2 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
3 * Copyright (c) 2010-2011 Juniper Networks, Inc.
6 * Portions of this software were developed by Robert N. M. Watson under
7 * contract to Juniper Networks, Inc.
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
17 * 3. Neither the name of the project nor the names of its contributors
18 * may be used to endorse or promote products derived from this software
19 * without specific prior written permission.
21 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33 * $KAME: in6_pcb.c,v 1.31 2001/05/21 05:45:10 jinmei Exp $
37 * Copyright (c) 1982, 1986, 1991, 1993
38 * The Regents of the University of California. All rights reserved.
40 * Redistribution and use in source and binary forms, with or without
41 * modification, are permitted provided that the following conditions
43 * 1. Redistributions of source code must retain the above copyright
44 * notice, this list of conditions and the following disclaimer.
45 * 2. Redistributions in binary form must reproduce the above copyright
46 * notice, this list of conditions and the following disclaimer in the
47 * documentation and/or other materials provided with the distribution.
48 * 4. Neither the name of the University nor the names of its contributors
49 * may be used to endorse or promote products derived from this software
50 * without specific prior written permission.
52 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
53 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
54 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
55 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
56 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
57 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
58 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
59 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
60 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
61 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
64 * @(#)in_pcb.c 8.2 (Berkeley) 1/4/94
67 #include <sys/cdefs.h>
68 __FBSDID("$FreeBSD$");
71 #include "opt_inet6.h"
72 #include "opt_ipsec.h"
73 #include "opt_pcbgroup.h"
76 #include <sys/param.h>
77 #include <sys/systm.h>
78 #include <sys/malloc.h>
80 #include <sys/domain.h>
81 #include <sys/protosw.h>
82 #include <sys/socket.h>
83 #include <sys/socketvar.h>
84 #include <sys/sockio.h>
85 #include <sys/errno.h>
94 #include <net/if_var.h>
95 #include <net/if_types.h>
96 #include <net/route.h>
98 #include <netinet/in.h>
99 #include <netinet/in_var.h>
100 #include <netinet/in_systm.h>
101 #include <netinet/tcp_var.h>
102 #include <netinet/ip6.h>
103 #include <netinet/ip_var.h>
105 #include <netinet6/ip6_var.h>
106 #include <netinet6/nd6.h>
107 #include <netinet/in_pcb.h>
108 #include <netinet6/in6_pcb.h>
109 #include <netinet6/scope6_var.h>
111 static struct inpcb *in6_pcblookup_hash_locked(struct inpcbinfo *,
112 struct in6_addr *, u_int, struct in6_addr *, u_int, int, struct ifnet *);
115 in6_pcbbind(register struct inpcb *inp, struct sockaddr *nam,
118 struct socket *so = inp->inp_socket;
119 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)NULL;
120 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
122 int error, lookupflags = 0;
123 int reuseport = (so->so_options & SO_REUSEPORT);
125 INP_WLOCK_ASSERT(inp);
126 INP_HASH_WLOCK_ASSERT(pcbinfo);
128 if (TAILQ_EMPTY(&V_in6_ifaddrhead)) /* XXX broken! */
129 return (EADDRNOTAVAIL);
130 if (inp->inp_lport || !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr))
132 if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0)
133 lookupflags = INPLOOKUP_WILDCARD;
135 if ((error = prison_local_ip6(cred, &inp->in6p_laddr,
136 ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0)
139 sin6 = (struct sockaddr_in6 *)nam;
140 if (nam->sa_len != sizeof(*sin6))
145 if (nam->sa_family != AF_INET6)
146 return (EAFNOSUPPORT);
148 if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0)
151 if ((error = prison_local_ip6(cred, &sin6->sin6_addr,
152 ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0)
155 lport = sin6->sin6_port;
156 if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
158 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
159 * allow compepte duplication of binding if
160 * SO_REUSEPORT is set, or if SO_REUSEADDR is set
161 * and a multicast address is bound on both
162 * new and duplicated sockets.
164 if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) != 0)
165 reuseport = SO_REUSEADDR|SO_REUSEPORT;
166 } else if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
169 sin6->sin6_port = 0; /* yech... */
170 if ((ifa = ifa_ifwithaddr((struct sockaddr *)sin6)) ==
172 (inp->inp_flags & INP_BINDANY) == 0) {
173 return (EADDRNOTAVAIL);
177 * XXX: bind to an anycast address might accidentally
178 * cause sending a packet with anycast source address.
