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
63 #include <linux/uaccess.h>
64 #include <linux/mroute6.h>
65
66 #include "ip6_offload.h"
67
68 MODULE_AUTHOR("Cast of dozens");
69 MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
70 MODULE_LICENSE("GPL");
71
72 /* The inetsw6 table contains everything that inet6_create needs to
73 * build a new socket.
74 */
75 static struct list_head inetsw6[SOCK_MAX];
76 static DEFINE_SPINLOCK(inetsw6_lock);
77
78 struct ipv6_params ipv6_defaults = {
79 .disable_ipv6 = 0,
80 .autoconf = 1,
81 };
82
83 static int disable_ipv6_mod;
84
85 module_param_named(disable, disable_ipv6_mod, int, 0444);
86 MODULE_PARM_DESC(disable, "Disable IPv6 module such that it is non-functional");
87
88 module_param_named(disable_ipv6, ipv6_defaults.disable_ipv6, int, 0444);
89 MODULE_PARM_DESC(disable_ipv6, "Disable IPv6 on all interfaces");
90
91 module_param_named(autoconf, ipv6_defaults.autoconf, int, 0444);
92 MODULE_PARM_DESC(autoconf, "Enable IPv6 address autoconfiguration on all interfaces");
93
ipv6_mod_enabled(void)94 bool ipv6_mod_enabled(void)
95 {
96 return disable_ipv6_mod == 0;
97 }
98 EXPORT_SYMBOL_GPL(ipv6_mod_enabled);
99
inet6_sk_generic(struct sock * sk)100 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
101 {
102 const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
103
104 return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
105 }
106
inet6_create(struct net * net,struct socket * sock,int protocol,int kern)107 static int inet6_create(struct net *net, struct socket *sock, int protocol,
108 int kern)
109 {
110 struct inet_sock *inet;
111 struct ipv6_pinfo *np;
112 struct sock *sk;
113 struct inet_protosw *answer;
114 struct proto *answer_prot;
115 unsigned char answer_flags;
116 int try_loading_module = 0;
117 int err;
118
119 if (protocol < 0 || protocol >= IPPROTO_MAX)
120 return -EINVAL;
121
122 /* Look for the requested type/protocol pair. */
123 lookup_protocol:
124 err = -ESOCKTNOSUPPORT;
125 rcu_read_lock();
126 list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
127
128 err = 0;
129 /* Check the non-wild match. */
130 if (protocol == answer->protocol) {
131 if (protocol != IPPROTO_IP)
132 break;
133 } else {
134 /* Check for the two wild cases. */
135 if (IPPROTO_IP == protocol) {
136 protocol = answer->protocol;
137 break;
138 }
139 if (IPPROTO_IP == answer->protocol)
140 break;
141 }
142 err = -EPROTONOSUPPORT;
143 }
144
145 if (err) {
146 if (try_loading_module < 2) {
147 rcu_read_unlock();
148 /*
149 * Be more specific, e.g. net-pf-10-proto-132-type-1
150 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
151 */
152 if (++try_loading_module == 1)
153 request_module("net-pf-%d-proto-%d-type-%d",
154 PF_INET6, protocol, sock->type);
155 /*
156 * Fall back to generic, e.g. net-pf-10-proto-132
157 * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
158 */
159 else
160 request_module("net-pf-%d-proto-%d",
161 PF_INET6, protocol);
162 goto lookup_protocol;
163 } else
164 goto out_rcu_unlock;
165 }
166
167 err = -EPERM;
168 if (sock->type == SOCK_RAW && !kern &&
169 !ns_capable(net->user_ns, CAP_NET_RAW))
170 goto out_rcu_unlock;
171
172 sock->ops = answer->ops;
173 answer_prot = answer->prot;
174 answer_flags = answer->flags;
175 rcu_read_unlock();
176
177 WARN_ON(!answer_prot->slab);
178
179 err = -ENOBUFS;
180 sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot, kern);
181 if (!sk)
182 goto out;
183
184 sock_init_data(sock, sk);
185
186 err = 0;
187 if (INET_PROTOSW_REUSE & answer_flags)
188 sk->sk_reuse = SK_CAN_REUSE;
189
190 inet = inet_sk(sk);
191 inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
192
193 if (SOCK_RAW == sock->type) {
194 inet->inet_num = protocol;
195 if (IPPROTO_RAW == protocol)
196 inet->hdrincl = 1;
197 }
198
199 sk->sk_destruct = inet_sock_destruct;
200 sk->sk_family = PF_INET6;
201 sk->sk_protocol = protocol;
202
203 sk->sk_backlog_rcv = answer->prot->backlog_rcv;
204
205 inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
206 np->hop_limit = -1;
207 np->mcast_hops = IPV6_DEFAULT_MCASTHOPS;
208 np->mc_loop = 1;
209 np->mc_all = 1;
210 np->pmtudisc = IPV6_PMTUDISC_WANT;
211 np->repflow = net->ipv6.sysctl.flowlabel_reflect & FLOWLABEL_REFLECT_ESTABLISHED;
212 sk->sk_ipv6only = net->ipv6.sysctl.bindv6only;
213
214 /* Init the ipv4 part of the socket since we can have sockets
215 * using v6 API for ipv4.
