xref: /netbsd-src/sys/netinet/in_pcb.c (revision 93bf6008f8b7982c1d1a9486e4a4a0e687fe36eb)
1 /*	$NetBSD: in_pcb.c,v 1.131 2009/04/14 21:25:20 elad 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 1.131 2009/04/14 21:25:20 elad 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/once.h>
110 #include <sys/pool.h>
111 #include <sys/proc.h>
112 #include <sys/kauth.h>
113 #include <sys/uidinfo.h>
114 
115 #include <net/if.h>
116 #include <net/route.h>
117 
118 #include <netinet/in.h>
119 #include <netinet/in_systm.h>
120 #include <netinet/ip.h>
121 #include <netinet/in_pcb.h>
122 #include <netinet/in_var.h>
123 #include <netinet/ip_var.h>
124 
125 #ifdef INET6
126 #include <netinet/ip6.h>
127 #include <netinet6/ip6_var.h>
128 #include <netinet6/in6_pcb.h>
129 #endif
130 
131 #ifdef IPSEC
132 #include <netinet6/ipsec.h>
133 #include <netkey/key.h>
134 #elif FAST_IPSEC
135 #include <netipsec/ipsec.h>
136 #include <netipsec/key.h>
137 #endif /* IPSEC */
138 
139 struct	in_addr zeroin_addr;
140 
141 #define	INPCBHASH_PORT(table, lport) \
142 	&(table)->inpt_porthashtbl[ntohs(lport) & (table)->inpt_porthash]
143 #define	INPCBHASH_BIND(table, laddr, lport) \
144 	&(table)->inpt_bindhashtbl[ \
145 	    ((ntohl((laddr).s_addr) + ntohs(lport))) & (table)->inpt_bindhash]
146 #define	INPCBHASH_CONNECT(table, faddr, fport, laddr, lport) \
147 	&(table)->inpt_connecthashtbl[ \
148 	    ((ntohl((faddr).s_addr) + ntohs(fport)) + \
149 	     (ntohl((laddr).s_addr) + ntohs(lport))) & (table)->inpt_connecthash]
150 
151 int	anonportmin = IPPORT_ANONMIN;
152 int	anonportmax = IPPORT_ANONMAX;
153 int	lowportmin  = IPPORT_RESERVEDMIN;
154 int	lowportmax  = IPPORT_RESERVEDMAX;
155 
156 static struct pool inpcb_pool;
157 
158 static int
159 inpcb_poolinit(void)
160 {
161 
162 	pool_init(&inpcb_pool, sizeof(struct inpcb), 0, 0, 0, "inpcbpl", NULL,
163 	    IPL_NET);
164 	return 0;
165 }
166 
167 void
168 in_pcbinit(struct inpcbtable *table, int bindhashsize, int connecthashsize)
169 {
170 	static ONCE_DECL(control);
171 
172 	CIRCLEQ_INIT(&table->inpt_queue);
173 	table->inpt_porthashtbl = hashinit(bindhashsize, HASH_LIST, true,
174 	    &table->inpt_porthash);
175 	table->inpt_bindhashtbl = hashinit(bindhashsize, HASH_LIST, true,
176 	    &table->inpt_bindhash);
177 	table->inpt_connecthashtbl = hashinit(connecthashsize, HASH_LIST, true,
178 	    &table->inpt_connecthash);
179 	table->inpt_lastlow = IPPORT_RESERVEDMAX;
180 	table->inpt_lastport = (u_int16_t)anonportmax;
181 
182 	RUN_ONCE(&control, inpcb_poolinit);
183 }
184 
185 int
186 in_pcballoc(struct socket *so, void *v)
187 {
188 	struct inpcbtable *table = v;
189 	struct inpcb *inp;
190 	int s;
191 #if defined(IPSEC) || defined(FAST_IPSEC)
192 	int error;
193 #endif
194 
195 	s = splnet();
196 	inp = pool_get(&inpcb_pool, PR_NOWAIT);
197 	splx(s);
198 	if (inp == NULL)
199 		return (ENOBUFS);
200 	memset((void *)inp, 0, sizeof(*inp));
201 	inp->inp_af = AF_INET;
202 	inp->inp_table = table;
203 	inp->inp_socket = so;
