xref: /dflybsd-src/sys/netinet/in_pcb.c (revision a9656fbcd49c376aba5e04370d8b0f1fa96e063c)
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 #define RANGECHK(var, min, max) \
135 	if ((var) < (min)) { (var) = (min); } \
136 	else if ((var) > (max)) { (var) = (max); }
137 
138 static int
139 sysctl_net_ipport_check(SYSCTL_HANDLER_ARGS)
140 {
141 	int error;
142 
143 	error = sysctl_handle_int(oidp, oidp->oid_arg1, oidp->oid_arg2, req);
144 	if (!error) {
145 		RANGECHK(ipport_lowfirstauto, 1, IPPORT_RESERVED - 1);
146 		RANGECHK(ipport_lowlastauto, 1, IPPORT_RESERVED - 1);
147 
148 		RANGECHK(ipport_firstauto, IPPORT_RESERVED, USHRT_MAX);
149 		RANGECHK(ipport_lastauto, IPPORT_RESERVED, USHRT_MAX);
150 
151 		RANGECHK(ipport_hifirstauto, IPPORT_RESERVED, USHRT_MAX);
152 		RANGECHK(ipport_hilastauto, IPPORT_RESERVED, USHRT_MAX);
153 	}
154 	return (error);
155 }
156 
157 #undef RANGECHK
158 
159 SYSCTL_NODE(_net_inet_ip, IPPROTO_IP, portrange, CTLFLAG_RW, 0, "IP Ports");
160 
161 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, lowfirst, CTLTYPE_INT|CTLFLAG_RW,
162 	   &ipport_lowfirstauto, 0, &sysctl_net_ipport_check, "I", "");
163 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, lowlast, CTLTYPE_INT|CTLFLAG_RW,
164 	   &ipport_lowlastauto, 0, &sysctl_net_ipport_check, "I", "");
165 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, first, CTLTYPE_INT|CTLFLAG_RW,
166 	   &ipport_firstauto, 0, &sysctl_net_ipport_check, "I", "");
167 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, last, CTLTYPE_INT|CTLFLAG_RW,
168 	   &ipport_lastauto, 0, &sysctl_net_ipport_check, "I", "");
169 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, hifirst, CTLTYPE_INT|CTLFLAG_RW,
170 	   &ipport_hifirstauto, 0, &sysctl_net_ipport_check, "I", "");
171 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, hilast, CTLTYPE_INT|CTLFLAG_RW,
172 	   &ipport_hilastauto, 0, &sysctl_net_ipport_check, "I", "");
173 
174 /*
175  * in_pcb.c: manage the Protocol Control Blocks.
176  *
177  * NOTE: It is assumed that most of these functions will be called from
178  * a critical section.  XXX - There are, unfortunately, a few exceptions
179  * to this rule that should be fixed.
180  *
181  * NOTE: The caller should initialize the cpu field to the cpu running the
182  * protocol stack associated with this inpcbinfo.
183  */
184 
185 void
186 in_pcbinfo_init(struct inpcbinfo *pcbinfo)
187 {
188 	LIST_INIT(&pcbinfo->pcblisthead);
189 	pcbinfo->cpu = -1;
190 }
191 
192 /*
193  * Allocate a PCB and associate it with the socket.
194  */
195 int
196 in_pcballoc(struct socket *so, struct inpcbinfo *pcbinfo)
197 {
198 	struct inpcb *inp;
199 #ifdef IPSEC
200 	int error;
201 #endif
202 
203 	inp = kmalloc(pcbinfo->ipi_size, M_PCB, M_WAITOK|M_ZERO);
204 	inp->inp_gencnt = ++pcbinfo->ipi_gencnt;
205 	inp->inp_pcbinfo = inp->inp_cpcbinfo = pcbinfo;
206 	inp->inp_socket = so;
207 #ifdef IPSEC
208 	error = ipsec_init_policy(so, &inp->inp_sp);
209 	if (error != 0) {
210 		kfree(inp, M_PCB);
211 		return (error);
212 	}
213 #endif
214 #ifdef INET6
215 	if (INP_SOCKAF(so) == AF_INET6 && ip6_v6only)
216 		inp->inp_flags |= IN6P_IPV6_V6ONLY;
217 	if (ip6_auto_flowlabel)
218 		inp->inp_flags |= IN6P_AUTOFLOWLABEL;
219 #endif
220 	so->so_pcb = inp;
221 	LIST_INSERT_HEAD(&pcbinfo->pcblisthead, inp, inp_list);
222 	pcbinfo->ipi_count++;
223 	return (0);
224 }
225 
226 int
227 in_pcbbind(struct inpcb *inp, struct sockaddr *nam, struct thread *td)
228 {
229 	struct socket *so = inp->inp_socket;
230 	struct proc *p = td->td_proc;
231 	unsigned short *lastport;
232 	struct sockaddr_in *sin;
233 	struct sockaddr_in jsin;
234 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
235 	struct ucred *cred = NULL;
236 	u_short lport = 0;
237 	int wild = 0, reuseport = (so->so_options & SO_REUSEPORT);
238 	int error;
239 
240 	KKASSERT(p);
241 
242 	if (TAILQ_EMPTY(&in_ifaddrheads[mycpuid])) /* XXX broken! */
243 		return (EADDRNOTAVAIL);
244 	if (inp->inp_lport != 0 || inp->inp_laddr.s_addr != INADDR_ANY)
245 		return (EINVAL);	/* already bound */
246 	if (!(so->so_options & (SO_REUSEADDR|SO_REUSEPORT)))
247 		wild = 1;    /* neither SO_REUSEADDR nor SO_REUSEPORT is set */
248 	if (p)
249 		cred = p->p_ucred;
250 	if (nam != NULL) {
251 		sin = (struct sockaddr_in *)nam;
252 		if (nam->sa_len != sizeof *sin)
253 			return (EINVAL);
254 #ifdef notdef
255 		/*
256 		 * We should check the family, but old programs
257 		 * incorrectly fail to initialize it.
