xref: /openbmc/linux/net/ipv6/af_inet6.c (revision e6c81cce)
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 	int try_loading_module = 0;
110 	int err;
111 
112 	/* Look for the requested type/protocol pair. */
113 lookup_protocol:
114 	err = -ESOCKTNOSUPPORT;
115 	rcu_read_lock();
116 	list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
117 
118 		err = 0;
119 		/* Check the non-wild match. */
120 		if (protocol == answer->protocol) {
121 			if (protocol != IPPROTO_IP)
122 				break;
123 		} else {
124 			/* Check for the two wild cases. */
125 			if (IPPROTO_IP == protocol) {
126 				protocol = answer->protocol;
127 				break;
128 			}
129 			if (IPPROTO_IP == answer->protocol)
130 				break;
131 		}
132 		err = -EPROTONOSUPPORT;
133 	}
134 
135 	if (err) {
136 		if (try_loading_module < 2) {
137 			rcu_read_unlock();
138 			/*
139 			 * Be more specific, e.g. net-pf-10-proto-132-type-1
140 			 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
141 			 */
142 			if (++try_loading_module == 1)
143 				request_module("net-pf-%d-proto-%d-type-%d",
144 						PF_INET6, protocol, sock->type);
145 			/*
146 			 * Fall back to generic, e.g. net-pf-10-proto-132
147 			 * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
148 			 */
149 			else
150 				request_module("net-pf-%d-proto-%d",
151 						PF_INET6, protocol);
152 			goto lookup_protocol;
153 		} else
154 			goto out_rcu_unlock;
155 	}
156 
157 	err = -EPERM;
158 	if (sock->type == SOCK_RAW && !kern &&
159 	    !ns_capable(net->user_ns, CAP_NET_RAW))
160 		goto out_rcu_unlock;
161 
162 	sock->ops = answer->ops;
163 	answer_prot = answer->prot;
164 	answer_flags = answer->flags;
165 	rcu_read_unlock();
166 
167 	WARN_ON(!answer_prot->slab);
168 
169 	err = -ENOBUFS;
170 	sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot);
171 	if (!sk)
172 		goto out;
173 
174 	sock_init_data(sock, sk);
175 
176 	err = 0;
177 	if (INET_PROTOSW_REUSE & answer_flags)
178 		sk->sk_reuse = SK_CAN_REUSE;
179 
180 	inet = inet_sk(sk);
181 	inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
182 
183 	if (SOCK_RAW == sock->type) {
184 		inet->inet_num = protocol;
185 		if (IPPROTO_RAW == protocol)
186 			inet->hdrincl = 1;
187 	}
188 
189 	sk->sk_destruct		= inet_sock_destruct;
190 	sk->sk_family		= PF_INET6;
191 	sk->sk_protocol		= protocol;
192 
193 	sk->sk_backlog_rcv	= answer->prot->backlog_rcv;
194 
195 	inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
196 	np->hop_limit	= -1;
197 	np->mcast_hops	= IPV6_DEFAULT_MCASTHOPS;
198 	np->mc_loop	= 1;
199 	np->pmtudisc	= IPV6_PMTUDISC_WANT;
200 	sk->sk_ipv6only	= net->ipv6.sysctl.bindv6only;
201 
202 	/* Init the ipv4 part of the socket since we can have sockets
203 	 * using v6 API for ipv4.
204 	 */
205 	inet->uc_ttl	= -1;
206 
207 	inet->mc_loop	= 1;
208 	inet->mc_ttl	= 1;
209 	inet->mc_index	= 0;
210 	inet->mc_list	= NULL;
211 	inet->rcv_tos	= 0;
212 
213 	if (net->ipv4.sysctl_ip_no_pmtu_disc)
214 		inet->pmtudisc = IP_PMTUDISC_DONT;
215 	else
216 		inet->pmtudisc = IP_PMTUDISC_WANT;
217 	/*
218 	 * Increment only the relevant sk_prot->socks debug field, this changes
219 	 * the previous behaviour of incrementing both the equivalent to
220 	 * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
221 	 *
222 	 * This allows better debug granularity as we'll know exactly how many
223 	 * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
224 	 * transport protocol socks. -acme
225 	 */
226 	sk_refcnt_debug_inc(sk);
227 
228 	if (inet->inet_num) {
229 		/* It assumes that any protocol which allows
230 		 * the user to assign a number at socket
231 		 * creation time automatically shares.
