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 void inet6_sock_destruct(struct sock *sk)
112 inet6_cleanup_sock(sk);
113 inet_sock_destruct(sk);
115 EXPORT_SYMBOL_GPL(inet6_sock_destruct);
117 static int inet6_create(struct net *net, struct socket *sock, int protocol,
120 struct inet_sock *inet;
121 struct ipv6_pinfo *np;
123 struct inet_protosw *answer;
124 struct proto *answer_prot;
125 unsigned char answer_flags;
126 int try_loading_module = 0;
129 if (protocol < 0 || protocol >= IPPROTO_MAX)
132 /* Look for the requested type/protocol pair. */
134 err = -ESOCKTNOSUPPORT;
136 list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
139 /* Check the non-wild match. */
140 if (protocol == answer->protocol) {
141 if (protocol != IPPROTO_IP)
144 /* Check for the two wild cases. */
145 if (IPPROTO_IP == protocol) {
146 protocol = answer->protocol;
149 if (IPPROTO_IP == answer->protocol)
152 err = -EPROTONOSUPPORT;
156 if (try_loading_module < 2) {
159 * Be more specific, e.g. net-pf-10-proto-132-type-1
160 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
162 if (++try_loading_module == 1)
163 request_module("net-pf-%d-proto-%d-type-%d",
164 PF_INET6, protocol, sock->type);
166 * Fall back to generic, e.g. net-pf-10-proto-132
167 * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
170 request_module("net-pf-%d-proto-%d",
172 goto lookup_protocol;
178 if (sock->type == SOCK_RAW && !kern &&
179 !ns_capable(net->user_ns, CAP_NET_RAW))
182 sock->ops = answer->ops;
183 answer_prot = answer->prot;
184 answer_flags = answer->flags;
187 WARN_ON(!answer_prot->slab);
190 sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot, kern);
194 sock_init_data(sock, sk);
197 if (INET_PROTOSW_REUSE & answer_flags)
198 sk->sk_reuse = SK_CAN_REUSE;
200 if (INET_PROTOSW_ICSK & answer_flags)
201 inet_init_csk_locks(sk);
204 inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
206 if (SOCK_RAW == sock->type) {
207 inet->inet_num = protocol;
208 if (IPPROTO_RAW == protocol)
212 sk->sk_destruct = inet6_sock_destruct;
213 sk->sk_family = PF_INET6;
214 sk->sk_protocol = protocol;
216 sk->sk_backlog_rcv = answer->prot->backlog_rcv;
218 inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
220 np->mcast_hops = IPV6_DEFAULT_MCASTHOPS;
223 np->pmtudisc = IPV6_PMTUDISC_WANT;
224 np->repflow = net->ipv6.sysctl.flowlabel_reflect & FLOWLABEL_REFLECT_ESTABLISHED;
225 sk->sk_ipv6only = net->ipv6.sysctl.bindv6only;
227 /* Init the ipv4 part of the socket since we can have sockets
228 * using v6 API for ipv4.
235 inet->mc_list = NULL;
238 if (READ_ONCE(net->ipv4.sysctl_ip_no_pmtu_disc))
239 inet->pmtudisc = IP_PMTUDISC_DONT;
241 inet->pmtudisc = IP_PMTUDISC_WANT;
243 * Increment only the relevant sk_prot->socks debug field, this changes
244 * the previous behaviour of incrementing both the equivalent to
245 * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
247 * This allows better debug granularity as we'll know exactly how many
248 * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
249 * transport protocol socks. -acme
251 sk_refcnt_debug_inc(sk);
253 if (inet->inet_num) {
254 /* It assumes that any protocol which allows
255 * the user to assign a number at socket
256 * creation time automatically shares.
