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>
128 * UDP protocol implementation.
129 * Per RFC 768, August, 1980.
132 extern struct protosw inetsw[];
133 static void udp6_detach(struct socket *so);
136 udp6_append(struct inpcb *inp, struct mbuf *n, int off,
137 struct sockaddr_in6 *fromsa)
140 struct mbuf *opts = NULL, *tmp_opts;
143 INP_LOCK_ASSERT(inp);
146 * Engage the tunneling protocol.
149 if (up->u_tun_func != NULL) {
152 (*up->u_tun_func)(n, off, inp, (struct sockaddr *)&fromsa[0],
155 return (in_pcbrele_rlocked(inp));
157 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
158 /* Check AH/ESP integrity. */
159 if (IPSEC_ENABLED(ipv6)) {
160 if (IPSEC_CHECK_POLICY(ipv6, n, inp) != 0) {
167 if (mac_inpcb_check_deliver(inp, n) != 0) {
173 if (inp->inp_flags & INP_CONTROLOPTS ||
174 inp->inp_socket->so_options & SO_TIMESTAMP)
175 ip6_savecontrol(inp, n, &opts);
176 if ((inp->inp_vflag & INP_IPV6) && (inp->inp_flags2 & INP_ORIGDSTADDR)) {
177 tmp_opts = sbcreatecontrol((caddr_t)&fromsa[1],
178 sizeof(struct sockaddr_in6), IPV6_ORIGDSTADDR, IPPROTO_IPV6);
181 tmp_opts->m_next = opts;
188 m_adj(n, off + sizeof(struct udphdr));
190 so = inp->inp_socket;
191 SOCKBUF_LOCK(&so->so_rcv);
192 if (sbappendaddr_locked(&so->so_rcv, (struct sockaddr *)&fromsa[0], n,
194 SOCKBUF_UNLOCK(&so->so_rcv);
198 UDPSTAT_INC(udps_fullsock);
200 sorwakeup_locked(so);
205 udp6_input(struct mbuf **mp, int *offp, int proto)
207 struct mbuf *m = *mp;
212 struct inpcbinfo *pcbinfo;
217 struct sockaddr_in6 fromsa[2];
218 struct m_tag *fwd_tag;
224 ifp = m->m_pkthdr.rcvif;
226 if (m->m_len < off + sizeof(struct udphdr)) {
227 m = m_pullup(m, off + sizeof(struct udphdr));
229 IP6STAT_INC(ip6s_exthdrtoolong);
231 return (IPPROTO_DONE);
234 ip6 = mtod(m, struct ip6_hdr *);
235 uh = (struct udphdr *)((caddr_t)ip6 + off);
237 UDPSTAT_INC(udps_ipackets);
240 * Destination port of 0 is illegal, based on RFC768.
242 if (uh->uh_dport == 0)
245 plen = ntohs(ip6->ip6_plen) - off + sizeof(*ip6);
246 ulen = ntohs((u_short)uh->uh_ulen);
249 cscov_partial = (nxt == IPPROTO_UDPLITE) ? 1 : 0;
250 if (nxt == IPPROTO_UDPLITE) {
251 /* Zero means checksum over the complete packet. */
256 if ((ulen < sizeof(struct udphdr)) || (ulen > plen)) {
257 /* XXX: What is the right UDPLite MIB counter? */
260 if (uh->uh_sum == 0) {
261 /* XXX: What is the right UDPLite MIB counter? */
265 if ((ulen < sizeof(struct udphdr)) || (plen != ulen)) {
266 UDPSTAT_INC(udps_badlen);
269 if (uh->uh_sum == 0) {
270 UDPSTAT_INC(udps_nosum);
275 if ((m->m_pkthdr.csum_flags & CSUM_DATA_VALID_IPV6) &&
277 if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR)
278 uh_sum = m->m_pkthdr.csum_data;
280 uh_sum = in6_cksum_pseudo(ip6, ulen, nxt,
281 m->m_pkthdr.csum_data);
284 uh_sum = in6_cksum_partial(m, nxt, off, plen, ulen);
287 UDPSTAT_INC(udps_badsum);
292 * Construct sockaddr format source address.
