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 inet->mc_list = NULL;
226 inet->rcv_tos = 0;
227
228 if (READ_ONCE(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 (snum && inet_is_local_unbindable_port(net, snum))
299 return -EPERM;
300 if (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_can_nonlocal_bind(net, inet) &&
341 v4addr != htonl(INADDR_ANY) &&
342 chk_addr_ret != RTN_LOCAL &&
343 chk_addr_ret != RTN_MULTICAST &&
344 chk_addr_ret != RTN_BROADCAST) {
345 err = -EADDRNOTAVAIL;
346 goto out;
347 }
348 } else {
349 if (addr_type != IPV6_ADDR_ANY) {
350 struct net_device *dev = NULL;
351
352 rcu_read_lock();
353 if (__ipv6_addr_needs_scope_id(addr_type)) {
354 if (addr_len >= sizeof(struct sockaddr_in6) &&
355 addr->sin6_scope_id) {
356 /* Override any existing binding, if another one
357 * is supplied by user.
358 */
359 sk->sk_bound_dev_if = addr->sin6_scope_id;
360 }
361
362 /* Binding to link-local address requires an interface */
363 if (!sk->sk_bound_dev_if) {
364 err = -EINVAL;
365 goto out_unlock;
366 }
367 }
368
369 if (sk->sk_bound_dev_if) {
370 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
371 if (!dev) {
372 err = -ENODEV;
373 goto out_unlock;
374 }
375 }
376
377 /* ipv4 addr of the socket is invalid. Only the
378 * unspecified and mapped address have a v4 equivalent.
379 */
380 v4addr = LOOPBACK4_IPV6;
381 if (!(addr_type & IPV6_ADDR_MULTICAST)) {
382 if (!ipv6_can_nonlocal_bind(net, inet) &&
383 !ipv6_chk_addr(net, &addr->sin6_addr,
384 dev, 0)) {
385 err = -EADDRNOTAVAIL;
386 goto out_unlock;
387 }
388 }
389 rcu_read_unlock();
390 }
391 }
392
393 inet->inet_rcv_saddr = v4addr;
394 inet->inet_saddr = v4addr;
395
396 sk->sk_v6_rcv_saddr = addr->sin6_addr;
397
398 if (!(addr_type & IPV6_ADDR_MULTICAST))
399 np->saddr = addr->sin6_addr;
400
401 saved_ipv6only = sk->sk_ipv6only;
402 if (addr_type != IPV6_ADDR_ANY && addr_type != IPV6_ADDR_MAPPED)
403 sk->sk_ipv6only = 1;
404
405 /* Make sure we are allowed to bind here. */
406 if (snum || !(inet->bind_address_no_port ||
407 (flags & BIND_FORCE_ADDRESS_NO_PORT))) {
408 if (sk->sk_prot->get_port(sk, snum)) {
409 sk->sk_ipv6only = saved_ipv6only;
410 inet_reset_saddr(sk);
411 err = -EADDRINUSE;
412 goto out;
413 }
414 if (!(flags & BIND_FROM_BPF)) {
415 err = BPF_CGROUP_RUN_PROG_INET6_POST_BIND(sk);
416 if (err) {
417 sk->sk_ipv6only = saved_ipv6only;
418 inet_reset_saddr(sk);
419 goto out;
420 }
421 }
422 }
423
424 if (addr_type != IPV6_ADDR_ANY)
425 sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
426 if (snum)
427 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
428 inet->inet_sport = htons(inet->inet_num);
429 inet->inet_dport = 0;
430 inet->inet_daddr = 0;
431 out:
432 if (flags & BIND_WITH_LOCK)
433 release_sock(sk);
434 return err;
435 out_unlock:
436 rcu_read_unlock();
437 goto out;
438 }
439
440 /* bind for INET6 API */
inet6_bind(struct socket * sock,struct sockaddr * uaddr,int addr_len)441 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
442 {
443 struct sock *sk = sock->sk;
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_INET6_BIND(sk, uaddr);
457 if (err)
458 return err;
459
460 return __inet6_bind(sk, uaddr, addr_len, BIND_WITH_LOCK);
461 }
462 EXPORT_SYMBOL(inet6_bind);
463
inet6_release(struct socket * sock)464 int inet6_release(struct socket *sock)
465 {
466 struct sock *sk = sock->sk;
467
468 if (!sk)
469 return -EINVAL;
470
471 /* Free mc lists */
472 ipv6_sock_mc_close(sk);
473
474 /* Free ac lists */
475 ipv6_sock_ac_close(sk);
476
477 return inet_release(sock);
478 }
479 EXPORT_SYMBOL(inet6_release);
480
inet6_destroy_sock(struct sock * sk)481 void inet6_destroy_sock(struct sock *sk)
482 {
483 struct ipv6_pinfo *np = inet6_sk(sk);
484 struct sk_buff *skb;
485 struct ipv6_txoptions *opt;
486
487 /* Release rx options */
488
489 skb = xchg(&np->pktoptions, NULL);
490 kfree_skb(skb);
491
492 skb = xchg(&np->rxpmtu, NULL);
493 kfree_skb(skb);
494
495 /* Free flowlabels */
496 fl6_free_socklist(sk);
497
498 /* Free tx options */
499
500 opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL);
501 if (opt) {
502 atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
503 txopt_put(opt);
504 }
505 }
506 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
507
508 /*
509 * This does both peername and sockname.
