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 /* netinet/udp_usrreqs.c */
136 pr_abort_t udp_abort;
137 pr_disconnect_t udp_disconnect;
141 * UDP protocol implementation.
142 * Per RFC 768, August, 1980.
145 static void udp6_detach(struct socket *so);
148 udp6_append(struct inpcb *inp, struct mbuf *n, int off,
149 struct sockaddr_in6 *fromsa)
152 struct mbuf *opts = NULL, *tmp_opts;
156 INP_LOCK_ASSERT(inp);
159 * Engage the tunneling protocol.
162 if (up->u_tun_func != NULL) {
165 filtered = (*up->u_tun_func)(n, off, inp,
166 (struct sockaddr *)&fromsa[0], up->u_tun_ctx);
169 return (in_pcbrele_rlocked(inp));
171 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
172 /* Check AH/ESP integrity. */
173 if (IPSEC_ENABLED(ipv6)) {
174 if (IPSEC_CHECK_POLICY(ipv6, n, inp) != 0) {
181 if (mac_inpcb_check_deliver(inp, n) != 0) {
187 if (inp->inp_flags & INP_CONTROLOPTS ||
188 inp->inp_socket->so_options & SO_TIMESTAMP)
189 ip6_savecontrol(inp, n, &opts);
190 if ((inp->inp_vflag & INP_IPV6) && (inp->inp_flags2 & INP_ORIGDSTADDR)) {
191 tmp_opts = sbcreatecontrol(&fromsa[1],
192 sizeof(struct sockaddr_in6), IPV6_ORIGDSTADDR,
193 IPPROTO_IPV6, M_NOWAIT);
196 tmp_opts->m_next = opts;
202 m_adj(n, off + sizeof(struct udphdr));
204 so = inp->inp_socket;
205 SOCKBUF_LOCK(&so->so_rcv);
206 if (sbappendaddr_locked(&so->so_rcv, (struct sockaddr *)&fromsa[0], n,
208 soroverflow_locked(so);
212 UDPSTAT_INC(udps_fullsock);
214 sorwakeup_locked(so);
218 struct udp6_multi_match_ctx {
224 udp6_multi_match(const struct inpcb *inp, void *v)
226 struct udp6_multi_match_ctx *ctx = v;
228 if ((inp->inp_vflag & INP_IPV6) == 0)
230 if (inp->inp_lport != ctx->uh->uh_dport)
232 if (inp->inp_fport != 0 && inp->inp_fport != ctx->uh->uh_sport)
234 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
235 !IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, &ctx->ip6->ip6_dst))
237 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) &&
238 (!IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &ctx->ip6->ip6_src) ||
239 inp->inp_fport != ctx->uh->uh_sport))
246 udp6_multi_input(struct mbuf *m, int off, int proto,
247 struct sockaddr_in6 *fromsa)
249 struct udp6_multi_match_ctx ctx;
250 struct inpcb_iterator inpi = INP_ITERATOR(udp_get_inpcbinfo(proto),
251 INPLOOKUP_RLOCKPCB, udp6_multi_match, &ctx);
253 struct ip6_moptions *imo;
258 * In the event that laddr should be set to the link-local
259 * address (this happens in RIPng), the multicast address
260 * specified in the received packet will not match laddr. To
261 * handle this situation, matching is relaxed if the
262 * receiving interface is the same as one specified in the
263 * socket and if the destination multicast address matches
264 * one of the multicast groups specified in the socket.
268 * KAME note: traditionally we dropped udpiphdr from mbuf
269 * here. We need udphdr for IPsec processing so we do that
272 ctx.ip6 = mtod(m, struct ip6_hdr *);
273 ctx.uh = (struct udphdr *)((char *)ctx.ip6 + off);
274 while ((inp = inp_next(&inpi)) != NULL) {
275 INP_RLOCK_ASSERT(inp);
277 * XXXRW: Because we weren't holding either the inpcb
278 * or the hash lock when we checked for a match
279 * before, we should probably recheck now that the
280 * inpcb lock is (supposed to be) held.
