1 /*	$NetBSD: in6_pcb.c,v 1.147 2016/07/15 07:40:09 ozaki-r Exp $	*/
2 /*	$KAME: in6_pcb.c,v 1.84 2001/02/08 18:02:08 itojun Exp $	*/
3 
4 /*
5  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the project nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 /*
34  * Copyright (c) 1982, 1986, 1991, 1993
35  *	The Regents of the University of California.  All rights reserved.
36  *
37  * Redistribution and use in source and binary forms, with or without
38  * modification, are permitted provided that the following conditions
39  * are met:
40  * 1. Redistributions of source code must retain the above copyright
41  *    notice, this list of conditions and the following disclaimer.
42  * 2. Redistributions in binary form must reproduce the above copyright
43  *    notice, this list of conditions and the following disclaimer in the
44  *    documentation and/or other materials provided with the distribution.
45  * 3. Neither the name of the University nor the names of its contributors
46  *    may be used to endorse or promote products derived from this software
47  *    without specific prior written permission.
48  *
49  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
50  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
51  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
52  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
53  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
54  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
55  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
56  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
57  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
58  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
59  * SUCH DAMAGE.
60  *
61  *	@(#)in_pcb.c	8.2 (Berkeley) 1/4/94
62  */
63 
64 #include <sys/cdefs.h>
65 __KERNEL_RCSID(0, "$NetBSD: in6_pcb.c,v 1.147 2016/07/15 07:40:09 ozaki-r Exp $");
66 
67 #ifdef _KERNEL_OPT
68 #include "opt_inet.h"
69 #include "opt_ipsec.h"
70 #endif
71 
72 #include <sys/param.h>
73 #include <sys/systm.h>
74 #include <sys/mbuf.h>
75 #include <sys/protosw.h>
76 #include <sys/socket.h>
77 #include <sys/socketvar.h>
78 #include <sys/ioctl.h>
79 #include <sys/errno.h>
80 #include <sys/time.h>
81 #include <sys/proc.h>
82 #include <sys/kauth.h>
83 #include <sys/domain.h>
84 #include <sys/once.h>
85 
86 #include <net/if.h>
87 #include <net/route.h>
88 
89 #include <netinet/in.h>
90 #include <netinet/in_var.h>
91 #include <netinet/in_systm.h>
92 #include <netinet/ip.h>
93 #include <netinet/in_pcb.h>
94 #include <netinet/ip6.h>
95 #include <netinet/portalgo.h>
96 #include <netinet6/ip6_var.h>
97 #include <netinet6/in6_pcb.h>
98 #include <netinet6/scope6_var.h>
99 #include <netinet6/nd6.h>
100 
101 #include "faith.h"
102 
103 #ifdef IPSEC
104 #include <netipsec/ipsec.h>
105 #include <netipsec/ipsec6.h>
106 #include <netipsec/key.h>
107 #endif /* IPSEC */
108 
109 #include <netinet/tcp_vtw.h>
110 
111 const struct in6_addr zeroin6_addr;
112 
113 #define	IN6PCBHASH_PORT(table, lport) \
114 	&(table)->inpt_porthashtbl[ntohs(lport) & (table)->inpt_porthash]
115 #define IN6PCBHASH_BIND(table, laddr, lport) \
116 	&(table)->inpt_bindhashtbl[ \
117 	    (((laddr)->s6_addr32[0] ^ (laddr)->s6_addr32[1] ^ \
118 	      (laddr)->s6_addr32[2] ^ (laddr)->s6_addr32[3]) + ntohs(lport)) & \
119 	    (table)->inpt_bindhash]
120 #define IN6PCBHASH_CONNECT(table, faddr, fport, laddr, lport) \
121 	&(table)->inpt_bindhashtbl[ \
122 	    ((((faddr)->s6_addr32[0] ^ (faddr)->s6_addr32[1] ^ \
123 	      (faddr)->s6_addr32[2] ^ (faddr)->s6_addr32[3]) + ntohs(fport)) + \
124 	     (((laddr)->s6_addr32[0] ^ (laddr)->s6_addr32[1] ^ \
125 	      (laddr)->s6_addr32[2] ^ (laddr)->s6_addr32[3]) + \
126 	      ntohs(lport))) & (table)->inpt_bindhash]
127 
128 int ip6_anonportmin = IPV6PORT_ANONMIN;
129 int ip6_anonportmax = IPV6PORT_ANONMAX;
130 int ip6_lowportmin  = IPV6PORT_RESERVEDMIN;
131 int ip6_lowportmax  = IPV6PORT_RESERVEDMAX;
132 
133 static struct pool in6pcb_pool;
134 
135 static int
in6pcb_poolinit(void)136 in6pcb_poolinit(void)
137 {
138 
139 	pool_init(&in6pcb_pool, sizeof(struct in6pcb), 0, 0, 0, "in6pcbpl",
140 	    NULL, IPL_SOFTNET);
141 	return 0;
142 }
143 
144 void
in6_pcbinit(struct inpcbtable * table,int bindhashsize,int connecthashsize)145 in6_pcbinit(struct inpcbtable *table, int bindhashsize, int connecthashsize)
146 {
147 	static ONCE_DECL(control);
148 
149 	in_pcbinit(table, bindhashsize, connecthashsize);
150 	table->inpt_lastport = (u_int16_t)ip6_anonportmax;
151 
152 	RUN_ONCE(&control, in6pcb_poolinit);
153 }
154 
155 int
in6_pcballoc(struct socket * so,void * v)156 in6_pcballoc(struct socket *so, void *v)
157 {
158 	struct inpcbtable *table = v;
159 	struct in6pcb *in6p;
160 	int s;
161 
162 	s = splnet();
163 	in6p = pool_get(&in6pcb_pool, PR_NOWAIT);
164 	splx(s);
165 	if (in6p == NULL)
166 		return (ENOBUFS);
167 	memset((void *)in6p, 0, sizeof(*in6p));
168 	in6p->in6p_af = AF_INET6;
169 	in6p->in6p_table = table;
170 	in6p->in6p_socket = so;
171 	in6p->in6p_hops = -1;	/* use kernel default */
172 	in6p->in6p_icmp6filt = NULL;
173 	in6p->in6p_portalgo = PORTALGO_DEFAULT;
174 	in6p->in6p_bindportonsend = false;
175 #if defined(IPSEC)
176 	if (ipsec_enabled) {
177 		int error = ipsec_init_pcbpolicy(so, &in6p->in6p_sp);
178 		if (error != 0) {
179 			s = splnet();
180 			pool_put(&in6pcb_pool, in6p);
181 			splx(s);
182 			return error;
183 		}
184 	}
185 #endif /* IPSEC */
186 	s = splnet();
187 	TAILQ_INSERT_HEAD(&table->inpt_queue, (struct inpcb_hdr*)in6p,
188 	    inph_queue);
189 	LIST_INSERT_HEAD(IN6PCBHASH_PORT(table, in6p->in6p_lport),
190 	    &in6p->in6p_head, inph_lhash);
191 	in6_pcbstate(in6p, IN6P_ATTACHED);
192 	splx(s);
193 	if (ip6_v6only)
194 		in6p->in6p_flags |= IN6P_IPV6_V6ONLY;
195 	so->so_pcb = (void *)in6p;
196 	return (0);
197 }
198 
199 /*
200  * Bind address from sin6 to in6p.