204 	inp->inp_errormtu = -1;
205 #if defined(IPSEC) || defined(FAST_IPSEC)
206 	error = ipsec_init_pcbpolicy(so, &inp->inp_sp);
207 	if (error != 0) {
208 		s = splnet();
209 		pool_put(&inpcb_pool, inp);
210 		splx(s);
211 		return error;
212 	}
213 #endif
214 	so->so_pcb = inp;
215 	s = splnet();
216 	CIRCLEQ_INSERT_HEAD(&table->inpt_queue, &inp->inp_head,
217 	    inph_queue);
218 	LIST_INSERT_HEAD(INPCBHASH_PORT(table, inp->inp_lport), &inp->inp_head,
219 	    inph_lhash);
220 	in_pcbstate(inp, INP_ATTACHED);
221 	splx(s);
222 	return (0);
223 }
224 
225 int
226 in_pcbbind(void *v, struct mbuf *nam, struct lwp *l)
227 {
228 	struct in_ifaddr *ia = NULL;
229 	struct inpcb *inp = v;
230 	struct socket *so = inp->inp_socket;
231 	struct inpcbtable *table = inp->inp_table;
232 	struct sockaddr_in *sin = NULL; /* XXXGCC */
233 	u_int16_t lport = 0;
234 	int wild = 0, reuseport = (so->so_options & SO_REUSEPORT);
235 	kauth_cred_t cred = l->l_cred;
236 
237 	if (inp->inp_af != AF_INET)
238 		return (EINVAL);
239 
240 	if (TAILQ_FIRST(&in_ifaddrhead) == 0)
241 		return (EADDRNOTAVAIL);
242 	if (inp->inp_lport || !in_nullhost(inp->inp_laddr))
243 		return (EINVAL);
244 	if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0)
245 		wild = 1;
246 	if (nam == 0)
247 		goto noname;
248 	sin = mtod(nam, struct sockaddr_in *);
249 	if (nam->m_len != sizeof (*sin))
250 		return (EINVAL);
251 	if (sin->sin_family != AF_INET)
252 		return (EAFNOSUPPORT);
253 	lport = sin->sin_port;
254 	if (IN_MULTICAST(sin->sin_addr.s_addr)) {
255 		/*
256 		 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
257 		 * allow complete duplication of binding if
258 		 * SO_REUSEPORT is set, or if SO_REUSEADDR is set
259 		 * and a multicast address is bound on both
260 		 * new and duplicated sockets.
261 		 */
262 		if (so->so_options & SO_REUSEADDR)
263 			reuseport = SO_REUSEADDR|SO_REUSEPORT;
264 	} else if (!in_nullhost(sin->sin_addr)) {
265 		INADDR_TO_IA(sin->sin_addr, ia);
266 		/* check for broadcast addresses */
267 		if (ia == NULL)
268 			ia = ifatoia(ifa_ifwithaddr(sintosa(sin)));
269 		if (ia == NULL)
270 			return (EADDRNOTAVAIL);
271 	}
272 	if (lport) {
273 		struct inpcb *t;
274 #ifdef INET6
275 		struct in6pcb *t6;
276 		struct in6_addr mapped;
277 #endif
278 #ifndef IPNOPRIVPORTS
279 		/* GROSS */
280 		if (ntohs(lport) < IPPORT_RESERVED &&
281 		    kauth_authorize_network(cred,
282 		    KAUTH_NETWORK_BIND,
283 		    KAUTH_REQ_NETWORK_BIND_PRIVPORT, so, sin,
284 		    NULL))
285 			return (EACCES);
286 #endif
287 #ifdef INET6
288 		memset(&mapped, 0, sizeof(mapped));
289 		mapped.s6_addr16[5] = 0xffff;
290 		memcpy(&mapped.s6_addr32[3], &sin->sin_addr,
291 		    sizeof(mapped.s6_addr32[3]));
292 		t6 = in6_pcblookup_port(table, &mapped, lport, wild);
293 		if (t6 && (reuseport & t6->in6p_socket->so_options) == 0)
294 			return (EADDRINUSE);
295 #endif
296 		if (so->so_uidinfo->ui_uid && !IN_MULTICAST(sin->sin_addr.s_addr)) {
