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("$FreeBSD$");
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 error = 0;
556 ia = NULL;
557 bzero(&sro, sizeof(sro));
558
559 sin = (struct sockaddr_in *)&sro.ro_dst;
560 sin->sin_family = AF_INET;
561 sin->sin_len = sizeof(struct sockaddr_in);
562 sin->sin_addr.s_addr = faddr->s_addr;
563
564 /*
565 * If route is known our src addr is taken from the i/f,
566 * else punt.
567 *
568 * Find out route to destination.
569 */
570 if ((inp->inp_socket->so_options & SO_DONTROUTE) == 0)
571 in_rtalloc_ign(&sro, RTF_CLONING, inp->inp_inc.inc_fibnum);
572
573 /*
574 * If we found a route, use the address corresponding to
575 * the outgoing interface.
576 *
577 * Otherwise assume faddr is reachable on a directly connected
578 * network and try to find a corresponding interface to take
579 * the source address from.
580 */
581 if (sro.ro_rt == NULL || sro.ro_rt->rt_ifp == NULL) {
582 struct ifnet *ifp;
583
584 ia = ifatoia(ifa_ifwithdstaddr((struct sockaddr *)sin));
585 if (ia == NULL)
586 ia = ifatoia(ifa_ifwithnet((struct sockaddr *)sin));
587 if (ia == NULL) {
588 error = ENETUNREACH;
589 goto done;
590 }
591
592 if (cred == NULL || !jailed(cred)) {
593 laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
594 goto done;
595 }
596
597 ifp = ia->ia_ifp;
598 ia = NULL;
599 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
600
601 sa = ifa->ifa_addr;
602 if (sa->sa_family != AF_INET)
603 continue;
604 sin = (struct sockaddr_in *)sa;
605 if (prison_check_ip4(cred, &sin->sin_addr) == 0) {
606 ia = (struct in_ifaddr *)ifa;
607 break;
608 }
609 }
610 if (ia != NULL) {
611 laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
612 goto done;
613 }
614
615 /* 3. As a last resort return the 'default' jail address. */
616 error = prison_get_ip4(cred, laddr);
617 goto done;
618 }
619
620 /*
621 * If the outgoing interface on the route found is not
622 * a loopback interface, use the address from that interface.
623 * In case of jails do those three steps:
624 * 1. check if the interface address belongs to the jail. If so use it.
625 * 2. check if we have any address on the outgoing interface
626 * belonging to this jail. If so use it.
627 * 3. as a last resort return the 'default' jail address.
628 */
629 if ((sro.ro_rt->rt_ifp->if_flags & IFF_LOOPBACK) == 0) {
630
631 /* If not jailed, use the default returned. */
632 if (cred == NULL || !jailed(cred)) {
633 ia = (struct in_ifaddr *)sro.ro_rt->rt_ifa;
634 laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
635 goto done;
636 }
637
638 /* Jailed. */
639 /* 1. Check if the iface address belongs to the jail. */
640 sin = (struct sockaddr_in *)sro.ro_rt->rt_ifa->ifa_addr;
641 if (prison_check_ip4(cred, &sin->sin_addr) == 0) {
642 ia = (struct in_ifaddr *)sro.ro_rt->rt_ifa;
643 laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
644 goto done;
645 }
646
647 /*
648 * 2. Check if we have any address on the outgoing interface
649 * belonging to this jail.
650 */
651 TAILQ_FOREACH(ifa, &sro.ro_rt->rt_ifp->if_addrhead, ifa_link) {
652
653 sa = ifa->ifa_addr;
654 if (sa->sa_family != AF_INET)
655 continue;
656 sin = (struct sockaddr_in *)sa;
657 if (prison_check_ip4(cred, &sin->sin_addr) == 0) {
658 ia = (struct in_ifaddr *)ifa;
659 break;
660 }
661 }
662 if (ia != NULL) {
663 laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
664 goto done;
665 }
666
667 /* 3. As a last resort return the 'default' jail address. */
668 error = prison_get_ip4(cred, laddr);
669 goto done;
670 }
671
672 /*
673 * The outgoing interface is marked with 'loopback net', so a route
674 * to ourselves is here.
675 * Try to find the interface of the destination address and then
676 * take the address from there. That interface is not necessarily
677 * a loopback interface.
678 * In case of jails, check that it is an address of the jail
679 * and if we cannot find, fall back to the 'default' jail address.
680 */
681 if ((sro.ro_rt->rt_ifp->if_flags & IFF_LOOPBACK) != 0) {
682 struct sockaddr_in sain;
683
684 bzero(&sain, sizeof(struct sockaddr_in));
685 sain.sin_family = AF_INET;
686 sain.sin_len = sizeof(struct sockaddr_in);
687 sain.sin_addr.s_addr = faddr->s_addr;
688
689 ia = ifatoia(ifa_ifwithdstaddr(sintosa(&sain)));
690 if (ia == NULL)
691 ia = ifatoia(ifa_ifwithnet(sintosa(&sain)));
692
693 if (cred == NULL || !jailed(cred)) {
694 #if __FreeBSD_version < 800000
695 if (ia == NULL)
696 ia = (struct in_ifaddr *)sro.ro_rt->rt_ifa;
697 #endif
698 if (ia == NULL) {
699 error = ENETUNREACH;
700 goto done;
701 }
702 laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
703 goto done;
704 }
705
706 /* Jailed. */
707 if (ia != NULL) {
708 struct ifnet *ifp;
709
710 ifp = ia->ia_ifp;
711 ia = NULL;
712 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
713
714 sa = ifa->ifa_addr;
715 if (sa->sa_family != AF_INET)
716 continue;
717 sin = (struct sockaddr_in *)sa;
718 if (prison_check_ip4(cred,
719 &sin->sin_addr) == 0) {
720 ia = (struct in_ifaddr *)ifa;
721 break;
722 }
723 }
724 if (ia != NULL) {
725 laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
726 goto done;
727 }
728 }
729
730 /* 3. As a last resort return the 'default' jail address. */
731 error = prison_get_ip4(cred, laddr);
732 goto done;
733 }
734
735 done:
736 if (sro.ro_rt != NULL)
737 RTFREE(sro.ro_rt);
738 return (error);
739 }
740
741 /*
742 * Set up for a connect from a socket to the specified address.
