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
67 #include <linux/uaccess.h>
68 #include <linux/mroute6.h>
69
70 #include "ip6_offload.h"
71
72 MODULE_AUTHOR("Cast of dozens");
73 MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
74 MODULE_LICENSE("GPL");
75
76 /* The inetsw6 table contains everything that inet6_create needs to
77 * build a new socket.
78 */
79 static struct list_head inetsw6[SOCK_MAX];
80 static DEFINE_SPINLOCK(inetsw6_lock);
81
82 struct ipv6_params ipv6_defaults = {
83 .disable_ipv6 = 0,
84 .autoconf = 1,
85 };
86
87 static int disable_ipv6_mod;
88
89 module_param_named(disable, disable_ipv6_mod, int, 0444);
90 MODULE_PARM_DESC(disable, "Disable IPv6 module such that it is non-functional");
91
92 module_param_named(disable_ipv6, ipv6_defaults.disable_ipv6, int, 0444);
93 MODULE_PARM_DESC(disable_ipv6, "Disable IPv6 on all interfaces");
94
95 module_param_named(autoconf, ipv6_defaults.autoconf, int, 0444);
96 MODULE_PARM_DESC(autoconf, "Enable IPv6 address autoconfiguration on all interfaces");
97
98 bool ipv6_mod_enabled(void)
99 {
100 return disable_ipv6_mod == 0;
101 }
102 EXPORT_SYMBOL_GPL(ipv6_mod_enabled);
103
104 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
105 {
106 const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
107
108 return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
109 }
110
111 static int inet6_create(struct net *net, struct socket *sock, int protocol,
112 int kern)
113 {
114 struct inet_sock *inet;
115 struct ipv6_pinfo *np;
116 struct sock *sk;
117 struct inet_protosw *answer;
118 struct proto *answer_prot;
119 unsigned char answer_flags;
120 int try_loading_module = 0;
121 int err;
122
123 if (protocol < 0 || protocol >= IPPROTO_MAX)
124 return -EINVAL;
125
126 /* Look for the requested type/protocol pair. */
127 lookup_protocol:
128 err = -ESOCKTNOSUPPORT;
129 rcu_read_lock();
130 list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
131
132 err = 0;
133 /* Check the non-wild match. */
134 if (protocol == answer->protocol) {
135 if (protocol != IPPROTO_IP)
136 break;
137 } else {
138 /* Check for the two wild cases. */
139 if (IPPROTO_IP == protocol) {
140 protocol = answer->protocol;
141 break;
142 }
143 if (IPPROTO_IP == answer->protocol)
144 break;
145 }
146 err = -EPROTONOSUPPORT;
147 }
148
149 if (err) {
150 if (try_loading_module < 2) {
151 rcu_read_unlock();
152 /*
153 * Be more specific, e.g. net-pf-10-proto-132-type-1
154 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
155 */
156 if (++try_loading_module == 1)
157 request_module("net-pf-%d-proto-%d-type-%d",
158 PF_INET6, protocol, sock->type);
159 /*
160 * Fall back to generic, e.g. net-pf-10-proto-132
161 * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
162 */
163 else
164 request_module("net-pf-%d-proto-%d",
165 PF_INET6, protocol);
166 goto lookup_protocol;
167 } else
168 goto out_rcu_unlock;
169 }
170
171 err = -EPERM;
172 if (sock->type == SOCK_RAW && !kern &&
173 !ns_capable(net->user_ns, CAP_NET_RAW))
174 goto out_rcu_unlock;
175
176 sock->ops = answer->ops;
177 answer_prot = answer->prot;
178 answer_flags = answer->flags;
179 rcu_read_unlock();
180
181 WARN_ON(!answer_prot->slab);
182
183 err = -ENOBUFS;
184 sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot, kern);
185 if (!sk)
186 goto out;
187
188 sock_init_data(sock, sk);
189
190 err = 0;
191 if (INET_PROTOSW_REUSE & answer_flags)
192 sk->sk_reuse = SK_CAN_REUSE;
193
194 inet = inet_sk(sk);
195 inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
196
197 if (SOCK_RAW == sock->type) {
198 inet->inet_num = protocol;
199 if (IPPROTO_RAW == protocol)
200 inet->hdrincl = 1;
201 }
202
203 sk->sk_destruct = inet_sock_destruct;
204 sk->sk_family = PF_INET6;
205 sk->sk_protocol = protocol;
206
207 sk->sk_backlog_rcv = answer->prot->backlog_rcv;
208
209 inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
210 np->hop_limit = -1;
211 np->mcast_hops = IPV6_DEFAULT_MCASTHOPS;
212 np->mc_loop = 1;
213 np->mc_all = 1;
214 np->pmtudisc = IPV6_PMTUDISC_WANT;
215 np->repflow = net->ipv6.sysctl.flowlabel_reflect & FLOWLABEL_REFLECT_ESTABLISHED;
216 sk->sk_ipv6only = net->ipv6.sysctl.bindv6only;
217
218 /* Init the ipv4 part of the socket since we can have sockets
219 * using v6 API for ipv4.