258 inet->inet_sport = htons(inet->inet_num);
259 err = sk->sk_prot->hash(sk);
261 sk_common_release(sk);
265 if (sk->sk_prot->init) {
266 err = sk->sk_prot->init(sk);
268 sk_common_release(sk);
274 err = BPF_CGROUP_RUN_PROG_INET_SOCK(sk);
276 sk_common_release(sk);
287 static int __inet6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len,
290 struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr;
291 struct inet_sock *inet = inet_sk(sk);
292 struct ipv6_pinfo *np = inet6_sk(sk);
293 struct net *net = sock_net(sk);
300 if (addr->sin6_family != AF_INET6)
301 return -EAFNOSUPPORT;
303 addr_type = ipv6_addr_type(&addr->sin6_addr);
304 if ((addr_type & IPV6_ADDR_MULTICAST) && sk->sk_type == SOCK_STREAM)
307 snum = ntohs(addr->sin6_port);
308 if (snum && inet_port_requires_bind_service(net, snum) &&
309 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
312 if (flags & BIND_WITH_LOCK)
315 /* Check these errors (active socket, double bind). */
316 if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
321 /* Check if the address belongs to the host. */
322 if (addr_type == IPV6_ADDR_MAPPED) {
323 struct net_device *dev = NULL;
326 /* Binding to v4-mapped address on a v6-only socket
329 if (sk->sk_ipv6only) {
335 if (sk->sk_bound_dev_if) {
336 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
343 /* Reproduce AF_INET checks to make the bindings consistent */
344 v4addr = addr->sin6_addr.s6_addr32[3];
345 chk_addr_ret = inet_addr_type_dev_table(net, dev, v4addr);
348 if (!inet_can_nonlocal_bind(net, inet) &&
349 v4addr != htonl(INADDR_ANY) &&
350 chk_addr_ret != RTN_LOCAL &&
351 chk_addr_ret != RTN_MULTICAST &&
352 chk_addr_ret != RTN_BROADCAST) {
353 err = -EADDRNOTAVAIL;
357 if (addr_type != IPV6_ADDR_ANY) {
358 struct net_device *dev = NULL;
361 if (__ipv6_addr_needs_scope_id(addr_type)) {
362 if (addr_len >= sizeof(struct sockaddr_in6) &&
363 addr->sin6_scope_id) {
364 /* Override any existing binding, if another one
365 * is supplied by user.
367 sk->sk_bound_dev_if = addr->sin6_scope_id;
370 /* Binding to link-local address requires an interface */
371 if (!sk->sk_bound_dev_if) {
377 if (sk->sk_bound_dev_if) {
378 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
385 /* ipv4 addr of the socket is invalid. Only the
386 * unspecified and mapped address have a v4 equivalent.
388 v4addr = LOOPBACK4_IPV6;
389 if (!(addr_type & IPV6_ADDR_MULTICAST)) {
390 if (!ipv6_can_nonlocal_bind(net, inet) &&
391 !ipv6_chk_addr(net, &addr->sin6_addr,
393 err = -EADDRNOTAVAIL;
401 inet->inet_rcv_saddr = v4addr;
402 inet->inet_saddr = v4addr;
404 sk->sk_v6_rcv_saddr = addr->sin6_addr;
406 if (!(addr_type & IPV6_ADDR_MULTICAST))
407 np->saddr = addr->sin6_addr;
409 saved_ipv6only = sk->sk_ipv6only;
410 if (addr_type != IPV6_ADDR_ANY && addr_type != IPV6_ADDR_MAPPED)
413 /* Make sure we are allowed to bind here. */
414 if (snum || !(inet->bind_address_no_port ||
415 (flags & BIND_FORCE_ADDRESS_NO_PORT))) {
416 if (sk->sk_prot->get_port(sk, snum)) {
417 sk->sk_ipv6only = saved_ipv6only;
418 inet_reset_saddr(sk);
422 if (!(flags & BIND_FROM_BPF)) {
423 err = BPF_CGROUP_RUN_PROG_INET6_POST_BIND(sk);
425 sk->sk_ipv6only = saved_ipv6only;
426 inet_reset_saddr(sk);
432 if (addr_type != IPV6_ADDR_ANY)
433 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
435 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
436 inet->inet_sport = htons(inet->inet_num);
437 inet->inet_dport = 0;
438 inet->inet_daddr = 0;
440 if (flags & BIND_WITH_LOCK)
448 /* bind for INET6 API */
449 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
451 struct sock *sk = sock->sk;
454 /* If the socket has its own bind function then use it. */
455 if (sk->sk_prot->bind)
456 return sk->sk_prot->bind(sk, uaddr, addr_len);
458 if (addr_len < SIN6_LEN_RFC2133)
461 /* BPF prog is run before any checks are done so that if the prog
462 * changes context in a wrong way it will be caught.
