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>
67 #include <linux/uaccess.h>
68 #include <linux/mroute6.h>
70 #include "ip6_offload.h"
72 MODULE_AUTHOR("Cast of dozens");
73 MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
74 MODULE_LICENSE("GPL");
76 /* The inetsw6 table contains everything that inet6_create needs to
79 static struct list_head inetsw6[SOCK_MAX];
80 static DEFINE_SPINLOCK(inetsw6_lock);
82 struct ipv6_params ipv6_defaults = {
87 static int disable_ipv6_mod;
89 module_param_named(disable, disable_ipv6_mod, int, 0444);
90 MODULE_PARM_DESC(disable, "Disable IPv6 module such that it is non-functional");
92 module_param_named(disable_ipv6, ipv6_defaults.disable_ipv6, int, 0444);
93 MODULE_PARM_DESC(disable_ipv6, "Disable IPv6 on all interfaces");
95 module_param_named(autoconf, ipv6_defaults.autoconf, int, 0444);
96 MODULE_PARM_DESC(autoconf, "Enable IPv6 address autoconfiguration on all interfaces");
98 bool ipv6_mod_enabled(void)
100 return disable_ipv6_mod == 0;
102 EXPORT_SYMBOL_GPL(ipv6_mod_enabled);
104 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
106 const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
108 return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
111 void inet6_sock_destruct(struct sock *sk)
113 inet6_cleanup_sock(sk);
114 inet_sock_destruct(sk);
116 EXPORT_SYMBOL_GPL(inet6_sock_destruct);
118 static int inet6_create(struct net *net, struct socket *sock, int protocol,
121 struct inet_sock *inet;
122 struct ipv6_pinfo *np;
124 struct inet_protosw *answer;
125 struct proto *answer_prot;
126 unsigned char answer_flags;
127 int try_loading_module = 0;
130 if (protocol < 0 || protocol >= IPPROTO_MAX)
133 /* Look for the requested type/protocol pair. */
135 err = -ESOCKTNOSUPPORT;
137 list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
140 /* Check the non-wild match. */
141 if (protocol == answer->protocol) {
142 if (protocol != IPPROTO_IP)
145 /* Check for the two wild cases. */
146 if (IPPROTO_IP == protocol) {
147 protocol = answer->protocol;
150 if (IPPROTO_IP == answer->protocol)
153 err = -EPROTONOSUPPORT;
157 if (try_loading_module < 2) {
160 * Be more specific, e.g. net-pf-10-proto-132-type-1
161 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
163 if (++try_loading_module == 1)
164 request_module("net-pf-%d-proto-%d-type-%d",
165 PF_INET6, protocol, sock->type);
167 * Fall back to generic, e.g. net-pf-10-proto-132
168 * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
171 request_module("net-pf-%d-proto-%d",
173 goto lookup_protocol;
179 if (sock->type == SOCK_RAW && !kern &&
180 !ns_capable(net->user_ns, CAP_NET_RAW))
183 sock->ops = answer->ops;
184 answer_prot = answer->prot;
185 answer_flags = answer->flags;
188 WARN_ON(!answer_prot->slab);
191 sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot, kern);
195 sock_init_data(sock, sk);
198 if (INET_PROTOSW_REUSE & answer_flags)
199 sk->sk_reuse = SK_CAN_REUSE;
202 inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
204 if (SOCK_RAW == sock->type) {
205 inet->inet_num = protocol;
206 if (IPPROTO_RAW == protocol)
210 sk->sk_destruct = inet6_sock_destruct;
211 sk->sk_family = PF_INET6;
212 sk->sk_protocol = protocol;
214 sk->sk_backlog_rcv = answer->prot->backlog_rcv;
216 inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
218 np->mcast_hops = IPV6_DEFAULT_MCASTHOPS;
221 np->pmtudisc = IPV6_PMTUDISC_WANT;
222 np->repflow = net->ipv6.sysctl.flowlabel_reflect & FLOWLABEL_REFLECT_ESTABLISHED;
223 sk->sk_ipv6only = net->ipv6.sysctl.bindv6only;
225 /* Init the ipv4 part of the socket since we can have sockets
226 * using v6 API for ipv4.