220 */
221 inet->uc_ttl = -1;
222
223 inet->mc_loop = 1;
224 inet->mc_ttl = 1;
225 inet->mc_index = 0;
226 RCU_INIT_POINTER(inet->mc_list, NULL);
227 inet->rcv_tos = 0;
228
229 if (net->ipv4.sysctl_ip_no_pmtu_disc)
230 inet->pmtudisc = IP_PMTUDISC_DONT;
231 else
232 inet->pmtudisc = IP_PMTUDISC_WANT;
233 /*
234 * Increment only the relevant sk_prot->socks debug field, this changes
235 * the previous behaviour of incrementing both the equivalent to
236 * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
237 *
238 * This allows better debug granularity as we'll know exactly how many
239 * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
240 * transport protocol socks. -acme
241 */
242 sk_refcnt_debug_inc(sk);
243
244 if (inet->inet_num) {
245 /* It assumes that any protocol which allows
246 * the user to assign a number at socket
247 * creation time automatically shares.
248 */
249 inet->inet_sport = htons(inet->inet_num);
250 err = sk->sk_prot->hash(sk);
251 if (err) {
252 sk_common_release(sk);
253 goto out;
254 }
255 }
256 if (sk->sk_prot->init) {
257 err = sk->sk_prot->init(sk);
258 if (err) {
259 sk_common_release(sk);
260 goto out;
261 }
262 }
263
264 if (!kern) {
265 err = BPF_CGROUP_RUN_PROG_INET_SOCK(sk);
266 if (err) {
267 sk_common_release(sk);
268 goto out;
269 }
270 }
271 out:
272 return err;
273 out_rcu_unlock:
274 rcu_read_unlock();
275 goto out;
276 }
277
278 static int __inet6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len,
279 u32 flags)
280 {
281 struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr;
282 struct inet_sock *inet = inet_sk(sk);
283 struct ipv6_pinfo *np = inet6_sk(sk);
284 struct net *net = sock_net(sk);
285 __be32 v4addr = 0;
286 unsigned short snum;
287 bool saved_ipv6only;
288 int addr_type = 0;
289 int err = 0;
290
291 if (addr->sin6_family != AF_INET6)
292 return -EAFNOSUPPORT;
293
294 addr_type = ipv6_addr_type(&addr->sin6_addr);
295 if ((addr_type & IPV6_ADDR_MULTICAST) && sk->sk_type == SOCK_STREAM)
296 return -EINVAL;
297
298 snum = ntohs(addr->sin6_port);
299 if (!(flags & BIND_NO_CAP_NET_BIND_SERVICE) &&
300 snum && inet_port_requires_bind_service(net, snum) &&
301 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
302 return -EACCES;
303
304 if (flags & BIND_WITH_LOCK)
305 lock_sock(sk);
306
307 /* Check these errors (active socket, double bind). */
308 if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
309 err = -EINVAL;
310 goto out;
311 }
312
313 /* Check if the address belongs to the host. */
314 if (addr_type == IPV6_ADDR_MAPPED) {
315 struct net_device *dev = NULL;
316 int chk_addr_ret;
317
318 /* Binding to v4-mapped address on a v6-only socket
319 * makes no sense
320 */
321 if (sk->sk_ipv6only) {
322 err = -EINVAL;
323 goto out;
324 }
325
326 rcu_read_lock();
327 if (sk->sk_bound_dev_if) {
328 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
329 if (!dev) {
330 err = -ENODEV;
331 goto out_unlock;
332 }
333 }
334
335 /* Reproduce AF_INET checks to make the bindings consistent */
336 v4addr = addr->sin6_addr.s6_addr32[3];
337 chk_addr_ret = inet_addr_type_dev_table(net, dev, v4addr);
338 rcu_read_unlock();
339
340 if (!inet_addr_valid_or_nonlocal(net, inet, v4addr,
341 chk_addr_ret)) {
342 err = -EADDRNOTAVAIL;
343 goto out;
344 }
345 } else {
346 if (addr_type != IPV6_ADDR_ANY) {
347 struct net_device *dev = NULL;
348
349 rcu_read_lock();
350 if (__ipv6_addr_needs_scope_id(addr_type)) {
351 if (addr_len >= sizeof(struct sockaddr_in6) &&
352 addr->sin6_scope_id) {
353 /* Override any existing binding, if another one
354 * is supplied by user.
355 */
356 sk->sk_bound_dev_if = addr->sin6_scope_id;
357 }
358
359 /* Binding to link-local address requires an interface */
360 if (!sk->sk_bound_dev_if) {
361 err = -EINVAL;
362 goto out_unlock;
363 }
364 }
365
366 if (sk->sk_bound_dev_if) {
367 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
368 if (!dev) {
369 err = -ENODEV;
370 goto out_unlock;
371 }
372 }
373
374 /* ipv4 addr of the socket is invalid. Only the
375 * unspecified and mapped address have a v4 equivalent.
