2 * SPDX-License-Identifier: BSD-3-Clause
4 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
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 * 3. Neither the name of the project 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 PROJECT 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 PROJECT 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 * $KAME: in6_pcb.c,v 1.31 2001/05/21 05:45:10 jinmei Exp $
39 * Copyright (c) 1982, 1986, 1991, 1993
40 * The Regents of the University of California. All rights reserved.
42 * Redistribution and use in source and binary forms, with or without
43 * modification, are permitted provided that the following conditions
45 * 1. Redistributions of source code must retain the above copyright
46 * notice, this list of conditions and the following disclaimer.
47 * 2. Redistributions in binary form must reproduce the above copyright
48 * notice, this list of conditions and the following disclaimer in the
49 * documentation and/or other materials provided with the distribution.
50 * 3. Neither the name of the University nor the names of its contributors
51 * may be used to endorse or promote products derived from this software
52 * without specific prior written permission.
54 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
55 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
56 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
57 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
58 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
59 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
60 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
61 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
62 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
63 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
66 * @(#)in_pcb.c 8.2 (Berkeley) 1/4/94
69 #include <sys/cdefs.h>
70 __FBSDID("$FreeBSD$");
73 #include "opt_inet6.h"
74 #include "opt_ipsec.h"
75 #include "opt_pcbgroup.h"
78 #include <sys/param.h>
79 #include <sys/systm.h>
80 #include <sys/malloc.h>
82 #include <sys/domain.h>
83 #include <sys/protosw.h>
84 #include <sys/socket.h>
85 #include <sys/socketvar.h>
86 #include <sys/sockio.h>
87 #include <sys/errno.h>
96 #include <net/if_var.h>
97 #include <net/if_llatbl.h>
98 #include <net/if_types.h>
99 #include <net/route.h>
101 #include <netinet/in.h>
102 #include <netinet/in_var.h>
103 #include <netinet/in_systm.h>
104 #include <netinet/tcp_var.h>
105 #include <netinet/ip6.h>
106 #include <netinet/ip_var.h>
108 #include <netinet6/ip6_var.h>
109 #include <netinet6/nd6.h>
110 #include <netinet/in_pcb.h>
111 #include <netinet6/in6_pcb.h>
112 #include <netinet6/scope6_var.h>
114 static struct inpcb *in6_pcblookup_hash_locked(struct inpcbinfo *,
115 struct in6_addr *, u_int, struct in6_addr *, u_int, int, struct ifnet *);
118 in6_pcbbind(struct inpcb *inp, struct sockaddr *nam,
121 struct socket *so = inp->inp_socket;
122 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)NULL;
123 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
125 int error, lookupflags = 0;
126 int reuseport = (so->so_options & SO_REUSEPORT);
129 * XXX: Maybe we could let SO_REUSEPORT_LB set SO_REUSEPORT bit here
130 * so that we don't have to add to the (already messy) code below.
132 int reuseport_lb = (so->so_options & SO_REUSEPORT_LB);
134 INP_WLOCK_ASSERT(inp);
135 INP_HASH_WLOCK_ASSERT(pcbinfo);
137 if (CK_STAILQ_EMPTY(&V_in6_ifaddrhead)) /* XXX broken! */
138 return (EADDRNOTAVAIL);
139 if (inp->inp_lport || !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr))
141 if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT|SO_REUSEPORT_LB)) == 0)
142 lookupflags = INPLOOKUP_WILDCARD;
144 if ((error = prison_local_ip6(cred, &inp->in6p_laddr,
145 ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0)
148 sin6 = (struct sockaddr_in6 *)nam;
149 if (nam->sa_len != sizeof(*sin6))
154 if (nam->sa_family != AF_INET6)
155 return (EAFNOSUPPORT);
157 if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0)
160 if ((error = prison_local_ip6(cred, &sin6->sin6_addr,
161 ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0)
164 lport = sin6->sin6_port;
165 if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
167 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
168 * allow compepte duplication of binding if
169 * SO_REUSEPORT is set, or if SO_REUSEADDR is set
170 * and a multicast address is bound on both
171 * new and duplicated sockets.
173 if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) != 0)
174 reuseport = SO_REUSEADDR|SO_REUSEPORT;
176 * XXX: How to deal with SO_REUSEPORT_LB here?
177 * Treat same as SO_REUSEPORT for now.
179 if ((so->so_options &
180 (SO_REUSEADDR|SO_REUSEPORT_LB)) != 0)
181 reuseport_lb = SO_REUSEADDR|SO_REUSEPORT_LB;
182 } else if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
185 sin6->sin6_port = 0; /* yech... */
187 if ((ifa = ifa_ifwithaddr((struct sockaddr *)sin6)) ==
189 (inp->inp_flags & INP_BINDANY) == 0) {
191 return (EADDRNOTAVAIL);
195 * XXX: bind to an anycast address might accidentally
196 * cause sending a packet with anycast source address.
