xref: /openbmc/linux/net/ipv6/af_inet6.c (revision 6a613ac6)
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, kern);
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 	np->autoflowlabel = ip6_default_np_autolabel(sock_net(sk));
201 	sk->sk_ipv6only	= net->ipv6.sysctl.bindv6only;
202 
203 	/* Init the ipv4 part of the socket since we can have sockets
204 	 * using v6 API for ipv4.
205 	 */
206 	inet->uc_ttl	= -1;
207 
208 	inet->mc_loop	= 1;
209 	inet->mc_ttl	= 1;
210 	inet->mc_index	= 0;
211 	inet->mc_list	= NULL;
212 	inet->rcv_tos	= 0;
213 
214 	if (net->ipv4.sysctl_ip_no_pmtu_disc)
215 		inet->pmtudisc = IP_PMTUDISC_DONT;
216 	else
217 		inet->pmtudisc = IP_PMTUDISC_WANT;
218 	/*
219 	 * Increment only the relevant sk_prot->socks debug field, this changes
220 	 * the previous behaviour of incrementing both the equivalent to
221 	 * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
222 	 *
223 	 * This allows better debug granularity as we'll know exactly how many
224 	 * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
225 	 * transport protocol socks. -acme
226 	 */
227 	sk_refcnt_debug_inc(sk);
228 
229 	if (inet->inet_num) {
230 		/* It assumes that any protocol which allows
231 		 * the user to assign a number at socket
232 		 * creation time automatically shares.
233 		 */
234 		inet->inet_sport = htons(inet->inet_num);
235 		sk->sk_prot->hash(sk);
236 	}
237 	if (sk->sk_prot->init) {
238 		err = sk->sk_prot->init(sk);
239 		if (err) {
240 			sk_common_release(sk);
241 			goto out;
242 		}
243 	}
244 out:
245 	return err;
246 out_rcu_unlock:
247 	rcu_read_unlock();
248 	goto out;
249 }
250 
251 
252 /* bind for INET6 API */
253 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
254 {
255 	struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr;
256 	struct sock *sk = sock->sk;
257 	struct inet_sock *inet = inet_sk(sk);
258 	struct ipv6_pinfo *np = inet6_sk(sk);
259 	struct net *net = sock_net(sk);
260 	__be32 v4addr = 0;
261 	unsigned short snum;
262 	int addr_type = 0;
263 	int err = 0;
264 
265 	/* If the socket has its own bind function then use it. */
266 	if (sk->sk_prot->bind)
267 		return sk->sk_prot->bind(sk, uaddr, addr_len);
268 
269 	if (addr_len < SIN6_LEN_RFC2133)
270 		return -EINVAL;
271 
272 	if (addr->sin6_family != AF_INET6)
273 		return -EAFNOSUPPORT;
274 
275 	addr_type = ipv6_addr_type(&addr->sin6_addr);
276 	if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
277 		return -EINVAL;
278 
279 	snum = ntohs(addr->sin6_port);
280 	if (snum && snum < PROT_SOCK && !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
281 		return -EACCES;
282 
283 	lock_sock(sk);
284 
285 	/* Check these errors (active socket, double bind). */
286 	if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
287 		err = -EINVAL;
288 		goto out;
289 	}
290 
291 	/* Check if the address belongs to the host. */
292 	if (addr_type == IPV6_ADDR_MAPPED) {
293 		int chk_addr_ret;
294 
295 		/* Binding to v4-mapped address on a v6-only socket
296 		 * makes no sense
297 		 */
298 		if (sk->sk_ipv6only) {
299 			err = -EINVAL;
300 			goto out;
301 		}
302 
303 		/* Reproduce AF_INET checks to make the bindings consistent */
304 		v4addr = addr->sin6_addr.s6_addr32[3];
305 		chk_addr_ret = inet_addr_type(net, v4addr);
306 		if (!net->ipv4.sysctl_ip_nonlocal_bind &&
307 		    !(inet->freebind || inet->transparent) &&
308 		    v4addr != htonl(INADDR_ANY) &&
309 		    chk_addr_ret != RTN_LOCAL &&
310 		    chk_addr_ret != RTN_MULTICAST &&
311 		    chk_addr_ret != RTN_BROADCAST) {
312 			err = -EADDRNOTAVAIL;
313 			goto out;
314 		}
315 	} else {
316 		if (addr_type != IPV6_ADDR_ANY) {
317 			struct net_device *dev = NULL;
318 
319 			rcu_read_lock();
320 			if (__ipv6_addr_needs_scope_id(addr_type)) {
321 				if (addr_len >= sizeof(struct sockaddr_in6) &&
322 				    addr->sin6_scope_id) {
323 					/* Override any existing binding, if another one
324 					 * is supplied by user.
