xref: /openbmc/linux/net/ipv6/af_inet6.c (revision c127f98ba9aba1818a6ca3a1da5a24653a10d966)
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 	int addr_type = 0;
288 	int err = 0;
289 
290 	/* If the socket has its own bind function then use it. */
291 	if (sk->sk_prot->bind)
292 		return sk->sk_prot->bind(sk, uaddr, addr_len);
293 
294 	if (addr_len < SIN6_LEN_RFC2133)
295 		return -EINVAL;
296 
297 	if (addr->sin6_family != AF_INET6)
298 		return -EAFNOSUPPORT;
299 
300 	addr_type = ipv6_addr_type(&addr->sin6_addr);
301 	if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
302 		return -EINVAL;
303 
304 	snum = ntohs(addr->sin6_port);
305 	if (snum && snum < inet_prot_sock(net) &&
306 	    !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
307 		return -EACCES;
308 
309 	lock_sock(sk);
310 
311 	/* Check these errors (active socket, double bind). */
312 	if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
313 		err = -EINVAL;
314 		goto out;
315 	}
316 
317 	/* Check if the address belongs to the host. */
318 	if (addr_type == IPV6_ADDR_MAPPED) {
319 		int chk_addr_ret;
320 
321 		/* Binding to v4-mapped address on a v6-only socket
322 		 * makes no sense
323 		 */
324 		if (sk->sk_ipv6only) {
325 			err = -EINVAL;
326 			goto out;
327 		}
328 
329 		/* Reproduce AF_INET checks to make the bindings consistent */
330 		v4addr = addr->sin6_addr.s6_addr32[3];
331 		chk_addr_ret = inet_addr_type(net, v4addr);
332 		if (!net->ipv4.sysctl_ip_nonlocal_bind &&
333 		    !(inet->freebind || inet->transparent) &&
334 		    v4addr != htonl(INADDR_ANY) &&
335 		    chk_addr_ret != RTN_LOCAL &&
336 		    chk_addr_ret != RTN_MULTICAST &&
337 		    chk_addr_ret != RTN_BROADCAST) {
338 			err = -EADDRNOTAVAIL;
339 			goto out;
340 		}
341 	} else {
342 		if (addr_type != IPV6_ADDR_ANY) {
343 			struct net_device *dev = NULL;
344 
345 			rcu_read_lock();
346 			if (__ipv6_addr_needs_scope_id(addr_type)) {
347 				if (addr_len >= sizeof(struct sockaddr_in6) &&
348 				    addr->sin6_scope_id) {
349 					/* Override any existing binding, if another one
350 					 * is supplied by user.
351 					 */
352 					sk->sk_bound_dev_if = addr->sin6_scope_id;
353 				}
354 
355 				/* Binding to link-local address requires an interface */
356 				if (!sk->sk_bound_dev_if) {
357 					err = -EINVAL;
358 					goto out_unlock;
359 				}
360 				dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
361 				if (!dev) {
362 					err = -ENODEV;
363 					goto out_unlock;
364 				}
365 			}
366 
367 			/* ipv4 addr of the socket is invalid.  Only the
368 			 * unspecified and mapped address have a v4 equivalent.
