xref: /freebsd/sys/netinet/in_pcb.c (revision 94942af266ac119ede0ca836f9aa5a5ac0582938)
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 #ifdef IPSEC
77 #include <netinet6/ipsec.h>
78 #include <netkey/key.h>
79 #endif /* IPSEC */
80 
81 #ifdef FAST_IPSEC
82 #if defined(IPSEC) || defined(IPSEC_ESP)
83 #error "Bad idea: don't compile with both IPSEC and FAST_IPSEC!"
84 #endif
85 
86 #include <netipsec/ipsec.h>
87 #include <netipsec/key.h>
88 #endif /* FAST_IPSEC */
89 
90 #include <security/mac/mac_framework.h>
91 
92 /*
93  * These configure the range of local port addresses assigned to
94  * "unspecified" outgoing connections/packets/whatever.
95  */
96 int	ipport_lowfirstauto  = IPPORT_RESERVED - 1;	/* 1023 */
97 int	ipport_lowlastauto = IPPORT_RESERVEDSTART;	/* 600 */
98 int	ipport_firstauto = IPPORT_HIFIRSTAUTO;		/* 49152 */
99 int	ipport_lastauto  = IPPORT_HILASTAUTO;		/* 65535 */
100 int	ipport_hifirstauto = IPPORT_HIFIRSTAUTO;	/* 49152 */
101 int	ipport_hilastauto  = IPPORT_HILASTAUTO;		/* 65535 */
102 
103 /*
104  * Reserved ports accessible only to root. There are significant
105  * security considerations that must be accounted for when changing these,
106  * but the security benefits can be great. Please be careful.
107  */
108 int	ipport_reservedhigh = IPPORT_RESERVED - 1;	/* 1023 */
109 int	ipport_reservedlow = 0;
110 
111 /* Variables dealing with random ephemeral port allocation. */
112 int	ipport_randomized = 1;	/* user controlled via sysctl */
113 int	ipport_randomcps = 10;	/* user controlled via sysctl */
114 int	ipport_randomtime = 45;	/* user controlled via sysctl */
115 int	ipport_stoprandom = 0;	/* toggled by ipport_tick */
116 int	ipport_tcpallocs;
117 int	ipport_tcplastcount;
118 
119 #define RANGECHK(var, min, max) \
120 	if ((var) < (min)) { (var) = (min); } \
121 	else if ((var) > (max)) { (var) = (max); }
122 
123 static int
124 sysctl_net_ipport_check(SYSCTL_HANDLER_ARGS)
125 {
126 	int error;
127 
128 	error = sysctl_handle_int(oidp, oidp->oid_arg1, oidp->oid_arg2, req);
129 	if (error == 0) {
130 		RANGECHK(ipport_lowfirstauto, 1, IPPORT_RESERVED - 1);
131 		RANGECHK(ipport_lowlastauto, 1, IPPORT_RESERVED - 1);
132 		RANGECHK(ipport_firstauto, IPPORT_RESERVED, IPPORT_MAX);
133 		RANGECHK(ipport_lastauto, IPPORT_RESERVED, IPPORT_MAX);
134 		RANGECHK(ipport_hifirstauto, IPPORT_RESERVED, IPPORT_MAX);
135 		RANGECHK(ipport_hilastauto, IPPORT_RESERVED, IPPORT_MAX);
136 	}
137 	return (error);
138 }
139 
140 #undef RANGECHK
141 
142 SYSCTL_NODE(_net_inet_ip, IPPROTO_IP, portrange, CTLFLAG_RW, 0, "IP Ports");
143 
144 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, lowfirst, CTLTYPE_INT|CTLFLAG_RW,
145 	   &ipport_lowfirstauto, 0, &sysctl_net_ipport_check, "I", "");
146 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, lowlast, CTLTYPE_INT|CTLFLAG_RW,
147 	   &ipport_lowlastauto, 0, &sysctl_net_ipport_check, "I", "");
148 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, first, CTLTYPE_INT|CTLFLAG_RW,
149 	   &ipport_firstauto, 0, &sysctl_net_ipport_check, "I", "");
150 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, last, CTLTYPE_INT|CTLFLAG_RW,
151 	   &ipport_lastauto, 0, &sysctl_net_ipport_check, "I", "");
152 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, hifirst, CTLTYPE_INT|CTLFLAG_RW,
153 	   &ipport_hifirstauto, 0, &sysctl_net_ipport_check, "I", "");
154 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, hilast, CTLTYPE_INT|CTLFLAG_RW,
155 	   &ipport_hilastauto, 0, &sysctl_net_ipport_check, "I", "");
156 SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, reservedhigh,
157 	   CTLFLAG_RW|CTLFLAG_SECURE, &ipport_reservedhigh, 0, "");
158 SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, reservedlow,
159 	   CTLFLAG_RW|CTLFLAG_SECURE, &ipport_reservedlow, 0, "");
160 SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, randomized, CTLFLAG_RW,
161 	   &ipport_randomized, 0, "Enable random port allocation");
162 SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, randomcps, CTLFLAG_RW,
163 	   &ipport_randomcps, 0, "Maximum number of random port "
164 	   "allocations before switching to a sequental one");
165 SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, randomtime, CTLFLAG_RW,
166 	   &ipport_randomtime, 0, "Minimum time to keep sequental port "
167 	   "allocation before switching to a random one");
168 
169 /*
170  * in_pcb.c: manage the Protocol Control Blocks.
171  *
172  * NOTE: It is assumed that most of these functions will be called with
173  * the pcbinfo lock held, and often, the inpcb lock held, as these utility
174  * functions often modify hash chains or addresses in pcbs.
175  */
176 
177 /*
178  * Allocate a PCB and associate it with the socket.
179  * On success return with the PCB locked.
