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);
242 * Destroy an inpcbinfo.
245 in_pcbinfo_destroy(struct inpcbinfo *pcbinfo)
248 KASSERT(pcbinfo->ipi_count == 0,
249 ("%s: ipi_count = %u", __func__, pcbinfo->ipi_count));
251 hashdestroy(pcbinfo->ipi_hashbase, M_PCB, pcbinfo->ipi_hashmask);
252 hashdestroy(pcbinfo->ipi_porthashbase, M_PCB,
253 pcbinfo->ipi_porthashmask);
255 in_pcbgroup_destroy(pcbinfo);
257 uma_zdestroy(pcbinfo->ipi_zone);
258 INP_HASH_LOCK_DESTROY(pcbinfo);
259 INP_INFO_LOCK_DESTROY(pcbinfo);
263 * Allocate a PCB and associate it with the socket.
264 * On success return with the PCB locked.
267 in_pcballoc(struct socket *so, struct inpcbinfo *pcbinfo)
272 INP_INFO_WLOCK_ASSERT(pcbinfo);
274 inp = uma_zalloc(pcbinfo->ipi_zone, M_NOWAIT);
277 bzero(inp, inp_zero_size);
278 inp->inp_pcbinfo = pcbinfo;
279 inp->inp_socket = so;
280 inp->inp_cred = crhold(so->so_cred);
281 inp->inp_inc.inc_fibnum = so->so_fibnum;
283 error = mac_inpcb_init(inp, M_NOWAIT);
286 mac_inpcb_create(so, inp);
289 error = ipsec_init_policy(so, &inp->inp_sp);
292 mac_inpcb_destroy(inp);
298 if (INP_SOCKAF(so) == AF_INET6) {
299 inp->inp_vflag |= INP_IPV6PROTO;
301 inp->inp_flags |= IN6P_IPV6_V6ONLY;
304 LIST_INSERT_HEAD(pcbinfo->ipi_listhead, inp, inp_list);
305 pcbinfo->ipi_count++;
306 so->so_pcb = (caddr_t)inp;
308 if (V_ip6_auto_flowlabel)
309 inp->inp_flags |= IN6P_AUTOFLOWLABEL;
312 inp->inp_gencnt = ++pcbinfo->ipi_gencnt;
313 refcount_init(&inp->inp_refcount, 1); /* Reference from inpcbinfo */
314 #if defined(IPSEC) || defined(MAC)
317 crfree(inp->inp_cred);
318 uma_zfree(pcbinfo->ipi_zone, inp);
326 in_pcbbind(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred)
330 INP_WLOCK_ASSERT(inp);
331 INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo);
333 if (inp->inp_lport != 0 || inp->inp_laddr.s_addr != INADDR_ANY)
335 anonport = inp->inp_lport == 0 && (nam == NULL ||
336 ((struct sockaddr_in *)nam)->sin_port == 0);
337 error = in_pcbbind_setup(inp, nam, &inp->inp_laddr.s_addr,
338 &inp->inp_lport, cred);
341 if (in_pcbinshash(inp) != 0) {
342 inp->inp_laddr.s_addr = INADDR_ANY;
347 inp->inp_flags |= INP_ANONPORT;
352 #if defined(INET) || defined(INET6)
354 in_pcb_lport(struct inpcb *inp, struct in_addr *laddrp, u_short *lportp,
355 struct ucred *cred, int lookupflags)
357 struct inpcbinfo *pcbinfo;
358 struct inpcb *tmpinp;
359 unsigned short *lastport;
360 int count, dorandom, error;
361 u_short aux, first, last, lport;
363 struct in_addr laddr;
366 pcbinfo = inp->inp_pcbinfo;
369 * Because no actual state changes occur here, a global write lock on
370 * the pcbinfo isn't required.
372 INP_LOCK_ASSERT(inp);
373 INP_HASH_LOCK_ASSERT(pcbinfo);
375 if (inp->inp_flags & INP_HIGHPORT) {
376 first = V_ipport_hifirstauto; /* sysctl */
377 last = V_ipport_hilastauto;
378 lastport = &pcbinfo->ipi_lasthi;
379 } else if (inp->inp_flags & INP_LOWPORT) {
380 error = priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT, 0);
383 first = V_ipport_lowfirstauto; /* 1023 */
384 last = V_ipport_lowlastauto; /* 600 */
385 lastport = &pcbinfo->ipi_lastlow;
387 first = V_ipport_firstauto; /* sysctl */
388 last = V_ipport_lastauto;
389 lastport = &pcbinfo->ipi_lastport;
392 * For UDP, use random port allocation as long as the user
393 * allows it. For TCP (and as of yet unknown) connections,
394 * use random port allocation only if the user allows it AND
395 * ipport_tick() allows it.
397 if (V_ipport_randomized &&
398 (!V_ipport_stoprandom || pcbinfo == &V_udbinfo))
403 * It makes no sense to do random port allocation if
404 * we have the only port available.
408 /* Make sure to not include UDP packets in the count. */
409 if (pcbinfo != &V_udbinfo)
410 V_ipport_tcpallocs++;
412 * Instead of having two loops further down counting up or down
413 * make sure that first is always <= last and go with only one
414 * code path implementing all logic.
423 /* Make the compiler happy. */
425 if ((inp->inp_vflag & (INP_IPV4|INP_IPV6)) == INP_IPV4) {
426 KASSERT(laddrp != NULL, ("%s: laddrp NULL for v4 inp %p",
431 tmpinp = NULL; /* Make compiler happy. */
435 *lastport = first + (arc4random() % (last - first));
437 count = last - first;
440 if (count-- < 0) /* completely used? */
441 return (EADDRNOTAVAIL);
443 if (*lastport < first || *lastport > last)
445 lport = htons(*lastport);
448 if ((inp->inp_vflag & INP_IPV6) != 0)
449 tmpinp = in6_pcblookup_local(pcbinfo,
450 &inp->in6p_laddr, lport, lookupflags, cred);
452 #if defined(INET) && defined(INET6)
456 tmpinp = in_pcblookup_local(pcbinfo, laddr,
457 lport, lookupflags, cred);
459 } while (tmpinp != NULL);
462 if ((inp->inp_vflag & (INP_IPV4|INP_IPV6)) == INP_IPV4)
463 laddrp->s_addr = laddr.s_addr;
471 * Return cached socket options.
474 inp_so_options(const struct inpcb *inp)
480 if ((inp->inp_flags2 & INP_REUSEPORT) != 0)
481 so_options |= SO_REUSEPORT;
482 if ((inp->inp_flags2 & INP_REUSEADDR) != 0)
483 so_options |= SO_REUSEADDR;
486 #endif /* INET || INET6 */
490 * Set up a bind operation on a PCB, performing port allocation
491 * as required, but do not actually modify the PCB. Callers can
492 * either complete the bind by setting inp_laddr/inp_lport and
493 * calling in_pcbinshash(), or they can just use the resulting
494 * port and address to authorise the sending of a once-off packet.
496 * On error, the values of *laddrp and *lportp are not changed.
499 in_pcbbind_setup(struct inpcb *inp, struct sockaddr *nam, in_addr_t *laddrp,
500 u_short *lportp, struct ucred *cred)
502 struct socket *so = inp->inp_socket;
503 struct sockaddr_in *sin;
504 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
505 struct in_addr laddr;
507 int lookupflags = 0, reuseport = (so->so_options & SO_REUSEPORT);
511 * No state changes, so read locks are sufficient here.
513 INP_LOCK_ASSERT(inp);
514 INP_HASH_LOCK_ASSERT(pcbinfo);
516 if (TAILQ_EMPTY(&V_in_ifaddrhead)) /* XXX broken! */
517 return (EADDRNOTAVAIL);
518 laddr.s_addr = *laddrp;
519 if (nam != NULL && laddr.s_addr != INADDR_ANY)
521 if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0)
522 lookupflags = INPLOOKUP_WILDCARD;
524 if ((error = prison_local_ip4(cred, &laddr)) != 0)
527 sin = (struct sockaddr_in *)nam;
528 if (nam->sa_len != sizeof (*sin))
532 * We should check the family, but old programs
533 * incorrectly fail to initialize it.
