xref: /openbmc/linux/net/ipv6/af_inet6.c (revision 3557b3fd)
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/ipv6_stubs.h>
60 #include <net/ndisc.h>
61 #ifdef CONFIG_IPV6_TUNNEL
62 #include <net/ip6_tunnel.h>
63 #endif
64 #include <net/calipso.h>
65 #include <net/seg6.h>
66 
67 #include <linux/uaccess.h>
68 #include <linux/mroute6.h>
69 
70 #include "ip6_offload.h"
71 
72 MODULE_AUTHOR("Cast of dozens");
73 MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
74 MODULE_LICENSE("GPL");
75 
76 /* The inetsw6 table contains everything that inet6_create needs to
77  * build a new socket.
78  */
79 static struct list_head inetsw6[SOCK_MAX];
80 static DEFINE_SPINLOCK(inetsw6_lock);
81 
82 struct ipv6_params ipv6_defaults = {
83 	.disable_ipv6 = 0,
84 	.autoconf = 1,
85 };
86 
87 static int disable_ipv6_mod;
88 
89 module_param_named(disable, disable_ipv6_mod, int, 0444);
90 MODULE_PARM_DESC(disable, "Disable IPv6 module such that it is non-functional");
91 
92 module_param_named(disable_ipv6, ipv6_defaults.disable_ipv6, int, 0444);
93 MODULE_PARM_DESC(disable_ipv6, "Disable IPv6 on all interfaces");
94 
95 module_param_named(autoconf, ipv6_defaults.autoconf, int, 0444);
96 MODULE_PARM_DESC(autoconf, "Enable IPv6 address autoconfiguration on all interfaces");
97 
98 bool ipv6_mod_enabled(void)
99 {
100 	return disable_ipv6_mod == 0;
101 }
102 EXPORT_SYMBOL_GPL(ipv6_mod_enabled);
103 
104 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk)
105 {
106 	const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo);
107 
108 	return (struct ipv6_pinfo *)(((u8 *)sk) + offset);
109 }
110 
111 static int inet6_create(struct net *net, struct socket *sock, int protocol,
112 			int kern)
113 {
114 	struct inet_sock *inet;
115 	struct ipv6_pinfo *np;
116 	struct sock *sk;
117 	struct inet_protosw *answer;
118 	struct proto *answer_prot;
119 	unsigned char answer_flags;
120 	int try_loading_module = 0;
121 	int err;
122 
123 	if (protocol < 0 || protocol >= IPPROTO_MAX)
124 		return -EINVAL;
125 
126 	/* Look for the requested type/protocol pair. */
127 lookup_protocol:
128 	err = -ESOCKTNOSUPPORT;
129 	rcu_read_lock();
130 	list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) {
131 
132 		err = 0;
133 		/* Check the non-wild match. */
134 		if (protocol == answer->protocol) {
135 			if (protocol != IPPROTO_IP)
136 				break;
137 		} else {
138 			/* Check for the two wild cases. */
139 			if (IPPROTO_IP == protocol) {
140 				protocol = answer->protocol;
141 				break;
142 			}
143 			if (IPPROTO_IP == answer->protocol)
144 				break;
145 		}
146 		err = -EPROTONOSUPPORT;
147 	}
148 
149 	if (err) {
150 		if (try_loading_module < 2) {
151 			rcu_read_unlock();
152 			/*
153 			 * Be more specific, e.g. net-pf-10-proto-132-type-1
154 			 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM)
155 			 */
156 			if (++try_loading_module == 1)
157 				request_module("net-pf-%d-proto-%d-type-%d",
158 						PF_INET6, protocol, sock->type);
159 			/*
160 			 * Fall back to generic, e.g. net-pf-10-proto-132
161 			 * (net-pf-PF_INET6-proto-IPPROTO_SCTP)
162 			 */
163 			else
164 				request_module("net-pf-%d-proto-%d",
165 						PF_INET6, protocol);
166 			goto lookup_protocol;
167 		} else
168 			goto out_rcu_unlock;
169 	}
170 
171 	err = -EPERM;
172 	if (sock->type == SOCK_RAW && !kern &&
173 	    !ns_capable(net->user_ns, CAP_NET_RAW))
174 		goto out_rcu_unlock;
175 
176 	sock->ops = answer->ops;
177 	answer_prot = answer->prot;
178 	answer_flags = answer->flags;
179 	rcu_read_unlock();
180 
181 	WARN_ON(!answer_prot->slab);
182 
183 	err = -ENOBUFS;
184 	sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot, kern);
185 	if (!sk)
186 		goto out;
187 
188 	sock_init_data(sock, sk);
189 
190 	err = 0;
191 	if (INET_PROTOSW_REUSE & answer_flags)
192 		sk->sk_reuse = SK_CAN_REUSE;
193 
194 	inet = inet_sk(sk);
195 	inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0;
196 
197 	if (SOCK_RAW == sock->type) {
198 		inet->inet_num = protocol;
199 		if (IPPROTO_RAW == protocol)
200 			inet->hdrincl = 1;
201 	}
202 
203 	sk->sk_destruct		= inet_sock_destruct;
204 	sk->sk_family		= PF_INET6;
205 	sk->sk_protocol		= protocol;
206 
207 	sk->sk_backlog_rcv	= answer->prot->backlog_rcv;
208 
209 	inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk);
210 	np->hop_limit	= -1;
211 	np->mcast_hops	= IPV6_DEFAULT_MCASTHOPS;
212 	np->mc_loop	= 1;
213 	np->mc_all	= 1;
214 	np->pmtudisc	= IPV6_PMTUDISC_WANT;
215 	np->repflow	= net->ipv6.sysctl.flowlabel_reflect;
216 	sk->sk_ipv6only	= net->ipv6.sysctl.bindv6only;
217 
218 	/* Init the ipv4 part of the socket since we can have sockets
219 	 * using v6 API for ipv4.
