xref: /linux/net/ipv6/af_inet6.c (revision b889fcf63cb62e7fdb7816565e28f44dbe4a76a5)
1 /*
2  *	PF_INET6 socket protocol family
3  *	Linux INET6 implementation
4  *
5  *	Authors:
6  *	Pedro Roque		<roque@di.fc.ul.pt>
7  *
8  *	Adapted from linux/net/ipv4/af_inet.c
9  *
10  * 	Fixes:
11  *	piggy, Karl Knutson	:	Socket protocol table
12  * 	Hideaki YOSHIFUJI	:	sin6_scope_id support
13  * 	Arnaldo Melo		: 	check proc_net_create return, cleanups
14  *
15  *	This program is free software; you can redistribute it and/or
16  *      modify it under the terms of the GNU General Public License
17  *      as published by the Free Software Foundation; either version
18  *      2 of the License, or (at your option) any later version.
19  */
20 
21 #define pr_fmt(fmt) "IPv6: " fmt
22 
23 #include <linux/module.h>
24 #include <linux/capability.h>
25 #include <linux/errno.h>
26 #include <linux/types.h>
27 #include <linux/socket.h>
28 #include <linux/in.h>
29 #include <linux/kernel.h>
30 #include <linux/timer.h>
31 #include <linux/string.h>
32 #include <linux/sockios.h>
33 #include <linux/net.h>
34 #include <linux/fcntl.h>
35 #include <linux/mm.h>
36 #include <linux/interrupt.h>
37 #include <linux/proc_fs.h>
38 #include <linux/stat.h>
39 #include <linux/init.h>
40 #include <linux/slab.h>
41 
42 #include <linux/inet.h>
43 #include <linux/netdevice.h>
44 #include <linux/icmpv6.h>
45 #include <linux/netfilter_ipv6.h>
46 
47 #include <net/ip.h>
48 #include <net/ipv6.h>
49 #include <net/udp.h>
50 #include <net/udplite.h>
51 #include <net/tcp.h>
52 #include <net/ipip.h>
53 #include <net/protocol.h>
54 #include <net/inet_common.h>
55 #include <net/route.h>
56 #include <net/transp_v6.h>
57 #include <net/ip6_route.h>
58 #include <net/addrconf.h>
59 #ifdef CONFIG_IPV6_TUNNEL
60 #include <net/ip6_tunnel.h>
61 #endif
62 
63 #include <asm/uaccess.h>
64 #include <linux/mroute6.h>
65 
66 MODULE_AUTHOR("Cast of dozens");
67 MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
68 MODULE_LICENSE("GPL");
69 
70 /* The inetsw6 table contains everything that inet6_create needs to
71  * build a new socket.
72  */
73 static struct list_head inetsw6[SOCK_MAX];
74 static DEFINE_SPINLOCK(inetsw6_lock);
75 
76 struct ipv6_params ipv6_defaults = {
77 	.disable_ipv6 = 0,
78 	.autoconf = 1,
79 };
80 
81 static int disable_ipv6_mod;
82 
83 module_param_named(disable, disable_ipv6_mod, int, 0444);
84 MODULE_PARM_DESC(disable, "Disable IPv6 module such that it is non-functional");
85 
86 module_param_named(disable_ipv6, ipv6_defaults.disable_ipv6, int, 0444);
87 MODULE_PARM_DESC(disable_ipv6, "Disable IPv6 on all interfaces");
88 
89 module_param_named(autoconf, ipv6_defaults.autoconf, int, 0444);
90 MODULE_PARM_DESC(autoconf, "Enable IPv6 address autoconfiguration on all interfaces");
91 
92 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
93 {
94 	const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
95 
96 	return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
97 }
98 
99 static int inet6_create(struct net *net, struct socket *sock, int protocol,
100 			int kern)
101 {
102 	struct inet_sock *inet;
103 	struct ipv6_pinfo *np;
104 	struct sock *sk;
105 	struct inet_protosw *answer;
106 	struct proto *answer_prot;
107 	unsigned char answer_flags;
108 	char answer_no_check;
109 	int try_loading_module = 0;
110 	int err;
111 
112 	if (sock->type != SOCK_RAW &&
113 	    sock->type != SOCK_DGRAM &&
114 	    !inet_ehash_secret)
115 		build_ehash_secret();
116 
117 	/* Look for the requested type/protocol pair. */