743 * On entry, *laddrp and *lportp should contain the current local
744 * address and port for the PCB; these are updated to the values
745 * that should be placed in inp_laddr and inp_lport to complete
746 * the connect.
747 *
748 * On success, *faddrp and *fportp will be set to the remote address
749 * and port. These are not updated in the error case.
750 *
751 * If the operation fails because the connection already exists,
752 * *oinpp will be set to the PCB of that connection so that the
753 * caller can decide to override it. In all other cases, *oinpp
754 * is set to NULL.
755 */
756 int
757 in_pcbconnect_setup(struct inpcb *inp, struct sockaddr *nam,
758 in_addr_t *laddrp, u_short *lportp, in_addr_t *faddrp, u_short *fportp,
759 struct inpcb **oinpp, struct ucred *cred)
760 {
761 struct sockaddr_in *sin = (struct sockaddr_in *)nam;
762 struct in_ifaddr *ia;
763 struct inpcb *oinp;
764 struct in_addr laddr, faddr;
765 u_short lport, fport;
766 int error;
767
768 /*
769 * Because a global state change doesn't actually occur here, a read
770 * lock is sufficient.
771 */
772 INP_INFO_LOCK_ASSERT(inp->inp_pcbinfo);
773 INP_LOCK_ASSERT(inp);
774
775 if (oinpp != NULL)
776 *oinpp = NULL;
777 if (nam->sa_len != sizeof (*sin))
778 return (EINVAL);
779 if (sin->sin_family != AF_INET)
780 return (EAFNOSUPPORT);
781 if (sin->sin_port == 0)
782 return (EADDRNOTAVAIL);
783 laddr.s_addr = *laddrp;
784 lport = *lportp;
785 faddr = sin->sin_addr;
786 fport = sin->sin_port;
787
788 if (!TAILQ_EMPTY(&in_ifaddrhead)) {
789 /*
790 * If the destination address is INADDR_ANY,
791 * use the primary local address.
792 * If the supplied address is INADDR_BROADCAST,
793 * and the primary interface supports broadcast,
794 * choose the broadcast address for that interface.
795 */
796 if (faddr.s_addr == INADDR_ANY) {
797 faddr =
798 IA_SIN(TAILQ_FIRST(&in_ifaddrhead))->sin_addr;
799 if (cred != NULL &&
800 (error = prison_get_ip4(cred, &faddr)) != 0)
801 return (error);
802 } else if (faddr.s_addr == (u_long)INADDR_BROADCAST &&
803 (TAILQ_FIRST(&in_ifaddrhead)->ia_ifp->if_flags &
804 IFF_BROADCAST))
805 faddr = satosin(&TAILQ_FIRST(
806 &in_ifaddrhead)->ia_broadaddr)->sin_addr;
807 }
808 if (laddr.s_addr == INADDR_ANY) {
809 error = in_pcbladdr(inp, &faddr, &laddr, cred);
810 if (error)
811 return (error);
812
813 /*
814 * If the destination address is multicast and an outgoing
815 * interface has been set as a multicast option, use the
816 * address of that interface as our source address.
817 */
818 if (IN_MULTICAST(ntohl(faddr.s_addr)) &&
819 inp->inp_moptions != NULL) {
820 struct ip_moptions *imo;
821 struct ifnet *ifp;
822
823 imo = inp->inp_moptions;
824 if (imo->imo_multicast_ifp != NULL) {
825 ifp = imo->imo_multicast_ifp;
826 TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link)
827 if (ia->ia_ifp == ifp)
828 break;
829 if (ia == NULL)
830 return (EADDRNOTAVAIL);
831 laddr = ia->ia_addr.sin_addr;
832 }
833 }
834 }
835
836 oinp = in_pcblookup_hash(inp->inp_pcbinfo, faddr, fport, laddr, lport,
837 0, NULL);
838 if (oinp != NULL) {
839 if (oinpp != NULL)
840 *oinpp = oinp;
841 return (EADDRINUSE);
842 }
843 if (lport == 0) {
844 error = in_pcbbind_setup(inp, NULL, &laddr.s_addr, &lport,
845 cred);
846 if (error)
847 return (error);
848 }
849 *laddrp = laddr.s_addr;
850 *lportp = lport;
851 *faddrp = faddr.s_addr;
852 *fportp = fport;
853 return (0);
854 }
855
856 void
857 in_pcbdisconnect(struct inpcb *inp)
858 {
859
860 INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
861 INP_WLOCK_ASSERT(inp);
862
863 inp->inp_faddr.s_addr = INADDR_ANY;
864 inp->inp_fport = 0;
865 in_pcbrehash(inp);
866 }
867
868 /*
869 * Historically, in_pcbdetach() included the functionality now found in
870 * in_pcbfree() and in_pcbdrop(). They are now broken out to reflect the
871 * more complex life cycle of TCP.
872 *
873 * in_pcbdetach() is responsibe for disconnecting the socket from an inpcb.
874 * For most protocols, this will be invoked immediately prior to calling
875 * in_pcbfree(). However, for TCP the inpcb may significantly outlive the
876 * socket, in which case in_pcbfree() may be deferred.
877 */
878 void
879 in_pcbdetach(struct inpcb *inp)
880 {
881
882 KASSERT(inp->inp_socket != NULL, ("%s: inp_socket == NULL", __func__));
883
884 inp->inp_socket->so_pcb = NULL;
885 inp->inp_socket = NULL;
886 }
887
888 /*
889 * in_pcbfree() is responsible for freeing an already-detached inpcb, as well
890 * as removing it from any global inpcb lists it might be on.
