xref: /dflybsd-src/sys/netinet/in_pcb.c (revision c9e3d8f96688a159959b1af2d4fef14b744173e3)
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 1.48 2008/11/08 03:38:23 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/priv.h>
84 #include <sys/jail.h>
85 #include <sys/kernel.h>
86 #include <sys/sysctl.h>
87 #include <sys/thread2.h>
88 
89 #include <machine/limits.h>
90 
91 #include <net/if.h>
92 #include <net/if_types.h>
93 #include <net/route.h>
94 
95 #include <netinet/in.h>
96 #include <netinet/in_pcb.h>
97 #include <netinet/in_var.h>
98 #include <netinet/ip_var.h>
99 #ifdef INET6
100 #include <netinet/ip6.h>
101 #include <netinet6/ip6_var.h>
102 #endif /* INET6 */
103 
104 #ifdef IPSEC
105 #include <netinet6/ipsec.h>
106 #include <netproto/key/key.h>
107 #endif
108 
109 #ifdef FAST_IPSEC
110 #if defined(IPSEC) || defined(IPSEC_ESP)
111 #error "Bad idea: don't compile with both IPSEC and FAST_IPSEC!"
112 #endif
113 
114 #include <netproto/ipsec/ipsec.h>
115 #include <netproto/ipsec/key.h>
116 #define	IPSEC
117 #endif /* FAST_IPSEC */
118 
119 struct in_addr zeroin_addr;
120 
121 /*
122  * These configure the range of local port addresses assigned to
123  * "unspecified" outgoing connections/packets/whatever.
124  */
125 int ipport_lowfirstauto = IPPORT_RESERVED - 1;	/* 1023 */
126 int ipport_lowlastauto = IPPORT_RESERVEDSTART;	/* 600 */
127 
128 int ipport_firstauto = IPPORT_RESERVED;		/* 1024 */
129 int ipport_lastauto = IPPORT_USERRESERVED;	/* 5000 */
130 
131 int ipport_hifirstauto = IPPORT_HIFIRSTAUTO;	/* 49152 */
132 int ipport_hilastauto = IPPORT_HILASTAUTO;	/* 65535 */
133 
134 static __inline void
135 RANGECHK(int var, int min, int max)
136 {
137 	if (var < min)
138 		var = min;
139 	else if (var > max)
140 		var = max;
141 }
142 
143 static int
144 sysctl_net_ipport_check(SYSCTL_HANDLER_ARGS)
145 {
146 	int error;
147 
148 	error = sysctl_handle_int(oidp, oidp->oid_arg1, oidp->oid_arg2, req);
149 	if (!error) {
150 		RANGECHK(ipport_lowfirstauto, 1, IPPORT_RESERVED - 1);
151 		RANGECHK(ipport_lowlastauto, 1, IPPORT_RESERVED - 1);
152 
153 		RANGECHK(ipport_firstauto, IPPORT_RESERVED, USHRT_MAX);
154 		RANGECHK(ipport_lastauto, IPPORT_RESERVED, USHRT_MAX);
155 
156 		RANGECHK(ipport_hifirstauto, IPPORT_RESERVED, USHRT_MAX);
157 		RANGECHK(ipport_hilastauto, IPPORT_RESERVED, USHRT_MAX);
158 	}
159 	return (error);
160 }
161 
162 SYSCTL_NODE(_net_inet_ip, IPPROTO_IP, portrange, CTLFLAG_RW, 0, "IP Ports");
163 
164 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, lowfirst, CTLTYPE_INT|CTLFLAG_RW,
165 	   &ipport_lowfirstauto, 0, &sysctl_net_ipport_check, "I", "");
166 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, lowlast, CTLTYPE_INT|CTLFLAG_RW,
167 	   &ipport_lowlastauto, 0, &sysctl_net_ipport_check, "I", "");
168 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, first, CTLTYPE_INT|CTLFLAG_RW,
169 	   &ipport_firstauto, 0, &sysctl_net_ipport_check, "I", "");
170 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, last, CTLTYPE_INT|CTLFLAG_RW,
171 	   &ipport_lastauto, 0, &sysctl_net_ipport_check, "I", "");
172 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, hifirst, CTLTYPE_INT|CTLFLAG_RW,
173 	   &ipport_hifirstauto, 0, &sysctl_net_ipport_check, "I", "");
174 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, hilast, CTLTYPE_INT|CTLFLAG_RW,
175 	   &ipport_hilastauto, 0, &sysctl_net_ipport_check, "I", "");
176 
177 /*
178  * in_pcb.c: manage the Protocol Control Blocks.
179  *
180  * NOTE: It is assumed that most of these functions will be called from
181  * a critical section.  XXX - There are, unfortunately, a few exceptions
182  * to this rule that should be fixed.
183  *
184  * NOTE: The caller should initialize the cpu field to the cpu running the
185  * protocol stack associated with this inpcbinfo.
186  */
187 
188 void
189 in_pcbinfo_init(struct inpcbinfo *pcbinfo)
190 {
191 	LIST_INIT(&pcbinfo->pcblisthead);
192 	pcbinfo->cpu = -1;
193 }
194 
195 /*
196  * Allocate a PCB and associate it with the socket.
