xref: /freebsd/sys/netinet/in_pcb.c (revision 35a04710d7286aa9538917fd7f8e417dbee95b82)
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_HIFIRSTAUTO;		/* 49152 */
93 int	ipport_lastauto  = IPPORT_HILASTAUTO;		/* 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 		goto out;
202 #endif /*IPSEC*/
203 #ifdef INET6
204 	if (INP_SOCKAF(so) == AF_INET6) {
205 		inp->inp_vflag |= INP_IPV6PROTO;
206 		if (ip6_v6only)
207 			inp->inp_flags |= IN6P_IPV6_V6ONLY;
208 	}
209 #endif
210 	LIST_INSERT_HEAD(pcbinfo->ipi_listhead, inp, inp_list);
211 	pcbinfo->ipi_count++;
212 	so->so_pcb = (caddr_t)inp;
213 #ifdef INET6
214 	if (ip6_auto_flowlabel)
215 		inp->inp_flags |= IN6P_AUTOFLOWLABEL;
216 #endif
217 	INP_LOCK(inp);
218 	inp->inp_gencnt = ++pcbinfo->ipi_gencnt;
219 
220 #if defined(IPSEC) || defined(MAC)
221 out:
222 	if (error != 0)
223 		uma_zfree(pcbinfo->ipi_zone, inp);
224 #endif
225 	return (error);
226 }
227 
228 int
229 in_pcbbind(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred)
230 {
231 	int anonport, error;
232 
233 	INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
234 	INP_LOCK_ASSERT(inp);
235 
236 	if (inp->inp_lport != 0 || inp->inp_laddr.s_addr != INADDR_ANY)
237 		return (EINVAL);
238 	anonport = inp->inp_lport == 0 && (nam == NULL ||
239 	    ((struct sockaddr_in *)nam)->sin_port == 0);
240 	error = in_pcbbind_setup(inp, nam, &inp->inp_laddr.s_addr,
241 	    &inp->inp_lport, cred);
242 	if (error)
243 		return (error);
244 	if (in_pcbinshash(inp) != 0) {
245 		inp->inp_laddr.s_addr = INADDR_ANY;
246 		inp->inp_lport = 0;
247 		return (EAGAIN);
248 	}
249 	if (anonport)
250 		inp->inp_flags |= INP_ANONPORT;
251 	return (0);
252 }
253 
254 /*
255  * Set up a bind operation on a PCB, performing port allocation
256  * as required, but do not actually modify the PCB. Callers can
257  * either complete the bind by setting inp_laddr/inp_lport and
258  * calling in_pcbinshash(), or they can just use the resulting
259  * port and address to authorise the sending of a once-off packet.
260  *
261  * On error, the values of *laddrp and *lportp are not changed.
262  */
263 int
264 in_pcbbind_setup(struct inpcb *inp, struct sockaddr *nam, in_addr_t *laddrp,
265     u_short *lportp, struct ucred *cred)
266 {
267 	struct socket *so = inp->inp_socket;
268 	unsigned short *lastport;
269 	struct sockaddr_in *sin;
270 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
271 	struct in_addr laddr;
272 	u_short lport = 0;
273 	int wild = 0, reuseport = (so->so_options & SO_REUSEPORT);
274 	int error, prison = 0;
275 	int dorandom;
276 
277 	INP_INFO_WLOCK_ASSERT(pcbinfo);
278 	INP_LOCK_ASSERT(inp);
279 
280 	if (TAILQ_EMPTY(&in_ifaddrhead)) /* XXX broken! */
281 		return (EADDRNOTAVAIL);
282 	laddr.s_addr = *laddrp;
283 	if (nam != NULL && laddr.s_addr != INADDR_ANY)
284 		return (EINVAL);
285 	if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0)
286 		wild = INPLOOKUP_WILDCARD;
287 	if (nam) {
288 		sin = (struct sockaddr_in *)nam;
289 		if (nam->sa_len != sizeof (*sin))
290 			return (EINVAL);
291 #ifdef notdef
292 		/*
293 		 * We should check the family, but old programs
294 		 * incorrectly fail to initialize it.
295 		 */
296 		if (sin->sin_family != AF_INET)
297 			return (EAFNOSUPPORT);
298 #endif
299 		if (sin->sin_addr.s_addr != INADDR_ANY)
300 			if (prison_ip(cred, 0, &sin->sin_addr.s_addr))
301 				return(EINVAL);
302 		if (sin->sin_port != *lportp) {
303 			/* Don't allow the port to change. */
304 			if (*lportp != 0)
305 				return (EINVAL);
306 			lport = sin->sin_port;
307 		}
308 		/* NB: lport is left as 0 if the port isn't being changed. */
309 		if (IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) {
310 			/*
311 			 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
312 			 * allow complete duplication of binding if
313 			 * SO_REUSEPORT is set, or if SO_REUSEADDR is set
314 			 * and a multicast address is bound on both
315 			 * new and duplicated sockets.
316 			 */
317 			if (so->so_options & SO_REUSEADDR)
318 				reuseport = SO_REUSEADDR|SO_REUSEPORT;
319 		} else if (sin->sin_addr.s_addr != INADDR_ANY) {
320 			sin->sin_port = 0;		/* yech... */
321 			bzero(&sin->sin_zero, sizeof(sin->sin_zero));
322 			if (ifa_ifwithaddr((struct sockaddr *)sin) == 0)
323 				return (EADDRNOTAVAIL);
324 		}
325 		laddr = sin->sin_addr;
326 		if (lport) {
327 			struct inpcb *t;
328 			struct tcptw *tw;
329 
330 			/* GROSS */
331 			if (ntohs(lport) <= ipport_reservedhigh &&
332 			    ntohs(lport) >= ipport_reservedlow &&
333 			    priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT,
334 			    0))
335 				return (EACCES);
336 			if (jailed(cred))
337 				prison = 1;
338 			if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)) &&
339 			    priv_check_cred(so->so_cred,
340 			    PRIV_NETINET_REUSEPORT, 0) != 0) {
341 				t = in_pcblookup_local(inp->inp_pcbinfo,
342 				    sin->sin_addr, lport,
343 				    prison ? 0 :  INPLOOKUP_WILDCARD);
344 	/*
345 	 * XXX
346 	 * This entire block sorely needs a rewrite.
