2 * Copyright (c) 1982, 1986, 1991, 1993, 1995
3 * The Regents of the University of California.
4 * Copyright (c) 2007 Robert N. M. Watson
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 * 4. Neither the name of the University nor the names of its contributors
16 * may be used to endorse or promote products derived from this software
17 * without specific prior written permission.
19 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31 * @(#)in_pcb.c 8.4 (Berkeley) 5/24/95
36 #include "opt_ipsec.h"
37 #include "opt_inet6.h"
40 #include <sys/param.h>
41 #include <sys/systm.h>
42 #include <sys/malloc.h>
44 #include <sys/domain.h>
45 #include <sys/protosw.h>
46 #include <sys/socket.h>
47 #include <sys/socketvar.h>
51 #include <sys/kernel.h>
52 #include <sys/sysctl.h>
61 #include <net/if_types.h>
62 #include <net/route.h>
64 #include <netinet/in.h>
65 #include <netinet/in_pcb.h>
66 #include <netinet/in_var.h>
67 #include <netinet/ip_var.h>
68 #include <netinet/tcp_var.h>
69 #include <netinet/udp.h>
70 #include <netinet/udp_var.h>
72 #include <netinet/ip6.h>
73 #include <netinet6/ip6_var.h>
77 #include <netinet6/ipsec.h>
78 #include <netkey/key.h>
82 #if defined(IPSEC) || defined(IPSEC_ESP)
83 #error "Bad idea: don't compile with both IPSEC and FAST_IPSEC!"
86 #include <netipsec/ipsec.h>
87 #include <netipsec/key.h>
88 #endif /* FAST_IPSEC */
90 #include <security/mac/mac_framework.h>
93 * These configure the range of local port addresses assigned to
94 * "unspecified" outgoing connections/packets/whatever.
96 int ipport_lowfirstauto = IPPORT_RESERVED - 1; /* 1023 */
97 int ipport_lowlastauto = IPPORT_RESERVEDSTART; /* 600 */
98 int ipport_firstauto = IPPORT_HIFIRSTAUTO; /* 49152 */
99 int ipport_lastauto = IPPORT_HILASTAUTO; /* 65535 */
100 int ipport_hifirstauto = IPPORT_HIFIRSTAUTO; /* 49152 */
101 int ipport_hilastauto = IPPORT_HILASTAUTO; /* 65535 */
104 * Reserved ports accessible only to root. There are significant
105 * security considerations that must be accounted for when changing these,
106 * but the security benefits can be great. Please be careful.
108 int ipport_reservedhigh = IPPORT_RESERVED - 1; /* 1023 */
109 int ipport_reservedlow = 0;
111 /* Variables dealing with random ephemeral port allocation. */
112 int ipport_randomized = 1; /* user controlled via sysctl */
113 int ipport_randomcps = 10; /* user controlled via sysctl */
114 int ipport_randomtime = 45; /* user controlled via sysctl */
115 int ipport_stoprandom = 0; /* toggled by ipport_tick */
116 int ipport_tcpallocs;
117 int ipport_tcplastcount;
119 #define RANGECHK(var, min, max) \
120 if ((var) < (min)) { (var) = (min); } \
121 else if ((var) > (max)) { (var) = (max); }
124 sysctl_net_ipport_check(SYSCTL_HANDLER_ARGS)
128 error = sysctl_handle_int(oidp, oidp->oid_arg1, oidp->oid_arg2, req);
130 RANGECHK(ipport_lowfirstauto, 1, IPPORT_RESERVED - 1);
131 RANGECHK(ipport_lowlastauto, 1, IPPORT_RESERVED - 1);
132 RANGECHK(ipport_firstauto, IPPORT_RESERVED, IPPORT_MAX);
133 RANGECHK(ipport_lastauto, IPPORT_RESERVED, IPPORT_MAX);
134 RANGECHK(ipport_hifirstauto, IPPORT_RESERVED, IPPORT_MAX);
135 RANGECHK(ipport_hilastauto, IPPORT_RESERVED, IPPORT_MAX);
142 SYSCTL_NODE(_net_inet_ip, IPPROTO_IP, portrange, CTLFLAG_RW, 0, "IP Ports");
144 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, lowfirst, CTLTYPE_INT|CTLFLAG_RW,
145 &ipport_lowfirstauto, 0, &sysctl_net_ipport_check, "I", "");
146 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, lowlast, CTLTYPE_INT|CTLFLAG_RW,
147 &ipport_lowlastauto, 0, &sysctl_net_ipport_check, "I", "");
148 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, first, CTLTYPE_INT|CTLFLAG_RW,
149 &ipport_firstauto, 0, &sysctl_net_ipport_check, "I", "");
150 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, last, CTLTYPE_INT|CTLFLAG_RW,
151 &ipport_lastauto, 0, &sysctl_net_ipport_check, "I", "");
152 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, hifirst, CTLTYPE_INT|CTLFLAG_RW,
153 &ipport_hifirstauto, 0, &sysctl_net_ipport_check, "I", "");
154 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, hilast, CTLTYPE_INT|CTLFLAG_RW,
155 &ipport_hilastauto, 0, &sysctl_net_ipport_check, "I", "");
156 SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, reservedhigh,
157 CTLFLAG_RW|CTLFLAG_SECURE, &ipport_reservedhigh, 0, "");
158 SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, reservedlow,
159 CTLFLAG_RW|CTLFLAG_SECURE, &ipport_reservedlow, 0, "");
160 SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, randomized, CTLFLAG_RW,
161 &ipport_randomized, 0, "Enable random port allocation");
162 SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, randomcps, CTLFLAG_RW,
163 &ipport_randomcps, 0, "Maximum number of random port "
164 "allocations before switching to a sequental one");
165 SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, randomtime, CTLFLAG_RW,
166 &ipport_randomtime, 0, "Minimum time to keep sequental port "
167 "allocation before switching to a random one");
170 * in_pcb.c: manage the Protocol Control Blocks.