118 lookup_protocol:
119 	err = -ESOCKTNOSUPPORT;
120 	rcu_read_lock();
121 	list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
122 
123 		err = 0;
124 		/* Check the non-wild match. */
125 		if (protocol == answer->protocol) {
126 			if (protocol != IPPROTO_IP)
127 				break;
128 		} else {
129 			/* Check for the two wild cases. */
130 			if (IPPROTO_IP == protocol) {
131 				protocol = answer->protocol;
132 				break;
133 			}
134 			if (IPPROTO_IP == answer->protocol)
135 				break;
136 		}
137 		err = -EPROTONOSUPPORT;
138 	}
139 
140 	if (err) {
141 		if (try_loading_module < 2) {
142 			rcu_read_unlock();
143 			/*
144 			 * Be more specific, e.g. net-pf-10-proto-132-type-1
145 			 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
146 			 */
147 			if (++try_loading_module == 1)
148 				request_module("net-pf-%d-proto-%d-type-%d",
149 						PF_INET6, protocol, sock->type);
150 			/*
151 			 * Fall back to generic, e.g. net-pf-10-proto-132
152 			 * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
153 			 */
154 			else
155 				request_module("net-pf-%d-proto-%d",
156 						PF_INET6, protocol);
157 			goto lookup_protocol;
158 		} else
159 			goto out_rcu_unlock;
160 	}
161 
162 	err = -EPERM;
163 	if (sock->type == SOCK_RAW && !kern &&
164 	    !ns_capable(net->user_ns, CAP_NET_RAW))
165 		goto out_rcu_unlock;
166 
167 	sock->ops = answer->ops;
168 	answer_prot = answer->prot;
169 	answer_no_check = answer->no_check;
170 	answer_flags = answer->flags;
171 	rcu_read_unlock();
172 
173 	WARN_ON(answer_prot->slab == NULL);
174 
175 	err = -ENOBUFS;
176 	sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot);
177 	if (sk == NULL)
178 		goto out;
179 
180 	sock_init_data(sock, sk);
181 
182 	err = 0;
183 	sk->sk_no_check = answer_no_check;
184 	if (INET_PROTOSW_REUSE & answer_flags)
185 		sk->sk_reuse = SK_CAN_REUSE;
186 
187 	inet = inet_sk(sk);
188 	inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
189 
190 	if (SOCK_RAW == sock->type) {
191 		inet->inet_num = protocol;
192 		if (IPPROTO_RAW == protocol)
193 			inet->hdrincl = 1;
194 	}
195 
196 	sk->sk_destruct		= inet_sock_destruct;
197 	sk->sk_family		= PF_INET6;
198 	sk->sk_protocol		= protocol;
199 
200 	sk->sk_backlog_rcv	= answer->prot->backlog_rcv;
201 
202 	inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
203 	np->hop_limit	= -1;
204 	np->mcast_hops	= IPV6_DEFAULT_MCASTHOPS;
205 	np->mc_loop	= 1;
206 	np->pmtudisc	= IPV6_PMTUDISC_WANT;
207 	np->ipv6only	= net->ipv6.sysctl.bindv6only;
208 
209 	/* Init the ipv4 part of the socket since we can have sockets
210 	 * using v6 API for ipv4.
211 	 */
212 	inet->uc_ttl	= -1;
213 
214 	inet->mc_loop	= 1;
215 	inet->mc_ttl	= 1;
216 	inet->mc_index	= 0;
217 	inet->mc_list	= NULL;
218 	inet->rcv_tos	= 0;
219 
220 	if (ipv4_config.no_pmtu_disc)
221 		inet->pmtudisc = IP_PMTUDISC_DONT;
222 	else
223 		inet->pmtudisc = IP_PMTUDISC_WANT;
224 	/*
225 	 * Increment only the relevant sk_prot->socks debug field, this changes
226 	 * the previous behaviour of incrementing both the equivalent to
227 	 * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
228 	 *
229 	 * This allows better debug granularity as we'll know exactly how many
230 	 * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
231 	 * transport protocol socks. -acme
232 	 */
233 	sk_refcnt_debug_inc(sk);
234 
235 	if (inet->inet_num) {
236 		/* It assumes that any protocol which allows
237 		 * the user to assign a number at socket
238 		 * creation time automatically shares.