283 * Handle socket delivery policy for any-source
284 * and source-specific multicast. [RFC3678]
286 if ((imo = inp->in6p_moptions) != NULL) {
287 struct sockaddr_in6 mcaddr;
290 bzero(&mcaddr, sizeof(struct sockaddr_in6));
291 mcaddr.sin6_len = sizeof(struct sockaddr_in6);
292 mcaddr.sin6_family = AF_INET6;
293 mcaddr.sin6_addr = ctx.ip6->ip6_dst;
295 blocked = im6o_mc_filter(imo, m->m_pkthdr.rcvif,
296 (struct sockaddr *)&mcaddr,
297 (struct sockaddr *)&fromsa[0]);
298 if (blocked != MCAST_PASS) {
299 if (blocked == MCAST_NOTGMEMBER)
300 IP6STAT_INC(ip6s_notmember);
301 if (blocked == MCAST_NOTSMEMBER ||
302 blocked == MCAST_MUTED)
303 UDPSTAT_INC(udps_filtermcast);
307 if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) != NULL) {
308 if (proto == IPPROTO_UDPLITE)
309 UDPLITE_PROBE(receive, NULL, inp, ctx.ip6,
312 UDP_PROBE(receive, NULL, inp, ctx.ip6, inp,
314 if (udp6_append(inp, n, off, fromsa)) {
321 * Don't look for additional matches if this one does
322 * not have either the SO_REUSEPORT or SO_REUSEADDR
323 * socket options set. This heuristic avoids
324 * searching through all pcbs in the common case of a
325 * non-shared port. It assumes that an application
326 * will never clear these options after setting them.
328 if ((inp->inp_socket->so_options &
329 (SO_REUSEPORT|SO_REUSEPORT_LB|SO_REUSEADDR)) == 0) {
338 * No matching pcb found; discard datagram. (No need
339 * to send an ICMP Port Unreachable for a broadcast
340 * or multicast datgram.)
342 UDPSTAT_INC(udps_noport);
343 UDPSTAT_INC(udps_noportmcast);
346 return (IPPROTO_DONE);
350 udp6_input(struct mbuf **mp, int *offp, int proto)
352 struct mbuf *m = *mp;
356 struct inpcbinfo *pcbinfo;
361 struct sockaddr_in6 fromsa[2];
362 struct m_tag *fwd_tag;
368 if (m->m_len < off + sizeof(struct udphdr)) {
369 m = m_pullup(m, off + sizeof(struct udphdr));
371 IP6STAT_INC(ip6s_exthdrtoolong);
373 return (IPPROTO_DONE);
376 ip6 = mtod(m, struct ip6_hdr *);
377 uh = (struct udphdr *)((caddr_t)ip6 + off);
379 UDPSTAT_INC(udps_ipackets);
382 * Destination port of 0 is illegal, based on RFC768.
384 if (uh->uh_dport == 0)
387 plen = ntohs(ip6->ip6_plen) - off + sizeof(*ip6);
388 ulen = ntohs((u_short)uh->uh_ulen);
391 cscov_partial = (nxt == IPPROTO_UDPLITE) ? 1 : 0;
392 if (nxt == IPPROTO_UDPLITE) {
393 /* Zero means checksum over the complete packet. */
398 if ((ulen < sizeof(struct udphdr)) || (ulen > plen)) {
399 /* XXX: What is the right UDPLite MIB counter? */
402 if (uh->uh_sum == 0) {
403 /* XXX: What is the right UDPLite MIB counter? */
407 if ((ulen < sizeof(struct udphdr)) || (plen != ulen)) {
408 UDPSTAT_INC(udps_badlen);
411 if (uh->uh_sum == 0) {
412 UDPSTAT_INC(udps_nosum);
414 * dport 0 was rejected earlier so this is OK even if
415 * zero_checksum_port is 0 (which is its default value).