376 */
377 v4addr = LOOPBACK4_IPV6;
378 if (!(addr_type & IPV6_ADDR_MULTICAST)) {
379 if (!ipv6_can_nonlocal_bind(net, inet) &&
380 !ipv6_chk_addr(net, &addr->sin6_addr,
381 dev, 0)) {
382 err = -EADDRNOTAVAIL;
383 goto out_unlock;
384 }
385 }
386 rcu_read_unlock();
387 }
388 }
389
390 inet->inet_rcv_saddr = v4addr;
391 inet->inet_saddr = v4addr;
392
393 sk->sk_v6_rcv_saddr = addr->sin6_addr;
394
395 if (!(addr_type & IPV6_ADDR_MULTICAST))
396 np->saddr = addr->sin6_addr;
397
398 saved_ipv6only = sk->sk_ipv6only;
399 if (addr_type != IPV6_ADDR_ANY && addr_type != IPV6_ADDR_MAPPED)
400 sk->sk_ipv6only = 1;
401
402 /* Make sure we are allowed to bind here. */
403 if (snum || !(inet->bind_address_no_port ||
404 (flags & BIND_FORCE_ADDRESS_NO_PORT))) {
405 if (sk->sk_prot->get_port(sk, snum)) {
406 sk->sk_ipv6only = saved_ipv6only;
407 inet_reset_saddr(sk);
408 err = -EADDRINUSE;
409 goto out;
410 }
411 if (!(flags & BIND_FROM_BPF)) {
412 err = BPF_CGROUP_RUN_PROG_INET6_POST_BIND(sk);
413 if (err) {
414 sk->sk_ipv6only = saved_ipv6only;
415 inet_reset_saddr(sk);
416 if (sk->sk_prot->put_port)
417 sk->sk_prot->put_port(sk);
418 goto out;
419 }
420 }
421 }
422
423 if (addr_type != IPV6_ADDR_ANY)
424 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
425 if (snum)
426 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
427 inet->inet_sport = htons(inet->inet_num);
428 inet->inet_dport = 0;
429 inet->inet_daddr = 0;
430 out:
431 if (flags & BIND_WITH_LOCK)
432 release_sock(sk);
433 return err;
434 out_unlock:
435 rcu_read_unlock();
436 goto out;
437 }
438
439 /* bind for INET6 API */
440 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
441 {
442 struct sock *sk = sock->sk;
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,
460 CGROUP_INET6_BIND, &flags);
461 if (err)
462 return err;
463
464 return __inet6_bind(sk, uaddr, addr_len, flags);
465 }
466 EXPORT_SYMBOL(inet6_bind);
467
468 int inet6_release(struct socket *sock)
469 {
470 struct sock *sk = sock->sk;
471
472 if (!sk)
473 return -EINVAL;
474
475 /* Free mc lists */
476 ipv6_sock_mc_close(sk);
477
478 /* Free ac lists */
479 ipv6_sock_ac_close(sk);
480
481 return inet_release(sock);
482 }
483 EXPORT_SYMBOL(inet6_release);
484
485 void inet6_destroy_sock(struct sock *sk)
486 {
487 struct ipv6_pinfo *np = inet6_sk(sk);
488 struct sk_buff *skb;
489 struct ipv6_txoptions *opt;
490
491 /* Release rx options */
492
493 skb = xchg(&np->pktoptions, NULL);
494 kfree_skb(skb);
495
496 skb = xchg(&np->rxpmtu, NULL);
497 kfree_skb(skb);
498
499 /* Free flowlabels */
500 fl6_free_socklist(sk);
501
502 /* Free tx options */
503
504 opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL);
505 if (opt) {
506 atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
507 txopt_put(opt);
508 }
509 }
510 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
511
512 /*
513 * This does both peername and sockname.
514 */
515 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
516 int peer)
517 {
518 struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
519 struct sock *sk = sock->sk;
520 struct inet_sock *inet = inet_sk(sk);
521 struct ipv6_pinfo *np = inet6_sk(sk);
522
523 sin->sin6_family = AF_INET6;
524 sin->sin6_flowinfo = 0;
525 sin->sin6_scope_id = 0;
526 lock_sock(sk);
527 if (peer) {
528 if (!inet->inet_dport ||
529 (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
530 peer == 1)) {
531 release_sock(sk);
532 return -ENOTCONN;
533 }
534 sin->sin6_port = inet->inet_dport;
535 sin->sin6_addr = sk->sk_v6_daddr;
536 if (np->sndflow)
537 sin->sin6_flowinfo = np->flow_label;
538 BPF_CGROUP_RUN_SA_PROG(sk, (struct sockaddr *)sin,
539 CGROUP_INET6_GETPEERNAME);
540 } else {
541 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
542 sin->sin6_addr = np->saddr;
543 else
544 sin->sin6_addr = sk->sk_v6_rcv_saddr;
545 sin->sin6_port = inet->inet_sport;
546 BPF_CGROUP_RUN_SA_PROG(sk, (struct sockaddr *)sin,
547 CGROUP_INET6_GETSOCKNAME);
548 }
549 sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
550 sk->sk_bound_dev_if);
551 release_sock(sk);
552 return sizeof(*sin);
553 }
554 EXPORT_SYMBOL(inet6_getname);
555
