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 struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
107 const int offset = sk->sk_prot->ipv6_pinfo_offset;
109 return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
112 void inet6_sock_destruct(struct sock *sk)
114 inet6_cleanup_sock(sk);
115 inet_sock_destruct(sk);
117 EXPORT_SYMBOL_GPL(inet6_sock_destruct);
119 static int inet6_create(struct net *net, struct socket *sock, int protocol,
122 struct inet_sock *inet;
123 struct ipv6_pinfo *np;
125 struct inet_protosw *answer;
126 struct proto *answer_prot;
127 unsigned char answer_flags;
128 int try_loading_module = 0;
131 if (protocol < 0 || protocol >= IPPROTO_MAX)
134 /* Look for the requested type/protocol pair. */
136 err = -ESOCKTNOSUPPORT;
138 list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
141 /* Check the non-wild match. */
142 if (protocol == answer->protocol) {
143 if (protocol != IPPROTO_IP)
146 /* Check for the two wild cases. */
147 if (IPPROTO_IP == protocol) {
148 protocol = answer->protocol;
151 if (IPPROTO_IP == answer->protocol)
154 err = -EPROTONOSUPPORT;
158 if (try_loading_module < 2) {
161 * Be more specific, e.g. net-pf-10-proto-132-type-1
162 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
164 if (++try_loading_module == 1)
165 request_module("net-pf-%d-proto-%d-type-%d",
166 PF_INET6, protocol, sock->type);
168 * Fall back to generic, e.g. net-pf-10-proto-132
169 * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
172 request_module("net-pf-%d-proto-%d",
174 goto lookup_protocol;
180 if (sock->type == SOCK_RAW && !kern &&
181 !ns_capable(net->user_ns, CAP_NET_RAW))
184 sock->ops = answer->ops;
185 answer_prot = answer->prot;
186 answer_flags = answer->flags;
189 WARN_ON(!answer_prot->slab);
192 sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot, kern);
196 sock_init_data(sock, sk);
199 if (INET_PROTOSW_REUSE & answer_flags)
200 sk->sk_reuse = SK_CAN_REUSE;
202 if (INET_PROTOSW_ICSK & answer_flags)
203 inet_init_csk_locks(sk);
206 inet_assign_bit(IS_ICSK, sk, INET_PROTOSW_ICSK & answer_flags);
208 if (SOCK_RAW == sock->type) {
209 inet->inet_num = protocol;
210 if (IPPROTO_RAW == protocol)
211 inet_set_bit(HDRINCL, sk);
214 sk->sk_destruct = inet6_sock_destruct;
215 sk->sk_family = PF_INET6;
216 sk->sk_protocol = protocol;
218 sk->sk_backlog_rcv = answer->prot->backlog_rcv;
220 inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
222 np->mcast_hops = IPV6_DEFAULT_MCASTHOPS;
225 np->pmtudisc = IPV6_PMTUDISC_WANT;
226 np->repflow = net->ipv6.sysctl.flowlabel_reflect & FLOWLABEL_REFLECT_ESTABLISHED;
227 sk->sk_ipv6only = net->ipv6.sysctl.bindv6only;
228 sk->sk_txrehash = READ_ONCE(net->core.sysctl_txrehash);
230 /* Init the ipv4 part of the socket since we can have sockets
231 * using v6 API for ipv4.
235 inet_set_bit(MC_LOOP, sk);
238 RCU_INIT_POINTER(inet->mc_list, NULL);
241 if (READ_ONCE(net->ipv4.sysctl_ip_no_pmtu_disc))
242 inet->pmtudisc = IP_PMTUDISC_DONT;
244 inet->pmtudisc = IP_PMTUDISC_WANT;
246 if (inet->inet_num) {
247 /* It assumes that any protocol which allows
248 * the user to assign a number at socket
249 * creation time automatically shares.
