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"
76 #include "opt_route.h"
79 #include <sys/param.h>
80 #include <sys/systm.h>
81 #include <sys/malloc.h>
83 #include <sys/domain.h>
84 #include <sys/protosw.h>
85 #include <sys/socket.h>
86 #include <sys/socketvar.h>
87 #include <sys/sockio.h>
88 #include <sys/errno.h>
97 #include <net/if_var.h>
98 #include <net/if_llatbl.h>
99 #include <net/if_types.h>
100 #include <net/route.h>
101 #include <net/route/nhop.h>
103 #include <netinet/in.h>
104 #include <netinet/in_var.h>
105 #include <netinet/in_systm.h>
106 #include <netinet/tcp_var.h>
107 #include <netinet/ip6.h>
108 #include <netinet/ip_var.h>
110 #include <netinet6/ip6_var.h>
111 #include <netinet6/nd6.h>
112 #include <netinet/in_pcb.h>
113 #include <netinet6/in6_pcb.h>
114 #include <netinet6/in6_fib.h>
115 #include <netinet6/scope6_var.h>
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)) {
183 struct epoch_tracker et;
186 sin6->sin6_port = 0; /* yech... */
188 if ((ifa = ifa_ifwithaddr((struct sockaddr *)sin6)) ==
190 (inp->inp_flags & INP_BINDANY) == 0) {
192 return (EADDRNOTAVAIL);
196 * XXX: bind to an anycast address might accidentally
197 * cause sending a packet with anycast source address.
198 * We should allow to bind to a deprecated address, since
199 * the application dares to use it.
202 ((struct in6_ifaddr *)ifa)->ia6_flags &
203 (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY|IN6_IFF_DETACHED)) {
205 return (EADDRNOTAVAIL);
214 if (ntohs(lport) <= V_ipport_reservedhigh &&
215 ntohs(lport) >= V_ipport_reservedlow &&
216 priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT))
218 if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr) &&
219 priv_check_cred(inp->inp_cred, PRIV_NETINET_REUSEPORT) != 0) {
220 t = in6_pcblookup_local(pcbinfo,
221 &sin6->sin6_addr, lport,
222 INPLOOKUP_WILDCARD, cred);
224 ((inp->inp_flags2 & INP_BINDMULTI) == 0) &&
225 ((t->inp_flags & INP_TIMEWAIT) == 0) &&
226 (so->so_type != SOCK_STREAM ||
227 IN6_IS_ADDR_UNSPECIFIED(&t->in6p_faddr)) &&
228 (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) ||
229 !IN6_IS_ADDR_UNSPECIFIED(&t->in6p_laddr) ||
230 (t->inp_flags2 & INP_REUSEPORT) ||
231 (t->inp_flags2 & INP_REUSEPORT_LB) == 0) &&
232 (inp->inp_cred->cr_uid !=
233 t->inp_cred->cr_uid))
237 * If the socket is a BINDMULTI socket, then
238 * the credentials need to match and the
239 * original socket also has to have been bound
242 if (t && (! in_pcbbind_check_bindmulti(inp, t)))
246 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0 &&
247 IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
248 struct sockaddr_in sin;
250 in6_sin6_2_sin(&sin, sin6);
251 t = in_pcblookup_local(pcbinfo,
253 INPLOOKUP_WILDCARD, cred);
255 ((inp->inp_flags2 & INP_BINDMULTI) == 0) &&
257 INP_TIMEWAIT) == 0) &&
258 (so->so_type != SOCK_STREAM ||
259 ntohl(t->inp_faddr.s_addr) ==
261 (inp->inp_cred->cr_uid !=
262 t->inp_cred->cr_uid))
265 if (t && (! in_pcbbind_check_bindmulti(inp, t)))
270 t = in6_pcblookup_local(pcbinfo, &sin6->sin6_addr,
271 lport, lookupflags, cred);
272 if (t && (t->inp_flags & INP_TIMEWAIT)) {
274 * XXXRW: If an incpb has had its timewait
275 * state recycled, we treat the address as
276 * being in use (for now). This is better
277 * than a panic, but not desirable.