510 */
inet6_getname(struct socket * sock,struct sockaddr * uaddr,int peer)511 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
512 int peer)
513 {
514 struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
515 struct sock *sk = sock->sk;
516 struct inet_sock *inet = inet_sk(sk);
517 struct ipv6_pinfo *np = inet6_sk(sk);
518
519 sin->sin6_family = AF_INET6;
520 sin->sin6_flowinfo = 0;
521 sin->sin6_scope_id = 0;
522 if (peer) {
523 if (!inet->inet_dport)
524 return -ENOTCONN;
525 if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
526 peer == 1)
527 return -ENOTCONN;
528 sin->sin6_port = inet->inet_dport;
529 sin->sin6_addr = sk->sk_v6_daddr;
530 if (np->sndflow)
531 sin->sin6_flowinfo = np->flow_label;
532 } else {
533 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
534 sin->sin6_addr = np->saddr;
535 else
536 sin->sin6_addr = sk->sk_v6_rcv_saddr;
537 sin->sin6_port = inet->inet_sport;
538 }
539 if (cgroup_bpf_enabled)
540 BPF_CGROUP_RUN_SA_PROG_LOCK(sk, (struct sockaddr *)sin,
541 peer ? BPF_CGROUP_INET6_GETPEERNAME :
542 BPF_CGROUP_INET6_GETSOCKNAME,
543 NULL);
544 sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
545 sk->sk_bound_dev_if);
546 return sizeof(*sin);
547 }
548 EXPORT_SYMBOL(inet6_getname);
549
inet6_ioctl(struct socket * sock,unsigned int cmd,unsigned long arg)550 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
551 {
552 void __user *argp = (void __user *)arg;
553 struct sock *sk = sock->sk;
554 struct net *net = sock_net(sk);
555
556 switch (cmd) {
557 case SIOCADDRT:
558 case SIOCDELRT: {
559 struct in6_rtmsg rtmsg;
560
561 if (copy_from_user(&rtmsg, argp, sizeof(rtmsg)))
562 return -EFAULT;
563 return ipv6_route_ioctl(net, cmd, &rtmsg);
564 }
565 case SIOCSIFADDR:
566 return addrconf_add_ifaddr(net, argp);
567 case SIOCDIFADDR:
568 return addrconf_del_ifaddr(net, argp);
569 case SIOCSIFDSTADDR:
570 return addrconf_set_dstaddr(net, argp);
571 default:
572 if (!sk->sk_prot->ioctl)
573 return -ENOIOCTLCMD;
574 return sk->sk_prot->ioctl(sk, cmd, arg);
575 }
576 /*NOTREACHED*/
577 return 0;
578 }
579 EXPORT_SYMBOL(inet6_ioctl);
580
581 #ifdef CONFIG_COMPAT
582 struct compat_in6_rtmsg {
583 struct in6_addr rtmsg_dst;
584 struct in6_addr rtmsg_src;
585 struct in6_addr rtmsg_gateway;
586 u32 rtmsg_type;
587 u16 rtmsg_dst_len;
588 u16 rtmsg_src_len;
589 u32 rtmsg_metric;
590 u32 rtmsg_info;
591 u32 rtmsg_flags;
592 s32 rtmsg_ifindex;
593 };
594
inet6_compat_routing_ioctl(struct sock * sk,unsigned int cmd,struct compat_in6_rtmsg __user * ur)595 static int inet6_compat_routing_ioctl(struct sock *sk, unsigned int cmd,
596 struct compat_in6_rtmsg __user *ur)
597 {
598 struct in6_rtmsg rt;
599
600 if (copy_from_user(&rt.rtmsg_dst, &ur->rtmsg_dst,
601 3 * sizeof(struct in6_addr)) ||
602 get_user(rt.rtmsg_type, &ur->rtmsg_type) ||
603 get_user(rt.rtmsg_dst_len, &ur->rtmsg_dst_len) ||
604 get_user(rt.rtmsg_src_len, &ur->rtmsg_src_len) ||
605 get_user(rt.rtmsg_metric, &ur->rtmsg_metric) ||
606 get_user(rt.rtmsg_info, &ur->rtmsg_info) ||
607 get_user(rt.rtmsg_flags, &ur->rtmsg_flags) ||
608 get_user(rt.rtmsg_ifindex, &ur->rtmsg_ifindex))
609 return -EFAULT;
610
611
612 return ipv6_route_ioctl(sock_net(sk), cmd, &rt);
613 }
614
inet6_compat_ioctl(struct socket * sock,unsigned int cmd,unsigned long arg)615 int inet6_compat_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
