xref: /freebsd/sys/netinet/in_pcb.c (revision b28624fde638caadd4a89f50c9b7e7da0f98c4d2)
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  * $FreeBSD$
33  */
34 
35 #include "opt_ddb.h"
36 #include "opt_ipsec.h"
37 #include "opt_inet6.h"
38 #include "opt_mac.h"
39 
40 #include <sys/param.h>
41 #include <sys/systm.h>
42 #include <sys/malloc.h>
43 #include <sys/mbuf.h>
44 #include <sys/domain.h>
45 #include <sys/protosw.h>
46 #include <sys/socket.h>
47 #include <sys/socketvar.h>
48 #include <sys/priv.h>
49 #include <sys/proc.h>
50 #include <sys/jail.h>
51 #include <sys/kernel.h>
52 #include <sys/sysctl.h>
53 
54 #ifdef DDB
55 #include <ddb/ddb.h>
56 #endif
57 
58 #include <vm/uma.h>
59 
60 #include <net/if.h>
61 #include <net/if_types.h>
62 #include <net/route.h>
63 
64 #include <netinet/in.h>
65 #include <netinet/in_pcb.h>
66 #include <netinet/in_var.h>
67 #include <netinet/ip_var.h>
68 #include <netinet/tcp_var.h>
69 #include <netinet/udp.h>
70 #include <netinet/udp_var.h>
71 #ifdef INET6
72 #include <netinet/ip6.h>
73 #include <netinet6/ip6_var.h>
74 #endif /* INET6 */
75 
76 
77 #ifdef IPSEC
78 #include <netipsec/ipsec.h>
79 #include <netipsec/key.h>
80 #endif /* IPSEC */
81 
82 #include <security/mac/mac_framework.h>
83 
84 /*
85  * These configure the range of local port addresses assigned to
86  * "unspecified" outgoing connections/packets/whatever.
87  */
88 int	ipport_lowfirstauto  = IPPORT_RESERVED - 1;	/* 1023 */
89 int	ipport_lowlastauto = IPPORT_RESERVEDSTART;	/* 600 */
90 int	ipport_firstauto = IPPORT_HIFIRSTAUTO;		/* 49152 */
91 int	ipport_lastauto  = IPPORT_HILASTAUTO;		/* 65535 */
92 int	ipport_hifirstauto = IPPORT_HIFIRSTAUTO;	/* 49152 */
93 int	ipport_hilastauto  = IPPORT_HILASTAUTO;		/* 65535 */
94 
95 /*
96  * Reserved ports accessible only to root. There are significant
97  * security considerations that must be accounted for when changing these,
98  * but the security benefits can be great. Please be careful.
99  */
100 int	ipport_reservedhigh = IPPORT_RESERVED - 1;	/* 1023 */
101 int	ipport_reservedlow = 0;
102 
103 /* Variables dealing with random ephemeral port allocation. */
104 int	ipport_randomized = 1;	/* user controlled via sysctl */
105 int	ipport_randomcps = 10;	/* user controlled via sysctl */
106 int	ipport_randomtime = 45;	/* user controlled via sysctl */
107 int	ipport_stoprandom = 0;	/* toggled by ipport_tick */
108 int	ipport_tcpallocs;
109 int	ipport_tcplastcount;
110 
111 #define RANGECHK(var, min, max) \
112 	if ((var) < (min)) { (var) = (min); } \
113 	else if ((var) > (max)) { (var) = (max); }
114 
115 static int
116 sysctl_net_ipport_check(SYSCTL_HANDLER_ARGS)
117 {
118 	int error;
119 
120 	error = sysctl_handle_int(oidp, oidp->oid_arg1, oidp->oid_arg2, req);
121 	if (error == 0) {
122 		RANGECHK(ipport_lowfirstauto, 1, IPPORT_RESERVED - 1);
123 		RANGECHK(ipport_lowlastauto, 1, IPPORT_RESERVED - 1);
124 		RANGECHK(ipport_firstauto, IPPORT_RESERVED, IPPORT_MAX);
125 		RANGECHK(ipport_lastauto, IPPORT_RESERVED, IPPORT_MAX);
126 		RANGECHK(ipport_hifirstauto, IPPORT_RESERVED, IPPORT_MAX);
127 		RANGECHK(ipport_hilastauto, IPPORT_RESERVED, IPPORT_MAX);
128 	}
129 	return (error);
130 }
131 
132 #undef RANGECHK
133 
134 SYSCTL_NODE(_net_inet_ip, IPPROTO_IP, portrange, CTLFLAG_RW, 0, "IP Ports");
135 
136 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, lowfirst, CTLTYPE_INT|CTLFLAG_RW,
137 	   &ipport_lowfirstauto, 0, &sysctl_net_ipport_check, "I", "");
138 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, lowlast, CTLTYPE_INT|CTLFLAG_RW,
139 	   &ipport_lowlastauto, 0, &sysctl_net_ipport_check, "I", "");
140 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, first, CTLTYPE_INT|CTLFLAG_RW,
141 	   &ipport_firstauto, 0, &sysctl_net_ipport_check, "I", "");
142 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, last, CTLTYPE_INT|CTLFLAG_RW,
143 	   &ipport_lastauto, 0, &sysctl_net_ipport_check, "I", "");
144 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, hifirst, CTLTYPE_INT|CTLFLAG_RW,
145 	   &ipport_hifirstauto, 0, &sysctl_net_ipport_check, "I", "");
146 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, hilast, CTLTYPE_INT|CTLFLAG_RW,
147 	   &ipport_hilastauto, 0, &sysctl_net_ipport_check, "I", "");
148 SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, reservedhigh,
149 	   CTLFLAG_RW|CTLFLAG_SECURE, &ipport_reservedhigh, 0, "");
150 SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, reservedlow,
151 	   CTLFLAG_RW|CTLFLAG_SECURE, &ipport_reservedlow, 0, "");
152 SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, randomized, CTLFLAG_RW,
153 	   &ipport_randomized, 0, "Enable random port allocation");
154 SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, randomcps, CTLFLAG_RW,
155 	   &ipport_randomcps, 0, "Maximum number of random port "
156 	   "allocations before switching to a sequental one");
157 SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, randomtime, CTLFLAG_RW,
158 	   &ipport_randomtime, 0, "Minimum time to keep sequental port "
159 	   "allocation before switching to a random one");
160 
161 /*
162  * in_pcb.c: manage the Protocol Control Blocks.
163  *
164  * NOTE: It is assumed that most of these functions will be called with
165  * the pcbinfo lock held, and often, the inpcb lock held, as these utility
166  * functions often modify hash chains or addresses in pcbs.
167  */
168 
169 /*
170  * Allocate a PCB and associate it with the socket.
171  * On success return with the PCB locked.