281 ((reuseport & tw->tw_so_options) == 0 &&
282 (reuseport_lb & tw->tw_so_options) == 0))
284 } else if (t && (reuseport & inp_so_options(t)) == 0 &&
285 (reuseport_lb & inp_so_options(t)) == 0) {
289 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0 &&
290 IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
291 struct sockaddr_in sin;
293 in6_sin6_2_sin(&sin, sin6);
294 t = in_pcblookup_local(pcbinfo, sin.sin_addr,
295 lport, lookupflags, cred);
296 if (t && t->inp_flags & INP_TIMEWAIT) {
300 if ((reuseport & tw->tw_so_options) == 0
301 && (reuseport_lb & tw->tw_so_options) == 0
302 && (ntohl(t->inp_laddr.s_addr) !=
303 INADDR_ANY || ((inp->inp_vflag &
305 (t->inp_vflag & INP_IPV6PROTO))))
308 (reuseport & inp_so_options(t)) == 0 &&
309 (reuseport_lb & inp_so_options(t)) == 0 &&
310 (ntohl(t->inp_laddr.s_addr) != INADDR_ANY ||
311 (t->inp_vflag & INP_IPV6PROTO) != 0)) {
317 inp->in6p_laddr = sin6->sin6_addr;
320 if ((error = in6_pcbsetport(&inp->in6p_laddr, inp, cred)) != 0) {
321 /* Undo an address bind that may have occurred. */
322 inp->in6p_laddr = in6addr_any;
326 inp->inp_lport = lport;
327 if (in_pcbinshash(inp) != 0) {
328 inp->in6p_laddr = in6addr_any;
337 * Transform old in6_pcbconnect() into an inner subroutine for new
338 * in6_pcbconnect(): Do some validity-checking on the remote
339 * address (in mbuf 'nam') and then determine local host address
340 * (i.e., which interface) to use to access that remote host.
342 * This preserves definition of in6_pcbconnect(), while supporting a
343 * slightly different version for T/TCP. (This is more than
344 * a bit of a kludge, but cleaning up the internal interfaces would
345 * have forced minor changes in every protocol).
348 in6_pcbladdr(struct inpcb *inp, struct sockaddr *nam,
349 struct in6_addr *plocal_addr6)
351 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam;
353 int scope_ambiguous = 0;
354 struct in6_addr in6a;
355 struct epoch_tracker et;
357 INP_WLOCK_ASSERT(inp);
358 INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo); /* XXXRW: why? */
360 if (nam->sa_len != sizeof (*sin6))
362 if (sin6->sin6_family != AF_INET6)
363 return (EAFNOSUPPORT);
364 if (sin6->sin6_port == 0)
365 return (EADDRNOTAVAIL);
367 if (sin6->sin6_scope_id == 0 && !V_ip6_use_defzone)
369 if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0)
372 if (!CK_STAILQ_EMPTY(&V_in6_ifaddrhead)) {
374 * If the destination address is UNSPECIFIED addr,
375 * use the loopback addr, e.g ::1.
377 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))
378 sin6->sin6_addr = in6addr_loopback;
380 if ((error = prison_remote_ip6(inp->inp_cred, &sin6->sin6_addr)) != 0)
384 error = in6_selectsrc_socket(sin6, inp->in6p_outputopts,
385 inp, inp->inp_cred, scope_ambiguous, &in6a, NULL);
391 * Do not update this earlier, in case we return with an error.
393 * XXX: this in6_selectsrc_socket result might replace the bound local
394 * address with the address specified by setsockopt(IPV6_PKTINFO).
395 * Is it the intended behavior?
397 *plocal_addr6 = in6a;
400 * Don't do pcblookup call here; return interface in
402 * and exit to caller, that will do the lookup.
410 * Connect from a socket to a specified address.
411 * Both address and port must be specified in argument sin.
412 * If don't have a local address for this socket yet,
416 in6_pcbconnect_mbuf(struct inpcb *inp, struct sockaddr *nam,
417 struct ucred *cred, struct mbuf *m, bool rehash)
419 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
420 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam;
421 struct sockaddr_in6 laddr6;
424 bzero(&laddr6, sizeof(laddr6));
425 laddr6.sin6_family = AF_INET6;
427 INP_WLOCK_ASSERT(inp);
428 INP_HASH_WLOCK_ASSERT(pcbinfo);
431 if (CALC_FLOWID_OUTBOUND) {
432 uint32_t hash_type, hash_val;
434 hash_val = fib6_calc_software_hash(&inp->in6p_laddr,
435 &sin6->sin6_addr, 0, sin6->sin6_port,
436 inp->inp_socket->so_proto->pr_protocol, &hash_type);
437 inp->inp_flowid = hash_val;
438 inp->inp_flowtype = hash_type;
442 * Call inner routine, to assign local interface address.
443 * in6_pcbladdr() may automatically fill in sin6_scope_id.