251 inet->inet_sport = htons(inet->inet_num);
252 err = sk->sk_prot->hash(sk);
254 sk_common_release(sk);
258 if (sk->sk_prot->init) {
259 err = sk->sk_prot->init(sk);
261 sk_common_release(sk);
267 err = BPF_CGROUP_RUN_PROG_INET_SOCK(sk);
269 sk_common_release(sk);
280 static int __inet6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len,
283 struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr;
284 struct inet_sock *inet = inet_sk(sk);
285 struct ipv6_pinfo *np = inet6_sk(sk);
286 struct net *net = sock_net(sk);
293 if (addr->sin6_family != AF_INET6)
294 return -EAFNOSUPPORT;
296 addr_type = ipv6_addr_type(&addr->sin6_addr);
297 if ((addr_type & IPV6_ADDR_MULTICAST) && sk->sk_type == SOCK_STREAM)
300 snum = ntohs(addr->sin6_port);
301 if (!(flags & BIND_NO_CAP_NET_BIND_SERVICE) &&
302 snum && inet_port_requires_bind_service(net, snum) &&
303 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
306 if (flags & BIND_WITH_LOCK)
309 /* Check these errors (active socket, double bind). */
310 if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
315 /* Check if the address belongs to the host. */
316 if (addr_type == IPV6_ADDR_MAPPED) {
317 struct net_device *dev = NULL;
320 /* Binding to v4-mapped address on a v6-only socket
323 if (ipv6_only_sock(sk)) {
329 if (sk->sk_bound_dev_if) {
330 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
337 /* Reproduce AF_INET checks to make the bindings consistent */
338 v4addr = addr->sin6_addr.s6_addr32[3];
339 chk_addr_ret = inet_addr_type_dev_table(net, dev, v4addr);
342 if (!inet_addr_valid_or_nonlocal(net, inet, v4addr,
344 err = -EADDRNOTAVAIL;
348 if (addr_type != IPV6_ADDR_ANY) {
349 struct net_device *dev = NULL;
352 if (__ipv6_addr_needs_scope_id(addr_type)) {
353 if (addr_len >= sizeof(struct sockaddr_in6) &&
354 addr->sin6_scope_id) {
355 /* Override any existing binding, if another one
356 * is supplied by user.
358 sk->sk_bound_dev_if = addr->sin6_scope_id;
361 /* Binding to link-local address requires an interface */
362 if (!sk->sk_bound_dev_if) {
368 if (sk->sk_bound_dev_if) {
369 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
376 /* ipv4 addr of the socket is invalid. Only the
377 * unspecified and mapped address have a v4 equivalent.
379 v4addr = LOOPBACK4_IPV6;
380 if (!(addr_type & IPV6_ADDR_MULTICAST)) {
381 if (!ipv6_can_nonlocal_bind(net, inet) &&
382 !ipv6_chk_addr(net, &addr->sin6_addr,
384 err = -EADDRNOTAVAIL;
392 inet->inet_rcv_saddr = v4addr;
393 inet->inet_saddr = v4addr;
395 sk->sk_v6_rcv_saddr = addr->sin6_addr;
397 if (!(addr_type & IPV6_ADDR_MULTICAST))
398 np->saddr = addr->sin6_addr;
400 saved_ipv6only = sk->sk_ipv6only;
401 if (addr_type != IPV6_ADDR_ANY && addr_type != IPV6_ADDR_MAPPED)
404 /* Make sure we are allowed to bind here. */
405 if (snum || !(inet_test_bit(BIND_ADDRESS_NO_PORT, sk) ||
406 (flags & BIND_FORCE_ADDRESS_NO_PORT))) {
407 err = sk->sk_prot->get_port(sk, snum);
409 sk->sk_ipv6only = saved_ipv6only;
410 inet_reset_saddr(sk);
413 if (!(flags & BIND_FROM_BPF)) {
414 err = BPF_CGROUP_RUN_PROG_INET6_POST_BIND(sk);
416 sk->sk_ipv6only = saved_ipv6only;
417 inet_reset_saddr(sk);
418 if (sk->sk_prot->put_port)
419 sk->sk_prot->put_port(sk);
425 if (addr_type != IPV6_ADDR_ANY)
426 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
428 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
429 inet->inet_sport = htons(inet->inet_num);
430 inet->inet_dport = 0;
431 inet->inet_daddr = 0;
433 if (flags & BIND_WITH_LOCK)
441 int inet6_bind_sk(struct sock *sk, struct sockaddr *uaddr, int addr_len)
443 u32 flags = BIND_WITH_LOCK;
444 const struct proto *prot;
447 /* IPV6_ADDRFORM can change sk->sk_prot under us. */
448 prot = READ_ONCE(sk->sk_prot);
449 /* If the socket has its own bind function then use it. */
451 return prot->bind(sk, uaddr, addr_len);
453 if (addr_len < SIN6_LEN_RFC2133)
456 /* BPF prog is run before any checks are done so that if the prog
457 * changes context in a wrong way it will be caught.
