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 .read_skb = tcp_read_skb,
706 .peek_len = tcp_peek_len,
708 .compat_ioctl = inet6_compat_ioctl,
710 .set_rcvlowat = tcp_set_rcvlowat,
713 const struct proto_ops inet6_dgram_ops = {
715 .owner = THIS_MODULE,
716 .release = inet6_release,
718 .connect = inet_dgram_connect, /* ok */
719 .socketpair = sock_no_socketpair, /* a do nothing */
720 .accept = sock_no_accept, /* a do nothing */
721 .getname = inet6_getname,
722 .poll = udp_poll, /* ok */
723 .ioctl = inet6_ioctl, /* must change */
724 .gettstamp = sock_gettstamp,
725 .listen = sock_no_listen, /* ok */
726 .shutdown = inet_shutdown, /* ok */
727 .setsockopt = sock_common_setsockopt, /* ok */
728 .getsockopt = sock_common_getsockopt, /* ok */
729 .sendmsg = inet6_sendmsg, /* retpoline's sake */
730 .recvmsg = inet6_recvmsg, /* retpoline's sake */
731 .read_skb = udp_read_skb,
732 .mmap = sock_no_mmap,
733 .sendpage = sock_no_sendpage,
734 .set_peek_off = sk_set_peek_off,
736 .compat_ioctl = inet6_compat_ioctl,
740 static const struct net_proto_family inet6_family_ops = {
742 .create = inet6_create,
743 .owner = THIS_MODULE,
746 int inet6_register_protosw(struct inet_protosw *p)
748 struct list_head *lh;
749 struct inet_protosw *answer;
750 struct list_head *last_perm;
751 int protocol = p->protocol;
754 spin_lock_bh(&inetsw6_lock);
757 if (p->type >= SOCK_MAX)
760 /* If we are trying to override a permanent protocol, bail. */
763 last_perm = &inetsw6[p->type];
764 list_for_each(lh, &inetsw6[p->type]) {
765 answer = list_entry(lh, struct inet_protosw, list);
767 /* Check only the non-wild match. */
768 if (INET_PROTOSW_PERMANENT & answer->flags) {
769 if (protocol == answer->protocol)
779 /* Add the new entry after the last permanent entry if any, so that
780 * the new entry does not override a permanent entry when matched with
781 * a wild-card protocol. But it is allowed to override any existing
782 * non-permanent entry. This means that when we remove this entry, the
783 * system automatically returns to the old behavior.
785 list_add_rcu(&p->list, last_perm);
788 spin_unlock_bh(&inetsw6_lock);
792 pr_err("Attempt to override permanent protocol %d\n", protocol);
796 pr_err("Ignoring attempt to register invalid socket type %d\n",
800 EXPORT_SYMBOL(inet6_register_protosw);
803 inet6_unregister_protosw(struct inet_protosw *p)
805 if (INET_PROTOSW_PERMANENT & p->flags) {
806 pr_err("Attempt to unregister permanent protocol %d\n",
809 spin_lock_bh(&inetsw6_lock);
810 list_del_rcu(&p->list);
811 spin_unlock_bh(&inetsw6_lock);
816 EXPORT_SYMBOL(inet6_unregister_protosw);
818 int inet6_sk_rebuild_header(struct sock *sk)
820 struct ipv6_pinfo *np = inet6_sk(sk);
821 struct dst_entry *dst;
823 dst = __sk_dst_check(sk, np->dst_cookie);
826 struct inet_sock *inet = inet_sk(sk);
827 struct in6_addr *final_p, final;
830 memset(&fl6, 0, sizeof(fl6));
831 fl6.flowi6_proto = sk->sk_protocol;
832 fl6.daddr = sk->sk_v6_daddr;
833 fl6.saddr = np->saddr;
834 fl6.flowlabel = np->flow_label;
835 fl6.flowi6_oif = sk->sk_bound_dev_if;
