xref: /dragonfly/sys/netinet/in_pcb.c (revision 1465342b)
1 /*
2  * Copyright (c) 2004 Jeffrey M. Hsu.  All rights reserved.
3  * Copyright (c) 2004 The DragonFly Project.  All rights reserved.
4  *
5  * This code is derived from software contributed to The DragonFly Project
6  * by Jeffrey M. Hsu.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of The DragonFly Project nor the names of its
17  *    contributors may be used to endorse or promote products derived
18  *    from this software without specific, prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
21  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
22  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
23  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE
24  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
25  * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
26  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
27  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
28  * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
29  * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
30  * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  */
33 
34 /*
35  * Copyright (c) 1982, 1986, 1991, 1993, 1995
36  *	The Regents of the University of California.  All rights reserved.
37  *
38  * Redistribution and use in source and binary forms, with or without
39  * modification, are permitted provided that the following conditions
40  * are met:
41  * 1. Redistributions of source code must retain the above copyright
42  *    notice, this list of conditions and the following disclaimer.
43  * 2. Redistributions in binary form must reproduce the above copyright
44  *    notice, this list of conditions and the following disclaimer in the
45  *    documentation and/or other materials provided with the distribution.
46  * 3. All advertising materials mentioning features or use of this software
47  *    must display the following acknowledgement:
48  *	This product includes software developed by the University of
49  *	California, Berkeley and its contributors.
50  * 4. Neither the name of the University nor the names of its contributors
51  *    may be used to endorse or promote products derived from this software
52  *    without specific prior written permission.
53  *
54  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
55  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
56  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
57  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
58  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
59  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
60  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
61  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
62  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
63  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
64  * SUCH DAMAGE.
65  *
66  *	@(#)in_pcb.c	8.4 (Berkeley) 5/24/95
67  * $FreeBSD: src/sys/netinet/in_pcb.c,v 1.59.2.27 2004/01/02 04:06:42 ambrisko Exp $
68  * $DragonFly: src/sys/netinet/in_pcb.c,v 1.44 2007/12/19 11:10:42 sephe Exp $
69  */
70 
71 #include "opt_ipsec.h"
72 #include "opt_inet6.h"
73 
74 #include <sys/param.h>
75 #include <sys/systm.h>
76 #include <sys/malloc.h>
77 #include <sys/mbuf.h>
78 #include <sys/domain.h>
79 #include <sys/protosw.h>
80 #include <sys/socket.h>
81 #include <sys/socketvar.h>
82 #include <sys/proc.h>
83 #include <sys/jail.h>
84 #include <sys/kernel.h>
85 #include <sys/sysctl.h>
86 #include <sys/thread2.h>
87 
88 #include <machine/limits.h>
89 
90 #include <vm/vm_zone.h>
91 
92 #include <net/if.h>
93 #include <net/if_types.h>
94 #include <net/route.h>
95 
96 #include <netinet/in.h>
97 #include <netinet/in_pcb.h>
98 #include <netinet/in_var.h>
99 #include <netinet/ip_var.h>
100 #ifdef INET6
101 #include <netinet/ip6.h>
102 #include <netinet6/ip6_var.h>
103 #endif /* INET6 */
104 
105 #ifdef IPSEC
106 #include <netinet6/ipsec.h>
107 #include <netproto/key/key.h>
108 #endif
109 
110 #ifdef FAST_IPSEC
111 #if defined(IPSEC) || defined(IPSEC_ESP)
112 #error "Bad idea: don't compile with both IPSEC and FAST_IPSEC!"
113 #endif
114 
115 #include <netproto/ipsec/ipsec.h>
116 #include <netproto/ipsec/key.h>
117 #define	IPSEC
118 #endif /* FAST_IPSEC */
119 
120 struct in_addr zeroin_addr;
121 
122 /*
123  * These configure the range of local port addresses assigned to
124  * "unspecified" outgoing connections/packets/whatever.
125  */
126 int ipport_lowfirstauto = IPPORT_RESERVED - 1;	/* 1023 */
127 int ipport_lowlastauto = IPPORT_RESERVEDSTART;	/* 600 */
128 
129 int ipport_firstauto = IPPORT_RESERVED;		/* 1024 */
130 int ipport_lastauto = IPPORT_USERRESERVED;	/* 5000 */
131 
132 int ipport_hifirstauto = IPPORT_HIFIRSTAUTO;	/* 49152 */
133 int ipport_hilastauto = IPPORT_HILASTAUTO;	/* 65535 */
134 
135 static __inline void
136 RANGECHK(int var, int min, int max)
137 {
138 	if (var < min)
139 		var = min;
140 	else if (var > max)
141 		var = max;
142 }
143 
144 static int
145 sysctl_net_ipport_check(SYSCTL_HANDLER_ARGS)
146 {
147 	int error;
148 
149 	error = sysctl_handle_int(oidp, oidp->oid_arg1, oidp->oid_arg2, req);
150 	if (!error) {
151 		RANGECHK(ipport_lowfirstauto, 1, IPPORT_RESERVED - 1);
152 		RANGECHK(ipport_lowlastauto, 1, IPPORT_RESERVED - 1);
153 
154 		RANGECHK(ipport_firstauto, IPPORT_RESERVED, USHRT_MAX);
155 		RANGECHK(ipport_lastauto, IPPORT_RESERVED, USHRT_MAX);
156 
157 		RANGECHK(ipport_hifirstauto, IPPORT_RESERVED, USHRT_MAX);
158 		RANGECHK(ipport_hilastauto, IPPORT_RESERVED, USHRT_MAX);
159 	}
160 	return (error);
161 }
162 
163 SYSCTL_NODE(_net_inet_ip, IPPROTO_IP, portrange, CTLFLAG_RW, 0, "IP Ports");
164 
165 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, lowfirst, CTLTYPE_INT|CTLFLAG_RW,
166 	   &ipport_lowfirstauto, 0, &sysctl_net_ipport_check, "I", "");
167 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, lowlast, CTLTYPE_INT|CTLFLAG_RW,
168 	   &ipport_lowlastauto, 0, &sysctl_net_ipport_check, "I", "");
169 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, first, CTLTYPE_INT|CTLFLAG_RW,
170 	   &ipport_firstauto, 0, &sysctl_net_ipport_check, "I", "");
171 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, last, CTLTYPE_INT|CTLFLAG_RW,
172 	   &ipport_lastauto, 0, &sysctl_net_ipport_check, "I", "");
173 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, hifirst, CTLTYPE_INT|CTLFLAG_RW,
174 	   &ipport_hifirstauto, 0, &sysctl_net_ipport_check, "I", "");
175 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, hilast, CTLTYPE_INT|CTLFLAG_RW,
176 	   &ipport_hilastauto, 0, &sysctl_net_ipport_check, "I", "");
177 
178 /*
179  * in_pcb.c: manage the Protocol Control Blocks.