459 err = BPF_CGROUP_RUN_PROG_INET_BIND_LOCK(sk, uaddr,
460 CGROUP_INET6_BIND, &flags);
464 return __inet6_bind(sk, uaddr, addr_len, flags);
467 /* bind for INET6 API */
468 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
470 return inet6_bind_sk(sock->sk, uaddr, addr_len);
472 EXPORT_SYMBOL(inet6_bind);
474 int inet6_release(struct socket *sock)
476 struct sock *sk = sock->sk;
482 ipv6_sock_mc_close(sk);
485 ipv6_sock_ac_close(sk);
487 return inet_release(sock);
489 EXPORT_SYMBOL(inet6_release);
491 void inet6_cleanup_sock(struct sock *sk)
493 struct ipv6_pinfo *np = inet6_sk(sk);
495 struct ipv6_txoptions *opt;
497 /* Release rx options */
499 skb = xchg(&np->pktoptions, NULL);
502 skb = xchg(&np->rxpmtu, NULL);
505 /* Free flowlabels */
506 fl6_free_socklist(sk);
508 /* Free tx options */
510 opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL);
512 atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
516 EXPORT_SYMBOL_GPL(inet6_cleanup_sock);
519 * This does both peername and sockname.
521 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
524 struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
525 struct sock *sk = sock->sk;
526 struct inet_sock *inet = inet_sk(sk);
527 struct ipv6_pinfo *np = inet6_sk(sk);
529 sin->sin6_family = AF_INET6;
530 sin->sin6_flowinfo = 0;
531 sin->sin6_scope_id = 0;
534 if (!inet->inet_dport ||
535 (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
540 sin->sin6_port = inet->inet_dport;
541 sin->sin6_addr = sk->sk_v6_daddr;
543 sin->sin6_flowinfo = np->flow_label;
544 BPF_CGROUP_RUN_SA_PROG(sk, (struct sockaddr *)sin,
545 CGROUP_INET6_GETPEERNAME);
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;
552 BPF_CGROUP_RUN_SA_PROG(sk, (struct sockaddr *)sin,
553 CGROUP_INET6_GETSOCKNAME);
555 sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
556 sk->sk_bound_dev_if);
560 EXPORT_SYMBOL(inet6_getname);
562 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
564 void __user *argp = (void __user *)arg;
565 struct sock *sk = sock->sk;
566 struct net *net = sock_net(sk);
567 const struct proto *prot;
572 struct in6_rtmsg rtmsg;
574 if (copy_from_user(&rtmsg, argp, sizeof(rtmsg)))
576 return ipv6_route_ioctl(net, cmd, &rtmsg);
579 return addrconf_add_ifaddr(net, argp);
581 return addrconf_del_ifaddr(net, argp);
583 return addrconf_set_dstaddr(net, argp);
585 /* IPV6_ADDRFORM can change sk->sk_prot under us. */
586 prot = READ_ONCE(sk->sk_prot);
589 return sk_ioctl(sk, cmd, (void __user *)arg);
594 EXPORT_SYMBOL(inet6_ioctl);
597 struct compat_in6_rtmsg {
598 struct in6_addr rtmsg_dst;
599 struct in6_addr rtmsg_src;
600 struct in6_addr rtmsg_gateway;
610 static int inet6_compat_routing_ioctl(struct sock *sk, unsigned int cmd,
611 struct compat_in6_rtmsg __user *ur)
615 if (copy_from_user(&rt.rtmsg_dst, &ur->rtmsg_dst,
616 3 * sizeof(struct in6_addr)) ||
617 get_user(rt.rtmsg_type, &ur->rtmsg_type) ||
618 get_user(rt.rtmsg_dst_len, &ur->rtmsg_dst_len) ||
619 get_user(rt.rtmsg_src_len, &ur->rtmsg_src_len) ||
