xref: /netbsd-src/sys/netinet/in_pcb.c (revision 4b71a66d0f279143147d63ebfcfd8a59499a3684)
1 /*	$NetBSD: in_pcb.c,v 1.125 2008/05/05 17:11:17 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  *
49  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
50  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
51  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
52  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
53  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
54  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
55  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
56  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
57  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
58  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
59  * POSSIBILITY OF SUCH DAMAGE.
60  */
61 
62 /*
63  * Copyright (c) 1982, 1986, 1991, 1993, 1995
64  *	The Regents of the University of California.  All rights reserved.
65  *
66  * Redistribution and use in source and binary forms, with or without
67  * modification, are permitted provided that the following conditions
68  * are met:
69  * 1. Redistributions of source code must retain the above copyright
70  *    notice, this list of conditions and the following disclaimer.
71  * 2. Redistributions in binary form must reproduce the above copyright
72  *    notice, this list of conditions and the following disclaimer in the
73  *    documentation and/or other materials provided with the distribution.
74  * 3. Neither the name of the University nor the names of its contributors
75  *    may be used to endorse or promote products derived from this software
76  *    without specific prior written permission.
77  *
78  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
79  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
80  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
81  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
82  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
83  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
84  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
85  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
86  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
87  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
88  * SUCH DAMAGE.
89  *
90  *	@(#)in_pcb.c	8.4 (Berkeley) 5/24/95
91  */
92 
93 #include <sys/cdefs.h>
94 __KERNEL_RCSID(0, "$NetBSD: in_pcb.c,v 1.125 2008/05/05 17:11:17 ad Exp $");
95 
96 #include "opt_inet.h"
97 #include "opt_ipsec.h"
98 
99 #include <sys/param.h>
100 #include <sys/systm.h>
101 #include <sys/malloc.h>
102 #include <sys/mbuf.h>
103 #include <sys/protosw.h>
104 #include <sys/socket.h>
105 #include <sys/socketvar.h>
106 #include <sys/ioctl.h>
107 #include <sys/errno.h>
108 #include <sys/time.h>
109 #include <sys/pool.h>
110 #include <sys/proc.h>
111 #include <sys/kauth.h>
112 
113 #include <net/if.h>
114 #include <net/route.h>
115 
116 #include <netinet/in.h>
117 #include <netinet/in_systm.h>
118 #include <netinet/ip.h>
119 #include <netinet/in_pcb.h>
120 #include <netinet/in_var.h>
121 #include <netinet/ip_var.h>
122 
123 #ifdef INET6
124 #include <netinet/ip6.h>
125 #include <netinet6/ip6_var.h>
126 #include <netinet6/in6_pcb.h>
127 #endif
128 
129 #ifdef IPSEC
130 #include <netinet6/ipsec.h>
131 #include <netkey/key.h>
132 #elif FAST_IPSEC
133 #include <netipsec/ipsec.h>
134 #include <netipsec/key.h>
135 #endif /* IPSEC */
136 
137 struct	in_addr zeroin_addr;
138 
139 #define	INPCBHASH_PORT(table, lport) \
