xref: /netbsd-src/sys/netinet/in_pcb.c (revision 481fca6e59249d8ffcf24fef7cfbe7b131bfb080)
1 /*	$NetBSD: in_pcb.c,v 1.66 2000/07/06 12:51:39 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 "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 
119 #include <net/if.h>
120 #include <net/route.h>
121 
122 #include <netinet/in.h>
123 #include <netinet/in_systm.h>
124 #include <netinet/ip.h>
125 #include <netinet/in_pcb.h>
126 #include <netinet/in_var.h>
127 #include <netinet/ip_var.h>
128 
129 #ifdef IPSEC
130 #include <netinet6/ipsec.h>
131 #include <netkey/key.h>
132 #endif /* IPSEC */
133 
134 struct	in_addr zeroin_addr;
135 
136 #define	INPCBHASH_BIND(table, laddr, lport) \
137 	&(table)->inpt_bindhashtbl[ \
138 	    ((ntohl((laddr).s_addr) + ntohs(lport))) & (table)->inpt_bindhash]
139 #define	INPCBHASH_CONNECT(table, faddr, fport, laddr, lport) \
140 	&(table)->inpt_connecthashtbl[ \
141 	    ((ntohl((faddr).s_addr) + ntohs(fport)) + \
142 	     (ntohl((laddr).s_addr) + ntohs(lport))) & (table)->inpt_connecthash]
143 
144 struct inpcb *
145 	in_pcblookup_port __P((struct inpcbtable *,
146 	    struct in_addr, u_int, int));
147 
148 int	anonportmin = IPPORT_ANONMIN;
149 int	anonportmax = IPPORT_ANONMAX;
150 
151 struct pool inpcb_pool;
152 
153 void
154 in_pcbinit(table, bindhashsize, connecthashsize)
155 	struct inpcbtable *table;
156 	int bindhashsize, connecthashsize;
157 {
158 	static int inpcb_pool_initialized;
159 
160 	if (inpcb_pool_initialized == 0) {
161 		pool_init(&inpcb_pool, sizeof(struct inpcb), 0, 0, 0,
162 		    "inpcbpl", 0, NULL, NULL, M_PCB);
163 		inpcb_pool_initialized = 1;
164 	}
165 
166 	CIRCLEQ_INIT(&table->inpt_queue);
167 	table->inpt_bindhashtbl =
168 	    hashinit(bindhashsize, M_PCB, M_WAITOK, &table->inpt_bindhash);
169 	table->inpt_connecthashtbl =
170 	    hashinit(connecthashsize, M_PCB, M_WAITOK, &table->inpt_connecthash);
171 	table->inpt_lastlow = IPPORT_RESERVEDMAX;
172 	table->inpt_lastport = (u_int16_t)anonportmax;
173 }
174 
175 int
176 in_pcballoc(so, v)
177 	struct socket *so;
178 	void *v;
179 {
180 	struct inpcbtable *table = v;
181 	struct inpcb *inp;
182 	int s;
183 
184 	inp = pool_get(&inpcb_pool, PR_NOWAIT);
185 	if (inp == NULL)
186 		return (ENOBUFS);
187 	bzero((caddr_t)inp, sizeof(*inp));
188 	inp->inp_table = table;
189 	inp->inp_socket = so;
190 	inp->inp_errormtu = -1;
191 	so->so_pcb = inp;
192 	s = splnet();
193 	CIRCLEQ_INSERT_HEAD(&table->inpt_queue, inp, inp_queue);
194 	in_pcbstate(inp, INP_ATTACHED);
195 	splx(s);
196 	return (0);
197 }
198 
199 int
200 in_pcbbind(v, nam, p)
201 	void *v;
202 	struct mbuf *nam;
203 	struct proc *p;
204 {
205 	struct inpcb *inp = v;
206 	struct socket *so = inp->inp_socket;
207 	struct inpcbtable *table = inp->inp_table;
208 	struct sockaddr_in *sin;
209 	u_int16_t lport = 0;
210 	int wild = 0, reuseport = (so->so_options & SO_REUSEPORT);
211 #ifndef IPNOPRIVPORTS
212 	int error;
213 #endif
214 
215 	if (in_ifaddr.tqh_first == 0)
216 		return (EADDRNOTAVAIL);
217 	if (inp->inp_lport || !in_nullhost(inp->inp_laddr))
218 		return (EINVAL);
219 	if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0)
220 		wild = 1;
221 	if (nam == 0)
222 		goto noname;
223 	sin = mtod(nam, struct sockaddr_in *);
224 	if (nam->m_len != sizeof (*sin))
225 		return (EINVAL);
226 #ifdef notdef
227 	/*
228 	 * We should check the family, but old programs
229 	 * incorrectly fail to initialize it.
