xref: /openbmc/linux/net/ipv6/af_inet6.c (revision 8e694cd2)
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 bool ipv6_mod_enabled(void)
96 {
97 	return disable_ipv6_mod == 0;
98 }
99 EXPORT_SYMBOL_GPL(ipv6_mod_enabled);
100 
101 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
102 {
103 	const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
104 
105 	return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
106 }
107 
108 static int inet6_create(struct net *net, struct socket *sock, int protocol,
109 			int kern)
110 {
111 	struct inet_sock *inet;
112 	struct ipv6_pinfo *np;
113 	struct sock *sk;
114 	struct inet_protosw *answer;
115 	struct proto *answer_prot;
116 	unsigned char answer_flags;
117 	int try_loading_module = 0;
118 	int err;
119 
120 	if (protocol < 0 || protocol >= IPPROTO_MAX)
121 		return -EINVAL;
122 
123 	/* Look for the requested type/protocol pair. */
124 lookup_protocol:
125 	err = -ESOCKTNOSUPPORT;
126 	rcu_read_lock();
127 	list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
128 
129 		err = 0;
130 		/* Check the non-wild match. */
131 		if (protocol == answer->protocol) {
132 			if (protocol != IPPROTO_IP)
133 				break;
134 		} else {
135 			/* Check for the two wild cases. */
136 			if (IPPROTO_IP == protocol) {
137 				protocol = answer->protocol;
138 				break;
139 			}
140 			if (IPPROTO_IP == answer->protocol)
141 				break;
142 		}
143 		err = -EPROTONOSUPPORT;
144 	}
145 
146 	if (err) {
147 		if (try_loading_module < 2) {
148 			rcu_read_unlock();
149 			/*
150 			 * Be more specific, e.g. net-pf-10-proto-132-type-1
151 			 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
152 			 */
153 			if (++try_loading_module == 1)
154 				request_module("net-pf-%d-proto-%d-type-%d",
155 						PF_INET6, protocol, sock->type);
156 			/*
157 			 * Fall back to generic, e.g. net-pf-10-proto-132
158 			 * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
159 			 */
160 			else
161 				request_module("net-pf-%d-proto-%d",
162 						PF_INET6, protocol);
163 			goto lookup_protocol;
164 		} else
165 			goto out_rcu_unlock;
166 	}
167 
168 	err = -EPERM;
169 	if (sock->type == SOCK_RAW && !kern &&
170 	    !ns_capable(net->user_ns, CAP_NET_RAW))
171 		goto out_rcu_unlock;
172 
173 	sock->ops = answer->ops;
174 	answer_prot = answer->prot;
175 	answer_flags = answer->flags;
176 	rcu_read_unlock();
177 
178 	WARN_ON(!answer_prot->slab);
179 
180 	err = -ENOBUFS;
181 	sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot, kern);
182 	if (!sk)
183 		goto out;
184 
185 	sock_init_data(sock, sk);
186 
187 	err = 0;
188 	if (INET_PROTOSW_REUSE & answer_flags)
189 		sk->sk_reuse = SK_CAN_REUSE;
190 
191 	inet = inet_sk(sk);
192 	inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
193 
194 	if (SOCK_RAW == sock->type) {
195 		inet->inet_num = protocol;
196 		if (IPPROTO_RAW == protocol)
197 			inet->hdrincl = 1;
198 	}
199 
200 	sk->sk_destruct		= inet_sock_destruct;
201 	sk->sk_family		= PF_INET6;
202 	sk->sk_protocol		= protocol;
203 
204 	sk->sk_backlog_rcv	= answer->prot->backlog_rcv;
205 
206 	inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
207 	np->hop_limit	= -1;
208 	np->mcast_hops	= IPV6_DEFAULT_MCASTHOPS;
209 	np->mc_loop	= 1;
210 	np->pmtudisc	= IPV6_PMTUDISC_WANT;
211 	np->autoflowlabel = ip6_default_np_autolabel(sock_net(sk));
212 	sk->sk_ipv6only	= net->ipv6.sysctl.bindv6only;
213 
214 	/* Init the ipv4 part of the socket since we can have sockets
215 	 * using v6 API for ipv4.