172 * NOTE: It is assumed that most of these functions will be called with
173 * the pcbinfo lock held, and often, the inpcb lock held, as these utility
174 * functions often modify hash chains or addresses in pcbs.
178 * Allocate a PCB and associate it with the socket.
179 * On success return with the PCB locked.
182 in_pcballoc(struct socket *so, struct inpcbinfo *pcbinfo)
187 INP_INFO_WLOCK_ASSERT(pcbinfo);
189 inp = uma_zalloc(pcbinfo->ipi_zone, M_NOWAIT);
192 bzero(inp, inp_zero_size);
193 inp->inp_pcbinfo = pcbinfo;
194 inp->inp_socket = so;
196 error = mac_init_inpcb(inp, M_NOWAIT);
200 mac_create_inpcb_from_socket(so, inp);
203 #if defined(IPSEC) || defined(FAST_IPSEC)
205 error = ipsec_init_policy(so, &inp->inp_sp);
207 error = ipsec_init_pcbpolicy(so, &inp->inp_sp);
213 if (INP_SOCKAF(so) == AF_INET6) {
214 inp->inp_vflag |= INP_IPV6PROTO;
216 inp->inp_flags |= IN6P_IPV6_V6ONLY;
219 LIST_INSERT_HEAD(pcbinfo->listhead, inp, inp_list);
220 pcbinfo->ipi_count++;
221 so->so_pcb = (caddr_t)inp;
223 if (ip6_auto_flowlabel)
224 inp->inp_flags |= IN6P_AUTOFLOWLABEL;
227 inp->inp_gencnt = ++pcbinfo->ipi_gencnt;
229 #if defined(IPSEC) || defined(FAST_IPSEC) || defined(MAC)
232 uma_zfree(pcbinfo->ipi_zone, inp);
238 in_pcbbind(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred)
242 INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
243 INP_LOCK_ASSERT(inp);
245 if (inp->inp_lport != 0 || inp->inp_laddr.s_addr != INADDR_ANY)
247 anonport = inp->inp_lport == 0 && (nam == NULL ||
248 ((struct sockaddr_in *)nam)->sin_port == 0);
249 error = in_pcbbind_setup(inp, nam, &inp->inp_laddr.s_addr,
250 &inp->inp_lport, cred);
253 if (in_pcbinshash(inp) != 0) {
254 inp->inp_laddr.s_addr = INADDR_ANY;
259 inp->inp_flags |= INP_ANONPORT;
264 * Set up a bind operation on a PCB, performing port allocation
265 * as required, but do not actually modify the PCB. Callers can
266 * either complete the bind by setting inp_laddr/inp_lport and
267 * calling in_pcbinshash(), or they can just use the resulting
268 * port and address to authorise the sending of a once-off packet.
270 * On error, the values of *laddrp and *lportp are not changed.
273 in_pcbbind_setup(struct inpcb *inp, struct sockaddr *nam, in_addr_t *laddrp,
274 u_short *lportp, struct ucred *cred)
276 struct socket *so = inp->inp_socket;
277 unsigned short *lastport;
278 struct sockaddr_in *sin;
279 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
280 struct in_addr laddr;
282 int wild = 0, reuseport = (so->so_options & SO_REUSEPORT);
283 int error, prison = 0;
286 INP_INFO_WLOCK_ASSERT(pcbinfo);
287 INP_LOCK_ASSERT(inp);
289 if (TAILQ_EMPTY(&in_ifaddrhead)) /* XXX broken! */
290 return (EADDRNOTAVAIL);
291 laddr.s_addr = *laddrp;
292 if (nam != NULL && laddr.s_addr != INADDR_ANY)
294 if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0)
295 wild = INPLOOKUP_WILDCARD;
297 sin = (struct sockaddr_in *)nam;
298 if (nam->sa_len != sizeof (*sin))
302 * We should check the family, but old programs
303 * incorrectly fail to initialize it.
305 if (sin->sin_family != AF_INET)
306 return (EAFNOSUPPORT);
308 if (sin->sin_addr.s_addr != INADDR_ANY)
309 if (prison_ip(cred, 0, &sin->sin_addr.s_addr))
311 if (sin->sin_port != *lportp) {
312 /* Don't allow the port to change. */
315 lport = sin->sin_port;
317 /* NB: lport is left as 0 if the port isn't being changed. */
318 if (IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) {
320 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
321 * allow complete duplication of binding if
322 * SO_REUSEPORT is set, or if SO_REUSEADDR is set
323 * and a multicast address is bound on both
324 * new and duplicated sockets.