347 	 */
348 				if (t &&
349 				    ((t->inp_vflag & INP_TIMEWAIT) == 0) &&
350 				    (so->so_type != SOCK_STREAM ||
351 				     ntohl(t->inp_faddr.s_addr) == INADDR_ANY) &&
352 				    (ntohl(sin->sin_addr.s_addr) != INADDR_ANY ||
353 				     ntohl(t->inp_laddr.s_addr) != INADDR_ANY ||
354 				     (t->inp_socket->so_options &
355 					 SO_REUSEPORT) == 0) &&
356 				    (so->so_cred->cr_uid !=
357 				     t->inp_socket->so_cred->cr_uid))
358 					return (EADDRINUSE);
359 			}
360 			if (prison && prison_ip(cred, 0, &sin->sin_addr.s_addr))
361 				return (EADDRNOTAVAIL);
362 			t = in_pcblookup_local(pcbinfo, sin->sin_addr,
363 			    lport, prison ? 0 : wild);
364 			if (t && (t->inp_vflag & INP_TIMEWAIT)) {
365 				/*
366 				 * XXXRW: If an incpb has had its timewait
367 				 * state recycled, we treat the address as
368 				 * being in use (for now).  This is better
369 				 * than a panic, but not desirable.
370 				 */
371 				tw = intotw(inp);
372 				if (tw == NULL ||
373 				    (reuseport & tw->tw_so_options) == 0)
374 					return (EADDRINUSE);
375 			} else if (t &&
376 			    (reuseport & t->inp_socket->so_options) == 0) {
377 #ifdef INET6
378 				if (ntohl(sin->sin_addr.s_addr) !=
379 				    INADDR_ANY ||
380 				    ntohl(t->inp_laddr.s_addr) !=
381 				    INADDR_ANY ||
382 				    INP_SOCKAF(so) ==
383 				    INP_SOCKAF(t->inp_socket))
384 #endif
385 				return (EADDRINUSE);
386 			}
387 		}
388 	}
389 	if (*lportp != 0)
390 		lport = *lportp;
391 	if (lport == 0) {
392 		u_short first, last;
393 		int count;
394 
395 		if (laddr.s_addr != INADDR_ANY)
396 			if (prison_ip(cred, 0, &laddr.s_addr))
397 				return (EINVAL);
398 
399 		if (inp->inp_flags & INP_HIGHPORT) {
400 			first = ipport_hifirstauto;	/* sysctl */
401 			last  = ipport_hilastauto;
402 			lastport = &pcbinfo->ipi_lasthi;
403 		} else if (inp->inp_flags & INP_LOWPORT) {
404 			error = priv_check_cred(cred,
405 			    PRIV_NETINET_RESERVEDPORT, 0);
406 			if (error)
407 				return error;
408 			first = ipport_lowfirstauto;	/* 1023 */
409 			last  = ipport_lowlastauto;	/* 600 */
410 			lastport = &pcbinfo->ipi_lastlow;
411 		} else {
412 			first = ipport_firstauto;	/* sysctl */
413 			last  = ipport_lastauto;
414 			lastport = &pcbinfo->ipi_lastport;
415 		}
416 		/*
417 		 * For UDP, use random port allocation as long as the user
418 		 * allows it.  For TCP (and as of yet unknown) connections,
419 		 * use random port allocation only if the user allows it AND
420 		 * ipport_tick() allows it.
421 		 */
422 		if (ipport_randomized &&
423 			(!ipport_stoprandom || pcbinfo == &udbinfo))
424 			dorandom = 1;
425 		else
426 			dorandom = 0;
427 		/*
428 		 * It makes no sense to do random port allocation if
429 		 * we have the only port available.
430 		 */
431 		if (first == last)
432 			dorandom = 0;
433 		/* Make sure to not include UDP packets in the count. */
434 		if (pcbinfo != &udbinfo)
435 			ipport_tcpallocs++;
436 		/*
437 		 * Simple check to ensure all ports are not used up causing
438 		 * a deadlock here.
439 		 *
440 		 * We split the two cases (up and down) so that the direction
441 		 * is not being tested on each round of the loop.
442 		 */
443 		if (first > last) {
444 			/*
445 			 * counting down
446 			 */
447 			if (dorandom)
448 				*lastport = first -
449 					    (arc4random() % (first - last));
450 			count = first - last;
451 
452 			do {
453 				if (count-- < 0)	/* completely used? */
454 					return (EADDRNOTAVAIL);
455 				--*lastport;
456 				if (*lastport > first || *lastport < last)
457 					*lastport = first;
458 				lport = htons(*lastport);
459 			} while (in_pcblookup_local(pcbinfo, laddr, lport,
460 			    wild));
461 		} else {
462 			/*
463 			 * counting up
464 			 */
465 			if (dorandom)
466 				*lastport = first +
467 					    (arc4random() % (last - first));
468 			count = last - first;
469 
470 			do {
471 				if (count-- < 0)	/* completely used? */
472 					return (EADDRNOTAVAIL);
473 				++*lastport;
474 				if (*lastport < first || *lastport > last)
475 					*lastport = first;
476 				lport = htons(*lastport);
477 			} while (in_pcblookup_local(pcbinfo, laddr, lport,
478 			    wild));
479 		}
480 	}
481 	if (prison_ip(cred, 0, &laddr.s_addr))
482 		return (EINVAL);
483 	*laddrp = laddr.s_addr;
484 	*lportp = lport;
485 	return (0);
486 }
487 
488 /*
489  * Connect from a socket to a specified address.
490  * Both address and port must be specified in argument sin.
491  * If don't have a local address for this socket yet,
492  * then pick one.