464 err = BPF_CGROUP_RUN_PROG_INET6_BIND(sk, uaddr);
468 return __inet6_bind(sk, uaddr, addr_len, BIND_WITH_LOCK);
470 EXPORT_SYMBOL(inet6_bind);
472 int inet6_release(struct socket *sock)
474 struct sock *sk = sock->sk;
480 ipv6_sock_mc_close(sk);
483 ipv6_sock_ac_close(sk);
485 return inet_release(sock);
487 EXPORT_SYMBOL(inet6_release);
489 void inet6_destroy_sock(struct sock *sk)
491 struct ipv6_pinfo *np = inet6_sk(sk);
493 struct ipv6_txoptions *opt;
495 /* Release rx options */
497 skb = xchg(&np->pktoptions, NULL);
500 skb = xchg(&np->rxpmtu, NULL);
503 /* Free flowlabels */
504 fl6_free_socklist(sk);
506 /* Free tx options */
508 opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL);
510 atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
514 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
516 void inet6_cleanup_sock(struct sock *sk)
518 inet6_destroy_sock(sk);
520 EXPORT_SYMBOL_GPL(inet6_cleanup_sock);
523 * This does both peername and sockname.
525 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
528 struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
529 struct sock *sk = sock->sk;
530 struct inet_sock *inet = inet_sk(sk);
531 struct ipv6_pinfo *np = inet6_sk(sk);
533 sin->sin6_family = AF_INET6;
534 sin->sin6_flowinfo = 0;
535 sin->sin6_scope_id = 0;
537 if (!inet->inet_dport)
539 if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
542 sin->sin6_port = inet->inet_dport;
543 sin->sin6_addr = sk->sk_v6_daddr;
545 sin->sin6_flowinfo = np->flow_label;
547 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
548 sin->sin6_addr = np->saddr;
550 sin->sin6_addr = sk->sk_v6_rcv_saddr;
551 sin->sin6_port = inet->inet_sport;
553 if (cgroup_bpf_enabled)
554 BPF_CGROUP_RUN_SA_PROG_LOCK(sk, (struct sockaddr *)sin,
555 peer ? BPF_CGROUP_INET6_GETPEERNAME :
556 BPF_CGROUP_INET6_GETSOCKNAME,
558 sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
559 sk->sk_bound_dev_if);
562 EXPORT_SYMBOL(inet6_getname);
564 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
566 void __user *argp = (void __user *)arg;
567 struct sock *sk = sock->sk;
568 struct net *net = sock_net(sk);
573 struct in6_rtmsg rtmsg;
575 if (copy_from_user(&rtmsg, argp, sizeof(rtmsg)))
577 return ipv6_route_ioctl(net, cmd, &rtmsg);
580 return addrconf_add_ifaddr(net, argp);
582 return addrconf_del_ifaddr(net, argp);
584 return addrconf_set_dstaddr(net, argp);
586 if (!sk->sk_prot->ioctl)
588 return sk->sk_prot->ioctl(sk, cmd, arg);
593 EXPORT_SYMBOL(inet6_ioctl);
596 struct compat_in6_rtmsg {
597 struct in6_addr rtmsg_dst;
598 struct in6_addr rtmsg_src;
599 struct in6_addr rtmsg_gateway;
609 static int inet6_compat_routing_ioctl(struct sock *sk, unsigned int cmd,
610 struct compat_in6_rtmsg __user *ur)
614 if (copy_from_user(&rt.rtmsg_dst, &ur->rtmsg_dst,