326 if (so->so_options & SO_REUSEADDR)
327 reuseport = SO_REUSEADDR|SO_REUSEPORT;
328 } else if (sin->sin_addr.s_addr != INADDR_ANY) {
329 sin->sin_port = 0; /* yech... */
330 bzero(&sin->sin_zero, sizeof(sin->sin_zero));
331 if (ifa_ifwithaddr((struct sockaddr *)sin) == 0)
332 return (EADDRNOTAVAIL);
334 laddr = sin->sin_addr;
340 if (ntohs(lport) <= ipport_reservedhigh &&
341 ntohs(lport) >= ipport_reservedlow &&
342 priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT,
347 if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)) &&
348 priv_check_cred(so->so_cred,
349 PRIV_NETINET_REUSEPORT, SUSER_ALLOWJAIL) != 0) {
350 t = in_pcblookup_local(inp->inp_pcbinfo,
351 sin->sin_addr, lport,
352 prison ? 0 : INPLOOKUP_WILDCARD);
355 * This entire block sorely needs a rewrite.
358 ((t->inp_vflag & INP_TIMEWAIT) == 0) &&
359 (so->so_type != SOCK_STREAM ||
360 ntohl(t->inp_faddr.s_addr) == INADDR_ANY) &&
361 (ntohl(sin->sin_addr.s_addr) != INADDR_ANY ||
362 ntohl(t->inp_laddr.s_addr) != INADDR_ANY ||
363 (t->inp_socket->so_options &
364 SO_REUSEPORT) == 0) &&
365 (so->so_cred->cr_uid !=
366 t->inp_socket->so_cred->cr_uid))
369 if (prison && prison_ip(cred, 0, &sin->sin_addr.s_addr))
370 return (EADDRNOTAVAIL);
371 t = in_pcblookup_local(pcbinfo, sin->sin_addr,
372 lport, prison ? 0 : wild);
373 if (t && (t->inp_vflag & INP_TIMEWAIT)) {
375 * XXXRW: If an incpb has had its timewait
376 * state recycled, we treat the address as
377 * being in use (for now). This is better
378 * than a panic, but not desirable.
382 (reuseport & tw->tw_so_options) == 0)
385 (reuseport & t->inp_socket->so_options) == 0) {
387 if (ntohl(sin->sin_addr.s_addr) !=
389 ntohl(t->inp_laddr.s_addr) !=
392 INP_SOCKAF(t->inp_socket))
404 if (laddr.s_addr != INADDR_ANY)
405 if (prison_ip(cred, 0, &laddr.s_addr))
408 if (inp->inp_flags & INP_HIGHPORT) {
409 first = ipport_hifirstauto; /* sysctl */
410 last = ipport_hilastauto;
411 lastport = &pcbinfo->lasthi;
412 } else if (inp->inp_flags & INP_LOWPORT) {
413 error = priv_check_cred(cred,
414 PRIV_NETINET_RESERVEDPORT, SUSER_ALLOWJAIL);
417 first = ipport_lowfirstauto; /* 1023 */
418 last = ipport_lowlastauto; /* 600 */
419 lastport = &pcbinfo->lastlow;
421 first = ipport_firstauto; /* sysctl */
422 last = ipport_lastauto;
423 lastport = &pcbinfo->lastport;
426 * For UDP, use random port allocation as long as the user
427 * allows it. For TCP (and as of yet unknown) connections,
428 * use random port allocation only if the user allows it AND
429 * ipport_tick() allows it.
431 if (ipport_randomized &&
432 (!ipport_stoprandom || pcbinfo == &udbinfo))
437 * It makes no sense to do random port allocation if
438 * we have the only port available.
442 /* Make sure to not include UDP packets in the count. */
443 if (pcbinfo != &udbinfo)
446 * Simple check to ensure all ports are not used up causing
449 * We split the two cases (up and down) so that the direction
450 * is not being tested on each round of the loop.
458 (arc4random() % (first - last));
459 count = first - last;
462 if (count-- < 0) /* completely used? */
463 return (EADDRNOTAVAIL);
465 if (*lastport > first || *lastport < last)
467 lport = htons(*lastport);
468 } while (in_pcblookup_local(pcbinfo, laddr, lport,
476 (arc4random() % (last - first));
477 count = last - first;
480 if (count-- < 0) /* completely used? */
481 return (EADDRNOTAVAIL);
483 if (*lastport < first || *lastport > last)
485 lport = htons(*lastport);
486 } while (in_pcblookup_local(pcbinfo, laddr, lport,
490 if (prison_ip(cred, 0, &laddr.s_addr))
492 *laddrp = laddr.s_addr;
498 * Connect from a socket to a specified address.
499 * Both address and port must be specified in argument sin.
500 * If don't have a local address for this socket yet,
504 in_pcbconnect(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred)
506 u_short lport, fport;
507 in_addr_t laddr, faddr;
510 INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
511 INP_LOCK_ASSERT(inp);
513 lport = inp->inp_lport;
514 laddr = inp->inp_laddr.s_addr;
515 anonport = (lport == 0);
516 error = in_pcbconnect_setup(inp, nam, &laddr, &lport, &faddr, &fport,
521 /* Do the initial binding of the local address if required. */
522 if (inp->inp_laddr.s_addr == INADDR_ANY && inp->inp_lport == 0) {
523 inp->inp_lport = lport;
524 inp->inp_laddr.s_addr = laddr;
525 if (in_pcbinshash(inp) != 0) {
526 inp->inp_laddr.s_addr = INADDR_ANY;
532 /* Commit the remaining changes. */
533 inp->inp_lport = lport;
534 inp->inp_laddr.s_addr = laddr;
535 inp->inp_faddr.s_addr = faddr;
536 inp->inp_fport = fport;
539 if (inp->inp_socket->so_type == SOCK_STREAM)
540 ipsec_pcbconn(inp->inp_sp);
543 inp->inp_flags |= INP_ANONPORT;
548 * Set up for a connect from a socket to the specified address.
