xref: /dragonfly/sys/netinet/in_pcb.c (revision e2f5ccfb)
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 1.46 2008/06/08 08:38:05 sephe Exp $
69  */
70 
71 #include "opt_ipsec.h"
72 #include "opt_inet6.h"
73 
74 #include <sys/param.h>
75 #include <sys/systm.h>
76 #include <sys/malloc.h>
77 #include <sys/mbuf.h>
78 #include <sys/domain.h>
79 #include <sys/protosw.h>
80 #include <sys/socket.h>
81 #include <sys/socketvar.h>
82 #include <sys/proc.h>
83 #include <sys/jail.h>
84 #include <sys/kernel.h>
85 #include <sys/sysctl.h>
86 #include <sys/thread2.h>
87 
88 #include <machine/limits.h>
89 
90 #include <vm/vm_zone.h>
91 
92 #include <net/if.h>
93 #include <net/if_types.h>
94 #include <net/route.h>
95 
96 #include <netinet/in.h>
97 #include <netinet/in_pcb.h>
98 #include <netinet/in_var.h>
99 #include <netinet/ip_var.h>
100 #ifdef INET6
101 #include <netinet/ip6.h>
102 #include <netinet6/ip6_var.h>
103 #endif /* INET6 */
104 
105 #ifdef IPSEC
106 #include <netinet6/ipsec.h>
107 #include <netproto/key/key.h>
108 #endif
109 
110 #ifdef FAST_IPSEC
111 #if defined(IPSEC) || defined(IPSEC_ESP)
112 #error "Bad idea: don't compile with both IPSEC and FAST_IPSEC!"
113 #endif
114 
115 #include <netproto/ipsec/ipsec.h>
116 #include <netproto/ipsec/key.h>
117 #define	IPSEC
118 #endif /* FAST_IPSEC */
119 
120 struct in_addr zeroin_addr;
121 
122 /*
123  * These configure the range of local port addresses assigned to
124  * "unspecified" outgoing connections/packets/whatever.
125  */
126 int ipport_lowfirstauto = IPPORT_RESERVED - 1;	/* 1023 */
127 int ipport_lowlastauto = IPPORT_RESERVEDSTART;	/* 600 */
128 
129 int ipport_firstauto = IPPORT_RESERVED;		/* 1024 */
130 int ipport_lastauto = IPPORT_USERRESERVED;	/* 5000 */
131 
132 int ipport_hifirstauto = IPPORT_HIFIRSTAUTO;	/* 49152 */
133 int ipport_hilastauto = IPPORT_HILASTAUTO;	/* 65535 */
134 
135 static __inline void
136 RANGECHK(int var, int min, int max)
137 {
138 	if (var < min)
139 		var = min;
140 	else if (var > max)
141 		var = max;
142 }
143 
144 static int
145 sysctl_net_ipport_check(SYSCTL_HANDLER_ARGS)
146 {
147 	int error;
148 
149 	error = sysctl_handle_int(oidp, oidp->oid_arg1, oidp->oid_arg2, req);
150 	if (!error) {
151 		RANGECHK(ipport_lowfirstauto, 1, IPPORT_RESERVED - 1);
152 		RANGECHK(ipport_lowlastauto, 1, IPPORT_RESERVED - 1);
153 
154 		RANGECHK(ipport_firstauto, IPPORT_RESERVED, USHRT_MAX);
155 		RANGECHK(ipport_lastauto, IPPORT_RESERVED, USHRT_MAX);
156 
157 		RANGECHK(ipport_hifirstauto, IPPORT_RESERVED, USHRT_MAX);
158 		RANGECHK(ipport_hilastauto, IPPORT_RESERVED, USHRT_MAX);
159 	}
160 	return (error);
161 }
162 
163 SYSCTL_NODE(_net_inet_ip, IPPROTO_IP, portrange, CTLFLAG_RW, 0, "IP Ports");
164 
165 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, lowfirst, CTLTYPE_INT|CTLFLAG_RW,
166 	   &ipport_lowfirstauto, 0, &sysctl_net_ipport_check, "I", "");
167 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, lowlast, CTLTYPE_INT|CTLFLAG_RW,
168 	   &ipport_lowlastauto, 0, &sysctl_net_ipport_check, "I", "");
169 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, first, CTLTYPE_INT|CTLFLAG_RW,
170 	   &ipport_firstauto, 0, &sysctl_net_ipport_check, "I", "");
171 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, last, CTLTYPE_INT|CTLFLAG_RW,
172 	   &ipport_lastauto, 0, &sysctl_net_ipport_check, "I", "");
173 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, hifirst, CTLTYPE_INT|CTLFLAG_RW,
174 	   &ipport_hifirstauto, 0, &sysctl_net_ipport_check, "I", "");
175 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, hilast, CTLTYPE_INT|CTLFLAG_RW,
176 	   &ipport_hilastauto, 0, &sysctl_net_ipport_check, "I", "");
177 
178 /*
179  * in_pcb.c: manage the Protocol Control Blocks.
180  *
181  * NOTE: It is assumed that most of these functions will be called from
182  * a critical section.  XXX - There are, unfortunately, a few exceptions
183  * to this rule that should be fixed.
184  *
185  * NOTE: The caller should initialize the cpu field to the cpu running the
186  * protocol stack associated with this inpcbinfo.
187  */
188 
189 void
190 in_pcbinfo_init(struct inpcbinfo *pcbinfo)
191 {
192 	LIST_INIT(&pcbinfo->pcblisthead);
193 	pcbinfo->cpu = -1;
194 }
195 
196 /*
197  * Allocate a PCB and associate it with the socket.
