xref: /dflybsd-src/sys/netinet/in_pcb.c (revision a78cd7563c9ec3cf7b234be12c4928541ec8bb41)
1 /*
2  * Copyright (c) 2004 Jeffrey Hsu.  All rights reserved.
3  * Copyright (c) 1982, 1986, 1991, 1993, 1995
4  *	The Regents of the University of California.  All rights reserved.
5  *
6  * Redistribution and use in source and binary forms, with or without
7  * modification, are permitted provided that the following conditions
8  * are met:
9  * 1. Redistributions of source code must retain the above copyright
10  *    notice, this list of conditions and the following disclaimer.
11  * 2. Redistributions in binary form must reproduce the above copyright
12  *    notice, this list of conditions and the following disclaimer in the
13  *    documentation and/or other materials provided with the distribution.
14  * 3. All advertising materials mentioning features or use of this software
15  *    must display the following acknowledgement:
16  *	This product includes software developed by the University of
17  *	California, Berkeley and its contributors.
18  * 4. Neither the name of the University nor the names of its contributors
19  *    may be used to endorse or promote products derived from this software
20  *    without specific prior written permission.
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  *
34  *	@(#)in_pcb.c	8.4 (Berkeley) 5/24/95
35  * $FreeBSD: src/sys/netinet/in_pcb.c,v 1.59.2.27 2004/01/02 04:06:42 ambrisko Exp $
36  * $DragonFly: src/sys/netinet/in_pcb.c,v 1.14 2004/03/06 05:00:41 hsu Exp $
37  */
38 
39 #include "opt_ipsec.h"
40 #include "opt_inet6.h"
41 
42 #include <sys/param.h>
43 #include <sys/systm.h>
44 #include <sys/malloc.h>
45 #include <sys/mbuf.h>
46 #include <sys/domain.h>
47 #include <sys/protosw.h>
48 #include <sys/socket.h>
49 #include <sys/socketvar.h>
50 #include <sys/proc.h>
51 #include <sys/jail.h>
52 #include <sys/kernel.h>
53 #include <sys/sysctl.h>
54 
55 #include <machine/limits.h>
56 
57 #include <vm/vm_zone.h>
58 
59 #include <net/if.h>
60 #include <net/if_types.h>
61 #include <net/route.h>
62 
63 #include <netinet/in.h>
64 #include <netinet/in_pcb.h>
65 #include <netinet/in_var.h>
66 #include <netinet/ip_var.h>
67 #ifdef INET6
68 #include <netinet/ip6.h>
69 #include <netinet6/ip6_var.h>
70 #endif /* INET6 */
71 
72 #ifdef IPSEC
73 #include <netinet6/ipsec.h>
74 #include <netproto/key/key.h>
75 #endif
76 
77 #ifdef FAST_IPSEC
78 #if defined(IPSEC) || defined(IPSEC_ESP)
79 #error "Bad idea: don't compile with both IPSEC and FAST_IPSEC!"
80 #endif
81 
82 #include <netipsec/ipsec.h>
83 #include <netipsec/key.h>
84 #define	IPSEC
85 #endif /* FAST_IPSEC */
86 
87 struct in_addr zeroin_addr;
88 
89 /*
90  * These configure the range of local port addresses assigned to
91  * "unspecified" outgoing connections/packets/whatever.
92  */
93 int ipport_lowfirstauto = IPPORT_RESERVED - 1;	/* 1023 */
94 int ipport_lowlastauto = IPPORT_RESERVEDSTART;	/* 600 */
95 
96 int ipport_firstauto = IPPORT_RESERVED;		/* 1024 */
97 int ipport_lastauto = IPPORT_USERRESERVED;	/* 5000 */
98 
99 int ipport_hifirstauto = IPPORT_HIFIRSTAUTO;	/* 49152 */
100 int ipport_hilastauto = IPPORT_HILASTAUTO;	/* 65535 */
101 
102 static __inline void
103 RANGECHK(int var, int min, int max)
104 {
105 	if (var < min)
106 		var = min;
107 	else if (var > max)
108 		var = max;
109 }
110 
111 static int
112 sysctl_net_ipport_check(SYSCTL_HANDLER_ARGS)
113 {
114 	int error;
115 
116 	error = sysctl_handle_int(oidp, oidp->oid_arg1, oidp->oid_arg2, req);
117 	if (!error) {
118 		RANGECHK(ipport_lowfirstauto, 1, IPPORT_RESERVED - 1);
119 		RANGECHK(ipport_lowlastauto, 1, IPPORT_RESERVED - 1);
120 
121 		RANGECHK(ipport_firstauto, IPPORT_RESERVED, USHRT_MAX);
122 		RANGECHK(ipport_lastauto, IPPORT_RESERVED, USHRT_MAX);
123 
124 		RANGECHK(ipport_hifirstauto, IPPORT_RESERVED, USHRT_MAX);
125 		RANGECHK(ipport_hilastauto, IPPORT_RESERVED, USHRT_MAX);
126 	}
127 	return (error);
128 }
129 
130 SYSCTL_NODE(_net_inet_ip, IPPROTO_IP, portrange, CTLFLAG_RW, 0, "IP Ports");
131 
132 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, lowfirst, CTLTYPE_INT|CTLFLAG_RW,
133 	   &ipport_lowfirstauto, 0, &sysctl_net_ipport_check, "I", "");
134 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, lowlast, CTLTYPE_INT|CTLFLAG_RW,
135 	   &ipport_lowlastauto, 0, &sysctl_net_ipport_check, "I", "");
136 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, first, CTLTYPE_INT|CTLFLAG_RW,
137 	   &ipport_firstauto, 0, &sysctl_net_ipport_check, "I", "");
138 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, last, CTLTYPE_INT|CTLFLAG_RW,
139 	   &ipport_lastauto, 0, &sysctl_net_ipport_check, "I", "");
140 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, hifirst, CTLTYPE_INT|CTLFLAG_RW,
141 	   &ipport_hifirstauto, 0, &sysctl_net_ipport_check, "I", "");
142 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, hilast, CTLTYPE_INT|CTLFLAG_RW,
143 	   &ipport_hilastauto, 0, &sysctl_net_ipport_check, "I", "");
144 
145 /*
146  * in_pcb.c: manage the Protocol Control Blocks.
