xref: /openbmc/linux/net/ipv6/af_inet6.c (revision f7777dcc)
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 	if (sock->type != SOCK_RAW &&
114 	    sock->type != SOCK_DGRAM &&
115 	    !inet_ehash_secret)
116 		build_ehash_secret();
117 
118 	/* Look for the requested type/protocol pair. */
119 lookup_protocol:
120 	err = -ESOCKTNOSUPPORT;
121 	rcu_read_lock();
122 	list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
123 
124 		err = 0;
125 		/* Check the non-wild match. */
126 		if (protocol == answer->protocol) {
127 			if (protocol != IPPROTO_IP)
128 				break;
129 		} else {
130 			/* Check for the two wild cases. */
131 			if (IPPROTO_IP == protocol) {
132 				protocol = answer->protocol;
133 				break;
134 			}
135 			if (IPPROTO_IP == answer->protocol)
136 				break;
137 		}
138 		err = -EPROTONOSUPPORT;
139 	}
140 
141 	if (err) {
142 		if (try_loading_module < 2) {
143 			rcu_read_unlock();
144 			/*
145 			 * Be more specific, e.g. net-pf-10-proto-132-type-1
146 			 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
147 			 */
148 			if (++try_loading_module == 1)
149 				request_module("net-pf-%d-proto-%d-type-%d",
150 						PF_INET6, protocol, sock->type);
151 			/*
152 			 * Fall back to generic, e.g. net-pf-10-proto-132
153 			 * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
154 			 */
155 			else
156 				request_module("net-pf-%d-proto-%d",
157 						PF_INET6, protocol);
158 			goto lookup_protocol;
159 		} else
160 			goto out_rcu_unlock;
161 	}
162 
163 	err = -EPERM;
164 	if (sock->type == SOCK_RAW && !kern &&
165 	    !ns_capable(net->user_ns, CAP_NET_RAW))
166 		goto out_rcu_unlock;
167 
168 	sock->ops = answer->ops;
169 	answer_prot = answer->prot;
170 	answer_no_check = answer->no_check;
171 	answer_flags = answer->flags;
172 	rcu_read_unlock();
173 
174 	WARN_ON(answer_prot->slab == NULL);
175 
176 	err = -ENOBUFS;
177 	sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot);
178 	if (sk == NULL)
179 		goto out;
180 
181 	sock_init_data(sock, sk);
182 
183 	err = 0;
184 	sk->sk_no_check = answer_no_check;
185 	if (INET_PROTOSW_REUSE & answer_flags)
186 		sk->sk_reuse = SK_CAN_REUSE;
187 
188 	inet = inet_sk(sk);
189 	inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
190 
191 	if (SOCK_RAW == sock->type) {
192 		inet->inet_num = protocol;
193 		if (IPPROTO_RAW == protocol)
194 			inet->hdrincl = 1;
195 	}
196 
197 	sk->sk_destruct		= inet_sock_destruct;
198 	sk->sk_family		= PF_INET6;
199 	sk->sk_protocol		= protocol;
200 
201 	sk->sk_backlog_rcv	= answer->prot->backlog_rcv;
202 
203 	inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
204 	np->hop_limit	= -1;
205 	np->mcast_hops	= IPV6_DEFAULT_MCASTHOPS;
206 	np->mc_loop	= 1;
207 	np->pmtudisc	= IPV6_PMTUDISC_WANT;
208 	np->ipv6only	= net->ipv6.sysctl.bindv6only;
209 
210 	/* Init the ipv4 part of the socket since we can have sockets
211 	 * using v6 API for ipv4.
212 	 */
213 	inet->uc_ttl	= -1;
214 
215 	inet->mc_loop	= 1;
216 	inet->mc_ttl	= 1;
217 	inet->mc_index	= 0;
218 	inet->mc_list	= NULL;
219 	inet->rcv_tos	= 0;
220 
221 	if (ipv4_config.no_pmtu_disc)
222 		inet->pmtudisc = IP_PMTUDISC_DONT;
223 	else
224 		inet->pmtudisc = IP_PMTUDISC_WANT;
225 	/*
226 	 * Increment only the relevant sk_prot->socks debug field, this changes
227 	 * the previous behaviour of incrementing both the equivalent to
228 	 * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
229 	 *
230 	 * This allows better debug granularity as we'll know exactly how many
231 	 * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
232 	 * transport protocol socks. -acme
233 	 */
234 	sk_refcnt_debug_inc(sk);
235 
236 	if (inet->inet_num) {
237 		/* It assumes that any protocol which allows
238 		 * the user to assign a number at socket
239 		 * creation time automatically shares.
