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
66 #include <linux/uaccess.h>
67 #include <linux/mroute6.h>
68
69 #include "ip6_offload.h"
70
71 MODULE_AUTHOR("Cast of dozens");
72 MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
73 MODULE_LICENSE("GPL");
74
75 /* The inetsw6 table contains everything that inet6_create needs to
76 * build a new socket.
77 */
78 static struct list_head inetsw6[SOCK_MAX];
79 static DEFINE_SPINLOCK(inetsw6_lock);
80
81 struct ipv6_params ipv6_defaults = {
82 .disable_ipv6 = 0,
83 .autoconf = 1,
84 };
85
86 static int disable_ipv6_mod;
87
88 module_param_named(disable, disable_ipv6_mod, int, 0444);
89 MODULE_PARM_DESC(disable, "Disable IPv6 module such that it is non-functional");
90
91 module_param_named(disable_ipv6, ipv6_defaults.disable_ipv6, int, 0444);
92 MODULE_PARM_DESC(disable_ipv6, "Disable IPv6 on all interfaces");
93
94 module_param_named(autoconf, ipv6_defaults.autoconf, int, 0444);
95 MODULE_PARM_DESC(autoconf, "Enable IPv6 address autoconfiguration on all interfaces");
96
ipv6_mod_enabled(void)97 bool ipv6_mod_enabled(void)
98 {
99 return disable_ipv6_mod == 0;
100 }
101 EXPORT_SYMBOL_GPL(ipv6_mod_enabled);
102
inet6_sk_generic(struct sock * sk)103 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
104 {
105 const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
106
107 return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
108 }
109
inet6_create(struct net * net,struct socket * sock,int protocol,int kern)110 static int inet6_create(struct net *net, struct socket *sock, int protocol,
111 int kern)
112 {
113 struct inet_sock *inet;
114 struct ipv6_pinfo *np;
115 struct sock *sk;
116 struct inet_protosw *answer;
117 struct proto *answer_prot;
118 unsigned char answer_flags;
119 int try_loading_module = 0;
120 int err;
121
122 if (protocol < 0 || protocol >= IPPROTO_MAX)
123 return -EINVAL;
124
125 /* Look for the requested type/protocol pair. */
126 lookup_protocol:
127 err = -ESOCKTNOSUPPORT;
128 rcu_read_lock();
129 list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
130
131 err = 0;
132 /* Check the non-wild match. */
133 if (protocol == answer->protocol) {
134 if (protocol != IPPROTO_IP)
135 break;
136 } else {
137 /* Check for the two wild cases. */
138 if (IPPROTO_IP == protocol) {
139 protocol = answer->protocol;
140 break;
141 }
142 if (IPPROTO_IP == answer->protocol)
143 break;
144 }
145 err = -EPROTONOSUPPORT;
146 }
147
148 if (err) {
149 if (try_loading_module < 2) {
150 rcu_read_unlock();
151 /*
152 * Be more specific, e.g. net-pf-10-proto-132-type-1
153 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
154 */
155 if (++try_loading_module == 1)
156 request_module("net-pf-%d-proto-%d-type-%d",
157 PF_INET6, protocol, sock->type);
158 /*
159 * Fall back to generic, e.g. net-pf-10-proto-132
160 * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
161 */
162 else
163 request_module("net-pf-%d-proto-%d",
164 PF_INET6, protocol);
165 goto lookup_protocol;
166 } else
167 goto out_rcu_unlock;
168 }
169
170 err = -EPERM;
171 if (sock->type == SOCK_RAW && !kern &&
172 !ns_capable(net->user_ns, CAP_NET_RAW))
173 goto out_rcu_unlock;
174
175 sock->ops = answer->ops;
176 answer_prot = answer->prot;
177 answer_flags = answer->flags;
178 rcu_read_unlock();
179
180 WARN_ON(!answer_prot->slab);
181
182 err = -ENOBUFS;
183 sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot, kern);
184 if (!sk)
185 goto out;
186
187 sock_init_data(sock, sk);
188
189 err = 0;
190 if (INET_PROTOSW_REUSE & answer_flags)
191 sk->sk_reuse = SK_CAN_REUSE;
192
193 inet = inet_sk(sk);
194 inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
195
196 if (SOCK_RAW == sock->type) {
197 inet->inet_num = protocol;
198 if (IPPROTO_RAW == protocol)
199 inet->hdrincl = 1;
200 }
201
202 sk->sk_destruct = inet_sock_destruct;
203 sk->sk_family = PF_INET6;
204 sk->sk_protocol = protocol;
205
206 sk->sk_backlog_rcv = answer->prot->backlog_rcv;
207
208 inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
209 np->hop_limit = -1;
210 np->mcast_hops = IPV6_DEFAULT_MCASTHOPS;
211 np->mc_loop = 1;
212 np->mc_all = 1;
213 np->pmtudisc = IPV6_PMTUDISC_WANT;
214 np->repflow = net->ipv6.sysctl.flowlabel_reflect & FLOWLABEL_REFLECT_ESTABLISHED;
215 sk->sk_ipv6only = net->ipv6.sysctl.bindv6only;
216
217 /* Init the ipv4 part of the socket since we can have sockets
218 * using v6 API for ipv4.
