xref: /netbsd-src/sys/netinet/in_pcb.c (revision df0caa2637da0538ecdf6b878c4d08e684b43d8f)
1 /*	$NetBSD: in_pcb.c,v 1.100 2005/05/29 21:41:23 christos 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. Neither the name of the University nor the names of its contributors
82  *    may be used to endorse or promote products derived from this software
83  *    without specific prior written permission.
84  *
85  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
86  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
87  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
88  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
89  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
90  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
91  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
92  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
93  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
94  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
95  * SUCH DAMAGE.
96  *
97  *	@(#)in_pcb.c	8.4 (Berkeley) 5/24/95
98  */
99 
100 #include <sys/cdefs.h>
101 __KERNEL_RCSID(0, "$NetBSD: in_pcb.c,v 1.100 2005/05/29 21:41:23 christos Exp $");
102 
103 #include "opt_inet.h"
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 INET6
130 #include <netinet/ip6.h>
131 #include <netinet6/ip6_var.h>
132 #include <netinet6/in6_pcb.h>
133 #endif
134 
135 #ifdef IPSEC
136 #include <netinet6/ipsec.h>
137 #include <netkey/key.h>
138 #elif FAST_IPSEC
139 #include <netipsec/ipsec.h>
140 #include <netipsec/key.h>
141 #endif /* IPSEC */
142 
143 struct	in_addr zeroin_addr;
144 
145 #define	INPCBHASH_PORT(table, lport) \
146 	&(table)->inpt_porthashtbl[ntohs(lport) & (table)->inpt_porthash]
147 #define	INPCBHASH_BIND(table, laddr, lport) \
148 	&(table)->inpt_bindhashtbl[ \
149 	    ((ntohl((laddr).s_addr) + ntohs(lport))) & (table)->inpt_bindhash]
150 #define	INPCBHASH_CONNECT(table, faddr, fport, laddr, lport) \
151 	&(table)->inpt_connecthashtbl[ \
152 	    ((ntohl((faddr).s_addr) + ntohs(fport)) + \
153 	     (ntohl((laddr).s_addr) + ntohs(lport))) & (table)->inpt_connecthash]
154 
155 int	anonportmin = IPPORT_ANONMIN;
156 int	anonportmax = IPPORT_ANONMAX;
157 int	lowportmin  = IPPORT_RESERVEDMIN;
158 int	lowportmax  = IPPORT_RESERVEDMAX;
159 
160 POOL_INIT(inpcb_pool, sizeof(struct inpcb), 0, 0, 0, "inpcbpl", NULL);
161 
162 void
163 in_pcbinit(struct inpcbtable *table, int bindhashsize, int connecthashsize)
164 {
165 
166 	CIRCLEQ_INIT(&table->inpt_queue);
167 	table->inpt_porthashtbl = hashinit(bindhashsize, HASH_LIST, M_PCB,
168 	    M_WAITOK, &table->inpt_porthash);
169 	table->inpt_bindhashtbl = hashinit(bindhashsize, HASH_LIST, M_PCB,
170 	    M_WAITOK, &table->inpt_bindhash);
171 	table->inpt_connecthashtbl = hashinit(connecthashsize, HASH_LIST,
172 	    M_PCB, M_WAITOK, &table->inpt_connecthash);
173 	table->inpt_lastlow = IPPORT_RESERVEDMAX;
174 	table->inpt_lastport = (u_int16_t)anonportmax;
175 }
176 
177 int
178 in_pcballoc(struct socket *so, void *v)
179 {
180 	struct inpcbtable *table = v;
181 	struct inpcb *inp;
182 	int s;
183 #if defined(IPSEC) || defined(FAST_IPSEC)
184 	int error;
185 #endif
186 
187 	inp = pool_get(&inpcb_pool, PR_NOWAIT);
188 	if (inp == NULL)
189 		return (ENOBUFS);
190 	bzero((caddr_t)inp, sizeof(*inp));
191 	inp->inp_af = AF_INET;
