xref: /freebsd/sys/netinet/in_pcb.c (revision c2bce4a2fcf3083607e00a1734b47c249751c8a8)
1 /*-
2  * Copyright (c) 1982, 1986, 1991, 1993, 1995
3  *	The Regents of the University of California.
4  * Copyright (c) 2007-2009 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_inet.h"
40 #include "opt_inet6.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/callout.h>
47 #include <sys/domain.h>
48 #include <sys/protosw.h>
49 #include <sys/socket.h>
50 #include <sys/socketvar.h>
51 #include <sys/priv.h>
52 #include <sys/proc.h>
53 #include <sys/jail.h>
54 #include <sys/kernel.h>
55 #include <sys/sysctl.h>
56 
57 #ifdef DDB
58 #include <ddb/ddb.h>
59 #endif
60 
61 #include <vm/uma.h>
62 
63 #include <net/if.h>
64 #include <net/if_types.h>
65 #include <net/route.h>
66 #include <net/vnet.h>
67 
68 #include <netinet/in.h>
69 #include <netinet/in_pcb.h>
70 #include <netinet/in_var.h>
71 #include <netinet/ip_var.h>
72 #include <netinet/tcp_var.h>
73 #include <netinet/udp.h>
74 #include <netinet/udp_var.h>
75 #ifdef INET6
76 #include <netinet/ip6.h>
77 #include <netinet6/ip6_var.h>
78 #include <netinet6/in6_pcb.h>
79 #endif /* INET6 */
80 
81 
82 #ifdef IPSEC
83 #include <netipsec/ipsec.h>
84 #include <netipsec/key.h>
85 #endif /* IPSEC */
86 
87 #include <security/mac/mac_framework.h>
88 
89 static struct callout	ipport_tick_callout;
90 
91 /*
92  * These configure the range of local port addresses assigned to
93  * "unspecified" outgoing connections/packets/whatever.
94  */
95 VNET_DEFINE(int, ipport_lowfirstauto) = IPPORT_RESERVED - 1;	/* 1023 */
96 VNET_DEFINE(int, ipport_lowlastauto) = IPPORT_RESERVEDSTART;	/* 600 */
97 VNET_DEFINE(int, ipport_firstauto) = IPPORT_EPHEMERALFIRST;	/* 10000 */
98 VNET_DEFINE(int, ipport_lastauto) = IPPORT_EPHEMERALLAST;	/* 65535 */
99 VNET_DEFINE(int, ipport_hifirstauto) = IPPORT_HIFIRSTAUTO;	/* 49152 */
100 VNET_DEFINE(int, ipport_hilastauto) = IPPORT_HILASTAUTO;	/* 65535 */
101 
102 /*
103  * Reserved ports accessible only to root. There are significant
104  * security considerations that must be accounted for when changing these,
105  * but the security benefits can be great. Please be careful.
106  */
107 VNET_DEFINE(int, ipport_reservedhigh) = IPPORT_RESERVED - 1;	/* 1023 */
108 VNET_DEFINE(int, ipport_reservedlow);
109 
110 /* Variables dealing with random ephemeral port allocation. */
111 VNET_DEFINE(int, ipport_randomized) = 1;	/* user controlled via sysctl */
112 VNET_DEFINE(int, ipport_randomcps) = 10;	/* user controlled via sysctl */
113 VNET_DEFINE(int, ipport_randomtime) = 45;	/* user controlled via sysctl */
114 VNET_DEFINE(int, ipport_stoprandom);		/* toggled by ipport_tick */
115 VNET_DEFINE(int, ipport_tcpallocs);
116 static VNET_DEFINE(int, ipport_tcplastcount);
117 
118 #define	V_ipport_tcplastcount		VNET(ipport_tcplastcount)
119 
120 #define RANGECHK(var, min, max) \
121 	if ((var) < (min)) { (var) = (min); } \
122 	else if ((var) > (max)) { (var) = (max); }
123 
124 static void	in_pcbremlists(struct inpcb *inp);
125 
126 static int
127 sysctl_net_ipport_check(SYSCTL_HANDLER_ARGS)
128 {
129 	int error;
130 
131 #ifdef VIMAGE
132 	error = vnet_sysctl_handle_int(oidp, arg1, arg2, req);
133 #else
134 	error = sysctl_handle_int(oidp, arg1, arg2, req);
135 #endif
136 	if (error == 0) {
137 		RANGECHK(V_ipport_lowfirstauto, 1, IPPORT_RESERVED - 1);
138 		RANGECHK(V_ipport_lowlastauto, 1, IPPORT_RESERVED - 1);
139 		RANGECHK(V_ipport_firstauto, IPPORT_RESERVED, IPPORT_MAX);
140 		RANGECHK(V_ipport_lastauto, IPPORT_RESERVED, IPPORT_MAX);
141 		RANGECHK(V_ipport_hifirstauto, IPPORT_RESERVED, IPPORT_MAX);
142 		RANGECHK(V_ipport_hilastauto, IPPORT_RESERVED, IPPORT_MAX);
143 	}
144 	return (error);
145 }
146 
147 #undef RANGECHK
148 
149 SYSCTL_NODE(_net_inet_ip, IPPROTO_IP, portrange, CTLFLAG_RW, 0, "IP Ports");
150 
151 SYSCTL_VNET_PROC(_net_inet_ip_portrange, OID_AUTO, lowfirst,
152 	CTLTYPE_INT|CTLFLAG_RW, &VNET_NAME(ipport_lowfirstauto), 0,
153 	&sysctl_net_ipport_check, "I", "");
154 SYSCTL_VNET_PROC(_net_inet_ip_portrange, OID_AUTO, lowlast,
155 	CTLTYPE_INT|CTLFLAG_RW, &VNET_NAME(ipport_lowlastauto), 0,
156 	&sysctl_net_ipport_check, "I", "");
157 SYSCTL_VNET_PROC(_net_inet_ip_portrange, OID_AUTO, first,
158 	CTLTYPE_INT|CTLFLAG_RW, &VNET_NAME(ipport_firstauto), 0,
159 	&sysctl_net_ipport_check, "I", "");
160 SYSCTL_VNET_PROC(_net_inet_ip_portrange, OID_AUTO, last,
161 	CTLTYPE_INT|CTLFLAG_RW, &VNET_NAME(ipport_lastauto), 0,
162 	&sysctl_net_ipport_check, "I", "");
163 SYSCTL_VNET_PROC(_net_inet_ip_portrange, OID_AUTO, hifirst,
164 	CTLTYPE_INT|CTLFLAG_RW, &VNET_NAME(ipport_hifirstauto), 0,
165 	&sysctl_net_ipport_check, "I", "");
166 SYSCTL_VNET_PROC(_net_inet_ip_portrange, OID_AUTO, hilast,
167 	CTLTYPE_INT|CTLFLAG_RW, &VNET_NAME(ipport_hilastauto), 0,
168 	&sysctl_net_ipport_check, "I", "");
169 SYSCTL_VNET_INT(_net_inet_ip_portrange, OID_AUTO, reservedhigh,
170 	CTLFLAG_RW|CTLFLAG_SECURE, &VNET_NAME(ipport_reservedhigh), 0, "");
171 SYSCTL_VNET_INT(_net_inet_ip_portrange, OID_AUTO, reservedlow,
172 	CTLFLAG_RW|CTLFLAG_SECURE, &VNET_NAME(ipport_reservedlow), 0, "");
173 SYSCTL_VNET_INT(_net_inet_ip_portrange, OID_AUTO, randomized, CTLFLAG_RW,
174 	&VNET_NAME(ipport_randomized), 0, "Enable random port allocation");
175 SYSCTL_VNET_INT(_net_inet_ip_portrange, OID_AUTO, randomcps, CTLFLAG_RW,
176 	&VNET_NAME(ipport_randomcps), 0, "Maximum number of random port "
177 	"allocations before switching to a sequental one");
178 SYSCTL_VNET_INT(_net_inet_ip_portrange, OID_AUTO, randomtime, CTLFLAG_RW,
179 	&VNET_NAME(ipport_randomtime), 0,
180 	"Minimum time to keep sequental port "
181 	"allocation before switching to a random one");
182 
183 /*
184  * in_pcb.c: manage the Protocol Control Blocks.
185  *
186  * NOTE: It is assumed that most of these functions will be called with
187  * the pcbinfo lock held, and often, the inpcb lock held, as these utility
188  * functions often modify hash chains or addresses in pcbs.
189  */
190 
191 /*
192  * Initialize an inpcbinfo -- we should be able to reduce the number of
193  * arguments in time.
194  */
195 void
196 in_pcbinfo_init(struct inpcbinfo *pcbinfo, const char *name,
197     struct inpcbhead *listhead, int hash_nelements, int porthash_nelements,
198     char *inpcbzone_name, uma_init inpcbzone_init, uma_fini inpcbzone_fini,
199     uint32_t inpcbzone_flags)
200 {
201 
202 	INP_INFO_LOCK_INIT(pcbinfo, name);
203 #ifdef VIMAGE
204 	pcbinfo->ipi_vnet = curvnet;
205 #endif
206 	pcbinfo->ipi_listhead = listhead;
207 	LIST_INIT(pcbinfo->ipi_listhead);
208 	pcbinfo->ipi_hashbase = hashinit(hash_nelements, M_PCB,
209 	    &pcbinfo->ipi_hashmask);
210 	pcbinfo->ipi_porthashbase = hashinit(porthash_nelements, M_PCB,
211 	    &pcbinfo->ipi_porthashmask);
212 	pcbinfo->ipi_zone = uma_zcreate(inpcbzone_name, sizeof(struct inpcb),
213 	    NULL, NULL, inpcbzone_init, inpcbzone_fini, UMA_ALIGN_PTR,
214 	    inpcbzone_flags);
215 	uma_zone_set_max(pcbinfo->ipi_zone, maxsockets);
216 }
217 
218 /*
219  * Destroy an inpcbinfo.
220  */
221 void
222 in_pcbinfo_destroy(struct inpcbinfo *pcbinfo)
223 {
224 
225 	hashdestroy(pcbinfo->ipi_hashbase, M_PCB, pcbinfo->ipi_hashmask);
226 	hashdestroy(pcbinfo->ipi_porthashbase, M_PCB,
227 	    pcbinfo->ipi_porthashmask);
228 	uma_zdestroy(pcbinfo->ipi_zone);
229 	INP_INFO_LOCK_DESTROY(pcbinfo);
230 }
231 
232 /*
233  * Allocate a PCB and associate it with the socket.
