xref: /openbmc/linux/net/ipv6/af_inet6.c (revision afc98d90)
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/ping.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 #include <net/ndisc.h>
60 #ifdef CONFIG_IPV6_TUNNEL
61 #include <net/ip6_tunnel.h>
62 #endif
63 
64 #include <asm/uaccess.h>
65 #include <linux/mroute6.h>
66 
67 MODULE_AUTHOR("Cast of dozens");
68 MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
69 MODULE_LICENSE("GPL");
70 
71 /* The inetsw6 table contains everything that inet6_create needs to
72  * build a new socket.
73  */
74 static struct list_head inetsw6[SOCK_MAX];
75 static DEFINE_SPINLOCK(inetsw6_lock);
76 
77 struct ipv6_params ipv6_defaults = {
78 	.disable_ipv6 = 0,
79 	.autoconf = 1,
80 };
81 
82 static int disable_ipv6_mod;
83 
84 module_param_named(disable, disable_ipv6_mod, int, 0444);
85 MODULE_PARM_DESC(disable, "Disable IPv6 module such that it is non-functional");
86 
87 module_param_named(disable_ipv6, ipv6_defaults.disable_ipv6, int, 0444);
88 MODULE_PARM_DESC(disable_ipv6, "Disable IPv6 on all interfaces");
89 
90 module_param_named(autoconf, ipv6_defaults.autoconf, int, 0444);
91 MODULE_PARM_DESC(autoconf, "Enable IPv6 address autoconfiguration on all interfaces");
92 
93 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
94 {
95 	const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
96 
97 	return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
98 }
99 
100 static int inet6_create(struct net *net, struct socket *sock, int protocol,
101 			int kern)
102 {
103 	struct inet_sock *inet;
104 	struct ipv6_pinfo *np;
105 	struct sock *sk;
106 	struct inet_protosw *answer;
107 	struct proto *answer_prot;
108 	unsigned char answer_flags;
109 	char answer_no_check;
110 	int try_loading_module = 0;
111 	int err;
112 
113 	/* Look for the requested type/protocol pair. */
114 lookup_protocol:
115 	err = -ESOCKTNOSUPPORT;
116 	rcu_read_lock();
117 	list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
118 
119 		err = 0;
120 		/* Check the non-wild match. */
121 		if (protocol == answer->protocol) {
122 			if (protocol != IPPROTO_IP)
123 				break;
124 		} else {
125 			/* Check for the two wild cases. */
126 			if (IPPROTO_IP == protocol) {
127 				protocol = answer->protocol;
128 				break;
129 			}
130 			if (IPPROTO_IP == answer->protocol)
131 				break;
132 		}
133 		err = -EPROTONOSUPPORT;
134 	}
135 
136 	if (err) {
137 		if (try_loading_module < 2) {
138 			rcu_read_unlock();
139 			/*
140 			 * Be more specific, e.g. net-pf-10-proto-132-type-1
141 			 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
142 			 */
143 			if (++try_loading_module == 1)
144 				request_module("net-pf-%d-proto-%d-type-%d",
145 						PF_INET6, protocol, sock->type);
146 			/*
147 			 * Fall back to generic, e.g. net-pf-10-proto-132
148 			 * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
149 			 */
150 			else
151 				request_module("net-pf-%d-proto-%d",
152 						PF_INET6, protocol);
153 			goto lookup_protocol;
154 		} else
155 			goto out_rcu_unlock;
156 	}
157 
158 	err = -EPERM;
159 	if (sock->type == SOCK_RAW && !kern &&
160 	    !ns_capable(net->user_ns, CAP_NET_RAW))
161 		goto out_rcu_unlock;
162 
163 	sock->ops = answer->ops;
164 	answer_prot = answer->prot;
165 	answer_no_check = answer->no_check;
166 	answer_flags = answer->flags;
167 	rcu_read_unlock();
168 
169 	WARN_ON(answer_prot->slab == NULL);
170 
171 	err = -ENOBUFS;
172 	sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot);
173 	if (sk == NULL)
174 		goto out;
175 
176 	sock_init_data(sock, sk);
177 
178 	err = 0;
179 	sk->sk_no_check = answer_no_check;
180 	if (INET_PROTOSW_REUSE & answer_flags)
181 		sk->sk_reuse = SK_CAN_REUSE;
182 
183 	inet = inet_sk(sk);
184 	inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
185 
186 	if (SOCK_RAW == sock->type) {
187 		inet->inet_num = protocol;
188 		if (IPPROTO_RAW == protocol)
189 			inet->hdrincl = 1;
190 	}
191 
192 	sk->sk_destruct		= inet_sock_destruct;
193 	sk->sk_family		= PF_INET6;
194 	sk->sk_protocol		= protocol;
195 
196 	sk->sk_backlog_rcv	= answer->prot->backlog_rcv;
197 
198 	inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
199 	np->hop_limit	= -1;
200 	np->mcast_hops	= IPV6_DEFAULT_MCASTHOPS;
201 	np->mc_loop	= 1;
202 	np->pmtudisc	= IPV6_PMTUDISC_WANT;
203 	np->ipv6only	= net->ipv6.sysctl.bindv6only;
204 
205 	/* Init the ipv4 part of the socket since we can have sockets
206 	 * using v6 API for ipv4.