192 	inp->inp_table = table;
193 	inp->inp_socket = so;
194 	inp->inp_errormtu = -1;
195 #if defined(IPSEC) || defined(FAST_IPSEC)
196 	error = ipsec_init_pcbpolicy(so, &inp->inp_sp);
197 	if (error != 0) {
198 		pool_put(&inpcb_pool, inp);
199 		return error;
200 	}
201 #endif
202 	so->so_pcb = inp;
203 	s = splnet();
204 	CIRCLEQ_INSERT_HEAD(&table->inpt_queue, &inp->inp_head,
205 	    inph_queue);
206 	LIST_INSERT_HEAD(INPCBHASH_PORT(table, inp->inp_lport), &inp->inp_head,
207 	    inph_lhash);
208 	in_pcbstate(inp, INP_ATTACHED);
209 	splx(s);
210 	return (0);
211 }
212 
213 int
214 in_pcbbind(void *v, struct mbuf *nam, struct proc *p)
215 {
216 	struct in_ifaddr *ia = NULL;
217 	struct inpcb *inp = v;
218 	struct socket *so = inp->inp_socket;
219 	struct inpcbtable *table = inp->inp_table;
220 	struct sockaddr_in *sin;
221 	u_int16_t lport = 0;
222 	int wild = 0, reuseport = (so->so_options & SO_REUSEPORT);
223 
224 	if (inp->inp_af != AF_INET)
225 		return (EINVAL);
226 
227 	if (TAILQ_FIRST(&in_ifaddrhead) == 0)
228 		return (EADDRNOTAVAIL);
229 	if (inp->inp_lport || !in_nullhost(inp->inp_laddr))
230 		return (EINVAL);
231 	if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0)
232 		wild = 1;
233 	if (nam == 0)
234 		goto noname;
235 	sin = mtod(nam, struct sockaddr_in *);
236 	if (nam->m_len != sizeof (*sin))
237 		return (EINVAL);
238 	if (sin->sin_family != AF_INET)
239 		return (EAFNOSUPPORT);
240 	lport = sin->sin_port;
241 	if (IN_MULTICAST(sin->sin_addr.s_addr)) {
242 		/*
243 		 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
244 		 * allow complete duplication of binding if
245 		 * SO_REUSEPORT is set, or if SO_REUSEADDR is set
246 		 * and a multicast address is bound on both
247 		 * new and duplicated sockets.
248 		 */
249 		if (so->so_options & SO_REUSEADDR)
250 			reuseport = SO_REUSEADDR|SO_REUSEPORT;
251 	} else if (!in_nullhost(sin->sin_addr)) {
252 		sin->sin_port = 0;		/* yech... */
253 		INADDR_TO_IA(sin->sin_addr, ia);
254 		/* check for broadcast addresses */
255 		if (ia == NULL)
256 			ia = ifatoia(ifa_ifwithaddr(sintosa(sin)));
257 		if (ia == NULL)
258 			return (EADDRNOTAVAIL);
259 	}
260 	if (lport) {
261 		struct inpcb *t;
262 #ifdef INET6
263 		struct in6pcb *t6;
264 		struct in6_addr mapped;
265 #endif
266 #ifndef IPNOPRIVPORTS
267 		/* GROSS */
268 		if (ntohs(lport) < IPPORT_RESERVED &&
269 		    (p == 0 || suser(p->p_ucred, &p->p_acflag)))
270 			return (EACCES);
271 #endif
272 #ifdef INET6
273 		memset(&mapped, 0, sizeof(mapped));
274 		mapped.s6_addr16[5] = 0xffff;
275 		memcpy(&mapped.s6_addr32[3], &sin->sin_addr,
276 		    sizeof(mapped.s6_addr32[3]));
277 		t6 = in6_pcblookup_port(table, &mapped, lport, wild);
278 		if (t6 && (reuseport & t6->in6p_socket->so_options) == 0)
279 			return (EADDRINUSE);
280 #endif
281 		if (so->so_uidinfo->ui_uid && !IN_MULTICAST(sin->sin_addr.s_addr)) {
282 			t = in_pcblookup_port(table, sin->sin_addr, lport, 1);
283 		/*
284 		 * XXX:	investigate ramifications of loosening this
285 		 *	restriction so that as long as both ports have
286 		 *	SO_REUSEPORT allow the bind
287 		 */
288 			if (t &&
289 			    (!in_nullhost(sin->sin_addr) ||
290 			     !in_nullhost(t->inp_laddr) ||
291 			     (t->inp_socket->so_options & SO_REUSEPORT) == 0)
292 			    && (so->so_uidinfo->ui_uid != t->inp_socket->so_uidinfo->ui_uid)) {
