xref: /linux/net/ipv6/af_inet6.c (revision 0883c2c06fb5bcf5b9e008270827e63c09a88c1e)
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 #include "ip6_offload.h"
68 
69 MODULE_AUTHOR("Cast of dozens");
70 MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
71 MODULE_LICENSE("GPL");
72 
73 /* The inetsw6 table contains everything that inet6_create needs to
74  * build a new socket.
75  */
76 static struct list_head inetsw6[SOCK_MAX];
77 static DEFINE_SPINLOCK(inetsw6_lock);
78 
79 struct ipv6_params ipv6_defaults = {
80 	.disable_ipv6 = 0,
81 	.autoconf = 1,
82 };
83 
84 static int disable_ipv6_mod;
85 
86 module_param_named(disable, disable_ipv6_mod, int, 0444);
87 MODULE_PARM_DESC(disable, "Disable IPv6 module such that it is non-functional");
88 
89 module_param_named(disable_ipv6, ipv6_defaults.disable_ipv6, int, 0444);
90 MODULE_PARM_DESC(disable_ipv6, "Disable IPv6 on all interfaces");
91 
92 module_param_named(autoconf, ipv6_defaults.autoconf, int, 0444);
93 MODULE_PARM_DESC(autoconf, "Enable IPv6 address autoconfiguration on all interfaces");
94 
95 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
96 {
97 	const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
98 
99 	return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
100 }
101 
102 static int inet6_create(struct net *net, struct socket *sock, int protocol,
103 			int kern)
104 {
105 	struct inet_sock *inet;
106 	struct ipv6_pinfo *np;
107 	struct sock *sk;
108 	struct inet_protosw *answer;
109 	struct proto *answer_prot;
110 	unsigned char answer_flags;
111 	int try_loading_module = 0;
112 	int err;
113 
114 	if (protocol < 0 || protocol >= IPPROTO_MAX)
115 		return -EINVAL;
116 
117 	/* Look for the requested type/protocol pair. */
118 lookup_protocol:
119 	err = -ESOCKTNOSUPPORT;
120 	rcu_read_lock();
121 	list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
122 
123 		err = 0;
124 		/* Check the non-wild match. */
125 		if (protocol == answer->protocol) {
126 			if (protocol != IPPROTO_IP)
127 				break;
128 		} else {
129 			/* Check for the two wild cases. */
130 			if (IPPROTO_IP == protocol) {
131 				protocol = answer->protocol;
132 				break;
133 			}
134 			if (IPPROTO_IP == answer->protocol)
135 				break;
136 		}
137 		err = -EPROTONOSUPPORT;
138 	}
139 
140 	if (err) {
141 		if (try_loading_module < 2) {
142 			rcu_read_unlock();
143 			/*
144 			 * Be more specific, e.g. net-pf-10-proto-132-type-1
145 			 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
146 			 */
147 			if (++try_loading_module == 1)
148 				request_module("net-pf-%d-proto-%d-type-%d",
149 						PF_INET6, protocol, sock->type);
150 			/*
151 			 * Fall back to generic, e.g. net-pf-10-proto-132
152 			 * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
153 			 */
154 			else
155 				request_module("net-pf-%d-proto-%d",
156 						PF_INET6, protocol);
157 			goto lookup_protocol;
158 		} else
159 			goto out_rcu_unlock;
160 	}
161 
162 	err = -EPERM;
163 	if (sock->type == SOCK_RAW && !kern &&
164 	    !ns_capable(net->user_ns, CAP_NET_RAW))
165 		goto out_rcu_unlock;
166 
167 	sock->ops = answer->ops;
168 	answer_prot = answer->prot;
169 	answer_flags = answer->flags;
170 	rcu_read_unlock();
171 
172 	WARN_ON(!answer_prot->slab);
173 
174 	err = -ENOBUFS;
175 	sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot, kern);
176 	if (!sk)
177 		goto out;
178 
179 	sock_init_data(sock, sk);
180 
181 	err = 0;
182 	if (INET_PROTOSW_REUSE & answer_flags)
183 		sk->sk_reuse = SK_CAN_REUSE;
184 
185 	inet = inet_sk(sk);
186 	inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
187 
188 	if (SOCK_RAW == sock->type) {
189 		inet->inet_num = protocol;
190 		if (IPPROTO_RAW == protocol)
191 			inet->hdrincl = 1;
192 	}
193 
194 	sk->sk_destruct		= inet_sock_destruct;
195 	sk->sk_family		= PF_INET6;
196 	sk->sk_protocol		= protocol;
197 
198 	sk->sk_backlog_rcv	= answer->prot->backlog_rcv;
199 
200 	inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
201 	np->hop_limit	= -1;
202 	np->mcast_hops	= IPV6_DEFAULT_MCASTHOPS;
203 	np->mc_loop	= 1;
204 	np->pmtudisc	= IPV6_PMTUDISC_WANT;
205 	np->autoflowlabel = ip6_default_np_autolabel(sock_net(sk));
206 	sk->sk_ipv6only	= net->ipv6.sysctl.bindv6only;
207 
208 	/* Init the ipv4 part of the socket since we can have sockets
209 	 * using v6 API for ipv4.
