xref: /linux/net/ipv6/af_inet6.c (revision 24ce659dcc02c21f8d6c0a7589c3320a4dfa8152)
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 
64 #include <linux/uaccess.h>
65 #include <linux/mroute6.h>
66 
67 #include "ip6_offload.h"
68 
69 MODULE_AUTHOR("Cast of dozens");
70 MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
71 MODULE_LICENSE("GPL");
72 
73 /* The inetsw6 table contains everything that inet6_create needs to
74  * build a new socket.
75  */
76 static struct list_head inetsw6[SOCK_MAX];
77 static DEFINE_SPINLOCK(inetsw6_lock);
78 
79 struct ipv6_params ipv6_defaults = {
80 	.disable_ipv6 = 0,
81 	.autoconf = 1,
82 };
83 
84 static int disable_ipv6_mod;
85 
86 module_param_named(disable, disable_ipv6_mod, int, 0444);
87 MODULE_PARM_DESC(disable, "Disable IPv6 module such that it is non-functional");
88 
89 module_param_named(disable_ipv6, ipv6_defaults.disable_ipv6, int, 0444);
90 MODULE_PARM_DESC(disable_ipv6, "Disable IPv6 on all interfaces");
91 
92 module_param_named(autoconf, ipv6_defaults.autoconf, int, 0444);
93 MODULE_PARM_DESC(autoconf, "Enable IPv6 address autoconfiguration on all interfaces");
94 
95 bool ipv6_mod_enabled(void)
96 {
97 	return disable_ipv6_mod == 0;
98 }
99 EXPORT_SYMBOL_GPL(ipv6_mod_enabled);
100 
101 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
102 {
103 	const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
104 
105 	return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
106 }
107 
108 static int inet6_create(struct net *net, struct socket *sock, int protocol,
109 			int kern)
110 {
111 	struct inet_sock *inet;
112 	struct ipv6_pinfo *np;
113 	struct sock *sk;
114 	struct inet_protosw *answer;
115 	struct proto *answer_prot;
116 	unsigned char answer_flags;
117 	int try_loading_module = 0;
118 	int err;
119 
120 	if (protocol < 0 || protocol >= IPPROTO_MAX)
121 		return -EINVAL;
122 
123 	/* Look for the requested type/protocol pair. */
124 lookup_protocol:
125 	err = -ESOCKTNOSUPPORT;
126 	rcu_read_lock();
127 	list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
128 
129 		err = 0;
130 		/* Check the non-wild match. */
131 		if (protocol == answer->protocol) {
132 			if (protocol != IPPROTO_IP)
133 				break;
134 		} else {
135 			/* Check for the two wild cases. */
136 			if (IPPROTO_IP == protocol) {
137 				protocol = answer->protocol;
138 				break;
139 			}
140 			if (IPPROTO_IP == answer->protocol)
141 				break;
142 		}
143 		err = -EPROTONOSUPPORT;
144 	}
145 
146 	if (err) {
147 		if (try_loading_module < 2) {
148 			rcu_read_unlock();
149 			/*
150 			 * Be more specific, e.g. net-pf-10-proto-132-type-1
151 			 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
152 			 */
153 			if (++try_loading_module == 1)
154 				request_module("net-pf-%d-proto-%d-type-%d",
155 						PF_INET6, protocol, sock->type);
156 			/*
157 			 * Fall back to generic, e.g. net-pf-10-proto-132
158 			 * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
159 			 */
160 			else
161 				request_module("net-pf-%d-proto-%d",
162 						PF_INET6, protocol);
163 			goto lookup_protocol;
164 		} else
165 			goto out_rcu_unlock;
166 	}
167 
168 	err = -EPERM;
169 	if (sock->type == SOCK_RAW && !kern &&
170 	    !ns_capable(net->user_ns, CAP_NET_RAW))
171 		goto out_rcu_unlock;
172 
173 	sock->ops = answer->ops;
174 	answer_prot = answer->prot;
175 	answer_flags = answer->flags;
176 	rcu_read_unlock();
177 
178 	WARN_ON(!answer_prot->slab);
179 
180 	err = -ENOBUFS;
181 	sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot, kern);
182 	if (!sk)
183 		goto out;
184 
185 	sock_init_data(sock, sk);
186 
187 	err = 0;
188 	if (INET_PROTOSW_REUSE & answer_flags)
189 		sk->sk_reuse = SK_CAN_REUSE;
190 
191 	inet = inet_sk(sk);
192 	inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
193 
194 	if (SOCK_RAW == sock->type) {
195 		inet->inet_num = protocol;
196 		if (IPPROTO_RAW == protocol)
197 			inet->hdrincl = 1;
198 	}
199 
200 	sk->sk_destruct		= inet_sock_destruct;
201 	sk->sk_family		= PF_INET6;
202 	sk->sk_protocol		= protocol;
203 
204 	sk->sk_backlog_rcv	= answer->prot->backlog_rcv;
205 
206 	inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
207 	np->hop_limit	= -1;
208 	np->mcast_hops	= IPV6_DEFAULT_MCASTHOPS;
209 	np->mc_loop	= 1;
210 	np->mc_all	= 1;
211 	np->pmtudisc	= IPV6_PMTUDISC_WANT;
212 	np->repflow	= net->ipv6.sysctl.flowlabel_reflect & FLOWLABEL_REFLECT_ESTABLISHED;
213 	sk->sk_ipv6only	= net->ipv6.sysctl.bindv6only;
214 
215 	/* Init the ipv4 part of the socket since we can have sockets
216 	 * using v6 API for ipv4.
