xref: /openbmc/linux/net/ipv6/af_inet6.c (revision 15d90a6a)
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->mc_all	= 1;
213 	np->pmtudisc	= IPV6_PMTUDISC_WANT;
214 	np->repflow	= net->ipv6.sysctl.flowlabel_reflect;
215 	sk->sk_ipv6only	= net->ipv6.sysctl.bindv6only;
216 
217 	/* Init the ipv4 part of the socket since we can have sockets
218 	 * using v6 API for ipv4.
219 	 */
220 	inet->uc_ttl	= -1;
221 
222 	inet->mc_loop	= 1;
223 	inet->mc_ttl	= 1;
224 	inet->mc_index	= 0;
225 	inet->mc_list	= NULL;
226 	inet->rcv_tos	= 0;
227 
228 	if (net->ipv4.sysctl_ip_no_pmtu_disc)
229 		inet->pmtudisc = IP_PMTUDISC_DONT;
230 	else
231 		inet->pmtudisc = IP_PMTUDISC_WANT;
232 	/*
233 	 * Increment only the relevant sk_prot->socks debug field, this changes
234 	 * the previous behaviour of incrementing both the equivalent to
235 	 * answer->prot->socks (inet6_sock_nr) and inet_sock_nr.
236 	 *
237 	 * This allows better debug granularity as we'll know exactly how many
238 	 * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6
239 	 * transport protocol socks. -acme
240 	 */
241 	sk_refcnt_debug_inc(sk);
242 
243 	if (inet->inet_num) {
244 		/* It assumes that any protocol which allows
245 		 * the user to assign a number at socket
246 		 * creation time automatically shares.
247 		 */
248 		inet->inet_sport = htons(inet->inet_num);
249 		err = sk->sk_prot->hash(sk);
250 		if (err) {
251 			sk_common_release(sk);
252 			goto out;
253 		}
254 	}
255 	if (sk->sk_prot->init) {
256 		err = sk->sk_prot->init(sk);
257 		if (err) {
258 			sk_common_release(sk);
259 			goto out;
260 		}
261 	}
262 
263 	if (!kern) {
264 		err = BPF_CGROUP_RUN_PROG_INET_SOCK(sk);
265 		if (err) {
266 			sk_common_release(sk);
267 			goto out;
268 		}
269 	}
270 out:
271 	return err;
272 out_rcu_unlock:
273 	rcu_read_unlock();
274 	goto out;
275 }
276 
277 static int __inet6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len,
278 			bool force_bind_address_no_port, bool with_lock)
279 {
280 	struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr;
281 	struct inet_sock *inet = inet_sk(sk);
282 	struct ipv6_pinfo *np = inet6_sk(sk);
283 	struct net *net = sock_net(sk);
284 	__be32 v4addr = 0;
285 	unsigned short snum;
286 	bool saved_ipv6only;
287 	int addr_type = 0;
288 	int err = 0;
289 
290 	if (addr->sin6_family != AF_INET6)
291 		return -EAFNOSUPPORT;
292 
293 	addr_type = ipv6_addr_type(&addr->sin6_addr);
294 	if ((addr_type & IPV6_ADDR_MULTICAST) && sk->sk_type == SOCK_STREAM)
295 		return -EINVAL;
296 
297 	snum = ntohs(addr->sin6_port);
298 	if (snum && snum < inet_prot_sock(net) &&
299 	    !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
300 		return -EACCES;
301 
302 	if (with_lock)
303 		lock_sock(sk);
304 
305 	/* Check these errors (active socket, double bind). */
306 	if (sk->sk_state != TCP_CLOSE || inet->inet_num) {
307 		err = -EINVAL;
308 		goto out;
309 	}
310 
311 	/* Check if the address belongs to the host. */
312 	if (addr_type == IPV6_ADDR_MAPPED) {
313 		struct net_device *dev = NULL;
314 		int chk_addr_ret;
315 
316 		/* Binding to v4-mapped address on a v6-only socket
317 		 * makes no sense
318 		 */
319 		if (sk->sk_ipv6only) {
320 			err = -EINVAL;
321 			goto out;
322 		}
323 
324 		rcu_read_lock();
325 		if (sk->sk_bound_dev_if) {
326 			dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
327 			if (!dev) {
328 				err = -ENODEV;
329 				goto out_unlock;
330 			}
331 		}
332 
333 		/* Reproduce AF_INET checks to make the bindings consistent */
334 		v4addr = addr->sin6_addr.s6_addr32[3];
335 		chk_addr_ret = inet_addr_type_dev_table(net, dev, v4addr);
336 		rcu_read_unlock();
337 
338 		if (!inet_can_nonlocal_bind(net, inet) &&
339 		    v4addr != htonl(INADDR_ANY) &&
340 		    chk_addr_ret != RTN_LOCAL &&
341 		    chk_addr_ret != RTN_MULTICAST &&
342 		    chk_addr_ret != RTN_BROADCAST) {
343 			err = -EADDRNOTAVAIL;
344 			goto out;
345 		}
346 	} else {
347 		if (addr_type != IPV6_ADDR_ANY) {
348 			struct net_device *dev = NULL;
349 
350 			rcu_read_lock();
351 			if (__ipv6_addr_needs_scope_id(addr_type)) {
352 				if (addr_len >= sizeof(struct sockaddr_in6) &&
353 				    addr->sin6_scope_id) {
354 					/* Override any existing binding, if another one
355 					 * is supplied by user.
