xref: /dragonfly/sys/netinet/in_pcb.c (revision 6e285212)
1 /*
2  * Copyright (c) 1982, 1986, 1991, 1993, 1995
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. All advertising materials mentioning features or use of this software
14  *    must display the following acknowledgement:
15  *	This product includes software developed by the University of
16  *	California, Berkeley and its contributors.
17  * 4. Neither the name of the University nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  *
33  *	@(#)in_pcb.c	8.4 (Berkeley) 5/24/95
34  * $FreeBSD: src/sys/netinet/in_pcb.c,v 1.59.2.26 2003/01/24 05:11:33 sam Exp $
35  * $DragonFly: src/sys/netinet/in_pcb.c,v 1.2 2003/06/17 04:28:51 dillon Exp $
36  */
37 
38 #include "opt_ipsec.h"
39 #include "opt_inet6.h"
40 
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/malloc.h>
44 #include <sys/mbuf.h>
45 #include <sys/domain.h>
46 #include <sys/protosw.h>
47 #include <sys/socket.h>
48 #include <sys/socketvar.h>
49 #include <sys/proc.h>
50 #include <sys/jail.h>
51 #include <sys/kernel.h>
52 #include <sys/sysctl.h>
53 
54 #include <machine/limits.h>
55 
56 #include <vm/vm_zone.h>
57 
58 #include <net/if.h>
59 #include <net/if_types.h>
60 #include <net/route.h>
61 
62 #include <netinet/in.h>
63 #include <netinet/in_pcb.h>
64 #include <netinet/in_var.h>
65 #include <netinet/ip_var.h>
66 #ifdef INET6
67 #include <netinet/ip6.h>
68 #include <netinet6/ip6_var.h>
69 #endif /* INET6 */
70 
71 #ifdef IPSEC
72 #include <netinet6/ipsec.h>
73 #include <netkey/key.h>
74 #endif /* IPSEC */
75 
76 #ifdef FAST_IPSEC
77 #if defined(IPSEC) || defined(IPSEC_ESP)
78 #error "Bad idea: don't compile with both IPSEC and FAST_IPSEC!"
79 #endif
80 
81 #include <netipsec/ipsec.h>
82 #include <netipsec/key.h>
83 #define	IPSEC
84 #endif /* FAST_IPSEC */
85 
86 struct	in_addr zeroin_addr;
87 
88 /*
89  * These configure the range of local port addresses assigned to
90  * "unspecified" outgoing connections/packets/whatever.
91  */
92 int	ipport_lowfirstauto  = IPPORT_RESERVED - 1;	/* 1023 */
93 int	ipport_lowlastauto = IPPORT_RESERVEDSTART;	/* 600 */
94 int	ipport_firstauto = IPPORT_RESERVED;		/* 1024 */
95 int	ipport_lastauto  = IPPORT_USERRESERVED;		/* 5000 */
96 int	ipport_hifirstauto = IPPORT_HIFIRSTAUTO;	/* 49152 */
97 int	ipport_hilastauto  = IPPORT_HILASTAUTO;		/* 65535 */
98 
99 #define RANGECHK(var, min, max) \
100 	if ((var) < (min)) { (var) = (min); } \
101 	else if ((var) > (max)) { (var) = (max); }
102 
103 static int
104 sysctl_net_ipport_check(SYSCTL_HANDLER_ARGS)
105 {
106 	int error = sysctl_handle_int(oidp,
107 		oidp->oid_arg1, oidp->oid_arg2, req);
108 	if (!error) {
109 		RANGECHK(ipport_lowfirstauto, 1, IPPORT_RESERVED - 1);
110 		RANGECHK(ipport_lowlastauto, 1, IPPORT_RESERVED - 1);
111 		RANGECHK(ipport_firstauto, IPPORT_RESERVED, USHRT_MAX);
112 		RANGECHK(ipport_lastauto, IPPORT_RESERVED, USHRT_MAX);
113 		RANGECHK(ipport_hifirstauto, IPPORT_RESERVED, USHRT_MAX);
114 		RANGECHK(ipport_hilastauto, IPPORT_RESERVED, USHRT_MAX);
115 	}
116 	return error;
117 }
118 
119 #undef RANGECHK
120 
121 SYSCTL_NODE(_net_inet_ip, IPPROTO_IP, portrange, CTLFLAG_RW, 0, "IP Ports");
122 
123 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, lowfirst, CTLTYPE_INT|CTLFLAG_RW,
124 	   &ipport_lowfirstauto, 0, &sysctl_net_ipport_check, "I", "");
125 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, lowlast, CTLTYPE_INT|CTLFLAG_RW,
126 	   &ipport_lowlastauto, 0, &sysctl_net_ipport_check, "I", "");
127 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, first, CTLTYPE_INT|CTLFLAG_RW,
128 	   &ipport_firstauto, 0, &sysctl_net_ipport_check, "I", "");
129 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, last, CTLTYPE_INT|CTLFLAG_RW,
130 	   &ipport_lastauto, 0, &sysctl_net_ipport_check, "I", "");
131 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, hifirst, CTLTYPE_INT|CTLFLAG_RW,
132 	   &ipport_hifirstauto, 0, &sysctl_net_ipport_check, "I", "");
133 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, hilast, CTLTYPE_INT|CTLFLAG_RW,
134 	   &ipport_hilastauto, 0, &sysctl_net_ipport_check, "I", "");
135 
136 /*
137  * in_pcb.c: manage the Protocol Control Blocks.
