FreeBSD/Linux Kernel Cross Reference
sys/netinet/in_pcb.c
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 * 4. Neither the name of the University nor the names of its contributors
14 * may be used to endorse or promote products derived from this software
15 * without specific prior written permission.
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27 * SUCH DAMAGE.
28 *
29 * @(#)in_pcb.c 8.4 (Berkeley) 5/24/95
30 * $FreeBSD: releng/5.3/sys/netinet/in_pcb.c 136738 2004-10-21 09:30:48Z rwatson $
31 */
32
33 #include "opt_ipsec.h"
34 #include "opt_inet6.h"
35 #include "opt_mac.h"
36
37 #include <sys/param.h>
38 #include <sys/systm.h>
39 #include <sys/mac.h>
40 #include <sys/malloc.h>
41 #include <sys/mbuf.h>
42 #include <sys/domain.h>
43 #include <sys/protosw.h>
44 #include <sys/socket.h>
45 #include <sys/socketvar.h>
46 #include <sys/proc.h>
47 #include <sys/jail.h>
48 #include <sys/kernel.h>
49 #include <sys/sysctl.h>
50
51 #include <vm/uma.h>
52
53 #include <net/if.h>
54 #include <net/if_types.h>
55 #include <net/route.h>
56
57 #include <netinet/in.h>
58 #include <netinet/in_pcb.h>
59 #include <netinet/in_var.h>
60 #include <netinet/ip_var.h>
61 #include <netinet/tcp_var.h>
62 #ifdef INET6
63 #include <netinet/ip6.h>
64 #include <netinet6/ip6_var.h>
65 #endif /* INET6 */
66
67 #ifdef IPSEC
68 #include <netinet6/ipsec.h>
69 #include <netkey/key.h>
70 #endif /* IPSEC */
71
72 #ifdef FAST_IPSEC
73 #if defined(IPSEC) || defined(IPSEC_ESP)
74 #error "Bad idea: don't compile with both IPSEC and FAST_IPSEC!"
75 #endif
76
77 #include <netipsec/ipsec.h>
78 #include <netipsec/key.h>
79 #endif /* FAST_IPSEC */
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_HIFIRSTAUTO; /* 49152 */
88 int ipport_lastauto = IPPORT_HILASTAUTO; /* 65535 */
89 int ipport_hifirstauto = IPPORT_HIFIRSTAUTO; /* 49152 */
90 int ipport_hilastauto = IPPORT_HILASTAUTO; /* 65535 */
91
92 /*
93 * Reserved ports accessible only to root. There are significant
94 * security considerations that must be accounted for when changing these,
95 * but the security benefits can be great. Please be careful.
96 */
97 int ipport_reservedhigh = IPPORT_RESERVED - 1; /* 1023 */
98 int ipport_reservedlow = 0;
99
100 /* Shall we allocate ephemeral ports in random order? */
101 int ipport_randomized = 1;
102
103 #define RANGECHK(var, min, max) \
104 if ((var) < (min)) { (var) = (min); } \
105 else if ((var) > (max)) { (var) = (max); }
106
107 static int
108 sysctl_net_ipport_check(SYSCTL_HANDLER_ARGS)
109 {
110 int error;
111
112 error = sysctl_handle_int(oidp, oidp->oid_arg1, oidp->oid_arg2, req);
113 if (error == 0) {
114 RANGECHK(ipport_lowfirstauto, 1, IPPORT_RESERVED - 1);
115 RANGECHK(ipport_lowlastauto, 1, IPPORT_RESERVED - 1);
116 RANGECHK(ipport_firstauto, IPPORT_RESERVED, IPPORT_MAX);
117 RANGECHK(ipport_lastauto, IPPORT_RESERVED, IPPORT_MAX);
118 RANGECHK(ipport_hifirstauto, IPPORT_RESERVED, IPPORT_MAX);
119 RANGECHK(ipport_hilastauto, IPPORT_RESERVED, IPPORT_MAX);
120 }
121 return (error);
122 }
123
124 #undef RANGECHK
125
126 SYSCTL_NODE(_net_inet_ip, IPPROTO_IP, portrange, CTLFLAG_RW, 0, "IP Ports");
127
128 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, lowfirst, CTLTYPE_INT|CTLFLAG_RW,
129 &ipport_lowfirstauto, 0, &sysctl_net_ipport_check, "I", "");
130 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, lowlast, CTLTYPE_INT|CTLFLAG_RW,
131 &ipport_lowlastauto, 0, &sysctl_net_ipport_check, "I", "");
132 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, first, CTLTYPE_INT|CTLFLAG_RW,
133 &ipport_firstauto, 0, &sysctl_net_ipport_check, "I", "");
134 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, last, CTLTYPE_INT|CTLFLAG_RW,
135 &ipport_lastauto, 0, &sysctl_net_ipport_check, "I", "");
136 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, hifirst, CTLTYPE_INT|CTLFLAG_RW,
137 &ipport_hifirstauto, 0, &sysctl_net_ipport_check, "I", "");
138 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, hilast, CTLTYPE_INT|CTLFLAG_RW,
139 &ipport_hilastauto, 0, &sysctl_net_ipport_check, "I", "");
140 SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, reservedhigh,
141 CTLFLAG_RW|CTLFLAG_SECURE, &ipport_reservedhigh, 0, "");
142 SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, reservedlow,
143 CTLFLAG_RW|CTLFLAG_SECURE, &ipport_reservedlow, 0, "");
144 SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, randomized,
145 CTLFLAG_RW, &ipport_randomized, 0, "");
146
147 /*
148 * in_pcb.c: manage the Protocol Control Blocks.
