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 * Copyright (c) 2010-2011 Juniper Networks, Inc.
8 * Portions of this software were developed by Robert N. M. Watson under
9 * contract to Juniper Networks, Inc.
11 * Redistribution and use in source and binary forms, with or without
12 * modification, are permitted provided that the following conditions
14 * 1. Redistributions of source code must retain the above copyright
15 * notice, this list of conditions and the following disclaimer.
16 * 2. Redistributions in binary form must reproduce the above copyright
17 * notice, this list of conditions and the following disclaimer in the
18 * documentation and/or other materials provided with the distribution.
19 * 4. Neither the name of the University nor the names of its contributors
20 * may be used to endorse or promote products derived from this software
21 * without specific prior written permission.
23 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
24 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
25 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
26 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
28 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
29 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
30 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
31 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
32 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
35 * @(#)in_pcb.c 8.4 (Berkeley) 5/24/95
38 #include <sys/cdefs.h>
39 __FBSDID("$FreeBSD$");
42 #include "opt_ipsec.h"
44 #include "opt_inet6.h"
45 #include "opt_pcbgroup.h"
47 #include <sys/param.h>
48 #include <sys/systm.h>
49 #include <sys/malloc.h>
51 #include <sys/callout.h>
52 #include <sys/domain.h>
53 #include <sys/protosw.h>
54 #include <sys/socket.h>
55 #include <sys/socketvar.h>
58 #include <sys/refcount.h>
60 #include <sys/kernel.h>
61 #include <sys/sysctl.h>
70 #include <net/if_types.h>
71 #include <net/route.h>
74 #if defined(INET) || defined(INET6)
75 #include <netinet/in.h>
76 #include <netinet/in_pcb.h>
77 #include <netinet/ip_var.h>
78 #include <netinet/tcp_var.h>
79 #include <netinet/udp.h>
80 #include <netinet/udp_var.h>
83 #include <netinet/in_var.h>
86 #include <netinet/ip6.h>
87 #include <netinet6/in6_pcb.h>
88 #include <netinet6/in6_var.h>
89 #include <netinet6/ip6_var.h>
94 #include <netipsec/ipsec.h>
95 #include <netipsec/key.h>
98 #include <security/mac/mac_framework.h>
100 static struct callout ipport_tick_callout;
103 * These configure the range of local port addresses assigned to
104 * "unspecified" outgoing connections/packets/whatever.
106 VNET_DEFINE(int, ipport_lowfirstauto) = IPPORT_RESERVED - 1; /* 1023 */
107 VNET_DEFINE(int, ipport_lowlastauto) = IPPORT_RESERVEDSTART; /* 600 */
108 VNET_DEFINE(int, ipport_firstauto) = IPPORT_EPHEMERALFIRST; /* 10000 */
109 VNET_DEFINE(int, ipport_lastauto) = IPPORT_EPHEMERALLAST; /* 65535 */
110 VNET_DEFINE(int, ipport_hifirstauto) = IPPORT_HIFIRSTAUTO; /* 49152 */
111 VNET_DEFINE(int, ipport_hilastauto) = IPPORT_HILASTAUTO; /* 65535 */
114 * Reserved ports accessible only to root. There are significant
115 * security considerations that must be accounted for when changing these,
116 * but the security benefits can be great. Please be careful.
118 VNET_DEFINE(int, ipport_reservedhigh) = IPPORT_RESERVED - 1; /* 1023 */
119 VNET_DEFINE(int, ipport_reservedlow);
121 /* Variables dealing with random ephemeral port allocation. */
122 VNET_DEFINE(int, ipport_randomized) = 1; /* user controlled via sysctl */
123 VNET_DEFINE(int, ipport_randomcps) = 10; /* user controlled via sysctl */
124 VNET_DEFINE(int, ipport_randomtime) = 45; /* user controlled via sysctl */
125 VNET_DEFINE(int, ipport_stoprandom); /* toggled by ipport_tick */
126 VNET_DEFINE(int, ipport_tcpallocs);
127 static VNET_DEFINE(int, ipport_tcplastcount);
129 #define V_ipport_tcplastcount VNET(ipport_tcplastcount)
131 static void in_pcbremlists(struct inpcb *inp);
133 static struct inpcb *in_pcblookup_hash_locked(struct inpcbinfo *pcbinfo,
134 struct in_addr faddr, u_int fport_arg,
135 struct in_addr laddr, u_int lport_arg,
136 int lookupflags, struct ifnet *ifp);
138 #define RANGECHK(var, min, max) \
139 if ((var) < (min)) { (var) = (min); } \
140 else if ((var) > (max)) { (var) = (max); }
143 sysctl_net_ipport_check(SYSCTL_HANDLER_ARGS)
148 error = vnet_sysctl_handle_int(oidp, arg1, arg2, req);
150 error = sysctl_handle_int(oidp, arg1, arg2, req);
153 RANGECHK(V_ipport_lowfirstauto, 1, IPPORT_RESERVED - 1);
154 RANGECHK(V_ipport_lowlastauto, 1, IPPORT_RESERVED - 1);
155 RANGECHK(V_ipport_firstauto, IPPORT_RESERVED, IPPORT_MAX);
156 RANGECHK(V_ipport_lastauto, IPPORT_RESERVED, IPPORT_MAX);
157 RANGECHK(V_ipport_hifirstauto, IPPORT_RESERVED, IPPORT_MAX);
158 RANGECHK(V_ipport_hilastauto, IPPORT_RESERVED, IPPORT_MAX);
165 static SYSCTL_NODE(_net_inet_ip, IPPROTO_IP, portrange, CTLFLAG_RW, 0,
168 SYSCTL_VNET_PROC(_net_inet_ip_portrange, OID_AUTO, lowfirst,
169 CTLTYPE_INT|CTLFLAG_RW, &VNET_NAME(ipport_lowfirstauto), 0,
170 &sysctl_net_ipport_check, "I", "");
171 SYSCTL_VNET_PROC(_net_inet_ip_portrange, OID_AUTO, lowlast,
172 CTLTYPE_INT|CTLFLAG_RW, &VNET_NAME(ipport_lowlastauto), 0,
173 &sysctl_net_ipport_check, "I", "");
174 SYSCTL_VNET_PROC(_net_inet_ip_portrange, OID_AUTO, first,
175 CTLTYPE_INT|CTLFLAG_RW, &VNET_NAME(ipport_firstauto), 0,
176 &sysctl_net_ipport_check, "I", "");
177 SYSCTL_VNET_PROC(_net_inet_ip_portrange, OID_AUTO, last,
178 CTLTYPE_INT|CTLFLAG_RW, &VNET_NAME(ipport_lastauto), 0,
179 &sysctl_net_ipport_check, "I", "");
180 SYSCTL_VNET_PROC(_net_inet_ip_portrange, OID_AUTO, hifirst,
181 CTLTYPE_INT|CTLFLAG_RW, &VNET_NAME(ipport_hifirstauto), 0,
182 &sysctl_net_ipport_check, "I", "");
183 SYSCTL_VNET_PROC(_net_inet_ip_portrange, OID_AUTO, hilast,
184 CTLTYPE_INT|CTLFLAG_RW, &VNET_NAME(ipport_hilastauto), 0,
185 &sysctl_net_ipport_check, "I", "");
186 SYSCTL_VNET_INT(_net_inet_ip_portrange, OID_AUTO, reservedhigh,
187 CTLFLAG_RW|CTLFLAG_SECURE, &VNET_NAME(ipport_reservedhigh), 0, "");
188 SYSCTL_VNET_INT(_net_inet_ip_portrange, OID_AUTO, reservedlow,
189 CTLFLAG_RW|CTLFLAG_SECURE, &VNET_NAME(ipport_reservedlow), 0, "");
190 SYSCTL_VNET_INT(_net_inet_ip_portrange, OID_AUTO, randomized, CTLFLAG_RW,
191 &VNET_NAME(ipport_randomized), 0, "Enable random port allocation");
192 SYSCTL_VNET_INT(_net_inet_ip_portrange, OID_AUTO, randomcps, CTLFLAG_RW,
193 &VNET_NAME(ipport_randomcps), 0, "Maximum number of random port "
194 "allocations before switching to a sequental one");
195 SYSCTL_VNET_INT(_net_inet_ip_portrange, OID_AUTO, randomtime, CTLFLAG_RW,
196 &VNET_NAME(ipport_randomtime), 0,
197 "Minimum time to keep sequental port "
198 "allocation before switching to a random one");
202 * in_pcb.c: manage the Protocol Control Blocks.
204 * NOTE: It is assumed that most of these functions will be called with
205 * the pcbinfo lock held, and often, the inpcb lock held, as these utility
206 * functions often modify hash chains or addresses in pcbs.