240 		 */
241 		inet->inet_sport = htons(inet->inet_num);
242 		sk->sk_prot->hash(sk);
243 	}
244 	if (sk->sk_prot->init) {
245 		err = sk->sk_prot->init(sk);
246 		if (err) {
247 			sk_common_release(sk);
248 			goto out;
249 		}
250 	}
251 out:
252 	return err;
253 out_rcu_unlock:
254 	rcu_read_unlock();
255 	goto out;
256 }
257 
258 
259 /* bind for INET6 API */
260 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
261 {
262 	struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr;
263 	struct sock *sk = sock->sk;
264 	struct inet_sock *inet = inet_sk(sk);
265 	struct ipv6_pinfo *np = inet6_sk(sk);
266 	struct net *net = sock_net(sk);
267 	__be32 v4addr = 0;
268 	unsigned short snum;
269 	int addr_type = 0;
270 	int err = 0;
271 
272 	/* If the socket has its own bind function then use it. */
273 	if (sk->sk_prot->bind)
274 		return sk->sk_prot->bind(sk, uaddr, addr_len);
275 
276 	if (addr_len < SIN6_LEN_RFC2133)
277 		return -EINVAL;
278 
279 	if (addr->sin6_family != AF_INET6)
280 		return -EAFNOSUPPORT;
281 
282 	addr_type = ipv6_addr_type(&addr->sin6_addr);
283 	if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
284 		return -EINVAL;
285 
286 	snum = ntohs(addr->sin6_port);
287 	if (snum && snum < PROT_SOCK && !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
288 		return -EACCES;
289 
290 	lock_sock(sk);
291 
292 	/* Check these errors (active socket, double bind). */
293 	if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
294 		err = -EINVAL;
295 		goto out;
296 	}
297 
298 	/* Check if the address belongs to the host. */
299 	if (addr_type == IPV6_ADDR_MAPPED) {
300 		int chk_addr_ret;
301 
302 		/* Binding to v4-mapped address on a v6-only socket
303 		 * makes no sense
304 		 */
305 		if (np->ipv6only) {
306 			err = -EINVAL;
307 			goto out;
308 		}
309 
310 		/* Reproduce AF_INET checks to make the bindings consistent */
311 		v4addr = addr->sin6_addr.s6_addr32[3];
312 		chk_addr_ret = inet_addr_type(net, v4addr);
313 		if (!sysctl_ip_nonlocal_bind &&
314 		    !(inet->freebind || inet->transparent) &&
315 		    v4addr != htonl(INADDR_ANY) &&
316 		    chk_addr_ret != RTN_LOCAL &&
317 		    chk_addr_ret != RTN_MULTICAST &&
318 		    chk_addr_ret != RTN_BROADCAST) {
319 			err = -EADDRNOTAVAIL;
320 			goto out;
321 		}
322 	} else {
323 		if (addr_type != IPV6_ADDR_ANY) {
324 			struct net_device *dev = NULL;
325 
326 			rcu_read_lock();
327 			if (__ipv6_addr_needs_scope_id(addr_type)) {
328 				if (addr_len >= sizeof(struct sockaddr_in6) &&
329 				    addr->sin6_scope_id) {
330 					/* Override any existing binding, if another one
331 					 * is supplied by user.
332 					 */
333 					sk->sk_bound_dev_if = addr->sin6_scope_id;
334 				}
335 
336 				/* Binding to link-local address requires an interface */
337 				if (!sk->sk_bound_dev_if) {
338 					err = -EINVAL;
339 					goto out_unlock;
340 				}
341 				dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
342 				if (!dev) {
343 					err = -ENODEV;
344 					goto out_unlock;
345 				}
346 			}
347 
348 			/* ipv4 addr of the socket is invalid.  Only the
349 			 * unspecified and mapped address have a v4 equivalent.
