xref: /freebsd/sys/netinet/in_pcb.c (revision 00a5db46de56179184c0f000eaacad695e2b0859)
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_inet.h"
39 #include "opt_ipsec.h"
40 #include "opt_inet6.h"
41 #include "opt_mac.h"
42 
43 #include <sys/param.h>
44 #include <sys/systm.h>
45 #include <sys/malloc.h>
46 #include <sys/mbuf.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 #include <sys/vimage.h>
57 
58 #ifdef DDB
59 #include <ddb/ddb.h>
60 #endif
61 
62 #include <vm/uma.h>
63 
64 #include <net/if.h>
65 #include <net/if_types.h>
66 #include <net/route.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 #include <netinet/vinet.h>
76 #ifdef INET6
77 #include <netinet/ip6.h>
78 #include <netinet6/ip6_var.h>
79 #include <netinet6/vinet6.h>
80 #endif /* INET6 */
81 
82 
83 #ifdef IPSEC
84 #include <netipsec/ipsec.h>
85 #include <netipsec/key.h>
86 #endif /* IPSEC */
87 
88 #include <security/mac/mac_framework.h>
89 
90 #ifdef VIMAGE_GLOBALS
91 /*
92  * These configure the range of local port addresses assigned to
93  * "unspecified" outgoing connections/packets/whatever.
94  */
95 int	ipport_lowfirstauto;
96 int	ipport_lowlastauto;
97 int	ipport_firstauto;
98 int	ipport_lastauto;
99 int	ipport_hifirstauto;
100 int	ipport_hilastauto;
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 int	ipport_reservedhigh;
108 int	ipport_reservedlow;
109 
110 /* Variables dealing with random ephemeral port allocation. */
111 int	ipport_randomized;
112 int	ipport_randomcps;
113 int	ipport_randomtime;
114 int	ipport_stoprandom;
115 int	ipport_tcpallocs;
116 int	ipport_tcplastcount;
117 #endif
118 
119 #define RANGECHK(var, min, max) \
120 	if ((var) < (min)) { (var) = (min); } \
121 	else if ((var) > (max)) { (var) = (max); }
122 
123 static void	in_pcbremlists(struct inpcb *inp);
124 
125 static int
126 sysctl_net_ipport_check(SYSCTL_HANDLER_ARGS)
127 {
128 	INIT_VNET_INET(curvnet);
129 	int error;
130 
131 	SYSCTL_RESOLVE_V_ARG1();
132 
133 	error = sysctl_handle_int(oidp, arg1, arg2, req);
134 	if (error == 0) {
135 		RANGECHK(V_ipport_lowfirstauto, 1, IPPORT_RESERVED - 1);
136 		RANGECHK(V_ipport_lowlastauto, 1, IPPORT_RESERVED - 1);
137 		RANGECHK(V_ipport_firstauto, IPPORT_RESERVED, IPPORT_MAX);
138 		RANGECHK(V_ipport_lastauto, IPPORT_RESERVED, IPPORT_MAX);
139 		RANGECHK(V_ipport_hifirstauto, IPPORT_RESERVED, IPPORT_MAX);
140 		RANGECHK(V_ipport_hilastauto, IPPORT_RESERVED, IPPORT_MAX);
141 	}
142 	return (error);
143 }
144 
145 #undef RANGECHK
146 
147 SYSCTL_NODE(_net_inet_ip, IPPROTO_IP, portrange, CTLFLAG_RW, 0, "IP Ports");
148 
149 SYSCTL_V_PROC(V_NET, vnet_inet, _net_inet_ip_portrange, OID_AUTO,
150 	lowfirst, CTLTYPE_INT|CTLFLAG_RW, ipport_lowfirstauto, 0,
151 	&sysctl_net_ipport_check, "I", "");
152 SYSCTL_V_PROC(V_NET, vnet_inet, _net_inet_ip_portrange, OID_AUTO,
153 	lowlast, CTLTYPE_INT|CTLFLAG_RW, ipport_lowlastauto, 0,
154 	&sysctl_net_ipport_check, "I", "");
155 SYSCTL_V_PROC(V_NET, vnet_inet, _net_inet_ip_portrange, OID_AUTO,
156 	first, CTLTYPE_INT|CTLFLAG_RW, ipport_firstauto, 0,
157 	&sysctl_net_ipport_check, "I", "");
158 SYSCTL_V_PROC(V_NET, vnet_inet, _net_inet_ip_portrange, OID_AUTO,
159 	last, CTLTYPE_INT|CTLFLAG_RW, ipport_lastauto, 0,
160 	&sysctl_net_ipport_check, "I", "");
161 SYSCTL_V_PROC(V_NET, vnet_inet, _net_inet_ip_portrange, OID_AUTO,
162 	hifirst, CTLTYPE_INT|CTLFLAG_RW, ipport_hifirstauto, 0,
163 	&sysctl_net_ipport_check, "I", "");
164 SYSCTL_V_PROC(V_NET, vnet_inet, _net_inet_ip_portrange, OID_AUTO,
165 	hilast, CTLTYPE_INT|CTLFLAG_RW, ipport_hilastauto, 0,
166 	&sysctl_net_ipport_check, "I", "");
167 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip_portrange, OID_AUTO,
168 	reservedhigh, CTLFLAG_RW|CTLFLAG_SECURE, ipport_reservedhigh, 0, "");
169 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip_portrange, OID_AUTO, reservedlow,
170 	CTLFLAG_RW|CTLFLAG_SECURE, ipport_reservedlow, 0, "");
171 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip_portrange, OID_AUTO, randomized,
172 	CTLFLAG_RW, ipport_randomized, 0, "Enable random port allocation");
173 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip_portrange, OID_AUTO, randomcps,
174 	CTLFLAG_RW, ipport_randomcps, 0, "Maximum number of random port "
175 	"allocations before switching to a sequental one");
176 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip_portrange, OID_AUTO, randomtime,
177 	CTLFLAG_RW, ipport_randomtime, 0,
178 	"Minimum time to keep sequental port "
179 	"allocation before switching to a random one");
180 
181 /*
182  * in_pcb.c: manage the Protocol Control Blocks.
183  *
184  * NOTE: It is assumed that most of these functions will be called with
185  * the pcbinfo lock held, and often, the inpcb lock held, as these utility
186  * functions often modify hash chains or addresses in pcbs.
187  */
188 
189 /*
190  * Allocate a PCB and associate it with the socket.
191  * On success return with the PCB locked.
192  */
193 int
194 in_pcballoc(struct socket *so, struct inpcbinfo *pcbinfo)
195 {
196 #ifdef INET6
197 	INIT_VNET_INET6(curvnet);
198 #endif
199 	struct inpcb *inp;
200 	int error;
201 
202 	INP_INFO_WLOCK_ASSERT(pcbinfo);
203 	error = 0;
204 	inp = uma_zalloc(pcbinfo->ipi_zone, M_NOWAIT);
205 	if (inp == NULL)
206 		return (ENOBUFS);
207 	bzero(inp, inp_zero_size);
208 	inp->inp_pcbinfo = pcbinfo;
209 	inp->inp_socket = so;
210 	inp->inp_cred = crhold(so->so_cred);
211 	inp->inp_inc.inc_fibnum = so->so_fibnum;
212 #ifdef MAC
213 	error = mac_inpcb_init(inp, M_NOWAIT);
214 	if (error != 0)
215 		goto out;
216 	SOCK_LOCK(so);
217 	mac_inpcb_create(so, inp);
218 	SOCK_UNLOCK(so);
219 #endif
220 #ifdef IPSEC
221 	error = ipsec_init_policy(so, &inp->inp_sp);
222 	if (error != 0) {
223 #ifdef MAC
224 		mac_inpcb_destroy(inp);
225 #endif
226 		goto out;
227 	}
228 #endif /*IPSEC*/
229 #ifdef INET6
230 	if (INP_SOCKAF(so) == AF_INET6) {
231 		inp->inp_vflag |= INP_IPV6PROTO;
232 		if (V_ip6_v6only)
233 			inp->inp_flags |= IN6P_IPV6_V6ONLY;
234 	}
235 #endif
236 	LIST_INSERT_HEAD(pcbinfo->ipi_listhead, inp, inp_list);
237 	pcbinfo->ipi_count++;
238 	so->so_pcb = (caddr_t)inp;
239 #ifdef INET6
240 	if (V_ip6_auto_flowlabel)
241 		inp->inp_flags |= IN6P_AUTOFLOWLABEL;
242 #endif
243 	INP_WLOCK(inp);
244 	inp->inp_gencnt = ++pcbinfo->ipi_gencnt;
245 	inp->inp_refcount = 1;	/* Reference from the inpcbinfo */
246 #if defined(IPSEC) || defined(MAC)
247 out:
248 	if (error != 0) {
249 		crfree(inp->inp_cred);
250 		uma_zfree(pcbinfo->ipi_zone, inp);
251 	}
252 #endif
253 	return (error);
254 }
255 
256 int
257 in_pcbbind(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred)
258 {
259 	int anonport, error;
260 
261 	INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
262 	INP_WLOCK_ASSERT(inp);
263 
264 	if (inp->inp_lport != 0 || inp->inp_laddr.s_addr != INADDR_ANY)
265 		return (EINVAL);
266 	anonport = inp->inp_lport == 0 && (nam == NULL ||
267 	    ((struct sockaddr_in *)nam)->sin_port == 0);
268 	error = in_pcbbind_setup(inp, nam, &inp->inp_laddr.s_addr,
269 	    &inp->inp_lport, cred);
270 	if (error)
271 		return (error);
272 	if (in_pcbinshash(inp) != 0) {
273 		inp->inp_laddr.s_addr = INADDR_ANY;
274 		inp->inp_lport = 0;
275 		return (EAGAIN);
276 	}
277 	if (anonport)
278 		inp->inp_flags |= INP_ANONPORT;
279 	return (0);
280 }
281 
282 /*
283  * Set up a bind operation on a PCB, performing port allocation
284  * as required, but do not actually modify the PCB. Callers can
285  * either complete the bind by setting inp_laddr/inp_lport and
286  * calling in_pcbinshash(), or they can just use the resulting
287  * port and address to authorise the sending of a once-off packet.