493  */
494 int
495 in_pcbconnect(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred)
496 {
497 	u_short lport, fport;
498 	in_addr_t laddr, faddr;
499 	int anonport, error;
500 
501 	INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
502 	INP_LOCK_ASSERT(inp);
503 
504 	lport = inp->inp_lport;
505 	laddr = inp->inp_laddr.s_addr;
506 	anonport = (lport == 0);
507 	error = in_pcbconnect_setup(inp, nam, &laddr, &lport, &faddr, &fport,
508 	    NULL, cred);
509 	if (error)
510 		return (error);
511 
512 	/* Do the initial binding of the local address if required. */
513 	if (inp->inp_laddr.s_addr == INADDR_ANY && inp->inp_lport == 0) {
514 		inp->inp_lport = lport;
515 		inp->inp_laddr.s_addr = laddr;
516 		if (in_pcbinshash(inp) != 0) {
517 			inp->inp_laddr.s_addr = INADDR_ANY;
518 			inp->inp_lport = 0;
519 			return (EAGAIN);
520 		}
521 	}
522 
523 	/* Commit the remaining changes. */
524 	inp->inp_lport = lport;
525 	inp->inp_laddr.s_addr = laddr;
526 	inp->inp_faddr.s_addr = faddr;
527 	inp->inp_fport = fport;
528 	in_pcbrehash(inp);
529 
530 	if (anonport)
531 		inp->inp_flags |= INP_ANONPORT;
532 	return (0);
533 }
534 
535 /*
536  * Set up for a connect from a socket to the specified address.
537  * On entry, *laddrp and *lportp should contain the current local
538  * address and port for the PCB; these are updated to the values
539  * that should be placed in inp_laddr and inp_lport to complete
540  * the connect.
541  *
542  * On success, *faddrp and *fportp will be set to the remote address
543  * and port. These are not updated in the error case.
544  *
545  * If the operation fails because the connection already exists,
546  * *oinpp will be set to the PCB of that connection so that the
547  * caller can decide to override it. In all other cases, *oinpp
548  * is set to NULL.
549  */
550 int
551 in_pcbconnect_setup(struct inpcb *inp, struct sockaddr *nam,
552     in_addr_t *laddrp, u_short *lportp, in_addr_t *faddrp, u_short *fportp,
553     struct inpcb **oinpp, struct ucred *cred)
554 {
555 	struct sockaddr_in *sin = (struct sockaddr_in *)nam;
556 	struct in_ifaddr *ia;
557 	struct sockaddr_in sa;
558 	struct ucred *socred;
559 	struct inpcb *oinp;
560 	struct in_addr laddr, faddr;
561 	u_short lport, fport;
562 	int error;
563 
564 	INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
565 	INP_LOCK_ASSERT(inp);
566 
567 	if (oinpp != NULL)
568 		*oinpp = NULL;
569 	if (nam->sa_len != sizeof (*sin))
570 		return (EINVAL);
571 	if (sin->sin_family != AF_INET)
572 		return (EAFNOSUPPORT);
573 	if (sin->sin_port == 0)
574 		return (EADDRNOTAVAIL);
575 	laddr.s_addr = *laddrp;
576 	lport = *lportp;
577 	faddr = sin->sin_addr;
578 	fport = sin->sin_port;
579 	socred = inp->inp_socket->so_cred;
580 	if (laddr.s_addr == INADDR_ANY && jailed(socred)) {
581 		bzero(&sa, sizeof(sa));
582 		sa.sin_addr.s_addr = htonl(prison_getip(socred));
583 		sa.sin_len = sizeof(sa);
584 		sa.sin_family = AF_INET;
585 		error = in_pcbbind_setup(inp, (struct sockaddr *)&sa,
586 		    &laddr.s_addr, &lport, cred);
587 		if (error)
588 			return (error);
589 	}
590 	if (!TAILQ_EMPTY(&in_ifaddrhead)) {
591 		/*
592 		 * If the destination address is INADDR_ANY,
593 		 * use the primary local address.
594 		 * If the supplied address is INADDR_BROADCAST,
595 		 * and the primary interface supports broadcast,
596 		 * choose the broadcast address for that interface.
597 		 */
598 		if (faddr.s_addr == INADDR_ANY)
599 			faddr = IA_SIN(TAILQ_FIRST(&in_ifaddrhead))->sin_addr;
600 		else if (faddr.s_addr == (u_long)INADDR_BROADCAST &&
601 		    (TAILQ_FIRST(&in_ifaddrhead)->ia_ifp->if_flags &
602 		    IFF_BROADCAST))
603 			faddr = satosin(&TAILQ_FIRST(
604 			    &in_ifaddrhead)->ia_broadaddr)->sin_addr;
605 	}
606 	if (laddr.s_addr == INADDR_ANY) {
607 		ia = (struct in_ifaddr *)0;
608 		/*
609 		 * If route is known our src addr is taken from the i/f,
610 		 * else punt.
611 		 *
612 		 * Find out route to destination
613 		 */
614 		if ((inp->inp_socket->so_options & SO_DONTROUTE) == 0)
615 			ia = ip_rtaddr(faddr);
616 		/*
617 		 * If we found a route, use the address corresponding to
618 		 * the outgoing interface.
619 		 *
620 		 * Otherwise assume faddr is reachable on a directly connected
621 		 * network and try to find a corresponding interface to take
622 		 * the source address from.
623 		 */
624 		if (ia == 0) {
625 			bzero(&sa, sizeof(sa));
626 			sa.sin_addr = faddr;
627 			sa.sin_len = sizeof(sa);
628 			sa.sin_family = AF_INET;
629 
630 			ia = ifatoia(ifa_ifwithdstaddr(sintosa(&sa)));
631 			if (ia == 0)
632 				ia = ifatoia(ifa_ifwithnet(sintosa(&sa)));
633 			if (ia == 0)
634 				return (ENETUNREACH);
635 		}
636 		/*
637 		 * If the destination address is multicast and an outgoing
638 		 * interface has been set as a multicast option, use the
639 		 * address of that interface as our source address.