535 if (sin->sin_family != AF_INET)
536 return (EAFNOSUPPORT);
538 error = prison_local_ip4(cred, &sin->sin_addr);
541 if (sin->sin_port != *lportp) {
542 /* Don't allow the port to change. */
545 lport = sin->sin_port;
547 /* NB: lport is left as 0 if the port isn't being changed. */
548 if (IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) {
550 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
551 * allow complete duplication of binding if
552 * SO_REUSEPORT is set, or if SO_REUSEADDR is set
553 * and a multicast address is bound on both
554 * new and duplicated sockets.
556 if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) != 0)
557 reuseport = SO_REUSEADDR|SO_REUSEPORT;
558 } else if (sin->sin_addr.s_addr != INADDR_ANY) {
559 sin->sin_port = 0; /* yech... */
560 bzero(&sin->sin_zero, sizeof(sin->sin_zero));
562 * Is the address a local IP address?
563 * If INP_BINDANY is set, then the socket may be bound
564 * to any endpoint address, local or not.
566 if ((inp->inp_flags & INP_BINDANY) == 0 &&
567 ifa_ifwithaddr_check((struct sockaddr *)sin) == 0)
568 return (EADDRNOTAVAIL);
570 laddr = sin->sin_addr;
576 if (ntohs(lport) <= V_ipport_reservedhigh &&
577 ntohs(lport) >= V_ipport_reservedlow &&
578 priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT,
581 if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)) &&
582 priv_check_cred(inp->inp_cred,
583 PRIV_NETINET_REUSEPORT, 0) != 0) {
584 t = in_pcblookup_local(pcbinfo, sin->sin_addr,
585 lport, INPLOOKUP_WILDCARD, cred);
588 * This entire block sorely needs a rewrite.
591 ((t->inp_flags & INP_TIMEWAIT) == 0) &&
592 (so->so_type != SOCK_STREAM ||
593 ntohl(t->inp_faddr.s_addr) == INADDR_ANY) &&
594 (ntohl(sin->sin_addr.s_addr) != INADDR_ANY ||
595 ntohl(t->inp_laddr.s_addr) != INADDR_ANY ||
596 (t->inp_flags2 & INP_REUSEPORT) == 0) &&
597 (inp->inp_cred->cr_uid !=
598 t->inp_cred->cr_uid))
601 t = in_pcblookup_local(pcbinfo, sin->sin_addr,
602 lport, lookupflags, cred);
603 if (t && (t->inp_flags & INP_TIMEWAIT)) {
605 * XXXRW: If an incpb has had its timewait
606 * state recycled, we treat the address as
607 * being in use (for now). This is better
608 * than a panic, but not desirable.
612 (reuseport & tw->tw_so_options) == 0)
614 } else if (t && (reuseport & inp_so_options(t)) == 0) {
616 if (ntohl(sin->sin_addr.s_addr) !=
618 ntohl(t->inp_laddr.s_addr) !=
620 (inp->inp_vflag & INP_IPV6PROTO) == 0 ||
621 (t->inp_vflag & INP_IPV6PROTO) == 0)
630 error = in_pcb_lport(inp, &laddr, &lport, cred, lookupflags);
635 *laddrp = laddr.s_addr;
641 * Connect from a socket to a specified address.
642 * Both address and port must be specified in argument sin.
643 * If don't have a local address for this socket yet,
647 in_pcbconnect_mbuf(struct inpcb *inp, struct sockaddr *nam,
648 struct ucred *cred, struct mbuf *m)
650 u_short lport, fport;
651 in_addr_t laddr, faddr;
654 INP_WLOCK_ASSERT(inp);
655 INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo);
657 lport = inp->inp_lport;
658 laddr = inp->inp_laddr.s_addr;
659 anonport = (lport == 0);
660 error = in_pcbconnect_setup(inp, nam, &laddr, &lport, &faddr, &fport,
665 /* Do the initial binding of the local address if required. */
666 if (inp->inp_laddr.s_addr == INADDR_ANY && inp->inp_lport == 0) {
667 inp->inp_lport = lport;
668 inp->inp_laddr.s_addr = laddr;
669 if (in_pcbinshash(inp) != 0) {
670 inp->inp_laddr.s_addr = INADDR_ANY;
676 /* Commit the remaining changes. */
677 inp->inp_lport = lport;
678 inp->inp_laddr.s_addr = laddr;
679 inp->inp_faddr.s_addr = faddr;
680 inp->inp_fport = fport;
681 in_pcbrehash_mbuf(inp, m);
684 inp->inp_flags |= INP_ANONPORT;
689 in_pcbconnect(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred)
692 return (in_pcbconnect_mbuf(inp, nam, cred, NULL));
696 * Do proper source address selection on an unbound socket in case
697 * of connect. Take jails into account as well.
700 in_pcbladdr(struct inpcb *inp, struct in_addr *faddr, struct in_addr *laddr,
705 struct sockaddr_in *sin;
709 KASSERT(laddr != NULL, ("%s: laddr NULL", __func__));
712 * Bypass source address selection and use the primary jail IP
715 if (cred != NULL && !prison_saddrsel_ip4(cred, laddr))
719 bzero(&sro, sizeof(sro));
721 sin = (struct sockaddr_in *)&sro.ro_dst;
722 sin->sin_family = AF_INET;
723 sin->sin_len = sizeof(struct sockaddr_in);
724 sin->sin_addr.s_addr = faddr->s_addr;
727 * If route is known our src addr is taken from the i/f,
730 * Find out route to destination.
732 if ((inp->inp_socket->so_options & SO_DONTROUTE) == 0)
733 in_rtalloc_ign(&sro, 0, inp->inp_inc.inc_fibnum);
736 * If we found a route, use the address corresponding to
737 * the outgoing interface.
739 * Otherwise assume faddr is reachable on a directly connected
740 * network and try to find a corresponding interface to take
741 * the source address from.
743 if (sro.ro_rt == NULL || sro.ro_rt->rt_ifp == NULL) {
744 struct in_ifaddr *ia;
747 ia = ifatoia(ifa_ifwithdstaddr((struct sockaddr *)sin));
749 ia = ifatoia(ifa_ifwithnet((struct sockaddr *)sin, 0));
755 if (cred == NULL || !prison_flag(cred, PR_IP4)) {
756 laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
757 ifa_free(&ia->ia_ifa);
762 ifa_free(&ia->ia_ifa);
765 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
768 if (sa->sa_family != AF_INET)
770 sin = (struct sockaddr_in *)sa;
771 if (prison_check_ip4(cred, &sin->sin_addr) == 0) {
772 ia = (struct in_ifaddr *)ifa;
777 laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
778 IF_ADDR_RUNLOCK(ifp);
781 IF_ADDR_RUNLOCK(ifp);
783 /* 3. As a last resort return the 'default' jail address. */
784 error = prison_get_ip4(cred, laddr);
789 * If the outgoing interface on the route found is not
790 * a loopback interface, use the address from that interface.
791 * In case of jails do those three steps:
792 * 1. check if the interface address belongs to the jail. If so use it.
793 * 2. check if we have any address on the outgoing interface
794 * belonging to this jail. If so use it.
795 * 3. as a last resort return the 'default' jail address.
797 if ((sro.ro_rt->rt_ifp->if_flags & IFF_LOOPBACK) == 0) {
798 struct in_ifaddr *ia;
801 /* If not jailed, use the default returned. */
802 if (cred == NULL || !prison_flag(cred, PR_IP4)) {
803 ia = (struct in_ifaddr *)sro.ro_rt->rt_ifa;
804 laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
809 /* 1. Check if the iface address belongs to the jail. */
810 sin = (struct sockaddr_in *)sro.ro_rt->rt_ifa->ifa_addr;
811 if (prison_check_ip4(cred, &sin->sin_addr) == 0) {
812 ia = (struct in_ifaddr *)sro.ro_rt->rt_ifa;
813 laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
818 * 2. Check if we have any address on the outgoing interface
819 * belonging to this jail.
