Loading...
1// SPDX-License-Identifier: GPL-2.0-or-later
2/*
3 * PF_INET6 socket protocol family
4 * Linux INET6 implementation
5 *
6 * Authors:
7 * Pedro Roque <roque@di.fc.ul.pt>
8 *
9 * Adapted from linux/net/ipv4/af_inet.c
10 *
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
15 */
16
17#define pr_fmt(fmt) "IPv6: " fmt
18
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>
37
38#include <linux/inet.h>
39#include <linux/netdevice.h>
40#include <linux/icmpv6.h>
41#include <linux/netfilter_ipv6.h>
42
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
68#include <linux/uaccess.h>
69#include <linux/mroute6.h>
70
71#include "ip6_offload.h"
72
73MODULE_AUTHOR("Cast of dozens");
74MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
75MODULE_LICENSE("GPL");
76
77/* The inetsw6 table contains everything that inet6_create needs to
78 * build a new socket.
79 */
80static struct list_head inetsw6[SOCK_MAX];
81static DEFINE_SPINLOCK(inetsw6_lock);
82
83struct ipv6_params ipv6_defaults = {
84 .disable_ipv6 = 0,
85 .autoconf = 1,
86};
87
88static int disable_ipv6_mod;
89
90module_param_named(disable, disable_ipv6_mod, int, 0444);
91MODULE_PARM_DESC(disable, "Disable IPv6 module such that it is non-functional");
92
93module_param_named(disable_ipv6, ipv6_defaults.disable_ipv6, int, 0444);
94MODULE_PARM_DESC(disable_ipv6, "Disable IPv6 on all interfaces");
95
96module_param_named(autoconf, ipv6_defaults.autoconf, int, 0444);
97MODULE_PARM_DESC(autoconf, "Enable IPv6 address autoconfiguration on all interfaces");
98
99bool ipv6_mod_enabled(void)
100{
101 return disable_ipv6_mod == 0;
102}
103EXPORT_SYMBOL_GPL(ipv6_mod_enabled);
104
105static struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
106{
107 const int offset = sk->sk_prot->ipv6_pinfo_offset;
108
109 return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
110}
111
112void inet6_sock_destruct(struct sock *sk)
113{
114 inet6_cleanup_sock(sk);
115 inet_sock_destruct(sk);
116}
117EXPORT_SYMBOL_GPL(inet6_sock_destruct);
118
119static int inet6_create(struct net *net, struct socket *sock, int protocol,
120 int kern)
121{
122 struct inet_sock *inet;
123 struct ipv6_pinfo *np;
124 struct sock *sk;
125 struct inet_protosw *answer;
126 struct proto *answer_prot;
127 unsigned char answer_flags;
128 int try_loading_module = 0;
129 int err;
130
131 if (protocol < 0 || protocol >= IPPROTO_MAX)
132 return -EINVAL;
133
134 /* Look for the requested type/protocol pair. */
135lookup_protocol:
136 err = -ESOCKTNOSUPPORT;
137 rcu_read_lock();
138 list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
139
140 err = 0;
141 /* Check the non-wild match. */
142 if (protocol == answer->protocol) {
143 if (protocol != IPPROTO_IP)
144 break;
145 } else {
146 /* Check for the two wild cases. */
147 if (IPPROTO_IP == protocol) {
148 protocol = answer->protocol;
149 break;
150 }
151 if (IPPROTO_IP == answer->protocol)
152 break;
153 }
154 err = -EPROTONOSUPPORT;
155 }
156
157 if (err) {
158 if (try_loading_module < 2) {
159 rcu_read_unlock();
160 /*
161 * Be more specific, e.g. net-pf-10-proto-132-type-1
162 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
163 */
164 if (++try_loading_module == 1)
165 request_module("net-pf-%d-proto-%d-type-%d",
166 PF_INET6, protocol, sock->type);
167 /*
168 * Fall back to generic, e.g. net-pf-10-proto-132
169 * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
170 */
171 else
172 request_module("net-pf-%d-proto-%d",
173 PF_INET6, protocol);
174 goto lookup_protocol;
175 } else
176 goto out_rcu_unlock;
177 }
178
179 err = -EPERM;
180 if (sock->type == SOCK_RAW && !kern &&
181 !ns_capable(net->user_ns, CAP_NET_RAW))
182 goto out_rcu_unlock;
183
184 sock->ops = answer->ops;
185 answer_prot = answer->prot;
186 answer_flags = answer->flags;
187 rcu_read_unlock();
188
189 WARN_ON(!answer_prot->slab);
190
191 err = -ENOBUFS;
192 sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot, kern);
193 if (!sk)
194 goto out;
195
196 sock_init_data(sock, sk);
197
198 err = 0;
199 if (INET_PROTOSW_REUSE & answer_flags)
200 sk->sk_reuse = SK_CAN_REUSE;
201
202 if (INET_PROTOSW_ICSK & answer_flags)
203 inet_init_csk_locks(sk);
204
205 inet = inet_sk(sk);
206 inet_assign_bit(IS_ICSK, sk, INET_PROTOSW_ICSK & answer_flags);
207
208 if (SOCK_RAW == sock->type) {
209 inet->inet_num = protocol;
210 if (IPPROTO_RAW == protocol)
211 inet_set_bit(HDRINCL, sk);
212 }
213
214 sk->sk_destruct = inet6_sock_destruct;
215 sk->sk_family = PF_INET6;
216 sk->sk_protocol = protocol;
217
218 sk->sk_backlog_rcv = answer->prot->backlog_rcv;
219
220 inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
221 np->hop_limit = -1;
222 np->mcast_hops = IPV6_DEFAULT_MCASTHOPS;
223 inet6_set_bit(MC6_LOOP, sk);
224 inet6_set_bit(MC6_ALL, sk);
225 np->pmtudisc = IPV6_PMTUDISC_WANT;
226 inet6_assign_bit(REPFLOW, sk, net->ipv6.sysctl.flowlabel_reflect &
227 FLOWLABEL_REFLECT_ESTABLISHED);
228 sk->sk_ipv6only = net->ipv6.sysctl.bindv6only;
229 sk->sk_txrehash = READ_ONCE(net->core.sysctl_txrehash);
230
231 /* Init the ipv4 part of the socket since we can have sockets
232 * using v6 API for ipv4.
233 */
234 inet->uc_ttl = -1;
235
236 inet_set_bit(MC_LOOP, sk);
237 inet->mc_ttl = 1;
238 inet->mc_index = 0;
239 RCU_INIT_POINTER(inet->mc_list, NULL);
240 inet->rcv_tos = 0;
241
242 if (READ_ONCE(net->ipv4.sysctl_ip_no_pmtu_disc))
243 inet->pmtudisc = IP_PMTUDISC_DONT;
244 else
245 inet->pmtudisc = IP_PMTUDISC_WANT;
246
247 if (inet->inet_num) {
248 /* It assumes that any protocol which allows
249 * the user to assign a number at socket
250 * creation time automatically shares.
251 */
252 inet->inet_sport = htons(inet->inet_num);
253 err = sk->sk_prot->hash(sk);
254 if (err) {
255 sk_common_release(sk);
256 goto out;
257 }
258 }
259 if (sk->sk_prot->init) {
260 err = sk->sk_prot->init(sk);
261 if (err) {
262 sk_common_release(sk);
263 goto out;
264 }
265 }
266
267 if (!kern) {
268 err = BPF_CGROUP_RUN_PROG_INET_SOCK(sk);
269 if (err) {
270 sk_common_release(sk);
271 goto out;
272 }
273 }
274out:
275 return err;
276out_rcu_unlock:
277 rcu_read_unlock();
278 goto out;
279}
280
281static int __inet6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len,
282 u32 flags)
283{
284 struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr;
285 struct inet_sock *inet = inet_sk(sk);
286 struct ipv6_pinfo *np = inet6_sk(sk);
287 struct net *net = sock_net(sk);
288 __be32 v4addr = 0;
289 unsigned short snum;
290 bool saved_ipv6only;
291 int addr_type = 0;
292 int err = 0;
293
294 if (addr->sin6_family != AF_INET6)
295 return -EAFNOSUPPORT;
296
297 addr_type = ipv6_addr_type(&addr->sin6_addr);
298 if ((addr_type & IPV6_ADDR_MULTICAST) && sk->sk_type == SOCK_STREAM)
299 return -EINVAL;
300
301 snum = ntohs(addr->sin6_port);
302 if (!(flags & BIND_NO_CAP_NET_BIND_SERVICE) &&
303 snum && inet_port_requires_bind_service(net, snum) &&
304 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
305 return -EACCES;
306
307 if (flags & BIND_WITH_LOCK)
308 lock_sock(sk);
309
310 /* Check these errors (active socket, double bind). */
311 if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
312 err = -EINVAL;
313 goto out;
314 }
315
316 /* Check if the address belongs to the host. */
317 if (addr_type == IPV6_ADDR_MAPPED) {
318 struct net_device *dev = NULL;
319 int chk_addr_ret;
320
321 /* Binding to v4-mapped address on a v6-only socket
322 * makes no sense
323 */
324 if (ipv6_only_sock(sk)) {
325 err = -EINVAL;
326 goto out;
327 }
328
329 rcu_read_lock();
330 if (sk->sk_bound_dev_if) {
331 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
332 if (!dev) {
333 err = -ENODEV;
334 goto out_unlock;
335 }
336 }
337
338 /* Reproduce AF_INET checks to make the bindings consistent */
339 v4addr = addr->sin6_addr.s6_addr32[3];
340 chk_addr_ret = inet_addr_type_dev_table(net, dev, v4addr);
341 rcu_read_unlock();
342
343 if (!inet_addr_valid_or_nonlocal(net, inet, v4addr,
344 chk_addr_ret)) {
345 err = -EADDRNOTAVAIL;
346 goto out;
347 }
348 } else {
349 if (addr_type != IPV6_ADDR_ANY) {
350 struct net_device *dev = NULL;
351
352 rcu_read_lock();
353 if (__ipv6_addr_needs_scope_id(addr_type)) {
354 if (addr_len >= sizeof(struct sockaddr_in6) &&
355 addr->sin6_scope_id) {
356 /* Override any existing binding, if another one
357 * is supplied by user.
358 */
359 sk->sk_bound_dev_if = addr->sin6_scope_id;
360 }
361
362 /* Binding to link-local address requires an interface */
363 if (!sk->sk_bound_dev_if) {
364 err = -EINVAL;
365 goto out_unlock;
366 }
367 }
368
369 if (sk->sk_bound_dev_if) {
370 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
371 if (!dev) {
372 err = -ENODEV;
373 goto out_unlock;
374 }
375 }
376
377 /* ipv4 addr of the socket is invalid. Only the
378 * unspecified and mapped address have a v4 equivalent.