210 	 */
211 	inet->uc_ttl	= -1;
212 
213 	inet->mc_loop	= 1;
214 	inet->mc_ttl	= 1;
215 	inet->mc_index	= 0;
216 	inet->mc_list	= NULL;
217 	inet->rcv_tos	= 0;
218 
219 	if (net->ipv4.sysctl_ip_no_pmtu_disc)
220 		inet->pmtudisc = IP_PMTUDISC_DONT;
221 	else
222 		inet->pmtudisc = IP_PMTUDISC_WANT;
223 	/*
224 	 * Increment only the relevant sk_prot->socks debug field, this changes
225 	 * the previous behaviour of incrementing both the equivalent to
226 	 * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
227 	 *
228 	 * This allows better debug granularity as we'll know exactly how many
229 	 * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
230 	 * transport protocol socks. -acme
231 	 */
232 	sk_refcnt_debug_inc(sk);
233 
234 	if (inet->inet_num) {
235 		/* It assumes that any protocol which allows
236 		 * the user to assign a number at socket
237 		 * creation time automatically shares.
238 		 */
239 		inet->inet_sport = htons(inet->inet_num);
240 		err = sk->sk_prot->hash(sk);
241 		if (err) {
242 			sk_common_release(sk);
243 			goto out;
244 		}
245 	}
246 	if (sk->sk_prot->init) {
247 		err = sk->sk_prot->init(sk);
248 		if (err) {
249 			sk_common_release(sk);
250 			goto out;
251 		}
252 	}
253 out:
254 	return err;
255 out_rcu_unlock:
256 	rcu_read_unlock();
257 	goto out;
258 }
259 
260 
261 /* bind for INET6 API */
262 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
263 {
264 	struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr;
265 	struct sock *sk = sock->sk;
266 	struct inet_sock *inet = inet_sk(sk);
267 	struct ipv6_pinfo *np = inet6_sk(sk);
268 	struct net *net = sock_net(sk);
269 	__be32 v4addr = 0;
270 	unsigned short snum;
271 	int addr_type = 0;
272 	int err = 0;
273 
274 	/* If the socket has its own bind function then use it. */
275 	if (sk->sk_prot->bind)
276 		return sk->sk_prot->bind(sk, uaddr, addr_len);
277 
278 	if (addr_len < SIN6_LEN_RFC2133)
279 		return -EINVAL;
280 
281 	if (addr->sin6_family != AF_INET6)
282 		return -EAFNOSUPPORT;
283 
284 	addr_type = ipv6_addr_type(&addr->sin6_addr);
285 	if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
286 		return -EINVAL;
287 
288 	snum = ntohs(addr->sin6_port);
289 	if (snum && snum < PROT_SOCK && !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
290 		return -EACCES;
291 
292 	lock_sock(sk);
293 
294 	/* Check these errors (active socket, double bind). */
295 	if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
296 		err = -EINVAL;
297 		goto out;
298 	}
299 
300 	/* Check if the address belongs to the host. */
301 	if (addr_type == IPV6_ADDR_MAPPED) {
302 		int chk_addr_ret;
303 
304 		/* Binding to v4-mapped address on a v6-only socket
305 		 * makes no sense
306 		 */
307 		if (sk->sk_ipv6only) {
308 			err = -EINVAL;
309 			goto out;
310 		}
311 
312 		/* Reproduce AF_INET checks to make the bindings consistent */
313 		v4addr = addr->sin6_addr.s6_addr32[3];
314 		chk_addr_ret = inet_addr_type(net, v4addr);
315 		if (!net->ipv4.sysctl_ip_nonlocal_bind &&
316 		    !(inet->freebind || inet->transparent) &&
317 		    v4addr != htonl(INADDR_ANY) &&
318 		    chk_addr_ret != RTN_LOCAL &&
319 		    chk_addr_ret != RTN_MULTICAST &&
320 		    chk_addr_ret != RTN_BROADCAST) {
321 			err = -EADDRNOTAVAIL;
322 			goto out;
323 		}
324 	} else {
325 		if (addr_type != IPV6_ADDR_ANY) {
326 			struct net_device *dev = NULL;
327 
328 			rcu_read_lock();
329 			if (__ipv6_addr_needs_scope_id(addr_type)) {
330 				if (addr_len >= sizeof(struct sockaddr_in6) &&
331 				    addr->sin6_scope_id) {
332 					/* Override any existing binding, if another one
333 					 * is supplied by user.