219 */
220 inet->uc_ttl = -1;
221
222 inet->mc_loop = 1;
223 inet->mc_ttl = 1;
224 inet->mc_index = 0;
225 RCU_INIT_POINTER(inet->mc_list, NULL);
226 inet->rcv_tos = 0;
227
228 if (net->ipv4.sysctl_ip_no_pmtu_disc)
229 inet->pmtudisc = IP_PMTUDISC_DONT;
230 else
231 inet->pmtudisc = IP_PMTUDISC_WANT;
232 /*
233 * Increment only the relevant sk_prot->socks debug field, this changes
234 * the previous behaviour of incrementing both the equivalent to
235 * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
236 *
237 * This allows better debug granularity as we'll know exactly how many
238 * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
239 * transport protocol socks. -acme
240 */
241 sk_refcnt_debug_inc(sk);
242
243 if (inet->inet_num) {
244 /* It assumes that any protocol which allows
245 * the user to assign a number at socket
246 * creation time automatically shares.
247 */
248 inet->inet_sport = htons(inet->inet_num);
249 err = sk->sk_prot->hash(sk);
250 if (err) {
251 sk_common_release(sk);
252 goto out;
253 }
254 }
255 if (sk->sk_prot->init) {
256 err = sk->sk_prot->init(sk);
257 if (err) {
258 sk_common_release(sk);
259 goto out;
260 }
261 }
262
263 if (!kern) {
264 err = BPF_CGROUP_RUN_PROG_INET_SOCK(sk);
265 if (err) {
266 sk_common_release(sk);
267 goto out;
268 }
269 }
270 out:
271 return err;
272 out_rcu_unlock:
273 rcu_read_unlock();
274 goto out;
275 }
276
__inet6_bind(struct sock * sk,struct sockaddr * uaddr,int addr_len,u32 flags)277 static int __inet6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len,
278 u32 flags)
279 {
280 struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr;
281 struct inet_sock *inet = inet_sk(sk);
282 struct ipv6_pinfo *np = inet6_sk(sk);
283 struct net *net = sock_net(sk);
284 __be32 v4addr = 0;
285 unsigned short snum;
286 bool saved_ipv6only;
287 int addr_type = 0;
288 int err = 0;
289
290 if (addr->sin6_family != AF_INET6)
291 return -EAFNOSUPPORT;
292
293 addr_type = ipv6_addr_type(&addr->sin6_addr);
294 if ((addr_type & IPV6_ADDR_MULTICAST) && sk->sk_type == SOCK_STREAM)
295 return -EINVAL;
296
297 snum = ntohs(addr->sin6_port);
298 if (!(flags & BIND_NO_CAP_NET_BIND_SERVICE) &&
299 snum && inet_port_requires_bind_service(net, snum) &&
300 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
301 return -EACCES;
302
303 if (flags & BIND_WITH_LOCK)
304 lock_sock(sk);
305
306 /* Check these errors (active socket, double bind). */
307 if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
308 err = -EINVAL;
309 goto out;
310 }
311
312 /* Check if the address belongs to the host. */
313 if (addr_type == IPV6_ADDR_MAPPED) {
314 struct net_device *dev = NULL;
315 int chk_addr_ret;
316
317 /* Binding to v4-mapped address on a v6-only socket
318 * makes no sense
319 */
320 if (sk->sk_ipv6only) {
321 err = -EINVAL;
322 goto out;
323 }
324
325 rcu_read_lock();
326 if (sk->sk_bound_dev_if) {
327 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
328 if (!dev) {
329 err = -ENODEV;
330 goto out_unlock;
331 }
332 }
333
334 /* Reproduce AF_INET checks to make the bindings consistent */
335 v4addr = addr->sin6_addr.s6_addr32[3];
336 chk_addr_ret = inet_addr_type_dev_table(net, dev, v4addr);
337 rcu_read_unlock();
338
339 if (!inet_can_nonlocal_bind(net, inet) &&
340 v4addr != htonl(INADDR_ANY) &&
341 chk_addr_ret != RTN_LOCAL &&
342 chk_addr_ret != RTN_MULTICAST &&
343 chk_addr_ret != RTN_BROADCAST) {
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->bind_address_no_port ||
406 (flags & BIND_FORCE_ADDRESS_NO_PORT))) {
407 if (sk->sk_prot->get_port(sk, snum)) {
408 sk->sk_ipv6only = saved_ipv6only;
409 inet_reset_saddr(sk);
410 err = -EADDRINUSE;
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 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 */
inet6_bind(struct socket * sock,struct sockaddr * uaddr,int addr_len)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 int err = 0;
445
446 /* If the socket has its own bind function then use it. */
447 if (sk->sk_prot->bind)
448 return sk->sk_prot->bind(sk, uaddr, addr_len);
449
450 if (addr_len < SIN6_LEN_RFC2133)
451 return -EINVAL;
452
453 /* BPF prog is run before any checks are done so that if the prog
454 * changes context in a wrong way it will be caught.
