xref: /freebsd/sys/netinet/in_pcb.c (revision 5129159789cc9d7bc514e4546b88e3427695002d)
1 /*
2  * Copyright (c) 1982, 1986, 1991, 1993, 1995
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. All advertising materials mentioning features or use of this software
14  *    must display the following acknowledgement:
15  *	This product includes software developed by the University of
16  *	California, Berkeley and its contributors.
17  * 4. Neither the name of the University nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  *
33  *	@(#)in_pcb.c	8.4 (Berkeley) 5/24/95
34  * $FreeBSD$
35  */
36 
37 #include "opt_inet6.h"
38 
39 #include <sys/param.h>
40 #include <sys/systm.h>
41 #include <sys/malloc.h>
42 #include <sys/mbuf.h>
43 #include <sys/domain.h>
44 #include <sys/protosw.h>
45 #include <sys/socket.h>
46 #include <sys/socketvar.h>
47 #include <sys/proc.h>
48 #include <sys/jail.h>
49 #include <sys/kernel.h>
50 #include <sys/sysctl.h>
51 
52 #include <machine/limits.h>
53 
54 #include <vm/vm_zone.h>
55 
56 #include <net/if.h>
57 #include <net/if_types.h>
58 #include <net/route.h>
59 
60 #include <netinet/in.h>
61 #include <netinet/in_pcb.h>
62 #include <netinet/in_var.h>
63 #include <netinet/ip_var.h>
64 #ifdef INET6
65 #include <netinet/ip6.h>
66 #include <netinet6/ip6_var.h>
67 #endif /* INET6 */
68 
69 #include "faith.h"
70 
71 #ifdef IPSEC
72 #include <netinet6/ipsec.h>
73 #include <netkey/key.h>
74 #include <netkey/key_debug.h>
75 #endif /* IPSEC */
76 
77 struct	in_addr zeroin_addr;
78 
79 static void	in_rtchange __P((struct inpcb *, int));
80 
81 /*
82  * These configure the range of local port addresses assigned to
83  * "unspecified" outgoing connections/packets/whatever.
84  */
85 int	ipport_lowfirstauto  = IPPORT_RESERVED - 1;	/* 1023 */
86 int	ipport_lowlastauto = IPPORT_RESERVEDSTART;	/* 600 */
87 int	ipport_firstauto = IPPORT_RESERVED;		/* 1024 */
88 int	ipport_lastauto  = IPPORT_USERRESERVED;		/* 5000 */
89 int	ipport_hifirstauto = IPPORT_HIFIRSTAUTO;	/* 49152 */
90 int	ipport_hilastauto  = IPPORT_HILASTAUTO;		/* 65535 */
91 
92 #define RANGECHK(var, min, max) \
93 	if ((var) < (min)) { (var) = (min); } \
94 	else if ((var) > (max)) { (var) = (max); }
95 
96 static int
97 sysctl_net_ipport_check SYSCTL_HANDLER_ARGS
98 {
99 	int error = sysctl_handle_int(oidp,
100 		oidp->oid_arg1, oidp->oid_arg2, req);
101 	if (!error) {
102 		RANGECHK(ipport_lowfirstauto, 1, IPPORT_RESERVED - 1);
103 		RANGECHK(ipport_lowlastauto, 1, IPPORT_RESERVED - 1);
104 		RANGECHK(ipport_firstauto, IPPORT_RESERVED, USHRT_MAX);
105 		RANGECHK(ipport_lastauto, IPPORT_RESERVED, USHRT_MAX);
106 		RANGECHK(ipport_hifirstauto, IPPORT_RESERVED, USHRT_MAX);
107 		RANGECHK(ipport_hilastauto, IPPORT_RESERVED, USHRT_MAX);
108 	}
109 	return error;
110 }
111 
112 #undef RANGECHK
113 
114 SYSCTL_NODE(_net_inet_ip, IPPROTO_IP, portrange, CTLFLAG_RW, 0, "IP Ports");
115 
116 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, lowfirst, CTLTYPE_INT|CTLFLAG_RW,
117 	   &ipport_lowfirstauto, 0, &sysctl_net_ipport_check, "I", "");
118 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, lowlast, CTLTYPE_INT|CTLFLAG_RW,
119 	   &ipport_lowlastauto, 0, &sysctl_net_ipport_check, "I", "");
120 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, first, CTLTYPE_INT|CTLFLAG_RW,
121 	   &ipport_firstauto, 0, &sysctl_net_ipport_check, "I", "");
122 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, last, CTLTYPE_INT|CTLFLAG_RW,
123 	   &ipport_lastauto, 0, &sysctl_net_ipport_check, "I", "");
124 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, hifirst, CTLTYPE_INT|CTLFLAG_RW,
125 	   &ipport_hifirstauto, 0, &sysctl_net_ipport_check, "I", "");
126 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, hilast, CTLTYPE_INT|CTLFLAG_RW,
127 	   &ipport_hilastauto, 0, &sysctl_net_ipport_check, "I", "");
128 
129 /*
130  * in_pcb.c: manage the Protocol Control Blocks.