379 */
380 v4addr = LOOPBACK4_IPV6;
381 if (!(addr_type & IPV6_ADDR_MULTICAST)) {
382 if (!ipv6_can_nonlocal_bind(net, inet) &&
383 !ipv6_chk_addr(net, &addr->sin6_addr,
384 dev, 0)) {
385 err = -EADDRNOTAVAIL;
386 goto out_unlock;
387 }
388 }
389 rcu_read_unlock();
390 }
391 }
392
393 inet->inet_rcv_saddr = v4addr;
394 inet->inet_saddr = v4addr;
395
396 sk->sk_v6_rcv_saddr = addr->sin6_addr;
397
398 if (!(addr_type & IPV6_ADDR_MULTICAST))
399 np->saddr = addr->sin6_addr;
400
401 saved_ipv6only = sk->sk_ipv6only;
402 if (addr_type != IPV6_ADDR_ANY && addr_type != IPV6_ADDR_MAPPED)
403 sk->sk_ipv6only = 1;
404
405 /* Make sure we are allowed to bind here. */
406 if (snum || !(inet_test_bit(BIND_ADDRESS_NO_PORT, sk) ||
407 (flags & BIND_FORCE_ADDRESS_NO_PORT))) {
408 err = sk->sk_prot->get_port(sk, snum);
409 if (err) {
410 sk->sk_ipv6only = saved_ipv6only;
411 inet_reset_saddr(sk);
412 goto out;
413 }
414 if (!(flags & BIND_FROM_BPF)) {
415 err = BPF_CGROUP_RUN_PROG_INET6_POST_BIND(sk);
416 if (err) {
417 sk->sk_ipv6only = saved_ipv6only;
418 inet_reset_saddr(sk);
419 if (sk->sk_prot->put_port)
420 sk->sk_prot->put_port(sk);
421 goto out;
422 }
423 }
424 }
425
426 if (addr_type != IPV6_ADDR_ANY)
427 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
428 if (snum)
429 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
430 inet->inet_sport = htons(inet->inet_num);
431 inet->inet_dport = 0;
432 inet->inet_daddr = 0;
433out:
434 if (flags & BIND_WITH_LOCK)
435 release_sock(sk);
436 return err;
437out_unlock:
438 rcu_read_unlock();
439 goto out;
440}
441
442int inet6_bind_sk(struct sock *sk, struct sockaddr *uaddr, int addr_len)
443{
444 u32 flags = BIND_WITH_LOCK;
445 const struct proto *prot;
446 int err = 0;
447
448 /* IPV6_ADDRFORM can change sk->sk_prot under us. */
449 prot = READ_ONCE(sk->sk_prot);
450 /* If the socket has its own bind function then use it. */
451 if (prot->bind)
452 return prot->bind(sk, uaddr, addr_len);
453
454 if (addr_len < SIN6_LEN_RFC2133)
455 return -EINVAL;
456
457 /* BPF prog is run before any checks are done so that if the prog
458 * changes context in a wrong way it will be caught.
459 */
460 err = BPF_CGROUP_RUN_PROG_INET_BIND_LOCK(sk, uaddr, &addr_len,
461 CGROUP_INET6_BIND, &flags);
462 if (err)
463 return err;
464
465 return __inet6_bind(sk, uaddr, addr_len, flags);
466}
467
468/* bind for INET6 API */
469int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
470{
471 return inet6_bind_sk(sock->sk, uaddr, addr_len);
472}
473EXPORT_SYMBOL(inet6_bind);
474
475int inet6_release(struct socket *sock)
476{
477 struct sock *sk = sock->sk;
478
479 if (!sk)
480 return -EINVAL;
481
482 /* Free mc lists */
483 ipv6_sock_mc_close(sk);
484
485 /* Free ac lists */
486 ipv6_sock_ac_close(sk);
487
488 return inet_release(sock);
489}
490EXPORT_SYMBOL(inet6_release);
491
492void inet6_cleanup_sock(struct sock *sk)
493{
494 struct ipv6_pinfo *np = inet6_sk(sk);
495 struct sk_buff *skb;
496 struct ipv6_txoptions *opt;
497
498 /* Release rx options */
499
500 skb = xchg(&np->pktoptions, NULL);
501 kfree_skb(skb);
502
503 skb = xchg(&np->rxpmtu, NULL);
504 kfree_skb(skb);
505
506 /* Free flowlabels */
507 fl6_free_socklist(sk);
508
509 /* Free tx options */
510
511 opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL);
512 if (opt) {
513 atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
514 txopt_put(opt);
515 }
516}
517EXPORT_SYMBOL_GPL(inet6_cleanup_sock);
518
519/*
520 * This does both peername and sockname.
521 */
522int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
523 int peer)
524{
525 struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
526 int sin_addr_len = sizeof(*sin);
527 struct sock *sk = sock->sk;
528 struct inet_sock *inet = inet_sk(sk);
529 struct ipv6_pinfo *np = inet6_sk(sk);
530
531 sin->sin6_family = AF_INET6;
532 sin->sin6_flowinfo = 0;
533 sin->sin6_scope_id = 0;
534 lock_sock(sk);
535 if (peer) {
536 if (!inet->inet_dport ||
537 (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
538 peer == 1)) {
539 release_sock(sk);
540 return -ENOTCONN;
541 }
542 sin->sin6_port = inet->inet_dport;
543 sin->sin6_addr = sk->sk_v6_daddr;
544 if (inet6_test_bit(SNDFLOW, sk))
545 sin->sin6_flowinfo = np->flow_label;
546 BPF_CGROUP_RUN_SA_PROG(sk, (struct sockaddr *)sin, &sin_addr_len,
547 CGROUP_INET6_GETPEERNAME);
548 } else {
549 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
550 sin->sin6_addr = np->saddr;
551 else
552 sin->sin6_addr = sk->sk_v6_rcv_saddr;
553 sin->sin6_port = inet->inet_sport;
554 BPF_CGROUP_RUN_SA_PROG(sk, (struct sockaddr *)sin, &sin_addr_len,
555 CGROUP_INET6_GETSOCKNAME);
556 }
557 sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
558 sk->sk_bound_dev_if);
559 release_sock(sk);
560 return sin_addr_len;
561}
562EXPORT_SYMBOL(inet6_getname);
563
564int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
565{
566 void __user *argp = (void __user *)arg;
567 struct sock *sk = sock->sk;
568 struct net *net = sock_net(sk);
569 const struct proto *prot;
570
571 switch (cmd) {
572 case SIOCADDRT:
573 case SIOCDELRT: {
574 struct in6_rtmsg rtmsg;
575
576 if (copy_from_user(&rtmsg, argp, sizeof(rtmsg)))
577 return -EFAULT;
578 return ipv6_route_ioctl(net, cmd, &rtmsg);
579 }
580 case SIOCSIFADDR:
581 return addrconf_add_ifaddr(net, argp);
582 case SIOCDIFADDR:
583 return addrconf_del_ifaddr(net, argp);
584 case SIOCSIFDSTADDR:
585 return addrconf_set_dstaddr(net, argp);
586 default:
587 /* IPV6_ADDRFORM can change sk->sk_prot under us. */
588 prot = READ_ONCE(sk->sk_prot);
589 if (!prot->ioctl)
590 return -ENOIOCTLCMD;
591 return sk_ioctl(sk, cmd, (void __user *)arg);
592 }
593 /*NOTREACHED*/
594 return 0;
595}
596EXPORT_SYMBOL(inet6_ioctl);
597
598#ifdef CONFIG_COMPAT
599struct compat_in6_rtmsg {
600 struct in6_addr rtmsg_dst;
601 struct in6_addr rtmsg_src;
602 struct in6_addr rtmsg_gateway;
603 u32 rtmsg_type;
604 u16 rtmsg_dst_len;
605 u16 rtmsg_src_len;
606 u32 rtmsg_metric;
607 u32 rtmsg_info;
608 u32 rtmsg_flags;
609 s32 rtmsg_ifindex;
610};
611
612static int inet6_compat_routing_ioctl(struct sock *sk, unsigned int cmd,
613 struct compat_in6_rtmsg __user *ur)
614{
615 struct in6_rtmsg rt;
616
617 if (copy_from_user(&rt.rtmsg_dst, &ur->rtmsg_dst,
618 3 * sizeof(struct in6_addr)) ||
619 get_user(rt.rtmsg_type, &ur->rtmsg_type) ||
620 get_user(rt.rtmsg_dst_len, &ur->rtmsg_dst_len) ||
621 get_user(rt.rtmsg_src_len, &ur->rtmsg_src_len) ||
622 get_user(rt.rtmsg_metric, &ur->rtmsg_metric) ||
623 get_user(rt.rtmsg_info, &ur->rtmsg_info) ||
624 get_user(rt.rtmsg_flags, &ur->rtmsg_flags) ||
625 get_user(rt.rtmsg_ifindex, &ur->rtmsg_ifindex))
626 return -EFAULT;
627
628
629 return ipv6_route_ioctl(sock_net(sk), cmd, &rt);
630}
631
632int inet6_compat_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
633{
634 void __user *argp = compat_ptr(arg);
635 struct sock *sk = sock->sk;
636
637 switch (cmd) {
638 case SIOCADDRT:
639 case SIOCDELRT:
640 return inet6_compat_routing_ioctl(sk, cmd, argp);
641 default:
642 return -ENOIOCTLCMD;
643 }
644}
645EXPORT_SYMBOL_GPL(inet6_compat_ioctl);
646#endif /* CONFIG_COMPAT */
647
648INDIRECT_CALLABLE_DECLARE(int udpv6_sendmsg(struct sock *, struct msghdr *,
649 size_t));
650int inet6_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
651{
652 struct sock *sk = sock->sk;
653 const struct proto *prot;
654
655 if (unlikely(inet_send_prepare(sk)))
656 return -EAGAIN;
657
658 /* IPV6_ADDRFORM can change sk->sk_prot under us. */
659 prot = READ_ONCE(sk->sk_prot);
660 return INDIRECT_CALL_2(prot->sendmsg, tcp_sendmsg, udpv6_sendmsg,
661 sk, msg, size);
662}
663
664INDIRECT_CALLABLE_DECLARE(int udpv6_recvmsg(struct sock *, struct msghdr *,
665 size_t, int, int *));
666int inet6_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
667 int flags)
668{
669 struct sock *sk = sock->sk;
670 const struct proto *prot;
671 int addr_len = 0;
672 int err;
673
674 if (likely(!(flags & MSG_ERRQUEUE)))
675 sock_rps_record_flow(sk);
676
677 /* IPV6_ADDRFORM can change sk->sk_prot under us. */
678 prot = READ_ONCE(sk->sk_prot);
679 err = INDIRECT_CALL_2(prot->recvmsg, tcp_recvmsg, udpv6_recvmsg,
680 sk, msg, size, flags, &addr_len);
681 if (err >= 0)
682 msg->msg_namelen = addr_len;
683 return err;
684}
685