325 					 */
326 					sk->sk_bound_dev_if = addr->sin6_scope_id;
327 				}
328 
329 				/* Binding to link-local address requires an interface */
330 				if (!sk->sk_bound_dev_if) {
331 					err = -EINVAL;
332 					goto out_unlock;
333 				}
334 				dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
335 				if (!dev) {
336 					err = -ENODEV;
337 					goto out_unlock;
338 				}
339 			}
340 
341 			/* ipv4 addr of the socket is invalid.  Only the
342 			 * unspecified and mapped address have a v4 equivalent.
343 			 */
344 			v4addr = LOOPBACK4_IPV6;
345 			if (!(addr_type & IPV6_ADDR_MULTICAST))	{
346 				if (!net->ipv6.sysctl.ip_nonlocal_bind &&
347 				    !(inet->freebind || inet->transparent) &&
348 				    !ipv6_chk_addr(net, &addr->sin6_addr,
349 						   dev, 0)) {
350 					err = -EADDRNOTAVAIL;
351 					goto out_unlock;
352 				}
353 			}
354 			rcu_read_unlock();
355 		}
356 	}
357 
358 	inet->inet_rcv_saddr = v4addr;
359 	inet->inet_saddr = v4addr;
360 
361 	sk->sk_v6_rcv_saddr = addr->sin6_addr;
362 
363 	if (!(addr_type & IPV6_ADDR_MULTICAST))
364 		np->saddr = addr->sin6_addr;
365 
366 	/* Make sure we are allowed to bind here. */
367 	if ((snum || !inet->bind_address_no_port) &&
368 	    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 			sk->sk_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)
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)
420 		kfree_skb(skb);
421 
422 	skb = xchg(&np->rxpmtu, NULL);
423 	if (skb)
424 		kfree_skb(skb);
425 
426 	/* Free flowlabels */
427 	fl6_free_socklist(sk);
428 
429 	/* Free tx options */
430 
431 	opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL);
432 	if (opt) {
433 		atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
434 		txopt_put(opt);
435 	}
436 }
437 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
438 
439 /*
440  *	This does both peername and sockname.