179 * We should allow to bind to a deprecated address, since
180 * the application dares to use it.
183 ((struct in6_ifaddr *)ifa)->ia6_flags &
184 (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY|IN6_IFF_DETACHED)) {
186 return (EADDRNOTAVAIL);
196 if (ntohs(lport) <= V_ipport_reservedhigh &&
197 ntohs(lport) >= V_ipport_reservedlow &&
198 priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT,
201 if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr) &&
202 priv_check_cred(inp->inp_cred,
203 PRIV_NETINET_REUSEPORT, 0) != 0) {
204 t = in6_pcblookup_local(pcbinfo,
205 &sin6->sin6_addr, lport,
206 INPLOOKUP_WILDCARD, cred);
208 ((inp->inp_flags2 & INP_BINDMULTI) == 0) &&
209 ((t->inp_flags & INP_TIMEWAIT) == 0) &&
210 (so->so_type != SOCK_STREAM ||
211 IN6_IS_ADDR_UNSPECIFIED(&t->in6p_faddr)) &&
212 (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) ||
213 !IN6_IS_ADDR_UNSPECIFIED(&t->in6p_laddr) ||
214 (t->inp_flags2 & INP_REUSEPORT) == 0) &&
215 (inp->inp_cred->cr_uid !=
216 t->inp_cred->cr_uid))
220 * If the socket is a BINDMULTI socket, then
221 * the credentials need to match and the
222 * original socket also has to have been bound
225 if (t && (! in_pcbbind_check_bindmulti(inp, t)))
229 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0 &&
230 IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
231 struct sockaddr_in sin;
233 in6_sin6_2_sin(&sin, sin6);
234 t = in_pcblookup_local(pcbinfo,
236 INPLOOKUP_WILDCARD, cred);
238 ((inp->inp_flags2 & INP_BINDMULTI) == 0) &&
240 INP_TIMEWAIT) == 0) &&
241 (so->so_type != SOCK_STREAM ||
242 ntohl(t->inp_faddr.s_addr) ==
244 (inp->inp_cred->cr_uid !=
245 t->inp_cred->cr_uid))
248 if (t && (! in_pcbbind_check_bindmulti(inp, t)))
253 t = in6_pcblookup_local(pcbinfo, &sin6->sin6_addr,
254 lport, lookupflags, cred);
255 if (t && (t->inp_flags & INP_TIMEWAIT)) {
257 * XXXRW: If an incpb has had its timewait
258 * state recycled, we treat the address as
259 * being in use (for now). This is better
260 * than a panic, but not desirable.
264 (reuseport & tw->tw_so_options) == 0)
266 } else if (t && (reuseport & inp_so_options(t)) == 0) {
270 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0 &&
271 IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
272 struct sockaddr_in sin;
274 in6_sin6_2_sin(&sin, sin6);
275 t = in_pcblookup_local(pcbinfo, sin.sin_addr,
276 lport, lookupflags, cred);
277 if (t && t->inp_flags & INP_TIMEWAIT) {
281 if ((reuseport & tw->tw_so_options) == 0
282 && (ntohl(t->inp_laddr.s_addr) !=
283 INADDR_ANY || ((inp->inp_vflag &
285 (t->inp_vflag & INP_IPV6PROTO))))
288 (reuseport & inp_so_options(t)) == 0 &&
289 (ntohl(t->inp_laddr.s_addr) != INADDR_ANY ||
290 (t->inp_vflag & INP_IPV6PROTO) != 0))
295 inp->in6p_laddr = sin6->sin6_addr;
298 if ((error = in6_pcbsetport(&inp->in6p_laddr, inp, cred)) != 0) {
299 /* Undo an address bind that may have occurred. */
300 inp->in6p_laddr = in6addr_any;
304 inp->inp_lport = lport;
305 if (in_pcbinshash(inp) != 0) {
306 inp->in6p_laddr = in6addr_any;
315 * Transform old in6_pcbconnect() into an inner subroutine for new
316 * in6_pcbconnect(): Do some validity-checking on the remote
317 * address (in mbuf 'nam') and then determine local host address
318 * (i.e., which interface) to use to access that remote host.
320 * This preserves definition of in6_pcbconnect(), while supporting a
321 * slightly different version for T/TCP. (This is more than
322 * a bit of a kludge, but cleaning up the internal interfaces would
323 * have forced minor changes in every protocol).