172  */
173 int
174 in_pcballoc(struct socket *so, struct inpcbinfo *pcbinfo)
175 {
176 	struct inpcb *inp;
177 	int error;
178 
179 	INP_INFO_WLOCK_ASSERT(pcbinfo);
180 	error = 0;
181 	inp = uma_zalloc(pcbinfo->ipi_zone, M_NOWAIT);
182 	if (inp == NULL)
183 		return (ENOBUFS);
184 	bzero(inp, inp_zero_size);
185 	inp->inp_pcbinfo = pcbinfo;
186 	inp->inp_socket = so;
187 #ifdef MAC
188 	error = mac_init_inpcb(inp, M_NOWAIT);
189 	if (error != 0)
190 		goto out;
191 	SOCK_LOCK(so);
192 	mac_create_inpcb_from_socket(so, inp);
193 	SOCK_UNLOCK(so);
194 #endif
195 
196 #ifdef IPSEC
197 	error = ipsec_init_policy(so, &inp->inp_sp);
198 	if (error != 0)
199 		goto out;
200 #endif /*IPSEC*/
201 #ifdef INET6
202 	if (INP_SOCKAF(so) == AF_INET6) {
203 		inp->inp_vflag |= INP_IPV6PROTO;
204 		if (ip6_v6only)
205 			inp->inp_flags |= IN6P_IPV6_V6ONLY;
206 	}
207 #endif
208 	LIST_INSERT_HEAD(pcbinfo->ipi_listhead, inp, inp_list);
209 	pcbinfo->ipi_count++;
210 	so->so_pcb = (caddr_t)inp;
211 #ifdef INET6
212 	if (ip6_auto_flowlabel)
213 		inp->inp_flags |= IN6P_AUTOFLOWLABEL;
214 #endif
215 	INP_LOCK(inp);
216 	inp->inp_gencnt = ++pcbinfo->ipi_gencnt;
217 
218 #if defined(IPSEC) || defined(MAC)
219 out:
220 	if (error != 0)
221 		uma_zfree(pcbinfo->ipi_zone, inp);
222 #endif
223 	return (error);
224 }
225 
226 int
227 in_pcbbind(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred)
228 {
229 	int anonport, error;
230 
231 	INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
232 	INP_LOCK_ASSERT(inp);
233 
234 	if (inp->inp_lport != 0 || inp->inp_laddr.s_addr != INADDR_ANY)
235 		return (EINVAL);
236 	anonport = inp->inp_lport == 0 && (nam == NULL ||
237 	    ((struct sockaddr_in *)nam)->sin_port == 0);
238 	error = in_pcbbind_setup(inp, nam, &inp->inp_laddr.s_addr,
239 	    &inp->inp_lport, cred);
240 	if (error)
241 		return (error);
242 	if (in_pcbinshash(inp) != 0) {
243 		inp->inp_laddr.s_addr = INADDR_ANY;
244 		inp->inp_lport = 0;
245 		return (EAGAIN);
246 	}
247 	if (anonport)
248 		inp->inp_flags |= INP_ANONPORT;
249 	return (0);
250 }
251 
252 /*
253  * Set up a bind operation on a PCB, performing port allocation
254  * as required, but do not actually modify the PCB. Callers can
255  * either complete the bind by setting inp_laddr/inp_lport and
256  * calling in_pcbinshash(), or they can just use the resulting
257  * port and address to authorise the sending of a once-off packet.
258  *
259  * On error, the values of *laddrp and *lportp are not changed.
260  */
261 int
262 in_pcbbind_setup(struct inpcb *inp, struct sockaddr *nam, in_addr_t *laddrp,
263     u_short *lportp, struct ucred *cred)
264 {
265 	struct socket *so = inp->inp_socket;
266 	unsigned short *lastport;
267 	struct sockaddr_in *sin;
268 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
269 	struct in_addr laddr;
270 	u_short lport = 0;
271 	int wild = 0, reuseport = (so->so_options & SO_REUSEPORT);
272 	int error, prison = 0;
273 	int dorandom;
274 
275 	INP_INFO_WLOCK_ASSERT(pcbinfo);
276 	INP_LOCK_ASSERT(inp);
277 
278 	if (TAILQ_EMPTY(&in_ifaddrhead)) /* XXX broken! */
279 		return (EADDRNOTAVAIL);
280 	laddr.s_addr = *laddrp;
281 	if (nam != NULL && laddr.s_addr != INADDR_ANY)
282 		return (EINVAL);
283 	if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0)
284 		wild = INPLOOKUP_WILDCARD;
285 	if (nam) {
286 		sin = (struct sockaddr_in *)nam;
287 		if (nam->sa_len != sizeof (*sin))
288 			return (EINVAL);
289 #ifdef notdef
290 		/*
291 		 * We should check the family, but old programs
292 		 * incorrectly fail to initialize it.
293 		 */
294 		if (sin->sin_family != AF_INET)
295 			return (EAFNOSUPPORT);
296 #endif
297 		if (sin->sin_addr.s_addr != INADDR_ANY)
298 			if (prison_ip(cred, 0, &sin->sin_addr.s_addr))
299 				return(EINVAL);
300 		if (sin->sin_port != *lportp) {
301 			/* Don't allow the port to change. */
302 			if (*lportp != 0)
303 				return (EINVAL);
304 			lport = sin->sin_port;
305 		}
306 		/* NB: lport is left as 0 if the port isn't being changed. */
307 		if (IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) {
308 			/*
309 			 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
310 			 * allow complete duplication of binding if
311 			 * SO_REUSEPORT is set, or if SO_REUSEADDR is set
312 			 * and a multicast address is bound on both
313 			 * new and duplicated sockets.
314 			 */
315 			if (so->so_options & SO_REUSEADDR)
316 				reuseport = SO_REUSEADDR|SO_REUSEPORT;
317 		} else if (sin->sin_addr.s_addr != INADDR_ANY) {
318 			sin->sin_port = 0;		/* yech... */
319 			bzero(&sin->sin_zero, sizeof(sin->sin_zero));
320 			if (ifa_ifwithaddr((struct sockaddr *)sin) == 0)
321 				return (EADDRNOTAVAIL);
322 		}
323 		laddr = sin->sin_addr;
324 		if (lport) {
325 			struct inpcb *t;
326 			struct tcptw *tw;
327 
328 			/* GROSS */
329 			if (ntohs(lport) <= ipport_reservedhigh &&
330 			    ntohs(lport) >= ipport_reservedlow &&
331 			    priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT,
332 			    0))
333 				return (EACCES);
334 			if (jailed(cred))
335 				prison = 1;
336 			if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)) &&
337 			    priv_check_cred(so->so_cred,
338 			    PRIV_NETINET_REUSEPORT, 0) != 0) {
339 				t = in_pcblookup_local(inp->inp_pcbinfo,
340 				    sin->sin_addr, lport,
341 				    prison ? 0 :  INPLOOKUP_WILDCARD);
342 	/*
343 	 * XXX
344 	 * This entire block sorely needs a rewrite.
345 	 */
346 				if (t &&
347 				    ((t->inp_vflag & INP_TIMEWAIT) == 0) &&
348 				    (so->so_type != SOCK_STREAM ||
349 				     ntohl(t->inp_faddr.s_addr) == INADDR_ANY) &&
350 				    (ntohl(sin->sin_addr.s_addr) != INADDR_ANY ||
351 				     ntohl(t->inp_laddr.s_addr) != INADDR_ANY ||
352 				     (t->inp_socket->so_options &
353 					 SO_REUSEPORT) == 0) &&
354 				    (so->so_cred->cr_uid !=
355 				     t->inp_socket->so_cred->cr_uid))
356 					return (EADDRINUSE);
357 			}
358 			if (prison && prison_ip(cred, 0, &sin->sin_addr.s_addr))
359 				return (EADDRNOTAVAIL);
360 			t = in_pcblookup_local(pcbinfo, sin->sin_addr,
361 			    lport, prison ? 0 : wild);
362 			if (t && (t->inp_vflag & INP_TIMEWAIT)) {
363 				/*
364 				 * XXXRW: If an incpb has had its timewait
365 				 * state recycled, we treat the address as
366 				 * being in use (for now).  This is better
367 				 * than a panic, but not desirable.
