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);
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;
224 sk->sk_txrehash = READ_ONCE(net->core.sysctl_txrehash);
226 /* Init the ipv4 part of the socket since we can have sockets
227 * using v6 API for ipv4.
234 RCU_INIT_POINTER(inet->mc_list, NULL);
237 if (READ_ONCE(net->ipv4.sysctl_ip_no_pmtu_disc))
238 inet->pmtudisc = IP_PMTUDISC_DONT;
240 inet->pmtudisc = IP_PMTUDISC_WANT;
242 * Increment only the relevant sk_prot->socks debug field, this changes
243 * the previous behaviour of incrementing both the equivalent to
244 * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
246 * This allows better debug granularity as we'll know exactly how many
247 * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
248 * transport protocol socks. -acme
250 sk_refcnt_debug_inc(sk);
252 if (inet->inet_num) {
253 /* It assumes that any protocol which allows
254 * the user to assign a number at socket
255 * creation time automatically shares.
257 inet->inet_sport = htons(inet->inet_num);
258 err = sk->sk_prot->hash(sk);
260 sk_common_release(sk);
264 if (sk->sk_prot->init) {
265 err = sk->sk_prot->init(sk);
267 sk_common_release(sk);
273 err = BPF_CGROUP_RUN_PROG_INET_SOCK(sk);
275 sk_common_release(sk);
286 static int __inet6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len,
289 struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr;
290 struct inet_sock *inet = inet_sk(sk);
291 struct ipv6_pinfo *np = inet6_sk(sk);
292 struct net *net = sock_net(sk);
299 if (addr->sin6_family != AF_INET6)
300 return -EAFNOSUPPORT;
302 addr_type = ipv6_addr_type(&addr->sin6_addr);
303 if ((addr_type & IPV6_ADDR_MULTICAST) && sk->sk_type == SOCK_STREAM)
306 snum = ntohs(addr->sin6_port);
307 if (!(flags & BIND_NO_CAP_NET_BIND_SERVICE) &&
308 snum && inet_port_requires_bind_service(net, snum) &&
309 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
312 if (flags & BIND_WITH_LOCK)
315 /* Check these errors (active socket, double bind). */
316 if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
321 /* Check if the address belongs to the host. */
322 if (addr_type == IPV6_ADDR_MAPPED) {
323 struct net_device *dev = NULL;
326 /* Binding to v4-mapped address on a v6-only socket
329 if (ipv6_only_sock(sk)) {
335 if (sk->sk_bound_dev_if) {
336 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
343 /* Reproduce AF_INET checks to make the bindings consistent */
344 v4addr = addr->sin6_addr.s6_addr32[3];
345 chk_addr_ret = inet_addr_type_dev_table(net, dev, v4addr);
348 if (!inet_addr_valid_or_nonlocal(net, inet, v4addr,
350 err = -EADDRNOTAVAIL;
354 if (addr_type != IPV6_ADDR_ANY) {
355 struct net_device *dev = NULL;
358 if (__ipv6_addr_needs_scope_id(addr_type)) {
359 if (addr_len >= sizeof(struct sockaddr_in6) &&
360 addr->sin6_scope_id) {
361 /* Override any existing binding, if another one
362 * is supplied by user.
364 sk->sk_bound_dev_if = addr->sin6_scope_id;
367 /* Binding to link-local address requires an interface */
368 if (!sk->sk_bound_dev_if) {
374 if (sk->sk_bound_dev_if) {
375 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
382 /* ipv4 addr of the socket is invalid. Only the
383 * unspecified and mapped address have a v4 equivalent.
