = ip6_mr_init();
1126 err = icmpv6_init();
1136 err = ipv6_netfilter_init();
1138 goto netfilter_fail;
1139 /* Create /proc/foo6 entries. */
1140 #ifdef CONFIG_PROC_FS
1142 if (raw6_proc_init())
1143 goto proc_raw6_fail;
1144 if (udplite6_proc_init())
1145 goto proc_udplite6_fail;
1146 if (ipv6_misc_proc_init())
1147 goto proc_misc6_fail;
1148 if (if6_proc_init())
1151 err = ip6_route_init();
1153 goto ip6_route_fail;
1154 err = ndisc_late_init();
1156 goto ndisc_late_fail;
1157 err = ip6_flowlabel_init();
1159 goto ip6_flowlabel_fail;
1160 err = ipv6_anycast_init();
1162 goto ipv6_anycast_fail;
1163 err = addrconf_init();
1167 /* Init v6 extension headers. */
1168 err = ipv6_exthdrs_init();
1170 goto ipv6_exthdrs_fail;
1172 err = ipv6_frag_init();
1174 goto ipv6_frag_fail;
1176 /* Init v6 transport protocols. */
1181 err = udplitev6_init();
1183 goto udplitev6_fail;
1185 err = udpv6_offload_init();
1187 goto udpv6_offload_fail;
1193 err = ipv6_packet_init();
1195 goto ipv6_packet_fail;
1197 err = pingv6_init();
1201 err = calipso_init();
1217 err = igmp6_late_init();
1219 goto igmp6_late_err;
1221 #ifdef CONFIG_SYSCTL
1222 err = ipv6_sysctl_register();
1227 /* ensure that ipv6 stubs are visible only after ipv6 is ready */
1229 ipv6_stub = &ipv6_stub_impl;
1230 ipv6_bpf_stub = &ipv6_bpf_stub_impl;
1234 #ifdef CONFIG_SYSCTL
1236 igmp6_late_cleanup();
1249 ipv6_packet_cleanup();
1253 udpv6_offload_exit();
1261 ipv6_exthdrs_exit();
1265 ipv6_anycast_cleanup();
1267 ip6_flowlabel_cleanup();
1269 ndisc_late_cleanup();
1271 ip6_route_cleanup();
1273 #ifdef CONFIG_PROC_FS
1276 ipv6_misc_proc_exit();
1278 udplite6_proc_exit();
1283 ipv6_netfilter_fini();
1293 unregister_pernet_subsys(&inet6_net_ops);
1294 register_pernet_fail:
1295 sock_unregister(PF_INET6);
1296 rtnl_unregister_all(PF_INET6);
1297 out_sock_register_fail:
1299 out_unregister_ping_proto:
1300 proto_unregister(&pingv6_prot);
1301 out_unregister_raw_proto:
1302 proto_unregister(&rawv6_prot);
1303 out_unregister_udplite_proto:
1304 proto_unregister(&udplitev6_prot);
1305 out_unregister_udp_proto:
1306 proto_unregister(&udpv6_prot);
1307 out_unregister_tcp_proto:
1308 proto_unregister(&tcpv6_prot);
1311 module_init(inet6_init);
1313 MODULE_ALIAS_NETPROTO(PF_INET6);