258 		 */
259 		if (sin->sin_family != AF_INET)
260 			return (EAFNOSUPPORT);
261 #endif
262 		if (!prison_replace_wildcards(td, nam))
263 				return (EINVAL);
264 		lport = sin->sin_port;
265 		if (IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) {
266 			/*
267 			 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
268 			 * allow complete duplication of binding if
269 			 * SO_REUSEPORT is set, or if SO_REUSEADDR is set
270 			 * and a multicast address is bound on both
271 			 * new and duplicated sockets.
272 			 */
273 			if (so->so_options & SO_REUSEADDR)
274 				reuseport = SO_REUSEADDR | SO_REUSEPORT;
275 		} else if (sin->sin_addr.s_addr != INADDR_ANY) {
276 			sin->sin_port = 0;		/* yech... */
277 			bzero(&sin->sin_zero, sizeof sin->sin_zero);
278 			if (ifa_ifwithaddr((struct sockaddr *)sin) == NULL)
279 				return (EADDRNOTAVAIL);
280 		}
281 		if (lport != 0) {
282 			struct inpcb *t;
283 
284 			/* GROSS */
285 			if (ntohs(lport) < IPPORT_RESERVED &&
286 			    cred && priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT, 0))
287 				return (EACCES);
288 			if (so->so_cred->cr_uid != 0 &&
289 			    !IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) {
290 				t = in_pcblookup_local(inp->inp_pcbinfo,
291 				    sin->sin_addr, lport,
292 				    INPLOOKUP_WILDCARD, cred);
293 				if (t &&
294 				    (!in_nullhost(sin->sin_addr) ||
295 				     !in_nullhost(t->inp_laddr) ||
296 				     (t->inp_socket->so_options &
297 					 SO_REUSEPORT) == 0) &&
298 				    (so->so_cred->cr_uid !=
299 				     t->inp_socket->so_cred->cr_uid)) {
300 #ifdef INET6
301 					if (!in_nullhost(sin->sin_addr) ||
302 					    !in_nullhost(t->inp_laddr) ||
303 					    INP_SOCKAF(so) ==
304 					    INP_SOCKAF(t->inp_socket))
305 #endif
306 					return (EADDRINUSE);
307 				}
308 			}
309 			if (cred && !prison_replace_wildcards(td, nam))
310 				return (EADDRNOTAVAIL);
311 			t = in_pcblookup_local(pcbinfo, sin->sin_addr, lport,
312 					       wild, cred);
313 			if (t && !(reuseport & t->inp_socket->so_options)) {
314 #ifdef INET6
315 				if (!in_nullhost(sin->sin_addr) ||
316 				    !in_nullhost(t->inp_laddr) ||
317 				    INP_SOCKAF(so) == INP_SOCKAF(t->inp_socket))
318 #endif
319 				return (EADDRINUSE);
320 			}
321 		}
322 		inp->inp_laddr = sin->sin_addr;
323 	}
324 	if (lport == 0) {
325 		ushort first, last;
326 		int count;
327 
328 		jsin.sin_family = AF_INET;
329 		jsin.sin_addr.s_addr = inp->inp_laddr.s_addr;
330 		if (!prison_replace_wildcards(td, (struct sockaddr *)&jsin)) {
331 			inp->inp_laddr.s_addr = INADDR_ANY;
332 			return (EINVAL);
333 		}
334 		inp->inp_laddr.s_addr = jsin.sin_addr.s_addr;
335 
336 		inp->inp_flags |= INP_ANONPORT;
337 
338 		if (inp->inp_flags & INP_HIGHPORT) {
339 			first = ipport_hifirstauto;	/* sysctl */
340 			last  = ipport_hilastauto;
341 			lastport = &pcbinfo->lasthi;
342 		} else if (inp->inp_flags & INP_LOWPORT) {
343 			if (cred &&
344 			    (error = priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT, 0))) {
345 				inp->inp_laddr.s_addr = INADDR_ANY;
346 				return (error);
347 			}
348 			first = ipport_lowfirstauto;	/* 1023 */
349 			last  = ipport_lowlastauto;	/* 600 */
350 			lastport = &pcbinfo->lastlow;
351 		} else {
352 			first = ipport_firstauto;	/* sysctl */
353 			last  = ipport_lastauto;
354 			lastport = &pcbinfo->lastport;
355 		}
356 		/*
357 		 * Simple check to ensure all ports are not used up causing
358 		 * a deadlock here.
359 		 *
360 		 * We split the two cases (up and down) so that the direction
361 		 * is not being tested on each round of the loop.
