xref: /netbsd-src/sys/netinet/in_pcb.c (revision 8ac07aec990b9d2e483062509d0a9fa5b4f57cf2)
1 /*	$NetBSD: in_pcb.c,v 1.123 2008/04/24 11:38:37 ad Exp $	*/
2 
3 /*
4  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the project nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  */
31 
32 /*-
33  * Copyright (c) 1998 The NetBSD Foundation, Inc.
34  * All rights reserved.
35  *
36  * This code is derived from software contributed to The NetBSD Foundation
37  * by Public Access Networks Corporation ("Panix").  It was developed under
38  * contract to Panix by Eric Haszlakiewicz and Thor Lancelot Simon.
39  *
40  * Redistribution and use in source and binary forms, with or without
41  * modification, are permitted provided that the following conditions
42  * are met:
43  * 1. Redistributions of source code must retain the above copyright
44  *    notice, this list of conditions and the following disclaimer.
45  * 2. Redistributions in binary form must reproduce the above copyright
46  *    notice, this list of conditions and the following disclaimer in the
47  *    documentation and/or other materials provided with the distribution.
48  * 3. All advertising materials mentioning features or use of this software
49  *    must display the following acknowledgement:
50  *	This product includes software developed by the NetBSD
51  *	Foundation, Inc. and its contributors.
52  * 4. Neither the name of The NetBSD Foundation nor the names of its
53  *    contributors may be used to endorse or promote products derived
54  *    from this software without specific prior written permission.
55  *
56  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
57  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
58  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
59  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
60  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
61  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
62  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
63  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
64  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
65  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
66  * POSSIBILITY OF SUCH DAMAGE.
67  */
68 
69 /*
70  * Copyright (c) 1982, 1986, 1991, 1993, 1995
71  *	The Regents of the University of California.  All rights reserved.
72  *
73  * Redistribution and use in source and binary forms, with or without
74  * modification, are permitted provided that the following conditions
75  * are met:
76  * 1. Redistributions of source code must retain the above copyright
77  *    notice, this list of conditions and the following disclaimer.
78  * 2. Redistributions in binary form must reproduce the above copyright
79  *    notice, this list of conditions and the following disclaimer in the
80  *    documentation and/or other materials provided with the distribution.
81  * 3. Neither the name of the University nor the names of its contributors
82  *    may be used to endorse or promote products derived from this software
83  *    without specific prior written permission.
84  *
85  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
86  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
87  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
88  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
89  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
90  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
91  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
92  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
93  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
94  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
95  * SUCH DAMAGE.
96  *
97  *	@(#)in_pcb.c	8.4 (Berkeley) 5/24/95
98  */
99 
100 #include <sys/cdefs.h>
101 __KERNEL_RCSID(0, "$NetBSD: in_pcb.c,v 1.123 2008/04/24 11:38:37 ad Exp $");
102 
103 #include "opt_inet.h"
104 #include "opt_ipsec.h"
105 
106 #include <sys/param.h>
107 #include <sys/systm.h>
108 #include <sys/malloc.h>
109 #include <sys/mbuf.h>
110 #include <sys/protosw.h>
111 #include <sys/socket.h>
112 #include <sys/socketvar.h>
113 #include <sys/ioctl.h>
114 #include <sys/errno.h>
115 #include <sys/time.h>
116 #include <sys/pool.h>
117 #include <sys/proc.h>
