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.
6 * Copyright (c) 2014 Kevin Lo
9 * Portions of this software were developed by Robert N. M. Watson under
10 * contract to Juniper Networks, Inc.
12 * Redistribution and use in source and binary forms, with or without
13 * modification, are permitted provided that the following conditions
15 * 1. Redistributions of source code must retain the above copyright
16 * notice, this list of conditions and the following disclaimer.
17 * 2. Redistributions in binary form must reproduce the above copyright
18 * notice, this list of conditions and the following disclaimer in the
19 * documentation and/or other materials provided with the distribution.
20 * 3. Neither the name of the project nor the names of its contributors
21 * may be used to endorse or promote products derived from this software
22 * without specific prior written permission.
24 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
36 * $KAME: udp6_usrreq.c,v 1.27 2001/05/21 05:45:10 jinmei Exp $
37 * $KAME: udp6_output.c,v 1.31 2001/05/21 16:39:15 jinmei Exp $
41 * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995
42 * The Regents of the University of California.
43 * All rights reserved.
45 * Redistribution and use in source and binary forms, with or without
46 * modification, are permitted provided that the following conditions
48 * 1. Redistributions of source code must retain the above copyright
49 * notice, this list of conditions and the following disclaimer.
50 * 2. Redistributions in binary form must reproduce the above copyright
51 * notice, this list of conditions and the following disclaimer in the
52 * documentation and/or other materials provided with the distribution.
53 * 3. Neither the name of the University nor the names of its contributors
54 * may be used to endorse or promote products derived from this software
55 * without specific prior written permission.
57 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
58 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
59 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
60 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
61 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
62 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
63 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
64 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
65 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
66 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
69 * @(#)udp_usrreq.c 8.6 (Berkeley) 5/23/95
72 #include <sys/cdefs.h>
73 __FBSDID("$FreeBSD$");
76 #include "opt_inet6.h"
77 #include "opt_ipsec.h"
78 #include "opt_route.h"
81 #include <sys/param.h>
83 #include <sys/kernel.h>
88 #include <sys/protosw.h>
90 #include <sys/signalvar.h>
91 #include <sys/socket.h>
92 #include <sys/socketvar.h>
94 #include <sys/sysctl.h>
95 #include <sys/syslog.h>
96 #include <sys/systm.h>
99 #include <net/if_var.h>
100 #include <net/if_types.h>
101 #include <net/route.h>
102 #include <net/rss_config.h>
104 #include <netinet/in.h>
105 #include <netinet/in_kdtrace.h>
106 #include <netinet/in_pcb.h>
107 #include <netinet/in_systm.h>
108 #include <netinet/in_var.h>
109 #include <netinet/ip.h>
110 #include <netinet/ip6.h>
111 #include <netinet/icmp6.h>
112 #include <netinet/ip_var.h>
113 #include <netinet/udp.h>
114 #include <netinet/udp_var.h>
115 #include <netinet/udplite.h>
117 #include <netinet6/ip6_var.h>
118 #include <netinet6/in6_fib.h>
119 #include <netinet6/in6_pcb.h>
120 #include <netinet6/in6_rss.h>
121 #include <netinet6/udp6_var.h>
122 #include <netinet6/scope6_var.h>
124 #include <netipsec/ipsec_support.h>
126 #include <security/mac/mac_framework.h>
128 VNET_DEFINE(int, zero_checksum_port) = 0;
129 #define V_zero_checksum_port VNET(zero_checksum_port)
130 SYSCTL_INT(_net_inet6_udp6, OID_AUTO, rfc6935_port, CTLFLAG_VNET | CTLFLAG_RW,
131 &VNET_NAME(zero_checksum_port), 0,
132 "Zero UDP checksum allowed for traffic to/from this port.");
135 * UDP protocol implementation.
136 * Per RFC 768, August, 1980.
139 static void udp6_detach(struct socket *so);
142 udp6_append(struct inpcb *inp, struct mbuf *n, int off,
143 struct sockaddr_in6 *fromsa)
146 struct mbuf *opts = NULL, *tmp_opts;
150 INP_LOCK_ASSERT(inp);
153 * Engage the tunneling protocol.
