xref: /openbsd-src/sys/netinet/in_pcb.c (revision ae3cb403620ab940fbaabb3055fac045a63d56b7)
1 /*	$OpenBSD: in_pcb.c,v 1.227 2017/12/04 13:40:34 bluhm Exp $	*/
2 /*	$NetBSD: in_pcb.c,v 1.25 1996/02/13 23:41:53 christos Exp $	*/
3 
4 /*
5  * Copyright (c) 1982, 1986, 1991, 1993
6  *	The Regents of the University of California.  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 University 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 REGENTS 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 REGENTS 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  *	@(#)COPYRIGHT	1.1 (NRL) 17 January 1995
33  *
34  * NRL grants permission for redistribution and use in source and binary
35  * forms, with or without modification, of the software and documentation
36  * created at NRL provided that the following conditions are met:
37  *
38  * 1. Redistributions of source code must retain the above copyright
39  *    notice, this list of conditions and the following disclaimer.
40  * 2. Redistributions in binary form must reproduce the above copyright
41  *    notice, this list of conditions and the following disclaimer in the
42  *    documentation and/or other materials provided with the distribution.
43  * 3. All advertising materials mentioning features or use of this software
44  *    must display the following acknowledgements:
45  *	This product includes software developed by the University of
46  *	California, Berkeley and its contributors.
47  *	This product includes software developed at the Information
48  *	Technology Division, US Naval Research Laboratory.
49  * 4. Neither the name of the NRL nor the names of its contributors
50  *    may be used to endorse or promote products derived from this software
51  *    without specific prior written permission.
52  *
53  * THE SOFTWARE PROVIDED BY NRL IS PROVIDED BY NRL AND CONTRIBUTORS ``AS
54  * IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
55  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A
56  * PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL NRL OR
57  * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
58  * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
59  * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
60  * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
61  * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
62  * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
63  * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
64  *
65  * The views and conclusions contained in the software and documentation
66  * are those of the authors and should not be interpreted as representing
67  * official policies, either expressed or implied, of the US Naval
68  * Research Laboratory (NRL).
69  */
70 
71 #include "pf.h"
72 
73 #include <sys/param.h>
74 #include <sys/systm.h>
75 #include <sys/mbuf.h>
76 #include <sys/protosw.h>
77 #include <sys/socket.h>
78 #include <sys/socketvar.h>
79 #include <sys/proc.h>
80 #include <sys/pledge.h>
81 #include <sys/domain.h>
82 #include <sys/pool.h>
83 
84 #include <net/if.h>
85 #include <net/if_var.h>
86 #include <net/route.h>
87 
88 #include <netinet/in.h>
89 #include <netinet/ip.h>
90 #include <netinet/in_pcb.h>
91 #include <netinet/in_var.h>
92 #include <netinet/ip_var.h>
93 
94 #include <net/pfvar.h>
95 
96 #include <sys/mount.h>
97 #include <nfs/nfsproto.h>
98 
99 #ifdef INET6
100 #include <netinet6/in6_var.h>
101 #include <netinet6/ip6_var.h>
102 #endif /* INET6 */
103 #ifdef IPSEC
104 #include <netinet/ip_esp.h>
105 #endif /* IPSEC */
106 
107 struct	in_addr zeroin_addr;
108 
109 union {
110 	struct in_addr	za_in;
111 	struct in6_addr	za_in6;
112 } zeroin46_addr;
113 
114 /*
115  * These configure the range of local port addresses assigned to
116  * "unspecified" outgoing connections/packets/whatever.
117  */
118 int ipport_firstauto = IPPORT_RESERVED;
119 int ipport_lastauto = IPPORT_USERRESERVED;
120 int ipport_hifirstauto = IPPORT_HIFIRSTAUTO;
121 int ipport_hilastauto = IPPORT_HILASTAUTO;
122 
123 struct baddynamicports baddynamicports;
124 struct baddynamicports rootonlyports;
125 struct pool inpcb_pool;
126 int inpcb_pool_initialized = 0;
127 
128 int in_pcbresize (struct inpcbtable *, int);
129 
130 #define	INPCBHASH_LOADFACTOR(_x)	(((_x) * 3) / 4)
131 
132 struct inpcbhead *in_pcbhash(struct inpcbtable *, int,
133     const struct in_addr *, u_short, const struct in_addr *, u_short);
134 struct inpcbhead *in6_pcbhash(struct inpcbtable *, int,
135     const struct in6_addr *, u_short, const struct in6_addr *, u_short);
136 struct inpcbhead *in_pcblhash(struct inpcbtable *, int, u_short);
137 
138 struct inpcbhead *
139 in_pcbhash(struct inpcbtable *table, int rdom,
140     const struct in_addr *faddr, u_short fport,
141     const struct in_addr *laddr, u_short lport)
142 {
143 	SIPHASH_CTX ctx;
144 	u_int32_t nrdom = htonl(rdom);
145 
146 	SipHash24_Init(&ctx, &table->inpt_key);
147 	SipHash24_Update(&ctx, &nrdom, sizeof(nrdom));
148 	SipHash24_Update(&ctx, faddr, sizeof(*faddr));
149 	SipHash24_Update(&ctx, &fport, sizeof(fport));
150 	SipHash24_Update(&ctx, laddr, sizeof(*laddr));
151 	SipHash24_Update(&ctx, &lport, sizeof(lport));
152 
153 	return (&table->inpt_hashtbl[SipHash24_End(&ctx) & table->inpt_hash]);
154 }
155 
156 #define	INPCBHASH(table, faddr, fport, laddr, lport, rdom) \
157 	in_pcbhash(table, rdom, faddr, fport, laddr, lport)
158 
159 struct inpcbhead *
160 in6_pcbhash(struct inpcbtable *table, int rdom,
161     const struct in6_addr *faddr, u_short fport,
162     const struct in6_addr *laddr, u_short lport)
163 {
164 	SIPHASH_CTX ctx;
165 	u_int32_t nrdom = htonl(rdom);
166 
167 	SipHash24_Init(&ctx, &table->inpt_key);
168 	SipHash24_Update(&ctx, &nrdom, sizeof(nrdom));
169 	SipHash24_Update(&ctx, faddr, sizeof(*faddr));
170 	SipHash24_Update(&ctx, &fport, sizeof(fport));
171 	SipHash24_Update(&ctx, laddr, sizeof(*laddr));
172 	SipHash24_Update(&ctx, &lport, sizeof(lport));
173 
174 	return (&table->inpt_hashtbl[SipHash24_End(&ctx) & table->inpt_hash]);
175 }
176 
177 #define	IN6PCBHASH(table, faddr, fport, laddr, lport, rdom) \
178 	in6_pcbhash(table, rdom, faddr, fport, laddr, lport)
179 
180 struct inpcbhead *
181 in_pcblhash(struct inpcbtable *table, int rdom, u_short lport)
182 {
183 	SIPHASH_CTX ctx;
184 	u_int32_t nrdom = htonl(rdom);
185 
186 	SipHash24_Init(&ctx, &table->inpt_key);
187 	SipHash24_Update(&ctx, &nrdom, sizeof(nrdom));
188 	SipHash24_Update(&ctx, &lport, sizeof(lport));
189 
190 	return (&table->inpt_lhashtbl[SipHash24_End(&ctx) & table->inpt_lhash]);
191 }
192 
193 #define	INPCBLHASH(table, lport, rdom) in_pcblhash(table, rdom, lport)
194 
195 void
196 in_pcbinit(struct inpcbtable *table, int hashsize)
197 {
198 
199 	TAILQ_INIT(&table->inpt_queue);
200 	table->inpt_hashtbl = hashinit(hashsize, M_PCB, M_NOWAIT,
201 	    &table->inpt_hash);
202 	if (table->inpt_hashtbl == NULL)
203 		panic("in_pcbinit: hashinit failed");
204 	table->inpt_lhashtbl = hashinit(hashsize, M_PCB, M_NOWAIT,
205 	    &table->inpt_lhash);
206 	if (table->inpt_lhashtbl == NULL)
207 		panic("in_pcbinit: hashinit failed for lport");
208 	table->inpt_count = 0;
209 	arc4random_buf(&table->inpt_key, sizeof(table->inpt_key));
210 }
211 
212 /*
213  * Check if the specified port is invalid for dynamic allocation.