217 	 */
218 	inet->uc_ttl	= -1;
219 
220 	inet->mc_loop	= 1;
221 	inet->mc_ttl	= 1;
222 	inet->mc_index	= 0;
223 	inet->mc_list	= NULL;
224 	inet->rcv_tos	= 0;
225 
226 	if (net->ipv4.sysctl_ip_no_pmtu_disc)
227 		inet->pmtudisc = IP_PMTUDISC_DONT;
228 	else
229 		inet->pmtudisc = IP_PMTUDISC_WANT;
230 	/*
231 	 * Increment only the relevant sk_prot->socks debug field, this changes
232 	 * the previous behaviour of incrementing both the equivalent to
233 	 * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
234 	 *
235 	 * This allows better debug granularity as we'll know exactly how many
236 	 * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
237 	 * transport protocol socks. -acme
238 	 */
239 	sk_refcnt_debug_inc(sk);
240 
241 	if (inet->inet_num) {
242 		/* It assumes that any protocol which allows
243 		 * the user to assign a number at socket
244 		 * creation time automatically shares.
245 		 */
246 		inet->inet_sport = htons(inet->inet_num);
247 		err = sk->sk_prot->hash(sk);
248 		if (err) {
249 			sk_common_release(sk);
250 			goto out;
251 		}
252 	}
253 	if (sk->sk_prot->init) {
254 		err = sk->sk_prot->init(sk);
255 		if (err) {
256 			sk_common_release(sk);
257 			goto out;
258 		}
259 	}
260 
261 	if (!kern) {
262 		err = BPF_CGROUP_RUN_PROG_INET_SOCK(sk);
263 		if (err) {
264 			sk_common_release(sk);
265 			goto out;
266 		}
267 	}
268 out:
269 	return err;
270 out_rcu_unlock:
271 	rcu_read_unlock();
272 	goto out;
273 }
274 
275 static int __inet6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len,
276 			bool force_bind_address_no_port, bool with_lock)
277 {
278 	struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr;
279 	struct inet_sock *inet = inet_sk(sk);
280 	struct ipv6_pinfo *np = inet6_sk(sk);
281 	struct net *net = sock_net(sk);
282 	__be32 v4addr = 0;
283 	unsigned short snum;
284 	bool saved_ipv6only;
285 	int addr_type = 0;
286 	int err = 0;
287 
288 	if (addr->sin6_family != AF_INET6)
289 		return -EAFNOSUPPORT;
290 
291 	addr_type = ipv6_addr_type(&addr->sin6_addr);
292 	if ((addr_type & IPV6_ADDR_MULTICAST) && sk->sk_type == SOCK_STREAM)
293 		return -EINVAL;
294 
295 	snum = ntohs(addr->sin6_port);
296 	if (snum && inet_port_requires_bind_service(net, snum) &&
297 	    !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
298 		return -EACCES;
299 
300 	if (with_lock)
301 		lock_sock(sk);
302 
303 	/* Check these errors (active socket, double bind). */
304 	if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
305 		err = -EINVAL;
306 		goto out;
307 	}
308 
309 	/* Check if the address belongs to the host. */
310 	if (addr_type == IPV6_ADDR_MAPPED) {
311 		struct net_device *dev = NULL;
312 		int chk_addr_ret;
313 
314 		/* Binding to v4-mapped address on a v6-only socket
315 		 * makes no sense
316 		 */
317 		if (sk->sk_ipv6only) {
318 			err = -EINVAL;
319 			goto out;
320 		}
321 
322 		rcu_read_lock();
323 		if (sk->sk_bound_dev_if) {
324 			dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
325 			if (!dev) {
326 				err = -ENODEV;
327 				goto out_unlock;
328 			}
329 		}
330 
331 		/* Reproduce AF_INET checks to make the bindings consistent */
332 		v4addr = addr->sin6_addr.s6_addr32[3];
333 		chk_addr_ret = inet_addr_type_dev_table(net, dev, v4addr);
334 		rcu_read_unlock();
335 
336 		if (!inet_can_nonlocal_bind(net, inet) &&
337 		    v4addr != htonl(INADDR_ANY) &&
338 		    chk_addr_ret != RTN_LOCAL &&
339 		    chk_addr_ret != RTN_MULTICAST &&
340 		    chk_addr_ret != RTN_BROADCAST) {
341 			err = -EADDRNOTAVAIL;
342 			goto out;
343 		}
344 	} else {
345 		if (addr_type != IPV6_ADDR_ANY) {
346 			struct net_device *dev = NULL;
347 
348 			rcu_read_lock();
349 			if (__ipv6_addr_needs_scope_id(addr_type)) {
350 				if (addr_len >= sizeof(struct sockaddr_in6) &&
351 				    addr->sin6_scope_id) {
352 					/* Override any existing binding, if another one
353 					 * is supplied by user.