362 		 */
363 		if (first > last) {
364 			/*
365 			 * counting down
366 			 */
367 			count = first - last;
368 
369 			do {
370 				if (count-- < 0) {	/* completely used? */
371 					inp->inp_laddr.s_addr = INADDR_ANY;
372 					return (EADDRNOTAVAIL);
373 				}
374 				--*lastport;
375 				if (*lastport > first || *lastport < last)
376 					*lastport = first;
377 				lport = htons(*lastport);
378 			} while (in_pcblookup_local(pcbinfo, inp->inp_laddr,
379 				 lport, wild, cred));
380 		} else {
381 			/*
382 			 * counting up
383 			 */
384 			count = last - first;
385 
386 			do {
387 				if (count-- < 0) {	/* completely used? */
388 					inp->inp_laddr.s_addr = INADDR_ANY;
389 					return (EADDRNOTAVAIL);
390 				}
391 				++*lastport;
392 				if (*lastport < first || *lastport > last)
393 					*lastport = first;
394 				lport = htons(*lastport);
395 			} while (in_pcblookup_local(pcbinfo, inp->inp_laddr,
396 				 lport, wild, cred));
397 		}
398 	}
399 	inp->inp_lport = lport;
400 
401 	jsin.sin_family = AF_INET;
402 	jsin.sin_addr.s_addr = inp->inp_laddr.s_addr;
403 	if (!prison_replace_wildcards(td, (struct sockaddr*)&jsin)) {
404 		inp->inp_laddr.s_addr = INADDR_ANY;
405 		inp->inp_lport = 0;
406 		return (EINVAL);
407 	}
408 	inp->inp_laddr.s_addr = jsin.sin_addr.s_addr;
409 
410 	if (in_pcbinsporthash(inp) != 0) {
411 		inp->inp_laddr.s_addr = INADDR_ANY;
412 		inp->inp_lport = 0;
413 		return (EAGAIN);
414 	}
415 	return (0);
416 }
417 
418 /*
419  *   Transform old in_pcbconnect() into an inner subroutine for new
420  *   in_pcbconnect(): Do some validity-checking on the remote
421  *   address (in mbuf 'nam') and then determine local host address
422  *   (i.e., which interface) to use to access that remote host.
423  *
424  *   This preserves definition of in_pcbconnect(), while supporting a
425  *   slightly different version for T/TCP.  (This is more than
426  *   a bit of a kludge, but cleaning up the internal interfaces would
427  *   have forced minor changes in every protocol).
428  */
429 int
430 in_pcbladdr(struct inpcb *inp, struct sockaddr *nam,
431 	struct sockaddr_in **plocal_sin, struct thread *td)
432 {
433 	struct in_ifaddr *ia;
434 	struct ucred *cred = NULL;
435 	struct sockaddr_in *sin = (struct sockaddr_in *)nam;
436 	struct sockaddr *jsin;
437 	int jailed = 0, alloc_route = 0;
438 
439 	if (nam->sa_len != sizeof *sin)
440 		return (EINVAL);
441 	if (sin->sin_family != AF_INET)
442 		return (EAFNOSUPPORT);
443 	if (sin->sin_port == 0)
444 		return (EADDRNOTAVAIL);
445 	if (td && td->td_proc && td->td_proc->p_ucred)
446 		cred = td->td_proc->p_ucred;
447 	if (cred && cred->cr_prison)
448 		jailed = 1;
449 	if (!TAILQ_EMPTY(&in_ifaddrheads[mycpuid])) {
450 		ia = TAILQ_FIRST(&in_ifaddrheads[mycpuid])->ia;
451 		/*
452 		 * If the destination address is INADDR_ANY,
453 		 * use the primary local address.
454 		 * If the supplied address is INADDR_BROADCAST,
455 		 * and the primary interface supports broadcast,
456 		 * choose the broadcast address for that interface.
457 		 */
458 		if (sin->sin_addr.s_addr == INADDR_ANY)
459 			sin->sin_addr = IA_SIN(ia)->sin_addr;
460 		else if (sin->sin_addr.s_addr == (u_long)INADDR_BROADCAST &&
461 		    (ia->ia_ifp->if_flags & IFF_BROADCAST))
462 			sin->sin_addr = satosin(&ia->ia_broadaddr)->sin_addr;
463 	}
464 	if (inp->inp_laddr.s_addr == INADDR_ANY) {
465 		struct route *ro;
466 
467 		ia = NULL;
468 		/*
469 		 * If route is known or can be allocated now,
470 		 * our src addr is taken from the i/f, else punt.
471 		 * Note that we should check the address family of the cached
472 		 * destination, in case of sharing the cache with IPv6.
473 		 */
474 		ro = &inp->inp_route;
475 		if (ro->ro_rt &&
476 		    (!(ro->ro_rt->rt_flags & RTF_UP) ||
477 		     ro->ro_dst.sa_family != AF_INET ||
478 		     satosin(&ro->ro_dst)->sin_addr.s_addr !=
479 				      sin->sin_addr.s_addr ||
480 		     inp->inp_socket->so_options & SO_DONTROUTE)) {
481 			RTFREE(ro->ro_rt);
482 			ro->ro_rt = NULL;
483 		}
484 		if (!(inp->inp_socket->so_options & SO_DONTROUTE) && /*XXX*/
485 		    (ro->ro_rt == NULL ||
486 		    ro->ro_rt->rt_ifp == NULL)) {
487 			/* No route yet, so try to acquire one */
488 			bzero(&ro->ro_dst, sizeof(struct sockaddr_in));
489 			ro->ro_dst.sa_family = AF_INET;
490 			ro->ro_dst.sa_len = sizeof(struct sockaddr_in);
491 			((struct sockaddr_in *) &ro->ro_dst)->sin_addr =
492 				sin->sin_addr;
493 			rtalloc(ro);
494 			alloc_route = 1;
495 		}
496 		/*
497 		 * If we found a route, use the address
498 		 * corresponding to the outgoing interface
499 		 * unless it is the loopback (in case a route
500 		 * to our address on another net goes to loopback).