220 	 */
221 	inet->uc_ttl	= -1;
222 
223 	inet->mc_loop	= 1;
224 	inet->mc_ttl	= 1;
225 	inet->mc_index	= 0;
226 	inet->mc_list	= NULL;
227 	inet->rcv_tos	= 0;
228 
229 	if (net->ipv4.sysctl_ip_no_pmtu_disc)
230 		inet->pmtudisc = IP_PMTUDISC_DONT;
231 	else
232 		inet->pmtudisc = IP_PMTUDISC_WANT;
233 	/*
234 	 * Increment only the relevant sk_prot->socks debug field, this changes
235 	 * the previous behaviour of incrementing both the equivalent to
236 	 * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
237 	 *
238 	 * This allows better debug granularity as we'll know exactly how many
239 	 * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
240 	 * transport protocol socks. -acme
241 	 */
242 	sk_refcnt_debug_inc(sk);
243 
244 	if (inet->inet_num) {
245 		/* It assumes that any protocol which allows
246 		 * the user to assign a number at socket
247 		 * creation time automatically shares.
248 		 */
249 		inet->inet_sport = htons(inet->inet_num);
250 		err = sk->sk_prot->hash(sk);
251 		if (err) {
252 			sk_common_release(sk);
253 			goto out;
254 		}
255 	}
256 	if (sk->sk_prot->init) {
257 		err = sk->sk_prot->init(sk);
258 		if (err) {
259 			sk_common_release(sk);
260 			goto out;
261 		}
262 	}
263 
264 	if (!kern) {
265 		err = BPF_CGROUP_RUN_PROG_INET_SOCK(sk);
266 		if (err) {
267 			sk_common_release(sk);
268 			goto out;
269 		}
270 	}
271 out:
272 	return err;
273 out_rcu_unlock:
274 	rcu_read_unlock();
275 	goto out;
276 }
277 
278 static int __inet6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len,
279 			bool force_bind_address_no_port, bool with_lock)
280 {
281 	struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr;
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 (addr->sin6_family != AF_INET6)
292 		return -EAFNOSUPPORT;
293 
294 	addr_type = ipv6_addr_type(&addr->sin6_addr);
295 	if ((addr_type & IPV6_ADDR_MULTICAST) && sk->sk_type == SOCK_STREAM)
296 		return -EINVAL;
297 
298 	snum = ntohs(addr->sin6_port);
299 	if (snum && snum < inet_prot_sock(net) &&
300 	    !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
301 		return -EACCES;
302 
303 	if (with_lock)
304 		lock_sock(sk);
305 
306 	/* Check these errors (active socket, double bind). */
307 	if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
308 		err = -EINVAL;
309 		goto out;
310 	}
311 
312 	/* Check if the address belongs to the host. */
313 	if (addr_type == IPV6_ADDR_MAPPED) {
314 		struct net_device *dev = NULL;
315 		int chk_addr_ret;
316 
317 		/* Binding to v4-mapped address on a v6-only socket
318 		 * makes no sense
319 		 */
320 		if (sk->sk_ipv6only) {
321 			err = -EINVAL;
322 			goto out;
323 		}
324 
325 		rcu_read_lock();
326 		if (sk->sk_bound_dev_if) {
327 			dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
328 			if (!dev) {
329 				err = -ENODEV;
330 				goto out_unlock;
331 			}
332 		}
333 
334 		/* Reproduce AF_INET checks to make the bindings consistent */
335 		v4addr = addr->sin6_addr.s6_addr32[3];
336 		chk_addr_ret = inet_addr_type_dev_table(net, dev, v4addr);
337 		rcu_read_unlock();
338 
339 		if (!inet_can_nonlocal_bind(net, inet) &&
340 		    v4addr != htonl(INADDR_ANY) &&
341 		    chk_addr_ret != RTN_LOCAL &&
342 		    chk_addr_ret != RTN_MULTICAST &&
343 		    chk_addr_ret != RTN_BROADCAST) {
344 			err = -EADDRNOTAVAIL;
345 			goto out;
346 		}
347 	} else {
348 		if (addr_type != IPV6_ADDR_ANY) {
349 			struct net_device *dev = NULL;
350 
351 			rcu_read_lock();
352 			if (__ipv6_addr_needs_scope_id(addr_type)) {
353 				if (addr_len >= sizeof(struct sockaddr_in6) &&
354 				    addr->sin6_scope_id) {
355 					/* Override any existing binding, if another one
356 					 * is supplied by user.
