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_PROC(_net_inet_ip_portrange, OID_AUTO, lowfirst,
168 CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW,
169 &VNET_NAME(ipport_lowfirstauto), 0, &sysctl_net_ipport_check, "I", "");
170 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, lowlast,
171 CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW,
172 &VNET_NAME(ipport_lowlastauto), 0, &sysctl_net_ipport_check, "I", "");
173 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, first,
174 CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW,
175 &VNET_NAME(ipport_firstauto), 0, &sysctl_net_ipport_check, "I", "");
176 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, last,
177 CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW,
178 &VNET_NAME(ipport_lastauto), 0, &sysctl_net_ipport_check, "I", "");
179 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, hifirst,
180 CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW,
181 &VNET_NAME(ipport_hifirstauto), 0, &sysctl_net_ipport_check, "I", "");
182 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, hilast,
183 CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW,
184 &VNET_NAME(ipport_hilastauto), 0, &sysctl_net_ipport_check, "I", "");
185 SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, reservedhigh,
186 CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE,
187 &VNET_NAME(ipport_reservedhigh), 0, "");
188 SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, reservedlow,
189 CTLFLAG_RW|CTLFLAG_SECURE, &VNET_NAME(ipport_reservedlow), 0, "");
190 SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, randomized,
191 CTLFLAG_VNET | CTLFLAG_RW,
192 &VNET_NAME(ipport_randomized), 0, "Enable random port allocation");
193 SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, randomcps,
194 CTLFLAG_VNET | CTLFLAG_RW,
195 &VNET_NAME(ipport_randomcps), 0, "Maximum number of random port "
196 "allocations before switching to a sequental one");
197 SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, randomtime,
198 CTLFLAG_VNET | CTLFLAG_RW,
199 &VNET_NAME(ipport_randomtime), 0,
200 "Minimum time to keep sequental port "
201 "allocation before switching to a random one");
205 * in_pcb.c: manage the Protocol Control Blocks.
207 * NOTE: It is assumed that most of these functions will be called with
208 * the pcbinfo lock held, and often, the inpcb lock held, as these utility
209 * functions often modify hash chains or addresses in pcbs.
213 * Initialize an inpcbinfo -- we should be able to reduce the number of
217 in_pcbinfo_init(struct inpcbinfo *pcbinfo, const char *name,
218 struct inpcbhead *listhead, int hash_nelements, int porthash_nelements,
219 char *inpcbzone_name, uma_init inpcbzone_init, uma_fini inpcbzone_fini,
220 uint32_t inpcbzone_flags, u_int hashfields)
223 INP_INFO_LOCK_INIT(pcbinfo, name);
224 INP_HASH_LOCK_INIT(pcbinfo, "pcbinfohash"); /* XXXRW: argument? */
226 pcbinfo->ipi_vnet = curvnet;
228 pcbinfo->ipi_listhead = listhead;
229 LIST_INIT(pcbinfo->ipi_listhead);
230 pcbinfo->ipi_count = 0;
231 pcbinfo->ipi_hashbase = hashinit(hash_nelements, M_PCB,
232 &pcbinfo->ipi_hashmask);
233 pcbinfo->ipi_porthashbase = hashinit(porthash_nelements, M_PCB,
234 &pcbinfo->ipi_porthashmask);
236 in_pcbgroup_init(pcbinfo, hashfields, hash_nelements);
238 pcbinfo->ipi_zone = uma_zcreate(inpcbzone_name, sizeof(struct inpcb),
239 NULL, NULL, inpcbzone_init, inpcbzone_fini, UMA_ALIGN_PTR,
241 uma_zone_set_max(pcbinfo->ipi_zone, maxsockets);
242 uma_zone_set_warning(pcbinfo->ipi_zone,
243 "kern.ipc.maxsockets limit reached");
247 * Destroy an inpcbinfo.
250 in_pcbinfo_destroy(struct inpcbinfo *pcbinfo)
253 KASSERT(pcbinfo->ipi_count == 0,
254 ("%s: ipi_count = %u", __func__, pcbinfo->ipi_count));
256 hashdestroy(pcbinfo->ipi_hashbase, M_PCB, pcbinfo->ipi_hashmask);
257 hashdestroy(pcbinfo->ipi_porthashbase, M_PCB,
258 pcbinfo->ipi_porthashmask);
260 in_pcbgroup_destroy(pcbinfo);
262 uma_zdestroy(pcbinfo->ipi_zone);
263 INP_HASH_LOCK_DESTROY(pcbinfo);
264 INP_INFO_LOCK_DESTROY(pcbinfo);
268 * Allocate a PCB and associate it with the socket.
269 * On success return with the PCB locked.
272 in_pcballoc(struct socket *so, struct inpcbinfo *pcbinfo)
277 INP_INFO_WLOCK_ASSERT(pcbinfo);
279 inp = uma_zalloc(pcbinfo->ipi_zone, M_NOWAIT);
282 bzero(inp, inp_zero_size);
283 inp->inp_pcbinfo = pcbinfo;
284 inp->inp_socket = so;
285 inp->inp_cred = crhold(so->so_cred);
286 inp->inp_inc.inc_fibnum = so->so_fibnum;
288 error = mac_inpcb_init(inp, M_NOWAIT);
291 mac_inpcb_create(so, inp);
294 error = ipsec_init_policy(so, &inp->inp_sp);
297 mac_inpcb_destroy(inp);
303 if (INP_SOCKAF(so) == AF_INET6) {
304 inp->inp_vflag |= INP_IPV6PROTO;
306 inp->inp_flags |= IN6P_IPV6_V6ONLY;
309 LIST_INSERT_HEAD(pcbinfo->ipi_listhead, inp, inp_list);
310 pcbinfo->ipi_count++;
311 so->so_pcb = (caddr_t)inp;
313 if (V_ip6_auto_flowlabel)
314 inp->inp_flags |= IN6P_AUTOFLOWLABEL;
317 inp->inp_gencnt = ++pcbinfo->ipi_gencnt;
318 refcount_init(&inp->inp_refcount, 1); /* Reference from inpcbinfo */
319 #if defined(IPSEC) || defined(MAC)
322 crfree(inp->inp_cred);
323 uma_zfree(pcbinfo->ipi_zone, inp);
331 in_pcbbind(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred)
335 INP_WLOCK_ASSERT(inp);
336 INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo);
338 if (inp->inp_lport != 0 || inp->inp_laddr.s_addr != INADDR_ANY)
340 anonport = nam == NULL || ((struct sockaddr_in *)nam)->sin_port == 0;
341 error = in_pcbbind_setup(inp, nam, &inp->inp_laddr.s_addr,
342 &inp->inp_lport, cred);
345 if (in_pcbinshash(inp) != 0) {
346 inp->inp_laddr.s_addr = INADDR_ANY;
351 inp->inp_flags |= INP_ANONPORT;
357 * Select a local port (number) to use.
359 #if defined(INET) || defined(INET6)
361 in_pcb_lport(struct inpcb *inp, struct in_addr *laddrp, u_short *lportp,
362 struct ucred *cred, int lookupflags)
364 struct inpcbinfo *pcbinfo;
365 struct inpcb *tmpinp;
366 unsigned short *lastport;
367 int count, dorandom, error;
368 u_short aux, first, last, lport;
370 struct in_addr laddr;
373 pcbinfo = inp->inp_pcbinfo;
376 * Because no actual state changes occur here, a global write lock on
377 * the pcbinfo isn't required.
379 INP_LOCK_ASSERT(inp);
380 INP_HASH_LOCK_ASSERT(pcbinfo);
382 if (inp->inp_flags & INP_HIGHPORT) {
383 first = V_ipport_hifirstauto; /* sysctl */
384 last = V_ipport_hilastauto;
385 lastport = &pcbinfo->ipi_lasthi;
386 } else if (inp->inp_flags & INP_LOWPORT) {
387 error = priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT, 0);
390 first = V_ipport_lowfirstauto; /* 1023 */
391 last = V_ipport_lowlastauto; /* 600 */
392 lastport = &pcbinfo->ipi_lastlow;
394 first = V_ipport_firstauto; /* sysctl */
395 last = V_ipport_lastauto;
396 lastport = &pcbinfo->ipi_lastport;
399 * For UDP(-Lite), use random port allocation as long as the user
400 * allows it. For TCP (and as of yet unknown) connections,
401 * use random port allocation only if the user allows it AND
402 * ipport_tick() allows it.
404 if (V_ipport_randomized &&
405 (!V_ipport_stoprandom || pcbinfo == &V_udbinfo ||
406 pcbinfo == &V_ulitecbinfo))
411 * It makes no sense to do random port allocation if
412 * we have the only port available.
416 /* Make sure to not include UDP(-Lite) packets in the count. */
417 if (pcbinfo != &V_udbinfo || pcbinfo != &V_ulitecbinfo)
418 V_ipport_tcpallocs++;
420 * Instead of having two loops further down counting up or down
421 * make sure that first is always <= last and go with only one
422 * code path implementing all logic.