501 		 */
502 		if (ro->ro_rt && !(ro->ro_rt->rt_ifp->if_flags & IFF_LOOPBACK)) {
503 			if (jailed) {
504 				if (jailed_ip(cred->cr_prison,
505 				    ro->ro_rt->rt_ifa->ifa_addr)) {
506 					ia = ifatoia(ro->ro_rt->rt_ifa);
507 				}
508 			} else {
509 				ia = ifatoia(ro->ro_rt->rt_ifa);
510 			}
511 		}
512 		if (ia == NULL) {
513 			u_short fport = sin->sin_port;
514 
515 			sin->sin_port = 0;
516 			ia = ifatoia(ifa_ifwithdstaddr(sintosa(sin)));
517 			if (ia && jailed && !jailed_ip(cred->cr_prison,
518 			    sintosa(&ia->ia_addr)))
519 				ia = NULL;
520 			if (ia == NULL)
521 				ia = ifatoia(ifa_ifwithnet(sintosa(sin)));
522 			if (ia && jailed && !jailed_ip(cred->cr_prison,
523 			    sintosa(&ia->ia_addr)))
524 				ia = NULL;
525 			sin->sin_port = fport;
526 			if (ia == NULL &&
527 			    !TAILQ_EMPTY(&in_ifaddrheads[mycpuid]))
528 				ia = TAILQ_FIRST(&in_ifaddrheads[mycpuid])->ia;
529 			if (ia && jailed && !jailed_ip(cred->cr_prison,
530 			    sintosa(&ia->ia_addr)))
531 				ia = NULL;
532 
533 			if (!jailed && ia == NULL)
534 				goto fail;
535 		}
536 		/*
537 		 * If the destination address is multicast and an outgoing
538 		 * interface has been set as a multicast option, use the
539 		 * address of that interface as our source address.
540 		 */
541 		if (!jailed && IN_MULTICAST(ntohl(sin->sin_addr.s_addr)) &&
542 		    inp->inp_moptions != NULL) {
543 			struct ip_moptions *imo;
544 			struct ifnet *ifp;
545 
546 			imo = inp->inp_moptions;
547 			if (imo->imo_multicast_ifp != NULL) {
548 				struct in_ifaddr_container *iac;
549 
550 				ifp = imo->imo_multicast_ifp;
551 				ia = NULL;
552 				TAILQ_FOREACH(iac,
553 				&in_ifaddrheads[mycpuid], ia_link) {
554 					if (iac->ia->ia_ifp == ifp) {
555 						ia = iac->ia;
556 						break;
557 					}
558 				}
559 				if (ia == NULL)
560 					goto fail;
561 			}
562 		}
563 		/*
564 		 * Don't do pcblookup call here; return interface in plocal_sin
565 		 * and exit to caller, that will do the lookup.
566 		 */
567 		if (ia == NULL && jailed) {
568 			if ((jsin = prison_get_nonlocal(cred->cr_prison, AF_INET, NULL)) != NULL ||
569 			    (jsin = prison_get_local(cred->cr_prison, AF_INET, NULL)) != NULL) {
570 				*plocal_sin = satosin(jsin);
571 			} else {
572 				/* IPv6 only Jail */
573 				goto fail;
574 			}
575 		} else {
576 			*plocal_sin = &ia->ia_addr;
577 		}
578 	}
579 	return (0);
580 fail:
581 	if (alloc_route) {
582 		struct route *ro = &inp->inp_route;
583 
584 		if (ro->ro_rt != NULL)
585 			RTFREE(ro->ro_rt);
586 		bzero(ro, sizeof(*ro));
587 	}
588 	return (EADDRNOTAVAIL);
589 }
590 
591 /*
592  * Outer subroutine:
593  * Connect from a socket to a specified address.
594  * Both address and port must be specified in argument sin.
595  * If don't have a local address for this socket yet,
596  * then pick one.
597  */
598 int
599 in_pcbconnect(struct inpcb *inp, struct sockaddr *nam, struct thread *td)
600 {
601 	struct sockaddr_in *if_sin;
602 	struct sockaddr_in *sin = (struct sockaddr_in *)nam;
603 	int error;
604 
605 	/* Call inner routine to assign local interface address. */
606 	if ((error = in_pcbladdr(inp, nam, &if_sin, td)) != 0)
607 		return (error);
608 
609 	if (in_pcblookup_hash(inp->inp_cpcbinfo, sin->sin_addr, sin->sin_port,
610 	    inp->inp_laddr.s_addr ? inp->inp_laddr : if_sin->sin_addr,
611 	    inp->inp_lport, FALSE, NULL) != NULL) {
612 		return (EADDRINUSE);
613 	}
614 	if (inp->inp_laddr.s_addr == INADDR_ANY) {
615 		if (inp->inp_lport == 0) {
616 			error = in_pcbbind(inp, NULL, td);
617 			if (error)
618 				return (error);
619 		}
620 		inp->inp_laddr = if_sin->sin_addr;
621 	}
622 	inp->inp_faddr = sin->sin_addr;
623 	inp->inp_fport = sin->sin_port;
624 	in_pcbinsconnhash(inp);
625 	return (0);
626 }
627 
628 void
629 in_pcbdisconnect(struct inpcb *inp)
630 {
631 
632 	inp->inp_faddr.s_addr = INADDR_ANY;
633 	inp->inp_fport = 0;
634 	in_pcbremconnhash(inp);
635 	if (inp->inp_socket->so_state & SS_NOFDREF)
636 		in_pcbdetach(inp);
637 }
638 
639 void
640 in_pcbdetach(struct inpcb *inp)
641 {
642 	struct socket *so = inp->inp_socket;
643 	struct inpcbinfo *ipi = inp->inp_pcbinfo;
644 
645 #ifdef IPSEC
646 	ipsec4_delete_pcbpolicy(inp);
647 #endif /*IPSEC*/
648 	inp->inp_gencnt = ++ipi->ipi_gencnt;
649 	in_pcbremlists(inp);
650 	so->so_pcb = NULL;
651 	sofree(so);
652 	if (inp->inp_options)
653 		m_free(inp->inp_options);
654 	if (inp->inp_route.ro_rt)
655 		rtfree(inp->inp_route.ro_rt);
656 	ip_freemoptions(inp->inp_moptions);
657 	inp->inp_vflag = 0;
658 	kfree(inp, M_PCB);
659 }
660 
661 /*
662  * The calling convention of in_setsockaddr() and in_setpeeraddr() was
663  * modified to match the pru_sockaddr() and pru_peeraddr() entry points
664  * in struct pr_usrreqs, so that protocols can just reference then directly
665  * without the need for a wrapper function.  The socket must have a valid
666  * (i.e., non-nil) PCB, but it should be impossible to get an invalid one
667  * except through a kernel programming error, so it is acceptable to panic
668  * (or in this case trap) if the PCB is invalid.  (Actually, we don't trap
669  * because there actually /is/ a programming error somewhere... XXX)
670  */
671 int
672 in_setsockaddr(struct socket *so, struct sockaddr **nam)
673 {
674 	struct inpcb *inp;
675 	struct sockaddr_in *sin;
676 
677 	/*
678 	 * Do the malloc first in case it blocks.