385 v4addr = LOOPBACK4_IPV6;
386 if (!(addr_type & IPV6_ADDR_MULTICAST)) {
387 if (!ipv6_can_nonlocal_bind(net, inet) &&
388 !ipv6_chk_addr(net, &addr->sin6_addr,
390 err = -EADDRNOTAVAIL;
398 inet->inet_rcv_saddr = v4addr;
399 inet->inet_saddr = v4addr;
401 sk->sk_v6_rcv_saddr = addr->sin6_addr;
403 if (!(addr_type & IPV6_ADDR_MULTICAST))
404 np->saddr = addr->sin6_addr;
406 saved_ipv6only = sk->sk_ipv6only;
407 if (addr_type != IPV6_ADDR_ANY && addr_type != IPV6_ADDR_MAPPED)
410 /* Make sure we are allowed to bind here. */
411 if (snum || !(inet->bind_address_no_port ||
412 (flags & BIND_FORCE_ADDRESS_NO_PORT))) {
413 err = sk->sk_prot->get_port(sk, snum);
415 sk->sk_ipv6only = saved_ipv6only;
416 inet_reset_saddr(sk);
419 if (!(flags & BIND_FROM_BPF)) {
420 err = BPF_CGROUP_RUN_PROG_INET6_POST_BIND(sk);
422 sk->sk_ipv6only = saved_ipv6only;
423 inet_reset_saddr(sk);
424 if (sk->sk_prot->put_port)
425 sk->sk_prot->put_port(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;
452 const struct proto *prot;
455 /* IPV6_ADDRFORM can change sk->sk_prot under us. */
456 prot = READ_ONCE(sk->sk_prot);
457 /* If the socket has its own bind function then use it. */
459 return prot->bind(sk, uaddr, addr_len);
461 if (addr_len < SIN6_LEN_RFC2133)
464 /* BPF prog is run before any checks are done so that if the prog
465 * changes context in a wrong way it will be caught.
467 err = BPF_CGROUP_RUN_PROG_INET_BIND_LOCK(sk, uaddr,
468 CGROUP_INET6_BIND, &flags);
472 return __inet6_bind(sk, uaddr, addr_len, flags);
474 EXPORT_SYMBOL(inet6_bind);
476 int inet6_release(struct socket *sock)
478 struct sock *sk = sock->sk;
484 ipv6_sock_mc_close(sk);
487 ipv6_sock_ac_close(sk);
489 return inet_release(sock);
491 EXPORT_SYMBOL(inet6_release);
493 void inet6_destroy_sock(struct sock *sk)
495 struct ipv6_pinfo *np = inet6_sk(sk);
497 struct ipv6_txoptions *opt;
499 /* Release rx options */
501 skb = xchg(&np->pktoptions, NULL);
504 skb = xchg(&np->rxpmtu, NULL);
507 /* Free flowlabels */
508 fl6_free_socklist(sk);
510 /* Free tx options */
512 opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL);
514 atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
518 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
520 void inet6_cleanup_sock(struct sock *sk)
522 inet6_destroy_sock(sk);
524 EXPORT_SYMBOL_GPL(inet6_cleanup_sock);
527 * This does both peername and sockname.
529 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
532 struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
533 struct sock *sk = sock->sk;
534 struct inet_sock *inet = inet_sk(sk);
535 struct ipv6_pinfo *np = inet6_sk(sk);
537 sin->sin6_family = AF_INET6;
538 sin->sin6_flowinfo = 0;
539 sin->sin6_scope_id = 0;
542 if (!inet->inet_dport ||
543 (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
548 sin->sin6_port = inet->inet_dport;
549 sin->sin6_addr = sk->sk_v6_daddr;
551 sin->sin6_flowinfo = np->flow_label;
552 BPF_CGROUP_RUN_SA_PROG(sk, (struct sockaddr *)sin,
553 CGROUP_INET6_GETPEERNAME);
555 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
556 sin->sin6_addr = np->saddr;
558 sin->sin6_addr = sk->sk_v6_rcv_saddr;
559 sin->sin6_port = inet->inet_sport;
560 BPF_CGROUP_RUN_SA_PROG(sk, (struct sockaddr *)sin,