180  */
181 int
182 in_pcballoc(struct socket *so, struct inpcbinfo *pcbinfo)
183 {
184 	struct inpcb *inp;
185 	int error;
186 
187 	INP_INFO_WLOCK_ASSERT(pcbinfo);
188 	error = 0;
189 	inp = uma_zalloc(pcbinfo->ipi_zone, M_NOWAIT);
190 	if (inp == NULL)
191 		return (ENOBUFS);
192 	bzero(inp, inp_zero_size);
193 	inp->inp_pcbinfo = pcbinfo;
194 	inp->inp_socket = so;
195 #ifdef MAC
196 	error = mac_init_inpcb(inp, M_NOWAIT);
197 	if (error != 0)
198 		goto out;
199 	SOCK_LOCK(so);
200 	mac_create_inpcb_from_socket(so, inp);
201 	SOCK_UNLOCK(so);
202 #endif
203 #if defined(IPSEC) || defined(FAST_IPSEC)
204 #ifdef FAST_IPSEC
205 	error = ipsec_init_policy(so, &inp->inp_sp);
206 #else
207 	error = ipsec_init_pcbpolicy(so, &inp->inp_sp);
208 #endif
209 	if (error != 0)
210 		goto out;
211 #endif /*IPSEC*/
212 #ifdef INET6
213 	if (INP_SOCKAF(so) == AF_INET6) {
214 		inp->inp_vflag |= INP_IPV6PROTO;
215 		if (ip6_v6only)
216 			inp->inp_flags |= IN6P_IPV6_V6ONLY;
217 	}
218 #endif
219 	LIST_INSERT_HEAD(pcbinfo->ipi_listhead, inp, inp_list);
220 	pcbinfo->ipi_count++;
221 	so->so_pcb = (caddr_t)inp;
222 #ifdef INET6
223 	if (ip6_auto_flowlabel)
224 		inp->inp_flags |= IN6P_AUTOFLOWLABEL;
225 #endif
226 	INP_LOCK(inp);
227 	inp->inp_gencnt = ++pcbinfo->ipi_gencnt;
228 
229 #if defined(IPSEC) || defined(FAST_IPSEC) || defined(MAC)
230 out:
231 	if (error != 0)
232 		uma_zfree(pcbinfo->ipi_zone, inp);
233 #endif
234 	return (error);
235 }
236 
237 int
238 in_pcbbind(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred)
239 {
240 	int anonport, error;
241 
242 	INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
243 	INP_LOCK_ASSERT(inp);
244 
245 	if (inp->inp_lport != 0 || inp->inp_laddr.s_addr != INADDR_ANY)
246 		return (EINVAL);
247 	anonport = inp->inp_lport == 0 && (nam == NULL ||
248 	    ((struct sockaddr_in *)nam)->sin_port == 0);
249 	error = in_pcbbind_setup(inp, nam, &inp->inp_laddr.s_addr,
250 	    &inp->inp_lport, cred);
251 	if (error)
252 		return (error);
253 	if (in_pcbinshash(inp) != 0) {
254 		inp->inp_laddr.s_addr = INADDR_ANY;
255 		inp->inp_lport = 0;
256 		return (EAGAIN);
257 	}
258 	if (anonport)
259 		inp->inp_flags |= INP_ANONPORT;
260 	return (0);
261 }
262 
263 /*
264  * Set up a bind operation on a PCB, performing port allocation
265  * as required, but do not actually modify the PCB. Callers can
266  * either complete the bind by setting inp_laddr/inp_lport and
267  * calling in_pcbinshash(), or they can just use the resulting
268  * port and address to authorise the sending of a once-off packet.
269  *
270  * On error, the values of *laddrp and *lportp are not changed.
271  */
272 int
273 in_pcbbind_setup(struct inpcb *inp, struct sockaddr *nam, in_addr_t *laddrp,
274     u_short *lportp, struct ucred *cred)
275 {
276 	struct socket *so = inp->inp_socket;
277 	unsigned short *lastport;
278 	struct sockaddr_in *sin;
279 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
280 	struct in_addr laddr;
281 	u_short lport = 0;
282 	int wild = 0, reuseport = (so->so_options & SO_REUSEPORT);
283 	int error, prison = 0;
284 	int dorandom;
285 
286 	INP_INFO_WLOCK_ASSERT(pcbinfo);
287 	INP_LOCK_ASSERT(inp);
288 
289 	if (TAILQ_EMPTY(&in_ifaddrhead)) /* XXX broken! */
290 		return (EADDRNOTAVAIL);
291 	laddr.s_addr = *laddrp;
292 	if (nam != NULL && laddr.s_addr != INADDR_ANY)
293 		return (EINVAL);
294 	if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0)
295 		wild = INPLOOKUP_WILDCARD;
296 	if (nam) {
297 		sin = (struct sockaddr_in *)nam;
298 		if (nam->sa_len != sizeof (*sin))
299 			return (EINVAL);
300 #ifdef notdef
301 		/*
302 		 * We should check the family, but old programs
303 		 * incorrectly fail to initialize it.
304 		 */
305 		if (sin->sin_family != AF_INET)
306 			return (EAFNOSUPPORT);
307 #endif
308 		if (sin->sin_addr.s_addr != INADDR_ANY)
309 			if (prison_ip(cred, 0, &sin->sin_addr.s_addr))
310 				return(EINVAL);
311 		if (sin->sin_port != *lportp) {
312 			/* Don't allow the port to change. */
313 			if (*lportp != 0)
314 				return (EINVAL);
315 			lport = sin->sin_port;
316 		}
317 		/* NB: lport is left as 0 if the port isn't being changed. */
318 		if (IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) {
319 			/*
320 			 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
321 			 * allow complete duplication of binding if
322 			 * SO_REUSEPORT is set, or if SO_REUSEADDR is set
323 			 * and a multicast address is bound on both
324 			 * new and duplicated sockets.
325 			 */
326 			if (so->so_options & SO_REUSEADDR)
327 				reuseport = SO_REUSEADDR|SO_REUSEPORT;
328 		} else if (sin->sin_addr.s_addr != INADDR_ANY) {
329 			sin->sin_port = 0;		/* yech... */
330 			bzero(&sin->sin_zero, sizeof(sin->sin_zero));
331 			if (ifa_ifwithaddr((struct sockaddr *)sin) == 0)
332 				return (EADDRNOTAVAIL);
333 		}
334 		laddr = sin->sin_addr;
335 		if (lport) {
336 			struct inpcb *t;
337 			struct tcptw *tw;
338 
339 			/* GROSS */
340 			if (ntohs(lport) <= ipport_reservedhigh &&
341 			    ntohs(lport) >= ipport_reservedlow &&
342 			    priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT,
343 			    SUSER_ALLOWJAIL))
344 				return (EACCES);
345 			if (jailed(cred))
346 				prison = 1;
347 			if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)) &&
348 			    priv_check_cred(so->so_cred,
349 			    PRIV_NETINET_REUSEPORT, SUSER_ALLOWJAIL) != 0) {
350 				t = in_pcblookup_local(inp->inp_pcbinfo,
351 				    sin->sin_addr, lport,
352 				    prison ? 0 :  INPLOOKUP_WILDCARD);
353 	/*
354 	 * XXX
355 	 * This entire block sorely needs a rewrite.
356 	 */
357 				if (t &&
358 				    ((t->inp_vflag & INP_TIMEWAIT) == 0) &&
359 				    (so->so_type != SOCK_STREAM ||
360 				     ntohl(t->inp_faddr.s_addr) == INADDR_ANY) &&
361 				    (ntohl(sin->sin_addr.s_addr) != INADDR_ANY ||
362 				     ntohl(t->inp_laddr.s_addr) != INADDR_ANY ||
363 				     (t->inp_socket->so_options &
364 					 SO_REUSEPORT) == 0) &&
365 				    (so->so_cred->cr_uid !=
366 				     t->inp_socket->so_cred->cr_uid))
367 					return (EADDRINUSE);
368 			}
369 			if (prison && prison_ip(cred, 0, &sin->sin_addr.s_addr))
370 				return (EADDRNOTAVAIL);
371 			t = in_pcblookup_local(pcbinfo, sin->sin_addr,
372 			    lport, prison ? 0 : wild);
373 			if (t && (t->inp_vflag & INP_TIMEWAIT)) {
374 				/*
375 				 * XXXRW: If an incpb has had its timewait
376 				 * state recycled, we treat the address as
377 				 * being in use (for now).  This is better
378 				 * than a panic, but not desirable.