836 fl6.flowi6_mark = sk->sk_mark;
837 fl6.fl6_dport = inet->inet_dport;
838 fl6.fl6_sport = inet->inet_sport;
839 fl6.flowi6_uid = sk->sk_uid;
840 security_sk_classify_flow(sk, flowi6_to_flowi_common(&fl6));
843 final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt),
847 dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p);
849 sk->sk_route_caps = 0;
850 sk->sk_err_soft = -PTR_ERR(dst);
854 ip6_dst_store(sk, dst, NULL, NULL);
859 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
861 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb,
862 const struct inet6_skb_parm *opt)
864 const struct ipv6_pinfo *np = inet6_sk(sk);
867 if (((opt->flags & IP6SKB_HOPBYHOP) &&
868 (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) ||
869 (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) &&
870 np->rxopt.bits.rxflow) ||
871 (opt->srcrt && (np->rxopt.bits.srcrt ||
872 np->rxopt.bits.osrcrt)) ||
873 ((opt->dst1 || opt->dst0) &&
874 (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
879 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
881 static struct packet_type ipv6_packet_type __read_mostly = {
882 .type = cpu_to_be16(ETH_P_IPV6),
884 .list_func = ipv6_list_rcv,
887 static int __init ipv6_packet_init(void)
889 dev_add_pack(&ipv6_packet_type);
893 static void ipv6_packet_cleanup(void)
895 dev_remove_pack(&ipv6_packet_type);
898 static int __net_init ipv6_init_mibs(struct net *net)
902 net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib);
903 if (!net->mib.udp_stats_in6)
905 net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib);
906 if (!net->mib.udplite_stats_in6)
907 goto err_udplite_mib;
908 net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib);
909 if (!net->mib.ipv6_statistics)
912 for_each_possible_cpu(i) {
913 struct ipstats_mib *af_inet6_stats;
914 af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i);
915 u64_stats_init(&af_inet6_stats->syncp);
919 net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib);
920 if (!net->mib.icmpv6_statistics)
922 net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
924 if (!net->mib.icmpv6msg_statistics)
925 goto err_icmpmsg_mib;
929 free_percpu(net->mib.icmpv6_statistics);
931 free_percpu(net->mib.ipv6_statistics);
933 free_percpu(net->mib.udplite_stats_in6);
935 free_percpu(net->mib.udp_stats_in6);
939 static void ipv6_cleanup_mibs(struct net *net)
941 free_percpu(net->mib.udp_stats_in6);
942 free_percpu(net->mib.udplite_stats_in6);
943 free_percpu(net->mib.ipv6_statistics);
944 free_percpu(net->mib.icmpv6_statistics);
945 kfree(net->mib.icmpv6msg_statistics);
948 static int __net_init inet6_net_init(struct net *net)
952 net->ipv6.sysctl.bindv6only = 0;
953 net->ipv6.sysctl.icmpv6_time = 1*HZ;
954 net->ipv6.sysctl.icmpv6_echo_ignore_all = 0;
955 net->ipv6.sysctl.icmpv6_echo_ignore_multicast = 0;
956 net->ipv6.sysctl.icmpv6_echo_ignore_anycast = 0;
958 /* By default, rate limit error messages.
959 * Except for pmtu discovery, it would break it.
960 * proc_do_large_bitmap needs pointer to the bitmap.
