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"
80 #include <sys/param.h>
82 #include <sys/kernel.h>
87 #include <sys/protosw.h>
89 #include <sys/signalvar.h>
90 #include <sys/socket.h>
91 #include <sys/socketvar.h>
93 #include <sys/sysctl.h>
94 #include <sys/syslog.h>
95 #include <sys/systm.h>
98 #include <net/if_var.h>
99 #include <net/if_types.h>
100 #include <net/route.h>
101 #include <net/rss_config.h>
103 #include <netinet/in.h>
104 #include <netinet/in_kdtrace.h>
105 #include <netinet/in_pcb.h>
106 #include <netinet/in_systm.h>
107 #include <netinet/in_var.h>
108 #include <netinet/ip.h>
109 #include <netinet/ip6.h>
110 #include <netinet/icmp6.h>
111 #include <netinet/ip_var.h>
112 #include <netinet/udp.h>
113 #include <netinet/udp_var.h>
114 #include <netinet/udplite.h>
116 #include <netinet6/ip6protosw.h>
117 #include <netinet6/ip6_var.h>
118 #include <netinet6/in6_pcb.h>
119 #include <netinet6/in6_rss.h>
120 #include <netinet6/udp6_var.h>
121 #include <netinet6/scope6_var.h>
123 #include <netipsec/ipsec_support.h>
125 #include <security/mac/mac_framework.h>
127 VNET_DEFINE(int, zero_checksum_port) = 0;
128 #define V_zero_checksum_port VNET(zero_checksum_port)
129 SYSCTL_INT(_net_inet6_udp6, OID_AUTO, rfc6935_port, CTLFLAG_VNET | CTLFLAG_RW,
130 &VNET_NAME(zero_checksum_port), 0,
131 "Zero UDP checksum allowed for traffic to/from this port.");
133 * UDP protocol implementation.
134 * Per RFC 768, August, 1980.
137 extern struct protosw inetsw[];
138 static void udp6_detach(struct socket *so);
141 udp6_append(struct inpcb *inp, struct mbuf *n, int off,
142 struct sockaddr_in6 *fromsa)
145 struct mbuf *opts = NULL, *tmp_opts;
148 INP_LOCK_ASSERT(inp);
151 * Engage the tunneling protocol.
154 if (up->u_tun_func != NULL) {
157 (*up->u_tun_func)(n, off, inp, (struct sockaddr *)&fromsa[0],
160 return (in_pcbrele_rlocked(inp));
162 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
163 /* Check AH/ESP integrity. */
164 if (IPSEC_ENABLED(ipv6)) {
165 if (IPSEC_CHECK_POLICY(ipv6, n, inp) != 0) {
172 if (mac_inpcb_check_deliver(inp, n) != 0) {
178 if (inp->inp_flags & INP_CONTROLOPTS ||
179 inp->inp_socket->so_options & SO_TIMESTAMP)
180 ip6_savecontrol(inp, n, &opts);
181 if ((inp->inp_vflag & INP_IPV6) && (inp->inp_flags2 & INP_ORIGDSTADDR)) {
182 tmp_opts = sbcreatecontrol((caddr_t)&fromsa[1],
183 sizeof(struct sockaddr_in6), IPV6_ORIGDSTADDR, IPPROTO_IPV6);
186 tmp_opts->m_next = opts;
192 m_adj(n, off + sizeof(struct udphdr));
194 so = inp->inp_socket;
195 SOCKBUF_LOCK(&so->so_rcv);
196 if (sbappendaddr_locked(&so->so_rcv, (struct sockaddr *)&fromsa[0], n,
198 SOCKBUF_UNLOCK(&so->so_rcv);
202 UDPSTAT_INC(udps_fullsock);
204 sorwakeup_locked(so);
209 udp6_input(struct mbuf **mp, int *offp, int proto)
211 struct mbuf *m = *mp;
216 struct inpcbinfo *pcbinfo;
221 struct sockaddr_in6 fromsa[2];
222 struct m_tag *fwd_tag;
228 ifp = m->m_pkthdr.rcvif;
230 if (m->m_len < off + sizeof(struct udphdr)) {
231 m = m_pullup(m, off + sizeof(struct udphdr));
233 IP6STAT_INC(ip6s_exthdrtoolong);
235 return (IPPROTO_DONE);
238 ip6 = mtod(m, struct ip6_hdr *);
239 uh = (struct udphdr *)((caddr_t)ip6 + off);
241 UDPSTAT_INC(udps_ipackets);
244 * Destination port of 0 is illegal, based on RFC768.