556 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
557 {
558 void __user *argp = (void __user *)arg;
559 struct sock *sk = sock->sk;
560 struct net *net = sock_net(sk);
561 const struct proto *prot;
562
563 switch (cmd) {
564 case SIOCADDRT:
565 case SIOCDELRT: {
566 struct in6_rtmsg rtmsg;
567
568 if (copy_from_user(&rtmsg, argp, sizeof(rtmsg)))
569 return -EFAULT;
570 return ipv6_route_ioctl(net, cmd, &rtmsg);
571 }
572 case SIOCSIFADDR:
573 return addrconf_add_ifaddr(net, argp);
574 case SIOCDIFADDR:
575 return addrconf_del_ifaddr(net, argp);
576 case SIOCSIFDSTADDR:
577 return addrconf_set_dstaddr(net, argp);
578 default:
579 /* IPV6_ADDRFORM can change sk->sk_prot under us. */
580 prot = READ_ONCE(sk->sk_prot);
581 if (!prot->ioctl)
582 return -ENOIOCTLCMD;
583 return prot->ioctl(sk, cmd, arg);
584 }
585 /*NOTREACHED*/
586 return 0;
587 }
588 EXPORT_SYMBOL(inet6_ioctl);
589
590 #ifdef CONFIG_COMPAT
591 struct compat_in6_rtmsg {
592 struct in6_addr rtmsg_dst;
593 struct in6_addr rtmsg_src;
594 struct in6_addr rtmsg_gateway;
595 u32 rtmsg_type;
596 u16 rtmsg_dst_len;
597 u16 rtmsg_src_len;
598 u32 rtmsg_metric;
599 u32 rtmsg_info;
600 u32 rtmsg_flags;
601 s32 rtmsg_ifindex;
602 };
603
604 static int inet6_compat_routing_ioctl(struct sock *sk, unsigned int cmd,
605 struct compat_in6_rtmsg __user *ur)
606 {
607 struct in6_rtmsg rt;
608
609 if (copy_from_user(&rt.rtmsg_dst, &ur->rtmsg_dst,
610 3 * sizeof(struct in6_addr)) ||
611 get_user(rt.rtmsg_type, &ur->rtmsg_type) ||
612 get_user(rt.rtmsg_dst_len, &ur->rtmsg_dst_len) ||
613 get_user(rt.rtmsg_src_len, &ur->rtmsg_src_len) ||
614 get_user(rt.rtmsg_metric, &ur->rtmsg_metric) ||
615 get_user(rt.rtmsg_info, &ur->rtmsg_info) ||
616 get_user(rt.rtmsg_flags, &ur->rtmsg_flags) ||
617 get_user(rt.rtmsg_ifindex, &ur->rtmsg_ifindex))
618 return -EFAULT;
619
620
621 return ipv6_route_ioctl(sock_net(sk), cmd, &rt);
622 }
623
624 int inet6_compat_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
625 {
626 void __user *argp = compat_ptr(arg);
627 struct sock *sk = sock->sk;
628
629 switch (cmd) {
630 case SIOCADDRT:
631 case SIOCDELRT:
632 return inet6_compat_routing_ioctl(sk, cmd, argp);
633 default:
634 return -ENOIOCTLCMD;
635 }
636 }
637 EXPORT_SYMBOL_GPL(inet6_compat_ioctl);
638 #endif /* CONFIG_COMPAT */
639
640 INDIRECT_CALLABLE_DECLARE(int udpv6_sendmsg(struct sock *, struct msghdr *,
641 size_t));
642 int inet6_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
643 {
644 struct sock *sk = sock->sk;
645 const struct proto *prot;
646
647 if (unlikely(inet_send_prepare(sk)))
648 return -EAGAIN;
649
650 /* IPV6_ADDRFORM can change sk->sk_prot under us. */
651 prot = READ_ONCE(sk->sk_prot);
652 return INDIRECT_CALL_2(prot->sendmsg, tcp_sendmsg, udpv6_sendmsg,
653 sk, msg, size);
654 }
655
656 INDIRECT_CALLABLE_DECLARE(int udpv6_recvmsg(struct sock *, struct msghdr *,
657 size_t, int, int, int *));
658 int inet6_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
659 int flags)
660 {
661 struct sock *sk = sock->sk;
662 const struct proto *prot;
663 int addr_len = 0;
664 int err;
665
666 if (likely(!(flags & MSG_ERRQUEUE)))
667 sock_rps_record_flow(sk);
668
669 /* IPV6_ADDRFORM can change sk->sk_prot under us. */
670 prot = READ_ONCE(sk->sk_prot);
671 err = INDIRECT_CALL_2(prot->recvmsg, tcp_recvmsg, udpv6_recvmsg,
672 sk, msg, size, flags & MSG_DONTWAIT,
673 flags & ~MSG_DONTWAIT, &addr_len);
674 if (err >= 0)
675 msg->msg_namelen = addr_len;
676 return err;
677 }
678
679 const struct proto_ops inet6_stream_ops = {
680 .family = PF_INET6,
681 .owner = THIS_MODULE,
682 .release = inet6_release,
683 .bind = inet6_bind,
684 .connect = inet_stream_connect, /* ok */
685 .socketpair = sock_no_socketpair, /* a do nothing */
686 .accept = inet_accept, /* ok */
687 .getname = inet6_getname,
688 .poll = tcp_poll, /* ok */
689 .ioctl = inet6_ioctl, /* must change */
690 .gettstamp = sock_gettstamp,
691 .listen = inet_listen, /* ok */
692 .shutdown = inet_shutdown, /* ok */
693 .setsockopt = sock_common_setsockopt, /* ok */
694 .getsockopt = sock_common_getsockopt, /* ok */
695 .sendmsg = inet6_sendmsg, /* retpoline's sake */
