xref: /openbmc/linux/net/ipv6/af_inet6.c (revision 133f9794)
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 #include <net/calipso.h>
64 #include <net/seg6.h>
65 
66 #include <linux/uaccess.h>
67 #include <linux/mroute6.h>
68 
69 #include "ip6_offload.h"
70 
71 MODULE_AUTHOR("Cast of dozens");
72 MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
73 MODULE_LICENSE("GPL");
74 
75 /* The inetsw6 table contains everything that inet6_create needs to
76  * build a new socket.
77  */
78 static struct list_head inetsw6[SOCK_MAX];
79 static DEFINE_SPINLOCK(inetsw6_lock);
80 
81 struct ipv6_params ipv6_defaults = {
82 	.disable_ipv6 = 0,
83 	.autoconf = 1,
84 };
85 
86 static int disable_ipv6_mod;
87 
88 module_param_named(disable, disable_ipv6_mod, int, 0444);
89 MODULE_PARM_DESC(disable, "Disable IPv6 module such that it is non-functional");
90 
91 module_param_named(disable_ipv6, ipv6_defaults.disable_ipv6, int, 0444);
92 MODULE_PARM_DESC(disable_ipv6, "Disable IPv6 on all interfaces");
93 
94 module_param_named(autoconf, ipv6_defaults.autoconf, int, 0444);
95 MODULE_PARM_DESC(autoconf, "Enable IPv6 address autoconfiguration on all interfaces");
96 
97 bool ipv6_mod_enabled(void)
98 {
99 	return disable_ipv6_mod == 0;
100 }
101 EXPORT_SYMBOL_GPL(ipv6_mod_enabled);
102 
103 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
104 {
105 	const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
106 
107 	return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
108 }
109 
110 static int inet6_create(struct net *net, struct socket *sock, int protocol,
111 			int kern)
112 {
113 	struct inet_sock *inet;
114 	struct ipv6_pinfo *np;
115 	struct sock *sk;
116 	struct inet_protosw *answer;
117 	struct proto *answer_prot;
118 	unsigned char answer_flags;
119 	int try_loading_module = 0;
120 	int err;
121 
122 	if (protocol < 0 || protocol >= IPPROTO_MAX)
123 		return -EINVAL;
124 
125 	/* Look for the requested type/protocol pair. */
126 lookup_protocol:
127 	err = -ESOCKTNOSUPPORT;
128 	rcu_read_lock();
129 	list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
130 
131 		err = 0;
132 		/* Check the non-wild match. */
133 		if (protocol == answer->protocol) {
134 			if (protocol != IPPROTO_IP)
135 				break;
136 		} else {
137 			/* Check for the two wild cases. */
138 			if (IPPROTO_IP == protocol) {
139 				protocol = answer->protocol;
140 				break;
141 			}
142 			if (IPPROTO_IP == answer->protocol)
143 				break;
144 		}
145 		err = -EPROTONOSUPPORT;
146 	}
147 
148 	if (err) {
149 		if (try_loading_module < 2) {
150 			rcu_read_unlock();
151 			/*
152 			 * Be more specific, e.g. net-pf-10-proto-132-type-1
153 			 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
154 			 */
155 			if (++try_loading_module == 1)
156 				request_module("net-pf-%d-proto-%d-type-%d",
157 						PF_INET6, protocol, sock->type);
158 			/*
159 			 * Fall back to generic, e.g. net-pf-10-proto-132
160 			 * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
161 			 */
162 			else
163 				request_module("net-pf-%d-proto-%d",
164 						PF_INET6, protocol);
165 			goto lookup_protocol;
166 		} else
167 			goto out_rcu_unlock;
168 	}
169 
170 	err = -EPERM;
171 	if (sock->type == SOCK_RAW && !kern &&
172 	    !ns_capable(net->user_ns, CAP_NET_RAW))
173 		goto out_rcu_unlock;
174 
175 	sock->ops = answer->ops;
176 	answer_prot = answer->prot;
177 	answer_flags = answer->flags;
178 	rcu_read_unlock();
179 
180 	WARN_ON(!answer_prot->slab);
181 
182 	err = -ENOBUFS;
183 	sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot, kern);
184 	if (!sk)
185 		goto out;
186 
187 	sock_init_data(sock, sk);
188 
189 	err = 0;
190 	if (INET_PROTOSW_REUSE & answer_flags)
191 		sk->sk_reuse = SK_CAN_REUSE;
192 
193 	inet = inet_sk(sk);
194 	inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
195 
196 	if (SOCK_RAW == sock->type) {
197 		inet->inet_num = protocol;
198 		if (IPPROTO_RAW == protocol)
199 			inet->hdrincl = 1;
200 	}
201 
202 	sk->sk_destruct		= inet_sock_destruct;
203 	sk->sk_family		= PF_INET6;
204 	sk->sk_protocol		= protocol;
205 
206 	sk->sk_backlog_rcv	= answer->prot->backlog_rcv;
207 
208 	inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
209 	np->hop_limit	= -1;
210 	np->mcast_hops	= IPV6_DEFAULT_MCASTHOPS;
211 	np->mc_loop	= 1;
212 	np->pmtudisc	= IPV6_PMTUDISC_WANT;
213 	np->repflow	= net->ipv6.sysctl.flowlabel_reflect;
214 	sk->sk_ipv6only	= net->ipv6.sysctl.bindv6only;
215 
216 	/* Init the ipv4 part of the socket since we can have sockets
217 	 * using v6 API for ipv4.