147  *
148  * NOTE: It is assumed that most of these functions will be called at
149  * splnet(). XXX - There are, unfortunately, a few exceptions to this
150  * rule that should be fixed.
151  */
152 
153 /*
154  * Allocate a PCB and associate it with the socket.
155  */
156 int
157 in_pcballoc(struct socket *so, struct inpcbinfo *pcbinfo)
158 {
159 	struct inpcb *inp;
160 #ifdef IPSEC
161 	int error;
162 #endif
163 
164 	inp = zalloc(pcbinfo->ipi_zone);
165 	if (inp == NULL)
166 		return (ENOBUFS);
167 	bzero((caddr_t)inp, sizeof *inp);
168 	inp->inp_gencnt = ++pcbinfo->ipi_gencnt;
169 	inp->inp_pcbinfo = pcbinfo;
170 	inp->inp_socket = so;
171 #ifdef IPSEC
172 	error = ipsec_init_policy(so, &inp->inp_sp);
173 	if (error != 0) {
174 		zfree(pcbinfo->ipi_zone, inp);
175 		return (error);
176 	}
177 #endif
178 #ifdef INET6
179 	if (INP_SOCKAF(so) == AF_INET6 && ip6_v6only)
180 		inp->inp_flags |= IN6P_IPV6_V6ONLY;
181 	if (ip6_auto_flowlabel)
182 		inp->inp_flags |= IN6P_AUTOFLOWLABEL;
183 #endif
184 	so->so_pcb = (caddr_t)inp;
185 	LIST_INSERT_HEAD(&pcbinfo->listhead, inp, inp_list);
186 	pcbinfo->ipi_count++;
187 	return (0);
188 }
189 
190 int
191 in_pcbbind(struct inpcb *inp, struct sockaddr *nam, struct thread *td)
192 {
193 	struct socket *so = inp->inp_socket;
194 	struct proc *p = td->td_proc;
195 	unsigned short *lastport;
196 	struct sockaddr_in *sin;
197 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
198 	u_short lport = 0;
199 	int wild = 0, reuseport = (so->so_options & SO_REUSEPORT);
200 	int error, prison = 0;
201 
202 	KKASSERT(p);
203 
204 	if (TAILQ_EMPTY(&in_ifaddrhead)) /* XXX broken! */
205 		return (EADDRNOTAVAIL);
206 	if (inp->inp_lport != 0 || inp->inp_laddr.s_addr != INADDR_ANY)
207 		return (EINVAL);	/* already bound */
208 	if (!(so->so_options & (SO_REUSEADDR|SO_REUSEPORT)))
209 		wild = 1;    /* neither SO_REUSEADDR nor SO_REUSEPORT is set */
210 	if (nam != NULL) {
211 		sin = (struct sockaddr_in *)nam;
212 		if (nam->sa_len != sizeof *sin)
213 			return (EINVAL);
214 #ifdef notdef
215 		/*
216 		 * We should check the family, but old programs
217 		 * incorrectly fail to initialize it.
218 		 */
219 		if (sin->sin_family != AF_INET)
220 			return (EAFNOSUPPORT);
221 #endif
222 		if (sin->sin_addr.s_addr != INADDR_ANY &&
223 		    prison_ip(td, 0, &sin->sin_addr.s_addr))
224 				return (EINVAL);
225 		lport = sin->sin_port;
226 		if (IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) {
227 			/*
228 			 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
229 			 * allow complete duplication of binding if
230 			 * SO_REUSEPORT is set, or if SO_REUSEADDR is set
231 			 * and a multicast address is bound on both
232 			 * new and duplicated sockets.
