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/ip6protosw.h>
118 #include <netinet6/ip6_var.h>
119 #include <netinet6/in6_fib.h>
120 #include <netinet6/in6_pcb.h>
121 #include <netinet6/in6_rss.h>
122 #include <netinet6/udp6_var.h>
123 #include <netinet6/scope6_var.h>
125 #include <netipsec/ipsec_support.h>
127 #include <security/mac/mac_framework.h>
129 VNET_DEFINE(int, zero_checksum_port) = 0;
130 #define V_zero_checksum_port VNET(zero_checksum_port)
131 SYSCTL_INT(_net_inet6_udp6, OID_AUTO, rfc6935_port, CTLFLAG_VNET | CTLFLAG_RW,
132 &VNET_NAME(zero_checksum_port), 0,
133 "Zero UDP checksum allowed for traffic to/from this port.");
135 * UDP protocol implementation.
136 * Per RFC 768, August, 1980.
139 extern struct protosw inetsw[];
140 static void udp6_detach(struct socket *so);
143 udp6_append(struct inpcb *inp, struct mbuf *n, int off,
144 struct sockaddr_in6 *fromsa)
147 struct mbuf *opts = NULL, *tmp_opts;
150 INP_LOCK_ASSERT(inp);
153 * Engage the tunneling protocol.
156 if (up->u_tun_func != NULL) {
159 (*up->u_tun_func)(n, off, inp, (struct sockaddr *)&fromsa[0],
162 return (in_pcbrele_rlocked(inp));
164 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
165 /* Check AH/ESP integrity. */
166 if (IPSEC_ENABLED(ipv6)) {
167 if (IPSEC_CHECK_POLICY(ipv6, n, inp) != 0) {
174 if (mac_inpcb_check_deliver(inp, n) != 0) {
180 if (inp->inp_flags & INP_CONTROLOPTS ||
181 inp->inp_socket->so_options & SO_TIMESTAMP)
182 ip6_savecontrol(inp, n, &opts);
183 if ((inp->inp_vflag & INP_IPV6) && (inp->inp_flags2 & INP_ORIGDSTADDR)) {
184 tmp_opts = sbcreatecontrol((caddr_t)&fromsa[1],
185 sizeof(struct sockaddr_in6), IPV6_ORIGDSTADDR, IPPROTO_IPV6);
188 tmp_opts->m_next = opts;
194 m_adj(n, off + sizeof(struct udphdr));
196 so = inp->inp_socket;
197 SOCKBUF_LOCK(&so->so_rcv);
198 if (sbappendaddr_locked(&so->so_rcv, (struct sockaddr *)&fromsa[0], n,
200 soroverflow_locked(so);
204 UDPSTAT_INC(udps_fullsock);
206 sorwakeup_locked(so);
210 struct udp6_multi_match_ctx {
216 udp6_multi_match(const struct inpcb *inp, void *v)
218 struct udp6_multi_match_ctx *ctx = v;
220 if ((inp->inp_vflag & INP_IPV6) == 0)
222 if (inp->inp_lport != ctx->uh->uh_dport)
224 if (inp->inp_fport != 0 && inp->inp_fport != ctx->uh->uh_sport)
226 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
227 !IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, &ctx->ip6->ip6_dst))
229 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) &&
230 (!IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &ctx->ip6->ip6_src) ||
231 inp->inp_fport != ctx->uh->uh_sport))
238 udp6_multi_input(struct mbuf *m, int off, int proto,
239 struct sockaddr_in6 *fromsa)
241 struct udp6_multi_match_ctx ctx;
242 struct inpcb_iterator inpi = INP_ITERATOR(udp_get_inpcbinfo(proto),
243 INPLOOKUP_RLOCKPCB, udp6_multi_match, &ctx);
245 struct ip6_moptions *imo;
250 * In the event that laddr should be set to the link-local
251 * address (this happens in RIPng), the multicast address
252 * specified in the received packet will not match laddr. To
253 * handle this situation, matching is relaxed if the
254 * receiving interface is the same as one specified in the
255 * socket and if the destination multicast address matches
256 * one of the multicast groups specified in the socket.
