xref: /freebsd/sys/netinet/in_pcb.c (revision ea906c4152774dff300bb26fbfc1e4188351c89a)
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 #include <sys/vimage.h>
56 
57 #ifdef DDB
58 #include <ddb/ddb.h>
59 #endif
60 
61 #include <vm/uma.h>
62 
63 #include <net/if.h>
64 #include <net/if_types.h>
65 #include <net/route.h>
66 
67 #include <netinet/in.h>
68 #include <netinet/in_pcb.h>
69 #include <netinet/in_var.h>
70 #include <netinet/ip_var.h>
71 #include <netinet/tcp_var.h>
72 #include <netinet/udp.h>
73 #include <netinet/udp_var.h>
74 #ifdef INET6
75 #include <netinet/ip6.h>
76 #include <netinet6/ip6_var.h>
77 #endif /* INET6 */
78 
79 
80 #ifdef IPSEC
81 #include <netipsec/ipsec.h>
82 #include <netipsec/key.h>
83 #endif /* IPSEC */
84 
85 #include <security/mac/mac_framework.h>
86 
87 /*
88  * These configure the range of local port addresses assigned to
89  * "unspecified" outgoing connections/packets/whatever.
90  */
91 int	ipport_lowfirstauto  = IPPORT_RESERVED - 1;	/* 1023 */
92 int	ipport_lowlastauto = IPPORT_RESERVEDSTART;	/* 600 */
93 int	ipport_firstauto = IPPORT_EPHEMERALFIRST;	/* 10000 */
94 int	ipport_lastauto  = IPPORT_EPHEMERALLAST;	/* 65535 */
95 int	ipport_hifirstauto = IPPORT_HIFIRSTAUTO;	/* 49152 */
96 int	ipport_hilastauto  = IPPORT_HILASTAUTO;		/* 65535 */
97 
98 /*
99  * Reserved ports accessible only to root. There are significant
100  * security considerations that must be accounted for when changing these,
101  * but the security benefits can be great. Please be careful.
102  */
103 int	ipport_reservedhigh = IPPORT_RESERVED - 1;	/* 1023 */
104 int	ipport_reservedlow = 0;
105 
106 /* Variables dealing with random ephemeral port allocation. */
107 int	ipport_randomized = 1;	/* user controlled via sysctl */
108 int	ipport_randomcps = 10;	/* user controlled via sysctl */
109 int	ipport_randomtime = 45;	/* user controlled via sysctl */
110 int	ipport_stoprandom = 0;	/* toggled by ipport_tick */
111 int	ipport_tcpallocs;
112 int	ipport_tcplastcount;
113 
114 #define RANGECHK(var, min, max) \
115 	if ((var) < (min)) { (var) = (min); } \
116 	else if ((var) > (max)) { (var) = (max); }
117 
118 static int
119 sysctl_net_ipport_check(SYSCTL_HANDLER_ARGS)
120 {
121 	int error;
122 
123 	error = sysctl_handle_int(oidp, oidp->oid_arg1, oidp->oid_arg2, req);
124 	if (error == 0) {
125 		RANGECHK(V_ipport_lowfirstauto, 1, IPPORT_RESERVED - 1);
126 		RANGECHK(V_ipport_lowlastauto, 1, IPPORT_RESERVED - 1);
127 		RANGECHK(V_ipport_firstauto, IPPORT_RESERVED, IPPORT_MAX);
128 		RANGECHK(V_ipport_lastauto, IPPORT_RESERVED, IPPORT_MAX);
129 		RANGECHK(V_ipport_hifirstauto, IPPORT_RESERVED, IPPORT_MAX);
130 		RANGECHK(V_ipport_hilastauto, IPPORT_RESERVED, IPPORT_MAX);
131 	}
132 	return (error);
133 }
134 
135 #undef RANGECHK
136 
137 SYSCTL_NODE(_net_inet_ip, IPPROTO_IP, portrange, CTLFLAG_RW, 0, "IP Ports");
138 
139 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, lowfirst, CTLTYPE_INT|CTLFLAG_RW,
140 	   &ipport_lowfirstauto, 0, &sysctl_net_ipport_check, "I", "");
141 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, lowlast, CTLTYPE_INT|CTLFLAG_RW,
142 	   &ipport_lowlastauto, 0, &sysctl_net_ipport_check, "I", "");
143 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, first, CTLTYPE_INT|CTLFLAG_RW,
144 	   &ipport_firstauto, 0, &sysctl_net_ipport_check, "I", "");
145 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, last, CTLTYPE_INT|CTLFLAG_RW,
146 	   &ipport_lastauto, 0, &sysctl_net_ipport_check, "I", "");
147 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, hifirst, CTLTYPE_INT|CTLFLAG_RW,
148 	   &ipport_hifirstauto, 0, &sysctl_net_ipport_check, "I", "");
149 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, hilast, CTLTYPE_INT|CTLFLAG_RW,
150 	   &ipport_hilastauto, 0, &sysctl_net_ipport_check, "I", "");
151 SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, reservedhigh,
152 	   CTLFLAG_RW|CTLFLAG_SECURE, &ipport_reservedhigh, 0, "");
153 SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, reservedlow,
154 	   CTLFLAG_RW|CTLFLAG_SECURE, &ipport_reservedlow, 0, "");
155 SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, randomized, CTLFLAG_RW,
156 	   &ipport_randomized, 0, "Enable random port allocation");
157 SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, randomcps, CTLFLAG_RW,
158 	   &ipport_randomcps, 0, "Maximum number of random port "
159 	   "allocations before switching to a sequental one");
160 SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, randomtime, CTLFLAG_RW,
161 	   &ipport_randomtime, 0, "Minimum time to keep sequental port "
162 	   "allocation before switching to a random one");
163 
164 /*
165  * in_pcb.c: manage the Protocol Control Blocks.
166  *
167  * NOTE: It is assumed that most of these functions will be called with
168  * the pcbinfo lock held, and often, the inpcb lock held, as these utility
169  * functions often modify hash chains or addresses in pcbs.
170  */
171 
172 /*
173  * Allocate a PCB and associate it with the socket.
174  * On success return with the PCB locked.
