xref: /freebsd/sys/netinet/in_pcb.c (revision 9336e0699bda8a301cd2bfa37106b6ec5e32012e)
1 /*-
2  * Copyright (c) 1982, 1986, 1991, 1993, 1995
3  *	The Regents of the University of California.
4  * Copyright (c) 2007 Robert N. M. Watson
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 4. Neither the name of the University nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	@(#)in_pcb.c	8.4 (Berkeley) 5/24/95
32  */
33 
34 #include <sys/cdefs.h>
35 __FBSDID("$FreeBSD$");
36 
37 #include "opt_ddb.h"
38 #include "opt_ipsec.h"
39 #include "opt_inet6.h"
40 #include "opt_mac.h"
41 
42 #include <sys/param.h>
43 #include <sys/systm.h>
44 #include <sys/malloc.h>
45 #include <sys/mbuf.h>
46 #include <sys/domain.h>
47 #include <sys/protosw.h>
48 #include <sys/socket.h>
49 #include <sys/socketvar.h>
50 #include <sys/priv.h>
51 #include <sys/proc.h>
52 #include <sys/jail.h>
53 #include <sys/kernel.h>
54 #include <sys/sysctl.h>
55 
56 #ifdef DDB
57 #include <ddb/ddb.h>
58 #endif
59 
60 #include <vm/uma.h>
61 
62 #include <net/if.h>
63 #include <net/if_types.h>
64 #include <net/route.h>
65 
66 #include <netinet/in.h>
67 #include <netinet/in_pcb.h>
68 #include <netinet/in_var.h>
69 #include <netinet/ip_var.h>
70 #include <netinet/tcp_var.h>
71 #include <netinet/udp.h>
72 #include <netinet/udp_var.h>
73 #ifdef INET6
74 #include <netinet/ip6.h>
75 #include <netinet6/ip6_var.h>
76 #endif /* INET6 */
77 
78 
79 #ifdef IPSEC
80 #include <netipsec/ipsec.h>
81 #include <netipsec/key.h>
82 #endif /* IPSEC */
83 
84 #include <security/mac/mac_framework.h>
85 
86 /*
87  * These configure the range of local port addresses assigned to
88  * "unspecified" outgoing connections/packets/whatever.
89  */
90 int	ipport_lowfirstauto  = IPPORT_RESERVED - 1;	/* 1023 */
91 int	ipport_lowlastauto = IPPORT_RESERVEDSTART;	/* 600 */
92 int	ipport_firstauto = IPPORT_HIFIRSTAUTO;		/* 49152 */
93 int	ipport_lastauto  = IPPORT_HILASTAUTO;		/* 65535 */
94 int	ipport_hifirstauto = IPPORT_HIFIRSTAUTO;	/* 49152 */
95 int	ipport_hilastauto  = IPPORT_HILASTAUTO;		/* 65535 */
96 
97 /*
98  * Reserved ports accessible only to root. There are significant
99  * security considerations that must be accounted for when changing these,
100  * but the security benefits can be great. Please be careful.
101  */
102 int	ipport_reservedhigh = IPPORT_RESERVED - 1;	/* 1023 */
103 int	ipport_reservedlow = 0;
104 
105 /* Variables dealing with random ephemeral port allocation. */
106 int	ipport_randomized = 1;	/* user controlled via sysctl */
107 int	ipport_randomcps = 10;	/* user controlled via sysctl */
108 int	ipport_randomtime = 45;	/* user controlled via sysctl */
109 int	ipport_stoprandom = 0;	/* toggled by ipport_tick */
110 int	ipport_tcpallocs;
111 int	ipport_tcplastcount;
112 
113 #define RANGECHK(var, min, max) \
114 	if ((var) < (min)) { (var) = (min); } \
115 	else if ((var) > (max)) { (var) = (max); }
116 
117 static int
118 sysctl_net_ipport_check(SYSCTL_HANDLER_ARGS)
119 {
120 	int error;
121 
122 	error = sysctl_handle_int(oidp, oidp->oid_arg1, oidp->oid_arg2, req);
123 	if (error == 0) {
124 		RANGECHK(ipport_lowfirstauto, 1, IPPORT_RESERVED - 1);
125 		RANGECHK(ipport_lowlastauto, 1, IPPORT_RESERVED - 1);
126 		RANGECHK(ipport_firstauto, IPPORT_RESERVED, IPPORT_MAX);
127 		RANGECHK(ipport_lastauto, IPPORT_RESERVED, IPPORT_MAX);
128 		RANGECHK(ipport_hifirstauto, IPPORT_RESERVED, IPPORT_MAX);
129 		RANGECHK(ipport_hilastauto, IPPORT_RESERVED, IPPORT_MAX);
130 	}
131 	return (error);
132 }
133 
134 #undef RANGECHK
135 
136 SYSCTL_NODE(_net_inet_ip, IPPROTO_IP, portrange, CTLFLAG_RW, 0, "IP Ports");
137 
138 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, lowfirst, CTLTYPE_INT|CTLFLAG_RW,
139 	   &ipport_lowfirstauto, 0, &sysctl_net_ipport_check, "I", "");
140 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, lowlast, CTLTYPE_INT|CTLFLAG_RW,
141 	   &ipport_lowlastauto, 0, &sysctl_net_ipport_check, "I", "");
142 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, first, CTLTYPE_INT|CTLFLAG_RW,
143 	   &ipport_firstauto, 0, &sysctl_net_ipport_check, "I", "");
144 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, last, CTLTYPE_INT|CTLFLAG_RW,
145 	   &ipport_lastauto, 0, &sysctl_net_ipport_check, "I", "");
146 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, hifirst, CTLTYPE_INT|CTLFLAG_RW,
147 	   &ipport_hifirstauto, 0, &sysctl_net_ipport_check, "I", "");
148 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, hilast, CTLTYPE_INT|CTLFLAG_RW,
149 	   &ipport_hilastauto, 0, &sysctl_net_ipport_check, "I", "");
150 SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, reservedhigh,
151 	   CTLFLAG_RW|CTLFLAG_SECURE, &ipport_reservedhigh, 0, "");
152 SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, reservedlow,
153 	   CTLFLAG_RW|CTLFLAG_SECURE, &ipport_reservedlow, 0, "");
154 SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, randomized, CTLFLAG_RW,
155 	   &ipport_randomized, 0, "Enable random port allocation");
156 SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, randomcps, CTLFLAG_RW,
157 	   &ipport_randomcps, 0, "Maximum number of random port "
158 	   "allocations before switching to a sequental one");
159 SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, randomtime, CTLFLAG_RW,
160 	   &ipport_randomtime, 0, "Minimum time to keep sequental port "
161 	   "allocation before switching to a random one");
162 
163 /*
164  * in_pcb.c: manage the Protocol Control Blocks.
165  *
166  * NOTE: It is assumed that most of these functions will be called with
167  * the pcbinfo lock held, and often, the inpcb lock held, as these utility
168  * functions often modify hash chains or addresses in pcbs.
169  */
170 
171 /*
172  * Allocate a PCB and associate it with the socket.
173  * On success return with the PCB locked.