369 			 */
370 			v4addr = LOOPBACK4_IPV6;
371 			if (!(addr_type & IPV6_ADDR_MULTICAST))	{
372 				if (!net->ipv6.sysctl.ip_nonlocal_bind &&
373 				    !(inet->freebind || inet->transparent) &&
374 				    !ipv6_chk_addr(net, &addr->sin6_addr,
375 						   dev, 0)) {
376 					err = -EADDRNOTAVAIL;
377 					goto out_unlock;
378 				}
379 			}
380 			rcu_read_unlock();
381 		}
382 	}
383 
384 	inet->inet_rcv_saddr = v4addr;
385 	inet->inet_saddr = v4addr;
386 
387 	sk->sk_v6_rcv_saddr = addr->sin6_addr;
388 
389 	if (!(addr_type & IPV6_ADDR_MULTICAST))
390 		np->saddr = addr->sin6_addr;
391 
392 	/* Make sure we are allowed to bind here. */
393 	if ((snum || !inet->bind_address_no_port) &&
394 	    sk->sk_prot->get_port(sk, snum)) {
395 		inet_reset_saddr(sk);
396 		err = -EADDRINUSE;
397 		goto out;
398 	}
399 
400 	if (addr_type != IPV6_ADDR_ANY) {
401 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
402 		if (addr_type != IPV6_ADDR_MAPPED)
403 			sk->sk_ipv6only = 1;
404 	}
405 	if (snum)
406 		sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
407 	inet->inet_sport = htons(inet->inet_num);
408 	inet->inet_dport = 0;
409 	inet->inet_daddr = 0;
410 out:
411 	release_sock(sk);
412 	return err;
413 out_unlock:
414 	rcu_read_unlock();
415 	goto out;
416 }
417 EXPORT_SYMBOL(inet6_bind);
418 
419 int inet6_release(struct socket *sock)
420 {
421 	struct sock *sk = sock->sk;
422 
423 	if (!sk)
424 		return -EINVAL;
425 
426 	/* Free mc lists */
427 	ipv6_sock_mc_close(sk);
428 
429 	/* Free ac lists */
430 	ipv6_sock_ac_close(sk);
431 
432 	return inet_release(sock);
433 }
434 EXPORT_SYMBOL(inet6_release);
435 
436 void inet6_destroy_sock(struct sock *sk)
437 {
438 	struct ipv6_pinfo *np = inet6_sk(sk);
439 	struct sk_buff *skb;
440 	struct ipv6_txoptions *opt;
441 
442 	/* Release rx options */
443 
444 	skb = xchg(&np->pktoptions, NULL);
445 	if (skb)
446 		kfree_skb(skb);
447 
448 	skb = xchg(&np->rxpmtu, NULL);
449 	if (skb)
450 		kfree_skb(skb);
451 
452 	/* Free flowlabels */
453 	fl6_free_socklist(sk);
454 
455 	/* Free tx options */
456 
457 	opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL);
458 	if (opt) {
459 		atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
460 		txopt_put(opt);
461 	}
462 }
463 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
464 
465 /*
466  *	This does both peername and sockname.
467  */
468 
469 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
470 		 int *uaddr_len, int peer)
471 {
472 	struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
473 	struct sock *sk = sock->sk;
474 	struct inet_sock *inet = inet_sk(sk);
475 	struct ipv6_pinfo *np = inet6_sk(sk);
476 
477 	sin->sin6_family = AF_INET6;
478 	sin->sin6_flowinfo = 0;
479 	sin->sin6_scope_id = 0;
480 	if (peer) {
481 		if (!inet->inet_dport)
482 			return -ENOTCONN;
483 		if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
484 		    peer == 1)
485 			return -ENOTCONN;
486 		sin->sin6_port = inet->inet_dport;
487 		sin->sin6_addr = sk->sk_v6_daddr;
488 		if (np->sndflow)
489 			sin->sin6_flowinfo = np->flow_label;
490 	} else {
491 		if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
492 			sin->sin6_addr = np->saddr;
493 		else
494 			sin->sin6_addr = sk->sk_v6_rcv_saddr;
495 
496 		sin->sin6_port = inet->inet_sport;
497 	}
498 	sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
499 						 sk->sk_bound_dev_if);
500 	*uaddr_len = sizeof(*sin);
501 	return 0;
502 }
503 EXPORT_SYMBOL(inet6_getname);
504 
505 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
506 {
507 	struct sock *sk = sock->sk;
508 	struct net *net = sock_net(sk);
509 
510 	switch (cmd) {
511 	case SIOCGSTAMP:
512 		return sock_get_timestamp(sk, (struct timeval __user *)arg);
513 
514 	case SIOCGSTAMPNS:
515 		return sock_get_timestampns(sk, (struct timespec __user *)arg);
516 
517 	case SIOCADDRT:
518 	case SIOCDELRT:
519 
520 		return ipv6_route_ioctl(net, cmd, (void __user *)arg);
521 
522 	case SIOCSIFADDR:
523 		return addrconf_add_ifaddr(net, (void __user *) arg);
524 	case SIOCDIFADDR:
525 		return addrconf_del_ifaddr(net, (void __user *) arg);
526 	case SIOCSIFDSTADDR:
527 		return addrconf_set_dstaddr(net, (void __user *) arg);
528 	default:
529 		if (!sk->sk_prot->ioctl)
530 			return -ENOIOCTLCMD;
531 		return sk->sk_prot->ioctl(sk, cmd, arg);
532 	}
533 	/*NOTREACHED*/
534 	return 0;
535 }
536 EXPORT_SYMBOL(inet6_ioctl);
537 
538 const struct proto_ops inet6_stream_ops = {
539 	.family		   = PF_INET6,
540 	.owner		   = THIS_MODULE,
541 	.release	   = inet6_release,
542 	.bind		   = inet6_bind,
543 	.connect	   = inet_stream_connect,	/* ok		*/
544 	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
545 	.accept		   = inet_accept,		/* ok		*/
546 	.getname	   = inet6_getname,
547 	.poll		   = tcp_poll,			/* ok		*/
548 	.ioctl		   = inet6_ioctl,		/* must change  */
549 	.listen		   = inet_listen,		/* ok		*/
550 	.shutdown	   = inet_shutdown,		/* ok		*/
551 	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
552 	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
553 	.sendmsg	   = inet_sendmsg,		/* ok		*/
554 	.recvmsg	   = inet_recvmsg,		/* ok		*/
555 	.mmap		   = sock_no_mmap,
556 	.sendpage	   = inet_sendpage,
557 	.sendmsg_locked    = tcp_sendmsg_locked,
558 	.sendpage_locked   = tcp_sendpage_locked,
559 	.splice_read	   = tcp_splice_read,
560 	.read_sock	   = tcp_read_sock,
561 	.peek_len	   = tcp_peek_len,
562 #ifdef CONFIG_COMPAT
563 	.compat_setsockopt = compat_sock_common_setsockopt,
564 	.compat_getsockopt = compat_sock_common_getsockopt,
565 #endif
566 };
567 
568 const struct proto_ops inet6_dgram_ops = {
569 	.family		   = PF_INET6,
570 	.owner		   = THIS_MODULE,
571 	.release	   = inet6_release,
572 	.bind		   = inet6_bind,
573 	.connect	   = inet_dgram_connect,	/* ok		*/
574 	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
575 	.accept		   = sock_no_accept,		/* a do nothing	*/
576 	.getname	   = inet6_getname,
577 	.poll		   = udp_poll,			/* ok		*/
578 	.ioctl		   = inet6_ioctl,		/* must change  */
579 	.listen		   = sock_no_listen,		/* ok		*/
580 	.shutdown	   = inet_shutdown,		/* ok		*/
581 	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
582 	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
583 	.sendmsg	   = inet_sendmsg,		/* ok		*/
584 	.recvmsg	   = inet_recvmsg,		/* ok		*/
585 	.mmap		   = sock_no_mmap,
586 	.sendpage	   = sock_no_sendpage,
587 	.set_peek_off	   = sk_set_peek_off,
588 #ifdef CONFIG_COMPAT
589 	.compat_setsockopt = compat_sock_common_setsockopt,
590 	.compat_getsockopt = compat_sock_common_getsockopt,
591 #endif
592 };
593 
594 static const struct net_proto_family inet6_family_ops = {
595 	.family = PF_INET6,
596 	.create = inet6_create,
597 	.owner	= THIS_MODULE,
598 };
599 
600 int inet6_register_protosw(struct inet_protosw *p)
601 {
602 	struct list_head *lh;
603 	struct inet_protosw *answer;
604 	struct list_head *last_perm;
605 	int protocol = p->protocol;
606 	int ret;
607 
608 	spin_lock_bh(&inetsw6_lock);
609 
610 	ret = -EINVAL;
611 	if (p->type >= SOCK_MAX)
612 		goto out_illegal;
613 
614 	/* If we are trying to override a permanent protocol, bail. */
615 	answer = NULL;
616 	ret = -EPERM;
617 	last_perm = &inetsw6[p->type];
618 	list_for_each(lh, &inetsw6[p->type]) {
619 		answer = list_entry(lh, struct inet_protosw, list);
620 
621 		/* Check only the non-wild match. */
622 		if (INET_PROTOSW_PERMANENT & answer->flags) {
623 			if (protocol == answer->protocol)
624 				break;
625 			last_perm = lh;
626 		}
627 
628 		answer = NULL;
629 	}
630 	if (answer)
631 		goto out_permanent;
632 
633 	/* Add the new entry after the last permanent entry if any, so that
634 	 * the new entry does not override a permanent entry when matched with
635 	 * a wild-card protocol. But it is allowed to override any existing
636 	 * non-permanent entry.  This means that when we remove this entry, the
637 	 * system automatically returns to the old behavior.