175  */
176 int
177 in_pcballoc(struct socket *so, struct inpcbinfo *pcbinfo)
178 {
179 	struct inpcb *inp;
180 	int error;
181 
182 	INP_INFO_WLOCK_ASSERT(pcbinfo);
183 	error = 0;
184 	inp = uma_zalloc(pcbinfo->ipi_zone, M_NOWAIT);
185 	if (inp == NULL)
186 		return (ENOBUFS);
187 	bzero(inp, inp_zero_size);
188 	inp->inp_pcbinfo = pcbinfo;
189 	inp->inp_socket = so;
190 	inp->inp_inc.inc_fibnum = so->so_fibnum;
191 #ifdef MAC
192 	error = mac_inpcb_init(inp, M_NOWAIT);
193 	if (error != 0)
194 		goto out;
195 	SOCK_LOCK(so);
196 	mac_inpcb_create(so, inp);
197 	SOCK_UNLOCK(so);
198 #endif
199 
200 #ifdef IPSEC
201 	error = ipsec_init_policy(so, &inp->inp_sp);
202 	if (error != 0) {
203 #ifdef MAC
204 		mac_inpcb_destroy(inp);
205 #endif
206 		goto out;
207 	}
208 #endif /*IPSEC*/
209 #ifdef INET6
210 	if (INP_SOCKAF(so) == AF_INET6) {
211 		inp->inp_vflag |= INP_IPV6PROTO;
212 		if (V_ip6_v6only)
213 			inp->inp_flags |= IN6P_IPV6_V6ONLY;
214 	}
215 #endif
216 	LIST_INSERT_HEAD(pcbinfo->ipi_listhead, inp, inp_list);
217 	pcbinfo->ipi_count++;
218 	so->so_pcb = (caddr_t)inp;
219 #ifdef INET6
220 	if (V_ip6_auto_flowlabel)
221 		inp->inp_flags |= IN6P_AUTOFLOWLABEL;
222 #endif
223 	INP_WLOCK(inp);
224 	inp->inp_gencnt = ++pcbinfo->ipi_gencnt;
225 
226 #if defined(IPSEC) || defined(MAC)
227 out:
228 	if (error != 0)
229 		uma_zfree(pcbinfo->ipi_zone, inp);
230 #endif
231 	return (error);
232 }
233 
234 int
235 in_pcbbind(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred)
236 {
237 	int anonport, error;
238 
239 	INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
240 	INP_WLOCK_ASSERT(inp);
241 
242 	if (inp->inp_lport != 0 || inp->inp_laddr.s_addr != INADDR_ANY)
243 		return (EINVAL);
244 	anonport = inp->inp_lport == 0 && (nam == NULL ||
245 	    ((struct sockaddr_in *)nam)->sin_port == 0);
246 	error = in_pcbbind_setup(inp, nam, &inp->inp_laddr.s_addr,
247 	    &inp->inp_lport, cred);
248 	if (error)
249 		return (error);
250 	if (in_pcbinshash(inp) != 0) {
251 		inp->inp_laddr.s_addr = INADDR_ANY;
252 		inp->inp_lport = 0;
253 		return (EAGAIN);
254 	}
255 	if (anonport)
256 		inp->inp_flags |= INP_ANONPORT;
257 	return (0);
258 }
259 
260 /*
261  * Set up a bind operation on a PCB, performing port allocation
262  * as required, but do not actually modify the PCB. Callers can
263  * either complete the bind by setting inp_laddr/inp_lport and
264  * calling in_pcbinshash(), or they can just use the resulting
265  * port and address to authorise the sending of a once-off packet.
266  *
267  * On error, the values of *laddrp and *lportp are not changed.
268  */
269 int
270 in_pcbbind_setup(struct inpcb *inp, struct sockaddr *nam, in_addr_t *laddrp,
271     u_short *lportp, struct ucred *cred)
272 {
273 	struct socket *so = inp->inp_socket;
274 	unsigned short *lastport;
275 	struct sockaddr_in *sin;
276 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
277 	struct in_addr laddr;
278 	u_short lport = 0;
279 	int wild = 0, reuseport = (so->so_options & SO_REUSEPORT);
280 	int error, prison = 0;
281 	int dorandom;
282 
283 	/*
284 	 * Because no actual state changes occur here, a global write lock on
285 	 * the pcbinfo isn't required.
286 	 */
287 	INP_INFO_LOCK_ASSERT(pcbinfo);
288 	INP_LOCK_ASSERT(inp);
289 
290 	if (TAILQ_EMPTY(&V_in_ifaddrhead)) /* XXX broken! */
291 		return (EADDRNOTAVAIL);
292 	laddr.s_addr = *laddrp;
293 	if (nam != NULL && laddr.s_addr != INADDR_ANY)
294 		return (EINVAL);
295 	if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0)
296 		wild = INPLOOKUP_WILDCARD;
297 	if (nam) {
298 		sin = (struct sockaddr_in *)nam;
299 		if (nam->sa_len != sizeof (*sin))
300 			return (EINVAL);
301 #ifdef notdef
302 		/*
303 		 * We should check the family, but old programs
304 		 * incorrectly fail to initialize it.
305 		 */
306 		if (sin->sin_family != AF_INET)
307 			return (EAFNOSUPPORT);
308 #endif
309 		if (sin->sin_addr.s_addr != INADDR_ANY)
310 			if (prison_ip(cred, 0, &sin->sin_addr.s_addr))
311 				return(EINVAL);
312 		if (sin->sin_port != *lportp) {
313 			/* Don't allow the port to change. */
314 			if (*lportp != 0)
315 				return (EINVAL);
316 			lport = sin->sin_port;
317 		}
318 		/* NB: lport is left as 0 if the port isn't being changed. */
319 		if (IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) {
320 			/*
321 			 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
322 			 * allow complete duplication of binding if
323 			 * SO_REUSEPORT is set, or if SO_REUSEADDR is set
324 			 * and a multicast address is bound on both
325 			 * new and duplicated sockets.
326 			 */
327 			if (so->so_options & SO_REUSEADDR)
328 				reuseport = SO_REUSEADDR|SO_REUSEPORT;
329 		} else if (sin->sin_addr.s_addr != INADDR_ANY) {
330 			sin->sin_port = 0;		/* yech... */
331 			bzero(&sin->sin_zero, sizeof(sin->sin_zero));
332 			if (ifa_ifwithaddr((struct sockaddr *)sin) == 0)
333 				return (EADDRNOTAVAIL);
334 		}
335 		laddr = sin->sin_addr;
336 		if (lport) {
337 			struct inpcb *t;
338 			struct tcptw *tw;
339 
340 			/* GROSS */
341 			if (ntohs(lport) <= V_ipport_reservedhigh &&
342 			    ntohs(lport) >= V_ipport_reservedlow &&
343 			    priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT,
344 			    0))
345 				return (EACCES);
346 			if (jailed(cred))
347 				prison = 1;
348 			if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)) &&
349 			    priv_check_cred(so->so_cred,
350 			    PRIV_NETINET_REUSEPORT, 0) != 0) {
351 				t = in_pcblookup_local(pcbinfo, sin->sin_addr,
352 				    lport, prison ? 0 : INPLOOKUP_WILDCARD,
353 				    cred);
354 	/*
355 	 * XXX
356 	 * This entire block sorely needs a rewrite.
357 	 */
358 				if (t &&
359 				    ((t->inp_vflag & INP_TIMEWAIT) == 0) &&
360 				    (so->so_type != SOCK_STREAM ||
361 				     ntohl(t->inp_faddr.s_addr) == INADDR_ANY) &&
362 				    (ntohl(sin->sin_addr.s_addr) != INADDR_ANY ||
363 				     ntohl(t->inp_laddr.s_addr) != INADDR_ANY ||
364 				     (t->inp_socket->so_options &
365 					 SO_REUSEPORT) == 0) &&
366 				    (so->so_cred->cr_uid !=
367 				     t->inp_socket->so_cred->cr_uid))
368 					return (EADDRINUSE);
369 			}
370 			if (prison && prison_ip(cred, 0, &sin->sin_addr.s_addr))
371 				return (EADDRNOTAVAIL);
372 			t = in_pcblookup_local(pcbinfo, sin->sin_addr,
373 			    lport, prison ? 0 : wild, cred);
374 			if (t && (t->inp_vflag & INP_TIMEWAIT)) {
375 				/*
376 				 * XXXRW: If an incpb has had its timewait
377 				 * state recycled, we treat the address as
378 				 * being in use (for now).  This is better
379 				 * than a panic, but not desirable.
