xref: /netbsd-src/sys/netinet/in_pcb.c (revision 001c68bd94f75ce9270b69227c4199fbf34ee396)
1 /*	$NetBSD: in_pcb.c,v 1.83 2003/06/26 00:19:13 itojun 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. All advertising materials mentioning features or use of this software
82  *    must display the following acknowledgement:
83  *	This product includes software developed by the University of
84  *	California, Berkeley and its contributors.
85  * 4. Neither the name of the University nor the names of its contributors
86  *    may be used to endorse or promote products derived from this software
87  *    without specific prior written permission.
88  *
89  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
90  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
91  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
92  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
93  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
94  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
95  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
96  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
97  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
98  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
99  * SUCH DAMAGE.
100  *
101  *	@(#)in_pcb.c	8.4 (Berkeley) 5/24/95
102  */
103 
104 #include <sys/cdefs.h>
105 __KERNEL_RCSID(0, "$NetBSD: in_pcb.c,v 1.83 2003/06/26 00:19:13 itojun Exp $");
106 
107 #include "opt_ipsec.h"
108 
109 #include <sys/param.h>
110 #include <sys/systm.h>
111 #include <sys/malloc.h>
112 #include <sys/mbuf.h>
113 #include <sys/protosw.h>
114 #include <sys/socket.h>
115 #include <sys/socketvar.h>
116 #include <sys/ioctl.h>
117 #include <sys/errno.h>
118 #include <sys/time.h>
119 #include <sys/pool.h>
120 #include <sys/proc.h>
121 
122 #include <net/if.h>
123 #include <net/route.h>
124 
125 #include <netinet/in.h>
126 #include <netinet/in_systm.h>
127 #include <netinet/ip.h>
128 #include <netinet/in_pcb.h>
129 #include <netinet/in_var.h>
130 #include <netinet/ip_var.h>
131 
132 #ifdef IPSEC
133 #include <netinet6/ipsec.h>
134 #include <netkey/key.h>
135 #endif /* IPSEC */
136 
137 struct	in_addr zeroin_addr;
138 
139 #define	INPCBHASH_BIND(table, laddr, lport) \
140 	&(table)->inpt_bindhashtbl[ \
141 	    ((ntohl((laddr).s_addr) + ntohs(lport))) & (table)->inpt_bindhash]
142 #define	INPCBHASH_CONNECT(table, faddr, fport, laddr, lport) \
143 	&(table)->inpt_connecthashtbl[ \
144 	    ((ntohl((faddr).s_addr) + ntohs(fport)) + \
145 	     (ntohl((laddr).s_addr) + ntohs(lport))) & (table)->inpt_connecthash]
146 
147 struct inpcb *
148 	in_pcblookup_port __P((struct inpcbtable *,
149 	    struct in_addr, u_int, int));
150 
151 int	anonportmin = IPPORT_ANONMIN;
152 int	anonportmax = IPPORT_ANONMAX;
153 int	lowportmin  = IPPORT_RESERVEDMIN;
154 int	lowportmax  = IPPORT_RESERVEDMAX;
155 
156 struct pool inpcb_pool;
157 
158 void
159 in_pcbinit(table, bindhashsize, connecthashsize)
160 	struct inpcbtable *table;
161 	int bindhashsize, connecthashsize;
162 {
163 	static int inpcb_pool_initialized;
164 
165 	if (inpcb_pool_initialized == 0) {
166 		pool_init(&inpcb_pool, sizeof(struct inpcb), 0, 0, 0,
167 		    "inpcbpl", NULL);
168 		inpcb_pool_initialized = 1;
169 	}
170 
171 	CIRCLEQ_INIT(&table->inpt_queue);