233 			 */
234 			if (so->so_options & SO_REUSEADDR)
235 				reuseport = SO_REUSEADDR | SO_REUSEPORT;
236 		} else if (sin->sin_addr.s_addr != INADDR_ANY) {
237 			sin->sin_port = 0;		/* yech... */
238 			bzero(&sin->sin_zero, sizeof sin->sin_zero);
239 			if (ifa_ifwithaddr((struct sockaddr *)sin) == NULL)
240 				return (EADDRNOTAVAIL);
241 		}
242 		if (lport != 0) {
243 			struct inpcb *t;
244 
245 			/* GROSS */
246 			if (ntohs(lport) < IPPORT_RESERVED &&
247 			    p && suser_cred(p->p_ucred, PRISON_ROOT))
248 				return (EACCES);
249 			if (p && p->p_ucred->cr_prison)
250 				prison = 1;
251 			if (so->so_cred->cr_uid != 0 &&
252 			    !IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) {
253 				t = in_pcblookup_local(inp->inp_pcbinfo,
254 				    sin->sin_addr, lport,
255 				    prison ? 0 : INPLOOKUP_WILDCARD);
256 				if (t &&
257 				    (!in_nullhost(sin->sin_addr) ||
258 				     !in_nullhost(t->inp_laddr) ||
259 				     (t->inp_socket->so_options &
260 					 SO_REUSEPORT) == 0) &&
261 				    (so->so_cred->cr_uid !=
262 				     t->inp_socket->so_cred->cr_uid)) {
263 #ifdef INET6
264 					if (!in_nullhost(sin->sin_addr) ||
265 					    !in_nullhost(t->inp_laddr) ||
266 					    INP_SOCKAF(so) ==
267 					    INP_SOCKAF(t->inp_socket))
268 #endif
269 					return (EADDRINUSE);
270 				}
271 			}
272 			if (prison && prison_ip(td, 0, &sin->sin_addr.s_addr))
273 				return (EADDRNOTAVAIL);
274 			t = in_pcblookup_local(pcbinfo, sin->sin_addr,
275 			    lport, prison ? 0 : wild);
276 			if (t && !(reuseport & t->inp_socket->so_options)) {
277 #ifdef INET6
278 				if (!in_nullhost(sin->sin_addr) ||
279 				    !in_nullhost(t->inp_laddr) ||
280 				    INP_SOCKAF(so) == INP_SOCKAF(t->inp_socket))
281 #endif
282 				return (EADDRINUSE);
283 			}
284 		}
285 		inp->inp_laddr = sin->sin_addr;
286 	}
287 	if (lport == 0) {
288 		ushort first, last;
289 		int count;
290 
291 		if (inp->inp_laddr.s_addr != INADDR_ANY &&
292 		    prison_ip(td, 0, &inp->inp_laddr.s_addr )) {
293 			inp->inp_laddr.s_addr = INADDR_ANY;
294 			return (EINVAL);
295 		}
296 		inp->inp_flags |= INP_ANONPORT;
297 
298 		if (inp->inp_flags & INP_HIGHPORT) {
299 			first = ipport_hifirstauto;	/* sysctl */
300 			last  = ipport_hilastauto;
301 			lastport = &pcbinfo->lasthi;
302 		} else if (inp->inp_flags & INP_LOWPORT) {
303 			if (p &&
304 			    (error = suser_cred(p->p_ucred, PRISON_ROOT))) {
305 				inp->inp_laddr.s_addr = INADDR_ANY;
306 				return (error);
307 			}
308 			first = ipport_lowfirstauto;	/* 1023 */
309 			last  = ipport_lowlastauto;	/* 600 */
310 			lastport = &pcbinfo->lastlow;
311 		} else {
312 			first = ipport_firstauto;	/* sysctl */
313 			last  = ipport_lastauto;
314 			lastport = &pcbinfo->lastport;
315 		}
316 		/*
317 		 * Simple check to ensure all ports are not used up causing
318 		 * a deadlock here.
319 		 *
320 		 * We split the two cases (up and down) so that the direction
321 		 * is not being tested on each round of the loop.
322 		 */
323 		if (first > last) {
324 			/*
325 			 * counting down
326 			 */
327 			count = first - last;
328 
329 			do {
330 				if (count-- < 0) {	/* completely used? */
331 					inp->inp_laddr.s_addr = INADDR_ANY;
332 					return (EADDRNOTAVAIL);
333 				}
334 				--*lastport;
335 				if (*lastport > first || *lastport < last)
336 					*lastport = first;
337 				lport = htons(*lastport);
338 			} while (in_pcblookup_local(pcbinfo,
339 				 inp->inp_laddr, lport, wild));
340 		} else {
341 			/*
342 			 * counting up
343 			 */
344 			count = last - first;
345 
346 			do {
347 				if (count-- < 0) {	/* completely used? */
348 					inp->inp_laddr.s_addr = INADDR_ANY;
349 					return (EADDRNOTAVAIL);
350 				}
351 				++*lastport;
352 				if (*lastport < first || *lastport > last)
353 					*lastport = first;
354 				lport = htons(*lastport);
355 			} while (in_pcblookup_local(pcbinfo,
356 				 inp->inp_laddr, lport, wild));
357 		}
358 	}
359 	inp->inp_lport = lport;
360 	if (prison_ip(td, 0, &inp->inp_laddr.s_addr)) {
361 		inp->inp_laddr.s_addr = INADDR_ANY;
362 		inp->inp_lport = 0;
363 		return (EINVAL);
364 	}
365 	if (in_pcbinsporthash(inp) != 0) {
366 		inp->inp_laddr.s_addr = INADDR_ANY;
367 		inp->inp_lport = 0;
368 		return (EAGAIN);
369 	}
370 	in_pcbinsbindhash(inp);
371 	return (0);
372 }
373 
374 /*
375  *   Transform old in_pcbconnect() into an inner subroutine for new
376  *   in_pcbconnect(): Do some validity-checking on the remote
377  *   address (in mbuf 'nam') and then determine local host address
378  *   (i.e., which interface) to use to access that remote host.
