xref: /freebsd/sys/netinet/in_pcb.c (revision 2be1a816b9ff69588e55be0a84cbe2a31efc0f2f)
1 /*-
2  * Copyright (c) 1982, 1986, 1991, 1993, 1995
3  *	The Regents of the University of California.
4  * Copyright (c) 2007 Robert N. M. Watson
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 4. Neither the name of the University nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	@(#)in_pcb.c	8.4 (Berkeley) 5/24/95
32  */
33 
34 #include <sys/cdefs.h>
35 __FBSDID("$FreeBSD$");
36 
37 #include "opt_ddb.h"
38 #include "opt_ipsec.h"
39 #include "opt_inet6.h"
40 #include "opt_mac.h"
41 
42 #include <sys/param.h>
43 #include <sys/systm.h>
44 #include <sys/malloc.h>
45 #include <sys/mbuf.h>
46 #include <sys/domain.h>
47 #include <sys/protosw.h>
48 #include <sys/socket.h>
49 #include <sys/socketvar.h>
50 #include <sys/priv.h>
51 #include <sys/proc.h>
52 #include <sys/jail.h>
53 #include <sys/kernel.h>
54 #include <sys/sysctl.h>
55 
56 #ifdef DDB
57 #include <ddb/ddb.h>
58 #endif
59 
60 #include <vm/uma.h>
61 
62 #include <net/if.h>
63 #include <net/if_types.h>
64 #include <net/route.h>
65 
66 #include <netinet/in.h>
67 #include <netinet/in_pcb.h>
68 #include <netinet/in_var.h>
69 #include <netinet/ip_var.h>
70 #include <netinet/tcp_var.h>
71 #include <netinet/udp.h>
72 #include <netinet/udp_var.h>
73 #ifdef INET6
74 #include <netinet/ip6.h>
75 #include <netinet6/ip6_var.h>
76 #endif /* INET6 */
77 
78 
79 #ifdef IPSEC
80 #include <netipsec/ipsec.h>
81 #include <netipsec/key.h>
82 #endif /* IPSEC */
83 
84 #include <security/mac/mac_framework.h>
85 
86 /*
87  * These configure the range of local port addresses assigned to
88  * "unspecified" outgoing connections/packets/whatever.
89  */
90 int	ipport_lowfirstauto  = IPPORT_RESERVED - 1;	/* 1023 */
91 int	ipport_lowlastauto = IPPORT_RESERVEDSTART;	/* 600 */
92 int	ipport_firstauto = IPPORT_EPHEMERALFIRST;	/* 10000 */
93 int	ipport_lastauto  = IPPORT_EPHEMERALLAST;	/* 65535 */
94 int	ipport_hifirstauto = IPPORT_HIFIRSTAUTO;	/* 49152 */
95 int	ipport_hilastauto  = IPPORT_HILASTAUTO;		/* 65535 */
96 
97 /*
98  * Reserved ports accessible only to root. There are significant
99  * security considerations that must be accounted for when changing these,
100  * but the security benefits can be great. Please be careful.
101  */
102 int	ipport_reservedhigh = IPPORT_RESERVED - 1;	/* 1023 */
103 int	ipport_reservedlow = 0;
104 
105 /* Variables dealing with random ephemeral port allocation. */
106 int	ipport_randomized = 1;	/* user controlled via sysctl */
107 int	ipport_randomcps = 10;	/* user controlled via sysctl */
108 int	ipport_randomtime = 45;	/* user controlled via sysctl */
109 int	ipport_stoprandom = 0;	/* toggled by ipport_tick */
110 int	ipport_tcpallocs;
111 int	ipport_tcplastcount;
112 
113 #define RANGECHK(var, min, max) \
114 	if ((var) < (min)) { (var) = (min); } \
115 	else if ((var) > (max)) { (var) = (max); }
116 
117 static int
118 sysctl_net_ipport_check(SYSCTL_HANDLER_ARGS)
119 {
120 	int error;
121 
122 	error = sysctl_handle_int(oidp, oidp->oid_arg1, oidp->oid_arg2, req);
123 	if (error == 0) {
124 		RANGECHK(ipport_lowfirstauto, 1, IPPORT_RESERVED - 1);
125 		RANGECHK(ipport_lowlastauto, 1, IPPORT_RESERVED - 1);
126 		RANGECHK(ipport_firstauto, IPPORT_RESERVED, IPPORT_MAX);
127 		RANGECHK(ipport_lastauto, IPPORT_RESERVED, IPPORT_MAX);
128 		RANGECHK(ipport_hifirstauto, IPPORT_RESERVED, IPPORT_MAX);
129 		RANGECHK(ipport_hilastauto, IPPORT_RESERVED, IPPORT_MAX);
130 	}
131 	return (error);
132 }
133 
134 #undef RANGECHK
135 
136 SYSCTL_NODE(_net_inet_ip, IPPROTO_IP, portrange, CTLFLAG_RW, 0, "IP Ports");
137 
138 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, lowfirst, CTLTYPE_INT|CTLFLAG_RW,
139 	   &ipport_lowfirstauto, 0, &sysctl_net_ipport_check, "I", "");
140 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, lowlast, CTLTYPE_INT|CTLFLAG_RW,
141 	   &ipport_lowlastauto, 0, &sysctl_net_ipport_check, "I", "");
142 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, first, CTLTYPE_INT|CTLFLAG_RW,
143 	   &ipport_firstauto, 0, &sysctl_net_ipport_check, "I", "");
144 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, last, CTLTYPE_INT|CTLFLAG_RW,
145 	   &ipport_lastauto, 0, &sysctl_net_ipport_check, "I", "");
146 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, hifirst, CTLTYPE_INT|CTLFLAG_RW,
147 	   &ipport_hifirstauto, 0, &sysctl_net_ipport_check, "I", "");
148 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, hilast, CTLTYPE_INT|CTLFLAG_RW,
149 	   &ipport_hilastauto, 0, &sysctl_net_ipport_check, "I", "");
150 SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, reservedhigh,
151 	   CTLFLAG_RW|CTLFLAG_SECURE, &ipport_reservedhigh, 0, "");
152 SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, reservedlow,
153 	   CTLFLAG_RW|CTLFLAG_SECURE, &ipport_reservedlow, 0, "");
154 SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, randomized, CTLFLAG_RW,
155 	   &ipport_randomized, 0, "Enable random port allocation");
156 SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, randomcps, CTLFLAG_RW,
157 	   &ipport_randomcps, 0, "Maximum number of random port "
158 	   "allocations before switching to a sequental one");
159 SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, randomtime, CTLFLAG_RW,
160 	   &ipport_randomtime, 0, "Minimum time to keep sequental port "
161 	   "allocation before switching to a random one");
162 
163 /*
164  * in_pcb.c: manage the Protocol Control Blocks.
165  *
166  * NOTE: It is assumed that most of these functions will be called with
167  * the pcbinfo lock held, and often, the inpcb lock held, as these utility
168  * functions often modify hash chains or addresses in pcbs.
169  */
170 
171 /*
172  * Allocate a PCB and associate it with the socket.
173  * On success return with the PCB locked.