172 	table->inpt_bindhashtbl = hashinit(bindhashsize, HASH_LIST, M_PCB,
173 	    M_WAITOK, &table->inpt_bindhash);
174 	table->inpt_connecthashtbl = hashinit(connecthashsize, HASH_LIST,
175 	    M_PCB, M_WAITOK, &table->inpt_connecthash);
176 	table->inpt_lastlow = IPPORT_RESERVEDMAX;
177 	table->inpt_lastport = (u_int16_t)anonportmax;
178 }
179 
180 int
181 in_pcballoc(so, v)
182 	struct socket *so;
183 	void *v;
184 {
185 	struct inpcbtable *table = v;
186 	struct inpcb *inp;
187 	int s;
188 #ifdef IPSEC
189 	int error;
190 #endif
191 
192 	inp = pool_get(&inpcb_pool, PR_NOWAIT);
193 	if (inp == NULL)
194 		return (ENOBUFS);
195 	bzero((caddr_t)inp, sizeof(*inp));
196 	inp->inp_table = table;
197 	inp->inp_socket = so;
198 	inp->inp_errormtu = -1;
199 #ifdef IPSEC
200 	error = ipsec_init_pcbpolicy(so, &inp->inp_sp);
201 	if (error != 0) {
202 		pool_put(&inpcb_pool, inp);
203 		return error;
204 	}
205 #endif
206 	so->so_pcb = inp;
207 	s = splnet();
208 	CIRCLEQ_INSERT_HEAD(&table->inpt_queue, inp, inp_queue);
209 	in_pcbstate(inp, INP_ATTACHED);
210 	splx(s);
211 	return (0);
212 }
213 
214 int
215 in_pcbbind(v, nam, p)
216 	void *v;
217 	struct mbuf *nam;
218 	struct proc *p;
219 {
220 	struct in_ifaddr *ia = NULL;
221 	struct inpcb *inp = v;
222 	struct socket *so = inp->inp_socket;
223 	struct inpcbtable *table = inp->inp_table;
224 	struct sockaddr_in *sin;
225 	u_int16_t lport = 0;
226 	int wild = 0, reuseport = (so->so_options & SO_REUSEPORT);
227 
228 	if (TAILQ_FIRST(&in_ifaddr) == 0)
229 		return (EADDRNOTAVAIL);
230 	if (inp->inp_lport || !in_nullhost(inp->inp_laddr))
231 		return (EINVAL);
232 	if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0)
233 		wild = 1;
234 	if (nam == 0)
235 		goto noname;
236 	sin = mtod(nam, struct sockaddr_in *);
237 	if (nam->m_len != sizeof (*sin))
238 		return (EINVAL);
239 	if (sin->sin_family != AF_INET)
240 		return (EAFNOSUPPORT);
241 	lport = sin->sin_port;
242 	if (IN_MULTICAST(sin->sin_addr.s_addr)) {
243 		/*
244 		 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
245 		 * allow complete duplication of binding if
246 		 * SO_REUSEPORT is set, or if SO_REUSEADDR is set
247 		 * and a multicast address is bound on both
248 		 * new and duplicated sockets.
249 		 */
250 		if (so->so_options & SO_REUSEADDR)
251 			reuseport = SO_REUSEADDR|SO_REUSEPORT;
252 	} else if (!in_nullhost(sin->sin_addr)) {
253 		sin->sin_port = 0;		/* yech... */
254 		INADDR_TO_IA(sin->sin_addr, ia);
255 		if (ia == NULL)
256 			return (EADDRNOTAVAIL);
257 	}
258 	if (lport) {
259 		struct inpcb *t;
260 #ifndef IPNOPRIVPORTS
261 		/* GROSS */
262 		if (ntohs(lport) < IPPORT_RESERVED &&
263 		    (p == 0 || suser(p->p_ucred, &p->p_acflag)))
264 			return (EACCES);
265 #endif
266 		if (so->so_uid && !IN_MULTICAST(sin->sin_addr.s_addr)) {
267 			t = in_pcblookup_port(table, sin->sin_addr, lport, 1);
268 		/*
269 		 * XXX:	investigate ramifications of loosening this
270 		 *	restriction so that as long as both ports have
271 		 *	SO_REUSEPORT allow the bind
272 		 */
273 			if (t &&
274 			    (!in_nullhost(sin->sin_addr) ||
275 			     !in_nullhost(t->inp_laddr) ||
276 			     (t->inp_socket->so_options & SO_REUSEPORT) == 0)
277 			    && (so->so_uid != t->inp_socket->so_uid)) {
278 				return (EADDRINUSE);
279 			}
280 		}
281 		t = in_pcblookup_port(table, sin->sin_addr, lport, wild);