230 	 */
231 	if (sin->sin_family != AF_INET)
232 		return (EAFNOSUPPORT);
233 #endif
234 	lport = sin->sin_port;
235 	if (IN_MULTICAST(sin->sin_addr.s_addr)) {
236 		/*
237 		 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
238 		 * allow complete duplication of binding if
239 		 * SO_REUSEPORT is set, or if SO_REUSEADDR is set
240 		 * and a multicast address is bound on both
241 		 * new and duplicated sockets.
242 		 */
243 		if (so->so_options & SO_REUSEADDR)
244 			reuseport = SO_REUSEADDR|SO_REUSEPORT;
245 	} else if (!in_nullhost(sin->sin_addr)) {
246 		sin->sin_port = 0;		/* yech... */
247 		if (ifa_ifwithaddr(sintosa(sin)) == 0)
248 			return (EADDRNOTAVAIL);
249 	}
250 	if (lport) {
251 		struct inpcb *t;
252 #ifndef IPNOPRIVPORTS
253 		/* GROSS */
254 		if (ntohs(lport) < IPPORT_RESERVED &&
255 		    (p == 0 || (error = suser(p->p_ucred, &p->p_acflag))))
256 			return (EACCES);
257 #endif
258 		if (so->so_uid && !IN_MULTICAST(sin->sin_addr.s_addr)) {
259 			t = in_pcblookup_port(table, sin->sin_addr, lport, 1);
260 		/*
261 		 * XXX:	investigate ramifications of loosening this
262 		 *	restriction so that as long as both ports have
263 		 *	SO_REUSEPORT allow the bind
264 		 */
265 			if (t &&
266 			    (!in_nullhost(sin->sin_addr) ||
267 			     !in_nullhost(t->inp_laddr) ||
268 			     (t->inp_socket->so_options & SO_REUSEPORT) == 0)
269 			    && (so->so_uid != t->inp_socket->so_uid)) {
270 				return (EADDRINUSE);
271 			}
272 		}
273 		t = in_pcblookup_port(table, sin->sin_addr, lport, wild);
274 		if (t && (reuseport & t->inp_socket->so_options) == 0)
275 			return (EADDRINUSE);
276 	}
277 	inp->inp_laddr = sin->sin_addr;
278 
279 noname:
280 	if (lport == 0) {
281 		int	   cnt;
282 		u_int16_t  min, max;
283 		u_int16_t *lastport;
284 
285 		if (inp->inp_flags & INP_LOWPORT) {
286 #ifndef IPNOPRIVPORTS
287 			if (p == 0 || (error = suser(p->p_ucred, &p->p_acflag)))
288 				return (EACCES);
289 #endif
290 			min = IPPORT_RESERVEDMIN;
291 			max = IPPORT_RESERVEDMAX;
292 			lastport = &table->inpt_lastlow;
293 		} else {
294 			min = anonportmin;
295 			max = anonportmax;
296 			lastport = &table->inpt_lastport;
297 		}
298 		if (min > max) {	/* sanity check */
299 			u_int16_t swp;
300 
301 			swp = min;
302 			min = max;
303 			max = swp;
304 		}
305 
306 		lport = *lastport - 1;
307 		for (cnt = max - min + 1; cnt; cnt--, lport--) {
308 			if (lport < min || lport > max)
309 				lport = max;
310 			if (!in_pcblookup_port(table, inp->inp_laddr,
311 			    htons(lport), 1))
312 				goto found;
313 		}
314 		if (!in_nullhost(inp->inp_laddr))
315 			inp->inp_laddr.s_addr = INADDR_ANY;
316 		return (EAGAIN);
317 	found:
318 		inp->inp_flags |= INP_ANONPORT;
319 		*lastport = lport;
320 		lport = htons(lport);
321 	}
322 	inp->inp_lport = lport;
323 	in_pcbstate(inp, INP_BOUND);
324 	return (0);
325 }
326 
327 /*
328  * Connect from a socket to a specified address.