616 {
617 void __user *argp = compat_ptr(arg);
618 struct sock *sk = sock->sk;
619
620 switch (cmd) {
621 case SIOCADDRT:
622 case SIOCDELRT:
623 return inet6_compat_routing_ioctl(sk, cmd, argp);
624 default:
625 return -ENOIOCTLCMD;
626 }
627 }
628 EXPORT_SYMBOL_GPL(inet6_compat_ioctl);
629 #endif /* CONFIG_COMPAT */
630
631 INDIRECT_CALLABLE_DECLARE(int udpv6_sendmsg(struct sock *, struct msghdr *,
632 size_t));
inet6_sendmsg(struct socket * sock,struct msghdr * msg,size_t size)633 int inet6_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
634 {
635 struct sock *sk = sock->sk;
636
637 if (unlikely(inet_send_prepare(sk)))
638 return -EAGAIN;
639
640 return INDIRECT_CALL_2(sk->sk_prot->sendmsg, tcp_sendmsg, udpv6_sendmsg,
641 sk, msg, size);
642 }
643
644 INDIRECT_CALLABLE_DECLARE(int udpv6_recvmsg(struct sock *, struct msghdr *,
645 size_t, int, int, int *));
inet6_recvmsg(struct socket * sock,struct msghdr * msg,size_t size,int flags)646 int inet6_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
647 int flags)
648 {
649 struct sock *sk = sock->sk;
650 int addr_len = 0;
651 int err;
652
653 if (likely(!(flags & MSG_ERRQUEUE)))
654 sock_rps_record_flow(sk);
655
656 err = INDIRECT_CALL_2(sk->sk_prot->recvmsg, tcp_recvmsg, udpv6_recvmsg,
657 sk, msg, size, flags & MSG_DONTWAIT,
658 flags & ~MSG_DONTWAIT, &addr_len);
659 if (err >= 0)
660 msg->msg_namelen = addr_len;
661 return err;
662 }
663
664 const struct proto_ops inet6_stream_ops = {
665 .family = PF_INET6,
666 .owner = THIS_MODULE,
667 .release = inet6_release,
668 .bind = inet6_bind,
669 .connect = inet_stream_connect, /* ok */
670 .socketpair = sock_no_socketpair, /* a do nothing */
671 .accept = inet_accept, /* ok */
672 .getname = inet6_getname,
673 .poll = tcp_poll, /* ok */
674 .ioctl = inet6_ioctl, /* must change */
675 .gettstamp = sock_gettstamp,
676 .listen = inet_listen, /* ok */
677 .shutdown = inet_shutdown, /* ok */
678 .setsockopt = sock_common_setsockopt, /* ok */
679 .getsockopt = sock_common_getsockopt, /* ok */
680 .sendmsg = inet6_sendmsg, /* retpoline's sake */
681 .recvmsg = inet6_recvmsg, /* retpoline's sake */
682 #ifdef CONFIG_MMU
683 .mmap = tcp_mmap,
684 #endif
685 .sendpage = inet_sendpage,
686 .sendmsg_locked = tcp_sendmsg_locked,
687 .sendpage_locked = tcp_sendpage_locked,
688 .splice_read = tcp_splice_read,
689 .read_sock = tcp_read_sock,
690 .peek_len = tcp_peek_len,
691 #ifdef CONFIG_COMPAT
692 .compat_ioctl = inet6_compat_ioctl,
693 #endif
694 .set_rcvlowat = tcp_set_rcvlowat,
695 };
696
697 const struct proto_ops inet6_dgram_ops = {
698 .family = PF_INET6,
699 .owner = THIS_MODULE,
700 .release = inet6_release,
701 .bind = inet6_bind,
702 .connect = inet_dgram_connect, /* ok */
703 .socketpair = sock_no_socketpair, /* a do nothing */
704 .accept = sock_no_accept, /* a do nothing */
705 .getname = inet6_getname,
706 .poll = udp_poll, /* ok */
707 .ioctl = inet6_ioctl, /* must change */
708 .gettstamp = sock_gettstamp,
709 .listen = sock_no_listen, /* ok */
710 .shutdown = inet_shutdown, /* ok */
711 .setsockopt = sock_common_setsockopt, /* ok */
712 .getsockopt = sock_common_getsockopt, /* ok */
713 .sendmsg = inet6_sendmsg, /* retpoline's sake */
714 .recvmsg = inet6_recvmsg, /* retpoline's sake */
715 .mmap = sock_no_mmap,
