xref: /freebsd/sys/netinet/in_pcb.c (revision 0572ccaa4543b0abef8ef81e384c1d04de9f3da1)
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  * Copyright (c) 2010-2011 Juniper Networks, Inc.
6  * All rights reserved.
7  *
8  * Portions of this software were developed by Robert N. M. Watson under
9  * contract to Juniper Networks, Inc.
10  *
11  * Redistribution and use in source and binary forms, with or without
12  * modification, are permitted provided that the following conditions
13  * are met:
14  * 1. Redistributions of source code must retain the above copyright
15  *    notice, this list of conditions and the following disclaimer.
16  * 2. Redistributions in binary form must reproduce the above copyright
17  *    notice, this list of conditions and the following disclaimer in the
18  *    documentation and/or other materials provided with the distribution.
19  * 4. Neither the name of the University nor the names of its contributors
20  *    may be used to endorse or promote products derived from this software
21  *    without specific prior written permission.
22  *
23  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
24  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
25  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
26  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
28  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
29  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
30  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
31  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
32  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33  * SUCH DAMAGE.
34  *
35  *	@(#)in_pcb.c	8.4 (Berkeley) 5/24/95
36  */
37 
38 #include <sys/cdefs.h>
39 __FBSDID("$FreeBSD$");
40 
41 #include "opt_ddb.h"
42 #include "opt_ipsec.h"
43 #include "opt_inet.h"
44 #include "opt_inet6.h"
45 #include "opt_pcbgroup.h"
46 #include "opt_rss.h"
47 
48 #include <sys/param.h>
49 #include <sys/systm.h>
50 #include <sys/malloc.h>
51 #include <sys/mbuf.h>
52 #include <sys/callout.h>
53 #include <sys/domain.h>
54 #include <sys/protosw.h>
55 #include <sys/socket.h>
56 #include <sys/socketvar.h>
57 #include <sys/priv.h>
58 #include <sys/proc.h>
59 #include <sys/refcount.h>
60 #include <sys/jail.h>
61 #include <sys/kernel.h>
62 #include <sys/sysctl.h>
63 
64 #ifdef DDB
65 #include <ddb/ddb.h>
66 #endif
67 
68 #include <vm/uma.h>
69 
70 #include <net/if.h>
71 #include <net/if_var.h>
72 #include <net/if_types.h>
73 #include <net/route.h>
74 #include <net/vnet.h>
75 
76 #if defined(INET) || defined(INET6)
77 #include <netinet/in.h>
78 #include <netinet/in_pcb.h>
79 #include <netinet/in_rss.h>
80 #include <netinet/ip_var.h>
81 #include <netinet/tcp_var.h>
82 #include <netinet/udp.h>
83 #include <netinet/udp_var.h>
84 #endif
85 #ifdef INET
86 #include <netinet/in_var.h>
87 #endif
88 #ifdef INET6
89 #include <netinet/ip6.h>
90 #include <netinet6/in6_pcb.h>
91 #include <netinet6/in6_var.h>
92 #include <netinet6/ip6_var.h>
93 #endif /* INET6 */
94 
95 
96 #ifdef IPSEC
97 #include <netipsec/ipsec.h>
98 #include <netipsec/key.h>
99 #endif /* IPSEC */
100 
101 #include <security/mac/mac_framework.h>
102 
103 static struct callout	ipport_tick_callout;
104 
105 /*
106  * These configure the range of local port addresses assigned to
107  * "unspecified" outgoing connections/packets/whatever.
108  */
109 VNET_DEFINE(int, ipport_lowfirstauto) = IPPORT_RESERVED - 1;	/* 1023 */
110 VNET_DEFINE(int, ipport_lowlastauto) = IPPORT_RESERVEDSTART;	/* 600 */
111 VNET_DEFINE(int, ipport_firstauto) = IPPORT_EPHEMERALFIRST;	/* 10000 */
112 VNET_DEFINE(int, ipport_lastauto) = IPPORT_EPHEMERALLAST;	/* 65535 */
113 VNET_DEFINE(int, ipport_hifirstauto) = IPPORT_HIFIRSTAUTO;	/* 49152 */
114 VNET_DEFINE(int, ipport_hilastauto) = IPPORT_HILASTAUTO;	/* 65535 */
115 
116 /*
117  * Reserved ports accessible only to root. There are significant
118  * security considerations that must be accounted for when changing these,
119  * but the security benefits can be great. Please be careful.
120  */
121 VNET_DEFINE(int, ipport_reservedhigh) = IPPORT_RESERVED - 1;	/* 1023 */
122 VNET_DEFINE(int, ipport_reservedlow);
123 
124 /* Variables dealing with random ephemeral port allocation. */
125 VNET_DEFINE(int, ipport_randomized) = 1;	/* user controlled via sysctl */
126 VNET_DEFINE(int, ipport_randomcps) = 10;	/* user controlled via sysctl */
127 VNET_DEFINE(int, ipport_randomtime) = 45;	/* user controlled via sysctl */
128 VNET_DEFINE(int, ipport_stoprandom);		/* toggled by ipport_tick */
129 VNET_DEFINE(int, ipport_tcpallocs);
130 static VNET_DEFINE(int, ipport_tcplastcount);
131 
132 #define	V_ipport_tcplastcount		VNET(ipport_tcplastcount)
133 
134 static void	in_pcbremlists(struct inpcb *inp);
135 #ifdef INET
136 static struct inpcb	*in_pcblookup_hash_locked(struct inpcbinfo *pcbinfo,
137 			    struct in_addr faddr, u_int fport_arg,
138 			    struct in_addr laddr, u_int lport_arg,
139 			    int lookupflags, struct ifnet *ifp);
140 
141 #define RANGECHK(var, min, max) \
142 	if ((var) < (min)) { (var) = (min); } \
143 	else if ((var) > (max)) { (var) = (max); }
144 
145 static int
146 sysctl_net_ipport_check(SYSCTL_HANDLER_ARGS)
147 {
148 	int error;
149 
150 	error = sysctl_handle_int(oidp, arg1, arg2, req);
151 	if (error == 0) {
152 		RANGECHK(V_ipport_lowfirstauto, 1, IPPORT_RESERVED - 1);
153 		RANGECHK(V_ipport_lowlastauto, 1, IPPORT_RESERVED - 1);
154 		RANGECHK(V_ipport_firstauto, IPPORT_RESERVED, IPPORT_MAX);
155 		RANGECHK(V_ipport_lastauto, IPPORT_RESERVED, IPPORT_MAX);
156 		RANGECHK(V_ipport_hifirstauto, IPPORT_RESERVED, IPPORT_MAX);
157 		RANGECHK(V_ipport_hilastauto, IPPORT_RESERVED, IPPORT_MAX);
158 	}
159 	return (error);
160 }
161 
162 #undef RANGECHK
163 
164 static SYSCTL_NODE(_net_inet_ip, IPPROTO_IP, portrange, CTLFLAG_RW, 0,
165     "IP Ports");
166 
167 SYSCTL_VNET_PROC(_net_inet_ip_portrange, OID_AUTO, lowfirst,
168 	CTLTYPE_INT|CTLFLAG_RW, &VNET_NAME(ipport_lowfirstauto), 0,
169 	&sysctl_net_ipport_check, "I", "");
170 SYSCTL_VNET_PROC(_net_inet_ip_portrange, OID_AUTO, lowlast,
171 	CTLTYPE_INT|CTLFLAG_RW, &VNET_NAME(ipport_lowlastauto), 0,
172 	&sysctl_net_ipport_check, "I", "");
173 SYSCTL_VNET_PROC(_net_inet_ip_portrange, OID_AUTO, first,
174 	CTLTYPE_INT|CTLFLAG_RW, &VNET_NAME(ipport_firstauto), 0,
175 	&sysctl_net_ipport_check, "I", "");
176 SYSCTL_VNET_PROC(_net_inet_ip_portrange, OID_AUTO, last,
177 	CTLTYPE_INT|CTLFLAG_RW, &VNET_NAME(ipport_lastauto), 0,
178 	&sysctl_net_ipport_check, "I", "");
179 SYSCTL_VNET_PROC(_net_inet_ip_portrange, OID_AUTO, hifirst,
180 	CTLTYPE_INT|CTLFLAG_RW, &VNET_NAME(ipport_hifirstauto), 0,
181 	&sysctl_net_ipport_check, "I", "");
182 SYSCTL_VNET_PROC(_net_inet_ip_portrange, OID_AUTO, hilast,
183 	CTLTYPE_INT|CTLFLAG_RW, &VNET_NAME(ipport_hilastauto), 0,
184 	&sysctl_net_ipport_check, "I", "");
185 SYSCTL_VNET_INT(_net_inet_ip_portrange, OID_AUTO, reservedhigh,
186 	CTLFLAG_RW|CTLFLAG_SECURE, &VNET_NAME(ipport_reservedhigh), 0, "");
187 SYSCTL_VNET_INT(_net_inet_ip_portrange, OID_AUTO, reservedlow,
188 	CTLFLAG_RW|CTLFLAG_SECURE, &VNET_NAME(ipport_reservedlow), 0, "");
189 SYSCTL_VNET_INT(_net_inet_ip_portrange, OID_AUTO, randomized, CTLFLAG_RW,
190 	&VNET_NAME(ipport_randomized), 0, "Enable random port allocation");
191 SYSCTL_VNET_INT(_net_inet_ip_portrange, OID_AUTO, randomcps, CTLFLAG_RW,
192 	&VNET_NAME(ipport_randomcps), 0, "Maximum number of random port "
193 	"allocations before switching to a sequental one");
194 SYSCTL_VNET_INT(_net_inet_ip_portrange, OID_AUTO, randomtime, CTLFLAG_RW,
195 	&VNET_NAME(ipport_randomtime), 0,
196 	"Minimum time to keep sequental port "
197 	"allocation before switching to a random one");
198 #endif /* INET */
199 
200 /*
201  * in_pcb.c: manage the Protocol Control Blocks.
202  *
203  * NOTE: It is assumed that most of these functions will be called with
204  * the pcbinfo lock held, and often, the inpcb lock held, as these utility
205  * functions often modify hash chains or addresses in pcbs.
206  */
207 
208 /*
209  * Initialize an inpcbinfo -- we should be able to reduce the number of
210  * arguments in time.
211  */
212 void
213 in_pcbinfo_init(struct inpcbinfo *pcbinfo, const char *name,
214     struct inpcbhead *listhead, int hash_nelements, int porthash_nelements,
215     char *inpcbzone_name, uma_init inpcbzone_init, uma_fini inpcbzone_fini,
216     uint32_t inpcbzone_flags, u_int hashfields)
217 {
218 
219 	INP_INFO_LOCK_INIT(pcbinfo, name);
220 	INP_HASH_LOCK_INIT(pcbinfo, "pcbinfohash");	/* XXXRW: argument? */
221 #ifdef VIMAGE
222 	pcbinfo->ipi_vnet = curvnet;
223 #endif
224 	pcbinfo->ipi_listhead = listhead;
225 	LIST_INIT(pcbinfo->ipi_listhead);
226 	pcbinfo->ipi_count = 0;
227 	pcbinfo->ipi_hashbase = hashinit(hash_nelements, M_PCB,
228 	    &pcbinfo->ipi_hashmask);
229 	pcbinfo->ipi_porthashbase = hashinit(porthash_nelements, M_PCB,
230 	    &pcbinfo->ipi_porthashmask);
231 #ifdef PCBGROUP
232 	in_pcbgroup_init(pcbinfo, hashfields, hash_nelements);
233 #endif
234 	pcbinfo->ipi_zone = uma_zcreate(inpcbzone_name, sizeof(struct inpcb),
235 	    NULL, NULL, inpcbzone_init, inpcbzone_fini, UMA_ALIGN_PTR,
236 	    inpcbzone_flags);
237 	uma_zone_set_max(pcbinfo->ipi_zone, maxsockets);
238 	uma_zone_set_warning(pcbinfo->ipi_zone,
239 	    "kern.ipc.maxsockets limit reached");
240 }
241 
242 /*
243  * Destroy an inpcbinfo.
244  */
245 void
246 in_pcbinfo_destroy(struct inpcbinfo *pcbinfo)
247 {
248 
249 	KASSERT(pcbinfo->ipi_count == 0,
250 	    ("%s: ipi_count = %u", __func__, pcbinfo->ipi_count));
251 
252 	hashdestroy(pcbinfo->ipi_hashbase, M_PCB, pcbinfo->ipi_hashmask);
253 	hashdestroy(pcbinfo->ipi_porthashbase, M_PCB,
254 	    pcbinfo->ipi_porthashmask);
255 #ifdef PCBGROUP
256 	in_pcbgroup_destroy(pcbinfo);
257 #endif
258 	uma_zdestroy(pcbinfo->ipi_zone);
259 	INP_HASH_LOCK_DESTROY(pcbinfo);
260 	INP_INFO_LOCK_DESTROY(pcbinfo);
261 }
262 
263 /*
264  * Allocate a PCB and associate it with the socket.