197 * We should allow to bind to a deprecated address, since
198 * the application dares to use it.
201 ((struct in6_ifaddr *)ifa)->ia6_flags &
202 (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY|IN6_IFF_DETACHED)) {
204 return (EADDRNOTAVAIL);
213 if (ntohs(lport) <= V_ipport_reservedhigh &&
214 ntohs(lport) >= V_ipport_reservedlow &&
215 priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT))
217 if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr) &&
218 priv_check_cred(inp->inp_cred, PRIV_NETINET_REUSEPORT) != 0) {
219 t = in6_pcblookup_local(pcbinfo,
220 &sin6->sin6_addr, lport,
221 INPLOOKUP_WILDCARD, cred);
223 ((inp->inp_flags2 & INP_BINDMULTI) == 0) &&
224 ((t->inp_flags & INP_TIMEWAIT) == 0) &&
225 (so->so_type != SOCK_STREAM ||
226 IN6_IS_ADDR_UNSPECIFIED(&t->in6p_faddr)) &&
227 (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) ||
228 !IN6_IS_ADDR_UNSPECIFIED(&t->in6p_laddr) ||
229 (t->inp_flags2 & INP_REUSEPORT) ||
230 (t->inp_flags2 & INP_REUSEPORT_LB) == 0) &&
231 (inp->inp_cred->cr_uid !=
232 t->inp_cred->cr_uid))
236 * If the socket is a BINDMULTI socket, then
237 * the credentials need to match and the
238 * original socket also has to have been bound
241 if (t && (! in_pcbbind_check_bindmulti(inp, t)))
245 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0 &&
246 IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
247 struct sockaddr_in sin;
249 in6_sin6_2_sin(&sin, sin6);
250 t = in_pcblookup_local(pcbinfo,
252 INPLOOKUP_WILDCARD, cred);
254 ((inp->inp_flags2 & INP_BINDMULTI) == 0) &&
256 INP_TIMEWAIT) == 0) &&
257 (so->so_type != SOCK_STREAM ||
258 ntohl(t->inp_faddr.s_addr) ==
260 (inp->inp_cred->cr_uid !=
261 t->inp_cred->cr_uid))
264 if (t && (! in_pcbbind_check_bindmulti(inp, t)))
269 t = in6_pcblookup_local(pcbinfo, &sin6->sin6_addr,
270 lport, lookupflags, cred);
271 if (t && (t->inp_flags & INP_TIMEWAIT)) {
273 * XXXRW: If an incpb has had its timewait
274 * state recycled, we treat the address as
275 * being in use (for now). This is better
276 * than a panic, but not desirable.
280 ((reuseport & tw->tw_so_options) == 0 &&
281 (reuseport_lb & tw->tw_so_options) == 0))
283 } else if (t && (reuseport & inp_so_options(t)) == 0 &&
284 (reuseport_lb & inp_so_options(t)) == 0) {
288 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0 &&
289 IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
290 struct sockaddr_in sin;
292 in6_sin6_2_sin(&sin, sin6);
293 t = in_pcblookup_local(pcbinfo, sin.sin_addr,
294 lport, lookupflags, cred);
295 if (t && t->inp_flags & INP_TIMEWAIT) {
299 if ((reuseport & tw->tw_so_options) == 0
300 && (reuseport_lb & tw->tw_so_options) == 0
301 && (ntohl(t->inp_laddr.s_addr) !=
302 INADDR_ANY || ((inp->inp_vflag &
304 (t->inp_vflag & INP_IPV6PROTO))))
307 (reuseport & inp_so_options(t)) == 0 &&
308 (reuseport_lb & inp_so_options(t)) == 0 &&
309 (ntohl(t->inp_laddr.s_addr) != INADDR_ANY ||
310 (t->inp_vflag & INP_IPV6PROTO) != 0)) {
316 inp->in6p_laddr = sin6->sin6_addr;
319 if ((error = in6_pcbsetport(&inp->in6p_laddr, inp, cred)) != 0) {
320 /* Undo an address bind that may have occurred. */
321 inp->in6p_laddr = in6addr_any;
325 inp->inp_lport = lport;
326 if (in_pcbinshash(inp) != 0) {
327 inp->in6p_laddr = in6addr_any;
336 * Transform old in6_pcbconnect() into an inner subroutine for new
337 * in6_pcbconnect(): Do some validity-checking on the remote
338 * address (in mbuf 'nam') and then determine local host address
339 * (i.e., which interface) to use to access that remote host.
341 * This preserves definition of in6_pcbconnect(), while supporting a
342 * slightly different version for T/TCP. (This is more than
343 * a bit of a kludge, but cleaning up the internal interfaces would
344 * have forced minor changes in every protocol).
