xref: /linux/net/ipv6/af_inet6.c (revision b454cc6636d254fbf6049b73e9560aee76fb04a3)
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  *	$Id: af_inet6.c,v 1.66 2002/02/01 22:01:04 davem Exp $
11  *
12  * 	Fixes:
13  *	piggy, Karl Knutson	:	Socket protocol table
14  * 	Hideaki YOSHIFUJI	:	sin6_scope_id support
15  * 	Arnaldo Melo		: 	check proc_net_create return, cleanups
16  *
17  *	This program is free software; you can redistribute it and/or
18  *      modify it under the terms of the GNU General Public License
19  *      as published by the Free Software Foundation; either version
20  *      2 of the License, or (at your option) any later version.
21  */
22 
23 
24 #include <linux/module.h>
25 #include <linux/capability.h>
26 #include <linux/errno.h>
27 #include <linux/types.h>
28 #include <linux/socket.h>
29 #include <linux/in.h>
30 #include <linux/kernel.h>
31 #include <linux/sched.h>
32 #include <linux/timer.h>
33 #include <linux/string.h>
34 #include <linux/sockios.h>
35 #include <linux/net.h>
36 #include <linux/fcntl.h>
37 #include <linux/mm.h>
38 #include <linux/interrupt.h>
39 #include <linux/proc_fs.h>
40 #include <linux/stat.h>
41 #include <linux/init.h>
42 
43 #include <linux/inet.h>
44 #include <linux/netdevice.h>
45 #include <linux/icmpv6.h>
46 #include <linux/smp_lock.h>
47 #include <linux/netfilter_ipv6.h>
48 
49 #include <net/ip.h>
50 #include <net/ipv6.h>
51 #include <net/udp.h>
52 #include <net/udplite.h>
53 #include <net/tcp.h>
54 #include <net/ipip.h>
55 #include <net/protocol.h>
56 #include <net/inet_common.h>
57 #include <net/transp_v6.h>
58 #include <net/ip6_route.h>
59 #include <net/addrconf.h>
60 #ifdef CONFIG_IPV6_TUNNEL
61 #include <net/ip6_tunnel.h>
62 #endif
63 #ifdef CONFIG_IPV6_MIP6
64 #include <net/mip6.h>
65 #endif
66 
67 #include <asm/uaccess.h>
68 #include <asm/system.h>
69 
70 MODULE_AUTHOR("Cast of dozens");
71 MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
72 MODULE_LICENSE("GPL");
73 
74 int sysctl_ipv6_bindv6only __read_mostly;
75 
76 /* The inetsw table contains everything that inet_create needs to
77  * build a new socket.
78  */
79 static struct list_head inetsw6[SOCK_MAX];
80 static DEFINE_SPINLOCK(inetsw6_lock);
81 
82 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
83 {
84 	const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
85 
86 	return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
87 }
88 
89 static int inet6_create(struct socket *sock, int protocol)
90 {
91 	struct inet_sock *inet;
92 	struct ipv6_pinfo *np;
93 	struct sock *sk;
94 	struct list_head *p;
95 	struct inet_protosw *answer;
96 	struct proto *answer_prot;
97 	unsigned char answer_flags;
98 	char answer_no_check;
99 	int try_loading_module = 0;
100 	int err;
101 
102 	/* Look for the requested type/protocol pair. */
103 	answer = NULL;
104 lookup_protocol:
105 	err = -ESOCKTNOSUPPORT;
106 	rcu_read_lock();
107 	list_for_each_rcu(p, &inetsw6[sock->type]) {
108 		answer = list_entry(p, struct inet_protosw, list);
109 
110 		/* Check the non-wild match. */
111 		if (protocol == answer->protocol) {
112 			if (protocol != IPPROTO_IP)
113 				break;
114 		} else {
