xref: /openbmc/linux/net/ipv6/af_inet6.c (revision 6dfcd296)
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 
65 #include <asm/uaccess.h>
66 #include <linux/mroute6.h>
67 
68 #include "ip6_offload.h"
69 
70 MODULE_AUTHOR("Cast of dozens");
71 MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
72 MODULE_LICENSE("GPL");
73 
74 /* The inetsw6 table contains everything that inet6_create needs to
75  * build a new socket.
76  */
77 static struct list_head inetsw6[SOCK_MAX];
78 static DEFINE_SPINLOCK(inetsw6_lock);
79 
80 struct ipv6_params ipv6_defaults = {
81 	.disable_ipv6 = 0,
82 	.autoconf = 1,
83 };
84 
85 static int disable_ipv6_mod;
86 
87 module_param_named(disable, disable_ipv6_mod, int, 0444);
88 MODULE_PARM_DESC(disable, "Disable IPv6 module such that it is non-functional");
89 
90 module_param_named(disable_ipv6, ipv6_defaults.disable_ipv6, int, 0444);
91 MODULE_PARM_DESC(disable_ipv6, "Disable IPv6 on all interfaces");
92 
93 module_param_named(autoconf, ipv6_defaults.autoconf, int, 0444);
94 MODULE_PARM_DESC(autoconf, "Enable IPv6 address autoconfiguration on all interfaces");
95 
96 bool ipv6_mod_enabled(void)
97 {
98 	return disable_ipv6_mod == 0;
99 }
100 EXPORT_SYMBOL_GPL(ipv6_mod_enabled);
101 
102 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
103 {
104 	const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
105 
106 	return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
107 }
108 
109 static int inet6_create(struct net *net, struct socket *sock, int protocol,
110 			int kern)
111 {
112 	struct inet_sock *inet;
113 	struct ipv6_pinfo *np;
114 	struct sock *sk;
115 	struct inet_protosw *answer;
116 	struct proto *answer_prot;
117 	unsigned char answer_flags;
118 	int try_loading_module = 0;
119 	int err;
120 
121 	if (protocol < 0 || protocol >= IPPROTO_MAX)
122 		return -EINVAL;
123 
124 	/* Look for the requested type/protocol pair. */
125 lookup_protocol:
126 	err = -ESOCKTNOSUPPORT;
127 	rcu_read_lock();
128 	list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
129 
130 		err = 0;
131 		/* Check the non-wild match. */
132 		if (protocol == answer->protocol) {
133 			if (protocol != IPPROTO_IP)
134 				break;
135 		} else {
136 			/* Check for the two wild cases. */
137 			if (IPPROTO_IP == protocol) {
138 				protocol = answer->protocol;
139 				break;
140 			}
141 			if (IPPROTO_IP == answer->protocol)
142 				break;
143 		}
144 		err = -EPROTONOSUPPORT;
145 	}
146 
147 	if (err) {
148 		if (try_loading_module < 2) {
149 			rcu_read_unlock();
150 			/*
151 			 * Be more specific, e.g. net-pf-10-proto-132-type-1
152 			 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
153 			 */
154 			if (++try_loading_module == 1)
155 				request_module("net-pf-%d-proto-%d-type-%d",
156 						PF_INET6, protocol, sock->type);
157 			/*
158 			 * Fall back to generic, e.g. net-pf-10-proto-132
159 			 * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
160 			 */
161 			else
162 				request_module("net-pf-%d-proto-%d",
163 						PF_INET6, protocol);
164 			goto lookup_protocol;
165 		} else
166 			goto out_rcu_unlock;
167 	}
168 
169 	err = -EPERM;
170 	if (sock->type == SOCK_RAW && !kern &&
171 	    !ns_capable(net->user_ns, CAP_NET_RAW))
172 		goto out_rcu_unlock;
173 
174 	sock->ops = answer->ops;
175 	answer_prot = answer->prot;
176 	answer_flags = answer->flags;
177 	rcu_read_unlock();
178 
179 	WARN_ON(!answer_prot->slab);
180 
181 	err = -ENOBUFS;
182 	sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot, kern);
183 	if (!sk)
184 		goto out;
185 
186 	sock_init_data(sock, sk);
187 
188 	err = 0;
189 	if (INET_PROTOSW_REUSE & answer_flags)
190 		sk->sk_reuse = SK_CAN_REUSE;
191 
192 	inet = inet_sk(sk);
193 	inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
194 
195 	if (SOCK_RAW == sock->type) {
196 		inet->inet_num = protocol;
197 		if (IPPROTO_RAW == protocol)
198 			inet->hdrincl = 1;
199 	}
200 
201 	sk->sk_destruct		= inet_sock_destruct;
202 	sk->sk_family		= PF_INET6;
203 	sk->sk_protocol		= protocol;
204 
205 	sk->sk_backlog_rcv	= answer->prot->backlog_rcv;
206 
207 	inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
208 	np->hop_limit	= -1;
209 	np->mcast_hops	= IPV6_DEFAULT_MCASTHOPS;
210 	np->mc_loop	= 1;
211 	np->pmtudisc	= IPV6_PMTUDISC_WANT;
212 	np->autoflowlabel = ip6_default_np_autolabel(sock_net(sk));
213 	sk->sk_ipv6only	= net->ipv6.sysctl.bindv6only;
214 
215 	/* Init the ipv4 part of the socket since we can have sockets
216 	 * using v6 API for ipv4.