260 * KAME note: traditionally we dropped udpiphdr from mbuf
261 * here. We need udphdr for IPsec processing so we do that
264 ctx.ip6 = mtod(m, struct ip6_hdr *);
265 ctx.uh = (struct udphdr *)((char *)ctx.ip6 + off);
266 while ((inp = inp_next(&inpi)) != NULL) {
267 INP_RLOCK_ASSERT(inp);
269 * XXXRW: Because we weren't holding either the inpcb
270 * or the hash lock when we checked for a match
271 * before, we should probably recheck now that the
272 * inpcb lock is (supposed to be) held.
275 * Handle socket delivery policy for any-source
276 * and source-specific multicast. [RFC3678]
278 if ((imo = inp->in6p_moptions) != NULL) {
279 struct sockaddr_in6 mcaddr;
282 bzero(&mcaddr, sizeof(struct sockaddr_in6));
283 mcaddr.sin6_len = sizeof(struct sockaddr_in6);
284 mcaddr.sin6_family = AF_INET6;
285 mcaddr.sin6_addr = ctx.ip6->ip6_dst;
287 blocked = im6o_mc_filter(imo, m->m_pkthdr.rcvif,
288 (struct sockaddr *)&mcaddr,
289 (struct sockaddr *)&fromsa[0]);
290 if (blocked != MCAST_PASS) {
291 if (blocked == MCAST_NOTGMEMBER)
292 IP6STAT_INC(ip6s_notmember);
293 if (blocked == MCAST_NOTSMEMBER ||
294 blocked == MCAST_MUTED)
295 UDPSTAT_INC(udps_filtermcast);
299 if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) != NULL) {
300 if (proto == IPPROTO_UDPLITE)
301 UDPLITE_PROBE(receive, NULL, inp, ctx.ip6,
304 UDP_PROBE(receive, NULL, inp, ctx.ip6, inp,
306 if (udp6_append(inp, n, off, fromsa)) {
313 * Don't look for additional matches if this one does
314 * not have either the SO_REUSEPORT or SO_REUSEADDR
315 * socket options set. This heuristic avoids
316 * searching through all pcbs in the common case of a
317 * non-shared port. It assumes that an application
318 * will never clear these options after setting them.
320 if ((inp->inp_socket->so_options &
321 (SO_REUSEPORT|SO_REUSEPORT_LB|SO_REUSEADDR)) == 0) {
330 * No matching pcb found; discard datagram. (No need
331 * to send an ICMP Port Unreachable for a broadcast
332 * or multicast datgram.)
334 UDPSTAT_INC(udps_noport);
335 UDPSTAT_INC(udps_noportmcast);
338 return (IPPROTO_DONE);
342 udp6_input(struct mbuf **mp, int *offp, int proto)
344 struct mbuf *m = *mp;
348 struct inpcbinfo *pcbinfo;
353 struct sockaddr_in6 fromsa[2];
354 struct m_tag *fwd_tag;
360 if (m->m_len < off + sizeof(struct udphdr)) {
361 m = m_pullup(m, off + sizeof(struct udphdr));
363 IP6STAT_INC(ip6s_exthdrtoolong);
365 return (IPPROTO_DONE);
368 ip6 = mtod(m, struct ip6_hdr *);
369 uh = (struct udphdr *)((caddr_t)ip6 + off);
371 UDPSTAT_INC(udps_ipackets);
374 * Destination port of 0 is illegal, based on RFC768.
376 if (uh->uh_dport == 0)
379 plen = ntohs(ip6->ip6_plen) - off + sizeof(*ip6);
380 ulen = ntohs((u_short)uh->uh_ulen);
383 cscov_partial = (nxt == IPPROTO_UDPLITE) ? 1 : 0;
384 if (nxt == IPPROTO_UDPLITE) {
385 /* Zero means checksum over the complete packet. */
390 if ((ulen < sizeof(struct udphdr)) || (ulen > plen)) {
391 /* XXX: What is the right UDPLite MIB counter? */
394 if (uh->uh_sum == 0) {
395 /* XXX: What is the right UDPLite MIB counter? */
399 if ((ulen < sizeof(struct udphdr)) || (plen != ulen)) {
400 UDPSTAT_INC(udps_badlen);
403 if (uh->uh_sum == 0) {
404 UDPSTAT_INC(udps_nosum);
406 * dport 0 was rejected earlier so this is OK even if
407 * zero_checksum_port is 0 (which is its default value).
