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>
51 #include <sys/malloc.h>
53 #include <sys/callout.h>
54 #include <sys/eventhandler.h>
55 #include <sys/domain.h>
56 #include <sys/protosw.h>
57 #include <sys/rmlock.h>
58 #include <sys/socket.h>
59 #include <sys/socketvar.h>
62 #include <sys/refcount.h>
64 #include <sys/kernel.h>
65 #include <sys/sysctl.h>
74 #include <net/if_var.h>
75 #include <net/if_types.h>
76 #include <net/if_llatbl.h>
77 #include <net/route.h>
78 #include <net/rss_config.h>
81 #if defined(INET) || defined(INET6)
82 #include <netinet/in.h>
83 #include <netinet/in_pcb.h>
84 #include <netinet/ip_var.h>
85 #include <netinet/tcp_var.h>
86 #include <netinet/udp.h>
87 #include <netinet/udp_var.h>
90 #include <netinet/in_var.h>
93 #include <netinet/ip6.h>
94 #include <netinet6/in6_pcb.h>
95 #include <netinet6/in6_var.h>
96 #include <netinet6/ip6_var.h>
99 #include <netipsec/ipsec_support.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? */
225 INP_LIST_LOCK_INIT(pcbinfo, "pcbinfolist");
227 pcbinfo->ipi_vnet = curvnet;
229 pcbinfo->ipi_listhead = listhead;
230 LIST_INIT(pcbinfo->ipi_listhead);
231 pcbinfo->ipi_count = 0;
232 pcbinfo->ipi_hashbase = hashinit(hash_nelements, M_PCB,
233 &pcbinfo->ipi_hashmask);
234 pcbinfo->ipi_porthashbase = hashinit(porthash_nelements, M_PCB,
235 &pcbinfo->ipi_porthashmask);
237 in_pcbgroup_init(pcbinfo, hashfields, hash_nelements);
239 pcbinfo->ipi_zone = uma_zcreate(inpcbzone_name, sizeof(struct inpcb),
240 NULL, NULL, inpcbzone_init, inpcbzone_fini, UMA_ALIGN_PTR,
242 uma_zone_set_max(pcbinfo->ipi_zone, maxsockets);
243 uma_zone_set_warning(pcbinfo->ipi_zone,
244 "kern.ipc.maxsockets limit reached");
248 * Destroy an inpcbinfo.
251 in_pcbinfo_destroy(struct inpcbinfo *pcbinfo)
254 KASSERT(pcbinfo->ipi_count == 0,
255 ("%s: ipi_count = %u", __func__, pcbinfo->ipi_count));
257 hashdestroy(pcbinfo->ipi_hashbase, M_PCB, pcbinfo->ipi_hashmask);
258 hashdestroy(pcbinfo->ipi_porthashbase, M_PCB,
259 pcbinfo->ipi_porthashmask);
261 in_pcbgroup_destroy(pcbinfo);
263 uma_zdestroy(pcbinfo->ipi_zone);
264 INP_LIST_LOCK_DESTROY(pcbinfo);
265 INP_HASH_LOCK_DESTROY(pcbinfo);
266 INP_INFO_LOCK_DESTROY(pcbinfo);
270 * Allocate a PCB and associate it with the socket.
271 * On success return with the PCB locked.
274 in_pcballoc(struct socket *so, struct inpcbinfo *pcbinfo)
280 if (pcbinfo == &V_tcbinfo) {
281 INP_INFO_RLOCK_ASSERT(pcbinfo);
283 INP_INFO_WLOCK_ASSERT(pcbinfo);
288 inp = uma_zalloc(pcbinfo->ipi_zone, M_NOWAIT);
291 bzero(inp, inp_zero_size);
292 inp->inp_pcbinfo = pcbinfo;
293 inp->inp_socket = so;
294 inp->inp_cred = crhold(so->so_cred);
295 inp->inp_inc.inc_fibnum = so->so_fibnum;
297 error = mac_inpcb_init(inp, M_NOWAIT);
300 mac_inpcb_create(so, inp);
302 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
303 error = ipsec_init_pcbpolicy(inp);
306 mac_inpcb_destroy(inp);
312 if (INP_SOCKAF(so) == AF_INET6) {
313 inp->inp_vflag |= INP_IPV6PROTO;
315 inp->inp_flags |= IN6P_IPV6_V6ONLY;
319 INP_LIST_WLOCK(pcbinfo);
320 LIST_INSERT_HEAD(pcbinfo->ipi_listhead, inp, inp_list);
321 pcbinfo->ipi_count++;
322 so->so_pcb = (caddr_t)inp;
324 if (V_ip6_auto_flowlabel)
325 inp->inp_flags |= IN6P_AUTOFLOWLABEL;
327 inp->inp_gencnt = ++pcbinfo->ipi_gencnt;
328 refcount_init(&inp->inp_refcount, 1); /* Reference from inpcbinfo */
331 * Routes in inpcb's can cache L2 as well; they are guaranteed
334 inp->inp_route.ro_flags = RT_LLE_CACHE;
335 INP_LIST_WUNLOCK(pcbinfo);
336 #if defined(IPSEC) || defined(IPSEC_SUPPORT) || defined(MAC)
339 crfree(inp->inp_cred);
340 uma_zfree(pcbinfo->ipi_zone, inp);
348 in_pcbbind(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred)
352 INP_WLOCK_ASSERT(inp);
353 INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo);
355 if (inp->inp_lport != 0 || inp->inp_laddr.s_addr != INADDR_ANY)
357 anonport = nam == NULL || ((struct sockaddr_in *)nam)->sin_port == 0;
358 error = in_pcbbind_setup(inp, nam, &inp->inp_laddr.s_addr,
359 &inp->inp_lport, cred);
362 if (in_pcbinshash(inp) != 0) {
363 inp->inp_laddr.s_addr = INADDR_ANY;
368 inp->inp_flags |= INP_ANONPORT;
374 * Select a local port (number) to use.
376 #if defined(INET) || defined(INET6)
378 in_pcb_lport(struct inpcb *inp, struct in_addr *laddrp, u_short *lportp,
379 struct ucred *cred, int lookupflags)
381 struct inpcbinfo *pcbinfo;
382 struct inpcb *tmpinp;
383 unsigned short *lastport;
384 int count, dorandom, error;
385 u_short aux, first, last, lport;
387 struct in_addr laddr;
390 pcbinfo = inp->inp_pcbinfo;
393 * Because no actual state changes occur here, a global write lock on
394 * the pcbinfo isn't required.
396 INP_LOCK_ASSERT(inp);
397 INP_HASH_LOCK_ASSERT(pcbinfo);
399 if (inp->inp_flags & INP_HIGHPORT) {
400 first = V_ipport_hifirstauto; /* sysctl */
401 last = V_ipport_hilastauto;
402 lastport = &pcbinfo->ipi_lasthi;
403 } else if (inp->inp_flags & INP_LOWPORT) {
404 error = priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT, 0);
407 first = V_ipport_lowfirstauto; /* 1023 */
408 last = V_ipport_lowlastauto; /* 600 */
409 lastport = &pcbinfo->ipi_lastlow;
411 first = V_ipport_firstauto; /* sysctl */
412 last = V_ipport_lastauto;
413 lastport = &pcbinfo->ipi_lastport;
416 * For UDP(-Lite), use random port allocation as long as the user
417 * allows it. For TCP (and as of yet unknown) connections,
418 * use random port allocation only if the user allows it AND
419 * ipport_tick() allows it.
421 if (V_ipport_randomized &&
422 (!V_ipport_stoprandom || pcbinfo == &V_udbinfo ||
423 pcbinfo == &V_ulitecbinfo))
428 * It makes no sense to do random port allocation if
429 * we have the only port available.
433 /* Make sure to not include UDP(-Lite) packets in the count. */
434 if (pcbinfo != &V_udbinfo || pcbinfo != &V_ulitecbinfo)
435 V_ipport_tcpallocs++;
437 * Instead of having two loops further down counting up or down
438 * make sure that first is always <= last and go with only one
439 * code path implementing all logic.
448 /* Make the compiler happy. */
450 if ((inp->inp_vflag & (INP_IPV4|INP_IPV6)) == INP_IPV4) {
451 KASSERT(laddrp != NULL, ("%s: laddrp NULL for v4 inp %p",
456 tmpinp = NULL; /* Make compiler happy. */
460 *lastport = first + (arc4random() % (last - first));
462 count = last - first;
465 if (count-- < 0) /* completely used? */
466 return (EADDRNOTAVAIL);
468 if (*lastport < first || *lastport > last)
470 lport = htons(*lastport);
473 if ((inp->inp_vflag & INP_IPV6) != 0)
474 tmpinp = in6_pcblookup_local(pcbinfo,
475 &inp->in6p_laddr, lport, lookupflags, cred);
477 #if defined(INET) && defined(INET6)
481 tmpinp = in_pcblookup_local(pcbinfo, laddr,
482 lport, lookupflags, cred);
484 } while (tmpinp != NULL);
487 if ((inp->inp_vflag & (INP_IPV4|INP_IPV6)) == INP_IPV4)
488 laddrp->s_addr = laddr.s_addr;
496 * Return cached socket options.
