2 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
3 * Copyright (c) 2010-2011 Juniper Networks, Inc.
4 * Copyright (c) 2014 Kevin Lo
7 * Portions of this software were developed by Robert N. M. Watson under
8 * contract to Juniper Networks, Inc.
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
13 * 1. Redistributions of source code must retain the above copyright
14 * notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 * notice, this list of conditions and the following disclaimer in the
17 * documentation and/or other materials provided with the distribution.
18 * 3. Neither the name of the project nor the names of its contributors
19 * may be used to endorse or promote products derived from this software
20 * without specific prior written permission.
22 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34 * $KAME: udp6_usrreq.c,v 1.27 2001/05/21 05:45:10 jinmei Exp $
35 * $KAME: udp6_output.c,v 1.31 2001/05/21 16:39:15 jinmei Exp $
39 * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995
40 * The Regents of the University of California.
41 * All rights reserved.
43 * Redistribution and use in source and binary forms, with or without
44 * modification, are permitted provided that the following conditions
46 * 1. Redistributions of source code must retain the above copyright
47 * notice, this list of conditions and the following disclaimer.
48 * 2. Redistributions in binary form must reproduce the above copyright
49 * notice, this list of conditions and the following disclaimer in the
50 * documentation and/or other materials provided with the distribution.
51 * 3. Neither the name of the University nor the names of its contributors
52 * may be used to endorse or promote products derived from this software
53 * without specific prior written permission.
55 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
56 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
57 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
58 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
59 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
60 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
61 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
62 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
63 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
64 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
67 * @(#)udp_usrreq.c 8.6 (Berkeley) 5/23/95
70 #include <sys/cdefs.h>
71 __FBSDID("$FreeBSD$");
74 #include "opt_inet6.h"
75 #include "opt_ipsec.h"
78 #include <sys/param.h>
80 #include <sys/kernel.h>
85 #include <sys/protosw.h>
87 #include <sys/signalvar.h>
88 #include <sys/socket.h>
89 #include <sys/socketvar.h>
91 #include <sys/sysctl.h>
92 #include <sys/syslog.h>
93 #include <sys/systm.h>
96 #include <net/if_var.h>
97 #include <net/if_types.h>
98 #include <net/route.h>
99 #include <net/rss_config.h>
101 #include <netinet/in.h>
102 #include <netinet/in_kdtrace.h>
103 #include <netinet/in_pcb.h>
104 #include <netinet/in_systm.h>
105 #include <netinet/in_var.h>
106 #include <netinet/ip.h>
107 #include <netinet/ip6.h>
108 #include <netinet/icmp6.h>
109 #include <netinet/ip_var.h>
110 #include <netinet/udp.h>
111 #include <netinet/udp_var.h>
112 #include <netinet/udplite.h>
114 #include <netinet6/ip6protosw.h>
115 #include <netinet6/ip6_var.h>
116 #include <netinet6/in6_pcb.h>
117 #include <netinet6/in6_rss.h>
118 #include <netinet6/udp6_var.h>
119 #include <netinet6/scope6_var.h>
121 #include <netipsec/ipsec_support.h>
123 #include <security/mac/mac_framework.h>
126 * UDP protocol implementation.
127 * Per RFC 768, August, 1980.
130 extern struct protosw inetsw[];
131 static void udp6_detach(struct socket *so);
134 udp6_append(struct inpcb *inp, struct mbuf *n, int off,
135 struct sockaddr_in6 *fromsa)
138 struct mbuf *opts = NULL, *tmp_opts;
141 INP_LOCK_ASSERT(inp);
144 * Engage the tunneling protocol.