409 if (ntohs(uh->uh_dport) == V_zero_checksum_port)
416 if ((m->m_pkthdr.csum_flags & CSUM_DATA_VALID_IPV6) &&
418 if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR)
419 uh_sum = m->m_pkthdr.csum_data;
421 uh_sum = in6_cksum_pseudo(ip6, ulen, nxt,
422 m->m_pkthdr.csum_data);
425 uh_sum = in6_cksum_partial(m, nxt, off, plen, ulen);
428 UDPSTAT_INC(udps_badsum);
434 * Construct sockaddr format source address.
436 init_sin6(&fromsa[0], m, 0);
437 fromsa[0].sin6_port = uh->uh_sport;
438 init_sin6(&fromsa[1], m, 1);
439 fromsa[1].sin6_port = uh->uh_dport;
441 pcbinfo = udp_get_inpcbinfo(nxt);
442 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
444 return (udp6_multi_input(m, off, proto, fromsa));
448 * Locate pcb for datagram.
452 * Grab info from PACKET_TAG_IPFORWARD tag prepended to the chain.
454 if ((m->m_flags & M_IP6_NEXTHOP) &&
455 (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
456 struct sockaddr_in6 *next_hop6;
458 next_hop6 = (struct sockaddr_in6 *)(fwd_tag + 1);
461 * Transparently forwarded. Pretend to be the destination.
462 * Already got one like this?
464 inp = in6_pcblookup_mbuf(pcbinfo, &ip6->ip6_src,
465 uh->uh_sport, &ip6->ip6_dst, uh->uh_dport,
466 INPLOOKUP_RLOCKPCB, m->m_pkthdr.rcvif, m);
469 * It's new. Try to find the ambushing socket.
470 * Because we've rewritten the destination address,
471 * any hardware-generated hash is ignored.
473 inp = in6_pcblookup(pcbinfo, &ip6->ip6_src,
474 uh->uh_sport, &next_hop6->sin6_addr,
475 next_hop6->sin6_port ? htons(next_hop6->sin6_port) :
476 uh->uh_dport, INPLOOKUP_WILDCARD |
477 INPLOOKUP_RLOCKPCB, m->m_pkthdr.rcvif);
479 /* Remove the tag from the packet. We don't need it anymore. */
480 m_tag_delete(m, fwd_tag);
481 m->m_flags &= ~M_IP6_NEXTHOP;
483 inp = in6_pcblookup_mbuf(pcbinfo, &ip6->ip6_src,
484 uh->uh_sport, &ip6->ip6_dst, uh->uh_dport,
485 INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB,
486 m->m_pkthdr.rcvif, m);
488 if (V_udp_log_in_vain) {
489 char ip6bufs[INET6_ADDRSTRLEN];
490 char ip6bufd[INET6_ADDRSTRLEN];
493 "Connection attempt to UDP [%s]:%d from [%s]:%d\n",
494 ip6_sprintf(ip6bufd, &ip6->ip6_dst),
496 ip6_sprintf(ip6bufs, &ip6->ip6_src),
497 ntohs(uh->uh_sport));
499 if (nxt == IPPROTO_UDPLITE)
500 UDPLITE_PROBE(receive, NULL, NULL, ip6, NULL, uh);
502 UDP_PROBE(receive, NULL, NULL, ip6, NULL, uh);
503 UDPSTAT_INC(udps_noport);
504 if (m->m_flags & M_MCAST) {
505 printf("UDP6: M_MCAST is set in a unicast packet.\n");
506 UDPSTAT_INC(udps_noportmcast);
509 if (V_udp_blackhole && (V_udp_blackhole_local ||
510 !in6_localaddr(&ip6->ip6_src)))
512 icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0);
514 return (IPPROTO_DONE);
516 INP_RLOCK_ASSERT(inp);
519 if (up->u_rxcslen == 0 || up->u_rxcslen > ulen) {
523 return (IPPROTO_DONE);
526 if (nxt == IPPROTO_UDPLITE)
527 UDPLITE_PROBE(receive, NULL, inp, ip6, inp, uh);
529 UDP_PROBE(receive, NULL, inp, ip6, inp, uh);
530 if (udp6_append(inp, m, off, fromsa) == 0)
533 return (IPPROTO_DONE);
538 return (IPPROTO_DONE);
542 udp6_common_ctlinput(int cmd, struct sockaddr *sa, void *d,
543 struct inpcbinfo *pcbinfo)
549 struct ip6ctlparam *ip6cp = NULL;
550 const struct sockaddr_in6 *sa6_src = NULL;