217 	 */
218 	inet->uc_ttl	= -1;
219 
220 	inet->mc_loop	= 1;
221 	inet->mc_ttl	= 1;
222 	inet->mc_index	= 0;
223 	inet->mc_list	= NULL;
224 	inet->rcv_tos	= 0;
225 
226 	if (net->ipv4.sysctl_ip_no_pmtu_disc)
227 		inet->pmtudisc = IP_PMTUDISC_DONT;
228 	else
229 		inet->pmtudisc = IP_PMTUDISC_WANT;
230 	/*
231 	 * Increment only the relevant sk_prot->socks debug field, this changes
232 	 * the previous behaviour of incrementing both the equivalent to
233 	 * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
234 	 *
235 	 * This allows better debug granularity as we'll know exactly how many
236 	 * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
237 	 * transport protocol socks. -acme
238 	 */
239 	sk_refcnt_debug_inc(sk);
240 
241 	if (inet->inet_num) {
242 		/* It assumes that any protocol which allows
243 		 * the user to assign a number at socket
244 		 * creation time automatically shares.
245 		 */
246 		inet->inet_sport = htons(inet->inet_num);
247 		err = sk->sk_prot->hash(sk);
248 		if (err) {
249 			sk_common_release(sk);
250 			goto out;
251 		}
252 	}
253 	if (sk->sk_prot->init) {
254 		err = sk->sk_prot->init(sk);
255 		if (err) {
256 			sk_common_release(sk);
257 			goto out;
258 		}
259 	}
260 out:
261 	return err;
262 out_rcu_unlock:
263 	rcu_read_unlock();
264 	goto out;
265 }
266 
267 
268 /* bind for INET6 API */
269 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
270 {
271 	struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr;
272 	struct sock *sk = sock->sk;
273 	struct inet_sock *inet = inet_sk(sk);
274 	struct ipv6_pinfo *np = inet6_sk(sk);
275 	struct net *net = sock_net(sk);
276 	__be32 v4addr = 0;
277 	unsigned short snum;
278 	int addr_type = 0;
279 	int err = 0;
280 
281 	/* If the socket has its own bind function then use it. */
282 	if (sk->sk_prot->bind)
283 		return sk->sk_prot->bind(sk, uaddr, addr_len);
284 
285 	if (addr_len < SIN6_LEN_RFC2133)
286 		return -EINVAL;
287 
288 	if (addr->sin6_family != AF_INET6)
289 		return -EAFNOSUPPORT;
290 
291 	addr_type = ipv6_addr_type(&addr->sin6_addr);
292 	if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
293 		return -EINVAL;
294 
295 	snum = ntohs(addr->sin6_port);
296 	if (snum && snum < PROT_SOCK && !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
297 		return -EACCES;
298 
299 	lock_sock(sk);
300 
301 	/* Check these errors (active socket, double bind). */
302 	if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
303 		err = -EINVAL;
304 		goto out;
305 	}
306 
307 	/* Check if the address belongs to the host. */
308 	if (addr_type == IPV6_ADDR_MAPPED) {
309 		int chk_addr_ret;
310 
311 		/* Binding to v4-mapped address on a v6-only socket
312 		 * makes no sense
313 		 */
314 		if (sk->sk_ipv6only) {
315 			err = -EINVAL;
316 			goto out;
317 		}
318 
319 		/* Reproduce AF_INET checks to make the bindings consistent */
320 		v4addr = addr->sin6_addr.s6_addr32[3];
321 		chk_addr_ret = inet_addr_type(net, v4addr);
322 		if (!net->ipv4.sysctl_ip_nonlocal_bind &&
323 		    !(inet->freebind || inet->transparent) &&
324 		    v4addr != htonl(INADDR_ANY) &&
325 		    chk_addr_ret != RTN_LOCAL &&
326 		    chk_addr_ret != RTN_MULTICAST &&
327 		    chk_addr_ret != RTN_BROADCAST) {
328 			err = -EADDRNOTAVAIL;
329 			goto out;
330 		}
331 	} else {
332 		if (addr_type != IPV6_ADDR_ANY) {
333 			struct net_device *dev = NULL;
334 
335 			rcu_read_lock();
336 			if (__ipv6_addr_needs_scope_id(addr_type)) {
337 				if (addr_len >= sizeof(struct sockaddr_in6) &&
338 				    addr->sin6_scope_id) {
339 					/* Override any existing binding, if another one
340 					 * is supplied by user.