174  */
175 int
176 in_pcballoc(struct socket *so, struct inpcbinfo *pcbinfo)
177 {
178 	struct inpcb *inp;
179 	int error;
180 
181 	INP_INFO_WLOCK_ASSERT(pcbinfo);
182 	error = 0;
183 	inp = uma_zalloc(pcbinfo->ipi_zone, M_NOWAIT);
184 	if (inp == NULL)
185 		return (ENOBUFS);
186 	bzero(inp, inp_zero_size);
187 	inp->inp_pcbinfo = pcbinfo;
188 	inp->inp_socket = so;
189 #ifdef MAC
190 	error = mac_inpcb_init(inp, M_NOWAIT);
191 	if (error != 0)
192 		goto out;
193 	SOCK_LOCK(so);
194 	mac_inpcb_create(so, inp);
195 	SOCK_UNLOCK(so);
196 #endif
197 
198 #ifdef IPSEC
199 	error = ipsec_init_policy(so, &inp->inp_sp);
200 	if (error != 0) {
201 #ifdef MAC
202 		mac_inpcb_destroy(inp);
203 #endif
204 		goto out;
205 	}
206 #endif /*IPSEC*/
207 #ifdef INET6
208 	if (INP_SOCKAF(so) == AF_INET6) {
209 		inp->inp_vflag |= INP_IPV6PROTO;
210 		if (ip6_v6only)
211 			inp->inp_flags |= IN6P_IPV6_V6ONLY;
212 	}
213 #endif
214 	LIST_INSERT_HEAD(pcbinfo->ipi_listhead, inp, inp_list);
215 	pcbinfo->ipi_count++;
216 	so->so_pcb = (caddr_t)inp;
217 #ifdef INET6
218 	if (ip6_auto_flowlabel)
219 		inp->inp_flags |= IN6P_AUTOFLOWLABEL;
220 #endif
221 	INP_WLOCK(inp);
222 	inp->inp_gencnt = ++pcbinfo->ipi_gencnt;
223 
224 #if defined(IPSEC) || defined(MAC)
225 out:
226 	if (error != 0)
227 		uma_zfree(pcbinfo->ipi_zone, inp);
228 #endif
229 	return (error);
230 }
231 
232 int
233 in_pcbbind(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred)
234 {
235 	int anonport, error;
236 
237 	INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
238 	INP_WLOCK_ASSERT(inp);
239 
240 	if (inp->inp_lport != 0 || inp->inp_laddr.s_addr != INADDR_ANY)
241 		return (EINVAL);
242 	anonport = inp->inp_lport == 0 && (nam == NULL ||
243 	    ((struct sockaddr_in *)nam)->sin_port == 0);
244 	error = in_pcbbind_setup(inp, nam, &inp->inp_laddr.s_addr,
245 	    &inp->inp_lport, cred);
246 	if (error)
247 		return (error);
248 	if (in_pcbinshash(inp) != 0) {
249 		inp->inp_laddr.s_addr = INADDR_ANY;
250 		inp->inp_lport = 0;
251 		return (EAGAIN);
252 	}
253 	if (anonport)
254 		inp->inp_flags |= INP_ANONPORT;
255 	return (0);
256 }
257 
258 /*
259  * Set up a bind operation on a PCB, performing port allocation
260  * as required, but do not actually modify the PCB. Callers can
261  * either complete the bind by setting inp_laddr/inp_lport and
262  * calling in_pcbinshash(), or they can just use the resulting
263  * port and address to authorise the sending of a once-off packet.
264  *
265  * On error, the values of *laddrp and *lportp are not changed.
266  */
267 int
268 in_pcbbind_setup(struct inpcb *inp, struct sockaddr *nam, in_addr_t *laddrp,
269     u_short *lportp, struct ucred *cred)
270 {
271 	struct socket *so = inp->inp_socket;
272 	unsigned short *lastport;
273 	struct sockaddr_in *sin;
274 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
275 	struct in_addr laddr;
276 	u_short lport = 0;
277 	int wild = 0, reuseport = (so->so_options & SO_REUSEPORT);
278 	int error, prison = 0;
279 	int dorandom;
280 
281 	/*
282 	 * Because no actual state changes occur here, a write global write
283 	 * lock on the pcbinfo isn't required.
284 	 */
285 	INP_INFO_LOCK_ASSERT(pcbinfo);
286 	INP_LOCK_ASSERT(inp);
287 
288 	if (TAILQ_EMPTY(&in_ifaddrhead)) /* XXX broken! */
289 		return (EADDRNOTAVAIL);
290 	laddr.s_addr = *laddrp;
291 	if (nam != NULL && laddr.s_addr != INADDR_ANY)
292 		return (EINVAL);
293 	if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0)
294 		wild = INPLOOKUP_WILDCARD;
295 	if (nam) {
296 		sin = (struct sockaddr_in *)nam;
297 		if (nam->sa_len != sizeof (*sin))
298 			return (EINVAL);
299 #ifdef notdef
300 		/*
301 		 * We should check the family, but old programs
302 		 * incorrectly fail to initialize it.
303 		 */
304 		if (sin->sin_family != AF_INET)
305 			return (EAFNOSUPPORT);
306 #endif
307 		if (sin->sin_addr.s_addr != INADDR_ANY)
308 			if (prison_ip(cred, 0, &sin->sin_addr.s_addr))
309 				return(EINVAL);
310 		if (sin->sin_port != *lportp) {
311 			/* Don't allow the port to change. */
312 			if (*lportp != 0)
313 				return (EINVAL);
314 			lport = sin->sin_port;
315 		}
316 		/* NB: lport is left as 0 if the port isn't being changed. */
317 		if (IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) {
318 			/*
319 			 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
320 			 * allow complete duplication of binding if
321 			 * SO_REUSEPORT is set, or if SO_REUSEADDR is set
322 			 * and a multicast address is bound on both
323 			 * new and duplicated sockets.
324 			 */
325 			if (so->so_options & SO_REUSEADDR)
326 				reuseport = SO_REUSEADDR|SO_REUSEPORT;
327 		} else if (sin->sin_addr.s_addr != INADDR_ANY) {
328 			sin->sin_port = 0;		/* yech... */
329 			bzero(&sin->sin_zero, sizeof(sin->sin_zero));
330 			if (ifa_ifwithaddr((struct sockaddr *)sin) == 0)
331 				return (EADDRNOTAVAIL);
332 		}
333 		laddr = sin->sin_addr;
334 		if (lport) {
335 			struct inpcb *t;
336 			struct tcptw *tw;
337 
338 			/* GROSS */
339 			if (ntohs(lport) <= ipport_reservedhigh &&
340 			    ntohs(lport) >= ipport_reservedlow &&
341 			    priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT,
342 			    0))
343 				return (EACCES);
344 			if (jailed(cred))
345 				prison = 1;
346 			if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)) &&
347 			    priv_check_cred(so->so_cred,
348 			    PRIV_NETINET_REUSEPORT, 0) != 0) {
349 				t = in_pcblookup_local(inp->inp_pcbinfo,
350 				    sin->sin_addr, lport,
351 				    prison ? 0 :  INPLOOKUP_WILDCARD);
352 	/*
353 	 * XXX
354 	 * This entire block sorely needs a rewrite.
