xref: /linux/net/ipv6/af_inet6.c (revision 3d7558bf2603159a51188842a0e221f2eaff5d72)
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 
94 bool ipv6_mod_enabled(void)
95 {
96 	return disable_ipv6_mod == 0;
97 }
98 EXPORT_SYMBOL_GPL(ipv6_mod_enabled);
99 
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 
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 
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_port_requires_bind_service(net, snum) &&
296 	    !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
297 		return -EACCES;
298 
299 	if (with_lock)
300 		lock_sock(sk);
301 
302 	/* Check these errors (active socket, double bind). */
303 	if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
304 		err = -EINVAL;
305 		goto out;
306 	}
307 
308 	/* Check if the address belongs to the host. */
309 	if (addr_type == IPV6_ADDR_MAPPED) {
310 		struct net_device *dev = NULL;
311 		int chk_addr_ret;
312 
313 		/* Binding to v4-mapped address on a v6-only socket
314 		 * makes no sense
315 		 */
316 		if (sk->sk_ipv6only) {
317 			err = -EINVAL;
318 			goto out;
319 		}
320 
321 		rcu_read_lock();
322 		if (sk->sk_bound_dev_if) {
323 			dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
324 			if (!dev) {
325 				err = -ENODEV;
326 				goto out_unlock;
327 			}
328 		}
329 
330 		/* Reproduce AF_INET checks to make the bindings consistent */
331 		v4addr = addr->sin6_addr.s6_addr32[3];
332 		chk_addr_ret = inet_addr_type_dev_table(net, dev, v4addr);
333 		rcu_read_unlock();
334 
335 		if (!inet_can_nonlocal_bind(net, inet) &&
336 		    v4addr != htonl(INADDR_ANY) &&
337 		    chk_addr_ret != RTN_LOCAL &&
338 		    chk_addr_ret != RTN_MULTICAST &&
339 		    chk_addr_ret != RTN_BROADCAST) {
340 			err = -EADDRNOTAVAIL;
341 			goto out;
342 		}
343 	} else {
344 		if (addr_type != IPV6_ADDR_ANY) {
345 			struct net_device *dev = NULL;
346 
347 			rcu_read_lock();
348 			if (__ipv6_addr_needs_scope_id(addr_type)) {
349 				if (addr_len >= sizeof(struct sockaddr_in6) &&
350 				    addr->sin6_scope_id) {
351 					/* Override any existing binding, if another one
352 					 * is supplied by user.
353 					 */
354 					sk->sk_bound_dev_if = addr->sin6_scope_id;
355 				}
356 
357 				/* Binding to link-local address requires an interface */
358 				if (!sk->sk_bound_dev_if) {
359 					err = -EINVAL;
360 					goto out_unlock;
361 				}
362 			}
363 
364 			if (sk->sk_bound_dev_if) {
365 				dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
366 				if (!dev) {
367 					err = -ENODEV;
368 					goto out_unlock;
369 				}
370 			}
371 
372 			/* ipv4 addr of the socket is invalid.  Only the
373 			 * unspecified and mapped address have a v4 equivalent.