239 		 */
240 		inet->inet_sport = htons(inet->inet_num);
241 		sk->sk_prot->hash(sk);
242 	}
243 	if (sk->sk_prot->init) {
244 		err = sk->sk_prot->init(sk);
245 		if (err) {
246 			sk_common_release(sk);
247 			goto out;
248 		}
249 	}
250 out:
251 	return err;
252 out_rcu_unlock:
253 	rcu_read_unlock();
254 	goto out;
255 }
256 
257 
258 /* bind for INET6 API */
259 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
260 {
261 	struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr;
262 	struct sock *sk = sock->sk;
263 	struct inet_sock *inet = inet_sk(sk);
264 	struct ipv6_pinfo *np = inet6_sk(sk);
265 	struct net *net = sock_net(sk);
266 	__be32 v4addr = 0;
267 	unsigned short snum;
268 	int addr_type = 0;
269 	int err = 0;
270 
271 	/* If the socket has its own bind function then use it. */
272 	if (sk->sk_prot->bind)
273 		return sk->sk_prot->bind(sk, uaddr, addr_len);
274 
275 	if (addr_len < SIN6_LEN_RFC2133)
276 		return -EINVAL;
277 
278 	if (addr->sin6_family != AF_INET6)
279 		return -EAFNOSUPPORT;
280 
281 	addr_type = ipv6_addr_type(&addr->sin6_addr);
282 	if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
283 		return -EINVAL;
284 
285 	snum = ntohs(addr->sin6_port);
286 	if (snum && snum < PROT_SOCK && !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
287 		return -EACCES;
288 
289 	lock_sock(sk);
290 
291 	/* Check these errors (active socket, double bind). */
292 	if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
293 		err = -EINVAL;
294 		goto out;
295 	}
296 
297 	/* Check if the address belongs to the host. */
298 	if (addr_type == IPV6_ADDR_MAPPED) {
299 		int chk_addr_ret;
300 
301 		/* Binding to v4-mapped address on a v6-only socket
302 		 * makes no sense
303 		 */
304 		if (np->ipv6only) {
305 			err = -EINVAL;
306 			goto out;
307 		}
308 
309 		/* Reproduce AF_INET checks to make the bindings consistent */
310 		v4addr = addr->sin6_addr.s6_addr32[3];
311 		chk_addr_ret = inet_addr_type(net, v4addr);
312 		if (!sysctl_ip_nonlocal_bind &&
313 		    !(inet->freebind || inet->transparent) &&
314 		    v4addr != htonl(INADDR_ANY) &&
315 		    chk_addr_ret != RTN_LOCAL &&
316 		    chk_addr_ret != RTN_MULTICAST &&
317 		    chk_addr_ret != RTN_BROADCAST) {
318 			err = -EADDRNOTAVAIL;
319 			goto out;
320 		}
321 	} else {
322 		if (addr_type != IPV6_ADDR_ANY) {
323 			struct net_device *dev = NULL;
324 
325 			rcu_read_lock();
326 			if (addr_type & IPV6_ADDR_LINKLOCAL) {
327 				if (addr_len >= sizeof(struct sockaddr_in6) &&
328 				    addr->sin6_scope_id) {
329 					/* Override any existing binding, if another one
330 					 * is supplied by user.
331 					 */
332 					sk->sk_bound_dev_if = addr->sin6_scope_id;
333 				}
334 
335 				/* Binding to link-local address requires an interface */
336 				if (!sk->sk_bound_dev_if) {
337 					err = -EINVAL;
338 					goto out_unlock;
339 				}
340 				dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
341 				if (!dev) {
342 					err = -ENODEV;
343 					goto out_unlock;
344 				}
345 			}
346 
347 			/* ipv4 addr of the socket is invalid.  Only the
348 			 * unspecified and mapped address have a v4 equivalent.
349 			 */
350 			v4addr = LOOPBACK4_IPV6;
351 			if (!(addr_type & IPV6_ADDR_MULTICAST))	{
352 				if (!(inet->freebind || inet->transparent) &&
353 				    !ipv6_chk_addr(net, &addr->sin6_addr,
354 						   dev, 0)) {
355 					err = -EADDRNOTAVAIL;
356 					goto out_unlock;
357 				}
358 			}
359 			rcu_read_unlock();
360 		}
361 	}
362 
363 	inet->inet_rcv_saddr = v4addr;
364 	inet->inet_saddr = v4addr;
365 
366 	np->rcv_saddr = addr->sin6_addr;
367 
368 	if (!(addr_type & IPV6_ADDR_MULTICAST))
369 		np->saddr = addr->sin6_addr;
370 
371 	/* Make sure we are allowed to bind here. */
372 	if (sk->sk_prot->get_port(sk, snum)) {
373 		inet_reset_saddr(sk);
374 		err = -EADDRINUSE;
375 		goto out;
376 	}
377 
378 	if (addr_type != IPV6_ADDR_ANY) {
379 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
380 		if (addr_type != IPV6_ADDR_MAPPED)
381 			np->ipv6only = 1;
382 	}
383 	if (snum)
384 		sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
385 	inet->inet_sport = htons(inet->inet_num);
386 	inet->inet_dport = 0;
387 	inet->inet_daddr = 0;
388 out:
389 	release_sock(sk);
390 	return err;
391 out_unlock:
392 	rcu_read_unlock();
393 	goto out;
394 }
395 EXPORT_SYMBOL(inet6_bind);
396 
397 int inet6_release(struct socket *sock)
398 {
399 	struct sock *sk = sock->sk;
400 
401 	if (sk == NULL)
402 		return -EINVAL;
403 
404 	/* Free mc lists */
405 	ipv6_sock_mc_close(sk);
406 
407 	/* Free ac lists */
408 	ipv6_sock_ac_close(sk);
409 
410 	return inet_release(sock);
411 }
412 EXPORT_SYMBOL(inet6_release);
413 
414 void inet6_destroy_sock(struct sock *sk)
415 {
416 	struct ipv6_pinfo *np = inet6_sk(sk);
417 	struct sk_buff *skb;
418 	struct ipv6_txoptions *opt;
419 
420 	/* Release rx options */
421 
422 	skb = xchg(&np->pktoptions, NULL);
423 	if (skb != NULL)
424 		kfree_skb(skb);
425 
426 	skb = xchg(&np->rxpmtu, NULL);
427 	if (skb != NULL)
428 		kfree_skb(skb);
429 
430 	/* Free flowlabels */
431 	fl6_free_socklist(sk);
432 
433 	/* Free tx options */
434 
435 	opt = xchg(&np->opt, NULL);
436 	if (opt != NULL)
437 		sock_kfree_s(sk, opt, opt->tot_len);
438 }
439 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
440 
441 /*
442  *	This does both peername and sockname.
