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"
48 #include <sys/param.h>
49 #include <sys/systm.h>
50 #include <sys/malloc.h>
52 #include <sys/callout.h>
53 #include <sys/domain.h>
54 #include <sys/protosw.h>
55 #include <sys/socket.h>
56 #include <sys/socketvar.h>
59 #include <sys/refcount.h>
61 #include <sys/kernel.h>
62 #include <sys/sysctl.h>
71 #include <net/if_var.h>
72 #include <net/if_types.h>
73 #include <net/route.h>
76 #if defined(INET) || defined(INET6)
77 #include <netinet/in.h>
78 #include <netinet/in_pcb.h>
79 #include <netinet/in_rss.h>
80 #include <netinet/ip_var.h>
81 #include <netinet/tcp_var.h>
82 #include <netinet/udp.h>
83 #include <netinet/udp_var.h>
86 #include <netinet/in_var.h>
89 #include <netinet/ip6.h>
90 #include <netinet6/in6_pcb.h>
91 #include <netinet6/in6_var.h>
92 #include <netinet6/ip6_var.h>
97 #include <netipsec/ipsec.h>
98 #include <netipsec/key.h>
101 #include <security/mac/mac_framework.h>
103 static struct callout ipport_tick_callout;
106 * These configure the range of local port addresses assigned to
107 * "unspecified" outgoing connections/packets/whatever.
109 VNET_DEFINE(int, ipport_lowfirstauto) = IPPORT_RESERVED - 1; /* 1023 */
110 VNET_DEFINE(int, ipport_lowlastauto) = IPPORT_RESERVEDSTART; /* 600 */
111 VNET_DEFINE(int, ipport_firstauto) = IPPORT_EPHEMERALFIRST; /* 10000 */
112 VNET_DEFINE(int, ipport_lastauto) = IPPORT_EPHEMERALLAST; /* 65535 */
113 VNET_DEFINE(int, ipport_hifirstauto) = IPPORT_HIFIRSTAUTO; /* 49152 */
114 VNET_DEFINE(int, ipport_hilastauto) = IPPORT_HILASTAUTO; /* 65535 */
117 * Reserved ports accessible only to root. There are significant
118 * security considerations that must be accounted for when changing these,
119 * but the security benefits can be great. Please be careful.
121 VNET_DEFINE(int, ipport_reservedhigh) = IPPORT_RESERVED - 1; /* 1023 */
122 VNET_DEFINE(int, ipport_reservedlow);
124 /* Variables dealing with random ephemeral port allocation. */
125 VNET_DEFINE(int, ipport_randomized) = 1; /* user controlled via sysctl */
126 VNET_DEFINE(int, ipport_randomcps) = 10; /* user controlled via sysctl */
127 VNET_DEFINE(int, ipport_randomtime) = 45; /* user controlled via sysctl */
128 VNET_DEFINE(int, ipport_stoprandom); /* toggled by ipport_tick */
129 VNET_DEFINE(int, ipport_tcpallocs);
130 static VNET_DEFINE(int, ipport_tcplastcount);
132 #define V_ipport_tcplastcount VNET(ipport_tcplastcount)
134 static void in_pcbremlists(struct inpcb *inp);
136 static struct inpcb *in_pcblookup_hash_locked(struct inpcbinfo *pcbinfo,
137 struct in_addr faddr, u_int fport_arg,
138 struct in_addr laddr, u_int lport_arg,
139 int lookupflags, struct ifnet *ifp);
141 #define RANGECHK(var, min, max) \
142 if ((var) < (min)) { (var) = (min); } \
143 else if ((var) > (max)) { (var) = (max); }
146 sysctl_net_ipport_check(SYSCTL_HANDLER_ARGS)
150 error = sysctl_handle_int(oidp, arg1, arg2, req);
152 RANGECHK(V_ipport_lowfirstauto, 1, IPPORT_RESERVED - 1);
153 RANGECHK(V_ipport_lowlastauto, 1, IPPORT_RESERVED - 1);
154 RANGECHK(V_ipport_firstauto, IPPORT_RESERVED, IPPORT_MAX);
155 RANGECHK(V_ipport_lastauto, IPPORT_RESERVED, IPPORT_MAX);
156 RANGECHK(V_ipport_hifirstauto, IPPORT_RESERVED, IPPORT_MAX);
157 RANGECHK(V_ipport_hilastauto, IPPORT_RESERVED, IPPORT_MAX);
164 static SYSCTL_NODE(_net_inet_ip, IPPROTO_IP, portrange, CTLFLAG_RW, 0,
167 SYSCTL_VNET_PROC(_net_inet_ip_portrange, OID_AUTO, lowfirst,
168 CTLTYPE_INT|CTLFLAG_RW, &VNET_NAME(ipport_lowfirstauto), 0,
169 &sysctl_net_ipport_check, "I", "");
170 SYSCTL_VNET_PROC(_net_inet_ip_portrange, OID_AUTO, lowlast,
171 CTLTYPE_INT|CTLFLAG_RW, &VNET_NAME(ipport_lowlastauto), 0,
172 &sysctl_net_ipport_check, "I", "");
173 SYSCTL_VNET_PROC(_net_inet_ip_portrange, OID_AUTO, first,
174 CTLTYPE_INT|CTLFLAG_RW, &VNET_NAME(ipport_firstauto), 0,
175 &sysctl_net_ipport_check, "I", "");
176 SYSCTL_VNET_PROC(_net_inet_ip_portrange, OID_AUTO, last,
177 CTLTYPE_INT|CTLFLAG_RW, &VNET_NAME(ipport_lastauto), 0,
178 &sysctl_net_ipport_check, "I", "");
179 SYSCTL_VNET_PROC(_net_inet_ip_portrange, OID_AUTO, hifirst,
180 CTLTYPE_INT|CTLFLAG_RW, &VNET_NAME(ipport_hifirstauto), 0,
181 &sysctl_net_ipport_check, "I", "");
182 SYSCTL_VNET_PROC(_net_inet_ip_portrange, OID_AUTO, hilast,
183 CTLTYPE_INT|CTLFLAG_RW, &VNET_NAME(ipport_hilastauto), 0,
184 &sysctl_net_ipport_check, "I", "");
185 SYSCTL_VNET_INT(_net_inet_ip_portrange, OID_AUTO, reservedhigh,
186 CTLFLAG_RW|CTLFLAG_SECURE, &VNET_NAME(ipport_reservedhigh), 0, "");
187 SYSCTL_VNET_INT(_net_inet_ip_portrange, OID_AUTO, reservedlow,
188 CTLFLAG_RW|CTLFLAG_SECURE, &VNET_NAME(ipport_reservedlow), 0, "");
189 SYSCTL_VNET_INT(_net_inet_ip_portrange, OID_AUTO, randomized, CTLFLAG_RW,
190 &VNET_NAME(ipport_randomized), 0, "Enable random port allocation");
191 SYSCTL_VNET_INT(_net_inet_ip_portrange, OID_AUTO, randomcps, CTLFLAG_RW,
192 &VNET_NAME(ipport_randomcps), 0, "Maximum number of random port "
193 "allocations before switching to a sequental one");
194 SYSCTL_VNET_INT(_net_inet_ip_portrange, OID_AUTO, randomtime, CTLFLAG_RW,
195 &VNET_NAME(ipport_randomtime), 0,
196 "Minimum time to keep sequental port "
197 "allocation before switching to a random one");
201 * in_pcb.c: manage the Protocol Control Blocks.
203 * NOTE: It is assumed that most of these functions will be called with
204 * the pcbinfo lock held, and often, the inpcb lock held, as these utility
205 * functions often modify hash chains or addresses in pcbs.
209 * Initialize an inpcbinfo -- we should be able to reduce the number of
213 in_pcbinfo_init(struct inpcbinfo *pcbinfo, const char *name,
214 struct inpcbhead *listhead, int hash_nelements, int porthash_nelements,
215 char *inpcbzone_name, uma_init inpcbzone_init, uma_fini inpcbzone_fini,
216 uint32_t inpcbzone_flags, u_int hashfields)
219 INP_INFO_LOCK_INIT(pcbinfo, name);
220 INP_HASH_LOCK_INIT(pcbinfo, "pcbinfohash"); /* XXXRW: argument? */
222 pcbinfo->ipi_vnet = curvnet;
224 pcbinfo->ipi_listhead = listhead;
225 LIST_INIT(pcbinfo->ipi_listhead);
226 pcbinfo->ipi_count = 0;
227 pcbinfo->ipi_hashbase = hashinit(hash_nelements, M_PCB,
228 &pcbinfo->ipi_hashmask);
229 pcbinfo->ipi_porthashbase = hashinit(porthash_nelements, M_PCB,
230 &pcbinfo->ipi_porthashmask);
232 in_pcbgroup_init(pcbinfo, hashfields, hash_nelements);
234 pcbinfo->ipi_zone = uma_zcreate(inpcbzone_name, sizeof(struct inpcb),
235 NULL, NULL, inpcbzone_init, inpcbzone_fini, UMA_ALIGN_PTR,
237 uma_zone_set_max(pcbinfo->ipi_zone, maxsockets);
238 uma_zone_set_warning(pcbinfo->ipi_zone,
239 "kern.ipc.maxsockets limit reached");
243 * Destroy an inpcbinfo.