696 .recvmsg = inet6_recvmsg, /* retpoline's sake */
697 #ifdef CONFIG_MMU
698 .mmap = tcp_mmap,
699 #endif
700 .sendpage = inet_sendpage,
701 .sendmsg_locked = tcp_sendmsg_locked,
702 .sendpage_locked = tcp_sendpage_locked,
703 .splice_read = tcp_splice_read,
704 .read_sock = tcp_read_sock,
705 .peek_len = tcp_peek_len,
706 #ifdef CONFIG_COMPAT
707 .compat_ioctl = inet6_compat_ioctl,
708 #endif
709 .set_rcvlowat = tcp_set_rcvlowat,
710 };
711
712 const struct proto_ops inet6_dgram_ops = {
713 .family = PF_INET6,
714 .owner = THIS_MODULE,
715 .release = inet6_release,
716 .bind = inet6_bind,
717 .connect = inet_dgram_connect, /* ok */
718 .socketpair = sock_no_socketpair, /* a do nothing */
719 .accept = sock_no_accept, /* a do nothing */
720 .getname = inet6_getname,
721 .poll = udp_poll, /* ok */
722 .ioctl = inet6_ioctl, /* must change */
723 .gettstamp = sock_gettstamp,
724 .listen = sock_no_listen, /* ok */
725 .shutdown = inet_shutdown, /* ok */
726 .setsockopt = sock_common_setsockopt, /* ok */
727 .getsockopt = sock_common_getsockopt, /* ok */
728 .sendmsg = inet6_sendmsg, /* retpoline's sake */
729 .recvmsg = inet6_recvmsg, /* retpoline's sake */
730 .read_sock = udp_read_sock,
731 .mmap = sock_no_mmap,
732 .sendpage = sock_no_sendpage,
733 .set_peek_off = sk_set_peek_off,
734 #ifdef CONFIG_COMPAT
735 .compat_ioctl = inet6_compat_ioctl,
736 #endif
737 };
738
739 static const struct net_proto_family inet6_family_ops = {
740 .family = PF_INET6,
741 .create = inet6_create,
742 .owner = THIS_MODULE,
743 };
744
745 int inet6_register_protosw(struct inet_protosw *p)
746 {
747 struct list_head *lh;
748 struct inet_protosw *answer;
749 struct list_head *last_perm;
750 int protocol = p->protocol;
751 int ret;
752
753 spin_lock_bh(&inetsw6_lock);
754
755 ret = -EINVAL;
756 if (p->type >= SOCK_MAX)
757 goto out_illegal;
758
759 /* If we are trying to override a permanent protocol, bail. */
760 answer = NULL;
761 ret = -EPERM;
762 last_perm = &inetsw6[p->type];
763 list_for_each(lh, &inetsw6[p->type]) {
764 answer = list_entry(lh, struct inet_protosw, list);
765
766 /* Check only the non-wild match. */
767 if (INET_PROTOSW_PERMANENT & answer->flags) {
768 if (protocol == answer->protocol)
769 break;
770 last_perm = lh;
771 }
772
773 answer = NULL;
774 }
775 if (answer)
776 goto out_permanent;
777
778 /* Add the new entry after the last permanent entry if any, so that
779 * the new entry does not override a permanent entry when matched with
780 * a wild-card protocol. But it is allowed to override any existing
781 * non-permanent entry. This means that when we remove this entry, the
782 * system automatically returns to the old behavior.
783 */
784 list_add_rcu(&p->list, last_perm);
785 ret = 0;
786 out:
787 spin_unlock_bh(&inetsw6_lock);
788 return ret;
789
790 out_permanent:
791 pr_err("Attempt to override permanent protocol %d\n", protocol);
792 goto out;
793
794 out_illegal:
795 pr_err("Ignoring attempt to register invalid socket type %d\n",
796 p->type);
797 goto out;
798 }
799 EXPORT_SYMBOL(inet6_register_protosw);
800
801 void
802 inet6_unregister_protosw(struct inet_protosw *p)
803 {
804 if (INET_PROTOSW_PERMANENT & p->flags) {
805 pr_err("Attempt to unregister permanent protocol %d\n",
806 p->protocol);
807 } else {
808 spin_lock_bh(&inetsw6_lock);
809 list_del_rcu(&p->list);
810 spin_unlock_bh(&inetsw6_lock);
811
812 synchronize_net();
813 }
814 }
815 EXPORT_SYMBOL(inet6_unregister_protosw);
816
817 int inet6_sk_rebuild_header(struct sock *sk)
818 {
819 struct ipv6_pinfo *np = inet6_sk(sk);
820 struct dst_entry *dst;
821
822 dst = __sk_dst_check(sk, np->dst_cookie);
823
824 if (!dst) {
825 struct inet_sock *inet = inet_sk(sk);
826 struct in6_addr *final_p, final;
827 struct flowi6 fl6;
828
829 memset(&fl6, 0, sizeof(fl6));
830 fl6.flowi6_proto = sk->sk_protocol;
831 fl6.daddr = sk->sk_v6_daddr;
832 fl6.saddr = np->saddr;
833 fl6.flowlabel = np->flow_label;
834 fl6.flowi6_oif = sk->sk_bound_dev_if;
835 fl6.flowi6_mark = sk->sk_mark;
836 fl6.fl6_dport = inet->inet_dport;
837 fl6.fl6_sport = inet->inet_sport;
838 fl6.flowi6_uid = sk->sk_uid;