174  */
175 int
176 in_pcballoc(struct socket *so, struct inpcbinfo *pcbinfo)
177 {
178 	struct inpcb *inp;
179 	int error;
180 
181 	INP_INFO_WLOCK_ASSERT(pcbinfo);
182 	error = 0;
183 	inp = uma_zalloc(pcbinfo->ipi_zone, M_NOWAIT);
184 	if (inp == NULL)
185 		return (ENOBUFS);
186 	bzero(inp, inp_zero_size);
187 	inp->inp_pcbinfo = pcbinfo;
188 	inp->inp_socket = so;
189 #ifdef MAC
190 	error = mac_inpcb_init(inp, M_NOWAIT);
191 	if (error != 0)
192 		goto out;
193 	SOCK_LOCK(so);
194 	mac_inpcb_create(so, inp);
195 	SOCK_UNLOCK(so);
196 #endif
197 
198 #ifdef IPSEC
199 	error = ipsec_init_policy(so, &inp->inp_sp);
200 	if (error != 0) {
201 #ifdef MAC
202 		mac_inpcb_destroy(inp);
203 #endif
204 		goto out;
205 }
206 #endif /*IPSEC*/
207 #ifdef INET6
208 	if (INP_SOCKAF(so) == AF_INET6) {
209 		inp->inp_vflag |= INP_IPV6PROTO;
210 		if (ip6_v6only)
211 			inp->inp_flags |= IN6P_IPV6_V6ONLY;
212 	}
213 #endif
214 	LIST_INSERT_HEAD(pcbinfo->ipi_listhead, inp, inp_list);
215 	pcbinfo->ipi_count++;
216 	so->so_pcb = (caddr_t)inp;
217 #ifdef INET6
218 	if (ip6_auto_flowlabel)
219 		inp->inp_flags |= IN6P_AUTOFLOWLABEL;
220 #endif
221 	INP_LOCK(inp);
222 	inp->inp_gencnt = ++pcbinfo->ipi_gencnt;
223 
224 #if defined(IPSEC) || defined(MAC)
225 out:
226 	if (error != 0)
227 		uma_zfree(pcbinfo->ipi_zone, inp);
228 #endif
229 	return (error);
230 }
231 
232 int
233 in_pcbbind(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred)
234 {
235 	int anonport, error;
236 
237 	INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
238 	INP_LOCK_ASSERT(inp);
239 
240 	if (inp->inp_lport != 0 || inp->inp_laddr.s_addr != INADDR_ANY)
241 		return (EINVAL);
242 	anonport = inp->inp_lport == 0 && (nam == NULL ||
243 	    ((struct sockaddr_in *)nam)->sin_port == 0);
244 	error = in_pcbbind_setup(inp, nam, &inp->inp_laddr.s_addr,
245 	    &inp->inp_lport, cred);
246 	if (error)
247 		return (error);
248 	if (in_pcbinshash(inp) != 0) {
249 		inp->inp_laddr.s_addr = INADDR_ANY;
250 		inp->inp_lport = 0;
251 		return (EAGAIN);
252 	}
253 	if (anonport)
254 		inp->inp_flags |= INP_ANONPORT;
255 	return (0);
256 }
257 
258 /*
259  * Set up a bind operation on a PCB, performing port allocation
260  * as required, but do not actually modify the PCB. Callers can
261  * either complete the bind by setting inp_laddr/inp_lport and
262  * calling in_pcbinshash(), or they can just use the resulting
263  * port and address to authorise the sending of a once-off packet.
264  *
265  * On error, the values of *laddrp and *lportp are not changed.
266  */
267 int
268 in_pcbbind_setup(struct inpcb *inp, struct sockaddr *nam, in_addr_t *laddrp,
269     u_short *lportp, struct ucred *cred)
270 {
271 	struct socket *so = inp->inp_socket;
272 	unsigned short *lastport;
273 	struct sockaddr_in *sin;
274 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
275 	struct in_addr laddr;
276 	u_short lport = 0;
277 	int wild = 0, reuseport = (so->so_options & SO_REUSEPORT);
278 	int error, prison = 0;
279 	int dorandom;
280 
281 	INP_INFO_WLOCK_ASSERT(pcbinfo);
282 	INP_LOCK_ASSERT(inp);
283 
284 	if (TAILQ_EMPTY(&in_ifaddrhead)) /* XXX broken! */
285 		return (EADDRNOTAVAIL);
286 	laddr.s_addr = *laddrp;
287 	if (nam != NULL && laddr.s_addr != INADDR_ANY)
288 		return (EINVAL);
289 	if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0)
290 		wild = INPLOOKUP_WILDCARD;
291 	if (nam) {
292 		sin = (struct sockaddr_in *)nam;
293 		if (nam->sa_len != sizeof (*sin))
294 			return (EINVAL);
295 #ifdef notdef
296 		/*
297 		 * We should check the family, but old programs
298 		 * incorrectly fail to initialize it.
299 		 */
300 		if (sin->sin_family != AF_INET)
301 			return (EAFNOSUPPORT);
302 #endif
303 		if (sin->sin_addr.s_addr != INADDR_ANY)
304 			if (prison_ip(cred, 0, &sin->sin_addr.s_addr))
305 				return(EINVAL);
306 		if (sin->sin_port != *lportp) {
307 			/* Don't allow the port to change. */
308 			if (*lportp != 0)
309 				return (EINVAL);
310 			lport = sin->sin_port;
311 		}
312 		/* NB: lport is left as 0 if the port isn't being changed. */
313 		if (IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) {
314 			/*
315 			 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
316 			 * allow complete duplication of binding if
317 			 * SO_REUSEPORT is set, or if SO_REUSEADDR is set
318 			 * and a multicast address is bound on both
319 			 * new and duplicated sockets.
320 			 */
321 			if (so->so_options & SO_REUSEADDR)
322 				reuseport = SO_REUSEADDR|SO_REUSEPORT;
323 		} else if (sin->sin_addr.s_addr != INADDR_ANY) {
324 			sin->sin_port = 0;		/* yech... */
325 			bzero(&sin->sin_zero, sizeof(sin->sin_zero));
326 			if (ifa_ifwithaddr((struct sockaddr *)sin) == 0)
327 				return (EADDRNOTAVAIL);
328 		}
329 		laddr = sin->sin_addr;
330 		if (lport) {
331 			struct inpcb *t;
332 			struct tcptw *tw;
333 
334 			/* GROSS */
335 			if (ntohs(lport) <= ipport_reservedhigh &&
336 			    ntohs(lport) >= ipport_reservedlow &&
337 			    priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT,
338 			    0))
339 				return (EACCES);
340 			if (jailed(cred))
341 				prison = 1;
342 			if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)) &&
343 			    priv_check_cred(so->so_cred,
344 			    PRIV_NETINET_REUSEPORT, 0) != 0) {
345 				t = in_pcblookup_local(inp->inp_pcbinfo,
346 				    sin->sin_addr, lport,
347 				    prison ? 0 :  INPLOOKUP_WILDCARD);
348 	/*
349 	 * XXX
350 	 * This entire block sorely needs a rewrite.