379  *
380  *   This preserves definition of in_pcbconnect(), while supporting a
381  *   slightly different version for T/TCP.  (This is more than
382  *   a bit of a kludge, but cleaning up the internal interfaces would
383  *   have forced minor changes in every protocol).
384  */
385 
386 int
387 in_pcbladdr(inp, nam, plocal_sin)
388 	struct inpcb *inp;
389 	struct sockaddr *nam;
390 	struct sockaddr_in **plocal_sin;
391 {
392 	struct in_ifaddr *ia;
393 	struct sockaddr_in *sin = (struct sockaddr_in *)nam;
394 
395 	if (nam->sa_len != sizeof *sin)
396 		return (EINVAL);
397 	if (sin->sin_family != AF_INET)
398 		return (EAFNOSUPPORT);
399 	if (sin->sin_port == 0)
400 		return (EADDRNOTAVAIL);
401 	if (!TAILQ_EMPTY(&in_ifaddrhead)) {
402 		ia = TAILQ_FIRST(&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(ia)->sin_addr;
412 		else if (sin->sin_addr.s_addr == (u_long)INADDR_BROADCAST &&
413 		    (ia->ia_ifp->if_flags & IFF_BROADCAST))
414 			sin->sin_addr = satosin(&ia->ia_broadaddr)->sin_addr;
415 	}
416 	if (inp->inp_laddr.s_addr == INADDR_ANY) {
417 		struct route *ro;
418 
419 		ia = (struct in_ifaddr *)NULL;
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_rt->rt_flags & RTF_UP) ||
429 		     ro->ro_dst.sa_family != AF_INET ||
430 		     satosin(&ro->ro_dst)->sin_addr.s_addr !=
431 		         sin->sin_addr.s_addr ||
432 		     inp->inp_socket->so_options & SO_DONTROUTE)) {
433 			RTFREE(ro->ro_rt);
434 			ro->ro_rt = (struct rtentry *)NULL;
435 		}
436 		if (!(inp->inp_socket->so_options & SO_DONTROUTE) && /*XXX*/
437 		    (ro->ro_rt == (struct rtentry *)NULL ||
438 		    ro->ro_rt->rt_ifp == (struct ifnet *)NULL)) {
439 			/* No route yet, so try to acquire one */
440 			bzero(&ro->ro_dst, sizeof(struct sockaddr_in));
441 			ro->ro_dst.sa_family = AF_INET;
442 			ro->ro_dst.sa_len = sizeof(struct sockaddr_in);
443 			((struct sockaddr_in *) &ro->ro_dst)->sin_addr =
444 				sin->sin_addr;
445 			rtalloc(ro);
446 		}
447 		/*
448 		 * If we found a route, use the address
449 		 * corresponding to the outgoing interface
450 		 * unless it is the loopback (in case a route
451 		 * to our address on another net goes to loopback).
452 		 */
453 		if (ro->ro_rt && !(ro->ro_rt->rt_ifp->if_flags & IFF_LOOPBACK))
454 			ia = ifatoia(ro->ro_rt->rt_ifa);
455 		if (ia == NULL) {
456 			u_short fport = sin->sin_port;
457 
458 			sin->sin_port = 0;
459 			ia = ifatoia(ifa_ifwithdstaddr(sintosa(sin)));
460 			if (ia == NULL)
461 				ia = ifatoia(ifa_ifwithnet(sintosa(sin)));
462 			sin->sin_port = fport;
463 			if (ia == NULL)
464 				ia = TAILQ_FIRST(&in_ifaddrhead);
465 			if (ia == NULL)
466 				return (EADDRNOTAVAIL);
467 		}
468 		/*
469 		 * If the destination address is multicast and an outgoing
470 		 * interface has been set as a multicast option, use the
471 		 * address of that interface as our source address.
472 		 */
473 		if (IN_MULTICAST(ntohl(sin->sin_addr.s_addr)) &&
474 		    inp->inp_moptions != NULL) {
475 			struct ip_moptions *imo;
476 			struct ifnet *ifp;
477 
478 			imo = inp->inp_moptions;
479 			if (imo->imo_multicast_ifp != NULL) {
480 				ifp = imo->imo_multicast_ifp;
481 				TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link)
482 					if (ia->ia_ifp == ifp)
483 						break;
484 				if (ia == NULL)
485 					return (EADDRNOTAVAIL);
486 			}
487 		}
488 		/*
489 		 * Don't do pcblookup call here; return interface in plocal_sin
490 		 * and exit to caller, that will do the lookup.
491 		 */
492 		*plocal_sin = &ia->ia_addr;
493 
494 	}
495 	return (0);
496 }
497 
498 /*
499  * Outer subroutine:
500  * Connect from a socket to a specified address.