354 					 */
355 					sk->sk_bound_dev_if = addr->sin6_scope_id;
356 				}
357 
358 				/* Binding to link-local address requires an interface */
359 				if (!sk->sk_bound_dev_if) {
360 					err = -EINVAL;
361 					goto out_unlock;
362 				}
363 			}
364 
365 			if (sk->sk_bound_dev_if) {
366 				dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
367 				if (!dev) {
368 					err = -ENODEV;
369 					goto out_unlock;
370 				}
371 			}
372 
373 			/* ipv4 addr of the socket is invalid.  Only the
374 			 * unspecified and mapped address have a v4 equivalent.
375 			 */
376 			v4addr = LOOPBACK4_IPV6;
377 			if (!(addr_type & IPV6_ADDR_MULTICAST))	{
378 				if (!ipv6_can_nonlocal_bind(net, inet) &&
379 				    !ipv6_chk_addr(net, &addr->sin6_addr,
380 						   dev, 0)) {
381 					err = -EADDRNOTAVAIL;
382 					goto out_unlock;
383 				}
384 			}
385 			rcu_read_unlock();
386 		}
387 	}
388 
389 	inet->inet_rcv_saddr = v4addr;
390 	inet->inet_saddr = v4addr;
391 
392 	sk->sk_v6_rcv_saddr = addr->sin6_addr;
393 
394 	if (!(addr_type & IPV6_ADDR_MULTICAST))
395 		np->saddr = addr->sin6_addr;
396 
397 	saved_ipv6only = sk->sk_ipv6only;
398 	if (addr_type != IPV6_ADDR_ANY && addr_type != IPV6_ADDR_MAPPED)
399 		sk->sk_ipv6only = 1;
400 
401 	/* Make sure we are allowed to bind here. */
402 	if (snum || !(inet->bind_address_no_port ||
403 		      force_bind_address_no_port)) {
404 		if (sk->sk_prot->get_port(sk, snum)) {
405 			sk->sk_ipv6only = saved_ipv6only;
406 			inet_reset_saddr(sk);
407 			err = -EADDRINUSE;
408 			goto out;
409 		}
410 		err = BPF_CGROUP_RUN_PROG_INET6_POST_BIND(sk);
411 		if (err) {
412 			sk->sk_ipv6only = saved_ipv6only;
413 			inet_reset_saddr(sk);
414 			goto out;
415 		}
416 	}
417 
418 	if (addr_type != IPV6_ADDR_ANY)
419 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
420 	if (snum)
421 		sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
422 	inet->inet_sport = htons(inet->inet_num);
423 	inet->inet_dport = 0;
424 	inet->inet_daddr = 0;
425 out:
426 	if (with_lock)
427 		release_sock(sk);
428 	return err;
429 out_unlock:
430 	rcu_read_unlock();
431 	goto out;
432 }
433 
434 /* bind for INET6 API */
435 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
436 {
437 	struct sock *sk = sock->sk;
438 	int err = 0;
439 
440 	/* If the socket has its own bind function then use it. */
441 	if (sk->sk_prot->bind)
442 		return sk->sk_prot->bind(sk, uaddr, addr_len);
443 
444 	if (addr_len < SIN6_LEN_RFC2133)
445 		return -EINVAL;
446 
447 	/* BPF prog is run before any checks are done so that if the prog
448 	 * changes context in a wrong way it will be caught.
449 	 */
450 	err = BPF_CGROUP_RUN_PROG_INET6_BIND(sk, uaddr);
451 	if (err)
452 		return err;
453 
454 	return __inet6_bind(sk, uaddr, addr_len, false, true);
455 }
456 EXPORT_SYMBOL(inet6_bind);
457 
458 int inet6_release(struct socket *sock)
459 {
460 	struct sock *sk = sock->sk;
461 
462 	if (!sk)
463 		return -EINVAL;
464 
465 	/* Free mc lists */
466 	ipv6_sock_mc_close(sk);
467 
468 	/* Free ac lists */
469 	ipv6_sock_ac_close(sk);
470 
471 	return inet_release(sock);
472 }
473 EXPORT_SYMBOL(inet6_release);
474 
475 void inet6_destroy_sock(struct sock *sk)
476 {
477 	struct ipv6_pinfo *np = inet6_sk(sk);
478 	struct sk_buff *skb;
479 	struct ipv6_txoptions *opt;
480 
481 	/* Release rx options */
482 
483 	skb = xchg(&np->pktoptions, NULL);
484 	kfree_skb(skb);
485 
486 	skb = xchg(&np->rxpmtu, NULL);
487 	kfree_skb(skb);
488 
489 	/* Free flowlabels */
490 	fl6_free_socklist(sk);
491 
492 	/* Free tx options */
493 
494 	opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL);
495 	if (opt) {
496 		atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
497 		txopt_put(opt);
498 	}
499 }
500 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
501 
502 /*
503  *	This does both peername and sockname.