350 			 */
351 			v4addr = LOOPBACK4_IPV6;
352 			if (!(addr_type & IPV6_ADDR_MULTICAST))	{
353 				if (!(inet->freebind || inet->transparent) &&
354 				    !ipv6_chk_addr(net, &addr->sin6_addr,
355 						   dev, 0)) {
356 					err = -EADDRNOTAVAIL;
357 					goto out_unlock;
358 				}
359 			}
360 			rcu_read_unlock();
361 		}
362 	}
363 
364 	inet->inet_rcv_saddr = v4addr;
365 	inet->inet_saddr = v4addr;
366 
367 	np->rcv_saddr = addr->sin6_addr;
368 
369 	if (!(addr_type & IPV6_ADDR_MULTICAST))
370 		np->saddr = addr->sin6_addr;
371 
372 	/* Make sure we are allowed to bind here. */
373 	if (sk->sk_prot->get_port(sk, snum)) {
374 		inet_reset_saddr(sk);
375 		err = -EADDRINUSE;
376 		goto out;
377 	}
378 
379 	if (addr_type != IPV6_ADDR_ANY) {
380 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
381 		if (addr_type != IPV6_ADDR_MAPPED)
382 			np->ipv6only = 1;
383 	}
384 	if (snum)
385 		sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
386 	inet->inet_sport = htons(inet->inet_num);
387 	inet->inet_dport = 0;
388 	inet->inet_daddr = 0;
389 out:
390 	release_sock(sk);
391 	return err;
392 out_unlock:
393 	rcu_read_unlock();
394 	goto out;
395 }
396 EXPORT_SYMBOL(inet6_bind);
397 
398 int inet6_release(struct socket *sock)
399 {
400 	struct sock *sk = sock->sk;
401 
402 	if (sk == NULL)
403 		return -EINVAL;
404 
405 	/* Free mc lists */
406 	ipv6_sock_mc_close(sk);
407 
408 	/* Free ac lists */
409 	ipv6_sock_ac_close(sk);
410 
411 	return inet_release(sock);
412 }
413 EXPORT_SYMBOL(inet6_release);
414 
415 void inet6_destroy_sock(struct sock *sk)
416 {
417 	struct ipv6_pinfo *np = inet6_sk(sk);
418 	struct sk_buff *skb;
419 	struct ipv6_txoptions *opt;
420 
421 	/* Release rx options */
422 
423 	skb = xchg(&np->pktoptions, NULL);
424 	if (skb != NULL)
425 		kfree_skb(skb);
426 
427 	skb = xchg(&np->rxpmtu, NULL);
428 	if (skb != NULL)
429 		kfree_skb(skb);
430 
431 	/* Free flowlabels */
432 	fl6_free_socklist(sk);
433 
434 	/* Free tx options */
435 
436 	opt = xchg(&np->opt, NULL);
437 	if (opt != NULL)
438 		sock_kfree_s(sk, opt, opt->tot_len);
439 }
440 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
441 
442 /*
443  *	This does both peername and sockname.
444  */
445 
446 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
447 		 int *uaddr_len, int peer)
448 {
449 	struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
450 	struct sock *sk = sock->sk;
451 	struct inet_sock *inet = inet_sk(sk);
452 	struct ipv6_pinfo *np = inet6_sk(sk);
453 
454 	sin->sin6_family = AF_INET6;
455 	sin->sin6_flowinfo = 0;
456 	sin->sin6_scope_id = 0;
457 	if (peer) {
458 		if (!inet->inet_dport)
459 			return -ENOTCONN;
460 		if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
461 		    peer == 1)
462 			return -ENOTCONN;
463 		sin->sin6_port = inet->inet_dport;
464 		sin->sin6_addr = np->daddr;
465 		if (np->sndflow)
466 			sin->sin6_flowinfo = np->flow_label;
467 	} else {
468 		if (ipv6_addr_any(&np->rcv_saddr))
469 			sin->sin6_addr = np->saddr;
470 		else
471 			sin->sin6_addr = np->rcv_saddr;
472 
473 		sin->sin6_port = inet->inet_sport;
474 	}
475 	sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
476 						 sk->sk_bound_dev_if);
477 	*uaddr_len = sizeof(*sin);
478 	return 0;
479 }
480 EXPORT_SYMBOL(inet6_getname);
481 
482 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
483 {
484 	struct sock *sk = sock->sk;
485 	struct net *net = sock_net(sk);
486 
487 	switch (cmd) {
488 	case SIOCGSTAMP:
489 		return sock_get_timestamp(sk, (struct timeval __user *)arg);
490 
491 	case SIOCGSTAMPNS:
492 		return sock_get_timestampns(sk, (struct timespec __user *)arg);
493 
494 	case SIOCADDRT:
495 	case SIOCDELRT:
496 
497 		return ipv6_route_ioctl(net, cmd, (void __user *)arg);
498 
499 	case SIOCSIFADDR:
500 		return addrconf_add_ifaddr(net, (void __user *) arg);
501 	case SIOCDIFADDR:
502 		return addrconf_del_ifaddr(net, (void __user *) arg);
503 	case SIOCSIFDSTADDR:
504 		return addrconf_set_dstaddr(net, (void __user *) arg);
505 	default:
506 		if (!sk->sk_prot->ioctl)
507 			return -ENOIOCTLCMD;
508 		return sk->sk_prot->ioctl(sk, cmd, arg);
509 	}
510 	/*NOTREACHED*/
511 	return 0;
512 }
513 EXPORT_SYMBOL(inet6_ioctl);
514 
515 const struct proto_ops inet6_stream_ops = {
516 	.family		   = PF_INET6,
517 	.owner		   = THIS_MODULE,
518 	.release	   = inet6_release,
519 	.bind		   = inet6_bind,
520 	.connect	   = inet_stream_connect,	/* ok		*/
521 	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
522 	.accept		   = inet_accept,		/* ok		*/
523 	.getname	   = inet6_getname,
524 	.poll		   = tcp_poll,			/* ok		*/
525 	.ioctl		   = inet6_ioctl,		/* must change  */
526 	.listen		   = inet_listen,		/* ok		*/
527 	.shutdown	   = inet_shutdown,		/* ok		*/
528 	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
529 	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
530 	.sendmsg	   = inet_sendmsg,		/* ok		*/
531 	.recvmsg	   = inet_recvmsg,		/* ok		*/
532 	.mmap		   = sock_no_mmap,
533 	.sendpage	   = inet_sendpage,
534 	.splice_read	   = tcp_splice_read,
535 #ifdef CONFIG_COMPAT
536 	.compat_setsockopt = compat_sock_common_setsockopt,
537 	.compat_getsockopt = compat_sock_common_getsockopt,
538 #endif
539 };
540 
541 const struct proto_ops inet6_dgram_ops = {
542 	.family		   = PF_INET6,
543 	.owner		   = THIS_MODULE,
544 	.release	   = inet6_release,
545 	.bind		   = inet6_bind,
546 	.connect	   = inet_dgram_connect,	/* ok		*/
547 	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
548 	.accept		   = sock_no_accept,		/* a do nothing	*/
549 	.getname	   = inet6_getname,
550 	.poll		   = udp_poll,			/* ok		*/
551 	.ioctl		   = inet6_ioctl,		/* must change  */
552 	.listen		   = sock_no_listen,		/* ok		*/
553 	.shutdown	   = inet_shutdown,		/* ok		*/
554 	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
555 	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
556 	.sendmsg	   = inet_sendmsg,		/* ok		*/
557 	.recvmsg	   = inet_recvmsg,		/* ok		*/
558 	.mmap		   = sock_no_mmap,
559 	.sendpage	   = sock_no_sendpage,
560 #ifdef CONFIG_COMPAT
561 	.compat_setsockopt = compat_sock_common_setsockopt,
562 	.compat_getsockopt = compat_sock_common_getsockopt,
563 #endif
564 };
565 
566 static const struct net_proto_family inet6_family_ops = {
567 	.family = PF_INET6,
568 	.create = inet6_create,
569 	.owner	= THIS_MODULE,
570 };
571 
572 int inet6_register_protosw(struct inet_protosw *p)
573 {
574 	struct list_head *lh;
575 	struct inet_protosw *answer;
576 	struct list_head *last_perm;
577 	int protocol = p->protocol;
578 	int ret;
579 
580 	spin_lock_bh(&inetsw6_lock);
581 
582 	ret = -EINVAL;
583 	if (p->type >= SOCK_MAX)
584 		goto out_illegal;
585 
586 	/* If we are trying to override a permanent protocol, bail. */
587 	answer = NULL;
588 	ret = -EPERM;
589 	last_perm = &inetsw6[p->type];
590 	list_for_each(lh, &inetsw6[p->type]) {
591 		answer = list_entry(lh, struct inet_protosw, list);
592 
593 		/* Check only the non-wild match. */
594 		if (INET_PROTOSW_PERMANENT & answer->flags) {
595 			if (protocol == answer->protocol)
596 				break;
597 			last_perm = lh;
598 		}
599 
600 		answer = NULL;
601 	}
602 	if (answer)
603 		goto out_permanent;
604 
605 	/* Add the new entry after the last permanent entry if any, so that
606 	 * the new entry does not override a permanent entry when matched with
607 	 * a wild-card protocol. But it is allowed to override any existing
608 	 * non-permanent entry.  This means that when we remove this entry, the
609 	 * system automatically returns to the old behavior.