379 				 */
380 				tw = intotw(inp);
381 				if (tw == NULL ||
382 				    (reuseport & tw->tw_so_options) == 0)
383 					return (EADDRINUSE);
384 			} else if (t &&
385 			    (reuseport & t->inp_socket->so_options) == 0) {
386 #ifdef INET6
387 				if (ntohl(sin->sin_addr.s_addr) !=
388 				    INADDR_ANY ||
389 				    ntohl(t->inp_laddr.s_addr) !=
390 				    INADDR_ANY ||
391 				    INP_SOCKAF(so) ==
392 				    INP_SOCKAF(t->inp_socket))
393 #endif
394 				return (EADDRINUSE);
395 			}
396 		}
397 	}
398 	if (*lportp != 0)
399 		lport = *lportp;
400 	if (lport == 0) {
401 		u_short first, last;
402 		int count;
403 
404 		if (laddr.s_addr != INADDR_ANY)
405 			if (prison_ip(cred, 0, &laddr.s_addr))
406 				return (EINVAL);
407 
408 		if (inp->inp_flags & INP_HIGHPORT) {
409 			first = ipport_hifirstauto;	/* sysctl */
410 			last  = ipport_hilastauto;
411 			lastport = &pcbinfo->ipi_lasthi;
412 		} else if (inp->inp_flags & INP_LOWPORT) {
413 			error = priv_check_cred(cred,
414 			    PRIV_NETINET_RESERVEDPORT, SUSER_ALLOWJAIL);
415 			if (error)
416 				return error;
417 			first = ipport_lowfirstauto;	/* 1023 */
418 			last  = ipport_lowlastauto;	/* 600 */
419 			lastport = &pcbinfo->ipi_lastlow;
420 		} else {
421 			first = ipport_firstauto;	/* sysctl */
422 			last  = ipport_lastauto;
423 			lastport = &pcbinfo->ipi_lastport;
424 		}
425 		/*
426 		 * For UDP, use random port allocation as long as the user
427 		 * allows it.  For TCP (and as of yet unknown) connections,
428 		 * use random port allocation only if the user allows it AND
429 		 * ipport_tick() allows it.
430 		 */
431 		if (ipport_randomized &&
432 			(!ipport_stoprandom || pcbinfo == &udbinfo))
433 			dorandom = 1;
434 		else
435 			dorandom = 0;
436 		/*
437 		 * It makes no sense to do random port allocation if
438 		 * we have the only port available.
439 		 */
440 		if (first == last)
441 			dorandom = 0;
442 		/* Make sure to not include UDP packets in the count. */
443 		if (pcbinfo != &udbinfo)
444 			ipport_tcpallocs++;
445 		/*
446 		 * Simple check to ensure all ports are not used up causing
447 		 * a deadlock here.
448 		 *
449 		 * We split the two cases (up and down) so that the direction
450 		 * is not being tested on each round of the loop.
451 		 */
452 		if (first > last) {
453 			/*
454 			 * counting down
455 			 */
456 			if (dorandom)
457 				*lastport = first -
458 					    (arc4random() % (first - last));
459 			count = first - last;
460 
461 			do {
462 				if (count-- < 0)	/* completely used? */
463 					return (EADDRNOTAVAIL);
464 				--*lastport;
465 				if (*lastport > first || *lastport < last)
466 					*lastport = first;
467 				lport = htons(*lastport);
468 			} while (in_pcblookup_local(pcbinfo, laddr, lport,
469 			    wild));
470 		} else {
471 			/*
472 			 * counting up
473 			 */
474 			if (dorandom)
475 				*lastport = first +
476 					    (arc4random() % (last - first));
477 			count = last - first;
478 
479 			do {
480 				if (count-- < 0)	/* completely used? */
481 					return (EADDRNOTAVAIL);
482 				++*lastport;
483 				if (*lastport < first || *lastport > last)
484 					*lastport = first;
485 				lport = htons(*lastport);
486 			} while (in_pcblookup_local(pcbinfo, laddr, lport,
487 			    wild));
488 		}
489 	}
490 	if (prison_ip(cred, 0, &laddr.s_addr))
491 		return (EINVAL);
492 	*laddrp = laddr.s_addr;
493 	*lportp = lport;
494 	return (0);
495 }
496 
497 /*
498  * Connect from a socket to a specified address.
499  * Both address and port must be specified in argument sin.
500  * If don't have a local address for this socket yet,
501  * then pick one.
502  */
503 int
504 in_pcbconnect(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred)
505 {
506 	u_short lport, fport;
507 	in_addr_t laddr, faddr;
508 	int anonport, error;
509 
510 	INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
511 	INP_LOCK_ASSERT(inp);
512 
513 	lport = inp->inp_lport;
514 	laddr = inp->inp_laddr.s_addr;
515 	anonport = (lport == 0);
516 	error = in_pcbconnect_setup(inp, nam, &laddr, &lport, &faddr, &fport,
517 	    NULL, cred);
518 	if (error)
519 		return (error);
520 
521 	/* Do the initial binding of the local address if required. */
522 	if (inp->inp_laddr.s_addr == INADDR_ANY && inp->inp_lport == 0) {
523 		inp->inp_lport = lport;
524 		inp->inp_laddr.s_addr = laddr;
525 		if (in_pcbinshash(inp) != 0) {
526 			inp->inp_laddr.s_addr = INADDR_ANY;
527 			inp->inp_lport = 0;
528 			return (EAGAIN);
529 		}
530 	}
531 
532 	/* Commit the remaining changes. */
533 	inp->inp_lport = lport;
534 	inp->inp_laddr.s_addr = laddr;
535 	inp->inp_faddr.s_addr = faddr;
536 	inp->inp_fport = fport;
537 	in_pcbrehash(inp);
538 #ifdef IPSEC
539 	if (inp->inp_socket->so_type == SOCK_STREAM)
540 		ipsec_pcbconn(inp->inp_sp);
541 #endif
542 	if (anonport)
543 		inp->inp_flags |= INP_ANONPORT;
544 	return (0);
545 }
546 
547 /*
548  * Set up for a connect from a socket to the specified address.
549  * On entry, *laddrp and *lportp should contain the current local
550  * address and port for the PCB; these are updated to the values
551  * that should be placed in inp_laddr and inp_lport to complete
552  * the connect.
553  *
554  * On success, *faddrp and *fportp will be set to the remote address
555  * and port. These are not updated in the error case.
556  *
557  * If the operation fails because the connection already exists,
558  * *oinpp will be set to the PCB of that connection so that the
559  * caller can decide to override it. In all other cases, *oinpp
560  * is set to NULL.