246 if (uh->uh_dport == 0)
249 plen = ntohs(ip6->ip6_plen) - off + sizeof(*ip6);
250 ulen = ntohs((u_short)uh->uh_ulen);
253 cscov_partial = (nxt == IPPROTO_UDPLITE) ? 1 : 0;
254 if (nxt == IPPROTO_UDPLITE) {
255 /* Zero means checksum over the complete packet. */
260 if ((ulen < sizeof(struct udphdr)) || (ulen > plen)) {
261 /* XXX: What is the right UDPLite MIB counter? */
264 if (uh->uh_sum == 0) {
265 /* XXX: What is the right UDPLite MIB counter? */
269 if ((ulen < sizeof(struct udphdr)) || (plen != ulen)) {
270 UDPSTAT_INC(udps_badlen);
273 if (uh->uh_sum == 0) {
274 UDPSTAT_INC(udps_nosum);
276 * dport 0 was rejected earlier so this is OK even if
277 * zero_checksum_port is 0 (which is its default value).
279 if (ntohs(uh->uh_dport) == V_zero_checksum_port)
286 if ((m->m_pkthdr.csum_flags & CSUM_DATA_VALID_IPV6) &&
288 if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR)
289 uh_sum = m->m_pkthdr.csum_data;
291 uh_sum = in6_cksum_pseudo(ip6, ulen, nxt,
292 m->m_pkthdr.csum_data);
295 uh_sum = in6_cksum_partial(m, nxt, off, plen, ulen);
298 UDPSTAT_INC(udps_badsum);
304 * Construct sockaddr format source address.
306 init_sin6(&fromsa[0], m, 0);
307 fromsa[0].sin6_port = uh->uh_sport;
308 init_sin6(&fromsa[1], m, 1);
309 fromsa[1].sin6_port = uh->uh_dport;
311 pcbinfo = udp_get_inpcbinfo(nxt);
312 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
314 struct inpcbhead *pcblist;
315 struct ip6_moptions *imo;
318 * In the event that laddr should be set to the link-local
319 * address (this happens in RIPng), the multicast address
320 * specified in the received packet will not match laddr. To
321 * handle this situation, matching is relaxed if the
322 * receiving interface is the same as one specified in the
323 * socket and if the destination multicast address matches
324 * one of the multicast groups specified in the socket.
328 * KAME note: traditionally we dropped udpiphdr from mbuf
329 * here. We need udphdr for IPsec processing so we do that
332 pcblist = udp_get_pcblist(nxt);
334 CK_LIST_FOREACH(inp, pcblist, inp_list) {
335 if ((inp->inp_vflag & INP_IPV6) == 0)
337 if (inp->inp_lport != uh->uh_dport)
339 if (inp->inp_fport != 0 &&
340 inp->inp_fport != uh->uh_sport)
342 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
343 if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr,
347 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
348 if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr,
350 inp->inp_fport != uh->uh_sport)
355 * XXXRW: Because we weren't holding either the inpcb
356 * or the hash lock when we checked for a match
357 * before, we should probably recheck now that the
358 * inpcb lock is (supposed to be) held.
