xref: /netbsd-src/sys/netinet/in_pcb.c (revision d710132b4b8ce7f7cccaaf660cb16aa16b4077a0)
1 /*	$NetBSD: in_pcb.c,v 1.82 2003/06/15 02:49:33 matt 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 1.82 2003/06/15 02:49:33 matt 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 	}
421 	if (in_pcblookup_connect(inp->inp_table, sin->sin_addr, sin->sin_port,
422 	    !in_nullhost(inp->inp_laddr) ? inp->inp_laddr : ifaddr->sin_addr,
423 	    inp->inp_lport) != 0)
424 		return (EADDRINUSE);
425 	if (in_nullhost(inp->inp_laddr)) {
426 		if (inp->inp_lport == 0) {
427 			error = in_pcbbind(inp, (struct mbuf *)0,
428 			    (struct proc *)0);
429 			/*
430 			 * This used to ignore the return value
431 			 * completely, but we need to check for
432 			 * ephemeral port shortage.
433 			 * XXX Should we check for other errors, too?
434 			 */
435 			if (error == EAGAIN)
436 				return (error);
437 		}
438 		KASSERT(inp->inp_ia == NULL);
439 		inp->inp_laddr = ia->ia_addr.sin_addr;
440 		inp->inp_ia = ia;
441 		LIST_INSERT_HEAD(&ia->ia_inpcbs, inp, inp_ialink);
442 		IFAREF(&ia->ia_ifa);
443 		inp->inp_laddr = ifaddr->sin_addr;
444 	}
445 	inp->inp_faddr = sin->sin_addr;
446 	inp->inp_fport = sin->sin_port;
447 	in_pcbstate(inp, INP_CONNECTED);
448 #ifdef IPSEC
449 	if (inp->inp_socket->so_type == SOCK_STREAM)
450 		ipsec_pcbconn(inp->inp_sp);
451 #endif
452 	return (0);
453 }
454 
455 void
456 in_pcbdisconnect(v)
457 	void *v;
458 {
459 	struct inpcb *inp = v;
460 
461 	inp->inp_faddr = zeroin_addr;
462 	inp->inp_fport = 0;
463 	in_pcbstate(inp, INP_BOUND);
464 	if (inp->inp_socket->so_state & SS_NOFDREF)
465 		in_pcbdetach(inp);
466 #ifdef IPSEC
467 	ipsec_pcbdisconn(inp->inp_sp);
468 #endif
469 }
470 
471 void
472 in_pcbdetach(v)
473 	void *v;
474 {
475 	struct inpcb *inp = v;
476 	struct socket *so = inp->inp_socket;
477 	int s;
478 
479 #ifdef IPSEC
480 	ipsec4_delete_pcbpolicy(inp);
481 #endif /*IPSEC*/
482 	so->so_pcb = 0;
483 	sofree(so);
484 	if (inp->inp_options)
485 		(void)m_free(inp->inp_options);
486 	if (inp->inp_route.ro_rt)
487 		rtfree(inp->inp_route.ro_rt);
488 	ip_freemoptions(inp->inp_moptions);
489 	if (inp->inp_ia != NULL) {
490 		LIST_REMOVE(inp, inp_ialink);
491 		IFAFREE(&inp->inp_ia->ia_ifa);
492 		inp->inp_ia = NULL;
493 	}
494 	s = splnet();
495 	in_pcbstate(inp, INP_ATTACHED);
496 	CIRCLEQ_REMOVE(&inp->inp_table->inpt_queue, inp, inp_queue);
497 	splx(s);
498 	pool_put(&inpcb_pool, inp);
499 }
500 
501 void
502 in_setsockaddr(inp, nam)
503 	struct inpcb *inp;
504 	struct mbuf *nam;
505 {
506 	struct sockaddr_in *sin;
507 
508 	nam->m_len = sizeof (*sin);
509 	sin = mtod(nam, struct sockaddr_in *);
510 	bzero((caddr_t)sin, sizeof (*sin));
511 	sin->sin_family = AF_INET;
512 	sin->sin_len = sizeof(*sin);
513 	sin->sin_port = inp->inp_lport;
514 	sin->sin_addr = inp->inp_laddr;
515 }
516 
517 void
518 in_setpeeraddr(inp, nam)
519 	struct inpcb *inp;
520 	struct mbuf *nam;
521 {
522 	struct sockaddr_in *sin;
523 
524 	nam->m_len = sizeof (*sin);
525 	sin = mtod(nam, struct sockaddr_in *);
526 	bzero((caddr_t)sin, sizeof (*sin));
527 	sin->sin_family = AF_INET;
528 	sin->sin_len = sizeof(*sin);
529 	sin->sin_port = inp->inp_fport;
530 	sin->sin_addr = inp->inp_faddr;
531 }
532 
533 /*
534  * Pass some notification to all connections of a protocol
535  * associated with address dst.  The local address and/or port numbers
536  * may be specified to limit the search.  The "usual action" will be
537  * taken, depending on the ctlinput cmd.  The caller must filter any
538  * cmds that are uninteresting (e.g., no error in the map).