115 			/* Check for the two wild cases. */
116 			if (IPPROTO_IP == protocol) {
117 				protocol = answer->protocol;
118 				break;
119 			}
120 			if (IPPROTO_IP == answer->protocol)
121 				break;
122 		}
123 		err = -EPROTONOSUPPORT;
124 		answer = NULL;
125 	}
126 
127 	if (!answer) {
128 		if (try_loading_module < 2) {
129 			rcu_read_unlock();
130 			/*
131 			 * Be more specific, e.g. net-pf-10-proto-132-type-1
132 			 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
133 			 */
134 			if (++try_loading_module == 1)
135 				request_module("net-pf-%d-proto-%d-type-%d",
136 						PF_INET6, protocol, sock->type);
137 			/*
138 			 * Fall back to generic, e.g. net-pf-10-proto-132
139 			 * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
140 			 */
141 			else
142 				request_module("net-pf-%d-proto-%d",
143 						PF_INET6, protocol);
144 			goto lookup_protocol;
145 		} else
146 			goto out_rcu_unlock;
147 	}
148 
149 	err = -EPERM;
150 	if (answer->capability > 0 && !capable(answer->capability))
151 		goto out_rcu_unlock;
152 
153 	sock->ops = answer->ops;
154 	answer_prot = answer->prot;
155 	answer_no_check = answer->no_check;
156 	answer_flags = answer->flags;
157 	rcu_read_unlock();
158 
159 	BUG_TRAP(answer_prot->slab != NULL);
160 
161 	err = -ENOBUFS;
162 	sk = sk_alloc(PF_INET6, GFP_KERNEL, answer_prot, 1);
163 	if (sk == NULL)
164 		goto out;
165 
166 	sock_init_data(sock, sk);
167 
168 	err = 0;
169 	sk->sk_no_check = answer_no_check;
170 	if (INET_PROTOSW_REUSE & answer_flags)
171 		sk->sk_reuse = 1;
172 
173 	inet = inet_sk(sk);
174 	inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
175 
176 	if (SOCK_RAW == sock->type) {
177 		inet->num = protocol;
178 		if (IPPROTO_RAW == protocol)
179 			inet->hdrincl = 1;
180 	}
181 
182 	sk->sk_destruct		= inet_sock_destruct;
183 	sk->sk_family		= PF_INET6;
184 	sk->sk_protocol		= protocol;
185 
186 	sk->sk_backlog_rcv	= answer->prot->backlog_rcv;
187 
188 	inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
189 	np->hop_limit	= -1;
190 	np->mcast_hops	= -1;
191 	np->mc_loop	= 1;
192 	np->pmtudisc	= IPV6_PMTUDISC_WANT;
193 	np->ipv6only	= sysctl_ipv6_bindv6only;
194 
195 	/* Init the ipv4 part of the socket since we can have sockets
196 	 * using v6 API for ipv4.
197 	 */
198 	inet->uc_ttl	= -1;
199 
200 	inet->mc_loop	= 1;
201 	inet->mc_ttl	= 1;
202 	inet->mc_index	= 0;
203 	inet->mc_list	= NULL;
204 
205 	if (ipv4_config.no_pmtu_disc)
206 		inet->pmtudisc = IP_PMTUDISC_DONT;
207 	else
208 		inet->pmtudisc = IP_PMTUDISC_WANT;
209 	/*
210 	 * Increment only the relevant sk_prot->socks debug field, this changes
211 	 * the previous behaviour of incrementing both the equivalent to
212 	 * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
213 	 *
214 	 * This allows better debug granularity as we'll know exactly how many
215 	 * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
216 	 * transport protocol socks. -acme
217 	 */
218 	sk_refcnt_debug_inc(sk);
219 
220 	if (inet->num) {
221 		/* It assumes that any protocol which allows
222 		 * the user to assign a number at socket
223 		 * creation time automatically shares.