198  */
199 int
200 in_pcballoc(struct socket *so, struct inpcbinfo *pcbinfo)
201 {
202 	struct inpcb *inp;
203 #ifdef IPSEC
204 	int error;
205 #endif
206 
207 	inp = zalloc(pcbinfo->ipi_zone);
208 	if (inp == NULL)
209 		return (ENOBUFS);
210 	bzero(inp, sizeof *inp);
211 	inp->inp_gencnt = ++pcbinfo->ipi_gencnt;
212 	inp->inp_pcbinfo = inp->inp_cpcbinfo = pcbinfo;
213 	inp->inp_socket = so;
214 #ifdef IPSEC
215 	error = ipsec_init_policy(so, &inp->inp_sp);
216 	if (error != 0) {
217 		zfree(pcbinfo->ipi_zone, inp);
218 		return (error);
219 	}
220 #endif
221 #ifdef INET6
222 	if (INP_SOCKAF(so) == AF_INET6 && ip6_v6only)
223 		inp->inp_flags |= IN6P_IPV6_V6ONLY;
224 	if (ip6_auto_flowlabel)
225 		inp->inp_flags |= IN6P_AUTOFLOWLABEL;
226 #endif
227 	so->so_pcb = inp;
228 	LIST_INSERT_HEAD(&pcbinfo->pcblisthead, inp, inp_list);
229 	pcbinfo->ipi_count++;
230 	return (0);
231 }
232 
233 int
234 in_pcbbind(struct inpcb *inp, struct sockaddr *nam, struct thread *td)
235 {
236 	struct socket *so = inp->inp_socket;
237 	struct proc *p = td->td_proc;
238 	unsigned short *lastport;
239 	struct sockaddr_in *sin;
240 	struct sockaddr_in jsin;
241 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
242 	struct ucred *cred = NULL;
243 	u_short lport = 0;
244 	int wild = 0, reuseport = (so->so_options & SO_REUSEPORT);
245 	int error;
246 
247 	KKASSERT(p);
248 
249 	if (TAILQ_EMPTY(&in_ifaddrheads[mycpuid])) /* XXX broken! */
250 		return (EADDRNOTAVAIL);
251 	if (inp->inp_lport != 0 || inp->inp_laddr.s_addr != INADDR_ANY)
252 		return (EINVAL);	/* already bound */
253 	if (!(so->so_options & (SO_REUSEADDR|SO_REUSEPORT)))
254 		wild = 1;    /* neither SO_REUSEADDR nor SO_REUSEPORT is set */
255 	if (p)
256 		cred = p->p_ucred;
257 	if (nam != NULL) {
258 		sin = (struct sockaddr_in *)nam;
259 		if (nam->sa_len != sizeof *sin)
260 			return (EINVAL);
261 #ifdef notdef
262 		/*
263 		 * We should check the family, but old programs
264 		 * incorrectly fail to initialize it.
265 		 */
266 		if (sin->sin_family != AF_INET)
267 			return (EAFNOSUPPORT);
268 #endif
269 		if (!prison_replace_wildcards(td, nam))
270 				return (EINVAL);
271 		lport = sin->sin_port;
272 		if (IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) {
273 			/*
274 			 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
275 			 * allow complete duplication of binding if
276 			 * SO_REUSEPORT is set, or if SO_REUSEADDR is set
277 			 * and a multicast address is bound on both
278 			 * new and duplicated sockets.
279 			 */
280 			if (so->so_options & SO_REUSEADDR)
281 				reuseport = SO_REUSEADDR | SO_REUSEPORT;
282 		} else if (sin->sin_addr.s_addr != INADDR_ANY) {
283 			sin->sin_port = 0;		/* yech... */
284 			bzero(&sin->sin_zero, sizeof sin->sin_zero);
285 			if (ifa_ifwithaddr((struct sockaddr *)sin) == NULL)
286 				return (EADDRNOTAVAIL);
287 		}
288 		if (lport != 0) {
289 			struct inpcb *t;
290 
291 			/* GROSS */
292 			if (ntohs(lport) < IPPORT_RESERVED &&
293 			    cred && suser_cred(cred, PRISON_ROOT))
294 				return (EACCES);
295 			if (so->so_cred->cr_uid != 0 &&
296 			    !IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) {
297 				t = in_pcblookup_local(inp->inp_pcbinfo,
298 				    sin->sin_addr, lport,
299 				    INPLOOKUP_WILDCARD, cred);
300 				if (t &&
301 				    (!in_nullhost(sin->sin_addr) ||
302 				     !in_nullhost(t->inp_laddr) ||
303 				     (t->inp_socket->so_options &
304 					 SO_REUSEPORT) == 0) &&
305 				    (so->so_cred->cr_uid !=
306 				     t->inp_socket->so_cred->cr_uid)) {
307 #ifdef INET6
308 					if (!in_nullhost(sin->sin_addr) ||
309 					    !in_nullhost(t->inp_laddr) ||
310 					    INP_SOCKAF(so) ==
311 					    INP_SOCKAF(t->inp_socket))
312 #endif
313 					return (EADDRINUSE);
314 				}
315 			}
316 			if (cred && !prison_replace_wildcards(td, nam))
317 				return (EADDRNOTAVAIL);
318 			t = in_pcblookup_local(pcbinfo, sin->sin_addr, lport,
319 					       wild, cred);
320 			if (t && !(reuseport & t->inp_socket->so_options)) {
321 #ifdef INET6
322 				if (!in_nullhost(sin->sin_addr) ||
323 				    !in_nullhost(t->inp_laddr) ||
324 				    INP_SOCKAF(so) == INP_SOCKAF(t->inp_socket))
325 #endif
326 				return (EADDRINUSE);
327 			}
328 		}
329 		inp->inp_laddr = sin->sin_addr;
330 	}
331 	if (lport == 0) {
332 		ushort first, last;
333 		int count;
334 
335 		jsin.sin_family = AF_INET;
336 		jsin.sin_addr.s_addr = inp->inp_laddr.s_addr;
337 		if (!prison_replace_wildcards(td, (struct sockaddr *)&jsin)) {
338 			inp->inp_laddr.s_addr = INADDR_ANY;
339 			return (EINVAL);
340 		}
341 		inp->inp_laddr.s_addr = jsin.sin_addr.s_addr;
342 
343 		inp->inp_flags |= INP_ANONPORT;
344 
345 		if (inp->inp_flags & INP_HIGHPORT) {
346 			first = ipport_hifirstauto;	/* sysctl */
347 			last  = ipport_hilastauto;
348 			lastport = &pcbinfo->lasthi;
349 		} else if (inp->inp_flags & INP_LOWPORT) {
350 			if (cred &&
351 			    (error = suser_cred(cred, PRISON_ROOT))) {
352 				inp->inp_laddr.s_addr = INADDR_ANY;
353 				return (error);
354 			}
355 			first = ipport_lowfirstauto;	/* 1023 */
356 			last  = ipport_lowlastauto;	/* 600 */
357 			lastport = &pcbinfo->lastlow;
358 		} else {
359 			first = ipport_firstauto;	/* sysctl */
360 			last  = ipport_lastauto;
361 			lastport = &pcbinfo->lastport;
362 		}
363 		/*
364 		 * Simple check to ensure all ports are not used up causing
365 		 * a deadlock here.