326 in6_pcbladdr(register struct inpcb *inp, struct sockaddr *nam,
327 struct in6_addr *plocal_addr6)
329 register struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam;
331 int scope_ambiguous = 0;
332 struct in6_addr in6a;
334 INP_WLOCK_ASSERT(inp);
335 INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo); /* XXXRW: why? */
337 if (nam->sa_len != sizeof (*sin6))
339 if (sin6->sin6_family != AF_INET6)
340 return (EAFNOSUPPORT);
341 if (sin6->sin6_port == 0)
342 return (EADDRNOTAVAIL);
344 if (sin6->sin6_scope_id == 0 && !V_ip6_use_defzone)
346 if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0)
349 if (!TAILQ_EMPTY(&V_in6_ifaddrhead)) {
351 * If the destination address is UNSPECIFIED addr,
352 * use the loopback addr, e.g ::1.
354 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))
355 sin6->sin6_addr = in6addr_loopback;
357 if ((error = prison_remote_ip6(inp->inp_cred, &sin6->sin6_addr)) != 0)
360 error = in6_selectsrc_socket(sin6, inp->in6p_outputopts,
361 inp, inp->inp_cred, scope_ambiguous, &in6a, NULL);
366 * Do not update this earlier, in case we return with an error.
368 * XXX: this in6_selectsrc_socket result might replace the bound local
369 * address with the address specified by setsockopt(IPV6_PKTINFO).
370 * Is it the intended behavior?
372 *plocal_addr6 = in6a;
375 * Don't do pcblookup call here; return interface in
377 * and exit to caller, that will do the lookup.
385 * Connect from a socket to a specified address.
386 * Both address and port must be specified in argument sin.
387 * If don't have a local address for this socket yet,
391 in6_pcbconnect_mbuf(register struct inpcb *inp, struct sockaddr *nam,
392 struct ucred *cred, struct mbuf *m)
394 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
395 register struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam;
396 struct in6_addr addr6;
399 INP_WLOCK_ASSERT(inp);
400 INP_HASH_WLOCK_ASSERT(pcbinfo);
403 * Call inner routine, to assign local interface address.
404 * in6_pcbladdr() may automatically fill in sin6_scope_id.
406 if ((error = in6_pcbladdr(inp, nam, &addr6)) != 0)
409 if (in6_pcblookup_hash_locked(pcbinfo, &sin6->sin6_addr,
411 IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)
412 ? &addr6 : &inp->in6p_laddr,
413 inp->inp_lport, 0, NULL) != NULL) {
416 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
417 if (inp->inp_lport == 0) {
418 error = in6_pcbbind(inp, (struct sockaddr *)0, cred);
422 inp->in6p_laddr = addr6;
424 inp->in6p_faddr = sin6->sin6_addr;
425 inp->inp_fport = sin6->sin6_port;
426 /* update flowinfo - draft-itojun-ipv6-flowlabel-api-00 */
427 inp->inp_flow &= ~IPV6_FLOWLABEL_MASK;
428 if (inp->inp_flags & IN6P_AUTOFLOWLABEL)
430 (htonl(ip6_randomflowlabel()) & IPV6_FLOWLABEL_MASK);
432 in_pcbrehash_mbuf(inp, m);
438 in6_pcbconnect(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred)
441 return (in6_pcbconnect_mbuf(inp, nam, cred, NULL));
445 in6_pcbdisconnect(struct inpcb *inp)
448 INP_WLOCK_ASSERT(inp);
449 INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo);
451 bzero((caddr_t)&inp->in6p_faddr, sizeof(inp->in6p_faddr));
453 /* clear flowinfo - draft-itojun-ipv6-flowlabel-api-00 */
454 inp->inp_flow &= ~IPV6_FLOWLABEL_MASK;
459 in6_sockaddr(in_port_t port, struct in6_addr *addr_p)
461 struct sockaddr_in6 *sin6;
463 sin6 = malloc(sizeof *sin6, M_SONAME, M_WAITOK);
464 bzero(sin6, sizeof *sin6);
465 sin6->sin6_family = AF_INET6;
466 sin6->sin6_len = sizeof(*sin6);
467 sin6->sin6_port = port;
468 sin6->sin6_addr = *addr_p;
469 (void)sa6_recoverscope(sin6); /* XXX: should catch errors */
471 return (struct sockaddr *)sin6;