329  * Both address and port must be specified in argument sin.
330  * If don't have a local address for this socket yet,
331  * then pick one.
332  */
333 int
334 in_pcbconnect(v, nam)
335 	void *v;
336 	struct mbuf *nam;
337 {
338 	struct inpcb *inp = v;
339 	struct in_ifaddr *ia;
340 	struct sockaddr_in *ifaddr = NULL;
341 	struct sockaddr_in *sin = mtod(nam, struct sockaddr_in *);
342 	int error;
343 
344 	if (nam->m_len != sizeof (*sin))
345 		return (EINVAL);
346 	if (sin->sin_family != AF_INET)
347 		return (EAFNOSUPPORT);
348 	if (sin->sin_port == 0)
349 		return (EADDRNOTAVAIL);
350 	if (in_ifaddr.tqh_first != 0) {
351 		/*
352 		 * If the destination address is INADDR_ANY,
353 		 * use any local address (likely loopback).
354 		 * If the supplied address is INADDR_BROADCAST,
355 		 * use the broadcast address of an interface
356 		 * which supports broadcast. (loopback does not)
357 		 */
358 
359 		if (in_nullhost(sin->sin_addr))
360 			sin->sin_addr = in_ifaddr.tqh_first->ia_addr.sin_addr;
361 		else if (sin->sin_addr.s_addr == INADDR_BROADCAST)
362 		    for (ia = in_ifaddr.tqh_first; ia != NULL;
363 		      ia = ia->ia_list.tqe_next)
364 			if (ia->ia_ifp->if_flags & IFF_BROADCAST) {
365 			    sin->sin_addr = ia->ia_broadaddr.sin_addr;
366 			    break;
367 			}
368 	}
369 	/*
370 	 * If we haven't bound which network number to use as ours,
371 	 * we will use the number of the outgoing interface.
372 	 * This depends on having done a routing lookup, which
373 	 * we will probably have to do anyway, so we might
374 	 * as well do it now.  On the other hand if we are
375 	 * sending to multiple destinations we may have already
376 	 * done the lookup, so see if we can use the route
377 	 * from before.  In any case, we only
378 	 * chose a port number once, even if sending to multiple
379 	 * destinations.
380 	 */
381 	if (in_nullhost(inp->inp_laddr)) {
382 #if 0
383 		struct route *ro;
384 
385 		ia = (struct in_ifaddr *)0;
386 		/*
387 		 * If route is known or can be allocated now,
388 		 * our src addr is taken from the i/f, else punt.
389 		 */
390 		ro = &inp->inp_route;
391 		if (ro->ro_rt &&
392 		    (!in_hosteq(satosin(&ro->ro_dst)->sin_addr,
393 			sin->sin_addr) ||
394 		    inp->inp_socket->so_options & SO_DONTROUTE)) {
395 			RTFREE(ro->ro_rt);
396 			ro->ro_rt = (struct rtentry *)0;
397 		}
398 		if ((inp->inp_socket->so_options & SO_DONTROUTE) == 0 && /*XXX*/
399 		    (ro->ro_rt == (struct rtentry *)0 ||
400 		    ro->ro_rt->rt_ifp == (struct ifnet *)0)) {
401 			/* No route yet, so try to acquire one */
402 			ro->ro_dst.sa_family = AF_INET;
403 			ro->ro_dst.sa_len = sizeof(struct sockaddr_in);
404 			satosin(&ro->ro_dst)->sin_addr = sin->sin_addr;
405 			rtalloc(ro);
406 		}
407 		/*
408 		 * If we found a route, use the address
409 		 * corresponding to the outgoing interface
410 		 * unless it is the loopback (in case a route
411 		 * to our address on another net goes to loopback).
412 		 *
413 		 * XXX Is this still true?  Do we care?