207 	 */
208 	inet->uc_ttl	= -1;
209 
210 	inet->mc_loop	= 1;
211 	inet->mc_ttl	= 1;
212 	inet->mc_index	= 0;
213 	inet->mc_list	= NULL;
214 	inet->rcv_tos	= 0;
215 
216 	if (net->ipv4.sysctl_ip_no_pmtu_disc)
217 		inet->pmtudisc = IP_PMTUDISC_DONT;
218 	else
219 		inet->pmtudisc = IP_PMTUDISC_WANT;
220 	/*
221 	 * Increment only the relevant sk_prot->socks debug field, this changes
222 	 * the previous behaviour of incrementing both the equivalent to
223 	 * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
224 	 *
225 	 * This allows better debug granularity as we'll know exactly how many
226 	 * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
227 	 * transport protocol socks. -acme
228 	 */
229 	sk_refcnt_debug_inc(sk);
230 
231 	if (inet->inet_num) {
232 		/* It assumes that any protocol which allows
233 		 * the user to assign a number at socket
234 		 * creation time automatically shares.
235 		 */
236 		inet->inet_sport = htons(inet->inet_num);
237 		sk->sk_prot->hash(sk);
238 	}
239 	if (sk->sk_prot->init) {
240 		err = sk->sk_prot->init(sk);
241 		if (err) {
242 			sk_common_release(sk);
243 			goto out;
244 		}
245 	}
246 out:
247 	return err;
248 out_rcu_unlock:
249 	rcu_read_unlock();
250 	goto out;
251 }
252 
253 
254 /* bind for INET6 API */
255 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
256 {
257 	struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr;
258 	struct sock *sk = sock->sk;
259 	struct inet_sock *inet = inet_sk(sk);
260 	struct ipv6_pinfo *np = inet6_sk(sk);
261 	struct net *net = sock_net(sk);
262 	__be32 v4addr = 0;
263 	unsigned short snum;
264 	int addr_type = 0;
265 	int err = 0;
266 
267 	/* If the socket has its own bind function then use it. */
268 	if (sk->sk_prot->bind)
269 		return sk->sk_prot->bind(sk, uaddr, addr_len);
270 
271 	if (addr_len < SIN6_LEN_RFC2133)
272 		return -EINVAL;
273 
274 	if (addr->sin6_family != AF_INET6)
275 		return -EAFNOSUPPORT;
276 
277 	addr_type = ipv6_addr_type(&addr->sin6_addr);
278 	if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
279 		return -EINVAL;
280 
281 	snum = ntohs(addr->sin6_port);
282 	if (snum && snum < PROT_SOCK && !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
283 		return -EACCES;
284 
285 	lock_sock(sk);
286 
287 	/* Check these errors (active socket, double bind). */
288 	if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
289 		err = -EINVAL;
290 		goto out;
291 	}
292 
293 	/* Check if the address belongs to the host. */
294 	if (addr_type == IPV6_ADDR_MAPPED) {
295 		int chk_addr_ret;
296 
297 		/* Binding to v4-mapped address on a v6-only socket
298 		 * makes no sense
299 		 */
300 		if (np->ipv6only) {
301 			err = -EINVAL;
302 			goto out;
303 		}
304 
305 		/* Reproduce AF_INET checks to make the bindings consistent */
306 		v4addr = addr->sin6_addr.s6_addr32[3];
307 		chk_addr_ret = inet_addr_type(net, v4addr);
308 		if (!sysctl_ip_nonlocal_bind &&
309 		    !(inet->freebind || inet->transparent) &&
310 		    v4addr != htonl(INADDR_ANY) &&
311 		    chk_addr_ret != RTN_LOCAL &&
312 		    chk_addr_ret != RTN_MULTICAST &&
313 		    chk_addr_ret != RTN_BROADCAST) {
314 			err = -EADDRNOTAVAIL;
315 			goto out;
316 		}
317 	} else {
318 		if (addr_type != IPV6_ADDR_ANY) {
319 			struct net_device *dev = NULL;
320 
321 			rcu_read_lock();
322 			if (__ipv6_addr_needs_scope_id(addr_type)) {
323 				if (addr_len >= sizeof(struct sockaddr_in6) &&
324 				    addr->sin6_scope_id) {
325 					/* Override any existing binding, if another one
326 					 * is supplied by user.