686const struct proto_ops inet6_stream_ops = {
687 .family = PF_INET6,
688 .owner = THIS_MODULE,
689 .release = inet6_release,
690 .bind = inet6_bind,
691 .connect = inet_stream_connect, /* ok */
692 .socketpair = sock_no_socketpair, /* a do nothing */
693 .accept = inet_accept, /* ok */
694 .getname = inet6_getname,
695 .poll = tcp_poll, /* ok */
696 .ioctl = inet6_ioctl, /* must change */
697 .gettstamp = sock_gettstamp,
698 .listen = inet_listen, /* ok */
699 .shutdown = inet_shutdown, /* ok */
700 .setsockopt = sock_common_setsockopt, /* ok */
701 .getsockopt = sock_common_getsockopt, /* ok */
702 .sendmsg = inet6_sendmsg, /* retpoline's sake */
703 .recvmsg = inet6_recvmsg, /* retpoline's sake */
704#ifdef CONFIG_MMU
705 .mmap = tcp_mmap,
706#endif
707 .splice_eof = inet_splice_eof,
708 .sendmsg_locked = tcp_sendmsg_locked,
709 .splice_read = tcp_splice_read,
710 .read_sock = tcp_read_sock,
711 .read_skb = tcp_read_skb,
712 .peek_len = tcp_peek_len,
713#ifdef CONFIG_COMPAT
714 .compat_ioctl = inet6_compat_ioctl,
715#endif
716 .set_rcvlowat = tcp_set_rcvlowat,
717};
718
719const struct proto_ops inet6_dgram_ops = {
720 .family = PF_INET6,
721 .owner = THIS_MODULE,
722 .release = inet6_release,
723 .bind = inet6_bind,
724 .connect = inet_dgram_connect, /* ok */
725 .socketpair = sock_no_socketpair, /* a do nothing */
726 .accept = sock_no_accept, /* a do nothing */
727 .getname = inet6_getname,
728 .poll = udp_poll, /* ok */
729 .ioctl = inet6_ioctl, /* must change */
730 .gettstamp = sock_gettstamp,
731 .listen = sock_no_listen, /* ok */
732 .shutdown = inet_shutdown, /* ok */
733 .setsockopt = sock_common_setsockopt, /* ok */
734 .getsockopt = sock_common_getsockopt, /* ok */
735 .sendmsg = inet6_sendmsg, /* retpoline's sake */
736 .recvmsg = inet6_recvmsg, /* retpoline's sake */
737 .read_skb = udp_read_skb,
738 .mmap = sock_no_mmap,
739 .set_peek_off = sk_set_peek_off,
740#ifdef CONFIG_COMPAT
741 .compat_ioctl = inet6_compat_ioctl,
742#endif
743};
744
745static const struct net_proto_family inet6_family_ops = {
746 .family = PF_INET6,
747 .create = inet6_create,
748 .owner = THIS_MODULE,
749};
750
751int inet6_register_protosw(struct inet_protosw *p)
752{
753 struct list_head *lh;
754 struct inet_protosw *answer;
755 struct list_head *last_perm;
756 int protocol = p->protocol;
757 int ret;
758
759 spin_lock_bh(&inetsw6_lock);
760
761 ret = -EINVAL;
762 if (p->type >= SOCK_MAX)
763 goto out_illegal;
764
765 /* If we are trying to override a permanent protocol, bail. */
766 answer = NULL;
767 ret = -EPERM;
768 last_perm = &inetsw6[p->type];
769 list_for_each(lh, &inetsw6[p->type]) {
770 answer = list_entry(lh, struct inet_protosw, list);
771
772 /* Check only the non-wild match. */
773 if (INET_PROTOSW_PERMANENT & answer->flags) {
774 if (protocol == answer->protocol)
775 break;
776 last_perm = lh;
777 }
778
779 answer = NULL;
780 }
781 if (answer)
782 goto out_permanent;
783
784 /* Add the new entry after the last permanent entry if any, so that
785 * the new entry does not override a permanent entry when matched with
786 * a wild-card protocol. But it is allowed to override any existing
787 * non-permanent entry. This means that when we remove this entry, the
788 * system automatically returns to the old behavior.
789 */
790 list_add_rcu(&p->list, last_perm);
791 ret = 0;
792out:
793 spin_unlock_bh(&inetsw6_lock);
794 return ret;
795
796out_permanent:
797 pr_err("Attempt to override permanent protocol %d\n", protocol);
798 goto out;
799
800out_illegal:
801 pr_err("Ignoring attempt to register invalid socket type %d\n",
802 p->type);
803 goto out;
804}
805EXPORT_SYMBOL(inet6_register_protosw);
806
807void
808inet6_unregister_protosw(struct inet_protosw *p)
809{
810 if (INET_PROTOSW_PERMANENT & p->flags) {
811 pr_err("Attempt to unregister permanent protocol %d\n",
812 p->protocol);
813 } else {
814 spin_lock_bh(&inetsw6_lock);
815 list_del_rcu(&p->list);
816 spin_unlock_bh(&inetsw6_lock);
817
818 synchronize_net();
819 }
820}
821EXPORT_SYMBOL(inet6_unregister_protosw);
822
823int inet6_sk_rebuild_header(struct sock *sk)
824{
825 struct ipv6_pinfo *np = inet6_sk(sk);
826 struct dst_entry *dst;
827
828 dst = __sk_dst_check(sk, np->dst_cookie);
829
830 if (!dst) {
831 struct inet_sock *inet = inet_sk(sk);
832 struct in6_addr *final_p, final;
833 struct flowi6 fl6;
834
835 memset(&fl6, 0, sizeof(fl6));
836 fl6.flowi6_proto = sk->sk_protocol;
837 fl6.daddr = sk->sk_v6_daddr;
838 fl6.saddr = np->saddr;
839 fl6.flowlabel = np->flow_label;
840 fl6.flowi6_oif = sk->sk_bound_dev_if;
841 fl6.flowi6_mark = sk->sk_mark;
842 fl6.fl6_dport = inet->inet_dport;
843 fl6.fl6_sport = inet->inet_sport;
844 fl6.flowi6_uid = sk->sk_uid;
845 security_sk_classify_flow(sk, flowi6_to_flowi_common(&fl6));
846
847 rcu_read_lock();
848 final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt),
849 &final);
850 rcu_read_unlock();
851
852 dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p);
853 if (IS_ERR(dst)) {
854 sk->sk_route_caps = 0;
855 WRITE_ONCE(sk->sk_err_soft, -PTR_ERR(dst));
856 return PTR_ERR(dst);
857 }
858
859 ip6_dst_store(sk, dst, NULL, NULL);
860 }
861
862 return 0;
863}
864EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
865
866bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb,
867 const struct inet6_skb_parm *opt)
868{
869 const struct ipv6_pinfo *np = inet6_sk(sk);
870
871 if (np->rxopt.all) {
872 if (((opt->flags & IP6SKB_HOPBYHOP) &&
873 (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) ||
874 (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) &&
875 np->rxopt.bits.rxflow) ||
876 (opt->srcrt && (np->rxopt.bits.srcrt ||
877 np->rxopt.bits.osrcrt)) ||
878 ((opt->dst1 || opt->dst0) &&
879 (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
880 return true;
881 }
882 return false;
883}
884EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
885
886static struct packet_type ipv6_packet_type __read_mostly = {
887 .type = cpu_to_be16(ETH_P_IPV6),
888 .func = ipv6_rcv,
889 .list_func = ipv6_list_rcv,
890};
891
892static int __init ipv6_packet_init(void)
893{
894 dev_add_pack(&ipv6_packet_type);
895 return 0;
896}
897
898static void ipv6_packet_cleanup(void)
899{
900 dev_remove_pack(&ipv6_packet_type);
901}
902
903static int __net_init ipv6_init_mibs(struct net *net)
904{
905 int i;
906
907 net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib);
908 if (!net->mib.udp_stats_in6)
909 return -ENOMEM;
910 net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib);
911 if (!net->mib.udplite_stats_in6)
912 goto err_udplite_mib;
913 net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib);
914 if (!net->mib.ipv6_statistics)
915 goto err_ip_mib;
916
917 for_each_possible_cpu(i) {
918 struct ipstats_mib *af_inet6_stats;
919 af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i);
920 u64_stats_init(&af_inet6_stats->syncp);
921 }
922
923
924 net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib);
925 if (!net->mib.icmpv6_statistics)
926 goto err_icmp_mib;
927 net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
928 GFP_KERNEL);
929 if (!net->mib.icmpv6msg_statistics)
930 goto err_icmpmsg_mib;
931 return 0;
932
933err_icmpmsg_mib:
934 free_percpu(net->mib.icmpv6_statistics);
935err_icmp_mib:
936 free_percpu(net->mib.ipv6_statistics);
937err_ip_mib:
938 free_percpu(net->mib.udplite_stats_in6);
939err_udplite_mib:
940 free_percpu(net->mib.udp_stats_in6);
941 return -ENOMEM;
942}
943
944static void ipv6_cleanup_mibs(struct net *net)
945{
946 free_percpu(net->mib.udp_stats_in6);
947 free_percpu(net->mib.udplite_stats_in6);
948 free_percpu(net->mib.ipv6_statistics);
949 free_percpu(net->mib.icmpv6_statistics);
950 kfree(net->mib.icmpv6msg_statistics);
951}
952
953static int __net_init inet6_net_init(struct net *net)
954{
955 int err = 0;
956
957 net->ipv6.sysctl.bindv6only = 0;
958 net->ipv6.sysctl.icmpv6_time = 1*HZ;
959 net->ipv6.sysctl.icmpv6_echo_ignore_all = 0;
960 net->ipv6.sysctl.icmpv6_echo_ignore_multicast = 0;
961 net->ipv6.sysctl.icmpv6_echo_ignore_anycast = 0;
962 net->ipv6.sysctl.icmpv6_error_anycast_as_unicast = 0;
963
964 /* By default, rate limit error messages.
965 * Except for pmtu discovery, it would break it.
966 * proc_do_large_bitmap needs pointer to the bitmap.