362 * Handle socket delivery policy for any-source
363 * and source-specific multicast. [RFC3678]
365 imo = inp->in6p_moptions;
366 if (imo && IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
367 struct sockaddr_in6 mcaddr;
371 if (__predict_false(inp->inp_flags2 & INP_FREED)) {
376 bzero(&mcaddr, sizeof(struct sockaddr_in6));
377 mcaddr.sin6_len = sizeof(struct sockaddr_in6);
378 mcaddr.sin6_family = AF_INET6;
379 mcaddr.sin6_addr = ip6->ip6_dst;
381 blocked = im6o_mc_filter(imo, ifp,
382 (struct sockaddr *)&mcaddr,
383 (struct sockaddr *)&fromsa[0]);
384 if (blocked != MCAST_PASS) {
385 if (blocked == MCAST_NOTGMEMBER)
386 IP6STAT_INC(ip6s_notmember);
387 if (blocked == MCAST_NOTSMEMBER ||
388 blocked == MCAST_MUTED)
389 UDPSTAT_INC(udps_filtermcast);
390 INP_RUNLOCK(inp); /* XXX */
399 if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) !=
402 if (__predict_true(last->inp_flags2 & INP_FREED) == 0) {
403 if (nxt == IPPROTO_UDPLITE)
404 UDPLITE_PROBE(receive, NULL, last,
407 UDP_PROBE(receive, NULL, last,
409 if (udp6_append(last, n, off, fromsa)) {
410 /* XXX-BZ do we leak m here? */
412 return (IPPROTO_DONE);
420 * Don't look for additional matches if this one does
421 * not have either the SO_REUSEPORT or SO_REUSEADDR
422 * socket options set. This heuristic avoids
423 * searching through all pcbs in the common case of a
424 * non-shared port. It assumes that an application
425 * will never clear these options after setting them.
427 if ((last->inp_socket->so_options &
428 (SO_REUSEPORT|SO_REUSEPORT_LB|SO_REUSEADDR)) == 0)
434 * No matching pcb found; discard datagram. (No need
435 * to send an ICMP Port Unreachable for a broadcast
436 * or multicast datgram.)
438 UDPSTAT_INC(udps_noport);
439 UDPSTAT_INC(udps_noportmcast);
443 if (__predict_true(last->inp_flags2 & INP_FREED) == 0) {
444 if (nxt == IPPROTO_UDPLITE)
445 UDPLITE_PROBE(receive, NULL, last, ip6, last, uh);
447 UDP_PROBE(receive, NULL, last, ip6, last, uh);
448 if (udp6_append(last, m, off, fromsa) == 0)
453 return (IPPROTO_DONE);
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);
519 icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0);
521 return (IPPROTO_DONE);
523 INP_RLOCK_ASSERT(inp);
526 if (up->u_rxcslen == 0 || up->u_rxcslen > ulen) {
530 return (IPPROTO_DONE);
533 if (nxt == IPPROTO_UDPLITE)
534 UDPLITE_PROBE(receive, NULL, inp, ip6, inp, uh);
536 UDP_PROBE(receive, NULL, inp, ip6, inp, uh);
537 if (udp6_append(inp, m, off, fromsa) == 0)
540 return (IPPROTO_DONE);
545 return (IPPROTO_DONE);
549 udp6_common_ctlinput(int cmd, struct sockaddr *sa, void *d,
550 struct inpcbinfo *pcbinfo)
556 struct ip6ctlparam *ip6cp = NULL;
557 const struct sockaddr_in6 *sa6_src = NULL;
559 struct inpcb *(*notify)(struct inpcb *, int) = udp_notify;
560 struct udp_portonly {
565 if (sa->sa_family != AF_INET6 ||
566 sa->sa_len != sizeof(struct sockaddr_in6))
569 if ((unsigned)cmd >= PRC_NCMDS)
571 if (PRC_IS_REDIRECT(cmd))
572 notify = in6_rtchange, d = NULL;
573 else if (cmd == PRC_HOSTDEAD)
575 else if (inet6ctlerrmap[cmd] == 0)
578 /* if the parameter is from icmp6, decode it. */
580 ip6cp = (struct ip6ctlparam *)d;
582 ip6 = ip6cp->ip6c_ip6;
583 off = ip6cp->ip6c_off;
584 cmdarg = ip6cp->ip6c_cmdarg;
585 sa6_src = ip6cp->ip6c_src;
595 * XXX: We assume that when IPV6 is non NULL,
596 * M and OFF are valid.