297 			t = in_pcblookup_port(table, sin->sin_addr, lport, 1);
298 		/*
299 		 * XXX:	investigate ramifications of loosening this
300 		 *	restriction so that as long as both ports have
301 		 *	SO_REUSEPORT allow the bind
302 		 */
303 			if (t &&
304 			    (!in_nullhost(sin->sin_addr) ||
305 			     !in_nullhost(t->inp_laddr) ||
306 			     (t->inp_socket->so_options & SO_REUSEPORT) == 0)
307 			    && (so->so_uidinfo->ui_uid != t->inp_socket->so_uidinfo->ui_uid)) {
308 				return (EADDRINUSE);
309 			}
310 		}
311 		t = in_pcblookup_port(table, sin->sin_addr, lport, wild);
312 		if (t && (reuseport & t->inp_socket->so_options) == 0)
313 			return (EADDRINUSE);
314 	}
315 	inp->inp_laddr = sin->sin_addr;
316 
317 noname:
318 	if (lport == 0) {
319 		int	   cnt;
320 		u_int16_t  mymin, mymax;
321 		u_int16_t *lastport;
322 
323 		if (inp->inp_flags & INP_LOWPORT) {
324 #ifndef IPNOPRIVPORTS
325 			if (kauth_authorize_network(cred,
326 			    KAUTH_NETWORK_BIND,
327 			    KAUTH_REQ_NETWORK_BIND_PRIVPORT, so,
328 			    sin, NULL))
329 				return (EACCES);
330 #endif
331 			mymin = lowportmin;
332 			mymax = lowportmax;
333 			lastport = &table->inpt_lastlow;
334 		} else {
335 			mymin = anonportmin;
336 			mymax = anonportmax;
337 			lastport = &table->inpt_lastport;
338 		}
339 		if (mymin > mymax) {	/* sanity check */
340 			u_int16_t swp;
341 
342 			swp = mymin;
343 			mymin = mymax;
344 			mymax = swp;
345 		}
346 
347 		lport = *lastport - 1;
348 		for (cnt = mymax - mymin + 1; cnt; cnt--, lport--) {
349 			if (lport < mymin || lport > mymax)
350 				lport = mymax;
351 			if (!in_pcblookup_port(table, inp->inp_laddr,
352 			    htons(lport), 1))
353 				goto found;
354 		}
355 		if (!in_nullhost(inp->inp_laddr))
356 			inp->inp_laddr.s_addr = INADDR_ANY;
357 		return (EAGAIN);
358 	found:
359 		inp->inp_flags |= INP_ANONPORT;
360 		*lastport = lport;
361 		lport = htons(lport);
362 	}
363 	inp->inp_lport = lport;
364 	LIST_REMOVE(&inp->inp_head, inph_lhash);
365 	LIST_INSERT_HEAD(INPCBHASH_PORT(table, inp->inp_lport), &inp->inp_head,
366 	    inph_lhash);
367 	in_pcbstate(inp, INP_BOUND);
368 	return (0);
369 }
370 
371 /*
372  * Connect from a socket to a specified address.
373  * Both address and port must be specified in argument sin.
374  * If don't have a local address for this socket yet,
375  * then pick one.
376  */
377 int
378 in_pcbconnect(void *v, struct mbuf *nam, struct lwp *l)
379 {
380 	struct inpcb *inp = v;
381 	struct in_ifaddr *ia = NULL;
382 	struct sockaddr_in *ifaddr = NULL;
383 	struct sockaddr_in *sin = mtod(nam, struct sockaddr_in *);
384 	int error;
385 
386 	if (inp->inp_af != AF_INET)
387 		return (EINVAL);
388 
389 	if (nam->m_len != sizeof (*sin))
390 		return (EINVAL);
391 	if (sin->sin_family != AF_INET)
392 		return (EAFNOSUPPORT);
393 	if (sin->sin_port == 0)
394 		return (EADDRNOTAVAIL);
395 	if (TAILQ_FIRST(&in_ifaddrhead) != 0) {
396 		/*
397 		 * If the destination address is INADDR_ANY,
398 		 * use any local address (likely loopback).