347 in6_pcbladdr(struct inpcb *inp, struct sockaddr *nam,
348 struct in6_addr *plocal_addr6)
350 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam;
352 int scope_ambiguous = 0;
353 struct in6_addr in6a;
355 INP_WLOCK_ASSERT(inp);
356 INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo); /* XXXRW: why? */
358 if (nam->sa_len != sizeof (*sin6))
360 if (sin6->sin6_family != AF_INET6)
361 return (EAFNOSUPPORT);
362 if (sin6->sin6_port == 0)
363 return (EADDRNOTAVAIL);
365 if (sin6->sin6_scope_id == 0 && !V_ip6_use_defzone)
367 if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0)
370 if (!CK_STAILQ_EMPTY(&V_in6_ifaddrhead)) {
372 * If the destination address is UNSPECIFIED addr,
373 * use the loopback addr, e.g ::1.
375 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))
376 sin6->sin6_addr = in6addr_loopback;
378 if ((error = prison_remote_ip6(inp->inp_cred, &sin6->sin6_addr)) != 0)
381 error = in6_selectsrc_socket(sin6, inp->in6p_outputopts,
382 inp, inp->inp_cred, scope_ambiguous, &in6a, NULL);
387 * Do not update this earlier, in case we return with an error.
389 * XXX: this in6_selectsrc_socket result might replace the bound local
390 * address with the address specified by setsockopt(IPV6_PKTINFO).
391 * Is it the intended behavior?
393 *plocal_addr6 = in6a;
396 * Don't do pcblookup call here; return interface in
398 * and exit to caller, that will do the lookup.
406 * Connect from a socket to a specified address.
407 * Both address and port must be specified in argument sin.
408 * If don't have a local address for this socket yet,
412 in6_pcbconnect_mbuf(struct inpcb *inp, struct sockaddr *nam,
413 struct ucred *cred, struct mbuf *m)
415 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
416 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam;
417 struct in6_addr addr6;
420 INP_WLOCK_ASSERT(inp);
421 INP_HASH_WLOCK_ASSERT(pcbinfo);
424 * Call inner routine, to assign local interface address.
425 * in6_pcbladdr() may automatically fill in sin6_scope_id.
427 if ((error = in6_pcbladdr(inp, nam, &addr6)) != 0)
430 if (in6_pcblookup_hash_locked(pcbinfo, &sin6->sin6_addr,
432 IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)
433 ? &addr6 : &inp->in6p_laddr,
434 inp->inp_lport, 0, NULL) != NULL) {
437 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
438 if (inp->inp_lport == 0) {
439 error = in6_pcbbind(inp, (struct sockaddr *)0, cred);
443 inp->in6p_laddr = addr6;
445 inp->in6p_faddr = sin6->sin6_addr;
446 inp->inp_fport = sin6->sin6_port;
447 /* update flowinfo - draft-itojun-ipv6-flowlabel-api-00 */
448 inp->inp_flow &= ~IPV6_FLOWLABEL_MASK;
449 if (inp->inp_flags & IN6P_AUTOFLOWLABEL)
451 (htonl(ip6_randomflowlabel()) & IPV6_FLOWLABEL_MASK);
453 in_pcbrehash_mbuf(inp, m);
459 in6_pcbconnect(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred)
462 return (in6_pcbconnect_mbuf(inp, nam, cred, NULL));
466 in6_pcbdisconnect(struct inpcb *inp)
469 INP_WLOCK_ASSERT(inp);
470 INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo);
472 bzero((caddr_t)&inp->in6p_faddr, sizeof(inp->in6p_faddr));
474 /* clear flowinfo - draft-itojun-ipv6-flowlabel-api-00 */
475 inp->inp_flow &= ~IPV6_FLOWLABEL_MASK;
480 in6_sockaddr(in_port_t port, struct in6_addr *addr_p)
482 struct sockaddr_in6 *sin6;
484 sin6 = malloc(sizeof *sin6, M_SONAME, M_WAITOK);
485 bzero(sin6, sizeof *sin6);
486 sin6->sin6_family = AF_INET6;
487 sin6->sin6_len = sizeof(*sin6);
488 sin6->sin6_port = port;
489 sin6->sin6_addr = *addr_p;
490 (void)sa6_recoverscope(sin6); /* XXX: should catch errors */
492 return (struct sockaddr *)sin6;