499 inp_so_options(const struct inpcb *inp)
505 if ((inp->inp_flags2 & INP_REUSEPORT) != 0)
506 so_options |= SO_REUSEPORT;
507 if ((inp->inp_flags2 & INP_REUSEADDR) != 0)
508 so_options |= SO_REUSEADDR;
511 #endif /* INET || INET6 */
514 * Check if a new BINDMULTI socket is allowed to be created.
516 * ni points to the new inp.
517 * oi points to the exisitng inp.
519 * This checks whether the existing inp also has BINDMULTI and
520 * whether the credentials match.
523 in_pcbbind_check_bindmulti(const struct inpcb *ni, const struct inpcb *oi)
525 /* Check permissions match */
526 if ((ni->inp_flags2 & INP_BINDMULTI) &&
527 (ni->inp_cred->cr_uid !=
528 oi->inp_cred->cr_uid))
531 /* Check the existing inp has BINDMULTI set */
532 if ((ni->inp_flags2 & INP_BINDMULTI) &&
533 ((oi->inp_flags2 & INP_BINDMULTI) == 0))
537 * We're okay - either INP_BINDMULTI isn't set on ni, or
538 * it is and it matches the checks.
545 * Set up a bind operation on a PCB, performing port allocation
546 * as required, but do not actually modify the PCB. Callers can
547 * either complete the bind by setting inp_laddr/inp_lport and
548 * calling in_pcbinshash(), or they can just use the resulting
549 * port and address to authorise the sending of a once-off packet.
551 * On error, the values of *laddrp and *lportp are not changed.
554 in_pcbbind_setup(struct inpcb *inp, struct sockaddr *nam, in_addr_t *laddrp,
555 u_short *lportp, struct ucred *cred)
557 struct socket *so = inp->inp_socket;
558 struct sockaddr_in *sin;
559 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
560 struct in_addr laddr;
562 int lookupflags = 0, reuseport = (so->so_options & SO_REUSEPORT);
566 * No state changes, so read locks are sufficient here.
568 INP_LOCK_ASSERT(inp);
569 INP_HASH_LOCK_ASSERT(pcbinfo);
571 if (TAILQ_EMPTY(&V_in_ifaddrhead)) /* XXX broken! */
572 return (EADDRNOTAVAIL);
573 laddr.s_addr = *laddrp;
574 if (nam != NULL && laddr.s_addr != INADDR_ANY)
576 if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0)
577 lookupflags = INPLOOKUP_WILDCARD;
579 if ((error = prison_local_ip4(cred, &laddr)) != 0)
582 sin = (struct sockaddr_in *)nam;
583 if (nam->sa_len != sizeof (*sin))
587 * We should check the family, but old programs
588 * incorrectly fail to initialize it.
590 if (sin->sin_family != AF_INET)
591 return (EAFNOSUPPORT);
593 error = prison_local_ip4(cred, &sin->sin_addr);
596 if (sin->sin_port != *lportp) {
597 /* Don't allow the port to change. */
600 lport = sin->sin_port;
602 /* NB: lport is left as 0 if the port isn't being changed. */
603 if (IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) {
605 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
606 * allow complete duplication of binding if
607 * SO_REUSEPORT is set, or if SO_REUSEADDR is set
608 * and a multicast address is bound on both
609 * new and duplicated sockets.
611 if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) != 0)
612 reuseport = SO_REUSEADDR|SO_REUSEPORT;
613 } else if (sin->sin_addr.s_addr != INADDR_ANY) {
614 sin->sin_port = 0; /* yech... */
615 bzero(&sin->sin_zero, sizeof(sin->sin_zero));
617 * Is the address a local IP address?
618 * If INP_BINDANY is set, then the socket may be bound
619 * to any endpoint address, local or not.
621 if ((inp->inp_flags & INP_BINDANY) == 0 &&
622 ifa_ifwithaddr_check((struct sockaddr *)sin) == 0)
623 return (EADDRNOTAVAIL);
625 laddr = sin->sin_addr;
631 if (ntohs(lport) <= V_ipport_reservedhigh &&
632 ntohs(lport) >= V_ipport_reservedlow &&
633 priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT,
636 if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)) &&
637 priv_check_cred(inp->inp_cred,
638 PRIV_NETINET_REUSEPORT, 0) != 0) {
639 t = in_pcblookup_local(pcbinfo, sin->sin_addr,
640 lport, INPLOOKUP_WILDCARD, cred);
643 * This entire block sorely needs a rewrite.
646 ((inp->inp_flags2 & INP_BINDMULTI) == 0) &&
647 ((t->inp_flags & INP_TIMEWAIT) == 0) &&
648 (so->so_type != SOCK_STREAM ||
649 ntohl(t->inp_faddr.s_addr) == INADDR_ANY) &&
650 (ntohl(sin->sin_addr.s_addr) != INADDR_ANY ||
651 ntohl(t->inp_laddr.s_addr) != INADDR_ANY ||
652 (t->inp_flags2 & INP_REUSEPORT) == 0) &&
653 (inp->inp_cred->cr_uid !=
654 t->inp_cred->cr_uid))
658 * If the socket is a BINDMULTI socket, then
659 * the credentials need to match and the
660 * original socket also has to have been bound
663 if (t && (! in_pcbbind_check_bindmulti(inp, t)))
666 t = in_pcblookup_local(pcbinfo, sin->sin_addr,
667 lport, lookupflags, cred);
668 if (t && (t->inp_flags & INP_TIMEWAIT)) {
670 * XXXRW: If an incpb has had its timewait
671 * state recycled, we treat the address as
672 * being in use (for now). This is better
673 * than a panic, but not desirable.
677 (reuseport & tw->tw_so_options) == 0)
680 ((inp->inp_flags2 & INP_BINDMULTI) == 0) &&
681 (reuseport & inp_so_options(t)) == 0) {
683 if (ntohl(sin->sin_addr.s_addr) !=
685 ntohl(t->inp_laddr.s_addr) !=
687 (inp->inp_vflag & INP_IPV6PROTO) == 0 ||
688 (t->inp_vflag & INP_IPV6PROTO) == 0)
691 if (t && (! in_pcbbind_check_bindmulti(inp, t)))
699 error = in_pcb_lport(inp, &laddr, &lport, cred, lookupflags);
704 *laddrp = laddr.s_addr;
710 * Connect from a socket to a specified address.
711 * Both address and port must be specified in argument sin.
712 * If don't have a local address for this socket yet,
716 in_pcbconnect_mbuf(struct inpcb *inp, struct sockaddr *nam,
717 struct ucred *cred, struct mbuf *m)
719 u_short lport, fport;
720 in_addr_t laddr, faddr;
723 INP_WLOCK_ASSERT(inp);
724 INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo);
726 lport = inp->inp_lport;
727 laddr = inp->inp_laddr.s_addr;
728 anonport = (lport == 0);
729 error = in_pcbconnect_setup(inp, nam, &laddr, &lport, &faddr, &fport,
734 /* Do the initial binding of the local address if required. */
735 if (inp->inp_laddr.s_addr == INADDR_ANY && inp->inp_lport == 0) {
736 inp->inp_lport = lport;
737 inp->inp_laddr.s_addr = laddr;
738 if (in_pcbinshash(inp) != 0) {
739 inp->inp_laddr.s_addr = INADDR_ANY;
745 /* Commit the remaining changes. */
746 inp->inp_lport = lport;
747 inp->inp_laddr.s_addr = laddr;
748 inp->inp_faddr.s_addr = faddr;
749 inp->inp_fport = fport;
750 in_pcbrehash_mbuf(inp, m);
753 inp->inp_flags |= INP_ANONPORT;
758 in_pcbconnect(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred)
761 return (in_pcbconnect_mbuf(inp, nam, cred, NULL));
765 * Do proper source address selection on an unbound socket in case
766 * of connect. Take jails into account as well.
769 in_pcbladdr(struct inpcb *inp, struct in_addr *faddr, struct in_addr *laddr,
774 struct sockaddr_in *sin;
778 KASSERT(laddr != NULL, ("%s: laddr NULL", __func__));
781 * Bypass source address selection and use the primary jail IP
784 if (cred != NULL && !prison_saddrsel_ip4(cred, laddr))
788 bzero(&sro, sizeof(sro));
790 sin = (struct sockaddr_in *)&sro.ro_dst;
791 sin->sin_family = AF_INET;
792 sin->sin_len = sizeof(struct sockaddr_in);
793 sin->sin_addr.s_addr = faddr->s_addr;
796 * If route is known our src addr is taken from the i/f,
799 * Find out route to destination.
801 if ((inp->inp_socket->so_options & SO_DONTROUTE) == 0)
802 in_rtalloc_ign(&sro, 0, inp->inp_inc.inc_fibnum);
805 * If we found a route, use the address corresponding to
806 * the outgoing interface.
808 * Otherwise assume faddr is reachable on a directly connected
809 * network and try to find a corresponding interface to take
810 * the source address from.