615 3 * sizeof(struct in6_addr)) ||
616 get_user(rt.rtmsg_type, &ur->rtmsg_type) ||
617 get_user(rt.rtmsg_dst_len, &ur->rtmsg_dst_len) ||
618 get_user(rt.rtmsg_src_len, &ur->rtmsg_src_len) ||
619 get_user(rt.rtmsg_metric, &ur->rtmsg_metric) ||
620 get_user(rt.rtmsg_info, &ur->rtmsg_info) ||
621 get_user(rt.rtmsg_flags, &ur->rtmsg_flags) ||
622 get_user(rt.rtmsg_ifindex, &ur->rtmsg_ifindex))
626 return ipv6_route_ioctl(sock_net(sk), cmd, &rt);
629 int inet6_compat_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
631 void __user *argp = compat_ptr(arg);
632 struct sock *sk = sock->sk;
637 return inet6_compat_routing_ioctl(sk, cmd, argp);
642 EXPORT_SYMBOL_GPL(inet6_compat_ioctl);
643 #endif /* CONFIG_COMPAT */
645 INDIRECT_CALLABLE_DECLARE(int udpv6_sendmsg(struct sock *, struct msghdr *,
647 int inet6_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
649 struct sock *sk = sock->sk;
651 if (unlikely(inet_send_prepare(sk)))
654 return INDIRECT_CALL_2(sk->sk_prot->sendmsg, tcp_sendmsg, udpv6_sendmsg,
658 INDIRECT_CALLABLE_DECLARE(int udpv6_recvmsg(struct sock *, struct msghdr *,
659 size_t, int, int, int *));
660 int inet6_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
663 struct sock *sk = sock->sk;
667 if (likely(!(flags & MSG_ERRQUEUE)))
668 sock_rps_record_flow(sk);
670 err = INDIRECT_CALL_2(sk->sk_prot->recvmsg, tcp_recvmsg, udpv6_recvmsg,
671 sk, msg, size, flags & MSG_DONTWAIT,
672 flags & ~MSG_DONTWAIT, &addr_len);
674 msg->msg_namelen = addr_len;
678 const struct proto_ops inet6_stream_ops = {
680 .owner = THIS_MODULE,
681 .release = inet6_release,
683 .connect = inet_stream_connect, /* ok */
684 .socketpair = sock_no_socketpair, /* a do nothing */
685 .accept = inet_accept, /* ok */
686 .getname = inet6_getname,
687 .poll = tcp_poll, /* ok */
688 .ioctl = inet6_ioctl, /* must change */
689 .gettstamp = sock_gettstamp,
690 .listen = inet_listen, /* ok */
691 .shutdown = inet_shutdown, /* ok */
692 .setsockopt = sock_common_setsockopt, /* ok */
693 .getsockopt = sock_common_getsockopt, /* ok */
694 .sendmsg = inet6_sendmsg, /* retpoline's sake */
695 .recvmsg = inet6_recvmsg, /* retpoline's sake */
699 .sendpage = inet_sendpage,
700 .sendmsg_locked = tcp_sendmsg_locked,
701 .sendpage_locked = tcp_sendpage_locked,
702 .splice_read = tcp_splice_read,
703 .read_sock = tcp_read_sock,
704 .peek_len = tcp_peek_len,
706 .compat_ioctl = inet6_compat_ioctl,
708 .set_rcvlowat = tcp_set_rcvlowat,
711 const struct proto_ops inet6_dgram_ops = {
713 .owner = THIS_MODULE,
714 .release = inet6_release,
716 .connect = inet_dgram_connect, /* ok */
717 .socketpair = sock_no_socketpair, /* a do nothing */
718 .accept = sock_no_accept, /* a do nothing */
719 .getname = inet6_getname,
720 .poll = udp_poll, /* ok */
