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.
4 * Copyright (c) 2007-2009 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: releng/7.3/sys/netinet/in_pcb.c 202924 2010-01-24 14:05:56Z bz $");
36
37 #include "opt_ddb.h"
38 #include "opt_ipsec.h"
39 #include "opt_inet6.h"
40 #include "opt_mac.h"
41
42 #include <sys/param.h>
43 #include <sys/systm.h>
44 #include <sys/malloc.h>
45 #include <sys/mbuf.h>
46 #include <sys/domain.h>
47 #include <sys/protosw.h>
48 #include <sys/socket.h>
49 #include <sys/socketvar.h>
50 #include <sys/priv.h>
51 #include <sys/proc.h>
52 #include <sys/jail.h>
53 #include <sys/kernel.h>
54 #include <sys/sysctl.h>
55
56 #ifdef DDB
57 #include <ddb/ddb.h>
58 #endif
59
60 #include <vm/uma.h>
61
62 #include <net/if.h>
63 #include <net/if_types.h>
64 #include <net/route.h>
65
66 #include <netinet/in.h>
67 #include <netinet/in_pcb.h>
68 #include <netinet/in_var.h>
69 #include <netinet/ip_var.h>
70 #include <netinet/tcp_var.h>
71 #include <netinet/udp.h>
72 #include <netinet/udp_var.h>
73 #ifdef INET6
74 #include <netinet/ip6.h>
75 #include <netinet6/ip6_var.h>
76 #endif /* INET6 */
77
78
79 #ifdef IPSEC
80 #include <netipsec/ipsec.h>
81 #include <netipsec/key.h>
82 #endif /* IPSEC */
83
84 #include <security/mac/mac_framework.h>
85
86 /*
87 * These configure the range of local port addresses assigned to
88 * "unspecified" outgoing connections/packets/whatever.
89 */
90 int ipport_lowfirstauto = IPPORT_RESERVED - 1; /* 1023 */
91 int ipport_lowlastauto = IPPORT_RESERVEDSTART; /* 600 */
92 int ipport_firstauto = IPPORT_HIFIRSTAUTO; /* 49152 */
93 int ipport_lastauto = IPPORT_HILASTAUTO; /* 65535 */
94 int ipport_hifirstauto = IPPORT_HIFIRSTAUTO; /* 49152 */
95 int ipport_hilastauto = IPPORT_HILASTAUTO; /* 65535 */
96
97 /*
98 * Reserved ports accessible only to root. There are significant
99 * security considerations that must be accounted for when changing these,
100 * but the security benefits can be great. Please be careful.
101 */
102 int ipport_reservedhigh = IPPORT_RESERVED - 1; /* 1023 */
103 int ipport_reservedlow = 0;
104
105 /* Variables dealing with random ephemeral port allocation. */
106 int ipport_randomized = 1; /* user controlled via sysctl */
107 int ipport_randomcps = 10; /* user controlled via sysctl */
108 int ipport_randomtime = 45; /* user controlled via sysctl */
109 int ipport_stoprandom = 0; /* toggled by ipport_tick */
110 int ipport_tcpallocs;
111 int ipport_tcplastcount;
112
113 #define RANGECHK(var, min, max) \
114 if ((var) < (min)) { (var) = (min); } \
115 else if ((var) > (max)) { (var) = (max); }
116
117 static int
118 sysctl_net_ipport_check(SYSCTL_HANDLER_ARGS)
119 {
120 int error;
121
122 error = sysctl_handle_int(oidp, oidp->oid_arg1, oidp->oid_arg2, req);
123 if (error == 0) {
124 RANGECHK(ipport_lowfirstauto, 1, IPPORT_RESERVED - 1);
125 RANGECHK(ipport_lowlastauto, 1, IPPORT_RESERVED - 1);
126 RANGECHK(ipport_firstauto, IPPORT_RESERVED, IPPORT_MAX);
127 RANGECHK(ipport_lastauto, IPPORT_RESERVED, IPPORT_MAX);
128 RANGECHK(ipport_hifirstauto, IPPORT_RESERVED, IPPORT_MAX);
129 RANGECHK(ipport_hilastauto, IPPORT_RESERVED, IPPORT_MAX);
130 }
131 return (error);
132 }
133
134 #undef RANGECHK
135
136 SYSCTL_NODE(_net_inet_ip, IPPROTO_IP, portrange, CTLFLAG_RW, 0, "IP Ports");
137
138 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, lowfirst, CTLTYPE_INT|CTLFLAG_RW,
139 &ipport_lowfirstauto, 0, &sysctl_net_ipport_check, "I", "");
140 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, lowlast, CTLTYPE_INT|CTLFLAG_RW,
141 &ipport_lowlastauto, 0, &sysctl_net_ipport_check, "I", "");
142 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, first, CTLTYPE_INT|CTLFLAG_RW,
143 &ipport_firstauto, 0, &sysctl_net_ipport_check, "I", "");
144 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, last, CTLTYPE_INT|CTLFLAG_RW,
145 &ipport_lastauto, 0, &sysctl_net_ipport_check, "I", "");
146 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, hifirst, CTLTYPE_INT|CTLFLAG_RW,
147 &ipport_hifirstauto, 0, &sysctl_net_ipport_check, "I", "");
148 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, hilast, CTLTYPE_INT|CTLFLAG_RW,
149 &ipport_hilastauto, 0, &sysctl_net_ipport_check, "I", "");
150 SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, reservedhigh,
151 CTLFLAG_RW|CTLFLAG_SECURE, &ipport_reservedhigh, 0, "");
152 SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, reservedlow,
153 CTLFLAG_RW|CTLFLAG_SECURE, &ipport_reservedlow, 0, "");
154 SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, randomized, CTLFLAG_RW,
155 &ipport_randomized, 0, "Enable random port allocation");
156 SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, randomcps, CTLFLAG_RW,
157 &ipport_randomcps, 0, "Maximum number of random port "
158 "allocations before switching to a sequental one");
159 SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, randomtime, CTLFLAG_RW,
160 &ipport_randomtime, 0, "Minimum time to keep sequental port "
161 "allocation before switching to a random one");
162
163 /*
164 * in_pcb.c: manage the Protocol Control Blocks.
165 *
166 * NOTE: It is assumed that most of these functions will be called with
167 * the pcbinfo lock held, and often, the inpcb lock held, as these utility
168 * functions often modify hash chains or addresses in pcbs.
169 */
170
171 /*
172 * Allocate a PCB and associate it with the socket.
173 * On success return with the PCB locked.
174 */
175 int
176 in_pcballoc(struct socket *so, struct inpcbinfo *pcbinfo)
177 {
178 struct inpcb *inp;
179 int error;
180
181 INP_INFO_WLOCK_ASSERT(pcbinfo);
182 error = 0;
183 inp = uma_zalloc(pcbinfo->ipi_zone, M_NOWAIT);
184 if (inp == NULL)
185 return (ENOBUFS);
186 bzero(inp, inp_zero_size);
187 inp->inp_pcbinfo = pcbinfo;
188 inp->inp_socket = so;
189 inp->inp_cred = crhold(so->so_cred);
190 inp->inp_inc.inc_fibnum = so->so_fibnum;
191 #ifdef MAC
192 error = mac_init_inpcb(inp, M_NOWAIT);
193 if (error != 0)
194 goto out;
195 SOCK_LOCK(so);
196 mac_create_inpcb_from_socket(so, inp);
197 SOCK_UNLOCK(so);
198 #endif
199 #ifdef IPSEC
200 error = ipsec_init_policy(so, &inp->inp_sp);
201 if (error != 0) {
202 #ifdef MAC
203 mac_destroy_inpcb(inp);
204 #endif
205 goto out;
206 }
207 #endif /*IPSEC*/
208 #ifdef INET6
209 if (INP_SOCKAF(so) == AF_INET6) {
210 inp->inp_vflag |= INP_IPV6PROTO;
211 if (ip6_v6only)
212 inp->inp_flags |= IN6P_IPV6_V6ONLY;
213 }
214 #endif
215 LIST_INSERT_HEAD(pcbinfo->ipi_listhead, inp, inp_list);
216 pcbinfo->ipi_count++;
217 so->so_pcb = (caddr_t)inp;
218 #ifdef INET6
219 if (ip6_auto_flowlabel)
220 inp->inp_flags |= IN6P_AUTOFLOWLABEL;
221 #endif
222 INP_WLOCK(inp);
223 inp->inp_gencnt = ++pcbinfo->ipi_gencnt;
224 #if defined(IPSEC) || defined(MAC)
225 out:
226 if (error != 0) {
227 crfree(inp->inp_cred);
228 uma_zfree(pcbinfo->ipi_zone, inp);
229 }
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;
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(&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 == NULL) {
298 if ((error = prison_local_ip4(cred, &laddr)) != 0)
299 return (error);
300 } else {
301 sin = (struct sockaddr_in *)nam;
302 if (nam->sa_len != sizeof (*sin))
303 return (EINVAL);
304 #ifdef notdef
305 /*
306 * We should check the family, but old programs
307 * incorrectly fail to initialize it.