246 in_pcbinfo_destroy(struct inpcbinfo *pcbinfo)
249 KASSERT(pcbinfo->ipi_count == 0,
250 ("%s: ipi_count = %u", __func__, pcbinfo->ipi_count));
252 hashdestroy(pcbinfo->ipi_hashbase, M_PCB, pcbinfo->ipi_hashmask);
253 hashdestroy(pcbinfo->ipi_porthashbase, M_PCB,
254 pcbinfo->ipi_porthashmask);
256 in_pcbgroup_destroy(pcbinfo);
258 uma_zdestroy(pcbinfo->ipi_zone);
259 INP_HASH_LOCK_DESTROY(pcbinfo);
260 INP_INFO_LOCK_DESTROY(pcbinfo);
264 * Allocate a PCB and associate it with the socket.
265 * On success return with the PCB locked.
268 in_pcballoc(struct socket *so, struct inpcbinfo *pcbinfo)
273 INP_INFO_WLOCK_ASSERT(pcbinfo);
275 inp = uma_zalloc(pcbinfo->ipi_zone, M_NOWAIT);
278 bzero(inp, inp_zero_size);
279 inp->inp_pcbinfo = pcbinfo;
280 inp->inp_socket = so;
281 inp->inp_cred = crhold(so->so_cred);
282 inp->inp_inc.inc_fibnum = so->so_fibnum;
284 error = mac_inpcb_init(inp, M_NOWAIT);
287 mac_inpcb_create(so, inp);
290 error = ipsec_init_policy(so, &inp->inp_sp);
293 mac_inpcb_destroy(inp);
299 if (INP_SOCKAF(so) == AF_INET6) {
300 inp->inp_vflag |= INP_IPV6PROTO;
302 inp->inp_flags |= IN6P_IPV6_V6ONLY;
305 LIST_INSERT_HEAD(pcbinfo->ipi_listhead, inp, inp_list);
306 pcbinfo->ipi_count++;
307 so->so_pcb = (caddr_t)inp;
309 if (V_ip6_auto_flowlabel)
310 inp->inp_flags |= IN6P_AUTOFLOWLABEL;
313 inp->inp_gencnt = ++pcbinfo->ipi_gencnt;
314 refcount_init(&inp->inp_refcount, 1); /* Reference from inpcbinfo */
315 #if defined(IPSEC) || defined(MAC)
318 crfree(inp->inp_cred);
319 uma_zfree(pcbinfo->ipi_zone, inp);
327 in_pcbbind(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred)
331 INP_WLOCK_ASSERT(inp);
332 INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo);
334 if (inp->inp_lport != 0 || inp->inp_laddr.s_addr != INADDR_ANY)
336 anonport = nam == NULL || ((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(-Lite), 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 ||
399 pcbinfo == &V_ulitecbinfo))
404 * It makes no sense to do random port allocation if
405 * we have the only port available.
409 /* Make sure to not include UDP(-Lite) packets in the count. */
410 if (pcbinfo != &V_udbinfo || pcbinfo != &V_ulitecbinfo)
411 V_ipport_tcpallocs++;
413 * Instead of having two loops further down counting up or down
414 * make sure that first is always <= last and go with only one
415 * code path implementing all logic.
424 /* Make the compiler happy. */
426 if ((inp->inp_vflag & (INP_IPV4|INP_IPV6)) == INP_IPV4) {
427 KASSERT(laddrp != NULL, ("%s: laddrp NULL for v4 inp %p",
432 tmpinp = NULL; /* Make compiler happy. */
436 *lastport = first + (arc4random() % (last - first));
438 count = last - first;
441 if (count-- < 0) /* completely used? */
442 return (EADDRNOTAVAIL);
444 if (*lastport < first || *lastport > last)
446 lport = htons(*lastport);
449 if ((inp->inp_vflag & INP_IPV6) != 0)
450 tmpinp = in6_pcblookup_local(pcbinfo,
451 &inp->in6p_laddr, lport, lookupflags, cred);
453 #if defined(INET) && defined(INET6)
457 tmpinp = in_pcblookup_local(pcbinfo, laddr,
458 lport, lookupflags, cred);
460 } while (tmpinp != NULL);
463 if ((inp->inp_vflag & (INP_IPV4|INP_IPV6)) == INP_IPV4)
464 laddrp->s_addr = laddr.s_addr;
472 * Return cached socket options.
475 inp_so_options(const struct inpcb *inp)
481 if ((inp->inp_flags2 & INP_REUSEPORT) != 0)
482 so_options |= SO_REUSEPORT;
483 if ((inp->inp_flags2 & INP_REUSEADDR) != 0)
484 so_options |= SO_REUSEADDR;
487 #endif /* INET || INET6 */
491 * Set up a bind operation on a PCB, performing port allocation
492 * as required, but do not actually modify the PCB. Callers can
493 * either complete the bind by setting inp_laddr/inp_lport and
494 * calling in_pcbinshash(), or they can just use the resulting
495 * port and address to authorise the sending of a once-off packet.
497 * On error, the values of *laddrp and *lportp are not changed.
500 in_pcbbind_setup(struct inpcb *inp, struct sockaddr *nam, in_addr_t *laddrp,
501 u_short *lportp, struct ucred *cred)
503 struct socket *so = inp->inp_socket;
504 struct sockaddr_in *sin;
505 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
506 struct in_addr laddr;
508 int lookupflags = 0, reuseport = (so->so_options & SO_REUSEPORT);
512 * No state changes, so read locks are sufficient here.
514 INP_LOCK_ASSERT(inp);
515 INP_HASH_LOCK_ASSERT(pcbinfo);
517 if (TAILQ_EMPTY(&V_in_ifaddrhead)) /* XXX broken! */
518 return (EADDRNOTAVAIL);
519 laddr.s_addr = *laddrp;
520 if (nam != NULL && laddr.s_addr != INADDR_ANY)
522 if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0)
523 lookupflags = INPLOOKUP_WILDCARD;
525 if ((error = prison_local_ip4(cred, &laddr)) != 0)
528 sin = (struct sockaddr_in *)nam;
529 if (nam->sa_len != sizeof (*sin))
533 * We should check the family, but old programs
534 * incorrectly fail to initialize it.
536 if (sin->sin_family != AF_INET)
537 return (EAFNOSUPPORT);
539 error = prison_local_ip4(cred, &sin->sin_addr);
542 if (sin->sin_port != *lportp) {
543 /* Don't allow the port to change. */
546 lport = sin->sin_port;
548 /* NB: lport is left as 0 if the port isn't being changed. */
549 if (IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) {
551 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
552 * allow complete duplication of binding if
553 * SO_REUSEPORT is set, or if SO_REUSEADDR is set
554 * and a multicast address is bound on both
555 * new and duplicated sockets.
557 if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) != 0)
558 reuseport = SO_REUSEADDR|SO_REUSEPORT;
559 } else if (sin->sin_addr.s_addr != INADDR_ANY) {
560 sin->sin_port = 0; /* yech... */
561 bzero(&sin->sin_zero, sizeof(sin->sin_zero));
563 * Is the address a local IP address?
564 * If INP_BINDANY is set, then the socket may be bound
565 * to any endpoint address, local or not.
567 if ((inp->inp_flags & INP_BINDANY) == 0 &&
568 ifa_ifwithaddr_check((struct sockaddr *)sin) == 0)
569 return (EADDRNOTAVAIL);
571 laddr = sin->sin_addr;
577 if (ntohs(lport) <= V_ipport_reservedhigh &&
578 ntohs(lport) >= V_ipport_reservedlow &&
579 priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT,
582 if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)) &&
583 priv_check_cred(inp->inp_cred,
584 PRIV_NETINET_REUSEPORT, 0) != 0) {
585 t = in_pcblookup_local(pcbinfo, sin->sin_addr,
586 lport, INPLOOKUP_WILDCARD, cred);
589 * This entire block sorely needs a rewrite.
592 ((t->inp_flags & INP_TIMEWAIT) == 0) &&
593 (so->so_type != SOCK_STREAM ||
594 ntohl(t->inp_faddr.s_addr) == INADDR_ANY) &&
595 (ntohl(sin->sin_addr.s_addr) != INADDR_ANY ||
596 ntohl(t->inp_laddr.s_addr) != INADDR_ANY ||
597 (t->inp_flags2 & INP_REUSEPORT) == 0) &&
598 (inp->inp_cred->cr_uid !=
599 t->inp_cred->cr_uid))
602 t = in_pcblookup_local(pcbinfo, sin->sin_addr,
603 lport, lookupflags, cred);
604 if (t && (t->inp_flags & INP_TIMEWAIT)) {
606 * XXXRW: If an incpb has had its timewait
607 * state recycled, we treat the address as
608 * being in use (for now). This is better
609 * than a panic, but not desirable.