417 if (ntohs(uh->uh_dport) == V_zero_checksum_port)
424 if ((m->m_pkthdr.csum_flags & CSUM_DATA_VALID_IPV6) &&
426 if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR)
427 uh_sum = m->m_pkthdr.csum_data;
429 uh_sum = in6_cksum_pseudo(ip6, ulen, nxt,
430 m->m_pkthdr.csum_data);
433 uh_sum = in6_cksum_partial(m, nxt, off, plen, ulen);
436 UDPSTAT_INC(udps_badsum);
442 * Construct sockaddr format source address.
444 init_sin6(&fromsa[0], m, 0);
445 fromsa[0].sin6_port = uh->uh_sport;
446 init_sin6(&fromsa[1], m, 1);
447 fromsa[1].sin6_port = uh->uh_dport;
449 pcbinfo = udp_get_inpcbinfo(nxt);
450 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
452 return (udp6_multi_input(m, off, proto, fromsa));
456 * Locate pcb for datagram.
460 * Grab info from PACKET_TAG_IPFORWARD tag prepended to the chain.
462 if ((m->m_flags & M_IP6_NEXTHOP) &&
463 (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
464 struct sockaddr_in6 *next_hop6;
466 next_hop6 = (struct sockaddr_in6 *)(fwd_tag + 1);
469 * Transparently forwarded. Pretend to be the destination.
470 * Already got one like this?
472 inp = in6_pcblookup_mbuf(pcbinfo, &ip6->ip6_src,
473 uh->uh_sport, &ip6->ip6_dst, uh->uh_dport,
474 INPLOOKUP_RLOCKPCB, m->m_pkthdr.rcvif, m);
477 * It's new. Try to find the ambushing socket.
478 * Because we've rewritten the destination address,
479 * any hardware-generated hash is ignored.
481 inp = in6_pcblookup(pcbinfo, &ip6->ip6_src,
482 uh->uh_sport, &next_hop6->sin6_addr,
483 next_hop6->sin6_port ? htons(next_hop6->sin6_port) :
484 uh->uh_dport, INPLOOKUP_WILDCARD |
485 INPLOOKUP_RLOCKPCB, m->m_pkthdr.rcvif);
487 /* Remove the tag from the packet. We don't need it anymore. */
488 m_tag_delete(m, fwd_tag);
489 m->m_flags &= ~M_IP6_NEXTHOP;
491 inp = in6_pcblookup_mbuf(pcbinfo, &ip6->ip6_src,
492 uh->uh_sport, &ip6->ip6_dst, uh->uh_dport,
493 INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB,
494 m->m_pkthdr.rcvif, m);
496 if (V_udp_log_in_vain) {
497 char ip6bufs[INET6_ADDRSTRLEN];
498 char ip6bufd[INET6_ADDRSTRLEN];
501 "Connection attempt to UDP [%s]:%d from [%s]:%d\n",
502 ip6_sprintf(ip6bufd, &ip6->ip6_dst),
504 ip6_sprintf(ip6bufs, &ip6->ip6_src),
505 ntohs(uh->uh_sport));
507 if (nxt == IPPROTO_UDPLITE)
508 UDPLITE_PROBE(receive, NULL, NULL, ip6, NULL, uh);
510 UDP_PROBE(receive, NULL, NULL, ip6, NULL, uh);
511 UDPSTAT_INC(udps_noport);
512 if (m->m_flags & M_MCAST) {
513 printf("UDP6: M_MCAST is set in a unicast packet.\n");
514 UDPSTAT_INC(udps_noportmcast);
517 if (V_udp_blackhole && (V_udp_blackhole_local ||
518 !in6_localaddr(&ip6->ip6_src)))
520 icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0);
522 return (IPPROTO_DONE);
524 INP_RLOCK_ASSERT(inp);
527 if (up->u_rxcslen == 0 || up->u_rxcslen > ulen) {
531 return (IPPROTO_DONE);