156 if (up->u_tun_func != NULL) {
159 filtered = (*up->u_tun_func)(n, off, inp,
160 (struct sockaddr *)&fromsa[0], up->u_tun_ctx);
163 return (in_pcbrele_rlocked(inp));
165 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
166 /* Check AH/ESP integrity. */
167 if (IPSEC_ENABLED(ipv6)) {
168 if (IPSEC_CHECK_POLICY(ipv6, n, inp) != 0) {
175 if (mac_inpcb_check_deliver(inp, n) != 0) {
181 if (inp->inp_flags & INP_CONTROLOPTS ||
182 inp->inp_socket->so_options & SO_TIMESTAMP)
183 ip6_savecontrol(inp, n, &opts);
184 if ((inp->inp_vflag & INP_IPV6) && (inp->inp_flags2 & INP_ORIGDSTADDR)) {
185 tmp_opts = sbcreatecontrol(&fromsa[1],
186 sizeof(struct sockaddr_in6), IPV6_ORIGDSTADDR,
187 IPPROTO_IPV6, M_NOWAIT);
190 tmp_opts->m_next = opts;
196 m_adj(n, off + sizeof(struct udphdr));
198 so = inp->inp_socket;
199 SOCKBUF_LOCK(&so->so_rcv);
200 if (sbappendaddr_locked(&so->so_rcv, (struct sockaddr *)&fromsa[0], n,
202 soroverflow_locked(so);
206 UDPSTAT_INC(udps_fullsock);
208 sorwakeup_locked(so);
212 struct udp6_multi_match_ctx {
218 udp6_multi_match(const struct inpcb *inp, void *v)
220 struct udp6_multi_match_ctx *ctx = v;
222 if ((inp->inp_vflag & INP_IPV6) == 0)
224 if (inp->inp_lport != ctx->uh->uh_dport)
226 if (inp->inp_fport != 0 && inp->inp_fport != ctx->uh->uh_sport)
228 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
229 !IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, &ctx->ip6->ip6_dst))
231 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) &&
232 (!IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &ctx->ip6->ip6_src) ||
233 inp->inp_fport != ctx->uh->uh_sport))
240 udp6_multi_input(struct mbuf *m, int off, int proto,
241 struct sockaddr_in6 *fromsa)
243 struct udp6_multi_match_ctx ctx;
244 struct inpcb_iterator inpi = INP_ITERATOR(udp_get_inpcbinfo(proto),
245 INPLOOKUP_RLOCKPCB, udp6_multi_match, &ctx);
247 struct ip6_moptions *imo;
252 * In the event that laddr should be set to the link-local
253 * address (this happens in RIPng), the multicast address
254 * specified in the received packet will not match laddr. To
255 * handle this situation, matching is relaxed if the
256 * receiving interface is the same as one specified in the
257 * socket and if the destination multicast address matches
258 * one of the multicast groups specified in the socket.
262 * KAME note: traditionally we dropped udpiphdr from mbuf
263 * here. We need udphdr for IPsec processing so we do that
266 ctx.ip6 = mtod(m, struct ip6_hdr *);
267 ctx.uh = (struct udphdr *)((char *)ctx.ip6 + off);
268 while ((inp = inp_next(&inpi)) != NULL) {
269 INP_RLOCK_ASSERT(inp);
271 * XXXRW: Because we weren't holding either the inpcb
272 * or the hash lock when we checked for a match
273 * before, we should probably recheck now that the
274 * inpcb lock is (supposed to be) held.
277 * Handle socket delivery policy for any-source
278 * and source-specific multicast. [RFC3678]
280 if ((imo = inp->in6p_moptions) != NULL) {
281 struct sockaddr_in6 mcaddr;
284 bzero(&mcaddr, sizeof(struct sockaddr_in6));
285 mcaddr.sin6_len = sizeof(struct sockaddr_in6);
286 mcaddr.sin6_family = AF_INET6;
287 mcaddr.sin6_addr = ctx.ip6->ip6_dst;
289 blocked = im6o_mc_filter(imo, m->m_pkthdr.rcvif,
290 (struct sockaddr *)&mcaddr,
291 (struct sockaddr *)&fromsa[0]);
292 if (blocked != MCAST_PASS) {
293 if (blocked == MCAST_NOTGMEMBER)
294 IP6STAT_INC(ip6s_notmember);
295 if (blocked == MCAST_NOTSMEMBER ||
296 blocked == MCAST_MUTED)
297 UDPSTAT_INC(udps_filtermcast);
301 if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) != NULL) {
302 if (proto == IPPROTO_UDPLITE)
303 UDPLITE_PROBE(receive, NULL, inp, ctx.ip6,
306 UDP_PROBE(receive, NULL, inp, ctx.ip6, inp,
308 if (udp6_append(inp, n, off, fromsa)) {
314 * Don't look for additional matches if this one does
315 * not have either the SO_REUSEPORT or SO_REUSEADDR
316 * socket options set. This heuristic avoids
317 * searching through all pcbs in the common case of a
318 * non-shared port. It assumes that an application
319 * will never clear these options after setting them.