501  * Both address and port must be specified in argument sin.
502  * If don't have a local address for this socket yet,
503  * then pick one.
504  */
505 int
506 in_pcbconnect(struct inpcb *inp, struct sockaddr *nam, struct thread *td)
507 {
508 	struct sockaddr_in *if_sin;
509 	struct sockaddr_in *sin = (struct sockaddr_in *)nam;
510 	struct sockaddr_in sa;
511 	struct ucred *cr = td->td_proc ? td->td_proc->p_ucred : NULL;
512 	int error;
513 
514 	if (cr && cr->cr_prison != NULL && in_nullhost(inp->inp_laddr)) {
515 		bzero(&sa, sizeof sa);
516 		sa.sin_addr.s_addr = htonl(cr->cr_prison->pr_ip);
517 		sa.sin_len = sizeof sa;
518 		sa.sin_family = AF_INET;
519 		error = in_pcbbind(inp, (struct sockaddr *)&sa, td);
520 		if (error)
521 			return (error);
522 	}
523 
524 	/* Call inner routine to assign local interface address. */
525 	if ((error = in_pcbladdr(inp, nam, &if_sin)) != 0)
526 		return (error);
527 
528 	if (in_pcblookup_hash(inp->inp_pcbinfo, sin->sin_addr, sin->sin_port,
529 	    inp->inp_laddr.s_addr ? inp->inp_laddr : if_sin->sin_addr,
530 	    inp->inp_lport, FALSE, NULL) != NULL) {
531 		return (EADDRINUSE);
532 	}
533 	if (inp->inp_laddr.s_addr == INADDR_ANY) {
534 		if (inp->inp_lport == 0) {
535 			error = in_pcbbind(inp, (struct sockaddr *)NULL, td);
536 			if (error)
537 				return (error);
538 		}
539 		inp->inp_laddr = if_sin->sin_addr;
540 	}
541 	inp->inp_faddr = sin->sin_addr;
542 	inp->inp_fport = sin->sin_port;
543 	in_pcbrehash(inp, INP_CONNECTED);
544 	return (0);
545 }
546 
547 void
548 in_pcbdisconnect(inp)
549 	struct inpcb *inp;
550 {
551 
552 	inp->inp_faddr.s_addr = INADDR_ANY;
553 	inp->inp_fport = 0;
554 	in_pcbremconnhash(inp);
555 	if (inp->inp_socket->so_state & SS_NOFDREF)
556 		in_pcbdetach(inp);
557 }
558 
559 void
560 in_pcbdetach(inp)
561 	struct inpcb *inp;
562 {
563 	struct socket *so = inp->inp_socket;
564 	struct inpcbinfo *ipi = inp->inp_pcbinfo;
565 
566 #ifdef IPSEC
567 	ipsec4_delete_pcbpolicy(inp);
568 #endif /*IPSEC*/
569 	inp->inp_gencnt = ++ipi->ipi_gencnt;
570 	in_pcbremlists(inp);
571 	so->so_pcb = 0;
572 	sofree(so);
573 	if (inp->inp_options)
574 		(void)m_free(inp->inp_options);
575 	if (inp->inp_route.ro_rt)
576 		rtfree(inp->inp_route.ro_rt);
577 	ip_freemoptions(inp->inp_moptions);
578 	inp->inp_vflag = 0;
579 	zfree(ipi->ipi_zone, inp);
580 }
581 
582 /*
583  * The calling convention of in_setsockaddr() and in_setpeeraddr() was
584  * modified to match the pru_sockaddr() and pru_peeraddr() entry points
585  * in struct pr_usrreqs, so that protocols can just reference then directly
586  * without the need for a wrapper function.  The socket must have a valid
587  * (i.e., non-nil) PCB, but it should be impossible to get an invalid one
588  * except through a kernel programming error, so it is acceptable to panic
589  * (or in this case trap) if the PCB is invalid.  (Actually, we don't trap
590  * because there actually /is/ a programming error somewhere... XXX)
591  */
592 int
593 in_setsockaddr(so, nam)
594 	struct socket *so;
595 	struct sockaddr **nam;
596 {
597 	int s;
598 	struct inpcb *inp;
599 	struct sockaddr_in *sin;
600 
601 	/*
602 	 * Do the malloc first in case it blocks.
603 	 */
604 	MALLOC(sin, struct sockaddr_in *, sizeof *sin, M_SONAME,
605 		M_WAITOK | M_ZERO);
606 	sin->sin_family = AF_INET;
607 	sin->sin_len = sizeof *sin;
608 
609 	s = splnet();
610 	inp = sotoinpcb(so);
611 	if (!inp) {
612 		splx(s);
613 		free(sin, M_SONAME);
614 		return (ECONNRESET);
615 	}
616 	sin->sin_port = inp->inp_lport;
617 	sin->sin_addr = inp->inp_laddr;
618 	splx(s);
619 
620 	*nam = (struct sockaddr *)sin;
621 	return (0);
622 }
623 
624 int
625 in_setpeeraddr(so, nam)
626 	struct socket *so;
627 	struct sockaddr **nam;
628 {
629 	int s;
630 	struct inpcb *inp;
631 	struct sockaddr_in *sin;
632 
633 	/*
634 	 * Do the malloc first in case it blocks.