294 init_sin6(&fromsa[0], m, 0);
295 fromsa[0].sin6_port = uh->uh_sport;
296 init_sin6(&fromsa[1], m, 1);
297 fromsa[1].sin6_port = uh->uh_dport;
299 pcbinfo = udp_get_inpcbinfo(nxt);
300 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
302 struct inpcbhead *pcblist;
303 struct ip6_moptions *imo;
306 * In the event that laddr should be set to the link-local
307 * address (this happens in RIPng), the multicast address
308 * specified in the received packet will not match laddr. To
309 * handle this situation, matching is relaxed if the
310 * receiving interface is the same as one specified in the
311 * socket and if the destination multicast address matches
312 * one of the multicast groups specified in the socket.
316 * KAME note: traditionally we dropped udpiphdr from mbuf
317 * here. We need udphdr for IPsec processing so we do that
320 pcblist = udp_get_pcblist(nxt);
322 CK_LIST_FOREACH(inp, pcblist, inp_list) {
323 if ((inp->inp_vflag & INP_IPV6) == 0)
325 if (inp->inp_lport != uh->uh_dport)
327 if (inp->inp_fport != 0 &&
328 inp->inp_fport != uh->uh_sport)
330 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
331 if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr,
335 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
336 if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr,
338 inp->inp_fport != uh->uh_sport)
343 * XXXRW: Because we weren't holding either the inpcb
344 * or the hash lock when we checked for a match
345 * before, we should probably recheck now that the
346 * inpcb lock is (supposed to be) held.
350 * Handle socket delivery policy for any-source
351 * and source-specific multicast. [RFC3678]
353 imo = inp->in6p_moptions;
354 if (imo && IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
355 struct sockaddr_in6 mcaddr;
359 if (__predict_false(inp->inp_flags2 & INP_FREED)) {
364 bzero(&mcaddr, sizeof(struct sockaddr_in6));
365 mcaddr.sin6_len = sizeof(struct sockaddr_in6);
366 mcaddr.sin6_family = AF_INET6;
367 mcaddr.sin6_addr = ip6->ip6_dst;
369 blocked = im6o_mc_filter(imo, ifp,
370 (struct sockaddr *)&mcaddr,
371 (struct sockaddr *)&fromsa[0]);
372 if (blocked != MCAST_PASS) {
373 if (blocked == MCAST_NOTGMEMBER)
374 IP6STAT_INC(ip6s_notmember);
375 if (blocked == MCAST_NOTSMEMBER ||
376 blocked == MCAST_MUTED)
377 UDPSTAT_INC(udps_filtermcast);
378 INP_RUNLOCK(inp); /* XXX */
387 if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) !=
390 if (__predict_true(last->inp_flags2 & INP_FREED) == 0) {
391 if (nxt == IPPROTO_UDPLITE)
392 UDPLITE_PROBE(receive, NULL, last,
395 UDP_PROBE(receive, NULL, last,
397 if (udp6_append(last, n, off, fromsa)) {
398 /* XXX-BZ do we leak m here? */
400 return (IPPROTO_DONE);
408 * Don't look for additional matches if this one does
409 * not have either the SO_REUSEPORT or SO_REUSEADDR
410 * socket options set. This heuristic avoids
411 * searching through all pcbs in the common case of a
412 * non-shared port. It assumes that an application
413 * will never clear these options after setting them.
415 if ((last->inp_socket->so_options &
416 (SO_REUSEPORT|SO_REUSEPORT_LB|SO_REUSEADDR)) == 0)
422 * No matching pcb found; discard datagram. (No need
423 * to send an ICMP Port Unreachable for a broadcast
424 * or multicast datgram.)
426 UDPSTAT_INC(udps_noport);
427 UDPSTAT_INC(udps_noportmcast);
431 if (__predict_true(last->inp_flags2 & INP_FREED) == 0) {
432 if (nxt == IPPROTO_UDPLITE)
433 UDPLITE_PROBE(receive, NULL, last, ip6, last, uh);
435 UDP_PROBE(receive, NULL, last, ip6, last, uh);
436 if (udp6_append(last, m, off, fromsa) == 0)
441 return (IPPROTO_DONE);
444 * Locate pcb for datagram.