234  * On success return with the PCB locked.
235  */
236 int
237 in_pcballoc(struct socket *so, struct inpcbinfo *pcbinfo)
238 {
239 	struct inpcb *inp;
240 	int error;
241 
242 	INP_INFO_WLOCK_ASSERT(pcbinfo);
243 	error = 0;
244 	inp = uma_zalloc(pcbinfo->ipi_zone, M_NOWAIT);
245 	if (inp == NULL)
246 		return (ENOBUFS);
247 	bzero(inp, inp_zero_size);
248 	inp->inp_pcbinfo = pcbinfo;
249 	inp->inp_socket = so;
250 	inp->inp_cred = crhold(so->so_cred);
251 	inp->inp_inc.inc_fibnum = so->so_fibnum;
252 #ifdef MAC
253 	error = mac_inpcb_init(inp, M_NOWAIT);
254 	if (error != 0)
255 		goto out;
256 	mac_inpcb_create(so, inp);
257 #endif
258 #ifdef IPSEC
259 	error = ipsec_init_policy(so, &inp->inp_sp);
260 	if (error != 0) {
261 #ifdef MAC
262 		mac_inpcb_destroy(inp);
263 #endif
264 		goto out;
265 	}
266 #endif /*IPSEC*/
267 #ifdef INET6
268 	if (INP_SOCKAF(so) == AF_INET6) {
269 		inp->inp_vflag |= INP_IPV6PROTO;
270 		if (V_ip6_v6only)
271 			inp->inp_flags |= IN6P_IPV6_V6ONLY;
272 	}
273 #endif
274 	LIST_INSERT_HEAD(pcbinfo->ipi_listhead, inp, inp_list);
275 	pcbinfo->ipi_count++;
276 	so->so_pcb = (caddr_t)inp;
277 #ifdef INET6
278 	if (V_ip6_auto_flowlabel)
279 		inp->inp_flags |= IN6P_AUTOFLOWLABEL;
280 #endif
281 	INP_WLOCK(inp);
282 	inp->inp_gencnt = ++pcbinfo->ipi_gencnt;
283 	inp->inp_refcount = 1;	/* Reference from the inpcbinfo */
284 #if defined(IPSEC) || defined(MAC)
285 out:
286 	if (error != 0) {
287 		crfree(inp->inp_cred);
288 		uma_zfree(pcbinfo->ipi_zone, inp);
289 	}
290 #endif
291 	return (error);
292 }
293 
294 int
295 in_pcbbind(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred)
296 {
297 	int anonport, error;
298 
299 	INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
300 	INP_WLOCK_ASSERT(inp);
301 
302 	if (inp->inp_lport != 0 || inp->inp_laddr.s_addr != INADDR_ANY)
303 		return (EINVAL);
304 	anonport = inp->inp_lport == 0 && (nam == NULL ||
305 	    ((struct sockaddr_in *)nam)->sin_port == 0);
306 	error = in_pcbbind_setup(inp, nam, &inp->inp_laddr.s_addr,
307 	    &inp->inp_lport, cred);
308 	if (error)
309 		return (error);
310 	if (in_pcbinshash(inp) != 0) {
311 		inp->inp_laddr.s_addr = INADDR_ANY;
312 		inp->inp_lport = 0;
313 		return (EAGAIN);
314 	}
315 	if (anonport)
316 		inp->inp_flags |= INP_ANONPORT;
317 	return (0);
318 }
319 
320 #if defined(INET) || defined(INET6)
321 int
322 in_pcb_lport(struct inpcb *inp, struct in_addr *laddrp, u_short *lportp,
323     struct ucred *cred, int wild)
324 {
325 	struct inpcbinfo *pcbinfo;
326 	struct inpcb *tmpinp;
327 	unsigned short *lastport;
328 	int count, dorandom, error;
329 	u_short aux, first, last, lport;
330 #ifdef INET
331 	struct in_addr laddr;
332 #endif
333 
334 	pcbinfo = inp->inp_pcbinfo;
335 
336 	/*
337 	 * Because no actual state changes occur here, a global write lock on
338 	 * the pcbinfo isn't required.
339 	 */
340 	INP_INFO_LOCK_ASSERT(pcbinfo);
341 	INP_LOCK_ASSERT(inp);
342 
343 	if (inp->inp_flags & INP_HIGHPORT) {
344 		first = V_ipport_hifirstauto;	/* sysctl */
345 		last  = V_ipport_hilastauto;
346 		lastport = &pcbinfo->ipi_lasthi;
347 	} else if (inp->inp_flags & INP_LOWPORT) {
348 		error = priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT, 0);
349 		if (error)
350 			return (error);
351 		first = V_ipport_lowfirstauto;	/* 1023 */
352 		last  = V_ipport_lowlastauto;	/* 600 */
353 		lastport = &pcbinfo->ipi_lastlow;
354 	} else {
355 		first = V_ipport_firstauto;	/* sysctl */
356 		last  = V_ipport_lastauto;
357 		lastport = &pcbinfo->ipi_lastport;
358 	}
359 	/*
360 	 * For UDP, use random port allocation as long as the user
361 	 * allows it.  For TCP (and as of yet unknown) connections,
362 	 * use random port allocation only if the user allows it AND
363 	 * ipport_tick() allows it.
364 	 */
365 	if (V_ipport_randomized &&
366 		(!V_ipport_stoprandom || pcbinfo == &V_udbinfo))
367 		dorandom = 1;
368 	else
369 		dorandom = 0;
370 	/*
371 	 * It makes no sense to do random port allocation if
372 	 * we have the only port available.
373 	 */
374 	if (first == last)
375 		dorandom = 0;
376 	/* Make sure to not include UDP packets in the count. */
377 	if (pcbinfo != &V_udbinfo)
378 		V_ipport_tcpallocs++;
379 	/*
380 	 * Instead of having two loops further down counting up or down
381 	 * make sure that first is always <= last and go with only one
382 	 * code path implementing all logic.
383 	 */
384 	if (first > last) {
385 		aux = first;
386 		first = last;
387 		last = aux;
388 	}
389 
390 #ifdef INET
391 	/* Make the compiler happy. */
392 	laddr.s_addr = 0;
393 	if ((inp->inp_vflag & (INP_IPV4|INP_IPV6)) == INP_IPV4) {
394 		KASSERT(laddrp != NULL, ("%s: laddrp NULL for v4 inp %p",
395 		    __func__, inp));
396 		laddr = *laddrp;
397 	}
398 #endif
399 	lport = *lportp;
400 
401 	if (dorandom)
402 		*lastport = first + (arc4random() % (last - first));
403 
404 	count = last - first;
405 
406 	do {
407 		if (count-- < 0)	/* completely used? */
408 			return (EADDRNOTAVAIL);
409 		++*lastport;
410 		if (*lastport < first || *lastport > last)
411 			*lastport = first;
412 		lport = htons(*lastport);
413 
414 #ifdef INET6
415 		if ((inp->inp_vflag & INP_IPV6) != 0)
416 			tmpinp = in6_pcblookup_local(pcbinfo,
417 			    &inp->in6p_laddr, lport, wild, cred);
418 #endif
419 #if defined(INET) && defined(INET6)
420 		else
421 #endif
422 #ifdef INET
423 			tmpinp = in_pcblookup_local(pcbinfo, laddr,
424 			    lport, wild, cred);
425 #endif
426 	} while (tmpinp != NULL);
427 
428 #ifdef INET
429 	if ((inp->inp_vflag & (INP_IPV4|INP_IPV6)) == INP_IPV4)
430 		laddrp->s_addr = laddr.s_addr;
431 #endif
432 	*lportp = lport;
433 
434 	return (0);
435 }
436 #endif /* INET || INET6 */
437 
438 /*
439  * Set up a bind operation on a PCB, performing port allocation
440  * as required, but do not actually modify the PCB. Callers can
441  * either complete the bind by setting inp_laddr/inp_lport and
442  * calling in_pcbinshash(), or they can just use the resulting
443  * port and address to authorise the sending of a once-off packet.
444  *
445  * On error, the values of *laddrp and *lportp are not changed.
446  */
447 int
448 in_pcbbind_setup(struct inpcb *inp, struct sockaddr *nam, in_addr_t *laddrp,
449     u_short *lportp, struct ucred *cred)
450 {
451 	struct socket *so = inp->inp_socket;
452 	struct sockaddr_in *sin;
453 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
454 	struct in_addr laddr;
455 	u_short lport = 0;
456 	int wild = 0, reuseport = (so->so_options & SO_REUSEPORT);
457 	int error;
458 
459 	/*
460 	 * Because no actual state changes occur here, a global write lock on
461 	 * the pcbinfo isn't required.
462 	 */
463 	INP_INFO_LOCK_ASSERT(pcbinfo);
464 	INP_LOCK_ASSERT(inp);
465 
466 	if (TAILQ_EMPTY(&V_in_ifaddrhead)) /* XXX broken! */
467 		return (EADDRNOTAVAIL);
468 	laddr.s_addr = *laddrp;
469 	if (nam != NULL && laddr.s_addr != INADDR_ANY)
470 		return (EINVAL);
471 	if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0)
472 		wild = INPLOOKUP_WILDCARD;
473 	if (nam == NULL) {
474 		if ((error = prison_local_ip4(cred, &laddr)) != 0)
475 			return (error);
476 	} else {
477 		sin = (struct sockaddr_in *)nam;
478 		if (nam->sa_len != sizeof (*sin))
479 			return (EINVAL);
480 #ifdef notdef
481 		/*
482 		 * We should check the family, but old programs
483 		 * incorrectly fail to initialize it.
484 		 */
485 		if (sin->sin_family != AF_INET)
486 			return (EAFNOSUPPORT);
487 #endif
488 		error = prison_local_ip4(cred, &sin->sin_addr);
489 		if (error)
490 			return (error);
491 		if (sin->sin_port != *lportp) {
492 			/* Don't allow the port to change. */
493 			if (*lportp != 0)
494 				return (EINVAL);
495 			lport = sin->sin_port;
496 		}
497 		/* NB: lport is left as 0 if the port isn't being changed. */
498 		if (IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) {
499 			/*
500 			 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
501 			 * allow complete duplication of binding if
502 			 * SO_REUSEPORT is set, or if SO_REUSEADDR is set
503 			 * and a multicast address is bound on both
504 			 * new and duplicated sockets.