445 if ((error = in6_pcbladdr(inp, nam, &laddr6.sin6_addr)) != 0)
448 if (in6_pcblookup_hash_locked(pcbinfo, &sin6->sin6_addr,
450 IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)
451 ? &laddr6.sin6_addr : &inp->in6p_laddr,
452 inp->inp_lport, 0, NULL, M_NODOM) != NULL) {
455 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
456 if (inp->inp_lport == 0) {
458 * rehash was required to be true in the past for
459 * this case; retain that convention. However,
460 * we now call in_pcb_lport_dest rather than
461 * in6_pcbbind; the former does not insert into
462 * the hash table, the latter does. Change rehash
463 * to false to do the in_pcbinshash below.
465 KASSERT(rehash == true,
466 ("Rehashing required for unbound inps"));
468 error = in_pcb_lport_dest(inp,
469 (struct sockaddr *) &laddr6, &inp->inp_lport,
470 (struct sockaddr *) sin6, sin6->sin6_port, cred,
475 inp->in6p_laddr = laddr6.sin6_addr;
477 inp->in6p_faddr = sin6->sin6_addr;
478 inp->inp_fport = sin6->sin6_port;
479 /* update flowinfo - draft-itojun-ipv6-flowlabel-api-00 */
480 inp->inp_flow &= ~IPV6_FLOWLABEL_MASK;
481 if (inp->inp_flags & IN6P_AUTOFLOWLABEL)
483 (htonl(ip6_randomflowlabel()) & IPV6_FLOWLABEL_MASK);
486 in_pcbrehash_mbuf(inp, m);
488 in_pcbinshash_mbuf(inp, m);
495 in6_pcbconnect(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred)
498 return (in6_pcbconnect_mbuf(inp, nam, cred, NULL, true));
502 in6_pcbdisconnect(struct inpcb *inp)
505 INP_WLOCK_ASSERT(inp);
506 INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo);
508 bzero((caddr_t)&inp->in6p_faddr, sizeof(inp->in6p_faddr));
510 /* clear flowinfo - draft-itojun-ipv6-flowlabel-api-00 */
511 inp->inp_flow &= ~IPV6_FLOWLABEL_MASK;
516 in6_sockaddr(in_port_t port, struct in6_addr *addr_p)
518 struct sockaddr_in6 *sin6;
520 sin6 = malloc(sizeof *sin6, M_SONAME, M_WAITOK);
521 bzero(sin6, sizeof *sin6);
522 sin6->sin6_family = AF_INET6;
523 sin6->sin6_len = sizeof(*sin6);
524 sin6->sin6_port = port;
525 sin6->sin6_addr = *addr_p;
526 (void)sa6_recoverscope(sin6); /* XXX: should catch errors */
528 return (struct sockaddr *)sin6;
532 in6_v4mapsin6_sockaddr(in_port_t port, struct in_addr *addr_p)
534 struct sockaddr_in sin;
535 struct sockaddr_in6 *sin6_p;
537 bzero(&sin, sizeof sin);
538 sin.sin_family = AF_INET;
539 sin.sin_len = sizeof(sin);
541 sin.sin_addr = *addr_p;
543 sin6_p = malloc(sizeof *sin6_p, M_SONAME,
545 in6_sin_2_v4mapsin6(&sin, sin6_p);
547 return (struct sockaddr *)sin6_p;
551 in6_getsockaddr(struct socket *so, struct sockaddr **nam)
554 struct in6_addr addr;
558 KASSERT(inp != NULL, ("in6_getsockaddr: inp == NULL"));
561 port = inp->inp_lport;
562 addr = inp->in6p_laddr;
565 *nam = in6_sockaddr(port, &addr);
570 in6_getpeeraddr(struct socket *so, struct sockaddr **nam)
573 struct in6_addr addr;
577 KASSERT(inp != NULL, ("in6_getpeeraddr: inp == NULL"));
580 port = inp->inp_fport;
581 addr = inp->in6p_faddr;
584 *nam = in6_sockaddr(port, &addr);
589 in6_mapped_sockaddr(struct socket *so, struct sockaddr **nam)
595 KASSERT(inp != NULL, ("in6_mapped_sockaddr: inp == NULL"));
598 if ((inp->inp_vflag & (INP_IPV4 | INP_IPV6)) == INP_IPV4) {
599 error = in_getsockaddr(so, nam);
601 in6_sin_2_v4mapsin6_in_sock(nam);
605 /* scope issues will be handled in in6_getsockaddr(). */
606 error = in6_getsockaddr(so, nam);
613 in6_mapped_peeraddr(struct socket *so, struct sockaddr **nam)
619 KASSERT(inp != NULL, ("in6_mapped_peeraddr: inp == NULL"));
622 if ((inp->inp_vflag & (INP_IPV4 | INP_IPV6)) == INP_IPV4) {
623 error = in_getpeeraddr(so, nam);
625 in6_sin_2_v4mapsin6_in_sock(nam);
628 /* scope issues will be handled in in6_getpeeraddr(). */
629 error = in6_getpeeraddr(so, nam);
635 * Pass some notification to all connections of a protocol
636 * associated with address dst. The local address and/or port numbers
637 * may be specified to limit the search. The "usual action" will be
638 * taken, depending on the ctlinput cmd. The caller must filter any
639 * cmds that are uninteresting (e.g., no error in the map).