812 if (sro.ro_rt == NULL || sro.ro_rt->rt_ifp == NULL) {
813 struct in_ifaddr *ia;
816 ia = ifatoia(ifa_ifwithdstaddr((struct sockaddr *)sin,
817 inp->inp_socket->so_fibnum));
819 ia = ifatoia(ifa_ifwithnet((struct sockaddr *)sin, 0,
820 inp->inp_socket->so_fibnum));
826 if (cred == NULL || !prison_flag(cred, PR_IP4)) {
827 laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
828 ifa_free(&ia->ia_ifa);
833 ifa_free(&ia->ia_ifa);
836 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
839 if (sa->sa_family != AF_INET)
841 sin = (struct sockaddr_in *)sa;
842 if (prison_check_ip4(cred, &sin->sin_addr) == 0) {
843 ia = (struct in_ifaddr *)ifa;
848 laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
849 IF_ADDR_RUNLOCK(ifp);
852 IF_ADDR_RUNLOCK(ifp);
854 /* 3. As a last resort return the 'default' jail address. */
855 error = prison_get_ip4(cred, laddr);
860 * If the outgoing interface on the route found is not
861 * a loopback interface, use the address from that interface.
862 * In case of jails do those three steps:
863 * 1. check if the interface address belongs to the jail. If so use it.
864 * 2. check if we have any address on the outgoing interface
865 * belonging to this jail. If so use it.
866 * 3. as a last resort return the 'default' jail address.
868 if ((sro.ro_rt->rt_ifp->if_flags & IFF_LOOPBACK) == 0) {
869 struct in_ifaddr *ia;
872 /* If not jailed, use the default returned. */
873 if (cred == NULL || !prison_flag(cred, PR_IP4)) {
874 ia = (struct in_ifaddr *)sro.ro_rt->rt_ifa;
875 laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
880 /* 1. Check if the iface address belongs to the jail. */
881 sin = (struct sockaddr_in *)sro.ro_rt->rt_ifa->ifa_addr;
882 if (prison_check_ip4(cred, &sin->sin_addr) == 0) {
883 ia = (struct in_ifaddr *)sro.ro_rt->rt_ifa;
884 laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
889 * 2. Check if we have any address on the outgoing interface
890 * belonging to this jail.
893 ifp = sro.ro_rt->rt_ifp;
895 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
897 if (sa->sa_family != AF_INET)
899 sin = (struct sockaddr_in *)sa;
900 if (prison_check_ip4(cred, &sin->sin_addr) == 0) {
901 ia = (struct in_ifaddr *)ifa;
906 laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
907 IF_ADDR_RUNLOCK(ifp);
910 IF_ADDR_RUNLOCK(ifp);
912 /* 3. As a last resort return the 'default' jail address. */
913 error = prison_get_ip4(cred, laddr);
918 * The outgoing interface is marked with 'loopback net', so a route
919 * to ourselves is here.
920 * Try to find the interface of the destination address and then
921 * take the address from there. That interface is not necessarily
922 * a loopback interface.
923 * In case of jails, check that it is an address of the jail
924 * and if we cannot find, fall back to the 'default' jail address.
926 if ((sro.ro_rt->rt_ifp->if_flags & IFF_LOOPBACK) != 0) {
927 struct sockaddr_in sain;
928 struct in_ifaddr *ia;
930 bzero(&sain, sizeof(struct sockaddr_in));
931 sain.sin_family = AF_INET;
932 sain.sin_len = sizeof(struct sockaddr_in);
933 sain.sin_addr.s_addr = faddr->s_addr;
935 ia = ifatoia(ifa_ifwithdstaddr(sintosa(&sain),
936 inp->inp_socket->so_fibnum));
938 ia = ifatoia(ifa_ifwithnet(sintosa(&sain), 0,
939 inp->inp_socket->so_fibnum));
941 ia = ifatoia(ifa_ifwithaddr(sintosa(&sain)));
943 if (cred == NULL || !prison_flag(cred, PR_IP4)) {
948 laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
949 ifa_free(&ia->ia_ifa);
958 ifa_free(&ia->ia_ifa);
961 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
964 if (sa->sa_family != AF_INET)
966 sin = (struct sockaddr_in *)sa;
967 if (prison_check_ip4(cred,
968 &sin->sin_addr) == 0) {
969 ia = (struct in_ifaddr *)ifa;
974 laddr->s_addr = ia->ia_addr.sin_addr.s_addr;
975 IF_ADDR_RUNLOCK(ifp);
978 IF_ADDR_RUNLOCK(ifp);
981 /* 3. As a last resort return the 'default' jail address. */
982 error = prison_get_ip4(cred, laddr);
987 if (sro.ro_rt != NULL)
993 * Set up for a connect from a socket to the specified address.
994 * On entry, *laddrp and *lportp should contain the current local
995 * address and port for the PCB; these are updated to the values
996 * that should be placed in inp_laddr and inp_lport to complete
999 * On success, *faddrp and *fportp will be set to the remote address
1000 * and port. These are not updated in the error case.
1002 * If the operation fails because the connection already exists,
1003 * *oinpp will be set to the PCB of that connection so that the
1004 * caller can decide to override it. In all other cases, *oinpp
1008 in_pcbconnect_setup(struct inpcb *inp, struct sockaddr *nam,
1009 in_addr_t *laddrp, u_short *lportp, in_addr_t *faddrp, u_short *fportp,
1010 struct inpcb **oinpp, struct ucred *cred)
1012 struct rm_priotracker in_ifa_tracker;
1013 struct sockaddr_in *sin = (struct sockaddr_in *)nam;
1014 struct in_ifaddr *ia;
1016 struct in_addr laddr, faddr;
1017 u_short lport, fport;
1021 * Because a global state change doesn't actually occur here, a read
1022 * lock is sufficient.
1024 INP_LOCK_ASSERT(inp);
1025 INP_HASH_LOCK_ASSERT(inp->inp_pcbinfo);
1029 if (nam->sa_len != sizeof (*sin))
1031 if (sin->sin_family != AF_INET)
1032 return (EAFNOSUPPORT);
1033 if (sin->sin_port == 0)
1034 return (EADDRNOTAVAIL);
1035 laddr.s_addr = *laddrp;
1037 faddr = sin->sin_addr;
1038 fport = sin->sin_port;
1040 if (!TAILQ_EMPTY(&V_in_ifaddrhead)) {
1042 * If the destination address is INADDR_ANY,
1043 * use the primary local address.
1044 * If the supplied address is INADDR_BROADCAST,
1045 * and the primary interface supports broadcast,
1046 * choose the broadcast address for that interface.
1048 if (faddr.s_addr == INADDR_ANY) {
1049 IN_IFADDR_RLOCK(&in_ifa_tracker);
1051 IA_SIN(TAILQ_FIRST(&V_in_ifaddrhead))->sin_addr;
1052 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
1054 (error = prison_get_ip4(cred, &faddr)) != 0)
1056 } else if (faddr.s_addr == (u_long)INADDR_BROADCAST) {
1057 IN_IFADDR_RLOCK(&in_ifa_tracker);
1058 if (TAILQ_FIRST(&V_in_ifaddrhead)->ia_ifp->if_flags &
1060 faddr = satosin(&TAILQ_FIRST(
1061 &V_in_ifaddrhead)->ia_broadaddr)->sin_addr;
1062 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
1065 if (laddr.s_addr == INADDR_ANY) {
1066 error = in_pcbladdr(inp, &faddr, &laddr, cred);
1068 * If the destination address is multicast and an outgoing
1069 * interface has been set as a multicast option, prefer the
1070 * address of that interface as our source address.
1072 if (IN_MULTICAST(ntohl(faddr.s_addr)) &&
1073 inp->inp_moptions != NULL) {
1074 struct ip_moptions *imo;
1077 imo = inp->inp_moptions;
1078 if (imo->imo_multicast_ifp != NULL) {
1079 ifp = imo->imo_multicast_ifp;
1080 IN_IFADDR_RLOCK(&in_ifa_tracker);
1081 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
1082 if ((ia->ia_ifp == ifp) &&
1084 prison_check_ip4(cred,
1085 &ia->ia_addr.sin_addr) == 0))
1089 error = EADDRNOTAVAIL;
1091 laddr = ia->ia_addr.sin_addr;
1094 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
1100 oinp = in_pcblookup_hash_locked(inp->inp_pcbinfo, faddr, fport,
1101 laddr, lport, 0, NULL);
1105 return (EADDRINUSE);
1108 error = in_pcbbind_setup(inp, NULL, &laddr.s_addr, &lport,
1113 *laddrp = laddr.s_addr;
1115 *faddrp = faddr.s_addr;
1121 in_pcbdisconnect(struct inpcb *inp)
1124 INP_WLOCK_ASSERT(inp);
1125 INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo);
1127 inp->inp_faddr.s_addr = INADDR_ANY;
1134 * in_pcbdetach() is responsibe for disassociating a socket from an inpcb.
1135 * For most protocols, this will be invoked immediately prior to calling
1136 * in_pcbfree(). However, with TCP the inpcb may significantly outlive the
1137 * socket, in which case in_pcbfree() is deferred.