561  */
562 int
563 in_pcbconnect_setup(struct inpcb *inp, struct sockaddr *nam,
564     in_addr_t *laddrp, u_short *lportp, in_addr_t *faddrp, u_short *fportp,
565     struct inpcb **oinpp, struct ucred *cred)
566 {
567 	struct sockaddr_in *sin = (struct sockaddr_in *)nam;
568 	struct in_ifaddr *ia;
569 	struct sockaddr_in sa;
570 	struct ucred *socred;
571 	struct inpcb *oinp;
572 	struct in_addr laddr, faddr;
573 	u_short lport, fport;
574 	int error;
575 
576 	INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
577 	INP_LOCK_ASSERT(inp);
578 
579 	if (oinpp != NULL)
580 		*oinpp = NULL;
581 	if (nam->sa_len != sizeof (*sin))
582 		return (EINVAL);
583 	if (sin->sin_family != AF_INET)
584 		return (EAFNOSUPPORT);
585 	if (sin->sin_port == 0)
586 		return (EADDRNOTAVAIL);
587 	laddr.s_addr = *laddrp;
588 	lport = *lportp;
589 	faddr = sin->sin_addr;
590 	fport = sin->sin_port;
591 	socred = inp->inp_socket->so_cred;
592 	if (laddr.s_addr == INADDR_ANY && jailed(socred)) {
593 		bzero(&sa, sizeof(sa));
594 		sa.sin_addr.s_addr = htonl(prison_getip(socred));
595 		sa.sin_len = sizeof(sa);
596 		sa.sin_family = AF_INET;
597 		error = in_pcbbind_setup(inp, (struct sockaddr *)&sa,
598 		    &laddr.s_addr, &lport, cred);
599 		if (error)
600 			return (error);
601 	}
602 	if (!TAILQ_EMPTY(&in_ifaddrhead)) {
603 		/*
604 		 * If the destination address is INADDR_ANY,
605 		 * use the primary local address.
606 		 * If the supplied address is INADDR_BROADCAST,
607 		 * and the primary interface supports broadcast,
608 		 * choose the broadcast address for that interface.
609 		 */
610 		if (faddr.s_addr == INADDR_ANY)
611 			faddr = IA_SIN(TAILQ_FIRST(&in_ifaddrhead))->sin_addr;
612 		else if (faddr.s_addr == (u_long)INADDR_BROADCAST &&
613 		    (TAILQ_FIRST(&in_ifaddrhead)->ia_ifp->if_flags &
614 		    IFF_BROADCAST))
615 			faddr = satosin(&TAILQ_FIRST(
616 			    &in_ifaddrhead)->ia_broadaddr)->sin_addr;
617 	}
618 	if (laddr.s_addr == INADDR_ANY) {
619 		ia = (struct in_ifaddr *)0;
620 		/*
621 		 * If route is known our src addr is taken from the i/f,
622 		 * else punt.
623 		 *
624 		 * Find out route to destination
625 		 */
626 		if ((inp->inp_socket->so_options & SO_DONTROUTE) == 0)
627 			ia = ip_rtaddr(faddr);
628 		/*
629 		 * If we found a route, use the address corresponding to
630 		 * the outgoing interface.
631 		 *
632 		 * Otherwise assume faddr is reachable on a directly connected
633 		 * network and try to find a corresponding interface to take
634 		 * the source address from.
635 		 */
636 		if (ia == 0) {
637 			bzero(&sa, sizeof(sa));
638 			sa.sin_addr = faddr;
639 			sa.sin_len = sizeof(sa);
640 			sa.sin_family = AF_INET;
641 
642 			ia = ifatoia(ifa_ifwithdstaddr(sintosa(&sa)));
643 			if (ia == 0)
644 				ia = ifatoia(ifa_ifwithnet(sintosa(&sa)));
645 			if (ia == 0)
646 				return (ENETUNREACH);
647 		}
648 		/*
649 		 * If the destination address is multicast and an outgoing
650 		 * interface has been set as a multicast option, use the
651 		 * address of that interface as our source address.
652 		 */
653 		if (IN_MULTICAST(ntohl(faddr.s_addr)) &&
654 		    inp->inp_moptions != NULL) {
655 			struct ip_moptions *imo;
656 			struct ifnet *ifp;
657 
658 			imo = inp->inp_moptions;
659 			if (imo->imo_multicast_ifp != NULL) {
660 				ifp = imo->imo_multicast_ifp;
661 				TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link)
662 					if (ia->ia_ifp == ifp)
663 						break;
664 				if (ia == 0)
665 					return (EADDRNOTAVAIL);
666 			}
667 		}
668 		laddr = ia->ia_addr.sin_addr;
669 	}
670 
671 	oinp = in_pcblookup_hash(inp->inp_pcbinfo, faddr, fport, laddr, lport,
672 	    0, NULL);
673 	if (oinp != NULL) {
674 		if (oinpp != NULL)
675 			*oinpp = oinp;
676 		return (EADDRINUSE);
677 	}
678 	if (lport == 0) {
679 		error = in_pcbbind_setup(inp, NULL, &laddr.s_addr, &lport,
680 		    cred);
681 		if (error)
682 			return (error);
683 	}
684 	*laddrp = laddr.s_addr;
685 	*lportp = lport;
686 	*faddrp = faddr.s_addr;
687 	*fportp = fport;
688 	return (0);
689 }
690 
691 void
692 in_pcbdisconnect(struct inpcb *inp)
693 {
694 
695 	INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
696 	INP_LOCK_ASSERT(inp);
697 
698 	inp->inp_faddr.s_addr = INADDR_ANY;
699 	inp->inp_fport = 0;
700 	in_pcbrehash(inp);
701 #ifdef IPSEC
702 	ipsec_pcbdisconn(inp->inp_sp);
703 #endif
704 }
705 
706 /*
707  * In the old world order, in_pcbdetach() served two functions: to detach the
708  * pcb from the socket/potentially free the socket, and to free the pcb
709  * itself.  In the new world order, the protocol code is responsible for
710  * managing the relationship with the socket, and this code simply frees the
711  * pcb.
712  */
713 void
714 in_pcbdetach(struct inpcb *inp)
715 {
716 
717 	KASSERT(inp->inp_socket != NULL, ("in_pcbdetach: inp_socket == NULL"));
718 	inp->inp_socket->so_pcb = NULL;
719 	inp->inp_socket = NULL;
720 }
721 
722 void
723 in_pcbfree(struct inpcb *inp)
724 {
725 	struct inpcbinfo *ipi = inp->inp_pcbinfo;
726 
727 	KASSERT(inp->inp_socket == NULL, ("in_pcbfree: inp_socket != NULL"));
728 	INP_INFO_WLOCK_ASSERT(ipi);
729 	INP_LOCK_ASSERT(inp);
730 
731 #if defined(IPSEC) || defined(FAST_IPSEC)
732 	ipsec4_delete_pcbpolicy(inp);
733 #endif /*IPSEC*/
734 	inp->inp_gencnt = ++ipi->ipi_gencnt;
735 	in_pcbremlists(inp);
736 	if (inp->inp_options)
737 		(void)m_free(inp->inp_options);
738 	ip_freemoptions(inp->inp_moptions);
739 	inp->inp_vflag = 0;
740 
741 #ifdef MAC
742 	mac_destroy_inpcb(inp);
743 #endif
744 	INP_UNLOCK(inp);
745 	uma_zfree(ipi->ipi_zone, inp);
746 }
747 
748 /*
749  * TCP needs to maintain its inpcb structure after the TCP connection has
750  * been torn down.  However, it must be disconnected from the inpcb hashes as
751  * it must not prevent binding of future connections to the same port/ip
752  * combination by other inpcbs.