265  * On success return with the PCB locked.
266  */
267 int
268 in_pcballoc(struct socket *so, struct inpcbinfo *pcbinfo)
269 {
270 	struct inpcb *inp;
271 	int error;
272 
273 	INP_INFO_WLOCK_ASSERT(pcbinfo);
274 	error = 0;
275 	inp = uma_zalloc(pcbinfo->ipi_zone, M_NOWAIT);
276 	if (inp == NULL)
277 		return (ENOBUFS);
278 	bzero(inp, inp_zero_size);
279 	inp->inp_pcbinfo = pcbinfo;
280 	inp->inp_socket = so;
281 	inp->inp_cred = crhold(so->so_cred);
282 	inp->inp_inc.inc_fibnum = so->so_fibnum;
283 #ifdef MAC
284 	error = mac_inpcb_init(inp, M_NOWAIT);
285 	if (error != 0)
286 		goto out;
287 	mac_inpcb_create(so, inp);
288 #endif
289 #ifdef IPSEC
290 	error = ipsec_init_policy(so, &inp->inp_sp);
291 	if (error != 0) {
292 #ifdef MAC
293 		mac_inpcb_destroy(inp);
294 #endif
295 		goto out;
296 	}
297 #endif /*IPSEC*/
298 #ifdef INET6
299 	if (INP_SOCKAF(so) == AF_INET6) {
300 		inp->inp_vflag |= INP_IPV6PROTO;
301 		if (V_ip6_v6only)
302 			inp->inp_flags |= IN6P_IPV6_V6ONLY;
303 	}
304 #endif
305 	LIST_INSERT_HEAD(pcbinfo->ipi_listhead, inp, inp_list);
306 	pcbinfo->ipi_count++;
307 	so->so_pcb = (caddr_t)inp;
308 #ifdef INET6
309 	if (V_ip6_auto_flowlabel)
310 		inp->inp_flags |= IN6P_AUTOFLOWLABEL;
311 #endif
312 	INP_WLOCK(inp);
313 	inp->inp_gencnt = ++pcbinfo->ipi_gencnt;
314 	refcount_init(&inp->inp_refcount, 1);	/* Reference from inpcbinfo */
315 #if defined(IPSEC) || defined(MAC)
316 out:
317 	if (error != 0) {
318 		crfree(inp->inp_cred);
319 		uma_zfree(pcbinfo->ipi_zone, inp);
320 	}
321 #endif
322 	return (error);
323 }
324 
325 #ifdef INET
326 int
327 in_pcbbind(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred)
328 {
329 	int anonport, error;
330 
331 	INP_WLOCK_ASSERT(inp);
332 	INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo);
333 
334 	if (inp->inp_lport != 0 || inp->inp_laddr.s_addr != INADDR_ANY)
335 		return (EINVAL);
336 	anonport = nam == NULL || ((struct sockaddr_in *)nam)->sin_port == 0;
337 	error = in_pcbbind_setup(inp, nam, &inp->inp_laddr.s_addr,
338 	    &inp->inp_lport, cred);
339 	if (error)
340 		return (error);
341 	if (in_pcbinshash(inp) != 0) {
342 		inp->inp_laddr.s_addr = INADDR_ANY;
343 		inp->inp_lport = 0;
344 		return (EAGAIN);
345 	}
346 	if (anonport)
347 		inp->inp_flags |= INP_ANONPORT;
348 	return (0);
349 }
350 #endif
351 
352 #if defined(INET) || defined(INET6)
353 int
354 in_pcb_lport(struct inpcb *inp, struct in_addr *laddrp, u_short *lportp,
355     struct ucred *cred, int lookupflags)
356 {
357 	struct inpcbinfo *pcbinfo;
358 	struct inpcb *tmpinp;
359 	unsigned short *lastport;
360 	int count, dorandom, error;
361 	u_short aux, first, last, lport;
362 #ifdef INET
363 	struct in_addr laddr;
364 #endif
365 
366 	pcbinfo = inp->inp_pcbinfo;
367 
368 	/*
369 	 * Because no actual state changes occur here, a global write lock on
370 	 * the pcbinfo isn't required.
371 	 */
372 	INP_LOCK_ASSERT(inp);
373 	INP_HASH_LOCK_ASSERT(pcbinfo);
374 
375 	if (inp->inp_flags & INP_HIGHPORT) {
376 		first = V_ipport_hifirstauto;	/* sysctl */
377 		last  = V_ipport_hilastauto;
378 		lastport = &pcbinfo->ipi_lasthi;
379 	} else if (inp->inp_flags & INP_LOWPORT) {
380 		error = priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT, 0);
381 		if (error)
382 			return (error);
383 		first = V_ipport_lowfirstauto;	/* 1023 */
384 		last  = V_ipport_lowlastauto;	/* 600 */
385 		lastport = &pcbinfo->ipi_lastlow;
386 	} else {
387 		first = V_ipport_firstauto;	/* sysctl */
388 		last  = V_ipport_lastauto;
389 		lastport = &pcbinfo->ipi_lastport;
390 	}
391 	/*
392 	 * For UDP(-Lite), use random port allocation as long as the user
393 	 * allows it.  For TCP (and as of yet unknown) connections,
394 	 * use random port allocation only if the user allows it AND
395 	 * ipport_tick() allows it.
396 	 */
397 	if (V_ipport_randomized &&
398 		(!V_ipport_stoprandom || pcbinfo == &V_udbinfo ||
399 		pcbinfo == &V_ulitecbinfo))
400 		dorandom = 1;
401 	else
402 		dorandom = 0;
403 	/*
404 	 * It makes no sense to do random port allocation if
405 	 * we have the only port available.
406 	 */
407 	if (first == last)
408 		dorandom = 0;
409 	/* Make sure to not include UDP(-Lite) packets in the count. */
410 	if (pcbinfo != &V_udbinfo || pcbinfo != &V_ulitecbinfo)
411 		V_ipport_tcpallocs++;
412 	/*
413 	 * Instead of having two loops further down counting up or down
414 	 * make sure that first is always <= last and go with only one
415 	 * code path implementing all logic.
416 	 */
417 	if (first > last) {
418 		aux = first;
419 		first = last;
420 		last = aux;
421 	}
422 
423 #ifdef INET
424 	/* Make the compiler happy. */
425 	laddr.s_addr = 0;
426 	if ((inp->inp_vflag & (INP_IPV4|INP_IPV6)) == INP_IPV4) {
427 		KASSERT(laddrp != NULL, ("%s: laddrp NULL for v4 inp %p",
428 		    __func__, inp));
429 		laddr = *laddrp;
430 	}
431 #endif
432 	tmpinp = NULL;	/* Make compiler happy. */
433 	lport = *lportp;
434 
435 	if (dorandom)
436 		*lastport = first + (arc4random() % (last - first));
437 
438 	count = last - first;
439 
440 	do {
441 		if (count-- < 0)	/* completely used? */
442 			return (EADDRNOTAVAIL);
443 		++*lastport;
444 		if (*lastport < first || *lastport > last)
445 			*lastport = first;
446 		lport = htons(*lastport);
447 
448 #ifdef INET6
449 		if ((inp->inp_vflag & INP_IPV6) != 0)
450 			tmpinp = in6_pcblookup_local(pcbinfo,
451 			    &inp->in6p_laddr, lport, lookupflags, cred);
452 #endif
453 #if defined(INET) && defined(INET6)
454 		else
455 #endif
456 #ifdef INET
457 			tmpinp = in_pcblookup_local(pcbinfo, laddr,
458 			    lport, lookupflags, cred);
459 #endif
460 	} while (tmpinp != NULL);
461 
462 #ifdef INET
463 	if ((inp->inp_vflag & (INP_IPV4|INP_IPV6)) == INP_IPV4)
464 		laddrp->s_addr = laddr.s_addr;
465 #endif
466 	*lportp = lport;
467 
468 	return (0);
469 }
470 
471 /*
472  * Return cached socket options.
473  */
474 short
475 inp_so_options(const struct inpcb *inp)
476 {
477    short so_options;
478 
479    so_options = 0;
480 
481    if ((inp->inp_flags2 & INP_REUSEPORT) != 0)
482 	   so_options |= SO_REUSEPORT;
483    if ((inp->inp_flags2 & INP_REUSEADDR) != 0)
484 	   so_options |= SO_REUSEADDR;
485    return (so_options);
486 }
487 #endif /* INET || INET6 */
488 
489 #ifdef INET
490 /*
491  * Set up a bind operation on a PCB, performing port allocation
492  * as required, but do not actually modify the PCB. Callers can
493  * either complete the bind by setting inp_laddr/inp_lport and
494  * calling in_pcbinshash(), or they can just use the resulting
495  * port and address to authorise the sending of a once-off packet.
496  *
497  * On error, the values of *laddrp and *lportp are not changed.
498  */
499 int
500 in_pcbbind_setup(struct inpcb *inp, struct sockaddr *nam, in_addr_t *laddrp,
501     u_short *lportp, struct ucred *cred)
502 {
503 	struct socket *so = inp->inp_socket;
504 	struct sockaddr_in *sin;
505 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
506 	struct in_addr laddr;
507 	u_short lport = 0;
508 	int lookupflags = 0, reuseport = (so->so_options & SO_REUSEPORT);
509 	int error;
510 
511 	/*
512 	 * No state changes, so read locks are sufficient here.
513 	 */
514 	INP_LOCK_ASSERT(inp);
515 	INP_HASH_LOCK_ASSERT(pcbinfo);
516 
517 	if (TAILQ_EMPTY(&V_in_ifaddrhead)) /* XXX broken! */
518 		return (EADDRNOTAVAIL);
519 	laddr.s_addr = *laddrp;
520 	if (nam != NULL && laddr.s_addr != INADDR_ANY)
521 		return (EINVAL);
522 	if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0)
523 		lookupflags = INPLOOKUP_WILDCARD;
524 	if (nam == NULL) {
525 		if ((error = prison_local_ip4(cred, &laddr)) != 0)
526 			return (error);
527 	} else {
528 		sin = (struct sockaddr_in *)nam;
529 		if (nam->sa_len != sizeof (*sin))
530 			return (EINVAL);
531 #ifdef notdef
532 		/*
533 		 * We should check the family, but old programs
534 		 * incorrectly fail to initialize it.
535 		 */
536 		if (sin->sin_family != AF_INET)
537 			return (EAFNOSUPPORT);
538 #endif
539 		error = prison_local_ip4(cred, &sin->sin_addr);
540 		if (error)
541 			return (error);
542 		if (sin->sin_port != *lportp) {
543 			/* Don't allow the port to change. */
544 			if (*lportp != 0)
545 				return (EINVAL);
546 			lport = sin->sin_port;
547 		}
548 		/* NB: lport is left as 0 if the port isn't being changed. */
549 		if (IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) {
550 			/*
551 			 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
552 			 * allow complete duplication of binding if
553 			 * SO_REUSEPORT is set, or if SO_REUSEADDR is set
554 			 * and a multicast address is bound on both
555 			 * new and duplicated sockets.
556 			 */
557 			if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) != 0)
558 				reuseport = SO_REUSEADDR|SO_REUSEPORT;
559 		} else if (sin->sin_addr.s_addr != INADDR_ANY) {
560 			sin->sin_port = 0;		/* yech... */
561 			bzero(&sin->sin_zero, sizeof(sin->sin_zero));
562 			/*
563 			 * Is the address a local IP address?
564 			 * If INP_BINDANY is set, then the socket may be bound
565 			 * to any endpoint address, local or not.
566 			 */
567 			if ((inp->inp_flags & INP_BINDANY) == 0 &&
568 			    ifa_ifwithaddr_check((struct sockaddr *)sin) == 0)
569 				return (EADDRNOTAVAIL);
570 		}
571 		laddr = sin->sin_addr;
572 		if (lport) {
573 			struct inpcb *t;
574 			struct tcptw *tw;
575 
576 			/* GROSS */
577 			if (ntohs(lport) <= V_ipport_reservedhigh &&
578 			    ntohs(lport) >= V_ipport_reservedlow &&
579 			    priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT,
580 			    0))
581 				return (EACCES);
582 			if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)) &&
583 			    priv_check_cred(inp->inp_cred,
584 			    PRIV_NETINET_REUSEPORT, 0) != 0) {
585 				t = in_pcblookup_local(pcbinfo, sin->sin_addr,
586 				    lport, INPLOOKUP_WILDCARD, cred);
587 	/*
588 	 * XXX
589 	 * This entire block sorely needs a rewrite.