201  */
202 static int
in6_pcbbind_addr(struct in6pcb * in6p,struct sockaddr_in6 * sin6,struct lwp * l)203 in6_pcbbind_addr(struct in6pcb *in6p, struct sockaddr_in6 *sin6, struct lwp *l)
204 {
205 	int error;
206 
207 	/*
208 	 * We should check the family, but old programs
209 	 * incorrectly fail to intialize it.
210 	 */
211 	if (sin6->sin6_family != AF_INET6)
212 		return (EAFNOSUPPORT);
213 
214 #ifndef INET
215 	if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr))
216 		return (EADDRNOTAVAIL);
217 #endif
218 
219 	if ((error = sa6_embedscope(sin6, ip6_use_defzone)) != 0)
220 		return (error);
221 
222 	if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
223 		if ((in6p->in6p_flags & IN6P_IPV6_V6ONLY) != 0)
224 			return (EINVAL);
225 		if (sin6->sin6_addr.s6_addr32[3]) {
226 			struct sockaddr_in sin;
227 
228 			memset(&sin, 0, sizeof(sin));
229 			sin.sin_len = sizeof(sin);
230 			sin.sin_family = AF_INET;
231 			bcopy(&sin6->sin6_addr.s6_addr32[3],
232 			    &sin.sin_addr, sizeof(sin.sin_addr));
233 			if (!IN_MULTICAST(sin.sin_addr.s_addr) &&
234 			    ifa_ifwithaddr((struct sockaddr *)&sin) == NULL)
235 				return EADDRNOTAVAIL;
236 		}
237 	} else if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
238 		// succeed
239 	} else if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
240 		struct ifaddr *ia = NULL;
241 
242 		if ((in6p->in6p_flags & IN6P_FAITH) == 0 &&
243 		    (ia = ifa_ifwithaddr(sin6tosa(sin6))) == NULL)
244 			return (EADDRNOTAVAIL);
245 
246 		/*
247 		 * bind to an anycast address might accidentally
248 		 * cause sending a packet with an anycast source
249 		 * address, so we forbid it.
250 		 *
251 		 * We should allow to bind to a deprecated address,
252 		 * since the application dare to use it.
253 		 * But, can we assume that they are careful enough
254 		 * to check if the address is deprecated or not?
255 		 * Maybe, as a safeguard, we should have a setsockopt
256 		 * flag to control the bind(2) behavior against
257 		 * deprecated addresses (default: forbid bind(2)).
258 		 */
259 		if (ia &&
260 		    ifatoia6(ia)->ia6_flags &
261 		    (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY|IN6_IFF_DETACHED))
262 			return (EADDRNOTAVAIL);
263 	}
264 
265 
266 	in6p->in6p_laddr = sin6->sin6_addr;
267 
268 
269 	return (0);
270 }
271 
272 /*
273  * Bind port from sin6 to in6p.
274  */
275 static int
in6_pcbbind_port(struct in6pcb * in6p,struct sockaddr_in6 * sin6,struct lwp * l)276 in6_pcbbind_port(struct in6pcb *in6p, struct sockaddr_in6 *sin6, struct lwp *l)
277 {
278 	struct inpcbtable *table = in6p->in6p_table;
279 	struct socket *so = in6p->in6p_socket;
280 	int wild = 0, reuseport = (so->so_options & SO_REUSEPORT);
281 	int error;
282 
283 	if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0 &&
284 	   ((so->so_proto->pr_flags & PR_CONNREQUIRED) == 0 ||
285 	    (so->so_options & SO_ACCEPTCONN) == 0))
286 		wild = 1;
287 
288 	if (sin6->sin6_port != 0) {
289 		enum kauth_network_req req;
290 
291 #ifndef IPNOPRIVPORTS
292 		if (ntohs(sin6->sin6_port) < IPV6PORT_RESERVED)
293 			req = KAUTH_REQ_NETWORK_BIND_PRIVPORT;
294 		else
295 #endif /* IPNOPRIVPORTS */
296 			req = KAUTH_REQ_NETWORK_BIND_PORT;
297 
298 		error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_BIND,
299 		    req, so, sin6, NULL);
300 		if (error)
301 			return (EACCES);
302 	}
303 
304 	if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
305 		/*
306 		 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
307 		 * allow compepte duplication of binding if
308 		 * SO_REUSEPORT is set, or if SO_REUSEADDR is set
309 		 * and a multicast address is bound on both
310 		 * new and duplicated sockets.
311 		 */
312 		if (so->so_options & (SO_REUSEADDR | SO_REUSEPORT))
313 			reuseport = SO_REUSEADDR|SO_REUSEPORT;
314 	}
315 
316 	if (sin6->sin6_port != 0) {
317 		if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
318 #ifdef INET
319 			struct inpcb *t;
320 			struct vestigial_inpcb vestige;
321 
322 			t = in_pcblookup_port(table,
323 			    *(struct in_addr *)&sin6->sin6_addr.s6_addr32[3],
324 			    sin6->sin6_port, wild, &vestige);
325 			if (t && (reuseport & t->inp_socket->so_options) == 0)
326 				return (EADDRINUSE);
327 			if (!t
328 			    && vestige.valid
329 			    && !(reuseport && vestige.reuse_port))
330 			    return EADDRINUSE;
331 #else
332 			return (EADDRNOTAVAIL);
333 #endif
334 		}
335 
336 		{
337 			struct in6pcb *t;
338 			struct vestigial_inpcb vestige;
339 
340 			t = in6_pcblookup_port(table, &sin6->sin6_addr,
341 			    sin6->sin6_port, wild, &vestige);
342 			if (t && (reuseport & t->in6p_socket->so_options) == 0)
343 				return (EADDRINUSE);
344 			if (!t
345 			    && vestige.valid
346 			    && !(reuseport && vestige.reuse_port))
347 			    return EADDRINUSE;
348 		}
349 	}
350 
351 	if (sin6->sin6_port == 0) {
352 		int e;
353 		e = in6_pcbsetport(sin6, in6p, l);
354 		if (e != 0)
355 			return (e);
356 	} else {
357 		in6p->in6p_lport = sin6->sin6_port;
358 		in6_pcbstate(in6p, IN6P_BOUND);
359 	}
360 
361 	LIST_REMOVE(&in6p->in6p_head, inph_lhash);
362 	LIST_INSERT_HEAD(IN6PCBHASH_PORT(table, in6p->in6p_lport),
363 	    &in6p->in6p_head, inph_lhash);
364 
365 	return (0);
366 }
367 
368 int
in6_pcbbind(void * v,struct sockaddr_in6 * sin6,struct lwp * l)369 in6_pcbbind(void *v, struct sockaddr_in6 *sin6, struct lwp *l)
370 {
371 	struct in6pcb *in6p = v;
372 	struct sockaddr_in6 lsin6;
373 	int error;
374 
375 	if (in6p->in6p_af != AF_INET6)
376 		return (EINVAL);
377 
378 	/*
379 	 * If we already have a local port or a local address it means we're
380 	 * bounded.
