xref: /openbmc/linux/net/ipv6/af_inet6.c (revision abfbd895)
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 	if (protocol < 0 || protocol >= IPPROTO_MAX)
113 		return -EINVAL;
114 
115 	/* Look for the requested type/protocol pair. */
116 lookup_protocol:
117 	err = -ESOCKTNOSUPPORT;
118 	rcu_read_lock();
119 	list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
120 
121 		err = 0;
122 		/* Check the non-wild match. */
123 		if (protocol == answer->protocol) {
124 			if (protocol != IPPROTO_IP)
125 				break;
126 		} else {
127 			/* Check for the two wild cases. */
128 			if (IPPROTO_IP == protocol) {
129 				protocol = answer->protocol;
130 				break;
131 			}
132 			if (IPPROTO_IP == answer->protocol)
133 				break;
134 		}
135 		err = -EPROTONOSUPPORT;
136 	}
137 
138 	if (err) {
139 		if (try_loading_module < 2) {
140 			rcu_read_unlock();
141 			/*
142 			 * Be more specific, e.g. net-pf-10-proto-132-type-1
143 			 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
144 			 */
145 			if (++try_loading_module == 1)
146 				request_module("net-pf-%d-proto-%d-type-%d",
147 						PF_INET6, protocol, sock->type);
148 			/*
149 			 * Fall back to generic, e.g. net-pf-10-proto-132
150 			 * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
151 			 */
152 			else
153 				request_module("net-pf-%d-proto-%d",
154 						PF_INET6, protocol);
155 			goto lookup_protocol;
156 		} else
157 			goto out_rcu_unlock;
158 	}
159 
160 	err = -EPERM;
161 	if (sock->type == SOCK_RAW && !kern &&
162 	    !ns_capable(net->user_ns, CAP_NET_RAW))
163 		goto out_rcu_unlock;
164 
165 	sock->ops = answer->ops;
166 	answer_prot = answer->prot;
167 	answer_flags = answer->flags;
168 	rcu_read_unlock();
169 
170 	WARN_ON(!answer_prot->slab);
171 
172 	err = -ENOBUFS;
173 	sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot, kern);
174 	if (!sk)
175 		goto out;
176 
177 	sock_init_data(sock, sk);
178 
179 	err = 0;
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->autoflowlabel = ip6_default_np_autolabel(sock_net(sk));
204 	sk->sk_ipv6only	= net->ipv6.sysctl.bindv6only;
205 
206 	/* Init the ipv4 part of the socket since we can have sockets
207 	 * using v6 API for ipv4.
208 	 */
209 	inet->uc_ttl	= -1;
210 
211 	inet->mc_loop	= 1;
212 	inet->mc_ttl	= 1;
213 	inet->mc_index	= 0;
214 	inet->mc_list	= NULL;
215 	inet->rcv_tos	= 0;
216 
217 	if (net->ipv4.sysctl_ip_no_pmtu_disc)
218 		inet->pmtudisc = IP_PMTUDISC_DONT;
219 	else
220 		inet->pmtudisc = IP_PMTUDISC_WANT;
221 	/*
222 	 * Increment only the relevant sk_prot->socks debug field, this changes
223 	 * the previous behaviour of incrementing both the equivalent to
224 	 * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
225 	 *
226 	 * This allows better debug granularity as we'll know exactly how many
227 	 * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
228 	 * transport protocol socks. -acme
229 	 */
230 	sk_refcnt_debug_inc(sk);
231 
232 	if (inet->inet_num) {
233 		/* It assumes that any protocol which allows
234 		 * the user to assign a number at socket
235 		 * creation time automatically shares.