891 */
892 void
893 in_pcbfree(struct inpcb *inp)
894 {
895 struct inpcbinfo *ipi = inp->inp_pcbinfo;
896
897 KASSERT(inp->inp_socket == NULL, ("%s: inp_socket != NULL", __func__));
898
899 INP_INFO_WLOCK_ASSERT(ipi);
900 INP_WLOCK_ASSERT(inp);
901
902 #ifdef IPSEC
903 if (inp->inp_sp != NULL)
904 ipsec_delete_pcbpolicy(inp);
905 #endif /* IPSEC */
906 inp->inp_gencnt = ++ipi->ipi_gencnt;
907 in_pcbremlists(inp);
908 #ifdef INET6
909 if (inp->inp_vflag & INP_IPV6PROTO) {
910 ip6_freepcbopts(inp->in6p_outputopts);
911 ip6_freemoptions(inp->in6p_moptions);
912 }
913 #endif
914 if (inp->inp_options)
915 (void)m_free(inp->inp_options);
916 if (inp->inp_moptions != NULL)
917 inp_freemoptions(inp->inp_moptions);
918 inp->inp_vflag = 0;
919 crfree(inp->inp_cred);
920
921 #ifdef MAC
922 mac_destroy_inpcb(inp);
923 #endif
924 INP_WUNLOCK(inp);
925 uma_zfree(ipi->ipi_zone, inp);
926 }
927
928 /*
929 * in_pcbdrop() removes an inpcb from hashed lists, releasing its address and
930 * port reservation, and preventing it from being returned by inpcb lookups.
931 *
932 * It is used by TCP to mark an inpcb as unused and avoid future packet
933 * delivery or event notification when a socket remains open but TCP has
934 * closed. This might occur as a result of a shutdown()-initiated TCP close
935 * or a RST on the wire, and allows the port binding to be reused while still
936 * maintaining the invariant that so_pcb always points to a valid inpcb until
937 * in_pcbdetach().
938 *
939 * XXXRW: An inp_lport of 0 is used to indicate that the inpcb is not on hash
940 * lists, but can lead to confusing netstat output, as open sockets with
941 * closed TCP connections will no longer appear to have their bound port
942 * number. An explicit flag would be better, as it would allow us to leave
943 * the port number intact after the connection is dropped.
944 *
945 * XXXRW: Possibly in_pcbdrop() should also prevent future notifications by
946 * in_pcbnotifyall() and in_pcbpurgeif0()?
947 */
948 void
949 in_pcbdrop(struct inpcb *inp)
950 {
951
952 INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
953 INP_WLOCK_ASSERT(inp);
954
955 inp->inp_flags |= INP_DROPPED;
956 if (inp->inp_flags & INP_INHASHLIST) {
957 struct inpcbport *phd = inp->inp_phd;
958
959 LIST_REMOVE(inp, inp_hash);
960 LIST_REMOVE(inp, inp_portlist);
961 if (LIST_FIRST(&phd->phd_pcblist) == NULL) {
962 LIST_REMOVE(phd, phd_hash);
963 free(phd, M_PCB);
964 }
965 inp->inp_flags &= ~INP_INHASHLIST;
966 }
967 }
968
969 /*
970 * Common routines to return the socket addresses associated with inpcbs.
971 */
972 struct sockaddr *
973 in_sockaddr(in_port_t port, struct in_addr *addr_p)
974 {
975 struct sockaddr_in *sin;
976
977 MALLOC(sin, struct sockaddr_in *, sizeof *sin, M_SONAME,
978 M_WAITOK | M_ZERO);
979 sin->sin_family = AF_INET;
980 sin->sin_len = sizeof(*sin);
981 sin->sin_addr = *addr_p;
982 sin->sin_port = port;
983
984 return (struct sockaddr *)sin;
985 }
986
987 int
988 in_getsockaddr(struct socket *so, struct sockaddr **nam)
989 {
990 struct inpcb *inp;
991 struct in_addr addr;
992 in_port_t port;
993
994 inp = sotoinpcb(so);
995 KASSERT(inp != NULL, ("in_getsockaddr: inp == NULL"));
996
997 INP_RLOCK(inp);
998 port = inp->inp_lport;
999 addr = inp->inp_laddr;
1000 INP_RUNLOCK(inp);
1001
1002 *nam = in_sockaddr(port, &addr);
1003 return 0;
1004 }
1005
1006 int
1007 in_getpeeraddr(struct socket *so, struct sockaddr **nam)
1008 {
1009 struct inpcb *inp;
1010 struct in_addr addr;
1011 in_port_t port;
1012
1013 inp = sotoinpcb(so);
1014 KASSERT(inp != NULL, ("in_getpeeraddr: inp == NULL"));
1015
1016 INP_RLOCK(inp);
1017 port = inp->inp_fport;
1018 addr = inp->inp_faddr;
1019 INP_RUNLOCK(inp);
1020
1021 *nam = in_sockaddr(port, &addr);
1022 return 0;
1023 }
1024
1025 void
1026 in_pcbnotifyall(struct inpcbinfo *pcbinfo, struct in_addr faddr, int errno,
1027 struct inpcb *(*notify)(struct inpcb *, int))
1028 {
1029 struct inpcb *inp, *inp_temp;
1030
1031 INP_INFO_WLOCK(pcbinfo);
1032 LIST_FOREACH_SAFE(inp, pcbinfo->ipi_listhead, inp_list, inp_temp) {
1033 INP_WLOCK(inp);
1034 #ifdef INET6
1035 if ((inp->inp_vflag & INP_IPV4) == 0) {
1036 INP_WUNLOCK(inp);
1037 continue;
1038 }
1039 #endif
1040 if (inp->inp_faddr.s_addr != faddr.s_addr ||
1041 inp->inp_socket == NULL) {
1042 INP_WUNLOCK(inp);
1043 continue;
1044 }
1045 if ((*notify)(inp, errno))
1046 INP_WUNLOCK(inp);
1047 }
1048 INP_INFO_WUNLOCK(pcbinfo);
1049 }
1050
1051 void
1052 in_pcbpurgeif0(struct inpcbinfo *pcbinfo, struct ifnet *ifp)
1053 {
1054 struct inpcb *inp;
1055 struct ip_moptions *imo;
1056 int i, gap;
1057
1058 INP_INFO_RLOCK(pcbinfo);
1059 LIST_FOREACH(inp, pcbinfo->ipi_listhead, inp_list) {
1060 INP_WLOCK(inp);
1061 imo = inp->inp_moptions;
1062 if ((inp->inp_vflag & INP_IPV4) &&
1063 imo != NULL) {
1064 /*
1065 * Unselect the outgoing interface if it is being
1066 * detached.