356 					 */
357 					sk->sk_bound_dev_if = addr->sin6_scope_id;
358 				}
359 
360 				/* Binding to link-local address requires an interface */
361 				if (!sk->sk_bound_dev_if) {
362 					err = -EINVAL;
363 					goto out_unlock;
364 				}
365 			}
366 
367 			if (sk->sk_bound_dev_if) {
368 				dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
369 				if (!dev) {
370 					err = -ENODEV;
371 					goto out_unlock;
372 				}
373 			}
374 
375 			/* ipv4 addr of the socket is invalid.  Only the
376 			 * unspecified and mapped address have a v4 equivalent.
377 			 */
378 			v4addr = LOOPBACK4_IPV6;
379 			if (!(addr_type & IPV6_ADDR_MULTICAST))	{
380 				if (!ipv6_can_nonlocal_bind(net, inet) &&
381 				    !ipv6_chk_addr(net, &addr->sin6_addr,
382 						   dev, 0)) {
383 					err = -EADDRNOTAVAIL;
384 					goto out_unlock;
385 				}
386 			}
387 			rcu_read_unlock();
388 		}
389 	}
390 
391 	inet->inet_rcv_saddr = v4addr;
392 	inet->inet_saddr = v4addr;
393 
394 	sk->sk_v6_rcv_saddr = addr->sin6_addr;
395 
396 	if (!(addr_type & IPV6_ADDR_MULTICAST))
397 		np->saddr = addr->sin6_addr;
398 
399 	saved_ipv6only = sk->sk_ipv6only;
400 	if (addr_type != IPV6_ADDR_ANY && addr_type != IPV6_ADDR_MAPPED)
401 		sk->sk_ipv6only = 1;
402 
403 	/* Make sure we are allowed to bind here. */
404 	if (snum || !(inet->bind_address_no_port ||
405 		      force_bind_address_no_port)) {
406 		if (sk->sk_prot->get_port(sk, snum)) {
407 			sk->sk_ipv6only = saved_ipv6only;
408 			inet_reset_saddr(sk);
409 			err = -EADDRINUSE;
410 			goto out;
411 		}
412 		err = BPF_CGROUP_RUN_PROG_INET6_POST_BIND(sk);
413 		if (err) {
414 			sk->sk_ipv6only = saved_ipv6only;
415 			inet_reset_saddr(sk);
416 			goto out;
417 		}
418 	}
419 
420 	if (addr_type != IPV6_ADDR_ANY)
421 		sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
422 	if (snum)
423 		sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
424 	inet->inet_sport = htons(inet->inet_num);
425 	inet->inet_dport = 0;
426 	inet->inet_daddr = 0;
427 out:
428 	if (with_lock)
429 		release_sock(sk);
430 	return err;
431 out_unlock:
432 	rcu_read_unlock();
433 	goto out;
434 }
435 
436 /* bind for INET6 API */
437 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
438 {
439 	struct sock *sk = sock->sk;
440 	int err = 0;
441 
442 	/* If the socket has its own bind function then use it. */
443 	if (sk->sk_prot->bind)
444 		return sk->sk_prot->bind(sk, uaddr, addr_len);
445 
446 	if (addr_len < SIN6_LEN_RFC2133)
447 		return -EINVAL;
448 
449 	/* BPF prog is run before any checks are done so that if the prog
450 	 * changes context in a wrong way it will be caught.