721 .ioctl = inet6_ioctl, /* must change */
722 .gettstamp = sock_gettstamp,
723 .listen = sock_no_listen, /* ok */
724 .shutdown = inet_shutdown, /* ok */
725 .setsockopt = sock_common_setsockopt, /* ok */
726 .getsockopt = sock_common_getsockopt, /* ok */
727 .sendmsg = inet6_sendmsg, /* retpoline's sake */
728 .recvmsg = inet6_recvmsg, /* retpoline's sake */
729 .mmap = sock_no_mmap,
730 .sendpage = sock_no_sendpage,
731 .set_peek_off = sk_set_peek_off,
733 .compat_ioctl = inet6_compat_ioctl,
737 static const struct net_proto_family inet6_family_ops = {
739 .create = inet6_create,
740 .owner = THIS_MODULE,
743 int inet6_register_protosw(struct inet_protosw *p)
745 struct list_head *lh;
746 struct inet_protosw *answer;
747 struct list_head *last_perm;
748 int protocol = p->protocol;
751 spin_lock_bh(&inetsw6_lock);
754 if (p->type >= SOCK_MAX)
757 /* If we are trying to override a permanent protocol, bail. */
760 last_perm = &inetsw6[p->type];
761 list_for_each(lh, &inetsw6[p->type]) {
762 answer = list_entry(lh, struct inet_protosw, list);
764 /* Check only the non-wild match. */
765 if (INET_PROTOSW_PERMANENT & answer->flags) {
766 if (protocol == answer->protocol)
776 /* Add the new entry after the last permanent entry if any, so that
777 * the new entry does not override a permanent entry when matched with
778 * a wild-card protocol. But it is allowed to override any existing
779 * non-permanent entry. This means that when we remove this entry, the
780 * system automatically returns to the old behavior.
782 list_add_rcu(&p->list, last_perm);
785 spin_unlock_bh(&inetsw6_lock);
789 pr_err("Attempt to override permanent protocol %d\n", protocol);
793 pr_err("Ignoring attempt to register invalid socket type %d\n",
797 EXPORT_SYMBOL(inet6_register_protosw);
800 inet6_unregister_protosw(struct inet_protosw *p)
802 if (INET_PROTOSW_PERMANENT & p->flags) {
803 pr_err("Attempt to unregister permanent protocol %d\n",
806 spin_lock_bh(&inetsw6_lock);
807 list_del_rcu(&p->list);
808 spin_unlock_bh(&inetsw6_lock);
813 EXPORT_SYMBOL(inet6_unregister_protosw);
815 int inet6_sk_rebuild_header(struct sock *sk)
817 struct ipv6_pinfo *np = inet6_sk(sk);
818 struct dst_entry *dst;
820 dst = __sk_dst_check(sk, np->dst_cookie);
823 struct inet_sock *inet = inet_sk(sk);
824 struct in6_addr *final_p, final;
827 memset(&fl6, 0, sizeof(fl6));
828 fl6.flowi6_proto = sk->sk_protocol;
829 fl6.daddr = sk->sk_v6_daddr;
830 fl6.saddr = np->saddr;
831 fl6.flowlabel = np->flow_label;
832 fl6.flowi6_oif = sk->sk_bound_dev_if;
833 fl6.flowi6_mark = sk->sk_mark;
834 fl6.fl6_dport = inet->inet_dport;
835 fl6.fl6_sport = inet->inet_sport;
836 fl6.flowi6_uid = sk->sk_uid;
837 security_sk_classify_flow(sk, flowi6_to_flowi_common(&fl6));
840 final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt),