138  *
139  * NOTE: It is assumed that most of these functions will be called at
140  * splnet(). XXX - There are, unfortunately, a few exceptions to this
141  * rule that should be fixed.
142  */
143 
144 /*
145  * Allocate a PCB and associate it with the socket.
146  */
147 int
148 in_pcballoc(so, pcbinfo, p)
149 	struct socket *so;
150 	struct inpcbinfo *pcbinfo;
151 	struct proc *p;
152 {
153 	register struct inpcb *inp;
154 #ifdef IPSEC
155 	int error;
156 #endif
157 
158 	inp = zalloci(pcbinfo->ipi_zone);
159 	if (inp == NULL)
160 		return (ENOBUFS);
161 	bzero((caddr_t)inp, sizeof(*inp));
162 	inp->inp_gencnt = ++pcbinfo->ipi_gencnt;
163 	inp->inp_pcbinfo = pcbinfo;
164 	inp->inp_socket = so;
165 #ifdef IPSEC
166 	error = ipsec_init_policy(so, &inp->inp_sp);
167 	if (error != 0) {
168 		zfreei(pcbinfo->ipi_zone, inp);
169 		return error;
170 	}
171 #endif /*IPSEC*/
172 #if defined(INET6)
173 	if (INP_SOCKAF(so) == AF_INET6 && ip6_v6only)
174 		inp->inp_flags |= IN6P_IPV6_V6ONLY;
175 #endif
176 	LIST_INSERT_HEAD(pcbinfo->listhead, inp, inp_list);
177 	pcbinfo->ipi_count++;
178 	so->so_pcb = (caddr_t)inp;
179 #ifdef INET6
180 	if (ip6_auto_flowlabel)
181 		inp->inp_flags |= IN6P_AUTOFLOWLABEL;
182 #endif
183 	return (0);
184 }
185 
186 int
187 in_pcbbind(inp, nam, p)
188 	register struct inpcb *inp;
189 	struct sockaddr *nam;
190 	struct proc *p;
191 {
192 	register struct socket *so = inp->inp_socket;
193 	unsigned short *lastport;
194 	struct sockaddr_in *sin;
195 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
196 	u_short lport = 0;
197 	int wild = 0, reuseport = (so->so_options & SO_REUSEPORT);
198 	int error, prison = 0;
199 
200 	if (TAILQ_EMPTY(&in_ifaddrhead)) /* XXX broken! */
201 		return (EADDRNOTAVAIL);
202 	if (inp->inp_lport || inp->inp_laddr.s_addr != INADDR_ANY)
203 		return (EINVAL);
204 	if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0)
205 		wild = 1;
206 	if (nam) {
207 		sin = (struct sockaddr_in *)nam;
208 		if (nam->sa_len != sizeof (*sin))
209 			return (EINVAL);
210 #ifdef notdef
211 		/*
212 		 * We should check the family, but old programs
213 		 * incorrectly fail to initialize it.
214 		 */
215 		if (sin->sin_family != AF_INET)
216 			return (EAFNOSUPPORT);
217 #endif
218 		if (sin->sin_addr.s_addr != INADDR_ANY)
219 			if (prison_ip(p, 0, &sin->sin_addr.s_addr))
220 				return(EINVAL);
221 		lport = sin->sin_port;
222 		if (IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) {
223 			/*
224 			 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
225 			 * allow complete duplication of binding if
226 			 * SO_REUSEPORT is set, or if SO_REUSEADDR is set
227 			 * and a multicast address is bound on both
228 			 * new and duplicated sockets.