197  */
198 int
199 in_pcballoc(struct socket *so, struct inpcbinfo *pcbinfo)
200 {
201 	struct inpcb *inp;
202 #ifdef IPSEC
203 	int error;
204 #endif
205 
206 	inp = kmalloc(pcbinfo->ipi_size, M_PCB, M_WAITOK|M_ZERO);
207 	inp->inp_gencnt = ++pcbinfo->ipi_gencnt;
208 	inp->inp_pcbinfo = inp->inp_cpcbinfo = pcbinfo;
209 	inp->inp_socket = so;
210 #ifdef IPSEC
211 	error = ipsec_init_policy(so, &inp->inp_sp);
212 	if (error != 0) {
213 		kfree(inp, M_PCB);
214 		return (error);
215 	}
216 #endif
217 #ifdef INET6
218 	if (INP_SOCKAF(so) == AF_INET6 && ip6_v6only)
219 		inp->inp_flags |= IN6P_IPV6_V6ONLY;
220 	if (ip6_auto_flowlabel)
221 		inp->inp_flags |= IN6P_AUTOFLOWLABEL;
222 #endif
223 	so->so_pcb = inp;
224 	LIST_INSERT_HEAD(&pcbinfo->pcblisthead, inp, inp_list);
225 	pcbinfo->ipi_count++;
226 	return (0);
227 }
228 
229 int
230 in_pcbbind(struct inpcb *inp, struct sockaddr *nam, struct thread *td)
231 {
232 	struct socket *so = inp->inp_socket;
233 	struct proc *p = td->td_proc;
234 	unsigned short *lastport;
235 	struct sockaddr_in *sin;
236 	struct sockaddr_in jsin;
237 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
238 	struct ucred *cred = NULL;
239 	u_short lport = 0;
240 	int wild = 0, reuseport = (so->so_options & SO_REUSEPORT);
241 	int error;
242 
243 	KKASSERT(p);
244 
245 	if (TAILQ_EMPTY(&in_ifaddrheads[mycpuid])) /* XXX broken! */
246 		return (EADDRNOTAVAIL);
247 	if (inp->inp_lport != 0 || inp->inp_laddr.s_addr != INADDR_ANY)
248 		return (EINVAL);	/* already bound */
249 	if (!(so->so_options & (SO_REUSEADDR|SO_REUSEPORT)))
250 		wild = 1;    /* neither SO_REUSEADDR nor SO_REUSEPORT is set */
251 	if (p)
252 		cred = p->p_ucred;
253 	if (nam != NULL) {
254 		sin = (struct sockaddr_in *)nam;
255 		if (nam->sa_len != sizeof *sin)
256 			return (EINVAL);
257 #ifdef notdef
258 		/*
259 		 * We should check the family, but old programs
260 		 * incorrectly fail to initialize it.
261 		 */
262 		if (sin->sin_family != AF_INET)
263 			return (EAFNOSUPPORT);
264 #endif
265 		if (!prison_replace_wildcards(td, nam))
266 				return (EINVAL);
267 		lport = sin->sin_port;
268 		if (IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) {
269 			/*
270 			 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
271 			 * allow complete duplication of binding if
272 			 * SO_REUSEPORT is set, or if SO_REUSEADDR is set
273 			 * and a multicast address is bound on both
274 			 * new and duplicated sockets.
275 			 */
276 			if (so->so_options & SO_REUSEADDR)
277 				reuseport = SO_REUSEADDR | SO_REUSEPORT;
278 		} else if (sin->sin_addr.s_addr != INADDR_ANY) {
279 			sin->sin_port = 0;		/* yech... */
280 			bzero(&sin->sin_zero, sizeof sin->sin_zero);
281 			if (ifa_ifwithaddr((struct sockaddr *)sin) == NULL)
282 				return (EADDRNOTAVAIL);
283 		}
284 		if (lport != 0) {
285 			struct inpcb *t;
286 
287 			/* GROSS */
288 			if (ntohs(lport) < IPPORT_RESERVED &&
289 			    cred && priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT, 0))
290 				return (EACCES);
291 			if (so->so_cred->cr_uid != 0 &&
292 			    !IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) {
293 				t = in_pcblookup_local(inp->inp_pcbinfo,
294 				    sin->sin_addr, lport,
295 				    INPLOOKUP_WILDCARD, cred);
296 				if (t &&
297 				    (!in_nullhost(sin->sin_addr) ||
298 				     !in_nullhost(t->inp_laddr) ||
299 				     (t->inp_socket->so_options &
300 					 SO_REUSEPORT) == 0) &&
301 				    (so->so_cred->cr_uid !=
302 				     t->inp_socket->so_cred->cr_uid)) {
303 #ifdef INET6
304 					if (!in_nullhost(sin->sin_addr) ||
305 					    !in_nullhost(t->inp_laddr) ||
306 					    INP_SOCKAF(so) ==
307 					    INP_SOCKAF(t->inp_socket))
308 #endif
309 					return (EADDRINUSE);
310 				}
311 			}
312 			if (cred && !prison_replace_wildcards(td, nam))
313 				return (EADDRNOTAVAIL);
314 			t = in_pcblookup_local(pcbinfo, sin->sin_addr, lport,
315 					       wild, cred);
316 			if (t && !(reuseport & t->inp_socket->so_options)) {
317 #ifdef INET6
318 				if (!in_nullhost(sin->sin_addr) ||
319 				    !in_nullhost(t->inp_laddr) ||
320 				    INP_SOCKAF(so) == INP_SOCKAF(t->inp_socket))
321 #endif
322 				return (EADDRINUSE);
323 			}
324 		}
325 		inp->inp_laddr = sin->sin_addr;
326 	}
327 	if (lport == 0) {
328 		ushort first, last;
329 		int count;
330 
331 		jsin.sin_family = AF_INET;
332 		jsin.sin_addr.s_addr = inp->inp_laddr.s_addr;
333 		if (!prison_replace_wildcards(td, (struct sockaddr *)&jsin)) {
334 			inp->inp_laddr.s_addr = INADDR_ANY;
335 			return (EINVAL);
336 		}
337 		inp->inp_laddr.s_addr = jsin.sin_addr.s_addr;
338 
339 		inp->inp_flags |= INP_ANONPORT;
340 
341 		if (inp->inp_flags & INP_HIGHPORT) {
342 			first = ipport_hifirstauto;	/* sysctl */
343 			last  = ipport_hilastauto;
344 			lastport = &pcbinfo->lasthi;
345 		} else if (inp->inp_flags & INP_LOWPORT) {
346 			if (cred &&
347 			    (error = priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT, 0))) {
348 				inp->inp_laddr.s_addr = INADDR_ANY;
349 				return (error);
350 			}
351 			first = ipport_lowfirstauto;	/* 1023 */
352 			last  = ipport_lowlastauto;	/* 600 */
353 			lastport = &pcbinfo->lastlow;
354 		} else {
355 			first = ipport_firstauto;	/* sysctl */
356 			last  = ipport_lastauto;
357 			lastport = &pcbinfo->lastport;
358 		}
359 		/*
360 		 * Simple check to ensure all ports are not used up causing
361 		 * a deadlock here.