147 if (up->u_tun_func != NULL) {
150 (*up->u_tun_func)(n, off, inp, (struct sockaddr *)&fromsa[0],
153 return (in_pcbrele_rlocked(inp));
155 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
156 /* Check AH/ESP integrity. */
157 if (IPSEC_ENABLED(ipv6)) {
158 if (IPSEC_CHECK_POLICY(ipv6, n, inp) != 0) {
165 if (mac_inpcb_check_deliver(inp, n) != 0) {
171 if (inp->inp_flags & INP_CONTROLOPTS ||
172 inp->inp_socket->so_options & SO_TIMESTAMP)
173 ip6_savecontrol(inp, n, &opts);
174 if ((inp->inp_vflag & INP_IPV6) && (inp->inp_flags2 & INP_ORIGDSTADDR)) {
175 tmp_opts = sbcreatecontrol((caddr_t)&fromsa[1],
176 sizeof(struct sockaddr_in6), IPV6_ORIGDSTADDR, IPPROTO_IPV6);
179 tmp_opts->m_next = opts;
186 m_adj(n, off + sizeof(struct udphdr));
188 so = inp->inp_socket;
189 SOCKBUF_LOCK(&so->so_rcv);
190 if (sbappendaddr_locked(&so->so_rcv, (struct sockaddr *)&fromsa[0], n,
192 SOCKBUF_UNLOCK(&so->so_rcv);
196 UDPSTAT_INC(udps_fullsock);
198 sorwakeup_locked(so);
203 udp6_input(struct mbuf **mp, int *offp, int proto)
205 struct mbuf *m = *mp;
210 struct inpcbinfo *pcbinfo;
215 struct sockaddr_in6 fromsa[2];
216 struct m_tag *fwd_tag;
220 ifp = m->m_pkthdr.rcvif;
221 ip6 = mtod(m, struct ip6_hdr *);
223 #ifndef PULLDOWN_TEST
224 IP6_EXTHDR_CHECK(m, off, sizeof(struct udphdr), IPPROTO_DONE);
225 ip6 = mtod(m, struct ip6_hdr *);
226 uh = (struct udphdr *)((caddr_t)ip6 + off);
228 IP6_EXTHDR_GET(uh, struct udphdr *, m, off, sizeof(*uh));
230 return (IPPROTO_DONE);
233 UDPSTAT_INC(udps_ipackets);
236 * Destination port of 0 is illegal, based on RFC768.
238 if (uh->uh_dport == 0)
241 plen = ntohs(ip6->ip6_plen) - off + sizeof(*ip6);
242 ulen = ntohs((u_short)uh->uh_ulen);
245 cscov_partial = (nxt == IPPROTO_UDPLITE) ? 1 : 0;
246 if (nxt == IPPROTO_UDPLITE) {
247 /* Zero means checksum over the complete packet. */
252 if ((ulen < sizeof(struct udphdr)) || (ulen > plen)) {
253 /* XXX: What is the right UDPLite MIB counter? */
256 if (uh->uh_sum == 0) {
257 /* XXX: What is the right UDPLite MIB counter? */
261 if ((ulen < sizeof(struct udphdr)) || (plen != ulen)) {
262 UDPSTAT_INC(udps_badlen);
265 if (uh->uh_sum == 0) {
266 UDPSTAT_INC(udps_nosum);
271 if ((m->m_pkthdr.csum_flags & CSUM_DATA_VALID_IPV6) &&
273 if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR)
274 uh_sum = m->m_pkthdr.csum_data;
276 uh_sum = in6_cksum_pseudo(ip6, ulen, nxt,
277 m->m_pkthdr.csum_data);
280 uh_sum = in6_cksum_partial(m, nxt, off, plen, ulen);
283 UDPSTAT_INC(udps_badsum);
288 * Construct sockaddr format source address.
290 init_sin6(&fromsa[0], m, 0);
291 fromsa[0].sin6_port = uh->uh_sport;
292 init_sin6(&fromsa[1], m, 1);
293 fromsa[1].sin6_port = uh->uh_dport;
295 pcbinfo = udp_get_inpcbinfo(nxt);
296 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
298 struct inpcbhead *pcblist;
299 struct ip6_moptions *imo;
301 INP_INFO_RLOCK(pcbinfo);
303 * In the event that laddr should be set to the link-local
304 * address (this happens in RIPng), the multicast address
305 * specified in the received packet will not match laddr. To
306 * handle this situation, matching is relaxed if the
307 * receiving interface is the same as one specified in the
308 * socket and if the destination multicast address matches
309 * one of the multicast groups specified in the socket.
313 * KAME note: traditionally we dropped udpiphdr from mbuf
314 * here. We need udphdr for IPsec processing so we do that
317 pcblist = udp_get_pcblist(nxt);
319 LIST_FOREACH(inp, pcblist, inp_list) {
320 if ((inp->inp_vflag & INP_IPV6) == 0)
322 if (inp->inp_lport != uh->uh_dport)
324 if (inp->inp_fport != 0 &&
325 inp->inp_fport != uh->uh_sport)
327 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
328 if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr,
332 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
333 if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr,
335 inp->inp_fport != uh->uh_sport)
340 * XXXRW: Because we weren't holding either the inpcb
341 * or the hash lock when we checked for a match
342 * before, we should probably recheck now that the
343 * inpcb lock is (supposed to be) held.
