xref: /netbsd-src/sys/netinet/in_pcb.c (revision 8b0f9554ff8762542c4defc4f70e1eb76fb508fa)
1 /*	$NetBSD: in_pcb.c,v 1.119 2007/08/21 08:34:33 dyoung 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 1.119 2007/08/21 08:34:33 dyoung 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 #include <sys/kauth.h>
119 
120 #include <net/if.h>
121 #include <net/route.h>
122 
123 #include <netinet/in.h>
124 #include <netinet/in_systm.h>
125 #include <netinet/ip.h>
126 #include <netinet/in_pcb.h>
127 #include <netinet/in_var.h>
128 #include <netinet/ip_var.h>
129 
130 #ifdef INET6
131 #include <netinet/ip6.h>
132 #include <netinet6/ip6_var.h>
133 #include <netinet6/in6_pcb.h>
134 #endif
135 
136 #ifdef IPSEC
137 #include <netinet6/ipsec.h>
138 #include <netkey/key.h>
139 #elif FAST_IPSEC
140 #include <netipsec/ipsec.h>
141 #include <netipsec/key.h>
142 #endif /* IPSEC */
143 
144 struct	in_addr zeroin_addr;
145 
146 #define	INPCBHASH_PORT(table, lport) \
147 	&(table)->inpt_porthashtbl[ntohs(lport) & (table)->inpt_porthash]
148 #define	INPCBHASH_BIND(table, laddr, lport) \
149 	&(table)->inpt_bindhashtbl[ \
150 	    ((ntohl((laddr).s_addr) + ntohs(lport))) & (table)->inpt_bindhash]
151 #define	INPCBHASH_CONNECT(table, faddr, fport, laddr, lport) \
152 	&(table)->inpt_connecthashtbl[ \
153 	    ((ntohl((faddr).s_addr) + ntohs(fport)) + \
154 	     (ntohl((laddr).s_addr) + ntohs(lport))) & (table)->inpt_connecthash]
155 
156 int	anonportmin = IPPORT_ANONMIN;
157 int	anonportmax = IPPORT_ANONMAX;
158 int	lowportmin  = IPPORT_RESERVEDMIN;
159 int	lowportmax  = IPPORT_RESERVEDMAX;
160 
161 POOL_INIT(inpcb_pool, sizeof(struct inpcb), 0, 0, 0, "inpcbpl", NULL,
162     IPL_NET);
163 
164 void
165 in_pcbinit(struct inpcbtable *table, int bindhashsize, int connecthashsize)
166 {
167 
168 	CIRCLEQ_INIT(&table->inpt_queue);
169 	table->inpt_porthashtbl = hashinit(bindhashsize, HASH_LIST, M_PCB,
170 	    M_WAITOK, &table->inpt_porthash);
171 	table->inpt_bindhashtbl = hashinit(bindhashsize, HASH_LIST, M_PCB,
172 	    M_WAITOK, &table->inpt_bindhash);
173 	table->inpt_connecthashtbl = hashinit(connecthashsize, HASH_LIST,
174 	    M_PCB, M_WAITOK, &table->inpt_connecthash);
175 	table->inpt_lastlow = IPPORT_RESERVEDMAX;
176 	table->inpt_lastport = (u_int16_t)anonportmax;
177 }
178 
179 int
180 in_pcballoc(struct socket *so, void *v)
181 {
182 	struct inpcbtable *table = v;
183 	struct inpcb *inp;
184 	int s;
185 #if defined(IPSEC) || defined(FAST_IPSEC)
186 	int error;
187 #endif
188 
189 	s = splnet();
190 	inp = pool_get(&inpcb_pool, PR_NOWAIT);
191 	splx(s);
192 	if (inp == NULL)
193 		return (ENOBUFS);
194 	bzero((void *)inp, sizeof(*inp));
195 	inp->inp_af = AF_INET;
196 	inp->inp_table = table;
197 	inp->inp_socket = so;
198 	inp->inp_errormtu = -1;
199 #if defined(IPSEC) || defined(FAST_IPSEC)
200 	error = ipsec_init_pcbpolicy(so, &inp->inp_sp);
201 	if (error != 0) {
202 		s = splnet();