334 					 */
335 					sk->sk_bound_dev_if = addr->sin6_scope_id;
336 				}
337 
338 				/* Binding to link-local address requires an interface */
339 				if (!sk->sk_bound_dev_if) {
340 					err = -EINVAL;
341 					goto out_unlock;
342 				}
343 				dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
344 				if (!dev) {
345 					err = -ENODEV;
346 					goto out_unlock;
347 				}
348 			}
349 
350 			/* ipv4 addr of the socket is invalid.  Only the
351 			 * unspecified and mapped address have a v4 equivalent.
352 			 */
353 			v4addr = LOOPBACK4_IPV6;
354 			if (!(addr_type & IPV6_ADDR_MULTICAST))	{
355 				if (!net->ipv6.sysctl.ip_nonlocal_bind &&
356 				    !(inet->freebind || inet->transparent) &&
357 				    !ipv6_chk_addr(net, &addr->sin6_addr,
358 						   dev, 0)) {
359 					err = -EADDRNOTAVAIL;
360 					goto out_unlock;
361 				}
362 			}
363 			rcu_read_unlock();
364 		}
365 	}
366 
367 	inet->inet_rcv_saddr = v4addr;
368 	inet->inet_saddr = v4addr;
369 
370 	sk->sk_v6_rcv_saddr = addr->sin6_addr;
371 
372 	if (!(addr_type & IPV6_ADDR_MULTICAST))
373 		np->saddr = addr->sin6_addr;
374 
375 	/* Make sure we are allowed to bind here. */
376 	if ((snum || !inet->bind_address_no_port) &&
377 	    sk->sk_prot->get_port(sk, snum)) {
378 		inet_reset_saddr(sk);
379 		err = -EADDRINUSE;
380 		goto out;
381 	}
382 
383 	if (addr_type != IPV6_ADDR_ANY) {
384 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
385 		if (addr_type != IPV6_ADDR_MAPPED)
386 			sk->sk_ipv6only = 1;
387 	}
388 	if (snum)
389 		sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
390 	inet->inet_sport = htons(inet->inet_num);
391 	inet->inet_dport = 0;
392 	inet->inet_daddr = 0;
393 out:
394 	release_sock(sk);
395 	return err;
396 out_unlock:
397 	rcu_read_unlock();
398 	goto out;
399 }
400 EXPORT_SYMBOL(inet6_bind);
401 
402 int inet6_release(struct socket *sock)
403 {
404 	struct sock *sk = sock->sk;
405 
406 	if (!sk)
407 		return -EINVAL;
408 
409 	/* Free mc lists */
410 	ipv6_sock_mc_close(sk);
411 
412 	/* Free ac lists */
413 	ipv6_sock_ac_close(sk);
414 
415 	return inet_release(sock);
416 }
417 EXPORT_SYMBOL(inet6_release);
418 
419 void inet6_destroy_sock(struct sock *sk)
420 {
421 	struct ipv6_pinfo *np = inet6_sk(sk);
422 	struct sk_buff *skb;
423 	struct ipv6_txoptions *opt;
424 
425 	/* Release rx options */
426 
427 	skb = xchg(&np->pktoptions, NULL);
428 	if (skb)
429 		kfree_skb(skb);
430 
431 	skb = xchg(&np->rxpmtu, NULL);
432 	if (skb)
433 		kfree_skb(skb);
434 
435 	/* Free flowlabels */
436 	fl6_free_socklist(sk);
437 
438 	/* Free tx options */
439 
440 	opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL);
441 	if (opt) {
442 		atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
443 		txopt_put(opt);
444 	}
445 }
446 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
447 
448 /*
449  *	This does both peername and sockname.