962 bitmap_set(net->ipv6.sysctl.icmpv6_ratemask, 0, ICMPV6_ERRMSG_MAX + 1);
963 bitmap_clear(net->ipv6.sysctl.icmpv6_ratemask, ICMPV6_PKT_TOOBIG, 1);
964 net->ipv6.sysctl.icmpv6_ratemask_ptr = net->ipv6.sysctl.icmpv6_ratemask;
966 net->ipv6.sysctl.flowlabel_consistency = 1;
967 net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS;
968 net->ipv6.sysctl.idgen_retries = 3;
969 net->ipv6.sysctl.idgen_delay = 1 * HZ;
970 net->ipv6.sysctl.flowlabel_state_ranges = 0;
971 net->ipv6.sysctl.max_dst_opts_cnt = IP6_DEFAULT_MAX_DST_OPTS_CNT;
972 net->ipv6.sysctl.max_hbh_opts_cnt = IP6_DEFAULT_MAX_HBH_OPTS_CNT;
973 net->ipv6.sysctl.max_dst_opts_len = IP6_DEFAULT_MAX_DST_OPTS_LEN;
974 net->ipv6.sysctl.max_hbh_opts_len = IP6_DEFAULT_MAX_HBH_OPTS_LEN;
975 net->ipv6.sysctl.fib_notify_on_flag_change = 0;
976 atomic_set(&net->ipv6.fib6_sernum, 1);
978 net->ipv6.sysctl.ioam6_id = IOAM6_DEFAULT_ID;
979 net->ipv6.sysctl.ioam6_id_wide = IOAM6_DEFAULT_ID_WIDE;
981 err = ipv6_init_mibs(net);
984 #ifdef CONFIG_PROC_FS
985 err = udp6_proc_init(net);
988 err = tcp6_proc_init(net);
991 err = ac6_proc_init(net);
997 #ifdef CONFIG_PROC_FS
1001 udp6_proc_exit(net);
1003 ipv6_cleanup_mibs(net);
1008 static void __net_exit inet6_net_exit(struct net *net)
1010 #ifdef CONFIG_PROC_FS
1011 udp6_proc_exit(net);
1012 tcp6_proc_exit(net);
1015 ipv6_cleanup_mibs(net);
1018 static struct pernet_operations inet6_net_ops = {
1019 .init = inet6_net_init,
1020 .exit = inet6_net_exit,
1023 static int ipv6_route_input(struct sk_buff *skb)
1025 ip6_route_input(skb);
1026 return skb_dst(skb)->error;
1029 static const struct ipv6_stub ipv6_stub_impl = {
1030 .ipv6_sock_mc_join = ipv6_sock_mc_join,
1031 .ipv6_sock_mc_drop = ipv6_sock_mc_drop,
1032 .ipv6_dst_lookup_flow = ip6_dst_lookup_flow,
1033 .ipv6_route_input = ipv6_route_input,
1034 .fib6_get_table = fib6_get_table,
1035 .fib6_table_lookup = fib6_table_lookup,
1036 .fib6_lookup = fib6_lookup,
1037 .fib6_select_path = fib6_select_path,
1038 .ip6_mtu_from_fib6 = ip6_mtu_from_fib6,
1039 .fib6_nh_init = fib6_nh_init,
1040 .fib6_nh_release = fib6_nh_release,
1041 .fib6_nh_release_dsts = fib6_nh_release_dsts,
1042 .fib6_update_sernum = fib6_update_sernum_stub,
1043 .fib6_rt_update = fib6_rt_update,
1044 .ip6_del_rt = ip6_del_rt,
1045 .udpv6_encap_enable = udpv6_encap_enable,
1046 .ndisc_send_na = ndisc_send_na,
1047 #if IS_ENABLED(CONFIG_XFRM)
1048 .xfrm6_local_rxpmtu = xfrm6_local_rxpmtu,
1049 .xfrm6_udp_encap_rcv = xfrm6_udp_encap_rcv,
1050 .xfrm6_rcv_encap = xfrm6_rcv_encap,
1053 .ipv6_fragment = ip6_fragment,
1054 .ipv6_dev_find = ipv6_dev_find,
1057 static const struct ipv6_bpf_stub ipv6_bpf_stub_impl = {
1058 .inet6_bind = __inet6_bind,
1059 .udp6_lib_lookup = __udp6_lib_lookup,
1060 .ipv6_setsockopt = do_ipv6_setsockopt,
1063 static int __init inet6_init(void)
1065 struct list_head *r;
1068 sock_skb_cb_check_size(sizeof(struct inet6_skb_parm));
1070 /* Register the socket-side information for inet6_create. */
1071 for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
1074 if (disable_ipv6_mod) {
1075 pr_info("Loaded, but administratively disabled, reboot required to enable\n");
1079 raw_hashinfo_init(&raw_v6_hashinfo);
1081 err = proto_register(&tcpv6_prot, 1);
1085 err = proto_register(&udpv6_prot, 1);
1087 goto out_unregister_tcp_proto;
1089 err = proto_register(&udplitev6_prot, 1);
1091 goto out_unregister_udp_proto;
1093 err = proto_register(&rawv6_prot, 1);
1095 goto out_unregister_udplite_proto;
1097 err = proto_register(&pingv6_prot, 1);
1099 goto out_unregister_raw_proto;
1101 /* We MUST register RAW sockets before we create the ICMP6,
1102 * IGMP6, or NDISC control sockets.