351 	 */
352 				if (t &&
353 				    ((t->inp_vflag & INP_TIMEWAIT) == 0) &&
354 				    (so->so_type != SOCK_STREAM ||
355 				     ntohl(t->inp_faddr.s_addr) == INADDR_ANY) &&
356 				    (ntohl(sin->sin_addr.s_addr) != INADDR_ANY ||
357 				     ntohl(t->inp_laddr.s_addr) != INADDR_ANY ||
358 				     (t->inp_socket->so_options &
359 					 SO_REUSEPORT) == 0) &&
360 				    (so->so_cred->cr_uid !=
361 				     t->inp_socket->so_cred->cr_uid))
362 					return (EADDRINUSE);
363 			}
364 			if (prison && prison_ip(cred, 0, &sin->sin_addr.s_addr))
365 				return (EADDRNOTAVAIL);
366 			t = in_pcblookup_local(pcbinfo, sin->sin_addr,
367 			    lport, prison ? 0 : wild);
368 			if (t && (t->inp_vflag & INP_TIMEWAIT)) {
369 				/*
370 				 * XXXRW: If an incpb has had its timewait
371 				 * state recycled, we treat the address as
372 				 * being in use (for now).  This is better
373 				 * than a panic, but not desirable.
374 				 */
375 				tw = intotw(inp);
376 				if (tw == NULL ||
377 				    (reuseport & tw->tw_so_options) == 0)
378 					return (EADDRINUSE);
379 			} else if (t &&
380 			    (reuseport & t->inp_socket->so_options) == 0) {
381 #ifdef INET6
382 				if (ntohl(sin->sin_addr.s_addr) !=
383 				    INADDR_ANY ||
384 				    ntohl(t->inp_laddr.s_addr) !=
385 				    INADDR_ANY ||
386 				    INP_SOCKAF(so) ==
387 				    INP_SOCKAF(t->inp_socket))
388 #endif
389 				return (EADDRINUSE);
390 			}
391 		}
392 	}
393 	if (*lportp != 0)
394 		lport = *lportp;
395 	if (lport == 0) {
396 		u_short first, last;
397 		int count;
398 
399 		if (laddr.s_addr != INADDR_ANY)
400 			if (prison_ip(cred, 0, &laddr.s_addr))
401 				return (EINVAL);
402 
403 		if (inp->inp_flags & INP_HIGHPORT) {
404 			first = ipport_hifirstauto;	/* sysctl */
405 			last  = ipport_hilastauto;
406 			lastport = &pcbinfo->ipi_lasthi;
407 		} else if (inp->inp_flags & INP_LOWPORT) {
408 			error = priv_check_cred(cred,
409 			    PRIV_NETINET_RESERVEDPORT, 0);
410 			if (error)
411 				return error;
412 			first = ipport_lowfirstauto;	/* 1023 */
413 			last  = ipport_lowlastauto;	/* 600 */
414 			lastport = &pcbinfo->ipi_lastlow;
415 		} else {
416 			first = ipport_firstauto;	/* sysctl */
417 			last  = ipport_lastauto;
418 			lastport = &pcbinfo->ipi_lastport;
419 		}
420 		/*
421 		 * For UDP, use random port allocation as long as the user
422 		 * allows it.  For TCP (and as of yet unknown) connections,
423 		 * use random port allocation only if the user allows it AND
424 		 * ipport_tick() allows it.
425 		 */
426 		if (ipport_randomized &&
427 			(!ipport_stoprandom || pcbinfo == &udbinfo))
428 			dorandom = 1;
429 		else
430 			dorandom = 0;
431 		/*
432 		 * It makes no sense to do random port allocation if
433 		 * we have the only port available.
434 		 */
435 		if (first == last)
436 			dorandom = 0;
437 		/* Make sure to not include UDP packets in the count. */
438 		if (pcbinfo != &udbinfo)
439 			ipport_tcpallocs++;
440 		/*
441 		 * Simple check to ensure all ports are not used up causing
442 		 * a deadlock here.
443 		 *
444 		 * We split the two cases (up and down) so that the direction
445 		 * is not being tested on each round of the loop.
446 		 */
447 		if (first > last) {
448 			/*
449 			 * counting down
450 			 */
451 			if (dorandom)
452 				*lastport = first -
453 					    (arc4random() % (first - last));
454 			count = first - last;
455 
456 			do {
457 				if (count-- < 0)	/* completely used? */
458 					return (EADDRNOTAVAIL);
459 				--*lastport;
460 				if (*lastport > first || *lastport < last)
461 					*lastport = first;
462 				lport = htons(*lastport);
463 			} while (in_pcblookup_local(pcbinfo, laddr, lport,
464 			    wild));
465 		} else {
466 			/*
467 			 * counting up
468 			 */
469 			if (dorandom)
470 				*lastport = first +
471 					    (arc4random() % (last - first));
472 			count = last - first;
473 
474 			do {
475 				if (count-- < 0)	/* completely used? */
476 					return (EADDRNOTAVAIL);
477 				++*lastport;
478 				if (*lastport < first || *lastport > last)
479 					*lastport = first;
480 				lport = htons(*lastport);
481 			} while (in_pcblookup_local(pcbinfo, laddr, lport,
482 			    wild));
483 		}
484 	}
485 	if (prison_ip(cred, 0, &laddr.s_addr))
486 		return (EINVAL);
487 	*laddrp = laddr.s_addr;
488 	*lportp = lport;
489 	return (0);
490 }
491 
492 /*
493  * Connect from a socket to a specified address.
494  * Both address and port must be specified in argument sin.
495  * If don't have a local address for this socket yet,
496  * then pick one.
497  */
498 int
499 in_pcbconnect(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred)
500 {
501 	u_short lport, fport;
502 	in_addr_t laddr, faddr;
503 	int anonport, error;
504 
505 	INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
506 	INP_LOCK_ASSERT(inp);
507 
508 	lport = inp->inp_lport;
509 	laddr = inp->inp_laddr.s_addr;
510 	anonport = (lport == 0);
511 	error = in_pcbconnect_setup(inp, nam, &laddr, &lport, &faddr, &fport,
512 	    NULL, cred);
513 	if (error)
514 		return (error);
515 
516 	/* Do the initial binding of the local address if required. */
517 	if (inp->inp_laddr.s_addr == INADDR_ANY && inp->inp_lport == 0) {
518 		inp->inp_lport = lport;
519 		inp->inp_laddr.s_addr = laddr;
520 		if (in_pcbinshash(inp) != 0) {
521 			inp->inp_laddr.s_addr = INADDR_ANY;
522 			inp->inp_lport = 0;
523 			return (EAGAIN);
524 		}
525 	}
526 
527 	/* Commit the remaining changes. */
528 	inp->inp_lport = lport;
529 	inp->inp_laddr.s_addr = laddr;
530 	inp->inp_faddr.s_addr = faddr;
531 	inp->inp_fport = fport;
532 	in_pcbrehash(inp);
533 
534 	if (anonport)
535 		inp->inp_flags |= INP_ANONPORT;
536 	return (0);
537 }
538 
539 /*
540  * Set up for a connect from a socket to the specified address.
541  * On entry, *laddrp and *lportp should contain the current local
542  * address and port for the PCB; these are updated to the values
543  * that should be placed in inp_laddr and inp_lport to complete
544  * the connect.
545  *
546  * On success, *faddrp and *fportp will be set to the remote address
547  * and port. These are not updated in the error case.
548  *
549  * If the operation fails because the connection already exists,
550  * *oinpp will be set to the PCB of that connection so that the
551  * caller can decide to override it. In all other cases, *oinpp
552  * is set to NULL.