613 (reuseport & tw->tw_so_options) == 0)
615 } else if (t && (reuseport & inp_so_options(t)) == 0) {
617 if (ntohl(sin->sin_addr.s_addr) !=
619 ntohl(t->inp_laddr.s_addr) !=
621 (inp->inp_vflag & INP_IPV6PROTO) == 0 ||
622 (t->inp_vflag & INP_IPV6PROTO) == 0)
631 error = in_pcb_lport(inp, &laddr, &lport, cred, lookupflags);
636 *laddrp = laddr.s_addr;
642 * Connect from a socket to a specified address.
643 * Both address and port must be specified in argument sin.
644 * If don't have a local address for this socket yet,
648 in_pcbconnect_mbuf(struct inpcb *inp, struct sockaddr *nam,
649 struct ucred *cred, struct mbuf *m)
651 u_short lport, fport;
652 in_addr_t laddr, faddr;
655 INP_WLOCK_ASSERT(inp);
656 INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo);
658 lport = inp->inp_lport;
659 laddr = inp->inp_laddr.s_addr;
660 anonport = (lport == 0);
661 error = in_pcbconnect_setup(inp, nam, &laddr, &lport, &faddr, &fport,
666 /* Do the initial binding of the local address if required. */
667 if (inp->inp_laddr.s_addr == INADDR_ANY && inp->inp_lport == 0) {
668 inp->inp_lport = lport;
669 inp->inp_laddr.s_addr = laddr;
670 if (in_pcbinshash(inp) != 0) {
671 inp->inp_laddr.s_addr = INADDR_ANY;
677 /* Commit the remaining changes. */
678 inp->inp_lport = lport;
679 inp->inp_laddr.s_addr = laddr;
680 inp->inp_faddr.s_addr = faddr;
681 inp->inp_fport = fport;
682 in_pcbrehash_mbuf(inp, m);
685 inp->inp_flags |= INP_ANONPORT;
690 in_pcbconnect(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred)
693 return (in_pcbconnect_mbuf(inp, nam, cred, NULL));
697 * Do proper source address selection on an unbound socket in case
698 * of connect. Take jails into account as well.
701 in_pcbladdr(struct inpcb *inp, struct in_addr *faddr, struct in_addr *laddr,
706 struct sockaddr_in *sin;
710 KASSERT(laddr != NULL, ("%s: laddr NULL", __func__));
713 * Bypass source address selection and use the primary jail IP
716 if (cred != NULL && !prison_saddrsel_ip4(cred, laddr))
720 bzero(&sro, sizeof(sro));
722 sin = (struct sockaddr_in *)&sro.ro_dst;
723 sin->sin_family = AF_INET;
724 sin->sin_len = sizeof(struct sockaddr_in);
725 sin->sin_addr.s_addr = faddr->s_addr;
728 * If route is known our src addr is taken from the i/f,
731 * Find out route to destination.
733 if ((inp->inp_socket->so_options & SO_DONTROUTE) == 0)
734 in_rtalloc_ign(&sro, 0, inp->inp_inc.inc_fibnum);
737 * If we found a route, use the address corresponding to
738 * the outgoing interface.
740 * Otherwise assume faddr is reachable on a directly connected
741 * network and try to find a corresponding interface to take
742 * the source address from.
744 if (sro.ro_rt == NULL || sro.ro_rt->rt_ifp == NULL) {
745 struct in_ifaddr *ia;
748 ia = ifatoia(ifa_ifwithdstaddr((struct sockaddr *)sin));
750 ia = ifatoia(ifa_ifwithnet((struct sockaddr *)sin, 0));
756 if (cred == NULL || !prison_flag(cred, PR_IP4)) {
757 laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
758 ifa_free(&ia->ia_ifa);
763 ifa_free(&ia->ia_ifa);
766 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
769 if (sa->sa_family != AF_INET)
771 sin = (struct sockaddr_in *)sa;
772 if (prison_check_ip4(cred, &sin->sin_addr) == 0) {
773 ia = (struct in_ifaddr *)ifa;
778 laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
779 IF_ADDR_RUNLOCK(ifp);
782 IF_ADDR_RUNLOCK(ifp);
784 /* 3. As a last resort return the 'default' jail address. */
785 error = prison_get_ip4(cred, laddr);
790 * If the outgoing interface on the route found is not
791 * a loopback interface, use the address from that interface.
792 * In case of jails do those three steps:
793 * 1. check if the interface address belongs to the jail. If so use it.
794 * 2. check if we have any address on the outgoing interface
795 * belonging to this jail. If so use it.
796 * 3. as a last resort return the 'default' jail address.
798 if ((sro.ro_rt->rt_ifp->if_flags & IFF_LOOPBACK) == 0) {
799 struct in_ifaddr *ia;
802 /* If not jailed, use the default returned. */
803 if (cred == NULL || !prison_flag(cred, PR_IP4)) {
804 ia = (struct in_ifaddr *)sro.ro_rt->rt_ifa;
805 laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
810 /* 1. Check if the iface address belongs to the jail. */
811 sin = (struct sockaddr_in *)sro.ro_rt->rt_ifa->ifa_addr;
812 if (prison_check_ip4(cred, &sin->sin_addr) == 0) {
813 ia = (struct in_ifaddr *)sro.ro_rt->rt_ifa;
814 laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
819 * 2. Check if we have any address on the outgoing interface
820 * belonging to this jail.
823 ifp = sro.ro_rt->rt_ifp;
825 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
827 if (sa->sa_family != AF_INET)
829 sin = (struct sockaddr_in *)sa;
830 if (prison_check_ip4(cred, &sin->sin_addr) == 0) {
831 ia = (struct in_ifaddr *)ifa;
836 laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
837 IF_ADDR_RUNLOCK(ifp);
840 IF_ADDR_RUNLOCK(ifp);
842 /* 3. As a last resort return the 'default' jail address. */
843 error = prison_get_ip4(cred, laddr);
848 * The outgoing interface is marked with 'loopback net', so a route
849 * to ourselves is here.
850 * Try to find the interface of the destination address and then
851 * take the address from there. That interface is not necessarily
852 * a loopback interface.
853 * In case of jails, check that it is an address of the jail
854 * and if we cannot find, fall back to the 'default' jail address.
856 if ((sro.ro_rt->rt_ifp->if_flags & IFF_LOOPBACK) != 0) {
857 struct sockaddr_in sain;
858 struct in_ifaddr *ia;
860 bzero(&sain, sizeof(struct sockaddr_in));
861 sain.sin_family = AF_INET;
862 sain.sin_len = sizeof(struct sockaddr_in);
863 sain.sin_addr.s_addr = faddr->s_addr;
865 ia = ifatoia(ifa_ifwithdstaddr(sintosa(&sain)));
867 ia = ifatoia(ifa_ifwithnet(sintosa(&sain), 0));
869 ia = ifatoia(ifa_ifwithaddr(sintosa(&sain)));
871 if (cred == NULL || !prison_flag(cred, PR_IP4)) {
876 laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
877 ifa_free(&ia->ia_ifa);
886 ifa_free(&ia->ia_ifa);
889 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
892 if (sa->sa_family != AF_INET)
894 sin = (struct sockaddr_in *)sa;
895 if (prison_check_ip4(cred,
896 &sin->sin_addr) == 0) {
897 ia = (struct in_ifaddr *)ifa;
902 laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
903 IF_ADDR_RUNLOCK(ifp);
906 IF_ADDR_RUNLOCK(ifp);
909 /* 3. As a last resort return the 'default' jail address. */
910 error = prison_get_ip4(cred, laddr);
915 if (sro.ro_rt != NULL)
921 * Set up for a connect from a socket to the specified address.
922 * On entry, *laddrp and *lportp should contain the current local
923 * address and port for the PCB; these are updated to the values
924 * that should be placed in inp_laddr and inp_lport to complete
927 * On success, *faddrp and *fportp will be set to the remote address
928 * and port. These are not updated in the error case.
930 * If the operation fails because the connection already exists,
931 * *oinpp will be set to the PCB of that connection so that the
932 * caller can decide to override it. In all other cases, *oinpp
936 in_pcbconnect_setup(struct inpcb *inp, struct sockaddr *nam,
937 in_addr_t *laddrp, u_short *lportp, in_addr_t *faddrp, u_short *fportp,
938 struct inpcb **oinpp, struct ucred *cred)
940 struct sockaddr_in *sin = (struct sockaddr_in *)nam;
941 struct in_ifaddr *ia;
943 struct in_addr laddr, faddr;
944 u_short lport, fport;
948 * Because a global state change doesn't actually occur here, a read
949 * lock is sufficient.