455 */
456 err = BPF_CGROUP_RUN_PROG_INET_BIND_LOCK(sk, uaddr,
457 BPF_CGROUP_INET6_BIND, &flags);
458 if (err)
459 return err;
460
461 return __inet6_bind(sk, uaddr, addr_len, flags);
462 }
463 EXPORT_SYMBOL(inet6_bind);
464
inet6_release(struct socket * sock)465 int inet6_release(struct socket *sock)
466 {
467 struct sock *sk = sock->sk;
468
469 if (!sk)
470 return -EINVAL;
471
472 /* Free mc lists */
473 ipv6_sock_mc_close(sk);
474
475 /* Free ac lists */
476 ipv6_sock_ac_close(sk);
477
478 return inet_release(sock);
479 }
480 EXPORT_SYMBOL(inet6_release);
481
inet6_destroy_sock(struct sock * sk)482 void inet6_destroy_sock(struct sock *sk)
483 {
484 struct ipv6_pinfo *np = inet6_sk(sk);
485 struct sk_buff *skb;
486 struct ipv6_txoptions *opt;
487
488 /* Release rx options */
489
490 skb = xchg(&np->pktoptions, NULL);
491 kfree_skb(skb);
492
493 skb = xchg(&np->rxpmtu, NULL);
494 kfree_skb(skb);
495
496 /* Free flowlabels */
497 fl6_free_socklist(sk);
498
499 /* Free tx options */
500
501 opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL);
502 if (opt) {
503 atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
504 txopt_put(opt);
505 }
506 }
507 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
508
509 /*
510 * This does both peername and sockname.
511 */
inet6_getname(struct socket * sock,struct sockaddr * uaddr,int peer)512 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
513 int peer)
514 {
515 struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
516 struct sock *sk = sock->sk;
517 struct inet_sock *inet = inet_sk(sk);
518 struct ipv6_pinfo *np = inet6_sk(sk);
519
520 sin->sin6_family = AF_INET6;
521 sin->sin6_flowinfo = 0;
522 sin->sin6_scope_id = 0;
523 if (peer) {
524 if (!inet->inet_dport)
525 return -ENOTCONN;
526 if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
527 peer == 1)
528 return -ENOTCONN;
529 sin->sin6_port = inet->inet_dport;
530 sin->sin6_addr = sk->sk_v6_daddr;
531 if (np->sndflow)
532 sin->sin6_flowinfo = np->flow_label;
533 BPF_CGROUP_RUN_SA_PROG_LOCK(sk, (struct sockaddr *)sin,
534 BPF_CGROUP_INET6_GETPEERNAME,
535 NULL);
536 } else {
537 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
538 sin->sin6_addr = np->saddr;
539 else
540 sin->sin6_addr = sk->sk_v6_rcv_saddr;
541 sin->sin6_port = inet->inet_sport;
542 BPF_CGROUP_RUN_SA_PROG_LOCK(sk, (struct sockaddr *)sin,
543 BPF_CGROUP_INET6_GETSOCKNAME,
544 NULL);
545 }
546 sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
547 sk->sk_bound_dev_if);
548 return sizeof(*sin);
549 }
550 EXPORT_SYMBOL(inet6_getname);
551
inet6_ioctl(struct socket * sock,unsigned int cmd,unsigned long arg)552 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
553 {
554 void __user *argp = (void __user *)arg;
555 struct sock *sk = sock->sk;
556 struct net *net = sock_net(sk);
557
558 switch (cmd) {
559 case SIOCADDRT:
560 case SIOCDELRT: {
561 struct in6_rtmsg rtmsg;
562
563 if (copy_from_user(&rtmsg, argp, sizeof(rtmsg)))
564 return -EFAULT;
565 return ipv6_route_ioctl(net, cmd, &rtmsg);
566 }
567 case SIOCSIFADDR:
568 return addrconf_add_ifaddr(net, argp);
569 case SIOCDIFADDR:
570 return addrconf_del_ifaddr(net, argp);
571 case SIOCSIFDSTADDR:
572 return addrconf_set_dstaddr(net, argp);
573 default:
574 if (!sk->sk_prot->ioctl)
575 return -ENOIOCTLCMD;
576 return sk->sk_prot->ioctl(sk, cmd, arg);
577 }
578 /*NOTREACHED*/
579 return 0;
580 }
581 EXPORT_SYMBOL(inet6_ioctl);
582
583 #ifdef CONFIG_COMPAT
584 struct compat_in6_rtmsg {
585 struct in6_addr rtmsg_dst;
586 struct in6_addr rtmsg_src;
587 struct in6_addr rtmsg_gateway;
588 u32 rtmsg_type;
589 u16 rtmsg_dst_len;
590 u16 rtmsg_src_len;
591 u32 rtmsg_metric;
592 u32 rtmsg_info;
593 u32 rtmsg_flags;
594 s32 rtmsg_ifindex;
595 };
596
inet6_compat_routing_ioctl(struct sock * sk,unsigned int cmd,struct compat_in6_rtmsg __user * ur)597 static int inet6_compat_routing_ioctl(struct sock *sk, unsigned int cmd,