293 				return (EADDRINUSE);
294 			}
295 		}
296 		t = in_pcblookup_port(table, sin->sin_addr, lport, wild);
297 		if (t && (reuseport & t->inp_socket->so_options) == 0)
298 			return (EADDRINUSE);
299 	}
300 	inp->inp_laddr = sin->sin_addr;
301 
302 noname:
303 	if (lport == 0) {
304 		int	   cnt;
305 		u_int16_t  mymin, mymax;
306 		u_int16_t *lastport;
307 
308 		if (inp->inp_flags & INP_LOWPORT) {
309 #ifndef IPNOPRIVPORTS
310 			if (p == 0 || suser(p->p_ucred, &p->p_acflag))
311 				return (EACCES);
312 #endif
313 			mymin = lowportmin;
314 			mymax = lowportmax;
315 			lastport = &table->inpt_lastlow;
316 		} else {
317 			mymin = anonportmin;
318 			mymax = anonportmax;
319 			lastport = &table->inpt_lastport;
320 		}
321 		if (mymin > mymax) {	/* sanity check */
322 			u_int16_t swp;
323 
324 			swp = mymin;
325 			mymin = mymax;
326 			mymax = swp;
327 		}
328 
329 		lport = *lastport - 1;
330 		for (cnt = mymax - mymin + 1; cnt; cnt--, lport--) {
331 			if (lport < mymin || lport > mymax)
332 				lport = mymax;
333 			if (!in_pcblookup_port(table, inp->inp_laddr,
334 			    htons(lport), 1))
335 				goto found;
336 		}
337 		if (!in_nullhost(inp->inp_laddr))
338 			inp->inp_laddr.s_addr = INADDR_ANY;
339 		return (EAGAIN);
340 	found:
341 		inp->inp_flags |= INP_ANONPORT;
342 		*lastport = lport;
343 		lport = htons(lport);
344 	}
345 	inp->inp_lport = lport;
346 	LIST_REMOVE(&inp->inp_head, inph_lhash);
347 	LIST_INSERT_HEAD(INPCBHASH_PORT(table, inp->inp_lport), &inp->inp_head,
348 	    inph_lhash);
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(void *v, struct mbuf *nam)
361 {
362 	struct inpcb *inp = v;
363 	struct in_ifaddr *ia = NULL;
364 	struct sockaddr_in *ifaddr = NULL;
365 	struct sockaddr_in *sin = mtod(nam, struct sockaddr_in *);
366 	int error;
367 
368 	if (inp->inp_af != AF_INET)
369 		return (EINVAL);
370 
371 	if (nam->m_len != sizeof (*sin))
372 		return (EINVAL);
373 	if (sin->sin_family != AF_INET)
374 		return (EAFNOSUPPORT);
375 	if (sin->sin_port == 0)
376 		return (EADDRNOTAVAIL);
377 	if (TAILQ_FIRST(&in_ifaddrhead) != 0) {
378 		/*
379 		 * If the destination address is INADDR_ANY,
380 		 * use any local address (likely loopback).
381 		 * If the supplied address is INADDR_BROADCAST,
382 		 * use the broadcast address of an interface
383 		 * which supports broadcast. (loopback does not)
384 		 */
385 
386 		if (in_nullhost(sin->sin_addr)) {
387 			sin->sin_addr =
388 			    TAILQ_FIRST(&in_ifaddrhead)->ia_addr.sin_addr;
389 		} else if (sin->sin_addr.s_addr == INADDR_BROADCAST) {
390 			TAILQ_FOREACH(ia, &in_ifaddrhead, ia_list) {
391 				if (ia->ia_ifp->if_flags & IFF_BROADCAST) {
392 					sin->sin_addr =
393 					    ia->ia_broadaddr.sin_addr;
394 					break;
395 				}
396 			}
397 		}
398 	}
399 	/*
400 	 * If we haven't bound which network number to use as ours,
401 	 * we will use the number of the outgoing interface.
402 	 * This depends on having done a routing lookup, which
403 	 * we will probably have to do anyway, so we might
404 	 * as well do it now.  On the other hand if we are
405 	 * sending to multiple destinations we may have already
406 	 * done the lookup, so see if we can use the route
407 	 * from before.  In any case, we only
408 	 * chose a port number once, even if sending to multiple
409 	 * destinations.
