Linux 6.12-rc7
[linux-stable.git] / net / ipv6 / af_inet6.c
blobba69b86f1c7d5e8d9c8b02f3e70a95d913082701
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3 * PF_INET6 socket protocol family
4 * Linux INET6 implementation
6 * Authors:
7 * Pedro Roque <roque@di.fc.ul.pt>
9 * Adapted from linux/net/ipv4/af_inet.c
11 * Fixes:
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>
24 #include <linux/in.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>
31 #include <linux/mm.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>
43 #include <net/ip.h>
44 #include <net/ipv6.h>
45 #include <net/udp.h>
46 #include <net/udplite.h>
47 #include <net/tcp.h>
48 #include <net/ping.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>
59 #endif
60 #include <net/calipso.h>
61 #include <net/seg6.h>
62 #include <net/rpl.h>
63 #include <net/compat.h>
64 #include <net/xfrm.h>
65 #include <net/ioam6.h>
66 #include <net/rawv6.h>
67 #include <net/rps.h>
69 #include <linux/uaccess.h>
70 #include <linux/mroute6.h>
72 #include "ip6_offload.h"
74 MODULE_AUTHOR("Cast of dozens");
75 MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
76 MODULE_LICENSE("GPL");
78 /* The inetsw6 table contains everything that inet6_create needs to
79 * build a new socket.
81 static struct list_head inetsw6[SOCK_MAX];
82 static DEFINE_SPINLOCK(inetsw6_lock);
84 struct ipv6_params ipv6_defaults = {
85 .disable_ipv6 = 0,
86 .autoconf = 1,
89 static int disable_ipv6_mod;
91 module_param_named(disable, disable_ipv6_mod, int, 0444);
92 MODULE_PARM_DESC(disable, "Disable IPv6 module such that it is non-functional");
94 module_param_named(disable_ipv6, ipv6_defaults.disable_ipv6, int, 0444);
95 MODULE_PARM_DESC(disable_ipv6, "Disable IPv6 on all interfaces");
97 module_param_named(autoconf, ipv6_defaults.autoconf, int, 0444);
98 MODULE_PARM_DESC(autoconf, "Enable IPv6 address autoconfiguration on all interfaces");
100 bool ipv6_mod_enabled(void)
102 return disable_ipv6_mod == 0;
104 EXPORT_SYMBOL_GPL(ipv6_mod_enabled);
106 static struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
108 const int offset = sk->sk_prot->ipv6_pinfo_offset;
110 return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
113 void inet6_sock_destruct(struct sock *sk)
115 inet6_cleanup_sock(sk);
116 inet_sock_destruct(sk);
118 EXPORT_SYMBOL_GPL(inet6_sock_destruct);
120 static int inet6_create(struct net *net, struct socket *sock, int protocol,
121 int kern)
123 struct inet_sock *inet;
124 struct ipv6_pinfo *np;
125 struct sock *sk;
126 struct inet_protosw *answer;
127 struct proto *answer_prot;
128 unsigned char answer_flags;
129 int try_loading_module = 0;
130 int err;
132 if (protocol < 0 || protocol >= IPPROTO_MAX)
133 return -EINVAL;
135 /* Look for the requested type/protocol pair. */
136 lookup_protocol:
137 err = -ESOCKTNOSUPPORT;
138 rcu_read_lock();
139 list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
141 err = 0;
142 /* Check the non-wild match. */
143 if (protocol == answer->protocol) {
144 if (protocol != IPPROTO_IP)
145 break;
146 } else {
147 /* Check for the two wild cases. */
148 if (IPPROTO_IP == protocol) {
149 protocol = answer->protocol;
150 break;
152 if (IPPROTO_IP == answer->protocol)
153 break;
155 err = -EPROTONOSUPPORT;
158 if (err) {
159 if (try_loading_module < 2) {
160 rcu_read_unlock();
162 * Be more specific, e.g. net-pf-10-proto-132-type-1
163 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
165 if (++try_loading_module == 1)
166 request_module("net-pf-%d-proto-%d-type-%d",
167 PF_INET6, protocol, sock->type);
169 * Fall back to generic, e.g. net-pf-10-proto-132
170 * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
172 else
173 request_module("net-pf-%d-proto-%d",
174 PF_INET6, protocol);
175 goto lookup_protocol;
176 } else
177 goto out_rcu_unlock;
180 err = -EPERM;
181 if (sock->type == SOCK_RAW && !kern &&
182 !ns_capable(net->user_ns, CAP_NET_RAW))
183 goto out_rcu_unlock;
185 sock->ops = answer->ops;
186 answer_prot = answer->prot;
187 answer_flags = answer->flags;
188 rcu_read_unlock();
190 WARN_ON(!answer_prot->slab);
192 err = -ENOBUFS;
193 sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot, kern);
194 if (!sk)
195 goto out;
197 sock_init_data(sock, sk);
199 err = 0;
200 if (INET_PROTOSW_REUSE & answer_flags)
201 sk->sk_reuse = SK_CAN_REUSE;
203 if (INET_PROTOSW_ICSK & answer_flags)
204 inet_init_csk_locks(sk);
206 inet = inet_sk(sk);
207 inet_assign_bit(IS_ICSK, sk, INET_PROTOSW_ICSK & answer_flags);
209 if (SOCK_RAW == sock->type) {
210 inet->inet_num = protocol;
211 if (IPPROTO_RAW == protocol)
212 inet_set_bit(HDRINCL, sk);
215 sk->sk_destruct = inet6_sock_destruct;
216 sk->sk_family = PF_INET6;
217 sk->sk_protocol = protocol;
219 sk->sk_backlog_rcv = answer->prot->backlog_rcv;
221 inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
222 np->hop_limit = -1;
223 np->mcast_hops = IPV6_DEFAULT_MCASTHOPS;
224 inet6_set_bit(MC6_LOOP, sk);
225 inet6_set_bit(MC6_ALL, sk);
226 np->pmtudisc = IPV6_PMTUDISC_WANT;
227 inet6_assign_bit(REPFLOW, sk, net->ipv6.sysctl.flowlabel_reflect &
228 FLOWLABEL_REFLECT_ESTABLISHED);
229 sk->sk_ipv6only = net->ipv6.sysctl.bindv6only;
230 sk->sk_txrehash = READ_ONCE(net->core.sysctl_txrehash);
232 /* Init the ipv4 part of the socket since we can have sockets
233 * using v6 API for ipv4.