414 		 */
415 		if (ro->ro_rt && !(ro->ro_rt->rt_ifp->if_flags & IFF_LOOPBACK))
416 			ia = ifatoia(ro->ro_rt->rt_ifa);
417 		if (ia == NULL) {
418 			u_int16_t fport = sin->sin_port;
419 
420 			sin->sin_port = 0;
421 			ia = ifatoia(ifa_ifwithladdr(sintosa(sin)));
422 			sin->sin_port = fport;
423 			if (ia == 0) {
424 				/* Find 1st non-loopback AF_INET address */
425 				for (ia = in_ifaddr.tqh_first ; ia != NULL;
426 				     ia = ia->ia_list.tqe_next) {
427 					if ((ia->ia_ifp->if_flags &
428 					     IFF_LOOPBACK) == 0)
429 						break;
430 				}
431 			}
432 			if (ia == NULL)
433 				return (EADDRNOTAVAIL);
434 		}
435 		/*
436 		 * If the destination address is multicast and an outgoing
437 		 * interface has been set as a multicast option, use the
438 		 * address of that interface as our source address.
439 		 */
440 		if (IN_MULTICAST(sin->sin_addr.s_addr) &&
441 		    inp->inp_moptions != NULL) {
442 			struct ip_moptions *imo;
443 			struct ifnet *ifp;
444 
445 			imo = inp->inp_moptions;
446 			if (imo->imo_multicast_ifp != NULL) {
447 				ifp = imo->imo_multicast_ifp;
448 				IFP_TO_IA(ifp, ia);		/* XXX */
449 				if (ia == 0)
450 					return (EADDRNOTAVAIL);
451 			}
452 		}
453 		ifaddr = satosin(&ia->ia_addr);
454 #else
455 		int error;
456 		ifaddr = in_selectsrc(sin, &inp->inp_route,
457 			inp->inp_socket->so_options, inp->inp_moptions, &error);
458 		if (ifaddr == NULL) {
459 			if (error == 0)
460 				error = EADDRNOTAVAIL;
461 			return error;
462 		}
463 #endif
464 	}
465 	if (in_pcblookup_connect(inp->inp_table, sin->sin_addr, sin->sin_port,
466 	    !in_nullhost(inp->inp_laddr) ? inp->inp_laddr : ifaddr->sin_addr,
467 	    inp->inp_lport) != 0)
468 		return (EADDRINUSE);
469 	if (in_nullhost(inp->inp_laddr)) {
470 		if (inp->inp_lport == 0) {
471 			error = in_pcbbind(inp, (struct mbuf *)0,
472 			    (struct proc *)0);
473 			/*
474 			 * This used to ignore the return value
475 			 * completely, but we need to check for
476 			 * ephemeral port shortage.
477 			 * XXX Should we check for other errors, too?
478 			 */
479 			if (error == EAGAIN)
480 				return (error);
481 		}
482 		inp->inp_laddr = ifaddr->sin_addr;
483 	}
484 	inp->inp_faddr = sin->sin_addr;
485 	inp->inp_fport = sin->sin_port;
486 	in_pcbstate(inp, INP_CONNECTED);
487 	return (0);
488 }
489 
490 void
491 in_pcbdisconnect(v)
492 	void *v;
493 {
494 	struct inpcb *inp = v;
495 
496 	inp->inp_faddr = zeroin_addr;
497 	inp->inp_fport = 0;
498 	in_pcbstate(inp, INP_BOUND);
499 	if (inp->inp_socket->so_state & SS_NOFDREF)
500 		in_pcbdetach(inp);
501 }
502 
503 void
504 in_pcbdetach(v)
505 	void *v;
506 {
507 	struct inpcb *inp = v;
508 	struct socket *so = inp->inp_socket;
509 	int s;
510 
511 #ifdef IPSEC
512 	ipsec4_delete_pcbpolicy(inp);
513 #endif /*IPSEC*/
514 	so->so_pcb = 0;
515 	sofree(so);
516 	if (inp->inp_options)
517 		(void)m_free(inp->inp_options);
518 	if (inp->inp_route.ro_rt)
519 		rtfree(inp->inp_route.ro_rt);
520 	ip_freemoptions(inp->inp_moptions);
521 	s = splnet();
522 	in_pcbstate(inp, INP_ATTACHED);
523 	CIRCLEQ_REMOVE(&inp->inp_table->inpt_queue, inp, inp_queue);
524 	splx(s);
525 	pool_put(&inpcb_pool, inp);
526 }
527 
528 void
529 in_setsockaddr(inp, nam)
530 	struct inpcb *inp;