553  */
554 int
555 in_pcbconnect_setup(struct inpcb *inp, struct sockaddr *nam,
556     in_addr_t *laddrp, u_short *lportp, in_addr_t *faddrp, u_short *fportp,
557     struct inpcb **oinpp, struct ucred *cred)
558 {
559 	struct sockaddr_in *sin = (struct sockaddr_in *)nam;
560 	struct in_ifaddr *ia;
561 	struct sockaddr_in sa;
562 	struct ucred *socred;
563 	struct inpcb *oinp;
564 	struct in_addr laddr, faddr;
565 	u_short lport, fport;
566 	int error;
567 
568 	INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
569 	INP_LOCK_ASSERT(inp);
570 
571 	if (oinpp != NULL)
572 		*oinpp = NULL;
573 	if (nam->sa_len != sizeof (*sin))
574 		return (EINVAL);
575 	if (sin->sin_family != AF_INET)
576 		return (EAFNOSUPPORT);
577 	if (sin->sin_port == 0)
578 		return (EADDRNOTAVAIL);
579 	laddr.s_addr = *laddrp;
580 	lport = *lportp;
581 	faddr = sin->sin_addr;
582 	fport = sin->sin_port;
583 	socred = inp->inp_socket->so_cred;
584 	if (laddr.s_addr == INADDR_ANY && jailed(socred)) {
585 		bzero(&sa, sizeof(sa));
586 		sa.sin_addr.s_addr = htonl(prison_getip(socred));
587 		sa.sin_len = sizeof(sa);
588 		sa.sin_family = AF_INET;
589 		error = in_pcbbind_setup(inp, (struct sockaddr *)&sa,
590 		    &laddr.s_addr, &lport, cred);
591 		if (error)
592 			return (error);
593 	}
594 	if (!TAILQ_EMPTY(&in_ifaddrhead)) {
595 		/*
596 		 * If the destination address is INADDR_ANY,
597 		 * use the primary local address.
598 		 * If the supplied address is INADDR_BROADCAST,
599 		 * and the primary interface supports broadcast,
600 		 * choose the broadcast address for that interface.
601 		 */
602 		if (faddr.s_addr == INADDR_ANY)
603 			faddr = IA_SIN(TAILQ_FIRST(&in_ifaddrhead))->sin_addr;
604 		else if (faddr.s_addr == (u_long)INADDR_BROADCAST &&
605 		    (TAILQ_FIRST(&in_ifaddrhead)->ia_ifp->if_flags &
606 		    IFF_BROADCAST))
607 			faddr = satosin(&TAILQ_FIRST(
608 			    &in_ifaddrhead)->ia_broadaddr)->sin_addr;
609 	}
610 	if (laddr.s_addr == INADDR_ANY) {
611 		ia = (struct in_ifaddr *)0;
612 		/*
613 		 * If route is known our src addr is taken from the i/f,
614 		 * else punt.
615 		 *
616 		 * Find out route to destination
617 		 */
618 		if ((inp->inp_socket->so_options & SO_DONTROUTE) == 0)
619 			ia = ip_rtaddr(faddr);
620 		/*
621 		 * If we found a route, use the address corresponding to
622 		 * the outgoing interface.
623 		 *
624 		 * Otherwise assume faddr is reachable on a directly connected
625 		 * network and try to find a corresponding interface to take
626 		 * the source address from.
627 		 */
628 		if (ia == 0) {
629 			bzero(&sa, sizeof(sa));
630 			sa.sin_addr = faddr;
631 			sa.sin_len = sizeof(sa);
632 			sa.sin_family = AF_INET;
633 
634 			ia = ifatoia(ifa_ifwithdstaddr(sintosa(&sa)));
635 			if (ia == 0)
636 				ia = ifatoia(ifa_ifwithnet(sintosa(&sa)));
637 			if (ia == 0)
638 				return (ENETUNREACH);
639 		}
640 		/*
641 		 * If the destination address is multicast and an outgoing
642 		 * interface has been set as a multicast option, use the
643 		 * address of that interface as our source address.
644 		 */
645 		if (IN_MULTICAST(ntohl(faddr.s_addr)) &&
646 		    inp->inp_moptions != NULL) {
647 			struct ip_moptions *imo;
648 			struct ifnet *ifp;
649 
650 			imo = inp->inp_moptions;
651 			if (imo->imo_multicast_ifp != NULL) {
652 				ifp = imo->imo_multicast_ifp;
653 				TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link)
654 					if (ia->ia_ifp == ifp)
655 						break;
656 				if (ia == 0)
657 					return (EADDRNOTAVAIL);
658 			}
659 		}
660 		laddr = ia->ia_addr.sin_addr;
661 	}
662 
663 	oinp = in_pcblookup_hash(inp->inp_pcbinfo, faddr, fport, laddr, lport,
664 	    0, NULL);
665 	if (oinp != NULL) {
666 		if (oinpp != NULL)
667 			*oinpp = oinp;
668 		return (EADDRINUSE);
669 	}
670 	if (lport == 0) {
671 		error = in_pcbbind_setup(inp, NULL, &laddr.s_addr, &lport,
672 		    cred);
673 		if (error)
674 			return (error);
675 	}
676 	*laddrp = laddr.s_addr;
677 	*lportp = lport;
678 	*faddrp = faddr.s_addr;
679 	*fportp = fport;
680 	return (0);
681 }
682 
683 void
684 in_pcbdisconnect(struct inpcb *inp)
685 {
686 
687 	INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
688 	INP_LOCK_ASSERT(inp);
689 
690 	inp->inp_faddr.s_addr = INADDR_ANY;
691 	inp->inp_fport = 0;
692 	in_pcbrehash(inp);
693 }
694 
695 /*
696  * In the old world order, in_pcbdetach() served two functions: to detach the
697  * pcb from the socket/potentially free the socket, and to free the pcb
698  * itself.  In the new world order, the protocol code is responsible for
699  * managing the relationship with the socket, and this code simply frees the
700  * pcb.
701  */
702 void
703 in_pcbdetach(struct inpcb *inp)
704 {
705 
706 	KASSERT(inp->inp_socket != NULL, ("in_pcbdetach: inp_socket == NULL"));
707 	inp->inp_socket->so_pcb = NULL;
708 	inp->inp_socket = NULL;
709 }
710 
711 void
712 in_pcbfree(struct inpcb *inp)
713 {
714 	struct inpcbinfo *ipi = inp->inp_pcbinfo;
715 
716 	KASSERT(inp->inp_socket == NULL, ("in_pcbfree: inp_socket != NULL"));
717 	INP_INFO_WLOCK_ASSERT(ipi);
718 	INP_LOCK_ASSERT(inp);
719 
720 #ifdef IPSEC
721 	ipsec4_delete_pcbpolicy(inp);
722 #endif /*IPSEC*/
723 	inp->inp_gencnt = ++ipi->ipi_gencnt;
724 	in_pcbremlists(inp);
725 	if (inp->inp_options)
726 		(void)m_free(inp->inp_options);
727 	if (inp->inp_moptions != NULL)
728 		inp_freemoptions(inp->inp_moptions);
729 	inp->inp_vflag = 0;
730 
731 #ifdef MAC
732 	mac_inpcb_destroy(inp);
733 #endif
734 	INP_UNLOCK(inp);
735 	uma_zfree(ipi->ipi_zone, inp);
736 }
737 
738 /*
739  * TCP needs to maintain its inpcb structure after the TCP connection has
740  * been torn down.  However, it must be disconnected from the inpcb hashes as
741  * it must not prevent binding of future connections to the same port/ip
742  * combination by other inpcbs.