374 			 */
375 			v4addr = LOOPBACK4_IPV6;
376 			if (!(addr_type & IPV6_ADDR_MULTICAST))	{
377 				if (!ipv6_can_nonlocal_bind(net, inet) &&
378 				    !ipv6_chk_addr(net, &addr->sin6_addr,
379 						   dev, 0)) {
380 					err = -EADDRNOTAVAIL;
381 					goto out_unlock;
382 				}
383 			}
384 			rcu_read_unlock();
385 		}
386 	}
387 
388 	inet->inet_rcv_saddr = v4addr;
389 	inet->inet_saddr = v4addr;
390 
391 	sk->sk_v6_rcv_saddr = addr->sin6_addr;
392 
393 	if (!(addr_type & IPV6_ADDR_MULTICAST))
394 		np->saddr = addr->sin6_addr;
395 
396 	saved_ipv6only = sk->sk_ipv6only;
397 	if (addr_type != IPV6_ADDR_ANY && addr_type != IPV6_ADDR_MAPPED)
398 		sk->sk_ipv6only = 1;
399 
400 	/* Make sure we are allowed to bind here. */
401 	if (snum || !(inet->bind_address_no_port ||
402 		      force_bind_address_no_port)) {
403 		if (sk->sk_prot->get_port(sk, snum)) {
404 			sk->sk_ipv6only = saved_ipv6only;
405 			inet_reset_saddr(sk);
406 			err = -EADDRINUSE;
407 			goto out;
408 		}
409 		err = BPF_CGROUP_RUN_PROG_INET6_POST_BIND(sk);
410 		if (err) {
411 			sk->sk_ipv6only = saved_ipv6only;
412 			inet_reset_saddr(sk);
413 			goto out;
414 		}
415 	}
416 
417 	if (addr_type != IPV6_ADDR_ANY)
418 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
419 	if (snum)
420 		sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
421 	inet->inet_sport = htons(inet->inet_num);
422 	inet->inet_dport = 0;
423 	inet->inet_daddr = 0;
424 out:
425 	if (with_lock)
426 		release_sock(sk);
427 	return err;
428 out_unlock:
429 	rcu_read_unlock();
430 	goto out;
431 }
432 
433 /* bind for INET6 API */
434 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
435 {
436 	struct sock *sk = sock->sk;
437 	int err = 0;
438 
439 	/* If the socket has its own bind function then use it. */
440 	if (sk->sk_prot->bind)
441 		return sk->sk_prot->bind(sk, uaddr, addr_len);
442 
443 	if (addr_len < SIN6_LEN_RFC2133)
444 		return -EINVAL;
445 
446 	/* BPF prog is run before any checks are done so that if the prog
447 	 * changes context in a wrong way it will be caught.
448 	 */
449 	err = BPF_CGROUP_RUN_PROG_INET6_BIND(sk, uaddr);
450 	if (err)
451 		return err;
452 
453 	return __inet6_bind(sk, uaddr, addr_len, false, true);
454 }
455 EXPORT_SYMBOL(inet6_bind);
456 
457 int inet6_release(struct socket *sock)
458 {
459 	struct sock *sk = sock->sk;
460 
461 	if (!sk)
462 		return -EINVAL;
463 
464 	/* Free mc lists */
465 	ipv6_sock_mc_close(sk);
466 
467 	/* Free ac lists */
468 	ipv6_sock_ac_close(sk);
469 
470 	return inet_release(sock);
471 }
472 EXPORT_SYMBOL(inet6_release);
473 
474 void inet6_destroy_sock(struct sock *sk)
475 {
476 	struct ipv6_pinfo *np = inet6_sk(sk);
477 	struct sk_buff *skb;
478 	struct ipv6_txoptions *opt;
479 
480 	/* Release rx options */
481 
482 	skb = xchg(&np->pktoptions, NULL);
483 	kfree_skb(skb);
484 
485 	skb = xchg(&np->rxpmtu, NULL);
486 	kfree_skb(skb);
487 
488 	/* Free flowlabels */
489 	fl6_free_socklist(sk);
490 
491 	/* Free tx options */
492 
493 	opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL);
494 	if (opt) {
495 		atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
496 		txopt_put(opt);
497 	}
498 }
499 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
500 
501 /*
502  *	This does both peername and sockname.