218 	 */
219 	inet->uc_ttl	= -1;
220 
221 	inet->mc_loop	= 1;
222 	inet->mc_ttl	= 1;
223 	inet->mc_index	= 0;
224 	inet->mc_list	= NULL;
225 	inet->rcv_tos	= 0;
226 
227 	if (net->ipv4.sysctl_ip_no_pmtu_disc)
228 		inet->pmtudisc = IP_PMTUDISC_DONT;
229 	else
230 		inet->pmtudisc = IP_PMTUDISC_WANT;
231 	/*
232 	 * Increment only the relevant sk_prot->socks debug field, this changes
233 	 * the previous behaviour of incrementing both the equivalent to
234 	 * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
235 	 *
236 	 * This allows better debug granularity as we'll know exactly how many
237 	 * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
238 	 * transport protocol socks. -acme
239 	 */
240 	sk_refcnt_debug_inc(sk);
241 
242 	if (inet->inet_num) {
243 		/* It assumes that any protocol which allows
244 		 * the user to assign a number at socket
245 		 * creation time automatically shares.
246 		 */
247 		inet->inet_sport = htons(inet->inet_num);
248 		err = sk->sk_prot->hash(sk);
249 		if (err) {
250 			sk_common_release(sk);
251 			goto out;
252 		}
253 	}
254 	if (sk->sk_prot->init) {
255 		err = sk->sk_prot->init(sk);
256 		if (err) {
257 			sk_common_release(sk);
258 			goto out;
259 		}
260 	}
261 
262 	if (!kern) {
263 		err = BPF_CGROUP_RUN_PROG_INET_SOCK(sk);
264 		if (err) {
265 			sk_common_release(sk);
266 			goto out;
267 		}
268 	}
269 out:
270 	return err;
271 out_rcu_unlock:
272 	rcu_read_unlock();
273 	goto out;
274 }
275 
276 
277 /* bind for INET6 API */
278 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
279 {
280 	struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr;
281 	struct sock *sk = sock->sk;
282 	struct inet_sock *inet = inet_sk(sk);
283 	struct ipv6_pinfo *np = inet6_sk(sk);
284 	struct net *net = sock_net(sk);
285 	__be32 v4addr = 0;
286 	unsigned short snum;
287 	bool saved_ipv6only;
288 	int addr_type = 0;
289 	int err = 0;
290 
291 	/* If the socket has its own bind function then use it. */
292 	if (sk->sk_prot->bind)
293 		return sk->sk_prot->bind(sk, uaddr, addr_len);
294 
295 	if (addr_len < SIN6_LEN_RFC2133)
296 		return -EINVAL;
297 
298 	if (addr->sin6_family != AF_INET6)
299 		return -EAFNOSUPPORT;
300 
301 	addr_type = ipv6_addr_type(&addr->sin6_addr);
302 	if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
303 		return -EINVAL;
304 
305 	snum = ntohs(addr->sin6_port);
306 	if (snum && snum < inet_prot_sock(net) &&
307 	    !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
308 		return -EACCES;
309 
310 	lock_sock(sk);
311 
312 	/* Check these errors (active socket, double bind). */
313 	if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
314 		err = -EINVAL;
315 		goto out;
316 	}
317 
318 	/* Check if the address belongs to the host. */
319 	if (addr_type == IPV6_ADDR_MAPPED) {
320 		int chk_addr_ret;
321 
322 		/* Binding to v4-mapped address on a v6-only socket
323 		 * makes no sense
324 		 */
325 		if (sk->sk_ipv6only) {
326 			err = -EINVAL;
327 			goto out;
328 		}
329 
330 		/* Reproduce AF_INET checks to make the bindings consistent */
331 		v4addr = addr->sin6_addr.s6_addr32[3];
332 		chk_addr_ret = inet_addr_type(net, v4addr);
333 		if (!net->ipv4.sysctl_ip_nonlocal_bind &&
334 		    !(inet->freebind || inet->transparent) &&
335 		    v4addr != htonl(INADDR_ANY) &&
336 		    chk_addr_ret != RTN_LOCAL &&
337 		    chk_addr_ret != RTN_MULTICAST &&
338 		    chk_addr_ret != RTN_BROADCAST) {
339 			err = -EADDRNOTAVAIL;
340 			goto out;
341 		}
342 	} else {
343 		if (addr_type != IPV6_ADDR_ANY) {
344 			struct net_device *dev = NULL;
345 
346 			rcu_read_lock();
347 			if (__ipv6_addr_needs_scope_id(addr_type)) {
348 				if (addr_len >= sizeof(struct sockaddr_in6) &&
349 				    addr->sin6_scope_id) {
350 					/* Override any existing binding, if another one
351 					 * is supplied by user.