321 if ((inp->inp_socket->so_options &
322 (SO_REUSEPORT|SO_REUSEPORT_LB|SO_REUSEADDR)) == 0) {
331 * No matching pcb found; discard datagram. (No need
332 * to send an ICMP Port Unreachable for a broadcast
333 * or multicast datgram.)
335 UDPSTAT_INC(udps_noport);
336 UDPSTAT_INC(udps_noportmcast);
339 return (IPPROTO_DONE);
343 udp6_input(struct mbuf **mp, int *offp, int proto)
345 struct mbuf *m = *mp;
349 struct inpcbinfo *pcbinfo;
354 struct sockaddr_in6 fromsa[2];
355 struct m_tag *fwd_tag;
361 if (m->m_len < off + sizeof(struct udphdr)) {
362 m = m_pullup(m, off + sizeof(struct udphdr));
364 IP6STAT_INC(ip6s_exthdrtoolong);
366 return (IPPROTO_DONE);
369 ip6 = mtod(m, struct ip6_hdr *);
370 uh = (struct udphdr *)((caddr_t)ip6 + off);
372 UDPSTAT_INC(udps_ipackets);
375 * Destination port of 0 is illegal, based on RFC768.
377 if (uh->uh_dport == 0)
380 plen = ntohs(ip6->ip6_plen) - off + sizeof(*ip6);
381 ulen = ntohs((u_short)uh->uh_ulen);
384 cscov_partial = (nxt == IPPROTO_UDPLITE) ? 1 : 0;
385 if (nxt == IPPROTO_UDPLITE) {
386 /* Zero means checksum over the complete packet. */
391 if ((ulen < sizeof(struct udphdr)) || (ulen > plen)) {
392 /* XXX: What is the right UDPLite MIB counter? */
395 if (uh->uh_sum == 0) {
396 /* XXX: What is the right UDPLite MIB counter? */
400 if ((ulen < sizeof(struct udphdr)) || (plen != ulen)) {
401 UDPSTAT_INC(udps_badlen);
404 if (uh->uh_sum == 0) {
405 UDPSTAT_INC(udps_nosum);
407 * dport 0 was rejected earlier so this is OK even if
408 * zero_checksum_port is 0 (which is its default value).
410 if (ntohs(uh->uh_dport) == V_zero_checksum_port)
417 if ((m->m_pkthdr.csum_flags & CSUM_DATA_VALID_IPV6) &&
419 if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR)
420 uh_sum = m->m_pkthdr.csum_data;
422 uh_sum = in6_cksum_pseudo(ip6, ulen, nxt,
423 m->m_pkthdr.csum_data);
426 uh_sum = in6_cksum_partial(m, nxt, off, plen, ulen);
429 UDPSTAT_INC(udps_badsum);
435 * Construct sockaddr format source address.
437 init_sin6(&fromsa[0], m, 0);
438 fromsa[0].sin6_port = uh->uh_sport;
439 init_sin6(&fromsa[1], m, 1);
440 fromsa[1].sin6_port = uh->uh_dport;
442 pcbinfo = udp_get_inpcbinfo(nxt);
443 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
445 return (udp6_multi_input(m, off, proto, fromsa));
449 * Locate pcb for datagram.
453 * Grab info from PACKET_TAG_IPFORWARD tag prepended to the chain.
455 if ((m->m_flags & M_IP6_NEXTHOP) &&
456 (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
457 struct sockaddr_in6 *next_hop6;
459 next_hop6 = (struct sockaddr_in6 *)(fwd_tag + 1);
462 * Transparently forwarded. Pretend to be the destination.
463 * Already got one like this?
465 inp = in6_pcblookup_mbuf(pcbinfo, &ip6->ip6_src,
466 uh->uh_sport, &ip6->ip6_dst, uh->uh_dport,
467 INPLOOKUP_RLOCKPCB, m->m_pkthdr.rcvif, m);
470 * It's new. Try to find the ambushing socket.