822 ifp = sro.ro_rt->rt_ifp;
824 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
826 if (sa->sa_family != AF_INET)
828 sin = (struct sockaddr_in *)sa;
829 if (prison_check_ip4(cred, &sin->sin_addr) == 0) {
830 ia = (struct in_ifaddr *)ifa;
835 laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
836 IF_ADDR_RUNLOCK(ifp);
839 IF_ADDR_RUNLOCK(ifp);
841 /* 3. As a last resort return the 'default' jail address. */
842 error = prison_get_ip4(cred, laddr);
847 * The outgoing interface is marked with 'loopback net', so a route
848 * to ourselves is here.
849 * Try to find the interface of the destination address and then
850 * take the address from there. That interface is not necessarily
851 * a loopback interface.
852 * In case of jails, check that it is an address of the jail
853 * and if we cannot find, fall back to the 'default' jail address.
855 if ((sro.ro_rt->rt_ifp->if_flags & IFF_LOOPBACK) != 0) {
856 struct sockaddr_in sain;
857 struct in_ifaddr *ia;
859 bzero(&sain, sizeof(struct sockaddr_in));
860 sain.sin_family = AF_INET;
861 sain.sin_len = sizeof(struct sockaddr_in);
862 sain.sin_addr.s_addr = faddr->s_addr;
864 ia = ifatoia(ifa_ifwithdstaddr(sintosa(&sain)));
866 ia = ifatoia(ifa_ifwithnet(sintosa(&sain), 0));
868 ia = ifatoia(ifa_ifwithaddr(sintosa(&sain)));
870 if (cred == NULL || !prison_flag(cred, PR_IP4)) {
875 laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
876 ifa_free(&ia->ia_ifa);
885 ifa_free(&ia->ia_ifa);
888 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
891 if (sa->sa_family != AF_INET)
893 sin = (struct sockaddr_in *)sa;
894 if (prison_check_ip4(cred,
895 &sin->sin_addr) == 0) {
896 ia = (struct in_ifaddr *)ifa;
901 laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
902 IF_ADDR_RUNLOCK(ifp);
905 IF_ADDR_RUNLOCK(ifp);
908 /* 3. As a last resort return the 'default' jail address. */
909 error = prison_get_ip4(cred, laddr);
914 if (sro.ro_rt != NULL)
920 * Set up for a connect from a socket to the specified address.
921 * On entry, *laddrp and *lportp should contain the current local
922 * address and port for the PCB; these are updated to the values
923 * that should be placed in inp_laddr and inp_lport to complete
926 * On success, *faddrp and *fportp will be set to the remote address
927 * and port. These are not updated in the error case.
929 * If the operation fails because the connection already exists,
930 * *oinpp will be set to the PCB of that connection so that the
931 * caller can decide to override it. In all other cases, *oinpp
935 in_pcbconnect_setup(struct inpcb *inp, struct sockaddr *nam,
936 in_addr_t *laddrp, u_short *lportp, in_addr_t *faddrp, u_short *fportp,
937 struct inpcb **oinpp, struct ucred *cred)
939 struct sockaddr_in *sin = (struct sockaddr_in *)nam;
940 struct in_ifaddr *ia;
942 struct in_addr laddr, faddr;
943 u_short lport, fport;
947 * Because a global state change doesn't actually occur here, a read
948 * lock is sufficient.
950 INP_LOCK_ASSERT(inp);
951 INP_HASH_LOCK_ASSERT(inp->inp_pcbinfo);
955 if (nam->sa_len != sizeof (*sin))
957 if (sin->sin_family != AF_INET)
958 return (EAFNOSUPPORT);
959 if (sin->sin_port == 0)
960 return (EADDRNOTAVAIL);
961 laddr.s_addr = *laddrp;
963 faddr = sin->sin_addr;
964 fport = sin->sin_port;
966 if (!TAILQ_EMPTY(&V_in_ifaddrhead)) {
968 * If the destination address is INADDR_ANY,
969 * use the primary local address.
970 * If the supplied address is INADDR_BROADCAST,
971 * and the primary interface supports broadcast,
972 * choose the broadcast address for that interface.
974 if (faddr.s_addr == INADDR_ANY) {
977 IA_SIN(TAILQ_FIRST(&V_in_ifaddrhead))->sin_addr;
980 (error = prison_get_ip4(cred, &faddr)) != 0)
982 } else if (faddr.s_addr == (u_long)INADDR_BROADCAST) {
984 if (TAILQ_FIRST(&V_in_ifaddrhead)->ia_ifp->if_flags &
986 faddr = satosin(&TAILQ_FIRST(
987 &V_in_ifaddrhead)->ia_broadaddr)->sin_addr;
991 if (laddr.s_addr == INADDR_ANY) {
992 error = in_pcbladdr(inp, &faddr, &laddr, cred);
994 * If the destination address is multicast and an outgoing
995 * interface has been set as a multicast option, prefer the
996 * address of that interface as our source address.
998 if (IN_MULTICAST(ntohl(faddr.s_addr)) &&
999 inp->inp_moptions != NULL) {
1000 struct ip_moptions *imo;
1003 imo = inp->inp_moptions;
1004 if (imo->imo_multicast_ifp != NULL) {
1005 ifp = imo->imo_multicast_ifp;
1007 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
1008 if ((ia->ia_ifp == ifp) &&
1010 prison_check_ip4(cred,
1011 &ia->ia_addr.sin_addr) == 0))
1015 error = EADDRNOTAVAIL;
1017 laddr = ia->ia_addr.sin_addr;
1020 IN_IFADDR_RUNLOCK();
1026 oinp = in_pcblookup_hash_locked(inp->inp_pcbinfo, faddr, fport,
1027 laddr, lport, 0, NULL);
1031 return (EADDRINUSE);
1034 error = in_pcbbind_setup(inp, NULL, &laddr.s_addr, &lport,
1039 *laddrp = laddr.s_addr;
1041 *faddrp = faddr.s_addr;
1047 in_pcbdisconnect(struct inpcb *inp)
1050 INP_WLOCK_ASSERT(inp);
1051 INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo);
1053 inp->inp_faddr.s_addr = INADDR_ANY;
1060 * in_pcbdetach() is responsibe for disassociating a socket from an inpcb.
1061 * For most protocols, this will be invoked immediately prior to calling
1062 * in_pcbfree(). However, with TCP the inpcb may significantly outlive the
1063 * socket, in which case in_pcbfree() is deferred.
1066 in_pcbdetach(struct inpcb *inp)
1069 KASSERT(inp->inp_socket != NULL, ("%s: inp_socket == NULL", __func__));
1071 inp->inp_socket->so_pcb = NULL;
1072 inp->inp_socket = NULL;
1076 * in_pcbref() bumps the reference count on an inpcb in order to maintain
1077 * stability of an inpcb pointer despite the inpcb lock being released. This
1078 * is used in TCP when the inpcbinfo lock needs to be acquired or upgraded,
1079 * but where the inpcb lock may already held, or when acquiring a reference
1082 * in_pcbref() should be used only to provide brief memory stability, and
1083 * must always be followed by a call to INP_WLOCK() and in_pcbrele() to
1084 * garbage collect the inpcb if it has been in_pcbfree()'d from another
1085 * context. Until in_pcbrele() has returned that the inpcb is still valid,
1086 * lock and rele are the *only* safe operations that may be performed on the
1089 * While the inpcb will not be freed, releasing the inpcb lock means that the
1090 * connection's state may change, so the caller should be careful to
1091 * revalidate any cached state on reacquiring the lock. Drop the reference
1092 * using in_pcbrele().