951 INP_LOCK_ASSERT(inp);
952 INP_HASH_LOCK_ASSERT(inp->inp_pcbinfo);
956 if (nam->sa_len != sizeof (*sin))
958 if (sin->sin_family != AF_INET)
959 return (EAFNOSUPPORT);
960 if (sin->sin_port == 0)
961 return (EADDRNOTAVAIL);
962 laddr.s_addr = *laddrp;
964 faddr = sin->sin_addr;
965 fport = sin->sin_port;
967 if (!TAILQ_EMPTY(&V_in_ifaddrhead)) {
969 * If the destination address is INADDR_ANY,
970 * use the primary local address.
971 * If the supplied address is INADDR_BROADCAST,
972 * and the primary interface supports broadcast,
973 * choose the broadcast address for that interface.
975 if (faddr.s_addr == INADDR_ANY) {
978 IA_SIN(TAILQ_FIRST(&V_in_ifaddrhead))->sin_addr;
981 (error = prison_get_ip4(cred, &faddr)) != 0)
983 } else if (faddr.s_addr == (u_long)INADDR_BROADCAST) {
985 if (TAILQ_FIRST(&V_in_ifaddrhead)->ia_ifp->if_flags &
987 faddr = satosin(&TAILQ_FIRST(
988 &V_in_ifaddrhead)->ia_broadaddr)->sin_addr;
992 if (laddr.s_addr == INADDR_ANY) {
993 error = in_pcbladdr(inp, &faddr, &laddr, cred);
995 * If the destination address is multicast and an outgoing
996 * interface has been set as a multicast option, prefer the
997 * address of that interface as our source address.
999 if (IN_MULTICAST(ntohl(faddr.s_addr)) &&
1000 inp->inp_moptions != NULL) {
1001 struct ip_moptions *imo;
1004 imo = inp->inp_moptions;
1005 if (imo->imo_multicast_ifp != NULL) {
1006 ifp = imo->imo_multicast_ifp;
1008 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
1009 if ((ia->ia_ifp == ifp) &&
1011 prison_check_ip4(cred,
1012 &ia->ia_addr.sin_addr) == 0))
1016 error = EADDRNOTAVAIL;
1018 laddr = ia->ia_addr.sin_addr;
1021 IN_IFADDR_RUNLOCK();
1027 oinp = in_pcblookup_hash_locked(inp->inp_pcbinfo, faddr, fport,
1028 laddr, lport, 0, NULL);
1032 return (EADDRINUSE);
1035 error = in_pcbbind_setup(inp, NULL, &laddr.s_addr, &lport,
1040 *laddrp = laddr.s_addr;
1042 *faddrp = faddr.s_addr;
1048 in_pcbdisconnect(struct inpcb *inp)
1051 INP_WLOCK_ASSERT(inp);
1052 INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo);
1054 inp->inp_faddr.s_addr = INADDR_ANY;
1061 * in_pcbdetach() is responsibe for disassociating a socket from an inpcb.
1062 * For most protocols, this will be invoked immediately prior to calling
1063 * in_pcbfree(). However, with TCP the inpcb may significantly outlive the
1064 * socket, in which case in_pcbfree() is deferred.
1067 in_pcbdetach(struct inpcb *inp)
1070 KASSERT(inp->inp_socket != NULL, ("%s: inp_socket == NULL", __func__));
1072 inp->inp_socket->so_pcb = NULL;
1073 inp->inp_socket = NULL;
1077 * in_pcbref() bumps the reference count on an inpcb in order to maintain
1078 * stability of an inpcb pointer despite the inpcb lock being released. This
1079 * is used in TCP when the inpcbinfo lock needs to be acquired or upgraded,
1080 * but where the inpcb lock may already held, or when acquiring a reference
1083 * in_pcbref() should be used only to provide brief memory stability, and
1084 * must always be followed by a call to INP_WLOCK() and in_pcbrele() to
1085 * garbage collect the inpcb if it has been in_pcbfree()'d from another
1086 * context. Until in_pcbrele() has returned that the inpcb is still valid,
1087 * lock and rele are the *only* safe operations that may be performed on the
1090 * While the inpcb will not be freed, releasing the inpcb lock means that the
1091 * connection's state may change, so the caller should be careful to
1092 * revalidate any cached state on reacquiring the lock. Drop the reference
1093 * using in_pcbrele().
1096 in_pcbref(struct inpcb *inp)
1099 KASSERT(inp->inp_refcount > 0, ("%s: refcount 0", __func__));
1101 refcount_acquire(&inp->inp_refcount);
1105 * Drop a refcount on an inpcb elevated using in_pcbref(); because a call to
1106 * in_pcbfree() may have been made between in_pcbref() and in_pcbrele(), we
1107 * return a flag indicating whether or not the inpcb remains valid. If it is
1108 * valid, we return with the inpcb lock held.
1110 * Notice that, unlike in_pcbref(), the inpcb lock must be held to drop a
1111 * reference on an inpcb. Historically more work was done here (actually, in
1112 * in_pcbfree_internal()) but has been moved to in_pcbfree() to avoid the
1113 * need for the pcbinfo lock in in_pcbrele(). Deferring the free is entirely
1114 * about memory stability (and continued use of the write lock).
1117 in_pcbrele_rlocked(struct inpcb *inp)
1119 struct inpcbinfo *pcbinfo;
1121 KASSERT(inp->inp_refcount > 0, ("%s: refcount 0", __func__));
1123 INP_RLOCK_ASSERT(inp);
1125 if (refcount_release(&inp->inp_refcount) == 0) {
1127 * If the inpcb has been freed, let the caller know, even if
1128 * this isn't the last reference.
1130 if (inp->inp_flags2 & INP_FREED) {
1137 KASSERT(inp->inp_socket == NULL, ("%s: inp_socket != NULL", __func__));
1140 pcbinfo = inp->inp_pcbinfo;
1141 uma_zfree(pcbinfo->ipi_zone, inp);
1146 in_pcbrele_wlocked(struct inpcb *inp)
1148 struct inpcbinfo *pcbinfo;
1150 KASSERT(inp->inp_refcount > 0, ("%s: refcount 0", __func__));
1152 INP_WLOCK_ASSERT(inp);
1154 if (refcount_release(&inp->inp_refcount) == 0)
1157 KASSERT(inp->inp_socket == NULL, ("%s: inp_socket != NULL", __func__));
1160 pcbinfo = inp->inp_pcbinfo;
1161 uma_zfree(pcbinfo->ipi_zone, inp);
1166 * Temporary wrapper.
1169 in_pcbrele(struct inpcb *inp)
1172 return (in_pcbrele_wlocked(inp));
1176 * Unconditionally schedule an inpcb to be freed by decrementing its
1177 * reference count, which should occur only after the inpcb has been detached
1178 * from its socket. If another thread holds a temporary reference (acquired
1179 * using in_pcbref()) then the free is deferred until that reference is
1180 * released using in_pcbrele(), but the inpcb is still unlocked. Almost all
1181 * work, including removal from global lists, is done in this context, where
1182 * the pcbinfo lock is held.
1185 in_pcbfree(struct inpcb *inp)
1187 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1189 KASSERT(inp->inp_socket == NULL, ("%s: inp_socket != NULL", __func__));
1191 INP_INFO_WLOCK_ASSERT(pcbinfo);
1192 INP_WLOCK_ASSERT(inp);
1194 /* XXXRW: Do as much as possible here. */
1196 if (inp->inp_sp != NULL)
1197 ipsec_delete_pcbpolicy(inp);
1199 inp->inp_gencnt = ++pcbinfo->ipi_gencnt;
1200 in_pcbremlists(inp);
1202 if (inp->inp_vflag & INP_IPV6PROTO) {
1203 ip6_freepcbopts(inp->in6p_outputopts);
1204 if (inp->in6p_moptions != NULL)
1205 ip6_freemoptions(inp->in6p_moptions);
1208 if (inp->inp_options)
1209 (void)m_free(inp->inp_options);
1211 if (inp->inp_moptions != NULL)
1212 inp_freemoptions(inp->inp_moptions);
1215 inp->inp_flags2 |= INP_FREED;
1216 crfree(inp->inp_cred);
1218 mac_inpcb_destroy(inp);
1220 if (!in_pcbrele_wlocked(inp))
1225 * in_pcbdrop() removes an inpcb from hashed lists, releasing its address and
1226 * port reservation, and preventing it from being returned by inpcb lookups.
1228 * It is used by TCP to mark an inpcb as unused and avoid future packet
1229 * delivery or event notification when a socket remains open but TCP has
1230 * closed. This might occur as a result of a shutdown()-initiated TCP close
1231 * or a RST on the wire, and allows the port binding to be reused while still
1232 * maintaining the invariant that so_pcb always points to a valid inpcb until
1235 * XXXRW: Possibly in_pcbdrop() should also prevent future notifications by
1236 * in_pcbnotifyall() and in_pcbpurgeif0()?