362 		 *
363 		 * We split the two cases (up and down) so that the direction
364 		 * is not being tested on each round of the loop.
365 		 */
366 		if (first > last) {
367 			/*
368 			 * counting down
369 			 */
370 			count = first - last;
371 
372 			do {
373 				if (count-- < 0) {	/* completely used? */
374 					inp->inp_laddr.s_addr = INADDR_ANY;
375 					return (EADDRNOTAVAIL);
376 				}
377 				--*lastport;
378 				if (*lastport > first || *lastport < last)
379 					*lastport = first;
380 				lport = htons(*lastport);
381 			} while (in_pcblookup_local(pcbinfo, inp->inp_laddr,
382 				 lport, wild, cred));
383 		} else {
384 			/*
385 			 * counting up
386 			 */
387 			count = last - first;
388 
389 			do {
390 				if (count-- < 0) {	/* completely used? */
391 					inp->inp_laddr.s_addr = INADDR_ANY;
392 					return (EADDRNOTAVAIL);
393 				}
394 				++*lastport;
395 				if (*lastport < first || *lastport > last)
396 					*lastport = first;
397 				lport = htons(*lastport);
398 			} while (in_pcblookup_local(pcbinfo, inp->inp_laddr,
399 				 lport, wild, cred));
400 		}
401 	}
402 	inp->inp_lport = lport;
403 
404 	jsin.sin_family = AF_INET;
405 	jsin.sin_addr.s_addr = inp->inp_laddr.s_addr;
406 	if (!prison_replace_wildcards(td, (struct sockaddr*)&jsin)) {
407 		inp->inp_laddr.s_addr = INADDR_ANY;
408 		inp->inp_lport = 0;
409 		return (EINVAL);
410 	}
411 	inp->inp_laddr.s_addr = jsin.sin_addr.s_addr;
412 
413 	if (in_pcbinsporthash(inp) != 0) {
414 		inp->inp_laddr.s_addr = INADDR_ANY;
415 		inp->inp_lport = 0;
416 		return (EAGAIN);
417 	}
418 	return (0);
419 }
420 
421 /*
422  *   Transform old in_pcbconnect() into an inner subroutine for new
423  *   in_pcbconnect(): Do some validity-checking on the remote
424  *   address (in mbuf 'nam') and then determine local host address
425  *   (i.e., which interface) to use to access that remote host.
426  *
427  *   This preserves definition of in_pcbconnect(), while supporting a
428  *   slightly different version for T/TCP.  (This is more than
429  *   a bit of a kludge, but cleaning up the internal interfaces would
430  *   have forced minor changes in every protocol).
431  */
432 int
433 in_pcbladdr(struct inpcb *inp, struct sockaddr *nam,
434 	struct sockaddr_in **plocal_sin, struct thread *td)
435 {
436 	struct in_ifaddr *ia;
437 	struct ucred *cred = NULL;
438 	struct sockaddr_in *sin = (struct sockaddr_in *)nam;
439 	struct sockaddr *jsin;
440 	int jailed = 0, alloc_route = 0;
441 
442 	if (nam->sa_len != sizeof *sin)
443 		return (EINVAL);
444 	if (sin->sin_family != AF_INET)
445 		return (EAFNOSUPPORT);
446 	if (sin->sin_port == 0)
447 		return (EADDRNOTAVAIL);
448 	if (td && td->td_proc && td->td_proc->p_ucred)
449 		cred = td->td_proc->p_ucred;
450 	if (cred && cred->cr_prison)
451 		jailed = 1;
452 	if (!TAILQ_EMPTY(&in_ifaddrheads[mycpuid])) {
453 		ia = TAILQ_FIRST(&in_ifaddrheads[mycpuid])->ia;
454 		/*
455 		 * If the destination address is INADDR_ANY,
456 		 * use the primary local address.
457 		 * If the supplied address is INADDR_BROADCAST,
458 		 * and the primary interface supports broadcast,
459 		 * choose the broadcast address for that interface.
460 		 */
461 		if (sin->sin_addr.s_addr == INADDR_ANY)
462 			sin->sin_addr = IA_SIN(ia)->sin_addr;
463 		else if (sin->sin_addr.s_addr == (u_long)INADDR_BROADCAST &&
464 		    (ia->ia_ifp->if_flags & IFF_BROADCAST))
465 			sin->sin_addr = satosin(&ia->ia_broadaddr)->sin_addr;
466 	}
467 	if (inp->inp_laddr.s_addr == INADDR_ANY) {
468 		struct route *ro;
469 
470 		ia = NULL;
471 		/*
472 		 * If route is known or can be allocated now,
473 		 * our src addr is taken from the i/f, else punt.
474 		 * Note that we should check the address family of the cached
475 		 * destination, in case of sharing the cache with IPv6.
476 		 */
477 		ro = &inp->inp_route;
478 		if (ro->ro_rt &&
479 		    (!(ro->ro_rt->rt_flags & RTF_UP) ||
480 		     ro->ro_dst.sa_family != AF_INET ||
481 		     satosin(&ro->ro_dst)->sin_addr.s_addr !=
482 				      sin->sin_addr.s_addr ||
483 		     inp->inp_socket->so_options & SO_DONTROUTE)) {
484 			RTFREE(ro->ro_rt);
485 			ro->ro_rt = NULL;
486 		}
487 		if (!(inp->inp_socket->so_options & SO_DONTROUTE) && /*XXX*/
488 		    (ro->ro_rt == NULL ||
489 		    ro->ro_rt->rt_ifp == NULL)) {
490 			/* No route yet, so try to acquire one */
491 			bzero(&ro->ro_dst, sizeof(struct sockaddr_in));
492 			ro->ro_dst.sa_family = AF_INET;
493 			ro->ro_dst.sa_len = sizeof(struct sockaddr_in);
494 			((struct sockaddr_in *) &ro->ro_dst)->sin_addr =
495 				sin->sin_addr;
496 			rtalloc(ro);
497 			alloc_route = 1;
498 		}
499 		/*
500 		 * If we found a route, use the address
501 		 * corresponding to the outgoing interface
502 		 * unless it is the loopback (in case a route
503 		 * to our address on another net goes to loopback).