282 		if (t && (reuseport & t->inp_socket->so_options) == 0)
283 			return (EADDRINUSE);
284 	}
285 	if (!in_nullhost(inp->inp_laddr)) {
286 		KASSERT(inp->inp_ia != NULL);
287 		LIST_REMOVE(inp, inp_ialink);
288 		IFAFREE(&inp->inp_ia->ia_ifa);
289 		inp->inp_ia = NULL;
290 	}
291 	inp->inp_laddr = sin->sin_addr;
292 	if (ia != NULL) {
293 		inp->inp_ia = ia;
294 		LIST_INSERT_HEAD(&ia->ia_inpcbs, inp, inp_ialink);
295 		IFAREF(&ia->ia_ifa);
296 	}
297 
298 
299 noname:
300 	if (lport == 0) {
301 		int	   cnt;
302 		u_int16_t  min, max;
303 		u_int16_t *lastport;
304 
305 		if (inp->inp_flags & INP_LOWPORT) {
306 #ifndef IPNOPRIVPORTS
307 			if (p == 0 || suser(p->p_ucred, &p->p_acflag))
308 				return (EACCES);
309 #endif
310 			min = lowportmin;
311 			max = lowportmax;
312 			lastport = &table->inpt_lastlow;
313 		} else {
314 			min = anonportmin;
315 			max = anonportmax;
316 			lastport = &table->inpt_lastport;
317 		}
318 		if (min > max) {	/* sanity check */
319 			u_int16_t swp;
320 
321 			swp = min;
322 			min = max;
323 			max = swp;
324 		}
325 
326 		lport = *lastport - 1;
327 		for (cnt = max - min + 1; cnt; cnt--, lport--) {
328 			if (lport < min || lport > max)
329 				lport = max;
330 			if (!in_pcblookup_port(table, inp->inp_laddr,
331 			    htons(lport), 1))
332 				goto found;
333 		}
334 		if (!in_nullhost(inp->inp_laddr)) {
335 		       if (inp->inp_ia != NULL) {
336 				LIST_REMOVE(inp, inp_ialink);
337 				IFAFREE(&inp->inp_ia->ia_ifa);
338 				inp->inp_ia = NULL;
339 			}
340 			inp->inp_laddr.s_addr = INADDR_ANY;
341 		}
342 		return (EAGAIN);
343 	found:
344 		inp->inp_flags |= INP_ANONPORT;
345 		*lastport = lport;
346 		lport = htons(lport);
347 	}
348 	inp->inp_lport = lport;
349 	in_pcbstate(inp, INP_BOUND);
350 	return (0);
351 }
352 
353 /*
354  * Connect from a socket to a specified address.
355  * Both address and port must be specified in argument sin.
356  * If don't have a local address for this socket yet,
357  * then pick one.
358  */
359 int
360 in_pcbconnect(v, nam)
361 	void *v;
362 	struct mbuf *nam;
363 {
364 	struct inpcb *inp = v;
365 	struct in_ifaddr *ia = NULL;
366 	struct sockaddr_in *ifaddr = NULL;
367 	struct sockaddr_in *sin = mtod(nam, struct sockaddr_in *);
368 	int error;
369 
370 	if (nam->m_len != sizeof (*sin))
371 		return (EINVAL);
372 	if (sin->sin_family != AF_INET)
373 		return (EAFNOSUPPORT);
374 	if (sin->sin_port == 0)
375 		return (EADDRNOTAVAIL);
376 	if (TAILQ_FIRST(&in_ifaddr) != 0) {
377 		/*
378 		 * If the destination address is INADDR_ANY,
379 		 * use any local address (likely loopback).
380 		 * If the supplied address is INADDR_BROADCAST,
381 		 * use the broadcast address of an interface
382 		 * which supports broadcast. (loopback does not)
383 		 */
384 
385 		if (in_nullhost(sin->sin_addr)) {
386 			sin->sin_addr =
387 			    TAILQ_FIRST(&in_ifaddr)->ia_addr.sin_addr;
388 		} else if (sin->sin_addr.s_addr == INADDR_BROADCAST) {
389 			TAILQ_FOREACH(ia, &in_ifaddr, ia_list) {
390 				if (ia->ia_ifp->if_flags & IFF_BROADCAST) {
391 					sin->sin_addr =
392 					    ia->ia_broadaddr.sin_addr;
393 					break;
394 				}
395 			}
396 		}
397 	}
398 	/*
399 	 * If we haven't bound which network number to use as ours,
400 	 * we will use the number of the outgoing interface.