598 struct compat_in6_rtmsg __user *ur)
599 {
600 struct in6_rtmsg rt;
601
602 if (copy_from_user(&rt.rtmsg_dst, &ur->rtmsg_dst,
603 3 * sizeof(struct in6_addr)) ||
604 get_user(rt.rtmsg_type, &ur->rtmsg_type) ||
605 get_user(rt.rtmsg_dst_len, &ur->rtmsg_dst_len) ||
606 get_user(rt.rtmsg_src_len, &ur->rtmsg_src_len) ||
607 get_user(rt.rtmsg_metric, &ur->rtmsg_metric) ||
608 get_user(rt.rtmsg_info, &ur->rtmsg_info) ||
609 get_user(rt.rtmsg_flags, &ur->rtmsg_flags) ||
610 get_user(rt.rtmsg_ifindex, &ur->rtmsg_ifindex))
611 return -EFAULT;
612
613
614 return ipv6_route_ioctl(sock_net(sk), cmd, &rt);
615 }
616
inet6_compat_ioctl(struct socket * sock,unsigned int cmd,unsigned long arg)617 int inet6_compat_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
618 {
619 void __user *argp = compat_ptr(arg);
620 struct sock *sk = sock->sk;
621
622 switch (cmd) {
623 case SIOCADDRT:
624 case SIOCDELRT:
625 return inet6_compat_routing_ioctl(sk, cmd, argp);
626 default:
627 return -ENOIOCTLCMD;
628 }
629 }
630 EXPORT_SYMBOL_GPL(inet6_compat_ioctl);
631 #endif /* CONFIG_COMPAT */
632
633 INDIRECT_CALLABLE_DECLARE(int udpv6_sendmsg(struct sock *, struct msghdr *,
634 size_t));
inet6_sendmsg(struct socket * sock,struct msghdr * msg,size_t size)635 int inet6_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
636 {
637 struct sock *sk = sock->sk;
638
639 if (unlikely(inet_send_prepare(sk)))
640 return -EAGAIN;
641
642 return INDIRECT_CALL_2(sk->sk_prot->sendmsg, tcp_sendmsg, udpv6_sendmsg,
643 sk, msg, size);
644 }
645
646 INDIRECT_CALLABLE_DECLARE(int udpv6_recvmsg(struct sock *, struct msghdr *,
647 size_t, int, int, int *));
inet6_recvmsg(struct socket * sock,struct msghdr * msg,size_t size,int flags)648 int inet6_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
649 int flags)
650 {
651 struct sock *sk = sock->sk;
652 int addr_len = 0;
653 int err;
654
655 if (likely(!(flags & MSG_ERRQUEUE)))
656 sock_rps_record_flow(sk);
657
658 err = INDIRECT_CALL_2(sk->sk_prot->recvmsg, tcp_recvmsg, udpv6_recvmsg,
659 sk, msg, size, flags & MSG_DONTWAIT,
660 flags & ~MSG_DONTWAIT, &addr_len);
661 if (err >= 0)
662 msg->msg_namelen = addr_len;
663 return err;
664 }
665
666 const struct proto_ops inet6_stream_ops = {
667 .family = PF_INET6,
668 .owner = THIS_MODULE,
669 .release = inet6_release,
670 .bind = inet6_bind,
671 .connect = inet_stream_connect, /* ok */
672 .socketpair = sock_no_socketpair, /* a do nothing */
673 .accept = inet_accept, /* ok */
674 .getname = inet6_getname,
675 .poll = tcp_poll, /* ok */
676 .ioctl = inet6_ioctl, /* must change */
677 .gettstamp = sock_gettstamp,
678 .listen = inet_listen, /* ok */
679 .shutdown = inet_shutdown, /* ok */
680 .setsockopt = sock_common_setsockopt, /* ok */
681 .getsockopt = sock_common_getsockopt, /* ok */
682 .sendmsg = inet6_sendmsg, /* retpoline's sake */
683 .recvmsg = inet6_recvmsg, /* retpoline's sake */
684 #ifdef CONFIG_MMU
685 .mmap = tcp_mmap,
686 #endif
687 .sendpage = inet_sendpage,
688 .sendmsg_locked = tcp_sendmsg_locked,
689 .sendpage_locked = tcp_sendpage_locked,
690 .splice_read = tcp_splice_read,
691 .read_sock = tcp_read_sock,
692 .peek_len = tcp_peek_len,
693 #ifdef CONFIG_COMPAT
694 .compat_ioctl = inet6_compat_ioctl,
695 #endif
696 .set_rcvlowat = tcp_set_rcvlowat,
697 };
698
699 const struct proto_ops inet6_dgram_ops = {
700 .family = PF_INET6,
701 .owner = THIS_MODULE,
702 .release = inet6_release,
703 .bind = inet6_bind,
704 .connect = inet_dgram_connect, /* ok */
705 .socketpair = sock_no_socketpair, /* a do nothing */
706 .accept = sock_no_accept, /* a do nothing */
707 .getname = inet6_getname,
708 .poll = udp_poll, /* ok */
709 .ioctl = inet6_ioctl, /* must change */
710 .gettstamp = sock_gettstamp,
711 .listen = sock_no_listen, /* ok */
712 .shutdown = inet_shutdown, /* ok */
713 .setsockopt = sock_common_setsockopt, /* ok */