496 in6_v4mapsin6_sockaddr(in_port_t port, struct in_addr *addr_p)
498 struct sockaddr_in sin;
499 struct sockaddr_in6 *sin6_p;
501 bzero(&sin, sizeof sin);
502 sin.sin_family = AF_INET;
503 sin.sin_len = sizeof(sin);
505 sin.sin_addr = *addr_p;
507 sin6_p = malloc(sizeof *sin6_p, M_SONAME,
509 in6_sin_2_v4mapsin6(&sin, sin6_p);
511 return (struct sockaddr *)sin6_p;
515 in6_getsockaddr(struct socket *so, struct sockaddr **nam)
518 struct in6_addr addr;
522 KASSERT(inp != NULL, ("in6_getsockaddr: inp == NULL"));
525 port = inp->inp_lport;
526 addr = inp->in6p_laddr;
529 *nam = in6_sockaddr(port, &addr);
534 in6_getpeeraddr(struct socket *so, struct sockaddr **nam)
537 struct in6_addr addr;
541 KASSERT(inp != NULL, ("in6_getpeeraddr: inp == NULL"));
544 port = inp->inp_fport;
545 addr = inp->in6p_faddr;
548 *nam = in6_sockaddr(port, &addr);
553 in6_mapped_sockaddr(struct socket *so, struct sockaddr **nam)
559 KASSERT(inp != NULL, ("in6_mapped_sockaddr: inp == NULL"));
562 if ((inp->inp_vflag & (INP_IPV4 | INP_IPV6)) == INP_IPV4) {
563 error = in_getsockaddr(so, nam);
565 in6_sin_2_v4mapsin6_in_sock(nam);
569 /* scope issues will be handled in in6_getsockaddr(). */
570 error = in6_getsockaddr(so, nam);
577 in6_mapped_peeraddr(struct socket *so, struct sockaddr **nam)
583 KASSERT(inp != NULL, ("in6_mapped_peeraddr: inp == NULL"));
586 if ((inp->inp_vflag & (INP_IPV4 | INP_IPV6)) == INP_IPV4) {
587 error = in_getpeeraddr(so, nam);
589 in6_sin_2_v4mapsin6_in_sock(nam);
592 /* scope issues will be handled in in6_getpeeraddr(). */
593 error = in6_getpeeraddr(so, nam);
599 * Pass some notification to all connections of a protocol
600 * associated with address dst. The local address and/or port numbers
601 * may be specified to limit the search. The "usual action" will be
602 * taken, depending on the ctlinput cmd. The caller must filter any
603 * cmds that are uninteresting (e.g., no error in the map).
604 * Call the protocol specific routine (if any) to report
605 * any errors for each matching socket.
608 in6_pcbnotify(struct inpcbinfo *pcbinfo, struct sockaddr *dst,
609 u_int fport_arg, const struct sockaddr *src, u_int lport_arg,
610 int cmd, void *cmdarg,
611 struct inpcb *(*notify)(struct inpcb *, int))
613 struct inpcb *inp, *inp_temp;
614 struct sockaddr_in6 sa6_src, *sa6_dst;
615 u_short fport = fport_arg, lport = lport_arg;
619 if ((unsigned)cmd >= PRC_NCMDS || dst->sa_family != AF_INET6)
622 sa6_dst = (struct sockaddr_in6 *)dst;
623 if (IN6_IS_ADDR_UNSPECIFIED(&sa6_dst->sin6_addr))
627 * note that src can be NULL when we get notify by local fragmentation.
629 sa6_src = (src == NULL) ? sa6_any : *(const struct sockaddr_in6 *)src;
630 flowinfo = sa6_src.sin6_flowinfo;
633 * Redirects go to all references to the destination,
634 * and use in6_rtchange to invalidate the route cache.
635 * Dead host indications: also use in6_rtchange to invalidate
636 * the cache, and deliver the error to all the sockets.
637 * Otherwise, if we have knowledge of the local port and address,
638 * deliver only to that socket.
640 if (PRC_IS_REDIRECT(cmd) || cmd == PRC_HOSTDEAD) {
643 bzero((caddr_t)&sa6_src.sin6_addr, sizeof(sa6_src.sin6_addr));
645 if (cmd != PRC_HOSTDEAD)
646 notify = in6_rtchange;
648 errno = inet6ctlerrmap[cmd];
649 INP_INFO_WLOCK(pcbinfo);
650 CK_LIST_FOREACH_SAFE(inp, pcbinfo->ipi_listhead, inp_list, inp_temp) {
652 if ((inp->inp_vflag & INP_IPV6) == 0) {
658 * If the error designates a new path MTU for a destination
659 * and the application (associated with this socket) wanted to
660 * know the value, notify.
661 * XXX: should we avoid to notify the value to TCP sockets?
