xref: /dragonfly/sys/netinet/in_pcb.c (revision 4a65f651)
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 <vm/vm_zone.h>
92 
93 #include <net/if.h>
94 #include <net/if_types.h>
95 #include <net/route.h>
96 
97 #include <netinet/in.h>
98 #include <netinet/in_pcb.h>
99 #include <netinet/in_var.h>
100 #include <netinet/ip_var.h>
101 #ifdef INET6
102 #include <netinet/ip6.h>
103 #include <netinet6/ip6_var.h>
104 #endif /* INET6 */
105 
106 #ifdef IPSEC
107 #include <netinet6/ipsec.h>
108 #include <netproto/key/key.h>
109 #endif
110 
111 #ifdef FAST_IPSEC
112 #if defined(IPSEC) || defined(IPSEC_ESP)
113 #error "Bad idea: don't compile with both IPSEC and FAST_IPSEC!"
114 #endif
115 
116 #include <netproto/ipsec/ipsec.h>
117 #include <netproto/ipsec/key.h>
118 #define	IPSEC
119 #endif /* FAST_IPSEC */
120 
121 struct in_addr zeroin_addr;
122 
123 /*
124  * These configure the range of local port addresses assigned to
125  * "unspecified" outgoing connections/packets/whatever.
126  */
127 int ipport_lowfirstauto = IPPORT_RESERVED - 1;	/* 1023 */
128 int ipport_lowlastauto = IPPORT_RESERVEDSTART;	/* 600 */
129 
130 int ipport_firstauto = IPPORT_RESERVED;		/* 1024 */
131 int ipport_lastauto = IPPORT_USERRESERVED;	/* 5000 */
132 
133 int ipport_hifirstauto = IPPORT_HIFIRSTAUTO;	/* 49152 */
134 int ipport_hilastauto = IPPORT_HILASTAUTO;	/* 65535 */
135 
136 static __inline void
137 RANGECHK(int var, int min, int max)
138 {
139 	if (var < min)
140 		var = min;
141 	else if (var > max)
142 		var = max;
143 }
144 
145 static int
146 sysctl_net_ipport_check(SYSCTL_HANDLER_ARGS)
147 {
148 	int error;
149 
150 	error = sysctl_handle_int(oidp, oidp->oid_arg1, oidp->oid_arg2, req);
151 	if (!error) {
152 		RANGECHK(ipport_lowfirstauto, 1, IPPORT_RESERVED - 1);
153 		RANGECHK(ipport_lowlastauto, 1, IPPORT_RESERVED - 1);
154 
155 		RANGECHK(ipport_firstauto, IPPORT_RESERVED, USHRT_MAX);
156 		RANGECHK(ipport_lastauto, IPPORT_RESERVED, USHRT_MAX);
157 
158 		RANGECHK(ipport_hifirstauto, IPPORT_RESERVED, USHRT_MAX);
159 		RANGECHK(ipport_hilastauto, IPPORT_RESERVED, USHRT_MAX);
160 	}
161 	return (error);
162 }
163 
164 SYSCTL_NODE(_net_inet_ip, IPPROTO_IP, portrange, CTLFLAG_RW, 0, "IP Ports");
165 
166 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, lowfirst, CTLTYPE_INT|CTLFLAG_RW,
167 	   &ipport_lowfirstauto, 0, &sysctl_net_ipport_check, "I", "");
168 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, lowlast, CTLTYPE_INT|CTLFLAG_RW,
169 	   &ipport_lowlastauto, 0, &sysctl_net_ipport_check, "I", "");
170 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, first, CTLTYPE_INT|CTLFLAG_RW,
171 	   &ipport_firstauto, 0, &sysctl_net_ipport_check, "I", "");
172 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, last, CTLTYPE_INT|CTLFLAG_RW,
173 	   &ipport_lastauto, 0, &sysctl_net_ipport_check, "I", "");
174 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, hifirst, CTLTYPE_INT|CTLFLAG_RW,
175 	   &ipport_hifirstauto, 0, &sysctl_net_ipport_check, "I", "");
176 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, hilast, CTLTYPE_INT|CTLFLAG_RW,
177 	   &ipport_hilastauto, 0, &sysctl_net_ipport_check, "I", "");
178 
179 /*
180  * in_pcb.c: manage the Protocol Control Blocks.
181  *
182  * NOTE: It is assumed that most of these functions will be called from
183  * a critical section.  XXX - There are, unfortunately, a few exceptions
184  * to this rule that should be fixed.
185  *
186  * NOTE: The caller should initialize the cpu field to the cpu running the
187  * protocol stack associated with this inpcbinfo.
188  */
189 
190 void
191 in_pcbinfo_init(struct inpcbinfo *pcbinfo)
192 {
193 	LIST_INIT(&pcbinfo->pcblisthead);
194 	pcbinfo->cpu = -1;
195 }
196 
197 /*
198  * Allocate a PCB and associate it with the socket.