743  */
744 void
745 in_pcbdrop(struct inpcb *inp)
746 {
747 
748 	INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
749 	INP_LOCK_ASSERT(inp);
750 
751 	inp->inp_vflag |= INP_DROPPED;
752 	if (inp->inp_lport) {
753 		struct inpcbport *phd = inp->inp_phd;
754 
755 		LIST_REMOVE(inp, inp_hash);
756 		LIST_REMOVE(inp, inp_portlist);
757 		if (LIST_FIRST(&phd->phd_pcblist) == NULL) {
758 			LIST_REMOVE(phd, phd_hash);
759 			free(phd, M_PCB);
760 		}
761 		inp->inp_lport = 0;
762 	}
763 }
764 
765 /*
766  * Common routines to return the socket addresses associated with inpcbs.
767  */
768 struct sockaddr *
769 in_sockaddr(in_port_t port, struct in_addr *addr_p)
770 {
771 	struct sockaddr_in *sin;
772 
773 	MALLOC(sin, struct sockaddr_in *, sizeof *sin, M_SONAME,
774 		M_WAITOK | M_ZERO);
775 	sin->sin_family = AF_INET;
776 	sin->sin_len = sizeof(*sin);
777 	sin->sin_addr = *addr_p;
778 	sin->sin_port = port;
779 
780 	return (struct sockaddr *)sin;
781 }
782 
783 int
784 in_getsockaddr(struct socket *so, struct sockaddr **nam)
785 {
786 	struct inpcb *inp;
787 	struct in_addr addr;
788 	in_port_t port;
789 
790 	inp = sotoinpcb(so);
791 	KASSERT(inp != NULL, ("in_getsockaddr: inp == NULL"));
792 
793 	INP_LOCK(inp);
794 	port = inp->inp_lport;
795 	addr = inp->inp_laddr;
796 	INP_UNLOCK(inp);
797 
798 	*nam = in_sockaddr(port, &addr);
799 	return 0;
800 }
801 
802 int
803 in_getpeeraddr(struct socket *so, struct sockaddr **nam)
804 {
805 	struct inpcb *inp;
806 	struct in_addr addr;
807 	in_port_t port;
808 
809 	inp = sotoinpcb(so);
810 	KASSERT(inp != NULL, ("in_getpeeraddr: inp == NULL"));
811 
812 	INP_LOCK(inp);
813 	port = inp->inp_fport;
814 	addr = inp->inp_faddr;
815 	INP_UNLOCK(inp);
816 
817 	*nam = in_sockaddr(port, &addr);
818 	return 0;
819 }
820 
821 void
822 in_pcbnotifyall(struct inpcbinfo *pcbinfo, struct in_addr faddr, int errno,
823     struct inpcb *(*notify)(struct inpcb *, int))
824 {
825 	struct inpcb *inp, *ninp;
826 	struct inpcbhead *head;
827 
828 	INP_INFO_WLOCK(pcbinfo);
829 	head = pcbinfo->ipi_listhead;
830 	for (inp = LIST_FIRST(head); inp != NULL; inp = ninp) {
831 		INP_LOCK(inp);
832 		ninp = LIST_NEXT(inp, inp_list);
833 #ifdef INET6
834 		if ((inp->inp_vflag & INP_IPV4) == 0) {
835 			INP_UNLOCK(inp);
836 			continue;
837 		}
838 #endif
839 		if (inp->inp_faddr.s_addr != faddr.s_addr ||
840 		    inp->inp_socket == NULL) {
841 			INP_UNLOCK(inp);
842 			continue;
843 		}
844 		if ((*notify)(inp, errno))
845 			INP_UNLOCK(inp);
846 	}
847 	INP_INFO_WUNLOCK(pcbinfo);
848 }
849 
850 void
851 in_pcbpurgeif0(struct inpcbinfo *pcbinfo, struct ifnet *ifp)
852 {
853 	struct inpcb *inp;
854 	struct ip_moptions *imo;
855 	int i, gap;
856 
857 	INP_INFO_RLOCK(pcbinfo);
858 	LIST_FOREACH(inp, pcbinfo->ipi_listhead, inp_list) {
859 		INP_LOCK(inp);
860 		imo = inp->inp_moptions;
861 		if ((inp->inp_vflag & INP_IPV4) &&
862 		    imo != NULL) {
863 			/*
864 			 * Unselect the outgoing interface if it is being
865 			 * detached.
866 			 */
867 			if (imo->imo_multicast_ifp == ifp)
868 				imo->imo_multicast_ifp = NULL;
869 
870 			/*
871 			 * Drop multicast group membership if we joined
872 			 * through the interface being detached.
873 			 */
874 			for (i = 0, gap = 0; i < imo->imo_num_memberships;
875 			    i++) {
876 				if (imo->imo_membership[i]->inm_ifp == ifp) {
877 					in_delmulti(imo->imo_membership[i]);
878 					gap++;
879 				} else if (gap != 0)
880 					imo->imo_membership[i - gap] =
881 					    imo->imo_membership[i];
882 			}
883 			imo->imo_num_memberships -= gap;
884 		}
885 		INP_UNLOCK(inp);
886 	}
887 	INP_INFO_RUNLOCK(pcbinfo);
888 }
889 
890 /*
891  * Lookup a PCB based on the local address and port.
892  */
893 #define INP_LOOKUP_MAPPED_PCB_COST	3
894 struct inpcb *
895 in_pcblookup_local(struct inpcbinfo *pcbinfo, struct in_addr laddr,
896     u_int lport_arg, int wild_okay)
897 {
898 	struct inpcb *inp;
899 #ifdef INET6
900 	int matchwild = 3 + INP_LOOKUP_MAPPED_PCB_COST;
901 #else
902 	int matchwild = 3;
903 #endif
904 	int wildcard;
905 	u_short lport = lport_arg;
906 
907 	INP_INFO_WLOCK_ASSERT(pcbinfo);
908 
909 	if (!wild_okay) {
910 		struct inpcbhead *head;
911 		/*
912 		 * Look for an unconnected (wildcard foreign addr) PCB that
913 		 * matches the local address and port we're looking for.
914 		 */
915 		head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport,
916 		    0, pcbinfo->ipi_hashmask)];
917 		LIST_FOREACH(inp, head, inp_hash) {
918 #ifdef INET6
919 			if ((inp->inp_vflag & INP_IPV4) == 0)
920 				continue;
921 #endif
922 			if (inp->inp_faddr.s_addr == INADDR_ANY &&
923 			    inp->inp_laddr.s_addr == laddr.s_addr &&
924 			    inp->inp_lport == lport) {
925 				/*
926 				 * Found.
927 				 */
928 				return (inp);
929 			}
930 		}
931 		/*
932 		 * Not found.
933 		 */
934 		return (NULL);
935 	} else {
936 		struct inpcbporthead *porthash;
937 		struct inpcbport *phd;
938 		struct inpcb *match = NULL;
939 		/*
940 		 * Best fit PCB lookup.