214  */
215 int
216 in_baddynamic(u_int16_t port, u_int16_t proto)
217 {
218 	switch (proto) {
219 	case IPPROTO_TCP:
220 		return (DP_ISSET(baddynamicports.tcp, port));
221 	case IPPROTO_UDP:
222 #ifdef IPSEC
223 		/* Cannot preset this as it is a sysctl */
224 		if (port == udpencap_port)
225 			return (1);
226 #endif
227 		return (DP_ISSET(baddynamicports.udp, port));
228 	default:
229 		return (0);
230 	}
231 }
232 
233 int
234 in_rootonly(u_int16_t port, u_int16_t proto)
235 {
236 	switch (proto) {
237 	case IPPROTO_TCP:
238 		return (port < IPPORT_RESERVED ||
239 		    DP_ISSET(rootonlyports.tcp, port));
240 	case IPPROTO_UDP:
241 		return (port < IPPORT_RESERVED ||
242 		    DP_ISSET(rootonlyports.udp, port));
243 	default:
244 		return (0);
245 	}
246 }
247 
248 int
249 in_pcballoc(struct socket *so, struct inpcbtable *table)
250 {
251 	struct inpcb *inp;
252 	struct inpcbhead *head;
253 
254 	NET_ASSERT_LOCKED();
255 
256 	if (inpcb_pool_initialized == 0) {
257 		pool_init(&inpcb_pool, sizeof(struct inpcb), 0,
258 		    IPL_SOFTNET, 0, "inpcbpl", NULL);
259 		inpcb_pool_initialized = 1;
260 	}
261 	inp = pool_get(&inpcb_pool, PR_NOWAIT|PR_ZERO);
262 	if (inp == NULL)
263 		return (ENOBUFS);
264 	inp->inp_table = table;
265 	inp->inp_socket = so;
266 	inp->inp_seclevel[SL_AUTH] = IPSEC_AUTH_LEVEL_DEFAULT;
267 	inp->inp_seclevel[SL_ESP_TRANS] = IPSEC_ESP_TRANS_LEVEL_DEFAULT;
268 	inp->inp_seclevel[SL_ESP_NETWORK] = IPSEC_ESP_NETWORK_LEVEL_DEFAULT;
269 	inp->inp_seclevel[SL_IPCOMP] = IPSEC_IPCOMP_LEVEL_DEFAULT;
270 	inp->inp_rtableid = curproc->p_p->ps_rtableid;
271 	if (table->inpt_hash != 0 &&
272 	    table->inpt_count++ > INPCBHASH_LOADFACTOR(table->inpt_hash))
273 		(void)in_pcbresize(table, (table->inpt_hash + 1) * 2);
274 	TAILQ_INSERT_HEAD(&table->inpt_queue, inp, inp_queue);
275 	head = INPCBLHASH(table, inp->inp_lport, inp->inp_rtableid);
276 	LIST_INSERT_HEAD(head, inp, inp_lhash);
277 #ifdef INET6
278 	if (sotopf(so) == PF_INET6)
279 		head = IN6PCBHASH(table, &inp->inp_faddr6, inp->inp_fport,
280 		    &inp->inp_laddr6, inp->inp_lport,
281 		    rtable_l2(inp->inp_rtableid));
282 	else
283 #endif /* INET6 */
284 		head = INPCBHASH(table, &inp->inp_faddr, inp->inp_fport,
285 		    &inp->inp_laddr, inp->inp_lport,
286 		    rtable_l2(inp->inp_rtableid));
287 	LIST_INSERT_HEAD(head, inp, inp_hash);
288 	so->so_pcb = inp;
289 	inp->inp_hops = -1;
290 
291 #ifdef INET6
292 	/*
293 	 * Small change in this function to set the INP_IPV6 flag so routines
294 	 * outside pcb-specific routines don't need to use sotopf(), and all
295 	 * of its pointer chasing, later.