561 CGROUP_INET6_GETSOCKNAME);
563 sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
564 sk->sk_bound_dev_if);
568 EXPORT_SYMBOL(inet6_getname);
570 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
572 void __user *argp = (void __user *)arg;
573 struct sock *sk = sock->sk;
574 struct net *net = sock_net(sk);
575 const struct proto *prot;
580 struct in6_rtmsg rtmsg;
582 if (copy_from_user(&rtmsg, argp, sizeof(rtmsg)))
584 return ipv6_route_ioctl(net, cmd, &rtmsg);
587 return addrconf_add_ifaddr(net, argp);
589 return addrconf_del_ifaddr(net, argp);
591 return addrconf_set_dstaddr(net, argp);
593 /* IPV6_ADDRFORM can change sk->sk_prot under us. */
594 prot = READ_ONCE(sk->sk_prot);
597 return prot->ioctl(sk, cmd, arg);
602 EXPORT_SYMBOL(inet6_ioctl);
605 struct compat_in6_rtmsg {
606 struct in6_addr rtmsg_dst;
607 struct in6_addr rtmsg_src;
608 struct in6_addr rtmsg_gateway;
618 static int inet6_compat_routing_ioctl(struct sock *sk, unsigned int cmd,
619 struct compat_in6_rtmsg __user *ur)
623 if (copy_from_user(&rt.rtmsg_dst, &ur->rtmsg_dst,
624 3 * sizeof(struct in6_addr)) ||
625 get_user(rt.rtmsg_type, &ur->rtmsg_type) ||
626 get_user(rt.rtmsg_dst_len, &ur->rtmsg_dst_len) ||
627 get_user(rt.rtmsg_src_len, &ur->rtmsg_src_len) ||
628 get_user(rt.rtmsg_metric, &ur->rtmsg_metric) ||
629 get_user(rt.rtmsg_info, &ur->rtmsg_info) ||
630 get_user(rt.rtmsg_flags, &ur->rtmsg_flags) ||
631 get_user(rt.rtmsg_ifindex, &ur->rtmsg_ifindex))
635 return ipv6_route_ioctl(sock_net(sk), cmd, &rt);
638 int inet6_compat_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
640 void __user *argp = compat_ptr(arg);
641 struct sock *sk = sock->sk;
646 return inet6_compat_routing_ioctl(sk, cmd, argp);
651 EXPORT_SYMBOL_GPL(inet6_compat_ioctl);
652 #endif /* CONFIG_COMPAT */
654 INDIRECT_CALLABLE_DECLARE(int udpv6_sendmsg(struct sock *, struct msghdr *,
656 int inet6_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
658 struct sock *sk = sock->sk;
659 const struct proto *prot;
661 if (unlikely(inet_send_prepare(sk)))
664 /* IPV6_ADDRFORM can change sk->sk_prot under us. */
665 prot = READ_ONCE(sk->sk_prot);
666 return INDIRECT_CALL_2(prot->sendmsg, tcp_sendmsg, udpv6_sendmsg,
670 INDIRECT_CALLABLE_DECLARE(int udpv6_recvmsg(struct sock *, struct msghdr *,
671 size_t, int, int *));
672 int inet6_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
675 struct sock *sk = sock->sk;
676 const struct proto *prot;
680 if (likely(!(flags & MSG_ERRQUEUE)))
681 sock_rps_record_flow(sk);
683 /* IPV6_ADDRFORM can change sk->sk_prot under us. */
684 prot = READ_ONCE(sk->sk_prot);
685 err = INDIRECT_CALL_2(prot->recvmsg, tcp_recvmsg, udpv6_recvmsg,
686 sk, msg, size, flags, &addr_len);
688 msg->msg_namelen = addr_len;
692 const struct proto_ops inet6_stream_ops = {
694 .owner = THIS_MODULE,
695 .release = inet6_release,
697 .connect = inet_stream_connect, /* ok */
698 .socketpair = sock_no_socketpair, /* a do nothing */
699 .accept = inet_accept, /* ok */
700 .getname = inet6_getname,
701 .poll = tcp_poll, /* ok */
702 .ioctl = inet6_ioctl, /* must change */
703 .gettstamp = sock_gettstamp,
704 .listen = inet_listen, /* ok */
705 .shutdown = inet_shutdown, /* ok */