451 	 */
452 	err = BPF_CGROUP_RUN_PROG_INET6_BIND(sk, uaddr);
453 	if (err)
454 		return err;
455 
456 	return __inet6_bind(sk, uaddr, addr_len, false, true);
457 }
458 EXPORT_SYMBOL(inet6_bind);
459 
460 int inet6_release(struct socket *sock)
461 {
462 	struct sock *sk = sock->sk;
463 
464 	if (!sk)
465 		return -EINVAL;
466 
467 	/* Free mc lists */
468 	ipv6_sock_mc_close(sk);
469 
470 	/* Free ac lists */
471 	ipv6_sock_ac_close(sk);
472 
473 	return inet_release(sock);
474 }
475 EXPORT_SYMBOL(inet6_release);
476 
477 void inet6_destroy_sock(struct sock *sk)
478 {
479 	struct ipv6_pinfo *np = inet6_sk(sk);
480 	struct sk_buff *skb;
481 	struct ipv6_txoptions *opt;
482 
483 	/* Release rx options */
484 
485 	skb = xchg(&np->pktoptions, NULL);
486 	kfree_skb(skb);
487 
488 	skb = xchg(&np->rxpmtu, NULL);
489 	kfree_skb(skb);
490 
491 	/* Free flowlabels */
492 	fl6_free_socklist(sk);
493 
494 	/* Free tx options */
495 
496 	opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL);
497 	if (opt) {
498 		atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
499 		txopt_put(opt);
500 	}
501 }
502 EXPORT_SYMBOL_GPL(inet6_destroy_sock);
503 
504 /*
505  *	This does both peername and sockname.
506  */
507 
508 int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
509 		 int peer)
510 {
511 	struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr;
512 	struct sock *sk = sock->sk;
513 	struct inet_sock *inet = inet_sk(sk);
514 	struct ipv6_pinfo *np = inet6_sk(sk);
515 
516 	sin->sin6_family = AF_INET6;
517 	sin->sin6_flowinfo = 0;
518 	sin->sin6_scope_id = 0;
519 	if (peer) {
520 		if (!inet->inet_dport)
521 			return -ENOTCONN;
522 		if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) &&
523 		    peer == 1)
524 			return -ENOTCONN;
525 		sin->sin6_port = inet->inet_dport;
526 		sin->sin6_addr = sk->sk_v6_daddr;
527 		if (np->sndflow)
528 			sin->sin6_flowinfo = np->flow_label;
529 	} else {
530 		if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
531 			sin->sin6_addr = np->saddr;
532 		else
533 			sin->sin6_addr = sk->sk_v6_rcv_saddr;
534 
535 		sin->sin6_port = inet->inet_sport;
536 	}
537 	sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
538 						 sk->sk_bound_dev_if);
539 	return sizeof(*sin);
540 }
541 EXPORT_SYMBOL(inet6_getname);
542 
543 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
544 {
545 	struct sock *sk = sock->sk;
546 	struct net *net = sock_net(sk);
547 
548 	switch (cmd) {
549 	case SIOCGSTAMP:
550 		return sock_get_timestamp(sk, (struct timeval __user *)arg);
551 
552 	case SIOCGSTAMPNS:
553 		return sock_get_timestampns(sk, (struct timespec __user *)arg);
554 
555 	case SIOCADDRT:
556 	case SIOCDELRT:
557 
558 		return ipv6_route_ioctl(net, cmd, (void __user *)arg);
559 
560 	case SIOCSIFADDR:
561 		return addrconf_add_ifaddr(net, (void __user *) arg);
562 	case SIOCDIFADDR:
563 		return addrconf_del_ifaddr(net, (void __user *) arg);
564 	case SIOCSIFDSTADDR:
565 		return addrconf_set_dstaddr(net, (void __user *) arg);
566 	default:
567 		if (!sk->sk_prot->ioctl)
568 			return -ENOIOCTLCMD;
569 		return sk->sk_prot->ioctl(sk, cmd, arg);
570 	}
571 	/*NOTREACHED*/
572 	return 0;
573 }
574 EXPORT_SYMBOL(inet6_ioctl);
575 
576 const struct proto_ops inet6_stream_ops = {
577 	.family		   = PF_INET6,
578 	.owner		   = THIS_MODULE,
579 	.release	   = inet6_release,
580 	.bind		   = inet6_bind,