475 in6_v4mapsin6_sockaddr(in_port_t port, struct in_addr *addr_p)
477 struct sockaddr_in sin;
478 struct sockaddr_in6 *sin6_p;
480 bzero(&sin, sizeof sin);
481 sin.sin_family = AF_INET;
482 sin.sin_len = sizeof(sin);
484 sin.sin_addr = *addr_p;
486 sin6_p = malloc(sizeof *sin6_p, M_SONAME,
488 in6_sin_2_v4mapsin6(&sin, sin6_p);
490 return (struct sockaddr *)sin6_p;
494 in6_getsockaddr(struct socket *so, struct sockaddr **nam)
496 register struct inpcb *inp;
497 struct in6_addr addr;
501 KASSERT(inp != NULL, ("in6_getsockaddr: inp == NULL"));
504 port = inp->inp_lport;
505 addr = inp->in6p_laddr;
508 *nam = in6_sockaddr(port, &addr);
513 in6_getpeeraddr(struct socket *so, struct sockaddr **nam)
516 struct in6_addr addr;
520 KASSERT(inp != NULL, ("in6_getpeeraddr: inp == NULL"));
523 port = inp->inp_fport;
524 addr = inp->in6p_faddr;
527 *nam = in6_sockaddr(port, &addr);
532 in6_mapped_sockaddr(struct socket *so, struct sockaddr **nam)
538 KASSERT(inp != NULL, ("in6_mapped_sockaddr: inp == NULL"));
541 if ((inp->inp_vflag & (INP_IPV4 | INP_IPV6)) == INP_IPV4) {
542 error = in_getsockaddr(so, nam);
544 in6_sin_2_v4mapsin6_in_sock(nam);
548 /* scope issues will be handled in in6_getsockaddr(). */
549 error = in6_getsockaddr(so, nam);
556 in6_mapped_peeraddr(struct socket *so, struct sockaddr **nam)
562 KASSERT(inp != NULL, ("in6_mapped_peeraddr: inp == NULL"));
565 if ((inp->inp_vflag & (INP_IPV4 | INP_IPV6)) == INP_IPV4) {
566 error = in_getpeeraddr(so, nam);
568 in6_sin_2_v4mapsin6_in_sock(nam);
571 /* scope issues will be handled in in6_getpeeraddr(). */
572 error = in6_getpeeraddr(so, nam);
578 * Pass some notification to all connections of a protocol
579 * associated with address dst. The local address and/or port numbers
580 * may be specified to limit the search. The "usual action" will be
581 * taken, depending on the ctlinput cmd. The caller must filter any
582 * cmds that are uninteresting (e.g., no error in the map).
583 * Call the protocol specific routine (if any) to report
584 * any errors for each matching socket.
587 in6_pcbnotify(struct inpcbinfo *pcbinfo, struct sockaddr *dst,
588 u_int fport_arg, const struct sockaddr *src, u_int lport_arg,
589 int cmd, void *cmdarg,
590 struct inpcb *(*notify)(struct inpcb *, int))
592 struct inpcb *inp, *inp_temp;
593 struct sockaddr_in6 sa6_src, *sa6_dst;
594 u_short fport = fport_arg, lport = lport_arg;
598 if ((unsigned)cmd >= PRC_NCMDS || dst->sa_family != AF_INET6)
601 sa6_dst = (struct sockaddr_in6 *)dst;
602 if (IN6_IS_ADDR_UNSPECIFIED(&sa6_dst->sin6_addr))
606 * note that src can be NULL when we get notify by local fragmentation.
608 sa6_src = (src == NULL) ? sa6_any : *(const struct sockaddr_in6 *)src;
609 flowinfo = sa6_src.sin6_flowinfo;
612 * Redirects go to all references to the destination,
613 * and use in6_rtchange to invalidate the route cache.
614 * Dead host indications: also use in6_rtchange to invalidate
615 * the cache, and deliver the error to all the sockets.
616 * Otherwise, if we have knowledge of the local port and address,
617 * deliver only to that socket.
619 if (PRC_IS_REDIRECT(cmd) || cmd == PRC_HOSTDEAD) {
622 bzero((caddr_t)&sa6_src.sin6_addr, sizeof(sa6_src.sin6_addr));
624 if (cmd != PRC_HOSTDEAD)
625 notify = in6_rtchange;
627 errno = inet6ctlerrmap[cmd];
628 INP_INFO_WLOCK(pcbinfo);
629 LIST_FOREACH_SAFE(inp, pcbinfo->ipi_listhead, inp_list, inp_temp) {
631 if ((inp->inp_vflag & INP_IPV6) == 0) {
637 * If the error designates a new path MTU for a destination
638 * and the application (associated with this socket) wanted to
639 * know the value, notify.