232 		 */
233 		inet->inet_sport = htons(inet->inet_num);
234 		sk->sk_prot->hash(sk);
235 	}
236 	if (sk->sk_prot->init) {
237 		err = sk->sk_prot->init(sk);
238 		if (err) {
239 			sk_common_release(sk);
240 			goto out;
241 		}
242 	}
243 out:
244 	return err;
245 out_rcu_unlock:
246 	rcu_read_unlock();
247 	goto out;
248 }
249 
250 
251 /* bind for INET6 API */
252 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
253 {
254 	struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr;
255 	struct sock *sk = sock->sk;
256 	struct inet_sock *inet = inet_sk(sk);
257 	struct ipv6_pinfo *np = inet6_sk(sk);
258 	struct net *net = sock_net(sk);
259 	__be32 v4addr = 0;
260 	unsigned short snum;
261 	int addr_type = 0;
262 	int err = 0;
263 
264 	/* If the socket has its own bind function then use it. */
265 	if (sk->sk_prot->bind)
266 		return sk->sk_prot->bind(sk, uaddr, addr_len);
267 
268 	if (addr_len < SIN6_LEN_RFC2133)
269 		return -EINVAL;
270 
271 	if (addr->sin6_family != AF_INET6)
272 		return -EAFNOSUPPORT;
273 
274 	addr_type = ipv6_addr_type(&addr->sin6_addr);
275 	if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
276 		return -EINVAL;
277 
278 	snum = ntohs(addr->sin6_port);
279 	if (snum && snum < PROT_SOCK && !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
280 		return -EACCES;
281 
282 	lock_sock(sk);
283 
284 	/* Check these errors (active socket, double bind). */
285 	if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
286 		err = -EINVAL;
287 		goto out;
288 	}
289 
290 	/* Check if the address belongs to the host. */
291 	if (addr_type == IPV6_ADDR_MAPPED) {
292 		int chk_addr_ret;
293 
294 		/* Binding to v4-mapped address on a v6-only socket
295 		 * makes no sense
296 		 */
297 		if (sk->sk_ipv6only) {
298 			err = -EINVAL;
299 			goto out;
300 		}
301 
302 		/* Reproduce AF_INET checks to make the bindings consistent */
303 		v4addr = addr->sin6_addr.s6_addr32[3];
304 		chk_addr_ret = inet_addr_type(net, v4addr);
305 		if (!net->ipv4.sysctl_ip_nonlocal_bind &&
306 		    !(inet->freebind || inet->transparent) &&
307 		    v4addr != htonl(INADDR_ANY) &&
308 		    chk_addr_ret != RTN_LOCAL &&
309 		    chk_addr_ret != RTN_MULTICAST &&
310 		    chk_addr_ret != RTN_BROADCAST) {
311 			err = -EADDRNOTAVAIL;
312 			goto out;
313 		}
314 	} else {
315 		if (addr_type != IPV6_ADDR_ANY) {
316 			struct net_device *dev = NULL;
317 
318 			rcu_read_lock();
319 			if (__ipv6_addr_needs_scope_id(addr_type)) {
320 				if (addr_len >= sizeof(struct sockaddr_in6) &&
321 				    addr->sin6_scope_id) {
322 					/* Override any existing binding, if another one
323 					 * is supplied by user.
324 					 */
325 					sk->sk_bound_dev_if = addr->sin6_scope_id;
326 				}
327 
328 				/* Binding to link-local address requires an interface */
329 				if (!sk->sk_bound_dev_if) {
330 					err = -EINVAL;
331 					goto out_unlock;
332 				}
333 				dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
334 				if (!dev) {
335 					err = -ENODEV;
336 					goto out_unlock;
337 				}
338 			}
339 
340 			/* ipv4 addr of the socket is invalid.  Only the
341 			 * unspecified and mapped address have a v4 equivalent.