118 #include <sys/kauth.h>
119 
120 #include <net/if.h>
121 #include <net/route.h>
122 
123 #include <netinet/in.h>
124 #include <netinet/in_systm.h>
125 #include <netinet/ip.h>
126 #include <netinet/in_pcb.h>
127 #include <netinet/in_var.h>
128 #include <netinet/ip_var.h>
129 
130 #ifdef INET6
131 #include <netinet/ip6.h>
132 #include <netinet6/ip6_var.h>
133 #include <netinet6/in6_pcb.h>
134 #endif
135 
136 #ifdef IPSEC
137 #include <netinet6/ipsec.h>
138 #include <netkey/key.h>
139 #elif FAST_IPSEC
140 #include <netipsec/ipsec.h>
141 #include <netipsec/key.h>
142 #endif /* IPSEC */
143 
144 struct	in_addr zeroin_addr;
145 
146 #define	INPCBHASH_PORT(table, lport) \
147 	&(table)->inpt_porthashtbl[ntohs(lport) & (table)->inpt_porthash]
148 #define	INPCBHASH_BIND(table, laddr, lport) \
149 	&(table)->inpt_bindhashtbl[ \
150 	    ((ntohl((laddr).s_addr) + ntohs(lport))) & (table)->inpt_bindhash]
151 #define	INPCBHASH_CONNECT(table, faddr, fport, laddr, lport) \
152 	&(table)->inpt_connecthashtbl[ \
153 	    ((ntohl((faddr).s_addr) + ntohs(fport)) + \
154 	     (ntohl((laddr).s_addr) + ntohs(lport))) & (table)->inpt_connecthash]
155 
156 int	anonportmin = IPPORT_ANONMIN;
157 int	anonportmax = IPPORT_ANONMAX;
158 int	lowportmin  = IPPORT_RESERVEDMIN;
159 int	lowportmax  = IPPORT_RESERVEDMAX;
160 
161 POOL_INIT(inpcb_pool, sizeof(struct inpcb), 0, 0, 0, "inpcbpl", NULL,
162     IPL_NET);
163 
164 void
165 in_pcbinit(struct inpcbtable *table, int bindhashsize, int connecthashsize)
166 {
167 
168 	CIRCLEQ_INIT(&table->inpt_queue);
169 	table->inpt_porthashtbl = hashinit(bindhashsize, HASH_LIST, M_PCB,
170 	    M_WAITOK, &table->inpt_porthash);
171 	table->inpt_bindhashtbl = hashinit(bindhashsize, HASH_LIST, M_PCB,
172 	    M_WAITOK, &table->inpt_bindhash);
173 	table->inpt_connecthashtbl = hashinit(connecthashsize, HASH_LIST,
174 	    M_PCB, M_WAITOK, &table->inpt_connecthash);
175 	table->inpt_lastlow = IPPORT_RESERVEDMAX;
176 	table->inpt_lastport = (u_int16_t)anonportmax;
177 }
178 
179 int
180 in_pcballoc(struct socket *so, void *v)
181 {
182 	struct inpcbtable *table = v;
183 	struct inpcb *inp;
184 	int s;
185 #if defined(IPSEC) || defined(FAST_IPSEC)
186 	int error;
187 #endif
188 
189 	s = splnet();
190 	inp = pool_get(&inpcb_pool, PR_NOWAIT);
191 	splx(s);
192 	if (inp == NULL)
193 		return (ENOBUFS);
194 	bzero((void *)inp, sizeof(*inp));
195 	inp->inp_af = AF_INET;
196 	inp->inp_table = table;
197 	inp->inp_socket = so;
198 	inp->inp_errormtu = -1;
199 #if defined(IPSEC) || defined(FAST_IPSEC)
200 	error = ipsec_init_pcbpolicy(so, &inp->inp_sp);
201 	if (error != 0) {
202 		s = splnet();
203 		pool_put(&inpcb_pool, inp);
204 		splx(s);
205 		return error;
206 	}
207 #endif
208 	so->so_pcb = inp;
209 	s = splnet();
210 	CIRCLEQ_INSERT_HEAD(&table->inpt_queue, &inp->inp_head,
211 	    inph_queue);
212 	LIST_INSERT_HEAD(INPCBHASH_PORT(table, inp->inp_lport), &inp->inp_head,
213 	    inph_lhash);
214 	in_pcbstate(inp, INP_ATTACHED);
215 	splx(s);
216 	return (0);
217 }
218 
219 int
220 in_pcbbind(void *v, struct mbuf *nam, struct lwp *l)
221 {
222 	struct in_ifaddr *ia = NULL;
223 	struct inpcb *inp = v;
224 	struct socket *so = inp->inp_socket;
225 	struct inpcbtable *table = inp->inp_table;
226 	struct sockaddr_in *sin = NULL; /* XXXGCC */
227 	u_int16_t lport = 0;
228 	int wild = 0, reuseport = (so->so_options & SO_REUSEPORT);
229 	kauth_cred_t cred = l->l_cred;
230 
231 	if (inp->inp_af != AF_INET)
232 		return (EINVAL);
233 
234 	if (TAILQ_FIRST(&in_ifaddrhead) == 0)
235 		return (EADDRNOTAVAIL);
236 	if (inp->inp_lport || !in_nullhost(inp->inp_laddr))
237 		return (EINVAL);
238 	if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0)
239 		wild = 1;
240 	if (nam == 0)
241 		goto noname;
242 	sin = mtod(nam, struct sockaddr_in *);
243 	if (nam->m_len != sizeof (*sin))
244 		return (EINVAL);
245 	if (sin->sin_family != AF_INET)
246 		return (EAFNOSUPPORT);
247 	lport = sin->sin_port;
248 	if (IN_MULTICAST(sin->sin_addr.s_addr)) {
249 		/*
250 		 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
251 		 * allow complete duplication of binding if
252 		 * SO_REUSEPORT is set, or if SO_REUSEADDR is set
253 		 * and a multicast address is bound on both
254 		 * new and duplicated sockets.