216 */
217 inet->uc_ttl = -1;
218
219 inet->mc_loop = 1;
220 inet->mc_ttl = 1;
221 inet->mc_index = 0;
222 inet->mc_list = NULL;
223 inet->rcv_tos = 0;
224
225 if (net->ipv4.sysctl_ip_no_pmtu_disc)
226 inet->pmtudisc = IP_PMTUDISC_DONT;
227 else
228 inet->pmtudisc = IP_PMTUDISC_WANT;
229 /*
230 * Increment only the relevant sk_prot->socks debug field, this changes
231 * the previous behaviour of incrementing both the equivalent to
232 * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
233 *
234 * This allows better debug granularity as we'll know exactly how many
235 * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
236 * transport protocol socks. -acme
237 */
238 sk_refcnt_debug_inc(sk);
239
240 if (inet->inet_num) {
241 /* It assumes that any protocol which allows
242 * the user to assign a number at socket
243 * creation time automatically shares.
244 */
245 inet->inet_sport = htons(inet->inet_num);
246 err = sk->sk_prot->hash(sk);
247 if (err) {
248 sk_common_release(sk);
249 goto out;
250 }
251 }
252 if (sk->sk_prot->init) {
253 err = sk->sk_prot->init(sk);
254 if (err) {
255 sk_common_release(sk);
256 goto out;
257 }
258 }
259
260 if (!kern) {
261 err = BPF_CGROUP_RUN_PROG_INET_SOCK(sk);
262 if (err) {
263 sk_common_release(sk);
264 goto out;
265 }
266 }
267 out:
268 return err;
269 out_rcu_unlock:
270 rcu_read_unlock();
271 goto out;
272 }
273
__inet6_bind(struct sock * sk,struct sockaddr * uaddr,int addr_len,bool force_bind_address_no_port,bool with_lock)274 static int __inet6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len,
275 bool force_bind_address_no_port, bool with_lock)
276 {
277 struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr;
278 struct inet_sock *inet = inet_sk(sk);
279 struct ipv6_pinfo *np = inet6_sk(sk);
280 struct net *net = sock_net(sk);
281 __be32 v4addr = 0;
282 unsigned short snum;
283 bool saved_ipv6only;
284 int addr_type = 0;
285 int err = 0;
286
287 if (addr->sin6_family != AF_INET6)
288 return -EAFNOSUPPORT;
289
290 addr_type = ipv6_addr_type(&addr->sin6_addr);
291 if ((addr_type & IPV6_ADDR_MULTICAST) && sk->sk_type == SOCK_STREAM)
292 return -EINVAL;
293
294 snum = ntohs(addr->sin6_port);
295 if (snum && inet_is_local_unbindable_port(net, snum))
296 return -EPERM;
297 if (snum && snum < inet_prot_sock(net) &&
298 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
299 return -EACCES;
300
301 if (with_lock)
302 lock_sock(sk);
303
304 /* Check these errors (active socket, double bind). */
305 if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
306 err = -EINVAL;
307 goto out;
308 }
309
310 /* Check if the address belongs to the host. */
311 if (addr_type == IPV6_ADDR_MAPPED) {
312 struct net_device *dev = NULL;
313 int chk_addr_ret;
314
315 /* Binding to v4-mapped address on a v6-only socket
316 * makes no sense
317 */
318 if (sk->sk_ipv6only) {
319 err = -EINVAL;
320 goto out;
321 }
322
323 rcu_read_lock();
324 if (sk->sk_bound_dev_if) {
325 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
326 if (!dev) {
327 err = -ENODEV;
328 goto out_unlock;
329 }
330 }
331
332 /* Reproduce AF_INET checks to make the bindings consistent */
333 v4addr = addr->sin6_addr.s6_addr32[3];
334 chk_addr_ret = inet_addr_type_dev_table(net, dev, v4addr);
335 rcu_read_unlock();
336
337 if (!inet_can_nonlocal_bind(net, inet) &&
338 v4addr != htonl(INADDR_ANY) &&
339 chk_addr_ret != RTN_LOCAL &&
340 chk_addr_ret != RTN_MULTICAST &&
341 chk_addr_ret != RTN_BROADCAST) {
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 force_bind_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 err = BPF_CGROUP_RUN_PROG_INET6_POST_BIND(sk);
412 if (err) {
413 sk->sk_ipv6only = saved_ipv6only;
414 inet_reset_saddr(sk);
415 goto out;
416 }
417 }
418
419 if (addr_type != IPV6_ADDR_ANY)
420 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
421 if (snum)
422 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
423 inet->inet_sport = htons(inet->inet_num);
424 inet->inet_dport = 0;
425 inet->inet_daddr = 0;
426 out:
427 if (with_lock)
428 release_sock(sk);
429 return err;
430 out_unlock:
431 rcu_read_unlock();
432 goto out;
433 }
434
435 /* bind for INET6 API */
inet6_bind(struct socket * sock,struct sockaddr * uaddr,int addr_len)436 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
437 {
438 struct sock *sk = sock->sk;
439 int err = 0;
440
441 /* If the socket has its own bind function then use it. */
442 if (sk->sk_prot->bind)
443 return sk->sk_prot->bind(sk, uaddr, addr_len);
444
445 if (addr_len < SIN6_LEN_RFC2133)
446 return -EINVAL;
447
448 /* BPF prog is run before any checks are done so that if the prog
449 * changes context in a wrong way it will be caught.