552 struct inpcb *(*notify)(struct inpcb *, int) = udp_notify;
553 struct udp_portonly {
558 if (sa->sa_family != AF_INET6 ||
559 sa->sa_len != sizeof(struct sockaddr_in6))
562 if ((unsigned)cmd >= PRC_NCMDS)
564 if (PRC_IS_REDIRECT(cmd))
565 notify = in6_rtchange, d = NULL;
566 else if (cmd == PRC_HOSTDEAD)
568 else if (inet6ctlerrmap[cmd] == 0)
571 /* if the parameter is from icmp6, decode it. */
573 ip6cp = (struct ip6ctlparam *)d;
575 ip6 = ip6cp->ip6c_ip6;
576 off = ip6cp->ip6c_off;
577 cmdarg = ip6cp->ip6c_cmdarg;
578 sa6_src = ip6cp->ip6c_src;
588 * XXX: We assume that when IPV6 is non NULL,
589 * M and OFF are valid.
592 /* Check if we can safely examine src and dst ports. */
593 if (m->m_pkthdr.len < off + sizeof(*uhp))
596 bzero(&uh, sizeof(uh));
597 m_copydata(m, off, sizeof(*uhp), (caddr_t)&uh);
599 if (!PRC_IS_REDIRECT(cmd)) {
600 /* Check to see if its tunneled */
602 inp = in6_pcblookup_mbuf(pcbinfo, &ip6->ip6_dst,
603 uh.uh_dport, &ip6->ip6_src, uh.uh_sport,
604 INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB,
605 m->m_pkthdr.rcvif, m);
610 if (up->u_icmp_func) {
613 (*up->u_icmp_func)(cmd, (struct sockaddr *)ip6cp->ip6c_src,
622 (void)in6_pcbnotify(pcbinfo, sa, uh.uh_dport,
623 (struct sockaddr *)ip6cp->ip6c_src, uh.uh_sport, cmd,
626 (void)in6_pcbnotify(pcbinfo, sa, 0,
627 (const struct sockaddr *)sa6_src, 0, cmd, cmdarg, notify);
631 udp6_ctlinput(int cmd, struct sockaddr *sa, void *d)
634 return (udp6_common_ctlinput(cmd, sa, d, &V_udbinfo));
638 udplite6_ctlinput(int cmd, struct sockaddr *sa, void *d)
641 return (udp6_common_ctlinput(cmd, sa, d, &V_ulitecbinfo));
645 udp6_getcred(SYSCTL_HANDLER_ARGS)
648 struct sockaddr_in6 addrs[2];
649 struct epoch_tracker et;
653 error = priv_check(req->td, PRIV_NETINET_GETCRED);
657 if (req->newlen != sizeof(addrs))
659 if (req->oldlen != sizeof(struct xucred))
661 error = SYSCTL_IN(req, addrs, sizeof(addrs));
664 if ((error = sa6_embedscope(&addrs[0], V_ip6_use_defzone)) != 0 ||
665 (error = sa6_embedscope(&addrs[1], V_ip6_use_defzone)) != 0) {
669 inp = in6_pcblookup(&V_udbinfo, &addrs[1].sin6_addr,
670 addrs[1].sin6_port, &addrs[0].sin6_addr, addrs[0].sin6_port,
671 INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB, NULL);
674 INP_RLOCK_ASSERT(inp);
675 if (inp->inp_socket == NULL)
678 error = cr_canseesocket(req->td->td_ucred,
681 cru2x(inp->inp_cred, &xuc);
686 error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred));
690 SYSCTL_PROC(_net_inet6_udp6, OID_AUTO, getcred,
691 CTLTYPE_OPAQUE | CTLFLAG_RW | CTLFLAG_MPSAFE,
692 0, 0, udp6_getcred, "S,xucred",
693 "Get the xucred of a UDP6 connection");
696 udp6_output(struct socket *so, int flags_arg, struct mbuf *m,
697 struct sockaddr *addr6, struct mbuf *control, struct thread *td)
702 struct in6_addr *laddr, *faddr, in6a;
703 struct ip6_pktopts *optp, opt;
704 struct sockaddr_in6 *sin6, tmp;
705 struct epoch_tracker et;
706 int cscov_partial, error, flags, hlen, scope_ambiguous;
707 u_int32_t ulen, plen;
712 /* addr6 has been validated in udp6_send(). */
713 sin6 = (struct sockaddr_in6 *)addr6;
716 * In contrast to to IPv4 we do not validate the max. packet length
717 * here due to IPv6 Jumbograms (RFC2675).