229 			 */
230 			if (so->so_options & SO_REUSEADDR)
231 				reuseport = SO_REUSEADDR|SO_REUSEPORT;
232 		} else if (sin->sin_addr.s_addr != INADDR_ANY) {
233 			sin->sin_port = 0;		/* yech... */
234 			bzero(&sin->sin_zero, sizeof(sin->sin_zero));
235 			if (ifa_ifwithaddr((struct sockaddr *)sin) == 0)
236 				return (EADDRNOTAVAIL);
237 		}
238 		if (lport) {
239 			struct inpcb *t;
240 
241 			/* GROSS */
242 			if (ntohs(lport) < IPPORT_RESERVED && p &&
243 			    suser_xxx(0, p, PRISON_ROOT))
244 				return (EACCES);
245 			if (p && p->p_prison)
246 				prison = 1;
247 			if (so->so_cred->cr_uid != 0 &&
248 			    !IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) {
249 				t = in_pcblookup_local(inp->inp_pcbinfo,
250 				    sin->sin_addr, lport,
251 				    prison ? 0 :  INPLOOKUP_WILDCARD);
252 				if (t &&
253 				    (ntohl(sin->sin_addr.s_addr) != INADDR_ANY ||
254 				     ntohl(t->inp_laddr.s_addr) != INADDR_ANY ||
255 				     (t->inp_socket->so_options &
256 					 SO_REUSEPORT) == 0) &&
257 				    (so->so_cred->cr_uid !=
258 				     t->inp_socket->so_cred->cr_uid)) {
259 #if defined(INET6)
260 					if (ntohl(sin->sin_addr.s_addr) !=
261 					    INADDR_ANY ||
262 					    ntohl(t->inp_laddr.s_addr) !=
263 					    INADDR_ANY ||
264 					    INP_SOCKAF(so) ==
265 					    INP_SOCKAF(t->inp_socket))
266 #endif /* defined(INET6) */
267 					return (EADDRINUSE);
268 				}
269 			}
270 			if (prison &&
271 			    prison_ip(p, 0, &sin->sin_addr.s_addr))
272 				return (EADDRNOTAVAIL);
273 			t = in_pcblookup_local(pcbinfo, sin->sin_addr,
274 			    lport, prison ? 0 : wild);
275 			if (t &&
276 			    (reuseport & t->inp_socket->so_options) == 0) {
277 #if defined(INET6)
278 				if (ntohl(sin->sin_addr.s_addr) !=
279 				    INADDR_ANY ||
280 				    ntohl(t->inp_laddr.s_addr) !=
281 				    INADDR_ANY ||
282 				    INP_SOCKAF(so) ==
283 				    INP_SOCKAF(t->inp_socket))
284 #endif /* defined(INET6) */
285 				return (EADDRINUSE);
286 			}
287 		}
288 		inp->inp_laddr = sin->sin_addr;
289 	}
290 	if (lport == 0) {
291 		ushort first, last;
292 		int count;
293 
294 		if (inp->inp_laddr.s_addr != INADDR_ANY)
295 			if (prison_ip(p, 0, &inp->inp_laddr.s_addr )) {
296 				inp->inp_laddr.s_addr = INADDR_ANY;
297 				return (EINVAL);
298 			}
299 		inp->inp_flags |= INP_ANONPORT;
300 
301 		if (inp->inp_flags & INP_HIGHPORT) {
302 			first = ipport_hifirstauto;	/* sysctl */
303 			last  = ipport_hilastauto;
304 			lastport = &pcbinfo->lasthi;
305 		} else if (inp->inp_flags & INP_LOWPORT) {
306 			if (p && (error = suser_xxx(0, p, PRISON_ROOT))) {
307 				inp->inp_laddr.s_addr = INADDR_ANY;
308 				return error;
309 			}
310 			first = ipport_lowfirstauto;	/* 1023 */
311 			last  = ipport_lowlastauto;	/* 600 */
312 			lastport = &pcbinfo->lastlow;
313 		} else {
314 			first = ipport_firstauto;	/* sysctl */
315 			last  = ipport_lastauto;
316 			lastport = &pcbinfo->lastport;
317 		}
318 		/*
319 		 * Simple check to ensure all ports are not used up causing
320 		 * a deadlock here.
321 		 *
322 		 * We split the two cases (up and down) so that the direction
323 		 * is not being tested on each round of the loop.
324 		 */
325 		if (first > last) {
326 			/*
327 			 * counting down
328 			 */
329 			count = first - last;
330 
331 			do {
332 				if (count-- < 0) {	/* completely used? */
333 					inp->inp_laddr.s_addr = INADDR_ANY;
334 					return (EADDRNOTAVAIL);
335 				}
336 				--*lastport;
337 				if (*lastport > first || *lastport < last)
338 					*lastport = first;
339 				lport = htons(*lastport);
340 			} while (in_pcblookup_local(pcbinfo,
341 				 inp->inp_laddr, lport, wild));
342 		} else {
343 			/*
344 			 * counting up
345 			 */
346 			count = last - first;
347 
348 			do {
349 				if (count-- < 0) {	/* completely used? */
350 					inp->inp_laddr.s_addr = INADDR_ANY;
351 					return (EADDRNOTAVAIL);
352 				}
353 				++*lastport;
354 				if (*lastport < first || *lastport > last)
355 					*lastport = first;
356 				lport = htons(*lastport);
357 			} while (in_pcblookup_local(pcbinfo,
358 				 inp->inp_laddr, lport, wild));
359 		}
360 	}
361 	inp->inp_lport = lport;
362 	if (prison_ip(p, 0, &inp->inp_laddr.s_addr)) {
363 		inp->inp_laddr.s_addr = INADDR_ANY;
364 		inp->inp_lport = 0;
365 		return (EINVAL);
366 	}
367 	if (in_pcbinshash(inp) != 0) {
368 		inp->inp_laddr.s_addr = INADDR_ANY;
369 		inp->inp_lport = 0;
370 		return (EAGAIN);
371 	}
372 	return (0);
373 }
374 
375 /*
376  *   Transform old in_pcbconnect() into an inner subroutine for new
377  *   in_pcbconnect(): Do some validity-checking on the remote
378  *   address (in mbuf 'nam') and then determine local host address
379  *   (i.e., which interface) to use to access that remote host.