199  */
200 int
201 in_pcballoc(struct socket *so, struct inpcbinfo *pcbinfo)
202 {
203 	struct inpcb *inp;
204 #ifdef IPSEC
205 	int error;
206 #endif
207 
208 	inp = zalloc(pcbinfo->ipi_zone);
209 	if (inp == NULL)
210 		return (ENOBUFS);
211 	bzero(inp, sizeof *inp);
212 	inp->inp_gencnt = ++pcbinfo->ipi_gencnt;
213 	inp->inp_pcbinfo = inp->inp_cpcbinfo = pcbinfo;
214 	inp->inp_socket = so;
215 #ifdef IPSEC
216 	error = ipsec_init_policy(so, &inp->inp_sp);
217 	if (error != 0) {
218 		zfree(pcbinfo->ipi_zone, inp);
219 		return (error);
220 	}
221 #endif
222 #ifdef INET6
223 	if (INP_SOCKAF(so) == AF_INET6 && ip6_v6only)
224 		inp->inp_flags |= IN6P_IPV6_V6ONLY;
225 	if (ip6_auto_flowlabel)
226 		inp->inp_flags |= IN6P_AUTOFLOWLABEL;
227 #endif
228 	so->so_pcb = inp;
229 	LIST_INSERT_HEAD(&pcbinfo->pcblisthead, inp, inp_list);
230 	pcbinfo->ipi_count++;
231 	return (0);
232 }
233 
234 int
235 in_pcbbind(struct inpcb *inp, struct sockaddr *nam, struct thread *td)
236 {
237 	struct socket *so = inp->inp_socket;
238 	struct proc *p = td->td_proc;
239 	unsigned short *lastport;
240 	struct sockaddr_in *sin;
241 	struct sockaddr_in jsin;
242 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
243 	struct ucred *cred = NULL;
244 	u_short lport = 0;
245 	int wild = 0, reuseport = (so->so_options & SO_REUSEPORT);
246 	int error;
247 
248 	KKASSERT(p);
249 
250 	if (TAILQ_EMPTY(&in_ifaddrheads[mycpuid])) /* XXX broken! */
251 		return (EADDRNOTAVAIL);
252 	if (inp->inp_lport != 0 || inp->inp_laddr.s_addr != INADDR_ANY)
253 		return (EINVAL);	/* already bound */
254 	if (!(so->so_options & (SO_REUSEADDR|SO_REUSEPORT)))
255 		wild = 1;    /* neither SO_REUSEADDR nor SO_REUSEPORT is set */
256 	if (p)
257 		cred = p->p_ucred;
258 	if (nam != NULL) {
259 		sin = (struct sockaddr_in *)nam;
260 		if (nam->sa_len != sizeof *sin)
261 			return (EINVAL);
262 #ifdef notdef
263 		/*
264 		 * We should check the family, but old programs
265 		 * incorrectly fail to initialize it.
266 		 */
267 		if (sin->sin_family != AF_INET)
268 			return (EAFNOSUPPORT);
269 #endif
270 		if (!prison_replace_wildcards(td, nam))
271 				return (EINVAL);
272 		lport = sin->sin_port;
273 		if (IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) {
274 			/*
275 			 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
276 			 * allow complete duplication of binding if
277 			 * SO_REUSEPORT is set, or if SO_REUSEADDR is set
278 			 * and a multicast address is bound on both
279 			 * new and duplicated sockets.
280 			 */
281 			if (so->so_options & SO_REUSEADDR)
282 				reuseport = SO_REUSEADDR | SO_REUSEPORT;
283 		} else if (sin->sin_addr.s_addr != INADDR_ANY) {
284 			sin->sin_port = 0;		/* yech... */
285 			bzero(&sin->sin_zero, sizeof sin->sin_zero);
286 			if (ifa_ifwithaddr((struct sockaddr *)sin) == NULL)
287 				return (EADDRNOTAVAIL);
288 		}
289 		if (lport != 0) {
290 			struct inpcb *t;
291 
292 			/* GROSS */
293 			if (ntohs(lport) < IPPORT_RESERVED &&
294 			    cred && priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT, 0))
295 				return (EACCES);
296 			if (so->so_cred->cr_uid != 0 &&
297 			    !IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) {
298 				t = in_pcblookup_local(inp->inp_pcbinfo,
299 				    sin->sin_addr, lport,
300 				    INPLOOKUP_WILDCARD, cred);
301 				if (t &&
302 				    (!in_nullhost(sin->sin_addr) ||
303 				     !in_nullhost(t->inp_laddr) ||
304 				     (t->inp_socket->so_options &
305 					 SO_REUSEPORT) == 0) &&
306 				    (so->so_cred->cr_uid !=
307 				     t->inp_socket->so_cred->cr_uid)) {
308 #ifdef INET6
309 					if (!in_nullhost(sin->sin_addr) ||
310 					    !in_nullhost(t->inp_laddr) ||
311 					    INP_SOCKAF(so) ==
312 					    INP_SOCKAF(t->inp_socket))
313 #endif
314 					return (EADDRINUSE);
315 				}
316 			}
317 			if (cred && !prison_replace_wildcards(td, nam))
318 				return (EADDRNOTAVAIL);
319 			t = in_pcblookup_local(pcbinfo, sin->sin_addr, lport,
320 					       wild, cred);
321 			if (t && !(reuseport & t->inp_socket->so_options)) {
322 #ifdef INET6
323 				if (!in_nullhost(sin->sin_addr) ||
324 				    !in_nullhost(t->inp_laddr) ||
325 				    INP_SOCKAF(so) == INP_SOCKAF(t->inp_socket))
326 #endif
327 				return (EADDRINUSE);
328 			}
329 		}
330 		inp->inp_laddr = sin->sin_addr;
331 	}
332 	if (lport == 0) {
333 		ushort first, last;
334 		int count;
335 
336 		jsin.sin_family = AF_INET;
337 		jsin.sin_addr.s_addr = inp->inp_laddr.s_addr;
338 		if (!prison_replace_wildcards(td, (struct sockaddr *)&jsin)) {
339 			inp->inp_laddr.s_addr = INADDR_ANY;
340 			return (EINVAL);
341 		}
342 		inp->inp_laddr.s_addr = jsin.sin_addr.s_addr;
343 
344 		inp->inp_flags |= INP_ANONPORT;
345 
346 		if (inp->inp_flags & INP_HIGHPORT) {
347 			first = ipport_hifirstauto;	/* sysctl */
348 			last  = ipport_hilastauto;
349 			lastport = &pcbinfo->lasthi;
350 		} else if (inp->inp_flags & INP_LOWPORT) {
351 			if (cred &&
352 			    (error = priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT, 0))) {
353 				inp->inp_laddr.s_addr = INADDR_ANY;
354 				return (error);
355 			}
356 			first = ipport_lowfirstauto;	/* 1023 */
357 			last  = ipport_lowlastauto;	/* 600 */
358 			lastport = &pcbinfo->lastlow;
359 		} else {
360 			first = ipport_firstauto;	/* sysctl */
361 			last  = ipport_lastauto;
362 			lastport = &pcbinfo->lastport;
363 		}
364 		/*
365 		 * Simple check to ensure all ports are not used up causing
366 		 * a deadlock here.