235 inet->uc_ttl = -1;
237 inet_set_bit(MC_LOOP, sk);
238 inet->mc_ttl = 1;
239 inet->mc_index = 0;
240 RCU_INIT_POINTER(inet->mc_list, NULL);
241 inet->rcv_tos = 0;
243 if (READ_ONCE(net->ipv4.sysctl_ip_no_pmtu_disc))
244 inet->pmtudisc = IP_PMTUDISC_DONT;
245 else
246 inet->pmtudisc = IP_PMTUDISC_WANT;
248 if (inet->inet_num) {
249 /* It assumes that any protocol which allows
250 * the user to assign a number at socket
251 * creation time automatically shares.
253 inet->inet_sport = htons(inet->inet_num);
254 err = sk->sk_prot->hash(sk);
255 if (err) {
256 sk_common_release(sk);
257 goto out;
260 if (sk->sk_prot->init) {
261 err = sk->sk_prot->init(sk);
262 if (err) {
263 sk_common_release(sk);
264 goto out;
268 if (!kern) {
269 err = BPF_CGROUP_RUN_PROG_INET_SOCK(sk);
270 if (err) {
271 sk_common_release(sk);
272 goto out;
275 out:
276 return err;
277 out_rcu_unlock:
278 rcu_read_unlock();
279 goto out;
282 static int __inet6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len,
283 u32 flags)
285 struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr;
286 struct inet_sock *inet = inet_sk(sk);
287 struct ipv6_pinfo *np = inet6_sk(sk);
288 struct net *net = sock_net(sk);
289 __be32 v4addr = 0;
290 unsigned short snum;
291 bool saved_ipv6only;
292 int addr_type = 0;
293 int err = 0;
295 if (addr->sin6_family != AF_INET6)
296 return -EAFNOSUPPORT;
298 addr_type = ipv6_addr_type(&addr->sin6_addr);
299 if ((addr_type & IPV6_ADDR_MULTICAST) && sk->sk_type == SOCK_STREAM)
300 return -EINVAL;
302 snum = ntohs(addr->sin6_port);
303 if (!(flags & BIND_NO_CAP_NET_BIND_SERVICE) &&
304 snum && inet_port_requires_bind_service(net, snum) &&
305 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
306 return -EACCES;
308 if (flags & BIND_WITH_LOCK)
309 lock_sock(sk);
311 /* Check these errors (active socket, double bind). */
312 if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
313 err = -EINVAL;
314 goto out;
317 /* Check if the address belongs to the host. */
318 if (addr_type == IPV6_ADDR_MAPPED) {
319 struct net_device *dev = NULL;
320 int chk_addr_ret;
322 /* Binding to v4-mapped address on a v6-only socket
323 * makes no sense
325 if (ipv6_only_sock(sk)) {
326 err = -EINVAL;
327 goto out;
330 rcu_read_lock();
331 if (sk->sk_bound_dev_if) {
332 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
333 if (!dev) {
334 err = -ENODEV;
335 goto out_unlock;
339 /* Reproduce AF_INET checks to make the bindings consistent */
340 v4addr = addr->sin6_addr.s6_addr32[3];
341 chk_addr_ret = inet_addr_type_dev_table(net, dev, v4addr);
342 rcu_read_unlock();
344 if (!inet_addr_valid_or_nonlocal(net, inet, v4addr,
345 chk_addr_ret)) {
346 err = -EADDRNOTAVAIL;
347 goto out;
349 } else {
350 if (addr_type != IPV6_ADDR_ANY) {
351 struct net_device *dev = NULL;
353 rcu_read_lock();
354 if (__ipv6_addr_needs_scope_id(addr_type)) {
355 if (addr_len >= sizeof(struct sockaddr_in6) &&
356 addr->sin6_scope_id) {
357 /* Override any existing binding, if another one
358 * is supplied by user.
360 sk->sk_bound_dev_if = addr->sin6_scope_id;
363 /* Binding to link-local address requires an interface */
364 if (!sk->sk_bound_dev_if) {
365 err = -EINVAL;
366 goto out_unlock;
370 if (sk->sk_bound_dev_if) {
371 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
372 if (!dev) {
373 err = -ENODEV;
374 goto out_unlock;
378 /* ipv4 addr of the socket is invalid. Only the
379 * unspecified and mapped address have a v4 equivalent.