620 get_user(rt.rtmsg_metric, &ur->rtmsg_metric) ||
621 get_user(rt.rtmsg_info, &ur->rtmsg_info) ||
622 get_user(rt.rtmsg_flags, &ur->rtmsg_flags) ||
623 get_user(rt.rtmsg_ifindex, &ur->rtmsg_ifindex))
627 return ipv6_route_ioctl(sock_net(sk), cmd, &rt);
630 int inet6_compat_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
632 void __user *argp = compat_ptr(arg);
633 struct sock *sk = sock->sk;
638 return inet6_compat_routing_ioctl(sk, cmd, argp);
643 EXPORT_SYMBOL_GPL(inet6_compat_ioctl);
644 #endif /* CONFIG_COMPAT */
646 INDIRECT_CALLABLE_DECLARE(int udpv6_sendmsg(struct sock *, struct msghdr *,
648 int inet6_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
650 struct sock *sk = sock->sk;
651 const struct proto *prot;
653 if (unlikely(inet_send_prepare(sk)))
656 /* IPV6_ADDRFORM can change sk->sk_prot under us. */
657 prot = READ_ONCE(sk->sk_prot);
658 return INDIRECT_CALL_2(prot->sendmsg, tcp_sendmsg, udpv6_sendmsg,
662 INDIRECT_CALLABLE_DECLARE(int udpv6_recvmsg(struct sock *, struct msghdr *,
663 size_t, int, int *));
664 int inet6_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
667 struct sock *sk = sock->sk;
668 const struct proto *prot;
672 if (likely(!(flags & MSG_ERRQUEUE)))
673 sock_rps_record_flow(sk);
675 /* IPV6_ADDRFORM can change sk->sk_prot under us. */
676 prot = READ_ONCE(sk->sk_prot);
677 err = INDIRECT_CALL_2(prot->recvmsg, tcp_recvmsg, udpv6_recvmsg,
678 sk, msg, size, flags, &addr_len);
680 msg->msg_namelen = addr_len;
684 const struct proto_ops inet6_stream_ops = {
686 .owner = THIS_MODULE,
687 .release = inet6_release,
689 .connect = inet_stream_connect, /* ok */
690 .socketpair = sock_no_socketpair, /* a do nothing */
691 .accept = inet_accept, /* ok */
692 .getname = inet6_getname,
693 .poll = tcp_poll, /* ok */
694 .ioctl = inet6_ioctl, /* must change */
695 .gettstamp = sock_gettstamp,
696 .listen = inet_listen, /* ok */
697 .shutdown = inet_shutdown, /* ok */
698 .setsockopt = sock_common_setsockopt, /* ok */
699 .getsockopt = sock_common_getsockopt, /* ok */
700 .sendmsg = inet6_sendmsg, /* retpoline's sake */
701 .recvmsg = inet6_recvmsg, /* retpoline's sake */
705 .splice_eof = inet_splice_eof,
706 .sendmsg_locked = tcp_sendmsg_locked,
707 .splice_read = tcp_splice_read,
708 .read_sock = tcp_read_sock,
709 .read_skb = tcp_read_skb,
710 .peek_len = tcp_peek_len,
712 .compat_ioctl = inet6_compat_ioctl,
714 .set_rcvlowat = tcp_set_rcvlowat,
717 const struct proto_ops inet6_dgram_ops = {
719 .owner = THIS_MODULE,
720 .release = inet6_release,
722 .connect = inet_dgram_connect, /* ok */
723 .socketpair = sock_no_socketpair, /* a do nothing */
724 .accept = sock_no_accept, /* a do nothing */
725 .getname = inet6_getname,
726 .poll = udp_poll, /* ok */
727 .ioctl = inet6_ioctl, /* must change */
728 .gettstamp = sock_gettstamp,
729 .listen = sock_no_listen, /* ok */
730 .shutdown = inet_shutdown, /* ok */
731 .setsockopt = sock_common_setsockopt, /* ok */
732 .getsockopt = sock_common_getsockopt, /* ok */
733 .sendmsg = inet6_sendmsg, /* retpoline's sake */