236 		 */
237 		inet->inet_sport = htons(inet->inet_num);
238 		sk->sk_prot->hash(sk);
239 	}
240 	if (sk->sk_prot->init) {
241 		err = sk->sk_prot->init(sk);
242 		if (err) {
243 			sk_common_release(sk);
244 			goto out;
245 		}
246 	}
247 out:
248 	return err;
249 out_rcu_unlock:
250 	rcu_read_unlock();
251 	goto out;
252 }
253 
254 
255 /* bind for INET6 API */
256 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
257 {
258 	struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr;
259 	struct sock *sk = sock->sk;
260 	struct inet_sock *inet = inet_sk(sk);
261 	struct ipv6_pinfo *np = inet6_sk(sk);
262 	struct net *net = sock_net(sk);
263 	__be32 v4addr = 0;
264 	unsigned short snum;
265 	int addr_type = 0;
266 	int err = 0;
267 
268 	/* If the socket has its own bind function then use it. */
269 	if (sk->sk_prot->bind)
270 		return sk->sk_prot->bind(sk, uaddr, addr_len);
271 
272 	if (addr_len < SIN6_LEN_RFC2133)
273 		return -EINVAL;
274 
275 	if (addr->sin6_family != AF_INET6)
276 		return -EAFNOSUPPORT;
277 
278 	addr_type = ipv6_addr_type(&addr->sin6_addr);
279 	if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
280 		return -EINVAL;
281 
282 	snum = ntohs(addr->sin6_port);
283 	if (snum && snum < PROT_SOCK && !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
284 		return -EACCES;
285 
286 	lock_sock(sk);
287 
288 	/* Check these errors (active socket, double bind). */
289 	if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
290 		err = -EINVAL;
291 		goto out;
292 	}
293 
294 	/* Check if the address belongs to the host. */
295 	if (addr_type == IPV6_ADDR_MAPPED) {
296 		int chk_addr_ret;
297 
298 		/* Binding to v4-mapped address on a v6-only socket
299 		 * makes no sense
300 		 */
301 		if (sk->sk_ipv6only) {
302 			err = -EINVAL;
303 			goto out;
304 		}
305 
306 		/* Reproduce AF_INET checks to make the bindings consistent */
307 		v4addr = addr->sin6_addr.s6_addr32[3];
308 		chk_addr_ret = inet_addr_type(net, v4addr);
309 		if (!net->ipv4.sysctl_ip_nonlocal_bind &&
310 		    !(inet->freebind || inet->transparent) &&
311 		    v4addr != htonl(INADDR_ANY) &&
312 		    chk_addr_ret != RTN_LOCAL &&
313 		    chk_addr_ret != RTN_MULTICAST &&
314 		    chk_addr_ret != RTN_BROADCAST) {
315 			err = -EADDRNOTAVAIL;
316 			goto out;
317 		}
318 	} else {
319 		if (addr_type != IPV6_ADDR_ANY) {
320 			struct net_device *dev = NULL;
321 
322 			rcu_read_lock();
323 			if (__ipv6_addr_needs_scope_id(addr_type)) {
324 				if (addr_len >= sizeof(struct sockaddr_in6) &&
325 				    addr->sin6_scope_id) {
326 					/* Override any existing binding, if another one
327 					 * is supplied by user.
328 					 */
329 					sk->sk_bound_dev_if = addr->sin6_scope_id;
330 				}
331 
332 				/* Binding to link-local address requires an interface */
333 				if (!sk->sk_bound_dev_if) {
334 					err = -EINVAL;
335 					goto out_unlock;
336 				}
337 				dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
338 				if (!dev) {
339 					err = -ENODEV;
340 					goto out_unlock;
341 				}
342 			}
343 
344 			/* ipv4 addr of the socket is invalid.  Only the
345 			 * unspecified and mapped address have a v4 equivalent.
346 			 */
347 			v4addr = LOOPBACK4_IPV6;
348 			if (!(addr_type & IPV6_ADDR_MULTICAST))	{
349 				if (!net->ipv6.sysctl.ip_nonlocal_bind &&
350 				    !(inet->freebind || inet->transparent) &&
351 				    !ipv6_chk_addr(net, &addr->sin6_addr,
352 						   dev, 0)) {
353 					err = -EADDRNOTAVAIL;
354 					goto out_unlock;
355 				}
356 			}
357 			rcu_read_unlock();
358 		}
359 	}
360 
361 	inet->inet_rcv_saddr = v4addr;
362 	inet->inet_saddr = v4addr;
363 
364 	sk->sk_v6_rcv_saddr = addr->sin6_addr;
365 
366 	if (!(addr_type & IPV6_ADDR_MULTICAST))
367 		np->saddr = addr->sin6_addr;
368 
369 	/* Make sure we are allowed to bind here. */
370 	if ((snum || !inet->bind_address_no_port) &&
371 	    sk->sk_prot->get_port(sk, snum)) {
372 		inet_reset_saddr(sk);