327 					 */
328 					sk->sk_bound_dev_if = addr->sin6_scope_id;
329 				}
330 
331 				/* Binding to link-local address requires an interface */
332 				if (!sk->sk_bound_dev_if) {
333 					err = -EINVAL;
334 					goto out_unlock;
335 				}
336 				dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
337 				if (!dev) {
338 					err = -ENODEV;
339 					goto out_unlock;
340 				}
341 			}
342 
343 			/* ipv4 addr of the socket is invalid.  Only the
344 			 * unspecified and mapped address have a v4 equivalent.
345 			 */
346 			v4addr = LOOPBACK4_IPV6;
347 			if (!(addr_type & IPV6_ADDR_MULTICAST))	{
348 				if (!(inet->freebind || inet->transparent) &&
349 				    !ipv6_chk_addr(net, &addr->sin6_addr,
350 						   dev, 0)) {
351 					err = -EADDRNOTAVAIL;
352 					goto out_unlock;
353 				}
354 			}
355 			rcu_read_unlock();
356 		}
357 	}
358 
359 	inet->inet_rcv_saddr = v4addr;
360 	inet->inet_saddr = v4addr;
361 
362 	sk->sk_v6_rcv_saddr = addr->sin6_addr;
363 
364 	if (!(addr_type & IPV6_ADDR_MULTICAST))
365 		np->saddr = addr->sin6_addr;
366 
367 	/* Make sure we are allowed to bind here. */
368 	if (sk->sk_prot->get_port(sk, snum)) {
369 		inet_reset_saddr(sk);
370 		err = -EADDRINUSE;
371 		goto out;
372 	}
373 
374 	if (addr_type != IPV6_ADDR_ANY) {
375 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
376 		if (addr_type != IPV6_ADDR_MAPPED)
377 			np->ipv6only = 1;
378 	}
379 	if (snum)
380 		sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
381 	inet->inet_sport = htons(inet->inet_num);
382 	inet->inet_dport = 0;
383 	inet->inet_daddr = 0;
384 out:
385 	release_sock(sk);
386 	return err;
387 out_unlock:
388 	rcu_read_unlock();
389 	goto out;
390 }
391 EXPORT_SYMBOL(inet6_bind);
392 
393 int inet6_release(struct socket *sock)
394 {
395 	struct sock *sk = sock->sk;
396 
397 	if (sk == NULL)
398 		return -EINVAL;
399 
400 	/* Free mc lists */
401 	ipv6_sock_mc_close(sk);
402 
403 	/* Free ac lists */
404 	ipv6_sock_ac_close(sk);
405 
406 	return inet_release(sock);
407 }
408 EXPORT_SYMBOL(inet6_release);
409 
410 void inet6_destroy_sock(struct sock *sk)
411 {
412 	struct ipv6_pinfo *np = inet6_sk(sk);
413 	struct sk_buff *skb;
414 	struct ipv6_txoptions *opt;
415 
416 	/* Release rx options */
417 
418 	skb = xchg(&np->pktoptions, NULL);
419 	if (skb != NULL)
420 		kfree_skb(skb);
421 
422 	skb = xchg(&np->rxpmtu, NULL);
423 	if (skb != NULL)
424 		kfree_skb(skb);
425 
426 	/* Free flowlabels */
427 	fl6_free_socklist(sk);
428 
429 	/* Free tx options */
430 
431 	opt = xchg(&np->opt, NULL);
432 	if (opt != NULL)
433 		sock_kfree_s(sk, opt, opt->tot_len);
434 }
435 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
436 
437 /*
438  *	This does both peername and sockname.