706 .setsockopt = sock_common_setsockopt, /* ok */
707 .getsockopt = sock_common_getsockopt, /* ok */
708 .sendmsg = inet6_sendmsg, /* retpoline's sake */
709 .recvmsg = inet6_recvmsg, /* retpoline's sake */
713 .sendpage = inet_sendpage,
714 .sendmsg_locked = tcp_sendmsg_locked,
715 .sendpage_locked = tcp_sendpage_locked,
716 .splice_read = tcp_splice_read,
717 .read_sock = tcp_read_sock,
718 .read_skb = tcp_read_skb,
719 .peek_len = tcp_peek_len,
721 .compat_ioctl = inet6_compat_ioctl,
723 .set_rcvlowat = tcp_set_rcvlowat,
726 const struct proto_ops inet6_dgram_ops = {
728 .owner = THIS_MODULE,
729 .release = inet6_release,
731 .connect = inet_dgram_connect, /* ok */
732 .socketpair = sock_no_socketpair, /* a do nothing */
733 .accept = sock_no_accept, /* a do nothing */
734 .getname = inet6_getname,
735 .poll = udp_poll, /* ok */
736 .ioctl = inet6_ioctl, /* must change */
737 .gettstamp = sock_gettstamp,
738 .listen = sock_no_listen, /* ok */
739 .shutdown = inet_shutdown, /* ok */
740 .setsockopt = sock_common_setsockopt, /* ok */
741 .getsockopt = sock_common_getsockopt, /* ok */
742 .sendmsg = inet6_sendmsg, /* retpoline's sake */
743 .recvmsg = inet6_recvmsg, /* retpoline's sake */
744 .read_skb = udp_read_skb,
745 .mmap = sock_no_mmap,
746 .sendpage = sock_no_sendpage,
747 .set_peek_off = sk_set_peek_off,
749 .compat_ioctl = inet6_compat_ioctl,
753 static const struct net_proto_family inet6_family_ops = {
755 .create = inet6_create,
756 .owner = THIS_MODULE,
759 int inet6_register_protosw(struct inet_protosw *p)
761 struct list_head *lh;
762 struct inet_protosw *answer;
763 struct list_head *last_perm;
764 int protocol = p->protocol;
767 spin_lock_bh(&inetsw6_lock);
770 if (p->type >= SOCK_MAX)
773 /* If we are trying to override a permanent protocol, bail. */
776 last_perm = &inetsw6[p->type];
777 list_for_each(lh, &inetsw6[p->type]) {
778 answer = list_entry(lh, struct inet_protosw, list);
780 /* Check only the non-wild match. */
781 if (INET_PROTOSW_PERMANENT & answer->flags) {
782 if (protocol == answer->protocol)
792 /* Add the new entry after the last permanent entry if any, so that
793 * the new entry does not override a permanent entry when matched with
794 * a wild-card protocol. But it is allowed to override any existing
795 * non-permanent entry. This means that when we remove this entry, the
796 * system automatically returns to the old behavior.
798 list_add_rcu(&p->list, last_perm);
801 spin_unlock_bh(&inetsw6_lock);
805 pr_err("Attempt to override permanent protocol %d\n", protocol);
809 pr_err("Ignoring attempt to register invalid socket type %d\n",
813 EXPORT_SYMBOL(inet6_register_protosw);
816 inet6_unregister_protosw(struct inet_protosw *p)
818 if (INET_PROTOSW_PERMANENT & p->flags) {
819 pr_err("Attempt to unregister permanent protocol %d\n",
822 spin_lock_bh(&inetsw6_lock);
823 list_del_rcu(&p->list);
824 spin_unlock_bh(&inetsw6_lock);
829 EXPORT_SYMBOL(inet6_unregister_protosw);
831 int inet6_sk_rebuild_header(struct sock *sk)
833 struct ipv6_pinfo *np = inet6_sk(sk);
834 struct dst_entry *dst;
836 dst = __sk_dst_check(sk, np->dst_cookie);
839 struct inet_sock *inet = inet_sk(sk);
840 struct in6_addr *final_p, final;
843 memset(&fl6, 0, sizeof(fl6));
