xref: /dflybsd-src/sys/netinet/in_pcb.c (revision 744c01d0dc2aa1481a40e5b0988d15691602f5c9)
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) 2004 Jeffrey M. Hsu.  All rights reserved.
36  *
37  * License terms: all terms for the DragonFly license above plus the following:
38  *
39  * 4. All advertising materials mentioning features or use of this software
40  *    must display the following acknowledgement:
41  *
42  *	This product includes software developed by Jeffrey M. Hsu
43  *	for the DragonFly Project.
44  *
45  *    This requirement may be waived with permission from Jeffrey Hsu.
46  *    This requirement will sunset and may be removed on July 8 2005,
47  *    after which the standard DragonFly license (as shown above) will
48  *    apply.
49  */
50 
51 /*
52  * Copyright (c) 1982, 1986, 1991, 1993, 1995
53  *	The Regents of the University of California.  All rights reserved.
54  *
55  * Redistribution and use in source and binary forms, with or without
56  * modification, are permitted provided that the following conditions
57  * are met:
58  * 1. Redistributions of source code must retain the above copyright
59  *    notice, this list of conditions and the following disclaimer.
60  * 2. Redistributions in binary form must reproduce the above copyright
61  *    notice, this list of conditions and the following disclaimer in the
62  *    documentation and/or other materials provided with the distribution.
63  * 3. All advertising materials mentioning features or use of this software
64  *    must display the following acknowledgement:
65  *	This product includes software developed by the University of
66  *	California, Berkeley and its contributors.
67  * 4. Neither the name of the University nor the names of its contributors
68  *    may be used to endorse or promote products derived from this software
69  *    without specific prior written permission.
70  *
71  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
72  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
73  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
74  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
75  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
76  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
77  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
78  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
79  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
80  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
81  * SUCH DAMAGE.
82  *
83  *	@(#)in_pcb.c	8.4 (Berkeley) 5/24/95
84  * $FreeBSD: src/sys/netinet/in_pcb.c,v 1.59.2.27 2004/01/02 04:06:42 ambrisko Exp $
85  * $DragonFly: src/sys/netinet/in_pcb.c,v 1.41 2006/12/29 18:02:56 victor Exp $
86  */
87 
88 #include "opt_ipsec.h"
89 #include "opt_inet6.h"
90 
91 #include <sys/param.h>
92 #include <sys/systm.h>
93 #include <sys/malloc.h>
94 #include <sys/mbuf.h>
95 #include <sys/domain.h>
96 #include <sys/protosw.h>
97 #include <sys/socket.h>
98 #include <sys/socketvar.h>
99 #include <sys/proc.h>
100 #include <sys/jail.h>
101 #include <sys/kernel.h>
102 #include <sys/sysctl.h>
103 #include <sys/thread2.h>
104 
105 #include <machine/limits.h>
106 
107 #include <vm/vm_zone.h>
108 
109 #include <net/if.h>
110 #include <net/if_types.h>
111 #include <net/route.h>
112 
113 #include <netinet/in.h>
114 #include <netinet/in_pcb.h>
115 #include <netinet/in_var.h>
116 #include <netinet/ip_var.h>
117 #ifdef INET6
118 #include <netinet/ip6.h>
119 #include <netinet6/ip6_var.h>
120 #endif /* INET6 */
121 
122 #ifdef IPSEC
123 #include <netinet6/ipsec.h>
124 #include <netproto/key/key.h>
125 #endif
126 
127 #ifdef FAST_IPSEC
128 #if defined(IPSEC) || defined(IPSEC_ESP)
129 #error "Bad idea: don't compile with both IPSEC and FAST_IPSEC!"
130 #endif
131 
132 #include <netproto/ipsec/ipsec.h>
133 #include <netproto/ipsec/key.h>
134 #define	IPSEC
135 #endif /* FAST_IPSEC */
136 
137 struct in_addr zeroin_addr;
138 
139 /*
140  * These configure the range of local port addresses assigned to
141  * "unspecified" outgoing connections/packets/whatever.
142  */
143 int ipport_lowfirstauto = IPPORT_RESERVED - 1;	/* 1023 */
144 int ipport_lowlastauto = IPPORT_RESERVEDSTART;	/* 600 */
145 
146 int ipport_firstauto = IPPORT_RESERVED;		/* 1024 */
147 int ipport_lastauto = IPPORT_USERRESERVED;	/* 5000 */
148 
149 int ipport_hifirstauto = IPPORT_HIFIRSTAUTO;	/* 49152 */
150 int ipport_hilastauto = IPPORT_HILASTAUTO;	/* 65535 */
151 
152 static __inline void
153 RANGECHK(int var, int min, int max)
154 {
155 	if (var < min)
156 		var = min;
157 	else if (var > max)
158 		var = max;
159 }
160 
161 static int
162 sysctl_net_ipport_check(SYSCTL_HANDLER_ARGS)
163 {
164 	int error;
165 
166 	error = sysctl_handle_int(oidp, oidp->oid_arg1, oidp->oid_arg2, req);
167 	if (!error) {
168 		RANGECHK(ipport_lowfirstauto, 1, IPPORT_RESERVED - 1);
169 		RANGECHK(ipport_lowlastauto, 1, IPPORT_RESERVED - 1);
170 
171 		RANGECHK(ipport_firstauto, IPPORT_RESERVED, USHRT_MAX);
172 		RANGECHK(ipport_lastauto, IPPORT_RESERVED, USHRT_MAX);
173 
174 		RANGECHK(ipport_hifirstauto, IPPORT_RESERVED, USHRT_MAX);
175 		RANGECHK(ipport_hilastauto, IPPORT_RESERVED, USHRT_MAX);
176 	}
177 	return (error);
178 }
179 
180 SYSCTL_NODE(_net_inet_ip, IPPROTO_IP, portrange, CTLFLAG_RW, 0, "IP Ports");
181 
182 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, lowfirst, CTLTYPE_INT|CTLFLAG_RW,
183 	   &ipport_lowfirstauto, 0, &sysctl_net_ipport_check, "I", "");
184 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, lowlast, CTLTYPE_INT|CTLFLAG_RW,
185 	   &ipport_lowlastauto, 0, &sysctl_net_ipport_check, "I", "");
186 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, first, CTLTYPE_INT|CTLFLAG_RW,
187 	   &ipport_firstauto, 0, &sysctl_net_ipport_check, "I", "");
188 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, last, CTLTYPE_INT|CTLFLAG_RW,
189 	   &ipport_lastauto, 0, &sysctl_net_ipport_check, "I", "");
190 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, hifirst, CTLTYPE_INT|CTLFLAG_RW,
191 	   &ipport_hifirstauto, 0, &sysctl_net_ipport_check, "I", "");
192 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, hilast, CTLTYPE_INT|CTLFLAG_RW,
193 	   &ipport_hilastauto, 0, &sysctl_net_ipport_check, "I", "");
194 
195 /*
196  * in_pcb.c: manage the Protocol Control Blocks.