288  *
289  * On error, the values of *laddrp and *lportp are not changed.
290  */
291 int
292 in_pcbbind_setup(struct inpcb *inp, struct sockaddr *nam, in_addr_t *laddrp,
293     u_short *lportp, struct ucred *cred)
294 {
295 	INIT_VNET_INET(inp->inp_vnet);
296 	struct socket *so = inp->inp_socket;
297 	unsigned short *lastport;
298 	struct sockaddr_in *sin;
299 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
300 	struct in_addr laddr;
301 	u_short lport = 0;
302 	int wild = 0, reuseport = (so->so_options & SO_REUSEPORT);
303 	int error;
304 	int dorandom;
305 
306 	/*
307 	 * Because no actual state changes occur here, a global write lock on
308 	 * the pcbinfo isn't required.
309 	 */
310 	INP_INFO_LOCK_ASSERT(pcbinfo);
311 	INP_LOCK_ASSERT(inp);
312 
313 	if (TAILQ_EMPTY(&V_in_ifaddrhead)) /* XXX broken! */
314 		return (EADDRNOTAVAIL);
315 	laddr.s_addr = *laddrp;
316 	if (nam != NULL && laddr.s_addr != INADDR_ANY)
317 		return (EINVAL);
318 	if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0)
319 		wild = INPLOOKUP_WILDCARD;
320 	if (nam == NULL) {
321 		if ((error = prison_local_ip4(cred, &laddr)) != 0)
322 			return (error);
323 	} else {
324 		sin = (struct sockaddr_in *)nam;
325 		if (nam->sa_len != sizeof (*sin))
326 			return (EINVAL);
327 #ifdef notdef
328 		/*
329 		 * We should check the family, but old programs
330 		 * incorrectly fail to initialize it.
331 		 */
332 		if (sin->sin_family != AF_INET)
333 			return (EAFNOSUPPORT);
334 #endif
335 		error = prison_local_ip4(cred, &sin->sin_addr);
336 		if (error)
337 			return (error);
338 		if (sin->sin_port != *lportp) {
339 			/* Don't allow the port to change. */
340 			if (*lportp != 0)
341 				return (EINVAL);
342 			lport = sin->sin_port;
343 		}
344 		/* NB: lport is left as 0 if the port isn't being changed. */
345 		if (IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) {
346 			/*
347 			 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
348 			 * allow complete duplication of binding if
349 			 * SO_REUSEPORT is set, or if SO_REUSEADDR is set
350 			 * and a multicast address is bound on both
351 			 * new and duplicated sockets.
352 			 */
353 			if (so->so_options & SO_REUSEADDR)
354 				reuseport = SO_REUSEADDR|SO_REUSEPORT;
355 		} else if (sin->sin_addr.s_addr != INADDR_ANY) {
356 			sin->sin_port = 0;		/* yech... */
357 			bzero(&sin->sin_zero, sizeof(sin->sin_zero));
358 			/*
359 			 * Is the address a local IP address?
360 			 * If INP_NONLOCALOK is set, then the socket may be bound
361 			 * to any endpoint address, local or not.
362 			 */
363 			if (
364 #if defined(IP_NONLOCALBIND)
365 			    ((inp->inp_flags & INP_NONLOCALOK) == 0) &&
366 #endif
367 			    (ifa_ifwithaddr((struct sockaddr *)sin) == 0))
368 				return (EADDRNOTAVAIL);
369 		}
370 		laddr = sin->sin_addr;
371 		if (lport) {
372 			struct inpcb *t;
373 			struct tcptw *tw;
374 
375 			/* GROSS */
376 			if (ntohs(lport) <= V_ipport_reservedhigh &&
377 			    ntohs(lport) >= V_ipport_reservedlow &&
378 			    priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT,
379 			    0))
380 				return (EACCES);
381 			if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)) &&
382 			    priv_check_cred(inp->inp_cred,
383 			    PRIV_NETINET_REUSEPORT, 0) != 0) {
384 				t = in_pcblookup_local(pcbinfo, sin->sin_addr,
385 				    lport, INPLOOKUP_WILDCARD, cred);
386 	/*
387 	 * XXX
388 	 * This entire block sorely needs a rewrite.
389 	 */
390 				if (t &&
391 				    ((t->inp_flags & INP_TIMEWAIT) == 0) &&
392 				    (so->so_type != SOCK_STREAM ||
393 				     ntohl(t->inp_faddr.s_addr) == INADDR_ANY) &&
394 				    (ntohl(sin->sin_addr.s_addr) != INADDR_ANY ||
395 				     ntohl(t->inp_laddr.s_addr) != INADDR_ANY ||
396 				     (t->inp_socket->so_options &
397 					 SO_REUSEPORT) == 0) &&
398 				    (inp->inp_cred->cr_uid !=
399 				     t->inp_cred->cr_uid))
400 					return (EADDRINUSE);
401 			}
402 			t = in_pcblookup_local(pcbinfo, sin->sin_addr,
403 			    lport, wild, cred);
404 			if (t && (t->inp_flags & INP_TIMEWAIT)) {
405 				/*
406 				 * XXXRW: If an incpb has had its timewait
407 				 * state recycled, we treat the address as
408 				 * being in use (for now).  This is better
409 				 * than a panic, but not desirable.
410 				 */
411 				tw = intotw(inp);
412 				if (tw == NULL ||
413 				    (reuseport & tw->tw_so_options) == 0)
414 					return (EADDRINUSE);
415 			} else if (t &&
416 			    (reuseport & t->inp_socket->so_options) == 0) {
417 #ifdef INET6
418 				if (ntohl(sin->sin_addr.s_addr) !=
419 				    INADDR_ANY ||
420 				    ntohl(t->inp_laddr.s_addr) !=
421 				    INADDR_ANY ||
422 				    INP_SOCKAF(so) ==
423 				    INP_SOCKAF(t->inp_socket))
424 #endif
425 				return (EADDRINUSE);
426 			}
427 		}
428 	}
429 	if (*lportp != 0)
430 		lport = *lportp;
431 	if (lport == 0) {
432 		u_short first, last, aux;
433 		int count;
434 
435 		if (inp->inp_flags & INP_HIGHPORT) {
436 			first = V_ipport_hifirstauto;	/* sysctl */
437 			last  = V_ipport_hilastauto;
438 			lastport = &pcbinfo->ipi_lasthi;
439 		} else if (inp->inp_flags & INP_LOWPORT) {
440 			error = priv_check_cred(cred,
441 			    PRIV_NETINET_RESERVEDPORT, 0);
442 			if (error)
443 				return error;
444 			first = V_ipport_lowfirstauto;	/* 1023 */
445 			last  = V_ipport_lowlastauto;	/* 600 */
446 			lastport = &pcbinfo->ipi_lastlow;
447 		} else {
448 			first = V_ipport_firstauto;	/* sysctl */
449 			last  = V_ipport_lastauto;
450 			lastport = &pcbinfo->ipi_lastport;
451 		}
452 		/*
453 		 * For UDP, use random port allocation as long as the user
454 		 * allows it.  For TCP (and as of yet unknown) connections,
455 		 * use random port allocation only if the user allows it AND
456 		 * ipport_tick() allows it.
457 		 */
458 		if (V_ipport_randomized &&
459 			(!V_ipport_stoprandom || pcbinfo == &V_udbinfo))
460 			dorandom = 1;
461 		else
462 			dorandom = 0;
463 		/*
464 		 * It makes no sense to do random port allocation if
465 		 * we have the only port available.
466 		 */
467 		if (first == last)
468 			dorandom = 0;
469 		/* Make sure to not include UDP packets in the count. */
470 		if (pcbinfo != &V_udbinfo)
471 			V_ipport_tcpallocs++;
472 		/*
473 		 * Instead of having two loops further down counting up or down
474 		 * make sure that first is always <= last and go with only one
475 		 * code path implementing all logic.
476 		 */
477 		if (first > last) {
478 			aux = first;
479 			first = last;
480 			last = aux;
481 		}
482 
483 		if (dorandom)
484 			*lastport = first +
485 				    (arc4random() % (last - first));
486 
487 		count = last - first;
488 
489 		do {
490 			if (count-- < 0)	/* completely used? */
491 				return (EADDRNOTAVAIL);
492 			++*lastport;
493 			if (*lastport < first || *lastport > last)
494 				*lastport = first;
495 			lport = htons(*lastport);
496 		} while (in_pcblookup_local(pcbinfo, laddr,
497 		    lport, wild, cred));
498 	}
499 	*laddrp = laddr.s_addr;
500 	*lportp = lport;
501 	return (0);
502 }
503 
504 /*
505  * Connect from a socket to a specified address.