347 * Handle socket delivery policy for any-source
348 * and source-specific multicast. [RFC3678]
350 imo = inp->in6p_moptions;
351 if (imo && IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
352 struct sockaddr_in6 mcaddr;
357 bzero(&mcaddr, sizeof(struct sockaddr_in6));
358 mcaddr.sin6_len = sizeof(struct sockaddr_in6);
359 mcaddr.sin6_family = AF_INET6;
360 mcaddr.sin6_addr = ip6->ip6_dst;
362 blocked = im6o_mc_filter(imo, ifp,
363 (struct sockaddr *)&mcaddr,
364 (struct sockaddr *)&fromsa[0]);
365 if (blocked != MCAST_PASS) {
366 if (blocked == MCAST_NOTGMEMBER)
367 IP6STAT_INC(ip6s_notmember);
368 if (blocked == MCAST_NOTSMEMBER ||
369 blocked == MCAST_MUTED)
370 UDPSTAT_INC(udps_filtermcast);
371 INP_RUNLOCK(inp); /* XXX */
380 if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) !=
383 UDP_PROBE(receive, NULL, last, ip6,
385 if (udp6_append(last, n, off, fromsa))
392 * Don't look for additional matches if this one does
393 * not have either the SO_REUSEPORT or SO_REUSEADDR
394 * socket options set. This heuristic avoids
395 * searching through all pcbs in the common case of a
396 * non-shared port. It assumes that an application
397 * will never clear these options after setting them.
399 if ((last->inp_socket->so_options &
400 (SO_REUSEPORT|SO_REUSEADDR)) == 0)
406 * No matching pcb found; discard datagram. (No need
407 * to send an ICMP Port Unreachable for a broadcast
408 * or multicast datgram.)
410 UDPSTAT_INC(udps_noport);
411 UDPSTAT_INC(udps_noportmcast);
415 INP_INFO_RUNLOCK(pcbinfo);
416 UDP_PROBE(receive, NULL, last, ip6, last, uh);
417 if (udp6_append(last, m, off, fromsa) == 0)
420 return (IPPROTO_DONE);
423 * Locate pcb for datagram.
427 * Grab info from PACKET_TAG_IPFORWARD tag prepended to the chain.
429 if ((m->m_flags & M_IP6_NEXTHOP) &&
430 (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
431 struct sockaddr_in6 *next_hop6;
433 next_hop6 = (struct sockaddr_in6 *)(fwd_tag + 1);
436 * Transparently forwarded. Pretend to be the destination.
437 * Already got one like this?
439 inp = in6_pcblookup_mbuf(pcbinfo, &ip6->ip6_src,
440 uh->uh_sport, &ip6->ip6_dst, uh->uh_dport,
441 INPLOOKUP_RLOCKPCB, m->m_pkthdr.rcvif, m);
444 * It's new. Try to find the ambushing socket.
445 * Because we've rewritten the destination address,
446 * any hardware-generated hash is ignored.