367 		 *
368 		 * We split the two cases (up and down) so that the direction
369 		 * is not being tested on each round of the loop.
370 		 */
371 		if (first > last) {
372 			/*
373 			 * counting down
374 			 */
375 			count = first - last;
376 
377 			do {
378 				if (count-- < 0) {	/* completely used? */
379 					inp->inp_laddr.s_addr = INADDR_ANY;
380 					return (EADDRNOTAVAIL);
381 				}
382 				--*lastport;
383 				if (*lastport > first || *lastport < last)
384 					*lastport = first;
385 				lport = htons(*lastport);
386 			} while (in_pcblookup_local(pcbinfo, inp->inp_laddr,
387 				 lport, wild, cred));
388 		} else {
389 			/*
390 			 * counting up
391 			 */
392 			count = last - first;
393 
394 			do {
395 				if (count-- < 0) {	/* completely used? */
396 					inp->inp_laddr.s_addr = INADDR_ANY;
397 					return (EADDRNOTAVAIL);
398 				}
399 				++*lastport;
400 				if (*lastport < first || *lastport > last)
401 					*lastport = first;
402 				lport = htons(*lastport);
403 			} while (in_pcblookup_local(pcbinfo, inp->inp_laddr,
404 				 lport, wild, cred));
405 		}
406 	}
407 	inp->inp_lport = lport;
408 
409 	jsin.sin_family = AF_INET;
410 	jsin.sin_addr.s_addr = inp->inp_laddr.s_addr;
411 	if (!prison_replace_wildcards(td, (struct sockaddr*)&jsin)) {
412 		inp->inp_laddr.s_addr = INADDR_ANY;
413 		inp->inp_lport = 0;
414 		return (EINVAL);
415 	}
416 	inp->inp_laddr.s_addr = jsin.sin_addr.s_addr;
417 
418 	if (in_pcbinsporthash(inp) != 0) {
419 		inp->inp_laddr.s_addr = INADDR_ANY;
420 		inp->inp_lport = 0;
421 		return (EAGAIN);
422 	}
423 	return (0);
424 }
425 
426 /*
427  *   Transform old in_pcbconnect() into an inner subroutine for new
428  *   in_pcbconnect(): Do some validity-checking on the remote
429  *   address (in mbuf 'nam') and then determine local host address
430  *   (i.e., which interface) to use to access that remote host.
431  *
432  *   This preserves definition of in_pcbconnect(), while supporting a
433  *   slightly different version for T/TCP.  (This is more than
434  *   a bit of a kludge, but cleaning up the internal interfaces would
435  *   have forced minor changes in every protocol).
436  */
437 int
438 in_pcbladdr(struct inpcb *inp, struct sockaddr *nam,
439 	struct sockaddr_in **plocal_sin, struct thread *td)
440 {
441 	struct in_ifaddr *ia;
442 	struct ucred *cred = NULL;
443 	struct sockaddr_in *sin = (struct sockaddr_in *)nam;
444 	struct sockaddr *jsin;
445 	int jailed = 0, alloc_route = 0;
446 
447 	if (nam->sa_len != sizeof *sin)
448 		return (EINVAL);
449 	if (sin->sin_family != AF_INET)
450 		return (EAFNOSUPPORT);
451 	if (sin->sin_port == 0)
452 		return (EADDRNOTAVAIL);
453 	if (td && td->td_proc && td->td_proc->p_ucred)
454 		cred = td->td_proc->p_ucred;
455 	if (cred && cred->cr_prison)
456 		jailed = 1;
457 	if (!TAILQ_EMPTY(&in_ifaddrheads[mycpuid])) {
458 		ia = TAILQ_FIRST(&in_ifaddrheads[mycpuid])->ia;
459 		/*
460 		 * If the destination address is INADDR_ANY,
461 		 * use the primary local address.
462 		 * If the supplied address is INADDR_BROADCAST,
463 		 * and the primary interface supports broadcast,
464 		 * choose the broadcast address for that interface.
465 		 */
466 		if (sin->sin_addr.s_addr == INADDR_ANY)
467 			sin->sin_addr = IA_SIN(ia)->sin_addr;
468 		else if (sin->sin_addr.s_addr == (u_long)INADDR_BROADCAST &&
469 		    (ia->ia_ifp->if_flags & IFF_BROADCAST))
470 			sin->sin_addr = satosin(&ia->ia_broadaddr)->sin_addr;
471 	}
472 	if (inp->inp_laddr.s_addr == INADDR_ANY) {
473 		struct route *ro;
474 
475 		ia = NULL;
476 		/*
477 		 * If route is known or can be allocated now,
478 		 * our src addr is taken from the i/f, else punt.
479 		 * Note that we should check the address family of the cached
480 		 * destination, in case of sharing the cache with IPv6.
481 		 */
482 		ro = &inp->inp_route;
483 		if (ro->ro_rt &&
484 		    (!(ro->ro_rt->rt_flags & RTF_UP) ||
485 		     ro->ro_dst.sa_family != AF_INET ||
486 		     satosin(&ro->ro_dst)->sin_addr.s_addr !=
487 				      sin->sin_addr.s_addr ||
488 		     inp->inp_socket->so_options & SO_DONTROUTE)) {
489 			RTFREE(ro->ro_rt);
490 			ro->ro_rt = NULL;
491 		}
492 		if (!(inp->inp_socket->so_options & SO_DONTROUTE) && /*XXX*/
493 		    (ro->ro_rt == NULL ||
494 		    ro->ro_rt->rt_ifp == NULL)) {
495 			/* No route yet, so try to acquire one */
496 			bzero(&ro->ro_dst, sizeof(struct sockaddr_in));
497 			ro->ro_dst.sa_family = AF_INET;
498 			ro->ro_dst.sa_len = sizeof(struct sockaddr_in);
499 			((struct sockaddr_in *) &ro->ro_dst)->sin_addr =
500 				sin->sin_addr;
501 			rtalloc(ro);
502 			alloc_route = 1;
503 		}
504 		/*
505 		 * If we found a route, use the address
506 		 * corresponding to the outgoing interface
507 		 * unless it is the loopback (in case a route
508 		 * to our address on another net goes to loopback).