399 		 * If the supplied address is INADDR_BROADCAST,
400 		 * use the broadcast address of an interface
401 		 * which supports broadcast. (loopback does not)
402 		 */
403 
404 		if (in_nullhost(sin->sin_addr)) {
405 			sin->sin_addr =
406 			    TAILQ_FIRST(&in_ifaddrhead)->ia_addr.sin_addr;
407 		} else if (sin->sin_addr.s_addr == INADDR_BROADCAST) {
408 			TAILQ_FOREACH(ia, &in_ifaddrhead, ia_list) {
409 				if (ia->ia_ifp->if_flags & IFF_BROADCAST) {
410 					sin->sin_addr =
411 					    ia->ia_broadaddr.sin_addr;
412 					break;
413 				}
414 			}
415 		}
416 	}
417 	/*
418 	 * If we haven't bound which network number to use as ours,
419 	 * we will use the number of the outgoing interface.
420 	 * This depends on having done a routing lookup, which
421 	 * we will probably have to do anyway, so we might
422 	 * as well do it now.  On the other hand if we are
423 	 * sending to multiple destinations we may have already
424 	 * done the lookup, so see if we can use the route
425 	 * from before.  In any case, we only
426 	 * chose a port number once, even if sending to multiple
427 	 * destinations.
428 	 */
429 	if (in_nullhost(inp->inp_laddr)) {
430 		int xerror;
431 		ifaddr = in_selectsrc(sin, &inp->inp_route,
432 		    inp->inp_socket->so_options, inp->inp_moptions, &xerror);
433 		if (ifaddr == NULL) {
434 			if (xerror == 0)
435 				xerror = EADDRNOTAVAIL;
436 			return xerror;
437 		}
438 		INADDR_TO_IA(ifaddr->sin_addr, ia);
439 		if (ia == NULL)
440 			return (EADDRNOTAVAIL);
441 	}
442 	if (in_pcblookup_connect(inp->inp_table, sin->sin_addr, sin->sin_port,
443 	    !in_nullhost(inp->inp_laddr) ? inp->inp_laddr : ifaddr->sin_addr,
444 	    inp->inp_lport) != 0)
445 		return (EADDRINUSE);
446 	if (in_nullhost(inp->inp_laddr)) {
447 		if (inp->inp_lport == 0) {
448 			error = in_pcbbind(inp, NULL, l);
449 			/*
450 			 * This used to ignore the return value
451 			 * completely, but we need to check for
452 			 * ephemeral port shortage.
453 			 * And attempts to request low ports if not root.
454 			 */
455 			if (error != 0)
456 				return (error);
457 		}
458 		inp->inp_laddr = ifaddr->sin_addr;
459 	}
460 	inp->inp_faddr = sin->sin_addr;
461 	inp->inp_fport = sin->sin_port;
462 	in_pcbstate(inp, INP_CONNECTED);
463 #if defined(IPSEC) || defined(FAST_IPSEC)
464 	if (inp->inp_socket->so_type == SOCK_STREAM)
465 		ipsec_pcbconn(inp->inp_sp);
466 #endif
467 	return (0);
468 }
469 
470 void
471 in_pcbdisconnect(void *v)
472 {
473 	struct inpcb *inp = v;
474 
475 	if (inp->inp_af != AF_INET)
476 		return;
477 
478 	inp->inp_faddr = zeroin_addr;
479 	inp->inp_fport = 0;
480 	in_pcbstate(inp, INP_BOUND);
481 #if defined(IPSEC) || defined(FAST_IPSEC)
482 	ipsec_pcbdisconn(inp->inp_sp);
483 #endif
484 	if (inp->inp_socket->so_state & SS_NOFDREF)
485 		in_pcbdetach(inp);
486 }
487 
488 void
489 in_pcbdetach(void *v)
490 {
491 	struct inpcb *inp = v;
492 	struct socket *so = inp->inp_socket;
493 	int s;
494 
495 	if (inp->inp_af != AF_INET)
496 		return;
497 
498 #if defined(IPSEC) || defined(FAST_IPSEC)
499 	ipsec4_delete_pcbpolicy(inp);
500 #endif /*IPSEC*/
501 	so->so_pcb = 0;
502 	if (inp->inp_options)
503 		(void)m_free(inp->inp_options);
504 	rtcache_free(&inp->inp_route);
505 	ip_freemoptions(inp->inp_moptions);
506 	s = splnet();
507 	in_pcbstate(inp, INP_ATTACHED);
508 	LIST_REMOVE(&inp->inp_head, inph_lhash);