149 *
150 * NOTE: It is assumed that most of these functions will be called at
151 * splnet(). XXX - There are, unfortunately, a few exceptions to this
152 * rule that should be fixed.
153 */
154
155 /*
156 * Allocate a PCB and associate it with the socket.
157 */
158 int
159 in_pcballoc(so, pcbinfo, type)
160 struct socket *so;
161 struct inpcbinfo *pcbinfo;
162 const char *type;
163 {
164 register struct inpcb *inp;
165 int error;
166
167 INP_INFO_WLOCK_ASSERT(pcbinfo);
168 error = 0;
169 inp = uma_zalloc(pcbinfo->ipi_zone, M_NOWAIT | M_ZERO);
170 if (inp == NULL)
171 return (ENOBUFS);
172 inp->inp_gencnt = ++pcbinfo->ipi_gencnt;
173 inp->inp_pcbinfo = pcbinfo;
174 inp->inp_socket = so;
175 #ifdef MAC
176 error = mac_init_inpcb(inp, M_NOWAIT);
177 if (error != 0)
178 goto out;
179 SOCK_LOCK(so);
180 mac_create_inpcb_from_socket(so, inp);
181 SOCK_UNLOCK(so);
182 #endif
183 #if defined(IPSEC) || defined(FAST_IPSEC)
184 #ifdef FAST_IPSEC
185 error = ipsec_init_policy(so, &inp->inp_sp);
186 #else
187 error = ipsec_init_pcbpolicy(so, &inp->inp_sp);
188 #endif
189 if (error != 0)
190 goto out;
191 #endif /*IPSEC*/
192 #if defined(INET6)
193 if (INP_SOCKAF(so) == AF_INET6) {
194 inp->inp_vflag |= INP_IPV6PROTO;
195 if (ip6_v6only)
196 inp->inp_flags |= IN6P_IPV6_V6ONLY;
197 }
198 #endif
199 LIST_INSERT_HEAD(pcbinfo->listhead, inp, inp_list);
200 pcbinfo->ipi_count++;
201 so->so_pcb = (caddr_t)inp;
202 INP_LOCK_INIT(inp, "inp", type);
203 #ifdef INET6
204 if (ip6_auto_flowlabel)
205 inp->inp_flags |= IN6P_AUTOFLOWLABEL;
206 #endif
207 #if defined(IPSEC) || defined(FAST_IPSEC) || defined(MAC)
208 out:
209 if (error != 0)
210 uma_zfree(pcbinfo->ipi_zone, inp);
211 #endif
212 return (error);
213 }
214
215 int
216 in_pcbbind(inp, nam, cred)
217 register struct inpcb *inp;
218 struct sockaddr *nam;
219 struct ucred *cred;
220 {
221 int anonport, error;
222
223 INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
224 INP_LOCK_ASSERT(inp);
225
226 if (inp->inp_lport != 0 || inp->inp_laddr.s_addr != INADDR_ANY)
227 return (EINVAL);
228 anonport = inp->inp_lport == 0 && (nam == NULL ||
229 ((struct sockaddr_in *)nam)->sin_port == 0);
230 error = in_pcbbind_setup(inp, nam, &inp->inp_laddr.s_addr,
231 &inp->inp_lport, cred);
232 if (error)
233 return (error);
234 if (in_pcbinshash(inp) != 0) {
235 inp->inp_laddr.s_addr = INADDR_ANY;
236 inp->inp_lport = 0;
237 return (EAGAIN);
238 }
239 if (anonport)
240 inp->inp_flags |= INP_ANONPORT;
241 return (0);
242 }
243
244 /*
245 * Set up a bind operation on a PCB, performing port allocation
246 * as required, but do not actually modify the PCB. Callers can
247 * either complete the bind by setting inp_laddr/inp_lport and
248 * calling in_pcbinshash(), or they can just use the resulting
249 * port and address to authorise the sending of a once-off packet.
250 *
251 * On error, the values of *laddrp and *lportp are not changed.
252 */
253 int
254 in_pcbbind_setup(inp, nam, laddrp, lportp, cred)
255 struct inpcb *inp;
256 struct sockaddr *nam;
257 in_addr_t *laddrp;
258 u_short *lportp;
259 struct ucred *cred;
260 {
261 struct socket *so = inp->inp_socket;
262 unsigned short *lastport;
263 struct sockaddr_in *sin;
264 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
265 struct in_addr laddr;
266 u_short lport = 0;
267 int wild = 0, reuseport = (so->so_options & SO_REUSEPORT);
268 int error, prison = 0;
269
270 INP_INFO_WLOCK_ASSERT(pcbinfo);
271 INP_LOCK_ASSERT(inp);
272
273 if (TAILQ_EMPTY(&in_ifaddrhead)) /* XXX broken! */
274 return (EADDRNOTAVAIL);
275 laddr.s_addr = *laddrp;
276 if (nam != NULL && laddr.s_addr != INADDR_ANY)
277 return (EINVAL);
278 if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0)
279 wild = 1;
280 if (nam) {
281 sin = (struct sockaddr_in *)nam;
282 if (nam->sa_len != sizeof (*sin))
283 return (EINVAL);
284 #ifdef notdef
285 /*
286 * We should check the family, but old programs
287 * incorrectly fail to initialize it.