380 				 */
381 				tw = intotw(inp);
382 				if (tw == NULL ||
383 				    (reuseport & tw->tw_so_options) == 0)
384 					return (EADDRINUSE);
385 			} else if (t &&
386 			    (reuseport & t->inp_socket->so_options) == 0) {
387 #ifdef INET6
388 				if (ntohl(sin->sin_addr.s_addr) !=
389 				    INADDR_ANY ||
390 				    ntohl(t->inp_laddr.s_addr) !=
391 				    INADDR_ANY ||
392 				    INP_SOCKAF(so) ==
393 				    INP_SOCKAF(t->inp_socket))
394 #endif
395 				return (EADDRINUSE);
396 			}
397 		}
398 	}
399 	if (*lportp != 0)
400 		lport = *lportp;
401 	if (lport == 0) {
402 		u_short first, last, aux;
403 		int count;
404 
405 		if (laddr.s_addr != INADDR_ANY)
406 			if (prison_ip(cred, 0, &laddr.s_addr))
407 				return (EINVAL);
408 
409 		if (inp->inp_flags & INP_HIGHPORT) {
410 			first = V_ipport_hifirstauto;	/* sysctl */
411 			last  = V_ipport_hilastauto;
412 			lastport = &pcbinfo->ipi_lasthi;
413 		} else if (inp->inp_flags & INP_LOWPORT) {
414 			error = priv_check_cred(cred,
415 			    PRIV_NETINET_RESERVEDPORT, 0);
416 			if (error)
417 				return error;
418 			first = V_ipport_lowfirstauto;	/* 1023 */
419 			last  = V_ipport_lowlastauto;	/* 600 */
420 			lastport = &pcbinfo->ipi_lastlow;
421 		} else {
422 			first = V_ipport_firstauto;	/* sysctl */
423 			last  = V_ipport_lastauto;
424 			lastport = &pcbinfo->ipi_lastport;
425 		}
426 		/*
427 		 * For UDP, use random port allocation as long as the user
428 		 * allows it.  For TCP (and as of yet unknown) connections,
429 		 * use random port allocation only if the user allows it AND
430 		 * ipport_tick() allows it.
431 		 */
432 		if (V_ipport_randomized &&
433 			(!V_ipport_stoprandom || pcbinfo == &V_udbinfo))
434 			dorandom = 1;
435 		else
436 			dorandom = 0;
437 		/*
438 		 * It makes no sense to do random port allocation if
439 		 * we have the only port available.
440 		 */
441 		if (first == last)
442 			dorandom = 0;
443 		/* Make sure to not include UDP packets in the count. */
444 		if (pcbinfo != &V_udbinfo)
445 			V_ipport_tcpallocs++;
446 		/*
447 		 * Simple check to ensure all ports are not used up causing
448 		 * a deadlock here.
449 		 */
450 		if (first > last) {
451 			aux = first;
452 			first = last;
453 			last = aux;
454 		}
455 
456 		if (dorandom)
457 			*lastport = first +
458 				    (arc4random() % (last - first));
459 
460 		count = last - first;
461 
462 		do {
463 			if (count-- < 0)	/* completely used? */
464 				return (EADDRNOTAVAIL);
465 			++*lastport;
466 			if (*lastport < first || *lastport > last)
467 				*lastport = first;
468 			lport = htons(*lastport);
469 		} while (in_pcblookup_local(pcbinfo, laddr,
470 		    lport, wild, cred));
471 	}
472 	if (prison_ip(cred, 0, &laddr.s_addr))
473 		return (EINVAL);
474 	*laddrp = laddr.s_addr;
475 	*lportp = lport;
476 	return (0);
477 }
478 
479 /*
480  * Connect from a socket to a specified address.
481  * Both address and port must be specified in argument sin.
482  * If don't have a local address for this socket yet,
483  * then pick one.
484  */
485 int
486 in_pcbconnect(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred)
487 {
488 	u_short lport, fport;
489 	in_addr_t laddr, faddr;
490 	int anonport, error;
491 
492 	INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
493 	INP_WLOCK_ASSERT(inp);
494 
495 	lport = inp->inp_lport;
496 	laddr = inp->inp_laddr.s_addr;
497 	anonport = (lport == 0);
498 	error = in_pcbconnect_setup(inp, nam, &laddr, &lport, &faddr, &fport,
499 	    NULL, cred);
500 	if (error)
501 		return (error);
502 
503 	/* Do the initial binding of the local address if required. */
504 	if (inp->inp_laddr.s_addr == INADDR_ANY && inp->inp_lport == 0) {
505 		inp->inp_lport = lport;
506 		inp->inp_laddr.s_addr = laddr;
507 		if (in_pcbinshash(inp) != 0) {
508 			inp->inp_laddr.s_addr = INADDR_ANY;
509 			inp->inp_lport = 0;
510 			return (EAGAIN);
511 		}
512 	}
513 
514 	/* Commit the remaining changes. */
515 	inp->inp_lport = lport;
516 	inp->inp_laddr.s_addr = laddr;
517 	inp->inp_faddr.s_addr = faddr;
518 	inp->inp_fport = fport;
519 	in_pcbrehash(inp);
520 
521 	if (anonport)
522 		inp->inp_flags |= INP_ANONPORT;
523 	return (0);
524 }
525 
526 /*
527  * Set up for a connect from a socket to the specified address.
528  * On entry, *laddrp and *lportp should contain the current local
529  * address and port for the PCB; these are updated to the values
530  * that should be placed in inp_laddr and inp_lport to complete
531  * the connect.
532  *
533  * On success, *faddrp and *fportp will be set to the remote address
534  * and port. These are not updated in the error case.
535  *
536  * If the operation fails because the connection already exists,
537  * *oinpp will be set to the PCB of that connection so that the
538  * caller can decide to override it. In all other cases, *oinpp
539  * is set to NULL.
540  */
541 int
542 in_pcbconnect_setup(struct inpcb *inp, struct sockaddr *nam,
543     in_addr_t *laddrp, u_short *lportp, in_addr_t *faddrp, u_short *fportp,
544     struct inpcb **oinpp, struct ucred *cred)
545 {
546 	struct sockaddr_in *sin = (struct sockaddr_in *)nam;
547 	struct in_ifaddr *ia;
548 	struct sockaddr_in sa;
549 	struct ucred *socred;
550 	struct inpcb *oinp;
551 	struct in_addr laddr, faddr;
552 	u_short lport, fport;
553 	int error;
554 
555 	/*
556 	 * Because a global state change doesn't actually occur here, a read
557 	 * lock is sufficient.
558 	 */
559 	INP_INFO_LOCK_ASSERT(inp->inp_pcbinfo);
560 	INP_LOCK_ASSERT(inp);
561 
562 	if (oinpp != NULL)
563 		*oinpp = NULL;
564 	if (nam->sa_len != sizeof (*sin))
565 		return (EINVAL);
566 	if (sin->sin_family != AF_INET)
567 		return (EAFNOSUPPORT);
568 	if (sin->sin_port == 0)
569 		return (EADDRNOTAVAIL);
570 	laddr.s_addr = *laddrp;
571 	lport = *lportp;
572 	faddr = sin->sin_addr;
573 	fport = sin->sin_port;
574 	socred = inp->inp_socket->so_cred;
575 	if (laddr.s_addr == INADDR_ANY && jailed(socred)) {
576 		bzero(&sa, sizeof(sa));
577 		sa.sin_addr.s_addr = htonl(prison_getip(socred));
578 		sa.sin_len = sizeof(sa);
579 		sa.sin_family = AF_INET;
580 		error = in_pcbbind_setup(inp, (struct sockaddr *)&sa,
581 		    &laddr.s_addr, &lport, cred);
582 		if (error)
583 			return (error);
584 	}
585 	if (!TAILQ_EMPTY(&V_in_ifaddrhead)) {
586 		/*
587 		 * If the destination address is INADDR_ANY,
588 		 * use the primary local address.