224 		 */
225 		inet->sport = htons(inet->num);
226 		sk->sk_prot->hash(sk);
227 	}
228 	if (sk->sk_prot->init) {
229 		err = sk->sk_prot->init(sk);
230 		if (err) {
231 			sk_common_release(sk);
232 			goto out;
233 		}
234 	}
235 out:
236 	return err;
237 out_rcu_unlock:
238 	rcu_read_unlock();
239 	goto out;
240 }
241 
242 
243 /* bind for INET6 API */
244 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
245 {
246 	struct sockaddr_in6 *addr=(struct sockaddr_in6 *)uaddr;
247 	struct sock *sk = sock->sk;
248 	struct inet_sock *inet = inet_sk(sk);
249 	struct ipv6_pinfo *np = inet6_sk(sk);
250 	__be32 v4addr = 0;
251 	unsigned short snum;
252 	int addr_type = 0;
253 	int err = 0;
254 
255 	/* If the socket has its own bind function then use it. */
256 	if (sk->sk_prot->bind)
257 		return sk->sk_prot->bind(sk, uaddr, addr_len);
258 
259 	if (addr_len < SIN6_LEN_RFC2133)
260 		return -EINVAL;
261 	addr_type = ipv6_addr_type(&addr->sin6_addr);
262 	if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
263 		return -EINVAL;
264 
265 	snum = ntohs(addr->sin6_port);
266 	if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
267 		return -EACCES;
268 
269 	lock_sock(sk);
270 
271 	/* Check these errors (active socket, double bind). */
272 	if (sk->sk_state != TCP_CLOSE || inet->num) {
273 		err = -EINVAL;
274 		goto out;
275 	}
276 
277 	/* Check if the address belongs to the host. */
278 	if (addr_type == IPV6_ADDR_MAPPED) {
279 		v4addr = addr->sin6_addr.s6_addr32[3];
280 		if (inet_addr_type(v4addr) != RTN_LOCAL) {
281 			err = -EADDRNOTAVAIL;
282 			goto out;
283 		}
284 	} else {
285 		if (addr_type != IPV6_ADDR_ANY) {
286 			struct net_device *dev = NULL;
287 
288 			if (addr_type & IPV6_ADDR_LINKLOCAL) {
289 				if (addr_len >= sizeof(struct sockaddr_in6) &&
290 				    addr->sin6_scope_id) {
291 					/* Override any existing binding, if another one
292 					 * is supplied by user.
293 					 */
294 					sk->sk_bound_dev_if = addr->sin6_scope_id;
295 				}
296 
297 				/* Binding to link-local address requires an interface */
298 				if (!sk->sk_bound_dev_if) {
299 					err = -EINVAL;
300 					goto out;
301 				}
302 				dev = dev_get_by_index(sk->sk_bound_dev_if);
303 				if (!dev) {
304 					err = -ENODEV;
305 					goto out;
306 				}
307 			}
308 
309 			/* ipv4 addr of the socket is invalid.  Only the
310 			 * unspecified and mapped address have a v4 equivalent.
311 			 */
312 			v4addr = LOOPBACK4_IPV6;
313 			if (!(addr_type & IPV6_ADDR_MULTICAST))	{
314 				if (!ipv6_chk_addr(&addr->sin6_addr, dev, 0)) {
315 					if (dev)
316 						dev_put(dev);
317 					err = -EADDRNOTAVAIL;
318 					goto out;
319 				}
320 			}
321 			if (dev)
322 				dev_put(dev);
323 		}
324 	}
325 
326 	inet->rcv_saddr = v4addr;
327 	inet->saddr = v4addr;
328 
329 	ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr);
330 
331 	if (!(addr_type & IPV6_ADDR_MULTICAST))
332 		ipv6_addr_copy(&np->saddr, &addr->sin6_addr);
333 
334 	/* Make sure we are allowed to bind here. */
335 	if (sk->sk_prot->get_port(sk, snum)) {
336 		inet_reset_saddr(sk);
337 		err = -EADDRINUSE;
338 		goto out;
339 	}
340 
341 	if (addr_type != IPV6_ADDR_ANY)
342 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
343 	if (snum)
344 		sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
345 	inet->sport = htons(inet->num);
346 	inet->dport = 0;
347 	inet->daddr = 0;
348 out:
349 	release_sock(sk);
350 	return err;
351 }
352 
353 int inet6_release(struct socket *sock)
354 {
355 	struct sock *sk = sock->sk;
356 
357 	if (sk == NULL)
358 		return -EINVAL;
359 
360 	/* Free mc lists */
361 	ipv6_sock_mc_close(sk);
362 
363 	/* Free ac lists */
364 	ipv6_sock_ac_close(sk);
365 
366 	return inet_release(sock);
367 }
368 
369 int inet6_destroy_sock(struct sock *sk)
370 {
371 	struct ipv6_pinfo *np = inet6_sk(sk);
372 	struct sk_buff *skb;
373 	struct ipv6_txoptions *opt;
374 
375 	/* Release rx options */
376 
377 	if ((skb = xchg(&np->pktoptions, NULL)) != NULL)
378 		kfree_skb(skb);
379 
380 	/* Free flowlabels */
381 	fl6_free_socklist(sk);
382 
383 	/* Free tx options */
384 
385 	if ((opt = xchg(&np->opt, NULL)) != NULL)
386 		sock_kfree_s(sk, opt, opt->tot_len);
387 
388 	return 0;
389 }
390 
391 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
392 
393 /*
394  *	This does both peername and sockname.