308 */
309 if (sin->sin_family != AF_INET)
310 return (EAFNOSUPPORT);
311 #endif
312 error = prison_local_ip4(cred, &sin->sin_addr);
313 if (error)
314 return (error);
315 if (sin->sin_port != *lportp) {
316 /* Don't allow the port to change. */
317 if (*lportp != 0)
318 return (EINVAL);
319 lport = sin->sin_port;
320 }
321 /* NB: lport is left as 0 if the port isn't being changed. */
322 if (IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) {
323 /*
324 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
325 * allow complete duplication of binding if
326 * SO_REUSEPORT is set, or if SO_REUSEADDR is set
327 * and a multicast address is bound on both
328 * new and duplicated sockets.
329 */
330 if (so->so_options & SO_REUSEADDR)
331 reuseport = SO_REUSEADDR|SO_REUSEPORT;
332 } else if (sin->sin_addr.s_addr != INADDR_ANY) {
333 sin->sin_port = 0; /* yech... */
334 bzero(&sin->sin_zero, sizeof(sin->sin_zero));
335 if (ifa_ifwithaddr((struct sockaddr *)sin) == 0)
336 return (EADDRNOTAVAIL);
337 }
338 laddr = sin->sin_addr;
339 if (lport) {
340 struct inpcb *t;
341 struct tcptw *tw;
342
343 /* GROSS */
344 if (ntohs(lport) <= ipport_reservedhigh &&
345 ntohs(lport) >= ipport_reservedlow &&
346 priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT,
347 0))
348 return (EACCES);
349 if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)) &&
350 priv_check_cred(inp->inp_cred,
351 PRIV_NETINET_REUSEPORT, 0) != 0) {
352 t = in_pcblookup_local(pcbinfo, sin->sin_addr,
353 lport, INPLOOKUP_WILDCARD, cred);
354 /*
355 * XXX
356 * This entire block sorely needs a rewrite.
357 */
358 if (t &&
359 ((t->inp_flags & 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 (inp->inp_cred->cr_uid !=
367 t->inp_cred->cr_uid))
368 return (EADDRINUSE);
369 }
370 t = in_pcblookup_local(pcbinfo, sin->sin_addr,
371 lport, wild, cred);
372 if (t && (t->inp_flags & INP_TIMEWAIT)) {
373 /*
374 * XXXRW: If an incpb has had its timewait
375 * state recycled, we treat the address as
376 * being in use (for now). This is better
377 * than a panic, but not desirable.
378 */
379 tw = intotw(inp);
380 if (tw == NULL ||
381 (reuseport & tw->tw_so_options) == 0)
382 return (EADDRINUSE);
383 } else if (t &&
384 (reuseport & t->inp_socket->so_options) == 0) {
385 #ifdef INET6
386 if (ntohl(sin->sin_addr.s_addr) !=
387 INADDR_ANY ||
388 ntohl(t->inp_laddr.s_addr) !=
389 INADDR_ANY ||
390 INP_SOCKAF(so) ==
391 INP_SOCKAF(t->inp_socket))
392 #endif
393 return (EADDRINUSE);
394 }
395 }
396 }
397 if (*lportp != 0)
398 lport = *lportp;
399 if (lport == 0) {
400 u_short first, last;
401 int count;
402
403 if (inp->inp_flags & INP_HIGHPORT) {
404 first = ipport_hifirstauto; /* sysctl */
405 last = ipport_hilastauto;
406 lastport = &pcbinfo->ipi_lasthi;
407 } else if (inp->inp_flags & INP_LOWPORT) {
408 error = priv_check_cred(cred,
409 PRIV_NETINET_RESERVEDPORT, 0);
410 if (error)
411 return error;
412 first = ipport_lowfirstauto; /* 1023 */
413 last = ipport_lowlastauto; /* 600 */
414 lastport = &pcbinfo->ipi_lastlow;
415 } else {
416 first = ipport_firstauto; /* sysctl */
417 last = ipport_lastauto;
418 lastport = &pcbinfo->ipi_lastport;
419 }
420 /*
421 * For UDP, use random port allocation as long as the user
422 * allows it. For TCP (and as of yet unknown) connections,
423 * use random port allocation only if the user allows it AND
424 * ipport_tick() allows it.
425 */
426 if (ipport_randomized &&
427 (!ipport_stoprandom || pcbinfo == &udbinfo))
428 dorandom = 1;
429 else
430 dorandom = 0;
431 /*
432 * It makes no sense to do random port allocation if
433 * we have the only port available.
434 */
435 if (first == last)
436 dorandom = 0;
437 /* Make sure to not include UDP packets in the count. */
438 if (pcbinfo != &udbinfo)
439 ipport_tcpallocs++;
440 /*
441 * Instead of having two loops further down counting up or down
442 * make sure that first is always <= last and go with only one
443 * code path implementing all logic.
444 *
445 * We split the two cases (up and down) so that the direction
446 * is not being tested on each round of the loop.
447 */
448 if (first > last) {
449 /*
450 * counting down
451 */
452 if (dorandom)
453 *lastport = first -
454 (arc4random() % (first - last));
455 count = first - last;
456
457 do {
458 if (count-- < 0) /* completely used? */
459 return (EADDRNOTAVAIL);
460 --*lastport;
461 if (*lastport > first || *lastport < last)
462 *lastport = first;
463 lport = htons(*lastport);
464 } while (in_pcblookup_local(pcbinfo, laddr, lport,
465 wild, cred));
466 } else {
467 /*
468 * counting up
469 */
470 if (dorandom)
471 *lastport = first +
472 (arc4random() % (last - first));
473 count = last - first;
474
475 do {
476 if (count-- < 0) /* completely used? */
477 return (EADDRNOTAVAIL);
478 ++*lastport;
479 if (*lastport < first || *lastport > last)
480 *lastport = first;
481 lport = htons(*lastport);
482 } while (in_pcblookup_local(pcbinfo, laddr, lport,
483 wild, cred));
484 }
485 }
486 *laddrp = laddr.s_addr;
487 *lportp = lport;
488 return (0);
489 }
490
491 /*
492 * Connect from a socket to a specified address.
493 * Both address and port must be specified in argument sin.
494 * If don't have a local address for this socket yet,
495 * then pick one.
496 */
497 int
498 in_pcbconnect(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred)
499 {
500 u_short lport, fport;
501 in_addr_t laddr, faddr;
502 int anonport, error;
503
504 INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
505 INP_WLOCK_ASSERT(inp);
506
507 lport = inp->inp_lport;
508 laddr = inp->inp_laddr.s_addr;
509 anonport = (lport == 0);
510 error = in_pcbconnect_setup(inp, nam, &laddr, &lport, &faddr, &fport,
511 NULL, cred);
512 if (error)
513 return (error);
514
515 /* Do the initial binding of the local address if required. */
516 if (inp->inp_laddr.s_addr == INADDR_ANY && inp->inp_lport == 0) {
517 inp->inp_lport = lport;
518 inp->inp_laddr.s_addr = laddr;
519 if (in_pcbinshash(inp) != 0) {
520 inp->inp_laddr.s_addr = INADDR_ANY;
521 inp->inp_lport = 0;
522 return (EAGAIN);
523 }
524 }
525
526 /* Commit the remaining changes. */
527 inp->inp_lport = lport;
528 inp->inp_laddr.s_addr = laddr;
529 inp->inp_faddr.s_addr = faddr;
530 inp->inp_fport = fport;
531 in_pcbrehash(inp);
532
533 if (anonport)
534 inp->inp_flags |= INP_ANONPORT;
535 return (0);
536 }
537
538 /*
539 * Do proper source address selection on an unbound socket in case
540 * of connect. Take jails into account as well.