581 	.connect	   = inet_stream_connect,	/* ok		*/
582 	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
583 	.accept		   = inet_accept,		/* ok		*/
584 	.getname	   = inet6_getname,
585 	.poll		   = tcp_poll,			/* ok		*/
586 	.ioctl		   = inet6_ioctl,		/* must change  */
587 	.listen		   = inet_listen,		/* ok		*/
588 	.shutdown	   = inet_shutdown,		/* ok		*/
589 	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
590 	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
591 	.sendmsg	   = inet_sendmsg,		/* ok		*/
592 	.recvmsg	   = inet_recvmsg,		/* ok		*/
593 #ifdef CONFIG_MMU
594 	.mmap		   = tcp_mmap,
595 #endif
596 	.sendpage	   = inet_sendpage,
597 	.sendmsg_locked    = tcp_sendmsg_locked,
598 	.sendpage_locked   = tcp_sendpage_locked,
599 	.splice_read	   = tcp_splice_read,
600 	.read_sock	   = tcp_read_sock,
601 	.peek_len	   = tcp_peek_len,
602 #ifdef CONFIG_COMPAT
603 	.compat_setsockopt = compat_sock_common_setsockopt,
604 	.compat_getsockopt = compat_sock_common_getsockopt,
605 #endif
606 	.set_rcvlowat	   = tcp_set_rcvlowat,
607 };
608 
609 const struct proto_ops inet6_dgram_ops = {
610 	.family		   = PF_INET6,
611 	.owner		   = THIS_MODULE,
612 	.release	   = inet6_release,
613 	.bind		   = inet6_bind,
614 	.connect	   = inet_dgram_connect,	/* ok		*/
615 	.socketpair	   = sock_no_socketpair,	/* a do nothing	*/
616 	.accept		   = sock_no_accept,		/* a do nothing	*/
617 	.getname	   = inet6_getname,
618 	.poll		   = udp_poll,			/* ok		*/
619 	.ioctl		   = inet6_ioctl,		/* must change  */
620 	.listen		   = sock_no_listen,		/* ok		*/
621 	.shutdown	   = inet_shutdown,		/* ok		*/
622 	.setsockopt	   = sock_common_setsockopt,	/* ok		*/
623 	.getsockopt	   = sock_common_getsockopt,	/* ok		*/
624 	.sendmsg	   = inet_sendmsg,		/* ok		*/
625 	.recvmsg	   = inet_recvmsg,		/* ok		*/
626 	.mmap		   = sock_no_mmap,
627 	.sendpage	   = sock_no_sendpage,
628 	.set_peek_off	   = sk_set_peek_off,
629 #ifdef CONFIG_COMPAT
630 	.compat_setsockopt = compat_sock_common_setsockopt,
631 	.compat_getsockopt = compat_sock_common_getsockopt,
632 #endif
633 };
634 
635 static const struct net_proto_family inet6_family_ops = {
636 	.family = PF_INET6,
637 	.create = inet6_create,
638 	.owner	= THIS_MODULE,
639 };
640 
641 int inet6_register_protosw(struct inet_protosw *p)
642 {
643 	struct list_head *lh;
644 	struct inet_protosw *answer;
645 	struct list_head *last_perm;
646 	int protocol = p->protocol;
647 	int ret;
648 
649 	spin_lock_bh(&inetsw6_lock);
650 
651 	ret = -EINVAL;
652 	if (p->type >= SOCK_MAX)
653 		goto out_illegal;
654 
655 	/* If we are trying to override a permanent protocol, bail. */
656 	answer = NULL;
657 	ret = -EPERM;
658 	last_perm = &inetsw6[p->type];
659 	list_for_each(lh, &inetsw6[p->type]) {
660 		answer = list_entry(lh, struct inet_protosw, list);
661 
662 		/* Check only the non-wild match. */
663 		if (INET_PROTOSW_PERMANENT & answer->flags) {
664 			if (protocol == answer->protocol)
665 				break;
666 			last_perm = lh;
667 		}
668 
669 		answer = NULL;
670 	}
671 	if (answer)
672 		goto out_permanent;
673 
674 	/* Add the new entry after the last permanent entry if any, so that
675 	 * the new entry does not override a permanent entry when matched with
676 	 * a wild-card protocol. But it is allowed to override any existing
677 	 * non-permanent entry.  This means that when we remove this entry, the
678 	 * system automatically returns to the old behavior.