640 		 */
641 		if (IN_MULTICAST(ntohl(faddr.s_addr)) &&
642 		    inp->inp_moptions != NULL) {
643 			struct ip_moptions *imo;
644 			struct ifnet *ifp;
645 
646 			imo = inp->inp_moptions;
647 			if (imo->imo_multicast_ifp != NULL) {
648 				ifp = imo->imo_multicast_ifp;
649 				TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link)
650 					if (ia->ia_ifp == ifp)
651 						break;
652 				if (ia == 0)
653 					return (EADDRNOTAVAIL);
654 			}
655 		}
656 		laddr = ia->ia_addr.sin_addr;
657 	}
658 
659 	oinp = in_pcblookup_hash(inp->inp_pcbinfo, faddr, fport, laddr, lport,
660 	    0, NULL);
661 	if (oinp != NULL) {
662 		if (oinpp != NULL)
663 			*oinpp = oinp;
664 		return (EADDRINUSE);
665 	}
666 	if (lport == 0) {
667 		error = in_pcbbind_setup(inp, NULL, &laddr.s_addr, &lport,
668 		    cred);
669 		if (error)
670 			return (error);
671 	}
672 	*laddrp = laddr.s_addr;
673 	*lportp = lport;
674 	*faddrp = faddr.s_addr;
675 	*fportp = fport;
676 	return (0);
677 }
678 
679 void
680 in_pcbdisconnect(struct inpcb *inp)
681 {
682 
683 	INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
684 	INP_LOCK_ASSERT(inp);
685 
686 	inp->inp_faddr.s_addr = INADDR_ANY;
687 	inp->inp_fport = 0;
688 	in_pcbrehash(inp);
689 }
690 
691 /*
692  * In the old world order, in_pcbdetach() served two functions: to detach the
693  * pcb from the socket/potentially free the socket, and to free the pcb
694  * itself.  In the new world order, the protocol code is responsible for
695  * managing the relationship with the socket, and this code simply frees the
696  * pcb.
697  */
698 void
699 in_pcbdetach(struct inpcb *inp)
700 {
701 
702 	KASSERT(inp->inp_socket != NULL, ("in_pcbdetach: inp_socket == NULL"));
703 	inp->inp_socket->so_pcb = NULL;
704 	inp->inp_socket = NULL;
705 }
706 
707 void
708 in_pcbfree(struct inpcb *inp)
709 {
710 	struct inpcbinfo *ipi = inp->inp_pcbinfo;
711 
712 	KASSERT(inp->inp_socket == NULL, ("in_pcbfree: inp_socket != NULL"));
713 	INP_INFO_WLOCK_ASSERT(ipi);
714 	INP_LOCK_ASSERT(inp);
715 
716 #ifdef IPSEC
717 	ipsec4_delete_pcbpolicy(inp);
718 #endif /*IPSEC*/
719 	inp->inp_gencnt = ++ipi->ipi_gencnt;
720 	in_pcbremlists(inp);
721 	if (inp->inp_options)
722 		(void)m_free(inp->inp_options);
723 	if (inp->inp_moptions != NULL)
724 		inp_freemoptions(inp->inp_moptions);
725 	inp->inp_vflag = 0;
726 
727 #ifdef MAC
728 	mac_inpcb_destroy(inp);
729 #endif
730 	INP_UNLOCK(inp);
731 	uma_zfree(ipi->ipi_zone, inp);
732 }
733 
734 /*
735  * TCP needs to maintain its inpcb structure after the TCP connection has
736  * been torn down.  However, it must be disconnected from the inpcb hashes as
737  * it must not prevent binding of future connections to the same port/ip
738  * combination by other inpcbs.
739  */
740 void
741 in_pcbdrop(struct inpcb *inp)
742 {
743 
744 	INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
745 	INP_LOCK_ASSERT(inp);
746 
747 	inp->inp_vflag |= INP_DROPPED;
748 	if (inp->inp_lport) {
749 		struct inpcbport *phd = inp->inp_phd;
750 
751 		LIST_REMOVE(inp, inp_hash);
752 		LIST_REMOVE(inp, inp_portlist);
753 		if (LIST_FIRST(&phd->phd_pcblist) == NULL) {
754 			LIST_REMOVE(phd, phd_hash);
755 			free(phd, M_PCB);
756 		}
757 		inp->inp_lport = 0;
758 	}
759 }
760 
761 /*
762  * Common routines to return the socket addresses associated with inpcbs.
763  */
764 struct sockaddr *
765 in_sockaddr(in_port_t port, struct in_addr *addr_p)
766 {
767 	struct sockaddr_in *sin;
768 
769 	MALLOC(sin, struct sockaddr_in *, sizeof *sin, M_SONAME,
770 		M_WAITOK | M_ZERO);
771 	sin->sin_family = AF_INET;
772 	sin->sin_len = sizeof(*sin);
773 	sin->sin_addr = *addr_p;
774 	sin->sin_port = port;
775 
776 	return (struct sockaddr *)sin;
777 }
778 
779 int
780 in_getsockaddr(struct socket *so, struct sockaddr **nam)
781 {
782 	struct inpcb *inp;
783 	struct in_addr addr;
784 	in_port_t port;
785 
786 	inp = sotoinpcb(so);
787 	KASSERT(inp != NULL, ("in_getsockaddr: inp == NULL"));
788 
789 	INP_LOCK(inp);
790 	port = inp->inp_lport;
791 	addr = inp->inp_laddr;
792 	INP_UNLOCK(inp);
793 
794 	*nam = in_sockaddr(port, &addr);
795 	return 0;
796 }
797 
798 int
799 in_getpeeraddr(struct socket *so, struct sockaddr **nam)
800 {
801 	struct inpcb *inp;
802 	struct in_addr addr;
803 	in_port_t port;
804 
805 	inp = sotoinpcb(so);
806 	KASSERT(inp != NULL, ("in_getpeeraddr: inp == NULL"));
807 
808 	INP_LOCK(inp);
809 	port = inp->inp_fport;
810 	addr = inp->inp_faddr;
811 	INP_UNLOCK(inp);
812 
813 	*nam = in_sockaddr(port, &addr);
814 	return 0;
815 }
816 
817 void
818 in_pcbnotifyall(struct inpcbinfo *pcbinfo, struct in_addr faddr, int errno,
819     struct inpcb *(*notify)(struct inpcb *, int))
820 {
821 	struct inpcb *inp, *ninp;
822 	struct inpcbhead *head;
823 
824 	INP_INFO_WLOCK(pcbinfo);
825 	head = pcbinfo->ipi_listhead;
826 	for (inp = LIST_FIRST(head); inp != NULL; inp = ninp) {
827 		INP_LOCK(inp);
828 		ninp = LIST_NEXT(inp, inp_list);
829 #ifdef INET6
830 		if ((inp->inp_vflag & INP_IPV4) == 0) {
831 			INP_UNLOCK(inp);
832 			continue;
833 		}
834 #endif
835 		if (inp->inp_faddr.s_addr != faddr.s_addr ||
836 		    inp->inp_socket == NULL) {
837 			INP_UNLOCK(inp);
838 			continue;
839 		}
840 		if ((*notify)(inp, errno))
841 			INP_UNLOCK(inp);
842 	}
843 	INP_INFO_WUNLOCK(pcbinfo);
844 }
845 
846 void
847 in_pcbpurgeif0(struct inpcbinfo *pcbinfo, struct ifnet *ifp)
848 {
849 	struct inpcb *inp;
850 	struct ip_moptions *imo;
851 	int i, gap;
852 
853 	INP_INFO_RLOCK(pcbinfo);
854 	LIST_FOREACH(inp, pcbinfo->ipi_listhead, inp_list) {
855 		INP_LOCK(inp);
856 		imo = inp->inp_moptions;
857 		if ((inp->inp_vflag & INP_IPV4) &&
858 		    imo != NULL) {
859 			/*
860 			 * Unselect the outgoing interface if it is being
861 			 * detached.