541 */
542 static int
543 in_pcbladdr(struct inpcb *inp, struct in_addr *faddr, struct in_addr *laddr,
544 struct ucred *cred)
545 {
546 struct in_ifaddr *ia;
547 struct ifaddr *ifa;
548 struct sockaddr *sa;
549 struct sockaddr_in *sin;
550 struct route sro;
551 int error;
552
553 KASSERT(laddr != NULL, ("%s: laddr NULL", __func__));
554
555 /*
556 * Bypass source address selection and use the primary jail IP
557 * if requested.
558 */
559 if (cred != NULL && !prison_saddrsel_ip4(cred, laddr))
560 return (0);
561
562 error = 0;
563 ia = NULL;
564 bzero(&sro, sizeof(sro));
565
566 sin = (struct sockaddr_in *)&sro.ro_dst;
567 sin->sin_family = AF_INET;
568 sin->sin_len = sizeof(struct sockaddr_in);
569 sin->sin_addr.s_addr = faddr->s_addr;
570
571 /*
572 * If route is known our src addr is taken from the i/f,
573 * else punt.
574 *
575 * Find out route to destination.
576 */
577 if ((inp->inp_socket->so_options & SO_DONTROUTE) == 0)
578 in_rtalloc_ign(&sro, RTF_CLONING, inp->inp_inc.inc_fibnum);
579
580 /*
581 * If we found a route, use the address corresponding to
582 * the outgoing interface.
583 *
584 * Otherwise assume faddr is reachable on a directly connected
585 * network and try to find a corresponding interface to take
586 * the source address from.
587 */
588 if (sro.ro_rt == NULL || sro.ro_rt->rt_ifp == NULL) {
589 struct ifnet *ifp;
590
591 ia = ifatoia(ifa_ifwithdstaddr((struct sockaddr *)sin));
592 if (ia == NULL)
593 ia = ifatoia(ifa_ifwithnet((struct sockaddr *)sin));
594 if (ia == NULL) {
595 error = ENETUNREACH;
596 goto done;
597 }
598
599 if (cred == NULL || !jailed(cred)) {
600 laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
601 goto done;
602 }
603
604 ifp = ia->ia_ifp;
605 ia = NULL;
606 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
607
608 sa = ifa->ifa_addr;
609 if (sa->sa_family != AF_INET)
610 continue;
611 sin = (struct sockaddr_in *)sa;
612 if (prison_check_ip4(cred, &sin->sin_addr) == 0) {
613 ia = (struct in_ifaddr *)ifa;
614 break;
615 }
616 }
617 if (ia != NULL) {
618 laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
619 goto done;
620 }
621
622 /* 3. As a last resort return the 'default' jail address. */
623 error = prison_get_ip4(cred, laddr);
624 goto done;
625 }
626
627 /*
628 * If the outgoing interface on the route found is not
629 * a loopback interface, use the address from that interface.
630 * In case of jails do those three steps:
631 * 1. check if the interface address belongs to the jail. If so use it.
632 * 2. check if we have any address on the outgoing interface
633 * belonging to this jail. If so use it.
634 * 3. as a last resort return the 'default' jail address.
635 */
636 if ((sro.ro_rt->rt_ifp->if_flags & IFF_LOOPBACK) == 0) {
637
638 /* If not jailed, use the default returned. */
639 if (cred == NULL || !jailed(cred)) {
640 ia = (struct in_ifaddr *)sro.ro_rt->rt_ifa;
641 laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
642 goto done;
643 }
644
645 /* Jailed. */
646 /* 1. Check if the iface address belongs to the jail. */
647 sin = (struct sockaddr_in *)sro.ro_rt->rt_ifa->ifa_addr;
648 if (prison_check_ip4(cred, &sin->sin_addr) == 0) {
649 ia = (struct in_ifaddr *)sro.ro_rt->rt_ifa;
650 laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
651 goto done;
652 }
653
654 /*
655 * 2. Check if we have any address on the outgoing interface
656 * belonging to this jail.
657 */
658 TAILQ_FOREACH(ifa, &sro.ro_rt->rt_ifp->if_addrhead, ifa_link) {
659
660 sa = ifa->ifa_addr;
661 if (sa->sa_family != AF_INET)
662 continue;
663 sin = (struct sockaddr_in *)sa;
664 if (prison_check_ip4(cred, &sin->sin_addr) == 0) {
665 ia = (struct in_ifaddr *)ifa;
666 break;
667 }
668 }
669 if (ia != NULL) {
670 laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
671 goto done;
672 }
673
674 /* 3. As a last resort return the 'default' jail address. */
675 error = prison_get_ip4(cred, laddr);
676 goto done;
677 }
678
679 /*
680 * The outgoing interface is marked with 'loopback net', so a route
681 * to ourselves is here.
682 * Try to find the interface of the destination address and then
683 * take the address from there. That interface is not necessarily
684 * a loopback interface.
685 * In case of jails, check that it is an address of the jail
686 * and if we cannot find, fall back to the 'default' jail address.
687 */
688 if ((sro.ro_rt->rt_ifp->if_flags & IFF_LOOPBACK) != 0) {
689 struct sockaddr_in sain;
690
691 bzero(&sain, sizeof(struct sockaddr_in));
692 sain.sin_family = AF_INET;
693 sain.sin_len = sizeof(struct sockaddr_in);
694 sain.sin_addr.s_addr = faddr->s_addr;
695
696 ia = ifatoia(ifa_ifwithdstaddr(sintosa(&sain)));
697 if (ia == NULL)
698 ia = ifatoia(ifa_ifwithnet(sintosa(&sain)));
699
700 if (cred == NULL || !jailed(cred)) {
701 #if __FreeBSD_version < 800000
702 if (ia == NULL)
703 ia = (struct in_ifaddr *)sro.ro_rt->rt_ifa;
704 #endif
705 if (ia == NULL) {
706 error = ENETUNREACH;
707 goto done;
708 }
709 laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
710 goto done;
711 }
712
713 /* Jailed. */
714 if (ia != NULL) {
715 struct ifnet *ifp;
716
717 ifp = ia->ia_ifp;
718 ia = NULL;
719 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
720
721 sa = ifa->ifa_addr;
722 if (sa->sa_family != AF_INET)
723 continue;
724 sin = (struct sockaddr_in *)sa;
725 if (prison_check_ip4(cred,
726 &sin->sin_addr) == 0) {
727 ia = (struct in_ifaddr *)ifa;
728 break;
729 }
730 }
731 if (ia != NULL) {
732 laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
733 goto done;
734 }
735 }
736
737 /* 3. As a last resort return the 'default' jail address. */
738 error = prison_get_ip4(cred, laddr);
739 goto done;
740 }
741
742 done:
743 if (sro.ro_rt != NULL)
744 RTFREE(sro.ro_rt);
745 return (error);
746 }
747
748 /*
749 * Set up for a connect from a socket to the specified address.
750 * On entry, *laddrp and *lportp should contain the current local
751 * address and port for the PCB; these are updated to the values
752 * that should be placed in inp_laddr and inp_lport to complete
753 * the connect.
754 *
755 * On success, *faddrp and *fportp will be set to the remote address
756 * and port. These are not updated in the error case.
757 *
758 * If the operation fails because the connection already exists,
759 * *oinpp will be set to the PCB of that connection so that the
760 * caller can decide to override it. In all other cases, *oinpp
761 * is set to NULL.