395  */
396 
397 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
398 		 int *uaddr_len, int peer)
399 {
400 	struct sockaddr_in6 *sin=(struct sockaddr_in6 *)uaddr;
401 	struct sock *sk = sock->sk;
402 	struct inet_sock *inet = inet_sk(sk);
403 	struct ipv6_pinfo *np = inet6_sk(sk);
404 
405 	sin->sin6_family = AF_INET6;
406 	sin->sin6_flowinfo = 0;
407 	sin->sin6_scope_id = 0;
408 	if (peer) {
409 		if (!inet->dport)
410 			return -ENOTCONN;
411 		if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
412 		    peer == 1)
413 			return -ENOTCONN;
414 		sin->sin6_port = inet->dport;
415 		ipv6_addr_copy(&sin->sin6_addr, &np->daddr);
416 		if (np->sndflow)
417 			sin->sin6_flowinfo = np->flow_label;
418 	} else {
419 		if (ipv6_addr_any(&np->rcv_saddr))
420 			ipv6_addr_copy(&sin->sin6_addr, &np->saddr);
421 		else
422 			ipv6_addr_copy(&sin->sin6_addr, &np->rcv_saddr);
423 
424 		sin->sin6_port = inet->sport;
425 	}
426 	if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL)
427 		sin->sin6_scope_id = sk->sk_bound_dev_if;
428 	*uaddr_len = sizeof(*sin);
429 	return(0);
430 }
431 
432 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
433 {
434 	struct sock *sk = sock->sk;
435 
436 	switch(cmd)
437 	{
438 	case SIOCGSTAMP:
439 		return sock_get_timestamp(sk, (struct timeval __user *)arg);
440 
441 	case SIOCADDRT:
442 	case SIOCDELRT:
443 
444 		return(ipv6_route_ioctl(cmd,(void __user *)arg));
445 
446 	case SIOCSIFADDR:
447 		return addrconf_add_ifaddr((void __user *) arg);
448 	case SIOCDIFADDR:
449 		return addrconf_del_ifaddr((void __user *) arg);
450 	case SIOCSIFDSTADDR:
451 		return addrconf_set_dstaddr((void __user *) arg);
452 	default:
453 		if (!sk->sk_prot->ioctl)
454 			return -ENOIOCTLCMD;
455 		return sk->sk_prot->ioctl(sk, cmd, arg);
456 	}
457 	/*NOTREACHED*/
458 	return(0);
459 }
460 
461 const struct proto_ops inet6_stream_ops = {
462 	.family		   = PF_INET6,
463 	.owner		   = THIS_MODULE,
464 	.release	   = inet6_release,
465 	.bind		   = inet6_bind,
466 	.connect	   = inet_stream_connect,	/* ok		*/
467 	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
468 	.accept		   = inet_accept,		/* ok		*/
469 	.getname	   = inet6_getname,
470 	.poll		   = tcp_poll,			/* ok		*/
471 	.ioctl		   = inet6_ioctl,		/* must change  */
472 	.listen		   = inet_listen,		/* ok		*/
473 	.shutdown	   = inet_shutdown,		/* ok		*/
474 	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
475 	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
476 	.sendmsg	   = inet_sendmsg,		/* ok		*/
477 	.recvmsg	   = sock_common_recvmsg,	/* ok		*/
478 	.mmap		   = sock_no_mmap,
479 	.sendpage	   = tcp_sendpage,
480 #ifdef CONFIG_COMPAT
481 	.compat_setsockopt = compat_sock_common_setsockopt,
482 	.compat_getsockopt = compat_sock_common_getsockopt,
483 #endif
484 };
485 
486 const struct proto_ops inet6_dgram_ops = {
487 	.family		   = PF_INET6,
488 	.owner		   = THIS_MODULE,
489 	.release	   = inet6_release,
490 	.bind		   = inet6_bind,
491 	.connect	   = inet_dgram_connect,	/* ok		*/
492 	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
493 	.accept		   = sock_no_accept,		/* a do nothing	*/
494 	.getname	   = inet6_getname,
495 	.poll		   = udp_poll,			/* ok		*/