509 	CIRCLEQ_REMOVE(&inp->inp_table->inpt_queue, &inp->inp_head,
510 	    inph_queue);
511 	pool_put(&inpcb_pool, inp);
512 	splx(s);
513 	sofree(so);			/* drops the socket's lock */
514 	mutex_enter(softnet_lock);	/* reacquire the softnet_lock */
515 }
516 
517 void
518 in_setsockaddr(struct inpcb *inp, struct mbuf *nam)
519 {
520 	struct sockaddr_in *sin;
521 
522 	if (inp->inp_af != AF_INET)
523 		return;
524 
525 	sin = mtod(nam, struct sockaddr_in *);
526 	sockaddr_in_init(sin, &inp->inp_laddr, inp->inp_lport);
527 	nam->m_len = sin->sin_len;
528 }
529 
530 void
531 in_setpeeraddr(struct inpcb *inp, struct mbuf *nam)
532 {
533 	struct sockaddr_in *sin;
534 
535 	if (inp->inp_af != AF_INET)
536 		return;
537 
538 	sin = mtod(nam, struct sockaddr_in *);
539 	sockaddr_in_init(sin, &inp->inp_faddr, inp->inp_fport);
540 	nam->m_len = sin->sin_len;
541 }
542 
543 /*
544  * Pass some notification to all connections of a protocol
545  * associated with address dst.  The local address and/or port numbers
546  * may be specified to limit the search.  The "usual action" will be
547  * taken, depending on the ctlinput cmd.  The caller must filter any
548  * cmds that are uninteresting (e.g., no error in the map).
549  * Call the protocol specific routine (if any) to report
550  * any errors for each matching socket.
551  *
552  * Must be called at splsoftnet.
553  */
554 int
555 in_pcbnotify(struct inpcbtable *table, struct in_addr faddr, u_int fport_arg,
556     struct in_addr laddr, u_int lport_arg, int errno,
557     void (*notify)(struct inpcb *, int))
558 {
559 	struct inpcbhead *head;
560 	struct inpcb *inp, *ninp;
561 	u_int16_t fport = fport_arg, lport = lport_arg;
562 	int nmatch;
563 
564 	if (in_nullhost(faddr) || notify == 0)
565 		return (0);
566 
567 	nmatch = 0;
568 	head = INPCBHASH_CONNECT(table, faddr, fport, laddr, lport);
569 	for (inp = (struct inpcb *)LIST_FIRST(head); inp != NULL; inp = ninp) {
570 		ninp = (struct inpcb *)LIST_NEXT(inp, inp_hash);
571 		if (inp->inp_af != AF_INET)
572 			continue;
573 		if (in_hosteq(inp->inp_faddr, faddr) &&
574 		    inp->inp_fport == fport &&
575 		    inp->inp_lport == lport &&
576 		    in_hosteq(inp->inp_laddr, laddr)) {
577 			(*notify)(inp, errno);
578 			nmatch++;
579 		}
580 	}
581 	return (nmatch);
582 }
583 
584 void
585 in_pcbnotifyall(struct inpcbtable *table, struct in_addr faddr, int errno,
586     void (*notify)(struct inpcb *, int))
587 {
588 	struct inpcb *inp, *ninp;
589 
590 	if (in_nullhost(faddr) || notify == 0)
591 		return;
592 
593 	for (inp = (struct inpcb *)CIRCLEQ_FIRST(&table->inpt_queue);
594 	    inp != (void *)&table->inpt_queue;
595 	    inp = ninp) {
596 		ninp = (struct inpcb *)CIRCLEQ_NEXT(inp, inp_queue);
597 		if (inp->inp_af != AF_INET)
598 			continue;
599 		if (in_hosteq(inp->inp_faddr, faddr))
600 			(*notify)(inp, errno);
601 	}
602 }
603 
604 void
605 in_pcbpurgeif0(struct inpcbtable *table, struct ifnet *ifp)
606 {
607 	struct inpcb *inp, *ninp;
608 	struct ip_moptions *imo;
609 	int i, gap;
610 
611 	for (inp = (struct inpcb *)CIRCLEQ_FIRST(&table->inpt_queue);
612 	    inp != (void *)&table->inpt_queue;
613 	    inp = ninp) {
614 		ninp = (struct inpcb *)CIRCLEQ_NEXT(inp, inp_queue);
615 		if (inp->inp_af != AF_INET)
616 			continue;
617 		imo = inp->inp_moptions;
618 		if (imo != NULL) {
619 			/*
620 			 * Unselect the outgoing interface if it is being
621 			 * detached.
