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 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->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
208 if (SOCK_RAW == sock->type) {
209 inet->inet_num = protocol;
210 if (IPPROTO_RAW == protocol)
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.
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 * Increment only the relevant sk_prot->socks debug field, this changes
247 * the previous behaviour of incrementing both the equivalent to
248 * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
250 * This allows better debug granularity as we'll know exactly how many
251 * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
252 * transport protocol socks. -acme
254 sk_refcnt_debug_inc(sk);
256 if (inet->inet_num) {
257 /* It assumes that any protocol which allows
258 * the user to assign a number at socket
259 * creation time automatically shares.
261 inet->inet_sport = htons(inet->inet_num);
262 err = sk->sk_prot->hash(sk);
264 sk_common_release(sk);
268 if (sk->sk_prot->init) {
269 err = sk->sk_prot->init(sk);
271 sk_common_release(sk);
277 err = BPF_CGROUP_RUN_PROG_INET_SOCK(sk);
279 sk_common_release(sk);
290 static int __inet6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len,
293 struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr;
294 struct inet_sock *inet = inet_sk(sk);
295 struct ipv6_pinfo *np = inet6_sk(sk);
296 struct net *net = sock_net(sk);
303 if (addr->sin6_family != AF_INET6)
304 return -EAFNOSUPPORT;
306 addr_type = ipv6_addr_type(&addr->sin6_addr);
307 if ((addr_type & IPV6_ADDR_MULTICAST) && sk->sk_type == SOCK_STREAM)
310 snum = ntohs(addr->sin6_port);
311 if (!(flags & BIND_NO_CAP_NET_BIND_SERVICE) &&
312 snum && inet_port_requires_bind_service(net, snum) &&
313 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
316 if (flags & BIND_WITH_LOCK)
319 /* Check these errors (active socket, double bind). */
320 if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
325 /* Check if the address belongs to the host. */
326 if (addr_type == IPV6_ADDR_MAPPED) {
327 struct net_device *dev = NULL;
330 /* Binding to v4-mapped address on a v6-only socket
333 if (ipv6_only_sock(sk)) {
339 if (sk->sk_bound_dev_if) {
340 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
347 /* Reproduce AF_INET checks to make the bindings consistent */
348 v4addr = addr->sin6_addr.s6_addr32[3];
349 chk_addr_ret = inet_addr_type_dev_table(net, dev, v4addr);
352 if (!inet_addr_valid_or_nonlocal(net, inet, v4addr,
354 err = -EADDRNOTAVAIL;
358 if (addr_type != IPV6_ADDR_ANY) {
359 struct net_device *dev = NULL;
362 if (__ipv6_addr_needs_scope_id(addr_type)) {
363 if (addr_len >= sizeof(struct sockaddr_in6) &&
364 addr->sin6_scope_id) {
365 /* Override any existing binding, if another one
366 * is supplied by user.
368 sk->sk_bound_dev_if = addr->sin6_scope_id;
371 /* Binding to link-local address requires an interface */
372 if (!sk->sk_bound_dev_if) {
378 if (sk->sk_bound_dev_if) {
379 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
386 /* ipv4 addr of the socket is invalid. Only the
387 * unspecified and mapped address have a v4 equivalent.