640 * XXX: should we avoid to notify the value to TCP sockets?
642 if (cmd == PRC_MSGSIZE && cmdarg != NULL)
643 ip6_notify_pmtu(inp, (struct sockaddr_in6 *)dst,
644 *(u_int32_t *)cmdarg);
647 * Detect if we should notify the error. If no source and
648 * destination ports are specifed, but non-zero flowinfo and
649 * local address match, notify the error. This is the case
650 * when the error is delivered with an encrypted buffer
651 * by ESP. Otherwise, just compare addresses and ports
654 if (lport == 0 && fport == 0 && flowinfo &&
655 inp->inp_socket != NULL &&
656 flowinfo == (inp->inp_flow & IPV6_FLOWLABEL_MASK) &&
657 IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, &sa6_src.sin6_addr))
659 else if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr,
660 &sa6_dst->sin6_addr) ||
661 inp->inp_socket == 0 ||
662 (lport && inp->inp_lport != lport) ||
663 (!IN6_IS_ADDR_UNSPECIFIED(&sa6_src.sin6_addr) &&
664 !IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr,
665 &sa6_src.sin6_addr)) ||
666 (fport && inp->inp_fport != fport)) {
673 if ((*notify)(inp, errno))
678 INP_INFO_WUNLOCK(pcbinfo);
682 * Lookup a PCB based on the local address and port. Caller must hold the
683 * hash lock. No inpcb locks or references are acquired.
686 in6_pcblookup_local(struct inpcbinfo *pcbinfo, struct in6_addr *laddr,
687 u_short lport, int lookupflags, struct ucred *cred)
689 register struct inpcb *inp;
690 int matchwild = 3, wildcard;
692 KASSERT((lookupflags & ~(INPLOOKUP_WILDCARD)) == 0,
693 ("%s: invalid lookup flags %d", __func__, lookupflags));
695 INP_HASH_WLOCK_ASSERT(pcbinfo);
697 if ((lookupflags & INPLOOKUP_WILDCARD) == 0) {
698 struct inpcbhead *head;
700 * Look for an unconnected (wildcard foreign addr) PCB that
701 * matches the local address and port we're looking for.
703 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(
704 INP6_PCBHASHKEY(&in6addr_any), lport, 0,
705 pcbinfo->ipi_hashmask)];
706 LIST_FOREACH(inp, head, inp_hash) {
707 /* XXX inp locking */
708 if ((inp->inp_vflag & INP_IPV6) == 0)
710 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) &&
711 IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr) &&
712 inp->inp_lport == lport) {
715 prison_equal_ip6(cred->cr_prison,
716 inp->inp_cred->cr_prison))
725 struct inpcbporthead *porthash;
726 struct inpcbport *phd;
727 struct inpcb *match = NULL;
729 * Best fit PCB lookup.
731 * First see if this local port is in use by looking on the
734 porthash = &pcbinfo->ipi_porthashbase[INP_PCBPORTHASH(lport,
735 pcbinfo->ipi_porthashmask)];
736 LIST_FOREACH(phd, porthash, phd_hash) {
737 if (phd->phd_port == lport)
742 * Port is in use by one or more PCBs. Look for best
745 LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) {
748 !prison_equal_ip6(cred->cr_prison,
749 inp->inp_cred->cr_prison))
751 /* XXX inp locking */
752 if ((inp->inp_vflag & INP_IPV6) == 0)
754 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr))
756 if (!IN6_IS_ADDR_UNSPECIFIED(
758 if (IN6_IS_ADDR_UNSPECIFIED(laddr))
760 else if (!IN6_ARE_ADDR_EQUAL(
761 &inp->in6p_laddr, laddr))
764 if (!IN6_IS_ADDR_UNSPECIFIED(laddr))
767 if (wildcard < matchwild) {
769 matchwild = wildcard;
780 in6_pcbpurgeif0(struct inpcbinfo *pcbinfo, struct ifnet *ifp)
783 struct ip6_moptions *im6o;
786 INP_INFO_WLOCK(pcbinfo);
787 LIST_FOREACH(in6p, pcbinfo->ipi_listhead, inp_list) {
789 im6o = in6p->in6p_moptions;
790 if ((in6p->inp_vflag & INP_IPV6) && im6o != NULL) {
792 * Unselect the outgoing ifp for multicast if it
795 if (im6o->im6o_multicast_ifp == ifp)
796 im6o->im6o_multicast_ifp = NULL;
798 * Drop multicast group membership if we joined
799 * through the interface being detached.