716 .sendpage = sock_no_sendpage,
717 .set_peek_off = sk_set_peek_off,
718 #ifdef CONFIG_COMPAT
719 .compat_ioctl = inet6_compat_ioctl,
720 #endif
721 };
722
723 static const struct net_proto_family inet6_family_ops = {
724 .family = PF_INET6,
725 .create = inet6_create,
726 .owner = THIS_MODULE,
727 };
728
inet6_register_protosw(struct inet_protosw * p)729 int inet6_register_protosw(struct inet_protosw *p)
730 {
731 struct list_head *lh;
732 struct inet_protosw *answer;
733 struct list_head *last_perm;
734 int protocol = p->protocol;
735 int ret;
736
737 spin_lock_bh(&inetsw6_lock);
738
739 ret = -EINVAL;
740 if (p->type >= SOCK_MAX)
741 goto out_illegal;
742
743 /* If we are trying to override a permanent protocol, bail. */
744 answer = NULL;
745 ret = -EPERM;
746 last_perm = &inetsw6[p->type];
747 list_for_each(lh, &inetsw6[p->type]) {
748 answer = list_entry(lh, struct inet_protosw, list);
749
750 /* Check only the non-wild match. */
751 if (INET_PROTOSW_PERMANENT & answer->flags) {
752 if (protocol == answer->protocol)
753 break;
754 last_perm = lh;
755 }
756
757 answer = NULL;
758 }
759 if (answer)
760 goto out_permanent;
761
762 /* Add the new entry after the last permanent entry if any, so that
763 * the new entry does not override a permanent entry when matched with
764 * a wild-card protocol. But it is allowed to override any existing
765 * non-permanent entry. This means that when we remove this entry, the
766 * system automatically returns to the old behavior.
767 */
768 list_add_rcu(&p->list, last_perm);
769 ret = 0;
770 out:
771 spin_unlock_bh(&inetsw6_lock);
772 return ret;
773
774 out_permanent:
775 pr_err("Attempt to override permanent protocol %d\n", protocol);
776 goto out;
777
778 out_illegal:
779 pr_err("Ignoring attempt to register invalid socket type %d\n",
780 p->type);
781 goto out;
782 }
783 EXPORT_SYMBOL(inet6_register_protosw);
784
785 void
inet6_unregister_protosw(struct inet_protosw * p)786 inet6_unregister_protosw(struct inet_protosw *p)
787 {
788 if (INET_PROTOSW_PERMANENT & p->flags) {
789 pr_err("Attempt to unregister permanent protocol %d\n",
790 p->protocol);
791 } else {
792 spin_lock_bh(&inetsw6_lock);
793 list_del_rcu(&p->list);
794 spin_unlock_bh(&inetsw6_lock);
795
796 synchronize_net();
797 }
798 }
799 EXPORT_SYMBOL(inet6_unregister_protosw);
800
inet6_sk_rebuild_header(struct sock * sk)801 int inet6_sk_rebuild_header(struct sock *sk)
802 {
803 struct ipv6_pinfo *np = inet6_sk(sk);
804 struct dst_entry *dst;
805
806 dst = __sk_dst_check(sk, np->dst_cookie);
807
808 if (!dst) {
809 struct inet_sock *inet = inet_sk(sk);
810 struct in6_addr *final_p, final;
811 struct flowi6 fl6;
812
813 memset(&fl6, 0, sizeof(fl6));
814 fl6.flowi6_proto = sk->sk_protocol;
815 fl6.daddr = sk->sk_v6_daddr;
816 fl6.saddr = np->saddr;
817 fl6.flowlabel = np->flow_label;
818 fl6.flowi6_oif = sk->sk_bound_dev_if;
819 fl6.flowi6_mark = sk->sk_mark;
820 fl6.fl6_dport = inet->inet_dport;
821 fl6.fl6_sport = inet->inet_sport;
822 fl6.flowi6_uid = sk->sk_uid;
823 security_sk_classify_flow(sk, flowi6_to_flowi_common(&fl6));
824
825 rcu_read_lock();
826 final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt),
827 &final);
828 rcu_read_unlock();
829
830 dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p);
831 if (IS_ERR(dst)) {
832 sk->sk_route_caps = 0;