534 if (nxt == IPPROTO_UDPLITE)
535 UDPLITE_PROBE(receive, NULL, inp, ip6, inp, uh);
537 UDP_PROBE(receive, NULL, inp, ip6, inp, uh);
538 if (udp6_append(inp, m, off, fromsa) == 0)
541 return (IPPROTO_DONE);
546 return (IPPROTO_DONE);
550 udp6_common_ctlinput(struct ip6ctlparam *ip6cp, struct inpcbinfo *pcbinfo)
557 struct udp_portonly {
562 if ((errno = icmp6_errmap(ip6cp->ip6c_icmp6)) == 0)
566 ip6 = ip6cp->ip6c_ip6;
567 off = ip6cp->ip6c_off;
569 /* Check if we can safely examine src and dst ports. */
570 if (m->m_pkthdr.len < off + sizeof(*uhp))
573 bzero(&uh, sizeof(uh));
574 m_copydata(m, off, sizeof(*uhp), (caddr_t)&uh);
576 /* Check to see if its tunneled */
577 inp = in6_pcblookup_mbuf(pcbinfo, &ip6->ip6_dst, uh.uh_dport,
578 &ip6->ip6_src, uh.uh_sport, INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB,
579 m->m_pkthdr.rcvif, m);
582 udp_tun_icmp_t *func;
585 func = up->u_icmp_func;
590 in6_pcbnotify(pcbinfo, ip6cp->ip6c_finaldst, uh.uh_dport,
591 ip6cp->ip6c_src, uh.uh_sport, errno, ip6cp->ip6c_cmdarg,
596 udp6_ctlinput(struct ip6ctlparam *ctl)
599 return (udp6_common_ctlinput(ctl, &V_udbinfo));
603 udplite6_ctlinput(struct ip6ctlparam *ctl)
606 return (udp6_common_ctlinput(ctl, &V_ulitecbinfo));
610 udp6_getcred(SYSCTL_HANDLER_ARGS)
613 struct sockaddr_in6 addrs[2];
614 struct epoch_tracker et;
618 error = priv_check(req->td, PRIV_NETINET_GETCRED);
622 if (req->newlen != sizeof(addrs))
624 if (req->oldlen != sizeof(struct xucred))
626 error = SYSCTL_IN(req, addrs, sizeof(addrs));
629 if ((error = sa6_embedscope(&addrs[0], V_ip6_use_defzone)) != 0 ||
630 (error = sa6_embedscope(&addrs[1], V_ip6_use_defzone)) != 0) {
634 inp = in6_pcblookup(&V_udbinfo, &addrs[1].sin6_addr,
635 addrs[1].sin6_port, &addrs[0].sin6_addr, addrs[0].sin6_port,
636 INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB, NULL);
639 INP_RLOCK_ASSERT(inp);
640 if (inp->inp_socket == NULL)
643 error = cr_canseesocket(req->td->td_ucred,
646 cru2x(inp->inp_cred, &xuc);
651 error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred));
655 SYSCTL_PROC(_net_inet6_udp6, OID_AUTO, getcred,
656 CTLTYPE_OPAQUE | CTLFLAG_RW | CTLFLAG_MPSAFE,
657 0, 0, udp6_getcred, "S,xucred",
658 "Get the xucred of a UDP6 connection");
661 udp6_output(struct socket *so, int flags_arg, struct mbuf *m,
662 struct sockaddr *addr6, struct mbuf *control, struct thread *td)
667 struct in6_addr *laddr, *faddr, in6a;
668 struct ip6_pktopts *optp, opt;
669 struct sockaddr_in6 *sin6, tmp;
670 struct epoch_tracker et;
671 int cscov_partial, error, flags, hlen, scope_ambiguous;
672 u_int32_t ulen, plen;
677 /* addr6 has been validated in udp6_send(). */
678 sin6 = (struct sockaddr_in6 *)addr6;
681 * In contrast to IPv4 we do not validate the max. packet length
682 * here due to IPv6 Jumbograms (RFC2675).
