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
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 * 4. Neither the name of the University nor the names of its contributors
16 * may be used to endorse or promote products derived from this software
17 * without specific prior written permission.
19 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31 * @(#)in_pcb.c 8.4 (Berkeley) 5/24/95
34 #include <sys/cdefs.h>
35 __FBSDID("$FreeBSD$");
38 #include "opt_ipsec.h"
39 #include "opt_inet6.h"
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/malloc.h>
45 #include <sys/domain.h>
46 #include <sys/protosw.h>
47 #include <sys/socket.h>
48 #include <sys/socketvar.h>
52 #include <sys/kernel.h>
53 #include <sys/sysctl.h>
62 #include <net/if_types.h>
63 #include <net/route.h>
66 #include <netinet/in.h>
67 #include <netinet/in_pcb.h>
68 #include <netinet/in_var.h>
69 #include <netinet/ip_var.h>
70 #include <netinet/tcp_var.h>
71 #include <netinet/udp.h>
72 #include <netinet/udp_var.h>
74 #include <netinet/ip6.h>
75 #include <netinet6/ip6_var.h>
80 #include <netipsec/ipsec.h>
81 #include <netipsec/key.h>
84 #include <security/mac/mac_framework.h>
87 * These configure the range of local port addresses assigned to
88 * "unspecified" outgoing connections/packets/whatever.
90 VNET_DEFINE(int, ipport_lowfirstauto) = IPPORT_RESERVED - 1; /* 1023 */
91 VNET_DEFINE(int, ipport_lowlastauto) = IPPORT_RESERVEDSTART; /* 600 */
92 VNET_DEFINE(int, ipport_firstauto) = IPPORT_EPHEMERALFIRST; /* 10000 */
93 VNET_DEFINE(int, ipport_lastauto) = IPPORT_EPHEMERALLAST; /* 65535 */
94 VNET_DEFINE(int, ipport_hifirstauto) = IPPORT_HIFIRSTAUTO; /* 49152 */
95 VNET_DEFINE(int, ipport_hilastauto) = IPPORT_HILASTAUTO; /* 65535 */
98 * Reserved ports accessible only to root. There are significant
99 * security considerations that must be accounted for when changing these,
100 * but the security benefits can be great. Please be careful.
102 VNET_DEFINE(int, ipport_reservedhigh) = IPPORT_RESERVED - 1; /* 1023 */
103 VNET_DEFINE(int, ipport_reservedlow);
105 /* Variables dealing with random ephemeral port allocation. */
106 VNET_DEFINE(int, ipport_randomized) = 1; /* user controlled via sysctl */
107 VNET_DEFINE(int, ipport_randomcps) = 10; /* user controlled via sysctl */
108 VNET_DEFINE(int, ipport_randomtime) = 45; /* user controlled via sysctl */
109 VNET_DEFINE(int, ipport_stoprandom); /* toggled by ipport_tick */
110 VNET_DEFINE(int, ipport_tcpallocs);
111 static VNET_DEFINE(int, ipport_tcplastcount);
113 #define V_ipport_tcplastcount VNET(ipport_tcplastcount)
115 #define RANGECHK(var, min, max) \
116 if ((var) < (min)) { (var) = (min); } \
117 else if ((var) > (max)) { (var) = (max); }
119 static void in_pcbremlists(struct inpcb *inp);
122 sysctl_net_ipport_check(SYSCTL_HANDLER_ARGS)
127 error = vnet_sysctl_handle_int(oidp, arg1, arg2, req);
129 error = sysctl_handle_int(oidp, arg1, arg2, req);
132 RANGECHK(V_ipport_lowfirstauto, 1, IPPORT_RESERVED - 1);
133 RANGECHK(V_ipport_lowlastauto, 1, IPPORT_RESERVED - 1);
134 RANGECHK(V_ipport_firstauto, IPPORT_RESERVED, IPPORT_MAX);
135 RANGECHK(V_ipport_lastauto, IPPORT_RESERVED, IPPORT_MAX);
136 RANGECHK(V_ipport_hifirstauto, IPPORT_RESERVED, IPPORT_MAX);
137 RANGECHK(V_ipport_hilastauto, IPPORT_RESERVED, IPPORT_MAX);
144 SYSCTL_NODE(_net_inet_ip, IPPROTO_IP, portrange, CTLFLAG_RW, 0, "IP Ports");
146 SYSCTL_VNET_PROC(_net_inet_ip_portrange, OID_AUTO, lowfirst,
147 CTLTYPE_INT|CTLFLAG_RW, &VNET_NAME(ipport_lowfirstauto), 0,
148 &sysctl_net_ipport_check, "I", "");
149 SYSCTL_VNET_PROC(_net_inet_ip_portrange, OID_AUTO, lowlast,
150 CTLTYPE_INT|CTLFLAG_RW, &VNET_NAME(ipport_lowlastauto), 0,
151 &sysctl_net_ipport_check, "I", "");
152 SYSCTL_VNET_PROC(_net_inet_ip_portrange, OID_AUTO, first,
153 CTLTYPE_INT|CTLFLAG_RW, &VNET_NAME(ipport_firstauto), 0,
154 &sysctl_net_ipport_check, "I", "");
155 SYSCTL_VNET_PROC(_net_inet_ip_portrange, OID_AUTO, last,
156 CTLTYPE_INT|CTLFLAG_RW, &VNET_NAME(ipport_lastauto), 0,
157 &sysctl_net_ipport_check, "I", "");
158 SYSCTL_VNET_PROC(_net_inet_ip_portrange, OID_AUTO, hifirst,
159 CTLTYPE_INT|CTLFLAG_RW, &VNET_NAME(ipport_hifirstauto), 0,
160 &sysctl_net_ipport_check, "I", "");
161 SYSCTL_VNET_PROC(_net_inet_ip_portrange, OID_AUTO, hilast,
162 CTLTYPE_INT|CTLFLAG_RW, &VNET_NAME(ipport_hilastauto), 0,
163 &sysctl_net_ipport_check, "I", "");
164 SYSCTL_VNET_INT(_net_inet_ip_portrange, OID_AUTO, reservedhigh,
165 CTLFLAG_RW|CTLFLAG_SECURE, &VNET_NAME(ipport_reservedhigh), 0, "");
166 SYSCTL_VNET_INT(_net_inet_ip_portrange, OID_AUTO, reservedlow,
167 CTLFLAG_RW|CTLFLAG_SECURE, &VNET_NAME(ipport_reservedlow), 0, "");
168 SYSCTL_VNET_INT(_net_inet_ip_portrange, OID_AUTO, randomized, CTLFLAG_RW,
169 &VNET_NAME(ipport_randomized), 0, "Enable random port allocation");
170 SYSCTL_VNET_INT(_net_inet_ip_portrange, OID_AUTO, randomcps, CTLFLAG_RW,
171 &VNET_NAME(ipport_randomcps), 0, "Maximum number of random port "
172 "allocations before switching to a sequental one");
173 SYSCTL_VNET_INT(_net_inet_ip_portrange, OID_AUTO, randomtime, CTLFLAG_RW,
174 &VNET_NAME(ipport_randomtime), 0,
175 "Minimum time to keep sequental port "
176 "allocation before switching to a random one");
179 * in_pcb.c: manage the Protocol Control Blocks.