401 	 * This depends on having done a routing lookup, which
402 	 * we will probably have to do anyway, so we might
403 	 * as well do it now.  On the other hand if we are
404 	 * sending to multiple destinations we may have already
405 	 * done the lookup, so see if we can use the route
406 	 * from before.  In any case, we only
407 	 * chose a port number once, even if sending to multiple
408 	 * destinations.
409 	 */
410 	if (in_nullhost(inp->inp_laddr)) {
411 		int error;
412 		ifaddr = in_selectsrc(sin, &inp->inp_route,
413 			inp->inp_socket->so_options, inp->inp_moptions, &error);
414 		if (ifaddr == NULL) {
415 			if (error == 0)
416 				error = EADDRNOTAVAIL;
417 			return error;
418 		}
419 		INADDR_TO_IA(ifaddr->sin_addr, ia);
420 		if (ia == NULL)
421 			return (EADDRNOTAVAIL);
422 	}
423 	if (in_pcblookup_connect(inp->inp_table, sin->sin_addr, sin->sin_port,
424 	    !in_nullhost(inp->inp_laddr) ? inp->inp_laddr : ifaddr->sin_addr,
425 	    inp->inp_lport) != 0)
426 		return (EADDRINUSE);
427 	if (in_nullhost(inp->inp_laddr)) {
428 		if (inp->inp_lport == 0) {
429 			error = in_pcbbind(inp, (struct mbuf *)0,
430 			    (struct proc *)0);
431 			/*
432 			 * This used to ignore the return value
433 			 * completely, but we need to check for
434 			 * ephemeral port shortage.
435 			 * XXX Should we check for other errors, too?
436 			 */
437 			if (error == EAGAIN)
438 				return (error);
439 		}
440 		KASSERT(inp->inp_ia == NULL);
441 		inp->inp_laddr = ia->ia_addr.sin_addr;
442 		inp->inp_ia = ia;
443 		LIST_INSERT_HEAD(&ia->ia_inpcbs, inp, inp_ialink);
444 		IFAREF(&ia->ia_ifa);
445 		inp->inp_laddr = ifaddr->sin_addr;
446 	}
447 	inp->inp_faddr = sin->sin_addr;
448 	inp->inp_fport = sin->sin_port;
449 	in_pcbstate(inp, INP_CONNECTED);
450 #ifdef IPSEC
451 	if (inp->inp_socket->so_type == SOCK_STREAM)
452 		ipsec_pcbconn(inp->inp_sp);
453 #endif
454 	return (0);
455 }
456 
457 void
458 in_pcbdisconnect(v)
459 	void *v;
460 {
461 	struct inpcb *inp = v;
462 
463 	inp->inp_faddr = zeroin_addr;
464 	inp->inp_fport = 0;
465 	in_pcbstate(inp, INP_BOUND);
466 	if (inp->inp_socket->so_state & SS_NOFDREF)
467 		in_pcbdetach(inp);
468 #ifdef IPSEC
469 	ipsec_pcbdisconn(inp->inp_sp);
470 #endif
471 }
472 
473 void
474 in_pcbdetach(v)
475 	void *v;
476 {
477 	struct inpcb *inp = v;
478 	struct socket *so = inp->inp_socket;
479 	int s;
480 
481 #ifdef IPSEC
482 	ipsec4_delete_pcbpolicy(inp);
483 #endif /*IPSEC*/
484 	so->so_pcb = 0;
485 	sofree(so);
486 	if (inp->inp_options)
487 		(void)m_free(inp->inp_options);
488 	if (inp->inp_route.ro_rt)
489 		rtfree(inp->inp_route.ro_rt);
490 	ip_freemoptions(inp->inp_moptions);
491 	if (inp->inp_ia != NULL) {
492 		LIST_REMOVE(inp, inp_ialink);
493 		IFAFREE(&inp->inp_ia->ia_ifa);
494 		inp->inp_ia = NULL;
495 	}
496 	s = splnet();
497 	in_pcbstate(inp, INP_ATTACHED);
498 	CIRCLEQ_REMOVE(&inp->inp_table->inpt_queue, inp, inp_queue);
499 	splx(s);
500 	pool_put(&inpcb_pool, inp);
501 }
502 
503 void
504 in_setsockaddr(inp, nam)
505 	struct inpcb *inp;
506 	struct mbuf *nam;
507 {
508 	struct sockaddr_in *sin;