504 		 */
505 		if (ro->ro_rt && !(ro->ro_rt->rt_ifp->if_flags & IFF_LOOPBACK)) {
506 			if (jailed) {
507 				if (jailed_ip(cred->cr_prison,
508 				    ro->ro_rt->rt_ifa->ifa_addr)) {
509 					ia = ifatoia(ro->ro_rt->rt_ifa);
510 				}
511 			} else {
512 				ia = ifatoia(ro->ro_rt->rt_ifa);
513 			}
514 		}
515 		if (ia == NULL) {
516 			u_short fport = sin->sin_port;
517 
518 			sin->sin_port = 0;
519 			ia = ifatoia(ifa_ifwithdstaddr(sintosa(sin)));
520 			if (ia && jailed && !jailed_ip(cred->cr_prison,
521 			    sintosa(&ia->ia_addr)))
522 				ia = NULL;
523 			if (ia == NULL)
524 				ia = ifatoia(ifa_ifwithnet(sintosa(sin)));
525 			if (ia && jailed && !jailed_ip(cred->cr_prison,
526 			    sintosa(&ia->ia_addr)))
527 				ia = NULL;
528 			sin->sin_port = fport;
529 			if (ia == NULL &&
530 			    !TAILQ_EMPTY(&in_ifaddrheads[mycpuid]))
531 				ia = TAILQ_FIRST(&in_ifaddrheads[mycpuid])->ia;
532 			if (ia && jailed && !jailed_ip(cred->cr_prison,
533 			    sintosa(&ia->ia_addr)))
534 				ia = NULL;
535 
536 			if (!jailed && ia == NULL)
537 				goto fail;
538 		}
539 		/*
540 		 * If the destination address is multicast and an outgoing
541 		 * interface has been set as a multicast option, use the
542 		 * address of that interface as our source address.
543 		 */
544 		if (!jailed && IN_MULTICAST(ntohl(sin->sin_addr.s_addr)) &&
545 		    inp->inp_moptions != NULL) {
546 			struct ip_moptions *imo;
547 			struct ifnet *ifp;
548 
549 			imo = inp->inp_moptions;
550 			if (imo->imo_multicast_ifp != NULL) {
551 				struct in_ifaddr_container *iac;
552 
553 				ifp = imo->imo_multicast_ifp;
554 				ia = NULL;
555 				TAILQ_FOREACH(iac,
556 				&in_ifaddrheads[mycpuid], ia_link) {
557 					if (iac->ia->ia_ifp == ifp) {
558 						ia = iac->ia;
559 						break;
560 					}
561 				}
562 				if (ia == NULL)
563 					goto fail;
564 			}
565 		}
566 		/*
567 		 * Don't do pcblookup call here; return interface in plocal_sin
568 		 * and exit to caller, that will do the lookup.
569 		 */
570 		if (ia == NULL && jailed) {
571 			if ((jsin = prison_get_nonlocal(cred->cr_prison, AF_INET, NULL)) != NULL ||
572 			    (jsin = prison_get_local(cred->cr_prison, AF_INET, NULL)) != NULL) {
573 				*plocal_sin = satosin(jsin);
574 			} else {
575 				/* IPv6 only Jail */
576 				goto fail;
577 			}
578 		} else {
579 			*plocal_sin = &ia->ia_addr;
580 		}
581 	}
582 	return (0);
583 fail:
584 	if (alloc_route) {
585 		struct route *ro = &inp->inp_route;
586 
587 		if (ro->ro_rt != NULL)
588 			RTFREE(ro->ro_rt);
589 		bzero(ro, sizeof(*ro));
590 	}
591 	return (EADDRNOTAVAIL);
592 }
593 
594 /*
595  * Outer subroutine:
596  * Connect from a socket to a specified address.
597  * Both address and port must be specified in argument sin.
598  * If don't have a local address for this socket yet,
599  * then pick one.
600  */
601 int
602 in_pcbconnect(struct inpcb *inp, struct sockaddr *nam, struct thread *td)
603 {
604 	struct sockaddr_in *if_sin;
605 	struct sockaddr_in *sin = (struct sockaddr_in *)nam;
606 	int error;
607 
608 	/* Call inner routine to assign local interface address. */
609 	if ((error = in_pcbladdr(inp, nam, &if_sin, td)) != 0)
610 		return (error);
611 
612 	if (in_pcblookup_hash(inp->inp_cpcbinfo, sin->sin_addr, sin->sin_port,
613 	    inp->inp_laddr.s_addr ? inp->inp_laddr : if_sin->sin_addr,
614 	    inp->inp_lport, FALSE, NULL) != NULL) {
615 		return (EADDRINUSE);
616 	}
617 	if (inp->inp_laddr.s_addr == INADDR_ANY) {
618 		if (inp->inp_lport == 0) {
619 			error = in_pcbbind(inp, NULL, td);
620 			if (error)
621 				return (error);
622 		}
623 		inp->inp_laddr = if_sin->sin_addr;
624 	}
625 	inp->inp_faddr = sin->sin_addr;
626 	inp->inp_fport = sin->sin_port;
627 	in_pcbinsconnhash(inp);
628 	return (0);
629 }
630 
631 void
632 in_pcbdisconnect(struct inpcb *inp)
633 {
634 
635 	inp->inp_faddr.s_addr = INADDR_ANY;
636 	inp->inp_fport = 0;
637 	in_pcbremconnhash(inp);
638 	if (inp->inp_socket->so_state & SS_NOFDREF)
639 		in_pcbdetach(inp);
640 }
641 
642 void
643 in_pcbdetach(struct inpcb *inp)
644 {
645 	struct socket *so = inp->inp_socket;
646 	struct inpcbinfo *ipi = inp->inp_pcbinfo;
647 
648 #ifdef IPSEC
649 	ipsec4_delete_pcbpolicy(inp);
650 #endif /*IPSEC*/
651 	inp->inp_gencnt = ++ipi->ipi_gencnt;
652 	in_pcbremlists(inp);
653 	so->so_pcb = NULL;
654 	sofree(so);
655 	if (inp->inp_options)
656 		m_free(inp->inp_options);
657 	if (inp->inp_route.ro_rt)
658 		rtfree(inp->inp_route.ro_rt);
659 	ip_freemoptions(inp->inp_moptions);
660 	inp->inp_vflag = 0;
661 	kfree(inp, M_PCB);
662 }
663 
664 /*
665  * The calling convention of in_setsockaddr() and in_setpeeraddr() was
666  * modified to match the pru_sockaddr() and pru_peeraddr() entry points
667  * in struct pr_usrreqs, so that protocols can just reference then directly
668  * without the need for a wrapper function.  The socket must have a valid
669  * (i.e., non-nil) PCB, but it should be impossible to get an invalid one
670  * except through a kernel programming error, so it is acceptable to panic
671  * (or in this case trap) if the PCB is invalid.  (Actually, we don't trap
672  * because there actually /is/ a programming error somewhere... XXX)
673  */
674 int
675 in_setsockaddr(struct socket *so, struct sockaddr **nam)
676 {
677 	struct inpcb *inp;
678 	struct sockaddr_in *sin;
679 
680 	/*
681 	 * Do the malloc first in case it blocks.