549 * On entry, *laddrp and *lportp should contain the current local
550 * address and port for the PCB; these are updated to the values
551 * that should be placed in inp_laddr and inp_lport to complete
554 * On success, *faddrp and *fportp will be set to the remote address
555 * and port. These are not updated in the error case.
557 * If the operation fails because the connection already exists,
558 * *oinpp will be set to the PCB of that connection so that the
559 * caller can decide to override it. In all other cases, *oinpp
563 in_pcbconnect_setup(struct inpcb *inp, struct sockaddr *nam,
564 in_addr_t *laddrp, u_short *lportp, in_addr_t *faddrp, u_short *fportp,
565 struct inpcb **oinpp, struct ucred *cred)
567 struct sockaddr_in *sin = (struct sockaddr_in *)nam;
568 struct in_ifaddr *ia;
569 struct sockaddr_in sa;
570 struct ucred *socred;
572 struct in_addr laddr, faddr;
573 u_short lport, fport;
576 INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
577 INP_LOCK_ASSERT(inp);
581 if (nam->sa_len != sizeof (*sin))
583 if (sin->sin_family != AF_INET)
584 return (EAFNOSUPPORT);
585 if (sin->sin_port == 0)
586 return (EADDRNOTAVAIL);
587 laddr.s_addr = *laddrp;
589 faddr = sin->sin_addr;
590 fport = sin->sin_port;
591 socred = inp->inp_socket->so_cred;
592 if (laddr.s_addr == INADDR_ANY && jailed(socred)) {
593 bzero(&sa, sizeof(sa));
594 sa.sin_addr.s_addr = htonl(prison_getip(socred));
595 sa.sin_len = sizeof(sa);
596 sa.sin_family = AF_INET;
597 error = in_pcbbind_setup(inp, (struct sockaddr *)&sa,
598 &laddr.s_addr, &lport, cred);
602 if (!TAILQ_EMPTY(&in_ifaddrhead)) {
604 * If the destination address is INADDR_ANY,
605 * use the primary local address.
606 * If the supplied address is INADDR_BROADCAST,
607 * and the primary interface supports broadcast,
608 * choose the broadcast address for that interface.
610 if (faddr.s_addr == INADDR_ANY)
611 faddr = IA_SIN(TAILQ_FIRST(&in_ifaddrhead))->sin_addr;
612 else if (faddr.s_addr == (u_long)INADDR_BROADCAST &&
613 (TAILQ_FIRST(&in_ifaddrhead)->ia_ifp->if_flags &
615 faddr = satosin(&TAILQ_FIRST(
616 &in_ifaddrhead)->ia_broadaddr)->sin_addr;
618 if (laddr.s_addr == INADDR_ANY) {
619 ia = (struct in_ifaddr *)0;
621 * If route is known our src addr is taken from the i/f,
624 * Find out route to destination
626 if ((inp->inp_socket->so_options & SO_DONTROUTE) == 0)
627 ia = ip_rtaddr(faddr);
629 * If we found a route, use the address corresponding to
630 * the outgoing interface.
632 * Otherwise assume faddr is reachable on a directly connected
633 * network and try to find a corresponding interface to take
634 * the source address from.
637 bzero(&sa, sizeof(sa));
639 sa.sin_len = sizeof(sa);
640 sa.sin_family = AF_INET;
642 ia = ifatoia(ifa_ifwithdstaddr(sintosa(&sa)));
644 ia = ifatoia(ifa_ifwithnet(sintosa(&sa)));
646 return (ENETUNREACH);
649 * If the destination address is multicast and an outgoing
650 * interface has been set as a multicast option, use the
651 * address of that interface as our source address.
653 if (IN_MULTICAST(ntohl(faddr.s_addr)) &&
654 inp->inp_moptions != NULL) {
655 struct ip_moptions *imo;
658 imo = inp->inp_moptions;
659 if (imo->imo_multicast_ifp != NULL) {
660 ifp = imo->imo_multicast_ifp;
661 TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link)
662 if (ia->ia_ifp == ifp)
665 return (EADDRNOTAVAIL);
668 laddr = ia->ia_addr.sin_addr;
671 oinp = in_pcblookup_hash(inp->inp_pcbinfo, faddr, fport, laddr, lport,
679 error = in_pcbbind_setup(inp, NULL, &laddr.s_addr, &lport,
684 *laddrp = laddr.s_addr;
686 *faddrp = faddr.s_addr;
692 in_pcbdisconnect(struct inpcb *inp)
695 INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
696 INP_LOCK_ASSERT(inp);
698 inp->inp_faddr.s_addr = INADDR_ANY;
702 ipsec_pcbdisconn(inp->inp_sp);
707 * In the old world order, in_pcbdetach() served two functions: to detach the
708 * pcb from the socket/potentially free the socket, and to free the pcb
709 * itself. In the new world order, the protocol code is responsible for
710 * managing the relationship with the socket, and this code simply frees the
714 in_pcbdetach(struct inpcb *inp)
717 KASSERT(inp->inp_socket != NULL, ("in_pcbdetach: inp_socket == NULL"));
718 inp->inp_socket->so_pcb = NULL;
719 inp->inp_socket = NULL;
723 in_pcbfree(struct inpcb *inp)
725 struct inpcbinfo *ipi = inp->inp_pcbinfo;
727 KASSERT(inp->inp_socket == NULL, ("in_pcbfree: inp_socket != NULL"));
728 INP_INFO_WLOCK_ASSERT(ipi);
729 INP_LOCK_ASSERT(inp);
731 #if defined(IPSEC) || defined(FAST_IPSEC)
732 ipsec4_delete_pcbpolicy(inp);
734 inp->inp_gencnt = ++ipi->ipi_gencnt;
736 if (inp->inp_options)
737 (void)m_free(inp->inp_options);
738 ip_freemoptions(inp->inp_moptions);
742 mac_destroy_inpcb(inp);
745 uma_zfree(ipi->ipi_zone, inp);
749 * TCP needs to maintain its inpcb structure after the TCP connection has
750 * been torn down. However, it must be disconnected from the inpcb hashes as
751 * it must not prevent binding of future connections to the same port/ip
752 * combination by other inpcbs.