503  */
504 
505 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
506 		 int peer)
507 {
508 	struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
509 	struct sock *sk = sock->sk;
510 	struct inet_sock *inet = inet_sk(sk);
511 	struct ipv6_pinfo *np = inet6_sk(sk);
512 
513 	sin->sin6_family = AF_INET6;
514 	sin->sin6_flowinfo = 0;
515 	sin->sin6_scope_id = 0;
516 	if (peer) {
517 		if (!inet->inet_dport)
518 			return -ENOTCONN;
519 		if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
520 		    peer == 1)
521 			return -ENOTCONN;
522 		sin->sin6_port = inet->inet_dport;
523 		sin->sin6_addr = sk->sk_v6_daddr;
524 		if (np->sndflow)
525 			sin->sin6_flowinfo = np->flow_label;
526 	} else {
527 		if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
528 			sin->sin6_addr = np->saddr;
529 		else
530 			sin->sin6_addr = sk->sk_v6_rcv_saddr;
531 
532 		sin->sin6_port = inet->inet_sport;
533 	}
534 	sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
535 						 sk->sk_bound_dev_if);
536 	return sizeof(*sin);
537 }
538 EXPORT_SYMBOL(inet6_getname);
539 
540 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
541 {
542 	struct sock *sk = sock->sk;
543 	struct net *net = sock_net(sk);
544 
545 	switch (cmd) {
546 	case SIOCADDRT:
547 	case SIOCDELRT:
548 
549 		return ipv6_route_ioctl(net, cmd, (void __user *)arg);
550 
551 	case SIOCSIFADDR:
552 		return addrconf_add_ifaddr(net, (void __user *) arg);
553 	case SIOCDIFADDR:
554 		return addrconf_del_ifaddr(net, (void __user *) arg);
555 	case SIOCSIFDSTADDR:
556 		return addrconf_set_dstaddr(net, (void __user *) arg);
557 	default:
558 		if (!sk->sk_prot->ioctl)
559 			return -ENOIOCTLCMD;
560 		return sk->sk_prot->ioctl(sk, cmd, arg);
561 	}
562 	/*NOTREACHED*/
563 	return 0;
564 }
565 EXPORT_SYMBOL(inet6_ioctl);
566 
567 INDIRECT_CALLABLE_DECLARE(int udpv6_sendmsg(struct sock *, struct msghdr *,
568 					    size_t));
569 int inet6_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
570 {
571 	struct sock *sk = sock->sk;
572 
573 	if (unlikely(inet_send_prepare(sk)))
574 		return -EAGAIN;
575 
576 	return INDIRECT_CALL_2(sk->sk_prot->sendmsg, tcp_sendmsg, udpv6_sendmsg,
577 			       sk, msg, size);
578 }
579 
580 INDIRECT_CALLABLE_DECLARE(int udpv6_recvmsg(struct sock *, struct msghdr *,
581 					    size_t, int, int, int *));
582 int inet6_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
583 		  int flags)
584 {
585 	struct sock *sk = sock->sk;
586 	int addr_len = 0;
587 	int err;
588 
589 	if (likely(!(flags & MSG_ERRQUEUE)))
590 		sock_rps_record_flow(sk);
591 
592 	err = INDIRECT_CALL_2(sk->sk_prot->recvmsg, tcp_recvmsg, udpv6_recvmsg,
593 			      sk, msg, size, flags & MSG_DONTWAIT,
594 			      flags & ~MSG_DONTWAIT, &addr_len);
595 	if (err >= 0)
596 		msg->msg_namelen = addr_len;
597 	return err;
598 }
599 
600 const struct proto_ops inet6_stream_ops = {
601 	.family		   = PF_INET6,
602 	.owner		   = THIS_MODULE,
603 	.release	   = inet6_release,
604 	.bind		   = inet6_bind,
605 	.connect	   = inet_stream_connect,	/* ok		*/
606 	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
607 	.accept		   = inet_accept,		/* ok		*/
608 	.getname	   = inet6_getname,
609 	.poll		   = tcp_poll,			/* ok		*/
610 	.ioctl		   = inet6_ioctl,		/* must change  */
611 	.gettstamp	   = sock_gettstamp,
612 	.listen		   = inet_listen,		/* ok		*/
613 	.shutdown	   = inet_shutdown,		/* ok		*/
614 	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
615 	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
616 	.sendmsg	   = inet6_sendmsg,		/* retpoline's sake */
617 	.recvmsg	   = inet6_recvmsg,		/* retpoline's sake */
618 #ifdef CONFIG_MMU
619 	.mmap		   = tcp_mmap,
620 #endif
621 	.sendpage	   = inet_sendpage,
622 	.sendmsg_locked    = tcp_sendmsg_locked,
623 	.sendpage_locked   = tcp_sendpage_locked,
624 	.splice_read	   = tcp_splice_read,
625 	.read_sock	   = tcp_read_sock,