380  *
381  *   This preserves definition of in_pcbconnect(), while supporting a
382  *   slightly different version for T/TCP.  (This is more than
383  *   a bit of a kludge, but cleaning up the internal interfaces would
384  *   have forced minor changes in every protocol).
385  */
386 
387 int
388 in_pcbladdr(inp, nam, plocal_sin)
389 	register struct inpcb *inp;
390 	struct sockaddr *nam;
391 	struct sockaddr_in **plocal_sin;
392 {
393 	struct in_ifaddr *ia;
394 	register struct sockaddr_in *sin = (struct sockaddr_in *)nam;
395 
396 	if (nam->sa_len != sizeof (*sin))
397 		return (EINVAL);
398 	if (sin->sin_family != AF_INET)
399 		return (EAFNOSUPPORT);
400 	if (sin->sin_port == 0)
401 		return (EADDRNOTAVAIL);
402 	if (!TAILQ_EMPTY(&in_ifaddrhead)) {
403 		/*
404 		 * If the destination address is INADDR_ANY,
405 		 * use the primary local address.
406 		 * If the supplied address is INADDR_BROADCAST,
407 		 * and the primary interface supports broadcast,
408 		 * choose the broadcast address for that interface.
409 		 */
410 		if (sin->sin_addr.s_addr == INADDR_ANY)
411 		    sin->sin_addr = IA_SIN(TAILQ_FIRST(&in_ifaddrhead))->sin_addr;
412 		else if (sin->sin_addr.s_addr == (u_long)INADDR_BROADCAST &&
413 		  (TAILQ_FIRST(&in_ifaddrhead)->ia_ifp->if_flags & IFF_BROADCAST))
414 		    sin->sin_addr = satosin(&TAILQ_FIRST(&in_ifaddrhead)->ia_broadaddr)->sin_addr;
415 	}
416 	if (inp->inp_laddr.s_addr == INADDR_ANY) {
417 		register struct route *ro;
418 
419 		ia = (struct in_ifaddr *)0;
420 		/*
421 		 * If route is known or can be allocated now,
422 		 * our src addr is taken from the i/f, else punt.
423 		 * Note that we should check the address family of the cached
424 		 * destination, in case of sharing the cache with IPv6.
425 		 */
426 		ro = &inp->inp_route;
427 		if (ro->ro_rt &&
428 		    (ro->ro_dst.sa_family != AF_INET ||
429 		     satosin(&ro->ro_dst)->sin_addr.s_addr !=
430 		     sin->sin_addr.s_addr ||
431 		     inp->inp_socket->so_options & SO_DONTROUTE)) {
432 			RTFREE(ro->ro_rt);
433 			ro->ro_rt = (struct rtentry *)0;
434 		}
435 		if ((inp->inp_socket->so_options & SO_DONTROUTE) == 0 && /*XXX*/
436 		    (ro->ro_rt == (struct rtentry *)0 ||
437 		    ro->ro_rt->rt_ifp == (struct ifnet *)0)) {
438 			/* No route yet, so try to acquire one */
439 			bzero(&ro->ro_dst, sizeof(struct sockaddr_in));
440 			ro->ro_dst.sa_family = AF_INET;
441 			ro->ro_dst.sa_len = sizeof(struct sockaddr_in);
442 			((struct sockaddr_in *) &ro->ro_dst)->sin_addr =
443 				sin->sin_addr;
444 			rtalloc(ro);
445 		}
446 		/*
447 		 * If we found a route, use the address
448 		 * corresponding to the outgoing interface
449 		 * unless it is the loopback (in case a route
450 		 * to our address on another net goes to loopback).
451 		 */
452 		if (ro->ro_rt && !(ro->ro_rt->rt_ifp->if_flags & IFF_LOOPBACK))
453 			ia = ifatoia(ro->ro_rt->rt_ifa);
454 		if (ia == 0) {
455 			u_short fport = sin->sin_port;
456 
457 			sin->sin_port = 0;
458 			ia = ifatoia(ifa_ifwithdstaddr(sintosa(sin)));
459 			if (ia == 0)
460 				ia = ifatoia(ifa_ifwithnet(sintosa(sin)));
461 			sin->sin_port = fport;
462 			if (ia == 0)
463 				ia = TAILQ_FIRST(&in_ifaddrhead);
464 			if (ia == 0)
465 				return (EADDRNOTAVAIL);
466 		}
467 		/*
468 		 * If the destination address is multicast and an outgoing
469 		 * interface has been set as a multicast option, use the
470 		 * address of that interface as our source address.
471 		 */
472 		if (IN_MULTICAST(ntohl(sin->sin_addr.s_addr)) &&
473 		    inp->inp_moptions != NULL) {
474 			struct ip_moptions *imo;
475 			struct ifnet *ifp;
476 
477 			imo = inp->inp_moptions;
478 			if (imo->imo_multicast_ifp != NULL) {
479 				ifp = imo->imo_multicast_ifp;
480 				TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link)
481 					if (ia->ia_ifp == ifp)
482 						break;
483 				if (ia == 0)
484 					return (EADDRNOTAVAIL);
485 			}
486 		}
487 	/*
488 	 * Don't do pcblookup call here; return interface in plocal_sin
489 	 * and exit to caller, that will do the lookup.