679 	 */
680 	MALLOC(sin, struct sockaddr_in *, sizeof *sin, M_SONAME,
681 		M_WAITOK | M_ZERO);
682 	sin->sin_family = AF_INET;
683 	sin->sin_len = sizeof *sin;
684 
685 	crit_enter();
686 	inp = so->so_pcb;
687 	if (!inp) {
688 		crit_exit();
689 		kfree(sin, M_SONAME);
690 		return (ECONNRESET);
691 	}
692 	sin->sin_port = inp->inp_lport;
693 	sin->sin_addr = inp->inp_laddr;
694 	crit_exit();
695 
696 	*nam = (struct sockaddr *)sin;
697 	return (0);
698 }
699 
700 int
701 in_setpeeraddr(struct socket *so, struct sockaddr **nam)
702 {
703 	struct inpcb *inp;
704 	struct sockaddr_in *sin;
705 
706 	/*
707 	 * Do the malloc first in case it blocks.
708 	 */
709 	MALLOC(sin, struct sockaddr_in *, sizeof *sin, M_SONAME,
710 		M_WAITOK | M_ZERO);
711 	sin->sin_family = AF_INET;
712 	sin->sin_len = sizeof *sin;
713 
714 	crit_enter();
715 	inp = so->so_pcb;
716 	if (!inp) {
717 		crit_exit();
718 		kfree(sin, M_SONAME);
719 		return (ECONNRESET);
720 	}
721 	sin->sin_port = inp->inp_fport;
722 	sin->sin_addr = inp->inp_faddr;
723 	crit_exit();
724 
725 	*nam = (struct sockaddr *)sin;
726 	return (0);
727 }
728 
729 void
730 in_pcbnotifyall(struct inpcbhead *head, struct in_addr faddr, int err,
731 		void (*notify)(struct inpcb *, int))
732 {
733 	struct inpcb *inp, *ninp;
734 
735 	/*
736 	 * note: if INP_PLACEMARKER is set we must ignore the rest of
737 	 * the structure and skip it.
738 	 */
739 	crit_enter();
740 	LIST_FOREACH_MUTABLE(inp, head, inp_list, ninp) {
741 		if (inp->inp_flags & INP_PLACEMARKER)
742 			continue;
743 #ifdef INET6
744 		if (!(inp->inp_vflag & INP_IPV4))
745 			continue;
746 #endif
747 		if (inp->inp_faddr.s_addr != faddr.s_addr ||
748 		    inp->inp_socket == NULL)
749 			continue;
750 		(*notify)(inp, err);		/* can remove inp from list! */
751 	}
752 	crit_exit();
753 }
754 
755 void
756 in_pcbpurgeif0(struct inpcb *head, struct ifnet *ifp)
757 {
758 	struct inpcb *inp;
759 	struct ip_moptions *imo;
760 	int i, gap;
761 
762 	for (inp = head; inp != NULL; inp = LIST_NEXT(inp, inp_list)) {
763 		if (inp->inp_flags & INP_PLACEMARKER)
764 			continue;
765 		imo = inp->inp_moptions;
766 		if ((inp->inp_vflag & INP_IPV4) && imo != NULL) {
767 			/*
768 			 * Unselect the outgoing interface if it is being
769 			 * detached.
770 			 */
771 			if (imo->imo_multicast_ifp == ifp)
772 				imo->imo_multicast_ifp = NULL;
773 
774 			/*
775 			 * Drop multicast group membership if we joined
776 			 * through the interface being detached.
777 			 */
778 			for (i = 0, gap = 0; i < imo->imo_num_memberships;
779 			    i++) {
780 				if (imo->imo_membership[i]->inm_ifp == ifp) {
781 					in_delmulti(imo->imo_membership[i]);
782 					gap++;
783 				} else if (gap != 0)
784 					imo->imo_membership[i - gap] =
785 					    imo->imo_membership[i];
786 			}
787 			imo->imo_num_memberships -= gap;
788 		}
789 	}
790 }
791 
792 /*
793  * Check for alternatives when higher level complains
794  * about service problems.  For now, invalidate cached
795  * routing information.  If the route was created dynamically
796  * (by a redirect), time to try a default gateway again.
797  */
798 void
799 in_losing(struct inpcb *inp)
800 {
801 	struct rtentry *rt;
802 	struct rt_addrinfo rtinfo;
803 
804 	if ((rt = inp->inp_route.ro_rt)) {
805 		bzero(&rtinfo, sizeof(struct rt_addrinfo));
806 		rtinfo.rti_info[RTAX_DST] = rt_key(rt);
807 		rtinfo.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
808 		rtinfo.rti_info[RTAX_NETMASK] = rt_mask(rt);
809 		rtinfo.rti_flags = rt->rt_flags;
810 		rt_missmsg(RTM_LOSING, &rtinfo, rt->rt_flags, 0);
811 		if (rt->rt_flags & RTF_DYNAMIC)
812 			rtrequest1_global(RTM_DELETE, &rtinfo, NULL, NULL);
813 		inp->inp_route.ro_rt = NULL;
814 		rtfree(rt);
815 		/*
816 		 * A new route can be allocated
817 		 * the next time output is attempted.