967 */
968 bitmap_set(net->ipv6.sysctl.icmpv6_ratemask, 0, ICMPV6_ERRMSG_MAX + 1);
969 bitmap_clear(net->ipv6.sysctl.icmpv6_ratemask, ICMPV6_PKT_TOOBIG, 1);
970 net->ipv6.sysctl.icmpv6_ratemask_ptr = net->ipv6.sysctl.icmpv6_ratemask;
971
972 net->ipv6.sysctl.flowlabel_consistency = 1;
973 net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS;
974 net->ipv6.sysctl.idgen_retries = 3;
975 net->ipv6.sysctl.idgen_delay = 1 * HZ;
976 net->ipv6.sysctl.flowlabel_state_ranges = 0;
977 net->ipv6.sysctl.max_dst_opts_cnt = IP6_DEFAULT_MAX_DST_OPTS_CNT;
978 net->ipv6.sysctl.max_hbh_opts_cnt = IP6_DEFAULT_MAX_HBH_OPTS_CNT;
979 net->ipv6.sysctl.max_dst_opts_len = IP6_DEFAULT_MAX_DST_OPTS_LEN;
980 net->ipv6.sysctl.max_hbh_opts_len = IP6_DEFAULT_MAX_HBH_OPTS_LEN;
981 net->ipv6.sysctl.fib_notify_on_flag_change = 0;
982 atomic_set(&net->ipv6.fib6_sernum, 1);
983
984 net->ipv6.sysctl.ioam6_id = IOAM6_DEFAULT_ID;
985 net->ipv6.sysctl.ioam6_id_wide = IOAM6_DEFAULT_ID_WIDE;
986
987 err = ipv6_init_mibs(net);
988 if (err)
989 return err;
990#ifdef CONFIG_PROC_FS
991 err = udp6_proc_init(net);
992 if (err)
993 goto out;
994 err = tcp6_proc_init(net);
995 if (err)
996 goto proc_tcp6_fail;
997 err = ac6_proc_init(net);
998 if (err)
999 goto proc_ac6_fail;
1000#endif
1001 return err;
1002
1003#ifdef CONFIG_PROC_FS
1004proc_ac6_fail:
1005 tcp6_proc_exit(net);
1006proc_tcp6_fail:
1007 udp6_proc_exit(net);
1008out:
1009 ipv6_cleanup_mibs(net);
1010 return err;
1011#endif
1012}
1013
1014static void __net_exit inet6_net_exit(struct net *net)
1015{
1016#ifdef CONFIG_PROC_FS
1017 udp6_proc_exit(net);
1018 tcp6_proc_exit(net);
1019 ac6_proc_exit(net);
1020#endif
1021 ipv6_cleanup_mibs(net);
1022}
1023
1024static struct pernet_operations inet6_net_ops = {
1025 .init = inet6_net_init,
1026 .exit = inet6_net_exit,
1027};
1028
1029static int ipv6_route_input(struct sk_buff *skb)
1030{
1031 ip6_route_input(skb);
1032 return skb_dst(skb)->error;
1033}
1034
1035static const struct ipv6_stub ipv6_stub_impl = {
1036 .ipv6_sock_mc_join = ipv6_sock_mc_join,
1037 .ipv6_sock_mc_drop = ipv6_sock_mc_drop,
1038 .ipv6_dst_lookup_flow = ip6_dst_lookup_flow,
1039 .ipv6_route_input = ipv6_route_input,
1040 .fib6_get_table = fib6_get_table,
1041 .fib6_table_lookup = fib6_table_lookup,
1042 .fib6_lookup = fib6_lookup,
1043 .fib6_select_path = fib6_select_path,
1044 .ip6_mtu_from_fib6 = ip6_mtu_from_fib6,
1045 .fib6_nh_init = fib6_nh_init,
1046 .fib6_nh_release = fib6_nh_release,
1047 .fib6_nh_release_dsts = fib6_nh_release_dsts,
1048 .fib6_update_sernum = fib6_update_sernum_stub,
1049 .fib6_rt_update = fib6_rt_update,
1050 .ip6_del_rt = ip6_del_rt,
1051 .udpv6_encap_enable = udpv6_encap_enable,
1052 .ndisc_send_na = ndisc_send_na,
1053#if IS_ENABLED(CONFIG_XFRM)
1054 .xfrm6_local_rxpmtu = xfrm6_local_rxpmtu,
1055 .xfrm6_udp_encap_rcv = xfrm6_udp_encap_rcv,
1056 .xfrm6_gro_udp_encap_rcv = xfrm6_gro_udp_encap_rcv,
1057 .xfrm6_rcv_encap = xfrm6_rcv_encap,
1058#endif
1059 .nd_tbl = &nd_tbl,
1060 .ipv6_fragment = ip6_fragment,
1061 .ipv6_dev_find = ipv6_dev_find,
1062};
1063
1064static const struct ipv6_bpf_stub ipv6_bpf_stub_impl = {
1065 .inet6_bind = __inet6_bind,
1066 .udp6_lib_lookup = __udp6_lib_lookup,
1067 .ipv6_setsockopt = do_ipv6_setsockopt,
1068 .ipv6_getsockopt = do_ipv6_getsockopt,
1069 .ipv6_dev_get_saddr = ipv6_dev_get_saddr,
1070};
1071
1072static int __init inet6_init(void)
1073{
1074 struct list_head *r;
1075 int err = 0;
1076
1077 sock_skb_cb_check_size(sizeof(struct inet6_skb_parm));
1078
1079 /* Register the socket-side information for inet6_create. */
1080 for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
1081 INIT_LIST_HEAD(r);
1082
1083 raw_hashinfo_init(&raw_v6_hashinfo);
1084
1085 if (disable_ipv6_mod) {
1086 pr_info("Loaded, but administratively disabled, reboot required to enable\n");
1087 goto out;
1088 }
1089
1090 err = proto_register(&tcpv6_prot, 1);
1091 if (err)
1092 goto out;
1093
1094 err = proto_register(&udpv6_prot, 1);
1095 if (err)
1096 goto out_unregister_tcp_proto;
1097
1098 err = proto_register(&udplitev6_prot, 1);
1099 if (err)
1100 goto out_unregister_udp_proto;
1101
1102 err = proto_register(&rawv6_prot, 1);
1103 if (err)
1104 goto out_unregister_udplite_proto;
1105
1106 err = proto_register(&pingv6_prot, 1);
1107 if (err)
1108 goto out_unregister_raw_proto;
1109
1110 /* We MUST register RAW sockets before we create the ICMP6,
1111 * IGMP6, or NDISC control sockets.
1112 */
1113 err = rawv6_init();
1114 if (err)
1115 goto out_unregister_ping_proto;
1116
1117 /* Register the family here so that the init calls below will
1118 * be able to create sockets. (?? is this dangerous ??)
1119 */
1120 err = sock_register(&inet6_family_ops);
1121 if (err)
1122 goto out_sock_register_fail;
1123
1124 /*
1125 * ipngwg API draft makes clear that the correct semantics
1126 * for TCP and UDP is to consider one TCP and UDP instance
1127 * in a host available by both INET and INET6 APIs and
1128 * able to communicate via both network protocols.
1129 */
1130
1131 err = register_pernet_subsys(&inet6_net_ops);
1132 if (err)
1133 goto register_pernet_fail;
1134 err = ip6_mr_init();
1135 if (err)
1136 goto ipmr_fail;
1137 err = icmpv6_init();
1138 if (err)
1139 goto icmp_fail;
1140 err = ndisc_init();
1141 if (err)
1142 goto ndisc_fail;
1143 err = igmp6_init();
1144 if (err)
1145 goto igmp_fail;
1146
1147 err = ipv6_netfilter_init();
1148 if (err)
1149 goto netfilter_fail;
1150 /* Create /proc/foo6 entries. */
1151#ifdef CONFIG_PROC_FS
1152 err = -ENOMEM;
1153 if (raw6_proc_init())
1154 goto proc_raw6_fail;
1155 if (udplite6_proc_init())
1156 goto proc_udplite6_fail;
1157 if (ipv6_misc_proc_init())
1158 goto proc_misc6_fail;
1159 if (if6_proc_init())
1160 goto proc_if6_fail;
1161#endif
1162 err = ip6_route_init();
1163 if (err)
1164 goto ip6_route_fail;
1165 err = ndisc_late_init();
1166 if (err)
1167 goto ndisc_late_fail;
1168 err = ip6_flowlabel_init();
1169 if (err)
1170 goto ip6_flowlabel_fail;
1171 err = ipv6_anycast_init();
1172 if (err)
1173 goto ipv6_anycast_fail;
1174 err = addrconf_init();
1175 if (err)
1176 goto addrconf_fail;
1177
1178 /* Init v6 extension headers. */
1179 err = ipv6_exthdrs_init();
1180 if (err)
1181 goto ipv6_exthdrs_fail;
1182
1183 err = ipv6_frag_init();
1184 if (err)
1185 goto ipv6_frag_fail;
1186
1187 /* Init v6 transport protocols. */
1188 err = udpv6_init();
1189 if (err)
1190 goto udpv6_fail;
1191
1192 err = udplitev6_init();
1193 if (err)
1194 goto udplitev6_fail;
1195
1196 err = udpv6_offload_init();
1197 if (err)
1198 goto udpv6_offload_fail;
1199
1200 err = tcpv6_init();
1201 if (err)
1202 goto tcpv6_fail;
1203
1204 err = ipv6_packet_init();
1205 if (err)
1206 goto ipv6_packet_fail;
1207
1208 err = pingv6_init();
1209 if (err)
1210 goto pingv6_fail;
1211
1212 err = calipso_init();
1213 if (err)
1214 goto calipso_fail;
1215
1216 err = seg6_init();
1217 if (err)
1218 goto seg6_fail;
1219
1220 err = rpl_init();
1221 if (err)
1222 goto rpl_fail;
1223
1224 err = ioam6_init();
1225 if (err)
1226 goto ioam6_fail;
1227
1228 err = igmp6_late_init();
1229 if (err)
1230 goto igmp6_late_err;
1231
1232#ifdef CONFIG_SYSCTL
1233 err = ipv6_sysctl_register();
1234 if (err)
1235 goto sysctl_fail;
1236#endif
1237
1238 /* ensure that ipv6 stubs are visible only after ipv6 is ready */
1239 wmb();
1240 ipv6_stub = &ipv6_stub_impl;
1241 ipv6_bpf_stub = &ipv6_bpf_stub_impl;
1242out:
1243 return err;
1244
1245#ifdef CONFIG_SYSCTL
1246sysctl_fail:
1247 igmp6_late_cleanup();
1248#endif
1249igmp6_late_err:
1250 ioam6_exit();
1251ioam6_fail:
1252 rpl_exit();
1253rpl_fail:
1254 seg6_exit();
1255seg6_fail:
1256 calipso_exit();
1257calipso_fail:
1258 pingv6_exit();
1259pingv6_fail:
1260 ipv6_packet_cleanup();
1261ipv6_packet_fail:
1262 tcpv6_exit();
1263tcpv6_fail:
1264 udpv6_offload_exit();
1265udpv6_offload_fail:
1266 udplitev6_exit();
1267udplitev6_fail:
1268 udpv6_exit();
1269udpv6_fail:
1270 ipv6_frag_exit();
1271ipv6_frag_fail:
1272 ipv6_exthdrs_exit();
1273ipv6_exthdrs_fail:
1274 addrconf_cleanup();
1275addrconf_fail:
1276 ipv6_anycast_cleanup();
1277ipv6_anycast_fail:
1278 ip6_flowlabel_cleanup();
1279ip6_flowlabel_fail:
1280 ndisc_late_cleanup();
1281ndisc_late_fail:
1282 ip6_route_cleanup();
1283ip6_route_fail:
1284#ifdef CONFIG_PROC_FS
1285 if6_proc_exit();
1286proc_if6_fail:
1287 ipv6_misc_proc_exit();
1288proc_misc6_fail:
1289 udplite6_proc_exit();
1290proc_udplite6_fail:
1291 raw6_proc_exit();
1292proc_raw6_fail:
1293#endif
1294 ipv6_netfilter_fini();
1295netfilter_fail:
1296 igmp6_cleanup();
1297igmp_fail:
1298 ndisc_cleanup();
1299ndisc_fail:
1300 icmpv6_cleanup();
1301icmp_fail:
1302 ip6_mr_cleanup();
1303ipmr_fail:
1304 unregister_pernet_subsys(&inet6_net_ops);
1305register_pernet_fail:
1306 sock_unregister(PF_INET6);
1307 rtnl_unregister_all(PF_INET6);
1308out_sock_register_fail:
1309 rawv6_exit();
1310out_unregister_ping_proto:
1311 proto_unregister(&pingv6_prot);
1312out_unregister_raw_proto:
1313 proto_unregister(&rawv6_prot);
1314out_unregister_udplite_proto:
1315 proto_unregister(&udplitev6_prot);
1316out_unregister_udp_proto:
1317 proto_unregister(&udpv6_prot);
1318out_unregister_tcp_proto:
1319 proto_unregister(&tcpv6_prot);
1320 goto out;
1321}
1322module_init(inet6_init);
1323
1324MODULE_ALIAS_NETPROTO(PF_INET6);
1// SPDX-License-Identifier: GPL-2.0-or-later
2/*
3 * PF_INET6 socket protocol family
4 * Linux INET6 implementation
5 *
6 * Authors:
7 * Pedro Roque <roque@di.fc.ul.pt>
8 *
9 * Adapted from linux/net/ipv4/af_inet.c
10 *
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
15 */
16
17#define pr_fmt(fmt) "IPv6: " fmt
18
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>
37
38#include <linux/inet.h>
39#include <linux/netdevice.h>
40#include <linux/icmpv6.h>
41#include <linux/netfilter_ipv6.h>
42
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>
68
69#include <linux/uaccess.h>
70#include <linux/mroute6.h>
71
72#include "ip6_offload.h"
73
74MODULE_AUTHOR("Cast of dozens");
75MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
76MODULE_LICENSE("GPL");
77
78/* The inetsw6 table contains everything that inet6_create needs to
79 * build a new socket.