381 	 */
382 	if (in6p->in6p_lport || !(IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) ||
383 	    (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr) &&
384 	      in6p->in6p_laddr.s6_addr32[3] == 0)))
385 		return (EINVAL);
386 
387 	if (NULL != sin6) {
388 		/* We were provided a sockaddr_in6 to use. */
389 		if (sin6->sin6_len != sizeof(*sin6))
390 			return (EINVAL);
391 	} else {
392 		/* We always bind to *something*, even if it's "anything". */
393 		lsin6 = *((const struct sockaddr_in6 *)
394 		    in6p->in6p_socket->so_proto->pr_domain->dom_sa_any);
395 		sin6 = &lsin6;
396 	}
397 
398 	/* Bind address. */
399 	error = in6_pcbbind_addr(in6p, sin6, l);
400 	if (error)
401 		return (error);
402 
403 	/* Bind port. */
404 	error = in6_pcbbind_port(in6p, sin6, l);
405 	if (error) {
406 		/*
407 		 * Reset the address here to "any" so we don't "leak" the
408 		 * in6pcb.
409 		 */
410 		in6p->in6p_laddr = in6addr_any;
411 
412 		return (error);
413 	}
414 
415 
416 #if 0
417 	in6p->in6p_flowinfo = 0;	/* XXX */
418 #endif
419 	return (0);
420 }
421 
422 /*
423  * Connect from a socket to a specified address.
424  * Both address and port must be specified in argument sin6.
425  * If don't have a local address for this socket yet,
426  * then pick one.
427  */
428 int
in6_pcbconnect(void * v,struct sockaddr_in6 * sin6,struct lwp * l)429 in6_pcbconnect(void *v, struct sockaddr_in6 *sin6, struct lwp *l)
430 {
431 	struct in6pcb *in6p = v;
432 	struct in6_addr *in6a = NULL;
433 	struct ifnet *ifp = NULL;	/* outgoing interface */
434 	int error = 0;
435 	int scope_ambiguous = 0;
436 #ifdef INET
437 	struct in6_addr mapped;
438 #endif
439 	struct sockaddr_in6 tmp;
440 	struct vestigial_inpcb vestige;
441 	struct psref psref;
442 	int bound;
443 
444 	(void)&in6a;				/* XXX fool gcc */
445 
446 	if (in6p->in6p_af != AF_INET6)
447 		return (EINVAL);
448 
449 	if (sin6->sin6_len != sizeof(*sin6))
450 		return (EINVAL);
451 	if (sin6->sin6_family != AF_INET6)
452 		return (EAFNOSUPPORT);
453 	if (sin6->sin6_port == 0)
454 		return (EADDRNOTAVAIL);
455 
456 	if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr) &&
457 	    in6p->in6p_socket->so_type == SOCK_STREAM)
458 		return EADDRNOTAVAIL;
459 
460 	if (sin6->sin6_scope_id == 0 && !ip6_use_defzone)
461 		scope_ambiguous = 1;
462 	if ((error = sa6_embedscope(sin6, ip6_use_defzone)) != 0)
463 		return(error);
464 
465 	/* sanity check for mapped address case */
466 	if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
467 		if ((in6p->in6p_flags & IN6P_IPV6_V6ONLY) != 0)
468 			return EINVAL;
469 		if (IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr))
470 			in6p->in6p_laddr.s6_addr16[5] = htons(0xffff);
471 		if (!IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr))
472 			return EINVAL;
473 	} else
474 	{
475 		if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr))
476 			return EINVAL;
477 	}
478 
479 	/* protect *sin6 from overwrites */
480 	tmp = *sin6;
481 	sin6 = &tmp;
482 
483 	bound = curlwp_bind();
484 	/* Source address selection. */
485 	if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr) &&
486 	    in6p->in6p_laddr.s6_addr32[3] == 0) {
487 #ifdef INET
488 		struct sockaddr_in sin, *sinp;
489 
490 		memset(&sin, 0, sizeof(sin));
491 		sin.sin_len = sizeof(sin);
492 		sin.sin_family = AF_INET;
493 		memcpy(&sin.sin_addr, &sin6->sin6_addr.s6_addr32[3],
494 			sizeof(sin.sin_addr));
495 		sinp = in_selectsrc(&sin, &in6p->in6p_route,
496 			in6p->in6p_socket->so_options, NULL, &error);
497 		if (sinp == NULL) {
498 			if (error == 0)
499 				error = EADDRNOTAVAIL;
500 			return (error);
501 		}
502 		memset(&mapped, 0, sizeof(mapped));
503 		mapped.s6_addr16[5] = htons(0xffff);
504 		memcpy(&mapped.s6_addr32[3], &sinp->sin_addr, sizeof(sinp->sin_addr));
505 		in6a = &mapped;
506 #else
507 		return EADDRNOTAVAIL;
508 #endif
509 	} else {
510 		/*
511 		 * XXX: in6_selectsrc might replace the bound local address
512 		 * with the address specified by setsockopt(IPV6_PKTINFO).
513 		 * Is it the intended behavior?