441  */
442 
443 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
444 		 int *uaddr_len, int peer)
445 {
446 	struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
447 	struct sock *sk = sock->sk;
448 	struct inet_sock *inet = inet_sk(sk);
449 	struct ipv6_pinfo *np = inet6_sk(sk);
450 
451 	sin->sin6_family = AF_INET6;
452 	sin->sin6_flowinfo = 0;
453 	sin->sin6_scope_id = 0;
454 	if (peer) {
455 		if (!inet->inet_dport)
456 			return -ENOTCONN;
457 		if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
458 		    peer == 1)
459 			return -ENOTCONN;
460 		sin->sin6_port = inet->inet_dport;
461 		sin->sin6_addr = sk->sk_v6_daddr;
462 		if (np->sndflow)
463 			sin->sin6_flowinfo = np->flow_label;
464 	} else {
465 		if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
466 			sin->sin6_addr = np->saddr;
467 		else
468 			sin->sin6_addr = sk->sk_v6_rcv_saddr;
469 
470 		sin->sin6_port = inet->inet_sport;
471 	}
472 	sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
473 						 sk->sk_bound_dev_if);
474 	*uaddr_len = sizeof(*sin);
475 	return 0;
476 }
477 EXPORT_SYMBOL(inet6_getname);
478 
479 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
480 {
481 	struct sock *sk = sock->sk;
482 	struct net *net = sock_net(sk);
483 
484 	switch (cmd) {
485 	case SIOCGSTAMP:
486 		return sock_get_timestamp(sk, (struct timeval __user *)arg);
487 
488 	case SIOCGSTAMPNS:
489 		return sock_get_timestampns(sk, (struct timespec __user *)arg);
490 
491 	case SIOCADDRT:
492 	case SIOCDELRT:
493 
494 		return ipv6_route_ioctl(net, cmd, (void __user *)arg);
495 
496 	case SIOCSIFADDR:
497 		return addrconf_add_ifaddr(net, (void __user *) arg);
498 	case SIOCDIFADDR:
499 		return addrconf_del_ifaddr(net, (void __user *) arg);
500 	case SIOCSIFDSTADDR:
501 		return addrconf_set_dstaddr(net, (void __user *) arg);
502 	default:
503 		if (!sk->sk_prot->ioctl)
504 			return -ENOIOCTLCMD;
505 		return sk->sk_prot->ioctl(sk, cmd, arg);
506 	}
507 	/*NOTREACHED*/
508 	return 0;
509 }
510 EXPORT_SYMBOL(inet6_ioctl);
511 
512 const struct proto_ops inet6_stream_ops = {
513 	.family		   = PF_INET6,
514 	.owner		   = THIS_MODULE,
515 	.release	   = inet6_release,
516 	.bind		   = inet6_bind,
517 	.connect	   = inet_stream_connect,	/* ok		*/
518 	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
519 	.accept		   = inet_accept,		/* ok		*/
520 	.getname	   = inet6_getname,
521 	.poll		   = tcp_poll,			/* ok		*/
522 	.ioctl		   = inet6_ioctl,		/* must change  */
523 	.listen		   = inet_listen,		/* ok		*/
524 	.shutdown	   = inet_shutdown,		/* ok		*/
525 	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
526 	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
527 	.sendmsg	   = inet_sendmsg,		/* ok		*/
528 	.recvmsg	   = inet_recvmsg,		/* ok		*/
529 	.mmap		   = sock_no_mmap,
530 	.sendpage	   = inet_sendpage,
531 	.splice_read	   = tcp_splice_read,
532 #ifdef CONFIG_COMPAT
533 	.compat_setsockopt = compat_sock_common_setsockopt,
534 	.compat_getsockopt = compat_sock_common_getsockopt,
535 #endif
536 };
537 
538 const struct proto_ops inet6_dgram_ops = {
539 	.family		   = PF_INET6,
540 	.owner		   = THIS_MODULE,
541 	.release	   = inet6_release,
542 	.bind		   = inet6_bind,
543 	.connect	   = inet_dgram_connect,	/* ok		*/