663 if (cmd == PRC_MSGSIZE && cmdarg != NULL)
664 ip6_notify_pmtu(inp, (struct sockaddr_in6 *)dst,
665 *(u_int32_t *)cmdarg);
668 * Detect if we should notify the error. If no source and
669 * destination ports are specifed, but non-zero flowinfo and
670 * local address match, notify the error. This is the case
671 * when the error is delivered with an encrypted buffer
672 * by ESP. Otherwise, just compare addresses and ports
675 if (lport == 0 && fport == 0 && flowinfo &&
676 inp->inp_socket != NULL &&
677 flowinfo == (inp->inp_flow & IPV6_FLOWLABEL_MASK) &&
678 IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, &sa6_src.sin6_addr))
680 else if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr,
681 &sa6_dst->sin6_addr) ||
682 inp->inp_socket == 0 ||
683 (lport && inp->inp_lport != lport) ||
684 (!IN6_IS_ADDR_UNSPECIFIED(&sa6_src.sin6_addr) &&
685 !IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr,
686 &sa6_src.sin6_addr)) ||
687 (fport && inp->inp_fport != fport)) {
694 if ((*notify)(inp, errno))
699 INP_INFO_WUNLOCK(pcbinfo);
703 * Lookup a PCB based on the local address and port. Caller must hold the
704 * hash lock. No inpcb locks or references are acquired.
707 in6_pcblookup_local(struct inpcbinfo *pcbinfo, struct in6_addr *laddr,
708 u_short lport, int lookupflags, struct ucred *cred)
711 int matchwild = 3, wildcard;
713 KASSERT((lookupflags & ~(INPLOOKUP_WILDCARD)) == 0,
714 ("%s: invalid lookup flags %d", __func__, lookupflags));
716 INP_HASH_WLOCK_ASSERT(pcbinfo);
718 if ((lookupflags & INPLOOKUP_WILDCARD) == 0) {
719 struct inpcbhead *head;
721 * Look for an unconnected (wildcard foreign addr) PCB that
722 * matches the local address and port we're looking for.
724 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(
725 INP6_PCBHASHKEY(&in6addr_any), lport, 0,
726 pcbinfo->ipi_hashmask)];
727 CK_LIST_FOREACH(inp, head, inp_hash) {
728 /* XXX inp locking */
729 if ((inp->inp_vflag & INP_IPV6) == 0)
731 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) &&
732 IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr) &&
733 inp->inp_lport == lport) {
736 prison_equal_ip6(cred->cr_prison,
737 inp->inp_cred->cr_prison))
746 struct inpcbporthead *porthash;
747 struct inpcbport *phd;
748 struct inpcb *match = NULL;
750 * Best fit PCB lookup.
752 * First see if this local port is in use by looking on the
755 porthash = &pcbinfo->ipi_porthashbase[INP_PCBPORTHASH(lport,
756 pcbinfo->ipi_porthashmask)];
757 CK_LIST_FOREACH(phd, porthash, phd_hash) {
758 if (phd->phd_port == lport)
763 * Port is in use by one or more PCBs. Look for best
766 CK_LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) {
769 !prison_equal_ip6(cred->cr_prison,
770 inp->inp_cred->cr_prison))
772 /* XXX inp locking */
773 if ((inp->inp_vflag & INP_IPV6) == 0)
775 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr))
777 if (!IN6_IS_ADDR_UNSPECIFIED(
779 if (IN6_IS_ADDR_UNSPECIFIED(laddr))
781 else if (!IN6_ARE_ADDR_EQUAL(
782 &inp->in6p_laddr, laddr))
785 if (!IN6_IS_ADDR_UNSPECIFIED(laddr))
788 if (wildcard < matchwild) {
790 matchwild = wildcard;
801 in6_pcbpurgeif0(struct inpcbinfo *pcbinfo, struct ifnet *ifp)
804 struct ip6_moptions *im6o;
807 INP_INFO_WLOCK(pcbinfo);
808 CK_LIST_FOREACH(in6p, pcbinfo->ipi_listhead, inp_list) {
810 if (__predict_false(in6p->inp_flags2 & INP_FREED)) {
814 im6o = in6p->in6p_moptions;
815 if ((in6p->inp_vflag & INP_IPV6) && im6o != NULL) {
817 * Unselect the outgoing ifp for multicast if it
820 if (im6o->im6o_multicast_ifp == ifp)
821 im6o->im6o_multicast_ifp = NULL;
823 * Drop multicast group membership if we joined
824 * through the interface being detached.
827 for (i = 0; i < im6o->im6o_num_memberships; i++) {
828 if (im6o->im6o_membership[i]->in6m_ifp ==
830 in6_leavegroup(im6o->im6o_membership[i], NULL);
832 } else if (gap != 0) {
833 im6o->im6o_membership[i - gap] =
834 im6o->im6o_membership[i];
837 im6o->im6o_num_memberships -= gap;
841 INP_INFO_WUNLOCK(pcbinfo);
845 * Check for alternatives when higher level complains
846 * about service problems. For now, invalidate cached
847 * routing information. If the route was created dynamically
848 * (by a redirect), time to try a default gateway again.
851 in6_losing(struct inpcb *inp)
854 RO_INVALIDATE_CACHE(&inp->inp_route6);
858 * After a routing change, flush old routing
859 * and allocate a (hopefully) better one.