640 * Call the protocol specific routine (if any) to report
641 * any errors for each matching socket.
644 in6_pcbnotify(struct inpcbinfo *pcbinfo, struct sockaddr *dst,
645 u_int fport_arg, const struct sockaddr *src, u_int lport_arg,
646 int cmd, void *cmdarg,
647 struct inpcb *(*notify)(struct inpcb *, int))
649 struct inpcb *inp, *inp_temp;
650 struct sockaddr_in6 sa6_src, *sa6_dst;
651 u_short fport = fport_arg, lport = lport_arg;
655 if ((unsigned)cmd >= PRC_NCMDS || dst->sa_family != AF_INET6)
658 sa6_dst = (struct sockaddr_in6 *)dst;
659 if (IN6_IS_ADDR_UNSPECIFIED(&sa6_dst->sin6_addr))
663 * note that src can be NULL when we get notify by local fragmentation.
665 sa6_src = (src == NULL) ? sa6_any : *(const struct sockaddr_in6 *)src;
666 flowinfo = sa6_src.sin6_flowinfo;
669 * Redirects go to all references to the destination,
670 * and use in6_rtchange to invalidate the route cache.
671 * Dead host indications: also use in6_rtchange to invalidate
672 * the cache, and deliver the error to all the sockets.
673 * Otherwise, if we have knowledge of the local port and address,
674 * deliver only to that socket.
676 if (PRC_IS_REDIRECT(cmd) || cmd == PRC_HOSTDEAD) {
679 bzero((caddr_t)&sa6_src.sin6_addr, sizeof(sa6_src.sin6_addr));
681 if (cmd != PRC_HOSTDEAD)
682 notify = in6_rtchange;
684 errno = inet6ctlerrmap[cmd];
685 INP_INFO_WLOCK(pcbinfo);
686 CK_LIST_FOREACH_SAFE(inp, pcbinfo->ipi_listhead, inp_list, inp_temp) {
688 if ((inp->inp_vflag & INP_IPV6) == 0) {
694 * If the error designates a new path MTU for a destination
695 * and the application (associated with this socket) wanted to
696 * know the value, notify.
697 * XXX: should we avoid to notify the value to TCP sockets?
699 if (cmd == PRC_MSGSIZE && cmdarg != NULL)
700 ip6_notify_pmtu(inp, (struct sockaddr_in6 *)dst,
701 *(u_int32_t *)cmdarg);
704 * Detect if we should notify the error. If no source and
705 * destination ports are specifed, but non-zero flowinfo and
706 * local address match, notify the error. This is the case
707 * when the error is delivered with an encrypted buffer
708 * by ESP. Otherwise, just compare addresses and ports
711 if (lport == 0 && fport == 0 && flowinfo &&
712 inp->inp_socket != NULL &&
713 flowinfo == (inp->inp_flow & IPV6_FLOWLABEL_MASK) &&
714 IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, &sa6_src.sin6_addr))
716 else if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr,
717 &sa6_dst->sin6_addr) ||
718 inp->inp_socket == 0 ||
719 (lport && inp->inp_lport != lport) ||
720 (!IN6_IS_ADDR_UNSPECIFIED(&sa6_src.sin6_addr) &&
721 !IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr,
722 &sa6_src.sin6_addr)) ||
723 (fport && inp->inp_fport != fport)) {
730 if ((*notify)(inp, errno))
735 INP_INFO_WUNLOCK(pcbinfo);
739 * Lookup a PCB based on the local address and port. Caller must hold the
740 * hash lock. No inpcb locks or references are acquired.
743 in6_pcblookup_local(struct inpcbinfo *pcbinfo, struct in6_addr *laddr,
744 u_short lport, int lookupflags, struct ucred *cred)
747 int matchwild = 3, wildcard;
749 KASSERT((lookupflags & ~(INPLOOKUP_WILDCARD)) == 0,
750 ("%s: invalid lookup flags %d", __func__, lookupflags));
752 INP_HASH_LOCK_ASSERT(pcbinfo);
754 if ((lookupflags & INPLOOKUP_WILDCARD) == 0) {
755 struct inpcbhead *head;
757 * Look for an unconnected (wildcard foreign addr) PCB that
758 * matches the local address and port we're looking for.