389 v4addr = LOOPBACK4_IPV6;
390 if (!(addr_type & IPV6_ADDR_MULTICAST)) {
391 if (!ipv6_can_nonlocal_bind(net, inet) &&
392 !ipv6_chk_addr(net, &addr->sin6_addr,
394 err = -EADDRNOTAVAIL;
402 inet->inet_rcv_saddr = v4addr;
403 inet->inet_saddr = v4addr;
405 sk->sk_v6_rcv_saddr = addr->sin6_addr;
407 if (!(addr_type & IPV6_ADDR_MULTICAST))
408 np->saddr = addr->sin6_addr;
410 saved_ipv6only = sk->sk_ipv6only;
411 if (addr_type != IPV6_ADDR_ANY && addr_type != IPV6_ADDR_MAPPED)
414 /* Make sure we are allowed to bind here. */
415 if (snum || !(inet->bind_address_no_port ||
416 (flags & BIND_FORCE_ADDRESS_NO_PORT))) {
417 err = sk->sk_prot->get_port(sk, snum);
419 sk->sk_ipv6only = saved_ipv6only;
420 inet_reset_saddr(sk);
423 if (!(flags & BIND_FROM_BPF)) {
424 err = BPF_CGROUP_RUN_PROG_INET6_POST_BIND(sk);
426 sk->sk_ipv6only = saved_ipv6only;
427 inet_reset_saddr(sk);
428 if (sk->sk_prot->put_port)
429 sk->sk_prot->put_port(sk);
435 if (addr_type != IPV6_ADDR_ANY)
436 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
438 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
439 inet->inet_sport = htons(inet->inet_num);
440 inet->inet_dport = 0;
441 inet->inet_daddr = 0;
443 if (flags & BIND_WITH_LOCK)
451 /* bind for INET6 API */
452 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
454 struct sock *sk = sock->sk;
455 u32 flags = BIND_WITH_LOCK;
456 const struct proto *prot;
459 /* IPV6_ADDRFORM can change sk->sk_prot under us. */
460 prot = READ_ONCE(sk->sk_prot);
461 /* If the socket has its own bind function then use it. */
463 return prot->bind(sk, uaddr, addr_len);
465 if (addr_len < SIN6_LEN_RFC2133)
468 /* BPF prog is run before any checks are done so that if the prog
469 * changes context in a wrong way it will be caught.
471 err = BPF_CGROUP_RUN_PROG_INET_BIND_LOCK(sk, uaddr,
472 CGROUP_INET6_BIND, &flags);
476 return __inet6_bind(sk, uaddr, addr_len, flags);
478 EXPORT_SYMBOL(inet6_bind);
480 int inet6_release(struct socket *sock)
482 struct sock *sk = sock->sk;
488 ipv6_sock_mc_close(sk);
491 ipv6_sock_ac_close(sk);
493 return inet_release(sock);
495 EXPORT_SYMBOL(inet6_release);
497 void inet6_destroy_sock(struct sock *sk)
499 struct ipv6_pinfo *np = inet6_sk(sk);
501 struct ipv6_txoptions *opt;
503 /* Release rx options */
505 skb = xchg(&np->pktoptions, NULL);
508 skb = xchg(&np->rxpmtu, NULL);
511 /* Free flowlabels */
512 fl6_free_socklist(sk);
514 /* Free tx options */
516 opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL);
518 atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
522 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
524 void inet6_cleanup_sock(struct sock *sk)
526 inet6_destroy_sock(sk);
528 EXPORT_SYMBOL_GPL(inet6_cleanup_sock);
531 * This does both peername and sockname.
533 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
536 struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
537 struct sock *sk = sock->sk;
538 struct inet_sock *inet = inet_sk(sk);
539 struct ipv6_pinfo *np = inet6_sk(sk);
541 sin->sin6_family = AF_INET6;
542 sin->sin6_flowinfo = 0;
543 sin->sin6_scope_id = 0;
546 if (!inet->inet_dport ||
547 (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
552 sin->sin6_port = inet->inet_dport;
553 sin->sin6_addr = sk->sk_v6_daddr;
555 sin->sin6_flowinfo = np->flow_label;