589 		 * If the supplied address is INADDR_BROADCAST,
590 		 * and the primary interface supports broadcast,
591 		 * choose the broadcast address for that interface.
592 		 */
593 		if (faddr.s_addr == INADDR_ANY)
594 			faddr = IA_SIN(TAILQ_FIRST(&V_in_ifaddrhead))->sin_addr;
595 		else if (faddr.s_addr == (u_long)INADDR_BROADCAST &&
596 		    (TAILQ_FIRST(&V_in_ifaddrhead)->ia_ifp->if_flags &
597 		    IFF_BROADCAST))
598 			faddr = satosin(&TAILQ_FIRST(
599 			    &V_in_ifaddrhead)->ia_broadaddr)->sin_addr;
600 	}
601 	if (laddr.s_addr == INADDR_ANY) {
602 		ia = NULL;
603 		/*
604 		 * If route is known our src addr is taken from the i/f,
605 		 * else punt.
606 		 *
607 		 * Find out route to destination
608 		 */
609 		if ((inp->inp_socket->so_options & SO_DONTROUTE) == 0)
610 			ia = ip_rtaddr(faddr, inp->inp_inc.inc_fibnum);
611 		/*
612 		 * If we found a route, use the address corresponding to
613 		 * the outgoing interface.
614 		 *
615 		 * Otherwise assume faddr is reachable on a directly connected
616 		 * network and try to find a corresponding interface to take
617 		 * the source address from.
618 		 */
619 		if (ia == NULL) {
620 			bzero(&sa, sizeof(sa));
621 			sa.sin_addr = faddr;
622 			sa.sin_len = sizeof(sa);
623 			sa.sin_family = AF_INET;
624 
625 			ia = ifatoia(ifa_ifwithdstaddr(sintosa(&sa)));
626 			if (ia == NULL)
627 				ia = ifatoia(ifa_ifwithnet(sintosa(&sa)));
628 			if (ia == NULL)
629 				return (ENETUNREACH);
630 		}
631 		/*
632 		 * If the destination address is multicast and an outgoing
633 		 * interface has been set as a multicast option, use the
634 		 * address of that interface as our source address.
635 		 */
636 		if (IN_MULTICAST(ntohl(faddr.s_addr)) &&
637 		    inp->inp_moptions != NULL) {
638 			struct ip_moptions *imo;
639 			struct ifnet *ifp;
640 
641 			imo = inp->inp_moptions;
642 			if (imo->imo_multicast_ifp != NULL) {
643 				ifp = imo->imo_multicast_ifp;
644 				TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link)
645 					if (ia->ia_ifp == ifp)
646 						break;
647 				if (ia == NULL)
648 					return (EADDRNOTAVAIL);
649 			}
650 		}
651 		laddr = ia->ia_addr.sin_addr;
652 	}
653 
654 	oinp = in_pcblookup_hash(inp->inp_pcbinfo, faddr, fport, laddr, lport,
655 	    0, NULL);
656 	if (oinp != NULL) {
657 		if (oinpp != NULL)
658 			*oinpp = oinp;
659 		return (EADDRINUSE);
660 	}
661 	if (lport == 0) {
662 		error = in_pcbbind_setup(inp, NULL, &laddr.s_addr, &lport,
663 		    cred);
664 		if (error)
665 			return (error);
666 	}
667 	*laddrp = laddr.s_addr;
668 	*lportp = lport;
669 	*faddrp = faddr.s_addr;
670 	*fportp = fport;
671 	return (0);
672 }
673 
674 void
675 in_pcbdisconnect(struct inpcb *inp)
676 {
677 
678 	INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
679 	INP_WLOCK_ASSERT(inp);
680 
681 	inp->inp_faddr.s_addr = INADDR_ANY;
682 	inp->inp_fport = 0;
683 	in_pcbrehash(inp);
684 }
685 
686 /*
687  * In the old world order, in_pcbdetach() served two functions: to detach the
688  * pcb from the socket/potentially free the socket, and to free the pcb
689  * itself.  In the new world order, the protocol code is responsible for
690  * managing the relationship with the socket, and this code simply frees the
691  * pcb.
692  */
693 void
694 in_pcbdetach(struct inpcb *inp)
695 {
696 
697 	KASSERT(inp->inp_socket != NULL, ("in_pcbdetach: inp_socket == NULL"));
698 	inp->inp_socket->so_pcb = NULL;
699 	inp->inp_socket = NULL;
700 }
701 
702 void
703 in_pcbfree(struct inpcb *inp)
704 {
705 	struct inpcbinfo *ipi = inp->inp_pcbinfo;
706 
707 	KASSERT(inp->inp_socket == NULL, ("in_pcbfree: inp_socket != NULL"));
708 
709 	INP_INFO_WLOCK_ASSERT(ipi);
710 	INP_WLOCK_ASSERT(inp);
711 
712 #ifdef IPSEC
713 	ipsec4_delete_pcbpolicy(inp);
714 #endif /*IPSEC*/
715 	inp->inp_gencnt = ++ipi->ipi_gencnt;
716 	in_pcbremlists(inp);
717 	if (inp->inp_options)
718 		(void)m_free(inp->inp_options);
719 	if (inp->inp_moptions != NULL)
720 		inp_freemoptions(inp->inp_moptions);
721 	inp->inp_vflag = 0;
722 
723 #ifdef MAC
724 	mac_inpcb_destroy(inp);
725 #endif
726 	INP_WUNLOCK(inp);
727 	uma_zfree(ipi->ipi_zone, inp);
728 }
729 
730 /*
731  * TCP needs to maintain its inpcb structure after the TCP connection has
732  * been torn down.  However, it must be disconnected from the inpcb hashes as
733  * it must not prevent binding of future connections to the same port/ip
734  * combination by other inpcbs.
735  */
736 void
737 in_pcbdrop(struct inpcb *inp)
738 {
739 
740 	INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
741 	INP_WLOCK_ASSERT(inp);
742 
743 	inp->inp_vflag |= INP_DROPPED;
744 	if (inp->inp_lport) {
745 		struct inpcbport *phd = inp->inp_phd;
746 
747 		LIST_REMOVE(inp, inp_hash);
748 		LIST_REMOVE(inp, inp_portlist);
749 		if (LIST_FIRST(&phd->phd_pcblist) == NULL) {
750 			LIST_REMOVE(phd, phd_hash);
751 			free(phd, M_PCB);
752 		}
753 		inp->inp_lport = 0;
754 	}
755 }
756 
757 /*
758  * Common routines to return the socket addresses associated with inpcbs.