131  *
132  * NOTE: It is assumed that most of these functions will be called at
133  * splnet(). XXX - There are, unfortunately, a few exceptions to this
134  * rule that should be fixed.
135  */
136 
137 /*
138  * Allocate a PCB and associate it with the socket.
139  */
140 int
141 in_pcballoc(so, pcbinfo, p)
142 	struct socket *so;
143 	struct inpcbinfo *pcbinfo;
144 	struct proc *p;
145 {
146 	register struct inpcb *inp;
147 
148 	inp = zalloci(pcbinfo->ipi_zone);
149 	if (inp == NULL)
150 		return (ENOBUFS);
151 	bzero((caddr_t)inp, sizeof(*inp));
152 	inp->inp_gencnt = ++pcbinfo->ipi_gencnt;
153 	inp->inp_pcbinfo = pcbinfo;
154 	inp->inp_socket = so;
155 	LIST_INSERT_HEAD(pcbinfo->listhead, inp, inp_list);
156 	pcbinfo->ipi_count++;
157 	so->so_pcb = (caddr_t)inp;
158 	return (0);
159 }
160 
161 int
162 in_pcbbind(inp, nam, p)
163 	register struct inpcb *inp;
164 	struct sockaddr *nam;
165 	struct proc *p;
166 {
167 	register struct socket *so = inp->inp_socket;
168 	unsigned short *lastport;
169 	struct sockaddr_in *sin;
170 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
171 	u_short lport = 0;
172 	int wild = 0, reuseport = (so->so_options & SO_REUSEPORT);
173 	int error, prison = 0;
174 
175 	if (TAILQ_EMPTY(&in_ifaddrhead)) /* XXX broken! */
176 		return (EADDRNOTAVAIL);
177 	if (inp->inp_lport || inp->inp_laddr.s_addr != INADDR_ANY)
178 		return (EINVAL);
179 	if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0)
180 		wild = 1;
181 	if (nam) {
182 		sin = (struct sockaddr_in *)nam;
183 		if (nam->sa_len != sizeof (*sin))
184 			return (EINVAL);
185 #ifdef notdef
186 		/*
187 		 * We should check the family, but old programs
188 		 * incorrectly fail to initialize it.
189 		 */
190 		if (sin->sin_family != AF_INET)
191 			return (EAFNOSUPPORT);
192 #endif
193 		if (prison_ip(p, 0, &sin->sin_addr.s_addr))
194 			return(EINVAL);
195 		lport = sin->sin_port;
196 		if (IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) {
197 			/*
198 			 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
199 			 * allow complete duplication of binding if
200 			 * SO_REUSEPORT is set, or if SO_REUSEADDR is set
201 			 * and a multicast address is bound on both
202 			 * new and duplicated sockets.
203 			 */
204 			if (so->so_options & SO_REUSEADDR)
205 				reuseport = SO_REUSEADDR|SO_REUSEPORT;
206 		} else if (sin->sin_addr.s_addr != INADDR_ANY) {
207 			sin->sin_port = 0;		/* yech... */
208 			if (ifa_ifwithaddr((struct sockaddr *)sin) == 0)
209 				return (EADDRNOTAVAIL);
210 		}
211 		if (lport) {
212 			struct inpcb *t;
213 
214 			/* GROSS */
215 			if (ntohs(lport) < IPPORT_RESERVED && p &&
216 			    suser_xxx(0, p, PRISON_ROOT))
217 				return (EACCES);
218 			if (p && p->p_prison)
219 				prison = 1;
220 			if (so->so_cred->cr_uid != 0 &&
221 			    !IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) {
222 				t = in_pcblookup_local(inp->inp_pcbinfo,
223 				    sin->sin_addr, lport,
224 				    prison ? 0 :  INPLOOKUP_WILDCARD);
225 				if (t &&
226 				    (ntohl(sin->sin_addr.s_addr) != INADDR_ANY ||
227 				     ntohl(t->inp_laddr.s_addr) != INADDR_ANY ||
228 				     (t->inp_socket->so_options &
229 					 SO_REUSEPORT) == 0) &&
230 				    (so->so_cred->cr_uid !=
231 				     t->inp_socket->so_cred->cr_uid)) {
232 #if defined(INET6)
233 					if (ip6_mapped_addr_on == 0 ||
234 					    ntohl(sin->sin_addr.s_addr) !=
235 					    INADDR_ANY ||
236 					    ntohl(t->inp_laddr.s_addr) !=
237 					    INADDR_ANY ||
238 					    INP_SOCKAF(so) ==
239 					    INP_SOCKAF(t->inp_socket))
240 #endif /* defined(INET6) */
241 					return (EADDRINUSE);
242 				}
243 			}
244 			t = in_pcblookup_local(pcbinfo, sin->sin_addr,
245 			    lport, prison ? 0 : wild);
246 			if (t &&
247 			    (reuseport & t->inp_socket->so_options) == 0) {
248 #if defined(INET6)
249 				if (ip6_mapped_addr_on == 0 ||
250 				    ntohl(sin->sin_addr.s_addr) !=