506  * Both address and port must be specified in argument sin.
507  * If don't have a local address for this socket yet,
508  * then pick one.
509  */
510 int
511 in_pcbconnect(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred)
512 {
513 	u_short lport, fport;
514 	in_addr_t laddr, faddr;
515 	int anonport, error;
516 
517 	INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
518 	INP_WLOCK_ASSERT(inp);
519 
520 	lport = inp->inp_lport;
521 	laddr = inp->inp_laddr.s_addr;
522 	anonport = (lport == 0);
523 	error = in_pcbconnect_setup(inp, nam, &laddr, &lport, &faddr, &fport,
524 	    NULL, cred);
525 	if (error)
526 		return (error);
527 
528 	/* Do the initial binding of the local address if required. */
529 	if (inp->inp_laddr.s_addr == INADDR_ANY && inp->inp_lport == 0) {
530 		inp->inp_lport = lport;
531 		inp->inp_laddr.s_addr = laddr;
532 		if (in_pcbinshash(inp) != 0) {
533 			inp->inp_laddr.s_addr = INADDR_ANY;
534 			inp->inp_lport = 0;
535 			return (EAGAIN);
536 		}
537 	}
538 
539 	/* Commit the remaining changes. */
540 	inp->inp_lport = lport;
541 	inp->inp_laddr.s_addr = laddr;
542 	inp->inp_faddr.s_addr = faddr;
543 	inp->inp_fport = fport;
544 	in_pcbrehash(inp);
545 
546 	if (anonport)
547 		inp->inp_flags |= INP_ANONPORT;
548 	return (0);
549 }
550 
551 /*
552  * Do proper source address selection on an unbound socket in case
553  * of connect. Take jails into account as well.
554  */
555 static int
556 in_pcbladdr(struct inpcb *inp, struct in_addr *faddr, struct in_addr *laddr,
557     struct ucred *cred)
558 {
559 	struct in_ifaddr *ia;
560 	struct ifaddr *ifa;
561 	struct sockaddr *sa;
562 	struct sockaddr_in *sin;
563 	struct route sro;
564 	int error;
565 
566 	KASSERT(laddr != NULL, ("%s: laddr NULL", __func__));
567 
568 	error = 0;
569 	ia = NULL;
570 	bzero(&sro, sizeof(sro));
571 
572 	sin = (struct sockaddr_in *)&sro.ro_dst;
573 	sin->sin_family = AF_INET;
574 	sin->sin_len = sizeof(struct sockaddr_in);
575 	sin->sin_addr.s_addr = faddr->s_addr;
576 
577 	/*
578 	 * If route is known our src addr is taken from the i/f,
579 	 * else punt.
580 	 *
581 	 * Find out route to destination.
582 	 */
583 	if ((inp->inp_socket->so_options & SO_DONTROUTE) == 0)
584 		in_rtalloc_ign(&sro, 0, inp->inp_inc.inc_fibnum);
585 
586 	/*
587 	 * If we found a route, use the address corresponding to
588 	 * the outgoing interface.
589 	 *
590 	 * Otherwise assume faddr is reachable on a directly connected
591 	 * network and try to find a corresponding interface to take
592 	 * the source address from.
593 	 */
594 	if (sro.ro_rt == NULL || sro.ro_rt->rt_ifp == NULL) {
595 		struct ifnet *ifp;
596 
597 		ia = ifatoia(ifa_ifwithdstaddr((struct sockaddr *)sin));
598 		if (ia == NULL)
599 			ia = ifatoia(ifa_ifwithnet((struct sockaddr *)sin));
600 		if (ia == NULL) {
601 			error = ENETUNREACH;
602 			goto done;
603 		}
604 
605 		if (cred == NULL || !jailed(cred)) {
606 			laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
607 			goto done;
608 		}
609 
610 		ifp = ia->ia_ifp;
611 		ia = NULL;
612 		IF_ADDR_LOCK(ifp);
613 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
614 
615 			sa = ifa->ifa_addr;
616 			if (sa->sa_family != AF_INET)
617 				continue;
618 			sin = (struct sockaddr_in *)sa;
619 			if (prison_check_ip4(cred, &sin->sin_addr) == 0) {
620 				ia = (struct in_ifaddr *)ifa;
621 				break;
622 			}
623 		}
624 		if (ia != NULL) {
625 			laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
626 			IF_ADDR_UNLOCK(ifp);
627 			goto done;
628 		}
629 		IF_ADDR_UNLOCK(ifp);
630 
631 		/* 3. As a last resort return the 'default' jail address. */
632 		error = prison_get_ip4(cred, laddr);
633 		goto done;
634 	}
635 
636 	/*
637 	 * If the outgoing interface on the route found is not
638 	 * a loopback interface, use the address from that interface.
639 	 * In case of jails do those three steps:
640 	 * 1. check if the interface address belongs to the jail. If so use it.
641 	 * 2. check if we have any address on the outgoing interface
642 	 *    belonging to this jail. If so use it.
643 	 * 3. as a last resort return the 'default' jail address.
644 	 */
645 	if ((sro.ro_rt->rt_ifp->if_flags & IFF_LOOPBACK) == 0) {
646 		struct ifnet *ifp;
647 
648 		/* If not jailed, use the default returned. */
649 		if (cred == NULL || !jailed(cred)) {
650 			ia = (struct in_ifaddr *)sro.ro_rt->rt_ifa;
651 			laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
652 			goto done;
653 		}
654 
655 		/* Jailed. */
656 		/* 1. Check if the iface address belongs to the jail. */
657 		sin = (struct sockaddr_in *)sro.ro_rt->rt_ifa->ifa_addr;
658 		if (prison_check_ip4(cred, &sin->sin_addr) == 0) {
659 			ia = (struct in_ifaddr *)sro.ro_rt->rt_ifa;
660 			laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
661 			goto done;
662 		}
663 
664 		/*
665 		 * 2. Check if we have any address on the outgoing interface
666 		 *    belonging to this jail.
667 		 */
668 		ifp = sro.ro_rt->rt_ifp;
669 		IF_ADDR_LOCK(ifp);
670 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
671 
672 			sa = ifa->ifa_addr;
673 			if (sa->sa_family != AF_INET)
674 				continue;
675 			sin = (struct sockaddr_in *)sa;
676 			if (prison_check_ip4(cred, &sin->sin_addr) == 0) {
677 				ia = (struct in_ifaddr *)ifa;
678 				break;
679 			}
680 		}
681 		if (ia != NULL) {
682 			laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
683 			IF_ADDR_UNLOCK(ifp);
684 			goto done;
685 		}
686 		IF_ADDR_UNLOCK(ifp);
687 
688 		/* 3. As a last resort return the 'default' jail address. */
689 		error = prison_get_ip4(cred, laddr);
690 		goto done;
691 	}
692 
693 	/*
694 	 * The outgoing interface is marked with 'loopback net', so a route
695 	 * to ourselves is here.
696 	 * Try to find the interface of the destination address and then
697 	 * take the address from there. That interface is not necessarily
698 	 * a loopback interface.
699 	 * In case of jails, check that it is an address of the jail
700 	 * and if we cannot find, fall back to the 'default' jail address.
701 	 */
702 	if ((sro.ro_rt->rt_ifp->if_flags & IFF_LOOPBACK) != 0) {
703 		struct sockaddr_in sain;
704 
705 		bzero(&sain, sizeof(struct sockaddr_in));
706 		sain.sin_family = AF_INET;
707 		sain.sin_len = sizeof(struct sockaddr_in);
708 		sain.sin_addr.s_addr = faddr->s_addr;
709 
710 		ia = ifatoia(ifa_ifwithdstaddr(sintosa(&sain)));
711 		if (ia == NULL)
712 			ia = ifatoia(ifa_ifwithnet(sintosa(&sain)));
713 
714 		if (cred == NULL || !jailed(cred)) {
715 #if __FreeBSD_version < 800000
716 			if (ia == NULL)
717 				ia = (struct in_ifaddr *)sro.ro_rt->rt_ifa;
718 #endif
719 			if (ia == NULL) {
720 				error = ENETUNREACH;
721 				goto done;
722 			}
723 			laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
724 			goto done;
725 		}
726 
727 		/* Jailed. */
728 		if (ia != NULL) {
729 			struct ifnet *ifp;
730 
731 			ifp = ia->ia_ifp;
732 			ia = NULL;
733 			IF_ADDR_LOCK(ifp);
734 			TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
735 
736 				sa = ifa->ifa_addr;
737 				if (sa->sa_family != AF_INET)
738 					continue;
739 				sin = (struct sockaddr_in *)sa;
740 				if (prison_check_ip4(cred,
741 				    &sin->sin_addr) == 0) {
742 					ia = (struct in_ifaddr *)ifa;
743 					break;
744 				}
745 			}
746 			if (ia != NULL) {
747 				laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
748 				IF_ADDR_UNLOCK(ifp);
749 				goto done;
750 			}
751 			IF_ADDR_UNLOCK(ifp);
752 		}
753 
754 		/* 3. As a last resort return the 'default' jail address. */
755 		error = prison_get_ip4(cred, laddr);
756 		goto done;
757 	}
758 
759 done:
760 	if (sro.ro_rt != NULL)
761 		RTFREE(sro.ro_rt);
762 	return (error);
763 }
764 
765 /*
766  * Set up for a connect from a socket to the specified address.