197  *
198  * NOTE: It is assumed that most of these functions will be called from
199  * a critical section.  XXX - There are, unfortunately, a few exceptions
200  * to this rule that should be fixed.
201  *
202  * NOTE: The caller should initialize the cpu field to the cpu running the
203  * protocol stack associated with this inpcbinfo.
204  */
205 
206 void
207 in_pcbinfo_init(struct inpcbinfo *pcbinfo)
208 {
209 	LIST_INIT(&pcbinfo->pcblisthead);
210 	pcbinfo->cpu = -1;
211 }
212 
213 /*
214  * Allocate a PCB and associate it with the socket.
215  */
216 int
217 in_pcballoc(struct socket *so, struct inpcbinfo *pcbinfo)
218 {
219 	struct inpcb *inp;
220 #ifdef IPSEC
221 	int error;
222 #endif
223 
224 	inp = zalloc(pcbinfo->ipi_zone);
225 	if (inp == NULL)
226 		return (ENOBUFS);
227 	bzero(inp, sizeof *inp);
228 	inp->inp_gencnt = ++pcbinfo->ipi_gencnt;
229 	inp->inp_pcbinfo = inp->inp_cpcbinfo = pcbinfo;
230 	inp->inp_socket = so;
231 #ifdef IPSEC
232 	error = ipsec_init_policy(so, &inp->inp_sp);
233 	if (error != 0) {
234 		zfree(pcbinfo->ipi_zone, inp);
235 		return (error);
236 	}
237 #endif
238 #ifdef INET6
239 	if (INP_SOCKAF(so) == AF_INET6 && ip6_v6only)
240 		inp->inp_flags |= IN6P_IPV6_V6ONLY;
241 	if (ip6_auto_flowlabel)
242 		inp->inp_flags |= IN6P_AUTOFLOWLABEL;
243 #endif
244 	so->so_pcb = inp;
245 	LIST_INSERT_HEAD(&pcbinfo->pcblisthead, inp, inp_list);
246 	pcbinfo->ipi_count++;
247 	return (0);
248 }
249 
250 int
251 in_pcbbind(struct inpcb *inp, struct sockaddr *nam, struct thread *td)
252 {
253 	struct socket *so = inp->inp_socket;
254 	struct proc *p = td->td_proc;
255 	unsigned short *lastport;
256 	struct sockaddr_in *sin;
257 	struct sockaddr_in jsin;
258 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
259 	struct ucred *cred = NULL;
260 	u_short lport = 0;
261 	int wild = 0, reuseport = (so->so_options & SO_REUSEPORT);
262 	int error;
263 
264 	KKASSERT(p);
265 
266 	if (TAILQ_EMPTY(&in_ifaddrhead)) /* XXX broken! */
267 		return (EADDRNOTAVAIL);
268 	if (inp->inp_lport != 0 || inp->inp_laddr.s_addr != INADDR_ANY)
269 		return (EINVAL);	/* already bound */
270 	if (!(so->so_options & (SO_REUSEADDR|SO_REUSEPORT)))
271 		wild = 1;    /* neither SO_REUSEADDR nor SO_REUSEPORT is set */
272 	if (p)
273 		cred = p->p_ucred;
274 	if (nam != NULL) {
275 		sin = (struct sockaddr_in *)nam;
276 		if (nam->sa_len != sizeof *sin)
277 			return (EINVAL);
278 #ifdef notdef
279 		/*
280 		 * We should check the family, but old programs
281 		 * incorrectly fail to initialize it.
282 		 */
283 		if (sin->sin_family != AF_INET)
284 			return (EAFNOSUPPORT);
285 #endif
286 		if (!prison_replace_wildcards(td, nam))
287 				return (EINVAL);
288 		lport = sin->sin_port;
289 		if (IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) {
290 			/*
291 			 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
292 			 * allow complete duplication of binding if
293 			 * SO_REUSEPORT is set, or if SO_REUSEADDR is set
294 			 * and a multicast address is bound on both
295 			 * new and duplicated sockets.
