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/8.0/sys/netinet/in_pcb.c 197229 2009-09-15 19:38:29Z qingli $");
36
37 #include "opt_ddb.h"
38 #include "opt_ipsec.h"
39 #include "opt_inet6.h"
40
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/malloc.h>
44 #include <sys/mbuf.h>
45 #include <sys/domain.h>
46 #include <sys/protosw.h>
47 #include <sys/socket.h>
48 #include <sys/socketvar.h>
49 #include <sys/priv.h>
50 #include <sys/proc.h>
51 #include <sys/jail.h>
52 #include <sys/kernel.h>
53 #include <sys/sysctl.h>
54
55 #ifdef DDB
56 #include <ddb/ddb.h>
57 #endif
58
59 #include <vm/uma.h>
60
61 #include <net/if.h>
62 #include <net/if_types.h>
63 #include <net/route.h>
64 #include <net/vnet.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 VNET_DEFINE(int, ipport_lowfirstauto) = IPPORT_RESERVED - 1; /* 1023 */
91 VNET_DEFINE(int, ipport_lowlastauto) = IPPORT_RESERVEDSTART; /* 600 */
92 VNET_DEFINE(int, ipport_firstauto) = IPPORT_EPHEMERALFIRST; /* 10000 */
93 VNET_DEFINE(int, ipport_lastauto) = IPPORT_EPHEMERALLAST; /* 65535 */
94 VNET_DEFINE(int, ipport_hifirstauto) = IPPORT_HIFIRSTAUTO; /* 49152 */
95 VNET_DEFINE(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 VNET_DEFINE(int, ipport_reservedhigh) = IPPORT_RESERVED - 1; /* 1023 */
103 VNET_DEFINE(int, ipport_reservedlow);
104
105 /* Variables dealing with random ephemeral port allocation. */
106 VNET_DEFINE(int, ipport_randomized) = 1; /* user controlled via sysctl */
107 VNET_DEFINE(int, ipport_randomcps) = 10; /* user controlled via sysctl */
108 VNET_DEFINE(int, ipport_randomtime) = 45; /* user controlled via sysctl */
109 VNET_DEFINE(int, ipport_stoprandom); /* toggled by ipport_tick */
110 VNET_DEFINE(int, ipport_tcpallocs);
111 static VNET_DEFINE(int, ipport_tcplastcount);
112
113 #define V_ipport_tcplastcount VNET(ipport_tcplastcount)
114
115 #define RANGECHK(var, min, max) \
116 if ((var) < (min)) { (var) = (min); } \
117 else if ((var) > (max)) { (var) = (max); }
118
119 static void in_pcbremlists(struct inpcb *inp);
120
121 static int
122 sysctl_net_ipport_check(SYSCTL_HANDLER_ARGS)
123 {
124 int error;
125
126 #ifdef VIMAGE
127 error = vnet_sysctl_handle_int(oidp, arg1, arg2, req);
128 #else
129 error = sysctl_handle_int(oidp, arg1, arg2, req);
130 #endif
131 if (error == 0) {
132 RANGECHK(V_ipport_lowfirstauto, 1, IPPORT_RESERVED - 1);
133 RANGECHK(V_ipport_lowlastauto, 1, IPPORT_RESERVED - 1);
134 RANGECHK(V_ipport_firstauto, IPPORT_RESERVED, IPPORT_MAX);
135 RANGECHK(V_ipport_lastauto, IPPORT_RESERVED, IPPORT_MAX);
136 RANGECHK(V_ipport_hifirstauto, IPPORT_RESERVED, IPPORT_MAX);
137 RANGECHK(V_ipport_hilastauto, IPPORT_RESERVED, IPPORT_MAX);
138 }
139 return (error);
140 }
141
142 #undef RANGECHK
143
144 SYSCTL_NODE(_net_inet_ip, IPPROTO_IP, portrange, CTLFLAG_RW, 0, "IP Ports");
145
146 SYSCTL_VNET_PROC(_net_inet_ip_portrange, OID_AUTO, lowfirst,
147 CTLTYPE_INT|CTLFLAG_RW, &VNET_NAME(ipport_lowfirstauto), 0,
148 &sysctl_net_ipport_check, "I", "");
149 SYSCTL_VNET_PROC(_net_inet_ip_portrange, OID_AUTO, lowlast,
150 CTLTYPE_INT|CTLFLAG_RW, &VNET_NAME(ipport_lowlastauto), 0,
151 &sysctl_net_ipport_check, "I", "");
152 SYSCTL_VNET_PROC(_net_inet_ip_portrange, OID_AUTO, first,
153 CTLTYPE_INT|CTLFLAG_RW, &VNET_NAME(ipport_firstauto), 0,
154 &sysctl_net_ipport_check, "I", "");
155 SYSCTL_VNET_PROC(_net_inet_ip_portrange, OID_AUTO, last,
156 CTLTYPE_INT|CTLFLAG_RW, &VNET_NAME(ipport_lastauto), 0,
157 &sysctl_net_ipport_check, "I", "");
158 SYSCTL_VNET_PROC(_net_inet_ip_portrange, OID_AUTO, hifirst,
159 CTLTYPE_INT|CTLFLAG_RW, &VNET_NAME(ipport_hifirstauto), 0,
160 &sysctl_net_ipport_check, "I", "");
161 SYSCTL_VNET_PROC(_net_inet_ip_portrange, OID_AUTO, hilast,
162 CTLTYPE_INT|CTLFLAG_RW, &VNET_NAME(ipport_hilastauto), 0,
163 &sysctl_net_ipport_check, "I", "");
164 SYSCTL_VNET_INT(_net_inet_ip_portrange, OID_AUTO, reservedhigh,
165 CTLFLAG_RW|CTLFLAG_SECURE, &VNET_NAME(ipport_reservedhigh), 0, "");
166 SYSCTL_VNET_INT(_net_inet_ip_portrange, OID_AUTO, reservedlow,
167 CTLFLAG_RW|CTLFLAG_SECURE, &VNET_NAME(ipport_reservedlow), 0, "");
168 SYSCTL_VNET_INT(_net_inet_ip_portrange, OID_AUTO, randomized, CTLFLAG_RW,
169 &VNET_NAME(ipport_randomized), 0, "Enable random port allocation");
170 SYSCTL_VNET_INT(_net_inet_ip_portrange, OID_AUTO, randomcps, CTLFLAG_RW,
171 &VNET_NAME(ipport_randomcps), 0, "Maximum number of random port "
172 "allocations before switching to a sequental one");
173 SYSCTL_VNET_INT(_net_inet_ip_portrange, OID_AUTO, randomtime, CTLFLAG_RW,
174 &VNET_NAME(ipport_randomtime), 0,
175 "Minimum time to keep sequental port "
176 "allocation before switching to a random one");
177
178 /*
179 * in_pcb.c: manage the Protocol Control Blocks.
180 *
181 * NOTE: It is assumed that most of these functions will be called with
182 * the pcbinfo lock held, and often, the inpcb lock held, as these utility
183 * functions often modify hash chains or addresses in pcbs.
184 */
185
186 /*
187 * Allocate a PCB and associate it with the socket.
188 * On success return with the PCB locked.