767  * On entry, *laddrp and *lportp should contain the current local
768  * address and port for the PCB; these are updated to the values
769  * that should be placed in inp_laddr and inp_lport to complete
770  * the connect.
771  *
772  * On success, *faddrp and *fportp will be set to the remote address
773  * and port. These are not updated in the error case.
774  *
775  * If the operation fails because the connection already exists,
776  * *oinpp will be set to the PCB of that connection so that the
777  * caller can decide to override it. In all other cases, *oinpp
778  * is set to NULL.
779  */
780 int
781 in_pcbconnect_setup(struct inpcb *inp, struct sockaddr *nam,
782     in_addr_t *laddrp, u_short *lportp, in_addr_t *faddrp, u_short *fportp,
783     struct inpcb **oinpp, struct ucred *cred)
784 {
785 	INIT_VNET_INET(inp->inp_vnet);
786 	struct sockaddr_in *sin = (struct sockaddr_in *)nam;
787 	struct in_ifaddr *ia;
788 	struct inpcb *oinp;
789 	struct in_addr laddr, faddr;
790 	u_short lport, fport;
791 	int error;
792 
793 	/*
794 	 * Because a global state change doesn't actually occur here, a read
795 	 * lock is sufficient.
796 	 */
797 	INP_INFO_LOCK_ASSERT(inp->inp_pcbinfo);
798 	INP_LOCK_ASSERT(inp);
799 
800 	if (oinpp != NULL)
801 		*oinpp = NULL;
802 	if (nam->sa_len != sizeof (*sin))
803 		return (EINVAL);
804 	if (sin->sin_family != AF_INET)
805 		return (EAFNOSUPPORT);
806 	if (sin->sin_port == 0)
807 		return (EADDRNOTAVAIL);
808 	laddr.s_addr = *laddrp;
809 	lport = *lportp;
810 	faddr = sin->sin_addr;
811 	fport = sin->sin_port;
812 
813 	if (!TAILQ_EMPTY(&V_in_ifaddrhead)) {
814 		/*
815 		 * If the destination address is INADDR_ANY,
816 		 * use the primary local address.
817 		 * If the supplied address is INADDR_BROADCAST,
818 		 * and the primary interface supports broadcast,
819 		 * choose the broadcast address for that interface.
820 		 */
821 		if (faddr.s_addr == INADDR_ANY) {
822 			faddr =
823 			    IA_SIN(TAILQ_FIRST(&V_in_ifaddrhead))->sin_addr;
824 			if (cred != NULL &&
825 			    (error = prison_get_ip4(cred, &faddr)) != 0)
826 				return (error);
827 		} else if (faddr.s_addr == (u_long)INADDR_BROADCAST &&
828 		    (TAILQ_FIRST(&V_in_ifaddrhead)->ia_ifp->if_flags &
829 		    IFF_BROADCAST))
830 			faddr = satosin(&TAILQ_FIRST(
831 			    &V_in_ifaddrhead)->ia_broadaddr)->sin_addr;
832 	}
833 	if (laddr.s_addr == INADDR_ANY) {
834 		error = in_pcbladdr(inp, &faddr, &laddr, cred);
835 		if (error)
836 			return (error);
837 
838 		/*
839 		 * If the destination address is multicast and an outgoing
840 		 * interface has been set as a multicast option, use the
841 		 * address of that interface as our source address.
842 		 */
843 		if (IN_MULTICAST(ntohl(faddr.s_addr)) &&
844 		    inp->inp_moptions != NULL) {
845 			struct ip_moptions *imo;
846 			struct ifnet *ifp;
847 
848 			imo = inp->inp_moptions;
849 			if (imo->imo_multicast_ifp != NULL) {
850 				ifp = imo->imo_multicast_ifp;
851 				TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link)
852 					if (ia->ia_ifp == ifp)
853 						break;
854 				if (ia == NULL)
855 					return (EADDRNOTAVAIL);
856 				laddr = ia->ia_addr.sin_addr;
857 			}
858 		}
859 	}
860 
861 	oinp = in_pcblookup_hash(inp->inp_pcbinfo, faddr, fport, laddr, lport,
862 	    0, NULL);
863 	if (oinp != NULL) {
864 		if (oinpp != NULL)
865 			*oinpp = oinp;
866 		return (EADDRINUSE);
867 	}
868 	if (lport == 0) {
869 		error = in_pcbbind_setup(inp, NULL, &laddr.s_addr, &lport,
870 		    cred);
871 		if (error)
872 			return (error);
873 	}
874 	*laddrp = laddr.s_addr;
875 	*lportp = lport;
876 	*faddrp = faddr.s_addr;
877 	*fportp = fport;
878 	return (0);
879 }
880 
881 void
882 in_pcbdisconnect(struct inpcb *inp)
883 {
884 
885 	INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
886 	INP_WLOCK_ASSERT(inp);
887 
888 	inp->inp_faddr.s_addr = INADDR_ANY;
889 	inp->inp_fport = 0;
890 	in_pcbrehash(inp);
891 }
892 
893 /*
894  * in_pcbdetach() is responsibe for disassociating a socket from an inpcb.
895  * For most protocols, this will be invoked immediately prior to calling
896  * in_pcbfree().  However, with TCP the inpcb may significantly outlive the
897  * socket, in which case in_pcbfree() is deferred.
898  */
899 void
900 in_pcbdetach(struct inpcb *inp)
901 {
902 
903 	KASSERT(inp->inp_socket != NULL, ("%s: inp_socket == NULL", __func__));
904 
905 	inp->inp_socket->so_pcb = NULL;
906 	inp->inp_socket = NULL;
907 }
908 
909 /*
910  * in_pcbfree_internal() frees an inpcb that has been detached from its
911  * socket, and whose reference count has reached 0.  It will also remove the
912  * inpcb from any global lists it might remain on.
913  */
914 static void
915 in_pcbfree_internal(struct inpcb *inp)
916 {
917 	struct inpcbinfo *ipi = inp->inp_pcbinfo;
918 
919 	KASSERT(inp->inp_socket == NULL, ("%s: inp_socket != NULL", __func__));
920 	KASSERT(inp->inp_refcount == 0, ("%s: refcount !0", __func__));
921 
922 	INP_INFO_WLOCK_ASSERT(ipi);
923 	INP_WLOCK_ASSERT(inp);
924 
925 #ifdef IPSEC
926 	if (inp->inp_sp != NULL)
927 		ipsec_delete_pcbpolicy(inp);
928 #endif /* IPSEC */
929 	inp->inp_gencnt = ++ipi->ipi_gencnt;
930 	in_pcbremlists(inp);
931 #ifdef INET6
932 	if (inp->inp_vflag & INP_IPV6PROTO) {
933 		ip6_freepcbopts(inp->in6p_outputopts);
934 		if (inp->in6p_moptions != NULL)
935 			ip6_freemoptions(inp->in6p_moptions);
936 	}
937 #endif
938 	if (inp->inp_options)
939 		(void)m_free(inp->inp_options);
940 	if (inp->inp_moptions != NULL)
941 		inp_freemoptions(inp->inp_moptions);
942 	inp->inp_vflag = 0;
943 	crfree(inp->inp_cred);
944 
945 #ifdef MAC
946 	mac_inpcb_destroy(inp);
947 #endif
948 	INP_WUNLOCK(inp);
949 	uma_zfree(ipi->ipi_zone, inp);
950 }
951 
952 /*
953  * in_pcbref() bumps the reference count on an inpcb in order to maintain
954  * stability of an inpcb pointer despite the inpcb lock being released.  This
955  * is used in TCP when the inpcbinfo lock needs to be acquired or upgraded,
956  * but where the inpcb lock is already held.
957  *
958  * While the inpcb will not be freed, releasing the inpcb lock means that the
959  * connection's state may change, so the caller should be careful to
960  * revalidate any cached state on reacquiring the lock.  Drop the reference
961  * using in_pcbrele().
962  */
963 void
964 in_pcbref(struct inpcb *inp)
965 {
966 
967 	INP_WLOCK_ASSERT(inp);
968 
969 	KASSERT(inp->inp_refcount > 0, ("%s: refcount 0", __func__));
970 
971 	inp->inp_refcount++;
972 }
973 
974 /*
975  * Drop a refcount on an inpcb elevated using in_pcbref(); because a call to
976  * in_pcbfree() may have been made between in_pcbref() and in_pcbrele(), we
977  * return a flag indicating whether or not the inpcb remains valid.  If it is
978  * valid, we return with the inpcb lock held.
979  */
980 int
981 in_pcbrele(struct inpcb *inp)
982 {
983 #ifdef INVARIANTS
984 	struct inpcbinfo *ipi = inp->inp_pcbinfo;
985 #endif
986 
987 	KASSERT(inp->inp_refcount > 0, ("%s: refcount 0", __func__));
988 
989 	INP_INFO_WLOCK_ASSERT(ipi);
990 	INP_WLOCK_ASSERT(inp);
991 
992 	inp->inp_refcount--;
993 	if (inp->inp_refcount > 0)
994 		return (0);
995 	in_pcbfree_internal(inp);
996 	return (1);
997 }
998 
999 /*
1000  * Unconditionally schedule an inpcb to be freed by decrementing its
1001  * reference count, which should occur only after the inpcb has been detached
1002  * from its socket.  If another thread holds a temporary reference (acquired
1003  * using in_pcbref()) then the free is deferred until that reference is
1004  * released using in_pcbrele(), but the inpcb is still unlocked.