531 	struct mbuf *nam;
532 {
533 	struct sockaddr_in *sin;
534 
535 	nam->m_len = sizeof (*sin);
536 	sin = mtod(nam, struct sockaddr_in *);
537 	bzero((caddr_t)sin, sizeof (*sin));
538 	sin->sin_family = AF_INET;
539 	sin->sin_len = sizeof(*sin);
540 	sin->sin_port = inp->inp_lport;
541 	sin->sin_addr = inp->inp_laddr;
542 }
543 
544 void
545 in_setpeeraddr(inp, nam)
546 	struct inpcb *inp;
547 	struct mbuf *nam;
548 {
549 	struct sockaddr_in *sin;
550 
551 	nam->m_len = sizeof (*sin);
552 	sin = mtod(nam, struct sockaddr_in *);
553 	bzero((caddr_t)sin, sizeof (*sin));
554 	sin->sin_family = AF_INET;
555 	sin->sin_len = sizeof(*sin);
556 	sin->sin_port = inp->inp_fport;
557 	sin->sin_addr = inp->inp_faddr;
558 }
559 
560 /*
561  * Pass some notification to all connections of a protocol
562  * associated with address dst.  The local address and/or port numbers
563  * may be specified to limit the search.  The "usual action" will be
564  * taken, depending on the ctlinput cmd.  The caller must filter any
565  * cmds that are uninteresting (e.g., no error in the map).
566  * Call the protocol specific routine (if any) to report
567  * any errors for each matching socket.
568  *
569  * Must be called at splsoftnet.
570  */
571 int
572 in_pcbnotify(table, faddr, fport_arg, laddr, lport_arg, errno, notify)
573 	struct inpcbtable *table;
574 	struct in_addr faddr, laddr;
575 	u_int fport_arg, lport_arg;
576 	int errno;
577 	void (*notify) __P((struct inpcb *, int));
578 {
579 	struct inpcbhead *head;
580 	struct inpcb *inp, *ninp;
581 	u_int16_t fport = fport_arg, lport = lport_arg;
582 	int nmatch;
583 
584 	if (in_nullhost(faddr) || notify == 0)
585 		return (0);
586 
587 	nmatch = 0;
588 	head = INPCBHASH_CONNECT(table, faddr, fport, laddr, lport);
589 	for (inp = head->lh_first; inp != NULL; inp = ninp) {
590 		ninp = inp->inp_hash.le_next;
591 		if (in_hosteq(inp->inp_faddr, faddr) &&
592 		    inp->inp_fport == fport &&
593 		    inp->inp_lport == lport &&
594 		    in_hosteq(inp->inp_laddr, laddr)) {
595 			(*notify)(inp, errno);
596 			nmatch++;
597 		}
598 	}
599 	return (nmatch);
600 }
601 
602 void
603 in_pcbnotifyall(table, faddr, errno, notify)
604 	struct inpcbtable *table;
605 	struct in_addr faddr;
606 	int errno;
607 	void (*notify) __P((struct inpcb *, int));
608 {
609 	struct inpcb *inp, *ninp;
610 
611 	if (in_nullhost(faddr) || notify == 0)
612 		return;
613 
614 	for (inp = table->inpt_queue.cqh_first;
615 	    inp != (struct inpcb *)&table->inpt_queue;
616 	    inp = ninp) {
617 		ninp = inp->inp_queue.cqe_next;
618 		if (in_hosteq(inp->inp_faddr, faddr))
619 			(*notify)(inp, errno);
620 	}
621 }
622 
623 void
624 in_pcbpurgeif(table, ifp)
625 	struct inpcbtable *table;
626 	struct ifnet *ifp;
627 {
628 	struct inpcb *inp, *ninp;
629 	struct ip_moptions *imo;
630 	int i, gap;
631 
632 	for (inp = table->inpt_queue.cqh_first;
633 	    inp != (struct inpcb *)&table->inpt_queue;
634 	    inp = ninp) {
635 		ninp = inp->inp_queue.cqe_next;
636 		if (inp->inp_route.ro_rt != NULL &&
637 		    inp->inp_route.ro_rt->rt_ifp == ifp)
638 			in_rtchange(inp, 0);
639 		imo = inp->inp_moptions;
640 		if (imo != NULL) {
641 			/*
642 			 * Unselect the outgoing interface if it is being
643 			 * detached.