366 		 *
367 		 * We split the two cases (up and down) so that the direction
368 		 * is not being tested on each round of the loop.
369 		 */
370 		if (first > last) {
371 			/*
372 			 * counting down
373 			 */
374 			count = first - last;
375 
376 			do {
377 				if (count-- < 0) {	/* completely used? */
378 					inp->inp_laddr.s_addr = INADDR_ANY;
379 					return (EADDRNOTAVAIL);
380 				}
381 				--*lastport;
382 				if (*lastport > first || *lastport < last)
383 					*lastport = first;
384 				lport = htons(*lastport);
385 			} while (in_pcblookup_local(pcbinfo, inp->inp_laddr,
386 				 lport, wild, cred));
387 		} else {
388 			/*
389 			 * counting up
390 			 */
391 			count = last - first;
392 
393 			do {
394 				if (count-- < 0) {	/* completely used? */
395 					inp->inp_laddr.s_addr = INADDR_ANY;
396 					return (EADDRNOTAVAIL);
397 				}
398 				++*lastport;
399 				if (*lastport < first || *lastport > last)
400 					*lastport = first;
401 				lport = htons(*lastport);
402 			} while (in_pcblookup_local(pcbinfo, inp->inp_laddr,
403 				 lport, wild, cred));
404 		}
405 	}
406 	inp->inp_lport = lport;
407 
408 	jsin.sin_family = AF_INET;
409 	jsin.sin_addr.s_addr = inp->inp_laddr.s_addr;
410 	if (!prison_replace_wildcards(td, (struct sockaddr*)&jsin)) {
411 		inp->inp_laddr.s_addr = INADDR_ANY;
412 		inp->inp_lport = 0;
413 		return (EINVAL);
414 	}
415 	inp->inp_laddr.s_addr = jsin.sin_addr.s_addr;
416 
417 	if (in_pcbinsporthash(inp) != 0) {
418 		inp->inp_laddr.s_addr = INADDR_ANY;
419 		inp->inp_lport = 0;
420 		return (EAGAIN);
421 	}
422 	return (0);
423 }
424 
425 /*
426  *   Transform old in_pcbconnect() into an inner subroutine for new
427  *   in_pcbconnect(): Do some validity-checking on the remote
428  *   address (in mbuf 'nam') and then determine local host address
429  *   (i.e., which interface) to use to access that remote host.
430  *
431  *   This preserves definition of in_pcbconnect(), while supporting a
432  *   slightly different version for T/TCP.  (This is more than
433  *   a bit of a kludge, but cleaning up the internal interfaces would
434  *   have forced minor changes in every protocol).
435  */
436 int
437 in_pcbladdr(struct inpcb *inp, struct sockaddr *nam,
438 	struct sockaddr_in **plocal_sin, struct thread *td)
439 {
440 	struct in_ifaddr *ia;
441 	struct ucred *cred = NULL;
442 	struct sockaddr_in *sin = (struct sockaddr_in *)nam;
443 	struct sockaddr *jsin;
444 	int jailed = 0, alloc_route = 0;
445 
446 	if (nam->sa_len != sizeof *sin)
447 		return (EINVAL);
448 	if (sin->sin_family != AF_INET)
449 		return (EAFNOSUPPORT);
450 	if (sin->sin_port == 0)
451 		return (EADDRNOTAVAIL);
452 	if (td && td->td_proc && td->td_proc->p_ucred)
453 		cred = td->td_proc->p_ucred;
454 	if (cred && cred->cr_prison)
455 		jailed = 1;
456 	if (!TAILQ_EMPTY(&in_ifaddrheads[mycpuid])) {
457 		ia = TAILQ_FIRST(&in_ifaddrheads[mycpuid])->ia;
458 		/*
459 		 * If the destination address is INADDR_ANY,
460 		 * use the primary local address.
461 		 * If the supplied address is INADDR_BROADCAST,
462 		 * and the primary interface supports broadcast,
463 		 * choose the broadcast address for that interface.
464 		 */
465 		if (sin->sin_addr.s_addr == INADDR_ANY)
466 			sin->sin_addr = IA_SIN(ia)->sin_addr;
467 		else if (sin->sin_addr.s_addr == (u_long)INADDR_BROADCAST &&
468 		    (ia->ia_ifp->if_flags & IFF_BROADCAST))
469 			sin->sin_addr = satosin(&ia->ia_broadaddr)->sin_addr;
470 	}
471 	if (inp->inp_laddr.s_addr == INADDR_ANY) {
472 		struct route *ro;
473 
474 		ia = (struct in_ifaddr *)NULL;
475 		/*
476 		 * If route is known or can be allocated now,
477 		 * our src addr is taken from the i/f, else punt.
478 		 * Note that we should check the address family of the cached
479 		 * destination, in case of sharing the cache with IPv6.