679 	 */
680 	list_add_rcu(&p->list, last_perm);
681 	ret = 0;
682 out:
683 	spin_unlock_bh(&inetsw6_lock);
684 	return ret;
685 
686 out_permanent:
687 	pr_err("Attempt to override permanent protocol %d\n", protocol);
688 	goto out;
689 
690 out_illegal:
691 	pr_err("Ignoring attempt to register invalid socket type %d\n",
692 	       p->type);
693 	goto out;
694 }
695 EXPORT_SYMBOL(inet6_register_protosw);
696 
697 void
698 inet6_unregister_protosw(struct inet_protosw *p)
699 {
700 	if (INET_PROTOSW_PERMANENT & p->flags) {
701 		pr_err("Attempt to unregister permanent protocol %d\n",
702 		       p->protocol);
703 	} else {
704 		spin_lock_bh(&inetsw6_lock);
705 		list_del_rcu(&p->list);
706 		spin_unlock_bh(&inetsw6_lock);
707 
708 		synchronize_net();
709 	}
710 }
711 EXPORT_SYMBOL(inet6_unregister_protosw);
712 
713 int inet6_sk_rebuild_header(struct sock *sk)
714 {
715 	struct ipv6_pinfo *np = inet6_sk(sk);
716 	struct dst_entry *dst;
717 
718 	dst = __sk_dst_check(sk, np->dst_cookie);
719 
720 	if (!dst) {
721 		struct inet_sock *inet = inet_sk(sk);
722 		struct in6_addr *final_p, final;
723 		struct flowi6 fl6;
724 
725 		memset(&fl6, 0, sizeof(fl6));
726 		fl6.flowi6_proto = sk->sk_protocol;
727 		fl6.daddr = sk->sk_v6_daddr;
728 		fl6.saddr = np->saddr;
729 		fl6.flowlabel = np->flow_label;
730 		fl6.flowi6_oif = sk->sk_bound_dev_if;
731 		fl6.flowi6_mark = sk->sk_mark;
732 		fl6.fl6_dport = inet->inet_dport;
733 		fl6.fl6_sport = inet->inet_sport;
734 		fl6.flowi6_uid = sk->sk_uid;
735 		security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
736 
737 		rcu_read_lock();
738 		final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt),
739 					 &final);
740 		rcu_read_unlock();
741 
742 		dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
743 		if (IS_ERR(dst)) {
744 			sk->sk_route_caps = 0;
745 			sk->sk_err_soft = -PTR_ERR(dst);
746 			return PTR_ERR(dst);
747 		}
748 
749 		ip6_dst_store(sk, dst, NULL, NULL);
750 	}
751 
752 	return 0;
753 }
754 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header);
755 
756 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb,
757 		       const struct inet6_skb_parm *opt)
758 {
759 	const struct ipv6_pinfo *np = inet6_sk(sk);
760 
761 	if (np->rxopt.all) {
762 		if (((opt->flags & IP6SKB_HOPBYHOP) &&
763 		     (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) ||
764 		    (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) &&
765 		     np->rxopt.bits.rxflow) ||
766 		    (opt->srcrt && (np->rxopt.bits.srcrt ||
767 		     np->rxopt.bits.osrcrt)) ||
768 		    ((opt->dst1 || opt->dst0) &&
769 		     (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts)))
770 			return true;
771 	}
772 	return false;
773 }
774 EXPORT_SYMBOL_GPL(ipv6_opt_accepted);
775 
776 static struct packet_type ipv6_packet_type __read_mostly = {
777 	.type = cpu_to_be16(ETH_P_IPV6),
778 	.func = ipv6_rcv,
779 	.list_func = ipv6_list_rcv,
780 };
781 
782 static int __init ipv6_packet_init(void)
783 {
784 	dev_add_pack(&ipv6_packet_type);
785 	return 0;
786 }
787 
788 static void ipv6_packet_cleanup(void)
789 {
790 	dev_remove_pack(&ipv6_packet_type);
791 }
792 
793 static int __net_init ipv6_init_mibs(struct net *net)
794 {
795 	int i;