802 for (i = 0; i < im6o->im6o_num_memberships; i++) {
803 if (im6o->im6o_membership[i]->in6m_ifp ==
805 in6_mc_leave(im6o->im6o_membership[i],
808 } else if (gap != 0) {
809 im6o->im6o_membership[i - gap] =
810 im6o->im6o_membership[i];
813 im6o->im6o_num_memberships -= gap;
817 INP_INFO_WUNLOCK(pcbinfo);
821 * Check for alternatives when higher level complains
822 * about service problems. For now, invalidate cached
823 * routing information. If the route was created dynamically
824 * (by a redirect), time to try a default gateway again.
827 in6_losing(struct inpcb *in6p)
831 * We don't store route pointers in the routing table anymore
837 * After a routing change, flush old routing
838 * and allocate a (hopefully) better one.
841 in6_rtchange(struct inpcb *inp, int errno)
844 * We don't store route pointers in the routing table anymore
851 * Lookup PCB in hash list, using pcbgroup tables.
853 static struct inpcb *
854 in6_pcblookup_group(struct inpcbinfo *pcbinfo, struct inpcbgroup *pcbgroup,
855 struct in6_addr *faddr, u_int fport_arg, struct in6_addr *laddr,
856 u_int lport_arg, int lookupflags, struct ifnet *ifp)
858 struct inpcbhead *head;
859 struct inpcb *inp, *tmpinp;
860 u_short fport = fport_arg, lport = lport_arg;
863 * First look for an exact match.
866 INP_GROUP_LOCK(pcbgroup);
867 head = &pcbgroup->ipg_hashbase[INP_PCBHASH(
868 INP6_PCBHASHKEY(faddr), lport, fport, pcbgroup->ipg_hashmask)];
869 LIST_FOREACH(inp, head, inp_pcbgrouphash) {
870 /* XXX inp locking */
871 if ((inp->inp_vflag & INP_IPV6) == 0)
873 if (IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, faddr) &&
874 IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr) &&
875 inp->inp_fport == fport &&
876 inp->inp_lport == lport) {
878 * XXX We should be able to directly return
879 * the inp here, without any checks.
880 * Well unless both bound with SO_REUSEPORT?
882 if (prison_flag(inp->inp_cred, PR_IP6))
888 if (tmpinp != NULL) {
894 * Then look for a wildcard match in the pcbgroup.
896 if ((lookupflags & INPLOOKUP_WILDCARD) != 0) {
897 struct inpcb *local_wild = NULL, *local_exact = NULL;
898 struct inpcb *jail_wild = NULL;
902 * Order of socket selection - we always prefer jails.
903 * 1. jailed, non-wild.
905 * 3. non-jailed, non-wild.
906 * 4. non-jailed, wild.
908 head = &pcbgroup->ipg_hashbase[
909 INP_PCBHASH(INADDR_ANY, lport, 0, pcbgroup->ipg_hashmask)];
910 LIST_FOREACH(inp, head, inp_pcbgrouphash) {
911 /* XXX inp locking */
912 if ((inp->inp_vflag & INP_IPV6) == 0)
915 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) ||
916 inp->inp_lport != lport) {
920 injail = prison_flag(inp->inp_cred, PR_IP6);
922 if (prison_check_ip6(inp->inp_cred,
926 if (local_exact != NULL)
930 if (IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr)) {
935 } else if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
955 * Then look for a wildcard match, if requested.
957 if ((lookupflags & INPLOOKUP_WILDCARD) != 0) {
958 struct inpcb *local_wild = NULL, *local_exact = NULL;
959 struct inpcb *jail_wild = NULL;
963 * Order of socket selection - we always prefer jails.
964 * 1. jailed, non-wild.
966 * 3. non-jailed, non-wild.
967 * 4. non-jailed, wild.