509 		 */
510 		if (ro->ro_rt && !(ro->ro_rt->rt_ifp->if_flags & IFF_LOOPBACK)) {
511 			if (jailed) {
512 				if (jailed_ip(cred->cr_prison,
513 				    ro->ro_rt->rt_ifa->ifa_addr)) {
514 					ia = ifatoia(ro->ro_rt->rt_ifa);
515 				}
516 			} else {
517 				ia = ifatoia(ro->ro_rt->rt_ifa);
518 			}
519 		}
520 		if (ia == NULL) {
521 			u_short fport = sin->sin_port;
522 
523 			sin->sin_port = 0;
524 			ia = ifatoia(ifa_ifwithdstaddr(sintosa(sin)));
525 			if (ia && jailed && !jailed_ip(cred->cr_prison,
526 			    sintosa(&ia->ia_addr)))
527 				ia = NULL;
528 			if (ia == NULL)
529 				ia = ifatoia(ifa_ifwithnet(sintosa(sin)));
530 			if (ia && jailed && !jailed_ip(cred->cr_prison,
531 			    sintosa(&ia->ia_addr)))
532 				ia = NULL;
533 			sin->sin_port = fport;
534 			if (ia == NULL &&
535 			    !TAILQ_EMPTY(&in_ifaddrheads[mycpuid]))
536 				ia = TAILQ_FIRST(&in_ifaddrheads[mycpuid])->ia;
537 			if (ia && jailed && !jailed_ip(cred->cr_prison,
538 			    sintosa(&ia->ia_addr)))
539 				ia = NULL;
540 
541 			if (!jailed && ia == NULL)
542 				goto fail;
543 		}
544 		/*
545 		 * If the destination address is multicast and an outgoing
546 		 * interface has been set as a multicast option, use the
547 		 * address of that interface as our source address.
548 		 */
549 		if (!jailed && IN_MULTICAST(ntohl(sin->sin_addr.s_addr)) &&
550 		    inp->inp_moptions != NULL) {
551 			struct ip_moptions *imo;
552 			struct ifnet *ifp;
553 
554 			imo = inp->inp_moptions;
555 			if (imo->imo_multicast_ifp != NULL) {
556 				struct in_ifaddr_container *iac;
557 
558 				ifp = imo->imo_multicast_ifp;
559 				ia = NULL;
560 				TAILQ_FOREACH(iac,
561 				&in_ifaddrheads[mycpuid], ia_link) {
562 					if (iac->ia->ia_ifp == ifp) {
563 						ia = iac->ia;
564 						break;
565 					}
566 				}
567 				if (ia == NULL)
568 					goto fail;
569 			}
570 		}
571 		/*
572 		 * Don't do pcblookup call here; return interface in plocal_sin
573 		 * and exit to caller, that will do the lookup.
574 		 */
575 		if (ia == NULL && jailed) {
576 			if ((jsin = prison_get_nonlocal(cred->cr_prison, AF_INET, NULL)) != NULL ||
577 			    (jsin = prison_get_local(cred->cr_prison, AF_INET, NULL)) != NULL) {
578 				*plocal_sin = satosin(jsin);
579 			} else {
580 				/* IPv6 only Jail */
581 				goto fail;
582 			}
583 		} else {
584 			*plocal_sin = &ia->ia_addr;
585 		}
586 	}
587 	return (0);
588 fail:
589 	if (alloc_route) {
590 		struct route *ro = &inp->inp_route;
591 
592 		if (ro->ro_rt != NULL)
593 			RTFREE(ro->ro_rt);
594 		bzero(ro, sizeof(*ro));
595 	}
596 	return (EADDRNOTAVAIL);
597 }
598 
599 /*
600  * Outer subroutine:
601  * Connect from a socket to a specified address.
602  * Both address and port must be specified in argument sin.
603  * If don't have a local address for this socket yet,
604  * then pick one.
605  */
606 int
607 in_pcbconnect(struct inpcb *inp, struct sockaddr *nam, struct thread *td)
608 {
609 	struct sockaddr_in *if_sin;
610 	struct sockaddr_in *sin = (struct sockaddr_in *)nam;
611 	int error;
612 
613 	/* Call inner routine to assign local interface address. */
614 	if ((error = in_pcbladdr(inp, nam, &if_sin, td)) != 0)
615 		return (error);
616 
617 	if (in_pcblookup_hash(inp->inp_cpcbinfo, sin->sin_addr, sin->sin_port,
618 	    inp->inp_laddr.s_addr ? inp->inp_laddr : if_sin->sin_addr,
619 	    inp->inp_lport, FALSE, NULL) != NULL) {
620 		return (EADDRINUSE);
621 	}
622 	if (inp->inp_laddr.s_addr == INADDR_ANY) {
623 		if (inp->inp_lport == 0) {
624 			error = in_pcbbind(inp, NULL, td);
625 			if (error)
626 				return (error);
627 		}
628 		inp->inp_laddr = if_sin->sin_addr;
629 	}
630 	inp->inp_faddr = sin->sin_addr;
631 	inp->inp_fport = sin->sin_port;
632 	in_pcbinsconnhash(inp);
633 	return (0);
634 }
635 
636 void
637 in_pcbdisconnect(struct inpcb *inp)
638 {
639 
640 	inp->inp_faddr.s_addr = INADDR_ANY;
641 	inp->inp_fport = 0;
642 	in_pcbremconnhash(inp);
643 	if (inp->inp_socket->so_state & SS_NOFDREF)
644 		in_pcbdetach(inp);
645 }
646 
647 void
648 in_pcbdetach(struct inpcb *inp)
649 {
650 	struct socket *so = inp->inp_socket;
651 	struct inpcbinfo *ipi = inp->inp_pcbinfo;
652 
653 #ifdef IPSEC
654 	ipsec4_delete_pcbpolicy(inp);
655 #endif /*IPSEC*/
656 	inp->inp_gencnt = ++ipi->ipi_gencnt;
657 	in_pcbremlists(inp);
658 	so->so_pcb = 0;
659 	sofree(so);
660 	if (inp->inp_options)
661 		m_free(inp->inp_options);
662 	if (inp->inp_route.ro_rt)
663 		rtfree(inp->inp_route.ro_rt);
664 	ip_freemoptions(inp->inp_moptions);
665 	inp->inp_vflag = 0;
666 	zfree(ipi->ipi_zone, inp);
667 }
668 
669 /*
670  * The calling convention of in_setsockaddr() and in_setpeeraddr() was
671  * modified to match the pru_sockaddr() and pru_peeraddr() entry points
672  * in struct pr_usrreqs, so that protocols can just reference then directly
673  * without the need for a wrapper function.  The socket must have a valid
674  * (i.e., non-nil) PCB, but it should be impossible to get an invalid one
675  * except through a kernel programming error, so it is acceptable to panic
676  * (or in this case trap) if the PCB is invalid.  (Actually, we don't trap
677  * because there actually /is/ a programming error somewhere... XXX)
678  */
679 int
680 in_setsockaddr(struct socket *so, struct sockaddr **nam)
681 {
682 	struct inpcb *inp;
683 	struct sockaddr_in *sin;
684 
685 	/*
686 	 * Do the malloc first in case it blocks.