844 fl6.flowi6_proto = sk->sk_protocol;
845 fl6.daddr = sk->sk_v6_daddr;
846 fl6.saddr = np->saddr;
847 fl6.flowlabel = np->flow_label;
848 fl6.flowi6_oif = sk->sk_bound_dev_if;
849 fl6.flowi6_mark = sk->sk_mark;
850 fl6.fl6_dport = inet->inet_dport;
851 fl6.fl6_sport = inet->inet_sport;
852 fl6.flowi6_uid = sk->sk_uid;
853 security_sk_classify_flow(sk, flowi6_to_flowi_common(&fl6));
856 final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt),
860 dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p);
862 sk->sk_route_caps = 0;
863 sk->sk_err_soft = -PTR_ERR(dst);
867 ip6_dst_store(sk, dst, NULL, NULL);
872 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
874 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb,
875 const struct inet6_skb_parm *opt)
877 const struct ipv6_pinfo *np = inet6_sk(sk);
880 if (((opt->flags & IP6SKB_HOPBYHOP) &&
881 (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) ||
882 (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) &&
883 np->rxopt.bits.rxflow) ||
884 (opt->srcrt && (np->rxopt.bits.srcrt ||
885 np->rxopt.bits.osrcrt)) ||
886 ((opt->dst1 || opt->dst0) &&
887 (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
892 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
894 static struct packet_type ipv6_packet_type __read_mostly = {
895 .type = cpu_to_be16(ETH_P_IPV6),
897 .list_func = ipv6_list_rcv,
900 static int __init ipv6_packet_init(void)
902 dev_add_pack(&ipv6_packet_type);
906 static void ipv6_packet_cleanup(void)
908 dev_remove_pack(&ipv6_packet_type);
911 static int __net_init ipv6_init_mibs(struct net *net)
915 net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib);
916 if (!net->mib.udp_stats_in6)
918 net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib);
919 if (!net->mib.udplite_stats_in6)
920 goto err_udplite_mib;
921 net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib);
922 if (!net->mib.ipv6_statistics)
925 for_each_possible_cpu(i) {
926 struct ipstats_mib *af_inet6_stats;
927 af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i);
928 u64_stats_init(&af_inet6_stats->syncp);
932 net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib);
933 if (!net->mib.icmpv6_statistics)
935 net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
937 if (!net->mib.icmpv6msg_statistics)
938 goto err_icmpmsg_mib;
942 free_percpu(net->mib.icmpv6_statistics);
944 free_percpu(net->mib.ipv6_statistics);
946 free_percpu(net->mib.udplite_stats_in6);
948 free_percpu(net->mib.udp_stats_in6);
952 static void ipv6_cleanup_mibs(struct net *net)
954 free_percpu(net->mib.udp_stats_in6);
955 free_percpu(net->mib.udplite_stats_in6);
956 free_percpu(net->mib.ipv6_statistics);
957 free_percpu(net->mib.icmpv6_statistics);
958 kfree(net->mib.icmpv6msg_statistics);
961 static int __net_init inet6_net_init(struct net *net)
965 net->ipv6.sysctl.bindv6only = 0;
966 net->ipv6.sysctl.icmpv6_time = 1*HZ;
967 net->ipv6.sysctl.icmpv6_echo_ignore_all = 0;
968 net->ipv6.sysctl.icmpv6_echo_ignore_multicast = 0;
969 net->ipv6.sysctl.icmpv6_echo_ignore_anycast = 0;
971 /* By default, rate limit error messages.
972 * Except for pmtu discovery, it would break it.
973 * proc_do_large_bitmap needs pointer to the bitmap.