431 /* Make the compiler happy. */
433 if ((inp->inp_vflag & (INP_IPV4|INP_IPV6)) == INP_IPV4) {
434 KASSERT(laddrp != NULL, ("%s: laddrp NULL for v4 inp %p",
439 tmpinp = NULL; /* Make compiler happy. */
443 *lastport = first + (arc4random() % (last - first));
445 count = last - first;
448 if (count-- < 0) /* completely used? */
449 return (EADDRNOTAVAIL);
451 if (*lastport < first || *lastport > last)
453 lport = htons(*lastport);
456 if ((inp->inp_vflag & INP_IPV6) != 0)
457 tmpinp = in6_pcblookup_local(pcbinfo,
458 &inp->in6p_laddr, lport, lookupflags, cred);
460 #if defined(INET) && defined(INET6)
464 tmpinp = in_pcblookup_local(pcbinfo, laddr,
465 lport, lookupflags, cred);
467 } while (tmpinp != NULL);
470 if ((inp->inp_vflag & (INP_IPV4|INP_IPV6)) == INP_IPV4)
471 laddrp->s_addr = laddr.s_addr;
479 * Return cached socket options.
482 inp_so_options(const struct inpcb *inp)
488 if ((inp->inp_flags2 & INP_REUSEPORT) != 0)
489 so_options |= SO_REUSEPORT;
490 if ((inp->inp_flags2 & INP_REUSEADDR) != 0)
491 so_options |= SO_REUSEADDR;
494 #endif /* INET || INET6 */
497 * Check if a new BINDMULTI socket is allowed to be created.
499 * ni points to the new inp.
500 * oi points to the exisitng inp.
502 * This checks whether the existing inp also has BINDMULTI and
503 * whether the credentials match.
506 in_pcbbind_check_bindmulti(const struct inpcb *ni, const struct inpcb *oi)
508 /* Check permissions match */
509 if ((ni->inp_flags2 & INP_BINDMULTI) &&
510 (ni->inp_cred->cr_uid !=
511 oi->inp_cred->cr_uid))
514 /* Check the existing inp has BINDMULTI set */
515 if ((ni->inp_flags2 & INP_BINDMULTI) &&
516 ((oi->inp_flags2 & INP_BINDMULTI) == 0))
520 * We're okay - either INP_BINDMULTI isn't set on ni, or
521 * it is and it matches the checks.
528 * Set up a bind operation on a PCB, performing port allocation
529 * as required, but do not actually modify the PCB. Callers can
530 * either complete the bind by setting inp_laddr/inp_lport and
531 * calling in_pcbinshash(), or they can just use the resulting
532 * port and address to authorise the sending of a once-off packet.
534 * On error, the values of *laddrp and *lportp are not changed.
537 in_pcbbind_setup(struct inpcb *inp, struct sockaddr *nam, in_addr_t *laddrp,
538 u_short *lportp, struct ucred *cred)
540 struct socket *so = inp->inp_socket;
541 struct sockaddr_in *sin;
542 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
543 struct in_addr laddr;
545 int lookupflags = 0, reuseport = (so->so_options & SO_REUSEPORT);
549 * No state changes, so read locks are sufficient here.
551 INP_LOCK_ASSERT(inp);
552 INP_HASH_LOCK_ASSERT(pcbinfo);
554 if (TAILQ_EMPTY(&V_in_ifaddrhead)) /* XXX broken! */
555 return (EADDRNOTAVAIL);
556 laddr.s_addr = *laddrp;
557 if (nam != NULL && laddr.s_addr != INADDR_ANY)
559 if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0)
560 lookupflags = INPLOOKUP_WILDCARD;
562 if ((error = prison_local_ip4(cred, &laddr)) != 0)
565 sin = (struct sockaddr_in *)nam;
566 if (nam->sa_len != sizeof (*sin))
570 * We should check the family, but old programs
571 * incorrectly fail to initialize it.
573 if (sin->sin_family != AF_INET)
574 return (EAFNOSUPPORT);
576 error = prison_local_ip4(cred, &sin->sin_addr);
579 if (sin->sin_port != *lportp) {
580 /* Don't allow the port to change. */
583 lport = sin->sin_port;
585 /* NB: lport is left as 0 if the port isn't being changed. */
586 if (IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) {
588 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
589 * allow complete duplication of binding if
590 * SO_REUSEPORT is set, or if SO_REUSEADDR is set
591 * and a multicast address is bound on both
592 * new and duplicated sockets.
594 if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) != 0)
595 reuseport = SO_REUSEADDR|SO_REUSEPORT;
596 } else if (sin->sin_addr.s_addr != INADDR_ANY) {
597 sin->sin_port = 0; /* yech... */
598 bzero(&sin->sin_zero, sizeof(sin->sin_zero));
600 * Is the address a local IP address?
601 * If INP_BINDANY is set, then the socket may be bound
602 * to any endpoint address, local or not.
604 if ((inp->inp_flags & INP_BINDANY) == 0 &&
605 ifa_ifwithaddr_check((struct sockaddr *)sin) == 0)
606 return (EADDRNOTAVAIL);
608 laddr = sin->sin_addr;
614 if (ntohs(lport) <= V_ipport_reservedhigh &&
615 ntohs(lport) >= V_ipport_reservedlow &&
616 priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT,
619 if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)) &&
620 priv_check_cred(inp->inp_cred,
621 PRIV_NETINET_REUSEPORT, 0) != 0) {
622 t = in_pcblookup_local(pcbinfo, sin->sin_addr,
623 lport, INPLOOKUP_WILDCARD, cred);
626 * This entire block sorely needs a rewrite.
629 ((inp->inp_flags2 & INP_BINDMULTI) == 0) &&
630 ((t->inp_flags & INP_TIMEWAIT) == 0) &&
631 (so->so_type != SOCK_STREAM ||
632 ntohl(t->inp_faddr.s_addr) == INADDR_ANY) &&
633 (ntohl(sin->sin_addr.s_addr) != INADDR_ANY ||
634 ntohl(t->inp_laddr.s_addr) != INADDR_ANY ||
635 (t->inp_flags2 & INP_REUSEPORT) == 0) &&
636 (inp->inp_cred->cr_uid !=
637 t->inp_cred->cr_uid))
641 * If the socket is a BINDMULTI socket, then
642 * the credentials need to match and the
643 * original socket also has to have been bound
646 if (t && (! in_pcbbind_check_bindmulti(inp, t)))
649 t = in_pcblookup_local(pcbinfo, sin->sin_addr,
650 lport, lookupflags, cred);
651 if (t && (t->inp_flags & INP_TIMEWAIT)) {
653 * XXXRW: If an incpb has had its timewait
654 * state recycled, we treat the address as
655 * being in use (for now). This is better
656 * than a panic, but not desirable.
660 (reuseport & tw->tw_so_options) == 0)
663 ((inp->inp_flags2 & INP_BINDMULTI) == 0) &&
664 (reuseport & inp_so_options(t)) == 0) {
666 if (ntohl(sin->sin_addr.s_addr) !=
668 ntohl(t->inp_laddr.s_addr) !=
670 (inp->inp_vflag & INP_IPV6PROTO) == 0 ||
671 (t->inp_vflag & INP_IPV6PROTO) == 0)
674 if (t && (! in_pcbbind_check_bindmulti(inp, t)))
682 error = in_pcb_lport(inp, &laddr, &lport, cred, lookupflags);
687 *laddrp = laddr.s_addr;
693 * Connect from a socket to a specified address.
694 * Both address and port must be specified in argument sin.
695 * If don't have a local address for this socket yet,
699 in_pcbconnect_mbuf(struct inpcb *inp, struct sockaddr *nam,
700 struct ucred *cred, struct mbuf *m)
702 u_short lport, fport;
703 in_addr_t laddr, faddr;
706 INP_WLOCK_ASSERT(inp);
707 INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo);
709 lport = inp->inp_lport;
710 laddr = inp->inp_laddr.s_addr;
711 anonport = (lport == 0);
712 error = in_pcbconnect_setup(inp, nam, &laddr, &lport, &faddr, &fport,
717 /* Do the initial binding of the local address if required. */
718 if (inp->inp_laddr.s_addr == INADDR_ANY && inp->inp_lport == 0) {
719 inp->inp_lport = lport;
720 inp->inp_laddr.s_addr = laddr;
721 if (in_pcbinshash(inp) != 0) {
722 inp->inp_laddr.s_addr = INADDR_ANY;
728 /* Commit the remaining changes. */
729 inp->inp_lport = lport;
730 inp->inp_laddr.s_addr = laddr;
731 inp->inp_faddr.s_addr = faddr;
732 inp->inp_fport = fport;
733 in_pcbrehash_mbuf(inp, m);
736 inp->inp_flags |= INP_ANONPORT;
741 in_pcbconnect(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred)
744 return (in_pcbconnect_mbuf(inp, nam, cred, NULL));
748 * Do proper source address selection on an unbound socket in case
749 * of connect. Take jails into account as well.