818 		 */
819 	}
820 }
821 
822 /*
823  * After a routing change, flush old routing
824  * and allocate a (hopefully) better one.
825  */
826 void
827 in_rtchange(struct inpcb *inp, int err)
828 {
829 	if (inp->inp_route.ro_rt) {
830 		rtfree(inp->inp_route.ro_rt);
831 		inp->inp_route.ro_rt = NULL;
832 		/*
833 		 * A new route can be allocated the next time
834 		 * output is attempted.
835 		 */
836 	}
837 }
838 
839 /*
840  * Lookup a PCB based on the local address and port.
841  */
842 struct inpcb *
843 in_pcblookup_local(struct inpcbinfo *pcbinfo, struct in_addr laddr,
844 		   u_int lport_arg, int wild_okay, struct ucred *cred)
845 {
846 	struct inpcb *inp;
847 	int matchwild = 3, wildcard;
848 	u_short lport = lport_arg;
849 
850 	struct inpcbporthead *porthash;
851 	struct inpcbport *phd;
852 	struct inpcb *match = NULL;
853 
854 	/*
855 	 * Best fit PCB lookup.
856 	 *
857 	 * First see if this local port is in use by looking on the
858 	 * port hash list.
859 	 */
860 	porthash = &pcbinfo->porthashbase[INP_PCBPORTHASH(lport,
861 	    pcbinfo->porthashmask)];
862 	LIST_FOREACH(phd, porthash, phd_hash) {
863 		if (phd->phd_port == lport)
864 			break;
865 	}
866 	if (phd != NULL) {
867 		/*
868 		 * Port is in use by one or more PCBs. Look for best
869 		 * fit.
870 		 */
871 		LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) {
872 			wildcard = 0;
873 #ifdef INET6
874 			if ((inp->inp_vflag & INP_IPV4) == 0)
875 				continue;
876 #endif
877 			if (inp->inp_faddr.s_addr != INADDR_ANY)
878 				wildcard++;
879 			if (inp->inp_laddr.s_addr != INADDR_ANY) {
880 				if (laddr.s_addr == INADDR_ANY)
881 					wildcard++;
882 				else if (inp->inp_laddr.s_addr != laddr.s_addr)
883 					continue;
884 			} else {
885 				if (laddr.s_addr != INADDR_ANY)
886 					wildcard++;
887 			}
888 			if (wildcard && !wild_okay)
889 				continue;
890 			if (wildcard < matchwild &&
891 			    (cred == NULL ||
892 			     cred->cr_prison ==
893 					inp->inp_socket->so_cred->cr_prison)) {
894 				match = inp;
895 				matchwild = wildcard;
896 				if (matchwild == 0) {
897 					break;
898 				}
899 			}
900 		}
901 	}
902 	return (match);
903 }
904 
905 /*
906  * Lookup PCB in hash list.
907  */
908 struct inpcb *
909 in_pcblookup_hash(struct inpcbinfo *pcbinfo, struct in_addr faddr, u_int fport_arg,
910 		  struct in_addr laddr, u_int lport_arg, boolean_t wildcard,
911 		  struct ifnet *ifp)
912 {
913 	struct inpcbhead *head;
914 	struct inpcb *inp, *jinp=NULL;
915 	u_short fport = fport_arg, lport = lport_arg;
916 
917 	/*
918 	 * First look for an exact match.
919 	 */
920 	head = &pcbinfo->hashbase[INP_PCBCONNHASH(faddr.s_addr, fport,
921 	    laddr.s_addr, lport, pcbinfo->hashmask)];
922 	LIST_FOREACH(inp, head, inp_hash) {
923 #ifdef INET6
924 		if (!(inp->inp_vflag & INP_IPV4))
925 			continue;
926 #endif
927 		if (in_hosteq(inp->inp_faddr, faddr) &&
928 		    in_hosteq(inp->inp_laddr, laddr) &&
929 		    inp->inp_fport == fport && inp->inp_lport == lport) {
930 			/* found */
931 			if (inp->inp_socket == NULL ||
932 			    inp->inp_socket->so_cred->cr_prison == NULL) {
933 				return (inp);
934 			} else {
935 				if  (jinp == NULL)
936 					jinp = inp;
937 			}
938 		}
939 	}
940 	if (jinp != NULL)
941 		return (jinp);
942 	if (wildcard) {
943 		struct inpcb *local_wild = NULL;
944 		struct inpcb *jinp_wild = NULL;
945 #ifdef INET6
946 		struct inpcb *local_wild_mapped = NULL;
947 #endif
948 		struct inpcontainer *ic;
949 		struct inpcontainerhead *chead;
950 		struct sockaddr_in jsin;
951 		struct ucred *cred;
952 
953 		/*
954 		 * Order of socket selection:
955 		 * 1. non-jailed, non-wild.
956 		 * 2. non-jailed, wild.
957 		 * 3. jailed, non-wild.
958 		 * 4. jailed, wild.