448 inp = in6_pcblookup(pcbinfo, &ip6->ip6_src,
449 uh->uh_sport, &next_hop6->sin6_addr,
450 next_hop6->sin6_port ? htons(next_hop6->sin6_port) :
451 uh->uh_dport, INPLOOKUP_WILDCARD |
452 INPLOOKUP_RLOCKPCB, m->m_pkthdr.rcvif);
454 /* Remove the tag from the packet. We don't need it anymore. */
455 m_tag_delete(m, fwd_tag);
456 m->m_flags &= ~M_IP6_NEXTHOP;
458 inp = in6_pcblookup_mbuf(pcbinfo, &ip6->ip6_src,
459 uh->uh_sport, &ip6->ip6_dst, uh->uh_dport,
460 INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB,
461 m->m_pkthdr.rcvif, m);
463 if (udp_log_in_vain) {
464 char ip6bufs[INET6_ADDRSTRLEN];
465 char ip6bufd[INET6_ADDRSTRLEN];
468 "Connection attempt to UDP [%s]:%d from [%s]:%d\n",
469 ip6_sprintf(ip6bufd, &ip6->ip6_dst),
471 ip6_sprintf(ip6bufs, &ip6->ip6_src),
472 ntohs(uh->uh_sport));
474 UDPSTAT_INC(udps_noport);
475 if (m->m_flags & M_MCAST) {
476 printf("UDP6: M_MCAST is set in a unicast packet.\n");
477 UDPSTAT_INC(udps_noportmcast);
482 icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0);
483 return (IPPROTO_DONE);
485 INP_RLOCK_ASSERT(inp);
488 if (up->u_rxcslen == 0 || up->u_rxcslen > ulen) {
491 return (IPPROTO_DONE);
494 UDP_PROBE(receive, NULL, inp, ip6, inp, uh);
495 if (udp6_append(inp, m, off, fromsa) == 0)
497 return (IPPROTO_DONE);
500 INP_INFO_RUNLOCK(pcbinfo);
504 return (IPPROTO_DONE);
508 udp6_common_ctlinput(int cmd, struct sockaddr *sa, void *d,
509 struct inpcbinfo *pcbinfo)
515 struct ip6ctlparam *ip6cp = NULL;
516 const struct sockaddr_in6 *sa6_src = NULL;
518 struct inpcb *(*notify)(struct inpcb *, int) = udp_notify;
519 struct udp_portonly {
524 if (sa->sa_family != AF_INET6 ||
525 sa->sa_len != sizeof(struct sockaddr_in6))
528 if ((unsigned)cmd >= PRC_NCMDS)
530 if (PRC_IS_REDIRECT(cmd))
531 notify = in6_rtchange, d = NULL;
532 else if (cmd == PRC_HOSTDEAD)
534 else if (inet6ctlerrmap[cmd] == 0)
537 /* if the parameter is from icmp6, decode it. */
539 ip6cp = (struct ip6ctlparam *)d;
541 ip6 = ip6cp->ip6c_ip6;
542 off = ip6cp->ip6c_off;
543 cmdarg = ip6cp->ip6c_cmdarg;
544 sa6_src = ip6cp->ip6c_src;
554 * XXX: We assume that when IPV6 is non NULL,
555 * M and OFF are valid.
558 /* Check if we can safely examine src and dst ports. */
559 if (m->m_pkthdr.len < off + sizeof(*uhp))
562 bzero(&uh, sizeof(uh));
563 m_copydata(m, off, sizeof(*uhp), (caddr_t)&uh);
565 if (!PRC_IS_REDIRECT(cmd)) {
566 /* Check to see if its tunneled */
568 inp = in6_pcblookup_mbuf(pcbinfo, &ip6->ip6_dst,
569 uh.uh_dport, &ip6->ip6_src, uh.uh_sport,
570 INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB,
571 m->m_pkthdr.rcvif, m);
576 if (up->u_icmp_func) {
579 (*up->u_icmp_func)(cmd, (struct sockaddr *)ip6cp->ip6c_src,
588 (void)in6_pcbnotify(pcbinfo, sa, uh.uh_dport,
589 (struct sockaddr *)ip6cp->ip6c_src, uh.uh_sport, cmd,
592 (void)in6_pcbnotify(pcbinfo, sa, 0,
593 (const struct sockaddr *)sa6_src, 0, cmd, cmdarg, notify);
597 udp6_ctlinput(int cmd, struct sockaddr *sa, void *d)
600 return (udp6_common_ctlinput(cmd, sa, d, &V_udbinfo));
604 udplite6_ctlinput(int cmd, struct sockaddr *sa, void *d)
607 return (udp6_common_ctlinput(cmd, sa, d, &V_ulitecbinfo));
611 udp6_getcred(SYSCTL_HANDLER_ARGS)
614 struct sockaddr_in6 addrs[2];
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) {
633 inp = in6_pcblookup(&V_udbinfo, &addrs[1].sin6_addr,
634 addrs[1].sin6_port, &addrs[0].sin6_addr, addrs[0].sin6_port,
635 INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB, NULL);
637 INP_RLOCK_ASSERT(inp);
638 if (inp->inp_socket == NULL)
641 error = cr_canseesocket(req->td->td_ucred,
644 cru2x(inp->inp_cred, &xuc);
649 error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred));