448 * Grab info from PACKET_TAG_IPFORWARD tag prepended to the chain.
450 if ((m->m_flags & M_IP6_NEXTHOP) &&
451 (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
452 struct sockaddr_in6 *next_hop6;
454 next_hop6 = (struct sockaddr_in6 *)(fwd_tag + 1);
457 * Transparently forwarded. Pretend to be the destination.
458 * Already got one like this?
460 inp = in6_pcblookup_mbuf(pcbinfo, &ip6->ip6_src,
461 uh->uh_sport, &ip6->ip6_dst, uh->uh_dport,
462 INPLOOKUP_RLOCKPCB, m->m_pkthdr.rcvif, m);
465 * It's new. Try to find the ambushing socket.
466 * Because we've rewritten the destination address,
467 * any hardware-generated hash is ignored.
469 inp = in6_pcblookup(pcbinfo, &ip6->ip6_src,
470 uh->uh_sport, &next_hop6->sin6_addr,
471 next_hop6->sin6_port ? htons(next_hop6->sin6_port) :
472 uh->uh_dport, INPLOOKUP_WILDCARD |
473 INPLOOKUP_RLOCKPCB, m->m_pkthdr.rcvif);
475 /* Remove the tag from the packet. We don't need it anymore. */
476 m_tag_delete(m, fwd_tag);
477 m->m_flags &= ~M_IP6_NEXTHOP;
479 inp = in6_pcblookup_mbuf(pcbinfo, &ip6->ip6_src,
480 uh->uh_sport, &ip6->ip6_dst, uh->uh_dport,
481 INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB,
482 m->m_pkthdr.rcvif, m);
484 if (udp_log_in_vain) {
485 char ip6bufs[INET6_ADDRSTRLEN];
486 char ip6bufd[INET6_ADDRSTRLEN];
489 "Connection attempt to UDP [%s]:%d from [%s]:%d\n",
490 ip6_sprintf(ip6bufd, &ip6->ip6_dst),
492 ip6_sprintf(ip6bufs, &ip6->ip6_src),
493 ntohs(uh->uh_sport));
495 if (nxt == IPPROTO_UDPLITE)
496 UDPLITE_PROBE(receive, NULL, NULL, ip6, NULL, uh);
498 UDP_PROBE(receive, NULL, NULL, ip6, NULL, uh);
499 UDPSTAT_INC(udps_noport);
500 if (m->m_flags & M_MCAST) {
501 printf("UDP6: M_MCAST is set in a unicast packet.\n");
502 UDPSTAT_INC(udps_noportmcast);
507 icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0);
509 return (IPPROTO_DONE);
511 INP_RLOCK_ASSERT(inp);
514 if (up->u_rxcslen == 0 || up->u_rxcslen > ulen) {
518 return (IPPROTO_DONE);
521 if (nxt == IPPROTO_UDPLITE)
522 UDPLITE_PROBE(receive, NULL, inp, ip6, inp, uh);
524 UDP_PROBE(receive, NULL, inp, ip6, inp, uh);
525 if (udp6_append(inp, m, off, fromsa) == 0)
528 return (IPPROTO_DONE);
533 return (IPPROTO_DONE);
537 udp6_common_ctlinput(int cmd, struct sockaddr *sa, void *d,
538 struct inpcbinfo *pcbinfo)
544 struct ip6ctlparam *ip6cp = NULL;
545 const struct sockaddr_in6 *sa6_src = NULL;
547 struct inpcb *(*notify)(struct inpcb *, int) = udp_notify;
548 struct udp_portonly {
553 if (sa->sa_family != AF_INET6 ||
554 sa->sa_len != sizeof(struct sockaddr_in6))
557 if ((unsigned)cmd >= PRC_NCMDS)
559 if (PRC_IS_REDIRECT(cmd))
560 notify = in6_rtchange, d = NULL;
561 else if (cmd == PRC_HOSTDEAD)
563 else if (inet6ctlerrmap[cmd] == 0)
566 /* if the parameter is from icmp6, decode it. */
568 ip6cp = (struct ip6ctlparam *)d;
570 ip6 = ip6cp->ip6c_ip6;
571 off = ip6cp->ip6c_off;
572 cmdarg = ip6cp->ip6c_cmdarg;
573 sa6_src = ip6cp->ip6c_src;
583 * XXX: We assume that when IPV6 is non NULL,
584 * M and OFF are valid.