181 * NOTE: It is assumed that most of these functions will be called with
182 * the pcbinfo lock held, and often, the inpcb lock held, as these utility
183 * functions often modify hash chains or addresses in pcbs.
187 * Allocate a PCB and associate it with the socket.
188 * On success return with the PCB locked.
191 in_pcballoc(struct socket *so, struct inpcbinfo *pcbinfo)
196 INP_INFO_WLOCK_ASSERT(pcbinfo);
198 inp = uma_zalloc(pcbinfo->ipi_zone, M_NOWAIT);
201 bzero(inp, inp_zero_size);
202 inp->inp_pcbinfo = pcbinfo;
203 inp->inp_socket = so;
204 inp->inp_cred = crhold(so->so_cred);
205 inp->inp_inc.inc_fibnum = so->so_fibnum;
207 error = mac_inpcb_init(inp, M_NOWAIT);
210 mac_inpcb_create(so, inp);
213 error = ipsec_init_policy(so, &inp->inp_sp);
216 mac_inpcb_destroy(inp);
222 if (INP_SOCKAF(so) == AF_INET6) {
223 inp->inp_vflag |= INP_IPV6PROTO;
225 inp->inp_flags |= IN6P_IPV6_V6ONLY;
228 LIST_INSERT_HEAD(pcbinfo->ipi_listhead, inp, inp_list);
229 pcbinfo->ipi_count++;
230 so->so_pcb = (caddr_t)inp;
232 if (V_ip6_auto_flowlabel)
233 inp->inp_flags |= IN6P_AUTOFLOWLABEL;
236 inp->inp_gencnt = ++pcbinfo->ipi_gencnt;
237 inp->inp_refcount = 1; /* Reference from the inpcbinfo */
238 #if defined(IPSEC) || defined(MAC)
241 crfree(inp->inp_cred);
242 uma_zfree(pcbinfo->ipi_zone, inp);
249 in_pcbbind(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred)
253 INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
254 INP_WLOCK_ASSERT(inp);
256 if (inp->inp_lport != 0 || inp->inp_laddr.s_addr != INADDR_ANY)
258 anonport = inp->inp_lport == 0 && (nam == NULL ||
259 ((struct sockaddr_in *)nam)->sin_port == 0);
260 error = in_pcbbind_setup(inp, nam, &inp->inp_laddr.s_addr,
261 &inp->inp_lport, cred);
264 if (in_pcbinshash(inp) != 0) {
265 inp->inp_laddr.s_addr = INADDR_ANY;
270 inp->inp_flags |= INP_ANONPORT;
275 * Set up a bind operation on a PCB, performing port allocation
276 * as required, but do not actually modify the PCB. Callers can
277 * either complete the bind by setting inp_laddr/inp_lport and
278 * calling in_pcbinshash(), or they can just use the resulting
279 * port and address to authorise the sending of a once-off packet.
281 * On error, the values of *laddrp and *lportp are not changed.
284 in_pcbbind_setup(struct inpcb *inp, struct sockaddr *nam, in_addr_t *laddrp,
285 u_short *lportp, struct ucred *cred)
287 struct socket *so = inp->inp_socket;
288 unsigned short *lastport;
289 struct sockaddr_in *sin;
290 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
291 struct in_addr laddr;
293 int wild = 0, reuseport = (so->so_options & SO_REUSEPORT);
298 * Because no actual state changes occur here, a global write lock on
299 * the pcbinfo isn't required.
301 INP_INFO_LOCK_ASSERT(pcbinfo);
302 INP_LOCK_ASSERT(inp);
304 if (TAILQ_EMPTY(&V_in_ifaddrhead)) /* XXX broken! */
305 return (EADDRNOTAVAIL);
306 laddr.s_addr = *laddrp;
307 if (nam != NULL && laddr.s_addr != INADDR_ANY)
309 if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0)
310 wild = INPLOOKUP_WILDCARD;
312 if ((error = prison_local_ip4(cred, &laddr)) != 0)
315 sin = (struct sockaddr_in *)nam;
316 if (nam->sa_len != sizeof (*sin))
320 * We should check the family, but old programs
321 * incorrectly fail to initialize it.
323 if (sin->sin_family != AF_INET)
324 return (EAFNOSUPPORT);
326 error = prison_local_ip4(cred, &sin->sin_addr);
329 if (sin->sin_port != *lportp) {
330 /* Don't allow the port to change. */
333 lport = sin->sin_port;
335 /* NB: lport is left as 0 if the port isn't being changed. */
336 if (IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) {
338 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
339 * allow complete duplication of binding if
340 * SO_REUSEPORT is set, or if SO_REUSEADDR is set
341 * and a multicast address is bound on both
342 * new and duplicated sockets.
344 if (so->so_options & SO_REUSEADDR)
345 reuseport = SO_REUSEADDR|SO_REUSEPORT;
346 } else if (sin->sin_addr.s_addr != INADDR_ANY) {
347 sin->sin_port = 0; /* yech... */
348 bzero(&sin->sin_zero, sizeof(sin->sin_zero));
350 * Is the address a local IP address?
351 * If INP_BINDANY is set, then the socket may be bound
352 * to any endpoint address, local or not.
354 if ((inp->inp_flags & INP_BINDANY) == 0 &&
355 ifa_ifwithaddr_check((struct sockaddr *)sin) == 0)
356 return (EADDRNOTAVAIL);
358 laddr = sin->sin_addr;
364 if (ntohs(lport) <= V_ipport_reservedhigh &&
365 ntohs(lport) >= V_ipport_reservedlow &&
366 priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT,
369 if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)) &&
370 priv_check_cred(inp->inp_cred,
371 PRIV_NETINET_REUSEPORT, 0) != 0) {
372 t = in_pcblookup_local(pcbinfo, sin->sin_addr,
373 lport, INPLOOKUP_WILDCARD, cred);
376 * This entire block sorely needs a rewrite.
379 ((t->inp_flags & INP_TIMEWAIT) == 0) &&
380 (so->so_type != SOCK_STREAM ||
381 ntohl(t->inp_faddr.s_addr) == INADDR_ANY) &&
382 (ntohl(sin->sin_addr.s_addr) != INADDR_ANY ||
383 ntohl(t->inp_laddr.s_addr) != INADDR_ANY ||
384 (t->inp_socket->so_options &
385 SO_REUSEPORT) == 0) &&
386 (inp->inp_cred->cr_uid !=
387 t->inp_cred->cr_uid))
390 t = in_pcblookup_local(pcbinfo, sin->sin_addr,
392 if (t && (t->inp_flags & INP_TIMEWAIT)) {
394 * XXXRW: If an incpb has had its timewait
395 * state recycled, we treat the address as
396 * being in use (for now). This is better
397 * than a panic, but not desirable.