1140 in_pcbdetach(struct inpcb *inp)
1143 KASSERT(inp->inp_socket != NULL, ("%s: inp_socket == NULL", __func__));
1145 inp->inp_socket->so_pcb = NULL;
1146 inp->inp_socket = NULL;
1150 * in_pcbref() bumps the reference count on an inpcb in order to maintain
1151 * stability of an inpcb pointer despite the inpcb lock being released. This
1152 * is used in TCP when the inpcbinfo lock needs to be acquired or upgraded,
1153 * but where the inpcb lock may already held, or when acquiring a reference
1156 * in_pcbref() should be used only to provide brief memory stability, and
1157 * must always be followed by a call to INP_WLOCK() and in_pcbrele() to
1158 * garbage collect the inpcb if it has been in_pcbfree()'d from another
1159 * context. Until in_pcbrele() has returned that the inpcb is still valid,
1160 * lock and rele are the *only* safe operations that may be performed on the
1163 * While the inpcb will not be freed, releasing the inpcb lock means that the
1164 * connection's state may change, so the caller should be careful to
1165 * revalidate any cached state on reacquiring the lock. Drop the reference
1166 * using in_pcbrele().
1169 in_pcbref(struct inpcb *inp)
1172 KASSERT(inp->inp_refcount > 0, ("%s: refcount 0", __func__));
1174 refcount_acquire(&inp->inp_refcount);
1178 * Drop a refcount on an inpcb elevated using in_pcbref(); because a call to
1179 * in_pcbfree() may have been made between in_pcbref() and in_pcbrele(), we
1180 * return a flag indicating whether or not the inpcb remains valid. If it is
1181 * valid, we return with the inpcb lock held.
1183 * Notice that, unlike in_pcbref(), the inpcb lock must be held to drop a
1184 * reference on an inpcb. Historically more work was done here (actually, in
1185 * in_pcbfree_internal()) but has been moved to in_pcbfree() to avoid the
1186 * need for the pcbinfo lock in in_pcbrele(). Deferring the free is entirely
1187 * about memory stability (and continued use of the write lock).
1190 in_pcbrele_rlocked(struct inpcb *inp)
1192 struct inpcbinfo *pcbinfo;
1194 KASSERT(inp->inp_refcount > 0, ("%s: refcount 0", __func__));
1196 INP_RLOCK_ASSERT(inp);
1198 if (refcount_release(&inp->inp_refcount) == 0) {
1200 * If the inpcb has been freed, let the caller know, even if
1201 * this isn't the last reference.
1203 if (inp->inp_flags2 & INP_FREED) {
1210 KASSERT(inp->inp_socket == NULL, ("%s: inp_socket != NULL", __func__));
1213 pcbinfo = inp->inp_pcbinfo;
1214 uma_zfree(pcbinfo->ipi_zone, inp);
1219 in_pcbrele_wlocked(struct inpcb *inp)
1221 struct inpcbinfo *pcbinfo;
1223 KASSERT(inp->inp_refcount > 0, ("%s: refcount 0", __func__));
1225 INP_WLOCK_ASSERT(inp);
1227 if (refcount_release(&inp->inp_refcount) == 0) {
1229 * If the inpcb has been freed, let the caller know, even if
1230 * this isn't the last reference.
1232 if (inp->inp_flags2 & INP_FREED) {
1239 KASSERT(inp->inp_socket == NULL, ("%s: inp_socket != NULL", __func__));
1242 pcbinfo = inp->inp_pcbinfo;
1243 uma_zfree(pcbinfo->ipi_zone, inp);
1248 * Temporary wrapper.
1251 in_pcbrele(struct inpcb *inp)
1254 return (in_pcbrele_wlocked(inp));
1258 * Unconditionally schedule an inpcb to be freed by decrementing its
1259 * reference count, which should occur only after the inpcb has been detached
1260 * from its socket. If another thread holds a temporary reference (acquired
1261 * using in_pcbref()) then the free is deferred until that reference is
1262 * released using in_pcbrele(), but the inpcb is still unlocked. Almost all
1263 * work, including removal from global lists, is done in this context, where
1264 * the pcbinfo lock is held.
1267 in_pcbfree(struct inpcb *inp)
1269 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1271 KASSERT(inp->inp_socket == NULL, ("%s: inp_socket != NULL", __func__));
1274 if (pcbinfo == &V_tcbinfo) {
1275 INP_INFO_LOCK_ASSERT(pcbinfo);
1277 INP_INFO_WLOCK_ASSERT(pcbinfo);
1280 INP_WLOCK_ASSERT(inp);
1282 /* XXXRW: Do as much as possible here. */
1283 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1284 if (inp->inp_sp != NULL)
1285 ipsec_delete_pcbpolicy(inp);
1287 INP_LIST_WLOCK(pcbinfo);
1288 inp->inp_gencnt = ++pcbinfo->ipi_gencnt;
1289 in_pcbremlists(inp);
1290 INP_LIST_WUNLOCK(pcbinfo);
1292 if (inp->inp_vflag & INP_IPV6PROTO) {
1293 ip6_freepcbopts(inp->in6p_outputopts);
1294 if (inp->in6p_moptions != NULL)
1295 ip6_freemoptions(inp->in6p_moptions);
1298 if (inp->inp_options)
1299 (void)m_free(inp->inp_options);
1301 if (inp->inp_moptions != NULL)
1302 inp_freemoptions(inp->inp_moptions);
1304 if (inp->inp_route.ro_rt) {
1305 RTFREE(inp->inp_route.ro_rt);
1306 inp->inp_route.ro_rt = (struct rtentry *)NULL;
1308 if (inp->inp_route.ro_lle)
1309 LLE_FREE(inp->inp_route.ro_lle); /* zeros ro_lle */
1312 inp->inp_flags2 |= INP_FREED;
1313 crfree(inp->inp_cred);
1315 mac_inpcb_destroy(inp);
1317 if (!in_pcbrele_wlocked(inp))
1322 * in_pcbdrop() removes an inpcb from hashed lists, releasing its address and
1323 * port reservation, and preventing it from being returned by inpcb lookups.
1325 * It is used by TCP to mark an inpcb as unused and avoid future packet
1326 * delivery or event notification when a socket remains open but TCP has
1327 * closed. This might occur as a result of a shutdown()-initiated TCP close
1328 * or a RST on the wire, and allows the port binding to be reused while still
1329 * maintaining the invariant that so_pcb always points to a valid inpcb until
1332 * XXXRW: Possibly in_pcbdrop() should also prevent future notifications by
1333 * in_pcbnotifyall() and in_pcbpurgeif0()?
1336 in_pcbdrop(struct inpcb *inp)
1339 INP_WLOCK_ASSERT(inp);
1342 * XXXRW: Possibly we should protect the setting of INP_DROPPED with
1345 inp->inp_flags |= INP_DROPPED;
1346 if (inp->inp_flags & INP_INHASHLIST) {
1347 struct inpcbport *phd = inp->inp_phd;
1349 INP_HASH_WLOCK(inp->inp_pcbinfo);
1350 LIST_REMOVE(inp, inp_hash);
1351 LIST_REMOVE(inp, inp_portlist);
1352 if (LIST_FIRST(&phd->phd_pcblist) == NULL) {
1353 LIST_REMOVE(phd, phd_hash);
1356 INP_HASH_WUNLOCK(inp->inp_pcbinfo);
1357 inp->inp_flags &= ~INP_INHASHLIST;
1359 in_pcbgroup_remove(inp);
1366 * Common routines to return the socket addresses associated with inpcbs.
1369 in_sockaddr(in_port_t port, struct in_addr *addr_p)
1371 struct sockaddr_in *sin;
1373 sin = malloc(sizeof *sin, M_SONAME,
1375 sin->sin_family = AF_INET;
1376 sin->sin_len = sizeof(*sin);
1377 sin->sin_addr = *addr_p;
1378 sin->sin_port = port;
1380 return (struct sockaddr *)sin;
1384 in_getsockaddr(struct socket *so, struct sockaddr **nam)
1387 struct in_addr addr;
1390 inp = sotoinpcb(so);
1391 KASSERT(inp != NULL, ("in_getsockaddr: inp == NULL"));
1394 port = inp->inp_lport;
1395 addr = inp->inp_laddr;
1398 *nam = in_sockaddr(port, &addr);
1403 in_getpeeraddr(struct socket *so, struct sockaddr **nam)
1406 struct in_addr addr;
1409 inp = sotoinpcb(so);
1410 KASSERT(inp != NULL, ("in_getpeeraddr: inp == NULL"));
1413 port = inp->inp_fport;
1414 addr = inp->inp_faddr;
1417 *nam = in_sockaddr(port, &addr);
1422 in_pcbnotifyall(struct inpcbinfo *pcbinfo, struct in_addr faddr, int errno,
1423 struct inpcb *(*notify)(struct inpcb *, int))
1425 struct inpcb *inp, *inp_temp;
1427 INP_INFO_WLOCK(pcbinfo);
1428 LIST_FOREACH_SAFE(inp, pcbinfo->ipi_listhead, inp_list, inp_temp) {
1431 if ((inp->inp_vflag & INP_IPV4) == 0) {
1436 if (inp->inp_faddr.s_addr != faddr.s_addr ||
1437 inp->inp_socket == NULL) {
1441 if ((*notify)(inp, errno))
1444 INP_INFO_WUNLOCK(pcbinfo);
1448 in_pcbpurgeif0(struct inpcbinfo *pcbinfo, struct ifnet *ifp)
1451 struct ip_moptions *imo;
1454 INP_INFO_WLOCK(pcbinfo);
1455 LIST_FOREACH(inp, pcbinfo->ipi_listhead, inp_list) {
1457 imo = inp->inp_moptions;
1458 if ((inp->inp_vflag & INP_IPV4) &&
1461 * Unselect the outgoing interface if it is being
1464 if (imo->imo_multicast_ifp == ifp)
1465 imo->imo_multicast_ifp = NULL;
1468 * Drop multicast group membership if we joined
1469 * through the interface being detached.