759  */
760 struct sockaddr *
761 in_sockaddr(in_port_t port, struct in_addr *addr_p)
762 {
763 	struct sockaddr_in *sin;
764 
765 	MALLOC(sin, struct sockaddr_in *, sizeof *sin, M_SONAME,
766 		M_WAITOK | M_ZERO);
767 	sin->sin_family = AF_INET;
768 	sin->sin_len = sizeof(*sin);
769 	sin->sin_addr = *addr_p;
770 	sin->sin_port = port;
771 
772 	return (struct sockaddr *)sin;
773 }
774 
775 int
776 in_getsockaddr(struct socket *so, struct sockaddr **nam)
777 {
778 	struct inpcb *inp;
779 	struct in_addr addr;
780 	in_port_t port;
781 
782 	inp = sotoinpcb(so);
783 	KASSERT(inp != NULL, ("in_getsockaddr: inp == NULL"));
784 
785 	INP_RLOCK(inp);
786 	port = inp->inp_lport;
787 	addr = inp->inp_laddr;
788 	INP_RUNLOCK(inp);
789 
790 	*nam = in_sockaddr(port, &addr);
791 	return 0;
792 }
793 
794 int
795 in_getpeeraddr(struct socket *so, struct sockaddr **nam)
796 {
797 	struct inpcb *inp;
798 	struct in_addr addr;
799 	in_port_t port;
800 
801 	inp = sotoinpcb(so);
802 	KASSERT(inp != NULL, ("in_getpeeraddr: inp == NULL"));
803 
804 	INP_RLOCK(inp);
805 	port = inp->inp_fport;
806 	addr = inp->inp_faddr;
807 	INP_RUNLOCK(inp);
808 
809 	*nam = in_sockaddr(port, &addr);
810 	return 0;
811 }
812 
813 void
814 in_pcbnotifyall(struct inpcbinfo *pcbinfo, struct in_addr faddr, int errno,
815     struct inpcb *(*notify)(struct inpcb *, int))
816 {
817 	struct inpcb *inp, *inp_temp;
818 
819 	INP_INFO_WLOCK(pcbinfo);
820 	LIST_FOREACH_SAFE(inp, pcbinfo->ipi_listhead, inp_list, inp_temp) {
821 		INP_WLOCK(inp);
822 #ifdef INET6
823 		if ((inp->inp_vflag & INP_IPV4) == 0) {
824 			INP_WUNLOCK(inp);
825 			continue;
826 		}
827 #endif
828 		if (inp->inp_faddr.s_addr != faddr.s_addr ||
829 		    inp->inp_socket == NULL) {
830 			INP_WUNLOCK(inp);
831 			continue;
832 		}
833 		if ((*notify)(inp, errno))
834 			INP_WUNLOCK(inp);
835 	}
836 	INP_INFO_WUNLOCK(pcbinfo);
837 }
838 
839 void
840 in_pcbpurgeif0(struct inpcbinfo *pcbinfo, struct ifnet *ifp)
841 {
842 	struct inpcb *inp;
843 	struct ip_moptions *imo;
844 	int i, gap;
845 
846 	INP_INFO_RLOCK(pcbinfo);
847 	LIST_FOREACH(inp, pcbinfo->ipi_listhead, inp_list) {
848 		INP_WLOCK(inp);
849 		imo = inp->inp_moptions;
850 		if ((inp->inp_vflag & INP_IPV4) &&
851 		    imo != NULL) {
852 			/*
853 			 * Unselect the outgoing interface if it is being
854 			 * detached.
855 			 */
856 			if (imo->imo_multicast_ifp == ifp)
857 				imo->imo_multicast_ifp = NULL;
858 
859 			/*
860 			 * Drop multicast group membership if we joined
861 			 * through the interface being detached.
862 			 */
863 			for (i = 0, gap = 0; i < imo->imo_num_memberships;
864 			    i++) {
865 				if (imo->imo_membership[i]->inm_ifp == ifp) {
866 					in_delmulti(imo->imo_membership[i]);
867 					gap++;
868 				} else if (gap != 0)
869 					imo->imo_membership[i - gap] =
870 					    imo->imo_membership[i];
871 			}
872 			imo->imo_num_memberships -= gap;
873 		}
874 		INP_WUNLOCK(inp);
875 	}
876 	INP_INFO_RUNLOCK(pcbinfo);
877 }
878 
879 /*
880  * Lookup a PCB based on the local address and port.
881  */
882 #define INP_LOOKUP_MAPPED_PCB_COST	3
883 struct inpcb *
884 in_pcblookup_local(struct inpcbinfo *pcbinfo, struct in_addr laddr,
885     u_short lport, int wild_okay, struct ucred *cred)
886 {
887 	struct inpcb *inp;
888 #ifdef INET6
889 	int matchwild = 3 + INP_LOOKUP_MAPPED_PCB_COST;
890 #else
891 	int matchwild = 3;
892 #endif
893 	int wildcard;
894 
895 	INP_INFO_LOCK_ASSERT(pcbinfo);
896 
897 	if (!wild_okay) {
898 		struct inpcbhead *head;
899 		/*
900 		 * Look for an unconnected (wildcard foreign addr) PCB that
901 		 * matches the local address and port we're looking for.
902 		 */
903 		head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport,
904 		    0, pcbinfo->ipi_hashmask)];
905 		LIST_FOREACH(inp, head, inp_hash) {
906 #ifdef INET6
907 			if ((inp->inp_vflag & INP_IPV4) == 0)
908 				continue;
909 #endif
910 			if (inp->inp_faddr.s_addr == INADDR_ANY &&
911 			    inp->inp_laddr.s_addr == laddr.s_addr &&
912 			    inp->inp_lport == lport) {
913 				/*
914 				 * Found.
915 				 */
916 				return (inp);
917 			}
918 		}
919 		/*
920 		 * Not found.
921 		 */
922 		return (NULL);
923 	} else {
924 		struct inpcbporthead *porthash;
925 		struct inpcbport *phd;
926 		struct inpcb *match = NULL;
927 		/*
928 		 * Best fit PCB lookup.
929 		 *
930 		 * First see if this local port is in use by looking on the
931 		 * port hash list.
932 		 */
933 		porthash = &pcbinfo->ipi_porthashbase[INP_PCBPORTHASH(lport,
934 		    pcbinfo->ipi_porthashmask)];
935 		LIST_FOREACH(phd, porthash, phd_hash) {
936 			if (phd->phd_port == lport)
937 				break;
938 		}
939 		if (phd != NULL) {
940 			/*
941 			 * Port is in use by one or more PCBs. Look for best
942 			 * fit.
943 			 */
944 			LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) {
945 				wildcard = 0;
946 #ifdef INET6
947 				if ((inp->inp_vflag & INP_IPV4) == 0)
948 					continue;
949 				/*
950 				 * We never select the PCB that has
951 				 * INP_IPV6 flag and is bound to :: if
952 				 * we have another PCB which is bound
953 				 * to 0.0.0.0.  If a PCB has the
954 				 * INP_IPV6 flag, then we set its cost
955 				 * higher than IPv4 only PCBs.
956 				 *
957 				 * Note that the case only happens
958 				 * when a socket is bound to ::, under
959 				 * the condition that the use of the
960 				 * mapped address is allowed.
961 				 */
962 				if ((inp->inp_vflag & INP_IPV6) != 0)
963 					wildcard += INP_LOOKUP_MAPPED_PCB_COST;
964 #endif
965 				if (inp->inp_faddr.s_addr != INADDR_ANY)
966 					wildcard++;
967 				if (inp->inp_laddr.s_addr != INADDR_ANY) {
968 					if (laddr.s_addr == INADDR_ANY)
969 						wildcard++;
970 					else if (inp->inp_laddr.s_addr != laddr.s_addr)
971 						continue;
972 				} else {
973 					if (laddr.s_addr != INADDR_ANY)
974 						wildcard++;
975 				}
976 				if (wildcard < matchwild) {
977 					match = inp;
978 					matchwild = wildcard;
979 					if (matchwild == 0) {
980 						break;
981 					}
982 				}
983 			}
984 		}
985 		return (match);
986 	}
987 }
988 #undef INP_LOOKUP_MAPPED_PCB_COST
989 
990 /*
991  * Lookup PCB in hash list.