471 * Because we've rewritten the destination address,
472 * any hardware-generated hash is ignored.
474 inp = in6_pcblookup(pcbinfo, &ip6->ip6_src,
475 uh->uh_sport, &next_hop6->sin6_addr,
476 next_hop6->sin6_port ? htons(next_hop6->sin6_port) :
477 uh->uh_dport, INPLOOKUP_WILDCARD |
478 INPLOOKUP_RLOCKPCB, m->m_pkthdr.rcvif);
480 /* Remove the tag from the packet. We don't need it anymore. */
481 m_tag_delete(m, fwd_tag);
482 m->m_flags &= ~M_IP6_NEXTHOP;
484 inp = in6_pcblookup_mbuf(pcbinfo, &ip6->ip6_src,
485 uh->uh_sport, &ip6->ip6_dst, uh->uh_dport,
486 INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB,
487 m->m_pkthdr.rcvif, m);
489 if (V_udp_log_in_vain) {
490 char ip6bufs[INET6_ADDRSTRLEN];
491 char ip6bufd[INET6_ADDRSTRLEN];
494 "Connection attempt to UDP [%s]:%d from [%s]:%d\n",
495 ip6_sprintf(ip6bufd, &ip6->ip6_dst),
497 ip6_sprintf(ip6bufs, &ip6->ip6_src),
498 ntohs(uh->uh_sport));
500 if (nxt == IPPROTO_UDPLITE)
501 UDPLITE_PROBE(receive, NULL, NULL, ip6, NULL, uh);
503 UDP_PROBE(receive, NULL, NULL, ip6, NULL, uh);
504 UDPSTAT_INC(udps_noport);
505 if (m->m_flags & M_MCAST) {
506 printf("UDP6: M_MCAST is set in a unicast packet.\n");
507 UDPSTAT_INC(udps_noportmcast);
510 if (V_udp_blackhole && (V_udp_blackhole_local ||
511 !in6_localaddr(&ip6->ip6_src)))
513 icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0);
515 return (IPPROTO_DONE);
517 INP_RLOCK_ASSERT(inp);
520 if (up->u_rxcslen == 0 || up->u_rxcslen > ulen) {
524 return (IPPROTO_DONE);
527 if (nxt == IPPROTO_UDPLITE)
528 UDPLITE_PROBE(receive, NULL, inp, ip6, inp, uh);
530 UDP_PROBE(receive, NULL, inp, ip6, inp, uh);
531 if (udp6_append(inp, m, off, fromsa) == 0)
534 return (IPPROTO_DONE);
539 return (IPPROTO_DONE);
543 udp6_common_ctlinput(struct ip6ctlparam *ip6cp, struct inpcbinfo *pcbinfo)
550 struct udp_portonly {
555 if ((errno = icmp6_errmap(ip6cp->ip6c_icmp6)) == 0)
559 ip6 = ip6cp->ip6c_ip6;
560 off = ip6cp->ip6c_off;
562 /* Check if we can safely examine src and dst ports. */
563 if (m->m_pkthdr.len < off + sizeof(*uhp))
566 bzero(&uh, sizeof(uh));
567 m_copydata(m, off, sizeof(*uhp), (caddr_t)&uh);
569 /* Check to see if its tunneled */
570 inp = in6_pcblookup_mbuf(pcbinfo, &ip6->ip6_dst, uh.uh_dport,
571 &ip6->ip6_src, uh.uh_sport, INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB,
572 m->m_pkthdr.rcvif, m);
575 udp_tun_icmp_t *func;
578 func = up->u_icmp_func;
583 in6_pcbnotify(pcbinfo, ip6cp->ip6c_finaldst, uh.uh_dport,
584 ip6cp->ip6c_src, uh.uh_sport, errno, ip6cp->ip6c_cmdarg,
589 udp6_ctlinput(struct ip6ctlparam *ctl)
592 return (udp6_common_ctlinput(ctl, &V_udbinfo));
596 udplite6_ctlinput(struct ip6ctlparam *ctl)
599 return (udp6_common_ctlinput(ctl, &V_ulitecbinfo));
603 udp6_getcred(SYSCTL_HANDLER_ARGS)
606 struct sockaddr_in6 addrs[2];
607 struct epoch_tracker et;
611 error = priv_check(req->td, PRIV_NETINET_GETCRED);
615 if (req->newlen != sizeof(addrs))
617 if (req->oldlen != sizeof(struct xucred))
619 error = SYSCTL_IN(req, addrs, sizeof(addrs));
622 if ((error = sa6_embedscope(&addrs[0], V_ip6_use_defzone)) != 0 ||
623 (error = sa6_embedscope(&addrs[1], V_ip6_use_defzone)) != 0) {
627 inp = in6_pcblookup(&V_udbinfo, &addrs[1].sin6_addr,
628 addrs[1].sin6_port, &addrs[0].sin6_addr, addrs[0].sin6_port,
629 INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB, NULL);
632 INP_RLOCK_ASSERT(inp);
633 if (inp->inp_socket == NULL)
636 error = cr_canseesocket(req->td->td_ucred,
639 cru2x(inp->inp_cred, &xuc);