342 			 */
343 			v4addr = LOOPBACK4_IPV6;
344 			if (!(addr_type & IPV6_ADDR_MULTICAST))	{
345 				if (!(inet->freebind || inet->transparent) &&
346 				    !ipv6_chk_addr(net, &addr->sin6_addr,
347 						   dev, 0)) {
348 					err = -EADDRNOTAVAIL;
349 					goto out_unlock;
350 				}
351 			}
352 			rcu_read_unlock();
353 		}
354 	}
355 
356 	inet->inet_rcv_saddr = v4addr;
357 	inet->inet_saddr = v4addr;
358 
359 	sk->sk_v6_rcv_saddr = addr->sin6_addr;
360 
361 	if (!(addr_type & IPV6_ADDR_MULTICAST))
362 		np->saddr = addr->sin6_addr;
363 
364 	/* Make sure we are allowed to bind here. */
365 	if (sk->sk_prot->get_port(sk, snum)) {
366 		inet_reset_saddr(sk);
367 		err = -EADDRINUSE;
368 		goto out;
369 	}
370 
371 	if (addr_type != IPV6_ADDR_ANY) {
372 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
373 		if (addr_type != IPV6_ADDR_MAPPED)
374 			sk->sk_ipv6only = 1;
375 	}
376 	if (snum)
377 		sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
378 	inet->inet_sport = htons(inet->inet_num);
379 	inet->inet_dport = 0;
380 	inet->inet_daddr = 0;
381 out:
382 	release_sock(sk);
383 	return err;
384 out_unlock:
385 	rcu_read_unlock();
386 	goto out;
387 }
388 EXPORT_SYMBOL(inet6_bind);
389 
390 int inet6_release(struct socket *sock)
391 {
392 	struct sock *sk = sock->sk;
393 
394 	if (!sk)
395 		return -EINVAL;
396 
397 	/* Free mc lists */
398 	ipv6_sock_mc_close(sk);
399 
400 	/* Free ac lists */
401 	ipv6_sock_ac_close(sk);
402 
403 	return inet_release(sock);
404 }
405 EXPORT_SYMBOL(inet6_release);
406 
407 void inet6_destroy_sock(struct sock *sk)
408 {
409 	struct ipv6_pinfo *np = inet6_sk(sk);
410 	struct sk_buff *skb;
411 	struct ipv6_txoptions *opt;
412 
413 	/* Release rx options */
414 
415 	skb = xchg(&np->pktoptions, NULL);
416 	if (skb)
417 		kfree_skb(skb);
418 
419 	skb = xchg(&np->rxpmtu, NULL);
420 	if (skb)
421 		kfree_skb(skb);
422 
423 	/* Free flowlabels */
424 	fl6_free_socklist(sk);
425 
426 	/* Free tx options */
427 
428 	opt = xchg(&np->opt, NULL);
429 	if (opt)
430 		sock_kfree_s(sk, opt, opt->tot_len);
431 }
432 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
433 
434 /*
435  *	This does both peername and sockname.