210 * Initialize an inpcbinfo -- we should be able to reduce the number of
214 in_pcbinfo_init(struct inpcbinfo *pcbinfo, const char *name,
215 struct inpcbhead *listhead, int hash_nelements, int porthash_nelements,
216 char *inpcbzone_name, uma_init inpcbzone_init, uma_fini inpcbzone_fini,
217 uint32_t inpcbzone_flags, u_int hashfields)
220 INP_INFO_LOCK_INIT(pcbinfo, name);
221 INP_HASH_LOCK_INIT(pcbinfo, "pcbinfohash"); /* XXXRW: argument? */
223 pcbinfo->ipi_vnet = curvnet;
225 pcbinfo->ipi_listhead = listhead;
226 LIST_INIT(pcbinfo->ipi_listhead);
227 pcbinfo->ipi_count = 0;
228 pcbinfo->ipi_hashbase = hashinit(hash_nelements, M_PCB,
229 &pcbinfo->ipi_hashmask);
230 pcbinfo->ipi_porthashbase = hashinit(porthash_nelements, M_PCB,
231 &pcbinfo->ipi_porthashmask);
233 in_pcbgroup_init(pcbinfo, hashfields, hash_nelements);
235 pcbinfo->ipi_zone = uma_zcreate(inpcbzone_name, sizeof(struct inpcb),
236 NULL, NULL, inpcbzone_init, inpcbzone_fini, UMA_ALIGN_PTR,
238 uma_zone_set_max(pcbinfo->ipi_zone, maxsockets);
239 uma_zone_set_warning(pcbinfo->ipi_zone,
240 "kern.ipc.maxsockets limit reached");
244 * Destroy an inpcbinfo.
247 in_pcbinfo_destroy(struct inpcbinfo *pcbinfo)
250 KASSERT(pcbinfo->ipi_count == 0,
251 ("%s: ipi_count = %u", __func__, pcbinfo->ipi_count));
253 hashdestroy(pcbinfo->ipi_hashbase, M_PCB, pcbinfo->ipi_hashmask);
254 hashdestroy(pcbinfo->ipi_porthashbase, M_PCB,
255 pcbinfo->ipi_porthashmask);
257 in_pcbgroup_destroy(pcbinfo);
259 uma_zdestroy(pcbinfo->ipi_zone);
260 INP_HASH_LOCK_DESTROY(pcbinfo);
261 INP_INFO_LOCK_DESTROY(pcbinfo);
265 * Allocate a PCB and associate it with the socket.
266 * On success return with the PCB locked.
269 in_pcballoc(struct socket *so, struct inpcbinfo *pcbinfo)
274 INP_INFO_WLOCK_ASSERT(pcbinfo);
276 inp = uma_zalloc(pcbinfo->ipi_zone, M_NOWAIT);
279 bzero(inp, inp_zero_size);
280 inp->inp_pcbinfo = pcbinfo;
281 inp->inp_socket = so;
282 inp->inp_cred = crhold(so->so_cred);
283 inp->inp_inc.inc_fibnum = so->so_fibnum;
285 error = mac_inpcb_init(inp, M_NOWAIT);
288 mac_inpcb_create(so, inp);
291 error = ipsec_init_policy(so, &inp->inp_sp);
294 mac_inpcb_destroy(inp);
300 if (INP_SOCKAF(so) == AF_INET6) {
301 inp->inp_vflag |= INP_IPV6PROTO;
303 inp->inp_flags |= IN6P_IPV6_V6ONLY;
306 LIST_INSERT_HEAD(pcbinfo->ipi_listhead, inp, inp_list);
307 pcbinfo->ipi_count++;
308 so->so_pcb = (caddr_t)inp;
310 if (V_ip6_auto_flowlabel)
311 inp->inp_flags |= IN6P_AUTOFLOWLABEL;
314 inp->inp_gencnt = ++pcbinfo->ipi_gencnt;
315 refcount_init(&inp->inp_refcount, 1); /* Reference from inpcbinfo */
316 #if defined(IPSEC) || defined(MAC)
319 crfree(inp->inp_cred);
320 uma_zfree(pcbinfo->ipi_zone, inp);
328 in_pcbbind(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred)
332 INP_WLOCK_ASSERT(inp);
333 INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo);
335 if (inp->inp_lport != 0 || inp->inp_laddr.s_addr != INADDR_ANY)
337 anonport = nam == NULL || ((struct sockaddr_in *)nam)->sin_port == 0;
338 error = in_pcbbind_setup(inp, nam, &inp->inp_laddr.s_addr,
339 &inp->inp_lport, cred);
342 if (in_pcbinshash(inp) != 0) {
343 inp->inp_laddr.s_addr = INADDR_ANY;
348 inp->inp_flags |= INP_ANONPORT;
353 #if defined(INET) || defined(INET6)
355 in_pcb_lport(struct inpcb *inp, struct in_addr *laddrp, u_short *lportp,
356 struct ucred *cred, int lookupflags)
358 struct inpcbinfo *pcbinfo;
359 struct inpcb *tmpinp;
360 unsigned short *lastport;
361 int count, dorandom, error;
362 u_short aux, first, last, lport;
364 struct in_addr laddr;
367 pcbinfo = inp->inp_pcbinfo;
370 * Because no actual state changes occur here, a global write lock on
371 * the pcbinfo isn't required.
373 INP_LOCK_ASSERT(inp);
374 INP_HASH_LOCK_ASSERT(pcbinfo);
376 if (inp->inp_flags & INP_HIGHPORT) {
377 first = V_ipport_hifirstauto; /* sysctl */
378 last = V_ipport_hilastauto;
379 lastport = &pcbinfo->ipi_lasthi;
380 } else if (inp->inp_flags & INP_LOWPORT) {
381 error = priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT, 0);
384 first = V_ipport_lowfirstauto; /* 1023 */
385 last = V_ipport_lowlastauto; /* 600 */
386 lastport = &pcbinfo->ipi_lastlow;
388 first = V_ipport_firstauto; /* sysctl */
389 last = V_ipport_lastauto;
390 lastport = &pcbinfo->ipi_lastport;
393 * For UDP, use random port allocation as long as the user
394 * allows it. For TCP (and as of yet unknown) connections,
395 * use random port allocation only if the user allows it AND
396 * ipport_tick() allows it.
398 if (V_ipport_randomized &&
399 (!V_ipport_stoprandom || pcbinfo == &V_udbinfo))
404 * It makes no sense to do random port allocation if
405 * we have the only port available.
409 /* Make sure to not include UDP packets in the count. */
410 if (pcbinfo != &V_udbinfo)
411 V_ipport_tcpallocs++;
413 * Instead of having two loops further down counting up or down
414 * make sure that first is always <= last and go with only one
415 * code path implementing all logic.
424 /* Make the compiler happy. */
426 if ((inp->inp_vflag & (INP_IPV4|INP_IPV6)) == INP_IPV4) {
427 KASSERT(laddrp != NULL, ("%s: laddrp NULL for v4 inp %p",
432 tmpinp = NULL; /* Make compiler happy. */
436 *lastport = first + (arc4random() % (last - first));
438 count = last - first;
441 if (count-- < 0) /* completely used? */
442 return (EADDRNOTAVAIL);
444 if (*lastport < first || *lastport > last)
446 lport = htons(*lastport);
449 if ((inp->inp_vflag & INP_IPV6) != 0)
450 tmpinp = in6_pcblookup_local(pcbinfo,
451 &inp->in6p_laddr, lport, lookupflags, cred);
453 #if defined(INET) && defined(INET6)
457 tmpinp = in_pcblookup_local(pcbinfo, laddr,
458 lport, lookupflags, cred);
460 } while (tmpinp != NULL);
463 if ((inp->inp_vflag & (INP_IPV4|INP_IPV6)) == INP_IPV4)
464 laddrp->s_addr = laddr.s_addr;
470 #endif /* INET || INET6 */
474 * Set up a bind operation on a PCB, performing port allocation
475 * as required, but do not actually modify the PCB. Callers can
476 * either complete the bind by setting inp_laddr/inp_lport and
477 * calling in_pcbinshash(), or they can just use the resulting
478 * port and address to authorise the sending of a once-off packet.
480 * On error, the values of *laddrp and *lportp are not changed.
483 in_pcbbind_setup(struct inpcb *inp, struct sockaddr *nam, in_addr_t *laddrp,
484 u_short *lportp, struct ucred *cred)
486 struct socket *so = inp->inp_socket;
487 struct sockaddr_in *sin;
488 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
489 struct in_addr laddr;
491 int lookupflags = 0, reuseport = (so->so_options & SO_REUSEPORT);
495 * No state changes, so read locks are sufficient here.
497 INP_LOCK_ASSERT(inp);
498 INP_HASH_LOCK_ASSERT(pcbinfo);
500 if (TAILQ_EMPTY(&V_in_ifaddrhead)) /* XXX broken! */
501 return (EADDRNOTAVAIL);
502 laddr.s_addr = *laddrp;
503 if (nam != NULL && laddr.s_addr != INADDR_ANY)
505 if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0)
506 lookupflags = INPLOOKUP_WILDCARD;
508 if ((error = prison_local_ip4(cred, &laddr)) != 0)
511 sin = (struct sockaddr_in *)nam;
512 if (nam->sa_len != sizeof (*sin))
516 * We should check the family, but old programs
517 * incorrectly fail to initialize it.
519 if (sin->sin_family != AF_INET)
520 return (EAFNOSUPPORT);
522 error = prison_local_ip4(cred, &sin->sin_addr);
525 if (sin->sin_port != *lportp) {
526 /* Don't allow the port to change. */
529 lport = sin->sin_port;
531 /* NB: lport is left as 0 if the port isn't being changed. */
532 if (IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) {
534 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
535 * allow complete duplication of binding if
536 * SO_REUSEPORT is set, or if SO_REUSEADDR is set
537 * and a multicast address is bound on both
538 * new and duplicated sockets.