839 security_sk_classify_flow(sk, flowi6_to_flowi_common(&fl6));
840
841 rcu_read_lock();
842 final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt),
843 &final);
844 rcu_read_unlock();
845
846 dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p);
847 if (IS_ERR(dst)) {
848 sk->sk_route_caps = 0;
849 sk->sk_err_soft = -PTR_ERR(dst);
850 return PTR_ERR(dst);
851 }
852
853 ip6_dst_store(sk, dst, NULL, NULL);
854 }
855
856 return 0;
857 }
858 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
859
860 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb,
861 const struct inet6_skb_parm *opt)
862 {
863 const struct ipv6_pinfo *np = inet6_sk(sk);
864
865 if (np->rxopt.all) {
866 if (((opt->flags & IP6SKB_HOPBYHOP) &&
867 (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) ||
868 (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) &&
869 np->rxopt.bits.rxflow) ||
870 (opt->srcrt && (np->rxopt.bits.srcrt ||
871 np->rxopt.bits.osrcrt)) ||
872 ((opt->dst1 || opt->dst0) &&
873 (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
874 return true;
875 }
876 return false;
877 }
878 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
879
880 static struct packet_type ipv6_packet_type __read_mostly = {
881 .type = cpu_to_be16(ETH_P_IPV6),
882 .func = ipv6_rcv,
883 .list_func = ipv6_list_rcv,
884 };
885
886 static int __init ipv6_packet_init(void)
887 {
888 dev_add_pack(&ipv6_packet_type);
889 return 0;
890 }
891
892 static void ipv6_packet_cleanup(void)
893 {
894 dev_remove_pack(&ipv6_packet_type);
895 }
896
897 static int __net_init ipv6_init_mibs(struct net *net)
898 {
899 int i;
900
901 net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib);
902 if (!net->mib.udp_stats_in6)
903 return -ENOMEM;
904 net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib);
905 if (!net->mib.udplite_stats_in6)
906 goto err_udplite_mib;
907 net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib);
908 if (!net->mib.ipv6_statistics)
909 goto err_ip_mib;
910
911 for_each_possible_cpu(i) {
912 struct ipstats_mib *af_inet6_stats;
913 af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i);
914 u64_stats_init(&af_inet6_stats->syncp);
915 }
916
917
918 net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib);
919 if (!net->mib.icmpv6_statistics)
920 goto err_icmp_mib;
921 net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
922 GFP_KERNEL);
923 if (!net->mib.icmpv6msg_statistics)
924 goto err_icmpmsg_mib;
925 return 0;
926
927 err_icmpmsg_mib:
928 free_percpu(net->mib.icmpv6_statistics);
929 err_icmp_mib:
930 free_percpu(net->mib.ipv6_statistics);
931 err_ip_mib:
932 free_percpu(net->mib.udplite_stats_in6);
933 err_udplite_mib:
934 free_percpu(net->mib.udp_stats_in6);
935 return -ENOMEM;
936 }
937
938 static void ipv6_cleanup_mibs(struct net *net)
939 {
940 free_percpu(net->mib.udp_stats_in6);
941 free_percpu(net->mib.udplite_stats_in6);
942 free_percpu(net->mib.ipv6_statistics);
943 free_percpu(net->mib.icmpv6_statistics);
944 kfree(net->mib.icmpv6msg_statistics);
945 }
946
947 static int __net_init inet6_net_init(struct net *net)
948 {
949 int err = 0;
950
951 net->ipv6.sysctl.bindv6only = 0;
952 net->ipv6.sysctl.icmpv6_time = 1*HZ;
953 net->ipv6.sysctl.icmpv6_echo_ignore_all = 0;
954 net->ipv6.sysctl.icmpv6_echo_ignore_multicast = 0;
955 net->ipv6.sysctl.icmpv6_echo_ignore_anycast = 0;
956
957 /* By default, rate limit error messages.
958 * Except for pmtu discovery, it would break it.
959 * proc_do_large_bitmap needs pointer to the bitmap.
960 */
961 bitmap_set(net->ipv6.sysctl.icmpv6_ratemask, 0, ICMPV6_ERRMSG_MAX + 1);
962 bitmap_clear(net->ipv6.sysctl.icmpv6_ratemask, ICMPV6_PKT_TOOBIG, 1);
963 net->ipv6.sysctl.icmpv6_ratemask_ptr = net->ipv6.sysctl.icmpv6_ratemask;
964
965 net->ipv6.sysctl.flowlabel_consistency = 1;
966 net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS;
967 net->ipv6.sysctl.idgen_retries = 3;
968 net->ipv6.sysctl.idgen_delay = 1 * HZ;
969 net->ipv6.sysctl.flowlabel_state_ranges = 0;
970 net->ipv6.sysctl.max_dst_opts_cnt = IP6_DEFAULT_MAX_DST_OPTS_CNT;
971 net->ipv6.sysctl.max_hbh_opts_cnt = IP6_DEFAULT_MAX_HBH_OPTS_CNT;