216 	 */
217 	inet->uc_ttl	= -1;
218 
219 	inet->mc_loop	= 1;
220 	inet->mc_ttl	= 1;
221 	inet->mc_index	= 0;
222 	inet->mc_list	= NULL;
223 	inet->rcv_tos	= 0;
224 
225 	if (net->ipv4.sysctl_ip_no_pmtu_disc)
226 		inet->pmtudisc = IP_PMTUDISC_DONT;
227 	else
228 		inet->pmtudisc = IP_PMTUDISC_WANT;
229 	/*
230 	 * Increment only the relevant sk_prot->socks debug field, this changes
231 	 * the previous behaviour of incrementing both the equivalent to
232 	 * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
233 	 *
234 	 * This allows better debug granularity as we'll know exactly how many
235 	 * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
236 	 * transport protocol socks. -acme
237 	 */
238 	sk_refcnt_debug_inc(sk);
239 
240 	if (inet->inet_num) {
241 		/* It assumes that any protocol which allows
242 		 * the user to assign a number at socket
243 		 * creation time automatically shares.
244 		 */
245 		inet->inet_sport = htons(inet->inet_num);
246 		err = sk->sk_prot->hash(sk);
247 		if (err) {
248 			sk_common_release(sk);
249 			goto out;
250 		}
251 	}
252 	if (sk->sk_prot->init) {
253 		err = sk->sk_prot->init(sk);
254 		if (err) {
255 			sk_common_release(sk);
256 			goto out;
257 		}
258 	}
259 out:
260 	return err;
261 out_rcu_unlock:
262 	rcu_read_unlock();
263 	goto out;
264 }
265 
266 
267 /* bind for INET6 API */
268 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
269 {
270 	struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr;
271 	struct sock *sk = sock->sk;
272 	struct inet_sock *inet = inet_sk(sk);
273 	struct ipv6_pinfo *np = inet6_sk(sk);
274 	struct net *net = sock_net(sk);
275 	__be32 v4addr = 0;
276 	unsigned short snum;
277 	int addr_type = 0;
278 	int err = 0;
279 
280 	/* If the socket has its own bind function then use it. */
281 	if (sk->sk_prot->bind)
282 		return sk->sk_prot->bind(sk, uaddr, addr_len);
283 
284 	if (addr_len < SIN6_LEN_RFC2133)
285 		return -EINVAL;
286 
287 	if (addr->sin6_family != AF_INET6)
288 		return -EAFNOSUPPORT;
289 
290 	addr_type = ipv6_addr_type(&addr->sin6_addr);
291 	if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
292 		return -EINVAL;
293 
294 	snum = ntohs(addr->sin6_port);
295 	if (snum && snum < PROT_SOCK && !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
296 		return -EACCES;
297 
298 	lock_sock(sk);
299 
300 	/* Check these errors (active socket, double bind). */
301 	if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
302 		err = -EINVAL;
303 		goto out;
304 	}
305 
306 	/* Check if the address belongs to the host. */
307 	if (addr_type == IPV6_ADDR_MAPPED) {
308 		int chk_addr_ret;
309 
310 		/* Binding to v4-mapped address on a v6-only socket
311 		 * makes no sense
312 		 */
313 		if (sk->sk_ipv6only) {
314 			err = -EINVAL;
315 			goto out;
316 		}
317 
318 		/* Reproduce AF_INET checks to make the bindings consistent */
319 		v4addr = addr->sin6_addr.s6_addr32[3];
320 		chk_addr_ret = inet_addr_type(net, v4addr);
321 		if (!net->ipv4.sysctl_ip_nonlocal_bind &&
322 		    !(inet->freebind || inet->transparent) &&
323 		    v4addr != htonl(INADDR_ANY) &&
324 		    chk_addr_ret != RTN_LOCAL &&
325 		    chk_addr_ret != RTN_MULTICAST &&
326 		    chk_addr_ret != RTN_BROADCAST) {
327 			err = -EADDRNOTAVAIL;
328 			goto out;
329 		}
330 	} else {
331 		if (addr_type != IPV6_ADDR_ANY) {
332 			struct net_device *dev = NULL;
333 
334 			rcu_read_lock();
335 			if (__ipv6_addr_needs_scope_id(addr_type)) {
336 				if (addr_len >= sizeof(struct sockaddr_in6) &&
337 				    addr->sin6_scope_id) {
338 					/* Override any existing binding, if another one
339 					 * is supplied by user.