288 */
289 if (sin->sin_family != AF_INET)
290 return (EAFNOSUPPORT);
291 #endif
292 if (sin->sin_addr.s_addr != INADDR_ANY)
293 if (prison_ip(cred, 0, &sin->sin_addr.s_addr))
294 return(EINVAL);
295 if (sin->sin_port != *lportp) {
296 /* Don't allow the port to change. */
297 if (*lportp != 0)
298 return (EINVAL);
299 lport = sin->sin_port;
300 }
301 /* NB: lport is left as 0 if the port isn't being changed. */
302 if (IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) {
303 /*
304 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
305 * allow complete duplication of binding if
306 * SO_REUSEPORT is set, or if SO_REUSEADDR is set
307 * and a multicast address is bound on both
308 * new and duplicated sockets.
309 */
310 if (so->so_options & SO_REUSEADDR)
311 reuseport = SO_REUSEADDR|SO_REUSEPORT;
312 } else if (sin->sin_addr.s_addr != INADDR_ANY) {
313 sin->sin_port = 0; /* yech... */
314 bzero(&sin->sin_zero, sizeof(sin->sin_zero));
315 if (ifa_ifwithaddr((struct sockaddr *)sin) == 0)
316 return (EADDRNOTAVAIL);
317 }
318 laddr = sin->sin_addr;
319 if (lport) {
320 struct inpcb *t;
321 /* GROSS */
322 if (ntohs(lport) <= ipport_reservedhigh &&
323 ntohs(lport) >= ipport_reservedlow &&
324 suser_cred(cred, SUSER_ALLOWJAIL))
325 return (EACCES);
326 if (jailed(cred))
327 prison = 1;
328 if (so->so_cred->cr_uid != 0 &&
329 !IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) {
330 t = in_pcblookup_local(inp->inp_pcbinfo,
331 sin->sin_addr, lport,
332 prison ? 0 : INPLOOKUP_WILDCARD);
333 /*
334 * XXX
335 * This entire block sorely needs a rewrite.
336 */
337 if (t &&
338 ((t->inp_vflag & INP_TIMEWAIT) == 0) &&
339 (so->so_type != SOCK_STREAM ||
340 ntohl(t->inp_faddr.s_addr) == INADDR_ANY) &&
341 (ntohl(sin->sin_addr.s_addr) != INADDR_ANY ||
342 ntohl(t->inp_laddr.s_addr) != INADDR_ANY ||
343 (t->inp_socket->so_options &
344 SO_REUSEPORT) == 0) &&
345 (so->so_cred->cr_uid !=
346 t->inp_socket->so_cred->cr_uid))
347 return (EADDRINUSE);
348 }
349 if (prison && prison_ip(cred, 0, &sin->sin_addr.s_addr))
350 return (EADDRNOTAVAIL);
351 t = in_pcblookup_local(pcbinfo, sin->sin_addr,
352 lport, prison ? 0 : wild);
353 if (t && (t->inp_vflag & INP_TIMEWAIT)) {
354 if ((reuseport & intotw(t)->tw_so_options) == 0)
355 return (EADDRINUSE);
356 } else
357 if (t &&
358 (reuseport & t->inp_socket->so_options) == 0) {
359 #if defined(INET6)
360 if (ntohl(sin->sin_addr.s_addr) !=
361 INADDR_ANY ||
362 ntohl(t->inp_laddr.s_addr) !=
363 INADDR_ANY ||
364 INP_SOCKAF(so) ==
365 INP_SOCKAF(t->inp_socket))
366 #endif /* defined(INET6) */
367 return (EADDRINUSE);
368 }
369 }
370 }
371 if (*lportp != 0)
372 lport = *lportp;
373 if (lport == 0) {
374 u_short first, last;
375 int count;
376
377 if (laddr.s_addr != INADDR_ANY)
378 if (prison_ip(cred, 0, &laddr.s_addr))
379 return (EINVAL);
380
381 if (inp->inp_flags & INP_HIGHPORT) {
382 first = ipport_hifirstauto; /* sysctl */
383 last = ipport_hilastauto;
384 lastport = &pcbinfo->lasthi;
385 } else if (inp->inp_flags & INP_LOWPORT) {
386 if ((error = suser_cred(cred, SUSER_ALLOWJAIL)) != 0)
387 return error;
388 first = ipport_lowfirstauto; /* 1023 */
389 last = ipport_lowlastauto; /* 600 */
390 lastport = &pcbinfo->lastlow;
391 } else {
392 first = ipport_firstauto; /* sysctl */
393 last = ipport_lastauto;
394 lastport = &pcbinfo->lastport;
395 }
396 /*
397 * Simple check to ensure all ports are not used up causing
398 * a deadlock here.
399 *
400 * We split the two cases (up and down) so that the direction
401 * is not being tested on each round of the loop.