381 v4addr = LOOPBACK4_IPV6;
382 if (!(addr_type & IPV6_ADDR_MULTICAST)) {
383 if (!ipv6_can_nonlocal_bind(net, inet) &&
384 !ipv6_chk_addr(net, &addr->sin6_addr,
385 dev, 0)) {
386 err = -EADDRNOTAVAIL;
387 goto out_unlock;
390 rcu_read_unlock();
394 inet->inet_rcv_saddr = v4addr;
395 inet->inet_saddr = v4addr;
397 sk->sk_v6_rcv_saddr = addr->sin6_addr;
399 if (!(addr_type & IPV6_ADDR_MULTICAST))
400 np->saddr = addr->sin6_addr;
402 saved_ipv6only = sk->sk_ipv6only;
403 if (addr_type != IPV6_ADDR_ANY && addr_type != IPV6_ADDR_MAPPED)
404 sk->sk_ipv6only = 1;
406 /* Make sure we are allowed to bind here. */
407 if (snum || !(inet_test_bit(BIND_ADDRESS_NO_PORT, sk) ||
408 (flags & BIND_FORCE_ADDRESS_NO_PORT))) {
409 err = sk->sk_prot->get_port(sk, snum);
410 if (err) {
411 sk->sk_ipv6only = saved_ipv6only;
412 inet_reset_saddr(sk);
413 goto out;
415 if (!(flags & BIND_FROM_BPF)) {
416 err = BPF_CGROUP_RUN_PROG_INET6_POST_BIND(sk);
417 if (err) {
418 sk->sk_ipv6only = saved_ipv6only;
419 inet_reset_saddr(sk);
420 if (sk->sk_prot->put_port)
421 sk->sk_prot->put_port(sk);
422 goto out;
427 if (addr_type != IPV6_ADDR_ANY)
428 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
429 if (snum)
430 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
431 inet->inet_sport = htons(inet->inet_num);
432 inet->inet_dport = 0;
433 inet->inet_daddr = 0;
434 out:
435 if (flags & BIND_WITH_LOCK)
436 release_sock(sk);
437 return err;
438 out_unlock:
439 rcu_read_unlock();
440 goto out;
443 int inet6_bind_sk(struct sock *sk, struct sockaddr *uaddr, int addr_len)
445 u32 flags = BIND_WITH_LOCK;
446 const struct proto *prot;
447 int err = 0;
449 /* IPV6_ADDRFORM can change sk->sk_prot under us. */
450 prot = READ_ONCE(sk->sk_prot);
451 /* If the socket has its own bind function then use it. */
452 if (prot->bind)
453 return prot->bind(sk, uaddr, addr_len);
455 if (addr_len < SIN6_LEN_RFC2133)
456 return -EINVAL;
458 /* BPF prog is run before any checks are done so that if the prog
459 * changes context in a wrong way it will be caught.
461 err = BPF_CGROUP_RUN_PROG_INET_BIND_LOCK(sk, uaddr, &addr_len,
462 CGROUP_INET6_BIND, &flags);
463 if (err)
464 return err;
466 return __inet6_bind(sk, uaddr, addr_len, flags);
469 /* bind for INET6 API */
470 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
472 return inet6_bind_sk(sock->sk, uaddr, addr_len);
474 EXPORT_SYMBOL(inet6_bind);
476 int inet6_release(struct socket *sock)
478 struct sock *sk = sock->sk;
480 if (!sk)
481 return -EINVAL;
483 /* Free mc lists */
484 ipv6_sock_mc_close(sk);
486 /* Free ac lists */
487 ipv6_sock_ac_close(sk);
489 return inet_release(sock);
491 EXPORT_SYMBOL(inet6_release);
493 void inet6_cleanup_sock(struct sock *sk)
495 struct ipv6_pinfo *np = inet6_sk(sk);
496 struct sk_buff *skb;
497 struct ipv6_txoptions *opt;
499 /* Release rx options */
501 skb = xchg(&np->pktoptions, NULL);
502 kfree_skb(skb);
504 skb = xchg(&np->rxpmtu, NULL);
505 kfree_skb(skb);
507 /* Free flowlabels */
508 fl6_free_socklist(sk);
510 /* Free tx options */
512 opt = unrcu_pointer(xchg(&np->opt, NULL));
513 if (opt) {
514 atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
515 txopt_put(opt);
518 EXPORT_SYMBOL_GPL(inet6_cleanup_sock);
521 * This does both peername and sockname.