450  */
451 
452 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
453 		 int *uaddr_len, int peer)
454 {
455 	struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
456 	struct sock *sk = sock->sk;
457 	struct inet_sock *inet = inet_sk(sk);
458 	struct ipv6_pinfo *np = inet6_sk(sk);
459 
460 	sin->sin6_family = AF_INET6;
461 	sin->sin6_flowinfo = 0;
462 	sin->sin6_scope_id = 0;
463 	if (peer) {
464 		if (!inet->inet_dport)
465 			return -ENOTCONN;
466 		if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
467 		    peer == 1)
468 			return -ENOTCONN;
469 		sin->sin6_port = inet->inet_dport;
470 		sin->sin6_addr = sk->sk_v6_daddr;
471 		if (np->sndflow)
472 			sin->sin6_flowinfo = np->flow_label;
473 	} else {
474 		if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
475 			sin->sin6_addr = np->saddr;
476 		else
477 			sin->sin6_addr = sk->sk_v6_rcv_saddr;
478 
479 		sin->sin6_port = inet->inet_sport;
480 	}
481 	sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
482 						 sk->sk_bound_dev_if);
483 	*uaddr_len = sizeof(*sin);
484 	return 0;
485 }
486 EXPORT_SYMBOL(inet6_getname);
487 
488 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
489 {
490 	struct sock *sk = sock->sk;
491 	struct net *net = sock_net(sk);
492 
493 	switch (cmd) {
494 	case SIOCGSTAMP:
495 		return sock_get_timestamp(sk, (struct timeval __user *)arg);
496 
497 	case SIOCGSTAMPNS:
498 		return sock_get_timestampns(sk, (struct timespec __user *)arg);
499 
500 	case SIOCADDRT:
501 	case SIOCDELRT:
502 
503 		return ipv6_route_ioctl(net, cmd, (void __user *)arg);
504 
505 	case SIOCSIFADDR:
506 		return addrconf_add_ifaddr(net, (void __user *) arg);
507 	case SIOCDIFADDR:
508 		return addrconf_del_ifaddr(net, (void __user *) arg);
509 	case SIOCSIFDSTADDR:
510 		return addrconf_set_dstaddr(net, (void __user *) arg);
511 	default:
512 		if (!sk->sk_prot->ioctl)
513 			return -ENOIOCTLCMD;
514 		return sk->sk_prot->ioctl(sk, cmd, arg);
515 	}
516 	/*NOTREACHED*/
517 	return 0;
518 }
519 EXPORT_SYMBOL(inet6_ioctl);
520 
521 const struct proto_ops inet6_stream_ops = {
522 	.family		   = PF_INET6,
523 	.owner		   = THIS_MODULE,
524 	.release	   = inet6_release,
525 	.bind		   = inet6_bind,
526 	.connect	   = inet_stream_connect,	/* ok		*/
527 	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
528 	.accept		   = inet_accept,		/* ok		*/
529 	.getname	   = inet6_getname,
530 	.poll		   = tcp_poll,			/* ok		*/
531 	.ioctl		   = inet6_ioctl,		/* must change  */
532 	.listen		   = inet_listen,		/* ok		*/
533 	.shutdown	   = inet_shutdown,		/* ok		*/
534 	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
535 	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
536 	.sendmsg	   = inet_sendmsg,		/* ok		*/
537 	.recvmsg	   = inet_recvmsg,		/* ok		*/
538 	.mmap		   = sock_no_mmap,
539 	.sendpage	   = inet_sendpage,
540 	.splice_read	   = tcp_splice_read,
541 #ifdef CONFIG_COMPAT
542 	.compat_setsockopt = compat_sock_common_setsockopt,
543 	.compat_getsockopt = compat_sock_common_getsockopt,
544 #endif
545 };
546 
547 const struct proto_ops inet6_dgram_ops = {
548 	.family		   = PF_INET6,
549 	.owner		   = THIS_MODULE,
550 	.release	   = inet6_release,
551 	.bind		   = inet6_bind,
552 	.connect	   = inet_dgram_connect,	/* ok		*/
553 	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