296 			 */
297 			if (so->so_options & SO_REUSEADDR)
298 				reuseport = SO_REUSEADDR | SO_REUSEPORT;
299 		} else if (sin->sin_addr.s_addr != INADDR_ANY) {
300 			sin->sin_port = 0;		/* yech... */
301 			bzero(&sin->sin_zero, sizeof sin->sin_zero);
302 			if (ifa_ifwithaddr((struct sockaddr *)sin) == NULL)
303 				return (EADDRNOTAVAIL);
304 		}
305 		if (lport != 0) {
306 			struct inpcb *t;
307 
308 			/* GROSS */
309 			if (ntohs(lport) < IPPORT_RESERVED &&
310 			    cred && suser_cred(cred, PRISON_ROOT))
311 				return (EACCES);
312 			if (so->so_cred->cr_uid != 0 &&
313 			    !IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) {
314 				t = in_pcblookup_local(inp->inp_pcbinfo,
315 				    sin->sin_addr, lport,
316 				    INPLOOKUP_WILDCARD, cred);
317 				if (t &&
318 				    (!in_nullhost(sin->sin_addr) ||
319 				     !in_nullhost(t->inp_laddr) ||
320 				     (t->inp_socket->so_options &
321 					 SO_REUSEPORT) == 0) &&
322 				    (so->so_cred->cr_uid !=
323 				     t->inp_socket->so_cred->cr_uid)) {
324 #ifdef INET6
325 					if (!in_nullhost(sin->sin_addr) ||
326 					    !in_nullhost(t->inp_laddr) ||
327 					    INP_SOCKAF(so) ==
328 					    INP_SOCKAF(t->inp_socket))
329 #endif
330 					return (EADDRINUSE);
331 				}
332 			}
333 			if (cred && !prison_replace_wildcards(td, nam))
334 				return (EADDRNOTAVAIL);
335 			t = in_pcblookup_local(pcbinfo, sin->sin_addr, lport,
336 					       wild, cred);
337 			if (t && !(reuseport & t->inp_socket->so_options)) {
338 #ifdef INET6
339 				if (!in_nullhost(sin->sin_addr) ||
340 				    !in_nullhost(t->inp_laddr) ||
341 				    INP_SOCKAF(so) == INP_SOCKAF(t->inp_socket))
342 #endif
343 				return (EADDRINUSE);
344 			}
345 		}
346 		inp->inp_laddr = sin->sin_addr;
347 	}
348 	if (lport == 0) {
349 		ushort first, last;
350 		int count;
351 
352 		jsin.sin_family = AF_INET;
353 		jsin.sin_addr.s_addr = inp->inp_laddr.s_addr;
354 		if (!prison_replace_wildcards(td, (struct sockaddr *)&jsin)) {
355 			inp->inp_laddr.s_addr = INADDR_ANY;
356 			return (EINVAL);
357 		}
358 		inp->inp_laddr.s_addr = jsin.sin_addr.s_addr;
359 
360 		inp->inp_flags |= INP_ANONPORT;
361 
362 		if (inp->inp_flags & INP_HIGHPORT) {
363 			first = ipport_hifirstauto;	/* sysctl */
364 			last  = ipport_hilastauto;
365 			lastport = &pcbinfo->lasthi;
366 		} else if (inp->inp_flags & INP_LOWPORT) {
367 			if (cred &&
368 			    (error = suser_cred(cred, PRISON_ROOT))) {
369 				inp->inp_laddr.s_addr = INADDR_ANY;
370 				return (error);
371 			}
372 			first = ipport_lowfirstauto;	/* 1023 */
373 			last  = ipport_lowlastauto;	/* 600 */
374 			lastport = &pcbinfo->lastlow;
375 		} else {
376 			first = ipport_firstauto;	/* sysctl */
377 			last  = ipport_lastauto;
378 			lastport = &pcbinfo->lastport;
379 		}
380 		/*
381 		 * Simple check to ensure all ports are not used up causing
382 		 * a deadlock here.
383 		 *
384 		 * We split the two cases (up and down) so that the direction
385 		 * is not being tested on each round of the loop.
386 		 */
387 		if (first > last) {
388 			/*
389 			 * counting down
390 			 */
391 			count = first - last;
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 		} else {
405 			/*
406 			 * counting up
407 			 */
408 			count = last - first;
409 
410 			do {
411 				if (count-- < 0) {	/* completely used? */
412 					inp->inp_laddr.s_addr = INADDR_ANY;
413 					return (EADDRNOTAVAIL);
414 				}
415 				++*lastport;
416 				if (*lastport < first || *lastport > last)
417 					*lastport = first;
418 				lport = htons(*lastport);
419 			} while (in_pcblookup_local(pcbinfo, inp->inp_laddr,
420 				 lport, wild, cred));
421 		}
422 	}
423 	inp->inp_lport = lport;
424 
425 	jsin.sin_family = AF_INET;
426 	jsin.sin_addr.s_addr = inp->inp_laddr.s_addr;
427 	if (!prison_replace_wildcards(td, (struct sockaddr*)&jsin)) {
428 		inp->inp_laddr.s_addr = INADDR_ANY;
429 		inp->inp_lport = 0;
430 		return (EINVAL);
431 	}
432 	inp->inp_laddr.s_addr = jsin.sin_addr.s_addr;
433 
434 	if (in_pcbinsporthash(inp) != 0) {
435 		inp->inp_laddr.s_addr = INADDR_ANY;
436 		inp->inp_lport = 0;
437 		return (EAGAIN);
438 	}
439 	return (0);
440 }
441 
442 /*
443  *   Transform old in_pcbconnect() into an inner subroutine for new
444  *   in_pcbconnect(): Do some validity-checking on the remote
445  *   address (in mbuf 'nam') and then determine local host address
446  *   (i.e., which interface) to use to access that remote host.
447  *
448  *   This preserves definition of in_pcbconnect(), while supporting a
449  *   slightly different version for T/TCP.  (This is more than
450  *   a bit of a kludge, but cleaning up the internal interfaces would
451  *   have forced minor changes in every protocol).
452  */
453 int
454 in_pcbladdr(struct inpcb *inp, struct sockaddr *nam,
455 	struct sockaddr_in **plocal_sin, struct thread *td)
456 {
457 	struct in_ifaddr *ia;
458 	struct ucred *cred = NULL;
459 	struct sockaddr_in *sin = (struct sockaddr_in *)nam;
460 	struct sockaddr_in jsin;
461 	int jailed = 0;
462 
463 	if (nam->sa_len != sizeof *sin)
464 		return (EINVAL);
465 	if (sin->sin_family != AF_INET)
466 		return (EAFNOSUPPORT);
467 	if (sin->sin_port == 0)
468 		return (EADDRNOTAVAIL);
469 	if (td && td->td_proc && td->td_proc->p_ucred)
470 		cred = td->td_proc->p_ucred;
471 	if (cred && cred->cr_prison)
472 		jailed = 1;
473 	if (!TAILQ_EMPTY(&in_ifaddrhead)) {
474 		ia = TAILQ_FIRST(&in_ifaddrhead);
475 		/*
476 		 * If the destination address is INADDR_ANY,
477 		 * use the primary local address.
478 		 * If the supplied address is INADDR_BROADCAST,
479 		 * and the primary interface supports broadcast,
480 		 * choose the broadcast address for that interface.