687 	 */
688 	MALLOC(sin, struct sockaddr_in *, sizeof *sin, M_SONAME,
689 		M_WAITOK | M_ZERO);
690 	sin->sin_family = AF_INET;
691 	sin->sin_len = sizeof *sin;
692 
693 	crit_enter();
694 	inp = so->so_pcb;
695 	if (!inp) {
696 		crit_exit();
697 		kfree(sin, M_SONAME);
698 		return (ECONNRESET);
699 	}
700 	sin->sin_port = inp->inp_lport;
701 	sin->sin_addr = inp->inp_laddr;
702 	crit_exit();
703 
704 	*nam = (struct sockaddr *)sin;
705 	return (0);
706 }
707 
708 int
709 in_setpeeraddr(struct socket *so, struct sockaddr **nam)
710 {
711 	struct inpcb *inp;
712 	struct sockaddr_in *sin;
713 
714 	/*
715 	 * Do the malloc first in case it blocks.
716 	 */
717 	MALLOC(sin, struct sockaddr_in *, sizeof *sin, M_SONAME,
718 		M_WAITOK | M_ZERO);
719 	sin->sin_family = AF_INET;
720 	sin->sin_len = sizeof *sin;
721 
722 	crit_enter();
723 	inp = so->so_pcb;
724 	if (!inp) {
725 		crit_exit();
726 		kfree(sin, M_SONAME);
727 		return (ECONNRESET);
728 	}
729 	sin->sin_port = inp->inp_fport;
730 	sin->sin_addr = inp->inp_faddr;
731 	crit_exit();
732 
733 	*nam = (struct sockaddr *)sin;
734 	return (0);
735 }
736 
737 void
738 in_pcbnotifyall(struct inpcbhead *head, struct in_addr faddr, int err,
739 		void (*notify)(struct inpcb *, int))
740 {
741 	struct inpcb *inp, *ninp;
742 
743 	/*
744 	 * note: if INP_PLACEMARKER is set we must ignore the rest of
745 	 * the structure and skip it.
746 	 */
747 	crit_enter();
748 	LIST_FOREACH_MUTABLE(inp, head, inp_list, ninp) {
749 		if (inp->inp_flags & INP_PLACEMARKER)
750 			continue;
751 #ifdef INET6
752 		if (!(inp->inp_vflag & INP_IPV4))
753 			continue;
754 #endif
755 		if (inp->inp_faddr.s_addr != faddr.s_addr ||
756 		    inp->inp_socket == NULL)
757 			continue;
758 		(*notify)(inp, err);		/* can remove inp from list! */
759 	}
760 	crit_exit();
761 }
762 
763 void
764 in_pcbpurgeif0(struct inpcb *head, struct ifnet *ifp)
765 {
766 	struct inpcb *inp;
767 	struct ip_moptions *imo;
768 	int i, gap;
769 
770 	for (inp = head; inp != NULL; inp = LIST_NEXT(inp, inp_list)) {
771 		if (inp->inp_flags & INP_PLACEMARKER)
772 			continue;
773 		imo = inp->inp_moptions;
774 		if ((inp->inp_vflag & INP_IPV4) && imo != NULL) {
775 			/*
776 			 * Unselect the outgoing interface if it is being
777 			 * detached.
778 			 */
779 			if (imo->imo_multicast_ifp == ifp)
780 				imo->imo_multicast_ifp = NULL;
781 
782 			/*
783 			 * Drop multicast group membership if we joined
784 			 * through the interface being detached.
785 			 */
786 			for (i = 0, gap = 0; i < imo->imo_num_memberships;
787 			    i++) {
788 				if (imo->imo_membership[i]->inm_ifp == ifp) {
789 					in_delmulti(imo->imo_membership[i]);
790 					gap++;
791 				} else if (gap != 0)
792 					imo->imo_membership[i - gap] =
793 					    imo->imo_membership[i];
794 			}
795 			imo->imo_num_memberships -= gap;
796 		}
797 	}
798 }
799 
800 /*
801  * Check for alternatives when higher level complains
802  * about service problems.  For now, invalidate cached
803  * routing information.  If the route was created dynamically
804  * (by a redirect), time to try a default gateway again.