752 in_pcbladdr(struct inpcb *inp, struct in_addr *faddr, struct in_addr *laddr,
757 struct sockaddr_in *sin;
761 KASSERT(laddr != NULL, ("%s: laddr NULL", __func__));
764 * Bypass source address selection and use the primary jail IP
767 if (cred != NULL && !prison_saddrsel_ip4(cred, laddr))
771 bzero(&sro, sizeof(sro));
773 sin = (struct sockaddr_in *)&sro.ro_dst;
774 sin->sin_family = AF_INET;
775 sin->sin_len = sizeof(struct sockaddr_in);
776 sin->sin_addr.s_addr = faddr->s_addr;
779 * If route is known our src addr is taken from the i/f,
782 * Find out route to destination.
784 if ((inp->inp_socket->so_options & SO_DONTROUTE) == 0)
785 in_rtalloc_ign(&sro, 0, inp->inp_inc.inc_fibnum);
788 * If we found a route, use the address corresponding to
789 * the outgoing interface.
791 * Otherwise assume faddr is reachable on a directly connected
792 * network and try to find a corresponding interface to take
793 * the source address from.
795 if (sro.ro_rt == NULL || sro.ro_rt->rt_ifp == NULL) {
796 struct in_ifaddr *ia;
799 ia = ifatoia(ifa_ifwithdstaddr((struct sockaddr *)sin,
800 inp->inp_socket->so_fibnum));
802 ia = ifatoia(ifa_ifwithnet((struct sockaddr *)sin, 0,
803 inp->inp_socket->so_fibnum));
809 if (cred == NULL || !prison_flag(cred, PR_IP4)) {
810 laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
811 ifa_free(&ia->ia_ifa);
816 ifa_free(&ia->ia_ifa);
819 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
822 if (sa->sa_family != AF_INET)
824 sin = (struct sockaddr_in *)sa;
825 if (prison_check_ip4(cred, &sin->sin_addr) == 0) {
826 ia = (struct in_ifaddr *)ifa;
831 laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
832 IF_ADDR_RUNLOCK(ifp);
835 IF_ADDR_RUNLOCK(ifp);
837 /* 3. As a last resort return the 'default' jail address. */
838 error = prison_get_ip4(cred, laddr);
843 * If the outgoing interface on the route found is not
844 * a loopback interface, use the address from that interface.
845 * In case of jails do those three steps:
846 * 1. check if the interface address belongs to the jail. If so use it.
847 * 2. check if we have any address on the outgoing interface
848 * belonging to this jail. If so use it.
849 * 3. as a last resort return the 'default' jail address.
851 if ((sro.ro_rt->rt_ifp->if_flags & IFF_LOOPBACK) == 0) {
852 struct in_ifaddr *ia;
855 /* If not jailed, use the default returned. */
856 if (cred == NULL || !prison_flag(cred, PR_IP4)) {
857 ia = (struct in_ifaddr *)sro.ro_rt->rt_ifa;
858 laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
863 /* 1. Check if the iface address belongs to the jail. */
864 sin = (struct sockaddr_in *)sro.ro_rt->rt_ifa->ifa_addr;
865 if (prison_check_ip4(cred, &sin->sin_addr) == 0) {
866 ia = (struct in_ifaddr *)sro.ro_rt->rt_ifa;
867 laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
872 * 2. Check if we have any address on the outgoing interface
873 * belonging to this jail.
876 ifp = sro.ro_rt->rt_ifp;
878 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
880 if (sa->sa_family != AF_INET)
882 sin = (struct sockaddr_in *)sa;
883 if (prison_check_ip4(cred, &sin->sin_addr) == 0) {
884 ia = (struct in_ifaddr *)ifa;
889 laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
890 IF_ADDR_RUNLOCK(ifp);
893 IF_ADDR_RUNLOCK(ifp);
895 /* 3. As a last resort return the 'default' jail address. */
896 error = prison_get_ip4(cred, laddr);
901 * The outgoing interface is marked with 'loopback net', so a route
902 * to ourselves is here.
903 * Try to find the interface of the destination address and then
904 * take the address from there. That interface is not necessarily
905 * a loopback interface.
906 * In case of jails, check that it is an address of the jail
907 * and if we cannot find, fall back to the 'default' jail address.
909 if ((sro.ro_rt->rt_ifp->if_flags & IFF_LOOPBACK) != 0) {
910 struct sockaddr_in sain;
911 struct in_ifaddr *ia;
913 bzero(&sain, sizeof(struct sockaddr_in));
914 sain.sin_family = AF_INET;
915 sain.sin_len = sizeof(struct sockaddr_in);
916 sain.sin_addr.s_addr = faddr->s_addr;
918 ia = ifatoia(ifa_ifwithdstaddr(sintosa(&sain),
919 inp->inp_socket->so_fibnum));
921 ia = ifatoia(ifa_ifwithnet(sintosa(&sain), 0,
922 inp->inp_socket->so_fibnum));
924 ia = ifatoia(ifa_ifwithaddr(sintosa(&sain)));
926 if (cred == NULL || !prison_flag(cred, PR_IP4)) {
931 laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
932 ifa_free(&ia->ia_ifa);
941 ifa_free(&ia->ia_ifa);
944 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
947 if (sa->sa_family != AF_INET)
949 sin = (struct sockaddr_in *)sa;
950 if (prison_check_ip4(cred,
951 &sin->sin_addr) == 0) {
952 ia = (struct in_ifaddr *)ifa;
957 laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
958 IF_ADDR_RUNLOCK(ifp);
961 IF_ADDR_RUNLOCK(ifp);
964 /* 3. As a last resort return the 'default' jail address. */
965 error = prison_get_ip4(cred, laddr);
970 if (sro.ro_rt != NULL)
976 * Set up for a connect from a socket to the specified address.
977 * On entry, *laddrp and *lportp should contain the current local
978 * address and port for the PCB; these are updated to the values
979 * that should be placed in inp_laddr and inp_lport to complete
982 * On success, *faddrp and *fportp will be set to the remote address
983 * and port. These are not updated in the error case.
985 * If the operation fails because the connection already exists,
986 * *oinpp will be set to the PCB of that connection so that the
987 * caller can decide to override it. In all other cases, *oinpp
991 in_pcbconnect_setup(struct inpcb *inp, struct sockaddr *nam,
992 in_addr_t *laddrp, u_short *lportp, in_addr_t *faddrp, u_short *fportp,
993 struct inpcb **oinpp, struct ucred *cred)
995 struct sockaddr_in *sin = (struct sockaddr_in *)nam;
996 struct in_ifaddr *ia;
998 struct in_addr laddr, faddr;
999 u_short lport, fport;
1003 * Because a global state change doesn't actually occur here, a read
1004 * lock is sufficient.
1006 INP_LOCK_ASSERT(inp);
1007 INP_HASH_LOCK_ASSERT(inp->inp_pcbinfo);
1011 if (nam->sa_len != sizeof (*sin))
1013 if (sin->sin_family != AF_INET)
1014 return (EAFNOSUPPORT);
1015 if (sin->sin_port == 0)
1016 return (EADDRNOTAVAIL);
1017 laddr.s_addr = *laddrp;
1019 faddr = sin->sin_addr;
1020 fport = sin->sin_port;
1022 if (!TAILQ_EMPTY(&V_in_ifaddrhead)) {
1024 * If the destination address is INADDR_ANY,
1025 * use the primary local address.
1026 * If the supplied address is INADDR_BROADCAST,
1027 * and the primary interface supports broadcast,
1028 * choose the broadcast address for that interface.
1030 if (faddr.s_addr == INADDR_ANY) {
1033 IA_SIN(TAILQ_FIRST(&V_in_ifaddrhead))->sin_addr;
1034 IN_IFADDR_RUNLOCK();
1036 (error = prison_get_ip4(cred, &faddr)) != 0)
1038 } else if (faddr.s_addr == (u_long)INADDR_BROADCAST) {
1040 if (TAILQ_FIRST(&V_in_ifaddrhead)->ia_ifp->if_flags &
1042 faddr = satosin(&TAILQ_FIRST(
1043 &V_in_ifaddrhead)->ia_broadaddr)->sin_addr;
1044 IN_IFADDR_RUNLOCK();
1047 if (laddr.s_addr == INADDR_ANY) {
1048 error = in_pcbladdr(inp, &faddr, &laddr, cred);
1050 * If the destination address is multicast and an outgoing
1051 * interface has been set as a multicast option, prefer the
1052 * address of that interface as our source address.