862 in6_rtchange(struct inpcb *inp, int errno __unused)
865 RO_INVALIDATE_CACHE(&inp->inp_route6);
869 static struct inpcb *
870 in6_pcblookup_lbgroup(const struct inpcbinfo *pcbinfo,
871 const struct in6_addr *laddr, uint16_t lport, const struct in6_addr *faddr,
872 uint16_t fport, int lookupflags)
874 struct inpcb *local_wild;
875 const struct inpcblbgrouphead *hdr;
876 struct inpcblbgroup *grp;
879 INP_HASH_LOCK_ASSERT(pcbinfo);
881 hdr = &pcbinfo->ipi_lbgrouphashbase[
882 INP_PCBPORTHASH(lport, pcbinfo->ipi_lbgrouphashmask)];
885 * Order of socket selection:
887 * 2. wild (if lookupflags contains INPLOOKUP_WILDCARD).
890 * - Load balanced group does not contain jailed sockets.
891 * - Load balanced does not contain IPv4 mapped INET6 wild sockets.
894 CK_LIST_FOREACH(grp, hdr, il_list) {
896 if (!(grp->il_vflag & INP_IPV6))
899 if (grp->il_lport != lport)
902 idx = INP_PCBLBGROUP_PKTHASH(INP6_PCBHASHKEY(faddr), lport,
903 fport) % grp->il_inpcnt;
904 if (IN6_ARE_ADDR_EQUAL(&grp->il6_laddr, laddr))
905 return (grp->il_inp[idx]);
906 if (IN6_IS_ADDR_UNSPECIFIED(&grp->il6_laddr) &&
907 (lookupflags & INPLOOKUP_WILDCARD) != 0)
908 local_wild = grp->il_inp[idx];
915 * Lookup PCB in hash list, using pcbgroup tables.
917 static struct inpcb *
918 in6_pcblookup_group(struct inpcbinfo *pcbinfo, struct inpcbgroup *pcbgroup,
919 struct in6_addr *faddr, u_int fport_arg, struct in6_addr *laddr,
920 u_int lport_arg, int lookupflags, struct ifnet *ifp)
922 struct inpcbhead *head;
923 struct inpcb *inp, *tmpinp;
924 u_short fport = fport_arg, lport = lport_arg;
928 * First look for an exact match.
931 INP_GROUP_LOCK(pcbgroup);
932 head = &pcbgroup->ipg_hashbase[INP_PCBHASH(
933 INP6_PCBHASHKEY(faddr), lport, fport, pcbgroup->ipg_hashmask)];
934 CK_LIST_FOREACH(inp, head, inp_pcbgrouphash) {
935 /* XXX inp locking */
936 if ((inp->inp_vflag & INP_IPV6) == 0)
938 if (IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, faddr) &&
939 IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr) &&
940 inp->inp_fport == fport &&
941 inp->inp_lport == lport) {
943 * XXX We should be able to directly return
944 * the inp here, without any checks.
945 * Well unless both bound with SO_REUSEPORT?
947 if (prison_flag(inp->inp_cred, PR_IP6))
953 if (tmpinp != NULL) {
959 * Then look for a wildcard match in the pcbgroup.
961 if ((lookupflags & INPLOOKUP_WILDCARD) != 0) {
962 struct inpcb *local_wild = NULL, *local_exact = NULL;
963 struct inpcb *jail_wild = NULL;
967 * Order of socket selection - we always prefer jails.
968 * 1. jailed, non-wild.
970 * 3. non-jailed, non-wild.
971 * 4. non-jailed, wild.
973 head = &pcbgroup->ipg_hashbase[
974 INP_PCBHASH(INADDR_ANY, lport, 0, pcbgroup->ipg_hashmask)];
975 CK_LIST_FOREACH(inp, head, inp_pcbgrouphash) {
976 /* XXX inp locking */
977 if ((inp->inp_vflag & INP_IPV6) == 0)
980 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) ||
981 inp->inp_lport != lport) {
985 injail = prison_flag(inp->inp_cred, PR_IP6);
987 if (prison_check_ip6(inp->inp_cred,
991 if (local_exact != NULL)
995 if (IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr)) {
1000 } else if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
1006 } /* LIST_FOREACH */
1020 * Then look for a wildcard match, if requested.
1022 if ((lookupflags & INPLOOKUP_WILDCARD) != 0) {
1023 struct inpcb *local_wild = NULL, *local_exact = NULL;
1024 struct inpcb *jail_wild = NULL;
1028 * Order of socket selection - we always prefer jails.
1029 * 1. jailed, non-wild.
1031 * 3. non-jailed, non-wild.
1032 * 4. non-jailed, wild.