203 		pool_put(&inpcb_pool, inp);
204 		splx(s);
205 		return error;
206 	}
207 #endif
208 	so->so_pcb = inp;
209 	s = splnet();
210 	CIRCLEQ_INSERT_HEAD(&table->inpt_queue, &inp->inp_head,
211 	    inph_queue);
212 	LIST_INSERT_HEAD(INPCBHASH_PORT(table, inp->inp_lport), &inp->inp_head,
213 	    inph_lhash);
214 	in_pcbstate(inp, INP_ATTACHED);
215 	splx(s);
216 	return (0);
217 }
218 
219 int
220 in_pcbbind(void *v, struct mbuf *nam, struct lwp *l)
221 {
222 	struct in_ifaddr *ia = NULL;
223 	struct inpcb *inp = v;
224 	struct socket *so = inp->inp_socket;
225 	struct inpcbtable *table = inp->inp_table;
226 	struct sockaddr_in *sin = NULL; /* XXXGCC */
227 	u_int16_t lport = 0;
228 	int wild = 0, reuseport = (so->so_options & SO_REUSEPORT);
229 
230 	if (inp->inp_af != AF_INET)
231 		return (EINVAL);
232 
233 	if (TAILQ_FIRST(&in_ifaddrhead) == 0)
234 		return (EADDRNOTAVAIL);
235 	if (inp->inp_lport || !in_nullhost(inp->inp_laddr))
236 		return (EINVAL);
237 	if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0)
238 		wild = 1;
239 	if (nam == 0)
240 		goto noname;
241 	sin = mtod(nam, struct sockaddr_in *);
242 	if (nam->m_len != sizeof (*sin))
243 		return (EINVAL);
244 	if (sin->sin_family != AF_INET)
245 		return (EAFNOSUPPORT);
246 	lport = sin->sin_port;
247 	if (IN_MULTICAST(sin->sin_addr.s_addr)) {
248 		/*
249 		 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
250 		 * allow complete duplication of binding if
251 		 * SO_REUSEPORT is set, or if SO_REUSEADDR is set
252 		 * and a multicast address is bound on both
253 		 * new and duplicated sockets.
254 		 */
255 		if (so->so_options & SO_REUSEADDR)
256 			reuseport = SO_REUSEADDR|SO_REUSEPORT;
257 	} else if (!in_nullhost(sin->sin_addr)) {
258 		sin->sin_port = 0;		/* yech... */
259 		INADDR_TO_IA(sin->sin_addr, ia);
260 		/* check for broadcast addresses */
261 		if (ia == NULL)
262 			ia = ifatoia(ifa_ifwithaddr(sintosa(sin)));
263 		if (ia == NULL)
264 			return (EADDRNOTAVAIL);
265 	}
266 	if (lport) {
267 		struct inpcb *t;
268 #ifdef INET6
269 		struct in6pcb *t6;
270 		struct in6_addr mapped;
271 #endif
272 #ifndef IPNOPRIVPORTS
273 		/* GROSS */
274 		if (ntohs(lport) < IPPORT_RESERVED &&
275 		    (l == 0 || kauth_authorize_network(l->l_cred,
276 		    KAUTH_NETWORK_BIND,
277 		    KAUTH_REQ_NETWORK_BIND_PRIVPORT, so, sin,
278 		    NULL)))
279 			return (EACCES);
280 #endif
281 #ifdef INET6
282 		memset(&mapped, 0, sizeof(mapped));
283 		mapped.s6_addr16[5] = 0xffff;
284 		memcpy(&mapped.s6_addr32[3], &sin->sin_addr,
285 		    sizeof(mapped.s6_addr32[3]));
286 		t6 = in6_pcblookup_port(table, &mapped, lport, wild);
287 		if (t6 && (reuseport & t6->in6p_socket->so_options) == 0)
288 			return (EADDRINUSE);
289 #endif
290 		if (so->so_uidinfo->ui_uid && !IN_MULTICAST(sin->sin_addr.s_addr)) {
291 			t = in_pcblookup_port(table, sin->sin_addr, lport, 1);
292 		/*
293 		 * XXX:	investigate ramifications of loosening this