1471 for (i = 0, gap = 0; i < imo->imo_num_memberships;
1473 if (imo->imo_membership[i]->inm_ifp == ifp) {
1474 in_delmulti(imo->imo_membership[i]);
1476 } else if (gap != 0)
1477 imo->imo_membership[i - gap] =
1478 imo->imo_membership[i];
1480 imo->imo_num_memberships -= gap;
1484 INP_INFO_WUNLOCK(pcbinfo);
1488 * Lookup a PCB based on the local address and port. Caller must hold the
1489 * hash lock. No inpcb locks or references are acquired.
1491 #define INP_LOOKUP_MAPPED_PCB_COST 3
1493 in_pcblookup_local(struct inpcbinfo *pcbinfo, struct in_addr laddr,
1494 u_short lport, int lookupflags, struct ucred *cred)
1498 int matchwild = 3 + INP_LOOKUP_MAPPED_PCB_COST;
1504 KASSERT((lookupflags & ~(INPLOOKUP_WILDCARD)) == 0,
1505 ("%s: invalid lookup flags %d", __func__, lookupflags));
1507 INP_HASH_LOCK_ASSERT(pcbinfo);
1509 if ((lookupflags & INPLOOKUP_WILDCARD) == 0) {
1510 struct inpcbhead *head;
1512 * Look for an unconnected (wildcard foreign addr) PCB that
1513 * matches the local address and port we're looking for.
1515 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport,
1516 0, pcbinfo->ipi_hashmask)];
1517 LIST_FOREACH(inp, head, inp_hash) {
1519 /* XXX inp locking */
1520 if ((inp->inp_vflag & INP_IPV4) == 0)
1523 if (inp->inp_faddr.s_addr == INADDR_ANY &&
1524 inp->inp_laddr.s_addr == laddr.s_addr &&
1525 inp->inp_lport == lport) {
1530 prison_equal_ip4(cred->cr_prison,
1531 inp->inp_cred->cr_prison))
1540 struct inpcbporthead *porthash;
1541 struct inpcbport *phd;
1542 struct inpcb *match = NULL;
1544 * Best fit PCB lookup.
1546 * First see if this local port is in use by looking on the
1549 porthash = &pcbinfo->ipi_porthashbase[INP_PCBPORTHASH(lport,
1550 pcbinfo->ipi_porthashmask)];
1551 LIST_FOREACH(phd, porthash, phd_hash) {
1552 if (phd->phd_port == lport)
1557 * Port is in use by one or more PCBs. Look for best
1560 LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) {
1563 !prison_equal_ip4(inp->inp_cred->cr_prison,
1567 /* XXX inp locking */
1568 if ((inp->inp_vflag & INP_IPV4) == 0)
1571 * We never select the PCB that has
1572 * INP_IPV6 flag and is bound to :: if
1573 * we have another PCB which is bound
1574 * to 0.0.0.0. If a PCB has the
1575 * INP_IPV6 flag, then we set its cost
1576 * higher than IPv4 only PCBs.
1578 * Note that the case only happens
1579 * when a socket is bound to ::, under
1580 * the condition that the use of the
1581 * mapped address is allowed.
1583 if ((inp->inp_vflag & INP_IPV6) != 0)
1584 wildcard += INP_LOOKUP_MAPPED_PCB_COST;
1586 if (inp->inp_faddr.s_addr != INADDR_ANY)
1588 if (inp->inp_laddr.s_addr != INADDR_ANY) {
1589 if (laddr.s_addr == INADDR_ANY)
1591 else if (inp->inp_laddr.s_addr != laddr.s_addr)
1594 if (laddr.s_addr != INADDR_ANY)
1597 if (wildcard < matchwild) {
1599 matchwild = wildcard;
1608 #undef INP_LOOKUP_MAPPED_PCB_COST
1612 * Lookup PCB in hash list, using pcbgroup tables.
1614 static struct inpcb *
1615 in_pcblookup_group(struct inpcbinfo *pcbinfo, struct inpcbgroup *pcbgroup,
1616 struct in_addr faddr, u_int fport_arg, struct in_addr laddr,
1617 u_int lport_arg, int lookupflags, struct ifnet *ifp)
1619 struct inpcbhead *head;
1620 struct inpcb *inp, *tmpinp;
1621 u_short fport = fport_arg, lport = lport_arg;
1625 * First look for an exact match.
1628 INP_GROUP_LOCK(pcbgroup);
1629 head = &pcbgroup->ipg_hashbase[INP_PCBHASH(faddr.s_addr, lport, fport,
1630 pcbgroup->ipg_hashmask)];
1631 LIST_FOREACH(inp, head, inp_pcbgrouphash) {
1633 /* XXX inp locking */
1634 if ((inp->inp_vflag & INP_IPV4) == 0)
1637 if (inp->inp_faddr.s_addr == faddr.s_addr &&
1638 inp->inp_laddr.s_addr == laddr.s_addr &&
1639 inp->inp_fport == fport &&
1640 inp->inp_lport == lport) {
1642 * XXX We should be able to directly return
1643 * the inp here, without any checks.
1644 * Well unless both bound with SO_REUSEPORT?
1646 if (prison_flag(inp->inp_cred, PR_IP4))
1652 if (tmpinp != NULL) {
1659 * For incoming connections, we may wish to do a wildcard
1660 * match for an RSS-local socket.
1662 if ((lookupflags & INPLOOKUP_WILDCARD) != 0) {
1663 struct inpcb *local_wild = NULL, *local_exact = NULL;
1665 struct inpcb *local_wild_mapped = NULL;
1667 struct inpcb *jail_wild = NULL;
1668 struct inpcbhead *head;
1672 * Order of socket selection - we always prefer jails.
1673 * 1. jailed, non-wild.
1675 * 3. non-jailed, non-wild.
1676 * 4. non-jailed, wild.
1679 head = &pcbgroup->ipg_hashbase[INP_PCBHASH(INADDR_ANY,
1680 lport, 0, pcbgroup->ipg_hashmask)];
1681 LIST_FOREACH(inp, head, inp_pcbgrouphash) {
1683 /* XXX inp locking */
1684 if ((inp->inp_vflag & INP_IPV4) == 0)
1687 if (inp->inp_faddr.s_addr != INADDR_ANY ||
1688 inp->inp_lport != lport)
1691 injail = prison_flag(inp->inp_cred, PR_IP4);
1693 if (prison_check_ip4(inp->inp_cred,
1697 if (local_exact != NULL)
1701 if (inp->inp_laddr.s_addr == laddr.s_addr) {
1706 } else if (inp->inp_laddr.s_addr == INADDR_ANY) {
1708 /* XXX inp locking, NULL check */
1709 if (inp->inp_vflag & INP_IPV6PROTO)
1710 local_wild_mapped = inp;
1718 } /* LIST_FOREACH */
1727 inp = local_wild_mapped;
1735 * Then look for a wildcard match, if requested.
1737 if ((lookupflags & INPLOOKUP_WILDCARD) != 0) {
1738 struct inpcb *local_wild = NULL, *local_exact = NULL;
1740 struct inpcb *local_wild_mapped = NULL;
1742 struct inpcb *jail_wild = NULL;
1743 struct inpcbhead *head;
1747 * Order of socket selection - we always prefer jails.
1748 * 1. jailed, non-wild.
1750 * 3. non-jailed, non-wild.
1751 * 4. non-jailed, wild.