180  *
181  * NOTE: It is assumed that most of these functions will be called from
182  * a critical section.  XXX - There are, unfortunately, a few exceptions
183  * to this rule that should be fixed.
184  *
185  * NOTE: The caller should initialize the cpu field to the cpu running the
186  * protocol stack associated with this inpcbinfo.
187  */
188 
189 void
190 in_pcbinfo_init(struct inpcbinfo *pcbinfo)
191 {
192 	LIST_INIT(&pcbinfo->pcblisthead);
193 	pcbinfo->cpu = -1;
194 }
195 
196 /*
197  * Allocate a PCB and associate it with the socket.
198  */
199 int
200 in_pcballoc(struct socket *so, struct inpcbinfo *pcbinfo)
201 {
202 	struct inpcb *inp;
203 #ifdef IPSEC
204 	int error;
205 #endif
206 
207 	inp = zalloc(pcbinfo->ipi_zone);
208 	if (inp == NULL)
209 		return (ENOBUFS);
210 	bzero(inp, sizeof *inp);
211 	inp->inp_gencnt = ++pcbinfo->ipi_gencnt;
212 	inp->inp_pcbinfo = inp->inp_cpcbinfo = pcbinfo;
213 	inp->inp_socket = so;
214 #ifdef IPSEC
215 	error = ipsec_init_policy(so, &inp->inp_sp);
216 	if (error != 0) {
217 		zfree(pcbinfo->ipi_zone, inp);
218 		return (error);
219 	}
220 #endif
221 #ifdef INET6
222 	if (INP_SOCKAF(so) == AF_INET6 && ip6_v6only)
223 		inp->inp_flags |= IN6P_IPV6_V6ONLY;
224 	if (ip6_auto_flowlabel)
225 		inp->inp_flags |= IN6P_AUTOFLOWLABEL;
226 #endif
227 	so->so_pcb = inp;
228 	LIST_INSERT_HEAD(&pcbinfo->pcblisthead, inp, inp_list);
229 	pcbinfo->ipi_count++;
230 	return (0);
231 }
232 
233 int
234 in_pcbbind(struct inpcb *inp, struct sockaddr *nam, struct thread *td)
235 {
236 	struct socket *so = inp->inp_socket;
237 	struct proc *p = td->td_proc;
238 	unsigned short *lastport;
239 	struct sockaddr_in *sin;
240 	struct sockaddr_in jsin;
241 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
242 	struct ucred *cred = NULL;
243 	u_short lport = 0;
244 	int wild = 0, reuseport = (so->so_options & SO_REUSEPORT);
245 	int error;
246 
247 	KKASSERT(p);
248 
249 	if (TAILQ_EMPTY(&in_ifaddrhead)) /* XXX broken! */
250 		return (EADDRNOTAVAIL);
251 	if (inp->inp_lport != 0 || inp->inp_laddr.s_addr != INADDR_ANY)
252 		return (EINVAL);	/* already bound */
253 	if (!(so->so_options & (SO_REUSEADDR|SO_REUSEPORT)))
254 		wild = 1;    /* neither SO_REUSEADDR nor SO_REUSEPORT is set */
255 	if (p)
256 		cred = p->p_ucred;
257 	if (nam != NULL) {
258 		sin = (struct sockaddr_in *)nam;
259 		if (nam->sa_len != sizeof *sin)
260 			return (EINVAL);
261 #ifdef notdef
262 		/*
263 		 * We should check the family, but old programs
264 		 * incorrectly fail to initialize it.
265 		 */
266 		if (sin->sin_family != AF_INET)
267 			return (EAFNOSUPPORT);
268 #endif
269 		if (!prison_replace_wildcards(td, nam))
270 				return (EINVAL);
271 		lport = sin->sin_port;
272 		if (IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) {
273 			/*
274 			 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
275 			 * allow complete duplication of binding if
276 			 * SO_REUSEPORT is set, or if SO_REUSEADDR is set
277 			 * and a multicast address is bound on both
278 			 * new and duplicated sockets.