496 	.ioctl		   = inet6_ioctl,		/* must change  */
497 	.listen		   = sock_no_listen,		/* ok		*/
498 	.shutdown	   = inet_shutdown,		/* ok		*/
499 	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
500 	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
501 	.sendmsg	   = inet_sendmsg,		/* ok		*/
502 	.recvmsg	   = sock_common_recvmsg,	/* ok		*/
503 	.mmap		   = sock_no_mmap,
504 	.sendpage	   = sock_no_sendpage,
505 #ifdef CONFIG_COMPAT
506 	.compat_setsockopt = compat_sock_common_setsockopt,
507 	.compat_getsockopt = compat_sock_common_getsockopt,
508 #endif
509 };
510 
511 static struct net_proto_family inet6_family_ops = {
512 	.family = PF_INET6,
513 	.create = inet6_create,
514 	.owner	= THIS_MODULE,
515 };
516 
517 /* Same as inet6_dgram_ops, sans udp_poll.  */
518 static const struct proto_ops inet6_sockraw_ops = {
519 	.family		   = PF_INET6,
520 	.owner		   = THIS_MODULE,
521 	.release	   = inet6_release,
522 	.bind		   = inet6_bind,
523 	.connect	   = inet_dgram_connect,	/* ok		*/
524 	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
525 	.accept		   = sock_no_accept,		/* a do nothing	*/
526 	.getname	   = inet6_getname,
527 	.poll		   = datagram_poll,		/* ok		*/
528 	.ioctl		   = inet6_ioctl,		/* must change  */
529 	.listen		   = sock_no_listen,		/* ok		*/
530 	.shutdown	   = inet_shutdown,		/* ok		*/
531 	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
532 	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
533 	.sendmsg	   = inet_sendmsg,		/* ok		*/
534 	.recvmsg	   = sock_common_recvmsg,	/* ok		*/
535 	.mmap		   = sock_no_mmap,
536 	.sendpage	   = sock_no_sendpage,
537 #ifdef CONFIG_COMPAT
538 	.compat_setsockopt = compat_sock_common_setsockopt,
539 	.compat_getsockopt = compat_sock_common_getsockopt,
540 #endif
541 };
542 
543 static struct inet_protosw rawv6_protosw = {
544 	.type		= SOCK_RAW,
545 	.protocol	= IPPROTO_IP,	/* wild card */
546 	.prot		= &rawv6_prot,
547 	.ops		= &inet6_sockraw_ops,
548 	.capability	= CAP_NET_RAW,
549 	.no_check	= UDP_CSUM_DEFAULT,
550 	.flags		= INET_PROTOSW_REUSE,
551 };
552 
553 void
554 inet6_register_protosw(struct inet_protosw *p)
555 {
556 	struct list_head *lh;
557 	struct inet_protosw *answer;
558 	int protocol = p->protocol;
559 	struct list_head *last_perm;
560 
561 	spin_lock_bh(&inetsw6_lock);
562 
563 	if (p->type >= SOCK_MAX)
564 		goto out_illegal;
565 
566 	/* If we are trying to override a permanent protocol, bail. */
567 	answer = NULL;
568 	last_perm = &inetsw6[p->type];
569 	list_for_each(lh, &inetsw6[p->type]) {
570 		answer = list_entry(lh, struct inet_protosw, list);
571 
572 		/* Check only the non-wild match. */
573 		if (INET_PROTOSW_PERMANENT & answer->flags) {
574 			if (protocol == answer->protocol)
575 				break;
576 			last_perm = lh;
577 		}
578 
579 		answer = NULL;
580 	}
581 	if (answer)
582 		goto out_permanent;
583 
584 	/* Add the new entry after the last permanent entry if any, so that
585 	 * the new entry does not override a permanent entry when matched with
586 	 * a wild-card protocol. But it is allowed to override any existing
587 	 * non-permanent entry.  This means that when we remove this entry, the
588 	 * system automatically returns to the old behavior.