410 	 */
411 	if (in_nullhost(inp->inp_laddr)) {
412 		int xerror;
413 		ifaddr = in_selectsrc(sin, &inp->inp_route,
414 		    inp->inp_socket->so_options, inp->inp_moptions, &xerror);
415 		if (ifaddr == NULL) {
416 			if (xerror == 0)
417 				xerror = EADDRNOTAVAIL;
418 			return xerror;
419 		}
420 		INADDR_TO_IA(ifaddr->sin_addr, ia);
421 		if (ia == NULL)
422 			return (EADDRNOTAVAIL);
423 	}
424 	if (in_pcblookup_connect(inp->inp_table, sin->sin_addr, sin->sin_port,
425 	    !in_nullhost(inp->inp_laddr) ? inp->inp_laddr : ifaddr->sin_addr,
426 	    inp->inp_lport) != 0)
427 		return (EADDRINUSE);
428 	if (in_nullhost(inp->inp_laddr)) {
429 		if (inp->inp_lport == 0) {
430 			error = in_pcbbind(inp, (struct mbuf *)0,
431 			    (struct proc *)0);
432 			/*
433 			 * This used to ignore the return value
434 			 * completely, but we need to check for
435 			 * ephemeral port shortage.
436 			 * XXX Should we check for other errors, too?
437 			 */
438 			if (error == EAGAIN)
439 				return (error);
440 		}
441 		inp->inp_laddr = ifaddr->sin_addr;
442 	}
443 	inp->inp_faddr = sin->sin_addr;
444 	inp->inp_fport = sin->sin_port;
445 	in_pcbstate(inp, INP_CONNECTED);
446 #if defined(IPSEC) || defined(FAST_IPSEC)
447 	if (inp->inp_socket->so_type == SOCK_STREAM)
448 		ipsec_pcbconn(inp->inp_sp);
449 #endif
450 	return (0);
451 }
452 
453 void
454 in_pcbdisconnect(void *v)
455 {
456 	struct inpcb *inp = v;
457 
458 	if (inp->inp_af != AF_INET)
459 		return;
460 
461 	inp->inp_faddr = zeroin_addr;
462 	inp->inp_fport = 0;
463 	in_pcbstate(inp, INP_BOUND);
464 #if defined(IPSEC) || defined(FAST_IPSEC)
465 	ipsec_pcbdisconn(inp->inp_sp);
466 #endif
467 	if (inp->inp_socket->so_state & SS_NOFDREF)
468 		in_pcbdetach(inp);
469 }
470 
471 void
472 in_pcbdetach(void *v)
473 {
474 	struct inpcb *inp = v;
475 	struct socket *so = inp->inp_socket;
476 	int s;
477 
478 	if (inp->inp_af != AF_INET)
479 		return;
480 
481 #if defined(IPSEC) || defined(FAST_IPSEC)
482 	ipsec4_delete_pcbpolicy(inp);
483 #endif /*IPSEC*/
484 	so->so_pcb = 0;
485 	sofree(so);
486 	if (inp->inp_options)
487 		(void)m_free(inp->inp_options);
488 	if (inp->inp_route.ro_rt)
489 		rtfree(inp->inp_route.ro_rt);
490 	ip_freemoptions(inp->inp_moptions);
491 	s = splnet();
492 	in_pcbstate(inp, INP_ATTACHED);
493 	LIST_REMOVE(&inp->inp_head, inph_lhash);
494 	CIRCLEQ_REMOVE(&inp->inp_table->inpt_queue, &inp->inp_head,
495 	    inph_queue);
496 	splx(s);
497 	pool_put(&inpcb_pool, inp);
498 }
499 
500 void
501 in_setsockaddr(struct inpcb *inp, struct mbuf *nam)
502 {
503 	struct sockaddr_in *sin;
504 
505 	if (inp->inp_af != AF_INET)
506 		return;
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(struct inpcb *inp, struct mbuf *nam)
519 {
520 	struct sockaddr_in *sin;
521 
522 	if (inp->inp_af != AF_INET)
523 		return;
524 
525 	nam->m_len = sizeof (*sin);
526 	sin = mtod(nam, struct sockaddr_in *);
527 	bzero((caddr_t)sin, sizeof (*sin));
528 	sin->sin_family = AF_INET;
529 	sin->sin_len = sizeof(*sin);
530 	sin->sin_port = inp->inp_fport;
531 	sin->sin_addr = inp->inp_faddr;
532 }
533 
534 /*
535  * Pass some notification to all connections of a protocol
536  * associated with address dst.  The local address and/or port numbers
537  * may be specified to limit the search.  The "usual action" will be
538  * taken, depending on the ctlinput cmd.  The caller must filter any
539  * cmds that are uninteresting (e.g., no error in the map).