540 if (so->so_options & SO_REUSEADDR)
541 reuseport = SO_REUSEADDR|SO_REUSEPORT;
542 } else if (sin->sin_addr.s_addr != INADDR_ANY) {
543 sin->sin_port = 0; /* yech... */
544 bzero(&sin->sin_zero, sizeof(sin->sin_zero));
546 * Is the address a local IP address?
547 * If INP_BINDANY is set, then the socket may be bound
548 * to any endpoint address, local or not.
550 if ((inp->inp_flags & INP_BINDANY) == 0 &&
551 ifa_ifwithaddr_check((struct sockaddr *)sin) == 0)
552 return (EADDRNOTAVAIL);
554 laddr = sin->sin_addr;
560 if (ntohs(lport) <= V_ipport_reservedhigh &&
561 ntohs(lport) >= V_ipport_reservedlow &&
562 priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT,
565 if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)) &&
566 priv_check_cred(inp->inp_cred,
567 PRIV_NETINET_REUSEPORT, 0) != 0) {
568 t = in_pcblookup_local(pcbinfo, sin->sin_addr,
569 lport, INPLOOKUP_WILDCARD, cred);
572 * This entire block sorely needs a rewrite.
575 ((t->inp_flags & INP_TIMEWAIT) == 0) &&
576 (so->so_type != SOCK_STREAM ||
577 ntohl(t->inp_faddr.s_addr) == INADDR_ANY) &&
578 (ntohl(sin->sin_addr.s_addr) != INADDR_ANY ||
579 ntohl(t->inp_laddr.s_addr) != INADDR_ANY ||
580 (t->inp_flags2 & INP_REUSEPORT) == 0) &&
581 (inp->inp_cred->cr_uid !=
582 t->inp_cred->cr_uid))
585 t = in_pcblookup_local(pcbinfo, sin->sin_addr,
586 lport, lookupflags, cred);
587 if (t && (t->inp_flags & INP_TIMEWAIT)) {
589 * XXXRW: If an incpb has had its timewait
590 * state recycled, we treat the address as
591 * being in use (for now). This is better
592 * than a panic, but not desirable.
596 (reuseport & tw->tw_so_options) == 0)
598 } else if (t && (reuseport == 0 ||
599 (t->inp_flags2 & INP_REUSEPORT) == 0)) {
601 if (ntohl(sin->sin_addr.s_addr) !=
603 ntohl(t->inp_laddr.s_addr) !=
605 (inp->inp_vflag & INP_IPV6PROTO) == 0 ||
606 (t->inp_vflag & INP_IPV6PROTO) == 0)
615 error = in_pcb_lport(inp, &laddr, &lport, cred, lookupflags);
620 *laddrp = laddr.s_addr;
626 * Connect from a socket to a specified address.
627 * Both address and port must be specified in argument sin.
628 * If don't have a local address for this socket yet,
632 in_pcbconnect_mbuf(struct inpcb *inp, struct sockaddr *nam,
633 struct ucred *cred, struct mbuf *m)
635 u_short lport, fport;
636 in_addr_t laddr, faddr;
639 INP_WLOCK_ASSERT(inp);
640 INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo);
642 lport = inp->inp_lport;
643 laddr = inp->inp_laddr.s_addr;
644 anonport = (lport == 0);
645 error = in_pcbconnect_setup(inp, nam, &laddr, &lport, &faddr, &fport,
650 /* Do the initial binding of the local address if required. */
651 if (inp->inp_laddr.s_addr == INADDR_ANY && inp->inp_lport == 0) {
652 inp->inp_lport = lport;
653 inp->inp_laddr.s_addr = laddr;
654 if (in_pcbinshash(inp) != 0) {
655 inp->inp_laddr.s_addr = INADDR_ANY;
661 /* Commit the remaining changes. */
662 inp->inp_lport = lport;
663 inp->inp_laddr.s_addr = laddr;
664 inp->inp_faddr.s_addr = faddr;
665 inp->inp_fport = fport;
666 in_pcbrehash_mbuf(inp, m);
669 inp->inp_flags |= INP_ANONPORT;
674 in_pcbconnect(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred)
677 return (in_pcbconnect_mbuf(inp, nam, cred, NULL));
681 * Do proper source address selection on an unbound socket in case
682 * of connect. Take jails into account as well.
685 in_pcbladdr(struct inpcb *inp, struct in_addr *faddr, struct in_addr *laddr,
690 struct sockaddr_in *sin;
694 KASSERT(laddr != NULL, ("%s: laddr NULL", __func__));
697 * Bypass source address selection and use the primary jail IP
700 if (cred != NULL && !prison_saddrsel_ip4(cred, laddr))
704 bzero(&sro, sizeof(sro));
706 sin = (struct sockaddr_in *)&sro.ro_dst;
707 sin->sin_family = AF_INET;
708 sin->sin_len = sizeof(struct sockaddr_in);
709 sin->sin_addr.s_addr = faddr->s_addr;
712 * If route is known our src addr is taken from the i/f,
715 * Find out route to destination.
717 if ((inp->inp_socket->so_options & SO_DONTROUTE) == 0)
718 in_rtalloc_ign(&sro, 0, inp->inp_inc.inc_fibnum);
721 * If we found a route, use the address corresponding to
722 * the outgoing interface.
724 * Otherwise assume faddr is reachable on a directly connected
725 * network and try to find a corresponding interface to take
726 * the source address from.
728 if (sro.ro_rt == NULL || sro.ro_rt->rt_ifp == NULL) {
729 struct in_ifaddr *ia;
732 ia = ifatoia(ifa_ifwithdstaddr((struct sockaddr *)sin));
734 ia = ifatoia(ifa_ifwithnet((struct sockaddr *)sin, 0));
740 if (cred == NULL || !prison_flag(cred, PR_IP4)) {
741 laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
742 ifa_free(&ia->ia_ifa);
747 ifa_free(&ia->ia_ifa);
750 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
753 if (sa->sa_family != AF_INET)
755 sin = (struct sockaddr_in *)sa;
756 if (prison_check_ip4(cred, &sin->sin_addr) == 0) {
757 ia = (struct in_ifaddr *)ifa;
762 laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
763 IF_ADDR_RUNLOCK(ifp);
766 IF_ADDR_RUNLOCK(ifp);
768 /* 3. As a last resort return the 'default' jail address. */
769 error = prison_get_ip4(cred, laddr);
774 * If the outgoing interface on the route found is not
775 * a loopback interface, use the address from that interface.
776 * In case of jails do those three steps:
777 * 1. check if the interface address belongs to the jail. If so use it.
778 * 2. check if we have any address on the outgoing interface
779 * belonging to this jail. If so use it.
780 * 3. as a last resort return the 'default' jail address.
782 if ((sro.ro_rt->rt_ifp->if_flags & IFF_LOOPBACK) == 0) {
783 struct in_ifaddr *ia;
786 /* If not jailed, use the default returned. */
787 if (cred == NULL || !prison_flag(cred, PR_IP4)) {
788 ia = (struct in_ifaddr *)sro.ro_rt->rt_ifa;
789 laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
794 /* 1. Check if the iface address belongs to the jail. */
795 sin = (struct sockaddr_in *)sro.ro_rt->rt_ifa->ifa_addr;
796 if (prison_check_ip4(cred, &sin->sin_addr) == 0) {
797 ia = (struct in_ifaddr *)sro.ro_rt->rt_ifa;
798 laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
803 * 2. Check if we have any address on the outgoing interface
804 * belonging to this jail.
807 ifp = sro.ro_rt->rt_ifp;
809 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
811 if (sa->sa_family != AF_INET)
813 sin = (struct sockaddr_in *)sa;
814 if (prison_check_ip4(cred, &sin->sin_addr) == 0) {
815 ia = (struct in_ifaddr *)ifa;
820 laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
821 IF_ADDR_RUNLOCK(ifp);
824 IF_ADDR_RUNLOCK(ifp);
826 /* 3. As a last resort return the 'default' jail address. */
827 error = prison_get_ip4(cred, laddr);
832 * The outgoing interface is marked with 'loopback net', so a route
833 * to ourselves is here.
834 * Try to find the interface of the destination address and then
835 * take the address from there. That interface is not necessarily
836 * a loopback interface.
837 * In case of jails, check that it is an address of the jail
838 * and if we cannot find, fall back to the 'default' jail address.