340 					 */
341 					sk->sk_bound_dev_if = addr->sin6_scope_id;
342 				}
343 
344 				/* Binding to link-local address requires an interface */
345 				if (!sk->sk_bound_dev_if) {
346 					err = -EINVAL;
347 					goto out_unlock;
348 				}
349 				dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
350 				if (!dev) {
351 					err = -ENODEV;
352 					goto out_unlock;
353 				}
354 			}
355 
356 			/* ipv4 addr of the socket is invalid.  Only the
357 			 * unspecified and mapped address have a v4 equivalent.
358 			 */
359 			v4addr = LOOPBACK4_IPV6;
360 			if (!(addr_type & IPV6_ADDR_MULTICAST))	{
361 				if (!net->ipv6.sysctl.ip_nonlocal_bind &&
362 				    !(inet->freebind || inet->transparent) &&
363 				    !ipv6_chk_addr(net, &addr->sin6_addr,
364 						   dev, 0)) {
365 					err = -EADDRNOTAVAIL;
366 					goto out_unlock;
367 				}
368 			}
369 			rcu_read_unlock();
370 		}
371 	}
372 
373 	inet->inet_rcv_saddr = v4addr;
374 	inet->inet_saddr = v4addr;
375 
376 	sk->sk_v6_rcv_saddr = addr->sin6_addr;
377 
378 	if (!(addr_type & IPV6_ADDR_MULTICAST))
379 		np->saddr = addr->sin6_addr;
380 
381 	/* Make sure we are allowed to bind here. */
382 	if ((snum || !inet->bind_address_no_port) &&
383 	    sk->sk_prot->get_port(sk, snum)) {
384 		inet_reset_saddr(sk);
385 		err = -EADDRINUSE;
386 		goto out;
387 	}
388 
389 	if (addr_type != IPV6_ADDR_ANY) {
390 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
391 		if (addr_type != IPV6_ADDR_MAPPED)
392 			sk->sk_ipv6only = 1;
393 	}
394 	if (snum)
395 		sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
396 	inet->inet_sport = htons(inet->inet_num);
397 	inet->inet_dport = 0;
398 	inet->inet_daddr = 0;
399 out:
400 	release_sock(sk);
401 	return err;
402 out_unlock:
403 	rcu_read_unlock();
404 	goto out;
405 }
406 EXPORT_SYMBOL(inet6_bind);
407 
408 int inet6_release(struct socket *sock)
409 {
410 	struct sock *sk = sock->sk;
411 
412 	if (!sk)
413 		return -EINVAL;
414 
415 	/* Free mc lists */
416 	ipv6_sock_mc_close(sk);
417 
418 	/* Free ac lists */
419 	ipv6_sock_ac_close(sk);
420 
421 	return inet_release(sock);
422 }
423 EXPORT_SYMBOL(inet6_release);
424 
425 void inet6_destroy_sock(struct sock *sk)
426 {
427 	struct ipv6_pinfo *np = inet6_sk(sk);
428 	struct sk_buff *skb;
429 	struct ipv6_txoptions *opt;
430 
431 	/* Release rx options */
432 
433 	skb = xchg(&np->pktoptions, NULL);
434 	if (skb)
435 		kfree_skb(skb);
436 
437 	skb = xchg(&np->rxpmtu, NULL);
438 	if (skb)
439 		kfree_skb(skb);
440 
441 	/* Free flowlabels */
442 	fl6_free_socklist(sk);
443 
444 	/* Free tx options */
445 
446 	opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL);
447 	if (opt) {
448 		atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
449 		txopt_put(opt);
450 	}
451 }
452 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
453 
454 /*
455  *	This does both peername and sockname.