255 		 */
256 		if (so->so_options & SO_REUSEADDR)
257 			reuseport = SO_REUSEADDR|SO_REUSEPORT;
258 	} else if (!in_nullhost(sin->sin_addr)) {
259 		sin->sin_port = 0;		/* yech... */
260 		INADDR_TO_IA(sin->sin_addr, ia);
261 		/* check for broadcast addresses */
262 		if (ia == NULL)
263 			ia = ifatoia(ifa_ifwithaddr(sintosa(sin)));
264 		if (ia == NULL)
265 			return (EADDRNOTAVAIL);
266 	}
267 	if (lport) {
268 		struct inpcb *t;
269 #ifdef INET6
270 		struct in6pcb *t6;
271 		struct in6_addr mapped;
272 #endif
273 #ifndef IPNOPRIVPORTS
274 		/* GROSS */
275 		if (ntohs(lport) < IPPORT_RESERVED &&
276 		    kauth_authorize_network(cred,
277 		    KAUTH_NETWORK_BIND,
278 		    KAUTH_REQ_NETWORK_BIND_PRIVPORT, so, sin,
279 		    NULL))
280 			return (EACCES);
281 #endif
282 #ifdef INET6
283 		memset(&mapped, 0, sizeof(mapped));
284 		mapped.s6_addr16[5] = 0xffff;
285 		memcpy(&mapped.s6_addr32[3], &sin->sin_addr,
286 		    sizeof(mapped.s6_addr32[3]));
287 		t6 = in6_pcblookup_port(table, &mapped, lport, wild);
288 		if (t6 && (reuseport & t6->in6p_socket->so_options) == 0)
289 			return (EADDRINUSE);
290 #endif
291 		if (so->so_uidinfo->ui_uid && !IN_MULTICAST(sin->sin_addr.s_addr)) {
292 			t = in_pcblookup_port(table, sin->sin_addr, lport, 1);
293 		/*
294 		 * XXX:	investigate ramifications of loosening this
295 		 *	restriction so that as long as both ports have
296 		 *	SO_REUSEPORT allow the bind
297 		 */
298 			if (t &&
299 			    (!in_nullhost(sin->sin_addr) ||
300 			     !in_nullhost(t->inp_laddr) ||
301 			     (t->inp_socket->so_options & SO_REUSEPORT) == 0)
302 			    && (so->so_uidinfo->ui_uid != t->inp_socket->so_uidinfo->ui_uid)) {
303 				return (EADDRINUSE);
304 			}
305 		}
306 		t = in_pcblookup_port(table, sin->sin_addr, lport, wild);
307 		if (t && (reuseport & t->inp_socket->so_options) == 0)
308 			return (EADDRINUSE);
309 	}
310 	inp->inp_laddr = sin->sin_addr;
311 
312 noname:
313 	if (lport == 0) {
314 		int	   cnt;
315 		u_int16_t  mymin, mymax;
316 		u_int16_t *lastport;
317 
318 		if (inp->inp_flags & INP_LOWPORT) {
319 #ifndef IPNOPRIVPORTS
320 			if (kauth_authorize_network(cred,
321 			    KAUTH_NETWORK_BIND,
322 			    KAUTH_REQ_NETWORK_BIND_PRIVPORT, so,
323 			    sin, NULL))
324 				return (EACCES);
325 #endif
326 			mymin = lowportmin;
327 			mymax = lowportmax;
328 			lastport = &table->inpt_lastlow;
329 		} else {
330 			mymin = anonportmin;
331 			mymax = anonportmax;
332 			lastport = &table->inpt_lastport;
333 		}
334 		if (mymin > mymax) {	/* sanity check */
335 			u_int16_t swp;
336 
337 			swp = mymin;
338 			mymin = mymax;
339 			mymax = swp;
340 		}
341 
342 		lport = *lastport - 1;
343 		for (cnt = mymax - mymin + 1; cnt; cnt--, lport--) {
344 			if (lport < mymin || lport > mymax)
345 				lport = mymax;
346 			if (!in_pcblookup_port(table, inp->inp_laddr,
347 			    htons(lport), 1))
348 				goto found;
349 		}
350 		if (!in_nullhost(inp->inp_laddr))
351 			inp->inp_laddr.s_addr = INADDR_ANY;
352 		return (EAGAIN);
353 	found:
354 		inp->inp_flags |= INP_ANONPORT;
355 		*lastport = lport;
356 		lport = htons(lport);
357 	}
358 	inp->inp_lport = lport;
359 	LIST_REMOVE(&inp->inp_head, inph_lhash);
360 	LIST_INSERT_HEAD(INPCBHASH_PORT(table, inp->inp_lport), &inp->inp_head,
361 	    inph_lhash);
362 	in_pcbstate(inp, INP_BOUND);
363 	return (0);
364 }
365 
366 /*
367  * Connect from a socket to a specified address.