975 bitmap_set(net->ipv6.sysctl.icmpv6_ratemask, 0, ICMPV6_ERRMSG_MAX + 1);
976 bitmap_clear(net->ipv6.sysctl.icmpv6_ratemask, ICMPV6_PKT_TOOBIG, 1);
977 net->ipv6.sysctl.icmpv6_ratemask_ptr = net->ipv6.sysctl.icmpv6_ratemask;
979 net->ipv6.sysctl.flowlabel_consistency = 1;
980 net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS;
981 net->ipv6.sysctl.idgen_retries = 3;
982 net->ipv6.sysctl.idgen_delay = 1 * HZ;
983 net->ipv6.sysctl.flowlabel_state_ranges = 0;
984 net->ipv6.sysctl.max_dst_opts_cnt = IP6_DEFAULT_MAX_DST_OPTS_CNT;
985 net->ipv6.sysctl.max_hbh_opts_cnt = IP6_DEFAULT_MAX_HBH_OPTS_CNT;
986 net->ipv6.sysctl.max_dst_opts_len = IP6_DEFAULT_MAX_DST_OPTS_LEN;
987 net->ipv6.sysctl.max_hbh_opts_len = IP6_DEFAULT_MAX_HBH_OPTS_LEN;
988 net->ipv6.sysctl.fib_notify_on_flag_change = 0;
989 atomic_set(&net->ipv6.fib6_sernum, 1);
991 net->ipv6.sysctl.ioam6_id = IOAM6_DEFAULT_ID;
992 net->ipv6.sysctl.ioam6_id_wide = IOAM6_DEFAULT_ID_WIDE;
994 err = ipv6_init_mibs(net);
997 #ifdef CONFIG_PROC_FS
998 err = udp6_proc_init(net);
1001 err = tcp6_proc_init(net);
1003 goto proc_tcp6_fail;
1004 err = ac6_proc_init(net);
1010 #ifdef CONFIG_PROC_FS
1012 tcp6_proc_exit(net);
1014 udp6_proc_exit(net);
1016 ipv6_cleanup_mibs(net);
1021 static void __net_exit inet6_net_exit(struct net *net)
1023 #ifdef CONFIG_PROC_FS
1024 udp6_proc_exit(net);
1025 tcp6_proc_exit(net);
1028 ipv6_cleanup_mibs(net);
1031 static struct pernet_operations inet6_net_ops = {
1032 .init = inet6_net_init,
1033 .exit = inet6_net_exit,
1036 static int ipv6_route_input(struct sk_buff *skb)
1038 ip6_route_input(skb);
1039 return skb_dst(skb)->error;
1042 static const struct ipv6_stub ipv6_stub_impl = {
1043 .ipv6_sock_mc_join = ipv6_sock_mc_join,
1044 .ipv6_sock_mc_drop = ipv6_sock_mc_drop,
1045 .ipv6_dst_lookup_flow = ip6_dst_lookup_flow,
1046 .ipv6_route_input = ipv6_route_input,
1047 .fib6_get_table = fib6_get_table,
1048 .fib6_table_lookup = fib6_table_lookup,
1049 .fib6_lookup = fib6_lookup,
1050 .fib6_select_path = fib6_select_path,
1051 .ip6_mtu_from_fib6 = ip6_mtu_from_fib6,
1052 .fib6_nh_init = fib6_nh_init,
1053 .fib6_nh_release = fib6_nh_release,
1054 .fib6_nh_release_dsts = fib6_nh_release_dsts,
1055 .fib6_update_sernum = fib6_update_sernum_stub,
1056 .fib6_rt_update = fib6_rt_update,
1057 .ip6_del_rt = ip6_del_rt,
1058 .udpv6_encap_enable = udpv6_encap_enable,
1059 .ndisc_send_na = ndisc_send_na,
1060 #if IS_ENABLED(CONFIG_XFRM)
1061 .xfrm6_local_rxpmtu = xfrm6_local_rxpmtu,
1062 .xfrm6_udp_encap_rcv = xfrm6_udp_encap_rcv,
1063 .xfrm6_rcv_encap = xfrm6_rcv_encap,
1066 .ipv6_fragment = ip6_fragment,
1067 .ipv6_dev_find = ipv6_dev_find,
1070 static const struct ipv6_bpf_stub ipv6_bpf_stub_impl = {
1071 .inet6_bind = __inet6_bind,
1072 .udp6_lib_lookup = __udp6_lib_lookup,
1073 .ipv6_setsockopt = do_ipv6_setsockopt,
1074 .ipv6_getsockopt = do_ipv6_getsockopt,
1077 static int __init inet6_init(void)
1079 struct list_head *r;
1082 sock_skb_cb_check_size(sizeof(struct inet6_skb_parm));
1084 /* Register the socket-side information for inet6_create. */
1085 for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
1088 raw_hashinfo_init(&raw_v6_hashinfo);
1090 if (disable_ipv6_mod) {
1091 pr_info("Loaded, but administratively disabled, reboot required to enable\n");
1095 err = proto_register(&tcpv6_prot, 1);
1099 err = proto_register(&udpv6_prot, 1);
1101 goto out_unregister_tcp_proto;
1103 err = proto_register(&udplitev6_prot, 1);
1105 goto out_unregister_udp_proto;
1107 err = proto_register(&rawv6_prot, 1);
1109 goto out_unregister_udplite_proto;
1111 err = proto_register(&pingv6_prot, 1);
1113 goto out_unregister_raw_proto;
1115 /* We MUST register RAW sockets before we create the ICMP6,
1116 * IGMP6, or NDISC control sockets.
