1 // SPDX-License-Identifier: GPL-2.0-or-later
3 * PF_INET6 socket protocol family
4 * Linux INET6 implementation
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 void inet6_sock_destruct(struct sock *sk)
114 inet6_cleanup_sock(sk);
115 inet_sock_destruct(sk);
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 (ipv6_only_sock(sk)) {
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_addr_valid_or_nonlocal(net, inet, v4addr,
349 err = -EADDRNOTAVAIL;
353 if (addr_type != IPV6_ADDR_ANY) {
354 struct net_device *dev = NULL;
357 if (__ipv6_addr_needs_scope_id(addr_type)) {
358 if (addr_len >= sizeof(struct sockaddr_in6) &&
359 addr->sin6_scope_id) {
360 /* Override any existing binding, if another one
361 * is supplied by user.
363 sk->sk_bound_dev_if = addr->sin6_scope_id;
366 /* Binding to link-local address requires an interface */
367 if (!sk->sk_bound_dev_if) {
373 if (sk->sk_bound_dev_if) {
374 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
381 /* ipv4 addr of the socket is invalid. Only the
382 * unspecified and mapped address have a v4 equivalent.
384 v4addr = LOOPBACK4_IPV6;
385 if (!(addr_type & IPV6_ADDR_MULTICAST)) {
386 if (!ipv6_can_nonlocal_bind(net, inet) &&
387 !ipv6_chk_addr(net, &addr->sin6_addr,
389 err = -EADDRNOTAVAIL;
397 inet->inet_rcv_saddr = v4addr;
398 inet->inet_saddr = v4addr;
400 sk->sk_v6_rcv_saddr = addr->sin6_addr;
402 if (!(addr_type & IPV6_ADDR_MULTICAST))
403 np->saddr = addr->sin6_addr;
405 saved_ipv6only = sk->sk_ipv6only;
406 if (addr_type != IPV6_ADDR_ANY && addr_type != IPV6_ADDR_MAPPED)
409 /* Make sure we are allowed to bind here. */
410 if (snum || !(inet->bind_address_no_port ||
411 (flags & BIND_FORCE_ADDRESS_NO_PORT))) {
412 if (sk->sk_prot->get_port(sk, snum)) {
413 sk->sk_ipv6only = saved_ipv6only;
414 inet_reset_saddr(sk);
418 if (!(flags & BIND_FROM_BPF)) {
419 err = BPF_CGROUP_RUN_PROG_INET6_POST_BIND(sk);
421 sk->sk_ipv6only = saved_ipv6only;
422 inet_reset_saddr(sk);
423 if (sk->sk_prot->put_port)
424 sk->sk_prot->put_port(sk);
430 if (addr_type != IPV6_ADDR_ANY)
431 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
433 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
434 inet->inet_sport = htons(inet->inet_num);
435 inet->inet_dport = 0;
436 inet->inet_daddr = 0;
438 if (flags & BIND_WITH_LOCK)
446 /* bind for INET6 API */
447 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
449 struct sock *sk = sock->sk;
450 u32 flags = BIND_WITH_LOCK;
451 const struct proto *prot;
454 /* IPV6_ADDRFORM can change sk->sk_prot under us. */
455 prot = READ_ONCE(sk->sk_prot);
456 /* If the socket has its own bind function then use it. */
458 return prot->bind(sk, uaddr, addr_len);
460 if (addr_len < SIN6_LEN_RFC2133)
463 /* BPF prog is run before any checks are done so that if the prog
464 * changes context in a wrong way it will be caught.