762 */
763 int
764 in_pcbconnect_setup(struct inpcb *inp, struct sockaddr *nam,
765 in_addr_t *laddrp, u_short *lportp, in_addr_t *faddrp, u_short *fportp,
766 struct inpcb **oinpp, struct ucred *cred)
767 {
768 struct sockaddr_in *sin = (struct sockaddr_in *)nam;
769 struct in_ifaddr *ia;
770 struct inpcb *oinp;
771 struct in_addr laddr, faddr;
772 u_short lport, fport;
773 int error;
774
775 /*
776 * Because a global state change doesn't actually occur here, a read
777 * lock is sufficient.
778 */
779 INP_INFO_LOCK_ASSERT(inp->inp_pcbinfo);
780 INP_LOCK_ASSERT(inp);
781
782 if (oinpp != NULL)
783 *oinpp = NULL;
784 if (nam->sa_len != sizeof (*sin))
785 return (EINVAL);
786 if (sin->sin_family != AF_INET)
787 return (EAFNOSUPPORT);
788 if (sin->sin_port == 0)
789 return (EADDRNOTAVAIL);
790 laddr.s_addr = *laddrp;
791 lport = *lportp;
792 faddr = sin->sin_addr;
793 fport = sin->sin_port;
794
795 if (!TAILQ_EMPTY(&in_ifaddrhead)) {
796 /*
797 * If the destination address is INADDR_ANY,
798 * use the primary local address.
799 * If the supplied address is INADDR_BROADCAST,
800 * and the primary interface supports broadcast,
801 * choose the broadcast address for that interface.
802 */
803 if (faddr.s_addr == INADDR_ANY) {
804 faddr =
805 IA_SIN(TAILQ_FIRST(&in_ifaddrhead))->sin_addr;
806 if (cred != NULL &&
807 (error = prison_get_ip4(cred, &faddr)) != 0)
808 return (error);
809 } else if (faddr.s_addr == (u_long)INADDR_BROADCAST &&
810 (TAILQ_FIRST(&in_ifaddrhead)->ia_ifp->if_flags &
811 IFF_BROADCAST))
812 faddr = satosin(&TAILQ_FIRST(
813 &in_ifaddrhead)->ia_broadaddr)->sin_addr;
814 }
815 if (laddr.s_addr == INADDR_ANY) {
816 error = in_pcbladdr(inp, &faddr, &laddr, cred);
817 if (error)
818 return (error);
819
820 /*
821 * If the destination address is multicast and an outgoing
822 * interface has been set as a multicast option, use the
823 * address of that interface as our source address.
824 */
825 if (IN_MULTICAST(ntohl(faddr.s_addr)) &&
826 inp->inp_moptions != NULL) {
827 struct ip_moptions *imo;
828 struct ifnet *ifp;
829
830 imo = inp->inp_moptions;
831 if (imo->imo_multicast_ifp != NULL) {
832 ifp = imo->imo_multicast_ifp;
833 TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link)
834 if (ia->ia_ifp == ifp)
835 break;
836 if (ia == NULL)
837 return (EADDRNOTAVAIL);
838 laddr = ia->ia_addr.sin_addr;
839 }
840 }
841 }
842
843 oinp = in_pcblookup_hash(inp->inp_pcbinfo, faddr, fport, laddr, lport,
844 0, NULL);
845 if (oinp != NULL) {
846 if (oinpp != NULL)
847 *oinpp = oinp;
848 return (EADDRINUSE);
849 }
850 if (lport == 0) {
851 error = in_pcbbind_setup(inp, NULL, &laddr.s_addr, &lport,
852 cred);
853 if (error)
854 return (error);
855 }
856 *laddrp = laddr.s_addr;
857 *lportp = lport;
858 *faddrp = faddr.s_addr;
859 *fportp = fport;
860 return (0);
861 }
862
863 void
864 in_pcbdisconnect(struct inpcb *inp)
865 {
866
867 INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
868 INP_WLOCK_ASSERT(inp);
869
870 inp->inp_faddr.s_addr = INADDR_ANY;
871 inp->inp_fport = 0;
872 in_pcbrehash(inp);
873 }
874
875 /*
876 * Historically, in_pcbdetach() included the functionality now found in
877 * in_pcbfree() and in_pcbdrop(). They are now broken out to reflect the
878 * more complex life cycle of TCP.
879 *
880 * in_pcbdetach() is responsibe for disconnecting the socket from an inpcb.
881 * For most protocols, this will be invoked immediately prior to calling
882 * in_pcbfree(). However, for TCP the inpcb may significantly outlive the
883 * socket, in which case in_pcbfree() may be deferred.
884 */
885 void
886 in_pcbdetach(struct inpcb *inp)
887 {
888
889 KASSERT(inp->inp_socket != NULL, ("%s: inp_socket == NULL", __func__));
890
891 inp->inp_socket->so_pcb = NULL;
892 inp->inp_socket = NULL;
893 }
894
895 /*
896 * in_pcbfree() is responsible for freeing an already-detached inpcb, as well
897 * as removing it from any global inpcb lists it might be on.
898 */
899 void
900 in_pcbfree(struct inpcb *inp)
901 {
902 struct inpcbinfo *ipi = inp->inp_pcbinfo;
903
904 KASSERT(inp->inp_socket == NULL, ("%s: inp_socket != NULL", __func__));
905
906 INP_INFO_WLOCK_ASSERT(ipi);
907 INP_WLOCK_ASSERT(inp);
908
909 #ifdef IPSEC
910 if (inp->inp_sp != NULL)
911 ipsec_delete_pcbpolicy(inp);
912 #endif /* IPSEC */
913 inp->inp_gencnt = ++ipi->ipi_gencnt;
914 in_pcbremlists(inp);
915 #ifdef INET6
916 if (inp->inp_vflag & INP_IPV6PROTO) {
917 ip6_freepcbopts(inp->in6p_outputopts);
918 ip6_freemoptions(inp->in6p_moptions);
919 }
920 #endif
921 if (inp->inp_options)
922 (void)m_free(inp->inp_options);
923 if (inp->inp_moptions != NULL)
924 inp_freemoptions(inp->inp_moptions);
925 inp->inp_vflag = 0;
926 crfree(inp->inp_cred);
927
928 #ifdef MAC
929 mac_destroy_inpcb(inp);
930 #endif
931 INP_WUNLOCK(inp);
932 uma_zfree(ipi->ipi_zone, inp);
933 }
934
935 /*
936 * in_pcbdrop() removes an inpcb from hashed lists, releasing its address and
937 * port reservation, and preventing it from being returned by inpcb lookups.
938 *
939 * It is used by TCP to mark an inpcb as unused and avoid future packet
940 * delivery or event notification when a socket remains open but TCP has
941 * closed. This might occur as a result of a shutdown()-initiated TCP close
942 * or a RST on the wire, and allows the port binding to be reused while still
943 * maintaining the invariant that so_pcb always points to a valid inpcb until
944 * in_pcbdetach().
945 *
946 * XXXRW: An inp_lport of 0 is used to indicate that the inpcb is not on hash
947 * lists, but can lead to confusing netstat output, as open sockets with
948 * closed TCP connections will no longer appear to have their bound port
949 * number. An explicit flag would be better, as it would allow us to leave
950 * the port number intact after the connection is dropped.
951 *
952 * XXXRW: Possibly in_pcbdrop() should also prevent future notifications by
953 * in_pcbnotifyall() and in_pcbpurgeif0()?
954 */
955 void
956 in_pcbdrop(struct inpcb *inp)
957 {
958
959 INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
960 INP_WLOCK_ASSERT(inp);
961
962 inp->inp_flags |= INP_DROPPED;
963 if (inp->inp_flags & INP_INHASHLIST) {
964 struct inpcbport *phd = inp->inp_phd;
965
966 LIST_REMOVE(inp, inp_hash);
967 LIST_REMOVE(inp, inp_portlist);
968 if (LIST_FIRST(&phd->phd_pcblist) == NULL) {
969 LIST_REMOVE(phd, phd_hash);
970 free(phd, M_PCB);
971 }
972 inp->inp_flags &= ~INP_INHASHLIST;
973 }
974 }
975
976 /*
977 * Common routines to return the socket addresses associated with inpcbs.