833 sk->sk_err_soft = -PTR_ERR(dst);
834 return PTR_ERR(dst);
835 }
836
837 ip6_dst_store(sk, dst, NULL, NULL);
838 }
839
840 return 0;
841 }
842 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
843
ipv6_opt_accepted(const struct sock * sk,const struct sk_buff * skb,const struct inet6_skb_parm * opt)844 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb,
845 const struct inet6_skb_parm *opt)
846 {
847 const struct ipv6_pinfo *np = inet6_sk(sk);
848
849 if (np->rxopt.all) {
850 if (((opt->flags & IP6SKB_HOPBYHOP) &&
851 (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) ||
852 (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) &&
853 np->rxopt.bits.rxflow) ||
854 (opt->srcrt && (np->rxopt.bits.srcrt ||
855 np->rxopt.bits.osrcrt)) ||
856 ((opt->dst1 || opt->dst0) &&
857 (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
858 return true;
859 }
860 return false;
861 }
862 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
863
864 static struct packet_type ipv6_packet_type __read_mostly = {
865 .type = cpu_to_be16(ETH_P_IPV6),
866 .func = ipv6_rcv,
867 .list_func = ipv6_list_rcv,
868 };
869
ipv6_packet_init(void)870 static int __init ipv6_packet_init(void)
871 {
872 dev_add_pack(&ipv6_packet_type);
873 return 0;
874 }
875
ipv6_packet_cleanup(void)876 static void ipv6_packet_cleanup(void)
877 {
878 dev_remove_pack(&ipv6_packet_type);
879 }
880
ipv6_init_mibs(struct net * net)881 static int __net_init ipv6_init_mibs(struct net *net)
882 {
883 int i;
884
885 net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib);
886 if (!net->mib.udp_stats_in6)
887 return -ENOMEM;
888 net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib);
889 if (!net->mib.udplite_stats_in6)
890 goto err_udplite_mib;
891 net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib);
892 if (!net->mib.ipv6_statistics)
893 goto err_ip_mib;
894
895 for_each_possible_cpu(i) {
896 struct ipstats_mib *af_inet6_stats;
897 af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i);
898 u64_stats_init(&af_inet6_stats->syncp);
899 }
900
901
902 net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib);
903 if (!net->mib.icmpv6_statistics)
904 goto err_icmp_mib;
905 net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
906 GFP_KERNEL);
907 if (!net->mib.icmpv6msg_statistics)
908 goto err_icmpmsg_mib;
909 return 0;
910
911 err_icmpmsg_mib:
912 free_percpu(net->mib.icmpv6_statistics);
913 err_icmp_mib:
914 free_percpu(net->mib.ipv6_statistics);
915 err_ip_mib:
916 free_percpu(net->mib.udplite_stats_in6);
917 err_udplite_mib:
918 free_percpu(net->mib.udp_stats_in6);
919 return -ENOMEM;
920 }
921
ipv6_cleanup_mibs(struct net * net)922 static void ipv6_cleanup_mibs(struct net *net)
923 {
924 free_percpu(net->mib.udp_stats_in6);
925 free_percpu(net->mib.udplite_stats_in6);
926 free_percpu(net->mib.ipv6_statistics);
927 free_percpu(net->mib.icmpv6_statistics);
928 kfree(net->mib.icmpv6msg_statistics);
929 }
930
inet6_net_init(struct net * net)931 static int __net_init inet6_net_init(struct net *net)
932 {
933 int err = 0;
934
935 net->ipv6.sysctl.bindv6only = 0;
936 net->ipv6.sysctl.icmpv6_time = 1*HZ;
937 net->ipv6.sysctl.icmpv6_echo_ignore_all = 0;
938 net->ipv6.sysctl.icmpv6_echo_ignore_multicast = 0;
939 net->ipv6.sysctl.icmpv6_echo_ignore_anycast = 0;
940
941 /* By default, rate limit error messages.