352 					 */
353 					sk->sk_bound_dev_if = addr->sin6_scope_id;
354 				}
355 
356 				/* Binding to link-local address requires an interface */
357 				if (!sk->sk_bound_dev_if) {
358 					err = -EINVAL;
359 					goto out_unlock;
360 				}
361 				dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
362 				if (!dev) {
363 					err = -ENODEV;
364 					goto out_unlock;
365 				}
366 			}
367 
368 			/* ipv4 addr of the socket is invalid.  Only the
369 			 * unspecified and mapped address have a v4 equivalent.
370 			 */
371 			v4addr = LOOPBACK4_IPV6;
372 			if (!(addr_type & IPV6_ADDR_MULTICAST))	{
373 				if (!net->ipv6.sysctl.ip_nonlocal_bind &&
374 				    !(inet->freebind || inet->transparent) &&
375 				    !ipv6_chk_addr(net, &addr->sin6_addr,
376 						   dev, 0)) {
377 					err = -EADDRNOTAVAIL;
378 					goto out_unlock;
379 				}
380 			}
381 			rcu_read_unlock();
382 		}
383 	}
384 
385 	inet->inet_rcv_saddr = v4addr;
386 	inet->inet_saddr = v4addr;
387 
388 	sk->sk_v6_rcv_saddr = addr->sin6_addr;
389 
390 	if (!(addr_type & IPV6_ADDR_MULTICAST))
391 		np->saddr = addr->sin6_addr;
392 
393 	saved_ipv6only = sk->sk_ipv6only;
394 	if (addr_type != IPV6_ADDR_ANY && addr_type != IPV6_ADDR_MAPPED)
395 		sk->sk_ipv6only = 1;
396 
397 	/* Make sure we are allowed to bind here. */
398 	if ((snum || !inet->bind_address_no_port) &&
399 	    sk->sk_prot->get_port(sk, snum)) {
400 		sk->sk_ipv6only = saved_ipv6only;
401 		inet_reset_saddr(sk);
402 		err = -EADDRINUSE;
403 		goto out;
404 	}
405 
406 	if (addr_type != IPV6_ADDR_ANY)
407 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
408 	if (snum)
409 		sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
410 	inet->inet_sport = htons(inet->inet_num);
411 	inet->inet_dport = 0;
412 	inet->inet_daddr = 0;
413 out:
414 	release_sock(sk);
415 	return err;
416 out_unlock:
417 	rcu_read_unlock();
418 	goto out;
419 }
420 EXPORT_SYMBOL(inet6_bind);
421 
422 int inet6_release(struct socket *sock)
423 {
424 	struct sock *sk = sock->sk;
425 
426 	if (!sk)
427 		return -EINVAL;
428 
429 	/* Free mc lists */
430 	ipv6_sock_mc_close(sk);
431 
432 	/* Free ac lists */
433 	ipv6_sock_ac_close(sk);
434 
435 	return inet_release(sock);
436 }
437 EXPORT_SYMBOL(inet6_release);
438 
439 void inet6_destroy_sock(struct sock *sk)
440 {
441 	struct ipv6_pinfo *np = inet6_sk(sk);
442 	struct sk_buff *skb;
443 	struct ipv6_txoptions *opt;
444 
445 	/* Release rx options */
446 
447 	skb = xchg(&np->pktoptions, NULL);
448 	if (skb)
449 		kfree_skb(skb);
450 
451 	skb = xchg(&np->rxpmtu, NULL);
452 	if (skb)
453 		kfree_skb(skb);
454 
455 	/* Free flowlabels */
456 	fl6_free_socklist(sk);
457 
458 	/* Free tx options */
459 
460 	opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL);
461 	if (opt) {
462 		atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
463 		txopt_put(opt);
464 	}
465 }
466 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
467 
468 /*
469  *	This does both peername and sockname.