972 net->ipv6.sysctl.max_dst_opts_len = IP6_DEFAULT_MAX_DST_OPTS_LEN;
973 net->ipv6.sysctl.max_hbh_opts_len = IP6_DEFAULT_MAX_HBH_OPTS_LEN;
974 net->ipv6.sysctl.fib_notify_on_flag_change = 0;
975 atomic_set(&net->ipv6.fib6_sernum, 1);
976
977 net->ipv6.sysctl.ioam6_id = IOAM6_DEFAULT_ID;
978 net->ipv6.sysctl.ioam6_id_wide = IOAM6_DEFAULT_ID_WIDE;
979
980 err = ipv6_init_mibs(net);
981 if (err)
982 return err;
983 #ifdef CONFIG_PROC_FS
984 err = udp6_proc_init(net);
985 if (err)
986 goto out;
987 err = tcp6_proc_init(net);
988 if (err)
989 goto proc_tcp6_fail;
990 err = ac6_proc_init(net);
991 if (err)
992 goto proc_ac6_fail;
993 #endif
994 return err;
995
996 #ifdef CONFIG_PROC_FS
997 proc_ac6_fail:
998 tcp6_proc_exit(net);
999 proc_tcp6_fail:
1000 udp6_proc_exit(net);
1001 out:
1002 ipv6_cleanup_mibs(net);
1003 return err;
1004 #endif
1005 }
1006
1007 static void __net_exit inet6_net_exit(struct net *net)
1008 {
1009 #ifdef CONFIG_PROC_FS
1010 udp6_proc_exit(net);
1011 tcp6_proc_exit(net);
1012 ac6_proc_exit(net);
1013 #endif
1014 ipv6_cleanup_mibs(net);
1015 }
1016
1017 static struct pernet_operations inet6_net_ops = {
1018 .init = inet6_net_init,
1019 .exit = inet6_net_exit,
1020 };
1021
1022 static int ipv6_route_input(struct sk_buff *skb)
1023 {
1024 ip6_route_input(skb);
1025 return skb_dst(skb)->error;
1026 }
1027
1028 static const struct ipv6_stub ipv6_stub_impl = {
1029 .ipv6_sock_mc_join = ipv6_sock_mc_join,
1030 .ipv6_sock_mc_drop = ipv6_sock_mc_drop,
1031 .ipv6_dst_lookup_flow = ip6_dst_lookup_flow,
1032 .ipv6_route_input = ipv6_route_input,
1033 .fib6_get_table = fib6_get_table,
1034 .fib6_table_lookup = fib6_table_lookup,
1035 .fib6_lookup = fib6_lookup,
1036 .fib6_select_path = fib6_select_path,
1037 .ip6_mtu_from_fib6 = ip6_mtu_from_fib6,
1038 .fib6_nh_init = fib6_nh_init,
1039 .fib6_nh_release = fib6_nh_release,
1040 .fib6_nh_release_dsts = fib6_nh_release_dsts,
1041 .fib6_update_sernum = fib6_update_sernum_stub,
1042 .fib6_rt_update = fib6_rt_update,
1043 .ip6_del_rt = ip6_del_rt,
1044 .udpv6_encap_enable = udpv6_encap_enable,
1045 .ndisc_send_na = ndisc_send_na,
1046 #if IS_ENABLED(CONFIG_XFRM)
1047 .xfrm6_local_rxpmtu = xfrm6_local_rxpmtu,
1048 .xfrm6_udp_encap_rcv = xfrm6_udp_encap_rcv,
1049 .xfrm6_rcv_encap = xfrm6_rcv_encap,
1050 #endif
1051 .nd_tbl = &nd_tbl,
1052 .ipv6_fragment = ip6_fragment,
1053 .ipv6_dev_find = ipv6_dev_find,
1054 };
1055
1056 static const struct ipv6_bpf_stub ipv6_bpf_stub_impl = {
1057 .inet6_bind = __inet6_bind,
1058 .udp6_lib_lookup = __udp6_lib_lookup,
1059 };
1060
1061 static int __init inet6_init(void)
1062 {
1063 struct list_head *r;
1064 int err = 0;
1065
1066 sock_skb_cb_check_size(sizeof(struct inet6_skb_parm));
1067
1068 /* Register the socket-side information for inet6_create. */
1069 for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
1070 INIT_LIST_HEAD(r);
1071
1072 if (disable_ipv6_mod) {
1073 pr_info("Loaded, but administratively disabled, reboot required to enable\n");
1074 goto out;
1075 }
1076
1077 err = proto_register(&tcpv6_prot, 1);
1078 if (err)
1079 goto out;
1080
1081 err = proto_register(&udpv6_prot, 1);
1082 if (err)
1083 goto out_unregister_tcp_proto;
1084
1085 err = proto_register(&udplitev6_prot, 1);
1086 if (err)
1087 goto out_unregister_udp_proto;
1088
1089 err = proto_register(&rawv6_prot, 1);
1090 if (err)
1091 goto out_unregister_udplite_proto;
1092
1093 err = proto_register(&pingv6_prot, 1);
1094 if (err)
1095 goto out_unregister_raw_proto;
1096
1097 /* We MUST register RAW sockets before we create the ICMP6,
1098 * IGMP6, or NDISC control sockets.
1099 */
1100 err = rawv6_init();
1101 if (err)
1102 goto out_unregister_ping_proto;
1103
1104 /* Register the family here so that the init calls below will
1105 * be able to create sockets. (?? is this dangerous ??)
1106 */
1107 err = sock_register(&inet6_family_ops);
1108 if (err)
1109 goto out_sock_register_fail;
1110
1111 /*
1112 * ipngwg API draft makes clear that the correct semantics
1113 * for TCP and UDP is to consider one TCP and UDP instance
1114 * in a host available by both INET and INET6 APIs and
1115 * able to communicate via both network protocols.