456  */
457 
458 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
459 		 int *uaddr_len, int peer)
460 {
461 	struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
462 	struct sock *sk = sock->sk;
463 	struct inet_sock *inet = inet_sk(sk);
464 	struct ipv6_pinfo *np = inet6_sk(sk);
465 
466 	sin->sin6_family = AF_INET6;
467 	sin->sin6_flowinfo = 0;
468 	sin->sin6_scope_id = 0;
469 	if (peer) {
470 		if (!inet->inet_dport)
471 			return -ENOTCONN;
472 		if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
473 		    peer == 1)
474 			return -ENOTCONN;
475 		sin->sin6_port = inet->inet_dport;
476 		sin->sin6_addr = sk->sk_v6_daddr;
477 		if (np->sndflow)
478 			sin->sin6_flowinfo = np->flow_label;
479 	} else {
480 		if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
481 			sin->sin6_addr = np->saddr;
482 		else
483 			sin->sin6_addr = sk->sk_v6_rcv_saddr;
484 
485 		sin->sin6_port = inet->inet_sport;
486 	}
487 	sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
488 						 sk->sk_bound_dev_if);
489 	*uaddr_len = sizeof(*sin);
490 	return 0;
491 }
492 EXPORT_SYMBOL(inet6_getname);
493 
494 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
495 {
496 	struct sock *sk = sock->sk;
497 	struct net *net = sock_net(sk);
498 
499 	switch (cmd) {
500 	case SIOCGSTAMP:
501 		return sock_get_timestamp(sk, (struct timeval __user *)arg);
502 
503 	case SIOCGSTAMPNS:
504 		return sock_get_timestampns(sk, (struct timespec __user *)arg);
505 
506 	case SIOCADDRT:
507 	case SIOCDELRT:
508 
509 		return ipv6_route_ioctl(net, cmd, (void __user *)arg);
510 
511 	case SIOCSIFADDR:
512 		return addrconf_add_ifaddr(net, (void __user *) arg);
513 	case SIOCDIFADDR:
514 		return addrconf_del_ifaddr(net, (void __user *) arg);
515 	case SIOCSIFDSTADDR:
516 		return addrconf_set_dstaddr(net, (void __user *) arg);
517 	default:
518 		if (!sk->sk_prot->ioctl)
519 			return -ENOIOCTLCMD;
520 		return sk->sk_prot->ioctl(sk, cmd, arg);
521 	}
522 	/*NOTREACHED*/
523 	return 0;
524 }
525 EXPORT_SYMBOL(inet6_ioctl);
526 
527 const struct proto_ops inet6_stream_ops = {
528 	.family		   = PF_INET6,
529 	.owner		   = THIS_MODULE,
530 	.release	   = inet6_release,
531 	.bind		   = inet6_bind,
532 	.connect	   = inet_stream_connect,	/* ok		*/
533 	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
534 	.accept		   = inet_accept,		/* ok		*/
535 	.getname	   = inet6_getname,
536 	.poll		   = tcp_poll,			/* ok		*/
537 	.ioctl		   = inet6_ioctl,		/* must change  */
538 	.listen		   = inet_listen,		/* ok		*/
539 	.shutdown	   = inet_shutdown,		/* ok		*/
540 	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
541 	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
542 	.sendmsg	   = inet_sendmsg,		/* ok		*/
543 	.recvmsg	   = inet_recvmsg,		/* ok		*/
544 	.mmap		   = sock_no_mmap,
545 	.sendpage	   = inet_sendpage,
546 	.splice_read	   = tcp_splice_read,
547 #ifdef CONFIG_COMPAT
548 	.compat_setsockopt = compat_sock_common_setsockopt,
549 	.compat_getsockopt = compat_sock_common_getsockopt,
550 #endif
551 };
552 
553 const struct proto_ops inet6_dgram_ops = {
554 	.family		   = PF_INET6,
555 	.owner		   = THIS_MODULE,
556 	.release	   = inet6_release,
557 	.bind		   = inet6_bind,
558 	.connect	   = inet_dgram_connect,	/* ok		*/
559 	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