622 			 */
623 			if (imo->imo_multicast_ifp == ifp)
624 				imo->imo_multicast_ifp = NULL;
625 
626 			/*
627 			 * Drop multicast group membership if we joined
628 			 * through the interface being detached.
629 			 */
630 			for (i = 0, gap = 0; i < imo->imo_num_memberships;
631 			    i++) {
632 				if (imo->imo_membership[i]->inm_ifp == ifp) {
633 					in_delmulti(imo->imo_membership[i]);
634 					gap++;
635 				} else if (gap != 0)
636 					imo->imo_membership[i - gap] =
637 					    imo->imo_membership[i];
638 			}
639 			imo->imo_num_memberships -= gap;
640 		}
641 	}
642 }
643 
644 void
645 in_pcbpurgeif(struct inpcbtable *table, struct ifnet *ifp)
646 {
647 	struct rtentry *rt;
648 	struct inpcb *inp, *ninp;
649 
650 	for (inp = (struct inpcb *)CIRCLEQ_FIRST(&table->inpt_queue);
651 	    inp != (void *)&table->inpt_queue;
652 	    inp = ninp) {
653 		ninp = (struct inpcb *)CIRCLEQ_NEXT(inp, inp_queue);
654 		if (inp->inp_af != AF_INET)
655 			continue;
656 		if ((rt = rtcache_validate(&inp->inp_route)) != NULL &&
657 		    rt->rt_ifp == ifp)
658 			in_rtchange(inp, 0);
659 	}
660 }
661 
662 /*
663  * Check for alternatives when higher level complains
664  * about service problems.  For now, invalidate cached
665  * routing information.  If the route was created dynamically
666  * (by a redirect), time to try a default gateway again.
667  */
668 void
669 in_losing(struct inpcb *inp)
670 {
671 	struct rtentry *rt;
672 	struct rt_addrinfo info;
673 
674 	if (inp->inp_af != AF_INET)
675 		return;
676 
677 	if ((rt = rtcache_validate(&inp->inp_route)) == NULL)
678 		return;
679 
680 	memset(&info, 0, sizeof(info));
681 	info.rti_info[RTAX_DST] = rtcache_getdst(&inp->inp_route);
682 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
683 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
684 	rt_missmsg(RTM_LOSING, &info, rt->rt_flags, 0);
685 	if (rt->rt_flags & RTF_DYNAMIC)
686 		(void) rtrequest(RTM_DELETE, rt_getkey(rt),
687 			rt->rt_gateway, rt_mask(rt), rt->rt_flags,
688 			NULL);
689 	/*
690 	 * A new route can be allocated
691 	 * the next time output is attempted.
692 	 */
693 	rtcache_free(&inp->inp_route);
694 }
695 
696 /*
697  * After a routing change, flush old routing.  A new route can be
698  * allocated the next time output is attempted.
699  */
700 void
701 in_rtchange(struct inpcb *inp, int errno)
702 {
703 
704 	if (inp->inp_af != AF_INET)
705 		return;
706 
707 	rtcache_free(&inp->inp_route);
708 
709 	/* XXX SHOULD NOTIFY HIGHER-LEVEL PROTOCOLS */
710 }
711 
712 struct inpcb *
713 in_pcblookup_port(struct inpcbtable *table, struct in_addr laddr,
714     u_int lport_arg, int lookup_wildcard)
715 {
716 	struct inpcbhead *head;
717 	struct inpcb_hdr *inph;
718 	struct inpcb *inp, *match = 0;
719 	int matchwild = 3, wildcard;
720 	u_int16_t lport = lport_arg;
721 
722 	head = INPCBHASH_PORT(table, lport);
723 	LIST_FOREACH(inph, head, inph_lhash) {
724 		inp = (struct inpcb *)inph;
725 		if (inp->inp_af != AF_INET)
726 			continue;
727 
728 		if (inp->inp_lport != lport)
729 			continue;
730 		wildcard = 0;
731 		if (!in_nullhost(inp->inp_faddr))
732 			wildcard++;
733 		if (in_nullhost(inp->inp_laddr)) {