514 		 */
515 		in6a = in6_selectsrc(sin6, in6p->in6p_outputopts,
516 				     in6p->in6p_moptions,
517 				     &in6p->in6p_route,
518 				     &in6p->in6p_laddr, &ifp, &psref, &error);
519 		if (ifp && scope_ambiguous &&
520 		    (error = in6_setscope(&sin6->sin6_addr, ifp, NULL)) != 0) {
521 			if_put(ifp, &psref);
522 			curlwp_bindx(bound);
523 			return(error);
524 		}
525 
526 		if (in6a == NULL) {
527 			if_put(ifp, &psref);
528 			curlwp_bindx(bound);
529 			if (error == 0)
530 				error = EADDRNOTAVAIL;
531 			return (error);
532 		}
533 	}
534 
535 	if (ifp != NULL) {
536 		in6p->in6p_ip6.ip6_hlim = (u_int8_t)in6_selecthlim(in6p, ifp);
537 		if_put(ifp, &psref);
538 	} else
539 		in6p->in6p_ip6.ip6_hlim = (u_int8_t)in6_selecthlim_rt(in6p);
540 	curlwp_bindx(bound);
541 
542 	if (in6_pcblookup_connect(in6p->in6p_table, &sin6->sin6_addr,
543 	    sin6->sin6_port,
544 	    IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) ? in6a : &in6p->in6p_laddr,
545 				  in6p->in6p_lport, 0, &vestige)
546 		|| vestige.valid)
547 		return (EADDRINUSE);
548 	if (IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) ||
549 	    (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr) &&
550 	     in6p->in6p_laddr.s6_addr32[3] == 0))
551 	{
552 		if (in6p->in6p_lport == 0) {
553 			error = in6_pcbbind(in6p, NULL, l);
554 			if (error != 0)
555 				return error;
556 		}
557 		in6p->in6p_laddr = *in6a;
558 	}
559 	in6p->in6p_faddr = sin6->sin6_addr;
560 	in6p->in6p_fport = sin6->sin6_port;
561 
562         /* Late bind, if needed */
563 	if (in6p->in6p_bindportonsend) {
564                struct sockaddr_in6 lsin = *((const struct sockaddr_in6 *)
565 		    in6p->in6p_socket->so_proto->pr_domain->dom_sa_any);
566 		lsin.sin6_addr = in6p->in6p_laddr;
567 		lsin.sin6_port = 0;
568 
569                if ((error = in6_pcbbind_port(in6p, &lsin, l)) != 0)
570                        return error;
571 	}
572 
573 	in6_pcbstate(in6p, IN6P_CONNECTED);
574 	in6p->in6p_flowinfo &= ~IPV6_FLOWLABEL_MASK;
575 	if (ip6_auto_flowlabel)
576 		in6p->in6p_flowinfo |=
577 		    (htonl(ip6_randomflowlabel()) & IPV6_FLOWLABEL_MASK);
578 #if defined(IPSEC)
579 	if (ipsec_enabled && in6p->in6p_socket->so_type == SOCK_STREAM)
580 		ipsec_pcbconn(in6p->in6p_sp);
581 #endif
582 	return (0);
583 }
584 
585 void
in6_pcbdisconnect(struct in6pcb * in6p)586 in6_pcbdisconnect(struct in6pcb *in6p)
587 {
588 	memset((void *)&in6p->in6p_faddr, 0, sizeof(in6p->in6p_faddr));
589 	in6p->in6p_fport = 0;
590 	in6_pcbstate(in6p, IN6P_BOUND);
591 	in6p->in6p_flowinfo &= ~IPV6_FLOWLABEL_MASK;
592 #if defined(IPSEC)
593 	if (ipsec_enabled)
594 		ipsec_pcbdisconn(in6p->in6p_sp);
595 #endif
596 	if (in6p->in6p_socket->so_state & SS_NOFDREF)
597 		in6_pcbdetach(in6p);
598 }
599 
600 void
in6_pcbdetach(struct in6pcb * in6p)601 in6_pcbdetach(struct in6pcb *in6p)
602 {
603 	struct socket *so = in6p->in6p_socket;
604 	int s;
605 
606 	if (in6p->in6p_af != AF_INET6)
607 		return;
608 
609 #if defined(IPSEC)
610 	if (ipsec_enabled)
611 		ipsec6_delete_pcbpolicy(in6p);
612 #endif
613 	so->so_pcb = NULL;
614 
615 	s = splnet();
616 	in6_pcbstate(in6p, IN6P_ATTACHED);
617 	LIST_REMOVE(&in6p->in6p_head, inph_lhash);
618 	TAILQ_REMOVE(&in6p->in6p_table->inpt_queue, &in6p->in6p_head,
619 	    inph_queue);
620 	splx(s);
621 
622 	if (in6p->in6p_options) {
623 		m_freem(in6p->in6p_options);
624 	}
625 	if (in6p->in6p_outputopts != NULL) {
626 		ip6_clearpktopts(in6p->in6p_outputopts, -1);
627 		free(in6p->in6p_outputopts, M_IP6OPT);
628 	}
629 	rtcache_free(&in6p->in6p_route);
630 	ip6_freemoptions(in6p->in6p_moptions);
631 	ip_freemoptions(in6p->in6p_v4moptions);
632 	sofree(so);				/* drops the socket's lock */
633 
634 	pool_put(&in6pcb_pool, in6p);
635 	mutex_enter(softnet_lock);		/* reacquire it */
636 }
637 
638 void
in6_setsockaddr(struct in6pcb * in6p,struct sockaddr_in6 * sin6)639 in6_setsockaddr(struct in6pcb *in6p, struct sockaddr_in6 *sin6)
640 {
641 
642 	if (in6p->in6p_af != AF_INET6)
643 		return;
644 
645 	sockaddr_in6_init(sin6, &in6p->in6p_laddr, in6p->in6p_lport, 0, 0);
646 	(void)sa6_recoverscope(sin6); /* XXX: should catch errors */
647 }
648 
649 void
in6_setpeeraddr(struct in6pcb * in6p,struct sockaddr_in6 * sin6)650 in6_setpeeraddr(struct in6pcb *in6p, struct sockaddr_in6 *sin6)
651 {
652 
653 	if (in6p->in6p_af != AF_INET6)
654 		return;
655 
656 	sockaddr_in6_init(sin6, &in6p->in6p_faddr, in6p->in6p_fport, 0, 0);
657 	(void)sa6_recoverscope(sin6); /* XXX: should catch errors */
658 }
659 
660 /*
661  * Pass some notification to all connections of a protocol
662  * associated with address dst.  The local address and/or port numbers
663  * may be specified to limit the search.  The "usual action" will be
664  * taken, depending on the ctlinput cmd.  The caller must filter any
665  * cmds that are uninteresting (e.g., no error in the map).
666  * Call the protocol specific routine (if any) to report
667  * any errors for each matching socket.
668  *
669  * Must be called at splsoftnet.