373 		err = -EADDRINUSE;
374 		goto out;
375 	}
376 
377 	if (addr_type != IPV6_ADDR_ANY) {
378 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
379 		if (addr_type != IPV6_ADDR_MAPPED)
380 			sk->sk_ipv6only = 1;
381 	}
382 	if (snum)
383 		sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
384 	inet->inet_sport = htons(inet->inet_num);
385 	inet->inet_dport = 0;
386 	inet->inet_daddr = 0;
387 out:
388 	release_sock(sk);
389 	return err;
390 out_unlock:
391 	rcu_read_unlock();
392 	goto out;
393 }
394 EXPORT_SYMBOL(inet6_bind);
395 
396 int inet6_release(struct socket *sock)
397 {
398 	struct sock *sk = sock->sk;
399 
400 	if (!sk)
401 		return -EINVAL;
402 
403 	/* Free mc lists */
404 	ipv6_sock_mc_close(sk);
405 
406 	/* Free ac lists */
407 	ipv6_sock_ac_close(sk);
408 
409 	return inet_release(sock);
410 }
411 EXPORT_SYMBOL(inet6_release);
412 
413 void inet6_destroy_sock(struct sock *sk)
414 {
415 	struct ipv6_pinfo *np = inet6_sk(sk);
416 	struct sk_buff *skb;
417 	struct ipv6_txoptions *opt;
418 
419 	/* Release rx options */
420 
421 	skb = xchg(&np->pktoptions, NULL);
422 	if (skb)
423 		kfree_skb(skb);
424 
425 	skb = xchg(&np->rxpmtu, NULL);
426 	if (skb)
427 		kfree_skb(skb);
428 
429 	/* Free flowlabels */
430 	fl6_free_socklist(sk);
431 
432 	/* Free tx options */
433 
434 	opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL);
435 	if (opt) {
436 		atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
437 		txopt_put(opt);
438 	}
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 = sk->sk_v6_daddr;
465 		if (np->sndflow)
466 			sin->sin6_flowinfo = np->flow_label;
467 	} else {
468 		if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
469 			sin->sin6_addr = np->saddr;
470 		else
471 			sin->sin6_addr = sk->sk_v6_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) {
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 = sk->sk_v6_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 		rcu_read_lock();
668 		final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt),
669 					 &final);
670 		rcu_read_unlock();
671 
672 		dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
673 		if (IS_ERR(dst)) {
674 			sk->sk_route_caps = 0;
675 			sk->sk_err_soft = -PTR_ERR(dst);
676 			return PTR_ERR(dst);
677 		}
678 
679 		ip6_dst_store(sk, dst, NULL, NULL);
680 	}
681 
682 	return 0;
683 }
684 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
685 
686 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb,
687 		       const struct inet6_skb_parm *opt)
688 {
689 	const struct ipv6_pinfo *np = inet6_sk(sk);
690 
691 	if (np->rxopt.all) {
692 		if (((opt->flags & IP6SKB_HOPBYHOP) &&
693 		     (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) ||
694 		    (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) &&
695 		     np->rxopt.bits.rxflow) ||
696 		    (opt->srcrt && (np->rxopt.bits.srcrt ||
697 		     np->rxopt.bits.osrcrt)) ||
698 		    ((opt->dst1 || opt->dst0) &&
699 		     (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
700 			return true;
701 	}
702 	return false;
703 }
704 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
705 
706 static struct packet_type ipv6_packet_type __read_mostly = {
707 	.type = cpu_to_be16(ETH_P_IPV6),
708 	.func = ipv6_rcv,
709 };
710 
711 static int __init ipv6_packet_init(void)
712 {
713 	dev_add_pack(&ipv6_packet_type);
714 	return 0;
715 }
716 
717 static void ipv6_packet_cleanup(void)
718 {
719 	dev_remove_pack(&ipv6_packet_type);
720 }
721 
722 static int __net_init ipv6_init_mibs(struct net *net)
723 {
724 	int i;
725 
726 	net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib);
727 	if (!net->mib.udp_stats_in6)
728 		return -ENOMEM;
729 	net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib);
730 	if (!net->mib.udplite_stats_in6)
731 		goto err_udplite_mib;
732 	net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib);
733 	if (!net->mib.ipv6_statistics)