687 /* Protect *addr6 from overwrites. */
692 * Application should provide a proper zone ID or the use of
693 * default zone IDs should be enabled. Unfortunately, some
694 * applications do not behave as it should, so we need a
695 * workaround. Even if an appropriate ID is not determined,
696 * we'll see if we can determine the outgoing interface. If we
697 * can, determine the zone ID based on the interface below.
699 if (sin6->sin6_scope_id == 0 && !V_ip6_use_defzone)
701 if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0) {
710 KASSERT(inp != NULL, ("%s: inp == NULL", __func__));
712 * In the following cases we want a write lock on the inp for either
713 * local operations or for possible route cache updates in the IPv6
715 * - on connected sockets (sin6 is NULL) for route cache updates,
716 * - when we are not bound to an address and source port (it is
717 * in6_pcbsetport() which will require the write lock).
719 * We check the inp fields before actually locking the inp, so
720 * here exists a race, and we may WLOCK the inp and end with already
721 * bound one by other thread. This is fine.
723 if (sin6 == NULL || (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
724 inp->inp_lport == 0))
729 nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ?
730 IPPROTO_UDP : IPPROTO_UDPLITE;
733 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
737 hasv4addr = (inp->inp_vflag & INP_IPV4);
739 hasv4addr = IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)
743 * XXXRW: We release UDP-layer locks before calling
744 * udp_send() in order to avoid recursion. However,
745 * this does mean there is a short window where inp's
746 * fields are unstable. Could this lead to a
747 * potential race in which the factors causing us to
748 * select the UDPv4 output routine are invalidated?
752 in6_sin6_2_sin_in_sock((struct sockaddr *)sin6);
753 /* addr will just be freed in sendit(). */
754 return (udp_send(so, flags_arg | PRUS_IPV6, m,
755 (struct sockaddr *)sin6, control, td));
759 if (sin6 && IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
761 * Given this is either an IPv6-only socket or no INET is
762 * supported we will fail the send if the given destination
763 * address is a v4mapped address.
765 * XXXGL: do we leak m and control?
773 if ((error = ip6_setpktopts(control, &opt,
774 inp->in6p_outputopts, td->td_ucred, nxt)) != 0) {
779 optp = inp->in6p_outputopts;
783 * Since we saw no essential reason for calling in_pcbconnect,
784 * we get rid of such kind of logic, and call in6_selectsrc
785 * and in6_pcbsetport in order to fill in the local address
786 * and the local port.
788 if (sin6->sin6_port == 0) {
789 error = EADDRNOTAVAIL;
793 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
794 /* how about ::ffff:0.0.0.0 case? */
800 * Given we handle the v4mapped case in the INET block above
801 * assert here that it must not happen anymore.
803 KASSERT(!IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr),
804 ("%s: sin6(%p)->sin6_addr is v4mapped which we "
805 "should have handled.", __func__, sin6));
807 /* This only requires read-locking. */
808 error = in6_selectsrc_socket(sin6, optp, inp,
809 td->td_ucred, scope_ambiguous, &in6a, NULL);
814 if (inp->inp_lport == 0) {
815 struct inpcbinfo *pcbinfo;
817 INP_WLOCK_ASSERT(inp);
819 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
820 INP_HASH_WLOCK(pcbinfo);
821 error = in6_pcbsetport(laddr, inp, td->td_ucred);
822 INP_HASH_WUNLOCK(pcbinfo);
824 /* Undo an address bind that may have occurred. */
825 inp->in6p_laddr = in6addr_any;
829 faddr = &sin6->sin6_addr;
830 fport = sin6->sin6_port; /* allow 0 port */
833 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
837 laddr = &inp->in6p_laddr;
838 faddr = &inp->in6p_faddr;
839 fport = inp->inp_fport;
842 ulen = m->m_pkthdr.len;
843 plen = sizeof(struct udphdr) + ulen;
844 hlen = sizeof(struct ip6_hdr);
847 * Calculate data length and get a mbuf
848 * for UDP and IP6 headers.