233 RCU_INIT_POINTER(inet->mc_list, NULL);
236 if (READ_ONCE(net->ipv4.sysctl_ip_no_pmtu_disc))
237 inet->pmtudisc = IP_PMTUDISC_DONT;
239 inet->pmtudisc = IP_PMTUDISC_WANT;
241 * Increment only the relevant sk_prot->socks debug field, this changes
242 * the previous behaviour of incrementing both the equivalent to
243 * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
245 * This allows better debug granularity as we'll know exactly how many
246 * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
247 * transport protocol socks. -acme
249 sk_refcnt_debug_inc(sk);
251 if (inet->inet_num) {
252 /* It assumes that any protocol which allows
253 * the user to assign a number at socket
254 * creation time automatically shares.
256 inet->inet_sport = htons(inet->inet_num);
257 err = sk->sk_prot->hash(sk);
259 sk_common_release(sk);
263 if (sk->sk_prot->init) {
264 err = sk->sk_prot->init(sk);
266 sk_common_release(sk);
272 err = BPF_CGROUP_RUN_PROG_INET_SOCK(sk);
274 sk_common_release(sk);
285 static int __inet6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len,
288 struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr;
289 struct inet_sock *inet = inet_sk(sk);
290 struct ipv6_pinfo *np = inet6_sk(sk);
291 struct net *net = sock_net(sk);
298 if (addr->sin6_family != AF_INET6)
299 return -EAFNOSUPPORT;
301 addr_type = ipv6_addr_type(&addr->sin6_addr);
302 if ((addr_type & IPV6_ADDR_MULTICAST) && sk->sk_type == SOCK_STREAM)
305 snum = ntohs(addr->sin6_port);
306 if (!(flags & BIND_NO_CAP_NET_BIND_SERVICE) &&
307 snum && inet_port_requires_bind_service(net, snum) &&
308 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
311 if (flags & BIND_WITH_LOCK)
314 /* Check these errors (active socket, double bind). */
315 if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
320 /* Check if the address belongs to the host. */
321 if (addr_type == IPV6_ADDR_MAPPED) {
322 struct net_device *dev = NULL;
325 /* Binding to v4-mapped address on a v6-only socket
328 if (sk->sk_ipv6only) {
334 if (sk->sk_bound_dev_if) {
335 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
342 /* Reproduce AF_INET checks to make the bindings consistent */
343 v4addr = addr->sin6_addr.s6_addr32[3];
344 chk_addr_ret = inet_addr_type_dev_table(net, dev, v4addr);
347 if (!inet_can_nonlocal_bind(net, inet) &&
348 v4addr != htonl(INADDR_ANY) &&
349 chk_addr_ret != RTN_LOCAL &&
350 chk_addr_ret != RTN_MULTICAST &&
351 chk_addr_ret != RTN_BROADCAST) {
352 err = -EADDRNOTAVAIL;
356 if (addr_type != IPV6_ADDR_ANY) {
357 struct net_device *dev = NULL;
360 if (__ipv6_addr_needs_scope_id(addr_type)) {
361 if (addr_len >= sizeof(struct sockaddr_in6) &&
362 addr->sin6_scope_id) {
363 /* Override any existing binding, if another one
364 * is supplied by user.
366 sk->sk_bound_dev_if = addr->sin6_scope_id;
369 /* Binding to link-local address requires an interface */
370 if (!sk->sk_bound_dev_if) {
376 if (sk->sk_bound_dev_if) {
377 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
384 /* ipv4 addr of the socket is invalid. Only the
385 * unspecified and mapped address have a v4 equivalent.