590 	 */
591 				if (t &&
592 				    ((t->inp_flags & INP_TIMEWAIT) == 0) &&
593 				    (so->so_type != SOCK_STREAM ||
594 				     ntohl(t->inp_faddr.s_addr) == INADDR_ANY) &&
595 				    (ntohl(sin->sin_addr.s_addr) != INADDR_ANY ||
596 				     ntohl(t->inp_laddr.s_addr) != INADDR_ANY ||
597 				     (t->inp_flags2 & INP_REUSEPORT) == 0) &&
598 				    (inp->inp_cred->cr_uid !=
599 				     t->inp_cred->cr_uid))
600 					return (EADDRINUSE);
601 			}
602 			t = in_pcblookup_local(pcbinfo, sin->sin_addr,
603 			    lport, lookupflags, cred);
604 			if (t && (t->inp_flags & INP_TIMEWAIT)) {
605 				/*
606 				 * XXXRW: If an incpb has had its timewait
607 				 * state recycled, we treat the address as
608 				 * being in use (for now).  This is better
609 				 * than a panic, but not desirable.
610 				 */
611 				tw = intotw(t);
612 				if (tw == NULL ||
613 				    (reuseport & tw->tw_so_options) == 0)
614 					return (EADDRINUSE);
615 			} else if (t && (reuseport & inp_so_options(t)) == 0) {
616 #ifdef INET6
617 				if (ntohl(sin->sin_addr.s_addr) !=
618 				    INADDR_ANY ||
619 				    ntohl(t->inp_laddr.s_addr) !=
620 				    INADDR_ANY ||
621 				    (inp->inp_vflag & INP_IPV6PROTO) == 0 ||
622 				    (t->inp_vflag & INP_IPV6PROTO) == 0)
623 #endif
624 				return (EADDRINUSE);
625 			}
626 		}
627 	}
628 	if (*lportp != 0)
629 		lport = *lportp;
630 	if (lport == 0) {
631 		error = in_pcb_lport(inp, &laddr, &lport, cred, lookupflags);
632 		if (error != 0)
633 			return (error);
634 
635 	}
636 	*laddrp = laddr.s_addr;
637 	*lportp = lport;
638 	return (0);
639 }
640 
641 /*
642  * Connect from a socket to a specified address.
643  * Both address and port must be specified in argument sin.
644  * If don't have a local address for this socket yet,
645  * then pick one.
646  */
647 int
648 in_pcbconnect_mbuf(struct inpcb *inp, struct sockaddr *nam,
649     struct ucred *cred, struct mbuf *m)
650 {
651 	u_short lport, fport;
652 	in_addr_t laddr, faddr;
653 	int anonport, error;
654 
655 	INP_WLOCK_ASSERT(inp);
656 	INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo);
657 
658 	lport = inp->inp_lport;
659 	laddr = inp->inp_laddr.s_addr;
660 	anonport = (lport == 0);
661 	error = in_pcbconnect_setup(inp, nam, &laddr, &lport, &faddr, &fport,
662 	    NULL, cred);
663 	if (error)
664 		return (error);
665 
666 	/* Do the initial binding of the local address if required. */
667 	if (inp->inp_laddr.s_addr == INADDR_ANY && inp->inp_lport == 0) {
668 		inp->inp_lport = lport;
669 		inp->inp_laddr.s_addr = laddr;
670 		if (in_pcbinshash(inp) != 0) {
671 			inp->inp_laddr.s_addr = INADDR_ANY;
672 			inp->inp_lport = 0;
673 			return (EAGAIN);
674 		}
675 	}
676 
677 	/* Commit the remaining changes. */
678 	inp->inp_lport = lport;
679 	inp->inp_laddr.s_addr = laddr;
680 	inp->inp_faddr.s_addr = faddr;
681 	inp->inp_fport = fport;
682 	in_pcbrehash_mbuf(inp, m);
683 
684 	if (anonport)
685 		inp->inp_flags |= INP_ANONPORT;
686 	return (0);
687 }
688 
689 int
690 in_pcbconnect(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred)
691 {
692 
693 	return (in_pcbconnect_mbuf(inp, nam, cred, NULL));
694 }
695 
696 /*
697  * Do proper source address selection on an unbound socket in case
698  * of connect. Take jails into account as well.
699  */
700 int
701 in_pcbladdr(struct inpcb *inp, struct in_addr *faddr, struct in_addr *laddr,
702     struct ucred *cred)
703 {
704 	struct ifaddr *ifa;
705 	struct sockaddr *sa;
706 	struct sockaddr_in *sin;
707 	struct route sro;
708 	int error;
709 
710 	KASSERT(laddr != NULL, ("%s: laddr NULL", __func__));
711 
712 	/*
713 	 * Bypass source address selection and use the primary jail IP
714 	 * if requested.
715 	 */
716 	if (cred != NULL && !prison_saddrsel_ip4(cred, laddr))
717 		return (0);
718 
719 	error = 0;
720 	bzero(&sro, sizeof(sro));
721 
722 	sin = (struct sockaddr_in *)&sro.ro_dst;
723 	sin->sin_family = AF_INET;
724 	sin->sin_len = sizeof(struct sockaddr_in);
725 	sin->sin_addr.s_addr = faddr->s_addr;
726 
727 	/*
728 	 * If route is known our src addr is taken from the i/f,
729 	 * else punt.
730 	 *
731 	 * Find out route to destination.
732 	 */
733 	if ((inp->inp_socket->so_options & SO_DONTROUTE) == 0)
734 		in_rtalloc_ign(&sro, 0, inp->inp_inc.inc_fibnum);
735 
736 	/*
737 	 * If we found a route, use the address corresponding to
738 	 * the outgoing interface.
739 	 *
740 	 * Otherwise assume faddr is reachable on a directly connected
741 	 * network and try to find a corresponding interface to take
742 	 * the source address from.
743 	 */
744 	if (sro.ro_rt == NULL || sro.ro_rt->rt_ifp == NULL) {
745 		struct in_ifaddr *ia;
746 		struct ifnet *ifp;
747 
748 		ia = ifatoia(ifa_ifwithdstaddr((struct sockaddr *)sin,
749 		    RT_DEFAULT_FIB));
750 		if (ia == NULL)
751 			ia = ifatoia(ifa_ifwithnet((struct sockaddr *)sin, 0,
752 			    RT_DEFAULT_FIB));
753 		if (ia == NULL) {
754 			error = ENETUNREACH;
755 			goto done;
756 		}
757 
758 		if (cred == NULL || !prison_flag(cred, PR_IP4)) {
759 			laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
760 			ifa_free(&ia->ia_ifa);
761 			goto done;
762 		}
763 
764 		ifp = ia->ia_ifp;
765 		ifa_free(&ia->ia_ifa);
766 		ia = NULL;
767 		IF_ADDR_RLOCK(ifp);
768 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
769 
770 			sa = ifa->ifa_addr;
771 			if (sa->sa_family != AF_INET)
772 				continue;
773 			sin = (struct sockaddr_in *)sa;
774 			if (prison_check_ip4(cred, &sin->sin_addr) == 0) {
775 				ia = (struct in_ifaddr *)ifa;
776 				break;
777 			}
778 		}
779 		if (ia != NULL) {
780 			laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
781 			IF_ADDR_RUNLOCK(ifp);
782 			goto done;
783 		}
784 		IF_ADDR_RUNLOCK(ifp);
785 
786 		/* 3. As a last resort return the 'default' jail address. */
787 		error = prison_get_ip4(cred, laddr);
788 		goto done;
789 	}
790 
791 	/*
792 	 * If the outgoing interface on the route found is not
793 	 * a loopback interface, use the address from that interface.
794 	 * In case of jails do those three steps:
795 	 * 1. check if the interface address belongs to the jail. If so use it.
796 	 * 2. check if we have any address on the outgoing interface
797 	 *    belonging to this jail. If so use it.
798 	 * 3. as a last resort return the 'default' jail address.
799 	 */
800 	if ((sro.ro_rt->rt_ifp->if_flags & IFF_LOOPBACK) == 0) {
801 		struct in_ifaddr *ia;
802 		struct ifnet *ifp;
803 
804 		/* If not jailed, use the default returned. */
805 		if (cred == NULL || !prison_flag(cred, PR_IP4)) {
806 			ia = (struct in_ifaddr *)sro.ro_rt->rt_ifa;
807 			laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
808 			goto done;
809 		}
810 
811 		/* Jailed. */
812 		/* 1. Check if the iface address belongs to the jail. */
813 		sin = (struct sockaddr_in *)sro.ro_rt->rt_ifa->ifa_addr;
814 		if (prison_check_ip4(cred, &sin->sin_addr) == 0) {
815 			ia = (struct in_ifaddr *)sro.ro_rt->rt_ifa;
816 			laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
817 			goto done;
818 		}
819 
820 		/*
821 		 * 2. Check if we have any address on the outgoing interface
822 		 *    belonging to this jail.
823 		 */
824 		ia = NULL;
825 		ifp = sro.ro_rt->rt_ifp;
826 		IF_ADDR_RLOCK(ifp);
827 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
828 			sa = ifa->ifa_addr;
829 			if (sa->sa_family != AF_INET)
830 				continue;
831 			sin = (struct sockaddr_in *)sa;
832 			if (prison_check_ip4(cred, &sin->sin_addr) == 0) {
833 				ia = (struct in_ifaddr *)ifa;
834 				break;
835 			}
836 		}
837 		if (ia != NULL) {
838 			laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
839 			IF_ADDR_RUNLOCK(ifp);
840 			goto done;
841 		}
842 		IF_ADDR_RUNLOCK(ifp);
843 
844 		/* 3. As a last resort return the 'default' jail address. */
845 		error = prison_get_ip4(cred, laddr);
846 		goto done;
847 	}
848 
849 	/*
850 	 * The outgoing interface is marked with 'loopback net', so a route
851 	 * to ourselves is here.
852 	 * Try to find the interface of the destination address and then
853 	 * take the address from there. That interface is not necessarily
854 	 * a loopback interface.
855 	 * In case of jails, check that it is an address of the jail
856 	 * and if we cannot find, fall back to the 'default' jail address.
857 	 */
858 	if ((sro.ro_rt->rt_ifp->if_flags & IFF_LOOPBACK) != 0) {
859 		struct sockaddr_in sain;
860 		struct in_ifaddr *ia;
861 
862 		bzero(&sain, sizeof(struct sockaddr_in));
863 		sain.sin_family = AF_INET;
864 		sain.sin_len = sizeof(struct sockaddr_in);
865 		sain.sin_addr.s_addr = faddr->s_addr;
866 
867 		ia = ifatoia(ifa_ifwithdstaddr(sintosa(&sain), RT_DEFAULT_FIB));
868 		if (ia == NULL)
869 			ia = ifatoia(ifa_ifwithnet(sintosa(&sain), 0,
870 			    RT_DEFAULT_FIB));
871 		if (ia == NULL)
872 			ia = ifatoia(ifa_ifwithaddr(sintosa(&sain)));
873 
874 		if (cred == NULL || !prison_flag(cred, PR_IP4)) {
875 			if (ia == NULL) {
876 				error = ENETUNREACH;
877 				goto done;
878 			}
879 			laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
880 			ifa_free(&ia->ia_ifa);
881 			goto done;
882 		}
883 
884 		/* Jailed. */
885 		if (ia != NULL) {
886 			struct ifnet *ifp;
887 
888 			ifp = ia->ia_ifp;
889 			ifa_free(&ia->ia_ifa);
890 			ia = NULL;
891 			IF_ADDR_RLOCK(ifp);
892 			TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
893 
894 				sa = ifa->ifa_addr;
895 				if (sa->sa_family != AF_INET)
896 					continue;
897 				sin = (struct sockaddr_in *)sa;
898 				if (prison_check_ip4(cred,
899 				    &sin->sin_addr) == 0) {
900 					ia = (struct in_ifaddr *)ifa;
901 					break;
902 				}
903 			}
904 			if (ia != NULL) {
905 				laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
906 				IF_ADDR_RUNLOCK(ifp);
907 				goto done;
908 			}
909 			IF_ADDR_RUNLOCK(ifp);
910 		}
911 
912 		/* 3. As a last resort return the 'default' jail address. */
913 		error = prison_get_ip4(cred, laddr);
914 		goto done;
915 	}
916 
917 done:
918 	if (sro.ro_rt != NULL)
919 		RTFREE(sro.ro_rt);
920 	return (error);
921 }
922 
923 /*
924  * Set up for a connect from a socket to the specified address.
925  * On entry, *laddrp and *lportp should contain the current local
926  * address and port for the PCB; these are updated to the values
927  * that should be placed in inp_laddr and inp_lport to complete
928  * the connect.