978 */
979 struct sockaddr *
980 in_sockaddr(in_port_t port, struct in_addr *addr_p)
981 {
982 struct sockaddr_in *sin;
983
984 MALLOC(sin, struct sockaddr_in *, sizeof *sin, M_SONAME,
985 M_WAITOK | M_ZERO);
986 sin->sin_family = AF_INET;
987 sin->sin_len = sizeof(*sin);
988 sin->sin_addr = *addr_p;
989 sin->sin_port = port;
990
991 return (struct sockaddr *)sin;
992 }
993
994 int
995 in_getsockaddr(struct socket *so, struct sockaddr **nam)
996 {
997 struct inpcb *inp;
998 struct in_addr addr;
999 in_port_t port;
1000
1001 inp = sotoinpcb(so);
1002 KASSERT(inp != NULL, ("in_getsockaddr: inp == NULL"));
1003
1004 INP_RLOCK(inp);
1005 port = inp->inp_lport;
1006 addr = inp->inp_laddr;
1007 INP_RUNLOCK(inp);
1008
1009 *nam = in_sockaddr(port, &addr);
1010 return 0;
1011 }
1012
1013 int
1014 in_getpeeraddr(struct socket *so, struct sockaddr **nam)
1015 {
1016 struct inpcb *inp;
1017 struct in_addr addr;
1018 in_port_t port;
1019
1020 inp = sotoinpcb(so);
1021 KASSERT(inp != NULL, ("in_getpeeraddr: inp == NULL"));
1022
1023 INP_RLOCK(inp);
1024 port = inp->inp_fport;
1025 addr = inp->inp_faddr;
1026 INP_RUNLOCK(inp);
1027
1028 *nam = in_sockaddr(port, &addr);
1029 return 0;
1030 }
1031
1032 void
1033 in_pcbnotifyall(struct inpcbinfo *pcbinfo, struct in_addr faddr, int errno,
1034 struct inpcb *(*notify)(struct inpcb *, int))
1035 {
1036 struct inpcb *inp, *inp_temp;
1037
1038 INP_INFO_WLOCK(pcbinfo);
1039 LIST_FOREACH_SAFE(inp, pcbinfo->ipi_listhead, inp_list, inp_temp) {
1040 INP_WLOCK(inp);
1041 #ifdef INET6
1042 if ((inp->inp_vflag & INP_IPV4) == 0) {
1043 INP_WUNLOCK(inp);
1044 continue;
1045 }
1046 #endif
1047 if (inp->inp_faddr.s_addr != faddr.s_addr ||
1048 inp->inp_socket == NULL) {
1049 INP_WUNLOCK(inp);
1050 continue;
1051 }
1052 if ((*notify)(inp, errno))
1053 INP_WUNLOCK(inp);
1054 }
1055 INP_INFO_WUNLOCK(pcbinfo);
1056 }
1057
1058 void
1059 in_pcbpurgeif0(struct inpcbinfo *pcbinfo, struct ifnet *ifp)
1060 {
1061 struct inpcb *inp;
1062 struct ip_moptions *imo;
1063 int i, gap;
1064
1065 INP_INFO_RLOCK(pcbinfo);
1066 LIST_FOREACH(inp, pcbinfo->ipi_listhead, inp_list) {
1067 INP_WLOCK(inp);
1068 imo = inp->inp_moptions;
1069 if ((inp->inp_vflag & INP_IPV4) &&
1070 imo != NULL) {
1071 /*
1072 * Unselect the outgoing interface if it is being
1073 * detached.
1074 */
1075 if (imo->imo_multicast_ifp == ifp)
1076 imo->imo_multicast_ifp = NULL;
1077
1078 /*
1079 * Drop multicast group membership if we joined
1080 * through the interface being detached.
1081 */
1082 for (i = 0, gap = 0; i < imo->imo_num_memberships;
1083 i++) {
1084 if (imo->imo_membership[i]->inm_ifp == ifp) {
1085 in_delmulti(imo->imo_membership[i]);
1086 gap++;
1087 } else if (gap != 0)
1088 imo->imo_membership[i - gap] =
1089 imo->imo_membership[i];
1090 }
1091 imo->imo_num_memberships -= gap;
1092 }
1093 INP_WUNLOCK(inp);
1094 }
1095 INP_INFO_RUNLOCK(pcbinfo);
1096 }
1097
1098 /*
1099 * Lookup a PCB based on the local address and port.
1100 */
1101 #define INP_LOOKUP_MAPPED_PCB_COST 3
1102 struct inpcb *
1103 in_pcblookup_local(struct inpcbinfo *pcbinfo, struct in_addr laddr,
1104 u_short lport, int wild_okay, struct ucred *cred)
1105 {
1106 struct inpcb *inp;
1107 #ifdef INET6
1108 int matchwild = 3 + INP_LOOKUP_MAPPED_PCB_COST;
1109 #else
1110 int matchwild = 3;
1111 #endif
1112 int wildcard;
1113
1114 INP_INFO_LOCK_ASSERT(pcbinfo);
1115
1116 if (!wild_okay) {
1117 struct inpcbhead *head;
1118 /*
1119 * Look for an unconnected (wildcard foreign addr) PCB that
1120 * matches the local address and port we're looking for.
1121 */
1122 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport,
1123 0, pcbinfo->ipi_hashmask)];
1124 LIST_FOREACH(inp, head, inp_hash) {
1125 #ifdef INET6
1126 /* XXX inp locking */
1127 if ((inp->inp_vflag & INP_IPV4) == 0)
1128 continue;
1129 #endif
1130 if (inp->inp_faddr.s_addr == INADDR_ANY &&
1131 inp->inp_laddr.s_addr == laddr.s_addr &&
1132 inp->inp_lport == lport) {
1133 /*
1134 * Found?
1135 */
1136 if (cred == NULL ||
1137 inp->inp_cred->cr_prison == cred->cr_prison)
1138 return (inp);
1139 }
1140 }
1141 /*
1142 * Not found.
1143 */
1144 return (NULL);
1145 } else {
1146 struct inpcbporthead *porthash;
1147 struct inpcbport *phd;
1148 struct inpcb *match = NULL;
1149 /*
1150 * Best fit PCB lookup.
1151 *
1152 * First see if this local port is in use by looking on the
1153 * port hash list.
1154 */
1155 porthash = &pcbinfo->ipi_porthashbase[INP_PCBPORTHASH(lport,
1156 pcbinfo->ipi_porthashmask)];
1157 LIST_FOREACH(phd, porthash, phd_hash) {
1158 if (phd->phd_port == lport)
1159 break;
1160 }
1161 if (phd != NULL) {
1162 /*
1163 * Port is in use by one or more PCBs. Look for best
1164 * fit.
1165 */
1166 LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) {
1167 wildcard = 0;
1168 if (cred != NULL &&
1169 inp->inp_cred->cr_prison != cred->cr_prison)
1170 continue;
1171 #ifdef INET6
1172 /* XXX inp locking */
1173 if ((inp->inp_vflag & INP_IPV4) == 0)
1174 continue;
1175 /*
1176 * We never select the PCB that has
1177 * INP_IPV6 flag and is bound to :: if
1178 * we have another PCB which is bound
1179 * to 0.0.0.0. If a PCB has the
1180 * INP_IPV6 flag, then we set its cost
1181 * higher than IPv4 only PCBs.
1182 *
1183 * Note that the case only happens
1184 * when a socket is bound to ::, under
1185 * the condition that the use of the
1186 * mapped address is allowed.
1187 */
1188 if ((inp->inp_vflag & INP_IPV6) != 0)
1189 wildcard += INP_LOOKUP_MAPPED_PCB_COST;
1190 #endif
1191 if (inp->inp_faddr.s_addr != INADDR_ANY)
1192 wildcard++;
1193 if (inp->inp_laddr.s_addr != INADDR_ANY) {
1194 if (laddr.s_addr == INADDR_ANY)
1195 wildcard++;
1196 else if (inp->inp_laddr.s_addr != laddr.s_addr)
1197 continue;
1198 } else {
1199 if (laddr.s_addr != INADDR_ANY)
1200 wildcard++;
1201 }
1202 if (wildcard < matchwild) {
1203 match = inp;
1204 matchwild = wildcard;
1205 if (matchwild == 0)
1206 break;
1207 }
1208 }
1209 }
1210 return (match);
1211 }
1212 }
1213 #undef INP_LOOKUP_MAPPED_PCB_COST
1214
1215 /*
1216 * Lookup PCB in hash list.