470  */
471 
472 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
473 		 int *uaddr_len, int peer)
474 {
475 	struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
476 	struct sock *sk = sock->sk;
477 	struct inet_sock *inet = inet_sk(sk);
478 	struct ipv6_pinfo *np = inet6_sk(sk);
479 
480 	sin->sin6_family = AF_INET6;
481 	sin->sin6_flowinfo = 0;
482 	sin->sin6_scope_id = 0;
483 	if (peer) {
484 		if (!inet->inet_dport)
485 			return -ENOTCONN;
486 		if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
487 		    peer == 1)
488 			return -ENOTCONN;
489 		sin->sin6_port = inet->inet_dport;
490 		sin->sin6_addr = sk->sk_v6_daddr;
491 		if (np->sndflow)
492 			sin->sin6_flowinfo = np->flow_label;
493 	} else {
494 		if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
495 			sin->sin6_addr = np->saddr;
496 		else
497 			sin->sin6_addr = sk->sk_v6_rcv_saddr;
498 
499 		sin->sin6_port = inet->inet_sport;
500 	}
501 	sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
502 						 sk->sk_bound_dev_if);
503 	*uaddr_len = sizeof(*sin);
504 	return 0;
505 }
506 EXPORT_SYMBOL(inet6_getname);
507 
508 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
509 {
510 	struct sock *sk = sock->sk;
511 	struct net *net = sock_net(sk);
512 
513 	switch (cmd) {
514 	case SIOCGSTAMP:
515 		return sock_get_timestamp(sk, (struct timeval __user *)arg);
516 
517 	case SIOCGSTAMPNS:
518 		return sock_get_timestampns(sk, (struct timespec __user *)arg);
519 
520 	case SIOCADDRT:
521 	case SIOCDELRT:
522 
523 		return ipv6_route_ioctl(net, cmd, (void __user *)arg);
524 
525 	case SIOCSIFADDR:
526 		return addrconf_add_ifaddr(net, (void __user *) arg);
527 	case SIOCDIFADDR:
528 		return addrconf_del_ifaddr(net, (void __user *) arg);
529 	case SIOCSIFDSTADDR:
530 		return addrconf_set_dstaddr(net, (void __user *) arg);
531 	default:
532 		if (!sk->sk_prot->ioctl)
533 			return -ENOIOCTLCMD;
534 		return sk->sk_prot->ioctl(sk, cmd, arg);
535 	}
536 	/*NOTREACHED*/
537 	return 0;
538 }
539 EXPORT_SYMBOL(inet6_ioctl);
540 
541 const struct proto_ops inet6_stream_ops = {
542 	.family		   = PF_INET6,
543 	.owner		   = THIS_MODULE,
544 	.release	   = inet6_release,
545 	.bind		   = inet6_bind,
546 	.connect	   = inet_stream_connect,	/* ok		*/
547 	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
548 	.accept		   = inet_accept,		/* ok		*/
549 	.getname	   = inet6_getname,
550 	.poll		   = tcp_poll,			/* ok		*/
551 	.ioctl		   = inet6_ioctl,		/* must change  */
552 	.listen		   = inet_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	   = inet_sendpage,
560 	.sendmsg_locked    = tcp_sendmsg_locked,
561 	.sendpage_locked   = tcp_sendpage_locked,
562 	.splice_read	   = tcp_splice_read,
563 	.read_sock	   = tcp_read_sock,
564 	.peek_len	   = tcp_peek_len,
565 #ifdef CONFIG_COMPAT
566 	.compat_setsockopt = compat_sock_common_setsockopt,
567 	.compat_getsockopt = compat_sock_common_getsockopt,
568 #endif
569 };
570 
571 const struct proto_ops inet6_dgram_ops = {
572 	.family		   = PF_INET6,
573 	.owner		   = THIS_MODULE,
574 	.release	   = inet6_release,
575 	.bind		   = inet6_bind,
576 	.connect	   = inet_dgram_connect,	/* ok		*/
577 	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
578 	.accept		   = sock_no_accept,		/* a do nothing	*/
579 	.getname	   = inet6_getname,
580 	.poll		   = udp_poll,			/* ok		*/
581 	.ioctl		   = inet6_ioctl,		/* must change  */