1034 head = &pcbinfo->ipi_wildbase[INP_PCBHASH(
1035 INP6_PCBHASHKEY(&in6addr_any), lport, 0,
1036 pcbinfo->ipi_wildmask)];
1037 CK_LIST_FOREACH(inp, head, inp_pcbgroup_wild) {
1038 /* XXX inp locking */
1039 if ((inp->inp_vflag & INP_IPV6) == 0)
1042 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) ||
1043 inp->inp_lport != lport) {
1047 injail = prison_flag(inp->inp_cred, PR_IP6);
1049 if (prison_check_ip6(inp->inp_cred,
1053 if (local_exact != NULL)
1057 if (IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr)) {
1062 } else if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
1068 } /* LIST_FOREACH */
1079 } /* if ((lookupflags & INPLOOKUP_WILDCARD) != 0) */
1080 INP_GROUP_UNLOCK(pcbgroup);
1084 if (lookupflags & INPLOOKUP_WLOCKPCB)
1085 locked = INP_TRY_WLOCK(inp);
1086 else if (lookupflags & INPLOOKUP_RLOCKPCB)
1087 locked = INP_TRY_RLOCK(inp);
1089 panic("%s: locking buf", __func__);
1092 INP_GROUP_UNLOCK(pcbgroup);
1094 if (lookupflags & INPLOOKUP_WLOCKPCB) {
1096 if (in_pcbrele_wlocked(inp))
1100 if (in_pcbrele_rlocked(inp))
1105 if (lookupflags & INPLOOKUP_WLOCKPCB)
1106 INP_WLOCK_ASSERT(inp);
1108 INP_RLOCK_ASSERT(inp);
1112 #endif /* PCBGROUP */
1115 * Lookup PCB in hash list.
1117 static struct inpcb *
1118 in6_pcblookup_hash_locked(struct inpcbinfo *pcbinfo, struct in6_addr *faddr,
1119 u_int fport_arg, struct in6_addr *laddr, u_int lport_arg,
1120 int lookupflags, struct ifnet *ifp)
1122 struct inpcbhead *head;
1123 struct inpcb *inp, *tmpinp;
1124 u_short fport = fport_arg, lport = lport_arg;
1126 KASSERT((lookupflags & ~(INPLOOKUP_WILDCARD)) == 0,
1127 ("%s: invalid lookup flags %d", __func__, lookupflags));
1129 INP_HASH_LOCK_ASSERT(pcbinfo);
1132 * First look for an exact match.
1135 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(
1136 INP6_PCBHASHKEY(faddr), lport, fport, pcbinfo->ipi_hashmask)];
1137 CK_LIST_FOREACH(inp, head, inp_hash) {
1138 /* XXX inp locking */
1139 if ((inp->inp_vflag & INP_IPV6) == 0)
1141 if (IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, faddr) &&
1142 IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr) &&
1143 inp->inp_fport == fport &&
1144 inp->inp_lport == lport) {
1146 * XXX We should be able to directly return
1147 * the inp here, without any checks.
1148 * Well unless both bound with SO_REUSEPORT?
1150 if (prison_flag(inp->inp_cred, PR_IP6))
1160 * Then look in lb group (for wildcard match).
1162 if ((lookupflags & INPLOOKUP_WILDCARD) != 0) {
1163 inp = in6_pcblookup_lbgroup(pcbinfo, laddr, lport, faddr,
1164 fport, lookupflags);
1170 * Then look for a wildcard match, if requested.
1172 if ((lookupflags & INPLOOKUP_WILDCARD) != 0) {
1173 struct inpcb *local_wild = NULL, *local_exact = NULL;
1174 struct inpcb *jail_wild = NULL;
1178 * Order of socket selection - we always prefer jails.
1179 * 1. jailed, non-wild.
1181 * 3. non-jailed, non-wild.
1182 * 4. non-jailed, wild.
1184 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(
1185 INP6_PCBHASHKEY(&in6addr_any), lport, 0,
1186 pcbinfo->ipi_hashmask)];
1187 CK_LIST_FOREACH(inp, head, inp_hash) {
1188 /* XXX inp locking */
1189 if ((inp->inp_vflag & INP_IPV6) == 0)
1192 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) ||
1193 inp->inp_lport != lport) {
1197 injail = prison_flag(inp->inp_cred, PR_IP6);
1199 if (prison_check_ip6(inp->inp_cred,
1203 if (local_exact != NULL)
1207 if (IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr)) {
1212 } else if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
1218 } /* LIST_FOREACH */
1220 if (jail_wild != NULL)
1222 if (local_exact != NULL)
1223 return (local_exact);
1224 if (local_wild != NULL)
1225 return (local_wild);
1226 } /* if ((lookupflags & INPLOOKUP_WILDCARD) != 0) */
1235 * Lookup PCB in hash list, using pcbinfo tables. This variation locks the
1236 * hash list lock, and will return the inpcb locked (i.e., requires
1237 * INPLOOKUP_LOCKPCB).