439  */
440 
441 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
442 		 int *uaddr_len, int peer)
443 {
444 	struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
445 	struct sock *sk = sock->sk;
446 	struct inet_sock *inet = inet_sk(sk);
447 	struct ipv6_pinfo *np = inet6_sk(sk);
448 
449 	sin->sin6_family = AF_INET6;
450 	sin->sin6_flowinfo = 0;
451 	sin->sin6_scope_id = 0;
452 	if (peer) {
453 		if (!inet->inet_dport)
454 			return -ENOTCONN;
455 		if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
456 		    peer == 1)
457 			return -ENOTCONN;
458 		sin->sin6_port = inet->inet_dport;
459 		sin->sin6_addr = sk->sk_v6_daddr;
460 		if (np->sndflow)
461 			sin->sin6_flowinfo = np->flow_label;
462 	} else {
463 		if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
464 			sin->sin6_addr = np->saddr;
465 		else
466 			sin->sin6_addr = sk->sk_v6_rcv_saddr;
467 
468 		sin->sin6_port = inet->inet_sport;
469 	}
470 	sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
471 						 sk->sk_bound_dev_if);
472 	*uaddr_len = sizeof(*sin);
473 	return 0;
474 }
475 EXPORT_SYMBOL(inet6_getname);
476 
477 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
478 {
479 	struct sock *sk = sock->sk;
480 	struct net *net = sock_net(sk);
481 
482 	switch (cmd) {
483 	case SIOCGSTAMP:
484 		return sock_get_timestamp(sk, (struct timeval __user *)arg);
485 
486 	case SIOCGSTAMPNS:
487 		return sock_get_timestampns(sk, (struct timespec __user *)arg);
488 
489 	case SIOCADDRT:
490 	case SIOCDELRT:
491 
492 		return ipv6_route_ioctl(net, cmd, (void __user *)arg);
493 
494 	case SIOCSIFADDR:
495 		return addrconf_add_ifaddr(net, (void __user *) arg);
496 	case SIOCDIFADDR:
497 		return addrconf_del_ifaddr(net, (void __user *) arg);
498 	case SIOCSIFDSTADDR:
499 		return addrconf_set_dstaddr(net, (void __user *) arg);
500 	default:
501 		if (!sk->sk_prot->ioctl)
502 			return -ENOIOCTLCMD;
503 		return sk->sk_prot->ioctl(sk, cmd, arg);
504 	}
505 	/*NOTREACHED*/
506 	return 0;
507 }
508 EXPORT_SYMBOL(inet6_ioctl);
509 
510 const struct proto_ops inet6_stream_ops = {
511 	.family		   = PF_INET6,
512 	.owner		   = THIS_MODULE,
513 	.release	   = inet6_release,
514 	.bind		   = inet6_bind,
515 	.connect	   = inet_stream_connect,	/* ok		*/
516 	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
517 	.accept		   = inet_accept,		/* ok		*/
518 	.getname	   = inet6_getname,
519 	.poll		   = tcp_poll,			/* ok		*/
520 	.ioctl		   = inet6_ioctl,		/* must change  */
521 	.listen		   = inet_listen,		/* ok		*/
522 	.shutdown	   = inet_shutdown,		/* ok		*/
523 	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
524 	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
525 	.sendmsg	   = inet_sendmsg,		/* ok		*/
526 	.recvmsg	   = inet_recvmsg,		/* ok		*/
527 	.mmap		   = sock_no_mmap,
528 	.sendpage	   = inet_sendpage,
529 	.splice_read	   = tcp_splice_read,
530 #ifdef CONFIG_COMPAT
531 	.compat_setsockopt = compat_sock_common_setsockopt,
532 	.compat_getsockopt = compat_sock_common_getsockopt,
533 #endif
534 };
535 
536 const struct proto_ops inet6_dgram_ops = {
537 	.family		   = PF_INET6,
538 	.owner		   = THIS_MODULE,
539 	.release	   = inet6_release,
540 	.bind		   = inet6_bind,
541 	.connect	   = inet_dgram_connect,	/* ok		*/
542 	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
543 	.accept		   = sock_no_accept,		/* a do nothing	*/