443  */
444 
445 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
446 		 int *uaddr_len, int peer)
447 {
448 	struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
449 	struct sock *sk = sock->sk;
450 	struct inet_sock *inet = inet_sk(sk);
451 	struct ipv6_pinfo *np = inet6_sk(sk);
452 
453 	sin->sin6_family = AF_INET6;
454 	sin->sin6_flowinfo = 0;
455 	sin->sin6_scope_id = 0;
456 	if (peer) {
457 		if (!inet->inet_dport)
458 			return -ENOTCONN;
459 		if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
460 		    peer == 1)
461 			return -ENOTCONN;
462 		sin->sin6_port = inet->inet_dport;
463 		sin->sin6_addr = np->daddr;
464 		if (np->sndflow)
465 			sin->sin6_flowinfo = np->flow_label;
466 	} else {
467 		if (ipv6_addr_any(&np->rcv_saddr))
468 			sin->sin6_addr = np->saddr;
469 		else
470 			sin->sin6_addr = np->rcv_saddr;
471 
472 		sin->sin6_port = inet->inet_sport;
473 	}
474 	if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL)
475 		sin->sin6_scope_id = sk->sk_bound_dev_if;
476 	*uaddr_len = sizeof(*sin);
477 	return 0;
478 }
479 EXPORT_SYMBOL(inet6_getname);
480 
481 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
482 {
483 	struct sock *sk = sock->sk;
484 	struct net *net = sock_net(sk);
485 
486 	switch (cmd) {
487 	case SIOCGSTAMP:
488 		return sock_get_timestamp(sk, (struct timeval __user *)arg);
489 
490 	case SIOCGSTAMPNS:
491 		return sock_get_timestampns(sk, (struct timespec __user *)arg);
492 
493 	case SIOCADDRT:
494 	case SIOCDELRT:
495 
496 		return ipv6_route_ioctl(net, cmd, (void __user *)arg);
497 
498 	case SIOCSIFADDR:
499 		return addrconf_add_ifaddr(net, (void __user *) arg);
500 	case SIOCDIFADDR:
501 		return addrconf_del_ifaddr(net, (void __user *) arg);
502 	case SIOCSIFDSTADDR:
503 		return addrconf_set_dstaddr(net, (void __user *) arg);
504 	default:
505 		if (!sk->sk_prot->ioctl)
506 			return -ENOIOCTLCMD;
507 		return sk->sk_prot->ioctl(sk, cmd, arg);
508 	}
509 	/*NOTREACHED*/
510 	return 0;
511 }
512 EXPORT_SYMBOL(inet6_ioctl);
513 
514 const struct proto_ops inet6_stream_ops = {
515 	.family		   = PF_INET6,
516 	.owner		   = THIS_MODULE,
517 	.release	   = inet6_release,
518 	.bind		   = inet6_bind,
519 	.connect	   = inet_stream_connect,	/* ok		*/
520 	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
521 	.accept		   = inet_accept,		/* ok		*/
522 	.getname	   = inet6_getname,
523 	.poll		   = tcp_poll,			/* ok		*/
524 	.ioctl		   = inet6_ioctl,		/* must change  */
525 	.listen		   = inet_listen,		/* ok		*/
526 	.shutdown	   = inet_shutdown,		/* ok		*/
527 	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
528 	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
529 	.sendmsg	   = inet_sendmsg,		/* ok		*/
530 	.recvmsg	   = inet_recvmsg,		/* ok		*/
531 	.mmap		   = sock_no_mmap,
532 	.sendpage	   = inet_sendpage,
533 	.splice_read	   = tcp_splice_read,
534 #ifdef CONFIG_COMPAT
535 	.compat_setsockopt = compat_sock_common_setsockopt,
536 	.compat_getsockopt = compat_sock_common_getsockopt,
537 #endif
538 };
539 
540 const struct proto_ops inet6_dgram_ops = {
541 	.family		   = PF_INET6,
542 	.owner		   = THIS_MODULE,
543 	.release	   = inet6_release,
544 	.bind		   = inet6_bind,
545 	.connect	   = inet_dgram_connect,	/* ok		*/
546 	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
547 	.accept		   = sock_no_accept,		/* a do nothing	*/
548 	.getname	   = inet6_getname,