294 		 *	restriction so that as long as both ports have
295 		 *	SO_REUSEPORT allow the bind
296 		 */
297 			if (t &&
298 			    (!in_nullhost(sin->sin_addr) ||
299 			     !in_nullhost(t->inp_laddr) ||
300 			     (t->inp_socket->so_options & SO_REUSEPORT) == 0)
301 			    && (so->so_uidinfo->ui_uid != t->inp_socket->so_uidinfo->ui_uid)) {
302 				return (EADDRINUSE);
303 			}
304 		}
305 		t = in_pcblookup_port(table, sin->sin_addr, lport, wild);
306 		if (t && (reuseport & t->inp_socket->so_options) == 0)
307 			return (EADDRINUSE);
308 	}
309 	inp->inp_laddr = sin->sin_addr;
310 
311 noname:
312 	if (lport == 0) {
313 		int	   cnt;
314 		u_int16_t  mymin, mymax;
315 		u_int16_t *lastport;
316 
317 		if (inp->inp_flags & INP_LOWPORT) {
318 #ifndef IPNOPRIVPORTS
319 			if (l == 0 || kauth_authorize_network(l->l_cred,
320 			    KAUTH_NETWORK_BIND,
321 			    KAUTH_REQ_NETWORK_BIND_PRIVPORT, so,
322 			    sin, NULL))
323 				return (EACCES);
324 #endif
325 			mymin = lowportmin;
326 			mymax = lowportmax;
327 			lastport = &table->inpt_lastlow;
328 		} else {
329 			mymin = anonportmin;
330 			mymax = anonportmax;
331 			lastport = &table->inpt_lastport;
332 		}
333 		if (mymin > mymax) {	/* sanity check */
334 			u_int16_t swp;
335 
336 			swp = mymin;
337 			mymin = mymax;
338 			mymax = swp;
339 		}
340 
341 		lport = *lastport - 1;
342 		for (cnt = mymax - mymin + 1; cnt; cnt--, lport--) {
343 			if (lport < mymin || lport > mymax)
344 				lport = mymax;
345 			if (!in_pcblookup_port(table, inp->inp_laddr,
346 			    htons(lport), 1))
347 				goto found;
348 		}
349 		if (!in_nullhost(inp->inp_laddr))
350 			inp->inp_laddr.s_addr = INADDR_ANY;
351 		return (EAGAIN);
352 	found:
353 		inp->inp_flags |= INP_ANONPORT;
354 		*lastport = lport;
355 		lport = htons(lport);
356 	}
357 	inp->inp_lport = lport;
358 	LIST_REMOVE(&inp->inp_head, inph_lhash);
359 	LIST_INSERT_HEAD(INPCBHASH_PORT(table, inp->inp_lport), &inp->inp_head,
360 	    inph_lhash);
361 	in_pcbstate(inp, INP_BOUND);
362 	return (0);
363 }
364 
365 /*
366  * Connect from a socket to a specified address.
367  * Both address and port must be specified in argument sin.
368  * If don't have a local address for this socket yet,
369  * then pick one.
370  */
371 int
372 in_pcbconnect(void *v, struct mbuf *nam, struct lwp *l)
373 {
374 	struct inpcb *inp = v;
375 	struct in_ifaddr *ia = NULL;
376 	struct sockaddr_in *ifaddr = NULL;
377 	struct sockaddr_in *sin = mtod(nam, struct sockaddr_in *);
378 	int error;
379 
380 	if (inp->inp_af != AF_INET)
381 		return (EINVAL);
382 
383 	if (nam->m_len != sizeof (*sin))
384 		return (EINVAL);
385 	if (sin->sin_family != AF_INET)
386 		return (EAFNOSUPPORT);
387 	if (sin->sin_port == 0)
388 		return (EADDRNOTAVAIL);
389 	if (TAILQ_FIRST(&in_ifaddrhead) != 0) {
390 		/*
391 		 * If the destination address is INADDR_ANY,
392 		 * use any local address (likely loopback).