368 				 */
369 				tw = intotw(inp);
370 				if (tw == NULL ||
371 				    (reuseport & tw->tw_so_options) == 0)
372 					return (EADDRINUSE);
373 			} else if (t &&
374 			    (reuseport & t->inp_socket->so_options) == 0) {
375 #ifdef INET6
376 				if (ntohl(sin->sin_addr.s_addr) !=
377 				    INADDR_ANY ||
378 				    ntohl(t->inp_laddr.s_addr) !=
379 				    INADDR_ANY ||
380 				    INP_SOCKAF(so) ==
381 				    INP_SOCKAF(t->inp_socket))
382 #endif
383 				return (EADDRINUSE);
384 			}
385 		}
386 	}
387 	if (*lportp != 0)
388 		lport = *lportp;
389 	if (lport == 0) {
390 		u_short first, last;
391 		int count;
392 
393 		if (laddr.s_addr != INADDR_ANY)
394 			if (prison_ip(cred, 0, &laddr.s_addr))
395 				return (EINVAL);
396 
397 		if (inp->inp_flags & INP_HIGHPORT) {
398 			first = ipport_hifirstauto;	/* sysctl */
399 			last  = ipport_hilastauto;
400 			lastport = &pcbinfo->ipi_lasthi;
401 		} else if (inp->inp_flags & INP_LOWPORT) {
402 			error = priv_check_cred(cred,
403 			    PRIV_NETINET_RESERVEDPORT, 0);
404 			if (error)
405 				return error;
406 			first = ipport_lowfirstauto;	/* 1023 */
407 			last  = ipport_lowlastauto;	/* 600 */
408 			lastport = &pcbinfo->ipi_lastlow;
409 		} else {
410 			first = ipport_firstauto;	/* sysctl */
411 			last  = ipport_lastauto;
412 			lastport = &pcbinfo->ipi_lastport;
413 		}
414 		/*
415 		 * For UDP, use random port allocation as long as the user
416 		 * allows it.  For TCP (and as of yet unknown) connections,
417 		 * use random port allocation only if the user allows it AND
418 		 * ipport_tick() allows it.
419 		 */
420 		if (ipport_randomized &&
421 			(!ipport_stoprandom || pcbinfo == &udbinfo))
422 			dorandom = 1;
423 		else
424 			dorandom = 0;
425 		/*
426 		 * It makes no sense to do random port allocation if
427 		 * we have the only port available.
428 		 */
429 		if (first == last)
430 			dorandom = 0;
431 		/* Make sure to not include UDP packets in the count. */
432 		if (pcbinfo != &udbinfo)
433 			ipport_tcpallocs++;
434 		/*
435 		 * Simple check to ensure all ports are not used up causing
436 		 * a deadlock here.
437 		 *
438 		 * We split the two cases (up and down) so that the direction
439 		 * is not being tested on each round of the loop.
440 		 */
441 		if (first > last) {
442 			/*
443 			 * counting down
444 			 */
445 			if (dorandom)
446 				*lastport = first -
447 					    (arc4random() % (first - last));
448 			count = first - last;
449 
450 			do {
451 				if (count-- < 0)	/* completely used? */
452 					return (EADDRNOTAVAIL);
453 				--*lastport;
454 				if (*lastport > first || *lastport < last)
455 					*lastport = first;
456 				lport = htons(*lastport);
457 			} while (in_pcblookup_local(pcbinfo, laddr, lport,
458 			    wild));
459 		} else {
460 			/*
461 			 * counting up
462 			 */
463 			if (dorandom)
464 				*lastport = first +
465 					    (arc4random() % (last - first));
466 			count = last - first;
467 
468 			do {
469 				if (count-- < 0)	/* completely used? */
470 					return (EADDRNOTAVAIL);
471 				++*lastport;
472 				if (*lastport < first || *lastport > last)
473 					*lastport = first;
474 				lport = htons(*lastport);
475 			} while (in_pcblookup_local(pcbinfo, laddr, lport,
476 			    wild));
477 		}
478 	}
479 	if (prison_ip(cred, 0, &laddr.s_addr))
480 		return (EINVAL);
481 	*laddrp = laddr.s_addr;
482 	*lportp = lport;
483 	return (0);
484 }
485 
486 /*
487  * Connect from a socket to a specified address.
488  * Both address and port must be specified in argument sin.
489  * If don't have a local address for this socket yet,
490  * then pick one.
491  */
492 int
493 in_pcbconnect(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred)
494 {
495 	u_short lport, fport;
496 	in_addr_t laddr, faddr;
497 	int anonport, error;
498 
499 	INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
500 	INP_LOCK_ASSERT(inp);
501 
502 	lport = inp->inp_lport;
503 	laddr = inp->inp_laddr.s_addr;
504 	anonport = (lport == 0);
505 	error = in_pcbconnect_setup(inp, nam, &laddr, &lport, &faddr, &fport,
506 	    NULL, cred);
507 	if (error)
508 		return (error);
509 
510 	/* Do the initial binding of the local address if required. */
511 	if (inp->inp_laddr.s_addr == INADDR_ANY && inp->inp_lport == 0) {
512 		inp->inp_lport = lport;
513 		inp->inp_laddr.s_addr = laddr;
514 		if (in_pcbinshash(inp) != 0) {
515 			inp->inp_laddr.s_addr = INADDR_ANY;
516 			inp->inp_lport = 0;
517 			return (EAGAIN);
518 		}
519 	}
520 
521 	/* Commit the remaining changes. */
522 	inp->inp_lport = lport;
523 	inp->inp_laddr.s_addr = laddr;
524 	inp->inp_faddr.s_addr = faddr;
525 	inp->inp_fport = fport;
526 	in_pcbrehash(inp);
527 
528 	if (anonport)
529 		inp->inp_flags |= INP_ANONPORT;
530 	return (0);
531 }
532 
533 /*
534  * Set up for a connect from a socket to the specified address.
535  * On entry, *laddrp and *lportp should contain the current local
536  * address and port for the PCB; these are updated to the values
537  * that should be placed in inp_laddr and inp_lport to complete
538  * the connect.
539  *
540  * On success, *faddrp and *fportp will be set to the remote address
541  * and port. These are not updated in the error case.
542  *
543  * If the operation fails because the connection already exists,
544  * *oinpp will be set to the PCB of that connection so that the
545  * caller can decide to override it. In all other cases, *oinpp
546  * is set to NULL.