549 	.poll		   = udp_poll,			/* ok		*/
550 	.ioctl		   = inet6_ioctl,		/* must change  */
551 	.listen		   = sock_no_listen,		/* ok		*/
552 	.shutdown	   = inet_shutdown,		/* ok		*/
553 	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
554 	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
555 	.sendmsg	   = inet_sendmsg,		/* ok		*/
556 	.recvmsg	   = inet_recvmsg,		/* ok		*/
557 	.mmap		   = sock_no_mmap,
558 	.sendpage	   = sock_no_sendpage,
559 #ifdef CONFIG_COMPAT
560 	.compat_setsockopt = compat_sock_common_setsockopt,
561 	.compat_getsockopt = compat_sock_common_getsockopt,
562 #endif
563 };
564 
565 static const struct net_proto_family inet6_family_ops = {
566 	.family = PF_INET6,
567 	.create = inet6_create,
568 	.owner	= THIS_MODULE,
569 };
570 
571 int inet6_register_protosw(struct inet_protosw *p)
572 {
573 	struct list_head *lh;
574 	struct inet_protosw *answer;
575 	struct list_head *last_perm;
576 	int protocol = p->protocol;
577 	int ret;
578 
579 	spin_lock_bh(&inetsw6_lock);
580 
581 	ret = -EINVAL;
582 	if (p->type >= SOCK_MAX)
583 		goto out_illegal;
584 
585 	/* If we are trying to override a permanent protocol, bail. */
586 	answer = NULL;
587 	ret = -EPERM;
588 	last_perm = &inetsw6[p->type];
589 	list_for_each(lh, &inetsw6[p->type]) {
590 		answer = list_entry(lh, struct inet_protosw, list);
591 
592 		/* Check only the non-wild match. */
593 		if (INET_PROTOSW_PERMANENT & answer->flags) {
594 			if (protocol == answer->protocol)
595 				break;
596 			last_perm = lh;
597 		}
598 
599 		answer = NULL;
600 	}
601 	if (answer)
602 		goto out_permanent;
603 
604 	/* Add the new entry after the last permanent entry if any, so that
605 	 * the new entry does not override a permanent entry when matched with
606 	 * a wild-card protocol. But it is allowed to override any existing
607 	 * non-permanent entry.  This means that when we remove this entry, the
608 	 * system automatically returns to the old behavior.
609 	 */
610 	list_add_rcu(&p->list, last_perm);
611 	ret = 0;
612 out:
613 	spin_unlock_bh(&inetsw6_lock);
614 	return ret;
615 
616 out_permanent:
617 	pr_err("Attempt to override permanent protocol %d\n", protocol);
618 	goto out;
619 
620 out_illegal:
621 	pr_err("Ignoring attempt to register invalid socket type %d\n",
622 	       p->type);
623 	goto out;
624 }
625 EXPORT_SYMBOL(inet6_register_protosw);
626 
627 void
628 inet6_unregister_protosw(struct inet_protosw *p)
629 {
630 	if (INET_PROTOSW_PERMANENT & p->flags) {
631 		pr_err("Attempt to unregister permanent protocol %d\n",
632 		       p->protocol);
633 	} else {
634 		spin_lock_bh(&inetsw6_lock);
635 		list_del_rcu(&p->list);
636 		spin_unlock_bh(&inetsw6_lock);
637 
638 		synchronize_net();
639 	}
640 }
641 EXPORT_SYMBOL(inet6_unregister_protosw);
642 
643 int inet6_sk_rebuild_header(struct sock *sk)
644 {
645 	struct ipv6_pinfo *np = inet6_sk(sk);
646 	struct dst_entry *dst;
647 
648 	dst = __sk_dst_check(sk, np->dst_cookie);
649 
650 	if (dst == NULL) {
651 		struct inet_sock *inet = inet_sk(sk);
652 		struct in6_addr *final_p, final;
653 		struct flowi6 fl6;
654 
655 		memset(&fl6, 0, sizeof(fl6));
656 		fl6.flowi6_proto = sk->sk_protocol;
657 		fl6.daddr = np->daddr;
658 		fl6.saddr = np->saddr;
659 		fl6.flowlabel = np->flow_label;
660 		fl6.flowi6_oif = sk->sk_bound_dev_if;
661 		fl6.flowi6_mark = sk->sk_mark;