544 	.getname	   = inet6_getname,
545 	.poll		   = udp_poll,			/* ok		*/
546 	.ioctl		   = inet6_ioctl,		/* must change  */
547 	.listen		   = sock_no_listen,		/* ok		*/
548 	.shutdown	   = inet_shutdown,		/* ok		*/
549 	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
550 	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
551 	.sendmsg	   = inet_sendmsg,		/* ok		*/
552 	.recvmsg	   = inet_recvmsg,		/* ok		*/
553 	.mmap		   = sock_no_mmap,
554 	.sendpage	   = sock_no_sendpage,
555 #ifdef CONFIG_COMPAT
556 	.compat_setsockopt = compat_sock_common_setsockopt,
557 	.compat_getsockopt = compat_sock_common_getsockopt,
558 #endif
559 };
560 
561 static const struct net_proto_family inet6_family_ops = {
562 	.family = PF_INET6,
563 	.create = inet6_create,
564 	.owner	= THIS_MODULE,
565 };
566 
567 int inet6_register_protosw(struct inet_protosw *p)
568 {
569 	struct list_head *lh;
570 	struct inet_protosw *answer;
571 	struct list_head *last_perm;
572 	int protocol = p->protocol;
573 	int ret;
574 
575 	spin_lock_bh(&inetsw6_lock);
576 
577 	ret = -EINVAL;
578 	if (p->type >= SOCK_MAX)
579 		goto out_illegal;
580 
581 	/* If we are trying to override a permanent protocol, bail. */
582 	answer = NULL;
583 	ret = -EPERM;
584 	last_perm = &inetsw6[p->type];
585 	list_for_each(lh, &inetsw6[p->type]) {
586 		answer = list_entry(lh, struct inet_protosw, list);
587 
588 		/* Check only the non-wild match. */
589 		if (INET_PROTOSW_PERMANENT & answer->flags) {
590 			if (protocol == answer->protocol)
591 				break;
592 			last_perm = lh;
593 		}
594 
595 		answer = NULL;
596 	}
597 	if (answer)
598 		goto out_permanent;
599 
600 	/* Add the new entry after the last permanent entry if any, so that
601 	 * the new entry does not override a permanent entry when matched with
602 	 * a wild-card protocol. But it is allowed to override any existing
603 	 * non-permanent entry.  This means that when we remove this entry, the
604 	 * system automatically returns to the old behavior.
605 	 */
606 	list_add_rcu(&p->list, last_perm);
607 	ret = 0;
608 out:
609 	spin_unlock_bh(&inetsw6_lock);
610 	return ret;
611 
612 out_permanent:
613 	pr_err("Attempt to override permanent protocol %d\n", protocol);
614 	goto out;
615 
616 out_illegal:
617 	pr_err("Ignoring attempt to register invalid socket type %d\n",
618 	       p->type);
619 	goto out;
620 }
621 EXPORT_SYMBOL(inet6_register_protosw);
622 
623 void
624 inet6_unregister_protosw(struct inet_protosw *p)
625 {
626 	if (INET_PROTOSW_PERMANENT & p->flags) {
627 		pr_err("Attempt to unregister permanent protocol %d\n",
628 		       p->protocol);
629 	} else {
630 		spin_lock_bh(&inetsw6_lock);
631 		list_del_rcu(&p->list);
632 		spin_unlock_bh(&inetsw6_lock);
633 
634 		synchronize_net();
635 	}
636 }
637 EXPORT_SYMBOL(inet6_unregister_protosw);
638 
639 int inet6_sk_rebuild_header(struct sock *sk)
640 {
641 	struct ipv6_pinfo *np = inet6_sk(sk);
642 	struct dst_entry *dst;
643 
644 	dst = __sk_dst_check(sk, np->dst_cookie);
645 
646 	if (dst == NULL) {
647 		struct inet_sock *inet = inet_sk(sk);
648 		struct in6_addr *final_p, final;
649 		struct flowi6 fl6;
650 
651 		memset(&fl6, 0, sizeof(fl6));
652 		fl6.flowi6_proto = sk->sk_protocol;
653 		fl6.daddr = sk->sk_v6_daddr;
654 		fl6.saddr = np->saddr;
655 		fl6.flowlabel = np->flow_label;
656 		fl6.flowi6_oif = sk->sk_bound_dev_if;