599 /* Check if we can safely examine src and dst ports. */
600 if (m->m_pkthdr.len < off + sizeof(*uhp))
603 bzero(&uh, sizeof(uh));
604 m_copydata(m, off, sizeof(*uhp), (caddr_t)&uh);
606 if (!PRC_IS_REDIRECT(cmd)) {
607 /* Check to see if its tunneled */
609 inp = in6_pcblookup_mbuf(pcbinfo, &ip6->ip6_dst,
610 uh.uh_dport, &ip6->ip6_src, uh.uh_sport,
611 INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB,
612 m->m_pkthdr.rcvif, m);
617 if (up->u_icmp_func) {
620 (*up->u_icmp_func)(cmd, (struct sockaddr *)ip6cp->ip6c_src,
629 (void)in6_pcbnotify(pcbinfo, sa, uh.uh_dport,
630 (struct sockaddr *)ip6cp->ip6c_src, uh.uh_sport, cmd,
633 (void)in6_pcbnotify(pcbinfo, sa, 0,
634 (const struct sockaddr *)sa6_src, 0, cmd, cmdarg, notify);
638 udp6_ctlinput(int cmd, struct sockaddr *sa, void *d)
641 return (udp6_common_ctlinput(cmd, sa, d, &V_udbinfo));
645 udplite6_ctlinput(int cmd, struct sockaddr *sa, void *d)
648 return (udp6_common_ctlinput(cmd, sa, d, &V_ulitecbinfo));
652 udp6_getcred(SYSCTL_HANDLER_ARGS)
655 struct sockaddr_in6 addrs[2];
656 struct epoch_tracker et;
660 error = priv_check(req->td, PRIV_NETINET_GETCRED);
664 if (req->newlen != sizeof(addrs))
666 if (req->oldlen != sizeof(struct xucred))
668 error = SYSCTL_IN(req, addrs, sizeof(addrs));
671 if ((error = sa6_embedscope(&addrs[0], V_ip6_use_defzone)) != 0 ||
672 (error = sa6_embedscope(&addrs[1], V_ip6_use_defzone)) != 0) {
676 inp = in6_pcblookup(&V_udbinfo, &addrs[1].sin6_addr,
677 addrs[1].sin6_port, &addrs[0].sin6_addr, addrs[0].sin6_port,
678 INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB, NULL);
681 INP_RLOCK_ASSERT(inp);
682 if (inp->inp_socket == NULL)
685 error = cr_canseesocket(req->td->td_ucred,
688 cru2x(inp->inp_cred, &xuc);
693 error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred));
697 SYSCTL_PROC(_net_inet6_udp6, OID_AUTO, getcred,
698 CTLTYPE_OPAQUE | CTLFLAG_RW | CTLFLAG_MPSAFE,
699 0, 0, udp6_getcred, "S,xucred",
700 "Get the xucred of a UDP6 connection");
703 udp6_output(struct socket *so, int flags_arg, struct mbuf *m,
704 struct sockaddr *addr6, struct mbuf *control, struct thread *td)
709 struct in6_addr *laddr, *faddr, in6a;
710 struct ip6_pktopts *optp, opt;
711 struct sockaddr_in6 *sin6, tmp;
712 struct epoch_tracker et;
713 int cscov_partial, error, flags, hlen, scope_ambiguous;
714 u_int32_t ulen, plen;
719 /* addr6 has been validated in udp6_send(). */
720 sin6 = (struct sockaddr_in6 *)addr6;
723 * In contrast to to IPv4 we do not validate the max. packet length
724 * here due to IPv6 Jumbograms (RFC2675).
729 /* Protect *addr6 from overwrites. */
734 * Application should provide a proper zone ID or the use of
735 * default zone IDs should be enabled. Unfortunately, some
736 * applications do not behave as it should, so we need a
737 * workaround. Even if an appropriate ID is not determined,
738 * we'll see if we can determine the outgoing interface. If we
739 * can, determine the zone ID based on the interface below.