1067 */
1068 if (imo->imo_multicast_ifp == ifp)
1069 imo->imo_multicast_ifp = NULL;
1070
1071 /*
1072 * Drop multicast group membership if we joined
1073 * through the interface being detached.
1074 */
1075 for (i = 0, gap = 0; i < imo->imo_num_memberships;
1076 i++) {
1077 if (imo->imo_membership[i]->inm_ifp == ifp) {
1078 in_delmulti(imo->imo_membership[i]);
1079 gap++;
1080 } else if (gap != 0)
1081 imo->imo_membership[i - gap] =
1082 imo->imo_membership[i];
1083 }
1084 imo->imo_num_memberships -= gap;
1085 }
1086 INP_WUNLOCK(inp);
1087 }
1088 INP_INFO_RUNLOCK(pcbinfo);
1089 }
1090
1091 /*
1092 * Lookup a PCB based on the local address and port.
1093 */
1094 #define INP_LOOKUP_MAPPED_PCB_COST 3
1095 struct inpcb *
1096 in_pcblookup_local(struct inpcbinfo *pcbinfo, struct in_addr laddr,
1097 u_short lport, int wild_okay, struct ucred *cred)
1098 {
1099 struct inpcb *inp;
1100 #ifdef INET6
1101 int matchwild = 3 + INP_LOOKUP_MAPPED_PCB_COST;
1102 #else
1103 int matchwild = 3;
1104 #endif
1105 int wildcard;
1106
1107 INP_INFO_LOCK_ASSERT(pcbinfo);
1108
1109 if (!wild_okay) {
1110 struct inpcbhead *head;
1111 /*
1112 * Look for an unconnected (wildcard foreign addr) PCB that
1113 * matches the local address and port we're looking for.
1114 */
1115 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport,
1116 0, pcbinfo->ipi_hashmask)];
1117 LIST_FOREACH(inp, head, inp_hash) {
1118 #ifdef INET6
1119 /* XXX inp locking */
1120 if ((inp->inp_vflag & INP_IPV4) == 0)
1121 continue;
1122 #endif
1123 if (inp->inp_faddr.s_addr == INADDR_ANY &&
1124 inp->inp_laddr.s_addr == laddr.s_addr &&
1125 inp->inp_lport == lport) {
1126 /*
1127 * Found?
1128 */
1129 if (cred == NULL ||
1130 inp->inp_cred->cr_prison == cred->cr_prison)
1131 return (inp);
1132 }
1133 }
1134 /*
1135 * Not found.
1136 */
1137 return (NULL);
1138 } else {
1139 struct inpcbporthead *porthash;
1140 struct inpcbport *phd;
1141 struct inpcb *match = NULL;
1142 /*
1143 * Best fit PCB lookup.
1144 *
1145 * First see if this local port is in use by looking on the
1146 * port hash list.
1147 */
1148 porthash = &pcbinfo->ipi_porthashbase[INP_PCBPORTHASH(lport,
1149 pcbinfo->ipi_porthashmask)];
1150 LIST_FOREACH(phd, porthash, phd_hash) {
1151 if (phd->phd_port == lport)
1152 break;
1153 }
1154 if (phd != NULL) {
1155 /*
1156 * Port is in use by one or more PCBs. Look for best
1157 * fit.
1158 */
1159 LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) {
1160 wildcard = 0;
1161 if (cred != NULL &&
1162 inp->inp_cred->cr_prison != cred->cr_prison)
1163 continue;
1164 #ifdef INET6
1165 /* XXX inp locking */
1166 if ((inp->inp_vflag & INP_IPV4) == 0)
1167 continue;
1168 /*
1169 * We never select the PCB that has
1170 * INP_IPV6 flag and is bound to :: if
1171 * we have another PCB which is bound
1172 * to 0.0.0.0. If a PCB has the
1173 * INP_IPV6 flag, then we set its cost
1174 * higher than IPv4 only PCBs.
1175 *
1176 * Note that the case only happens
1177 * when a socket is bound to ::, under
1178 * the condition that the use of the
1179 * mapped address is allowed.
1180 */
1181 if ((inp->inp_vflag & INP_IPV6) != 0)
1182 wildcard += INP_LOOKUP_MAPPED_PCB_COST;
1183 #endif
1184 if (inp->inp_faddr.s_addr != INADDR_ANY)
1185 wildcard++;
1186 if (inp->inp_laddr.s_addr != INADDR_ANY) {
1187 if (laddr.s_addr == INADDR_ANY)
1188 wildcard++;
1189 else if (inp->inp_laddr.s_addr != laddr.s_addr)
1190 continue;
1191 } else {
1192 if (laddr.s_addr != INADDR_ANY)
1193 wildcard++;
1194 }
1195 if (wildcard < matchwild) {
1196 match = inp;
1197 matchwild = wildcard;
1198 if (matchwild == 0)
1199 break;
1200 }
1201 }
1202 }
1203 return (match);
1204 }
1205 }
1206 #undef INP_LOOKUP_MAPPED_PCB_COST
1207
1208 /*
1209 * Lookup PCB in hash list.