1217 */
1218 struct inpcb *
1219 in_pcblookup_hash(struct inpcbinfo *pcbinfo, struct in_addr faddr,
1220 u_int fport_arg, struct in_addr laddr, u_int lport_arg, int wildcard,
1221 struct ifnet *ifp)
1222 {
1223 struct inpcbhead *head;
1224 struct inpcb *inp, *tmpinp;
1225 u_short fport = fport_arg, lport = lport_arg;
1226
1227 INP_INFO_LOCK_ASSERT(pcbinfo);
1228
1229 /*
1230 * First look for an exact match.
1231 */
1232 tmpinp = NULL;
1233 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(faddr.s_addr, lport, fport,
1234 pcbinfo->ipi_hashmask)];
1235 LIST_FOREACH(inp, head, inp_hash) {
1236 #ifdef INET6
1237 /* XXX inp locking */
1238 if ((inp->inp_vflag & INP_IPV4) == 0)
1239 continue;
1240 #endif
1241 if (inp->inp_faddr.s_addr == faddr.s_addr &&
1242 inp->inp_laddr.s_addr == laddr.s_addr &&
1243 inp->inp_fport == fport &&
1244 inp->inp_lport == lport) {
1245 /*
1246 * XXX We should be able to directly return
1247 * the inp here, without any checks.
1248 * Well unless both bound with SO_REUSEPORT?
1249 */
1250 if (jailed(inp->inp_cred))
1251 return (inp);
1252 if (tmpinp == NULL)
1253 tmpinp = inp;
1254 }
1255 }
1256 if (tmpinp != NULL)
1257 return (tmpinp);
1258
1259 /*
1260 * Then look for a wildcard match, if requested.
1261 */
1262 if (wildcard == INPLOOKUP_WILDCARD) {
1263 struct inpcb *local_wild = NULL, *local_exact = NULL;
1264 #ifdef INET6
1265 struct inpcb *local_wild_mapped = NULL;
1266 #endif
1267 struct inpcb *jail_wild = NULL;
1268 int injail;
1269
1270 /*
1271 * Order of socket selection - we always prefer jails.
1272 * 1. jailed, non-wild.
1273 * 2. jailed, wild.
1274 * 3. non-jailed, non-wild.
1275 * 4. non-jailed, wild.
1276 */
1277
1278 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport,
1279 0, pcbinfo->ipi_hashmask)];
1280 LIST_FOREACH(inp, head, inp_hash) {
1281 #ifdef INET6
1282 /* XXX inp locking */
1283 if ((inp->inp_vflag & INP_IPV4) == 0)
1284 continue;
1285 #endif
1286 if (inp->inp_faddr.s_addr != INADDR_ANY ||
1287 inp->inp_lport != lport)
1288 continue;
1289
1290 /* XXX inp locking */
1291 if (ifp && ifp->if_type == IFT_FAITH &&
1292 (inp->inp_flags & INP_FAITH) == 0)
1293 continue;
1294
1295 injail = jailed(inp->inp_cred);
1296 if (injail) {
1297 if (prison_check_ip4(inp->inp_cred,
1298 &laddr) != 0)
1299 continue;
1300 } else {
1301 if (local_exact != NULL)
1302 continue;
1303 }
1304
1305 if (inp->inp_laddr.s_addr == laddr.s_addr) {
1306 if (injail)
1307 return (inp);
1308 else
1309 local_exact = inp;
1310 } else if (inp->inp_laddr.s_addr == INADDR_ANY) {
1311 #ifdef INET6
1312 /* XXX inp locking, NULL check */
1313 if (inp->inp_vflag & INP_IPV6PROTO)
1314 local_wild_mapped = inp;
1315 else
1316 #endif /* INET6 */
1317 if (injail)
1318 jail_wild = inp;
1319 else
1320 local_wild = inp;
1321 }
1322 } /* LIST_FOREACH */
1323 if (jail_wild != NULL)
1324 return (jail_wild);
1325 if (local_exact != NULL)
1326 return (local_exact);
1327 if (local_wild != NULL)
1328 return (local_wild);
1329 #ifdef INET6
1330 if (local_wild_mapped != NULL)
1331 return (local_wild_mapped);
1332 #endif /* defined(INET6) */
1333 } /* if (wildcard == INPLOOKUP_WILDCARD) */
1334
1335 return (NULL);
1336 }
1337
1338 /*
1339 * Insert PCB onto various hash lists.
1340 */
1341 int
1342 in_pcbinshash(struct inpcb *inp)
1343 {
1344 struct inpcbhead *pcbhash;
1345 struct inpcbporthead *pcbporthash;
1346 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1347 struct inpcbport *phd;
1348 u_int32_t hashkey_faddr;
1349
1350 INP_INFO_WLOCK_ASSERT(pcbinfo);
1351 INP_WLOCK_ASSERT(inp);
1352 KASSERT((inp->inp_flags & INP_INHASHLIST) == 0,
1353 ("in_pcbinshash: INP_INHASHLIST"));
1354
1355 #ifdef INET6
1356 if (inp->inp_vflag & INP_IPV6)
1357 hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX */;
1358 else
1359 #endif /* INET6 */
1360 hashkey_faddr = inp->inp_faddr.s_addr;
1361
1362 pcbhash = &pcbinfo->ipi_hashbase[INP_PCBHASH(hashkey_faddr,
1363 inp->inp_lport, inp->inp_fport, pcbinfo->ipi_hashmask)];
1364
1365 pcbporthash = &pcbinfo->ipi_porthashbase[
1366 INP_PCBPORTHASH(inp->inp_lport, pcbinfo->ipi_porthashmask)];
1367
1368 /*
1369 * Go through port list and look for a head for this lport.
1370 */
1371 LIST_FOREACH(phd, pcbporthash, phd_hash) {
1372 if (phd->phd_port == inp->inp_lport)
1373 break;
1374 }
1375 /*
1376 * If none exists, malloc one and tack it on.
1377 */
1378 if (phd == NULL) {
1379 MALLOC(phd, struct inpcbport *, sizeof(struct inpcbport), M_PCB, M_NOWAIT);
1380 if (phd == NULL) {
1381 return (ENOBUFS); /* XXX */
1382 }
1383 phd->phd_port = inp->inp_lport;
1384 LIST_INIT(&phd->phd_pcblist);
1385 LIST_INSERT_HEAD(pcbporthash, phd, phd_hash);
1386 }
1387 inp->inp_phd = phd;
1388 LIST_INSERT_HEAD(&phd->phd_pcblist, inp, inp_portlist);
1389 LIST_INSERT_HEAD(pcbhash, inp, inp_hash);
1390 inp->inp_flags |= INP_INHASHLIST;
1391 return (0);
1392 }
1393
1394 /*
1395 * Move PCB to the proper hash bucket when { faddr, fport } have been
1396 * changed. NOTE: This does not handle the case of the lport changing (the
1397 * hashed port list would have to be updated as well), so the lport must
1398 * not change after in_pcbinshash() has been called.
1399 */
1400 void
1401 in_pcbrehash(struct inpcb *inp)
1402 {
1403 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1404 struct inpcbhead *head;
1405 u_int32_t hashkey_faddr;
1406
1407 INP_INFO_WLOCK_ASSERT(pcbinfo);
1408 INP_WLOCK_ASSERT(inp);
1409 KASSERT(inp->inp_flags & INP_INHASHLIST,
1410 ("in_pcbrehash: !INP_INHASHLIST"));
1411
1412 #ifdef INET6
1413 if (inp->inp_vflag & INP_IPV6)
1414 hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX */;
1415 else
1416 #endif /* INET6 */
1417 hashkey_faddr = inp->inp_faddr.s_addr;
1418
1419 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(hashkey_faddr,
1420 inp->inp_lport, inp->inp_fport, pcbinfo->ipi_hashmask)];
1421
1422 LIST_REMOVE(inp, inp_hash);
1423 LIST_INSERT_HEAD(head, inp, inp_hash);
1424 }
1425
1426 /*
1427 * Remove PCB from various lists.