644 			 */
645 			if (imo->imo_multicast_ifp == ifp)
646 				imo->imo_multicast_ifp = NULL;
647 
648 			/*
649 			 * Drop multicast group membership if we joined
650 			 * through the interface being detached.
651 			 */
652 			for (i = 0, gap = 0; i < imo->imo_num_memberships;
653 			    i++) {
654 				if (imo->imo_membership[i]->inm_ifp == ifp) {
655 					in_delmulti(imo->imo_membership[i]);
656 					gap++;
657 				} else if (gap != 0)
658 					imo->imo_membership[i - gap] =
659 					    imo->imo_membership[i];
660 			}
661 			imo->imo_num_memberships -= gap;
662 		}
663 	}
664 }
665 
666 /*
667  * Check for alternatives when higher level complains
668  * about service problems.  For now, invalidate cached
669  * routing information.  If the route was created dynamically
670  * (by a redirect), time to try a default gateway again.
671  */
672 void
673 in_losing(inp)
674 	struct inpcb *inp;
675 {
676 	struct rtentry *rt;
677 	struct rt_addrinfo info;
678 
679 	if ((rt = inp->inp_route.ro_rt)) {
680 		inp->inp_route.ro_rt = 0;
681 		bzero((caddr_t)&info, sizeof(info));
682 		info.rti_info[RTAX_DST] = &inp->inp_route.ro_dst;
683 		info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
684 		info.rti_info[RTAX_NETMASK] = rt_mask(rt);
685 		rt_missmsg(RTM_LOSING, &info, rt->rt_flags, 0);
686 		if (rt->rt_flags & RTF_DYNAMIC)
687 			(void) rtrequest(RTM_DELETE, rt_key(rt),
688 				rt->rt_gateway, rt_mask(rt), rt->rt_flags,
689 				(struct rtentry **)0);
690 		else
691 		/*
692 		 * A new route can be allocated
693 		 * the next time output is attempted.
694 		 */
695 			rtfree(rt);
696 	}
697 }
698 
699 /*
700  * After a routing change, flush old routing
701  * and allocate a (hopefully) better one.
702  */
703 void
704 in_rtchange(inp, errno)
705 	struct inpcb *inp;
706 	int errno;
707 {
708 
709 	if (inp->inp_route.ro_rt) {
710 		rtfree(inp->inp_route.ro_rt);
711 		inp->inp_route.ro_rt = 0;
712 		/*
713 		 * A new route can be allocated the next time
714 		 * output is attempted.
715 		 */
716 	}
717 	/* XXX SHOULD NOTIFY HIGHER-LEVEL PROTOCOLS */
718 }
719 
720 struct inpcb *
721 in_pcblookup_port(table, laddr, lport_arg, lookup_wildcard)
722 	struct inpcbtable *table;
723 	struct in_addr laddr;
724 	u_int lport_arg;
725 	int lookup_wildcard;
726 {
727 	struct inpcb *inp, *match = 0;
728 	int matchwild = 3, wildcard;
729 	u_int16_t lport = lport_arg;
730 
731 	for (inp = table->inpt_queue.cqh_first;
732 	    inp != (struct inpcb *)&table->inpt_queue;
733 	    inp = inp->inp_queue.cqe_next) {
734 		if (inp->inp_lport != lport)
735 			continue;
736 		wildcard = 0;
737 		if (!in_nullhost(inp->inp_faddr))
738 			wildcard++;
739 		if (in_nullhost(inp->inp_laddr)) {
740 			if (!in_nullhost(laddr))
741 				wildcard++;
742 		} else {
743 			if (in_nullhost(laddr))
744 				wildcard++;
745 			else {
746 				if (!in_hosteq(inp->inp_laddr, laddr))
747 					continue;
748 			}
749 		}
750 		if (wildcard && !lookup_wildcard)
751 			continue;
752 		if (wildcard < matchwild) {
753 			match = inp;
754 			matchwild = wildcard;
755 			if (matchwild == 0)
756 				break;
757 		}
758 	}