540  * Call the protocol specific routine (if any) to report
541  * any errors for each matching socket.
542  *
543  * Must be called at splsoftnet.
544  */
545 int
546 in_pcbnotify(struct inpcbtable *table, struct in_addr faddr, u_int fport_arg,
547     struct in_addr laddr, u_int lport_arg, int errno,
548     void (*notify)(struct inpcb *, int))
549 {
550 	struct inpcbhead *head;
551 	struct inpcb *inp, *ninp;
552 	u_int16_t fport = fport_arg, lport = lport_arg;
553 	int nmatch;
554 
555 	if (in_nullhost(faddr) || notify == 0)
556 		return (0);
557 
558 	nmatch = 0;
559 	head = INPCBHASH_CONNECT(table, faddr, fport, laddr, lport);
560 	for (inp = (struct inpcb *)LIST_FIRST(head); inp != NULL; inp = ninp) {
561 		ninp = (struct inpcb *)LIST_NEXT(inp, inp_hash);
562 		if (inp->inp_af != AF_INET)
563 			continue;
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(struct inpcbtable *table, struct in_addr faddr, int errno,
577     void (*notify)(struct inpcb *, int))
578 {
579 	struct inpcb *inp, *ninp;
580 
581 	if (in_nullhost(faddr) || notify == 0)
582 		return;
583 
584 	for (inp = (struct inpcb *)CIRCLEQ_FIRST(&table->inpt_queue);
585 	    inp != (void *)&table->inpt_queue;
586 	    inp = ninp) {
587 		ninp = (struct inpcb *)CIRCLEQ_NEXT(inp, inp_queue);
588 		if (inp->inp_af != AF_INET)
589 			continue;
590 		if (in_hosteq(inp->inp_faddr, faddr))
591 			(*notify)(inp, errno);
592 	}
593 }
594 
595 void
596 in_pcbpurgeif0(struct inpcbtable *table, struct ifnet *ifp)
597 {
598 	struct inpcb *inp, *ninp;
599 	struct ip_moptions *imo;
600 	int i, gap;
601 
602 	for (inp = (struct inpcb *)CIRCLEQ_FIRST(&table->inpt_queue);
603 	    inp != (void *)&table->inpt_queue;
604 	    inp = ninp) {
605 		ninp = (struct inpcb *)CIRCLEQ_NEXT(inp, inp_queue);
606 		if (inp->inp_af != AF_INET)
607 			continue;
608 		imo = inp->inp_moptions;
609 		if (imo != NULL) {
610 			/*
611 			 * Unselect the outgoing interface if it is being
612 			 * detached.
613 			 */
614 			if (imo->imo_multicast_ifp == ifp)
615 				imo->imo_multicast_ifp = NULL;
616 
617 			/*
618 			 * Drop multicast group membership if we joined
619 			 * through the interface being detached.
620 			 */
621 			for (i = 0, gap = 0; i < imo->imo_num_memberships;
622 			    i++) {
623 				if (imo->imo_membership[i]->inm_ifp == ifp) {
624 					in_delmulti(imo->imo_membership[i]);
625 					gap++;
626 				} else if (gap != 0)
627 					imo->imo_membership[i - gap] =
628 					    imo->imo_membership[i];
629 			}
630 			imo->imo_num_memberships -= gap;
631 		}
632 	}
633 }
634 
635 void
636 in_pcbpurgeif(struct inpcbtable *table, struct ifnet *ifp)
637 {
638 	struct inpcb *inp, *ninp;
639 
640 	for (inp = (struct inpcb *)CIRCLEQ_FIRST(&table->inpt_queue);
641 	    inp != (void *)&table->inpt_queue;
642 	    inp = ninp) {
643 		ninp = (struct inpcb *)CIRCLEQ_NEXT(inp, inp_queue);
644 		if (inp->inp_af != AF_INET)
645 			continue;
646 		if (inp->inp_route.ro_rt != NULL &&
647 		    inp->inp_route.ro_rt->rt_ifp == ifp)
648 			in_rtchange(inp, 0);
649 	}
650 }
651 
652 /*
653  * Check for alternatives when higher level complains
654  * about service problems.  For now, invalidate cached
655  * routing information.  If the route was created dynamically
656  * (by a redirect), time to try a default gateway again.