844 dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p);
846 sk->sk_route_caps = 0;
847 sk->sk_err_soft = -PTR_ERR(dst);
851 ip6_dst_store(sk, dst, NULL, NULL);
856 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
858 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb,
859 const struct inet6_skb_parm *opt)
861 const struct ipv6_pinfo *np = inet6_sk(sk);
864 if (((opt->flags & IP6SKB_HOPBYHOP) &&
865 (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) ||
866 (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) &&
867 np->rxopt.bits.rxflow) ||
868 (opt->srcrt && (np->rxopt.bits.srcrt ||
869 np->rxopt.bits.osrcrt)) ||
870 ((opt->dst1 || opt->dst0) &&
871 (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
876 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
878 static struct packet_type ipv6_packet_type __read_mostly = {
879 .type = cpu_to_be16(ETH_P_IPV6),
881 .list_func = ipv6_list_rcv,
884 static int __init ipv6_packet_init(void)
886 dev_add_pack(&ipv6_packet_type);
890 static void ipv6_packet_cleanup(void)
892 dev_remove_pack(&ipv6_packet_type);
895 static int __net_init ipv6_init_mibs(struct net *net)
899 net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib);
900 if (!net->mib.udp_stats_in6)
902 net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib);
903 if (!net->mib.udplite_stats_in6)
904 goto err_udplite_mib;
905 net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib);
906 if (!net->mib.ipv6_statistics)
909 for_each_possible_cpu(i) {
910 struct ipstats_mib *af_inet6_stats;
911 af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i);
912 u64_stats_init(&af_inet6_stats->syncp);
916 net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib);
917 if (!net->mib.icmpv6_statistics)
919 net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
921 if (!net->mib.icmpv6msg_statistics)
922 goto err_icmpmsg_mib;
926 free_percpu(net->mib.icmpv6_statistics);
928 free_percpu(net->mib.ipv6_statistics);
930 free_percpu(net->mib.udplite_stats_in6);
932 free_percpu(net->mib.udp_stats_in6);
936 static void ipv6_cleanup_mibs(struct net *net)
938 free_percpu(net->mib.udp_stats_in6);
939 free_percpu(net->mib.udplite_stats_in6);
940 free_percpu(net->mib.ipv6_statistics);
941 free_percpu(net->mib.icmpv6_statistics);
942 kfree(net->mib.icmpv6msg_statistics);
945 static int __net_init inet6_net_init(struct net *net)
949 net->ipv6.sysctl.bindv6only = 0;
950 net->ipv6.sysctl.icmpv6_time = 1*HZ;
951 net->ipv6.sysctl.icmpv6_echo_ignore_all = 0;
952 net->ipv6.sysctl.icmpv6_echo_ignore_multicast = 0;
953 net->ipv6.sysctl.icmpv6_echo_ignore_anycast = 0;
955 /* By default, rate limit error messages.
956 * Except for pmtu discovery, it would break it.
957 * proc_do_large_bitmap needs pointer to the bitmap.