929  *
930  * On success, *faddrp and *fportp will be set to the remote address
931  * and port. These are not updated in the error case.
932  *
933  * If the operation fails because the connection already exists,
934  * *oinpp will be set to the PCB of that connection so that the
935  * caller can decide to override it. In all other cases, *oinpp
936  * is set to NULL.
937  */
938 int
939 in_pcbconnect_setup(struct inpcb *inp, struct sockaddr *nam,
940     in_addr_t *laddrp, u_short *lportp, in_addr_t *faddrp, u_short *fportp,
941     struct inpcb **oinpp, struct ucred *cred)
942 {
943 	struct sockaddr_in *sin = (struct sockaddr_in *)nam;
944 	struct in_ifaddr *ia;
945 	struct inpcb *oinp;
946 	struct in_addr laddr, faddr;
947 	u_short lport, fport;
948 	int error;
949 
950 	/*
951 	 * Because a global state change doesn't actually occur here, a read
952 	 * lock is sufficient.
953 	 */
954 	INP_LOCK_ASSERT(inp);
955 	INP_HASH_LOCK_ASSERT(inp->inp_pcbinfo);
956 
957 	if (oinpp != NULL)
958 		*oinpp = NULL;
959 	if (nam->sa_len != sizeof (*sin))
960 		return (EINVAL);
961 	if (sin->sin_family != AF_INET)
962 		return (EAFNOSUPPORT);
963 	if (sin->sin_port == 0)
964 		return (EADDRNOTAVAIL);
965 	laddr.s_addr = *laddrp;
966 	lport = *lportp;
967 	faddr = sin->sin_addr;
968 	fport = sin->sin_port;
969 
970 	if (!TAILQ_EMPTY(&V_in_ifaddrhead)) {
971 		/*
972 		 * If the destination address is INADDR_ANY,
973 		 * use the primary local address.
974 		 * If the supplied address is INADDR_BROADCAST,
975 		 * and the primary interface supports broadcast,
976 		 * choose the broadcast address for that interface.
977 		 */
978 		if (faddr.s_addr == INADDR_ANY) {
979 			IN_IFADDR_RLOCK();
980 			faddr =
981 			    IA_SIN(TAILQ_FIRST(&V_in_ifaddrhead))->sin_addr;
982 			IN_IFADDR_RUNLOCK();
983 			if (cred != NULL &&
984 			    (error = prison_get_ip4(cred, &faddr)) != 0)
985 				return (error);
986 		} else if (faddr.s_addr == (u_long)INADDR_BROADCAST) {
987 			IN_IFADDR_RLOCK();
988 			if (TAILQ_FIRST(&V_in_ifaddrhead)->ia_ifp->if_flags &
989 			    IFF_BROADCAST)
990 				faddr = satosin(&TAILQ_FIRST(
991 				    &V_in_ifaddrhead)->ia_broadaddr)->sin_addr;
992 			IN_IFADDR_RUNLOCK();
993 		}
994 	}
995 	if (laddr.s_addr == INADDR_ANY) {
996 		error = in_pcbladdr(inp, &faddr, &laddr, cred);
997 		/*
998 		 * If the destination address is multicast and an outgoing
999 		 * interface has been set as a multicast option, prefer the
1000 		 * address of that interface as our source address.
1001 		 */
1002 		if (IN_MULTICAST(ntohl(faddr.s_addr)) &&
1003 		    inp->inp_moptions != NULL) {
1004 			struct ip_moptions *imo;
1005 			struct ifnet *ifp;
1006 
1007 			imo = inp->inp_moptions;
1008 			if (imo->imo_multicast_ifp != NULL) {
1009 				ifp = imo->imo_multicast_ifp;
1010 				IN_IFADDR_RLOCK();
1011 				TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
1012 					if ((ia->ia_ifp == ifp) &&
1013 					    (cred == NULL ||
1014 					    prison_check_ip4(cred,
1015 					    &ia->ia_addr.sin_addr) == 0))
1016 						break;
1017 				}
1018 				if (ia == NULL)
1019 					error = EADDRNOTAVAIL;
1020 				else {
1021 					laddr = ia->ia_addr.sin_addr;
1022 					error = 0;
1023 				}
1024 				IN_IFADDR_RUNLOCK();
1025 			}
1026 		}
1027 		if (error)
1028 			return (error);
1029 	}
1030 	oinp = in_pcblookup_hash_locked(inp->inp_pcbinfo, faddr, fport,
1031 	    laddr, lport, 0, NULL);
1032 	if (oinp != NULL) {
1033 		if (oinpp != NULL)
1034 			*oinpp = oinp;
1035 		return (EADDRINUSE);
1036 	}
1037 	if (lport == 0) {
1038 		error = in_pcbbind_setup(inp, NULL, &laddr.s_addr, &lport,
1039 		    cred);
1040 		if (error)
1041 			return (error);
1042 	}
1043 	*laddrp = laddr.s_addr;
1044 	*lportp = lport;
1045 	*faddrp = faddr.s_addr;
1046 	*fportp = fport;
1047 	return (0);
1048 }
1049 
1050 void
1051 in_pcbdisconnect(struct inpcb *inp)
1052 {
1053 
1054 	INP_WLOCK_ASSERT(inp);
1055 	INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo);
1056 
1057 	inp->inp_faddr.s_addr = INADDR_ANY;
1058 	inp->inp_fport = 0;
1059 	in_pcbrehash(inp);
1060 }
1061 #endif /* INET */
1062 
1063 /*
1064  * in_pcbdetach() is responsibe for disassociating a socket from an inpcb.
1065  * For most protocols, this will be invoked immediately prior to calling
1066  * in_pcbfree().  However, with TCP the inpcb may significantly outlive the
1067  * socket, in which case in_pcbfree() is deferred.
1068  */
1069 void
1070 in_pcbdetach(struct inpcb *inp)
1071 {
1072 
1073 	KASSERT(inp->inp_socket != NULL, ("%s: inp_socket == NULL", __func__));
1074 
1075 	inp->inp_socket->so_pcb = NULL;
1076 	inp->inp_socket = NULL;
1077 }
1078 
1079 /*
1080  * in_pcbref() bumps the reference count on an inpcb in order to maintain
1081  * stability of an inpcb pointer despite the inpcb lock being released.  This
1082  * is used in TCP when the inpcbinfo lock needs to be acquired or upgraded,
1083  * but where the inpcb lock may already held, or when acquiring a reference
1084  * via a pcbgroup.
1085  *
1086  * in_pcbref() should be used only to provide brief memory stability, and
1087  * must always be followed by a call to INP_WLOCK() and in_pcbrele() to
1088  * garbage collect the inpcb if it has been in_pcbfree()'d from another
1089  * context.  Until in_pcbrele() has returned that the inpcb is still valid,
1090  * lock and rele are the *only* safe operations that may be performed on the
1091  * inpcb.
1092  *
1093  * While the inpcb will not be freed, releasing the inpcb lock means that the
1094  * connection's state may change, so the caller should be careful to
1095  * revalidate any cached state on reacquiring the lock.  Drop the reference
1096  * using in_pcbrele().
1097  */
1098 void
1099 in_pcbref(struct inpcb *inp)
1100 {
1101 
1102 	KASSERT(inp->inp_refcount > 0, ("%s: refcount 0", __func__));
1103 
1104 	refcount_acquire(&inp->inp_refcount);
1105 }
1106 
1107 /*
1108  * Drop a refcount on an inpcb elevated using in_pcbref(); because a call to
1109  * in_pcbfree() may have been made between in_pcbref() and in_pcbrele(), we
1110  * return a flag indicating whether or not the inpcb remains valid.  If it is
1111  * valid, we return with the inpcb lock held.
1112  *
1113  * Notice that, unlike in_pcbref(), the inpcb lock must be held to drop a
1114  * reference on an inpcb.  Historically more work was done here (actually, in
1115  * in_pcbfree_internal()) but has been moved to in_pcbfree() to avoid the
1116  * need for the pcbinfo lock in in_pcbrele().  Deferring the free is entirely
1117  * about memory stability (and continued use of the write lock).
1118  */
1119 int
1120 in_pcbrele_rlocked(struct inpcb *inp)
1121 {
1122 	struct inpcbinfo *pcbinfo;
1123 
1124 	KASSERT(inp->inp_refcount > 0, ("%s: refcount 0", __func__));
1125 
1126 	INP_RLOCK_ASSERT(inp);
1127 
1128 	if (refcount_release(&inp->inp_refcount) == 0) {
1129 		/*
1130 		 * If the inpcb has been freed, let the caller know, even if
1131 		 * this isn't the last reference.
1132 		 */
1133 		if (inp->inp_flags2 & INP_FREED) {
1134 			INP_RUNLOCK(inp);
1135 			return (1);
1136 		}
1137 		return (0);
1138 	}
1139 
1140 	KASSERT(inp->inp_socket == NULL, ("%s: inp_socket != NULL", __func__));
1141 
1142 	INP_RUNLOCK(inp);
1143 	pcbinfo = inp->inp_pcbinfo;
1144 	uma_zfree(pcbinfo->ipi_zone, inp);
1145 	return (1);
1146 }
1147 
1148 int
1149 in_pcbrele_wlocked(struct inpcb *inp)
1150 {
1151 	struct inpcbinfo *pcbinfo;
1152 
1153 	KASSERT(inp->inp_refcount > 0, ("%s: refcount 0", __func__));
1154 
1155 	INP_WLOCK_ASSERT(inp);
1156 
1157 	if (refcount_release(&inp->inp_refcount) == 0)
1158 		return (0);
1159 
1160 	KASSERT(inp->inp_socket == NULL, ("%s: inp_socket != NULL", __func__));
1161 
1162 	INP_WUNLOCK(inp);
1163 	pcbinfo = inp->inp_pcbinfo;
1164 	uma_zfree(pcbinfo->ipi_zone, inp);
1165 	return (1);
1166 }
1167 
1168 /*
1169  * Temporary wrapper.
1170  */
1171 int
1172 in_pcbrele(struct inpcb *inp)
1173 {
1174 
1175 	return (in_pcbrele_wlocked(inp));
1176 }
1177 
1178 /*
1179  * Unconditionally schedule an inpcb to be freed by decrementing its
1180  * reference count, which should occur only after the inpcb has been detached
1181  * from its socket.  If another thread holds a temporary reference (acquired
1182  * using in_pcbref()) then the free is deferred until that reference is
1183  * released using in_pcbrele(), but the inpcb is still unlocked.  Almost all
1184  * work, including removal from global lists, is done in this context, where
1185  * the pcbinfo lock is held.
1186  */
1187 void
1188 in_pcbfree(struct inpcb *inp)
1189 {
1190 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1191 
1192 	KASSERT(inp->inp_socket == NULL, ("%s: inp_socket != NULL", __func__));
1193 
1194 	INP_INFO_WLOCK_ASSERT(pcbinfo);
1195 	INP_WLOCK_ASSERT(inp);
1196 
1197 	/* XXXRW: Do as much as possible here. */
1198 #ifdef IPSEC
1199 	if (inp->inp_sp != NULL)
1200 		ipsec_delete_pcbpolicy(inp);
1201 #endif
1202 	inp->inp_gencnt = ++pcbinfo->ipi_gencnt;
1203 	in_pcbremlists(inp);
1204 #ifdef INET6
1205 	if (inp->inp_vflag & INP_IPV6PROTO) {
1206 		ip6_freepcbopts(inp->in6p_outputopts);
1207 		if (inp->in6p_moptions != NULL)
1208 			ip6_freemoptions(inp->in6p_moptions);
1209 	}
1210 #endif
1211 	if (inp->inp_options)
1212 		(void)m_free(inp->inp_options);
1213 #ifdef INET
1214 	if (inp->inp_moptions != NULL)
1215 		inp_freemoptions(inp->inp_moptions);
1216 #endif
1217 	inp->inp_vflag = 0;
1218 	inp->inp_flags2 |= INP_FREED;
1219 	crfree(inp->inp_cred);
1220 #ifdef MAC
1221 	mac_inpcb_destroy(inp);
1222 #endif
1223 	if (!in_pcbrele_wlocked(inp))
1224 		INP_WUNLOCK(inp);
1225 }
1226 
1227 /*
1228  * in_pcbdrop() removes an inpcb from hashed lists, releasing its address and
1229  * port reservation, and preventing it from being returned by inpcb lookups.
1230  *
1231  * It is used by TCP to mark an inpcb as unused and avoid future packet
1232  * delivery or event notification when a socket remains open but TCP has
1233  * closed.  This might occur as a result of a shutdown()-initiated TCP close
1234  * or a RST on the wire, and allows the port binding to be reused while still
1235  * maintaining the invariant that so_pcb always points to a valid inpcb until
1236  * in_pcbdetach().