840 if ((sro.ro_rt->rt_ifp->if_flags & IFF_LOOPBACK) != 0) {
841 struct sockaddr_in sain;
842 struct in_ifaddr *ia;
844 bzero(&sain, sizeof(struct sockaddr_in));
845 sain.sin_family = AF_INET;
846 sain.sin_len = sizeof(struct sockaddr_in);
847 sain.sin_addr.s_addr = faddr->s_addr;
849 ia = ifatoia(ifa_ifwithdstaddr(sintosa(&sain)));
851 ia = ifatoia(ifa_ifwithnet(sintosa(&sain), 0));
853 ia = ifatoia(ifa_ifwithaddr(sintosa(&sain)));
855 if (cred == NULL || !prison_flag(cred, PR_IP4)) {
860 laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
861 ifa_free(&ia->ia_ifa);
870 ifa_free(&ia->ia_ifa);
873 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
876 if (sa->sa_family != AF_INET)
878 sin = (struct sockaddr_in *)sa;
879 if (prison_check_ip4(cred,
880 &sin->sin_addr) == 0) {
881 ia = (struct in_ifaddr *)ifa;
886 laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
887 IF_ADDR_RUNLOCK(ifp);
890 IF_ADDR_RUNLOCK(ifp);
893 /* 3. As a last resort return the 'default' jail address. */
894 error = prison_get_ip4(cred, laddr);
899 if (sro.ro_rt != NULL)
905 * Set up for a connect from a socket to the specified address.
906 * On entry, *laddrp and *lportp should contain the current local
907 * address and port for the PCB; these are updated to the values
908 * that should be placed in inp_laddr and inp_lport to complete
911 * On success, *faddrp and *fportp will be set to the remote address
912 * and port. These are not updated in the error case.
914 * If the operation fails because the connection already exists,
915 * *oinpp will be set to the PCB of that connection so that the
916 * caller can decide to override it. In all other cases, *oinpp
920 in_pcbconnect_setup(struct inpcb *inp, struct sockaddr *nam,
921 in_addr_t *laddrp, u_short *lportp, in_addr_t *faddrp, u_short *fportp,
922 struct inpcb **oinpp, struct ucred *cred)
924 struct sockaddr_in *sin = (struct sockaddr_in *)nam;
925 struct in_ifaddr *ia;
927 struct in_addr laddr, faddr;
928 u_short lport, fport;
932 * Because a global state change doesn't actually occur here, a read
933 * lock is sufficient.
935 INP_LOCK_ASSERT(inp);
936 INP_HASH_LOCK_ASSERT(inp->inp_pcbinfo);
940 if (nam->sa_len != sizeof (*sin))
942 if (sin->sin_family != AF_INET)
943 return (EAFNOSUPPORT);
944 if (sin->sin_port == 0)
945 return (EADDRNOTAVAIL);
946 laddr.s_addr = *laddrp;
948 faddr = sin->sin_addr;
949 fport = sin->sin_port;
951 if (!TAILQ_EMPTY(&V_in_ifaddrhead)) {
953 * If the destination address is INADDR_ANY,
954 * use the primary local address.
955 * If the supplied address is INADDR_BROADCAST,
956 * and the primary interface supports broadcast,
957 * choose the broadcast address for that interface.
959 if (faddr.s_addr == INADDR_ANY) {
962 IA_SIN(TAILQ_FIRST(&V_in_ifaddrhead))->sin_addr;
965 (error = prison_get_ip4(cred, &faddr)) != 0)
967 } else if (faddr.s_addr == (u_long)INADDR_BROADCAST) {
969 if (TAILQ_FIRST(&V_in_ifaddrhead)->ia_ifp->if_flags &
971 faddr = satosin(&TAILQ_FIRST(
972 &V_in_ifaddrhead)->ia_broadaddr)->sin_addr;
976 if (laddr.s_addr == INADDR_ANY) {
977 error = in_pcbladdr(inp, &faddr, &laddr, cred);
979 * If the destination address is multicast and an outgoing
980 * interface has been set as a multicast option, prefer the
981 * address of that interface as our source address.
983 if (IN_MULTICAST(ntohl(faddr.s_addr)) &&
984 inp->inp_moptions != NULL) {
985 struct ip_moptions *imo;
988 imo = inp->inp_moptions;
989 if (imo->imo_multicast_ifp != NULL) {
990 ifp = imo->imo_multicast_ifp;
992 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
993 if ((ia->ia_ifp == ifp) &&
995 prison_check_ip4(cred,
996 &ia->ia_addr.sin_addr) == 0))
1000 error = EADDRNOTAVAIL;
1002 laddr = ia->ia_addr.sin_addr;
1005 IN_IFADDR_RUNLOCK();
1011 oinp = in_pcblookup_hash_locked(inp->inp_pcbinfo, faddr, fport,
1012 laddr, lport, 0, NULL);
1016 return (EADDRINUSE);
1019 error = in_pcbbind_setup(inp, NULL, &laddr.s_addr, &lport,
1024 *laddrp = laddr.s_addr;
1026 *faddrp = faddr.s_addr;
1032 in_pcbdisconnect(struct inpcb *inp)
1035 INP_WLOCK_ASSERT(inp);
1036 INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo);
1038 inp->inp_faddr.s_addr = INADDR_ANY;
1045 * in_pcbdetach() is responsibe for disassociating a socket from an inpcb.
1046 * For most protocols, this will be invoked immediately prior to calling
1047 * in_pcbfree(). However, with TCP the inpcb may significantly outlive the
1048 * socket, in which case in_pcbfree() is deferred.
1051 in_pcbdetach(struct inpcb *inp)
1054 KASSERT(inp->inp_socket != NULL, ("%s: inp_socket == NULL", __func__));
1056 inp->inp_socket->so_pcb = NULL;
1057 inp->inp_socket = NULL;
1061 * in_pcbref() bumps the reference count on an inpcb in order to maintain
1062 * stability of an inpcb pointer despite the inpcb lock being released. This
1063 * is used in TCP when the inpcbinfo lock needs to be acquired or upgraded,
1064 * but where the inpcb lock may already held, or when acquiring a reference
1067 * in_pcbref() should be used only to provide brief memory stability, and
1068 * must always be followed by a call to INP_WLOCK() and in_pcbrele() to
1069 * garbage collect the inpcb if it has been in_pcbfree()'d from another
1070 * context. Until in_pcbrele() has returned that the inpcb is still valid,
1071 * lock and rele are the *only* safe operations that may be performed on the
1074 * While the inpcb will not be freed, releasing the inpcb lock means that the
1075 * connection's state may change, so the caller should be careful to
1076 * revalidate any cached state on reacquiring the lock. Drop the reference
1077 * using in_pcbrele().
1080 in_pcbref(struct inpcb *inp)
1083 KASSERT(inp->inp_refcount > 0, ("%s: refcount 0", __func__));
1085 refcount_acquire(&inp->inp_refcount);
1089 * Drop a refcount on an inpcb elevated using in_pcbref(); because a call to
1090 * in_pcbfree() may have been made between in_pcbref() and in_pcbrele(), we
1091 * return a flag indicating whether or not the inpcb remains valid. If it is
1092 * valid, we return with the inpcb lock held.
1094 * Notice that, unlike in_pcbref(), the inpcb lock must be held to drop a
1095 * reference on an inpcb. Historically more work was done here (actually, in
1096 * in_pcbfree_internal()) but has been moved to in_pcbfree() to avoid the
1097 * need for the pcbinfo lock in in_pcbrele(). Deferring the free is entirely
1098 * about memory stability (and continued use of the write lock).
1101 in_pcbrele_rlocked(struct inpcb *inp)
1103 struct inpcbinfo *pcbinfo;
1105 KASSERT(inp->inp_refcount > 0, ("%s: refcount 0", __func__));
1107 INP_RLOCK_ASSERT(inp);
1109 if (refcount_release(&inp->inp_refcount) == 0) {
1111 * If the inpcb has been freed, let the caller know, even if
1112 * this isn't the last reference.
1114 if (inp->inp_flags2 & INP_FREED) {
1121 KASSERT(inp->inp_socket == NULL, ("%s: inp_socket != NULL", __func__));
1124 pcbinfo = inp->inp_pcbinfo;
1125 uma_zfree(pcbinfo->ipi_zone, inp);
1130 in_pcbrele_wlocked(struct inpcb *inp)
1132 struct inpcbinfo *pcbinfo;
1134 KASSERT(inp->inp_refcount > 0, ("%s: refcount 0", __func__));
1136 INP_WLOCK_ASSERT(inp);
1138 if (refcount_release(&inp->inp_refcount) == 0)
1141 KASSERT(inp->inp_socket == NULL, ("%s: inp_socket != NULL", __func__));
1144 pcbinfo = inp->inp_pcbinfo;
1145 uma_zfree(pcbinfo->ipi_zone, inp);
1150 * Temporary wrapper.
1153 in_pcbrele(struct inpcb *inp)
1156 return (in_pcbrele_wlocked(inp));
1160 * Unconditionally schedule an inpcb to be freed by decrementing its
1161 * reference count, which should occur only after the inpcb has been detached
1162 * from its socket. If another thread holds a temporary reference (acquired
1163 * using in_pcbref()) then the free is deferred until that reference is
1164 * released using in_pcbrele(), but the inpcb is still unlocked. Almost all
1165 * work, including removal from global lists, is done in this context, where
1166 * the pcbinfo lock is held.