189 */
190 int
191 in_pcballoc(struct socket *so, struct inpcbinfo *pcbinfo)
192 {
193 struct inpcb *inp;
194 int error;
195
196 INP_INFO_WLOCK_ASSERT(pcbinfo);
197 error = 0;
198 inp = uma_zalloc(pcbinfo->ipi_zone, M_NOWAIT);
199 if (inp == NULL)
200 return (ENOBUFS);
201 bzero(inp, inp_zero_size);
202 inp->inp_pcbinfo = pcbinfo;
203 inp->inp_socket = so;
204 inp->inp_cred = crhold(so->so_cred);
205 inp->inp_inc.inc_fibnum = so->so_fibnum;
206 #ifdef MAC
207 error = mac_inpcb_init(inp, M_NOWAIT);
208 if (error != 0)
209 goto out;
210 mac_inpcb_create(so, inp);
211 #endif
212 #ifdef IPSEC
213 error = ipsec_init_policy(so, &inp->inp_sp);
214 if (error != 0) {
215 #ifdef MAC
216 mac_inpcb_destroy(inp);
217 #endif
218 goto out;
219 }
220 #endif /*IPSEC*/
221 #ifdef INET6
222 if (INP_SOCKAF(so) == AF_INET6) {
223 inp->inp_vflag |= INP_IPV6PROTO;
224 if (V_ip6_v6only)
225 inp->inp_flags |= IN6P_IPV6_V6ONLY;
226 }
227 #endif
228 LIST_INSERT_HEAD(pcbinfo->ipi_listhead, inp, inp_list);
229 pcbinfo->ipi_count++;
230 so->so_pcb = (caddr_t)inp;
231 #ifdef INET6
232 if (V_ip6_auto_flowlabel)
233 inp->inp_flags |= IN6P_AUTOFLOWLABEL;
234 #endif
235 INP_WLOCK(inp);
236 inp->inp_gencnt = ++pcbinfo->ipi_gencnt;
237 inp->inp_refcount = 1; /* Reference from the inpcbinfo */
238 #if defined(IPSEC) || defined(MAC)
239 out:
240 if (error != 0) {
241 crfree(inp->inp_cred);
242 uma_zfree(pcbinfo->ipi_zone, inp);
243 }
244 #endif
245 return (error);
246 }
247
248 int
249 in_pcbbind(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred)
250 {
251 int anonport, error;
252
253 INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
254 INP_WLOCK_ASSERT(inp);
255
256 if (inp->inp_lport != 0 || inp->inp_laddr.s_addr != INADDR_ANY)
257 return (EINVAL);
258 anonport = inp->inp_lport == 0 && (nam == NULL ||
259 ((struct sockaddr_in *)nam)->sin_port == 0);
260 error = in_pcbbind_setup(inp, nam, &inp->inp_laddr.s_addr,
261 &inp->inp_lport, cred);
262 if (error)
263 return (error);
264 if (in_pcbinshash(inp) != 0) {
265 inp->inp_laddr.s_addr = INADDR_ANY;
266 inp->inp_lport = 0;
267 return (EAGAIN);
268 }
269 if (anonport)
270 inp->inp_flags |= INP_ANONPORT;
271 return (0);
272 }
273
274 /*
275 * Set up a bind operation on a PCB, performing port allocation
276 * as required, but do not actually modify the PCB. Callers can
277 * either complete the bind by setting inp_laddr/inp_lport and
278 * calling in_pcbinshash(), or they can just use the resulting
279 * port and address to authorise the sending of a once-off packet.
280 *
281 * On error, the values of *laddrp and *lportp are not changed.
282 */
283 int
284 in_pcbbind_setup(struct inpcb *inp, struct sockaddr *nam, in_addr_t *laddrp,
285 u_short *lportp, struct ucred *cred)
286 {
287 struct socket *so = inp->inp_socket;
288 unsigned short *lastport;
289 struct sockaddr_in *sin;
290 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
291 struct in_addr laddr;
292 u_short lport = 0;
293 int wild = 0, reuseport = (so->so_options & SO_REUSEPORT);
294 int error;
295 int dorandom;
296
297 /*
298 * Because no actual state changes occur here, a global write lock on
299 * the pcbinfo isn't required.
300 */
301 INP_INFO_LOCK_ASSERT(pcbinfo);
302 INP_LOCK_ASSERT(inp);
303
304 if (TAILQ_EMPTY(&V_in_ifaddrhead)) /* XXX broken! */
305 return (EADDRNOTAVAIL);
306 laddr.s_addr = *laddrp;
307 if (nam != NULL && laddr.s_addr != INADDR_ANY)
308 return (EINVAL);
309 if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0)
310 wild = INPLOOKUP_WILDCARD;
311 if (nam == NULL) {
312 if ((error = prison_local_ip4(cred, &laddr)) != 0)
313 return (error);
314 } else {
315 sin = (struct sockaddr_in *)nam;
316 if (nam->sa_len != sizeof (*sin))
317 return (EINVAL);
318 #ifdef notdef
319 /*
320 * We should check the family, but old programs
321 * incorrectly fail to initialize it.
322 */
323 if (sin->sin_family != AF_INET)
324 return (EAFNOSUPPORT);
325 #endif
326 error = prison_local_ip4(cred, &sin->sin_addr);
327 if (error)
328 return (error);
329 if (sin->sin_port != *lportp) {
330 /* Don't allow the port to change. */
331 if (*lportp != 0)
332 return (EINVAL);
333 lport = sin->sin_port;
334 }
335 /* NB: lport is left as 0 if the port isn't being changed. */
336 if (IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) {
337 /*
338 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
339 * allow complete duplication of binding if
340 * SO_REUSEPORT is set, or if SO_REUSEADDR is set
341 * and a multicast address is bound on both
342 * new and duplicated sockets.
343 */
344 if (so->so_options & SO_REUSEADDR)
345 reuseport = SO_REUSEADDR|SO_REUSEPORT;
346 } else if (sin->sin_addr.s_addr != INADDR_ANY) {
347 sin->sin_port = 0; /* yech... */
348 bzero(&sin->sin_zero, sizeof(sin->sin_zero));
349 /*
350 * Is the address a local IP address?
351 * If INP_BINDANY is set, then the socket may be bound
352 * to any endpoint address, local or not.
353 */
354 if ((inp->inp_flags & INP_BINDANY) == 0 &&
355 ifa_ifwithaddr_check((struct sockaddr *)sin) == 0)
356 return (EADDRNOTAVAIL);
357 }
358 laddr = sin->sin_addr;
359 if (lport) {
360 struct inpcb *t;
361 struct tcptw *tw;
362
363 /* GROSS */
364 if (ntohs(lport) <= V_ipport_reservedhigh &&
365 ntohs(lport) >= V_ipport_reservedlow &&
366 priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT,
367 0))
368 return (EACCES);
369 if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)) &&
370 priv_check_cred(inp->inp_cred,
371 PRIV_NETINET_REUSEPORT, 0) != 0) {
372 t = in_pcblookup_local(pcbinfo, sin->sin_addr,
373 lport, INPLOOKUP_WILDCARD, cred);
374 /*
375 * XXX
376 * This entire block sorely needs a rewrite.
377 */
378 if (t &&
379 ((t->inp_flags & INP_TIMEWAIT) == 0) &&
380 (so->so_type != SOCK_STREAM ||
381 ntohl(t->inp_faddr.s_addr) == INADDR_ANY) &&
382 (ntohl(sin->sin_addr.s_addr) != INADDR_ANY ||
383 ntohl(t->inp_laddr.s_addr) != INADDR_ANY ||
384 (t->inp_socket->so_options &
385 SO_REUSEPORT) == 0) &&
386 (inp->inp_cred->cr_uid !=
387 t->inp_cred->cr_uid))
388 return (EADDRINUSE);
389 }
390 t = in_pcblookup_local(pcbinfo, sin->sin_addr,
391 lport, wild, cred);
392 if (t && (t->inp_flags & INP_TIMEWAIT)) {
393 /*
394 * XXXRW: If an incpb has had its timewait
395 * state recycled, we treat the address as
396 * being in use (for now). This is better
397 * than a panic, but not desirable.
398 */
399 tw = intotw(inp);
400 if (tw == NULL ||
401 (reuseport & tw->tw_so_options) == 0)
402 return (EADDRINUSE);
403 } else if (t &&
404 (reuseport & t->inp_socket->so_options) == 0) {
405 #ifdef INET6
406 if (ntohl(sin->sin_addr.s_addr) !=
407 INADDR_ANY ||
408 ntohl(t->inp_laddr.s_addr) !=
409 INADDR_ANY ||
410 INP_SOCKAF(so) ==
411 INP_SOCKAF(t->inp_socket))
412 #endif
413 return (EADDRINUSE);
414 }
415 }
416 }
417 if (*lportp != 0)
418 lport = *lportp;
419 if (lport == 0) {
420 u_short first, last, aux;
421 int count;
422
423 if (inp->inp_flags & INP_HIGHPORT) {
424 first = V_ipport_hifirstauto; /* sysctl */
425 last = V_ipport_hilastauto;
426 lastport = &pcbinfo->ipi_lasthi;
427 } else if (inp->inp_flags & INP_LOWPORT) {
428 error = priv_check_cred(cred,
429 PRIV_NETINET_RESERVEDPORT, 0);
430 if (error)
431 return error;
432 first = V_ipport_lowfirstauto; /* 1023 */
433 last = V_ipport_lowlastauto; /* 600 */
434 lastport = &pcbinfo->ipi_lastlow;
435 } else {
436 first = V_ipport_firstauto; /* sysctl */
437 last = V_ipport_lastauto;
438 lastport = &pcbinfo->ipi_lastport;
439 }
440 /*
441 * For UDP, use random port allocation as long as the user
442 * allows it. For TCP (and as of yet unknown) connections,
443 * use random port allocation only if the user allows it AND
444 * ipport_tick() allows it.
445 */
446 if (V_ipport_randomized &&
447 (!V_ipport_stoprandom || pcbinfo == &V_udbinfo))
448 dorandom = 1;
449 else
450 dorandom = 0;
451 /*
452 * It makes no sense to do random port allocation if
453 * we have the only port available.
454 */
455 if (first == last)
456 dorandom = 0;
457 /* Make sure to not include UDP packets in the count. */
458 if (pcbinfo != &V_udbinfo)
459 V_ipport_tcpallocs++;
460 /*
461 * Instead of having two loops further down counting up or down
462 * make sure that first is always <= last and go with only one
463 * code path implementing all logic.