862 			 */
863 			if (imo->imo_multicast_ifp == ifp)
864 				imo->imo_multicast_ifp = NULL;
865 
866 			/*
867 			 * Drop multicast group membership if we joined
868 			 * through the interface being detached.
869 			 */
870 			for (i = 0, gap = 0; i < imo->imo_num_memberships;
871 			    i++) {
872 				if (imo->imo_membership[i]->inm_ifp == ifp) {
873 					in_delmulti(imo->imo_membership[i]);
874 					gap++;
875 				} else if (gap != 0)
876 					imo->imo_membership[i - gap] =
877 					    imo->imo_membership[i];
878 			}
879 			imo->imo_num_memberships -= gap;
880 		}
881 		INP_UNLOCK(inp);
882 	}
883 	INP_INFO_RUNLOCK(pcbinfo);
884 }
885 
886 /*
887  * Lookup a PCB based on the local address and port.
888  */
889 #define INP_LOOKUP_MAPPED_PCB_COST	3
890 struct inpcb *
891 in_pcblookup_local(struct inpcbinfo *pcbinfo, struct in_addr laddr,
892     u_int lport_arg, int wild_okay)
893 {
894 	struct inpcb *inp;
895 #ifdef INET6
896 	int matchwild = 3 + INP_LOOKUP_MAPPED_PCB_COST;
897 #else
898 	int matchwild = 3;
899 #endif
900 	int wildcard;
901 	u_short lport = lport_arg;
902 
903 	INP_INFO_WLOCK_ASSERT(pcbinfo);
904 
905 	if (!wild_okay) {
906 		struct inpcbhead *head;
907 		/*
908 		 * Look for an unconnected (wildcard foreign addr) PCB that
909 		 * matches the local address and port we're looking for.
910 		 */
911 		head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport,
912 		    0, pcbinfo->ipi_hashmask)];
913 		LIST_FOREACH(inp, head, inp_hash) {
914 #ifdef INET6
915 			if ((inp->inp_vflag & INP_IPV4) == 0)
916 				continue;
917 #endif
918 			if (inp->inp_faddr.s_addr == INADDR_ANY &&
919 			    inp->inp_laddr.s_addr == laddr.s_addr &&
920 			    inp->inp_lport == lport) {
921 				/*
922 				 * Found.
923 				 */
924 				return (inp);
925 			}
926 		}
927 		/*
928 		 * Not found.
929 		 */
930 		return (NULL);
931 	} else {
932 		struct inpcbporthead *porthash;
933 		struct inpcbport *phd;
934 		struct inpcb *match = NULL;
935 		/*
936 		 * Best fit PCB lookup.
937 		 *
938 		 * First see if this local port is in use by looking on the
939 		 * port hash list.
940 		 */
941 		porthash = &pcbinfo->ipi_porthashbase[INP_PCBPORTHASH(lport,
942 		    pcbinfo->ipi_porthashmask)];
943 		LIST_FOREACH(phd, porthash, phd_hash) {
944 			if (phd->phd_port == lport)
945 				break;
946 		}
947 		if (phd != NULL) {
948 			/*
949 			 * Port is in use by one or more PCBs. Look for best
950 			 * fit.
951 			 */
952 			LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) {
953 				wildcard = 0;
954 #ifdef INET6
955 				if ((inp->inp_vflag & INP_IPV4) == 0)
956 					continue;
957 				/*
958 				 * We never select the PCB that has
959 				 * INP_IPV6 flag and is bound to :: if
960 				 * we have another PCB which is bound
961 				 * to 0.0.0.0.  If a PCB has the
962 				 * INP_IPV6 flag, then we set its cost
963 				 * higher than IPv4 only PCBs.
964 				 *
965 				 * Note that the case only happens
966 				 * when a socket is bound to ::, under
967 				 * the condition that the use of the
968 				 * mapped address is allowed.
969 				 */
970 				if ((inp->inp_vflag & INP_IPV6) != 0)
971 					wildcard += INP_LOOKUP_MAPPED_PCB_COST;
972 #endif
973 				if (inp->inp_faddr.s_addr != INADDR_ANY)
974 					wildcard++;
975 				if (inp->inp_laddr.s_addr != INADDR_ANY) {
976 					if (laddr.s_addr == INADDR_ANY)
977 						wildcard++;
978 					else if (inp->inp_laddr.s_addr != laddr.s_addr)
979 						continue;
980 				} else {
981 					if (laddr.s_addr != INADDR_ANY)
982 						wildcard++;
983 				}
984 				if (wildcard < matchwild) {
985 					match = inp;
986 					matchwild = wildcard;
987 					if (matchwild == 0) {
988 						break;
989 					}
990 				}
991 			}
992 		}
993 		return (match);
994 	}
995 }
996 #undef INP_LOOKUP_MAPPED_PCB_COST
997 
998 /*
999  * Lookup PCB in hash list.
1000  */
1001 struct inpcb *
1002 in_pcblookup_hash(struct inpcbinfo *pcbinfo, struct in_addr faddr,
1003     u_int fport_arg, struct in_addr laddr, u_int lport_arg, int wildcard,
1004     struct ifnet *ifp)
1005 {
1006 	struct inpcbhead *head;
1007 	struct inpcb *inp;
1008 	u_short fport = fport_arg, lport = lport_arg;
1009 
1010 	INP_INFO_RLOCK_ASSERT(pcbinfo);
1011 
1012 	/*
1013 	 * First look for an exact match.