341 					 */
342 					sk->sk_bound_dev_if = addr->sin6_scope_id;
343 				}
344 
345 				/* Binding to link-local address requires an interface */
346 				if (!sk->sk_bound_dev_if) {
347 					err = -EINVAL;
348 					goto out_unlock;
349 				}
350 				dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
351 				if (!dev) {
352 					err = -ENODEV;
353 					goto out_unlock;
354 				}
355 			}
356 
357 			/* ipv4 addr of the socket is invalid.  Only the
358 			 * unspecified and mapped address have a v4 equivalent.
359 			 */
360 			v4addr = LOOPBACK4_IPV6;
361 			if (!(addr_type & IPV6_ADDR_MULTICAST))	{
362 				if (!net->ipv6.sysctl.ip_nonlocal_bind &&
363 				    !(inet->freebind || inet->transparent) &&
364 				    !ipv6_chk_addr(net, &addr->sin6_addr,
365 						   dev, 0)) {
366 					err = -EADDRNOTAVAIL;
367 					goto out_unlock;
368 				}
369 			}
370 			rcu_read_unlock();
371 		}
372 	}
373 
374 	inet->inet_rcv_saddr = v4addr;
375 	inet->inet_saddr = v4addr;
376 
377 	sk->sk_v6_rcv_saddr = addr->sin6_addr;
378 
379 	if (!(addr_type & IPV6_ADDR_MULTICAST))
380 		np->saddr = addr->sin6_addr;
381 
382 	/* Make sure we are allowed to bind here. */
383 	if ((snum || !inet->bind_address_no_port) &&
384 	    sk->sk_prot->get_port(sk, snum)) {
385 		inet_reset_saddr(sk);
386 		err = -EADDRINUSE;
387 		goto out;
388 	}
389 
390 	if (addr_type != IPV6_ADDR_ANY) {
391 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
392 		if (addr_type != IPV6_ADDR_MAPPED)
393 			sk->sk_ipv6only = 1;
394 	}
395 	if (snum)
396 		sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
397 	inet->inet_sport = htons(inet->inet_num);
398 	inet->inet_dport = 0;
399 	inet->inet_daddr = 0;
400 out:
401 	release_sock(sk);
402 	return err;
403 out_unlock:
404 	rcu_read_unlock();
405 	goto out;
406 }
407 EXPORT_SYMBOL(inet6_bind);
408 
409 int inet6_release(struct socket *sock)
410 {
411 	struct sock *sk = sock->sk;
412 
413 	if (!sk)
414 		return -EINVAL;
415 
416 	/* Free mc lists */
417 	ipv6_sock_mc_close(sk);
418 
419 	/* Free ac lists */
420 	ipv6_sock_ac_close(sk);
421 
422 	return inet_release(sock);
423 }
424 EXPORT_SYMBOL(inet6_release);
425 
426 void inet6_destroy_sock(struct sock *sk)
427 {
428 	struct ipv6_pinfo *np = inet6_sk(sk);
429 	struct sk_buff *skb;
430 	struct ipv6_txoptions *opt;
431 
432 	/* Release rx options */
433 
434 	skb = xchg(&np->pktoptions, NULL);
435 	if (skb)
436 		kfree_skb(skb);
437 
438 	skb = xchg(&np->rxpmtu, NULL);
439 	if (skb)
440 		kfree_skb(skb);
441 
442 	/* Free flowlabels */
443 	fl6_free_socklist(sk);
444 
445 	/* Free tx options */
446 
447 	opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL);
448 	if (opt) {
449 		atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
450 		txopt_put(opt);
451 	}
452 }
453 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
454 
455 /*
456  *	This does both peername and sockname.