539  * Call the protocol specific routine (if any) to report
540  * any errors for each matching socket.
541  *
542  * Must be called at splsoftnet.
543  */
544 int
545 in_pcbnotify(table, faddr, fport_arg, laddr, lport_arg, errno, notify)
546 	struct inpcbtable *table;
547 	struct in_addr faddr, laddr;
548 	u_int fport_arg, lport_arg;
549 	int errno;
550 	void (*notify) __P((struct inpcb *, int));
551 {
552 	struct inpcbhead *head;
553 	struct inpcb *inp, *ninp;
554 	u_int16_t fport = fport_arg, lport = lport_arg;
555 	int nmatch;
556 
557 	if (in_nullhost(faddr) || notify == 0)
558 		return (0);
559 
560 	nmatch = 0;
561 	head = INPCBHASH_CONNECT(table, faddr, fport, laddr, lport);
562 	for (inp = LIST_FIRST(head); inp != NULL; inp = ninp) {
563 		ninp = LIST_NEXT(inp, inp_hash);
564 		if (in_hosteq(inp->inp_faddr, faddr) &&
565 		    inp->inp_fport == fport &&
566 		    inp->inp_lport == lport &&
567 		    in_hosteq(inp->inp_laddr, laddr)) {
568 			(*notify)(inp, errno);
569 			nmatch++;
570 		}
571 	}
572 	return (nmatch);
573 }
574 
575 void
576 in_pcbnotifyall(table, faddr, errno, notify)
577 	struct inpcbtable *table;
578 	struct in_addr faddr;
579 	int errno;
580 	void (*notify) __P((struct inpcb *, int));
581 {
582 	struct inpcb *inp, *ninp;
583 
584 	if (in_nullhost(faddr) || notify == 0)
585 		return;
586 
587 	for (inp = CIRCLEQ_FIRST(&table->inpt_queue);
588 	    inp != (void *)&table->inpt_queue;
589 	    inp = ninp) {
590 		ninp = CIRCLEQ_NEXT(inp, inp_queue);
591 		if (in_hosteq(inp->inp_faddr, faddr))
592 			(*notify)(inp, errno);
593 	}
594 }
595 
596 void
597 in_pcbpurgeif0(table, ifp)
598 	struct inpcbtable *table;
599 	struct ifnet *ifp;
600 {
601 	struct inpcb *inp, *ninp;
602 	struct ip_moptions *imo;
603 	int i, gap;
604 
605 	for (inp = CIRCLEQ_FIRST(&table->inpt_queue);
606 	    inp != (void *)&table->inpt_queue;
607 	    inp = ninp) {
608 		ninp = CIRCLEQ_NEXT(inp, inp_queue);
609 		imo = inp->inp_moptions;
610 		if (imo != NULL) {
611 			/*
612 			 * Unselect the outgoing interface if it is being
613 			 * detached.
614 			 */
615 			if (imo->imo_multicast_ifp == ifp)
616 				imo->imo_multicast_ifp = NULL;
617 
618 			/*
619 			 * Drop multicast group membership if we joined
620 			 * through the interface being detached.