393 		 * If the supplied address is INADDR_BROADCAST,
394 		 * use the broadcast address of an interface
395 		 * which supports broadcast. (loopback does not)
396 		 */
397 
398 		if (in_nullhost(sin->sin_addr)) {
399 			sin->sin_addr =
400 			    TAILQ_FIRST(&in_ifaddrhead)->ia_addr.sin_addr;
401 		} else if (sin->sin_addr.s_addr == INADDR_BROADCAST) {
402 			TAILQ_FOREACH(ia, &in_ifaddrhead, ia_list) {
403 				if (ia->ia_ifp->if_flags & IFF_BROADCAST) {
404 					sin->sin_addr =
405 					    ia->ia_broadaddr.sin_addr;
406 					break;
407 				}
408 			}
409 		}
410 	}
411 	/*
412 	 * If we haven't bound which network number to use as ours,
413 	 * we will use the number of the outgoing interface.
414 	 * This depends on having done a routing lookup, which
415 	 * we will probably have to do anyway, so we might
416 	 * as well do it now.  On the other hand if we are
417 	 * sending to multiple destinations we may have already
418 	 * done the lookup, so see if we can use the route
419 	 * from before.  In any case, we only
420 	 * chose a port number once, even if sending to multiple
421 	 * destinations.
422 	 */
423 	if (in_nullhost(inp->inp_laddr)) {
424 		int xerror;
425 		ifaddr = in_selectsrc(sin, &inp->inp_route,
426 		    inp->inp_socket->so_options, inp->inp_moptions, &xerror);
427 		if (ifaddr == NULL) {
428 			if (xerror == 0)
429 				xerror = EADDRNOTAVAIL;
430 			return xerror;
431 		}
432 		INADDR_TO_IA(ifaddr->sin_addr, ia);
433 		if (ia == NULL)
434 			return (EADDRNOTAVAIL);
435 	}
436 	if (in_pcblookup_connect(inp->inp_table, sin->sin_addr, sin->sin_port,
437 	    !in_nullhost(inp->inp_laddr) ? inp->inp_laddr : ifaddr->sin_addr,
438 	    inp->inp_lport) != 0)
439 		return (EADDRINUSE);
440 	if (in_nullhost(inp->inp_laddr)) {
441 		if (inp->inp_lport == 0) {
442 			error = in_pcbbind(inp, NULL, l);
443 			/*
444 			 * This used to ignore the return value
445 			 * completely, but we need to check for
446 			 * ephemeral port shortage.
447 			 * And attempts to request low ports if not root.
448 			 */
449 			if (error != 0)
450 				return (error);
451 		}
452 		inp->inp_laddr = ifaddr->sin_addr;
453 	}
454 	inp->inp_faddr = sin->sin_addr;
455 	inp->inp_fport = sin->sin_port;
456 	in_pcbstate(inp, INP_CONNECTED);
457 #if defined(IPSEC) || defined(FAST_IPSEC)
458 	if (inp->inp_socket->so_type == SOCK_STREAM)
459 		ipsec_pcbconn(inp->inp_sp);
460 #endif
461 	return (0);
462 }
463 
464 void
465 in_pcbdisconnect(void *v)
466 {
467 	struct inpcb *inp = v;
468 
469 	if (inp->inp_af != AF_INET)
470 		return;
471 
472 	inp->inp_faddr = zeroin_addr;
473 	inp->inp_fport = 0;
474 	in_pcbstate(inp, INP_BOUND);
475 #if defined(IPSEC) || defined(FAST_IPSEC)
476 	ipsec_pcbdisconn(inp->inp_sp);
477 #endif
478 	if (inp->inp_socket->so_state & SS_NOFDREF)
479 		in_pcbdetach(inp);
480 }
481 
482 void
483 in_pcbdetach(void *v)
484 {
485 	struct inpcb *inp = v;
486 	struct socket *so = inp->inp_socket;
487 	int s;
488 
489 	if (inp->inp_af != AF_INET)
490 		return;
491 
492 #if defined(IPSEC) || defined(FAST_IPSEC)
493 	ipsec4_delete_pcbpolicy(inp);
494 #endif /*IPSEC*/
495 	so->so_pcb = 0;
496 	sofree(so);
497 	if (inp->inp_options)
498 		(void)m_free(inp->inp_options);
499 	rtcache_free(&inp->inp_route);
500 	ip_freemoptions(inp->inp_moptions);
501 	s = splnet();
502 	in_pcbstate(inp, INP_ATTACHED);