80 */
81static struct list_head inetsw6[SOCK_MAX];
82static DEFINE_SPINLOCK(inetsw6_lock);
83
84struct ipv6_params ipv6_defaults = {
85 .disable_ipv6 = 0,
86 .autoconf = 1,
87};
88
89static int disable_ipv6_mod;
90
91module_param_named(disable, disable_ipv6_mod, int, 0444);
92MODULE_PARM_DESC(disable, "Disable IPv6 module such that it is non-functional");
93
94module_param_named(disable_ipv6, ipv6_defaults.disable_ipv6, int, 0444);
95MODULE_PARM_DESC(disable_ipv6, "Disable IPv6 on all interfaces");
96
97module_param_named(autoconf, ipv6_defaults.autoconf, int, 0444);
98MODULE_PARM_DESC(autoconf, "Enable IPv6 address autoconfiguration on all interfaces");
99
100bool ipv6_mod_enabled(void)
101{
102 return disable_ipv6_mod == 0;
103}
104EXPORT_SYMBOL_GPL(ipv6_mod_enabled);
105
106static struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
107{
108 const int offset = sk->sk_prot->ipv6_pinfo_offset;
109
110 return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
111}
112
113void inet6_sock_destruct(struct sock *sk)
114{
115 inet6_cleanup_sock(sk);
116 inet_sock_destruct(sk);
117}
118EXPORT_SYMBOL_GPL(inet6_sock_destruct);
119
120static int inet6_create(struct net *net, struct socket *sock, int protocol,
121 int kern)
122{
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;
131
132 if (protocol < 0 || protocol >= IPPROTO_MAX)
133 return -EINVAL;
134
135 /* Look for the requested type/protocol pair. */
136lookup_protocol:
137 err = -ESOCKTNOSUPPORT;
138 rcu_read_lock();
139 list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
140
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;
151 }
152 if (IPPROTO_IP == answer->protocol)
153 break;
154 }
155 err = -EPROTONOSUPPORT;
156 }
157
158 if (err) {
159 if (try_loading_module < 2) {
160 rcu_read_unlock();
161 /*
162 * Be more specific, e.g. net-pf-10-proto-132-type-1
163 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
164 */
165 if (++try_loading_module == 1)
166 request_module("net-pf-%d-proto-%d-type-%d",
167 PF_INET6, protocol, sock->type);
168 /*
169 * Fall back to generic, e.g. net-pf-10-proto-132
170 * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
171 */
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;
178 }
179
180 err = -EPERM;
181 if (sock->type == SOCK_RAW && !kern &&
182 !ns_capable(net->user_ns, CAP_NET_RAW))
183 goto out_rcu_unlock;
184
185 sock->ops = answer->ops;
186 answer_prot = answer->prot;
187 answer_flags = answer->flags;
188 rcu_read_unlock();
189
190 WARN_ON(!answer_prot->slab);
191
192 err = -ENOBUFS;
193 sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot, kern);
194 if (!sk)
195 goto out;
196
197 sock_init_data(sock, sk);
198
199 err = 0;
200 if (INET_PROTOSW_REUSE & answer_flags)
201 sk->sk_reuse = SK_CAN_REUSE;
202
203 if (INET_PROTOSW_ICSK & answer_flags)
204 inet_init_csk_locks(sk);
205
206 inet = inet_sk(sk);
207 inet_assign_bit(IS_ICSK, sk, INET_PROTOSW_ICSK & answer_flags);
208
209 if (SOCK_RAW == sock->type) {
210 inet->inet_num = protocol;
211 if (IPPROTO_RAW == protocol)
212 inet_set_bit(HDRINCL, sk);
213 }
214
215 sk->sk_destruct = inet6_sock_destruct;
216 sk->sk_family = PF_INET6;
217 sk->sk_protocol = protocol;
218
219 sk->sk_backlog_rcv = answer->prot->backlog_rcv;
220
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);
231
232 /* Init the ipv4 part of the socket since we can have sockets
233 * using v6 API for ipv4.
234 */
235 inet->uc_ttl = -1;
236
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;
242
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;
247
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.
252 */
253 inet->inet_sport = htons(inet->inet_num);
254 err = sk->sk_prot->hash(sk);
255 if (err)
256 goto out_sk_release;
257 }
258 if (sk->sk_prot->init) {
259 err = sk->sk_prot->init(sk);
260 if (err)
261 goto out_sk_release;
262 }
263
264 if (!kern) {
265 err = BPF_CGROUP_RUN_PROG_INET_SOCK(sk);
266 if (err)
267 goto out_sk_release;
268 }
269out:
270 return err;
271out_rcu_unlock:
272 rcu_read_unlock();
273 goto out;
274out_sk_release:
275 sk_common_release(sk);
276 sock->sk = NULL;
277 goto out;
278}
279
280static int __inet6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len,
281 u32 flags)
282{
283 struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr;
284 struct inet_sock *inet = inet_sk(sk);
285 struct ipv6_pinfo *np = inet6_sk(sk);
286 struct net *net = sock_net(sk);
287 __be32 v4addr = 0;
288 unsigned short snum;
289 bool saved_ipv6only;
290 int addr_type = 0;
291 int err = 0;
292
293 if (addr->sin6_family != AF_INET6)
294 return -EAFNOSUPPORT;
295
296 addr_type = ipv6_addr_type(&addr->sin6_addr);
297 if ((addr_type & IPV6_ADDR_MULTICAST) && sk->sk_type == SOCK_STREAM)
298 return -EINVAL;
299
300 snum = ntohs(addr->sin6_port);
301 if (!(flags & BIND_NO_CAP_NET_BIND_SERVICE) &&
302 snum && inet_port_requires_bind_service(net, snum) &&
303 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
304 return -EACCES;
305
306 if (flags & BIND_WITH_LOCK)
307 lock_sock(sk);
308
309 /* Check these errors (active socket, double bind). */
310 if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
311 err = -EINVAL;
312 goto out;
313 }
314
315 /* Check if the address belongs to the host. */
316 if (addr_type == IPV6_ADDR_MAPPED) {
317 struct net_device *dev = NULL;
318 int chk_addr_ret;
319
320 /* Binding to v4-mapped address on a v6-only socket
321 * makes no sense
322 */
323 if (ipv6_only_sock(sk)) {
324 err = -EINVAL;
325 goto out;
326 }
327
328 rcu_read_lock();
329 if (sk->sk_bound_dev_if) {
330 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
331 if (!dev) {
332 err = -ENODEV;
333 goto out_unlock;
334 }
335 }
336
337 /* Reproduce AF_INET checks to make the bindings consistent */
338 v4addr = addr->sin6_addr.s6_addr32[3];
339 chk_addr_ret = inet_addr_type_dev_table(net, dev, v4addr);
340 rcu_read_unlock();
341
342 if (!inet_addr_valid_or_nonlocal(net, inet, v4addr,
343 chk_addr_ret)) {
344 err = -EADDRNOTAVAIL;
345 goto out;
346 }
347 } else {
348 if (addr_type != IPV6_ADDR_ANY) {
349 struct net_device *dev = NULL;
350
351 rcu_read_lock();
352 if (__ipv6_addr_needs_scope_id(addr_type)) {
353 if (addr_len >= sizeof(struct sockaddr_in6) &&
354 addr->sin6_scope_id) {
355 /* Override any existing binding, if another one
356 * is supplied by user.
357 */
358 sk->sk_bound_dev_if = addr->sin6_scope_id;
359 }
360
361 /* Binding to link-local address requires an interface */
362 if (!sk->sk_bound_dev_if) {
363 err = -EINVAL;
364 goto out_unlock;
365 }
366 }
367
368 if (sk->sk_bound_dev_if) {
369 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
370 if (!dev) {
371 err = -ENODEV;
372 goto out_unlock;
373 }
374 }
375
376 /* ipv4 addr of the socket is invalid. Only the
377 * unspecified and mapped address have a v4 equivalent.