560 	.accept		   = sock_no_accept,		/* a do nothing	*/
561 	.getname	   = inet6_getname,
562 	.poll		   = udp_poll,			/* ok		*/
563 	.ioctl		   = inet6_ioctl,		/* must change  */
564 	.listen		   = sock_no_listen,		/* ok		*/
565 	.shutdown	   = inet_shutdown,		/* ok		*/
566 	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
567 	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
568 	.sendmsg	   = inet_sendmsg,		/* ok		*/
569 	.recvmsg	   = inet_recvmsg,		/* ok		*/
570 	.mmap		   = sock_no_mmap,
571 	.sendpage	   = sock_no_sendpage,
572 	.set_peek_off	   = sk_set_peek_off,
573 #ifdef CONFIG_COMPAT
574 	.compat_setsockopt = compat_sock_common_setsockopt,
575 	.compat_getsockopt = compat_sock_common_getsockopt,
576 #endif
577 };
578 
579 static const struct net_proto_family inet6_family_ops = {
580 	.family = PF_INET6,
581 	.create = inet6_create,
582 	.owner	= THIS_MODULE,
583 };
584 
585 int inet6_register_protosw(struct inet_protosw *p)
586 {
587 	struct list_head *lh;
588 	struct inet_protosw *answer;
589 	struct list_head *last_perm;
590 	int protocol = p->protocol;
591 	int ret;
592 
593 	spin_lock_bh(&inetsw6_lock);
594 
595 	ret = -EINVAL;
596 	if (p->type >= SOCK_MAX)
597 		goto out_illegal;
598 
599 	/* If we are trying to override a permanent protocol, bail. */
600 	answer = NULL;
601 	ret = -EPERM;
602 	last_perm = &inetsw6[p->type];
603 	list_for_each(lh, &inetsw6[p->type]) {
604 		answer = list_entry(lh, struct inet_protosw, list);
605 
606 		/* Check only the non-wild match. */
607 		if (INET_PROTOSW_PERMANENT & answer->flags) {
608 			if (protocol == answer->protocol)
609 				break;
610 			last_perm = lh;
611 		}
612 
613 		answer = NULL;
614 	}
615 	if (answer)
616 		goto out_permanent;
617 
618 	/* Add the new entry after the last permanent entry if any, so that
619 	 * the new entry does not override a permanent entry when matched with
620 	 * a wild-card protocol. But it is allowed to override any existing
621 	 * non-permanent entry.  This means that when we remove this entry, the
622 	 * system automatically returns to the old behavior.
623 	 */
624 	list_add_rcu(&p->list, last_perm);
625 	ret = 0;
626 out:
627 	spin_unlock_bh(&inetsw6_lock);
628 	return ret;
629 
630 out_permanent:
631 	pr_err("Attempt to override permanent protocol %d\n", protocol);
632 	goto out;
633 
634 out_illegal:
635 	pr_err("Ignoring attempt to register invalid socket type %d\n",
636 	       p->type);
637 	goto out;
638 }
639 EXPORT_SYMBOL(inet6_register_protosw);
640 
641 void
642 inet6_unregister_protosw(struct inet_protosw *p)
643 {
644 	if (INET_PROTOSW_PERMANENT & p->flags) {
645 		pr_err("Attempt to unregister permanent protocol %d\n",
646 		       p->protocol);
647 	} else {
648 		spin_lock_bh(&inetsw6_lock);
649 		list_del_rcu(&p->list);
650 		spin_unlock_bh(&inetsw6_lock);
651 
652 		synchronize_net();
653 	}
654 }
655 EXPORT_SYMBOL(inet6_unregister_protosw);
656 
657 int inet6_sk_rebuild_header(struct sock *sk)
658 {
659 	struct ipv6_pinfo *np = inet6_sk(sk);
660 	struct dst_entry *dst;
661 
662 	dst = __sk_dst_check(sk, np->dst_cookie);
663 
664 	if (!dst) {
665 		struct inet_sock *inet = inet_sk(sk);
666 		struct in6_addr *final_p, final;
667 		struct flowi6 fl6;
668 
669 		memset(&fl6, 0, sizeof(fl6));
670 		fl6.flowi6_proto = sk->sk_protocol;
671 		fl6.daddr = sk->sk_v6_daddr;