1239 in_pcbdrop(struct inpcb *inp)
1242 INP_WLOCK_ASSERT(inp);
1245 * XXXRW: Possibly we should protect the setting of INP_DROPPED with
1248 inp->inp_flags |= INP_DROPPED;
1249 if (inp->inp_flags & INP_INHASHLIST) {
1250 struct inpcbport *phd = inp->inp_phd;
1252 INP_HASH_WLOCK(inp->inp_pcbinfo);
1253 LIST_REMOVE(inp, inp_hash);
1254 LIST_REMOVE(inp, inp_portlist);
1255 if (LIST_FIRST(&phd->phd_pcblist) == NULL) {
1256 LIST_REMOVE(phd, phd_hash);
1259 INP_HASH_WUNLOCK(inp->inp_pcbinfo);
1260 inp->inp_flags &= ~INP_INHASHLIST;
1262 in_pcbgroup_remove(inp);
1269 * Common routines to return the socket addresses associated with inpcbs.
1272 in_sockaddr(in_port_t port, struct in_addr *addr_p)
1274 struct sockaddr_in *sin;
1276 sin = malloc(sizeof *sin, M_SONAME,
1278 sin->sin_family = AF_INET;
1279 sin->sin_len = sizeof(*sin);
1280 sin->sin_addr = *addr_p;
1281 sin->sin_port = port;
1283 return (struct sockaddr *)sin;
1287 in_getsockaddr(struct socket *so, struct sockaddr **nam)
1290 struct in_addr addr;
1293 inp = sotoinpcb(so);
1294 KASSERT(inp != NULL, ("in_getsockaddr: inp == NULL"));
1297 port = inp->inp_lport;
1298 addr = inp->inp_laddr;
1301 *nam = in_sockaddr(port, &addr);
1306 in_getpeeraddr(struct socket *so, struct sockaddr **nam)
1309 struct in_addr addr;
1312 inp = sotoinpcb(so);
1313 KASSERT(inp != NULL, ("in_getpeeraddr: inp == NULL"));
1316 port = inp->inp_fport;
1317 addr = inp->inp_faddr;
1320 *nam = in_sockaddr(port, &addr);
1325 in_pcbnotifyall(struct inpcbinfo *pcbinfo, struct in_addr faddr, int errno,
1326 struct inpcb *(*notify)(struct inpcb *, int))
1328 struct inpcb *inp, *inp_temp;
1330 INP_INFO_WLOCK(pcbinfo);
1331 LIST_FOREACH_SAFE(inp, pcbinfo->ipi_listhead, inp_list, inp_temp) {
1334 if ((inp->inp_vflag & INP_IPV4) == 0) {
1339 if (inp->inp_faddr.s_addr != faddr.s_addr ||
1340 inp->inp_socket == NULL) {
1344 if ((*notify)(inp, errno))
1347 INP_INFO_WUNLOCK(pcbinfo);
1351 in_pcbpurgeif0(struct inpcbinfo *pcbinfo, struct ifnet *ifp)
1354 struct ip_moptions *imo;
1357 INP_INFO_RLOCK(pcbinfo);
1358 LIST_FOREACH(inp, pcbinfo->ipi_listhead, inp_list) {
1360 imo = inp->inp_moptions;
1361 if ((inp->inp_vflag & INP_IPV4) &&
1364 * Unselect the outgoing interface if it is being
1367 if (imo->imo_multicast_ifp == ifp)
1368 imo->imo_multicast_ifp = NULL;
1371 * Drop multicast group membership if we joined
1372 * through the interface being detached.
1374 for (i = 0, gap = 0; i < imo->imo_num_memberships;
1376 if (imo->imo_membership[i]->inm_ifp == ifp) {
1377 in_delmulti(imo->imo_membership[i]);
1379 } else if (gap != 0)
1380 imo->imo_membership[i - gap] =
1381 imo->imo_membership[i];
1383 imo->imo_num_memberships -= gap;
1387 INP_INFO_RUNLOCK(pcbinfo);
1391 * Lookup a PCB based on the local address and port. Caller must hold the
1392 * hash lock. No inpcb locks or references are acquired.
1394 #define INP_LOOKUP_MAPPED_PCB_COST 3
1396 in_pcblookup_local(struct inpcbinfo *pcbinfo, struct in_addr laddr,
1397 u_short lport, int lookupflags, struct ucred *cred)
1401 int matchwild = 3 + INP_LOOKUP_MAPPED_PCB_COST;
1407 KASSERT((lookupflags & ~(INPLOOKUP_WILDCARD)) == 0,
1408 ("%s: invalid lookup flags %d", __func__, lookupflags));
1410 INP_HASH_LOCK_ASSERT(pcbinfo);
1412 if ((lookupflags & INPLOOKUP_WILDCARD) == 0) {
1413 struct inpcbhead *head;
1415 * Look for an unconnected (wildcard foreign addr) PCB that
1416 * matches the local address and port we're looking for.
1418 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport,
1419 0, pcbinfo->ipi_hashmask)];
1420 LIST_FOREACH(inp, head, inp_hash) {
1422 /* XXX inp locking */
1423 if ((inp->inp_vflag & INP_IPV4) == 0)
1426 if (inp->inp_faddr.s_addr == INADDR_ANY &&
1427 inp->inp_laddr.s_addr == laddr.s_addr &&
1428 inp->inp_lport == lport) {
1433 prison_equal_ip4(cred->cr_prison,
1434 inp->inp_cred->cr_prison))
1443 struct inpcbporthead *porthash;
1444 struct inpcbport *phd;
1445 struct inpcb *match = NULL;
1447 * Best fit PCB lookup.
1449 * First see if this local port is in use by looking on the
1452 porthash = &pcbinfo->ipi_porthashbase[INP_PCBPORTHASH(lport,
1453 pcbinfo->ipi_porthashmask)];
1454 LIST_FOREACH(phd, porthash, phd_hash) {
1455 if (phd->phd_port == lport)
1460 * Port is in use by one or more PCBs. Look for best
1463 LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) {
1466 !prison_equal_ip4(inp->inp_cred->cr_prison,
1470 /* XXX inp locking */
1471 if ((inp->inp_vflag & INP_IPV4) == 0)
1474 * We never select the PCB that has
1475 * INP_IPV6 flag and is bound to :: if
1476 * we have another PCB which is bound
1477 * to 0.0.0.0. If a PCB has the
1478 * INP_IPV6 flag, then we set its cost
1479 * higher than IPv4 only PCBs.
1481 * Note that the case only happens
1482 * when a socket is bound to ::, under
1483 * the condition that the use of the
1484 * mapped address is allowed.
1486 if ((inp->inp_vflag & INP_IPV6) != 0)
1487 wildcard += INP_LOOKUP_MAPPED_PCB_COST;
1489 if (inp->inp_faddr.s_addr != INADDR_ANY)
1491 if (inp->inp_laddr.s_addr != INADDR_ANY) {
1492 if (laddr.s_addr == INADDR_ANY)
1494 else if (inp->inp_laddr.s_addr != laddr.s_addr)
1497 if (laddr.s_addr != INADDR_ANY)
1500 if (wildcard < matchwild) {
1502 matchwild = wildcard;
1511 #undef INP_LOOKUP_MAPPED_PCB_COST
1515 * Lookup PCB in hash list, using pcbgroup tables.
1517 static struct inpcb *
1518 in_pcblookup_group(struct inpcbinfo *pcbinfo, struct inpcbgroup *pcbgroup,
1519 struct in_addr faddr, u_int fport_arg, struct in_addr laddr,
1520 u_int lport_arg, int lookupflags, struct ifnet *ifp)
1522 struct inpcbhead *head;
1523 struct inpcb *inp, *tmpinp;
1524 u_short fport = fport_arg, lport = lport_arg;
1527 * First look for an exact match.
1530 INP_GROUP_LOCK(pcbgroup);
1531 head = &pcbgroup->ipg_hashbase[INP_PCBHASH(faddr.s_addr, lport, fport,
1532 pcbgroup->ipg_hashmask)];
1533 LIST_FOREACH(inp, head, inp_pcbgrouphash) {
1535 /* XXX inp locking */
1536 if ((inp->inp_vflag & INP_IPV4) == 0)
1539 if (inp->inp_faddr.s_addr == faddr.s_addr &&
1540 inp->inp_laddr.s_addr == laddr.s_addr &&
1541 inp->inp_fport == fport &&
1542 inp->inp_lport == lport) {
1544 * XXX We should be able to directly return
1545 * the inp here, without any checks.
1546 * Well unless both bound with SO_REUSEPORT?
1548 if (prison_flag(inp->inp_cred, PR_IP4))
1554 if (tmpinp != NULL) {
1560 * Then look for a wildcard match, if requested.
1562 if ((lookupflags & INPLOOKUP_WILDCARD) != 0) {
1563 struct inpcb *local_wild = NULL, *local_exact = NULL;
1565 struct inpcb *local_wild_mapped = NULL;
1567 struct inpcb *jail_wild = NULL;
1568 struct inpcbhead *head;
1572 * Order of socket selection - we always prefer jails.
1573 * 1. jailed, non-wild.
1575 * 3. non-jailed, non-wild.
1576 * 4. non-jailed, wild.