523 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
524 int peer)
526 struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
527 int sin_addr_len = sizeof(*sin);
528 struct sock *sk = sock->sk;
529 struct inet_sock *inet = inet_sk(sk);
530 struct ipv6_pinfo *np = inet6_sk(sk);
532 sin->sin6_family = AF_INET6;
533 sin->sin6_flowinfo = 0;
534 sin->sin6_scope_id = 0;
535 lock_sock(sk);
536 if (peer) {
537 if (!inet->inet_dport ||
538 (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
539 peer == 1)) {
540 release_sock(sk);
541 return -ENOTCONN;
543 sin->sin6_port = inet->inet_dport;
544 sin->sin6_addr = sk->sk_v6_daddr;
545 if (inet6_test_bit(SNDFLOW, sk))
546 sin->sin6_flowinfo = np->flow_label;
547 BPF_CGROUP_RUN_SA_PROG(sk, (struct sockaddr *)sin, &sin_addr_len,
548 CGROUP_INET6_GETPEERNAME);
549 } else {
550 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
551 sin->sin6_addr = np->saddr;
552 else
553 sin->sin6_addr = sk->sk_v6_rcv_saddr;
554 sin->sin6_port = inet->inet_sport;
555 BPF_CGROUP_RUN_SA_PROG(sk, (struct sockaddr *)sin, &sin_addr_len,
556 CGROUP_INET6_GETSOCKNAME);
558 sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
559 sk->sk_bound_dev_if);
560 release_sock(sk);
561 return sin_addr_len;
563 EXPORT_SYMBOL(inet6_getname);
565 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
567 void __user *argp = (void __user *)arg;
568 struct sock *sk = sock->sk;
569 struct net *net = sock_net(sk);
570 const struct proto *prot;
572 switch (cmd) {
573 case SIOCADDRT:
574 case SIOCDELRT: {
575 struct in6_rtmsg rtmsg;
577 if (copy_from_user(&rtmsg, argp, sizeof(rtmsg)))
578 return -EFAULT;
579 return ipv6_route_ioctl(net, cmd, &rtmsg);
581 case SIOCSIFADDR:
582 return addrconf_add_ifaddr(net, argp);
583 case SIOCDIFADDR:
584 return addrconf_del_ifaddr(net, argp);
585 case SIOCSIFDSTADDR:
586 return addrconf_set_dstaddr(net, argp);
587 default:
588 /* IPV6_ADDRFORM can change sk->sk_prot under us. */
589 prot = READ_ONCE(sk->sk_prot);
590 if (!prot->ioctl)
591 return -ENOIOCTLCMD;
592 return sk_ioctl(sk, cmd, (void __user *)arg);
594 /*NOTREACHED*/
595 return 0;
597 EXPORT_SYMBOL(inet6_ioctl);
599 #ifdef CONFIG_COMPAT
600 struct compat_in6_rtmsg {
601 struct in6_addr rtmsg_dst;
602 struct in6_addr rtmsg_src;
603 struct in6_addr rtmsg_gateway;
604 u32 rtmsg_type;
605 u16 rtmsg_dst_len;
606 u16 rtmsg_src_len;
607 u32 rtmsg_metric;
608 u32 rtmsg_info;
609 u32 rtmsg_flags;
610 s32 rtmsg_ifindex;
613 static int inet6_compat_routing_ioctl(struct sock *sk, unsigned int cmd,
614 struct compat_in6_rtmsg __user *ur)
616 struct in6_rtmsg rt;
618 if (copy_from_user(&rt.rtmsg_dst, &ur->rtmsg_dst,
619 3 * sizeof(struct in6_addr)) ||
620 get_user(rt.rtmsg_type, &ur->rtmsg_type) ||
621 get_user(rt.rtmsg_dst_len, &ur->rtmsg_dst_len) ||
622 get_user(rt.rtmsg_src_len, &ur->rtmsg_src_len) ||
623 get_user(rt.rtmsg_metric, &ur->rtmsg_metric) ||
624 get_user(rt.rtmsg_info, &ur->rtmsg_info) ||
625 get_user(rt.rtmsg_flags, &ur->rtmsg_flags) ||
626 get_user(rt.rtmsg_ifindex, &ur->rtmsg_ifindex))
627 return -EFAULT;
630 return ipv6_route_ioctl(sock_net(sk), cmd, &rt);
633 int inet6_compat_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
635 void __user *argp = compat_ptr(arg);
636 struct sock *sk = sock->sk;
638 switch (cmd) {
639 case SIOCADDRT:
640 case SIOCDELRT:
641 return inet6_compat_routing_ioctl(sk, cmd, argp);
642 default:
643 return -ENOIOCTLCMD;
646 EXPORT_SYMBOL_GPL(inet6_compat_ioctl);
647 #endif /* CONFIG_COMPAT */
649 INDIRECT_CALLABLE_DECLARE(int udpv6_sendmsg(struct sock *, struct msghdr *,
650 size_t));
651 int inet6_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
653 struct sock *sk = sock->sk;
654 const struct proto *prot;
656 if (unlikely(inet_send_prepare(sk)))
657 return -EAGAIN;
659 /* IPV6_ADDRFORM can change sk->sk_prot under us. */
660 prot = READ_ONCE(sk->sk_prot);
661 return INDIRECT_CALL_2(prot->sendmsg, tcp_sendmsg, udpv6_sendmsg,
662 sk, msg, size);
665 INDIRECT_CALLABLE_DECLARE(int udpv6_recvmsg(struct sock *, struct msghdr *,
666 size_t, int, int *));
667 int inet6_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
668 int flags)
670 struct sock *sk = sock->sk;
671 const struct proto *prot;
672 int addr_len = 0;
673 int err;
675 if (likely(!(flags & MSG_ERRQUEUE)))
676 sock_rps_record_flow(sk);
678 /* IPV6_ADDRFORM can change sk->sk_prot under us. */