140 	&(table)->inpt_porthashtbl[ntohs(lport) & (table)->inpt_porthash]
141 #define	INPCBHASH_BIND(table, laddr, lport) \
142 	&(table)->inpt_bindhashtbl[ \
143 	    ((ntohl((laddr).s_addr) + ntohs(lport))) & (table)->inpt_bindhash]
144 #define	INPCBHASH_CONNECT(table, faddr, fport, laddr, lport) \
145 	&(table)->inpt_connecthashtbl[ \
146 	    ((ntohl((faddr).s_addr) + ntohs(fport)) + \
147 	     (ntohl((laddr).s_addr) + ntohs(lport))) & (table)->inpt_connecthash]
148 
149 int	anonportmin = IPPORT_ANONMIN;
150 int	anonportmax = IPPORT_ANONMAX;
151 int	lowportmin  = IPPORT_RESERVEDMIN;
152 int	lowportmax  = IPPORT_RESERVEDMAX;
153 
154 POOL_INIT(inpcb_pool, sizeof(struct inpcb), 0, 0, 0, "inpcbpl", NULL,
155     IPL_NET);
156 
157 void
158 in_pcbinit(struct inpcbtable *table, int bindhashsize, int connecthashsize)
159 {
160 
161 	CIRCLEQ_INIT(&table->inpt_queue);
162 	table->inpt_porthashtbl = hashinit(bindhashsize, HASH_LIST, true,
163 	    &table->inpt_porthash);
164 	table->inpt_bindhashtbl = hashinit(bindhashsize, HASH_LIST, true,
165 	    &table->inpt_bindhash);
166 	table->inpt_connecthashtbl = hashinit(connecthashsize, HASH_LIST, true,
167 	    &table->inpt_connecthash);
168 	table->inpt_lastlow = IPPORT_RESERVEDMAX;
169 	table->inpt_lastport = (u_int16_t)anonportmax;
170 }
171 
172 int
173 in_pcballoc(struct socket *so, void *v)
174 {
175 	struct inpcbtable *table = v;
176 	struct inpcb *inp;
177 	int s;
178 #if defined(IPSEC) || defined(FAST_IPSEC)
179 	int error;
180 #endif
181 
182 	s = splnet();
183 	inp = pool_get(&inpcb_pool, PR_NOWAIT);
184 	splx(s);
185 	if (inp == NULL)
186 		return (ENOBUFS);
187 	bzero((void *)inp, sizeof(*inp));
188 	inp->inp_af = AF_INET;
189 	inp->inp_table = table;
190 	inp->inp_socket = so;
191 	inp->inp_errormtu = -1;
192 #if defined(IPSEC) || defined(FAST_IPSEC)
193 	error = ipsec_init_pcbpolicy(so, &inp->inp_sp);
194 	if (error != 0) {
195 		s = splnet();
196 		pool_put(&inpcb_pool, inp);
197 		splx(s);
198 		return error;
199 	}
200 #endif
201 	so->so_pcb = inp;
202 	s = splnet();
203 	CIRCLEQ_INSERT_HEAD(&table->inpt_queue, &inp->inp_head,
204 	    inph_queue);
205 	LIST_INSERT_HEAD(INPCBHASH_PORT(table, inp->inp_lport), &inp->inp_head,
206 	    inph_lhash);
207 	in_pcbstate(inp, INP_ATTACHED);
208 	splx(s);
209 	return (0);
210 }
211 
212 int
213 in_pcbbind(void *v, struct mbuf *nam, struct lwp *l)
214 {
215 	struct in_ifaddr *ia = NULL;
216 	struct inpcb *inp = v;
217 	struct socket *so = inp->inp_socket;
218 	struct inpcbtable *table = inp->inp_table;
219 	struct sockaddr_in *sin = NULL; /* XXXGCC */
220 	u_int16_t lport = 0;
221 	int wild = 0, reuseport = (so->so_options & SO_REUSEPORT);
222 	kauth_cred_t cred = l->l_cred;
223 
224 	if (inp->inp_af != AF_INET)
225 		return (EINVAL);
226 
227 	if (TAILQ_FIRST(&in_ifaddrhead) == 0)
228 		return (EADDRNOTAVAIL);
229 	if (inp->inp_lport || !in_nullhost(inp->inp_laddr))
230 		return (EINVAL);
231 	if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0)
232 		wild = 1;
233 	if (nam == 0)
234 		goto noname;
235 	sin = mtod(nam, struct sockaddr_in *);
236 	if (nam->m_len != sizeof (*sin))
237 		return (EINVAL);
238 	if (sin->sin_family != AF_INET)
239 		return (EAFNOSUPPORT);
240 	lport = sin->sin_port;
241 	if (IN_MULTICAST(sin->sin_addr.s_addr)) {
242 		/*
243 		 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
244 		 * allow complete duplication of binding if
245 		 * SO_REUSEPORT is set, or if SO_REUSEADDR is set
246 		 * and a multicast address is bound on both
247 		 * new and duplicated sockets.