1120 goto out_unregister_ping_proto;
1122 /* Register the family here so that the init calls below will
1123 * be able to create sockets. (?? is this dangerous ??)
1125 err = sock_register(&inet6_family_ops);
1127 goto out_sock_register_fail;
1130 * ipngwg API draft makes clear that the correct semantics
1131 * for TCP and UDP is to consider one TCP and UDP instance
1132 * in a host available by both INET and INET6 APIs and
1133 * able to communicate via both network protocols.
1136 err = register_pernet_subsys(&inet6_net_ops);
1138 goto register_pernet_fail;
1139 err = ip6_mr_init();
1142 err = icmpv6_init();
1152 err = ipv6_netfilter_init();
1154 goto netfilter_fail;
1155 /* Create /proc/foo6 entries. */
1156 #ifdef CONFIG_PROC_FS
1158 if (raw6_proc_init())
1159 goto proc_raw6_fail;
1160 if (udplite6_proc_init())
1161 goto proc_udplite6_fail;
1162 if (ipv6_misc_proc_init())
1163 goto proc_misc6_fail;
1164 if (if6_proc_init())
1167 err = ip6_route_init();
1169 goto ip6_route_fail;
1170 err = ndisc_late_init();
1172 goto ndisc_late_fail;
1173 err = ip6_flowlabel_init();
1175 goto ip6_flowlabel_fail;
1176 err = ipv6_anycast_init();
1178 goto ipv6_anycast_fail;
1179 err = addrconf_init();
1183 /* Init v6 extension headers. */
1184 err = ipv6_exthdrs_init();
1186 goto ipv6_exthdrs_fail;
1188 err = ipv6_frag_init();
1190 goto ipv6_frag_fail;
1192 /* Init v6 transport protocols. */
1197 err = udplitev6_init();
1199 goto udplitev6_fail;
1201 err = udpv6_offload_init();
1203 goto udpv6_offload_fail;
1209 err = ipv6_packet_init();
1211 goto ipv6_packet_fail;
1213 err = pingv6_init();
1217 err = calipso_init();
1233 err = igmp6_late_init();
1235 goto igmp6_late_err;
1237 #ifdef CONFIG_SYSCTL
1238 err = ipv6_sysctl_register();
1243 /* ensure that ipv6 stubs are visible only after ipv6 is ready */
1245 ipv6_stub = &ipv6_stub_impl;
1246 ipv6_bpf_stub = &ipv6_bpf_stub_impl;
1250 #ifdef CONFIG_SYSCTL
1252 igmp6_late_cleanup();
1265 ipv6_packet_cleanup();
1269 udpv6_offload_exit();
1277 ipv6_exthdrs_exit();
1281 ipv6_anycast_cleanup();
1283 ip6_flowlabel_cleanup();
1285 ndisc_late_cleanup();
1287 ip6_route_cleanup();
1289 #ifdef CONFIG_PROC_FS
1292 ipv6_misc_proc_exit();
1294 udplite6_proc_exit();
1299 ipv6_netfilter_fini();
1309 unregister_pernet_subsys(&inet6_net_ops);
1310 register_pernet_fail:
1311 sock_unregister(PF_INET6);
1312 rtnl_unregister_all(PF_INET6);
1313 out_sock_register_fail:
1315 out_unregister_ping_proto:
1316 proto_unregister(&pingv6_prot);
1317 out_unregister_raw_proto:
1318 proto_unregister(&rawv6_prot);
1319 out_unregister_udplite_proto:
1320 proto_unregister(&udplitev6_prot);
1321 out_unregister_udp_proto:
1322 proto_unregister(&udpv6_prot);
1323 out_unregister_tcp_proto:
1324 proto_unregister(&tcpv6_prot);
1327 module_init(inet6_init);
1329 MODULE_ALIAS_NETPROTO(PF_INET6);