657 		fl6.flowi6_mark = sk->sk_mark;
658 		fl6.fl6_dport = inet->inet_dport;
659 		fl6.fl6_sport = inet->inet_sport;
660 		security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
661 
662 		final_p = fl6_update_dst(&fl6, np->opt, &final);
663 
664 		dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
665 		if (IS_ERR(dst)) {
666 			sk->sk_route_caps = 0;
667 			sk->sk_err_soft = -PTR_ERR(dst);
668 			return PTR_ERR(dst);
669 		}
670 
671 		__ip6_dst_store(sk, dst, NULL, NULL);
672 	}
673 
674 	return 0;
675 }
676 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
677 
678 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb)
679 {
680 	const struct ipv6_pinfo *np = inet6_sk(sk);
681 	const struct inet6_skb_parm *opt = IP6CB(skb);
682 
683 	if (np->rxopt.all) {
684 		if ((opt->hop && (np->rxopt.bits.hopopts ||
685 				  np->rxopt.bits.ohopopts)) ||
686 		    (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) &&
687 		     np->rxopt.bits.rxflow) ||
688 		    (opt->srcrt && (np->rxopt.bits.srcrt ||
689 		     np->rxopt.bits.osrcrt)) ||
690 		    ((opt->dst1 || opt->dst0) &&
691 		     (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
692 			return true;
693 	}
694 	return false;
695 }
696 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
697 
698 static struct packet_type ipv6_packet_type __read_mostly = {
699 	.type = cpu_to_be16(ETH_P_IPV6),
700 	.func = ipv6_rcv,
701 };
702 
703 static int __init ipv6_packet_init(void)
704 {
705 	dev_add_pack(&ipv6_packet_type);
706 	return 0;
707 }
708 
709 static void ipv6_packet_cleanup(void)
710 {
711 	dev_remove_pack(&ipv6_packet_type);
712 }
713 
714 static int __net_init ipv6_init_mibs(struct net *net)
715 {
716 	int i;
717 
718 	if (snmp_mib_init((void __percpu **)net->mib.udp_stats_in6,
719 			  sizeof(struct udp_mib),
720 			  __alignof__(struct udp_mib)) < 0)
721 		return -ENOMEM;
722 	if (snmp_mib_init((void __percpu **)net->mib.udplite_stats_in6,
723 			  sizeof(struct udp_mib),
724 			  __alignof__(struct udp_mib)) < 0)
725 		goto err_udplite_mib;
726 	if (snmp_mib_init((void __percpu **)net->mib.ipv6_statistics,
727 			  sizeof(struct ipstats_mib),
728 			  __alignof__(struct ipstats_mib)) < 0)
729 		goto err_ip_mib;
730 
731 	for_each_possible_cpu(i) {
732 		struct ipstats_mib *af_inet6_stats;
733 		af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics[0], i);
734 		u64_stats_init(&af_inet6_stats->syncp);
735 #if SNMP_ARRAY_SZ == 2
736 		af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics[1], i);
737 		u64_stats_init(&af_inet6_stats->syncp);
738 #endif
739 	}
740 
741 
742 	if (snmp_mib_init((void __percpu **)net->mib.icmpv6_statistics,
743 			  sizeof(struct icmpv6_mib),
744 			  __alignof__(struct icmpv6_mib)) < 0)
745 		goto err_icmp_mib;
746 	net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
747 						GFP_KERNEL);
748 	if (!net->mib.icmpv6msg_statistics)
749 		goto err_icmpmsg_mib;
750 	return 0;
751 
752 err_icmpmsg_mib:
753 	snmp_mib_free((void __percpu **)net->mib.icmpv6_statistics);
754 err_icmp_mib:
755 	snmp_mib_free((void __percpu **)net->mib.ipv6_statistics);
756 err_ip_mib:
757 	snmp_mib_free((void __percpu **)net->mib.udplite_stats_in6);
758 err_udplite_mib:
759 	snmp_mib_free((void __percpu **)net->mib.udp_stats_in6);
760 	return -ENOMEM;
761 }
762 
763 static void ipv6_cleanup_mibs(struct net *net)
764 {