734 		goto err_ip_mib;
735 
736 	for_each_possible_cpu(i) {
737 		struct ipstats_mib *af_inet6_stats;
738 		af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i);
739 		u64_stats_init(&af_inet6_stats->syncp);
740 	}
741 
742 
743 	net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib);
744 	if (!net->mib.icmpv6_statistics)
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 	free_percpu(net->mib.icmpv6_statistics);
754 err_icmp_mib:
755 	free_percpu(net->mib.ipv6_statistics);
756 err_ip_mib:
757 	free_percpu(net->mib.udplite_stats_in6);
758 err_udplite_mib:
759 	free_percpu(net->mib.udp_stats_in6);
760 	return -ENOMEM;
761 }
762 
763 static void ipv6_cleanup_mibs(struct net *net)
764 {
765 	free_percpu(net->mib.udp_stats_in6);
766 	free_percpu(net->mib.udplite_stats_in6);
767 	free_percpu(net->mib.ipv6_statistics);
768 	free_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 	net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS;
780 	net->ipv6.sysctl.idgen_retries = 3;
781 	net->ipv6.sysctl.idgen_delay = 1 * HZ;
782 	net->ipv6.sysctl.flowlabel_state_ranges = 0;
783 	atomic_set(&net->ipv6.fib6_sernum, 1);
784 
785 	err = ipv6_init_mibs(net);
786 	if (err)
787 		return err;
788 #ifdef CONFIG_PROC_FS
789 	err = udp6_proc_init(net);
790 	if (err)
791 		goto out;
792 	err = tcp6_proc_init(net);
793 	if (err)
794 		goto proc_tcp6_fail;
795 	err = ac6_proc_init(net);
796 	if (err)
797 		goto proc_ac6_fail;
798 #endif
799 	return err;
800 
801 #ifdef CONFIG_PROC_FS
802 proc_ac6_fail:
803 	tcp6_proc_exit(net);
804 proc_tcp6_fail:
805 	udp6_proc_exit(net);
806 out:
807 	ipv6_cleanup_mibs(net);
808 	return err;
809 #endif
810 }
811 
812 static void __net_exit inet6_net_exit(struct net *net)
813 {
814 #ifdef CONFIG_PROC_FS
815 	udp6_proc_exit(net);
816 	tcp6_proc_exit(net);
817 	ac6_proc_exit(net);
818 #endif
819 	ipv6_cleanup_mibs(net);
820 }
821 
822 static struct pernet_operations inet6_net_ops = {
823 	.init = inet6_net_init,
824 	.exit = inet6_net_exit,
825 };
826 
827 static const struct ipv6_stub ipv6_stub_impl = {
828 	.ipv6_sock_mc_join = ipv6_sock_mc_join,
829 	.ipv6_sock_mc_drop = ipv6_sock_mc_drop,
830 	.ipv6_dst_lookup = ip6_dst_lookup,
831 	.udpv6_encap_enable = udpv6_encap_enable,
832 	.ndisc_send_na = ndisc_send_na,
833 	.nd_tbl	= &nd_tbl,
834 };
835 
836 static int __init inet6_init(void)
837 {
838 	struct list_head *r;
839 	int err = 0;
840 
841 	sock_skb_cb_check_size(sizeof(struct inet6_skb_parm));
842 
843 	/* Register the socket-side information for inet6_create.  */
844 	for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
845 		INIT_LIST_HEAD(r);
846 
847 	if (disable_ipv6_mod) {
848 		pr_info("Loaded, but administratively disabled, reboot required to enable\n");
849 		goto out;
850 	}
851 
852 	err = proto_register(&tcpv6_prot, 1);
853 	if (err)
854 		goto out;
855 
856 	err = proto_register(&udpv6_prot, 1);
857 	if (err)
858 		goto out_unregister_tcp_proto;
859 
860 	err = proto_register(&udplitev6_prot, 1);
861 	if (err)
862 		goto out_unregister_udp_proto;
863 
864 	err = proto_register(&rawv6_prot, 1);
865 	if (err)
866 		goto out_unregister_udplite_proto;
867 
868 	err = proto_register(&pingv6_prot, 1);
869 	if (err)
870 		goto out_unregister_ping_proto;
871 
872 	/* We MUST register RAW sockets before we create the ICMP6,
873 	 * IGMP6, or NDISC control sockets.
874 	 */
875 	err = rawv6_init();
876 	if (err)
877 		goto out_unregister_raw_proto;
878 
879 	/* Register the family here so that the init calls below will
880 	 * be able to create sockets. (?? is this dangerous ??)
881 	 */
882 	err = sock_register(&inet6_family_ops);
883 	if (err)
884 		goto out_sock_register_fail;
885 
886 	/*
887 	 *	ipngwg API draft makes clear that the correct semantics
888 	 *	for TCP and UDP is to consider one TCP and UDP instance
889 	 *	in a host available by both INET and INET6 APIs and
890 	 *	able to communicate via both network protocols.