734 .recvmsg = inet6_recvmsg, /* retpoline's sake */
735 .read_skb = udp_read_skb,
736 .mmap = sock_no_mmap,
737 .set_peek_off = sk_set_peek_off,
739 .compat_ioctl = inet6_compat_ioctl,
743 static const struct net_proto_family inet6_family_ops = {
745 .create = inet6_create,
746 .owner = THIS_MODULE,
749 int inet6_register_protosw(struct inet_protosw *p)
751 struct list_head *lh;
752 struct inet_protosw *answer;
753 struct list_head *last_perm;
754 int protocol = p->protocol;
757 spin_lock_bh(&inetsw6_lock);
760 if (p->type >= SOCK_MAX)
763 /* If we are trying to override a permanent protocol, bail. */
766 last_perm = &inetsw6[p->type];
767 list_for_each(lh, &inetsw6[p->type]) {
768 answer = list_entry(lh, struct inet_protosw, list);
770 /* Check only the non-wild match. */
771 if (INET_PROTOSW_PERMANENT & answer->flags) {
772 if (protocol == answer->protocol)
782 /* Add the new entry after the last permanent entry if any, so that
783 * the new entry does not override a permanent entry when matched with
784 * a wild-card protocol. But it is allowed to override any existing
785 * non-permanent entry. This means that when we remove this entry, the
786 * system automatically returns to the old behavior.
788 list_add_rcu(&p->list, last_perm);
791 spin_unlock_bh(&inetsw6_lock);
795 pr_err("Attempt to override permanent protocol %d\n", protocol);
799 pr_err("Ignoring attempt to register invalid socket type %d\n",
803 EXPORT_SYMBOL(inet6_register_protosw);
806 inet6_unregister_protosw(struct inet_protosw *p)
808 if (INET_PROTOSW_PERMANENT & p->flags) {
809 pr_err("Attempt to unregister permanent protocol %d\n",
812 spin_lock_bh(&inetsw6_lock);
813 list_del_rcu(&p->list);
814 spin_unlock_bh(&inetsw6_lock);
819 EXPORT_SYMBOL(inet6_unregister_protosw);
821 int inet6_sk_rebuild_header(struct sock *sk)
823 struct ipv6_pinfo *np = inet6_sk(sk);
824 struct dst_entry *dst;
826 dst = __sk_dst_check(sk, np->dst_cookie);
829 struct inet_sock *inet = inet_sk(sk);
830 struct in6_addr *final_p, final;
833 memset(&fl6, 0, sizeof(fl6));
834 fl6.flowi6_proto = sk->sk_protocol;
835 fl6.daddr = sk->sk_v6_daddr;
836 fl6.saddr = np->saddr;
837 fl6.flowlabel = np->flow_label;
838 fl6.flowi6_oif = sk->sk_bound_dev_if;
839 fl6.flowi6_mark = sk->sk_mark;
840 fl6.fl6_dport = inet->inet_dport;
841 fl6.fl6_sport = inet->inet_sport;
842 fl6.flowi6_uid = sk->sk_uid;
843 security_sk_classify_flow(sk, flowi6_to_flowi_common(&fl6));
846 final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt),
850 dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p);
852 sk->sk_route_caps = 0;
853 WRITE_ONCE(sk->sk_err_soft, -PTR_ERR(dst));
857 ip6_dst_store(sk, dst, NULL, NULL);
862 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
864 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb,
865 const struct inet6_skb_parm *opt)
867 const struct ipv6_pinfo *np = inet6_sk(sk);
870 if (((opt->flags & IP6SKB_HOPBYHOP) &&
871 (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) ||