296 	 */
297 	if (sotopf(so) == PF_INET6)
298 		inp->inp_flags = INP_IPV6;
299 	inp->inp_cksum6 = -1;
300 #endif /* INET6 */
301 	return (0);
302 }
303 
304 int
305 in_pcbbind(struct inpcb *inp, struct mbuf *nam, struct proc *p)
306 {
307 	struct socket *so = inp->inp_socket;
308 	u_int16_t lport = 0;
309 	int wild = 0;
310 	void *laddr = &zeroin46_addr;
311 	int error;
312 
313 	if (inp->inp_lport)
314 		return (EINVAL);
315 
316 	if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0 &&
317 	    ((so->so_proto->pr_flags & PR_CONNREQUIRED) == 0 ||
318 	     (so->so_options & SO_ACCEPTCONN) == 0))
319 		wild = INPLOOKUP_WILDCARD;
320 
321 	switch (sotopf(so)) {
322 #ifdef INET6
323 	case PF_INET6:
324 		if (!IN6_IS_ADDR_UNSPECIFIED(&inp->inp_laddr6))
325 			return (EINVAL);
326 		wild |= INPLOOKUP_IPV6;
327 
328 		if (nam) {
329 			struct sockaddr_in6 *sin6;
330 
331 			if ((error = in6_nam2sin6(nam, &sin6)))
332 				return (error);
333 			if ((error = in6_pcbaddrisavail(inp, sin6, wild, p)))
334 				return (error);
335 			laddr = &sin6->sin6_addr;
336 			lport = sin6->sin6_port;
337 		}
338 		break;
339 #endif
340 	case PF_INET:
341 		if (inp->inp_laddr.s_addr != INADDR_ANY)
342 			return (EINVAL);
343 
344 		if (nam) {
345 			struct sockaddr_in *sin;
346 
347 			if ((error = in_nam2sin(nam, &sin)))
348 				return (error);
349 			if ((error = in_pcbaddrisavail(inp, sin, wild, p)))
350 				return (error);
351 			laddr = &sin->sin_addr;
352 			lport = sin->sin_port;
353 		}
354 		break;
355 	default:
356 		return (EINVAL);
357 	}
358 
359 	if (lport == 0) {
360 		if ((error = in_pcbpickport(&lport, laddr, wild, inp, p)))
361 			return (error);
362 	} else {
363 		if (in_rootonly(ntohs(lport), so->so_proto->pr_protocol) &&
364 		    suser(p, 0) != 0)
365 			return (EACCES);
366 	}
367 	if (nam) {
368 		switch (sotopf(so)) {
369 #ifdef INET6
370 		case PF_INET6:
371 			inp->inp_laddr6 = *(struct in6_addr *)laddr;
372 			break;
373 #endif
374 		case PF_INET:
375 			inp->inp_laddr = *(struct in_addr *)laddr;
376 			break;
377 		}
378 	}
379 	inp->inp_lport = lport;
380 	in_pcbrehash(inp);
381 	return (0);
382 }
383 
384 int
385 in_pcbaddrisavail(struct inpcb *inp, struct sockaddr_in *sin, int wild,
386     struct proc *p)
387 {
388 	struct socket *so = inp->inp_socket;
389 	struct inpcbtable *table = inp->inp_table;
390 	u_int16_t lport = sin->sin_port;
391 	int reuseport = (so->so_options & SO_REUSEPORT);
392 
393 	if (IN_MULTICAST(sin->sin_addr.s_addr)) {
394 		/*
395 		 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
396 		 * allow complete duplication of binding if
397 		 * SO_REUSEPORT is set, or if SO_REUSEADDR is set
398 		 * and a multicast address is bound on both
399 		 * new and duplicated sockets.
400 		 */
401 		if (so->so_options & (SO_REUSEADDR|SO_REUSEPORT))
402 			reuseport = SO_REUSEADDR|SO_REUSEPORT;
403 	} else if (sin->sin_addr.s_addr != INADDR_ANY) {
404 		/*
405 		 * we must check that we are binding to an address we
406 		 * own except when:
407 		 * - SO_BINDANY is set or
408 		 * - we are binding a UDP socket to 255.255.255.255 or
409 		 * - we are binding a UDP socket to one of our broadcast
410 		 *   addresses
411 		 */
412 		if (!ISSET(so->so_options, SO_BINDANY) &&
413 		    !(so->so_type == SOCK_DGRAM &&
414 		    sin->sin_addr.s_addr == INADDR_BROADCAST) &&
415 		    !(so->so_type == SOCK_DGRAM &&
416 		    in_broadcast(sin->sin_addr, inp->inp_rtableid))) {
417 			struct ifaddr *ia;
418 
419 			sin->sin_port = 0;
420 			memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
421 			ia = ifa_ifwithaddr(sintosa(sin), inp->inp_rtableid);
422 			sin->sin_port = lport;
423 
424 			if (ia == NULL)
425 				return (EADDRNOTAVAIL);
426 		}
427 	}
428 	if (lport) {
429 		struct inpcb *t;
430 
431 		if (so->so_euid) {
432 			t = in_pcblookup_local(table, &sin->sin_addr, lport,
433 			    INPLOOKUP_WILDCARD, inp->inp_rtableid);
434 			if (t && (so->so_euid != t->inp_socket->so_euid))
435 				return (EADDRINUSE);
436 		}
437 		t = in_pcblookup_local(table, &sin->sin_addr, lport,
438 		    wild, inp->inp_rtableid);
439 		if (t && (reuseport & t->inp_socket->so_options) == 0)
440 			return (EADDRINUSE);
441 	}
442 
443 	return (0);
444 }
445 
446 int
447 in_pcbpickport(u_int16_t *lport, void *laddr, int wild, struct inpcb *inp,
448     struct proc *p)
449 {
450 	struct socket *so = inp->inp_socket;
451 	struct inpcbtable *table = inp->inp_table;
452 	u_int16_t first, last, lower, higher, candidate, localport;
453 	int count;
454 
455 	if (inp->inp_flags & INP_HIGHPORT) {
456 		first = ipport_hifirstauto;	/* sysctl */
457 		last = ipport_hilastauto;
458 	} else if (inp->inp_flags & INP_LOWPORT) {
459 		if (suser(p, 0))
460 			return (EACCES);
461 		first = IPPORT_RESERVED-1; /* 1023 */
462 		last = 600;		   /* not IPPORT_RESERVED/2 */
463 	} else {
464 		first = ipport_firstauto;	/* sysctl */
465 		last = ipport_lastauto;
466 	}
467 	if (first < last) {
468 		lower = first;
469 		higher = last;
470 	} else {
471 		lower = last;
472 		higher = first;
473 	}
474 
475 	/*
476 	 * Simple check to ensure all ports are not used up causing
477 	 * a deadlock here.
478 	 */
479 
480 	count = higher - lower;
481 	candidate = lower + arc4random_uniform(count);
482 
483 	do {
484 		if (count-- < 0)	/* completely used? */
485 			return (EADDRNOTAVAIL);
486 		++candidate;
487 		if (candidate < lower || candidate > higher)
488 			candidate = lower;
489 		localport = htons(candidate);
490 	} while (in_baddynamic(candidate, so->so_proto->pr_protocol) ||
491 	    in_pcblookup_local(table, laddr, localport, wild,
492 	    inp->inp_rtableid));
493 	*lport = localport;
494 
495 	return (0);
496 }
497 
498 /*
499  * Connect from a socket to a specified address.
500  * Both address and port must be specified in argument sin.
501  * If don't have a local address for this socket yet,
502  * then pick one.