504  */
505 
506 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
507 		 int peer)
508 {
509 	struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
510 	struct sock *sk = sock->sk;
511 	struct inet_sock *inet = inet_sk(sk);
512 	struct ipv6_pinfo *np = inet6_sk(sk);
513 
514 	sin->sin6_family = AF_INET6;
515 	sin->sin6_flowinfo = 0;
516 	sin->sin6_scope_id = 0;
517 	if (peer) {
518 		if (!inet->inet_dport)
519 			return -ENOTCONN;
520 		if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
521 		    peer == 1)
522 			return -ENOTCONN;
523 		sin->sin6_port = inet->inet_dport;
524 		sin->sin6_addr = sk->sk_v6_daddr;
525 		if (np->sndflow)
526 			sin->sin6_flowinfo = np->flow_label;
527 	} else {
528 		if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
529 			sin->sin6_addr = np->saddr;
530 		else
531 			sin->sin6_addr = sk->sk_v6_rcv_saddr;
532 
533 		sin->sin6_port = inet->inet_sport;
534 	}
535 	sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
536 						 sk->sk_bound_dev_if);
537 	return sizeof(*sin);
538 }
539 EXPORT_SYMBOL(inet6_getname);
540 
541 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
542 {
543 	struct sock *sk = sock->sk;
544 	struct net *net = sock_net(sk);
545 
546 	switch (cmd) {
547 	case SIOCADDRT:
548 	case SIOCDELRT:
549 
550 		return ipv6_route_ioctl(net, cmd, (void __user *)arg);
551 
552 	case SIOCSIFADDR:
553 		return addrconf_add_ifaddr(net, (void __user *) arg);
554 	case SIOCDIFADDR:
555 		return addrconf_del_ifaddr(net, (void __user *) arg);
556 	case SIOCSIFDSTADDR:
557 		return addrconf_set_dstaddr(net, (void __user *) arg);
558 	default:
559 		if (!sk->sk_prot->ioctl)
560 			return -ENOIOCTLCMD;
561 		return sk->sk_prot->ioctl(sk, cmd, arg);
562 	}
563 	/*NOTREACHED*/
564 	return 0;
565 }
566 EXPORT_SYMBOL(inet6_ioctl);
567 
568 INDIRECT_CALLABLE_DECLARE(int udpv6_sendmsg(struct sock *, struct msghdr *,
569 					    size_t));
570 int inet6_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
571 {
572 	struct sock *sk = sock->sk;
573 
574 	if (unlikely(inet_send_prepare(sk)))
575 		return -EAGAIN;
576 
577 	return INDIRECT_CALL_2(sk->sk_prot->sendmsg, tcp_sendmsg, udpv6_sendmsg,
578 			       sk, msg, size);
579 }
580 
581 INDIRECT_CALLABLE_DECLARE(int udpv6_recvmsg(struct sock *, struct msghdr *,
582 					    size_t, int, int, int *));
583 int inet6_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
584 		  int flags)
585 {
586 	struct sock *sk = sock->sk;
587 	int addr_len = 0;
588 	int err;
589 
590 	if (likely(!(flags & MSG_ERRQUEUE)))
591 		sock_rps_record_flow(sk);
592 
593 	err = INDIRECT_CALL_2(sk->sk_prot->recvmsg, tcp_recvmsg, udpv6_recvmsg,
594 			      sk, msg, size, flags & MSG_DONTWAIT,
595 			      flags & ~MSG_DONTWAIT, &addr_len);
596 	if (err >= 0)
597 		msg->msg_namelen = addr_len;
598 	return err;
599 }
600 
601 const struct proto_ops inet6_stream_ops = {
602 	.family		   = PF_INET6,
603 	.owner		   = THIS_MODULE,
604 	.release	   = inet6_release,
605 	.bind		   = inet6_bind,
606 	.connect	   = inet_stream_connect,	/* ok		*/
607 	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
608 	.accept		   = inet_accept,		/* ok		*/
609 	.getname	   = inet6_getname,
610 	.poll		   = tcp_poll,			/* ok		*/
611 	.ioctl		   = inet6_ioctl,		/* must change  */
612 	.gettstamp	   = sock_gettstamp,
613 	.listen		   = inet_listen,		/* ok		*/
614 	.shutdown	   = inet_shutdown,		/* ok		*/
615 	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
616 	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
617 	.sendmsg	   = inet6_sendmsg,		/* retpoline's sake */
618 	.recvmsg	   = inet6_recvmsg,		/* retpoline's sake */
619 #ifdef CONFIG_MMU
620 	.mmap		   = tcp_mmap,
621 #endif
622 	.sendpage	   = inet_sendpage,
623 	.sendmsg_locked    = tcp_sendmsg_locked,
624 	.sendpage_locked   = tcp_sendpage_locked,
625 	.splice_read	   = tcp_splice_read,
626 	.read_sock	   = tcp_read_sock,
627 	.peek_len	   = tcp_peek_len,
628 #ifdef CONFIG_COMPAT
629 	.compat_setsockopt = compat_sock_common_setsockopt,