682 	 */
683 	MALLOC(sin, struct sockaddr_in *, sizeof *sin, M_SONAME,
684 		M_WAITOK | M_ZERO);
685 	sin->sin_family = AF_INET;
686 	sin->sin_len = sizeof *sin;
687 
688 	crit_enter();
689 	inp = so->so_pcb;
690 	if (!inp) {
691 		crit_exit();
692 		kfree(sin, M_SONAME);
693 		return (ECONNRESET);
694 	}
695 	sin->sin_port = inp->inp_lport;
696 	sin->sin_addr = inp->inp_laddr;
697 	crit_exit();
698 
699 	*nam = (struct sockaddr *)sin;
700 	return (0);
701 }
702 
703 int
704 in_setpeeraddr(struct socket *so, struct sockaddr **nam)
705 {
706 	struct inpcb *inp;
707 	struct sockaddr_in *sin;
708 
709 	/*
710 	 * Do the malloc first in case it blocks.
711 	 */
712 	MALLOC(sin, struct sockaddr_in *, sizeof *sin, M_SONAME,
713 		M_WAITOK | M_ZERO);
714 	sin->sin_family = AF_INET;
715 	sin->sin_len = sizeof *sin;
716 
717 	crit_enter();
718 	inp = so->so_pcb;
719 	if (!inp) {
720 		crit_exit();
721 		kfree(sin, M_SONAME);
722 		return (ECONNRESET);
723 	}
724 	sin->sin_port = inp->inp_fport;
725 	sin->sin_addr = inp->inp_faddr;
726 	crit_exit();
727 
728 	*nam = (struct sockaddr *)sin;
729 	return (0);
730 }
731 
732 void
733 in_pcbnotifyall(struct inpcbhead *head, struct in_addr faddr, int err,
734 		void (*notify)(struct inpcb *, int))
735 {
736 	struct inpcb *inp, *ninp;
737 
738 	/*
739 	 * note: if INP_PLACEMARKER is set we must ignore the rest of
740 	 * the structure and skip it.
741 	 */
742 	crit_enter();
743 	LIST_FOREACH_MUTABLE(inp, head, inp_list, ninp) {
744 		if (inp->inp_flags & INP_PLACEMARKER)
745 			continue;
746 #ifdef INET6
747 		if (!(inp->inp_vflag & INP_IPV4))
748 			continue;
749 #endif
750 		if (inp->inp_faddr.s_addr != faddr.s_addr ||
751 		    inp->inp_socket == NULL)
752 			continue;
753 		(*notify)(inp, err);		/* can remove inp from list! */
754 	}
755 	crit_exit();
756 }
757 
758 void
759 in_pcbpurgeif0(struct inpcb *head, struct ifnet *ifp)
760 {
761 	struct inpcb *inp;
762 	struct ip_moptions *imo;
763 	int i, gap;
764 
765 	for (inp = head; inp != NULL; inp = LIST_NEXT(inp, inp_list)) {
766 		if (inp->inp_flags & INP_PLACEMARKER)
767 			continue;
768 		imo = inp->inp_moptions;
769 		if ((inp->inp_vflag & INP_IPV4) && imo != NULL) {
770 			/*
771 			 * Unselect the outgoing interface if it is being
772 			 * detached.
773 			 */
774 			if (imo->imo_multicast_ifp == ifp)
775 				imo->imo_multicast_ifp = NULL;
776 
777 			/*
778 			 * Drop multicast group membership if we joined
779 			 * through the interface being detached.
780 			 */
781 			for (i = 0, gap = 0; i < imo->imo_num_memberships;
782 			    i++) {
783 				if (imo->imo_membership[i]->inm_ifp == ifp) {
784 					in_delmulti(imo->imo_membership[i]);
785 					gap++;
786 				} else if (gap != 0)
787 					imo->imo_membership[i - gap] =
788 					    imo->imo_membership[i];
789 			}
790 			imo->imo_num_memberships -= gap;
791 		}
792 	}
793 }
794 
795 /*
796  * Check for alternatives when higher level complains
797  * about service problems.  For now, invalidate cached
798  * routing information.  If the route was created dynamically
799  * (by a redirect), time to try a default gateway again.