753  */
754 void
755 in_pcbdrop(struct inpcb *inp)
756 {
757 
758 	INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
759 	INP_LOCK_ASSERT(inp);
760 
761 	inp->inp_vflag |= INP_DROPPED;
762 	if (inp->inp_lport) {
763 		struct inpcbport *phd = inp->inp_phd;
764 
765 		LIST_REMOVE(inp, inp_hash);
766 		LIST_REMOVE(inp, inp_portlist);
767 		if (LIST_FIRST(&phd->phd_pcblist) == NULL) {
768 			LIST_REMOVE(phd, phd_hash);
769 			free(phd, M_PCB);
770 		}
771 		inp->inp_lport = 0;
772 	}
773 }
774 
775 /*
776  * Common routines to return the socket addresses associated with inpcbs.
777  */
778 struct sockaddr *
779 in_sockaddr(in_port_t port, struct in_addr *addr_p)
780 {
781 	struct sockaddr_in *sin;
782 
783 	MALLOC(sin, struct sockaddr_in *, sizeof *sin, M_SONAME,
784 		M_WAITOK | M_ZERO);
785 	sin->sin_family = AF_INET;
786 	sin->sin_len = sizeof(*sin);
787 	sin->sin_addr = *addr_p;
788 	sin->sin_port = port;
789 
790 	return (struct sockaddr *)sin;
791 }
792 
793 int
794 in_getsockaddr(struct socket *so, struct sockaddr **nam)
795 {
796 	struct inpcb *inp;
797 	struct in_addr addr;
798 	in_port_t port;
799 
800 	inp = sotoinpcb(so);
801 	KASSERT(inp != NULL, ("in_getsockaddr: inp == NULL"));
802 
803 	INP_LOCK(inp);
804 	port = inp->inp_lport;
805 	addr = inp->inp_laddr;
806 	INP_UNLOCK(inp);
807 
808 	*nam = in_sockaddr(port, &addr);
809 	return 0;
810 }
811 
812 int
813 in_getpeeraddr(struct socket *so, struct sockaddr **nam)
814 {
815 	struct inpcb *inp;
816 	struct in_addr addr;
817 	in_port_t port;
818 
819 	inp = sotoinpcb(so);
820 	KASSERT(inp != NULL, ("in_getpeeraddr: inp == NULL"));
821 
822 	INP_LOCK(inp);
823 	port = inp->inp_fport;
824 	addr = inp->inp_faddr;
825 	INP_UNLOCK(inp);
826 
827 	*nam = in_sockaddr(port, &addr);
828 	return 0;
829 }
830 
831 void
832 in_pcbnotifyall(struct inpcbinfo *pcbinfo, struct in_addr faddr, int errno,
833     struct inpcb *(*notify)(struct inpcb *, int))
834 {
835 	struct inpcb *inp, *ninp;
836 	struct inpcbhead *head;
837 
838 	INP_INFO_WLOCK(pcbinfo);
839 	head = pcbinfo->ipi_listhead;
840 	for (inp = LIST_FIRST(head); inp != NULL; inp = ninp) {
841 		INP_LOCK(inp);
842 		ninp = LIST_NEXT(inp, inp_list);
843 #ifdef INET6
844 		if ((inp->inp_vflag & INP_IPV4) == 0) {
845 			INP_UNLOCK(inp);
846 			continue;
847 		}
848 #endif
849 		if (inp->inp_faddr.s_addr != faddr.s_addr ||
850 		    inp->inp_socket == NULL) {
851 			INP_UNLOCK(inp);
852 			continue;
853 		}
854 		if ((*notify)(inp, errno))
855 			INP_UNLOCK(inp);
856 	}
857 	INP_INFO_WUNLOCK(pcbinfo);
858 }
859 
860 void
861 in_pcbpurgeif0(struct inpcbinfo *pcbinfo, struct ifnet *ifp)
862 {
863 	struct inpcb *inp;
864 	struct ip_moptions *imo;
865 	int i, gap;
866 
867 	INP_INFO_RLOCK(pcbinfo);
868 	LIST_FOREACH(inp, pcbinfo->ipi_listhead, inp_list) {
869 		INP_LOCK(inp);
870 		imo = inp->inp_moptions;
871 		if ((inp->inp_vflag & INP_IPV4) &&
872 		    imo != NULL) {
873 			/*
874 			 * Unselect the outgoing interface if it is being
875 			 * detached.
876 			 */
877 			if (imo->imo_multicast_ifp == ifp)
878 				imo->imo_multicast_ifp = NULL;
879 
880 			/*
881 			 * Drop multicast group membership if we joined
882 			 * through the interface being detached.
883 			 */
884 			for (i = 0, gap = 0; i < imo->imo_num_memberships;
885 			    i++) {
886 				if (imo->imo_membership[i]->inm_ifp == ifp) {
887 					in_delmulti(imo->imo_membership[i]);
888 					gap++;
889 				} else if (gap != 0)
890 					imo->imo_membership[i - gap] =
891 					    imo->imo_membership[i];
892 			}
893 			imo->imo_num_memberships -= gap;
894 		}
895 		INP_UNLOCK(inp);
896 	}
897 	INP_INFO_RUNLOCK(pcbinfo);
898 }
899 
900 /*
901  * Lookup a PCB based on the local address and port.
902  */
903 #define INP_LOOKUP_MAPPED_PCB_COST	3
904 struct inpcb *
905 in_pcblookup_local(struct inpcbinfo *pcbinfo, struct in_addr laddr,
906     u_int lport_arg, int wild_okay)
907 {
908 	struct inpcb *inp;
909 #ifdef INET6
910 	int matchwild = 3 + INP_LOOKUP_MAPPED_PCB_COST;
911 #else
912 	int matchwild = 3;
913 #endif
914 	int wildcard;
915 	u_short lport = lport_arg;
916 
917 	INP_INFO_WLOCK_ASSERT(pcbinfo);
918 
919 	if (!wild_okay) {
920 		struct inpcbhead *head;
921 		/*
922 		 * Look for an unconnected (wildcard foreign addr) PCB that
923 		 * matches the local address and port we're looking for.
924 		 */
925 		head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport,
926 		    0, pcbinfo->ipi_hashmask)];
927 		LIST_FOREACH(inp, head, inp_hash) {
928 #ifdef INET6
929 			if ((inp->inp_vflag & INP_IPV4) == 0)
930 				continue;
931 #endif
932 			if (inp->inp_faddr.s_addr == INADDR_ANY &&
933 			    inp->inp_laddr.s_addr == laddr.s_addr &&
934 			    inp->inp_lport == lport) {
935 				/*
936 				 * Found.
937 				 */
938 				return (inp);
939 			}
940 		}
941 		/*
942 		 * Not found.
943 		 */
944 		return (NULL);
945 	} else {
946 		struct inpcbporthead *porthash;
947 		struct inpcbport *phd;
948 		struct inpcb *match = NULL;
949 		/*
950 		 * Best fit PCB lookup.
951 		 *
952 		 * First see if this local port is in use by looking on the
953 		 * port hash list.
954 		 */
955 		porthash = &pcbinfo->ipi_porthashbase[INP_PCBPORTHASH(lport,
956 		    pcbinfo->ipi_porthashmask)];
957 		LIST_FOREACH(phd, porthash, phd_hash) {
958 			if (phd->phd_port == lport)
959 				break;
960 		}
961 		if (phd != NULL) {
962 			/*
963 			 * Port is in use by one or more PCBs. Look for best
964 			 * fit.