480 		 */
481 		ro = &inp->inp_route;
482 		if (ro->ro_rt &&
483 		    (!(ro->ro_rt->rt_flags & RTF_UP) ||
484 		     ro->ro_dst.sa_family != AF_INET ||
485 		     satosin(&ro->ro_dst)->sin_addr.s_addr !=
486 				      sin->sin_addr.s_addr ||
487 		     inp->inp_socket->so_options & SO_DONTROUTE)) {
488 			RTFREE(ro->ro_rt);
489 			ro->ro_rt = (struct rtentry *)NULL;
490 		}
491 		if (!(inp->inp_socket->so_options & SO_DONTROUTE) && /*XXX*/
492 		    (ro->ro_rt == (struct rtentry *)NULL ||
493 		    ro->ro_rt->rt_ifp == (struct ifnet *)NULL)) {
494 			/* No route yet, so try to acquire one */
495 			bzero(&ro->ro_dst, sizeof(struct sockaddr_in));
496 			ro->ro_dst.sa_family = AF_INET;
497 			ro->ro_dst.sa_len = sizeof(struct sockaddr_in);
498 			((struct sockaddr_in *) &ro->ro_dst)->sin_addr =
499 				sin->sin_addr;
500 			rtalloc(ro);
501 			alloc_route = 1;
502 		}
503 		/*
504 		 * If we found a route, use the address
505 		 * corresponding to the outgoing interface
506 		 * unless it is the loopback (in case a route
507 		 * to our address on another net goes to loopback).
508 		 */
509 		if (ro->ro_rt && !(ro->ro_rt->rt_ifp->if_flags & IFF_LOOPBACK)) {
510 			if (jailed) {
511 				if (jailed_ip(cred->cr_prison,
512 				    ro->ro_rt->rt_ifa->ifa_addr)) {
513 					ia = ifatoia(ro->ro_rt->rt_ifa);
514 				}
515 			} else {
516 				ia = ifatoia(ro->ro_rt->rt_ifa);
517 			}
518 		}
519 		if (ia == NULL) {
520 			u_short fport = sin->sin_port;
521 
522 			sin->sin_port = 0;
523 			ia = ifatoia(ifa_ifwithdstaddr(sintosa(sin)));
524 			if (ia && jailed && !jailed_ip(cred->cr_prison,
525 			    sintosa(&ia->ia_addr)))
526 				ia = NULL;
527 			if (ia == NULL)
528 				ia = ifatoia(ifa_ifwithnet(sintosa(sin)));
529 			if (ia && jailed && !jailed_ip(cred->cr_prison,
530 			    sintosa(&ia->ia_addr)))
531 				ia = NULL;
532 			sin->sin_port = fport;
533 			if (ia == NULL &&
534 			    !TAILQ_EMPTY(&in_ifaddrheads[mycpuid]))
535 				ia = TAILQ_FIRST(&in_ifaddrheads[mycpuid])->ia;
536 			if (ia && jailed && !jailed_ip(cred->cr_prison,
537 			    sintosa(&ia->ia_addr)))
538 				ia = NULL;
539 
540 			if (!jailed && ia == NULL)
541 				goto fail;
542 		}
543 		/*
544 		 * If the destination address is multicast and an outgoing
545 		 * interface has been set as a multicast option, use the
546 		 * address of that interface as our source address.
547 		 */
548 		if (!jailed && IN_MULTICAST(ntohl(sin->sin_addr.s_addr)) &&
549 		    inp->inp_moptions != NULL) {
550 			struct ip_moptions *imo;
551 			struct ifnet *ifp;
552 
553 			imo = inp->inp_moptions;
554 			if (imo->imo_multicast_ifp != NULL) {
555 				struct in_ifaddr_container *iac;
556 
557 				ifp = imo->imo_multicast_ifp;
558 				ia = NULL;
559 				TAILQ_FOREACH(iac,
560 				&in_ifaddrheads[mycpuid], ia_link) {
561 					if (iac->ia->ia_ifp == ifp) {
562 						ia = iac->ia;
563 						break;
564 					}
565 				}
566 				if (ia == NULL)
567 					goto fail;
568 			}
569 		}
570 		/*
571 		 * Don't do pcblookup call here; return interface in plocal_sin
572 		 * and exit to caller, that will do the lookup.
573 		 */
574 		if (ia == NULL && jailed) {
575 			if ((jsin = prison_get_nonlocal(cred->cr_prison, AF_INET, NULL)) != NULL ||
576 			    (jsin = prison_get_local(cred->cr_prison, AF_INET, NULL)) != NULL) {
577 				*plocal_sin = satosin(jsin);
578 			} else {
579 				/* IPv6 only Jail */
580 				goto fail;
581 			}
582 		} else {
583 			*plocal_sin = &ia->ia_addr;
584 		}
585 	}
586 	return (0);
587 fail:
588 	if (alloc_route) {
589 		struct route *ro = &inp->inp_route;
590 
591 		if (ro->ro_rt != NULL)
592 			RTFREE(ro->ro_rt);
593 		bzero(ro, sizeof(*ro));
594 	}
595 	return (EADDRNOTAVAIL);
596 }
597 
598 /*
599  * Outer subroutine:
600  * Connect from a socket to a specified address.
601  * Both address and port must be specified in argument sin.
602  * If don't have a local address for this socket yet,
603  * then pick one.