1239 static struct inpcb *
1240 in6_pcblookup_hash(struct inpcbinfo *pcbinfo, struct in6_addr *faddr,
1241 u_int fport, struct in6_addr *laddr, u_int lport, int lookupflags,
1246 INP_HASH_RLOCK(pcbinfo);
1247 inp = in6_pcblookup_hash_locked(pcbinfo, faddr, fport, laddr, lport,
1248 (lookupflags & ~(INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)), ifp);
1250 if (lookupflags & INPLOOKUP_WLOCKPCB) {
1252 if (__predict_false(inp->inp_flags2 & INP_FREED)) {
1256 } else if (lookupflags & INPLOOKUP_RLOCKPCB) {
1258 if (__predict_false(inp->inp_flags2 & INP_FREED)) {
1263 panic("%s: locking bug", __func__);
1266 if (lookupflags & INPLOOKUP_WLOCKPCB)
1267 INP_WLOCK_ASSERT(inp);
1269 INP_RLOCK_ASSERT(inp);
1273 INP_HASH_RUNLOCK(pcbinfo);
1278 * Public inpcb lookup routines, accepting a 4-tuple, and optionally, an mbuf
1279 * from which a pre-calculated hash value may be extracted.
1281 * Possibly more of this logic should be in in6_pcbgroup.c.
1284 in6_pcblookup(struct inpcbinfo *pcbinfo, struct in6_addr *faddr, u_int fport,
1285 struct in6_addr *laddr, u_int lport, int lookupflags, struct ifnet *ifp)
1287 #if defined(PCBGROUP) && !defined(RSS)
1288 struct inpcbgroup *pcbgroup;
1291 KASSERT((lookupflags & ~INPLOOKUP_MASK) == 0,
1292 ("%s: invalid lookup flags %d", __func__, lookupflags));
1293 KASSERT((lookupflags & (INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)) != 0,
1294 ("%s: LOCKPCB not set", __func__));
1297 * When not using RSS, use connection groups in preference to the
1298 * reservation table when looking up 4-tuples. When using RSS, just
1299 * use the reservation table, due to the cost of the Toeplitz hash
1302 * XXXRW: This policy belongs in the pcbgroup code, as in principle
1303 * we could be doing RSS with a non-Toeplitz hash that is affordable
1306 #if defined(PCBGROUP) && !defined(RSS)
1307 if (in_pcbgroup_enabled(pcbinfo)) {
1308 pcbgroup = in6_pcbgroup_bytuple(pcbinfo, laddr, lport, faddr,
1310 return (in6_pcblookup_group(pcbinfo, pcbgroup, faddr, fport,
1311 laddr, lport, lookupflags, ifp));
1314 return (in6_pcblookup_hash(pcbinfo, faddr, fport, laddr, lport,
1319 in6_pcblookup_mbuf(struct inpcbinfo *pcbinfo, struct in6_addr *faddr,
1320 u_int fport, struct in6_addr *laddr, u_int lport, int lookupflags,
1321 struct ifnet *ifp, struct mbuf *m)
1324 struct inpcbgroup *pcbgroup;
1327 KASSERT((lookupflags & ~INPLOOKUP_MASK) == 0,
1328 ("%s: invalid lookup flags %d", __func__, lookupflags));
1329 KASSERT((lookupflags & (INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)) != 0,
1330 ("%s: LOCKPCB not set", __func__));
1334 * If we can use a hardware-generated hash to look up the connection
1335 * group, use that connection group to find the inpcb. Otherwise
1336 * fall back on a software hash -- or the reservation table if we're
1339 * XXXRW: As above, that policy belongs in the pcbgroup code.
1341 if (in_pcbgroup_enabled(pcbinfo) &&
1342 M_HASHTYPE_TEST(m, M_HASHTYPE_NONE) == 0) {
1343 pcbgroup = in6_pcbgroup_byhash(pcbinfo, M_HASHTYPE_GET(m),
1344 m->m_pkthdr.flowid);
1345 if (pcbgroup != NULL)
1346 return (in6_pcblookup_group(pcbinfo, pcbgroup, faddr,
1347 fport, laddr, lport, lookupflags, ifp));
1349 pcbgroup = in6_pcbgroup_bytuple(pcbinfo, laddr, lport, faddr,
1351 return (in6_pcblookup_group(pcbinfo, pcbgroup, faddr, fport,
1352 laddr, lport, lookupflags, ifp));
1356 return (in6_pcblookup_hash(pcbinfo, faddr, fport, laddr, lport,
1361 init_sin6(struct sockaddr_in6 *sin6, struct mbuf *m, int srcordst)
1365 ip = mtod(m, struct ip6_hdr *);
1366 bzero(sin6, sizeof(*sin6));
1367 sin6->sin6_len = sizeof(*sin6);
1368 sin6->sin6_family = AF_INET6;
1369 sin6->sin6_addr = srcordst ? ip->ip6_dst : ip->ip6_src;
1371 (void)sa6_recoverscope(sin6); /* XXX: should catch errors... */