679 prot = READ_ONCE(sk->sk_prot);
680 err = INDIRECT_CALL_2(prot->recvmsg, tcp_recvmsg, udpv6_recvmsg,
681 sk, msg, size, flags, &addr_len);
682 if (err >= 0)
683 msg->msg_namelen = addr_len;
684 return err;
687 const struct proto_ops inet6_stream_ops = {
688 .family = PF_INET6,
689 .owner = THIS_MODULE,
690 .release = inet6_release,
691 .bind = inet6_bind,
692 .connect = inet_stream_connect, /* ok */
693 .socketpair = sock_no_socketpair, /* a do nothing */
694 .accept = inet_accept, /* ok */
695 .getname = inet6_getname,
696 .poll = tcp_poll, /* ok */
697 .ioctl = inet6_ioctl, /* must change */
698 .gettstamp = sock_gettstamp,
699 .listen = inet_listen, /* ok */
700 .shutdown = inet_shutdown, /* ok */
701 .setsockopt = sock_common_setsockopt, /* ok */
702 .getsockopt = sock_common_getsockopt, /* ok */
703 .sendmsg = inet6_sendmsg, /* retpoline's sake */
704 .recvmsg = inet6_recvmsg, /* retpoline's sake */
705 #ifdef CONFIG_MMU
706 .mmap = tcp_mmap,
707 #endif
708 .splice_eof = inet_splice_eof,
709 .sendmsg_locked = tcp_sendmsg_locked,
710 .splice_read = tcp_splice_read,
711 .set_peek_off = sk_set_peek_off,
712 .read_sock = tcp_read_sock,
713 .read_skb = tcp_read_skb,
714 .peek_len = tcp_peek_len,
715 #ifdef CONFIG_COMPAT
716 .compat_ioctl = inet6_compat_ioctl,
717 #endif
718 .set_rcvlowat = tcp_set_rcvlowat,
721 const struct proto_ops inet6_dgram_ops = {
722 .family = PF_INET6,
723 .owner = THIS_MODULE,
724 .release = inet6_release,
725 .bind = inet6_bind,
726 .connect = inet_dgram_connect, /* ok */
727 .socketpair = sock_no_socketpair, /* a do nothing */
728 .accept = sock_no_accept, /* a do nothing */
729 .getname = inet6_getname,
730 .poll = udp_poll, /* ok */
731 .ioctl = inet6_ioctl, /* must change */
732 .gettstamp = sock_gettstamp,
733 .listen = sock_no_listen, /* ok */
734 .shutdown = inet_shutdown, /* ok */
735 .setsockopt = sock_common_setsockopt, /* ok */
736 .getsockopt = sock_common_getsockopt, /* ok */
737 .sendmsg = inet6_sendmsg, /* retpoline's sake */
738 .recvmsg = inet6_recvmsg, /* retpoline's sake */
739 .read_skb = udp_read_skb,
740 .mmap = sock_no_mmap,
741 .set_peek_off = udp_set_peek_off,
742 #ifdef CONFIG_COMPAT
743 .compat_ioctl = inet6_compat_ioctl,
744 #endif
747 static const struct net_proto_family inet6_family_ops = {
748 .family = PF_INET6,
749 .create = inet6_create,
750 .owner = THIS_MODULE,
753 int inet6_register_protosw(struct inet_protosw *p)
755 struct list_head *lh;
756 struct inet_protosw *answer;
757 struct list_head *last_perm;
758 int protocol = p->protocol;
759 int ret;
761 spin_lock_bh(&inetsw6_lock);
763 ret = -EINVAL;
764 if (p->type >= SOCK_MAX)
765 goto out_illegal;
767 /* If we are trying to override a permanent protocol, bail. */
768 answer = NULL;
769 ret = -EPERM;
770 last_perm = &inetsw6[p->type];
771 list_for_each(lh, &inetsw6[p->type]) {
772 answer = list_entry(lh, struct inet_protosw, list);
774 /* Check only the non-wild match. */
775 if (INET_PROTOSW_PERMANENT & answer->flags) {
776 if (protocol == answer->protocol)
777 break;
778 last_perm = lh;
781 answer = NULL;
783 if (answer)
784 goto out_permanent;
786 /* Add the new entry after the last permanent entry if any, so that
787 * the new entry does not override a permanent entry when matched with
788 * a wild-card protocol. But it is allowed to override any existing
789 * non-permanent entry. This means that when we remove this entry, the
790 * system automatically returns to the old behavior.
792 list_add_rcu(&p->list, last_perm);
793 ret = 0;
794 out:
795 spin_unlock_bh(&inetsw6_lock);
796 return ret;
798 out_permanent:
799 pr_err("Attempt to override permanent protocol %d\n", protocol);
800 goto out;
802 out_illegal:
803 pr_err("Ignoring attempt to register invalid socket type %d\n",
804 p->type);
805 goto out;
807 EXPORT_SYMBOL(inet6_register_protosw);
809 void
810 inet6_unregister_protosw(struct inet_protosw *p)
812 if (INET_PROTOSW_PERMANENT & p->flags) {
813 pr_err("Attempt to unregister permanent protocol %d\n",
814 p->protocol);
815 } else {
816 spin_lock_bh(&inetsw6_lock);
817 list_del_rcu(&p->list);
818 spin_unlock_bh(&inetsw6_lock);
820 synchronize_net();
823 EXPORT_SYMBOL(inet6_unregister_protosw);
825 int inet6_sk_rebuild_header(struct sock *sk)
827 struct ipv6_pinfo *np = inet6_sk(sk);
828 struct dst_entry *dst;
830 dst = __sk_dst_check(sk, np->dst_cookie);