872 (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) &&
873 np->rxopt.bits.rxflow) ||
874 (opt->srcrt && (np->rxopt.bits.srcrt ||
875 np->rxopt.bits.osrcrt)) ||
876 ((opt->dst1 || opt->dst0) &&
877 (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
882 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
884 static struct packet_type ipv6_packet_type __read_mostly = {
885 .type = cpu_to_be16(ETH_P_IPV6),
887 .list_func = ipv6_list_rcv,
890 static int __init ipv6_packet_init(void)
892 dev_add_pack(&ipv6_packet_type);
896 static void ipv6_packet_cleanup(void)
898 dev_remove_pack(&ipv6_packet_type);
901 static int __net_init ipv6_init_mibs(struct net *net)
905 net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib);
906 if (!net->mib.udp_stats_in6)
908 net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib);
909 if (!net->mib.udplite_stats_in6)
910 goto err_udplite_mib;
911 net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib);
912 if (!net->mib.ipv6_statistics)
915 for_each_possible_cpu(i) {
916 struct ipstats_mib *af_inet6_stats;
917 af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i);
918 u64_stats_init(&af_inet6_stats->syncp);
922 net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib);
923 if (!net->mib.icmpv6_statistics)
925 net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
927 if (!net->mib.icmpv6msg_statistics)
928 goto err_icmpmsg_mib;
932 free_percpu(net->mib.icmpv6_statistics);
934 free_percpu(net->mib.ipv6_statistics);
936 free_percpu(net->mib.udplite_stats_in6);
938 free_percpu(net->mib.udp_stats_in6);
942 static void ipv6_cleanup_mibs(struct net *net)
944 free_percpu(net->mib.udp_stats_in6);
945 free_percpu(net->mib.udplite_stats_in6);
946 free_percpu(net->mib.ipv6_statistics);
947 free_percpu(net->mib.icmpv6_statistics);
948 kfree(net->mib.icmpv6msg_statistics);
951 static int __net_init inet6_net_init(struct net *net)
955 net->ipv6.sysctl.bindv6only = 0;
956 net->ipv6.sysctl.icmpv6_time = 1*HZ;
957 net->ipv6.sysctl.icmpv6_echo_ignore_all = 0;
958 net->ipv6.sysctl.icmpv6_echo_ignore_multicast = 0;
959 net->ipv6.sysctl.icmpv6_echo_ignore_anycast = 0;
960 net->ipv6.sysctl.icmpv6_error_anycast_as_unicast = 0;
962 /* By default, rate limit error messages.
963 * Except for pmtu discovery, it would break it.
964 * proc_do_large_bitmap needs pointer to the bitmap.
966 bitmap_set(net->ipv6.sysctl.icmpv6_ratemask, 0, ICMPV6_ERRMSG_MAX + 1);
967 bitmap_clear(net->ipv6.sysctl.icmpv6_ratemask, ICMPV6_PKT_TOOBIG, 1);
968 net->ipv6.sysctl.icmpv6_ratemask_ptr = net->ipv6.sysctl.icmpv6_ratemask;
970 net->ipv6.sysctl.flowlabel_consistency = 1;
971 net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS;
972 net->ipv6.sysctl.idgen_retries = 3;
973 net->ipv6.sysctl.idgen_delay = 1 * HZ;
974 net->ipv6.sysctl.flowlabel_state_ranges = 0;
975 net->ipv6.sysctl.max_dst_opts_cnt = IP6_DEFAULT_MAX_DST_OPTS_CNT;
976 net->ipv6.sysctl.max_hbh_opts_cnt = IP6_DEFAULT_MAX_HBH_OPTS_CNT;
977 net->ipv6.sysctl.max_dst_opts_len = IP6_DEFAULT_MAX_DST_OPTS_LEN;