1237  *
1238  * XXXRW: Possibly in_pcbdrop() should also prevent future notifications by
1239  * in_pcbnotifyall() and in_pcbpurgeif0()?
1240  */
1241 void
1242 in_pcbdrop(struct inpcb *inp)
1243 {
1244 
1245 	INP_WLOCK_ASSERT(inp);
1246 
1247 	/*
1248 	 * XXXRW: Possibly we should protect the setting of INP_DROPPED with
1249 	 * the hash lock...?
1250 	 */
1251 	inp->inp_flags |= INP_DROPPED;
1252 	if (inp->inp_flags & INP_INHASHLIST) {
1253 		struct inpcbport *phd = inp->inp_phd;
1254 
1255 		INP_HASH_WLOCK(inp->inp_pcbinfo);
1256 		LIST_REMOVE(inp, inp_hash);
1257 		LIST_REMOVE(inp, inp_portlist);
1258 		if (LIST_FIRST(&phd->phd_pcblist) == NULL) {
1259 			LIST_REMOVE(phd, phd_hash);
1260 			free(phd, M_PCB);
1261 		}
1262 		INP_HASH_WUNLOCK(inp->inp_pcbinfo);
1263 		inp->inp_flags &= ~INP_INHASHLIST;
1264 #ifdef PCBGROUP
1265 		in_pcbgroup_remove(inp);
1266 #endif
1267 	}
1268 }
1269 
1270 #ifdef INET
1271 /*
1272  * Common routines to return the socket addresses associated with inpcbs.
1273  */
1274 struct sockaddr *
1275 in_sockaddr(in_port_t port, struct in_addr *addr_p)
1276 {
1277 	struct sockaddr_in *sin;
1278 
1279 	sin = malloc(sizeof *sin, M_SONAME,
1280 		M_WAITOK | M_ZERO);
1281 	sin->sin_family = AF_INET;
1282 	sin->sin_len = sizeof(*sin);
1283 	sin->sin_addr = *addr_p;
1284 	sin->sin_port = port;
1285 
1286 	return (struct sockaddr *)sin;
1287 }
1288 
1289 int
1290 in_getsockaddr(struct socket *so, struct sockaddr **nam)
1291 {
1292 	struct inpcb *inp;
1293 	struct in_addr addr;
1294 	in_port_t port;
1295 
1296 	inp = sotoinpcb(so);
1297 	KASSERT(inp != NULL, ("in_getsockaddr: inp == NULL"));
1298 
1299 	INP_RLOCK(inp);
1300 	port = inp->inp_lport;
1301 	addr = inp->inp_laddr;
1302 	INP_RUNLOCK(inp);
1303 
1304 	*nam = in_sockaddr(port, &addr);
1305 	return 0;
1306 }
1307 
1308 int
1309 in_getpeeraddr(struct socket *so, struct sockaddr **nam)
1310 {
1311 	struct inpcb *inp;
1312 	struct in_addr addr;
1313 	in_port_t port;
1314 
1315 	inp = sotoinpcb(so);
1316 	KASSERT(inp != NULL, ("in_getpeeraddr: inp == NULL"));
1317 
1318 	INP_RLOCK(inp);
1319 	port = inp->inp_fport;
1320 	addr = inp->inp_faddr;
1321 	INP_RUNLOCK(inp);
1322 
1323 	*nam = in_sockaddr(port, &addr);
1324 	return 0;
1325 }
1326 
1327 void
1328 in_pcbnotifyall(struct inpcbinfo *pcbinfo, struct in_addr faddr, int errno,
1329     struct inpcb *(*notify)(struct inpcb *, int))
1330 {
1331 	struct inpcb *inp, *inp_temp;
1332 
1333 	INP_INFO_WLOCK(pcbinfo);
1334 	LIST_FOREACH_SAFE(inp, pcbinfo->ipi_listhead, inp_list, inp_temp) {
1335 		INP_WLOCK(inp);
1336 #ifdef INET6
1337 		if ((inp->inp_vflag & INP_IPV4) == 0) {
1338 			INP_WUNLOCK(inp);
1339 			continue;
1340 		}
1341 #endif
1342 		if (inp->inp_faddr.s_addr != faddr.s_addr ||
1343 		    inp->inp_socket == NULL) {
1344 			INP_WUNLOCK(inp);
1345 			continue;
1346 		}
1347 		if ((*notify)(inp, errno))
1348 			INP_WUNLOCK(inp);
1349 	}
1350 	INP_INFO_WUNLOCK(pcbinfo);
1351 }
1352 
1353 void
1354 in_pcbpurgeif0(struct inpcbinfo *pcbinfo, struct ifnet *ifp)
1355 {
1356 	struct inpcb *inp;
1357 	struct ip_moptions *imo;
1358 	int i, gap;
1359 
1360 	INP_INFO_RLOCK(pcbinfo);
1361 	LIST_FOREACH(inp, pcbinfo->ipi_listhead, inp_list) {
1362 		INP_WLOCK(inp);
1363 		imo = inp->inp_moptions;
1364 		if ((inp->inp_vflag & INP_IPV4) &&
1365 		    imo != NULL) {
1366 			/*
1367 			 * Unselect the outgoing interface if it is being
1368 			 * detached.
1369 			 */
1370 			if (imo->imo_multicast_ifp == ifp)
1371 				imo->imo_multicast_ifp = NULL;
1372 
1373 			/*
1374 			 * Drop multicast group membership if we joined
1375 			 * through the interface being detached.
1376 			 */
1377 			for (i = 0, gap = 0; i < imo->imo_num_memberships;
1378 			    i++) {
1379 				if (imo->imo_membership[i]->inm_ifp == ifp) {
1380 					in_delmulti(imo->imo_membership[i]);
1381 					gap++;
1382 				} else if (gap != 0)
1383 					imo->imo_membership[i - gap] =
1384 					    imo->imo_membership[i];
1385 			}
1386 			imo->imo_num_memberships -= gap;
1387 		}
1388 		INP_WUNLOCK(inp);
1389 	}
1390 	INP_INFO_RUNLOCK(pcbinfo);
1391 }
1392 
1393 /*
1394  * Lookup a PCB based on the local address and port.  Caller must hold the
1395  * hash lock.  No inpcb locks or references are acquired.
1396  */
1397 #define INP_LOOKUP_MAPPED_PCB_COST	3
1398 struct inpcb *
1399 in_pcblookup_local(struct inpcbinfo *pcbinfo, struct in_addr laddr,
1400     u_short lport, int lookupflags, struct ucred *cred)
1401 {
1402 	struct inpcb *inp;
1403 #ifdef INET6
1404 	int matchwild = 3 + INP_LOOKUP_MAPPED_PCB_COST;
1405 #else
1406 	int matchwild = 3;
1407 #endif
1408 	int wildcard;
1409 
1410 	KASSERT((lookupflags & ~(INPLOOKUP_WILDCARD)) == 0,
1411 	    ("%s: invalid lookup flags %d", __func__, lookupflags));
1412 
1413 	INP_HASH_LOCK_ASSERT(pcbinfo);
1414 
1415 	if ((lookupflags & INPLOOKUP_WILDCARD) == 0) {
1416 		struct inpcbhead *head;
1417 		/*
1418 		 * Look for an unconnected (wildcard foreign addr) PCB that
1419 		 * matches the local address and port we're looking for.
1420 		 */
1421 		head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport,
1422 		    0, pcbinfo->ipi_hashmask)];
1423 		LIST_FOREACH(inp, head, inp_hash) {
1424 #ifdef INET6
1425 			/* XXX inp locking */
1426 			if ((inp->inp_vflag & INP_IPV4) == 0)
1427 				continue;
1428 #endif
1429 			if (inp->inp_faddr.s_addr == INADDR_ANY &&
1430 			    inp->inp_laddr.s_addr == laddr.s_addr &&
1431 			    inp->inp_lport == lport) {
1432 				/*
1433 				 * Found?
1434 				 */
1435 				if (cred == NULL ||
1436 				    prison_equal_ip4(cred->cr_prison,
1437 					inp->inp_cred->cr_prison))
1438 					return (inp);
1439 			}
1440 		}
1441 		/*
1442 		 * Not found.
1443 		 */
1444 		return (NULL);
1445 	} else {
1446 		struct inpcbporthead *porthash;
1447 		struct inpcbport *phd;
1448 		struct inpcb *match = NULL;
1449 		/*
1450 		 * Best fit PCB lookup.
1451 		 *
1452 		 * First see if this local port is in use by looking on the
1453 		 * port hash list.
1454 		 */
1455 		porthash = &pcbinfo->ipi_porthashbase[INP_PCBPORTHASH(lport,
1456 		    pcbinfo->ipi_porthashmask)];
1457 		LIST_FOREACH(phd, porthash, phd_hash) {
1458 			if (phd->phd_port == lport)
1459 				break;
1460 		}
1461 		if (phd != NULL) {
1462 			/*
1463 			 * Port is in use by one or more PCBs. Look for best
1464 			 * fit.
1465 			 */
1466 			LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) {
1467 				wildcard = 0;
1468 				if (cred != NULL &&
1469 				    !prison_equal_ip4(inp->inp_cred->cr_prison,
1470 					cred->cr_prison))
1471 					continue;
1472 #ifdef INET6
1473 				/* XXX inp locking */
1474 				if ((inp->inp_vflag & INP_IPV4) == 0)
1475 					continue;
1476 				/*
1477 				 * We never select the PCB that has
1478 				 * INP_IPV6 flag and is bound to :: if
1479 				 * we have another PCB which is bound
1480 				 * to 0.0.0.0.  If a PCB has the
1481 				 * INP_IPV6 flag, then we set its cost
1482 				 * higher than IPv4 only PCBs.
1483 				 *
1484 				 * Note that the case only happens
1485 				 * when a socket is bound to ::, under
1486 				 * the condition that the use of the
1487 				 * mapped address is allowed.
1488 				 */
1489 				if ((inp->inp_vflag & INP_IPV6) != 0)
1490 					wildcard += INP_LOOKUP_MAPPED_PCB_COST;
1491 #endif
1492 				if (inp->inp_faddr.s_addr != INADDR_ANY)
1493 					wildcard++;
1494 				if (inp->inp_laddr.s_addr != INADDR_ANY) {
1495 					if (laddr.s_addr == INADDR_ANY)
1496 						wildcard++;
1497 					else if (inp->inp_laddr.s_addr != laddr.s_addr)
1498 						continue;
1499 				} else {
1500 					if (laddr.s_addr != INADDR_ANY)
1501 						wildcard++;
1502 				}
1503 				if (wildcard < matchwild) {
1504 					match = inp;
1505 					matchwild = wildcard;
1506 					if (matchwild == 0)
1507 						break;
1508 				}
1509 			}
1510 		}
1511 		return (match);
1512 	}
1513 }
1514 #undef INP_LOOKUP_MAPPED_PCB_COST
1515 
1516 #ifdef PCBGROUP
1517 /*
1518  * Lookup PCB in hash list, using pcbgroup tables.
1519  */
1520 static struct inpcb *
1521 in_pcblookup_group(struct inpcbinfo *pcbinfo, struct inpcbgroup *pcbgroup,
1522     struct in_addr faddr, u_int fport_arg, struct in_addr laddr,
1523     u_int lport_arg, int lookupflags, struct ifnet *ifp)
1524 {
1525 	struct inpcbhead *head;
1526 	struct inpcb *inp, *tmpinp;
1527 	u_short fport = fport_arg, lport = lport_arg;
1528 
1529 	/*
1530 	 * First look for an exact match.
1531 	 */
1532 	tmpinp = NULL;
1533 	INP_GROUP_LOCK(pcbgroup);
1534 	head = &pcbgroup->ipg_hashbase[INP_PCBHASH(faddr.s_addr, lport, fport,
1535 	    pcbgroup->ipg_hashmask)];
1536 	LIST_FOREACH(inp, head, inp_pcbgrouphash) {
1537 #ifdef INET6
1538 		/* XXX inp locking */
1539 		if ((inp->inp_vflag & INP_IPV4) == 0)
1540 			continue;
1541 #endif
1542 		if (inp->inp_faddr.s_addr == faddr.s_addr &&
1543 		    inp->inp_laddr.s_addr == laddr.s_addr &&
1544 		    inp->inp_fport == fport &&
1545 		    inp->inp_lport == lport) {
1546 			/*
1547 			 * XXX We should be able to directly return
1548 			 * the inp here, without any checks.
1549 			 * Well unless both bound with SO_REUSEPORT?