401 (reuseport & tw->tw_so_options) == 0)
404 (reuseport & t->inp_socket->so_options) == 0) {
406 if (ntohl(sin->sin_addr.s_addr) !=
408 ntohl(t->inp_laddr.s_addr) !=
411 INP_SOCKAF(t->inp_socket))
420 u_short first, last, aux;
423 if (inp->inp_flags & INP_HIGHPORT) {
424 first = V_ipport_hifirstauto; /* sysctl */
425 last = V_ipport_hilastauto;
426 lastport = &pcbinfo->ipi_lasthi;
427 } else if (inp->inp_flags & INP_LOWPORT) {
428 error = priv_check_cred(cred,
429 PRIV_NETINET_RESERVEDPORT, 0);
432 first = V_ipport_lowfirstauto; /* 1023 */
433 last = V_ipport_lowlastauto; /* 600 */
434 lastport = &pcbinfo->ipi_lastlow;
436 first = V_ipport_firstauto; /* sysctl */
437 last = V_ipport_lastauto;
438 lastport = &pcbinfo->ipi_lastport;
441 * For UDP, use random port allocation as long as the user
442 * allows it. For TCP (and as of yet unknown) connections,
443 * use random port allocation only if the user allows it AND
444 * ipport_tick() allows it.
446 if (V_ipport_randomized &&
447 (!V_ipport_stoprandom || pcbinfo == &V_udbinfo))
452 * It makes no sense to do random port allocation if
453 * we have the only port available.
457 /* Make sure to not include UDP packets in the count. */
458 if (pcbinfo != &V_udbinfo)
459 V_ipport_tcpallocs++;
461 * Instead of having two loops further down counting up or down
462 * make sure that first is always <= last and go with only one
463 * code path implementing all logic.
473 (arc4random() % (last - first));
475 count = last - first;
478 if (count-- < 0) /* completely used? */
479 return (EADDRNOTAVAIL);
481 if (*lastport < first || *lastport > last)
483 lport = htons(*lastport);
484 } while (in_pcblookup_local(pcbinfo, laddr,
487 *laddrp = laddr.s_addr;
493 * Connect from a socket to a specified address.
494 * Both address and port must be specified in argument sin.
495 * If don't have a local address for this socket yet,
499 in_pcbconnect(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred)
501 u_short lport, fport;
502 in_addr_t laddr, faddr;
505 INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
506 INP_WLOCK_ASSERT(inp);
508 lport = inp->inp_lport;
509 laddr = inp->inp_laddr.s_addr;
510 anonport = (lport == 0);
511 error = in_pcbconnect_setup(inp, nam, &laddr, &lport, &faddr, &fport,
516 /* Do the initial binding of the local address if required. */
517 if (inp->inp_laddr.s_addr == INADDR_ANY && inp->inp_lport == 0) {
518 inp->inp_lport = lport;
519 inp->inp_laddr.s_addr = laddr;
520 if (in_pcbinshash(inp) != 0) {
521 inp->inp_laddr.s_addr = INADDR_ANY;
527 /* Commit the remaining changes. */
528 inp->inp_lport = lport;
529 inp->inp_laddr.s_addr = laddr;
530 inp->inp_faddr.s_addr = faddr;
531 inp->inp_fport = fport;
535 inp->inp_flags |= INP_ANONPORT;
540 * Do proper source address selection on an unbound socket in case
541 * of connect. Take jails into account as well.
544 in_pcbladdr(struct inpcb *inp, struct in_addr *faddr, struct in_addr *laddr,
549 struct sockaddr_in *sin;
553 KASSERT(laddr != NULL, ("%s: laddr NULL", __func__));
556 bzero(&sro, sizeof(sro));
558 sin = (struct sockaddr_in *)&sro.ro_dst;
559 sin->sin_family = AF_INET;
560 sin->sin_len = sizeof(struct sockaddr_in);
561 sin->sin_addr.s_addr = faddr->s_addr;
564 * If route is known our src addr is taken from the i/f,
567 * Find out route to destination.
569 if ((inp->inp_socket->so_options & SO_DONTROUTE) == 0)
570 in_rtalloc_ign(&sro, 0, inp->inp_inc.inc_fibnum);
573 * If we found a route, use the address corresponding to
574 * the outgoing interface.
576 * Otherwise assume faddr is reachable on a directly connected
577 * network and try to find a corresponding interface to take
578 * the source address from.
580 if (sro.ro_rt == NULL || sro.ro_rt->rt_ifp == NULL) {
581 struct in_ifaddr *ia;
584 ia = ifatoia(ifa_ifwithdstaddr((struct sockaddr *)sin));
586 ia = ifatoia(ifa_ifwithnet((struct sockaddr *)sin));
592 if (cred == NULL || !prison_flag(cred, PR_IP4)) {
593 laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
594 ifa_free(&ia->ia_ifa);
599 ifa_free(&ia->ia_ifa);
602 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
605 if (sa->sa_family != AF_INET)
607 sin = (struct sockaddr_in *)sa;
608 if (prison_check_ip4(cred, &sin->sin_addr) == 0) {
609 ia = (struct in_ifaddr *)ifa;
614 laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
620 /* 3. As a last resort return the 'default' jail address. */
621 error = prison_get_ip4(cred, laddr);
626 * If the outgoing interface on the route found is not
627 * a loopback interface, use the address from that interface.
628 * In case of jails do those three steps:
629 * 1. check if the interface address belongs to the jail. If so use it.
630 * 2. check if we have any address on the outgoing interface
631 * belonging to this jail. If so use it.
632 * 3. as a last resort return the 'default' jail address.
634 if ((sro.ro_rt->rt_ifp->if_flags & IFF_LOOPBACK) == 0) {
635 struct in_ifaddr *ia;
638 /* If not jailed, use the default returned. */
639 if (cred == NULL || !prison_flag(cred, PR_IP4)) {
640 ia = (struct in_ifaddr *)sro.ro_rt->rt_ifa;
641 laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
646 /* 1. Check if the iface address belongs to the jail. */
647 sin = (struct sockaddr_in *)sro.ro_rt->rt_ifa->ifa_addr;
648 if (prison_check_ip4(cred, &sin->sin_addr) == 0) {
649 ia = (struct in_ifaddr *)sro.ro_rt->rt_ifa;
650 laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
655 * 2. Check if we have any address on the outgoing interface
656 * belonging to this jail.
659 ifp = sro.ro_rt->rt_ifp;
661 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
663 if (sa->sa_family != AF_INET)
665 sin = (struct sockaddr_in *)sa;
666 if (prison_check_ip4(cred, &sin->sin_addr) == 0) {
667 ia = (struct in_ifaddr *)ifa;
672 laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
678 /* 3. As a last resort return the 'default' jail address. */
679 error = prison_get_ip4(cred, laddr);
684 * The outgoing interface is marked with 'loopback net', so a route
685 * to ourselves is here.