626 	.peek_len	   = tcp_peek_len,
627 #ifdef CONFIG_COMPAT
628 	.compat_setsockopt = compat_sock_common_setsockopt,
629 	.compat_getsockopt = compat_sock_common_getsockopt,
630 #endif
631 	.set_rcvlowat	   = tcp_set_rcvlowat,
632 };
633 
634 const struct proto_ops inet6_dgram_ops = {
635 	.family		   = PF_INET6,
636 	.owner		   = THIS_MODULE,
637 	.release	   = inet6_release,
638 	.bind		   = inet6_bind,
639 	.connect	   = inet_dgram_connect,	/* ok		*/
640 	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
641 	.accept		   = sock_no_accept,		/* a do nothing	*/
642 	.getname	   = inet6_getname,
643 	.poll		   = udp_poll,			/* ok		*/
644 	.ioctl		   = inet6_ioctl,		/* must change  */
645 	.gettstamp	   = sock_gettstamp,
646 	.listen		   = sock_no_listen,		/* ok		*/
647 	.shutdown	   = inet_shutdown,		/* ok		*/
648 	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
649 	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
650 	.sendmsg	   = inet6_sendmsg,		/* retpoline's sake */
651 	.recvmsg	   = inet6_recvmsg,		/* retpoline's sake */
652 	.mmap		   = sock_no_mmap,
653 	.sendpage	   = sock_no_sendpage,
654 	.set_peek_off	   = sk_set_peek_off,
655 #ifdef CONFIG_COMPAT
656 	.compat_setsockopt = compat_sock_common_setsockopt,
657 	.compat_getsockopt = compat_sock_common_getsockopt,
658 #endif
659 };
660 
661 static const struct net_proto_family inet6_family_ops = {
662 	.family = PF_INET6,
663 	.create = inet6_create,
664 	.owner	= THIS_MODULE,
665 };
666 
667 int inet6_register_protosw(struct inet_protosw *p)
668 {
669 	struct list_head *lh;
670 	struct inet_protosw *answer;
671 	struct list_head *last_perm;
672 	int protocol = p->protocol;
673 	int ret;
674 
675 	spin_lock_bh(&inetsw6_lock);
676 
677 	ret = -EINVAL;
678 	if (p->type >= SOCK_MAX)
679 		goto out_illegal;
680 
681 	/* If we are trying to override a permanent protocol, bail. */
682 	answer = NULL;
683 	ret = -EPERM;
684 	last_perm = &inetsw6[p->type];
685 	list_for_each(lh, &inetsw6[p->type]) {
686 		answer = list_entry(lh, struct inet_protosw, list);
687 
688 		/* Check only the non-wild match. */
689 		if (INET_PROTOSW_PERMANENT & answer->flags) {
690 			if (protocol == answer->protocol)
691 				break;
692 			last_perm = lh;
693 		}
694 
695 		answer = NULL;
696 	}
697 	if (answer)
698 		goto out_permanent;
699 
700 	/* Add the new entry after the last permanent entry if any, so that
701 	 * the new entry does not override a permanent entry when matched with
702 	 * a wild-card protocol. But it is allowed to override any existing
703 	 * non-permanent entry.  This means that when we remove this entry, the
704 	 * system automatically returns to the old behavior.
705 	 */
706 	list_add_rcu(&p->list, last_perm);
707 	ret = 0;
708 out:
709 	spin_unlock_bh(&inetsw6_lock);
710 	return ret;
711 
712 out_permanent:
713 	pr_err("Attempt to override permanent protocol %d\n", protocol);
714 	goto out;
715 
716 out_illegal:
717 	pr_err("Ignoring attempt to register invalid socket type %d\n",
718 	       p->type);
719 	goto out;
720 }
721 EXPORT_SYMBOL(inet6_register_protosw);
722 
723 void
724 inet6_unregister_protosw(struct inet_protosw *p)
725 {
726 	if (INET_PROTOSW_PERMANENT & p->flags) {
727 		pr_err("Attempt to unregister permanent protocol %d\n",
728 		       p->protocol);
729 	} else {
730 		spin_lock_bh(&inetsw6_lock);
731 		list_del_rcu(&p->list);
732 		spin_unlock_bh(&inetsw6_lock);
733 
734 		synchronize_net();
735 	}
736 }
737 EXPORT_SYMBOL(inet6_unregister_protosw);
738 
739 int inet6_sk_rebuild_header(struct sock *sk)
740 {
741 	struct ipv6_pinfo *np = inet6_sk(sk);
742 	struct dst_entry *dst;
743 
744 	dst = __sk_dst_check(sk, np->dst_cookie);
745 
746 	if (!dst) {
747 		struct inet_sock *inet = inet_sk(sk);
748 		struct in6_addr *final_p, final;
749 		struct flowi6 fl6;
750 
751 		memset(&fl6, 0, sizeof(fl6));
752 		fl6.flowi6_proto = sk->sk_protocol;