402 */
403 if (first > last) {
404 /*
405 * counting down
406 */
407 if (ipport_randomized)
408 *lastport = first -
409 (arc4random() % (first - last));
410 count = first - last;
411
412 do {
413 if (count-- < 0) /* completely used? */
414 return (EADDRNOTAVAIL);
415 --*lastport;
416 if (*lastport > first || *lastport < last)
417 *lastport = first;
418 lport = htons(*lastport);
419 } while (in_pcblookup_local(pcbinfo, laddr, lport,
420 wild));
421 } else {
422 /*
423 * counting up
424 */
425 if (ipport_randomized)
426 *lastport = first +
427 (arc4random() % (last - first));
428 count = last - first;
429
430 do {
431 if (count-- < 0) /* completely used? */
432 return (EADDRNOTAVAIL);
433 ++*lastport;
434 if (*lastport < first || *lastport > last)
435 *lastport = first;
436 lport = htons(*lastport);
437 } while (in_pcblookup_local(pcbinfo, laddr, lport,
438 wild));
439 }
440 }
441 if (prison_ip(cred, 0, &laddr.s_addr))
442 return (EINVAL);
443 *laddrp = laddr.s_addr;
444 *lportp = lport;
445 return (0);
446 }
447
448 /*
449 * Connect from a socket to a specified address.
450 * Both address and port must be specified in argument sin.
451 * If don't have a local address for this socket yet,
452 * then pick one.
453 */
454 int
455 in_pcbconnect(inp, nam, cred)
456 register struct inpcb *inp;
457 struct sockaddr *nam;
458 struct ucred *cred;
459 {
460 u_short lport, fport;
461 in_addr_t laddr, faddr;
462 int anonport, error;
463
464 INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
465 INP_LOCK_ASSERT(inp);
466
467 lport = inp->inp_lport;
468 laddr = inp->inp_laddr.s_addr;
469 anonport = (lport == 0);
470 error = in_pcbconnect_setup(inp, nam, &laddr, &lport, &faddr, &fport,
471 NULL, cred);
472 if (error)
473 return (error);
474
475 /* Do the initial binding of the local address if required. */
476 if (inp->inp_laddr.s_addr == INADDR_ANY && inp->inp_lport == 0) {
477 inp->inp_lport = lport;
478 inp->inp_laddr.s_addr = laddr;
479 if (in_pcbinshash(inp) != 0) {
480 inp->inp_laddr.s_addr = INADDR_ANY;
481 inp->inp_lport = 0;
482 return (EAGAIN);
483 }
484 }
485
486 /* Commit the remaining changes. */
487 inp->inp_lport = lport;
488 inp->inp_laddr.s_addr = laddr;
489 inp->inp_faddr.s_addr = faddr;
490 inp->inp_fport = fport;
491 in_pcbrehash(inp);
492 #ifdef IPSEC
493 if (inp->inp_socket->so_type == SOCK_STREAM)
494 ipsec_pcbconn(inp->inp_sp);
495 #endif
496 if (anonport)
497 inp->inp_flags |= INP_ANONPORT;
498 return (0);
499 }
500
501 /*
502 * Set up for a connect from a socket to the specified address.
503 * On entry, *laddrp and *lportp should contain the current local
504 * address and port for the PCB; these are updated to the values
505 * that should be placed in inp_laddr and inp_lport to complete
506 * the connect.
507 *
508 * On success, *faddrp and *fportp will be set to the remote address
509 * and port. These are not updated in the error case.
510 *
511 * If the operation fails because the connection already exists,
512 * *oinpp will be set to the PCB of that connection so that the
513 * caller can decide to override it. In all other cases, *oinpp
514 * is set to NULL.
515 */
516 int
517 in_pcbconnect_setup(inp, nam, laddrp, lportp, faddrp, fportp, oinpp, cred)
518 register struct inpcb *inp;
519 struct sockaddr *nam;
520 in_addr_t *laddrp;
521 u_short *lportp;
522 in_addr_t *faddrp;
523 u_short *fportp;
524 struct inpcb **oinpp;
525 struct ucred *cred;
526 {
527 struct sockaddr_in *sin = (struct sockaddr_in *)nam;
528 struct in_ifaddr *ia;
529 struct sockaddr_in sa;
530 struct ucred *socred;
531 struct inpcb *oinp;
532 struct in_addr laddr, faddr;
533 u_short lport, fport;
534 int error;
535
536 INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
537 INP_LOCK_ASSERT(inp);
538
539 if (oinpp != NULL)
540 *oinpp = NULL;
541 if (nam->sa_len != sizeof (*sin))
542 return (EINVAL);
543 if (sin->sin_family != AF_INET)
544 return (EAFNOSUPPORT);
545 if (sin->sin_port == 0)
546 return (EADDRNOTAVAIL);
547 laddr.s_addr = *laddrp;
548 lport = *lportp;
549 faddr = sin->sin_addr;
550 fport = sin->sin_port;
551 socred = inp->inp_socket->so_cred;
552 if (laddr.s_addr == INADDR_ANY && jailed(socred)) {
553 bzero(&sa, sizeof(sa));
554 sa.sin_addr.s_addr = htonl(prison_getip(socred));
555 sa.sin_len = sizeof(sa);
556 sa.sin_family = AF_INET;
557 error = in_pcbbind_setup(inp, (struct sockaddr *)&sa,
558 &laddr.s_addr, &lport, cred);
559 if (error)
560 return (error);
561 }
562 if (!TAILQ_EMPTY(&in_ifaddrhead)) {
563 /*
564 * If the destination address is INADDR_ANY,
565 * use the primary local address.
566 * If the supplied address is INADDR_BROADCAST,
567 * and the primary interface supports broadcast,
568 * choose the broadcast address for that interface.