610 	 */
611 	list_add_rcu(&p->list, last_perm);
612 	ret = 0;
613 out:
614 	spin_unlock_bh(&inetsw6_lock);
615 	return ret;
616 
617 out_permanent:
618 	pr_err("Attempt to override permanent protocol %d\n", protocol);
619 	goto out;
620 
621 out_illegal:
622 	pr_err("Ignoring attempt to register invalid socket type %d\n",
623 	       p->type);
624 	goto out;
625 }
626 EXPORT_SYMBOL(inet6_register_protosw);
627 
628 void
629 inet6_unregister_protosw(struct inet_protosw *p)
630 {
631 	if (INET_PROTOSW_PERMANENT & p->flags) {
632 		pr_err("Attempt to unregister permanent protocol %d\n",
633 		       p->protocol);
634 	} else {
635 		spin_lock_bh(&inetsw6_lock);
636 		list_del_rcu(&p->list);
637 		spin_unlock_bh(&inetsw6_lock);
638 
639 		synchronize_net();
640 	}
641 }
642 EXPORT_SYMBOL(inet6_unregister_protosw);
643 
644 int inet6_sk_rebuild_header(struct sock *sk)
645 {
646 	struct ipv6_pinfo *np = inet6_sk(sk);
647 	struct dst_entry *dst;
648 
649 	dst = __sk_dst_check(sk, np->dst_cookie);
650 
651 	if (dst == NULL) {
652 		struct inet_sock *inet = inet_sk(sk);
653 		struct in6_addr *final_p, final;
654 		struct flowi6 fl6;
655 
656 		memset(&fl6, 0, sizeof(fl6));
657 		fl6.flowi6_proto = sk->sk_protocol;
658 		fl6.daddr = np->daddr;
659 		fl6.saddr = np->saddr;
660 		fl6.flowlabel = np->flow_label;
661 		fl6.flowi6_oif = sk->sk_bound_dev_if;
662 		fl6.flowi6_mark = sk->sk_mark;
663 		fl6.fl6_dport = inet->inet_dport;
664 		fl6.fl6_sport = inet->inet_sport;
665 		security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
666 
667 		final_p = fl6_update_dst(&fl6, np->opt, &final);
668 
669 		dst = ip6_dst_lookup_flow(sk, &fl6, final_p, false);
670 		if (IS_ERR(dst)) {
671 			sk->sk_route_caps = 0;
672 			sk->sk_err_soft = -PTR_ERR(dst);
673 			return PTR_ERR(dst);
674 		}
675 
676 		__ip6_dst_store(sk, dst, NULL, NULL);
677 	}
678 
679 	return 0;
680 }
681 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
682 
683 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb)
684 {
685 	const struct ipv6_pinfo *np = inet6_sk(sk);
686 	const struct inet6_skb_parm *opt = IP6CB(skb);
687 
688 	if (np->rxopt.all) {
689 		if ((opt->hop && (np->rxopt.bits.hopopts ||
690 				  np->rxopt.bits.ohopopts)) ||
691 		    ((IPV6_FLOWINFO_MASK &
692 		      *(__be32 *)skb_network_header(skb)) &&
693 		     np->rxopt.bits.rxflow) ||
694 		    (opt->srcrt && (np->rxopt.bits.srcrt ||
695 		     np->rxopt.bits.osrcrt)) ||
696 		    ((opt->dst1 || opt->dst0) &&
697 		     (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
698 			return true;
699 	}
700 	return false;
701 }
702 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
703 
704 static struct packet_type ipv6_packet_type __read_mostly = {
705 	.type = cpu_to_be16(ETH_P_IPV6),
706 	.func = ipv6_rcv,
707 };
708 
709 static int __init ipv6_packet_init(void)
710 {
711 	dev_add_pack(&ipv6_packet_type);
712 	return 0;
713 }
714 
715 static void ipv6_packet_cleanup(void)
716 {
717 	dev_remove_pack(&ipv6_packet_type);
718 }
719 
720 static int __net_init ipv6_init_mibs(struct net *net)
721 {
722 	if (snmp_mib_init((void __percpu **)net->mib.udp_stats_in6,
723 			  sizeof(struct udp_mib),
724 			  __alignof__(struct udp_mib)) < 0)
725 		return -ENOMEM;
726 	if (snmp_mib_init((void __percpu **)net->mib.udplite_stats_in6,
727 			  sizeof(struct udp_mib),
728 			  __alignof__(struct udp_mib)) < 0)
729 		goto err_udplite_mib;
730 	if (snmp_mib_init((void __percpu **)net->mib.ipv6_statistics,
731 			  sizeof(struct ipstats_mib),
732 			  __alignof__(struct ipstats_mib)) < 0)