1005  */
1006 void
1007 in_pcbfree(struct inpcb *inp)
1008 {
1009 #ifdef INVARIANTS
1010 	struct inpcbinfo *ipi = inp->inp_pcbinfo;
1011 #endif
1012 
1013 	KASSERT(inp->inp_socket == NULL, ("%s: inp_socket != NULL",
1014 	    __func__));
1015 
1016 	INP_INFO_WLOCK_ASSERT(ipi);
1017 	INP_WLOCK_ASSERT(inp);
1018 
1019 	if (!in_pcbrele(inp))
1020 		INP_WUNLOCK(inp);
1021 }
1022 
1023 /*
1024  * in_pcbdrop() removes an inpcb from hashed lists, releasing its address and
1025  * port reservation, and preventing it from being returned by inpcb lookups.
1026  *
1027  * It is used by TCP to mark an inpcb as unused and avoid future packet
1028  * delivery or event notification when a socket remains open but TCP has
1029  * closed.  This might occur as a result of a shutdown()-initiated TCP close
1030  * or a RST on the wire, and allows the port binding to be reused while still
1031  * maintaining the invariant that so_pcb always points to a valid inpcb until
1032  * in_pcbdetach().
1033  *
1034  * XXXRW: An inp_lport of 0 is used to indicate that the inpcb is not on hash
1035  * lists, but can lead to confusing netstat output, as open sockets with
1036  * closed TCP connections will no longer appear to have their bound port
1037  * number.  An explicit flag would be better, as it would allow us to leave
1038  * the port number intact after the connection is dropped.
1039  *
1040  * XXXRW: Possibly in_pcbdrop() should also prevent future notifications by
1041  * in_pcbnotifyall() and in_pcbpurgeif0()?
1042  */
1043 void
1044 in_pcbdrop(struct inpcb *inp)
1045 {
1046 
1047 	INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
1048 	INP_WLOCK_ASSERT(inp);
1049 
1050 	inp->inp_flags |= INP_DROPPED;
1051 	if (inp->inp_flags & INP_INHASHLIST) {
1052 		struct inpcbport *phd = inp->inp_phd;
1053 
1054 		LIST_REMOVE(inp, inp_hash);
1055 		LIST_REMOVE(inp, inp_portlist);
1056 		if (LIST_FIRST(&phd->phd_pcblist) == NULL) {
1057 			LIST_REMOVE(phd, phd_hash);
1058 			free(phd, M_PCB);
1059 		}
1060 		inp->inp_flags &= ~INP_INHASHLIST;
1061 	}
1062 }
1063 
1064 /*
1065  * Common routines to return the socket addresses associated with inpcbs.
1066  */
1067 struct sockaddr *
1068 in_sockaddr(in_port_t port, struct in_addr *addr_p)
1069 {
1070 	struct sockaddr_in *sin;
1071 
1072 	sin = malloc(sizeof *sin, M_SONAME,
1073 		M_WAITOK | M_ZERO);
1074 	sin->sin_family = AF_INET;
1075 	sin->sin_len = sizeof(*sin);
1076 	sin->sin_addr = *addr_p;
1077 	sin->sin_port = port;
1078 
1079 	return (struct sockaddr *)sin;
1080 }
1081 
1082 int
1083 in_getsockaddr(struct socket *so, struct sockaddr **nam)
1084 {
1085 	struct inpcb *inp;
1086 	struct in_addr addr;
1087 	in_port_t port;
1088 
1089 	inp = sotoinpcb(so);
1090 	KASSERT(inp != NULL, ("in_getsockaddr: inp == NULL"));
1091 
1092 	INP_RLOCK(inp);
1093 	port = inp->inp_lport;
1094 	addr = inp->inp_laddr;
1095 	INP_RUNLOCK(inp);
1096 
1097 	*nam = in_sockaddr(port, &addr);
1098 	return 0;
1099 }
1100 
1101 int
1102 in_getpeeraddr(struct socket *so, struct sockaddr **nam)
1103 {
1104 	struct inpcb *inp;
1105 	struct in_addr addr;
1106 	in_port_t port;
1107 
1108 	inp = sotoinpcb(so);
1109 	KASSERT(inp != NULL, ("in_getpeeraddr: inp == NULL"));
1110 
1111 	INP_RLOCK(inp);
1112 	port = inp->inp_fport;
1113 	addr = inp->inp_faddr;
1114 	INP_RUNLOCK(inp);
1115 
1116 	*nam = in_sockaddr(port, &addr);
1117 	return 0;
1118 }
1119 
1120 void
1121 in_pcbnotifyall(struct inpcbinfo *pcbinfo, struct in_addr faddr, int errno,
1122     struct inpcb *(*notify)(struct inpcb *, int))
1123 {
1124 	struct inpcb *inp, *inp_temp;
1125 
1126 	INP_INFO_WLOCK(pcbinfo);
1127 	LIST_FOREACH_SAFE(inp, pcbinfo->ipi_listhead, inp_list, inp_temp) {
1128 		INP_WLOCK(inp);
1129 #ifdef INET6
1130 		if ((inp->inp_vflag & INP_IPV4) == 0) {
1131 			INP_WUNLOCK(inp);
1132 			continue;
1133 		}
1134 #endif
1135 		if (inp->inp_faddr.s_addr != faddr.s_addr ||
1136 		    inp->inp_socket == NULL) {
1137 			INP_WUNLOCK(inp);
1138 			continue;
1139 		}
1140 		if ((*notify)(inp, errno))
1141 			INP_WUNLOCK(inp);
1142 	}
1143 	INP_INFO_WUNLOCK(pcbinfo);
1144 }
1145 
1146 void
1147 in_pcbpurgeif0(struct inpcbinfo *pcbinfo, struct ifnet *ifp)
1148 {
1149 	struct inpcb *inp;
1150 	struct ip_moptions *imo;
1151 	int i, gap;
1152 
1153 	INP_INFO_RLOCK(pcbinfo);
1154 	LIST_FOREACH(inp, pcbinfo->ipi_listhead, inp_list) {
1155 		INP_WLOCK(inp);
1156 		imo = inp->inp_moptions;
1157 		if ((inp->inp_vflag & INP_IPV4) &&
1158 		    imo != NULL) {
1159 			/*
1160 			 * Unselect the outgoing interface if it is being
1161 			 * detached.
1162 			 */
1163 			if (imo->imo_multicast_ifp == ifp)
1164 				imo->imo_multicast_ifp = NULL;
1165 
1166 			/*
1167 			 * Drop multicast group membership if we joined
1168 			 * through the interface being detached.
1169 			 */
1170 			for (i = 0, gap = 0; i < imo->imo_num_memberships;
1171 			    i++) {
1172 				if (imo->imo_membership[i]->inm_ifp == ifp) {
1173 					in_delmulti(imo->imo_membership[i]);
1174 					gap++;
1175 				} else if (gap != 0)
1176 					imo->imo_membership[i - gap] =
1177 					    imo->imo_membership[i];
1178 			}
1179 			imo->imo_num_memberships -= gap;
1180 		}
1181 		INP_WUNLOCK(inp);
1182 	}
1183 	INP_INFO_RUNLOCK(pcbinfo);
1184 }
1185 
1186 /*
1187  * Lookup a PCB based on the local address and port.
1188  */
1189 #define INP_LOOKUP_MAPPED_PCB_COST	3
1190 struct inpcb *
1191 in_pcblookup_local(struct inpcbinfo *pcbinfo, struct in_addr laddr,
1192     u_short lport, int wild_okay, struct ucred *cred)
1193 {
1194 	struct inpcb *inp;
1195 #ifdef INET6
1196 	int matchwild = 3 + INP_LOOKUP_MAPPED_PCB_COST;
1197 #else
1198 	int matchwild = 3;
1199 #endif
1200 	int wildcard;
1201 
1202 	INP_INFO_LOCK_ASSERT(pcbinfo);
1203 
1204 	if (!wild_okay) {
1205 		struct inpcbhead *head;
1206 		/*
1207 		 * Look for an unconnected (wildcard foreign addr) PCB that
1208 		 * matches the local address and port we're looking for.
1209 		 */
1210 		head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport,
1211 		    0, pcbinfo->ipi_hashmask)];
1212 		LIST_FOREACH(inp, head, inp_hash) {
1213 #ifdef INET6
1214 			/* XXX inp locking */
1215 			if ((inp->inp_vflag & INP_IPV4) == 0)
1216 				continue;
1217 #endif
1218 			if (inp->inp_faddr.s_addr == INADDR_ANY &&
1219 			    inp->inp_laddr.s_addr == laddr.s_addr &&
1220 			    inp->inp_lport == lport) {
1221 				/*
1222 				 * Found?
1223 				 */
1224 				if (cred == NULL ||
1225 				    inp->inp_cred->cr_prison == cred->cr_prison)
1226 					return (inp);
1227 			}
1228 		}
1229 		/*
1230 		 * Not found.
1231 		 */
1232 		return (NULL);
1233 	} else {
1234 		struct inpcbporthead *porthash;
1235 		struct inpcbport *phd;
1236 		struct inpcb *match = NULL;
1237 		/*
1238 		 * Best fit PCB lookup.