638 	 */
639 	list_add_rcu(&p->list, last_perm);
640 	ret = 0;
641 out:
642 	spin_unlock_bh(&inetsw6_lock);
643 	return ret;
644 
645 out_permanent:
646 	pr_err("Attempt to override permanent protocol %d\n", protocol);
647 	goto out;
648 
649 out_illegal:
650 	pr_err("Ignoring attempt to register invalid socket type %d\n",
651 	       p->type);
652 	goto out;
653 }
654 EXPORT_SYMBOL(inet6_register_protosw);
655 
656 void
657 inet6_unregister_protosw(struct inet_protosw *p)
658 {
659 	if (INET_PROTOSW_PERMANENT & p->flags) {
660 		pr_err("Attempt to unregister permanent protocol %d\n",
661 		       p->protocol);
662 	} else {
663 		spin_lock_bh(&inetsw6_lock);
664 		list_del_rcu(&p->list);
665 		spin_unlock_bh(&inetsw6_lock);
666 
667 		synchronize_net();
668 	}
669 }
670 EXPORT_SYMBOL(inet6_unregister_protosw);
671 
672 int inet6_sk_rebuild_header(struct sock *sk)
673 {
674 	struct ipv6_pinfo *np = inet6_sk(sk);
675 	struct dst_entry *dst;
676 
677 	dst = __sk_dst_check(sk, np->dst_cookie);
678 
679 	if (!dst) {
680 		struct inet_sock *inet = inet_sk(sk);
681 		struct in6_addr *final_p, final;
682 		struct flowi6 fl6;
683 
684 		memset(&fl6, 0, sizeof(fl6));
685 		fl6.flowi6_proto = sk->sk_protocol;
686 		fl6.daddr = sk->sk_v6_daddr;
687 		fl6.saddr = np->saddr;
688 		fl6.flowlabel = np->flow_label;
689 		fl6.flowi6_oif = sk->sk_bound_dev_if;
690 		fl6.flowi6_mark = sk->sk_mark;
691 		fl6.fl6_dport = inet->inet_dport;
692 		fl6.fl6_sport = inet->inet_sport;
693 		fl6.flowi6_uid = sk->sk_uid;
694 		security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
695 
696 		rcu_read_lock();
697 		final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt),
698 					 &final);
699 		rcu_read_unlock();
700 
701 		dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
702 		if (IS_ERR(dst)) {
703 			sk->sk_route_caps = 0;
704 			sk->sk_err_soft = -PTR_ERR(dst);
705 			return PTR_ERR(dst);
706 		}
707 
708 		ip6_dst_store(sk, dst, NULL, NULL);
709 	}
710 
711 	return 0;
712 }
713 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
714 
715 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb,
716 		       const struct inet6_skb_parm *opt)
717 {
718 	const struct ipv6_pinfo *np = inet6_sk(sk);
719 
720 	if (np->rxopt.all) {
721 		if (((opt->flags & IP6SKB_HOPBYHOP) &&
722 		     (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) ||
723 		    (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) &&
724 		     np->rxopt.bits.rxflow) ||
725 		    (opt->srcrt && (np->rxopt.bits.srcrt ||
726 		     np->rxopt.bits.osrcrt)) ||
727 		    ((opt->dst1 || opt->dst0) &&
728 		     (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
729 			return true;
730 	}
731 	return false;
732 }
733 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
734 
735 static struct packet_type ipv6_packet_type __read_mostly = {
736 	.type = cpu_to_be16(ETH_P_IPV6),
737 	.func = ipv6_rcv,
738 };
739 
740 static int __init ipv6_packet_init(void)
741 {
742 	dev_add_pack(&ipv6_packet_type);
743 	return 0;
744 }
745 
746 static void ipv6_packet_cleanup(void)