554 	.accept		   = sock_no_accept,		/* a do nothing	*/
555 	.getname	   = inet6_getname,
556 	.poll		   = udp_poll,			/* ok		*/
557 	.ioctl		   = inet6_ioctl,		/* must change  */
558 	.listen		   = sock_no_listen,		/* ok		*/
559 	.shutdown	   = inet_shutdown,		/* ok		*/
560 	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
561 	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
562 	.sendmsg	   = inet_sendmsg,		/* ok		*/
563 	.recvmsg	   = inet_recvmsg,		/* ok		*/
564 	.mmap		   = sock_no_mmap,
565 	.sendpage	   = sock_no_sendpage,
566 	.set_peek_off	   = sk_set_peek_off,
567 #ifdef CONFIG_COMPAT
568 	.compat_setsockopt = compat_sock_common_setsockopt,
569 	.compat_getsockopt = compat_sock_common_getsockopt,
570 #endif
571 };
572 
573 static const struct net_proto_family inet6_family_ops = {
574 	.family = PF_INET6,
575 	.create = inet6_create,
576 	.owner	= THIS_MODULE,
577 };
578 
579 int inet6_register_protosw(struct inet_protosw *p)
580 {
581 	struct list_head *lh;
582 	struct inet_protosw *answer;
583 	struct list_head *last_perm;
584 	int protocol = p->protocol;
585 	int ret;
586 
587 	spin_lock_bh(&inetsw6_lock);
588 
589 	ret = -EINVAL;
590 	if (p->type >= SOCK_MAX)
591 		goto out_illegal;
592 
593 	/* If we are trying to override a permanent protocol, bail. */
594 	answer = NULL;
595 	ret = -EPERM;
596 	last_perm = &inetsw6[p->type];
597 	list_for_each(lh, &inetsw6[p->type]) {
598 		answer = list_entry(lh, struct inet_protosw, list);
599 
600 		/* Check only the non-wild match. */
601 		if (INET_PROTOSW_PERMANENT & answer->flags) {
602 			if (protocol == answer->protocol)
603 				break;
604 			last_perm = lh;
605 		}
606 
607 		answer = NULL;
608 	}
609 	if (answer)
610 		goto out_permanent;
611 
612 	/* Add the new entry after the last permanent entry if any, so that
613 	 * the new entry does not override a permanent entry when matched with
614 	 * a wild-card protocol. But it is allowed to override any existing
615 	 * non-permanent entry.  This means that when we remove this entry, the
616 	 * system automatically returns to the old behavior.
617 	 */
618 	list_add_rcu(&p->list, last_perm);
619 	ret = 0;
620 out:
621 	spin_unlock_bh(&inetsw6_lock);
622 	return ret;
623 
624 out_permanent:
625 	pr_err("Attempt to override permanent protocol %d\n", protocol);
626 	goto out;
627 
628 out_illegal:
629 	pr_err("Ignoring attempt to register invalid socket type %d\n",
630 	       p->type);
631 	goto out;
632 }
633 EXPORT_SYMBOL(inet6_register_protosw);
634 
635 void
636 inet6_unregister_protosw(struct inet_protosw *p)
637 {
638 	if (INET_PROTOSW_PERMANENT & p->flags) {
639 		pr_err("Attempt to unregister permanent protocol %d\n",
640 		       p->protocol);
641 	} else {
642 		spin_lock_bh(&inetsw6_lock);
643 		list_del_rcu(&p->list);
644 		spin_unlock_bh(&inetsw6_lock);
645 
646 		synchronize_net();
647 	}
648 }
649 EXPORT_SYMBOL(inet6_unregister_protosw);
650 
651 int inet6_sk_rebuild_header(struct sock *sk)
652 {
653 	struct ipv6_pinfo *np = inet6_sk(sk);
654 	struct dst_entry *dst;
655 
656 	dst = __sk_dst_check(sk, np->dst_cookie);
657 
658 	if (!dst) {
659 		struct inet_sock *inet = inet_sk(sk);
660 		struct in6_addr *final_p, final;
661 		struct flowi6 fl6;
662 
663 		memset(&fl6, 0, sizeof(fl6));
664 		fl6.flowi6_proto = sk->sk_protocol;
665 		fl6.daddr = sk->sk_v6_daddr;