503  */
504 int
505 in_pcbconnect(struct inpcb *inp, struct mbuf *nam)
506 {
507 	struct in_addr *ina = NULL;
508 	struct sockaddr_in *sin;
509 	int error;
510 
511 #ifdef INET6
512 	if (sotopf(inp->inp_socket) == PF_INET6)
513 		return (in6_pcbconnect(inp, nam));
514 	if ((inp->inp_flags & INP_IPV6) != 0)
515 		panic("IPv6 pcb passed into in_pcbconnect");
516 #endif /* INET6 */
517 
518 	if ((error = in_nam2sin(nam, &sin)))
519 		return (error);
520 	if (sin->sin_port == 0)
521 		return (EADDRNOTAVAIL);
522 	error = in_pcbselsrc(&ina, sin, inp);
523 	if (error)
524 		return (error);
525 
526 	if (in_pcbhashlookup(inp->inp_table, sin->sin_addr, sin->sin_port,
527 	    *ina, inp->inp_lport, inp->inp_rtableid) != NULL)
528 		return (EADDRINUSE);
529 
530 	KASSERT(inp->inp_laddr.s_addr == INADDR_ANY || inp->inp_lport);
531 
532 	if (inp->inp_laddr.s_addr == INADDR_ANY) {
533 		if (inp->inp_lport == 0) {
534 			error = in_pcbbind(inp, NULL, curproc);
535 			if (error)
536 				return (error);
537 			if (in_pcbhashlookup(inp->inp_table, sin->sin_addr,
538 			    sin->sin_port, *ina, inp->inp_lport,
539 			    inp->inp_rtableid) != NULL) {
540 				inp->inp_lport = 0;
541 				return (EADDRINUSE);
542 			}
543 		}
544 		inp->inp_laddr = *ina;
545 	}
546 	inp->inp_faddr = sin->sin_addr;
547 	inp->inp_fport = sin->sin_port;
548 	in_pcbrehash(inp);
549 #ifdef IPSEC
550 	{
551 		/* Cause an IPsec SA to be established. */
552 		/* error is just ignored */
553 		ipsp_spd_inp(NULL, AF_INET, 0, &error, IPSP_DIRECTION_OUT,
554 		    NULL, inp, NULL);
555 	}
556 #endif
557 	return (0);
558 }
559 
560 void
561 in_pcbdisconnect(struct inpcb *inp)
562 {
563 	switch (sotopf(inp->inp_socket)) {
564 #ifdef INET6
565 	case PF_INET6:
566 		inp->inp_faddr6 = in6addr_any;
567 		break;
568 #endif
569 	case PF_INET:
570 		inp->inp_faddr.s_addr = INADDR_ANY;
571 		break;
572 	}
573 
574 	inp->inp_fport = 0;
575 	in_pcbrehash(inp);
576 	if (inp->inp_socket->so_state & SS_NOFDREF)
577 		in_pcbdetach(inp);
578 }
579 
580 void
581 in_pcbdetach(struct inpcb *inp)
582 {
583 	struct socket *so = inp->inp_socket;
584 
585 	NET_ASSERT_LOCKED();
586 
587 	so->so_pcb = 0;
588 	sofree(so);
589 	m_freem(inp->inp_options);
590 	if (inp->inp_route.ro_rt) {
591 		rtfree(inp->inp_route.ro_rt);
592 		inp->inp_route.ro_rt = NULL;
593 	}
594 #ifdef INET6
595 	if (inp->inp_flags & INP_IPV6) {
596 		ip6_freepcbopts(inp->inp_outputopts6);
597 		ip6_freemoptions(inp->inp_moptions6);
598 	} else
599 #endif
600 		ip_freemoptions(inp->inp_moptions);
601 #if NPF > 0
602 	if (inp->inp_pf_sk) {
603 		pf_remove_divert_state(inp->inp_pf_sk);
604 		/* pf_remove_divert_state() may have detached the state */
605 		pf_inp_unlink(inp);
606 	}
607 #endif
608 	LIST_REMOVE(inp, inp_lhash);
609 	LIST_REMOVE(inp, inp_hash);
610 	TAILQ_REMOVE(&inp->inp_table->inpt_queue, inp, inp_queue);
611 	inp->inp_table->inpt_count--;
612 	pool_put(&inpcb_pool, inp);
613 }
614 
615 void
616 in_setsockaddr(struct inpcb *inp, struct mbuf *nam)
617 {
618 	struct sockaddr_in *sin;
619 
620 	nam->m_len = sizeof(*sin);
621 	sin = mtod(nam, struct sockaddr_in *);
622 	memset(sin, 0, sizeof(*sin));
623 	sin->sin_family = AF_INET;
624 	sin->sin_len = sizeof(*sin);
625 	sin->sin_port = inp->inp_lport;
626 	sin->sin_addr = inp->inp_laddr;
627 }
628 
629 void
630 in_setpeeraddr(struct inpcb *inp, struct mbuf *nam)
631 {
632 	struct sockaddr_in *sin;
633 
634 #ifdef INET6
635 	if (sotopf(inp->inp_socket) == PF_INET6) {
636 		in6_setpeeraddr(inp, nam);
637 		return;
638 	}
639 #endif /* INET6 */
640 
641 	nam->m_len = sizeof(*sin);
642 	sin = mtod(nam, struct sockaddr_in *);
643 	memset(sin, 0, sizeof(*sin));
644 	sin->sin_family = AF_INET;
645 	sin->sin_len = sizeof(*sin);
646 	sin->sin_port = inp->inp_fport;
647 	sin->sin_addr = inp->inp_faddr;
648 }
649 
650 /*
651  * Pass some notification to all connections of a protocol
652  * associated with address dst.  The "usual action" will be
653  * taken, depending on the ctlinput cmd.  The caller must filter any
654  * cmds that are uninteresting (e.g., no error in the map).
655  * Call the protocol specific routine (if any) to report
656  * any errors for each matching socket.
657  */
658 void
659 in_pcbnotifyall(struct inpcbtable *table, struct sockaddr *dst, u_int rdomain,
660     int errno, void (*notify)(struct inpcb *, int))
661 {
662 	struct inpcb *inp, *ninp;
663 	struct in_addr faddr;
664 
665 	NET_ASSERT_LOCKED();
666 
667 #ifdef INET6
668 	/*
669 	 * See in6_pcbnotify() for IPv6 codepath.  By the time this
670 	 * gets called, the addresses passed are either definitely IPv4 or
671 	 * IPv6; *_pcbnotify() never gets called with v4-mapped v6 addresses.