490 	 */
491 		*plocal_sin = &ia->ia_addr;
492 
493 	}
494 	return(0);
495 }
496 
497 /*
498  * Outer subroutine:
499  * Connect from a socket to a specified address.
500  * Both address and port must be specified in argument sin.
501  * If don't have a local address for this socket yet,
502  * then pick one.
503  */
504 int
505 in_pcbconnect(inp, nam, p)
506 	register struct inpcb *inp;
507 	struct sockaddr *nam;
508 	struct proc *p;
509 {
510 	struct sockaddr_in *ifaddr;
511 	struct sockaddr_in *sin = (struct sockaddr_in *)nam;
512 	struct sockaddr_in sa;
513 	int error;
514 
515 	if (inp->inp_laddr.s_addr == INADDR_ANY && p->p_prison != NULL) {
516 		bzero(&sa, sizeof (sa));
517 		sa.sin_addr.s_addr = htonl(p->p_prison->pr_ip);
518 		sa.sin_len=sizeof (sa);
519 		sa.sin_family = AF_INET;
520 		error = in_pcbbind(inp, (struct sockaddr *)&sa, p);
521 		if (error)
522 			return (error);
523 	}
524 	/*
525 	 *   Call inner routine, to assign local interface address.
526 	 */
527 	if ((error = in_pcbladdr(inp, nam, &ifaddr)) != 0)
528 		return(error);
529 
530 	if (in_pcblookup_hash(inp->inp_pcbinfo, sin->sin_addr, sin->sin_port,
531 	    inp->inp_laddr.s_addr ? inp->inp_laddr : ifaddr->sin_addr,
532 	    inp->inp_lport, 0, NULL) != NULL) {
533 		return (EADDRINUSE);
534 	}
535 	if (inp->inp_laddr.s_addr == INADDR_ANY) {
536 		if (inp->inp_lport == 0) {
537 			error = in_pcbbind(inp, (struct sockaddr *)0, p);
538 			if (error)
539 			    return (error);
540 		}
541 		inp->inp_laddr = ifaddr->sin_addr;
542 	}
543 	inp->inp_faddr = sin->sin_addr;
544 	inp->inp_fport = sin->sin_port;
545 	in_pcbrehash(inp);
546 	return (0);
547 }
548 
549 void
550 in_pcbdisconnect(inp)
551 	struct inpcb *inp;
552 {
553 
554 	inp->inp_faddr.s_addr = INADDR_ANY;
555 	inp->inp_fport = 0;
556 	in_pcbrehash(inp);
557 	if (inp->inp_socket->so_state & SS_NOFDREF)
558 		in_pcbdetach(inp);
559 }
560 
561 void
562 in_pcbdetach(inp)
563 	struct inpcb *inp;
564 {
565 	struct socket *so = inp->inp_socket;
566 	struct inpcbinfo *ipi = inp->inp_pcbinfo;
567 
568 #ifdef IPSEC
569 	ipsec4_delete_pcbpolicy(inp);
570 #endif /*IPSEC*/
571 	inp->inp_gencnt = ++ipi->ipi_gencnt;
572 	in_pcbremlists(inp);
573 	so->so_pcb = 0;
574 	sofree(so);
575 	if (inp->inp_options)
576 		(void)m_free(inp->inp_options);
577 	if (inp->inp_route.ro_rt)
578 		rtfree(inp->inp_route.ro_rt);
579 	ip_freemoptions(inp->inp_moptions);
580 	inp->inp_vflag = 0;
581 	zfreei(ipi->ipi_zone, inp);
582 }
583 
584 /*
585  * The calling convention of in_setsockaddr() and in_setpeeraddr() was
586  * modified to match the pru_sockaddr() and pru_peeraddr() entry points
587  * in struct pr_usrreqs, so that protocols can just reference then directly
588  * without the need for a wrapper function.  The socket must have a valid
589  * (i.e., non-nil) PCB, but it should be impossible to get an invalid one
590  * except through a kernel programming error, so it is acceptable to panic
591  * (or in this case trap) if the PCB is invalid.  (Actually, we don't trap
592  * because there actually /is/ a programming error somewhere... XXX)
593  */
594 int
595 in_setsockaddr(so, nam)
596 	struct socket *so;
597 	struct sockaddr **nam;
598 {
599 	int s;
600 	register struct inpcb *inp;
601 	register struct sockaddr_in *sin;
602 
603 	/*
604 	 * Do the malloc first in case it blocks.
605 	 */
606 	MALLOC(sin, struct sockaddr_in *, sizeof *sin, M_SONAME,
607 		M_WAITOK | M_ZERO);
608 	sin->sin_family = AF_INET;
609 	sin->sin_len = sizeof(*sin);
610 
611 	s = splnet();
612 	inp = sotoinpcb(so);
613 	if (!inp) {
614 		splx(s);
615 		free(sin, M_SONAME);
616 		return ECONNRESET;
617 	}
618 	sin->sin_port = inp->inp_lport;
619 	sin->sin_addr = inp->inp_laddr;
620 	splx(s);
621 
622 	*nam = (struct sockaddr *)sin;
623 	return 0;
624 }
625 
626 int
627 in_setpeeraddr(so, nam)
628 	struct socket *so;
629 	struct sockaddr **nam;
630 {
631 	int s;
632 	struct inpcb *inp;
633 	register struct sockaddr_in *sin;
634 
635 	/*
636 	 * Do the malloc first in case it blocks.