796 
797 	net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib);
798 	if (!net->mib.udp_stats_in6)
799 		return -ENOMEM;
800 	net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib);
801 	if (!net->mib.udplite_stats_in6)
802 		goto err_udplite_mib;
803 	net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib);
804 	if (!net->mib.ipv6_statistics)
805 		goto err_ip_mib;
806 
807 	for_each_possible_cpu(i) {
808 		struct ipstats_mib *af_inet6_stats;
809 		af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i);
810 		u64_stats_init(&af_inet6_stats->syncp);
811 	}
812 
813 
814 	net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib);
815 	if (!net->mib.icmpv6_statistics)
816 		goto err_icmp_mib;
817 	net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib),
818 						GFP_KERNEL);
819 	if (!net->mib.icmpv6msg_statistics)
820 		goto err_icmpmsg_mib;
821 	return 0;
822 
823 err_icmpmsg_mib:
824 	free_percpu(net->mib.icmpv6_statistics);
825 err_icmp_mib:
826 	free_percpu(net->mib.ipv6_statistics);
827 err_ip_mib:
828 	free_percpu(net->mib.udplite_stats_in6);
829 err_udplite_mib:
830 	free_percpu(net->mib.udp_stats_in6);
831 	return -ENOMEM;
832 }
833 
834 static void ipv6_cleanup_mibs(struct net *net)
835 {
836 	free_percpu(net->mib.udp_stats_in6);
837 	free_percpu(net->mib.udplite_stats_in6);
838 	free_percpu(net->mib.ipv6_statistics);
839 	free_percpu(net->mib.icmpv6_statistics);
840 	kfree(net->mib.icmpv6msg_statistics);
841 }
842 
843 static int __net_init inet6_net_init(struct net *net)
844 {
845 	int err = 0;
846 
847 	net->ipv6.sysctl.bindv6only = 0;
848 	net->ipv6.sysctl.icmpv6_time = 1*HZ;
849 	net->ipv6.sysctl.icmpv6_echo_ignore_all = 0;
850 	net->ipv6.sysctl.flowlabel_consistency = 1;
851 	net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS;
852 	net->ipv6.sysctl.idgen_retries = 3;
853 	net->ipv6.sysctl.idgen_delay = 1 * HZ;
854 	net->ipv6.sysctl.flowlabel_state_ranges = 0;
855 	net->ipv6.sysctl.max_dst_opts_cnt = IP6_DEFAULT_MAX_DST_OPTS_CNT;
856 	net->ipv6.sysctl.max_hbh_opts_cnt = IP6_DEFAULT_MAX_HBH_OPTS_CNT;
857 	net->ipv6.sysctl.max_dst_opts_len = IP6_DEFAULT_MAX_DST_OPTS_LEN;
858 	net->ipv6.sysctl.max_hbh_opts_len = IP6_DEFAULT_MAX_HBH_OPTS_LEN;
859 	atomic_set(&net->ipv6.fib6_sernum, 1);
860 
861 	err = ipv6_init_mibs(net);
862 	if (err)
863 		return err;
864 #ifdef CONFIG_PROC_FS
865 	err = udp6_proc_init(net);
866 	if (err)
867 		goto out;
868 	err = tcp6_proc_init(net);
869 	if (err)
870 		goto proc_tcp6_fail;
871 	err = ac6_proc_init(net);
872 	if (err)
873 		goto proc_ac6_fail;
874 #endif
875 	return err;
876 
877 #ifdef CONFIG_PROC_FS
878 proc_ac6_fail:
879 	tcp6_proc_exit(net);
880 proc_tcp6_fail:
881 	udp6_proc_exit(net);
882 out:
883 	ipv6_cleanup_mibs(net);
884 	return err;
885 #endif
886 }
887 
888 static void __net_exit inet6_net_exit(struct net *net)
889 {
890 #ifdef CONFIG_PROC_FS
891 	udp6_proc_exit(net);
892 	tcp6_proc_exit(net);
893 	ac6_proc_exit(net);
894 #endif
895 	ipv6_cleanup_mibs(net);
896 }
897 
898 static struct pernet_operations inet6_net_ops = {
899 	.init = inet6_net_init,
900 	.exit = inet6_net_exit,
901 };
902 
903 static const struct ipv6_stub ipv6_stub_impl = {
904 	.ipv6_sock_mc_join = ipv6_sock_mc_join,