589 	 */
590 	list_add_rcu(&p->list, last_perm);
591 out:
592 	spin_unlock_bh(&inetsw6_lock);
593 	return;
594 
595 out_permanent:
596 	printk(KERN_ERR "Attempt to override permanent protocol %d.\n",
597 	       protocol);
598 	goto out;
599 
600 out_illegal:
601 	printk(KERN_ERR
602 	       "Ignoring attempt to register invalid socket type %d.\n",
603 	       p->type);
604 	goto out;
605 }
606 
607 void
608 inet6_unregister_protosw(struct inet_protosw *p)
609 {
610 	if (INET_PROTOSW_PERMANENT & p->flags) {
611 		printk(KERN_ERR
612 		       "Attempt to unregister permanent protocol %d.\n",
613 		       p->protocol);
614 	} else {
615 		spin_lock_bh(&inetsw6_lock);
616 		list_del_rcu(&p->list);
617 		spin_unlock_bh(&inetsw6_lock);
618 
619 		synchronize_net();
620 	}
621 }
622 
623 int inet6_sk_rebuild_header(struct sock *sk)
624 {
625 	int err;
626 	struct dst_entry *dst;
627 	struct ipv6_pinfo *np = inet6_sk(sk);
628 
629 	dst = __sk_dst_check(sk, np->dst_cookie);
630 
631 	if (dst == NULL) {
632 		struct inet_sock *inet = inet_sk(sk);
633 		struct in6_addr *final_p = NULL, final;
634 		struct flowi fl;
635 
636 		memset(&fl, 0, sizeof(fl));
637 		fl.proto = sk->sk_protocol;
638 		ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
639 		ipv6_addr_copy(&fl.fl6_src, &np->saddr);
640 		fl.fl6_flowlabel = np->flow_label;
641 		fl.oif = sk->sk_bound_dev_if;
642 		fl.fl_ip_dport = inet->dport;
643 		fl.fl_ip_sport = inet->sport;
644 		security_sk_classify_flow(sk, &fl);
645 
646 		if (np->opt && np->opt->srcrt) {
647 			struct rt0_hdr *rt0 = (struct rt0_hdr *) np->opt->srcrt;
648 			ipv6_addr_copy(&final, &fl.fl6_dst);
649 			ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
650 			final_p = &final;
651 		}
652 
653 		err = ip6_dst_lookup(sk, &dst, &fl);
654 		if (err) {
655 			sk->sk_route_caps = 0;
656 			return err;
657 		}
658 		if (final_p)
659 			ipv6_addr_copy(&fl.fl6_dst, final_p);
660 
661 		if ((err = xfrm_lookup(&dst, &fl, sk, 0)) < 0) {
662 			sk->sk_err_soft = -err;
663 			return err;
664 		}
665 
666 		__ip6_dst_store(sk, dst, NULL, NULL);
667 	}
668 
669 	return 0;
670 }
671 
672 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
673 
674 int ipv6_opt_accepted(struct sock *sk, struct sk_buff *skb)
675 {
676 	struct ipv6_pinfo *np = inet6_sk(sk);
677 	struct inet6_skb_parm *opt = IP6CB(skb);
678 
679 	if (np->rxopt.all) {
680 		if ((opt->hop && (np->rxopt.bits.hopopts ||
681 				  np->rxopt.bits.ohopopts)) ||
682 		    ((IPV6_FLOWINFO_MASK & *(__be32*)skb->nh.raw) &&
683 		     np->rxopt.bits.rxflow) ||
684 		    (opt->srcrt && (np->rxopt.bits.srcrt ||
685 		     np->rxopt.bits.osrcrt)) ||
686 		    ((opt->dst1 || opt->dst0) &&
687 		     (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
688 			return 1;
689 	}
690 	return 0;
691 }
692 
693 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
694 
695 int
696 snmp6_mib_init(void *ptr[2], size_t mibsize, size_t mibalign)
697 {
698 	if (ptr == NULL)
699 		return -EINVAL;
700 
701 	ptr[0] = __alloc_percpu(mibsize);
702 	if (!ptr[0])
703 		goto err0;
704 
705 	ptr[1] = __alloc_percpu(mibsize);
706 	if (!ptr[1])
707 		goto err1;
708 
709 	return 0;