505 			 */
506 			if (so->so_options & SO_REUSEADDR)
507 				reuseport = SO_REUSEADDR|SO_REUSEPORT;
508 		} else if (sin->sin_addr.s_addr != INADDR_ANY) {
509 			sin->sin_port = 0;		/* yech... */
510 			bzero(&sin->sin_zero, sizeof(sin->sin_zero));
511 			/*
512 			 * Is the address a local IP address?
513 			 * If INP_BINDANY is set, then the socket may be bound
514 			 * to any endpoint address, local or not.
515 			 */
516 			if ((inp->inp_flags & INP_BINDANY) == 0 &&
517 			    ifa_ifwithaddr_check((struct sockaddr *)sin) == 0)
518 				return (EADDRNOTAVAIL);
519 		}
520 		laddr = sin->sin_addr;
521 		if (lport) {
522 			struct inpcb *t;
523 			struct tcptw *tw;
524 
525 			/* GROSS */
526 			if (ntohs(lport) <= V_ipport_reservedhigh &&
527 			    ntohs(lport) >= V_ipport_reservedlow &&
528 			    priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT,
529 			    0))
530 				return (EACCES);
531 			if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)) &&
532 			    priv_check_cred(inp->inp_cred,
533 			    PRIV_NETINET_REUSEPORT, 0) != 0) {
534 				t = in_pcblookup_local(pcbinfo, sin->sin_addr,
535 				    lport, INPLOOKUP_WILDCARD, cred);
536 	/*
537 	 * XXX
538 	 * This entire block sorely needs a rewrite.
539 	 */
540 				if (t &&
541 				    ((t->inp_flags & INP_TIMEWAIT) == 0) &&
542 				    (so->so_type != SOCK_STREAM ||
543 				     ntohl(t->inp_faddr.s_addr) == INADDR_ANY) &&
544 				    (ntohl(sin->sin_addr.s_addr) != INADDR_ANY ||
545 				     ntohl(t->inp_laddr.s_addr) != INADDR_ANY ||
546 				     (t->inp_socket->so_options &
547 					 SO_REUSEPORT) == 0) &&
548 				    (inp->inp_cred->cr_uid !=
549 				     t->inp_cred->cr_uid))
550 					return (EADDRINUSE);
551 			}
552 			t = in_pcblookup_local(pcbinfo, sin->sin_addr,
553 			    lport, wild, cred);
554 			if (t && (t->inp_flags & INP_TIMEWAIT)) {
555 				/*
556 				 * XXXRW: If an incpb has had its timewait
557 				 * state recycled, we treat the address as
558 				 * being in use (for now).  This is better
559 				 * than a panic, but not desirable.
560 				 */
561 				tw = intotw(inp);
562 				if (tw == NULL ||
563 				    (reuseport & tw->tw_so_options) == 0)
564 					return (EADDRINUSE);
565 			} else if (t &&
566 			    (reuseport & t->inp_socket->so_options) == 0) {
567 #ifdef INET6
568 				if (ntohl(sin->sin_addr.s_addr) !=
569 				    INADDR_ANY ||
570 				    ntohl(t->inp_laddr.s_addr) !=
571 				    INADDR_ANY ||
572 				    INP_SOCKAF(so) ==
573 				    INP_SOCKAF(t->inp_socket))
574 #endif
575 				return (EADDRINUSE);
576 			}
577 		}
578 	}
579 	if (*lportp != 0)
580 		lport = *lportp;
581 	if (lport == 0) {
582 		error = in_pcb_lport(inp, &laddr, &lport, cred, wild);
583 		if (error != 0)
584 			return (error);
585 
586 	}
587 	*laddrp = laddr.s_addr;
588 	*lportp = lport;
589 	return (0);
590 }
591 
592 /*
593  * Connect from a socket to a specified address.
594  * Both address and port must be specified in argument sin.
595  * If don't have a local address for this socket yet,
596  * then pick one.
597  */
598 int
599 in_pcbconnect(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred)
600 {
601 	u_short lport, fport;
602 	in_addr_t laddr, faddr;
603 	int anonport, error;
604 
605 	INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
606 	INP_WLOCK_ASSERT(inp);
607 
608 	lport = inp->inp_lport;
609 	laddr = inp->inp_laddr.s_addr;
610 	anonport = (lport == 0);
611 	error = in_pcbconnect_setup(inp, nam, &laddr, &lport, &faddr, &fport,
612 	    NULL, cred);
613 	if (error)
614 		return (error);
615 
616 	/* Do the initial binding of the local address if required. */
617 	if (inp->inp_laddr.s_addr == INADDR_ANY && inp->inp_lport == 0) {
618 		inp->inp_lport = lport;
619 		inp->inp_laddr.s_addr = laddr;
620 		if (in_pcbinshash(inp) != 0) {
621 			inp->inp_laddr.s_addr = INADDR_ANY;
622 			inp->inp_lport = 0;
623 			return (EAGAIN);
624 		}
625 	}
626 
627 	/* Commit the remaining changes. */
628 	inp->inp_lport = lport;
629 	inp->inp_laddr.s_addr = laddr;
630 	inp->inp_faddr.s_addr = faddr;
631 	inp->inp_fport = fport;
632 	in_pcbrehash(inp);
633 
634 	if (anonport)
635 		inp->inp_flags |= INP_ANONPORT;
636 	return (0);
637 }
638 
639 /*
640  * Do proper source address selection on an unbound socket in case
641  * of connect. Take jails into account as well.
642  */
643 static int
644 in_pcbladdr(struct inpcb *inp, struct in_addr *faddr, struct in_addr *laddr,
645     struct ucred *cred)
646 {
647 	struct ifaddr *ifa;
648 	struct sockaddr *sa;
649 	struct sockaddr_in *sin;
650 	struct route sro;
651 	int error;
652 
653 	KASSERT(laddr != NULL, ("%s: laddr NULL", __func__));
654 
655 	/*
656 	 * Bypass source address selection and use the primary jail IP
657 	 * if requested.
658 	 */
659 	if (cred != NULL && !prison_saddrsel_ip4(cred, laddr))
660 		return (0);
661 
662 	error = 0;
663 	bzero(&sro, sizeof(sro));
664 
665 	sin = (struct sockaddr_in *)&sro.ro_dst;
666 	sin->sin_family = AF_INET;
667 	sin->sin_len = sizeof(struct sockaddr_in);
668 	sin->sin_addr.s_addr = faddr->s_addr;
669 
670 	/*
671 	 * If route is known our src addr is taken from the i/f,
672 	 * else punt.
673 	 *
674 	 * Find out route to destination.
675 	 */
676 	if ((inp->inp_socket->so_options & SO_DONTROUTE) == 0)
677 		in_rtalloc_ign(&sro, 0, inp->inp_inc.inc_fibnum);
678 
679 	/*
680 	 * If we found a route, use the address corresponding to
681 	 * the outgoing interface.
682 	 *
683 	 * Otherwise assume faddr is reachable on a directly connected
684 	 * network and try to find a corresponding interface to take
685 	 * the source address from.
686 	 */
687 	if (sro.ro_rt == NULL || sro.ro_rt->rt_ifp == NULL) {
688 		struct in_ifaddr *ia;
689 		struct ifnet *ifp;
690 
691 		ia = ifatoia(ifa_ifwithdstaddr((struct sockaddr *)sin));
692 		if (ia == NULL)
693 			ia = ifatoia(ifa_ifwithnet((struct sockaddr *)sin, 0));
694 		if (ia == NULL) {
695 			error = ENETUNREACH;
696 			goto done;
697 		}
698 
699 		if (cred == NULL || !prison_flag(cred, PR_IP4)) {
700 			laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
701 			ifa_free(&ia->ia_ifa);
702 			goto done;
703 		}
704 
705 		ifp = ia->ia_ifp;
706 		ifa_free(&ia->ia_ifa);
707 		ia = NULL;
708 		IF_ADDR_LOCK(ifp);
709 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
710 
711 			sa = ifa->ifa_addr;
712 			if (sa->sa_family != AF_INET)
713 				continue;
714 			sin = (struct sockaddr_in *)sa;
715 			if (prison_check_ip4(cred, &sin->sin_addr) == 0) {
716 				ia = (struct in_ifaddr *)ifa;
717 				break;
718 			}
719 		}
720 		if (ia != NULL) {
721 			laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
722 			IF_ADDR_UNLOCK(ifp);
723 			goto done;
724 		}
725 		IF_ADDR_UNLOCK(ifp);
726 
727 		/* 3. As a last resort return the 'default' jail address. */
728 		error = prison_get_ip4(cred, laddr);
729 		goto done;
730 	}
731 
732 	/*
733 	 * If the outgoing interface on the route found is not
734 	 * a loopback interface, use the address from that interface.
735 	 * In case of jails do those three steps:
736 	 * 1. check if the interface address belongs to the jail. If so use it.
737 	 * 2. check if we have any address on the outgoing interface
738 	 *    belonging to this jail. If so use it.
739 	 * 3. as a last resort return the 'default' jail address.
740 	 */
741 	if ((sro.ro_rt->rt_ifp->if_flags & IFF_LOOPBACK) == 0) {
742 		struct in_ifaddr *ia;
743 		struct ifnet *ifp;
744 
745 		/* If not jailed, use the default returned. */
746 		if (cred == NULL || !prison_flag(cred, PR_IP4)) {
747 			ia = (struct in_ifaddr *)sro.ro_rt->rt_ifa;
748 			laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
749 			goto done;
750 		}
751 
752 		/* Jailed. */
753 		/* 1. Check if the iface address belongs to the jail. */
754 		sin = (struct sockaddr_in *)sro.ro_rt->rt_ifa->ifa_addr;
755 		if (prison_check_ip4(cred, &sin->sin_addr) == 0) {
756 			ia = (struct in_ifaddr *)sro.ro_rt->rt_ifa;
757 			laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
758 			goto done;
759 		}
760 
761 		/*
762 		 * 2. Check if we have any address on the outgoing interface
763 		 *    belonging to this jail.