544 	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
545 	.accept		   = sock_no_accept,		/* a do nothing	*/
546 	.getname	   = inet6_getname,
547 	.poll		   = udp_poll,			/* ok		*/
548 	.ioctl		   = inet6_ioctl,		/* must change  */
549 	.listen		   = sock_no_listen,		/* ok		*/
550 	.shutdown	   = inet_shutdown,		/* ok		*/
551 	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
552 	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
553 	.sendmsg	   = inet_sendmsg,		/* ok		*/
554 	.recvmsg	   = inet_recvmsg,		/* ok		*/
555 	.mmap		   = sock_no_mmap,
556 	.sendpage	   = sock_no_sendpage,
557 #ifdef CONFIG_COMPAT
558 	.compat_setsockopt = compat_sock_common_setsockopt,
559 	.compat_getsockopt = compat_sock_common_getsockopt,
560 #endif
561 };
562 
563 static const struct net_proto_family inet6_family_ops = {
564 	.family = PF_INET6,
565 	.create = inet6_create,
566 	.owner	= THIS_MODULE,
567 };
568 
569 int inet6_register_protosw(struct inet_protosw *p)
570 {
571 	struct list_head *lh;
572 	struct inet_protosw *answer;
573 	struct list_head *last_perm;
574 	int protocol = p->protocol;
575 	int ret;
576 
577 	spin_lock_bh(&inetsw6_lock);
578 
579 	ret = -EINVAL;
580 	if (p->type >= SOCK_MAX)
581 		goto out_illegal;
582 
583 	/* If we are trying to override a permanent protocol, bail. */
584 	answer = NULL;
585 	ret = -EPERM;
586 	last_perm = &inetsw6[p->type];
587 	list_for_each(lh, &inetsw6[p->type]) {
588 		answer = list_entry(lh, struct inet_protosw, list);
589 
590 		/* Check only the non-wild match. */
591 		if (INET_PROTOSW_PERMANENT & answer->flags) {
592 			if (protocol == answer->protocol)
593 				break;
594 			last_perm = lh;
595 		}
596 
597 		answer = NULL;
598 	}
599 	if (answer)
600 		goto out_permanent;
601 
602 	/* Add the new entry after the last permanent entry if any, so that
603 	 * the new entry does not override a permanent entry when matched with
604 	 * a wild-card protocol. But it is allowed to override any existing
605 	 * non-permanent entry.  This means that when we remove this entry, the
606 	 * system automatically returns to the old behavior.
607 	 */
608 	list_add_rcu(&p->list, last_perm);
609 	ret = 0;
610 out:
611 	spin_unlock_bh(&inetsw6_lock);
612 	return ret;
613 
614 out_permanent:
615 	pr_err("Attempt to override permanent protocol %d\n", protocol);
616 	goto out;
617 
618 out_illegal:
619 	pr_err("Ignoring attempt to register invalid socket type %d\n",
620 	       p->type);
621 	goto out;
622 }
623 EXPORT_SYMBOL(inet6_register_protosw);
624 
625 void
626 inet6_unregister_protosw(struct inet_protosw *p)
627 {
628 	if (INET_PROTOSW_PERMANENT & p->flags) {
629 		pr_err("Attempt to unregister permanent protocol %d\n",
630 		       p->protocol);
631 	} else {
632 		spin_lock_bh(&inetsw6_lock);
633 		list_del_rcu(&p->list);
634 		spin_unlock_bh(&inetsw6_lock);
635 
636 		synchronize_net();
637 	}
638 }
639 EXPORT_SYMBOL(inet6_unregister_protosw);
640 
641 int inet6_sk_rebuild_header(struct sock *sk)
642 {
643 	struct ipv6_pinfo *np = inet6_sk(sk);
644 	struct dst_entry *dst;
645 
646 	dst = __sk_dst_check(sk, np->dst_cookie);
647 
648 	if (!dst) {
649 		struct inet_sock *inet = inet_sk(sk);
650 		struct in6_addr *final_p, final;