357 					 */
358 					sk->sk_bound_dev_if = addr->sin6_scope_id;
359 				}
360 
361 				/* Binding to link-local address requires an interface */
362 				if (!sk->sk_bound_dev_if) {
363 					err = -EINVAL;
364 					goto out_unlock;
365 				}
366 			}
367 
368 			if (sk->sk_bound_dev_if) {
369 				dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
370 				if (!dev) {
371 					err = -ENODEV;
372 					goto out_unlock;
373 				}
374 			}
375 
376 			/* ipv4 addr of the socket is invalid.  Only the
377 			 * unspecified and mapped address have a v4 equivalent.
378 			 */
379 			v4addr = LOOPBACK4_IPV6;
380 			if (!(addr_type & IPV6_ADDR_MULTICAST))	{
381 				if (!ipv6_can_nonlocal_bind(net, inet) &&
382 				    !ipv6_chk_addr(net, &addr->sin6_addr,
383 						   dev, 0)) {
384 					err = -EADDRNOTAVAIL;
385 					goto out_unlock;
386 				}
387 			}
388 			rcu_read_unlock();
389 		}
390 	}
391 
392 	inet->inet_rcv_saddr = v4addr;
393 	inet->inet_saddr = v4addr;
394 
395 	sk->sk_v6_rcv_saddr = addr->sin6_addr;
396 
397 	if (!(addr_type & IPV6_ADDR_MULTICAST))
398 		np->saddr = addr->sin6_addr;
399 
400 	saved_ipv6only = sk->sk_ipv6only;
401 	if (addr_type != IPV6_ADDR_ANY && addr_type != IPV6_ADDR_MAPPED)
402 		sk->sk_ipv6only = 1;
403 
404 	/* Make sure we are allowed to bind here. */
405 	if (snum || !(inet->bind_address_no_port ||
406 		      force_bind_address_no_port)) {
407 		if (sk->sk_prot->get_port(sk, snum)) {
408 			sk->sk_ipv6only = saved_ipv6only;
409 			inet_reset_saddr(sk);
410 			err = -EADDRINUSE;
411 			goto out;
412 		}
413 		err = BPF_CGROUP_RUN_PROG_INET6_POST_BIND(sk);
414 		if (err) {
415 			sk->sk_ipv6only = saved_ipv6only;
416 			inet_reset_saddr(sk);
417 			goto out;
418 		}
419 	}
420 
421 	if (addr_type != IPV6_ADDR_ANY)
422 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
423 	if (snum)
424 		sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
425 	inet->inet_sport = htons(inet->inet_num);
426 	inet->inet_dport = 0;
427 	inet->inet_daddr = 0;
428 out:
429 	if (with_lock)
430 		release_sock(sk);
431 	return err;
432 out_unlock:
433 	rcu_read_unlock();
434 	goto out;
435 }
436 
437 /* bind for INET6 API */
438 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
439 {
440 	struct sock *sk = sock->sk;
441 	int err = 0;
442 
443 	/* If the socket has its own bind function then use it. */
444 	if (sk->sk_prot->bind)
445 		return sk->sk_prot->bind(sk, uaddr, addr_len);
446 
447 	if (addr_len < SIN6_LEN_RFC2133)
448 		return -EINVAL;
449 
450 	/* BPF prog is run before any checks are done so that if the prog
451 	 * changes context in a wrong way it will be caught.
452 	 */
453 	err = BPF_CGROUP_RUN_PROG_INET6_BIND(sk, uaddr);
454 	if (err)
455 		return err;
456 
457 	return __inet6_bind(sk, uaddr, addr_len, false, true);
458 }
459 EXPORT_SYMBOL(inet6_bind);
460 
461 int inet6_release(struct socket *sock)
462 {
463 	struct sock *sk = sock->sk;
464 
465 	if (!sk)
466 		return -EINVAL;
467 
468 	/* Free mc lists */
469 	ipv6_sock_mc_close(sk);
470 
471 	/* Free ac lists */
472 	ipv6_sock_ac_close(sk);
473 
474 	return inet_release(sock);
475 }
476 EXPORT_SYMBOL(inet6_release);
477 
478 void inet6_destroy_sock(struct sock *sk)
479 {
480 	struct ipv6_pinfo *np = inet6_sk(sk);
481 	struct sk_buff *skb;
482 	struct ipv6_txoptions *opt;
483 
484 	/* Release rx options */
485 
486 	skb = xchg(&np->pktoptions, NULL);
487 	kfree_skb(skb);
488 
489 	skb = xchg(&np->rxpmtu, NULL);
490 	kfree_skb(skb);
491 
492 	/* Free flowlabels */
493 	fl6_free_socklist(sk);
494 
495 	/* Free tx options */
496 
497 	opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL);
498 	if (opt) {
499 		atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
500 		txopt_put(opt);
501 	}
502 }
503 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
504 
505 /*
506  *	This does both peername and sockname.