481 		 */
482 		if (sin->sin_addr.s_addr == INADDR_ANY)
483 			sin->sin_addr = IA_SIN(ia)->sin_addr;
484 		else if (sin->sin_addr.s_addr == (u_long)INADDR_BROADCAST &&
485 		    (ia->ia_ifp->if_flags & IFF_BROADCAST))
486 			sin->sin_addr = satosin(&ia->ia_broadaddr)->sin_addr;
487 	}
488 	if (inp->inp_laddr.s_addr == INADDR_ANY) {
489 		struct route *ro;
490 
491 		ia = (struct in_ifaddr *)NULL;
492 		/*
493 		 * If route is known or can be allocated now,
494 		 * our src addr is taken from the i/f, else punt.
495 		 * Note that we should check the address family of the cached
496 		 * destination, in case of sharing the cache with IPv6.
497 		 */
498 		ro = &inp->inp_route;
499 		if (ro->ro_rt &&
500 		    (!(ro->ro_rt->rt_flags & RTF_UP) ||
501 		     ro->ro_dst.sa_family != AF_INET ||
502 		     satosin(&ro->ro_dst)->sin_addr.s_addr !=
503 				      sin->sin_addr.s_addr ||
504 		     inp->inp_socket->so_options & SO_DONTROUTE)) {
505 			RTFREE(ro->ro_rt);
506 			ro->ro_rt = (struct rtentry *)NULL;
507 		}
508 		if (!(inp->inp_socket->so_options & SO_DONTROUTE) && /*XXX*/
509 		    (ro->ro_rt == (struct rtentry *)NULL ||
510 		    ro->ro_rt->rt_ifp == (struct ifnet *)NULL)) {
511 			/* No route yet, so try to acquire one */
512 			bzero(&ro->ro_dst, sizeof(struct sockaddr_in));
513 			ro->ro_dst.sa_family = AF_INET;
514 			ro->ro_dst.sa_len = sizeof(struct sockaddr_in);
515 			((struct sockaddr_in *) &ro->ro_dst)->sin_addr =
516 				sin->sin_addr;
517 			rtalloc(ro);
518 		}
519 		/*
520 		 * If we found a route, use the address
521 		 * corresponding to the outgoing interface
522 		 * unless it is the loopback (in case a route
523 		 * to our address on another net goes to loopback).
524 		 */
525 		if (ro->ro_rt && !(ro->ro_rt->rt_ifp->if_flags & IFF_LOOPBACK)) {
526 			if (jailed) {
527 				if (jailed_ip(cred->cr_prison,
528 				    ro->ro_rt->rt_ifa->ifa_addr)) {
529 					ia = ifatoia(ro->ro_rt->rt_ifa);
530 				}
531 			} else {
532 				ia = ifatoia(ro->ro_rt->rt_ifa);
533 			}
534 		}
535 		if (ia == NULL) {
536 			u_short fport = sin->sin_port;
537 
538 			sin->sin_port = 0;
539 			ia = ifatoia(ifa_ifwithdstaddr(sintosa(sin)));
540 			if (ia && jailed && !jailed_ip(cred->cr_prison,
541 			    sintosa(&ia->ia_addr)))
542 				ia = NULL;
543 			if (ia == NULL)
544 				ia = ifatoia(ifa_ifwithnet(sintosa(sin)));
545 			if (ia && jailed && !jailed_ip(cred->cr_prison,
546 			    sintosa(&ia->ia_addr)))
547 				ia = NULL;
548 			sin->sin_port = fport;
549 			if (ia == NULL)
550 				ia = TAILQ_FIRST(&in_ifaddrhead);
551 			if (ia && jailed && !jailed_ip(cred->cr_prison,
552 			    sintosa(&ia->ia_addr)))
553 				ia = NULL;
554 
555 			if (!jailed && ia == NULL)
556 				return (EADDRNOTAVAIL);
557 		}
558 		/*
559 		 * If the destination address is multicast and an outgoing
560 		 * interface has been set as a multicast option, use the
561 		 * address of that interface as our source address.
562 		 */
563 		if (!jailed && IN_MULTICAST(ntohl(sin->sin_addr.s_addr)) &&
564 		    inp->inp_moptions != NULL) {
565 			struct ip_moptions *imo;
566 			struct ifnet *ifp;
567 
568 			imo = inp->inp_moptions;
569 			if (imo->imo_multicast_ifp != NULL) {
570 				ifp = imo->imo_multicast_ifp;
571 				TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link)
572 					if (ia->ia_ifp == ifp)
573 						break;
574 				if (ia == NULL)
575 					return (EADDRNOTAVAIL);
576 			}
577 		}
578 		/*
579 		 * Don't do pcblookup call here; return interface in plocal_sin
580 		 * and exit to caller, that will do the lookup.
581 		 */
582 		if (jailed) {
583 			jsin.sin_family = AF_INET;
584 			if (!prison_get_nonlocal(cred->cr_prison,
585 			    sintosa(&jsin)) &&
586 			    !prison_get_local(cred->cr_prison, sintosa(&jsin)))
587 					/* IPv6 only Jail */
588 					return (EADDRNOTAVAIL);
589 			*plocal_sin = &jsin;
590 		} else {
591 			*plocal_sin = &ia->ia_addr;
592 		}
593 	}
594 	return (0);
595 }
596 
597 /*
598  * Outer subroutine:
599  * Connect from a socket to a specified address.
600  * Both address and port must be specified in argument sin.
601  * If don't have a local address for this socket yet,
602  * then pick one.