509 
510 	nam->m_len = sizeof (*sin);
511 	sin = mtod(nam, struct sockaddr_in *);
512 	bzero((caddr_t)sin, sizeof (*sin));
513 	sin->sin_family = AF_INET;
514 	sin->sin_len = sizeof(*sin);
515 	sin->sin_port = inp->inp_lport;
516 	sin->sin_addr = inp->inp_laddr;
517 }
518 
519 void
520 in_setpeeraddr(inp, nam)
521 	struct inpcb *inp;
522 	struct mbuf *nam;
523 {
524 	struct sockaddr_in *sin;
525 
526 	nam->m_len = sizeof (*sin);
527 	sin = mtod(nam, struct sockaddr_in *);
528 	bzero((caddr_t)sin, sizeof (*sin));
529 	sin->sin_family = AF_INET;
530 	sin->sin_len = sizeof(*sin);
531 	sin->sin_port = inp->inp_fport;
532 	sin->sin_addr = inp->inp_faddr;
533 }
534 
535 /*
536  * Pass some notification to all connections of a protocol
537  * associated with address dst.  The local address and/or port numbers
538  * may be specified to limit the search.  The "usual action" will be
539  * taken, depending on the ctlinput cmd.  The caller must filter any
540  * cmds that are uninteresting (e.g., no error in the map).
541  * Call the protocol specific routine (if any) to report
542  * any errors for each matching socket.
543  *
544  * Must be called at splsoftnet.
545  */
546 int
547 in_pcbnotify(table, faddr, fport_arg, laddr, lport_arg, errno, notify)
548 	struct inpcbtable *table;
549 	struct in_addr faddr, laddr;
550 	u_int fport_arg, lport_arg;
551 	int errno;
552 	void (*notify) __P((struct inpcb *, int));
553 {
554 	struct inpcbhead *head;
555 	struct inpcb *inp, *ninp;
556 	u_int16_t fport = fport_arg, lport = lport_arg;
557 	int nmatch;
558 
559 	if (in_nullhost(faddr) || notify == 0)
560 		return (0);
561 
562 	nmatch = 0;
563 	head = INPCBHASH_CONNECT(table, faddr, fport, laddr, lport);
564 	for (inp = LIST_FIRST(head); inp != NULL; inp = ninp) {
565 		ninp = LIST_NEXT(inp, inp_hash);
566 		if (in_hosteq(inp->inp_faddr, faddr) &&
567 		    inp->inp_fport == fport &&
568 		    inp->inp_lport == lport &&
569 		    in_hosteq(inp->inp_laddr, laddr)) {
570 			(*notify)(inp, errno);
571 			nmatch++;
572 		}
573 	}
574 	return (nmatch);
575 }
576 
577 void
578 in_pcbnotifyall(table, faddr, errno, notify)
579 	struct inpcbtable *table;
580 	struct in_addr faddr;
581 	int errno;
582 	void (*notify) __P((struct inpcb *, int));
583 {
584 	struct inpcb *inp, *ninp;
585 
586 	if (in_nullhost(faddr) || notify == 0)
587 		return;
588 
589 	for (inp = CIRCLEQ_FIRST(&table->inpt_queue);
590 	    inp != (void *)&table->inpt_queue;
591 	    inp = ninp) {
592 		ninp = CIRCLEQ_NEXT(inp, inp_queue);
593 		if (in_hosteq(inp->inp_faddr, faddr))
594 			(*notify)(inp, errno);
595 	}
596 }
597 
598 void
599 in_pcbpurgeif0(table, ifp)
600 	struct inpcbtable *table;
601 	struct ifnet *ifp;
602 {
603 	struct inpcb *inp, *ninp;
604 	struct ip_moptions *imo;
605 	int i, gap;
606 
607 	for (inp = CIRCLEQ_FIRST(&table->inpt_queue);
608 	    inp != (void *)&table->inpt_queue;
609 	    inp = ninp) {
610 		ninp = CIRCLEQ_NEXT(inp, inp_queue);
611 		imo = inp->inp_moptions;
612 		if (imo != NULL) {
613 			/*
614 			 * Unselect the outgoing interface if it is being
615 			 * detached.