722 /* Protect *addr6 from overwrites. */
727 * Application should provide a proper zone ID or the use of
728 * default zone IDs should be enabled. Unfortunately, some
729 * applications do not behave as it should, so we need a
730 * workaround. Even if an appropriate ID is not determined,
731 * we'll see if we can determine the outgoing interface. If we
732 * can, determine the zone ID based on the interface below.
734 if (sin6->sin6_scope_id == 0 && !V_ip6_use_defzone)
736 if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0) {
745 KASSERT(inp != NULL, ("%s: inp == NULL", __func__));
747 * In the following cases we want a write lock on the inp for either
748 * local operations or for possible route cache updates in the IPv6
750 * - on connected sockets (sin6 is NULL) for route cache updates,
751 * - when we are not bound to an address and source port (it is
752 * in6_pcbsetport() which will require the write lock).
754 * We check the inp fields before actually locking the inp, so
755 * here exists a race, and we may WLOCK the inp and end with already
756 * bound one by other thread. This is fine.
758 if (sin6 == NULL || (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
759 inp->inp_lport == 0))
764 nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ?
765 IPPROTO_UDP : IPPROTO_UDPLITE;
768 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
772 hasv4addr = (inp->inp_vflag & INP_IPV4);
774 hasv4addr = IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)
777 struct pr_usrreqs *pru;
780 * XXXRW: We release UDP-layer locks before calling
781 * udp_send() in order to avoid recursion. However,
782 * this does mean there is a short window where inp's
783 * fields are unstable. Could this lead to a
784 * potential race in which the factors causing us to
785 * select the UDPv4 output routine are invalidated?
789 in6_sin6_2_sin_in_sock((struct sockaddr *)sin6);
790 pru = inetsw[ip_protox[nxt]].pr_usrreqs;
791 /* addr will just be freed in sendit(). */
792 return ((*pru->pru_send)(so, flags_arg | PRUS_IPV6, m,
793 (struct sockaddr *)sin6, control, td));
797 if (sin6 && IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
799 * Given this is either an IPv6-only socket or no INET is
800 * supported we will fail the send if the given destination
801 * address is a v4mapped address.
803 * XXXGL: do we leak m and control?
811 if ((error = ip6_setpktopts(control, &opt,
812 inp->in6p_outputopts, td->td_ucred, nxt)) != 0) {
817 optp = inp->in6p_outputopts;
821 * Since we saw no essential reason for calling in_pcbconnect,
822 * we get rid of such kind of logic, and call in6_selectsrc
823 * and in6_pcbsetport in order to fill in the local address
824 * and the local port.
826 if (sin6->sin6_port == 0) {
827 error = EADDRNOTAVAIL;
831 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
832 /* how about ::ffff:0.0.0.0 case? */
838 * Given we handle the v4mapped case in the INET block above
839 * assert here that it must not happen anymore.
841 KASSERT(!IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr),
842 ("%s: sin6(%p)->sin6_addr is v4mapped which we "
843 "should have handled.", __func__, sin6));
845 /* This only requires read-locking. */
846 error = in6_selectsrc_socket(sin6, optp, inp,
847 td->td_ucred, scope_ambiguous, &in6a, NULL);
852 if (inp->inp_lport == 0) {
853 struct inpcbinfo *pcbinfo;
855 INP_WLOCK_ASSERT(inp);
857 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
858 INP_HASH_WLOCK(pcbinfo);
859 error = in6_pcbsetport(laddr, inp, td->td_ucred);
860 INP_HASH_WUNLOCK(pcbinfo);
862 /* Undo an address bind that may have occurred. */
863 inp->in6p_laddr = in6addr_any;
867 faddr = &sin6->sin6_addr;
868 fport = sin6->sin6_port; /* allow 0 port */
871 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
875 laddr = &inp->in6p_laddr;
876 faddr = &inp->in6p_faddr;
877 fport = inp->inp_fport;
880 ulen = m->m_pkthdr.len;
881 plen = sizeof(struct udphdr) + ulen;
882 hlen = sizeof(struct ip6_hdr);
885 * Calculate data length and get a mbuf
886 * for UDP and IP6 headers.