653 SYSCTL_PROC(_net_inet6_udp6, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW, 0,
654 0, udp6_getcred, "S,xucred", "Get the xucred of a UDP6 connection");
657 udp6_output(struct inpcb *inp, struct mbuf *m, struct sockaddr *addr6,
658 struct mbuf *control, struct thread *td)
660 u_int32_t ulen = m->m_pkthdr.len;
661 u_int32_t plen = sizeof(struct udphdr) + ulen;
664 struct in6_addr *laddr, *faddr, in6a;
665 struct sockaddr_in6 *sin6 = NULL;
666 int cscov_partial = 0;
667 int scope_ambiguous = 0;
672 struct ip6_pktopts *optp, opt;
673 int af = AF_INET6, hlen = sizeof(struct ip6_hdr);
675 struct sockaddr_in6 tmp;
677 INP_WLOCK_ASSERT(inp);
678 INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo);
681 /* addr6 has been validated in udp6_send(). */
682 sin6 = (struct sockaddr_in6 *)addr6;
684 /* protect *sin6 from overwrites */
689 * Application should provide a proper zone ID or the use of
690 * default zone IDs should be enabled. Unfortunately, some
691 * applications do not behave as it should, so we need a
692 * workaround. Even if an appropriate ID is not determined,
693 * we'll see if we can determine the outgoing interface. If we
694 * can, determine the zone ID based on the interface below.
696 if (sin6->sin6_scope_id == 0 && !V_ip6_use_defzone)
698 if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0)
702 nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ?
703 IPPROTO_UDP : IPPROTO_UDPLITE;
705 if ((error = ip6_setpktopts(control, &opt,
706 inp->in6p_outputopts, td->td_ucred, nxt)) != 0)
710 optp = inp->in6p_outputopts;
713 faddr = &sin6->sin6_addr;
716 * Since we saw no essential reason for calling in_pcbconnect,
717 * we get rid of such kind of logic, and call in6_selectsrc
718 * and in6_pcbsetport in order to fill in the local address
719 * and the local port.
721 if (sin6->sin6_port == 0) {
722 error = EADDRNOTAVAIL;
726 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
727 /* how about ::ffff:0.0.0.0 case? */
732 fport = sin6->sin6_port; /* allow 0 port */
734 if (IN6_IS_ADDR_V4MAPPED(faddr)) {
735 if ((inp->inp_flags & IN6P_IPV6_V6ONLY)) {
737 * I believe we should explicitly discard the
738 * packet when mapped addresses are disabled,
739 * rather than send the packet as an IPv6 one.
740 * If we chose the latter approach, the packet
741 * might be sent out on the wire based on the
742 * default route, the situation which we'd
743 * probably want to avoid.
744 * (20010421 jinmei@kame.net)
749 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
750 !IN6_IS_ADDR_V4MAPPED(&inp->in6p_laddr)) {
752 * when remote addr is an IPv4-mapped address,
753 * local addr should not be an IPv6 address,
754 * since you cannot determine how to map IPv6
755 * source address to IPv4.
764 if (!IN6_IS_ADDR_V4MAPPED(faddr)) {
765 error = in6_selectsrc_socket(sin6, optp, inp,
766 td->td_ucred, scope_ambiguous, &in6a, NULL);
771 laddr = &inp->in6p_laddr; /* XXX */
774 error = EADDRNOTAVAIL;
777 if (inp->inp_lport == 0 &&
778 (error = in6_pcbsetport(laddr, inp, td->td_ucred)) != 0) {
779 /* Undo an address bind that may have occurred. */
780 inp->in6p_laddr = in6addr_any;
784 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
788 if (IN6_IS_ADDR_V4MAPPED(&inp->in6p_faddr)) {
789 if ((inp->inp_flags & IN6P_IPV6_V6ONLY)) {
791 * XXX: this case would happen when the
792 * application sets the V6ONLY flag after
793 * connecting the foreign address.
794 * Such applications should be fixed,
797 log(LOG_INFO, "udp6_output: IPV6_V6ONLY "
798 "option was set for a connected socket\n");
804 laddr = &inp->in6p_laddr;
805 faddr = &inp->in6p_faddr;
806 fport = inp->inp_fport;
810 hlen = sizeof(struct ip);
813 * Calculate data length and get a mbuf
814 * for UDP and IP6 headers.