760 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(
761 INP6_PCBHASHKEY(&in6addr_any), lport, 0,
762 pcbinfo->ipi_hashmask)];
763 CK_LIST_FOREACH(inp, head, inp_hash) {
764 /* XXX inp locking */
765 if ((inp->inp_vflag & INP_IPV6) == 0)
767 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) &&
768 IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr) &&
769 inp->inp_lport == lport) {
772 prison_equal_ip6(cred->cr_prison,
773 inp->inp_cred->cr_prison))
782 struct inpcbporthead *porthash;
783 struct inpcbport *phd;
784 struct inpcb *match = NULL;
786 * Best fit PCB lookup.
788 * First see if this local port is in use by looking on the
791 porthash = &pcbinfo->ipi_porthashbase[INP_PCBPORTHASH(lport,
792 pcbinfo->ipi_porthashmask)];
793 CK_LIST_FOREACH(phd, porthash, phd_hash) {
794 if (phd->phd_port == lport)
799 * Port is in use by one or more PCBs. Look for best
802 CK_LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) {
805 !prison_equal_ip6(cred->cr_prison,
806 inp->inp_cred->cr_prison))
808 /* XXX inp locking */
809 if ((inp->inp_vflag & INP_IPV6) == 0)
811 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr))
813 if (!IN6_IS_ADDR_UNSPECIFIED(
815 if (IN6_IS_ADDR_UNSPECIFIED(laddr))
817 else if (!IN6_ARE_ADDR_EQUAL(
818 &inp->in6p_laddr, laddr))
821 if (!IN6_IS_ADDR_UNSPECIFIED(laddr))
824 if (wildcard < matchwild) {
826 matchwild = wildcard;
837 in6_pcbpurgeif0(struct inpcbinfo *pcbinfo, struct ifnet *ifp)
840 struct in6_multi *inm;
841 struct in6_mfilter *imf;
842 struct ip6_moptions *im6o;
844 INP_INFO_WLOCK(pcbinfo);
845 CK_LIST_FOREACH(inp, pcbinfo->ipi_listhead, inp_list) {
847 if (__predict_false(inp->inp_flags2 & INP_FREED)) {
851 im6o = inp->in6p_moptions;
852 if ((inp->inp_vflag & INP_IPV6) && im6o != NULL) {
854 * Unselect the outgoing ifp for multicast if it
857 if (im6o->im6o_multicast_ifp == ifp)
858 im6o->im6o_multicast_ifp = NULL;
860 * Drop multicast group membership if we joined
861 * through the interface being detached.
864 IP6_MFILTER_FOREACH(imf, &im6o->im6o_head) {
865 if ((inm = imf->im6f_in6m) == NULL)
867 if (inm->in6m_ifp != ifp)
869 ip6_mfilter_remove(&im6o->im6o_head, imf);
870 IN6_MULTI_LOCK_ASSERT();
871 in6_leavegroup_locked(inm, NULL);
872 ip6_mfilter_free(imf);
878 INP_INFO_WUNLOCK(pcbinfo);
882 * Check for alternatives when higher level complains
883 * about service problems. For now, invalidate cached
884 * routing information. If the route was created dynamically
885 * (by a redirect), time to try a default gateway again.
888 in6_losing(struct inpcb *inp)
891 RO_INVALIDATE_CACHE(&inp->inp_route6);
895 * After a routing change, flush old routing
896 * and allocate a (hopefully) better one.
899 in6_rtchange(struct inpcb *inp, int errno __unused)
902 RO_INVALIDATE_CACHE(&inp->inp_route6);
906 static struct inpcb *
907 in6_pcblookup_lbgroup(const struct inpcbinfo *pcbinfo,
908 const struct in6_addr *laddr, uint16_t lport, const struct in6_addr *faddr,
909 uint16_t fport, int lookupflags, uint8_t numa_domain)
911 struct inpcb *local_wild, *numa_wild;
912 const struct inpcblbgrouphead *hdr;
913 struct inpcblbgroup *grp;
916 INP_HASH_LOCK_ASSERT(pcbinfo);
918 hdr = &pcbinfo->ipi_lbgrouphashbase[
919 INP_PCBPORTHASH(lport, pcbinfo->ipi_lbgrouphashmask)];
922 * Order of socket selection:
924 * 2. wild (if lookupflags contains INPLOOKUP_WILDCARD).
927 * - Load balanced group does not contain jailed sockets.
928 * - Load balanced does not contain IPv4 mapped INET6 wild sockets.