710 
711 err1:
712 	free_percpu(ptr[0]);
713 	ptr[0] = NULL;
714 err0:
715 	return -ENOMEM;
716 }
717 
718 void
719 snmp6_mib_free(void *ptr[2])
720 {
721 	if (ptr == NULL)
722 		return;
723 	free_percpu(ptr[0]);
724 	free_percpu(ptr[1]);
725 	ptr[0] = ptr[1] = NULL;
726 }
727 
728 static int __init init_ipv6_mibs(void)
729 {
730 	if (snmp6_mib_init((void **)ipv6_statistics, sizeof (struct ipstats_mib),
731 			   __alignof__(struct ipstats_mib)) < 0)
732 		goto err_ip_mib;
733 	if (snmp6_mib_init((void **)icmpv6_statistics, sizeof (struct icmpv6_mib),
734 			   __alignof__(struct icmpv6_mib)) < 0)
735 		goto err_icmp_mib;
736 	if (snmp6_mib_init((void **)udp_stats_in6, sizeof (struct udp_mib),
737 			   __alignof__(struct udp_mib)) < 0)
738 		goto err_udp_mib;
739 	if (snmp6_mib_init((void **)udplite_stats_in6, sizeof (struct udp_mib),
740 			   __alignof__(struct udp_mib)) < 0)
741 		goto err_udplite_mib;
742 	return 0;
743 
744 err_udplite_mib:
745 	snmp6_mib_free((void **)udp_stats_in6);
746 err_udp_mib:
747 	snmp6_mib_free((void **)icmpv6_statistics);
748 err_icmp_mib:
749 	snmp6_mib_free((void **)ipv6_statistics);
750 err_ip_mib:
751 	return -ENOMEM;
752 
753 }
754 
755 static void cleanup_ipv6_mibs(void)
756 {
757 	snmp6_mib_free((void **)ipv6_statistics);
758 	snmp6_mib_free((void **)icmpv6_statistics);
759 	snmp6_mib_free((void **)udp_stats_in6);
760 	snmp6_mib_free((void **)udplite_stats_in6);
761 }
762 
763 static int __init inet6_init(void)
764 {
765 	struct sk_buff *dummy_skb;
766         struct list_head *r;
767 	int err;
768 
769 	BUILD_BUG_ON(sizeof(struct inet6_skb_parm) > sizeof(dummy_skb->cb));
770 
771 #ifdef MODULE
772 #if 0 /* FIXME --RR */
773 	if (!mod_member_present(&__this_module, can_unload))
774 	  return -EINVAL;
775 
776 	__this_module.can_unload = &ipv6_unload;
777 #endif
778 #endif
779 
780 	err = proto_register(&tcpv6_prot, 1);
781 	if (err)
782 		goto out;
783 
784 	err = proto_register(&udpv6_prot, 1);
785 	if (err)
786 		goto out_unregister_tcp_proto;
787 
788 	err = proto_register(&udplitev6_prot, 1);
789 	if (err)
790 		goto out_unregister_udp_proto;
791 
792 	err = proto_register(&rawv6_prot, 1);
793 	if (err)
794 		goto out_unregister_udplite_proto;
795 
796 
797 	/* Register the socket-side information for inet6_create.  */
798 	for(r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
799 		INIT_LIST_HEAD(r);
800 
801 	/* We MUST register RAW sockets before we create the ICMP6,
802 	 * IGMP6, or NDISC control sockets.
803 	 */
804 	inet6_register_protosw(&rawv6_protosw);
805 
806 	/* Register the family here so that the init calls below will
807 	 * be able to create sockets. (?? is this dangerous ??)
808 	 */
809 	err = sock_register(&inet6_family_ops);
810 	if (err)
811 		goto out_unregister_raw_proto;
812 
813 	/* Initialise ipv6 mibs */
814 	err = init_ipv6_mibs();
815 	if (err)
816 		goto out_unregister_sock;
817 
818 	/*
819 	 *	ipngwg API draft makes clear that the correct semantics
820 	 *	for TCP and UDP is to consider one TCP and UDP instance
821 	 *	in a host availiable by both INET and INET6 APIs and
822 	 *	able to communicate via both network protocols.