464 */
465 if (first > last) {
466 aux = first;
467 first = last;
468 last = aux;
469 }
470
471 if (dorandom)
472 *lastport = first +
473 (arc4random() % (last - first));
474
475 count = last - first;
476
477 do {
478 if (count-- < 0) /* completely used? */
479 return (EADDRNOTAVAIL);
480 ++*lastport;
481 if (*lastport < first || *lastport > last)
482 *lastport = first;
483 lport = htons(*lastport);
484 } while (in_pcblookup_local(pcbinfo, laddr,
485 lport, wild, cred));
486 }
487 *laddrp = laddr.s_addr;
488 *lportp = lport;
489 return (0);
490 }
491
492 /*
493 * Connect from a socket to a specified address.
494 * Both address and port must be specified in argument sin.
495 * If don't have a local address for this socket yet,
496 * then pick one.
497 */
498 int
499 in_pcbconnect(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred)
500 {
501 u_short lport, fport;
502 in_addr_t laddr, faddr;
503 int anonport, error;
504
505 INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
506 INP_WLOCK_ASSERT(inp);
507
508 lport = inp->inp_lport;
509 laddr = inp->inp_laddr.s_addr;
510 anonport = (lport == 0);
511 error = in_pcbconnect_setup(inp, nam, &laddr, &lport, &faddr, &fport,
512 NULL, cred);
513 if (error)
514 return (error);
515
516 /* Do the initial binding of the local address if required. */
517 if (inp->inp_laddr.s_addr == INADDR_ANY && inp->inp_lport == 0) {
518 inp->inp_lport = lport;
519 inp->inp_laddr.s_addr = laddr;
520 if (in_pcbinshash(inp) != 0) {
521 inp->inp_laddr.s_addr = INADDR_ANY;
522 inp->inp_lport = 0;
523 return (EAGAIN);
524 }
525 }
526
527 /* Commit the remaining changes. */
528 inp->inp_lport = lport;
529 inp->inp_laddr.s_addr = laddr;
530 inp->inp_faddr.s_addr = faddr;
531 inp->inp_fport = fport;
532 in_pcbrehash(inp);
533
534 if (anonport)
535 inp->inp_flags |= INP_ANONPORT;
536 return (0);
537 }
538
539 /*
540 * Do proper source address selection on an unbound socket in case
541 * of connect. Take jails into account as well.
542 */
543 static int
544 in_pcbladdr(struct inpcb *inp, struct in_addr *faddr, struct in_addr *laddr,
545 struct ucred *cred)
546 {
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 bzero(&sro, sizeof(sro));
557
558 sin = (struct sockaddr_in *)&sro.ro_dst;
559 sin->sin_family = AF_INET;
560 sin->sin_len = sizeof(struct sockaddr_in);
561 sin->sin_addr.s_addr = faddr->s_addr;
562
563 /*
564 * If route is known our src addr is taken from the i/f,
565 * else punt.
566 *
567 * Find out route to destination.
568 */
569 if ((inp->inp_socket->so_options & SO_DONTROUTE) == 0)
570 in_rtalloc_ign(&sro, 0, inp->inp_inc.inc_fibnum);
571
572 /*
573 * If we found a route, use the address corresponding to
574 * the outgoing interface.
575 *
576 * Otherwise assume faddr is reachable on a directly connected
577 * network and try to find a corresponding interface to take
578 * the source address from.
579 */
580 if (sro.ro_rt == NULL || sro.ro_rt->rt_ifp == NULL) {
581 struct in_ifaddr *ia;
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 || !prison_flag(cred, PR_IP4)) {
593 laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
594 ifa_free(&ia->ia_ifa);
595 goto done;
596 }
597
598 ifp = ia->ia_ifp;
599 ifa_free(&ia->ia_ifa);
600 ia = NULL;
601 IF_ADDR_LOCK(ifp);
602 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
603
604 sa = ifa->ifa_addr;
605 if (sa->sa_family != AF_INET)
606 continue;
607 sin = (struct sockaddr_in *)sa;
608 if (prison_check_ip4(cred, &sin->sin_addr) == 0) {
609 ia = (struct in_ifaddr *)ifa;
610 break;
611 }
612 }
613 if (ia != NULL) {
614 laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
615 IF_ADDR_UNLOCK(ifp);
616 goto done;
617 }
618 IF_ADDR_UNLOCK(ifp);
619
620 /* 3. As a last resort return the 'default' jail address. */
621 error = prison_get_ip4(cred, laddr);
622 goto done;
623 }
624
625 /*
626 * If the outgoing interface on the route found is not
627 * a loopback interface, use the address from that interface.
628 * In case of jails do those three steps:
629 * 1. check if the interface address belongs to the jail. If so use it.
630 * 2. check if we have any address on the outgoing interface
631 * belonging to this jail. If so use it.
632 * 3. as a last resort return the 'default' jail address.
633 */
634 if ((sro.ro_rt->rt_ifp->if_flags & IFF_LOOPBACK) == 0) {
635 struct in_ifaddr *ia;
636 struct ifnet *ifp;
637
638 /* If not jailed, use the default returned. */
639 if (cred == NULL || !prison_flag(cred, PR_IP4)) {
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 ia = NULL;
659 ifp = sro.ro_rt->rt_ifp;
660 IF_ADDR_LOCK(ifp);
661 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
662 sa = ifa->ifa_addr;
663 if (sa->sa_family != AF_INET)
664 continue;
665 sin = (struct sockaddr_in *)sa;
666 if (prison_check_ip4(cred, &sin->sin_addr) == 0) {
667 ia = (struct in_ifaddr *)ifa;
668 break;
669 }
670 }
671 if (ia != NULL) {
672 laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
673 IF_ADDR_UNLOCK(ifp);
674 goto done;
675 }
676 IF_ADDR_UNLOCK(ifp);
677
678 /* 3. As a last resort return the 'default' jail address. */
679 error = prison_get_ip4(cred, laddr);
680 goto done;
681 }
682
683 /*
684 * The outgoing interface is marked with 'loopback net', so a route
685 * to ourselves is here.
686 * Try to find the interface of the destination address and then
687 * take the address from there. That interface is not necessarily
688 * a loopback interface.
689 * In case of jails, check that it is an address of the jail
690 * and if we cannot find, fall back to the 'default' jail address.
691 */
692 if ((sro.ro_rt->rt_ifp->if_flags & IFF_LOOPBACK) != 0) {
693 struct sockaddr_in sain;
694 struct in_ifaddr *ia;
695
696 bzero(&sain, sizeof(struct sockaddr_in));
697 sain.sin_family = AF_INET;
698 sain.sin_len = sizeof(struct sockaddr_in);
699 sain.sin_addr.s_addr = faddr->s_addr;
700
701 ia = ifatoia(ifa_ifwithdstaddr(sintosa(&sain)));
702 if (ia == NULL)
703 ia = ifatoia(ifa_ifwithnet(sintosa(&sain)));
704 if (ia == NULL)
705 ia = ifatoia(ifa_ifwithaddr(sintosa(&sain)));
706
707 if (cred == NULL || !prison_flag(cred, PR_IP4)) {
708 if (ia == NULL) {
709 error = ENETUNREACH;
710 goto done;
711 }
712 laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
713 ifa_free(&ia->ia_ifa);
714 goto done;
715 }
716
717 /* Jailed. */
718 if (ia != NULL) {
719 struct ifnet *ifp;
720
721 ifp = ia->ia_ifp;
722 ifa_free(&ia->ia_ifa);
723 ia = NULL;
724 IF_ADDR_LOCK(ifp);
725 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
726
727 sa = ifa->ifa_addr;
728 if (sa->sa_family != AF_INET)
729 continue;
730 sin = (struct sockaddr_in *)sa;
731 if (prison_check_ip4(cred,
732 &sin->sin_addr) == 0) {
733 ia = (struct in_ifaddr *)ifa;
734 break;
735 }
736 }
737 if (ia != NULL) {
738 laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
739 IF_ADDR_UNLOCK(ifp);
740 goto done;
741 }
742 IF_ADDR_UNLOCK(ifp);
743 }
744
745 /* 3. As a last resort return the 'default' jail address. */
746 error = prison_get_ip4(cred, laddr);
747 goto done;
748 }
749
750 done:
751 if (sro.ro_rt != NULL)
752 RTFREE(sro.ro_rt);
753 return (error);
754 }
755
756 /*
757 * Set up for a connect from a socket to the specified address.
758 * On entry, *laddrp and *lportp should contain the current local
759 * address and port for the PCB; these are updated to the values
760 * that should be placed in inp_laddr and inp_lport to complete
761 * the connect.
762 *
763 * On success, *faddrp and *fportp will be set to the remote address
764 * and port. These are not updated in the error case.