816 M_PREPEND(m, hlen + sizeof(struct udphdr), M_NOWAIT);
823 * Stuff checksum and output datagram.
825 udp6 = (struct udphdr *)(mtod(m, caddr_t) + hlen);
826 udp6->uh_sport = inp->inp_lport; /* lport is always set in the PCB */
827 udp6->uh_dport = fport;
828 if (nxt == IPPROTO_UDPLITE) {
832 cscov = up->u_txcslen;
835 udp6->uh_ulen = htons(cscov);
837 * For UDP-Lite, checksum coverage length of zero means
838 * the entire UDPLite packet is covered by the checksum.
840 cscov_partial = (cscov == 0) ? 0 : 1;
841 } else if (plen <= 0xffff)
842 udp6->uh_ulen = htons((u_short)plen);
849 ip6 = mtod(m, struct ip6_hdr *);
850 ip6->ip6_flow = inp->inp_flow & IPV6_FLOWINFO_MASK;
851 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
852 ip6->ip6_vfc |= IPV6_VERSION;
853 ip6->ip6_plen = htons((u_short)plen);
855 ip6->ip6_hlim = in6_selecthlim(inp, NULL);
856 ip6->ip6_src = *laddr;
857 ip6->ip6_dst = *faddr;
860 if ((udp6->uh_sum = in6_cksum_partial(m, nxt,
861 sizeof(struct ip6_hdr), plen, cscov)) == 0)
862 udp6->uh_sum = 0xffff;
864 udp6->uh_sum = in6_cksum_pseudo(ip6, plen, nxt, 0);
865 m->m_pkthdr.csum_flags = CSUM_UDP_IPV6;
866 m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
871 uint32_t hash_val, hash_type;
874 pr = inp->inp_socket->so_proto->pr_protocol;
876 * Calculate an appropriate RSS hash for UDP and
879 * The called function will take care of figuring out
880 * whether a 2-tuple or 4-tuple hash is required based
881 * on the currently configured scheme.
883 * Later later on connected socket values should be
884 * cached in the inpcb and reused, rather than constantly
887 * UDP Lite is a different protocol number and will
888 * likely end up being hashed as a 2-tuple until
889 * RSS / NICs grow UDP Lite protocol awareness.
891 if (rss_proto_software_hash_v6(faddr, laddr, fport,
892 inp->inp_lport, pr, &hash_val, &hash_type) == 0) {
893 m->m_pkthdr.flowid = hash_val;
894 M_HASHTYPE_SET(m, hash_type);
901 * Don't override with the inp cached flowid.
903 * Until the whole UDP path is vetted, it may actually
906 flags |= IP_NODEFAULTFLOWID;
909 UDP_PROBE(send, NULL, inp, ip6, inp, udp6);
910 UDPSTAT_INC(udps_opackets);
911 error = ip6_output(m, optp, &inp->inp_route6, flags,
912 inp->in6p_moptions, NULL, inp);
915 error = EAFNOSUPPORT;
925 ip6_clearpktopts(&opt, -1);
932 udp6_abort(struct socket *so)
935 struct inpcbinfo *pcbinfo;
937 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
939 KASSERT(inp != NULL, ("udp6_abort: inp == NULL"));
943 if (inp->inp_vflag & INP_IPV4) {
944 struct pr_usrreqs *pru;
947 nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ?
948 IPPROTO_UDP : IPPROTO_UDPLITE;
950 pru = inetsw[ip_protox[nxt]].pr_usrreqs;
951 (*pru->pru_abort)(so);
956 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
957 INP_HASH_WLOCK(pcbinfo);
958 in6_pcbdisconnect(inp);
959 inp->in6p_laddr = in6addr_any;
960 INP_HASH_WUNLOCK(pcbinfo);
961 soisdisconnected(so);
967 udp6_attach(struct socket *so, int proto, struct thread *td)
970 struct inpcbinfo *pcbinfo;
973 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
975 KASSERT(inp == NULL, ("udp6_attach: inp != NULL"));
977 if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
978 error = soreserve(so, udp_sendspace, udp_recvspace);
982 INP_INFO_WLOCK(pcbinfo);
983 error = in_pcballoc(so, pcbinfo);
985 INP_INFO_WUNLOCK(pcbinfo);
988 inp = (struct inpcb *)so->so_pcb;
989 inp->inp_vflag |= INP_IPV6;
990 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0)
991 inp->inp_vflag |= INP_IPV4;
992 inp->in6p_hops = -1; /* use kernel default */
993 inp->in6p_cksum = -1; /* just to be sure */
996 * IPv4 TTL initialization is necessary for an IPv6 socket as well,
997 * because the socket may be bound to an IPv6 wildcard address,
998 * which may match an IPv4-mapped IPv6 address.