755 in_pcbdrop(struct inpcb *inp)
758 INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
759 INP_LOCK_ASSERT(inp);
761 inp->inp_vflag |= INP_DROPPED;
762 if (inp->inp_lport) {
763 struct inpcbport *phd = inp->inp_phd;
765 LIST_REMOVE(inp, inp_hash);
766 LIST_REMOVE(inp, inp_portlist);
767 if (LIST_FIRST(&phd->phd_pcblist) == NULL) {
768 LIST_REMOVE(phd, phd_hash);
776 in_sockaddr(in_port_t port, struct in_addr *addr_p)
778 struct sockaddr_in *sin;
780 MALLOC(sin, struct sockaddr_in *, sizeof *sin, M_SONAME,
782 sin->sin_family = AF_INET;
783 sin->sin_len = sizeof(*sin);
784 sin->sin_addr = *addr_p;
785 sin->sin_port = port;
787 return (struct sockaddr *)sin;
791 * The wrapper function will pass down the pcbinfo for this function to lock.
792 * The socket must have a valid
793 * (i.e., non-nil) PCB, but it should be impossible to get an invalid one
794 * except through a kernel programming error, so it is acceptable to panic
795 * (or in this case trap) if the PCB is invalid. (Actually, we don't trap
796 * because there actually /is/ a programming error somewhere... XXX)
799 in_setsockaddr(struct socket *so, struct sockaddr **nam,
800 struct inpcbinfo *pcbinfo)
807 KASSERT(inp != NULL, ("in_setsockaddr: inp == NULL"));
810 port = inp->inp_lport;
811 addr = inp->inp_laddr;
814 *nam = in_sockaddr(port, &addr);
819 * The wrapper function will pass down the pcbinfo for this function to lock.
822 in_setpeeraddr(struct socket *so, struct sockaddr **nam,
823 struct inpcbinfo *pcbinfo)
830 KASSERT(inp != NULL, ("in_setpeeraddr: inp == NULL"));
833 port = inp->inp_fport;
834 addr = inp->inp_faddr;
837 *nam = in_sockaddr(port, &addr);
842 in_pcbnotifyall(struct inpcbinfo *pcbinfo, struct in_addr faddr, int errno,
843 struct inpcb *(*notify)(struct inpcb *, int))
845 struct inpcb *inp, *ninp;
846 struct inpcbhead *head;
848 INP_INFO_WLOCK(pcbinfo);
849 head = pcbinfo->listhead;
850 for (inp = LIST_FIRST(head); inp != NULL; inp = ninp) {
852 ninp = LIST_NEXT(inp, inp_list);
854 if ((inp->inp_vflag & INP_IPV4) == 0) {
859 if (inp->inp_faddr.s_addr != faddr.s_addr ||
860 inp->inp_socket == NULL) {
864 if ((*notify)(inp, errno))
867 INP_INFO_WUNLOCK(pcbinfo);
871 in_pcbpurgeif0(struct inpcbinfo *pcbinfo, struct ifnet *ifp)
874 struct ip_moptions *imo;
877 INP_INFO_RLOCK(pcbinfo);
878 LIST_FOREACH(inp, pcbinfo->listhead, inp_list) {
880 imo = inp->inp_moptions;
881 if ((inp->inp_vflag & INP_IPV4) &&
884 * Unselect the outgoing interface if it is being
887 if (imo->imo_multicast_ifp == ifp)
888 imo->imo_multicast_ifp = NULL;
891 * Drop multicast group membership if we joined
892 * through the interface being detached.
894 for (i = 0, gap = 0; i < imo->imo_num_memberships;
896 if (imo->imo_membership[i]->inm_ifp == ifp) {
897 in_delmulti(imo->imo_membership[i]);
900 imo->imo_membership[i - gap] =
901 imo->imo_membership[i];
903 imo->imo_num_memberships -= gap;
907 INP_INFO_RUNLOCK(pcbinfo);
911 * Lookup a PCB based on the local address and port.
913 #define INP_LOOKUP_MAPPED_PCB_COST 3
915 in_pcblookup_local(struct inpcbinfo *pcbinfo, struct in_addr laddr,
916 u_int lport_arg, int wild_okay)
920 int matchwild = 3 + INP_LOOKUP_MAPPED_PCB_COST;
925 u_short lport = lport_arg;
927 INP_INFO_WLOCK_ASSERT(pcbinfo);
930 struct inpcbhead *head;
932 * Look for an unconnected (wildcard foreign addr) PCB that
933 * matches the local address and port we're looking for.