764 		 */
765 		ia = NULL;
766 		ifp = sro.ro_rt->rt_ifp;
767 		IF_ADDR_LOCK(ifp);
768 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
769 			sa = ifa->ifa_addr;
770 			if (sa->sa_family != AF_INET)
771 				continue;
772 			sin = (struct sockaddr_in *)sa;
773 			if (prison_check_ip4(cred, &sin->sin_addr) == 0) {
774 				ia = (struct in_ifaddr *)ifa;
775 				break;
776 			}
777 		}
778 		if (ia != NULL) {
779 			laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
780 			IF_ADDR_UNLOCK(ifp);
781 			goto done;
782 		}
783 		IF_ADDR_UNLOCK(ifp);
784 
785 		/* 3. As a last resort return the 'default' jail address. */
786 		error = prison_get_ip4(cred, laddr);
787 		goto done;
788 	}
789 
790 	/*
791 	 * The outgoing interface is marked with 'loopback net', so a route
792 	 * to ourselves is here.
793 	 * Try to find the interface of the destination address and then
794 	 * take the address from there. That interface is not necessarily
795 	 * a loopback interface.
796 	 * In case of jails, check that it is an address of the jail
797 	 * and if we cannot find, fall back to the 'default' jail address.
798 	 */
799 	if ((sro.ro_rt->rt_ifp->if_flags & IFF_LOOPBACK) != 0) {
800 		struct sockaddr_in sain;
801 		struct in_ifaddr *ia;
802 
803 		bzero(&sain, sizeof(struct sockaddr_in));
804 		sain.sin_family = AF_INET;
805 		sain.sin_len = sizeof(struct sockaddr_in);
806 		sain.sin_addr.s_addr = faddr->s_addr;
807 
808 		ia = ifatoia(ifa_ifwithdstaddr(sintosa(&sain)));
809 		if (ia == NULL)
810 			ia = ifatoia(ifa_ifwithnet(sintosa(&sain), 0));
811 		if (ia == NULL)
812 			ia = ifatoia(ifa_ifwithaddr(sintosa(&sain)));
813 
814 		if (cred == NULL || !prison_flag(cred, PR_IP4)) {
815 			if (ia == NULL) {
816 				error = ENETUNREACH;
817 				goto done;
818 			}
819 			laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
820 			ifa_free(&ia->ia_ifa);
821 			goto done;
822 		}
823 
824 		/* Jailed. */
825 		if (ia != NULL) {
826 			struct ifnet *ifp;
827 
828 			ifp = ia->ia_ifp;
829 			ifa_free(&ia->ia_ifa);
830 			ia = NULL;
831 			IF_ADDR_LOCK(ifp);
832 			TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
833 
834 				sa = ifa->ifa_addr;
835 				if (sa->sa_family != AF_INET)
836 					continue;
837 				sin = (struct sockaddr_in *)sa;
838 				if (prison_check_ip4(cred,
839 				    &sin->sin_addr) == 0) {
840 					ia = (struct in_ifaddr *)ifa;
841 					break;
842 				}
843 			}
844 			if (ia != NULL) {
845 				laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
846 				IF_ADDR_UNLOCK(ifp);
847 				goto done;
848 			}
849 			IF_ADDR_UNLOCK(ifp);
850 		}
851 
852 		/* 3. As a last resort return the 'default' jail address. */
853 		error = prison_get_ip4(cred, laddr);
854 		goto done;
855 	}
856 
857 done:
858 	if (sro.ro_rt != NULL)
859 		RTFREE(sro.ro_rt);
860 	return (error);
861 }
862 
863 /*
864  * Set up for a connect from a socket to the specified address.
865  * On entry, *laddrp and *lportp should contain the current local
866  * address and port for the PCB; these are updated to the values
867  * that should be placed in inp_laddr and inp_lport to complete
868  * the connect.
869  *
870  * On success, *faddrp and *fportp will be set to the remote address
871  * and port. These are not updated in the error case.
872  *
873  * If the operation fails because the connection already exists,
874  * *oinpp will be set to the PCB of that connection so that the
875  * caller can decide to override it. In all other cases, *oinpp
876  * is set to NULL.
877  */
878 int
879 in_pcbconnect_setup(struct inpcb *inp, struct sockaddr *nam,
880     in_addr_t *laddrp, u_short *lportp, in_addr_t *faddrp, u_short *fportp,
881     struct inpcb **oinpp, struct ucred *cred)
882 {
883 	struct sockaddr_in *sin = (struct sockaddr_in *)nam;
884 	struct in_ifaddr *ia;
885 	struct inpcb *oinp;
886 	struct in_addr laddr, faddr;
887 	u_short lport, fport;
888 	int error;
889 
890 	/*
891 	 * Because a global state change doesn't actually occur here, a read
892 	 * lock is sufficient.
893 	 */
894 	INP_INFO_LOCK_ASSERT(inp->inp_pcbinfo);
895 	INP_LOCK_ASSERT(inp);
896 
897 	if (oinpp != NULL)
898 		*oinpp = NULL;
899 	if (nam->sa_len != sizeof (*sin))
900 		return (EINVAL);
901 	if (sin->sin_family != AF_INET)
902 		return (EAFNOSUPPORT);
903 	if (sin->sin_port == 0)
904 		return (EADDRNOTAVAIL);
905 	laddr.s_addr = *laddrp;
906 	lport = *lportp;
907 	faddr = sin->sin_addr;
908 	fport = sin->sin_port;
909 
910 	if (!TAILQ_EMPTY(&V_in_ifaddrhead)) {
911 		/*
912 		 * If the destination address is INADDR_ANY,
913 		 * use the primary local address.
914 		 * If the supplied address is INADDR_BROADCAST,
915 		 * and the primary interface supports broadcast,
916 		 * choose the broadcast address for that interface.
917 		 */
918 		if (faddr.s_addr == INADDR_ANY) {
919 			IN_IFADDR_RLOCK();
920 			faddr =
921 			    IA_SIN(TAILQ_FIRST(&V_in_ifaddrhead))->sin_addr;
922 			IN_IFADDR_RUNLOCK();
923 			if (cred != NULL &&
924 			    (error = prison_get_ip4(cred, &faddr)) != 0)
925 				return (error);
926 		} else if (faddr.s_addr == (u_long)INADDR_BROADCAST) {
927 			IN_IFADDR_RLOCK();
928 			if (TAILQ_FIRST(&V_in_ifaddrhead)->ia_ifp->if_flags &
929 			    IFF_BROADCAST)
930 				faddr = satosin(&TAILQ_FIRST(
931 				    &V_in_ifaddrhead)->ia_broadaddr)->sin_addr;
932 			IN_IFADDR_RUNLOCK();
933 		}
934 	}
935 	if (laddr.s_addr == INADDR_ANY) {
936 		error = in_pcbladdr(inp, &faddr, &laddr, cred);
937 		/*
938 		 * If the destination address is multicast and an outgoing
939 		 * interface has been set as a multicast option, prefer the
940 		 * address of that interface as our source address.
941 		 */
942 		if (IN_MULTICAST(ntohl(faddr.s_addr)) &&
943 		    inp->inp_moptions != NULL) {
944 			struct ip_moptions *imo;
945 			struct ifnet *ifp;
946 
947 			imo = inp->inp_moptions;
948 			if (imo->imo_multicast_ifp != NULL) {
949 				ifp = imo->imo_multicast_ifp;
950 				IN_IFADDR_RLOCK();
951 				TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
952 					if ((ia->ia_ifp == ifp) &&
953 					    (cred == NULL ||
954 					    prison_check_ip4(cred,
955 					    &ia->ia_addr.sin_addr) == 0))
956 						break;
957 				}
958 				if (ia == NULL)
959 					error = EADDRNOTAVAIL;
960 				else {
961 					laddr = ia->ia_addr.sin_addr;
962 					error = 0;
963 				}
964 				IN_IFADDR_RUNLOCK();
965 			}
966 		}
967 		if (error)
968 			return (error);
969 	}
970 	oinp = in_pcblookup_hash(inp->inp_pcbinfo, faddr, fport, laddr, lport,
971 	    0, NULL);
972 	if (oinp != NULL) {
973 		if (oinpp != NULL)
974 			*oinpp = oinp;
975 		return (EADDRINUSE);
976 	}
977 	if (lport == 0) {
978 		error = in_pcbbind_setup(inp, NULL, &laddr.s_addr, &lport,
979 		    cred);
980 		if (error)
981 			return (error);
982 	}
983 	*laddrp = laddr.s_addr;
984 	*lportp = lport;
985 	*faddrp = faddr.s_addr;
986 	*fportp = fport;
987 	return (0);
988 }
989 
990 void
991 in_pcbdisconnect(struct inpcb *inp)
992 {
993 
994 	INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
995 	INP_WLOCK_ASSERT(inp);
996 
997 	inp->inp_faddr.s_addr = INADDR_ANY;
998 	inp->inp_fport = 0;
999 	in_pcbrehash(inp);
1000 }
1001 
1002 /*
1003  * in_pcbdetach() is responsibe for disassociating a socket from an inpcb.
1004  * For most protocols, this will be invoked immediately prior to calling
1005  * in_pcbfree().  However, with TCP the inpcb may significantly outlive the
1006  * socket, in which case in_pcbfree() is deferred.
1007  */
1008 void
1009 in_pcbdetach(struct inpcb *inp)
1010 {
1011 
1012 	KASSERT(inp->inp_socket != NULL, ("%s: inp_socket == NULL", __func__));
1013 
1014 	inp->inp_socket->so_pcb = NULL;
1015 	inp->inp_socket = NULL;
1016 }
1017 
1018 /*
1019  * in_pcbfree_internal() frees an inpcb that has been detached from its
1020  * socket, and whose reference count has reached 0.  It will also remove the
1021  * inpcb from any global lists it might remain on.