753 		fl6.daddr = sk->sk_v6_daddr;
754 		fl6.saddr = np->saddr;
755 		fl6.flowlabel = np->flow_label;
756 		fl6.flowi6_oif = sk->sk_bound_dev_if;
757 		fl6.flowi6_mark = sk->sk_mark;
758 		fl6.fl6_dport = inet->inet_dport;
759 		fl6.fl6_sport = inet->inet_sport;
760 		fl6.flowi6_uid = sk->sk_uid;
761 		security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
762 
763 		rcu_read_lock();
764 		final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt),
765 					 &final);
766 		rcu_read_unlock();
767 
768 		dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p);
769 		if (IS_ERR(dst)) {
770 			sk->sk_route_caps = 0;
771 			sk->sk_err_soft = -PTR_ERR(dst);
772 			return PTR_ERR(dst);
773 		}
774 
775 		ip6_dst_store(sk, dst, NULL, NULL);
776 	}
777 
778 	return 0;
779 }
780 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
781 
782 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb,
783 		       const struct inet6_skb_parm *opt)
784 {
785 	const struct ipv6_pinfo *np = inet6_sk(sk);
786 
787 	if (np->rxopt.all) {
788 		if (((opt->flags & IP6SKB_HOPBYHOP) &&
789 		     (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) ||
790 		    (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) &&
791 		     np->rxopt.bits.rxflow) ||
792 		    (opt->srcrt && (np->rxopt.bits.srcrt ||
793 		     np->rxopt.bits.osrcrt)) ||
794 		    ((opt->dst1 || opt->dst0) &&
795 		     (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
796 			return true;
797 	}
798 	return false;
799 }
800 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
801 
802 static struct packet_type ipv6_packet_type __read_mostly = {
803 	.type = cpu_to_be16(ETH_P_IPV6),
804 	.func = ipv6_rcv,
805 	.list_func = ipv6_list_rcv,
806 };
807 
808 static int __init ipv6_packet_init(void)
809 {
810 	dev_add_pack(&ipv6_packet_type);
811 	return 0;
812 }
813 
814 static void ipv6_packet_cleanup(void)
815 {
816 	dev_remove_pack(&ipv6_packet_type);
817 }
818 
819 static int __net_init ipv6_init_mibs(struct net *net)
820 {
821 	int i;
822 
823 	net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib);
824 	if (!net->mib.udp_stats_in6)
825 		return -ENOMEM;
826 	net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib);
827 	if (!net->mib.udplite_stats_in6)
828 		goto err_udplite_mib;
829 	net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib);
830 	if (!net->mib.ipv6_statistics)
831 		goto err_ip_mib;
832 
833 	for_each_possible_cpu(i) {
834 		struct ipstats_mib *af_inet6_stats;
835 		af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i);
836 		u64_stats_init(&af_inet6_stats->syncp);
837 	}
838 
839 
840 	net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib);
841 	if (!net->mib.icmpv6_statistics)
842 		goto err_icmp_mib;
843 	net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
844 						GFP_KERNEL);
845 	if (!net->mib.icmpv6msg_statistics)
846 		goto err_icmpmsg_mib;
847 	return 0;
848 
849 err_icmpmsg_mib:
850 	free_percpu(net->mib.icmpv6_statistics);
851 err_icmp_mib:
852 	free_percpu(net->mib.ipv6_statistics);
853 err_ip_mib:
854 	free_percpu(net->mib.udplite_stats_in6);
855 err_udplite_mib:
856 	free_percpu(net->mib.udp_stats_in6);
857 	return -ENOMEM;
858 }
859 
860 static void ipv6_cleanup_mibs(struct net *net)
861 {
862 	free_percpu(net->mib.udp_stats_in6);
863 	free_percpu(net->mib.udplite_stats_in6);
864 	free_percpu(net->mib.ipv6_statistics);
865 	free_percpu(net->mib.icmpv6_statistics);
866 	kfree(net->mib.icmpv6msg_statistics);
867 }
868 
869 static int __net_init inet6_net_init(struct net *net)
870 {
871 	int err = 0;
872 
873 	net->ipv6.sysctl.bindv6only = 0;
874 	net->ipv6.sysctl.icmpv6_time = 1*HZ;
875 	net->ipv6.sysctl.icmpv6_echo_ignore_all = 0;
876 	net->ipv6.sysctl.icmpv6_echo_ignore_multicast = 0;
877 	net->ipv6.sysctl.icmpv6_echo_ignore_anycast = 0;
878 
879 	/* By default, rate limit error messages.