965 			 */
966 			LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) {
967 				wildcard = 0;
968 #ifdef INET6
969 				if ((inp->inp_vflag & INP_IPV4) == 0)
970 					continue;
971 				/*
972 				 * We never select the PCB that has
973 				 * INP_IPV6 flag and is bound to :: if
974 				 * we have another PCB which is bound
975 				 * to 0.0.0.0.  If a PCB has the
976 				 * INP_IPV6 flag, then we set its cost
977 				 * higher than IPv4 only PCBs.
978 				 *
979 				 * Note that the case only happens
980 				 * when a socket is bound to ::, under
981 				 * the condition that the use of the
982 				 * mapped address is allowed.
983 				 */
984 				if ((inp->inp_vflag & INP_IPV6) != 0)
985 					wildcard += INP_LOOKUP_MAPPED_PCB_COST;
986 #endif
987 				if (inp->inp_faddr.s_addr != INADDR_ANY)
988 					wildcard++;
989 				if (inp->inp_laddr.s_addr != INADDR_ANY) {
990 					if (laddr.s_addr == INADDR_ANY)
991 						wildcard++;
992 					else if (inp->inp_laddr.s_addr != laddr.s_addr)
993 						continue;
994 				} else {
995 					if (laddr.s_addr != INADDR_ANY)
996 						wildcard++;
997 				}
998 				if (wildcard < matchwild) {
999 					match = inp;
1000 					matchwild = wildcard;
1001 					if (matchwild == 0) {
1002 						break;
1003 					}
1004 				}
1005 			}
1006 		}
1007 		return (match);
1008 	}
1009 }
1010 #undef INP_LOOKUP_MAPPED_PCB_COST
1011 
1012 /*
1013  * Lookup PCB in hash list.
1014  */
1015 struct inpcb *
1016 in_pcblookup_hash(struct inpcbinfo *pcbinfo, struct in_addr faddr,
1017     u_int fport_arg, struct in_addr laddr, u_int lport_arg, int wildcard,
1018     struct ifnet *ifp)
1019 {
1020 	struct inpcbhead *head;
1021 	struct inpcb *inp;
1022 	u_short fport = fport_arg, lport = lport_arg;
1023 
1024 	INP_INFO_RLOCK_ASSERT(pcbinfo);
1025 
1026 	/*
1027 	 * First look for an exact match.
1028 	 */
1029 	head = &pcbinfo->ipi_hashbase[INP_PCBHASH(faddr.s_addr, lport, fport,
1030 	    pcbinfo->ipi_hashmask)];
1031 	LIST_FOREACH(inp, head, inp_hash) {
1032 #ifdef INET6
1033 		if ((inp->inp_vflag & INP_IPV4) == 0)
1034 			continue;
1035 #endif
1036 		if (inp->inp_faddr.s_addr == faddr.s_addr &&
1037 		    inp->inp_laddr.s_addr == laddr.s_addr &&
1038 		    inp->inp_fport == fport &&
1039 		    inp->inp_lport == lport)
1040 			return (inp);
1041 	}
1042 
1043 	/*
1044 	 * Then look for a wildcard match, if requested.
1045 	 */
1046 	if (wildcard) {
1047 		struct inpcb *local_wild = NULL;
1048 #ifdef INET6
1049 		struct inpcb *local_wild_mapped = NULL;
1050 #endif
1051 
1052 		head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport,
1053 		    0, pcbinfo->ipi_hashmask)];
1054 		LIST_FOREACH(inp, head, inp_hash) {
1055 #ifdef INET6
1056 			if ((inp->inp_vflag & INP_IPV4) == 0)
1057 				continue;
1058 #endif
1059 			if (inp->inp_faddr.s_addr == INADDR_ANY &&
1060 			    inp->inp_lport == lport) {
1061 				if (ifp && ifp->if_type == IFT_FAITH &&
1062 				    (inp->inp_flags & INP_FAITH) == 0)
1063 					continue;
1064 				if (inp->inp_laddr.s_addr == laddr.s_addr)
1065 					return (inp);
1066 				else if (inp->inp_laddr.s_addr == INADDR_ANY) {
1067 #ifdef INET6
1068 					if (INP_CHECK_SOCKAF(inp->inp_socket,
1069 							     AF_INET6))
1070 						local_wild_mapped = inp;
1071 					else
1072 #endif
1073 						local_wild = inp;
1074 				}
1075 			}
1076 		}
1077 #ifdef INET6
1078 		if (local_wild == NULL)
1079 			return (local_wild_mapped);
1080 #endif
1081 		return (local_wild);
1082 	}
1083 	return (NULL);
1084 }
1085 
1086 /*
1087  * Insert PCB onto various hash lists.
1088  */
1089 int
1090 in_pcbinshash(struct inpcb *inp)
1091 {
1092 	struct inpcbhead *pcbhash;
1093 	struct inpcbporthead *pcbporthash;
1094 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1095 	struct inpcbport *phd;
1096 	u_int32_t hashkey_faddr;
1097 
1098 	INP_INFO_WLOCK_ASSERT(pcbinfo);
1099 	INP_LOCK_ASSERT(inp);
1100 
1101 #ifdef INET6
1102 	if (inp->inp_vflag & INP_IPV6)
1103 		hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX */;
1104 	else
1105 #endif /* INET6 */
1106 	hashkey_faddr = inp->inp_faddr.s_addr;
1107 
1108 	pcbhash = &pcbinfo->ipi_hashbase[INP_PCBHASH(hashkey_faddr,
1109 		 inp->inp_lport, inp->inp_fport, pcbinfo->ipi_hashmask)];
1110 
1111 	pcbporthash = &pcbinfo->ipi_porthashbase[
1112 	    INP_PCBPORTHASH(inp->inp_lport, pcbinfo->ipi_porthashmask)];
1113 
1114 	/*
1115 	 * Go through port list and look for a head for this lport.
1116 	 */
1117 	LIST_FOREACH(phd, pcbporthash, phd_hash) {
1118 		if (phd->phd_port == inp->inp_lport)
1119 			break;
1120 	}
1121 	/*
1122 	 * If none exists, malloc one and tack it on.
1123 	 */
1124 	if (phd == NULL) {
1125 		MALLOC(phd, struct inpcbport *, sizeof(struct inpcbport), M_PCB, M_NOWAIT);
1126 		if (phd == NULL) {
1127 			return (ENOBUFS); /* XXX */
1128 		}
1129 		phd->phd_port = inp->inp_lport;
1130 		LIST_INIT(&phd->phd_pcblist);
1131 		LIST_INSERT_HEAD(pcbporthash, phd, phd_hash);
1132 	}
1133 	inp->inp_phd = phd;
1134 	LIST_INSERT_HEAD(&phd->phd_pcblist, inp, inp_portlist);
1135 	LIST_INSERT_HEAD(pcbhash, inp, inp_hash);
1136 	return (0);
1137 }
1138 
1139 /*
1140  * Move PCB to the proper hash bucket when { faddr, fport } have  been
1141  * changed. NOTE: This does not handle the case of the lport changing (the
1142  * hashed port list would have to be updated as well), so the lport must
1143  * not change after in_pcbinshash() has been called.