630 	.compat_getsockopt = compat_sock_common_getsockopt,
631 #endif
632 	.set_rcvlowat	   = tcp_set_rcvlowat,
633 };
634 
635 const struct proto_ops inet6_dgram_ops = {
636 	.family		   = PF_INET6,
637 	.owner		   = THIS_MODULE,
638 	.release	   = inet6_release,
639 	.bind		   = inet6_bind,
640 	.connect	   = inet_dgram_connect,	/* ok		*/
641 	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
642 	.accept		   = sock_no_accept,		/* a do nothing	*/
643 	.getname	   = inet6_getname,
644 	.poll		   = udp_poll,			/* ok		*/
645 	.ioctl		   = inet6_ioctl,		/* must change  */
646 	.gettstamp	   = sock_gettstamp,
647 	.listen		   = sock_no_listen,		/* ok		*/
648 	.shutdown	   = inet_shutdown,		/* ok		*/
649 	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
650 	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
651 	.sendmsg	   = inet6_sendmsg,		/* retpoline's sake */
652 	.recvmsg	   = inet6_recvmsg,		/* retpoline's sake */
653 	.mmap		   = sock_no_mmap,
654 	.sendpage	   = sock_no_sendpage,
655 	.set_peek_off	   = sk_set_peek_off,
656 #ifdef CONFIG_COMPAT
657 	.compat_setsockopt = compat_sock_common_setsockopt,
658 	.compat_getsockopt = compat_sock_common_getsockopt,
659 #endif
660 };
661 
662 static const struct net_proto_family inet6_family_ops = {
663 	.family = PF_INET6,
664 	.create = inet6_create,
665 	.owner	= THIS_MODULE,
666 };
667 
668 int inet6_register_protosw(struct inet_protosw *p)
669 {
670 	struct list_head *lh;
671 	struct inet_protosw *answer;
672 	struct list_head *last_perm;
673 	int protocol = p->protocol;
674 	int ret;
675 
676 	spin_lock_bh(&inetsw6_lock);
677 
678 	ret = -EINVAL;
679 	if (p->type >= SOCK_MAX)
680 		goto out_illegal;
681 
682 	/* If we are trying to override a permanent protocol, bail. */
683 	answer = NULL;
684 	ret = -EPERM;
685 	last_perm = &inetsw6[p->type];
686 	list_for_each(lh, &inetsw6[p->type]) {
687 		answer = list_entry(lh, struct inet_protosw, list);
688 
689 		/* Check only the non-wild match. */
690 		if (INET_PROTOSW_PERMANENT & answer->flags) {
691 			if (protocol == answer->protocol)
692 				break;
693 			last_perm = lh;
694 		}
695 
696 		answer = NULL;
697 	}
698 	if (answer)
699 		goto out_permanent;
700 
701 	/* Add the new entry after the last permanent entry if any, so that
702 	 * the new entry does not override a permanent entry when matched with
703 	 * a wild-card protocol. But it is allowed to override any existing
704 	 * non-permanent entry.  This means that when we remove this entry, the
705 	 * system automatically returns to the old behavior.
706 	 */
707 	list_add_rcu(&p->list, last_perm);
708 	ret = 0;
709 out:
710 	spin_unlock_bh(&inetsw6_lock);
711 	return ret;
712 
713 out_permanent:
714 	pr_err("Attempt to override permanent protocol %d\n", protocol);
715 	goto out;
716 
717 out_illegal:
718 	pr_err("Ignoring attempt to register invalid socket type %d\n",
719 	       p->type);
720 	goto out;
721 }
722 EXPORT_SYMBOL(inet6_register_protosw);
723 
724 void
725 inet6_unregister_protosw(struct inet_protosw *p)
726 {
727 	if (INET_PROTOSW_PERMANENT & p->flags) {
728 		pr_err("Attempt to unregister permanent protocol %d\n",
729 		       p->protocol);
730 	} else {
731 		spin_lock_bh(&inetsw6_lock);
732 		list_del_rcu(&p->list);
733 		spin_unlock_bh(&inetsw6_lock);
734 
735 		synchronize_net();
736 	}
737 }
738 EXPORT_SYMBOL(inet6_unregister_protosw);
739 
740 int inet6_sk_rebuild_header(struct sock *sk)
741 {
742 	struct ipv6_pinfo *np = inet6_sk(sk);
743 	struct dst_entry *dst;
744 
745 	dst = __sk_dst_check(sk, np->dst_cookie);
746 
747 	if (!dst) {
748 		struct inet_sock *inet = inet_sk(sk);
749 		struct in6_addr *final_p, final;
750 		struct flowi6 fl6;
751 
752 		memset(&fl6, 0, sizeof(fl6));
753 		fl6.flowi6_proto = sk->sk_protocol;
754 		fl6.daddr = sk->sk_v6_daddr;
755 		fl6.saddr = np->saddr;
756 		fl6.flowlabel = np->flow_label;
757 		fl6.flowi6_oif = sk->sk_bound_dev_if;
758 		fl6.flowi6_mark = sk->sk_mark;
759 		fl6.fl6_dport = inet->inet_dport;
760 		fl6.fl6_sport = inet->inet_sport;
761 		fl6.flowi6_uid = sk->sk_uid;