1753 head = &pcbinfo->ipi_wildbase[INP_PCBHASH(INADDR_ANY, lport,
1754 0, pcbinfo->ipi_wildmask)];
1755 LIST_FOREACH(inp, head, inp_pcbgroup_wild) {
1757 /* XXX inp locking */
1758 if ((inp->inp_vflag & INP_IPV4) == 0)
1761 if (inp->inp_faddr.s_addr != INADDR_ANY ||
1762 inp->inp_lport != lport)
1765 injail = prison_flag(inp->inp_cred, PR_IP4);
1767 if (prison_check_ip4(inp->inp_cred,
1771 if (local_exact != NULL)
1775 if (inp->inp_laddr.s_addr == laddr.s_addr) {
1780 } else if (inp->inp_laddr.s_addr == INADDR_ANY) {
1782 /* XXX inp locking, NULL check */
1783 if (inp->inp_vflag & INP_IPV6PROTO)
1784 local_wild_mapped = inp;
1792 } /* LIST_FOREACH */
1800 inp = local_wild_mapped;
1804 } /* if (lookupflags & INPLOOKUP_WILDCARD) */
1805 INP_GROUP_UNLOCK(pcbgroup);
1809 if (lookupflags & INPLOOKUP_WLOCKPCB)
1810 locked = INP_TRY_WLOCK(inp);
1811 else if (lookupflags & INPLOOKUP_RLOCKPCB)
1812 locked = INP_TRY_RLOCK(inp);
1814 panic("%s: locking bug", __func__);
1817 INP_GROUP_UNLOCK(pcbgroup);
1819 if (lookupflags & INPLOOKUP_WLOCKPCB) {
1821 if (in_pcbrele_wlocked(inp))
1825 if (in_pcbrele_rlocked(inp))
1830 if (lookupflags & INPLOOKUP_WLOCKPCB)
1831 INP_WLOCK_ASSERT(inp);
1833 INP_RLOCK_ASSERT(inp);
1837 #endif /* PCBGROUP */
1840 * Lookup PCB in hash list, using pcbinfo tables. This variation assumes
1841 * that the caller has locked the hash list, and will not perform any further
1842 * locking or reference operations on either the hash list or the connection.
1844 static struct inpcb *
1845 in_pcblookup_hash_locked(struct inpcbinfo *pcbinfo, struct in_addr faddr,
1846 u_int fport_arg, struct in_addr laddr, u_int lport_arg, int lookupflags,
1849 struct inpcbhead *head;
1850 struct inpcb *inp, *tmpinp;
1851 u_short fport = fport_arg, lport = lport_arg;
1853 KASSERT((lookupflags & ~(INPLOOKUP_WILDCARD)) == 0,
1854 ("%s: invalid lookup flags %d", __func__, lookupflags));
1856 INP_HASH_LOCK_ASSERT(pcbinfo);
1859 * First look for an exact match.
1862 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(faddr.s_addr, lport, fport,
1863 pcbinfo->ipi_hashmask)];
1864 LIST_FOREACH(inp, head, inp_hash) {
1866 /* XXX inp locking */
1867 if ((inp->inp_vflag & INP_IPV4) == 0)
1870 if (inp->inp_faddr.s_addr == faddr.s_addr &&
1871 inp->inp_laddr.s_addr == laddr.s_addr &&
1872 inp->inp_fport == fport &&
1873 inp->inp_lport == lport) {
1875 * XXX We should be able to directly return
1876 * the inp here, without any checks.
1877 * Well unless both bound with SO_REUSEPORT?
1879 if (prison_flag(inp->inp_cred, PR_IP4))
1889 * Then look for a wildcard match, if requested.
1891 if ((lookupflags & INPLOOKUP_WILDCARD) != 0) {
1892 struct inpcb *local_wild = NULL, *local_exact = NULL;
1894 struct inpcb *local_wild_mapped = NULL;
1896 struct inpcb *jail_wild = NULL;
1900 * Order of socket selection - we always prefer jails.
1901 * 1. jailed, non-wild.
1903 * 3. non-jailed, non-wild.
1904 * 4. non-jailed, wild.
1907 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport,
1908 0, pcbinfo->ipi_hashmask)];
1909 LIST_FOREACH(inp, head, inp_hash) {
1911 /* XXX inp locking */
1912 if ((inp->inp_vflag & INP_IPV4) == 0)
1915 if (inp->inp_faddr.s_addr != INADDR_ANY ||
1916 inp->inp_lport != lport)
1919 injail = prison_flag(inp->inp_cred, PR_IP4);
1921 if (prison_check_ip4(inp->inp_cred,
1925 if (local_exact != NULL)
1929 if (inp->inp_laddr.s_addr == laddr.s_addr) {
1934 } else if (inp->inp_laddr.s_addr == INADDR_ANY) {
1936 /* XXX inp locking, NULL check */
1937 if (inp->inp_vflag & INP_IPV6PROTO)
1938 local_wild_mapped = inp;
1946 } /* LIST_FOREACH */
1947 if (jail_wild != NULL)
1949 if (local_exact != NULL)
1950 return (local_exact);
1951 if (local_wild != NULL)
1952 return (local_wild);
1954 if (local_wild_mapped != NULL)
1955 return (local_wild_mapped);
1957 } /* if ((lookupflags & INPLOOKUP_WILDCARD) != 0) */
1963 * Lookup PCB in hash list, using pcbinfo tables. This variation locks the
1964 * hash list lock, and will return the inpcb locked (i.e., requires
1965 * INPLOOKUP_LOCKPCB).
1967 static struct inpcb *
1968 in_pcblookup_hash(struct inpcbinfo *pcbinfo, struct in_addr faddr,
1969 u_int fport, struct in_addr laddr, u_int lport, int lookupflags,
1975 INP_HASH_RLOCK(pcbinfo);
1976 inp = in_pcblookup_hash_locked(pcbinfo, faddr, fport, laddr, lport,
1977 (lookupflags & ~(INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)), ifp);
1979 if (lookupflags & INPLOOKUP_WLOCKPCB)
1980 locked = INP_TRY_WLOCK(inp);
1981 else if (lookupflags & INPLOOKUP_RLOCKPCB)
1982 locked = INP_TRY_RLOCK(inp);
1984 panic("%s: locking bug", __func__);
1987 INP_HASH_RUNLOCK(pcbinfo);
1989 if (lookupflags & INPLOOKUP_WLOCKPCB) {
1991 if (in_pcbrele_wlocked(inp))
1995 if (in_pcbrele_rlocked(inp))
2000 if (lookupflags & INPLOOKUP_WLOCKPCB)
2001 INP_WLOCK_ASSERT(inp);
2003 INP_RLOCK_ASSERT(inp);
2006 INP_HASH_RUNLOCK(pcbinfo);
2011 * Public inpcb lookup routines, accepting a 4-tuple, and optionally, an mbuf
2012 * from which a pre-calculated hash value may be extracted.
2014 * Possibly more of this logic should be in in_pcbgroup.c.
2017 in_pcblookup(struct inpcbinfo *pcbinfo, struct in_addr faddr, u_int fport,
2018 struct in_addr laddr, u_int lport, int lookupflags, struct ifnet *ifp)
2020 #if defined(PCBGROUP) && !defined(RSS)
2021 struct inpcbgroup *pcbgroup;
2024 KASSERT((lookupflags & ~INPLOOKUP_MASK) == 0,
2025 ("%s: invalid lookup flags %d", __func__, lookupflags));
2026 KASSERT((lookupflags & (INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)) != 0,
2027 ("%s: LOCKPCB not set", __func__));
2030 * When not using RSS, use connection groups in preference to the
2031 * reservation table when looking up 4-tuples. When using RSS, just
2032 * use the reservation table, due to the cost of the Toeplitz hash
2035 * XXXRW: This policy belongs in the pcbgroup code, as in principle
2036 * we could be doing RSS with a non-Toeplitz hash that is affordable
2039 #if defined(PCBGROUP) && !defined(RSS)
2040 if (in_pcbgroup_enabled(pcbinfo)) {
2041 pcbgroup = in_pcbgroup_bytuple(pcbinfo, laddr, lport, faddr,
2043 return (in_pcblookup_group(pcbinfo, pcbgroup, faddr, fport,
2044 laddr, lport, lookupflags, ifp));
2047 return (in_pcblookup_hash(pcbinfo, faddr, fport, laddr, lport,
2052 in_pcblookup_mbuf(struct inpcbinfo *pcbinfo, struct in_addr faddr,
2053 u_int fport, struct in_addr laddr, u_int lport, int lookupflags,
2054 struct ifnet *ifp, struct mbuf *m)
2057 struct inpcbgroup *pcbgroup;
2060 KASSERT((lookupflags & ~INPLOOKUP_MASK) == 0,
2061 ("%s: invalid lookup flags %d", __func__, lookupflags));
2062 KASSERT((lookupflags & (INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)) != 0,
2063 ("%s: LOCKPCB not set", __func__));
2067 * If we can use a hardware-generated hash to look up the connection
2068 * group, use that connection group to find the inpcb. Otherwise
2069 * fall back on a software hash -- or the reservation table if we're
2072 * XXXRW: As above, that policy belongs in the pcbgroup code.
2074 if (in_pcbgroup_enabled(pcbinfo) &&
2075 !(M_HASHTYPE_TEST(m, M_HASHTYPE_NONE))) {
2076 pcbgroup = in_pcbgroup_byhash(pcbinfo, M_HASHTYPE_GET(m),
2077 m->m_pkthdr.flowid);
2078 if (pcbgroup != NULL)
2079 return (in_pcblookup_group(pcbinfo, pcbgroup, faddr,
2080 fport, laddr, lport, lookupflags, ifp));
2082 pcbgroup = in_pcbgroup_bytuple(pcbinfo, laddr, lport, faddr,
2084 return (in_pcblookup_group(pcbinfo, pcbgroup, faddr, fport,
2085 laddr, lport, lookupflags, ifp));
2089 return (in_pcblookup_hash(pcbinfo, faddr, fport, laddr, lport,
2095 * Insert PCB onto various hash lists.