905 	.ipv6_sock_mc_drop = ipv6_sock_mc_drop,
906 	.ipv6_dst_lookup   = ip6_dst_lookup,
907 	.fib6_get_table	   = fib6_get_table,
908 	.fib6_table_lookup = fib6_table_lookup,
909 	.fib6_lookup       = fib6_lookup,
910 	.fib6_multipath_select = fib6_multipath_select,
911 	.ip6_mtu_from_fib6 = ip6_mtu_from_fib6,
912 	.udpv6_encap_enable = udpv6_encap_enable,
913 	.ndisc_send_na = ndisc_send_na,
914 	.nd_tbl	= &nd_tbl,
915 };
916 
917 static const struct ipv6_bpf_stub ipv6_bpf_stub_impl = {
918 	.inet6_bind = __inet6_bind,
919 	.udp6_lib_lookup = __udp6_lib_lookup,
920 };
921 
922 static int __init inet6_init(void)
923 {
924 	struct list_head *r;
925 	int err = 0;
926 
927 	sock_skb_cb_check_size(sizeof(struct inet6_skb_parm));
928 
929 	/* Register the socket-side information for inet6_create.  */
930 	for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r)
931 		INIT_LIST_HEAD(r);
932 
933 	if (disable_ipv6_mod) {
934 		pr_info("Loaded, but administratively disabled, reboot required to enable\n");
935 		goto out;
936 	}
937 
938 	err = proto_register(&tcpv6_prot, 1);
939 	if (err)
940 		goto out;
941 
942 	err = proto_register(&udpv6_prot, 1);
943 	if (err)
944 		goto out_unregister_tcp_proto;
945 
946 	err = proto_register(&udplitev6_prot, 1);
947 	if (err)
948 		goto out_unregister_udp_proto;
949 
950 	err = proto_register(&rawv6_prot, 1);
951 	if (err)
952 		goto out_unregister_udplite_proto;
953 
954 	err = proto_register(&pingv6_prot, 1);
955 	if (err)
956 		goto out_unregister_raw_proto;
957 
958 	/* We MUST register RAW sockets before we create the ICMP6,
959 	 * IGMP6, or NDISC control sockets.
960 	 */
961 	err = rawv6_init();
962 	if (err)
963 		goto out_unregister_ping_proto;
964 
965 	/* Register the family here so that the init calls below will
966 	 * be able to create sockets. (?? is this dangerous ??)
967 	 */
968 	err = sock_register(&inet6_family_ops);
969 	if (err)
970 		goto out_sock_register_fail;
971 
972 	/*
973 	 *	ipngwg API draft makes clear that the correct semantics
974 	 *	for TCP and UDP is to consider one TCP and UDP instance
975 	 *	in a host available by both INET and INET6 APIs and
976 	 *	able to communicate via both network protocols.
977 	 */
978 
979 	err = register_pernet_subsys(&inet6_net_ops);
980 	if (err)
981 		goto register_pernet_fail;
982 	err = ip6_mr_init();
983 	if (err)
984 		goto ipmr_fail;
985 	err = icmpv6_init();
986 	if (err)
987 		goto icmp_fail;
988 	err = ndisc_init();
989 	if (err)
990 		goto ndisc_fail;
991 	err = igmp6_init();
992 	if (err)
993 		goto igmp_fail;
994 
995 	err = ipv6_netfilter_init();
996 	if (err)
997 		goto netfilter_fail;
998 	/* Create /proc/foo6 entries. */
999 #ifdef CONFIG_PROC_FS
1000 	err = -ENOMEM;
1001 	if (raw6_proc_init())
1002 		goto proc_raw6_fail;
1003 	if (udplite6_proc_init())
1004 		goto proc_udplite6_fail;
1005 	if (ipv6_misc_proc_init())
1006 		goto proc_misc6_fail;
1007 	if (if6_proc_init())
1008 		goto proc_if6_fail;
1009 #endif
1010 	err = ip6_route_init();
1011 	if (err)
1012 		goto ip6_route_fail;
1013 	err = ndisc_late_init();
1014 	if (err)
1015 		goto ndisc_late_fail;
1016 	err = ip6_flowlabel_init();
1017 	if (err)
1018 		goto ip6_flowlabel_fail;
1019 	err = ipv6_anycast_init();