436  */
437 
438 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
439 		 int *uaddr_len, int peer)
440 {
441 	struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
442 	struct sock *sk = sock->sk;
443 	struct inet_sock *inet = inet_sk(sk);
444 	struct ipv6_pinfo *np = inet6_sk(sk);
445 
446 	sin->sin6_family = AF_INET6;
447 	sin->sin6_flowinfo = 0;
448 	sin->sin6_scope_id = 0;
449 	if (peer) {
450 		if (!inet->inet_dport)
451 			return -ENOTCONN;
452 		if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
453 		    peer == 1)
454 			return -ENOTCONN;
455 		sin->sin6_port = inet->inet_dport;
456 		sin->sin6_addr = sk->sk_v6_daddr;
457 		if (np->sndflow)
458 			sin->sin6_flowinfo = np->flow_label;
459 	} else {
460 		if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
461 			sin->sin6_addr = np->saddr;
462 		else
463 			sin->sin6_addr = sk->sk_v6_rcv_saddr;
464 
465 		sin->sin6_port = inet->inet_sport;
466 	}
467 	sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
468 						 sk->sk_bound_dev_if);
469 	*uaddr_len = sizeof(*sin);
470 	return 0;
471 }
472 EXPORT_SYMBOL(inet6_getname);
473 
474 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
475 {
476 	struct sock *sk = sock->sk;
477 	struct net *net = sock_net(sk);
478 
479 	switch (cmd) {
480 	case SIOCGSTAMP:
481 		return sock_get_timestamp(sk, (struct timeval __user *)arg);
482 
483 	case SIOCGSTAMPNS:
484 		return sock_get_timestampns(sk, (struct timespec __user *)arg);
485 
486 	case SIOCADDRT:
487 	case SIOCDELRT:
488 
489 		return ipv6_route_ioctl(net, cmd, (void __user *)arg);
490 
491 	case SIOCSIFADDR:
492 		return addrconf_add_ifaddr(net, (void __user *) arg);
493 	case SIOCDIFADDR:
494 		return addrconf_del_ifaddr(net, (void __user *) arg);
495 	case SIOCSIFDSTADDR:
496 		return addrconf_set_dstaddr(net, (void __user *) arg);
497 	default:
498 		if (!sk->sk_prot->ioctl)
499 			return -ENOIOCTLCMD;
500 		return sk->sk_prot->ioctl(sk, cmd, arg);
501 	}
502 	/*NOTREACHED*/
503 	return 0;
504 }
505 EXPORT_SYMBOL(inet6_ioctl);
506 
507 const struct proto_ops inet6_stream_ops = {
508 	.family		   = PF_INET6,
509 	.owner		   = THIS_MODULE,
510 	.release	   = inet6_release,
511 	.bind		   = inet6_bind,
512 	.connect	   = inet_stream_connect,	/* ok		*/
513 	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
514 	.accept		   = inet_accept,		/* ok		*/
515 	.getname	   = inet6_getname,
516 	.poll		   = tcp_poll,			/* ok		*/
517 	.ioctl		   = inet6_ioctl,		/* must change  */
518 	.listen		   = inet_listen,		/* ok		*/
519 	.shutdown	   = inet_shutdown,		/* ok		*/
520 	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
521 	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
522 	.sendmsg	   = inet_sendmsg,		/* ok		*/
523 	.recvmsg	   = inet_recvmsg,		/* ok		*/
524 	.mmap		   = sock_no_mmap,
525 	.sendpage	   = inet_sendpage,
526 	.splice_read	   = tcp_splice_read,
527 #ifdef CONFIG_COMPAT
528 	.compat_setsockopt = compat_sock_common_setsockopt,
529 	.compat_getsockopt = compat_sock_common_getsockopt,
530 #endif
531 };
532 
533 const struct proto_ops inet6_dgram_ops = {
534 	.family		   = PF_INET6,
535 	.owner		   = THIS_MODULE,
536 	.release	   = inet6_release,
537 	.bind		   = inet6_bind,
538 	.connect	   = inet_dgram_connect,	/* ok		*/