1169 in_pcbfree(struct inpcb *inp)
1171 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1173 KASSERT(inp->inp_socket == NULL, ("%s: inp_socket != NULL", __func__));
1175 INP_INFO_WLOCK_ASSERT(pcbinfo);
1176 INP_WLOCK_ASSERT(inp);
1178 /* XXXRW: Do as much as possible here. */
1180 if (inp->inp_sp != NULL)
1181 ipsec_delete_pcbpolicy(inp);
1183 inp->inp_gencnt = ++pcbinfo->ipi_gencnt;
1184 in_pcbremlists(inp);
1186 if (inp->inp_vflag & INP_IPV6PROTO) {
1187 ip6_freepcbopts(inp->in6p_outputopts);
1188 if (inp->in6p_moptions != NULL)
1189 ip6_freemoptions(inp->in6p_moptions);
1192 if (inp->inp_options)
1193 (void)m_free(inp->inp_options);
1195 if (inp->inp_moptions != NULL)
1196 inp_freemoptions(inp->inp_moptions);
1199 inp->inp_flags2 |= INP_FREED;
1200 crfree(inp->inp_cred);
1202 mac_inpcb_destroy(inp);
1204 if (!in_pcbrele_wlocked(inp))
1209 * in_pcbdrop() removes an inpcb from hashed lists, releasing its address and
1210 * port reservation, and preventing it from being returned by inpcb lookups.
1212 * It is used by TCP to mark an inpcb as unused and avoid future packet
1213 * delivery or event notification when a socket remains open but TCP has
1214 * closed. This might occur as a result of a shutdown()-initiated TCP close
1215 * or a RST on the wire, and allows the port binding to be reused while still
1216 * maintaining the invariant that so_pcb always points to a valid inpcb until
1219 * XXXRW: Possibly in_pcbdrop() should also prevent future notifications by
1220 * in_pcbnotifyall() and in_pcbpurgeif0()?
1223 in_pcbdrop(struct inpcb *inp)
1226 INP_WLOCK_ASSERT(inp);
1229 * XXXRW: Possibly we should protect the setting of INP_DROPPED with
1232 inp->inp_flags |= INP_DROPPED;
1233 if (inp->inp_flags & INP_INHASHLIST) {
1234 struct inpcbport *phd = inp->inp_phd;
1236 INP_HASH_WLOCK(inp->inp_pcbinfo);
1237 LIST_REMOVE(inp, inp_hash);
1238 LIST_REMOVE(inp, inp_portlist);
1239 if (LIST_FIRST(&phd->phd_pcblist) == NULL) {
1240 LIST_REMOVE(phd, phd_hash);
1243 INP_HASH_WUNLOCK(inp->inp_pcbinfo);
1244 inp->inp_flags &= ~INP_INHASHLIST;
1246 in_pcbgroup_remove(inp);
1253 * Common routines to return the socket addresses associated with inpcbs.
1256 in_sockaddr(in_port_t port, struct in_addr *addr_p)
1258 struct sockaddr_in *sin;
1260 sin = malloc(sizeof *sin, M_SONAME,
1262 sin->sin_family = AF_INET;
1263 sin->sin_len = sizeof(*sin);
1264 sin->sin_addr = *addr_p;
1265 sin->sin_port = port;
1267 return (struct sockaddr *)sin;
1271 in_getsockaddr(struct socket *so, struct sockaddr **nam)
1274 struct in_addr addr;
1277 inp = sotoinpcb(so);
1278 KASSERT(inp != NULL, ("in_getsockaddr: inp == NULL"));
1281 port = inp->inp_lport;
1282 addr = inp->inp_laddr;
1285 *nam = in_sockaddr(port, &addr);
1290 in_getpeeraddr(struct socket *so, struct sockaddr **nam)
1293 struct in_addr addr;
1296 inp = sotoinpcb(so);
1297 KASSERT(inp != NULL, ("in_getpeeraddr: inp == NULL"));
1300 port = inp->inp_fport;
1301 addr = inp->inp_faddr;
1304 *nam = in_sockaddr(port, &addr);
1309 in_pcbnotifyall(struct inpcbinfo *pcbinfo, struct in_addr faddr, int errno,
1310 struct inpcb *(*notify)(struct inpcb *, int))
1312 struct inpcb *inp, *inp_temp;
1314 INP_INFO_WLOCK(pcbinfo);
1315 LIST_FOREACH_SAFE(inp, pcbinfo->ipi_listhead, inp_list, inp_temp) {
1318 if ((inp->inp_vflag & INP_IPV4) == 0) {
1323 if (inp->inp_faddr.s_addr != faddr.s_addr ||
1324 inp->inp_socket == NULL) {
1328 if ((*notify)(inp, errno))
1331 INP_INFO_WUNLOCK(pcbinfo);
1335 in_pcbpurgeif0(struct inpcbinfo *pcbinfo, struct ifnet *ifp)
1338 struct ip_moptions *imo;
1341 INP_INFO_RLOCK(pcbinfo);
1342 LIST_FOREACH(inp, pcbinfo->ipi_listhead, inp_list) {
1344 imo = inp->inp_moptions;
1345 if ((inp->inp_vflag & INP_IPV4) &&
1348 * Unselect the outgoing interface if it is being
1351 if (imo->imo_multicast_ifp == ifp)
1352 imo->imo_multicast_ifp = NULL;
1355 * Drop multicast group membership if we joined
1356 * through the interface being detached.
1358 for (i = 0, gap = 0; i < imo->imo_num_memberships;
1360 if (imo->imo_membership[i]->inm_ifp == ifp) {
1361 in_delmulti(imo->imo_membership[i]);
1363 } else if (gap != 0)
1364 imo->imo_membership[i - gap] =
1365 imo->imo_membership[i];
1367 imo->imo_num_memberships -= gap;
1371 INP_INFO_RUNLOCK(pcbinfo);
1375 * Lookup a PCB based on the local address and port. Caller must hold the
1376 * hash lock. No inpcb locks or references are acquired.
1378 #define INP_LOOKUP_MAPPED_PCB_COST 3
1380 in_pcblookup_local(struct inpcbinfo *pcbinfo, struct in_addr laddr,
1381 u_short lport, int lookupflags, struct ucred *cred)
1385 int matchwild = 3 + INP_LOOKUP_MAPPED_PCB_COST;
1391 KASSERT((lookupflags & ~(INPLOOKUP_WILDCARD)) == 0,
1392 ("%s: invalid lookup flags %d", __func__, lookupflags));
1394 INP_HASH_LOCK_ASSERT(pcbinfo);
1396 if ((lookupflags & INPLOOKUP_WILDCARD) == 0) {
1397 struct inpcbhead *head;
1399 * Look for an unconnected (wildcard foreign addr) PCB that
1400 * matches the local address and port we're looking for.
1402 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport,
1403 0, pcbinfo->ipi_hashmask)];
1404 LIST_FOREACH(inp, head, inp_hash) {
1406 /* XXX inp locking */
1407 if ((inp->inp_vflag & INP_IPV4) == 0)
1410 if (inp->inp_faddr.s_addr == INADDR_ANY &&
1411 inp->inp_laddr.s_addr == laddr.s_addr &&
1412 inp->inp_lport == lport) {
1417 prison_equal_ip4(cred->cr_prison,
1418 inp->inp_cred->cr_prison))
1427 struct inpcbporthead *porthash;
1428 struct inpcbport *phd;
1429 struct inpcb *match = NULL;
1431 * Best fit PCB lookup.
1433 * First see if this local port is in use by looking on the
1436 porthash = &pcbinfo->ipi_porthashbase[INP_PCBPORTHASH(lport,
1437 pcbinfo->ipi_porthashmask)];
1438 LIST_FOREACH(phd, porthash, phd_hash) {
1439 if (phd->phd_port == lport)
1444 * Port is in use by one or more PCBs. Look for best
1447 LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) {
1450 !prison_equal_ip4(inp->inp_cred->cr_prison,
1454 /* XXX inp locking */
1455 if ((inp->inp_vflag & INP_IPV4) == 0)
1458 * We never select the PCB that has
1459 * INP_IPV6 flag and is bound to :: if
1460 * we have another PCB which is bound
1461 * to 0.0.0.0. If a PCB has the
1462 * INP_IPV6 flag, then we set its cost
1463 * higher than IPv4 only PCBs.
1465 * Note that the case only happens
1466 * when a socket is bound to ::, under
1467 * the condition that the use of the
1468 * mapped address is allowed.
1470 if ((inp->inp_vflag & INP_IPV6) != 0)
1471 wildcard += INP_LOOKUP_MAPPED_PCB_COST;
1473 if (inp->inp_faddr.s_addr != INADDR_ANY)
1475 if (inp->inp_laddr.s_addr != INADDR_ANY) {
1476 if (laddr.s_addr == INADDR_ANY)
1478 else if (inp->inp_laddr.s_addr != laddr.s_addr)
1481 if (laddr.s_addr != INADDR_ANY)
1484 if (wildcard < matchwild) {
1486 matchwild = wildcard;
1495 #undef INP_LOOKUP_MAPPED_PCB_COST
1499 * Lookup PCB in hash list, using pcbgroup tables.
1501 static struct inpcb *
1502 in_pcblookup_group(struct inpcbinfo *pcbinfo, struct inpcbgroup *pcbgroup,
1503 struct in_addr faddr, u_int fport_arg, struct in_addr laddr,
1504 u_int lport_arg, int lookupflags, struct ifnet *ifp)
1506 struct inpcbhead *head;
1507 struct inpcb *inp, *tmpinp;
1508 u_short fport = fport_arg, lport = lport_arg;
1511 * First look for an exact match.