805  */
806 void
807 in_losing(struct inpcb *inp)
808 {
809 	struct rtentry *rt;
810 	struct rt_addrinfo rtinfo;
811 
812 	if ((rt = inp->inp_route.ro_rt)) {
813 		bzero(&rtinfo, sizeof(struct rt_addrinfo));
814 		rtinfo.rti_info[RTAX_DST] = rt_key(rt);
815 		rtinfo.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
816 		rtinfo.rti_info[RTAX_NETMASK] = rt_mask(rt);
817 		rtinfo.rti_flags = rt->rt_flags;
818 		rt_missmsg(RTM_LOSING, &rtinfo, rt->rt_flags, 0);
819 		if (rt->rt_flags & RTF_DYNAMIC)
820 			rtrequest1_global(RTM_DELETE, &rtinfo, NULL, NULL);
821 		inp->inp_route.ro_rt = NULL;
822 		rtfree(rt);
823 		/*
824 		 * A new route can be allocated
825 		 * the next time output is attempted.
826 		 */
827 	}
828 }
829 
830 /*
831  * After a routing change, flush old routing
832  * and allocate a (hopefully) better one.
833  */
834 void
835 in_rtchange(struct inpcb *inp, int err)
836 {
837 	if (inp->inp_route.ro_rt) {
838 		rtfree(inp->inp_route.ro_rt);
839 		inp->inp_route.ro_rt = NULL;
840 		/*
841 		 * A new route can be allocated the next time
842 		 * output is attempted.
843 		 */
844 	}
845 }
846 
847 /*
848  * Lookup a PCB based on the local address and port.
849  */
850 struct inpcb *
851 in_pcblookup_local(struct inpcbinfo *pcbinfo, struct in_addr laddr,
852 		   u_int lport_arg, int wild_okay, struct ucred *cred)
853 {
854 	struct inpcb *inp;
855 	int matchwild = 3, wildcard;
856 	u_short lport = lport_arg;
857 
858 	struct inpcbporthead *porthash;
859 	struct inpcbport *phd;
860 	struct inpcb *match = NULL;
861 
862 	/*
863 	 * Best fit PCB lookup.
864 	 *
865 	 * First see if this local port is in use by looking on the
866 	 * port hash list.
867 	 */
868 	porthash = &pcbinfo->porthashbase[INP_PCBPORTHASH(lport,
869 	    pcbinfo->porthashmask)];
870 	LIST_FOREACH(phd, porthash, phd_hash) {
871 		if (phd->phd_port == lport)
872 			break;
873 	}
874 	if (phd != NULL) {
875 		/*
876 		 * Port is in use by one or more PCBs. Look for best
877 		 * fit.
878 		 */
879 		LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) {
880 			wildcard = 0;
881 #ifdef INET6
882 			if ((inp->inp_vflag & INP_IPV4) == 0)
883 				continue;
884 #endif
885 			if (inp->inp_faddr.s_addr != INADDR_ANY)
886 				wildcard++;
887 			if (inp->inp_laddr.s_addr != INADDR_ANY) {
888 				if (laddr.s_addr == INADDR_ANY)
889 					wildcard++;
890 				else if (inp->inp_laddr.s_addr != laddr.s_addr)
891 					continue;
892 			} else {
893 				if (laddr.s_addr != INADDR_ANY)
894 					wildcard++;
895 			}
896 			if (wildcard && !wild_okay)
897 				continue;
898 			if (wildcard < matchwild &&
899 			    (cred == NULL ||
900 			     cred->cr_prison ==
901 					inp->inp_socket->so_cred->cr_prison)) {
902 				match = inp;
903 				matchwild = wildcard;
904 				if (matchwild == 0) {
905 					break;
906 				}
907 			}
908 		}
909 	}
910 	return (match);
911 }
912 
913 /*
914  * Lookup PCB in hash list.
915  */
916 struct inpcb *
917 in_pcblookup_hash(struct inpcbinfo *pcbinfo, struct in_addr faddr, u_int fport_arg,
918 		  struct in_addr laddr, u_int lport_arg, boolean_t wildcard,
919 		  struct ifnet *ifp)
920 {
921 	struct inpcbhead *head;
922 	struct inpcb *inp, *jinp=NULL;
923 	u_short fport = fport_arg, lport = lport_arg;
924 
925 	/*
926 	 * First look for an exact match.
927 	 */
928 	head = &pcbinfo->hashbase[INP_PCBCONNHASH(faddr.s_addr, fport,
929 	    laddr.s_addr, lport, pcbinfo->hashmask)];
930 	LIST_FOREACH(inp, head, inp_hash) {
931 #ifdef INET6
932 		if (!(inp->inp_vflag & INP_IPV4))
933 			continue;
934 #endif
935 		if (in_hosteq(inp->inp_faddr, faddr) &&
936 		    in_hosteq(inp->inp_laddr, laddr) &&
937 		    inp->inp_fport == fport && inp->inp_lport == lport) {
938 			/* found */
939 			if (inp->inp_socket == NULL ||
940 			    inp->inp_socket->so_cred->cr_prison == NULL) {
941 				return (inp);
942 			} else {
943 				if  (jinp == NULL)
944 					jinp = inp;
945 			}
946 		}
947 	}
948 	if (jinp != NULL)
949 		return (jinp);
950 	if (wildcard) {
951 		struct inpcb *local_wild = NULL;
952 		struct inpcb *jinp_wild = NULL;
953 #ifdef INET6
954 		struct inpcb *local_wild_mapped = NULL;
955 #endif
956 		struct inpcontainer *ic;
957 		struct inpcontainerhead *chead;
958 		struct sockaddr_in jsin;
959 		struct ucred *cred;
960 
961 		/*
962 		 * Order of socket selection:
963 		 * 1. non-jailed, non-wild.
964 		 * 2. non-jailed, wild.
965 		 * 3. jailed, non-wild.
966 		 * 4. jailed, wild.