741 if (sin6->sin6_scope_id == 0 && !V_ip6_use_defzone)
743 if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0) {
752 KASSERT(inp != NULL, ("%s: inp == NULL", __func__));
754 * In the following cases we want a write lock on the inp for either
755 * local operations or for possible route cache updates in the IPv6
757 * - on connected sockets (sin6 is NULL) for route cache updates,
758 * - when we are not bound to an address and source port (it is
759 * in6_pcbsetport() which will require the write lock).
761 * We check the inp fields before actually locking the inp, so
762 * here exists a race, and we may WLOCK the inp and end with already
763 * bound one by other thread. This is fine.
765 if (sin6 == NULL || (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
766 inp->inp_lport == 0))
771 nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ?
772 IPPROTO_UDP : IPPROTO_UDPLITE;
775 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
779 hasv4addr = (inp->inp_vflag & INP_IPV4);
781 hasv4addr = IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)
784 struct pr_usrreqs *pru;
787 * XXXRW: We release UDP-layer locks before calling
788 * udp_send() in order to avoid recursion. However,
789 * this does mean there is a short window where inp's
790 * fields are unstable. Could this lead to a
791 * potential race in which the factors causing us to
792 * select the UDPv4 output routine are invalidated?
796 in6_sin6_2_sin_in_sock((struct sockaddr *)sin6);
797 pru = inetsw[ip_protox[nxt]].pr_usrreqs;
798 /* addr will just be freed in sendit(). */
799 return ((*pru->pru_send)(so, flags_arg | PRUS_IPV6, m,
800 (struct sockaddr *)sin6, control, td));
804 if (sin6 && IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
806 * Given this is either an IPv6-only socket or no INET is
807 * supported we will fail the send if the given destination
808 * address is a v4mapped address.
810 * XXXGL: do we leak m and control?
817 if ((error = ip6_setpktopts(control, &opt,
818 inp->in6p_outputopts, td->td_ucred, nxt)) != 0) {
820 ip6_clearpktopts(&opt, -1);
828 optp = inp->in6p_outputopts;
833 * Since we saw no essential reason for calling in_pcbconnect,
834 * we get rid of such kind of logic, and call in6_selectsrc
835 * and in6_pcbsetport in order to fill in the local address
836 * and the local port.
838 if (sin6->sin6_port == 0) {
839 error = EADDRNOTAVAIL;
843 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
844 /* how about ::ffff:0.0.0.0 case? */
850 * Given we handle the v4mapped case in the INET block above
851 * assert here that it must not happen anymore.
853 KASSERT(!IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr),
854 ("%s: sin6(%p)->sin6_addr is v4mapped which we "
855 "should have handled.", __func__, sin6));
857 /* This only requires read-locking. */
858 error = in6_selectsrc_socket(sin6, optp, inp,
859 td->td_ucred, scope_ambiguous, &in6a, NULL);
864 if (inp->inp_lport == 0) {
865 struct inpcbinfo *pcbinfo;
867 INP_WLOCK_ASSERT(inp);
869 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
870 INP_HASH_WLOCK(pcbinfo);
871 error = in6_pcbsetport(laddr, inp, td->td_ucred);
872 INP_HASH_WUNLOCK(pcbinfo);
874 /* Undo an address bind that may have occurred. */
875 inp->in6p_laddr = in6addr_any;
879 faddr = &sin6->sin6_addr;
880 fport = sin6->sin6_port; /* allow 0 port */
883 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
887 laddr = &inp->in6p_laddr;
888 faddr = &inp->in6p_faddr;
889 fport = inp->inp_fport;
892 ulen = m->m_pkthdr.len;
893 plen = sizeof(struct udphdr) + ulen;
894 hlen = sizeof(struct ip6_hdr);
897 * Calculate data length and get a mbuf
898 * for UDP and IP6 headers.
900 M_PREPEND(m, hlen + sizeof(struct udphdr), M_NOWAIT);
907 * Stuff checksum and output datagram.
909 cscov = cscov_partial = 0;
910 udp6 = (struct udphdr *)(mtod(m, caddr_t) + hlen);
911 udp6->uh_sport = inp->inp_lport; /* lport is always set in the PCB */
912 udp6->uh_dport = fport;
913 if (nxt == IPPROTO_UDPLITE) {
917 cscov = up->u_txcslen;
920 udp6->uh_ulen = htons(cscov);
922 * For UDP-Lite, checksum coverage length of zero means
923 * the entire UDPLite packet is covered by the checksum.