466 err = BPF_CGROUP_RUN_PROG_INET_BIND_LOCK(sk, uaddr,
467 CGROUP_INET6_BIND, &flags);
471 return __inet6_bind(sk, uaddr, addr_len, flags);
473 EXPORT_SYMBOL(inet6_bind);
475 int inet6_release(struct socket *sock)
477 struct sock *sk = sock->sk;
483 ipv6_sock_mc_close(sk);
486 ipv6_sock_ac_close(sk);
488 return inet_release(sock);
490 EXPORT_SYMBOL(inet6_release);
492 void inet6_destroy_sock(struct sock *sk)
494 struct ipv6_pinfo *np = inet6_sk(sk);
496 struct ipv6_txoptions *opt;
498 /* Release rx options */
500 skb = xchg(&np->pktoptions, NULL);
503 skb = xchg(&np->rxpmtu, NULL);
506 /* Free flowlabels */
507 fl6_free_socklist(sk);
509 /* Free tx options */
511 opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL);
513 atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
517 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
519 void inet6_cleanup_sock(struct sock *sk)
521 inet6_destroy_sock(sk);
523 EXPORT_SYMBOL_GPL(inet6_cleanup_sock);
526 * This does both peername and sockname.
528 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
531 struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
532 struct sock *sk = sock->sk;
533 struct inet_sock *inet = inet_sk(sk);
534 struct ipv6_pinfo *np = inet6_sk(sk);
536 sin->sin6_family = AF_INET6;
537 sin->sin6_flowinfo = 0;
538 sin->sin6_scope_id = 0;
541 if (!inet->inet_dport ||
542 (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
547 sin->sin6_port = inet->inet_dport;
548 sin->sin6_addr = sk->sk_v6_daddr;
550 sin->sin6_flowinfo = np->flow_label;
551 BPF_CGROUP_RUN_SA_PROG(sk, (struct sockaddr *)sin,
552 CGROUP_INET6_GETPEERNAME);
554 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
555 sin->sin6_addr = np->saddr;
557 sin->sin6_addr = sk->sk_v6_rcv_saddr;
558 sin->sin6_port = inet->inet_sport;
559 BPF_CGROUP_RUN_SA_PROG(sk, (struct sockaddr *)sin,
560 CGROUP_INET6_GETSOCKNAME);
562 sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
563 sk->sk_bound_dev_if);
567 EXPORT_SYMBOL(inet6_getname);
569 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
571 void __user *argp = (void __user *)arg;
572 struct sock *sk = sock->sk;
573 struct net *net = sock_net(sk);
574 const struct proto *prot;
579 struct in6_rtmsg rtmsg;
581 if (copy_from_user(&rtmsg, argp, sizeof(rtmsg)))
583 return ipv6_route_ioctl(net, cmd, &rtmsg);
586 return addrconf_add_ifaddr(net, argp);
588 return addrconf_del_ifaddr(net, argp);
590 return addrconf_set_dstaddr(net, argp);
592 /* IPV6_ADDRFORM can change sk->sk_prot under us. */
593 prot = READ_ONCE(sk->sk_prot);
596 return prot->ioctl(sk, cmd, arg);
601 EXPORT_SYMBOL(inet6_ioctl);
604 struct compat_in6_rtmsg {
605 struct in6_addr rtmsg_dst;
606 struct in6_addr rtmsg_src;
607 struct in6_addr rtmsg_gateway;
617 static int inet6_compat_routing_ioctl(struct sock *sk, unsigned int cmd,
618 struct compat_in6_rtmsg __user *ur)
622 if (copy_from_user(&rt.rtmsg_dst, &ur->rtmsg_dst,
623 3 * sizeof(struct in6_addr)) ||
624 get_user(rt.rtmsg_type, &ur->rtmsg_type) ||
625 get_user(rt.rtmsg_dst_len, &ur->rtmsg_dst_len) ||