248 		 */
249 		if (so->so_options & SO_REUSEADDR)
250 			reuseport = SO_REUSEADDR|SO_REUSEPORT;
251 	} else if (!in_nullhost(sin->sin_addr)) {
252 		sin->sin_port = 0;		/* yech... */
253 		INADDR_TO_IA(sin->sin_addr, ia);
254 		/* check for broadcast addresses */
255 		if (ia == NULL)
256 			ia = ifatoia(ifa_ifwithaddr(sintosa(sin)));
257 		if (ia == NULL)
258 			return (EADDRNOTAVAIL);
259 	}
260 	if (lport) {
261 		struct inpcb *t;
262 #ifdef INET6
263 		struct in6pcb *t6;
264 		struct in6_addr mapped;
265 #endif
266 #ifndef IPNOPRIVPORTS
267 		/* GROSS */
268 		if (ntohs(lport) < IPPORT_RESERVED &&
269 		    kauth_authorize_network(cred,
270 		    KAUTH_NETWORK_BIND,
271 		    KAUTH_REQ_NETWORK_BIND_PRIVPORT, so, sin,
272 		    NULL))
273 			return (EACCES);
274 #endif
275 #ifdef INET6
276 		memset(&mapped, 0, sizeof(mapped));
277 		mapped.s6_addr16[5] = 0xffff;
278 		memcpy(&mapped.s6_addr32[3], &sin->sin_addr,
279 		    sizeof(mapped.s6_addr32[3]));
280 		t6 = in6_pcblookup_port(table, &mapped, lport, wild);
281 		if (t6 && (reuseport & t6->in6p_socket->so_options) == 0)
282 			return (EADDRINUSE);
283 #endif
284 		if (so->so_uidinfo->ui_uid && !IN_MULTICAST(sin->sin_addr.s_addr)) {
285 			t = in_pcblookup_port(table, sin->sin_addr, lport, 1);
286 		/*
287 		 * XXX:	investigate ramifications of loosening this
288 		 *	restriction so that as long as both ports have
289 		 *	SO_REUSEPORT allow the bind
290 		 */
291 			if (t &&
292 			    (!in_nullhost(sin->sin_addr) ||
293 			     !in_nullhost(t->inp_laddr) ||
294 			     (t->inp_socket->so_options & SO_REUSEPORT) == 0)
295 			    && (so->so_uidinfo->ui_uid != t->inp_socket->so_uidinfo->ui_uid)) {
296 				return (EADDRINUSE);
297 			}
298 		}
299 		t = in_pcblookup_port(table, sin->sin_addr, lport, wild);
300 		if (t && (reuseport & t->inp_socket->so_options) == 0)
301 			return (EADDRINUSE);
302 	}
303 	inp->inp_laddr = sin->sin_addr;
304 
305 noname:
306 	if (lport == 0) {
307 		int	   cnt;
308 		u_int16_t  mymin, mymax;
309 		u_int16_t *lastport;
310 
311 		if (inp->inp_flags & INP_LOWPORT) {
312 #ifndef IPNOPRIVPORTS
313 			if (kauth_authorize_network(cred,
314 			    KAUTH_NETWORK_BIND,
315 			    KAUTH_REQ_NETWORK_BIND_PRIVPORT, so,
316 			    sin, NULL))
317 				return (EACCES);
318 #endif
319 			mymin = lowportmin;
320 			mymax = lowportmax;
321 			lastport = &table->inpt_lastlow;
322 		} else {
323 			mymin = anonportmin;
324 			mymax = anonportmax;
325 			lastport = &table->inpt_lastport;
326 		}
327 		if (mymin > mymax) {	/* sanity check */
328 			u_int16_t swp;
329 
330 			swp = mymin;
331 			mymin = mymax;
332 			mymax = swp;
333 		}
334 
335 		lport = *lastport - 1;
336 		for (cnt = mymax - mymin + 1; cnt; cnt--, lport--) {
337 			if (lport < mymin || lport > mymax)
338 				lport = mymax;
339 			if (!in_pcblookup_port(table, inp->inp_laddr,
340 			    htons(lport), 1))
341 				goto found;
342 		}
343 		if (!in_nullhost(inp->inp_laddr))
344 			inp->inp_laddr.s_addr = INADDR_ANY;
345 		return (EAGAIN);
346 	found:
347 		inp->inp_flags |= INP_ANONPORT;
348 		*lastport = lport;
349 		lport = htons(lport);
350 	}
351 	inp->inp_lport = lport;
352 	LIST_REMOVE(&inp->inp_head, inph_lhash);
353 	LIST_INSERT_HEAD(INPCBHASH_PORT(table, inp->inp_lport), &inp->inp_head,
354 	    inph_lhash);
355 	in_pcbstate(inp, INP_BOUND);
356 	return (0);
357 }
358 
359 /*
360  * Connect from a socket to a specified address.
361  * Both address and port must be specified in argument sin.
362  * If don't have a local address for this socket yet,
363  * then pick one.