978 net->ipv6.sysctl.max_hbh_opts_len = IP6_DEFAULT_MAX_HBH_OPTS_LEN;
979 net->ipv6.sysctl.fib_notify_on_flag_change = 0;
980 atomic_set(&net->ipv6.fib6_sernum, 1);
982 net->ipv6.sysctl.ioam6_id = IOAM6_DEFAULT_ID;
983 net->ipv6.sysctl.ioam6_id_wide = IOAM6_DEFAULT_ID_WIDE;
985 err = ipv6_init_mibs(net);
988 #ifdef CONFIG_PROC_FS
989 err = udp6_proc_init(net);
992 err = tcp6_proc_init(net);
995 err = ac6_proc_init(net);
1001 #ifdef CONFIG_PROC_FS
1003 tcp6_proc_exit(net);
1005 udp6_proc_exit(net);
1007 ipv6_cleanup_mibs(net);
1012 static void __net_exit inet6_net_exit(struct net *net)
1014 #ifdef CONFIG_PROC_FS
1015 udp6_proc_exit(net);
1016 tcp6_proc_exit(net);
1019 ipv6_cleanup_mibs(net);
1022 static struct pernet_operations inet6_net_ops = {
1023 .init = inet6_net_init,
1024 .exit = inet6_net_exit,
1027 static int ipv6_route_input(struct sk_buff *skb)
1029 ip6_route_input(skb);
1030 return skb_dst(skb)->error;
1033 static const struct ipv6_stub ipv6_stub_impl = {
1034 .ipv6_sock_mc_join = ipv6_sock_mc_join,
1035 .ipv6_sock_mc_drop = ipv6_sock_mc_drop,
1036 .ipv6_dst_lookup_flow = ip6_dst_lookup_flow,
1037 .ipv6_route_input = ipv6_route_input,
1038 .fib6_get_table = fib6_get_table,
1039 .fib6_table_lookup = fib6_table_lookup,
1040 .fib6_lookup = fib6_lookup,
1041 .fib6_select_path = fib6_select_path,
1042 .ip6_mtu_from_fib6 = ip6_mtu_from_fib6,
1043 .fib6_nh_init = fib6_nh_init,
1044 .fib6_nh_release = fib6_nh_release,
1045 .fib6_nh_release_dsts = fib6_nh_release_dsts,
1046 .fib6_update_sernum = fib6_update_sernum_stub,
1047 .fib6_rt_update = fib6_rt_update,
1048 .ip6_del_rt = ip6_del_rt,
1049 .udpv6_encap_enable = udpv6_encap_enable,
1050 .ndisc_send_na = ndisc_send_na,
1051 #if IS_ENABLED(CONFIG_XFRM)
1052 .xfrm6_local_rxpmtu = xfrm6_local_rxpmtu,
1053 .xfrm6_udp_encap_rcv = xfrm6_udp_encap_rcv,
1054 .xfrm6_rcv_encap = xfrm6_rcv_encap,
1057 .ipv6_fragment = ip6_fragment,
1058 .ipv6_dev_find = ipv6_dev_find,
1061 static const struct ipv6_bpf_stub ipv6_bpf_stub_impl = {
1062 .inet6_bind = __inet6_bind,
1063 .udp6_lib_lookup = __udp6_lib_lookup,
1064 .ipv6_setsockopt = do_ipv6_setsockopt,
1065 .ipv6_getsockopt = do_ipv6_getsockopt,
1068 static int __init inet6_init(void)
1070 struct list_head *r;
1073 sock_skb_cb_check_size(sizeof(struct inet6_skb_parm));
1075 /* Register the socket-side information for inet6_create. */
1076 for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
1079 raw_hashinfo_init(&raw_v6_hashinfo);
1081 if (disable_ipv6_mod) {
1082 pr_info("Loaded, but administratively disabled, reboot required to enable\n");
1086 err = proto_register(&tcpv6_prot, 1);
1090 err = proto_register(&udpv6_prot, 1);
1092 goto out_unregister_tcp_proto;
1094 err = proto_register(&udplitev6_prot, 1);
1096 goto out_unregister_udp_proto;
1098 err = proto_register(&rawv6_prot, 1);
1100 goto out_unregister_udplite_proto;
1102 err = proto_register(&pingv6_prot, 1);
1104 goto out_unregister_raw_proto;
1106 /* We MUST register RAW sockets before we create the ICMP6,
1107 * IGMP6, or NDISC control sockets.