666 		fl6.saddr = np->saddr;
667 		fl6.flowlabel = np->flow_label;
668 		fl6.flowi6_oif = sk->sk_bound_dev_if;
669 		fl6.flowi6_mark = sk->sk_mark;
670 		fl6.fl6_dport = inet->inet_dport;
671 		fl6.fl6_sport = inet->inet_sport;
672 		security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
673 
674 		rcu_read_lock();
675 		final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt),
676 					 &final);
677 		rcu_read_unlock();
678 
679 		dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
680 		if (IS_ERR(dst)) {
681 			sk->sk_route_caps = 0;
682 			sk->sk_err_soft = -PTR_ERR(dst);
683 			return PTR_ERR(dst);
684 		}
685 
686 		ip6_dst_store(sk, dst, NULL, NULL);
687 	}
688 
689 	return 0;
690 }
691 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
692 
693 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb,
694 		       const struct inet6_skb_parm *opt)
695 {
696 	const struct ipv6_pinfo *np = inet6_sk(sk);
697 
698 	if (np->rxopt.all) {
699 		if (((opt->flags & IP6SKB_HOPBYHOP) &&
700 		     (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) ||
701 		    (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) &&
702 		     np->rxopt.bits.rxflow) ||
703 		    (opt->srcrt && (np->rxopt.bits.srcrt ||
704 		     np->rxopt.bits.osrcrt)) ||
705 		    ((opt->dst1 || opt->dst0) &&
706 		     (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
707 			return true;
708 	}
709 	return false;
710 }
711 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
712 
713 static struct packet_type ipv6_packet_type __read_mostly = {
714 	.type = cpu_to_be16(ETH_P_IPV6),
715 	.func = ipv6_rcv,
716 };
717 
718 static int __init ipv6_packet_init(void)
719 {
720 	dev_add_pack(&ipv6_packet_type);
721 	return 0;
722 }
723 
724 static void ipv6_packet_cleanup(void)
725 {
726 	dev_remove_pack(&ipv6_packet_type);
727 }
728 
729 static int __net_init ipv6_init_mibs(struct net *net)
730 {
731 	int i;
732 
733 	net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib);
734 	if (!net->mib.udp_stats_in6)
735 		return -ENOMEM;
736 	net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib);
737 	if (!net->mib.udplite_stats_in6)
738 		goto err_udplite_mib;
739 	net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib);
740 	if (!net->mib.ipv6_statistics)
741 		goto err_ip_mib;
742 
743 	for_each_possible_cpu(i) {
744 		struct ipstats_mib *af_inet6_stats;
745 		af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i);
746 		u64_stats_init(&af_inet6_stats->syncp);
747 	}
748 
749 
750 	net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib);
751 	if (!net->mib.icmpv6_statistics)
752 		goto err_icmp_mib;
753 	net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
754 						GFP_KERNEL);
755 	if (!net->mib.icmpv6msg_statistics)
756 		goto err_icmpmsg_mib;
757 	return 0;
758 
759 err_icmpmsg_mib:
760 	free_percpu(net->mib.icmpv6_statistics);
761 err_icmp_mib:
762 	free_percpu(net->mib.ipv6_statistics);
763 err_ip_mib:
764 	free_percpu(net->mib.udplite_stats_in6);
765 err_udplite_mib:
766 	free_percpu(net->mib.udp_stats_in6);
767 	return -ENOMEM;
768 }
769 
770 static void ipv6_cleanup_mibs(struct net *net)
771 {
772 	free_percpu(net->mib.udp_stats_in6);
773 	free_percpu(net->mib.udplite_stats_in6);
774 	free_percpu(net->mib.ipv6_statistics);
775 	free_percpu(net->mib.icmpv6_statistics);