637 	 */
638 	MALLOC(sin, struct sockaddr_in *, sizeof *sin, M_SONAME,
639 		M_WAITOK | M_ZERO);
640 	sin->sin_family = AF_INET;
641 	sin->sin_len = sizeof(*sin);
642 
643 	s = splnet();
644 	inp = sotoinpcb(so);
645 	if (!inp) {
646 		splx(s);
647 		free(sin, M_SONAME);
648 		return ECONNRESET;
649 	}
650 	sin->sin_port = inp->inp_fport;
651 	sin->sin_addr = inp->inp_faddr;
652 	splx(s);
653 
654 	*nam = (struct sockaddr *)sin;
655 	return 0;
656 }
657 
658 void
659 in_pcbnotifyall(head, faddr, errno, notify)
660 	struct inpcbhead *head;
661 	struct in_addr faddr;
662 	void (*notify) __P((struct inpcb *, int));
663 {
664 	struct inpcb *inp, *ninp;
665 	int s;
666 
667 	s = splnet();
668 	for (inp = LIST_FIRST(head); inp != NULL; inp = ninp) {
669 		ninp = LIST_NEXT(inp, inp_list);
670 #ifdef INET6
671 		if ((inp->inp_vflag & INP_IPV4) == 0)
672 			continue;
673 #endif
674 		if (inp->inp_faddr.s_addr != faddr.s_addr ||
675 		    inp->inp_socket == NULL)
676 				continue;
677 		(*notify)(inp, errno);
678 	}
679 	splx(s);
680 }
681 
682 void
683 in_pcbpurgeif0(head, ifp)
684 	struct inpcb *head;
685 	struct ifnet *ifp;
686 {
687 	struct inpcb *inp;
688 	struct ip_moptions *imo;
689 	int i, gap;
690 
691 	for (inp = head; inp != NULL; inp = LIST_NEXT(inp, inp_list)) {
692 		imo = inp->inp_moptions;
693 		if ((inp->inp_vflag & INP_IPV4) &&
694 		    imo != NULL) {
695 			/*
696 			 * Unselect the outgoing interface if it is being
697 			 * detached.
698 			 */
699 			if (imo->imo_multicast_ifp == ifp)
700 				imo->imo_multicast_ifp = NULL;
701 
702 			/*
703 			 * Drop multicast group membership if we joined
704 			 * through the interface being detached.
705 			 */
706 			for (i = 0, gap = 0; i < imo->imo_num_memberships;
707 			    i++) {
708 				if (imo->imo_membership[i]->inm_ifp == ifp) {
709 					in_delmulti(imo->imo_membership[i]);
710 					gap++;
711 				} else if (gap != 0)
712 					imo->imo_membership[i - gap] =
713 					    imo->imo_membership[i];
714 			}
715 			imo->imo_num_memberships -= gap;
716 		}
717 	}
718 }
719 
720 /*
721  * Check for alternatives when higher level complains
722  * about service problems.  For now, invalidate cached
723  * routing information.  If the route was created dynamically
724  * (by a redirect), time to try a default gateway again.
725  */
726 void
727 in_losing(inp)
728 	struct inpcb *inp;
729 {
730 	register struct rtentry *rt;
731 	struct rt_addrinfo info;
732 
733 	if ((rt = inp->inp_route.ro_rt)) {
734 		bzero((caddr_t)&info, sizeof(info));
735 		info.rti_flags = rt->rt_flags;
736 		info.rti_info[RTAX_DST] = rt_key(rt);
737 		info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
738 		info.rti_info[RTAX_NETMASK] = rt_mask(rt);
739 		rt_missmsg(RTM_LOSING, &info, rt->rt_flags, 0);
740 		if (rt->rt_flags & RTF_DYNAMIC)
741 			(void) rtrequest1(RTM_DELETE, &info, NULL);
742 		inp->inp_route.ro_rt = NULL;
743 		rtfree(rt);
744 		/*
745 		 * A new route can be allocated
746 		 * the next time output is attempted.
747 		 */
748 	}
749 }
750 
751 /*
752  * After a routing change, flush old routing
753  * and allocate a (hopefully) better one.
754  */
755 void
756 in_rtchange(inp, errno)
757 	register struct inpcb *inp;
758 	int errno;
759 {
760 	if (inp->inp_route.ro_rt) {
761 		rtfree(inp->inp_route.ro_rt);
762 		inp->inp_route.ro_rt = 0;
763 		/*
764 		 * A new route can be allocated the next time
765 		 * output is attempted.
766 		 */
767 	}
768 }
769 
770 /*
771  * Lookup a PCB based on the local address and port.