503 	LIST_REMOVE(&inp->inp_head, inph_lhash);
504 	CIRCLEQ_REMOVE(&inp->inp_table->inpt_queue, &inp->inp_head,
505 	    inph_queue);
506 	pool_put(&inpcb_pool, inp);
507 	splx(s);
508 }
509 
510 void
511 in_setsockaddr(struct inpcb *inp, struct mbuf *nam)
512 {
513 	struct sockaddr_in *sin;
514 
515 	if (inp->inp_af != AF_INET)
516 		return;
517 
518 	sin = mtod(nam, struct sockaddr_in *);
519 	sockaddr_in_init(sin, &inp->inp_laddr, inp->inp_lport);
520 	nam->m_len = sin->sin_len;
521 }
522 
523 void
524 in_setpeeraddr(struct inpcb *inp, struct mbuf *nam)
525 {
526 	struct sockaddr_in *sin;
527 
528 	if (inp->inp_af != AF_INET)
529 		return;
530 
531 	sin = mtod(nam, struct sockaddr_in *);
532 	sockaddr_in_init(sin, &inp->inp_faddr, inp->inp_fport);
533 	nam->m_len = sin->sin_len;
534 }
535 
536 /*
537  * Pass some notification to all connections of a protocol
538  * associated with address dst.  The local address and/or port numbers
539  * may be specified to limit the search.  The "usual action" will be
540  * taken, depending on the ctlinput cmd.  The caller must filter any
541  * cmds that are uninteresting (e.g., no error in the map).
542  * Call the protocol specific routine (if any) to report
543  * any errors for each matching socket.
544  *
545  * Must be called at splsoftnet.
546  */
547 int
548 in_pcbnotify(struct inpcbtable *table, struct in_addr faddr, u_int fport_arg,
549     struct in_addr laddr, u_int lport_arg, int errno,
550     void (*notify)(struct inpcb *, int))
551 {
552 	struct inpcbhead *head;
553 	struct inpcb *inp, *ninp;
554 	u_int16_t fport = fport_arg, lport = lport_arg;
555 	int nmatch;
556 
557 	if (in_nullhost(faddr) || notify == 0)
558 		return (0);
559 
560 	nmatch = 0;
561 	head = INPCBHASH_CONNECT(table, faddr, fport, laddr, lport);
562 	for (inp = (struct inpcb *)LIST_FIRST(head); inp != NULL; inp = ninp) {
563 		ninp = (struct inpcb *)LIST_NEXT(inp, inp_hash);
564 		if (inp->inp_af != AF_INET)
565 			continue;
566 		if (in_hosteq(inp->inp_faddr, faddr) &&
567 		    inp->inp_fport == fport &&
568 		    inp->inp_lport == lport &&
569 		    in_hosteq(inp->inp_laddr, laddr)) {
570 			(*notify)(inp, errno);
571 			nmatch++;
572 		}
573 	}
574 	return (nmatch);
575 }
576 
577 void
578 in_pcbnotifyall(struct inpcbtable *table, struct in_addr faddr, int errno,
579     void (*notify)(struct inpcb *, int))
580 {
581 	struct inpcb *inp, *ninp;
582 
583 	if (in_nullhost(faddr) || notify == 0)
584 		return;
585 
586 	for (inp = (struct inpcb *)CIRCLEQ_FIRST(&table->inpt_queue);
587 	    inp != (void *)&table->inpt_queue;
588 	    inp = ninp) {
589 		ninp = (struct inpcb *)CIRCLEQ_NEXT(inp, inp_queue);
590 		if (inp->inp_af != AF_INET)
591 			continue;
592 		if (in_hosteq(inp->inp_faddr, faddr))
593 			(*notify)(inp, errno);
594 	}
595 }
596 
597 void
598 in_pcbpurgeif0(struct inpcbtable *table, struct ifnet *ifp)
599 {
600 	struct inpcb *inp, *ninp;
601 	struct ip_moptions *imo;
602 	int i, gap;
603 
604 	for (inp = (struct inpcb *)CIRCLEQ_FIRST(&table->inpt_queue);
605 	    inp != (void *)&table->inpt_queue;
606 	    inp = ninp) {
607 		ninp = (struct inpcb *)CIRCLEQ_NEXT(inp, inp_queue);
608 		if (inp->inp_af != AF_INET)
609 			continue;
610 		imo = inp->inp_moptions;
611 		if (imo != NULL) {
612 			/*
613 			 * Unselect the outgoing interface if it is being
614 			 * detached.