604  */
605 int
606 in_pcbconnect(struct inpcb *inp, struct sockaddr *nam, struct thread *td)
607 {
608 	struct sockaddr_in *if_sin;
609 	struct sockaddr_in *sin = (struct sockaddr_in *)nam;
610 	int error;
611 
612 	/* Call inner routine to assign local interface address. */
613 	if ((error = in_pcbladdr(inp, nam, &if_sin, td)) != 0)
614 		return (error);
615 
616 	if (in_pcblookup_hash(inp->inp_cpcbinfo, sin->sin_addr, sin->sin_port,
617 	    inp->inp_laddr.s_addr ? inp->inp_laddr : if_sin->sin_addr,
618 	    inp->inp_lport, FALSE, NULL) != NULL) {
619 		return (EADDRINUSE);
620 	}
621 	if (inp->inp_laddr.s_addr == INADDR_ANY) {
622 		if (inp->inp_lport == 0) {
623 			error = in_pcbbind(inp, (struct sockaddr *)NULL, td);
624 			if (error)
625 				return (error);
626 		}
627 		inp->inp_laddr = if_sin->sin_addr;
628 	}
629 	inp->inp_faddr = sin->sin_addr;
630 	inp->inp_fport = sin->sin_port;
631 	in_pcbinsconnhash(inp);
632 	return (0);
633 }
634 
635 void
636 in_pcbdisconnect(struct inpcb *inp)
637 {
638 
639 	inp->inp_faddr.s_addr = INADDR_ANY;
640 	inp->inp_fport = 0;
641 	in_pcbremconnhash(inp);
642 	if (inp->inp_socket->so_state & SS_NOFDREF)
643 		in_pcbdetach(inp);
644 }
645 
646 void
647 in_pcbdetach(struct inpcb *inp)
648 {
649 	struct socket *so = inp->inp_socket;
650 	struct inpcbinfo *ipi = inp->inp_pcbinfo;
651 
652 #ifdef IPSEC
653 	ipsec4_delete_pcbpolicy(inp);
654 #endif /*IPSEC*/
655 	inp->inp_gencnt = ++ipi->ipi_gencnt;
656 	in_pcbremlists(inp);
657 	so->so_pcb = 0;
658 	sofree(so);
659 	if (inp->inp_options)
660 		m_free(inp->inp_options);
661 	if (inp->inp_route.ro_rt)
662 		rtfree(inp->inp_route.ro_rt);
663 	ip_freemoptions(inp->inp_moptions);
664 	inp->inp_vflag = 0;
665 	zfree(ipi->ipi_zone, inp);
666 }
667 
668 /*
669  * The calling convention of in_setsockaddr() and in_setpeeraddr() was
670  * modified to match the pru_sockaddr() and pru_peeraddr() entry points
671  * in struct pr_usrreqs, so that protocols can just reference then directly
672  * without the need for a wrapper function.  The socket must have a valid
673  * (i.e., non-nil) PCB, but it should be impossible to get an invalid one
674  * except through a kernel programming error, so it is acceptable to panic
675  * (or in this case trap) if the PCB is invalid.  (Actually, we don't trap
676  * because there actually /is/ a programming error somewhere... XXX)
677  */
678 int
679 in_setsockaddr(struct socket *so, struct sockaddr **nam)
680 {
681 	struct inpcb *inp;
682 	struct sockaddr_in *sin;
683 
684 	/*
685 	 * Do the malloc first in case it blocks.
686 	 */
687 	MALLOC(sin, struct sockaddr_in *, sizeof *sin, M_SONAME,
688 		M_WAITOK | M_ZERO);
689 	sin->sin_family = AF_INET;
690 	sin->sin_len = sizeof *sin;
691 
692 	crit_enter();
693 	inp = so->so_pcb;
694 	if (!inp) {
695 		crit_exit();
696 		kfree(sin, M_SONAME);
697 		return (ECONNRESET);
698 	}
699 	sin->sin_port = inp->inp_lport;
700 	sin->sin_addr = inp->inp_laddr;
701 	crit_exit();
702 
703 	*nam = (struct sockaddr *)sin;
704 	return (0);
705 }
706 
707 int
708 in_setpeeraddr(struct socket *so, struct sockaddr **nam)
709 {
710 	struct inpcb *inp;
711 	struct sockaddr_in *sin;
712 
713 	/*
714 	 * Do the malloc first in case it blocks.
715 	 */
716 	MALLOC(sin, struct sockaddr_in *, sizeof *sin, M_SONAME,
717 		M_WAITOK | M_ZERO);
718 	sin->sin_family = AF_INET;
719 	sin->sin_len = sizeof *sin;
720 
721 	crit_enter();
722 	inp = so->so_pcb;
723 	if (!inp) {
724 		crit_exit();
725 		kfree(sin, M_SONAME);
726 		return (ECONNRESET);
727 	}
728 	sin->sin_port = inp->inp_fport;
729 	sin->sin_addr = inp->inp_faddr;
730 	crit_exit();
731 
732 	*nam = (struct sockaddr *)sin;
733 	return (0);
734 }
735 
736 void
737 in_pcbnotifyall(struct inpcbhead *head, struct in_addr faddr, int err,
738 		void (*notify)(struct inpcb *, int))
739 {
740 	struct inpcb *inp, *ninp;
741 
742 	/*
743 	 * note: if INP_PLACEMARKER is set we must ignore the rest of
744 	 * the structure and skip it.
745 	 */
746 	crit_enter();
747 	LIST_FOREACH_MUTABLE(inp, head, inp_list, ninp) {
748 		if (inp->inp_flags & INP_PLACEMARKER)
749 			continue;
750 #ifdef INET6
751 		if (!(inp->inp_vflag & INP_IPV4))
752 			continue;
753 #endif
754 		if (inp->inp_faddr.s_addr != faddr.s_addr ||
755 		    inp->inp_socket == NULL)
756 			continue;
757 		(*notify)(inp, err);		/* can remove inp from list! */
758 	}
759 	crit_exit();
760 }
761 
762 void
763 in_pcbpurgeif0(struct inpcb *head, struct ifnet *ifp)
764 {
765 	struct inpcb *inp;
766 	struct ip_moptions *imo;
767 	int i, gap;
768 
769 	for (inp = head; inp != NULL; inp = LIST_NEXT(inp, inp_list)) {
770 		if (inp->inp_flags & INP_PLACEMARKER)
771 			continue;
772 		imo = inp->inp_moptions;
773 		if ((inp->inp_vflag & INP_IPV4) && imo != NULL) {
774 			/*
775 			 * Unselect the outgoing interface if it is being
776 			 * detached.
777 			 */
778 			if (imo->imo_multicast_ifp == ifp)
779 				imo->imo_multicast_ifp = NULL;
780 
781 			/*
782 			 * Drop multicast group membership if we joined
783 			 * through the interface being detached.
784 			 */
785 			for (i = 0, gap = 0; i < imo->imo_num_memberships;
786 			    i++) {
787 				if (imo->imo_membership[i]->inm_ifp == ifp) {
788 					in_delmulti(imo->imo_membership[i]);
789 					gap++;
790 				} else if (gap != 0)
791 					imo->imo_membership[i - gap] =
792 					    imo->imo_membership[i];
793 			}
794 			imo->imo_num_memberships -= gap;
795 		}
796 	}
797 }
798 
799 /*
800  * Check for alternatives when higher level complains
801  * about service problems.  For now, invalidate cached
802  * routing information.  If the route was created dynamically
803  * (by a redirect), time to try a default gateway again.