539 	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
540 	.accept		   = sock_no_accept,		/* a do nothing	*/
541 	.getname	   = inet6_getname,
542 	.poll		   = udp_poll,			/* ok		*/
543 	.ioctl		   = inet6_ioctl,		/* must change  */
544 	.listen		   = sock_no_listen,		/* ok		*/
545 	.shutdown	   = inet_shutdown,		/* ok		*/
546 	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
547 	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
548 	.sendmsg	   = inet_sendmsg,		/* ok		*/
549 	.recvmsg	   = inet_recvmsg,		/* ok		*/
550 	.mmap		   = sock_no_mmap,
551 	.sendpage	   = sock_no_sendpage,
552 #ifdef CONFIG_COMPAT
553 	.compat_setsockopt = compat_sock_common_setsockopt,
554 	.compat_getsockopt = compat_sock_common_getsockopt,
555 #endif
556 };
557 
558 static const struct net_proto_family inet6_family_ops = {
559 	.family = PF_INET6,
560 	.create = inet6_create,
561 	.owner	= THIS_MODULE,
562 };
563 
564 int inet6_register_protosw(struct inet_protosw *p)
565 {
566 	struct list_head *lh;
567 	struct inet_protosw *answer;
568 	struct list_head *last_perm;
569 	int protocol = p->protocol;
570 	int ret;
571 
572 	spin_lock_bh(&inetsw6_lock);
573 
574 	ret = -EINVAL;
575 	if (p->type >= SOCK_MAX)
576 		goto out_illegal;
577 
578 	/* If we are trying to override a permanent protocol, bail. */
579 	answer = NULL;
580 	ret = -EPERM;
581 	last_perm = &inetsw6[p->type];
582 	list_for_each(lh, &inetsw6[p->type]) {
583 		answer = list_entry(lh, struct inet_protosw, list);
584 
585 		/* Check only the non-wild match. */
586 		if (INET_PROTOSW_PERMANENT & answer->flags) {
587 			if (protocol == answer->protocol)
588 				break;
589 			last_perm = lh;
590 		}
591 
592 		answer = NULL;
593 	}
594 	if (answer)
595 		goto out_permanent;
596 
597 	/* Add the new entry after the last permanent entry if any, so that
598 	 * the new entry does not override a permanent entry when matched with
599 	 * a wild-card protocol. But it is allowed to override any existing
600 	 * non-permanent entry.  This means that when we remove this entry, the
601 	 * system automatically returns to the old behavior.
602 	 */
603 	list_add_rcu(&p->list, last_perm);
604 	ret = 0;
605 out:
606 	spin_unlock_bh(&inetsw6_lock);
607 	return ret;
608 
609 out_permanent:
610 	pr_err("Attempt to override permanent protocol %d\n", protocol);
611 	goto out;
612 
613 out_illegal:
614 	pr_err("Ignoring attempt to register invalid socket type %d\n",
615 	       p->type);
616 	goto out;
617 }
618 EXPORT_SYMBOL(inet6_register_protosw);
619 
620 void
621 inet6_unregister_protosw(struct inet_protosw *p)
622 {
623 	if (INET_PROTOSW_PERMANENT & p->flags) {
624 		pr_err("Attempt to unregister permanent protocol %d\n",
625 		       p->protocol);
626 	} else {
627 		spin_lock_bh(&inetsw6_lock);
628 		list_del_rcu(&p->list);
629 		spin_unlock_bh(&inetsw6_lock);
630 
631 		synchronize_net();
632 	}
633 }
634 EXPORT_SYMBOL(inet6_unregister_protosw);
635 
636 int inet6_sk_rebuild_header(struct sock *sk)
637 {
638 	struct ipv6_pinfo *np = inet6_sk(sk);
639 	struct dst_entry *dst;
640 
641 	dst = __sk_dst_check(sk, np->dst_cookie);
642 
643 	if (!dst) {
644 		struct inet_sock *inet = inet_sk(sk);
645 		struct in6_addr *final_p, final;
646 		struct flowi6 fl6;
647 
648 		memset(&fl6, 0, sizeof(fl6));