657  */
658 void
659 in_losing(struct inpcb *inp)
660 {
661 	struct rtentry *rt;
662 	struct rt_addrinfo info;
663 
664 	if (inp->inp_af != AF_INET)
665 		return;
666 
667 	if ((rt = inp->inp_route.ro_rt)) {
668 		inp->inp_route.ro_rt = 0;
669 		bzero((caddr_t)&info, sizeof(info));
670 		info.rti_info[RTAX_DST] = &inp->inp_route.ro_dst;
671 		info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
672 		info.rti_info[RTAX_NETMASK] = rt_mask(rt);
673 		rt_missmsg(RTM_LOSING, &info, rt->rt_flags, 0);
674 		if (rt->rt_flags & RTF_DYNAMIC)
675 			(void) rtrequest(RTM_DELETE, rt_key(rt),
676 				rt->rt_gateway, rt_mask(rt), rt->rt_flags,
677 				(struct rtentry **)0);
678 		else
679 		/*
680 		 * A new route can be allocated
681 		 * the next time output is attempted.
682 		 */
683 			rtfree(rt);
684 	}
685 }
686 
687 /*
688  * After a routing change, flush old routing
689  * and allocate a (hopefully) better one.
690  */
691 void
692 in_rtchange(struct inpcb *inp, int errno)
693 {
694 
695 	if (inp->inp_af != AF_INET)
696 		return;
697 
698 	if (inp->inp_route.ro_rt) {
699 		rtfree(inp->inp_route.ro_rt);
700 		inp->inp_route.ro_rt = 0;
701 		/*
702 		 * A new route can be allocated the next time
703 		 * output is attempted.
704 		 */
705 	}
706 	/* XXX SHOULD NOTIFY HIGHER-LEVEL PROTOCOLS */
707 }
708 
709 struct inpcb *
710 in_pcblookup_port(struct inpcbtable *table, struct in_addr laddr,
711     u_int lport_arg, int lookup_wildcard)
712 {
713 	struct inpcbhead *head;
714 	struct inpcb_hdr *inph;
715 	struct inpcb *inp, *match = 0;
716 	int matchwild = 3, wildcard;
717 	u_int16_t lport = lport_arg;
718 
719 	head = INPCBHASH_PORT(table, lport);
720 	LIST_FOREACH(inph, head, inph_lhash) {
721 		inp = (struct inpcb *)inph;
722 		if (inp->inp_af != AF_INET)
723 			continue;
724 
725 		if (inp->inp_lport != lport)
726 			continue;
727 		wildcard = 0;
728 		if (!in_nullhost(inp->inp_faddr))
729 			wildcard++;
730 		if (in_nullhost(inp->inp_laddr)) {
731 			if (!in_nullhost(laddr))
732 				wildcard++;
733 		} else {
734 			if (in_nullhost(laddr))
735 				wildcard++;
736 			else {
737 				if (!in_hosteq(inp->inp_laddr, laddr))
738 					continue;
739 			}
740 		}
741 		if (wildcard && !lookup_wildcard)
742 			continue;
743 		if (wildcard < matchwild) {
744 			match = inp;
745 			matchwild = wildcard;
746 			if (matchwild == 0)
747 				break;
748 		}
749 	}
750 	return (match);
751 }
752 
753 #ifdef DIAGNOSTIC
754 int	in_pcbnotifymiss = 0;
755 #endif
756 
757 struct inpcb *
758 in_pcblookup_connect(struct inpcbtable *table,
759     struct in_addr faddr, u_int fport_arg,
760     struct in_addr laddr, u_int lport_arg)
761 {
762 	struct inpcbhead *head;
763 	struct inpcb_hdr *inph;
764 	struct inpcb *inp;
765 	u_int16_t fport = fport_arg, lport = lport_arg;
766 
767 	head = INPCBHASH_CONNECT(table, faddr, fport, laddr, lport);
768 	LIST_FOREACH(inph, head, inph_hash) {
769 		inp = (struct inpcb *)inph;
770 		if (inp->inp_af != AF_INET)
771 			continue;
772 
773 		if (in_hosteq(inp->inp_faddr, faddr) &&
774 		    inp->inp_fport == fport &&
775 		    inp->inp_lport == lport &&
776 		    in_hosteq(inp->inp_laddr, laddr))
777 			goto out;
778 	}
779 #ifdef DIAGNOSTIC
780 	if (in_pcbnotifymiss) {
781 		printf("in_pcblookup_connect: faddr=%08x fport=%d laddr=%08x lport=%d\n",
782 		    ntohl(faddr.s_addr), ntohs(fport),