888 M_PREPEND(m, hlen + sizeof(struct udphdr), M_NOWAIT);
895 * Stuff checksum and output datagram.
897 cscov = cscov_partial = 0;
898 udp6 = (struct udphdr *)(mtod(m, caddr_t) + hlen);
899 udp6->uh_sport = inp->inp_lport; /* lport is always set in the PCB */
900 udp6->uh_dport = fport;
901 if (nxt == IPPROTO_UDPLITE) {
905 cscov = up->u_txcslen;
908 udp6->uh_ulen = htons(cscov);
910 * For UDP-Lite, checksum coverage length of zero means
911 * the entire UDPLite packet is covered by the checksum.
913 cscov_partial = (cscov == 0) ? 0 : 1;
914 } else if (plen <= 0xffff)
915 udp6->uh_ulen = htons((u_short)plen);
920 ip6 = mtod(m, struct ip6_hdr *);
921 ip6->ip6_flow = inp->inp_flow & IPV6_FLOWINFO_MASK;
922 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
923 ip6->ip6_vfc |= IPV6_VERSION;
924 ip6->ip6_plen = htons((u_short)plen);
926 ip6->ip6_hlim = in6_selecthlim(inp, NULL);
927 ip6->ip6_src = *laddr;
928 ip6->ip6_dst = *faddr;
931 mac_inpcb_create_mbuf(inp, m);
935 if ((udp6->uh_sum = in6_cksum_partial(m, nxt,
936 sizeof(struct ip6_hdr), plen, cscov)) == 0)
937 udp6->uh_sum = 0xffff;
939 udp6->uh_sum = in6_cksum_pseudo(ip6, plen, nxt, 0);
940 m->m_pkthdr.csum_flags = CSUM_UDP_IPV6;
941 m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
945 #if defined(ROUTE_MPATH) || defined(RSS)
946 if (CALC_FLOWID_OUTBOUND_SENDTO) {
947 uint32_t hash_type, hash_val;
950 pr = inp->inp_socket->so_proto->pr_protocol;
952 hash_val = fib6_calc_packet_hash(laddr, faddr,
953 inp->inp_lport, fport, pr, &hash_type);
954 m->m_pkthdr.flowid = hash_val;
955 M_HASHTYPE_SET(m, hash_type);
957 /* do not use inp flowid */
958 flags |= IP_NODEFAULTFLOWID;
961 UDPSTAT_INC(udps_opackets);
962 if (nxt == IPPROTO_UDPLITE)
963 UDPLITE_PROBE(send, NULL, inp, ip6, inp, udp6);
965 UDP_PROBE(send, NULL, inp, ip6, inp, udp6);
966 error = ip6_output(m, optp,
967 INP_WLOCKED(inp) ? &inp->inp_route6 : NULL, flags,
968 inp->in6p_moptions, NULL, inp);
973 ip6_clearpktopts(&opt, -1);
982 ip6_clearpktopts(&opt, -1);
991 udp6_abort(struct socket *so)
994 struct inpcbinfo *pcbinfo;
996 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
998 KASSERT(inp != NULL, ("udp6_abort: inp == NULL"));
1002 if (inp->inp_vflag & INP_IPV4) {
1003 struct pr_usrreqs *pru;
1006 nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ?