387 v4addr = LOOPBACK4_IPV6;
388 if (!(addr_type & IPV6_ADDR_MULTICAST)) {
389 if (!ipv6_can_nonlocal_bind(net, inet) &&
390 !ipv6_chk_addr(net, &addr->sin6_addr,
392 err = -EADDRNOTAVAIL;
400 inet->inet_rcv_saddr = v4addr;
401 inet->inet_saddr = v4addr;
403 sk->sk_v6_rcv_saddr = addr->sin6_addr;
405 if (!(addr_type & IPV6_ADDR_MULTICAST))
406 np->saddr = addr->sin6_addr;
408 saved_ipv6only = sk->sk_ipv6only;
409 if (addr_type != IPV6_ADDR_ANY && addr_type != IPV6_ADDR_MAPPED)
412 /* Make sure we are allowed to bind here. */
413 if (snum || !(inet->bind_address_no_port ||
414 (flags & BIND_FORCE_ADDRESS_NO_PORT))) {
415 if (sk->sk_prot->get_port(sk, snum)) {
416 sk->sk_ipv6only = saved_ipv6only;
417 inet_reset_saddr(sk);
421 if (!(flags & BIND_FROM_BPF)) {
422 err = BPF_CGROUP_RUN_PROG_INET6_POST_BIND(sk);
424 sk->sk_ipv6only = saved_ipv6only;
425 inet_reset_saddr(sk);
431 if (addr_type != IPV6_ADDR_ANY)
432 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
434 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
435 inet->inet_sport = htons(inet->inet_num);
436 inet->inet_dport = 0;
437 inet->inet_daddr = 0;
439 if (flags & BIND_WITH_LOCK)
447 /* bind for INET6 API */
448 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
450 struct sock *sk = sock->sk;
451 u32 flags = BIND_WITH_LOCK;
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_INET_BIND_LOCK(sk, uaddr,
465 CGROUP_INET6_BIND, &flags);
469 return __inet6_bind(sk, uaddr, addr_len, flags);
471 EXPORT_SYMBOL(inet6_bind);
473 int inet6_release(struct socket *sock)
475 struct sock *sk = sock->sk;
481 ipv6_sock_mc_close(sk);
484 ipv6_sock_ac_close(sk);
486 return inet_release(sock);
488 EXPORT_SYMBOL(inet6_release);
490 void inet6_destroy_sock(struct sock *sk)
492 struct ipv6_pinfo *np = inet6_sk(sk);
494 struct ipv6_txoptions *opt;
496 /* Release rx options */
498 skb = xchg(&np->pktoptions, NULL);
501 skb = xchg(&np->rxpmtu, NULL);
504 /* Free flowlabels */
505 fl6_free_socklist(sk);
507 /* Free tx options */
509 opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL);
511 atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
515 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
517 void inet6_cleanup_sock(struct sock *sk)
519 inet6_destroy_sock(sk);
521 EXPORT_SYMBOL_GPL(inet6_cleanup_sock);
524 * This does both peername and sockname.