1428 */
1429 void
1430 in_pcbremlists(struct inpcb *inp)
1431 {
1432 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1433
1434 INP_INFO_WLOCK_ASSERT(pcbinfo);
1435 INP_WLOCK_ASSERT(inp);
1436
1437 inp->inp_gencnt = ++pcbinfo->ipi_gencnt;
1438 if (inp->inp_flags & INP_INHASHLIST) {
1439 struct inpcbport *phd = inp->inp_phd;
1440
1441 LIST_REMOVE(inp, inp_hash);
1442 LIST_REMOVE(inp, inp_portlist);
1443 if (LIST_FIRST(&phd->phd_pcblist) == NULL) {
1444 LIST_REMOVE(phd, phd_hash);
1445 free(phd, M_PCB);
1446 }
1447 inp->inp_flags &= ~INP_INHASHLIST;
1448 }
1449 LIST_REMOVE(inp, inp_list);
1450 pcbinfo->ipi_count--;
1451 }
1452
1453 /*
1454 * A set label operation has occurred at the socket layer, propagate the
1455 * label change into the in_pcb for the socket.
1456 */
1457 void
1458 in_pcbsosetlabel(struct socket *so)
1459 {
1460 #ifdef MAC
1461 struct inpcb *inp;
1462
1463 inp = sotoinpcb(so);
1464 KASSERT(inp != NULL, ("in_pcbsosetlabel: so->so_pcb == NULL"));
1465
1466 INP_WLOCK(inp);
1467 SOCK_LOCK(so);
1468 mac_inpcb_sosetlabel(so, inp);
1469 SOCK_UNLOCK(so);
1470 INP_WUNLOCK(inp);
1471 #endif
1472 }
1473
1474 /*
1475 * ipport_tick runs once per second, determining if random port allocation
1476 * should be continued. If more than ipport_randomcps ports have been
1477 * allocated in the last second, then we return to sequential port
1478 * allocation. We return to random allocation only once we drop below
1479 * ipport_randomcps for at least ipport_randomtime seconds.
1480 */
1481 void
1482 ipport_tick(void *xtp)
1483 {
1484
1485 if (ipport_tcpallocs <= ipport_tcplastcount + ipport_randomcps) {
1486 if (ipport_stoprandom > 0)
1487 ipport_stoprandom--;
1488 } else
1489 ipport_stoprandom = ipport_randomtime;
1490 ipport_tcplastcount = ipport_tcpallocs;
1491 callout_reset(&ipport_tick_callout, hz, ipport_tick, NULL);
1492 }
1493
1494 void
1495 inp_apply_all(void (*func)(struct inpcb *, void *), void *arg)
1496 {
1497 struct inpcb *inp;
1498
1499 INP_INFO_RLOCK(&tcbinfo);
1500 LIST_FOREACH(inp, tcbinfo.ipi_listhead, inp_list) {
1501 INP_WLOCK(inp);
1502 func(inp, arg);
1503 INP_WUNLOCK(inp);
1504 }
1505 INP_INFO_RUNLOCK(&tcbinfo);
1506 }
1507
1508 struct socket *
1509 inp_inpcbtosocket(struct inpcb *inp)
1510 {
1511
1512 INP_WLOCK_ASSERT(inp);
1513 return (inp->inp_socket);
1514 }
1515
1516 struct tcpcb *
1517 inp_inpcbtotcpcb(struct inpcb *inp)
1518 {
1519
1520 INP_WLOCK_ASSERT(inp);
1521 return ((struct tcpcb *)inp->inp_ppcb);
1522 }
1523
1524 int
1525 inp_ip_tos_get(const struct inpcb *inp)
1526 {
1527
1528 return (inp->inp_ip_tos);
1529 }
1530
1531 void
1532 inp_ip_tos_set(struct inpcb *inp, int val)
1533 {
1534
1535 inp->inp_ip_tos = val;
1536 }
1537
1538 void
1539 inp_4tuple_get(struct inpcb *inp, uint32_t *laddr, uint16_t *lp,
1540 uint32_t *faddr, uint16_t *fp)
1541 {
1542
1543 INP_LOCK_ASSERT(inp);
1544 *laddr = inp->inp_laddr.s_addr;
1545 *faddr = inp->inp_faddr.s_addr;
1546 *lp = inp->inp_lport;
1547 *fp = inp->inp_fport;
1548 }
1549
1550 struct inpcb *
1551 so_sotoinpcb(struct socket *so)
1552 {
1553
1554 return (sotoinpcb(so));
1555 }
1556
1557 struct tcpcb *
1558 so_sototcpcb(struct socket *so)
1559 {
1560
1561 return (sototcpcb(so));
1562 }
1563
1564 void
1565 inp_wlock(struct inpcb *inp)
1566 {
1567
1568 INP_WLOCK(inp);
1569 }
1570
1571 void
1572 inp_wunlock(struct inpcb *inp)
1573 {
1574
1575 INP_WUNLOCK(inp);
1576 }
1577
1578 void
1579 inp_rlock(struct inpcb *inp)
1580 {
1581
1582 INP_RLOCK(inp);
1583 }
1584
1585 void
1586 inp_runlock(struct inpcb *inp)
1587 {
1588
1589 INP_RUNLOCK(inp);
1590 }
1591
1592 #ifdef INVARIANTS
1593 void
1594 inp_wlock_assert(struct inpcb *inp)
1595 {
1596
1597 INP_WLOCK_ASSERT(inp);
1598 }
1599
1600 void
1601 inp_rlock_assert(struct inpcb *inp)
1602 {
1603
1604 INP_RLOCK_ASSERT(inp);
1605 }
1606
1607 void
1608 inp_lock_assert(struct inpcb *inp)
1609 {
1610
1611 INP_LOCK_ASSERT(inp);
1612 }
1613
1614 void
1615 inp_unlock_assert(struct inpcb *inp)
1616 {
1617
1618 INP_UNLOCK_ASSERT(inp);
1619 }
1620
1621 #endif
1622
1623 #ifdef DDB
1624 static void
1625 db_print_indent(int indent)
1626 {
1627 int i;
1628
1629 for (i = 0; i < indent; i++)
1630 db_printf(" ");
1631 }
1632
1633 static void
1634 db_print_inconninfo(struct in_conninfo *inc, const char *name, int indent)
1635 {
1636 char faddr_str[48], laddr_str[48];
1637
1638 db_print_indent(indent);
1639 db_printf("%s at %p\n", name, inc);
1640
1641 indent += 2;
1642
1643 #ifdef INET6
1644 if (inc->inc_flags & INC_ISIPV6) {
1645 /* IPv6. */
1646 ip6_sprintf(laddr_str, &inc->inc6_laddr);
1647 ip6_sprintf(faddr_str, &inc->inc6_faddr);
1648 } else {
1649 #endif
1650 /* IPv4. */
1651 inet_ntoa_r(inc->inc_laddr, laddr_str);
1652 inet_ntoa_r(inc->inc_faddr, faddr_str);
1653 #ifdef INET6
1654 }
1655 #endif
1656 db_print_indent(indent);
1657 db_printf("inc_laddr %s inc_lport %u\n", laddr_str,
1658 ntohs(inc->inc_lport));
1659 db_print_indent(indent);
1660 db_printf("inc_faddr %s inc_fport %u\n", faddr_str,
1661 ntohs(inc->inc_fport));
1662 }
1663
1664 static void
1665 db_print_inpflags(int inp_flags)
1666 {
1667 int comma;
1668
1669 comma = 0;
1670 if (inp_flags & INP_RECVOPTS) {
1671 db_printf("%sINP_RECVOPTS", comma ? ", " : "");
1672 comma = 1;
1673 }
1674 if (inp_flags & INP_RECVRETOPTS) {
1675 db_printf("%sINP_RECVRETOPTS", comma ? ", " : "");