662 		fl6.fl6_dport = inet->inet_dport;
663 		fl6.fl6_sport = inet->inet_sport;
664 		security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
665 
666 		final_p = fl6_update_dst(&fl6, np->opt, &final);
667 
668 		dst = ip6_dst_lookup_flow(sk, &fl6, final_p, false);
669 		if (IS_ERR(dst)) {
670 			sk->sk_route_caps = 0;
671 			sk->sk_err_soft = -PTR_ERR(dst);
672 			return PTR_ERR(dst);
673 		}
674 
675 		__ip6_dst_store(sk, dst, NULL, NULL);
676 	}
677 
678 	return 0;
679 }
680 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
681 
682 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb)
683 {
684 	const struct ipv6_pinfo *np = inet6_sk(sk);
685 	const struct inet6_skb_parm *opt = IP6CB(skb);
686 
687 	if (np->rxopt.all) {
688 		if ((opt->hop && (np->rxopt.bits.hopopts ||
689 				  np->rxopt.bits.ohopopts)) ||
690 		    ((IPV6_FLOWINFO_MASK &
691 		      *(__be32 *)skb_network_header(skb)) &&
692 		     np->rxopt.bits.rxflow) ||
693 		    (opt->srcrt && (np->rxopt.bits.srcrt ||
694 		     np->rxopt.bits.osrcrt)) ||
695 		    ((opt->dst1 || opt->dst0) &&
696 		     (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
697 			return true;
698 	}
699 	return false;
700 }
701 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
702 
703 static struct packet_type ipv6_packet_type __read_mostly = {
704 	.type = cpu_to_be16(ETH_P_IPV6),
705 	.func = ipv6_rcv,
706 };
707 
708 static int __init ipv6_packet_init(void)
709 {
710 	dev_add_pack(&ipv6_packet_type);
711 	return 0;
712 }
713 
714 static void ipv6_packet_cleanup(void)
715 {
716 	dev_remove_pack(&ipv6_packet_type);
717 }
718 
719 static int __net_init ipv6_init_mibs(struct net *net)
720 {
721 	if (snmp_mib_init((void __percpu **)net->mib.udp_stats_in6,
722 			  sizeof(struct udp_mib),
723 			  __alignof__(struct udp_mib)) < 0)
724 		return -ENOMEM;
725 	if (snmp_mib_init((void __percpu **)net->mib.udplite_stats_in6,
726 			  sizeof(struct udp_mib),
727 			  __alignof__(struct udp_mib)) < 0)
728 		goto err_udplite_mib;
729 	if (snmp_mib_init((void __percpu **)net->mib.ipv6_statistics,
730 			  sizeof(struct ipstats_mib),
731 			  __alignof__(struct ipstats_mib)) < 0)
732 		goto err_ip_mib;
733 	if (snmp_mib_init((void __percpu **)net->mib.icmpv6_statistics,
734 			  sizeof(struct icmpv6_mib),
735 			  __alignof__(struct icmpv6_mib)) < 0)
736 		goto err_icmp_mib;
737 	net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
738 						GFP_KERNEL);
739 	if (!net->mib.icmpv6msg_statistics)
740 		goto err_icmpmsg_mib;
741 	return 0;
742 
743 err_icmpmsg_mib:
744 	snmp_mib_free((void __percpu **)net->mib.icmpv6_statistics);
745 err_icmp_mib:
746 	snmp_mib_free((void __percpu **)net->mib.ipv6_statistics);
747 err_ip_mib:
748 	snmp_mib_free((void __percpu **)net->mib.udplite_stats_in6);
749 err_udplite_mib:
750 	snmp_mib_free((void __percpu **)net->mib.udp_stats_in6);
751 	return -ENOMEM;
752 }
753 
754 static void ipv6_cleanup_mibs(struct net *net)
755 {
756 	snmp_mib_free((void __percpu **)net->mib.udp_stats_in6);
757 	snmp_mib_free((void __percpu **)net->mib.udplite_stats_in6);
758 	snmp_mib_free((void __percpu **)net->mib.ipv6_statistics);
759 	snmp_mib_free((void __percpu **)net->mib.icmpv6_statistics);
760 	kfree(net->mib.icmpv6msg_statistics);
761 }
762 
763 static int __net_init inet6_net_init(struct net *net)
764 {
765 	int err = 0;
766 
767 	net->ipv6.sysctl.bindv6only = 0;