251 				    INADDR_ANY ||
252 				    ntohl(t->inp_laddr.s_addr) !=
253 				    INADDR_ANY ||
254 				    INP_SOCKAF(so) ==
255 				    INP_SOCKAF(t->inp_socket))
256 #endif /* defined(INET6) */
257 				return (EADDRINUSE);
258 			}
259 		}
260 		inp->inp_laddr = sin->sin_addr;
261 	}
262 	if (lport == 0) {
263 		ushort first, last;
264 		int count;
265 
266 		if (prison_ip(p, 0, &inp->inp_laddr.s_addr ))
267 			return (EINVAL);
268 		inp->inp_flags |= INP_ANONPORT;
269 
270 		if (inp->inp_flags & INP_HIGHPORT) {
271 			first = ipport_hifirstauto;	/* sysctl */
272 			last  = ipport_hilastauto;
273 			lastport = &pcbinfo->lasthi;
274 		} else if (inp->inp_flags & INP_LOWPORT) {
275 			if (p && (error = suser_xxx(0, p, PRISON_ROOT)))
276 				return error;
277 			first = ipport_lowfirstauto;	/* 1023 */
278 			last  = ipport_lowlastauto;	/* 600 */
279 			lastport = &pcbinfo->lastlow;
280 		} else {
281 			first = ipport_firstauto;	/* sysctl */
282 			last  = ipport_lastauto;
283 			lastport = &pcbinfo->lastport;
284 		}
285 		/*
286 		 * Simple check to ensure all ports are not used up causing
287 		 * a deadlock here.
288 		 *
289 		 * We split the two cases (up and down) so that the direction
290 		 * is not being tested on each round of the loop.
291 		 */
292 		if (first > last) {
293 			/*
294 			 * counting down
295 			 */
296 			count = first - last;
297 
298 			do {
299 				if (count-- < 0) {	/* completely used? */
300 					/*
301 					 * Undo any address bind that may have
302 					 * occurred above.
303 					 */
304 					inp->inp_laddr.s_addr = INADDR_ANY;
305 					return (EAGAIN);
306 				}
307 				--*lastport;
308 				if (*lastport > first || *lastport < last)
309 					*lastport = first;
310 				lport = htons(*lastport);
311 			} while (in_pcblookup_local(pcbinfo,
312 				 inp->inp_laddr, lport, wild));
313 		} else {
314 			/*
315 			 * counting up
316 			 */
317 			count = last - first;
318 
319 			do {
320 				if (count-- < 0) {	/* completely used? */
321 					/*
322 					 * Undo any address bind that may have
323 					 * occurred above.
324 					 */
325 					inp->inp_laddr.s_addr = INADDR_ANY;
326 					return (EAGAIN);
327 				}
328 				++*lastport;
329 				if (*lastport < first || *lastport > last)
330 					*lastport = first;
331 				lport = htons(*lastport);
332 			} while (in_pcblookup_local(pcbinfo,
333 				 inp->inp_laddr, lport, wild));
334 		}
335 	}
336 	inp->inp_lport = lport;
337 	if (in_pcbinshash(inp) != 0) {
338 		inp->inp_laddr.s_addr = INADDR_ANY;
339 		inp->inp_lport = 0;
340 		return (EAGAIN);
341 	}
342 	return (0);
343 }
344 
345 /*
346  *   Transform old in_pcbconnect() into an inner subroutine for new
347  *   in_pcbconnect(): Do some validity-checking on the remote
348  *   address (in mbuf 'nam') and then determine local host address
349  *   (i.e., which interface) to use to access that remote host.
350  *
351  *   This preserves definition of in_pcbconnect(), while supporting a
352  *   slightly different version for T/TCP.  (This is more than
353  *   a bit of a kludge, but cleaning up the internal interfaces would
354  *   have forced minor changes in every protocol).
355  */
356 
357 int
358 in_pcbladdr(inp, nam, plocal_sin)
359 	register struct inpcb *inp;
360 	struct sockaddr *nam;
361 	struct sockaddr_in **plocal_sin;
362 {
363 	struct in_ifaddr *ia;
364 	register struct sockaddr_in *sin = (struct sockaddr_in *)nam;
365 
366 	if (nam->sa_len != sizeof (*sin))
367 		return (EINVAL);
368 	if (sin->sin_family != AF_INET)
369 		return (EAFNOSUPPORT);
370 	if (sin->sin_port == 0)
371 		return (EADDRNOTAVAIL);
372 	if (!TAILQ_EMPTY(&in_ifaddrhead)) {
373 		/*
374 		 * If the destination address is INADDR_ANY,
375 		 * use the primary local address.
376 		 * If the supplied address is INADDR_BROADCAST,
377 		 * and the primary interface supports broadcast,
378 		 * choose the broadcast address for that interface.