279 			 */
280 			if (so->so_options & SO_REUSEADDR)
281 				reuseport = SO_REUSEADDR | SO_REUSEPORT;
282 		} else if (sin->sin_addr.s_addr != INADDR_ANY) {
283 			sin->sin_port = 0;		/* yech... */
284 			bzero(&sin->sin_zero, sizeof sin->sin_zero);
285 			if (ifa_ifwithaddr((struct sockaddr *)sin) == NULL)
286 				return (EADDRNOTAVAIL);
287 		}
288 		if (lport != 0) {
289 			struct inpcb *t;
290 
291 			/* GROSS */
292 			if (ntohs(lport) < IPPORT_RESERVED &&
293 			    cred && suser_cred(cred, PRISON_ROOT))
294 				return (EACCES);
295 			if (so->so_cred->cr_uid != 0 &&
296 			    !IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) {
297 				t = in_pcblookup_local(inp->inp_pcbinfo,
298 				    sin->sin_addr, lport,
299 				    INPLOOKUP_WILDCARD, cred);
300 				if (t &&
301 				    (!in_nullhost(sin->sin_addr) ||
302 				     !in_nullhost(t->inp_laddr) ||
303 				     (t->inp_socket->so_options &
304 					 SO_REUSEPORT) == 0) &&
305 				    (so->so_cred->cr_uid !=
306 				     t->inp_socket->so_cred->cr_uid)) {
307 #ifdef INET6
308 					if (!in_nullhost(sin->sin_addr) ||
309 					    !in_nullhost(t->inp_laddr) ||
310 					    INP_SOCKAF(so) ==
311 					    INP_SOCKAF(t->inp_socket))
312 #endif
313 					return (EADDRINUSE);
314 				}
315 			}
316 			if (cred && !prison_replace_wildcards(td, nam))
317 				return (EADDRNOTAVAIL);
318 			t = in_pcblookup_local(pcbinfo, sin->sin_addr, lport,
319 					       wild, cred);
320 			if (t && !(reuseport & t->inp_socket->so_options)) {
321 #ifdef INET6
322 				if (!in_nullhost(sin->sin_addr) ||
323 				    !in_nullhost(t->inp_laddr) ||
324 				    INP_SOCKAF(so) == INP_SOCKAF(t->inp_socket))
325 #endif
326 				return (EADDRINUSE);
327 			}
328 		}
329 		inp->inp_laddr = sin->sin_addr;
330 	}
331 	if (lport == 0) {
332 		ushort first, last;
333 		int count;
334 
335 		jsin.sin_family = AF_INET;
336 		jsin.sin_addr.s_addr = inp->inp_laddr.s_addr;
337 		if (!prison_replace_wildcards(td, (struct sockaddr *)&jsin)) {
338 			inp->inp_laddr.s_addr = INADDR_ANY;
339 			return (EINVAL);
340 		}
341 		inp->inp_laddr.s_addr = jsin.sin_addr.s_addr;
342 
343 		inp->inp_flags |= INP_ANONPORT;
344 
345 		if (inp->inp_flags & INP_HIGHPORT) {
346 			first = ipport_hifirstauto;	/* sysctl */
347 			last  = ipport_hilastauto;
348 			lastport = &pcbinfo->lasthi;
349 		} else if (inp->inp_flags & INP_LOWPORT) {
350 			if (cred &&
351 			    (error = suser_cred(cred, PRISON_ROOT))) {
352 				inp->inp_laddr.s_addr = INADDR_ANY;
353 				return (error);
354 			}
355 			first = ipport_lowfirstauto;	/* 1023 */
356 			last  = ipport_lowlastauto;	/* 600 */
357 			lastport = &pcbinfo->lastlow;
358 		} else {
359 			first = ipport_firstauto;	/* sysctl */
360 			last  = ipport_lastauto;
361 			lastport = &pcbinfo->lastport;
362 		}
363 		/*
364 		 * Simple check to ensure all ports are not used up causing
365 		 * a deadlock here.
366 		 *
367 		 * We split the two cases (up and down) so that the direction
368 		 * is not being tested on each round of the loop.
369 		 */
370 		if (first > last) {
371 			/*
372 			 * counting down
373 			 */
374 			count = first - last;
375 
376 			do {
377 				if (count-- < 0) {	/* completely used? */
378 					inp->inp_laddr.s_addr = INADDR_ANY;
379 					return (EADDRNOTAVAIL);
380 				}
381 				--*lastport;
382 				if (*lastport > first || *lastport < last)
383 					*lastport = first;
384 				lport = htons(*lastport);
385 			} while (in_pcblookup_local(pcbinfo, inp->inp_laddr,
386 				 lport, wild, cred));
387 		} else {
388 			/*
389 			 * counting up
390 			 */
391 			count = last - first;
392 
393 			do {
394 				if (count-- < 0) {	/* completely used? */
395 					inp->inp_laddr.s_addr = INADDR_ANY;
396 					return (EADDRNOTAVAIL);
397 				}
398 				++*lastport;
399 				if (*lastport < first || *lastport > last)
400 					*lastport = first;
401 				lport = htons(*lastport);
402 			} while (in_pcblookup_local(pcbinfo, inp->inp_laddr,
403 				 lport, wild, cred));
404 		}
405 	}
406 	inp->inp_lport = lport;
407 
408 	jsin.sin_family = AF_INET;
409 	jsin.sin_addr.s_addr = inp->inp_laddr.s_addr;
410 	if (!prison_replace_wildcards(td, (struct sockaddr*)&jsin)) {
411 		inp->inp_laddr.s_addr = INADDR_ANY;
412 		inp->inp_lport = 0;
413 		return (EINVAL);
414 	}
415 	inp->inp_laddr.s_addr = jsin.sin_addr.s_addr;
416 
417 	if (in_pcbinsporthash(inp) != 0) {
418 		inp->inp_laddr.s_addr = INADDR_ANY;
419 		inp->inp_lport = 0;
420 		return (EAGAIN);
421 	}
422 	return (0);
423 }
424 
425 /*
426  *   Transform old in_pcbconnect() into an inner subroutine for new
427  *   in_pcbconnect(): Do some validity-checking on the remote
428  *   address (in mbuf 'nam') and then determine local host address
429  *   (i.e., which interface) to use to access that remote host.
430  *
431  *   This preserves definition of in_pcbconnect(), while supporting a
432  *   slightly different version for T/TCP.  (This is more than
433  *   a bit of a kludge, but cleaning up the internal interfaces would
434  *   have forced minor changes in every protocol).
435  */
436 int
437 in_pcbladdr(struct inpcb *inp, struct sockaddr *nam,
438 	struct sockaddr_in **plocal_sin, struct thread *td)
439 {
440 	struct in_ifaddr *ia;
441 	struct ucred *cred = NULL;
442 	struct sockaddr_in *sin = (struct sockaddr_in *)nam;
443 	struct sockaddr *jsin;
444 	int jailed = 0;
445 
446 	if (nam->sa_len != sizeof *sin)
447 		return (EINVAL);
448 	if (sin->sin_family != AF_INET)
449 		return (EAFNOSUPPORT);
450 	if (sin->sin_port == 0)
451 		return (EADDRNOTAVAIL);
452 	if (td && td->td_proc && td->td_proc->p_ucred)
453 		cred = td->td_proc->p_ucred;
454 	if (cred && cred->cr_prison)
455 		jailed = 1;
456 	if (!TAILQ_EMPTY(&in_ifaddrhead)) {
457 		ia = TAILQ_FIRST(&in_ifaddrhead);
458 		/*
459 		 * If the destination address is INADDR_ANY,
460 		 * use the primary local address.
461 		 * If the supplied address is INADDR_BROADCAST,
462 		 * and the primary interface supports broadcast,
463 		 * choose the broadcast address for that interface.