603  */
604 int
605 in_pcbconnect(struct inpcb *inp, struct sockaddr *nam, struct thread *td)
606 {
607 	struct sockaddr_in *if_sin;
608 	struct sockaddr_in *sin = (struct sockaddr_in *)nam;
609 	int error;
610 
611 	/* Call inner routine to assign local interface address. */
612 	if ((error = in_pcbladdr(inp, nam, &if_sin, td)) != 0)
613 		return (error);
614 
615 	if (in_pcblookup_hash(inp->inp_cpcbinfo, sin->sin_addr, sin->sin_port,
616 	    inp->inp_laddr.s_addr ? inp->inp_laddr : if_sin->sin_addr,
617 	    inp->inp_lport, FALSE, NULL) != NULL) {
618 		return (EADDRINUSE);
619 	}
620 	if (inp->inp_laddr.s_addr == INADDR_ANY) {
621 		if (inp->inp_lport == 0) {
622 			error = in_pcbbind(inp, (struct sockaddr *)NULL, td);
623 			if (error)
624 				return (error);
625 		}
626 		inp->inp_laddr = if_sin->sin_addr;
627 	}
628 	inp->inp_faddr = sin->sin_addr;
629 	inp->inp_fport = sin->sin_port;
630 	in_pcbinsconnhash(inp);
631 	return (0);
632 }
633 
634 void
635 in_pcbdisconnect(struct inpcb *inp)
636 {
637 
638 	inp->inp_faddr.s_addr = INADDR_ANY;
639 	inp->inp_fport = 0;
640 	in_pcbremconnhash(inp);
641 	if (inp->inp_socket->so_state & SS_NOFDREF)
642 		in_pcbdetach(inp);
643 }
644 
645 void
646 in_pcbdetach(struct inpcb *inp)
647 {
648 	struct socket *so = inp->inp_socket;
649 	struct inpcbinfo *ipi = inp->inp_pcbinfo;
650 
651 #ifdef IPSEC
652 	ipsec4_delete_pcbpolicy(inp);
653 #endif /*IPSEC*/
654 	inp->inp_gencnt = ++ipi->ipi_gencnt;
655 	in_pcbremlists(inp);
656 	so->so_pcb = 0;
657 	sofree(so);
658 	if (inp->inp_options)
659 		m_free(inp->inp_options);
660 	if (inp->inp_route.ro_rt)
661 		rtfree(inp->inp_route.ro_rt);
662 	ip_freemoptions(inp->inp_moptions);
663 	inp->inp_vflag = 0;
664 	zfree(ipi->ipi_zone, inp);
665 }
666 
667 /*
668  * The calling convention of in_setsockaddr() and in_setpeeraddr() was
669  * modified to match the pru_sockaddr() and pru_peeraddr() entry points
670  * in struct pr_usrreqs, so that protocols can just reference then directly
671  * without the need for a wrapper function.  The socket must have a valid
672  * (i.e., non-nil) PCB, but it should be impossible to get an invalid one
673  * except through a kernel programming error, so it is acceptable to panic
674  * (or in this case trap) if the PCB is invalid.  (Actually, we don't trap
675  * because there actually /is/ a programming error somewhere... XXX)
676  */
677 int
678 in_setsockaddr(struct socket *so, struct sockaddr **nam)
679 {
680 	struct inpcb *inp;
681 	struct sockaddr_in *sin;
682 
683 	/*
684 	 * Do the malloc first in case it blocks.
685 	 */
686 	MALLOC(sin, struct sockaddr_in *, sizeof *sin, M_SONAME,
687 		M_WAITOK | M_ZERO);
688 	sin->sin_family = AF_INET;
689 	sin->sin_len = sizeof *sin;
690 
691 	crit_enter();
692 	inp = so->so_pcb;
693 	if (!inp) {
694 		crit_exit();
695 		kfree(sin, M_SONAME);
696 		return (ECONNRESET);
697 	}
698 	sin->sin_port = inp->inp_lport;
699 	sin->sin_addr = inp->inp_laddr;
700 	crit_exit();
701 
702 	*nam = (struct sockaddr *)sin;
703 	return (0);
704 }
705 
706 int
707 in_setpeeraddr(struct socket *so, struct sockaddr **nam)
708 {
709 	struct inpcb *inp;
710 	struct sockaddr_in *sin;
711 
712 	/*
713 	 * Do the malloc first in case it blocks.
714 	 */
715 	MALLOC(sin, struct sockaddr_in *, sizeof *sin, M_SONAME,
716 		M_WAITOK | M_ZERO);
717 	sin->sin_family = AF_INET;
718 	sin->sin_len = sizeof *sin;
719 
720 	crit_enter();
721 	inp = so->so_pcb;
722 	if (!inp) {
723 		crit_exit();
724 		kfree(sin, M_SONAME);
725 		return (ECONNRESET);
726 	}
727 	sin->sin_port = inp->inp_fport;
728 	sin->sin_addr = inp->inp_faddr;
729 	crit_exit();
730 
731 	*nam = (struct sockaddr *)sin;
732 	return (0);
733 }
734 
735 void
736 in_pcbnotifyall(struct inpcbhead *head, struct in_addr faddr, int err,
737 		void (*notify)(struct inpcb *, int))
738 {
739 	struct inpcb *inp, *ninp;
740 
741 	/*
742 	 * note: if INP_PLACEMARKER is set we must ignore the rest of
743 	 * the structure and skip it.
744 	 */
745 	crit_enter();
746 	LIST_FOREACH_MUTABLE(inp, head, inp_list, ninp) {
747 		if (inp->inp_flags & INP_PLACEMARKER)
748 			continue;
749 #ifdef INET6
750 		if (!(inp->inp_vflag & INP_IPV4))
751 			continue;
752 #endif
753 		if (inp->inp_faddr.s_addr != faddr.s_addr ||
754 		    inp->inp_socket == NULL)
755 			continue;
756 		(*notify)(inp, err);		/* can remove inp from list! */
757 	}
758 	crit_exit();
759 }
760 
761 void
762 in_pcbpurgeif0(struct inpcb *head, struct ifnet *ifp)
763 {
764 	struct inpcb *inp;
765 	struct ip_moptions *imo;
766 	int i, gap;
767 
768 	for (inp = head; inp != NULL; inp = LIST_NEXT(inp, inp_list)) {
769 		if (inp->inp_flags & INP_PLACEMARKER)
770 			continue;
771 		imo = inp->inp_moptions;
772 		if ((inp->inp_vflag & INP_IPV4) && imo != NULL) {
773 			/*
774 			 * Unselect the outgoing interface if it is being
775 			 * detached.
776 			 */
777 			if (imo->imo_multicast_ifp == ifp)
778 				imo->imo_multicast_ifp = NULL;
779 
780 			/*
781 			 * Drop multicast group membership if we joined
782 			 * through the interface being detached.
783 			 */
784 			for (i = 0, gap = 0; i < imo->imo_num_memberships;
785 			    i++) {
786 				if (imo->imo_membership[i]->inm_ifp == ifp) {
787 					in_delmulti(imo->imo_membership[i]);
788 					gap++;
789 				} else if (gap != 0)
790 					imo->imo_membership[i - gap] =
791 					    imo->imo_membership[i];
792 			}
793 			imo->imo_num_memberships -= gap;
794 		}
795 	}
796 }
797 
798 /*
799  * Check for alternatives when higher level complains
800  * about service problems.  For now, invalidate cached
801  * routing information.  If the route was created dynamically
802  * (by a redirect), time to try a default gateway again.