747 {
748 	dev_remove_pack(&ipv6_packet_type);
749 }
750 
751 static int __net_init ipv6_init_mibs(struct net *net)
752 {
753 	int i;
754 
755 	net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib);
756 	if (!net->mib.udp_stats_in6)
757 		return -ENOMEM;
758 	net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib);
759 	if (!net->mib.udplite_stats_in6)
760 		goto err_udplite_mib;
761 	net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib);
762 	if (!net->mib.ipv6_statistics)
763 		goto err_ip_mib;
764 
765 	for_each_possible_cpu(i) {
766 		struct ipstats_mib *af_inet6_stats;
767 		af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i);
768 		u64_stats_init(&af_inet6_stats->syncp);
769 	}
770 
771 
772 	net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib);
773 	if (!net->mib.icmpv6_statistics)
774 		goto err_icmp_mib;
775 	net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
776 						GFP_KERNEL);
777 	if (!net->mib.icmpv6msg_statistics)
778 		goto err_icmpmsg_mib;
779 	return 0;
780 
781 err_icmpmsg_mib:
782 	free_percpu(net->mib.icmpv6_statistics);
783 err_icmp_mib:
784 	free_percpu(net->mib.ipv6_statistics);
785 err_ip_mib:
786 	free_percpu(net->mib.udplite_stats_in6);
787 err_udplite_mib:
788 	free_percpu(net->mib.udp_stats_in6);
789 	return -ENOMEM;
790 }
791 
792 static void ipv6_cleanup_mibs(struct net *net)
793 {
794 	free_percpu(net->mib.udp_stats_in6);
795 	free_percpu(net->mib.udplite_stats_in6);
796 	free_percpu(net->mib.ipv6_statistics);
797 	free_percpu(net->mib.icmpv6_statistics);
798 	kfree(net->mib.icmpv6msg_statistics);
799 }
800 
801 static int __net_init inet6_net_init(struct net *net)
802 {
803 	int err = 0;
804 
805 	net->ipv6.sysctl.bindv6only = 0;
806 	net->ipv6.sysctl.icmpv6_time = 1*HZ;
807 	net->ipv6.sysctl.flowlabel_consistency = 1;
808 	net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS;
809 	net->ipv6.sysctl.idgen_retries = 3;
810 	net->ipv6.sysctl.idgen_delay = 1 * HZ;
811 	net->ipv6.sysctl.flowlabel_state_ranges = 0;
812 	net->ipv6.sysctl.max_dst_opts_cnt = IP6_DEFAULT_MAX_DST_OPTS_CNT;
813 	net->ipv6.sysctl.max_hbh_opts_cnt = IP6_DEFAULT_MAX_HBH_OPTS_CNT;
814 	net->ipv6.sysctl.max_dst_opts_len = IP6_DEFAULT_MAX_DST_OPTS_LEN;
815 	net->ipv6.sysctl.max_hbh_opts_len = IP6_DEFAULT_MAX_HBH_OPTS_LEN;
816 	atomic_set(&net->ipv6.fib6_sernum, 1);
817 
818 	err = ipv6_init_mibs(net);
819 	if (err)
820 		return err;
821 #ifdef CONFIG_PROC_FS
822 	err = udp6_proc_init(net);
823 	if (err)
824 		goto out;
825 	err = tcp6_proc_init(net);
826 	if (err)
827 		goto proc_tcp6_fail;
828 	err = ac6_proc_init(net);
829 	if (err)
830 		goto proc_ac6_fail;
831 #endif
832 	return err;
833 
834 #ifdef CONFIG_PROC_FS
835 proc_ac6_fail:
836 	tcp6_proc_exit(net);
837 proc_tcp6_fail:
838 	udp6_proc_exit(net);
839 out:
840 	ipv6_cleanup_mibs(net);
841 	return err;
842 #endif
843 }
844 
845 static void __net_exit inet6_net_exit(struct net *net)
846 {
847 #ifdef CONFIG_PROC_FS
848 	udp6_proc_exit(net);
849 	tcp6_proc_exit(net);
850 	ac6_proc_exit(net);
851 #endif