649 		fl6.flowi6_proto = sk->sk_protocol;
650 		fl6.daddr = sk->sk_v6_daddr;
651 		fl6.saddr = np->saddr;
652 		fl6.flowlabel = np->flow_label;
653 		fl6.flowi6_oif = sk->sk_bound_dev_if;
654 		fl6.flowi6_mark = sk->sk_mark;
655 		fl6.fl6_dport = inet->inet_dport;
656 		fl6.fl6_sport = inet->inet_sport;
657 		security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
658 
659 		final_p = fl6_update_dst(&fl6, np->opt, &final);
660 
661 		dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
662 		if (IS_ERR(dst)) {
663 			sk->sk_route_caps = 0;
664 			sk->sk_err_soft = -PTR_ERR(dst);
665 			return PTR_ERR(dst);
666 		}
667 
668 		__ip6_dst_store(sk, dst, NULL, NULL);
669 	}
670 
671 	return 0;
672 }
673 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
674 
675 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb,
676 		       const struct inet6_skb_parm *opt)
677 {
678 	const struct ipv6_pinfo *np = inet6_sk(sk);
679 
680 	if (np->rxopt.all) {
681 		if ((opt->hop && (np->rxopt.bits.hopopts ||
682 				  np->rxopt.bits.ohopopts)) ||
683 		    (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) &&
684 		     np->rxopt.bits.rxflow) ||
685 		    (opt->srcrt && (np->rxopt.bits.srcrt ||
686 		     np->rxopt.bits.osrcrt)) ||
687 		    ((opt->dst1 || opt->dst0) &&
688 		     (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
689 			return true;
690 	}
691 	return false;
692 }
693 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
694 
695 static struct packet_type ipv6_packet_type __read_mostly = {
696 	.type = cpu_to_be16(ETH_P_IPV6),
697 	.func = ipv6_rcv,
698 };
699 
700 static int __init ipv6_packet_init(void)
701 {
702 	dev_add_pack(&ipv6_packet_type);
703 	return 0;
704 }
705 
706 static void ipv6_packet_cleanup(void)
707 {
708 	dev_remove_pack(&ipv6_packet_type);
709 }
710 
711 static int __net_init ipv6_init_mibs(struct net *net)
712 {
713 	int i;
714 
715 	net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib);
716 	if (!net->mib.udp_stats_in6)
717 		return -ENOMEM;
718 	net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib);
719 	if (!net->mib.udplite_stats_in6)
720 		goto err_udplite_mib;
721 	net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib);
722 	if (!net->mib.ipv6_statistics)
723 		goto err_ip_mib;
724 
725 	for_each_possible_cpu(i) {
726 		struct ipstats_mib *af_inet6_stats;
727 		af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i);
728 		u64_stats_init(&af_inet6_stats->syncp);
729 	}
730 
731 
732 	net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib);
733 	if (!net->mib.icmpv6_statistics)
734 		goto err_icmp_mib;
735 	net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
736 						GFP_KERNEL);
737 	if (!net->mib.icmpv6msg_statistics)
738 		goto err_icmpmsg_mib;
739 	return 0;
740 
741 err_icmpmsg_mib:
742 	free_percpu(net->mib.icmpv6_statistics);
743 err_icmp_mib:
744 	free_percpu(net->mib.ipv6_statistics);
745 err_ip_mib:
746 	free_percpu(net->mib.udplite_stats_in6);
747 err_udplite_mib:
748 	free_percpu(net->mib.udp_stats_in6);
749 	return -ENOMEM;
750 }
751 
752 static void ipv6_cleanup_mibs(struct net *net)
753 {
754 	free_percpu(net->mib.udp_stats_in6);
755 	free_percpu(net->mib.udplite_stats_in6);
756 	free_percpu(net->mib.ipv6_statistics);