526 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
529 struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
530 struct sock *sk = sock->sk;
531 struct inet_sock *inet = inet_sk(sk);
532 struct ipv6_pinfo *np = inet6_sk(sk);
534 sin->sin6_family = AF_INET6;
535 sin->sin6_flowinfo = 0;
536 sin->sin6_scope_id = 0;
539 if (!inet->inet_dport ||
540 (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
545 sin->sin6_port = inet->inet_dport;
546 sin->sin6_addr = sk->sk_v6_daddr;
548 sin->sin6_flowinfo = np->flow_label;
549 BPF_CGROUP_RUN_SA_PROG(sk, (struct sockaddr *)sin,
550 CGROUP_INET6_GETPEERNAME);
552 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
553 sin->sin6_addr = np->saddr;
555 sin->sin6_addr = sk->sk_v6_rcv_saddr;
556 sin->sin6_port = inet->inet_sport;
557 BPF_CGROUP_RUN_SA_PROG(sk, (struct sockaddr *)sin,
558 CGROUP_INET6_GETSOCKNAME);
560 sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
561 sk->sk_bound_dev_if);
565 EXPORT_SYMBOL(inet6_getname);
567 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
569 void __user *argp = (void __user *)arg;
570 struct sock *sk = sock->sk;
571 struct net *net = sock_net(sk);
576 struct in6_rtmsg rtmsg;
578 if (copy_from_user(&rtmsg, argp, sizeof(rtmsg)))
580 return ipv6_route_ioctl(net, cmd, &rtmsg);
583 return addrconf_add_ifaddr(net, argp);
585 return addrconf_del_ifaddr(net, argp);
587 return addrconf_set_dstaddr(net, argp);
589 if (!sk->sk_prot->ioctl)
591 return sk->sk_prot->ioctl(sk, cmd, arg);
596 EXPORT_SYMBOL(inet6_ioctl);
599 struct compat_in6_rtmsg {
600 struct in6_addr rtmsg_dst;
601 struct in6_addr rtmsg_src;
602 struct in6_addr rtmsg_gateway;
612 static int inet6_compat_routing_ioctl(struct sock *sk, unsigned int cmd,
613 struct compat_in6_rtmsg __user *ur)
617 if (copy_from_user(&rt.rtmsg_dst, &ur->rtmsg_dst,
618 3 * sizeof(struct in6_addr)) ||
619 get_user(rt.rtmsg_type, &ur->rtmsg_type) ||
620 get_user(rt.rtmsg_dst_len, &ur->rtmsg_dst_len) ||
621 get_user(rt.rtmsg_src_len, &ur->rtmsg_src_len) ||
622 get_user(rt.rtmsg_metric, &ur->rtmsg_metric) ||
623 get_user(rt.rtmsg_info, &ur->rtmsg_info) ||
624 get_user(rt.rtmsg_flags, &ur->rtmsg_flags) ||
625 get_user(rt.rtmsg_ifindex, &ur->rtmsg_ifindex))
629 return ipv6_route_ioctl(sock_net(sk), cmd, &rt);
632 int inet6_compat_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
634 void __user *argp = compat_ptr(arg);
635 struct sock *sk = sock->sk;
640 return inet6_compat_routing_ioctl(sk, cmd, argp);
645 EXPORT_SYMBOL_GPL(inet6_compat_ioctl);
646 #endif /* CONFIG_COMPAT */
648 INDIRECT_CALLABLE_DECLARE(int udpv6_sendmsg(struct sock *, struct msghdr *,
650 int inet6_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
652 struct sock *sk = sock->sk;
654 if (unlikely(inet_send_prepare(sk)))
657 return INDIRECT_CALL_2(sk->sk_prot->sendmsg, tcp_sendmsg, udpv6_sendmsg,
661 INDIRECT_CALLABLE_DECLARE(int udpv6_recvmsg(struct sock *, struct msghdr *,
662 size_t, int, int, int *));
663 int inet6_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
666 struct sock *sk = sock->sk;
670 if (likely(!(flags & MSG_ERRQUEUE)))
671 sock_rps_record_flow(sk);
673 err = INDIRECT_CALL_2(sk->sk_prot->recvmsg, tcp_recvmsg, udpv6_recvmsg,
674 sk, msg, size, flags & MSG_DONTWAIT,
675 flags & ~MSG_DONTWAIT, &addr_len);
677 msg->msg_namelen = addr_len;
681 const struct proto_ops inet6_stream_ops = {
683 .owner = THIS_MODULE,