355 	 */
356 				if (t &&
357 				    ((t->inp_vflag & INP_TIMEWAIT) == 0) &&
358 				    (so->so_type != SOCK_STREAM ||
359 				     ntohl(t->inp_faddr.s_addr) == INADDR_ANY) &&
360 				    (ntohl(sin->sin_addr.s_addr) != INADDR_ANY ||
361 				     ntohl(t->inp_laddr.s_addr) != INADDR_ANY ||
362 				     (t->inp_socket->so_options &
363 					 SO_REUSEPORT) == 0) &&
364 				    (so->so_cred->cr_uid !=
365 				     t->inp_socket->so_cred->cr_uid))
366 					return (EADDRINUSE);
367 			}
368 			if (prison && prison_ip(cred, 0, &sin->sin_addr.s_addr))
369 				return (EADDRNOTAVAIL);
370 			t = in_pcblookup_local(pcbinfo, sin->sin_addr,
371 			    lport, prison ? 0 : wild);
372 			if (t && (t->inp_vflag & INP_TIMEWAIT)) {
373 				/*
374 				 * XXXRW: If an incpb has had its timewait
375 				 * state recycled, we treat the address as
376 				 * being in use (for now).  This is better
377 				 * than a panic, but not desirable.
378 				 */
379 				tw = intotw(inp);
380 				if (tw == NULL ||
381 				    (reuseport & tw->tw_so_options) == 0)
382 					return (EADDRINUSE);
383 			} else if (t &&
384 			    (reuseport & t->inp_socket->so_options) == 0) {
385 #ifdef INET6
386 				if (ntohl(sin->sin_addr.s_addr) !=
387 				    INADDR_ANY ||
388 				    ntohl(t->inp_laddr.s_addr) !=
389 				    INADDR_ANY ||
390 				    INP_SOCKAF(so) ==
391 				    INP_SOCKAF(t->inp_socket))
392 #endif
393 				return (EADDRINUSE);
394 			}
395 		}
396 	}
397 	if (*lportp != 0)
398 		lport = *lportp;
399 	if (lport == 0) {
400 		u_short first, last, aux;
401 		int count;
402 
403 		if (laddr.s_addr != INADDR_ANY)
404 			if (prison_ip(cred, 0, &laddr.s_addr))
405 				return (EINVAL);
406 
407 		if (inp->inp_flags & INP_HIGHPORT) {
408 			first = ipport_hifirstauto;	/* sysctl */
409 			last  = ipport_hilastauto;
410 			lastport = &pcbinfo->ipi_lasthi;
411 		} else if (inp->inp_flags & INP_LOWPORT) {
412 			error = priv_check_cred(cred,
413 			    PRIV_NETINET_RESERVEDPORT, 0);
414 			if (error)
415 				return error;
416 			first = ipport_lowfirstauto;	/* 1023 */
417 			last  = ipport_lowlastauto;	/* 600 */
418 			lastport = &pcbinfo->ipi_lastlow;
419 		} else {
420 			first = ipport_firstauto;	/* sysctl */
421 			last  = ipport_lastauto;
422 			lastport = &pcbinfo->ipi_lastport;
423 		}
424 		/*
425 		 * For UDP, use random port allocation as long as the user
426 		 * allows it.  For TCP (and as of yet unknown) connections,
427 		 * use random port allocation only if the user allows it AND
428 		 * ipport_tick() allows it.
429 		 */
430 		if (ipport_randomized &&
431 			(!ipport_stoprandom || pcbinfo == &udbinfo))
432 			dorandom = 1;
433 		else
434 			dorandom = 0;
435 		/*
436 		 * It makes no sense to do random port allocation if
437 		 * we have the only port available.
438 		 */
439 		if (first == last)
440 			dorandom = 0;
441 		/* Make sure to not include UDP packets in the count. */
442 		if (pcbinfo != &udbinfo)
443 			ipport_tcpallocs++;
444 		/*
445 		 * Simple check to ensure all ports are not used up causing
446 		 * a deadlock here.
447 		 */
448 		if (first > last) {
449 			aux = first;
450 			first = last;
451 			last = aux;
452 		}
453 
454 		if (dorandom)
455 			*lastport = first +
456 				    (arc4random() % (last - first));
457 
458 		count = last - first;
459 
460 		do {
461 			if (count-- < 0)	/* completely used? */
462 				return (EADDRNOTAVAIL);
463 			++*lastport;
464 			if (*lastport < first || *lastport > last)
465 				*lastport = first;
466 			lport = htons(*lastport);
467 		} while (in_pcblookup_local(pcbinfo, laddr, lport,
468 		    wild));
469 	}
470 	if (prison_ip(cred, 0, &laddr.s_addr))
471 		return (EINVAL);
472 	*laddrp = laddr.s_addr;
473 	*lportp = lport;
474 	return (0);
475 }
476 
477 /*
478  * Connect from a socket to a specified address.
479  * Both address and port must be specified in argument sin.
480  * If don't have a local address for this socket yet,
481  * then pick one.
482  */
483 int
484 in_pcbconnect(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred)
485 {
486 	u_short lport, fport;
487 	in_addr_t laddr, faddr;
488 	int anonport, error;
489 
490 	INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
491 	INP_WLOCK_ASSERT(inp);
492 
493 	lport = inp->inp_lport;
494 	laddr = inp->inp_laddr.s_addr;
495 	anonport = (lport == 0);
496 	error = in_pcbconnect_setup(inp, nam, &laddr, &lport, &faddr, &fport,
497 	    NULL, cred);
498 	if (error)
499 		return (error);
500 
501 	/* Do the initial binding of the local address if required. */
502 	if (inp->inp_laddr.s_addr == INADDR_ANY && inp->inp_lport == 0) {
503 		inp->inp_lport = lport;
504 		inp->inp_laddr.s_addr = laddr;
505 		if (in_pcbinshash(inp) != 0) {
506 			inp->inp_laddr.s_addr = INADDR_ANY;
507 			inp->inp_lport = 0;
508 			return (EAGAIN);
509 		}
510 	}
511 
512 	/* Commit the remaining changes. */
513 	inp->inp_lport = lport;
514 	inp->inp_laddr.s_addr = laddr;
515 	inp->inp_faddr.s_addr = faddr;
516 	inp->inp_fport = fport;
517 	in_pcbrehash(inp);
518 
519 	if (anonport)
520 		inp->inp_flags |= INP_ANONPORT;
521 	return (0);
522 }
523 
524 /*
525  * Set up for a connect from a socket to the specified address.
526  * On entry, *laddrp and *lportp should contain the current local
527  * address and port for the PCB; these are updated to the values
528  * that should be placed in inp_laddr and inp_lport to complete
529  * the connect.
530  *
531  * On success, *faddrp and *fportp will be set to the remote address
532  * and port. These are not updated in the error case.
533  *
534  * If the operation fails because the connection already exists,
535  * *oinpp will be set to the PCB of that connection so that the
536  * caller can decide to override it. In all other cases, *oinpp
537  * is set to NULL.
538  */
539 int
540 in_pcbconnect_setup(struct inpcb *inp, struct sockaddr *nam,
541     in_addr_t *laddrp, u_short *lportp, in_addr_t *faddrp, u_short *fportp,
542     struct inpcb **oinpp, struct ucred *cred)
543 {
544 	struct sockaddr_in *sin = (struct sockaddr_in *)nam;
545 	struct in_ifaddr *ia;
546 	struct sockaddr_in sa;
547 	struct ucred *socred;
548 	struct inpcb *oinp;
549 	struct in_addr laddr, faddr;
550 	u_short lport, fport;
551 	int error;
552 
553 	/*
554 	 * Because a global state change doesn't actually occur here, a read
555 	 * lock is sufficient.