992  */
993 struct inpcb *
994 in_pcblookup_hash(struct inpcbinfo *pcbinfo, struct in_addr faddr,
995     u_int fport_arg, struct in_addr laddr, u_int lport_arg, int wildcard,
996     struct ifnet *ifp)
997 {
998 	struct inpcbhead *head;
999 	struct inpcb *inp;
1000 	u_short fport = fport_arg, lport = lport_arg;
1001 
1002 	INP_INFO_LOCK_ASSERT(pcbinfo);
1003 
1004 	/*
1005 	 * First look for an exact match.
1006 	 */
1007 	head = &pcbinfo->ipi_hashbase[INP_PCBHASH(faddr.s_addr, lport, fport,
1008 	    pcbinfo->ipi_hashmask)];
1009 	LIST_FOREACH(inp, head, inp_hash) {
1010 #ifdef INET6
1011 		if ((inp->inp_vflag & INP_IPV4) == 0)
1012 			continue;
1013 #endif
1014 		if (inp->inp_faddr.s_addr == faddr.s_addr &&
1015 		    inp->inp_laddr.s_addr == laddr.s_addr &&
1016 		    inp->inp_fport == fport &&
1017 		    inp->inp_lport == lport)
1018 			return (inp);
1019 	}
1020 
1021 	/*
1022 	 * Then look for a wildcard match, if requested.
1023 	 */
1024 	if (wildcard) {
1025 		struct inpcb *local_wild = NULL;
1026 #ifdef INET6
1027 		struct inpcb *local_wild_mapped = NULL;
1028 #endif
1029 
1030 		head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport,
1031 		    0, pcbinfo->ipi_hashmask)];
1032 		LIST_FOREACH(inp, head, inp_hash) {
1033 #ifdef INET6
1034 			if ((inp->inp_vflag & INP_IPV4) == 0)
1035 				continue;
1036 #endif
1037 			if (inp->inp_faddr.s_addr == INADDR_ANY &&
1038 			    inp->inp_lport == lport) {
1039 				if (ifp && ifp->if_type == IFT_FAITH &&
1040 				    (inp->inp_flags & INP_FAITH) == 0)
1041 					continue;
1042 				if (inp->inp_laddr.s_addr == laddr.s_addr)
1043 					return (inp);
1044 				else if (inp->inp_laddr.s_addr == INADDR_ANY) {
1045 #ifdef INET6
1046 					if (INP_CHECK_SOCKAF(inp->inp_socket,
1047 							     AF_INET6))
1048 						local_wild_mapped = inp;
1049 					else
1050 #endif
1051 						local_wild = inp;
1052 				}
1053 			}
1054 		}
1055 #ifdef INET6
1056 		if (local_wild == NULL)
1057 			return (local_wild_mapped);
1058 #endif
1059 		return (local_wild);
1060 	}
1061 	return (NULL);
1062 }
1063 
1064 /*
1065  * Insert PCB onto various hash lists.
1066  */
1067 int
1068 in_pcbinshash(struct inpcb *inp)
1069 {
1070 	struct inpcbhead *pcbhash;
1071 	struct inpcbporthead *pcbporthash;
1072 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1073 	struct inpcbport *phd;
1074 	u_int32_t hashkey_faddr;
1075 
1076 	INP_INFO_WLOCK_ASSERT(pcbinfo);
1077 	INP_WLOCK_ASSERT(inp);
1078 
1079 #ifdef INET6
1080 	if (inp->inp_vflag & INP_IPV6)
1081 		hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX */;
1082 	else
1083 #endif /* INET6 */
1084 	hashkey_faddr = inp->inp_faddr.s_addr;
1085 
1086 	pcbhash = &pcbinfo->ipi_hashbase[INP_PCBHASH(hashkey_faddr,
1087 		 inp->inp_lport, inp->inp_fport, pcbinfo->ipi_hashmask)];
1088 
1089 	pcbporthash = &pcbinfo->ipi_porthashbase[
1090 	    INP_PCBPORTHASH(inp->inp_lport, pcbinfo->ipi_porthashmask)];
1091 
1092 	/*
1093 	 * Go through port list and look for a head for this lport.
1094 	 */
1095 	LIST_FOREACH(phd, pcbporthash, phd_hash) {
1096 		if (phd->phd_port == inp->inp_lport)
1097 			break;
1098 	}
1099 	/*
1100 	 * If none exists, malloc one and tack it on.
1101 	 */
1102 	if (phd == NULL) {
1103 		MALLOC(phd, struct inpcbport *, sizeof(struct inpcbport), M_PCB, M_NOWAIT);
1104 		if (phd == NULL) {
1105 			return (ENOBUFS); /* XXX */
1106 		}
1107 		phd->phd_port = inp->inp_lport;
1108 		LIST_INIT(&phd->phd_pcblist);
1109 		LIST_INSERT_HEAD(pcbporthash, phd, phd_hash);
1110 	}
1111 	inp->inp_phd = phd;
1112 	LIST_INSERT_HEAD(&phd->phd_pcblist, inp, inp_portlist);
1113 	LIST_INSERT_HEAD(pcbhash, inp, inp_hash);
1114 	return (0);
1115 }
1116 
1117 /*
1118  * Move PCB to the proper hash bucket when { faddr, fport } have  been
1119  * changed. NOTE: This does not handle the case of the lport changing (the
1120  * hashed port list would have to be updated as well), so the lport must
1121  * not change after in_pcbinshash() has been called.
1122  */
1123 void
1124 in_pcbrehash(struct inpcb *inp)
1125 {
1126 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1127 	struct inpcbhead *head;
1128 	u_int32_t hashkey_faddr;
1129 
1130 	INP_INFO_WLOCK_ASSERT(pcbinfo);
1131 	INP_WLOCK_ASSERT(inp);
1132 
1133 #ifdef INET6
1134 	if (inp->inp_vflag & INP_IPV6)
1135 		hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX */;
1136 	else
1137 #endif /* INET6 */
1138 	hashkey_faddr = inp->inp_faddr.s_addr;
1139 
1140 	head = &pcbinfo->ipi_hashbase[INP_PCBHASH(hashkey_faddr,
1141 		inp->inp_lport, inp->inp_fport, pcbinfo->ipi_hashmask)];
1142 
1143 	LIST_REMOVE(inp, inp_hash);
1144 	LIST_INSERT_HEAD(head, inp, inp_hash);
1145 }
1146 
1147 /*
1148  * Remove PCB from various lists.
1149  */
1150 void
1151 in_pcbremlists(struct inpcb *inp)
1152 {
1153 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1154 
1155 	INP_INFO_WLOCK_ASSERT(pcbinfo);
1156 	INP_WLOCK_ASSERT(inp);
1157 
1158 	inp->inp_gencnt = ++pcbinfo->ipi_gencnt;
1159 	if (inp->inp_lport) {
1160 		struct inpcbport *phd = inp->inp_phd;
1161 
1162 		LIST_REMOVE(inp, inp_hash);
1163 		LIST_REMOVE(inp, inp_portlist);
1164 		if (LIST_FIRST(&phd->phd_pcblist) == NULL) {
1165 			LIST_REMOVE(phd, phd_hash);
1166 			free(phd, M_PCB);
1167 		}
1168 	}
1169 	LIST_REMOVE(inp, inp_list);
1170 	pcbinfo->ipi_count--;
1171 }
1172 
1173 /*
1174  * A set label operation has occurred at the socket layer, propagate the
1175  * label change into the in_pcb for the socket.