757 	free_percpu(net->mib.icmpv6_statistics);
758 	kfree(net->mib.icmpv6msg_statistics);
759 }
760 
761 static int __net_init inet6_net_init(struct net *net)
762 {
763 	int err = 0;
764 
765 	net->ipv6.sysctl.bindv6only = 0;
766 	net->ipv6.sysctl.icmpv6_time = 1*HZ;
767 	net->ipv6.sysctl.flowlabel_consistency = 1;
768 	net->ipv6.sysctl.auto_flowlabels = 0;
769 	net->ipv6.sysctl.idgen_retries = 3;
770 	net->ipv6.sysctl.idgen_delay = 1 * HZ;
771 	atomic_set(&net->ipv6.fib6_sernum, 1);
772 
773 	err = ipv6_init_mibs(net);
774 	if (err)
775 		return err;
776 #ifdef CONFIG_PROC_FS
777 	err = udp6_proc_init(net);
778 	if (err)
779 		goto out;
780 	err = tcp6_proc_init(net);
781 	if (err)
782 		goto proc_tcp6_fail;
783 	err = ac6_proc_init(net);
784 	if (err)
785 		goto proc_ac6_fail;
786 #endif
787 	return err;
788 
789 #ifdef CONFIG_PROC_FS
790 proc_ac6_fail:
791 	tcp6_proc_exit(net);
792 proc_tcp6_fail:
793 	udp6_proc_exit(net);
794 out:
795 	ipv6_cleanup_mibs(net);
796 	return err;
797 #endif
798 }
799 
800 static void __net_exit inet6_net_exit(struct net *net)
801 {
802 #ifdef CONFIG_PROC_FS
803 	udp6_proc_exit(net);
804 	tcp6_proc_exit(net);
805 	ac6_proc_exit(net);
806 #endif
807 	ipv6_cleanup_mibs(net);
808 }
809 
810 static struct pernet_operations inet6_net_ops = {
811 	.init = inet6_net_init,
812 	.exit = inet6_net_exit,
813 };
814 
815 static const struct ipv6_stub ipv6_stub_impl = {
816 	.ipv6_sock_mc_join = ipv6_sock_mc_join,
817 	.ipv6_sock_mc_drop = ipv6_sock_mc_drop,
818 	.ipv6_dst_lookup = ip6_dst_lookup,
819 	.udpv6_encap_enable = udpv6_encap_enable,
820 	.ndisc_send_na = ndisc_send_na,
821 	.nd_tbl	= &nd_tbl,
822 };
823 
824 static int __init inet6_init(void)
825 {
826 	struct list_head *r;
827 	int err = 0;
828 
829 	sock_skb_cb_check_size(sizeof(struct inet6_skb_parm));
830 
831 	/* Register the socket-side information for inet6_create.  */
832 	for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
833 		INIT_LIST_HEAD(r);
834 
835 	if (disable_ipv6_mod) {
836 		pr_info("Loaded, but administratively disabled, reboot required to enable\n");
837 		goto out;
838 	}
839 
840 	err = proto_register(&tcpv6_prot, 1);
841 	if (err)
842 		goto out;
843 
844 	err = proto_register(&udpv6_prot, 1);
845 	if (err)
846 		goto out_unregister_tcp_proto;
847 
848 	err = proto_register(&udplitev6_prot, 1);
849 	if (err)
850 		goto out_unregister_udp_proto;
851 
852 	err = proto_register(&rawv6_prot, 1);
853 	if (err)
854 		goto out_unregister_udplite_proto;
855 
856 	err = proto_register(&pingv6_prot, 1);
857 	if (err)
858 		goto out_unregister_ping_proto;
859 
860 	/* We MUST register RAW sockets before we create the ICMP6,
861 	 * IGMP6, or NDISC control sockets.
862 	 */
863 	err = rawv6_init();
864 	if (err)
865 		goto out_unregister_raw_proto;
866 
867 	/* Register the family here so that the init calls below will
868 	 * be able to create sockets. (?? is this dangerous ??)
869 	 */
870 	err = sock_register(&inet6_family_ops);
871 	if (err)
872 		goto out_sock_register_fail;
873 
874 	/*
875 	 *	ipngwg API draft makes clear that the correct semantics
876 	 *	for TCP and UDP is to consider one TCP and UDP instance
877 	 *	in a host available by both INET and INET6 APIs and
878 	 *	able to communicate via both network protocols.