684 .release = inet6_release,
686 .connect = inet_stream_connect, /* ok */
687 .socketpair = sock_no_socketpair, /* a do nothing */
688 .accept = inet_accept, /* ok */
689 .getname = inet6_getname,
690 .poll = tcp_poll, /* ok */
691 .ioctl = inet6_ioctl, /* must change */
692 .gettstamp = sock_gettstamp,
693 .listen = inet_listen, /* ok */
694 .shutdown = inet_shutdown, /* ok */
695 .setsockopt = sock_common_setsockopt, /* ok */
696 .getsockopt = sock_common_getsockopt, /* ok */
697 .sendmsg = inet6_sendmsg, /* retpoline's sake */
698 .recvmsg = inet6_recvmsg, /* retpoline's sake */
702 .sendpage = inet_sendpage,
703 .sendmsg_locked = tcp_sendmsg_locked,
704 .sendpage_locked = tcp_sendpage_locked,
705 .splice_read = tcp_splice_read,
706 .read_sock = tcp_read_sock,
707 .peek_len = tcp_peek_len,
709 .compat_ioctl = inet6_compat_ioctl,
711 .set_rcvlowat = tcp_set_rcvlowat,
714 const struct proto_ops inet6_dgram_ops = {
716 .owner = THIS_MODULE,
717 .release = inet6_release,
719 .connect = inet_dgram_connect, /* ok */
720 .socketpair = sock_no_socketpair, /* a do nothing */
721 .accept = sock_no_accept, /* a do nothing */
722 .getname = inet6_getname,
723 .poll = udp_poll, /* ok */
724 .ioctl = inet6_ioctl, /* must change */
725 .gettstamp = sock_gettstamp,
726 .listen = sock_no_listen, /* ok */
727 .shutdown = inet_shutdown, /* ok */
728 .setsockopt = sock_common_setsockopt, /* ok */
729 .getsockopt = sock_common_getsockopt, /* ok */
730 .sendmsg = inet6_sendmsg, /* retpoline's sake */
731 .recvmsg = inet6_recvmsg, /* retpoline's sake */
732 .read_sock = udp_read_sock,
733 .mmap = sock_no_mmap,
734 .sendpage = sock_no_sendpage,
735 .set_peek_off = sk_set_peek_off,
737 .compat_ioctl = inet6_compat_ioctl,
741 static const struct net_proto_family inet6_family_ops = {
743 .create = inet6_create,
744 .owner = THIS_MODULE,
747 int inet6_register_protosw(struct inet_protosw *p)
749 struct list_head *lh;
750 struct inet_protosw *answer;
751 struct list_head *last_perm;
752 int protocol = p->protocol;
755 spin_lock_bh(&inetsw6_lock);
758 if (p->type >= SOCK_MAX)
761 /* If we are trying to override a permanent protocol, bail. */
764 last_perm = &inetsw6[p->type];
765 list_for_each(lh, &inetsw6[p->type]) {
766 answer = list_entry(lh, struct inet_protosw, list);
768 /* Check only the non-wild match. */
769 if (INET_PROTOSW_PERMANENT & answer->flags) {
770 if (protocol == answer->protocol)
780 /* Add the new entry after the last permanent entry if any, so that
781 * the new entry does not override a permanent entry when matched with
782 * a wild-card protocol. But it is allowed to override any existing
783 * non-permanent entry. This means that when we remove this entry, the
784 * system automatically returns to the old behavior.
786 list_add_rcu(&p->list, last_perm);
789 spin_unlock_bh(&inetsw6_lock);
793 pr_err("Attempt to override permanent protocol %d\n", protocol);
797 pr_err("Ignoring attempt to register invalid socket type %d\n",
801 EXPORT_SYMBOL(inet6_register_protosw);
804 inet6_unregister_protosw(struct inet_protosw *p)
806 if (INET_PROTOSW_PERMANENT & p->flags) {
807 pr_err("Attempt to unregister permanent protocol %d\n",
810 spin_lock_bh(&inetsw6_lock);
811 list_del_rcu(&p->list);
812 spin_unlock_bh(&inetsw6_lock);
817 EXPORT_SYMBOL(inet6_unregister_protosw);
819 int inet6_sk_rebuild_header(struct sock *sk)
821 struct ipv6_pinfo *np = inet6_sk(sk);