762 		security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
763 
764 		rcu_read_lock();
765 		final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt),
766 					 &final);
767 		rcu_read_unlock();
768 
769 		dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p);
770 		if (IS_ERR(dst)) {
771 			sk->sk_route_caps = 0;
772 			sk->sk_err_soft = -PTR_ERR(dst);
773 			return PTR_ERR(dst);
774 		}
775 
776 		ip6_dst_store(sk, dst, NULL, NULL);
777 	}
778 
779 	return 0;
780 }
781 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
782 
783 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb,
784 		       const struct inet6_skb_parm *opt)
785 {
786 	const struct ipv6_pinfo *np = inet6_sk(sk);
787 
788 	if (np->rxopt.all) {
789 		if (((opt->flags & IP6SKB_HOPBYHOP) &&
790 		     (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) ||
791 		    (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) &&
792 		     np->rxopt.bits.rxflow) ||
793 		    (opt->srcrt && (np->rxopt.bits.srcrt ||
794 		     np->rxopt.bits.osrcrt)) ||
795 		    ((opt->dst1 || opt->dst0) &&
796 		     (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
797 			return true;
798 	}
799 	return false;
800 }
801 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
802 
803 static struct packet_type ipv6_packet_type __read_mostly = {
804 	.type = cpu_to_be16(ETH_P_IPV6),
805 	.func = ipv6_rcv,
806 	.list_func = ipv6_list_rcv,
807 };
808 
809 static int __init ipv6_packet_init(void)
810 {
811 	dev_add_pack(&ipv6_packet_type);
812 	return 0;
813 }
814 
815 static void ipv6_packet_cleanup(void)
816 {
817 	dev_remove_pack(&ipv6_packet_type);
818 }
819 
820 static int __net_init ipv6_init_mibs(struct net *net)
821 {
822 	int i;
823 
824 	net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib);
825 	if (!net->mib.udp_stats_in6)
826 		return -ENOMEM;
827 	net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib);
828 	if (!net->mib.udplite_stats_in6)
829 		goto err_udplite_mib;
830 	net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib);
831 	if (!net->mib.ipv6_statistics)
832 		goto err_ip_mib;
833 
834 	for_each_possible_cpu(i) {
835 		struct ipstats_mib *af_inet6_stats;
836 		af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i);
837 		u64_stats_init(&af_inet6_stats->syncp);
838 	}
839 
840 
841 	net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib);
842 	if (!net->mib.icmpv6_statistics)
843 		goto err_icmp_mib;
844 	net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
845 						GFP_KERNEL);
846 	if (!net->mib.icmpv6msg_statistics)
847 		goto err_icmpmsg_mib;
848 	return 0;
849 
850 err_icmpmsg_mib:
851 	free_percpu(net->mib.icmpv6_statistics);
852 err_icmp_mib:
853 	free_percpu(net->mib.ipv6_statistics);
854 err_ip_mib:
855 	free_percpu(net->mib.udplite_stats_in6);
856 err_udplite_mib:
857 	free_percpu(net->mib.udp_stats_in6);
858 	return -ENOMEM;
859 }
860 
861 static void ipv6_cleanup_mibs(struct net *net)
862 {
863 	free_percpu(net->mib.udp_stats_in6);
864 	free_percpu(net->mib.udplite_stats_in6);
865 	free_percpu(net->mib.ipv6_statistics);
866 	free_percpu(net->mib.icmpv6_statistics);
867 	kfree(net->mib.icmpv6msg_statistics);
868 }
869 
870 static int __net_init inet6_net_init(struct net *net)
871 {
872 	int err = 0;
873 
874 	net->ipv6.sysctl.bindv6only = 0;
875 	net->ipv6.sysctl.icmpv6_time = 1*HZ;
876 	net->ipv6.sysctl.icmpv6_echo_ignore_all = 0;
877 	net->ipv6.sysctl.icmpv6_echo_ignore_multicast = 0;
878 	net->ipv6.sysctl.icmpv6_echo_ignore_anycast = 0;
879 
880 	/* By default, rate limit error messages.
881 	 * Except for pmtu discovery, it would break it.
882 	 * proc_do_large_bitmap needs pointer to the bitmap.
883 	 */
884 	bitmap_set(net->ipv6.sysctl.icmpv6_ratemask, 0, ICMPV6_ERRMSG_MAX + 1);
885 	bitmap_clear(net->ipv6.sysctl.icmpv6_ratemask, ICMPV6_PKT_TOOBIG, 1);
886 	net->ipv6.sysctl.icmpv6_ratemask_ptr = net->ipv6.sysctl.icmpv6_ratemask;
887 
888 	net->ipv6.sysctl.flowlabel_consistency = 1;
889 	net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS;