672 	 */
673 #endif /* INET6 */
674 
675 	if (dst->sa_family != AF_INET)
676 		return;
677 	faddr = satosin(dst)->sin_addr;
678 	if (faddr.s_addr == INADDR_ANY)
679 		return;
680 
681 	rdomain = rtable_l2(rdomain);
682 	TAILQ_FOREACH_SAFE(inp, &table->inpt_queue, inp_queue, ninp) {
683 #ifdef INET6
684 		if (inp->inp_flags & INP_IPV6)
685 			continue;
686 #endif
687 		if (inp->inp_faddr.s_addr != faddr.s_addr ||
688 		    rtable_l2(inp->inp_rtableid) != rdomain ||
689 		    inp->inp_socket == 0) {
690 			continue;
691 		}
692 		if (notify)
693 			(*notify)(inp, errno);
694 	}
695 }
696 
697 /*
698  * Check for alternatives when higher level complains
699  * about service problems.  For now, invalidate cached
700  * routing information.  If the route was created dynamically
701  * (by a redirect), time to try a default gateway again.
702  */
703 void
704 in_losing(struct inpcb *inp)
705 {
706 	struct rtentry *rt;
707 	struct rt_addrinfo info;
708 	struct sockaddr_in6 sa_mask;
709 
710 	if ((rt = inp->inp_route.ro_rt)) {
711 		inp->inp_route.ro_rt = 0;
712 
713 		memset(&info, 0, sizeof(info));
714 		info.rti_flags = rt->rt_flags;
715 		info.rti_info[RTAX_DST] = &inp->inp_route.ro_dst;
716 		info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
717 		info.rti_info[RTAX_NETMASK] = rt_plen2mask(rt, &sa_mask);
718 
719 		KERNEL_LOCK();
720 		rtm_miss(RTM_LOSING, &info, rt->rt_flags, rt->rt_priority,
721 		    rt->rt_ifidx, 0, inp->inp_rtableid);
722 		KERNEL_UNLOCK();
723 		if (rt->rt_flags & RTF_DYNAMIC) {
724 			struct ifnet *ifp;
725 
726 			ifp = if_get(rt->rt_ifidx);
727 			/*
728 			 * If the interface is gone, all its attached
729 			 * route entries have been removed from the table,
730 			 * so we're dealing with a stale cache and have
731 			 * nothing to do.
732 			 */
733 			if (ifp != NULL) {
734 				rtrequest_delete(&info, rt->rt_priority, ifp,
735 				    NULL, inp->inp_rtableid);
736 			}
737 			if_put(ifp);
738 		}
739 		/*
740 		 * A new route can be allocated
741 		 * the next time output is attempted.
742 		 * rtfree() needs to be called in anycase because the inp
743 		 * is still holding a reference to rt.
744 		 */
745 		rtfree(rt);
746 	}
747 }
748 
749 /*
750  * After a routing change, flush old routing
751  * and allocate a (hopefully) better one.
752  */
753 void
754 in_rtchange(struct inpcb *inp, int errno)
755 {
756 	if (inp->inp_route.ro_rt) {
757 		rtfree(inp->inp_route.ro_rt);
758 		inp->inp_route.ro_rt = 0;
759 		/*
760 		 * A new route can be allocated the next time
761 		 * output is attempted.
762 		 */
763 	}
764 }
765 
766 struct inpcb *
767 in_pcblookup_local(struct inpcbtable *table, void *laddrp, u_int lport_arg,
768     int flags, u_int rdomain)
769 {
770 	struct inpcb *inp, *match = NULL;
771 	int matchwild = 3, wildcard;
772 	u_int16_t lport = lport_arg;
773 	struct in_addr laddr = *(struct in_addr *)laddrp;
774 #ifdef INET6
775 	struct in6_addr *laddr6 = (struct in6_addr *)laddrp;
776 #endif
777 	struct inpcbhead *head;
778 
779 	rdomain = rtable_l2(rdomain);	/* convert passed rtableid to rdomain */
780 	head = INPCBLHASH(table, lport, rdomain);
781 	LIST_FOREACH(inp, head, inp_lhash) {
782 		if (rtable_l2(inp->inp_rtableid) != rdomain)
783 			continue;
784 		if (inp->inp_lport != lport)
785 			continue;
786 		wildcard = 0;
787 #ifdef INET6
788 		if (ISSET(flags, INPLOOKUP_IPV6)) {
789 			if (!ISSET(inp->inp_flags, INP_IPV6))
790 				continue;
791 
792 			if (!IN6_IS_ADDR_UNSPECIFIED(&inp->inp_faddr6))
793 				wildcard++;
794 
795 			if (!IN6_ARE_ADDR_EQUAL(&inp->inp_laddr6, laddr6)) {
796 				if (IN6_IS_ADDR_UNSPECIFIED(&inp->inp_laddr6) ||
797 				    IN6_IS_ADDR_UNSPECIFIED(laddr6))
798 					wildcard++;
799 				else
800 					continue;
801 			}
802 
803 		} else
804 #endif /* INET6 */
805 		{
806 #ifdef INET6
807 			if (ISSET(inp->inp_flags, INP_IPV6))
808 				continue;
809 #endif /* INET6 */
810 
811 			if (inp->inp_faddr.s_addr != INADDR_ANY)
812 				wildcard++;
813 
814 			if (inp->inp_laddr.s_addr != laddr.s_addr) {
815 				if (inp->inp_laddr.s_addr == INADDR_ANY ||
816 				    laddr.s_addr == INADDR_ANY)
817 					wildcard++;
818 				else
819 					continue;
820 			}
821 
822 		}
823 		if ((!wildcard || (flags & INPLOOKUP_WILDCARD)) &&
824 		    wildcard < matchwild) {
825 			match = inp;
826 			if ((matchwild = wildcard) == 0)
827 				break;
828 		}
829 	}
830 	return (match);
831 }
832 
833 struct rtentry *
834 in_pcbrtentry(struct inpcb *inp)
835 {
836 	struct route *ro;
837 
838 	ro = &inp->inp_route;
839 
840 	/* check if route is still valid */
841 	if (!rtisvalid(ro->ro_rt)) {
842 		rtfree(ro->ro_rt);
843 		ro->ro_rt = NULL;
844 	}
845 
846 	/*
847 	 * No route yet, so try to acquire one.