879 	 */
880 
881 	err = register_pernet_subsys(&inet6_net_ops);
882 	if (err)
883 		goto register_pernet_fail;
884 	err = icmpv6_init();
885 	if (err)
886 		goto icmp_fail;
887 	err = ip6_mr_init();
888 	if (err)
889 		goto ipmr_fail;
890 	err = ndisc_init();
891 	if (err)
892 		goto ndisc_fail;
893 	err = igmp6_init();
894 	if (err)
895 		goto igmp_fail;
896 
897 	ipv6_stub = &ipv6_stub_impl;
898 
899 	err = ipv6_netfilter_init();
900 	if (err)
901 		goto netfilter_fail;
902 	/* Create /proc/foo6 entries. */
903 #ifdef CONFIG_PROC_FS
904 	err = -ENOMEM;
905 	if (raw6_proc_init())
906 		goto proc_raw6_fail;
907 	if (udplite6_proc_init())
908 		goto proc_udplite6_fail;
909 	if (ipv6_misc_proc_init())
910 		goto proc_misc6_fail;
911 	if (if6_proc_init())
912 		goto proc_if6_fail;
913 #endif
914 	err = ip6_route_init();
915 	if (err)
916 		goto ip6_route_fail;
917 	err = ndisc_late_init();
918 	if (err)
919 		goto ndisc_late_fail;
920 	err = ip6_flowlabel_init();
921 	if (err)
922 		goto ip6_flowlabel_fail;
923 	err = addrconf_init();
924 	if (err)
925 		goto addrconf_fail;
926 
927 	/* Init v6 extension headers. */
928 	err = ipv6_exthdrs_init();
929 	if (err)
930 		goto ipv6_exthdrs_fail;
931 
932 	err = ipv6_frag_init();
933 	if (err)
934 		goto ipv6_frag_fail;
935 
936 	/* Init v6 transport protocols. */
937 	err = udpv6_init();
938 	if (err)
939 		goto udpv6_fail;
940 
941 	err = udplitev6_init();
942 	if (err)
943 		goto udplitev6_fail;
944 
945 	err = tcpv6_init();
946 	if (err)
947 		goto tcpv6_fail;
948 
949 	err = ipv6_packet_init();
950 	if (err)
951 		goto ipv6_packet_fail;
952 
953 	err = pingv6_init();
954 	if (err)
955 		goto pingv6_fail;
956 
957 #ifdef CONFIG_SYSCTL
958 	err = ipv6_sysctl_register();
959 	if (err)
960 		goto sysctl_fail;
961 #endif
962 out:
963 	return err;
964 
965 #ifdef CONFIG_SYSCTL
966 sysctl_fail:
967 	pingv6_exit();
968 #endif
969 pingv6_fail:
970 	ipv6_packet_cleanup();
971 ipv6_packet_fail:
972 	tcpv6_exit();
973 tcpv6_fail:
974 	udplitev6_exit();
975 udplitev6_fail:
976 	udpv6_exit();
977 udpv6_fail:
978 	ipv6_frag_exit();
979 ipv6_frag_fail:
980 	ipv6_exthdrs_exit();
981 ipv6_exthdrs_fail:
982 	addrconf_cleanup();
983 addrconf_fail:
984 	ip6_flowlabel_cleanup();
985 ip6_flowlabel_fail:
986 	ndisc_late_cleanup();
987 ndisc_late_fail:
988 	ip6_route_cleanup();
989 ip6_route_fail:
990 #ifdef CONFIG_PROC_FS
991 	if6_proc_exit();
992 proc_if6_fail:
993 	ipv6_misc_proc_exit();
994 proc_misc6_fail:
995 	udplite6_proc_exit();
996 proc_udplite6_fail:
997 	raw6_proc_exit();
998 proc_raw6_fail:
999 #endif
1000 	ipv6_netfilter_fini();
1001 netfilter_fail:
1002 	igmp6_cleanup();
1003 igmp_fail:
1004 	ndisc_cleanup();
1005 ndisc_fail:
1006 	ip6_mr_cleanup();
1007 ipmr_fail:
1008 	icmpv6_cleanup();
1009 icmp_fail:
1010 	unregister_pernet_subsys(&inet6_net_ops);
1011 register_pernet_fail:
1012 	sock_unregister(PF_INET6);
1013 	rtnl_unregister_all(PF_INET6);
1014 out_sock_register_fail:
1015 	rawv6_exit();
1016 out_unregister_ping_proto:
1017 	proto_unregister(&pingv6_prot);
1018 out_unregister_raw_proto:
1019 	proto_unregister(&rawv6_prot);
1020 out_unregister_udplite_proto:
1021 	proto_unregister(&udplitev6_prot);
1022 out_unregister_udp_proto:
1023 	proto_unregister(&udpv6_prot);
1024 out_unregister_tcp_proto:
1025 	proto_unregister(&tcpv6_prot);
1026 	goto out;
1027 }
1028 module_init(inet6_init);
1029 
1030 MODULE_ALIAS_NETPROTO(PF_INET6);
1031