556 BPF_CGROUP_RUN_SA_PROG(sk, (struct sockaddr *)sin,
557 CGROUP_INET6_GETPEERNAME);
559 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
560 sin->sin6_addr = np->saddr;
562 sin->sin6_addr = sk->sk_v6_rcv_saddr;
563 sin->sin6_port = inet->inet_sport;
564 BPF_CGROUP_RUN_SA_PROG(sk, (struct sockaddr *)sin,
565 CGROUP_INET6_GETSOCKNAME);
567 sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
568 sk->sk_bound_dev_if);
572 EXPORT_SYMBOL(inet6_getname);
574 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
576 void __user *argp = (void __user *)arg;
577 struct sock *sk = sock->sk;
578 struct net *net = sock_net(sk);
579 const struct proto *prot;
584 struct in6_rtmsg rtmsg;
586 if (copy_from_user(&rtmsg, argp, sizeof(rtmsg)))
588 return ipv6_route_ioctl(net, cmd, &rtmsg);
591 return addrconf_add_ifaddr(net, argp);
593 return addrconf_del_ifaddr(net, argp);
595 return addrconf_set_dstaddr(net, argp);
597 /* IPV6_ADDRFORM can change sk->sk_prot under us. */
598 prot = READ_ONCE(sk->sk_prot);
601 return prot->ioctl(sk, cmd, arg);
606 EXPORT_SYMBOL(inet6_ioctl);
609 struct compat_in6_rtmsg {
610 struct in6_addr rtmsg_dst;
611 struct in6_addr rtmsg_src;
612 struct in6_addr rtmsg_gateway;
622 static int inet6_compat_routing_ioctl(struct sock *sk, unsigned int cmd,
623 struct compat_in6_rtmsg __user *ur)
627 if (copy_from_user(&rt.rtmsg_dst, &ur->rtmsg_dst,
628 3 * sizeof(struct in6_addr)) ||
629 get_user(rt.rtmsg_type, &ur->rtmsg_type) ||
630 get_user(rt.rtmsg_dst_len, &ur->rtmsg_dst_len) ||
631 get_user(rt.rtmsg_src_len, &ur->rtmsg_src_len) ||
632 get_user(rt.rtmsg_metric, &ur->rtmsg_metric) ||
633 get_user(rt.rtmsg_info, &ur->rtmsg_info) ||
634 get_user(rt.rtmsg_flags, &ur->rtmsg_flags) ||
635 get_user(rt.rtmsg_ifindex, &ur->rtmsg_ifindex))
639 return ipv6_route_ioctl(sock_net(sk), cmd, &rt);
642 int inet6_compat_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
644 void __user *argp = compat_ptr(arg);
645 struct sock *sk = sock->sk;
650 return inet6_compat_routing_ioctl(sk, cmd, argp);
655 EXPORT_SYMBOL_GPL(inet6_compat_ioctl);
656 #endif /* CONFIG_COMPAT */
658 INDIRECT_CALLABLE_DECLARE(int udpv6_sendmsg(struct sock *, struct msghdr *,
660 int inet6_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
662 struct sock *sk = sock->sk;
663 const struct proto *prot;
665 if (unlikely(inet_send_prepare(sk)))
668 /* IPV6_ADDRFORM can change sk->sk_prot under us. */
669 prot = READ_ONCE(sk->sk_prot);
670 return INDIRECT_CALL_2(prot->sendmsg, tcp_sendmsg, udpv6_sendmsg,
674 INDIRECT_CALLABLE_DECLARE(int udpv6_recvmsg(struct sock *, struct msghdr *,
675 size_t, int, int *));
676 int inet6_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
679 struct sock *sk = sock->sk;
680 const struct proto *prot;
684 if (likely(!(flags & MSG_ERRQUEUE)))
685 sock_rps_record_flow(sk);
687 /* IPV6_ADDRFORM can change sk->sk_prot under us. */
688 prot = READ_ONCE(sk->sk_prot);
689 err = INDIRECT_CALL_2(prot->recvmsg, tcp_recvmsg, udpv6_recvmsg,
690 sk, msg, size, flags, &addr_len);
692 msg->msg_namelen = addr_len;
696 const struct proto_ops inet6_stream_ops = {
698 .owner = THIS_MODULE,
699 .release = inet6_release,
701 .connect = inet_stream_connect, /* ok */
702 .socketpair = sock_no_socketpair, /* a do nothing */
703 .accept = inet_accept, /* ok */
704 .getname = inet6_getname,
705 .poll = tcp_poll, /* ok */