621 			 */
622 			for (i = 0, gap = 0; i < imo->imo_num_memberships;
623 			    i++) {
624 				if (imo->imo_membership[i]->inm_ifp == ifp) {
625 					in_delmulti(imo->imo_membership[i]);
626 					gap++;
627 				} else if (gap != 0)
628 					imo->imo_membership[i - gap] =
629 					    imo->imo_membership[i];
630 			}
631 			imo->imo_num_memberships -= gap;
632 		}
633 	}
634 }
635 
636 void
637 in_pcbpurgeif(table, ifp)
638 	struct inpcbtable *table;
639 	struct ifnet *ifp;
640 {
641 	struct inpcb *inp, *ninp;
642 
643 	for (inp = CIRCLEQ_FIRST(&table->inpt_queue);
644 	    inp != (void *)&table->inpt_queue;
645 	    inp = ninp) {
646 		ninp = CIRCLEQ_NEXT(inp, inp_queue);
647 		if (inp->inp_route.ro_rt != NULL &&
648 		    inp->inp_route.ro_rt->rt_ifp == ifp)
649 			in_rtchange(inp, 0);
650 	}
651 }
652 
653 /*
654  * Check for alternatives when higher level complains
655  * about service problems.  For now, invalidate cached
656  * routing information.  If the route was created dynamically
657  * (by a redirect), time to try a default gateway again.
658  */
659 void
660 in_losing(inp)
661 	struct inpcb *inp;
662 {
663 	struct rtentry *rt;
664 	struct rt_addrinfo info;
665 
666 	if ((rt = inp->inp_route.ro_rt)) {
667 		inp->inp_route.ro_rt = 0;
668 		bzero((caddr_t)&info, sizeof(info));
669 		info.rti_info[RTAX_DST] = &inp->inp_route.ro_dst;
670 		info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
671 		info.rti_info[RTAX_NETMASK] = rt_mask(rt);
672 		rt_missmsg(RTM_LOSING, &info, rt->rt_flags, 0);
673 		if (rt->rt_flags & RTF_DYNAMIC)
674 			(void) rtrequest(RTM_DELETE, rt_key(rt),
675 				rt->rt_gateway, rt_mask(rt), rt->rt_flags,
676 				(struct rtentry **)0);
677 		else
678 		/*
679 		 * A new route can be allocated
680 		 * the next time output is attempted.
681 		 */
682 			rtfree(rt);
683 	}
684 }
685 
686 /*
687  * After a routing change, flush old routing
688  * and allocate a (hopefully) better one.
689  */
690 void
691 in_rtchange(inp, errno)
692 	struct inpcb *inp;
693 	int errno;
694 {
695 
696 	if (inp->inp_route.ro_rt) {
697 		rtfree(inp->inp_route.ro_rt);
698 		inp->inp_route.ro_rt = 0;
699 		/*
700 		 * A new route can be allocated the next time
701 		 * output is attempted.
702 		 */
703 	}
704 	/* XXX SHOULD NOTIFY HIGHER-LEVEL PROTOCOLS */
705 }
706 
707 struct inpcb *
708 in_pcblookup_port(table, laddr, lport_arg, lookup_wildcard)
709 	struct inpcbtable *table;
710 	struct in_addr laddr;
711 	u_int lport_arg;
712 	int lookup_wildcard;
713 {
714 	struct inpcb *inp, *match = 0;
715 	int matchwild = 3, wildcard;
716 	u_int16_t lport = lport_arg;
717 
718 	CIRCLEQ_FOREACH(inp, &table->inpt_queue, inp_queue) {
719 		if (inp->inp_lport != lport)
720 			continue;
721 		wildcard = 0;
722 		if (!in_nullhost(inp->inp_faddr))
723 			wildcard++;
724 		if (in_nullhost(inp->inp_laddr)) {
725 			if (!in_nullhost(laddr))
726 				wildcard++;
727 		} else {
728 			if (in_nullhost(laddr))
729 				wildcard++;
730 			else {
731 				if (!in_hosteq(inp->inp_laddr, laddr))
732 					continue;
733 			}
734 		}
735 		if (wildcard && !lookup_wildcard)
736 			continue;
737 		if (wildcard < matchwild) {
738 			match = inp;
739 			matchwild = wildcard;
740 			if (matchwild == 0)
741 				break;
742 		}
743 	}
744 	return (match);
745 }
746 
747 #ifdef DIAGNOSTIC
748 int	in_pcbnotifymiss = 0;
749 #endif
750 
751 struct inpcb *
752 in_pcblookup_connect(table, faddr, fport_arg, laddr, lport_arg)
753 	struct inpcbtable *table;
754 	struct in_addr faddr, laddr;
755 	u_int fport_arg, lport_arg;
756 {