932 CK_LIST_FOREACH(grp, hdr, il_list) {
934 if (!(grp->il_vflag & INP_IPV6))
937 if (grp->il_lport != lport)
940 idx = INP_PCBLBGROUP_PKTHASH(INP6_PCBHASHKEY(faddr), lport,
941 fport) % grp->il_inpcnt;
942 if (IN6_ARE_ADDR_EQUAL(&grp->il6_laddr, laddr)) {
943 if (numa_domain == M_NODOM ||
944 grp->il_numa_domain == numa_domain) {
945 return (grp->il_inp[idx]);
948 numa_wild = grp->il_inp[idx];
950 if (IN6_IS_ADDR_UNSPECIFIED(&grp->il6_laddr) &&
951 (lookupflags & INPLOOKUP_WILDCARD) != 0 &&
952 (local_wild == NULL || numa_domain == M_NODOM ||
953 grp->il_numa_domain == numa_domain)) {
954 local_wild = grp->il_inp[idx];
957 if (numa_wild != NULL)
964 * Lookup PCB in hash list, using pcbgroup tables.
966 static struct inpcb *
967 in6_pcblookup_group(struct inpcbinfo *pcbinfo, struct inpcbgroup *pcbgroup,
968 struct in6_addr *faddr, u_int fport_arg, struct in6_addr *laddr,
969 u_int lport_arg, int lookupflags, struct ifnet *ifp)
971 struct inpcbhead *head;
972 struct inpcb *inp, *tmpinp;
973 u_short fport = fport_arg, lport = lport_arg;
977 * First look for an exact match.
980 INP_GROUP_LOCK(pcbgroup);
981 head = &pcbgroup->ipg_hashbase[INP_PCBHASH(
982 INP6_PCBHASHKEY(faddr), lport, fport, pcbgroup->ipg_hashmask)];
983 CK_LIST_FOREACH(inp, head, inp_pcbgrouphash) {
984 /* XXX inp locking */
985 if ((inp->inp_vflag & INP_IPV6) == 0)
987 if (IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, faddr) &&
988 IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr) &&
989 inp->inp_fport == fport &&
990 inp->inp_lport == lport) {
992 * XXX We should be able to directly return
993 * the inp here, without any checks.
994 * Well unless both bound with SO_REUSEPORT?
996 if (prison_flag(inp->inp_cred, PR_IP6))
1002 if (tmpinp != NULL) {
1008 * Then look for a wildcard match in the pcbgroup.
1010 if ((lookupflags & INPLOOKUP_WILDCARD) != 0) {
1011 struct inpcb *local_wild = NULL, *local_exact = NULL;
1012 struct inpcb *jail_wild = NULL;
1016 * Order of socket selection - we always prefer jails.
1017 * 1. jailed, non-wild.
1019 * 3. non-jailed, non-wild.
1020 * 4. non-jailed, wild.
1022 head = &pcbgroup->ipg_hashbase[
1023 INP_PCBHASH(INADDR_ANY, lport, 0, pcbgroup->ipg_hashmask)];
1024 CK_LIST_FOREACH(inp, head, inp_pcbgrouphash) {
1025 /* XXX inp locking */
1026 if ((inp->inp_vflag & INP_IPV6) == 0)
1029 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) ||
1030 inp->inp_lport != lport) {
1034 injail = prison_flag(inp->inp_cred, PR_IP6);
1036 if (prison_check_ip6(inp->inp_cred,
1040 if (local_exact != NULL)
1044 if (IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr)) {
1049 } else if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
1055 } /* LIST_FOREACH */
1069 * Then look for a wildcard match, if requested.
1071 if ((lookupflags & INPLOOKUP_WILDCARD) != 0) {
1072 struct inpcb *local_wild = NULL, *local_exact = NULL;
1073 struct inpcb *jail_wild = NULL;
1077 * Order of socket selection - we always prefer jails.
1078 * 1. jailed, non-wild.
1080 * 3. non-jailed, non-wild.
1081 * 4. non-jailed, wild.