651 		struct flowi6 fl6;
652 
653 		memset(&fl6, 0, sizeof(fl6));
654 		fl6.flowi6_proto = sk->sk_protocol;
655 		fl6.daddr = sk->sk_v6_daddr;
656 		fl6.saddr = np->saddr;
657 		fl6.flowlabel = np->flow_label;
658 		fl6.flowi6_oif = sk->sk_bound_dev_if;
659 		fl6.flowi6_mark = sk->sk_mark;
660 		fl6.fl6_dport = inet->inet_dport;
661 		fl6.fl6_sport = inet->inet_sport;
662 		security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
663 
664 		rcu_read_lock();
665 		final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt),
666 					 &final);
667 		rcu_read_unlock();
668 
669 		dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
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 		       const struct inet6_skb_parm *opt)
685 {
686 	const struct ipv6_pinfo *np = inet6_sk(sk);
687 
688 	if (np->rxopt.all) {
689 		if (((opt->flags & IP6SKB_HOPBYHOP) &&
690 		     (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) ||
691 		    (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) &&
692 		     np->rxopt.bits.rxflow) ||
693 		    (opt->srcrt && (np->rxopt.bits.srcrt ||
694 		     np->rxopt.bits.osrcrt)) ||
695 		    ((opt->dst1 || opt->dst0) &&
696 		     (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
697 			return true;
698 	}
699 	return false;
700 }
701 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
702 
703 static struct packet_type ipv6_packet_type __read_mostly = {
704 	.type = cpu_to_be16(ETH_P_IPV6),
705 	.func = ipv6_rcv,
706 };
707 
708 static int __init ipv6_packet_init(void)
709 {
710 	dev_add_pack(&ipv6_packet_type);
711 	return 0;
712 }
713 
714 static void ipv6_packet_cleanup(void)
715 {
716 	dev_remove_pack(&ipv6_packet_type);
717 }
718 
719 static int __net_init ipv6_init_mibs(struct net *net)
720 {
721 	int i;
722 
723 	net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib);
724 	if (!net->mib.udp_stats_in6)
725 		return -ENOMEM;
726 	net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib);
727 	if (!net->mib.udplite_stats_in6)
728 		goto err_udplite_mib;
729 	net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib);
730 	if (!net->mib.ipv6_statistics)
731 		goto err_ip_mib;
732 
733 	for_each_possible_cpu(i) {
734 		struct ipstats_mib *af_inet6_stats;
735 		af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i);
736 		u64_stats_init(&af_inet6_stats->syncp);
737 	}
738 
739 
740 	net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib);
741 	if (!net->mib.icmpv6_statistics)
742 		goto err_icmp_mib;
743 	net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
744 						GFP_KERNEL);
745 	if (!net->mib.icmpv6msg_statistics)
746 		goto err_icmpmsg_mib;
747 	return 0;
748 
749 err_icmpmsg_mib:
750 	free_percpu(net->mib.icmpv6_statistics);
751 err_icmp_mib:
752 	free_percpu(net->mib.ipv6_statistics);
753 err_ip_mib:
754 	free_percpu(net->mib.udplite_stats_in6);
755 err_udplite_mib:
756 	free_percpu(net->mib.udp_stats_in6);
757 	return -ENOMEM;
758 }
759 
760 static void ipv6_cleanup_mibs(struct net *net)
761 {
762 	free_percpu(net->mib.udp_stats_in6);
763 	free_percpu(net->mib.udplite_stats_in6);
764 	free_percpu(net->mib.ipv6_statistics);
765 	free_percpu(net->mib.icmpv6_statistics);
766 	kfree(net->mib.icmpv6msg_statistics);
767 }
768 