2098 in_pcbinshash_internal(struct inpcb *inp, int do_pcbgroup_update)
2100 struct inpcbhead *pcbhash;
2101 struct inpcbporthead *pcbporthash;
2102 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
2103 struct inpcbport *phd;
2104 u_int32_t hashkey_faddr;
2106 INP_WLOCK_ASSERT(inp);
2107 INP_HASH_WLOCK_ASSERT(pcbinfo);
2109 KASSERT((inp->inp_flags & INP_INHASHLIST) == 0,
2110 ("in_pcbinshash: INP_INHASHLIST"));
2113 if (inp->inp_vflag & INP_IPV6)
2114 hashkey_faddr = INP6_PCBHASHKEY(&inp->in6p_faddr);
2117 hashkey_faddr = inp->inp_faddr.s_addr;
2119 pcbhash = &pcbinfo->ipi_hashbase[INP_PCBHASH(hashkey_faddr,
2120 inp->inp_lport, inp->inp_fport, pcbinfo->ipi_hashmask)];
2122 pcbporthash = &pcbinfo->ipi_porthashbase[
2123 INP_PCBPORTHASH(inp->inp_lport, pcbinfo->ipi_porthashmask)];
2126 * Go through port list and look for a head for this lport.
2128 LIST_FOREACH(phd, pcbporthash, phd_hash) {
2129 if (phd->phd_port == inp->inp_lport)
2133 * If none exists, malloc one and tack it on.
2136 phd = malloc(sizeof(struct inpcbport), M_PCB, M_NOWAIT);
2138 return (ENOBUFS); /* XXX */
2140 phd->phd_port = inp->inp_lport;
2141 LIST_INIT(&phd->phd_pcblist);
2142 LIST_INSERT_HEAD(pcbporthash, phd, phd_hash);
2145 LIST_INSERT_HEAD(&phd->phd_pcblist, inp, inp_portlist);
2146 LIST_INSERT_HEAD(pcbhash, inp, inp_hash);
2147 inp->inp_flags |= INP_INHASHLIST;
2149 if (do_pcbgroup_update)
2150 in_pcbgroup_update(inp);
2156 * For now, there are two public interfaces to insert an inpcb into the hash
2157 * lists -- one that does update pcbgroups, and one that doesn't. The latter
2158 * is used only in the TCP syncache, where in_pcbinshash is called before the
2159 * full 4-tuple is set for the inpcb, and we don't want to install in the
2160 * pcbgroup until later.
2162 * XXXRW: This seems like a misfeature. in_pcbinshash should always update
2163 * connection groups, and partially initialised inpcbs should not be exposed
2164 * to either reservation hash tables or pcbgroups.
2167 in_pcbinshash(struct inpcb *inp)
2170 return (in_pcbinshash_internal(inp, 1));
2174 in_pcbinshash_nopcbgroup(struct inpcb *inp)
2177 return (in_pcbinshash_internal(inp, 0));
2181 * Move PCB to the proper hash bucket when { faddr, fport } have been
2182 * changed. NOTE: This does not handle the case of the lport changing (the
2183 * hashed port list would have to be updated as well), so the lport must
2184 * not change after in_pcbinshash() has been called.
2187 in_pcbrehash_mbuf(struct inpcb *inp, struct mbuf *m)
2189 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
2190 struct inpcbhead *head;
2191 u_int32_t hashkey_faddr;
2193 INP_WLOCK_ASSERT(inp);
2194 INP_HASH_WLOCK_ASSERT(pcbinfo);
2196 KASSERT(inp->inp_flags & INP_INHASHLIST,
2197 ("in_pcbrehash: !INP_INHASHLIST"));
2200 if (inp->inp_vflag & INP_IPV6)
2201 hashkey_faddr = INP6_PCBHASHKEY(&inp->in6p_faddr);
2204 hashkey_faddr = inp->inp_faddr.s_addr;
2206 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(hashkey_faddr,
2207 inp->inp_lport, inp->inp_fport, pcbinfo->ipi_hashmask)];
2209 LIST_REMOVE(inp, inp_hash);
2210 LIST_INSERT_HEAD(head, inp, inp_hash);
2214 in_pcbgroup_update_mbuf(inp, m);
2216 in_pcbgroup_update(inp);
2221 in_pcbrehash(struct inpcb *inp)
2224 in_pcbrehash_mbuf(inp, NULL);
2228 * Remove PCB from various lists.
2231 in_pcbremlists(struct inpcb *inp)
2233 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
2236 if (pcbinfo == &V_tcbinfo) {
2237 INP_INFO_RLOCK_ASSERT(pcbinfo);
2239 INP_INFO_WLOCK_ASSERT(pcbinfo);
2243 INP_WLOCK_ASSERT(inp);
2244 INP_LIST_WLOCK_ASSERT(pcbinfo);
2246 inp->inp_gencnt = ++pcbinfo->ipi_gencnt;
2247 if (inp->inp_flags & INP_INHASHLIST) {
2248 struct inpcbport *phd = inp->inp_phd;
2250 INP_HASH_WLOCK(pcbinfo);
2251 LIST_REMOVE(inp, inp_hash);
2252 LIST_REMOVE(inp, inp_portlist);
2253 if (LIST_FIRST(&phd->phd_pcblist) == NULL) {
2254 LIST_REMOVE(phd, phd_hash);
2257 INP_HASH_WUNLOCK(pcbinfo);
2258 inp->inp_flags &= ~INP_INHASHLIST;
2260 LIST_REMOVE(inp, inp_list);
2261 pcbinfo->ipi_count--;
2263 in_pcbgroup_remove(inp);
2268 * Check for alternatives when higher level complains
2269 * about service problems. For now, invalidate cached
2270 * routing information. If the route was created dynamically
2271 * (by a redirect), time to try a default gateway again.
2274 in_losing(struct inpcb *inp)
2277 if (inp->inp_route.ro_rt) {
2278 RTFREE(inp->inp_route.ro_rt);
2279 inp->inp_route.ro_rt = (struct rtentry *)NULL;
2281 if (inp->inp_route.ro_lle)
2282 LLE_FREE(inp->inp_route.ro_lle); /* zeros ro_lle */
2287 * A set label operation has occurred at the socket layer, propagate the
2288 * label change into the in_pcb for the socket.
2291 in_pcbsosetlabel(struct socket *so)
2296 inp = sotoinpcb(so);
2297 KASSERT(inp != NULL, ("in_pcbsosetlabel: so->so_pcb == NULL"));
2301 mac_inpcb_sosetlabel(so, inp);
2308 * ipport_tick runs once per second, determining if random port allocation
2309 * should be continued. If more than ipport_randomcps ports have been
2310 * allocated in the last second, then we return to sequential port
2311 * allocation. We return to random allocation only once we drop below
2312 * ipport_randomcps for at least ipport_randomtime seconds.
2315 ipport_tick(void *xtp)
2317 VNET_ITERATOR_DECL(vnet_iter);
2319 VNET_LIST_RLOCK_NOSLEEP();
2320 VNET_FOREACH(vnet_iter) {
2321 CURVNET_SET(vnet_iter); /* XXX appease INVARIANTS here */
2322 if (V_ipport_tcpallocs <=
2323 V_ipport_tcplastcount + V_ipport_randomcps) {
2324 if (V_ipport_stoprandom > 0)
2325 V_ipport_stoprandom--;
2327 V_ipport_stoprandom = V_ipport_randomtime;
2328 V_ipport_tcplastcount = V_ipport_tcpallocs;
2331 VNET_LIST_RUNLOCK_NOSLEEP();
2332 callout_reset(&ipport_tick_callout, hz, ipport_tick, NULL);
2339 callout_stop(&ipport_tick_callout);
2343 * The ipport_callout should start running at about the time we attach the
2344 * inet or inet6 domains.