850 M_PREPEND(m, hlen + sizeof(struct udphdr), M_NOWAIT);
857 * Stuff checksum and output datagram.
859 cscov = cscov_partial = 0;
860 udp6 = (struct udphdr *)(mtod(m, caddr_t) + hlen);
861 udp6->uh_sport = inp->inp_lport; /* lport is always set in the PCB */
862 udp6->uh_dport = fport;
863 if (nxt == IPPROTO_UDPLITE) {
867 cscov = up->u_txcslen;
870 udp6->uh_ulen = htons(cscov);
872 * For UDP-Lite, checksum coverage length of zero means
873 * the entire UDPLite packet is covered by the checksum.
875 cscov_partial = (cscov == 0) ? 0 : 1;
876 } else if (plen <= 0xffff)
877 udp6->uh_ulen = htons((u_short)plen);
882 ip6 = mtod(m, struct ip6_hdr *);
883 ip6->ip6_flow = inp->inp_flow & IPV6_FLOWINFO_MASK;
884 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
885 ip6->ip6_vfc |= IPV6_VERSION;
886 ip6->ip6_plen = htons((u_short)plen);
888 ip6->ip6_hlim = in6_selecthlim(inp, NULL);
889 ip6->ip6_src = *laddr;
890 ip6->ip6_dst = *faddr;
893 mac_inpcb_create_mbuf(inp, m);
897 if ((udp6->uh_sum = in6_cksum_partial(m, nxt,
898 sizeof(struct ip6_hdr), plen, cscov)) == 0)
899 udp6->uh_sum = 0xffff;
901 udp6->uh_sum = in6_cksum_pseudo(ip6, plen, nxt, 0);
902 m->m_pkthdr.csum_flags = CSUM_UDP_IPV6;
903 m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
907 #if defined(ROUTE_MPATH) || defined(RSS)
908 if (CALC_FLOWID_OUTBOUND_SENDTO) {
909 uint32_t hash_type, hash_val;
912 pr = inp->inp_socket->so_proto->pr_protocol;
914 hash_val = fib6_calc_packet_hash(laddr, faddr,
915 inp->inp_lport, fport, pr, &hash_type);
916 m->m_pkthdr.flowid = hash_val;
917 M_HASHTYPE_SET(m, hash_type);
919 /* do not use inp flowid */
920 flags |= IP_NODEFAULTFLOWID;
923 UDPSTAT_INC(udps_opackets);
924 if (nxt == IPPROTO_UDPLITE)
925 UDPLITE_PROBE(send, NULL, inp, ip6, inp, udp6);
927 UDP_PROBE(send, NULL, inp, ip6, inp, udp6);
928 error = ip6_output(m, optp,
929 INP_WLOCKED(inp) ? &inp->inp_route6 : NULL, flags,
930 inp->in6p_moptions, NULL, inp);
935 ip6_clearpktopts(&opt, -1);
944 ip6_clearpktopts(&opt, -1);
953 udp6_abort(struct socket *so)
956 struct inpcbinfo *pcbinfo;
958 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
960 KASSERT(inp != NULL, ("udp6_abort: inp == NULL"));
964 if (inp->inp_vflag & INP_IPV4) {
971 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
972 INP_HASH_WLOCK(pcbinfo);
973 in6_pcbdisconnect(inp);
974 inp->in6p_laddr = in6addr_any;
975 INP_HASH_WUNLOCK(pcbinfo);
976 soisdisconnected(so);
982 udp6_attach(struct socket *so, int proto, struct thread *td)
985 struct inpcbinfo *pcbinfo;
988 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
990 KASSERT(inp == NULL, ("udp6_attach: inp != NULL"));
992 if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
993 error = soreserve(so, udp_sendspace, udp_recvspace);
997 error = in_pcballoc(so, pcbinfo);
1000 inp = (struct inpcb *)so->so_pcb;
1001 inp->in6p_cksum = -1; /* just to be sure */
1004 * IPv4 TTL initialization is necessary for an IPv6 socket as well,
1005 * because the socket may be bound to an IPv6 wildcard address,
1006 * which may match an IPv4-mapped IPv6 address.