368  * Both address and port must be specified in argument sin.
369  * If don't have a local address for this socket yet,
370  * then pick one.
371  */
372 int
373 in_pcbconnect(void *v, struct mbuf *nam, struct lwp *l)
374 {
375 	struct inpcb *inp = v;
376 	struct in_ifaddr *ia = NULL;
377 	struct sockaddr_in *ifaddr = NULL;
378 	struct sockaddr_in *sin = mtod(nam, struct sockaddr_in *);
379 	int error;
380 
381 	if (inp->inp_af != AF_INET)
382 		return (EINVAL);
383 
384 	if (nam->m_len != sizeof (*sin))
385 		return (EINVAL);
386 	if (sin->sin_family != AF_INET)
387 		return (EAFNOSUPPORT);
388 	if (sin->sin_port == 0)
389 		return (EADDRNOTAVAIL);
390 	if (TAILQ_FIRST(&in_ifaddrhead) != 0) {
391 		/*
392 		 * If the destination address is INADDR_ANY,
393 		 * use any local address (likely loopback).
394 		 * If the supplied address is INADDR_BROADCAST,
395 		 * use the broadcast address of an interface
396 		 * which supports broadcast. (loopback does not)
397 		 */
398 
399 		if (in_nullhost(sin->sin_addr)) {
400 			sin->sin_addr =
401 			    TAILQ_FIRST(&in_ifaddrhead)->ia_addr.sin_addr;
402 		} else if (sin->sin_addr.s_addr == INADDR_BROADCAST) {
403 			TAILQ_FOREACH(ia, &in_ifaddrhead, ia_list) {
404 				if (ia->ia_ifp->if_flags & IFF_BROADCAST) {
405 					sin->sin_addr =
406 					    ia->ia_broadaddr.sin_addr;
407 					break;
408 				}
409 			}
410 		}
411 	}
412 	/*
413 	 * If we haven't bound which network number to use as ours,
414 	 * we will use the number of the outgoing interface.
415 	 * This depends on having done a routing lookup, which
416 	 * we will probably have to do anyway, so we might
417 	 * as well do it now.  On the other hand if we are
418 	 * sending to multiple destinations we may have already
419 	 * done the lookup, so see if we can use the route
420 	 * from before.  In any case, we only
421 	 * chose a port number once, even if sending to multiple
422 	 * destinations.
423 	 */
424 	if (in_nullhost(inp->inp_laddr)) {
425 		int xerror;
426 		ifaddr = in_selectsrc(sin, &inp->inp_route,
427 		    inp->inp_socket->so_options, inp->inp_moptions, &xerror);
428 		if (ifaddr == NULL) {
429 			if (xerror == 0)
430 				xerror = EADDRNOTAVAIL;
431 			return xerror;
432 		}
433 		INADDR_TO_IA(ifaddr->sin_addr, ia);
434 		if (ia == NULL)
435 			return (EADDRNOTAVAIL);
436 	}
437 	if (in_pcblookup_connect(inp->inp_table, sin->sin_addr, sin->sin_port,
438 	    !in_nullhost(inp->inp_laddr) ? inp->inp_laddr : ifaddr->sin_addr,
439 	    inp->inp_lport) != 0)
440 		return (EADDRINUSE);
441 	if (in_nullhost(inp->inp_laddr)) {
442 		if (inp->inp_lport == 0) {
443 			error = in_pcbbind(inp, NULL, l);
444 			/*
445 			 * This used to ignore the return value
446 			 * completely, but we need to check for
447 			 * ephemeral port shortage.
448 			 * And attempts to request low ports if not root.