772  */
773 struct inpcb *
774 in_pcblookup_local(pcbinfo, laddr, lport_arg, wild_okay)
775 	struct inpcbinfo *pcbinfo;
776 	struct in_addr laddr;
777 	u_int lport_arg;
778 	int wild_okay;
779 {
780 	register struct inpcb *inp;
781 	int matchwild = 3, wildcard;
782 	u_short lport = lport_arg;
783 
784 	if (!wild_okay) {
785 		struct inpcbhead *head;
786 		/*
787 		 * Look for an unconnected (wildcard foreign addr) PCB that
788 		 * matches the local address and port we're looking for.
789 		 */
790 		head = &pcbinfo->hashbase[INP_PCBHASH(INADDR_ANY, lport, 0, pcbinfo->hashmask)];
791 		LIST_FOREACH(inp, head, inp_hash) {
792 #ifdef INET6
793 			if ((inp->inp_vflag & INP_IPV4) == 0)
794 				continue;
795 #endif
796 			if (inp->inp_faddr.s_addr == INADDR_ANY &&
797 			    inp->inp_laddr.s_addr == laddr.s_addr &&
798 			    inp->inp_lport == lport) {
799 				/*
800 				 * Found.
801 				 */
802 				return (inp);
803 			}
804 		}
805 		/*
806 		 * Not found.
807 		 */
808 		return (NULL);
809 	} else {
810 		struct inpcbporthead *porthash;
811 		struct inpcbport *phd;
812 		struct inpcb *match = NULL;
813 		/*
814 		 * Best fit PCB lookup.
815 		 *
816 		 * First see if this local port is in use by looking on the
817 		 * port hash list.
818 		 */
819 		porthash = &pcbinfo->porthashbase[INP_PCBPORTHASH(lport,
820 		    pcbinfo->porthashmask)];
821 		LIST_FOREACH(phd, porthash, phd_hash) {
822 			if (phd->phd_port == lport)
823 				break;
824 		}
825 		if (phd != NULL) {
826 			/*
827 			 * Port is in use by one or more PCBs. Look for best
828 			 * fit.
829 			 */
830 			LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) {
831 				wildcard = 0;
832 #ifdef INET6
833 				if ((inp->inp_vflag & INP_IPV4) == 0)
834 					continue;
835 #endif
836 				if (inp->inp_faddr.s_addr != INADDR_ANY)
837 					wildcard++;
838 				if (inp->inp_laddr.s_addr != INADDR_ANY) {
839 					if (laddr.s_addr == INADDR_ANY)
840 						wildcard++;
841 					else if (inp->inp_laddr.s_addr != laddr.s_addr)
842 						continue;
843 				} else {
844 					if (laddr.s_addr != INADDR_ANY)
845 						wildcard++;
846 				}
847 				if (wildcard < matchwild) {
848 					match = inp;
849 					matchwild = wildcard;
850 					if (matchwild == 0) {
851 						break;
852 					}
853 				}
854 			}
855 		}
856 		return (match);
857 	}
858 }
859 
860 /*
861  * Lookup PCB in hash list.
862  */
863 struct inpcb *
864 in_pcblookup_hash(pcbinfo, faddr, fport_arg, laddr, lport_arg, wildcard,
865 		  ifp)
866 	struct inpcbinfo *pcbinfo;
867 	struct in_addr faddr, laddr;
868 	u_int fport_arg, lport_arg;
869 	int wildcard;
870 	struct ifnet *ifp;
871 {
872 	struct inpcbhead *head;
873 	register struct inpcb *inp;
874 	u_short fport = fport_arg, lport = lport_arg;
875 
876 	/*
877 	 * First look for an exact match.
878 	 */
879 	head = &pcbinfo->hashbase[INP_PCBHASH(faddr.s_addr, lport, fport, pcbinfo->hashmask)];
880 	LIST_FOREACH(inp, head, inp_hash) {
881 #ifdef INET6
882 		if ((inp->inp_vflag & INP_IPV4) == 0)
883 			continue;
884 #endif
885 		if (inp->inp_faddr.s_addr == faddr.s_addr &&
886 		    inp->inp_laddr.s_addr == laddr.s_addr &&
887 		    inp->inp_fport == fport &&
888 		    inp->inp_lport == lport) {
889 			/*
890 			 * Found.
891 			 */
892 			return (inp);
893 		}
894 	}
895 	if (wildcard) {
896 		struct inpcb *local_wild = NULL;
897 #if defined(INET6)
898 		struct inpcb *local_wild_mapped = NULL;
899 #endif /* defined(INET6) */
900 
901 		head = &pcbinfo->hashbase[INP_PCBHASH(INADDR_ANY, lport, 0, pcbinfo->hashmask)];
902 		LIST_FOREACH(inp, head, inp_hash) {
903 #ifdef INET6
904 			if ((inp->inp_vflag & INP_IPV4) == 0)
905 				continue;
906 #endif
907 			if (inp->inp_faddr.s_addr == INADDR_ANY &&
908 			    inp->inp_lport == lport) {
909 				if (ifp && ifp->if_type == IFT_FAITH &&
910 				    (inp->inp_flags & INP_FAITH) == 0)
911 					continue;
912 				if (inp->inp_laddr.s_addr == laddr.s_addr)
913 					return (inp);
914 				else if (inp->inp_laddr.s_addr == INADDR_ANY) {
915 #if defined(INET6)
916 					if (INP_CHECK_SOCKAF(inp->inp_socket,
917 							     AF_INET6))
918 						local_wild_mapped = inp;
919 					else
920 #endif /* defined(INET6) */
921 					local_wild = inp;
922 				}
923 			}
924 		}
925 #if defined(INET6)
926 		if (local_wild == NULL)
927 			return (local_wild_mapped);
928 #endif /* defined(INET6) */
929 		return (local_wild);
930 	}
931 
932 	/*
933 	 * Not found.