733 		goto err_ip_mib;
734 	if (snmp_mib_init((void __percpu **)net->mib.icmpv6_statistics,
735 			  sizeof(struct icmpv6_mib),
736 			  __alignof__(struct icmpv6_mib)) < 0)
737 		goto err_icmp_mib;
738 	net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
739 						GFP_KERNEL);
740 	if (!net->mib.icmpv6msg_statistics)
741 		goto err_icmpmsg_mib;
742 	return 0;
743 
744 err_icmpmsg_mib:
745 	snmp_mib_free((void __percpu **)net->mib.icmpv6_statistics);
746 err_icmp_mib:
747 	snmp_mib_free((void __percpu **)net->mib.ipv6_statistics);
748 err_ip_mib:
749 	snmp_mib_free((void __percpu **)net->mib.udplite_stats_in6);
750 err_udplite_mib:
751 	snmp_mib_free((void __percpu **)net->mib.udp_stats_in6);
752 	return -ENOMEM;
753 }
754 
755 static void ipv6_cleanup_mibs(struct net *net)
756 {
757 	snmp_mib_free((void __percpu **)net->mib.udp_stats_in6);
758 	snmp_mib_free((void __percpu **)net->mib.udplite_stats_in6);
759 	snmp_mib_free((void __percpu **)net->mib.ipv6_statistics);
760 	snmp_mib_free((void __percpu **)net->mib.icmpv6_statistics);
761 	kfree(net->mib.icmpv6msg_statistics);
762 }
763 
764 static int __net_init inet6_net_init(struct net *net)
765 {
766 	int err = 0;
767 
768 	net->ipv6.sysctl.bindv6only = 0;
769 	net->ipv6.sysctl.icmpv6_time = 1*HZ;
770 	atomic_set(&net->ipv6.rt_genid, 0);
771 
772 	err = ipv6_init_mibs(net);
773 	if (err)
774 		return err;
775 #ifdef CONFIG_PROC_FS
776 	err = udp6_proc_init(net);
777 	if (err)
778 		goto out;
779 	err = tcp6_proc_init(net);
780 	if (err)
781 		goto proc_tcp6_fail;
782 	err = ac6_proc_init(net);
783 	if (err)
784 		goto proc_ac6_fail;
785 #endif
786 	return err;
787 
788 #ifdef CONFIG_PROC_FS
789 proc_ac6_fail:
790 	tcp6_proc_exit(net);
791 proc_tcp6_fail:
792 	udp6_proc_exit(net);
793 out:
794 	ipv6_cleanup_mibs(net);
795 	return err;
796 #endif
797 }
798 
799 static void __net_exit inet6_net_exit(struct net *net)
800 {
801 #ifdef CONFIG_PROC_FS
802 	udp6_proc_exit(net);
803 	tcp6_proc_exit(net);
804 	ac6_proc_exit(net);
805 #endif
806 	ipv6_cleanup_mibs(net);
807 }
808 
809 static struct pernet_operations inet6_net_ops = {
810 	.init = inet6_net_init,
811 	.exit = inet6_net_exit,
812 };
813 
814 static const struct ipv6_stub ipv6_stub_impl = {
815 	.ipv6_sock_mc_join = ipv6_sock_mc_join,
816 	.ipv6_sock_mc_drop = ipv6_sock_mc_drop,
817 	.ipv6_dst_lookup = ip6_dst_lookup,
818 	.udpv6_encap_enable = udpv6_encap_enable,
819 	.ndisc_send_na = ndisc_send_na,
820 	.nd_tbl	= &nd_tbl,
821 };
822 
823 static int __init inet6_init(void)
824 {
825 	struct list_head *r;
826 	int err = 0;
827 
828 	BUILD_BUG_ON(sizeof(struct inet6_skb_parm) > FIELD_SIZEOF(struct sk_buff, cb));
829 
830 	/* Register the socket-side information for inet6_create.  */
831 	for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
832 		INIT_LIST_HEAD(r);
833 
834 	if (disable_ipv6_mod) {
835 		pr_info("Loaded, but administratively disabled, reboot required to enable\n");
836 		goto out;
837 	}
838 
839 	err = proto_register(&tcpv6_prot, 1);
840 	if (err)
841 		goto out;
842 
843 	err = proto_register(&udpv6_prot, 1);
844 	if (err)
845 		goto out_unregister_tcp_proto;
846 
847 	err = proto_register(&udplitev6_prot, 1);
848 	if (err)
849 		goto out_unregister_udp_proto;
850 
851 	err = proto_register(&rawv6_prot, 1);
852 	if (err)
853 		goto out_unregister_udplite_proto;
854 
855 	err = proto_register(&pingv6_prot, 1);
856 	if (err)
857 		goto out_unregister_ping_proto;
858 
859 	/* We MUST register RAW sockets before we create the ICMP6,
860 	 * IGMP6, or NDISC control sockets.