635 	 */
636 	MALLOC(sin, struct sockaddr_in *, sizeof *sin, M_SONAME,
637 		M_WAITOK | M_ZERO);
638 	sin->sin_family = AF_INET;
639 	sin->sin_len = sizeof *sin;
640 
641 	s = splnet();
642 	inp = sotoinpcb(so);
643 	if (!inp) {
644 		splx(s);
645 		free(sin, M_SONAME);
646 		return (ECONNRESET);
647 	}
648 	sin->sin_port = inp->inp_fport;
649 	sin->sin_addr = inp->inp_faddr;
650 	splx(s);
651 
652 	*nam = (struct sockaddr *)sin;
653 	return (0);
654 }
655 
656 void
657 in_pcbnotifyall(head, faddr, errno, notify)
658 	struct inpcbhead *head;
659 	struct in_addr faddr;
660 	void (*notify) (struct inpcb *, int);
661 {
662 	struct inpcb *inp, *ninp;
663 	int s;
664 
665 	s = splnet();
666 	for (inp = LIST_FIRST(head); inp != NULL; inp = ninp) {
667 		ninp = LIST_NEXT(inp, inp_list);
668 #ifdef INET6
669 		if (!(inp->inp_vflag & INP_IPV4))
670 			continue;
671 #endif
672 		if (inp->inp_faddr.s_addr != faddr.s_addr ||
673 		    inp->inp_socket == NULL)
674 			continue;
675 		(*notify)(inp, errno);
676 	}
677 	splx(s);
678 }
679 
680 void
681 in_pcbpurgeif0(head, ifp)
682 	struct inpcb *head;
683 	struct ifnet *ifp;
684 {
685 	struct inpcb *inp;
686 	struct ip_moptions *imo;
687 	int i, gap;
688 
689 	for (inp = head; inp != NULL; inp = LIST_NEXT(inp, inp_list)) {
690 		imo = inp->inp_moptions;
691 		if ((inp->inp_vflag & INP_IPV4) && imo != NULL) {
692 			/*
693 			 * Unselect the outgoing interface if it is being
694 			 * detached.
695 			 */
696 			if (imo->imo_multicast_ifp == ifp)
697 				imo->imo_multicast_ifp = NULL;
698 
699 			/*
700 			 * Drop multicast group membership if we joined
701 			 * through the interface being detached.
702 			 */
703 			for (i = 0, gap = 0; i < imo->imo_num_memberships;
704 			    i++) {
705 				if (imo->imo_membership[i]->inm_ifp == ifp) {
706 					in_delmulti(imo->imo_membership[i]);
707 					gap++;
708 				} else if (gap != 0)
709 					imo->imo_membership[i - gap] =
710 					    imo->imo_membership[i];
711 			}
712 			imo->imo_num_memberships -= gap;
713 		}
714 	}
715 }
716 
717 /*
718  * Check for alternatives when higher level complains
719  * about service problems.  For now, invalidate cached
720  * routing information.  If the route was created dynamically
721  * (by a redirect), time to try a default gateway again.
722  */
723 void
724 in_losing(inp)
725 	struct inpcb *inp;
726 {
727 	struct rtentry *rt;
728 	struct rt_addrinfo info;
729 
730 	if ((rt = inp->inp_route.ro_rt)) {
731 		bzero((caddr_t)&info, sizeof info);
732 		info.rti_flags = rt->rt_flags;
733 		info.rti_info[RTAX_DST] = rt_key(rt);
734 		info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
735 		info.rti_info[RTAX_NETMASK] = rt_mask(rt);
736 		rt_missmsg(RTM_LOSING, &info, rt->rt_flags, 0);
737 		if (rt->rt_flags & RTF_DYNAMIC)
738 			(void) rtrequest1(RTM_DELETE, &info, NULL);
739 		inp->inp_route.ro_rt = NULL;
740 		rtfree(rt);
741 		/*
742 		 * A new route can be allocated
743 		 * the next time output is attempted.
744 		 */
745 	}
746 }
747 
748 /*
749  * After a routing change, flush old routing
750  * and allocate a (hopefully) better one.
751  */
752 void
753 in_rtchange(inp, errno)
754 	struct inpcb *inp;
755 	int errno;
756 {
757 	if (inp->inp_route.ro_rt) {
758 		rtfree(inp->inp_route.ro_rt);
759 		inp->inp_route.ro_rt = 0;
760 		/*
761 		 * A new route can be allocated the next time
762 		 * output is attempted.
763 		 */
764 	}
765 }
766 
767 /*
768  * Lookup a PCB based on the local address and port.
769  */
770 struct inpcb *
771 in_pcblookup_local(pcbinfo, laddr, lport_arg, wild_okay)
772 	struct inpcbinfo *pcbinfo;
773 	struct in_addr laddr;
774 	u_int lport_arg;
775 	int wild_okay;
776 {
777 	struct inpcb *inp;
778 	int matchwild = 3, wildcard;
779 	u_short lport = lport_arg;
780 
781 	struct inpcbporthead *porthash;
782 	struct inpcbport *phd;
783 	struct inpcb *match = NULL;
784 
785 	/*
786 	 * Best fit PCB lookup.