464 		 */
465 		if (sin->sin_addr.s_addr == INADDR_ANY)
466 			sin->sin_addr = IA_SIN(ia)->sin_addr;
467 		else if (sin->sin_addr.s_addr == (u_long)INADDR_BROADCAST &&
468 		    (ia->ia_ifp->if_flags & IFF_BROADCAST))
469 			sin->sin_addr = satosin(&ia->ia_broadaddr)->sin_addr;
470 	}
471 	if (inp->inp_laddr.s_addr == INADDR_ANY) {
472 		struct route *ro;
473 
474 		ia = (struct in_ifaddr *)NULL;
475 		/*
476 		 * If route is known or can be allocated now,
477 		 * our src addr is taken from the i/f, else punt.
478 		 * Note that we should check the address family of the cached
479 		 * destination, in case of sharing the cache with IPv6.
480 		 */
481 		ro = &inp->inp_route;
482 		if (ro->ro_rt &&
483 		    (!(ro->ro_rt->rt_flags & RTF_UP) ||
484 		     ro->ro_dst.sa_family != AF_INET ||
485 		     satosin(&ro->ro_dst)->sin_addr.s_addr !=
486 				      sin->sin_addr.s_addr ||
487 		     inp->inp_socket->so_options & SO_DONTROUTE)) {
488 			RTFREE(ro->ro_rt);
489 			ro->ro_rt = (struct rtentry *)NULL;
490 		}
491 		if (!(inp->inp_socket->so_options & SO_DONTROUTE) && /*XXX*/
492 		    (ro->ro_rt == (struct rtentry *)NULL ||
493 		    ro->ro_rt->rt_ifp == (struct ifnet *)NULL)) {
494 			/* No route yet, so try to acquire one */
495 			bzero(&ro->ro_dst, sizeof(struct sockaddr_in));
496 			ro->ro_dst.sa_family = AF_INET;
497 			ro->ro_dst.sa_len = sizeof(struct sockaddr_in);
498 			((struct sockaddr_in *) &ro->ro_dst)->sin_addr =
499 				sin->sin_addr;
500 			rtalloc(ro);
501 		}
502 		/*
503 		 * If we found a route, use the address
504 		 * corresponding to the outgoing interface
505 		 * unless it is the loopback (in case a route
506 		 * to our address on another net goes to loopback).
507 		 */
508 		if (ro->ro_rt && !(ro->ro_rt->rt_ifp->if_flags & IFF_LOOPBACK)) {
509 			if (jailed) {
510 				if (jailed_ip(cred->cr_prison,
511 				    ro->ro_rt->rt_ifa->ifa_addr)) {
512 					ia = ifatoia(ro->ro_rt->rt_ifa);
513 				}
514 			} else {
515 				ia = ifatoia(ro->ro_rt->rt_ifa);
516 			}
517 		}
518 		if (ia == NULL) {
519 			u_short fport = sin->sin_port;
520 
521 			sin->sin_port = 0;
522 			ia = ifatoia(ifa_ifwithdstaddr(sintosa(sin)));
523 			if (ia && jailed && !jailed_ip(cred->cr_prison,
524 			    sintosa(&ia->ia_addr)))
525 				ia = NULL;
526 			if (ia == NULL)
527 				ia = ifatoia(ifa_ifwithnet(sintosa(sin)));
528 			if (ia && jailed && !jailed_ip(cred->cr_prison,
529 			    sintosa(&ia->ia_addr)))
530 				ia = NULL;
531 			sin->sin_port = fport;
532 			if (ia == NULL)
533 				ia = TAILQ_FIRST(&in_ifaddrhead);
534 			if (ia && jailed && !jailed_ip(cred->cr_prison,
535 			    sintosa(&ia->ia_addr)))
536 				ia = NULL;
537 
538 			if (!jailed && ia == NULL)
539 				return (EADDRNOTAVAIL);
540 		}
541 		/*
542 		 * If the destination address is multicast and an outgoing
543 		 * interface has been set as a multicast option, use the
544 		 * address of that interface as our source address.
545 		 */
546 		if (!jailed && IN_MULTICAST(ntohl(sin->sin_addr.s_addr)) &&
547 		    inp->inp_moptions != NULL) {
548 			struct ip_moptions *imo;
549 			struct ifnet *ifp;
550 
551 			imo = inp->inp_moptions;
552 			if (imo->imo_multicast_ifp != NULL) {
553 				ifp = imo->imo_multicast_ifp;
554 				TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link)
555 					if (ia->ia_ifp == ifp)
556 						break;
557 				if (ia == NULL)
558 					return (EADDRNOTAVAIL);
559 			}
560 		}
561 		/*
562 		 * Don't do pcblookup call here; return interface in plocal_sin
563 		 * and exit to caller, that will do the lookup.
564 		 */
565 		if (ia == NULL && jailed) {
566 			if ((jsin = prison_get_nonlocal(cred->cr_prison, AF_INET, NULL)) != NULL ||
567 			    (jsin = prison_get_local(cred->cr_prison, AF_INET, NULL)) != NULL)
568 				*plocal_sin = satosin(jsin);
569 			else
570 				/* IPv6 only Jail */
571 				return (EADDRNOTAVAIL);
572 		} else {
573 			*plocal_sin = &ia->ia_addr;
574 		}
575 	}
576 	return (0);
577 }
578 
579 /*
580  * Outer subroutine:
581  * Connect from a socket to a specified address.
582  * Both address and port must be specified in argument sin.
583  * If don't have a local address for this socket yet,
584  * then pick one.