757 	struct inpcbhead *head;
758 	struct inpcb *inp;
759 	u_int16_t fport = fport_arg, lport = lport_arg;
760 
761 	head = INPCBHASH_CONNECT(table, faddr, fport, laddr, lport);
762 	LIST_FOREACH(inp, head, inp_hash) {
763 		if (in_hosteq(inp->inp_faddr, faddr) &&
764 		    inp->inp_fport == fport &&
765 		    inp->inp_lport == lport &&
766 		    in_hosteq(inp->inp_laddr, laddr))
767 			goto out;
768 	}
769 #ifdef DIAGNOSTIC
770 	if (in_pcbnotifymiss) {
771 		printf("in_pcblookup_connect: faddr=%08x fport=%d laddr=%08x lport=%d\n",
772 		    ntohl(faddr.s_addr), ntohs(fport),
773 		    ntohl(laddr.s_addr), ntohs(lport));
774 	}
775 #endif
776 	return (0);
777 
778 out:
779 	/* Move this PCB to the head of hash chain. */
780 	if (inp != LIST_FIRST(head)) {
781 		LIST_REMOVE(inp, inp_hash);
782 		LIST_INSERT_HEAD(head, inp, inp_hash);
783 	}
784 	return (inp);
785 }
786 
787 struct inpcb *
788 in_pcblookup_bind(table, laddr, lport_arg)
789 	struct inpcbtable *table;
790 	struct in_addr laddr;
791 	u_int lport_arg;
792 {
793 	struct inpcbhead *head;
794 	struct inpcb *inp;
795 	u_int16_t lport = lport_arg;
796 
797 	head = INPCBHASH_BIND(table, laddr, lport);
798 	LIST_FOREACH(inp, head, inp_hash) {
799 		if (inp->inp_lport == lport &&
800 		    in_hosteq(inp->inp_laddr, laddr))
801 			goto out;
802 	}
803 	head = INPCBHASH_BIND(table, zeroin_addr, lport);
804 	LIST_FOREACH(inp, head, inp_hash) {
805 		if (inp->inp_lport == lport &&
806 		    in_hosteq(inp->inp_laddr, zeroin_addr))
807 			goto out;
808 	}
809 #ifdef DIAGNOSTIC
810 	if (in_pcbnotifymiss) {
811 		printf("in_pcblookup_bind: laddr=%08x lport=%d\n",
812 		    ntohl(laddr.s_addr), ntohs(lport));
813 	}
814 #endif
815 	return (0);
816 
817 out:
818 	/* Move this PCB to the head of hash chain. */
819 	if (inp != LIST_FIRST(head)) {
820 		LIST_REMOVE(inp, inp_hash);
821 		LIST_INSERT_HEAD(head, inp, inp_hash);
822 	}
823 	return (inp);
824 }
825 
826 void
827 in_pcbstate(inp, state)
828 	struct inpcb *inp;
829 	int state;
830 {
831 
832 	if (inp->inp_state > INP_ATTACHED)
833 		LIST_REMOVE(inp, inp_hash);
834 
835 	switch (state) {
836 	case INP_BOUND:
837 		LIST_INSERT_HEAD(INPCBHASH_BIND(inp->inp_table,
838 		    inp->inp_laddr, inp->inp_lport), inp, inp_hash);
839 		break;
840 	case INP_CONNECTED:
841 		LIST_INSERT_HEAD(INPCBHASH_CONNECT(inp->inp_table,
842 		    inp->inp_faddr, inp->inp_fport,
843 		    inp->inp_laddr, inp->inp_lport), inp, inp_hash);
844 		break;
845 	}
846 
847 	inp->inp_state = state;
848 }
849 
850 struct rtentry *
851 in_pcbrtentry(inp)
852 	struct inpcb *inp;
853 {
854 	struct route *ro;
855 
856 	ro = &inp->inp_route;
857 
858 	if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
859 	    !in_hosteq(satosin(&ro->ro_dst)->sin_addr, inp->inp_faddr))) {
860 		RTFREE(ro->ro_rt);
861 		ro->ro_rt = (struct rtentry *)NULL;
862 	}
863 	if (ro->ro_rt == (struct rtentry *)NULL &&
864 	    !in_nullhost(inp->inp_faddr)) {
865 		bzero(&ro->ro_dst, sizeof(struct sockaddr_in));
866 		ro->ro_dst.sa_family = AF_INET;
867 		ro->ro_dst.sa_len = sizeof(ro->ro_dst);
868 		satosin(&ro->ro_dst)->sin_addr = inp->inp_faddr;
869 		rtalloc(ro);
870 	}
871 	return (ro->ro_rt);
872 }
873 
874 struct sockaddr_in *
875 in_selectsrc(sin, ro, soopts, mopts, errorp)
876 	struct sockaddr_in *sin;
877 	struct route *ro;
878 	int soopts;
879 	struct ip_moptions *mopts;
880 	int *errorp;
881 {
882 	struct in_ifaddr *ia;
883 
884 	ia = (struct in_ifaddr *)0;
885 	/*
886 	 * If route is known or can be allocated now,
887 	 * our src addr is taken from the i/f, else punt.