556 	 */
557 	INP_INFO_LOCK_ASSERT(inp->inp_pcbinfo);
558 	INP_LOCK_ASSERT(inp);
559 
560 	if (oinpp != NULL)
561 		*oinpp = NULL;
562 	if (nam->sa_len != sizeof (*sin))
563 		return (EINVAL);
564 	if (sin->sin_family != AF_INET)
565 		return (EAFNOSUPPORT);
566 	if (sin->sin_port == 0)
567 		return (EADDRNOTAVAIL);
568 	laddr.s_addr = *laddrp;
569 	lport = *lportp;
570 	faddr = sin->sin_addr;
571 	fport = sin->sin_port;
572 	socred = inp->inp_socket->so_cred;
573 	if (laddr.s_addr == INADDR_ANY && jailed(socred)) {
574 		bzero(&sa, sizeof(sa));
575 		sa.sin_addr.s_addr = htonl(prison_getip(socred));
576 		sa.sin_len = sizeof(sa);
577 		sa.sin_family = AF_INET;
578 		error = in_pcbbind_setup(inp, (struct sockaddr *)&sa,
579 		    &laddr.s_addr, &lport, cred);
580 		if (error)
581 			return (error);
582 	}
583 	if (!TAILQ_EMPTY(&in_ifaddrhead)) {
584 		/*
585 		 * If the destination address is INADDR_ANY,
586 		 * use the primary local address.
587 		 * If the supplied address is INADDR_BROADCAST,
588 		 * and the primary interface supports broadcast,
589 		 * choose the broadcast address for that interface.
590 		 */
591 		if (faddr.s_addr == INADDR_ANY)
592 			faddr = IA_SIN(TAILQ_FIRST(&in_ifaddrhead))->sin_addr;
593 		else if (faddr.s_addr == (u_long)INADDR_BROADCAST &&
594 		    (TAILQ_FIRST(&in_ifaddrhead)->ia_ifp->if_flags &
595 		    IFF_BROADCAST))
596 			faddr = satosin(&TAILQ_FIRST(
597 			    &in_ifaddrhead)->ia_broadaddr)->sin_addr;
598 	}
599 	if (laddr.s_addr == INADDR_ANY) {
600 		ia = (struct in_ifaddr *)0;
601 		/*
602 		 * If route is known our src addr is taken from the i/f,
603 		 * else punt.
604 		 *
605 		 * Find out route to destination
606 		 */
607 		if ((inp->inp_socket->so_options & SO_DONTROUTE) == 0)
608 			ia = ip_rtaddr(faddr);
609 		/*
610 		 * If we found a route, use the address corresponding to
611 		 * the outgoing interface.
612 		 *
613 		 * Otherwise assume faddr is reachable on a directly connected
614 		 * network and try to find a corresponding interface to take
615 		 * the source address from.
616 		 */
617 		if (ia == 0) {
618 			bzero(&sa, sizeof(sa));
619 			sa.sin_addr = faddr;
620 			sa.sin_len = sizeof(sa);
621 			sa.sin_family = AF_INET;
622 
623 			ia = ifatoia(ifa_ifwithdstaddr(sintosa(&sa)));
624 			if (ia == 0)
625 				ia = ifatoia(ifa_ifwithnet(sintosa(&sa)));
626 			if (ia == 0)
627 				return (ENETUNREACH);
628 		}
629 		/*
630 		 * If the destination address is multicast and an outgoing
631 		 * interface has been set as a multicast option, use the
632 		 * address of that interface as our source address.
633 		 */
634 		if (IN_MULTICAST(ntohl(faddr.s_addr)) &&
635 		    inp->inp_moptions != NULL) {
636 			struct ip_moptions *imo;
637 			struct ifnet *ifp;
638 
639 			imo = inp->inp_moptions;
640 			if (imo->imo_multicast_ifp != NULL) {
641 				ifp = imo->imo_multicast_ifp;
642 				TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link)
643 					if (ia->ia_ifp == ifp)
644 						break;
645 				if (ia == 0)
646 					return (EADDRNOTAVAIL);
647 			}
648 		}
649 		laddr = ia->ia_addr.sin_addr;
650 	}
651 
652 	oinp = in_pcblookup_hash(inp->inp_pcbinfo, faddr, fport, laddr, lport,
653 	    0, NULL);
654 	if (oinp != NULL) {
655 		if (oinpp != NULL)
656 			*oinpp = oinp;
657 		return (EADDRINUSE);
658 	}
659 	if (lport == 0) {
660 		error = in_pcbbind_setup(inp, NULL, &laddr.s_addr, &lport,
661 		    cred);
662 		if (error)
663 			return (error);
664 	}
665 	*laddrp = laddr.s_addr;
666 	*lportp = lport;
667 	*faddrp = faddr.s_addr;
668 	*fportp = fport;
669 	return (0);
670 }
671 
672 void
673 in_pcbdisconnect(struct inpcb *inp)
674 {
675 
676 	INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
677 	INP_WLOCK_ASSERT(inp);
678 
679 	inp->inp_faddr.s_addr = INADDR_ANY;
680 	inp->inp_fport = 0;
681 	in_pcbrehash(inp);
682 }
683 
684 /*
685  * In the old world order, in_pcbdetach() served two functions: to detach the
686  * pcb from the socket/potentially free the socket, and to free the pcb
687  * itself.  In the new world order, the protocol code is responsible for
688  * managing the relationship with the socket, and this code simply frees the
689  * pcb.
690  */
691 void
692 in_pcbdetach(struct inpcb *inp)
693 {
694 
695 	KASSERT(inp->inp_socket != NULL, ("in_pcbdetach: inp_socket == NULL"));
696 	inp->inp_socket->so_pcb = NULL;
697 	inp->inp_socket = NULL;
698 }
699 
700 void
701 in_pcbfree(struct inpcb *inp)
702 {
703 	struct inpcbinfo *ipi = inp->inp_pcbinfo;
704 
705 	KASSERT(inp->inp_socket == NULL, ("in_pcbfree: inp_socket != NULL"));
706 
707 	INP_INFO_WLOCK_ASSERT(ipi);
708 	INP_WLOCK_ASSERT(inp);
709 
710 #ifdef IPSEC
711 	ipsec4_delete_pcbpolicy(inp);
712 #endif /*IPSEC*/
713 	inp->inp_gencnt = ++ipi->ipi_gencnt;
714 	in_pcbremlists(inp);
715 	if (inp->inp_options)
716 		(void)m_free(inp->inp_options);
717 	if (inp->inp_moptions != NULL)
718 		inp_freemoptions(inp->inp_moptions);
719 	inp->inp_vflag = 0;
720 
721 #ifdef MAC
722 	mac_inpcb_destroy(inp);
723 #endif
724 	INP_WUNLOCK(inp);
725 	uma_zfree(ipi->ipi_zone, inp);
726 }
727 
728 /*
729  * TCP needs to maintain its inpcb structure after the TCP connection has
730  * been torn down.  However, it must be disconnected from the inpcb hashes as
731  * it must not prevent binding of future connections to the same port/ip
732  * combination by other inpcbs.