776 	kfree(net->mib.icmpv6msg_statistics);
777 }
778 
779 static int __net_init inet6_net_init(struct net *net)
780 {
781 	int err = 0;
782 
783 	net->ipv6.sysctl.bindv6only = 0;
784 	net->ipv6.sysctl.icmpv6_time = 1*HZ;
785 	net->ipv6.sysctl.flowlabel_consistency = 1;
786 	net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS;
787 	net->ipv6.sysctl.idgen_retries = 3;
788 	net->ipv6.sysctl.idgen_delay = 1 * HZ;
789 	net->ipv6.sysctl.flowlabel_state_ranges = 0;
790 	atomic_set(&net->ipv6.fib6_sernum, 1);
791 
792 	err = ipv6_init_mibs(net);
793 	if (err)
794 		return err;
795 #ifdef CONFIG_PROC_FS
796 	err = udp6_proc_init(net);
797 	if (err)
798 		goto out;
799 	err = tcp6_proc_init(net);
800 	if (err)
801 		goto proc_tcp6_fail;
802 	err = ac6_proc_init(net);
803 	if (err)
804 		goto proc_ac6_fail;
805 #endif
806 	return err;
807 
808 #ifdef CONFIG_PROC_FS
809 proc_ac6_fail:
810 	tcp6_proc_exit(net);
811 proc_tcp6_fail:
812 	udp6_proc_exit(net);
813 out:
814 	ipv6_cleanup_mibs(net);
815 	return err;
816 #endif
817 }
818 
819 static void __net_exit inet6_net_exit(struct net *net)
820 {
821 #ifdef CONFIG_PROC_FS
822 	udp6_proc_exit(net);
823 	tcp6_proc_exit(net);
824 	ac6_proc_exit(net);
825 #endif
826 	ipv6_cleanup_mibs(net);
827 }
828 
829 static struct pernet_operations inet6_net_ops = {
830 	.init = inet6_net_init,
831 	.exit = inet6_net_exit,
832 };
833 
834 static const struct ipv6_stub ipv6_stub_impl = {
835 	.ipv6_sock_mc_join = ipv6_sock_mc_join,
836 	.ipv6_sock_mc_drop = ipv6_sock_mc_drop,
837 	.ipv6_dst_lookup = ip6_dst_lookup,
838 	.udpv6_encap_enable = udpv6_encap_enable,
839 	.ndisc_send_na = ndisc_send_na,
840 	.nd_tbl	= &nd_tbl,
841 };
842 
843 static int __init inet6_init(void)
844 {
845 	struct list_head *r;
846 	int err = 0;
847 
848 	sock_skb_cb_check_size(sizeof(struct inet6_skb_parm));
849 
850 	/* Register the socket-side information for inet6_create.  */
851 	for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
852 		INIT_LIST_HEAD(r);
853 
854 	if (disable_ipv6_mod) {
855 		pr_info("Loaded, but administratively disabled, reboot required to enable\n");
856 		goto out;
857 	}
858 
859 	err = proto_register(&tcpv6_prot, 1);
860 	if (err)
861 		goto out;
862 
863 	err = proto_register(&udpv6_prot, 1);
864 	if (err)
865 		goto out_unregister_tcp_proto;
866 
867 	err = proto_register(&udplitev6_prot, 1);
868 	if (err)
869 		goto out_unregister_udp_proto;
870 
871 	err = proto_register(&rawv6_prot, 1);
872 	if (err)
873 		goto out_unregister_udplite_proto;
874 
875 	err = proto_register(&pingv6_prot, 1);
876 	if (err)
877 		goto out_unregister_ping_proto;
878 
879 	/* We MUST register RAW sockets before we create the ICMP6,
880 	 * IGMP6, or NDISC control sockets.
881 	 */
882 	err = rawv6_init();
883 	if (err)
884 		goto out_unregister_raw_proto;
885 
886 	/* Register the family here so that the init calls below will
887 	 * be able to create sockets. (?? is this dangerous ??)
888 	 */
889 	err = sock_register(&inet6_family_ops);
890 	if (err)
891 		goto out_sock_register_fail;
892 
893 	/*
894 	 *	ipngwg API draft makes clear that the correct semantics
895 	 *	for TCP and UDP is to consider one TCP and UDP instance
896 	 *	in a host available by both INET and INET6 APIs and
897 	 *	able to communicate via both network protocols.