787 	 *
788 	 * First see if this local port is in use by looking on the
789 	 * port hash list.
790 	 */
791 	porthash = &pcbinfo->porthashbase[INP_PCBPORTHASH(lport,
792 	    pcbinfo->porthashmask)];
793 	LIST_FOREACH(phd, porthash, phd_hash) {
794 		if (phd->phd_port == lport)
795 			break;
796 	}
797 	if (phd != NULL) {
798 		/*
799 		 * Port is in use by one or more PCBs. Look for best
800 		 * fit.
801 		 */
802 		LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) {
803 			wildcard = 0;
804 #ifdef INET6
805 			if ((inp->inp_vflag & INP_IPV4) == 0)
806 				continue;
807 #endif
808 			if (inp->inp_faddr.s_addr != INADDR_ANY)
809 				wildcard++;
810 			if (inp->inp_laddr.s_addr != INADDR_ANY) {
811 				if (laddr.s_addr == INADDR_ANY)
812 					wildcard++;
813 				else if (inp->inp_laddr.s_addr != laddr.s_addr)
814 					continue;
815 			} else {
816 				if (laddr.s_addr != INADDR_ANY)
817 					wildcard++;
818 			}
819 			if (wildcard && !wild_okay)
820 				continue;
821 			if (wildcard < matchwild) {
822 				match = inp;
823 				matchwild = wildcard;
824 				if (matchwild == 0) {
825 					break;
826 				}
827 			}
828 		}
829 	}
830 	return (match);
831 }
832 
833 /*
834  * Lookup PCB in hash list.
835  */
836 struct inpcb *
837 in_pcblookup_hash(pcbinfo, faddr, fport_arg, laddr, lport_arg, wildcard, ifp)
838 	struct inpcbinfo *pcbinfo;
839 	struct in_addr faddr, laddr;
840 	u_int fport_arg, lport_arg;
841 	boolean_t wildcard;
842 	struct ifnet *ifp;
843 {
844 	struct inpcbhead *head;
845 	struct inpcb *inp;
846 	u_short fport = fport_arg, lport = lport_arg;
847 
848 	/*
849 	 * First look for an exact match.
850 	 */
851 	head = &pcbinfo->hashbase[INP_PCBCONNHASH(faddr.s_addr, fport,
852 	    laddr.s_addr, lport, pcbinfo->hashmask)];
853 	LIST_FOREACH(inp, head, inp_hash) {
854 #ifdef INET6
855 		if (!(inp->inp_vflag & INP_IPV4))
856 			continue;
857 #endif
858 		if (in_hosteq(inp->inp_faddr, faddr) &&
859 		    in_hosteq(inp->inp_laddr, laddr) &&
860 		    inp->inp_fport == fport && inp->inp_lport == lport) {
861 		    	/* found */
862 			return (inp);
863 		}
864 	}
865 
866 	if (wildcard) {
867 		struct inpcb *local_wild = NULL;
868 #ifdef INET6
869 		struct inpcb *local_wild_mapped = NULL;
870 #endif
871 
872 		head = &pcbinfo->bindhashbase[INP_PCBBINDHASH(lport,
873 		    pcbinfo->bindhashmask)];
874 		LIST_FOREACH(inp, head, inp_hash) {
875 #ifdef INET6
876 			if (!(inp->inp_vflag & INP_IPV4))
877 				continue;
878 #endif
879 			if (inp->inp_lport == lport) {
880 				if (ifp && ifp->if_type == IFT_FAITH &&
881 				    !(inp->inp_flags & INP_FAITH))
882 					continue;
883 				if (inp->inp_laddr.s_addr == laddr.s_addr)
884 					return (inp);
885 				if (inp->inp_laddr.s_addr == INADDR_ANY) {
886 #ifdef INET6
887 					if (INP_CHECK_SOCKAF(inp->inp_socket,
888 							     AF_INET6))
889 						local_wild_mapped = inp;
890 					else
891 #endif
892 						local_wild = inp;
893 				}
894 			}
895 		}
896 #ifdef INET6
897 		if (local_wild == NULL)
898 			return (local_wild_mapped);
899 #endif
900 		return (local_wild);
901 	}
902 
903 	/*
904 	 * Not found.
905 	 */
906 	return (NULL);
907 }
908 
909 /*
910  * Insert PCB into connection hash table.
911  */
912 void
913 in_pcbinsconnhash(struct inpcb *inp)
914 {
915 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
916 	struct inpcbhead *bucket;
917 	u_int32_t hashkey_faddr, hashkey_laddr;
918 
919 #ifdef INET6
920 	if (inp->inp_vflag & INP_IPV6) {
921 		hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX JH */;
922 		hashkey_laddr = inp->in6p_laddr.s6_addr32[3] /* XXX JH */;
923 	} else {
924 #endif
925 		hashkey_faddr = inp->inp_faddr.s_addr;
926 		hashkey_laddr = inp->inp_laddr.s_addr;
927 #ifdef INET6
928 	}
929 #endif
930 
931 	inp->inp_flags |= INP_CONNECTED;
932 
933 	/*
934 	 * Insert into the connection hash table.