706 .ioctl = inet6_ioctl, /* must change */
707 .gettstamp = sock_gettstamp,
708 .listen = inet_listen, /* ok */
709 .shutdown = inet_shutdown, /* ok */
710 .setsockopt = sock_common_setsockopt, /* ok */
711 .getsockopt = sock_common_getsockopt, /* ok */
712 .sendmsg = inet6_sendmsg, /* retpoline's sake */
713 .recvmsg = inet6_recvmsg, /* retpoline's sake */
717 .splice_eof = inet_splice_eof,
718 .sendpage = inet_sendpage,
719 .sendmsg_locked = tcp_sendmsg_locked,
720 .sendpage_locked = tcp_sendpage_locked,
721 .splice_read = tcp_splice_read,
722 .read_sock = tcp_read_sock,
723 .read_skb = tcp_read_skb,
724 .peek_len = tcp_peek_len,
726 .compat_ioctl = inet6_compat_ioctl,
728 .set_rcvlowat = tcp_set_rcvlowat,
731 const struct proto_ops inet6_dgram_ops = {
733 .owner = THIS_MODULE,
734 .release = inet6_release,
736 .connect = inet_dgram_connect, /* ok */
737 .socketpair = sock_no_socketpair, /* a do nothing */
738 .accept = sock_no_accept, /* a do nothing */
739 .getname = inet6_getname,
740 .poll = udp_poll, /* ok */
741 .ioctl = inet6_ioctl, /* must change */
742 .gettstamp = sock_gettstamp,
743 .listen = sock_no_listen, /* ok */
744 .shutdown = inet_shutdown, /* ok */
745 .setsockopt = sock_common_setsockopt, /* ok */
746 .getsockopt = sock_common_getsockopt, /* ok */
747 .sendmsg = inet6_sendmsg, /* retpoline's sake */
748 .recvmsg = inet6_recvmsg, /* retpoline's sake */
749 .read_skb = udp_read_skb,
750 .mmap = sock_no_mmap,
751 .sendpage = sock_no_sendpage,
752 .set_peek_off = sk_set_peek_off,
754 .compat_ioctl = inet6_compat_ioctl,
758 static const struct net_proto_family inet6_family_ops = {
760 .create = inet6_create,
761 .owner = THIS_MODULE,
764 int inet6_register_protosw(struct inet_protosw *p)
766 struct list_head *lh;
767 struct inet_protosw *answer;
768 struct list_head *last_perm;
769 int protocol = p->protocol;
772 spin_lock_bh(&inetsw6_lock);
775 if (p->type >= SOCK_MAX)
778 /* If we are trying to override a permanent protocol, bail. */
781 last_perm = &inetsw6[p->type];
782 list_for_each(lh, &inetsw6[p->type]) {
783 answer = list_entry(lh, struct inet_protosw, list);
785 /* Check only the non-wild match. */
786 if (INET_PROTOSW_PERMANENT & answer->flags) {
787 if (protocol == answer->protocol)
797 /* Add the new entry after the last permanent entry if any, so that
798 * the new entry does not override a permanent entry when matched with
799 * a wild-card protocol. But it is allowed to override any existing
800 * non-permanent entry. This means that when we remove this entry, the
801 * system automatically returns to the old behavior.
803 list_add_rcu(&p->list, last_perm);
806 spin_unlock_bh(&inetsw6_lock);
810 pr_err("Attempt to override permanent protocol %d\n", protocol);
814 pr_err("Ignoring attempt to register invalid socket type %d\n",
818 EXPORT_SYMBOL(inet6_register_protosw);
821 inet6_unregister_protosw(struct inet_protosw *p)
823 if (INET_PROTOSW_PERMANENT & p->flags) {
824 pr_err("Attempt to unregister permanent protocol %d\n",
827 spin_lock_bh(&inetsw6_lock);
828 list_del_rcu(&p->list);
829 spin_unlock_bh(&inetsw6_lock);
834 EXPORT_SYMBOL(inet6_unregister_protosw);
836 int inet6_sk_rebuild_header(struct sock *sk)
838 struct ipv6_pinfo *np = inet6_sk(sk);
839 struct dst_entry *dst;
841 dst = __sk_dst_check(sk, np->dst_cookie);