585  */
586 int
587 in_pcbconnect(struct inpcb *inp, struct sockaddr *nam, struct thread *td)
588 {
589 	struct sockaddr_in *if_sin;
590 	struct sockaddr_in *sin = (struct sockaddr_in *)nam;
591 	int error;
592 
593 	/* Call inner routine to assign local interface address. */
594 	if ((error = in_pcbladdr(inp, nam, &if_sin, td)) != 0)
595 		return (error);
596 
597 	if (in_pcblookup_hash(inp->inp_cpcbinfo, sin->sin_addr, sin->sin_port,
598 	    inp->inp_laddr.s_addr ? inp->inp_laddr : if_sin->sin_addr,
599 	    inp->inp_lport, FALSE, NULL) != NULL) {
600 		return (EADDRINUSE);
601 	}
602 	if (inp->inp_laddr.s_addr == INADDR_ANY) {
603 		if (inp->inp_lport == 0) {
604 			error = in_pcbbind(inp, (struct sockaddr *)NULL, td);
605 			if (error)
606 				return (error);
607 		}
608 		inp->inp_laddr = if_sin->sin_addr;
609 	}
610 	inp->inp_faddr = sin->sin_addr;
611 	inp->inp_fport = sin->sin_port;
612 	in_pcbinsconnhash(inp);
613 	return (0);
614 }
615 
616 void
617 in_pcbdisconnect(struct inpcb *inp)
618 {
619 
620 	inp->inp_faddr.s_addr = INADDR_ANY;
621 	inp->inp_fport = 0;
622 	in_pcbremconnhash(inp);
623 	if (inp->inp_socket->so_state & SS_NOFDREF)
624 		in_pcbdetach(inp);
625 }
626 
627 void
628 in_pcbdetach(struct inpcb *inp)
629 {
630 	struct socket *so = inp->inp_socket;
631 	struct inpcbinfo *ipi = inp->inp_pcbinfo;
632 
633 #ifdef IPSEC
634 	ipsec4_delete_pcbpolicy(inp);
635 #endif /*IPSEC*/
636 	inp->inp_gencnt = ++ipi->ipi_gencnt;
637 	in_pcbremlists(inp);
638 	so->so_pcb = 0;
639 	sofree(so);
640 	if (inp->inp_options)
641 		m_free(inp->inp_options);
642 	if (inp->inp_route.ro_rt)
643 		rtfree(inp->inp_route.ro_rt);
644 	ip_freemoptions(inp->inp_moptions);
645 	inp->inp_vflag = 0;
646 	zfree(ipi->ipi_zone, inp);
647 }
648 
649 /*
650  * The calling convention of in_setsockaddr() and in_setpeeraddr() was
651  * modified to match the pru_sockaddr() and pru_peeraddr() entry points
652  * in struct pr_usrreqs, so that protocols can just reference then directly
653  * without the need for a wrapper function.  The socket must have a valid
654  * (i.e., non-nil) PCB, but it should be impossible to get an invalid one
655  * except through a kernel programming error, so it is acceptable to panic
656  * (or in this case trap) if the PCB is invalid.  (Actually, we don't trap
657  * because there actually /is/ a programming error somewhere... XXX)
658  */
659 int
660 in_setsockaddr(struct socket *so, struct sockaddr **nam)
661 {
662 	struct inpcb *inp;
663 	struct sockaddr_in *sin;
664 
665 	/*
666 	 * Do the malloc first in case it blocks.
667 	 */
668 	MALLOC(sin, struct sockaddr_in *, sizeof *sin, M_SONAME,
669 		M_WAITOK | M_ZERO);
670 	sin->sin_family = AF_INET;
671 	sin->sin_len = sizeof *sin;
672 
673 	crit_enter();
674 	inp = so->so_pcb;
675 	if (!inp) {
676 		crit_exit();
677 		kfree(sin, M_SONAME);
678 		return (ECONNRESET);
679 	}
680 	sin->sin_port = inp->inp_lport;
681 	sin->sin_addr = inp->inp_laddr;
682 	crit_exit();
683 
684 	*nam = (struct sockaddr *)sin;
685 	return (0);
686 }
687 
688 int
689 in_setpeeraddr(struct socket *so, struct sockaddr **nam)
690 {
691 	struct inpcb *inp;
692 	struct sockaddr_in *sin;
693 
694 	/*
695 	 * Do the malloc first in case it blocks.
696 	 */
697 	MALLOC(sin, struct sockaddr_in *, sizeof *sin, M_SONAME,
698 		M_WAITOK | M_ZERO);
699 	sin->sin_family = AF_INET;
700 	sin->sin_len = sizeof *sin;
701 
702 	crit_enter();
703 	inp = so->so_pcb;
704 	if (!inp) {
705 		crit_exit();
706 		kfree(sin, M_SONAME);
707 		return (ECONNRESET);
708 	}
709 	sin->sin_port = inp->inp_fport;
710 	sin->sin_addr = inp->inp_faddr;
711 	crit_exit();
712 
713 	*nam = (struct sockaddr *)sin;
714 	return (0);
715 }
716 
717 void
718 in_pcbnotifyall(struct inpcbhead *head, struct in_addr faddr, int err,
719 		void (*notify)(struct inpcb *, int))
720 {
721 	struct inpcb *inp, *ninp;
722 
723 	/*
724 	 * note: if INP_PLACEMARKER is set we must ignore the rest of
725 	 * the structure and skip it.
726 	 */
727 	crit_enter();
728 	LIST_FOREACH_MUTABLE(inp, head, inp_list, ninp) {
729 		if (inp->inp_flags & INP_PLACEMARKER)
730 			continue;
731 #ifdef INET6
732 		if (!(inp->inp_vflag & INP_IPV4))
733 			continue;
734 #endif
735 		if (inp->inp_faddr.s_addr != faddr.s_addr ||
736 		    inp->inp_socket == NULL)
737 			continue;
738 		(*notify)(inp, err);		/* can remove inp from list! */
739 	}
740 	crit_exit();
741 }
742 
743 void
744 in_pcbpurgeif0(struct inpcb *head, struct ifnet *ifp)
745 {
746 	struct inpcb *inp;
747 	struct ip_moptions *imo;
748 	int i, gap;
749 
750 	for (inp = head; inp != NULL; inp = LIST_NEXT(inp, inp_list)) {
751 		if (inp->inp_flags & INP_PLACEMARKER)
752 			continue;
753 		imo = inp->inp_moptions;
754 		if ((inp->inp_vflag & INP_IPV4) && imo != NULL) {
755 			/*
756 			 * Unselect the outgoing interface if it is being
757 			 * detached.
758 			 */
759 			if (imo->imo_multicast_ifp == ifp)
760 				imo->imo_multicast_ifp = NULL;
761 
762 			/*
763 			 * Drop multicast group membership if we joined
764 			 * through the interface being detached.