379 		 */
380 #define	satosin(sa)	((struct sockaddr_in *)(sa))
381 #define sintosa(sin)	((struct sockaddr *)(sin))
382 #define ifatoia(ifa)	((struct in_ifaddr *)(ifa))
383 		if (sin->sin_addr.s_addr == INADDR_ANY)
384 		    sin->sin_addr = IA_SIN(in_ifaddrhead.tqh_first)->sin_addr;
385 		else if (sin->sin_addr.s_addr == (u_long)INADDR_BROADCAST &&
386 		  (in_ifaddrhead.tqh_first->ia_ifp->if_flags & IFF_BROADCAST))
387 		    sin->sin_addr = satosin(&in_ifaddrhead.tqh_first->ia_broadaddr)->sin_addr;
388 	}
389 	if (inp->inp_laddr.s_addr == INADDR_ANY) {
390 		register struct route *ro;
391 
392 		ia = (struct in_ifaddr *)0;
393 		/*
394 		 * If route is known or can be allocated now,
395 		 * our src addr is taken from the i/f, else punt.
396 		 */
397 		ro = &inp->inp_route;
398 		if (ro->ro_rt &&
399 		    (satosin(&ro->ro_dst)->sin_addr.s_addr !=
400 			sin->sin_addr.s_addr ||
401 		    inp->inp_socket->so_options & SO_DONTROUTE)) {
402 			RTFREE(ro->ro_rt);
403 			ro->ro_rt = (struct rtentry *)0;
404 		}
405 		if ((inp->inp_socket->so_options & SO_DONTROUTE) == 0 && /*XXX*/
406 		    (ro->ro_rt == (struct rtentry *)0 ||
407 		    ro->ro_rt->rt_ifp == (struct ifnet *)0)) {
408 			/* No route yet, so try to acquire one */
409 			ro->ro_dst.sa_family = AF_INET;
410 			ro->ro_dst.sa_len = sizeof(struct sockaddr_in);
411 			((struct sockaddr_in *) &ro->ro_dst)->sin_addr =
412 				sin->sin_addr;
413 			rtalloc(ro);
414 		}
415 		/*
416 		 * If we found a route, use the address
417 		 * corresponding to the outgoing interface
418 		 * unless it is the loopback (in case a route
419 		 * to our address on another net goes to loopback).
420 		 */
421 		if (ro->ro_rt && !(ro->ro_rt->rt_ifp->if_flags & IFF_LOOPBACK))
422 			ia = ifatoia(ro->ro_rt->rt_ifa);
423 		if (ia == 0) {
424 			u_short fport = sin->sin_port;
425 
426 			sin->sin_port = 0;
427 			ia = ifatoia(ifa_ifwithdstaddr(sintosa(sin)));
428 			if (ia == 0)
429 				ia = ifatoia(ifa_ifwithnet(sintosa(sin)));
430 			sin->sin_port = fport;
431 			if (ia == 0)
432 				ia = in_ifaddrhead.tqh_first;
433 			if (ia == 0)
434 				return (EADDRNOTAVAIL);
435 		}
436 		/*
437 		 * If the destination address is multicast and an outgoing
438 		 * interface has been set as a multicast option, use the
439 		 * address of that interface as our source address.
440 		 */
441 		if (IN_MULTICAST(ntohl(sin->sin_addr.s_addr)) &&
442 		    inp->inp_moptions != NULL) {
443 			struct ip_moptions *imo;
444 			struct ifnet *ifp;
445 
446 			imo = inp->inp_moptions;
447 			if (imo->imo_multicast_ifp != NULL) {
448 				ifp = imo->imo_multicast_ifp;
449 				for (ia = in_ifaddrhead.tqh_first; ia;
450 				     ia = ia->ia_link.tqe_next)
451 					if (ia->ia_ifp == ifp)
452 						break;
453 				if (ia == 0)
454 					return (EADDRNOTAVAIL);
455 			}
456 		}
457 	/*
458 	 * Don't do pcblookup call here; return interface in plocal_sin
459 	 * and exit to caller, that will do the lookup.
460 	 */
461 		*plocal_sin = &ia->ia_addr;
462 
463 	}
464 	return(0);
465 }
466 
467 /*
468  * Outer subroutine:
469  * Connect from a socket to a specified address.
470  * Both address and port must be specified in argument sin.
471  * If don't have a local address for this socket yet,
472  * then pick one.
473  */
474 int
475 in_pcbconnect(inp, nam, p)
476 	register struct inpcb *inp;
477 	struct sockaddr *nam;
478 	struct proc *p;
479 {
480 	struct sockaddr_in *ifaddr;
481 	register struct sockaddr_in *sin = (struct sockaddr_in *)nam;
482 	int error;
483 
484 	/*
485 	 *   Call inner routine, to assign local interface address.