1578 head = &pcbinfo->ipi_wildbase[INP_PCBHASH(INADDR_ANY, lport,
1579 0, pcbinfo->ipi_wildmask)];
1580 LIST_FOREACH(inp, head, inp_pcbgroup_wild) {
1582 /* XXX inp locking */
1583 if ((inp->inp_vflag & INP_IPV4) == 0)
1586 if (inp->inp_faddr.s_addr != INADDR_ANY ||
1587 inp->inp_lport != lport)
1590 /* XXX inp locking */
1591 if (ifp && ifp->if_type == IFT_FAITH &&
1592 (inp->inp_flags & INP_FAITH) == 0)
1595 injail = prison_flag(inp->inp_cred, PR_IP4);
1597 if (prison_check_ip4(inp->inp_cred,
1601 if (local_exact != NULL)
1605 if (inp->inp_laddr.s_addr == laddr.s_addr) {
1610 } else if (inp->inp_laddr.s_addr == INADDR_ANY) {
1612 /* XXX inp locking, NULL check */
1613 if (inp->inp_vflag & INP_IPV6PROTO)
1614 local_wild_mapped = inp;
1622 } /* LIST_FOREACH */
1630 inp = local_wild_mapped;
1634 } /* if (lookupflags & INPLOOKUP_WILDCARD) */
1635 INP_GROUP_UNLOCK(pcbgroup);
1640 INP_GROUP_UNLOCK(pcbgroup);
1641 if (lookupflags & INPLOOKUP_WLOCKPCB) {
1643 if (in_pcbrele_wlocked(inp))
1645 } else if (lookupflags & INPLOOKUP_RLOCKPCB) {
1647 if (in_pcbrele_rlocked(inp))
1650 panic("%s: locking bug", __func__);
1653 #endif /* PCBGROUP */
1656 * Lookup PCB in hash list, using pcbinfo tables. This variation assumes
1657 * that the caller has locked the hash list, and will not perform any further
1658 * locking or reference operations on either the hash list or the connection.
1660 static struct inpcb *
1661 in_pcblookup_hash_locked(struct inpcbinfo *pcbinfo, struct in_addr faddr,
1662 u_int fport_arg, struct in_addr laddr, u_int lport_arg, int lookupflags,
1665 struct inpcbhead *head;
1666 struct inpcb *inp, *tmpinp;
1667 u_short fport = fport_arg, lport = lport_arg;
1669 KASSERT((lookupflags & ~(INPLOOKUP_WILDCARD)) == 0,
1670 ("%s: invalid lookup flags %d", __func__, lookupflags));
1672 INP_HASH_LOCK_ASSERT(pcbinfo);
1675 * First look for an exact match.
1678 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(faddr.s_addr, lport, fport,
1679 pcbinfo->ipi_hashmask)];
1680 LIST_FOREACH(inp, head, inp_hash) {
1682 /* XXX inp locking */
1683 if ((inp->inp_vflag & INP_IPV4) == 0)
1686 if (inp->inp_faddr.s_addr == faddr.s_addr &&
1687 inp->inp_laddr.s_addr == laddr.s_addr &&
1688 inp->inp_fport == fport &&
1689 inp->inp_lport == lport) {
1691 * XXX We should be able to directly return
1692 * the inp here, without any checks.
1693 * Well unless both bound with SO_REUSEPORT?
1695 if (prison_flag(inp->inp_cred, PR_IP4))
1705 * Then look for a wildcard match, if requested.
1707 if ((lookupflags & INPLOOKUP_WILDCARD) != 0) {
1708 struct inpcb *local_wild = NULL, *local_exact = NULL;
1710 struct inpcb *local_wild_mapped = NULL;
1712 struct inpcb *jail_wild = NULL;
1716 * Order of socket selection - we always prefer jails.
1717 * 1. jailed, non-wild.
1719 * 3. non-jailed, non-wild.
1720 * 4. non-jailed, wild.
1723 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport,
1724 0, pcbinfo->ipi_hashmask)];
1725 LIST_FOREACH(inp, head, inp_hash) {
1727 /* XXX inp locking */
1728 if ((inp->inp_vflag & INP_IPV4) == 0)
1731 if (inp->inp_faddr.s_addr != INADDR_ANY ||
1732 inp->inp_lport != lport)
1735 /* XXX inp locking */
1736 if (ifp && ifp->if_type == IFT_FAITH &&
1737 (inp->inp_flags & INP_FAITH) == 0)
1740 injail = prison_flag(inp->inp_cred, PR_IP4);
1742 if (prison_check_ip4(inp->inp_cred,
1746 if (local_exact != NULL)
1750 if (inp->inp_laddr.s_addr == laddr.s_addr) {
1755 } else if (inp->inp_laddr.s_addr == INADDR_ANY) {
1757 /* XXX inp locking, NULL check */
1758 if (inp->inp_vflag & INP_IPV6PROTO)
1759 local_wild_mapped = inp;
1767 } /* LIST_FOREACH */
1768 if (jail_wild != NULL)
1770 if (local_exact != NULL)
1771 return (local_exact);
1772 if (local_wild != NULL)
1773 return (local_wild);
1775 if (local_wild_mapped != NULL)
1776 return (local_wild_mapped);
1778 } /* if ((lookupflags & INPLOOKUP_WILDCARD) != 0) */
1784 * Lookup PCB in hash list, using pcbinfo tables. This variation locks the
1785 * hash list lock, and will return the inpcb locked (i.e., requires
1786 * INPLOOKUP_LOCKPCB).
1788 static struct inpcb *
1789 in_pcblookup_hash(struct inpcbinfo *pcbinfo, struct in_addr faddr,
1790 u_int fport, struct in_addr laddr, u_int lport, int lookupflags,
1795 INP_HASH_RLOCK(pcbinfo);
1796 inp = in_pcblookup_hash_locked(pcbinfo, faddr, fport, laddr, lport,
1797 (lookupflags & ~(INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)), ifp);
1800 INP_HASH_RUNLOCK(pcbinfo);
1801 if (lookupflags & INPLOOKUP_WLOCKPCB) {
1803 if (in_pcbrele_wlocked(inp))
1805 } else if (lookupflags & INPLOOKUP_RLOCKPCB) {
1807 if (in_pcbrele_rlocked(inp))
1810 panic("%s: locking bug", __func__);
1812 INP_HASH_RUNLOCK(pcbinfo);
1817 * Public inpcb lookup routines, accepting a 4-tuple, and optionally, an mbuf
1818 * from which a pre-calculated hash value may be extracted.
1820 * Possibly more of this logic should be in in_pcbgroup.c.
1823 in_pcblookup(struct inpcbinfo *pcbinfo, struct in_addr faddr, u_int fport,
1824 struct in_addr laddr, u_int lport, int lookupflags, struct ifnet *ifp)
1826 #if defined(PCBGROUP) && !defined(RSS)
1827 struct inpcbgroup *pcbgroup;
1830 KASSERT((lookupflags & ~INPLOOKUP_MASK) == 0,
1831 ("%s: invalid lookup flags %d", __func__, lookupflags));
1832 KASSERT((lookupflags & (INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)) != 0,
1833 ("%s: LOCKPCB not set", __func__));
1836 * When not using RSS, use connection groups in preference to the
1837 * reservation table when looking up 4-tuples. When using RSS, just
1838 * use the reservation table, due to the cost of the Toeplitz hash
1841 * XXXRW: This policy belongs in the pcbgroup code, as in principle
1842 * we could be doing RSS with a non-Toeplitz hash that is affordable
1845 #if defined(PCBGROUP) && !defined(RSS)
1846 if (in_pcbgroup_enabled(pcbinfo)) {
1847 pcbgroup = in_pcbgroup_bytuple(pcbinfo, laddr, lport, faddr,
1849 return (in_pcblookup_group(pcbinfo, pcbgroup, faddr, fport,
1850 laddr, lport, lookupflags, ifp));
1853 return (in_pcblookup_hash(pcbinfo, faddr, fport, laddr, lport,
1858 in_pcblookup_mbuf(struct inpcbinfo *pcbinfo, struct in_addr faddr,
1859 u_int fport, struct in_addr laddr, u_int lport, int lookupflags,
1860 struct ifnet *ifp, struct mbuf *m)
1863 struct inpcbgroup *pcbgroup;
1866 KASSERT((lookupflags & ~INPLOOKUP_MASK) == 0,
1867 ("%s: invalid lookup flags %d", __func__, lookupflags));
1868 KASSERT((lookupflags & (INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)) != 0,
1869 ("%s: LOCKPCB not set", __func__));
1873 * If we can use a hardware-generated hash to look up the connection
1874 * group, use that connection group to find the inpcb. Otherwise
1875 * fall back on a software hash -- or the reservation table if we're
1878 * XXXRW: As above, that policy belongs in the pcbgroup code.