1144  */
1145 void
1146 in_pcbrehash(struct inpcb *inp)
1147 {
1148 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1149 	struct inpcbhead *head;
1150 	u_int32_t hashkey_faddr;
1151 
1152 	INP_INFO_WLOCK_ASSERT(pcbinfo);
1153 	INP_LOCK_ASSERT(inp);
1154 
1155 #ifdef INET6
1156 	if (inp->inp_vflag & INP_IPV6)
1157 		hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX */;
1158 	else
1159 #endif /* INET6 */
1160 	hashkey_faddr = inp->inp_faddr.s_addr;
1161 
1162 	head = &pcbinfo->ipi_hashbase[INP_PCBHASH(hashkey_faddr,
1163 		inp->inp_lport, inp->inp_fport, pcbinfo->ipi_hashmask)];
1164 
1165 	LIST_REMOVE(inp, inp_hash);
1166 	LIST_INSERT_HEAD(head, inp, inp_hash);
1167 }
1168 
1169 /*
1170  * Remove PCB from various lists.
1171  */
1172 void
1173 in_pcbremlists(struct inpcb *inp)
1174 {
1175 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1176 
1177 	INP_INFO_WLOCK_ASSERT(pcbinfo);
1178 	INP_LOCK_ASSERT(inp);
1179 
1180 	inp->inp_gencnt = ++pcbinfo->ipi_gencnt;
1181 	if (inp->inp_lport) {
1182 		struct inpcbport *phd = inp->inp_phd;
1183 
1184 		LIST_REMOVE(inp, inp_hash);
1185 		LIST_REMOVE(inp, inp_portlist);
1186 		if (LIST_FIRST(&phd->phd_pcblist) == NULL) {
1187 			LIST_REMOVE(phd, phd_hash);
1188 			free(phd, M_PCB);
1189 		}
1190 	}
1191 	LIST_REMOVE(inp, inp_list);
1192 	pcbinfo->ipi_count--;
1193 }
1194 
1195 /*
1196  * A set label operation has occurred at the socket layer, propagate the
1197  * label change into the in_pcb for the socket.
1198  */
1199 void
1200 in_pcbsosetlabel(struct socket *so)
1201 {
1202 #ifdef MAC
1203 	struct inpcb *inp;
1204 
1205 	inp = sotoinpcb(so);
1206 	KASSERT(inp != NULL, ("in_pcbsosetlabel: so->so_pcb == NULL"));
1207 
1208 	INP_LOCK(inp);
1209 	SOCK_LOCK(so);
1210 	mac_inpcb_sosetlabel(so, inp);
1211 	SOCK_UNLOCK(so);
1212 	INP_UNLOCK(inp);
1213 #endif
1214 }
1215 
1216 /*
1217  * ipport_tick runs once per second, determining if random port allocation
1218  * should be continued.  If more than ipport_randomcps ports have been
1219  * allocated in the last second, then we return to sequential port
1220  * allocation. We return to random allocation only once we drop below
1221  * ipport_randomcps for at least ipport_randomtime seconds.
1222  */
1223 void
1224 ipport_tick(void *xtp)
1225 {
1226 
1227 	if (ipport_tcpallocs <= ipport_tcplastcount + ipport_randomcps) {
1228 		if (ipport_stoprandom > 0)
1229 			ipport_stoprandom--;
1230 	} else
1231 		ipport_stoprandom = ipport_randomtime;
1232 	ipport_tcplastcount = ipport_tcpallocs;
1233 	callout_reset(&ipport_tick_callout, hz, ipport_tick, NULL);
1234 }
1235 
1236 #ifdef DDB
1237 static void
1238 db_print_indent(int indent)
1239 {
1240 	int i;
1241 
1242 	for (i = 0; i < indent; i++)
1243 		db_printf(" ");
1244 }
1245 
1246 static void
1247 db_print_inconninfo(struct in_conninfo *inc, const char *name, int indent)
1248 {
1249 	char faddr_str[48], laddr_str[48];
1250 
1251 	db_print_indent(indent);
1252 	db_printf("%s at %p\n", name, inc);
1253 
1254 	indent += 2;
1255 
1256 #ifdef INET6
1257 	if (inc->inc_flags == 1) {
1258 		/* IPv6. */
1259 		ip6_sprintf(laddr_str, &inc->inc6_laddr);
1260 		ip6_sprintf(faddr_str, &inc->inc6_faddr);
1261 	} else {
1262 #endif
1263 		/* IPv4. */
1264 		inet_ntoa_r(inc->inc_laddr, laddr_str);
1265 		inet_ntoa_r(inc->inc_faddr, faddr_str);
1266 #ifdef INET6
1267 	}
1268 #endif
1269 	db_print_indent(indent);
1270 	db_printf("inc_laddr %s   inc_lport %u\n", laddr_str,
1271 	    ntohs(inc->inc_lport));
1272 	db_print_indent(indent);
1273 	db_printf("inc_faddr %s   inc_fport %u\n", faddr_str,
1274 	    ntohs(inc->inc_fport));
1275 }
1276 
1277 static void
1278 db_print_inpflags(int inp_flags)
1279 {
1280 	int comma;
1281 
1282 	comma = 0;
1283 	if (inp_flags & INP_RECVOPTS) {
1284 		db_printf("%sINP_RECVOPTS", comma ? ", " : "");
1285 		comma = 1;
1286 	}
1287 	if (inp_flags & INP_RECVRETOPTS) {
1288 		db_printf("%sINP_RECVRETOPTS", comma ? ", " : "");
1289 		comma = 1;
1290 	}
1291 	if (inp_flags & INP_RECVDSTADDR) {
1292 		db_printf("%sINP_RECVDSTADDR", comma ? ", " : "");
1293 		comma = 1;
1294 	}
1295 	if (inp_flags & INP_HDRINCL) {
1296 		db_printf("%sINP_HDRINCL", comma ? ", " : "");
1297 		comma = 1;
1298 	}
1299 	if (inp_flags & INP_HIGHPORT) {
1300 		db_printf("%sINP_HIGHPORT", comma ? ", " : "");
1301 		comma = 1;
1302 	}
1303 	if (inp_flags & INP_LOWPORT) {
1304 		db_printf("%sINP_LOWPORT", comma ? ", " : "");
1305 		comma = 1;
1306 	}
1307 	if (inp_flags & INP_ANONPORT) {
1308 		db_printf("%sINP_ANONPORT", comma ? ", " : "");
1309 		comma = 1;
1310 	}
1311 	if (inp_flags & INP_RECVIF) {
1312 		db_printf("%sINP_RECVIF", comma ? ", " : "");
1313 		comma = 1;
1314 	}
1315 	if (inp_flags & INP_MTUDISC) {
1316 		db_printf("%sINP_MTUDISC", comma ? ", " : "");