765 *
766 * If the operation fails because the connection already exists,
767 * *oinpp will be set to the PCB of that connection so that the
768 * caller can decide to override it. In all other cases, *oinpp
769 * is set to NULL.
770 */
771 int
772 in_pcbconnect_setup(struct inpcb *inp, struct sockaddr *nam,
773 in_addr_t *laddrp, u_short *lportp, in_addr_t *faddrp, u_short *fportp,
774 struct inpcb **oinpp, struct ucred *cred)
775 {
776 struct sockaddr_in *sin = (struct sockaddr_in *)nam;
777 struct in_ifaddr *ia;
778 struct inpcb *oinp;
779 struct in_addr laddr, faddr;
780 u_short lport, fport;
781 int error;
782
783 /*
784 * Because a global state change doesn't actually occur here, a read
785 * lock is sufficient.
786 */
787 INP_INFO_LOCK_ASSERT(inp->inp_pcbinfo);
788 INP_LOCK_ASSERT(inp);
789
790 if (oinpp != NULL)
791 *oinpp = NULL;
792 if (nam->sa_len != sizeof (*sin))
793 return (EINVAL);
794 if (sin->sin_family != AF_INET)
795 return (EAFNOSUPPORT);
796 if (sin->sin_port == 0)
797 return (EADDRNOTAVAIL);
798 laddr.s_addr = *laddrp;
799 lport = *lportp;
800 faddr = sin->sin_addr;
801 fport = sin->sin_port;
802
803 if (!TAILQ_EMPTY(&V_in_ifaddrhead)) {
804 /*
805 * If the destination address is INADDR_ANY,
806 * use the primary local address.
807 * If the supplied address is INADDR_BROADCAST,
808 * and the primary interface supports broadcast,
809 * choose the broadcast address for that interface.
810 */
811 if (faddr.s_addr == INADDR_ANY) {
812 IN_IFADDR_RLOCK();
813 faddr =
814 IA_SIN(TAILQ_FIRST(&V_in_ifaddrhead))->sin_addr;
815 IN_IFADDR_RUNLOCK();
816 if (cred != NULL &&
817 (error = prison_get_ip4(cred, &faddr)) != 0)
818 return (error);
819 } else if (faddr.s_addr == (u_long)INADDR_BROADCAST) {
820 IN_IFADDR_RLOCK();
821 if (TAILQ_FIRST(&V_in_ifaddrhead)->ia_ifp->if_flags &
822 IFF_BROADCAST)
823 faddr = satosin(&TAILQ_FIRST(
824 &V_in_ifaddrhead)->ia_broadaddr)->sin_addr;
825 IN_IFADDR_RUNLOCK();
826 }
827 }
828 if (laddr.s_addr == INADDR_ANY) {
829 error = in_pcbladdr(inp, &faddr, &laddr, cred);
830 if (error)
831 return (error);
832
833 /*
834 * If the destination address is multicast and an outgoing
835 * interface has been set as a multicast option, use the
836 * address of that interface as our source address.
837 */
838 if (IN_MULTICAST(ntohl(faddr.s_addr)) &&
839 inp->inp_moptions != NULL) {
840 struct ip_moptions *imo;
841 struct ifnet *ifp;
842
843 imo = inp->inp_moptions;
844 if (imo->imo_multicast_ifp != NULL) {
845 ifp = imo->imo_multicast_ifp;
846 IN_IFADDR_RLOCK();
847 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link)
848 if (ia->ia_ifp == ifp)
849 break;
850 if (ia == NULL) {
851 IN_IFADDR_RUNLOCK();
852 return (EADDRNOTAVAIL);
853 }
854 laddr = ia->ia_addr.sin_addr;
855 IN_IFADDR_RUNLOCK();
856 }
857 }
858 }
859
860 oinp = in_pcblookup_hash(inp->inp_pcbinfo, faddr, fport, laddr, lport,
861 0, NULL);
862 if (oinp != NULL) {
863 if (oinpp != NULL)
864 *oinpp = oinp;
865 return (EADDRINUSE);
866 }
867 if (lport == 0) {
868 error = in_pcbbind_setup(inp, NULL, &laddr.s_addr, &lport,
869 cred);
870 if (error)
871 return (error);
872 }
873 *laddrp = laddr.s_addr;
874 *lportp = lport;
875 *faddrp = faddr.s_addr;
876 *fportp = fport;
877 return (0);
878 }
879
880 void
881 in_pcbdisconnect(struct inpcb *inp)
882 {
883
884 INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
885 INP_WLOCK_ASSERT(inp);
886
887 inp->inp_faddr.s_addr = INADDR_ANY;
888 inp->inp_fport = 0;
889 in_pcbrehash(inp);
890 }
891
892 /*
893 * in_pcbdetach() is responsibe for disassociating a socket from an inpcb.
894 * For most protocols, this will be invoked immediately prior to calling
895 * in_pcbfree(). However, with TCP the inpcb may significantly outlive the
896 * socket, in which case in_pcbfree() is deferred.
897 */
898 void
899 in_pcbdetach(struct inpcb *inp)
900 {
901
902 KASSERT(inp->inp_socket != NULL, ("%s: inp_socket == NULL", __func__));
903
904 inp->inp_socket->so_pcb = NULL;
905 inp->inp_socket = NULL;
906 }
907
908 /*
909 * in_pcbfree_internal() frees an inpcb that has been detached from its
910 * socket, and whose reference count has reached 0. It will also remove the
911 * inpcb from any global lists it might remain on.
912 */
913 static void
914 in_pcbfree_internal(struct inpcb *inp)
915 {
916 struct inpcbinfo *ipi = inp->inp_pcbinfo;
917
918 KASSERT(inp->inp_socket == NULL, ("%s: inp_socket != NULL", __func__));
919 KASSERT(inp->inp_refcount == 0, ("%s: refcount !0", __func__));
920
921 INP_INFO_WLOCK_ASSERT(ipi);
922 INP_WLOCK_ASSERT(inp);
923
924 #ifdef IPSEC
925 if (inp->inp_sp != NULL)
926 ipsec_delete_pcbpolicy(inp);
927 #endif /* IPSEC */
928 inp->inp_gencnt = ++ipi->ipi_gencnt;
929 in_pcbremlists(inp);
930 #ifdef INET6
931 if (inp->inp_vflag & INP_IPV6PROTO) {
932 ip6_freepcbopts(inp->in6p_outputopts);
933 if (inp->in6p_moptions != NULL)
934 ip6_freemoptions(inp->in6p_moptions);
935 }
936 #endif
937 if (inp->inp_options)
938 (void)m_free(inp->inp_options);
939 if (inp->inp_moptions != NULL)
940 inp_freemoptions(inp->inp_moptions);
941 inp->inp_vflag = 0;
942 crfree(inp->inp_cred);
943
944 #ifdef MAC
945 mac_inpcb_destroy(inp);
946 #endif
947 INP_WUNLOCK(inp);
948 uma_zfree(ipi->ipi_zone, inp);
949 }
950
951 /*
952 * in_pcbref() bumps the reference count on an inpcb in order to maintain
953 * stability of an inpcb pointer despite the inpcb lock being released. This
954 * is used in TCP when the inpcbinfo lock needs to be acquired or upgraded,
955 * but where the inpcb lock is already held.
956 *
957 * While the inpcb will not be freed, releasing the inpcb lock means that the
958 * connection's state may change, so the caller should be careful to
959 * revalidate any cached state on reacquiring the lock. Drop the reference
960 * using in_pcbrele().
961 */
962 void
963 in_pcbref(struct inpcb *inp)
964 {
965
966 INP_WLOCK_ASSERT(inp);
967
968 KASSERT(inp->inp_refcount > 0, ("%s: refcount 0", __func__));
969
970 inp->inp_refcount++;
971 }
972
973 /*
974 * Drop a refcount on an inpcb elevated using in_pcbref(); because a call to
975 * in_pcbfree() may have been made between in_pcbref() and in_pcbrele(), we
976 * return a flag indicating whether or not the inpcb remains valid. If it is
977 * valid, we return with the inpcb lock held.
978 */
979 int
980 in_pcbrele(struct inpcb *inp)
981 {
982 #ifdef INVARIANTS
983 struct inpcbinfo *ipi = inp->inp_pcbinfo;
984 #endif
985
986 KASSERT(inp->inp_refcount > 0, ("%s: refcount 0", __func__));
987
988 INP_INFO_WLOCK_ASSERT(ipi);
989 INP_WLOCK_ASSERT(inp);
990
991 inp->inp_refcount--;
992 if (inp->inp_refcount > 0)
993 return (0);
994 in_pcbfree_internal(inp);
995 return (1);
996 }
997
998 /*
999 * Unconditionally schedule an inpcb to be freed by decrementing its
1000 * reference count, which should occur only after the inpcb has been detached
1001 * from its socket. If another thread holds a temporary reference (acquired
1002 * using in_pcbref()) then the free is deferred until that reference is
1003 * released using in_pcbrele(), but the inpcb is still unlocked.