670  *
671  * Note: src (4th arg) carries the flowlabel value on the original IPv6
672  * header, in sin6_flowinfo member.
673  */
674 int
in6_pcbnotify(struct inpcbtable * table,const struct sockaddr * dst,u_int fport_arg,const struct sockaddr * src,u_int lport_arg,int cmd,void * cmdarg,void (* notify)(struct in6pcb *,int))675 in6_pcbnotify(struct inpcbtable *table, const struct sockaddr *dst,
676     u_int fport_arg, const struct sockaddr *src, u_int lport_arg, int cmd,
677     void *cmdarg, void (*notify)(struct in6pcb *, int))
678 {
679 	struct rtentry *rt;
680 	struct inpcb_hdr *inph, *ninph;
681 	struct sockaddr_in6 sa6_src;
682 	const struct sockaddr_in6 *sa6_dst;
683 	u_int16_t fport = fport_arg, lport = lport_arg;
684 	int errno;
685 	int nmatch = 0;
686 	u_int32_t flowinfo;
687 
688 	if ((unsigned)cmd >= PRC_NCMDS || dst->sa_family != AF_INET6)
689 		return 0;
690 
691 	sa6_dst = (const struct sockaddr_in6 *)dst;
692 	if (IN6_IS_ADDR_UNSPECIFIED(&sa6_dst->sin6_addr))
693 		return 0;
694 
695 	/*
696 	 * note that src can be NULL when we get notify by local fragmentation.
697 	 */
698 	sa6_src = (src == NULL) ? sa6_any : *(const struct sockaddr_in6 *)src;
699 	flowinfo = sa6_src.sin6_flowinfo;
700 
701 	/*
702 	 * Redirects go to all references to the destination,
703 	 * and use in6_rtchange to invalidate the route cache.
704 	 * Dead host indications: also use in6_rtchange to invalidate
705 	 * the cache, and deliver the error to all the sockets.
706 	 * Otherwise, if we have knowledge of the local port and address,
707 	 * deliver only to that socket.
708 	 */
709 	if (PRC_IS_REDIRECT(cmd) || cmd == PRC_HOSTDEAD) {
710 		fport = 0;
711 		lport = 0;
712 		memset((void *)&sa6_src.sin6_addr, 0, sizeof(sa6_src.sin6_addr));
713 
714 		if (cmd != PRC_HOSTDEAD)
715 			notify = in6_rtchange;
716 	}
717 
718 	errno = inet6ctlerrmap[cmd];
719 	TAILQ_FOREACH_SAFE(inph, &table->inpt_queue, inph_queue, ninph) {
720 		struct in6pcb *in6p = (struct in6pcb *)inph;
721 		if (in6p->in6p_af != AF_INET6)
722 			continue;
723 
724 		/*
725 		 * Under the following condition, notify of redirects
726 		 * to the pcb, without making address matches against inpcb.
727 		 * - redirect notification is arrived.
728 		 * - the inpcb is unconnected.
729 		 * - the inpcb is caching !RTF_HOST routing entry.
730 		 * - the ICMPv6 notification is from the gateway cached in the
731 		 *   inpcb.  i.e. ICMPv6 notification is from nexthop gateway
732 		 *   the inpcb used very recently.
733 		 *
734 		 * This is to improve interaction between netbsd/openbsd
735 		 * redirect handling code, and inpcb route cache code.
736 		 * without the clause, !RTF_HOST routing entry (which carries
737 		 * gateway used by inpcb right before the ICMPv6 redirect)
738 		 * will be cached forever in unconnected inpcb.
739 		 *
740 		 * There still is a question regarding to what is TRT:
741 		 * - On bsdi/freebsd, RTF_HOST (cloned) routing entry will be
742 		 *   generated on packet output.  inpcb will always cache
743 		 *   RTF_HOST routing entry so there's no need for the clause
744 		 *   (ICMPv6 redirect will update RTF_HOST routing entry,
745 		 *   and inpcb is caching it already).
746 		 *   However, bsdi/freebsd are vulnerable to local DoS attacks
747 		 *   due to the cloned routing entries.
748 		 * - Specwise, "destination cache" is mentioned in RFC2461.
749 		 *   Jinmei says that it implies bsdi/freebsd behavior, itojun
750 		 *   is not really convinced.
751 		 * - Having hiwat/lowat on # of cloned host route (redirect/
752 		 *   pmtud) may be a good idea.  netbsd/openbsd has it.  see
753 		 *   icmp6_mtudisc_update().
754 		 */
755 		if ((PRC_IS_REDIRECT(cmd) || cmd == PRC_HOSTDEAD) &&
756 		    IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
757 		    (rt = rtcache_validate(&in6p->in6p_route)) != NULL &&
758 		    !(rt->rt_flags & RTF_HOST)) {
759 			const struct sockaddr_in6 *dst6;
760 
761 			dst6 = (const struct sockaddr_in6 *)
762 			    rtcache_getdst(&in6p->in6p_route);
763 			if (dst6 == NULL)
764 				;
765 			else if (IN6_ARE_ADDR_EQUAL(&dst6->sin6_addr,
766 			    &sa6_dst->sin6_addr))
767 				goto do_notify;
768 		}
769 
770 		/*
771 		 * If the error designates a new path MTU for a destination
772 		 * and the application (associated with this socket) wanted to
773 		 * know the value, notify. Note that we notify for all
774 		 * disconnected sockets if the corresponding application
775 		 * wanted. This is because some UDP applications keep sending
776 		 * sockets disconnected.
777 		 * XXX: should we avoid to notify the value to TCP sockets?
778 		 */
779 		if (cmd == PRC_MSGSIZE && (in6p->in6p_flags & IN6P_MTU) != 0 &&
780 		    (IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) ||
781 		     IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &sa6_dst->sin6_addr))) {
782 			ip6_notify_pmtu(in6p, (const struct sockaddr_in6 *)dst,
783 					(u_int32_t *)cmdarg);
784 		}
785 
786 		/*
787 		 * Detect if we should notify the error. If no source and
788 		 * destination ports are specified, but non-zero flowinfo and
789 		 * local address match, notify the error. This is the case
790 		 * when the error is delivered with an encrypted buffer
791 		 * by ESP. Otherwise, just compare addresses and ports
792 		 * as usual.