507  */
508 
509 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
510 		 int peer)
511 {
512 	struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
513 	struct sock *sk = sock->sk;
514 	struct inet_sock *inet = inet_sk(sk);
515 	struct ipv6_pinfo *np = inet6_sk(sk);
516 
517 	sin->sin6_family = AF_INET6;
518 	sin->sin6_flowinfo = 0;
519 	sin->sin6_scope_id = 0;
520 	if (peer) {
521 		if (!inet->inet_dport)
522 			return -ENOTCONN;
523 		if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
524 		    peer == 1)
525 			return -ENOTCONN;
526 		sin->sin6_port = inet->inet_dport;
527 		sin->sin6_addr = sk->sk_v6_daddr;
528 		if (np->sndflow)
529 			sin->sin6_flowinfo = np->flow_label;
530 	} else {
531 		if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
532 			sin->sin6_addr = np->saddr;
533 		else
534 			sin->sin6_addr = sk->sk_v6_rcv_saddr;
535 
536 		sin->sin6_port = inet->inet_sport;
537 	}
538 	sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
539 						 sk->sk_bound_dev_if);
540 	return sizeof(*sin);
541 }
542 EXPORT_SYMBOL(inet6_getname);
543 
544 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
545 {
546 	struct sock *sk = sock->sk;
547 	struct net *net = sock_net(sk);
548 
549 	switch (cmd) {
550 	case SIOCADDRT:
551 	case SIOCDELRT:
552 
553 		return ipv6_route_ioctl(net, cmd, (void __user *)arg);
554 
555 	case SIOCSIFADDR:
556 		return addrconf_add_ifaddr(net, (void __user *) arg);
557 	case SIOCDIFADDR:
558 		return addrconf_del_ifaddr(net, (void __user *) arg);
559 	case SIOCSIFDSTADDR:
560 		return addrconf_set_dstaddr(net, (void __user *) arg);
561 	default:
562 		if (!sk->sk_prot->ioctl)
563 			return -ENOIOCTLCMD;
564 		return sk->sk_prot->ioctl(sk, cmd, arg);
565 	}
566 	/*NOTREACHED*/
567 	return 0;
568 }
569 EXPORT_SYMBOL(inet6_ioctl);
570 
571 const struct proto_ops inet6_stream_ops = {
572 	.family		   = PF_INET6,
573 	.owner		   = THIS_MODULE,
574 	.release	   = inet6_release,
575 	.bind		   = inet6_bind,
576 	.connect	   = inet_stream_connect,	/* ok		*/
577 	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
578 	.accept		   = inet_accept,		/* ok		*/
579 	.getname	   = inet6_getname,
580 	.poll		   = tcp_poll,			/* ok		*/
581 	.ioctl		   = inet6_ioctl,		/* must change  */
582 	.gettstamp	   = sock_gettstamp,
583 	.listen		   = inet_listen,		/* ok		*/
584 	.shutdown	   = inet_shutdown,		/* ok		*/
585 	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
586 	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
587 	.sendmsg	   = inet_sendmsg,		/* ok		*/
588 	.recvmsg	   = inet_recvmsg,		/* ok		*/
589 #ifdef CONFIG_MMU
590 	.mmap		   = tcp_mmap,
591 #endif
592 	.sendpage	   = inet_sendpage,
593 	.sendmsg_locked    = tcp_sendmsg_locked,
594 	.sendpage_locked   = tcp_sendpage_locked,
595 	.splice_read	   = tcp_splice_read,
596 	.read_sock	   = tcp_read_sock,
597 	.peek_len	   = tcp_peek_len,
598 #ifdef CONFIG_COMPAT
599 	.compat_setsockopt = compat_sock_common_setsockopt,
600 	.compat_getsockopt = compat_sock_common_getsockopt,
601 #endif
602 	.set_rcvlowat	   = tcp_set_rcvlowat,
603 };
604 
605 const struct proto_ops inet6_dgram_ops = {
606 	.family		   = PF_INET6,
607 	.owner		   = THIS_MODULE,
608 	.release	   = inet6_release,
609 	.bind		   = inet6_bind,
610 	.connect	   = inet_dgram_connect,	/* ok		*/
611 	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
612 	.accept		   = sock_no_accept,		/* a do nothing	*/
613 	.getname	   = inet6_getname,
614 	.poll		   = udp_poll,			/* ok		*/
615 	.ioctl		   = inet6_ioctl,		/* must change  */
616 	.gettstamp	   = sock_gettstamp,
617 	.listen		   = sock_no_listen,		/* ok		*/
618 	.shutdown	   = inet_shutdown,		/* ok		*/
619 	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