967 		 */
968 		jsin.sin_family = AF_INET;
969 		chead = &pcbinfo->wildcardhashbase[
970 		    INP_PCBWILDCARDHASH(lport, pcbinfo->wildcardhashmask)];
971 		LIST_FOREACH(ic, chead, ic_list) {
972 			inp = ic->ic_inp;
973 			jsin.sin_addr.s_addr = laddr.s_addr;
974 #ifdef INET6
975 			if (!(inp->inp_vflag & INP_IPV4))
976 				continue;
977 #endif
978 			if (inp->inp_socket != NULL)
979 				cred = inp->inp_socket->so_cred;
980 			else
981 				cred = NULL;
982 			if (cred != NULL && jailed(cred)) {
983 				if (jinp != NULL)
984 					continue;
985 				else
986 					if (!jailed_ip(cred->cr_prison,
987 					    (struct sockaddr *)&jsin))
988 						continue;
989 			}
990 			if (inp->inp_lport == lport) {
991 				if (ifp && ifp->if_type == IFT_FAITH &&
992 				    !(inp->inp_flags & INP_FAITH))
993 					continue;
994 				if (inp->inp_laddr.s_addr == laddr.s_addr) {
995 					if (cred != NULL && jailed(cred))
996 						jinp = inp;
997 					else
998 						return (inp);
999 				}
1000 				if (inp->inp_laddr.s_addr == INADDR_ANY) {
1001 #ifdef INET6
1002 					if (INP_CHECK_SOCKAF(inp->inp_socket,
1003 							     AF_INET6))
1004 						local_wild_mapped = inp;
1005 					else
1006 #endif
1007 						if (cred != NULL &&
1008 						    jailed(cred))
1009 							jinp_wild = inp;
1010 						else
1011 							local_wild = inp;
1012 				}
1013 			}
1014 		}
1015 		if (local_wild != NULL)
1016 			return (local_wild);
1017 #ifdef INET6
1018 		if (local_wild_mapped != NULL)
1019 			return (local_wild_mapped);
1020 #endif
1021 		if (jinp != NULL)
1022 			return (jinp);
1023 		return (jinp_wild);
1024 	}
1025 
1026 	/*
1027 	 * Not found.
1028 	 */
1029 	return (NULL);
1030 }
1031 
1032 /*
1033  * Insert PCB into connection hash table.
1034  */
1035 void
1036 in_pcbinsconnhash(struct inpcb *inp)
1037 {
1038 	struct inpcbinfo *pcbinfo = inp->inp_cpcbinfo;
1039 	struct inpcbhead *bucket;
1040 	u_int32_t hashkey_faddr, hashkey_laddr;
1041 
1042 #ifdef INET6
1043 	if (inp->inp_vflag & INP_IPV6) {
1044 		hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX JH */;
1045 		hashkey_laddr = inp->in6p_laddr.s6_addr32[3] /* XXX JH */;
1046 	} else {
1047 #endif
1048 		hashkey_faddr = inp->inp_faddr.s_addr;
1049 		hashkey_laddr = inp->inp_laddr.s_addr;
1050 #ifdef INET6
1051 	}
1052 #endif
1053 
1054 	KASSERT(!(inp->inp_flags & INP_CONNECTED), ("already on hash list"));
1055 	inp->inp_flags |= INP_CONNECTED;
1056 
1057 	/*
1058 	 * Insert into the connection hash table.
1059 	 */
1060 	bucket = &pcbinfo->hashbase[INP_PCBCONNHASH(hashkey_faddr,
1061 	    inp->inp_fport, hashkey_laddr, inp->inp_lport, pcbinfo->hashmask)];
1062 	LIST_INSERT_HEAD(bucket, inp, inp_hash);
1063 }
1064 
1065 /*
1066  * Remove PCB from connection hash table.
1067  */
1068 void
1069 in_pcbremconnhash(struct inpcb *inp)
1070 {
1071 	KASSERT(inp->inp_flags & INP_CONNECTED, ("inp not connected"));
1072 	LIST_REMOVE(inp, inp_hash);
1073 	inp->inp_flags &= ~INP_CONNECTED;
1074 }
1075 
1076 /*
1077  * Insert PCB into port hash table.
1078  */
1079 int
1080 in_pcbinsporthash(struct inpcb *inp)
1081 {
1082 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1083 	struct inpcbporthead *pcbporthash;
1084 	struct inpcbport *phd;
1085 
1086 	/*
1087 	 * Insert into the port hash table.
1088 	 */
1089 	pcbporthash = &pcbinfo->porthashbase[
1090 	    INP_PCBPORTHASH(inp->inp_lport, pcbinfo->porthashmask)];
1091 
1092 	/* Go through port list and look for a head for this lport. */
1093 	LIST_FOREACH(phd, pcbporthash, phd_hash)
1094 		if (phd->phd_port == inp->inp_lport)
1095 			break;
1096 
1097 	/* If none exists, malloc one and tack it on. */
1098 	if (phd == NULL) {
1099 		MALLOC(phd, struct inpcbport *, sizeof(struct inpcbport),
1100 		    M_PCB, M_INTWAIT | M_NULLOK);
1101 		if (phd == NULL)
1102 			return (ENOBUFS); /* XXX */
1103 		phd->phd_port = inp->inp_lport;
1104 		LIST_INIT(&phd->phd_pcblist);
1105 		LIST_INSERT_HEAD(pcbporthash, phd, phd_hash);
1106 	}
1107 
1108 	inp->inp_phd = phd;
1109 	LIST_INSERT_HEAD(&phd->phd_pcblist, inp, inp_portlist);
1110 
1111 	return (0);
1112 }
1113 
1114 void
1115 in_pcbinswildcardhash_oncpu(struct inpcb *inp, struct inpcbinfo *pcbinfo)
1116 {
1117 	struct inpcontainer *ic;
1118 	struct inpcontainerhead *bucket;
1119 
1120 	bucket = &pcbinfo->wildcardhashbase[
1121 	    INP_PCBWILDCARDHASH(inp->inp_lport, pcbinfo->wildcardhashmask)];
1122 
1123 	ic = kmalloc(sizeof(struct inpcontainer), M_TEMP, M_INTWAIT);
1124 	ic->ic_inp = inp;
1125 	LIST_INSERT_HEAD(bucket, ic, ic_list);
1126 }
1127 
1128 /*
1129  * Insert PCB into wildcard hash table.