1317 		comma = 1;
1318 	}
1319 	if (inp_flags & INP_FAITH) {
1320 		db_printf("%sINP_FAITH", comma ? ", " : "");
1321 		comma = 1;
1322 	}
1323 	if (inp_flags & INP_RECVTTL) {
1324 		db_printf("%sINP_RECVTTL", comma ? ", " : "");
1325 		comma = 1;
1326 	}
1327 	if (inp_flags & INP_DONTFRAG) {
1328 		db_printf("%sINP_DONTFRAG", comma ? ", " : "");
1329 		comma = 1;
1330 	}
1331 	if (inp_flags & IN6P_IPV6_V6ONLY) {
1332 		db_printf("%sIN6P_IPV6_V6ONLY", comma ? ", " : "");
1333 		comma = 1;
1334 	}
1335 	if (inp_flags & IN6P_PKTINFO) {
1336 		db_printf("%sIN6P_PKTINFO", comma ? ", " : "");
1337 		comma = 1;
1338 	}
1339 	if (inp_flags & IN6P_HOPLIMIT) {
1340 		db_printf("%sIN6P_HOPLIMIT", comma ? ", " : "");
1341 		comma = 1;
1342 	}
1343 	if (inp_flags & IN6P_HOPOPTS) {
1344 		db_printf("%sIN6P_HOPOPTS", comma ? ", " : "");
1345 		comma = 1;
1346 	}
1347 	if (inp_flags & IN6P_DSTOPTS) {
1348 		db_printf("%sIN6P_DSTOPTS", comma ? ", " : "");
1349 		comma = 1;
1350 	}
1351 	if (inp_flags & IN6P_RTHDR) {
1352 		db_printf("%sIN6P_RTHDR", comma ? ", " : "");
1353 		comma = 1;
1354 	}
1355 	if (inp_flags & IN6P_RTHDRDSTOPTS) {
1356 		db_printf("%sIN6P_RTHDRDSTOPTS", comma ? ", " : "");
1357 		comma = 1;
1358 	}
1359 	if (inp_flags & IN6P_TCLASS) {
1360 		db_printf("%sIN6P_TCLASS", comma ? ", " : "");
1361 		comma = 1;
1362 	}
1363 	if (inp_flags & IN6P_AUTOFLOWLABEL) {
1364 		db_printf("%sIN6P_AUTOFLOWLABEL", comma ? ", " : "");
1365 		comma = 1;
1366 	}
1367 	if (inp_flags & IN6P_RFC2292) {
1368 		db_printf("%sIN6P_RFC2292", comma ? ", " : "");
1369 		comma = 1;
1370 	}
1371 	if (inp_flags & IN6P_MTU) {
1372 		db_printf("IN6P_MTU%s", comma ? ", " : "");
1373 		comma = 1;
1374 	}
1375 }
1376 
1377 static void
1378 db_print_inpvflag(u_char inp_vflag)
1379 {
1380 	int comma;
1381 
1382 	comma = 0;
1383 	if (inp_vflag & INP_IPV4) {
1384 		db_printf("%sINP_IPV4", comma ? ", " : "");
1385 		comma  = 1;
1386 	}
1387 	if (inp_vflag & INP_IPV6) {
1388 		db_printf("%sINP_IPV6", comma ? ", " : "");
1389 		comma  = 1;
1390 	}
1391 	if (inp_vflag & INP_IPV6PROTO) {
1392 		db_printf("%sINP_IPV6PROTO", comma ? ", " : "");
1393 		comma  = 1;
1394 	}
1395 	if (inp_vflag & INP_TIMEWAIT) {
1396 		db_printf("%sINP_TIMEWAIT", comma ? ", " : "");
1397 		comma  = 1;
1398 	}
1399 	if (inp_vflag & INP_ONESBCAST) {
1400 		db_printf("%sINP_ONESBCAST", comma ? ", " : "");
1401 		comma  = 1;
1402 	}
1403 	if (inp_vflag & INP_DROPPED) {
1404 		db_printf("%sINP_DROPPED", comma ? ", " : "");
1405 		comma  = 1;
1406 	}
1407 	if (inp_vflag & INP_SOCKREF) {
1408 		db_printf("%sINP_SOCKREF", comma ? ", " : "");
1409 		comma  = 1;
1410 	}
1411 }
1412 
1413 void
1414 db_print_inpcb(struct inpcb *inp, const char *name, int indent)
1415 {
1416 
1417 	db_print_indent(indent);
1418 	db_printf("%s at %p\n", name, inp);
1419 
1420 	indent += 2;
1421 
1422 	db_print_indent(indent);
1423 	db_printf("inp_flow: 0x%x\n", inp->inp_flow);
1424 
1425 	db_print_inconninfo(&inp->inp_inc, "inp_conninfo", indent);
1426 
1427 	db_print_indent(indent);
1428 	db_printf("inp_ppcb: %p   inp_pcbinfo: %p   inp_socket: %p\n",
1429 	    inp->inp_ppcb, inp->inp_pcbinfo, inp->inp_socket);
1430 
1431 	db_print_indent(indent);
1432 	db_printf("inp_label: %p   inp_flags: 0x%x (",
1433 	   inp->inp_label, inp->inp_flags);
1434 	db_print_inpflags(inp->inp_flags);
1435 	db_printf(")\n");
1436 
1437 	db_print_indent(indent);
1438 	db_printf("inp_sp: %p   inp_vflag: 0x%x (", inp->inp_sp,
1439 	    inp->inp_vflag);
1440 	db_print_inpvflag(inp->inp_vflag);
1441 	db_printf(")\n");
1442 
1443 	db_print_indent(indent);
1444 	db_printf("inp_ip_ttl: %d   inp_ip_p: %d   inp_ip_minttl: %d\n",
1445 	    inp->inp_ip_ttl, inp->inp_ip_p, inp->inp_ip_minttl);
1446 
1447 	db_print_indent(indent);
1448 #ifdef INET6
1449 	if (inp->inp_vflag & INP_IPV6) {
1450 		db_printf("in6p_options: %p   in6p_outputopts: %p   "
1451 		    "in6p_moptions: %p\n", inp->in6p_options,
1452 		    inp->in6p_outputopts, inp->in6p_moptions);
1453 		db_printf("in6p_icmp6filt: %p   in6p_cksum %d   "
1454 		    "in6p_hops %u\n", inp->in6p_icmp6filt, inp->in6p_cksum,
1455 		    inp->in6p_hops);
1456 	} else
1457 #endif
1458 	{
1459 		db_printf("inp_ip_tos: %d   inp_ip_options: %p   "
1460 		    "inp_ip_moptions: %p\n", inp->inp_ip_tos,
1461 		    inp->inp_options, inp->inp_moptions);
1462 	}
1463 
1464 	db_print_indent(indent);
1465 	db_printf("inp_phd: %p   inp_gencnt: %ju\n", inp->inp_phd,
1466 	    (uintmax_t)inp->inp_gencnt);
1467 }
1468 
1469 DB_SHOW_COMMAND(inpcb, db_show_inpcb)
1470 {
1471 	struct inpcb *inp;
1472 
1473 	if (!have_addr) {
1474 		db_printf("usage: show inpcb <addr>\n");
1475 		return;
1476 	}
1477 	inp = (struct inpcb *)addr;
1478 
1479 	db_print_inpcb(inp, "inpcb", 0);
1480 }
1481 #endif
1482