672 		fl6.saddr = np->saddr;
673 		fl6.flowlabel = np->flow_label;
674 		fl6.flowi6_oif = sk->sk_bound_dev_if;
675 		fl6.flowi6_mark = sk->sk_mark;
676 		fl6.fl6_dport = inet->inet_dport;
677 		fl6.fl6_sport = inet->inet_sport;
678 		security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
679 
680 		rcu_read_lock();
681 		final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt),
682 					 &final);
683 		rcu_read_unlock();
684 
685 		dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
686 		if (IS_ERR(dst)) {
687 			sk->sk_route_caps = 0;
688 			sk->sk_err_soft = -PTR_ERR(dst);
689 			return PTR_ERR(dst);
690 		}
691 
692 		ip6_dst_store(sk, dst, NULL, NULL);
693 	}
694 
695 	return 0;
696 }
697 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
698 
699 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb,
700 		       const struct inet6_skb_parm *opt)
701 {
702 	const struct ipv6_pinfo *np = inet6_sk(sk);
703 
704 	if (np->rxopt.all) {
705 		if (((opt->flags & IP6SKB_HOPBYHOP) &&
706 		     (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) ||
707 		    (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) &&
708 		     np->rxopt.bits.rxflow) ||
709 		    (opt->srcrt && (np->rxopt.bits.srcrt ||
710 		     np->rxopt.bits.osrcrt)) ||
711 		    ((opt->dst1 || opt->dst0) &&
712 		     (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
713 			return true;
714 	}
715 	return false;
716 }
717 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
718 
719 static struct packet_type ipv6_packet_type __read_mostly = {
720 	.type = cpu_to_be16(ETH_P_IPV6),
721 	.func = ipv6_rcv,
722 };
723 
724 static int __init ipv6_packet_init(void)
725 {
726 	dev_add_pack(&ipv6_packet_type);
727 	return 0;
728 }
729 
730 static void ipv6_packet_cleanup(void)
731 {
732 	dev_remove_pack(&ipv6_packet_type);
733 }
734 
735 static int __net_init ipv6_init_mibs(struct net *net)
736 {
737 	int i;
738 
739 	net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib);
740 	if (!net->mib.udp_stats_in6)
741 		return -ENOMEM;
742 	net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib);
743 	if (!net->mib.udplite_stats_in6)
744 		goto err_udplite_mib;
745 	net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib);
746 	if (!net->mib.ipv6_statistics)
747 		goto err_ip_mib;
748 
749 	for_each_possible_cpu(i) {
750 		struct ipstats_mib *af_inet6_stats;
751 		af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i);
752 		u64_stats_init(&af_inet6_stats->syncp);
753 	}
754 
755 
756 	net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib);
757 	if (!net->mib.icmpv6_statistics)
758 		goto err_icmp_mib;
759 	net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
760 						GFP_KERNEL);
761 	if (!net->mib.icmpv6msg_statistics)
762 		goto err_icmpmsg_mib;
763 	return 0;
764 
765 err_icmpmsg_mib:
766 	free_percpu(net->mib.icmpv6_statistics);
767 err_icmp_mib:
768 	free_percpu(net->mib.ipv6_statistics);
769 err_ip_mib:
770 	free_percpu(net->mib.udplite_stats_in6);
771 err_udplite_mib:
772 	free_percpu(net->mib.udp_stats_in6);
773 	return -ENOMEM;
774 }
775 
776 static void ipv6_cleanup_mibs(struct net *net)
777 {
778 	free_percpu(net->mib.udp_stats_in6);
779 	free_percpu(net->mib.udplite_stats_in6);
780 	free_percpu(net->mib.ipv6_statistics);
781 	free_percpu(net->mib.icmpv6_statistics);