1550 			 */
1551 			if (prison_flag(inp->inp_cred, PR_IP4))
1552 				goto found;
1553 			if (tmpinp == NULL)
1554 				tmpinp = inp;
1555 		}
1556 	}
1557 	if (tmpinp != NULL) {
1558 		inp = tmpinp;
1559 		goto found;
1560 	}
1561 
1562 	/*
1563 	 * Then look for a wildcard match, if requested.
1564 	 */
1565 	if ((lookupflags & INPLOOKUP_WILDCARD) != 0) {
1566 		struct inpcb *local_wild = NULL, *local_exact = NULL;
1567 #ifdef INET6
1568 		struct inpcb *local_wild_mapped = NULL;
1569 #endif
1570 		struct inpcb *jail_wild = NULL;
1571 		struct inpcbhead *head;
1572 		int injail;
1573 
1574 		/*
1575 		 * Order of socket selection - we always prefer jails.
1576 		 *      1. jailed, non-wild.
1577 		 *      2. jailed, wild.
1578 		 *      3. non-jailed, non-wild.
1579 		 *      4. non-jailed, wild.
1580 		 */
1581 		head = &pcbinfo->ipi_wildbase[INP_PCBHASH(INADDR_ANY, lport,
1582 		    0, pcbinfo->ipi_wildmask)];
1583 		LIST_FOREACH(inp, head, inp_pcbgroup_wild) {
1584 #ifdef INET6
1585 			/* XXX inp locking */
1586 			if ((inp->inp_vflag & INP_IPV4) == 0)
1587 				continue;
1588 #endif
1589 			if (inp->inp_faddr.s_addr != INADDR_ANY ||
1590 			    inp->inp_lport != lport)
1591 				continue;
1592 
1593 			/* XXX inp locking */
1594 			if (ifp && ifp->if_type == IFT_FAITH &&
1595 			    (inp->inp_flags & INP_FAITH) == 0)
1596 				continue;
1597 
1598 			injail = prison_flag(inp->inp_cred, PR_IP4);
1599 			if (injail) {
1600 				if (prison_check_ip4(inp->inp_cred,
1601 				    &laddr) != 0)
1602 					continue;
1603 			} else {
1604 				if (local_exact != NULL)
1605 					continue;
1606 			}
1607 
1608 			if (inp->inp_laddr.s_addr == laddr.s_addr) {
1609 				if (injail)
1610 					goto found;
1611 				else
1612 					local_exact = inp;
1613 			} else if (inp->inp_laddr.s_addr == INADDR_ANY) {
1614 #ifdef INET6
1615 				/* XXX inp locking, NULL check */
1616 				if (inp->inp_vflag & INP_IPV6PROTO)
1617 					local_wild_mapped = inp;
1618 				else
1619 #endif
1620 					if (injail)
1621 						jail_wild = inp;
1622 					else
1623 						local_wild = inp;
1624 			}
1625 		} /* LIST_FOREACH */
1626 		inp = jail_wild;
1627 		if (inp == NULL)
1628 			inp = local_exact;
1629 		if (inp == NULL)
1630 			inp = local_wild;
1631 #ifdef INET6
1632 		if (inp == NULL)
1633 			inp = local_wild_mapped;
1634 #endif
1635 		if (inp != NULL)
1636 			goto found;
1637 	} /* if (lookupflags & INPLOOKUP_WILDCARD) */
1638 	INP_GROUP_UNLOCK(pcbgroup);
1639 	return (NULL);
1640 
1641 found:
1642 	in_pcbref(inp);
1643 	INP_GROUP_UNLOCK(pcbgroup);
1644 	if (lookupflags & INPLOOKUP_WLOCKPCB) {
1645 		INP_WLOCK(inp);
1646 		if (in_pcbrele_wlocked(inp))
1647 			return (NULL);
1648 	} else if (lookupflags & INPLOOKUP_RLOCKPCB) {
1649 		INP_RLOCK(inp);
1650 		if (in_pcbrele_rlocked(inp))
1651 			return (NULL);
1652 	} else
1653 		panic("%s: locking bug", __func__);
1654 	return (inp);
1655 }
1656 #endif /* PCBGROUP */
1657 
1658 /*
1659  * Lookup PCB in hash list, using pcbinfo tables.  This variation assumes
1660  * that the caller has locked the hash list, and will not perform any further
1661  * locking or reference operations on either the hash list or the connection.
1662  */
1663 static struct inpcb *
1664 in_pcblookup_hash_locked(struct inpcbinfo *pcbinfo, struct in_addr faddr,
1665     u_int fport_arg, struct in_addr laddr, u_int lport_arg, int lookupflags,
1666     struct ifnet *ifp)
1667 {
1668 	struct inpcbhead *head;
1669 	struct inpcb *inp, *tmpinp;
1670 	u_short fport = fport_arg, lport = lport_arg;
1671 
1672 	KASSERT((lookupflags & ~(INPLOOKUP_WILDCARD)) == 0,
1673 	    ("%s: invalid lookup flags %d", __func__, lookupflags));
1674 
1675 	INP_HASH_LOCK_ASSERT(pcbinfo);
1676 
1677 	/*
1678 	 * First look for an exact match.
1679 	 */
1680 	tmpinp = NULL;
1681 	head = &pcbinfo->ipi_hashbase[INP_PCBHASH(faddr.s_addr, lport, fport,
1682 	    pcbinfo->ipi_hashmask)];
1683 	LIST_FOREACH(inp, head, inp_hash) {
1684 #ifdef INET6
1685 		/* XXX inp locking */
1686 		if ((inp->inp_vflag & INP_IPV4) == 0)
1687 			continue;
1688 #endif
1689 		if (inp->inp_faddr.s_addr == faddr.s_addr &&
1690 		    inp->inp_laddr.s_addr == laddr.s_addr &&
1691 		    inp->inp_fport == fport &&
1692 		    inp->inp_lport == lport) {
1693 			/*
1694 			 * XXX We should be able to directly return
1695 			 * the inp here, without any checks.
1696 			 * Well unless both bound with SO_REUSEPORT?
1697 			 */
1698 			if (prison_flag(inp->inp_cred, PR_IP4))
1699 				return (inp);
1700 			if (tmpinp == NULL)
1701 				tmpinp = inp;
1702 		}
1703 	}
1704 	if (tmpinp != NULL)
1705 		return (tmpinp);
1706 
1707 	/*
1708 	 * Then look for a wildcard match, if requested.
1709 	 */
1710 	if ((lookupflags & INPLOOKUP_WILDCARD) != 0) {
1711 		struct inpcb *local_wild = NULL, *local_exact = NULL;
1712 #ifdef INET6
1713 		struct inpcb *local_wild_mapped = NULL;
1714 #endif
1715 		struct inpcb *jail_wild = NULL;
1716 		int injail;
1717 
1718 		/*
1719 		 * Order of socket selection - we always prefer jails.
1720 		 *      1. jailed, non-wild.
1721 		 *      2. jailed, wild.
1722 		 *      3. non-jailed, non-wild.
1723 		 *      4. non-jailed, wild.
1724 		 */
1725 
1726 		head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport,
1727 		    0, pcbinfo->ipi_hashmask)];
1728 		LIST_FOREACH(inp, head, inp_hash) {
1729 #ifdef INET6
1730 			/* XXX inp locking */
1731 			if ((inp->inp_vflag & INP_IPV4) == 0)
1732 				continue;
1733 #endif
1734 			if (inp->inp_faddr.s_addr != INADDR_ANY ||
1735 			    inp->inp_lport != lport)
1736 				continue;
1737 
1738 			/* XXX inp locking */
1739 			if (ifp && ifp->if_type == IFT_FAITH &&
1740 			    (inp->inp_flags & INP_FAITH) == 0)
1741 				continue;
1742 
1743 			injail = prison_flag(inp->inp_cred, PR_IP4);
1744 			if (injail) {
1745 				if (prison_check_ip4(inp->inp_cred,
1746 				    &laddr) != 0)
1747 					continue;
1748 			} else {
1749 				if (local_exact != NULL)
1750 					continue;
1751 			}
1752 
1753 			if (inp->inp_laddr.s_addr == laddr.s_addr) {
1754 				if (injail)
1755 					return (inp);
1756 				else
1757 					local_exact = inp;
1758 			} else if (inp->inp_laddr.s_addr == INADDR_ANY) {
1759 #ifdef INET6
1760 				/* XXX inp locking, NULL check */
1761 				if (inp->inp_vflag & INP_IPV6PROTO)
1762 					local_wild_mapped = inp;
1763 				else
1764 #endif
1765 					if (injail)
1766 						jail_wild = inp;
1767 					else
1768 						local_wild = inp;
1769 			}
1770 		} /* LIST_FOREACH */
1771 		if (jail_wild != NULL)
1772 			return (jail_wild);
1773 		if (local_exact != NULL)
1774 			return (local_exact);
1775 		if (local_wild != NULL)
1776 			return (local_wild);
1777 #ifdef INET6
1778 		if (local_wild_mapped != NULL)
1779 			return (local_wild_mapped);
1780 #endif
1781 	} /* if ((lookupflags & INPLOOKUP_WILDCARD) != 0) */
1782 
1783 	return (NULL);
1784 }
1785 
1786 /*
1787  * Lookup PCB in hash list, using pcbinfo tables.  This variation locks the
1788  * hash list lock, and will return the inpcb locked (i.e., requires
1789  * INPLOOKUP_LOCKPCB).
1790  */
1791 static struct inpcb *
1792 in_pcblookup_hash(struct inpcbinfo *pcbinfo, struct in_addr faddr,
1793     u_int fport, struct in_addr laddr, u_int lport, int lookupflags,
1794     struct ifnet *ifp)
1795 {
1796 	struct inpcb *inp;
1797 
1798 	INP_HASH_RLOCK(pcbinfo);
1799 	inp = in_pcblookup_hash_locked(pcbinfo, faddr, fport, laddr, lport,
1800 	    (lookupflags & ~(INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)), ifp);
1801 	if (inp != NULL) {
1802 		in_pcbref(inp);
1803 		INP_HASH_RUNLOCK(pcbinfo);
1804 		if (lookupflags & INPLOOKUP_WLOCKPCB) {
1805 			INP_WLOCK(inp);
1806 			if (in_pcbrele_wlocked(inp))
1807 				return (NULL);
1808 		} else if (lookupflags & INPLOOKUP_RLOCKPCB) {
1809 			INP_RLOCK(inp);
1810 			if (in_pcbrele_rlocked(inp))
1811 				return (NULL);
1812 		} else
1813 			panic("%s: locking bug", __func__);
1814 	} else
1815 		INP_HASH_RUNLOCK(pcbinfo);
1816 	return (inp);
1817 }
1818 
1819 /*
1820  * Public inpcb lookup routines, accepting a 4-tuple, and optionally, an mbuf
1821  * from which a pre-calculated hash value may be extracted.
1822  *
1823  * Possibly more of this logic should be in in_pcbgroup.c.
1824  */
1825 struct inpcb *
1826 in_pcblookup(struct inpcbinfo *pcbinfo, struct in_addr faddr, u_int fport,
1827     struct in_addr laddr, u_int lport, int lookupflags, struct ifnet *ifp)
1828 {
1829 #if defined(PCBGROUP) && !defined(RSS)
1830 	struct inpcbgroup *pcbgroup;
1831 #endif
1832 
1833 	KASSERT((lookupflags & ~INPLOOKUP_MASK) == 0,
1834 	    ("%s: invalid lookup flags %d", __func__, lookupflags));
1835 	KASSERT((lookupflags & (INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)) != 0,
1836 	    ("%s: LOCKPCB not set", __func__));
1837 
1838 	/*
1839 	 * When not using RSS, use connection groups in preference to the
1840 	 * reservation table when looking up 4-tuples.  When using RSS, just
1841 	 * use the reservation table, due to the cost of the Toeplitz hash
1842 	 * in software.
1843 	 *
1844 	 * XXXRW: This policy belongs in the pcbgroup code, as in principle
1845 	 * we could be doing RSS with a non-Toeplitz hash that is affordable
1846 	 * in software.
1847 	 */
1848 #if defined(PCBGROUP) && !defined(RSS)
1849 	if (in_pcbgroup_enabled(pcbinfo)) {
1850 		pcbgroup = in_pcbgroup_bytuple(pcbinfo, laddr, lport, faddr,
1851 		    fport);
1852 		return (in_pcblookup_group(pcbinfo, pcbgroup, faddr, fport,
1853 		    laddr, lport, lookupflags, ifp));
1854 	}
1855 #endif
1856 	return (in_pcblookup_hash(pcbinfo, faddr, fport, laddr, lport,
1857 	    lookupflags, ifp));
1858 }
1859 
1860 struct inpcb *
1861 in_pcblookup_mbuf(struct inpcbinfo *pcbinfo, struct in_addr faddr,
1862     u_int fport, struct in_addr laddr, u_int lport, int lookupflags,
1863     struct ifnet *ifp, struct mbuf *m)
1864 {
1865 #ifdef PCBGROUP
1866 	struct inpcbgroup *pcbgroup;
1867 #endif
1868 
1869 	KASSERT((lookupflags & ~INPLOOKUP_MASK) == 0,
1870 	    ("%s: invalid lookup flags %d", __func__, lookupflags));
1871 	KASSERT((lookupflags & (INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)) != 0,
1872 	    ("%s: LOCKPCB not set", __func__));
1873 
1874 #ifdef PCBGROUP
1875 	/*
1876 	 * If we can use a hardware-generated hash to look up the connection
1877 	 * group, use that connection group to find the inpcb.  Otherwise
1878 	 * fall back on a software hash -- or the reservation table if we're
1879 	 * using RSS.