959 bitmap_set(net->ipv6.sysctl.icmpv6_ratemask, 0, ICMPV6_ERRMSG_MAX + 1);
960 bitmap_clear(net->ipv6.sysctl.icmpv6_ratemask, ICMPV6_PKT_TOOBIG, 1);
961 net->ipv6.sysctl.icmpv6_ratemask_ptr = net->ipv6.sysctl.icmpv6_ratemask;
963 net->ipv6.sysctl.flowlabel_consistency = 1;
964 net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS;
965 net->ipv6.sysctl.idgen_retries = 3;
966 net->ipv6.sysctl.idgen_delay = 1 * HZ;
967 net->ipv6.sysctl.flowlabel_state_ranges = 0;
968 net->ipv6.sysctl.max_dst_opts_cnt = IP6_DEFAULT_MAX_DST_OPTS_CNT;
969 net->ipv6.sysctl.max_hbh_opts_cnt = IP6_DEFAULT_MAX_HBH_OPTS_CNT;
970 net->ipv6.sysctl.max_dst_opts_len = IP6_DEFAULT_MAX_DST_OPTS_LEN;
971 net->ipv6.sysctl.max_hbh_opts_len = IP6_DEFAULT_MAX_HBH_OPTS_LEN;
972 atomic_set(&net->ipv6.fib6_sernum, 1);
974 err = ipv6_init_mibs(net);
977 #ifdef CONFIG_PROC_FS
978 err = udp6_proc_init(net);
981 err = tcp6_proc_init(net);
984 err = ac6_proc_init(net);
990 #ifdef CONFIG_PROC_FS
996 ipv6_cleanup_mibs(net);
1001 static void __net_exit inet6_net_exit(struct net *net)
1003 #ifdef CONFIG_PROC_FS
1004 udp6_proc_exit(net);
1005 tcp6_proc_exit(net);
1008 ipv6_cleanup_mibs(net);
1011 static struct pernet_operations inet6_net_ops = {
1012 .init = inet6_net_init,
1013 .exit = inet6_net_exit,
1016 static int ipv6_route_input(struct sk_buff *skb)
1018 ip6_route_input(skb);
1019 return skb_dst(skb)->error;
1022 static const struct ipv6_stub ipv6_stub_impl = {
1023 .ipv6_sock_mc_join = ipv6_sock_mc_join,
1024 .ipv6_sock_mc_drop = ipv6_sock_mc_drop,
1025 .ipv6_dst_lookup_flow = ip6_dst_lookup_flow,
1026 .ipv6_route_input = ipv6_route_input,
1027 .fib6_get_table = fib6_get_table,
1028 .fib6_table_lookup = fib6_table_lookup,
1029 .fib6_lookup = fib6_lookup,
1030 .fib6_select_path = fib6_select_path,
1031 .ip6_mtu_from_fib6 = ip6_mtu_from_fib6,
1032 .fib6_nh_init = fib6_nh_init,
1033 .fib6_nh_release = fib6_nh_release,
1034 .fib6_nh_release_dsts = fib6_nh_release_dsts,
1035 .fib6_update_sernum = fib6_update_sernum_stub,
1036 .fib6_rt_update = fib6_rt_update,
1037 .ip6_del_rt = ip6_del_rt,
1038 .udpv6_encap_enable = udpv6_encap_enable,
1039 .ndisc_send_na = ndisc_send_na,
1040 #if IS_ENABLED(CONFIG_XFRM)
1041 .xfrm6_local_rxpmtu = xfrm6_local_rxpmtu,
1042 .xfrm6_udp_encap_rcv = xfrm6_udp_encap_rcv,
1043 .xfrm6_rcv_encap = xfrm6_rcv_encap,
1046 .ipv6_fragment = ip6_fragment,
1049 static const struct ipv6_bpf_stub ipv6_bpf_stub_impl = {
1050 .inet6_bind = __inet6_bind,
1051 .udp6_lib_lookup = __udp6_lib_lookup,
1054 static int __init inet6_init(void)
1056 struct list_head *r;
1059 sock_skb_cb_check_size(sizeof(struct inet6_skb_parm));
1061 /* Register the socket-side information for inet6_create. */
1062 for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
1065 if (disable_ipv6_mod) {
1066 pr_info("Loaded, but administratively disabled, reboot required to enable\n");
1070 err = proto_register(&tcpv6_prot, 1);
1074 err = proto_register(&udpv6_prot, 1);
1076 goto out_unregister_tcp_proto;
1078 err = proto_register(&udplitev6_prot, 1);
1080 goto out_unregister_udp_proto;
1082 err = proto_register(&rawv6_prot, 1);
1084 goto out_unregister_udplite_proto;
1086 err = proto_register(&pingv6_prot, 1);
1088 goto out_unregister_raw_proto;
1090 /* We MUST register RAW sockets before we create the ICMP6,
1091 * IGMP6, or NDISC control sockets.