1239 		 *
1240 		 * First see if this local port is in use by looking on the
1241 		 * port hash list.
1242 		 */
1243 		porthash = &pcbinfo->ipi_porthashbase[INP_PCBPORTHASH(lport,
1244 		    pcbinfo->ipi_porthashmask)];
1245 		LIST_FOREACH(phd, porthash, phd_hash) {
1246 			if (phd->phd_port == lport)
1247 				break;
1248 		}
1249 		if (phd != NULL) {
1250 			/*
1251 			 * Port is in use by one or more PCBs. Look for best
1252 			 * fit.
1253 			 */
1254 			LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) {
1255 				wildcard = 0;
1256 				if (cred != NULL &&
1257 				    inp->inp_cred->cr_prison != cred->cr_prison)
1258 					continue;
1259 #ifdef INET6
1260 				/* XXX inp locking */
1261 				if ((inp->inp_vflag & INP_IPV4) == 0)
1262 					continue;
1263 				/*
1264 				 * We never select the PCB that has
1265 				 * INP_IPV6 flag and is bound to :: if
1266 				 * we have another PCB which is bound
1267 				 * to 0.0.0.0.  If a PCB has the
1268 				 * INP_IPV6 flag, then we set its cost
1269 				 * higher than IPv4 only PCBs.
1270 				 *
1271 				 * Note that the case only happens
1272 				 * when a socket is bound to ::, under
1273 				 * the condition that the use of the
1274 				 * mapped address is allowed.
1275 				 */
1276 				if ((inp->inp_vflag & INP_IPV6) != 0)
1277 					wildcard += INP_LOOKUP_MAPPED_PCB_COST;
1278 #endif
1279 				if (inp->inp_faddr.s_addr != INADDR_ANY)
1280 					wildcard++;
1281 				if (inp->inp_laddr.s_addr != INADDR_ANY) {
1282 					if (laddr.s_addr == INADDR_ANY)
1283 						wildcard++;
1284 					else if (inp->inp_laddr.s_addr != laddr.s_addr)
1285 						continue;
1286 				} else {
1287 					if (laddr.s_addr != INADDR_ANY)
1288 						wildcard++;
1289 				}
1290 				if (wildcard < matchwild) {
1291 					match = inp;
1292 					matchwild = wildcard;
1293 					if (matchwild == 0)
1294 						break;
1295 				}
1296 			}
1297 		}
1298 		return (match);
1299 	}
1300 }
1301 #undef INP_LOOKUP_MAPPED_PCB_COST
1302 
1303 /*
1304  * Lookup PCB in hash list.
1305  */
1306 struct inpcb *
1307 in_pcblookup_hash(struct inpcbinfo *pcbinfo, struct in_addr faddr,
1308     u_int fport_arg, struct in_addr laddr, u_int lport_arg, int wildcard,
1309     struct ifnet *ifp)
1310 {
1311 	struct inpcbhead *head;
1312 	struct inpcb *inp, *tmpinp;
1313 	u_short fport = fport_arg, lport = lport_arg;
1314 
1315 	INP_INFO_LOCK_ASSERT(pcbinfo);
1316 
1317 	/*
1318 	 * First look for an exact match.
1319 	 */
1320 	tmpinp = NULL;
1321 	head = &pcbinfo->ipi_hashbase[INP_PCBHASH(faddr.s_addr, lport, fport,
1322 	    pcbinfo->ipi_hashmask)];
1323 	LIST_FOREACH(inp, head, inp_hash) {
1324 #ifdef INET6
1325 		/* XXX inp locking */
1326 		if ((inp->inp_vflag & INP_IPV4) == 0)
1327 			continue;
1328 #endif
1329 		if (inp->inp_faddr.s_addr == faddr.s_addr &&
1330 		    inp->inp_laddr.s_addr == laddr.s_addr &&
1331 		    inp->inp_fport == fport &&
1332 		    inp->inp_lport == lport) {
1333 			/*
1334 			 * XXX We should be able to directly return
1335 			 * the inp here, without any checks.
1336 			 * Well unless both bound with SO_REUSEPORT?
1337 			 */
1338 			if (jailed(inp->inp_cred))
1339 				return (inp);
1340 			if (tmpinp == NULL)
1341 				tmpinp = inp;
1342 		}
1343 	}
1344 	if (tmpinp != NULL)
1345 		return (tmpinp);
1346 
1347 	/*
1348 	 * Then look for a wildcard match, if requested.
1349 	 */
1350 	if (wildcard == INPLOOKUP_WILDCARD) {
1351 		struct inpcb *local_wild = NULL, *local_exact = NULL;
1352 #ifdef INET6
1353 		struct inpcb *local_wild_mapped = NULL;
1354 #endif
1355 		struct inpcb *jail_wild = NULL;
1356 		int injail;
1357 
1358 		/*
1359 		 * Order of socket selection - we always prefer jails.
1360 		 *      1. jailed, non-wild.
1361 		 *      2. jailed, wild.
1362 		 *      3. non-jailed, non-wild.
1363 		 *      4. non-jailed, wild.
1364 		 */
1365 
1366 		head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport,
1367 		    0, pcbinfo->ipi_hashmask)];
1368 		LIST_FOREACH(inp, head, inp_hash) {
1369 #ifdef INET6
1370 			/* XXX inp locking */
1371 			if ((inp->inp_vflag & INP_IPV4) == 0)
1372 				continue;
1373 #endif
1374 			if (inp->inp_faddr.s_addr != INADDR_ANY ||
1375 			    inp->inp_lport != lport)
1376 				continue;
1377 
1378 			/* XXX inp locking */
1379 			if (ifp && ifp->if_type == IFT_FAITH &&
1380 			    (inp->inp_flags & INP_FAITH) == 0)
1381 				continue;
1382 
1383 			injail = jailed(inp->inp_cred);
1384 			if (injail) {
1385 				if (prison_check_ip4(inp->inp_cred,
1386 				    &laddr) != 0)
1387 					continue;
1388 			} else {
1389 				if (local_exact != NULL)
1390 					continue;
1391 			}
1392 
1393 			if (inp->inp_laddr.s_addr == laddr.s_addr) {
1394 				if (injail)
1395 					return (inp);
1396 				else
1397 					local_exact = inp;
1398 			} else if (inp->inp_laddr.s_addr == INADDR_ANY) {
1399 #ifdef INET6
1400 				/* XXX inp locking, NULL check */
1401 				if (inp->inp_vflag & INP_IPV6PROTO)
1402 					local_wild_mapped = inp;
1403 				else
1404 #endif /* INET6 */
1405 					if (injail)
1406 						jail_wild = inp;
1407 					else
1408 						local_wild = inp;
1409 			}
1410 		} /* LIST_FOREACH */
1411 		if (jail_wild != NULL)
1412 			return (jail_wild);
1413 		if (local_exact != NULL)
1414 			return (local_exact);
1415 		if (local_wild != NULL)
1416 			return (local_wild);
1417 #ifdef INET6
1418 		if (local_wild_mapped != NULL)
1419 			return (local_wild_mapped);
1420 #endif /* defined(INET6) */
1421 	} /* if (wildcard == INPLOOKUP_WILDCARD) */
1422 
1423 	return (NULL);
1424 }
1425 
1426 /*
1427  * Insert PCB onto various hash lists.
1428  */
1429 int
1430 in_pcbinshash(struct inpcb *inp)
1431 {
1432 	struct inpcbhead *pcbhash;
1433 	struct inpcbporthead *pcbporthash;
1434 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1435 	struct inpcbport *phd;
1436 	u_int32_t hashkey_faddr;
1437 
1438 	INP_INFO_WLOCK_ASSERT(pcbinfo);
1439 	INP_WLOCK_ASSERT(inp);
1440 	KASSERT((inp->inp_flags & INP_INHASHLIST) == 0,
1441 	    ("in_pcbinshash: INP_INHASHLIST"));
1442 
1443 #ifdef INET6
1444 	if (inp->inp_vflag & INP_IPV6)
1445 		hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX */;
1446 	else
1447 #endif /* INET6 */
1448 	hashkey_faddr = inp->inp_faddr.s_addr;
1449 
1450 	pcbhash = &pcbinfo->ipi_hashbase[INP_PCBHASH(hashkey_faddr,
1451 		 inp->inp_lport, inp->inp_fport, pcbinfo->ipi_hashmask)];
1452 
1453 	pcbporthash = &pcbinfo->ipi_porthashbase[
1454 	    INP_PCBPORTHASH(inp->inp_lport, pcbinfo->ipi_porthashmask)];
1455 
1456 	/*
1457 	 * Go through port list and look for a head for this lport.
1458 	 */
1459 	LIST_FOREACH(phd, pcbporthash, phd_hash) {
1460 		if (phd->phd_port == inp->inp_lport)
1461 			break;
1462 	}
1463 	/*
1464 	 * If none exists, malloc one and tack it on.