378 */
379 v4addr = LOOPBACK4_IPV6;
380 if (!(addr_type & IPV6_ADDR_MULTICAST)) {
381 if (!ipv6_can_nonlocal_bind(net, inet) &&
382 !ipv6_chk_addr(net, &addr->sin6_addr,
383 dev, 0)) {
384 err = -EADDRNOTAVAIL;
385 goto out_unlock;
386 }
387 }
388 rcu_read_unlock();
389 }
390 }
391
392 inet->inet_rcv_saddr = v4addr;
393 inet->inet_saddr = v4addr;
394
395 sk->sk_v6_rcv_saddr = addr->sin6_addr;
396
397 if (!(addr_type & IPV6_ADDR_MULTICAST))
398 np->saddr = addr->sin6_addr;
399
400 saved_ipv6only = sk->sk_ipv6only;
401 if (addr_type != IPV6_ADDR_ANY && addr_type != IPV6_ADDR_MAPPED)
402 sk->sk_ipv6only = 1;
403
404 /* Make sure we are allowed to bind here. */
405 if (snum || !(inet_test_bit(BIND_ADDRESS_NO_PORT, sk) ||
406 (flags & BIND_FORCE_ADDRESS_NO_PORT))) {
407 err = sk->sk_prot->get_port(sk, snum);
408 if (err) {
409 sk->sk_ipv6only = saved_ipv6only;
410 inet_reset_saddr(sk);
411 goto out;
412 }
413 if (!(flags & BIND_FROM_BPF)) {
414 err = BPF_CGROUP_RUN_PROG_INET6_POST_BIND(sk);
415 if (err) {
416 sk->sk_ipv6only = saved_ipv6only;
417 inet_reset_saddr(sk);
418 if (sk->sk_prot->put_port)
419 sk->sk_prot->put_port(sk);
420 goto out;
421 }
422 }
423 }
424
425 if (addr_type != IPV6_ADDR_ANY)
426 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
427 if (snum)
428 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
429 inet->inet_sport = htons(inet->inet_num);
430 inet->inet_dport = 0;
431 inet->inet_daddr = 0;
432out:
433 if (flags & BIND_WITH_LOCK)
434 release_sock(sk);
435 return err;
436out_unlock:
437 rcu_read_unlock();
438 goto out;
439}
440
441int inet6_bind_sk(struct sock *sk, struct sockaddr *uaddr, int addr_len)
442{
443 u32 flags = BIND_WITH_LOCK;
444 const struct proto *prot;
445 int err = 0;
446
447 /* IPV6_ADDRFORM can change sk->sk_prot under us. */
448 prot = READ_ONCE(sk->sk_prot);
449 /* If the socket has its own bind function then use it. */
450 if (prot->bind)
451 return prot->bind(sk, uaddr, addr_len);
452
453 if (addr_len < SIN6_LEN_RFC2133)
454 return -EINVAL;
455
456 /* BPF prog is run before any checks are done so that if the prog
457 * changes context in a wrong way it will be caught.
458 */
459 err = BPF_CGROUP_RUN_PROG_INET_BIND_LOCK(sk, uaddr, &addr_len,
460 CGROUP_INET6_BIND, &flags);
461 if (err)
462 return err;
463
464 return __inet6_bind(sk, uaddr, addr_len, flags);
465}
466
467/* bind for INET6 API */
468int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
469{
470 return inet6_bind_sk(sock->sk, uaddr, addr_len);
471}
472EXPORT_SYMBOL(inet6_bind);
473
474int inet6_release(struct socket *sock)
475{
476 struct sock *sk = sock->sk;
477
478 if (!sk)
479 return -EINVAL;
480
481 /* Free mc lists */
482 ipv6_sock_mc_close(sk);
483
484 /* Free ac lists */
485 ipv6_sock_ac_close(sk);
486
487 return inet_release(sock);
488}
489EXPORT_SYMBOL(inet6_release);
490
491void inet6_cleanup_sock(struct sock *sk)
492{
493 struct ipv6_pinfo *np = inet6_sk(sk);
494 struct sk_buff *skb;
495 struct ipv6_txoptions *opt;
496
497 /* Release rx options */
498
499 skb = xchg(&np->pktoptions, NULL);
500 kfree_skb(skb);
501
502 skb = xchg(&np->rxpmtu, NULL);
503 kfree_skb(skb);
504
505 /* Free flowlabels */
506 fl6_free_socklist(sk);
507
508 /* Free tx options */
509
510 opt = unrcu_pointer(xchg(&np->opt, NULL));
511 if (opt) {
512 atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
513 txopt_put(opt);
514 }
515}
516EXPORT_SYMBOL_GPL(inet6_cleanup_sock);
517
518/*
519 * This does both peername and sockname.
520 */
521int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
522 int peer)
523{
524 struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
525 int sin_addr_len = sizeof(*sin);
526 struct sock *sk = sock->sk;
527 struct inet_sock *inet = inet_sk(sk);
528 struct ipv6_pinfo *np = inet6_sk(sk);
529
530 sin->sin6_family = AF_INET6;
531 sin->sin6_flowinfo = 0;
532 sin->sin6_scope_id = 0;
533 lock_sock(sk);
534 if (peer) {
535 if (!inet->inet_dport ||
536 (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
537 peer == 1)) {
538 release_sock(sk);
539 return -ENOTCONN;
540 }
541 sin->sin6_port = inet->inet_dport;
542 sin->sin6_addr = sk->sk_v6_daddr;
543 if (inet6_test_bit(SNDFLOW, sk))
544 sin->sin6_flowinfo = np->flow_label;
545 BPF_CGROUP_RUN_SA_PROG(sk, (struct sockaddr *)sin, &sin_addr_len,
546 CGROUP_INET6_GETPEERNAME);
547 } else {
548 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
549 sin->sin6_addr = np->saddr;
550 else
551 sin->sin6_addr = sk->sk_v6_rcv_saddr;
552 sin->sin6_port = inet->inet_sport;
553 BPF_CGROUP_RUN_SA_PROG(sk, (struct sockaddr *)sin, &sin_addr_len,
554 CGROUP_INET6_GETSOCKNAME);
555 }
556 sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
557 sk->sk_bound_dev_if);
558 release_sock(sk);
559 return sin_addr_len;
560}
561EXPORT_SYMBOL(inet6_getname);
562
563int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
564{
565 void __user *argp = (void __user *)arg;
566 struct sock *sk = sock->sk;
567 struct net *net = sock_net(sk);
568 const struct proto *prot;
569
570 switch (cmd) {
571 case SIOCADDRT:
572 case SIOCDELRT: {
573 struct in6_rtmsg rtmsg;
574
575 if (copy_from_user(&rtmsg, argp, sizeof(rtmsg)))
576 return -EFAULT;
577 return ipv6_route_ioctl(net, cmd, &rtmsg);
578 }
579 case SIOCSIFADDR:
580 return addrconf_add_ifaddr(net, argp);
581 case SIOCDIFADDR:
582 return addrconf_del_ifaddr(net, argp);
583 case SIOCSIFDSTADDR:
584 return addrconf_set_dstaddr(net, argp);
585 default:
586 /* IPV6_ADDRFORM can change sk->sk_prot under us. */
587 prot = READ_ONCE(sk->sk_prot);
588 if (!prot->ioctl)
589 return -ENOIOCTLCMD;
590 return sk_ioctl(sk, cmd, (void __user *)arg);
591 }
592 /*NOTREACHED*/
593 return 0;
594}
595EXPORT_SYMBOL(inet6_ioctl);
596
597#ifdef CONFIG_COMPAT
598struct compat_in6_rtmsg {
599 struct in6_addr rtmsg_dst;
600 struct in6_addr rtmsg_src;
601 struct in6_addr rtmsg_gateway;
602 u32 rtmsg_type;
603 u16 rtmsg_dst_len;
604 u16 rtmsg_src_len;
605 u32 rtmsg_metric;
606 u32 rtmsg_info;
607 u32 rtmsg_flags;
608 s32 rtmsg_ifindex;
609};
610
611static int inet6_compat_routing_ioctl(struct sock *sk, unsigned int cmd,
612 struct compat_in6_rtmsg __user *ur)
613{
614 struct in6_rtmsg rt;
615
616 if (copy_from_user(&rt.rtmsg_dst, &ur->rtmsg_dst,
617 3 * sizeof(struct in6_addr)) ||
618 get_user(rt.rtmsg_type, &ur->rtmsg_type) ||
619 get_user(rt.rtmsg_dst_len, &ur->rtmsg_dst_len) ||
620 get_user(rt.rtmsg_src_len, &ur->rtmsg_src_len) ||
621 get_user(rt.rtmsg_metric, &ur->rtmsg_metric) ||
622 get_user(rt.rtmsg_info, &ur->rtmsg_info) ||
623 get_user(rt.rtmsg_flags, &ur->rtmsg_flags) ||
624 get_user(rt.rtmsg_ifindex, &ur->rtmsg_ifindex))
625 return -EFAULT;
626
627
628 return ipv6_route_ioctl(sock_net(sk), cmd, &rt);
629}
630
631int inet6_compat_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
632{
633 void __user *argp = compat_ptr(arg);
634 struct sock *sk = sock->sk;
635
636 switch (cmd) {
637 case SIOCADDRT:
638 case SIOCDELRT:
639 return inet6_compat_routing_ioctl(sk, cmd, argp);
640 default:
641 return -ENOIOCTLCMD;
642 }
643}
644EXPORT_SYMBOL_GPL(inet6_compat_ioctl);
645#endif /* CONFIG_COMPAT */
646
647INDIRECT_CALLABLE_DECLARE(int udpv6_sendmsg(struct sock *, struct msghdr *,
648 size_t));
649int inet6_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
650{
651 struct sock *sk = sock->sk;
652 const struct proto *prot;
653
654 if (unlikely(inet_send_prepare(sk)))
655 return -EAGAIN;
656
657 /* IPV6_ADDRFORM can change sk->sk_prot under us. */
658 prot = READ_ONCE(sk->sk_prot);
659 return INDIRECT_CALL_2(prot->sendmsg, tcp_sendmsg, udpv6_sendmsg,
660 sk, msg, size);
661}
662
663INDIRECT_CALLABLE_DECLARE(int udpv6_recvmsg(struct sock *, struct msghdr *,
664 size_t, int, int *));
665int inet6_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
666 int flags)
667{
668 struct sock *sk = sock->sk;
669 const struct proto *prot;
670 int addr_len = 0;
671 int err;
672
673 if (likely(!(flags & MSG_ERRQUEUE)))
674 sock_rps_record_flow(sk);
675
676 /* IPV6_ADDRFORM can change sk->sk_prot under us. */
677 prot = READ_ONCE(sk->sk_prot);
678 err = INDIRECT_CALL_2(prot->recvmsg, tcp_recvmsg, udpv6_recvmsg,
679 sk, msg, size, flags, &addr_len);
680 if (err >= 0)
681 msg->msg_namelen = addr_len;
682 return err;
683}
684
685const struct proto_ops inet6_stream_ops = {
686 .family = PF_INET6,
687 .owner = THIS_MODULE,
688 .release = inet6_release,
689 .bind = inet6_bind,
690 .connect = inet_stream_connect, /* ok */
691 .socketpair = sock_no_socketpair, /* a do nothing */
692 .accept = inet_accept, /* ok */
693 .getname = inet6_getname,
694 .poll = tcp_poll, /* ok */
695 .ioctl = inet6_ioctl, /* must change */
696 .gettstamp = sock_gettstamp,
697 .listen = inet_listen, /* ok */
698 .shutdown = inet_shutdown, /* ok */
699 .setsockopt = sock_common_setsockopt, /* ok */
700 .getsockopt = sock_common_getsockopt, /* ok */
701 .sendmsg = inet6_sendmsg, /* retpoline's sake */
702 .recvmsg = inet6_recvmsg, /* retpoline's sake */
703#ifdef CONFIG_MMU
704 .mmap = tcp_mmap,
705#endif
706 .splice_eof = inet_splice_eof,
707 .sendmsg_locked = tcp_sendmsg_locked,
708 .splice_read = tcp_splice_read,
709 .set_peek_off = sk_set_peek_off,
710 .read_sock = tcp_read_sock,
711 .read_skb = tcp_read_skb,
712 .peek_len = tcp_peek_len,
713#ifdef CONFIG_COMPAT
714 .compat_ioctl = inet6_compat_ioctl,
715#endif
716 .set_rcvlowat = tcp_set_rcvlowat,
717};
718
719const struct proto_ops inet6_dgram_ops = {
720 .family = PF_INET6,
721 .owner = THIS_MODULE,
722 .release = inet6_release,
723 .bind = inet6_bind,
724 .connect = inet_dgram_connect, /* ok */
725 .socketpair = sock_no_socketpair, /* a do nothing */
726 .accept = sock_no_accept, /* a do nothing */
727 .getname = inet6_getname,
728 .poll = udp_poll, /* ok */
729 .ioctl = inet6_ioctl, /* must change */
730 .gettstamp = sock_gettstamp,
731 .listen = sock_no_listen, /* ok */
732 .shutdown = inet_shutdown, /* ok */
733 .setsockopt = sock_common_setsockopt, /* ok */
734 .getsockopt = sock_common_getsockopt, /* ok */
735 .sendmsg = inet6_sendmsg, /* retpoline's sake */
736 .recvmsg = inet6_recvmsg, /* retpoline's sake */
737 .read_skb = udp_read_skb,
738 .mmap = sock_no_mmap,
739 .set_peek_off = udp_set_peek_off,
740#ifdef CONFIG_COMPAT
741 .compat_ioctl = inet6_compat_ioctl,
742#endif
743};
744
745static const struct net_proto_family inet6_family_ops = {
746 .family = PF_INET6,
747 .create = inet6_create,
748 .owner = THIS_MODULE,
749};
750
751int inet6_register_protosw(struct inet_protosw *p)
752{
753 struct list_head *lh;
754 struct inet_protosw *answer;
755 struct list_head *last_perm;
756 int protocol = p->protocol;
757 int ret;
758
759 spin_lock_bh(&inetsw6_lock);
760
761 ret = -EINVAL;
762 if (p->type >= SOCK_MAX)
763 goto out_illegal;
764
765 /* If we are trying to override a permanent protocol, bail. */
766 answer = NULL;
767 ret = -EPERM;
768 last_perm = &inetsw6[p->type];
769 list_for_each(lh, &inetsw6[p->type]) {
770 answer = list_entry(lh, struct inet_protosw, list);
771
772 /* Check only the non-wild match. */
773 if (INET_PROTOSW_PERMANENT & answer->flags) {
774 if (protocol == answer->protocol)
775 break;
776 last_perm = lh;
777 }
778
779 answer = NULL;
780 }
781 if (answer)
782 goto out_permanent;
783
784 /* Add the new entry after the last permanent entry if any, so that
785 * the new entry does not override a permanent entry when matched with
786 * a wild-card protocol. But it is allowed to override any existing
787 * non-permanent entry. This means that when we remove this entry, the
788 * system automatically returns to the old behavior.