890 	net->ipv6.sysctl.idgen_retries = 3;
891 	net->ipv6.sysctl.idgen_delay = 1 * HZ;
892 	net->ipv6.sysctl.flowlabel_state_ranges = 0;
893 	net->ipv6.sysctl.max_dst_opts_cnt = IP6_DEFAULT_MAX_DST_OPTS_CNT;
894 	net->ipv6.sysctl.max_hbh_opts_cnt = IP6_DEFAULT_MAX_HBH_OPTS_CNT;
895 	net->ipv6.sysctl.max_dst_opts_len = IP6_DEFAULT_MAX_DST_OPTS_LEN;
896 	net->ipv6.sysctl.max_hbh_opts_len = IP6_DEFAULT_MAX_HBH_OPTS_LEN;
897 	atomic_set(&net->ipv6.fib6_sernum, 1);
898 
899 	err = ipv6_init_mibs(net);
900 	if (err)
901 		return err;
902 #ifdef CONFIG_PROC_FS
903 	err = udp6_proc_init(net);
904 	if (err)
905 		goto out;
906 	err = tcp6_proc_init(net);
907 	if (err)
908 		goto proc_tcp6_fail;
909 	err = ac6_proc_init(net);
910 	if (err)
911 		goto proc_ac6_fail;
912 #endif
913 	return err;
914 
915 #ifdef CONFIG_PROC_FS
916 proc_ac6_fail:
917 	tcp6_proc_exit(net);
918 proc_tcp6_fail:
919 	udp6_proc_exit(net);
920 out:
921 	ipv6_cleanup_mibs(net);
922 	return err;
923 #endif
924 }
925 
926 static void __net_exit inet6_net_exit(struct net *net)
927 {
928 #ifdef CONFIG_PROC_FS
929 	udp6_proc_exit(net);
930 	tcp6_proc_exit(net);
931 	ac6_proc_exit(net);
932 #endif
933 	ipv6_cleanup_mibs(net);
934 }
935 
936 static struct pernet_operations inet6_net_ops = {
937 	.init = inet6_net_init,
938 	.exit = inet6_net_exit,
939 };
940 
941 static int ipv6_route_input(struct sk_buff *skb)
942 {
943 	ip6_route_input(skb);
944 	return skb_dst(skb)->error;
945 }
946 
947 static const struct ipv6_stub ipv6_stub_impl = {
948 	.ipv6_sock_mc_join = ipv6_sock_mc_join,
949 	.ipv6_sock_mc_drop = ipv6_sock_mc_drop,
950 	.ipv6_dst_lookup_flow = ip6_dst_lookup_flow,
951 	.ipv6_route_input  = ipv6_route_input,
952 	.fib6_get_table	   = fib6_get_table,
953 	.fib6_table_lookup = fib6_table_lookup,
954 	.fib6_lookup       = fib6_lookup,
955 	.fib6_select_path  = fib6_select_path,
956 	.ip6_mtu_from_fib6 = ip6_mtu_from_fib6,
957 	.fib6_nh_init	   = fib6_nh_init,
958 	.fib6_nh_release   = fib6_nh_release,
959 	.fib6_update_sernum = fib6_update_sernum_stub,
960 	.fib6_rt_update	   = fib6_rt_update,
961 	.ip6_del_rt	   = ip6_del_rt,
962 	.udpv6_encap_enable = udpv6_encap_enable,
963 	.ndisc_send_na = ndisc_send_na,
964 	.nd_tbl	= &nd_tbl,
965 };
966 
967 static const struct ipv6_bpf_stub ipv6_bpf_stub_impl = {
968 	.inet6_bind = __inet6_bind,
969 	.udp6_lib_lookup = __udp6_lib_lookup,
970 };
971 
972 static int __init inet6_init(void)
973 {
974 	struct list_head *r;
975 	int err = 0;
976 
977 	sock_skb_cb_check_size(sizeof(struct inet6_skb_parm));
978 
979 	/* Register the socket-side information for inet6_create.  */
980 	for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
981 		INIT_LIST_HEAD(r);
982 
983 	if (disable_ipv6_mod) {
984 		pr_info("Loaded, but administratively disabled, reboot required to enable\n");
985 		goto out;
986 	}
987 
988 	err = proto_register(&tcpv6_prot, 1);
989 	if (err)
990 		goto out;
991 
992 	err = proto_register(&udpv6_prot, 1);
993 	if (err)
994 		goto out_unregister_tcp_proto;
995 
996 	err = proto_register(&udplitev6_prot, 1);
997 	if (err)
998 		goto out_unregister_udp_proto;
999 
1000 	err = proto_register(&rawv6_prot, 1);
1001 	if (err)
1002 		goto out_unregister_udplite_proto;
1003 
1004 	err = proto_register(&pingv6_prot, 1);
1005 	if (err)
1006 		goto out_unregister_raw_proto;
1007 
1008 	/* We MUST register RAW sockets before we create the ICMP6,
1009 	 * IGMP6, or NDISC control sockets.
1010 	 */
1011 	err = rawv6_init();
1012 	if (err)
1013 		goto out_unregister_ping_proto;
1014 
1015 	/* Register the family here so that the init calls below will
1016 	 * be able to create sockets. (?? is this dangerous ??)
1017 	 */
1018 	err = sock_register(&inet6_family_ops);
1019 	if (err)
1020 		goto out_sock_register_fail;
1021 
1022 	/*
1023 	 *	ipngwg API draft makes clear that the correct semantics
1024 	 *	for TCP and UDP is to consider one TCP and UDP instance
1025 	 *	in a host available by both INET and INET6 APIs and
1026 	 *	able to communicate via both network protocols.