942 * Except for pmtu discovery, it would break it.
943 * proc_do_large_bitmap needs pointer to the bitmap.
944 */
945 bitmap_set(net->ipv6.sysctl.icmpv6_ratemask, 0, ICMPV6_ERRMSG_MAX + 1);
946 bitmap_clear(net->ipv6.sysctl.icmpv6_ratemask, ICMPV6_PKT_TOOBIG, 1);
947 net->ipv6.sysctl.icmpv6_ratemask_ptr = net->ipv6.sysctl.icmpv6_ratemask;
948
949 net->ipv6.sysctl.flowlabel_consistency = 1;
950 net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS;
951 net->ipv6.sysctl.idgen_retries = 3;
952 net->ipv6.sysctl.idgen_delay = 1 * HZ;
953 net->ipv6.sysctl.flowlabel_state_ranges = 0;
954 net->ipv6.sysctl.max_dst_opts_cnt = IP6_DEFAULT_MAX_DST_OPTS_CNT;
955 net->ipv6.sysctl.max_hbh_opts_cnt = IP6_DEFAULT_MAX_HBH_OPTS_CNT;
956 net->ipv6.sysctl.max_dst_opts_len = IP6_DEFAULT_MAX_DST_OPTS_LEN;
957 net->ipv6.sysctl.max_hbh_opts_len = IP6_DEFAULT_MAX_HBH_OPTS_LEN;
958 atomic_set(&net->ipv6.fib6_sernum, 1);
959
960 err = ipv6_init_mibs(net);
961 if (err)
962 return err;
963 #ifdef CONFIG_PROC_FS
964 err = udp6_proc_init(net);
965 if (err)
966 goto out;
967 err = tcp6_proc_init(net);
968 if (err)
969 goto proc_tcp6_fail;
970 err = ac6_proc_init(net);
971 if (err)
972 goto proc_ac6_fail;
973 #endif
974 return err;
975
976 #ifdef CONFIG_PROC_FS
977 proc_ac6_fail:
978 tcp6_proc_exit(net);
979 proc_tcp6_fail:
980 udp6_proc_exit(net);
981 out:
982 ipv6_cleanup_mibs(net);
983 return err;
984 #endif
985 }
986
inet6_net_exit(struct net * net)987 static void __net_exit inet6_net_exit(struct net *net)
988 {
989 #ifdef CONFIG_PROC_FS
990 udp6_proc_exit(net);
991 tcp6_proc_exit(net);
992 ac6_proc_exit(net);
993 #endif
994 ipv6_cleanup_mibs(net);
995 }
996
997 static struct pernet_operations inet6_net_ops = {
998 .init = inet6_net_init,
999 .exit = inet6_net_exit,
1000 };
1001
ipv6_route_input(struct sk_buff * skb)1002 static int ipv6_route_input(struct sk_buff *skb)
1003 {
1004 ip6_route_input(skb);
1005 return skb_dst(skb)->error;
1006 }
1007
1008 static const struct ipv6_stub ipv6_stub_impl = {
1009 .ipv6_sock_mc_join = ipv6_sock_mc_join,
1010 .ipv6_sock_mc_drop = ipv6_sock_mc_drop,
1011 .ipv6_dst_lookup_flow = ip6_dst_lookup_flow,
1012 .ipv6_route_input = ipv6_route_input,
1013 .fib6_get_table = fib6_get_table,
1014 .fib6_table_lookup = fib6_table_lookup,
1015 .fib6_lookup = fib6_lookup,
1016 .fib6_select_path = fib6_select_path,
1017 .ip6_mtu_from_fib6 = ip6_mtu_from_fib6,
1018 .fib6_nh_init = fib6_nh_init,
1019 .fib6_nh_release = fib6_nh_release,
1020 .fib6_update_sernum = fib6_update_sernum_stub,
1021 .fib6_rt_update = fib6_rt_update,
1022 .ip6_del_rt = ip6_del_rt,
1023 .udpv6_encap_enable = udpv6_encap_enable,
1024 .ndisc_send_na = ndisc_send_na,
1025 #if IS_ENABLED(CONFIG_XFRM)
1026 .xfrm6_local_rxpmtu = xfrm6_local_rxpmtu,
1027 .xfrm6_udp_encap_rcv = xfrm6_udp_encap_rcv,
1028 .xfrm6_rcv_encap = xfrm6_rcv_encap,
1029 #endif
1030 .nd_tbl = &nd_tbl,
1031 .ipv6_fragment = ip6_fragment,
1032 };
1033
1034 static const struct ipv6_bpf_stub ipv6_bpf_stub_impl = {