800  */
801 void
802 in_losing(struct inpcb *inp)
803 {
804 	struct rtentry *rt;
805 	struct rt_addrinfo rtinfo;
806 
807 	if ((rt = inp->inp_route.ro_rt)) {
808 		bzero(&rtinfo, sizeof(struct rt_addrinfo));
809 		rtinfo.rti_info[RTAX_DST] = rt_key(rt);
810 		rtinfo.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
811 		rtinfo.rti_info[RTAX_NETMASK] = rt_mask(rt);
812 		rtinfo.rti_flags = rt->rt_flags;
813 		rt_missmsg(RTM_LOSING, &rtinfo, rt->rt_flags, 0);
814 		if (rt->rt_flags & RTF_DYNAMIC)
815 			rtrequest1_global(RTM_DELETE, &rtinfo, NULL, NULL);
816 		inp->inp_route.ro_rt = NULL;
817 		rtfree(rt);
818 		/*
819 		 * A new route can be allocated
820 		 * the next time output is attempted.
821 		 */
822 	}
823 }
824 
825 /*
826  * After a routing change, flush old routing
827  * and allocate a (hopefully) better one.
828  */
829 void
830 in_rtchange(struct inpcb *inp, int err)
831 {
832 	if (inp->inp_route.ro_rt) {
833 		rtfree(inp->inp_route.ro_rt);
834 		inp->inp_route.ro_rt = NULL;
835 		/*
836 		 * A new route can be allocated the next time
837 		 * output is attempted.
838 		 */
839 	}
840 }
841 
842 /*
843  * Lookup a PCB based on the local address and port.
844  */
845 struct inpcb *
846 in_pcblookup_local(struct inpcbinfo *pcbinfo, struct in_addr laddr,
847 		   u_int lport_arg, int wild_okay, struct ucred *cred)
848 {
849 	struct inpcb *inp;
850 	int matchwild = 3, wildcard;
851 	u_short lport = lport_arg;
852 
853 	struct inpcbporthead *porthash;
854 	struct inpcbport *phd;
855 	struct inpcb *match = NULL;
856 
857 	/*
858 	 * Best fit PCB lookup.
859 	 *
860 	 * First see if this local port is in use by looking on the
861 	 * port hash list.
862 	 */
863 	porthash = &pcbinfo->porthashbase[INP_PCBPORTHASH(lport,
864 	    pcbinfo->porthashmask)];
865 	LIST_FOREACH(phd, porthash, phd_hash) {
866 		if (phd->phd_port == lport)
867 			break;
868 	}
869 	if (phd != NULL) {
870 		/*
871 		 * Port is in use by one or more PCBs. Look for best
872 		 * fit.
873 		 */
874 		LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) {
875 			wildcard = 0;
876 #ifdef INET6
877 			if ((inp->inp_vflag & INP_IPV4) == 0)
878 				continue;
879 #endif
880 			if (inp->inp_faddr.s_addr != INADDR_ANY)
881 				wildcard++;
882 			if (inp->inp_laddr.s_addr != INADDR_ANY) {
883 				if (laddr.s_addr == INADDR_ANY)
884 					wildcard++;
885 				else if (inp->inp_laddr.s_addr != laddr.s_addr)
886 					continue;
887 			} else {
888 				if (laddr.s_addr != INADDR_ANY)
889 					wildcard++;
890 			}
891 			if (wildcard && !wild_okay)
892 				continue;
893 			if (wildcard < matchwild &&
894 			    (cred == NULL ||
895 			     cred->cr_prison ==
896 					inp->inp_socket->so_cred->cr_prison)) {
897 				match = inp;
898 				matchwild = wildcard;
899 				if (matchwild == 0) {
900 					break;
901 				}
902 			}
903 		}
904 	}
905 	return (match);
906 }
907 
908 /*
909  * Lookup PCB in hash list.
910  */
911 struct inpcb *
912 in_pcblookup_hash(struct inpcbinfo *pcbinfo, struct in_addr faddr, u_int fport_arg,
913 		  struct in_addr laddr, u_int lport_arg, boolean_t wildcard,
914 		  struct ifnet *ifp)
915 {
916 	struct inpcbhead *head;
917 	struct inpcb *inp, *jinp=NULL;
918 	u_short fport = fport_arg, lport = lport_arg;
919 
920 	/*
921 	 * First look for an exact match.
922 	 */
923 	head = &pcbinfo->hashbase[INP_PCBCONNHASH(faddr.s_addr, fport,
924 	    laddr.s_addr, lport, pcbinfo->hashmask)];
925 	LIST_FOREACH(inp, head, inp_hash) {
926 #ifdef INET6
927 		if (!(inp->inp_vflag & INP_IPV4))
928 			continue;
929 #endif
930 		if (in_hosteq(inp->inp_faddr, faddr) &&
931 		    in_hosteq(inp->inp_laddr, laddr) &&
932 		    inp->inp_fport == fport && inp->inp_lport == lport) {
933 			/* found */
934 			if (inp->inp_socket == NULL ||
935 			    inp->inp_socket->so_cred->cr_prison == NULL) {
936 				return (inp);
937 			} else {
938 				if  (jinp == NULL)
939 					jinp = inp;
940 			}
941 		}
942 	}
943 	if (jinp != NULL)
944 		return (jinp);
945 	if (wildcard) {
946 		struct inpcb *local_wild = NULL;
947 		struct inpcb *jinp_wild = NULL;
948 #ifdef INET6
949 		struct inpcb *local_wild_mapped = NULL;
950 #endif
951 		struct inpcontainer *ic;
952 		struct inpcontainerhead *chead;
953 		struct sockaddr_in jsin;
954 		struct ucred *cred;
955 
956 		/*
957 		 * Order of socket selection:
958 		 * 1. non-jailed, non-wild.
959 		 * 2. non-jailed, wild.
960 		 * 3. jailed, non-wild.
961 		 * 4. jailed, wild.