765 			 */
766 			for (i = 0, gap = 0; i < imo->imo_num_memberships;
767 			    i++) {
768 				if (imo->imo_membership[i]->inm_ifp == ifp) {
769 					in_delmulti(imo->imo_membership[i]);
770 					gap++;
771 				} else if (gap != 0)
772 					imo->imo_membership[i - gap] =
773 					    imo->imo_membership[i];
774 			}
775 			imo->imo_num_memberships -= gap;
776 		}
777 	}
778 }
779 
780 /*
781  * Check for alternatives when higher level complains
782  * about service problems.  For now, invalidate cached
783  * routing information.  If the route was created dynamically
784  * (by a redirect), time to try a default gateway again.
785  */
786 void
787 in_losing(struct inpcb *inp)
788 {
789 	struct rtentry *rt;
790 	struct rt_addrinfo rtinfo;
791 
792 	if ((rt = inp->inp_route.ro_rt)) {
793 		bzero(&rtinfo, sizeof(struct rt_addrinfo));
794 		rtinfo.rti_info[RTAX_DST] = rt_key(rt);
795 		rtinfo.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
796 		rtinfo.rti_info[RTAX_NETMASK] = rt_mask(rt);
797 		rtinfo.rti_flags = rt->rt_flags;
798 		rt_missmsg(RTM_LOSING, &rtinfo, rt->rt_flags, 0);
799 		if (rt->rt_flags & RTF_DYNAMIC)
800 			rtrequest1_global(RTM_DELETE, &rtinfo, NULL, NULL);
801 		inp->inp_route.ro_rt = NULL;
802 		rtfree(rt);
803 		/*
804 		 * A new route can be allocated
805 		 * the next time output is attempted.
806 		 */
807 	}
808 }
809 
810 /*
811  * After a routing change, flush old routing
812  * and allocate a (hopefully) better one.
813  */
814 void
815 in_rtchange(struct inpcb *inp, int err)
816 {
817 	if (inp->inp_route.ro_rt) {
818 		rtfree(inp->inp_route.ro_rt);
819 		inp->inp_route.ro_rt = NULL;
820 		/*
821 		 * A new route can be allocated the next time
822 		 * output is attempted.
823 		 */
824 	}
825 }
826 
827 /*
828  * Lookup a PCB based on the local address and port.
829  */
830 struct inpcb *
831 in_pcblookup_local(struct inpcbinfo *pcbinfo, struct in_addr laddr,
832 		   u_int lport_arg, int wild_okay, struct ucred *cred)
833 {
834 	struct inpcb *inp;
835 	int matchwild = 3, wildcard;
836 	u_short lport = lport_arg;
837 
838 	struct inpcbporthead *porthash;
839 	struct inpcbport *phd;
840 	struct inpcb *match = NULL;
841 
842 	/*
843 	 * Best fit PCB lookup.
844 	 *
845 	 * First see if this local port is in use by looking on the
846 	 * port hash list.
847 	 */
848 	porthash = &pcbinfo->porthashbase[INP_PCBPORTHASH(lport,
849 	    pcbinfo->porthashmask)];
850 	LIST_FOREACH(phd, porthash, phd_hash) {
851 		if (phd->phd_port == lport)
852 			break;
853 	}
854 	if (phd != NULL) {
855 		/*
856 		 * Port is in use by one or more PCBs. Look for best
857 		 * fit.
858 		 */
859 		LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) {
860 			wildcard = 0;
861 #ifdef INET6
862 			if ((inp->inp_vflag & INP_IPV4) == 0)
863 				continue;
864 #endif
865 			if (inp->inp_faddr.s_addr != INADDR_ANY)
866 				wildcard++;
867 			if (inp->inp_laddr.s_addr != INADDR_ANY) {
868 				if (laddr.s_addr == INADDR_ANY)
869 					wildcard++;
870 				else if (inp->inp_laddr.s_addr != laddr.s_addr)
871 					continue;
872 			} else {
873 				if (laddr.s_addr != INADDR_ANY)
874 					wildcard++;
875 			}
876 			if (wildcard && !wild_okay)
877 				continue;
878 			if (wildcard < matchwild &&
879 			    (cred == NULL ||
880 			     cred->cr_prison ==
881 					inp->inp_socket->so_cred->cr_prison)) {
882 				match = inp;
883 				matchwild = wildcard;
884 				if (matchwild == 0) {
885 					break;
886 				}
887 			}
888 		}
889 	}
890 	return (match);
891 }
892 
893 /*
894  * Lookup PCB in hash list.
895  */
896 struct inpcb *
897 in_pcblookup_hash(struct inpcbinfo *pcbinfo, struct in_addr faddr, u_int fport_arg,
898 		  struct in_addr laddr, u_int lport_arg, boolean_t wildcard,
899 		  struct ifnet *ifp)
900 {
901 	struct inpcbhead *head;
902 	struct inpcb *inp, *jinp=NULL;
903 	u_short fport = fport_arg, lport = lport_arg;
904 
905 	/*
906 	 * First look for an exact match.
907 	 */
908 	head = &pcbinfo->hashbase[INP_PCBCONNHASH(faddr.s_addr, fport,
909 	    laddr.s_addr, lport, pcbinfo->hashmask)];
910 	LIST_FOREACH(inp, head, inp_hash) {
911 #ifdef INET6
912 		if (!(inp->inp_vflag & INP_IPV4))
913 			continue;
914 #endif
915 		if (in_hosteq(inp->inp_faddr, faddr) &&
916 		    in_hosteq(inp->inp_laddr, laddr) &&
917 		    inp->inp_fport == fport && inp->inp_lport == lport) {
918 			/* found */
919 			if (inp->inp_socket == NULL ||
920 			    inp->inp_socket->so_cred->cr_prison == NULL) {
921 				return (inp);
922 			} else {
923 				if  (jinp == NULL)
924 					jinp = inp;
925 			}
926 		}
927 	}
928 	if (jinp != NULL)
929 		return (jinp);
930 	if (wildcard) {
931 		struct inpcb *local_wild = NULL;
932 		struct inpcb *jinp_wild = NULL;
933 #ifdef INET6
934 		struct inpcb *local_wild_mapped = NULL;
935 #endif
936 		struct inpcontainer *ic;
937 		struct inpcontainerhead *chead;
938 		struct sockaddr_in jsin;
939 		struct ucred *cred;
940 
941 		/*
942 		 * Order of socket selection:
943 		 * 1. non-jailed, non-wild.
944 		 * 2. non-jailed, wild.