1022  */
1023 static void
1024 in_pcbfree_internal(struct inpcb *inp)
1025 {
1026 	struct inpcbinfo *ipi = inp->inp_pcbinfo;
1027 
1028 	KASSERT(inp->inp_socket == NULL, ("%s: inp_socket != NULL", __func__));
1029 	KASSERT(inp->inp_refcount == 0, ("%s: refcount !0", __func__));
1030 
1031 	INP_INFO_WLOCK_ASSERT(ipi);
1032 	INP_WLOCK_ASSERT(inp);
1033 
1034 #ifdef IPSEC
1035 	if (inp->inp_sp != NULL)
1036 		ipsec_delete_pcbpolicy(inp);
1037 #endif /* IPSEC */
1038 	inp->inp_gencnt = ++ipi->ipi_gencnt;
1039 	in_pcbremlists(inp);
1040 #ifdef INET6
1041 	if (inp->inp_vflag & INP_IPV6PROTO) {
1042 		ip6_freepcbopts(inp->in6p_outputopts);
1043 		if (inp->in6p_moptions != NULL)
1044 			ip6_freemoptions(inp->in6p_moptions);
1045 	}
1046 #endif
1047 	if (inp->inp_options)
1048 		(void)m_free(inp->inp_options);
1049 	if (inp->inp_moptions != NULL)
1050 		inp_freemoptions(inp->inp_moptions);
1051 	inp->inp_vflag = 0;
1052 	crfree(inp->inp_cred);
1053 
1054 #ifdef MAC
1055 	mac_inpcb_destroy(inp);
1056 #endif
1057 	INP_WUNLOCK(inp);
1058 	uma_zfree(ipi->ipi_zone, inp);
1059 }
1060 
1061 /*
1062  * in_pcbref() bumps the reference count on an inpcb in order to maintain
1063  * stability of an inpcb pointer despite the inpcb lock being released.  This
1064  * is used in TCP when the inpcbinfo lock needs to be acquired or upgraded,
1065  * but where the inpcb lock is already held.
1066  *
1067  * While the inpcb will not be freed, releasing the inpcb lock means that the
1068  * connection's state may change, so the caller should be careful to
1069  * revalidate any cached state on reacquiring the lock.  Drop the reference
1070  * using in_pcbrele().
1071  */
1072 void
1073 in_pcbref(struct inpcb *inp)
1074 {
1075 
1076 	INP_WLOCK_ASSERT(inp);
1077 
1078 	KASSERT(inp->inp_refcount > 0, ("%s: refcount 0", __func__));
1079 
1080 	inp->inp_refcount++;
1081 }
1082 
1083 /*
1084  * Drop a refcount on an inpcb elevated using in_pcbref(); because a call to
1085  * in_pcbfree() may have been made between in_pcbref() and in_pcbrele(), we
1086  * return a flag indicating whether or not the inpcb remains valid.  If it is
1087  * valid, we return with the inpcb lock held.
1088  */
1089 int
1090 in_pcbrele(struct inpcb *inp)
1091 {
1092 #ifdef INVARIANTS
1093 	struct inpcbinfo *ipi = inp->inp_pcbinfo;
1094 #endif
1095 
1096 	KASSERT(inp->inp_refcount > 0, ("%s: refcount 0", __func__));
1097 
1098 	INP_INFO_WLOCK_ASSERT(ipi);
1099 	INP_WLOCK_ASSERT(inp);
1100 
1101 	inp->inp_refcount--;
1102 	if (inp->inp_refcount > 0)
1103 		return (0);
1104 	in_pcbfree_internal(inp);
1105 	return (1);
1106 }
1107 
1108 /*
1109  * Unconditionally schedule an inpcb to be freed by decrementing its
1110  * reference count, which should occur only after the inpcb has been detached
1111  * from its socket.  If another thread holds a temporary reference (acquired
1112  * using in_pcbref()) then the free is deferred until that reference is
1113  * released using in_pcbrele(), but the inpcb is still unlocked.
1114  */
1115 void
1116 in_pcbfree(struct inpcb *inp)
1117 {
1118 #ifdef INVARIANTS
1119 	struct inpcbinfo *ipi = inp->inp_pcbinfo;
1120 #endif
1121 
1122 	KASSERT(inp->inp_socket == NULL, ("%s: inp_socket != NULL",
1123 	    __func__));
1124 
1125 	INP_INFO_WLOCK_ASSERT(ipi);
1126 	INP_WLOCK_ASSERT(inp);
1127 
1128 	if (!in_pcbrele(inp))
1129 		INP_WUNLOCK(inp);
1130 }
1131 
1132 /*
1133  * in_pcbdrop() removes an inpcb from hashed lists, releasing its address and
1134  * port reservation, and preventing it from being returned by inpcb lookups.
1135  *
1136  * It is used by TCP to mark an inpcb as unused and avoid future packet
1137  * delivery or event notification when a socket remains open but TCP has
1138  * closed.  This might occur as a result of a shutdown()-initiated TCP close
1139  * or a RST on the wire, and allows the port binding to be reused while still
1140  * maintaining the invariant that so_pcb always points to a valid inpcb until
1141  * in_pcbdetach().
1142  *
1143  * XXXRW: Possibly in_pcbdrop() should also prevent future notifications by
1144  * in_pcbnotifyall() and in_pcbpurgeif0()?
1145  */
1146 void
1147 in_pcbdrop(struct inpcb *inp)
1148 {
1149 
1150 	INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
1151 	INP_WLOCK_ASSERT(inp);
1152 
1153 	inp->inp_flags |= INP_DROPPED;
1154 	if (inp->inp_flags & INP_INHASHLIST) {
1155 		struct inpcbport *phd = inp->inp_phd;
1156 
1157 		LIST_REMOVE(inp, inp_hash);
1158 		LIST_REMOVE(inp, inp_portlist);
1159 		if (LIST_FIRST(&phd->phd_pcblist) == NULL) {
1160 			LIST_REMOVE(phd, phd_hash);
1161 			free(phd, M_PCB);
1162 		}
1163 		inp->inp_flags &= ~INP_INHASHLIST;
1164 	}
1165 }
1166 
1167 /*
1168  * Common routines to return the socket addresses associated with inpcbs.
1169  */
1170 struct sockaddr *
1171 in_sockaddr(in_port_t port, struct in_addr *addr_p)
1172 {
1173 	struct sockaddr_in *sin;
1174 
1175 	sin = malloc(sizeof *sin, M_SONAME,
1176 		M_WAITOK | M_ZERO);
1177 	sin->sin_family = AF_INET;
1178 	sin->sin_len = sizeof(*sin);
1179 	sin->sin_addr = *addr_p;
1180 	sin->sin_port = port;
1181 
1182 	return (struct sockaddr *)sin;
1183 }
1184 
1185 int
1186 in_getsockaddr(struct socket *so, struct sockaddr **nam)
1187 {
1188 	struct inpcb *inp;
1189 	struct in_addr addr;
1190 	in_port_t port;
1191 
1192 	inp = sotoinpcb(so);
1193 	KASSERT(inp != NULL, ("in_getsockaddr: inp == NULL"));
1194 
1195 	INP_RLOCK(inp);
1196 	port = inp->inp_lport;
1197 	addr = inp->inp_laddr;
1198 	INP_RUNLOCK(inp);
1199 
1200 	*nam = in_sockaddr(port, &addr);
1201 	return 0;
1202 }
1203 
1204 int
1205 in_getpeeraddr(struct socket *so, struct sockaddr **nam)
1206 {
1207 	struct inpcb *inp;
1208 	struct in_addr addr;
1209 	in_port_t port;
1210 
1211 	inp = sotoinpcb(so);
1212 	KASSERT(inp != NULL, ("in_getpeeraddr: inp == NULL"));
1213 
1214 	INP_RLOCK(inp);
1215 	port = inp->inp_fport;
1216 	addr = inp->inp_faddr;
1217 	INP_RUNLOCK(inp);
1218 
1219 	*nam = in_sockaddr(port, &addr);
1220 	return 0;
1221 }
1222 
1223 void
1224 in_pcbnotifyall(struct inpcbinfo *pcbinfo, struct in_addr faddr, int errno,
1225     struct inpcb *(*notify)(struct inpcb *, int))
1226 {
1227 	struct inpcb *inp, *inp_temp;
1228 
1229 	INP_INFO_WLOCK(pcbinfo);
1230 	LIST_FOREACH_SAFE(inp, pcbinfo->ipi_listhead, inp_list, inp_temp) {
1231 		INP_WLOCK(inp);
1232 #ifdef INET6
1233 		if ((inp->inp_vflag & INP_IPV4) == 0) {
1234 			INP_WUNLOCK(inp);
1235 			continue;
1236 		}
1237 #endif
1238 		if (inp->inp_faddr.s_addr != faddr.s_addr ||
1239 		    inp->inp_socket == NULL) {
1240 			INP_WUNLOCK(inp);
1241 			continue;
1242 		}
1243 		if ((*notify)(inp, errno))
1244 			INP_WUNLOCK(inp);
1245 	}
1246 	INP_INFO_WUNLOCK(pcbinfo);
1247 }
1248 
1249 void
1250 in_pcbpurgeif0(struct inpcbinfo *pcbinfo, struct ifnet *ifp)
1251 {
1252 	struct inpcb *inp;
1253 	struct ip_moptions *imo;
1254 	int i, gap;
1255 
1256 	INP_INFO_RLOCK(pcbinfo);
1257 	LIST_FOREACH(inp, pcbinfo->ipi_listhead, inp_list) {
1258 		INP_WLOCK(inp);
1259 		imo = inp->inp_moptions;
1260 		if ((inp->inp_vflag & INP_IPV4) &&
1261 		    imo != NULL) {
1262 			/*
1263 			 * Unselect the outgoing interface if it is being
1264 			 * detached.
1265 			 */
1266 			if (imo->imo_multicast_ifp == ifp)
1267 				imo->imo_multicast_ifp = NULL;
1268 
1269 			/*
1270 			 * Drop multicast group membership if we joined
1271 			 * through the interface being detached.
1272 			 */
1273 			for (i = 0, gap = 0; i < imo->imo_num_memberships;
1274 			    i++) {
1275 				if (imo->imo_membership[i]->inm_ifp == ifp) {
1276 					in_delmulti(imo->imo_membership[i]);
1277 					gap++;
1278 				} else if (gap != 0)
1279 					imo->imo_membership[i - gap] =
1280 					    imo->imo_membership[i];
1281 			}
1282 			imo->imo_num_memberships -= gap;
1283 		}
1284 		INP_WUNLOCK(inp);
1285 	}
1286 	INP_INFO_RUNLOCK(pcbinfo);
1287 }
1288 
1289 /*
1290  * Lookup a PCB based on the local address and port.