941 		 *
942 		 * First see if this local port is in use by looking on the
943 		 * port hash list.
944 		 */
945 		porthash = &pcbinfo->ipi_porthashbase[INP_PCBPORTHASH(lport,
946 		    pcbinfo->ipi_porthashmask)];
947 		LIST_FOREACH(phd, porthash, phd_hash) {
948 			if (phd->phd_port == lport)
949 				break;
950 		}
951 		if (phd != NULL) {
952 			/*
953 			 * Port is in use by one or more PCBs. Look for best
954 			 * fit.
955 			 */
956 			LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) {
957 				wildcard = 0;
958 #ifdef INET6
959 				if ((inp->inp_vflag & INP_IPV4) == 0)
960 					continue;
961 				/*
962 				 * We never select the PCB that has
963 				 * INP_IPV6 flag and is bound to :: if
964 				 * we have another PCB which is bound
965 				 * to 0.0.0.0.  If a PCB has the
966 				 * INP_IPV6 flag, then we set its cost
967 				 * higher than IPv4 only PCBs.
968 				 *
969 				 * Note that the case only happens
970 				 * when a socket is bound to ::, under
971 				 * the condition that the use of the
972 				 * mapped address is allowed.
973 				 */
974 				if ((inp->inp_vflag & INP_IPV6) != 0)
975 					wildcard += INP_LOOKUP_MAPPED_PCB_COST;
976 #endif
977 				if (inp->inp_faddr.s_addr != INADDR_ANY)
978 					wildcard++;
979 				if (inp->inp_laddr.s_addr != INADDR_ANY) {
980 					if (laddr.s_addr == INADDR_ANY)
981 						wildcard++;
982 					else if (inp->inp_laddr.s_addr != laddr.s_addr)
983 						continue;
984 				} else {
985 					if (laddr.s_addr != INADDR_ANY)
986 						wildcard++;
987 				}
988 				if (wildcard < matchwild) {
989 					match = inp;
990 					matchwild = wildcard;
991 					if (matchwild == 0) {
992 						break;
993 					}
994 				}
995 			}
996 		}
997 		return (match);
998 	}
999 }
1000 #undef INP_LOOKUP_MAPPED_PCB_COST
1001 
1002 /*
1003  * Lookup PCB in hash list.
1004  */
1005 struct inpcb *
1006 in_pcblookup_hash(struct inpcbinfo *pcbinfo, struct in_addr faddr,
1007     u_int fport_arg, struct in_addr laddr, u_int lport_arg, int wildcard,
1008     struct ifnet *ifp)
1009 {
1010 	struct inpcbhead *head;
1011 	struct inpcb *inp;
1012 	u_short fport = fport_arg, lport = lport_arg;
1013 
1014 	INP_INFO_RLOCK_ASSERT(pcbinfo);
1015 
1016 	/*
1017 	 * First look for an exact match.
1018 	 */
1019 	head = &pcbinfo->ipi_hashbase[INP_PCBHASH(faddr.s_addr, lport, fport,
1020 	    pcbinfo->ipi_hashmask)];
1021 	LIST_FOREACH(inp, head, inp_hash) {
1022 #ifdef INET6
1023 		if ((inp->inp_vflag & INP_IPV4) == 0)
1024 			continue;
1025 #endif
1026 		if (inp->inp_faddr.s_addr == faddr.s_addr &&
1027 		    inp->inp_laddr.s_addr == laddr.s_addr &&
1028 		    inp->inp_fport == fport &&
1029 		    inp->inp_lport == lport)
1030 			return (inp);
1031 	}
1032 
1033 	/*
1034 	 * Then look for a wildcard match, if requested.
1035 	 */
1036 	if (wildcard) {
1037 		struct inpcb *local_wild = NULL;
1038 #ifdef INET6
1039 		struct inpcb *local_wild_mapped = NULL;
1040 #endif
1041 
1042 		head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport,
1043 		    0, pcbinfo->ipi_hashmask)];
1044 		LIST_FOREACH(inp, head, inp_hash) {
1045 #ifdef INET6
1046 			if ((inp->inp_vflag & INP_IPV4) == 0)
1047 				continue;
1048 #endif
1049 			if (inp->inp_faddr.s_addr == INADDR_ANY &&
1050 			    inp->inp_lport == lport) {
1051 				if (ifp && ifp->if_type == IFT_FAITH &&
1052 				    (inp->inp_flags & INP_FAITH) == 0)
1053 					continue;
1054 				if (inp->inp_laddr.s_addr == laddr.s_addr)
1055 					return (inp);
1056 				else if (inp->inp_laddr.s_addr == INADDR_ANY) {
1057 #ifdef INET6
1058 					if (INP_CHECK_SOCKAF(inp->inp_socket,
1059 							     AF_INET6))
1060 						local_wild_mapped = inp;
1061 					else
1062 #endif
1063 						local_wild = inp;
1064 				}
1065 			}
1066 		}
1067 #ifdef INET6
1068 		if (local_wild == NULL)
1069 			return (local_wild_mapped);
1070 #endif
1071 		return (local_wild);
1072 	}
1073 	return (NULL);
1074 }
1075 
1076 /*
1077  * Insert PCB onto various hash lists.
1078  */
1079 int
1080 in_pcbinshash(struct inpcb *inp)
1081 {
1082 	struct inpcbhead *pcbhash;
1083 	struct inpcbporthead *pcbporthash;
1084 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1085 	struct inpcbport *phd;
1086 	u_int32_t hashkey_faddr;
1087 
1088 	INP_INFO_WLOCK_ASSERT(pcbinfo);
1089 	INP_LOCK_ASSERT(inp);
1090 
1091 #ifdef INET6
1092 	if (inp->inp_vflag & INP_IPV6)
1093 		hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX */;
1094 	else
1095 #endif /* INET6 */
1096 	hashkey_faddr = inp->inp_faddr.s_addr;
1097 
1098 	pcbhash = &pcbinfo->ipi_hashbase[INP_PCBHASH(hashkey_faddr,
1099 		 inp->inp_lport, inp->inp_fport, pcbinfo->ipi_hashmask)];
1100 
1101 	pcbporthash = &pcbinfo->ipi_porthashbase[
1102 	    INP_PCBPORTHASH(inp->inp_lport, pcbinfo->ipi_porthashmask)];
1103 
1104 	/*
1105 	 * Go through port list and look for a head for this lport.
1106 	 */
1107 	LIST_FOREACH(phd, pcbporthash, phd_hash) {
1108 		if (phd->phd_port == inp->inp_lport)
1109 			break;
1110 	}
1111 	/*
1112 	 * If none exists, malloc one and tack it on.
1113 	 */
1114 	if (phd == NULL) {
1115 		MALLOC(phd, struct inpcbport *, sizeof(struct inpcbport), M_PCB, M_NOWAIT);
1116 		if (phd == NULL) {
1117 			return (ENOBUFS); /* XXX */
1118 		}
1119 		phd->phd_port = inp->inp_lport;
1120 		LIST_INIT(&phd->phd_pcblist);
1121 		LIST_INSERT_HEAD(pcbporthash, phd, phd_hash);
1122 	}
1123 	inp->inp_phd = phd;
1124 	LIST_INSERT_HEAD(&phd->phd_pcblist, inp, inp_portlist);
1125 	LIST_INSERT_HEAD(pcbhash, inp, inp_hash);
1126 	return (0);
1127 }
1128 
1129 /*
1130  * Move PCB to the proper hash bucket when { faddr, fport } have  been
1131  * changed. NOTE: This does not handle the case of the lport changing (the
1132  * hashed port list would have to be updated as well), so the lport must
1133  * not change after in_pcbinshash() has been called.