733  */
734 void
735 in_pcbdrop(struct inpcb *inp)
736 {
737 
738 	INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
739 	INP_WLOCK_ASSERT(inp);
740 
741 	inp->inp_vflag |= INP_DROPPED;
742 	if (inp->inp_lport) {
743 		struct inpcbport *phd = inp->inp_phd;
744 
745 		LIST_REMOVE(inp, inp_hash);
746 		LIST_REMOVE(inp, inp_portlist);
747 		if (LIST_FIRST(&phd->phd_pcblist) == NULL) {
748 			LIST_REMOVE(phd, phd_hash);
749 			free(phd, M_PCB);
750 		}
751 		inp->inp_lport = 0;
752 	}
753 }
754 
755 /*
756  * Common routines to return the socket addresses associated with inpcbs.
757  */
758 struct sockaddr *
759 in_sockaddr(in_port_t port, struct in_addr *addr_p)
760 {
761 	struct sockaddr_in *sin;
762 
763 	MALLOC(sin, struct sockaddr_in *, sizeof *sin, M_SONAME,
764 		M_WAITOK | M_ZERO);
765 	sin->sin_family = AF_INET;
766 	sin->sin_len = sizeof(*sin);
767 	sin->sin_addr = *addr_p;
768 	sin->sin_port = port;
769 
770 	return (struct sockaddr *)sin;
771 }
772 
773 int
774 in_getsockaddr(struct socket *so, struct sockaddr **nam)
775 {
776 	struct inpcb *inp;
777 	struct in_addr addr;
778 	in_port_t port;
779 
780 	inp = sotoinpcb(so);
781 	KASSERT(inp != NULL, ("in_getsockaddr: inp == NULL"));
782 
783 	INP_RLOCK(inp);
784 	port = inp->inp_lport;
785 	addr = inp->inp_laddr;
786 	INP_RUNLOCK(inp);
787 
788 	*nam = in_sockaddr(port, &addr);
789 	return 0;
790 }
791 
792 int
793 in_getpeeraddr(struct socket *so, struct sockaddr **nam)
794 {
795 	struct inpcb *inp;
796 	struct in_addr addr;
797 	in_port_t port;
798 
799 	inp = sotoinpcb(so);
800 	KASSERT(inp != NULL, ("in_getpeeraddr: inp == NULL"));
801 
802 	INP_RLOCK(inp);
803 	port = inp->inp_fport;
804 	addr = inp->inp_faddr;
805 	INP_RUNLOCK(inp);
806 
807 	*nam = in_sockaddr(port, &addr);
808 	return 0;
809 }
810 
811 void
812 in_pcbnotifyall(struct inpcbinfo *pcbinfo, struct in_addr faddr, int errno,
813     struct inpcb *(*notify)(struct inpcb *, int))
814 {
815 	struct inpcb *inp, *inp_temp;
816 
817 	INP_INFO_WLOCK(pcbinfo);
818 	LIST_FOREACH_SAFE(inp, pcbinfo->ipi_listhead, inp_list, inp_temp) {
819 		INP_WLOCK(inp);
820 #ifdef INET6
821 		if ((inp->inp_vflag & INP_IPV4) == 0) {
822 			INP_WUNLOCK(inp);
823 			continue;
824 		}
825 #endif
826 		if (inp->inp_faddr.s_addr != faddr.s_addr ||
827 		    inp->inp_socket == NULL) {
828 			INP_WUNLOCK(inp);
829 			continue;
830 		}
831 		if ((*notify)(inp, errno))
832 			INP_WUNLOCK(inp);
833 	}
834 	INP_INFO_WUNLOCK(pcbinfo);
835 }
836 
837 void
838 in_pcbpurgeif0(struct inpcbinfo *pcbinfo, struct ifnet *ifp)
839 {
840 	struct inpcb *inp;
841 	struct ip_moptions *imo;
842 	int i, gap;
843 
844 	INP_INFO_RLOCK(pcbinfo);
845 	LIST_FOREACH(inp, pcbinfo->ipi_listhead, inp_list) {
846 		INP_WLOCK(inp);
847 		imo = inp->inp_moptions;
848 		if ((inp->inp_vflag & INP_IPV4) &&
849 		    imo != NULL) {
850 			/*
851 			 * Unselect the outgoing interface if it is being
852 			 * detached.
853 			 */
854 			if (imo->imo_multicast_ifp == ifp)
855 				imo->imo_multicast_ifp = NULL;
856 
857 			/*
858 			 * Drop multicast group membership if we joined
859 			 * through the interface being detached.
860 			 */
861 			for (i = 0, gap = 0; i < imo->imo_num_memberships;
862 			    i++) {
863 				if (imo->imo_membership[i]->inm_ifp == ifp) {
864 					in_delmulti(imo->imo_membership[i]);
865 					gap++;
866 				} else if (gap != 0)
867 					imo->imo_membership[i - gap] =
868 					    imo->imo_membership[i];
869 			}
870 			imo->imo_num_memberships -= gap;
871 		}
872 		INP_WUNLOCK(inp);
873 	}
874 	INP_INFO_RUNLOCK(pcbinfo);
875 }
876 
877 /*
878  * Lookup a PCB based on the local address and port.
879  */
880 #define INP_LOOKUP_MAPPED_PCB_COST	3
881 struct inpcb *
882 in_pcblookup_local(struct inpcbinfo *pcbinfo, struct in_addr laddr,
883     u_int lport_arg, int wild_okay)
884 {
885 	struct inpcb *inp;
886 #ifdef INET6
887 	int matchwild = 3 + INP_LOOKUP_MAPPED_PCB_COST;
888 #else
889 	int matchwild = 3;
890 #endif
891 	int wildcard;
892 	u_short lport = lport_arg;
893 
894 	INP_INFO_LOCK_ASSERT(pcbinfo);
895 
896 	if (!wild_okay) {
897 		struct inpcbhead *head;
898 		/*
899 		 * Look for an unconnected (wildcard foreign addr) PCB that
900 		 * matches the local address and port we're looking for.
901 		 */
902 		head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport,
903 		    0, pcbinfo->ipi_hashmask)];
904 		LIST_FOREACH(inp, head, inp_hash) {
905 #ifdef INET6
906 			if ((inp->inp_vflag & INP_IPV4) == 0)
907 				continue;
908 #endif
909 			if (inp->inp_faddr.s_addr == INADDR_ANY &&
910 			    inp->inp_laddr.s_addr == laddr.s_addr &&
911 			    inp->inp_lport == lport) {
912 				/*
913 				 * Found.
914 				 */
915 				return (inp);
916 			}
917 		}
918 		/*
919 		 * Not found.
920 		 */
921 		return (NULL);
922 	} else {
923 		struct inpcbporthead *porthash;
924 		struct inpcbport *phd;
925 		struct inpcb *match = NULL;
926 		/*
927 		 * Best fit PCB lookup.
928 		 *
929 		 * First see if this local port is in use by looking on the
930 		 * port hash list.
931 		 */
932 		porthash = &pcbinfo->ipi_porthashbase[INP_PCBPORTHASH(lport,
933 		    pcbinfo->ipi_porthashmask)];
934 		LIST_FOREACH(phd, porthash, phd_hash) {
935 			if (phd->phd_port == lport)
936 				break;
937 		}
938 		if (phd != NULL) {
939 			/*
940 			 * Port is in use by one or more PCBs. Look for best
941 			 * fit.