457  */
458 
459 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
460 		 int *uaddr_len, int peer)
461 {
462 	struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
463 	struct sock *sk = sock->sk;
464 	struct inet_sock *inet = inet_sk(sk);
465 	struct ipv6_pinfo *np = inet6_sk(sk);
466 
467 	sin->sin6_family = AF_INET6;
468 	sin->sin6_flowinfo = 0;
469 	sin->sin6_scope_id = 0;
470 	if (peer) {
471 		if (!inet->inet_dport)
472 			return -ENOTCONN;
473 		if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
474 		    peer == 1)
475 			return -ENOTCONN;
476 		sin->sin6_port = inet->inet_dport;
477 		sin->sin6_addr = sk->sk_v6_daddr;
478 		if (np->sndflow)
479 			sin->sin6_flowinfo = np->flow_label;
480 	} else {
481 		if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
482 			sin->sin6_addr = np->saddr;
483 		else
484 			sin->sin6_addr = sk->sk_v6_rcv_saddr;
485 
486 		sin->sin6_port = inet->inet_sport;
487 	}
488 	sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
489 						 sk->sk_bound_dev_if);
490 	*uaddr_len = sizeof(*sin);
491 	return 0;
492 }
493 EXPORT_SYMBOL(inet6_getname);
494 
495 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
496 {
497 	struct sock *sk = sock->sk;
498 	struct net *net = sock_net(sk);
499 
500 	switch (cmd) {
501 	case SIOCGSTAMP:
502 		return sock_get_timestamp(sk, (struct timeval __user *)arg);
503 
504 	case SIOCGSTAMPNS:
505 		return sock_get_timestampns(sk, (struct timespec __user *)arg);
506 
507 	case SIOCADDRT:
508 	case SIOCDELRT:
509 
510 		return ipv6_route_ioctl(net, cmd, (void __user *)arg);
511 
512 	case SIOCSIFADDR:
513 		return addrconf_add_ifaddr(net, (void __user *) arg);
514 	case SIOCDIFADDR:
515 		return addrconf_del_ifaddr(net, (void __user *) arg);
516 	case SIOCSIFDSTADDR:
517 		return addrconf_set_dstaddr(net, (void __user *) arg);
518 	default:
519 		if (!sk->sk_prot->ioctl)
520 			return -ENOIOCTLCMD;
521 		return sk->sk_prot->ioctl(sk, cmd, arg);
522 	}
523 	/*NOTREACHED*/
524 	return 0;
525 }
526 EXPORT_SYMBOL(inet6_ioctl);
527 
528 const struct proto_ops inet6_stream_ops = {
529 	.family		   = PF_INET6,
530 	.owner		   = THIS_MODULE,
531 	.release	   = inet6_release,
532 	.bind		   = inet6_bind,
533 	.connect	   = inet_stream_connect,	/* ok		*/
534 	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
535 	.accept		   = inet_accept,		/* ok		*/
536 	.getname	   = inet6_getname,
537 	.poll		   = tcp_poll,			/* ok		*/
538 	.ioctl		   = inet6_ioctl,		/* must change  */
539 	.listen		   = inet_listen,		/* ok		*/
540 	.shutdown	   = inet_shutdown,		/* ok		*/
541 	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
542 	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
543 	.sendmsg	   = inet_sendmsg,		/* ok		*/
544 	.recvmsg	   = inet_recvmsg,		/* ok		*/
545 	.mmap		   = sock_no_mmap,
546 	.sendpage	   = inet_sendpage,
547 	.splice_read	   = tcp_splice_read,
548 	.read_sock	   = tcp_read_sock,
549 	.peek_len	   = tcp_peek_len,
550 #ifdef CONFIG_COMPAT
551 	.compat_setsockopt = compat_sock_common_setsockopt,
552 	.compat_getsockopt = compat_sock_common_getsockopt,
553 #endif
554 };
555 
556 const struct proto_ops inet6_dgram_ops = {
557 	.family		   = PF_INET6,
558 	.owner		   = THIS_MODULE,
559 	.release	   = inet6_release,
560 	.bind		   = inet6_bind,
561 	.connect	   = inet_dgram_connect,	/* ok		*/