852 	ipv6_cleanup_mibs(net);
853 }
854 
855 static struct pernet_operations inet6_net_ops = {
856 	.init = inet6_net_init,
857 	.exit = inet6_net_exit,
858 };
859 
860 static const struct ipv6_stub ipv6_stub_impl = {
861 	.ipv6_sock_mc_join = ipv6_sock_mc_join,
862 	.ipv6_sock_mc_drop = ipv6_sock_mc_drop,
863 	.ipv6_dst_lookup = ip6_dst_lookup,
864 	.udpv6_encap_enable = udpv6_encap_enable,
865 	.ndisc_send_na = ndisc_send_na,
866 	.nd_tbl	= &nd_tbl,
867 };
868 
869 static int __init inet6_init(void)
870 {
871 	struct list_head *r;
872 	int err = 0;
873 
874 	sock_skb_cb_check_size(sizeof(struct inet6_skb_parm));
875 
876 	/* Register the socket-side information for inet6_create.  */
877 	for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
878 		INIT_LIST_HEAD(r);
879 
880 	if (disable_ipv6_mod) {
881 		pr_info("Loaded, but administratively disabled, reboot required to enable\n");
882 		goto out;
883 	}
884 
885 	err = proto_register(&tcpv6_prot, 1);
886 	if (err)
887 		goto out;
888 
889 	err = proto_register(&udpv6_prot, 1);
890 	if (err)
891 		goto out_unregister_tcp_proto;
892 
893 	err = proto_register(&udplitev6_prot, 1);
894 	if (err)
895 		goto out_unregister_udp_proto;
896 
897 	err = proto_register(&rawv6_prot, 1);
898 	if (err)
899 		goto out_unregister_udplite_proto;
900 
901 	err = proto_register(&pingv6_prot, 1);
902 	if (err)
903 		goto out_unregister_ping_proto;
904 
905 	/* We MUST register RAW sockets before we create the ICMP6,
906 	 * IGMP6, or NDISC control sockets.
907 	 */
908 	err = rawv6_init();
909 	if (err)
910 		goto out_unregister_raw_proto;
911 
912 	/* Register the family here so that the init calls below will
913 	 * be able to create sockets. (?? is this dangerous ??)
914 	 */
915 	err = sock_register(&inet6_family_ops);
916 	if (err)
917 		goto out_sock_register_fail;
918 
919 	/*
920 	 *	ipngwg API draft makes clear that the correct semantics
921 	 *	for TCP and UDP is to consider one TCP and UDP instance
922 	 *	in a host available by both INET and INET6 APIs and
923 	 *	able to communicate via both network protocols.
924 	 */
925 
926 	err = register_pernet_subsys(&inet6_net_ops);
927 	if (err)
928 		goto register_pernet_fail;
929 	err = ip6_mr_init();
930 	if (err)
931 		goto ipmr_fail;
932 	err = icmpv6_init();
933 	if (err)
934 		goto icmp_fail;
935 	err = ndisc_init();
936 	if (err)
937 		goto ndisc_fail;
938 	err = igmp6_init();
939 	if (err)
940 		goto igmp_fail;
941 
942 	err = ipv6_netfilter_init();
943 	if (err)
944 		goto netfilter_fail;
945 	/* Create /proc/foo6 entries. */
946 #ifdef CONFIG_PROC_FS
947 	err = -ENOMEM;
948 	if (raw6_proc_init())
949 		goto proc_raw6_fail;
950 	if (udplite6_proc_init())
951 		goto proc_udplite6_fail;
952 	if (ipv6_misc_proc_init())
953 		goto proc_misc6_fail;
954 	if (if6_proc_init())
955 		goto proc_if6_fail;
956 #endif
957 	err = ip6_route_init();
958 	if (err)
959 		goto ip6_route_fail;
960 	err = ndisc_late_init();
961 	if (err)
962 		goto ndisc_late_fail;
963 	err = ip6_flowlabel_init();
964 	if (err)
965 		goto ip6_flowlabel_fail;