782 	kfree(net->mib.icmpv6msg_statistics);
783 }
784 
785 static int __net_init inet6_net_init(struct net *net)
786 {
787 	int err = 0;
788 
789 	net->ipv6.sysctl.bindv6only = 0;
790 	net->ipv6.sysctl.icmpv6_time = 1*HZ;
791 	net->ipv6.sysctl.flowlabel_consistency = 1;
792 	net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS;
793 	net->ipv6.sysctl.idgen_retries = 3;
794 	net->ipv6.sysctl.idgen_delay = 1 * HZ;
795 	net->ipv6.sysctl.flowlabel_state_ranges = 0;
796 	atomic_set(&net->ipv6.fib6_sernum, 1);
797 
798 	err = ipv6_init_mibs(net);
799 	if (err)
800 		return err;
801 #ifdef CONFIG_PROC_FS
802 	err = udp6_proc_init(net);
803 	if (err)
804 		goto out;
805 	err = tcp6_proc_init(net);
806 	if (err)
807 		goto proc_tcp6_fail;
808 	err = ac6_proc_init(net);
809 	if (err)
810 		goto proc_ac6_fail;
811 #endif
812 	return err;
813 
814 #ifdef CONFIG_PROC_FS
815 proc_ac6_fail:
816 	tcp6_proc_exit(net);
817 proc_tcp6_fail:
818 	udp6_proc_exit(net);
819 out:
820 	ipv6_cleanup_mibs(net);
821 	return err;
822 #endif
823 }
824 
825 static void __net_exit inet6_net_exit(struct net *net)
826 {
827 #ifdef CONFIG_PROC_FS
828 	udp6_proc_exit(net);
829 	tcp6_proc_exit(net);
830 	ac6_proc_exit(net);
831 #endif
832 	ipv6_cleanup_mibs(net);
833 }
834 
835 static struct pernet_operations inet6_net_ops = {
836 	.init = inet6_net_init,
837 	.exit = inet6_net_exit,
838 };
839 
840 static const struct ipv6_stub ipv6_stub_impl = {
841 	.ipv6_sock_mc_join = ipv6_sock_mc_join,
842 	.ipv6_sock_mc_drop = ipv6_sock_mc_drop,
843 	.ipv6_dst_lookup = ip6_dst_lookup,
844 	.udpv6_encap_enable = udpv6_encap_enable,
845 	.ndisc_send_na = ndisc_send_na,
846 	.nd_tbl	= &nd_tbl,
847 };
848 
849 static int __init inet6_init(void)
850 {
851 	struct list_head *r;
852 	int err = 0;
853 
854 	sock_skb_cb_check_size(sizeof(struct inet6_skb_parm));
855 
856 	/* Register the socket-side information for inet6_create.  */
857 	for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
858 		INIT_LIST_HEAD(r);
859 
860 	if (disable_ipv6_mod) {
861 		pr_info("Loaded, but administratively disabled, reboot required to enable\n");
862 		goto out;
863 	}
864 
865 	err = proto_register(&tcpv6_prot, 1);
866 	if (err)
867 		goto out;
868 
869 	err = proto_register(&udpv6_prot, 1);
870 	if (err)
871 		goto out_unregister_tcp_proto;
872 
873 	err = proto_register(&udplitev6_prot, 1);
874 	if (err)
875 		goto out_unregister_udp_proto;
876 
877 	err = proto_register(&rawv6_prot, 1);
878 	if (err)
879 		goto out_unregister_udplite_proto;
880 
881 	err = proto_register(&pingv6_prot, 1);
882 	if (err)
883 		goto out_unregister_ping_proto;
884 
885 	/* We MUST register RAW sockets before we create the ICMP6,
886 	 * IGMP6, or NDISC control sockets.
887 	 */
888 	err = rawv6_init();
889 	if (err)
890 		goto out_unregister_raw_proto;
891 
892 	/* Register the family here so that the init calls below will
893 	 * be able to create sockets. (?? is this dangerous ??)
894 	 */
895 	err = sock_register(&inet6_family_ops);
896 	if (err)
897 		goto out_sock_register_fail;
898 
899 	/*
900 	 *	ipngwg API draft makes clear that the correct semantics
901 	 *	for TCP and UDP is to consider one TCP and UDP instance
902 	 *	in a host available by both INET and INET6 APIs and
903 	 *	able to communicate via both network protocols.