486 	 */
487 	if ((error = in_pcbladdr(inp, nam, &ifaddr)) != 0)
488 		return(error);
489 
490 	if (in_pcblookup_hash(inp->inp_pcbinfo, sin->sin_addr, sin->sin_port,
491 	    inp->inp_laddr.s_addr ? inp->inp_laddr : ifaddr->sin_addr,
492 	    inp->inp_lport, 0, NULL) != NULL) {
493 		return (EADDRINUSE);
494 	}
495 	if (inp->inp_laddr.s_addr == INADDR_ANY) {
496 		if (inp->inp_lport == 0) {
497 			error = in_pcbbind(inp, (struct sockaddr *)0, p);
498 			if (error)
499 			    return (error);
500 		}
501 		inp->inp_laddr = ifaddr->sin_addr;
502 	}
503 	inp->inp_faddr = sin->sin_addr;
504 	inp->inp_fport = sin->sin_port;
505 	in_pcbrehash(inp);
506 	return (0);
507 }
508 
509 void
510 in_pcbdisconnect(inp)
511 	struct inpcb *inp;
512 {
513 
514 	inp->inp_faddr.s_addr = INADDR_ANY;
515 	inp->inp_fport = 0;
516 	in_pcbrehash(inp);
517 	if (inp->inp_socket->so_state & SS_NOFDREF)
518 		in_pcbdetach(inp);
519 }
520 
521 void
522 in_pcbdetach(inp)
523 	struct inpcb *inp;
524 {
525 	struct socket *so = inp->inp_socket;
526 	struct inpcbinfo *ipi = inp->inp_pcbinfo;
527 
528 #ifdef IPSEC
529 	ipsec4_delete_pcbpolicy(inp);
530 #endif /*IPSEC*/
531 	inp->inp_gencnt = ++ipi->ipi_gencnt;
532 	in_pcbremlists(inp);
533 	so->so_pcb = 0;
534 	sofree(so);
535 	if (inp->inp_options)
536 		(void)m_free(inp->inp_options);
537 	if (inp->inp_route.ro_rt)
538 		rtfree(inp->inp_route.ro_rt);
539 	ip_freemoptions(inp->inp_moptions);
540 	inp->inp_vflag = 0;
541 	zfreei(ipi->ipi_zone, inp);
542 }
543 
544 /*
545  * The calling convention of in_setsockaddr() and in_setpeeraddr() was
546  * modified to match the pru_sockaddr() and pru_peeraddr() entry points
547  * in struct pr_usrreqs, so that protocols can just reference then directly
548  * without the need for a wrapper function.  The socket must have a valid
549  * (i.e., non-nil) PCB, but it should be impossible to get an invalid one
550  * except through a kernel programming error, so it is acceptable to panic
551  * (or in this case trap) if the PCB is invalid.  (Actually, we don't trap
552  * because there actually /is/ a programming error somewhere... XXX)
553  */
554 int
555 in_setsockaddr(so, nam)
556 	struct socket *so;
557 	struct sockaddr **nam;
558 {
559 	int s;
560 	register struct inpcb *inp;
561 	register struct sockaddr_in *sin;
562 
563 	/*
564 	 * Do the malloc first in case it blocks.
565 	 */
566 	MALLOC(sin, struct sockaddr_in *, sizeof *sin, M_SONAME, M_WAITOK);
567 	bzero(sin, sizeof *sin);
568 	sin->sin_family = AF_INET;
569 	sin->sin_len = sizeof(*sin);
570 
571 	s = splnet();
572 	inp = sotoinpcb(so);
573 	if (!inp) {
574 		splx(s);
575 		free(sin, M_SONAME);
576 		return EINVAL;
577 	}
578 	sin->sin_port = inp->inp_lport;
579 	sin->sin_addr = inp->inp_laddr;
580 	splx(s);
581 
582 	*nam = (struct sockaddr *)sin;
583 	return 0;
584 }
585 
586 int
587 in_setpeeraddr(so, nam)
588 	struct socket *so;
589 	struct sockaddr **nam;
590 {
591 	int s;
592 	struct inpcb *inp;
593 	register struct sockaddr_in *sin;
594 
595 	/*
596 	 * Do the malloc first in case it blocks.
597 	 */
598 	MALLOC(sin, struct sockaddr_in *, sizeof *sin, M_SONAME, M_WAITOK);
599 	bzero((caddr_t)sin, sizeof (*sin));
600 	sin->sin_family = AF_INET;
601 	sin->sin_len = sizeof(*sin);
602 
603 	s = splnet();
604 	inp = sotoinpcb(so);
605 	if (!inp) {
606 		splx(s);
607 		free(sin, M_SONAME);
608 		return EINVAL;
609 	}
610 	sin->sin_port = inp->inp_fport;
611 	sin->sin_addr = inp->inp_faddr;
612 	splx(s);
613 
614 	*nam = (struct sockaddr *)sin;
615 	return 0;
616 }
617 
618 /*
619  * Pass some notification to all connections of a protocol
620  * associated with address dst.  The local address and/or port numbers
621  * may be specified to limit the search.  The "usual action" will be
622  * taken, depending on the ctlinput cmd.  The caller must filter any
623  * cmds that are uninteresting (e.g., no error in the map).