848 	 */
849 	if (ro->ro_rt == NULL) {
850 #ifdef INET6
851 		memset(ro, 0, sizeof(struct route_in6));
852 #else
853 		memset(ro, 0, sizeof(struct route));
854 #endif
855 
856 		switch(sotopf(inp->inp_socket)) {
857 #ifdef INET6
858 		case PF_INET6:
859 			if (IN6_IS_ADDR_UNSPECIFIED(&inp->inp_faddr6))
860 				break;
861 			ro->ro_dst.sa_family = AF_INET6;
862 			ro->ro_dst.sa_len = sizeof(struct sockaddr_in6);
863 			satosin6(&ro->ro_dst)->sin6_addr = inp->inp_faddr6;
864 			ro->ro_tableid = inp->inp_rtableid;
865 			ro->ro_rt = rtalloc_mpath(&ro->ro_dst,
866 			    &inp->inp_laddr6.s6_addr32[0], ro->ro_tableid);
867 			break;
868 #endif /* INET6 */
869 		case PF_INET:
870 			if (inp->inp_faddr.s_addr == INADDR_ANY)
871 				break;
872 			ro->ro_dst.sa_family = AF_INET;
873 			ro->ro_dst.sa_len = sizeof(struct sockaddr_in);
874 			satosin(&ro->ro_dst)->sin_addr = inp->inp_faddr;
875 			ro->ro_tableid = inp->inp_rtableid;
876 			ro->ro_rt = rtalloc_mpath(&ro->ro_dst,
877 			    &inp->inp_laddr.s_addr, ro->ro_tableid);
878 			break;
879 		}
880 	}
881 	return (ro->ro_rt);
882 }
883 
884 /*
885  * Return an IPv4 address, which is the most appropriate for a given
886  * destination.
887  * If necessary, this function lookups the routing table and returns
888  * an entry to the caller for later use.
889  */
890 int
891 in_pcbselsrc(struct in_addr **insrc, struct sockaddr_in *sin,
892     struct inpcb *inp)
893 {
894 	struct ip_moptions *mopts = inp->inp_moptions;
895 	struct route *ro = &inp->inp_route;
896 	struct in_addr *laddr = &inp->inp_laddr;
897 	u_int rtableid = inp->inp_rtableid;
898 
899 	struct sockaddr_in *sin2;
900 	struct in_ifaddr *ia = NULL;
901 
902 	/*
903 	 * If the socket(if any) is already bound, use that bound address
904 	 * unless it is INADDR_ANY or INADDR_BROADCAST.
905 	 */
906 	if (laddr && laddr->s_addr != INADDR_ANY &&
907 	    laddr->s_addr != INADDR_BROADCAST) {
908 		*insrc = laddr;
909 		return (0);
910 	}
911 
912 	/*
913 	 * If the destination address is multicast and an outgoing
914 	 * interface has been set as a multicast option, use the
915 	 * address of that interface as our source address.
916 	 */
917 	if (IN_MULTICAST(sin->sin_addr.s_addr) && mopts != NULL) {
918 		struct ifnet *ifp;
919 
920 		ifp = if_get(mopts->imo_ifidx);
921 		if (ifp != NULL) {
922 			if (ifp->if_rdomain == rtable_l2(rtableid))
923 				IFP_TO_IA(ifp, ia);
924 			if (ia == NULL) {
925 				if_put(ifp);
926 				return (EADDRNOTAVAIL);
927 			}
928 
929 			*insrc = &ia->ia_addr.sin_addr;
930 			if_put(ifp);
931 			return (0);
932 		}
933 	}
934 	/*
935 	 * If route is known or can be allocated now,
936 	 * our src addr is taken from the i/f, else punt.
937 	 */
938 	if (!rtisvalid(ro->ro_rt) || (ro->ro_tableid != rtableid) ||
939 	    (satosin(&ro->ro_dst)->sin_addr.s_addr != sin->sin_addr.s_addr)) {
940 		rtfree(ro->ro_rt);
941 		ro->ro_rt = NULL;
942 	}
943 	if (ro->ro_rt == NULL) {
944 		/* No route yet, so try to acquire one */
945 		ro->ro_dst.sa_family = AF_INET;
946 		ro->ro_dst.sa_len = sizeof(struct sockaddr_in);
947 		satosin(&ro->ro_dst)->sin_addr = sin->sin_addr;
948 		ro->ro_tableid = rtableid;
949 		ro->ro_rt = rtalloc_mpath(&ro->ro_dst, NULL, ro->ro_tableid);
950 
951 		/*
952 		 * It is important to zero out the rest of the
953 		 * struct sockaddr_in when mixing v6 & v4!
954 		 */
955 		sin2 = satosin(&ro->ro_dst);
956 		memset(sin2->sin_zero, 0, sizeof(sin2->sin_zero));
957 	}
958 	/*
959 	 * If we found a route, use the address
960 	 * corresponding to the outgoing interface.
961 	 */
962 	if (ro->ro_rt != NULL)
963 		ia = ifatoia(ro->ro_rt->rt_ifa);
964 
965 	if (ia == NULL)
966 		return (EADDRNOTAVAIL);
967 
968 	*insrc = &ia->ia_addr.sin_addr;
969 	return (0);
970 }
971 
972 void
973 in_pcbrehash(struct inpcb *inp)
974 {
975 	struct inpcbtable *table = inp->inp_table;
976 	struct inpcbhead *head;
977 
978 	NET_ASSERT_LOCKED();
979 
980 	LIST_REMOVE(inp, inp_lhash);
981 	head = INPCBLHASH(table, inp->inp_lport, inp->inp_rtableid);
982 	LIST_INSERT_HEAD(head, inp, inp_lhash);
983 	LIST_REMOVE(inp, inp_hash);
984 #ifdef INET6
985 	if (inp->inp_flags & INP_IPV6)
986 		head = IN6PCBHASH(table, &inp->inp_faddr6, inp->inp_fport,
987 		    &inp->inp_laddr6, inp->inp_lport,
988 		    rtable_l2(inp->inp_rtableid));
989 	else
990 #endif /* INET6 */
991 		head = INPCBHASH(table, &inp->inp_faddr, inp->inp_fport,
992 		    &inp->inp_laddr, inp->inp_lport,
993 		    rtable_l2(inp->inp_rtableid));
994 	LIST_INSERT_HEAD(head, inp, inp_hash);
995 }
996 
997 int
998 in_pcbresize(struct inpcbtable *table, int hashsize)
999 {
1000 	u_long nhash, nlhash;
1001 	void *nhashtbl, *nlhashtbl, *ohashtbl, *olhashtbl;
1002 	struct inpcb *inp0, *inp1;
1003 
1004 	ohashtbl = table->inpt_hashtbl;
1005 	olhashtbl = table->inpt_lhashtbl;
1006 
1007 	nhashtbl = hashinit(hashsize, M_PCB, M_NOWAIT, &nhash);
1008 	nlhashtbl = hashinit(hashsize, M_PCB, M_NOWAIT, &nlhash);
1009 	if (nhashtbl == NULL || nlhashtbl == NULL) {
1010 		if (nhashtbl != NULL)
1011 			free(nhashtbl, M_PCB, 0);
1012 		if (nlhashtbl != NULL)
1013 			free(nlhashtbl, M_PCB, 0);
1014 		return (ENOBUFS);
1015 	}
1016 	table->inpt_hashtbl = nhashtbl;
1017 	table->inpt_lhashtbl = nlhashtbl;
1018 	table->inpt_hash = nhash;
1019 	table->inpt_lhash = nlhash;
1020 	arc4random_buf(&table->inpt_key, sizeof(table->inpt_key));
1021 
1022 	TAILQ_FOREACH_SAFE(inp0, &table->inpt_queue, inp_queue, inp1) {
1023 		in_pcbrehash(inp0);
1024 	}
1025 	free(ohashtbl, M_PCB, 0);
1026 	free(olhashtbl, M_PCB, 0);
1027 
1028 	return (0);
1029 }
1030 
1031 #ifdef DIAGNOSTIC
1032 int	in_pcbnotifymiss = 0;
1033 #endif
1034 
1035 /*
1036  * The in(6)_pcbhashlookup functions are used to locate connected sockets
1037  * quickly:
1038  *     faddr.fport <-> laddr.lport
1039  * No wildcard matching is done so that listening sockets are not found.