1035 .inet6_bind = __inet6_bind,
1036 .udp6_lib_lookup = __udp6_lib_lookup,
1037 };
1038
inet6_init(void)1039 static int __init inet6_init(void)
1040 {
1041 struct list_head *r;
1042 int err = 0;
1043
1044 sock_skb_cb_check_size(sizeof(struct inet6_skb_parm));
1045
1046 /* Register the socket-side information for inet6_create. */
1047 for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
1048 INIT_LIST_HEAD(r);
1049
1050 if (disable_ipv6_mod) {
1051 pr_info("Loaded, but administratively disabled, reboot required to enable\n");
1052 goto out;
1053 }
1054
1055 err = proto_register(&tcpv6_prot, 1);
1056 if (err)
1057 goto out;
1058
1059 err = proto_register(&udpv6_prot, 1);
1060 if (err)
1061 goto out_unregister_tcp_proto;
1062
1063 err = proto_register(&udplitev6_prot, 1);
1064 if (err)
1065 goto out_unregister_udp_proto;
1066
1067 err = proto_register(&rawv6_prot, 1);
1068 if (err)
1069 goto out_unregister_udplite_proto;
1070
1071 err = proto_register(&pingv6_prot, 1);
1072 if (err)
1073 goto out_unregister_raw_proto;
1074
1075 /* We MUST register RAW sockets before we create the ICMP6,
1076 * IGMP6, or NDISC control sockets.
1077 */
1078 err = rawv6_init();
1079 if (err)
1080 goto out_unregister_ping_proto;
1081
1082 /* Register the family here so that the init calls below will
1083 * be able to create sockets. (?? is this dangerous ??)
1084 */
1085 err = sock_register(&inet6_family_ops);
1086 if (err)
1087 goto out_sock_register_fail;
1088
1089 /*
1090 * ipngwg API draft makes clear that the correct semantics
1091 * for TCP and UDP is to consider one TCP and UDP instance
1092 * in a host available by both INET and INET6 APIs and
1093 * able to communicate via both network protocols.
1094 */
1095
1096 err = register_pernet_subsys(&inet6_net_ops);
1097 if (err)
1098 goto register_pernet_fail;
1099 err = ip6_mr_init();
1100 if (err)
1101 goto ipmr_fail;
1102 err = icmpv6_init();
1103 if (err)
1104 goto icmp_fail;
1105 err = ndisc_init();
1106 if (err)
1107 goto ndisc_fail;
1108 err = igmp6_init();
1109 if (err)
1110 goto igmp_fail;
1111
1112 err = ipv6_netfilter_init();
1113 if (err)
1114 goto netfilter_fail;
1115 /* Create /proc/foo6 entries. */
1116 #ifdef CONFIG_PROC_FS
1117 err = -ENOMEM;
1118 if (raw6_proc_init())
1119 goto proc_raw6_fail;
1120 if (udplite6_proc_init())
1121 goto proc_udplite6_fail;
1122 if (ipv6_misc_proc_init())
1123 goto proc_misc6_fail;
1124 if (if6_proc_init())
1125 goto proc_if6_fail;
1126 #endif
1127 err = ip6_route_init();
1128 if (err)
1129 goto ip6_route_fail;
1130 err = ndisc_late_init();
1131 if (err)
1132 goto ndisc_late_fail;
1133 err = ip6_flowlabel_init();
1134 if (err)
1135 goto ip6_flowlabel_fail;
1136 err = ipv6_anycast_init();
1137 if (err)
1138 goto ipv6_anycast_fail;
1139 err = addrconf_init();
1140 if (err)
1141 goto addrconf_fail;
1142
1143 /* Init v6 extension headers. */
1144 err = ipv6_exthdrs_init();
1145 if (err)
1146 goto ipv6_exthdrs_fail;
1147
1148 err = ipv6_frag_init();