1083 head = &pcbinfo->ipi_wildbase[INP_PCBHASH(
1084 INP6_PCBHASHKEY(&in6addr_any), lport, 0,
1085 pcbinfo->ipi_wildmask)];
1086 CK_LIST_FOREACH(inp, head, inp_pcbgroup_wild) {
1087 /* XXX inp locking */
1088 if ((inp->inp_vflag & INP_IPV6) == 0)
1091 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) ||
1092 inp->inp_lport != lport) {
1096 injail = prison_flag(inp->inp_cred, PR_IP6);
1098 if (prison_check_ip6(inp->inp_cred,
1102 if (local_exact != NULL)
1106 if (IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr)) {
1111 } else if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
1117 } /* LIST_FOREACH */
1128 } /* if ((lookupflags & INPLOOKUP_WILDCARD) != 0) */
1129 INP_GROUP_UNLOCK(pcbgroup);
1133 if (lookupflags & INPLOOKUP_WLOCKPCB)
1134 locked = INP_TRY_WLOCK(inp);
1135 else if (lookupflags & INPLOOKUP_RLOCKPCB)
1136 locked = INP_TRY_RLOCK(inp);
1138 panic("%s: locking buf", __func__);
1141 INP_GROUP_UNLOCK(pcbgroup);
1143 if (lookupflags & INPLOOKUP_WLOCKPCB) {
1145 if (in_pcbrele_wlocked(inp))
1149 if (in_pcbrele_rlocked(inp))
1154 if (lookupflags & INPLOOKUP_WLOCKPCB)
1155 INP_WLOCK_ASSERT(inp);
1157 INP_RLOCK_ASSERT(inp);
1161 #endif /* PCBGROUP */
1164 * Lookup PCB in hash list. Used in in_pcb.c as well as here.
1167 in6_pcblookup_hash_locked(struct inpcbinfo *pcbinfo, struct in6_addr *faddr,
1168 u_int fport_arg, struct in6_addr *laddr, u_int lport_arg,
1169 int lookupflags, struct ifnet *ifp, uint8_t numa_domain)
1171 struct inpcbhead *head;
1172 struct inpcb *inp, *tmpinp;
1173 u_short fport = fport_arg, lport = lport_arg;
1175 KASSERT((lookupflags & ~(INPLOOKUP_WILDCARD)) == 0,
1176 ("%s: invalid lookup flags %d", __func__, lookupflags));
1178 INP_HASH_LOCK_ASSERT(pcbinfo);
1181 * First look for an exact match.
1184 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(
1185 INP6_PCBHASHKEY(faddr), lport, fport, pcbinfo->ipi_hashmask)];
1186 CK_LIST_FOREACH(inp, head, inp_hash) {
1187 /* XXX inp locking */
1188 if ((inp->inp_vflag & INP_IPV6) == 0)
1190 if (IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, faddr) &&
1191 IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr) &&
1192 inp->inp_fport == fport &&
1193 inp->inp_lport == lport) {
1195 * XXX We should be able to directly return
1196 * the inp here, without any checks.
1197 * Well unless both bound with SO_REUSEPORT?
1199 if (prison_flag(inp->inp_cred, PR_IP6))
1209 * Then look in lb group (for wildcard match).
1211 if ((lookupflags & INPLOOKUP_WILDCARD) != 0) {
1212 inp = in6_pcblookup_lbgroup(pcbinfo, laddr, lport, faddr,
1213 fport, lookupflags, numa_domain);
1219 * Then look for a wildcard match, if requested.
1221 if ((lookupflags & INPLOOKUP_WILDCARD) != 0) {
1222 struct inpcb *local_wild = NULL, *local_exact = NULL;
1223 struct inpcb *jail_wild = NULL;
1227 * Order of socket selection - we always prefer jails.
1228 * 1. jailed, non-wild.
1230 * 3. non-jailed, non-wild.
1231 * 4. non-jailed, wild.
1233 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(
1234 INP6_PCBHASHKEY(&in6addr_any), lport, 0,
1235 pcbinfo->ipi_hashmask)];
1236 CK_LIST_FOREACH(inp, head, inp_hash) {
1237 /* XXX inp locking */
1238 if ((inp->inp_vflag & INP_IPV6) == 0)
1241 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) ||
1242 inp->inp_lport != lport) {
1246 injail = prison_flag(inp->inp_cred, PR_IP6);
1248 if (prison_check_ip6(inp->inp_cred,
1252 if (local_exact != NULL)
1256 if (IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr)) {
1261 } else if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
1267 } /* LIST_FOREACH */
1269 if (jail_wild != NULL)
1271 if (local_exact != NULL)
1272 return (local_exact);
1273 if (local_wild != NULL)
1274 return (local_wild);
1275 } /* if ((lookupflags & INPLOOKUP_WILDCARD) != 0) */
1284 * Lookup PCB in hash list, using pcbinfo tables. This variation locks the
1285 * hash list lock, and will return the inpcb locked (i.e., requires
1286 * INPLOOKUP_LOCKPCB).