793 		 */
794 		if (lport == 0 && fport == 0 && flowinfo &&
795 		    in6p->in6p_socket != NULL &&
796 		    flowinfo == (in6p->in6p_flowinfo & IPV6_FLOWLABEL_MASK) &&
797 		    IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &sa6_src.sin6_addr))
798 			goto do_notify;
799 		else if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr,
800 					     &sa6_dst->sin6_addr) ||
801 		    in6p->in6p_socket == NULL ||
802 		    (lport && in6p->in6p_lport != lport) ||
803 		    (!IN6_IS_ADDR_UNSPECIFIED(&sa6_src.sin6_addr) &&
804 		     !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr,
805 					 &sa6_src.sin6_addr)) ||
806 		    (fport && in6p->in6p_fport != fport))
807 			continue;
808 
809 	  do_notify:
810 		if (notify)
811 			(*notify)(in6p, errno);
812 		nmatch++;
813 	}
814 	return nmatch;
815 }
816 
817 void
in6_pcbpurgeif0(struct inpcbtable * table,struct ifnet * ifp)818 in6_pcbpurgeif0(struct inpcbtable *table, struct ifnet *ifp)
819 {
820 	struct inpcb_hdr *inph, *ninph;
821 	struct ip6_moptions *im6o;
822 	struct in6_multi_mship *imm, *nimm;
823 
824 	KASSERT(ifp != NULL);
825 
826 	TAILQ_FOREACH_SAFE(inph, &table->inpt_queue, inph_queue, ninph) {
827 		struct in6pcb *in6p = (struct in6pcb *)inph;
828 		if (in6p->in6p_af != AF_INET6)
829 			continue;
830 
831 		im6o = in6p->in6p_moptions;
832 		if (im6o) {
833 			/*
834 			 * Unselect the outgoing interface if it is being
835 			 * detached.
836 			 */
837 			if (im6o->im6o_multicast_if_index == ifp->if_index)
838 				im6o->im6o_multicast_if_index = 0;
839 
840 			/*
841 			 * Drop multicast group membership if we joined
842 			 * through the interface being detached.
843 			 * XXX controversial - is it really legal for kernel
844 			 * to force this?
845 			 */
846 			for (imm = im6o->im6o_memberships.lh_first;
847 			     imm != NULL; imm = nimm) {
848 				nimm = imm->i6mm_chain.le_next;
849 				if (imm->i6mm_maddr->in6m_ifp == ifp) {
850 					LIST_REMOVE(imm, i6mm_chain);
851 					in6_leavegroup(imm);
852 				}
853 			}
854 		}
855 		in_purgeifmcast(in6p->in6p_v4moptions, ifp);
856 	}
857 }
858 
859 void
in6_pcbpurgeif(struct inpcbtable * table,struct ifnet * ifp)860 in6_pcbpurgeif(struct inpcbtable *table, struct ifnet *ifp)
861 {
862 	struct rtentry *rt;
863 	struct inpcb_hdr *inph, *ninph;
864 
865 	TAILQ_FOREACH_SAFE(inph, &table->inpt_queue, inph_queue, ninph) {
866 		struct in6pcb *in6p = (struct in6pcb *)inph;
867 		if (in6p->in6p_af != AF_INET6)
868 			continue;
869 		if ((rt = rtcache_validate(&in6p->in6p_route)) != NULL &&
870 		    rt->rt_ifp == ifp)
871 			in6_rtchange(in6p, 0);
872 	}
873 }
874 
875 /*
876  * Check for alternatives when higher level complains
877  * about service problems.  For now, invalidate cached
878  * routing information.  If the route was created dynamically
879  * (by a redirect), time to try a default gateway again.
880  */
881 void
in6_losing(struct in6pcb * in6p)882 in6_losing(struct in6pcb *in6p)
883 {
884 	struct rtentry *rt;
885 	struct rt_addrinfo info;
886 
887 	if (in6p->in6p_af != AF_INET6)
888 		return;
889 
890 	if ((rt = rtcache_validate(&in6p->in6p_route)) == NULL)
891 		return;
892 
893 	memset(&info, 0, sizeof(info));
894 	info.rti_info[RTAX_DST] = rtcache_getdst(&in6p->in6p_route);
895 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
896 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
897 	rt_missmsg(RTM_LOSING, &info, rt->rt_flags, 0);
898 	if (rt->rt_flags & RTF_DYNAMIC) {
899 		(void)rtrequest(RTM_DELETE, rt_getkey(rt),
900 		    rt->rt_gateway, rt_mask(rt), rt->rt_flags, NULL);
901 	}
902 	/*
903 	 * A new route can be allocated
904 	 * the next time output is attempted.
905 	 */
906 	rtcache_free(&in6p->in6p_route);
907 }
908 
909 /*
910  * After a routing change, flush old routing.  A new route can be
911  * allocated the next time output is attempted.
912  */
913 void
in6_rtchange(struct in6pcb * in6p,int errno)914 in6_rtchange(struct in6pcb *in6p, int errno)
915 {
916 	if (in6p->in6p_af != AF_INET6)
917 		return;
918 
919 	rtcache_free(&in6p->in6p_route);
920 	/*
921 	 * A new route can be allocated the next time
922 	 * output is attempted.