569 */
570 if (faddr.s_addr == INADDR_ANY)
571 faddr = IA_SIN(TAILQ_FIRST(&in_ifaddrhead))->sin_addr;
572 else if (faddr.s_addr == (u_long)INADDR_BROADCAST &&
573 (TAILQ_FIRST(&in_ifaddrhead)->ia_ifp->if_flags &
574 IFF_BROADCAST))
575 faddr = satosin(&TAILQ_FIRST(
576 &in_ifaddrhead)->ia_broadaddr)->sin_addr;
577 }
578 if (laddr.s_addr == INADDR_ANY) {
579 struct route sro;
580
581 bzero(&sro, sizeof(sro));
582 ia = (struct in_ifaddr *)0;
583 /*
584 * If route is known our src addr is taken from the i/f,
585 * else punt.
586 */
587 if ((inp->inp_socket->so_options & SO_DONTROUTE) == 0) {
588 /* Find out route to destination */
589 sro.ro_dst.sa_family = AF_INET;
590 sro.ro_dst.sa_len = sizeof(struct sockaddr_in);
591 ((struct sockaddr_in *)&sro.ro_dst)->sin_addr = faddr;
592 rtalloc_ign(&sro, RTF_CLONING);
593 }
594 /*
595 * If we found a route, use the address
596 * corresponding to the outgoing interface
597 * unless it is the loopback (in case a route
598 * to our address on another net goes to loopback).
599 */
600 if (sro.ro_rt && !(sro.ro_rt->rt_ifp->if_flags & IFF_LOOPBACK))
601 ia = ifatoia(sro.ro_rt->rt_ifa);
602 if (sro.ro_rt)
603 RTFREE(sro.ro_rt);
604 if (ia == 0) {
605 bzero(&sa, sizeof(sa));
606 sa.sin_addr = faddr;
607 sa.sin_len = sizeof(sa);
608 sa.sin_family = AF_INET;
609
610 ia = ifatoia(ifa_ifwithdstaddr(sintosa(&sa)));
611 if (ia == 0)
612 ia = ifatoia(ifa_ifwithnet(sintosa(&sa)));
613 if (ia == 0)
614 return (ENETUNREACH);
615 }
616 /*
617 * If the destination address is multicast and an outgoing
618 * interface has been set as a multicast option, use the
619 * address of that interface as our source address.
620 */
621 if (IN_MULTICAST(ntohl(faddr.s_addr)) &&
622 inp->inp_moptions != NULL) {
623 struct ip_moptions *imo;
624 struct ifnet *ifp;
625
626 imo = inp->inp_moptions;
627 if (imo->imo_multicast_ifp != NULL) {
628 ifp = imo->imo_multicast_ifp;
629 TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link)
630 if (ia->ia_ifp == ifp)
631 break;
632 if (ia == 0)
633 return (EADDRNOTAVAIL);
634 }
635 }
636 laddr = ia->ia_addr.sin_addr;
637 }
638
639 oinp = in_pcblookup_hash(inp->inp_pcbinfo, faddr, fport, laddr, lport,
640 0, NULL);
641 if (oinp != NULL) {
642 if (oinpp != NULL)
643 *oinpp = oinp;
644 return (EADDRINUSE);
645 }
646 if (lport == 0) {
647 error = in_pcbbind_setup(inp, NULL, &laddr.s_addr, &lport,
648 cred);
649 if (error)
650 return (error);
651 }
652 *laddrp = laddr.s_addr;
653 *lportp = lport;
654 *faddrp = faddr.s_addr;
655 *fportp = fport;
656 return (0);
657 }
658
659 void
660 in_pcbdisconnect(inp)
661 struct inpcb *inp;
662 {
663 INP_LOCK_ASSERT(inp);
664
665 inp->inp_faddr.s_addr = INADDR_ANY;
666 inp->inp_fport = 0;
667 in_pcbrehash(inp);
668 #ifdef IPSEC
669 ipsec_pcbdisconn(inp->inp_sp);
670 #endif
671 if (inp->inp_socket->so_state & SS_NOFDREF)
672 in_pcbdetach(inp);
673 }
674
675 void
676 in_pcbdetach(inp)
677 struct inpcb *inp;
678 {
679 struct socket *so = inp->inp_socket;
680 struct inpcbinfo *ipi = inp->inp_pcbinfo;
681
682 INP_LOCK_ASSERT(inp);
683
684 #if defined(IPSEC) || defined(FAST_IPSEC)
685 ipsec4_delete_pcbpolicy(inp);
686 #endif /*IPSEC*/
687 inp->inp_gencnt = ++ipi->ipi_gencnt;
688 in_pcbremlists(inp);
689 if (so) {
690 ACCEPT_LOCK();
691 SOCK_LOCK(so);
692 so->so_pcb = 0;
693 sotryfree(so);
694 }
695 if (inp->inp_options)
696 (void)m_free(inp->inp_options);
697 ip_freemoptions(inp->inp_moptions);
698 inp->inp_vflag = 0;
699 INP_LOCK_DESTROY(inp);
700 #ifdef MAC
701 mac_destroy_inpcb(inp);
702 #endif
703 uma_zfree(ipi->ipi_zone, inp);
704 }
705
706 struct sockaddr *
707 in_sockaddr(port, addr_p)
708 in_port_t port;
709 struct in_addr *addr_p;
710 {
711 struct sockaddr_in *sin;
712
713 MALLOC(sin, struct sockaddr_in *, sizeof *sin, M_SONAME,
714 M_WAITOK | M_ZERO);
715 sin->sin_family = AF_INET;
716 sin->sin_len = sizeof(*sin);
717 sin->sin_addr = *addr_p;
718 sin->sin_port = port;
719
720 return (struct sockaddr *)sin;
721 }
722
723 /*
724 * The wrapper function will pass down the pcbinfo for this function to lock.