942 			 */
943 			LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) {
944 				wildcard = 0;
945 #ifdef INET6
946 				if ((inp->inp_vflag & INP_IPV4) == 0)
947 					continue;
948 				/*
949 				 * We never select the PCB that has
950 				 * INP_IPV6 flag and is bound to :: if
951 				 * we have another PCB which is bound
952 				 * to 0.0.0.0.  If a PCB has the
953 				 * INP_IPV6 flag, then we set its cost
954 				 * higher than IPv4 only PCBs.
955 				 *
956 				 * Note that the case only happens
957 				 * when a socket is bound to ::, under
958 				 * the condition that the use of the
959 				 * mapped address is allowed.
960 				 */
961 				if ((inp->inp_vflag & INP_IPV6) != 0)
962 					wildcard += INP_LOOKUP_MAPPED_PCB_COST;
963 #endif
964 				if (inp->inp_faddr.s_addr != INADDR_ANY)
965 					wildcard++;
966 				if (inp->inp_laddr.s_addr != INADDR_ANY) {
967 					if (laddr.s_addr == INADDR_ANY)
968 						wildcard++;
969 					else if (inp->inp_laddr.s_addr != laddr.s_addr)
970 						continue;
971 				} else {
972 					if (laddr.s_addr != INADDR_ANY)
973 						wildcard++;
974 				}
975 				if (wildcard < matchwild) {
976 					match = inp;
977 					matchwild = wildcard;
978 					if (matchwild == 0) {
979 						break;
980 					}
981 				}
982 			}
983 		}
984 		return (match);
985 	}
986 }
987 #undef INP_LOOKUP_MAPPED_PCB_COST
988 
989 /*
990  * Lookup PCB in hash list.
991  */
992 struct inpcb *
993 in_pcblookup_hash(struct inpcbinfo *pcbinfo, struct in_addr faddr,
994     u_int fport_arg, struct in_addr laddr, u_int lport_arg, int wildcard,
995     struct ifnet *ifp)
996 {
997 	struct inpcbhead *head;
998 	struct inpcb *inp;
999 	u_short fport = fport_arg, lport = lport_arg;
1000 
1001 	INP_INFO_LOCK_ASSERT(pcbinfo);
1002 
1003 	/*
1004 	 * First look for an exact match.
1005 	 */
1006 	head = &pcbinfo->ipi_hashbase[INP_PCBHASH(faddr.s_addr, lport, fport,
1007 	    pcbinfo->ipi_hashmask)];
1008 	LIST_FOREACH(inp, head, inp_hash) {
1009 #ifdef INET6
1010 		if ((inp->inp_vflag & INP_IPV4) == 0)
1011 			continue;
1012 #endif
1013 		if (inp->inp_faddr.s_addr == faddr.s_addr &&
1014 		    inp->inp_laddr.s_addr == laddr.s_addr &&
1015 		    inp->inp_fport == fport &&
1016 		    inp->inp_lport == lport)
1017 			return (inp);
1018 	}
1019 
1020 	/*
1021 	 * Then look for a wildcard match, if requested.
1022 	 */
1023 	if (wildcard) {
1024 		struct inpcb *local_wild = NULL;
1025 #ifdef INET6
1026 		struct inpcb *local_wild_mapped = NULL;
1027 #endif
1028 
1029 		head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport,
1030 		    0, pcbinfo->ipi_hashmask)];
1031 		LIST_FOREACH(inp, head, inp_hash) {
1032 #ifdef INET6
1033 			if ((inp->inp_vflag & INP_IPV4) == 0)
1034 				continue;
1035 #endif
1036 			if (inp->inp_faddr.s_addr == INADDR_ANY &&
1037 			    inp->inp_lport == lport) {
1038 				if (ifp && ifp->if_type == IFT_FAITH &&
1039 				    (inp->inp_flags & INP_FAITH) == 0)
1040 					continue;
1041 				if (inp->inp_laddr.s_addr == laddr.s_addr)
1042 					return (inp);
1043 				else if (inp->inp_laddr.s_addr == INADDR_ANY) {
1044 #ifdef INET6
1045 					if (INP_CHECK_SOCKAF(inp->inp_socket,
1046 							     AF_INET6))
1047 						local_wild_mapped = inp;
1048 					else
1049 #endif
1050 						local_wild = inp;
1051 				}
1052 			}
1053 		}
1054 #ifdef INET6
1055 		if (local_wild == NULL)
1056 			return (local_wild_mapped);
1057 #endif
1058 		return (local_wild);
1059 	}
1060 	return (NULL);
1061 }
1062 
1063 /*
1064  * Insert PCB onto various hash lists.
1065  */
1066 int
1067 in_pcbinshash(struct inpcb *inp)
1068 {
1069 	struct inpcbhead *pcbhash;
1070 	struct inpcbporthead *pcbporthash;
1071 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1072 	struct inpcbport *phd;
1073 	u_int32_t hashkey_faddr;
1074 
1075 	INP_INFO_WLOCK_ASSERT(pcbinfo);
1076 	INP_WLOCK_ASSERT(inp);
1077 
1078 #ifdef INET6
1079 	if (inp->inp_vflag & INP_IPV6)
1080 		hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX */;
1081 	else
1082 #endif /* INET6 */
1083 	hashkey_faddr = inp->inp_faddr.s_addr;
1084 
1085 	pcbhash = &pcbinfo->ipi_hashbase[INP_PCBHASH(hashkey_faddr,
1086 		 inp->inp_lport, inp->inp_fport, pcbinfo->ipi_hashmask)];
1087 
1088 	pcbporthash = &pcbinfo->ipi_porthashbase[
1089 	    INP_PCBPORTHASH(inp->inp_lport, pcbinfo->ipi_porthashmask)];
1090 
1091 	/*
1092 	 * Go through port list and look for a head for this lport.
1093 	 */
1094 	LIST_FOREACH(phd, pcbporthash, phd_hash) {
1095 		if (phd->phd_port == inp->inp_lport)
1096 			break;
1097 	}
1098 	/*
1099 	 * If none exists, malloc one and tack it on.
1100 	 */
1101 	if (phd == NULL) {
1102 		MALLOC(phd, struct inpcbport *, sizeof(struct inpcbport), M_PCB, M_NOWAIT);
1103 		if (phd == NULL) {
1104 			return (ENOBUFS); /* XXX */
1105 		}
1106 		phd->phd_port = inp->inp_lport;
1107 		LIST_INIT(&phd->phd_pcblist);
1108 		LIST_INSERT_HEAD(pcbporthash, phd, phd_hash);
1109 	}
1110 	inp->inp_phd = phd;
1111 	LIST_INSERT_HEAD(&phd->phd_pcblist, inp, inp_portlist);
1112 	LIST_INSERT_HEAD(pcbhash, inp, inp_hash);
1113 	return (0);
1114 }
1115 
1116 /*
1117  * Move PCB to the proper hash bucket when { faddr, fport } have  been
1118  * changed. NOTE: This does not handle the case of the lport changing (the
1119  * hashed port list would have to be updated as well), so the lport must
1120  * not change after in_pcbinshash() has been called.