587 /* Check if we can safely examine src and dst ports. */
588 if (m->m_pkthdr.len < off + sizeof(*uhp))
591 bzero(&uh, sizeof(uh));
592 m_copydata(m, off, sizeof(*uhp), (caddr_t)&uh);
594 if (!PRC_IS_REDIRECT(cmd)) {
595 /* Check to see if its tunneled */
597 inp = in6_pcblookup_mbuf(pcbinfo, &ip6->ip6_dst,
598 uh.uh_dport, &ip6->ip6_src, uh.uh_sport,
599 INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB,
600 m->m_pkthdr.rcvif, m);
605 if (up->u_icmp_func) {
608 (*up->u_icmp_func)(cmd, (struct sockaddr *)ip6cp->ip6c_src,
617 (void)in6_pcbnotify(pcbinfo, sa, uh.uh_dport,
618 (struct sockaddr *)ip6cp->ip6c_src, uh.uh_sport, cmd,
621 (void)in6_pcbnotify(pcbinfo, sa, 0,
622 (const struct sockaddr *)sa6_src, 0, cmd, cmdarg, notify);
626 udp6_ctlinput(int cmd, struct sockaddr *sa, void *d)
629 return (udp6_common_ctlinput(cmd, sa, d, &V_udbinfo));
633 udplite6_ctlinput(int cmd, struct sockaddr *sa, void *d)
636 return (udp6_common_ctlinput(cmd, sa, d, &V_ulitecbinfo));
640 udp6_getcred(SYSCTL_HANDLER_ARGS)
643 struct sockaddr_in6 addrs[2];
644 struct epoch_tracker et;
648 error = priv_check(req->td, PRIV_NETINET_GETCRED);
652 if (req->newlen != sizeof(addrs))
654 if (req->oldlen != sizeof(struct xucred))
656 error = SYSCTL_IN(req, addrs, sizeof(addrs));
659 if ((error = sa6_embedscope(&addrs[0], V_ip6_use_defzone)) != 0 ||
660 (error = sa6_embedscope(&addrs[1], V_ip6_use_defzone)) != 0) {
664 inp = in6_pcblookup(&V_udbinfo, &addrs[1].sin6_addr,
665 addrs[1].sin6_port, &addrs[0].sin6_addr, addrs[0].sin6_port,
666 INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB, NULL);
669 INP_RLOCK_ASSERT(inp);
670 if (inp->inp_socket == NULL)
673 error = cr_canseesocket(req->td->td_ucred,
676 cru2x(inp->inp_cred, &xuc);
681 error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred));
685 SYSCTL_PROC(_net_inet6_udp6, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW, 0,
686 0, udp6_getcred, "S,xucred", "Get the xucred of a UDP6 connection");
689 udp6_output(struct socket *so, int flags_arg, struct mbuf *m,
690 struct sockaddr *addr6, struct mbuf *control, struct thread *td)
695 struct in6_addr *laddr, *faddr, in6a;
696 struct ip6_pktopts *optp, opt;
697 struct sockaddr_in6 *sin6, tmp;
698 struct epoch_tracker et;
699 int cscov_partial, error, flags, hlen, scope_ambiguous;
700 u_int32_t ulen, plen;
705 /* addr6 has been validated in udp6_send(). */
706 sin6 = (struct sockaddr_in6 *)addr6;
709 * In contrast to to IPv4 we do not validate the max. packet length
710 * here due to IPv6 Jumbograms (RFC2675).
715 /* Protect *addr6 from overwrites. */
720 * Application should provide a proper zone ID or the use of
721 * default zone IDs should be enabled. Unfortunately, some
722 * applications do not behave as it should, so we need a
723 * workaround. Even if an appropriate ID is not determined,
724 * we'll see if we can determine the outgoing interface. If we
725 * can, determine the zone ID based on the interface below.