844 struct inet_sock *inet = inet_sk(sk);
845 struct in6_addr *final_p, final;
848 memset(&fl6, 0, sizeof(fl6));
849 fl6.flowi6_proto = sk->sk_protocol;
850 fl6.daddr = sk->sk_v6_daddr;
851 fl6.saddr = np->saddr;
852 fl6.flowlabel = np->flow_label;
853 fl6.flowi6_oif = sk->sk_bound_dev_if;
854 fl6.flowi6_mark = sk->sk_mark;
855 fl6.fl6_dport = inet->inet_dport;
856 fl6.fl6_sport = inet->inet_sport;
857 fl6.flowi6_uid = sk->sk_uid;
858 security_sk_classify_flow(sk, flowi6_to_flowi_common(&fl6));
861 final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt),
865 dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p);
867 sk->sk_route_caps = 0;
868 sk->sk_err_soft = -PTR_ERR(dst);
872 ip6_dst_store(sk, dst, NULL, NULL);
877 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
879 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb,
880 const struct inet6_skb_parm *opt)
882 const struct ipv6_pinfo *np = inet6_sk(sk);
885 if (((opt->flags & IP6SKB_HOPBYHOP) &&
886 (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) ||
887 (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) &&
888 np->rxopt.bits.rxflow) ||
889 (opt->srcrt && (np->rxopt.bits.srcrt ||
890 np->rxopt.bits.osrcrt)) ||
891 ((opt->dst1 || opt->dst0) &&
892 (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
897 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
899 static struct packet_type ipv6_packet_type __read_mostly = {
900 .type = cpu_to_be16(ETH_P_IPV6),
902 .list_func = ipv6_list_rcv,
905 static int __init ipv6_packet_init(void)
907 dev_add_pack(&ipv6_packet_type);
911 static void ipv6_packet_cleanup(void)
913 dev_remove_pack(&ipv6_packet_type);
916 static int __net_init ipv6_init_mibs(struct net *net)
920 net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib);
921 if (!net->mib.udp_stats_in6)
923 net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib);
924 if (!net->mib.udplite_stats_in6)
925 goto err_udplite_mib;
926 net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib);
927 if (!net->mib.ipv6_statistics)
930 for_each_possible_cpu(i) {
931 struct ipstats_mib *af_inet6_stats;
932 af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i);
933 u64_stats_init(&af_inet6_stats->syncp);
937 net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib);
938 if (!net->mib.icmpv6_statistics)
940 net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
942 if (!net->mib.icmpv6msg_statistics)
943 goto err_icmpmsg_mib;
947 free_percpu(net->mib.icmpv6_statistics);
949 free_percpu(net->mib.ipv6_statistics);
951 free_percpu(net->mib.udplite_stats_in6);
953 free_percpu(net->mib.udp_stats_in6);
957 static void ipv6_cleanup_mibs(struct net *net)
959 free_percpu(net->mib.udp_stats_in6);
960 free_percpu(net->mib.udplite_stats_in6);
961 free_percpu(net->mib.ipv6_statistics);
962 free_percpu(net->mib.icmpv6_statistics);
963 kfree(net->mib.icmpv6msg_statistics);
966 static int __net_init inet6_net_init(struct net *net)
970 net->ipv6.sysctl.bindv6only = 0;
971 net->ipv6.sysctl.icmpv6_time = 1*HZ;
972 net->ipv6.sysctl.icmpv6_echo_ignore_all = 0;
973 net->ipv6.sysctl.icmpv6_echo_ignore_multicast = 0;
974 net->ipv6.sysctl.icmpv6_echo_ignore_anycast = 0;
976 /* By default, rate limit error messages.
977 * Except for pmtu discovery, it would break it.
978 * proc_do_large_bitmap needs pointer to the bitmap.