1291  */
1292 #define INP_LOOKUP_MAPPED_PCB_COST	3
1293 struct inpcb *
1294 in_pcblookup_local(struct inpcbinfo *pcbinfo, struct in_addr laddr,
1295     u_short lport, int wild_okay, struct ucred *cred)
1296 {
1297 	struct inpcb *inp;
1298 #ifdef INET6
1299 	int matchwild = 3 + INP_LOOKUP_MAPPED_PCB_COST;
1300 #else
1301 	int matchwild = 3;
1302 #endif
1303 	int wildcard;
1304 
1305 	INP_INFO_LOCK_ASSERT(pcbinfo);
1306 
1307 	if (!wild_okay) {
1308 		struct inpcbhead *head;
1309 		/*
1310 		 * Look for an unconnected (wildcard foreign addr) PCB that
1311 		 * matches the local address and port we're looking for.
1312 		 */
1313 		head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport,
1314 		    0, pcbinfo->ipi_hashmask)];
1315 		LIST_FOREACH(inp, head, inp_hash) {
1316 #ifdef INET6
1317 			/* XXX inp locking */
1318 			if ((inp->inp_vflag & INP_IPV4) == 0)
1319 				continue;
1320 #endif
1321 			if (inp->inp_faddr.s_addr == INADDR_ANY &&
1322 			    inp->inp_laddr.s_addr == laddr.s_addr &&
1323 			    inp->inp_lport == lport) {
1324 				/*
1325 				 * Found?
1326 				 */
1327 				if (cred == NULL ||
1328 				    prison_equal_ip4(cred->cr_prison,
1329 					inp->inp_cred->cr_prison))
1330 					return (inp);
1331 			}
1332 		}
1333 		/*
1334 		 * Not found.
1335 		 */
1336 		return (NULL);
1337 	} else {
1338 		struct inpcbporthead *porthash;
1339 		struct inpcbport *phd;
1340 		struct inpcb *match = NULL;
1341 		/*
1342 		 * Best fit PCB lookup.
1343 		 *
1344 		 * First see if this local port is in use by looking on the
1345 		 * port hash list.
1346 		 */
1347 		porthash = &pcbinfo->ipi_porthashbase[INP_PCBPORTHASH(lport,
1348 		    pcbinfo->ipi_porthashmask)];
1349 		LIST_FOREACH(phd, porthash, phd_hash) {
1350 			if (phd->phd_port == lport)
1351 				break;
1352 		}
1353 		if (phd != NULL) {
1354 			/*
1355 			 * Port is in use by one or more PCBs. Look for best
1356 			 * fit.
1357 			 */
1358 			LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) {
1359 				wildcard = 0;
1360 				if (cred != NULL &&
1361 				    !prison_equal_ip4(inp->inp_cred->cr_prison,
1362 					cred->cr_prison))
1363 					continue;
1364 #ifdef INET6
1365 				/* XXX inp locking */
1366 				if ((inp->inp_vflag & INP_IPV4) == 0)
1367 					continue;
1368 				/*
1369 				 * We never select the PCB that has
1370 				 * INP_IPV6 flag and is bound to :: if
1371 				 * we have another PCB which is bound
1372 				 * to 0.0.0.0.  If a PCB has the
1373 				 * INP_IPV6 flag, then we set its cost
1374 				 * higher than IPv4 only PCBs.
1375 				 *
1376 				 * Note that the case only happens
1377 				 * when a socket is bound to ::, under
1378 				 * the condition that the use of the
1379 				 * mapped address is allowed.
1380 				 */
1381 				if ((inp->inp_vflag & INP_IPV6) != 0)
1382 					wildcard += INP_LOOKUP_MAPPED_PCB_COST;
1383 #endif
1384 				if (inp->inp_faddr.s_addr != INADDR_ANY)
1385 					wildcard++;
1386 				if (inp->inp_laddr.s_addr != INADDR_ANY) {
1387 					if (laddr.s_addr == INADDR_ANY)
1388 						wildcard++;
1389 					else if (inp->inp_laddr.s_addr != laddr.s_addr)
1390 						continue;
1391 				} else {
1392 					if (laddr.s_addr != INADDR_ANY)
1393 						wildcard++;
1394 				}
1395 				if (wildcard < matchwild) {
1396 					match = inp;
1397 					matchwild = wildcard;
1398 					if (matchwild == 0)
1399 						break;
1400 				}
1401 			}
1402 		}
1403 		return (match);
1404 	}
1405 }
1406 #undef INP_LOOKUP_MAPPED_PCB_COST
1407 
1408 /*
1409  * Lookup PCB in hash list.
1410  */
1411 struct inpcb *
1412 in_pcblookup_hash(struct inpcbinfo *pcbinfo, struct in_addr faddr,
1413     u_int fport_arg, struct in_addr laddr, u_int lport_arg, int wildcard,
1414     struct ifnet *ifp)
1415 {
1416 	struct inpcbhead *head;
1417 	struct inpcb *inp, *tmpinp;
1418 	u_short fport = fport_arg, lport = lport_arg;
1419 
1420 	INP_INFO_LOCK_ASSERT(pcbinfo);
1421 
1422 	/*
1423 	 * First look for an exact match.
1424 	 */
1425 	tmpinp = NULL;
1426 	head = &pcbinfo->ipi_hashbase[INP_PCBHASH(faddr.s_addr, lport, fport,
1427 	    pcbinfo->ipi_hashmask)];
1428 	LIST_FOREACH(inp, head, inp_hash) {
1429 #ifdef INET6
1430 		/* XXX inp locking */
1431 		if ((inp->inp_vflag & INP_IPV4) == 0)
1432 			continue;
1433 #endif
1434 		if (inp->inp_faddr.s_addr == faddr.s_addr &&
1435 		    inp->inp_laddr.s_addr == laddr.s_addr &&
1436 		    inp->inp_fport == fport &&
1437 		    inp->inp_lport == lport) {
1438 			/*
1439 			 * XXX We should be able to directly return
1440 			 * the inp here, without any checks.
1441 			 * Well unless both bound with SO_REUSEPORT?
1442 			 */
1443 			if (prison_flag(inp->inp_cred, PR_IP4))
1444 				return (inp);
1445 			if (tmpinp == NULL)
1446 				tmpinp = inp;
1447 		}
1448 	}
1449 	if (tmpinp != NULL)
1450 		return (tmpinp);
1451 
1452 	/*
1453 	 * Then look for a wildcard match, if requested.
1454 	 */
1455 	if (wildcard == INPLOOKUP_WILDCARD) {
1456 		struct inpcb *local_wild = NULL, *local_exact = NULL;
1457 #ifdef INET6
1458 		struct inpcb *local_wild_mapped = NULL;
1459 #endif
1460 		struct inpcb *jail_wild = NULL;
1461 		int injail;
1462 
1463 		/*
1464 		 * Order of socket selection - we always prefer jails.
1465 		 *      1. jailed, non-wild.
1466 		 *      2. jailed, wild.
1467 		 *      3. non-jailed, non-wild.
1468 		 *      4. non-jailed, wild.
1469 		 */
1470 
1471 		head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport,
1472 		    0, pcbinfo->ipi_hashmask)];
1473 		LIST_FOREACH(inp, head, inp_hash) {
1474 #ifdef INET6
1475 			/* XXX inp locking */
1476 			if ((inp->inp_vflag & INP_IPV4) == 0)
1477 				continue;
1478 #endif
1479 			if (inp->inp_faddr.s_addr != INADDR_ANY ||
1480 			    inp->inp_lport != lport)
1481 				continue;
1482 
1483 			/* XXX inp locking */
1484 			if (ifp && ifp->if_type == IFT_FAITH &&
1485 			    (inp->inp_flags & INP_FAITH) == 0)
1486 				continue;
1487 
1488 			injail = prison_flag(inp->inp_cred, PR_IP4);
1489 			if (injail) {
1490 				if (prison_check_ip4(inp->inp_cred,
1491 				    &laddr) != 0)
1492 					continue;
1493 			} else {
1494 				if (local_exact != NULL)
1495 					continue;
1496 			}
1497 
1498 			if (inp->inp_laddr.s_addr == laddr.s_addr) {
1499 				if (injail)
1500 					return (inp);
1501 				else
1502 					local_exact = inp;
1503 			} else if (inp->inp_laddr.s_addr == INADDR_ANY) {
1504 #ifdef INET6
1505 				/* XXX inp locking, NULL check */
1506 				if (inp->inp_vflag & INP_IPV6PROTO)
1507 					local_wild_mapped = inp;
1508 				else
1509 #endif /* INET6 */
1510 					if (injail)
1511 						jail_wild = inp;
1512 					else
1513 						local_wild = inp;
1514 			}
1515 		} /* LIST_FOREACH */
1516 		if (jail_wild != NULL)
1517 			return (jail_wild);
1518 		if (local_exact != NULL)
1519 			return (local_exact);
1520 		if (local_wild != NULL)
1521 			return (local_wild);
1522 #ifdef INET6
1523 		if (local_wild_mapped != NULL)
1524 			return (local_wild_mapped);
1525 #endif /* defined(INET6) */
1526 	} /* if (wildcard == INPLOOKUP_WILDCARD) */
1527 
1528 	return (NULL);
1529 }
1530 
1531 /*
1532  * Insert PCB onto various hash lists.
1533  */
1534 int
1535 in_pcbinshash(struct inpcb *inp)
1536 {
1537 	struct inpcbhead *pcbhash;
1538 	struct inpcbporthead *pcbporthash;
1539 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1540 	struct inpcbport *phd;
1541 	u_int32_t hashkey_faddr;
1542 
1543 	INP_INFO_WLOCK_ASSERT(pcbinfo);
1544 	INP_WLOCK_ASSERT(inp);
1545 	KASSERT((inp->inp_flags & INP_INHASHLIST) == 0,
1546 	    ("in_pcbinshash: INP_INHASHLIST"));
1547 
1548 #ifdef INET6
1549 	if (inp->inp_vflag & INP_IPV6)
1550 		hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX */;
1551 	else
1552 #endif /* INET6 */
1553 	hashkey_faddr = inp->inp_faddr.s_addr;
1554 
1555 	pcbhash = &pcbinfo->ipi_hashbase[INP_PCBHASH(hashkey_faddr,
1556 		 inp->inp_lport, inp->inp_fport, pcbinfo->ipi_hashmask)];
1557 
1558 	pcbporthash = &pcbinfo->ipi_porthashbase[
1559 	    INP_PCBPORTHASH(inp->inp_lport, pcbinfo->ipi_porthashmask)];
1560 
1561 	/*
1562 	 * Go through port list and look for a head for this lport.