1465 	 */
1466 	if (phd == NULL) {
1467 		phd = malloc(sizeof(struct inpcbport), M_PCB, M_NOWAIT);
1468 		if (phd == NULL) {
1469 			return (ENOBUFS); /* XXX */
1470 		}
1471 		phd->phd_port = inp->inp_lport;
1472 		LIST_INIT(&phd->phd_pcblist);
1473 		LIST_INSERT_HEAD(pcbporthash, phd, phd_hash);
1474 	}
1475 	inp->inp_phd = phd;
1476 	LIST_INSERT_HEAD(&phd->phd_pcblist, inp, inp_portlist);
1477 	LIST_INSERT_HEAD(pcbhash, inp, inp_hash);
1478 	inp->inp_flags |= INP_INHASHLIST;
1479 	return (0);
1480 }
1481 
1482 /*
1483  * Move PCB to the proper hash bucket when { faddr, fport } have  been
1484  * changed. NOTE: This does not handle the case of the lport changing (the
1485  * hashed port list would have to be updated as well), so the lport must
1486  * not change after in_pcbinshash() has been called.
1487  */
1488 void
1489 in_pcbrehash(struct inpcb *inp)
1490 {
1491 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1492 	struct inpcbhead *head;
1493 	u_int32_t hashkey_faddr;
1494 
1495 	INP_INFO_WLOCK_ASSERT(pcbinfo);
1496 	INP_WLOCK_ASSERT(inp);
1497 	KASSERT(inp->inp_flags & INP_INHASHLIST,
1498 	    ("in_pcbrehash: !INP_INHASHLIST"));
1499 
1500 #ifdef INET6
1501 	if (inp->inp_vflag & INP_IPV6)
1502 		hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX */;
1503 	else
1504 #endif /* INET6 */
1505 	hashkey_faddr = inp->inp_faddr.s_addr;
1506 
1507 	head = &pcbinfo->ipi_hashbase[INP_PCBHASH(hashkey_faddr,
1508 		inp->inp_lport, inp->inp_fport, pcbinfo->ipi_hashmask)];
1509 
1510 	LIST_REMOVE(inp, inp_hash);
1511 	LIST_INSERT_HEAD(head, inp, inp_hash);
1512 }
1513 
1514 /*
1515  * Remove PCB from various lists.
1516  */
1517 static void
1518 in_pcbremlists(struct inpcb *inp)
1519 {
1520 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1521 
1522 	INP_INFO_WLOCK_ASSERT(pcbinfo);
1523 	INP_WLOCK_ASSERT(inp);
1524 
1525 	inp->inp_gencnt = ++pcbinfo->ipi_gencnt;
1526 	if (inp->inp_flags & INP_INHASHLIST) {
1527 		struct inpcbport *phd = inp->inp_phd;
1528 
1529 		LIST_REMOVE(inp, inp_hash);
1530 		LIST_REMOVE(inp, inp_portlist);
1531 		if (LIST_FIRST(&phd->phd_pcblist) == NULL) {
1532 			LIST_REMOVE(phd, phd_hash);
1533 			free(phd, M_PCB);
1534 		}
1535 		inp->inp_flags &= ~INP_INHASHLIST;
1536 	}
1537 	LIST_REMOVE(inp, inp_list);
1538 	pcbinfo->ipi_count--;
1539 }
1540 
1541 /*
1542  * A set label operation has occurred at the socket layer, propagate the
1543  * label change into the in_pcb for the socket.
1544  */
1545 void
1546 in_pcbsosetlabel(struct socket *so)
1547 {
1548 #ifdef MAC
1549 	struct inpcb *inp;
1550 
1551 	inp = sotoinpcb(so);
1552 	KASSERT(inp != NULL, ("in_pcbsosetlabel: so->so_pcb == NULL"));
1553 
1554 	INP_WLOCK(inp);
1555 	SOCK_LOCK(so);
1556 	mac_inpcb_sosetlabel(so, inp);
1557 	SOCK_UNLOCK(so);
1558 	INP_WUNLOCK(inp);
1559 #endif
1560 }
1561 
1562 /*
1563  * ipport_tick runs once per second, determining if random port allocation
1564  * should be continued.  If more than ipport_randomcps ports have been
1565  * allocated in the last second, then we return to sequential port
1566  * allocation. We return to random allocation only once we drop below
1567  * ipport_randomcps for at least ipport_randomtime seconds.
1568  */
1569 void
1570 ipport_tick(void *xtp)
1571 {
1572 	VNET_ITERATOR_DECL(vnet_iter);
1573 
1574 	VNET_LIST_RLOCK();
1575 	VNET_FOREACH(vnet_iter) {
1576 		CURVNET_SET(vnet_iter);	/* XXX appease INVARIANTS here */
1577 		INIT_VNET_INET(vnet_iter);
1578 		if (V_ipport_tcpallocs <=
1579 		    V_ipport_tcplastcount + V_ipport_randomcps) {
1580 			if (V_ipport_stoprandom > 0)
1581 				V_ipport_stoprandom--;
1582 		} else
1583 			V_ipport_stoprandom = V_ipport_randomtime;
1584 		V_ipport_tcplastcount = V_ipport_tcpallocs;
1585 		CURVNET_RESTORE();
1586 	}
1587 	VNET_LIST_RUNLOCK();
1588 	callout_reset(&ipport_tick_callout, hz, ipport_tick, NULL);
1589 }
1590 
1591 void
1592 inp_wlock(struct inpcb *inp)
1593 {
1594 
1595 	INP_WLOCK(inp);
1596 }
1597 
1598 void
1599 inp_wunlock(struct inpcb *inp)
1600 {
1601 
1602 	INP_WUNLOCK(inp);
1603 }
1604 
1605 void
1606 inp_rlock(struct inpcb *inp)
1607 {
1608 
1609 	INP_RLOCK(inp);
1610 }
1611 
1612 void
1613 inp_runlock(struct inpcb *inp)
1614 {
1615 
1616 	INP_RUNLOCK(inp);
1617 }
1618 
1619 #ifdef INVARIANTS
1620 void
1621 inp_lock_assert(struct inpcb *inp)
1622 {
1623 
1624 	INP_WLOCK_ASSERT(inp);
1625 }
1626 
1627 void
1628 inp_unlock_assert(struct inpcb *inp)
1629 {
1630 
1631 	INP_UNLOCK_ASSERT(inp);
1632 }
1633 #endif
1634 
1635 void
1636 inp_apply_all(void (*func)(struct inpcb *, void *), void *arg)
1637 {
1638 	INIT_VNET_INET(curvnet);
1639 	struct inpcb *inp;
1640 
1641 	INP_INFO_RLOCK(&V_tcbinfo);
1642 	LIST_FOREACH(inp, V_tcbinfo.ipi_listhead, inp_list) {
1643 		INP_WLOCK(inp);
1644 		func(inp, arg);
1645 		INP_WUNLOCK(inp);
1646 	}
1647 	INP_INFO_RUNLOCK(&V_tcbinfo);
1648 }
1649 
1650 struct socket *
1651 inp_inpcbtosocket(struct inpcb *inp)
1652 {
1653 
1654 	INP_WLOCK_ASSERT(inp);
1655 	return (inp->inp_socket);
1656 }
1657 
1658 struct tcpcb *
1659 inp_inpcbtotcpcb(struct inpcb *inp)
1660 {
1661 
1662 	INP_WLOCK_ASSERT(inp);
1663 	return ((struct tcpcb *)inp->inp_ppcb);
1664 }
1665 
1666 int
1667 inp_ip_tos_get(const struct inpcb *inp)
1668 {
1669 
1670 	return (inp->inp_ip_tos);
1671 }
1672 
1673 void
1674 inp_ip_tos_set(struct inpcb *inp, int val)
1675 {
1676 
1677 	inp->inp_ip_tos = val;
1678 }
1679 
1680 void
1681 inp_4tuple_get(struct inpcb *inp, uint32_t *laddr, uint16_t *lp,
1682     uint32_t *faddr, uint16_t *fp)
1683 {
1684 
1685 	INP_LOCK_ASSERT(inp);
1686 	*laddr = inp->inp_laddr.s_addr;
1687 	*faddr = inp->inp_faddr.s_addr;
1688 	*lp = inp->inp_lport;
1689 	*fp = inp->inp_fport;
1690 }
1691 
1692 struct inpcb *
1693 so_sotoinpcb(struct socket *so)
1694 {
1695 
1696 	return (sotoinpcb(so));
1697 }
1698 
1699 struct tcpcb *
1700 so_sototcpcb(struct socket *so)
1701 {
1702 
1703 	return (sototcpcb(so));
1704 }
1705 
1706 #ifdef DDB
1707 static void
1708 db_print_indent(int indent)
1709 {
1710 	int i;
1711 
1712 	for (i = 0; i < indent; i++)
1713 		db_printf(" ");
1714 }
1715 
1716 static void
1717 db_print_inconninfo(struct in_conninfo *inc, const char *name, int indent)
1718 {
1719 	char faddr_str[48], laddr_str[48];
1720 
1721 	db_print_indent(indent);
1722 	db_printf("%s at %p\n", name, inc);
1723 
1724 	indent += 2;
1725 
1726 #ifdef INET6
1727 	if (inc->inc_flags & INC_ISIPV6) {
1728 		/* IPv6. */