582 	.listen		   = sock_no_listen,		/* ok		*/
583 	.shutdown	   = inet_shutdown,		/* ok		*/
584 	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
585 	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
586 	.sendmsg	   = inet_sendmsg,		/* ok		*/
587 	.recvmsg	   = inet_recvmsg,		/* ok		*/
588 	.mmap		   = sock_no_mmap,
589 	.sendpage	   = sock_no_sendpage,
590 	.set_peek_off	   = sk_set_peek_off,
591 #ifdef CONFIG_COMPAT
592 	.compat_setsockopt = compat_sock_common_setsockopt,
593 	.compat_getsockopt = compat_sock_common_getsockopt,
594 #endif
595 };
596 
597 static const struct net_proto_family inet6_family_ops = {
598 	.family = PF_INET6,
599 	.create = inet6_create,
600 	.owner	= THIS_MODULE,
601 };
602 
603 int inet6_register_protosw(struct inet_protosw *p)
604 {
605 	struct list_head *lh;
606 	struct inet_protosw *answer;
607 	struct list_head *last_perm;
608 	int protocol = p->protocol;
609 	int ret;
610 
611 	spin_lock_bh(&inetsw6_lock);
612 
613 	ret = -EINVAL;
614 	if (p->type >= SOCK_MAX)
615 		goto out_illegal;
616 
617 	/* If we are trying to override a permanent protocol, bail. */
618 	answer = NULL;
619 	ret = -EPERM;
620 	last_perm = &inetsw6[p->type];
621 	list_for_each(lh, &inetsw6[p->type]) {
622 		answer = list_entry(lh, struct inet_protosw, list);
623 
624 		/* Check only the non-wild match. */
625 		if (INET_PROTOSW_PERMANENT & answer->flags) {
626 			if (protocol == answer->protocol)
627 				break;
628 			last_perm = lh;
629 		}
630 
631 		answer = NULL;
632 	}
633 	if (answer)
634 		goto out_permanent;
635 
636 	/* Add the new entry after the last permanent entry if any, so that
637 	 * the new entry does not override a permanent entry when matched with
638 	 * a wild-card protocol. But it is allowed to override any existing
639 	 * non-permanent entry.  This means that when we remove this entry, the
640 	 * system automatically returns to the old behavior.
641 	 */
642 	list_add_rcu(&p->list, last_perm);
643 	ret = 0;
644 out:
645 	spin_unlock_bh(&inetsw6_lock);
646 	return ret;
647 
648 out_permanent:
649 	pr_err("Attempt to override permanent protocol %d\n", protocol);
650 	goto out;
651 
652 out_illegal:
653 	pr_err("Ignoring attempt to register invalid socket type %d\n",
654 	       p->type);
655 	goto out;
656 }
657 EXPORT_SYMBOL(inet6_register_protosw);
658 
659 void
660 inet6_unregister_protosw(struct inet_protosw *p)
661 {
662 	if (INET_PROTOSW_PERMANENT & p->flags) {
663 		pr_err("Attempt to unregister permanent protocol %d\n",
664 		       p->protocol);
665 	} else {
666 		spin_lock_bh(&inetsw6_lock);
667 		list_del_rcu(&p->list);
668 		spin_unlock_bh(&inetsw6_lock);
669 
670 		synchronize_net();
671 	}
672 }
673 EXPORT_SYMBOL(inet6_unregister_protosw);
674 
675 int inet6_sk_rebuild_header(struct sock *sk)
676 {
677 	struct ipv6_pinfo *np = inet6_sk(sk);
678 	struct dst_entry *dst;
679 
680 	dst = __sk_dst_check(sk, np->dst_cookie);
681 
682 	if (!dst) {
683 		struct inet_sock *inet = inet_sk(sk);
684 		struct in6_addr *final_p, final;
685 		struct flowi6 fl6;
686 
687 		memset(&fl6, 0, sizeof(fl6));
688 		fl6.flowi6_proto = sk->sk_protocol;
689 		fl6.daddr = sk->sk_v6_daddr;
690 		fl6.saddr = np->saddr;
691 		fl6.flowlabel = np->flow_label;
692 		fl6.flowi6_oif = sk->sk_bound_dev_if;
693 		fl6.flowi6_mark = sk->sk_mark;
694 		fl6.fl6_dport = inet->inet_dport;