725 * The socket must have a valid
726 * (i.e., non-nil) PCB, but it should be impossible to get an invalid one
727 * except through a kernel programming error, so it is acceptable to panic
728 * (or in this case trap) if the PCB is invalid. (Actually, we don't trap
729 * because there actually /is/ a programming error somewhere... XXX)
730 */
731 int
732 in_setsockaddr(so, nam, pcbinfo)
733 struct socket *so;
734 struct sockaddr **nam;
735 struct inpcbinfo *pcbinfo;
736 {
737 int s;
738 register struct inpcb *inp;
739 struct in_addr addr;
740 in_port_t port;
741
742 s = splnet();
743 INP_INFO_RLOCK(pcbinfo);
744 inp = sotoinpcb(so);
745 if (!inp) {
746 INP_INFO_RUNLOCK(pcbinfo);
747 splx(s);
748 return ECONNRESET;
749 }
750 INP_LOCK(inp);
751 port = inp->inp_lport;
752 addr = inp->inp_laddr;
753 INP_UNLOCK(inp);
754 INP_INFO_RUNLOCK(pcbinfo);
755 splx(s);
756
757 *nam = in_sockaddr(port, &addr);
758 return 0;
759 }
760
761 /*
762 * The wrapper function will pass down the pcbinfo for this function to lock.
763 */
764 int
765 in_setpeeraddr(so, nam, pcbinfo)
766 struct socket *so;
767 struct sockaddr **nam;
768 struct inpcbinfo *pcbinfo;
769 {
770 int s;
771 register struct inpcb *inp;
772 struct in_addr addr;
773 in_port_t port;
774
775 s = splnet();
776 INP_INFO_RLOCK(pcbinfo);
777 inp = sotoinpcb(so);
778 if (!inp) {
779 INP_INFO_RUNLOCK(pcbinfo);
780 splx(s);
781 return ECONNRESET;
782 }
783 INP_LOCK(inp);
784 port = inp->inp_fport;
785 addr = inp->inp_faddr;
786 INP_UNLOCK(inp);
787 INP_INFO_RUNLOCK(pcbinfo);
788 splx(s);
789
790 *nam = in_sockaddr(port, &addr);
791 return 0;
792 }
793
794 void
795 in_pcbnotifyall(pcbinfo, faddr, errno, notify)
796 struct inpcbinfo *pcbinfo;
797 struct in_addr faddr;
798 int errno;
799 struct inpcb *(*notify)(struct inpcb *, int);
800 {
801 struct inpcb *inp, *ninp;
802 struct inpcbhead *head;
803 int s;
804
805 s = splnet();
806 INP_INFO_WLOCK(pcbinfo);
807 head = pcbinfo->listhead;
808 for (inp = LIST_FIRST(head); inp != NULL; inp = ninp) {
809 INP_LOCK(inp);
810 ninp = LIST_NEXT(inp, inp_list);
811 #ifdef INET6
812 if ((inp->inp_vflag & INP_IPV4) == 0) {
813 INP_UNLOCK(inp);
814 continue;
815 }
816 #endif
817 if (inp->inp_faddr.s_addr != faddr.s_addr ||
818 inp->inp_socket == NULL) {
819 INP_UNLOCK(inp);
820 continue;
821 }
822 if ((*notify)(inp, errno))
823 INP_UNLOCK(inp);
824 }
825 INP_INFO_WUNLOCK(pcbinfo);
826 splx(s);
827 }
828
829 void
830 in_pcbpurgeif0(pcbinfo, ifp)
831 struct inpcbinfo *pcbinfo;
832 struct ifnet *ifp;
833 {
834 struct inpcb *inp;
835 struct ip_moptions *imo;
836 int i, gap;
837
838 /* why no splnet here? XXX */
839 INP_INFO_RLOCK(pcbinfo);
840 LIST_FOREACH(inp, pcbinfo->listhead, inp_list) {
841 INP_LOCK(inp);
842 imo = inp->inp_moptions;
843 if ((inp->inp_vflag & INP_IPV4) &&
844 imo != NULL) {
845 /*
846 * Unselect the outgoing interface if it is being
847 * detached.
848 */
849 if (imo->imo_multicast_ifp == ifp)
850 imo->imo_multicast_ifp = NULL;
851
852 /*
853 * Drop multicast group membership if we joined
854 * through the interface being detached.
855 */
856 for (i = 0, gap = 0; i < imo->imo_num_memberships;
857 i++) {
858 if (imo->imo_membership[i]->inm_ifp == ifp) {
859 in_delmulti(imo->imo_membership[i]);
860 gap++;
861 } else if (gap != 0)
862 imo->imo_membership[i - gap] =
863 imo->imo_membership[i];
864 }
865 imo->imo_num_memberships -= gap;
866 }
867 INP_UNLOCK(inp);
868 }
869 INP_INFO_RUNLOCK(pcbinfo);
870 }
871
872 /*
873 * Lookup a PCB based on the local address and port.
874 */
875 struct inpcb *
876 in_pcblookup_local(pcbinfo, laddr, lport_arg, wild_okay)
877 struct inpcbinfo *pcbinfo;
878 struct in_addr laddr;
879 u_int lport_arg;
880 int wild_okay;
881 {
882 register struct inpcb *inp;
883 int matchwild = 3, wildcard;
884 u_short lport = lport_arg;
885
886 INP_INFO_WLOCK_ASSERT(pcbinfo);
887
888 if (!wild_okay) {
889 struct inpcbhead *head;
890 /*
891 * Look for an unconnected (wildcard foreign addr) PCB that
892 * matches the local address and port we're looking for.