686 * Try to find the interface of the destination address and then
687 * take the address from there. That interface is not necessarily
688 * a loopback interface.
689 * In case of jails, check that it is an address of the jail
690 * and if we cannot find, fall back to the 'default' jail address.
692 if ((sro.ro_rt->rt_ifp->if_flags & IFF_LOOPBACK) != 0) {
693 struct sockaddr_in sain;
694 struct in_ifaddr *ia;
696 bzero(&sain, sizeof(struct sockaddr_in));
697 sain.sin_family = AF_INET;
698 sain.sin_len = sizeof(struct sockaddr_in);
699 sain.sin_addr.s_addr = faddr->s_addr;
701 ia = ifatoia(ifa_ifwithdstaddr(sintosa(&sain)));
703 ia = ifatoia(ifa_ifwithnet(sintosa(&sain)));
705 if (cred == NULL || !prison_flag(cred, PR_IP4)) {
710 laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
711 ifa_free(&ia->ia_ifa);
720 ifa_free(&ia->ia_ifa);
723 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
726 if (sa->sa_family != AF_INET)
728 sin = (struct sockaddr_in *)sa;
729 if (prison_check_ip4(cred,
730 &sin->sin_addr) == 0) {
731 ia = (struct in_ifaddr *)ifa;
736 laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
743 /* 3. As a last resort return the 'default' jail address. */
744 error = prison_get_ip4(cred, laddr);
749 if (sro.ro_rt != NULL)
755 * Set up for a connect from a socket to the specified address.
756 * On entry, *laddrp and *lportp should contain the current local
757 * address and port for the PCB; these are updated to the values
758 * that should be placed in inp_laddr and inp_lport to complete
761 * On success, *faddrp and *fportp will be set to the remote address
762 * and port. These are not updated in the error case.
764 * If the operation fails because the connection already exists,
765 * *oinpp will be set to the PCB of that connection so that the
766 * caller can decide to override it. In all other cases, *oinpp
770 in_pcbconnect_setup(struct inpcb *inp, struct sockaddr *nam,
771 in_addr_t *laddrp, u_short *lportp, in_addr_t *faddrp, u_short *fportp,
772 struct inpcb **oinpp, struct ucred *cred)
774 struct sockaddr_in *sin = (struct sockaddr_in *)nam;
775 struct in_ifaddr *ia;
777 struct in_addr laddr, faddr;
778 u_short lport, fport;
782 * Because a global state change doesn't actually occur here, a read
783 * lock is sufficient.
785 INP_INFO_LOCK_ASSERT(inp->inp_pcbinfo);
786 INP_LOCK_ASSERT(inp);
790 if (nam->sa_len != sizeof (*sin))
792 if (sin->sin_family != AF_INET)
793 return (EAFNOSUPPORT);
794 if (sin->sin_port == 0)
795 return (EADDRNOTAVAIL);
796 laddr.s_addr = *laddrp;
798 faddr = sin->sin_addr;
799 fport = sin->sin_port;
801 if (!TAILQ_EMPTY(&V_in_ifaddrhead)) {
803 * If the destination address is INADDR_ANY,
804 * use the primary local address.
805 * If the supplied address is INADDR_BROADCAST,
806 * and the primary interface supports broadcast,
807 * choose the broadcast address for that interface.
809 if (faddr.s_addr == INADDR_ANY) {
812 IA_SIN(TAILQ_FIRST(&V_in_ifaddrhead))->sin_addr;
815 (error = prison_get_ip4(cred, &faddr)) != 0)
817 } else if (faddr.s_addr == (u_long)INADDR_BROADCAST) {
819 if (TAILQ_FIRST(&V_in_ifaddrhead)->ia_ifp->if_flags &
821 faddr = satosin(&TAILQ_FIRST(
822 &V_in_ifaddrhead)->ia_broadaddr)->sin_addr;
826 if (laddr.s_addr == INADDR_ANY) {
827 error = in_pcbladdr(inp, &faddr, &laddr, cred);
832 * If the destination address is multicast and an outgoing
833 * interface has been set as a multicast option, use the
834 * address of that interface as our source address.
836 if (IN_MULTICAST(ntohl(faddr.s_addr)) &&
837 inp->inp_moptions != NULL) {
838 struct ip_moptions *imo;
841 imo = inp->inp_moptions;
842 if (imo->imo_multicast_ifp != NULL) {
843 ifp = imo->imo_multicast_ifp;
845 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link)
846 if (ia->ia_ifp == ifp)
850 return (EADDRNOTAVAIL);
852 laddr = ia->ia_addr.sin_addr;
858 oinp = in_pcblookup_hash(inp->inp_pcbinfo, faddr, fport, laddr, lport,
866 error = in_pcbbind_setup(inp, NULL, &laddr.s_addr, &lport,
871 *laddrp = laddr.s_addr;
873 *faddrp = faddr.s_addr;
879 in_pcbdisconnect(struct inpcb *inp)
882 INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
883 INP_WLOCK_ASSERT(inp);
885 inp->inp_faddr.s_addr = INADDR_ANY;
891 * in_pcbdetach() is responsibe for disassociating a socket from an inpcb.
892 * For most protocols, this will be invoked immediately prior to calling
893 * in_pcbfree(). However, with TCP the inpcb may significantly outlive the
894 * socket, in which case in_pcbfree() is deferred.
897 in_pcbdetach(struct inpcb *inp)
900 KASSERT(inp->inp_socket != NULL, ("%s: inp_socket == NULL", __func__));
902 inp->inp_socket->so_pcb = NULL;
903 inp->inp_socket = NULL;
907 * in_pcbfree_internal() frees an inpcb that has been detached from its
908 * socket, and whose reference count has reached 0. It will also remove the
909 * inpcb from any global lists it might remain on.
912 in_pcbfree_internal(struct inpcb *inp)
914 struct inpcbinfo *ipi = inp->inp_pcbinfo;
916 KASSERT(inp->inp_socket == NULL, ("%s: inp_socket != NULL", __func__));
917 KASSERT(inp->inp_refcount == 0, ("%s: refcount !0", __func__));
919 INP_INFO_WLOCK_ASSERT(ipi);
920 INP_WLOCK_ASSERT(inp);
923 if (inp->inp_sp != NULL)
924 ipsec_delete_pcbpolicy(inp);
926 inp->inp_gencnt = ++ipi->ipi_gencnt;
929 if (inp->inp_vflag & INP_IPV6PROTO) {
930 ip6_freepcbopts(inp->in6p_outputopts);
931 if (inp->in6p_moptions != NULL)
932 ip6_freemoptions(inp->in6p_moptions);
935 if (inp->inp_options)
936 (void)m_free(inp->inp_options);
937 if (inp->inp_moptions != NULL)
938 inp_freemoptions(inp->inp_moptions);
940 crfree(inp->inp_cred);
943 mac_inpcb_destroy(inp);
946 uma_zfree(ipi->ipi_zone, inp);
950 * in_pcbref() bumps the reference count on an inpcb in order to maintain
951 * stability of an inpcb pointer despite the inpcb lock being released. This
952 * is used in TCP when the inpcbinfo lock needs to be acquired or upgraded,
953 * but where the inpcb lock is already held.