1121  */
1122 void
1123 in_pcbrehash(struct inpcb *inp)
1124 {
1125 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1126 	struct inpcbhead *head;
1127 	u_int32_t hashkey_faddr;
1128 
1129 	INP_INFO_WLOCK_ASSERT(pcbinfo);
1130 	INP_WLOCK_ASSERT(inp);
1131 
1132 #ifdef INET6
1133 	if (inp->inp_vflag & INP_IPV6)
1134 		hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX */;
1135 	else
1136 #endif /* INET6 */
1137 	hashkey_faddr = inp->inp_faddr.s_addr;
1138 
1139 	head = &pcbinfo->ipi_hashbase[INP_PCBHASH(hashkey_faddr,
1140 		inp->inp_lport, inp->inp_fport, pcbinfo->ipi_hashmask)];
1141 
1142 	LIST_REMOVE(inp, inp_hash);
1143 	LIST_INSERT_HEAD(head, inp, inp_hash);
1144 }
1145 
1146 /*
1147  * Remove PCB from various lists.
1148  */
1149 void
1150 in_pcbremlists(struct inpcb *inp)
1151 {
1152 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1153 
1154 	INP_INFO_WLOCK_ASSERT(pcbinfo);
1155 	INP_WLOCK_ASSERT(inp);
1156 
1157 	inp->inp_gencnt = ++pcbinfo->ipi_gencnt;
1158 	if (inp->inp_lport) {
1159 		struct inpcbport *phd = inp->inp_phd;
1160 
1161 		LIST_REMOVE(inp, inp_hash);
1162 		LIST_REMOVE(inp, inp_portlist);
1163 		if (LIST_FIRST(&phd->phd_pcblist) == NULL) {
1164 			LIST_REMOVE(phd, phd_hash);
1165 			free(phd, M_PCB);
1166 		}
1167 	}
1168 	LIST_REMOVE(inp, inp_list);
1169 	pcbinfo->ipi_count--;
1170 }
1171 
1172 /*
1173  * A set label operation has occurred at the socket layer, propagate the
1174  * label change into the in_pcb for the socket.
1175  */
1176 void
1177 in_pcbsosetlabel(struct socket *so)
1178 {
1179 #ifdef MAC
1180 	struct inpcb *inp;
1181 
1182 	inp = sotoinpcb(so);
1183 	KASSERT(inp != NULL, ("in_pcbsosetlabel: so->so_pcb == NULL"));
1184 
1185 	INP_WLOCK(inp);
1186 	SOCK_LOCK(so);
1187 	mac_inpcb_sosetlabel(so, inp);
1188 	SOCK_UNLOCK(so);
1189 	INP_WUNLOCK(inp);
1190 #endif
1191 }
1192 
1193 /*
1194  * ipport_tick runs once per second, determining if random port allocation
1195  * should be continued.  If more than ipport_randomcps ports have been
1196  * allocated in the last second, then we return to sequential port
1197  * allocation. We return to random allocation only once we drop below
1198  * ipport_randomcps for at least ipport_randomtime seconds.
1199  */
1200 void
1201 ipport_tick(void *xtp)
1202 {
1203 
1204 	if (ipport_tcpallocs <= ipport_tcplastcount + ipport_randomcps) {
1205 		if (ipport_stoprandom > 0)
1206 			ipport_stoprandom--;
1207 	} else
1208 		ipport_stoprandom = ipport_randomtime;
1209 	ipport_tcplastcount = ipport_tcpallocs;
1210 	callout_reset(&ipport_tick_callout, hz, ipport_tick, NULL);
1211 }
1212 
1213 void
1214 inp_wlock(struct inpcb *inp)
1215 {
1216 
1217 	INP_WLOCK(inp);
1218 }
1219 
1220 void
1221 inp_wunlock(struct inpcb *inp)
1222 {
1223 
1224 	INP_WUNLOCK(inp);
1225 }
1226 
1227 void
1228 inp_rlock(struct inpcb *inp)
1229 {
1230 
1231 	INP_RLOCK(inp);
1232 }
1233 
1234 void
1235 inp_runlock(struct inpcb *inp)
1236 {
1237 
1238 	INP_RUNLOCK(inp);
1239 }
1240 
1241 #ifdef INVARIANTS
1242 void
1243 inp_lock_assert(struct inpcb *inp)
1244 {
1245 
1246 	INP_WLOCK_ASSERT(inp);
1247 }
1248 
1249 void
1250 inp_unlock_assert(struct inpcb *inp)
1251 {
1252 
1253 	INP_UNLOCK_ASSERT(inp);
1254 }
1255 #endif
1256 
1257 #ifdef DDB
1258 static void
1259 db_print_indent(int indent)
1260 {
1261 	int i;
1262 
1263 	for (i = 0; i < indent; i++)
1264 		db_printf(" ");
1265 }
1266 
1267 static void
1268 db_print_inconninfo(struct in_conninfo *inc, const char *name, int indent)
1269 {
1270 	char faddr_str[48], laddr_str[48];
1271 
1272 	db_print_indent(indent);
1273 	db_printf("%s at %p\n", name, inc);
1274 
1275 	indent += 2;
1276 
1277 #ifdef INET6
1278 	if (inc->inc_flags == 1) {
1279 		/* IPv6. */
1280 		ip6_sprintf(laddr_str, &inc->inc6_laddr);
1281 		ip6_sprintf(faddr_str, &inc->inc6_faddr);
1282 	} else {
1283 #endif
1284 		/* IPv4. */
1285 		inet_ntoa_r(inc->inc_laddr, laddr_str);
1286 		inet_ntoa_r(inc->inc_faddr, faddr_str);
1287 #ifdef INET6
1288 	}
1289 #endif
1290 	db_print_indent(indent);
1291 	db_printf("inc_laddr %s   inc_lport %u\n", laddr_str,
1292 	    ntohs(inc->inc_lport));
1293 	db_print_indent(indent);
1294 	db_printf("inc_faddr %s   inc_fport %u\n", faddr_str,
1295 	    ntohs(inc->inc_fport));
1296 }
1297 
1298 static void
1299 db_print_inpflags(int inp_flags)
1300 {
1301 	int comma;
1302 
1303 	comma = 0;
1304 	if (inp_flags & INP_RECVOPTS) {
1305 		db_printf("%sINP_RECVOPTS", comma ? ", " : "");
1306 		comma = 1;
1307 	}
1308 	if (inp_flags & INP_RECVRETOPTS) {
1309 		db_printf("%sINP_RECVRETOPTS", comma ? ", " : "");
1310 		comma = 1;
1311 	}
1312 	if (inp_flags & INP_RECVDSTADDR) {
1313 		db_printf("%sINP_RECVDSTADDR", comma ? ", " : "");
1314 		comma = 1;
1315 	}
1316 	if (inp_flags & INP_HDRINCL) {
1317 		db_printf("%sINP_HDRINCL", comma ? ", " : "");
1318 		comma = 1;
1319 	}
1320 	if (inp_flags & INP_HIGHPORT) {
1321 		db_printf("%sINP_HIGHPORT", comma ? ", " : "");
1322 		comma = 1;
1323 	}
1324 	if (inp_flags & INP_LOWPORT) {
1325 		db_printf("%sINP_LOWPORT", comma ? ", " : "");
1326 		comma = 1;
1327 	}