626 get_user(rt.rtmsg_src_len, &ur->rtmsg_src_len) ||
627 get_user(rt.rtmsg_metric, &ur->rtmsg_metric) ||
628 get_user(rt.rtmsg_info, &ur->rtmsg_info) ||
629 get_user(rt.rtmsg_flags, &ur->rtmsg_flags) ||
630 get_user(rt.rtmsg_ifindex, &ur->rtmsg_ifindex))
634 return ipv6_route_ioctl(sock_net(sk), cmd, &rt);
637 int inet6_compat_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
639 void __user *argp = compat_ptr(arg);
640 struct sock *sk = sock->sk;
645 return inet6_compat_routing_ioctl(sk, cmd, argp);
650 EXPORT_SYMBOL_GPL(inet6_compat_ioctl);
651 #endif /* CONFIG_COMPAT */
653 INDIRECT_CALLABLE_DECLARE(int udpv6_sendmsg(struct sock *, struct msghdr *,
655 int inet6_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
657 struct sock *sk = sock->sk;
658 const struct proto *prot;
660 if (unlikely(inet_send_prepare(sk)))
663 /* IPV6_ADDRFORM can change sk->sk_prot under us. */
664 prot = READ_ONCE(sk->sk_prot);
665 return INDIRECT_CALL_2(prot->sendmsg, tcp_sendmsg, udpv6_sendmsg,
669 INDIRECT_CALLABLE_DECLARE(int udpv6_recvmsg(struct sock *, struct msghdr *,
670 size_t, int, int *));
671 int inet6_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
674 struct sock *sk = sock->sk;
675 const struct proto *prot;
679 if (likely(!(flags & MSG_ERRQUEUE)))
680 sock_rps_record_flow(sk);
682 /* IPV6_ADDRFORM can change sk->sk_prot under us. */
683 prot = READ_ONCE(sk->sk_prot);
684 err = INDIRECT_CALL_2(prot->recvmsg, tcp_recvmsg, udpv6_recvmsg,
685 sk, msg, size, flags, &addr_len);
687 msg->msg_namelen = addr_len;
691 const struct proto_ops inet6_stream_ops = {
693 .owner = THIS_MODULE,
694 .release = inet6_release,
696 .connect = inet_stream_connect, /* ok */
697 .socketpair = sock_no_socketpair, /* a do nothing */
698 .accept = inet_accept, /* ok */
699 .getname = inet6_getname,
700 .poll = tcp_poll, /* ok */
701 .ioctl = inet6_ioctl, /* must change */
702 .gettstamp = sock_gettstamp,
703 .listen = inet_listen, /* ok */
704 .shutdown = inet_shutdown, /* ok */
705 .setsockopt = sock_common_setsockopt, /* ok */
706 .getsockopt = sock_common_getsockopt, /* ok */
707 .sendmsg = inet6_sendmsg, /* retpoline's sake */
708 .recvmsg = inet6_recvmsg, /* retpoline's sake */
712 .sendpage = inet_sendpage,
713 .sendmsg_locked = tcp_sendmsg_locked,
714 .sendpage_locked = tcp_sendpage_locked,
715 .splice_read = tcp_splice_read,
716 .read_sock = tcp_read_sock,
717 .read_skb = tcp_read_skb,
718 .peek_len = tcp_peek_len,
720 .compat_ioctl = inet6_compat_ioctl,
722 .set_rcvlowat = tcp_set_rcvlowat,
725 const struct proto_ops inet6_dgram_ops = {
727 .owner = THIS_MODULE,
728 .release = inet6_release,
730 .connect = inet_dgram_connect, /* ok */
731 .socketpair = sock_no_socketpair, /* a do nothing */
732 .accept = sock_no_accept, /* a do nothing */
733 .getname = inet6_getname,
734 .poll = udp_poll, /* ok */
735 .ioctl = inet6_ioctl, /* must change */
736 .gettstamp = sock_gettstamp,
737 .listen = sock_no_listen, /* ok */
738 .shutdown = inet_shutdown, /* ok */
739 .setsockopt = sock_common_setsockopt, /* ok */