449 			 */
450 			if (error != 0)
451 				return (error);
452 		}
453 		inp->inp_laddr = ifaddr->sin_addr;
454 	}
455 	inp->inp_faddr = sin->sin_addr;
456 	inp->inp_fport = sin->sin_port;
457 	in_pcbstate(inp, INP_CONNECTED);
458 #if defined(IPSEC) || defined(FAST_IPSEC)
459 	if (inp->inp_socket->so_type == SOCK_STREAM)
460 		ipsec_pcbconn(inp->inp_sp);
461 #endif
462 	return (0);
463 }
464 
465 void
466 in_pcbdisconnect(void *v)
467 {
468 	struct inpcb *inp = v;
469 
470 	if (inp->inp_af != AF_INET)
471 		return;
472 
473 	inp->inp_faddr = zeroin_addr;
474 	inp->inp_fport = 0;
475 	in_pcbstate(inp, INP_BOUND);
476 #if defined(IPSEC) || defined(FAST_IPSEC)
477 	ipsec_pcbdisconn(inp->inp_sp);
478 #endif
479 	if (inp->inp_socket->so_state & SS_NOFDREF)
480 		in_pcbdetach(inp);
481 }
482 
483 void
484 in_pcbdetach(void *v)
485 {
486 	struct inpcb *inp = v;
487 	struct socket *so = inp->inp_socket;
488 	int s;
489 
490 	if (inp->inp_af != AF_INET)
491 		return;
492 
493 #if defined(IPSEC) || defined(FAST_IPSEC)
494 	ipsec4_delete_pcbpolicy(inp);
495 #endif /*IPSEC*/
496 	so->so_pcb = 0;
497 	/* sofree drop's the socket's lock */
498 	sofree(so);
499 	mutex_enter(softnet_lock);
500 	if (inp->inp_options)
501 		(void)m_free(inp->inp_options);
502 	rtcache_free(&inp->inp_route);
503 	ip_freemoptions(inp->inp_moptions);
504 	s = splnet();
505 	in_pcbstate(inp, INP_ATTACHED);
506 	LIST_REMOVE(&inp->inp_head, inph_lhash);
507 	CIRCLEQ_REMOVE(&inp->inp_table->inpt_queue, &inp->inp_head,
508 	    inph_queue);
509 	pool_put(&inpcb_pool, inp);
510 	splx(s);
511 }
512 
513 void
514 in_setsockaddr(struct inpcb *inp, struct mbuf *nam)
515 {
516 	struct sockaddr_in *sin;
517 
518 	if (inp->inp_af != AF_INET)
519 		return;
520 
521 	sin = mtod(nam, struct sockaddr_in *);
522 	sockaddr_in_init(sin, &inp->inp_laddr, inp->inp_lport);
523 	nam->m_len = sin->sin_len;
524 }
525 
526 void
527 in_setpeeraddr(struct inpcb *inp, struct mbuf *nam)
528 {
529 	struct sockaddr_in *sin;
530 
531 	if (inp->inp_af != AF_INET)
532 		return;
533 
534 	sin = mtod(nam, struct sockaddr_in *);
535 	sockaddr_in_init(sin, &inp->inp_faddr, inp->inp_fport);
536 	nam->m_len = sin->sin_len;
537 }
538 
539 /*
540  * Pass some notification to all connections of a protocol
541  * associated with address dst.  The local address and/or port numbers
542  * may be specified to limit the search.  The "usual action" will be
543  * taken, depending on the ctlinput cmd.  The caller must filter any
544  * cmds that are uninteresting (e.g., no error in the map).
545  * Call the protocol specific routine (if any) to report
546  * any errors for each matching socket.
547  *
548  * Must be called at splsoftnet.
549  */
550 int
551 in_pcbnotify(struct inpcbtable *table, struct in_addr faddr, u_int fport_arg,
552     struct in_addr laddr, u_int lport_arg, int errno,
553     void (*notify)(struct inpcb *, int))
554 {
555 	struct inpcbhead *head;
556 	struct inpcb *inp, *ninp;
557 	u_int16_t fport = fport_arg, lport = lport_arg;
558 	int nmatch;
559 
560 	if (in_nullhost(faddr) || notify == 0)
561 		return (0);
562 
563 	nmatch = 0;
564 	head = INPCBHASH_CONNECT(table, faddr, fport, laddr, lport);
565 	for (inp = (struct inpcb *)LIST_FIRST(head); inp != NULL; inp = ninp) {
566 		ninp = (struct inpcb *)LIST_NEXT(inp, inp_hash);
567 		if (inp->inp_af != AF_INET)
568 			continue;
569 		if (in_hosteq(inp->inp_faddr, faddr) &&
570 		    inp->inp_fport == fport &&
571 		    inp->inp_lport == lport &&
572 		    in_hosteq(inp->inp_laddr, laddr)) {
573 			(*notify)(inp, errno);
574 			nmatch++;
575 		}
576 	}
577 	return (nmatch);
578 }
579 
580 void
581 in_pcbnotifyall(struct inpcbtable *table, struct in_addr faddr, int errno,
582     void (*notify)(struct inpcb *, int))
583 {
584 	struct inpcb *inp, *ninp;
585 
586 	if (in_nullhost(faddr) || notify == 0)
587 		return;
588 
589 	for (inp = (struct inpcb *)CIRCLEQ_FIRST(&table->inpt_queue);
590 	    inp != (void *)&table->inpt_queue;
591 	    inp = ninp) {
592 		ninp = (struct inpcb *)CIRCLEQ_NEXT(inp, inp_queue);
593 		if (inp->inp_af != AF_INET)
594 			continue;
595 		if (in_hosteq(inp->inp_faddr, faddr))
596 			(*notify)(inp, errno);
597 	}
598 }
599 
600 void
601 in_pcbpurgeif0(struct inpcbtable *table, struct ifnet *ifp)
602 {
603 	struct inpcb *inp, *ninp;
604 	struct ip_moptions *imo;
605 	int i, gap;
606 
607 	for (inp = (struct inpcb *)CIRCLEQ_FIRST(&table->inpt_queue);
608 	    inp != (void *)&table->inpt_queue;
609 	    inp = ninp) {
610 		ninp = (struct inpcb *)CIRCLEQ_NEXT(inp, inp_queue);
611 		if (inp->inp_af != AF_INET)
612 			continue;
613 		imo = inp->inp_moptions;
614 		if (imo != NULL) {
615 			/*
616 			 * Unselect the outgoing interface if it is being
617 			 * detached.