955 * While the inpcb will not be freed, releasing the inpcb lock means that the
956 * connection's state may change, so the caller should be careful to
957 * revalidate any cached state on reacquiring the lock. Drop the reference
958 * using in_pcbrele().
961 in_pcbref(struct inpcb *inp)
964 INP_WLOCK_ASSERT(inp);
966 KASSERT(inp->inp_refcount > 0, ("%s: refcount 0", __func__));
972 * Drop a refcount on an inpcb elevated using in_pcbref(); because a call to
973 * in_pcbfree() may have been made between in_pcbref() and in_pcbrele(), we
974 * return a flag indicating whether or not the inpcb remains valid. If it is
975 * valid, we return with the inpcb lock held.
978 in_pcbrele(struct inpcb *inp)
981 struct inpcbinfo *ipi = inp->inp_pcbinfo;
984 KASSERT(inp->inp_refcount > 0, ("%s: refcount 0", __func__));
986 INP_INFO_WLOCK_ASSERT(ipi);
987 INP_WLOCK_ASSERT(inp);
990 if (inp->inp_refcount > 0)
992 in_pcbfree_internal(inp);
997 * Unconditionally schedule an inpcb to be freed by decrementing its
998 * reference count, which should occur only after the inpcb has been detached
999 * from its socket. If another thread holds a temporary reference (acquired
1000 * using in_pcbref()) then the free is deferred until that reference is
1001 * released using in_pcbrele(), but the inpcb is still unlocked.
1004 in_pcbfree(struct inpcb *inp)
1007 struct inpcbinfo *ipi = inp->inp_pcbinfo;
1010 KASSERT(inp->inp_socket == NULL, ("%s: inp_socket != NULL",
1013 INP_INFO_WLOCK_ASSERT(ipi);
1014 INP_WLOCK_ASSERT(inp);
1016 if (!in_pcbrele(inp))
1021 * in_pcbdrop() removes an inpcb from hashed lists, releasing its address and
1022 * port reservation, and preventing it from being returned by inpcb lookups.
1024 * It is used by TCP to mark an inpcb as unused and avoid future packet
1025 * delivery or event notification when a socket remains open but TCP has
1026 * closed. This might occur as a result of a shutdown()-initiated TCP close
1027 * or a RST on the wire, and allows the port binding to be reused while still
1028 * maintaining the invariant that so_pcb always points to a valid inpcb until
1031 * XXXRW: An inp_lport of 0 is used to indicate that the inpcb is not on hash
1032 * lists, but can lead to confusing netstat output, as open sockets with
1033 * closed TCP connections will no longer appear to have their bound port
1034 * number. An explicit flag would be better, as it would allow us to leave
1035 * the port number intact after the connection is dropped.
1037 * XXXRW: Possibly in_pcbdrop() should also prevent future notifications by
1038 * in_pcbnotifyall() and in_pcbpurgeif0()?
1041 in_pcbdrop(struct inpcb *inp)
1044 INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
1045 INP_WLOCK_ASSERT(inp);
1047 inp->inp_flags |= INP_DROPPED;
1048 if (inp->inp_flags & INP_INHASHLIST) {
1049 struct inpcbport *phd = inp->inp_phd;
1051 LIST_REMOVE(inp, inp_hash);
1052 LIST_REMOVE(inp, inp_portlist);
1053 if (LIST_FIRST(&phd->phd_pcblist) == NULL) {
1054 LIST_REMOVE(phd, phd_hash);
1057 inp->inp_flags &= ~INP_INHASHLIST;
1062 * Common routines to return the socket addresses associated with inpcbs.
1065 in_sockaddr(in_port_t port, struct in_addr *addr_p)
1067 struct sockaddr_in *sin;
1069 sin = malloc(sizeof *sin, M_SONAME,
1071 sin->sin_family = AF_INET;
1072 sin->sin_len = sizeof(*sin);
1073 sin->sin_addr = *addr_p;
1074 sin->sin_port = port;
1076 return (struct sockaddr *)sin;
1080 in_getsockaddr(struct socket *so, struct sockaddr **nam)
1083 struct in_addr addr;
1086 inp = sotoinpcb(so);
1087 KASSERT(inp != NULL, ("in_getsockaddr: inp == NULL"));
1090 port = inp->inp_lport;
1091 addr = inp->inp_laddr;
1094 *nam = in_sockaddr(port, &addr);
1099 in_getpeeraddr(struct socket *so, struct sockaddr **nam)
1102 struct in_addr addr;
1105 inp = sotoinpcb(so);
1106 KASSERT(inp != NULL, ("in_getpeeraddr: inp == NULL"));
1109 port = inp->inp_fport;
1110 addr = inp->inp_faddr;
1113 *nam = in_sockaddr(port, &addr);
1118 in_pcbnotifyall(struct inpcbinfo *pcbinfo, struct in_addr faddr, int errno,
1119 struct inpcb *(*notify)(struct inpcb *, int))
1121 struct inpcb *inp, *inp_temp;
1123 INP_INFO_WLOCK(pcbinfo);
1124 LIST_FOREACH_SAFE(inp, pcbinfo->ipi_listhead, inp_list, inp_temp) {
1127 if ((inp->inp_vflag & INP_IPV4) == 0) {
1132 if (inp->inp_faddr.s_addr != faddr.s_addr ||
1133 inp->inp_socket == NULL) {
1137 if ((*notify)(inp, errno))
1140 INP_INFO_WUNLOCK(pcbinfo);
1144 in_pcbpurgeif0(struct inpcbinfo *pcbinfo, struct ifnet *ifp)
1147 struct ip_moptions *imo;
1150 INP_INFO_RLOCK(pcbinfo);
1151 LIST_FOREACH(inp, pcbinfo->ipi_listhead, inp_list) {
1153 imo = inp->inp_moptions;
1154 if ((inp->inp_vflag & INP_IPV4) &&
1157 * Unselect the outgoing interface if it is being
1160 if (imo->imo_multicast_ifp == ifp)
1161 imo->imo_multicast_ifp = NULL;
1164 * Drop multicast group membership if we joined
1165 * through the interface being detached.
1167 for (i = 0, gap = 0; i < imo->imo_num_memberships;
1169 if (imo->imo_membership[i]->inm_ifp == ifp) {
1170 in_delmulti(imo->imo_membership[i]);
1172 } else if (gap != 0)
1173 imo->imo_membership[i - gap] =
1174 imo->imo_membership[i];
1176 imo->imo_num_memberships -= gap;
1180 INP_INFO_RUNLOCK(pcbinfo);
1184 * Lookup a PCB based on the local address and port.