1514 INP_GROUP_LOCK(pcbgroup);
1515 head = &pcbgroup->ipg_hashbase[INP_PCBHASH(faddr.s_addr, lport, fport,
1516 pcbgroup->ipg_hashmask)];
1517 LIST_FOREACH(inp, head, inp_pcbgrouphash) {
1519 /* XXX inp locking */
1520 if ((inp->inp_vflag & INP_IPV4) == 0)
1523 if (inp->inp_faddr.s_addr == faddr.s_addr &&
1524 inp->inp_laddr.s_addr == laddr.s_addr &&
1525 inp->inp_fport == fport &&
1526 inp->inp_lport == lport) {
1528 * XXX We should be able to directly return
1529 * the inp here, without any checks.
1530 * Well unless both bound with SO_REUSEPORT?
1532 if (prison_flag(inp->inp_cred, PR_IP4))
1538 if (tmpinp != NULL) {
1544 * Then look for a wildcard match, if requested.
1546 if ((lookupflags & INPLOOKUP_WILDCARD) != 0) {
1547 struct inpcb *local_wild = NULL, *local_exact = NULL;
1549 struct inpcb *local_wild_mapped = NULL;
1551 struct inpcb *jail_wild = NULL;
1552 struct inpcbhead *head;
1556 * Order of socket selection - we always prefer jails.
1557 * 1. jailed, non-wild.
1559 * 3. non-jailed, non-wild.
1560 * 4. non-jailed, wild.
1562 head = &pcbinfo->ipi_wildbase[INP_PCBHASH(INADDR_ANY, lport,
1563 0, pcbinfo->ipi_wildmask)];
1564 LIST_FOREACH(inp, head, inp_pcbgroup_wild) {
1566 /* XXX inp locking */
1567 if ((inp->inp_vflag & INP_IPV4) == 0)
1570 if (inp->inp_faddr.s_addr != INADDR_ANY ||
1571 inp->inp_lport != lport)
1574 /* XXX inp locking */
1575 if (ifp && ifp->if_type == IFT_FAITH &&
1576 (inp->inp_flags & INP_FAITH) == 0)
1579 injail = prison_flag(inp->inp_cred, PR_IP4);
1581 if (prison_check_ip4(inp->inp_cred,
1585 if (local_exact != NULL)
1589 if (inp->inp_laddr.s_addr == laddr.s_addr) {
1594 } else if (inp->inp_laddr.s_addr == INADDR_ANY) {
1596 /* XXX inp locking, NULL check */
1597 if (inp->inp_vflag & INP_IPV6PROTO)
1598 local_wild_mapped = inp;
1606 } /* LIST_FOREACH */
1614 inp = local_wild_mapped;
1618 } /* if (lookupflags & INPLOOKUP_WILDCARD) */
1619 INP_GROUP_UNLOCK(pcbgroup);
1624 INP_GROUP_UNLOCK(pcbgroup);
1625 if (lookupflags & INPLOOKUP_WLOCKPCB) {
1627 if (in_pcbrele_wlocked(inp))
1629 } else if (lookupflags & INPLOOKUP_RLOCKPCB) {
1631 if (in_pcbrele_rlocked(inp))
1634 panic("%s: locking bug", __func__);
1637 #endif /* PCBGROUP */
1640 * Lookup PCB in hash list, using pcbinfo tables. This variation assumes
1641 * that the caller has locked the hash list, and will not perform any further
1642 * locking or reference operations on either the hash list or the connection.
1644 static struct inpcb *
1645 in_pcblookup_hash_locked(struct inpcbinfo *pcbinfo, struct in_addr faddr,
1646 u_int fport_arg, struct in_addr laddr, u_int lport_arg, int lookupflags,
1649 struct inpcbhead *head;
1650 struct inpcb *inp, *tmpinp;
1651 u_short fport = fport_arg, lport = lport_arg;
1653 KASSERT((lookupflags & ~(INPLOOKUP_WILDCARD)) == 0,
1654 ("%s: invalid lookup flags %d", __func__, lookupflags));
1656 INP_HASH_LOCK_ASSERT(pcbinfo);
1659 * First look for an exact match.
1662 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(faddr.s_addr, lport, fport,
1663 pcbinfo->ipi_hashmask)];
1664 LIST_FOREACH(inp, head, inp_hash) {
1666 /* XXX inp locking */
1667 if ((inp->inp_vflag & INP_IPV4) == 0)
1670 if (inp->inp_faddr.s_addr == faddr.s_addr &&
1671 inp->inp_laddr.s_addr == laddr.s_addr &&
1672 inp->inp_fport == fport &&
1673 inp->inp_lport == lport) {
1675 * XXX We should be able to directly return
1676 * the inp here, without any checks.
1677 * Well unless both bound with SO_REUSEPORT?
1679 if (prison_flag(inp->inp_cred, PR_IP4))
1689 * Then look for a wildcard match, if requested.
1691 if ((lookupflags & INPLOOKUP_WILDCARD) != 0) {
1692 struct inpcb *local_wild = NULL, *local_exact = NULL;
1694 struct inpcb *local_wild_mapped = NULL;
1696 struct inpcb *jail_wild = NULL;
1700 * Order of socket selection - we always prefer jails.
1701 * 1. jailed, non-wild.
1703 * 3. non-jailed, non-wild.
1704 * 4. non-jailed, wild.
1707 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport,
1708 0, pcbinfo->ipi_hashmask)];
1709 LIST_FOREACH(inp, head, inp_hash) {
1711 /* XXX inp locking */
1712 if ((inp->inp_vflag & INP_IPV4) == 0)
1715 if (inp->inp_faddr.s_addr != INADDR_ANY ||
1716 inp->inp_lport != lport)
1719 /* XXX inp locking */
1720 if (ifp && ifp->if_type == IFT_FAITH &&
1721 (inp->inp_flags & INP_FAITH) == 0)
1724 injail = prison_flag(inp->inp_cred, PR_IP4);
1726 if (prison_check_ip4(inp->inp_cred,
1730 if (local_exact != NULL)
1734 if (inp->inp_laddr.s_addr == laddr.s_addr) {
1739 } else if (inp->inp_laddr.s_addr == INADDR_ANY) {
1741 /* XXX inp locking, NULL check */
1742 if (inp->inp_vflag & INP_IPV6PROTO)
1743 local_wild_mapped = inp;
1751 } /* LIST_FOREACH */
1752 if (jail_wild != NULL)
1754 if (local_exact != NULL)
1755 return (local_exact);
1756 if (local_wild != NULL)
1757 return (local_wild);
1759 if (local_wild_mapped != NULL)
1760 return (local_wild_mapped);
1762 } /* if ((lookupflags & INPLOOKUP_WILDCARD) != 0) */
1768 * Lookup PCB in hash list, using pcbinfo tables. This variation locks the
1769 * hash list lock, and will return the inpcb locked (i.e., requires
1770 * INPLOOKUP_LOCKPCB).
1772 static struct inpcb *
1773 in_pcblookup_hash(struct inpcbinfo *pcbinfo, struct in_addr faddr,
1774 u_int fport, struct in_addr laddr, u_int lport, int lookupflags,
1779 INP_HASH_RLOCK(pcbinfo);
1780 inp = in_pcblookup_hash_locked(pcbinfo, faddr, fport, laddr, lport,
1781 (lookupflags & ~(INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)), ifp);
1784 INP_HASH_RUNLOCK(pcbinfo);
1785 if (lookupflags & INPLOOKUP_WLOCKPCB) {
1787 if (in_pcbrele_wlocked(inp))
1789 } else if (lookupflags & INPLOOKUP_RLOCKPCB) {
1791 if (in_pcbrele_rlocked(inp))
1794 panic("%s: locking bug", __func__);
1796 INP_HASH_RUNLOCK(pcbinfo);
1801 * Public inpcb lookup routines, accepting a 4-tuple, and optionally, an mbuf
1802 * from which a pre-calculated hash value may be extracted.
1804 * Possibly more of this logic should be in in_pcbgroup.c.