547  */
548 int
549 in_pcbconnect_setup(struct inpcb *inp, struct sockaddr *nam,
550     in_addr_t *laddrp, u_short *lportp, in_addr_t *faddrp, u_short *fportp,
551     struct inpcb **oinpp, struct ucred *cred)
552 {
553 	struct sockaddr_in *sin = (struct sockaddr_in *)nam;
554 	struct in_ifaddr *ia;
555 	struct sockaddr_in sa;
556 	struct ucred *socred;
557 	struct inpcb *oinp;
558 	struct in_addr laddr, faddr;
559 	u_short lport, fport;
560 	int error;
561 
562 	INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
563 	INP_LOCK_ASSERT(inp);
564 
565 	if (oinpp != NULL)
566 		*oinpp = NULL;
567 	if (nam->sa_len != sizeof (*sin))
568 		return (EINVAL);
569 	if (sin->sin_family != AF_INET)
570 		return (EAFNOSUPPORT);
571 	if (sin->sin_port == 0)
572 		return (EADDRNOTAVAIL);
573 	laddr.s_addr = *laddrp;
574 	lport = *lportp;
575 	faddr = sin->sin_addr;
576 	fport = sin->sin_port;
577 	socred = inp->inp_socket->so_cred;
578 	if (laddr.s_addr == INADDR_ANY && jailed(socred)) {
579 		bzero(&sa, sizeof(sa));
580 		sa.sin_addr.s_addr = htonl(prison_getip(socred));
581 		sa.sin_len = sizeof(sa);
582 		sa.sin_family = AF_INET;
583 		error = in_pcbbind_setup(inp, (struct sockaddr *)&sa,
584 		    &laddr.s_addr, &lport, cred);
585 		if (error)
586 			return (error);
587 	}
588 	if (!TAILQ_EMPTY(&in_ifaddrhead)) {
589 		/*
590 		 * If the destination address is INADDR_ANY,
591 		 * use the primary local address.
592 		 * If the supplied address is INADDR_BROADCAST,
593 		 * and the primary interface supports broadcast,
594 		 * choose the broadcast address for that interface.
595 		 */
596 		if (faddr.s_addr == INADDR_ANY)
597 			faddr = IA_SIN(TAILQ_FIRST(&in_ifaddrhead))->sin_addr;
598 		else if (faddr.s_addr == (u_long)INADDR_BROADCAST &&
599 		    (TAILQ_FIRST(&in_ifaddrhead)->ia_ifp->if_flags &
600 		    IFF_BROADCAST))
601 			faddr = satosin(&TAILQ_FIRST(
602 			    &in_ifaddrhead)->ia_broadaddr)->sin_addr;
603 	}
604 	if (laddr.s_addr == INADDR_ANY) {
605 		ia = (struct in_ifaddr *)0;
606 		/*
607 		 * If route is known our src addr is taken from the i/f,
608 		 * else punt.
609 		 *
610 		 * Find out route to destination
611 		 */
612 		if ((inp->inp_socket->so_options & SO_DONTROUTE) == 0)
613 			ia = ip_rtaddr(faddr);
614 		/*
615 		 * If we found a route, use the address corresponding to
616 		 * the outgoing interface.
617 		 *
618 		 * Otherwise assume faddr is reachable on a directly connected
619 		 * network and try to find a corresponding interface to take
620 		 * the source address from.
621 		 */
622 		if (ia == 0) {
623 			bzero(&sa, sizeof(sa));
624 			sa.sin_addr = faddr;
625 			sa.sin_len = sizeof(sa);
626 			sa.sin_family = AF_INET;
627 
628 			ia = ifatoia(ifa_ifwithdstaddr(sintosa(&sa)));
629 			if (ia == 0)
630 				ia = ifatoia(ifa_ifwithnet(sintosa(&sa)));
631 			if (ia == 0)
632 				return (ENETUNREACH);
633 		}
634 		/*
635 		 * If the destination address is multicast and an outgoing
636 		 * interface has been set as a multicast option, use the
637 		 * address of that interface as our source address.
638 		 */
639 		if (IN_MULTICAST(ntohl(faddr.s_addr)) &&
640 		    inp->inp_moptions != NULL) {
641 			struct ip_moptions *imo;
642 			struct ifnet *ifp;
643 
644 			imo = inp->inp_moptions;
645 			if (imo->imo_multicast_ifp != NULL) {
646 				ifp = imo->imo_multicast_ifp;
647 				TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link)
648 					if (ia->ia_ifp == ifp)
649 						break;
650 				if (ia == 0)
651 					return (EADDRNOTAVAIL);
652 			}
653 		}
654 		laddr = ia->ia_addr.sin_addr;
655 	}
656 
657 	oinp = in_pcblookup_hash(inp->inp_pcbinfo, faddr, fport, laddr, lport,
658 	    0, NULL);
659 	if (oinp != NULL) {
660 		if (oinpp != NULL)
661 			*oinpp = oinp;
662 		return (EADDRINUSE);
663 	}
664 	if (lport == 0) {
665 		error = in_pcbbind_setup(inp, NULL, &laddr.s_addr, &lport,
666 		    cred);
667 		if (error)
668 			return (error);
669 	}
670 	*laddrp = laddr.s_addr;
671 	*lportp = lport;
672 	*faddrp = faddr.s_addr;
673 	*fportp = fport;
674 	return (0);
675 }
676 
677 void
678 in_pcbdisconnect(struct inpcb *inp)
679 {
680 
681 	INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
682 	INP_LOCK_ASSERT(inp);
683 
684 	inp->inp_faddr.s_addr = INADDR_ANY;
685 	inp->inp_fport = 0;
686 	in_pcbrehash(inp);
687 }
688 
689 /*
690  * In the old world order, in_pcbdetach() served two functions: to detach the
691  * pcb from the socket/potentially free the socket, and to free the pcb
692  * itself.  In the new world order, the protocol code is responsible for
693  * managing the relationship with the socket, and this code simply frees the
694  * pcb.
695  */
696 void
697 in_pcbdetach(struct inpcb *inp)
698 {
699 
700 	KASSERT(inp->inp_socket != NULL, ("in_pcbdetach: inp_socket == NULL"));
701 	inp->inp_socket->so_pcb = NULL;
702 	inp->inp_socket = NULL;
703 }
704 
705 void
706 in_pcbfree(struct inpcb *inp)
707 {
708 	struct inpcbinfo *ipi = inp->inp_pcbinfo;
709 
710 	KASSERT(inp->inp_socket == NULL, ("in_pcbfree: inp_socket != NULL"));
711 	INP_INFO_WLOCK_ASSERT(ipi);
712 	INP_LOCK_ASSERT(inp);
713 
714 #ifdef IPSEC
715 	ipsec4_delete_pcbpolicy(inp);
716 #endif /*IPSEC*/
717 	inp->inp_gencnt = ++ipi->ipi_gencnt;
718 	in_pcbremlists(inp);
719 	if (inp->inp_options)
720 		(void)m_free(inp->inp_options);
721 	if (inp->inp_moptions != NULL)
722 		inp_freemoptions(inp->inp_moptions);
723 	inp->inp_vflag = 0;
724 
725 #ifdef MAC
726 	mac_destroy_inpcb(inp);
727 #endif
728 	INP_UNLOCK(inp);
729 	uma_zfree(ipi->ipi_zone, inp);
730 }
731 
732 /*
733  * TCP needs to maintain its inpcb structure after the TCP connection has
734  * been torn down.  However, it must be disconnected from the inpcb hashes as
735  * it must not prevent binding of future connections to the same port/ip
736  * combination by other inpcbs.