1007 IPPROTO_UDP : IPPROTO_UDPLITE;
1009 pru = inetsw[ip_protox[nxt]].pr_usrreqs;
1010 (*pru->pru_abort)(so);
1015 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1016 INP_HASH_WLOCK(pcbinfo);
1017 in6_pcbdisconnect(inp);
1018 inp->in6p_laddr = in6addr_any;
1019 INP_HASH_WUNLOCK(pcbinfo);
1020 soisdisconnected(so);
1026 udp6_attach(struct socket *so, int proto, struct thread *td)
1029 struct inpcbinfo *pcbinfo;
1032 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1033 inp = sotoinpcb(so);
1034 KASSERT(inp == NULL, ("udp6_attach: inp != NULL"));
1036 if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
1037 error = soreserve(so, udp_sendspace, udp_recvspace);
1041 error = in_pcballoc(so, pcbinfo);
1044 inp = (struct inpcb *)so->so_pcb;
1045 inp->inp_vflag |= INP_IPV6;
1046 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0)
1047 inp->inp_vflag |= INP_IPV4;
1048 inp->in6p_hops = -1; /* use kernel default */
1049 inp->in6p_cksum = -1; /* just to be sure */
1052 * IPv4 TTL initialization is necessary for an IPv6 socket as well,
1053 * because the socket may be bound to an IPv6 wildcard address,
1054 * which may match an IPv4-mapped IPv6 address.
1056 inp->inp_ip_ttl = V_ip_defttl;
1058 error = udp_newudpcb(inp);
1069 udp6_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
1072 struct inpcbinfo *pcbinfo;
1076 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1077 inp = sotoinpcb(so);
1078 KASSERT(inp != NULL, ("udp6_bind: inp == NULL"));
1080 if (nam->sa_family != AF_INET6)
1081 return (EAFNOSUPPORT);
1082 if (nam->sa_len != sizeof(struct sockaddr_in6))
1086 INP_HASH_WLOCK(pcbinfo);
1087 vflagsav = inp->inp_vflag;
1088 inp->inp_vflag &= ~INP_IPV4;
1089 inp->inp_vflag |= INP_IPV6;
1090 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
1091 struct sockaddr_in6 *sin6_p;
1093 sin6_p = (struct sockaddr_in6 *)nam;
1095 if (IN6_IS_ADDR_UNSPECIFIED(&sin6_p->sin6_addr))
1096 inp->inp_vflag |= INP_IPV4;
1098 else if (IN6_IS_ADDR_V4MAPPED(&sin6_p->sin6_addr)) {
1099 struct sockaddr_in sin;
1101 in6_sin6_2_sin(&sin, sin6_p);
1102 inp->inp_vflag |= INP_IPV4;
1103 inp->inp_vflag &= ~INP_IPV6;
1104 error = in_pcbbind(inp, (struct sockaddr *)&sin,
1111 error = in6_pcbbind(inp, nam, td->td_ucred);
1116 inp->inp_vflag = vflagsav;
1117 INP_HASH_WUNLOCK(pcbinfo);
1123 udp6_close(struct socket *so)
1126 struct inpcbinfo *pcbinfo;
1128 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1129 inp = sotoinpcb(so);
1130 KASSERT(inp != NULL, ("udp6_close: inp == NULL"));
1134 if (inp->inp_vflag & INP_IPV4) {
1135 struct pr_usrreqs *pru;
1138 nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ?
1139 IPPROTO_UDP : IPPROTO_UDPLITE;
1141 pru = inetsw[ip_protox[nxt]].pr_usrreqs;
1142 (*pru->pru_disconnect)(so);
1146 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1147 INP_HASH_WLOCK(pcbinfo);
1148 in6_pcbdisconnect(inp);
1149 inp->in6p_laddr = in6addr_any;
1150 INP_HASH_WUNLOCK(pcbinfo);
1151 soisdisconnected(so);
1157 udp6_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
1160 struct epoch_tracker et;
1163 struct inpcbinfo *pcbinfo;
1164 struct sockaddr_in6 *sin6;
1168 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1169 inp = sotoinpcb(so);
1170 KASSERT(inp != NULL, ("udp6_connect: inp == NULL"));
1172 sin6 = (struct sockaddr_in6 *)nam;
1173 if (sin6->sin6_family != AF_INET6)
1174 return (EAFNOSUPPORT);
1175 if (sin6->sin6_len != sizeof(*sin6))
1179 * XXXRW: Need to clarify locking of v4/v6 flags.