1027 	 */
1028 
1029 	err = register_pernet_subsys(&inet6_net_ops);
1030 	if (err)
1031 		goto register_pernet_fail;
1032 	err = ip6_mr_init();
1033 	if (err)
1034 		goto ipmr_fail;
1035 	err = icmpv6_init();
1036 	if (err)
1037 		goto icmp_fail;
1038 	err = ndisc_init();
1039 	if (err)
1040 		goto ndisc_fail;
1041 	err = igmp6_init();
1042 	if (err)
1043 		goto igmp_fail;
1044 
1045 	err = ipv6_netfilter_init();
1046 	if (err)
1047 		goto netfilter_fail;
1048 	/* Create /proc/foo6 entries. */
1049 #ifdef CONFIG_PROC_FS
1050 	err = -ENOMEM;
1051 	if (raw6_proc_init())
1052 		goto proc_raw6_fail;
1053 	if (udplite6_proc_init())
1054 		goto proc_udplite6_fail;
1055 	if (ipv6_misc_proc_init())
1056 		goto proc_misc6_fail;
1057 	if (if6_proc_init())
1058 		goto proc_if6_fail;
1059 #endif
1060 	err = ip6_route_init();
1061 	if (err)
1062 		goto ip6_route_fail;
1063 	err = ndisc_late_init();
1064 	if (err)
1065 		goto ndisc_late_fail;
1066 	err = ip6_flowlabel_init();
1067 	if (err)
1068 		goto ip6_flowlabel_fail;
1069 	err = ipv6_anycast_init();
1070 	if (err)
1071 		goto ipv6_anycast_fail;
1072 	err = addrconf_init();
1073 	if (err)
1074 		goto addrconf_fail;
1075 
1076 	/* Init v6 extension headers. */
1077 	err = ipv6_exthdrs_init();
1078 	if (err)
1079 		goto ipv6_exthdrs_fail;
1080 
1081 	err = ipv6_frag_init();
1082 	if (err)
1083 		goto ipv6_frag_fail;
1084 
1085 	/* Init v6 transport protocols. */
1086 	err = udpv6_init();
1087 	if (err)
1088 		goto udpv6_fail;
1089 
1090 	err = udplitev6_init();
1091 	if (err)
1092 		goto udplitev6_fail;
1093 
1094 	err = udpv6_offload_init();
1095 	if (err)
1096 		goto udpv6_offload_fail;
1097 
1098 	err = tcpv6_init();
1099 	if (err)
1100 		goto tcpv6_fail;
1101 
1102 	err = ipv6_packet_init();
1103 	if (err)
1104 		goto ipv6_packet_fail;
1105 
1106 	err = pingv6_init();
1107 	if (err)
1108 		goto pingv6_fail;
1109 
1110 	err = calipso_init();
1111 	if (err)
1112 		goto calipso_fail;
1113 
1114 	err = seg6_init();
1115 	if (err)
1116 		goto seg6_fail;
1117 
1118 	err = rpl_init();
1119 	if (err)
1120 		goto rpl_fail;
1121 
1122 	err = igmp6_late_init();
1123 	if (err)
1124 		goto igmp6_late_err;
1125 
1126 #ifdef CONFIG_SYSCTL
1127 	err = ipv6_sysctl_register();
1128 	if (err)
1129 		goto sysctl_fail;
1130 #endif
1131 
1132 	/* ensure that ipv6 stubs are visible only after ipv6 is ready */
1133 	wmb();
1134 	ipv6_stub = &ipv6_stub_impl;
1135 	ipv6_bpf_stub = &ipv6_bpf_stub_impl;
1136 out:
1137 	return err;
1138 
1139 #ifdef CONFIG_SYSCTL
1140 sysctl_fail:
1141 	igmp6_late_cleanup();
1142 #endif
1143 igmp6_late_err:
1144 	rpl_exit();
1145 rpl_fail:
1146 	seg6_exit();
1147 seg6_fail:
1148 	calipso_exit();
1149 calipso_fail:
1150 	pingv6_exit();
1151 pingv6_fail:
1152 	ipv6_packet_cleanup();
1153 ipv6_packet_fail:
1154 	tcpv6_exit();
1155 tcpv6_fail:
1156 	udpv6_offload_exit();
1157 udpv6_offload_fail:
1158 	udplitev6_exit();
1159 udplitev6_fail:
1160 	udpv6_exit();
1161 udpv6_fail:
1162 	ipv6_frag_exit();
1163 ipv6_frag_fail:
1164 	ipv6_exthdrs_exit();
1165 ipv6_exthdrs_fail:
1166 	addrconf_cleanup();
1167 addrconf_fail:
1168 	ipv6_anycast_cleanup();
1169 ipv6_anycast_fail:
1170 	ip6_flowlabel_cleanup();
1171 ip6_flowlabel_fail:
1172 	ndisc_late_cleanup();
1173 ndisc_late_fail:
1174 	ip6_route_cleanup();
1175 ip6_route_fail:
1176 #ifdef CONFIG_PROC_FS
1177 	if6_proc_exit();
1178 proc_if6_fail:
1179 	ipv6_misc_proc_exit();
1180 proc_misc6_fail:
1181 	udplite6_proc_exit();
1182 proc_udplite6_fail:
1183 	raw6_proc_exit();
1184 proc_raw6_fail:
1185 #endif
1186 	ipv6_netfilter_fini();
1187 netfilter_fail:
1188 	igmp6_cleanup();
1189 igmp_fail:
1190 	ndisc_cleanup();
1191 ndisc_fail:
1192 	icmpv6_cleanup();
1193 icmp_fail:
1194 	ip6_mr_cleanup();
1195 ipmr_fail:
1196 	unregister_pernet_subsys(&inet6_net_ops);
1197 register_pernet_fail:
1198 	sock_unregister(PF_INET6);
1199 	rtnl_unregister_all(PF_INET6);
1200 out_sock_register_fail:
1201 	rawv6_exit();
1202 out_unregister_ping_proto:
1203 	proto_unregister(&pingv6_prot);
1204 out_unregister_raw_proto:
1205 	proto_unregister(&rawv6_prot);
1206 out_unregister_udplite_proto:
1207 	proto_unregister(&udplitev6_prot);
1208 out_unregister_udp_proto:
1209 	proto_unregister(&udpv6_prot);
1210 out_unregister_tcp_proto:
1211 	proto_unregister(&tcpv6_prot);
1212 	goto out;
1213 }
1214 module_init(inet6_init);
1215 
1216 MODULE_ALIAS_NETPROTO(PF_INET6);
1217