1134  */
1135 void
1136 in_pcbrehash(struct inpcb *inp)
1137 {
1138 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1139 	struct inpcbhead *head;
1140 	u_int32_t hashkey_faddr;
1141 
1142 	INP_INFO_WLOCK_ASSERT(pcbinfo);
1143 	INP_LOCK_ASSERT(inp);
1144 
1145 #ifdef INET6
1146 	if (inp->inp_vflag & INP_IPV6)
1147 		hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX */;
1148 	else
1149 #endif /* INET6 */
1150 	hashkey_faddr = inp->inp_faddr.s_addr;
1151 
1152 	head = &pcbinfo->ipi_hashbase[INP_PCBHASH(hashkey_faddr,
1153 		inp->inp_lport, inp->inp_fport, pcbinfo->ipi_hashmask)];
1154 
1155 	LIST_REMOVE(inp, inp_hash);
1156 	LIST_INSERT_HEAD(head, inp, inp_hash);
1157 }
1158 
1159 /*
1160  * Remove PCB from various lists.
1161  */
1162 void
1163 in_pcbremlists(struct inpcb *inp)
1164 {
1165 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1166 
1167 	INP_INFO_WLOCK_ASSERT(pcbinfo);
1168 	INP_LOCK_ASSERT(inp);
1169 
1170 	inp->inp_gencnt = ++pcbinfo->ipi_gencnt;
1171 	if (inp->inp_lport) {
1172 		struct inpcbport *phd = inp->inp_phd;
1173 
1174 		LIST_REMOVE(inp, inp_hash);
1175 		LIST_REMOVE(inp, inp_portlist);
1176 		if (LIST_FIRST(&phd->phd_pcblist) == NULL) {
1177 			LIST_REMOVE(phd, phd_hash);
1178 			free(phd, M_PCB);
1179 		}
1180 	}
1181 	LIST_REMOVE(inp, inp_list);
1182 	pcbinfo->ipi_count--;
1183 }
1184 
1185 /*
1186  * A set label operation has occurred at the socket layer, propagate the
1187  * label change into the in_pcb for the socket.
1188  */
1189 void
1190 in_pcbsosetlabel(struct socket *so)
1191 {
1192 #ifdef MAC
1193 	struct inpcb *inp;
1194 
1195 	inp = sotoinpcb(so);
1196 	KASSERT(inp != NULL, ("in_pcbsosetlabel: so->so_pcb == NULL"));
1197 
1198 	INP_LOCK(inp);
1199 	SOCK_LOCK(so);
1200 	mac_inpcb_sosetlabel(so, inp);
1201 	SOCK_UNLOCK(so);
1202 	INP_UNLOCK(inp);
1203 #endif
1204 }
1205 
1206 /*
1207  * ipport_tick runs once per second, determining if random port allocation
1208  * should be continued.  If more than ipport_randomcps ports have been
1209  * allocated in the last second, then we return to sequential port
1210  * allocation. We return to random allocation only once we drop below
1211  * ipport_randomcps for at least ipport_randomtime seconds.
1212  */
1213 void
1214 ipport_tick(void *xtp)
1215 {
1216 
1217 	if (ipport_tcpallocs <= ipport_tcplastcount + ipport_randomcps) {
1218 		if (ipport_stoprandom > 0)
1219 			ipport_stoprandom--;
1220 	} else
1221 		ipport_stoprandom = ipport_randomtime;
1222 	ipport_tcplastcount = ipport_tcpallocs;
1223 	callout_reset(&ipport_tick_callout, hz, ipport_tick, NULL);
1224 }
1225 
1226 #ifdef DDB
1227 static void
1228 db_print_indent(int indent)
1229 {
1230 	int i;
1231 
1232 	for (i = 0; i < indent; i++)
1233 		db_printf(" ");
1234 }
1235 
1236 static void
1237 db_print_inconninfo(struct in_conninfo *inc, const char *name, int indent)
1238 {
1239 	char faddr_str[48], laddr_str[48];
1240 
1241 	db_print_indent(indent);
1242 	db_printf("%s at %p\n", name, inc);
1243 
1244 	indent += 2;
1245 
1246 #ifdef INET6
1247 	if (inc->inc_flags == 1) {
1248 		/* IPv6. */
1249 		ip6_sprintf(laddr_str, &inc->inc6_laddr);
1250 		ip6_sprintf(faddr_str, &inc->inc6_faddr);
1251 	} else {
1252 #endif
1253 		/* IPv4. */
1254 		inet_ntoa_r(inc->inc_laddr, laddr_str);
1255 		inet_ntoa_r(inc->inc_faddr, faddr_str);
1256 #ifdef INET6
1257 	}
1258 #endif
1259 	db_print_indent(indent);
1260 	db_printf("inc_laddr %s   inc_lport %u\n", laddr_str,
1261 	    ntohs(inc->inc_lport));
1262 	db_print_indent(indent);
1263 	db_printf("inc_faddr %s   inc_fport %u\n", faddr_str,
1264 	    ntohs(inc->inc_fport));
1265 }
1266 
1267 static void
1268 db_print_inpflags(int inp_flags)
1269 {
1270 	int comma;
1271 
1272 	comma = 0;
1273 	if (inp_flags & INP_RECVOPTS) {
1274 		db_printf("%sINP_RECVOPTS", comma ? ", " : "");
1275 		comma = 1;
1276 	}
1277 	if (inp_flags & INP_RECVRETOPTS) {
1278 		db_printf("%sINP_RECVRETOPTS", comma ? ", " : "");
1279 		comma = 1;
1280 	}
1281 	if (inp_flags & INP_RECVDSTADDR) {
1282 		db_printf("%sINP_RECVDSTADDR", comma ? ", " : "");
1283 		comma = 1;
1284 	}
1285 	if (inp_flags & INP_HDRINCL) {
1286 		db_printf("%sINP_HDRINCL", comma ? ", " : "");
1287 		comma = 1;
1288 	}
1289 	if (inp_flags & INP_HIGHPORT) {
1290 		db_printf("%sINP_HIGHPORT", comma ? ", " : "");
1291 		comma = 1;
1292 	}
1293 	if (inp_flags & INP_LOWPORT) {
1294 		db_printf("%sINP_LOWPORT", comma ? ", " : "");
1295 		comma = 1;
1296 	}
1297 	if (inp_flags & INP_ANONPORT) {
1298 		db_printf("%sINP_ANONPORT", comma ? ", " : "");
1299 		comma = 1;
1300 	}
1301 	if (inp_flags & INP_RECVIF) {
1302 		db_printf("%sINP_RECVIF", comma ? ", " : "");
1303 		comma = 1;
1304 	}
1305 	if (inp_flags & INP_MTUDISC) {
1306 		db_printf("%sINP_MTUDISC", comma ? ", " : "");