759 	return (match);
760 }
761 
762 #ifdef DIAGNOSTIC
763 int	in_pcbnotifymiss = 0;
764 #endif
765 
766 struct inpcb *
767 in_pcblookup_connect(table, faddr, fport_arg, laddr, lport_arg)
768 	struct inpcbtable *table;
769 	struct in_addr faddr, laddr;
770 	u_int fport_arg, lport_arg;
771 {
772 	struct inpcbhead *head;
773 	struct inpcb *inp;
774 	u_int16_t fport = fport_arg, lport = lport_arg;
775 
776 	head = INPCBHASH_CONNECT(table, faddr, fport, laddr, lport);
777 	for (inp = head->lh_first; inp != NULL; inp = inp->inp_hash.le_next) {
778 		if (in_hosteq(inp->inp_faddr, faddr) &&
779 		    inp->inp_fport == fport &&
780 		    inp->inp_lport == lport &&
781 		    in_hosteq(inp->inp_laddr, laddr))
782 			goto out;
783 	}
784 #ifdef DIAGNOSTIC
785 	if (in_pcbnotifymiss) {
786 		printf("in_pcblookup_connect: faddr=%08x fport=%d laddr=%08x lport=%d\n",
787 		    ntohl(faddr.s_addr), ntohs(fport),
788 		    ntohl(laddr.s_addr), ntohs(lport));
789 	}
790 #endif
791 	return (0);
792 
793 out:
794 	/* Move this PCB to the head of hash chain. */
795 	if (inp != head->lh_first) {
796 		LIST_REMOVE(inp, inp_hash);
797 		LIST_INSERT_HEAD(head, inp, inp_hash);
798 	}
799 	return (inp);
800 }
801 
802 struct inpcb *
803 in_pcblookup_bind(table, laddr, lport_arg)
804 	struct inpcbtable *table;
805 	struct in_addr laddr;
806 	u_int lport_arg;
807 {
808 	struct inpcbhead *head;
809 	struct inpcb *inp;
810 	u_int16_t lport = lport_arg;
811 
812 	head = INPCBHASH_BIND(table, laddr, lport);
813 	for (inp = head->lh_first; inp != NULL; inp = inp->inp_hash.le_next) {
814 		if (inp->inp_lport == lport &&
815 		    in_hosteq(inp->inp_laddr, laddr))
816 			goto out;
817 	}
818 	head = INPCBHASH_BIND(table, zeroin_addr, lport);
819 	for (inp = head->lh_first; inp != NULL; inp = inp->inp_hash.le_next) {
820 		if (inp->inp_lport == lport &&
821 		    in_hosteq(inp->inp_laddr, zeroin_addr))
822 			goto out;
823 	}
824 #ifdef DIAGNOSTIC
825 	if (in_pcbnotifymiss) {
826 		printf("in_pcblookup_bind: laddr=%08x lport=%d\n",
827 		    ntohl(laddr.s_addr), ntohs(lport));
828 	}
829 #endif
830 	return (0);
831 
832 out:
833 	/* Move this PCB to the head of hash chain. */
834 	if (inp != head->lh_first) {
835 		LIST_REMOVE(inp, inp_hash);
836 		LIST_INSERT_HEAD(head, inp, inp_hash);
837 	}
838 	return (inp);
839 }
840 
841 void
842 in_pcbstate(inp, state)
843 	struct inpcb *inp;
844 	int state;
845 {
846 
847 	if (inp->inp_state > INP_ATTACHED)
848 		LIST_REMOVE(inp, inp_hash);
849 
850 	switch (state) {
851 	case INP_BOUND:
852 		LIST_INSERT_HEAD(INPCBHASH_BIND(inp->inp_table,
853 		    inp->inp_laddr, inp->inp_lport), inp, inp_hash);
854 		break;
855 	case INP_CONNECTED:
856 		LIST_INSERT_HEAD(INPCBHASH_CONNECT(inp->inp_table,
857 		    inp->inp_faddr, inp->inp_fport,
858 		    inp->inp_laddr, inp->inp_lport), inp, inp_hash);
859 		break;
860 	}
861 
862 	inp->inp_state = state;
863 }
864 
865 struct rtentry *
866 in_pcbrtentry(inp)
867 	struct inpcb *inp;
868 {
869 	struct route *ro;
870 
871 	ro = &inp->inp_route;
872 
873 	if (ro->ro_rt == NULL) {
874 		/*
875 		 * No route yet, so try to acquire one.