935 	 */
936 	bucket = &pcbinfo->hashbase[INP_PCBCONNHASH(hashkey_faddr,
937 	    inp->inp_fport, hashkey_laddr,
938 	    inp->inp_lport, pcbinfo->hashmask)];
939 	LIST_INSERT_HEAD(bucket, inp, inp_hash);
940 }
941 
942 /*
943  * Remove PCB from connection hash table.
944  */
945 void
946 in_pcbremconnhash(struct inpcb *inp)
947 {
948 	KASSERT(inp->inp_flags & INP_CONNECTED, ("inp not connected"));
949 	LIST_REMOVE(inp, inp_hash);
950 	inp->inp_flags &= ~INP_CONNECTED;
951 }
952 
953 /*
954  * Insert PCB into port hash table.
955  */
956 int
957 in_pcbinsporthash(struct inpcb *inp)
958 {
959 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
960 	struct inpcbporthead *pcbporthash;
961 	struct inpcbport *phd;
962 
963 	/*
964 	 * Insert into the port hash table.
965 	 */
966 	pcbporthash = &pcbinfo->porthashbase[
967 	    INP_PCBPORTHASH(inp->inp_lport, pcbinfo->porthashmask)];
968 
969 	/* Go through port list and look for a head for this lport. */
970 	LIST_FOREACH(phd, pcbporthash, phd_hash)
971 		if (phd->phd_port == inp->inp_lport)
972 			break;
973 
974 	/* If none exists, malloc one and tack it on. */
975 	if (phd == NULL) {
976 		MALLOC(phd, struct inpcbport *, sizeof(struct inpcbport),
977 		    M_PCB, M_NOWAIT);
978 		if (phd == NULL)
979 			return (ENOBUFS); /* XXX */
980 		phd->phd_port = inp->inp_lport;
981 		LIST_INIT(&phd->phd_pcblist);
982 		LIST_INSERT_HEAD(pcbporthash, phd, phd_hash);
983 	}
984 
985 	inp->inp_phd = phd;
986 	LIST_INSERT_HEAD(&phd->phd_pcblist, inp, inp_portlist);
987 
988 	return (0);
989 }
990 
991 /*
992  * Insert PCB into bind hash table.
993  */
994 void
995 in_pcbinsbindhash(struct inpcb *inp)
996 {
997 	struct inpcbhead *bucket;
998 
999 	bucket = &inp->inp_pcbinfo->bindhashbase[INP_PCBBINDHASH(inp->inp_lport,
1000 	    inp->inp_pcbinfo->porthashmask)];
1001 
1002 	inp->inp_flags |= INP_BOUND;
1003 	LIST_INSERT_HEAD(bucket, inp, inp_hash);
1004 }
1005 
1006 /*
1007  * Remove PCB from bind hash table.
1008  */
1009 void
1010 in_pcbrembindhash(struct inpcb *inp)
1011 {
1012 	KASSERT(inp->inp_flags & INP_BOUND, ("inp not bound"));
1013 	LIST_REMOVE(inp, inp_hash);
1014 	inp->inp_flags &= ~INP_BOUND;
1015 }
1016 
1017 static void
1018 in_pcbremhash(struct inpcb *inp)
1019 {
1020 	if (inp->inp_flags & (INP_BOUND | INP_CONNECTED)) {
1021 		LIST_REMOVE(inp, inp_hash);
1022 		inp->inp_flags &= ~(INP_BOUND | INP_CONNECTED);
1023 	}
1024 }
1025 
1026 void
1027 in_pcbrehash(struct inpcb *inp, int state)
1028 {
1029 	in_pcbremhash(inp);
1030 	if (state == INP_BOUND)
1031 		in_pcbinsbindhash(inp);
1032 	else
1033 		in_pcbinsconnhash(inp);
1034 }
1035 
1036 /*
1037  * Remove PCB from various lists.
1038  */
1039 void
1040 in_pcbremlists(inp)
1041 	struct inpcb *inp;
1042 {
1043 	if (inp->inp_lport) {
1044 		struct inpcbport *phd = inp->inp_phd;
1045 
1046 		LIST_REMOVE(inp, inp_portlist);
1047 		if (LIST_FIRST(&phd->phd_pcblist) == NULL) {
1048 			LIST_REMOVE(phd, phd_hash);
1049 			free(phd, M_PCB);
1050 		}
1051 	}
1052 	in_pcbremhash(inp);
1053 	LIST_REMOVE(inp, inp_list);
1054 	inp->inp_pcbinfo->ipi_count--;
1055 }
1056 
1057 int
1058 prison_xinpcb(struct thread *td, struct inpcb *inp)
1059 {
1060 	struct ucred *cr;
1061 
1062 	if (td->td_proc == NULL)
1063 		return (0);
1064 	cr = td->td_proc->p_ucred;
1065 	if (cr->cr_prison == NULL)
1066 		return (0);
1067 	if (ntohl(inp->inp_laddr.s_addr) == cr->cr_prison->pr_ip)
1068 		return (0);
1069 	return (1);
1070 }
1071