1004 */
1005 void
1006 in_pcbfree(struct inpcb *inp)
1007 {
1008 #ifdef INVARIANTS
1009 struct inpcbinfo *ipi = inp->inp_pcbinfo;
1010 #endif
1011
1012 KASSERT(inp->inp_socket == NULL, ("%s: inp_socket != NULL",
1013 __func__));
1014
1015 INP_INFO_WLOCK_ASSERT(ipi);
1016 INP_WLOCK_ASSERT(inp);
1017
1018 if (!in_pcbrele(inp))
1019 INP_WUNLOCK(inp);
1020 }
1021
1022 /*
1023 * in_pcbdrop() removes an inpcb from hashed lists, releasing its address and
1024 * port reservation, and preventing it from being returned by inpcb lookups.
1025 *
1026 * It is used by TCP to mark an inpcb as unused and avoid future packet
1027 * delivery or event notification when a socket remains open but TCP has
1028 * closed. This might occur as a result of a shutdown()-initiated TCP close
1029 * or a RST on the wire, and allows the port binding to be reused while still
1030 * maintaining the invariant that so_pcb always points to a valid inpcb until
1031 * in_pcbdetach().
1032 *
1033 * XXXRW: An inp_lport of 0 is used to indicate that the inpcb is not on hash
1034 * lists, but can lead to confusing netstat output, as open sockets with
1035 * closed TCP connections will no longer appear to have their bound port
1036 * number. An explicit flag would be better, as it would allow us to leave
1037 * the port number intact after the connection is dropped.
1038 *
1039 * XXXRW: Possibly in_pcbdrop() should also prevent future notifications by
1040 * in_pcbnotifyall() and in_pcbpurgeif0()?
1041 */
1042 void
1043 in_pcbdrop(struct inpcb *inp)
1044 {
1045
1046 INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
1047 INP_WLOCK_ASSERT(inp);
1048
1049 inp->inp_flags |= INP_DROPPED;
1050 if (inp->inp_flags & INP_INHASHLIST) {
1051 struct inpcbport *phd = inp->inp_phd;
1052
1053 LIST_REMOVE(inp, inp_hash);
1054 LIST_REMOVE(inp, inp_portlist);
1055 if (LIST_FIRST(&phd->phd_pcblist) == NULL) {
1056 LIST_REMOVE(phd, phd_hash);
1057 free(phd, M_PCB);
1058 }
1059 inp->inp_flags &= ~INP_INHASHLIST;
1060 }
1061 }
1062
1063 /*
1064 * Common routines to return the socket addresses associated with inpcbs.
1065 */
1066 struct sockaddr *
1067 in_sockaddr(in_port_t port, struct in_addr *addr_p)
1068 {
1069 struct sockaddr_in *sin;
1070
1071 sin = malloc(sizeof *sin, M_SONAME,
1072 M_WAITOK | M_ZERO);
1073 sin->sin_family = AF_INET;
1074 sin->sin_len = sizeof(*sin);
1075 sin->sin_addr = *addr_p;
1076 sin->sin_port = port;
1077
1078 return (struct sockaddr *)sin;
1079 }
1080
1081 int
1082 in_getsockaddr(struct socket *so, struct sockaddr **nam)
1083 {
1084 struct inpcb *inp;
1085 struct in_addr addr;
1086 in_port_t port;
1087
1088 inp = sotoinpcb(so);
1089 KASSERT(inp != NULL, ("in_getsockaddr: inp == NULL"));
1090
1091 INP_RLOCK(inp);
1092 port = inp->inp_lport;
1093 addr = inp->inp_laddr;
1094 INP_RUNLOCK(inp);
1095
1096 *nam = in_sockaddr(port, &addr);
1097 return 0;
1098 }
1099
1100 int
1101 in_getpeeraddr(struct socket *so, struct sockaddr **nam)
1102 {
1103 struct inpcb *inp;
1104 struct in_addr addr;
1105 in_port_t port;
1106
1107 inp = sotoinpcb(so);
1108 KASSERT(inp != NULL, ("in_getpeeraddr: inp == NULL"));
1109
1110 INP_RLOCK(inp);
1111 port = inp->inp_fport;
1112 addr = inp->inp_faddr;
1113 INP_RUNLOCK(inp);
1114
1115 *nam = in_sockaddr(port, &addr);
1116 return 0;
1117 }
1118
1119 void
1120 in_pcbnotifyall(struct inpcbinfo *pcbinfo, struct in_addr faddr, int errno,
1121 struct inpcb *(*notify)(struct inpcb *, int))
1122 {
1123 struct inpcb *inp, *inp_temp;
1124
1125 INP_INFO_WLOCK(pcbinfo);
1126 LIST_FOREACH_SAFE(inp, pcbinfo->ipi_listhead, inp_list, inp_temp) {
1127 INP_WLOCK(inp);
1128 #ifdef INET6
1129 if ((inp->inp_vflag & INP_IPV4) == 0) {
1130 INP_WUNLOCK(inp);
1131 continue;
1132 }
1133 #endif
1134 if (inp->inp_faddr.s_addr != faddr.s_addr ||
1135 inp->inp_socket == NULL) {
1136 INP_WUNLOCK(inp);
1137 continue;
1138 }
1139 if ((*notify)(inp, errno))
1140 INP_WUNLOCK(inp);
1141 }
1142 INP_INFO_WUNLOCK(pcbinfo);
1143 }
1144
1145 void
1146 in_pcbpurgeif0(struct inpcbinfo *pcbinfo, struct ifnet *ifp)
1147 {
1148 struct inpcb *inp;
1149 struct ip_moptions *imo;
1150 int i, gap;
1151
1152 INP_INFO_RLOCK(pcbinfo);
1153 LIST_FOREACH(inp, pcbinfo->ipi_listhead, inp_list) {
1154 INP_WLOCK(inp);
1155 imo = inp->inp_moptions;
1156 if ((inp->inp_vflag & INP_IPV4) &&
1157 imo != NULL) {
1158 /*
1159 * Unselect the outgoing interface if it is being
1160 * detached.
1161 */
1162 if (imo->imo_multicast_ifp == ifp)
1163 imo->imo_multicast_ifp = NULL;
1164
1165 /*
1166 * Drop multicast group membership if we joined
1167 * through the interface being detached.
1168 */
1169 for (i = 0, gap = 0; i < imo->imo_num_memberships;
1170 i++) {
1171 if (imo->imo_membership[i]->inm_ifp == ifp) {
1172 in_delmulti(imo->imo_membership[i]);
1173 gap++;
1174 } else if (gap != 0)
1175 imo->imo_membership[i - gap] =
1176 imo->imo_membership[i];
1177 }
1178 imo->imo_num_memberships -= gap;
1179 }
1180 INP_WUNLOCK(inp);
1181 }
1182 INP_INFO_RUNLOCK(pcbinfo);
1183 }
1184
1185 /*
1186 * Lookup a PCB based on the local address and port.
1187 */
1188 #define INP_LOOKUP_MAPPED_PCB_COST 3
1189 struct inpcb *
1190 in_pcblookup_local(struct inpcbinfo *pcbinfo, struct in_addr laddr,
1191 u_short lport, int wild_okay, struct ucred *cred)
1192 {
1193 struct inpcb *inp;
1194 #ifdef INET6
1195 int matchwild = 3 + INP_LOOKUP_MAPPED_PCB_COST;
1196 #else
1197 int matchwild = 3;
1198 #endif
1199 int wildcard;
1200
1201 INP_INFO_LOCK_ASSERT(pcbinfo);
1202
1203 if (!wild_okay) {
1204 struct inpcbhead *head;
1205 /*
1206 * Look for an unconnected (wildcard foreign addr) PCB that
1207 * matches the local address and port we're looking for.
1208 */
1209 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport,
1210 0, pcbinfo->ipi_hashmask)];
1211 LIST_FOREACH(inp, head, inp_hash) {
1212 #ifdef INET6
1213 /* XXX inp locking */
1214 if ((inp->inp_vflag & INP_IPV4) == 0)
1215 continue;
1216 #endif
1217 if (inp->inp_faddr.s_addr == INADDR_ANY &&
1218 inp->inp_laddr.s_addr == laddr.s_addr &&
1219 inp->inp_lport == lport) {
1220 /*
1221 * Found?
1222 */
1223 if (cred == NULL ||
1224 prison_equal_ip4(cred->cr_prison,
1225 inp->inp_cred->cr_prison))
1226 return (inp);
1227 }
1228 }
1229 /*
1230 * Not found.
1231 */
1232 return (NULL);
1233 } else {
1234 struct inpcbporthead *porthash;
1235 struct inpcbport *phd;
1236 struct inpcb *match = NULL;
1237 /*
1238 * Best fit PCB lookup.