969 head = &pcbinfo->ipi_wildbase[INP_PCBHASH(
970 INP6_PCBHASHKEY(&in6addr_any), lport, 0,
971 pcbinfo->ipi_wildmask)];
972 LIST_FOREACH(inp, head, inp_pcbgroup_wild) {
973 /* XXX inp locking */
974 if ((inp->inp_vflag & INP_IPV6) == 0)
977 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) ||
978 inp->inp_lport != lport) {
982 injail = prison_flag(inp->inp_cred, PR_IP6);
984 if (prison_check_ip6(inp->inp_cred,
988 if (local_exact != NULL)
992 if (IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr)) {
997 } else if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
1003 } /* LIST_FOREACH */
1014 } /* if ((lookupflags & INPLOOKUP_WILDCARD) != 0) */
1015 INP_GROUP_UNLOCK(pcbgroup);
1020 INP_GROUP_UNLOCK(pcbgroup);
1021 if (lookupflags & INPLOOKUP_WLOCKPCB) {
1023 if (in_pcbrele_wlocked(inp))
1025 } else if (lookupflags & INPLOOKUP_RLOCKPCB) {
1027 if (in_pcbrele_rlocked(inp))
1030 panic("%s: locking buf", __func__);
1033 #endif /* PCBGROUP */
1036 * Lookup PCB in hash list.
1038 static struct inpcb *
1039 in6_pcblookup_hash_locked(struct inpcbinfo *pcbinfo, struct in6_addr *faddr,
1040 u_int fport_arg, struct in6_addr *laddr, u_int lport_arg,
1041 int lookupflags, struct ifnet *ifp)
1043 struct inpcbhead *head;
1044 struct inpcb *inp, *tmpinp;
1045 u_short fport = fport_arg, lport = lport_arg;
1047 KASSERT((lookupflags & ~(INPLOOKUP_WILDCARD)) == 0,
1048 ("%s: invalid lookup flags %d", __func__, lookupflags));
1050 INP_HASH_LOCK_ASSERT(pcbinfo);
1053 * First look for an exact match.
1056 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(
1057 INP6_PCBHASHKEY(faddr), lport, fport, pcbinfo->ipi_hashmask)];
1058 LIST_FOREACH(inp, head, inp_hash) {
1059 /* XXX inp locking */
1060 if ((inp->inp_vflag & INP_IPV6) == 0)
1062 if (IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, faddr) &&
1063 IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr) &&
1064 inp->inp_fport == fport &&
1065 inp->inp_lport == lport) {
1067 * XXX We should be able to directly return
1068 * the inp here, without any checks.
1069 * Well unless both bound with SO_REUSEPORT?
1071 if (prison_flag(inp->inp_cred, PR_IP6))
1081 * Then look for a wildcard match, if requested.
1083 if ((lookupflags & INPLOOKUP_WILDCARD) != 0) {
1084 struct inpcb *local_wild = NULL, *local_exact = NULL;
1085 struct inpcb *jail_wild = NULL;
1089 * Order of socket selection - we always prefer jails.
1090 * 1. jailed, non-wild.
1092 * 3. non-jailed, non-wild.
1093 * 4. non-jailed, wild.
1095 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(
1096 INP6_PCBHASHKEY(&in6addr_any), lport, 0,
1097 pcbinfo->ipi_hashmask)];
1098 LIST_FOREACH(inp, head, inp_hash) {
1099 /* XXX inp locking */
1100 if ((inp->inp_vflag & INP_IPV6) == 0)
1103 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) ||
1104 inp->inp_lport != lport) {
1108 injail = prison_flag(inp->inp_cred, PR_IP6);
1110 if (prison_check_ip6(inp->inp_cred,
1114 if (local_exact != NULL)
1118 if (IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr)) {
1123 } else if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
1129 } /* LIST_FOREACH */
1131 if (jail_wild != NULL)
1133 if (local_exact != NULL)
1134 return (local_exact);
1135 if (local_wild != NULL)
1136 return (local_wild);
1137 } /* if ((lookupflags & INPLOOKUP_WILDCARD) != 0) */
1146 * Lookup PCB in hash list, using pcbinfo tables. This variation locks the
1147 * hash list lock, and will return the inpcb locked (i.e., requires
1148 * INPLOOKUP_LOCKPCB).