945 		 * 3. jailed, non-wild.
946 		 * 4. jailed, wild.
947 		 */
948 		jsin.sin_family = AF_INET;
949 		chead = &pcbinfo->wildcardhashbase[
950 		    INP_PCBWILDCARDHASH(lport, pcbinfo->wildcardhashmask)];
951 		LIST_FOREACH(ic, chead, ic_list) {
952 			inp = ic->ic_inp;
953 			jsin.sin_addr.s_addr = laddr.s_addr;
954 #ifdef INET6
955 			if (!(inp->inp_vflag & INP_IPV4))
956 				continue;
957 #endif
958 			if (inp->inp_socket != NULL)
959 				cred = inp->inp_socket->so_cred;
960 			else
961 				cred = NULL;
962 			if (cred != NULL && jailed(cred)) {
963 				if (jinp != NULL)
964 					continue;
965 				else
966 					if (!jailed_ip(cred->cr_prison,
967 					    (struct sockaddr *)&jsin))
968 						continue;
969 			}
970 			if (inp->inp_lport == lport) {
971 				if (ifp && ifp->if_type == IFT_FAITH &&
972 				    !(inp->inp_flags & INP_FAITH))
973 					continue;
974 				if (inp->inp_laddr.s_addr == laddr.s_addr) {
975 					if (cred != NULL && jailed(cred))
976 						jinp = inp;
977 					else
978 						return (inp);
979 				}
980 				if (inp->inp_laddr.s_addr == INADDR_ANY) {
981 #ifdef INET6
982 					if (INP_CHECK_SOCKAF(inp->inp_socket,
983 							     AF_INET6))
984 						local_wild_mapped = inp;
985 					else
986 #endif
987 						if (cred != NULL &&
988 						    jailed(cred))
989 							jinp_wild = inp;
990 						else
991 							local_wild = inp;
992 				}
993 			}
994 		}
995 		if (local_wild != NULL)
996 			return (local_wild);
997 #ifdef INET6
998 		if (local_wild_mapped != NULL)
999 			return (local_wild_mapped);
1000 #endif
1001 		if (jinp != NULL)
1002 			return (jinp);
1003 		return (jinp_wild);
1004 	}
1005 
1006 	/*
1007 	 * Not found.
1008 	 */
1009 	return (NULL);
1010 }
1011 
1012 /*
1013  * Insert PCB into connection hash table.
1014  */
1015 void
1016 in_pcbinsconnhash(struct inpcb *inp)
1017 {
1018 	struct inpcbinfo *pcbinfo = inp->inp_cpcbinfo;
1019 	struct inpcbhead *bucket;
1020 	u_int32_t hashkey_faddr, hashkey_laddr;
1021 
1022 #ifdef INET6
1023 	if (inp->inp_vflag & INP_IPV6) {
1024 		hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX JH */;
1025 		hashkey_laddr = inp->in6p_laddr.s6_addr32[3] /* XXX JH */;
1026 	} else {
1027 #endif
1028 		hashkey_faddr = inp->inp_faddr.s_addr;
1029 		hashkey_laddr = inp->inp_laddr.s_addr;
1030 #ifdef INET6
1031 	}
1032 #endif
1033 
1034 	KASSERT(!(inp->inp_flags & INP_CONNECTED), ("already on hash list"));
1035 	inp->inp_flags |= INP_CONNECTED;
1036 
1037 	/*
1038 	 * Insert into the connection hash table.
1039 	 */
1040 	bucket = &pcbinfo->hashbase[INP_PCBCONNHASH(hashkey_faddr,
1041 	    inp->inp_fport, hashkey_laddr, inp->inp_lport, pcbinfo->hashmask)];
1042 	LIST_INSERT_HEAD(bucket, inp, inp_hash);
1043 }
1044 
1045 /*
1046  * Remove PCB from connection hash table.
1047  */
1048 void
1049 in_pcbremconnhash(struct inpcb *inp)
1050 {
1051 	KASSERT(inp->inp_flags & INP_CONNECTED, ("inp not connected"));
1052 	LIST_REMOVE(inp, inp_hash);
1053 	inp->inp_flags &= ~INP_CONNECTED;
1054 }
1055 
1056 /*
1057  * Insert PCB into port hash table.
1058  */
1059 int
1060 in_pcbinsporthash(struct inpcb *inp)
1061 {
1062 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1063 	struct inpcbporthead *pcbporthash;
1064 	struct inpcbport *phd;
1065 
1066 	/*
1067 	 * Insert into the port hash table.
1068 	 */
1069 	pcbporthash = &pcbinfo->porthashbase[
1070 	    INP_PCBPORTHASH(inp->inp_lport, pcbinfo->porthashmask)];
1071 
1072 	/* Go through port list and look for a head for this lport. */
1073 	LIST_FOREACH(phd, pcbporthash, phd_hash)
1074 		if (phd->phd_port == inp->inp_lport)
1075 			break;
1076 
1077 	/* If none exists, malloc one and tack it on. */
1078 	if (phd == NULL) {
1079 		MALLOC(phd, struct inpcbport *, sizeof(struct inpcbport),
1080 		    M_PCB, M_INTWAIT | M_NULLOK);
1081 		if (phd == NULL)
1082 			return (ENOBUFS); /* XXX */
1083 		phd->phd_port = inp->inp_lport;
1084 		LIST_INIT(&phd->phd_pcblist);
1085 		LIST_INSERT_HEAD(pcbporthash, phd, phd_hash);
1086 	}
1087 
1088 	inp->inp_phd = phd;
1089 	LIST_INSERT_HEAD(&phd->phd_pcblist, inp, inp_portlist);
1090 
1091 	return (0);
1092 }
1093 
1094 void
1095 in_pcbinswildcardhash_oncpu(struct inpcb *inp, struct inpcbinfo *pcbinfo)
1096 {
1097 	struct inpcontainer *ic;
1098 	struct inpcontainerhead *bucket;
1099 
1100 	bucket = &pcbinfo->wildcardhashbase[
1101 	    INP_PCBWILDCARDHASH(inp->inp_lport, pcbinfo->wildcardhashmask)];
1102 
1103 	ic = kmalloc(sizeof(struct inpcontainer), M_TEMP, M_INTWAIT);
1104 	ic->ic_inp = inp;
1105 	LIST_INSERT_HEAD(bucket, ic, ic_list);
1106 }
1107 
1108 /*
1109  * Insert PCB into wildcard hash table.