1116 */
1117
1118 err = register_pernet_subsys(&inet6_net_ops);
1119 if (err)
1120 goto register_pernet_fail;
1121 err = ip6_mr_init();
1122 if (err)
1123 goto ipmr_fail;
1124 err = icmpv6_init();
1125 if (err)
1126 goto icmp_fail;
1127 err = ndisc_init();
1128 if (err)
1129 goto ndisc_fail;
1130 err = igmp6_init();
1131 if (err)
1132 goto igmp_fail;
1133
1134 err = ipv6_netfilter_init();
1135 if (err)
1136 goto netfilter_fail;
1137 /* Create /proc/foo6 entries. */
1138 #ifdef CONFIG_PROC_FS
1139 err = -ENOMEM;
1140 if (raw6_proc_init())
1141 goto proc_raw6_fail;
1142 if (udplite6_proc_init())
1143 goto proc_udplite6_fail;
1144 if (ipv6_misc_proc_init())
1145 goto proc_misc6_fail;
1146 if (if6_proc_init())
1147 goto proc_if6_fail;
1148 #endif
1149 err = ip6_route_init();
1150 if (err)
1151 goto ip6_route_fail;
1152 err = ndisc_late_init();
1153 if (err)
1154 goto ndisc_late_fail;
1155 err = ip6_flowlabel_init();
1156 if (err)
1157 goto ip6_flowlabel_fail;
1158 err = ipv6_anycast_init();
1159 if (err)
1160 goto ipv6_anycast_fail;
1161 err = addrconf_init();
1162 if (err)
1163 goto addrconf_fail;
1164
1165 /* Init v6 extension headers. */
1166 err = ipv6_exthdrs_init();
1167 if (err)
1168 goto ipv6_exthdrs_fail;
1169
1170 err = ipv6_frag_init();
1171 if (err)
1172 goto ipv6_frag_fail;
1173
1174 /* Init v6 transport protocols. */
1175 err = udpv6_init();
1176 if (err)
1177 goto udpv6_fail;
1178
1179 err = udplitev6_init();
1180 if (err)
1181 goto udplitev6_fail;
1182
1183 err = udpv6_offload_init();
1184 if (err)
1185 goto udpv6_offload_fail;
1186
1187 err = tcpv6_init();
1188 if (err)
1189 goto tcpv6_fail;
1190
1191 err = ipv6_packet_init();
1192 if (err)
1193 goto ipv6_packet_fail;
1194
1195 err = pingv6_init();
1196 if (err)
1197 goto pingv6_fail;
1198
1199 err = calipso_init();
1200 if (err)
1201 goto calipso_fail;
1202
1203 err = seg6_init();
1204 if (err)
1205 goto seg6_fail;
1206
1207 err = rpl_init();
1208 if (err)
1209 goto rpl_fail;
1210
1211 err = ioam6_init();
1212 if (err)
1213 goto ioam6_fail;
1214
1215 err = igmp6_late_init();
1216 if (err)
1217 goto igmp6_late_err;
1218
1219 #ifdef CONFIG_SYSCTL
1220 err = ipv6_sysctl_register();
1221 if (err)
1222 goto sysctl_fail;
1223 #endif
1224
1225 /* ensure that ipv6 stubs are visible only after ipv6 is ready */
1226 wmb();
1227 ipv6_stub = &ipv6_stub_impl;
1228 ipv6_bpf_stub = &ipv6_bpf_stub_impl;
1229 out:
1230 return err;
1231
1232 #ifdef CONFIG_SYSCTL
1233 sysctl_fail:
1234 igmp6_late_cleanup();
1235 #endif
1236 igmp6_late_err:
1237 ioam6_exit();
1238 ioam6_fail:
1239 rpl_exit();
1240 rpl_fail:
1241 seg6_exit();
1242 seg6_fail:
1243 calipso_exit();
1244 calipso_fail:
1245 pingv6_exit();
1246 pingv6_fail:
1247 ipv6_packet_cleanup();
1248 ipv6_packet_fail:
1249 tcpv6_exit();
1250 tcpv6_fail:
1251 udpv6_offload_exit();
1252 udpv6_offload_fail:
1253 udplitev6_exit();
1254 udplitev6_fail:
1255 udpv6_exit();
1256 udpv6_fail:
1257 ipv6_frag_exit();
1258 ipv6_frag_fail:
1259 ipv6_exthdrs_exit();
1260 ipv6_exthdrs_fail:
1261 addrconf_cleanup();
1262 addrconf_fail:
1263 ipv6_anycast_cleanup();
1264 ipv6_anycast_fail:
1265 ip6_flowlabel_cleanup();
1266 ip6_flowlabel_fail:
1267 ndisc_late_cleanup();
1268 ndisc_late_fail:
1269 ip6_route_cleanup();
1270 ip6_route_fail:
1271 #ifdef CONFIG_PROC_FS
1272 if6_proc_exit();
1273 proc_if6_fail:
1274 ipv6_misc_proc_exit();
1275 proc_misc6_fail:
1276 udplite6_proc_exit();
1277 proc_udplite6_fail:
1278 raw6_proc_exit();
1279 proc_raw6_fail:
1280 #endif
1281 ipv6_netfilter_fini();
1282 netfilter_fail:
1283 igmp6_cleanup();
1284 igmp_fail:
1285 ndisc_cleanup();
1286 ndisc_fail:
1287 icmpv6_cleanup();
1288 icmp_fail:
1289 ip6_mr_cleanup();
1290 ipmr_fail:
1291 unregister_pernet_subsys(&inet6_net_ops);
1292 register_pernet_fail:
1293 sock_unregister(PF_INET6);
1294 rtnl_unregister_all(PF_INET6);
1295 out_sock_register_fail:
1296 rawv6_exit();
1297 out_unregister_ping_proto:
1298 proto_unregister(&pingv6_prot);
1299 out_unregister_raw_proto:
1300 proto_unregister(&rawv6_prot);
1301 out_unregister_udplite_proto:
1302 proto_unregister(&udplitev6_prot);
1303 out_unregister_udp_proto:
1304 proto_unregister(&udpv6_prot);
1305 out_unregister_tcp_proto:
1306 proto_unregister(&tcpv6_prot);
1307 goto out;
1308 }
1309 module_init(inet6_init);
1310
1311 MODULE_ALIAS_NETPROTO(PF_INET6);
1312
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.