1186 #define INP_LOOKUP_MAPPED_PCB_COST 3
1188 in_pcblookup_local(struct inpcbinfo *pcbinfo, struct in_addr laddr,
1189 u_short lport, int wild_okay, struct ucred *cred)
1193 int matchwild = 3 + INP_LOOKUP_MAPPED_PCB_COST;
1199 INP_INFO_LOCK_ASSERT(pcbinfo);
1202 struct inpcbhead *head;
1204 * Look for an unconnected (wildcard foreign addr) PCB that
1205 * matches the local address and port we're looking for.
1207 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport,
1208 0, pcbinfo->ipi_hashmask)];
1209 LIST_FOREACH(inp, head, inp_hash) {
1211 /* XXX inp locking */
1212 if ((inp->inp_vflag & INP_IPV4) == 0)
1215 if (inp->inp_faddr.s_addr == INADDR_ANY &&
1216 inp->inp_laddr.s_addr == laddr.s_addr &&
1217 inp->inp_lport == lport) {
1222 prison_equal_ip4(cred->cr_prison,
1223 inp->inp_cred->cr_prison))
1232 struct inpcbporthead *porthash;
1233 struct inpcbport *phd;
1234 struct inpcb *match = NULL;
1236 * Best fit PCB lookup.
1238 * First see if this local port is in use by looking on the
1241 porthash = &pcbinfo->ipi_porthashbase[INP_PCBPORTHASH(lport,
1242 pcbinfo->ipi_porthashmask)];
1243 LIST_FOREACH(phd, porthash, phd_hash) {
1244 if (phd->phd_port == lport)
1249 * Port is in use by one or more PCBs. Look for best
1252 LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) {
1255 !prison_equal_ip4(inp->inp_cred->cr_prison,
1259 /* XXX inp locking */
1260 if ((inp->inp_vflag & INP_IPV4) == 0)
1263 * We never select the PCB that has
1264 * INP_IPV6 flag and is bound to :: if
1265 * we have another PCB which is bound
1266 * to 0.0.0.0. If a PCB has the
1267 * INP_IPV6 flag, then we set its cost
1268 * higher than IPv4 only PCBs.
1270 * Note that the case only happens
1271 * when a socket is bound to ::, under
1272 * the condition that the use of the
1273 * mapped address is allowed.
1275 if ((inp->inp_vflag & INP_IPV6) != 0)
1276 wildcard += INP_LOOKUP_MAPPED_PCB_COST;
1278 if (inp->inp_faddr.s_addr != INADDR_ANY)
1280 if (inp->inp_laddr.s_addr != INADDR_ANY) {
1281 if (laddr.s_addr == INADDR_ANY)
1283 else if (inp->inp_laddr.s_addr != laddr.s_addr)
1286 if (laddr.s_addr != INADDR_ANY)
1289 if (wildcard < matchwild) {
1291 matchwild = wildcard;
1300 #undef INP_LOOKUP_MAPPED_PCB_COST
1303 * Lookup PCB in hash list.
1306 in_pcblookup_hash(struct inpcbinfo *pcbinfo, struct in_addr faddr,
1307 u_int fport_arg, struct in_addr laddr, u_int lport_arg, int wildcard,
1310 struct inpcbhead *head;
1311 struct inpcb *inp, *tmpinp;
1312 u_short fport = fport_arg, lport = lport_arg;
1314 INP_INFO_LOCK_ASSERT(pcbinfo);
1317 * First look for an exact match.
1320 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(faddr.s_addr, lport, fport,
1321 pcbinfo->ipi_hashmask)];
1322 LIST_FOREACH(inp, head, inp_hash) {
1324 /* XXX inp locking */
1325 if ((inp->inp_vflag & INP_IPV4) == 0)
1328 if (inp->inp_faddr.s_addr == faddr.s_addr &&
1329 inp->inp_laddr.s_addr == laddr.s_addr &&
1330 inp->inp_fport == fport &&
1331 inp->inp_lport == lport) {
1333 * XXX We should be able to directly return
1334 * the inp here, without any checks.
1335 * Well unless both bound with SO_REUSEPORT?
1337 if (prison_flag(inp->inp_cred, PR_IP4))
1347 * Then look for a wildcard match, if requested.
1349 if (wildcard == INPLOOKUP_WILDCARD) {
1350 struct inpcb *local_wild = NULL, *local_exact = NULL;
1352 struct inpcb *local_wild_mapped = NULL;
1354 struct inpcb *jail_wild = NULL;
1358 * Order of socket selection - we always prefer jails.
1359 * 1. jailed, non-wild.
1361 * 3. non-jailed, non-wild.
1362 * 4. non-jailed, wild.
1365 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport,
1366 0, pcbinfo->ipi_hashmask)];
1367 LIST_FOREACH(inp, head, inp_hash) {
1369 /* XXX inp locking */
1370 if ((inp->inp_vflag & INP_IPV4) == 0)
1373 if (inp->inp_faddr.s_addr != INADDR_ANY ||
1374 inp->inp_lport != lport)
1377 /* XXX inp locking */
1378 if (ifp && ifp->if_type == IFT_FAITH &&
1379 (inp->inp_flags & INP_FAITH) == 0)
1382 injail = prison_flag(inp->inp_cred, PR_IP4);
1384 if (prison_check_ip4(inp->inp_cred,
1388 if (local_exact != NULL)
1392 if (inp->inp_laddr.s_addr == laddr.s_addr) {
1397 } else if (inp->inp_laddr.s_addr == INADDR_ANY) {
1399 /* XXX inp locking, NULL check */
1400 if (inp->inp_vflag & INP_IPV6PROTO)
1401 local_wild_mapped = inp;
1409 } /* LIST_FOREACH */
1410 if (jail_wild != NULL)
1412 if (local_exact != NULL)
1413 return (local_exact);
1414 if (local_wild != NULL)
1415 return (local_wild);
1417 if (local_wild_mapped != NULL)
1418 return (local_wild_mapped);
1419 #endif /* defined(INET6) */
1420 } /* if (wildcard == INPLOOKUP_WILDCARD) */
1426 * Insert PCB onto various hash lists.
1429 in_pcbinshash(struct inpcb *inp)
1431 struct inpcbhead *pcbhash;
1432 struct inpcbporthead *pcbporthash;
1433 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1434 struct inpcbport *phd;
1435 u_int32_t hashkey_faddr;
1437 INP_INFO_WLOCK_ASSERT(pcbinfo);
1438 INP_WLOCK_ASSERT(inp);
1439 KASSERT((inp->inp_flags & INP_INHASHLIST) == 0,
1440 ("in_pcbinshash: INP_INHASHLIST"));
1443 if (inp->inp_vflag & INP_IPV6)
1444 hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX */;
1447 hashkey_faddr = inp->inp_faddr.s_addr;
1449 pcbhash = &pcbinfo->ipi_hashbase[INP_PCBHASH(hashkey_faddr,
1450 inp->inp_lport, inp->inp_fport, pcbinfo->ipi_hashmask)];
1452 pcbporthash = &pcbinfo->ipi_porthashbase[
1453 INP_PCBPORTHASH(inp->inp_lport, pcbinfo->ipi_porthashmask)];
1456 * Go through port list and look for a head for this lport.