893 */
894 head = &pcbinfo->hashbase[INP_PCBHASH(INADDR_ANY, lport, 0, pcbinfo->hashmask)];
895 LIST_FOREACH(inp, head, inp_hash) {
896 #ifdef INET6
897 if ((inp->inp_vflag & INP_IPV4) == 0)
898 continue;
899 #endif
900 if (inp->inp_faddr.s_addr == INADDR_ANY &&
901 inp->inp_laddr.s_addr == laddr.s_addr &&
902 inp->inp_lport == lport) {
903 /*
904 * Found.
905 */
906 return (inp);
907 }
908 }
909 /*
910 * Not found.
911 */
912 return (NULL);
913 } else {
914 struct inpcbporthead *porthash;
915 struct inpcbport *phd;
916 struct inpcb *match = NULL;
917 /*
918 * Best fit PCB lookup.
919 *
920 * First see if this local port is in use by looking on the
921 * port hash list.
922 */
923 retrylookup:
924 porthash = &pcbinfo->porthashbase[INP_PCBPORTHASH(lport,
925 pcbinfo->porthashmask)];
926 LIST_FOREACH(phd, porthash, phd_hash) {
927 if (phd->phd_port == lport)
928 break;
929 }
930 if (phd != NULL) {
931 /*
932 * Port is in use by one or more PCBs. Look for best
933 * fit.
934 */
935 LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) {
936 wildcard = 0;
937 #ifdef INET6
938 if ((inp->inp_vflag & INP_IPV4) == 0)
939 continue;
940 #endif
941 /*
942 * Clean out old time_wait sockets if they
943 * are clogging up needed local ports.
944 */
945 if ((inp->inp_vflag & INP_TIMEWAIT) != 0) {
946 if (tcp_twrecycleable((struct tcptw *)inp->inp_ppcb)) {
947 INP_LOCK(inp);
948 tcp_twclose((struct tcptw *)inp->inp_ppcb, 0);
949 match = NULL;
950 goto retrylookup;
951 }
952 }
953 if (inp->inp_faddr.s_addr != INADDR_ANY)
954 wildcard++;
955 if (inp->inp_laddr.s_addr != INADDR_ANY) {
956 if (laddr.s_addr == INADDR_ANY)
957 wildcard++;
958 else if (inp->inp_laddr.s_addr != laddr.s_addr)
959 continue;
960 } else {
961 if (laddr.s_addr != INADDR_ANY)
962 wildcard++;
963 }
964 if (wildcard < matchwild) {
965 match = inp;
966 matchwild = wildcard;
967 if (matchwild == 0) {
968 break;
969 }
970 }
971 }
972 }
973 return (match);
974 }
975 }
976
977 /*
978 * Lookup PCB in hash list.
979 */
980 struct inpcb *
981 in_pcblookup_hash(pcbinfo, faddr, fport_arg, laddr, lport_arg, wildcard,
982 ifp)
983 struct inpcbinfo *pcbinfo;
984 struct in_addr faddr, laddr;
985 u_int fport_arg, lport_arg;
986 int wildcard;
987 struct ifnet *ifp;
988 {
989 struct inpcbhead *head;
990 register struct inpcb *inp;
991 u_short fport = fport_arg, lport = lport_arg;
992
993 INP_INFO_RLOCK_ASSERT(pcbinfo);
994 /*
995 * First look for an exact match.
996 */
997 head = &pcbinfo->hashbase[INP_PCBHASH(faddr.s_addr, lport, fport, pcbinfo->hashmask)];
998 LIST_FOREACH(inp, head, inp_hash) {
999 #ifdef INET6
1000 if ((inp->inp_vflag & INP_IPV4) == 0)
1001 continue;
1002 #endif
1003 if (inp->inp_faddr.s_addr == faddr.s_addr &&
1004 inp->inp_laddr.s_addr == laddr.s_addr &&
1005 inp->inp_fport == fport &&
1006 inp->inp_lport == lport) {
1007 /*
1008 * Found.
1009 */
1010 return (inp);
1011 }
1012 }
1013 if (wildcard) {
1014 struct inpcb *local_wild = NULL;
1015 #if defined(INET6)
1016 struct inpcb *local_wild_mapped = NULL;
1017 #endif /* defined(INET6) */
1018
1019 head = &pcbinfo->hashbase[INP_PCBHASH(INADDR_ANY, lport, 0, pcbinfo->hashmask)];
1020 LIST_FOREACH(inp, head, inp_hash) {
1021 #ifdef INET6
1022 if ((inp->inp_vflag & INP_IPV4) == 0)
1023 continue;
1024 #endif
1025 if (inp->inp_faddr.s_addr == INADDR_ANY &&
1026 inp->inp_lport == lport) {
1027 if (ifp && ifp->if_type == IFT_FAITH &&
1028 (inp->inp_flags & INP_FAITH) == 0)
1029 continue;
1030 if (inp->inp_laddr.s_addr == laddr.s_addr)
1031 return (inp);
1032 else if (inp->inp_laddr.s_addr == INADDR_ANY) {
1033 #if defined(INET6)
1034 if (INP_CHECK_SOCKAF(inp->inp_socket,
1035 AF_INET6))
1036 local_wild_mapped = inp;
1037 else
1038 #endif /* defined(INET6) */
1039 local_wild = inp;
1040 }
1041 }
1042 }
1043 #if defined(INET6)
1044 if (local_wild == NULL)
1045 return (local_wild_mapped);
1046 #endif /* defined(INET6) */
1047 return (local_wild);
1048 }
1049
1050 /*
1051 * Not found.