1880 	 *
1881 	 * XXXRW: As above, that policy belongs in the pcbgroup code.
1882 	 */
1883 	if (in_pcbgroup_enabled(pcbinfo) &&
1884 	    !(M_HASHTYPE_TEST(m, M_HASHTYPE_NONE))) {
1885 		pcbgroup = in_pcbgroup_byhash(pcbinfo, M_HASHTYPE_GET(m),
1886 		    m->m_pkthdr.flowid);
1887 		if (pcbgroup != NULL)
1888 			return (in_pcblookup_group(pcbinfo, pcbgroup, faddr,
1889 			    fport, laddr, lport, lookupflags, ifp));
1890 #ifndef RSS
1891 		pcbgroup = in_pcbgroup_bytuple(pcbinfo, laddr, lport, faddr,
1892 		    fport);
1893 		return (in_pcblookup_group(pcbinfo, pcbgroup, faddr, fport,
1894 		    laddr, lport, lookupflags, ifp));
1895 #endif
1896 	}
1897 #endif
1898 	return (in_pcblookup_hash(pcbinfo, faddr, fport, laddr, lport,
1899 	    lookupflags, ifp));
1900 }
1901 #endif /* INET */
1902 
1903 /*
1904  * Insert PCB onto various hash lists.
1905  */
1906 static int
1907 in_pcbinshash_internal(struct inpcb *inp, int do_pcbgroup_update)
1908 {
1909 	struct inpcbhead *pcbhash;
1910 	struct inpcbporthead *pcbporthash;
1911 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1912 	struct inpcbport *phd;
1913 	u_int32_t hashkey_faddr;
1914 
1915 	INP_WLOCK_ASSERT(inp);
1916 	INP_HASH_WLOCK_ASSERT(pcbinfo);
1917 
1918 	KASSERT((inp->inp_flags & INP_INHASHLIST) == 0,
1919 	    ("in_pcbinshash: INP_INHASHLIST"));
1920 
1921 #ifdef INET6
1922 	if (inp->inp_vflag & INP_IPV6)
1923 		hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX */;
1924 	else
1925 #endif
1926 	hashkey_faddr = inp->inp_faddr.s_addr;
1927 
1928 	pcbhash = &pcbinfo->ipi_hashbase[INP_PCBHASH(hashkey_faddr,
1929 		 inp->inp_lport, inp->inp_fport, pcbinfo->ipi_hashmask)];
1930 
1931 	pcbporthash = &pcbinfo->ipi_porthashbase[
1932 	    INP_PCBPORTHASH(inp->inp_lport, pcbinfo->ipi_porthashmask)];
1933 
1934 	/*
1935 	 * Go through port list and look for a head for this lport.
1936 	 */
1937 	LIST_FOREACH(phd, pcbporthash, phd_hash) {
1938 		if (phd->phd_port == inp->inp_lport)
1939 			break;
1940 	}
1941 	/*
1942 	 * If none exists, malloc one and tack it on.
1943 	 */
1944 	if (phd == NULL) {
1945 		phd = malloc(sizeof(struct inpcbport), M_PCB, M_NOWAIT);
1946 		if (phd == NULL) {
1947 			return (ENOBUFS); /* XXX */
1948 		}
1949 		phd->phd_port = inp->inp_lport;
1950 		LIST_INIT(&phd->phd_pcblist);
1951 		LIST_INSERT_HEAD(pcbporthash, phd, phd_hash);
1952 	}
1953 	inp->inp_phd = phd;
1954 	LIST_INSERT_HEAD(&phd->phd_pcblist, inp, inp_portlist);
1955 	LIST_INSERT_HEAD(pcbhash, inp, inp_hash);
1956 	inp->inp_flags |= INP_INHASHLIST;
1957 #ifdef PCBGROUP
1958 	if (do_pcbgroup_update)
1959 		in_pcbgroup_update(inp);
1960 #endif
1961 	return (0);
1962 }
1963 
1964 /*
1965  * For now, there are two public interfaces to insert an inpcb into the hash
1966  * lists -- one that does update pcbgroups, and one that doesn't.  The latter
1967  * is used only in the TCP syncache, where in_pcbinshash is called before the
1968  * full 4-tuple is set for the inpcb, and we don't want to install in the
1969  * pcbgroup until later.
1970  *
1971  * XXXRW: This seems like a misfeature.  in_pcbinshash should always update
1972  * connection groups, and partially initialised inpcbs should not be exposed
1973  * to either reservation hash tables or pcbgroups.
1974  */
1975 int
1976 in_pcbinshash(struct inpcb *inp)
1977 {
1978 
1979 	return (in_pcbinshash_internal(inp, 1));
1980 }
1981 
1982 int
1983 in_pcbinshash_nopcbgroup(struct inpcb *inp)
1984 {
1985 
1986 	return (in_pcbinshash_internal(inp, 0));
1987 }
1988 
1989 /*
1990  * Move PCB to the proper hash bucket when { faddr, fport } have  been
1991  * changed. NOTE: This does not handle the case of the lport changing (the
1992  * hashed port list would have to be updated as well), so the lport must
1993  * not change after in_pcbinshash() has been called.
1994  */
1995 void
1996 in_pcbrehash_mbuf(struct inpcb *inp, struct mbuf *m)
1997 {
1998 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1999 	struct inpcbhead *head;
2000 	u_int32_t hashkey_faddr;
2001 
2002 	INP_WLOCK_ASSERT(inp);
2003 	INP_HASH_WLOCK_ASSERT(pcbinfo);
2004 
2005 	KASSERT(inp->inp_flags & INP_INHASHLIST,
2006 	    ("in_pcbrehash: !INP_INHASHLIST"));
2007 
2008 #ifdef INET6
2009 	if (inp->inp_vflag & INP_IPV6)
2010 		hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX */;
2011 	else
2012 #endif
2013 	hashkey_faddr = inp->inp_faddr.s_addr;
2014 
2015 	head = &pcbinfo->ipi_hashbase[INP_PCBHASH(hashkey_faddr,
2016 		inp->inp_lport, inp->inp_fport, pcbinfo->ipi_hashmask)];
2017 
2018 	LIST_REMOVE(inp, inp_hash);
2019 	LIST_INSERT_HEAD(head, inp, inp_hash);
2020 
2021 #ifdef PCBGROUP
2022 	if (m != NULL)
2023 		in_pcbgroup_update_mbuf(inp, m);
2024 	else
2025 		in_pcbgroup_update(inp);
2026 #endif
2027 }
2028 
2029 void
2030 in_pcbrehash(struct inpcb *inp)
2031 {
2032 
2033 	in_pcbrehash_mbuf(inp, NULL);
2034 }
2035 
2036 /*
2037  * Remove PCB from various lists.
2038  */
2039 static void
2040 in_pcbremlists(struct inpcb *inp)
2041 {
2042 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
2043 
2044 	INP_INFO_WLOCK_ASSERT(pcbinfo);
2045 	INP_WLOCK_ASSERT(inp);
2046 
2047 	inp->inp_gencnt = ++pcbinfo->ipi_gencnt;
2048 	if (inp->inp_flags & INP_INHASHLIST) {
2049 		struct inpcbport *phd = inp->inp_phd;
2050 
2051 		INP_HASH_WLOCK(pcbinfo);
2052 		LIST_REMOVE(inp, inp_hash);
2053 		LIST_REMOVE(inp, inp_portlist);
2054 		if (LIST_FIRST(&phd->phd_pcblist) == NULL) {
2055 			LIST_REMOVE(phd, phd_hash);
2056 			free(phd, M_PCB);
2057 		}
2058 		INP_HASH_WUNLOCK(pcbinfo);
2059 		inp->inp_flags &= ~INP_INHASHLIST;
2060 	}
2061 	LIST_REMOVE(inp, inp_list);
2062 	pcbinfo->ipi_count--;
2063 #ifdef PCBGROUP
2064 	in_pcbgroup_remove(inp);
2065 #endif
2066 }
2067 
2068 /*
2069  * A set label operation has occurred at the socket layer, propagate the
2070  * label change into the in_pcb for the socket.
2071  */
2072 void
2073 in_pcbsosetlabel(struct socket *so)
2074 {
2075 #ifdef MAC
2076 	struct inpcb *inp;
2077 
2078 	inp = sotoinpcb(so);
2079 	KASSERT(inp != NULL, ("in_pcbsosetlabel: so->so_pcb == NULL"));
2080 
2081 	INP_WLOCK(inp);
2082 	SOCK_LOCK(so);
2083 	mac_inpcb_sosetlabel(so, inp);
2084 	SOCK_UNLOCK(so);
2085 	INP_WUNLOCK(inp);
2086 #endif
2087 }
2088 
2089 /*
2090  * ipport_tick runs once per second, determining if random port allocation
2091  * should be continued.  If more than ipport_randomcps ports have been
2092  * allocated in the last second, then we return to sequential port
2093  * allocation. We return to random allocation only once we drop below
2094  * ipport_randomcps for at least ipport_randomtime seconds.
2095  */
2096 static void
2097 ipport_tick(void *xtp)
2098 {
2099 	VNET_ITERATOR_DECL(vnet_iter);
2100 
2101 	VNET_LIST_RLOCK_NOSLEEP();
2102 	VNET_FOREACH(vnet_iter) {
2103 		CURVNET_SET(vnet_iter);	/* XXX appease INVARIANTS here */
2104 		if (V_ipport_tcpallocs <=
2105 		    V_ipport_tcplastcount + V_ipport_randomcps) {
2106 			if (V_ipport_stoprandom > 0)
2107 				V_ipport_stoprandom--;
2108 		} else
2109 			V_ipport_stoprandom = V_ipport_randomtime;
2110 		V_ipport_tcplastcount = V_ipport_tcpallocs;
2111 		CURVNET_RESTORE();
2112 	}
2113 	VNET_LIST_RUNLOCK_NOSLEEP();
2114 	callout_reset(&ipport_tick_callout, hz, ipport_tick, NULL);
2115 }
2116 
2117 static void
2118 ip_fini(void *xtp)
2119 {
2120 
2121 	callout_stop(&ipport_tick_callout);
2122 }
2123 
2124 /*
2125  * The ipport_callout should start running at about the time we attach the
2126  * inet or inet6 domains.