1111 goto out_unregister_ping_proto;
1113 /* Register the family here so that the init calls below will
1114 * be able to create sockets. (?? is this dangerous ??)
1116 err = sock_register(&inet6_family_ops);
1118 goto out_sock_register_fail;
1121 * ipngwg API draft makes clear that the correct semantics
1122 * for TCP and UDP is to consider one TCP and UDP instance
1123 * in a host available by both INET and INET6 APIs and
1124 * able to communicate via both network protocols.
1127 err = register_pernet_subsys(&inet6_net_ops);
1129 goto register_pernet_fail;
1130 err = ip6_mr_init();
1133 err = icmpv6_init();
1143 err = ipv6_netfilter_init();
1145 goto netfilter_fail;
1146 /* Create /proc/foo6 entries. */
1147 #ifdef CONFIG_PROC_FS
1149 if (raw6_proc_init())
1150 goto proc_raw6_fail;
1151 if (udplite6_proc_init())
1152 goto proc_udplite6_fail;
1153 if (ipv6_misc_proc_init())
1154 goto proc_misc6_fail;
1155 if (if6_proc_init())
1158 err = ip6_route_init();
1160 goto ip6_route_fail;
1161 err = ndisc_late_init();
1163 goto ndisc_late_fail;
1164 err = ip6_flowlabel_init();
1166 goto ip6_flowlabel_fail;
1167 err = ipv6_anycast_init();
1169 goto ipv6_anycast_fail;
1170 err = addrconf_init();
1174 /* Init v6 extension headers. */
1175 err = ipv6_exthdrs_init();
1177 goto ipv6_exthdrs_fail;
1179 err = ipv6_frag_init();
1181 goto ipv6_frag_fail;
1183 /* Init v6 transport protocols. */
1188 err = udplitev6_init();
1190 goto udplitev6_fail;
1192 err = udpv6_offload_init();
1194 goto udpv6_offload_fail;
1200 err = ipv6_packet_init();
1202 goto ipv6_packet_fail;
1204 err = pingv6_init();
1208 err = calipso_init();
1224 err = igmp6_late_init();
1226 goto igmp6_late_err;
1228 #ifdef CONFIG_SYSCTL
1229 err = ipv6_sysctl_register();
1234 /* ensure that ipv6 stubs are visible only after ipv6 is ready */
1236 ipv6_stub = &ipv6_stub_impl;
1237 ipv6_bpf_stub = &ipv6_bpf_stub_impl;
1241 #ifdef CONFIG_SYSCTL
1243 igmp6_late_cleanup();
1256 ipv6_packet_cleanup();
1260 udpv6_offload_exit();
1268 ipv6_exthdrs_exit();
1272 ipv6_anycast_cleanup();
1274 ip6_flowlabel_cleanup();
1276 ndisc_late_cleanup();
1278 ip6_route_cleanup();
1280 #ifdef CONFIG_PROC_FS
1283 ipv6_misc_proc_exit();
1285 udplite6_proc_exit();
1290 ipv6_netfilter_fini();
1300 unregister_pernet_subsys(&inet6_net_ops);
1301 register_pernet_fail:
1302 sock_unregister(PF_INET6);
1303 rtnl_unregister_all(PF_INET6);
1304 out_sock_register_fail:
1306 out_unregister_ping_proto:
1307 proto_unregister(&pingv6_prot);
1308 out_unregister_raw_proto:
1309 proto_unregister(&rawv6_prot);
1310 out_unregister_udplite_proto:
1311 proto_unregister(&udplitev6_prot);
1312 out_unregister_udp_proto:
1313 proto_unregister(&udpv6_prot);
1314 out_unregister_tcp_proto:
1315 proto_unregister(&tcpv6_prot);
1318 module_init(inet6_init);
1320 MODULE_ALIAS_NETPROTO(PF_INET6);