364  */
365 int
366 in_pcbconnect(void *v, struct mbuf *nam, struct lwp *l)
367 {
368 	struct inpcb *inp = v;
369 	struct in_ifaddr *ia = NULL;
370 	struct sockaddr_in *ifaddr = NULL;
371 	struct sockaddr_in *sin = mtod(nam, struct sockaddr_in *);
372 	int error;
373 
374 	if (inp->inp_af != AF_INET)
375 		return (EINVAL);
376 
377 	if (nam->m_len != sizeof (*sin))
378 		return (EINVAL);
379 	if (sin->sin_family != AF_INET)
380 		return (EAFNOSUPPORT);
381 	if (sin->sin_port == 0)
382 		return (EADDRNOTAVAIL);
383 	if (TAILQ_FIRST(&in_ifaddrhead) != 0) {
384 		/*
385 		 * If the destination address is INADDR_ANY,
386 		 * use any local address (likely loopback).
387 		 * If the supplied address is INADDR_BROADCAST,
388 		 * use the broadcast address of an interface
389 		 * which supports broadcast. (loopback does not)
390 		 */
391 
392 		if (in_nullhost(sin->sin_addr)) {
393 			sin->sin_addr =
394 			    TAILQ_FIRST(&in_ifaddrhead)->ia_addr.sin_addr;
395 		} else if (sin->sin_addr.s_addr == INADDR_BROADCAST) {
396 			TAILQ_FOREACH(ia, &in_ifaddrhead, ia_list) {
397 				if (ia->ia_ifp->if_flags & IFF_BROADCAST) {
398 					sin->sin_addr =
399 					    ia->ia_broadaddr.sin_addr;
400 					break;
401 				}
402 			}
403 		}
404 	}
405 	/*
406 	 * If we haven't bound which network number to use as ours,
407 	 * we will use the number of the outgoing interface.
408 	 * This depends on having done a routing lookup, which
409 	 * we will probably have to do anyway, so we might
410 	 * as well do it now.  On the other hand if we are
411 	 * sending to multiple destinations we may have already
412 	 * done the lookup, so see if we can use the route
413 	 * from before.  In any case, we only
414 	 * chose a port number once, even if sending to multiple
415 	 * destinations.
416 	 */
417 	if (in_nullhost(inp->inp_laddr)) {
418 		int xerror;
419 		ifaddr = in_selectsrc(sin, &inp->inp_route,
420 		    inp->inp_socket->so_options, inp->inp_moptions, &xerror);
421 		if (ifaddr == NULL) {
422 			if (xerror == 0)
423 				xerror = EADDRNOTAVAIL;
424 			return xerror;
425 		}
426 		INADDR_TO_IA(ifaddr->sin_addr, ia);
427 		if (ia == NULL)
428 			return (EADDRNOTAVAIL);
429 	}
430 	if (in_pcblookup_connect(inp->inp_table, sin->sin_addr, sin->sin_port,
431 	    !in_nullhost(inp->inp_laddr) ? inp->inp_laddr : ifaddr->sin_addr,
432 	    inp->inp_lport) != 0)
433 		return (EADDRINUSE);
434 	if (in_nullhost(inp->inp_laddr)) {
435 		if (inp->inp_lport == 0) {
436 			error = in_pcbbind(inp, NULL, l);
437 			/*
438 			 * This used to ignore the return value
439 			 * completely, but we need to check for
440 			 * ephemeral port shortage.
441 			 * And attempts to request low ports if not root.
442 			 */
443 			if (error != 0)
444 				return (error);
445 		}
446 		inp->inp_laddr = ifaddr->sin_addr;
447 	}
448 	inp->inp_faddr = sin->sin_addr;
449 	inp->inp_fport = sin->sin_port;
450 	in_pcbstate(inp, INP_CONNECTED);
451 #if defined(IPSEC) || defined(FAST_IPSEC)
452 	if (inp->inp_socket->so_type == SOCK_STREAM)
453 		ipsec_pcbconn(inp->inp_sp);
454 #endif
455 	return (0);
456 }
457 
458 void
459 in_pcbdisconnect(void *v)
460 {
461 	struct inpcb *inp = v;
462 
463 	if (inp->inp_af != AF_INET)
464 		return;
465 
466 	inp->inp_faddr = zeroin_addr;
467 	inp->inp_fport = 0;
468 	in_pcbstate(inp, INP_BOUND);
469 #if defined(IPSEC) || defined(FAST_IPSEC)
470 	ipsec_pcbdisconn(inp->inp_sp);
471 #endif
472 	if (inp->inp_socket->so_state & SS_NOFDREF)
473 		in_pcbdetach(inp);
474 }
475 
476 void
477 in_pcbdetach(void *v)
478 {
479 	struct inpcb *inp = v;
480 	struct socket *so = inp->inp_socket;
481 	int s;
482 
483 	if (inp->inp_af != AF_INET)
484 		return;
485 
486 #if defined(IPSEC) || defined(FAST_IPSEC)