1052 */
1053 return (NULL);
1054 }
1055
1056 /*
1057 * Insert PCB onto various hash lists.
1058 */
1059 int
1060 in_pcbinshash(inp)
1061 struct inpcb *inp;
1062 {
1063 struct inpcbhead *pcbhash;
1064 struct inpcbporthead *pcbporthash;
1065 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1066 struct inpcbport *phd;
1067 u_int32_t hashkey_faddr;
1068
1069 INP_INFO_WLOCK_ASSERT(pcbinfo);
1070 #ifdef INET6
1071 if (inp->inp_vflag & INP_IPV6)
1072 hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX */;
1073 else
1074 #endif /* INET6 */
1075 hashkey_faddr = inp->inp_faddr.s_addr;
1076
1077 pcbhash = &pcbinfo->hashbase[INP_PCBHASH(hashkey_faddr,
1078 inp->inp_lport, inp->inp_fport, pcbinfo->hashmask)];
1079
1080 pcbporthash = &pcbinfo->porthashbase[INP_PCBPORTHASH(inp->inp_lport,
1081 pcbinfo->porthashmask)];
1082
1083 /*
1084 * Go through port list and look for a head for this lport.
1085 */
1086 LIST_FOREACH(phd, pcbporthash, phd_hash) {
1087 if (phd->phd_port == inp->inp_lport)
1088 break;
1089 }
1090 /*
1091 * If none exists, malloc one and tack it on.
1092 */
1093 if (phd == NULL) {
1094 MALLOC(phd, struct inpcbport *, sizeof(struct inpcbport), M_PCB, M_NOWAIT);
1095 if (phd == NULL) {
1096 return (ENOBUFS); /* XXX */
1097 }
1098 phd->phd_port = inp->inp_lport;
1099 LIST_INIT(&phd->phd_pcblist);
1100 LIST_INSERT_HEAD(pcbporthash, phd, phd_hash);
1101 }
1102 inp->inp_phd = phd;
1103 LIST_INSERT_HEAD(&phd->phd_pcblist, inp, inp_portlist);
1104 LIST_INSERT_HEAD(pcbhash, inp, inp_hash);
1105 return (0);
1106 }
1107
1108 /*
1109 * Move PCB to the proper hash bucket when { faddr, fport } have been
1110 * changed. NOTE: This does not handle the case of the lport changing (the
1111 * hashed port list would have to be updated as well), so the lport must
1112 * not change after in_pcbinshash() has been called.
1113 */
1114 void
1115 in_pcbrehash(inp)
1116 struct inpcb *inp;
1117 {
1118 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1119 struct inpcbhead *head;
1120 u_int32_t hashkey_faddr;
1121
1122 INP_INFO_WLOCK_ASSERT(pcbinfo);
1123 INP_LOCK_ASSERT(inp);
1124 #ifdef INET6
1125 if (inp->inp_vflag & INP_IPV6)
1126 hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX */;
1127 else
1128 #endif /* INET6 */
1129 hashkey_faddr = inp->inp_faddr.s_addr;
1130
1131 head = &pcbinfo->hashbase[INP_PCBHASH(hashkey_faddr,
1132 inp->inp_lport, inp->inp_fport, pcbinfo->hashmask)];
1133
1134 LIST_REMOVE(inp, inp_hash);
1135 LIST_INSERT_HEAD(head, inp, inp_hash);
1136 }
1137
1138 /*
1139 * Remove PCB from various lists.
1140 */
1141 void
1142 in_pcbremlists(inp)
1143 struct inpcb *inp;
1144 {
1145 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1146
1147 INP_INFO_WLOCK_ASSERT(pcbinfo);
1148 INP_LOCK_ASSERT(inp);
1149
1150 inp->inp_gencnt = ++pcbinfo->ipi_gencnt;
1151 if (inp->inp_lport) {
1152 struct inpcbport *phd = inp->inp_phd;
1153
1154 LIST_REMOVE(inp, inp_hash);
1155 LIST_REMOVE(inp, inp_portlist);
1156 if (LIST_FIRST(&phd->phd_pcblist) == NULL) {
1157 LIST_REMOVE(phd, phd_hash);
1158 free(phd, M_PCB);
1159 }
1160 }
1161 LIST_REMOVE(inp, inp_list);
1162 pcbinfo->ipi_count--;
1163 }
1164
1165 /*
1166 * A set label operation has occurred at the socket layer, propagate the
1167 * label change into the in_pcb for the socket.
1168 */
1169 void
1170 in_pcbsosetlabel(so)
1171 struct socket *so;
1172 {
1173 #ifdef MAC
1174 struct inpcb *inp;
1175
1176 inp = (struct inpcb *)so->so_pcb;
1177 INP_LOCK(inp);
1178 SOCK_LOCK(so);
1179 mac_inpcb_sosetlabel(so, inp);
1180 SOCK_UNLOCK(so);
1181 INP_UNLOCK(inp);
1182 #endif
1183 }
Cache object: 78d7b4442b1bc83dd57529a97243aac5
|