740 .getsockopt = sock_common_getsockopt, /* ok */
741 .sendmsg = inet6_sendmsg, /* retpoline's sake */
742 .recvmsg = inet6_recvmsg, /* retpoline's sake */
743 .read_skb = udp_read_skb,
744 .mmap = sock_no_mmap,
745 .sendpage = sock_no_sendpage,
746 .set_peek_off = sk_set_peek_off,
748 .compat_ioctl = inet6_compat_ioctl,
752 static const struct net_proto_family inet6_family_ops = {
754 .create = inet6_create,
755 .owner = THIS_MODULE,
758 int inet6_register_protosw(struct inet_protosw *p)
760 struct list_head *lh;
761 struct inet_protosw *answer;
762 struct list_head *last_perm;
763 int protocol = p->protocol;
766 spin_lock_bh(&inetsw6_lock);
769 if (p->type >= SOCK_MAX)
772 /* If we are trying to override a permanent protocol, bail. */
775 last_perm = &inetsw6[p->type];
776 list_for_each(lh, &inetsw6[p->type]) {
777 answer = list_entry(lh, struct inet_protosw, list);
779 /* Check only the non-wild match. */
780 if (INET_PROTOSW_PERMANENT & answer->flags) {
781 if (protocol == answer->protocol)
791 /* Add the new entry after the last permanent entry if any, so that
792 * the new entry does not override a permanent entry when matched with
793 * a wild-card protocol. But it is allowed to override any existing
794 * non-permanent entry. This means that when we remove this entry, the
795 * system automatically returns to the old behavior.
797 list_add_rcu(&p->list, last_perm);
800 spin_unlock_bh(&inetsw6_lock);
804 pr_err("Attempt to override permanent protocol %d\n", protocol);
808 pr_err("Ignoring attempt to register invalid socket type %d\n",
812 EXPORT_SYMBOL(inet6_register_protosw);
815 inet6_unregister_protosw(struct inet_protosw *p)
817 if (INET_PROTOSW_PERMANENT & p->flags) {
818 pr_err("Attempt to unregister permanent protocol %d\n",
821 spin_lock_bh(&inetsw6_lock);
822 list_del_rcu(&p->list);
823 spin_unlock_bh(&inetsw6_lock);
828 EXPORT_SYMBOL(inet6_unregister_protosw);
830 int inet6_sk_rebuild_header(struct sock *sk)
832 struct ipv6_pinfo *np = inet6_sk(sk);
833 struct dst_entry *dst;
835 dst = __sk_dst_check(sk, np->dst_cookie);
838 struct inet_sock *inet = inet_sk(sk);
839 struct in6_addr *final_p, final;
842 memset(&fl6, 0, sizeof(fl6));
843 fl6.flowi6_proto = sk->sk_protocol;
844 fl6.daddr = sk->sk_v6_daddr;
845 fl6.saddr = np->saddr;
846 fl6.flowlabel = np->flow_label;
847 fl6.flowi6_oif = sk->sk_bound_dev_if;
848 fl6.flowi6_mark = sk->sk_mark;
849 fl6.fl6_dport = inet->inet_dport;
850 fl6.fl6_sport = inet->inet_sport;
851 fl6.flowi6_uid = sk->sk_uid;
852 security_sk_classify_flow(sk, flowi6_to_flowi_common(&fl6));
855 final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt),
859 dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p);
861 sk->sk_route_caps = 0;
862 sk->sk_err_soft = -PTR_ERR(dst);
866 ip6_dst_store(sk, dst, NULL, NULL);
871 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
873 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb,
874 const struct inet6_skb_parm *opt)
876 const struct ipv6_pinfo *np = inet6_sk(sk);
879 if (((opt->flags & IP6SKB_HOPBYHOP) &&