450 */
451 err = BPF_CGROUP_RUN_PROG_INET6_BIND(sk, uaddr);
452 if (err)
453 return err;
454
455 return __inet6_bind(sk, uaddr, addr_len, false, true);
456 }
457 EXPORT_SYMBOL(inet6_bind);
458
inet6_release(struct socket * sock)459 int inet6_release(struct socket *sock)
460 {
461 struct sock *sk = sock->sk;
462
463 if (!sk)
464 return -EINVAL;
465
466 /* Free mc lists */
467 ipv6_sock_mc_close(sk);
468
469 /* Free ac lists */
470 ipv6_sock_ac_close(sk);
471
472 return inet_release(sock);
473 }
474 EXPORT_SYMBOL(inet6_release);
475
inet6_destroy_sock(struct sock * sk)476 void inet6_destroy_sock(struct sock *sk)
477 {
478 struct ipv6_pinfo *np = inet6_sk(sk);
479 struct sk_buff *skb;
480 struct ipv6_txoptions *opt;
481
482 /* Release rx options */
483
484 skb = xchg(&np->pktoptions, NULL);
485 kfree_skb(skb);
486
487 skb = xchg(&np->rxpmtu, NULL);
488 kfree_skb(skb);
489
490 /* Free flowlabels */
491 fl6_free_socklist(sk);
492
493 /* Free tx options */
494
495 opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL);
496 if (opt) {
497 atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
498 txopt_put(opt);
499 }
500 }
501 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
502
503 /*
504 * This does both peername and sockname.
505 */
506
inet6_getname(struct socket * sock,struct sockaddr * uaddr,int peer)507 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
508 int peer)
509 {
510 struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
511 struct sock *sk = sock->sk;
512 struct inet_sock *inet = inet_sk(sk);
513 struct ipv6_pinfo *np = inet6_sk(sk);
514
515 sin->sin6_family = AF_INET6;
516 sin->sin6_flowinfo = 0;
517 sin->sin6_scope_id = 0;
518 if (peer) {
519 if (!inet->inet_dport)
520 return -ENOTCONN;
521 if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
522 peer == 1)
523 return -ENOTCONN;
524 sin->sin6_port = inet->inet_dport;
525 sin->sin6_addr = sk->sk_v6_daddr;
526 if (np->sndflow)
527 sin->sin6_flowinfo = np->flow_label;
528 } else {
529 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
530 sin->sin6_addr = np->saddr;
531 else
532 sin->sin6_addr = sk->sk_v6_rcv_saddr;
533
534 sin->sin6_port = inet->inet_sport;
535 }
536 sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
537 sk->sk_bound_dev_if);
538 return sizeof(*sin);
539 }
540 EXPORT_SYMBOL(inet6_getname);
541
inet6_ioctl(struct socket * sock,unsigned int cmd,unsigned long arg)542 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
543 {
544 struct sock *sk = sock->sk;
545 struct net *net = sock_net(sk);
546
547 switch (cmd) {
548 case SIOCADDRT:
549 case SIOCDELRT:
550
551 return ipv6_route_ioctl(net, cmd, (void __user *)arg);
552
553 case SIOCSIFADDR:
554 return addrconf_add_ifaddr(net, (void __user *) arg);
555 case SIOCDIFADDR:
556 return addrconf_del_ifaddr(net, (void __user *) arg);
557 case SIOCSIFDSTADDR:
558 return addrconf_set_dstaddr(net, (void __user *) arg);
559 default:
560 if (!sk->sk_prot->ioctl)
561 return -ENOIOCTLCMD;
562 return sk->sk_prot->ioctl(sk, cmd, arg);
563 }
564 /*NOTREACHED*/
565 return 0;
566 }
567 EXPORT_SYMBOL(inet6_ioctl);
568
569 INDIRECT_CALLABLE_DECLARE(int udpv6_sendmsg(struct sock *, struct msghdr *,
570 size_t));
inet6_sendmsg(struct socket * sock,struct msghdr * msg,size_t size)571 int inet6_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
572 {
573 struct sock *sk = sock->sk;
574
575 if (unlikely(inet_send_prepare(sk)))
576 return -EAGAIN;
577
578 return INDIRECT_CALL_2(sk->sk_prot->sendmsg, tcp_sendmsg, udpv6_sendmsg,
579 sk, msg, size);
580 }
581
582 INDIRECT_CALLABLE_DECLARE(int udpv6_recvmsg(struct sock *, struct msghdr *,
583 size_t, int, int, int *));
inet6_recvmsg(struct socket * sock,struct msghdr * msg,size_t size,int flags)584 int inet6_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
585 int flags)
586 {
587 struct sock *sk = sock->sk;