487 	ipsec4_delete_pcbpolicy(inp);
488 #endif /*IPSEC*/
489 	so->so_pcb = 0;
490 	/* sofree drop's the socket's lock */
491 	sofree(so);
492 	mutex_enter(softnet_lock);
493 	if (inp->inp_options)
494 		(void)m_free(inp->inp_options);
495 	rtcache_free(&inp->inp_route);
496 	ip_freemoptions(inp->inp_moptions);
497 	s = splnet();
498 	in_pcbstate(inp, INP_ATTACHED);
499 	LIST_REMOVE(&inp->inp_head, inph_lhash);
500 	CIRCLEQ_REMOVE(&inp->inp_table->inpt_queue, &inp->inp_head,
501 	    inph_queue);
502 	pool_put(&inpcb_pool, inp);
503 	splx(s);
504 }
505 
506 void
507 in_setsockaddr(struct inpcb *inp, struct mbuf *nam)
508 {
509 	struct sockaddr_in *sin;
510 
511 	if (inp->inp_af != AF_INET)
512 		return;
513 
514 	sin = mtod(nam, struct sockaddr_in *);
515 	sockaddr_in_init(sin, &inp->inp_laddr, inp->inp_lport);
516 	nam->m_len = sin->sin_len;
517 }
518 
519 void
520 in_setpeeraddr(struct inpcb *inp, struct mbuf *nam)
521 {
522 	struct sockaddr_in *sin;
523 
524 	if (inp->inp_af != AF_INET)
525 		return;
526 
527 	sin = mtod(nam, struct sockaddr_in *);
528 	sockaddr_in_init(sin, &inp->inp_faddr, inp->inp_fport);
529 	nam->m_len = sin->sin_len;
530 }
531 
532 /*
533  * Pass some notification to all connections of a protocol
534  * associated with address dst.  The local address and/or port numbers
535  * may be specified to limit the search.  The "usual action" will be
536  * taken, depending on the ctlinput cmd.  The caller must filter any
537  * cmds that are uninteresting (e.g., no error in the map).
538  * Call the protocol specific routine (if any) to report
539  * any errors for each matching socket.
540  *
541  * Must be called at splsoftnet.
542  */
543 int
544 in_pcbnotify(struct inpcbtable *table, struct in_addr faddr, u_int fport_arg,
545     struct in_addr laddr, u_int lport_arg, int errno,
546     void (*notify)(struct inpcb *, int))
547 {
548 	struct inpcbhead *head;
549 	struct inpcb *inp, *ninp;
550 	u_int16_t fport = fport_arg, lport = lport_arg;
551 	int nmatch;
552 
553 	if (in_nullhost(faddr) || notify == 0)
554 		return (0);
555 
556 	nmatch = 0;
557 	head = INPCBHASH_CONNECT(table, faddr, fport, laddr, lport);
558 	for (inp = (struct inpcb *)LIST_FIRST(head); inp != NULL; inp = ninp) {
559 		ninp = (struct inpcb *)LIST_NEXT(inp, inp_hash);
560 		if (inp->inp_af != AF_INET)
561 			continue;
562 		if (in_hosteq(inp->inp_faddr, faddr) &&
563 		    inp->inp_fport == fport &&
564 		    inp->inp_lport == lport &&
565 		    in_hosteq(inp->inp_laddr, laddr)) {
566 			(*notify)(inp, errno);
567 			nmatch++;
568 		}
569 	}
570 	return (nmatch);
571 }
572 
573 void
574 in_pcbnotifyall(struct inpcbtable *table, struct in_addr faddr, int errno,
575     void (*notify)(struct inpcb *, int))
576 {
577 	struct inpcb *inp, *ninp;
578 
579 	if (in_nullhost(faddr) || notify == 0)
580 		return;
581 
582 	for (inp = (struct inpcb *)CIRCLEQ_FIRST(&table->inpt_queue);
583 	    inp != (void *)&table->inpt_queue;
584 	    inp = ninp) {
585 		ninp = (struct inpcb *)CIRCLEQ_NEXT(inp, inp_queue);
586 		if (inp->inp_af != AF_INET)
587 			continue;
588 		if (in_hosteq(inp->inp_faddr, faddr))
589 			(*notify)(inp, errno);
590 	}
591 }
592 
593 void
594 in_pcbpurgeif0(struct inpcbtable *table, struct ifnet *ifp)
595 {
596 	struct inpcb *inp, *ninp;
597 	struct ip_moptions *imo;
598 	int i, gap;
599 
600 	for (inp = (struct inpcb *)CIRCLEQ_FIRST(&table->inpt_queue);
601 	    inp != (void *)&table->inpt_queue;
602 	    inp = ninp) {
603 		ninp = (struct inpcb *)CIRCLEQ_NEXT(inp, inp_queue);
604 		if (inp->inp_af != AF_INET)
605 			continue;
606 		imo = inp->inp_moptions;
607 		if (imo != NULL) {
608 			/*
609 			 * Unselect the outgoing interface if it is being
610 			 * detached.