618 			 */
619 			if (imo->imo_multicast_ifp == ifp)
620 				imo->imo_multicast_ifp = NULL;
621 
622 			/*
623 			 * Drop multicast group membership if we joined
624 			 * through the interface being detached.
625 			 */
626 			for (i = 0, gap = 0; i < imo->imo_num_memberships;
627 			    i++) {
628 				if (imo->imo_membership[i]->inm_ifp == ifp) {
629 					in_delmulti(imo->imo_membership[i]);
630 					gap++;
631 				} else if (gap != 0)
632 					imo->imo_membership[i - gap] =
633 					    imo->imo_membership[i];
634 			}
635 			imo->imo_num_memberships -= gap;
636 		}
637 	}
638 }
639 
640 void
641 in_pcbpurgeif(struct inpcbtable *table, struct ifnet *ifp)
642 {
643 	struct rtentry *rt;
644 	struct inpcb *inp, *ninp;
645 
646 	for (inp = (struct inpcb *)CIRCLEQ_FIRST(&table->inpt_queue);
647 	    inp != (void *)&table->inpt_queue;
648 	    inp = ninp) {
649 		ninp = (struct inpcb *)CIRCLEQ_NEXT(inp, inp_queue);
650 		if (inp->inp_af != AF_INET)
651 			continue;
652 		if ((rt = rtcache_validate(&inp->inp_route)) != NULL &&
653 		    rt->rt_ifp == ifp)
654 			in_rtchange(inp, 0);
655 	}
656 }
657 
658 /*
659  * Check for alternatives when higher level complains
660  * about service problems.  For now, invalidate cached
661  * routing information.  If the route was created dynamically
662  * (by a redirect), time to try a default gateway again.
663  */
664 void
665 in_losing(struct inpcb *inp)
666 {
667 	struct rtentry *rt;
668 	struct rt_addrinfo info;
669 
670 	if (inp->inp_af != AF_INET)
671 		return;
672 
673 	if ((rt = rtcache_validate(&inp->inp_route)) == NULL)
674 		return;
675 
676 	memset(&info, 0, sizeof(info));
677 	info.rti_info[RTAX_DST] = rtcache_getdst(&inp->inp_route);
678 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
679 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
680 	rt_missmsg(RTM_LOSING, &info, rt->rt_flags, 0);
681 	if (rt->rt_flags & RTF_DYNAMIC)
682 		(void) rtrequest(RTM_DELETE, rt_getkey(rt),
683 			rt->rt_gateway, rt_mask(rt), rt->rt_flags,
684 			NULL);
685 	/*
686 	 * A new route can be allocated
687 	 * the next time output is attempted.
688 	 */
689 	rtcache_free(&inp->inp_route);
690 }
691 
692 /*
693  * After a routing change, flush old routing.  A new route can be
694  * allocated the next time output is attempted.