765 	snmp_mib_free((void __percpu **)net->mib.udp_stats_in6);
766 	snmp_mib_free((void __percpu **)net->mib.udplite_stats_in6);
767 	snmp_mib_free((void __percpu **)net->mib.ipv6_statistics);
768 	snmp_mib_free((void __percpu **)net->mib.icmpv6_statistics);
769 	kfree(net->mib.icmpv6msg_statistics);
770 }
771 
772 static int __net_init inet6_net_init(struct net *net)
773 {
774 	int err = 0;
775 
776 	net->ipv6.sysctl.bindv6only = 0;
777 	net->ipv6.sysctl.icmpv6_time = 1*HZ;
778 	net->ipv6.sysctl.flowlabel_consistency = 1;
779 	atomic_set(&net->ipv6.rt_genid, 0);
780 
781 	err = ipv6_init_mibs(net);
782 	if (err)
783 		return err;
784 #ifdef CONFIG_PROC_FS
785 	err = udp6_proc_init(net);
786 	if (err)
787 		goto out;
788 	err = tcp6_proc_init(net);
789 	if (err)
790 		goto proc_tcp6_fail;
791 	err = ac6_proc_init(net);
792 	if (err)
793 		goto proc_ac6_fail;
794 #endif
795 	return err;
796 
797 #ifdef CONFIG_PROC_FS
798 proc_ac6_fail:
799 	tcp6_proc_exit(net);
800 proc_tcp6_fail:
801 	udp6_proc_exit(net);
802 out:
803 	ipv6_cleanup_mibs(net);
804 	return err;
805 #endif
806 }
807 
808 static void __net_exit inet6_net_exit(struct net *net)
809 {
810 #ifdef CONFIG_PROC_FS
811 	udp6_proc_exit(net);
812 	tcp6_proc_exit(net);
813 	ac6_proc_exit(net);
814 #endif
815 	ipv6_cleanup_mibs(net);
816 }
817 
818 static struct pernet_operations inet6_net_ops = {
819 	.init = inet6_net_init,
820 	.exit = inet6_net_exit,
821 };
822 
823 static const struct ipv6_stub ipv6_stub_impl = {
824 	.ipv6_sock_mc_join = ipv6_sock_mc_join,
825 	.ipv6_sock_mc_drop = ipv6_sock_mc_drop,
826 	.ipv6_dst_lookup = ip6_dst_lookup,
827 	.udpv6_encap_enable = udpv6_encap_enable,
828 	.ndisc_send_na = ndisc_send_na,
829 	.nd_tbl	= &nd_tbl,
830 };
831 
832 static int __init inet6_init(void)
833 {
834 	struct list_head *r;
835 	int err = 0;
836 
837 	BUILD_BUG_ON(sizeof(struct inet6_skb_parm) > FIELD_SIZEOF(struct sk_buff, cb));
838 
839 	/* Register the socket-side information for inet6_create.  */
840 	for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
841 		INIT_LIST_HEAD(r);
842 
843 	if (disable_ipv6_mod) {
844 		pr_info("Loaded, but administratively disabled, reboot required to enable\n");
845 		goto out;
846 	}
847 
848 	err = proto_register(&tcpv6_prot, 1);
849 	if (err)
850 		goto out;
851 
852 	err = proto_register(&udpv6_prot, 1);
853 	if (err)
854 		goto out_unregister_tcp_proto;
855 
856 	err = proto_register(&udplitev6_prot, 1);
857 	if (err)
858 		goto out_unregister_udp_proto;
859 
860 	err = proto_register(&rawv6_prot, 1);
861 	if (err)
862 		goto out_unregister_udplite_proto;
863 
864 	err = proto_register(&pingv6_prot, 1);
865 	if (err)
866 		goto out_unregister_ping_proto;
867 
868 	/* We MUST register RAW sockets before we create the ICMP6,
869 	 * IGMP6, or NDISC control sockets.
870 	 */
871 	err = rawv6_init();
872 	if (err)
873 		goto out_unregister_raw_proto;
874 
875 	/* Register the family here so that the init calls below will
876 	 * be able to create sockets. (?? is this dangerous ??)
877 	 */
878 	err = sock_register(&inet6_family_ops);
879 	if (err)
880 		goto out_sock_register_fail;
881 
882 	/*
883 	 *	ipngwg API draft makes clear that the correct semantics
884 	 *	for TCP and UDP is to consider one TCP and UDP instance
885 	 *	in a host available by both INET and INET6 APIs and
886 	 *	able to communicate via both network protocols.