644 error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred));
648 SYSCTL_PROC(_net_inet6_udp6, OID_AUTO, getcred,
649 CTLTYPE_OPAQUE | CTLFLAG_RW | CTLFLAG_MPSAFE,
650 0, 0, udp6_getcred, "S,xucred",
651 "Get the xucred of a UDP6 connection");
654 udp6_send(struct socket *so, int flags_arg, struct mbuf *m,
655 struct sockaddr *addr6, struct mbuf *control, struct thread *td)
660 struct in6_addr *laddr, *faddr, in6a;
661 struct ip6_pktopts *optp, opt;
662 struct sockaddr_in6 *sin6, tmp;
663 struct epoch_tracker et;
664 int cscov_partial, error, flags, hlen, scope_ambiguous;
665 u_int32_t ulen, plen;
672 if (addr6->sa_family != AF_INET6)
673 error = EAFNOSUPPORT;
674 else if (addr6->sa_len != sizeof(struct sockaddr_in6))
676 if (__predict_false(error != 0)) {
683 sin6 = (struct sockaddr_in6 *)addr6;
686 * In contrast to IPv4 we do not validate the max. packet length
687 * here due to IPv6 Jumbograms (RFC2675).
692 /* Protect *addr6 from overwrites. */
697 * Application should provide a proper zone ID or the use of
698 * default zone IDs should be enabled. Unfortunately, some
699 * applications do not behave as it should, so we need a
700 * workaround. Even if an appropriate ID is not determined,
701 * we'll see if we can determine the outgoing interface. If we
702 * can, determine the zone ID based on the interface below.
704 if (sin6->sin6_scope_id == 0 && !V_ip6_use_defzone)
706 if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0) {
715 KASSERT(inp != NULL, ("%s: inp == NULL", __func__));
717 * In the following cases we want a write lock on the inp for either
718 * local operations or for possible route cache updates in the IPv6
720 * - on connected sockets (sin6 is NULL) for route cache updates,
721 * - when we are not bound to an address and source port (it is
722 * in6_pcbsetport() which will require the write lock).
724 * We check the inp fields before actually locking the inp, so
725 * here exists a race, and we may WLOCK the inp and end with already
726 * bound one by other thread. This is fine.
728 if (sin6 == NULL || (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
729 inp->inp_lport == 0))
734 nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ?
735 IPPROTO_UDP : IPPROTO_UDPLITE;
738 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
742 hasv4addr = (inp->inp_vflag & INP_IPV4);
744 hasv4addr = IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)
748 * XXXRW: We release UDP-layer locks before calling
749 * udp_send() in order to avoid recursion. However,
750 * this does mean there is a short window where inp's
751 * fields are unstable. Could this lead to a
752 * potential race in which the factors causing us to
753 * select the UDPv4 output routine are invalidated?
757 in6_sin6_2_sin_in_sock((struct sockaddr *)sin6);
758 /* addr will just be freed in sendit(). */
759 return (udp_send(so, flags_arg | PRUS_IPV6, m,
760 (struct sockaddr *)sin6, control, td));
764 if (sin6 && IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
766 * Given this is either an IPv6-only socket or no INET is
767 * supported we will fail the send if the given destination
768 * address is a v4mapped address.
778 if ((error = ip6_setpktopts(control, &opt,
779 inp->in6p_outputopts, td->td_ucred, nxt)) != 0) {
784 optp = inp->in6p_outputopts;
788 * Since we saw no essential reason for calling in_pcbconnect,
789 * we get rid of such kind of logic, and call in6_selectsrc
790 * and in6_pcbsetport in order to fill in the local address
791 * and the local port.