1008 inp->inp_ip_ttl = V_ip_defttl;
1010 error = udp_newudpcb(inp);
1021 udp6_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
1024 struct inpcbinfo *pcbinfo;
1028 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1029 inp = sotoinpcb(so);
1030 KASSERT(inp != NULL, ("udp6_bind: inp == NULL"));
1032 if (nam->sa_family != AF_INET6)
1033 return (EAFNOSUPPORT);
1034 if (nam->sa_len != sizeof(struct sockaddr_in6))
1038 INP_HASH_WLOCK(pcbinfo);
1039 vflagsav = inp->inp_vflag;
1040 inp->inp_vflag &= ~INP_IPV4;
1041 inp->inp_vflag |= INP_IPV6;
1042 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
1043 struct sockaddr_in6 *sin6_p;
1045 sin6_p = (struct sockaddr_in6 *)nam;
1047 if (IN6_IS_ADDR_UNSPECIFIED(&sin6_p->sin6_addr))
1048 inp->inp_vflag |= INP_IPV4;
1050 else if (IN6_IS_ADDR_V4MAPPED(&sin6_p->sin6_addr)) {
1051 struct sockaddr_in sin;
1053 in6_sin6_2_sin(&sin, sin6_p);
1054 inp->inp_vflag |= INP_IPV4;
1055 inp->inp_vflag &= ~INP_IPV6;
1056 error = in_pcbbind(inp, (struct sockaddr *)&sin,
1063 error = in6_pcbbind(inp, nam, td->td_ucred);
1068 inp->inp_vflag = vflagsav;
1069 INP_HASH_WUNLOCK(pcbinfo);
1075 udp6_close(struct socket *so)
1078 struct inpcbinfo *pcbinfo;
1080 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1081 inp = sotoinpcb(so);
1082 KASSERT(inp != NULL, ("udp6_close: inp == NULL"));
1086 if (inp->inp_vflag & INP_IPV4) {
1088 (void)udp_disconnect(so);
1092 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1093 INP_HASH_WLOCK(pcbinfo);
1094 in6_pcbdisconnect(inp);
1095 inp->in6p_laddr = in6addr_any;
1096 INP_HASH_WUNLOCK(pcbinfo);
1097 soisdisconnected(so);
1103 udp6_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
1106 struct epoch_tracker et;
1109 struct inpcbinfo *pcbinfo;
1110 struct sockaddr_in6 *sin6;
1114 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1115 inp = sotoinpcb(so);
1116 KASSERT(inp != NULL, ("udp6_connect: inp == NULL"));
1118 sin6 = (struct sockaddr_in6 *)nam;
1119 if (sin6->sin6_family != AF_INET6)
1120 return (EAFNOSUPPORT);
1121 if (sin6->sin6_len != sizeof(*sin6))
1125 * XXXRW: Need to clarify locking of v4/v6 flags.