925 cscov_partial = (cscov == 0) ? 0 : 1;
926 } else if (plen <= 0xffff)
927 udp6->uh_ulen = htons((u_short)plen);
932 ip6 = mtod(m, struct ip6_hdr *);
933 ip6->ip6_flow = inp->inp_flow & IPV6_FLOWINFO_MASK;
934 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
935 ip6->ip6_vfc |= IPV6_VERSION;
936 ip6->ip6_plen = htons((u_short)plen);
938 ip6->ip6_hlim = in6_selecthlim(inp, NULL);
939 ip6->ip6_src = *laddr;
940 ip6->ip6_dst = *faddr;
943 mac_inpcb_create_mbuf(inp, m);
947 if ((udp6->uh_sum = in6_cksum_partial(m, nxt,
948 sizeof(struct ip6_hdr), plen, cscov)) == 0)
949 udp6->uh_sum = 0xffff;
951 udp6->uh_sum = in6_cksum_pseudo(ip6, plen, nxt, 0);
952 m->m_pkthdr.csum_flags = CSUM_UDP_IPV6;
953 m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
959 uint32_t hash_val, hash_type;
962 pr = inp->inp_socket->so_proto->pr_protocol;
964 * Calculate an appropriate RSS hash for UDP and
967 * The called function will take care of figuring out
968 * whether a 2-tuple or 4-tuple hash is required based
969 * on the currently configured scheme.
971 * Later later on connected socket values should be
972 * cached in the inpcb and reused, rather than constantly
975 * UDP Lite is a different protocol number and will
976 * likely end up being hashed as a 2-tuple until
977 * RSS / NICs grow UDP Lite protocol awareness.
979 if (rss_proto_software_hash_v6(faddr, laddr, fport,
980 inp->inp_lport, pr, &hash_val, &hash_type) == 0) {
981 m->m_pkthdr.flowid = hash_val;
982 M_HASHTYPE_SET(m, hash_type);
986 * Don't override with the inp cached flowid.
988 * Until the whole UDP path is vetted, it may actually
991 flags |= IP_NODEFAULTFLOWID;
995 UDPSTAT_INC(udps_opackets);
996 if (nxt == IPPROTO_UDPLITE)
997 UDPLITE_PROBE(send, NULL, inp, ip6, inp, udp6);
999 UDP_PROBE(send, NULL, inp, ip6, inp, udp6);
1000 error = ip6_output(m, optp,
1001 INP_WLOCKED(inp) ? &inp->inp_route6 : NULL, flags,
1002 inp->in6p_moptions, NULL, inp);
1007 ip6_clearpktopts(&opt, -1);
1016 ip6_clearpktopts(&opt, -1);
1025 udp6_abort(struct socket *so)
1028 struct inpcbinfo *pcbinfo;
1030 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1031 inp = sotoinpcb(so);
1032 KASSERT(inp != NULL, ("udp6_abort: inp == NULL"));
1036 if (inp->inp_vflag & INP_IPV4) {
1037 struct pr_usrreqs *pru;
1040 nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ?
1041 IPPROTO_UDP : IPPROTO_UDPLITE;
1043 pru = inetsw[ip_protox[nxt]].pr_usrreqs;
1044 (*pru->pru_abort)(so);
1049 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1050 INP_HASH_WLOCK(pcbinfo);
1051 in6_pcbdisconnect(inp);
1052 inp->in6p_laddr = in6addr_any;
1053 INP_HASH_WUNLOCK(pcbinfo);
1054 soisdisconnected(so);
1060 udp6_attach(struct socket *so, int proto, struct thread *td)
1063 struct inpcbinfo *pcbinfo;
1066 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1067 inp = sotoinpcb(so);
1068 KASSERT(inp == NULL, ("udp6_attach: inp != NULL"));
1070 if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
1071 error = soreserve(so, udp_sendspace, udp_recvspace);
1075 INP_INFO_WLOCK(pcbinfo);
1076 error = in_pcballoc(so, pcbinfo);
1078 INP_INFO_WUNLOCK(pcbinfo);
1081 inp = (struct inpcb *)so->so_pcb;
1082 inp->inp_vflag |= INP_IPV6;
1083 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0)
1084 inp->inp_vflag |= INP_IPV4;
1085 inp->in6p_hops = -1; /* use kernel default */
1086 inp->in6p_cksum = -1; /* just to be sure */
1089 * IPv4 TTL initialization is necessary for an IPv6 socket as well,
1090 * because the socket may be bound to an IPv6 wildcard address,
1091 * which may match an IPv4-mapped IPv6 address.