803  */
804 void
805 in_losing(struct inpcb *inp)
806 {
807 	struct rtentry *rt;
808 	struct rt_addrinfo rtinfo;
809 
810 	if ((rt = inp->inp_route.ro_rt)) {
811 		bzero(&rtinfo, sizeof(struct rt_addrinfo));
812 		rtinfo.rti_info[RTAX_DST] = rt_key(rt);
813 		rtinfo.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
814 		rtinfo.rti_info[RTAX_NETMASK] = rt_mask(rt);
815 		rtinfo.rti_flags = rt->rt_flags;
816 		rt_missmsg(RTM_LOSING, &rtinfo, rt->rt_flags, 0);
817 		if (rt->rt_flags & RTF_DYNAMIC)
818 			rtrequest1_global(RTM_DELETE, &rtinfo, NULL, NULL);
819 		inp->inp_route.ro_rt = NULL;
820 		rtfree(rt);
821 		/*
822 		 * A new route can be allocated
823 		 * the next time output is attempted.
824 		 */
825 	}
826 }
827 
828 /*
829  * After a routing change, flush old routing
830  * and allocate a (hopefully) better one.
831  */
832 void
833 in_rtchange(struct inpcb *inp, int err)
834 {
835 	if (inp->inp_route.ro_rt) {
836 		rtfree(inp->inp_route.ro_rt);
837 		inp->inp_route.ro_rt = NULL;
838 		/*
839 		 * A new route can be allocated the next time
840 		 * output is attempted.
841 		 */
842 	}
843 }
844 
845 /*
846  * Lookup a PCB based on the local address and port.
847  */
848 struct inpcb *
849 in_pcblookup_local(struct inpcbinfo *pcbinfo, struct in_addr laddr,
850 		   u_int lport_arg, int wild_okay, struct ucred *cred)
851 {
852 	struct inpcb *inp;
853 	int matchwild = 3, wildcard;
854 	u_short lport = lport_arg;
855 
856 	struct inpcbporthead *porthash;
857 	struct inpcbport *phd;
858 	struct inpcb *match = NULL;
859 
860 	/*
861 	 * Best fit PCB lookup.
862 	 *
863 	 * First see if this local port is in use by looking on the
864 	 * port hash list.
865 	 */
866 	porthash = &pcbinfo->porthashbase[INP_PCBPORTHASH(lport,
867 	    pcbinfo->porthashmask)];
868 	LIST_FOREACH(phd, porthash, phd_hash) {
869 		if (phd->phd_port == lport)
870 			break;
871 	}
872 	if (phd != NULL) {
873 		/*
874 		 * Port is in use by one or more PCBs. Look for best
875 		 * fit.
876 		 */
877 		LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) {
878 			wildcard = 0;
879 #ifdef INET6
880 			if ((inp->inp_vflag & INP_IPV4) == 0)
881 				continue;
882 #endif
883 			if (inp->inp_faddr.s_addr != INADDR_ANY)
884 				wildcard++;
885 			if (inp->inp_laddr.s_addr != INADDR_ANY) {
886 				if (laddr.s_addr == INADDR_ANY)
887 					wildcard++;
888 				else if (inp->inp_laddr.s_addr != laddr.s_addr)
889 					continue;
890 			} else {
891 				if (laddr.s_addr != INADDR_ANY)
892 					wildcard++;
893 			}
894 			if (wildcard && !wild_okay)
895 				continue;
896 			if (wildcard < matchwild &&
897 			    (cred == NULL ||
898 			     cred->cr_prison ==
899 					inp->inp_socket->so_cred->cr_prison)) {
900 				match = inp;
901 				matchwild = wildcard;
902 				if (matchwild == 0) {
903 					break;
904 				}
905 			}
906 		}
907 	}
908 	return (match);
909 }
910 
911 /*
912  * Lookup PCB in hash list.
913  */
914 struct inpcb *
915 in_pcblookup_hash(struct inpcbinfo *pcbinfo, struct in_addr faddr, u_int fport_arg,
916 		  struct in_addr laddr, u_int lport_arg, boolean_t wildcard,
917 		  struct ifnet *ifp)
918 {
919 	struct inpcbhead *head;
920 	struct inpcb *inp, *jinp=NULL;
921 	u_short fport = fport_arg, lport = lport_arg;
922 
923 	/*
924 	 * First look for an exact match.
925 	 */
926 	head = &pcbinfo->hashbase[INP_PCBCONNHASH(faddr.s_addr, fport,
927 	    laddr.s_addr, lport, pcbinfo->hashmask)];
928 	LIST_FOREACH(inp, head, inp_hash) {
929 #ifdef INET6
930 		if (!(inp->inp_vflag & INP_IPV4))
931 			continue;
932 #endif
933 		if (in_hosteq(inp->inp_faddr, faddr) &&
934 		    in_hosteq(inp->inp_laddr, laddr) &&
935 		    inp->inp_fport == fport && inp->inp_lport == lport) {
936 			/* found */
937 			if (inp->inp_socket == NULL ||
938 			    inp->inp_socket->so_cred->cr_prison == NULL) {
939 				return (inp);
940 			} else {
941 				if  (jinp == NULL)
942 					jinp = inp;
943 			}
944 		}
945 	}
946 	if (jinp != NULL)
947 		return (jinp);
948 	if (wildcard) {
949 		struct inpcb *local_wild = NULL;
950 		struct inpcb *jinp_wild = NULL;
951 #ifdef INET6
952 		struct inpcb *local_wild_mapped = NULL;
953 #endif
954 		struct inpcontainer *ic;
955 		struct inpcontainerhead *chead;
956 		struct sockaddr_in jsin;
957 		struct ucred *cred;
958 
959 		/*
960 		 * Order of socket selection:
961 		 * 1. non-jailed, non-wild.
962 		 * 2. non-jailed, wild.
963 		 * 3. jailed, non-wild.
964 		 * 4. jailed, wild.