880 (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) ||
881 (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) &&
882 np->rxopt.bits.rxflow) ||
883 (opt->srcrt && (np->rxopt.bits.srcrt ||
884 np->rxopt.bits.osrcrt)) ||
885 ((opt->dst1 || opt->dst0) &&
886 (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
891 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
893 static struct packet_type ipv6_packet_type __read_mostly = {
894 .type = cpu_to_be16(ETH_P_IPV6),
896 .list_func = ipv6_list_rcv,
899 static int __init ipv6_packet_init(void)
901 dev_add_pack(&ipv6_packet_type);
905 static void ipv6_packet_cleanup(void)
907 dev_remove_pack(&ipv6_packet_type);
910 static int __net_init ipv6_init_mibs(struct net *net)
914 net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib);
915 if (!net->mib.udp_stats_in6)
917 net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib);
918 if (!net->mib.udplite_stats_in6)
919 goto err_udplite_mib;
920 net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib);
921 if (!net->mib.ipv6_statistics)
924 for_each_possible_cpu(i) {
925 struct ipstats_mib *af_inet6_stats;
926 af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i);
927 u64_stats_init(&af_inet6_stats->syncp);
931 net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib);
932 if (!net->mib.icmpv6_statistics)
934 net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
936 if (!net->mib.icmpv6msg_statistics)
937 goto err_icmpmsg_mib;
941 free_percpu(net->mib.icmpv6_statistics);
943 free_percpu(net->mib.ipv6_statistics);
945 free_percpu(net->mib.udplite_stats_in6);
947 free_percpu(net->mib.udp_stats_in6);
951 static void ipv6_cleanup_mibs(struct net *net)
953 free_percpu(net->mib.udp_stats_in6);
954 free_percpu(net->mib.udplite_stats_in6);
955 free_percpu(net->mib.ipv6_statistics);
956 free_percpu(net->mib.icmpv6_statistics);
957 kfree(net->mib.icmpv6msg_statistics);
960 static int __net_init inet6_net_init(struct net *net)
964 net->ipv6.sysctl.bindv6only = 0;
965 net->ipv6.sysctl.icmpv6_time = 1*HZ;
966 net->ipv6.sysctl.icmpv6_echo_ignore_all = 0;
967 net->ipv6.sysctl.icmpv6_echo_ignore_multicast = 0;
968 net->ipv6.sysctl.icmpv6_echo_ignore_anycast = 0;
970 /* By default, rate limit error messages.
971 * Except for pmtu discovery, it would break it.
972 * proc_do_large_bitmap needs pointer to the bitmap.
974 bitmap_set(net->ipv6.sysctl.icmpv6_ratemask, 0, ICMPV6_ERRMSG_MAX + 1);
975 bitmap_clear(net->ipv6.sysctl.icmpv6_ratemask, ICMPV6_PKT_TOOBIG, 1);
976 net->ipv6.sysctl.icmpv6_ratemask_ptr = net->ipv6.sysctl.icmpv6_ratemask;
978 net->ipv6.sysctl.flowlabel_consistency = 1;
979 net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS;
980 net->ipv6.sysctl.idgen_retries = 3;
981 net->ipv6.sysctl.idgen_delay = 1 * HZ;
982 net->ipv6.sysctl.flowlabel_state_ranges = 0;
983 net->ipv6.sysctl.max_dst_opts_cnt = IP6_DEFAULT_MAX_DST_OPTS_CNT;
984 net->ipv6.sysctl.max_hbh_opts_cnt = IP6_DEFAULT_MAX_HBH_OPTS_CNT;
985 net->ipv6.sysctl.max_dst_opts_len = IP6_DEFAULT_MAX_DST_OPTS_LEN;