1110  */
1111 void
1112 in_pcbinswildcardhash(struct inpcb *inp)
1113 {
1114 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1115 
1116 	KKASSERT(pcbinfo != NULL);
1117 
1118 	in_pcbinswildcardhash_oncpu(inp, pcbinfo);
1119 	inp->inp_flags |= INP_WILDCARD;
1120 }
1121 
1122 void
1123 in_pcbremwildcardhash_oncpu(struct inpcb *inp, struct inpcbinfo *pcbinfo)
1124 {
1125 	struct inpcontainer *ic;
1126 	struct inpcontainerhead *head;
1127 
1128 	/* find bucket */
1129 	head = &pcbinfo->wildcardhashbase[
1130 	    INP_PCBWILDCARDHASH(inp->inp_lport, pcbinfo->wildcardhashmask)];
1131 
1132 	LIST_FOREACH(ic, head, ic_list) {
1133 		if (ic->ic_inp == inp)
1134 			goto found;
1135 	}
1136 	return;			/* not found! */
1137 
1138 found:
1139 	LIST_REMOVE(ic, ic_list);	/* remove container from bucket chain */
1140 	kfree(ic, M_TEMP);		/* deallocate container */
1141 }
1142 
1143 /*
1144  * Remove PCB from wildcard hash table.
1145  */
1146 void
1147 in_pcbremwildcardhash(struct inpcb *inp)
1148 {
1149 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1150 
1151 	KASSERT(inp->inp_flags & INP_WILDCARD, ("inp not wildcard"));
1152 	in_pcbremwildcardhash_oncpu(inp, pcbinfo);
1153 	inp->inp_flags &= ~INP_WILDCARD;
1154 }
1155 
1156 /*
1157  * Remove PCB from various lists.
1158  */
1159 void
1160 in_pcbremlists(struct inpcb *inp)
1161 {
1162 	if (inp->inp_lport) {
1163 		struct inpcbport *phd = inp->inp_phd;
1164 
1165 		LIST_REMOVE(inp, inp_portlist);
1166 		if (LIST_FIRST(&phd->phd_pcblist) == NULL) {
1167 			LIST_REMOVE(phd, phd_hash);
1168 			kfree(phd, M_PCB);
1169 		}
1170 	}
1171 	if (inp->inp_flags & INP_WILDCARD) {
1172 		in_pcbremwildcardhash(inp);
1173 	} else if (inp->inp_flags & INP_CONNECTED) {
1174 		in_pcbremconnhash(inp);
1175 	}
1176 	LIST_REMOVE(inp, inp_list);
1177 	inp->inp_pcbinfo->ipi_count--;
1178 }
1179 
1180 int
1181 prison_xinpcb(struct thread *td, struct inpcb *inp)
1182 {
1183 	struct ucred *cr;
1184 
1185 	if (td->td_proc == NULL)
1186 		return (0);
1187 	cr = td->td_proc->p_ucred;
1188 	if (cr->cr_prison == NULL)
1189 		return (0);
1190 	if (inp->inp_socket && inp->inp_socket->so_cred &&
1191 	    inp->inp_socket->so_cred->cr_prison &&
1192 	    cr->cr_prison == inp->inp_socket->so_cred->cr_prison)
1193 		return (0);
1194 	return (1);
1195 }
1196 
1197 int
1198 in_pcblist_global(SYSCTL_HANDLER_ARGS)
1199 {
1200 	struct inpcbinfo *pcbinfo = arg1;
1201 	struct inpcb *inp, *marker;
1202 	struct xinpcb xi;
1203 	int error, i, n;
1204 	inp_gen_t gencnt;
1205 
1206 	/*
1207 	 * The process of preparing the TCB list is too time-consuming and
1208 	 * resource-intensive to repeat twice on every request.
1209 	 */
1210 	if (req->oldptr == NULL) {
1211 		n = pcbinfo->ipi_count;
1212 		req->oldidx = (n + n/8 + 10) * sizeof(struct xinpcb);
1213 		return 0;
1214 	}
1215 
1216 	if (req->newptr != NULL)
1217 		return EPERM;
1218 
1219 	/*
1220 	 * OK, now we're committed to doing something.  Re-fetch ipi_count
1221 	 * after obtaining the generation count.
1222 	 */
1223 	gencnt = pcbinfo->ipi_gencnt;
1224 	n = pcbinfo->ipi_count;
1225 
1226 	marker = kmalloc(sizeof(struct inpcb), M_TEMP, M_WAITOK|M_ZERO);
1227 	marker->inp_flags |= INP_PLACEMARKER;
1228 	LIST_INSERT_HEAD(&pcbinfo->pcblisthead, marker, inp_list);
1229 
1230 	i = 0;
1231 	error = 0;
1232 
1233 	while ((inp = LIST_NEXT(marker, inp_list)) != NULL && i < n) {
1234 		LIST_REMOVE(marker, inp_list);
1235 		LIST_INSERT_AFTER(inp, marker, inp_list);
1236 
1237 		if (inp->inp_flags & INP_PLACEMARKER)
1238 			continue;
1239 		if (inp->inp_gencnt > gencnt)
1240 			continue;
1241 		if (prison_xinpcb(req->td, inp))
1242 			continue;
1243 		bzero(&xi, sizeof xi);
1244 		xi.xi_len = sizeof xi;
1245 		bcopy(inp, &xi.xi_inp, sizeof *inp);
1246 		if (inp->inp_socket)
1247 			sotoxsocket(inp->inp_socket, &xi.xi_socket);
1248 		if ((error = SYSCTL_OUT(req, &xi, sizeof xi)) != 0)
1249 			break;
1250 		++i;
1251 	}
1252 	LIST_REMOVE(marker, inp_list);
1253 	if (error == 0 && i < n) {
1254 		bzero(&xi, sizeof xi);
1255 		xi.xi_len = sizeof xi;
1256 		while (i < n) {
1257 			error = SYSCTL_OUT(req, &xi, sizeof xi);
1258 			++i;
1259 		}
1260 	}
1261 	kfree(marker, M_TEMP);
1262 	return(error);
1263 }
1264