965 		 */
966 		jsin.sin_family = AF_INET;
967 		chead = &pcbinfo->wildcardhashbase[
968 		    INP_PCBWILDCARDHASH(lport, pcbinfo->wildcardhashmask)];
969 		LIST_FOREACH(ic, chead, ic_list) {
970 			inp = ic->ic_inp;
971 			jsin.sin_addr.s_addr = laddr.s_addr;
972 #ifdef INET6
973 			if (!(inp->inp_vflag & INP_IPV4))
974 				continue;
975 #endif
976 			if (inp->inp_socket != NULL)
977 				cred = inp->inp_socket->so_cred;
978 			else
979 				cred = NULL;
980 			if (cred != NULL && jailed(cred)) {
981 				if (jinp != NULL)
982 					continue;
983 				else
984 					if (!jailed_ip(cred->cr_prison,
985 					    (struct sockaddr *)&jsin))
986 						continue;
987 			}
988 			if (inp->inp_lport == lport) {
989 				if (ifp && ifp->if_type == IFT_FAITH &&
990 				    !(inp->inp_flags & INP_FAITH))
991 					continue;
992 				if (inp->inp_laddr.s_addr == laddr.s_addr) {
993 					if (cred != NULL && jailed(cred))
994 						jinp = inp;
995 					else
996 						return (inp);
997 				}
998 				if (inp->inp_laddr.s_addr == INADDR_ANY) {
999 #ifdef INET6
1000 					if (INP_CHECK_SOCKAF(inp->inp_socket,
1001 							     AF_INET6))
1002 						local_wild_mapped = inp;
1003 					else
1004 #endif
1005 						if (cred != NULL &&
1006 						    jailed(cred))
1007 							jinp_wild = inp;
1008 						else
1009 							local_wild = inp;
1010 				}
1011 			}
1012 		}
1013 		if (local_wild != NULL)
1014 			return (local_wild);
1015 #ifdef INET6
1016 		if (local_wild_mapped != NULL)
1017 			return (local_wild_mapped);
1018 #endif
1019 		if (jinp != NULL)
1020 			return (jinp);
1021 		return (jinp_wild);
1022 	}
1023 
1024 	/*
1025 	 * Not found.
1026 	 */
1027 	return (NULL);
1028 }
1029 
1030 /*
1031  * Insert PCB into connection hash table.
1032  */
1033 void
1034 in_pcbinsconnhash(struct inpcb *inp)
1035 {
1036 	struct inpcbinfo *pcbinfo = inp->inp_cpcbinfo;
1037 	struct inpcbhead *bucket;
1038 	u_int32_t hashkey_faddr, hashkey_laddr;
1039 
1040 #ifdef INET6
1041 	if (inp->inp_vflag & INP_IPV6) {
1042 		hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX JH */;
1043 		hashkey_laddr = inp->in6p_laddr.s6_addr32[3] /* XXX JH */;
1044 	} else {
1045 #endif
1046 		hashkey_faddr = inp->inp_faddr.s_addr;
1047 		hashkey_laddr = inp->inp_laddr.s_addr;
1048 #ifdef INET6
1049 	}
1050 #endif
1051 
1052 	KASSERT(!(inp->inp_flags & INP_CONNECTED), ("already on hash list"));
1053 	inp->inp_flags |= INP_CONNECTED;
1054 
1055 	/*
1056 	 * Insert into the connection hash table.
1057 	 */
1058 	bucket = &pcbinfo->hashbase[INP_PCBCONNHASH(hashkey_faddr,
1059 	    inp->inp_fport, hashkey_laddr, inp->inp_lport, pcbinfo->hashmask)];
1060 	LIST_INSERT_HEAD(bucket, inp, inp_hash);
1061 }
1062 
1063 /*
1064  * Remove PCB from connection hash table.
1065  */
1066 void
1067 in_pcbremconnhash(struct inpcb *inp)
1068 {
1069 	KASSERT(inp->inp_flags & INP_CONNECTED, ("inp not connected"));
1070 	LIST_REMOVE(inp, inp_hash);
1071 	inp->inp_flags &= ~INP_CONNECTED;
1072 }
1073 
1074 /*
1075  * Insert PCB into port hash table.
1076  */
1077 int
1078 in_pcbinsporthash(struct inpcb *inp)
1079 {
1080 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1081 	struct inpcbporthead *pcbporthash;
1082 	struct inpcbport *phd;
1083 
1084 	/*
1085 	 * Insert into the port hash table.
1086 	 */
1087 	pcbporthash = &pcbinfo->porthashbase[
1088 	    INP_PCBPORTHASH(inp->inp_lport, pcbinfo->porthashmask)];
1089 
1090 	/* Go through port list and look for a head for this lport. */
1091 	LIST_FOREACH(phd, pcbporthash, phd_hash)
1092 		if (phd->phd_port == inp->inp_lport)
1093 			break;
1094 
1095 	/* If none exists, malloc one and tack it on. */
1096 	if (phd == NULL) {
1097 		MALLOC(phd, struct inpcbport *, sizeof(struct inpcbport),
1098 		    M_PCB, M_INTWAIT | M_NULLOK);
1099 		if (phd == NULL)
1100 			return (ENOBUFS); /* XXX */
1101 		phd->phd_port = inp->inp_lport;
1102 		LIST_INIT(&phd->phd_pcblist);
1103 		LIST_INSERT_HEAD(pcbporthash, phd, phd_hash);
1104 	}
1105 
1106 	inp->inp_phd = phd;
1107 	LIST_INSERT_HEAD(&phd->phd_pcblist, inp, inp_portlist);
1108 
1109 	return (0);
1110 }
1111 
1112 void
1113 in_pcbinswildcardhash_oncpu(struct inpcb *inp, struct inpcbinfo *pcbinfo)
1114 {
1115 	struct inpcontainer *ic;
1116 	struct inpcontainerhead *bucket;
1117 
1118 	bucket = &pcbinfo->wildcardhashbase[
1119 	    INP_PCBWILDCARDHASH(inp->inp_lport, pcbinfo->wildcardhashmask)];
1120 
1121 	ic = kmalloc(sizeof(struct inpcontainer), M_TEMP, M_INTWAIT);
1122 	ic->ic_inp = inp;
1123 	LIST_INSERT_HEAD(bucket, ic, ic_list);
1124 }
1125 
1126 /*
1127  * Insert PCB into wildcard hash table.