1210 */
1211 struct inpcb *
1212 in_pcblookup_hash(struct inpcbinfo *pcbinfo, struct in_addr faddr,
1213 u_int fport_arg, struct in_addr laddr, u_int lport_arg, int wildcard,
1214 struct ifnet *ifp)
1215 {
1216 struct inpcbhead *head;
1217 struct inpcb *inp, *tmpinp;
1218 u_short fport = fport_arg, lport = lport_arg;
1219
1220 INP_INFO_LOCK_ASSERT(pcbinfo);
1221
1222 /*
1223 * First look for an exact match.
1224 */
1225 tmpinp = NULL;
1226 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(faddr.s_addr, lport, fport,
1227 pcbinfo->ipi_hashmask)];
1228 LIST_FOREACH(inp, head, inp_hash) {
1229 #ifdef INET6
1230 /* XXX inp locking */
1231 if ((inp->inp_vflag & INP_IPV4) == 0)
1232 continue;
1233 #endif
1234 if (inp->inp_faddr.s_addr == faddr.s_addr &&
1235 inp->inp_laddr.s_addr == laddr.s_addr &&
1236 inp->inp_fport == fport &&
1237 inp->inp_lport == lport) {
1238 /*
1239 * XXX We should be able to directly return
1240 * the inp here, without any checks.
1241 * Well unless both bound with SO_REUSEPORT?
1242 */
1243 if (jailed(inp->inp_cred))
1244 return (inp);
1245 if (tmpinp == NULL)
1246 tmpinp = inp;
1247 }
1248 }
1249 if (tmpinp != NULL)
1250 return (tmpinp);
1251
1252 /*
1253 * Then look for a wildcard match, if requested.
1254 */
1255 if (wildcard == INPLOOKUP_WILDCARD) {
1256 struct inpcb *local_wild = NULL, *local_exact = NULL;
1257 #ifdef INET6
1258 struct inpcb *local_wild_mapped = NULL;
1259 #endif
1260 struct inpcb *jail_wild = NULL;
1261 int injail;
1262
1263 /*
1264 * Order of socket selection - we always prefer jails.
1265 * 1. jailed, non-wild.
1266 * 2. jailed, wild.
1267 * 3. non-jailed, non-wild.
1268 * 4. non-jailed, wild.
1269 */
1270
1271 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport,
1272 0, pcbinfo->ipi_hashmask)];
1273 LIST_FOREACH(inp, head, inp_hash) {
1274 #ifdef INET6
1275 /* XXX inp locking */
1276 if ((inp->inp_vflag & INP_IPV4) == 0)
1277 continue;
1278 #endif
1279 if (inp->inp_faddr.s_addr != INADDR_ANY ||
1280 inp->inp_lport != lport)
1281 continue;
1282
1283 /* XXX inp locking */
1284 if (ifp && ifp->if_type == IFT_FAITH &&
1285 (inp->inp_flags & INP_FAITH) == 0)
1286 continue;
1287
1288 injail = jailed(inp->inp_cred);
1289 if (injail) {
1290 if (prison_check_ip4(inp->inp_cred,
1291 &laddr) != 0)
1292 continue;
1293 } else {
1294 if (local_exact != NULL)
1295 continue;
1296 }
1297
1298 if (inp->inp_laddr.s_addr == laddr.s_addr) {
1299 if (injail)
1300 return (inp);
1301 else
1302 local_exact = inp;
1303 } else if (inp->inp_laddr.s_addr == INADDR_ANY) {
1304 #ifdef INET6
1305 /* XXX inp locking, NULL check */
1306 if (inp->inp_vflag & INP_IPV6PROTO)
1307 local_wild_mapped = inp;
1308 else
1309 #endif /* INET6 */
1310 if (injail)
1311 jail_wild = inp;
1312 else
1313 local_wild = inp;
1314 }
1315 } /* LIST_FOREACH */
1316 if (jail_wild != NULL)
1317 return (jail_wild);
1318 if (local_exact != NULL)
1319 return (local_exact);
1320 if (local_wild != NULL)
1321 return (local_wild);
1322 #ifdef INET6
1323 if (local_wild_mapped != NULL)
1324 return (local_wild_mapped);
1325 #endif /* defined(INET6) */
1326 } /* if (wildcard == INPLOOKUP_WILDCARD) */
1327
1328 return (NULL);
1329 }
1330
1331 /*
1332 * Insert PCB onto various hash lists.
1333 */
1334 int
1335 in_pcbinshash(struct inpcb *inp)
1336 {
1337 struct inpcbhead *pcbhash;
1338 struct inpcbporthead *pcbporthash;
1339 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1340 struct inpcbport *phd;
1341 u_int32_t hashkey_faddr;
1342
1343 INP_INFO_WLOCK_ASSERT(pcbinfo);
1344 INP_WLOCK_ASSERT(inp);
1345 KASSERT((inp->inp_flags & INP_INHASHLIST) == 0,
1346 ("in_pcbinshash: INP_INHASHLIST"));
1347
1348 #ifdef INET6
1349 if (inp->inp_vflag & INP_IPV6)
1350 hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX */;
1351 else
1352 #endif /* INET6 */
1353 hashkey_faddr = inp->inp_faddr.s_addr;
1354
1355 pcbhash = &pcbinfo->ipi_hashbase[INP_PCBHASH(hashkey_faddr,
1356 inp->inp_lport, inp->inp_fport, pcbinfo->ipi_hashmask)];
1357
1358 pcbporthash = &pcbinfo->ipi_porthashbase[
1359 INP_PCBPORTHASH(inp->inp_lport, pcbinfo->ipi_porthashmask)];
1360
1361 /*
1362 * Go through port list and look for a head for this lport.
1363 */
1364 LIST_FOREACH(phd, pcbporthash, phd_hash) {
1365 if (phd->phd_port == inp->inp_lport)
1366 break;
1367 }
1368 /*
1369 * If none exists, malloc one and tack it on.