793 if (sin6->sin6_port == 0) {
794 error = EADDRNOTAVAIL;
798 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
799 /* how about ::ffff:0.0.0.0 case? */
805 * Given we handle the v4mapped case in the INET block above
806 * assert here that it must not happen anymore.
808 KASSERT(!IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr),
809 ("%s: sin6(%p)->sin6_addr is v4mapped which we "
810 "should have handled.", __func__, sin6));
812 /* This only requires read-locking. */
813 error = in6_selectsrc_socket(sin6, optp, inp,
814 td->td_ucred, scope_ambiguous, &in6a, NULL);
819 if (inp->inp_lport == 0) {
820 struct inpcbinfo *pcbinfo;
822 INP_WLOCK_ASSERT(inp);
824 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
825 INP_HASH_WLOCK(pcbinfo);
826 error = in6_pcbsetport(laddr, inp, td->td_ucred);
827 INP_HASH_WUNLOCK(pcbinfo);
829 /* Undo an address bind that may have occurred. */
830 inp->in6p_laddr = in6addr_any;
834 faddr = &sin6->sin6_addr;
835 fport = sin6->sin6_port; /* allow 0 port */
838 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
842 laddr = &inp->in6p_laddr;
843 faddr = &inp->in6p_faddr;
844 fport = inp->inp_fport;
847 ulen = m->m_pkthdr.len;
848 plen = sizeof(struct udphdr) + ulen;
849 hlen = sizeof(struct ip6_hdr);
852 * Calculate data length and get a mbuf
853 * for UDP and IP6 headers.
855 M_PREPEND(m, hlen + sizeof(struct udphdr), M_NOWAIT);
862 * Stuff checksum and output datagram.
864 cscov = cscov_partial = 0;
865 udp6 = (struct udphdr *)(mtod(m, caddr_t) + hlen);
866 udp6->uh_sport = inp->inp_lport; /* lport is always set in the PCB */
867 udp6->uh_dport = fport;
868 if (nxt == IPPROTO_UDPLITE) {
872 cscov = up->u_txcslen;
875 udp6->uh_ulen = htons(cscov);
877 * For UDP-Lite, checksum coverage length of zero means
878 * the entire UDPLite packet is covered by the checksum.
880 cscov_partial = (cscov == 0) ? 0 : 1;
881 } else if (plen <= 0xffff)
882 udp6->uh_ulen = htons((u_short)plen);
887 ip6 = mtod(m, struct ip6_hdr *);
888 ip6->ip6_flow = inp->inp_flow & IPV6_FLOWINFO_MASK;
889 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
890 ip6->ip6_vfc |= IPV6_VERSION;
891 ip6->ip6_plen = htons((u_short)plen);
893 ip6->ip6_hlim = in6_selecthlim(inp, NULL);
894 ip6->ip6_src = *laddr;
895 ip6->ip6_dst = *faddr;
898 mac_inpcb_create_mbuf(inp, m);
902 if ((udp6->uh_sum = in6_cksum_partial(m, nxt,
903 sizeof(struct ip6_hdr), plen, cscov)) == 0)
904 udp6->uh_sum = 0xffff;
906 udp6->uh_sum = in6_cksum_pseudo(ip6, plen, nxt, 0);
907 m->m_pkthdr.csum_flags = CSUM_UDP_IPV6;
908 m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
912 #if defined(ROUTE_MPATH) || defined(RSS)
913 if (CALC_FLOWID_OUTBOUND_SENDTO) {
914 uint32_t hash_type, hash_val;
917 pr = inp->inp_socket->so_proto->pr_protocol;
919 hash_val = fib6_calc_packet_hash(laddr, faddr,
920 inp->inp_lport, fport, pr, &hash_type);
921 m->m_pkthdr.flowid = hash_val;
922 M_HASHTYPE_SET(m, hash_type);
924 /* do not use inp flowid */
925 flags |= IP_NODEFAULTFLOWID;
928 UDPSTAT_INC(udps_opackets);
929 if (nxt == IPPROTO_UDPLITE)
930 UDPLITE_PROBE(send, NULL, inp, ip6, inp, udp6);
932 UDP_PROBE(send, NULL, inp, ip6, inp, udp6);
933 error = ip6_output(m, optp,
934 INP_WLOCKED(inp) ? &inp->inp_route6 : NULL, flags,