986 net->ipv6.sysctl.max_hbh_opts_len = IP6_DEFAULT_MAX_HBH_OPTS_LEN;
987 net->ipv6.sysctl.fib_notify_on_flag_change = 0;
988 atomic_set(&net->ipv6.fib6_sernum, 1);
990 net->ipv6.sysctl.ioam6_id = IOAM6_DEFAULT_ID;
991 net->ipv6.sysctl.ioam6_id_wide = IOAM6_DEFAULT_ID_WIDE;
993 err = ipv6_init_mibs(net);
996 #ifdef CONFIG_PROC_FS
997 err = udp6_proc_init(net);
1000 err = tcp6_proc_init(net);
1002 goto proc_tcp6_fail;
1003 err = ac6_proc_init(net);
1009 #ifdef CONFIG_PROC_FS
1011 tcp6_proc_exit(net);
1013 udp6_proc_exit(net);
1015 ipv6_cleanup_mibs(net);
1020 static void __net_exit inet6_net_exit(struct net *net)
1022 #ifdef CONFIG_PROC_FS
1023 udp6_proc_exit(net);
1024 tcp6_proc_exit(net);
1027 ipv6_cleanup_mibs(net);
1030 static struct pernet_operations inet6_net_ops = {
1031 .init = inet6_net_init,
1032 .exit = inet6_net_exit,
1035 static int ipv6_route_input(struct sk_buff *skb)
1037 ip6_route_input(skb);
1038 return skb_dst(skb)->error;
1041 static const struct ipv6_stub ipv6_stub_impl = {
1042 .ipv6_sock_mc_join = ipv6_sock_mc_join,
1043 .ipv6_sock_mc_drop = ipv6_sock_mc_drop,
1044 .ipv6_dst_lookup_flow = ip6_dst_lookup_flow,
1045 .ipv6_route_input = ipv6_route_input,
1046 .fib6_get_table = fib6_get_table,
1047 .fib6_table_lookup = fib6_table_lookup,
1048 .fib6_lookup = fib6_lookup,
1049 .fib6_select_path = fib6_select_path,
1050 .ip6_mtu_from_fib6 = ip6_mtu_from_fib6,
1051 .fib6_nh_init = fib6_nh_init,
1052 .fib6_nh_release = fib6_nh_release,
1053 .fib6_nh_release_dsts = fib6_nh_release_dsts,
1054 .fib6_update_sernum = fib6_update_sernum_stub,
1055 .fib6_rt_update = fib6_rt_update,
1056 .ip6_del_rt = ip6_del_rt,
1057 .udpv6_encap_enable = udpv6_encap_enable,
1058 .ndisc_send_na = ndisc_send_na,
1059 #if IS_ENABLED(CONFIG_XFRM)
1060 .xfrm6_local_rxpmtu = xfrm6_local_rxpmtu,
1061 .xfrm6_udp_encap_rcv = xfrm6_udp_encap_rcv,
1062 .xfrm6_rcv_encap = xfrm6_rcv_encap,
1065 .ipv6_fragment = ip6_fragment,
1066 .ipv6_dev_find = ipv6_dev_find,
1069 static const struct ipv6_bpf_stub ipv6_bpf_stub_impl = {
1070 .inet6_bind = __inet6_bind,
1071 .udp6_lib_lookup = __udp6_lib_lookup,
1072 .ipv6_setsockopt = do_ipv6_setsockopt,
1073 .ipv6_getsockopt = do_ipv6_getsockopt,
1076 static int __init inet6_init(void)
1078 struct list_head *r;
1081 sock_skb_cb_check_size(sizeof(struct inet6_skb_parm));
1083 /* Register the socket-side information for inet6_create. */
1084 for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
1087 raw_hashinfo_init(&raw_v6_hashinfo);
1089 if (disable_ipv6_mod) {
1090 pr_info("Loaded, but administratively disabled, reboot required to enable\n");
1094 err = proto_register(&tcpv6_prot, 1);
1098 err = proto_register(&udpv6_prot, 1);
1100 goto out_unregister_tcp_proto;
1102 err = proto_register(&udplitev6_prot, 1);
1104 goto out_unregister_udp_proto;
1106 err = proto_register(&rawv6_prot, 1);
1108 goto out_unregister_udplite_proto;
1110 err = proto_register(&pingv6_prot, 1);
1112 goto out_unregister_raw_proto;
1114 /* We MUST register RAW sockets before we create the ICMP6,
1115 * IGMP6, or NDISC control sockets.