1370 */
1371 if (phd == NULL) {
1372 MALLOC(phd, struct inpcbport *, sizeof(struct inpcbport), M_PCB, M_NOWAIT);
1373 if (phd == NULL) {
1374 return (ENOBUFS); /* XXX */
1375 }
1376 phd->phd_port = inp->inp_lport;
1377 LIST_INIT(&phd->phd_pcblist);
1378 LIST_INSERT_HEAD(pcbporthash, phd, phd_hash);
1379 }
1380 inp->inp_phd = phd;
1381 LIST_INSERT_HEAD(&phd->phd_pcblist, inp, inp_portlist);
1382 LIST_INSERT_HEAD(pcbhash, inp, inp_hash);
1383 inp->inp_flags |= INP_INHASHLIST;
1384 return (0);
1385 }
1386
1387 /*
1388 * Move PCB to the proper hash bucket when { faddr, fport } have been
1389 * changed. NOTE: This does not handle the case of the lport changing (the
1390 * hashed port list would have to be updated as well), so the lport must
1391 * not change after in_pcbinshash() has been called.
1392 */
1393 void
1394 in_pcbrehash(struct inpcb *inp)
1395 {
1396 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1397 struct inpcbhead *head;
1398 u_int32_t hashkey_faddr;
1399
1400 INP_INFO_WLOCK_ASSERT(pcbinfo);
1401 INP_WLOCK_ASSERT(inp);
1402 KASSERT(inp->inp_flags & INP_INHASHLIST,
1403 ("in_pcbrehash: !INP_INHASHLIST"));
1404
1405 #ifdef INET6
1406 if (inp->inp_vflag & INP_IPV6)
1407 hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX */;
1408 else
1409 #endif /* INET6 */
1410 hashkey_faddr = inp->inp_faddr.s_addr;
1411
1412 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(hashkey_faddr,
1413 inp->inp_lport, inp->inp_fport, pcbinfo->ipi_hashmask)];
1414
1415 LIST_REMOVE(inp, inp_hash);
1416 LIST_INSERT_HEAD(head, inp, inp_hash);
1417 }
1418
1419 /*
1420 * Remove PCB from various lists.
1421 */
1422 void
1423 in_pcbremlists(struct inpcb *inp)
1424 {
1425 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1426
1427 INP_INFO_WLOCK_ASSERT(pcbinfo);
1428 INP_WLOCK_ASSERT(inp);
1429
1430 inp->inp_gencnt = ++pcbinfo->ipi_gencnt;
1431 if (inp->inp_flags & INP_INHASHLIST) {
1432 struct inpcbport *phd = inp->inp_phd;
1433
1434 LIST_REMOVE(inp, inp_hash);
1435 LIST_REMOVE(inp, inp_portlist);
1436 if (LIST_FIRST(&phd->phd_pcblist) == NULL) {
1437 LIST_REMOVE(phd, phd_hash);
1438 free(phd, M_PCB);
1439 }
1440 inp->inp_flags &= ~INP_INHASHLIST;
1441 }
1442 LIST_REMOVE(inp, inp_list);
1443 pcbinfo->ipi_count--;
1444 }
1445
1446 /*
1447 * A set label operation has occurred at the socket layer, propagate the
1448 * label change into the in_pcb for the socket.
1449 */
1450 void
1451 in_pcbsosetlabel(struct socket *so)
1452 {
1453 #ifdef MAC
1454 struct inpcb *inp;
1455
1456 inp = sotoinpcb(so);
1457 KASSERT(inp != NULL, ("in_pcbsosetlabel: so->so_pcb == NULL"));
1458
1459 INP_WLOCK(inp);
1460 SOCK_LOCK(so);
1461 mac_inpcb_sosetlabel(so, inp);
1462 SOCK_UNLOCK(so);
1463 INP_WUNLOCK(inp);
1464 #endif
1465 }
1466
1467 /*
1468 * ipport_tick runs once per second, determining if random port allocation
1469 * should be continued. If more than ipport_randomcps ports have been
1470 * allocated in the last second, then we return to sequential port
1471 * allocation. We return to random allocation only once we drop below
1472 * ipport_randomcps for at least ipport_randomtime seconds.
1473 */
1474 void
1475 ipport_tick(void *xtp)
1476 {
1477
1478 if (ipport_tcpallocs <= ipport_tcplastcount + ipport_randomcps) {
1479 if (ipport_stoprandom > 0)
1480 ipport_stoprandom--;
1481 } else
1482 ipport_stoprandom = ipport_randomtime;
1483 ipport_tcplastcount = ipport_tcpallocs;
1484 callout_reset(&ipport_tick_callout, hz, ipport_tick, NULL);
1485 }
1486
1487 void
1488 inp_apply_all(void (*func)(struct inpcb *, void *), void *arg)
1489 {
1490 struct inpcb *inp;
1491
1492 INP_INFO_RLOCK(&tcbinfo);
1493 LIST_FOREACH(inp, tcbinfo.ipi_listhead, inp_list) {
1494 INP_WLOCK(inp);
1495 func(inp, arg);
1496 INP_WUNLOCK(inp);
1497 }
1498 INP_INFO_RUNLOCK(&tcbinfo);
1499 }
1500
1501 struct socket *
1502 inp_inpcbtosocket(struct inpcb *inp)
1503 {
1504
1505 INP_WLOCK_ASSERT(inp);
1506 return (inp->inp_socket);
1507 }
1508
1509 struct tcpcb *
1510 inp_inpcbtotcpcb(struct inpcb *inp)
1511 {
1512
1513 INP_WLOCK_ASSERT(inp);
1514 return ((struct tcpcb *)inp->inp_ppcb);