1095 in_pcbref(struct inpcb *inp)
1098 KASSERT(inp->inp_refcount > 0, ("%s: refcount 0", __func__));
1100 refcount_acquire(&inp->inp_refcount);
1104 * Drop a refcount on an inpcb elevated using in_pcbref(); because a call to
1105 * in_pcbfree() may have been made between in_pcbref() and in_pcbrele(), we
1106 * return a flag indicating whether or not the inpcb remains valid. If it is
1107 * valid, we return with the inpcb lock held.
1109 * Notice that, unlike in_pcbref(), the inpcb lock must be held to drop a
1110 * reference on an inpcb. Historically more work was done here (actually, in
1111 * in_pcbfree_internal()) but has been moved to in_pcbfree() to avoid the
1112 * need for the pcbinfo lock in in_pcbrele(). Deferring the free is entirely
1113 * about memory stability (and continued use of the write lock).
1116 in_pcbrele_rlocked(struct inpcb *inp)
1118 struct inpcbinfo *pcbinfo;
1120 KASSERT(inp->inp_refcount > 0, ("%s: refcount 0", __func__));
1122 INP_RLOCK_ASSERT(inp);
1124 if (refcount_release(&inp->inp_refcount) == 0) {
1126 * If the inpcb has been freed, let the caller know, even if
1127 * this isn't the last reference.
1129 if (inp->inp_flags2 & INP_FREED) {
1136 KASSERT(inp->inp_socket == NULL, ("%s: inp_socket != NULL", __func__));
1139 pcbinfo = inp->inp_pcbinfo;
1140 uma_zfree(pcbinfo->ipi_zone, inp);
1145 in_pcbrele_wlocked(struct inpcb *inp)
1147 struct inpcbinfo *pcbinfo;
1149 KASSERT(inp->inp_refcount > 0, ("%s: refcount 0", __func__));
1151 INP_WLOCK_ASSERT(inp);
1153 if (refcount_release(&inp->inp_refcount) == 0)
1156 KASSERT(inp->inp_socket == NULL, ("%s: inp_socket != NULL", __func__));
1159 pcbinfo = inp->inp_pcbinfo;
1160 uma_zfree(pcbinfo->ipi_zone, inp);
1165 * Temporary wrapper.
1168 in_pcbrele(struct inpcb *inp)
1171 return (in_pcbrele_wlocked(inp));
1175 * Unconditionally schedule an inpcb to be freed by decrementing its
1176 * reference count, which should occur only after the inpcb has been detached
1177 * from its socket. If another thread holds a temporary reference (acquired
1178 * using in_pcbref()) then the free is deferred until that reference is
1179 * released using in_pcbrele(), but the inpcb is still unlocked. Almost all
1180 * work, including removal from global lists, is done in this context, where
1181 * the pcbinfo lock is held.
1184 in_pcbfree(struct inpcb *inp)
1186 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1188 KASSERT(inp->inp_socket == NULL, ("%s: inp_socket != NULL", __func__));
1190 INP_INFO_WLOCK_ASSERT(pcbinfo);
1191 INP_WLOCK_ASSERT(inp);
1193 /* XXXRW: Do as much as possible here. */
1195 if (inp->inp_sp != NULL)
1196 ipsec_delete_pcbpolicy(inp);
1198 inp->inp_gencnt = ++pcbinfo->ipi_gencnt;
1199 in_pcbremlists(inp);
1201 if (inp->inp_vflag & INP_IPV6PROTO) {
1202 ip6_freepcbopts(inp->in6p_outputopts);
1203 if (inp->in6p_moptions != NULL)
1204 ip6_freemoptions(inp->in6p_moptions);
1207 if (inp->inp_options)
1208 (void)m_free(inp->inp_options);
1210 if (inp->inp_moptions != NULL)
1211 inp_freemoptions(inp->inp_moptions);
1214 inp->inp_flags2 |= INP_FREED;
1215 crfree(inp->inp_cred);
1217 mac_inpcb_destroy(inp);
1219 if (!in_pcbrele_wlocked(inp))
1224 * in_pcbdrop() removes an inpcb from hashed lists, releasing its address and
1225 * port reservation, and preventing it from being returned by inpcb lookups.
1227 * It is used by TCP to mark an inpcb as unused and avoid future packet
1228 * delivery or event notification when a socket remains open but TCP has
1229 * closed. This might occur as a result of a shutdown()-initiated TCP close
1230 * or a RST on the wire, and allows the port binding to be reused while still
1231 * maintaining the invariant that so_pcb always points to a valid inpcb until
1234 * XXXRW: Possibly in_pcbdrop() should also prevent future notifications by
1235 * in_pcbnotifyall() and in_pcbpurgeif0()?
1238 in_pcbdrop(struct inpcb *inp)
1241 INP_WLOCK_ASSERT(inp);
1244 * XXXRW: Possibly we should protect the setting of INP_DROPPED with
1247 inp->inp_flags |= INP_DROPPED;
1248 if (inp->inp_flags & INP_INHASHLIST) {
1249 struct inpcbport *phd = inp->inp_phd;
1251 INP_HASH_WLOCK(inp->inp_pcbinfo);
1252 LIST_REMOVE(inp, inp_hash);
1253 LIST_REMOVE(inp, inp_portlist);
1254 if (LIST_FIRST(&phd->phd_pcblist) == NULL) {
1255 LIST_REMOVE(phd, phd_hash);
1258 INP_HASH_WUNLOCK(inp->inp_pcbinfo);
1259 inp->inp_flags &= ~INP_INHASHLIST;
1261 in_pcbgroup_remove(inp);
1268 * Common routines to return the socket addresses associated with inpcbs.
1271 in_sockaddr(in_port_t port, struct in_addr *addr_p)
1273 struct sockaddr_in *sin;
1275 sin = malloc(sizeof *sin, M_SONAME,
1277 sin->sin_family = AF_INET;
1278 sin->sin_len = sizeof(*sin);
1279 sin->sin_addr = *addr_p;
1280 sin->sin_port = port;
1282 return (struct sockaddr *)sin;
1286 in_getsockaddr(struct socket *so, struct sockaddr **nam)
1289 struct in_addr addr;
1292 inp = sotoinpcb(so);
1293 KASSERT(inp != NULL, ("in_getsockaddr: inp == NULL"));
1296 port = inp->inp_lport;
1297 addr = inp->inp_laddr;
1300 *nam = in_sockaddr(port, &addr);
1305 in_getpeeraddr(struct socket *so, struct sockaddr **nam)
1308 struct in_addr addr;
1311 inp = sotoinpcb(so);
1312 KASSERT(inp != NULL, ("in_getpeeraddr: inp == NULL"));
1315 port = inp->inp_fport;
1316 addr = inp->inp_faddr;
1319 *nam = in_sockaddr(port, &addr);
1324 in_pcbnotifyall(struct inpcbinfo *pcbinfo, struct in_addr faddr, int errno,
1325 struct inpcb *(*notify)(struct inpcb *, int))
1327 struct inpcb *inp, *inp_temp;
1329 INP_INFO_WLOCK(pcbinfo);
1330 LIST_FOREACH_SAFE(inp, pcbinfo->ipi_listhead, inp_list, inp_temp) {
1333 if ((inp->inp_vflag & INP_IPV4) == 0) {
1338 if (inp->inp_faddr.s_addr != faddr.s_addr ||
1339 inp->inp_socket == NULL) {
1343 if ((*notify)(inp, errno))
1346 INP_INFO_WUNLOCK(pcbinfo);
1350 in_pcbpurgeif0(struct inpcbinfo *pcbinfo, struct ifnet *ifp)
1353 struct ip_moptions *imo;
1356 INP_INFO_RLOCK(pcbinfo);
1357 LIST_FOREACH(inp, pcbinfo->ipi_listhead, inp_list) {
1359 imo = inp->inp_moptions;
1360 if ((inp->inp_vflag & INP_IPV4) &&
1363 * Unselect the outgoing interface if it is being
1366 if (imo->imo_multicast_ifp == ifp)
1367 imo->imo_multicast_ifp = NULL;
1370 * Drop multicast group membership if we joined
1371 * through the interface being detached.