1130  */
1131 void
1132 in_pcbinswildcardhash(struct inpcb *inp)
1133 {
1134 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1135 
1136 	KKASSERT(pcbinfo != NULL);
1137 
1138 	in_pcbinswildcardhash_oncpu(inp, pcbinfo);
1139 	inp->inp_flags |= INP_WILDCARD;
1140 }
1141 
1142 void
1143 in_pcbremwildcardhash_oncpu(struct inpcb *inp, struct inpcbinfo *pcbinfo)
1144 {
1145 	struct inpcontainer *ic;
1146 	struct inpcontainerhead *head;
1147 
1148 	/* find bucket */
1149 	head = &pcbinfo->wildcardhashbase[
1150 	    INP_PCBWILDCARDHASH(inp->inp_lport, pcbinfo->wildcardhashmask)];
1151 
1152 	LIST_FOREACH(ic, head, ic_list) {
1153 		if (ic->ic_inp == inp)
1154 			goto found;
1155 	}
1156 	return;			/* not found! */
1157 
1158 found:
1159 	LIST_REMOVE(ic, ic_list);	/* remove container from bucket chain */
1160 	kfree(ic, M_TEMP);		/* deallocate container */
1161 }
1162 
1163 /*
1164  * Remove PCB from wildcard hash table.
1165  */
1166 void
1167 in_pcbremwildcardhash(struct inpcb *inp)
1168 {
1169 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1170 
1171 	KASSERT(inp->inp_flags & INP_WILDCARD, ("inp not wildcard"));
1172 	in_pcbremwildcardhash_oncpu(inp, pcbinfo);
1173 	inp->inp_flags &= ~INP_WILDCARD;
1174 }
1175 
1176 /*
1177  * Remove PCB from various lists.
1178  */
1179 void
1180 in_pcbremlists(struct inpcb *inp)
1181 {
1182 	if (inp->inp_lport) {
1183 		struct inpcbport *phd = inp->inp_phd;
1184 
1185 		LIST_REMOVE(inp, inp_portlist);
1186 		if (LIST_FIRST(&phd->phd_pcblist) == NULL) {
1187 			LIST_REMOVE(phd, phd_hash);
1188 			kfree(phd, M_PCB);
1189 		}
1190 	}
1191 	if (inp->inp_flags & INP_WILDCARD) {
1192 		in_pcbremwildcardhash(inp);
1193 	} else if (inp->inp_flags & INP_CONNECTED) {
1194 		in_pcbremconnhash(inp);
1195 	}
1196 	LIST_REMOVE(inp, inp_list);
1197 	inp->inp_pcbinfo->ipi_count--;
1198 }
1199 
1200 int
1201 prison_xinpcb(struct thread *td, struct inpcb *inp)
1202 {
1203 	struct ucred *cr;
1204 
1205 	if (td->td_proc == NULL)
1206 		return (0);
1207 	cr = td->td_proc->p_ucred;
1208 	if (cr->cr_prison == NULL)
1209 		return (0);
1210 	if (inp->inp_socket && inp->inp_socket->so_cred &&
1211 	    inp->inp_socket->so_cred->cr_prison &&
1212 	    cr->cr_prison == inp->inp_socket->so_cred->cr_prison)
1213 		return (0);
1214 	return (1);
1215 }
1216 
1217 int
1218 in_pcblist_global(SYSCTL_HANDLER_ARGS)
1219 {
1220 	struct inpcbinfo *pcbinfo = arg1;
1221 	struct inpcb *inp, *marker;
1222 	struct xinpcb xi;
1223 	int error, i, n;
1224 	inp_gen_t gencnt;
1225 
1226 	/*
1227 	 * The process of preparing the TCB list is too time-consuming and
1228 	 * resource-intensive to repeat twice on every request.
1229 	 */
1230 	if (req->oldptr == NULL) {
1231 		n = pcbinfo->ipi_count;
1232 		req->oldidx = (n + n/8 + 10) * sizeof(struct xinpcb);
1233 		return 0;
1234 	}
1235 
1236 	if (req->newptr != NULL)
1237 		return EPERM;
1238 
1239 	/*
1240 	 * OK, now we're committed to doing something.  Re-fetch ipi_count
1241 	 * after obtaining the generation count.
1242 	 */
1243 	gencnt = pcbinfo->ipi_gencnt;
1244 	n = pcbinfo->ipi_count;
1245 
1246 	marker = kmalloc(sizeof(struct inpcb), M_TEMP, M_WAITOK|M_ZERO);
1247 	marker->inp_flags |= INP_PLACEMARKER;
1248 	LIST_INSERT_HEAD(&pcbinfo->pcblisthead, marker, inp_list);
1249 
1250 	i = 0;
1251 	error = 0;
1252 
1253 	while ((inp = LIST_NEXT(marker, inp_list)) != NULL && i < n) {
1254 		LIST_REMOVE(marker, inp_list);
1255 		LIST_INSERT_AFTER(inp, marker, inp_list);
1256 
1257 		if (inp->inp_flags & INP_PLACEMARKER)
1258 			continue;
1259 		if (inp->inp_gencnt > gencnt)
1260 			continue;
1261 		if (prison_xinpcb(req->td, inp))
1262 			continue;
1263 		bzero(&xi, sizeof xi);
1264 		xi.xi_len = sizeof xi;
1265 		bcopy(inp, &xi.xi_inp, sizeof *inp);
1266 		if (inp->inp_socket)
1267 			sotoxsocket(inp->inp_socket, &xi.xi_socket);
1268 		if ((error = SYSCTL_OUT(req, &xi, sizeof xi)) != 0)
1269 			break;
1270 		++i;
1271 	}
1272 	LIST_REMOVE(marker, inp_list);
1273 	if (error == 0 && i < n) {
1274 		bzero(&xi, sizeof xi);
1275 		xi.xi_len = sizeof xi;
1276 		while (i < n) {
1277 			error = SYSCTL_OUT(req, &xi, sizeof xi);
1278 			++i;
1279 		}
1280 	}
1281 	kfree(marker, M_TEMP);
1282 	return(error);
1283 }
1284