804  */
805 void
806 in_losing(struct inpcb *inp)
807 {
808 	struct rtentry *rt;
809 	struct rt_addrinfo rtinfo;
810 
811 	if ((rt = inp->inp_route.ro_rt)) {
812 		bzero(&rtinfo, sizeof(struct rt_addrinfo));
813 		rtinfo.rti_info[RTAX_DST] = rt_key(rt);
814 		rtinfo.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
815 		rtinfo.rti_info[RTAX_NETMASK] = rt_mask(rt);
816 		rtinfo.rti_flags = rt->rt_flags;
817 		rt_missmsg(RTM_LOSING, &rtinfo, rt->rt_flags, 0);
818 		if (rt->rt_flags & RTF_DYNAMIC)
819 			rtrequest1_global(RTM_DELETE, &rtinfo, NULL, NULL);
820 		inp->inp_route.ro_rt = NULL;
821 		rtfree(rt);
822 		/*
823 		 * A new route can be allocated
824 		 * the next time output is attempted.
825 		 */
826 	}
827 }
828 
829 /*
830  * After a routing change, flush old routing
831  * and allocate a (hopefully) better one.
832  */
833 void
834 in_rtchange(struct inpcb *inp, int err)
835 {
836 	if (inp->inp_route.ro_rt) {
837 		rtfree(inp->inp_route.ro_rt);
838 		inp->inp_route.ro_rt = NULL;
839 		/*
840 		 * A new route can be allocated the next time
841 		 * output is attempted.
842 		 */
843 	}
844 }
845 
846 /*
847  * Lookup a PCB based on the local address and port.
848  */
849 struct inpcb *
850 in_pcblookup_local(struct inpcbinfo *pcbinfo, struct in_addr laddr,
851 		   u_int lport_arg, int wild_okay, struct ucred *cred)
852 {
853 	struct inpcb *inp;
854 	int matchwild = 3, wildcard;
855 	u_short lport = lport_arg;
856 
857 	struct inpcbporthead *porthash;
858 	struct inpcbport *phd;
859 	struct inpcb *match = NULL;
860 
861 	/*
862 	 * Best fit PCB lookup.
863 	 *
864 	 * First see if this local port is in use by looking on the
865 	 * port hash list.
866 	 */
867 	porthash = &pcbinfo->porthashbase[INP_PCBPORTHASH(lport,
868 	    pcbinfo->porthashmask)];
869 	LIST_FOREACH(phd, porthash, phd_hash) {
870 		if (phd->phd_port == lport)
871 			break;
872 	}
873 	if (phd != NULL) {
874 		/*
875 		 * Port is in use by one or more PCBs. Look for best
876 		 * fit.
877 		 */
878 		LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) {
879 			wildcard = 0;
880 #ifdef INET6
881 			if ((inp->inp_vflag & INP_IPV4) == 0)
882 				continue;
883 #endif
884 			if (inp->inp_faddr.s_addr != INADDR_ANY)
885 				wildcard++;
886 			if (inp->inp_laddr.s_addr != INADDR_ANY) {
887 				if (laddr.s_addr == INADDR_ANY)
888 					wildcard++;
889 				else if (inp->inp_laddr.s_addr != laddr.s_addr)
890 					continue;
891 			} else {
892 				if (laddr.s_addr != INADDR_ANY)
893 					wildcard++;
894 			}
895 			if (wildcard && !wild_okay)
896 				continue;
897 			if (wildcard < matchwild &&
898 			    (cred == NULL ||
899 			     cred->cr_prison ==
900 					inp->inp_socket->so_cred->cr_prison)) {
901 				match = inp;
902 				matchwild = wildcard;
903 				if (matchwild == 0) {
904 					break;
905 				}
906 			}
907 		}
908 	}
909 	return (match);
910 }
911 
912 /*
913  * Lookup PCB in hash list.
914  */
915 struct inpcb *
916 in_pcblookup_hash(struct inpcbinfo *pcbinfo, struct in_addr faddr, u_int fport_arg,
917 		  struct in_addr laddr, u_int lport_arg, boolean_t wildcard,
918 		  struct ifnet *ifp)
919 {
920 	struct inpcbhead *head;
921 	struct inpcb *inp, *jinp=NULL;
922 	u_short fport = fport_arg, lport = lport_arg;
923 
924 	/*
925 	 * First look for an exact match.
926 	 */
927 	head = &pcbinfo->hashbase[INP_PCBCONNHASH(faddr.s_addr, fport,
928 	    laddr.s_addr, lport, pcbinfo->hashmask)];
929 	LIST_FOREACH(inp, head, inp_hash) {
930 #ifdef INET6
931 		if (!(inp->inp_vflag & INP_IPV4))
932 			continue;
933 #endif
934 		if (in_hosteq(inp->inp_faddr, faddr) &&
935 		    in_hosteq(inp->inp_laddr, laddr) &&
936 		    inp->inp_fport == fport && inp->inp_lport == lport) {
937 			/* found */
938 			if (inp->inp_socket == NULL ||
939 			    inp->inp_socket->so_cred->cr_prison == NULL) {
940 				return (inp);
941 			} else {
942 				if  (jinp == NULL)
943 					jinp = inp;
944 			}
945 		}
946 	}
947 	if (jinp != NULL)
948 		return (jinp);
949 	if (wildcard) {
950 		struct inpcb *local_wild = NULL;
951 		struct inpcb *jinp_wild = NULL;
952 #ifdef INET6
953 		struct inpcb *local_wild_mapped = NULL;
954 #endif
955 		struct inpcontainer *ic;
956 		struct inpcontainerhead *chead;
957 		struct sockaddr_in jsin;
958 		struct ucred *cred;
959 
960 		/*
961 		 * Order of socket selection:
962 		 * 1. non-jailed, non-wild.
963 		 * 2. non-jailed, wild.
964 		 * 3. jailed, non-wild.
965 		 * 4. jailed, wild.