935 inp->in6p_moptions, NULL, inp);
940 ip6_clearpktopts(&opt, -1);
949 ip6_clearpktopts(&opt, -1);
958 udp6_abort(struct socket *so)
961 struct inpcbinfo *pcbinfo;
963 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
965 KASSERT(inp != NULL, ("udp6_abort: inp == NULL"));
969 if (inp->inp_vflag & INP_IPV4) {
976 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
977 INP_HASH_WLOCK(pcbinfo);
978 in6_pcbdisconnect(inp);
979 INP_HASH_WUNLOCK(pcbinfo);
980 soisdisconnected(so);
986 udp6_attach(struct socket *so, int proto, struct thread *td)
988 struct inpcbinfo *pcbinfo;
993 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
995 KASSERT(inp == NULL, ("udp6_attach: inp != NULL"));
997 if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
998 error = soreserve(so, udp_sendspace, udp_recvspace);
1002 error = in_pcballoc(so, pcbinfo);
1005 inp = (struct inpcb *)so->so_pcb;
1006 inp->in6p_cksum = -1; /* just to be sure */
1009 * IPv4 TTL initialization is necessary for an IPv6 socket as well,
1010 * because the socket may be bound to an IPv6 wildcard address,
1011 * which may match an IPv4-mapped IPv6 address.
1013 inp->inp_ip_ttl = V_ip_defttl;
1014 up = intoudpcb(inp);
1015 bzero(&up->u_start_zero, u_zero_size);
1021 udp6_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
1023 struct sockaddr_in6 *sin6_p;
1025 struct inpcbinfo *pcbinfo;
1029 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1030 inp = sotoinpcb(so);
1031 KASSERT(inp != NULL, ("udp6_bind: inp == NULL"));
1033 if (nam->sa_family != AF_INET6)
1034 return (EAFNOSUPPORT);
1035 if (nam->sa_len != sizeof(struct sockaddr_in6))
1038 sin6_p = (struct sockaddr_in6 *)nam;
1041 INP_HASH_WLOCK(pcbinfo);
1042 vflagsav = inp->inp_vflag;
1043 inp->inp_vflag &= ~INP_IPV4;
1044 inp->inp_vflag |= INP_IPV6;
1045 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
1046 if (IN6_IS_ADDR_UNSPECIFIED(&sin6_p->sin6_addr))
1047 inp->inp_vflag |= INP_IPV4;
1049 else if (IN6_IS_ADDR_V4MAPPED(&sin6_p->sin6_addr)) {
1050 struct sockaddr_in sin;
1052 in6_sin6_2_sin(&sin, sin6_p);
1053 inp->inp_vflag |= INP_IPV4;
1054 inp->inp_vflag &= ~INP_IPV6;
1055 error = in_pcbbind(inp, &sin, td->td_ucred);
1061 error = in6_pcbbind(inp, sin6_p, td->td_ucred);
1066 inp->inp_vflag = vflagsav;
1067 INP_HASH_WUNLOCK(pcbinfo);
1073 udp6_close(struct socket *so)
1076 struct inpcbinfo *pcbinfo;
1078 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1079 inp = sotoinpcb(so);
1080 KASSERT(inp != NULL, ("udp6_close: inp == NULL"));
1084 if (inp->inp_vflag & INP_IPV4) {
1086 (void)udp_disconnect(so);
1090 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1091 INP_HASH_WLOCK(pcbinfo);
1092 in6_pcbdisconnect(inp);
1093 INP_HASH_WUNLOCK(pcbinfo);
1094 soisdisconnected(so);
1100 udp6_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
1102 struct epoch_tracker et;
1104 struct inpcbinfo *pcbinfo;
1105 struct sockaddr_in6 *sin6;
1109 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1110 inp = sotoinpcb(so);
1111 KASSERT(inp != NULL, ("udp6_connect: inp == NULL"));
1113 sin6 = (struct sockaddr_in6 *)nam;
1114 if (sin6->sin6_family != AF_INET6)
1115 return (EAFNOSUPPORT);
1116 if (sin6->sin6_len != sizeof(*sin6))
1120 * XXXRW: Need to clarify locking of v4/v6 flags.