1328 	if (inp_flags & INP_ANONPORT) {
1329 		db_printf("%sINP_ANONPORT", comma ? ", " : "");
1330 		comma = 1;
1331 	}
1332 	if (inp_flags & INP_RECVIF) {
1333 		db_printf("%sINP_RECVIF", comma ? ", " : "");
1334 		comma = 1;
1335 	}
1336 	if (inp_flags & INP_MTUDISC) {
1337 		db_printf("%sINP_MTUDISC", comma ? ", " : "");
1338 		comma = 1;
1339 	}
1340 	if (inp_flags & INP_FAITH) {
1341 		db_printf("%sINP_FAITH", comma ? ", " : "");
1342 		comma = 1;
1343 	}
1344 	if (inp_flags & INP_RECVTTL) {
1345 		db_printf("%sINP_RECVTTL", comma ? ", " : "");
1346 		comma = 1;
1347 	}
1348 	if (inp_flags & INP_DONTFRAG) {
1349 		db_printf("%sINP_DONTFRAG", comma ? ", " : "");
1350 		comma = 1;
1351 	}
1352 	if (inp_flags & IN6P_IPV6_V6ONLY) {
1353 		db_printf("%sIN6P_IPV6_V6ONLY", comma ? ", " : "");
1354 		comma = 1;
1355 	}
1356 	if (inp_flags & IN6P_PKTINFO) {
1357 		db_printf("%sIN6P_PKTINFO", comma ? ", " : "");
1358 		comma = 1;
1359 	}
1360 	if (inp_flags & IN6P_HOPLIMIT) {
1361 		db_printf("%sIN6P_HOPLIMIT", comma ? ", " : "");
1362 		comma = 1;
1363 	}
1364 	if (inp_flags & IN6P_HOPOPTS) {
1365 		db_printf("%sIN6P_HOPOPTS", comma ? ", " : "");
1366 		comma = 1;
1367 	}
1368 	if (inp_flags & IN6P_DSTOPTS) {
1369 		db_printf("%sIN6P_DSTOPTS", comma ? ", " : "");
1370 		comma = 1;
1371 	}
1372 	if (inp_flags & IN6P_RTHDR) {
1373 		db_printf("%sIN6P_RTHDR", comma ? ", " : "");
1374 		comma = 1;
1375 	}
1376 	if (inp_flags & IN6P_RTHDRDSTOPTS) {
1377 		db_printf("%sIN6P_RTHDRDSTOPTS", comma ? ", " : "");
1378 		comma = 1;
1379 	}
1380 	if (inp_flags & IN6P_TCLASS) {
1381 		db_printf("%sIN6P_TCLASS", comma ? ", " : "");
1382 		comma = 1;
1383 	}
1384 	if (inp_flags & IN6P_AUTOFLOWLABEL) {
1385 		db_printf("%sIN6P_AUTOFLOWLABEL", comma ? ", " : "");
1386 		comma = 1;
1387 	}
1388 	if (inp_flags & IN6P_RFC2292) {
1389 		db_printf("%sIN6P_RFC2292", comma ? ", " : "");
1390 		comma = 1;
1391 	}
1392 	if (inp_flags & IN6P_MTU) {
1393 		db_printf("IN6P_MTU%s", comma ? ", " : "");
1394 		comma = 1;
1395 	}
1396 }
1397 
1398 static void
1399 db_print_inpvflag(u_char inp_vflag)
1400 {
1401 	int comma;
1402 
1403 	comma = 0;
1404 	if (inp_vflag & INP_IPV4) {
1405 		db_printf("%sINP_IPV4", comma ? ", " : "");
1406 		comma  = 1;
1407 	}
1408 	if (inp_vflag & INP_IPV6) {
1409 		db_printf("%sINP_IPV6", comma ? ", " : "");
1410 		comma  = 1;
1411 	}
1412 	if (inp_vflag & INP_IPV6PROTO) {
1413 		db_printf("%sINP_IPV6PROTO", comma ? ", " : "");
1414 		comma  = 1;
1415 	}
1416 	if (inp_vflag & INP_TIMEWAIT) {
1417 		db_printf("%sINP_TIMEWAIT", comma ? ", " : "");
1418 		comma  = 1;
1419 	}
1420 	if (inp_vflag & INP_ONESBCAST) {
1421 		db_printf("%sINP_ONESBCAST", comma ? ", " : "");
1422 		comma  = 1;
1423 	}
1424 	if (inp_vflag & INP_DROPPED) {
1425 		db_printf("%sINP_DROPPED", comma ? ", " : "");
1426 		comma  = 1;
1427 	}
1428 	if (inp_vflag & INP_SOCKREF) {
1429 		db_printf("%sINP_SOCKREF", comma ? ", " : "");
1430 		comma  = 1;
1431 	}
1432 }
1433 
1434 void
1435 db_print_inpcb(struct inpcb *inp, const char *name, int indent)
1436 {
1437 
1438 	db_print_indent(indent);
1439 	db_printf("%s at %p\n", name, inp);
1440 
1441 	indent += 2;
1442 
1443 	db_print_indent(indent);
1444 	db_printf("inp_flow: 0x%x\n", inp->inp_flow);
1445 
1446 	db_print_inconninfo(&inp->inp_inc, "inp_conninfo", indent);
1447 
1448 	db_print_indent(indent);
1449 	db_printf("inp_ppcb: %p   inp_pcbinfo: %p   inp_socket: %p\n",
1450 	    inp->inp_ppcb, inp->inp_pcbinfo, inp->inp_socket);
1451 
1452 	db_print_indent(indent);
1453 	db_printf("inp_label: %p   inp_flags: 0x%x (",
1454 	   inp->inp_label, inp->inp_flags);
1455 	db_print_inpflags(inp->inp_flags);
1456 	db_printf(")\n");
1457 
1458 	db_print_indent(indent);
1459 	db_printf("inp_sp: %p   inp_vflag: 0x%x (", inp->inp_sp,
1460 	    inp->inp_vflag);
1461 	db_print_inpvflag(inp->inp_vflag);
1462 	db_printf(")\n");
1463 
1464 	db_print_indent(indent);
1465 	db_printf("inp_ip_ttl: %d   inp_ip_p: %d   inp_ip_minttl: %d\n",
1466 	    inp->inp_ip_ttl, inp->inp_ip_p, inp->inp_ip_minttl);
1467 
1468 	db_print_indent(indent);
1469 #ifdef INET6
1470 	if (inp->inp_vflag & INP_IPV6) {
1471 		db_printf("in6p_options: %p   in6p_outputopts: %p   "
1472 		    "in6p_moptions: %p\n", inp->in6p_options,
1473 		    inp->in6p_outputopts, inp->in6p_moptions);
1474 		db_printf("in6p_icmp6filt: %p   in6p_cksum %d   "
1475 		    "in6p_hops %u\n", inp->in6p_icmp6filt, inp->in6p_cksum,
1476 		    inp->in6p_hops);
1477 	} else
1478 #endif
1479 	{
1480 		db_printf("inp_ip_tos: %d   inp_ip_options: %p   "
1481 		    "inp_ip_moptions: %p\n", inp->inp_ip_tos,
1482 		    inp->inp_options, inp->inp_moptions);
1483 	}
1484 
1485 	db_print_indent(indent);
1486 	db_printf("inp_phd: %p   inp_gencnt: %ju\n", inp->inp_phd,
1487 	    (uintmax_t)inp->inp_gencnt);
1488 }
1489 
1490 DB_SHOW_COMMAND(inpcb, db_show_inpcb)
1491 {
1492 	struct inpcb *inp;
1493 
1494 	if (!have_addr) {
1495 		db_printf("usage: show inpcb <addr>\n");
1496 		return;
1497 	}
1498 	inp = (struct inpcb *)addr;
1499 
1500 	db_print_inpcb(inp, "inpcb", 0);
1501 }
1502 #endif
1503