620 	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
621 	.sendmsg	   = inet_sendmsg,		/* ok		*/
622 	.recvmsg	   = inet_recvmsg,		/* ok		*/
623 	.mmap		   = sock_no_mmap,
624 	.sendpage	   = sock_no_sendpage,
625 	.set_peek_off	   = sk_set_peek_off,
626 #ifdef CONFIG_COMPAT
627 	.compat_setsockopt = compat_sock_common_setsockopt,
628 	.compat_getsockopt = compat_sock_common_getsockopt,
629 #endif
630 };
631 
632 static const struct net_proto_family inet6_family_ops = {
633 	.family = PF_INET6,
634 	.create = inet6_create,
635 	.owner	= THIS_MODULE,
636 };
637 
638 int inet6_register_protosw(struct inet_protosw *p)
639 {
640 	struct list_head *lh;
641 	struct inet_protosw *answer;
642 	struct list_head *last_perm;
643 	int protocol = p->protocol;
644 	int ret;
645 
646 	spin_lock_bh(&inetsw6_lock);
647 
648 	ret = -EINVAL;
649 	if (p->type >= SOCK_MAX)
650 		goto out_illegal;
651 
652 	/* If we are trying to override a permanent protocol, bail. */
653 	answer = NULL;
654 	ret = -EPERM;
655 	last_perm = &inetsw6[p->type];
656 	list_for_each(lh, &inetsw6[p->type]) {
657 		answer = list_entry(lh, struct inet_protosw, list);
658 
659 		/* Check only the non-wild match. */
660 		if (INET_PROTOSW_PERMANENT & answer->flags) {
661 			if (protocol == answer->protocol)
662 				break;
663 			last_perm = lh;
664 		}
665 
666 		answer = NULL;
667 	}
668 	if (answer)
669 		goto out_permanent;
670 
671 	/* Add the new entry after the last permanent entry if any, so that
672 	 * the new entry does not override a permanent entry when matched with
673 	 * a wild-card protocol. But it is allowed to override any existing
674 	 * non-permanent entry.  This means that when we remove this entry, the
675 	 * system automatically returns to the old behavior.
676 	 */
677 	list_add_rcu(&p->list, last_perm);
678 	ret = 0;
679 out:
680 	spin_unlock_bh(&inetsw6_lock);
681 	return ret;
682 
683 out_permanent:
684 	pr_err("Attempt to override permanent protocol %d\n", protocol);
685 	goto out;
686 
687 out_illegal:
688 	pr_err("Ignoring attempt to register invalid socket type %d\n",
689 	       p->type);
690 	goto out;
691 }
692 EXPORT_SYMBOL(inet6_register_protosw);
693 
694 void
695 inet6_unregister_protosw(struct inet_protosw *p)
696 {
697 	if (INET_PROTOSW_PERMANENT & p->flags) {
698 		pr_err("Attempt to unregister permanent protocol %d\n",
699 		       p->protocol);
700 	} else {
701 		spin_lock_bh(&inetsw6_lock);
702 		list_del_rcu(&p->list);
703 		spin_unlock_bh(&inetsw6_lock);
704 
705 		synchronize_net();
706 	}
707 }
708 EXPORT_SYMBOL(inet6_unregister_protosw);
709 
710 int inet6_sk_rebuild_header(struct sock *sk)
711 {
712 	struct ipv6_pinfo *np = inet6_sk(sk);
713 	struct dst_entry *dst;
714 
715 	dst = __sk_dst_check(sk, np->dst_cookie);
716 
717 	if (!dst) {
718 		struct inet_sock *inet = inet_sk(sk);
719 		struct in6_addr *final_p, final;
720 		struct flowi6 fl6;
721 
722 		memset(&fl6, 0, sizeof(fl6));
723 		fl6.flowi6_proto = sk->sk_protocol;
724 		fl6.daddr = sk->sk_v6_daddr;
725 		fl6.saddr = np->saddr;
726 		fl6.flowlabel = np->flow_label;
727 		fl6.flowi6_oif = sk->sk_bound_dev_if;
728 		fl6.flowi6_mark = sk->sk_mark;
729 		fl6.fl6_dport = inet->inet_dport;
730 		fl6.fl6_sport = inet->inet_sport;
731 		fl6.flowi6_uid = sk->sk_uid;
732 		security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
733 
734 		rcu_read_lock();
735 		final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt),
736 					 &final);
737 		rcu_read_unlock();
738 
739 		dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
740 		if (IS_ERR(dst)) {
741 			sk->sk_route_caps = 0;
742 			sk->sk_err_soft = -PTR_ERR(dst);
743 			return PTR_ERR(dst);
744 		}
745 
746 		ip6_dst_store(sk, dst, NULL, NULL);