904 	 */
905 
906 	err = register_pernet_subsys(&inet6_net_ops);
907 	if (err)
908 		goto register_pernet_fail;
909 	err = icmpv6_init();
910 	if (err)
911 		goto icmp_fail;
912 	err = ip6_mr_init();
913 	if (err)
914 		goto ipmr_fail;
915 	err = ndisc_init();
916 	if (err)
917 		goto ndisc_fail;
918 	err = igmp6_init();
919 	if (err)
920 		goto igmp_fail;
921 
922 	ipv6_stub = &ipv6_stub_impl;
923 
924 	err = ipv6_netfilter_init();
925 	if (err)
926 		goto netfilter_fail;
927 	/* Create /proc/foo6 entries. */
928 #ifdef CONFIG_PROC_FS
929 	err = -ENOMEM;
930 	if (raw6_proc_init())
931 		goto proc_raw6_fail;
932 	if (udplite6_proc_init())
933 		goto proc_udplite6_fail;
934 	if (ipv6_misc_proc_init())
935 		goto proc_misc6_fail;
936 	if (if6_proc_init())
937 		goto proc_if6_fail;
938 #endif
939 	err = ip6_route_init();
940 	if (err)
941 		goto ip6_route_fail;
942 	err = ndisc_late_init();
943 	if (err)
944 		goto ndisc_late_fail;
945 	err = ip6_flowlabel_init();
946 	if (err)
947 		goto ip6_flowlabel_fail;
948 	err = addrconf_init();
949 	if (err)
950 		goto addrconf_fail;
951 
952 	/* Init v6 extension headers. */
953 	err = ipv6_exthdrs_init();
954 	if (err)
955 		goto ipv6_exthdrs_fail;
956 
957 	err = ipv6_frag_init();
958 	if (err)
959 		goto ipv6_frag_fail;
960 
961 	/* Init v6 transport protocols. */
962 	err = udpv6_init();
963 	if (err)
964 		goto udpv6_fail;
965 
966 	err = udplitev6_init();
967 	if (err)
968 		goto udplitev6_fail;
969 
970 	err = udpv6_offload_init();
971 	if (err)
972 		goto udpv6_offload_fail;
973 
974 	err = tcpv6_init();
975 	if (err)
976 		goto tcpv6_fail;
977 
978 	err = ipv6_packet_init();
979 	if (err)
980 		goto ipv6_packet_fail;
981 
982 	err = pingv6_init();
983 	if (err)
984 		goto pingv6_fail;
985 
986 #ifdef CONFIG_SYSCTL
987 	err = ipv6_sysctl_register();
988 	if (err)
989 		goto sysctl_fail;
990 #endif
991 out:
992 	return err;
993 
994 #ifdef CONFIG_SYSCTL
995 sysctl_fail:
996 	pingv6_exit();
997 #endif
998 pingv6_fail:
999 	ipv6_packet_cleanup();
1000 ipv6_packet_fail:
1001 	tcpv6_exit();
1002 tcpv6_fail:
1003 	udpv6_offload_exit();
1004 udpv6_offload_fail:
1005 	udplitev6_exit();
1006 udplitev6_fail:
1007 	udpv6_exit();
1008 udpv6_fail:
1009 	ipv6_frag_exit();
1010 ipv6_frag_fail:
1011 	ipv6_exthdrs_exit();
1012 ipv6_exthdrs_fail:
1013 	addrconf_cleanup();
1014 addrconf_fail:
1015 	ip6_flowlabel_cleanup();
1016 ip6_flowlabel_fail:
1017 	ndisc_late_cleanup();
1018 ndisc_late_fail:
1019 	ip6_route_cleanup();
1020 ip6_route_fail:
1021 #ifdef CONFIG_PROC_FS
1022 	if6_proc_exit();
1023 proc_if6_fail:
1024 	ipv6_misc_proc_exit();
1025 proc_misc6_fail:
1026 	udplite6_proc_exit();
1027 proc_udplite6_fail:
1028 	raw6_proc_exit();
1029 proc_raw6_fail:
1030 #endif
1031 	ipv6_netfilter_fini();
1032 netfilter_fail:
1033 	igmp6_cleanup();
1034 igmp_fail:
1035 	ndisc_cleanup();
1036 ndisc_fail:
1037 	ip6_mr_cleanup();
1038 ipmr_fail:
1039 	icmpv6_cleanup();
1040 icmp_fail:
1041 	unregister_pernet_subsys(&inet6_net_ops);
1042 register_pernet_fail:
1043 	sock_unregister(PF_INET6);
1044 	rtnl_unregister_all(PF_INET6);
1045 out_sock_register_fail:
1046 	rawv6_exit();
1047 out_unregister_ping_proto:
1048 	proto_unregister(&pingv6_prot);
1049 out_unregister_raw_proto:
1050 	proto_unregister(&rawv6_prot);
1051 out_unregister_udplite_proto:
1052 	proto_unregister(&udplitev6_prot);
1053 out_unregister_udp_proto:
1054 	proto_unregister(&udpv6_prot);
1055 out_unregister_tcp_proto:
1056 	proto_unregister(&tcpv6_prot);
1057 	goto out;
1058 }
1059 module_init(inet6_init);
1060 
1061 MODULE_ALIAS_NETPROTO(PF_INET6);
1062