935 head = &pcbinfo->hashbase[INP_PCBHASH(INADDR_ANY, lport, 0, pcbinfo->hashmask)];
936 LIST_FOREACH(inp, head, inp_hash) {
938 if ((inp->inp_vflag & INP_IPV4) == 0)
941 if (inp->inp_faddr.s_addr == INADDR_ANY &&
942 inp->inp_laddr.s_addr == laddr.s_addr &&
943 inp->inp_lport == lport) {
955 struct inpcbporthead *porthash;
956 struct inpcbport *phd;
957 struct inpcb *match = NULL;
959 * Best fit PCB lookup.
961 * First see if this local port is in use by looking on the
964 porthash = &pcbinfo->porthashbase[INP_PCBPORTHASH(lport,
965 pcbinfo->porthashmask)];
966 LIST_FOREACH(phd, porthash, phd_hash) {
967 if (phd->phd_port == lport)
972 * Port is in use by one or more PCBs. Look for best
975 LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) {
978 if ((inp->inp_vflag & INP_IPV4) == 0)
981 * We never select the PCB that has
982 * INP_IPV6 flag and is bound to :: if
983 * we have another PCB which is bound
984 * to 0.0.0.0. If a PCB has the
985 * INP_IPV6 flag, then we set its cost
986 * higher than IPv4 only PCBs.
988 * Note that the case only happens
989 * when a socket is bound to ::, under
990 * the condition that the use of the
991 * mapped address is allowed.
993 if ((inp->inp_vflag & INP_IPV6) != 0)
994 wildcard += INP_LOOKUP_MAPPED_PCB_COST;
996 if (inp->inp_faddr.s_addr != INADDR_ANY)
998 if (inp->inp_laddr.s_addr != INADDR_ANY) {
999 if (laddr.s_addr == INADDR_ANY)
1001 else if (inp->inp_laddr.s_addr != laddr.s_addr)
1004 if (laddr.s_addr != INADDR_ANY)
1007 if (wildcard < matchwild) {
1009 matchwild = wildcard;
1010 if (matchwild == 0) {
1019 #undef INP_LOOKUP_MAPPED_PCB_COST
1022 * Lookup PCB in hash list.
1025 in_pcblookup_hash(struct inpcbinfo *pcbinfo, struct in_addr faddr,
1026 u_int fport_arg, struct in_addr laddr, u_int lport_arg, int wildcard,
1029 struct inpcbhead *head;
1031 u_short fport = fport_arg, lport = lport_arg;
1033 INP_INFO_RLOCK_ASSERT(pcbinfo);
1036 * First look for an exact match.
1038 head = &pcbinfo->hashbase[INP_PCBHASH(faddr.s_addr, lport, fport,
1039 pcbinfo->hashmask)];
1040 LIST_FOREACH(inp, head, inp_hash) {
1042 if ((inp->inp_vflag & INP_IPV4) == 0)
1045 if (inp->inp_faddr.s_addr == faddr.s_addr &&
1046 inp->inp_laddr.s_addr == laddr.s_addr &&
1047 inp->inp_fport == fport &&
1048 inp->inp_lport == lport)
1053 * Then look for a wildcard match, if requested.
1056 struct inpcb *local_wild = NULL;
1058 struct inpcb *local_wild_mapped = NULL;
1061 head = &pcbinfo->hashbase[INP_PCBHASH(INADDR_ANY, lport, 0,
1062 pcbinfo->hashmask)];
1063 LIST_FOREACH(inp, head, inp_hash) {
1065 if ((inp->inp_vflag & INP_IPV4) == 0)
1068 if (inp->inp_faddr.s_addr == INADDR_ANY &&
1069 inp->inp_lport == lport) {
1070 if (ifp && ifp->if_type == IFT_FAITH &&
1071 (inp->inp_flags & INP_FAITH) == 0)
1073 if (inp->inp_laddr.s_addr == laddr.s_addr)
1075 else if (inp->inp_laddr.s_addr == INADDR_ANY) {
1077 if (INP_CHECK_SOCKAF(inp->inp_socket,
1079 local_wild_mapped = inp;
1087 if (local_wild == NULL)
1088 return (local_wild_mapped);
1090 return (local_wild);
1096 * Insert PCB onto various hash lists.
1099 in_pcbinshash(struct inpcb *inp)
1101 struct inpcbhead *pcbhash;
1102 struct inpcbporthead *pcbporthash;
1103 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1104 struct inpcbport *phd;
1105 u_int32_t hashkey_faddr;
1107 INP_INFO_WLOCK_ASSERT(pcbinfo);
1108 INP_LOCK_ASSERT(inp);
1111 if (inp->inp_vflag & INP_IPV6)
1112 hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX */;
1115 hashkey_faddr = inp->inp_faddr.s_addr;
1117 pcbhash = &pcbinfo->hashbase[INP_PCBHASH(hashkey_faddr,
1118 inp->inp_lport, inp->inp_fport, pcbinfo->hashmask)];
1120 pcbporthash = &pcbinfo->porthashbase[INP_PCBPORTHASH(inp->inp_lport,
1121 pcbinfo->porthashmask)];
1124 * Go through port list and look for a head for this lport.
1126 LIST_FOREACH(phd, pcbporthash, phd_hash) {
1127 if (phd->phd_port == inp->inp_lport)
1131 * If none exists, malloc one and tack it on.