832 if (!dst) {
833 struct inet_sock *inet = inet_sk(sk);
834 struct in6_addr *final_p, final;
835 struct flowi6 fl6;
837 memset(&fl6, 0, sizeof(fl6));
838 fl6.flowi6_proto = sk->sk_protocol;
839 fl6.daddr = sk->sk_v6_daddr;
840 fl6.saddr = np->saddr;
841 fl6.flowlabel = np->flow_label;
842 fl6.flowi6_oif = sk->sk_bound_dev_if;
843 fl6.flowi6_mark = sk->sk_mark;
844 fl6.fl6_dport = inet->inet_dport;
845 fl6.fl6_sport = inet->inet_sport;
846 fl6.flowi6_uid = sk->sk_uid;
847 security_sk_classify_flow(sk, flowi6_to_flowi_common(&fl6));
849 rcu_read_lock();
850 final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt),
851 &final);
852 rcu_read_unlock();
854 dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p);
855 if (IS_ERR(dst)) {
856 sk->sk_route_caps = 0;
857 WRITE_ONCE(sk->sk_err_soft, -PTR_ERR(dst));
858 return PTR_ERR(dst);
861 ip6_dst_store(sk, dst, NULL, NULL);
864 return 0;
866 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
868 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb,
869 const struct inet6_skb_parm *opt)
871 const struct ipv6_pinfo *np = inet6_sk(sk);
873 if (np->rxopt.all) {
874 if (((opt->flags & IP6SKB_HOPBYHOP) &&
875 (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) ||
876 (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) &&
877 np->rxopt.bits.rxflow) ||
878 (opt->srcrt && (np->rxopt.bits.srcrt ||
879 np->rxopt.bits.osrcrt)) ||
880 ((opt->dst1 || opt->dst0) &&
881 (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
882 return true;
884 return false;
886 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
888 static struct packet_type ipv6_packet_type __read_mostly = {
889 .type = cpu_to_be16(ETH_P_IPV6),
890 .func = ipv6_rcv,
891 .list_func = ipv6_list_rcv,
894 static int __init ipv6_packet_init(void)
896 dev_add_pack(&ipv6_packet_type);
897 return 0;
900 static void ipv6_packet_cleanup(void)
902 dev_remove_pack(&ipv6_packet_type);
905 static int __net_init ipv6_init_mibs(struct net *net)
907 int i;
909 net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib);
910 if (!net->mib.udp_stats_in6)
911 return -ENOMEM;
912 net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib);
913 if (!net->mib.udplite_stats_in6)
914 goto err_udplite_mib;
915 net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib);
916 if (!net->mib.ipv6_statistics)
917 goto err_ip_mib;
919 for_each_possible_cpu(i) {
920 struct ipstats_mib *af_inet6_stats;
921 af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i);
922 u64_stats_init(&af_inet6_stats->syncp);
926 net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib);
927 if (!net->mib.icmpv6_statistics)
928 goto err_icmp_mib;
929 net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
930 GFP_KERNEL);
931 if (!net->mib.icmpv6msg_statistics)
932 goto err_icmpmsg_mib;
933 return 0;
935 err_icmpmsg_mib:
936 free_percpu(net->mib.icmpv6_statistics);
937 err_icmp_mib:
938 free_percpu(net->mib.ipv6_statistics);
939 err_ip_mib:
940 free_percpu(net->mib.udplite_stats_in6);
941 err_udplite_mib:
942 free_percpu(net->mib.udp_stats_in6);
943 return -ENOMEM;
946 static void ipv6_cleanup_mibs(struct net *net)
948 free_percpu(net->mib.udp_stats_in6);
949 free_percpu(net->mib.udplite_stats_in6);
950 free_percpu(net->mib.ipv6_statistics);
951 free_percpu(net->mib.icmpv6_statistics);
952 kfree(net->mib.icmpv6msg_statistics);
955 static int __net_init inet6_net_init(struct net *net)
957 int err = 0;
959 net->ipv6.sysctl.bindv6only = 0;
960 net->ipv6.sysctl.icmpv6_time = 1*HZ;
961 net->ipv6.sysctl.icmpv6_echo_ignore_all = 0;
962 net->ipv6.sysctl.icmpv6_echo_ignore_multicast = 0;
963 net->ipv6.sysctl.icmpv6_echo_ignore_anycast = 0;
964 net->ipv6.sysctl.icmpv6_error_anycast_as_unicast = 0;
966 /* By default, rate limit error messages.
967 * Except for pmtu discovery, it would break it.
968 * proc_do_large_bitmap needs pointer to the bitmap.
970 bitmap_set(net->ipv6.sysctl.icmpv6_ratemask, 0, ICMPV6_ERRMSG_MAX + 1);
971 bitmap_clear(net->ipv6.sysctl.icmpv6_ratemask, ICMPV6_PKT_TOOBIG, 1);
972 net->ipv6.sysctl.icmpv6_ratemask_ptr = net->ipv6.sysctl.icmpv6_ratemask;
974 net->ipv6.sysctl.flowlabel_consistency = 1;
975 net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS;
976 net->ipv6.sysctl.idgen_retries = 3;
977 net->ipv6.sysctl.idgen_delay = 1 * HZ;
978 net->ipv6.sysctl.flowlabel_state_ranges = 0;