1054 if (IN_MULTICAST(ntohl(faddr.s_addr)) &&
1055 inp->inp_moptions != NULL) {
1056 struct ip_moptions *imo;
1059 imo = inp->inp_moptions;
1060 if (imo->imo_multicast_ifp != NULL) {
1061 ifp = imo->imo_multicast_ifp;
1063 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
1064 if ((ia->ia_ifp == ifp) &&
1066 prison_check_ip4(cred,
1067 &ia->ia_addr.sin_addr) == 0))
1071 error = EADDRNOTAVAIL;
1073 laddr = ia->ia_addr.sin_addr;
1076 IN_IFADDR_RUNLOCK();
1082 oinp = in_pcblookup_hash_locked(inp->inp_pcbinfo, faddr, fport,
1083 laddr, lport, 0, NULL);
1087 return (EADDRINUSE);
1090 error = in_pcbbind_setup(inp, NULL, &laddr.s_addr, &lport,
1095 *laddrp = laddr.s_addr;
1097 *faddrp = faddr.s_addr;
1103 in_pcbdisconnect(struct inpcb *inp)
1106 INP_WLOCK_ASSERT(inp);
1107 INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo);
1109 inp->inp_faddr.s_addr = INADDR_ANY;
1116 * in_pcbdetach() is responsibe for disassociating a socket from an inpcb.
1117 * For most protocols, this will be invoked immediately prior to calling
1118 * in_pcbfree(). However, with TCP the inpcb may significantly outlive the
1119 * socket, in which case in_pcbfree() is deferred.
1122 in_pcbdetach(struct inpcb *inp)
1125 KASSERT(inp->inp_socket != NULL, ("%s: inp_socket == NULL", __func__));
1127 inp->inp_socket->so_pcb = NULL;
1128 inp->inp_socket = NULL;
1132 * in_pcbref() bumps the reference count on an inpcb in order to maintain
1133 * stability of an inpcb pointer despite the inpcb lock being released. This
1134 * is used in TCP when the inpcbinfo lock needs to be acquired or upgraded,
1135 * but where the inpcb lock may already held, or when acquiring a reference
1138 * in_pcbref() should be used only to provide brief memory stability, and
1139 * must always be followed by a call to INP_WLOCK() and in_pcbrele() to
1140 * garbage collect the inpcb if it has been in_pcbfree()'d from another
1141 * context. Until in_pcbrele() has returned that the inpcb is still valid,
1142 * lock and rele are the *only* safe operations that may be performed on the
1145 * While the inpcb will not be freed, releasing the inpcb lock means that the
1146 * connection's state may change, so the caller should be careful to
1147 * revalidate any cached state on reacquiring the lock. Drop the reference
1148 * using in_pcbrele().
1151 in_pcbref(struct inpcb *inp)
1154 KASSERT(inp->inp_refcount > 0, ("%s: refcount 0", __func__));
1156 refcount_acquire(&inp->inp_refcount);
1160 * Drop a refcount on an inpcb elevated using in_pcbref(); because a call to
1161 * in_pcbfree() may have been made between in_pcbref() and in_pcbrele(), we
1162 * return a flag indicating whether or not the inpcb remains valid. If it is
1163 * valid, we return with the inpcb lock held.
1165 * Notice that, unlike in_pcbref(), the inpcb lock must be held to drop a
1166 * reference on an inpcb. Historically more work was done here (actually, in
1167 * in_pcbfree_internal()) but has been moved to in_pcbfree() to avoid the
1168 * need for the pcbinfo lock in in_pcbrele(). Deferring the free is entirely
1169 * about memory stability (and continued use of the write lock).
1172 in_pcbrele_rlocked(struct inpcb *inp)
1174 struct inpcbinfo *pcbinfo;
1176 KASSERT(inp->inp_refcount > 0, ("%s: refcount 0", __func__));
1178 INP_RLOCK_ASSERT(inp);
1180 if (refcount_release(&inp->inp_refcount) == 0) {
1182 * If the inpcb has been freed, let the caller know, even if
1183 * this isn't the last reference.
1185 if (inp->inp_flags2 & INP_FREED) {
1192 KASSERT(inp->inp_socket == NULL, ("%s: inp_socket != NULL", __func__));
1195 pcbinfo = inp->inp_pcbinfo;
1196 uma_zfree(pcbinfo->ipi_zone, inp);
1201 in_pcbrele_wlocked(struct inpcb *inp)
1203 struct inpcbinfo *pcbinfo;
1205 KASSERT(inp->inp_refcount > 0, ("%s: refcount 0", __func__));
1207 INP_WLOCK_ASSERT(inp);
1209 if (refcount_release(&inp->inp_refcount) == 0)
1212 KASSERT(inp->inp_socket == NULL, ("%s: inp_socket != NULL", __func__));
1215 pcbinfo = inp->inp_pcbinfo;
1216 uma_zfree(pcbinfo->ipi_zone, inp);
1221 * Temporary wrapper.
1224 in_pcbrele(struct inpcb *inp)
1227 return (in_pcbrele_wlocked(inp));
1231 * Unconditionally schedule an inpcb to be freed by decrementing its
1232 * reference count, which should occur only after the inpcb has been detached
1233 * from its socket. If another thread holds a temporary reference (acquired
1234 * using in_pcbref()) then the free is deferred until that reference is
1235 * released using in_pcbrele(), but the inpcb is still unlocked. Almost all
1236 * work, including removal from global lists, is done in this context, where
1237 * the pcbinfo lock is held.
1240 in_pcbfree(struct inpcb *inp)
1242 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1244 KASSERT(inp->inp_socket == NULL, ("%s: inp_socket != NULL", __func__));
1246 INP_INFO_WLOCK_ASSERT(pcbinfo);
1247 INP_WLOCK_ASSERT(inp);
1249 /* XXXRW: Do as much as possible here. */
1251 if (inp->inp_sp != NULL)
1252 ipsec_delete_pcbpolicy(inp);
1254 inp->inp_gencnt = ++pcbinfo->ipi_gencnt;
1255 in_pcbremlists(inp);
1257 if (inp->inp_vflag & INP_IPV6PROTO) {
1258 ip6_freepcbopts(inp->in6p_outputopts);
1259 if (inp->in6p_moptions != NULL)
1260 ip6_freemoptions(inp->in6p_moptions);
1263 if (inp->inp_options)
1264 (void)m_free(inp->inp_options);
1266 if (inp->inp_moptions != NULL)
1267 inp_freemoptions(inp->inp_moptions);
1270 inp->inp_flags2 |= INP_FREED;
1271 crfree(inp->inp_cred);
1273 mac_inpcb_destroy(inp);
1275 if (!in_pcbrele_wlocked(inp))
1280 * in_pcbdrop() removes an inpcb from hashed lists, releasing its address and
1281 * port reservation, and preventing it from being returned by inpcb lookups.
1283 * It is used by TCP to mark an inpcb as unused and avoid future packet
1284 * delivery or event notification when a socket remains open but TCP has
1285 * closed. This might occur as a result of a shutdown()-initiated TCP close
1286 * or a RST on the wire, and allows the port binding to be reused while still
1287 * maintaining the invariant that so_pcb always points to a valid inpcb until
1290 * XXXRW: Possibly in_pcbdrop() should also prevent future notifications by
1291 * in_pcbnotifyall() and in_pcbpurgeif0()?
1294 in_pcbdrop(struct inpcb *inp)
1297 INP_WLOCK_ASSERT(inp);
1300 * XXXRW: Possibly we should protect the setting of INP_DROPPED with
1303 inp->inp_flags |= INP_DROPPED;
1304 if (inp->inp_flags & INP_INHASHLIST) {
1305 struct inpcbport *phd = inp->inp_phd;
1307 INP_HASH_WLOCK(inp->inp_pcbinfo);
1308 LIST_REMOVE(inp, inp_hash);
1309 LIST_REMOVE(inp, inp_portlist);
1310 if (LIST_FIRST(&phd->phd_pcblist) == NULL) {
1311 LIST_REMOVE(phd, phd_hash);
1314 INP_HASH_WUNLOCK(inp->inp_pcbinfo);
1315 inp->inp_flags &= ~INP_INHASHLIST;
1317 in_pcbgroup_remove(inp);
1324 * Common routines to return the socket addresses associated with inpcbs.