1458 LIST_FOREACH(phd, pcbporthash, phd_hash) {
1459 if (phd->phd_port == inp->inp_lport)
1463 * If none exists, malloc one and tack it on.
1466 phd = malloc(sizeof(struct inpcbport), M_PCB, M_NOWAIT);
1468 return (ENOBUFS); /* XXX */
1470 phd->phd_port = inp->inp_lport;
1471 LIST_INIT(&phd->phd_pcblist);
1472 LIST_INSERT_HEAD(pcbporthash, phd, phd_hash);
1475 LIST_INSERT_HEAD(&phd->phd_pcblist, inp, inp_portlist);
1476 LIST_INSERT_HEAD(pcbhash, inp, inp_hash);
1477 inp->inp_flags |= INP_INHASHLIST;
1482 * Move PCB to the proper hash bucket when { faddr, fport } have been
1483 * changed. NOTE: This does not handle the case of the lport changing (the
1484 * hashed port list would have to be updated as well), so the lport must
1485 * not change after in_pcbinshash() has been called.
1488 in_pcbrehash(struct inpcb *inp)
1490 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1491 struct inpcbhead *head;
1492 u_int32_t hashkey_faddr;
1494 INP_INFO_WLOCK_ASSERT(pcbinfo);
1495 INP_WLOCK_ASSERT(inp);
1496 KASSERT(inp->inp_flags & INP_INHASHLIST,
1497 ("in_pcbrehash: !INP_INHASHLIST"));
1500 if (inp->inp_vflag & INP_IPV6)
1501 hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX */;
1504 hashkey_faddr = inp->inp_faddr.s_addr;
1506 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(hashkey_faddr,
1507 inp->inp_lport, inp->inp_fport, pcbinfo->ipi_hashmask)];
1509 LIST_REMOVE(inp, inp_hash);
1510 LIST_INSERT_HEAD(head, inp, inp_hash);
1514 * Remove PCB from various lists.
1517 in_pcbremlists(struct inpcb *inp)
1519 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1521 INP_INFO_WLOCK_ASSERT(pcbinfo);
1522 INP_WLOCK_ASSERT(inp);
1524 inp->inp_gencnt = ++pcbinfo->ipi_gencnt;
1525 if (inp->inp_flags & INP_INHASHLIST) {
1526 struct inpcbport *phd = inp->inp_phd;
1528 LIST_REMOVE(inp, inp_hash);
1529 LIST_REMOVE(inp, inp_portlist);
1530 if (LIST_FIRST(&phd->phd_pcblist) == NULL) {
1531 LIST_REMOVE(phd, phd_hash);
1534 inp->inp_flags &= ~INP_INHASHLIST;
1536 LIST_REMOVE(inp, inp_list);
1537 pcbinfo->ipi_count--;
1541 * A set label operation has occurred at the socket layer, propagate the
1542 * label change into the in_pcb for the socket.
1545 in_pcbsosetlabel(struct socket *so)
1550 inp = sotoinpcb(so);
1551 KASSERT(inp != NULL, ("in_pcbsosetlabel: so->so_pcb == NULL"));
1555 mac_inpcb_sosetlabel(so, inp);
1562 * ipport_tick runs once per second, determining if random port allocation
1563 * should be continued. If more than ipport_randomcps ports have been
1564 * allocated in the last second, then we return to sequential port
1565 * allocation. We return to random allocation only once we drop below
1566 * ipport_randomcps for at least ipport_randomtime seconds.
1569 ipport_tick(void *xtp)
1571 VNET_ITERATOR_DECL(vnet_iter);
1573 VNET_LIST_RLOCK_NOSLEEP();
1574 VNET_FOREACH(vnet_iter) {
1575 CURVNET_SET(vnet_iter); /* XXX appease INVARIANTS here */
1576 if (V_ipport_tcpallocs <=
1577 V_ipport_tcplastcount + V_ipport_randomcps) {
1578 if (V_ipport_stoprandom > 0)
1579 V_ipport_stoprandom--;
1581 V_ipport_stoprandom = V_ipport_randomtime;
1582 V_ipport_tcplastcount = V_ipport_tcpallocs;
1585 VNET_LIST_RUNLOCK_NOSLEEP();
1586 callout_reset(&ipport_tick_callout, hz, ipport_tick, NULL);
1590 inp_wlock(struct inpcb *inp)
1597 inp_wunlock(struct inpcb *inp)
1604 inp_rlock(struct inpcb *inp)
1611 inp_runlock(struct inpcb *inp)
1619 inp_lock_assert(struct inpcb *inp)
1622 INP_WLOCK_ASSERT(inp);
1626 inp_unlock_assert(struct inpcb *inp)
1629 INP_UNLOCK_ASSERT(inp);
1634 inp_apply_all(void (*func)(struct inpcb *, void *), void *arg)
1638 INP_INFO_RLOCK(&V_tcbinfo);
1639 LIST_FOREACH(inp, V_tcbinfo.ipi_listhead, inp_list) {
1644 INP_INFO_RUNLOCK(&V_tcbinfo);
1648 inp_inpcbtosocket(struct inpcb *inp)
1651 INP_WLOCK_ASSERT(inp);
1652 return (inp->inp_socket);
1656 inp_inpcbtotcpcb(struct inpcb *inp)
1659 INP_WLOCK_ASSERT(inp);
1660 return ((struct tcpcb *)inp->inp_ppcb);
1664 inp_ip_tos_get(const struct inpcb *inp)
1667 return (inp->inp_ip_tos);
1671 inp_ip_tos_set(struct inpcb *inp, int val)
1674 inp->inp_ip_tos = val;
1678 inp_4tuple_get(struct inpcb *inp, uint32_t *laddr, uint16_t *lp,
1679 uint32_t *faddr, uint16_t *fp)
1682 INP_LOCK_ASSERT(inp);
1683 *laddr = inp->inp_laddr.s_addr;
1684 *faddr = inp->inp_faddr.s_addr;
1685 *lp = inp->inp_lport;
1686 *fp = inp->inp_fport;
1690 so_sotoinpcb(struct socket *so)
1693 return (sotoinpcb(so));
1697 so_sototcpcb(struct socket *so)
1700 return (sototcpcb(so));
1705 db_print_indent(int indent)
1709 for (i = 0; i < indent; i++)
1714 db_print_inconninfo(struct in_conninfo *inc, const char *name, int indent)
1716 char faddr_str[48], laddr_str[48];
1718 db_print_indent(indent);
1719 db_printf("%s at %p\n", name, inc);
1724 if (inc->inc_flags & INC_ISIPV6) {
1726 ip6_sprintf(laddr_str, &inc->inc6_laddr);