1020 	if (err)
1021 		goto ipv6_anycast_fail;
1022 	err = addrconf_init();
1023 	if (err)
1024 		goto addrconf_fail;
1025 
1026 	/* Init v6 extension headers. */
1027 	err = ipv6_exthdrs_init();
1028 	if (err)
1029 		goto ipv6_exthdrs_fail;
1030 
1031 	err = ipv6_frag_init();
1032 	if (err)
1033 		goto ipv6_frag_fail;
1034 
1035 	/* Init v6 transport protocols. */
1036 	err = udpv6_init();
1037 	if (err)
1038 		goto udpv6_fail;
1039 
1040 	err = udplitev6_init();
1041 	if (err)
1042 		goto udplitev6_fail;
1043 
1044 	err = udpv6_offload_init();
1045 	if (err)
1046 		goto udpv6_offload_fail;
1047 
1048 	err = tcpv6_init();
1049 	if (err)
1050 		goto tcpv6_fail;
1051 
1052 	err = ipv6_packet_init();
1053 	if (err)
1054 		goto ipv6_packet_fail;
1055 
1056 	err = pingv6_init();
1057 	if (err)
1058 		goto pingv6_fail;
1059 
1060 	err = calipso_init();
1061 	if (err)
1062 		goto calipso_fail;
1063 
1064 	err = seg6_init();
1065 	if (err)
1066 		goto seg6_fail;
1067 
1068 	err = igmp6_late_init();
1069 	if (err)
1070 		goto igmp6_late_err;
1071 
1072 #ifdef CONFIG_SYSCTL
1073 	err = ipv6_sysctl_register();
1074 	if (err)
1075 		goto sysctl_fail;
1076 #endif
1077 
1078 	/* ensure that ipv6 stubs are visible only after ipv6 is ready */
1079 	wmb();
1080 	ipv6_stub = &ipv6_stub_impl;
1081 	ipv6_bpf_stub = &ipv6_bpf_stub_impl;
1082 out:
1083 	return err;
1084 
1085 #ifdef CONFIG_SYSCTL
1086 sysctl_fail:
1087 	igmp6_late_cleanup();
1088 #endif
1089 igmp6_late_err:
1090 	seg6_exit();
1091 seg6_fail:
1092 	calipso_exit();
1093 calipso_fail:
1094 	pingv6_exit();
1095 pingv6_fail:
1096 	ipv6_packet_cleanup();
1097 ipv6_packet_fail:
1098 	tcpv6_exit();
1099 tcpv6_fail:
1100 	udpv6_offload_exit();
1101 udpv6_offload_fail:
1102 	udplitev6_exit();
1103 udplitev6_fail:
1104 	udpv6_exit();
1105 udpv6_fail:
1106 	ipv6_frag_exit();
1107 ipv6_frag_fail:
1108 	ipv6_exthdrs_exit();
1109 ipv6_exthdrs_fail:
1110 	addrconf_cleanup();
1111 addrconf_fail:
1112 	ipv6_anycast_cleanup();
1113 ipv6_anycast_fail:
1114 	ip6_flowlabel_cleanup();
1115 ip6_flowlabel_fail:
1116 	ndisc_late_cleanup();
1117 ndisc_late_fail:
1118 	ip6_route_cleanup();
1119 ip6_route_fail:
1120 #ifdef CONFIG_PROC_FS
1121 	if6_proc_exit();
1122 proc_if6_fail:
1123 	ipv6_misc_proc_exit();
1124 proc_misc6_fail:
1125 	udplite6_proc_exit();
1126 proc_udplite6_fail:
1127 	raw6_proc_exit();
1128 proc_raw6_fail:
1129 #endif
1130 	ipv6_netfilter_fini();
1131 netfilter_fail:
1132 	igmp6_cleanup();
1133 igmp_fail:
1134 	ndisc_cleanup();
1135 ndisc_fail:
1136 	icmpv6_cleanup();
1137 icmp_fail:
1138 	ip6_mr_cleanup();
1139 ipmr_fail:
1140 	unregister_pernet_subsys(&inet6_net_ops);
1141 register_pernet_fail:
1142 	sock_unregister(PF_INET6);
1143 	rtnl_unregister_all(PF_INET6);
1144 out_sock_register_fail:
1145 	rawv6_exit();
1146 out_unregister_ping_proto:
1147 	proto_unregister(&pingv6_prot);
1148 out_unregister_raw_proto:
1149 	proto_unregister(&rawv6_prot);
1150 out_unregister_udplite_proto:
1151 	proto_unregister(&udplitev6_prot);
1152 out_unregister_udp_proto:
1153 	proto_unregister(&udpv6_prot);
1154 out_unregister_tcp_proto:
1155 	proto_unregister(&tcpv6_prot);
1156 	goto out;
1157 }
1158 module_init(inet6_init);
1159 
1160 MODULE_ALIAS_NETPROTO(PF_INET6);
1161