616 			 */
617 			if (imo->imo_multicast_ifp == ifp)
618 				imo->imo_multicast_ifp = NULL;
619 
620 			/*
621 			 * Drop multicast group membership if we joined
622 			 * through the interface being detached.
623 			 */
624 			for (i = 0, gap = 0; i < imo->imo_num_memberships;
625 			    i++) {
626 				if (imo->imo_membership[i]->inm_ifp == ifp) {
627 					in_delmulti(imo->imo_membership[i]);
628 					gap++;
629 				} else if (gap != 0)
630 					imo->imo_membership[i - gap] =
631 					    imo->imo_membership[i];
632 			}
633 			imo->imo_num_memberships -= gap;
634 		}
635 	}
636 }
637 
638 void
639 in_pcbpurgeif(table, ifp)
640 	struct inpcbtable *table;
641 	struct ifnet *ifp;
642 {
643 	struct inpcb *inp, *ninp;
644 
645 	for (inp = CIRCLEQ_FIRST(&table->inpt_queue);
646 	    inp != (void *)&table->inpt_queue;
647 	    inp = ninp) {
648 		ninp = CIRCLEQ_NEXT(inp, inp_queue);
649 		if (inp->inp_route.ro_rt != NULL &&
650 		    inp->inp_route.ro_rt->rt_ifp == ifp)
651 			in_rtchange(inp, 0);
652 	}
653 }
654 
655 /*
656  * Check for alternatives when higher level complains
657  * about service problems.  For now, invalidate cached
658  * routing information.  If the route was created dynamically
659  * (by a redirect), time to try a default gateway again.
660  */
661 void
662 in_losing(inp)
663 	struct inpcb *inp;
664 {
665 	struct rtentry *rt;
666 	struct rt_addrinfo info;
667 
668 	if ((rt = inp->inp_route.ro_rt)) {
669 		inp->inp_route.ro_rt = 0;
670 		bzero((caddr_t)&info, sizeof(info));
671 		info.rti_info[RTAX_DST] = &inp->inp_route.ro_dst;
672 		info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
673 		info.rti_info[RTAX_NETMASK] = rt_mask(rt);
674 		rt_missmsg(RTM_LOSING, &info, rt->rt_flags, 0);
675 		if (rt->rt_flags & RTF_DYNAMIC)
676 			(void) rtrequest(RTM_DELETE, rt_key(rt),
677 				rt->rt_gateway, rt_mask(rt), rt->rt_flags,
678 				(struct rtentry **)0);
679 		else
680 		/*
681 		 * A new route can be allocated
682 		 * the next time output is attempted.
683 		 */
684 			rtfree(rt);
685 	}
686 }
687 
688 /*
689  * After a routing change, flush old routing
690  * and allocate a (hopefully) better one.
691  */
692 void
693 in_rtchange(inp, errno)
694 	struct inpcb *inp;
695 	int errno;
696 {
697 
698 	if (inp->inp_route.ro_rt) {
699 		rtfree(inp->inp_route.ro_rt);
700 		inp->inp_route.ro_rt = 0;
701 		/*
702 		 * A new route can be allocated the next time
703 		 * output is attempted.