768 	net->ipv6.sysctl.icmpv6_time = 1*HZ;
769 
770 	err = ipv6_init_mibs(net);
771 	if (err)
772 		return err;
773 #ifdef CONFIG_PROC_FS
774 	err = udp6_proc_init(net);
775 	if (err)
776 		goto out;
777 	err = tcp6_proc_init(net);
778 	if (err)
779 		goto proc_tcp6_fail;
780 	err = ac6_proc_init(net);
781 	if (err)
782 		goto proc_ac6_fail;
783 #endif
784 	return err;
785 
786 #ifdef CONFIG_PROC_FS
787 proc_ac6_fail:
788 	tcp6_proc_exit(net);
789 proc_tcp6_fail:
790 	udp6_proc_exit(net);
791 out:
792 	ipv6_cleanup_mibs(net);
793 	return err;
794 #endif
795 }
796 
797 static void __net_exit inet6_net_exit(struct net *net)
798 {
799 #ifdef CONFIG_PROC_FS
800 	udp6_proc_exit(net);
801 	tcp6_proc_exit(net);
802 	ac6_proc_exit(net);
803 #endif
804 	ipv6_cleanup_mibs(net);
805 }
806 
807 static struct pernet_operations inet6_net_ops = {
808 	.init = inet6_net_init,
809 	.exit = inet6_net_exit,
810 };
811 
812 static int __init inet6_init(void)
813 {
814 	struct sk_buff *dummy_skb;
815 	struct list_head *r;
816 	int err = 0;
817 
818 	BUILD_BUG_ON(sizeof(struct inet6_skb_parm) > sizeof(dummy_skb->cb));
819 
820 	/* Register the socket-side information for inet6_create.  */
821 	for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
822 		INIT_LIST_HEAD(r);
823 
824 	if (disable_ipv6_mod) {
825 		pr_info("Loaded, but administratively disabled, reboot required to enable\n");
826 		goto out;
827 	}
828 
829 	err = proto_register(&tcpv6_prot, 1);
830 	if (err)
831 		goto out;
832 
833 	err = proto_register(&udpv6_prot, 1);
834 	if (err)
835 		goto out_unregister_tcp_proto;
836 
837 	err = proto_register(&udplitev6_prot, 1);
838 	if (err)
839 		goto out_unregister_udp_proto;
840 
841 	err = proto_register(&rawv6_prot, 1);
842 	if (err)
843 		goto out_unregister_udplite_proto;
844 
845 
846 	/* We MUST register RAW sockets before we create the ICMP6,
847 	 * IGMP6, or NDISC control sockets.
848 	 */
849 	err = rawv6_init();
850 	if (err)
851 		goto out_unregister_raw_proto;
852 
853 	/* Register the family here so that the init calls below will
854 	 * be able to create sockets. (?? is this dangerous ??)
855 	 */
856 	err = sock_register(&inet6_family_ops);
857 	if (err)
858 		goto out_sock_register_fail;
859 
860 	tcpv6_prot.sysctl_mem = init_net.ipv4.sysctl_tcp_mem;
861 
862 	/*
863 	 *	ipngwg API draft makes clear that the correct semantics
864 	 *	for TCP and UDP is to consider one TCP and UDP instance
865 	 *	in a host available by both INET and INET6 APIs and
866 	 *	able to communicate via both network protocols.
867 	 */
868 
869 	err = register_pernet_subsys(&inet6_net_ops);
870 	if (err)
871 		goto register_pernet_fail;
872 	err = icmpv6_init();
873 	if (err)
874 		goto icmp_fail;
875 	err = ip6_mr_init();
876 	if (err)
877 		goto ipmr_fail;
878 	err = ndisc_init();
879 	if (err)
880 		goto ndisc_fail;
881 	err = igmp6_init();
882 	if (err)
883 		goto igmp_fail;
884 	err = ipv6_netfilter_init();
885 	if (err)
886 		goto netfilter_fail;
887 	/* Create /proc/foo6 entries. */
888 #ifdef CONFIG_PROC_FS
889 	err = -ENOMEM;
890 	if (raw6_proc_init())
891 		goto proc_raw6_fail;
892 	if (udplite6_proc_init())
893 		goto proc_udplite6_fail;
894 	if (ipv6_misc_proc_init())
895 		goto proc_misc6_fail;
896 	if (if6_proc_init())
897 		goto proc_if6_fail;
898 #endif
899 	err = ip6_route_init();
900 	if (err)
901 		goto ip6_route_fail;
902 	err = ip6_flowlabel_init();