1239 *
1240 * First see if this local port is in use by looking on the
1241 * port hash list.
1242 */
1243 porthash = &pcbinfo->ipi_porthashbase[INP_PCBPORTHASH(lport,
1244 pcbinfo->ipi_porthashmask)];
1245 LIST_FOREACH(phd, porthash, phd_hash) {
1246 if (phd->phd_port == lport)
1247 break;
1248 }
1249 if (phd != NULL) {
1250 /*
1251 * Port is in use by one or more PCBs. Look for best
1252 * fit.
1253 */
1254 LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) {
1255 wildcard = 0;
1256 if (cred != NULL &&
1257 !prison_equal_ip4(inp->inp_cred->cr_prison,
1258 cred->cr_prison))
1259 continue;
1260 #ifdef INET6
1261 /* XXX inp locking */
1262 if ((inp->inp_vflag & INP_IPV4) == 0)
1263 continue;
1264 /*
1265 * We never select the PCB that has
1266 * INP_IPV6 flag and is bound to :: if
1267 * we have another PCB which is bound
1268 * to 0.0.0.0. If a PCB has the
1269 * INP_IPV6 flag, then we set its cost
1270 * higher than IPv4 only PCBs.
1271 *
1272 * Note that the case only happens
1273 * when a socket is bound to ::, under
1274 * the condition that the use of the
1275 * mapped address is allowed.
1276 */
1277 if ((inp->inp_vflag & INP_IPV6) != 0)
1278 wildcard += INP_LOOKUP_MAPPED_PCB_COST;
1279 #endif
1280 if (inp->inp_faddr.s_addr != INADDR_ANY)
1281 wildcard++;
1282 if (inp->inp_laddr.s_addr != INADDR_ANY) {
1283 if (laddr.s_addr == INADDR_ANY)
1284 wildcard++;
1285 else if (inp->inp_laddr.s_addr != laddr.s_addr)
1286 continue;
1287 } else {
1288 if (laddr.s_addr != INADDR_ANY)
1289 wildcard++;
1290 }
1291 if (wildcard < matchwild) {
1292 match = inp;
1293 matchwild = wildcard;
1294 if (matchwild == 0)
1295 break;
1296 }
1297 }
1298 }
1299 return (match);
1300 }
1301 }
1302 #undef INP_LOOKUP_MAPPED_PCB_COST
1303
1304 /*
1305 * Lookup PCB in hash list.
1306 */
1307 struct inpcb *
1308 in_pcblookup_hash(struct inpcbinfo *pcbinfo, struct in_addr faddr,
1309 u_int fport_arg, struct in_addr laddr, u_int lport_arg, int wildcard,
1310 struct ifnet *ifp)
1311 {
1312 struct inpcbhead *head;
1313 struct inpcb *inp, *tmpinp;
1314 u_short fport = fport_arg, lport = lport_arg;
1315
1316 INP_INFO_LOCK_ASSERT(pcbinfo);
1317
1318 /*
1319 * First look for an exact match.
1320 */
1321 tmpinp = NULL;
1322 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(faddr.s_addr, lport, fport,
1323 pcbinfo->ipi_hashmask)];
1324 LIST_FOREACH(inp, head, inp_hash) {
1325 #ifdef INET6
1326 /* XXX inp locking */
1327 if ((inp->inp_vflag & INP_IPV4) == 0)
1328 continue;
1329 #endif
1330 if (inp->inp_faddr.s_addr == faddr.s_addr &&
1331 inp->inp_laddr.s_addr == laddr.s_addr &&
1332 inp->inp_fport == fport &&
1333 inp->inp_lport == lport) {
1334 /*
1335 * XXX We should be able to directly return
1336 * the inp here, without any checks.
1337 * Well unless both bound with SO_REUSEPORT?
1338 */
1339 if (prison_flag(inp->inp_cred, PR_IP4))
1340 return (inp);
1341 if (tmpinp == NULL)
1342 tmpinp = inp;
1343 }
1344 }
1345 if (tmpinp != NULL)
1346 return (tmpinp);
1347
1348 /*
1349 * Then look for a wildcard match, if requested.
1350 */
1351 if (wildcard == INPLOOKUP_WILDCARD) {
1352 struct inpcb *local_wild = NULL, *local_exact = NULL;
1353 #ifdef INET6
1354 struct inpcb *local_wild_mapped = NULL;
1355 #endif
1356 struct inpcb *jail_wild = NULL;
1357 int injail;
1358
1359 /*
1360 * Order of socket selection - we always prefer jails.
1361 * 1. jailed, non-wild.
1362 * 2. jailed, wild.
1363 * 3. non-jailed, non-wild.
1364 * 4. non-jailed, wild.
1365 */
1366
1367 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport,
1368 0, pcbinfo->ipi_hashmask)];
1369 LIST_FOREACH(inp, head, inp_hash) {
1370 #ifdef INET6
1371 /* XXX inp locking */
1372 if ((inp->inp_vflag & INP_IPV4) == 0)
1373 continue;
1374 #endif
1375 if (inp->inp_faddr.s_addr != INADDR_ANY ||
1376 inp->inp_lport != lport)
1377 continue;
1378
1379 /* XXX inp locking */
1380 if (ifp && ifp->if_type == IFT_FAITH &&
1381 (inp->inp_flags & INP_FAITH) == 0)
1382 continue;
1383
1384 injail = prison_flag(inp->inp_cred, PR_IP4);
1385 if (injail) {
1386 if (prison_check_ip4(inp->inp_cred,
1387 &laddr) != 0)
1388 continue;
1389 } else {
1390 if (local_exact != NULL)
1391 continue;
1392 }
1393
1394 if (inp->inp_laddr.s_addr == laddr.s_addr) {
1395 if (injail)
1396 return (inp);
1397 else
1398 local_exact = inp;
1399 } else if (inp->inp_laddr.s_addr == INADDR_ANY) {
1400 #ifdef INET6
1401 /* XXX inp locking, NULL check */
1402 if (inp->inp_vflag & INP_IPV6PROTO)
1403 local_wild_mapped = inp;
1404 else
1405 #endif /* INET6 */
1406 if (injail)
1407 jail_wild = inp;
1408 else
1409 local_wild = inp;
1410 }
1411 } /* LIST_FOREACH */
1412 if (jail_wild != NULL)
1413 return (jail_wild);
1414 if (local_exact != NULL)
1415 return (local_exact);
1416 if (local_wild != NULL)
1417 return (local_wild);
1418 #ifdef INET6
1419 if (local_wild_mapped != NULL)
1420 return (local_wild_mapped);
1421 #endif /* defined(INET6) */
1422 } /* if (wildcard == INPLOOKUP_WILDCARD) */
1423
1424 return (NULL);
1425 }
1426
1427 /*
1428 * Insert PCB onto various hash lists.
1429 */
1430 int
1431 in_pcbinshash(struct inpcb *inp)
1432 {
1433 struct inpcbhead *pcbhash;
1434 struct inpcbporthead *pcbporthash;
1435 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1436 struct inpcbport *phd;
1437 u_int32_t hashkey_faddr;
1438
1439 INP_INFO_WLOCK_ASSERT(pcbinfo);
1440 INP_WLOCK_ASSERT(inp);
1441 KASSERT((inp->inp_flags & INP_INHASHLIST) == 0,
1442 ("in_pcbinshash: INP_INHASHLIST"));
1443
1444 #ifdef INET6
1445 if (inp->inp_vflag & INP_IPV6)
1446 hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX */;
1447 else
1448 #endif /* INET6 */
1449 hashkey_faddr = inp->inp_faddr.s_addr;
1450
1451 pcbhash = &pcbinfo->ipi_hashbase[INP_PCBHASH(hashkey_faddr,
1452 inp->inp_lport, inp->inp_fport, pcbinfo->ipi_hashmask)];
1453
1454 pcbporthash = &pcbinfo->ipi_porthashbase[
1455 INP_PCBPORTHASH(inp->inp_lport, pcbinfo->ipi_porthashmask)];
1456
1457 /*
1458 * Go through port list and look for a head for this lport.
1459 */
1460 LIST_FOREACH(phd, pcbporthash, phd_hash) {
1461 if (phd->phd_port == inp->inp_lport)
1462 break;
1463 }
1464 /*
1465 * If none exists, malloc one and tack it on.
1466 */
1467 if (phd == NULL) {
1468 phd = malloc(sizeof(struct inpcbport), M_PCB, M_NOWAIT);
1469 if (phd == NULL) {
1470 return (ENOBUFS); /* XXX */
1471 }
1472 phd->phd_port = inp->inp_lport;
1473 LIST_INIT(&phd->phd_pcblist);
1474 LIST_INSERT_HEAD(pcbporthash, phd, phd_hash);
1475 }
1476 inp->inp_phd = phd;
1477 LIST_INSERT_HEAD(&phd->phd_pcblist, inp, inp_portlist);
1478 LIST_INSERT_HEAD(pcbhash, inp, inp_hash);
1479 inp->inp_flags |= INP_INHASHLIST;
1480 return (0);
1481 }
1482
1483 /*
1484 * Move PCB to the proper hash bucket when { faddr, fport } have been
1485 * changed. NOTE: This does not handle the case of the lport changing (the
1486 * hashed port list would have to be updated as well), so the lport must
1487 * not change after in_pcbinshash() has been called.