2347 ipport_tick_init(const void *unused __unused)
2350 /* Start ipport_tick. */
2351 callout_init(&ipport_tick_callout, 1);
2352 callout_reset(&ipport_tick_callout, 1, ipport_tick, NULL);
2353 EVENTHANDLER_REGISTER(shutdown_pre_sync, ip_fini, NULL,
2354 SHUTDOWN_PRI_DEFAULT);
2356 SYSINIT(ipport_tick_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE,
2357 ipport_tick_init, NULL);
2360 inp_wlock(struct inpcb *inp)
2367 inp_wunlock(struct inpcb *inp)
2374 inp_rlock(struct inpcb *inp)
2381 inp_runlock(struct inpcb *inp)
2389 inp_lock_assert(struct inpcb *inp)
2392 INP_WLOCK_ASSERT(inp);
2396 inp_unlock_assert(struct inpcb *inp)
2399 INP_UNLOCK_ASSERT(inp);
2404 inp_apply_all(void (*func)(struct inpcb *, void *), void *arg)
2408 INP_INFO_WLOCK(&V_tcbinfo);
2409 LIST_FOREACH(inp, V_tcbinfo.ipi_listhead, inp_list) {
2414 INP_INFO_WUNLOCK(&V_tcbinfo);
2418 inp_inpcbtosocket(struct inpcb *inp)
2421 INP_WLOCK_ASSERT(inp);
2422 return (inp->inp_socket);
2426 inp_inpcbtotcpcb(struct inpcb *inp)
2429 INP_WLOCK_ASSERT(inp);
2430 return ((struct tcpcb *)inp->inp_ppcb);
2434 inp_ip_tos_get(const struct inpcb *inp)
2437 return (inp->inp_ip_tos);
2441 inp_ip_tos_set(struct inpcb *inp, int val)
2444 inp->inp_ip_tos = val;
2448 inp_4tuple_get(struct inpcb *inp, uint32_t *laddr, uint16_t *lp,
2449 uint32_t *faddr, uint16_t *fp)
2452 INP_LOCK_ASSERT(inp);
2453 *laddr = inp->inp_laddr.s_addr;
2454 *faddr = inp->inp_faddr.s_addr;
2455 *lp = inp->inp_lport;
2456 *fp = inp->inp_fport;
2460 so_sotoinpcb(struct socket *so)
2463 return (sotoinpcb(so));
2467 so_sototcpcb(struct socket *so)
2470 return (sototcpcb(so));
2475 db_print_indent(int indent)
2479 for (i = 0; i < indent; i++)
2484 db_print_inconninfo(struct in_conninfo *inc, const char *name, int indent)
2486 char faddr_str[48], laddr_str[48];
2488 db_print_indent(indent);
2489 db_printf("%s at %p\n", name, inc);
2494 if (inc->inc_flags & INC_ISIPV6) {
2496 ip6_sprintf(laddr_str, &inc->inc6_laddr);
2497 ip6_sprintf(faddr_str, &inc->inc6_faddr);
2502 inet_ntoa_r(inc->inc_laddr, laddr_str);
2503 inet_ntoa_r(inc->inc_faddr, faddr_str);
2505 db_print_indent(indent);
2506 db_printf("inc_laddr %s inc_lport %u\n", laddr_str,
2507 ntohs(inc->inc_lport));
2508 db_print_indent(indent);
2509 db_printf("inc_faddr %s inc_fport %u\n", faddr_str,
2510 ntohs(inc->inc_fport));
2514 db_print_inpflags(int inp_flags)
2519 if (inp_flags & INP_RECVOPTS) {
2520 db_printf("%sINP_RECVOPTS", comma ? ", " : "");
2523 if (inp_flags & INP_RECVRETOPTS) {
2524 db_printf("%sINP_RECVRETOPTS", comma ? ", " : "");
2527 if (inp_flags & INP_RECVDSTADDR) {
2528 db_printf("%sINP_RECVDSTADDR", comma ? ", " : "");
2531 if (inp_flags & INP_HDRINCL) {
2532 db_printf("%sINP_HDRINCL", comma ? ", " : "");
2535 if (inp_flags & INP_HIGHPORT) {
2536 db_printf("%sINP_HIGHPORT", comma ? ", " : "");
2539 if (inp_flags & INP_LOWPORT) {
2540 db_printf("%sINP_LOWPORT", comma ? ", " : "");
2543 if (inp_flags & INP_ANONPORT) {
2544 db_printf("%sINP_ANONPORT", comma ? ", " : "");
2547 if (inp_flags & INP_RECVIF) {
2548 db_printf("%sINP_RECVIF", comma ? ", " : "");
2551 if (inp_flags & INP_MTUDISC) {
2552 db_printf("%sINP_MTUDISC", comma ? ", " : "");
2555 if (inp_flags & INP_RECVTTL) {
2556 db_printf("%sINP_RECVTTL", comma ? ", " : "");
2559 if (inp_flags & INP_DONTFRAG) {
2560 db_printf("%sINP_DONTFRAG", comma ? ", " : "");
2563 if (inp_flags & INP_RECVTOS) {
2564 db_printf("%sINP_RECVTOS", comma ? ", " : "");
2567 if (inp_flags & IN6P_IPV6_V6ONLY) {
2568 db_printf("%sIN6P_IPV6_V6ONLY", comma ? ", " : "");
2571 if (inp_flags & IN6P_PKTINFO) {
2572 db_printf("%sIN6P_PKTINFO", comma ? ", " : "");
2575 if (inp_flags & IN6P_HOPLIMIT) {
2576 db_printf("%sIN6P_HOPLIMIT", comma ? ", " : "");
2579 if (inp_flags & IN6P_HOPOPTS) {
2580 db_printf("%sIN6P_HOPOPTS", comma ? ", " : "");
2583 if (inp_flags & IN6P_DSTOPTS) {
2584 db_printf("%sIN6P_DSTOPTS", comma ? ", " : "");
2587 if (inp_flags & IN6P_RTHDR) {
2588 db_printf("%sIN6P_RTHDR", comma ? ", " : "");
2591 if (inp_flags & IN6P_RTHDRDSTOPTS) {
2592 db_printf("%sIN6P_RTHDRDSTOPTS", comma ? ", " : "");
2595 if (inp_flags & IN6P_TCLASS) {
2596 db_printf("%sIN6P_TCLASS", comma ? ", " : "");
2599 if (inp_flags & IN6P_AUTOFLOWLABEL) {
2600 db_printf("%sIN6P_AUTOFLOWLABEL", comma ? ", " : "");
2603 if (inp_flags & INP_TIMEWAIT) {
2604 db_printf("%sINP_TIMEWAIT", comma ? ", " : "");
2607 if (inp_flags & INP_ONESBCAST) {
2608 db_printf("%sINP_ONESBCAST", comma ? ", " : "");
2611 if (inp_flags & INP_DROPPED) {
2612 db_printf("%sINP_DROPPED", comma ? ", " : "");
2615 if (inp_flags & INP_SOCKREF) {
2616 db_printf("%sINP_SOCKREF", comma ? ", " : "");
2619 if (inp_flags & IN6P_RFC2292) {
2620 db_printf("%sIN6P_RFC2292", comma ? ", " : "");
2623 if (inp_flags & IN6P_MTU) {
2624 db_printf("IN6P_MTU%s", comma ? ", " : "");
2630 db_print_inpvflag(u_char inp_vflag)
2635 if (inp_vflag & INP_IPV4) {
2636 db_printf("%sINP_IPV4", comma ? ", " : "");
2639 if (inp_vflag & INP_IPV6) {
2640 db_printf("%sINP_IPV6", comma ? ", " : "");
2643 if (inp_vflag & INP_IPV6PROTO) {
2644 db_printf("%sINP_IPV6PROTO", comma ? ", " : "");
2650 db_print_inpcb(struct inpcb *inp, const char *name, int indent)
2653 db_print_indent(indent);
2654 db_printf("%s at %p\n", name, inp);
2658 db_print_indent(indent);
2659 db_printf("inp_flow: 0x%x\n", inp->inp_flow);
2661 db_print_inconninfo(&inp->inp_inc, "inp_conninfo", indent);
2663 db_print_indent(indent);
2664 db_printf("inp_ppcb: %p inp_pcbinfo: %p inp_socket: %p\n",
2665 inp->inp_ppcb, inp->inp_pcbinfo, inp->inp_socket);
2667 db_print_indent(indent);
2668 db_printf("inp_label: %p inp_flags: 0x%x (",
2669 inp->inp_label, inp->inp_flags);
2670 db_print_inpflags(inp->inp_flags);
2673 db_print_indent(indent);
2674 db_printf("inp_sp: %p inp_vflag: 0x%x (", inp->inp_sp,
2676 db_print_inpvflag(inp->inp_vflag);
2679 db_print_indent(indent);
2680 db_printf("inp_ip_ttl: %d inp_ip_p: %d inp_ip_minttl: %d\n",
2681 inp->inp_ip_ttl, inp->inp_ip_p, inp->inp_ip_minttl);
2683 db_print_indent(indent);
2685 if (inp->inp_vflag & INP_IPV6) {
2686 db_printf("in6p_options: %p in6p_outputopts: %p "
2687 "in6p_moptions: %p\n", inp->in6p_options,
2688 inp->in6p_outputopts, inp->in6p_moptions);
2689 db_printf("in6p_icmp6filt: %p in6p_cksum %d "
2690 "in6p_hops %u\n", inp->in6p_icmp6filt, inp->in6p_cksum,
2695 db_printf("inp_ip_tos: %d inp_ip_options: %p "
2696 "inp_ip_moptions: %p\n", inp->inp_ip_tos,
2697 inp->inp_options, inp->inp_moptions);
2700 db_print_indent(indent);
2701 db_printf("inp_phd: %p inp_gencnt: %ju\n", inp->inp_phd,
2702 (uintmax_t)inp->inp_gencnt);
2705 DB_SHOW_COMMAND(inpcb, db_show_inpcb)
2710 db_printf("usage: show inpcb <addr>\n");
2713 inp = (struct inpcb *)addr;
2715 db_print_inpcb(inp, "inpcb", 0);