822 struct dst_entry *dst;
824 dst = __sk_dst_check(sk, np->dst_cookie);
827 struct inet_sock *inet = inet_sk(sk);
828 struct in6_addr *final_p, final;
831 memset(&fl6, 0, sizeof(fl6));
832 fl6.flowi6_proto = sk->sk_protocol;
833 fl6.daddr = sk->sk_v6_daddr;
834 fl6.saddr = np->saddr;
835 fl6.flowlabel = np->flow_label;
836 fl6.flowi6_oif = sk->sk_bound_dev_if;
837 fl6.flowi6_mark = sk->sk_mark;
838 fl6.fl6_dport = inet->inet_dport;
839 fl6.fl6_sport = inet->inet_sport;
840 fl6.flowi6_uid = sk->sk_uid;
841 security_sk_classify_flow(sk, flowi6_to_flowi_common(&fl6));
844 final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt),
848 dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p);
850 sk->sk_route_caps = 0;
851 sk->sk_err_soft = -PTR_ERR(dst);
855 ip6_dst_store(sk, dst, NULL, NULL);
860 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
862 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb,
863 const struct inet6_skb_parm *opt)
865 const struct ipv6_pinfo *np = inet6_sk(sk);
868 if (((opt->flags & IP6SKB_HOPBYHOP) &&
869 (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) ||
870 (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) &&
871 np->rxopt.bits.rxflow) ||
872 (opt->srcrt && (np->rxopt.bits.srcrt ||
873 np->rxopt.bits.osrcrt)) ||
874 ((opt->dst1 || opt->dst0) &&
875 (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
880 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
882 static struct packet_type ipv6_packet_type __read_mostly = {
883 .type = cpu_to_be16(ETH_P_IPV6),
885 .list_func = ipv6_list_rcv,
888 static int __init ipv6_packet_init(void)
890 dev_add_pack(&ipv6_packet_type);
894 static void ipv6_packet_cleanup(void)
896 dev_remove_pack(&ipv6_packet_type);
899 static int __net_init ipv6_init_mibs(struct net *net)
903 net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib);
904 if (!net->mib.udp_stats_in6)
906 net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib);
907 if (!net->mib.udplite_stats_in6)
908 goto err_udplite_mib;
909 net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib);
910 if (!net->mib.ipv6_statistics)
913 for_each_possible_cpu(i) {
914 struct ipstats_mib *af_inet6_stats;
915 af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i);
916 u64_stats_init(&af_inet6_stats->syncp);
920 net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib);
921 if (!net->mib.icmpv6_statistics)
923 net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
925 if (!net->mib.icmpv6msg_statistics)
926 goto err_icmpmsg_mib;
930 free_percpu(net->mib.icmpv6_statistics);
932 free_percpu(net->mib.ipv6_statistics);
934 free_percpu(net->mib.udplite_stats_in6);
936 free_percpu(net->mib.udp_stats_in6);
940 static void ipv6_cleanup_mibs(struct net *net)
942 free_percpu(net->mib.udp_stats_in6);
943 free_percpu(net->mib.udplite_stats_in6);
944 free_percpu(net->mib.ipv6_statistics);
945 free_percpu(net->mib.icmpv6_statistics);
946 kfree(net->mib.icmpv6msg_statistics);
949 static int __net_init inet6_net_init(struct net *net)
953 net->ipv6.sysctl.bindv6only = 0;
954 net->ipv6.sysctl.icmpv6_time = 1*HZ;
955 net->ipv6.sysctl.icmpv6_echo_ignore_all = 0;
956 net->ipv6.sysctl.icmpv6_echo_ignore_multicast = 0;
957 net->ipv6.sysctl.icmpv6_echo_ignore_anycast = 0;
959 /* By default, rate limit error messages.
960 * Except for pmtu discovery, it would break it.
961 * proc_do_large_bitmap needs pointer to the bitmap.