1488 */
1489 void
1490 in_pcbrehash(struct inpcb *inp)
1491 {
1492 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1493 struct inpcbhead *head;
1494 u_int32_t hashkey_faddr;
1495
1496 INP_INFO_WLOCK_ASSERT(pcbinfo);
1497 INP_WLOCK_ASSERT(inp);
1498 KASSERT(inp->inp_flags & INP_INHASHLIST,
1499 ("in_pcbrehash: !INP_INHASHLIST"));
1500
1501 #ifdef INET6
1502 if (inp->inp_vflag & INP_IPV6)
1503 hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX */;
1504 else
1505 #endif /* INET6 */
1506 hashkey_faddr = inp->inp_faddr.s_addr;
1507
1508 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(hashkey_faddr,
1509 inp->inp_lport, inp->inp_fport, pcbinfo->ipi_hashmask)];
1510
1511 LIST_REMOVE(inp, inp_hash);
1512 LIST_INSERT_HEAD(head, inp, inp_hash);
1513 }
1514
1515 /*
1516 * Remove PCB from various lists.
1517 */
1518 static void
1519 in_pcbremlists(struct inpcb *inp)
1520 {
1521 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1522
1523 INP_INFO_WLOCK_ASSERT(pcbinfo);
1524 INP_WLOCK_ASSERT(inp);
1525
1526 inp->inp_gencnt = ++pcbinfo->ipi_gencnt;
1527 if (inp->inp_flags & INP_INHASHLIST) {
1528 struct inpcbport *phd = inp->inp_phd;
1529
1530 LIST_REMOVE(inp, inp_hash);
1531 LIST_REMOVE(inp, inp_portlist);
1532 if (LIST_FIRST(&phd->phd_pcblist) == NULL) {
1533 LIST_REMOVE(phd, phd_hash);
1534 free(phd, M_PCB);
1535 }
1536 inp->inp_flags &= ~INP_INHASHLIST;
1537 }
1538 LIST_REMOVE(inp, inp_list);
1539 pcbinfo->ipi_count--;
1540 }
1541
1542 /*
1543 * A set label operation has occurred at the socket layer, propagate the
1544 * label change into the in_pcb for the socket.
1545 */
1546 void
1547 in_pcbsosetlabel(struct socket *so)
1548 {
1549 #ifdef MAC
1550 struct inpcb *inp;
1551
1552 inp = sotoinpcb(so);
1553 KASSERT(inp != NULL, ("in_pcbsosetlabel: so->so_pcb == NULL"));
1554
1555 INP_WLOCK(inp);
1556 SOCK_LOCK(so);
1557 mac_inpcb_sosetlabel(so, inp);
1558 SOCK_UNLOCK(so);
1559 INP_WUNLOCK(inp);
1560 #endif
1561 }
1562
1563 /*
1564 * ipport_tick runs once per second, determining if random port allocation
1565 * should be continued. If more than ipport_randomcps ports have been
1566 * allocated in the last second, then we return to sequential port
1567 * allocation. We return to random allocation only once we drop below
1568 * ipport_randomcps for at least ipport_randomtime seconds.
1569 */
1570 void
1571 ipport_tick(void *xtp)
1572 {
1573 VNET_ITERATOR_DECL(vnet_iter);
1574
1575 VNET_LIST_RLOCK_NOSLEEP();
1576 VNET_FOREACH(vnet_iter) {
1577 CURVNET_SET(vnet_iter); /* XXX appease INVARIANTS here */
1578 if (V_ipport_tcpallocs <=
1579 V_ipport_tcplastcount + V_ipport_randomcps) {
1580 if (V_ipport_stoprandom > 0)
1581 V_ipport_stoprandom--;
1582 } else
1583 V_ipport_stoprandom = V_ipport_randomtime;
1584 V_ipport_tcplastcount = V_ipport_tcpallocs;
1585 CURVNET_RESTORE();
1586 }
1587 VNET_LIST_RUNLOCK_NOSLEEP();
1588 callout_reset(&ipport_tick_callout, hz, ipport_tick, NULL);
1589 }
1590
1591 void
1592 inp_wlock(struct inpcb *inp)
1593 {
1594
1595 INP_WLOCK(inp);
1596 }
1597
1598 void
1599 inp_wunlock(struct inpcb *inp)
1600 {
1601
1602 INP_WUNLOCK(inp);
1603 }
1604
1605 void
1606 inp_rlock(struct inpcb *inp)
1607 {
1608
1609 INP_RLOCK(inp);
1610 }
1611
1612 void
1613 inp_runlock(struct inpcb *inp)
1614 {
1615
1616 INP_RUNLOCK(inp);
1617 }
1618
1619 #ifdef INVARIANTS
1620 void
1621 inp_lock_assert(struct inpcb *inp)
1622 {
1623
1624 INP_WLOCK_ASSERT(inp);
1625 }
1626
1627 void
1628 inp_unlock_assert(struct inpcb *inp)
1629 {
1630
1631 INP_UNLOCK_ASSERT(inp);
1632 }
1633 #endif
1634
1635 void
1636 inp_apply_all(void (*func)(struct inpcb *, void *), void *arg)
1637 {
1638 struct inpcb *inp;
1639
1640 INP_INFO_RLOCK(&V_tcbinfo);
1641 LIST_FOREACH(inp, V_tcbinfo.ipi_listhead, inp_list) {
1642 INP_WLOCK(inp);
1643 func(inp, arg);
1644 INP_WUNLOCK(inp);
1645 }
1646 INP_INFO_RUNLOCK(&V_tcbinfo);
1647 }
1648
1649 struct socket *
1650 inp_inpcbtosocket(struct inpcb *inp)
1651 {
1652
1653 INP_WLOCK_ASSERT(inp);
1654 return (inp->inp_socket);
1655 }
1656
1657 struct tcpcb *
1658 inp_inpcbtotcpcb(struct inpcb *inp)
1659 {
1660
1661 INP_WLOCK_ASSERT(inp);
1662 return ((struct tcpcb *)inp->inp_ppcb);
1663 }
1664
1665 int
1666 inp_ip_tos_get(const struct inpcb *inp)
1667 {
1668
1669 return (inp->inp_ip_tos);
1670 }
1671
1672 void
1673 inp_ip_tos_set(struct inpcb *inp, int val)
1674 {
1675
1676 inp->inp_ip_tos = val;
1677 }
1678
1679 void
1680 inp_4tuple_get(struct inpcb *inp, uint32_t *laddr, uint16_t *lp,
1681 uint32_t *faddr, uint16_t *fp)
1682 {
1683
1684 INP_LOCK_ASSERT(inp);
1685 *laddr = inp->inp_laddr.s_addr;
1686 *faddr = inp->inp_faddr.s_addr;
1687 *lp = inp->inp_lport;
1688 *fp = inp->inp_fport;
1689 }
1690
1691 struct inpcb *
1692 so_sotoinpcb(struct socket *so)
1693 {
1694
1695 return (sotoinpcb(so));
1696 }
1697
1698 struct tcpcb *
1699 so_sototcpcb(struct socket *so)
1700 {
1701
1702 return (sototcpcb(so));
1703 }
1704
1705 #ifdef DDB
1706 static void
1707 db_print_indent(int indent)
1708 {
1709 int i;
1710
1711 for (i = 0; i < indent; i++)
1712 db_printf(" ");
1713 }
1714
1715 static void
1716 db_print_inconninfo(struct in_conninfo *inc, const char *name, int indent)
1717 {
1718 char faddr_str[48], laddr_str[48];
1719
1720 db_print_indent(indent);
1721 db_printf("%s at %p\n", name, inc);
1722
1723 indent += 2;
1724
1725 #ifdef INET6
1726 if (inc->inc_flags & INC_ISIPV6) {
1727 /* IPv6. */
1728 ip6_sprintf(laddr_str, &inc->inc6_laddr);
1729 ip6_sprintf(faddr_str, &inc->inc6_faddr);
1730 } else {
1731 #endif
1732 /* IPv4. */
1733 inet_ntoa_r(inc->inc_laddr, laddr_str);
1734 inet_ntoa_r(inc->inc_faddr, faddr_str);
1735 #ifdef INET6
1736 }
1737 #endif
1738 db_print_indent(indent);
1739 db_printf("inc_laddr %s inc_lport %u\n", laddr_str,
1740 ntohs(inc->inc_lport));