704 		 */
705 	}
706 	/* XXX SHOULD NOTIFY HIGHER-LEVEL PROTOCOLS */
707 }
708 
709 struct inpcb *
710 in_pcblookup_port(table, laddr, lport_arg, lookup_wildcard)
711 	struct inpcbtable *table;
712 	struct in_addr laddr;
713 	u_int lport_arg;
714 	int lookup_wildcard;
715 {
716 	struct inpcb *inp, *match = 0;
717 	int matchwild = 3, wildcard;
718 	u_int16_t lport = lport_arg;
719 
720 	CIRCLEQ_FOREACH(inp, &table->inpt_queue, inp_queue) {
721 		if (inp->inp_lport != lport)
722 			continue;
723 		wildcard = 0;
724 		if (!in_nullhost(inp->inp_faddr))
725 			wildcard++;
726 		if (in_nullhost(inp->inp_laddr)) {
727 			if (!in_nullhost(laddr))
728 				wildcard++;
729 		} else {
730 			if (in_nullhost(laddr))
731 				wildcard++;
732 			else {
733 				if (!in_hosteq(inp->inp_laddr, laddr))
734 					continue;
735 			}
736 		}
737 		if (wildcard && !lookup_wildcard)
738 			continue;
739 		if (wildcard < matchwild) {
740 			match = inp;
741 			matchwild = wildcard;
742 			if (matchwild == 0)
743 				break;
744 		}
745 	}
746 	return (match);
747 }
748 
749 #ifdef DIAGNOSTIC
750 int	in_pcbnotifymiss = 0;
751 #endif
752 
753 struct inpcb *
754 in_pcblookup_connect(table, faddr, fport_arg, laddr, lport_arg)
755 	struct inpcbtable *table;
756 	struct in_addr faddr, laddr;
757 	u_int fport_arg, lport_arg;
758 {
759 	struct inpcbhead *head;
760 	struct inpcb *inp;
761 	u_int16_t fport = fport_arg, lport = lport_arg;
762 
763 	head = INPCBHASH_CONNECT(table, faddr, fport, laddr, lport);
764 	LIST_FOREACH(inp, head, inp_hash) {
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 	if (inp != LIST_FIRST(head)) {
783 		LIST_REMOVE(inp, inp_hash);
784 		LIST_INSERT_HEAD(head, inp, inp_hash);
785 	}
786 	return (inp);
787 }
788 
789 struct inpcb *
790 in_pcblookup_bind(table, laddr, lport_arg)
791 	struct inpcbtable *table;
792 	struct in_addr laddr;
793 	u_int lport_arg;
794 {
795 	struct inpcbhead *head;
796 	struct inpcb *inp;
797 	u_int16_t lport = lport_arg;
798 
799 	head = INPCBHASH_BIND(table, laddr, lport);
800 	LIST_FOREACH(inp, head, inp_hash) {
801 		if (inp->inp_lport == lport &&
802 		    in_hosteq(inp->inp_laddr, laddr))
803 			goto out;
804 	}
805 	head = INPCBHASH_BIND(table, zeroin_addr, lport);
806 	LIST_FOREACH(inp, head, inp_hash) {
807 		if (inp->inp_lport == lport &&
808 		    in_hosteq(inp->inp_laddr, zeroin_addr))
809 			goto out;
810 	}
811 #ifdef DIAGNOSTIC
812 	if (in_pcbnotifymiss) {
813 		printf("in_pcblookup_bind: laddr=%08x lport=%d\n",
814 		    ntohl(laddr.s_addr), ntohs(lport));
815 	}
816 #endif
817 	return (0);
818 
819 out:
820 	/* Move this PCB to the head of hash chain. */
821 	if (inp != LIST_FIRST(head)) {
822 		LIST_REMOVE(inp, inp_hash);
823 		LIST_INSERT_HEAD(head, inp, inp_hash);
824 	}
825 	return (inp);
826 }
827 
828 void
829 in_pcbstate(inp, state)
830 	struct inpcb *inp;
831 	int state;
832 {
833 
834 	if (inp->inp_state > INP_ATTACHED)
835 		LIST_REMOVE(inp, inp_hash);
836 
837 	switch (state) {
838 	case INP_BOUND:
839 		LIST_INSERT_HEAD(INPCBHASH_BIND(inp->inp_table,
840 		    inp->inp_laddr, inp->inp_lport), inp, inp_hash);
841 		break;
842 	case INP_CONNECTED:
843 		LIST_INSERT_HEAD(INPCBHASH_CONNECT(inp->inp_table,
844 		    inp->inp_faddr, inp->inp_fport,
845 		    inp->inp_laddr, inp->inp_lport), inp, inp_hash);
846 		break;
847 	}
848 
849 	inp->inp_state = state;
850 }
851 
852 struct rtentry *
853 in_pcbrtentry(inp)
854 	struct inpcb *inp;
855 {
856 	struct route *ro;
857 
858 	ro = &inp->inp_route;
859 
860 	if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
861 	    !in_hosteq(satosin(&ro->ro_dst)->sin_addr, inp->inp_faddr))) {
862 		RTFREE(ro->ro_rt);
863 		ro->ro_rt = (struct rtentry *)NULL;
864 	}
865 	if (ro->ro_rt == (struct rtentry *)NULL &&
866 	    !in_nullhost(inp->inp_faddr)) {
867 		bzero(&ro->ro_dst, sizeof(struct sockaddr_in));
868 		ro->ro_dst.sa_family = AF_INET;
869 		ro->ro_dst.sa_len = sizeof(ro->ro_dst);
870 		satosin(&ro->ro_dst)->sin_addr = inp->inp_faddr;
871 		rtalloc(ro);
872 	}
873 	return (ro->ro_rt);
874 }
875 
876 struct sockaddr_in *
877 in_selectsrc(sin, ro, soopts, mopts, errorp)
878 	struct sockaddr_in *sin;
879 	struct route *ro;
880 	int soopts;
881 	struct ip_moptions *mopts;
882 	int *errorp;
883 {
884 	struct in_ifaddr *ia;
885 
886 	ia = (struct in_ifaddr *)0;
887 	/*
888 	 * If route is known or can be allocated now,
889 	 * our src addr is taken from the i/f, else punt.