1807 in_pcblookup(struct inpcbinfo *pcbinfo, struct in_addr faddr, u_int fport,
1808 struct in_addr laddr, u_int lport, int lookupflags, struct ifnet *ifp)
1810 #if defined(PCBGROUP)
1811 struct inpcbgroup *pcbgroup;
1814 KASSERT((lookupflags & ~INPLOOKUP_MASK) == 0,
1815 ("%s: invalid lookup flags %d", __func__, lookupflags));
1816 KASSERT((lookupflags & (INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)) != 0,
1817 ("%s: LOCKPCB not set", __func__));
1819 #if defined(PCBGROUP)
1820 if (in_pcbgroup_enabled(pcbinfo)) {
1821 pcbgroup = in_pcbgroup_bytuple(pcbinfo, laddr, lport, faddr,
1823 return (in_pcblookup_group(pcbinfo, pcbgroup, faddr, fport,
1824 laddr, lport, lookupflags, ifp));
1827 return (in_pcblookup_hash(pcbinfo, faddr, fport, laddr, lport,
1832 in_pcblookup_mbuf(struct inpcbinfo *pcbinfo, struct in_addr faddr,
1833 u_int fport, struct in_addr laddr, u_int lport, int lookupflags,
1834 struct ifnet *ifp, struct mbuf *m)
1837 struct inpcbgroup *pcbgroup;
1840 KASSERT((lookupflags & ~INPLOOKUP_MASK) == 0,
1841 ("%s: invalid lookup flags %d", __func__, lookupflags));
1842 KASSERT((lookupflags & (INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)) != 0,
1843 ("%s: LOCKPCB not set", __func__));
1846 if (in_pcbgroup_enabled(pcbinfo)) {
1847 pcbgroup = in_pcbgroup_byhash(pcbinfo, M_HASHTYPE_GET(m),
1848 m->m_pkthdr.flowid);
1849 if (pcbgroup != NULL)
1850 return (in_pcblookup_group(pcbinfo, pcbgroup, faddr,
1851 fport, laddr, lport, lookupflags, ifp));
1852 pcbgroup = in_pcbgroup_bytuple(pcbinfo, laddr, lport, faddr,
1854 return (in_pcblookup_group(pcbinfo, pcbgroup, faddr, fport,
1855 laddr, lport, lookupflags, ifp));
1858 return (in_pcblookup_hash(pcbinfo, faddr, fport, laddr, lport,
1864 * Insert PCB onto various hash lists.
1867 in_pcbinshash_internal(struct inpcb *inp, int do_pcbgroup_update)
1869 struct inpcbhead *pcbhash;
1870 struct inpcbporthead *pcbporthash;
1871 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1872 struct inpcbport *phd;
1873 u_int32_t hashkey_faddr;
1875 INP_WLOCK_ASSERT(inp);
1876 INP_HASH_WLOCK_ASSERT(pcbinfo);
1878 KASSERT((inp->inp_flags & INP_INHASHLIST) == 0,
1879 ("in_pcbinshash: INP_INHASHLIST"));
1882 if (inp->inp_vflag & INP_IPV6)
1883 hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX */;
1886 hashkey_faddr = inp->inp_faddr.s_addr;
1888 pcbhash = &pcbinfo->ipi_hashbase[INP_PCBHASH(hashkey_faddr,
1889 inp->inp_lport, inp->inp_fport, pcbinfo->ipi_hashmask)];
1891 pcbporthash = &pcbinfo->ipi_porthashbase[
1892 INP_PCBPORTHASH(inp->inp_lport, pcbinfo->ipi_porthashmask)];
1895 * Go through port list and look for a head for this lport.
1897 LIST_FOREACH(phd, pcbporthash, phd_hash) {
1898 if (phd->phd_port == inp->inp_lport)
1902 * If none exists, malloc one and tack it on.
1905 phd = malloc(sizeof(struct inpcbport), M_PCB, M_NOWAIT);
1907 return (ENOBUFS); /* XXX */
1909 phd->phd_port = inp->inp_lport;
1910 LIST_INIT(&phd->phd_pcblist);
1911 LIST_INSERT_HEAD(pcbporthash, phd, phd_hash);
1914 LIST_INSERT_HEAD(&phd->phd_pcblist, inp, inp_portlist);
1915 LIST_INSERT_HEAD(pcbhash, inp, inp_hash);
1916 inp->inp_flags |= INP_INHASHLIST;
1918 if (do_pcbgroup_update)
1919 in_pcbgroup_update(inp);
1925 * For now, there are two public interfaces to insert an inpcb into the hash
1926 * lists -- one that does update pcbgroups, and one that doesn't. The latter
1927 * is used only in the TCP syncache, where in_pcbinshash is called before the
1928 * full 4-tuple is set for the inpcb, and we don't want to install in the
1929 * pcbgroup until later.
1931 * XXXRW: This seems like a misfeature. in_pcbinshash should always update
1932 * connection groups, and partially initialised inpcbs should not be exposed
1933 * to either reservation hash tables or pcbgroups.
1936 in_pcbinshash(struct inpcb *inp)
1939 return (in_pcbinshash_internal(inp, 1));
1943 in_pcbinshash_nopcbgroup(struct inpcb *inp)
1946 return (in_pcbinshash_internal(inp, 0));
1950 * Move PCB to the proper hash bucket when { faddr, fport } have been
1951 * changed. NOTE: This does not handle the case of the lport changing (the
1952 * hashed port list would have to be updated as well), so the lport must
1953 * not change after in_pcbinshash() has been called.
1956 in_pcbrehash_mbuf(struct inpcb *inp, struct mbuf *m)
1958 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1959 struct inpcbhead *head;
1960 u_int32_t hashkey_faddr;
1962 INP_WLOCK_ASSERT(inp);
1963 INP_HASH_WLOCK_ASSERT(pcbinfo);
1965 KASSERT(inp->inp_flags & INP_INHASHLIST,
1966 ("in_pcbrehash: !INP_INHASHLIST"));
1969 if (inp->inp_vflag & INP_IPV6)
1970 hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX */;
1973 hashkey_faddr = inp->inp_faddr.s_addr;
1975 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(hashkey_faddr,
1976 inp->inp_lport, inp->inp_fport, pcbinfo->ipi_hashmask)];
1978 LIST_REMOVE(inp, inp_hash);
1979 LIST_INSERT_HEAD(head, inp, inp_hash);
1983 in_pcbgroup_update_mbuf(inp, m);
1985 in_pcbgroup_update(inp);
1990 in_pcbrehash(struct inpcb *inp)
1993 in_pcbrehash_mbuf(inp, NULL);
1997 * Remove PCB from various lists.
2000 in_pcbremlists(struct inpcb *inp)
2002 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
2004 INP_INFO_WLOCK_ASSERT(pcbinfo);
2005 INP_WLOCK_ASSERT(inp);
2007 inp->inp_gencnt = ++pcbinfo->ipi_gencnt;
2008 if (inp->inp_flags & INP_INHASHLIST) {
2009 struct inpcbport *phd = inp->inp_phd;
2011 INP_HASH_WLOCK(pcbinfo);
2012 LIST_REMOVE(inp, inp_hash);
2013 LIST_REMOVE(inp, inp_portlist);
2014 if (LIST_FIRST(&phd->phd_pcblist) == NULL) {
2015 LIST_REMOVE(phd, phd_hash);
2018 INP_HASH_WUNLOCK(pcbinfo);
2019 inp->inp_flags &= ~INP_INHASHLIST;
2021 LIST_REMOVE(inp, inp_list);
2022 pcbinfo->ipi_count--;
2024 in_pcbgroup_remove(inp);
2029 * A set label operation has occurred at the socket layer, propagate the
2030 * label change into the in_pcb for the socket.
2033 in_pcbsosetlabel(struct socket *so)
2038 inp = sotoinpcb(so);
2039 KASSERT(inp != NULL, ("in_pcbsosetlabel: so->so_pcb == NULL"));
2043 mac_inpcb_sosetlabel(so, inp);
2050 * ipport_tick runs once per second, determining if random port allocation
2051 * should be continued. If more than ipport_randomcps ports have been
2052 * allocated in the last second, then we return to sequential port
2053 * allocation. We return to random allocation only once we drop below
2054 * ipport_randomcps for at least ipport_randomtime seconds.
2057 ipport_tick(void *xtp)
2059 VNET_ITERATOR_DECL(vnet_iter);
2061 VNET_LIST_RLOCK_NOSLEEP();
2062 VNET_FOREACH(vnet_iter) {
2063 CURVNET_SET(vnet_iter); /* XXX appease INVARIANTS here */
2064 if (V_ipport_tcpallocs <=
2065 V_ipport_tcplastcount + V_ipport_randomcps) {
2066 if (V_ipport_stoprandom > 0)
2067 V_ipport_stoprandom--;
2069 V_ipport_stoprandom = V_ipport_randomtime;
2070 V_ipport_tcplastcount = V_ipport_tcpallocs;
2073 VNET_LIST_RUNLOCK_NOSLEEP();
2074 callout_reset(&ipport_tick_callout, hz, ipport_tick, NULL);
2081 callout_stop(&ipport_tick_callout);
2085 * The ipport_callout should start running at about the time we attach the
2086 * inet or inet6 domains.