923 	 */
924 }
925 
926 struct in6pcb *
in6_pcblookup_port(struct inpcbtable * table,struct in6_addr * laddr6,u_int lport_arg,int lookup_wildcard,struct vestigial_inpcb * vp)927 in6_pcblookup_port(struct inpcbtable *table, struct in6_addr *laddr6,
928 		   u_int lport_arg, int lookup_wildcard, struct vestigial_inpcb *vp)
929 {
930 	struct inpcbhead *head;
931 	struct inpcb_hdr *inph;
932 	struct in6pcb *in6p, *match = NULL;
933 	int matchwild = 3, wildcard;
934 	u_int16_t lport = lport_arg;
935 
936 	if (vp)
937 		vp->valid = 0;
938 
939 	head = IN6PCBHASH_PORT(table, lport);
940 	LIST_FOREACH(inph, head, inph_lhash) {
941 		in6p = (struct in6pcb *)inph;
942 		if (in6p->in6p_af != AF_INET6)
943 			continue;
944 
945 		if (in6p->in6p_lport != lport)
946 			continue;
947 		wildcard = 0;
948 		if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_faddr)) {
949 			if ((in6p->in6p_flags & IN6P_IPV6_V6ONLY) != 0)
950 				continue;
951 		}
952 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr))
953 			wildcard++;
954 		if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr)) {
955 			if ((in6p->in6p_flags & IN6P_IPV6_V6ONLY) != 0)
956 				continue;
957 			if (!IN6_IS_ADDR_V4MAPPED(laddr6))
958 				continue;
959 
960 			/* duplicate of IPv4 logic */
961 			wildcard = 0;
962 			if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_faddr) &&
963 			    in6p->in6p_faddr.s6_addr32[3])
964 				wildcard++;
965 			if (!in6p->in6p_laddr.s6_addr32[3]) {
966 				if (laddr6->s6_addr32[3])
967 					wildcard++;
968 			} else {
969 				if (!laddr6->s6_addr32[3])
970 					wildcard++;
971 				else {
972 					if (in6p->in6p_laddr.s6_addr32[3] !=
973 					    laddr6->s6_addr32[3])
974 						continue;
975 				}
976 			}
977 		} else if (IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
978 			if (IN6_IS_ADDR_V4MAPPED(laddr6)) {
979 				if ((in6p->in6p_flags & IN6P_IPV6_V6ONLY) != 0)
980 					continue;
981 			}
982 			if (!IN6_IS_ADDR_UNSPECIFIED(laddr6))
983 				wildcard++;
984 		} else {
985 			if (IN6_IS_ADDR_V4MAPPED(laddr6)) {
986 				if ((in6p->in6p_flags & IN6P_IPV6_V6ONLY) != 0)
987 					continue;
988 			}
989 			if (IN6_IS_ADDR_UNSPECIFIED(laddr6))
990 				wildcard++;
991 			else {
992 				if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr,
993 				    laddr6))
994 					continue;
995 			}
996 		}
997 		if (wildcard && !lookup_wildcard)
998 			continue;
999 		if (wildcard < matchwild) {
1000 			match = in6p;
1001 			matchwild = wildcard;
1002 			if (matchwild == 0)
1003 				break;
1004 		}
1005 	}
1006 	if (match && matchwild == 0)
1007 		return match;
1008 
1009 	if (vp && table->vestige && table->vestige->init_ports6) {
1010 		struct vestigial_inpcb better;
1011 		void *state;
1012 
1013 		state = (*table->vestige->init_ports6)(laddr6,
1014 						       lport_arg,
1015 						       lookup_wildcard);
1016 		while (table->vestige
1017 		       && (*table->vestige->next_port6)(state, vp)) {
1018 
1019 			if (vp->lport != lport)
1020 				continue;
1021 			wildcard = 0;
1022 			if (!IN6_IS_ADDR_UNSPECIFIED(&vp->faddr.v6))
1023 				wildcard++;
1024 			if (IN6_IS_ADDR_UNSPECIFIED(&vp->laddr.v6)) {
1025 				if (!IN6_IS_ADDR_UNSPECIFIED(laddr6))
1026 					wildcard++;
1027 			} else {
1028 				if (IN6_IS_ADDR_V4MAPPED(laddr6)) {
1029 					if (vp->v6only)
1030 						continue;
1031 				}
1032 				if (IN6_IS_ADDR_UNSPECIFIED(laddr6))
1033 					wildcard++;
1034 				else {
1035 					if (!IN6_ARE_ADDR_EQUAL(&vp->laddr.v6, laddr6))
1036 						continue;
1037 				}
1038 			}
1039 			if (wildcard && !lookup_wildcard)
1040 				continue;
1041 			if (wildcard < matchwild) {
1042 				better = *vp;
1043 				match  = (void*)&better;
1044 
1045 				matchwild = wildcard;
1046 				if (matchwild == 0)
1047 					break;
1048 			}
1049 		}
1050 
1051 		if (match) {
1052 			if (match != (void*)&better)
1053 				return match;
1054 			else {
1055 				*vp = better;
1056 				return 0;
1057 			}
1058 		}
1059 	}
1060 	return (match);
1061 }
1062 
1063 /*
1064  * WARNING: return value (rtentry) could be IPv4 one if in6pcb is connected to
1065  * IPv4 mapped address.
1066  */
1067 struct rtentry *
in6_pcbrtentry(struct in6pcb * in6p)1068 in6_pcbrtentry(struct in6pcb *in6p)
1069 {
1070 	struct rtentry *rt;
1071 	struct route *ro;
1072 	union {
1073 		const struct sockaddr *sa;
1074 		const struct sockaddr_in6 *sa6;
1075 #ifdef INET
1076 		const struct sockaddr_in *sa4;
1077 #endif
1078 	} cdst;
1079 
1080 	ro = &in6p->in6p_route;
1081 
1082 	if (in6p->in6p_af != AF_INET6)
1083 		return (NULL);
1084 
1085 	cdst.sa = rtcache_getdst(ro);
1086 	if (cdst.sa == NULL)
1087 		;
1088 #ifdef INET
1089 	else if (cdst.sa->sa_family == AF_INET) {
1090 		KASSERT(IN6_IS_ADDR_V4MAPPED(&in6p->in6p_faddr));
1091 		if (cdst.sa4->sin_addr.s_addr != in6p->in6p_faddr.s6_addr32[3])
1092 			rtcache_free(ro);
1093 	}
1094 #endif
1095 	else {
1096 		if (!IN6_ARE_ADDR_EQUAL(&cdst.sa6->sin6_addr,
1097 					&in6p->in6p_faddr))
1098 			rtcache_free(ro);
1099 	}
1100 	if ((rt = rtcache_validate(ro)) == NULL)
1101 		rt = rtcache_update(ro, 1);
1102 #ifdef INET
1103 	if (rt == NULL && IN6_IS_ADDR_V4MAPPED(&in6p->in6p_faddr)) {
1104 		union {
1105 			struct sockaddr		dst;
1106 			struct sockaddr_in	dst4;
1107 		} u;
1108 		struct in_addr addr;
1109 
1110 		addr.s_addr = in6p->in6p_faddr.s6_addr32[3];
1111 
1112 		sockaddr_in_init(&u.dst4, &addr, 0);
1113 		if (rtcache_setdst(ro, &u.dst) != 0)
1114 			return NULL;
1115 
1116 		rt = rtcache_init(ro);
1117 	} else
1118 #endif