962 		 */
963 		jsin.sin_family = AF_INET;
964 		chead = &pcbinfo->wildcardhashbase[
965 		    INP_PCBWILDCARDHASH(lport, pcbinfo->wildcardhashmask)];
966 		LIST_FOREACH(ic, chead, ic_list) {
967 			inp = ic->ic_inp;
968 			jsin.sin_addr.s_addr = laddr.s_addr;
969 #ifdef INET6
970 			if (!(inp->inp_vflag & INP_IPV4))
971 				continue;
972 #endif
973 			if (inp->inp_socket != NULL)
974 				cred = inp->inp_socket->so_cred;
975 			else
976 				cred = NULL;
977 			if (cred != NULL && jailed(cred)) {
978 				if (jinp != NULL)
979 					continue;
980 				else
981 					if (!jailed_ip(cred->cr_prison,
982 					    (struct sockaddr *)&jsin))
983 						continue;
984 			}
985 			if (inp->inp_lport == lport) {
986 				if (ifp && ifp->if_type == IFT_FAITH &&
987 				    !(inp->inp_flags & INP_FAITH))
988 					continue;
989 				if (inp->inp_laddr.s_addr == laddr.s_addr) {
990 					if (cred != NULL && jailed(cred))
991 						jinp = inp;
992 					else
993 						return (inp);
994 				}
995 				if (inp->inp_laddr.s_addr == INADDR_ANY) {
996 #ifdef INET6
997 					if (INP_CHECK_SOCKAF(inp->inp_socket,
998 							     AF_INET6))
999 						local_wild_mapped = inp;
1000 					else
1001 #endif
1002 						if (cred != NULL &&
1003 						    jailed(cred))
1004 							jinp_wild = inp;
1005 						else
1006 							local_wild = inp;
1007 				}
1008 			}
1009 		}
1010 		if (local_wild != NULL)
1011 			return (local_wild);
1012 #ifdef INET6
1013 		if (local_wild_mapped != NULL)
1014 			return (local_wild_mapped);
1015 #endif
1016 		if (jinp != NULL)
1017 			return (jinp);
1018 		return (jinp_wild);
1019 	}
1020 
1021 	/*
1022 	 * Not found.
1023 	 */
1024 	return (NULL);
1025 }
1026 
1027 /*
1028  * Insert PCB into connection hash table.
1029  */
1030 void
1031 in_pcbinsconnhash(struct inpcb *inp)
1032 {
1033 	struct inpcbinfo *pcbinfo = inp->inp_cpcbinfo;
1034 	struct inpcbhead *bucket;
1035 	u_int32_t hashkey_faddr, hashkey_laddr;
1036 
1037 #ifdef INET6
1038 	if (inp->inp_vflag & INP_IPV6) {
1039 		hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX JH */;
1040 		hashkey_laddr = inp->in6p_laddr.s6_addr32[3] /* XXX JH */;
1041 	} else {
1042 #endif
1043 		hashkey_faddr = inp->inp_faddr.s_addr;
1044 		hashkey_laddr = inp->inp_laddr.s_addr;
1045 #ifdef INET6
1046 	}
1047 #endif
1048 
1049 	KASSERT(!(inp->inp_flags & INP_CONNECTED), ("already on hash list"));
1050 	inp->inp_flags |= INP_CONNECTED;
1051 
1052 	/*
1053 	 * Insert into the connection hash table.
1054 	 */
1055 	bucket = &pcbinfo->hashbase[INP_PCBCONNHASH(hashkey_faddr,
1056 	    inp->inp_fport, hashkey_laddr, inp->inp_lport, pcbinfo->hashmask)];
1057 	LIST_INSERT_HEAD(bucket, inp, inp_hash);
1058 }
1059 
1060 /*
1061  * Remove PCB from connection hash table.
1062  */
1063 void
1064 in_pcbremconnhash(struct inpcb *inp)
1065 {
1066 	KASSERT(inp->inp_flags & INP_CONNECTED, ("inp not connected"));
1067 	LIST_REMOVE(inp, inp_hash);
1068 	inp->inp_flags &= ~INP_CONNECTED;
1069 }
1070 
1071 /*
1072  * Insert PCB into port hash table.
1073  */
1074 int
1075 in_pcbinsporthash(struct inpcb *inp)
1076 {
1077 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1078 	struct inpcbporthead *pcbporthash;
1079 	struct inpcbport *phd;
1080 
1081 	/*
1082 	 * Insert into the port hash table.
1083 	 */
1084 	pcbporthash = &pcbinfo->porthashbase[
1085 	    INP_PCBPORTHASH(inp->inp_lport, pcbinfo->porthashmask)];
1086 
1087 	/* Go through port list and look for a head for this lport. */
1088 	LIST_FOREACH(phd, pcbporthash, phd_hash)
1089 		if (phd->phd_port == inp->inp_lport)
1090 			break;
1091 
1092 	/* If none exists, malloc one and tack it on. */
1093 	if (phd == NULL) {
1094 		MALLOC(phd, struct inpcbport *, sizeof(struct inpcbport),
1095 		    M_PCB, M_INTWAIT | M_NULLOK);
1096 		if (phd == NULL)
1097 			return (ENOBUFS); /* XXX */
1098 		phd->phd_port = inp->inp_lport;
1099 		LIST_INIT(&phd->phd_pcblist);
1100 		LIST_INSERT_HEAD(pcbporthash, phd, phd_hash);
1101 	}
1102 
1103 	inp->inp_phd = phd;
1104 	LIST_INSERT_HEAD(&phd->phd_pcblist, inp, inp_portlist);
1105 
1106 	return (0);
1107 }
1108 
1109 void
1110 in_pcbinswildcardhash_oncpu(struct inpcb *inp, struct inpcbinfo *pcbinfo)
1111 {
1112 	struct inpcontainer *ic;
1113 	struct inpcontainerhead *bucket;
1114 
1115 	bucket = &pcbinfo->wildcardhashbase[
1116 	    INP_PCBWILDCARDHASH(inp->inp_lport, pcbinfo->wildcardhashmask)];
1117 
1118 	ic = kmalloc(sizeof(struct inpcontainer), M_TEMP, M_INTWAIT);
1119 	ic->ic_inp = inp;
1120 	LIST_INSERT_HEAD(bucket, ic, ic_list);
1121 }
1122 
1123 /*
1124  * Insert PCB into wildcard hash table.