714 .getsockopt = sock_common_getsockopt, /* ok */
715 .sendmsg = inet6_sendmsg, /* retpoline's sake */
716 .recvmsg = inet6_recvmsg, /* retpoline's sake */
717 .read_sock = udp_read_sock,
718 .mmap = sock_no_mmap,
719 .sendpage = sock_no_sendpage,
720 .set_peek_off = sk_set_peek_off,
721 #ifdef CONFIG_COMPAT
722 .compat_ioctl = inet6_compat_ioctl,
723 #endif
724 };
725
726 static const struct net_proto_family inet6_family_ops = {
727 .family = PF_INET6,
728 .create = inet6_create,
729 .owner = THIS_MODULE,
730 };
731
inet6_register_protosw(struct inet_protosw * p)732 int inet6_register_protosw(struct inet_protosw *p)
733 {
734 struct list_head *lh;
735 struct inet_protosw *answer;
736 struct list_head *last_perm;
737 int protocol = p->protocol;
738 int ret;
739
740 spin_lock_bh(&inetsw6_lock);
741
742 ret = -EINVAL;
743 if (p->type >= SOCK_MAX)
744 goto out_illegal;
745
746 /* If we are trying to override a permanent protocol, bail. */
747 answer = NULL;
748 ret = -EPERM;
749 last_perm = &inetsw6[p->type];
750 list_for_each(lh, &inetsw6[p->type]) {
751 answer = list_entry(lh, struct inet_protosw, list);
752
753 /* Check only the non-wild match. */
754 if (INET_PROTOSW_PERMANENT & answer->flags) {
755 if (protocol == answer->protocol)
756 break;
757 last_perm = lh;
758 }
759
760 answer = NULL;
761 }
762 if (answer)
763 goto out_permanent;
764
765 /* Add the new entry after the last permanent entry if any, so that
766 * the new entry does not override a permanent entry when matched with
767 * a wild-card protocol. But it is allowed to override any existing
768 * non-permanent entry. This means that when we remove this entry, the
769 * system automatically returns to the old behavior.
770 */
771 list_add_rcu(&p->list, last_perm);
772 ret = 0;
773 out:
774 spin_unlock_bh(&inetsw6_lock);
775 return ret;
776
777 out_permanent:
778 pr_err("Attempt to override permanent protocol %d\n", protocol);
779 goto out;
780
781 out_illegal:
782 pr_err("Ignoring attempt to register invalid socket type %d\n",
783 p->type);
784 goto out;
785 }
786 EXPORT_SYMBOL(inet6_register_protosw);
787
788 void
inet6_unregister_protosw(struct inet_protosw * p)789 inet6_unregister_protosw(struct inet_protosw *p)
790 {
791 if (INET_PROTOSW_PERMANENT & p->flags) {
792 pr_err("Attempt to unregister permanent protocol %d\n",
793 p->protocol);
794 } else {
795 spin_lock_bh(&inetsw6_lock);
796 list_del_rcu(&p->list);
797 spin_unlock_bh(&inetsw6_lock);
798
799 synchronize_net();
800 }
801 }
802 EXPORT_SYMBOL(inet6_unregister_protosw);
803
inet6_sk_rebuild_header(struct sock * sk)804 int inet6_sk_rebuild_header(struct sock *sk)
805 {
806 struct ipv6_pinfo *np = inet6_sk(sk);
807 struct dst_entry *dst;
808
809 dst = __sk_dst_check(sk, np->dst_cookie);
810
811 if (!dst) {
812 struct inet_sock *inet = inet_sk(sk);
813 struct in6_addr *final_p, final;
814 struct flowi6 fl6;
815
816 memset(&fl6, 0, sizeof(fl6));
817 fl6.flowi6_proto = sk->sk_protocol;
818 fl6.daddr = sk->sk_v6_daddr;
819 fl6.saddr = np->saddr;
820 fl6.flowlabel = np->flow_label;
821 fl6.flowi6_oif = sk->sk_bound_dev_if;
822 fl6.flowi6_mark = sk->sk_mark;
823 fl6.fl6_dport = inet->inet_dport;
824 fl6.fl6_sport = inet->inet_sport;
825 fl6.flowi6_uid = sk->sk_uid;
826 security_sk_classify_flow(sk, flowi6_to_flowi_common(&fl6));
827
828 rcu_read_lock();
829 final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt),
830 &final);
831 rcu_read_unlock();
832
833 dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p);
834 if (IS_ERR(dst)) {
835 sk->sk_route_caps = 0;
836 sk->sk_err_soft = -PTR_ERR(dst);
837 return PTR_ERR(dst);
838 }
839
840 ip6_dst_store(sk, dst, NULL, NULL);
841 }
842
843 return 0;
844 }
845 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
846
ipv6_opt_accepted(const struct sock * sk,const struct sk_buff * skb,const struct inet6_skb_parm * opt)847 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb,