888 	 * Note that we should check the address family of the cached
889 	 * destination, in case of sharing the cache with IPv6.
890 	 */
891 	if (ro->ro_rt &&
892 	    (ro->ro_dst.sa_family != AF_INET ||
893 	    !in_hosteq(satosin(&ro->ro_dst)->sin_addr, sin->sin_addr) ||
894 	    soopts & SO_DONTROUTE)) {
895 		RTFREE(ro->ro_rt);
896 		ro->ro_rt = (struct rtentry *)0;
897 	}
898 	if ((soopts & SO_DONTROUTE) == 0 && /*XXX*/
899 	    (ro->ro_rt == (struct rtentry *)0 ||
900 	     ro->ro_rt->rt_ifp == (struct ifnet *)0)) {
901 		/* No route yet, so try to acquire one */
902 		bzero(&ro->ro_dst, sizeof(struct sockaddr_in));
903 		ro->ro_dst.sa_family = AF_INET;
904 		ro->ro_dst.sa_len = sizeof(struct sockaddr_in);
905 		satosin(&ro->ro_dst)->sin_addr = sin->sin_addr;
906 		rtalloc(ro);
907 	}
908 	/*
909 	 * If we found a route, use the address
910 	 * corresponding to the outgoing interface
911 	 * unless it is the loopback (in case a route
912 	 * to our address on another net goes to loopback).
913 	 *
914 	 * XXX Is this still true?  Do we care?
915 	 */
916 	if (ro->ro_rt && !(ro->ro_rt->rt_ifp->if_flags & IFF_LOOPBACK))
917 		ia = ifatoia(ro->ro_rt->rt_ifa);
918 	if (ia == NULL) {
919 		u_int16_t fport = sin->sin_port;
920 
921 		sin->sin_port = 0;
922 		ia = ifatoia(ifa_ifwithladdr(sintosa(sin)));
923 		sin->sin_port = fport;
924 		if (ia == 0) {
925 			/* Find 1st non-loopback AF_INET address */
926 			TAILQ_FOREACH(ia, &in_ifaddr, ia_list) {
927 				if (!(ia->ia_ifp->if_flags & IFF_LOOPBACK))
928 					break;
929 			}
930 		}
931 		if (ia == NULL) {
932 			*errorp = EADDRNOTAVAIL;
933 			return NULL;
934 		}
935 	}
936 	/*
937 	 * If the destination address is multicast and an outgoing
938 	 * interface has been set as a multicast option, use the
939 	 * address of that interface as our source address.
940 	 */
941 	if (IN_MULTICAST(sin->sin_addr.s_addr) && mopts != NULL) {
942 		struct ip_moptions *imo;
943 		struct ifnet *ifp;
944 
945 		imo = mopts;
946 		if (imo->imo_multicast_ifp != NULL) {
947 			ifp = imo->imo_multicast_ifp;
948 			IFP_TO_IA(ifp, ia);		/* XXX */
949 			if (ia == 0) {
950 				*errorp = EADDRNOTAVAIL;
951 				return NULL;
952 			}
953 		}
954 	}
955 	return satosin(&ia->ia_addr);
956 }
957