1093 inp->inp_ip_ttl = V_ip_defttl;
1095 error = udp_newudpcb(inp);
1099 INP_INFO_WUNLOCK(pcbinfo);
1103 INP_INFO_WUNLOCK(pcbinfo);
1108 udp6_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
1111 struct inpcbinfo *pcbinfo;
1115 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1116 inp = sotoinpcb(so);
1117 KASSERT(inp != NULL, ("udp6_bind: inp == NULL"));
1120 INP_HASH_WLOCK(pcbinfo);
1121 vflagsav = inp->inp_vflag;
1122 inp->inp_vflag &= ~INP_IPV4;
1123 inp->inp_vflag |= INP_IPV6;
1124 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
1125 struct sockaddr_in6 *sin6_p;
1127 sin6_p = (struct sockaddr_in6 *)nam;
1129 if (IN6_IS_ADDR_UNSPECIFIED(&sin6_p->sin6_addr))
1130 inp->inp_vflag |= INP_IPV4;
1132 else if (IN6_IS_ADDR_V4MAPPED(&sin6_p->sin6_addr)) {
1133 struct sockaddr_in sin;
1135 in6_sin6_2_sin(&sin, sin6_p);
1136 inp->inp_vflag |= INP_IPV4;
1137 inp->inp_vflag &= ~INP_IPV6;
1138 error = in_pcbbind(inp, (struct sockaddr *)&sin,
1145 error = in6_pcbbind(inp, nam, td->td_ucred);
1150 inp->inp_vflag = vflagsav;
1151 INP_HASH_WUNLOCK(pcbinfo);
1157 udp6_close(struct socket *so)
1160 struct inpcbinfo *pcbinfo;
1162 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1163 inp = sotoinpcb(so);
1164 KASSERT(inp != NULL, ("udp6_close: inp == NULL"));
1168 if (inp->inp_vflag & INP_IPV4) {
1169 struct pr_usrreqs *pru;
1172 nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ?
1173 IPPROTO_UDP : IPPROTO_UDPLITE;
1175 pru = inetsw[ip_protox[nxt]].pr_usrreqs;
1176 (*pru->pru_disconnect)(so);
1180 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1181 INP_HASH_WLOCK(pcbinfo);
1182 in6_pcbdisconnect(inp);
1183 inp->in6p_laddr = in6addr_any;
1184 INP_HASH_WUNLOCK(pcbinfo);
1185 soisdisconnected(so);
1191 udp6_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
1194 struct epoch_tracker et;
1197 struct inpcbinfo *pcbinfo;
1198 struct sockaddr_in6 *sin6;
1202 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1203 inp = sotoinpcb(so);
1204 sin6 = (struct sockaddr_in6 *)nam;
1205 KASSERT(inp != NULL, ("udp6_connect: inp == NULL"));
1208 * XXXRW: Need to clarify locking of v4/v6 flags.