1563 	 */
1564 	LIST_FOREACH(phd, pcbporthash, phd_hash) {
1565 		if (phd->phd_port == inp->inp_lport)
1566 			break;
1567 	}
1568 	/*
1569 	 * If none exists, malloc one and tack it on.
1570 	 */
1571 	if (phd == NULL) {
1572 		phd = malloc(sizeof(struct inpcbport), M_PCB, M_NOWAIT);
1573 		if (phd == NULL) {
1574 			return (ENOBUFS); /* XXX */
1575 		}
1576 		phd->phd_port = inp->inp_lport;
1577 		LIST_INIT(&phd->phd_pcblist);
1578 		LIST_INSERT_HEAD(pcbporthash, phd, phd_hash);
1579 	}
1580 	inp->inp_phd = phd;
1581 	LIST_INSERT_HEAD(&phd->phd_pcblist, inp, inp_portlist);
1582 	LIST_INSERT_HEAD(pcbhash, inp, inp_hash);
1583 	inp->inp_flags |= INP_INHASHLIST;
1584 	return (0);
1585 }
1586 
1587 /*
1588  * Move PCB to the proper hash bucket when { faddr, fport } have  been
1589  * changed. NOTE: This does not handle the case of the lport changing (the
1590  * hashed port list would have to be updated as well), so the lport must
1591  * not change after in_pcbinshash() has been called.
1592  */
1593 void
1594 in_pcbrehash(struct inpcb *inp)
1595 {
1596 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1597 	struct inpcbhead *head;
1598 	u_int32_t hashkey_faddr;
1599 
1600 	INP_INFO_WLOCK_ASSERT(pcbinfo);
1601 	INP_WLOCK_ASSERT(inp);
1602 	KASSERT(inp->inp_flags & INP_INHASHLIST,
1603 	    ("in_pcbrehash: !INP_INHASHLIST"));
1604 
1605 #ifdef INET6
1606 	if (inp->inp_vflag & INP_IPV6)
1607 		hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX */;
1608 	else
1609 #endif /* INET6 */
1610 	hashkey_faddr = inp->inp_faddr.s_addr;
1611 
1612 	head = &pcbinfo->ipi_hashbase[INP_PCBHASH(hashkey_faddr,
1613 		inp->inp_lport, inp->inp_fport, pcbinfo->ipi_hashmask)];
1614 
1615 	LIST_REMOVE(inp, inp_hash);
1616 	LIST_INSERT_HEAD(head, inp, inp_hash);
1617 }
1618 
1619 /*
1620  * Remove PCB from various lists.
1621  */
1622 static void
1623 in_pcbremlists(struct inpcb *inp)
1624 {
1625 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1626 
1627 	INP_INFO_WLOCK_ASSERT(pcbinfo);
1628 	INP_WLOCK_ASSERT(inp);
1629 
1630 	inp->inp_gencnt = ++pcbinfo->ipi_gencnt;
1631 	if (inp->inp_flags & INP_INHASHLIST) {
1632 		struct inpcbport *phd = inp->inp_phd;
1633 
1634 		LIST_REMOVE(inp, inp_hash);
1635 		LIST_REMOVE(inp, inp_portlist);
1636 		if (LIST_FIRST(&phd->phd_pcblist) == NULL) {
1637 			LIST_REMOVE(phd, phd_hash);
1638 			free(phd, M_PCB);
1639 		}
1640 		inp->inp_flags &= ~INP_INHASHLIST;
1641 	}
1642 	LIST_REMOVE(inp, inp_list);
1643 	pcbinfo->ipi_count--;
1644 }
1645 
1646 /*
1647  * A set label operation has occurred at the socket layer, propagate the
1648  * label change into the in_pcb for the socket.
1649  */
1650 void
1651 in_pcbsosetlabel(struct socket *so)
1652 {
1653 #ifdef MAC
1654 	struct inpcb *inp;
1655 
1656 	inp = sotoinpcb(so);
1657 	KASSERT(inp != NULL, ("in_pcbsosetlabel: so->so_pcb == NULL"));
1658 
1659 	INP_WLOCK(inp);
1660 	SOCK_LOCK(so);
1661 	mac_inpcb_sosetlabel(so, inp);
1662 	SOCK_UNLOCK(so);
1663 	INP_WUNLOCK(inp);
1664 #endif
1665 }
1666 
1667 /*
1668  * ipport_tick runs once per second, determining if random port allocation
1669  * should be continued.  If more than ipport_randomcps ports have been
1670  * allocated in the last second, then we return to sequential port
1671  * allocation. We return to random allocation only once we drop below
1672  * ipport_randomcps for at least ipport_randomtime seconds.
1673  */
1674 static void
1675 ipport_tick(void *xtp)
1676 {
1677 	VNET_ITERATOR_DECL(vnet_iter);
1678 
1679 	VNET_LIST_RLOCK_NOSLEEP();
1680 	VNET_FOREACH(vnet_iter) {
1681 		CURVNET_SET(vnet_iter);	/* XXX appease INVARIANTS here */
1682 		if (V_ipport_tcpallocs <=
1683 		    V_ipport_tcplastcount + V_ipport_randomcps) {
1684 			if (V_ipport_stoprandom > 0)
1685 				V_ipport_stoprandom--;
1686 		} else
1687 			V_ipport_stoprandom = V_ipport_randomtime;
1688 		V_ipport_tcplastcount = V_ipport_tcpallocs;
1689 		CURVNET_RESTORE();
1690 	}
1691 	VNET_LIST_RUNLOCK_NOSLEEP();
1692 	callout_reset(&ipport_tick_callout, hz, ipport_tick, NULL);
1693 }
1694 
1695 static void
1696 ip_fini(void *xtp)
1697 {
1698 
1699 	callout_stop(&ipport_tick_callout);
1700 }
1701 
1702 /*
1703  * The ipport_callout should start running at about the time we attach the
1704  * inet or inet6 domains.
1705  */
1706 static void
1707 ipport_tick_init(const void *unused __unused)
1708 {
1709 
1710 	/* Start ipport_tick. */
1711 	callout_init(&ipport_tick_callout, CALLOUT_MPSAFE);
1712 	callout_reset(&ipport_tick_callout, 1, ipport_tick, NULL);
1713 	EVENTHANDLER_REGISTER(shutdown_pre_sync, ip_fini, NULL,
1714 		SHUTDOWN_PRI_DEFAULT);
1715 }
1716 SYSINIT(ipport_tick_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE,
1717     ipport_tick_init, NULL);
1718 
1719 void
1720 inp_wlock(struct inpcb *inp)
1721 {
1722 
1723 	INP_WLOCK(inp);
1724 }
1725 
1726 void
1727 inp_wunlock(struct inpcb *inp)
1728 {
1729 
1730 	INP_WUNLOCK(inp);
1731 }
1732 
1733 void
1734 inp_rlock(struct inpcb *inp)
1735 {
1736 
1737 	INP_RLOCK(inp);
1738 }
1739 
1740 void
1741 inp_runlock(struct inpcb *inp)
1742 {
1743 
1744 	INP_RUNLOCK(inp);
1745 }
1746 
1747 #ifdef INVARIANTS
1748 void
1749 inp_lock_assert(struct inpcb *inp)
1750 {
1751 
1752 	INP_WLOCK_ASSERT(inp);
1753 }
1754 
1755 void
1756 inp_unlock_assert(struct inpcb *inp)
1757 {
1758 
1759 	INP_UNLOCK_ASSERT(inp);
1760 }
1761 #endif
1762 
1763 void
1764 inp_apply_all(void (*func)(struct inpcb *, void *), void *arg)
1765 {
1766 	struct inpcb *inp;
1767 
1768 	INP_INFO_RLOCK(&V_tcbinfo);
1769 	LIST_FOREACH(inp, V_tcbinfo.ipi_listhead, inp_list) {
1770 		INP_WLOCK(inp);
1771 		func(inp, arg);
1772 		INP_WUNLOCK(inp);
1773 	}
1774 	INP_INFO_RUNLOCK(&V_tcbinfo);
1775 }
1776 
1777 struct socket *
1778 inp_inpcbtosocket(struct inpcb *inp)
1779 {
1780 
1781 	INP_WLOCK_ASSERT(inp);
1782 	return (inp->inp_socket);
1783 }
1784 
1785 struct tcpcb *
1786 inp_inpcbtotcpcb(struct inpcb *inp)
1787 {
1788 
1789 	INP_WLOCK_ASSERT(inp);
1790 	return ((struct tcpcb *)inp->inp_ppcb);
1791 }
1792 
1793 int
1794 inp_ip_tos_get(const struct inpcb *inp)
1795 {
1796 
1797 	return (inp->inp_ip_tos);
1798 }
1799 
1800 void
1801 inp_ip_tos_set(struct inpcb *inp, int val)
1802 {
1803 
1804 	inp->inp_ip_tos = val;
1805 }
1806 
1807 void
1808 inp_4tuple_get(struct inpcb *inp, uint32_t *laddr, uint16_t *lp,
1809     uint32_t *faddr, uint16_t *fp)
1810 {
1811 
1812 	INP_LOCK_ASSERT(inp);
1813 	*laddr = inp->inp_laddr.s_addr;
1814 	*faddr = inp->inp_faddr.s_addr;
1815 	*lp = inp->inp_lport;
1816 	*fp = inp->inp_fport;
1817 }
1818 
1819 struct inpcb *
1820 so_sotoinpcb(struct socket *so)
1821 {
1822 
1823 	return (sotoinpcb(so));
1824 }
1825 
1826 struct tcpcb *
1827 so_sototcpcb(struct socket *so)
1828 {
1829 
1830 	return (sototcpcb(so));
1831 }
1832 
1833 #ifdef DDB
1834 static void
1835 db_print_indent(int indent)
1836 {
1837 	int i;
1838 