783 		    ntohl(laddr.s_addr), ntohs(lport));
784 	}
785 #endif
786 	return (0);
787 
788 out:
789 	/* Move this PCB to the head of hash chain. */
790 	inph = &inp->inp_head;
791 	if (inph != LIST_FIRST(head)) {
792 		LIST_REMOVE(inph, inph_hash);
793 		LIST_INSERT_HEAD(head, inph, inph_hash);
794 	}
795 	return (inp);
796 }
797 
798 struct inpcb *
799 in_pcblookup_bind(struct inpcbtable *table,
800     struct in_addr laddr, u_int lport_arg)
801 {
802 	struct inpcbhead *head;
803 	struct inpcb_hdr *inph;
804 	struct inpcb *inp;
805 	u_int16_t lport = lport_arg;
806 
807 	head = INPCBHASH_BIND(table, laddr, lport);
808 	LIST_FOREACH(inph, head, inph_hash) {
809 		inp = (struct inpcb *)inph;
810 		if (inp->inp_af != AF_INET)
811 			continue;
812 
813 		if (inp->inp_lport == lport &&
814 		    in_hosteq(inp->inp_laddr, laddr))
815 			goto out;
816 	}
817 	head = INPCBHASH_BIND(table, zeroin_addr, lport);
818 	LIST_FOREACH(inph, head, inph_hash) {
819 		inp = (struct inpcb *)inph;
820 		if (inp->inp_af != AF_INET)
821 			continue;
822 
823 		if (inp->inp_lport == lport &&
824 		    in_hosteq(inp->inp_laddr, zeroin_addr))
825 			goto out;
826 	}
827 #ifdef DIAGNOSTIC
828 	if (in_pcbnotifymiss) {
829 		printf("in_pcblookup_bind: laddr=%08x lport=%d\n",
830 		    ntohl(laddr.s_addr), ntohs(lport));
831 	}
832 #endif
833 	return (0);
834 
835 out:
836 	/* Move this PCB to the head of hash chain. */
837 	inph = &inp->inp_head;
838 	if (inph != LIST_FIRST(head)) {
839 		LIST_REMOVE(inph, inph_hash);
840 		LIST_INSERT_HEAD(head, inph, inph_hash);
841 	}
842 	return (inp);
843 }
844 
845 void
846 in_pcbstate(struct inpcb *inp, int state)
847 {
848 
849 	if (inp->inp_af != AF_INET)
850 		return;
851 
852 	if (inp->inp_state > INP_ATTACHED)
853 		LIST_REMOVE(&inp->inp_head, inph_hash);
854 
855 	switch (state) {
856 	case INP_BOUND:
857 		LIST_INSERT_HEAD(INPCBHASH_BIND(inp->inp_table,
858 		    inp->inp_laddr, inp->inp_lport), &inp->inp_head,
859 		    inph_hash);
860 		break;
861 	case INP_CONNECTED:
862 		LIST_INSERT_HEAD(INPCBHASH_CONNECT(inp->inp_table,
863 		    inp->inp_faddr, inp->inp_fport,
864 		    inp->inp_laddr, inp->inp_lport), &inp->inp_head,
865 		    inph_hash);
866 		break;
867 	}
868 
869 	inp->inp_state = state;
870 }
871 
872 struct rtentry *
873 in_pcbrtentry(struct inpcb *inp)
874 {
875 	struct route *ro;
876 
877 	if (inp->inp_af != AF_INET)
878 		return (NULL);
879 
880 	ro = &inp->inp_route;
881 
882 	if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
883 	    !in_hosteq(satosin(&ro->ro_dst)->sin_addr, inp->inp_faddr))) {
884 		RTFREE(ro->ro_rt);
885 		ro->ro_rt = (struct rtentry *)NULL;
886 	}
887 	if (ro->ro_rt == (struct rtentry *)NULL &&
888 	    !in_nullhost(inp->inp_faddr)) {
889 		bzero(&ro->ro_dst, sizeof(struct sockaddr_in));
890 		ro->ro_dst.sa_family = AF_INET;
891 		ro->ro_dst.sa_len = sizeof(ro->ro_dst);
892 		satosin(&ro->ro_dst)->sin_addr = inp->inp_faddr;
893 		rtalloc(ro);
894 	}
895 	return (ro->ro_rt);
896 }
897 
898 struct sockaddr_in *
899 in_selectsrc(struct sockaddr_in *sin, struct route *ro,
900     int soopts, struct ip_moptions *mopts, int *errorp)
901 {
902 	struct in_ifaddr *ia;
903 
904 	ia = (struct in_ifaddr *)0;
905 	/*
906 	 * If route is known or can be allocated now,
907 	 * our src addr is taken from the i/f, else punt.