747 	}
748 
749 	return 0;
750 }
751 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
752 
753 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb,
754 		       const struct inet6_skb_parm *opt)
755 {
756 	const struct ipv6_pinfo *np = inet6_sk(sk);
757 
758 	if (np->rxopt.all) {
759 		if (((opt->flags & IP6SKB_HOPBYHOP) &&
760 		     (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) ||
761 		    (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) &&
762 		     np->rxopt.bits.rxflow) ||
763 		    (opt->srcrt && (np->rxopt.bits.srcrt ||
764 		     np->rxopt.bits.osrcrt)) ||
765 		    ((opt->dst1 || opt->dst0) &&
766 		     (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
767 			return true;
768 	}
769 	return false;
770 }
771 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
772 
773 static struct packet_type ipv6_packet_type __read_mostly = {
774 	.type = cpu_to_be16(ETH_P_IPV6),
775 	.func = ipv6_rcv,
776 	.list_func = ipv6_list_rcv,
777 };
778 
779 static int __init ipv6_packet_init(void)
780 {
781 	dev_add_pack(&ipv6_packet_type);
782 	return 0;
783 }
784 
785 static void ipv6_packet_cleanup(void)
786 {
787 	dev_remove_pack(&ipv6_packet_type);
788 }
789 
790 static int __net_init ipv6_init_mibs(struct net *net)
791 {
792 	int i;
793 
794 	net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib);
795 	if (!net->mib.udp_stats_in6)
796 		return -ENOMEM;
797 	net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib);
798 	if (!net->mib.udplite_stats_in6)
799 		goto err_udplite_mib;
800 	net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib);
801 	if (!net->mib.ipv6_statistics)
802 		goto err_ip_mib;
803 
804 	for_each_possible_cpu(i) {
805 		struct ipstats_mib *af_inet6_stats;
806 		af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i);
807 		u64_stats_init(&af_inet6_stats->syncp);
808 	}
809 
810 
811 	net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib);
812 	if (!net->mib.icmpv6_statistics)
813 		goto err_icmp_mib;
814 	net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
815 						GFP_KERNEL);
816 	if (!net->mib.icmpv6msg_statistics)
817 		goto err_icmpmsg_mib;
818 	return 0;
819 
820 err_icmpmsg_mib:
821 	free_percpu(net->mib.icmpv6_statistics);
822 err_icmp_mib:
823 	free_percpu(net->mib.ipv6_statistics);
824 err_ip_mib:
825 	free_percpu(net->mib.udplite_stats_in6);
826 err_udplite_mib:
827 	free_percpu(net->mib.udp_stats_in6);
828 	return -ENOMEM;
829 }
830 
831 static void ipv6_cleanup_mibs(struct net *net)
832 {
833 	free_percpu(net->mib.udp_stats_in6);
834 	free_percpu(net->mib.udplite_stats_in6);
835 	free_percpu(net->mib.ipv6_statistics);
836 	free_percpu(net->mib.icmpv6_statistics);
837 	kfree(net->mib.icmpv6msg_statistics);
838 }
839 
840 static int __net_init inet6_net_init(struct net *net)
841 {
842 	int err = 0;
843 
844 	net->ipv6.sysctl.bindv6only = 0;
845 	net->ipv6.sysctl.icmpv6_time = 1*HZ;
846 	net->ipv6.sysctl.icmpv6_echo_ignore_all = 0;
847 	net->ipv6.sysctl.icmpv6_echo_ignore_multicast = 0;
848 	net->ipv6.sysctl.icmpv6_echo_ignore_anycast = 0;
849 
850 	/* By default, rate limit error messages.
851 	 * Except for pmtu discovery, it would break it.
852 	 * proc_do_large_bitmap needs pointer to the bitmap.
853 	 */
854 	bitmap_set(net->ipv6.sysctl.icmpv6_ratemask, 0, ICMPV6_ERRMSG_MAX + 1);
855 	bitmap_clear(net->ipv6.sysctl.icmpv6_ratemask, ICMPV6_PKT_TOOBIG, 1);
856 	net->ipv6.sysctl.icmpv6_ratemask_ptr = net->ipv6.sysctl.icmpv6_ratemask;
857 
858 	net->ipv6.sysctl.flowlabel_consistency = 1;
859 	net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS;
860 	net->ipv6.sysctl.idgen_retries = 3;
861 	net->ipv6.sysctl.idgen_delay = 1 * HZ;
862 	net->ipv6.sysctl.flowlabel_state_ranges = 0;