1839 	for (i = 0; i < indent; i++)
1840 		db_printf(" ");
1841 }
1842 
1843 static void
1844 db_print_inconninfo(struct in_conninfo *inc, const char *name, int indent)
1845 {
1846 	char faddr_str[48], laddr_str[48];
1847 
1848 	db_print_indent(indent);
1849 	db_printf("%s at %p\n", name, inc);
1850 
1851 	indent += 2;
1852 
1853 #ifdef INET6
1854 	if (inc->inc_flags & INC_ISIPV6) {
1855 		/* IPv6. */
1856 		ip6_sprintf(laddr_str, &inc->inc6_laddr);
1857 		ip6_sprintf(faddr_str, &inc->inc6_faddr);
1858 	} else {
1859 #endif
1860 		/* IPv4. */
1861 		inet_ntoa_r(inc->inc_laddr, laddr_str);
1862 		inet_ntoa_r(inc->inc_faddr, faddr_str);
1863 #ifdef INET6
1864 	}
1865 #endif
1866 	db_print_indent(indent);
1867 	db_printf("inc_laddr %s   inc_lport %u\n", laddr_str,
1868 	    ntohs(inc->inc_lport));
1869 	db_print_indent(indent);
1870 	db_printf("inc_faddr %s   inc_fport %u\n", faddr_str,
1871 	    ntohs(inc->inc_fport));
1872 }
1873 
1874 static void
1875 db_print_inpflags(int inp_flags)
1876 {
1877 	int comma;
1878 
1879 	comma = 0;
1880 	if (inp_flags & INP_RECVOPTS) {
1881 		db_printf("%sINP_RECVOPTS", comma ? ", " : "");
1882 		comma = 1;
1883 	}
1884 	if (inp_flags & INP_RECVRETOPTS) {
1885 		db_printf("%sINP_RECVRETOPTS", comma ? ", " : "");
1886 		comma = 1;
1887 	}
1888 	if (inp_flags & INP_RECVDSTADDR) {
1889 		db_printf("%sINP_RECVDSTADDR", comma ? ", " : "");
1890 		comma = 1;
1891 	}
1892 	if (inp_flags & INP_HDRINCL) {
1893 		db_printf("%sINP_HDRINCL", comma ? ", " : "");
1894 		comma = 1;
1895 	}
1896 	if (inp_flags & INP_HIGHPORT) {
1897 		db_printf("%sINP_HIGHPORT", comma ? ", " : "");
1898 		comma = 1;
1899 	}
1900 	if (inp_flags & INP_LOWPORT) {
1901 		db_printf("%sINP_LOWPORT", comma ? ", " : "");
1902 		comma = 1;
1903 	}
1904 	if (inp_flags & INP_ANONPORT) {
1905 		db_printf("%sINP_ANONPORT", comma ? ", " : "");
1906 		comma = 1;
1907 	}
1908 	if (inp_flags & INP_RECVIF) {
1909 		db_printf("%sINP_RECVIF", comma ? ", " : "");
1910 		comma = 1;
1911 	}
1912 	if (inp_flags & INP_MTUDISC) {
1913 		db_printf("%sINP_MTUDISC", comma ? ", " : "");
1914 		comma = 1;
1915 	}
1916 	if (inp_flags & INP_FAITH) {
1917 		db_printf("%sINP_FAITH", comma ? ", " : "");
1918 		comma = 1;
1919 	}
1920 	if (inp_flags & INP_RECVTTL) {
1921 		db_printf("%sINP_RECVTTL", comma ? ", " : "");
1922 		comma = 1;
1923 	}
1924 	if (inp_flags & INP_DONTFRAG) {
1925 		db_printf("%sINP_DONTFRAG", comma ? ", " : "");
1926 		comma = 1;
1927 	}
1928 	if (inp_flags & IN6P_IPV6_V6ONLY) {
1929 		db_printf("%sIN6P_IPV6_V6ONLY", comma ? ", " : "");
1930 		comma = 1;
1931 	}
1932 	if (inp_flags & IN6P_PKTINFO) {
1933 		db_printf("%sIN6P_PKTINFO", comma ? ", " : "");
1934 		comma = 1;
1935 	}
1936 	if (inp_flags & IN6P_HOPLIMIT) {
1937 		db_printf("%sIN6P_HOPLIMIT", comma ? ", " : "");
1938 		comma = 1;
1939 	}
1940 	if (inp_flags & IN6P_HOPOPTS) {
1941 		db_printf("%sIN6P_HOPOPTS", comma ? ", " : "");
1942 		comma = 1;
1943 	}
1944 	if (inp_flags & IN6P_DSTOPTS) {
1945 		db_printf("%sIN6P_DSTOPTS", comma ? ", " : "");
1946 		comma = 1;
1947 	}
1948 	if (inp_flags & IN6P_RTHDR) {
1949 		db_printf("%sIN6P_RTHDR", comma ? ", " : "");
1950 		comma = 1;
1951 	}
1952 	if (inp_flags & IN6P_RTHDRDSTOPTS) {
1953 		db_printf("%sIN6P_RTHDRDSTOPTS", comma ? ", " : "");
1954 		comma = 1;
1955 	}
1956 	if (inp_flags & IN6P_TCLASS) {
1957 		db_printf("%sIN6P_TCLASS", comma ? ", " : "");
1958 		comma = 1;
1959 	}
1960 	if (inp_flags & IN6P_AUTOFLOWLABEL) {
1961 		db_printf("%sIN6P_AUTOFLOWLABEL", comma ? ", " : "");
1962 		comma = 1;
1963 	}
1964 	if (inp_flags & INP_TIMEWAIT) {
1965 		db_printf("%sINP_TIMEWAIT", comma ? ", " : "");
1966 		comma  = 1;
1967 	}
1968 	if (inp_flags & INP_ONESBCAST) {
1969 		db_printf("%sINP_ONESBCAST", comma ? ", " : "");
1970 		comma  = 1;
1971 	}
1972 	if (inp_flags & INP_DROPPED) {
1973 		db_printf("%sINP_DROPPED", comma ? ", " : "");
1974 		comma  = 1;
1975 	}
1976 	if (inp_flags & INP_SOCKREF) {
1977 		db_printf("%sINP_SOCKREF", comma ? ", " : "");
1978 		comma  = 1;
1979 	}
1980 	if (inp_flags & IN6P_RFC2292) {
1981 		db_printf("%sIN6P_RFC2292", comma ? ", " : "");
1982 		comma = 1;
1983 	}
1984 	if (inp_flags & IN6P_MTU) {
1985 		db_printf("IN6P_MTU%s", comma ? ", " : "");
1986 		comma = 1;
1987 	}
1988 }
1989 
1990 static void
1991 db_print_inpvflag(u_char inp_vflag)
1992 {
1993 	int comma;
1994 
1995 	comma = 0;
1996 	if (inp_vflag & INP_IPV4) {
1997 		db_printf("%sINP_IPV4", comma ? ", " : "");
1998 		comma  = 1;
1999 	}
2000 	if (inp_vflag & INP_IPV6) {
2001 		db_printf("%sINP_IPV6", comma ? ", " : "");
2002 		comma  = 1;
2003 	}
2004 	if (inp_vflag & INP_IPV6PROTO) {
2005 		db_printf("%sINP_IPV6PROTO", comma ? ", " : "");
2006 		comma  = 1;
2007 	}
2008 }
2009 
2010 static void
2011 db_print_inpcb(struct inpcb *inp, const char *name, int indent)
2012 {
2013 
2014 	db_print_indent(indent);
2015 	db_printf("%s at %p\n", name, inp);
2016 
2017 	indent += 2;
2018 
2019 	db_print_indent(indent);
2020 	db_printf("inp_flow: 0x%x\n", inp->inp_flow);
2021 
2022 	db_print_inconninfo(&inp->inp_inc, "inp_conninfo", indent);
2023 
2024 	db_print_indent(indent);
2025 	db_printf("inp_ppcb: %p   inp_pcbinfo: %p   inp_socket: %p\n",
2026 	    inp->inp_ppcb, inp->inp_pcbinfo, inp->inp_socket);
2027 
2028 	db_print_indent(indent);
2029 	db_printf("inp_label: %p   inp_flags: 0x%x (",
2030 	   inp->inp_label, inp->inp_flags);
2031 	db_print_inpflags(inp->inp_flags);
2032 	db_printf(")\n");
2033 
2034 	db_print_indent(indent);
2035 	db_printf("inp_sp: %p   inp_vflag: 0x%x (", inp->inp_sp,
2036 	    inp->inp_vflag);
2037 	db_print_inpvflag(inp->inp_vflag);
2038 	db_printf(")\n");
2039 
2040 	db_print_indent(indent);
2041 	db_printf("inp_ip_ttl: %d   inp_ip_p: %d   inp_ip_minttl: %d\n",
2042 	    inp->inp_ip_ttl, inp->inp_ip_p, inp->inp_ip_minttl);
2043 
2044 	db_print_indent(indent);
2045 #ifdef INET6
2046 	if (inp->inp_vflag & INP_IPV6) {
2047 		db_printf("in6p_options: %p   in6p_outputopts: %p   "
2048 		    "in6p_moptions: %p\n", inp->in6p_options,
2049 		    inp->in6p_outputopts, inp->in6p_moptions);
2050 		db_printf("in6p_icmp6filt: %p   in6p_cksum %d   "
2051 		    "in6p_hops %u\n", inp->in6p_icmp6filt, inp->in6p_cksum,
2052 		    inp->in6p_hops);
2053 	} else
2054 #endif
2055 	{
2056 		db_printf("inp_ip_tos: %d   inp_ip_options: %p   "
2057 		    "inp_ip_moptions: %p\n", inp->inp_ip_tos,
2058 		    inp->inp_options, inp->inp_moptions);
2059 	}
2060 
2061 	db_print_indent(indent);
2062 	db_printf("inp_phd: %p   inp_gencnt: %ju\n", inp->inp_phd,
2063 	    (uintmax_t)inp->inp_gencnt);
2064 }
2065 
2066 DB_SHOW_COMMAND(inpcb, db_show_inpcb)
2067 {
2068 	struct inpcb *inp;
2069 
2070 	if (!have_addr) {
2071 		db_printf("usage: show inpcb <addr>\n");
2072 		return;
2073 	}
2074 	inp = (struct inpcb *)addr;
2075 
2076 	db_print_inpcb(inp, "inpcb", 0);
2077 }
2078 #endif
2079