1373 for (i = 0, gap = 0; i < imo->imo_num_memberships;
1375 if (imo->imo_membership[i]->inm_ifp == ifp) {
1376 in_delmulti(imo->imo_membership[i]);
1378 } else if (gap != 0)
1379 imo->imo_membership[i - gap] =
1380 imo->imo_membership[i];
1382 imo->imo_num_memberships -= gap;
1386 INP_INFO_RUNLOCK(pcbinfo);
1390 * Lookup a PCB based on the local address and port. Caller must hold the
1391 * hash lock. No inpcb locks or references are acquired.
1393 #define INP_LOOKUP_MAPPED_PCB_COST 3
1395 in_pcblookup_local(struct inpcbinfo *pcbinfo, struct in_addr laddr,
1396 u_short lport, int lookupflags, struct ucred *cred)
1400 int matchwild = 3 + INP_LOOKUP_MAPPED_PCB_COST;
1406 KASSERT((lookupflags & ~(INPLOOKUP_WILDCARD)) == 0,
1407 ("%s: invalid lookup flags %d", __func__, lookupflags));
1409 INP_HASH_LOCK_ASSERT(pcbinfo);
1411 if ((lookupflags & INPLOOKUP_WILDCARD) == 0) {
1412 struct inpcbhead *head;
1414 * Look for an unconnected (wildcard foreign addr) PCB that
1415 * matches the local address and port we're looking for.
1417 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport,
1418 0, pcbinfo->ipi_hashmask)];
1419 LIST_FOREACH(inp, head, inp_hash) {
1421 /* XXX inp locking */
1422 if ((inp->inp_vflag & INP_IPV4) == 0)
1425 if (inp->inp_faddr.s_addr == INADDR_ANY &&
1426 inp->inp_laddr.s_addr == laddr.s_addr &&
1427 inp->inp_lport == lport) {
1432 prison_equal_ip4(cred->cr_prison,
1433 inp->inp_cred->cr_prison))
1442 struct inpcbporthead *porthash;
1443 struct inpcbport *phd;
1444 struct inpcb *match = NULL;
1446 * Best fit PCB lookup.
1448 * First see if this local port is in use by looking on the
1451 porthash = &pcbinfo->ipi_porthashbase[INP_PCBPORTHASH(lport,
1452 pcbinfo->ipi_porthashmask)];
1453 LIST_FOREACH(phd, porthash, phd_hash) {
1454 if (phd->phd_port == lport)
1459 * Port is in use by one or more PCBs. Look for best
1462 LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) {
1465 !prison_equal_ip4(inp->inp_cred->cr_prison,
1469 /* XXX inp locking */
1470 if ((inp->inp_vflag & INP_IPV4) == 0)
1473 * We never select the PCB that has
1474 * INP_IPV6 flag and is bound to :: if
1475 * we have another PCB which is bound
1476 * to 0.0.0.0. If a PCB has the
1477 * INP_IPV6 flag, then we set its cost
1478 * higher than IPv4 only PCBs.
1480 * Note that the case only happens
1481 * when a socket is bound to ::, under
1482 * the condition that the use of the
1483 * mapped address is allowed.
1485 if ((inp->inp_vflag & INP_IPV6) != 0)
1486 wildcard += INP_LOOKUP_MAPPED_PCB_COST;
1488 if (inp->inp_faddr.s_addr != INADDR_ANY)
1490 if (inp->inp_laddr.s_addr != INADDR_ANY) {
1491 if (laddr.s_addr == INADDR_ANY)
1493 else if (inp->inp_laddr.s_addr != laddr.s_addr)
1496 if (laddr.s_addr != INADDR_ANY)
1499 if (wildcard < matchwild) {
1501 matchwild = wildcard;
1510 #undef INP_LOOKUP_MAPPED_PCB_COST
1514 * Lookup PCB in hash list, using pcbgroup tables.
1516 static struct inpcb *
1517 in_pcblookup_group(struct inpcbinfo *pcbinfo, struct inpcbgroup *pcbgroup,
1518 struct in_addr faddr, u_int fport_arg, struct in_addr laddr,
1519 u_int lport_arg, int lookupflags, struct ifnet *ifp)
1521 struct inpcbhead *head;
1522 struct inpcb *inp, *tmpinp;
1523 u_short fport = fport_arg, lport = lport_arg;
1526 * First look for an exact match.
1529 INP_GROUP_LOCK(pcbgroup);
1530 head = &pcbgroup->ipg_hashbase[INP_PCBHASH(faddr.s_addr, lport, fport,
1531 pcbgroup->ipg_hashmask)];
1532 LIST_FOREACH(inp, head, inp_pcbgrouphash) {
1534 /* XXX inp locking */
1535 if ((inp->inp_vflag & INP_IPV4) == 0)
1538 if (inp->inp_faddr.s_addr == faddr.s_addr &&
1539 inp->inp_laddr.s_addr == laddr.s_addr &&
1540 inp->inp_fport == fport &&
1541 inp->inp_lport == lport) {
1543 * XXX We should be able to directly return
1544 * the inp here, without any checks.
1545 * Well unless both bound with SO_REUSEPORT?
1547 if (prison_flag(inp->inp_cred, PR_IP4))
1553 if (tmpinp != NULL) {
1559 * Then look for a wildcard match, if requested.
1561 if ((lookupflags & INPLOOKUP_WILDCARD) != 0) {
1562 struct inpcb *local_wild = NULL, *local_exact = NULL;
1564 struct inpcb *local_wild_mapped = NULL;
1566 struct inpcb *jail_wild = NULL;
1567 struct inpcbhead *head;
1571 * Order of socket selection - we always prefer jails.
1572 * 1. jailed, non-wild.
1574 * 3. non-jailed, non-wild.
1575 * 4. non-jailed, wild.
1577 head = &pcbinfo->ipi_wildbase[INP_PCBHASH(INADDR_ANY, lport,
1578 0, pcbinfo->ipi_wildmask)];
1579 LIST_FOREACH(inp, head, inp_pcbgroup_wild) {
1581 /* XXX inp locking */
1582 if ((inp->inp_vflag & INP_IPV4) == 0)
1585 if (inp->inp_faddr.s_addr != INADDR_ANY ||
1586 inp->inp_lport != lport)
1589 /* XXX inp locking */
1590 if (ifp && ifp->if_type == IFT_FAITH &&
1591 (inp->inp_flags & INP_FAITH) == 0)
1594 injail = prison_flag(inp->inp_cred, PR_IP4);
1596 if (prison_check_ip4(inp->inp_cred,
1600 if (local_exact != NULL)
1604 if (inp->inp_laddr.s_addr == laddr.s_addr) {
1609 } else if (inp->inp_laddr.s_addr == INADDR_ANY) {
1611 /* XXX inp locking, NULL check */
1612 if (inp->inp_vflag & INP_IPV6PROTO)
1613 local_wild_mapped = inp;
1621 } /* LIST_FOREACH */
1629 inp = local_wild_mapped;
1630 #endif /* defined(INET6) */
1633 } /* if (lookupflags & INPLOOKUP_WILDCARD) */
1634 INP_GROUP_UNLOCK(pcbgroup);
1639 INP_GROUP_UNLOCK(pcbgroup);
1640 if (lookupflags & INPLOOKUP_WLOCKPCB) {
1642 if (in_pcbrele_wlocked(inp))
1644 } else if (lookupflags & INPLOOKUP_RLOCKPCB) {
1646 if (in_pcbrele_rlocked(inp))
1649 panic("%s: locking bug", __func__);
1652 #endif /* PCBGROUP */
1655 * Lookup PCB in hash list, using pcbinfo tables. This variation assumes
1656 * that the caller has locked the hash list, and will not perform any further
1657 * locking or reference operations on either the hash list or the connection.