959 		 */
960 		jsin.sin_family = AF_INET;
961 		chead = &pcbinfo->wildcardhashbase[
962 		    INP_PCBWILDCARDHASH(lport, pcbinfo->wildcardhashmask)];
963 		LIST_FOREACH(ic, chead, ic_list) {
964 			inp = ic->ic_inp;
965 			jsin.sin_addr.s_addr = laddr.s_addr;
966 #ifdef INET6
967 			if (!(inp->inp_vflag & INP_IPV4))
968 				continue;
969 #endif
970 			if (inp->inp_socket != NULL)
971 				cred = inp->inp_socket->so_cred;
972 			else
973 				cred = NULL;
974 			if (cred != NULL && jailed(cred)) {
975 				if (jinp != NULL)
976 					continue;
977 				else
978 					if (!jailed_ip(cred->cr_prison,
979 					    (struct sockaddr *)&jsin))
980 						continue;
981 			}
982 			if (inp->inp_lport == lport) {
983 				if (ifp && ifp->if_type == IFT_FAITH &&
984 				    !(inp->inp_flags & INP_FAITH))
985 					continue;
986 				if (inp->inp_laddr.s_addr == laddr.s_addr) {
987 					if (cred != NULL && jailed(cred))
988 						jinp = inp;
989 					else
990 						return (inp);
991 				}
992 				if (inp->inp_laddr.s_addr == INADDR_ANY) {
993 #ifdef INET6
994 					if (INP_CHECK_SOCKAF(inp->inp_socket,
995 							     AF_INET6))
996 						local_wild_mapped = inp;
997 					else
998 #endif
999 						if (cred != NULL &&
1000 						    jailed(cred))
1001 							jinp_wild = inp;
1002 						else
1003 							local_wild = inp;
1004 				}
1005 			}
1006 		}
1007 		if (local_wild != NULL)
1008 			return (local_wild);
1009 #ifdef INET6
1010 		if (local_wild_mapped != NULL)
1011 			return (local_wild_mapped);
1012 #endif
1013 		if (jinp != NULL)
1014 			return (jinp);
1015 		return (jinp_wild);
1016 	}
1017 
1018 	/*
1019 	 * Not found.
1020 	 */
1021 	return (NULL);
1022 }
1023 
1024 /*
1025  * Insert PCB into connection hash table.
1026  */
1027 void
1028 in_pcbinsconnhash(struct inpcb *inp)
1029 {
1030 	struct inpcbinfo *pcbinfo = inp->inp_cpcbinfo;
1031 	struct inpcbhead *bucket;
1032 	u_int32_t hashkey_faddr, hashkey_laddr;
1033 
1034 #ifdef INET6
1035 	if (inp->inp_vflag & INP_IPV6) {
1036 		hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX JH */;
1037 		hashkey_laddr = inp->in6p_laddr.s6_addr32[3] /* XXX JH */;
1038 	} else {
1039 #endif
1040 		hashkey_faddr = inp->inp_faddr.s_addr;
1041 		hashkey_laddr = inp->inp_laddr.s_addr;
1042 #ifdef INET6
1043 	}
1044 #endif
1045 
1046 	KASSERT(!(inp->inp_flags & INP_CONNECTED), ("already on hash list"));
1047 	inp->inp_flags |= INP_CONNECTED;
1048 
1049 	/*
1050 	 * Insert into the connection hash table.
1051 	 */
1052 	bucket = &pcbinfo->hashbase[INP_PCBCONNHASH(hashkey_faddr,
1053 	    inp->inp_fport, hashkey_laddr, inp->inp_lport, pcbinfo->hashmask)];
1054 	LIST_INSERT_HEAD(bucket, inp, inp_hash);
1055 }
1056 
1057 /*
1058  * Remove PCB from connection hash table.
1059  */
1060 void
1061 in_pcbremconnhash(struct inpcb *inp)
1062 {
1063 	KASSERT(inp->inp_flags & INP_CONNECTED, ("inp not connected"));
1064 	LIST_REMOVE(inp, inp_hash);
1065 	inp->inp_flags &= ~INP_CONNECTED;
1066 }
1067 
1068 /*
1069  * Insert PCB into port hash table.
1070  */
1071 int
1072 in_pcbinsporthash(struct inpcb *inp)
1073 {
1074 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1075 	struct inpcbporthead *pcbporthash;
1076 	struct inpcbport *phd;
1077 
1078 	/*
1079 	 * Insert into the port hash table.
1080 	 */
1081 	pcbporthash = &pcbinfo->porthashbase[
1082 	    INP_PCBPORTHASH(inp->inp_lport, pcbinfo->porthashmask)];
1083 
1084 	/* Go through port list and look for a head for this lport. */
1085 	LIST_FOREACH(phd, pcbporthash, phd_hash)
1086 		if (phd->phd_port == inp->inp_lport)
1087 			break;
1088 
1089 	/* If none exists, malloc one and tack it on. */
1090 	if (phd == NULL) {
1091 		MALLOC(phd, struct inpcbport *, sizeof(struct inpcbport),
1092 		    M_PCB, M_INTWAIT | M_NULLOK);
1093 		if (phd == NULL)
1094 			return (ENOBUFS); /* XXX */
1095 		phd->phd_port = inp->inp_lport;
1096 		LIST_INIT(&phd->phd_pcblist);
1097 		LIST_INSERT_HEAD(pcbporthash, phd, phd_hash);
1098 	}
1099 
1100 	inp->inp_phd = phd;
1101 	LIST_INSERT_HEAD(&phd->phd_pcblist, inp, inp_portlist);
1102 
1103 	return (0);
1104 }
1105 
1106 void
1107 in_pcbinswildcardhash_oncpu(struct inpcb *inp, struct inpcbinfo *pcbinfo)
1108 {
1109 	struct inpcontainer *ic;
1110 	struct inpcontainerhead *bucket;
1111 
1112 	bucket = &pcbinfo->wildcardhashbase[
1113 	    INP_PCBWILDCARDHASH(inp->inp_lport, pcbinfo->wildcardhashmask)];
1114 
1115 	ic = kmalloc(sizeof(struct inpcontainer), M_TEMP, M_INTWAIT);
1116 	ic->ic_inp = inp;
1117 	LIST_INSERT_HEAD(bucket, ic, ic_list);
1118 }
1119 
1120 /*
1121  * Insert PCB into wildcard hash table.