1515 }
1516
1517 int
1518 inp_ip_tos_get(const struct inpcb *inp)
1519 {
1520
1521 return (inp->inp_ip_tos);
1522 }
1523
1524 void
1525 inp_ip_tos_set(struct inpcb *inp, int val)
1526 {
1527
1528 inp->inp_ip_tos = val;
1529 }
1530
1531 void
1532 inp_4tuple_get(struct inpcb *inp, uint32_t *laddr, uint16_t *lp,
1533 uint32_t *faddr, uint16_t *fp)
1534 {
1535
1536 INP_LOCK_ASSERT(inp);
1537 *laddr = inp->inp_laddr.s_addr;
1538 *faddr = inp->inp_faddr.s_addr;
1539 *lp = inp->inp_lport;
1540 *fp = inp->inp_fport;
1541 }
1542
1543 struct inpcb *
1544 so_sotoinpcb(struct socket *so)
1545 {
1546
1547 return (sotoinpcb(so));
1548 }
1549
1550 struct tcpcb *
1551 so_sototcpcb(struct socket *so)
1552 {
1553
1554 return (sototcpcb(so));
1555 }
1556
1557 void
1558 inp_wlock(struct inpcb *inp)
1559 {
1560
1561 INP_WLOCK(inp);
1562 }
1563
1564 void
1565 inp_wunlock(struct inpcb *inp)
1566 {
1567
1568 INP_WUNLOCK(inp);
1569 }
1570
1571 void
1572 inp_rlock(struct inpcb *inp)
1573 {
1574
1575 INP_RLOCK(inp);
1576 }
1577
1578 void
1579 inp_runlock(struct inpcb *inp)
1580 {
1581
1582 INP_RUNLOCK(inp);
1583 }
1584
1585 #ifdef INVARIANTS
1586 void
1587 inp_wlock_assert(struct inpcb *inp)
1588 {
1589
1590 INP_WLOCK_ASSERT(inp);
1591 }
1592
1593 void
1594 inp_rlock_assert(struct inpcb *inp)
1595 {
1596
1597 INP_RLOCK_ASSERT(inp);
1598 }
1599
1600 void
1601 inp_lock_assert(struct inpcb *inp)
1602 {
1603
1604 INP_LOCK_ASSERT(inp);
1605 }
1606
1607 void
1608 inp_unlock_assert(struct inpcb *inp)
1609 {
1610
1611 INP_UNLOCK_ASSERT(inp);
1612 }
1613
1614 #endif
1615
1616 #ifdef DDB
1617 static void
1618 db_print_indent(int indent)
1619 {
1620 int i;
1621
1622 for (i = 0; i < indent; i++)
1623 db_printf(" ");
1624 }
1625
1626 static void
1627 db_print_inconninfo(struct in_conninfo *inc, const char *name, int indent)
1628 {
1629 char faddr_str[48], laddr_str[48];
1630
1631 db_print_indent(indent);
1632 db_printf("%s at %p\n", name, inc);
1633
1634 indent += 2;
1635
1636 #ifdef INET6
1637 if (inc->inc_flags & INC_ISIPV6) {
1638 /* IPv6. */
1639 ip6_sprintf(laddr_str, &inc->inc6_laddr);
1640 ip6_sprintf(faddr_str, &inc->inc6_faddr);
1641 } else {
1642 #endif
1643 /* IPv4. */
1644 inet_ntoa_r(inc->inc_laddr, laddr_str);
1645 inet_ntoa_r(inc->inc_faddr, faddr_str);
1646 #ifdef INET6
1647 }
1648 #endif
1649 db_print_indent(indent);
1650 db_printf("inc_laddr %s inc_lport %u\n", laddr_str,
1651 ntohs(inc->inc_lport));
1652 db_print_indent(indent);
1653 db_printf("inc_faddr %s inc_fport %u\n", faddr_str,
1654 ntohs(inc->inc_fport));
1655 }
1656
1657 static void
1658 db_print_inpflags(int inp_flags)
1659 {
1660 int comma;
1661
1662 comma = 0;
1663 if (inp_flags & INP_RECVOPTS) {
1664 db_printf("%sINP_RECVOPTS", comma ? ", " : "");
1665 comma = 1;
1666 }
1667 if (inp_flags & INP_RECVRETOPTS) {
1668 db_printf("%sINP_RECVRETOPTS", comma ? ", " : "");
1669 comma = 1;
1670 }
1671 if (inp_flags & INP_RECVDSTADDR) {
1672 db_printf("%sINP_RECVDSTADDR", comma ? ", " : "");
1673 comma = 1;
1674 }
1675 if (inp_flags & INP_HDRINCL) {
1676 db_printf("%sINP_HDRINCL", comma ? ", " : "");
1677 comma = 1;
1678 }
1679 if (inp_flags & INP_HIGHPORT) {
1680 db_printf("%sINP_HIGHPORT", comma ? ", " : "");
1681 comma = 1;
1682 }
1683 if (inp_flags & INP_LOWPORT) {
1684 db_printf("%sINP_LOWPORT", comma ? ", " : "");
1685 comma = 1;
1686 }
1687 if (inp_flags & INP_ANONPORT) {
1688 db_printf("%sINP_ANONPORT", comma ? ", " : "");
1689 comma = 1;
1690 }
1691 if (inp_flags & INP_RECVIF) {
1692 db_printf("%sINP_RECVIF", comma ? ", " : "");
1693 comma = 1;
1694 }
1695 if (inp_flags & INP_MTUDISC) {
1696 db_printf("%sINP_MTUDISC", comma ? ", " : "");
1697 comma = 1;
1698 }
1699 if (inp_flags & INP_FAITH) {
1700 db_printf("%sINP_FAITH", comma ? ", " : "");
1701 comma = 1;
1702 }
1703 if (inp_flags & INP_RECVTTL) {
1704 db_printf("%sINP_RECVTTL", comma ? ", " : "");
1705 comma = 1;
1706 }
1707 if (inp_flags & INP_DONTFRAG) {
1708 db_printf("%sINP_DONTFRAG", comma ? ", " : "");
1709 comma = 1;
1710 }
1711 if (inp_flags & IN6P_IPV6_V6ONLY) {
1712 db_printf("%sIN6P_IPV6_V6ONLY", comma ? ", " : "");