611 			 */
612 			if (imo->imo_multicast_ifp == ifp)
613 				imo->imo_multicast_ifp = NULL;
614 
615 			/*
616 			 * Drop multicast group membership if we joined
617 			 * through the interface being detached.
618 			 */
619 			for (i = 0, gap = 0; i < imo->imo_num_memberships;
620 			    i++) {
621 				if (imo->imo_membership[i]->inm_ifp == ifp) {
622 					in_delmulti(imo->imo_membership[i]);
623 					gap++;
624 				} else if (gap != 0)
625 					imo->imo_membership[i - gap] =
626 					    imo->imo_membership[i];
627 			}
628 			imo->imo_num_memberships -= gap;
629 		}
630 	}
631 }
632 
633 void
634 in_pcbpurgeif(struct inpcbtable *table, struct ifnet *ifp)
635 {
636 	struct rtentry *rt;
637 	struct inpcb *inp, *ninp;
638 
639 	for (inp = (struct inpcb *)CIRCLEQ_FIRST(&table->inpt_queue);
640 	    inp != (void *)&table->inpt_queue;
641 	    inp = ninp) {
642 		ninp = (struct inpcb *)CIRCLEQ_NEXT(inp, inp_queue);
643 		if (inp->inp_af != AF_INET)
644 			continue;
645 		if ((rt = rtcache_validate(&inp->inp_route)) != NULL &&
646 		    rt->rt_ifp == ifp)
647 			in_rtchange(inp, 0);
648 	}
649 }
650 
651 /*
652  * Check for alternatives when higher level complains
653  * about service problems.  For now, invalidate cached
654  * routing information.  If the route was created dynamically
655  * (by a redirect), time to try a default gateway again.
656  */
657 void
658 in_losing(struct inpcb *inp)
659 {
660 	struct rtentry *rt;
661 	struct rt_addrinfo info;
662 
663 	if (inp->inp_af != AF_INET)
664 		return;
665 
666 	if ((rt = rtcache_validate(&inp->inp_route)) == NULL)
667 		return;
668 
669 	memset(&info, 0, sizeof(info));
670 	info.rti_info[RTAX_DST] = rtcache_getdst(&inp->inp_route);
671 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
672 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
673 	rt_missmsg(RTM_LOSING, &info, rt->rt_flags, 0);
674 	if (rt->rt_flags & RTF_DYNAMIC)
675 		(void) rtrequest(RTM_DELETE, rt_getkey(rt),
676 			rt->rt_gateway, rt_mask(rt), rt->rt_flags,
677 			NULL);
678 	/*
679 	 * A new route can be allocated
680 	 * the next time output is attempted.
681 	 */
682 	rtcache_free(&inp->inp_route);
683 }
684 
685 /*
686  * After a routing change, flush old routing.  A new route can be
687  * allocated the next time output is attempted.