1040  * If the functions return NULL in(6)_pcblookup_listen can be used to
1041  * find a listening/bound socket that may accept the connection.
1042  * After those two lookups no other are necessary.
1043  */
1044 struct inpcb *
1045 in_pcbhashlookup(struct inpcbtable *table, struct in_addr faddr,
1046     u_int fport_arg, struct in_addr laddr, u_int lport_arg, u_int rdomain)
1047 {
1048 	struct inpcbhead *head;
1049 	struct inpcb *inp;
1050 	u_int16_t fport = fport_arg, lport = lport_arg;
1051 
1052 	rdomain = rtable_l2(rdomain);	/* convert passed rtableid to rdomain */
1053 	head = INPCBHASH(table, &faddr, fport, &laddr, lport, rdomain);
1054 	LIST_FOREACH(inp, head, inp_hash) {
1055 #ifdef INET6
1056 		if (inp->inp_flags & INP_IPV6)
1057 			continue;	/*XXX*/
1058 #endif
1059 		if (inp->inp_faddr.s_addr == faddr.s_addr &&
1060 		    inp->inp_fport == fport && inp->inp_lport == lport &&
1061 		    inp->inp_laddr.s_addr == laddr.s_addr &&
1062 		    rtable_l2(inp->inp_rtableid) == rdomain) {
1063 			/*
1064 			 * Move this PCB to the head of hash chain so that
1065 			 * repeated accesses are quicker.  This is analogous to
1066 			 * the historic single-entry PCB cache.
1067 			 */
1068 			if (inp != LIST_FIRST(head)) {
1069 				LIST_REMOVE(inp, inp_hash);
1070 				LIST_INSERT_HEAD(head, inp, inp_hash);
1071 			}
1072 			break;
1073 		}
1074 	}
1075 #ifdef DIAGNOSTIC
1076 	if (inp == NULL && in_pcbnotifymiss) {
1077 		printf("%s: faddr=%08x fport=%d laddr=%08x lport=%d rdom=%d\n",
1078 		    __func__, ntohl(faddr.s_addr), ntohs(fport),
1079 		    ntohl(laddr.s_addr), ntohs(lport), rdomain);
1080 	}
1081 #endif
1082 	return (inp);
1083 }
1084 
1085 #ifdef INET6
1086 struct inpcb *
1087 in6_pcbhashlookup(struct inpcbtable *table, const struct in6_addr *faddr,
1088     u_int fport_arg, const struct in6_addr *laddr, u_int lport_arg,
1089     u_int rtable)
1090 {
1091 	struct inpcbhead *head;
1092 	struct inpcb *inp;
1093 	u_int16_t fport = fport_arg, lport = lport_arg;
1094 
1095 	rtable = rtable_l2(rtable);	/* convert passed rtableid to rdomain */
1096 	head = IN6PCBHASH(table, faddr, fport, laddr, lport, rtable);
1097 	LIST_FOREACH(inp, head, inp_hash) {
1098 		if (!(inp->inp_flags & INP_IPV6))
1099 			continue;
1100 		if (IN6_ARE_ADDR_EQUAL(&inp->inp_faddr6, faddr) &&
1101 		    inp->inp_fport == fport && inp->inp_lport == lport &&
1102 		    IN6_ARE_ADDR_EQUAL(&inp->inp_laddr6, laddr) &&
1103 		    rtable_l2(inp->inp_rtableid) == rtable) {
1104 			/*
1105 			 * Move this PCB to the head of hash chain so that
1106 			 * repeated accesses are quicker.  This is analogous to
1107 			 * the historic single-entry PCB cache.