908 	 * Note that we should check the address family of the cached
909 	 * destination, in case of sharing the cache with IPv6.
910 	 */
911 	if (ro->ro_rt &&
912 	    (ro->ro_dst.sa_family != AF_INET ||
913 	    !in_hosteq(satosin(&ro->ro_dst)->sin_addr, sin->sin_addr) ||
914 	    soopts & SO_DONTROUTE)) {
915 		RTFREE(ro->ro_rt);
916 		ro->ro_rt = (struct rtentry *)0;
917 	}
918 	if ((soopts & SO_DONTROUTE) == 0 && /*XXX*/
919 	    (ro->ro_rt == (struct rtentry *)0 ||
920 	     ro->ro_rt->rt_ifp == (struct ifnet *)0)) {
921 		/* No route yet, so try to acquire one */
922 		bzero(&ro->ro_dst, sizeof(struct sockaddr_in));
923 		ro->ro_dst.sa_family = AF_INET;
924 		ro->ro_dst.sa_len = sizeof(struct sockaddr_in);
925 		satosin(&ro->ro_dst)->sin_addr = sin->sin_addr;
926 		rtalloc(ro);
927 	}
928 	/*
929 	 * If we found a route, use the address
930 	 * corresponding to the outgoing interface
931 	 * unless it is the loopback (in case a route
932 	 * to our address on another net goes to loopback).
933 	 *
934 	 * XXX Is this still true?  Do we care?
935 	 */
936 	if (ro->ro_rt && !(ro->ro_rt->rt_ifp->if_flags & IFF_LOOPBACK))
937 		ia = ifatoia(ro->ro_rt->rt_ifa);
938 	if (ia == NULL) {
939 		u_int16_t fport = sin->sin_port;
940 
941 		sin->sin_port = 0;
942 		ia = ifatoia(ifa_ifwithladdr(sintosa(sin)));
943 		sin->sin_port = fport;
944 		if (ia == 0) {
945 			/* Find 1st non-loopback AF_INET address */
946 			TAILQ_FOREACH(ia, &in_ifaddrhead, ia_list) {
947 				if (!(ia->ia_ifp->if_flags & IFF_LOOPBACK))
948 					break;
949 			}
950 		}
951 		if (ia == NULL) {
952 			*errorp = EADDRNOTAVAIL;
953 			return NULL;
954 		}
955 	}
956 	/*
957 	 * If the destination address is multicast and an outgoing
958 	 * interface has been set as a multicast option, use the
959 	 * address of that interface as our source address.
960 	 */
961 	if (IN_MULTICAST(sin->sin_addr.s_addr) && mopts != NULL) {
962 		struct ip_moptions *imo;
963 		struct ifnet *ifp;
964 
965 		imo = mopts;
966 		if (imo->imo_multicast_ifp != NULL) {
967 			ifp = imo->imo_multicast_ifp;
968 			IFP_TO_IA(ifp, ia);		/* XXX */
969 			if (ia == 0) {
970 				*errorp = EADDRNOTAVAIL;
971 				return NULL;
972 			}
973 		}
974 	}
975 	return satosin(&ia->ia_addr);
976 }
977