880 	 * Except for pmtu discovery, it would break it.
881 	 * proc_do_large_bitmap needs pointer to the bitmap.
882 	 */
883 	bitmap_set(net->ipv6.sysctl.icmpv6_ratemask, 0, ICMPV6_ERRMSG_MAX + 1);
884 	bitmap_clear(net->ipv6.sysctl.icmpv6_ratemask, ICMPV6_PKT_TOOBIG, 1);
885 	net->ipv6.sysctl.icmpv6_ratemask_ptr = net->ipv6.sysctl.icmpv6_ratemask;
886 
887 	net->ipv6.sysctl.flowlabel_consistency = 1;
888 	net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS;
889 	net->ipv6.sysctl.idgen_retries = 3;
890 	net->ipv6.sysctl.idgen_delay = 1 * HZ;
891 	net->ipv6.sysctl.flowlabel_state_ranges = 0;
892 	net->ipv6.sysctl.max_dst_opts_cnt = IP6_DEFAULT_MAX_DST_OPTS_CNT;
893 	net->ipv6.sysctl.max_hbh_opts_cnt = IP6_DEFAULT_MAX_HBH_OPTS_CNT;
894 	net->ipv6.sysctl.max_dst_opts_len = IP6_DEFAULT_MAX_DST_OPTS_LEN;
895 	net->ipv6.sysctl.max_hbh_opts_len = IP6_DEFAULT_MAX_HBH_OPTS_LEN;
896 	atomic_set(&net->ipv6.fib6_sernum, 1);
897 
898 	err = ipv6_init_mibs(net);
899 	if (err)
900 		return err;
901 #ifdef CONFIG_PROC_FS
902 	err = udp6_proc_init(net);
903 	if (err)
904 		goto out;
905 	err = tcp6_proc_init(net);
906 	if (err)
907 		goto proc_tcp6_fail;
908 	err = ac6_proc_init(net);
909 	if (err)
910 		goto proc_ac6_fail;
911 #endif
912 	return err;
913 
914 #ifdef CONFIG_PROC_FS
915 proc_ac6_fail:
916 	tcp6_proc_exit(net);
917 proc_tcp6_fail:
918 	udp6_proc_exit(net);
919 out:
920 	ipv6_cleanup_mibs(net);
921 	return err;
922 #endif
923 }
924 
925 static void __net_exit inet6_net_exit(struct net *net)
926 {
927 #ifdef CONFIG_PROC_FS
928 	udp6_proc_exit(net);
929 	tcp6_proc_exit(net);
930 	ac6_proc_exit(net);
931 #endif
932 	ipv6_cleanup_mibs(net);
933 }
934 
935 static struct pernet_operations inet6_net_ops = {
936 	.init = inet6_net_init,
937 	.exit = inet6_net_exit,
938 };
939 
940 static int ipv6_route_input(struct sk_buff *skb)
941 {
942 	ip6_route_input(skb);
943 	return skb_dst(skb)->error;
944 }
945 
946 static const struct ipv6_stub ipv6_stub_impl = {
947 	.ipv6_sock_mc_join = ipv6_sock_mc_join,
948 	.ipv6_sock_mc_drop = ipv6_sock_mc_drop,
949 	.ipv6_dst_lookup_flow = ip6_dst_lookup_flow,
950 	.ipv6_route_input  = ipv6_route_input,
951 	.fib6_get_table	   = fib6_get_table,
952 	.fib6_table_lookup = fib6_table_lookup,
953 	.fib6_lookup       = fib6_lookup,
954 	.fib6_select_path  = fib6_select_path,
955 	.ip6_mtu_from_fib6 = ip6_mtu_from_fib6,
956 	.fib6_nh_init	   = fib6_nh_init,
957 	.fib6_nh_release   = fib6_nh_release,
958 	.fib6_update_sernum = fib6_update_sernum_stub,
959 	.fib6_rt_update	   = fib6_rt_update,
960 	.ip6_del_rt	   = ip6_del_rt,
961 	.udpv6_encap_enable = udpv6_encap_enable,
962 	.ndisc_send_na = ndisc_send_na,
963 	.nd_tbl	= &nd_tbl,
964 };
965 
966 static const struct ipv6_bpf_stub ipv6_bpf_stub_impl = {
967 	.inet6_bind = __inet6_bind,
968 	.udp6_lib_lookup = __udp6_lib_lookup,
969 };
970 
971 static int __init inet6_init(void)
972 {
973 	struct list_head *r;