1106 goto out_unregister_ping_proto;
1108 /* Register the family here so that the init calls below will
1109 * be able to create sockets. (?? is this dangerous ??)
1111 err = sock_register(&inet6_family_ops);
1113 goto out_sock_register_fail;
1116 * ipngwg API draft makes clear that the correct semantics
1117 * for TCP and UDP is to consider one TCP and UDP instance
1118 * in a host available by both INET and INET6 APIs and
1119 * able to communicate via both network protocols.
1122 err = register_pernet_subsys(&inet6_net_ops);
1124 goto register_pernet_fail;
1125 err = ip6_mr_init();
1128 err = icmpv6_init();
1138 err = ipv6_netfilter_init();
1140 goto netfilter_fail;
1141 /* Create /proc/foo6 entries. */
1142 #ifdef CONFIG_PROC_FS
1144 if (raw6_proc_init())
1145 goto proc_raw6_fail;
1146 if (udplite6_proc_init())
1147 goto proc_udplite6_fail;
1148 if (ipv6_misc_proc_init())
1149 goto proc_misc6_fail;
1150 if (if6_proc_init())
1153 err = ip6_route_init();
1155 goto ip6_route_fail;
1156 err = ndisc_late_init();
1158 goto ndisc_late_fail;
1159 err = ip6_flowlabel_init();
1161 goto ip6_flowlabel_fail;
1162 err = ipv6_anycast_init();
1164 goto ipv6_anycast_fail;
1165 err = addrconf_init();
1169 /* Init v6 extension headers. */
1170 err = ipv6_exthdrs_init();
1172 goto ipv6_exthdrs_fail;
1174 err = ipv6_frag_init();
1176 goto ipv6_frag_fail;
1178 /* Init v6 transport protocols. */
1183 err = udplitev6_init();
1185 goto udplitev6_fail;
1187 err = udpv6_offload_init();
1189 goto udpv6_offload_fail;
1195 err = ipv6_packet_init();
1197 goto ipv6_packet_fail;
1199 err = pingv6_init();
1203 err = calipso_init();
1219 err = igmp6_late_init();
1221 goto igmp6_late_err;
1223 #ifdef CONFIG_SYSCTL
1224 err = ipv6_sysctl_register();
1229 /* ensure that ipv6 stubs are visible only after ipv6 is ready */
1231 ipv6_stub = &ipv6_stub_impl;
1232 ipv6_bpf_stub = &ipv6_bpf_stub_impl;
1236 #ifdef CONFIG_SYSCTL
1238 igmp6_late_cleanup();
1251 ipv6_packet_cleanup();
1255 udpv6_offload_exit();
1263 ipv6_exthdrs_exit();
1267 ipv6_anycast_cleanup();
1269 ip6_flowlabel_cleanup();
1271 ndisc_late_cleanup();
1273 ip6_route_cleanup();
1275 #ifdef CONFIG_PROC_FS
1278 ipv6_misc_proc_exit();
1280 udplite6_proc_exit();
1285 ipv6_netfilter_fini();
1295 unregister_pernet_subsys(&inet6_net_ops);
1296 register_pernet_fail:
1297 sock_unregister(PF_INET6);
1298 rtnl_unregister_all(PF_INET6);
1299 out_sock_register_fail:
1301 out_unregister_ping_proto:
1302 proto_unregister(&pingv6_prot);
1303 out_unregister_raw_proto:
1304 proto_unregister(&rawv6_prot);
1305 out_unregister_udplite_proto:
1306 proto_unregister(&udplitev6_prot);
1307 out_unregister_udp_proto:
1308 proto_unregister(&udpv6_prot);
1309 out_unregister_tcp_proto:
1310 proto_unregister(&tcpv6_prot);
1313 module_init(inet6_init);
1315 MODULE_ALIAS_NETPROTO(PF_INET6);