695  */
696 void
697 in_rtchange(struct inpcb *inp, int errno)
698 {
699 
700 	if (inp->inp_af != AF_INET)
701 		return;
702 
703 	rtcache_free(&inp->inp_route);
704 
705 	/* XXX SHOULD NOTIFY HIGHER-LEVEL PROTOCOLS */
706 }
707 
708 struct inpcb *
709 in_pcblookup_port(struct inpcbtable *table, struct in_addr laddr,
710     u_int lport_arg, int lookup_wildcard)
711 {
712 	struct inpcbhead *head;
713 	struct inpcb_hdr *inph;
714 	struct inpcb *inp, *match = 0;
715 	int matchwild = 3, wildcard;
716 	u_int16_t lport = lport_arg;
717 
718 	head = INPCBHASH_PORT(table, lport);
719 	LIST_FOREACH(inph, head, inph_lhash) {
720 		inp = (struct inpcb *)inph;
721 		if (inp->inp_af != AF_INET)
722 			continue;
723 
724 		if (inp->inp_lport != lport)
725 			continue;
726 		wildcard = 0;
727 		if (!in_nullhost(inp->inp_faddr))
728 			wildcard++;
729 		if (in_nullhost(inp->inp_laddr)) {
730 			if (!in_nullhost(laddr))
731 				wildcard++;
732 		} else {
733 			if (in_nullhost(laddr))
734 				wildcard++;
735 			else {
736 				if (!in_hosteq(inp->inp_laddr, laddr))
737 					continue;
738 			}
739 		}
740 		if (wildcard && !lookup_wildcard)
741 			continue;
742 		if (wildcard < matchwild) {
743 			match = inp;
744 			matchwild = wildcard;
745 			if (matchwild == 0)
746 				break;
747 		}
748 	}
749 	return (match);
750 }
751 
752 #ifdef DIAGNOSTIC
753 int	in_pcbnotifymiss = 0;
754 #endif
755 
756 struct inpcb *
757 in_pcblookup_connect(struct inpcbtable *table,
758     struct in_addr faddr, u_int fport_arg,
759     struct in_addr laddr, u_int lport_arg)
760 {
761 	struct inpcbhead *head;
762 	struct inpcb_hdr *inph;
763 	struct inpcb *inp;
764 	u_int16_t fport = fport_arg, lport = lport_arg;
765 
766 	head = INPCBHASH_CONNECT(table, faddr, fport, laddr, lport);
767 	LIST_FOREACH(inph, head, inph_hash) {
768 		inp = (struct inpcb *)inph;
769 		if (inp->inp_af != AF_INET)
770 			continue;
771 
772 		if (in_hosteq(inp->inp_faddr, faddr) &&
773 		    inp->inp_fport == fport &&
774 		    inp->inp_lport == lport &&
775 		    in_hosteq(inp->inp_laddr, laddr))
776 			goto out;
777 	}
778 #ifdef DIAGNOSTIC
779 	if (in_pcbnotifymiss) {
780 		printf("in_pcblookup_connect: faddr=%08x fport=%d laddr=%08x lport=%d\n",
781 		    ntohl(faddr.s_addr), ntohs(fport),
782 		    ntohl(laddr.s_addr), ntohs(lport));
783 	}
784 #endif
785 	return (0);
786 
787 out:
788 	/* Move this PCB to the head of hash chain. */
789 	inph = &inp->inp_head;
790 	if (inph != LIST_FIRST(head)) {
791 		LIST_REMOVE(inph, inph_hash);
792 		LIST_INSERT_HEAD(head, inph, inph_hash);
793 	}
794 	return (inp);
795 }
796 
797 struct inpcb *
798 in_pcblookup_bind(struct inpcbtable *table,
799     struct in_addr laddr, u_int lport_arg)
800 {
801 	struct inpcbhead *head;
802 	struct inpcb_hdr *inph;
803 	struct inpcb *inp;
804 	u_int16_t lport = lport_arg;
805 
806 	head = INPCBHASH_BIND(table, laddr, lport);
807 	LIST_FOREACH(inph, head, inph_hash) {
808 		inp = (struct inpcb *)inph;
809 		if (inp->inp_af != AF_INET)
810 			continue;
811 
812 		if (inp->inp_lport == lport &&
813 		    in_hosteq(inp->inp_laddr, laddr))
814 			goto out;
815 	}
816 	head = INPCBHASH_BIND(table, zeroin_addr, lport);
817 	LIST_FOREACH(inph, head, inph_hash) {
818 		inp = (struct inpcb *)inph;
819 		if (inp->inp_af != AF_INET)
820 			continue;
821 
822 		if (inp->inp_lport == lport &&
823 		    in_hosteq(inp->inp_laddr, zeroin_addr))
824 			goto out;
825 	}
826 #ifdef DIAGNOSTIC
827 	if (in_pcbnotifymiss) {
828 		printf("in_pcblookup_bind: laddr=%08x lport=%d\n",
829 		    ntohl(laddr.s_addr), ntohs(lport));
830 	}
831 #endif
832 	return (0);
833 
834 out:
835 	/* Move this PCB to the head of hash chain. */
836 	inph = &inp->inp_head;
837 	if (inph != LIST_FIRST(head)) {
838 		LIST_REMOVE(inph, inph_hash);
839 		LIST_INSERT_HEAD(head, inph, inph_hash);
840 	}
841 	return (inp);
842 }
843 
844 void
845 in_pcbstate(struct inpcb *inp, int state)
846 {
847 
848 	if (inp->inp_af != AF_INET)
849 		return;
850 
851 	if (inp->inp_state > INP_ATTACHED)
852 		LIST_REMOVE(&inp->inp_head, inph_hash);
853 
854 	switch (state) {
855 	case INP_BOUND:
856 		LIST_INSERT_HEAD(INPCBHASH_BIND(inp->inp_table,
857 		    inp->inp_laddr, inp->inp_lport), &inp->inp_head,
858 		    inph_hash);
859 		break;
860 	case INP_CONNECTED:
861 		LIST_INSERT_HEAD(INPCBHASH_CONNECT(inp->inp_table,
862 		    inp->inp_faddr, inp->inp_fport,
863 		    inp->inp_laddr, inp->inp_lport), &inp->inp_head,
864 		    inph_hash);
865 		break;
866 	}
867 
868 	inp->inp_state = state;
869 }
870 
871 struct rtentry *
872 in_pcbrtentry(struct inpcb *inp)
873 {
874 	struct route *ro;
875 	union {
876 		struct sockaddr		dst;
877 		struct sockaddr_in	dst4;
878 	} u;
879 
880 	if (inp->inp_af != AF_INET)
881 		return (NULL);
882 
883 	ro = &inp->inp_route;
884 
885 	sockaddr_in_init(&u.dst4, &inp->inp_faddr, 0);
886 	return rtcache_lookup(ro, &u.dst);
887 }
888 
889 struct sockaddr_in *
890 in_selectsrc(struct sockaddr_in *sin, struct route *ro,
891     int soopts, struct ip_moptions *mopts, int *errorp)
892 {
893 	struct rtentry *rt = NULL;
894 	struct in_ifaddr *ia = NULL;
895 
896 	/*
897          * If route is known or can be allocated now, take the
898          * source address from the interface.  Otherwise, punt.
899 	 */
900 	if ((soopts & SO_DONTROUTE) != 0)
901 		rtcache_free(ro);
902 	else {
903 		union {
904 			struct sockaddr		dst;
905 			struct sockaddr_in	dst4;
906 		} u;
907 
908 		sockaddr_in_init(&u.dst4, &sin->sin_addr, 0);
909 		rt = rtcache_lookup(ro, &u.dst);
910 	}
911 	/*
912 	 * If we found a route, use the address
913 	 * corresponding to the outgoing interface
914 	 * unless it is the loopback (in case a route
915 	 * to our address on another net goes to loopback).
916 	 *
917 	 * XXX Is this still true?  Do we care?
918 	 */
919 	if (rt != NULL && (rt->rt_ifp->if_flags & IFF_LOOPBACK) == 0)
920 		ia = ifatoia(rt->rt_ifa);
921 	if (ia == NULL) {
922 		u_int16_t fport = sin->sin_port;
923 
924 		sin->sin_port = 0;
925 		ia = ifatoia(ifa_ifwithladdr(sintosa(sin)));
926 		sin->sin_port = fport;
927 		if (ia == NULL) {
928 			/* Find 1st non-loopback AF_INET address */
929 			TAILQ_FOREACH(ia, &in_ifaddrhead, ia_list) {
930 				if (!(ia->ia_ifp->if_flags & IFF_LOOPBACK))
931 					break;
932 			}
933 		}
934 		if (ia == NULL) {
935 			*errorp = EADDRNOTAVAIL;
936 			return NULL;
937 		}
938 	}
939 	/*
940 	 * If the destination address is multicast and an outgoing
941 	 * interface has been set as a multicast option, use the
942 	 * address of that interface as our source address.
943 	 */
944 	if (IN_MULTICAST(sin->sin_addr.s_addr) && mopts != NULL) {
945 		struct ip_moptions *imo;
946 		struct ifnet *ifp;
947 
948 		imo = mopts;
949 		if (imo->imo_multicast_ifp != NULL) {
950 			ifp = imo->imo_multicast_ifp;
951 			IFP_TO_IA(ifp, ia);		/* XXX */
952 			if (ia == 0) {
953 				*errorp = EADDRNOTAVAIL;
954 				return NULL;
955 			}
956 		}
957 	}
958 	if (ia->ia_ifa.ifa_getifa != NULL) {
959 		ia = ifatoia((*ia->ia_ifa.ifa_getifa)(&ia->ia_ifa,
960 		                                      sintosa(sin)));
961 	}
962 #ifdef GETIFA_DEBUG
963 	else
964 		printf("%s: missing ifa_getifa\n", __func__);
965 #endif
966 	return satosin(&ia->ia_addr);
967 }
968