1288 static struct inpcb *
1289 in6_pcblookup_hash(struct inpcbinfo *pcbinfo, struct in6_addr *faddr,
1290 u_int fport, struct in6_addr *laddr, u_int lport, int lookupflags,
1291 struct ifnet *ifp, uint8_t numa_domain)
1295 inp = in6_pcblookup_hash_locked(pcbinfo, faddr, fport, laddr, lport,
1296 (lookupflags & ~(INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)), ifp,
1299 if (lookupflags & INPLOOKUP_WLOCKPCB) {
1301 if (__predict_false(inp->inp_flags2 & INP_FREED)) {
1305 } else if (lookupflags & INPLOOKUP_RLOCKPCB) {
1307 if (__predict_false(inp->inp_flags2 & INP_FREED)) {
1312 panic("%s: locking bug", __func__);
1315 if (lookupflags & INPLOOKUP_WLOCKPCB)
1316 INP_WLOCK_ASSERT(inp);
1318 INP_RLOCK_ASSERT(inp);
1326 * Public inpcb lookup routines, accepting a 4-tuple, and optionally, an mbuf
1327 * from which a pre-calculated hash value may be extracted.
1329 * Possibly more of this logic should be in in6_pcbgroup.c.
1332 in6_pcblookup(struct inpcbinfo *pcbinfo, struct in6_addr *faddr, u_int fport,
1333 struct in6_addr *laddr, u_int lport, int lookupflags, struct ifnet *ifp)
1335 #if defined(PCBGROUP) && !defined(RSS)
1336 struct inpcbgroup *pcbgroup;
1339 KASSERT((lookupflags & ~INPLOOKUP_MASK) == 0,
1340 ("%s: invalid lookup flags %d", __func__, lookupflags));
1341 KASSERT((lookupflags & (INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)) != 0,
1342 ("%s: LOCKPCB not set", __func__));
1345 * When not using RSS, use connection groups in preference to the
1346 * reservation table when looking up 4-tuples. When using RSS, just
1347 * use the reservation table, due to the cost of the Toeplitz hash
1350 * XXXRW: This policy belongs in the pcbgroup code, as in principle
1351 * we could be doing RSS with a non-Toeplitz hash that is affordable
1354 #if defined(PCBGROUP) && !defined(RSS)
1355 if (in_pcbgroup_enabled(pcbinfo)) {
1356 pcbgroup = in6_pcbgroup_bytuple(pcbinfo, laddr, lport, faddr,
1358 return (in6_pcblookup_group(pcbinfo, pcbgroup, faddr, fport,
1359 laddr, lport, lookupflags, ifp));
1362 return (in6_pcblookup_hash(pcbinfo, faddr, fport, laddr, lport,
1363 lookupflags, ifp, M_NODOM));
1367 in6_pcblookup_mbuf(struct inpcbinfo *pcbinfo, struct in6_addr *faddr,
1368 u_int fport, struct in6_addr *laddr, u_int lport, int lookupflags,
1369 struct ifnet *ifp, struct mbuf *m)
1372 struct inpcbgroup *pcbgroup;
1375 KASSERT((lookupflags & ~INPLOOKUP_MASK) == 0,
1376 ("%s: invalid lookup flags %d", __func__, lookupflags));
1377 KASSERT((lookupflags & (INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)) != 0,
1378 ("%s: LOCKPCB not set", __func__));
1382 * If we can use a hardware-generated hash to look up the connection
1383 * group, use that connection group to find the inpcb. Otherwise
1384 * fall back on a software hash -- or the reservation table if we're
1387 * XXXRW: As above, that policy belongs in the pcbgroup code.
1389 if (in_pcbgroup_enabled(pcbinfo) &&
1390 M_HASHTYPE_TEST(m, M_HASHTYPE_NONE) == 0) {
1391 pcbgroup = in6_pcbgroup_byhash(pcbinfo, M_HASHTYPE_GET(m),
1392 m->m_pkthdr.flowid);
1393 if (pcbgroup != NULL)
1394 return (in6_pcblookup_group(pcbinfo, pcbgroup, faddr,
1395 fport, laddr, lport, lookupflags, ifp));
1397 pcbgroup = in6_pcbgroup_bytuple(pcbinfo, laddr, lport, faddr,
1399 return (in6_pcblookup_group(pcbinfo, pcbgroup, faddr, fport,
1400 laddr, lport, lookupflags, ifp));
1404 return (in6_pcblookup_hash(pcbinfo, faddr, fport, laddr, lport,
1405 lookupflags, ifp, m->m_pkthdr.numa_domain));
1409 init_sin6(struct sockaddr_in6 *sin6, struct mbuf *m, int srcordst)
1413 ip = mtod(m, struct ip6_hdr *);
1414 bzero(sin6, sizeof(*sin6));
1415 sin6->sin6_len = sizeof(*sin6);
1416 sin6->sin6_family = AF_INET6;
1417 sin6->sin6_addr = srcordst ? ip->ip6_dst : ip->ip6_src;
1419 (void)sa6_recoverscope(sin6); /* XXX: should catch errors... */