588 int addr_len = 0;
589 int err;
590
591 if (likely(!(flags & MSG_ERRQUEUE)))
592 sock_rps_record_flow(sk);
593
594 err = INDIRECT_CALL_2(sk->sk_prot->recvmsg, tcp_recvmsg, udpv6_recvmsg,
595 sk, msg, size, flags & MSG_DONTWAIT,
596 flags & ~MSG_DONTWAIT, &addr_len);
597 if (err >= 0)
598 msg->msg_namelen = addr_len;
599 return err;
600 }
601
602 const struct proto_ops inet6_stream_ops = {
603 .family = PF_INET6,
604 .owner = THIS_MODULE,
605 .release = inet6_release,
606 .bind = inet6_bind,
607 .connect = inet_stream_connect, /* ok */
608 .socketpair = sock_no_socketpair, /* a do nothing */
609 .accept = inet_accept, /* ok */
610 .getname = inet6_getname,
611 .poll = tcp_poll, /* ok */
612 .ioctl = inet6_ioctl, /* must change */
613 .gettstamp = sock_gettstamp,
614 .listen = inet_listen, /* ok */
615 .shutdown = inet_shutdown, /* ok */
616 .setsockopt = sock_common_setsockopt, /* ok */
617 .getsockopt = sock_common_getsockopt, /* ok */
618 .sendmsg = inet6_sendmsg, /* retpoline's sake */
619 .recvmsg = inet6_recvmsg, /* retpoline's sake */
620 #ifdef CONFIG_MMU
621 .mmap = tcp_mmap,
622 #endif
623 .sendpage = inet_sendpage,
624 .sendmsg_locked = tcp_sendmsg_locked,
625 .sendpage_locked = tcp_sendpage_locked,
626 .splice_read = tcp_splice_read,
627 .read_sock = tcp_read_sock,
628 .peek_len = tcp_peek_len,
629 #ifdef CONFIG_COMPAT
630 .compat_setsockopt = compat_sock_common_setsockopt,
631 .compat_getsockopt = compat_sock_common_getsockopt,
632 #endif
633 .set_rcvlowat = tcp_set_rcvlowat,
634 };
635
636 const struct proto_ops inet6_dgram_ops = {
637 .family = PF_INET6,
638 .owner = THIS_MODULE,
639 .release = inet6_release,
640 .bind = inet6_bind,
641 .connect = inet_dgram_connect, /* ok */
642 .socketpair = sock_no_socketpair, /* a do nothing */
643 .accept = sock_no_accept, /* a do nothing */
644 .getname = inet6_getname,
645 .poll = udp_poll, /* ok */
646 .ioctl = inet6_ioctl, /* must change */
647 .gettstamp = sock_gettstamp,
648 .listen = sock_no_listen, /* ok */
649 .shutdown = inet_shutdown, /* ok */
650 .setsockopt = sock_common_setsockopt, /* ok */
651 .getsockopt = sock_common_getsockopt, /* ok */
652 .sendmsg = inet6_sendmsg, /* retpoline's sake */
653 .recvmsg = inet6_recvmsg, /* retpoline's sake */
654 .mmap = sock_no_mmap,
655 .sendpage = sock_no_sendpage,
656 .set_peek_off = sk_set_peek_off,
657 #ifdef CONFIG_COMPAT
658 .compat_setsockopt = compat_sock_common_setsockopt,
659 .compat_getsockopt = compat_sock_common_getsockopt,
660 #endif
661 };
662
663 static const struct net_proto_family inet6_family_ops = {
664 .family = PF_INET6,
665 .create = inet6_create,
666 .owner = THIS_MODULE,
667 };
668
inet6_register_protosw(struct inet_protosw * p)669 int inet6_register_protosw(struct inet_protosw *p)
670 {
671 struct list_head *lh;
672 struct inet_protosw *answer;
673 struct list_head *last_perm;
674 int protocol = p->protocol;
675 int ret;
676
677 spin_lock_bh(&inetsw6_lock);
678
679 ret = -EINVAL;
680 if (p->type >= SOCK_MAX)
681 goto out_illegal;
682
683 /* If we are trying to override a permanent protocol, bail. */
684 answer = NULL;
685 ret = -EPERM;
686 last_perm = &inetsw6[p->type];
687 list_for_each(lh, &inetsw6[p->type]) {
688 answer = list_entry(lh, struct inet_protosw, list);
689
690 /* Check only the non-wild match. */
691 if (INET_PROTOSW_PERMANENT & answer->flags) {
692 if (protocol == answer->protocol)
693 break;
694 last_perm = lh;
695 }
696
697 answer = NULL;
698 }
699 if (answer)
700 goto out_permanent;
701
702 /* Add the new entry after the last permanent entry if any, so that
703 * the new entry does not override a permanent entry when matched with
704 * a wild-card protocol. But it is allowed to override any existing
705 * non-permanent entry. This means that when we remove this entry, the
706 * system automatically returns to the old behavior.