615 			 */
616 			if (imo->imo_multicast_ifp == ifp)
617 				imo->imo_multicast_ifp = NULL;
618 
619 			/*
620 			 * Drop multicast group membership if we joined
621 			 * through the interface being detached.
622 			 */
623 			for (i = 0, gap = 0; i < imo->imo_num_memberships;
624 			    i++) {
625 				if (imo->imo_membership[i]->inm_ifp == ifp) {
626 					in_delmulti(imo->imo_membership[i]);
627 					gap++;
628 				} else if (gap != 0)
629 					imo->imo_membership[i - gap] =
630 					    imo->imo_membership[i];
631 			}
632 			imo->imo_num_memberships -= gap;
633 		}
634 	}
635 }
636 
637 void
638 in_pcbpurgeif(struct inpcbtable *table, struct ifnet *ifp)
639 {
640 	struct inpcb *inp, *ninp;
641 
642 	for (inp = (struct inpcb *)CIRCLEQ_FIRST(&table->inpt_queue);
643 	    inp != (void *)&table->inpt_queue;
644 	    inp = ninp) {
645 		ninp = (struct inpcb *)CIRCLEQ_NEXT(inp, inp_queue);
646 		if (inp->inp_af != AF_INET)
647 			continue;
648 		if (inp->inp_route.ro_rt != NULL &&
649 		    inp->inp_route.ro_rt->rt_ifp == ifp)
650 			in_rtchange(inp, 0);
651 	}
652 }
653 
654 /*
655  * Check for alternatives when higher level complains
656  * about service problems.  For now, invalidate cached
657  * routing information.  If the route was created dynamically
658  * (by a redirect), time to try a default gateway again.
659  */
660 void
661 in_losing(struct inpcb *inp)
662 {
663 	struct rtentry *rt;
664 	struct rt_addrinfo info;
665 
666 	if (inp->inp_af != AF_INET)
667 		return;
668 
669 	if ((rt = inp->inp_route.ro_rt) != NULL) {
670 		memset(&info, 0, sizeof(info));
671 		info.rti_info[RTAX_DST] = rtcache_getdst(&inp->inp_route);
672 		info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
673 		info.rti_info[RTAX_NETMASK] = rt_mask(rt);
674 		rt_missmsg(RTM_LOSING, &info, rt->rt_flags, 0);
675 		if (rt->rt_flags & RTF_DYNAMIC)
676 			(void) rtrequest(RTM_DELETE, rt_getkey(rt),
677 				rt->rt_gateway, rt_mask(rt), rt->rt_flags,
678 				NULL);
679 		/*
680 		 * A new route can be allocated
681 		 * the next time output is attempted.
682 		 */
683 		rtcache_free(&inp->inp_route);
684 	}
685 }
686 
687 /*
688  * After a routing change, flush old routing.  A new route can be
689  * allocated the next time output is attempted.
690  */
691 void
692 in_rtchange(struct inpcb *inp, int errno)
693 {
694 
695 	if (inp->inp_af != AF_INET)
696 		return;
697 
698 	rtcache_free(&inp->inp_route);
699 
700 	/* XXX SHOULD NOTIFY HIGHER-LEVEL PROTOCOLS */
701 }
702 
703 struct inpcb *
704 in_pcblookup_port(struct inpcbtable *table, struct in_addr laddr,
705     u_int lport_arg, int lookup_wildcard)
706 {
707 	struct inpcbhead *head;
708 	struct inpcb_hdr *inph;
709 	struct inpcb *inp, *match = 0;
710 	int matchwild = 3, wildcard;
711 	u_int16_t lport = lport_arg;
712 
713 	head = INPCBHASH_PORT(table, lport);
714 	LIST_FOREACH(inph, head, inph_lhash) {
715 		inp = (struct inpcb *)inph;
716 		if (inp->inp_af != AF_INET)
717 			continue;
718 
719 		if (inp->inp_lport != lport)
720 			continue;
721 		wildcard = 0;
722 		if (!in_nullhost(inp->inp_faddr))
723 			wildcard++;