1327 in_sockaddr(in_port_t port, struct in_addr *addr_p)
1329 struct sockaddr_in *sin;
1331 sin = malloc(sizeof *sin, M_SONAME,
1333 sin->sin_family = AF_INET;
1334 sin->sin_len = sizeof(*sin);
1335 sin->sin_addr = *addr_p;
1336 sin->sin_port = port;
1338 return (struct sockaddr *)sin;
1342 in_getsockaddr(struct socket *so, struct sockaddr **nam)
1345 struct in_addr addr;
1348 inp = sotoinpcb(so);
1349 KASSERT(inp != NULL, ("in_getsockaddr: inp == NULL"));
1352 port = inp->inp_lport;
1353 addr = inp->inp_laddr;
1356 *nam = in_sockaddr(port, &addr);
1361 in_getpeeraddr(struct socket *so, struct sockaddr **nam)
1364 struct in_addr addr;
1367 inp = sotoinpcb(so);
1368 KASSERT(inp != NULL, ("in_getpeeraddr: inp == NULL"));
1371 port = inp->inp_fport;
1372 addr = inp->inp_faddr;
1375 *nam = in_sockaddr(port, &addr);
1380 in_pcbnotifyall(struct inpcbinfo *pcbinfo, struct in_addr faddr, int errno,
1381 struct inpcb *(*notify)(struct inpcb *, int))
1383 struct inpcb *inp, *inp_temp;
1385 INP_INFO_WLOCK(pcbinfo);
1386 LIST_FOREACH_SAFE(inp, pcbinfo->ipi_listhead, inp_list, inp_temp) {
1389 if ((inp->inp_vflag & INP_IPV4) == 0) {
1394 if (inp->inp_faddr.s_addr != faddr.s_addr ||
1395 inp->inp_socket == NULL) {
1399 if ((*notify)(inp, errno))
1402 INP_INFO_WUNLOCK(pcbinfo);
1406 in_pcbpurgeif0(struct inpcbinfo *pcbinfo, struct ifnet *ifp)
1409 struct ip_moptions *imo;
1412 INP_INFO_RLOCK(pcbinfo);
1413 LIST_FOREACH(inp, pcbinfo->ipi_listhead, inp_list) {
1415 imo = inp->inp_moptions;
1416 if ((inp->inp_vflag & INP_IPV4) &&
1419 * Unselect the outgoing interface if it is being
1422 if (imo->imo_multicast_ifp == ifp)
1423 imo->imo_multicast_ifp = NULL;
1426 * Drop multicast group membership if we joined
1427 * through the interface being detached.
1429 for (i = 0, gap = 0; i < imo->imo_num_memberships;
1431 if (imo->imo_membership[i]->inm_ifp == ifp) {
1432 in_delmulti(imo->imo_membership[i]);
1434 } else if (gap != 0)
1435 imo->imo_membership[i - gap] =
1436 imo->imo_membership[i];
1438 imo->imo_num_memberships -= gap;
1442 INP_INFO_RUNLOCK(pcbinfo);
1446 * Lookup a PCB based on the local address and port. Caller must hold the
1447 * hash lock. No inpcb locks or references are acquired.
1449 #define INP_LOOKUP_MAPPED_PCB_COST 3
1451 in_pcblookup_local(struct inpcbinfo *pcbinfo, struct in_addr laddr,
1452 u_short lport, int lookupflags, struct ucred *cred)
1456 int matchwild = 3 + INP_LOOKUP_MAPPED_PCB_COST;
1462 KASSERT((lookupflags & ~(INPLOOKUP_WILDCARD)) == 0,
1463 ("%s: invalid lookup flags %d", __func__, lookupflags));
1465 INP_HASH_LOCK_ASSERT(pcbinfo);
1467 if ((lookupflags & INPLOOKUP_WILDCARD) == 0) {
1468 struct inpcbhead *head;
1470 * Look for an unconnected (wildcard foreign addr) PCB that
1471 * matches the local address and port we're looking for.
1473 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport,
1474 0, pcbinfo->ipi_hashmask)];
1475 LIST_FOREACH(inp, head, inp_hash) {
1477 /* XXX inp locking */
1478 if ((inp->inp_vflag & INP_IPV4) == 0)
1481 if (inp->inp_faddr.s_addr == INADDR_ANY &&
1482 inp->inp_laddr.s_addr == laddr.s_addr &&
1483 inp->inp_lport == lport) {
1488 prison_equal_ip4(cred->cr_prison,
1489 inp->inp_cred->cr_prison))
1498 struct inpcbporthead *porthash;
1499 struct inpcbport *phd;
1500 struct inpcb *match = NULL;
1502 * Best fit PCB lookup.
1504 * First see if this local port is in use by looking on the
1507 porthash = &pcbinfo->ipi_porthashbase[INP_PCBPORTHASH(lport,
1508 pcbinfo->ipi_porthashmask)];
1509 LIST_FOREACH(phd, porthash, phd_hash) {
1510 if (phd->phd_port == lport)
1515 * Port is in use by one or more PCBs. Look for best
1518 LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) {
1521 !prison_equal_ip4(inp->inp_cred->cr_prison,
1525 /* XXX inp locking */
1526 if ((inp->inp_vflag & INP_IPV4) == 0)
1529 * We never select the PCB that has
1530 * INP_IPV6 flag and is bound to :: if
1531 * we have another PCB which is bound
1532 * to 0.0.0.0. If a PCB has the
1533 * INP_IPV6 flag, then we set its cost
1534 * higher than IPv4 only PCBs.
1536 * Note that the case only happens
1537 * when a socket is bound to ::, under
1538 * the condition that the use of the
1539 * mapped address is allowed.
1541 if ((inp->inp_vflag & INP_IPV6) != 0)
1542 wildcard += INP_LOOKUP_MAPPED_PCB_COST;
1544 if (inp->inp_faddr.s_addr != INADDR_ANY)
1546 if (inp->inp_laddr.s_addr != INADDR_ANY) {
1547 if (laddr.s_addr == INADDR_ANY)
1549 else if (inp->inp_laddr.s_addr != laddr.s_addr)
1552 if (laddr.s_addr != INADDR_ANY)
1555 if (wildcard < matchwild) {
1557 matchwild = wildcard;
1566 #undef INP_LOOKUP_MAPPED_PCB_COST
1570 * Lookup PCB in hash list, using pcbgroup tables.
1572 static struct inpcb *
1573 in_pcblookup_group(struct inpcbinfo *pcbinfo, struct inpcbgroup *pcbgroup,
1574 struct in_addr faddr, u_int fport_arg, struct in_addr laddr,
1575 u_int lport_arg, int lookupflags, struct ifnet *ifp)
1577 struct inpcbhead *head;
1578 struct inpcb *inp, *tmpinp;
1579 u_short fport = fport_arg, lport = lport_arg;
1582 * First look for an exact match.
1585 INP_GROUP_LOCK(pcbgroup);
1586 head = &pcbgroup->ipg_hashbase[INP_PCBHASH(faddr.s_addr, lport, fport,
1587 pcbgroup->ipg_hashmask)];
1588 LIST_FOREACH(inp, head, inp_pcbgrouphash) {
1590 /* XXX inp locking */
1591 if ((inp->inp_vflag & INP_IPV4) == 0)
1594 if (inp->inp_faddr.s_addr == faddr.s_addr &&
1595 inp->inp_laddr.s_addr == laddr.s_addr &&
1596 inp->inp_fport == fport &&
1597 inp->inp_lport == lport) {
1599 * XXX We should be able to directly return
1600 * the inp here, without any checks.
1601 * Well unless both bound with SO_REUSEPORT?
1603 if (prison_flag(inp->inp_cred, PR_IP4))
1609 if (tmpinp != NULL) {
1616 * For incoming connections, we may wish to do a wildcard
1617 * match for an RSS-local socket.
1619 if ((lookupflags & INPLOOKUP_WILDCARD) != 0) {
1620 struct inpcb *local_wild = NULL, *local_exact = NULL;
1622 struct inpcb *local_wild_mapped = NULL;
1624 struct inpcb *jail_wild = NULL;
1625 struct inpcbhead *head;
1629 * Order of socket selection - we always prefer jails.
1630 * 1. jailed, non-wild.
1632 * 3. non-jailed, non-wild.
1633 * 4. non-jailed, wild.
1636 head = &pcbgroup->ipg_hashbase[INP_PCBHASH(INADDR_ANY,
1637 lport, 0, pcbgroup->ipg_hashmask)];
1638 LIST_FOREACH(inp, head, inp_pcbgrouphash) {
1640 /* XXX inp locking */
1641 if ((inp->inp_vflag & INP_IPV4) == 0)
1644 if (inp->inp_faddr.s_addr != INADDR_ANY ||
1645 inp->inp_lport != lport)
1648 injail = prison_flag(inp->inp_cred, PR_IP4);
1650 if (prison_check_ip4(inp->inp_cred,
1654 if (local_exact != NULL)
1658 if (inp->inp_laddr.s_addr == laddr.s_addr) {
1663 } else if (inp->inp_laddr.s_addr == INADDR_ANY) {
1665 /* XXX inp locking, NULL check */
1666 if (inp->inp_vflag & INP_IPV6PROTO)
1667 local_wild_mapped = inp;
1675 } /* LIST_FOREACH */
1684 inp = local_wild_mapped;
1692 * Then look for a wildcard match, if requested.
1694 if ((lookupflags & INPLOOKUP_WILDCARD) != 0) {
1695 struct inpcb *local_wild = NULL, *local_exact = NULL;
1697 struct inpcb *local_wild_mapped = NULL;
1699 struct inpcb *jail_wild = NULL;
1700 struct inpcbhead *head;
1704 * Order of socket selection - we always prefer jails.