707 */
708 list_add_rcu(&p->list, last_perm);
709 ret = 0;
710 out:
711 spin_unlock_bh(&inetsw6_lock);
712 return ret;
713
714 out_permanent:
715 pr_err("Attempt to override permanent protocol %d\n", protocol);
716 goto out;
717
718 out_illegal:
719 pr_err("Ignoring attempt to register invalid socket type %d\n",
720 p->type);
721 goto out;
722 }
723 EXPORT_SYMBOL(inet6_register_protosw);
724
725 void
inet6_unregister_protosw(struct inet_protosw * p)726 inet6_unregister_protosw(struct inet_protosw *p)
727 {
728 if (INET_PROTOSW_PERMANENT & p->flags) {
729 pr_err("Attempt to unregister permanent protocol %d\n",
730 p->protocol);
731 } else {
732 spin_lock_bh(&inetsw6_lock);
733 list_del_rcu(&p->list);
734 spin_unlock_bh(&inetsw6_lock);
735
736 synchronize_net();
737 }
738 }
739 EXPORT_SYMBOL(inet6_unregister_protosw);
740
inet6_sk_rebuild_header(struct sock * sk)741 int inet6_sk_rebuild_header(struct sock *sk)
742 {
743 struct ipv6_pinfo *np = inet6_sk(sk);
744 struct dst_entry *dst;
745
746 dst = __sk_dst_check(sk, np->dst_cookie);
747
748 if (!dst) {
749 struct inet_sock *inet = inet_sk(sk);
750 struct in6_addr *final_p, final;
751 struct flowi6 fl6;
752
753 memset(&fl6, 0, sizeof(fl6));
754 fl6.flowi6_proto = sk->sk_protocol;
755 fl6.daddr = sk->sk_v6_daddr;
756 fl6.saddr = np->saddr;
757 fl6.flowlabel = np->flow_label;
758 fl6.flowi6_oif = sk->sk_bound_dev_if;
759 fl6.flowi6_mark = sk->sk_mark;
760 fl6.fl6_dport = inet->inet_dport;
761 fl6.fl6_sport = inet->inet_sport;
762 fl6.flowi6_uid = sk->sk_uid;
763 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
764
765 rcu_read_lock();
766 final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt),
767 &final);
768 rcu_read_unlock();
769
770 dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p);
771 if (IS_ERR(dst)) {
772 sk->sk_route_caps = 0;
773 sk->sk_err_soft = -PTR_ERR(dst);
774 return PTR_ERR(dst);
775 }
776
777 ip6_dst_store(sk, dst, NULL, NULL);
778 }
779
780 return 0;
781 }
782 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
783
ipv6_opt_accepted(const struct sock * sk,const struct sk_buff * skb,const struct inet6_skb_parm * opt)784 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb,
785 const struct inet6_skb_parm *opt)
786 {
787 const struct ipv6_pinfo *np = inet6_sk(sk);
788
789 if (np->rxopt.all) {
790 if (((opt->flags & IP6SKB_HOPBYHOP) &&
791 (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) ||
792 (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) &&
793 np->rxopt.bits.rxflow) ||
794 (opt->srcrt && (np->rxopt.bits.srcrt ||
795 np->rxopt.bits.osrcrt)) ||
796 ((opt->dst1 || opt->dst0) &&
797 (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
798 return true;
799 }
800 return false;
801 }
802 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
803
804 static struct packet_type ipv6_packet_type __read_mostly = {
805 .type = cpu_to_be16(ETH_P_IPV6),
806 .func = ipv6_rcv,
807 .list_func = ipv6_list_rcv,
808 };
809
ipv6_packet_init(void)810 static int __init ipv6_packet_init(void)
811 {
812 dev_add_pack(&ipv6_packet_type);
813 return 0;
814 }
815
ipv6_packet_cleanup(void)816 static void ipv6_packet_cleanup(void)
817 {
818 dev_remove_pack(&ipv6_packet_type);