980 bitmap_set(net->ipv6.sysctl.icmpv6_ratemask, 0, ICMPV6_ERRMSG_MAX + 1);
981 bitmap_clear(net->ipv6.sysctl.icmpv6_ratemask, ICMPV6_PKT_TOOBIG, 1);
982 net->ipv6.sysctl.icmpv6_ratemask_ptr = net->ipv6.sysctl.icmpv6_ratemask;
984 net->ipv6.sysctl.flowlabel_consistency = 1;
985 net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS;
986 net->ipv6.sysctl.idgen_retries = 3;
987 net->ipv6.sysctl.idgen_delay = 1 * HZ;
988 net->ipv6.sysctl.flowlabel_state_ranges = 0;
989 net->ipv6.sysctl.max_dst_opts_cnt = IP6_DEFAULT_MAX_DST_OPTS_CNT;
990 net->ipv6.sysctl.max_hbh_opts_cnt = IP6_DEFAULT_MAX_HBH_OPTS_CNT;
991 net->ipv6.sysctl.max_dst_opts_len = IP6_DEFAULT_MAX_DST_OPTS_LEN;
992 net->ipv6.sysctl.max_hbh_opts_len = IP6_DEFAULT_MAX_HBH_OPTS_LEN;
993 net->ipv6.sysctl.fib_notify_on_flag_change = 0;
994 atomic_set(&net->ipv6.fib6_sernum, 1);
996 net->ipv6.sysctl.ioam6_id = IOAM6_DEFAULT_ID;
997 net->ipv6.sysctl.ioam6_id_wide = IOAM6_DEFAULT_ID_WIDE;
999 err = ipv6_init_mibs(net);
1002 #ifdef CONFIG_PROC_FS
1003 err = udp6_proc_init(net);
1006 err = tcp6_proc_init(net);
1008 goto proc_tcp6_fail;
1009 err = ac6_proc_init(net);
1015 #ifdef CONFIG_PROC_FS
1017 tcp6_proc_exit(net);
1019 udp6_proc_exit(net);
1021 ipv6_cleanup_mibs(net);
1026 static void __net_exit inet6_net_exit(struct net *net)
1028 #ifdef CONFIG_PROC_FS
1029 udp6_proc_exit(net);
1030 tcp6_proc_exit(net);
1033 ipv6_cleanup_mibs(net);
1036 static struct pernet_operations inet6_net_ops = {
1037 .init = inet6_net_init,
1038 .exit = inet6_net_exit,
1041 static int ipv6_route_input(struct sk_buff *skb)
1043 ip6_route_input(skb);
1044 return skb_dst(skb)->error;
1047 static const struct ipv6_stub ipv6_stub_impl = {
1048 .ipv6_sock_mc_join = ipv6_sock_mc_join,
1049 .ipv6_sock_mc_drop = ipv6_sock_mc_drop,
1050 .ipv6_dst_lookup_flow = ip6_dst_lookup_flow,
1051 .ipv6_route_input = ipv6_route_input,
1052 .fib6_get_table = fib6_get_table,
1053 .fib6_table_lookup = fib6_table_lookup,
1054 .fib6_lookup = fib6_lookup,
1055 .fib6_select_path = fib6_select_path,
1056 .ip6_mtu_from_fib6 = ip6_mtu_from_fib6,
1057 .fib6_nh_init = fib6_nh_init,
1058 .fib6_nh_release = fib6_nh_release,
1059 .fib6_nh_release_dsts = fib6_nh_release_dsts,
1060 .fib6_update_sernum = fib6_update_sernum_stub,
1061 .fib6_rt_update = fib6_rt_update,
1062 .ip6_del_rt = ip6_del_rt,
1063 .udpv6_encap_enable = udpv6_encap_enable,
1064 .ndisc_send_na = ndisc_send_na,
1065 #if IS_ENABLED(CONFIG_XFRM)
1066 .xfrm6_local_rxpmtu = xfrm6_local_rxpmtu,
1067 .xfrm6_udp_encap_rcv = xfrm6_udp_encap_rcv,
1068 .xfrm6_rcv_encap = xfrm6_rcv_encap,
1071 .ipv6_fragment = ip6_fragment,
1072 .ipv6_dev_find = ipv6_dev_find,
1075 static const struct ipv6_bpf_stub ipv6_bpf_stub_impl = {
1076 .inet6_bind = __inet6_bind,
1077 .udp6_lib_lookup = __udp6_lib_lookup,
1078 .ipv6_setsockopt = do_ipv6_setsockopt,
1079 .ipv6_getsockopt = do_ipv6_getsockopt,
1080 .ipv6_dev_get_saddr = ipv6_dev_get_saddr,
1083 static int __init inet6_init(void)
1085 struct list_head *r;
1088 sock_skb_cb_check_size(sizeof(struct inet6_skb_parm));
1090 /* Register the socket-side information for inet6_create. */
1091 for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
1094 raw_hashinfo_init(&raw_v6_hashinfo);
1096 if (disable_ipv6_mod) {
1097 pr_info("Loaded, but administratively disabled, reboot required to enable\n");
1101 err = proto_register(&tcpv6_prot, 1);
1105 err = proto_register(&udpv6_prot, 1);
1107 goto out_unregister_tcp_proto;
1109 err = proto_register(&udplitev6_prot, 1);
1111 goto out_unregister_udp_proto;
1113 err = proto_register(&rawv6_prot, 1);
1115 goto out_unregister_udplite_proto;
1117 err = proto_register(&pingv6_prot, 1);
1119 goto out_unregister_raw_proto;
1121 /* We MUST register RAW sockets before we create the ICMP6,
1122 * IGMP6, or NDISC control sockets.