1705 * 1. jailed, non-wild.
1707 * 3. non-jailed, non-wild.
1708 * 4. non-jailed, wild.
1710 head = &pcbinfo->ipi_wildbase[INP_PCBHASH(INADDR_ANY, lport,
1711 0, pcbinfo->ipi_wildmask)];
1712 LIST_FOREACH(inp, head, inp_pcbgroup_wild) {
1714 /* XXX inp locking */
1715 if ((inp->inp_vflag & INP_IPV4) == 0)
1718 if (inp->inp_faddr.s_addr != INADDR_ANY ||
1719 inp->inp_lport != lport)
1722 injail = prison_flag(inp->inp_cred, PR_IP4);
1724 if (prison_check_ip4(inp->inp_cred,
1728 if (local_exact != NULL)
1732 if (inp->inp_laddr.s_addr == laddr.s_addr) {
1737 } else if (inp->inp_laddr.s_addr == INADDR_ANY) {
1739 /* XXX inp locking, NULL check */
1740 if (inp->inp_vflag & INP_IPV6PROTO)
1741 local_wild_mapped = inp;
1749 } /* LIST_FOREACH */
1757 inp = local_wild_mapped;
1761 } /* if (lookupflags & INPLOOKUP_WILDCARD) */
1762 INP_GROUP_UNLOCK(pcbgroup);
1767 INP_GROUP_UNLOCK(pcbgroup);
1768 if (lookupflags & INPLOOKUP_WLOCKPCB) {
1770 if (in_pcbrele_wlocked(inp))
1772 } else if (lookupflags & INPLOOKUP_RLOCKPCB) {
1774 if (in_pcbrele_rlocked(inp))
1777 panic("%s: locking bug", __func__);
1780 #endif /* PCBGROUP */
1783 * Lookup PCB in hash list, using pcbinfo tables. This variation assumes
1784 * that the caller has locked the hash list, and will not perform any further
1785 * locking or reference operations on either the hash list or the connection.
1787 static struct inpcb *
1788 in_pcblookup_hash_locked(struct inpcbinfo *pcbinfo, struct in_addr faddr,
1789 u_int fport_arg, struct in_addr laddr, u_int lport_arg, int lookupflags,
1792 struct inpcbhead *head;
1793 struct inpcb *inp, *tmpinp;
1794 u_short fport = fport_arg, lport = lport_arg;
1796 KASSERT((lookupflags & ~(INPLOOKUP_WILDCARD)) == 0,
1797 ("%s: invalid lookup flags %d", __func__, lookupflags));
1799 INP_HASH_LOCK_ASSERT(pcbinfo);
1802 * First look for an exact match.
1805 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(faddr.s_addr, lport, fport,
1806 pcbinfo->ipi_hashmask)];
1807 LIST_FOREACH(inp, head, inp_hash) {
1809 /* XXX inp locking */
1810 if ((inp->inp_vflag & INP_IPV4) == 0)
1813 if (inp->inp_faddr.s_addr == faddr.s_addr &&
1814 inp->inp_laddr.s_addr == laddr.s_addr &&
1815 inp->inp_fport == fport &&
1816 inp->inp_lport == lport) {
1818 * XXX We should be able to directly return
1819 * the inp here, without any checks.
1820 * Well unless both bound with SO_REUSEPORT?
1822 if (prison_flag(inp->inp_cred, PR_IP4))
1832 * Then look for a wildcard match, if requested.
1834 if ((lookupflags & INPLOOKUP_WILDCARD) != 0) {
1835 struct inpcb *local_wild = NULL, *local_exact = NULL;
1837 struct inpcb *local_wild_mapped = NULL;
1839 struct inpcb *jail_wild = NULL;
1843 * Order of socket selection - we always prefer jails.
1844 * 1. jailed, non-wild.
1846 * 3. non-jailed, non-wild.
1847 * 4. non-jailed, wild.
1850 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport,
1851 0, pcbinfo->ipi_hashmask)];
1852 LIST_FOREACH(inp, head, inp_hash) {
1854 /* XXX inp locking */
1855 if ((inp->inp_vflag & INP_IPV4) == 0)
1858 if (inp->inp_faddr.s_addr != INADDR_ANY ||
1859 inp->inp_lport != lport)
1862 injail = prison_flag(inp->inp_cred, PR_IP4);
1864 if (prison_check_ip4(inp->inp_cred,
1868 if (local_exact != NULL)
1872 if (inp->inp_laddr.s_addr == laddr.s_addr) {
1877 } else if (inp->inp_laddr.s_addr == INADDR_ANY) {
1879 /* XXX inp locking, NULL check */
1880 if (inp->inp_vflag & INP_IPV6PROTO)
1881 local_wild_mapped = inp;
1889 } /* LIST_FOREACH */
1890 if (jail_wild != NULL)
1892 if (local_exact != NULL)
1893 return (local_exact);
1894 if (local_wild != NULL)
1895 return (local_wild);
1897 if (local_wild_mapped != NULL)
1898 return (local_wild_mapped);
1900 } /* if ((lookupflags & INPLOOKUP_WILDCARD) != 0) */
1906 * Lookup PCB in hash list, using pcbinfo tables. This variation locks the
1907 * hash list lock, and will return the inpcb locked (i.e., requires
1908 * INPLOOKUP_LOCKPCB).
1910 static struct inpcb *
1911 in_pcblookup_hash(struct inpcbinfo *pcbinfo, struct in_addr faddr,
1912 u_int fport, struct in_addr laddr, u_int lport, int lookupflags,
1917 INP_HASH_RLOCK(pcbinfo);
1918 inp = in_pcblookup_hash_locked(pcbinfo, faddr, fport, laddr, lport,
1919 (lookupflags & ~(INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)), ifp);
1922 INP_HASH_RUNLOCK(pcbinfo);
1923 if (lookupflags & INPLOOKUP_WLOCKPCB) {
1925 if (in_pcbrele_wlocked(inp))
1927 } else if (lookupflags & INPLOOKUP_RLOCKPCB) {
1929 if (in_pcbrele_rlocked(inp))
1932 panic("%s: locking bug", __func__);
1934 INP_HASH_RUNLOCK(pcbinfo);
1939 * Public inpcb lookup routines, accepting a 4-tuple, and optionally, an mbuf
1940 * from which a pre-calculated hash value may be extracted.
1942 * Possibly more of this logic should be in in_pcbgroup.c.
1945 in_pcblookup(struct inpcbinfo *pcbinfo, struct in_addr faddr, u_int fport,
1946 struct in_addr laddr, u_int lport, int lookupflags, struct ifnet *ifp)
1948 #if defined(PCBGROUP) && !defined(RSS)
1949 struct inpcbgroup *pcbgroup;
1952 KASSERT((lookupflags & ~INPLOOKUP_MASK) == 0,
1953 ("%s: invalid lookup flags %d", __func__, lookupflags));
1954 KASSERT((lookupflags & (INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)) != 0,
1955 ("%s: LOCKPCB not set", __func__));
1958 * When not using RSS, use connection groups in preference to the
1959 * reservation table when looking up 4-tuples. When using RSS, just
1960 * use the reservation table, due to the cost of the Toeplitz hash
1963 * XXXRW: This policy belongs in the pcbgroup code, as in principle
1964 * we could be doing RSS with a non-Toeplitz hash that is affordable
1967 #if defined(PCBGROUP) && !defined(RSS)
1968 if (in_pcbgroup_enabled(pcbinfo)) {
1969 pcbgroup = in_pcbgroup_bytuple(pcbinfo, laddr, lport, faddr,
1971 return (in_pcblookup_group(pcbinfo, pcbgroup, faddr, fport,
1972 laddr, lport, lookupflags, ifp));
1975 return (in_pcblookup_hash(pcbinfo, faddr, fport, laddr, lport,
1980 in_pcblookup_mbuf(struct inpcbinfo *pcbinfo, struct in_addr faddr,
1981 u_int fport, struct in_addr laddr, u_int lport, int lookupflags,
1982 struct ifnet *ifp, struct mbuf *m)
1985 struct inpcbgroup *pcbgroup;
1988 KASSERT((lookupflags & ~INPLOOKUP_MASK) == 0,
1989 ("%s: invalid lookup flags %d", __func__, lookupflags));
1990 KASSERT((lookupflags & (INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)) != 0,
1991 ("%s: LOCKPCB not set", __func__));
1995 * If we can use a hardware-generated hash to look up the connection
1996 * group, use that connection group to find the inpcb. Otherwise
1997 * fall back on a software hash -- or the reservation table if we're
2000 * XXXRW: As above, that policy belongs in the pcbgroup code.