1122  */
1123 void
1124 in_pcbinswildcardhash(struct inpcb *inp)
1125 {
1126 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1127 
1128 	KKASSERT(pcbinfo != NULL);
1129 
1130 	in_pcbinswildcardhash_oncpu(inp, pcbinfo);
1131 	inp->inp_flags |= INP_WILDCARD;
1132 }
1133 
1134 void
1135 in_pcbremwildcardhash_oncpu(struct inpcb *inp, struct inpcbinfo *pcbinfo)
1136 {
1137 	struct inpcontainer *ic;
1138 	struct inpcontainerhead *head;
1139 
1140 	/* find bucket */
1141 	head = &pcbinfo->wildcardhashbase[
1142 	    INP_PCBWILDCARDHASH(inp->inp_lport, pcbinfo->wildcardhashmask)];
1143 
1144 	LIST_FOREACH(ic, head, ic_list) {
1145 		if (ic->ic_inp == inp)
1146 			goto found;
1147 	}
1148 	return;			/* not found! */
1149 
1150 found:
1151 	LIST_REMOVE(ic, ic_list);	/* remove container from bucket chain */
1152 	kfree(ic, M_TEMP);		/* deallocate container */
1153 }
1154 
1155 /*
1156  * Remove PCB from wildcard hash table.
1157  */
1158 void
1159 in_pcbremwildcardhash(struct inpcb *inp)
1160 {
1161 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1162 
1163 	KASSERT(inp->inp_flags & INP_WILDCARD, ("inp not wildcard"));
1164 	in_pcbremwildcardhash_oncpu(inp, pcbinfo);
1165 	inp->inp_flags &= ~INP_WILDCARD;
1166 }
1167 
1168 /*
1169  * Remove PCB from various lists.
1170  */
1171 void
1172 in_pcbremlists(struct inpcb *inp)
1173 {
1174 	if (inp->inp_lport) {
1175 		struct inpcbport *phd = inp->inp_phd;
1176 
1177 		LIST_REMOVE(inp, inp_portlist);
1178 		if (LIST_FIRST(&phd->phd_pcblist) == NULL) {
1179 			LIST_REMOVE(phd, phd_hash);
1180 			kfree(phd, M_PCB);
1181 		}
1182 	}
1183 	if (inp->inp_flags & INP_WILDCARD) {
1184 		in_pcbremwildcardhash(inp);
1185 	} else if (inp->inp_flags & INP_CONNECTED) {
1186 		in_pcbremconnhash(inp);
1187 	}
1188 	LIST_REMOVE(inp, inp_list);
1189 	inp->inp_pcbinfo->ipi_count--;
1190 }
1191 
1192 int
1193 prison_xinpcb(struct thread *td, struct inpcb *inp)
1194 {
1195 	struct ucred *cr;
1196 
1197 	if (td->td_proc == NULL)
1198 		return (0);
1199 	cr = td->td_proc->p_ucred;
1200 	if (cr->cr_prison == NULL)
1201 		return (0);
1202 	if (inp->inp_socket && inp->inp_socket->so_cred &&
1203 	    inp->inp_socket->so_cred->cr_prison &&
1204 	    cr->cr_prison == inp->inp_socket->so_cred->cr_prison)
1205 		return (0);
1206 	return (1);
1207 }
1208 
1209 int
1210 in_pcblist_global(SYSCTL_HANDLER_ARGS)
1211 {
1212 	struct inpcbinfo *pcbinfo = arg1;
1213 	struct inpcb *inp, *marker;
1214 	struct xinpcb xi;
1215 	int error, i, n;
1216 	inp_gen_t gencnt;
1217 
1218 	/*
1219 	 * The process of preparing the TCB list is too time-consuming and
1220 	 * resource-intensive to repeat twice on every request.
1221 	 */
1222 	if (req->oldptr == NULL) {
1223 		n = pcbinfo->ipi_count;
1224 		req->oldidx = (n + n/8 + 10) * sizeof(struct xinpcb);
1225 		return 0;
1226 	}
1227 
1228 	if (req->newptr != NULL)
1229 		return EPERM;
1230 
1231 	/*
1232 	 * OK, now we're committed to doing something.  Re-fetch ipi_count
1233 	 * after obtaining the generation count.
1234 	 */
1235 	gencnt = pcbinfo->ipi_gencnt;
1236 	n = pcbinfo->ipi_count;
1237 
1238 	marker = kmalloc(sizeof(struct inpcb), M_TEMP, M_WAITOK|M_ZERO);
1239 	marker->inp_flags |= INP_PLACEMARKER;
1240 	LIST_INSERT_HEAD(&pcbinfo->pcblisthead, marker, inp_list);
1241 
1242 	i = 0;
1243 	error = 0;
1244 
1245 	while ((inp = LIST_NEXT(marker, inp_list)) != NULL && i < n) {
1246 		LIST_REMOVE(marker, inp_list);
1247 		LIST_INSERT_AFTER(inp, marker, inp_list);
1248 
1249 		if (inp->inp_flags & INP_PLACEMARKER)
1250 			continue;
1251 		if (inp->inp_gencnt > gencnt)
1252 			continue;
1253 		if (prison_xinpcb(req->td, inp))
1254 			continue;
1255 		bzero(&xi, sizeof xi);
1256 		xi.xi_len = sizeof xi;
1257 		bcopy(inp, &xi.xi_inp, sizeof *inp);
1258 		if (inp->inp_socket)
1259 			sotoxsocket(inp->inp_socket, &xi.xi_socket);
1260 		if ((error = SYSCTL_OUT(req, &xi, sizeof xi)) != 0)
1261 			break;
1262 		++i;
1263 	}
1264 	LIST_REMOVE(marker, inp_list);
1265 	if (error == 0 && i < n) {
1266 		bzero(&xi, sizeof xi);
1267 		xi.xi_len = sizeof xi;
1268 		while (i < n) {
1269 			error = SYSCTL_OUT(req, &xi, sizeof xi);
1270 			++i;
1271 		}
1272 	}
1273 	kfree(marker, M_TEMP);
1274 	return(error);
1275 }
1276