1124 if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
1125 struct sockaddr_in sin;
1127 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) {
1131 if ((inp->inp_vflag & INP_IPV4) == 0) {
1132 error = EAFNOSUPPORT;
1135 if (inp->inp_faddr.s_addr != INADDR_ANY) {
1139 in6_sin6_2_sin(&sin, sin6);
1140 error = prison_remote_ip4(td->td_ucred, &sin.sin_addr);
1143 vflagsav = inp->inp_vflag;
1144 inp->inp_vflag |= INP_IPV4;
1145 inp->inp_vflag &= ~INP_IPV6;
1146 NET_EPOCH_ENTER(et);
1147 INP_HASH_WLOCK(pcbinfo);
1148 error = in_pcbconnect(inp, &sin, td->td_ucred, true);
1149 INP_HASH_WUNLOCK(pcbinfo);
1152 * If connect succeeds, mark socket as connected. If
1153 * connect fails and socket is unbound, reset inp_vflag
1158 else if (inp->inp_laddr.s_addr == INADDR_ANY &&
1159 inp->inp_lport == 0)
1160 inp->inp_vflag = vflagsav;
1163 if ((inp->inp_vflag & INP_IPV6) == 0) {
1164 error = EAFNOSUPPORT;
1169 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1173 error = prison_remote_ip6(td->td_ucred, &sin6->sin6_addr);
1176 vflagsav = inp->inp_vflag;
1177 inp->inp_vflag &= ~INP_IPV4;
1178 inp->inp_vflag |= INP_IPV6;
1179 NET_EPOCH_ENTER(et);
1180 INP_HASH_WLOCK(pcbinfo);
1181 error = in6_pcbconnect(inp, sin6, td->td_ucred, true);
1182 INP_HASH_WUNLOCK(pcbinfo);
1185 * If connect succeeds, mark socket as connected. If
1186 * connect fails and socket is unbound, reset inp_vflag
1191 else if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
1192 inp->inp_lport == 0)
1193 inp->inp_vflag = vflagsav;
1200 udp6_detach(struct socket *so)
1204 inp = sotoinpcb(so);
1205 KASSERT(inp != NULL, ("udp6_detach: inp == NULL"));
1213 udp6_disconnect(struct socket *so)
1216 struct inpcbinfo *pcbinfo;
1218 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1219 inp = sotoinpcb(so);
1220 KASSERT(inp != NULL, ("udp6_disconnect: inp == NULL"));
1224 if (inp->inp_vflag & INP_IPV4) {
1226 (void)udp_disconnect(so);
1231 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1236 INP_HASH_WLOCK(pcbinfo);
1237 in6_pcbdisconnect(inp);
1238 INP_HASH_WUNLOCK(pcbinfo);
1240 so->so_state &= ~SS_ISCONNECTED; /* XXX */
1246 #define UDP6_PROTOSW \
1247 .pr_type = SOCK_DGRAM, \
1248 .pr_flags = PR_ATOMIC|PR_ADDR|PR_CAPATTACH, \
1249 .pr_ctloutput = ip6_ctloutput, \
1250 .pr_abort = udp6_abort, \
1251 .pr_attach = udp6_attach, \
1252 .pr_bind = udp6_bind, \
1253 .pr_connect = udp6_connect, \
1254 .pr_control = in6_control, \
1255 .pr_detach = udp6_detach, \
1256 .pr_disconnect = udp6_disconnect, \
1257 .pr_peeraddr = in6_mapped_peeraddr, \
1258 .pr_send = udp6_send, \
1259 .pr_shutdown = udp_shutdown, \
1260 .pr_sockaddr = in6_mapped_sockaddr, \
1261 .pr_soreceive = soreceive_dgram, \
1262 .pr_sosend = sosend_dgram, \
1263 .pr_sosetlabel = in_pcbsosetlabel, \
1264 .pr_close = udp6_close
1266 struct protosw udp6_protosw = {
1267 .pr_protocol = IPPROTO_UDP,
1271 struct protosw udplite6_protosw = {
1272 .pr_protocol = IPPROTO_UDPLITE,
1277 udp6_init(void *arg __unused)
1280 IP6PROTO_REGISTER(IPPROTO_UDP, udp6_input, udp6_ctlinput);
1281 IP6PROTO_REGISTER(IPPROTO_UDPLITE, udp6_input, udplite6_ctlinput);
1283 SYSINIT(udp6_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, udp6_init, NULL);