1126 goto out_unregister_ping_proto;
1128 /* Register the family here so that the init calls below will
1129 * be able to create sockets. (?? is this dangerous ??)
1131 err = sock_register(&inet6_family_ops);
1133 goto out_sock_register_fail;
1136 * ipngwg API draft makes clear that the correct semantics
1137 * for TCP and UDP is to consider one TCP and UDP instance
1138 * in a host available by both INET and INET6 APIs and
1139 * able to communicate via both network protocols.
1142 err = register_pernet_subsys(&inet6_net_ops);
1144 goto register_pernet_fail;
1145 err = ip6_mr_init();
1148 err = icmpv6_init();
1158 err = ipv6_netfilter_init();
1160 goto netfilter_fail;
1161 /* Create /proc/foo6 entries. */
1162 #ifdef CONFIG_PROC_FS
1164 if (raw6_proc_init())
1165 goto proc_raw6_fail;
1166 if (udplite6_proc_init())
1167 goto proc_udplite6_fail;
1168 if (ipv6_misc_proc_init())
1169 goto proc_misc6_fail;
1170 if (if6_proc_init())
1173 err = ip6_route_init();
1175 goto ip6_route_fail;
1176 err = ndisc_late_init();
1178 goto ndisc_late_fail;
1179 err = ip6_flowlabel_init();
1181 goto ip6_flowlabel_fail;
1182 err = ipv6_anycast_init();
1184 goto ipv6_anycast_fail;
1185 err = addrconf_init();
1189 /* Init v6 extension headers. */
1190 err = ipv6_exthdrs_init();
1192 goto ipv6_exthdrs_fail;
1194 err = ipv6_frag_init();
1196 goto ipv6_frag_fail;
1198 /* Init v6 transport protocols. */
1203 err = udplitev6_init();
1205 goto udplitev6_fail;
1207 err = udpv6_offload_init();
1209 goto udpv6_offload_fail;
1215 err = ipv6_packet_init();
1217 goto ipv6_packet_fail;
1219 err = pingv6_init();
1223 err = calipso_init();
1239 err = igmp6_late_init();
1241 goto igmp6_late_err;
1243 #ifdef CONFIG_SYSCTL
1244 err = ipv6_sysctl_register();
1249 /* ensure that ipv6 stubs are visible only after ipv6 is ready */
1251 ipv6_stub = &ipv6_stub_impl;
1252 ipv6_bpf_stub = &ipv6_bpf_stub_impl;
1256 #ifdef CONFIG_SYSCTL
1258 igmp6_late_cleanup();
1271 ipv6_packet_cleanup();
1275 udpv6_offload_exit();
1283 ipv6_exthdrs_exit();
1287 ipv6_anycast_cleanup();
1289 ip6_flowlabel_cleanup();
1291 ndisc_late_cleanup();
1293 ip6_route_cleanup();
1295 #ifdef CONFIG_PROC_FS
1298 ipv6_misc_proc_exit();
1300 udplite6_proc_exit();
1305 ipv6_netfilter_fini();
1315 unregister_pernet_subsys(&inet6_net_ops);
1316 register_pernet_fail:
1317 sock_unregister(PF_INET6);
1318 rtnl_unregister_all(PF_INET6);
1319 out_sock_register_fail:
1321 out_unregister_ping_proto:
1322 proto_unregister(&pingv6_prot);
1323 out_unregister_raw_proto:
1324 proto_unregister(&rawv6_prot);
1325 out_unregister_udplite_proto:
1326 proto_unregister(&udplitev6_prot);
1327 out_unregister_udp_proto:
1328 proto_unregister(&udpv6_prot);
1329 out_unregister_tcp_proto:
1330 proto_unregister(&tcpv6_prot);
1333 module_init(inet6_init);
1335 MODULE_ALIAS_NETPROTO(PF_INET6);