1134 MALLOC(phd, struct inpcbport *, sizeof(struct inpcbport), M_PCB, M_NOWAIT);
1136 return (ENOBUFS); /* XXX */
1138 phd->phd_port = inp->inp_lport;
1139 LIST_INIT(&phd->phd_pcblist);
1140 LIST_INSERT_HEAD(pcbporthash, phd, phd_hash);
1143 LIST_INSERT_HEAD(&phd->phd_pcblist, inp, inp_portlist);
1144 LIST_INSERT_HEAD(pcbhash, inp, inp_hash);
1149 * Move PCB to the proper hash bucket when { faddr, fport } have been
1150 * changed. NOTE: This does not handle the case of the lport changing (the
1151 * hashed port list would have to be updated as well), so the lport must
1152 * not change after in_pcbinshash() has been called.
1155 in_pcbrehash(struct inpcb *inp)
1157 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1158 struct inpcbhead *head;
1159 u_int32_t hashkey_faddr;
1161 INP_INFO_WLOCK_ASSERT(pcbinfo);
1162 INP_LOCK_ASSERT(inp);
1165 if (inp->inp_vflag & INP_IPV6)
1166 hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX */;
1169 hashkey_faddr = inp->inp_faddr.s_addr;
1171 head = &pcbinfo->hashbase[INP_PCBHASH(hashkey_faddr,
1172 inp->inp_lport, inp->inp_fport, pcbinfo->hashmask)];
1174 LIST_REMOVE(inp, inp_hash);
1175 LIST_INSERT_HEAD(head, inp, inp_hash);
1179 * Remove PCB from various lists.
1182 in_pcbremlists(struct inpcb *inp)
1184 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1186 INP_INFO_WLOCK_ASSERT(pcbinfo);
1187 INP_LOCK_ASSERT(inp);
1189 inp->inp_gencnt = ++pcbinfo->ipi_gencnt;
1190 if (inp->inp_lport) {
1191 struct inpcbport *phd = inp->inp_phd;
1193 LIST_REMOVE(inp, inp_hash);
1194 LIST_REMOVE(inp, inp_portlist);
1195 if (LIST_FIRST(&phd->phd_pcblist) == NULL) {
1196 LIST_REMOVE(phd, phd_hash);
1200 LIST_REMOVE(inp, inp_list);
1201 pcbinfo->ipi_count--;
1205 * A set label operation has occurred at the socket layer, propagate the
1206 * label change into the in_pcb for the socket.
1209 in_pcbsosetlabel(struct socket *so)
1214 inp = sotoinpcb(so);
1215 KASSERT(inp != NULL, ("in_pcbsosetlabel: so->so_pcb == NULL"));
1219 mac_inpcb_sosetlabel(so, inp);
1226 * ipport_tick runs once per second, determining if random port allocation
1227 * should be continued. If more than ipport_randomcps ports have been
1228 * allocated in the last second, then we return to sequential port
1229 * allocation. We return to random allocation only once we drop below
1230 * ipport_randomcps for at least ipport_randomtime seconds.
1233 ipport_tick(void *xtp)
1236 if (ipport_tcpallocs <= ipport_tcplastcount + ipport_randomcps) {
1237 if (ipport_stoprandom > 0)
1238 ipport_stoprandom--;
1240 ipport_stoprandom = ipport_randomtime;
1241 ipport_tcplastcount = ipport_tcpallocs;
1242 callout_reset(&ipport_tick_callout, hz, ipport_tick, NULL);
1247 db_print_indent(int indent)
1251 for (i = 0; i < indent; i++)
1256 db_print_inconninfo(struct in_conninfo *inc, const char *name, int indent)
1258 char faddr_str[48], laddr_str[48];
1260 db_print_indent(indent);
1261 db_printf("%s at %p\n", name, inc);
1266 if (inc->inc_flags == 1) {
1268 ip6_sprintf(laddr_str, &inc->inc6_laddr);
1269 ip6_sprintf(faddr_str, &inc->inc6_faddr);
1273 inet_ntoa_r(inc->inc_laddr, laddr_str);
1274 inet_ntoa_r(inc->inc_faddr, faddr_str);
1278 db_print_indent(indent);
1279 db_printf("inc_laddr %s inc_lport %u\n", laddr_str,
1280 ntohs(inc->inc_lport));
1281 db_print_indent(indent);
1282 db_printf("inc_faddr %s inc_fport %u\n", faddr_str,
1283 ntohs(inc->inc_fport));
1287 db_print_inpflags(int inp_flags)
1292 if (inp_flags & INP_RECVOPTS) {
1293 db_printf("%sINP_RECVOPTS", comma ? ", " : "");
1296 if (inp_flags & INP_RECVRETOPTS) {
1297 db_printf("%sINP_RECVRETOPTS", comma ? ", " : "");
1300 if (inp_flags & INP_RECVDSTADDR) {
1301 db_printf("%sINP_RECVDSTADDR", comma ? ", " : "");
1304 if (inp_flags & INP_HDRINCL) {
1305 db_printf("%sINP_HDRINCL", comma ? ", " : "");
1308 if (inp_flags & INP_HIGHPORT) {
1309 db_printf("%sINP_HIGHPORT", comma ? ", " : "");