1307 		comma = 1;
1308 	}
1309 	if (inp_flags & INP_FAITH) {
1310 		db_printf("%sINP_FAITH", comma ? ", " : "");
1311 		comma = 1;
1312 	}
1313 	if (inp_flags & INP_RECVTTL) {
1314 		db_printf("%sINP_RECVTTL", comma ? ", " : "");
1315 		comma = 1;
1316 	}
1317 	if (inp_flags & INP_DONTFRAG) {
1318 		db_printf("%sINP_DONTFRAG", comma ? ", " : "");
1319 		comma = 1;
1320 	}
1321 	if (inp_flags & IN6P_IPV6_V6ONLY) {
1322 		db_printf("%sIN6P_IPV6_V6ONLY", comma ? ", " : "");
1323 		comma = 1;
1324 	}
1325 	if (inp_flags & IN6P_PKTINFO) {
1326 		db_printf("%sIN6P_PKTINFO", comma ? ", " : "");
1327 		comma = 1;
1328 	}
1329 	if (inp_flags & IN6P_HOPLIMIT) {
1330 		db_printf("%sIN6P_HOPLIMIT", comma ? ", " : "");
1331 		comma = 1;
1332 	}
1333 	if (inp_flags & IN6P_HOPOPTS) {
1334 		db_printf("%sIN6P_HOPOPTS", comma ? ", " : "");
1335 		comma = 1;
1336 	}
1337 	if (inp_flags & IN6P_DSTOPTS) {
1338 		db_printf("%sIN6P_DSTOPTS", comma ? ", " : "");
1339 		comma = 1;
1340 	}
1341 	if (inp_flags & IN6P_RTHDR) {
1342 		db_printf("%sIN6P_RTHDR", comma ? ", " : "");
1343 		comma = 1;
1344 	}
1345 	if (inp_flags & IN6P_RTHDRDSTOPTS) {
1346 		db_printf("%sIN6P_RTHDRDSTOPTS", comma ? ", " : "");
1347 		comma = 1;
1348 	}
1349 	if (inp_flags & IN6P_TCLASS) {
1350 		db_printf("%sIN6P_TCLASS", comma ? ", " : "");
1351 		comma = 1;
1352 	}
1353 	if (inp_flags & IN6P_AUTOFLOWLABEL) {
1354 		db_printf("%sIN6P_AUTOFLOWLABEL", comma ? ", " : "");
1355 		comma = 1;
1356 	}
1357 	if (inp_flags & IN6P_RFC2292) {
1358 		db_printf("%sIN6P_RFC2292", comma ? ", " : "");
1359 		comma = 1;
1360 	}
1361 	if (inp_flags & IN6P_MTU) {
1362 		db_printf("IN6P_MTU%s", comma ? ", " : "");
1363 		comma = 1;
1364 	}
1365 }
1366 
1367 static void
1368 db_print_inpvflag(u_char inp_vflag)
1369 {
1370 	int comma;
1371 
1372 	comma = 0;
1373 	if (inp_vflag & INP_IPV4) {
1374 		db_printf("%sINP_IPV4", comma ? ", " : "");
1375 		comma  = 1;
1376 	}
1377 	if (inp_vflag & INP_IPV6) {
1378 		db_printf("%sINP_IPV6", comma ? ", " : "");
1379 		comma  = 1;
1380 	}
1381 	if (inp_vflag & INP_IPV6PROTO) {
1382 		db_printf("%sINP_IPV6PROTO", comma ? ", " : "");
1383 		comma  = 1;
1384 	}
1385 	if (inp_vflag & INP_TIMEWAIT) {
1386 		db_printf("%sINP_TIMEWAIT", comma ? ", " : "");
1387 		comma  = 1;
1388 	}
1389 	if (inp_vflag & INP_ONESBCAST) {
1390 		db_printf("%sINP_ONESBCAST", comma ? ", " : "");
1391 		comma  = 1;
1392 	}
1393 	if (inp_vflag & INP_DROPPED) {
1394 		db_printf("%sINP_DROPPED", comma ? ", " : "");
1395 		comma  = 1;
1396 	}
1397 	if (inp_vflag & INP_SOCKREF) {
1398 		db_printf("%sINP_SOCKREF", comma ? ", " : "");
1399 		comma  = 1;
1400 	}
1401 }
1402 
1403 void
1404 db_print_inpcb(struct inpcb *inp, const char *name, int indent)
1405 {
1406 
1407 	db_print_indent(indent);
1408 	db_printf("%s at %p\n", name, inp);
1409 
1410 	indent += 2;
1411 
1412 	db_print_indent(indent);
1413 	db_printf("inp_flow: 0x%x\n", inp->inp_flow);
1414 
1415 	db_print_inconninfo(&inp->inp_inc, "inp_conninfo", indent);
1416 
1417 	db_print_indent(indent);
1418 	db_printf("inp_ppcb: %p   inp_pcbinfo: %p   inp_socket: %p\n",
1419 	    inp->inp_ppcb, inp->inp_pcbinfo, inp->inp_socket);
1420 
1421 	db_print_indent(indent);
1422 	db_printf("inp_label: %p   inp_flags: 0x%x (",
1423 	   inp->inp_label, inp->inp_flags);
1424 	db_print_inpflags(inp->inp_flags);
1425 	db_printf(")\n");
1426 
1427 	db_print_indent(indent);
1428 	db_printf("inp_sp: %p   inp_vflag: 0x%x (", inp->inp_sp,
1429 	    inp->inp_vflag);
1430 	db_print_inpvflag(inp->inp_vflag);
1431 	db_printf(")\n");
1432 
1433 	db_print_indent(indent);
1434 	db_printf("inp_ip_ttl: %d   inp_ip_p: %d   inp_ip_minttl: %d\n",
1435 	    inp->inp_ip_ttl, inp->inp_ip_p, inp->inp_ip_minttl);
1436 
1437 	db_print_indent(indent);
1438 #ifdef INET6
1439 	if (inp->inp_vflag & INP_IPV6) {
1440 		db_printf("in6p_options: %p   in6p_outputopts: %p   "
1441 		    "in6p_moptions: %p\n", inp->in6p_options,
1442 		    inp->in6p_outputopts, inp->in6p_moptions);
1443 		db_printf("in6p_icmp6filt: %p   in6p_cksum %d   "
1444 		    "in6p_hops %u\n", inp->in6p_icmp6filt, inp->in6p_cksum,
1445 		    inp->in6p_hops);
1446 	} else
1447 #endif
1448 	{
1449 		db_printf("inp_ip_tos: %d   inp_ip_options: %p   "
1450 		    "inp_ip_moptions: %p\n", inp->inp_ip_tos,
1451 		    inp->inp_options, inp->inp_moptions);
1452 	}
1453 
1454 	db_print_indent(indent);
1455 	db_printf("inp_phd: %p   inp_gencnt: %ju\n", inp->inp_phd,
1456 	    (uintmax_t)inp->inp_gencnt);
1457 }
1458 
1459 DB_SHOW_COMMAND(inpcb, db_show_inpcb)
1460 {
1461 	struct inpcb *inp;
1462 
1463 	if (!have_addr) {
1464 		db_printf("usage: show inpcb <addr>\n");
1465 		return;
1466 	}
1467 	inp = (struct inpcb *)addr;
1468 
1469 	db_print_inpcb(inp, "inpcb", 0);
1470 }
1471 #endif
1472