1014 	 */
1015 	head = &pcbinfo->ipi_hashbase[INP_PCBHASH(faddr.s_addr, lport, fport,
1016 	    pcbinfo->ipi_hashmask)];
1017 	LIST_FOREACH(inp, head, inp_hash) {
1018 #ifdef INET6
1019 		if ((inp->inp_vflag & INP_IPV4) == 0)
1020 			continue;
1021 #endif
1022 		if (inp->inp_faddr.s_addr == faddr.s_addr &&
1023 		    inp->inp_laddr.s_addr == laddr.s_addr &&
1024 		    inp->inp_fport == fport &&
1025 		    inp->inp_lport == lport)
1026 			return (inp);
1027 	}
1028 
1029 	/*
1030 	 * Then look for a wildcard match, if requested.
1031 	 */
1032 	if (wildcard) {
1033 		struct inpcb *local_wild = NULL;
1034 #ifdef INET6
1035 		struct inpcb *local_wild_mapped = NULL;
1036 #endif
1037 
1038 		head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport,
1039 		    0, pcbinfo->ipi_hashmask)];
1040 		LIST_FOREACH(inp, head, inp_hash) {
1041 #ifdef INET6
1042 			if ((inp->inp_vflag & INP_IPV4) == 0)
1043 				continue;
1044 #endif
1045 			if (inp->inp_faddr.s_addr == INADDR_ANY &&
1046 			    inp->inp_lport == lport) {
1047 				if (ifp && ifp->if_type == IFT_FAITH &&
1048 				    (inp->inp_flags & INP_FAITH) == 0)
1049 					continue;
1050 				if (inp->inp_laddr.s_addr == laddr.s_addr)
1051 					return (inp);
1052 				else if (inp->inp_laddr.s_addr == INADDR_ANY) {
1053 #ifdef INET6
1054 					if (INP_CHECK_SOCKAF(inp->inp_socket,
1055 							     AF_INET6))
1056 						local_wild_mapped = inp;
1057 					else
1058 #endif
1059 						local_wild = inp;
1060 				}
1061 			}
1062 		}
1063 #ifdef INET6
1064 		if (local_wild == NULL)
1065 			return (local_wild_mapped);
1066 #endif
1067 		return (local_wild);
1068 	}
1069 	return (NULL);
1070 }
1071 
1072 /*
1073  * Insert PCB onto various hash lists.
1074  */
1075 int
1076 in_pcbinshash(struct inpcb *inp)
1077 {
1078 	struct inpcbhead *pcbhash;
1079 	struct inpcbporthead *pcbporthash;
1080 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1081 	struct inpcbport *phd;
1082 	u_int32_t hashkey_faddr;
1083 
1084 	INP_INFO_WLOCK_ASSERT(pcbinfo);
1085 	INP_LOCK_ASSERT(inp);
1086 
1087 #ifdef INET6
1088 	if (inp->inp_vflag & INP_IPV6)
1089 		hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX */;
1090 	else
1091 #endif /* INET6 */
1092 	hashkey_faddr = inp->inp_faddr.s_addr;
1093 
1094 	pcbhash = &pcbinfo->ipi_hashbase[INP_PCBHASH(hashkey_faddr,
1095 		 inp->inp_lport, inp->inp_fport, pcbinfo->ipi_hashmask)];
1096 
1097 	pcbporthash = &pcbinfo->ipi_porthashbase[
1098 	    INP_PCBPORTHASH(inp->inp_lport, pcbinfo->ipi_porthashmask)];
1099 
1100 	/*
1101 	 * Go through port list and look for a head for this lport.
1102 	 */
1103 	LIST_FOREACH(phd, pcbporthash, phd_hash) {
1104 		if (phd->phd_port == inp->inp_lport)
1105 			break;
1106 	}
1107 	/*
1108 	 * If none exists, malloc one and tack it on.
1109 	 */
1110 	if (phd == NULL) {
1111 		MALLOC(phd, struct inpcbport *, sizeof(struct inpcbport), M_PCB, M_NOWAIT);
1112 		if (phd == NULL) {
1113 			return (ENOBUFS); /* XXX */
1114 		}
1115 		phd->phd_port = inp->inp_lport;
1116 		LIST_INIT(&phd->phd_pcblist);
1117 		LIST_INSERT_HEAD(pcbporthash, phd, phd_hash);
1118 	}
1119 	inp->inp_phd = phd;
1120 	LIST_INSERT_HEAD(&phd->phd_pcblist, inp, inp_portlist);
1121 	LIST_INSERT_HEAD(pcbhash, inp, inp_hash);
1122 	return (0);
1123 }
1124 
1125 /*
1126  * Move PCB to the proper hash bucket when { faddr, fport } have  been
1127  * changed. NOTE: This does not handle the case of the lport changing (the
1128  * hashed port list would have to be updated as well), so the lport must
1129  * not change after in_pcbinshash() has been called.
1130  */
1131 void
1132 in_pcbrehash(struct inpcb *inp)
1133 {
1134 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1135 	struct inpcbhead *head;
1136 	u_int32_t hashkey_faddr;
1137 
1138 	INP_INFO_WLOCK_ASSERT(pcbinfo);
1139 	INP_LOCK_ASSERT(inp);
1140 
1141 #ifdef INET6
1142 	if (inp->inp_vflag & INP_IPV6)
1143 		hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX */;
1144 	else
1145 #endif /* INET6 */
1146 	hashkey_faddr = inp->inp_faddr.s_addr;
1147 
1148 	head = &pcbinfo->ipi_hashbase[INP_PCBHASH(hashkey_faddr,
1149 		inp->inp_lport, inp->inp_fport, pcbinfo->ipi_hashmask)];
1150 
1151 	LIST_REMOVE(inp, inp_hash);
1152 	LIST_INSERT_HEAD(head, inp, inp_hash);
1153 }
1154 
1155 /*
1156  * Remove PCB from various lists.
1157  */
1158 void
1159 in_pcbremlists(struct inpcb *inp)
1160 {
1161 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1162 
1163 	INP_INFO_WLOCK_ASSERT(pcbinfo);
1164 	INP_LOCK_ASSERT(inp);
1165 
1166 	inp->inp_gencnt = ++pcbinfo->ipi_gencnt;
1167 	if (inp->inp_lport) {
1168 		struct inpcbport *phd = inp->inp_phd;
1169 
1170 		LIST_REMOVE(inp, inp_hash);
1171 		LIST_REMOVE(inp, inp_portlist);
1172 		if (LIST_FIRST(&phd->phd_pcblist) == NULL) {
1173 			LIST_REMOVE(phd, phd_hash);
1174 			free(phd, M_PCB);
1175 		}
1176 	}
1177 	LIST_REMOVE(inp, inp_list);
1178 	pcbinfo->ipi_count--;
1179 }
1180 
1181 /*
1182  * A set label operation has occurred at the socket layer, propagate the
1183  * label change into the in_pcb for the socket.