979 net->ipv6.sysctl.max_dst_opts_cnt = IP6_DEFAULT_MAX_DST_OPTS_CNT;
980 net->ipv6.sysctl.max_hbh_opts_cnt = IP6_DEFAULT_MAX_HBH_OPTS_CNT;
981 net->ipv6.sysctl.max_dst_opts_len = IP6_DEFAULT_MAX_DST_OPTS_LEN;
982 net->ipv6.sysctl.max_hbh_opts_len = IP6_DEFAULT_MAX_HBH_OPTS_LEN;
983 net->ipv6.sysctl.fib_notify_on_flag_change = 0;
984 atomic_set(&net->ipv6.fib6_sernum, 1);
986 net->ipv6.sysctl.ioam6_id = IOAM6_DEFAULT_ID;
987 net->ipv6.sysctl.ioam6_id_wide = IOAM6_DEFAULT_ID_WIDE;
989 err = ipv6_init_mibs(net);
990 if (err)
991 return err;
992 #ifdef CONFIG_PROC_FS
993 err = udp6_proc_init(net);
994 if (err)
995 goto out;
996 err = tcp6_proc_init(net);
997 if (err)
998 goto proc_tcp6_fail;
999 err = ac6_proc_init(net);
1000 if (err)
1001 goto proc_ac6_fail;
1002 #endif
1003 return err;
1005 #ifdef CONFIG_PROC_FS
1006 proc_ac6_fail:
1007 tcp6_proc_exit(net);
1008 proc_tcp6_fail:
1009 udp6_proc_exit(net);
1010 out:
1011 ipv6_cleanup_mibs(net);
1012 return err;
1013 #endif
1016 static void __net_exit inet6_net_exit(struct net *net)
1018 #ifdef CONFIG_PROC_FS
1019 udp6_proc_exit(net);
1020 tcp6_proc_exit(net);
1021 ac6_proc_exit(net);
1022 #endif
1023 ipv6_cleanup_mibs(net);
1026 static struct pernet_operations inet6_net_ops = {
1027 .init = inet6_net_init,
1028 .exit = inet6_net_exit,
1031 static int ipv6_route_input(struct sk_buff *skb)
1033 ip6_route_input(skb);
1034 return skb_dst(skb)->error;
1037 static const struct ipv6_stub ipv6_stub_impl = {
1038 .ipv6_sock_mc_join = ipv6_sock_mc_join,
1039 .ipv6_sock_mc_drop = ipv6_sock_mc_drop,
1040 .ipv6_dst_lookup_flow = ip6_dst_lookup_flow,
1041 .ipv6_route_input = ipv6_route_input,
1042 .fib6_get_table = fib6_get_table,
1043 .fib6_table_lookup = fib6_table_lookup,
1044 .fib6_lookup = fib6_lookup,
1045 .fib6_select_path = fib6_select_path,
1046 .ip6_mtu_from_fib6 = ip6_mtu_from_fib6,
1047 .fib6_nh_init = fib6_nh_init,
1048 .fib6_nh_release = fib6_nh_release,
1049 .fib6_nh_release_dsts = fib6_nh_release_dsts,
1050 .fib6_update_sernum = fib6_update_sernum_stub,
1051 .fib6_rt_update = fib6_rt_update,
1052 .ip6_del_rt = ip6_del_rt,
1053 .udpv6_encap_enable = udpv6_encap_enable,
1054 .ndisc_send_na = ndisc_send_na,
1055 #if IS_ENABLED(CONFIG_XFRM)
1056 .xfrm6_local_rxpmtu = xfrm6_local_rxpmtu,
1057 .xfrm6_udp_encap_rcv = xfrm6_udp_encap_rcv,
1058 .xfrm6_gro_udp_encap_rcv = xfrm6_gro_udp_encap_rcv,
1059 .xfrm6_rcv_encap = xfrm6_rcv_encap,
1060 #endif
1061 .nd_tbl = &nd_tbl,
1062 .ipv6_fragment = ip6_fragment,
1063 .ipv6_dev_find = ipv6_dev_find,
1064 .ip6_xmit = ip6_xmit,
1067 static const struct ipv6_bpf_stub ipv6_bpf_stub_impl = {
1068 .inet6_bind = __inet6_bind,
1069 .udp6_lib_lookup = __udp6_lib_lookup,
1070 .ipv6_setsockopt = do_ipv6_setsockopt,
1071 .ipv6_getsockopt = do_ipv6_getsockopt,
1072 .ipv6_dev_get_saddr = ipv6_dev_get_saddr,
1075 static int __init inet6_init(void)
1077 struct list_head *r;
1078 int err = 0;
1080 sock_skb_cb_check_size(sizeof(struct inet6_skb_parm));
1082 /* Register the socket-side information for inet6_create. */
1083 for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
1084 INIT_LIST_HEAD(r);
1086 raw_hashinfo_init(&raw_v6_hashinfo);
1088 if (disable_ipv6_mod) {
1089 pr_info("Loaded, but administratively disabled, reboot required to enable\n");
1090 goto out;
1093 err = proto_register(&tcpv6_prot, 1);
1094 if (err)
1095 goto out;
1097 err = proto_register(&udpv6_prot, 1);
1098 if (err)
1099 goto out_unregister_tcp_proto;
1101 err = proto_register(&udplitev6_prot, 1);
1102 if (err)
1103 goto out_unregister_udp_proto;
1105 err = proto_register(&rawv6_prot, 1);
1106 if (err)
1107 goto out_unregister_udplite_proto;
1109 err = proto_register(&pingv6_prot, 1);
1110 if (err)
1111 goto out_unregister_raw_proto;
1113 /* We MUST register RAW sockets before we create the ICMP6,
1114 * IGMP6, or NDISC control sockets.
1116 err = rawv6_init();
1117 if (err)
1118 goto out_unregister_ping_proto;
1120 /* Register the family here so that the init calls below will
1121 * be able to create sockets. (?? is this dangerous ??)
1123 err = sock_register(&inet6_family_ops);
1124 if (err)
1125 goto out_sock_register_fail;
1128 * ipngwg API draft makes clear that the correct semantics
1129 * for TCP and UDP is to consider one TCP and UDP instance
1130 * in a host available by both INET and INET6 APIs and
1131 * able to communicate via both network protocols.