769 static int __net_init inet6_net_init(struct net *net)
770 {
771 	int err = 0;
772 
773 	net->ipv6.sysctl.bindv6only = 0;
774 	net->ipv6.sysctl.icmpv6_time = 1*HZ;
775 	net->ipv6.sysctl.flowlabel_consistency = 1;
776 	net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS;
777 	net->ipv6.sysctl.idgen_retries = 3;
778 	net->ipv6.sysctl.idgen_delay = 1 * HZ;
779 	net->ipv6.sysctl.flowlabel_state_ranges = 0;
780 	atomic_set(&net->ipv6.fib6_sernum, 1);
781 
782 	err = ipv6_init_mibs(net);
783 	if (err)
784 		return err;
785 #ifdef CONFIG_PROC_FS
786 	err = udp6_proc_init(net);
787 	if (err)
788 		goto out;
789 	err = tcp6_proc_init(net);
790 	if (err)
791 		goto proc_tcp6_fail;
792 	err = ac6_proc_init(net);
793 	if (err)
794 		goto proc_ac6_fail;
795 #endif
796 	return err;
797 
798 #ifdef CONFIG_PROC_FS
799 proc_ac6_fail:
800 	tcp6_proc_exit(net);
801 proc_tcp6_fail:
802 	udp6_proc_exit(net);
803 out:
804 	ipv6_cleanup_mibs(net);
805 	return err;
806 #endif
807 }
808 
809 static void __net_exit inet6_net_exit(struct net *net)
810 {
811 #ifdef CONFIG_PROC_FS
812 	udp6_proc_exit(net);
813 	tcp6_proc_exit(net);
814 	ac6_proc_exit(net);
815 #endif
816 	ipv6_cleanup_mibs(net);
817 }
818 
819 static struct pernet_operations inet6_net_ops = {
820 	.init = inet6_net_init,
821 	.exit = inet6_net_exit,
822 };
823 
824 static const struct ipv6_stub ipv6_stub_impl = {
825 	.ipv6_sock_mc_join = ipv6_sock_mc_join,
826 	.ipv6_sock_mc_drop = ipv6_sock_mc_drop,
827 	.ipv6_dst_lookup = ip6_dst_lookup,
828 	.udpv6_encap_enable = udpv6_encap_enable,
829 	.ndisc_send_na = ndisc_send_na,
830 	.nd_tbl	= &nd_tbl,
831 };
832 
833 static int __init inet6_init(void)
834 {
835 	struct list_head *r;
836 	int err = 0;
837 
838 	sock_skb_cb_check_size(sizeof(struct inet6_skb_parm));
839 
840 	/* Register the socket-side information for inet6_create.  */
841 	for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
842 		INIT_LIST_HEAD(r);
843 
844 	if (disable_ipv6_mod) {
845 		pr_info("Loaded, but administratively disabled, reboot required to enable\n");
846 		goto out;
847 	}
848 
849 	err = proto_register(&tcpv6_prot, 1);
850 	if (err)
851 		goto out;
852 
853 	err = proto_register(&udpv6_prot, 1);
854 	if (err)
855 		goto out_unregister_tcp_proto;
856 
857 	err = proto_register(&udplitev6_prot, 1);
858 	if (err)
859 		goto out_unregister_udp_proto;
860 
861 	err = proto_register(&rawv6_prot, 1);
862 	if (err)
863 		goto out_unregister_udplite_proto;
864 
865 	err = proto_register(&pingv6_prot, 1);
866 	if (err)
867 		goto out_unregister_ping_proto;
868 
869 	/* We MUST register RAW sockets before we create the ICMP6,
870 	 * IGMP6, or NDISC control sockets.
871 	 */
872 	err = rawv6_init();
873 	if (err)
874 		goto out_unregister_raw_proto;
875 
876 	/* Register the family here so that the init calls below will
877 	 * be able to create sockets. (?? is this dangerous ??)
878 	 */
879 	err = sock_register(&inet6_family_ops);
880 	if (err)
881 		goto out_sock_register_fail;
882 
883 	/*
884 	 *	ipngwg API draft makes clear that the correct semantics
885 	 *	for TCP and UDP is to consider one TCP and UDP instance
886 	 *	in a host available by both INET and INET6 APIs and
887 	 *	able to communicate via both network protocols.