1108 			 */
1109 			if (inp != LIST_FIRST(head)) {
1110 				LIST_REMOVE(inp, inp_hash);
1111 				LIST_INSERT_HEAD(head, inp, inp_hash);
1112 			}
1113 			break;
1114 		}
1115 	}
1116 #ifdef DIAGNOSTIC
1117 	if (inp == NULL && in_pcbnotifymiss) {
1118 		printf("in6_pcbhashlookup: faddr=");
1119 		printf(" fport=%d laddr=", ntohs(fport));
1120 		printf(" lport=%d\n", ntohs(lport));
1121 	}
1122 #endif
1123 	return (inp);
1124 }
1125 #endif /* INET6 */
1126 
1127 /*
1128  * The in(6)_pcblookup_listen functions are used to locate listening
1129  * sockets quickly.  This are sockets with unspecified foreign address
1130  * and port:
1131  *		*.*     <-> laddr.lport
1132  *		*.*     <->     *.lport
1133  */
1134 struct inpcb *
1135 in_pcblookup_listen(struct inpcbtable *table, struct in_addr laddr,
1136     u_int lport_arg, struct mbuf *m, u_int rdomain)
1137 {
1138 	struct inpcbhead *head;
1139 	struct in_addr *key1, *key2;
1140 	struct inpcb *inp;
1141 	u_int16_t lport = lport_arg;
1142 
1143 	rdomain = rtable_l2(rdomain);	/* convert passed rtableid to rdomain */
1144 	key1 = &laddr;
1145 	key2 = &zeroin_addr;
1146 #if NPF > 0
1147 	if (m && m->m_pkthdr.pf.flags & PF_TAG_DIVERTED) {
1148 		struct pf_divert *divert;
1149 
1150 		divert = pf_find_divert(m);
1151 		KASSERT(divert != NULL);
1152 		switch (divert->type) {
1153 		case PF_DIVERT_TO:
1154 			key1 = key2 = &divert->addr.v4;
1155 			lport = divert->port;
1156 			break;
1157 		case PF_DIVERT_REPLY:
1158 			return (NULL);
1159 		default:
1160 			panic("%s: unknown divert type %d, mbuf %p, divert %p",
1161 			    __func__, divert->type, m, divert);
1162 		}
1163 	} else if (m && m->m_pkthdr.pf.flags & PF_TAG_TRANSLATE_LOCALHOST) {
1164 		key1 = &zeroin_addr;
1165 		key2 = &laddr;
1166 	}
1167 #endif
1168 
1169 	head = INPCBHASH(table, &zeroin_addr, 0, key1, lport, rdomain);
1170 	LIST_FOREACH(inp, head, inp_hash) {
1171 #ifdef INET6
1172 		if (inp->inp_flags & INP_IPV6)
1173 			continue;	/*XXX*/
1174 #endif
1175 		if (inp->inp_lport == lport && inp->inp_fport == 0 &&
1176 		    inp->inp_laddr.s_addr == key1->s_addr &&
1177 		    inp->inp_faddr.s_addr == INADDR_ANY &&
1178 		    rtable_l2(inp->inp_rtableid) == rdomain)
1179 			break;
1180 	}
1181 	if (inp == NULL && key1->s_addr != key2->s_addr) {
1182 		head = INPCBHASH(table, &zeroin_addr, 0, key2, lport, rdomain);
1183 		LIST_FOREACH(inp, head, inp_hash) {
1184 #ifdef INET6
1185 			if (inp->inp_flags & INP_IPV6)
1186 				continue;	/*XXX*/
1187 #endif
1188 			if (inp->inp_lport == lport && inp->inp_fport == 0 &&
1189 			    inp->inp_laddr.s_addr == key2->s_addr &&
1190 			    inp->inp_faddr.s_addr == INADDR_ANY &&
1191 			    rtable_l2(inp->inp_rtableid) == rdomain)
1192 				break;
1193 		}
1194 	}
1195 #ifdef DIAGNOSTIC
1196 	if (inp == NULL && in_pcbnotifymiss) {
1197 		printf("in_pcblookup_listen: laddr=%08x lport=%d\n",
1198 		    ntohl(laddr.s_addr), ntohs(lport));
1199 	}
1200 #endif
1201 	/*
1202 	 * Move this PCB to the head of hash chain so that
1203 	 * repeated accesses are quicker.  This is analogous to
1204 	 * the historic single-entry PCB cache.
1205 	 */
1206 	if (inp != NULL && inp != LIST_FIRST(head)) {
1207 		LIST_REMOVE(inp, inp_hash);
1208 		LIST_INSERT_HEAD(head, inp, inp_hash);
1209 	}
1210 	return (inp);
1211 }
1212 
1213 #ifdef INET6
1214 struct inpcb *
1215 in6_pcblookup_listen(struct inpcbtable *table, struct in6_addr *laddr,
1216     u_int lport_arg, struct mbuf *m, u_int rtable)
1217 {
1218 	struct inpcbhead *head;
1219 	struct in6_addr *key1, *key2;
1220 	struct inpcb *inp;
1221 	u_int16_t lport = lport_arg;
1222 
1223 	rtable = rtable_l2(rtable);	/* convert passed rtableid to rdomain */
1224 	key1 = laddr;
1225 	key2 = &zeroin6_addr;
1226 #if NPF > 0
1227 	if (m && m->m_pkthdr.pf.flags & PF_TAG_DIVERTED) {
1228 		struct pf_divert *divert;
1229 
1230 		divert = pf_find_divert(m);
1231 		KASSERT(divert != NULL);
1232 		switch (divert->type) {
1233 		case PF_DIVERT_TO:
1234 			key1 = key2 = &divert->addr.v6;
1235 			lport = divert->port;
1236 			break;
1237 		case PF_DIVERT_REPLY:
1238 			return (NULL);
1239 		default:
1240 			panic("%s: unknown divert type %d, mbuf %p, divert %p",
1241 			    __func__, divert->type, m, divert);
1242 		}
1243 	} else if (m && m->m_pkthdr.pf.flags & PF_TAG_TRANSLATE_LOCALHOST) {
1244 		key1 = &zeroin6_addr;
1245 		key2 = laddr;
1246 	}
1247 #endif
1248 
1249 	head = IN6PCBHASH(table, &zeroin6_addr, 0, key1, lport, rtable);
1250 	LIST_FOREACH(inp, head, inp_hash) {
1251 		if (!(inp->inp_flags & INP_IPV6))
1252 			continue;
1253 		if (inp->inp_lport == lport && inp->inp_fport == 0 &&
1254 		    IN6_ARE_ADDR_EQUAL(&inp->inp_laddr6, key1) &&
1255 		    IN6_IS_ADDR_UNSPECIFIED(&inp->inp_faddr6) &&
1256 		    rtable_l2(inp->inp_rtableid) == rtable)
1257 			break;
1258 	}
1259 	if (inp == NULL && ! IN6_ARE_ADDR_EQUAL(key1, key2)) {
1260 		head = IN6PCBHASH(table, &zeroin6_addr, 0, key2, lport, rtable);
1261 		LIST_FOREACH(inp, head, inp_hash) {
1262 			if (!(inp->inp_flags & INP_IPV6))
1263 				continue;
1264 			if (inp->inp_lport == lport && inp->inp_fport == 0 &&
1265 			    IN6_ARE_ADDR_EQUAL(&inp->inp_laddr6, key2) &&
1266 			    IN6_IS_ADDR_UNSPECIFIED(&inp->inp_faddr6) &&
1267 			    rtable_l2(inp->inp_rtableid) == rtable)
1268 				break;
1269 		}
1270 	}
1271 #ifdef DIAGNOSTIC
1272 	if (inp == NULL && in_pcbnotifymiss) {
1273 		printf("in6_pcblookup_listen: laddr= lport=%d\n",
1274 		    ntohs(lport));
1275 	}
1276 #endif
1277 	/*
1278 	 * Move this PCB to the head of hash chain so that
1279 	 * repeated accesses are quicker.  This is analogous to
1280 	 * the historic single-entry PCB cache.
1281 	 */
1282 	if (inp != NULL && inp != LIST_FIRST(head)) {
1283 		LIST_REMOVE(inp, inp_hash);
1284 		LIST_INSERT_HEAD(head, inp, inp_hash);
1285 	}
1286 	return (inp);
1287 }
1288 #endif /* INET6 */
1289