1176  */
1177 void
1178 in_pcbsosetlabel(struct socket *so)
1179 {
1180 #ifdef MAC
1181 	struct inpcb *inp;
1182 
1183 	inp = sotoinpcb(so);
1184 	KASSERT(inp != NULL, ("in_pcbsosetlabel: so->so_pcb == NULL"));
1185 
1186 	INP_WLOCK(inp);
1187 	SOCK_LOCK(so);
1188 	mac_inpcb_sosetlabel(so, inp);
1189 	SOCK_UNLOCK(so);
1190 	INP_WUNLOCK(inp);
1191 #endif
1192 }
1193 
1194 /*
1195  * ipport_tick runs once per second, determining if random port allocation
1196  * should be continued.  If more than ipport_randomcps ports have been
1197  * allocated in the last second, then we return to sequential port
1198  * allocation. We return to random allocation only once we drop below
1199  * ipport_randomcps for at least ipport_randomtime seconds.
1200  */
1201 void
1202 ipport_tick(void *xtp)
1203 {
1204 
1205 	if (V_ipport_tcpallocs <= V_ipport_tcplastcount + V_ipport_randomcps) {
1206 		if (V_ipport_stoprandom > 0)
1207 			V_ipport_stoprandom--;
1208 	} else
1209 		V_ipport_stoprandom = V_ipport_randomtime;
1210 	V_ipport_tcplastcount = V_ipport_tcpallocs;
1211 	callout_reset(&ipport_tick_callout, hz, ipport_tick, NULL);
1212 }
1213 
1214 void
1215 inp_wlock(struct inpcb *inp)
1216 {
1217 
1218 	INP_WLOCK(inp);
1219 }
1220 
1221 void
1222 inp_wunlock(struct inpcb *inp)
1223 {
1224 
1225 	INP_WUNLOCK(inp);
1226 }
1227 
1228 void
1229 inp_rlock(struct inpcb *inp)
1230 {
1231 
1232 	INP_RLOCK(inp);
1233 }
1234 
1235 void
1236 inp_runlock(struct inpcb *inp)
1237 {
1238 
1239 	INP_RUNLOCK(inp);
1240 }
1241 
1242 #ifdef INVARIANTS
1243 void
1244 inp_lock_assert(struct inpcb *inp)
1245 {
1246 
1247 	INP_WLOCK_ASSERT(inp);
1248 }
1249 
1250 void
1251 inp_unlock_assert(struct inpcb *inp)
1252 {
1253 
1254 	INP_UNLOCK_ASSERT(inp);
1255 }
1256 #endif
1257 
1258 void
1259 inp_apply_all(void (*func)(struct inpcb *, void *), void *arg)
1260 {
1261 	struct inpcb *inp;
1262 
1263 	INP_INFO_RLOCK(&V_tcbinfo);
1264 	LIST_FOREACH(inp, tcbinfo.ipi_listhead, inp_list) {
1265 		INP_WLOCK(inp);
1266 		func(inp, arg);
1267 		INP_WUNLOCK(inp);
1268 	}
1269 	INP_INFO_RUNLOCK(&V_tcbinfo);
1270 }
1271 
1272 struct socket *
1273 inp_inpcbtosocket(struct inpcb *inp)
1274 {
1275 
1276 	INP_WLOCK_ASSERT(inp);
1277 	return (inp->inp_socket);
1278 }
1279 
1280 struct tcpcb *
1281 inp_inpcbtotcpcb(struct inpcb *inp)
1282 {
1283 
1284 	INP_WLOCK_ASSERT(inp);
1285 	return ((struct tcpcb *)inp->inp_ppcb);
1286 }
1287 
1288 int
1289 inp_ip_tos_get(const struct inpcb *inp)
1290 {
1291 
1292 	return (inp->inp_ip_tos);
1293 }
1294 
1295 void
1296 inp_ip_tos_set(struct inpcb *inp, int val)
1297 {
1298 
1299 	inp->inp_ip_tos = val;
1300 }
1301 
1302 void
1303 inp_4tuple_get(struct inpcb *inp, uint32_t *laddr, uint16_t *lp,
1304     uint32_t *faddr, uint16_t *fp)
1305 {
1306 
1307 	INP_LOCK_ASSERT(inp);
1308 	*laddr = inp->inp_laddr.s_addr;
1309 	*faddr = inp->inp_faddr.s_addr;
1310 	*lp = inp->inp_lport;
1311 	*fp = inp->inp_fport;
1312 }
1313 
1314 struct inpcb *
1315 so_sotoinpcb(struct socket *so)
1316 {
1317 
1318 	return (sotoinpcb(so));
1319 }
1320 
1321 struct tcpcb *
1322 so_sototcpcb(struct socket *so)
1323 {
1324 
1325 	return (sototcpcb(so));
1326 }
1327 
1328 #ifdef DDB
1329 static void
1330 db_print_indent(int indent)
1331 {
1332 	int i;
1333 
1334 	for (i = 0; i < indent; i++)
1335 		db_printf(" ");
1336 }
1337 
1338 static void
1339 db_print_inconninfo(struct in_conninfo *inc, const char *name, int indent)
1340 {
1341 	char faddr_str[48], laddr_str[48];
1342 
1343 	db_print_indent(indent);
1344 	db_printf("%s at %p\n", name, inc);
1345 
1346 	indent += 2;
1347 
1348 #ifdef INET6
1349 	if (inc->inc_flags == 1) {
1350 		/* IPv6. */
1351 		ip6_sprintf(laddr_str, &inc->inc6_laddr);
1352 		ip6_sprintf(faddr_str, &inc->inc6_faddr);
1353 	} else {
1354 #endif
1355 		/* IPv4. */
1356 		inet_ntoa_r(inc->inc_laddr, laddr_str);
1357 		inet_ntoa_r(inc->inc_faddr, faddr_str);
1358 #ifdef INET6
1359 	}
1360 #endif
1361 	db_print_indent(indent);
1362 	db_printf("inc_laddr %s   inc_lport %u\n", laddr_str,
1363 	    ntohs(inc->inc_lport));
1364 	db_print_indent(indent);
1365 	db_printf("inc_faddr %s   inc_fport %u\n", faddr_str,
1366 	    ntohs(inc->inc_fport));
1367 }
1368 
1369 static void
1370 db_print_inpflags(int inp_flags)
1371 {
1372 	int comma;
1373 
1374 	comma = 0;
1375 	if (inp_flags & INP_RECVOPTS) {
1376 		db_printf("%sINP_RECVOPTS", comma ? ", " : "");
1377 		comma = 1;
1378 	}
1379 	if (inp_flags & INP_RECVRETOPTS) {
1380 		db_printf("%sINP_RECVRETOPTS", comma ? ", " : "");
1381 		comma = 1;
1382 	}
1383 	if (inp_flags & INP_RECVDSTADDR) {
1384 		db_printf("%sINP_RECVDSTADDR", comma ? ", " : "");
1385 		comma = 1;
1386 	}
1387 	if (inp_flags & INP_HDRINCL) {
1388 		db_printf("%sINP_HDRINCL", comma ? ", " : "");
1389 		comma = 1;
1390 	}
1391 	if (inp_flags & INP_HIGHPORT) {
1392 		db_printf("%sINP_HIGHPORT", comma ? ", " : "");
1393 		comma = 1;
1394 	}
1395 	if (inp_flags & INP_LOWPORT) {
1396 		db_printf("%sINP_LOWPORT", comma ? ", " : "");
1397 		comma = 1;
1398 	}
1399 	if (inp_flags & INP_ANONPORT) {