1184  */
1185 void
1186 in_pcbsosetlabel(struct socket *so)
1187 {
1188 #ifdef MAC
1189 	struct inpcb *inp;
1190 
1191 	inp = sotoinpcb(so);
1192 	KASSERT(inp != NULL, ("in_pcbsosetlabel: so->so_pcb == NULL"));
1193 
1194 	INP_LOCK(inp);
1195 	SOCK_LOCK(so);
1196 	mac_inpcb_sosetlabel(so, inp);
1197 	SOCK_UNLOCK(so);
1198 	INP_UNLOCK(inp);
1199 #endif
1200 }
1201 
1202 /*
1203  * ipport_tick runs once per second, determining if random port allocation
1204  * should be continued.  If more than ipport_randomcps ports have been
1205  * allocated in the last second, then we return to sequential port
1206  * allocation. We return to random allocation only once we drop below
1207  * ipport_randomcps for at least ipport_randomtime seconds.
1208  */
1209 void
1210 ipport_tick(void *xtp)
1211 {
1212 
1213 	if (ipport_tcpallocs <= ipport_tcplastcount + ipport_randomcps) {
1214 		if (ipport_stoprandom > 0)
1215 			ipport_stoprandom--;
1216 	} else
1217 		ipport_stoprandom = ipport_randomtime;
1218 	ipport_tcplastcount = ipport_tcpallocs;
1219 	callout_reset(&ipport_tick_callout, hz, ipport_tick, NULL);
1220 }
1221 
1222 #ifdef DDB
1223 static void
1224 db_print_indent(int indent)
1225 {
1226 	int i;
1227 
1228 	for (i = 0; i < indent; i++)
1229 		db_printf(" ");
1230 }
1231 
1232 static void
1233 db_print_inconninfo(struct in_conninfo *inc, const char *name, int indent)
1234 {
1235 	char faddr_str[48], laddr_str[48];
1236 
1237 	db_print_indent(indent);
1238 	db_printf("%s at %p\n", name, inc);
1239 
1240 	indent += 2;
1241 
1242 #ifdef INET6
1243 	if (inc->inc_flags == 1) {
1244 		/* IPv6. */
1245 		ip6_sprintf(laddr_str, &inc->inc6_laddr);
1246 		ip6_sprintf(faddr_str, &inc->inc6_faddr);
1247 	} else {
1248 #endif
1249 		/* IPv4. */
1250 		inet_ntoa_r(inc->inc_laddr, laddr_str);
1251 		inet_ntoa_r(inc->inc_faddr, faddr_str);
1252 #ifdef INET6
1253 	}
1254 #endif
1255 	db_print_indent(indent);
1256 	db_printf("inc_laddr %s   inc_lport %u\n", laddr_str,
1257 	    ntohs(inc->inc_lport));
1258 	db_print_indent(indent);
1259 	db_printf("inc_faddr %s   inc_fport %u\n", faddr_str,
1260 	    ntohs(inc->inc_fport));
1261 }
1262 
1263 static void
1264 db_print_inpflags(int inp_flags)
1265 {
1266 	int comma;
1267 
1268 	comma = 0;
1269 	if (inp_flags & INP_RECVOPTS) {
1270 		db_printf("%sINP_RECVOPTS", comma ? ", " : "");
1271 		comma = 1;
1272 	}
1273 	if (inp_flags & INP_RECVRETOPTS) {
1274 		db_printf("%sINP_RECVRETOPTS", comma ? ", " : "");
1275 		comma = 1;
1276 	}
1277 	if (inp_flags & INP_RECVDSTADDR) {
1278 		db_printf("%sINP_RECVDSTADDR", comma ? ", " : "");
1279 		comma = 1;
1280 	}
1281 	if (inp_flags & INP_HDRINCL) {
1282 		db_printf("%sINP_HDRINCL", comma ? ", " : "");
1283 		comma = 1;
1284 	}
1285 	if (inp_flags & INP_HIGHPORT) {
1286 		db_printf("%sINP_HIGHPORT", comma ? ", " : "");
1287 		comma = 1;
1288 	}
1289 	if (inp_flags & INP_LOWPORT) {
1290 		db_printf("%sINP_LOWPORT", comma ? ", " : "");
1291 		comma = 1;
1292 	}
1293 	if (inp_flags & INP_ANONPORT) {
1294 		db_printf("%sINP_ANONPORT", comma ? ", " : "");
1295 		comma = 1;
1296 	}
1297 	if (inp_flags & INP_RECVIF) {
1298 		db_printf("%sINP_RECVIF", comma ? ", " : "");
1299 		comma = 1;
1300 	}
1301 	if (inp_flags & INP_MTUDISC) {
1302 		db_printf("%sINP_MTUDISC", comma ? ", " : "");
1303 		comma = 1;
1304 	}
1305 	if (inp_flags & INP_FAITH) {
1306 		db_printf("%sINP_FAITH", comma ? ", " : "");
1307 		comma = 1;
1308 	}
1309 	if (inp_flags & INP_RECVTTL) {
1310 		db_printf("%sINP_RECVTTL", comma ? ", " : "");
1311 		comma = 1;
1312 	}
1313 	if (inp_flags & INP_DONTFRAG) {
1314 		db_printf("%sINP_DONTFRAG", comma ? ", " : "");
1315 		comma = 1;
1316 	}
1317 	if (inp_flags & IN6P_IPV6_V6ONLY) {
1318 		db_printf("%sIN6P_IPV6_V6ONLY", comma ? ", " : "");
1319 		comma = 1;
1320 	}
1321 	if (inp_flags & IN6P_PKTINFO) {
1322 		db_printf("%sIN6P_PKTINFO", comma ? ", " : "");
1323 		comma = 1;
1324 	}
1325 	if (inp_flags & IN6P_HOPLIMIT) {
1326 		db_printf("%sIN6P_HOPLIMIT", comma ? ", " : "");
1327 		comma = 1;
1328 	}