695 		fl6.fl6_sport = inet->inet_sport;
696 		fl6.flowi6_uid = sk->sk_uid;
697 		security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
698 
699 		rcu_read_lock();
700 		final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt),
701 					 &final);
702 		rcu_read_unlock();
703 
704 		dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
705 		if (IS_ERR(dst)) {
706 			sk->sk_route_caps = 0;
707 			sk->sk_err_soft = -PTR_ERR(dst);
708 			return PTR_ERR(dst);
709 		}
710 
711 		ip6_dst_store(sk, dst, NULL, NULL);
712 	}
713 
714 	return 0;
715 }
716 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
717 
718 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb,
719 		       const struct inet6_skb_parm *opt)
720 {
721 	const struct ipv6_pinfo *np = inet6_sk(sk);
722 
723 	if (np->rxopt.all) {
724 		if (((opt->flags & IP6SKB_HOPBYHOP) &&
725 		     (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) ||
726 		    (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) &&
727 		     np->rxopt.bits.rxflow) ||
728 		    (opt->srcrt && (np->rxopt.bits.srcrt ||
729 		     np->rxopt.bits.osrcrt)) ||
730 		    ((opt->dst1 || opt->dst0) &&
731 		     (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
732 			return true;
733 	}
734 	return false;
735 }
736 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
737 
738 static struct packet_type ipv6_packet_type __read_mostly = {
739 	.type = cpu_to_be16(ETH_P_IPV6),
740 	.func = ipv6_rcv,
741 };
742 
743 static int __init ipv6_packet_init(void)
744 {
745 	dev_add_pack(&ipv6_packet_type);
746 	return 0;
747 }
748 
749 static void ipv6_packet_cleanup(void)
750 {
751 	dev_remove_pack(&ipv6_packet_type);
752 }
753 
754 static int __net_init ipv6_init_mibs(struct net *net)
755 {
756 	int i;
757 
758 	net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib);
759 	if (!net->mib.udp_stats_in6)
760 		return -ENOMEM;
761 	net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib);
762 	if (!net->mib.udplite_stats_in6)
763 		goto err_udplite_mib;
764 	net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib);
765 	if (!net->mib.ipv6_statistics)
766 		goto err_ip_mib;
767 
768 	for_each_possible_cpu(i) {
769 		struct ipstats_mib *af_inet6_stats;
770 		af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i);
771 		u64_stats_init(&af_inet6_stats->syncp);
772 	}
773 
774 
775 	net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib);
776 	if (!net->mib.icmpv6_statistics)
777 		goto err_icmp_mib;
778 	net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
779 						GFP_KERNEL);
780 	if (!net->mib.icmpv6msg_statistics)
781 		goto err_icmpmsg_mib;
782 	return 0;
783 
784 err_icmpmsg_mib:
785 	free_percpu(net->mib.icmpv6_statistics);
786 err_icmp_mib:
787 	free_percpu(net->mib.ipv6_statistics);
788 err_ip_mib:
789 	free_percpu(net->mib.udplite_stats_in6);
790 err_udplite_mib:
791 	free_percpu(net->mib.udp_stats_in6);
792 	return -ENOMEM;
793 }
794 
795 static void ipv6_cleanup_mibs(struct net *net)
796 {
797 	free_percpu(net->mib.udp_stats_in6);
798 	free_percpu(net->mib.udplite_stats_in6);
799 	free_percpu(net->mib.ipv6_statistics);
800 	free_percpu(net->mib.icmpv6_statistics);
801 	kfree(net->mib.icmpv6msg_statistics);
802 }
803 
804 static int __net_init inet6_net_init(struct net *net)
805 {
806 	int err = 0;
807 
808 	net->ipv6.sysctl.bindv6only = 0;
809 	net->ipv6.sysctl.icmpv6_time = 1*HZ;