734 			if (!in_nullhost(laddr))
735 				wildcard++;
736 		} else {
737 			if (in_nullhost(laddr))
738 				wildcard++;
739 			else {
740 				if (!in_hosteq(inp->inp_laddr, laddr))
741 					continue;
742 			}
743 		}
744 		if (wildcard && !lookup_wildcard)
745 			continue;
746 		if (wildcard < matchwild) {
747 			match = inp;
748 			matchwild = wildcard;
749 			if (matchwild == 0)
750 				break;
751 		}
752 	}
753 	return (match);
754 }
755 
756 #ifdef DIAGNOSTIC
757 int	in_pcbnotifymiss = 0;
758 #endif
759 
760 struct inpcb *
761 in_pcblookup_connect(struct inpcbtable *table,
762     struct in_addr faddr, u_int fport_arg,
763     struct in_addr laddr, u_int lport_arg)
764 {
765 	struct inpcbhead *head;
766 	struct inpcb_hdr *inph;
767 	struct inpcb *inp;
768 	u_int16_t fport = fport_arg, lport = lport_arg;
769 
770 	head = INPCBHASH_CONNECT(table, faddr, fport, laddr, lport);
771 	LIST_FOREACH(inph, head, inph_hash) {
772 		inp = (struct inpcb *)inph;
773 		if (inp->inp_af != AF_INET)
774 			continue;
775 
776 		if (in_hosteq(inp->inp_faddr, faddr) &&
777 		    inp->inp_fport == fport &&
778 		    inp->inp_lport == lport &&
779 		    in_hosteq(inp->inp_laddr, laddr))
780 			goto out;
781 	}
782 #ifdef DIAGNOSTIC
783 	if (in_pcbnotifymiss) {
784 		printf("in_pcblookup_connect: faddr=%08x fport=%d laddr=%08x lport=%d\n",
785 		    ntohl(faddr.s_addr), ntohs(fport),
786 		    ntohl(laddr.s_addr), ntohs(lport));
787 	}
788 #endif
789 	return (0);
790 
791 out:
792 	/* Move this PCB to the head of hash chain. */
793 	inph = &inp->inp_head;
794 	if (inph != LIST_FIRST(head)) {
795 		LIST_REMOVE(inph, inph_hash);
796 		LIST_INSERT_HEAD(head, inph, inph_hash);
797 	}
798 	return (inp);
799 }
800 
801 struct inpcb *
802 in_pcblookup_bind(struct inpcbtable *table,
803     struct in_addr laddr, u_int lport_arg)
804 {
805 	struct inpcbhead *head;
806 	struct inpcb_hdr *inph;
807 	struct inpcb *inp;
808 	u_int16_t lport = lport_arg;
809 
810 	head = INPCBHASH_BIND(table, laddr, lport);
811 	LIST_FOREACH(inph, head, inph_hash) {
812 		inp = (struct inpcb *)inph;
813 		if (inp->inp_af != AF_INET)
814 			continue;
815 
816 		if (inp->inp_lport == lport &&
817 		    in_hosteq(inp->inp_laddr, laddr))
818 			goto out;
819 	}
820 	head = INPCBHASH_BIND(table, zeroin_addr, lport);
821 	LIST_FOREACH(inph, head, inph_hash) {
822 		inp = (struct inpcb *)inph;
823 		if (inp->inp_af != AF_INET)
824 			continue;
825 
826 		if (inp->inp_lport == lport &&
827 		    in_hosteq(inp->inp_laddr, zeroin_addr))
828 			goto out;
829 	}
830 #ifdef DIAGNOSTIC
831 	if (in_pcbnotifymiss) {
832 		printf("in_pcblookup_bind: laddr=%08x lport=%d\n",
833 		    ntohl(laddr.s_addr), ntohs(lport));
834 	}
835 #endif
836 	return (0);
837 
838 out:
839 	/* Move this PCB to the head of hash chain. */
840 	inph = &inp->inp_head;
841 	if (inph != LIST_FIRST(head)) {
842 		LIST_REMOVE(inph, inph_hash);
843 		LIST_INSERT_HEAD(head, inph, inph_hash);
844 	}
845 	return (inp);
846 }
847 
848 void
849 in_pcbstate(struct inpcb *inp, int state)
850 {
851 
852 	if (inp->inp_af != AF_INET)
853 		return;
854 
855 	if (inp->inp_state > INP_ATTACHED)
856 		LIST_REMOVE(&inp->inp_head, inph_hash);