974 	int err = 0;
975 
976 	sock_skb_cb_check_size(sizeof(struct inet6_skb_parm));
977 
978 	/* Register the socket-side information for inet6_create.  */
979 	for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
980 		INIT_LIST_HEAD(r);
981 
982 	if (disable_ipv6_mod) {
983 		pr_info("Loaded, but administratively disabled, reboot required to enable\n");
984 		goto out;
985 	}
986 
987 	err = proto_register(&tcpv6_prot, 1);
988 	if (err)
989 		goto out;
990 
991 	err = proto_register(&udpv6_prot, 1);
992 	if (err)
993 		goto out_unregister_tcp_proto;
994 
995 	err = proto_register(&udplitev6_prot, 1);
996 	if (err)
997 		goto out_unregister_udp_proto;
998 
999 	err = proto_register(&rawv6_prot, 1);
1000 	if (err)
1001 		goto out_unregister_udplite_proto;
1002 
1003 	err = proto_register(&pingv6_prot, 1);
1004 	if (err)
1005 		goto out_unregister_raw_proto;
1006 
1007 	/* We MUST register RAW sockets before we create the ICMP6,
1008 	 * IGMP6, or NDISC control sockets.
1009 	 */
1010 	err = rawv6_init();
1011 	if (err)
1012 		goto out_unregister_ping_proto;
1013 
1014 	/* Register the family here so that the init calls below will
1015 	 * be able to create sockets. (?? is this dangerous ??)
1016 	 */
1017 	err = sock_register(&inet6_family_ops);
1018 	if (err)
1019 		goto out_sock_register_fail;
1020 
1021 	/*
1022 	 *	ipngwg API draft makes clear that the correct semantics
1023 	 *	for TCP and UDP is to consider one TCP and UDP instance
1024 	 *	in a host available by both INET and INET6 APIs and
1025 	 *	able to communicate via both network protocols.
1026 	 */
1027 
1028 	err = register_pernet_subsys(&inet6_net_ops);
1029 	if (err)
1030 		goto register_pernet_fail;
1031 	err = ip6_mr_init();
1032 	if (err)
1033 		goto ipmr_fail;
1034 	err = icmpv6_init();
1035 	if (err)
1036 		goto icmp_fail;
1037 	err = ndisc_init();
1038 	if (err)
1039 		goto ndisc_fail;
1040 	err = igmp6_init();
1041 	if (err)
1042 		goto igmp_fail;
1043 
1044 	err = ipv6_netfilter_init();
1045 	if (err)
1046 		goto netfilter_fail;
1047 	/* Create /proc/foo6 entries. */
1048 #ifdef CONFIG_PROC_FS
1049 	err = -ENOMEM;
1050 	if (raw6_proc_init())
1051 		goto proc_raw6_fail;
1052 	if (udplite6_proc_init())
1053 		goto proc_udplite6_fail;
1054 	if (ipv6_misc_proc_init())
1055 		goto proc_misc6_fail;
1056 	if (if6_proc_init())
1057 		goto proc_if6_fail;
1058 #endif
1059 	err = ip6_route_init();
1060 	if (err)
1061 		goto ip6_route_fail;
1062 	err = ndisc_late_init();
1063 	if (err)
1064 		goto ndisc_late_fail;
1065 	err = ip6_flowlabel_init();
1066 	if (err)
1067 		goto ip6_flowlabel_fail;
1068 	err = ipv6_anycast_init();
1069 	if (err)
1070 		goto ipv6_anycast_fail;
1071 	err = addrconf_init();
1072 	if (err)
1073 		goto addrconf_fail;
1074 
1075 	/* Init v6 extension headers. */
1076 	err = ipv6_exthdrs_init();
1077 	if (err)
1078 		goto ipv6_exthdrs_fail;
1079 
1080 	err = ipv6_frag_init();
1081 	if (err)
1082 		goto ipv6_frag_fail;