1741 db_print_indent(indent);
1742 db_printf("inc_faddr %s inc_fport %u\n", faddr_str,
1743 ntohs(inc->inc_fport));
1744 }
1745
1746 static void
1747 db_print_inpflags(int inp_flags)
1748 {
1749 int comma;
1750
1751 comma = 0;
1752 if (inp_flags & INP_RECVOPTS) {
1753 db_printf("%sINP_RECVOPTS", comma ? ", " : "");
1754 comma = 1;
1755 }
1756 if (inp_flags & INP_RECVRETOPTS) {
1757 db_printf("%sINP_RECVRETOPTS", comma ? ", " : "");
1758 comma = 1;
1759 }
1760 if (inp_flags & INP_RECVDSTADDR) {
1761 db_printf("%sINP_RECVDSTADDR", comma ? ", " : "");
1762 comma = 1;
1763 }
1764 if (inp_flags & INP_HDRINCL) {
1765 db_printf("%sINP_HDRINCL", comma ? ", " : "");
1766 comma = 1;
1767 }
1768 if (inp_flags & INP_HIGHPORT) {
1769 db_printf("%sINP_HIGHPORT", comma ? ", " : "");
1770 comma = 1;
1771 }
1772 if (inp_flags & INP_LOWPORT) {
1773 db_printf("%sINP_LOWPORT", comma ? ", " : "");
1774 comma = 1;
1775 }
1776 if (inp_flags & INP_ANONPORT) {
1777 db_printf("%sINP_ANONPORT", comma ? ", " : "");
1778 comma = 1;
1779 }
1780 if (inp_flags & INP_RECVIF) {
1781 db_printf("%sINP_RECVIF", comma ? ", " : "");
1782 comma = 1;
1783 }
1784 if (inp_flags & INP_MTUDISC) {
1785 db_printf("%sINP_MTUDISC", comma ? ", " : "");
1786 comma = 1;
1787 }
1788 if (inp_flags & INP_FAITH) {
1789 db_printf("%sINP_FAITH", comma ? ", " : "");
1790 comma = 1;
1791 }
1792 if (inp_flags & INP_RECVTTL) {
1793 db_printf("%sINP_RECVTTL", comma ? ", " : "");
1794 comma = 1;
1795 }
1796 if (inp_flags & INP_DONTFRAG) {
1797 db_printf("%sINP_DONTFRAG", comma ? ", " : "");
1798 comma = 1;
1799 }
1800 if (inp_flags & IN6P_IPV6_V6ONLY) {
1801 db_printf("%sIN6P_IPV6_V6ONLY", comma ? ", " : "");
1802 comma = 1;
1803 }
1804 if (inp_flags & IN6P_PKTINFO) {
1805 db_printf("%sIN6P_PKTINFO", comma ? ", " : "");
1806 comma = 1;
1807 }
1808 if (inp_flags & IN6P_HOPLIMIT) {
1809 db_printf("%sIN6P_HOPLIMIT", comma ? ", " : "");
1810 comma = 1;
1811 }
1812 if (inp_flags & IN6P_HOPOPTS) {
1813 db_printf("%sIN6P_HOPOPTS", comma ? ", " : "");
1814 comma = 1;
1815 }
1816 if (inp_flags & IN6P_DSTOPTS) {
1817 db_printf("%sIN6P_DSTOPTS", comma ? ", " : "");
1818 comma = 1;
1819 }
1820 if (inp_flags & IN6P_RTHDR) {
1821 db_printf("%sIN6P_RTHDR", comma ? ", " : "");
1822 comma = 1;
1823 }
1824 if (inp_flags & IN6P_RTHDRDSTOPTS) {
1825 db_printf("%sIN6P_RTHDRDSTOPTS", comma ? ", " : "");
1826 comma = 1;
1827 }
1828 if (inp_flags & IN6P_TCLASS) {
1829 db_printf("%sIN6P_TCLASS", comma ? ", " : "");
1830 comma = 1;
1831 }
1832 if (inp_flags & IN6P_AUTOFLOWLABEL) {
1833 db_printf("%sIN6P_AUTOFLOWLABEL", comma ? ", " : "");
1834 comma = 1;
1835 }
1836 if (inp_flags & INP_TIMEWAIT) {
1837 db_printf("%sINP_TIMEWAIT", comma ? ", " : "");
1838 comma = 1;
1839 }
1840 if (inp_flags & INP_ONESBCAST) {
1841 db_printf("%sINP_ONESBCAST", comma ? ", " : "");
1842 comma = 1;
1843 }
1844 if (inp_flags & INP_DROPPED) {
1845 db_printf("%sINP_DROPPED", comma ? ", " : "");
1846 comma = 1;
1847 }
1848 if (inp_flags & INP_SOCKREF) {
1849 db_printf("%sINP_SOCKREF", comma ? ", " : "");
1850 comma = 1;
1851 }
1852 if (inp_flags & IN6P_RFC2292) {
1853 db_printf("%sIN6P_RFC2292", comma ? ", " : "");
1854 comma = 1;
1855 }
1856 if (inp_flags & IN6P_MTU) {
1857 db_printf("IN6P_MTU%s", comma ? ", " : "");
1858 comma = 1;
1859 }
1860 }
1861
1862 static void
1863 db_print_inpvflag(u_char inp_vflag)
1864 {
1865 int comma;
1866
1867 comma = 0;
1868 if (inp_vflag & INP_IPV4) {
1869 db_printf("%sINP_IPV4", comma ? ", " : "");
1870 comma = 1;
1871 }
1872 if (inp_vflag & INP_IPV6) {
1873 db_printf("%sINP_IPV6", comma ? ", " : "");
1874 comma = 1;
1875 }
1876 if (inp_vflag & INP_IPV6PROTO) {
1877 db_printf("%sINP_IPV6PROTO", comma ? ", " : "");
1878 comma = 1;
1879 }
1880 }
1881
1882 static void
1883 db_print_inpcb(struct inpcb *inp, const char *name, int indent)
1884 {
1885
1886 db_print_indent(indent);
1887 db_printf("%s at %p\n", name, inp);
1888
1889 indent += 2;
1890
1891 db_print_indent(indent);
1892 db_printf("inp_flow: 0x%x\n", inp->inp_flow);
1893
1894 db_print_inconninfo(&inp->inp_inc, "inp_conninfo", indent);
1895
1896 db_print_indent(indent);
1897 db_printf("inp_ppcb: %p inp_pcbinfo: %p inp_socket: %p\n",
1898 inp->inp_ppcb, inp->inp_pcbinfo, inp->inp_socket);
1899
1900 db_print_indent(indent);
1901 db_printf("inp_label: %p inp_flags: 0x%x (",
1902 inp->inp_label, inp->inp_flags);
1903 db_print_inpflags(inp->inp_flags);
1904 db_printf(")\n");
1905
1906 db_print_indent(indent);
1907 db_printf("inp_sp: %p inp_vflag: 0x%x (", inp->inp_sp,
1908 inp->inp_vflag);
1909 db_print_inpvflag(inp->inp_vflag);
1910 db_printf(")\n");
1911
1912 db_print_indent(indent);
1913 db_printf("inp_ip_ttl: %d inp_ip_p: %d inp_ip_minttl: %d\n",
1914 inp->inp_ip_ttl, inp->inp_ip_p, inp->inp_ip_minttl);
1915
1916 db_print_indent(indent);
1917 #ifdef INET6
1918 if (inp->inp_vflag & INP_IPV6) {
1919 db_printf("in6p_options: %p in6p_outputopts: %p "
1920 "in6p_moptions: %p\n", inp->in6p_options,
1921 inp->in6p_outputopts, inp->in6p_moptions);
1922 db_printf("in6p_icmp6filt: %p in6p_cksum %d "
1923 "in6p_hops %u\n", inp->in6p_icmp6filt, inp->in6p_cksum,
1924 inp->in6p_hops);
1925 } else
1926 #endif
1927 {
1928 db_printf("inp_ip_tos: %d inp_ip_options: %p "
1929 "inp_ip_moptions: %p\n", inp->inp_ip_tos,
1930 inp->inp_options, inp->inp_moptions);
1931 }
1932
1933 db_print_indent(indent);
1934 db_printf("inp_phd: %p inp_gencnt: %ju\n", inp->inp_phd,
1935 (uintmax_t)inp->inp_gencnt);
1936 }
1937
1938 DB_SHOW_COMMAND(inpcb, db_show_inpcb)
1939 {
1940 struct inpcb *inp;
1941
1942 if (!have_addr) {
1943 db_printf("usage: show inpcb <addr>\n");
1944 return;
1945 }
1946 inp = (struct inpcb *)addr;
1947
1948 db_print_inpcb(inp, "inpcb", 0);
1949 }
1950 #endif
Cache object: d3539b56456ec4f2960d40750102dfa7
|