888 	 */
889 
890 	err = register_pernet_subsys(&inet6_net_ops);
891 	if (err)
892 		goto register_pernet_fail;
893 	err = icmpv6_init();
894 	if (err)
895 		goto icmp_fail;
896 	err = ip6_mr_init();
897 	if (err)
898 		goto ipmr_fail;
899 	err = ndisc_init();
900 	if (err)
901 		goto ndisc_fail;
902 	err = igmp6_init();
903 	if (err)
904 		goto igmp_fail;
905 
906 	ipv6_stub = &ipv6_stub_impl;
907 
908 	err = ipv6_netfilter_init();
909 	if (err)
910 		goto netfilter_fail;
911 	/* Create /proc/foo6 entries. */
912 #ifdef CONFIG_PROC_FS
913 	err = -ENOMEM;
914 	if (raw6_proc_init())
915 		goto proc_raw6_fail;
916 	if (udplite6_proc_init())
917 		goto proc_udplite6_fail;
918 	if (ipv6_misc_proc_init())
919 		goto proc_misc6_fail;
920 	if (if6_proc_init())
921 		goto proc_if6_fail;
922 #endif
923 	err = ip6_route_init();
924 	if (err)
925 		goto ip6_route_fail;
926 	err = ndisc_late_init();
927 	if (err)
928 		goto ndisc_late_fail;
929 	err = ip6_flowlabel_init();
930 	if (err)
931 		goto ip6_flowlabel_fail;
932 	err = addrconf_init();
933 	if (err)
934 		goto addrconf_fail;
935 
936 	/* Init v6 extension headers. */
937 	err = ipv6_exthdrs_init();
938 	if (err)
939 		goto ipv6_exthdrs_fail;
940 
941 	err = ipv6_frag_init();
942 	if (err)
943 		goto ipv6_frag_fail;
944 
945 	/* Init v6 transport protocols. */
946 	err = udpv6_init();
947 	if (err)
948 		goto udpv6_fail;
949 
950 	err = udplitev6_init();
951 	if (err)
952 		goto udplitev6_fail;
953 
954 	err = tcpv6_init();
955 	if (err)
956 		goto tcpv6_fail;
957 
958 	err = ipv6_packet_init();
959 	if (err)
960 		goto ipv6_packet_fail;
961 
962 	err = pingv6_init();
963 	if (err)
964 		goto pingv6_fail;
965 
966 #ifdef CONFIG_SYSCTL
967 	err = ipv6_sysctl_register();
968 	if (err)
969 		goto sysctl_fail;
970 #endif
971 out:
972 	return err;
973 
974 #ifdef CONFIG_SYSCTL
975 sysctl_fail:
976 	pingv6_exit();
977 #endif
978 pingv6_fail:
979 	ipv6_packet_cleanup();
980 ipv6_packet_fail:
981 	tcpv6_exit();
982 tcpv6_fail:
983 	udplitev6_exit();
984 udplitev6_fail:
985 	udpv6_exit();
986 udpv6_fail:
987 	ipv6_frag_exit();
988 ipv6_frag_fail:
989 	ipv6_exthdrs_exit();
990 ipv6_exthdrs_fail:
991 	addrconf_cleanup();
992 addrconf_fail:
993 	ip6_flowlabel_cleanup();
994 ip6_flowlabel_fail:
995 	ndisc_late_cleanup();
996 ndisc_late_fail:
997 	ip6_route_cleanup();
998 ip6_route_fail:
999 #ifdef CONFIG_PROC_FS
1000 	if6_proc_exit();
1001 proc_if6_fail:
1002 	ipv6_misc_proc_exit();
1003 proc_misc6_fail:
1004 	udplite6_proc_exit();
1005 proc_udplite6_fail:
1006 	raw6_proc_exit();
1007 proc_raw6_fail:
1008 #endif
1009 	ipv6_netfilter_fini();
1010 netfilter_fail:
1011 	igmp6_cleanup();
1012 igmp_fail:
1013 	ndisc_cleanup();
1014 ndisc_fail:
1015 	ip6_mr_cleanup();
1016 ipmr_fail:
1017 	icmpv6_cleanup();
1018 icmp_fail:
1019 	unregister_pernet_subsys(&inet6_net_ops);
1020 register_pernet_fail:
1021 	sock_unregister(PF_INET6);
1022 	rtnl_unregister_all(PF_INET6);
1023 out_sock_register_fail:
1024 	rawv6_exit();
1025 out_unregister_ping_proto:
1026 	proto_unregister(&pingv6_prot);
1027 out_unregister_raw_proto:
1028 	proto_unregister(&rawv6_prot);
1029 out_unregister_udplite_proto:
1030 	proto_unregister(&udplitev6_prot);
1031 out_unregister_udp_proto:
1032 	proto_unregister(&udpv6_prot);
1033 out_unregister_tcp_proto:
1034 	proto_unregister(&tcpv6_prot);
1035 	goto out;
1036 }
1037 module_init(inet6_init);
1038 
1039 MODULE_ALIAS_NETPROTO(PF_INET6);
1040