891 	 */
892 
893 	err = register_pernet_subsys(&inet6_net_ops);
894 	if (err)
895 		goto register_pernet_fail;
896 	err = icmpv6_init();
897 	if (err)
898 		goto icmp_fail;
899 	err = ip6_mr_init();
900 	if (err)
901 		goto ipmr_fail;
902 	err = ndisc_init();
903 	if (err)
904 		goto ndisc_fail;
905 	err = igmp6_init();
906 	if (err)
907 		goto igmp_fail;
908 
909 	ipv6_stub = &ipv6_stub_impl;
910 
911 	err = ipv6_netfilter_init();
912 	if (err)
913 		goto netfilter_fail;
914 	/* Create /proc/foo6 entries. */
915 #ifdef CONFIG_PROC_FS
916 	err = -ENOMEM;
917 	if (raw6_proc_init())
918 		goto proc_raw6_fail;
919 	if (udplite6_proc_init())
920 		goto proc_udplite6_fail;
921 	if (ipv6_misc_proc_init())
922 		goto proc_misc6_fail;
923 	if (if6_proc_init())
924 		goto proc_if6_fail;
925 #endif
926 	err = ip6_route_init();
927 	if (err)
928 		goto ip6_route_fail;
929 	err = ndisc_late_init();
930 	if (err)
931 		goto ndisc_late_fail;
932 	err = ip6_flowlabel_init();
933 	if (err)
934 		goto ip6_flowlabel_fail;
935 	err = addrconf_init();
936 	if (err)
937 		goto addrconf_fail;
938 
939 	/* Init v6 extension headers. */
940 	err = ipv6_exthdrs_init();
941 	if (err)
942 		goto ipv6_exthdrs_fail;
943 
944 	err = ipv6_frag_init();
945 	if (err)
946 		goto ipv6_frag_fail;
947 
948 	/* Init v6 transport protocols. */
949 	err = udpv6_init();
950 	if (err)
951 		goto udpv6_fail;
952 
953 	err = udplitev6_init();
954 	if (err)
955 		goto udplitev6_fail;
956 
957 	err = tcpv6_init();
958 	if (err)
959 		goto tcpv6_fail;
960 
961 	err = ipv6_packet_init();
962 	if (err)
963 		goto ipv6_packet_fail;
964 
965 	err = pingv6_init();
966 	if (err)
967 		goto pingv6_fail;
968 
969 #ifdef CONFIG_SYSCTL
970 	err = ipv6_sysctl_register();
971 	if (err)
972 		goto sysctl_fail;
973 #endif
974 out:
975 	return err;
976 
977 #ifdef CONFIG_SYSCTL
978 sysctl_fail:
979 	pingv6_exit();
980 #endif
981 pingv6_fail:
982 	ipv6_packet_cleanup();
983 ipv6_packet_fail:
984 	tcpv6_exit();
985 tcpv6_fail:
986 	udplitev6_exit();
987 udplitev6_fail:
988 	udpv6_exit();
989 udpv6_fail:
990 	ipv6_frag_exit();
991 ipv6_frag_fail:
992 	ipv6_exthdrs_exit();
993 ipv6_exthdrs_fail:
994 	addrconf_cleanup();
995 addrconf_fail:
996 	ip6_flowlabel_cleanup();
997 ip6_flowlabel_fail:
998 	ndisc_late_cleanup();
999 ndisc_late_fail:
1000 	ip6_route_cleanup();
1001 ip6_route_fail:
1002 #ifdef CONFIG_PROC_FS
1003 	if6_proc_exit();
1004 proc_if6_fail:
1005 	ipv6_misc_proc_exit();
1006 proc_misc6_fail:
1007 	udplite6_proc_exit();
1008 proc_udplite6_fail:
1009 	raw6_proc_exit();
1010 proc_raw6_fail:
1011 #endif
1012 	ipv6_netfilter_fini();
1013 netfilter_fail:
1014 	igmp6_cleanup();
1015 igmp_fail:
1016 	ndisc_cleanup();
1017 ndisc_fail:
1018 	ip6_mr_cleanup();
1019 ipmr_fail:
1020 	icmpv6_cleanup();
1021 icmp_fail:
1022 	unregister_pernet_subsys(&inet6_net_ops);
1023 register_pernet_fail:
1024 	sock_unregister(PF_INET6);
1025 	rtnl_unregister_all(PF_INET6);
1026 out_sock_register_fail:
1027 	rawv6_exit();
1028 out_unregister_ping_proto:
1029 	proto_unregister(&pingv6_prot);
1030 out_unregister_raw_proto:
1031 	proto_unregister(&rawv6_prot);
1032 out_unregister_udplite_proto:
1033 	proto_unregister(&udplitev6_prot);
1034 out_unregister_udp_proto:
1035 	proto_unregister(&udpv6_prot);
1036 out_unregister_tcp_proto:
1037 	proto_unregister(&tcpv6_prot);
1038 	goto out;
1039 }
1040 module_init(inet6_init);
1041 
1042 MODULE_ALIAS_NETPROTO(PF_INET6);
1043