966 		 */
967 		jsin.sin_family = AF_INET;
968 		chead = &pcbinfo->wildcardhashbase[
969 		    INP_PCBWILDCARDHASH(lport, pcbinfo->wildcardhashmask)];
970 		LIST_FOREACH(ic, chead, ic_list) {
971 			inp = ic->ic_inp;
972 			jsin.sin_addr.s_addr = laddr.s_addr;
973 #ifdef INET6
974 			if (!(inp->inp_vflag & INP_IPV4))
975 				continue;
976 #endif
977 			if (inp->inp_socket != NULL)
978 				cred = inp->inp_socket->so_cred;
979 			else
980 				cred = NULL;
981 			if (cred != NULL && jailed(cred)) {
982 				if (jinp != NULL)
983 					continue;
984 				else
985 					if (!jailed_ip(cred->cr_prison,
986 					    (struct sockaddr *)&jsin))
987 						continue;
988 			}
989 			if (inp->inp_lport == lport) {
990 				if (ifp && ifp->if_type == IFT_FAITH &&
991 				    !(inp->inp_flags & INP_FAITH))
992 					continue;
993 				if (inp->inp_laddr.s_addr == laddr.s_addr) {
994 					if (cred != NULL && jailed(cred))
995 						jinp = inp;
996 					else
997 						return (inp);
998 				}
999 				if (inp->inp_laddr.s_addr == INADDR_ANY) {
1000 #ifdef INET6
1001 					if (INP_CHECK_SOCKAF(inp->inp_socket,
1002 							     AF_INET6))
1003 						local_wild_mapped = inp;
1004 					else
1005 #endif
1006 						if (cred != NULL &&
1007 						    jailed(cred))
1008 							jinp_wild = inp;
1009 						else
1010 							local_wild = inp;
1011 				}
1012 			}
1013 		}
1014 		if (local_wild != NULL)
1015 			return (local_wild);
1016 #ifdef INET6
1017 		if (local_wild_mapped != NULL)
1018 			return (local_wild_mapped);
1019 #endif
1020 		if (jinp != NULL)
1021 			return (jinp);
1022 		return (jinp_wild);
1023 	}
1024 
1025 	/*
1026 	 * Not found.
1027 	 */
1028 	return (NULL);
1029 }
1030 
1031 /*
1032  * Insert PCB into connection hash table.
1033  */
1034 void
1035 in_pcbinsconnhash(struct inpcb *inp)
1036 {
1037 	struct inpcbinfo *pcbinfo = inp->inp_cpcbinfo;
1038 	struct inpcbhead *bucket;
1039 	u_int32_t hashkey_faddr, hashkey_laddr;
1040 
1041 #ifdef INET6
1042 	if (inp->inp_vflag & INP_IPV6) {
1043 		hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX JH */;
1044 		hashkey_laddr = inp->in6p_laddr.s6_addr32[3] /* XXX JH */;
1045 	} else {
1046 #endif
1047 		hashkey_faddr = inp->inp_faddr.s_addr;
1048 		hashkey_laddr = inp->inp_laddr.s_addr;
1049 #ifdef INET6
1050 	}
1051 #endif
1052 
1053 	KASSERT(!(inp->inp_flags & INP_CONNECTED), ("already on hash list"));
1054 	inp->inp_flags |= INP_CONNECTED;
1055 
1056 	/*
1057 	 * Insert into the connection hash table.
1058 	 */
1059 	bucket = &pcbinfo->hashbase[INP_PCBCONNHASH(hashkey_faddr,
1060 	    inp->inp_fport, hashkey_laddr, inp->inp_lport, pcbinfo->hashmask)];
1061 	LIST_INSERT_HEAD(bucket, inp, inp_hash);
1062 }
1063 
1064 /*
1065  * Remove PCB from connection hash table.
1066  */
1067 void
1068 in_pcbremconnhash(struct inpcb *inp)
1069 {
1070 	KASSERT(inp->inp_flags & INP_CONNECTED, ("inp not connected"));
1071 	LIST_REMOVE(inp, inp_hash);
1072 	inp->inp_flags &= ~INP_CONNECTED;
1073 }
1074 
1075 /*
1076  * Insert PCB into port hash table.
1077  */
1078 int
1079 in_pcbinsporthash(struct inpcb *inp)
1080 {
1081 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1082 	struct inpcbporthead *pcbporthash;
1083 	struct inpcbport *phd;
1084 
1085 	/*
1086 	 * Insert into the port hash table.
1087 	 */
1088 	pcbporthash = &pcbinfo->porthashbase[
1089 	    INP_PCBPORTHASH(inp->inp_lport, pcbinfo->porthashmask)];
1090 
1091 	/* Go through port list and look for a head for this lport. */
1092 	LIST_FOREACH(phd, pcbporthash, phd_hash)
1093 		if (phd->phd_port == inp->inp_lport)
1094 			break;
1095 
1096 	/* If none exists, malloc one and tack it on. */
1097 	if (phd == NULL) {
1098 		MALLOC(phd, struct inpcbport *, sizeof(struct inpcbport),
1099 		    M_PCB, M_INTWAIT | M_NULLOK);
1100 		if (phd == NULL)
1101 			return (ENOBUFS); /* XXX */
1102 		phd->phd_port = inp->inp_lport;
1103 		LIST_INIT(&phd->phd_pcblist);
1104 		LIST_INSERT_HEAD(pcbporthash, phd, phd_hash);
1105 	}
1106 
1107 	inp->inp_phd = phd;
1108 	LIST_INSERT_HEAD(&phd->phd_pcblist, inp, inp_portlist);
1109 
1110 	return (0);
1111 }
1112 
1113 void
1114 in_pcbinswildcardhash_oncpu(struct inpcb *inp, struct inpcbinfo *pcbinfo)
1115 {
1116 	struct inpcontainer *ic;
1117 	struct inpcontainerhead *bucket;
1118 
1119 	bucket = &pcbinfo->wildcardhashbase[
1120 	    INP_PCBWILDCARDHASH(inp->inp_lport, pcbinfo->wildcardhashmask)];
1121 
1122 	ic = kmalloc(sizeof(struct inpcontainer), M_TEMP, M_INTWAIT);
1123 	ic->ic_inp = inp;
1124 	LIST_INSERT_HEAD(bucket, ic, ic_list);
1125 }
1126 
1127 /*
1128  * Insert PCB into wildcard hash table.