861 	 */
862 	err = rawv6_init();
863 	if (err)
864 		goto out_unregister_raw_proto;
865 
866 	/* Register the family here so that the init calls below will
867 	 * be able to create sockets. (?? is this dangerous ??)
868 	 */
869 	err = sock_register(&inet6_family_ops);
870 	if (err)
871 		goto out_sock_register_fail;
872 
873 	tcpv6_prot.sysctl_mem = init_net.ipv4.sysctl_tcp_mem;
874 
875 	/*
876 	 *	ipngwg API draft makes clear that the correct semantics
877 	 *	for TCP and UDP is to consider one TCP and UDP instance
878 	 *	in a host available by both INET and INET6 APIs and
879 	 *	able to communicate via both network protocols.
880 	 */
881 
882 	err = register_pernet_subsys(&inet6_net_ops);
883 	if (err)
884 		goto register_pernet_fail;
885 	err = icmpv6_init();
886 	if (err)
887 		goto icmp_fail;
888 	err = ip6_mr_init();
889 	if (err)
890 		goto ipmr_fail;
891 	err = ndisc_init();
892 	if (err)
893 		goto ndisc_fail;
894 	err = igmp6_init();
895 	if (err)
896 		goto igmp_fail;
897 
898 	ipv6_stub = &ipv6_stub_impl;
899 
900 	err = ipv6_netfilter_init();
901 	if (err)
902 		goto netfilter_fail;
903 	/* Create /proc/foo6 entries. */
904 #ifdef CONFIG_PROC_FS
905 	err = -ENOMEM;
906 	if (raw6_proc_init())
907 		goto proc_raw6_fail;
908 	if (udplite6_proc_init())
909 		goto proc_udplite6_fail;
910 	if (ipv6_misc_proc_init())
911 		goto proc_misc6_fail;
912 	if (if6_proc_init())
913 		goto proc_if6_fail;
914 #endif
915 	err = ip6_route_init();
916 	if (err)
917 		goto ip6_route_fail;
918 	err = ndisc_late_init();
919 	if (err)
920 		goto ndisc_late_fail;
921 	err = ip6_flowlabel_init();
922 	if (err)
923 		goto ip6_flowlabel_fail;
924 	err = addrconf_init();
925 	if (err)
926 		goto addrconf_fail;
927 
928 	/* Init v6 extension headers. */
929 	err = ipv6_exthdrs_init();
930 	if (err)
931 		goto ipv6_exthdrs_fail;
932 
933 	err = ipv6_frag_init();
934 	if (err)
935 		goto ipv6_frag_fail;
936 
937 	/* Init v6 transport protocols. */
938 	err = udpv6_init();
939 	if (err)
940 		goto udpv6_fail;
941 
942 	err = udplitev6_init();
943 	if (err)
944 		goto udplitev6_fail;
945 
946 	err = tcpv6_init();
947 	if (err)
948 		goto tcpv6_fail;
949 
950 	err = ipv6_packet_init();
951 	if (err)
952 		goto ipv6_packet_fail;
953 
954 	err = pingv6_init();
955 	if (err)
956 		goto pingv6_fail;
957 
958 #ifdef CONFIG_SYSCTL
959 	err = ipv6_sysctl_register();
960 	if (err)
961 		goto sysctl_fail;
962 #endif
963 out:
964 	return err;
965 
966 #ifdef CONFIG_SYSCTL
967 sysctl_fail:
968 	ipv6_packet_cleanup();
969 #endif
970 pingv6_fail:
971 	pingv6_exit();
972 ipv6_packet_fail:
973 	tcpv6_exit();
974 tcpv6_fail:
975 	udplitev6_exit();