2002 if (in_pcbgroup_enabled(pcbinfo) &&
2003 !(M_HASHTYPE_TEST(m, M_HASHTYPE_NONE))) {
2004 pcbgroup = in_pcbgroup_byhash(pcbinfo, M_HASHTYPE_GET(m),
2005 m->m_pkthdr.flowid);
2006 if (pcbgroup != NULL)
2007 return (in_pcblookup_group(pcbinfo, pcbgroup, faddr,
2008 fport, laddr, lport, lookupflags, ifp));
2010 pcbgroup = in_pcbgroup_bytuple(pcbinfo, laddr, lport, faddr,
2012 return (in_pcblookup_group(pcbinfo, pcbgroup, faddr, fport,
2013 laddr, lport, lookupflags, ifp));
2017 return (in_pcblookup_hash(pcbinfo, faddr, fport, laddr, lport,
2023 * Insert PCB onto various hash lists.
2026 in_pcbinshash_internal(struct inpcb *inp, int do_pcbgroup_update)
2028 struct inpcbhead *pcbhash;
2029 struct inpcbporthead *pcbporthash;
2030 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
2031 struct inpcbport *phd;
2032 u_int32_t hashkey_faddr;
2034 INP_WLOCK_ASSERT(inp);
2035 INP_HASH_WLOCK_ASSERT(pcbinfo);
2037 KASSERT((inp->inp_flags & INP_INHASHLIST) == 0,
2038 ("in_pcbinshash: INP_INHASHLIST"));
2041 if (inp->inp_vflag & INP_IPV6)
2042 hashkey_faddr = INP6_PCBHASHKEY(&inp->in6p_faddr);
2045 hashkey_faddr = inp->inp_faddr.s_addr;
2047 pcbhash = &pcbinfo->ipi_hashbase[INP_PCBHASH(hashkey_faddr,
2048 inp->inp_lport, inp->inp_fport, pcbinfo->ipi_hashmask)];
2050 pcbporthash = &pcbinfo->ipi_porthashbase[
2051 INP_PCBPORTHASH(inp->inp_lport, pcbinfo->ipi_porthashmask)];
2054 * Go through port list and look for a head for this lport.
2056 LIST_FOREACH(phd, pcbporthash, phd_hash) {
2057 if (phd->phd_port == inp->inp_lport)
2061 * If none exists, malloc one and tack it on.
2064 phd = malloc(sizeof(struct inpcbport), M_PCB, M_NOWAIT);
2066 return (ENOBUFS); /* XXX */
2068 phd->phd_port = inp->inp_lport;
2069 LIST_INIT(&phd->phd_pcblist);
2070 LIST_INSERT_HEAD(pcbporthash, phd, phd_hash);
2073 LIST_INSERT_HEAD(&phd->phd_pcblist, inp, inp_portlist);
2074 LIST_INSERT_HEAD(pcbhash, inp, inp_hash);
2075 inp->inp_flags |= INP_INHASHLIST;
2077 if (do_pcbgroup_update)
2078 in_pcbgroup_update(inp);
2084 * For now, there are two public interfaces to insert an inpcb into the hash
2085 * lists -- one that does update pcbgroups, and one that doesn't. The latter
2086 * is used only in the TCP syncache, where in_pcbinshash is called before the
2087 * full 4-tuple is set for the inpcb, and we don't want to install in the
2088 * pcbgroup until later.
2090 * XXXRW: This seems like a misfeature. in_pcbinshash should always update
2091 * connection groups, and partially initialised inpcbs should not be exposed
2092 * to either reservation hash tables or pcbgroups.
2095 in_pcbinshash(struct inpcb *inp)
2098 return (in_pcbinshash_internal(inp, 1));
2102 in_pcbinshash_nopcbgroup(struct inpcb *inp)
2105 return (in_pcbinshash_internal(inp, 0));
2109 * Move PCB to the proper hash bucket when { faddr, fport } have been
2110 * changed. NOTE: This does not handle the case of the lport changing (the
2111 * hashed port list would have to be updated as well), so the lport must
2112 * not change after in_pcbinshash() has been called.
2115 in_pcbrehash_mbuf(struct inpcb *inp, struct mbuf *m)
2117 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
2118 struct inpcbhead *head;
2119 u_int32_t hashkey_faddr;
2121 INP_WLOCK_ASSERT(inp);
2122 INP_HASH_WLOCK_ASSERT(pcbinfo);
2124 KASSERT(inp->inp_flags & INP_INHASHLIST,
2125 ("in_pcbrehash: !INP_INHASHLIST"));
2128 if (inp->inp_vflag & INP_IPV6)
2129 hashkey_faddr = INP6_PCBHASHKEY(&inp->in6p_faddr);
2132 hashkey_faddr = inp->inp_faddr.s_addr;
2134 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(hashkey_faddr,
2135 inp->inp_lport, inp->inp_fport, pcbinfo->ipi_hashmask)];
2137 LIST_REMOVE(inp, inp_hash);
2138 LIST_INSERT_HEAD(head, inp, inp_hash);
2142 in_pcbgroup_update_mbuf(inp, m);
2144 in_pcbgroup_update(inp);
2149 in_pcbrehash(struct inpcb *inp)
2152 in_pcbrehash_mbuf(inp, NULL);
2156 * Remove PCB from various lists.
2159 in_pcbremlists(struct inpcb *inp)
2161 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
2163 INP_INFO_WLOCK_ASSERT(pcbinfo);
2164 INP_WLOCK_ASSERT(inp);
2166 inp->inp_gencnt = ++pcbinfo->ipi_gencnt;
2167 if (inp->inp_flags & INP_INHASHLIST) {
2168 struct inpcbport *phd = inp->inp_phd;
2170 INP_HASH_WLOCK(pcbinfo);
2171 LIST_REMOVE(inp, inp_hash);
2172 LIST_REMOVE(inp, inp_portlist);
2173 if (LIST_FIRST(&phd->phd_pcblist) == NULL) {
2174 LIST_REMOVE(phd, phd_hash);
2177 INP_HASH_WUNLOCK(pcbinfo);
2178 inp->inp_flags &= ~INP_INHASHLIST;
2180 LIST_REMOVE(inp, inp_list);
2181 pcbinfo->ipi_count--;
2183 in_pcbgroup_remove(inp);
2188 * A set label operation has occurred at the socket layer, propagate the
2189 * label change into the in_pcb for the socket.
2192 in_pcbsosetlabel(struct socket *so)
2197 inp = sotoinpcb(so);
2198 KASSERT(inp != NULL, ("in_pcbsosetlabel: so->so_pcb == NULL"));
2202 mac_inpcb_sosetlabel(so, inp);
2209 * ipport_tick runs once per second, determining if random port allocation
2210 * should be continued. If more than ipport_randomcps ports have been
2211 * allocated in the last second, then we return to sequential port
2212 * allocation. We return to random allocation only once we drop below
2213 * ipport_randomcps for at least ipport_randomtime seconds.
2216 ipport_tick(void *xtp)
2218 VNET_ITERATOR_DECL(vnet_iter);
2220 VNET_LIST_RLOCK_NOSLEEP();
2221 VNET_FOREACH(vnet_iter) {
2222 CURVNET_SET(vnet_iter); /* XXX appease INVARIANTS here */
2223 if (V_ipport_tcpallocs <=
2224 V_ipport_tcplastcount + V_ipport_randomcps) {
2225 if (V_ipport_stoprandom > 0)
2226 V_ipport_stoprandom--;
2228 V_ipport_stoprandom = V_ipport_randomtime;
2229 V_ipport_tcplastcount = V_ipport_tcpallocs;
2232 VNET_LIST_RUNLOCK_NOSLEEP();
2233 callout_reset(&ipport_tick_callout, hz, ipport_tick, NULL);
2240 callout_stop(&ipport_tick_callout);
2244 * The ipport_callout should start running at about the time we attach the
2245 * inet or inet6 domains.