737  */
738 void
739 in_pcbdrop(struct inpcb *inp)
740 {
741 
742 	INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
743 	INP_LOCK_ASSERT(inp);
744 
745 	inp->inp_vflag |= INP_DROPPED;
746 	if (inp->inp_lport) {
747 		struct inpcbport *phd = inp->inp_phd;
748 
749 		LIST_REMOVE(inp, inp_hash);
750 		LIST_REMOVE(inp, inp_portlist);
751 		if (LIST_FIRST(&phd->phd_pcblist) == NULL) {
752 			LIST_REMOVE(phd, phd_hash);
753 			free(phd, M_PCB);
754 		}
755 		inp->inp_lport = 0;
756 	}
757 }
758 
759 /*
760  * Common routines to return the socket addresses associated with inpcbs.
761  */
762 struct sockaddr *
763 in_sockaddr(in_port_t port, struct in_addr *addr_p)
764 {
765 	struct sockaddr_in *sin;
766 
767 	MALLOC(sin, struct sockaddr_in *, sizeof *sin, M_SONAME,
768 		M_WAITOK | M_ZERO);
769 	sin->sin_family = AF_INET;
770 	sin->sin_len = sizeof(*sin);
771 	sin->sin_addr = *addr_p;
772 	sin->sin_port = port;
773 
774 	return (struct sockaddr *)sin;
775 }
776 
777 int
778 in_getsockaddr(struct socket *so, struct sockaddr **nam)
779 {
780 	struct inpcb *inp;
781 	struct in_addr addr;
782 	in_port_t port;
783 
784 	inp = sotoinpcb(so);
785 	KASSERT(inp != NULL, ("in_getsockaddr: inp == NULL"));
786 
787 	INP_LOCK(inp);
788 	port = inp->inp_lport;
789 	addr = inp->inp_laddr;
790 	INP_UNLOCK(inp);
791 
792 	*nam = in_sockaddr(port, &addr);
793 	return 0;
794 }
795 
796 int
797 in_getpeeraddr(struct socket *so, struct sockaddr **nam)
798 {
799 	struct inpcb *inp;
800 	struct in_addr addr;
801 	in_port_t port;
802 
803 	inp = sotoinpcb(so);
804 	KASSERT(inp != NULL, ("in_getpeeraddr: inp == NULL"));
805 
806 	INP_LOCK(inp);
807 	port = inp->inp_fport;
808 	addr = inp->inp_faddr;
809 	INP_UNLOCK(inp);
810 
811 	*nam = in_sockaddr(port, &addr);
812 	return 0;
813 }
814 
815 void
816 in_pcbnotifyall(struct inpcbinfo *pcbinfo, struct in_addr faddr, int errno,
817     struct inpcb *(*notify)(struct inpcb *, int))
818 {
819 	struct inpcb *inp, *ninp;
820 	struct inpcbhead *head;
821 
822 	INP_INFO_WLOCK(pcbinfo);
823 	head = pcbinfo->ipi_listhead;
824 	for (inp = LIST_FIRST(head); inp != NULL; inp = ninp) {
825 		INP_LOCK(inp);
826 		ninp = LIST_NEXT(inp, inp_list);
827 #ifdef INET6
828 		if ((inp->inp_vflag & INP_IPV4) == 0) {
829 			INP_UNLOCK(inp);
830 			continue;
831 		}
832 #endif
833 		if (inp->inp_faddr.s_addr != faddr.s_addr ||
834 		    inp->inp_socket == NULL) {
835 			INP_UNLOCK(inp);
836 			continue;
837 		}
838 		if ((*notify)(inp, errno))
839 			INP_UNLOCK(inp);
840 	}
841 	INP_INFO_WUNLOCK(pcbinfo);
842 }
843 
844 void
845 in_pcbpurgeif0(struct inpcbinfo *pcbinfo, struct ifnet *ifp)
846 {
847 	struct inpcb *inp;
848 	struct ip_moptions *imo;
849 	int i, gap;
850 
851 	INP_INFO_RLOCK(pcbinfo);
852 	LIST_FOREACH(inp, pcbinfo->ipi_listhead, inp_list) {
853 		INP_LOCK(inp);
854 		imo = inp->inp_moptions;
855 		if ((inp->inp_vflag & INP_IPV4) &&
856 		    imo != NULL) {
857 			/*
858 			 * Unselect the outgoing interface if it is being
859 			 * detached.
860 			 */
861 			if (imo->imo_multicast_ifp == ifp)
862 				imo->imo_multicast_ifp = NULL;
863 
864 			/*
865 			 * Drop multicast group membership if we joined
866 			 * through the interface being detached.
867 			 */
868 			for (i = 0, gap = 0; i < imo->imo_num_memberships;
869 			    i++) {
870 				if (imo->imo_membership[i]->inm_ifp == ifp) {
871 					in_delmulti(imo->imo_membership[i]);
872 					gap++;
873 				} else if (gap != 0)
874 					imo->imo_membership[i - gap] =
875 					    imo->imo_membership[i];
876 			}
877 			imo->imo_num_memberships -= gap;
878 		}
879 		INP_UNLOCK(inp);
880 	}
881 	INP_INFO_RUNLOCK(pcbinfo);
882 }
883 
884 /*
885  * Lookup a PCB based on the local address and port.
886  */
887 #define INP_LOOKUP_MAPPED_PCB_COST	3
888 struct inpcb *
889 in_pcblookup_local(struct inpcbinfo *pcbinfo, struct in_addr laddr,
890     u_int lport_arg, int wild_okay)
891 {
892 	struct inpcb *inp;
893 #ifdef INET6
894 	int matchwild = 3 + INP_LOOKUP_MAPPED_PCB_COST;
895 #else
896 	int matchwild = 3;
897 #endif
898 	int wildcard;
899 	u_short lport = lport_arg;
900 
901 	INP_INFO_WLOCK_ASSERT(pcbinfo);
902 
903 	if (!wild_okay) {
904 		struct inpcbhead *head;
905 		/*
906 		 * Look for an unconnected (wildcard foreign addr) PCB that
907 		 * matches the local address and port we're looking for.
908 		 */
909 		head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport,
910 		    0, pcbinfo->ipi_hashmask)];
911 		LIST_FOREACH(inp, head, inp_hash) {
912 #ifdef INET6
913 			if ((inp->inp_vflag & INP_IPV4) == 0)
914 				continue;
915 #endif
916 			if (inp->inp_faddr.s_addr == INADDR_ANY &&
917 			    inp->inp_laddr.s_addr == laddr.s_addr &&
918 			    inp->inp_lport == lport) {
919 				/*
920 				 * Found.
921 				 */
922 				return (inp);
923 			}
924 		}
925 		/*
926 		 * Not found.
927 		 */
928 		return (NULL);
929 	} else {
930 		struct inpcbporthead *porthash;
931 		struct inpcbport *phd;
932 		struct inpcb *match = NULL;
933 		/*
934 		 * Best fit PCB lookup.
935 		 *
936 		 * First see if this local port is in use by looking on the
937 		 * port hash list.