562 	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
563 	.accept		   = sock_no_accept,		/* a do nothing	*/
564 	.getname	   = inet6_getname,
565 	.poll		   = udp_poll,			/* ok		*/
566 	.ioctl		   = inet6_ioctl,		/* must change  */
567 	.listen		   = sock_no_listen,		/* ok		*/
568 	.shutdown	   = inet_shutdown,		/* ok		*/
569 	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
570 	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
571 	.sendmsg	   = inet_sendmsg,		/* ok		*/
572 	.recvmsg	   = inet_recvmsg,		/* ok		*/
573 	.mmap		   = sock_no_mmap,
574 	.sendpage	   = sock_no_sendpage,
575 	.set_peek_off	   = sk_set_peek_off,
576 #ifdef CONFIG_COMPAT
577 	.compat_setsockopt = compat_sock_common_setsockopt,
578 	.compat_getsockopt = compat_sock_common_getsockopt,
579 #endif
580 };
581 
582 static const struct net_proto_family inet6_family_ops = {
583 	.family = PF_INET6,
584 	.create = inet6_create,
585 	.owner	= THIS_MODULE,
586 };
587 
588 int inet6_register_protosw(struct inet_protosw *p)
589 {
590 	struct list_head *lh;
591 	struct inet_protosw *answer;
592 	struct list_head *last_perm;
593 	int protocol = p->protocol;
594 	int ret;
595 
596 	spin_lock_bh(&inetsw6_lock);
597 
598 	ret = -EINVAL;
599 	if (p->type >= SOCK_MAX)
600 		goto out_illegal;
601 
602 	/* If we are trying to override a permanent protocol, bail. */
603 	answer = NULL;
604 	ret = -EPERM;
605 	last_perm = &inetsw6[p->type];
606 	list_for_each(lh, &inetsw6[p->type]) {
607 		answer = list_entry(lh, struct inet_protosw, list);
608 
609 		/* Check only the non-wild match. */
610 		if (INET_PROTOSW_PERMANENT & answer->flags) {
611 			if (protocol == answer->protocol)
612 				break;
613 			last_perm = lh;
614 		}
615 
616 		answer = NULL;
617 	}
618 	if (answer)
619 		goto out_permanent;
620 
621 	/* Add the new entry after the last permanent entry if any, so that
622 	 * the new entry does not override a permanent entry when matched with
623 	 * a wild-card protocol. But it is allowed to override any existing
624 	 * non-permanent entry.  This means that when we remove this entry, the
625 	 * system automatically returns to the old behavior.
626 	 */
627 	list_add_rcu(&p->list, last_perm);
628 	ret = 0;
629 out:
630 	spin_unlock_bh(&inetsw6_lock);
631 	return ret;
632 
633 out_permanent:
634 	pr_err("Attempt to override permanent protocol %d\n", protocol);
635 	goto out;
636 
637 out_illegal:
638 	pr_err("Ignoring attempt to register invalid socket type %d\n",
639 	       p->type);
640 	goto out;
641 }
642 EXPORT_SYMBOL(inet6_register_protosw);
643 
644 void
645 inet6_unregister_protosw(struct inet_protosw *p)
646 {
647 	if (INET_PROTOSW_PERMANENT & p->flags) {
648 		pr_err("Attempt to unregister permanent protocol %d\n",
649 		       p->protocol);
650 	} else {
651 		spin_lock_bh(&inetsw6_lock);
652 		list_del_rcu(&p->list);
653 		spin_unlock_bh(&inetsw6_lock);
654 
655 		synchronize_net();
656 	}
657 }
658 EXPORT_SYMBOL(inet6_unregister_protosw);
659 
660 int inet6_sk_rebuild_header(struct sock *sk)
661 {
662 	struct ipv6_pinfo *np = inet6_sk(sk);
663 	struct dst_entry *dst;
664 
665 	dst = __sk_dst_check(sk, np->dst_cookie);
666 
667 	if (!dst) {
668 		struct inet_sock *inet = inet_sk(sk);
669 		struct in6_addr *final_p, final;
670 		struct flowi6 fl6;
671 
672 		memset(&fl6, 0, sizeof(fl6));
673 		fl6.flowi6_proto = sk->sk_protocol;