857 
858 	switch (state) {
859 	case INP_BOUND:
860 		LIST_INSERT_HEAD(INPCBHASH_BIND(inp->inp_table,
861 		    inp->inp_laddr, inp->inp_lport), &inp->inp_head,
862 		    inph_hash);
863 		break;
864 	case INP_CONNECTED:
865 		LIST_INSERT_HEAD(INPCBHASH_CONNECT(inp->inp_table,
866 		    inp->inp_faddr, inp->inp_fport,
867 		    inp->inp_laddr, inp->inp_lport), &inp->inp_head,
868 		    inph_hash);
869 		break;
870 	}
871 
872 	inp->inp_state = state;
873 }
874 
875 struct rtentry *
876 in_pcbrtentry(struct inpcb *inp)
877 {
878 	struct route *ro;
879 	union {
880 		struct sockaddr		dst;
881 		struct sockaddr_in	dst4;
882 	} u;
883 
884 	if (inp->inp_af != AF_INET)
885 		return (NULL);
886 
887 	ro = &inp->inp_route;
888 
889 	sockaddr_in_init(&u.dst4, &inp->inp_faddr, 0);
890 	return rtcache_lookup(ro, &u.dst);
891 }
892 
893 struct sockaddr_in *
894 in_selectsrc(struct sockaddr_in *sin, struct route *ro,
895     int soopts, struct ip_moptions *mopts, int *errorp)
896 {
897 	struct rtentry *rt = NULL;
898 	struct in_ifaddr *ia = NULL;
899 
900 	/*
901          * If route is known or can be allocated now, take the
902          * source address from the interface.  Otherwise, punt.
903 	 */
904 	if ((soopts & SO_DONTROUTE) != 0)
905 		rtcache_free(ro);
906 	else {
907 		union {
908 			struct sockaddr		dst;
909 			struct sockaddr_in	dst4;
910 		} u;
911 
912 		sockaddr_in_init(&u.dst4, &sin->sin_addr, 0);
913 		rt = rtcache_lookup(ro, &u.dst);
914 	}
915 	/*
916 	 * If we found a route, use the address
917 	 * corresponding to the outgoing interface
918 	 * unless it is the loopback (in case a route
919 	 * to our address on another net goes to loopback).
920 	 *
921 	 * XXX Is this still true?  Do we care?
922 	 */
923 	if (rt != NULL && (rt->rt_ifp->if_flags & IFF_LOOPBACK) == 0)
924 		ia = ifatoia(rt->rt_ifa);
925 	if (ia == NULL) {
926 		u_int16_t fport = sin->sin_port;
927 
928 		sin->sin_port = 0;
929 		ia = ifatoia(ifa_ifwithladdr(sintosa(sin)));
930 		sin->sin_port = fport;
931 		if (ia == NULL) {
932 			/* Find 1st non-loopback AF_INET address */
933 			TAILQ_FOREACH(ia, &in_ifaddrhead, ia_list) {
934 				if (!(ia->ia_ifp->if_flags & IFF_LOOPBACK))
935 					break;
936 			}
937 		}
938 		if (ia == NULL) {
939 			*errorp = EADDRNOTAVAIL;
940 			return NULL;
941 		}
942 	}
943 	/*
944 	 * If the destination address is multicast and an outgoing
945 	 * interface has been set as a multicast option, use the
946 	 * address of that interface as our source address.
947 	 */
948 	if (IN_MULTICAST(sin->sin_addr.s_addr) && mopts != NULL) {
949 		struct ip_moptions *imo;
950 		struct ifnet *ifp;
951 
952 		imo = mopts;
953 		if (imo->imo_multicast_ifp != NULL) {
954 			ifp = imo->imo_multicast_ifp;
955 			IFP_TO_IA(ifp, ia);		/* XXX */
956 			if (ia == 0) {
957 				*errorp = EADDRNOTAVAIL;
958 				return NULL;
959 			}
960 		}
961 	}
962 	if (ia->ia_ifa.ifa_getifa != NULL) {
963 		ia = ifatoia((*ia->ia_ifa.ifa_getifa)(&ia->ia_ifa,
964 		                                      sintosa(sin)));
965 	}
966 #ifdef GETIFA_DEBUG
967 	else
968 		printf("%s: missing ifa_getifa\n", __func__);
969 #endif
970 	return satosin(&ia->ia_addr);
971 }
972