624  * Call the protocol specific routine (if any) to report
625  * any errors for each matching socket.
626  */
627 void
628 in_pcbnotify(head, dst, fport_arg, laddr, lport_arg, cmd, notify)
629 	struct inpcbhead *head;
630 	struct sockaddr *dst;
631 	u_int fport_arg, lport_arg;
632 	struct in_addr laddr;
633 	int cmd;
634 	void (*notify) __P((struct inpcb *, int));
635 {
636 	register struct inpcb *inp, *oinp;
637 	struct in_addr faddr;
638 	u_short fport = fport_arg, lport = lport_arg;
639 	int errno, s;
640 
641 	if ((unsigned)cmd > PRC_NCMDS || dst->sa_family != AF_INET)
642 		return;
643 	faddr = ((struct sockaddr_in *)dst)->sin_addr;
644 	if (faddr.s_addr == INADDR_ANY)
645 		return;
646 
647 	/*
648 	 * Redirects go to all references to the destination,
649 	 * and use in_rtchange to invalidate the route cache.
650 	 * Dead host indications: notify all references to the destination.
651 	 * Otherwise, if we have knowledge of the local port and address,
652 	 * deliver only to that socket.
653 	 */
654 	if (PRC_IS_REDIRECT(cmd) || cmd == PRC_HOSTDEAD) {
655 		fport = 0;
656 		lport = 0;
657 		laddr.s_addr = 0;
658 		if (cmd != PRC_HOSTDEAD)
659 			notify = in_rtchange;
660 	}
661 	errno = inetctlerrmap[cmd];
662 	s = splnet();
663 	for (inp = head->lh_first; inp != NULL;) {
664 #ifdef INET6
665 		if ((inp->inp_vflag & INP_IPV4) == NULL) {
666 			inp = LIST_NEXT(inp, inp_list);
667 			continue;
668 		}
669 #endif
670 		if (inp->inp_faddr.s_addr != faddr.s_addr ||
671 		    inp->inp_socket == 0 ||
672 		    (lport && inp->inp_lport != lport) ||
673 		    (laddr.s_addr && inp->inp_laddr.s_addr != laddr.s_addr) ||
674 		    (fport && inp->inp_fport != fport)) {
675 			inp = inp->inp_list.le_next;
676 			continue;
677 		}
678 		oinp = inp;
679 		inp = inp->inp_list.le_next;
680 		if (notify)
681 			(*notify)(oinp, errno);
682 	}
683 	splx(s);
684 }
685 
686 /*
687  * Check for alternatives when higher level complains
688  * about service problems.  For now, invalidate cached
689  * routing information.  If the route was created dynamically
690  * (by a redirect), time to try a default gateway again.
691  */
692 void
693 in_losing(inp)
694 	struct inpcb *inp;
695 {
696 	register struct rtentry *rt;
697 	struct rt_addrinfo info;
698 
699 	if ((rt = inp->inp_route.ro_rt)) {
700 		inp->inp_route.ro_rt = 0;
701 		bzero((caddr_t)&info, sizeof(info));
702 		info.rti_info[RTAX_DST] =
703 			(struct sockaddr *)&inp->inp_route.ro_dst;
704 		info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
705 		info.rti_info[RTAX_NETMASK] = rt_mask(rt);
706 		rt_missmsg(RTM_LOSING, &info, rt->rt_flags, 0);
707 		if (rt->rt_flags & RTF_DYNAMIC)
708 			(void) rtrequest(RTM_DELETE, rt_key(rt),
709 				rt->rt_gateway, rt_mask(rt), rt->rt_flags,
710 				(struct rtentry **)0);
711 		else
712 		/*
713 		 * A new route can be allocated
714 		 * the next time output is attempted.
715 		 */
716 			rtfree(rt);
717 	}
718 }
719 
720 /*
721  * After a routing change, flush old routing
722  * and allocate a (hopefully) better one.
723  */
724 static void
725 in_rtchange(inp, errno)
726 	register struct inpcb *inp;
727 	int errno;
728 {
729 	if (inp->inp_route.ro_rt) {
730 		rtfree(inp->inp_route.ro_rt);
731 		inp->inp_route.ro_rt = 0;
732 		/*
733 		 * A new route can be allocated the next time
734 		 * output is attempted.
735 		 */
736 	}
737 }
738 
739 /*
740  * Lookup a PCB based on the local address and port.
741  */
742 struct inpcb *
743 in_pcblookup_local(pcbinfo, laddr, lport_arg, wild_okay)
744 	struct inpcbinfo *pcbinfo;
745 	struct in_addr laddr;
746 	u_int lport_arg;
747 	int wild_okay;
748 {
749 	register struct inpcb *inp;
750 	int matchwild = 3, wildcard;
751 	u_short lport = lport_arg;
752 
753 	if (!wild_okay) {
754 		struct inpcbhead *head;
755 		/*
756 		 * Look for an unconnected (wildcard foreign addr) PCB that
757 		 * matches the local address and port we're looking for.