903 	if (err)
904 		goto ip6_flowlabel_fail;
905 	err = addrconf_init();
906 	if (err)
907 		goto addrconf_fail;
908 
909 	/* Init v6 extension headers. */
910 	err = ipv6_exthdrs_init();
911 	if (err)
912 		goto ipv6_exthdrs_fail;
913 
914 	err = ipv6_frag_init();
915 	if (err)
916 		goto ipv6_frag_fail;
917 
918 	/* Init v6 transport protocols. */
919 	err = udpv6_init();
920 	if (err)
921 		goto udpv6_fail;
922 
923 	err = udplitev6_init();
924 	if (err)
925 		goto udplitev6_fail;
926 
927 	err = tcpv6_init();
928 	if (err)
929 		goto tcpv6_fail;
930 
931 	err = ipv6_packet_init();
932 	if (err)
933 		goto ipv6_packet_fail;
934 
935 #ifdef CONFIG_SYSCTL
936 	err = ipv6_sysctl_register();
937 	if (err)
938 		goto sysctl_fail;
939 #endif
940 out:
941 	return err;
942 
943 #ifdef CONFIG_SYSCTL
944 sysctl_fail:
945 	ipv6_packet_cleanup();
946 #endif
947 ipv6_packet_fail:
948 	tcpv6_exit();
949 tcpv6_fail:
950 	udplitev6_exit();
951 udplitev6_fail:
952 	udpv6_exit();
953 udpv6_fail:
954 	ipv6_frag_exit();
955 ipv6_frag_fail:
956 	ipv6_exthdrs_exit();
957 ipv6_exthdrs_fail:
958 	addrconf_cleanup();
959 addrconf_fail:
960 	ip6_flowlabel_cleanup();
961 ip6_flowlabel_fail:
962 	ip6_route_cleanup();
963 ip6_route_fail:
964 #ifdef CONFIG_PROC_FS
965 	if6_proc_exit();
966 proc_if6_fail:
967 	ipv6_misc_proc_exit();
968 proc_misc6_fail:
969 	udplite6_proc_exit();
970 proc_udplite6_fail:
971 	raw6_proc_exit();
972 proc_raw6_fail:
973 #endif
974 	ipv6_netfilter_fini();
975 netfilter_fail:
976 	igmp6_cleanup();
977 igmp_fail:
978 	ndisc_cleanup();
979 ndisc_fail:
980 	ip6_mr_cleanup();
981 ipmr_fail:
982 	icmpv6_cleanup();
983 icmp_fail:
984 	unregister_pernet_subsys(&inet6_net_ops);
985 register_pernet_fail:
986 	sock_unregister(PF_INET6);
987 	rtnl_unregister_all(PF_INET6);
988 out_sock_register_fail:
989 	rawv6_exit();
990 out_unregister_raw_proto:
991 	proto_unregister(&rawv6_prot);
992 out_unregister_udplite_proto:
993 	proto_unregister(&udplitev6_prot);
994 out_unregister_udp_proto:
995 	proto_unregister(&udpv6_prot);
996 out_unregister_tcp_proto:
997 	proto_unregister(&tcpv6_prot);
998 	goto out;
999 }
1000 module_init(inet6_init);
1001 
1002 static void __exit inet6_exit(void)
1003 {
1004 	if (disable_ipv6_mod)
1005 		return;
1006 
1007 	/* First of all disallow new sockets creation. */
1008 	sock_unregister(PF_INET6);
1009 	/* Disallow any further netlink messages */
1010 	rtnl_unregister_all(PF_INET6);
1011 
1012 	udpv6_exit();
1013 	udplitev6_exit();
1014 	tcpv6_exit();
1015 
1016 	/* Cleanup code parts. */
1017 	ipv6_packet_cleanup();
1018 	ipv6_frag_exit();
1019 	ipv6_exthdrs_exit();
1020 	addrconf_cleanup();
1021 	ip6_flowlabel_cleanup();
1022 	ip6_route_cleanup();
1023 #ifdef CONFIG_PROC_FS
1024 
1025 	/* Cleanup code parts. */
1026 	if6_proc_exit();
1027 	ipv6_misc_proc_exit();
1028 	udplite6_proc_exit();
1029 	raw6_proc_exit();
1030 #endif
1031 	ipv6_netfilter_fini();
1032 	igmp6_cleanup();
1033 	ndisc_cleanup();
1034 	ip6_mr_cleanup();
1035 	icmpv6_cleanup();
1036 	rawv6_exit();
1037 
1038 	unregister_pernet_subsys(&inet6_net_ops);
1039 	proto_unregister(&rawv6_prot);
1040 	proto_unregister(&udplitev6_prot);
1041 	proto_unregister(&udpv6_prot);
1042 	proto_unregister(&tcpv6_prot);
1043 
1044 	rcu_barrier(); /* Wait for completion of call_rcu()'s */
1045 }
1046 module_exit(inet6_exit);
1047 
1048 MODULE_ALIAS_NETPROTO(PF_INET6);
1049