938 		 */
939 		porthash = &pcbinfo->ipi_porthashbase[INP_PCBPORTHASH(lport,
940 		    pcbinfo->ipi_porthashmask)];
941 		LIST_FOREACH(phd, porthash, phd_hash) {
942 			if (phd->phd_port == lport)
943 				break;
944 		}
945 		if (phd != NULL) {
946 			/*
947 			 * Port is in use by one or more PCBs. Look for best
948 			 * fit.
949 			 */
950 			LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) {
951 				wildcard = 0;
952 #ifdef INET6
953 				if ((inp->inp_vflag & INP_IPV4) == 0)
954 					continue;
955 				/*
956 				 * We never select the PCB that has
957 				 * INP_IPV6 flag and is bound to :: if
958 				 * we have another PCB which is bound
959 				 * to 0.0.0.0.  If a PCB has the
960 				 * INP_IPV6 flag, then we set its cost
961 				 * higher than IPv4 only PCBs.
962 				 *
963 				 * Note that the case only happens
964 				 * when a socket is bound to ::, under
965 				 * the condition that the use of the
966 				 * mapped address is allowed.
967 				 */
968 				if ((inp->inp_vflag & INP_IPV6) != 0)
969 					wildcard += INP_LOOKUP_MAPPED_PCB_COST;
970 #endif
971 				if (inp->inp_faddr.s_addr != INADDR_ANY)
972 					wildcard++;
973 				if (inp->inp_laddr.s_addr != INADDR_ANY) {
974 					if (laddr.s_addr == INADDR_ANY)
975 						wildcard++;
976 					else if (inp->inp_laddr.s_addr != laddr.s_addr)
977 						continue;
978 				} else {
979 					if (laddr.s_addr != INADDR_ANY)
980 						wildcard++;
981 				}
982 				if (wildcard < matchwild) {
983 					match = inp;
984 					matchwild = wildcard;
985 					if (matchwild == 0) {
986 						break;
987 					}
988 				}
989 			}
990 		}
991 		return (match);
992 	}
993 }
994 #undef INP_LOOKUP_MAPPED_PCB_COST
995 
996 /*
997  * Lookup PCB in hash list.
998  */
999 struct inpcb *
1000 in_pcblookup_hash(struct inpcbinfo *pcbinfo, struct in_addr faddr,
1001     u_int fport_arg, struct in_addr laddr, u_int lport_arg, int wildcard,
1002     struct ifnet *ifp)
1003 {
1004 	struct inpcbhead *head;
1005 	struct inpcb *inp;
1006 	u_short fport = fport_arg, lport = lport_arg;
1007 
1008 	INP_INFO_RLOCK_ASSERT(pcbinfo);
1009 
1010 	/*
1011 	 * First look for an exact match.
1012 	 */
1013 	head = &pcbinfo->ipi_hashbase[INP_PCBHASH(faddr.s_addr, lport, fport,
1014 	    pcbinfo->ipi_hashmask)];
1015 	LIST_FOREACH(inp, head, inp_hash) {
1016 #ifdef INET6
1017 		if ((inp->inp_vflag & INP_IPV4) == 0)
1018 			continue;
1019 #endif
1020 		if (inp->inp_faddr.s_addr == faddr.s_addr &&
1021 		    inp->inp_laddr.s_addr == laddr.s_addr &&
1022 		    inp->inp_fport == fport &&
1023 		    inp->inp_lport == lport)
1024 			return (inp);
1025 	}
1026 
1027 	/*
1028 	 * Then look for a wildcard match, if requested.
1029 	 */
1030 	if (wildcard) {
1031 		struct inpcb *local_wild = NULL;
1032 #ifdef INET6
1033 		struct inpcb *local_wild_mapped = NULL;
1034 #endif
1035 
1036 		head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport,
1037 		    0, pcbinfo->ipi_hashmask)];
1038 		LIST_FOREACH(inp, head, inp_hash) {
1039 #ifdef INET6
1040 			if ((inp->inp_vflag & INP_IPV4) == 0)
1041 				continue;
1042 #endif
1043 			if (inp->inp_faddr.s_addr == INADDR_ANY &&
1044 			    inp->inp_lport == lport) {
1045 				if (ifp && ifp->if_type == IFT_FAITH &&
1046 				    (inp->inp_flags & INP_FAITH) == 0)
1047 					continue;
1048 				if (inp->inp_laddr.s_addr == laddr.s_addr)
1049 					return (inp);
1050 				else if (inp->inp_laddr.s_addr == INADDR_ANY) {
1051 #ifdef INET6
1052 					if (INP_CHECK_SOCKAF(inp->inp_socket,
1053 							     AF_INET6))
1054 						local_wild_mapped = inp;
1055 					else
1056 #endif
1057 						local_wild = inp;
1058 				}
1059 			}
1060 		}
1061 #ifdef INET6
1062 		if (local_wild == NULL)
1063 			return (local_wild_mapped);
1064 #endif
1065 		return (local_wild);
1066 	}
1067 	return (NULL);
1068 }
1069 
1070 /*
1071  * Insert PCB onto various hash lists.
1072  */
1073 int
1074 in_pcbinshash(struct inpcb *inp)
1075 {
1076 	struct inpcbhead *pcbhash;
1077 	struct inpcbporthead *pcbporthash;
1078 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1079 	struct inpcbport *phd;
1080 	u_int32_t hashkey_faddr;
1081 
1082 	INP_INFO_WLOCK_ASSERT(pcbinfo);
1083 	INP_LOCK_ASSERT(inp);
1084 
1085 #ifdef INET6
1086 	if (inp->inp_vflag & INP_IPV6)
1087 		hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX */;
1088 	else
1089 #endif /* INET6 */
1090 	hashkey_faddr = inp->inp_faddr.s_addr;
1091 
1092 	pcbhash = &pcbinfo->ipi_hashbase[INP_PCBHASH(hashkey_faddr,
1093 		 inp->inp_lport, inp->inp_fport, pcbinfo->ipi_hashmask)];
1094 
1095 	pcbporthash = &pcbinfo->ipi_porthashbase[
1096 	    INP_PCBPORTHASH(inp->inp_lport, pcbinfo->ipi_porthashmask)];
1097 
1098 	/*
1099 	 * Go through port list and look for a head for this lport.
1100 	 */
1101 	LIST_FOREACH(phd, pcbporthash, phd_hash) {
1102 		if (phd->phd_port == inp->inp_lport)
1103 			break;
1104 	}
1105 	/*
1106 	 * If none exists, malloc one and tack it on.
1107 	 */
1108 	if (phd == NULL) {
1109 		MALLOC(phd, struct inpcbport *, sizeof(struct inpcbport), M_PCB, M_NOWAIT);
1110 		if (phd == NULL) {
1111 			return (ENOBUFS); /* XXX */
1112 		}
1113 		phd->phd_port = inp->inp_lport;
1114 		LIST_INIT(&phd->phd_pcblist);
1115 		LIST_INSERT_HEAD(pcbporthash, phd, phd_hash);
1116 	}
1117 	inp->inp_phd = phd;
1118 	LIST_INSERT_HEAD(&phd->phd_pcblist, inp, inp_portlist);
1119 	LIST_INSERT_HEAD(pcbhash, inp, inp_hash);
1120 	return (0);
1121 }
1122 
1123 /*
1124  * Move PCB to the proper hash bucket when { faddr, fport } have  been
1125  * changed. NOTE: This does not handle the case of the lport changing (the
1126  * hashed port list would have to be updated as well), so the lport must
1127  * not change after in_pcbinshash() has been called.