1125  */
1126 void
1127 in_pcbinswildcardhash(struct inpcb *inp)
1128 {
1129 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1130 
1131 	KKASSERT(pcbinfo != NULL);
1132 
1133 	in_pcbinswildcardhash_oncpu(inp, pcbinfo);
1134 	inp->inp_flags |= INP_WILDCARD;
1135 }
1136 
1137 void
1138 in_pcbremwildcardhash_oncpu(struct inpcb *inp, struct inpcbinfo *pcbinfo)
1139 {
1140 	struct inpcontainer *ic;
1141 	struct inpcontainerhead *head;
1142 
1143 	/* find bucket */
1144 	head = &pcbinfo->wildcardhashbase[
1145 	    INP_PCBWILDCARDHASH(inp->inp_lport, pcbinfo->wildcardhashmask)];
1146 
1147 	LIST_FOREACH(ic, head, ic_list) {
1148 		if (ic->ic_inp == inp)
1149 			goto found;
1150 	}
1151 	return;			/* not found! */
1152 
1153 found:
1154 	LIST_REMOVE(ic, ic_list);	/* remove container from bucket chain */
1155 	kfree(ic, M_TEMP);		/* deallocate container */
1156 }
1157 
1158 /*
1159  * Remove PCB from wildcard hash table.
1160  */
1161 void
1162 in_pcbremwildcardhash(struct inpcb *inp)
1163 {
1164 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1165 
1166 	KASSERT(inp->inp_flags & INP_WILDCARD, ("inp not wildcard"));
1167 	in_pcbremwildcardhash_oncpu(inp, pcbinfo);
1168 	inp->inp_flags &= ~INP_WILDCARD;
1169 }
1170 
1171 /*
1172  * Remove PCB from various lists.
1173  */
1174 void
1175 in_pcbremlists(struct inpcb *inp)
1176 {
1177 	if (inp->inp_lport) {
1178 		struct inpcbport *phd = inp->inp_phd;
1179 
1180 		LIST_REMOVE(inp, inp_portlist);
1181 		if (LIST_FIRST(&phd->phd_pcblist) == NULL) {
1182 			LIST_REMOVE(phd, phd_hash);
1183 			kfree(phd, M_PCB);
1184 		}
1185 	}
1186 	if (inp->inp_flags & INP_WILDCARD) {
1187 		in_pcbremwildcardhash(inp);
1188 	} else if (inp->inp_flags & INP_CONNECTED) {
1189 		in_pcbremconnhash(inp);
1190 	}
1191 	LIST_REMOVE(inp, inp_list);
1192 	inp->inp_pcbinfo->ipi_count--;
1193 }
1194 
1195 int
1196 prison_xinpcb(struct thread *td, struct inpcb *inp)
1197 {
1198 	struct ucred *cr;
1199 
1200 	if (td->td_proc == NULL)
1201 		return (0);
1202 	cr = td->td_proc->p_ucred;
1203 	if (cr->cr_prison == NULL)
1204 		return (0);
1205 	if (inp->inp_socket && inp->inp_socket->so_cred &&
1206 	    inp->inp_socket->so_cred->cr_prison &&
1207 	    cr->cr_prison == inp->inp_socket->so_cred->cr_prison)
1208 		return (0);
1209 	return (1);
1210 }
1211 
1212 int
1213 in_pcblist_global(SYSCTL_HANDLER_ARGS)
1214 {
1215 	struct inpcbinfo *pcbinfo = arg1;
1216 	struct inpcb *inp, *marker;
1217 	struct xinpcb xi;
1218 	int error, i, n;
1219 	inp_gen_t gencnt;
1220 
1221 	/*
1222 	 * The process of preparing the TCB list is too time-consuming and
1223 	 * resource-intensive to repeat twice on every request.
1224 	 */
1225 	if (req->oldptr == NULL) {
1226 		n = pcbinfo->ipi_count;
1227 		req->oldidx = (n + n/8 + 10) * sizeof(struct xinpcb);
1228 		return 0;
1229 	}
1230 
1231 	if (req->newptr != NULL)
1232 		return EPERM;
1233 
1234 	/*
1235 	 * OK, now we're committed to doing something.  Re-fetch ipi_count
1236 	 * after obtaining the generation count.
1237 	 */
1238 	gencnt = pcbinfo->ipi_gencnt;
1239 	n = pcbinfo->ipi_count;
1240 
1241 	marker = kmalloc(sizeof(struct inpcb), M_TEMP, M_WAITOK|M_ZERO);
1242 	marker->inp_flags |= INP_PLACEMARKER;
1243 	LIST_INSERT_HEAD(&pcbinfo->pcblisthead, marker, inp_list);
1244 
1245 	i = 0;
1246 	error = 0;
1247 
1248 	while ((inp = LIST_NEXT(marker, inp_list)) != NULL && i < n) {
1249 		LIST_REMOVE(marker, inp_list);
1250 		LIST_INSERT_AFTER(inp, marker, inp_list);
1251 
1252 		if (inp->inp_flags & INP_PLACEMARKER)
1253 			continue;
1254 		if (inp->inp_gencnt > gencnt)
1255 			continue;
1256 		if (prison_xinpcb(req->td, inp))
1257 			continue;
1258 		bzero(&xi, sizeof xi);
1259 		xi.xi_len = sizeof xi;
1260 		bcopy(inp, &xi.xi_inp, sizeof *inp);
1261 		if (inp->inp_socket)
1262 			sotoxsocket(inp->inp_socket, &xi.xi_socket);
1263 		if ((error = SYSCTL_OUT(req, &xi, sizeof xi)) != 0)
1264 			break;
1265 		++i;
1266 	}
1267 	LIST_REMOVE(marker, inp_list);
1268 	if (error == 0 && i < n) {
1269 		bzero(&xi, sizeof xi);
1270 		xi.xi_len = sizeof xi;
1271 		while (i < n) {
1272 			error = SYSCTL_OUT(req, &xi, sizeof xi);
1273 			++i;
1274 		}
1275 	}
1276 	kfree(marker, M_TEMP);
1277 	return(error);
1278 }
1279