1000 inp->inp_ip_ttl = V_ip_defttl;
1002 error = udp_newudpcb(inp);
1006 INP_INFO_WUNLOCK(pcbinfo);
1010 INP_INFO_WUNLOCK(pcbinfo);
1015 udp6_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
1018 struct inpcbinfo *pcbinfo;
1021 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1022 inp = sotoinpcb(so);
1023 KASSERT(inp != NULL, ("udp6_bind: inp == NULL"));
1026 INP_HASH_WLOCK(pcbinfo);
1027 inp->inp_vflag &= ~INP_IPV4;
1028 inp->inp_vflag |= INP_IPV6;
1029 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
1030 struct sockaddr_in6 *sin6_p;
1032 sin6_p = (struct sockaddr_in6 *)nam;
1034 if (IN6_IS_ADDR_UNSPECIFIED(&sin6_p->sin6_addr))
1035 inp->inp_vflag |= INP_IPV4;
1037 else if (IN6_IS_ADDR_V4MAPPED(&sin6_p->sin6_addr)) {
1038 struct sockaddr_in sin;
1040 in6_sin6_2_sin(&sin, sin6_p);
1041 inp->inp_vflag |= INP_IPV4;
1042 inp->inp_vflag &= ~INP_IPV6;
1043 error = in_pcbbind(inp, (struct sockaddr *)&sin,
1050 error = in6_pcbbind(inp, nam, td->td_ucred);
1054 INP_HASH_WUNLOCK(pcbinfo);
1060 udp6_close(struct socket *so)
1063 struct inpcbinfo *pcbinfo;
1065 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1066 inp = sotoinpcb(so);
1067 KASSERT(inp != NULL, ("udp6_close: inp == NULL"));
1071 if (inp->inp_vflag & INP_IPV4) {
1072 struct pr_usrreqs *pru;
1075 nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ?
1076 IPPROTO_UDP : IPPROTO_UDPLITE;
1078 pru = inetsw[ip_protox[nxt]].pr_usrreqs;
1079 (*pru->pru_disconnect)(so);
1083 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1084 INP_HASH_WLOCK(pcbinfo);
1085 in6_pcbdisconnect(inp);
1086 inp->in6p_laddr = in6addr_any;
1087 INP_HASH_WUNLOCK(pcbinfo);
1088 soisdisconnected(so);
1094 udp6_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
1097 struct inpcbinfo *pcbinfo;
1098 struct sockaddr_in6 *sin6;
1101 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1102 inp = sotoinpcb(so);
1103 sin6 = (struct sockaddr_in6 *)nam;
1104 KASSERT(inp != NULL, ("udp6_connect: inp == NULL"));
1107 * XXXRW: Need to clarify locking of v4/v6 flags.