724 		if (in_nullhost(inp->inp_laddr)) {
725 			if (!in_nullhost(laddr))
726 				wildcard++;
727 		} else {
728 			if (in_nullhost(laddr))
729 				wildcard++;
730 			else {
731 				if (!in_hosteq(inp->inp_laddr, laddr))
732 					continue;
733 			}
734 		}
735 		if (wildcard && !lookup_wildcard)
736 			continue;
737 		if (wildcard < matchwild) {
738 			match = inp;
739 			matchwild = wildcard;
740 			if (matchwild == 0)
741 				break;
742 		}
743 	}
744 	return (match);
745 }
746 
747 #ifdef DIAGNOSTIC
748 int	in_pcbnotifymiss = 0;
749 #endif
750 
751 struct inpcb *
752 in_pcblookup_connect(struct inpcbtable *table,
753     struct in_addr faddr, u_int fport_arg,
754     struct in_addr laddr, u_int lport_arg)
755 {
756 	struct inpcbhead *head;
757 	struct inpcb_hdr *inph;
758 	struct inpcb *inp;
759 	u_int16_t fport = fport_arg, lport = lport_arg;
760 
761 	head = INPCBHASH_CONNECT(table, faddr, fport, laddr, lport);
762 	LIST_FOREACH(inph, head, inph_hash) {
763 		inp = (struct inpcb *)inph;
764 		if (inp->inp_af != AF_INET)
765 			continue;
766 
767 		if (in_hosteq(inp->inp_faddr, faddr) &&
768 		    inp->inp_fport == fport &&
769 		    inp->inp_lport == lport &&
770 		    in_hosteq(inp->inp_laddr, laddr))
771 			goto out;
772 	}
773 #ifdef DIAGNOSTIC
774 	if (in_pcbnotifymiss) {
775 		printf("in_pcblookup_connect: faddr=%08x fport=%d laddr=%08x lport=%d\n",
776 		    ntohl(faddr.s_addr), ntohs(fport),
777 		    ntohl(laddr.s_addr), ntohs(lport));
778 	}
779 #endif
780 	return (0);
781 
782 out:
783 	/* Move this PCB to the head of hash chain. */
784 	inph = &inp->inp_head;
785 	if (inph != LIST_FIRST(head)) {
786 		LIST_REMOVE(inph, inph_hash);
787 		LIST_INSERT_HEAD(head, inph, inph_hash);
788 	}
789 	return (inp);
790 }
791 
792 struct inpcb *
793 in_pcblookup_bind(struct inpcbtable *table,
794     struct in_addr laddr, u_int lport_arg)
795 {
796 	struct inpcbhead *head;
797 	struct inpcb_hdr *inph;
798 	struct inpcb *inp;
799 	u_int16_t lport = lport_arg;
800 
801 	head = INPCBHASH_BIND(table, laddr, lport);
802 	LIST_FOREACH(inph, head, inph_hash) {
803 		inp = (struct inpcb *)inph;
804 		if (inp->inp_af != AF_INET)
805 			continue;
806 
807 		if (inp->inp_lport == lport &&
808 		    in_hosteq(inp->inp_laddr, laddr))
809 			goto out;
810 	}
811 	head = INPCBHASH_BIND(table, zeroin_addr, lport);
812 	LIST_FOREACH(inph, head, inph_hash) {
813 		inp = (struct inpcb *)inph;
814 		if (inp->inp_af != AF_INET)
815 			continue;
816 
817 		if (inp->inp_lport == lport &&
818 		    in_hosteq(inp->inp_laddr, zeroin_addr))
819 			goto out;
820 	}
821 #ifdef DIAGNOSTIC
822 	if (in_pcbnotifymiss) {
823 		printf("in_pcblookup_bind: laddr=%08x lport=%d\n",
824 		    ntohl(laddr.s_addr), ntohs(lport));
825 	}
826 #endif
827 	return (0);
828 
829 out:
830 	/* Move this PCB to the head of hash chain. */
831 	inph = &inp->inp_head;
832 	if (inph != LIST_FIRST(head)) {
833 		LIST_REMOVE(inph, inph_hash);
834 		LIST_INSERT_HEAD(head, inph, inph_hash);
835 	}
836 	return (inp);
837 }
838 
839 void
840 in_pcbstate(struct inpcb *inp, int state)
841 {
842 
843 	if (inp->inp_af != AF_INET)
844 		return;