1119 goto out_unregister_ping_proto;
1121 /* Register the family here so that the init calls below will
1122 * be able to create sockets. (?? is this dangerous ??)
1124 err = sock_register(&inet6_family_ops);
1126 goto out_sock_register_fail;
1129 * ipngwg API draft makes clear that the correct semantics
1130 * for TCP and UDP is to consider one TCP and UDP instance
1131 * in a host available by both INET and INET6 APIs and
1132 * able to communicate via both network protocols.
1135 err = register_pernet_subsys(&inet6_net_ops);
1137 goto register_pernet_fail;
1138 err = ip6_mr_init();
1141 err = icmpv6_init();
1151 err = ipv6_netfilter_init();
1153 goto netfilter_fail;
1154 /* Create /proc/foo6 entries. */
1155 #ifdef CONFIG_PROC_FS
1157 if (raw6_proc_init())
1158 goto proc_raw6_fail;
1159 if (udplite6_proc_init())
1160 goto proc_udplite6_fail;
1161 if (ipv6_misc_proc_init())
1162 goto proc_misc6_fail;
1163 if (if6_proc_init())
1166 err = ip6_route_init();
1168 goto ip6_route_fail;
1169 err = ndisc_late_init();
1171 goto ndisc_late_fail;
1172 err = ip6_flowlabel_init();
1174 goto ip6_flowlabel_fail;
1175 err = ipv6_anycast_init();
1177 goto ipv6_anycast_fail;
1178 err = addrconf_init();
1182 /* Init v6 extension headers. */
1183 err = ipv6_exthdrs_init();
1185 goto ipv6_exthdrs_fail;
1187 err = ipv6_frag_init();
1189 goto ipv6_frag_fail;
1191 /* Init v6 transport protocols. */
1196 err = udplitev6_init();
1198 goto udplitev6_fail;
1200 err = udpv6_offload_init();
1202 goto udpv6_offload_fail;
1208 err = ipv6_packet_init();
1210 goto ipv6_packet_fail;
1212 err = pingv6_init();
1216 err = calipso_init();
1232 err = igmp6_late_init();
1234 goto igmp6_late_err;
1236 #ifdef CONFIG_SYSCTL
1237 err = ipv6_sysctl_register();
1242 /* ensure that ipv6 stubs are visible only after ipv6 is ready */
1244 ipv6_stub = &ipv6_stub_impl;
1245 ipv6_bpf_stub = &ipv6_bpf_stub_impl;
1249 #ifdef CONFIG_SYSCTL
1251 igmp6_late_cleanup();
1264 ipv6_packet_cleanup();
1268 udpv6_offload_exit();
1276 ipv6_exthdrs_exit();
1280 ipv6_anycast_cleanup();
1282 ip6_flowlabel_cleanup();
1284 ndisc_late_cleanup();
1286 ip6_route_cleanup();
1288 #ifdef CONFIG_PROC_FS
1291 ipv6_misc_proc_exit();
1293 udplite6_proc_exit();
1298 ipv6_netfilter_fini();
1308 unregister_pernet_subsys(&inet6_net_ops);
1309 register_pernet_fail:
1310 sock_unregister(PF_INET6);
1311 rtnl_unregister_all(PF_INET6);
1312 out_sock_register_fail:
1314 out_unregister_ping_proto:
1315 proto_unregister(&pingv6_prot);
1316 out_unregister_raw_proto:
1317 proto_unregister(&rawv6_prot);
1318 out_unregister_udplite_proto:
1319 proto_unregister(&udplitev6_prot);
1320 out_unregister_udp_proto:
1321 proto_unregister(&udpv6_prot);
1322 out_unregister_tcp_proto:
1323 proto_unregister(&tcpv6_prot);
1326 module_init(inet6_init);
1328 MODULE_ALIAS_NETPROTO(PF_INET6);