1312 if (inp_flags & INP_LOWPORT) {
1313 db_printf("%sINP_LOWPORT", comma ? ", " : "");
1316 if (inp_flags & INP_ANONPORT) {
1317 db_printf("%sINP_ANONPORT", comma ? ", " : "");
1320 if (inp_flags & INP_RECVIF) {
1321 db_printf("%sINP_RECVIF", comma ? ", " : "");
1324 if (inp_flags & INP_MTUDISC) {
1325 db_printf("%sINP_MTUDISC", comma ? ", " : "");
1328 if (inp_flags & INP_FAITH) {
1329 db_printf("%sINP_FAITH", comma ? ", " : "");
1332 if (inp_flags & INP_RECVTTL) {
1333 db_printf("%sINP_RECVTTL", comma ? ", " : "");
1336 if (inp_flags & INP_DONTFRAG) {
1337 db_printf("%sINP_DONTFRAG", comma ? ", " : "");
1340 if (inp_flags & IN6P_IPV6_V6ONLY) {
1341 db_printf("%sIN6P_IPV6_V6ONLY", comma ? ", " : "");
1344 if (inp_flags & IN6P_PKTINFO) {
1345 db_printf("%sIN6P_PKTINFO", comma ? ", " : "");
1348 if (inp_flags & IN6P_HOPLIMIT) {
1349 db_printf("%sIN6P_HOPLIMIT", comma ? ", " : "");
1352 if (inp_flags & IN6P_HOPOPTS) {
1353 db_printf("%sIN6P_HOPOPTS", comma ? ", " : "");
1356 if (inp_flags & IN6P_DSTOPTS) {
1357 db_printf("%sIN6P_DSTOPTS", comma ? ", " : "");
1360 if (inp_flags & IN6P_RTHDR) {
1361 db_printf("%sIN6P_RTHDR", comma ? ", " : "");
1364 if (inp_flags & IN6P_RTHDRDSTOPTS) {
1365 db_printf("%sIN6P_RTHDRDSTOPTS", comma ? ", " : "");
1368 if (inp_flags & IN6P_TCLASS) {
1369 db_printf("%sIN6P_TCLASS", comma ? ", " : "");
1372 if (inp_flags & IN6P_AUTOFLOWLABEL) {
1373 db_printf("%sIN6P_AUTOFLOWLABEL", comma ? ", " : "");
1376 if (inp_flags & IN6P_RFC2292) {
1377 db_printf("%sIN6P_RFC2292", comma ? ", " : "");
1380 if (inp_flags & IN6P_MTU) {
1381 db_printf("IN6P_MTU%s", comma ? ", " : "");
1387 db_print_inpvflag(u_char inp_vflag)
1392 if (inp_vflag & INP_IPV4) {
1393 db_printf("%sINP_IPV4", comma ? ", " : "");
1396 if (inp_vflag & INP_IPV6) {
1397 db_printf("%sINP_IPV6", comma ? ", " : "");
1400 if (inp_vflag & INP_IPV6PROTO) {
1401 db_printf("%sINP_IPV6PROTO", comma ? ", " : "");
1404 if (inp_vflag & INP_TIMEWAIT) {
1405 db_printf("%sINP_TIMEWAIT", comma ? ", " : "");
1408 if (inp_vflag & INP_ONESBCAST) {
1409 db_printf("%sINP_ONESBCAST", comma ? ", " : "");
1412 if (inp_vflag & INP_DROPPED) {
1413 db_printf("%sINP_DROPPED", comma ? ", " : "");
1416 if (inp_vflag & INP_SOCKREF) {
1417 db_printf("%sINP_SOCKREF", comma ? ", " : "");
1423 db_print_inpcb(struct inpcb *inp, const char *name, int indent)
1426 db_print_indent(indent);
1427 db_printf("%s at %p\n", name, inp);
1431 db_print_indent(indent);
1432 db_printf("inp_flow: 0x%x\n", inp->inp_flow);
1434 db_print_inconninfo(&inp->inp_inc, "inp_conninfo", indent);
1436 db_print_indent(indent);
1437 db_printf("inp_ppcb: %p inp_pcbinfo: %p inp_socket: %p\n",
1438 inp->inp_ppcb, inp->inp_pcbinfo, inp->inp_socket);
1440 db_print_indent(indent);
1441 db_printf("inp_label: %p inp_flags: 0x%x (",
1442 inp->inp_label, inp->inp_flags);
1443 db_print_inpflags(inp->inp_flags);
1446 db_print_indent(indent);
1447 db_printf("inp_sp: %p inp_vflag: 0x%x (", inp->inp_sp,
1449 db_print_inpvflag(inp->inp_vflag);
1452 db_print_indent(indent);
1453 db_printf("inp_ip_ttl: %d inp_ip_p: %d inp_ip_minttl: %d\n",
1454 inp->inp_ip_ttl, inp->inp_ip_p, inp->inp_ip_minttl);
1456 db_print_indent(indent);
1458 if (inp->inp_vflag & INP_IPV6) {
1459 db_printf("in6p_options: %p in6p_outputopts: %p "
1460 "in6p_moptions: %p\n", inp->in6p_options,
1461 inp->in6p_outputopts, inp->in6p_moptions);
1462 db_printf("in6p_icmp6filt: %p in6p_cksum %d "
1463 "in6p_hops %u\n", inp->in6p_icmp6filt, inp->in6p_cksum,
1468 db_printf("inp_ip_tos: %d inp_ip_options: %p "
1469 "inp_ip_moptions: %p\n", inp->inp_ip_tos,
1470 inp->inp_options, inp->inp_moptions);
1473 db_print_indent(indent);
1474 db_printf("inp_phd: %p inp_gencnt: %ju\n", inp->inp_phd,
1475 (uintmax_t)inp->inp_gencnt);
1478 DB_SHOW_COMMAND(inpcb, db_show_inpcb)
1483 db_printf("usage: show inpcb <addr>\n");
1486 inp = (struct inpcb *)addr;
1488 db_print_inpcb(inp, "inpcb", 0);