2089 ipport_tick_init(const void *unused __unused)
2092 /* Start ipport_tick. */
2093 callout_init(&ipport_tick_callout, CALLOUT_MPSAFE);
2094 callout_reset(&ipport_tick_callout, 1, ipport_tick, NULL);
2095 EVENTHANDLER_REGISTER(shutdown_pre_sync, ip_fini, NULL,
2096 SHUTDOWN_PRI_DEFAULT);
2098 SYSINIT(ipport_tick_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE,
2099 ipport_tick_init, NULL);
2102 inp_wlock(struct inpcb *inp)
2109 inp_wunlock(struct inpcb *inp)
2116 inp_rlock(struct inpcb *inp)
2123 inp_runlock(struct inpcb *inp)
2131 inp_lock_assert(struct inpcb *inp)
2134 INP_WLOCK_ASSERT(inp);
2138 inp_unlock_assert(struct inpcb *inp)
2141 INP_UNLOCK_ASSERT(inp);
2146 inp_apply_all(void (*func)(struct inpcb *, void *), void *arg)
2150 INP_INFO_RLOCK(&V_tcbinfo);
2151 LIST_FOREACH(inp, V_tcbinfo.ipi_listhead, inp_list) {
2156 INP_INFO_RUNLOCK(&V_tcbinfo);
2160 inp_inpcbtosocket(struct inpcb *inp)
2163 INP_WLOCK_ASSERT(inp);
2164 return (inp->inp_socket);
2168 inp_inpcbtotcpcb(struct inpcb *inp)
2171 INP_WLOCK_ASSERT(inp);
2172 return ((struct tcpcb *)inp->inp_ppcb);
2176 inp_ip_tos_get(const struct inpcb *inp)
2179 return (inp->inp_ip_tos);
2183 inp_ip_tos_set(struct inpcb *inp, int val)
2186 inp->inp_ip_tos = val;
2190 inp_4tuple_get(struct inpcb *inp, uint32_t *laddr, uint16_t *lp,
2191 uint32_t *faddr, uint16_t *fp)
2194 INP_LOCK_ASSERT(inp);
2195 *laddr = inp->inp_laddr.s_addr;
2196 *faddr = inp->inp_faddr.s_addr;
2197 *lp = inp->inp_lport;
2198 *fp = inp->inp_fport;
2202 so_sotoinpcb(struct socket *so)
2205 return (sotoinpcb(so));
2209 so_sototcpcb(struct socket *so)
2212 return (sototcpcb(so));
2217 db_print_indent(int indent)
2221 for (i = 0; i < indent; i++)
2226 db_print_inconninfo(struct in_conninfo *inc, const char *name, int indent)
2228 char faddr_str[48], laddr_str[48];
2230 db_print_indent(indent);
2231 db_printf("%s at %p\n", name, inc);
2236 if (inc->inc_flags & INC_ISIPV6) {
2238 ip6_sprintf(laddr_str, &inc->inc6_laddr);
2239 ip6_sprintf(faddr_str, &inc->inc6_faddr);
2244 inet_ntoa_r(inc->inc_laddr, laddr_str);
2245 inet_ntoa_r(inc->inc_faddr, faddr_str);
2247 db_print_indent(indent);
2248 db_printf("inc_laddr %s inc_lport %u\n", laddr_str,
2249 ntohs(inc->inc_lport));
2250 db_print_indent(indent);
2251 db_printf("inc_faddr %s inc_fport %u\n", faddr_str,
2252 ntohs(inc->inc_fport));
2256 db_print_inpflags(int inp_flags)
2261 if (inp_flags & INP_RECVOPTS) {
2262 db_printf("%sINP_RECVOPTS", comma ? ", " : "");
2265 if (inp_flags & INP_RECVRETOPTS) {
2266 db_printf("%sINP_RECVRETOPTS", comma ? ", " : "");
2269 if (inp_flags & INP_RECVDSTADDR) {
2270 db_printf("%sINP_RECVDSTADDR", comma ? ", " : "");
2273 if (inp_flags & INP_HDRINCL) {
2274 db_printf("%sINP_HDRINCL", comma ? ", " : "");
2277 if (inp_flags & INP_HIGHPORT) {
2278 db_printf("%sINP_HIGHPORT", comma ? ", " : "");
2281 if (inp_flags & INP_LOWPORT) {
2282 db_printf("%sINP_LOWPORT", comma ? ", " : "");
2285 if (inp_flags & INP_ANONPORT) {
2286 db_printf("%sINP_ANONPORT", comma ? ", " : "");
2289 if (inp_flags & INP_RECVIF) {
2290 db_printf("%sINP_RECVIF", comma ? ", " : "");
2293 if (inp_flags & INP_MTUDISC) {
2294 db_printf("%sINP_MTUDISC", comma ? ", " : "");
2297 if (inp_flags & INP_FAITH) {
2298 db_printf("%sINP_FAITH", comma ? ", " : "");
2301 if (inp_flags & INP_RECVTTL) {
2302 db_printf("%sINP_RECVTTL", comma ? ", " : "");
2305 if (inp_flags & INP_DONTFRAG) {
2306 db_printf("%sINP_DONTFRAG", comma ? ", " : "");
2309 if (inp_flags & INP_RECVTOS) {
2310 db_printf("%sINP_RECVTOS", comma ? ", " : "");
2313 if (inp_flags & IN6P_IPV6_V6ONLY) {
2314 db_printf("%sIN6P_IPV6_V6ONLY", comma ? ", " : "");
2317 if (inp_flags & IN6P_PKTINFO) {
2318 db_printf("%sIN6P_PKTINFO", comma ? ", " : "");
2321 if (inp_flags & IN6P_HOPLIMIT) {
2322 db_printf("%sIN6P_HOPLIMIT", comma ? ", " : "");
2325 if (inp_flags & IN6P_HOPOPTS) {
2326 db_printf("%sIN6P_HOPOPTS", comma ? ", " : "");
2329 if (inp_flags & IN6P_DSTOPTS) {
2330 db_printf("%sIN6P_DSTOPTS", comma ? ", " : "");
2333 if (inp_flags & IN6P_RTHDR) {
2334 db_printf("%sIN6P_RTHDR", comma ? ", " : "");
2337 if (inp_flags & IN6P_RTHDRDSTOPTS) {
2338 db_printf("%sIN6P_RTHDRDSTOPTS", comma ? ", " : "");
2341 if (inp_flags & IN6P_TCLASS) {
2342 db_printf("%sIN6P_TCLASS", comma ? ", " : "");
2345 if (inp_flags & IN6P_AUTOFLOWLABEL) {
2346 db_printf("%sIN6P_AUTOFLOWLABEL", comma ? ", " : "");
2349 if (inp_flags & INP_TIMEWAIT) {
2350 db_printf("%sINP_TIMEWAIT", comma ? ", " : "");
2353 if (inp_flags & INP_ONESBCAST) {
2354 db_printf("%sINP_ONESBCAST", comma ? ", " : "");
2357 if (inp_flags & INP_DROPPED) {
2358 db_printf("%sINP_DROPPED", comma ? ", " : "");
2361 if (inp_flags & INP_SOCKREF) {
2362 db_printf("%sINP_SOCKREF", comma ? ", " : "");
2365 if (inp_flags & IN6P_RFC2292) {
2366 db_printf("%sIN6P_RFC2292", comma ? ", " : "");
2369 if (inp_flags & IN6P_MTU) {
2370 db_printf("IN6P_MTU%s", comma ? ", " : "");
2376 db_print_inpvflag(u_char inp_vflag)
2381 if (inp_vflag & INP_IPV4) {
2382 db_printf("%sINP_IPV4", comma ? ", " : "");
2385 if (inp_vflag & INP_IPV6) {
2386 db_printf("%sINP_IPV6", comma ? ", " : "");
2389 if (inp_vflag & INP_IPV6PROTO) {
2390 db_printf("%sINP_IPV6PROTO", comma ? ", " : "");
2396 db_print_inpcb(struct inpcb *inp, const char *name, int indent)
2399 db_print_indent(indent);
2400 db_printf("%s at %p\n", name, inp);
2404 db_print_indent(indent);
2405 db_printf("inp_flow: 0x%x\n", inp->inp_flow);
2407 db_print_inconninfo(&inp->inp_inc, "inp_conninfo", indent);
2409 db_print_indent(indent);
2410 db_printf("inp_ppcb: %p inp_pcbinfo: %p inp_socket: %p\n",
2411 inp->inp_ppcb, inp->inp_pcbinfo, inp->inp_socket);
2413 db_print_indent(indent);
2414 db_printf("inp_label: %p inp_flags: 0x%x (",
2415 inp->inp_label, inp->inp_flags);
2416 db_print_inpflags(inp->inp_flags);
2419 db_print_indent(indent);
2420 db_printf("inp_sp: %p inp_vflag: 0x%x (", inp->inp_sp,
2422 db_print_inpvflag(inp->inp_vflag);
2425 db_print_indent(indent);
2426 db_printf("inp_ip_ttl: %d inp_ip_p: %d inp_ip_minttl: %d\n",
2427 inp->inp_ip_ttl, inp->inp_ip_p, inp->inp_ip_minttl);
2429 db_print_indent(indent);
2431 if (inp->inp_vflag & INP_IPV6) {
2432 db_printf("in6p_options: %p in6p_outputopts: %p "
2433 "in6p_moptions: %p\n", inp->in6p_options,
2434 inp->in6p_outputopts, inp->in6p_moptions);
2435 db_printf("in6p_icmp6filt: %p in6p_cksum %d "
2436 "in6p_hops %u\n", inp->in6p_icmp6filt, inp->in6p_cksum,
2441 db_printf("inp_ip_tos: %d inp_ip_options: %p "
2442 "inp_ip_moptions: %p\n", inp->inp_ip_tos,
2443 inp->inp_options, inp->inp_moptions);
2446 db_print_indent(indent);
2447 db_printf("inp_phd: %p inp_gencnt: %ju\n", inp->inp_phd,
2448 (uintmax_t)inp->inp_gencnt);
2451 DB_SHOW_COMMAND(inpcb, db_show_inpcb)
2456 db_printf("usage: show inpcb <addr>\n");
2459 inp = (struct inpcb *)addr;
2461 db_print_inpcb(inp, "inpcb", 0);