1659 static struct inpcb *
1660 in_pcblookup_hash_locked(struct inpcbinfo *pcbinfo, struct in_addr faddr,
1661 u_int fport_arg, struct in_addr laddr, u_int lport_arg, int lookupflags,
1664 struct inpcbhead *head;
1665 struct inpcb *inp, *tmpinp;
1666 u_short fport = fport_arg, lport = lport_arg;
1668 KASSERT((lookupflags & ~(INPLOOKUP_WILDCARD)) == 0,
1669 ("%s: invalid lookup flags %d", __func__, lookupflags));
1671 INP_HASH_LOCK_ASSERT(pcbinfo);
1674 * First look for an exact match.
1677 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(faddr.s_addr, lport, fport,
1678 pcbinfo->ipi_hashmask)];
1679 LIST_FOREACH(inp, head, inp_hash) {
1681 /* XXX inp locking */
1682 if ((inp->inp_vflag & INP_IPV4) == 0)
1685 if (inp->inp_faddr.s_addr == faddr.s_addr &&
1686 inp->inp_laddr.s_addr == laddr.s_addr &&
1687 inp->inp_fport == fport &&
1688 inp->inp_lport == lport) {
1690 * XXX We should be able to directly return
1691 * the inp here, without any checks.
1692 * Well unless both bound with SO_REUSEPORT?
1694 if (prison_flag(inp->inp_cred, PR_IP4))
1704 * Then look for a wildcard match, if requested.
1706 if ((lookupflags & INPLOOKUP_WILDCARD) != 0) {
1707 struct inpcb *local_wild = NULL, *local_exact = NULL;
1709 struct inpcb *local_wild_mapped = NULL;
1711 struct inpcb *jail_wild = NULL;
1715 * Order of socket selection - we always prefer jails.
1716 * 1. jailed, non-wild.
1718 * 3. non-jailed, non-wild.
1719 * 4. non-jailed, wild.
1722 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport,
1723 0, pcbinfo->ipi_hashmask)];
1724 LIST_FOREACH(inp, head, inp_hash) {
1726 /* XXX inp locking */
1727 if ((inp->inp_vflag & INP_IPV4) == 0)
1730 if (inp->inp_faddr.s_addr != INADDR_ANY ||
1731 inp->inp_lport != lport)
1734 /* XXX inp locking */
1735 if (ifp && ifp->if_type == IFT_FAITH &&
1736 (inp->inp_flags & INP_FAITH) == 0)
1739 injail = prison_flag(inp->inp_cred, PR_IP4);
1741 if (prison_check_ip4(inp->inp_cred,
1745 if (local_exact != NULL)
1749 if (inp->inp_laddr.s_addr == laddr.s_addr) {
1754 } else if (inp->inp_laddr.s_addr == INADDR_ANY) {
1756 /* XXX inp locking, NULL check */
1757 if (inp->inp_vflag & INP_IPV6PROTO)
1758 local_wild_mapped = inp;
1766 } /* LIST_FOREACH */
1767 if (jail_wild != NULL)
1769 if (local_exact != NULL)
1770 return (local_exact);
1771 if (local_wild != NULL)
1772 return (local_wild);
1774 if (local_wild_mapped != NULL)
1775 return (local_wild_mapped);
1776 #endif /* defined(INET6) */
1777 } /* if ((lookupflags & INPLOOKUP_WILDCARD) != 0) */
1783 * Lookup PCB in hash list, using pcbinfo tables. This variation locks the
1784 * hash list lock, and will return the inpcb locked (i.e., requires
1785 * INPLOOKUP_LOCKPCB).
1787 static struct inpcb *
1788 in_pcblookup_hash(struct inpcbinfo *pcbinfo, struct in_addr faddr,
1789 u_int fport, struct in_addr laddr, u_int lport, int lookupflags,
1794 INP_HASH_RLOCK(pcbinfo);
1795 inp = in_pcblookup_hash_locked(pcbinfo, faddr, fport, laddr, lport,
1796 (lookupflags & ~(INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)), ifp);
1799 INP_HASH_RUNLOCK(pcbinfo);
1800 if (lookupflags & INPLOOKUP_WLOCKPCB) {
1802 if (in_pcbrele_wlocked(inp))
1804 } else if (lookupflags & INPLOOKUP_RLOCKPCB) {
1806 if (in_pcbrele_rlocked(inp))
1809 panic("%s: locking bug", __func__);
1811 INP_HASH_RUNLOCK(pcbinfo);
1816 * Public inpcb lookup routines, accepting a 4-tuple, and optionally, an mbuf
1817 * from which a pre-calculated hash value may be extracted.
1819 * Possibly more of this logic should be in in_pcbgroup.c.
1822 in_pcblookup(struct inpcbinfo *pcbinfo, struct in_addr faddr, u_int fport,
1823 struct in_addr laddr, u_int lport, int lookupflags, struct ifnet *ifp)
1825 #if defined(PCBGROUP)
1826 struct inpcbgroup *pcbgroup;
1829 KASSERT((lookupflags & ~INPLOOKUP_MASK) == 0,
1830 ("%s: invalid lookup flags %d", __func__, lookupflags));
1831 KASSERT((lookupflags & (INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)) != 0,
1832 ("%s: LOCKPCB not set", __func__));
1834 #if defined(PCBGROUP)
1835 if (in_pcbgroup_enabled(pcbinfo)) {
1836 pcbgroup = in_pcbgroup_bytuple(pcbinfo, laddr, lport, faddr,
1838 return (in_pcblookup_group(pcbinfo, pcbgroup, faddr, fport,
1839 laddr, lport, lookupflags, ifp));
1842 return (in_pcblookup_hash(pcbinfo, faddr, fport, laddr, lport,
1847 in_pcblookup_mbuf(struct inpcbinfo *pcbinfo, struct in_addr faddr,
1848 u_int fport, struct in_addr laddr, u_int lport, int lookupflags,
1849 struct ifnet *ifp, struct mbuf *m)
1852 struct inpcbgroup *pcbgroup;
1855 KASSERT((lookupflags & ~INPLOOKUP_MASK) == 0,
1856 ("%s: invalid lookup flags %d", __func__, lookupflags));
1857 KASSERT((lookupflags & (INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)) != 0,
1858 ("%s: LOCKPCB not set", __func__));
1861 if (in_pcbgroup_enabled(pcbinfo)) {
1862 pcbgroup = in_pcbgroup_byhash(pcbinfo, M_HASHTYPE_GET(m),
1863 m->m_pkthdr.flowid);
1864 if (pcbgroup != NULL)
1865 return (in_pcblookup_group(pcbinfo, pcbgroup, faddr,
1866 fport, laddr, lport, lookupflags, ifp));
1867 pcbgroup = in_pcbgroup_bytuple(pcbinfo, laddr, lport, faddr,
1869 return (in_pcblookup_group(pcbinfo, pcbgroup, faddr, fport,
1870 laddr, lport, lookupflags, ifp));
1873 return (in_pcblookup_hash(pcbinfo, faddr, fport, laddr, lport,
1879 * Insert PCB onto various hash lists.
1882 in_pcbinshash_internal(struct inpcb *inp, int do_pcbgroup_update)
1884 struct inpcbhead *pcbhash;
1885 struct inpcbporthead *pcbporthash;
1886 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1887 struct inpcbport *phd;
1888 u_int32_t hashkey_faddr;
1890 INP_WLOCK_ASSERT(inp);
1891 INP_HASH_WLOCK_ASSERT(pcbinfo);
1893 KASSERT((inp->inp_flags & INP_INHASHLIST) == 0,
1894 ("in_pcbinshash: INP_INHASHLIST"));
1897 if (inp->inp_vflag & INP_IPV6)
1898 hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX */;
1901 hashkey_faddr = inp->inp_faddr.s_addr;
1903 pcbhash = &pcbinfo->ipi_hashbase[INP_PCBHASH(hashkey_faddr,
1904 inp->inp_lport, inp->inp_fport, pcbinfo->ipi_hashmask)];
1906 pcbporthash = &pcbinfo->ipi_porthashbase[
1907 INP_PCBPORTHASH(inp->inp_lport, pcbinfo->ipi_porthashmask)];
1910 * Go through port list and look for a head for this lport.