848 const struct inet6_skb_parm *opt)
849 {
850 const struct ipv6_pinfo *np = inet6_sk(sk);
851
852 if (np->rxopt.all) {
853 if (((opt->flags & IP6SKB_HOPBYHOP) &&
854 (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) ||
855 (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) &&
856 np->rxopt.bits.rxflow) ||
857 (opt->srcrt && (np->rxopt.bits.srcrt ||
858 np->rxopt.bits.osrcrt)) ||
859 ((opt->dst1 || opt->dst0) &&
860 (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
861 return true;
862 }
863 return false;
864 }
865 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
866
867 static struct packet_type ipv6_packet_type __read_mostly = {
868 .type = cpu_to_be16(ETH_P_IPV6),
869 .func = ipv6_rcv,
870 .list_func = ipv6_list_rcv,
871 };
872
ipv6_packet_init(void)873 static int __init ipv6_packet_init(void)
874 {
875 dev_add_pack(&ipv6_packet_type);
876 return 0;
877 }
878
ipv6_packet_cleanup(void)879 static void ipv6_packet_cleanup(void)
880 {
881 dev_remove_pack(&ipv6_packet_type);
882 }
883
ipv6_init_mibs(struct net * net)884 static int __net_init ipv6_init_mibs(struct net *net)
885 {
886 int i;
887
888 net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib);
889 if (!net->mib.udp_stats_in6)
890 return -ENOMEM;
891 net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib);
892 if (!net->mib.udplite_stats_in6)
893 goto err_udplite_mib;
894 net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib);
895 if (!net->mib.ipv6_statistics)
896 goto err_ip_mib;
897
898 for_each_possible_cpu(i) {
899 struct ipstats_mib *af_inet6_stats;
900 af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i);
901 u64_stats_init(&af_inet6_stats->syncp);
902 }
903
904
905 net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib);
906 if (!net->mib.icmpv6_statistics)
907 goto err_icmp_mib;
908 net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
909 GFP_KERNEL);
910 if (!net->mib.icmpv6msg_statistics)
911 goto err_icmpmsg_mib;
912 return 0;
913
914 err_icmpmsg_mib:
915 free_percpu(net->mib.icmpv6_statistics);
916 err_icmp_mib:
917 free_percpu(net->mib.ipv6_statistics);
918 err_ip_mib:
919 free_percpu(net->mib.udplite_stats_in6);
920 err_udplite_mib:
921 free_percpu(net->mib.udp_stats_in6);
922 return -ENOMEM;
923 }
924
ipv6_cleanup_mibs(struct net * net)925 static void ipv6_cleanup_mibs(struct net *net)
926 {
927 free_percpu(net->mib.udp_stats_in6);
928 free_percpu(net->mib.udplite_stats_in6);
929 free_percpu(net->mib.ipv6_statistics);
930 free_percpu(net->mib.icmpv6_statistics);
931 kfree(net->mib.icmpv6msg_statistics);
932 }
933
inet6_net_init(struct net * net)934 static int __net_init inet6_net_init(struct net *net)
935 {
936 int err = 0;
937
938 net->ipv6.sysctl.bindv6only = 0;
939 net->ipv6.sysctl.icmpv6_time = 1*HZ;
940 net->ipv6.sysctl.icmpv6_echo_ignore_all = 0;
941 net->ipv6.sysctl.icmpv6_echo_ignore_multicast = 0;
942 net->ipv6.sysctl.icmpv6_echo_ignore_anycast = 0;
943
944 /* By default, rate limit error messages.
945 * Except for pmtu discovery, it would break it.
946 * proc_do_large_bitmap needs pointer to the bitmap.
947 */
948 bitmap_set(net->ipv6.sysctl.icmpv6_ratemask, 0, ICMPV6_ERRMSG_MAX + 1);
949 bitmap_clear(net->ipv6.sysctl.icmpv6_ratemask, ICMPV6_PKT_TOOBIG, 1);
950 net->ipv6.sysctl.icmpv6_ratemask_ptr = net->ipv6.sysctl.icmpv6_ratemask;
951
952 net->ipv6.sysctl.flowlabel_consistency = 1;
953 net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS;
954 net->ipv6.sysctl.idgen_retries = 3;
955 net->ipv6.sysctl.idgen_delay = 1 * HZ;
956 net->ipv6.sysctl.flowlabel_state_ranges = 0;
957 net->ipv6.sysctl.max_dst_opts_cnt = IP6_DEFAULT_MAX_DST_OPTS_CNT;
958 net->ipv6.sysctl.max_hbh_opts_cnt = IP6_DEFAULT_MAX_HBH_OPTS_CNT;
959 net->ipv6.sysctl.max_dst_opts_len = IP6_DEFAULT_MAX_DST_OPTS_LEN;
960 net->ipv6.sysctl.max_hbh_opts_len = IP6_DEFAULT_MAX_HBH_OPTS_LEN;
961 net->ipv6.sysctl.fib_notify_on_flag_change = 0;