789 */
790 list_add_rcu(&p->list, last_perm);
791 ret = 0;
792out:
793 spin_unlock_bh(&inetsw6_lock);
794 return ret;
795
796out_permanent:
797 pr_err("Attempt to override permanent protocol %d\n", protocol);
798 goto out;
799
800out_illegal:
801 pr_err("Ignoring attempt to register invalid socket type %d\n",
802 p->type);
803 goto out;
804}
805EXPORT_SYMBOL(inet6_register_protosw);
806
807void
808inet6_unregister_protosw(struct inet_protosw *p)
809{
810 if (INET_PROTOSW_PERMANENT & p->flags) {
811 pr_err("Attempt to unregister permanent protocol %d\n",
812 p->protocol);
813 } else {
814 spin_lock_bh(&inetsw6_lock);
815 list_del_rcu(&p->list);
816 spin_unlock_bh(&inetsw6_lock);
817
818 synchronize_net();
819 }
820}
821EXPORT_SYMBOL(inet6_unregister_protosw);
822
823int inet6_sk_rebuild_header(struct sock *sk)
824{
825 struct ipv6_pinfo *np = inet6_sk(sk);
826 struct dst_entry *dst;
827
828 dst = __sk_dst_check(sk, np->dst_cookie);
829
830 if (!dst) {
831 struct inet_sock *inet = inet_sk(sk);
832 struct in6_addr *final_p, final;
833 struct flowi6 fl6;
834
835 memset(&fl6, 0, sizeof(fl6));
836 fl6.flowi6_proto = sk->sk_protocol;
837 fl6.daddr = sk->sk_v6_daddr;
838 fl6.saddr = np->saddr;
839 fl6.flowlabel = np->flow_label;
840 fl6.flowi6_oif = sk->sk_bound_dev_if;
841 fl6.flowi6_mark = sk->sk_mark;
842 fl6.fl6_dport = inet->inet_dport;
843 fl6.fl6_sport = inet->inet_sport;
844 fl6.flowi6_uid = sk->sk_uid;
845 security_sk_classify_flow(sk, flowi6_to_flowi_common(&fl6));
846
847 rcu_read_lock();
848 final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt),
849 &final);
850 rcu_read_unlock();
851
852 dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p);
853 if (IS_ERR(dst)) {
854 sk->sk_route_caps = 0;
855 WRITE_ONCE(sk->sk_err_soft, -PTR_ERR(dst));
856 return PTR_ERR(dst);
857 }
858
859 ip6_dst_store(sk, dst, NULL, NULL);
860 }
861
862 return 0;
863}
864EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
865
866bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb,
867 const struct inet6_skb_parm *opt)
868{
869 const struct ipv6_pinfo *np = inet6_sk(sk);
870
871 if (np->rxopt.all) {
872 if (((opt->flags & IP6SKB_HOPBYHOP) &&
873 (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) ||
874 (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) &&
875 np->rxopt.bits.rxflow) ||
876 (opt->srcrt && (np->rxopt.bits.srcrt ||
877 np->rxopt.bits.osrcrt)) ||
878 ((opt->dst1 || opt->dst0) &&
879 (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
880 return true;
881 }
882 return false;
883}
884EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
885
886static struct packet_type ipv6_packet_type __read_mostly = {
887 .type = cpu_to_be16(ETH_P_IPV6),
888 .func = ipv6_rcv,
889 .list_func = ipv6_list_rcv,
890};
891
892static int __init ipv6_packet_init(void)
893{
894 dev_add_pack(&ipv6_packet_type);
895 return 0;
896}
897
898static void ipv6_packet_cleanup(void)
899{
900 dev_remove_pack(&ipv6_packet_type);
901}
902
903static int __net_init ipv6_init_mibs(struct net *net)
904{
905 int i;
906
907 net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib);
908 if (!net->mib.udp_stats_in6)
909 return -ENOMEM;
910 net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib);
911 if (!net->mib.udplite_stats_in6)
912 goto err_udplite_mib;
913 net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib);
914 if (!net->mib.ipv6_statistics)
915 goto err_ip_mib;
916
917 for_each_possible_cpu(i) {
918 struct ipstats_mib *af_inet6_stats;
919 af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i);
920 u64_stats_init(&af_inet6_stats->syncp);
921 }
922
923
924 net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib);
925 if (!net->mib.icmpv6_statistics)
926 goto err_icmp_mib;
927 net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
928 GFP_KERNEL);
929 if (!net->mib.icmpv6msg_statistics)
930 goto err_icmpmsg_mib;
931 return 0;
932
933err_icmpmsg_mib:
934 free_percpu(net->mib.icmpv6_statistics);
935err_icmp_mib:
936 free_percpu(net->mib.ipv6_statistics);
937err_ip_mib:
938 free_percpu(net->mib.udplite_stats_in6);
939err_udplite_mib:
940 free_percpu(net->mib.udp_stats_in6);
941 return -ENOMEM;
942}
943
944static void ipv6_cleanup_mibs(struct net *net)
945{
946 free_percpu(net->mib.udp_stats_in6);
947 free_percpu(net->mib.udplite_stats_in6);
948 free_percpu(net->mib.ipv6_statistics);
949 free_percpu(net->mib.icmpv6_statistics);
950 kfree(net->mib.icmpv6msg_statistics);
951}
952
953static int __net_init inet6_net_init(struct net *net)
954{
955 int err = 0;
956
957 net->ipv6.sysctl.bindv6only = 0;
958 net->ipv6.sysctl.icmpv6_time = 1*HZ;
959 net->ipv6.sysctl.icmpv6_echo_ignore_all = 0;
960 net->ipv6.sysctl.icmpv6_echo_ignore_multicast = 0;
961 net->ipv6.sysctl.icmpv6_echo_ignore_anycast = 0;
962 net->ipv6.sysctl.icmpv6_error_anycast_as_unicast = 0;
963
964 /* By default, rate limit error messages.
965 * Except for pmtu discovery, it would break it.
966 * proc_do_large_bitmap needs pointer to the bitmap.