887 	 */
888 
889 	err = register_pernet_subsys(&inet6_net_ops);
890 	if (err)
891 		goto register_pernet_fail;
892 	err = icmpv6_init();
893 	if (err)
894 		goto icmp_fail;
895 	err = ip6_mr_init();
896 	if (err)
897 		goto ipmr_fail;
898 	err = ndisc_init();
899 	if (err)
900 		goto ndisc_fail;
901 	err = igmp6_init();
902 	if (err)
903 		goto igmp_fail;
904 
905 	ipv6_stub = &ipv6_stub_impl;
906 
907 	err = ipv6_netfilter_init();
908 	if (err)
909 		goto netfilter_fail;
910 	/* Create /proc/foo6 entries. */
911 #ifdef CONFIG_PROC_FS
912 	err = -ENOMEM;
913 	if (raw6_proc_init())
914 		goto proc_raw6_fail;
915 	if (udplite6_proc_init())
916 		goto proc_udplite6_fail;
917 	if (ipv6_misc_proc_init())
918 		goto proc_misc6_fail;
919 	if (if6_proc_init())
920 		goto proc_if6_fail;
921 #endif
922 	err = ip6_route_init();
923 	if (err)
924 		goto ip6_route_fail;
925 	err = ndisc_late_init();
926 	if (err)
927 		goto ndisc_late_fail;
928 	err = ip6_flowlabel_init();
929 	if (err)
930 		goto ip6_flowlabel_fail;
931 	err = addrconf_init();
932 	if (err)
933 		goto addrconf_fail;
934 
935 	/* Init v6 extension headers. */
936 	err = ipv6_exthdrs_init();
937 	if (err)
938 		goto ipv6_exthdrs_fail;
939 
940 	err = ipv6_frag_init();
941 	if (err)
942 		goto ipv6_frag_fail;
943 
944 	/* Init v6 transport protocols. */
945 	err = udpv6_init();
946 	if (err)
947 		goto udpv6_fail;
948 
949 	err = udplitev6_init();
950 	if (err)
951 		goto udplitev6_fail;
952 
953 	err = tcpv6_init();
954 	if (err)
955 		goto tcpv6_fail;
956 
957 	err = ipv6_packet_init();
958 	if (err)
959 		goto ipv6_packet_fail;
960 
961 	err = pingv6_init();
962 	if (err)
963 		goto pingv6_fail;
964 
965 #ifdef CONFIG_SYSCTL
966 	err = ipv6_sysctl_register();
967 	if (err)
968 		goto sysctl_fail;
969 #endif
970 out:
971 	return err;
972 
973 #ifdef CONFIG_SYSCTL
974 sysctl_fail:
975 	pingv6_exit();
976 #endif
977 pingv6_fail:
978 	ipv6_packet_cleanup();
979 ipv6_packet_fail:
980 	tcpv6_exit();
981 tcpv6_fail:
982 	udplitev6_exit();
983 udplitev6_fail:
984 	udpv6_exit();
985 udpv6_fail:
986 	ipv6_frag_exit();
987 ipv6_frag_fail:
988 	ipv6_exthdrs_exit();
989 ipv6_exthdrs_fail:
990 	addrconf_cleanup();
991 addrconf_fail:
992 	ip6_flowlabel_cleanup();
993 ip6_flowlabel_fail:
994 	ndisc_late_cleanup();
995 ndisc_late_fail:
996 	ip6_route_cleanup();
997 ip6_route_fail:
998 #ifdef CONFIG_PROC_FS
999 	if6_proc_exit();
1000 proc_if6_fail:
1001 	ipv6_misc_proc_exit();
1002 proc_misc6_fail:
1003 	udplite6_proc_exit();
1004 proc_udplite6_fail:
1005 	raw6_proc_exit();
1006 proc_raw6_fail:
1007 #endif
1008 	ipv6_netfilter_fini();
1009 netfilter_fail:
1010 	igmp6_cleanup();
1011 igmp_fail:
1012 	ndisc_cleanup();
1013 ndisc_fail:
1014 	ip6_mr_cleanup();
1015 ipmr_fail:
1016 	icmpv6_cleanup();
1017 icmp_fail:
1018 	unregister_pernet_subsys(&inet6_net_ops);
1019 register_pernet_fail:
1020 	sock_unregister(PF_INET6);
1021 	rtnl_unregister_all(PF_INET6);
1022 out_sock_register_fail:
1023 	rawv6_exit();
1024 out_unregister_ping_proto:
1025 	proto_unregister(&pingv6_prot);
1026 out_unregister_raw_proto:
1027 	proto_unregister(&rawv6_prot);
1028 out_unregister_udplite_proto:
1029 	proto_unregister(&udplitev6_prot);
1030 out_unregister_udp_proto:
1031 	proto_unregister(&udpv6_prot);
1032 out_unregister_tcp_proto:
1033 	proto_unregister(&tcpv6_prot);
1034 	goto out;
1035 }
1036 module_init(inet6_init);
1037 
1038 MODULE_ALIAS_NETPROTO(PF_INET6);
1039