688  */
689 void
690 in_rtchange(struct inpcb *inp, int errno)
691 {
692 
693 	if (inp->inp_af != AF_INET)
694 		return;
695 
696 	rtcache_free(&inp->inp_route);
697 
698 	/* XXX SHOULD NOTIFY HIGHER-LEVEL PROTOCOLS */
699 }
700 
701 struct inpcb *
702 in_pcblookup_port(struct inpcbtable *table, struct in_addr laddr,
703     u_int lport_arg, int lookup_wildcard)
704 {
705 	struct inpcbhead *head;
706 	struct inpcb_hdr *inph;
707 	struct inpcb *inp, *match = 0;
708 	int matchwild = 3, wildcard;
709 	u_int16_t lport = lport_arg;
710 
711 	head = INPCBHASH_PORT(table, lport);
712 	LIST_FOREACH(inph, head, inph_lhash) {
713 		inp = (struct inpcb *)inph;
714 		if (inp->inp_af != AF_INET)
715 			continue;
716 
717 		if (inp->inp_lport != lport)
718 			continue;
719 		wildcard = 0;
720 		if (!in_nullhost(inp->inp_faddr))
721 			wildcard++;
722 		if (in_nullhost(inp->inp_laddr)) {
723 			if (!in_nullhost(laddr))
724 				wildcard++;
725 		} else {
726 			if (in_nullhost(laddr))
727 				wildcard++;
728 			else {
729 				if (!in_hosteq(inp->inp_laddr, laddr))
730 					continue;
731 			}
732 		}
733 		if (wildcard && !lookup_wildcard)
734 			continue;
735 		if (wildcard < matchwild) {
736 			match = inp;
737 			matchwild = wildcard;
738 			if (matchwild == 0)
739 				break;
740 		}
741 	}
742 	return (match);
743 }
744 
745 #ifdef DIAGNOSTIC
746 int	in_pcbnotifymiss = 0;
747 #endif
748 
749 struct inpcb *
750 in_pcblookup_connect(struct inpcbtable *table,
751     struct in_addr faddr, u_int fport_arg,
752     struct in_addr laddr, u_int lport_arg)
753 {
754 	struct inpcbhead *head;
755 	struct inpcb_hdr *inph;
756 	struct inpcb *inp;
757 	u_int16_t fport = fport_arg, lport = lport_arg;
758 
759 	head = INPCBHASH_CONNECT(table, faddr, fport, laddr, lport);
760 	LIST_FOREACH(inph, head, inph_hash) {
761 		inp = (struct inpcb *)inph;
762 		if (inp->inp_af != AF_INET)
763 			continue;
764 
765 		if (in_hosteq(inp->inp_faddr, faddr) &&
766 		    inp->inp_fport == fport &&
767 		    inp->inp_lport == lport &&
768 		    in_hosteq(inp->inp_laddr, laddr))
769 			goto out;
770 	}
771 #ifdef DIAGNOSTIC
772 	if (in_pcbnotifymiss) {
773 		printf("in_pcblookup_connect: faddr=%08x fport=%d laddr=%08x lport=%d\n",
774 		    ntohl(faddr.s_addr), ntohs(fport),
775 		    ntohl(laddr.s_addr), ntohs(lport));
776 	}
777 #endif
778 	return (0);
779 
780 out:
781 	/* Move this PCB to the head of hash chain. */
782 	inph = &inp->inp_head;
783 	if (inph != LIST_FIRST(head)) {
784 		LIST_REMOVE(inph, inph_hash);
785 		LIST_INSERT_HEAD(head, inph, inph_hash);
786 	}
787 	return (inp);
788 }
789 
790 struct inpcb *
791 in_pcblookup_bind(struct inpcbtable *table,
792     struct in_addr laddr, u_int lport_arg)
793 {
794 	struct inpcbhead *head;
795 	struct inpcb_hdr *inph;
796 	struct inpcb *inp;
797 	u_int16_t lport = lport_arg;
798 
799 	head = INPCBHASH_BIND(table, laddr, lport);
800 	LIST_FOREACH(inph, head, inph_hash) {
801 		inp = (struct inpcb *)inph;
802 		if (inp->inp_af != AF_INET)
803 			continue;
804 
805 		if (inp->inp_lport == lport &&
806 		    in_hosteq(inp->inp_laddr, laddr))
807 			goto out;
808 	}
809 	head = INPCBHASH_BIND(table, zeroin_addr, lport);
810 	LIST_FOREACH(inph, head, inph_hash) {
811 		inp = (struct inpcb *)inph;
812 		if (inp->inp_af != AF_INET)
813 			continue;
814 
815 		if (inp->inp_lport == lport &&
816 		    in_hosteq(inp->inp_laddr, zeroin_addr))
817 			goto out;
818 	}
819 #ifdef DIAGNOSTIC
820 	if (in_pcbnotifymiss) {
821 		printf("in_pcblookup_bind: laddr=%08x lport=%d\n",
822 		    ntohl(laddr.s_addr), ntohs(lport));
823 	}
824 #endif
825 	return (0);
826 
827 out:
828 	/* Move this PCB to the head of hash chain. */
829 	inph = &inp->inp_head;
830 	if (inph != LIST_FIRST(head)) {
831 		LIST_REMOVE(inph, inph_hash);
832 		LIST_INSERT_HEAD(head, inph, inph_hash);
833 	}
834 	return (inp);
835 }
836 
837 void
838 in_pcbstate(struct inpcb *inp, int state)
839 {
840 
841 	if (inp->inp_af != AF_INET)
842 		return;
843 
844 	if (inp->inp_state > INP_ATTACHED)
845 		LIST_REMOVE(&inp->inp_head, inph_hash);
846 
847 	switch (state) {
848 	case INP_BOUND:
849 		LIST_INSERT_HEAD(INPCBHASH_BIND(inp->inp_table,
850 		    inp->inp_laddr, inp->inp_lport), &inp->inp_head,
851 		    inph_hash);
852 		break;
853 	case INP_CONNECTED:
854 		LIST_INSERT_HEAD(INPCBHASH_CONNECT(inp->inp_table,
855 		    inp->inp_faddr, inp->inp_fport,
856 		    inp->inp_laddr, inp->inp_lport), &inp->inp_head,
857 		    inph_hash);
858 		break;
859 	}
860 
861 	inp->inp_state = state;
862 }
863 
864 struct rtentry *
865 in_pcbrtentry(struct inpcb *inp)
866 {
867 	struct route *ro;
868 	union {
869 		struct sockaddr		dst;
870 		struct sockaddr_in	dst4;
871 	} u;
872 
873 	if (inp->inp_af != AF_INET)
874 		return (NULL);
875 
876 	ro = &inp->inp_route;
877 
878 	sockaddr_in_init(&u.dst4, &inp->inp_faddr, 0);
879 	return rtcache_lookup(ro, &u.dst);
880 }
881 
882 struct sockaddr_in *
883 in_selectsrc(struct sockaddr_in *sin, struct route *ro,
884     int soopts, struct ip_moptions *mopts, int *errorp)
885 {
886 	struct rtentry *rt = NULL;
887 	struct in_ifaddr *ia = NULL;
888 
889 	/*
890          * If route is known or can be allocated now, take the
891          * source address from the interface.  Otherwise, punt.
892 	 */
893 	if ((soopts & SO_DONTROUTE) != 0)
894 		rtcache_free(ro);
895 	else {
896 		union {
897 			struct sockaddr		dst;
898 			struct sockaddr_in	dst4;
899 		} u;
900 
901 		sockaddr_in_init(&u.dst4, &sin->sin_addr, 0);
902 		rt = rtcache_lookup(ro, &u.dst);
903 	}
904 	/*
905 	 * If we found a route, use the address
906 	 * corresponding to the outgoing interface
907 	 * unless it is the loopback (in case a route
908 	 * to our address on another net goes to loopback).
909 	 *
910 	 * XXX Is this still true?  Do we care?
911 	 */
912 	if (rt != NULL && (rt->rt_ifp->if_flags & IFF_LOOPBACK) == 0)
913 		ia = ifatoia(rt->rt_ifa);
914 	if (ia == NULL) {
915 		u_int16_t fport = sin->sin_port;
916 
917 		sin->sin_port = 0;
918 		ia = ifatoia(ifa_ifwithladdr(sintosa(sin)));
919 		sin->sin_port = fport;
920 		if (ia == NULL) {
921 			/* Find 1st non-loopback AF_INET address */
922 			TAILQ_FOREACH(ia, &in_ifaddrhead, ia_list) {
923 				if (!(ia->ia_ifp->if_flags & IFF_LOOPBACK))
924 					break;
925 			}
926 		}
927 		if (ia == NULL) {
928 			*errorp = EADDRNOTAVAIL;
929 			return NULL;
930 		}
931 	}
932 	/*
933 	 * If the destination address is multicast and an outgoing
934 	 * interface has been set as a multicast option, use the
935 	 * address of that interface as our source address.
936 	 */
937 	if (IN_MULTICAST(sin->sin_addr.s_addr) && mopts != NULL) {
938 		struct ip_moptions *imo;
939 		struct ifnet *ifp;
940 
941 		imo = mopts;
942 		if (imo->imo_multicast_ifp != NULL) {
943 			ifp = imo->imo_multicast_ifp;
944 			IFP_TO_IA(ifp, ia);		/* XXX */
945 			if (ia == 0) {
946 				*errorp = EADDRNOTAVAIL;
947 				return NULL;
948 			}
949 		}
950 	}
951 	if (ia->ia_ifa.ifa_getifa != NULL) {
952 		ia = ifatoia((*ia->ia_ifa.ifa_getifa)(&ia->ia_ifa,
953 		                                      sintosa(sin)));
954 	}
955 #ifdef GETIFA_DEBUG
956 	else
957 		printf("%s: missing ifa_getifa\n", __func__);
958 #endif
959 	return satosin(&ia->ia_addr);
960 }
961