976 udplitev6_fail:
977 	udpv6_exit();
978 udpv6_fail:
979 	ipv6_frag_exit();
980 ipv6_frag_fail:
981 	ipv6_exthdrs_exit();
982 ipv6_exthdrs_fail:
983 	addrconf_cleanup();
984 addrconf_fail:
985 	ip6_flowlabel_cleanup();
986 ip6_flowlabel_fail:
987 	ndisc_late_cleanup();
988 ndisc_late_fail:
989 	ip6_route_cleanup();
990 ip6_route_fail:
991 #ifdef CONFIG_PROC_FS
992 	if6_proc_exit();
993 proc_if6_fail:
994 	ipv6_misc_proc_exit();
995 proc_misc6_fail:
996 	udplite6_proc_exit();
997 proc_udplite6_fail:
998 	raw6_proc_exit();
999 proc_raw6_fail:
1000 #endif
1001 	ipv6_netfilter_fini();
1002 netfilter_fail:
1003 	igmp6_cleanup();
1004 igmp_fail:
1005 	ndisc_cleanup();
1006 ndisc_fail:
1007 	ip6_mr_cleanup();
1008 ipmr_fail:
1009 	icmpv6_cleanup();
1010 icmp_fail:
1011 	unregister_pernet_subsys(&inet6_net_ops);
1012 register_pernet_fail:
1013 	sock_unregister(PF_INET6);
1014 	rtnl_unregister_all(PF_INET6);
1015 out_sock_register_fail:
1016 	rawv6_exit();
1017 out_unregister_ping_proto:
1018 	proto_unregister(&pingv6_prot);
1019 out_unregister_raw_proto:
1020 	proto_unregister(&rawv6_prot);
1021 out_unregister_udplite_proto:
1022 	proto_unregister(&udplitev6_prot);
1023 out_unregister_udp_proto:
1024 	proto_unregister(&udpv6_prot);
1025 out_unregister_tcp_proto:
1026 	proto_unregister(&tcpv6_prot);
1027 	goto out;
1028 }
1029 module_init(inet6_init);
1030 
1031 static void __exit inet6_exit(void)
1032 {
1033 	if (disable_ipv6_mod)
1034 		return;
1035 
1036 	/* First of all disallow new sockets creation. */
1037 	sock_unregister(PF_INET6);
1038 	/* Disallow any further netlink messages */
1039 	rtnl_unregister_all(PF_INET6);
1040 
1041 	udpv6_exit();
1042 	udplitev6_exit();
1043 	tcpv6_exit();
1044 
1045 	/* Cleanup code parts. */
1046 	ipv6_packet_cleanup();
1047 	ipv6_frag_exit();
1048 	ipv6_exthdrs_exit();
1049 	addrconf_cleanup();
1050 	ip6_flowlabel_cleanup();
1051 	ndisc_late_cleanup();
1052 	ip6_route_cleanup();
1053 #ifdef CONFIG_PROC_FS
1054 
1055 	/* Cleanup code parts. */
1056 	if6_proc_exit();
1057 	ipv6_misc_proc_exit();
1058 	udplite6_proc_exit();
1059 	raw6_proc_exit();
1060 #endif
1061 	ipv6_netfilter_fini();
1062 	ipv6_stub = NULL;
1063 	igmp6_cleanup();
1064 	ndisc_cleanup();
1065 	ip6_mr_cleanup();
1066 	icmpv6_cleanup();
1067 	rawv6_exit();
1068 
1069 	unregister_pernet_subsys(&inet6_net_ops);
1070 	proto_unregister(&rawv6_prot);
1071 	proto_unregister(&udplitev6_prot);
1072 	proto_unregister(&udpv6_prot);
1073 	proto_unregister(&tcpv6_prot);
1074 
1075 	rcu_barrier(); /* Wait for completion of call_rcu()'s */
1076 }
1077 module_exit(inet6_exit);
1078 
1079 MODULE_ALIAS_NETPROTO(PF_INET6);
1080