863 	net->ipv6.sysctl.max_dst_opts_cnt = IP6_DEFAULT_MAX_DST_OPTS_CNT;
864 	net->ipv6.sysctl.max_hbh_opts_cnt = IP6_DEFAULT_MAX_HBH_OPTS_CNT;
865 	net->ipv6.sysctl.max_dst_opts_len = IP6_DEFAULT_MAX_DST_OPTS_LEN;
866 	net->ipv6.sysctl.max_hbh_opts_len = IP6_DEFAULT_MAX_HBH_OPTS_LEN;
867 	atomic_set(&net->ipv6.fib6_sernum, 1);
868 
869 	err = ipv6_init_mibs(net);
870 	if (err)
871 		return err;
872 #ifdef CONFIG_PROC_FS
873 	err = udp6_proc_init(net);
874 	if (err)
875 		goto out;
876 	err = tcp6_proc_init(net);
877 	if (err)
878 		goto proc_tcp6_fail;
879 	err = ac6_proc_init(net);
880 	if (err)
881 		goto proc_ac6_fail;
882 #endif
883 	return err;
884 
885 #ifdef CONFIG_PROC_FS
886 proc_ac6_fail:
887 	tcp6_proc_exit(net);
888 proc_tcp6_fail:
889 	udp6_proc_exit(net);
890 out:
891 	ipv6_cleanup_mibs(net);
892 	return err;
893 #endif
894 }
895 
896 static void __net_exit inet6_net_exit(struct net *net)
897 {
898 #ifdef CONFIG_PROC_FS
899 	udp6_proc_exit(net);
900 	tcp6_proc_exit(net);
901 	ac6_proc_exit(net);
902 #endif
903 	ipv6_cleanup_mibs(net);
904 }
905 
906 static struct pernet_operations inet6_net_ops = {
907 	.init = inet6_net_init,
908 	.exit = inet6_net_exit,
909 };
910 
911 static int ipv6_route_input(struct sk_buff *skb)
912 {
913 	ip6_route_input(skb);
914 	return skb_dst(skb)->error;
915 }
916 
917 static const struct ipv6_stub ipv6_stub_impl = {
918 	.ipv6_sock_mc_join = ipv6_sock_mc_join,
919 	.ipv6_sock_mc_drop = ipv6_sock_mc_drop,
920 	.ipv6_dst_lookup   = ip6_dst_lookup,
921 	.ipv6_route_input  = ipv6_route_input,
922 	.fib6_get_table	   = fib6_get_table,
923 	.fib6_table_lookup = fib6_table_lookup,
924 	.fib6_lookup       = fib6_lookup,
925 	.fib6_select_path  = fib6_select_path,
926 	.ip6_mtu_from_fib6 = ip6_mtu_from_fib6,
927 	.fib6_nh_init	   = fib6_nh_init,
928 	.fib6_nh_release   = fib6_nh_release,
929 	.udpv6_encap_enable = udpv6_encap_enable,
930 	.ndisc_send_na = ndisc_send_na,
931 	.nd_tbl	= &nd_tbl,
932 };
933 
934 static const struct ipv6_bpf_stub ipv6_bpf_stub_impl = {
935 	.inet6_bind = __inet6_bind,
936 	.udp6_lib_lookup = __udp6_lib_lookup,
937 };
938 
939 static int __init inet6_init(void)
940 {
941 	struct list_head *r;
942 	int err = 0;
943 
944 	sock_skb_cb_check_size(sizeof(struct inet6_skb_parm));
945 
946 	/* Register the socket-side information for inet6_create.  */
947 	for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
948 		INIT_LIST_HEAD(r);
949 
950 	if (disable_ipv6_mod) {
951 		pr_info("Loaded, but administratively disabled, reboot required to enable\n");
952 		goto out;
953 	}
954 
955 	err = proto_register(&tcpv6_prot, 1);
956 	if (err)
957 		goto out;
958 
959 	err = proto_register(&udpv6_prot, 1);
960 	if (err)
961 		goto out_unregister_tcp_proto;
962 
963 	err = proto_register(&udplitev6_prot, 1);
964 	if (err)
965 		goto out_unregister_udp_proto;
966 
967 	err = proto_register(&rawv6_prot, 1);
968 	if (err)
969 		goto out_unregister_udplite_proto;
970 
971 	err = proto_register(&pingv6_prot, 1);
972 	if (err)
973 		goto out_unregister_raw_proto;
974 
975 	/* We MUST register RAW sockets before we create the ICMP6,
976 	 * IGMP6, or NDISC control sockets.
977 	 */
978 	err = rawv6_init();
979 	if (err)
980 		goto out_unregister_ping_proto;
981 
982 	/* Register the family here so that the init calls below will
983 	 * be able to create sockets. (?? is this dangerous ??)
984 	 */
985 	err = sock_register(&inet6_family_ops);
986 	if (err)
987 		goto out_sock_register_fail;
988 
989 	/*
990 	 *	ipngwg API draft makes clear that the correct semantics
991 	 *	for TCP and UDP is to consider one TCP and UDP instance
992 	 *	in a host available by both INET and INET6 APIs and
993 	 *	able to communicate via both network protocols.