758 		 */
759 		head = &pcbinfo->hashbase[INP_PCBHASH(INADDR_ANY, lport, 0, pcbinfo->hashmask)];
760 		for (inp = head->lh_first; inp != NULL; inp = inp->inp_hash.le_next) {
761 #ifdef INET6
762 			if ((inp->inp_vflag & INP_IPV4) == NULL)
763 				continue;
764 #endif
765 			if (inp->inp_faddr.s_addr == INADDR_ANY &&
766 			    inp->inp_laddr.s_addr == laddr.s_addr &&
767 			    inp->inp_lport == lport) {
768 				/*
769 				 * Found.
770 				 */
771 				return (inp);
772 			}
773 		}
774 		/*
775 		 * Not found.
776 		 */
777 		return (NULL);
778 	} else {
779 		struct inpcbporthead *porthash;
780 		struct inpcbport *phd;
781 		struct inpcb *match = NULL;
782 		/*
783 		 * Best fit PCB lookup.
784 		 *
785 		 * First see if this local port is in use by looking on the
786 		 * port hash list.
787 		 */
788 		porthash = &pcbinfo->porthashbase[INP_PCBPORTHASH(lport,
789 		    pcbinfo->porthashmask)];
790 		for (phd = porthash->lh_first; phd != NULL; phd = phd->phd_hash.le_next) {
791 			if (phd->phd_port == lport)
792 				break;
793 		}
794 		if (phd != NULL) {
795 			/*
796 			 * Port is in use by one or more PCBs. Look for best
797 			 * fit.
798 			 */
799 			for (inp = phd->phd_pcblist.lh_first; inp != NULL;
800 			    inp = inp->inp_portlist.le_next) {
801 				wildcard = 0;
802 #ifdef INET6
803 				if ((inp->inp_vflag & INP_IPV4) == NULL)
804 					continue;
805 #endif
806 				if (inp->inp_faddr.s_addr != INADDR_ANY)
807 					wildcard++;
808 				if (inp->inp_laddr.s_addr != INADDR_ANY) {
809 					if (laddr.s_addr == INADDR_ANY)
810 						wildcard++;
811 					else if (inp->inp_laddr.s_addr != laddr.s_addr)
812 						continue;
813 				} else {
814 					if (laddr.s_addr != INADDR_ANY)
815 						wildcard++;
816 				}
817 				if (wildcard < matchwild) {
818 					match = inp;
819 					matchwild = wildcard;
820 					if (matchwild == 0) {
821 						break;
822 					}
823 				}
824 			}
825 		}
826 		return (match);
827 	}
828 }
829 
830 /*
831  * Lookup PCB in hash list.
832  */
833 struct inpcb *
834 in_pcblookup_hash(pcbinfo, faddr, fport_arg, laddr, lport_arg, wildcard,
835 		  ifp)
836 	struct inpcbinfo *pcbinfo;
837 	struct in_addr faddr, laddr;
838 	u_int fport_arg, lport_arg;
839 	int wildcard;
840 	struct ifnet *ifp;
841 {
842 	struct inpcbhead *head;
843 	register struct inpcb *inp;
844 	u_short fport = fport_arg, lport = lport_arg;
845 
846 	/*
847 	 * First look for an exact match.
848 	 */
849 	head = &pcbinfo->hashbase[INP_PCBHASH(faddr.s_addr, lport, fport, pcbinfo->hashmask)];
850 	for (inp = head->lh_first; inp != NULL; inp = inp->inp_hash.le_next) {
851 #ifdef INET6
852 		if ((inp->inp_vflag & INP_IPV4) == NULL)
853 			continue;
854 #endif
855 		if (inp->inp_faddr.s_addr == faddr.s_addr &&
856 		    inp->inp_laddr.s_addr == laddr.s_addr &&
857 		    inp->inp_fport == fport &&
858 		    inp->inp_lport == lport) {
859 			/*
860 			 * Found.
861 			 */
862 			return (inp);
863 		}
864 	}
865 	if (wildcard) {
866 		struct inpcb *local_wild = NULL;
867 #if defined(INET6)
868 		struct inpcb *local_wild_mapped = NULL;
869 #endif /* defined(INET6) */
870 
871 		head = &pcbinfo->hashbase[INP_PCBHASH(INADDR_ANY, lport, 0, pcbinfo->hashmask)];
872 		for (inp = head->lh_first; inp != NULL; inp = inp->inp_hash.le_next) {
873 #ifdef INET6
874 			if ((inp->inp_vflag & INP_IPV4) == NULL)
875 				continue;
876 #endif
877 			if (inp->inp_faddr.s_addr == INADDR_ANY &&
878 			    inp->inp_lport == lport) {
879 #if defined(NFAITH) && NFAITH > 0
880 				if (ifp && ifp->if_type == IFT_FAITH &&
881 				    (inp->inp_flags & INP_FAITH) == 0)
882 					continue;
883 #endif
884 				if (inp->inp_laddr.s_addr == laddr.s_addr)
885 					return (inp);
886 				else if (inp->inp_laddr.s_addr == INADDR_ANY) {
887 #if defined(INET6)
888 					if (INP_CHECK_SOCKAF(inp->inp_socket,
889 							     AF_INET6))
890 						local_wild_mapped = inp;
891 					else
892 #endif /* defined(INET6) */
893 					local_wild = inp;
894 				}
895 			}
896 		}
897 #if defined(INET6)
898 		if (local_wild == NULL)
899 			return (local_wild_mapped);
900 #endif /* defined(INET6) */
901 		return (local_wild);
902 	}
903 
904 	/*
905 	 * Not found.