967 */
968 bitmap_set(net->ipv6.sysctl.icmpv6_ratemask, 0, ICMPV6_ERRMSG_MAX + 1);
969 bitmap_clear(net->ipv6.sysctl.icmpv6_ratemask, ICMPV6_PKT_TOOBIG, 1);
970 net->ipv6.sysctl.icmpv6_ratemask_ptr = net->ipv6.sysctl.icmpv6_ratemask;
971
972 net->ipv6.sysctl.flowlabel_consistency = 1;
973 net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS;
974 net->ipv6.sysctl.idgen_retries = 3;
975 net->ipv6.sysctl.idgen_delay = 1 * HZ;
976 net->ipv6.sysctl.flowlabel_state_ranges = 0;
977 net->ipv6.sysctl.max_dst_opts_cnt = IP6_DEFAULT_MAX_DST_OPTS_CNT;
978 net->ipv6.sysctl.max_hbh_opts_cnt = IP6_DEFAULT_MAX_HBH_OPTS_CNT;
979 net->ipv6.sysctl.max_dst_opts_len = IP6_DEFAULT_MAX_DST_OPTS_LEN;
980 net->ipv6.sysctl.max_hbh_opts_len = IP6_DEFAULT_MAX_HBH_OPTS_LEN;
981 net->ipv6.sysctl.fib_notify_on_flag_change = 0;
982 atomic_set(&net->ipv6.fib6_sernum, 1);
983
984 net->ipv6.sysctl.ioam6_id = IOAM6_DEFAULT_ID;
985 net->ipv6.sysctl.ioam6_id_wide = IOAM6_DEFAULT_ID_WIDE;
986
987 err = ipv6_init_mibs(net);
988 if (err)
989 return err;
990#ifdef CONFIG_PROC_FS
991 err = udp6_proc_init(net);
992 if (err)
993 goto out;
994 err = tcp6_proc_init(net);
995 if (err)
996 goto proc_tcp6_fail;
997 err = ac6_proc_init(net);
998 if (err)
999 goto proc_ac6_fail;
1000#endif
1001 return err;
1002
1003#ifdef CONFIG_PROC_FS
1004proc_ac6_fail:
1005 tcp6_proc_exit(net);
1006proc_tcp6_fail:
1007 udp6_proc_exit(net);
1008out:
1009 ipv6_cleanup_mibs(net);
1010 return err;
1011#endif
1012}
1013
1014static void __net_exit inet6_net_exit(struct net *net)
1015{
1016#ifdef CONFIG_PROC_FS
1017 udp6_proc_exit(net);
1018 tcp6_proc_exit(net);
1019 ac6_proc_exit(net);
1020#endif
1021 ipv6_cleanup_mibs(net);
1022}
1023
1024static struct pernet_operations inet6_net_ops = {
1025 .init = inet6_net_init,
1026 .exit = inet6_net_exit,
1027};
1028
1029static int ipv6_route_input(struct sk_buff *skb)
1030{
1031 ip6_route_input(skb);
1032 return skb_dst(skb)->error;
1033}
1034
1035static const struct ipv6_stub ipv6_stub_impl = {
1036 .ipv6_sock_mc_join = ipv6_sock_mc_join,
1037 .ipv6_sock_mc_drop = ipv6_sock_mc_drop,
1038 .ipv6_dst_lookup_flow = ip6_dst_lookup_flow,
1039 .ipv6_route_input = ipv6_route_input,
1040 .fib6_get_table = fib6_get_table,
1041 .fib6_table_lookup = fib6_table_lookup,
1042 .fib6_lookup = fib6_lookup,
1043 .fib6_select_path = fib6_select_path,
1044 .ip6_mtu_from_fib6 = ip6_mtu_from_fib6,
1045 .fib6_nh_init = fib6_nh_init,
1046 .fib6_nh_release = fib6_nh_release,
1047 .fib6_nh_release_dsts = fib6_nh_release_dsts,
1048 .fib6_update_sernum = fib6_update_sernum_stub,
1049 .fib6_rt_update = fib6_rt_update,
1050 .ip6_del_rt = ip6_del_rt,
1051 .udpv6_encap_enable = udpv6_encap_enable,
1052 .ndisc_send_na = ndisc_send_na,
1053#if IS_ENABLED(CONFIG_XFRM)
1054 .xfrm6_local_rxpmtu = xfrm6_local_rxpmtu,
1055 .xfrm6_udp_encap_rcv = xfrm6_udp_encap_rcv,
1056 .xfrm6_gro_udp_encap_rcv = xfrm6_gro_udp_encap_rcv,
1057 .xfrm6_rcv_encap = xfrm6_rcv_encap,
1058#endif
1059 .nd_tbl = &nd_tbl,
1060 .ipv6_fragment = ip6_fragment,
1061 .ipv6_dev_find = ipv6_dev_find,
1062 .ip6_xmit = ip6_xmit,
1063};
1064
1065static const struct ipv6_bpf_stub ipv6_bpf_stub_impl = {
1066 .inet6_bind = __inet6_bind,
1067 .udp6_lib_lookup = __udp6_lib_lookup,
1068 .ipv6_setsockopt = do_ipv6_setsockopt,
1069 .ipv6_getsockopt = do_ipv6_getsockopt,
1070 .ipv6_dev_get_saddr = ipv6_dev_get_saddr,
1071};
1072
1073static int __init inet6_init(void)
1074{
1075 struct list_head *r;
1076 int err = 0;
1077
1078 sock_skb_cb_check_size(sizeof(struct inet6_skb_parm));
1079
1080 /* Register the socket-side information for inet6_create. */
1081 for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
1082 INIT_LIST_HEAD(r);
1083
1084 raw_hashinfo_init(&raw_v6_hashinfo);
1085
1086 if (disable_ipv6_mod) {
1087 pr_info("Loaded, but administratively disabled, reboot required to enable\n");
1088 goto out;
1089 }
1090
1091 err = proto_register(&tcpv6_prot, 1);
1092 if (err)
1093 goto out;
1094
1095 err = proto_register(&udpv6_prot, 1);
1096 if (err)
1097 goto out_unregister_tcp_proto;
1098
1099 err = proto_register(&udplitev6_prot, 1);
1100 if (err)
1101 goto out_unregister_udp_proto;
1102
1103 err = proto_register(&rawv6_prot, 1);
1104 if (err)
1105 goto out_unregister_udplite_proto;
1106
1107 err = proto_register(&pingv6_prot, 1);
1108 if (err)
1109 goto out_unregister_raw_proto;
1110
1111 /* We MUST register RAW sockets before we create the ICMP6,
1112 * IGMP6, or NDISC control sockets.
1113 */
1114 err = rawv6_init();
1115 if (err)
1116 goto out_unregister_ping_proto;
1117
1118 /* Register the family here so that the init calls below will
1119 * be able to create sockets. (?? is this dangerous ??)
1120 */
1121 err = sock_register(&inet6_family_ops);
1122 if (err)
1123 goto out_sock_register_fail;
1124
1125 /*
1126 * ipngwg API draft makes clear that the correct semantics
1127 * for TCP and UDP is to consider one TCP and UDP instance
1128 * in a host available by both INET and INET6 APIs and
1129 * able to communicate via both network protocols.
1130 */
1131
1132 err = register_pernet_subsys(&inet6_net_ops);
1133 if (err)
1134 goto register_pernet_fail;
1135 err = ip6_mr_init();
1136 if (err)
1137 goto ipmr_fail;
1138 err = icmpv6_init();
1139 if (err)
1140 goto icmp_fail;
1141 err = ndisc_init();
1142 if (err)
1143 goto ndisc_fail;
1144 err = igmp6_init();
1145 if (err)
1146 goto igmp_fail;
1147
1148 err = ipv6_netfilter_init();
1149 if (err)
1150 goto netfilter_fail;
1151 /* Create /proc/foo6 entries. */
1152#ifdef CONFIG_PROC_FS
1153 err = -ENOMEM;
1154 if (raw6_proc_init())
1155 goto proc_raw6_fail;
1156 if (udplite6_proc_init())
1157 goto proc_udplite6_fail;
1158 if (ipv6_misc_proc_init())
1159 goto proc_misc6_fail;
1160 if (if6_proc_init())
1161 goto proc_if6_fail;
1162#endif
1163 err = ip6_route_init();
1164 if (err)
1165 goto ip6_route_fail;
1166 err = ndisc_late_init();
1167 if (err)
1168 goto ndisc_late_fail;
1169 err = ip6_flowlabel_init();
1170 if (err)
1171 goto ip6_flowlabel_fail;
1172 err = ipv6_anycast_init();
1173 if (err)
1174 goto ipv6_anycast_fail;
1175 err = addrconf_init();
1176 if (err)
1177 goto addrconf_fail;
1178
1179 /* Init v6 extension headers. */
1180 err = ipv6_exthdrs_init();
1181 if (err)
1182 goto ipv6_exthdrs_fail;
1183
1184 err = ipv6_frag_init();
1185 if (err)
1186 goto ipv6_frag_fail;
1187
1188 /* Init v6 transport protocols. */
1189 err = udpv6_init();
1190 if (err)
1191 goto udpv6_fail;
1192
1193 err = udplitev6_init();
1194 if (err)
1195 goto udplitev6_fail;
1196
1197 err = udpv6_offload_init();
1198 if (err)
1199 goto udpv6_offload_fail;
1200
1201 err = tcpv6_init();
1202 if (err)
1203 goto tcpv6_fail;
1204
1205 err = ipv6_packet_init();
1206 if (err)
1207 goto ipv6_packet_fail;
1208
1209 err = pingv6_init();
1210 if (err)
1211 goto pingv6_fail;
1212
1213 err = calipso_init();
1214 if (err)
1215 goto calipso_fail;
1216
1217 err = seg6_init();
1218 if (err)
1219 goto seg6_fail;
1220
1221 err = rpl_init();
1222 if (err)
1223 goto rpl_fail;
1224
1225 err = ioam6_init();
1226 if (err)
1227 goto ioam6_fail;
1228
1229 err = igmp6_late_init();
1230 if (err)
1231 goto igmp6_late_err;
1232
1233#ifdef CONFIG_SYSCTL
1234 err = ipv6_sysctl_register();
1235 if (err)
1236 goto sysctl_fail;
1237#endif
1238
1239 /* ensure that ipv6 stubs are visible only after ipv6 is ready */
1240 wmb();
1241 ipv6_stub = &ipv6_stub_impl;
1242 ipv6_bpf_stub = &ipv6_bpf_stub_impl;
1243out:
1244 return err;
1245
1246#ifdef CONFIG_SYSCTL
1247sysctl_fail:
1248 igmp6_late_cleanup();
1249#endif
1250igmp6_late_err:
1251 ioam6_exit();
1252ioam6_fail:
1253 rpl_exit();
1254rpl_fail:
1255 seg6_exit();
1256seg6_fail:
1257 calipso_exit();
1258calipso_fail:
1259 pingv6_exit();
1260pingv6_fail:
1261 ipv6_packet_cleanup();
1262ipv6_packet_fail:
1263 tcpv6_exit();
1264tcpv6_fail:
1265 udpv6_offload_exit();
1266udpv6_offload_fail:
1267 udplitev6_exit();
1268udplitev6_fail:
1269 udpv6_exit();
1270udpv6_fail:
1271 ipv6_frag_exit();
1272ipv6_frag_fail:
1273 ipv6_exthdrs_exit();
1274ipv6_exthdrs_fail:
1275 addrconf_cleanup();
1276addrconf_fail:
1277 ipv6_anycast_cleanup();
1278ipv6_anycast_fail:
1279 ip6_flowlabel_cleanup();
1280ip6_flowlabel_fail:
1281 ndisc_late_cleanup();
1282ndisc_late_fail:
1283 ip6_route_cleanup();
1284ip6_route_fail:
1285#ifdef CONFIG_PROC_FS
1286 if6_proc_exit();
1287proc_if6_fail:
1288 ipv6_misc_proc_exit();
1289proc_misc6_fail:
1290 udplite6_proc_exit();
1291proc_udplite6_fail:
1292 raw6_proc_exit();
1293proc_raw6_fail:
1294#endif
1295 ipv6_netfilter_fini();
1296netfilter_fail:
1297 igmp6_cleanup();
1298igmp_fail:
1299 ndisc_cleanup();
1300ndisc_fail:
1301 icmpv6_cleanup();
1302icmp_fail:
1303 ip6_mr_cleanup();
1304ipmr_fail:
1305 unregister_pernet_subsys(&inet6_net_ops);
1306register_pernet_fail:
1307 sock_unregister(PF_INET6);
1308 rtnl_unregister_all(PF_INET6);
1309out_sock_register_fail:
1310 rawv6_exit();
1311out_unregister_ping_proto:
1312 proto_unregister(&pingv6_prot);
1313out_unregister_raw_proto:
1314 proto_unregister(&rawv6_prot);
1315out_unregister_udplite_proto:
1316 proto_unregister(&udplitev6_prot);
1317out_unregister_udp_proto:
1318 proto_unregister(&udpv6_prot);
1319out_unregister_tcp_proto:
1320 proto_unregister(&tcpv6_prot);
1321 goto out;
1322}
1323module_init(inet6_init);
1324
1325MODULE_ALIAS_NETPROTO(PF_INET6);