1727 ip6_sprintf(faddr_str, &inc->inc6_faddr);
1731 inet_ntoa_r(inc->inc_laddr, laddr_str);
1732 inet_ntoa_r(inc->inc_faddr, faddr_str);
1736 db_print_indent(indent);
1737 db_printf("inc_laddr %s inc_lport %u\n", laddr_str,
1738 ntohs(inc->inc_lport));
1739 db_print_indent(indent);
1740 db_printf("inc_faddr %s inc_fport %u\n", faddr_str,
1741 ntohs(inc->inc_fport));
1745 db_print_inpflags(int inp_flags)
1750 if (inp_flags & INP_RECVOPTS) {
1751 db_printf("%sINP_RECVOPTS", comma ? ", " : "");
1754 if (inp_flags & INP_RECVRETOPTS) {
1755 db_printf("%sINP_RECVRETOPTS", comma ? ", " : "");
1758 if (inp_flags & INP_RECVDSTADDR) {
1759 db_printf("%sINP_RECVDSTADDR", comma ? ", " : "");
1762 if (inp_flags & INP_HDRINCL) {
1763 db_printf("%sINP_HDRINCL", comma ? ", " : "");
1766 if (inp_flags & INP_HIGHPORT) {
1767 db_printf("%sINP_HIGHPORT", comma ? ", " : "");
1770 if (inp_flags & INP_LOWPORT) {
1771 db_printf("%sINP_LOWPORT", comma ? ", " : "");
1774 if (inp_flags & INP_ANONPORT) {
1775 db_printf("%sINP_ANONPORT", comma ? ", " : "");
1778 if (inp_flags & INP_RECVIF) {
1779 db_printf("%sINP_RECVIF", comma ? ", " : "");
1782 if (inp_flags & INP_MTUDISC) {
1783 db_printf("%sINP_MTUDISC", comma ? ", " : "");
1786 if (inp_flags & INP_FAITH) {
1787 db_printf("%sINP_FAITH", comma ? ", " : "");
1790 if (inp_flags & INP_RECVTTL) {
1791 db_printf("%sINP_RECVTTL", comma ? ", " : "");
1794 if (inp_flags & INP_DONTFRAG) {
1795 db_printf("%sINP_DONTFRAG", comma ? ", " : "");
1798 if (inp_flags & IN6P_IPV6_V6ONLY) {
1799 db_printf("%sIN6P_IPV6_V6ONLY", comma ? ", " : "");
1802 if (inp_flags & IN6P_PKTINFO) {
1803 db_printf("%sIN6P_PKTINFO", comma ? ", " : "");
1806 if (inp_flags & IN6P_HOPLIMIT) {
1807 db_printf("%sIN6P_HOPLIMIT", comma ? ", " : "");
1810 if (inp_flags & IN6P_HOPOPTS) {
1811 db_printf("%sIN6P_HOPOPTS", comma ? ", " : "");
1814 if (inp_flags & IN6P_DSTOPTS) {
1815 db_printf("%sIN6P_DSTOPTS", comma ? ", " : "");
1818 if (inp_flags & IN6P_RTHDR) {
1819 db_printf("%sIN6P_RTHDR", comma ? ", " : "");
1822 if (inp_flags & IN6P_RTHDRDSTOPTS) {
1823 db_printf("%sIN6P_RTHDRDSTOPTS", comma ? ", " : "");
1826 if (inp_flags & IN6P_TCLASS) {
1827 db_printf("%sIN6P_TCLASS", comma ? ", " : "");
1830 if (inp_flags & IN6P_AUTOFLOWLABEL) {
1831 db_printf("%sIN6P_AUTOFLOWLABEL", comma ? ", " : "");
1834 if (inp_flags & INP_TIMEWAIT) {
1835 db_printf("%sINP_TIMEWAIT", comma ? ", " : "");
1838 if (inp_flags & INP_ONESBCAST) {
1839 db_printf("%sINP_ONESBCAST", comma ? ", " : "");
1842 if (inp_flags & INP_DROPPED) {
1843 db_printf("%sINP_DROPPED", comma ? ", " : "");
1846 if (inp_flags & INP_SOCKREF) {
1847 db_printf("%sINP_SOCKREF", comma ? ", " : "");
1850 if (inp_flags & IN6P_RFC2292) {
1851 db_printf("%sIN6P_RFC2292", comma ? ", " : "");
1854 if (inp_flags & IN6P_MTU) {
1855 db_printf("IN6P_MTU%s", comma ? ", " : "");
1861 db_print_inpvflag(u_char inp_vflag)
1866 if (inp_vflag & INP_IPV4) {
1867 db_printf("%sINP_IPV4", comma ? ", " : "");
1870 if (inp_vflag & INP_IPV6) {
1871 db_printf("%sINP_IPV6", comma ? ", " : "");
1874 if (inp_vflag & INP_IPV6PROTO) {
1875 db_printf("%sINP_IPV6PROTO", comma ? ", " : "");
1881 db_print_inpcb(struct inpcb *inp, const char *name, int indent)
1884 db_print_indent(indent);
1885 db_printf("%s at %p\n", name, inp);
1889 db_print_indent(indent);
1890 db_printf("inp_flow: 0x%x\n", inp->inp_flow);
1892 db_print_inconninfo(&inp->inp_inc, "inp_conninfo", indent);
1894 db_print_indent(indent);
1895 db_printf("inp_ppcb: %p inp_pcbinfo: %p inp_socket: %p\n",
1896 inp->inp_ppcb, inp->inp_pcbinfo, inp->inp_socket);
1898 db_print_indent(indent);
1899 db_printf("inp_label: %p inp_flags: 0x%x (",
1900 inp->inp_label, inp->inp_flags);
1901 db_print_inpflags(inp->inp_flags);
1904 db_print_indent(indent);
1905 db_printf("inp_sp: %p inp_vflag: 0x%x (", inp->inp_sp,
1907 db_print_inpvflag(inp->inp_vflag);
1910 db_print_indent(indent);
1911 db_printf("inp_ip_ttl: %d inp_ip_p: %d inp_ip_minttl: %d\n",
1912 inp->inp_ip_ttl, inp->inp_ip_p, inp->inp_ip_minttl);
1914 db_print_indent(indent);
1916 if (inp->inp_vflag & INP_IPV6) {
1917 db_printf("in6p_options: %p in6p_outputopts: %p "
1918 "in6p_moptions: %p\n", inp->in6p_options,
1919 inp->in6p_outputopts, inp->in6p_moptions);
1920 db_printf("in6p_icmp6filt: %p in6p_cksum %d "
1921 "in6p_hops %u\n", inp->in6p_icmp6filt, inp->in6p_cksum,
1926 db_printf("inp_ip_tos: %d inp_ip_options: %p "
1927 "inp_ip_moptions: %p\n", inp->inp_ip_tos,
1928 inp->inp_options, inp->inp_moptions);
1931 db_print_indent(indent);
1932 db_printf("inp_phd: %p inp_gencnt: %ju\n", inp->inp_phd,
1933 (uintmax_t)inp->inp_gencnt);
1936 DB_SHOW_COMMAND(inpcb, db_show_inpcb)
1941 db_printf("usage: show inpcb <addr>\n");
1944 inp = (struct inpcb *)addr;
1946 db_print_inpcb(inp, "inpcb", 0);