810 	net->ipv6.sysctl.flowlabel_consistency = 1;
811 	net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS;
812 	net->ipv6.sysctl.idgen_retries = 3;
813 	net->ipv6.sysctl.idgen_delay = 1 * HZ;
814 	net->ipv6.sysctl.flowlabel_state_ranges = 0;
815 	net->ipv6.sysctl.max_dst_opts_cnt = IP6_DEFAULT_MAX_DST_OPTS_CNT;
816 	net->ipv6.sysctl.max_hbh_opts_cnt = IP6_DEFAULT_MAX_HBH_OPTS_CNT;
817 	net->ipv6.sysctl.max_dst_opts_len = IP6_DEFAULT_MAX_DST_OPTS_LEN;
818 	net->ipv6.sysctl.max_hbh_opts_len = IP6_DEFAULT_MAX_HBH_OPTS_LEN;
819 	atomic_set(&net->ipv6.fib6_sernum, 1);
820 
821 	err = ipv6_init_mibs(net);
822 	if (err)
823 		return err;
824 #ifdef CONFIG_PROC_FS
825 	err = udp6_proc_init(net);
826 	if (err)
827 		goto out;
828 	err = tcp6_proc_init(net);
829 	if (err)
830 		goto proc_tcp6_fail;
831 	err = ac6_proc_init(net);
832 	if (err)
833 		goto proc_ac6_fail;
834 #endif
835 	return err;
836 
837 #ifdef CONFIG_PROC_FS
838 proc_ac6_fail:
839 	tcp6_proc_exit(net);
840 proc_tcp6_fail:
841 	udp6_proc_exit(net);
842 out:
843 	ipv6_cleanup_mibs(net);
844 	return err;
845 #endif
846 }
847 
848 static void __net_exit inet6_net_exit(struct net *net)
849 {
850 #ifdef CONFIG_PROC_FS
851 	udp6_proc_exit(net);
852 	tcp6_proc_exit(net);
853 	ac6_proc_exit(net);
854 #endif
855 	ipv6_cleanup_mibs(net);
856 }
857 
858 static struct pernet_operations inet6_net_ops = {
859 	.init = inet6_net_init,
860 	.exit = inet6_net_exit,
861 };
862 
863 static const struct ipv6_stub ipv6_stub_impl = {
864 	.ipv6_sock_mc_join = ipv6_sock_mc_join,
865 	.ipv6_sock_mc_drop = ipv6_sock_mc_drop,
866 	.ipv6_dst_lookup = ip6_dst_lookup,
867 	.udpv6_encap_enable = udpv6_encap_enable,
868 	.ndisc_send_na = ndisc_send_na,
869 	.nd_tbl	= &nd_tbl,
870 };
871 
872 static int __init inet6_init(void)
873 {
874 	struct list_head *r;
875 	int err = 0;
876 
877 	sock_skb_cb_check_size(sizeof(struct inet6_skb_parm));
878 
879 	/* Register the socket-side information for inet6_create.  */
880 	for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
881 		INIT_LIST_HEAD(r);
882 
883 	if (disable_ipv6_mod) {
884 		pr_info("Loaded, but administratively disabled, reboot required to enable\n");
885 		goto out;
886 	}
887 
888 	err = proto_register(&tcpv6_prot, 1);
889 	if (err)
890 		goto out;
891 
892 	err = proto_register(&udpv6_prot, 1);
893 	if (err)
894 		goto out_unregister_tcp_proto;
895 
896 	err = proto_register(&udplitev6_prot, 1);
897 	if (err)
898 		goto out_unregister_udp_proto;
899 
900 	err = proto_register(&rawv6_prot, 1);
901 	if (err)
902 		goto out_unregister_udplite_proto;
903 
904 	err = proto_register(&pingv6_prot, 1);
905 	if (err)
906 		goto out_unregister_ping_proto;
907 
908 	/* We MUST register RAW sockets before we create the ICMP6,
909 	 * IGMP6, or NDISC control sockets.
910 	 */
911 	err = rawv6_init();
912 	if (err)
913 		goto out_unregister_raw_proto;
914 
915 	/* Register the family here so that the init calls below will
916 	 * be able to create sockets. (?? is this dangerous ??)
917 	 */
918 	err = sock_register(&inet6_family_ops);
919 	if (err)
920 		goto out_sock_register_fail;
921 
922 	/*
923 	 *	ipngwg API draft makes clear that the correct semantics
924 	 *	for TCP and UDP is to consider one TCP and UDP instance
925 	 *	in a host available by both INET and INET6 APIs and
926 	 *	able to communicate via both network protocols.