1713 comma = 1;
1714 }
1715 if (inp_flags & IN6P_PKTINFO) {
1716 db_printf("%sIN6P_PKTINFO", comma ? ", " : "");
1717 comma = 1;
1718 }
1719 if (inp_flags & IN6P_HOPLIMIT) {
1720 db_printf("%sIN6P_HOPLIMIT", comma ? ", " : "");
1721 comma = 1;
1722 }
1723 if (inp_flags & IN6P_HOPOPTS) {
1724 db_printf("%sIN6P_HOPOPTS", comma ? ", " : "");
1725 comma = 1;
1726 }
1727 if (inp_flags & IN6P_DSTOPTS) {
1728 db_printf("%sIN6P_DSTOPTS", comma ? ", " : "");
1729 comma = 1;
1730 }
1731 if (inp_flags & IN6P_RTHDR) {
1732 db_printf("%sIN6P_RTHDR", comma ? ", " : "");
1733 comma = 1;
1734 }
1735 if (inp_flags & IN6P_RTHDRDSTOPTS) {
1736 db_printf("%sIN6P_RTHDRDSTOPTS", comma ? ", " : "");
1737 comma = 1;
1738 }
1739 if (inp_flags & IN6P_TCLASS) {
1740 db_printf("%sIN6P_TCLASS", comma ? ", " : "");
1741 comma = 1;
1742 }
1743 if (inp_flags & IN6P_AUTOFLOWLABEL) {
1744 db_printf("%sIN6P_AUTOFLOWLABEL", comma ? ", " : "");
1745 comma = 1;
1746 }
1747 if (inp_flags & INP_TIMEWAIT) {
1748 db_printf("%sINP_TIMEWAIT", comma ? ", " : "");
1749 comma = 1;
1750 }
1751 if (inp_flags & INP_ONESBCAST) {
1752 db_printf("%sINP_ONESBCAST", comma ? ", " : "");
1753 comma = 1;
1754 }
1755 if (inp_flags & INP_DROPPED) {
1756 db_printf("%sINP_DROPPED", comma ? ", " : "");
1757 comma = 1;
1758 }
1759 if (inp_flags & INP_SOCKREF) {
1760 db_printf("%sINP_SOCKREF", comma ? ", " : "");
1761 comma = 1;
1762 }
1763 if (inp_flags & IN6P_RFC2292) {
1764 db_printf("%sIN6P_RFC2292", comma ? ", " : "");
1765 comma = 1;
1766 }
1767 if (inp_flags & IN6P_MTU) {
1768 db_printf("IN6P_MTU%s", comma ? ", " : "");
1769 comma = 1;
1770 }
1771 }
1772
1773 static void
1774 db_print_inpvflag(u_char inp_vflag)
1775 {
1776 int comma;
1777
1778 comma = 0;
1779 if (inp_vflag & INP_IPV4) {
1780 db_printf("%sINP_IPV4", comma ? ", " : "");
1781 comma = 1;
1782 }
1783 if (inp_vflag & INP_IPV6) {
1784 db_printf("%sINP_IPV6", comma ? ", " : "");
1785 comma = 1;
1786 }
1787 if (inp_vflag & INP_IPV6PROTO) {
1788 db_printf("%sINP_IPV6PROTO", comma ? ", " : "");
1789 comma = 1;
1790 }
1791 }
1792
1793 void
1794 db_print_inpcb(struct inpcb *inp, const char *name, int indent)
1795 {
1796
1797 db_print_indent(indent);
1798 db_printf("%s at %p\n", name, inp);
1799
1800 indent += 2;
1801
1802 db_print_indent(indent);
1803 db_printf("inp_flow: 0x%x\n", inp->inp_flow);
1804
1805 db_print_inconninfo(&inp->inp_inc, "inp_conninfo", indent);
1806
1807 db_print_indent(indent);
1808 db_printf("inp_ppcb: %p inp_pcbinfo: %p inp_socket: %p\n",
1809 inp->inp_ppcb, inp->inp_pcbinfo, inp->inp_socket);
1810
1811 db_print_indent(indent);
1812 db_printf("inp_label: %p inp_flags: 0x%x (",
1813 inp->inp_label, inp->inp_flags);
1814 db_print_inpflags(inp->inp_flags);
1815 db_printf(")\n");
1816
1817 db_print_indent(indent);
1818 db_printf("inp_sp: %p inp_vflag: 0x%x (", inp->inp_sp,
1819 inp->inp_vflag);
1820 db_print_inpvflag(inp->inp_vflag);
1821 db_printf(")\n");
1822
1823 db_print_indent(indent);
1824 db_printf("inp_ip_ttl: %d inp_ip_p: %d inp_ip_minttl: %d\n",
1825 inp->inp_ip_ttl, inp->inp_ip_p, inp->inp_ip_minttl);
1826
1827 db_print_indent(indent);
1828 #ifdef INET6
1829 if (inp->inp_vflag & INP_IPV6) {
1830 db_printf("in6p_options: %p in6p_outputopts: %p "
1831 "in6p_moptions: %p\n", inp->in6p_options,
1832 inp->in6p_outputopts, inp->in6p_moptions);
1833 db_printf("in6p_icmp6filt: %p in6p_cksum %d "
1834 "in6p_hops %u\n", inp->in6p_icmp6filt, inp->in6p_cksum,
1835 inp->in6p_hops);
1836 } else
1837 #endif
1838 {
1839 db_printf("inp_ip_tos: %d inp_ip_options: %p "
1840 "inp_ip_moptions: %p\n", inp->inp_ip_tos,
1841 inp->inp_options, inp->inp_moptions);
1842 }
1843
1844 db_print_indent(indent);
1845 db_printf("inp_phd: %p inp_gencnt: %ju\n", inp->inp_phd,
1846 (uintmax_t)inp->inp_gencnt);
1847 }
1848
1849 DB_SHOW_COMMAND(inpcb, db_show_inpcb)
1850 {
1851 struct inpcb *inp;
1852
1853 if (!have_addr) {
1854 db_printf("usage: show inpcb <addr>\n");
1855 return;
1856 }
1857 inp = (struct inpcb *)addr;
1858
1859 db_print_inpcb(inp, "inpcb", 0);
1860 }
1861 #endif
Cache object: 73a6d9c25c716a826aa3d4fd30c87e15
|