674 		fl6.daddr = sk->sk_v6_daddr;
675 		fl6.saddr = np->saddr;
676 		fl6.flowlabel = np->flow_label;
677 		fl6.flowi6_oif = sk->sk_bound_dev_if;
678 		fl6.flowi6_mark = sk->sk_mark;
679 		fl6.fl6_dport = inet->inet_dport;
680 		fl6.fl6_sport = inet->inet_sport;
681 		security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
682 
683 		rcu_read_lock();
684 		final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt),
685 					 &final);
686 		rcu_read_unlock();
687 
688 		dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
689 		if (IS_ERR(dst)) {
690 			sk->sk_route_caps = 0;
691 			sk->sk_err_soft = -PTR_ERR(dst);
692 			return PTR_ERR(dst);
693 		}
694 
695 		ip6_dst_store(sk, dst, NULL, NULL);
696 	}
697 
698 	return 0;
699 }
700 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
701 
702 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb,
703 		       const struct inet6_skb_parm *opt)
704 {
705 	const struct ipv6_pinfo *np = inet6_sk(sk);
706 
707 	if (np->rxopt.all) {
708 		if (((opt->flags & IP6SKB_HOPBYHOP) &&
709 		     (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) ||
710 		    (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) &&
711 		     np->rxopt.bits.rxflow) ||
712 		    (opt->srcrt && (np->rxopt.bits.srcrt ||
713 		     np->rxopt.bits.osrcrt)) ||
714 		    ((opt->dst1 || opt->dst0) &&
715 		     (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
716 			return true;
717 	}
718 	return false;
719 }
720 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
721 
722 static struct packet_type ipv6_packet_type __read_mostly = {
723 	.type = cpu_to_be16(ETH_P_IPV6),
724 	.func = ipv6_rcv,
725 };
726 
727 static int __init ipv6_packet_init(void)
728 {
729 	dev_add_pack(&ipv6_packet_type);
730 	return 0;
731 }
732 
733 static void ipv6_packet_cleanup(void)
734 {
735 	dev_remove_pack(&ipv6_packet_type);
736 }
737 
738 static int __net_init ipv6_init_mibs(struct net *net)
739 {
740 	int i;
741 
742 	net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib);
743 	if (!net->mib.udp_stats_in6)
744 		return -ENOMEM;
745 	net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib);
746 	if (!net->mib.udplite_stats_in6)
747 		goto err_udplite_mib;
748 	net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib);
749 	if (!net->mib.ipv6_statistics)
750 		goto err_ip_mib;
751 
752 	for_each_possible_cpu(i) {
753 		struct ipstats_mib *af_inet6_stats;
754 		af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i);
755 		u64_stats_init(&af_inet6_stats->syncp);
756 	}
757 
758 
759 	net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib);
760 	if (!net->mib.icmpv6_statistics)
761 		goto err_icmp_mib;
762 	net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
763 						GFP_KERNEL);
764 	if (!net->mib.icmpv6msg_statistics)
765 		goto err_icmpmsg_mib;
766 	return 0;
767 
768 err_icmpmsg_mib:
769 	free_percpu(net->mib.icmpv6_statistics);
770 err_icmp_mib:
771 	free_percpu(net->mib.ipv6_statistics);
772 err_ip_mib:
773 	free_percpu(net->mib.udplite_stats_in6);
774 err_udplite_mib:
775 	free_percpu(net->mib.udp_stats_in6);
776 	return -ENOMEM;
777 }
778 
779 static void ipv6_cleanup_mibs(struct net *net)
780 {
781 	free_percpu(net->mib.udp_stats_in6);
782 	free_percpu(net->mib.udplite_stats_in6);
783 	free_percpu(net->mib.ipv6_statistics);