1128  */
1129 void
1130 in_pcbinswildcardhash(struct inpcb *inp)
1131 {
1132 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1133 
1134 	KKASSERT(pcbinfo != NULL);
1135 
1136 	in_pcbinswildcardhash_oncpu(inp, pcbinfo);
1137 	inp->inp_flags |= INP_WILDCARD;
1138 }
1139 
1140 void
1141 in_pcbremwildcardhash_oncpu(struct inpcb *inp, struct inpcbinfo *pcbinfo)
1142 {
1143 	struct inpcontainer *ic;
1144 	struct inpcontainerhead *head;
1145 
1146 	/* find bucket */
1147 	head = &pcbinfo->wildcardhashbase[
1148 	    INP_PCBWILDCARDHASH(inp->inp_lport, pcbinfo->wildcardhashmask)];
1149 
1150 	LIST_FOREACH(ic, head, ic_list) {
1151 		if (ic->ic_inp == inp)
1152 			goto found;
1153 	}
1154 	return;			/* not found! */
1155 
1156 found:
1157 	LIST_REMOVE(ic, ic_list);	/* remove container from bucket chain */
1158 	kfree(ic, M_TEMP);		/* deallocate container */
1159 }
1160 
1161 /*
1162  * Remove PCB from wildcard hash table.
1163  */
1164 void
1165 in_pcbremwildcardhash(struct inpcb *inp)
1166 {
1167 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1168 
1169 	KASSERT(inp->inp_flags & INP_WILDCARD, ("inp not wildcard"));
1170 	in_pcbremwildcardhash_oncpu(inp, pcbinfo);
1171 	inp->inp_flags &= ~INP_WILDCARD;
1172 }
1173 
1174 /*
1175  * Remove PCB from various lists.
1176  */
1177 void
1178 in_pcbremlists(struct inpcb *inp)
1179 {
1180 	if (inp->inp_lport) {
1181 		struct inpcbport *phd = inp->inp_phd;
1182 
1183 		LIST_REMOVE(inp, inp_portlist);
1184 		if (LIST_FIRST(&phd->phd_pcblist) == NULL) {
1185 			LIST_REMOVE(phd, phd_hash);
1186 			kfree(phd, M_PCB);
1187 		}
1188 	}
1189 	if (inp->inp_flags & INP_WILDCARD) {
1190 		in_pcbremwildcardhash(inp);
1191 	} else if (inp->inp_flags & INP_CONNECTED) {
1192 		in_pcbremconnhash(inp);
1193 	}
1194 	LIST_REMOVE(inp, inp_list);
1195 	inp->inp_pcbinfo->ipi_count--;
1196 }
1197 
1198 int
1199 prison_xinpcb(struct thread *td, struct inpcb *inp)
1200 {
1201 	struct ucred *cr;
1202 
1203 	if (td->td_proc == NULL)
1204 		return (0);
1205 	cr = td->td_proc->p_ucred;
1206 	if (cr->cr_prison == NULL)
1207 		return (0);
1208 	if (inp->inp_socket && inp->inp_socket->so_cred &&
1209 	    inp->inp_socket->so_cred->cr_prison &&
1210 	    cr->cr_prison == inp->inp_socket->so_cred->cr_prison)
1211 		return (0);
1212 	return (1);
1213 }
1214 
1215 int
1216 in_pcblist_global(SYSCTL_HANDLER_ARGS)
1217 {
1218 	struct inpcbinfo *pcbinfo = arg1;
1219 	struct inpcb *inp, *marker;
1220 	struct xinpcb xi;
1221 	int error, i, n;
1222 	inp_gen_t gencnt;
1223 
1224 	/*
1225 	 * The process of preparing the TCB list is too time-consuming and
1226 	 * resource-intensive to repeat twice on every request.
1227 	 */
1228 	if (req->oldptr == NULL) {
1229 		n = pcbinfo->ipi_count;
1230 		req->oldidx = (n + n/8 + 10) * sizeof(struct xinpcb);
1231 		return 0;
1232 	}
1233 
1234 	if (req->newptr != NULL)
1235 		return EPERM;
1236 
1237 	/*
1238 	 * OK, now we're committed to doing something.  Re-fetch ipi_count
1239 	 * after obtaining the generation count.
1240 	 */
1241 	gencnt = pcbinfo->ipi_gencnt;
1242 	n = pcbinfo->ipi_count;
1243 
1244 	marker = kmalloc(sizeof(struct inpcb), M_TEMP, M_WAITOK|M_ZERO);
1245 	marker->inp_flags |= INP_PLACEMARKER;
1246 	LIST_INSERT_HEAD(&pcbinfo->pcblisthead, marker, inp_list);
1247 
1248 	i = 0;
1249 	error = 0;
1250 
1251 	while ((inp = LIST_NEXT(marker, inp_list)) != NULL && i < n) {
1252 		LIST_REMOVE(marker, inp_list);
1253 		LIST_INSERT_AFTER(inp, marker, inp_list);
1254 
1255 		if (inp->inp_flags & INP_PLACEMARKER)
1256 			continue;
1257 		if (inp->inp_gencnt > gencnt)
1258 			continue;
1259 		if (prison_xinpcb(req->td, inp))
1260 			continue;
1261 		bzero(&xi, sizeof xi);
1262 		xi.xi_len = sizeof xi;
1263 		bcopy(inp, &xi.xi_inp, sizeof *inp);
1264 		if (inp->inp_socket)
1265 			sotoxsocket(inp->inp_socket, &xi.xi_socket);
1266 		if ((error = SYSCTL_OUT(req, &xi, sizeof xi)) != 0)
1267 			break;
1268 		++i;
1269 	}
1270 	LIST_REMOVE(marker, inp_list);
1271 	if (error == 0 && i < n) {
1272 		bzero(&xi, sizeof xi);
1273 		xi.xi_len = sizeof xi;
1274 		while (i < n) {
1275 			error = SYSCTL_OUT(req, &xi, sizeof xi);
1276 			++i;
1277 		}
1278 	}
1279 	kfree(marker, M_TEMP);
1280 	return(error);
1281 }
1282