2248 ipport_tick_init(const void *unused __unused)
2251 /* Start ipport_tick. */
2252 callout_init(&ipport_tick_callout, CALLOUT_MPSAFE);
2253 callout_reset(&ipport_tick_callout, 1, ipport_tick, NULL);
2254 EVENTHANDLER_REGISTER(shutdown_pre_sync, ip_fini, NULL,
2255 SHUTDOWN_PRI_DEFAULT);
2257 SYSINIT(ipport_tick_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE,
2258 ipport_tick_init, NULL);
2261 inp_wlock(struct inpcb *inp)
2268 inp_wunlock(struct inpcb *inp)
2275 inp_rlock(struct inpcb *inp)
2282 inp_runlock(struct inpcb *inp)
2290 inp_lock_assert(struct inpcb *inp)
2293 INP_WLOCK_ASSERT(inp);
2297 inp_unlock_assert(struct inpcb *inp)
2300 INP_UNLOCK_ASSERT(inp);
2305 inp_apply_all(void (*func)(struct inpcb *, void *), void *arg)
2309 INP_INFO_RLOCK(&V_tcbinfo);
2310 LIST_FOREACH(inp, V_tcbinfo.ipi_listhead, inp_list) {
2315 INP_INFO_RUNLOCK(&V_tcbinfo);
2319 inp_inpcbtosocket(struct inpcb *inp)
2322 INP_WLOCK_ASSERT(inp);
2323 return (inp->inp_socket);
2327 inp_inpcbtotcpcb(struct inpcb *inp)
2330 INP_WLOCK_ASSERT(inp);
2331 return ((struct tcpcb *)inp->inp_ppcb);
2335 inp_ip_tos_get(const struct inpcb *inp)
2338 return (inp->inp_ip_tos);
2342 inp_ip_tos_set(struct inpcb *inp, int val)
2345 inp->inp_ip_tos = val;
2349 inp_4tuple_get(struct inpcb *inp, uint32_t *laddr, uint16_t *lp,
2350 uint32_t *faddr, uint16_t *fp)
2353 INP_LOCK_ASSERT(inp);
2354 *laddr = inp->inp_laddr.s_addr;
2355 *faddr = inp->inp_faddr.s_addr;
2356 *lp = inp->inp_lport;
2357 *fp = inp->inp_fport;
2361 so_sotoinpcb(struct socket *so)
2364 return (sotoinpcb(so));
2368 so_sototcpcb(struct socket *so)
2371 return (sototcpcb(so));
2376 db_print_indent(int indent)
2380 for (i = 0; i < indent; i++)
2385 db_print_inconninfo(struct in_conninfo *inc, const char *name, int indent)
2387 char faddr_str[48], laddr_str[48];
2389 db_print_indent(indent);
2390 db_printf("%s at %p\n", name, inc);
2395 if (inc->inc_flags & INC_ISIPV6) {
2397 ip6_sprintf(laddr_str, &inc->inc6_laddr);
2398 ip6_sprintf(faddr_str, &inc->inc6_faddr);
2403 inet_ntoa_r(inc->inc_laddr, laddr_str);
2404 inet_ntoa_r(inc->inc_faddr, faddr_str);
2406 db_print_indent(indent);
2407 db_printf("inc_laddr %s inc_lport %u\n", laddr_str,
2408 ntohs(inc->inc_lport));
2409 db_print_indent(indent);
2410 db_printf("inc_faddr %s inc_fport %u\n", faddr_str,
2411 ntohs(inc->inc_fport));
2415 db_print_inpflags(int inp_flags)
2420 if (inp_flags & INP_RECVOPTS) {
2421 db_printf("%sINP_RECVOPTS", comma ? ", " : "");
2424 if (inp_flags & INP_RECVRETOPTS) {
2425 db_printf("%sINP_RECVRETOPTS", comma ? ", " : "");
2428 if (inp_flags & INP_RECVDSTADDR) {
2429 db_printf("%sINP_RECVDSTADDR", comma ? ", " : "");
2432 if (inp_flags & INP_HDRINCL) {
2433 db_printf("%sINP_HDRINCL", comma ? ", " : "");
2436 if (inp_flags & INP_HIGHPORT) {
2437 db_printf("%sINP_HIGHPORT", comma ? ", " : "");
2440 if (inp_flags & INP_LOWPORT) {
2441 db_printf("%sINP_LOWPORT", comma ? ", " : "");
2444 if (inp_flags & INP_ANONPORT) {
2445 db_printf("%sINP_ANONPORT", comma ? ", " : "");
2448 if (inp_flags & INP_RECVIF) {
2449 db_printf("%sINP_RECVIF", comma ? ", " : "");
2452 if (inp_flags & INP_MTUDISC) {
2453 db_printf("%sINP_MTUDISC", comma ? ", " : "");
2456 if (inp_flags & INP_RECVTTL) {
2457 db_printf("%sINP_RECVTTL", comma ? ", " : "");
2460 if (inp_flags & INP_DONTFRAG) {
2461 db_printf("%sINP_DONTFRAG", comma ? ", " : "");
2464 if (inp_flags & INP_RECVTOS) {
2465 db_printf("%sINP_RECVTOS", comma ? ", " : "");
2468 if (inp_flags & IN6P_IPV6_V6ONLY) {
2469 db_printf("%sIN6P_IPV6_V6ONLY", comma ? ", " : "");
2472 if (inp_flags & IN6P_PKTINFO) {
2473 db_printf("%sIN6P_PKTINFO", comma ? ", " : "");
2476 if (inp_flags & IN6P_HOPLIMIT) {
2477 db_printf("%sIN6P_HOPLIMIT", comma ? ", " : "");
2480 if (inp_flags & IN6P_HOPOPTS) {
2481 db_printf("%sIN6P_HOPOPTS", comma ? ", " : "");
2484 if (inp_flags & IN6P_DSTOPTS) {
2485 db_printf("%sIN6P_DSTOPTS", comma ? ", " : "");
2488 if (inp_flags & IN6P_RTHDR) {
2489 db_printf("%sIN6P_RTHDR", comma ? ", " : "");
2492 if (inp_flags & IN6P_RTHDRDSTOPTS) {
2493 db_printf("%sIN6P_RTHDRDSTOPTS", comma ? ", " : "");
2496 if (inp_flags & IN6P_TCLASS) {
2497 db_printf("%sIN6P_TCLASS", comma ? ", " : "");
2500 if (inp_flags & IN6P_AUTOFLOWLABEL) {
2501 db_printf("%sIN6P_AUTOFLOWLABEL", comma ? ", " : "");
2504 if (inp_flags & INP_TIMEWAIT) {
2505 db_printf("%sINP_TIMEWAIT", comma ? ", " : "");
2508 if (inp_flags & INP_ONESBCAST) {
2509 db_printf("%sINP_ONESBCAST", comma ? ", " : "");
2512 if (inp_flags & INP_DROPPED) {
2513 db_printf("%sINP_DROPPED", comma ? ", " : "");
2516 if (inp_flags & INP_SOCKREF) {
2517 db_printf("%sINP_SOCKREF", comma ? ", " : "");
2520 if (inp_flags & IN6P_RFC2292) {
2521 db_printf("%sIN6P_RFC2292", comma ? ", " : "");
2524 if (inp_flags & IN6P_MTU) {
2525 db_printf("IN6P_MTU%s", comma ? ", " : "");
2531 db_print_inpvflag(u_char inp_vflag)
2536 if (inp_vflag & INP_IPV4) {
2537 db_printf("%sINP_IPV4", comma ? ", " : "");
2540 if (inp_vflag & INP_IPV6) {
2541 db_printf("%sINP_IPV6", comma ? ", " : "");
2544 if (inp_vflag & INP_IPV6PROTO) {
2545 db_printf("%sINP_IPV6PROTO", comma ? ", " : "");
2551 db_print_inpcb(struct inpcb *inp, const char *name, int indent)
2554 db_print_indent(indent);
2555 db_printf("%s at %p\n", name, inp);
2559 db_print_indent(indent);
2560 db_printf("inp_flow: 0x%x\n", inp->inp_flow);
2562 db_print_inconninfo(&inp->inp_inc, "inp_conninfo", indent);
2564 db_print_indent(indent);
2565 db_printf("inp_ppcb: %p inp_pcbinfo: %p inp_socket: %p\n",
2566 inp->inp_ppcb, inp->inp_pcbinfo, inp->inp_socket);
2568 db_print_indent(indent);
2569 db_printf("inp_label: %p inp_flags: 0x%x (",
2570 inp->inp_label, inp->inp_flags);
2571 db_print_inpflags(inp->inp_flags);
2574 db_print_indent(indent);
2575 db_printf("inp_sp: %p inp_vflag: 0x%x (", inp->inp_sp,
2577 db_print_inpvflag(inp->inp_vflag);
2580 db_print_indent(indent);
2581 db_printf("inp_ip_ttl: %d inp_ip_p: %d inp_ip_minttl: %d\n",
2582 inp->inp_ip_ttl, inp->inp_ip_p, inp->inp_ip_minttl);
2584 db_print_indent(indent);
2586 if (inp->inp_vflag & INP_IPV6) {
2587 db_printf("in6p_options: %p in6p_outputopts: %p "
2588 "in6p_moptions: %p\n", inp->in6p_options,
2589 inp->in6p_outputopts, inp->in6p_moptions);
2590 db_printf("in6p_icmp6filt: %p in6p_cksum %d "
2591 "in6p_hops %u\n", inp->in6p_icmp6filt, inp->in6p_cksum,
2596 db_printf("inp_ip_tos: %d inp_ip_options: %p "
2597 "inp_ip_moptions: %p\n", inp->inp_ip_tos,
2598 inp->inp_options, inp->inp_moptions);
2601 db_print_indent(indent);
2602 db_printf("inp_phd: %p inp_gencnt: %ju\n", inp->inp_phd,
2603 (uintmax_t)inp->inp_gencnt);
2606 DB_SHOW_COMMAND(inpcb, db_show_inpcb)
2611 db_printf("usage: show inpcb <addr>\n");
2614 inp = (struct inpcb *)addr;
2616 db_print_inpcb(inp, "inpcb", 0);