1128  */
1129 void
1130 in_pcbrehash(struct inpcb *inp)
1131 {
1132 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1133 	struct inpcbhead *head;
1134 	u_int32_t hashkey_faddr;
1135 
1136 	INP_INFO_WLOCK_ASSERT(pcbinfo);
1137 	INP_LOCK_ASSERT(inp);
1138 
1139 #ifdef INET6
1140 	if (inp->inp_vflag & INP_IPV6)
1141 		hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX */;
1142 	else
1143 #endif /* INET6 */
1144 	hashkey_faddr = inp->inp_faddr.s_addr;
1145 
1146 	head = &pcbinfo->ipi_hashbase[INP_PCBHASH(hashkey_faddr,
1147 		inp->inp_lport, inp->inp_fport, pcbinfo->ipi_hashmask)];
1148 
1149 	LIST_REMOVE(inp, inp_hash);
1150 	LIST_INSERT_HEAD(head, inp, inp_hash);
1151 }
1152 
1153 /*
1154  * Remove PCB from various lists.
1155  */
1156 void
1157 in_pcbremlists(struct inpcb *inp)
1158 {
1159 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1160 
1161 	INP_INFO_WLOCK_ASSERT(pcbinfo);
1162 	INP_LOCK_ASSERT(inp);
1163 
1164 	inp->inp_gencnt = ++pcbinfo->ipi_gencnt;
1165 	if (inp->inp_lport) {
1166 		struct inpcbport *phd = inp->inp_phd;
1167 
1168 		LIST_REMOVE(inp, inp_hash);
1169 		LIST_REMOVE(inp, inp_portlist);
1170 		if (LIST_FIRST(&phd->phd_pcblist) == NULL) {
1171 			LIST_REMOVE(phd, phd_hash);
1172 			free(phd, M_PCB);
1173 		}
1174 	}
1175 	LIST_REMOVE(inp, inp_list);
1176 	pcbinfo->ipi_count--;
1177 }
1178 
1179 /*
1180  * A set label operation has occurred at the socket layer, propagate the
1181  * label change into the in_pcb for the socket.
1182  */
1183 void
1184 in_pcbsosetlabel(struct socket *so)
1185 {
1186 #ifdef MAC
1187 	struct inpcb *inp;
1188 
1189 	inp = sotoinpcb(so);
1190 	KASSERT(inp != NULL, ("in_pcbsosetlabel: so->so_pcb == NULL"));
1191 
1192 	INP_LOCK(inp);
1193 	SOCK_LOCK(so);
1194 	mac_inpcb_sosetlabel(so, inp);
1195 	SOCK_UNLOCK(so);
1196 	INP_UNLOCK(inp);
1197 #endif
1198 }
1199 
1200 /*
1201  * ipport_tick runs once per second, determining if random port allocation
1202  * should be continued.  If more than ipport_randomcps ports have been
1203  * allocated in the last second, then we return to sequential port
1204  * allocation. We return to random allocation only once we drop below
1205  * ipport_randomcps for at least ipport_randomtime seconds.
1206  */
1207 void
1208 ipport_tick(void *xtp)
1209 {
1210 
1211 	if (ipport_tcpallocs <= ipport_tcplastcount + ipport_randomcps) {
1212 		if (ipport_stoprandom > 0)
1213 			ipport_stoprandom--;
1214 	} else
1215 		ipport_stoprandom = ipport_randomtime;
1216 	ipport_tcplastcount = ipport_tcpallocs;
1217 	callout_reset(&ipport_tick_callout, hz, ipport_tick, NULL);
1218 }
1219 
1220 #ifdef DDB
1221 static void
1222 db_print_indent(int indent)
1223 {
1224 	int i;
1225 
1226 	for (i = 0; i < indent; i++)
1227 		db_printf(" ");
1228 }
1229 
1230 static void
1231 db_print_inconninfo(struct in_conninfo *inc, const char *name, int indent)
1232 {
1233 	char faddr_str[48], laddr_str[48];
1234 
1235 	db_print_indent(indent);
1236 	db_printf("%s at %p\n", name, inc);
1237 
1238 	indent += 2;
1239 
1240 #ifdef INET6
1241 	if (inc->inc_flags == 1) {
1242 		/* IPv6. */
1243 		ip6_sprintf(laddr_str, &inc->inc6_laddr);
1244 		ip6_sprintf(faddr_str, &inc->inc6_faddr);
1245 	} else {
1246 #endif
1247 		/* IPv4. */
1248 		inet_ntoa_r(inc->inc_laddr, laddr_str);
1249 		inet_ntoa_r(inc->inc_faddr, faddr_str);
1250 #ifdef INET6
1251 	}
1252 #endif
1253 	db_print_indent(indent);
1254 	db_printf("inc_laddr %s   inc_lport %u\n", laddr_str,
1255 	    ntohs(inc->inc_lport));
1256 	db_print_indent(indent);
1257 	db_printf("inc_faddr %s   inc_fport %u\n", faddr_str,
1258 	    ntohs(inc->inc_fport));
1259 }
1260 
1261 static void
1262 db_print_inpflags(int inp_flags)
1263 {
1264 	int comma;
1265 
1266 	comma = 0;
1267 	if (inp_flags & INP_RECVOPTS) {
1268 		db_printf("%sINP_RECVOPTS", comma ? ", " : "");
1269 		comma = 1;
1270 	}
1271 	if (inp_flags & INP_RECVRETOPTS) {
1272 		db_printf("%sINP_RECVRETOPTS", comma ? ", " : "");
1273 		comma = 1;
1274 	}
1275 	if (inp_flags & INP_RECVDSTADDR) {
1276 		db_printf("%sINP_RECVDSTADDR", comma ? ", " : "");
1277 		comma = 1;
1278 	}
1279 	if (inp_flags & INP_HDRINCL) {
1280 		db_printf("%sINP_HDRINCL", comma ? ", " : "");
1281 		comma = 1;
1282 	}
1283 	if (inp_flags & INP_HIGHPORT) {
1284 		db_printf("%sINP_HIGHPORT", comma ? ", " : "");
1285 		comma = 1;
1286 	}
1287 	if (inp_flags & INP_LOWPORT) {
1288 		db_printf("%sINP_LOWPORT", comma ? ", " : "");
1289 		comma = 1;
1290 	}
1291 	if (inp_flags & INP_ANONPORT) {
1292 		db_printf("%sINP_ANONPORT", comma ? ", " : "");
1293 		comma = 1;
1294 	}
1295 	if (inp_flags & INP_RECVIF) {
1296 		db_printf("%sINP_RECVIF", comma ? ", " : "");
1297 		comma = 1;
1298 	}
1299 	if (inp_flags & INP_MTUDISC) {
1300 		db_printf("%sINP_MTUDISC", comma ? ", " : "");