962 atomic_set(&net->ipv6.fib6_sernum, 1);
963
964 err = ipv6_init_mibs(net);
965 if (err)
966 return err;
967 #ifdef CONFIG_PROC_FS
968 err = udp6_proc_init(net);
969 if (err)
970 goto out;
971 err = tcp6_proc_init(net);
972 if (err)
973 goto proc_tcp6_fail;
974 err = ac6_proc_init(net);
975 if (err)
976 goto proc_ac6_fail;
977 #endif
978 return err;
979
980 #ifdef CONFIG_PROC_FS
981 proc_ac6_fail:
982 tcp6_proc_exit(net);
983 proc_tcp6_fail:
984 udp6_proc_exit(net);
985 out:
986 ipv6_cleanup_mibs(net);
987 return err;
988 #endif
989 }
990
inet6_net_exit(struct net * net)991 static void __net_exit inet6_net_exit(struct net *net)
992 {
993 #ifdef CONFIG_PROC_FS
994 udp6_proc_exit(net);
995 tcp6_proc_exit(net);
996 ac6_proc_exit(net);
997 #endif
998 ipv6_cleanup_mibs(net);
999 }
1000
1001 static struct pernet_operations inet6_net_ops = {
1002 .init = inet6_net_init,
1003 .exit = inet6_net_exit,
1004 };
1005
ipv6_route_input(struct sk_buff * skb)1006 static int ipv6_route_input(struct sk_buff *skb)
1007 {
1008 ip6_route_input(skb);
1009 return skb_dst(skb)->error;
1010 }
1011
1012 static const struct ipv6_stub ipv6_stub_impl = {
1013 .ipv6_sock_mc_join = ipv6_sock_mc_join,
1014 .ipv6_sock_mc_drop = ipv6_sock_mc_drop,
1015 .ipv6_dst_lookup_flow = ip6_dst_lookup_flow,
1016 .ipv6_route_input = ipv6_route_input,
1017 .fib6_get_table = fib6_get_table,
1018 .fib6_table_lookup = fib6_table_lookup,
1019 .fib6_lookup = fib6_lookup,
1020 .fib6_select_path = fib6_select_path,
1021 .ip6_mtu_from_fib6 = ip6_mtu_from_fib6,
1022 .fib6_nh_init = fib6_nh_init,
1023 .fib6_nh_release = fib6_nh_release,
1024 .fib6_update_sernum = fib6_update_sernum_stub,
1025 .fib6_rt_update = fib6_rt_update,
1026 .ip6_del_rt = ip6_del_rt,
1027 .udpv6_encap_enable = udpv6_encap_enable,
1028 .ndisc_send_na = ndisc_send_na,
1029 #if IS_ENABLED(CONFIG_XFRM)
1030 .xfrm6_local_rxpmtu = xfrm6_local_rxpmtu,
1031 .xfrm6_udp_encap_rcv = xfrm6_udp_encap_rcv,
1032 .xfrm6_rcv_encap = xfrm6_rcv_encap,
1033 #endif
1034 .nd_tbl = &nd_tbl,
1035 .ipv6_fragment = ip6_fragment,
1036 .ipv6_dev_find = ipv6_dev_find,
1037 };
1038
1039 static const struct ipv6_bpf_stub ipv6_bpf_stub_impl = {
1040 .inet6_bind = __inet6_bind,
1041 .udp6_lib_lookup = __udp6_lib_lookup,
1042 };
1043
inet6_init(void)1044 static int __init inet6_init(void)
1045 {
1046 struct list_head *r;
1047 int err = 0;
1048
1049 sock_skb_cb_check_size(sizeof(struct inet6_skb_parm));
1050
1051 /* Register the socket-side information for inet6_create. */
1052 for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
1053 INIT_LIST_HEAD(r);
1054
1055 if (disable_ipv6_mod) {
1056 pr_info("Loaded, but administratively disabled, reboot required to enable\n");
1057 goto out;
1058 }
1059
1060 err = proto_register(&tcpv6_prot, 1);
1061 if (err)
1062 goto out;
1063
1064 err = proto_register(&udpv6_prot, 1);
1065 if (err)
1066 goto out_unregister_tcp_proto;
1067
1068 err = proto_register(&udplitev6_prot, 1);
1069 if (err)
1070 goto out_unregister_udp_proto;
1071
1072 err = proto_register(&rawv6_prot, 1);
1073 if (err)
1074 goto out_unregister_udplite_proto;
1075
1076 err = proto_register(&pingv6_prot, 1);
1077 if (err)
1078 goto out_unregister_raw_proto;
1079
1080 /* We MUST register RAW sockets before we create the ICMP6,
1081 * IGMP6, or NDISC control sockets.
1082 */
1083 err = rawv6_init();
1084 if (err)
1085 goto out_unregister_ping_proto;
1086
1087 /* Register the family here so that the init calls below will
1088 * be able to create sockets. (?? is this dangerous ??)
1089 */
1090 err = sock_register(&inet6_family_ops);
1091 if (err)
1092 goto out_sock_register_fail;
1093
1094 /*
1095 * ipngwg API draft makes clear that the correct semantics
1096 * for TCP and UDP is to consider one TCP and UDP instance
1097 * in a host available by both INET and INET6 APIs and
1098 * able to communicate via both network protocols.