1212 if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
1213 struct sockaddr_in sin;
1215 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) {
1219 if ((inp->inp_vflag & INP_IPV4) == 0) {
1220 error = EAFNOSUPPORT;
1223 if (inp->inp_faddr.s_addr != INADDR_ANY) {
1227 in6_sin6_2_sin(&sin, sin6);
1228 error = prison_remote_ip4(td->td_ucred, &sin.sin_addr);
1231 vflagsav = inp->inp_vflag;
1232 inp->inp_vflag |= INP_IPV4;
1233 inp->inp_vflag &= ~INP_IPV6;
1234 NET_EPOCH_ENTER(et);
1235 INP_HASH_WLOCK(pcbinfo);
1236 error = in_pcbconnect(inp, (struct sockaddr *)&sin,
1238 INP_HASH_WUNLOCK(pcbinfo);
1241 * If connect succeeds, mark socket as connected. If
1242 * connect fails and socket is unbound, reset inp_vflag
1247 else if (inp->inp_laddr.s_addr == INADDR_ANY &&
1248 inp->inp_lport == 0)
1249 inp->inp_vflag = vflagsav;
1252 if ((inp->inp_vflag & INP_IPV6) == 0) {
1253 error = EAFNOSUPPORT;
1258 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1262 error = prison_remote_ip6(td->td_ucred, &sin6->sin6_addr);
1265 vflagsav = inp->inp_vflag;
1266 inp->inp_vflag &= ~INP_IPV4;
1267 inp->inp_vflag |= INP_IPV6;
1268 INP_HASH_WLOCK(pcbinfo);
1269 error = in6_pcbconnect(inp, nam, td->td_ucred);
1270 INP_HASH_WUNLOCK(pcbinfo);
1272 * If connect succeeds, mark socket as connected. If
1273 * connect fails and socket is unbound, reset inp_vflag
1278 else if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
1279 inp->inp_lport == 0)
1280 inp->inp_vflag = vflagsav;
1287 udp6_detach(struct socket *so)
1290 struct inpcbinfo *pcbinfo;
1293 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1294 inp = sotoinpcb(so);
1295 KASSERT(inp != NULL, ("udp6_detach: inp == NULL"));
1297 INP_INFO_WLOCK(pcbinfo);
1299 up = intoudpcb(inp);
1300 KASSERT(up != NULL, ("%s: up == NULL", __func__));
1303 INP_INFO_WUNLOCK(pcbinfo);
1308 udp6_disconnect(struct socket *so)
1311 struct inpcbinfo *pcbinfo;
1313 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1314 inp = sotoinpcb(so);
1315 KASSERT(inp != NULL, ("udp6_disconnect: inp == NULL"));
1319 if (inp->inp_vflag & INP_IPV4) {
1320 struct pr_usrreqs *pru;
1323 nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ?
1324 IPPROTO_UDP : IPPROTO_UDPLITE;
1326 pru = inetsw[ip_protox[nxt]].pr_usrreqs;
1327 (void)(*pru->pru_disconnect)(so);
1332 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1337 INP_HASH_WLOCK(pcbinfo);
1338 in6_pcbdisconnect(inp);
1339 inp->in6p_laddr = in6addr_any;
1340 INP_HASH_WUNLOCK(pcbinfo);
1342 so->so_state &= ~SS_ISCONNECTED; /* XXX */
1349 udp6_send(struct socket *so, int flags, struct mbuf *m,
1350 struct sockaddr *addr, struct mbuf *control, struct thread *td)
1355 if (addr->sa_len != sizeof(struct sockaddr_in6)) {
1359 if (addr->sa_family != AF_INET6) {
1360 error = EAFNOSUPPORT;
1365 return (udp6_output(so, flags, m, addr, control, td));
1374 struct pr_usrreqs udp6_usrreqs = {
1375 .pru_abort = udp6_abort,
1376 .pru_attach = udp6_attach,
1377 .pru_bind = udp6_bind,
1378 .pru_connect = udp6_connect,
1379 .pru_control = in6_control,
1380 .pru_detach = udp6_detach,
1381 .pru_disconnect = udp6_disconnect,
1382 .pru_peeraddr = in6_mapped_peeraddr,
1383 .pru_send = udp6_send,
1384 .pru_shutdown = udp_shutdown,
1385 .pru_sockaddr = in6_mapped_sockaddr,
1386 .pru_soreceive = soreceive_dgram,
1387 .pru_sosend = sosend_dgram,
1388 .pru_sosetlabel = in_pcbsosetlabel,
1389 .pru_close = udp6_close