927 	 */
928 
929 	err = register_pernet_subsys(&inet6_net_ops);
930 	if (err)
931 		goto register_pernet_fail;
932 	err = ip6_mr_init();
933 	if (err)
934 		goto ipmr_fail;
935 	err = icmpv6_init();
936 	if (err)
937 		goto icmp_fail;
938 	err = ndisc_init();
939 	if (err)
940 		goto ndisc_fail;
941 	err = igmp6_init();
942 	if (err)
943 		goto igmp_fail;
944 
945 	err = ipv6_netfilter_init();
946 	if (err)
947 		goto netfilter_fail;
948 	/* Create /proc/foo6 entries. */
949 #ifdef CONFIG_PROC_FS
950 	err = -ENOMEM;
951 	if (raw6_proc_init())
952 		goto proc_raw6_fail;
953 	if (udplite6_proc_init())
954 		goto proc_udplite6_fail;
955 	if (ipv6_misc_proc_init())
956 		goto proc_misc6_fail;
957 	if (if6_proc_init())
958 		goto proc_if6_fail;
959 #endif
960 	err = ip6_route_init();
961 	if (err)
962 		goto ip6_route_fail;
963 	err = ndisc_late_init();
964 	if (err)
965 		goto ndisc_late_fail;
966 	err = ip6_flowlabel_init();
967 	if (err)
968 		goto ip6_flowlabel_fail;
969 	err = addrconf_init();
970 	if (err)
971 		goto addrconf_fail;
972 
973 	/* Init v6 extension headers. */
974 	err = ipv6_exthdrs_init();
975 	if (err)
976 		goto ipv6_exthdrs_fail;
977 
978 	err = ipv6_frag_init();
979 	if (err)
980 		goto ipv6_frag_fail;
981 
982 	/* Init v6 transport protocols. */
983 	err = udpv6_init();
984 	if (err)
985 		goto udpv6_fail;
986 
987 	err = udplitev6_init();
988 	if (err)
989 		goto udplitev6_fail;
990 
991 	err = udpv6_offload_init();
992 	if (err)
993 		goto udpv6_offload_fail;
994 
995 	err = tcpv6_init();
996 	if (err)
997 		goto tcpv6_fail;
998 
999 	err = ipv6_packet_init();
1000 	if (err)
1001 		goto ipv6_packet_fail;
1002 
1003 	err = pingv6_init();
1004 	if (err)
1005 		goto pingv6_fail;
1006 
1007 	err = calipso_init();
1008 	if (err)
1009 		goto calipso_fail;
1010 
1011 	err = seg6_init();
1012 	if (err)
1013 		goto seg6_fail;
1014 
1015 	err = igmp6_late_init();
1016 	if (err)
1017 		goto igmp6_late_err;
1018 
1019 #ifdef CONFIG_SYSCTL
1020 	err = ipv6_sysctl_register();
1021 	if (err)
1022 		goto sysctl_fail;
1023 #endif
1024 
1025 	/* ensure that ipv6 stubs are visible only after ipv6 is ready */
1026 	wmb();
1027 	ipv6_stub = &ipv6_stub_impl;
1028 out:
1029 	return err;
1030 
1031 #ifdef CONFIG_SYSCTL
1032 sysctl_fail:
1033 	igmp6_late_cleanup();
1034 #endif
1035 igmp6_late_err:
1036 	seg6_exit();
1037 seg6_fail:
1038 	calipso_exit();
1039 calipso_fail:
1040 	pingv6_exit();
1041 pingv6_fail:
1042 	ipv6_packet_cleanup();
1043 ipv6_packet_fail:
1044 	tcpv6_exit();
1045 tcpv6_fail:
1046 	udpv6_offload_exit();
1047 udpv6_offload_fail:
1048 	udplitev6_exit();
1049 udplitev6_fail:
1050 	udpv6_exit();
1051 udpv6_fail:
1052 	ipv6_frag_exit();
1053 ipv6_frag_fail:
1054 	ipv6_exthdrs_exit();
1055 ipv6_exthdrs_fail:
1056 	addrconf_cleanup();
1057 addrconf_fail:
1058 	ip6_flowlabel_cleanup();
1059 ip6_flowlabel_fail:
1060 	ndisc_late_cleanup();
1061 ndisc_late_fail:
1062 	ip6_route_cleanup();
1063 ip6_route_fail:
1064 #ifdef CONFIG_PROC_FS
1065 	if6_proc_exit();
1066 proc_if6_fail:
1067 	ipv6_misc_proc_exit();
1068 proc_misc6_fail:
1069 	udplite6_proc_exit();
1070 proc_udplite6_fail:
1071 	raw6_proc_exit();
1072 proc_raw6_fail:
1073 #endif
1074 	ipv6_netfilter_fini();
1075 netfilter_fail:
1076 	igmp6_cleanup();
1077 igmp_fail:
1078 	ndisc_cleanup();
1079 ndisc_fail:
1080 	ip6_mr_cleanup();
1081 icmp_fail:
1082 	unregister_pernet_subsys(&inet6_net_ops);
1083 ipmr_fail:
1084 	icmpv6_cleanup();
1085 register_pernet_fail:
1086 	sock_unregister(PF_INET6);
1087 	rtnl_unregister_all(PF_INET6);
1088 out_sock_register_fail:
1089 	rawv6_exit();
1090 out_unregister_ping_proto:
1091 	proto_unregister(&pingv6_prot);
1092 out_unregister_raw_proto:
1093 	proto_unregister(&rawv6_prot);
1094 out_unregister_udplite_proto:
1095 	proto_unregister(&udplitev6_prot);
1096 out_unregister_udp_proto:
1097 	proto_unregister(&udpv6_prot);
1098 out_unregister_tcp_proto:
1099 	proto_unregister(&tcpv6_prot);
1100 	goto out;
1101 }
1102 module_init(inet6_init);
1103 
1104 MODULE_ALIAS_NETPROTO(PF_INET6);
1105