784 	free_percpu(net->mib.icmpv6_statistics);
785 	kfree(net->mib.icmpv6msg_statistics);
786 }
787 
788 static int __net_init inet6_net_init(struct net *net)
789 {
790 	int err = 0;
791 
792 	net->ipv6.sysctl.bindv6only = 0;
793 	net->ipv6.sysctl.icmpv6_time = 1*HZ;
794 	net->ipv6.sysctl.flowlabel_consistency = 1;
795 	net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS;
796 	net->ipv6.sysctl.idgen_retries = 3;
797 	net->ipv6.sysctl.idgen_delay = 1 * HZ;
798 	net->ipv6.sysctl.flowlabel_state_ranges = 0;
799 	atomic_set(&net->ipv6.fib6_sernum, 1);
800 
801 	err = ipv6_init_mibs(net);
802 	if (err)
803 		return err;
804 #ifdef CONFIG_PROC_FS
805 	err = udp6_proc_init(net);
806 	if (err)
807 		goto out;
808 	err = tcp6_proc_init(net);
809 	if (err)
810 		goto proc_tcp6_fail;
811 	err = ac6_proc_init(net);
812 	if (err)
813 		goto proc_ac6_fail;
814 #endif
815 	return err;
816 
817 #ifdef CONFIG_PROC_FS
818 proc_ac6_fail:
819 	tcp6_proc_exit(net);
820 proc_tcp6_fail:
821 	udp6_proc_exit(net);
822 out:
823 	ipv6_cleanup_mibs(net);
824 	return err;
825 #endif
826 }
827 
828 static void __net_exit inet6_net_exit(struct net *net)
829 {
830 #ifdef CONFIG_PROC_FS
831 	udp6_proc_exit(net);
832 	tcp6_proc_exit(net);
833 	ac6_proc_exit(net);
834 #endif
835 	ipv6_cleanup_mibs(net);
836 }
837 
838 static struct pernet_operations inet6_net_ops = {
839 	.init = inet6_net_init,
840 	.exit = inet6_net_exit,
841 };
842 
843 static const struct ipv6_stub ipv6_stub_impl = {
844 	.ipv6_sock_mc_join = ipv6_sock_mc_join,
845 	.ipv6_sock_mc_drop = ipv6_sock_mc_drop,
846 	.ipv6_dst_lookup = ip6_dst_lookup,
847 	.udpv6_encap_enable = udpv6_encap_enable,
848 	.ndisc_send_na = ndisc_send_na,
849 	.nd_tbl	= &nd_tbl,
850 };
851 
852 static int __init inet6_init(void)
853 {
854 	struct list_head *r;
855 	int err = 0;
856 
857 	sock_skb_cb_check_size(sizeof(struct inet6_skb_parm));
858 
859 	/* Register the socket-side information for inet6_create.  */
860 	for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
861 		INIT_LIST_HEAD(r);
862 
863 	if (disable_ipv6_mod) {
864 		pr_info("Loaded, but administratively disabled, reboot required to enable\n");
865 		goto out;
866 	}
867 
868 	err = proto_register(&tcpv6_prot, 1);
869 	if (err)
870 		goto out;
871 
872 	err = proto_register(&udpv6_prot, 1);
873 	if (err)
874 		goto out_unregister_tcp_proto;
875 
876 	err = proto_register(&udplitev6_prot, 1);
877 	if (err)
878 		goto out_unregister_udp_proto;
879 
880 	err = proto_register(&rawv6_prot, 1);
881 	if (err)
882 		goto out_unregister_udplite_proto;
883 
884 	err = proto_register(&pingv6_prot, 1);
885 	if (err)
886 		goto out_unregister_ping_proto;
887 
888 	/* We MUST register RAW sockets before we create the ICMP6,
889 	 * IGMP6, or NDISC control sockets.
890 	 */
891 	err = rawv6_init();
892 	if (err)
893 		goto out_unregister_raw_proto;
894 
895 	/* Register the family here so that the init calls below will
896 	 * be able to create sockets. (?? is this dangerous ??)
897 	 */
898 	err = sock_register(&inet6_family_ops);
899 	if (err)
900 		goto out_sock_register_fail;
901 
902 	/*
903 	 *	ipngwg API draft makes clear that the correct semantics
904 	 *	for TCP and UDP is to consider one TCP and UDP instance
905 	 *	in a host available by both INET and INET6 APIs and
906 	 *	able to communicate via both network protocols.