1329 	if (inp_flags & IN6P_HOPOPTS) {
1330 		db_printf("%sIN6P_HOPOPTS", comma ? ", " : "");
1331 		comma = 1;
1332 	}
1333 	if (inp_flags & IN6P_DSTOPTS) {
1334 		db_printf("%sIN6P_DSTOPTS", comma ? ", " : "");
1335 		comma = 1;
1336 	}
1337 	if (inp_flags & IN6P_RTHDR) {
1338 		db_printf("%sIN6P_RTHDR", comma ? ", " : "");
1339 		comma = 1;
1340 	}
1341 	if (inp_flags & IN6P_RTHDRDSTOPTS) {
1342 		db_printf("%sIN6P_RTHDRDSTOPTS", comma ? ", " : "");
1343 		comma = 1;
1344 	}
1345 	if (inp_flags & IN6P_TCLASS) {
1346 		db_printf("%sIN6P_TCLASS", comma ? ", " : "");
1347 		comma = 1;
1348 	}
1349 	if (inp_flags & IN6P_AUTOFLOWLABEL) {
1350 		db_printf("%sIN6P_AUTOFLOWLABEL", comma ? ", " : "");
1351 		comma = 1;
1352 	}
1353 	if (inp_flags & IN6P_RFC2292) {
1354 		db_printf("%sIN6P_RFC2292", comma ? ", " : "");
1355 		comma = 1;
1356 	}
1357 	if (inp_flags & IN6P_MTU) {
1358 		db_printf("IN6P_MTU%s", comma ? ", " : "");
1359 		comma = 1;
1360 	}
1361 }
1362 
1363 static void
1364 db_print_inpvflag(u_char inp_vflag)
1365 {
1366 	int comma;
1367 
1368 	comma = 0;
1369 	if (inp_vflag & INP_IPV4) {
1370 		db_printf("%sINP_IPV4", comma ? ", " : "");
1371 		comma  = 1;
1372 	}
1373 	if (inp_vflag & INP_IPV6) {
1374 		db_printf("%sINP_IPV6", comma ? ", " : "");
1375 		comma  = 1;
1376 	}
1377 	if (inp_vflag & INP_IPV6PROTO) {
1378 		db_printf("%sINP_IPV6PROTO", comma ? ", " : "");
1379 		comma  = 1;
1380 	}
1381 	if (inp_vflag & INP_TIMEWAIT) {
1382 		db_printf("%sINP_TIMEWAIT", comma ? ", " : "");
1383 		comma  = 1;
1384 	}
1385 	if (inp_vflag & INP_ONESBCAST) {
1386 		db_printf("%sINP_ONESBCAST", comma ? ", " : "");
1387 		comma  = 1;
1388 	}
1389 	if (inp_vflag & INP_DROPPED) {
1390 		db_printf("%sINP_DROPPED", comma ? ", " : "");
1391 		comma  = 1;
1392 	}
1393 	if (inp_vflag & INP_SOCKREF) {
1394 		db_printf("%sINP_SOCKREF", comma ? ", " : "");
1395 		comma  = 1;
1396 	}
1397 }
1398 
1399 void
1400 db_print_inpcb(struct inpcb *inp, const char *name, int indent)
1401 {
1402 
1403 	db_print_indent(indent);
1404 	db_printf("%s at %p\n", name, inp);
1405 
1406 	indent += 2;
1407 
1408 	db_print_indent(indent);
1409 	db_printf("inp_flow: 0x%x\n", inp->inp_flow);
1410 
1411 	db_print_inconninfo(&inp->inp_inc, "inp_conninfo", indent);
1412 
1413 	db_print_indent(indent);
1414 	db_printf("inp_ppcb: %p   inp_pcbinfo: %p   inp_socket: %p\n",
1415 	    inp->inp_ppcb, inp->inp_pcbinfo, inp->inp_socket);
1416 
1417 	db_print_indent(indent);
1418 	db_printf("inp_label: %p   inp_flags: 0x%x (",
1419 	   inp->inp_label, inp->inp_flags);
1420 	db_print_inpflags(inp->inp_flags);
1421 	db_printf(")\n");
1422 
1423 	db_print_indent(indent);
1424 	db_printf("inp_sp: %p   inp_vflag: 0x%x (", inp->inp_sp,
1425 	    inp->inp_vflag);
1426 	db_print_inpvflag(inp->inp_vflag);
1427 	db_printf(")\n");
1428 
1429 	db_print_indent(indent);
1430 	db_printf("inp_ip_ttl: %d   inp_ip_p: %d   inp_ip_minttl: %d\n",
1431 	    inp->inp_ip_ttl, inp->inp_ip_p, inp->inp_ip_minttl);
1432 
1433 	db_print_indent(indent);
1434 #ifdef INET6
1435 	if (inp->inp_vflag & INP_IPV6) {
1436 		db_printf("in6p_options: %p   in6p_outputopts: %p   "
1437 		    "in6p_moptions: %p\n", inp->in6p_options,
1438 		    inp->in6p_outputopts, inp->in6p_moptions);
1439 		db_printf("in6p_icmp6filt: %p   in6p_cksum %d   "
1440 		    "in6p_hops %u\n", inp->in6p_icmp6filt, inp->in6p_cksum,
1441 		    inp->in6p_hops);
1442 	} else
1443 #endif
1444 	{
1445 		db_printf("inp_ip_tos: %d   inp_ip_options: %p   "
1446 		    "inp_ip_moptions: %p\n", inp->inp_ip_tos,
1447 		    inp->inp_options, inp->inp_moptions);
1448 	}
1449 
1450 	db_print_indent(indent);
1451 	db_printf("inp_phd: %p   inp_gencnt: %ju\n", inp->inp_phd,
1452 	    (uintmax_t)inp->inp_gencnt);
1453 }
1454 
1455 DB_SHOW_COMMAND(inpcb, db_show_inpcb)
1456 {
1457 	struct inpcb *inp;
1458 
1459 	if (!have_addr) {
1460 		db_printf("usage: show inpcb <addr>\n");
1461 		return;
1462 	}
1463 	inp = (struct inpcb *)addr;
1464 
1465 	db_print_inpcb(inp, "inpcb", 0);
1466 }
1467 #endif
1468