1400 		db_printf("%sINP_ANONPORT", comma ? ", " : "");
1401 		comma = 1;
1402 	}
1403 	if (inp_flags & INP_RECVIF) {
1404 		db_printf("%sINP_RECVIF", comma ? ", " : "");
1405 		comma = 1;
1406 	}
1407 	if (inp_flags & INP_MTUDISC) {
1408 		db_printf("%sINP_MTUDISC", comma ? ", " : "");
1409 		comma = 1;
1410 	}
1411 	if (inp_flags & INP_FAITH) {
1412 		db_printf("%sINP_FAITH", comma ? ", " : "");
1413 		comma = 1;
1414 	}
1415 	if (inp_flags & INP_RECVTTL) {
1416 		db_printf("%sINP_RECVTTL", comma ? ", " : "");
1417 		comma = 1;
1418 	}
1419 	if (inp_flags & INP_DONTFRAG) {
1420 		db_printf("%sINP_DONTFRAG", comma ? ", " : "");
1421 		comma = 1;
1422 	}
1423 	if (inp_flags & IN6P_IPV6_V6ONLY) {
1424 		db_printf("%sIN6P_IPV6_V6ONLY", comma ? ", " : "");
1425 		comma = 1;
1426 	}
1427 	if (inp_flags & IN6P_PKTINFO) {
1428 		db_printf("%sIN6P_PKTINFO", comma ? ", " : "");
1429 		comma = 1;
1430 	}
1431 	if (inp_flags & IN6P_HOPLIMIT) {
1432 		db_printf("%sIN6P_HOPLIMIT", comma ? ", " : "");
1433 		comma = 1;
1434 	}
1435 	if (inp_flags & IN6P_HOPOPTS) {
1436 		db_printf("%sIN6P_HOPOPTS", comma ? ", " : "");
1437 		comma = 1;
1438 	}
1439 	if (inp_flags & IN6P_DSTOPTS) {
1440 		db_printf("%sIN6P_DSTOPTS", comma ? ", " : "");
1441 		comma = 1;
1442 	}
1443 	if (inp_flags & IN6P_RTHDR) {
1444 		db_printf("%sIN6P_RTHDR", comma ? ", " : "");
1445 		comma = 1;
1446 	}
1447 	if (inp_flags & IN6P_RTHDRDSTOPTS) {
1448 		db_printf("%sIN6P_RTHDRDSTOPTS", comma ? ", " : "");
1449 		comma = 1;
1450 	}
1451 	if (inp_flags & IN6P_TCLASS) {
1452 		db_printf("%sIN6P_TCLASS", comma ? ", " : "");
1453 		comma = 1;
1454 	}
1455 	if (inp_flags & IN6P_AUTOFLOWLABEL) {
1456 		db_printf("%sIN6P_AUTOFLOWLABEL", comma ? ", " : "");
1457 		comma = 1;
1458 	}
1459 	if (inp_flags & IN6P_RFC2292) {
1460 		db_printf("%sIN6P_RFC2292", comma ? ", " : "");
1461 		comma = 1;
1462 	}
1463 	if (inp_flags & IN6P_MTU) {
1464 		db_printf("IN6P_MTU%s", comma ? ", " : "");
1465 		comma = 1;
1466 	}
1467 }
1468 
1469 static void
1470 db_print_inpvflag(u_char inp_vflag)
1471 {
1472 	int comma;
1473 
1474 	comma = 0;
1475 	if (inp_vflag & INP_IPV4) {
1476 		db_printf("%sINP_IPV4", comma ? ", " : "");
1477 		comma  = 1;
1478 	}
1479 	if (inp_vflag & INP_IPV6) {
1480 		db_printf("%sINP_IPV6", comma ? ", " : "");
1481 		comma  = 1;
1482 	}
1483 	if (inp_vflag & INP_IPV6PROTO) {
1484 		db_printf("%sINP_IPV6PROTO", comma ? ", " : "");
1485 		comma  = 1;
1486 	}
1487 	if (inp_vflag & INP_TIMEWAIT) {
1488 		db_printf("%sINP_TIMEWAIT", comma ? ", " : "");
1489 		comma  = 1;
1490 	}
1491 	if (inp_vflag & INP_ONESBCAST) {
1492 		db_printf("%sINP_ONESBCAST", comma ? ", " : "");
1493 		comma  = 1;
1494 	}
1495 	if (inp_vflag & INP_DROPPED) {
1496 		db_printf("%sINP_DROPPED", comma ? ", " : "");
1497 		comma  = 1;
1498 	}
1499 	if (inp_vflag & INP_SOCKREF) {
1500 		db_printf("%sINP_SOCKREF", comma ? ", " : "");
1501 		comma  = 1;
1502 	}
1503 }
1504 
1505 void
1506 db_print_inpcb(struct inpcb *inp, const char *name, int indent)
1507 {
1508 
1509 	db_print_indent(indent);
1510 	db_printf("%s at %p\n", name, inp);
1511 
1512 	indent += 2;
1513 
1514 	db_print_indent(indent);
1515 	db_printf("inp_flow: 0x%x\n", inp->inp_flow);
1516 
1517 	db_print_inconninfo(&inp->inp_inc, "inp_conninfo", indent);
1518 
1519 	db_print_indent(indent);
1520 	db_printf("inp_ppcb: %p   inp_pcbinfo: %p   inp_socket: %p\n",
1521 	    inp->inp_ppcb, inp->inp_pcbinfo, inp->inp_socket);
1522 
1523 	db_print_indent(indent);
1524 	db_printf("inp_label: %p   inp_flags: 0x%x (",
1525 	   inp->inp_label, inp->inp_flags);
1526 	db_print_inpflags(inp->inp_flags);
1527 	db_printf(")\n");
1528 
1529 	db_print_indent(indent);
1530 	db_printf("inp_sp: %p   inp_vflag: 0x%x (", inp->inp_sp,
1531 	    inp->inp_vflag);
1532 	db_print_inpvflag(inp->inp_vflag);
1533 	db_printf(")\n");
1534 
1535 	db_print_indent(indent);
1536 	db_printf("inp_ip_ttl: %d   inp_ip_p: %d   inp_ip_minttl: %d\n",
1537 	    inp->inp_ip_ttl, inp->inp_ip_p, inp->inp_ip_minttl);
1538 
1539 	db_print_indent(indent);
1540 #ifdef INET6
1541 	if (inp->inp_vflag & INP_IPV6) {
1542 		db_printf("in6p_options: %p   in6p_outputopts: %p   "
1543 		    "in6p_moptions: %p\n", inp->in6p_options,
1544 		    inp->in6p_outputopts, inp->in6p_moptions);
1545 		db_printf("in6p_icmp6filt: %p   in6p_cksum %d   "
1546 		    "in6p_hops %u\n", inp->in6p_icmp6filt, inp->in6p_cksum,
1547 		    inp->in6p_hops);
1548 	} else
1549 #endif
1550 	{
1551 		db_printf("inp_ip_tos: %d   inp_ip_options: %p   "
1552 		    "inp_ip_moptions: %p\n", inp->inp_ip_tos,
1553 		    inp->inp_options, inp->inp_moptions);
1554 	}
1555 
1556 	db_print_indent(indent);
1557 	db_printf("inp_phd: %p   inp_gencnt: %ju\n", inp->inp_phd,
1558 	    (uintmax_t)inp->inp_gencnt);
1559 }
1560 
1561 DB_SHOW_COMMAND(inpcb, db_show_inpcb)
1562 {
1563 	struct inpcb *inp;
1564 
1565 	if (!have_addr) {
1566 		db_printf("usage: show inpcb <addr>\n");
1567 		return;
1568 	}
1569 	inp = (struct inpcb *)addr;
1570 
1571 	db_print_inpcb(inp, "inpcb", 0);
1572 }
1573 #endif
1574