819 }
820
ipv6_init_mibs(struct net * net)821 static int __net_init ipv6_init_mibs(struct net *net)
822 {
823 int i;
824
825 net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib);
826 if (!net->mib.udp_stats_in6)
827 return -ENOMEM;
828 net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib);
829 if (!net->mib.udplite_stats_in6)
830 goto err_udplite_mib;
831 net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib);
832 if (!net->mib.ipv6_statistics)
833 goto err_ip_mib;
834
835 for_each_possible_cpu(i) {
836 struct ipstats_mib *af_inet6_stats;
837 af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i);
838 u64_stats_init(&af_inet6_stats->syncp);
839 }
840
841
842 net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib);
843 if (!net->mib.icmpv6_statistics)
844 goto err_icmp_mib;
845 net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
846 GFP_KERNEL);
847 if (!net->mib.icmpv6msg_statistics)
848 goto err_icmpmsg_mib;
849 return 0;
850
851 err_icmpmsg_mib:
852 free_percpu(net->mib.icmpv6_statistics);
853 err_icmp_mib:
854 free_percpu(net->mib.ipv6_statistics);
855 err_ip_mib:
856 free_percpu(net->mib.udplite_stats_in6);
857 err_udplite_mib:
858 free_percpu(net->mib.udp_stats_in6);
859 return -ENOMEM;
860 }
861
ipv6_cleanup_mibs(struct net * net)862 static void ipv6_cleanup_mibs(struct net *net)
863 {
864 free_percpu(net->mib.udp_stats_in6);
865 free_percpu(net->mib.udplite_stats_in6);
866 free_percpu(net->mib.ipv6_statistics);
867 free_percpu(net->mib.icmpv6_statistics);
868 kfree(net->mib.icmpv6msg_statistics);
869 }
870
inet6_net_init(struct net * net)871 static int __net_init inet6_net_init(struct net *net)
872 {
873 int err = 0;
874
875 net->ipv6.sysctl.bindv6only = 0;
876 net->ipv6.sysctl.icmpv6_time = 1*HZ;
877 net->ipv6.sysctl.icmpv6_echo_ignore_all = 0;
878 net->ipv6.sysctl.icmpv6_echo_ignore_multicast = 0;
879 net->ipv6.sysctl.icmpv6_echo_ignore_anycast = 0;
880
881 /* By default, rate limit error messages.
882 * Except for pmtu discovery, it would break it.
883 * proc_do_large_bitmap needs pointer to the bitmap.
884 */
885 bitmap_set(net->ipv6.sysctl.icmpv6_ratemask, 0, ICMPV6_ERRMSG_MAX + 1);
886 bitmap_clear(net->ipv6.sysctl.icmpv6_ratemask, ICMPV6_PKT_TOOBIG, 1);
887 net->ipv6.sysctl.icmpv6_ratemask_ptr = net->ipv6.sysctl.icmpv6_ratemask;
888
889 net->ipv6.sysctl.flowlabel_consistency = 1;
890 net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS;
891 net->ipv6.sysctl.idgen_retries = 3;
892 net->ipv6.sysctl.idgen_delay = 1 * HZ;
893 net->ipv6.sysctl.flowlabel_state_ranges = 0;
894 net->ipv6.sysctl.max_dst_opts_cnt = IP6_DEFAULT_MAX_DST_OPTS_CNT;
895 net->ipv6.sysctl.max_hbh_opts_cnt = IP6_DEFAULT_MAX_HBH_OPTS_CNT;
896 net->ipv6.sysctl.max_dst_opts_len = IP6_DEFAULT_MAX_DST_OPTS_LEN;
897 net->ipv6.sysctl.max_hbh_opts_len = IP6_DEFAULT_MAX_HBH_OPTS_LEN;
898 atomic_set(&net->ipv6.fib6_sernum, 1);
899
900 err = ipv6_init_mibs(net);
901 if (err)
902 return err;
903 #ifdef CONFIG_PROC_FS
904 err = udp6_proc_init(net);
905 if (err)
906 goto out;
907 err = tcp6_proc_init(net);
908 if (err)
909 goto proc_tcp6_fail;
910 err = ac6_proc_init(net);
911 if (err)
912 goto proc_ac6_fail;
913 #endif
914 return err;
915