898 	 */
899 
900 	err = register_pernet_subsys(&inet6_net_ops);
901 	if (err)
902 		goto register_pernet_fail;
903 	err = icmpv6_init();
904 	if (err)
905 		goto icmp_fail;
906 	err = ip6_mr_init();
907 	if (err)
908 		goto ipmr_fail;
909 	err = ndisc_init();
910 	if (err)
911 		goto ndisc_fail;
912 	err = igmp6_init();
913 	if (err)
914 		goto igmp_fail;
915 
916 	ipv6_stub = &ipv6_stub_impl;
917 
918 	err = ipv6_netfilter_init();
919 	if (err)
920 		goto netfilter_fail;
921 	/* Create /proc/foo6 entries. */
922 #ifdef CONFIG_PROC_FS
923 	err = -ENOMEM;
924 	if (raw6_proc_init())
925 		goto proc_raw6_fail;
926 	if (udplite6_proc_init())
927 		goto proc_udplite6_fail;
928 	if (ipv6_misc_proc_init())
929 		goto proc_misc6_fail;
930 	if (if6_proc_init())
931 		goto proc_if6_fail;
932 #endif
933 	err = ip6_route_init();
934 	if (err)
935 		goto ip6_route_fail;
936 	err = ndisc_late_init();
937 	if (err)
938 		goto ndisc_late_fail;
939 	err = ip6_flowlabel_init();
940 	if (err)
941 		goto ip6_flowlabel_fail;
942 	err = addrconf_init();
943 	if (err)
944 		goto addrconf_fail;
945 
946 	/* Init v6 extension headers. */
947 	err = ipv6_exthdrs_init();
948 	if (err)
949 		goto ipv6_exthdrs_fail;
950 
951 	err = ipv6_frag_init();
952 	if (err)
953 		goto ipv6_frag_fail;
954 
955 	/* Init v6 transport protocols. */
956 	err = udpv6_init();
957 	if (err)
958 		goto udpv6_fail;
959 
960 	err = udplitev6_init();
961 	if (err)
962 		goto udplitev6_fail;
963 
964 	err = udpv6_offload_init();
965 	if (err)
966 		goto udpv6_offload_fail;
967 
968 	err = tcpv6_init();
969 	if (err)
970 		goto tcpv6_fail;
971 
972 	err = ipv6_packet_init();
973 	if (err)
974 		goto ipv6_packet_fail;
975 
976 	err = pingv6_init();
977 	if (err)
978 		goto pingv6_fail;
979 
980 #ifdef CONFIG_SYSCTL
981 	err = ipv6_sysctl_register();
982 	if (err)
983 		goto sysctl_fail;
984 #endif
985 out:
986 	return err;
987 
988 #ifdef CONFIG_SYSCTL
989 sysctl_fail:
990 	pingv6_exit();
991 #endif
992 pingv6_fail:
993 	ipv6_packet_cleanup();
994 ipv6_packet_fail:
995 	tcpv6_exit();
996 tcpv6_fail:
997 	udpv6_offload_exit();
998 udpv6_offload_fail:
999 	udplitev6_exit();
1000 udplitev6_fail:
1001 	udpv6_exit();
1002 udpv6_fail:
1003 	ipv6_frag_exit();
1004 ipv6_frag_fail:
1005 	ipv6_exthdrs_exit();
1006 ipv6_exthdrs_fail:
1007 	addrconf_cleanup();
1008 addrconf_fail:
1009 	ip6_flowlabel_cleanup();
1010 ip6_flowlabel_fail:
1011 	ndisc_late_cleanup();
1012 ndisc_late_fail:
1013 	ip6_route_cleanup();
1014 ip6_route_fail:
1015 #ifdef CONFIG_PROC_FS
1016 	if6_proc_exit();
1017 proc_if6_fail:
1018 	ipv6_misc_proc_exit();
1019 proc_misc6_fail:
1020 	udplite6_proc_exit();
1021 proc_udplite6_fail:
1022 	raw6_proc_exit();
1023 proc_raw6_fail:
1024 #endif
1025 	ipv6_netfilter_fini();
1026 netfilter_fail:
1027 	igmp6_cleanup();
1028 igmp_fail:
1029 	ndisc_cleanup();
1030 ndisc_fail:
1031 	ip6_mr_cleanup();
1032 ipmr_fail:
1033 	icmpv6_cleanup();
1034 icmp_fail:
1035 	unregister_pernet_subsys(&inet6_net_ops);
1036 register_pernet_fail:
1037 	sock_unregister(PF_INET6);
1038 	rtnl_unregister_all(PF_INET6);
1039 out_sock_register_fail:
1040 	rawv6_exit();
1041 out_unregister_ping_proto:
1042 	proto_unregister(&pingv6_prot);
1043 out_unregister_raw_proto:
1044 	proto_unregister(&rawv6_prot);
1045 out_unregister_udplite_proto:
1046 	proto_unregister(&udplitev6_prot);
1047 out_unregister_udp_proto:
1048 	proto_unregister(&udpv6_prot);
1049 out_unregister_tcp_proto:
1050 	proto_unregister(&tcpv6_prot);
1051 	goto out;
1052 }
1053 module_init(inet6_init);
1054 
1055 MODULE_ALIAS_NETPROTO(PF_INET6);
1056