1880 if (in_pcbgroup_enabled(pcbinfo) &&
1881 !(M_HASHTYPE_TEST(m, M_HASHTYPE_NONE))) {
1882 pcbgroup = in_pcbgroup_byhash(pcbinfo, M_HASHTYPE_GET(m),
1883 m->m_pkthdr.flowid);
1884 if (pcbgroup != NULL)
1885 return (in_pcblookup_group(pcbinfo, pcbgroup, faddr,
1886 fport, laddr, lport, lookupflags, ifp));
1888 pcbgroup = in_pcbgroup_bytuple(pcbinfo, laddr, lport, faddr,
1890 return (in_pcblookup_group(pcbinfo, pcbgroup, faddr, fport,
1891 laddr, lport, lookupflags, ifp));
1895 return (in_pcblookup_hash(pcbinfo, faddr, fport, laddr, lport,
1901 * Insert PCB onto various hash lists.
1904 in_pcbinshash_internal(struct inpcb *inp, int do_pcbgroup_update)
1906 struct inpcbhead *pcbhash;
1907 struct inpcbporthead *pcbporthash;
1908 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1909 struct inpcbport *phd;
1910 u_int32_t hashkey_faddr;
1912 INP_WLOCK_ASSERT(inp);
1913 INP_HASH_WLOCK_ASSERT(pcbinfo);
1915 KASSERT((inp->inp_flags & INP_INHASHLIST) == 0,
1916 ("in_pcbinshash: INP_INHASHLIST"));
1919 if (inp->inp_vflag & INP_IPV6)
1920 hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX */;
1923 hashkey_faddr = inp->inp_faddr.s_addr;
1925 pcbhash = &pcbinfo->ipi_hashbase[INP_PCBHASH(hashkey_faddr,
1926 inp->inp_lport, inp->inp_fport, pcbinfo->ipi_hashmask)];
1928 pcbporthash = &pcbinfo->ipi_porthashbase[
1929 INP_PCBPORTHASH(inp->inp_lport, pcbinfo->ipi_porthashmask)];
1932 * Go through port list and look for a head for this lport.
1934 LIST_FOREACH(phd, pcbporthash, phd_hash) {
1935 if (phd->phd_port == inp->inp_lport)
1939 * If none exists, malloc one and tack it on.
1942 phd = malloc(sizeof(struct inpcbport), M_PCB, M_NOWAIT);
1944 return (ENOBUFS); /* XXX */
1946 phd->phd_port = inp->inp_lport;
1947 LIST_INIT(&phd->phd_pcblist);
1948 LIST_INSERT_HEAD(pcbporthash, phd, phd_hash);
1951 LIST_INSERT_HEAD(&phd->phd_pcblist, inp, inp_portlist);
1952 LIST_INSERT_HEAD(pcbhash, inp, inp_hash);
1953 inp->inp_flags |= INP_INHASHLIST;
1955 if (do_pcbgroup_update)
1956 in_pcbgroup_update(inp);
1962 * For now, there are two public interfaces to insert an inpcb into the hash
1963 * lists -- one that does update pcbgroups, and one that doesn't. The latter
1964 * is used only in the TCP syncache, where in_pcbinshash is called before the
1965 * full 4-tuple is set for the inpcb, and we don't want to install in the
1966 * pcbgroup until later.
1968 * XXXRW: This seems like a misfeature. in_pcbinshash should always update
1969 * connection groups, and partially initialised inpcbs should not be exposed
1970 * to either reservation hash tables or pcbgroups.
1973 in_pcbinshash(struct inpcb *inp)
1976 return (in_pcbinshash_internal(inp, 1));
1980 in_pcbinshash_nopcbgroup(struct inpcb *inp)
1983 return (in_pcbinshash_internal(inp, 0));
1987 * Move PCB to the proper hash bucket when { faddr, fport } have been
1988 * changed. NOTE: This does not handle the case of the lport changing (the
1989 * hashed port list would have to be updated as well), so the lport must
1990 * not change after in_pcbinshash() has been called.
1993 in_pcbrehash_mbuf(struct inpcb *inp, struct mbuf *m)
1995 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1996 struct inpcbhead *head;
1997 u_int32_t hashkey_faddr;
1999 INP_WLOCK_ASSERT(inp);
2000 INP_HASH_WLOCK_ASSERT(pcbinfo);
2002 KASSERT(inp->inp_flags & INP_INHASHLIST,
2003 ("in_pcbrehash: !INP_INHASHLIST"));
2006 if (inp->inp_vflag & INP_IPV6)
2007 hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX */;
2010 hashkey_faddr = inp->inp_faddr.s_addr;
2012 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(hashkey_faddr,
2013 inp->inp_lport, inp->inp_fport, pcbinfo->ipi_hashmask)];
2015 LIST_REMOVE(inp, inp_hash);
2016 LIST_INSERT_HEAD(head, inp, inp_hash);
2020 in_pcbgroup_update_mbuf(inp, m);
2022 in_pcbgroup_update(inp);
2027 in_pcbrehash(struct inpcb *inp)
2030 in_pcbrehash_mbuf(inp, NULL);
2034 * Remove PCB from various lists.
2037 in_pcbremlists(struct inpcb *inp)
2039 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
2041 INP_INFO_WLOCK_ASSERT(pcbinfo);
2042 INP_WLOCK_ASSERT(inp);
2044 inp->inp_gencnt = ++pcbinfo->ipi_gencnt;
2045 if (inp->inp_flags & INP_INHASHLIST) {
2046 struct inpcbport *phd = inp->inp_phd;
2048 INP_HASH_WLOCK(pcbinfo);
2049 LIST_REMOVE(inp, inp_hash);
2050 LIST_REMOVE(inp, inp_portlist);
2051 if (LIST_FIRST(&phd->phd_pcblist) == NULL) {
2052 LIST_REMOVE(phd, phd_hash);
2055 INP_HASH_WUNLOCK(pcbinfo);
2056 inp->inp_flags &= ~INP_INHASHLIST;
2058 LIST_REMOVE(inp, inp_list);
2059 pcbinfo->ipi_count--;
2061 in_pcbgroup_remove(inp);
2066 * A set label operation has occurred at the socket layer, propagate the
2067 * label change into the in_pcb for the socket.
2070 in_pcbsosetlabel(struct socket *so)
2075 inp = sotoinpcb(so);
2076 KASSERT(inp != NULL, ("in_pcbsosetlabel: so->so_pcb == NULL"));
2080 mac_inpcb_sosetlabel(so, inp);
2087 * ipport_tick runs once per second, determining if random port allocation
2088 * should be continued. If more than ipport_randomcps ports have been
2089 * allocated in the last second, then we return to sequential port
2090 * allocation. We return to random allocation only once we drop below
2091 * ipport_randomcps for at least ipport_randomtime seconds.
2094 ipport_tick(void *xtp)
2096 VNET_ITERATOR_DECL(vnet_iter);
2098 VNET_LIST_RLOCK_NOSLEEP();
2099 VNET_FOREACH(vnet_iter) {
2100 CURVNET_SET(vnet_iter); /* XXX appease INVARIANTS here */
2101 if (V_ipport_tcpallocs <=
2102 V_ipport_tcplastcount + V_ipport_randomcps) {
2103 if (V_ipport_stoprandom > 0)
2104 V_ipport_stoprandom--;
2106 V_ipport_stoprandom = V_ipport_randomtime;
2107 V_ipport_tcplastcount = V_ipport_tcpallocs;
2110 VNET_LIST_RUNLOCK_NOSLEEP();
2111 callout_reset(&ipport_tick_callout, hz, ipport_tick, NULL);
2118 callout_stop(&ipport_tick_callout);
2122 * The ipport_callout should start running at about the time we attach the
2123 * inet or inet6 domains.