994 	 */
995 
996 	err = register_pernet_subsys(&inet6_net_ops);
997 	if (err)
998 		goto register_pernet_fail;
999 	err = ip6_mr_init();
1000 	if (err)
1001 		goto ipmr_fail;
1002 	err = icmpv6_init();
1003 	if (err)
1004 		goto icmp_fail;
1005 	err = ndisc_init();
1006 	if (err)
1007 		goto ndisc_fail;
1008 	err = igmp6_init();
1009 	if (err)
1010 		goto igmp_fail;
1011 
1012 	err = ipv6_netfilter_init();
1013 	if (err)
1014 		goto netfilter_fail;
1015 	/* Create /proc/foo6 entries. */
1016 #ifdef CONFIG_PROC_FS
1017 	err = -ENOMEM;
1018 	if (raw6_proc_init())
1019 		goto proc_raw6_fail;
1020 	if (udplite6_proc_init())
1021 		goto proc_udplite6_fail;
1022 	if (ipv6_misc_proc_init())
1023 		goto proc_misc6_fail;
1024 	if (if6_proc_init())
1025 		goto proc_if6_fail;
1026 #endif
1027 	err = ip6_route_init();
1028 	if (err)
1029 		goto ip6_route_fail;
1030 	err = ndisc_late_init();
1031 	if (err)
1032 		goto ndisc_late_fail;
1033 	err = ip6_flowlabel_init();
1034 	if (err)
1035 		goto ip6_flowlabel_fail;
1036 	err = ipv6_anycast_init();
1037 	if (err)
1038 		goto ipv6_anycast_fail;
1039 	err = addrconf_init();
1040 	if (err)
1041 		goto addrconf_fail;
1042 
1043 	/* Init v6 extension headers. */
1044 	err = ipv6_exthdrs_init();
1045 	if (err)
1046 		goto ipv6_exthdrs_fail;
1047 
1048 	err = ipv6_frag_init();
1049 	if (err)
1050 		goto ipv6_frag_fail;
1051 
1052 	/* Init v6 transport protocols. */
1053 	err = udpv6_init();
1054 	if (err)
1055 		goto udpv6_fail;
1056 
1057 	err = udplitev6_init();
1058 	if (err)
1059 		goto udplitev6_fail;
1060 
1061 	err = udpv6_offload_init();
1062 	if (err)
1063 		goto udpv6_offload_fail;
1064 
1065 	err = tcpv6_init();
1066 	if (err)
1067 		goto tcpv6_fail;
1068 
1069 	err = ipv6_packet_init();
1070 	if (err)
1071 		goto ipv6_packet_fail;
1072 
1073 	err = pingv6_init();
1074 	if (err)
1075 		goto pingv6_fail;
1076 
1077 	err = calipso_init();
1078 	if (err)
1079 		goto calipso_fail;
1080 
1081 	err = seg6_init();
1082 	if (err)
1083 		goto seg6_fail;
1084 
1085 	err = igmp6_late_init();
1086 	if (err)
1087 		goto igmp6_late_err;
1088 
1089 #ifdef CONFIG_SYSCTL
1090 	err = ipv6_sysctl_register();
1091 	if (err)
1092 		goto sysctl_fail;
1093 #endif
1094 
1095 	/* ensure that ipv6 stubs are visible only after ipv6 is ready */
1096 	wmb();
1097 	ipv6_stub = &ipv6_stub_impl;
1098 	ipv6_bpf_stub = &ipv6_bpf_stub_impl;
1099 out:
1100 	return err;
1101 
1102 #ifdef CONFIG_SYSCTL
1103 sysctl_fail:
1104 	igmp6_late_cleanup();
1105 #endif
1106 igmp6_late_err:
1107 	seg6_exit();
1108 seg6_fail:
1109 	calipso_exit();
1110 calipso_fail:
1111 	pingv6_exit();
1112 pingv6_fail:
1113 	ipv6_packet_cleanup();
1114 ipv6_packet_fail:
1115 	tcpv6_exit();
1116 tcpv6_fail:
1117 	udpv6_offload_exit();
1118 udpv6_offload_fail:
1119 	udplitev6_exit();
1120 udplitev6_fail:
1121 	udpv6_exit();
1122 udpv6_fail:
1123 	ipv6_frag_exit();
1124 ipv6_frag_fail:
1125 	ipv6_exthdrs_exit();
1126 ipv6_exthdrs_fail:
1127 	addrconf_cleanup();
1128 addrconf_fail:
1129 	ipv6_anycast_cleanup();
1130 ipv6_anycast_fail:
1131 	ip6_flowlabel_cleanup();
1132 ip6_flowlabel_fail:
1133 	ndisc_late_cleanup();
1134 ndisc_late_fail:
1135 	ip6_route_cleanup();
1136 ip6_route_fail:
1137 #ifdef CONFIG_PROC_FS
1138 	if6_proc_exit();
1139 proc_if6_fail:
1140 	ipv6_misc_proc_exit();
1141 proc_misc6_fail:
1142 	udplite6_proc_exit();
1143 proc_udplite6_fail:
1144 	raw6_proc_exit();
1145 proc_raw6_fail:
1146 #endif
1147 	ipv6_netfilter_fini();
1148 netfilter_fail:
1149 	igmp6_cleanup();
1150 igmp_fail:
1151 	ndisc_cleanup();
1152 ndisc_fail:
1153 	icmpv6_cleanup();
1154 icmp_fail:
1155 	ip6_mr_cleanup();
1156 ipmr_fail:
1157 	unregister_pernet_subsys(&inet6_net_ops);
1158 register_pernet_fail:
1159 	sock_unregister(PF_INET6);
1160 	rtnl_unregister_all(PF_INET6);
1161 out_sock_register_fail:
1162 	rawv6_exit();
1163 out_unregister_ping_proto:
1164 	proto_unregister(&pingv6_prot);
1165 out_unregister_raw_proto:
1166 	proto_unregister(&rawv6_prot);
1167 out_unregister_udplite_proto:
1168 	proto_unregister(&udplitev6_prot);
1169 out_unregister_udp_proto:
1170 	proto_unregister(&udpv6_prot);
1171 out_unregister_tcp_proto:
1172 	proto_unregister(&tcpv6_prot);
1173 	goto out;
1174 }
1175 module_init(inet6_init);
1176 
1177 MODULE_ALIAS_NETPROTO(PF_INET6);
1178