890 	 * Note that we should check the address family of the cached
891 	 * destination, in case of sharing the cache with IPv6.
892 	 */
893 	if (ro->ro_rt &&
894 	    (ro->ro_dst.sa_family != AF_INET ||
895 	    !in_hosteq(satosin(&ro->ro_dst)->sin_addr, sin->sin_addr) ||
896 	    soopts & SO_DONTROUTE)) {
897 		RTFREE(ro->ro_rt);
898 		ro->ro_rt = (struct rtentry *)0;
899 	}
900 	if ((soopts & SO_DONTROUTE) == 0 && /*XXX*/
901 	    (ro->ro_rt == (struct rtentry *)0 ||
902 	     ro->ro_rt->rt_ifp == (struct ifnet *)0)) {
903 		/* No route yet, so try to acquire one */
904 		bzero(&ro->ro_dst, sizeof(struct sockaddr_in));
905 		ro->ro_dst.sa_family = AF_INET;
906 		ro->ro_dst.sa_len = sizeof(struct sockaddr_in);
907 		satosin(&ro->ro_dst)->sin_addr = sin->sin_addr;
908 		rtalloc(ro);
909 	}
910 	/*
911 	 * If we found a route, use the address
912 	 * corresponding to the outgoing interface
913 	 * unless it is the loopback (in case a route
914 	 * to our address on another net goes to loopback).
915 	 *
916 	 * XXX Is this still true?  Do we care?
917 	 */
918 	if (ro->ro_rt && !(ro->ro_rt->rt_ifp->if_flags & IFF_LOOPBACK))
919 		ia = ifatoia(ro->ro_rt->rt_ifa);
920 	if (ia == NULL) {
921 		u_int16_t fport = sin->sin_port;
922 
923 		sin->sin_port = 0;
924 		ia = ifatoia(ifa_ifwithladdr(sintosa(sin)));
925 		sin->sin_port = fport;
926 		if (ia == 0) {
927 			/* Find 1st non-loopback AF_INET address */
928 			TAILQ_FOREACH(ia, &in_ifaddr, ia_list) {
929 				if (!(ia->ia_ifp->if_flags & IFF_LOOPBACK))
930 					break;
931 			}
932 		}
933 		if (ia == NULL) {
934 			*errorp = EADDRNOTAVAIL;
935 			return NULL;
936 		}
937 	}
938 	/*
939 	 * If the destination address is multicast and an outgoing
940 	 * interface has been set as a multicast option, use the
941 	 * address of that interface as our source address.
942 	 */
943 	if (IN_MULTICAST(sin->sin_addr.s_addr) && mopts != NULL) {
944 		struct ip_moptions *imo;
945 		struct ifnet *ifp;
946 
947 		imo = mopts;
948 		if (imo->imo_multicast_ifp != NULL) {
949 			ifp = imo->imo_multicast_ifp;
950 			IFP_TO_IA(ifp, ia);		/* XXX */
951 			if (ia == 0) {
952 				*errorp = EADDRNOTAVAIL;
953 				return NULL;
954 			}
955 		}
956 	}
957 	return satosin(&ia->ia_addr);
958 }
959