1095 goto out_unregister_ping_proto;
1097 /* Register the family here so that the init calls below will
1098 * be able to create sockets. (?? is this dangerous ??)
1100 err = sock_register(&inet6_family_ops);
1102 goto out_sock_register_fail;
1105 * ipngwg API draft makes clear that the correct semantics
1106 * for TCP and UDP is to consider one TCP and UDP instance
1107 * in a host available by both INET and INET6 APIs and
1108 * able to communicate via both network protocols.
1111 err = register_pernet_subsys(&inet6_net_ops);
1113 goto register_pernet_fail;
1114 err = ip6_mr_init();
1117 err = icmpv6_init();
1127 err = ipv6_netfilter_init();
1129 goto netfilter_fail;
1130 /* Create /proc/foo6 entries. */
1131 #ifdef CONFIG_PROC_FS
1133 if (raw6_proc_init())
1134 goto proc_raw6_fail;
1135 if (udplite6_proc_init())
1136 goto proc_udplite6_fail;
1137 if (ipv6_misc_proc_init())
1138 goto proc_misc6_fail;
1139 if (if6_proc_init())
1142 err = ip6_route_init();
1144 goto ip6_route_fail;
1145 err = ndisc_late_init();
1147 goto ndisc_late_fail;
1148 err = ip6_flowlabel_init();
1150 goto ip6_flowlabel_fail;
1151 err = ipv6_anycast_init();
1153 goto ipv6_anycast_fail;
1154 err = addrconf_init();
1158 /* Init v6 extension headers. */
1159 err = ipv6_exthdrs_init();
1161 goto ipv6_exthdrs_fail;
1163 err = ipv6_frag_init();
1165 goto ipv6_frag_fail;
1167 /* Init v6 transport protocols. */
1172 err = udplitev6_init();
1174 goto udplitev6_fail;
1176 err = udpv6_offload_init();
1178 goto udpv6_offload_fail;
1184 err = ipv6_packet_init();
1186 goto ipv6_packet_fail;
1188 err = pingv6_init();
1192 err = calipso_init();
1204 err = igmp6_late_init();
1206 goto igmp6_late_err;
1208 #ifdef CONFIG_SYSCTL
1209 err = ipv6_sysctl_register();
1214 /* ensure that ipv6 stubs are visible only after ipv6 is ready */
1216 ipv6_stub = &ipv6_stub_impl;
1217 ipv6_bpf_stub = &ipv6_bpf_stub_impl;
1221 #ifdef CONFIG_SYSCTL
1223 igmp6_late_cleanup();
1234 ipv6_packet_cleanup();
1238 udpv6_offload_exit();
1246 ipv6_exthdrs_exit();
1250 ipv6_anycast_cleanup();
1252 ip6_flowlabel_cleanup();
1254 ndisc_late_cleanup();
1256 ip6_route_cleanup();
1258 #ifdef CONFIG_PROC_FS
1261 ipv6_misc_proc_exit();
1263 udplite6_proc_exit();
1268 ipv6_netfilter_fini();
1278 unregister_pernet_subsys(&inet6_net_ops);
1279 register_pernet_fail:
1280 sock_unregister(PF_INET6);
1281 rtnl_unregister_all(PF_INET6);
1282 out_sock_register_fail:
1284 out_unregister_ping_proto:
1285 proto_unregister(&pingv6_prot);
1286 out_unregister_raw_proto:
1287 proto_unregister(&rawv6_prot);
1288 out_unregister_udplite_proto:
1289 proto_unregister(&udplitev6_prot);
1290 out_unregister_udp_proto:
1291 proto_unregister(&udpv6_prot);
1292 out_unregister_tcp_proto:
1293 proto_unregister(&tcpv6_prot);
1296 module_init(inet6_init);
1298 MODULE_ALIAS_NETPROTO(PF_INET6);