1111 if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
1112 struct sockaddr_in sin;
1114 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) {
1118 if ((inp->inp_vflag & INP_IPV4) == 0) {
1119 error = EAFNOSUPPORT;
1122 if (inp->inp_faddr.s_addr != INADDR_ANY) {
1126 in6_sin6_2_sin(&sin, sin6);
1127 inp->inp_vflag |= INP_IPV4;
1128 inp->inp_vflag &= ~INP_IPV6;
1129 error = prison_remote_ip4(td->td_ucred, &sin.sin_addr);
1132 INP_HASH_WLOCK(pcbinfo);
1133 error = in_pcbconnect(inp, (struct sockaddr *)&sin,
1135 INP_HASH_WUNLOCK(pcbinfo);
1140 if ((inp->inp_vflag & INP_IPV6) == 0) {
1141 error = EAFNOSUPPORT;
1146 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1150 inp->inp_vflag &= ~INP_IPV4;
1151 inp->inp_vflag |= INP_IPV6;
1152 error = prison_remote_ip6(td->td_ucred, &sin6->sin6_addr);
1155 INP_HASH_WLOCK(pcbinfo);
1156 error = in6_pcbconnect(inp, nam, td->td_ucred);
1157 INP_HASH_WUNLOCK(pcbinfo);
1166 udp6_detach(struct socket *so)
1169 struct inpcbinfo *pcbinfo;
1172 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1173 inp = sotoinpcb(so);
1174 KASSERT(inp != NULL, ("udp6_detach: inp == NULL"));
1176 INP_INFO_WLOCK(pcbinfo);
1178 up = intoudpcb(inp);
1179 KASSERT(up != NULL, ("%s: up == NULL", __func__));
1182 INP_INFO_WUNLOCK(pcbinfo);
1187 udp6_disconnect(struct socket *so)
1190 struct inpcbinfo *pcbinfo;
1193 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1194 inp = sotoinpcb(so);
1195 KASSERT(inp != NULL, ("udp6_disconnect: inp == NULL"));
1199 if (inp->inp_vflag & INP_IPV4) {
1200 struct pr_usrreqs *pru;
1203 nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ?
1204 IPPROTO_UDP : IPPROTO_UDPLITE;
1206 pru = inetsw[ip_protox[nxt]].pr_usrreqs;
1207 (void)(*pru->pru_disconnect)(so);
1212 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1217 INP_HASH_WLOCK(pcbinfo);
1218 in6_pcbdisconnect(inp);
1219 inp->in6p_laddr = in6addr_any;
1220 INP_HASH_WUNLOCK(pcbinfo);
1222 so->so_state &= ~SS_ISCONNECTED; /* XXX */
1230 udp6_send(struct socket *so, int flags, struct mbuf *m,
1231 struct sockaddr *addr, struct mbuf *control, struct thread *td)
1234 struct inpcbinfo *pcbinfo;
1237 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1238 inp = sotoinpcb(so);
1239 KASSERT(inp != NULL, ("udp6_send: inp == NULL"));
1243 if (addr->sa_len != sizeof(struct sockaddr_in6)) {
1247 if (addr->sa_family != AF_INET6) {
1248 error = EAFNOSUPPORT;
1254 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
1256 struct sockaddr_in6 *sin6 = NULL;
1259 hasv4addr = (inp->inp_vflag & INP_IPV4);
1261 sin6 = (struct sockaddr_in6 *)addr;
1262 hasv4addr = IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)
1266 struct pr_usrreqs *pru;
1269 nxt = (inp->inp_socket->so_proto->pr_protocol ==
1270 IPPROTO_UDP) ? IPPROTO_UDP : IPPROTO_UDPLITE;
1272 * XXXRW: We release UDP-layer locks before calling
1273 * udp_send() in order to avoid recursion. However,
1274 * this does mean there is a short window where inp's
1275 * fields are unstable. Could this lead to a
1276 * potential race in which the factors causing us to
1277 * select the UDPv4 output routine are invalidated?
1281 in6_sin6_2_sin_in_sock(addr);
1282 pru = inetsw[ip_protox[nxt]].pr_usrreqs;
1283 /* addr will just be freed in sendit(). */
1284 return ((*pru->pru_send)(so, flags, m, addr, control,
1290 mac_inpcb_create_mbuf(inp, m);
1292 INP_HASH_WLOCK(pcbinfo);
1293 error = udp6_output(inp, m, addr, control, td);
1294 INP_HASH_WUNLOCK(pcbinfo);
1304 struct pr_usrreqs udp6_usrreqs = {
1305 .pru_abort = udp6_abort,
1306 .pru_attach = udp6_attach,
1307 .pru_bind = udp6_bind,
1308 .pru_connect = udp6_connect,
1309 .pru_control = in6_control,
1310 .pru_detach = udp6_detach,
1311 .pru_disconnect = udp6_disconnect,
1312 .pru_peeraddr = in6_mapped_peeraddr,
1313 .pru_send = udp6_send,
1314 .pru_shutdown = udp_shutdown,
1315 .pru_sockaddr = in6_mapped_sockaddr,
1316 .pru_soreceive = soreceive_dgram,
1317 .pru_sosend = sosend_dgram,
1318 .pru_sosetlabel = in_pcbsosetlabel,
1319 .pru_close = udp6_close