966 	err = addrconf_init();
967 	if (err)
968 		goto addrconf_fail;
969 
970 	/* Init v6 extension headers. */
971 	err = ipv6_exthdrs_init();
972 	if (err)
973 		goto ipv6_exthdrs_fail;
974 
975 	err = ipv6_frag_init();
976 	if (err)
977 		goto ipv6_frag_fail;
978 
979 	/* Init v6 transport protocols. */
980 	err = udpv6_init();
981 	if (err)
982 		goto udpv6_fail;
983 
984 	err = udplitev6_init();
985 	if (err)
986 		goto udplitev6_fail;
987 
988 	err = udpv6_offload_init();
989 	if (err)
990 		goto udpv6_offload_fail;
991 
992 	err = tcpv6_init();
993 	if (err)
994 		goto tcpv6_fail;
995 
996 	err = ipv6_packet_init();
997 	if (err)
998 		goto ipv6_packet_fail;
999 
1000 	err = pingv6_init();
1001 	if (err)
1002 		goto pingv6_fail;
1003 
1004 	err = calipso_init();
1005 	if (err)
1006 		goto calipso_fail;
1007 
1008 	err = seg6_init();
1009 	if (err)
1010 		goto seg6_fail;
1011 
1012 	err = igmp6_late_init();
1013 	if (err)
1014 		goto igmp6_late_err;
1015 
1016 #ifdef CONFIG_SYSCTL
1017 	err = ipv6_sysctl_register();
1018 	if (err)
1019 		goto sysctl_fail;
1020 #endif
1021 
1022 	/* ensure that ipv6 stubs are visible only after ipv6 is ready */
1023 	wmb();
1024 	ipv6_stub = &ipv6_stub_impl;
1025 out:
1026 	return err;
1027 
1028 #ifdef CONFIG_SYSCTL
1029 sysctl_fail:
1030 	igmp6_late_cleanup();
1031 #endif
1032 igmp6_late_err:
1033 	seg6_exit();
1034 seg6_fail:
1035 	calipso_exit();
1036 calipso_fail:
1037 	pingv6_exit();
1038 pingv6_fail:
1039 	ipv6_packet_cleanup();
1040 ipv6_packet_fail:
1041 	tcpv6_exit();
1042 tcpv6_fail:
1043 	udpv6_offload_exit();
1044 udpv6_offload_fail:
1045 	udplitev6_exit();
1046 udplitev6_fail:
1047 	udpv6_exit();
1048 udpv6_fail:
1049 	ipv6_frag_exit();
1050 ipv6_frag_fail:
1051 	ipv6_exthdrs_exit();
1052 ipv6_exthdrs_fail:
1053 	addrconf_cleanup();
1054 addrconf_fail:
1055 	ip6_flowlabel_cleanup();
1056 ip6_flowlabel_fail:
1057 	ndisc_late_cleanup();
1058 ndisc_late_fail:
1059 	ip6_route_cleanup();
1060 ip6_route_fail:
1061 #ifdef CONFIG_PROC_FS
1062 	if6_proc_exit();
1063 proc_if6_fail:
1064 	ipv6_misc_proc_exit();
1065 proc_misc6_fail:
1066 	udplite6_proc_exit();
1067 proc_udplite6_fail:
1068 	raw6_proc_exit();
1069 proc_raw6_fail:
1070 #endif
1071 	ipv6_netfilter_fini();
1072 netfilter_fail:
1073 	igmp6_cleanup();
1074 igmp_fail:
1075 	ndisc_cleanup();
1076 ndisc_fail:
1077 	ip6_mr_cleanup();
1078 icmp_fail:
1079 	unregister_pernet_subsys(&inet6_net_ops);
1080 ipmr_fail:
1081 	icmpv6_cleanup();
1082 register_pernet_fail:
1083 	sock_unregister(PF_INET6);
1084 	rtnl_unregister_all(PF_INET6);
1085 out_sock_register_fail:
1086 	rawv6_exit();
1087 out_unregister_ping_proto:
1088 	proto_unregister(&pingv6_prot);
1089 out_unregister_raw_proto:
1090 	proto_unregister(&rawv6_prot);
1091 out_unregister_udplite_proto:
1092 	proto_unregister(&udplitev6_prot);
1093 out_unregister_udp_proto:
1094 	proto_unregister(&udpv6_prot);
1095 out_unregister_tcp_proto:
1096 	proto_unregister(&tcpv6_prot);
1097 	goto out;
1098 }
1099 module_init(inet6_init);
1100 
1101 MODULE_ALIAS_NETPROTO(PF_INET6);
1102