1183 if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
1184 struct sockaddr_in sin;
1186 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) {
1190 if ((inp->inp_vflag & INP_IPV4) == 0) {
1191 error = EAFNOSUPPORT;
1194 if (inp->inp_faddr.s_addr != INADDR_ANY) {
1198 in6_sin6_2_sin(&sin, sin6);
1199 error = prison_remote_ip4(td->td_ucred, &sin.sin_addr);
1202 vflagsav = inp->inp_vflag;
1203 inp->inp_vflag |= INP_IPV4;
1204 inp->inp_vflag &= ~INP_IPV6;
1205 NET_EPOCH_ENTER(et);
1206 INP_HASH_WLOCK(pcbinfo);
1207 error = in_pcbconnect(inp, (struct sockaddr *)&sin,
1208 td->td_ucred, true);
1209 INP_HASH_WUNLOCK(pcbinfo);
1212 * If connect succeeds, mark socket as connected. If
1213 * connect fails and socket is unbound, reset inp_vflag
1218 else if (inp->inp_laddr.s_addr == INADDR_ANY &&
1219 inp->inp_lport == 0)
1220 inp->inp_vflag = vflagsav;
1223 if ((inp->inp_vflag & INP_IPV6) == 0) {
1224 error = EAFNOSUPPORT;
1229 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1233 error = prison_remote_ip6(td->td_ucred, &sin6->sin6_addr);
1236 vflagsav = inp->inp_vflag;
1237 inp->inp_vflag &= ~INP_IPV4;
1238 inp->inp_vflag |= INP_IPV6;
1239 INP_HASH_WLOCK(pcbinfo);
1240 error = in6_pcbconnect(inp, nam, td->td_ucred);
1241 INP_HASH_WUNLOCK(pcbinfo);
1243 * If connect succeeds, mark socket as connected. If
1244 * connect fails and socket is unbound, reset inp_vflag
1249 else if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
1250 inp->inp_lport == 0)
1251 inp->inp_vflag = vflagsav;
1258 udp6_detach(struct socket *so)
1263 inp = sotoinpcb(so);
1264 KASSERT(inp != NULL, ("udp6_detach: inp == NULL"));
1267 up = intoudpcb(inp);
1268 KASSERT(up != NULL, ("%s: up == NULL", __func__));
1275 udp6_disconnect(struct socket *so)
1278 struct inpcbinfo *pcbinfo;
1280 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1281 inp = sotoinpcb(so);
1282 KASSERT(inp != NULL, ("udp6_disconnect: inp == NULL"));
1286 if (inp->inp_vflag & INP_IPV4) {
1287 struct pr_usrreqs *pru;
1290 nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ?
1291 IPPROTO_UDP : IPPROTO_UDPLITE;
1293 pru = inetsw[ip_protox[nxt]].pr_usrreqs;
1294 (void)(*pru->pru_disconnect)(so);
1299 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1304 INP_HASH_WLOCK(pcbinfo);
1305 in6_pcbdisconnect(inp);
1306 inp->in6p_laddr = in6addr_any;
1307 INP_HASH_WUNLOCK(pcbinfo);
1309 so->so_state &= ~SS_ISCONNECTED; /* XXX */
1316 udp6_send(struct socket *so, int flags, struct mbuf *m,
1317 struct sockaddr *addr, struct mbuf *control, struct thread *td)
1322 if (addr->sa_len != sizeof(struct sockaddr_in6)) {
1326 if (addr->sa_family != AF_INET6) {
1327 error = EAFNOSUPPORT;
1332 return (udp6_output(so, flags, m, addr, control, td));
1341 struct pr_usrreqs udp6_usrreqs = {
1342 .pru_abort = udp6_abort,
1343 .pru_attach = udp6_attach,
1344 .pru_bind = udp6_bind,
1345 .pru_connect = udp6_connect,
1346 .pru_control = in6_control,
1347 .pru_detach = udp6_detach,
1348 .pru_disconnect = udp6_disconnect,
1349 .pru_peeraddr = in6_mapped_peeraddr,
1350 .pru_send = udp6_send,
1351 .pru_shutdown = udp_shutdown,
1352 .pru_sockaddr = in6_mapped_sockaddr,
1353 .pru_soreceive = soreceive_dgram,
1354 .pru_sosend = sosend_dgram,
1355 .pru_sosetlabel = in_pcbsosetlabel,
1356 .pru_close = udp6_close