845 
846 	if (inp->inp_state > INP_ATTACHED)
847 		LIST_REMOVE(&inp->inp_head, inph_hash);
848 
849 	switch (state) {
850 	case INP_BOUND:
851 		LIST_INSERT_HEAD(INPCBHASH_BIND(inp->inp_table,
852 		    inp->inp_laddr, inp->inp_lport), &inp->inp_head,
853 		    inph_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_head,
859 		    inph_hash);
860 		break;
861 	}
862 
863 	inp->inp_state = state;
864 }
865 
866 struct rtentry *
867 in_pcbrtentry(struct inpcb *inp)
868 {
869 	struct route *ro;
870 	union {
871 		struct sockaddr		dst;
872 		struct sockaddr_in	dst4;
873 	} u;
874 
875 	if (inp->inp_af != AF_INET)
876 		return (NULL);
877 
878 	ro = &inp->inp_route;
879 
880 	sockaddr_in_init(&u.dst4, &inp->inp_faddr, 0);
881 	return rtcache_lookup(ro, &u.dst);
882 }
883 
884 struct sockaddr_in *
885 in_selectsrc(struct sockaddr_in *sin, struct route *ro,
886     int soopts, struct ip_moptions *mopts, int *errorp)
887 {
888 	struct in_ifaddr *ia = NULL;
889 
890 	/*
891          * If route is known or can be allocated now, take the
892          * source address from the interface.  Otherwise, punt.
893 	 */
894 	if ((soopts & SO_DONTROUTE) != 0)
895 		rtcache_free(ro);
896 	else {
897 		union {
898 			struct sockaddr		dst;
899 			struct sockaddr_in	dst4;
900 		} u;
901 
902 		sockaddr_in_init(&u.dst4, &sin->sin_addr, 0);
903 		(void)rtcache_lookup(ro, &u.dst);
904 	}
905 	/*
906 	 * If we found a route, use the address
907 	 * corresponding to the outgoing interface
908 	 * unless it is the loopback (in case a route
909 	 * to our address on another net goes to loopback).
910 	 *
911 	 * XXX Is this still true?  Do we care?
912 	 */
913 	if (ro->ro_rt != NULL && !(ro->ro_rt->rt_ifp->if_flags & IFF_LOOPBACK))
914 		ia = ifatoia(ro->ro_rt->rt_ifa);
915 	if (ia == NULL) {
916 		u_int16_t fport = sin->sin_port;
917 
918 		sin->sin_port = 0;
919 		ia = ifatoia(ifa_ifwithladdr(sintosa(sin)));
920 		sin->sin_port = fport;
921 		if (ia == NULL) {
922 			/* Find 1st non-loopback AF_INET address */
923 			TAILQ_FOREACH(ia, &in_ifaddrhead, ia_list) {
924 				if (!(ia->ia_ifp->if_flags & IFF_LOOPBACK))
925 					break;
926 			}
927 		}
928 		if (ia == NULL) {
929 			*errorp = EADDRNOTAVAIL;
930 			return NULL;
931 		}
932 	}
933 	/*
934 	 * If the destination address is multicast and an outgoing
935 	 * interface has been set as a multicast option, use the
936 	 * address of that interface as our source address.
937 	 */
938 	if (IN_MULTICAST(sin->sin_addr.s_addr) && mopts != NULL) {
939 		struct ip_moptions *imo;
940 		struct ifnet *ifp;
941 
942 		imo = mopts;
943 		if (imo->imo_multicast_ifp != NULL) {
944 			ifp = imo->imo_multicast_ifp;
945 			IFP_TO_IA(ifp, ia);		/* XXX */
946 			if (ia == 0) {
947 				*errorp = EADDRNOTAVAIL;
948 				return NULL;
949 			}
950 		}
951 	}
952 	if (ia->ia_ifa.ifa_getifa != NULL) {
953 		ia = ifatoia((*ia->ia_ifa.ifa_getifa)(&ia->ia_ifa,
954 		                                      sintosa(sin)));
955 	}
956 #ifdef GETIFA_DEBUG
957 	else
958 		printf("%s: missing ifa_getifa\n", __func__);
959 #endif
960 	return satosin(&ia->ia_addr);
961 }
962