1912 LIST_FOREACH(phd, pcbporthash, phd_hash) {
1913 if (phd->phd_port == inp->inp_lport)
1917 * If none exists, malloc one and tack it on.
1920 phd = malloc(sizeof(struct inpcbport), M_PCB, M_NOWAIT);
1922 return (ENOBUFS); /* XXX */
1924 phd->phd_port = inp->inp_lport;
1925 LIST_INIT(&phd->phd_pcblist);
1926 LIST_INSERT_HEAD(pcbporthash, phd, phd_hash);
1929 LIST_INSERT_HEAD(&phd->phd_pcblist, inp, inp_portlist);
1930 LIST_INSERT_HEAD(pcbhash, inp, inp_hash);
1931 inp->inp_flags |= INP_INHASHLIST;
1933 if (do_pcbgroup_update)
1934 in_pcbgroup_update(inp);
1940 * For now, there are two public interfaces to insert an inpcb into the hash
1941 * lists -- one that does update pcbgroups, and one that doesn't. The latter
1942 * is used only in the TCP syncache, where in_pcbinshash is called before the
1943 * full 4-tuple is set for the inpcb, and we don't want to install in the
1944 * pcbgroup until later.
1946 * XXXRW: This seems like a misfeature. in_pcbinshash should always update
1947 * connection groups, and partially initialised inpcbs should not be exposed
1948 * to either reservation hash tables or pcbgroups.
1951 in_pcbinshash(struct inpcb *inp)
1954 return (in_pcbinshash_internal(inp, 1));
1958 in_pcbinshash_nopcbgroup(struct inpcb *inp)
1961 return (in_pcbinshash_internal(inp, 0));
1965 * Move PCB to the proper hash bucket when { faddr, fport } have been
1966 * changed. NOTE: This does not handle the case of the lport changing (the
1967 * hashed port list would have to be updated as well), so the lport must
1968 * not change after in_pcbinshash() has been called.
1971 in_pcbrehash_mbuf(struct inpcb *inp, struct mbuf *m)
1973 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1974 struct inpcbhead *head;
1975 u_int32_t hashkey_faddr;
1977 INP_WLOCK_ASSERT(inp);
1978 INP_HASH_WLOCK_ASSERT(pcbinfo);
1980 KASSERT(inp->inp_flags & INP_INHASHLIST,
1981 ("in_pcbrehash: !INP_INHASHLIST"));
1984 if (inp->inp_vflag & INP_IPV6)
1985 hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX */;
1988 hashkey_faddr = inp->inp_faddr.s_addr;
1990 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(hashkey_faddr,
1991 inp->inp_lport, inp->inp_fport, pcbinfo->ipi_hashmask)];
1993 LIST_REMOVE(inp, inp_hash);
1994 LIST_INSERT_HEAD(head, inp, inp_hash);
1998 in_pcbgroup_update_mbuf(inp, m);
2000 in_pcbgroup_update(inp);
2005 in_pcbrehash(struct inpcb *inp)
2008 in_pcbrehash_mbuf(inp, NULL);
2012 * Remove PCB from various lists.
2015 in_pcbremlists(struct inpcb *inp)
2017 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
2019 INP_INFO_WLOCK_ASSERT(pcbinfo);
2020 INP_WLOCK_ASSERT(inp);
2022 inp->inp_gencnt = ++pcbinfo->ipi_gencnt;
2023 if (inp->inp_flags & INP_INHASHLIST) {
2024 struct inpcbport *phd = inp->inp_phd;
2026 INP_HASH_WLOCK(pcbinfo);
2027 LIST_REMOVE(inp, inp_hash);
2028 LIST_REMOVE(inp, inp_portlist);
2029 if (LIST_FIRST(&phd->phd_pcblist) == NULL) {
2030 LIST_REMOVE(phd, phd_hash);
2033 INP_HASH_WUNLOCK(pcbinfo);
2034 inp->inp_flags &= ~INP_INHASHLIST;
2036 LIST_REMOVE(inp, inp_list);
2037 pcbinfo->ipi_count--;
2039 in_pcbgroup_remove(inp);
2044 * A set label operation has occurred at the socket layer, propagate the
2045 * label change into the in_pcb for the socket.
2048 in_pcbsosetlabel(struct socket *so)
2053 inp = sotoinpcb(so);
2054 KASSERT(inp != NULL, ("in_pcbsosetlabel: so->so_pcb == NULL"));
2058 mac_inpcb_sosetlabel(so, inp);
2065 * ipport_tick runs once per second, determining if random port allocation
2066 * should be continued. If more than ipport_randomcps ports have been
2067 * allocated in the last second, then we return to sequential port
2068 * allocation. We return to random allocation only once we drop below
2069 * ipport_randomcps for at least ipport_randomtime seconds.
2072 ipport_tick(void *xtp)
2074 VNET_ITERATOR_DECL(vnet_iter);
2076 VNET_LIST_RLOCK_NOSLEEP();
2077 VNET_FOREACH(vnet_iter) {
2078 CURVNET_SET(vnet_iter); /* XXX appease INVARIANTS here */
2079 if (V_ipport_tcpallocs <=
2080 V_ipport_tcplastcount + V_ipport_randomcps) {
2081 if (V_ipport_stoprandom > 0)
2082 V_ipport_stoprandom--;
2084 V_ipport_stoprandom = V_ipport_randomtime;
2085 V_ipport_tcplastcount = V_ipport_tcpallocs;
2088 VNET_LIST_RUNLOCK_NOSLEEP();
2089 callout_reset(&ipport_tick_callout, hz, ipport_tick, NULL);
2096 callout_stop(&ipport_tick_callout);
2100 * The ipport_callout should start running at about the time we attach the
2101 * inet or inet6 domains.