823 	 */
824 
825 #ifdef CONFIG_SYSCTL
826 	ipv6_sysctl_register();
827 #endif
828 	err = icmpv6_init(&inet6_family_ops);
829 	if (err)
830 		goto icmp_fail;
831 	err = ndisc_init(&inet6_family_ops);
832 	if (err)
833 		goto ndisc_fail;
834 	err = igmp6_init(&inet6_family_ops);
835 	if (err)
836 		goto igmp_fail;
837 	err = ipv6_netfilter_init();
838 	if (err)
839 		goto netfilter_fail;
840 	/* Create /proc/foo6 entries. */
841 #ifdef CONFIG_PROC_FS
842 	err = -ENOMEM;
843 	if (raw6_proc_init())
844 		goto proc_raw6_fail;
845 	if (tcp6_proc_init())
846 		goto proc_tcp6_fail;
847 	if (udp6_proc_init())
848 		goto proc_udp6_fail;
849 	if (udplite6_proc_init())
850 		goto proc_udplite6_fail;
851 	if (ipv6_misc_proc_init())
852 		goto proc_misc6_fail;
853 
854 	if (ac6_proc_init())
855 		goto proc_anycast6_fail;
856 	if (if6_proc_init())
857 		goto proc_if6_fail;
858 #endif
859 	ip6_route_init();
860 	ip6_flowlabel_init();
861 	err = addrconf_init();
862 	if (err)
863 		goto addrconf_fail;
864 
865 	/* Init v6 extension headers. */
866 	ipv6_rthdr_init();
867 	ipv6_frag_init();
868 	ipv6_nodata_init();
869 	ipv6_destopt_init();
870 #ifdef CONFIG_IPV6_MIP6
871 	mip6_init();
872 #endif
873 
874 	/* Init v6 transport protocols. */
875 	udpv6_init();
876 	udplitev6_init();
877 	tcpv6_init();
878 
879 	ipv6_packet_init();
880 	err = 0;
881 out:
882 	return err;
883 
884 addrconf_fail:
885 	ip6_flowlabel_cleanup();
886 	ip6_route_cleanup();
887 #ifdef CONFIG_PROC_FS
888 	if6_proc_exit();
889 proc_if6_fail:
890 	ac6_proc_exit();
891 proc_anycast6_fail:
892 	ipv6_misc_proc_exit();
893 proc_misc6_fail:
894 	udplite6_proc_exit();
895 proc_udplite6_fail:
896 	udp6_proc_exit();
897 proc_udp6_fail:
898 	tcp6_proc_exit();
899 proc_tcp6_fail:
900 	raw6_proc_exit();
901 proc_raw6_fail:
902 #endif
903 	ipv6_netfilter_fini();
904 netfilter_fail:
905 	igmp6_cleanup();
906 igmp_fail:
907 	ndisc_cleanup();
908 ndisc_fail:
909 	icmpv6_cleanup();
910 icmp_fail:
911 #ifdef CONFIG_SYSCTL
912 	ipv6_sysctl_unregister();
913 #endif
914 	cleanup_ipv6_mibs();
915 out_unregister_sock:
916 	sock_unregister(PF_INET6);
917 out_unregister_raw_proto:
918 	proto_unregister(&rawv6_prot);
919 out_unregister_udplite_proto:
920 	proto_unregister(&udplitev6_prot);
921 out_unregister_udp_proto:
922 	proto_unregister(&udpv6_prot);
923 out_unregister_tcp_proto:
924 	proto_unregister(&tcpv6_prot);
925 	goto out;
926 }
927 module_init(inet6_init);
928 
929 static void __exit inet6_exit(void)
930 {
931 	/* First of all disallow new sockets creation. */
932 	sock_unregister(PF_INET6);
933 #ifdef CONFIG_PROC_FS
934 	if6_proc_exit();
935 	ac6_proc_exit();
936  	ipv6_misc_proc_exit();
937  	udp6_proc_exit();
938  	udplite6_proc_exit();
939  	tcp6_proc_exit();
940  	raw6_proc_exit();
941 #endif
942 #ifdef CONFIG_IPV6_MIP6
943 	mip6_fini();
944 #endif
945 	/* Cleanup code parts. */
946 	ip6_flowlabel_cleanup();
947 	addrconf_cleanup();
948 	ip6_route_cleanup();
949 	ipv6_packet_cleanup();
950 	igmp6_cleanup();
951 	ipv6_netfilter_fini();
952 	ndisc_cleanup();
953 	icmpv6_cleanup();
954 #ifdef CONFIG_SYSCTL
955 	ipv6_sysctl_unregister();
956 #endif
957 	cleanup_ipv6_mibs();
958 	proto_unregister(&rawv6_prot);
959 	proto_unregister(&udpv6_prot);
960 	proto_unregister(&tcpv6_prot);
961 }
962 module_exit(inet6_exit);
963 
964 MODULE_ALIAS_NETPROTO(PF_INET6);
965