1134 err = register_pernet_subsys(&inet6_net_ops);
1135 if (err)
1136 goto register_pernet_fail;
1137 err = ip6_mr_init();
1138 if (err)
1139 goto ipmr_fail;
1140 err = icmpv6_init();
1141 if (err)
1142 goto icmp_fail;
1143 err = ndisc_init();
1144 if (err)
1145 goto ndisc_fail;
1146 err = igmp6_init();
1147 if (err)
1148 goto igmp_fail;
1150 err = ipv6_netfilter_init();
1151 if (err)
1152 goto netfilter_fail;
1153 /* Create /proc/foo6 entries. */
1154 #ifdef CONFIG_PROC_FS
1155 err = -ENOMEM;
1156 if (raw6_proc_init())
1157 goto proc_raw6_fail;
1158 if (udplite6_proc_init())
1159 goto proc_udplite6_fail;
1160 if (ipv6_misc_proc_init())
1161 goto proc_misc6_fail;
1162 if (if6_proc_init())
1163 goto proc_if6_fail;
1164 #endif
1165 err = ip6_route_init();
1166 if (err)
1167 goto ip6_route_fail;
1168 err = ndisc_late_init();
1169 if (err)
1170 goto ndisc_late_fail;
1171 err = ip6_flowlabel_init();
1172 if (err)
1173 goto ip6_flowlabel_fail;
1174 err = ipv6_anycast_init();
1175 if (err)
1176 goto ipv6_anycast_fail;
1177 err = addrconf_init();
1178 if (err)
1179 goto addrconf_fail;
1181 /* Init v6 extension headers. */
1182 err = ipv6_exthdrs_init();
1183 if (err)
1184 goto ipv6_exthdrs_fail;
1186 err = ipv6_frag_init();
1187 if (err)
1188 goto ipv6_frag_fail;
1190 /* Init v6 transport protocols. */
1191 err = udpv6_init();
1192 if (err)
1193 goto udpv6_fail;
1195 err = udplitev6_init();
1196 if (err)
1197 goto udplitev6_fail;
1199 err = udpv6_offload_init();
1200 if (err)
1201 goto udpv6_offload_fail;
1203 err = tcpv6_init();
1204 if (err)
1205 goto tcpv6_fail;
1207 err = ipv6_packet_init();
1208 if (err)
1209 goto ipv6_packet_fail;
1211 err = pingv6_init();
1212 if (err)
1213 goto pingv6_fail;
1215 err = calipso_init();
1216 if (err)
1217 goto calipso_fail;
1219 err = seg6_init();
1220 if (err)
1221 goto seg6_fail;
1223 err = rpl_init();
1224 if (err)
1225 goto rpl_fail;
1227 err = ioam6_init();
1228 if (err)
1229 goto ioam6_fail;
1231 err = igmp6_late_init();
1232 if (err)
1233 goto igmp6_late_err;
1235 #ifdef CONFIG_SYSCTL
1236 err = ipv6_sysctl_register();
1237 if (err)
1238 goto sysctl_fail;
1239 #endif
1241 /* ensure that ipv6 stubs are visible only after ipv6 is ready */
1242 wmb();
1243 ipv6_stub = &ipv6_stub_impl;
1244 ipv6_bpf_stub = &ipv6_bpf_stub_impl;
1245 out:
1246 return err;
1248 #ifdef CONFIG_SYSCTL
1249 sysctl_fail:
1250 igmp6_late_cleanup();
1251 #endif
1252 igmp6_late_err:
1253 ioam6_exit();
1254 ioam6_fail:
1255 rpl_exit();
1256 rpl_fail:
1257 seg6_exit();
1258 seg6_fail:
1259 calipso_exit();
1260 calipso_fail:
1261 pingv6_exit();
1262 pingv6_fail:
1263 ipv6_packet_cleanup();
1264 ipv6_packet_fail:
1265 tcpv6_exit();
1266 tcpv6_fail:
1267 udpv6_offload_exit();
1268 udpv6_offload_fail:
1269 udplitev6_exit();
1270 udplitev6_fail:
1271 udpv6_exit();
1272 udpv6_fail:
1273 ipv6_frag_exit();
1274 ipv6_frag_fail:
1275 ipv6_exthdrs_exit();
1276 ipv6_exthdrs_fail:
1277 addrconf_cleanup();
1278 addrconf_fail:
1279 ipv6_anycast_cleanup();
1280 ipv6_anycast_fail:
1281 ip6_flowlabel_cleanup();
1282 ip6_flowlabel_fail:
1283 ndisc_late_cleanup();
1284 ndisc_late_fail:
1285 ip6_route_cleanup();
1286 ip6_route_fail:
1287 #ifdef CONFIG_PROC_FS
1288 if6_proc_exit();
1289 proc_if6_fail:
1290 ipv6_misc_proc_exit();
1291 proc_misc6_fail:
1292 udplite6_proc_exit();
1293 proc_udplite6_fail:
1294 raw6_proc_exit();
1295 proc_raw6_fail:
1296 #endif
1297 ipv6_netfilter_fini();
1298 netfilter_fail:
1299 igmp6_cleanup();
1300 igmp_fail:
1301 ndisc_cleanup();
1302 ndisc_fail:
1303 icmpv6_cleanup();
1304 icmp_fail:
1305 ip6_mr_cleanup();
1306 ipmr_fail:
1307 unregister_pernet_subsys(&inet6_net_ops);
1308 register_pernet_fail:
1309 sock_unregister(PF_INET6);
1310 rtnl_unregister_all(PF_INET6);
1311 out_sock_register_fail:
1312 rawv6_exit();
1313 out_unregister_ping_proto:
1314 proto_unregister(&pingv6_prot);
1315 out_unregister_raw_proto:
1316 proto_unregister(&rawv6_prot);
1317 out_unregister_udplite_proto:
1318 proto_unregister(&udplitev6_prot);
1319 out_unregister_udp_proto:
1320 proto_unregister(&udpv6_prot);
1321 out_unregister_tcp_proto:
1322 proto_unregister(&tcpv6_prot);
1323 goto out;
1325 module_init(inet6_init);
1327 MODULE_ALIAS_NETPROTO(PF_INET6);