876 		 */
877 		if (!in_nullhost(inp->inp_faddr)) {
878 			ro->ro_dst.sa_family = AF_INET;
879 			ro->ro_dst.sa_len = sizeof(ro->ro_dst);
880 			satosin(&ro->ro_dst)->sin_addr = inp->inp_faddr;
881 			rtalloc(ro);
882 		}
883 	}
884 	return (ro->ro_rt);
885 }
886 
887 struct sockaddr_in *
888 in_selectsrc(sin, ro, soopts, mopts, errorp)
889 	struct sockaddr_in *sin;
890 	struct route *ro;
891 	int soopts;
892 	struct ip_moptions *mopts;
893 	int *errorp;
894 {
895 	struct in_ifaddr *ia;
896 
897 	ia = (struct in_ifaddr *)0;
898 	/*
899 	 * If route is known or can be allocated now,
900 	 * our src addr is taken from the i/f, else punt.
901 	 */
902 	if (ro->ro_rt &&
903 	    (!in_hosteq(satosin(&ro->ro_dst)->sin_addr, sin->sin_addr) ||
904 	    soopts & SO_DONTROUTE)) {
905 		RTFREE(ro->ro_rt);
906 		ro->ro_rt = (struct rtentry *)0;
907 	}
908 	if ((soopts & SO_DONTROUTE) == 0 && /*XXX*/
909 	    (ro->ro_rt == (struct rtentry *)0 ||
910 	    ro->ro_rt->rt_ifp == (struct ifnet *)0)) {
911 		/* No route yet, so try to acquire one */
912 		ro->ro_dst.sa_family = AF_INET;
913 		ro->ro_dst.sa_len = sizeof(struct sockaddr_in);
914 		satosin(&ro->ro_dst)->sin_addr = sin->sin_addr;
915 		rtalloc(ro);
916 	}
917 	/*
918 	 * If we found a route, use the address
919 	 * corresponding to the outgoing interface
920 	 * unless it is the loopback (in case a route
921 	 * to our address on another net goes to loopback).
922 	 *
923 	 * XXX Is this still true?  Do we care?
924 	 */
925 	if (ro->ro_rt && !(ro->ro_rt->rt_ifp->if_flags & IFF_LOOPBACK))
926 		ia = ifatoia(ro->ro_rt->rt_ifa);
927 	if (ia == NULL) {
928 		u_int16_t fport = sin->sin_port;
929 
930 		sin->sin_port = 0;
931 		ia = ifatoia(ifa_ifwithladdr(sintosa(sin)));
932 		sin->sin_port = fport;
933 		if (ia == 0) {
934 			/* Find 1st non-loopback AF_INET address */
935 			for (ia = in_ifaddr.tqh_first;
936 			     ia != NULL;
937 			     ia = ia->ia_list.tqe_next) {
938 				if (!(ia->ia_ifp->if_flags & IFF_LOOPBACK))
939 					break;
940 			}
941 		}
942 		if (ia == NULL) {
943 			*errorp = EADDRNOTAVAIL;
944 			return NULL;
945 		}
946 	}
947 	/*
948 	 * If the destination address is multicast and an outgoing
949 	 * interface has been set as a multicast option, use the
950 	 * address of that interface as our source address.
951 	 */
952 	if (IN_MULTICAST(sin->sin_addr.s_addr) && mopts != NULL) {
953 		struct ip_moptions *imo;
954 		struct ifnet *ifp;
955 
956 		imo = mopts;
957 		if (imo->imo_multicast_ifp != NULL) {
958 			ifp = imo->imo_multicast_ifp;
959 			IFP_TO_IA(ifp, ia);		/* XXX */
960 			if (ia == 0) {
961 				*errorp = EADDRNOTAVAIL;
962 				return NULL;
963 			}
964 		}
965 	}
966 	return satosin(&ia->ia_addr);
967 }
968