916 #ifdef CONFIG_PROC_FS
917 proc_ac6_fail:
918 tcp6_proc_exit(net);
919 proc_tcp6_fail:
920 udp6_proc_exit(net);
921 out:
922 ipv6_cleanup_mibs(net);
923 return err;
924 #endif
925 }
926
inet6_net_exit(struct net * net)927 static void __net_exit inet6_net_exit(struct net *net)
928 {
929 #ifdef CONFIG_PROC_FS
930 udp6_proc_exit(net);
931 tcp6_proc_exit(net);
932 ac6_proc_exit(net);
933 #endif
934 ipv6_cleanup_mibs(net);
935 }
936
937 static struct pernet_operations inet6_net_ops = {
938 .init = inet6_net_init,
939 .exit = inet6_net_exit,
940 };
941
ipv6_route_input(struct sk_buff * skb)942 static int ipv6_route_input(struct sk_buff *skb)
943 {
944 ip6_route_input(skb);
945 return skb_dst(skb)->error;
946 }
947
948 static const struct ipv6_stub ipv6_stub_impl = {
949 .ipv6_sock_mc_join = ipv6_sock_mc_join,
950 .ipv6_sock_mc_drop = ipv6_sock_mc_drop,
951 .ipv6_dst_lookup_flow = ip6_dst_lookup_flow,
952 .ipv6_route_input = ipv6_route_input,
953 .fib6_get_table = fib6_get_table,
954 .fib6_table_lookup = fib6_table_lookup,
955 .fib6_lookup = fib6_lookup,
956 .fib6_select_path = fib6_select_path,
957 .ip6_mtu_from_fib6 = ip6_mtu_from_fib6,
958 .fib6_nh_init = fib6_nh_init,
959 .fib6_nh_release = fib6_nh_release,
960 .fib6_update_sernum = fib6_update_sernum_stub,
961 .fib6_rt_update = fib6_rt_update,
962 .ip6_del_rt = ip6_del_rt,
963 .udpv6_encap_enable = udpv6_encap_enable,
964 .ndisc_send_na = ndisc_send_na,
965 .nd_tbl = &nd_tbl,
966 };
967
968 static const struct ipv6_bpf_stub ipv6_bpf_stub_impl = {
969 .inet6_bind = __inet6_bind,
970 .udp6_lib_lookup = __udp6_lib_lookup,
971 };
972
inet6_init(void)973 static int __init inet6_init(void)
974 {
975 struct list_head *r;
976 int err = 0;
977
978 sock_skb_cb_check_size(sizeof(struct inet6_skb_parm));
979
980 /* Register the socket-side information for inet6_create. */
981 for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
982 INIT_LIST_HEAD(r);
983
984 if (disable_ipv6_mod) {
985 pr_info("Loaded, but administratively disabled, reboot required to enable\n");
986 goto out;
987 }
988
989 err = proto_register(&tcpv6_prot, 1);
990 if (err)
991 goto out;
992
993 err = proto_register(&udpv6_prot, 1);
994 if (err)
995 goto out_unregister_tcp_proto;
996
997 err = proto_register(&udplitev6_prot, 1);
998 if (err)
999 goto out_unregister_udp_proto;
1000
1001 err = proto_register(&rawv6_prot, 1);
1002 if (err)
1003 goto out_unregister_udplite_proto;
1004
1005 err = proto_register(&pingv6_prot, 1);
1006 if (err)
1007 goto out_unregister_raw_proto;
1008
1009 /* We MUST register RAW sockets before we create the ICMP6,
1010 * IGMP6, or NDISC control sockets.
1011 */
1012 err = rawv6_init();
1013 if (err)
1014 goto out_unregister_ping_proto;
1015
1016 /* Register the family here so that the init calls below will
1017 * be able to create sockets. (?? is this dangerous ??)
1018 */
1019 err = sock_register(&inet6_family_ops);
1020 if (err)
1021 goto out_sock_register_fail;
1022
1023 /*
1024 * ipngwg API draft makes clear that the correct semantics
1025 * for TCP and UDP is to consider one TCP and UDP instance
1026 * in a host available by both INET and INET6 APIs and
1027 * able to communicate via both network protocols.