1083 
1084 	/* Init v6 transport protocols. */
1085 	err = udpv6_init();
1086 	if (err)
1087 		goto udpv6_fail;
1088 
1089 	err = udplitev6_init();
1090 	if (err)
1091 		goto udplitev6_fail;
1092 
1093 	err = udpv6_offload_init();
1094 	if (err)
1095 		goto udpv6_offload_fail;
1096 
1097 	err = tcpv6_init();
1098 	if (err)
1099 		goto tcpv6_fail;
1100 
1101 	err = ipv6_packet_init();
1102 	if (err)
1103 		goto ipv6_packet_fail;
1104 
1105 	err = pingv6_init();
1106 	if (err)
1107 		goto pingv6_fail;
1108 
1109 	err = calipso_init();
1110 	if (err)
1111 		goto calipso_fail;
1112 
1113 	err = seg6_init();
1114 	if (err)
1115 		goto seg6_fail;
1116 
1117 	err = igmp6_late_init();
1118 	if (err)
1119 		goto igmp6_late_err;
1120 
1121 #ifdef CONFIG_SYSCTL
1122 	err = ipv6_sysctl_register();
1123 	if (err)
1124 		goto sysctl_fail;
1125 #endif
1126 
1127 	/* ensure that ipv6 stubs are visible only after ipv6 is ready */
1128 	wmb();
1129 	ipv6_stub = &ipv6_stub_impl;
1130 	ipv6_bpf_stub = &ipv6_bpf_stub_impl;
1131 out:
1132 	return err;
1133 
1134 #ifdef CONFIG_SYSCTL
1135 sysctl_fail:
1136 	igmp6_late_cleanup();
1137 #endif
1138 igmp6_late_err:
1139 	seg6_exit();
1140 seg6_fail:
1141 	calipso_exit();
1142 calipso_fail:
1143 	pingv6_exit();
1144 pingv6_fail:
1145 	ipv6_packet_cleanup();
1146 ipv6_packet_fail:
1147 	tcpv6_exit();
1148 tcpv6_fail:
1149 	udpv6_offload_exit();
1150 udpv6_offload_fail:
1151 	udplitev6_exit();
1152 udplitev6_fail:
1153 	udpv6_exit();
1154 udpv6_fail:
1155 	ipv6_frag_exit();
1156 ipv6_frag_fail:
1157 	ipv6_exthdrs_exit();
1158 ipv6_exthdrs_fail:
1159 	addrconf_cleanup();
1160 addrconf_fail:
1161 	ipv6_anycast_cleanup();
1162 ipv6_anycast_fail:
1163 	ip6_flowlabel_cleanup();
1164 ip6_flowlabel_fail:
1165 	ndisc_late_cleanup();
1166 ndisc_late_fail:
1167 	ip6_route_cleanup();
1168 ip6_route_fail:
1169 #ifdef CONFIG_PROC_FS
1170 	if6_proc_exit();
1171 proc_if6_fail:
1172 	ipv6_misc_proc_exit();
1173 proc_misc6_fail:
1174 	udplite6_proc_exit();
1175 proc_udplite6_fail:
1176 	raw6_proc_exit();
1177 proc_raw6_fail:
1178 #endif
1179 	ipv6_netfilter_fini();
1180 netfilter_fail:
1181 	igmp6_cleanup();
1182 igmp_fail:
1183 	ndisc_cleanup();
1184 ndisc_fail:
1185 	icmpv6_cleanup();
1186 icmp_fail:
1187 	ip6_mr_cleanup();
1188 ipmr_fail:
1189 	unregister_pernet_subsys(&inet6_net_ops);
1190 register_pernet_fail:
1191 	sock_unregister(PF_INET6);
1192 	rtnl_unregister_all(PF_INET6);
1193 out_sock_register_fail:
1194 	rawv6_exit();
1195 out_unregister_ping_proto:
1196 	proto_unregister(&pingv6_prot);
1197 out_unregister_raw_proto:
1198 	proto_unregister(&rawv6_prot);
1199 out_unregister_udplite_proto:
1200 	proto_unregister(&udplitev6_prot);
1201 out_unregister_udp_proto:
1202 	proto_unregister(&udpv6_prot);
1203 out_unregister_tcp_proto:
1204 	proto_unregister(&tcpv6_prot);
1205 	goto out;
1206 }
1207 module_init(inet6_init);
1208 
1209 MODULE_ALIAS_NETPROTO(PF_INET6);
1210