1149 if (err)
1150 goto ipv6_frag_fail;
1151
1152 /* Init v6 transport protocols. */
1153 err = udpv6_init();
1154 if (err)
1155 goto udpv6_fail;
1156
1157 err = udplitev6_init();
1158 if (err)
1159 goto udplitev6_fail;
1160
1161 err = udpv6_offload_init();
1162 if (err)
1163 goto udpv6_offload_fail;
1164
1165 err = tcpv6_init();
1166 if (err)
1167 goto tcpv6_fail;
1168
1169 err = ipv6_packet_init();
1170 if (err)
1171 goto ipv6_packet_fail;
1172
1173 err = pingv6_init();
1174 if (err)
1175 goto pingv6_fail;
1176
1177 err = calipso_init();
1178 if (err)
1179 goto calipso_fail;
1180
1181 err = seg6_init();
1182 if (err)
1183 goto seg6_fail;
1184
1185 err = rpl_init();
1186 if (err)
1187 goto rpl_fail;
1188
1189 err = igmp6_late_init();
1190 if (err)
1191 goto igmp6_late_err;
1192
1193 #ifdef CONFIG_SYSCTL
1194 err = ipv6_sysctl_register();
1195 if (err)
1196 goto sysctl_fail;
1197 #endif
1198
1199 /* ensure that ipv6 stubs are visible only after ipv6 is ready */
1200 wmb();
1201 ipv6_stub = &ipv6_stub_impl;
1202 ipv6_bpf_stub = &ipv6_bpf_stub_impl;
1203 out:
1204 return err;
1205
1206 #ifdef CONFIG_SYSCTL
1207 sysctl_fail:
1208 igmp6_late_cleanup();
1209 #endif
1210 igmp6_late_err:
1211 rpl_exit();
1212 rpl_fail:
1213 seg6_exit();
1214 seg6_fail:
1215 calipso_exit();
1216 calipso_fail:
1217 pingv6_exit();
1218 pingv6_fail:
1219 ipv6_packet_cleanup();
1220 ipv6_packet_fail:
1221 tcpv6_exit();
1222 tcpv6_fail:
1223 udpv6_offload_exit();
1224 udpv6_offload_fail:
1225 udplitev6_exit();
1226 udplitev6_fail:
1227 udpv6_exit();
1228 udpv6_fail:
1229 ipv6_frag_exit();
1230 ipv6_frag_fail:
1231 ipv6_exthdrs_exit();
1232 ipv6_exthdrs_fail:
1233 addrconf_cleanup();
1234 addrconf_fail:
1235 ipv6_anycast_cleanup();
1236 ipv6_anycast_fail:
1237 ip6_flowlabel_cleanup();
1238 ip6_flowlabel_fail:
1239 ndisc_late_cleanup();
1240 ndisc_late_fail:
1241 ip6_route_cleanup();
1242 ip6_route_fail:
1243 #ifdef CONFIG_PROC_FS
1244 if6_proc_exit();
1245 proc_if6_fail:
1246 ipv6_misc_proc_exit();
1247 proc_misc6_fail:
1248 udplite6_proc_exit();
1249 proc_udplite6_fail:
1250 raw6_proc_exit();
1251 proc_raw6_fail:
1252 #endif
1253 ipv6_netfilter_fini();
1254 netfilter_fail:
1255 igmp6_cleanup();
1256 igmp_fail:
1257 ndisc_cleanup();
1258 ndisc_fail:
1259 icmpv6_cleanup();
1260 icmp_fail:
1261 ip6_mr_cleanup();
1262 ipmr_fail:
1263 unregister_pernet_subsys(&inet6_net_ops);
1264 register_pernet_fail:
1265 sock_unregister(PF_INET6);
1266 rtnl_unregister_all(PF_INET6);
1267 out_sock_register_fail:
1268 rawv6_exit();
1269 out_unregister_ping_proto:
1270 proto_unregister(&pingv6_prot);
1271 out_unregister_raw_proto:
1272 proto_unregister(&rawv6_prot);
1273 out_unregister_udplite_proto:
1274 proto_unregister(&udplitev6_prot);
1275 out_unregister_udp_proto:
1276 proto_unregister(&udpv6_prot);
1277 out_unregister_tcp_proto:
1278 proto_unregister(&tcpv6_prot);
1279 goto out;
1280 }
1281 module_init(inet6_init);
1282
1283 MODULE_ALIAS_NETPROTO(PF_INET6);
1284