1028 */
1029
1030 err = register_pernet_subsys(&inet6_net_ops);
1031 if (err)
1032 goto register_pernet_fail;
1033 err = ip6_mr_init();
1034 if (err)
1035 goto ipmr_fail;
1036 err = icmpv6_init();
1037 if (err)
1038 goto icmp_fail;
1039 err = ndisc_init();
1040 if (err)
1041 goto ndisc_fail;
1042 err = igmp6_init();
1043 if (err)
1044 goto igmp_fail;
1045
1046 err = ipv6_netfilter_init();
1047 if (err)
1048 goto netfilter_fail;
1049 /* Create /proc/foo6 entries. */
1050 #ifdef CONFIG_PROC_FS
1051 err = -ENOMEM;
1052 if (raw6_proc_init())
1053 goto proc_raw6_fail;
1054 if (udplite6_proc_init())
1055 goto proc_udplite6_fail;
1056 if (ipv6_misc_proc_init())
1057 goto proc_misc6_fail;
1058 if (if6_proc_init())
1059 goto proc_if6_fail;
1060 #endif
1061 err = ip6_route_init();
1062 if (err)
1063 goto ip6_route_fail;
1064 err = ndisc_late_init();
1065 if (err)
1066 goto ndisc_late_fail;
1067 err = ip6_flowlabel_init();
1068 if (err)
1069 goto ip6_flowlabel_fail;
1070 err = ipv6_anycast_init();
1071 if (err)
1072 goto ipv6_anycast_fail;
1073 err = addrconf_init();
1074 if (err)
1075 goto addrconf_fail;
1076
1077 /* Init v6 extension headers. */
1078 err = ipv6_exthdrs_init();
1079 if (err)
1080 goto ipv6_exthdrs_fail;
1081
1082 err = ipv6_frag_init();
1083 if (err)
1084 goto ipv6_frag_fail;
1085
1086 /* Init v6 transport protocols. */
1087 err = udpv6_init();
1088 if (err)
1089 goto udpv6_fail;
1090
1091 err = udplitev6_init();
1092 if (err)
1093 goto udplitev6_fail;
1094
1095 err = udpv6_offload_init();
1096 if (err)
1097 goto udpv6_offload_fail;
1098
1099 err = tcpv6_init();
1100 if (err)
1101 goto tcpv6_fail;
1102
1103 err = ipv6_packet_init();
1104 if (err)
1105 goto ipv6_packet_fail;
1106
1107 err = pingv6_init();
1108 if (err)
1109 goto pingv6_fail;
1110
1111 err = calipso_init();
1112 if (err)
1113 goto calipso_fail;
1114
1115 err = seg6_init();
1116 if (err)
1117 goto seg6_fail;
1118
1119 err = igmp6_late_init();
1120 if (err)
1121 goto igmp6_late_err;
1122
1123 #ifdef CONFIG_SYSCTL
1124 err = ipv6_sysctl_register();
1125 if (err)
1126 goto sysctl_fail;
1127 #endif
1128
1129 /* ensure that ipv6 stubs are visible only after ipv6 is ready */
1130 wmb();
1131 ipv6_stub = &ipv6_stub_impl;
1132 ipv6_bpf_stub = &ipv6_bpf_stub_impl;
1133 out:
1134 return err;
1135
1136 #ifdef CONFIG_SYSCTL
1137 sysctl_fail:
1138 igmp6_late_cleanup();
1139 #endif
1140 igmp6_late_err:
1141 seg6_exit();
1142 seg6_fail:
1143 calipso_exit();
1144 calipso_fail:
1145 pingv6_exit();
1146 pingv6_fail:
1147 ipv6_packet_cleanup();
1148 ipv6_packet_fail:
1149 tcpv6_exit();
1150 tcpv6_fail:
1151 udpv6_offload_exit();
1152 udpv6_offload_fail:
1153 udplitev6_exit();
1154 udplitev6_fail:
1155 udpv6_exit();
1156 udpv6_fail:
1157 ipv6_frag_exit();
1158 ipv6_frag_fail:
1159 ipv6_exthdrs_exit();
1160 ipv6_exthdrs_fail:
1161 addrconf_cleanup();
1162 addrconf_fail:
1163 ipv6_anycast_cleanup();
1164 ipv6_anycast_fail:
1165 ip6_flowlabel_cleanup();
1166 ip6_flowlabel_fail:
1167 ndisc_late_cleanup();
1168 ndisc_late_fail:
1169 ip6_route_cleanup();
1170 ip6_route_fail:
1171 #ifdef CONFIG_PROC_FS
1172 if6_proc_exit();
1173 proc_if6_fail:
1174 ipv6_misc_proc_exit();
1175 proc_misc6_fail:
1176 udplite6_proc_exit();
1177 proc_udplite6_fail:
1178 raw6_proc_exit();
1179 proc_raw6_fail:
1180 #endif
1181 ipv6_netfilter_fini();
1182 netfilter_fail:
1183 igmp6_cleanup();
1184 igmp_fail:
1185 ndisc_cleanup();
1186 ndisc_fail:
1187 icmpv6_cleanup();
1188 icmp_fail:
1189 ip6_mr_cleanup();
1190 ipmr_fail:
1191 unregister_pernet_subsys(&inet6_net_ops);
1192 register_pernet_fail:
1193 sock_unregister(PF_INET6);
1194 rtnl_unregister_all(PF_INET6);
1195 out_sock_register_fail:
1196 rawv6_exit();
1197 out_unregister_ping_proto:
1198 proto_unregister(&pingv6_prot);
1199 out_unregister_raw_proto:
1200 proto_unregister(&rawv6_prot);
1201 out_unregister_udplite_proto:
1202 proto_unregister(&udplitev6_prot);
1203 out_unregister_udp_proto:
1204 proto_unregister(&udpv6_prot);
1205 out_unregister_tcp_proto:
1206 proto_unregister(&tcpv6_prot);
1207 goto out;
1208 }
1209 module_init(inet6_init);
1210
1211 MODULE_ALIAS_NETPROTO(PF_INET6);
1212