1301 		comma = 1;
1302 	}
1303 	if (inp_flags & INP_FAITH) {
1304 		db_printf("%sINP_FAITH", comma ? ", " : "");
1305 		comma = 1;
1306 	}
1307 	if (inp_flags & INP_RECVTTL) {
1308 		db_printf("%sINP_RECVTTL", comma ? ", " : "");
1309 		comma = 1;
1310 	}
1311 	if (inp_flags & INP_DONTFRAG) {
1312 		db_printf("%sINP_DONTFRAG", comma ? ", " : "");
1313 		comma = 1;
1314 	}
1315 	if (inp_flags & IN6P_IPV6_V6ONLY) {
1316 		db_printf("%sIN6P_IPV6_V6ONLY", comma ? ", " : "");
1317 		comma = 1;
1318 	}
1319 	if (inp_flags & IN6P_PKTINFO) {
1320 		db_printf("%sIN6P_PKTINFO", comma ? ", " : "");
1321 		comma = 1;
1322 	}
1323 	if (inp_flags & IN6P_HOPLIMIT) {
1324 		db_printf("%sIN6P_HOPLIMIT", comma ? ", " : "");
1325 		comma = 1;
1326 	}
1327 	if (inp_flags & IN6P_HOPOPTS) {
1328 		db_printf("%sIN6P_HOPOPTS", comma ? ", " : "");
1329 		comma = 1;
1330 	}
1331 	if (inp_flags & IN6P_DSTOPTS) {
1332 		db_printf("%sIN6P_DSTOPTS", comma ? ", " : "");
1333 		comma = 1;
1334 	}
1335 	if (inp_flags & IN6P_RTHDR) {
1336 		db_printf("%sIN6P_RTHDR", comma ? ", " : "");
1337 		comma = 1;
1338 	}
1339 	if (inp_flags & IN6P_RTHDRDSTOPTS) {
1340 		db_printf("%sIN6P_RTHDRDSTOPTS", comma ? ", " : "");
1341 		comma = 1;
1342 	}
1343 	if (inp_flags & IN6P_TCLASS) {
1344 		db_printf("%sIN6P_TCLASS", comma ? ", " : "");
1345 		comma = 1;
1346 	}
1347 	if (inp_flags & IN6P_AUTOFLOWLABEL) {
1348 		db_printf("%sIN6P_AUTOFLOWLABEL", comma ? ", " : "");
1349 		comma = 1;
1350 	}
1351 	if (inp_flags & IN6P_RFC2292) {
1352 		db_printf("%sIN6P_RFC2292", comma ? ", " : "");
1353 		comma = 1;
1354 	}
1355 	if (inp_flags & IN6P_MTU) {
1356 		db_printf("IN6P_MTU%s", comma ? ", " : "");
1357 		comma = 1;
1358 	}
1359 }
1360 
1361 static void
1362 db_print_inpvflag(u_char inp_vflag)
1363 {
1364 	int comma;
1365 
1366 	comma = 0;
1367 	if (inp_vflag & INP_IPV4) {
1368 		db_printf("%sINP_IPV4", comma ? ", " : "");
1369 		comma  = 1;
1370 	}
1371 	if (inp_vflag & INP_IPV6) {
1372 		db_printf("%sINP_IPV6", comma ? ", " : "");
1373 		comma  = 1;
1374 	}
1375 	if (inp_vflag & INP_IPV6PROTO) {
1376 		db_printf("%sINP_IPV6PROTO", comma ? ", " : "");
1377 		comma  = 1;
1378 	}
1379 	if (inp_vflag & INP_TIMEWAIT) {
1380 		db_printf("%sINP_TIMEWAIT", comma ? ", " : "");
1381 		comma  = 1;
1382 	}
1383 	if (inp_vflag & INP_ONESBCAST) {
1384 		db_printf("%sINP_ONESBCAST", comma ? ", " : "");
1385 		comma  = 1;
1386 	}
1387 	if (inp_vflag & INP_DROPPED) {
1388 		db_printf("%sINP_DROPPED", comma ? ", " : "");
1389 		comma  = 1;
1390 	}
1391 	if (inp_vflag & INP_SOCKREF) {
1392 		db_printf("%sINP_SOCKREF", comma ? ", " : "");
1393 		comma  = 1;
1394 	}
1395 }
1396 
1397 void
1398 db_print_inpcb(struct inpcb *inp, const char *name, int indent)
1399 {
1400 
1401 	db_print_indent(indent);
1402 	db_printf("%s at %p\n", name, inp);
1403 
1404 	indent += 2;
1405 
1406 	db_print_indent(indent);
1407 	db_printf("inp_flow: 0x%x\n", inp->inp_flow);
1408 
1409 	db_print_inconninfo(&inp->inp_inc, "inp_conninfo", indent);
1410 
1411 	db_print_indent(indent);
1412 	db_printf("inp_ppcb: %p   inp_pcbinfo: %p   inp_socket: %p\n",
1413 	    inp->inp_ppcb, inp->inp_pcbinfo, inp->inp_socket);
1414 
1415 	db_print_indent(indent);
1416 	db_printf("inp_label: %p   inp_flags: 0x%x (",
1417 	   inp->inp_label, inp->inp_flags);
1418 	db_print_inpflags(inp->inp_flags);
1419 	db_printf(")\n");
1420 
1421 	db_print_indent(indent);
1422 	db_printf("inp_sp: %p   inp_vflag: 0x%x (", inp->inp_sp,
1423 	    inp->inp_vflag);
1424 	db_print_inpvflag(inp->inp_vflag);
1425 	db_printf(")\n");
1426 
1427 	db_print_indent(indent);
1428 	db_printf("inp_ip_ttl: %d   inp_ip_p: %d   inp_ip_minttl: %d\n",
1429 	    inp->inp_ip_ttl, inp->inp_ip_p, inp->inp_ip_minttl);
1430 
1431 	db_print_indent(indent);
1432 #ifdef INET6
1433 	if (inp->inp_vflag & INP_IPV6) {
1434 		db_printf("in6p_options: %p   in6p_outputopts: %p   "
1435 		    "in6p_moptions: %p\n", inp->in6p_options,
1436 		    inp->in6p_outputopts, inp->in6p_moptions);
1437 		db_printf("in6p_icmp6filt: %p   in6p_cksum %d   "
1438 		    "in6p_hops %u\n", inp->in6p_icmp6filt, inp->in6p_cksum,
1439 		    inp->in6p_hops);
1440 	} else
1441 #endif
1442 	{
1443 		db_printf("inp_ip_tos: %d   inp_ip_options: %p   "
1444 		    "inp_ip_moptions: %p\n", inp->inp_ip_tos,
1445 		    inp->inp_options, inp->inp_moptions);
1446 	}
1447 
1448 	db_print_indent(indent);
1449 	db_printf("inp_phd: %p   inp_gencnt: %ju\n", inp->inp_phd,
1450 	    (uintmax_t)inp->inp_gencnt);
1451 }
1452 
1453 DB_SHOW_COMMAND(inpcb, db_show_inpcb)
1454 {
1455 	struct inpcb *inp;
1456 
1457 	if (!have_addr) {
1458 		db_printf("usage: show inpcb <addr>\n");
1459 		return;
1460 	}
1461 	inp = (struct inpcb *)addr;
1462 
1463 	db_print_inpcb(inp, "inpcb", 0);
1464 }
1465 #endif
1466