1150 static struct inpcb *
1151 in6_pcblookup_hash(struct inpcbinfo *pcbinfo, struct in6_addr *faddr,
1152 u_int fport, struct in6_addr *laddr, u_int lport, int lookupflags,
1157 INP_HASH_RLOCK(pcbinfo);
1158 inp = in6_pcblookup_hash_locked(pcbinfo, faddr, fport, laddr, lport,
1159 (lookupflags & ~(INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)), ifp);
1162 INP_HASH_RUNLOCK(pcbinfo);
1163 if (lookupflags & INPLOOKUP_WLOCKPCB) {
1165 if (in_pcbrele_wlocked(inp))
1167 } else if (lookupflags & INPLOOKUP_RLOCKPCB) {
1169 if (in_pcbrele_rlocked(inp))
1172 panic("%s: locking bug", __func__);
1174 INP_HASH_RUNLOCK(pcbinfo);
1179 * Public inpcb lookup routines, accepting a 4-tuple, and optionally, an mbuf
1180 * from which a pre-calculated hash value may be extracted.
1182 * Possibly more of this logic should be in in6_pcbgroup.c.
1185 in6_pcblookup(struct inpcbinfo *pcbinfo, struct in6_addr *faddr, u_int fport,
1186 struct in6_addr *laddr, u_int lport, int lookupflags, struct ifnet *ifp)
1188 #if defined(PCBGROUP) && !defined(RSS)
1189 struct inpcbgroup *pcbgroup;
1192 KASSERT((lookupflags & ~INPLOOKUP_MASK) == 0,
1193 ("%s: invalid lookup flags %d", __func__, lookupflags));
1194 KASSERT((lookupflags & (INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)) != 0,
1195 ("%s: LOCKPCB not set", __func__));
1198 * When not using RSS, use connection groups in preference to the
1199 * reservation table when looking up 4-tuples. When using RSS, just
1200 * use the reservation table, due to the cost of the Toeplitz hash
1203 * XXXRW: This policy belongs in the pcbgroup code, as in principle
1204 * we could be doing RSS with a non-Toeplitz hash that is affordable
1207 #if defined(PCBGROUP) && !defined(RSS)
1208 if (in_pcbgroup_enabled(pcbinfo)) {
1209 pcbgroup = in6_pcbgroup_bytuple(pcbinfo, laddr, lport, faddr,
1211 return (in6_pcblookup_group(pcbinfo, pcbgroup, faddr, fport,
1212 laddr, lport, lookupflags, ifp));
1215 return (in6_pcblookup_hash(pcbinfo, faddr, fport, laddr, lport,
1220 in6_pcblookup_mbuf(struct inpcbinfo *pcbinfo, struct in6_addr *faddr,
1221 u_int fport, struct in6_addr *laddr, u_int lport, int lookupflags,
1222 struct ifnet *ifp, struct mbuf *m)
1225 struct inpcbgroup *pcbgroup;
1228 KASSERT((lookupflags & ~INPLOOKUP_MASK) == 0,
1229 ("%s: invalid lookup flags %d", __func__, lookupflags));
1230 KASSERT((lookupflags & (INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)) != 0,
1231 ("%s: LOCKPCB not set", __func__));
1235 * If we can use a hardware-generated hash to look up the connection
1236 * group, use that connection group to find the inpcb. Otherwise
1237 * fall back on a software hash -- or the reservation table if we're
1240 * XXXRW: As above, that policy belongs in the pcbgroup code.
1242 if (in_pcbgroup_enabled(pcbinfo) &&
1243 M_HASHTYPE_TEST(m, M_HASHTYPE_NONE) == 0) {
1244 pcbgroup = in6_pcbgroup_byhash(pcbinfo, M_HASHTYPE_GET(m),
1245 m->m_pkthdr.flowid);
1246 if (pcbgroup != NULL)
1247 return (in6_pcblookup_group(pcbinfo, pcbgroup, faddr,
1248 fport, laddr, lport, lookupflags, ifp));
1250 pcbgroup = in6_pcbgroup_bytuple(pcbinfo, laddr, lport, faddr,
1252 return (in6_pcblookup_group(pcbinfo, pcbgroup, faddr, fport,
1253 laddr, lport, lookupflags, ifp));
1257 return (in6_pcblookup_hash(pcbinfo, faddr, fport, laddr, lport,
1262 init_sin6(struct sockaddr_in6 *sin6, struct mbuf *m)
1266 ip = mtod(m, struct ip6_hdr *);
1267 bzero(sin6, sizeof(*sin6));
1268 sin6->sin6_len = sizeof(*sin6);
1269 sin6->sin6_family = AF_INET6;
1270 sin6->sin6_addr = ip->ip6_src;
1272 (void)sa6_recoverscope(sin6); /* XXX: should catch errors... */