1676 comma = 1;
1677 }
1678 if (inp_flags & INP_RECVDSTADDR) {
1679 db_printf("%sINP_RECVDSTADDR", comma ? ", " : "");
1680 comma = 1;
1681 }
1682 if (inp_flags & INP_HDRINCL) {
1683 db_printf("%sINP_HDRINCL", comma ? ", " : "");
1684 comma = 1;
1685 }
1686 if (inp_flags & INP_HIGHPORT) {
1687 db_printf("%sINP_HIGHPORT", comma ? ", " : "");
1688 comma = 1;
1689 }
1690 if (inp_flags & INP_LOWPORT) {
1691 db_printf("%sINP_LOWPORT", comma ? ", " : "");
1692 comma = 1;
1693 }
1694 if (inp_flags & INP_ANONPORT) {
1695 db_printf("%sINP_ANONPORT", comma ? ", " : "");
1696 comma = 1;
1697 }
1698 if (inp_flags & INP_RECVIF) {
1699 db_printf("%sINP_RECVIF", comma ? ", " : "");
1700 comma = 1;
1701 }
1702 if (inp_flags & INP_MTUDISC) {
1703 db_printf("%sINP_MTUDISC", comma ? ", " : "");
1704 comma = 1;
1705 }
1706 if (inp_flags & INP_FAITH) {
1707 db_printf("%sINP_FAITH", comma ? ", " : "");
1708 comma = 1;
1709 }
1710 if (inp_flags & INP_RECVTTL) {
1711 db_printf("%sINP_RECVTTL", comma ? ", " : "");
1712 comma = 1;
1713 }
1714 if (inp_flags & INP_DONTFRAG) {
1715 db_printf("%sINP_DONTFRAG", comma ? ", " : "");
1716 comma = 1;
1717 }
1718 if (inp_flags & IN6P_IPV6_V6ONLY) {
1719 db_printf("%sIN6P_IPV6_V6ONLY", comma ? ", " : "");
1720 comma = 1;
1721 }
1722 if (inp_flags & IN6P_PKTINFO) {
1723 db_printf("%sIN6P_PKTINFO", comma ? ", " : "");
1724 comma = 1;
1725 }
1726 if (inp_flags & IN6P_HOPLIMIT) {
1727 db_printf("%sIN6P_HOPLIMIT", comma ? ", " : "");
1728 comma = 1;
1729 }
1730 if (inp_flags & IN6P_HOPOPTS) {
1731 db_printf("%sIN6P_HOPOPTS", comma ? ", " : "");
1732 comma = 1;
1733 }
1734 if (inp_flags & IN6P_DSTOPTS) {
1735 db_printf("%sIN6P_DSTOPTS", comma ? ", " : "");
1736 comma = 1;
1737 }
1738 if (inp_flags & IN6P_RTHDR) {
1739 db_printf("%sIN6P_RTHDR", comma ? ", " : "");
1740 comma = 1;
1741 }
1742 if (inp_flags & IN6P_RTHDRDSTOPTS) {
1743 db_printf("%sIN6P_RTHDRDSTOPTS", comma ? ", " : "");
1744 comma = 1;
1745 }
1746 if (inp_flags & IN6P_TCLASS) {
1747 db_printf("%sIN6P_TCLASS", comma ? ", " : "");
1748 comma = 1;
1749 }
1750 if (inp_flags & IN6P_AUTOFLOWLABEL) {
1751 db_printf("%sIN6P_AUTOFLOWLABEL", comma ? ", " : "");
1752 comma = 1;
1753 }
1754 if (inp_flags & INP_TIMEWAIT) {
1755 db_printf("%sINP_TIMEWAIT", comma ? ", " : "");
1756 comma = 1;
1757 }
1758 if (inp_flags & INP_ONESBCAST) {
1759 db_printf("%sINP_ONESBCAST", comma ? ", " : "");
1760 comma = 1;
1761 }
1762 if (inp_flags & INP_DROPPED) {
1763 db_printf("%sINP_DROPPED", comma ? ", " : "");
1764 comma = 1;
1765 }
1766 if (inp_flags & INP_SOCKREF) {
1767 db_printf("%sINP_SOCKREF", comma ? ", " : "");
1768 comma = 1;
1769 }
1770 if (inp_flags & IN6P_RFC2292) {
1771 db_printf("%sIN6P_RFC2292", comma ? ", " : "");
1772 comma = 1;
1773 }
1774 if (inp_flags & IN6P_MTU) {
1775 db_printf("IN6P_MTU%s", comma ? ", " : "");
1776 comma = 1;
1777 }
1778 }
1779
1780 static void
1781 db_print_inpvflag(u_char inp_vflag)
1782 {
1783 int comma;
1784
1785 comma = 0;
1786 if (inp_vflag & INP_IPV4) {
1787 db_printf("%sINP_IPV4", comma ? ", " : "");
1788 comma = 1;
1789 }
1790 if (inp_vflag & INP_IPV6) {
1791 db_printf("%sINP_IPV6", comma ? ", " : "");
1792 comma = 1;
1793 }
1794 if (inp_vflag & INP_IPV6PROTO) {
1795 db_printf("%sINP_IPV6PROTO", comma ? ", " : "");
1796 comma = 1;
1797 }
1798 }
1799
1800 void
1801 db_print_inpcb(struct inpcb *inp, const char *name, int indent)
1802 {
1803
1804 db_print_indent(indent);
1805 db_printf("%s at %p\n", name, inp);
1806
1807 indent += 2;
1808
1809 db_print_indent(indent);
1810 db_printf("inp_flow: 0x%x\n", inp->inp_flow);
1811
1812 db_print_inconninfo(&inp->inp_inc, "inp_conninfo", indent);
1813
1814 db_print_indent(indent);
1815 db_printf("inp_ppcb: %p inp_pcbinfo: %p inp_socket: %p\n",
1816 inp->inp_ppcb, inp->inp_pcbinfo, inp->inp_socket);
1817
1818 db_print_indent(indent);
1819 db_printf("inp_label: %p inp_flags: 0x%x (",
1820 inp->inp_label, inp->inp_flags);
1821 db_print_inpflags(inp->inp_flags);
1822 db_printf(")\n");
1823
1824 db_print_indent(indent);
1825 db_printf("inp_sp: %p inp_vflag: 0x%x (", inp->inp_sp,
1826 inp->inp_vflag);
1827 db_print_inpvflag(inp->inp_vflag);
1828 db_printf(")\n");
1829
1830 db_print_indent(indent);
1831 db_printf("inp_ip_ttl: %d inp_ip_p: %d inp_ip_minttl: %d\n",
1832 inp->inp_ip_ttl, inp->inp_ip_p, inp->inp_ip_minttl);
1833
1834 db_print_indent(indent);
1835 #ifdef INET6
1836 if (inp->inp_vflag & INP_IPV6) {
1837 db_printf("in6p_options: %p in6p_outputopts: %p "
1838 "in6p_moptions: %p\n", inp->in6p_options,
1839 inp->in6p_outputopts, inp->in6p_moptions);
1840 db_printf("in6p_icmp6filt: %p in6p_cksum %d "
1841 "in6p_hops %u\n", inp->in6p_icmp6filt, inp->in6p_cksum,
1842 inp->in6p_hops);
1843 } else
1844 #endif
1845 {
1846 db_printf("inp_ip_tos: %d inp_ip_options: %p "
1847 "inp_ip_moptions: %p\n", inp->inp_ip_tos,
1848 inp->inp_options, inp->inp_moptions);
1849 }
1850
1851 db_print_indent(indent);
1852 db_printf("inp_phd: %p inp_gencnt: %ju\n", inp->inp_phd,
1853 (uintmax_t)inp->inp_gencnt);
1854 }
1855
1856 DB_SHOW_COMMAND(inpcb, db_show_inpcb)
1857 {
1858 struct inpcb *inp;
1859
1860 if (!have_addr) {
1861 db_printf("usage: show inpcb <addr>\n");
1862 return;
1863 }
1864 inp = (struct inpcb *)addr;
1865
1866 db_print_inpcb(inp, "inpcb", 0);
1867 }
1868 #endif
Cache object: 4cea5816234c11e6d0e806fc3633b488
|