1099 */
1100
1101 err = register_pernet_subsys(&inet6_net_ops);
1102 if (err)
1103 goto register_pernet_fail;
1104 err = ip6_mr_init();
1105 if (err)
1106 goto ipmr_fail;
1107 err = icmpv6_init();
1108 if (err)
1109 goto icmp_fail;
1110 err = ndisc_init();
1111 if (err)
1112 goto ndisc_fail;
1113 err = igmp6_init();
1114 if (err)
1115 goto igmp_fail;
1116
1117 err = ipv6_netfilter_init();
1118 if (err)
1119 goto netfilter_fail;
1120 /* Create /proc/foo6 entries. */
1121 #ifdef CONFIG_PROC_FS
1122 err = -ENOMEM;
1123 if (raw6_proc_init())
1124 goto proc_raw6_fail;
1125 if (udplite6_proc_init())
1126 goto proc_udplite6_fail;
1127 if (ipv6_misc_proc_init())
1128 goto proc_misc6_fail;
1129 if (if6_proc_init())
1130 goto proc_if6_fail;
1131 #endif
1132 err = ip6_route_init();
1133 if (err)
1134 goto ip6_route_fail;
1135 err = ndisc_late_init();
1136 if (err)
1137 goto ndisc_late_fail;
1138 err = ip6_flowlabel_init();
1139 if (err)
1140 goto ip6_flowlabel_fail;
1141 err = ipv6_anycast_init();
1142 if (err)
1143 goto ipv6_anycast_fail;
1144 err = addrconf_init();
1145 if (err)
1146 goto addrconf_fail;
1147
1148 /* Init v6 extension headers. */
1149 err = ipv6_exthdrs_init();
1150 if (err)
1151 goto ipv6_exthdrs_fail;
1152
1153 err = ipv6_frag_init();
1154 if (err)
1155 goto ipv6_frag_fail;
1156
1157 /* Init v6 transport protocols. */
1158 err = udpv6_init();
1159 if (err)
1160 goto udpv6_fail;
1161
1162 err = udplitev6_init();
1163 if (err)
1164 goto udplitev6_fail;
1165
1166 err = udpv6_offload_init();
1167 if (err)
1168 goto udpv6_offload_fail;
1169
1170 err = tcpv6_init();
1171 if (err)
1172 goto tcpv6_fail;
1173
1174 err = ipv6_packet_init();
1175 if (err)
1176 goto ipv6_packet_fail;
1177
1178 err = pingv6_init();
1179 if (err)
1180 goto pingv6_fail;
1181
1182 err = calipso_init();
1183 if (err)
1184 goto calipso_fail;
1185
1186 err = seg6_init();
1187 if (err)
1188 goto seg6_fail;
1189
1190 err = rpl_init();
1191 if (err)
1192 goto rpl_fail;
1193
1194 err = igmp6_late_init();
1195 if (err)
1196 goto igmp6_late_err;
1197
1198 #ifdef CONFIG_SYSCTL
1199 err = ipv6_sysctl_register();
1200 if (err)
1201 goto sysctl_fail;
1202 #endif
1203
1204 /* ensure that ipv6 stubs are visible only after ipv6 is ready */
1205 wmb();
1206 ipv6_stub = &ipv6_stub_impl;
1207 ipv6_bpf_stub = &ipv6_bpf_stub_impl;
1208 out:
1209 return err;
1210
1211 #ifdef CONFIG_SYSCTL
1212 sysctl_fail:
1213 igmp6_late_cleanup();
1214 #endif
1215 igmp6_late_err:
1216 rpl_exit();
1217 rpl_fail:
1218 seg6_exit();
1219 seg6_fail:
1220 calipso_exit();
1221 calipso_fail:
1222 pingv6_exit();
1223 pingv6_fail:
1224 ipv6_packet_cleanup();
1225 ipv6_packet_fail:
1226 tcpv6_exit();
1227 tcpv6_fail:
1228 udpv6_offload_exit();
1229 udpv6_offload_fail:
1230 udplitev6_exit();
1231 udplitev6_fail:
1232 udpv6_exit();
1233 udpv6_fail:
1234 ipv6_frag_exit();
1235 ipv6_frag_fail:
1236 ipv6_exthdrs_exit();
1237 ipv6_exthdrs_fail:
1238 addrconf_cleanup();
1239 addrconf_fail:
1240 ipv6_anycast_cleanup();
1241 ipv6_anycast_fail:
1242 ip6_flowlabel_cleanup();
1243 ip6_flowlabel_fail:
1244 ndisc_late_cleanup();
1245 ndisc_late_fail:
1246 ip6_route_cleanup();
1247 ip6_route_fail:
1248 #ifdef CONFIG_PROC_FS
1249 if6_proc_exit();
1250 proc_if6_fail:
1251 ipv6_misc_proc_exit();
1252 proc_misc6_fail:
1253 udplite6_proc_exit();
1254 proc_udplite6_fail:
1255 raw6_proc_exit();
1256 proc_raw6_fail:
1257 #endif
1258 ipv6_netfilter_fini();
1259 netfilter_fail:
1260 igmp6_cleanup();
1261 igmp_fail:
1262 ndisc_cleanup();
1263 ndisc_fail:
1264 icmpv6_cleanup();
1265 icmp_fail:
1266 ip6_mr_cleanup();
1267 ipmr_fail:
1268 unregister_pernet_subsys(&inet6_net_ops);
1269 register_pernet_fail:
1270 sock_unregister(PF_INET6);
1271 rtnl_unregister_all(PF_INET6);
1272 out_sock_register_fail:
1273 rawv6_exit();
1274 out_unregister_ping_proto:
1275 proto_unregister(&pingv6_prot);
1276 out_unregister_raw_proto:
1277 proto_unregister(&rawv6_prot);
1278 out_unregister_udplite_proto:
1279 proto_unregister(&udplitev6_prot);
1280 out_unregister_udp_proto:
1281 proto_unregister(&udpv6_prot);
1282 out_unregister_tcp_proto:
1283 proto_unregister(&tcpv6_prot);
1284 goto out;
1285 }
1286 module_init(inet6_init);
1287
1288 MODULE_ALIAS_NETPROTO(PF_INET6);
1289