2104 ipport_tick_init(const void *unused __unused)
2107 /* Start ipport_tick. */
2108 callout_init(&ipport_tick_callout, CALLOUT_MPSAFE);
2109 callout_reset(&ipport_tick_callout, 1, ipport_tick, NULL);
2110 EVENTHANDLER_REGISTER(shutdown_pre_sync, ip_fini, NULL,
2111 SHUTDOWN_PRI_DEFAULT);
2113 SYSINIT(ipport_tick_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE,
2114 ipport_tick_init, NULL);
2117 inp_wlock(struct inpcb *inp)
2124 inp_wunlock(struct inpcb *inp)
2131 inp_rlock(struct inpcb *inp)
2138 inp_runlock(struct inpcb *inp)
2146 inp_lock_assert(struct inpcb *inp)
2149 INP_WLOCK_ASSERT(inp);
2153 inp_unlock_assert(struct inpcb *inp)
2156 INP_UNLOCK_ASSERT(inp);
2161 inp_apply_all(void (*func)(struct inpcb *, void *), void *arg)
2165 INP_INFO_RLOCK(&V_tcbinfo);
2166 LIST_FOREACH(inp, V_tcbinfo.ipi_listhead, inp_list) {
2171 INP_INFO_RUNLOCK(&V_tcbinfo);
2175 inp_inpcbtosocket(struct inpcb *inp)
2178 INP_WLOCK_ASSERT(inp);
2179 return (inp->inp_socket);
2183 inp_inpcbtotcpcb(struct inpcb *inp)
2186 INP_WLOCK_ASSERT(inp);
2187 return ((struct tcpcb *)inp->inp_ppcb);
2191 inp_ip_tos_get(const struct inpcb *inp)
2194 return (inp->inp_ip_tos);
2198 inp_ip_tos_set(struct inpcb *inp, int val)
2201 inp->inp_ip_tos = val;
2205 inp_4tuple_get(struct inpcb *inp, uint32_t *laddr, uint16_t *lp,
2206 uint32_t *faddr, uint16_t *fp)
2209 INP_LOCK_ASSERT(inp);
2210 *laddr = inp->inp_laddr.s_addr;
2211 *faddr = inp->inp_faddr.s_addr;
2212 *lp = inp->inp_lport;
2213 *fp = inp->inp_fport;
2217 so_sotoinpcb(struct socket *so)
2220 return (sotoinpcb(so));
2224 so_sototcpcb(struct socket *so)
2227 return (sototcpcb(so));
2232 db_print_indent(int indent)
2236 for (i = 0; i < indent; i++)
2241 db_print_inconninfo(struct in_conninfo *inc, const char *name, int indent)
2243 char faddr_str[48], laddr_str[48];
2245 db_print_indent(indent);
2246 db_printf("%s at %p\n", name, inc);
2251 if (inc->inc_flags & INC_ISIPV6) {
2253 ip6_sprintf(laddr_str, &inc->inc6_laddr);
2254 ip6_sprintf(faddr_str, &inc->inc6_faddr);
2258 inet_ntoa_r(inc->inc_laddr, laddr_str);
2259 inet_ntoa_r(inc->inc_faddr, faddr_str);
2263 db_print_indent(indent);
2264 db_printf("inc_laddr %s inc_lport %u\n", laddr_str,
2265 ntohs(inc->inc_lport));
2266 db_print_indent(indent);
2267 db_printf("inc_faddr %s inc_fport %u\n", faddr_str,
2268 ntohs(inc->inc_fport));
2272 db_print_inpflags(int inp_flags)
2277 if (inp_flags & INP_RECVOPTS) {
2278 db_printf("%sINP_RECVOPTS", comma ? ", " : "");
2281 if (inp_flags & INP_RECVRETOPTS) {
2282 db_printf("%sINP_RECVRETOPTS", comma ? ", " : "");
2285 if (inp_flags & INP_RECVDSTADDR) {
2286 db_printf("%sINP_RECVDSTADDR", comma ? ", " : "");
2289 if (inp_flags & INP_HDRINCL) {
2290 db_printf("%sINP_HDRINCL", comma ? ", " : "");
2293 if (inp_flags & INP_HIGHPORT) {
2294 db_printf("%sINP_HIGHPORT", comma ? ", " : "");
2297 if (inp_flags & INP_LOWPORT) {
2298 db_printf("%sINP_LOWPORT", comma ? ", " : "");
2301 if (inp_flags & INP_ANONPORT) {
2302 db_printf("%sINP_ANONPORT", comma ? ", " : "");
2305 if (inp_flags & INP_RECVIF) {
2306 db_printf("%sINP_RECVIF", comma ? ", " : "");
2309 if (inp_flags & INP_MTUDISC) {
2310 db_printf("%sINP_MTUDISC", comma ? ", " : "");
2313 if (inp_flags & INP_FAITH) {
2314 db_printf("%sINP_FAITH", comma ? ", " : "");
2317 if (inp_flags & INP_RECVTTL) {
2318 db_printf("%sINP_RECVTTL", comma ? ", " : "");
2321 if (inp_flags & INP_DONTFRAG) {
2322 db_printf("%sINP_DONTFRAG", comma ? ", " : "");
2325 if (inp_flags & INP_RECVTOS) {
2326 db_printf("%sINP_RECVTOS", comma ? ", " : "");
2329 if (inp_flags & IN6P_IPV6_V6ONLY) {
2330 db_printf("%sIN6P_IPV6_V6ONLY", comma ? ", " : "");
2333 if (inp_flags & IN6P_PKTINFO) {
2334 db_printf("%sIN6P_PKTINFO", comma ? ", " : "");
2337 if (inp_flags & IN6P_HOPLIMIT) {
2338 db_printf("%sIN6P_HOPLIMIT", comma ? ", " : "");
2341 if (inp_flags & IN6P_HOPOPTS) {
2342 db_printf("%sIN6P_HOPOPTS", comma ? ", " : "");
2345 if (inp_flags & IN6P_DSTOPTS) {
2346 db_printf("%sIN6P_DSTOPTS", comma ? ", " : "");
2349 if (inp_flags & IN6P_RTHDR) {
2350 db_printf("%sIN6P_RTHDR", comma ? ", " : "");
2353 if (inp_flags & IN6P_RTHDRDSTOPTS) {
2354 db_printf("%sIN6P_RTHDRDSTOPTS", comma ? ", " : "");
2357 if (inp_flags & IN6P_TCLASS) {
2358 db_printf("%sIN6P_TCLASS", comma ? ", " : "");
2361 if (inp_flags & IN6P_AUTOFLOWLABEL) {
2362 db_printf("%sIN6P_AUTOFLOWLABEL", comma ? ", " : "");
2365 if (inp_flags & INP_TIMEWAIT) {
2366 db_printf("%sINP_TIMEWAIT", comma ? ", " : "");
2369 if (inp_flags & INP_ONESBCAST) {
2370 db_printf("%sINP_ONESBCAST", comma ? ", " : "");
2373 if (inp_flags & INP_DROPPED) {
2374 db_printf("%sINP_DROPPED", comma ? ", " : "");
2377 if (inp_flags & INP_SOCKREF) {
2378 db_printf("%sINP_SOCKREF", comma ? ", " : "");
2381 if (inp_flags & IN6P_RFC2292) {
2382 db_printf("%sIN6P_RFC2292", comma ? ", " : "");
2385 if (inp_flags & IN6P_MTU) {
2386 db_printf("IN6P_MTU%s", comma ? ", " : "");
2392 db_print_inpvflag(u_char inp_vflag)
2397 if (inp_vflag & INP_IPV4) {
2398 db_printf("%sINP_IPV4", comma ? ", " : "");
2401 if (inp_vflag & INP_IPV6) {
2402 db_printf("%sINP_IPV6", comma ? ", " : "");
2405 if (inp_vflag & INP_IPV6PROTO) {
2406 db_printf("%sINP_IPV6PROTO", comma ? ", " : "");
2412 db_print_inpcb(struct inpcb *inp, const char *name, int indent)
2415 db_print_indent(indent);
2416 db_printf("%s at %p\n", name, inp);
2420 db_print_indent(indent);
2421 db_printf("inp_flow: 0x%x\n", inp->inp_flow);
2423 db_print_inconninfo(&inp->inp_inc, "inp_conninfo", indent);
2425 db_print_indent(indent);
2426 db_printf("inp_ppcb: %p inp_pcbinfo: %p inp_socket: %p\n",
2427 inp->inp_ppcb, inp->inp_pcbinfo, inp->inp_socket);
2429 db_print_indent(indent);
2430 db_printf("inp_label: %p inp_flags: 0x%x (",
2431 inp->inp_label, inp->inp_flags);
2432 db_print_inpflags(inp->inp_flags);
2435 db_print_indent(indent);
2436 db_printf("inp_sp: %p inp_vflag: 0x%x (", inp->inp_sp,
2438 db_print_inpvflag(inp->inp_vflag);
2441 db_print_indent(indent);
2442 db_printf("inp_ip_ttl: %d inp_ip_p: %d inp_ip_minttl: %d\n",
2443 inp->inp_ip_ttl, inp->inp_ip_p, inp->inp_ip_minttl);
2445 db_print_indent(indent);
2447 if (inp->inp_vflag & INP_IPV6) {
2448 db_printf("in6p_options: %p in6p_outputopts: %p "
2449 "in6p_moptions: %p\n", inp->in6p_options,
2450 inp->in6p_outputopts, inp->in6p_moptions);
2451 db_printf("in6p_icmp6filt: %p in6p_cksum %d "
2452 "in6p_hops %u\n", inp->in6p_icmp6filt, inp->in6p_cksum,
2457 db_printf("inp_ip_tos: %d inp_ip_options: %p "
2458 "inp_ip_moptions: %p\n", inp->inp_ip_tos,
2459 inp->inp_options, inp->inp_moptions);
2462 db_print_indent(indent);
2463 db_printf("inp_phd: %p inp_gencnt: %ju\n", inp->inp_phd,
2464 (uintmax_t)inp->inp_gencnt);
2467 DB_SHOW_COMMAND(inpcb, db_show_inpcb)
2472 db_printf("usage: show inpcb <addr>\n");
2475 inp = (struct inpcb *)addr;
2477 db_print_inpcb(inp, "inpcb", 0);