2127  */
2128 static void
2129 ipport_tick_init(const void *unused __unused)
2130 {
2131 
2132 	/* Start ipport_tick. */
2133 	callout_init(&ipport_tick_callout, CALLOUT_MPSAFE);
2134 	callout_reset(&ipport_tick_callout, 1, ipport_tick, NULL);
2135 	EVENTHANDLER_REGISTER(shutdown_pre_sync, ip_fini, NULL,
2136 		SHUTDOWN_PRI_DEFAULT);
2137 }
2138 SYSINIT(ipport_tick_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE,
2139     ipport_tick_init, NULL);
2140 
2141 void
2142 inp_wlock(struct inpcb *inp)
2143 {
2144 
2145 	INP_WLOCK(inp);
2146 }
2147 
2148 void
2149 inp_wunlock(struct inpcb *inp)
2150 {
2151 
2152 	INP_WUNLOCK(inp);
2153 }
2154 
2155 void
2156 inp_rlock(struct inpcb *inp)
2157 {
2158 
2159 	INP_RLOCK(inp);
2160 }
2161 
2162 void
2163 inp_runlock(struct inpcb *inp)
2164 {
2165 
2166 	INP_RUNLOCK(inp);
2167 }
2168 
2169 #ifdef INVARIANTS
2170 void
2171 inp_lock_assert(struct inpcb *inp)
2172 {
2173 
2174 	INP_WLOCK_ASSERT(inp);
2175 }
2176 
2177 void
2178 inp_unlock_assert(struct inpcb *inp)
2179 {
2180 
2181 	INP_UNLOCK_ASSERT(inp);
2182 }
2183 #endif
2184 
2185 void
2186 inp_apply_all(void (*func)(struct inpcb *, void *), void *arg)
2187 {
2188 	struct inpcb *inp;
2189 
2190 	INP_INFO_RLOCK(&V_tcbinfo);
2191 	LIST_FOREACH(inp, V_tcbinfo.ipi_listhead, inp_list) {
2192 		INP_WLOCK(inp);
2193 		func(inp, arg);
2194 		INP_WUNLOCK(inp);
2195 	}
2196 	INP_INFO_RUNLOCK(&V_tcbinfo);
2197 }
2198 
2199 struct socket *
2200 inp_inpcbtosocket(struct inpcb *inp)
2201 {
2202 
2203 	INP_WLOCK_ASSERT(inp);
2204 	return (inp->inp_socket);
2205 }
2206 
2207 struct tcpcb *
2208 inp_inpcbtotcpcb(struct inpcb *inp)
2209 {
2210 
2211 	INP_WLOCK_ASSERT(inp);
2212 	return ((struct tcpcb *)inp->inp_ppcb);
2213 }
2214 
2215 int
2216 inp_ip_tos_get(const struct inpcb *inp)
2217 {
2218 
2219 	return (inp->inp_ip_tos);
2220 }
2221 
2222 void
2223 inp_ip_tos_set(struct inpcb *inp, int val)
2224 {
2225 
2226 	inp->inp_ip_tos = val;
2227 }
2228 
2229 void
2230 inp_4tuple_get(struct inpcb *inp, uint32_t *laddr, uint16_t *lp,
2231     uint32_t *faddr, uint16_t *fp)
2232 {
2233 
2234 	INP_LOCK_ASSERT(inp);
2235 	*laddr = inp->inp_laddr.s_addr;
2236 	*faddr = inp->inp_faddr.s_addr;
2237 	*lp = inp->inp_lport;
2238 	*fp = inp->inp_fport;
2239 }
2240 
2241 struct inpcb *
2242 so_sotoinpcb(struct socket *so)
2243 {
2244 
2245 	return (sotoinpcb(so));
2246 }
2247 
2248 struct tcpcb *
2249 so_sototcpcb(struct socket *so)
2250 {
2251 
2252 	return (sototcpcb(so));
2253 }
2254 
2255 #ifdef DDB
2256 static void
2257 db_print_indent(int indent)
2258 {
2259 	int i;
2260 
2261 	for (i = 0; i < indent; i++)
2262 		db_printf(" ");
2263 }
2264 
2265 static void
2266 db_print_inconninfo(struct in_conninfo *inc, const char *name, int indent)
2267 {
2268 	char faddr_str[48], laddr_str[48];
2269 
2270 	db_print_indent(indent);
2271 	db_printf("%s at %p\n", name, inc);
2272 
2273 	indent += 2;
2274 
2275 #ifdef INET6
2276 	if (inc->inc_flags & INC_ISIPV6) {
2277 		/* IPv6. */
2278 		ip6_sprintf(laddr_str, &inc->inc6_laddr);
2279 		ip6_sprintf(faddr_str, &inc->inc6_faddr);
2280 	} else
2281 #endif
2282 	{
2283 		/* IPv4. */
2284 		inet_ntoa_r(inc->inc_laddr, laddr_str);
2285 		inet_ntoa_r(inc->inc_faddr, faddr_str);
2286 	}
2287 	db_print_indent(indent);
2288 	db_printf("inc_laddr %s   inc_lport %u\n", laddr_str,
2289 	    ntohs(inc->inc_lport));
2290 	db_print_indent(indent);
2291 	db_printf("inc_faddr %s   inc_fport %u\n", faddr_str,
2292 	    ntohs(inc->inc_fport));
2293 }
2294 
2295 static void
2296 db_print_inpflags(int inp_flags)
2297 {
2298 	int comma;
2299 
2300 	comma = 0;
2301 	if (inp_flags & INP_RECVOPTS) {
2302 		db_printf("%sINP_RECVOPTS", comma ? ", " : "");
2303 		comma = 1;
2304 	}
2305 	if (inp_flags & INP_RECVRETOPTS) {
2306 		db_printf("%sINP_RECVRETOPTS", comma ? ", " : "");
2307 		comma = 1;
2308 	}
2309 	if (inp_flags & INP_RECVDSTADDR) {
2310 		db_printf("%sINP_RECVDSTADDR", comma ? ", " : "");
2311 		comma = 1;
2312 	}
2313 	if (inp_flags & INP_HDRINCL) {
2314 		db_printf("%sINP_HDRINCL", comma ? ", " : "");
2315 		comma = 1;
2316 	}
2317 	if (inp_flags & INP_HIGHPORT) {
2318 		db_printf("%sINP_HIGHPORT", comma ? ", " : "");
2319 		comma = 1;
2320 	}
2321 	if (inp_flags & INP_LOWPORT) {
2322 		db_printf("%sINP_LOWPORT", comma ? ", " : "");
2323 		comma = 1;
2324 	}
2325 	if (inp_flags & INP_ANONPORT) {
2326 		db_printf("%sINP_ANONPORT", comma ? ", " : "");
2327 		comma = 1;
2328 	}
2329 	if (inp_flags & INP_RECVIF) {
2330 		db_printf("%sINP_RECVIF", comma ? ", " : "");
2331 		comma = 1;
2332 	}
2333 	if (inp_flags & INP_MTUDISC) {
2334 		db_printf("%sINP_MTUDISC", comma ? ", " : "");
2335 		comma = 1;
2336 	}
2337 	if (inp_flags & INP_FAITH) {
2338 		db_printf("%sINP_FAITH", comma ? ", " : "");
2339 		comma = 1;
2340 	}
2341 	if (inp_flags & INP_RECVTTL) {
2342 		db_printf("%sINP_RECVTTL", comma ? ", " : "");
2343 		comma = 1;
2344 	}
2345 	if (inp_flags & INP_DONTFRAG) {
2346 		db_printf("%sINP_DONTFRAG", comma ? ", " : "");
2347 		comma = 1;
2348 	}
2349 	if (inp_flags & INP_RECVTOS) {
2350 		db_printf("%sINP_RECVTOS", comma ? ", " : "");
2351 		comma = 1;
2352 	}
2353 	if (inp_flags & IN6P_IPV6_V6ONLY) {
2354 		db_printf("%sIN6P_IPV6_V6ONLY", comma ? ", " : "");
2355 		comma = 1;
2356 	}
2357 	if (inp_flags & IN6P_PKTINFO) {
2358 		db_printf("%sIN6P_PKTINFO", comma ? ", " : "");
2359 		comma = 1;
2360 	}
2361 	if (inp_flags & IN6P_HOPLIMIT) {
2362 		db_printf("%sIN6P_HOPLIMIT", comma ? ", " : "");
2363 		comma = 1;
2364 	}
2365 	if (inp_flags & IN6P_HOPOPTS) {
2366 		db_printf("%sIN6P_HOPOPTS", comma ? ", " : "");
2367 		comma = 1;
2368 	}
2369 	if (inp_flags & IN6P_DSTOPTS) {
2370 		db_printf("%sIN6P_DSTOPTS", comma ? ", " : "");
2371 		comma = 1;
2372 	}
2373 	if (inp_flags & IN6P_RTHDR) {
2374 		db_printf("%sIN6P_RTHDR", comma ? ", " : "");
2375 		comma = 1;
2376 	}
2377 	if (inp_flags & IN6P_RTHDRDSTOPTS) {
2378 		db_printf("%sIN6P_RTHDRDSTOPTS", comma ? ", " : "");
2379 		comma = 1;
2380 	}
2381 	if (inp_flags & IN6P_TCLASS) {
2382 		db_printf("%sIN6P_TCLASS", comma ? ", " : "");
2383 		comma = 1;
2384 	}
2385 	if (inp_flags & IN6P_AUTOFLOWLABEL) {
2386 		db_printf("%sIN6P_AUTOFLOWLABEL", comma ? ", " : "");
2387 		comma = 1;
2388 	}
2389 	if (inp_flags & INP_TIMEWAIT) {
2390 		db_printf("%sINP_TIMEWAIT", comma ? ", " : "");
2391 		comma  = 1;
2392 	}
2393 	if (inp_flags & INP_ONESBCAST) {
2394 		db_printf("%sINP_ONESBCAST", comma ? ", " : "");
2395 		comma  = 1;
2396 	}
2397 	if (inp_flags & INP_DROPPED) {
2398 		db_printf("%sINP_DROPPED", comma ? ", " : "");
2399 		comma  = 1;
2400 	}
2401 	if (inp_flags & INP_SOCKREF) {
2402 		db_printf("%sINP_SOCKREF", comma ? ", " : "");
2403 		comma  = 1;
2404 	}
2405 	if (inp_flags & IN6P_RFC2292) {
2406 		db_printf("%sIN6P_RFC2292", comma ? ", " : "");
2407 		comma = 1;
2408 	}
2409 	if (inp_flags & IN6P_MTU) {
2410 		db_printf("IN6P_MTU%s", comma ? ", " : "");
2411 		comma = 1;
2412 	}
2413 }
2414 
2415 static void
2416 db_print_inpvflag(u_char inp_vflag)
2417 {
2418 	int comma;
2419 
2420 	comma = 0;
2421 	if (inp_vflag & INP_IPV4) {
2422 		db_printf("%sINP_IPV4", comma ? ", " : "");
2423 		comma  = 1;
2424 	}
2425 	if (inp_vflag & INP_IPV6) {
2426 		db_printf("%sINP_IPV6", comma ? ", " : "");
2427 		comma  = 1;
2428 	}
2429 	if (inp_vflag & INP_IPV6PROTO) {
2430 		db_printf("%sINP_IPV6PROTO", comma ? ", " : "");
2431 		comma  = 1;
2432 	}
2433 }
2434 
2435 static void
2436 db_print_inpcb(struct inpcb *inp, const char *name, int indent)
2437 {
2438 
2439 	db_print_indent(indent);
2440 	db_printf("%s at %p\n", name, inp);
2441 
2442 	indent += 2;
2443 
2444 	db_print_indent(indent);
2445 	db_printf("inp_flow: 0x%x\n", inp->inp_flow);
2446 
2447 	db_print_inconninfo(&inp->inp_inc, "inp_conninfo", indent);
2448 
2449 	db_print_indent(indent);
2450 	db_printf("inp_ppcb: %p   inp_pcbinfo: %p   inp_socket: %p\n",
2451 	    inp->inp_ppcb, inp->inp_pcbinfo, inp->inp_socket);
2452 
2453 	db_print_indent(indent);
2454 	db_printf("inp_label: %p   inp_flags: 0x%x (",
2455 	   inp->inp_label, inp->inp_flags);
2456 	db_print_inpflags(inp->inp_flags);
2457 	db_printf(")\n");
2458 
2459 	db_print_indent(indent);
2460 	db_printf("inp_sp: %p   inp_vflag: 0x%x (", inp->inp_sp,
2461 	    inp->inp_vflag);
2462 	db_print_inpvflag(inp->inp_vflag);
2463 	db_printf(")\n");
2464 
2465 	db_print_indent(indent);
2466 	db_printf("inp_ip_ttl: %d   inp_ip_p: %d   inp_ip_minttl: %d\n",
2467 	    inp->inp_ip_ttl, inp->inp_ip_p, inp->inp_ip_minttl);
2468 
2469 	db_print_indent(indent);
2470 #ifdef INET6
2471 	if (inp->inp_vflag & INP_IPV6) {
2472 		db_printf("in6p_options: %p   in6p_outputopts: %p   "
2473 		    "in6p_moptions: %p\n", inp->in6p_options,
2474 		    inp->in6p_outputopts, inp->in6p_moptions);
2475 		db_printf("in6p_icmp6filt: %p   in6p_cksum %d   "
2476 		    "in6p_hops %u\n", inp->in6p_icmp6filt, inp->in6p_cksum,
2477 		    inp->in6p_hops);
2478 	} else
2479 #endif
2480 	{
2481 		db_printf("inp_ip_tos: %d   inp_ip_options: %p   "
2482 		    "inp_ip_moptions: %p\n", inp->inp_ip_tos,
2483 		    inp->inp_options, inp->inp_moptions);
2484 	}
2485 
2486 	db_print_indent(indent);
2487 	db_printf("inp_phd: %p   inp_gencnt: %ju\n", inp->inp_phd,
2488 	    (uintmax_t)inp->inp_gencnt);
2489 }
2490 
2491 DB_SHOW_COMMAND(inpcb, db_show_inpcb)
2492 {
2493 	struct inpcb *inp;
2494 
2495 	if (!have_addr) {
2496 		db_printf("usage: show inpcb <addr>\n");
2497 		return;
2498 	}
2499 	inp = (struct inpcb *)addr;
2500 
2501 	db_print_inpcb(inp, "inpcb", 0);
2502 }
2503 #endif /* DDB */
2504