1729 		ip6_sprintf(laddr_str, &inc->inc6_laddr);
1730 		ip6_sprintf(faddr_str, &inc->inc6_faddr);
1731 	} else {
1732 #endif
1733 		/* IPv4. */
1734 		inet_ntoa_r(inc->inc_laddr, laddr_str);
1735 		inet_ntoa_r(inc->inc_faddr, faddr_str);
1736 #ifdef INET6
1737 	}
1738 #endif
1739 	db_print_indent(indent);
1740 	db_printf("inc_laddr %s   inc_lport %u\n", laddr_str,
1741 	    ntohs(inc->inc_lport));
1742 	db_print_indent(indent);
1743 	db_printf("inc_faddr %s   inc_fport %u\n", faddr_str,
1744 	    ntohs(inc->inc_fport));
1745 }
1746 
1747 static void
1748 db_print_inpflags(int inp_flags)
1749 {
1750 	int comma;
1751 
1752 	comma = 0;
1753 	if (inp_flags & INP_RECVOPTS) {
1754 		db_printf("%sINP_RECVOPTS", comma ? ", " : "");
1755 		comma = 1;
1756 	}
1757 	if (inp_flags & INP_RECVRETOPTS) {
1758 		db_printf("%sINP_RECVRETOPTS", comma ? ", " : "");
1759 		comma = 1;
1760 	}
1761 	if (inp_flags & INP_RECVDSTADDR) {
1762 		db_printf("%sINP_RECVDSTADDR", comma ? ", " : "");
1763 		comma = 1;
1764 	}
1765 	if (inp_flags & INP_HDRINCL) {
1766 		db_printf("%sINP_HDRINCL", comma ? ", " : "");
1767 		comma = 1;
1768 	}
1769 	if (inp_flags & INP_HIGHPORT) {
1770 		db_printf("%sINP_HIGHPORT", comma ? ", " : "");
1771 		comma = 1;
1772 	}
1773 	if (inp_flags & INP_LOWPORT) {
1774 		db_printf("%sINP_LOWPORT", comma ? ", " : "");
1775 		comma = 1;
1776 	}
1777 	if (inp_flags & INP_ANONPORT) {
1778 		db_printf("%sINP_ANONPORT", comma ? ", " : "");
1779 		comma = 1;
1780 	}
1781 	if (inp_flags & INP_RECVIF) {
1782 		db_printf("%sINP_RECVIF", comma ? ", " : "");
1783 		comma = 1;
1784 	}
1785 	if (inp_flags & INP_MTUDISC) {
1786 		db_printf("%sINP_MTUDISC", comma ? ", " : "");
1787 		comma = 1;
1788 	}
1789 	if (inp_flags & INP_FAITH) {
1790 		db_printf("%sINP_FAITH", comma ? ", " : "");
1791 		comma = 1;
1792 	}
1793 	if (inp_flags & INP_RECVTTL) {
1794 		db_printf("%sINP_RECVTTL", comma ? ", " : "");
1795 		comma = 1;
1796 	}
1797 	if (inp_flags & INP_DONTFRAG) {
1798 		db_printf("%sINP_DONTFRAG", comma ? ", " : "");
1799 		comma = 1;
1800 	}
1801 	if (inp_flags & IN6P_IPV6_V6ONLY) {
1802 		db_printf("%sIN6P_IPV6_V6ONLY", comma ? ", " : "");
1803 		comma = 1;
1804 	}
1805 	if (inp_flags & IN6P_PKTINFO) {
1806 		db_printf("%sIN6P_PKTINFO", comma ? ", " : "");
1807 		comma = 1;
1808 	}
1809 	if (inp_flags & IN6P_HOPLIMIT) {
1810 		db_printf("%sIN6P_HOPLIMIT", comma ? ", " : "");
1811 		comma = 1;
1812 	}
1813 	if (inp_flags & IN6P_HOPOPTS) {
1814 		db_printf("%sIN6P_HOPOPTS", comma ? ", " : "");
1815 		comma = 1;
1816 	}
1817 	if (inp_flags & IN6P_DSTOPTS) {
1818 		db_printf("%sIN6P_DSTOPTS", comma ? ", " : "");
1819 		comma = 1;
1820 	}
1821 	if (inp_flags & IN6P_RTHDR) {
1822 		db_printf("%sIN6P_RTHDR", comma ? ", " : "");
1823 		comma = 1;
1824 	}
1825 	if (inp_flags & IN6P_RTHDRDSTOPTS) {
1826 		db_printf("%sIN6P_RTHDRDSTOPTS", comma ? ", " : "");
1827 		comma = 1;
1828 	}
1829 	if (inp_flags & IN6P_TCLASS) {
1830 		db_printf("%sIN6P_TCLASS", comma ? ", " : "");
1831 		comma = 1;
1832 	}
1833 	if (inp_flags & IN6P_AUTOFLOWLABEL) {
1834 		db_printf("%sIN6P_AUTOFLOWLABEL", comma ? ", " : "");
1835 		comma = 1;
1836 	}
1837 	if (inp_flags & INP_TIMEWAIT) {
1838 		db_printf("%sINP_TIMEWAIT", comma ? ", " : "");
1839 		comma  = 1;
1840 	}
1841 	if (inp_flags & INP_ONESBCAST) {
1842 		db_printf("%sINP_ONESBCAST", comma ? ", " : "");
1843 		comma  = 1;
1844 	}
1845 	if (inp_flags & INP_DROPPED) {
1846 		db_printf("%sINP_DROPPED", comma ? ", " : "");
1847 		comma  = 1;
1848 	}
1849 	if (inp_flags & INP_SOCKREF) {
1850 		db_printf("%sINP_SOCKREF", comma ? ", " : "");
1851 		comma  = 1;
1852 	}
1853 	if (inp_flags & IN6P_RFC2292) {
1854 		db_printf("%sIN6P_RFC2292", comma ? ", " : "");
1855 		comma = 1;
1856 	}
1857 	if (inp_flags & IN6P_MTU) {
1858 		db_printf("IN6P_MTU%s", comma ? ", " : "");
1859 		comma = 1;
1860 	}
1861 }
1862 
1863 static void
1864 db_print_inpvflag(u_char inp_vflag)
1865 {
1866 	int comma;
1867 
1868 	comma = 0;
1869 	if (inp_vflag & INP_IPV4) {
1870 		db_printf("%sINP_IPV4", comma ? ", " : "");
1871 		comma  = 1;
1872 	}
1873 	if (inp_vflag & INP_IPV6) {
1874 		db_printf("%sINP_IPV6", comma ? ", " : "");
1875 		comma  = 1;
1876 	}
1877 	if (inp_vflag & INP_IPV6PROTO) {
1878 		db_printf("%sINP_IPV6PROTO", comma ? ", " : "");
1879 		comma  = 1;
1880 	}
1881 }
1882 
1883 static void
1884 db_print_inpcb(struct inpcb *inp, const char *name, int indent)
1885 {
1886 
1887 	db_print_indent(indent);
1888 	db_printf("%s at %p\n", name, inp);
1889 
1890 	indent += 2;
1891 
1892 	db_print_indent(indent);
1893 	db_printf("inp_flow: 0x%x\n", inp->inp_flow);
1894 
1895 	db_print_inconninfo(&inp->inp_inc, "inp_conninfo", indent);
1896 
1897 	db_print_indent(indent);
1898 	db_printf("inp_ppcb: %p   inp_pcbinfo: %p   inp_socket: %p\n",
1899 	    inp->inp_ppcb, inp->inp_pcbinfo, inp->inp_socket);
1900 
1901 	db_print_indent(indent);
1902 	db_printf("inp_label: %p   inp_flags: 0x%x (",
1903 	   inp->inp_label, inp->inp_flags);
1904 	db_print_inpflags(inp->inp_flags);
1905 	db_printf(")\n");
1906 
1907 	db_print_indent(indent);
1908 	db_printf("inp_sp: %p   inp_vflag: 0x%x (", inp->inp_sp,
1909 	    inp->inp_vflag);
1910 	db_print_inpvflag(inp->inp_vflag);
1911 	db_printf(")\n");
1912 
1913 	db_print_indent(indent);
1914 	db_printf("inp_ip_ttl: %d   inp_ip_p: %d   inp_ip_minttl: %d\n",
1915 	    inp->inp_ip_ttl, inp->inp_ip_p, inp->inp_ip_minttl);
1916 
1917 	db_print_indent(indent);
1918 #ifdef INET6
1919 	if (inp->inp_vflag & INP_IPV6) {
1920 		db_printf("in6p_options: %p   in6p_outputopts: %p   "
1921 		    "in6p_moptions: %p\n", inp->in6p_options,
1922 		    inp->in6p_outputopts, inp->in6p_moptions);
1923 		db_printf("in6p_icmp6filt: %p   in6p_cksum %d   "
1924 		    "in6p_hops %u\n", inp->in6p_icmp6filt, inp->in6p_cksum,
1925 		    inp->in6p_hops);
1926 	} else
1927 #endif
1928 	{
1929 		db_printf("inp_ip_tos: %d   inp_ip_options: %p   "
1930 		    "inp_ip_moptions: %p\n", inp->inp_ip_tos,
1931 		    inp->inp_options, inp->inp_moptions);
1932 	}
1933 
1934 	db_print_indent(indent);
1935 	db_printf("inp_phd: %p   inp_gencnt: %ju\n", inp->inp_phd,
1936 	    (uintmax_t)inp->inp_gencnt);
1937 }
1938 
1939 DB_SHOW_COMMAND(inpcb, db_show_inpcb)
1940 {
1941 	struct inpcb *inp;
1942 
1943 	if (!have_addr) {
1944 		db_printf("usage: show inpcb <addr>\n");
1945 		return;
1946 	}
1947 	inp = (struct inpcb *)addr;
1948 
1949 	db_print_inpcb(inp, "inpcb", 0);
1950 }
1951 #endif
1952