1129 if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
1130 struct sockaddr_in sin;
1132 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) {
1136 if ((inp->inp_vflag & INP_IPV4) == 0) {
1137 error = EAFNOSUPPORT;
1140 if (inp->inp_faddr.s_addr != INADDR_ANY) {
1144 in6_sin6_2_sin(&sin, sin6);
1145 error = prison_remote_ip4(td->td_ucred, &sin.sin_addr);
1148 vflagsav = inp->inp_vflag;
1149 inp->inp_vflag |= INP_IPV4;
1150 inp->inp_vflag &= ~INP_IPV6;
1151 NET_EPOCH_ENTER(et);
1152 INP_HASH_WLOCK(pcbinfo);
1153 error = in_pcbconnect(inp, (struct sockaddr *)&sin,
1154 td->td_ucred, true);
1155 INP_HASH_WUNLOCK(pcbinfo);
1158 * If connect succeeds, mark socket as connected. If
1159 * connect fails and socket is unbound, reset inp_vflag
1164 else if (inp->inp_laddr.s_addr == INADDR_ANY &&
1165 inp->inp_lport == 0)
1166 inp->inp_vflag = vflagsav;
1169 if ((inp->inp_vflag & INP_IPV6) == 0) {
1170 error = EAFNOSUPPORT;
1175 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1179 error = prison_remote_ip6(td->td_ucred, &sin6->sin6_addr);
1182 vflagsav = inp->inp_vflag;
1183 inp->inp_vflag &= ~INP_IPV4;
1184 inp->inp_vflag |= INP_IPV6;
1185 INP_HASH_WLOCK(pcbinfo);
1186 error = in6_pcbconnect(inp, nam, td->td_ucred);
1187 INP_HASH_WUNLOCK(pcbinfo);
1189 * If connect succeeds, mark socket as connected. If
1190 * connect fails and socket is unbound, reset inp_vflag
1195 else if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
1196 inp->inp_lport == 0)
1197 inp->inp_vflag = vflagsav;
1204 udp6_detach(struct socket *so)
1209 inp = sotoinpcb(so);
1210 KASSERT(inp != NULL, ("udp6_detach: inp == NULL"));
1213 up = intoudpcb(inp);
1214 KASSERT(up != NULL, ("%s: up == NULL", __func__));
1221 udp6_disconnect(struct socket *so)
1224 struct inpcbinfo *pcbinfo;
1226 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1227 inp = sotoinpcb(so);
1228 KASSERT(inp != NULL, ("udp6_disconnect: inp == NULL"));
1232 if (inp->inp_vflag & INP_IPV4) {
1234 (void)udp_disconnect(so);
1239 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1244 INP_HASH_WLOCK(pcbinfo);
1245 in6_pcbdisconnect(inp);
1246 inp->in6p_laddr = in6addr_any;
1247 INP_HASH_WUNLOCK(pcbinfo);
1249 so->so_state &= ~SS_ISCONNECTED; /* XXX */
1256 udp6_send(struct socket *so, int flags, struct mbuf *m,
1257 struct sockaddr *addr, struct mbuf *control, struct thread *td)
1262 if (addr->sa_len != sizeof(struct sockaddr_in6)) {
1266 if (addr->sa_family != AF_INET6) {
1267 error = EAFNOSUPPORT;
1272 return (udp6_output(so, flags, m, addr, control, td));
1281 #define UDP6_PROTOSW \
1282 .pr_type = SOCK_DGRAM, \
1283 .pr_flags = PR_ATOMIC|PR_ADDR|PR_CAPATTACH, \
1284 .pr_ctloutput = ip6_ctloutput, \
1285 .pr_abort = udp6_abort, \
1286 .pr_attach = udp6_attach, \
1287 .pr_bind = udp6_bind, \
1288 .pr_connect = udp6_connect, \
1289 .pr_control = in6_control, \
1290 .pr_detach = udp6_detach, \
1291 .pr_disconnect = udp6_disconnect, \
1292 .pr_peeraddr = in6_mapped_peeraddr, \
1293 .pr_send = udp6_send, \
1294 .pr_shutdown = udp_shutdown, \
1295 .pr_sockaddr = in6_mapped_sockaddr, \
1296 .pr_soreceive = soreceive_dgram, \
1297 .pr_sosend = sosend_dgram, \
1298 .pr_sosetlabel = in_pcbsosetlabel, \
1299 .pr_close = udp6_close
1301 struct protosw udp6_protosw = {
1302 .pr_protocol = IPPROTO_UDP,
1306 struct protosw udplite6_protosw = {
1307 .pr_protocol = IPPROTO_UDPLITE,
1312 udp6_init(void *arg __unused)
1315 IP6PROTO_REGISTER(IPPROTO_UDP, udp6_input, udp6_ctlinput);
1316 IP6PROTO_REGISTER(IPPROTO_UDPLITE, udp6_input, udplite6_ctlinput);
1318 SYSINIT(udp6_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, udp6_init, NULL);