906 	 */
907 	return (NULL);
908 }
909 
910 /*
911  * Insert PCB onto various hash lists.
912  */
913 int
914 in_pcbinshash(inp)
915 	struct inpcb *inp;
916 {
917 	struct inpcbhead *pcbhash;
918 	struct inpcbporthead *pcbporthash;
919 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
920 	struct inpcbport *phd;
921 	u_int32_t hashkey_faddr;
922 
923 #ifdef INET6
924 	if (inp->inp_vflag & INP_IPV6)
925 		hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX */;
926 	else
927 #endif /* INET6 */
928 	hashkey_faddr = inp->inp_faddr.s_addr;
929 
930 	pcbhash = &pcbinfo->hashbase[INP_PCBHASH(hashkey_faddr,
931 		 inp->inp_lport, inp->inp_fport, pcbinfo->hashmask)];
932 
933 	pcbporthash = &pcbinfo->porthashbase[INP_PCBPORTHASH(inp->inp_lport,
934 	    pcbinfo->porthashmask)];
935 
936 	/*
937 	 * Go through port list and look for a head for this lport.
938 	 */
939 	for (phd = pcbporthash->lh_first; phd != NULL; phd = phd->phd_hash.le_next) {
940 		if (phd->phd_port == inp->inp_lport)
941 			break;
942 	}
943 	/*
944 	 * If none exists, malloc one and tack it on.
945 	 */
946 	if (phd == NULL) {
947 		MALLOC(phd, struct inpcbport *, sizeof(struct inpcbport), M_PCB, M_NOWAIT);
948 		if (phd == NULL) {
949 			return (ENOBUFS); /* XXX */
950 		}
951 		phd->phd_port = inp->inp_lport;
952 		LIST_INIT(&phd->phd_pcblist);
953 		LIST_INSERT_HEAD(pcbporthash, phd, phd_hash);
954 	}
955 	inp->inp_phd = phd;
956 	LIST_INSERT_HEAD(&phd->phd_pcblist, inp, inp_portlist);
957 	LIST_INSERT_HEAD(pcbhash, inp, inp_hash);
958 	return (0);
959 }
960 
961 /*
962  * Move PCB to the proper hash bucket when { faddr, fport } have  been
963  * changed. NOTE: This does not handle the case of the lport changing (the
964  * hashed port list would have to be updated as well), so the lport must
965  * not change after in_pcbinshash() has been called.
966  */
967 void
968 in_pcbrehash(inp)
969 	struct inpcb *inp;
970 {
971 	struct inpcbhead *head;
972 	u_int32_t hashkey_faddr;
973 
974 #ifdef INET6
975 	if (inp->inp_vflag & INP_IPV6)
976 		hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX */;
977 	else
978 #endif /* INET6 */
979 	hashkey_faddr = inp->inp_faddr.s_addr;
980 
981 	head = &inp->inp_pcbinfo->hashbase[INP_PCBHASH(hashkey_faddr,
982 		inp->inp_lport, inp->inp_fport, inp->inp_pcbinfo->hashmask)];
983 
984 	LIST_REMOVE(inp, inp_hash);
985 	LIST_INSERT_HEAD(head, inp, inp_hash);
986 }
987 
988 /*
989  * Remove PCB from various lists.
990  */
991 void
992 in_pcbremlists(inp)
993 	struct inpcb *inp;
994 {
995 	inp->inp_gencnt = ++inp->inp_pcbinfo->ipi_gencnt;
996 	if (inp->inp_lport) {
997 		struct inpcbport *phd = inp->inp_phd;
998 
999 		LIST_REMOVE(inp, inp_hash);
1000 		LIST_REMOVE(inp, inp_portlist);
1001 		if (phd->phd_pcblist.lh_first == NULL) {
1002 			LIST_REMOVE(phd, phd_hash);
1003 			free(phd, M_PCB);
1004 		}
1005 	}
1006 	LIST_REMOVE(inp, inp_list);
1007 	inp->inp_pcbinfo->ipi_count--;
1008 }
1009 
1010 int
1011 prison_xinpcb(struct proc *p, struct inpcb *inp)
1012 {
1013 	if (!p->p_prison)
1014 		return (0);
1015 	if (ntohl(inp->inp_laddr.s_addr) == p->p_prison->pr_ip)
1016 		return (0);
1017 	return (1);
1018 }
1019