1129  */
1130 void
1131 in_pcbinswildcardhash(struct inpcb *inp)
1132 {
1133 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1134 
1135 	KKASSERT(pcbinfo != NULL);
1136 
1137 	in_pcbinswildcardhash_oncpu(inp, pcbinfo);
1138 	inp->inp_flags |= INP_WILDCARD;
1139 }
1140 
1141 void
1142 in_pcbremwildcardhash_oncpu(struct inpcb *inp, struct inpcbinfo *pcbinfo)
1143 {
1144 	struct inpcontainer *ic;
1145 	struct inpcontainerhead *head;
1146 
1147 	/* find bucket */
1148 	head = &pcbinfo->wildcardhashbase[
1149 	    INP_PCBWILDCARDHASH(inp->inp_lport, pcbinfo->wildcardhashmask)];
1150 
1151 	LIST_FOREACH(ic, head, ic_list) {
1152 		if (ic->ic_inp == inp)
1153 			goto found;
1154 	}
1155 	return;			/* not found! */
1156 
1157 found:
1158 	LIST_REMOVE(ic, ic_list);	/* remove container from bucket chain */
1159 	kfree(ic, M_TEMP);		/* deallocate container */
1160 }
1161 
1162 /*
1163  * Remove PCB from wildcard hash table.
1164  */
1165 void
1166 in_pcbremwildcardhash(struct inpcb *inp)
1167 {
1168 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1169 
1170 	KASSERT(inp->inp_flags & INP_WILDCARD, ("inp not wildcard"));
1171 	in_pcbremwildcardhash_oncpu(inp, pcbinfo);
1172 	inp->inp_flags &= ~INP_WILDCARD;
1173 }
1174 
1175 /*
1176  * Remove PCB from various lists.
1177  */
1178 void
1179 in_pcbremlists(struct inpcb *inp)
1180 {
1181 	if (inp->inp_lport) {
1182 		struct inpcbport *phd = inp->inp_phd;
1183 
1184 		LIST_REMOVE(inp, inp_portlist);
1185 		if (LIST_FIRST(&phd->phd_pcblist) == NULL) {
1186 			LIST_REMOVE(phd, phd_hash);
1187 			kfree(phd, M_PCB);
1188 		}
1189 	}
1190 	if (inp->inp_flags & INP_WILDCARD) {
1191 		in_pcbremwildcardhash(inp);
1192 	} else if (inp->inp_flags & INP_CONNECTED) {
1193 		in_pcbremconnhash(inp);
1194 	}
1195 	LIST_REMOVE(inp, inp_list);
1196 	inp->inp_pcbinfo->ipi_count--;
1197 }
1198 
1199 int
1200 prison_xinpcb(struct thread *td, struct inpcb *inp)
1201 {
1202 	struct ucred *cr;
1203 
1204 	if (td->td_proc == NULL)
1205 		return (0);
1206 	cr = td->td_proc->p_ucred;
1207 	if (cr->cr_prison == NULL)
1208 		return (0);
1209 	if (inp->inp_socket && inp->inp_socket->so_cred &&
1210 	    inp->inp_socket->so_cred->cr_prison &&
1211 	    cr->cr_prison == inp->inp_socket->so_cred->cr_prison)
1212 		return (0);
1213 	return (1);
1214 }
1215 
1216 int
1217 in_pcblist_global(SYSCTL_HANDLER_ARGS)
1218 {
1219 	struct inpcbinfo *pcbinfo = arg1;
1220 	struct inpcb *inp, *marker;
1221 	struct xinpcb xi;
1222 	int error, i, n;
1223 	inp_gen_t gencnt;
1224 
1225 	/*
1226 	 * The process of preparing the TCB list is too time-consuming and
1227 	 * resource-intensive to repeat twice on every request.
1228 	 */
1229 	if (req->oldptr == NULL) {
1230 		n = pcbinfo->ipi_count;
1231 		req->oldidx = (n + n/8 + 10) * sizeof(struct xinpcb);
1232 		return 0;
1233 	}
1234 
1235 	if (req->newptr != NULL)
1236 		return EPERM;
1237 
1238 	/*
1239 	 * OK, now we're committed to doing something.  Re-fetch ipi_count
1240 	 * after obtaining the generation count.
1241 	 */
1242 	gencnt = pcbinfo->ipi_gencnt;
1243 	n = pcbinfo->ipi_count;
1244 
1245 	marker = kmalloc(sizeof(struct inpcb), M_TEMP, M_WAITOK|M_ZERO);
1246 	marker->inp_flags |= INP_PLACEMARKER;
1247 	LIST_INSERT_HEAD(&pcbinfo->pcblisthead, marker, inp_list);
1248 
1249 	i = 0;
1250 	error = 0;
1251 
1252 	while ((inp = LIST_NEXT(marker, inp_list)) != NULL && i < n) {
1253 		LIST_REMOVE(marker, inp_list);
1254 		LIST_INSERT_AFTER(inp, marker, inp_list);
1255 
1256 		if (inp->inp_flags & INP_PLACEMARKER)
1257 			continue;
1258 		if (inp->inp_gencnt > gencnt)
1259 			continue;
1260 		if (prison_xinpcb(req->td, inp))
1261 			continue;
1262 		bzero(&xi, sizeof xi);
1263 		xi.xi_len = sizeof xi;
1264 		bcopy(inp, &xi.xi_inp, sizeof *inp);
1265 		if (inp->inp_socket)
1266 			sotoxsocket(inp->inp_socket, &xi.xi_socket);
1267 		if ((error = SYSCTL_OUT(req, &xi, sizeof xi)) != 0)
1268 			break;
1269 		++i;
1270 	}
1271 	LIST_REMOVE(marker, inp_list);
1272 	if (error == 0 && i < n) {
1273 		bzero(&xi, sizeof xi);
1274 		xi.xi_len = sizeof xi;
1275 		while (i < n) {
1276 			error = SYSCTL_OUT(req, &xi, sizeof xi);
1277 			++i;
1278 		}
1279 	}
1280 	kfree(marker, M_TEMP);
1281 	return(error);
1282 }
1283