934 	 */
935 	return (NULL);
936 }
937 
938 /*
939  * Insert PCB onto various hash lists.
940  */
941 int
942 in_pcbinshash(inp)
943 	struct inpcb *inp;
944 {
945 	struct inpcbhead *pcbhash;
946 	struct inpcbporthead *pcbporthash;
947 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
948 	struct inpcbport *phd;
949 	u_int32_t hashkey_faddr;
950 
951 #ifdef INET6
952 	if (inp->inp_vflag & INP_IPV6)
953 		hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX */;
954 	else
955 #endif /* INET6 */
956 	hashkey_faddr = inp->inp_faddr.s_addr;
957 
958 	pcbhash = &pcbinfo->hashbase[INP_PCBHASH(hashkey_faddr,
959 		 inp->inp_lport, inp->inp_fport, pcbinfo->hashmask)];
960 
961 	pcbporthash = &pcbinfo->porthashbase[INP_PCBPORTHASH(inp->inp_lport,
962 	    pcbinfo->porthashmask)];
963 
964 	/*
965 	 * Go through port list and look for a head for this lport.
966 	 */
967 	LIST_FOREACH(phd, pcbporthash, phd_hash) {
968 		if (phd->phd_port == inp->inp_lport)
969 			break;
970 	}
971 	/*
972 	 * If none exists, malloc one and tack it on.
973 	 */
974 	if (phd == NULL) {
975 		MALLOC(phd, struct inpcbport *, sizeof(struct inpcbport), M_PCB, M_NOWAIT);
976 		if (phd == NULL) {
977 			return (ENOBUFS); /* XXX */
978 		}
979 		phd->phd_port = inp->inp_lport;
980 		LIST_INIT(&phd->phd_pcblist);
981 		LIST_INSERT_HEAD(pcbporthash, phd, phd_hash);
982 	}
983 	inp->inp_phd = phd;
984 	LIST_INSERT_HEAD(&phd->phd_pcblist, inp, inp_portlist);
985 	LIST_INSERT_HEAD(pcbhash, inp, inp_hash);
986 	return (0);
987 }
988 
989 /*
990  * Move PCB to the proper hash bucket when { faddr, fport } have  been
991  * changed. NOTE: This does not handle the case of the lport changing (the
992  * hashed port list would have to be updated as well), so the lport must
993  * not change after in_pcbinshash() has been called.
994  */
995 void
996 in_pcbrehash(inp)
997 	struct inpcb *inp;
998 {
999 	struct inpcbhead *head;
1000 	u_int32_t hashkey_faddr;
1001 
1002 #ifdef INET6
1003 	if (inp->inp_vflag & INP_IPV6)
1004 		hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX */;
1005 	else
1006 #endif /* INET6 */
1007 	hashkey_faddr = inp->inp_faddr.s_addr;
1008 
1009 	head = &inp->inp_pcbinfo->hashbase[INP_PCBHASH(hashkey_faddr,
1010 		inp->inp_lport, inp->inp_fport, inp->inp_pcbinfo->hashmask)];
1011 
1012 	LIST_REMOVE(inp, inp_hash);
1013 	LIST_INSERT_HEAD(head, inp, inp_hash);
1014 }
1015 
1016 /*
1017  * Remove PCB from various lists.
1018  */
1019 void
1020 in_pcbremlists(inp)
1021 	struct inpcb *inp;
1022 {
1023 	inp->inp_gencnt = ++inp->inp_pcbinfo->ipi_gencnt;
1024 	if (inp->inp_lport) {
1025 		struct inpcbport *phd = inp->inp_phd;
1026 
1027 		LIST_REMOVE(inp, inp_hash);
1028 		LIST_REMOVE(inp, inp_portlist);
1029 		if (LIST_FIRST(&phd->phd_pcblist) == NULL) {
1030 			LIST_REMOVE(phd, phd_hash);
1031 			free(phd, M_PCB);
1032 		}
1033 	}
1034 	LIST_REMOVE(inp, inp_list);
1035 	inp->inp_pcbinfo->ipi_count--;
1036 }
1037 
1038 int
1039 prison_xinpcb(struct proc *p, struct inpcb *inp)
1040 {
1041 	if (!p->p_prison)
1042 		return (0);
1043 	if (ntohl(inp->inp_laddr.s_addr) == p->p_prison->pr_ip)
1044 		return (0);
1045 	return (1);
1046 }
1047