963 bitmap_set(net->ipv6.sysctl.icmpv6_ratemask, 0, ICMPV6_ERRMSG_MAX + 1);
964 bitmap_clear(net->ipv6.sysctl.icmpv6_ratemask, ICMPV6_PKT_TOOBIG, 1);
965 net->ipv6.sysctl.icmpv6_ratemask_ptr = net->ipv6.sysctl.icmpv6_ratemask;
967 net->ipv6.sysctl.flowlabel_consistency = 1;
968 net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS;
969 net->ipv6.sysctl.idgen_retries = 3;
970 net->ipv6.sysctl.idgen_delay = 1 * HZ;
971 net->ipv6.sysctl.flowlabel_state_ranges = 0;
972 net->ipv6.sysctl.max_dst_opts_cnt = IP6_DEFAULT_MAX_DST_OPTS_CNT;
973 net->ipv6.sysctl.max_hbh_opts_cnt = IP6_DEFAULT_MAX_HBH_OPTS_CNT;
974 net->ipv6.sysctl.max_dst_opts_len = IP6_DEFAULT_MAX_DST_OPTS_LEN;
975 net->ipv6.sysctl.max_hbh_opts_len = IP6_DEFAULT_MAX_HBH_OPTS_LEN;
976 net->ipv6.sysctl.fib_notify_on_flag_change = 0;
977 atomic_set(&net->ipv6.fib6_sernum, 1);
979 net->ipv6.sysctl.ioam6_id = IOAM6_DEFAULT_ID;
980 net->ipv6.sysctl.ioam6_id_wide = IOAM6_DEFAULT_ID_WIDE;
982 err = ipv6_init_mibs(net);
985 #ifdef CONFIG_PROC_FS
986 err = udp6_proc_init(net);
989 err = tcp6_proc_init(net);
992 err = ac6_proc_init(net);
998 #ifdef CONFIG_PROC_FS
1000 tcp6_proc_exit(net);
1002 udp6_proc_exit(net);
1004 ipv6_cleanup_mibs(net);
1009 static void __net_exit inet6_net_exit(struct net *net)
1011 #ifdef CONFIG_PROC_FS
1012 udp6_proc_exit(net);
1013 tcp6_proc_exit(net);
1016 ipv6_cleanup_mibs(net);
1019 static struct pernet_operations inet6_net_ops = {
1020 .init = inet6_net_init,
1021 .exit = inet6_net_exit,
1024 static int ipv6_route_input(struct sk_buff *skb)
1026 ip6_route_input(skb);
1027 return skb_dst(skb)->error;
1030 static const struct ipv6_stub ipv6_stub_impl = {
1031 .ipv6_sock_mc_join = ipv6_sock_mc_join,
1032 .ipv6_sock_mc_drop = ipv6_sock_mc_drop,
1033 .ipv6_dst_lookup_flow = ip6_dst_lookup_flow,
1034 .ipv6_route_input = ipv6_route_input,
1035 .fib6_get_table = fib6_get_table,
1036 .fib6_table_lookup = fib6_table_lookup,
1037 .fib6_lookup = fib6_lookup,
1038 .fib6_select_path = fib6_select_path,
1039 .ip6_mtu_from_fib6 = ip6_mtu_from_fib6,
1040 .fib6_nh_init = fib6_nh_init,
1041 .fib6_nh_release = fib6_nh_release,
1042 .fib6_nh_release_dsts = fib6_nh_release_dsts,
1043 .fib6_update_sernum = fib6_update_sernum_stub,
1044 .fib6_rt_update = fib6_rt_update,
1045 .ip6_del_rt = ip6_del_rt,
1046 .udpv6_encap_enable = udpv6_encap_enable,
1047 .ndisc_send_na = ndisc_send_na,
1048 #if IS_ENABLED(CONFIG_XFRM)
1049 .xfrm6_local_rxpmtu = xfrm6_local_rxpmtu,
1050 .xfrm6_udp_encap_rcv = xfrm6_udp_encap_rcv,
1051 .xfrm6_rcv_encap = xfrm6_rcv_encap,
1054 .ipv6_fragment = ip6_fragment,
1055 .ipv6_dev_find = ipv6_dev_find,
1058 static const struct ipv6_bpf_stub ipv6_bpf_stub_impl = {
1059 .inet6_bind = __inet6_bind,
1060 .udp6_lib_lookup = __udp6_lib_lookup,
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 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);