907 	 */
908 
909 	err = register_pernet_subsys(&inet6_net_ops);
910 	if (err)
911 		goto register_pernet_fail;
912 	err = icmpv6_init();
913 	if (err)
914 		goto icmp_fail;
915 	err = ip6_mr_init();
916 	if (err)
917 		goto ipmr_fail;
918 	err = ndisc_init();
919 	if (err)
920 		goto ndisc_fail;
921 	err = igmp6_init();
922 	if (err)
923 		goto igmp_fail;
924 
925 	ipv6_stub = &ipv6_stub_impl;
926 
927 	err = ipv6_netfilter_init();
928 	if (err)
929 		goto netfilter_fail;
930 	/* Create /proc/foo6 entries. */
931 #ifdef CONFIG_PROC_FS
932 	err = -ENOMEM;
933 	if (raw6_proc_init())
934 		goto proc_raw6_fail;
935 	if (udplite6_proc_init())
936 		goto proc_udplite6_fail;
937 	if (ipv6_misc_proc_init())
938 		goto proc_misc6_fail;
939 	if (if6_proc_init())
940 		goto proc_if6_fail;
941 #endif
942 	err = ip6_route_init();
943 	if (err)
944 		goto ip6_route_fail;
945 	err = ndisc_late_init();
946 	if (err)
947 		goto ndisc_late_fail;
948 	err = ip6_flowlabel_init();
949 	if (err)
950 		goto ip6_flowlabel_fail;
951 	err = addrconf_init();
952 	if (err)
953 		goto addrconf_fail;
954 
955 	/* Init v6 extension headers. */
956 	err = ipv6_exthdrs_init();
957 	if (err)
958 		goto ipv6_exthdrs_fail;
959 
960 	err = ipv6_frag_init();
961 	if (err)
962 		goto ipv6_frag_fail;
963 
964 	/* Init v6 transport protocols. */
965 	err = udpv6_init();
966 	if (err)
967 		goto udpv6_fail;
968 
969 	err = udplitev6_init();
970 	if (err)
971 		goto udplitev6_fail;
972 
973 	err = udpv6_offload_init();
974 	if (err)
975 		goto udpv6_offload_fail;
976 
977 	err = tcpv6_init();
978 	if (err)
979 		goto tcpv6_fail;
980 
981 	err = ipv6_packet_init();
982 	if (err)
983 		goto ipv6_packet_fail;
984 
985 	err = pingv6_init();
986 	if (err)
987 		goto pingv6_fail;
988 
989 	err = calipso_init();
990 	if (err)
991 		goto calipso_fail;
992 
993 #ifdef CONFIG_SYSCTL
994 	err = ipv6_sysctl_register();
995 	if (err)
996 		goto sysctl_fail;
997 #endif
998 out:
999 	return err;
1000 
1001 #ifdef CONFIG_SYSCTL
1002 sysctl_fail:
1003 	calipso_exit();
1004 #endif
1005 calipso_fail:
1006 	pingv6_exit();
1007 pingv6_fail:
1008 	ipv6_packet_cleanup();
1009 ipv6_packet_fail:
1010 	tcpv6_exit();
1011 tcpv6_fail:
1012 	udpv6_offload_exit();
1013 udpv6_offload_fail:
1014 	udplitev6_exit();
1015 udplitev6_fail:
1016 	udpv6_exit();
1017 udpv6_fail:
1018 	ipv6_frag_exit();
1019 ipv6_frag_fail:
1020 	ipv6_exthdrs_exit();
1021 ipv6_exthdrs_fail:
1022 	addrconf_cleanup();
1023 addrconf_fail:
1024 	ip6_flowlabel_cleanup();
1025 ip6_flowlabel_fail:
1026 	ndisc_late_cleanup();
1027 ndisc_late_fail:
1028 	ip6_route_cleanup();
1029 ip6_route_fail:
1030 #ifdef CONFIG_PROC_FS
1031 	if6_proc_exit();
1032 proc_if6_fail:
1033 	ipv6_misc_proc_exit();
1034 proc_misc6_fail:
1035 	udplite6_proc_exit();
1036 proc_udplite6_fail:
1037 	raw6_proc_exit();
1038 proc_raw6_fail:
1039 #endif
1040 	ipv6_netfilter_fini();
1041 netfilter_fail:
1042 	igmp6_cleanup();
1043 igmp_fail:
1044 	ndisc_cleanup();
1045 ndisc_fail:
1046 	ip6_mr_cleanup();
1047 ipmr_fail:
1048 	icmpv6_cleanup();
1049 icmp_fail:
1050 	unregister_pernet_subsys(&inet6_net_ops);
1051 register_pernet_fail:
1052 	sock_unregister(PF_INET6);
1053 	rtnl_unregister_all(PF_INET6);
1054 out_sock_register_fail:
1055 	rawv6_exit();
1056 out_unregister_ping_proto:
1057 	proto_unregister(&pingv6_prot);
1058 out_unregister_raw_proto:
1059 	proto_unregister(&rawv6_prot);
1060 out_unregister_udplite_proto:
1061 	proto_unregister(&udplitev6_prot);
1062 out_unregister_udp_proto:
1063 	proto_unregister(&udpv6_prot);
1064 out_unregister_tcp_proto:
1065 	proto_unregister(&tcpv6_prot);
1066 	goto out;
1067 }
1068 module_init(inet6_init);
1069 
1070 MODULE_ALIAS_NETPROTO(PF_INET6);
1071