1119 	if (rt == NULL && !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) {
1120 		union {
1121 			struct sockaddr		dst;
1122 			struct sockaddr_in6	dst6;
1123 		} u;
1124 
1125 		sockaddr_in6_init(&u.dst6, &in6p->in6p_faddr, 0, 0, 0);
1126 		if (rtcache_setdst(ro, &u.dst) != 0)
1127 			return NULL;
1128 
1129 		rt = rtcache_init(ro);
1130 	}
1131 	return rt;
1132 }
1133 
1134 struct in6pcb *
in6_pcblookup_connect(struct inpcbtable * table,const struct in6_addr * faddr6,u_int fport_arg,const struct in6_addr * laddr6,u_int lport_arg,int faith,struct vestigial_inpcb * vp)1135 in6_pcblookup_connect(struct inpcbtable *table, const struct in6_addr *faddr6,
1136 		      u_int fport_arg, const struct in6_addr *laddr6, u_int lport_arg,
1137 		      int faith,
1138 		      struct vestigial_inpcb *vp)
1139 {
1140 	struct inpcbhead *head;
1141 	struct inpcb_hdr *inph;
1142 	struct in6pcb *in6p;
1143 	u_int16_t fport = fport_arg, lport = lport_arg;
1144 
1145 	if (vp)
1146 		vp->valid = 0;
1147 
1148 	head = IN6PCBHASH_CONNECT(table, faddr6, fport, laddr6, lport);
1149 	LIST_FOREACH(inph, head, inph_hash) {
1150 		in6p = (struct in6pcb *)inph;
1151 		if (in6p->in6p_af != AF_INET6)
1152 			continue;
1153 
1154 		/* find exact match on both source and dest */
1155 		if (in6p->in6p_fport != fport)
1156 			continue;
1157 		if (in6p->in6p_lport != lport)
1158 			continue;
1159 		if (IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr))
1160 			continue;
1161 		if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, faddr6))
1162 			continue;
1163 		if (IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr))
1164 			continue;
1165 		if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, laddr6))
1166 			continue;
1167 		if ((IN6_IS_ADDR_V4MAPPED(laddr6) ||
1168 		     IN6_IS_ADDR_V4MAPPED(faddr6)) &&
1169 		    (in6p->in6p_flags & IN6P_IPV6_V6ONLY))
1170 			continue;
1171 		return in6p;
1172 	}
1173 	if (vp && table->vestige) {
1174 		if ((*table->vestige->lookup6)(faddr6, fport_arg,
1175 					       laddr6, lport_arg, vp))
1176 			return NULL;
1177 	}
1178 
1179 	return NULL;
1180 }
1181 
1182 struct in6pcb *
in6_pcblookup_bind(struct inpcbtable * table,const struct in6_addr * laddr6,u_int lport_arg,int faith)1183 in6_pcblookup_bind(struct inpcbtable *table, const struct in6_addr *laddr6,
1184 	u_int lport_arg, int faith)
1185 {
1186 	struct inpcbhead *head;
1187 	struct inpcb_hdr *inph;
1188 	struct in6pcb *in6p;
1189 	u_int16_t lport = lport_arg;
1190 #ifdef INET
1191 	struct in6_addr zero_mapped;
1192 #endif
1193 
1194 	head = IN6PCBHASH_BIND(table, laddr6, lport);
1195 	LIST_FOREACH(inph, head, inph_hash) {
1196 		in6p = (struct in6pcb *)inph;
1197 		if (in6p->in6p_af != AF_INET6)
1198 			continue;
1199 
1200 		if (faith && (in6p->in6p_flags & IN6P_FAITH) == 0)
1201 			continue;
1202 		if (in6p->in6p_fport != 0)
1203 			continue;
1204 		if (in6p->in6p_lport != lport)
1205 			continue;
1206 		if (IN6_IS_ADDR_V4MAPPED(laddr6) &&
1207 		    (in6p->in6p_flags & IN6P_IPV6_V6ONLY) != 0)
1208 			continue;
1209 		if (IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, laddr6))
1210 			goto out;
1211 	}
1212 #ifdef INET
1213 	if (IN6_IS_ADDR_V4MAPPED(laddr6)) {
1214 		memset(&zero_mapped, 0, sizeof(zero_mapped));
1215 		zero_mapped.s6_addr16[5] = 0xffff;
1216 		head = IN6PCBHASH_BIND(table, &zero_mapped, lport);
1217 		LIST_FOREACH(inph, head, inph_hash) {
1218 			in6p = (struct in6pcb *)inph;
1219 			if (in6p->in6p_af != AF_INET6)
1220 				continue;
1221 
1222 			if (faith && (in6p->in6p_flags & IN6P_FAITH) == 0)
1223 				continue;
1224 			if (in6p->in6p_fport != 0)
1225 				continue;
1226 			if (in6p->in6p_lport != lport)
1227 				continue;
1228 			if ((in6p->in6p_flags & IN6P_IPV6_V6ONLY) != 0)
1229 				continue;
1230 			if (IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &zero_mapped))
1231 				goto out;
1232 		}
1233 	}
1234 #endif
1235 	head = IN6PCBHASH_BIND(table, &zeroin6_addr, lport);
1236 	LIST_FOREACH(inph, head, inph_hash) {
1237 		in6p = (struct in6pcb *)inph;
1238 		if (in6p->in6p_af != AF_INET6)
1239 			continue;
1240 
1241 		if (faith && (in6p->in6p_flags & IN6P_FAITH) == 0)
1242 			continue;
1243 		if (in6p->in6p_fport != 0)
1244 			continue;
1245 		if (in6p->in6p_lport != lport)
1246 			continue;
1247 		if (IN6_IS_ADDR_V4MAPPED(laddr6) &&
1248 		    (in6p->in6p_flags & IN6P_IPV6_V6ONLY) != 0)
1249 			continue;
1250 		if (IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &zeroin6_addr))
1251 			goto out;
1252 	}
1253 	return (NULL);
1254 
1255 out:
1256 	inph = &in6p->in6p_head;
1257 	if (inph != LIST_FIRST(head)) {
1258 		LIST_REMOVE(inph, inph_hash);
1259 		LIST_INSERT_HEAD(head, inph, inph_hash);
1260 	}
1261 	return in6p;
1262 }
1263 
1264 void
in6_pcbstate(struct in6pcb * in6p,int state)1265 in6_pcbstate(struct in6pcb *in6p, int state)
1266 {
1267 
1268 	if (in6p->in6p_af != AF_INET6)
1269 		return;
1270 
1271 	if (in6p->in6p_state > IN6P_ATTACHED)
1272 		LIST_REMOVE(&in6p->in6p_head, inph_hash);
1273 
1274 	switch (state) {
1275 	case IN6P_BOUND:
1276 		LIST_INSERT_HEAD(IN6PCBHASH_BIND(in6p->in6p_table,
1277 		    &in6p->in6p_laddr, in6p->in6p_lport), &in6p->in6p_head,
1278 		    inph_hash);
1279 		break;
1280 	case IN6P_CONNECTED:
1281 		LIST_INSERT_HEAD(IN6PCBHASH_CONNECT(in6p->in6p_table,
1282 		    &in6p->in6p_faddr, in6p->in6p_fport,
1283 		    &in6p->in6p_laddr, in6p->in6p_lport), &in6p->in6p_head,
1284 		    inph_hash);
1285 		break;
1286 	}
1287 
1288 	in6p->in6p_state = state;
1289 }
1290