2126 ipport_tick_init(const void *unused __unused)
2129 /* Start ipport_tick. */
2130 callout_init(&ipport_tick_callout, CALLOUT_MPSAFE);
2131 callout_reset(&ipport_tick_callout, 1, ipport_tick, NULL);
2132 EVENTHANDLER_REGISTER(shutdown_pre_sync, ip_fini, NULL,
2133 SHUTDOWN_PRI_DEFAULT);
2135 SYSINIT(ipport_tick_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE,
2136 ipport_tick_init, NULL);
2139 inp_wlock(struct inpcb *inp)
2146 inp_wunlock(struct inpcb *inp)
2153 inp_rlock(struct inpcb *inp)
2160 inp_runlock(struct inpcb *inp)
2168 inp_lock_assert(struct inpcb *inp)
2171 INP_WLOCK_ASSERT(inp);
2175 inp_unlock_assert(struct inpcb *inp)
2178 INP_UNLOCK_ASSERT(inp);
2183 inp_apply_all(void (*func)(struct inpcb *, void *), void *arg)
2187 INP_INFO_RLOCK(&V_tcbinfo);
2188 LIST_FOREACH(inp, V_tcbinfo.ipi_listhead, inp_list) {
2193 INP_INFO_RUNLOCK(&V_tcbinfo);
2197 inp_inpcbtosocket(struct inpcb *inp)
2200 INP_WLOCK_ASSERT(inp);
2201 return (inp->inp_socket);
2205 inp_inpcbtotcpcb(struct inpcb *inp)
2208 INP_WLOCK_ASSERT(inp);
2209 return ((struct tcpcb *)inp->inp_ppcb);
2213 inp_ip_tos_get(const struct inpcb *inp)
2216 return (inp->inp_ip_tos);
2220 inp_ip_tos_set(struct inpcb *inp, int val)
2223 inp->inp_ip_tos = val;
2227 inp_4tuple_get(struct inpcb *inp, uint32_t *laddr, uint16_t *lp,
2228 uint32_t *faddr, uint16_t *fp)
2231 INP_LOCK_ASSERT(inp);
2232 *laddr = inp->inp_laddr.s_addr;
2233 *faddr = inp->inp_faddr.s_addr;
2234 *lp = inp->inp_lport;
2235 *fp = inp->inp_fport;
2239 so_sotoinpcb(struct socket *so)
2242 return (sotoinpcb(so));
2246 so_sototcpcb(struct socket *so)
2249 return (sototcpcb(so));
2254 db_print_indent(int indent)
2258 for (i = 0; i < indent; i++)
2263 db_print_inconninfo(struct in_conninfo *inc, const char *name, int indent)
2265 char faddr_str[48], laddr_str[48];
2267 db_print_indent(indent);
2268 db_printf("%s at %p\n", name, inc);
2273 if (inc->inc_flags & INC_ISIPV6) {
2275 ip6_sprintf(laddr_str, &inc->inc6_laddr);
2276 ip6_sprintf(faddr_str, &inc->inc6_faddr);
2281 inet_ntoa_r(inc->inc_laddr, laddr_str);
2282 inet_ntoa_r(inc->inc_faddr, faddr_str);
2284 db_print_indent(indent);
2285 db_printf("inc_laddr %s inc_lport %u\n", laddr_str,
2286 ntohs(inc->inc_lport));
2287 db_print_indent(indent);
2288 db_printf("inc_faddr %s inc_fport %u\n", faddr_str,
2289 ntohs(inc->inc_fport));
2293 db_print_inpflags(int inp_flags)
2298 if (inp_flags & INP_RECVOPTS) {
2299 db_printf("%sINP_RECVOPTS", comma ? ", " : "");
2302 if (inp_flags & INP_RECVRETOPTS) {
2303 db_printf("%sINP_RECVRETOPTS", comma ? ", " : "");
2306 if (inp_flags & INP_RECVDSTADDR) {
2307 db_printf("%sINP_RECVDSTADDR", comma ? ", " : "");
2310 if (inp_flags & INP_HDRINCL) {
2311 db_printf("%sINP_HDRINCL", comma ? ", " : "");
2314 if (inp_flags & INP_HIGHPORT) {
2315 db_printf("%sINP_HIGHPORT", comma ? ", " : "");
2318 if (inp_flags & INP_LOWPORT) {
2319 db_printf("%sINP_LOWPORT", comma ? ", " : "");
2322 if (inp_flags & INP_ANONPORT) {
2323 db_printf("%sINP_ANONPORT", comma ? ", " : "");
2326 if (inp_flags & INP_RECVIF) {
2327 db_printf("%sINP_RECVIF", comma ? ", " : "");
2330 if (inp_flags & INP_MTUDISC) {
2331 db_printf("%sINP_MTUDISC", comma ? ", " : "");
2334 if (inp_flags & INP_FAITH) {
2335 db_printf("%sINP_FAITH", comma ? ", " : "");
2338 if (inp_flags & INP_RECVTTL) {
2339 db_printf("%sINP_RECVTTL", comma ? ", " : "");
2342 if (inp_flags & INP_DONTFRAG) {
2343 db_printf("%sINP_DONTFRAG", comma ? ", " : "");
2346 if (inp_flags & INP_RECVTOS) {
2347 db_printf("%sINP_RECVTOS", comma ? ", " : "");
2350 if (inp_flags & IN6P_IPV6_V6ONLY) {
2351 db_printf("%sIN6P_IPV6_V6ONLY", comma ? ", " : "");
2354 if (inp_flags & IN6P_PKTINFO) {
2355 db_printf("%sIN6P_PKTINFO", comma ? ", " : "");
2358 if (inp_flags & IN6P_HOPLIMIT) {
2359 db_printf("%sIN6P_HOPLIMIT", comma ? ", " : "");
2362 if (inp_flags & IN6P_HOPOPTS) {
2363 db_printf("%sIN6P_HOPOPTS", comma ? ", " : "");
2366 if (inp_flags & IN6P_DSTOPTS) {
2367 db_printf("%sIN6P_DSTOPTS", comma ? ", " : "");
2370 if (inp_flags & IN6P_RTHDR) {
2371 db_printf("%sIN6P_RTHDR", comma ? ", " : "");
2374 if (inp_flags & IN6P_RTHDRDSTOPTS) {
2375 db_printf("%sIN6P_RTHDRDSTOPTS", comma ? ", " : "");
2378 if (inp_flags & IN6P_TCLASS) {
2379 db_printf("%sIN6P_TCLASS", comma ? ", " : "");
2382 if (inp_flags & IN6P_AUTOFLOWLABEL) {
2383 db_printf("%sIN6P_AUTOFLOWLABEL", comma ? ", " : "");
2386 if (inp_flags & INP_TIMEWAIT) {
2387 db_printf("%sINP_TIMEWAIT", comma ? ", " : "");
2390 if (inp_flags & INP_ONESBCAST) {
2391 db_printf("%sINP_ONESBCAST", comma ? ", " : "");
2394 if (inp_flags & INP_DROPPED) {
2395 db_printf("%sINP_DROPPED", comma ? ", " : "");
2398 if (inp_flags & INP_SOCKREF) {
2399 db_printf("%sINP_SOCKREF", comma ? ", " : "");
2402 if (inp_flags & IN6P_RFC2292) {
2403 db_printf("%sIN6P_RFC2292", comma ? ", " : "");
2406 if (inp_flags & IN6P_MTU) {
2407 db_printf("IN6P_MTU%s", comma ? ", " : "");
2413 db_print_inpvflag(u_char inp_vflag)
2418 if (inp_vflag & INP_IPV4) {
2419 db_printf("%sINP_IPV4", comma ? ", " : "");
2422 if (inp_vflag & INP_IPV6) {
2423 db_printf("%sINP_IPV6", comma ? ", " : "");
2426 if (inp_vflag & INP_IPV6PROTO) {
2427 db_printf("%sINP_IPV6PROTO", comma ? ", " : "");
2433 db_print_inpcb(struct inpcb *inp, const char *name, int indent)
2436 db_print_indent(indent);
2437 db_printf("%s at %p\n", name, inp);
2441 db_print_indent(indent);
2442 db_printf("inp_flow: 0x%x\n", inp->inp_flow);
2444 db_print_inconninfo(&inp->inp_inc, "inp_conninfo", indent);
2446 db_print_indent(indent);
2447 db_printf("inp_ppcb: %p inp_pcbinfo: %p inp_socket: %p\n",
2448 inp->inp_ppcb, inp->inp_pcbinfo, inp->inp_socket);
2450 db_print_indent(indent);
2451 db_printf("inp_label: %p inp_flags: 0x%x (",
2452 inp->inp_label, inp->inp_flags);
2453 db_print_inpflags(inp->inp_flags);
2456 db_print_indent(indent);
2457 db_printf("inp_sp: %p inp_vflag: 0x%x (", inp->inp_sp,
2459 db_print_inpvflag(inp->inp_vflag);
2462 db_print_indent(indent);
2463 db_printf("inp_ip_ttl: %d inp_ip_p: %d inp_ip_minttl: %d\n",
2464 inp->inp_ip_ttl, inp->inp_ip_p, inp->inp_ip_minttl);
2466 db_print_indent(indent);
2468 if (inp->inp_vflag & INP_IPV6) {
2469 db_printf("in6p_options: %p in6p_outputopts: %p "
2470 "in6p_moptions: %p\n", inp->in6p_options,
2471 inp->in6p_outputopts, inp->in6p_moptions);
2472 db_printf("in6p_icmp6filt: %p in6p_cksum %d "
2473 "in6p_hops %u\n", inp->in6p_icmp6filt, inp->in6p_cksum,
2478 db_printf("inp_ip_tos: %d inp_ip_options: %p "
2479 "inp_ip_moptions: %p\n", inp->inp_ip_tos,
2480 inp->inp_options, inp->inp_moptions);
2483 db_print_indent(indent);
2484 db_printf("inp_phd: %p inp_gencnt: %ju\n", inp->inp_phd,
2485 (uintmax_t)inp->inp_gencnt);
2488 DB_SHOW_COMMAND(inpcb, db_show_inpcb)
2493 db_printf("usage: show inpcb <addr>\n");
2496 inp = (struct inpcb *)addr;
2498 db_print_inpcb(inp, "inpcb", 0);