727 if (sin6->sin6_scope_id == 0 && !V_ip6_use_defzone)
729 if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0) {
738 KASSERT(inp != NULL, ("%s: inp == NULL", __func__));
740 * In the following cases we want a write lock on the inp for either
741 * local operations or for possible route cache updates in the IPv6
743 * - on connected sockets (sin6 is NULL) for route cache updates,
744 * - when we are not bound to an address and source port (it is
745 * in6_pcbsetport() which will require the write lock).
747 * We check the inp fields before actually locking the inp, so
748 * here exists a race, and we may WLOCK the inp and end with already
749 * bound one by other thread. This is fine.
751 if (sin6 == NULL || (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
752 inp->inp_lport == 0))
757 nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ?
758 IPPROTO_UDP : IPPROTO_UDPLITE;
761 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
765 hasv4addr = (inp->inp_vflag & INP_IPV4);
767 hasv4addr = IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)
770 struct pr_usrreqs *pru;
773 * XXXRW: We release UDP-layer locks before calling
774 * udp_send() in order to avoid recursion. However,
775 * this does mean there is a short window where inp's
776 * fields are unstable. Could this lead to a
777 * potential race in which the factors causing us to
778 * select the UDPv4 output routine are invalidated?
782 in6_sin6_2_sin_in_sock((struct sockaddr *)sin6);
783 pru = inetsw[ip_protox[nxt]].pr_usrreqs;
784 /* addr will just be freed in sendit(). */
785 return ((*pru->pru_send)(so, flags_arg, m,
786 (struct sockaddr *)sin6, control, td));
790 if (sin6 && IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
792 * Given this is either an IPv6-only socket or no INET is
793 * supported we will fail the send if the given destination
794 * address is a v4mapped address.
796 * XXXGL: do we leak m and control?
803 if ((error = ip6_setpktopts(control, &opt,
804 inp->in6p_outputopts, td->td_ucred, nxt)) != 0) {
806 ip6_clearpktopts(&opt, -1);
814 optp = inp->in6p_outputopts;
820 * Since we saw no essential reason for calling in_pcbconnect,
821 * we get rid of such kind of logic, and call in6_selectsrc
822 * and in6_pcbsetport in order to fill in the local address
823 * and the local port.
825 if (sin6->sin6_port == 0) {
826 error = EADDRNOTAVAIL;
830 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
831 /* how about ::ffff:0.0.0.0 case? */
837 * Given we handle the v4mapped case in the INET block above
838 * assert here that it must not happen anymore.
840 KASSERT(!IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr),
841 ("%s: sin6(%p)->sin6_addr is v4mapped which we "
842 "should have handled.", __func__, sin6));
844 /* This only requires read-locking. */
845 error = in6_selectsrc_socket(sin6, optp, inp,
846 td->td_ucred, scope_ambiguous, &in6a, NULL);
851 if (inp->inp_lport == 0) {
852 struct inpcbinfo *pcbinfo;
854 INP_WLOCK_ASSERT(inp);
856 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
857 INP_HASH_WLOCK(pcbinfo);
858 error = in6_pcbsetport(laddr, inp, td->td_ucred);
859 INP_HASH_WUNLOCK(pcbinfo);
861 /* Undo an address bind that may have occurred. */
862 inp->in6p_laddr = in6addr_any;
866 faddr = &sin6->sin6_addr;
867 fport = sin6->sin6_port; /* allow 0 port */
870 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
874 laddr = &inp->in6p_laddr;
875 faddr = &inp->in6p_faddr;
876 fport = inp->inp_fport;
879 ulen = m->m_pkthdr.len;
880 plen = sizeof(struct udphdr) + ulen;
881 hlen = sizeof(struct ip6_hdr);
884 * Calculate data length and get a mbuf
885 * for UDP and IP6 headers.
887 M_PREPEND(m, hlen + sizeof(struct udphdr), M_NOWAIT);
894 * Stuff checksum and output datagram.
896 cscov = cscov_partial = 0;
897 udp6 = (struct udphdr *)(mtod(m, caddr_t) + hlen);
898 udp6->uh_sport = inp->inp_lport; /* lport is always set in the PCB */
899 udp6->uh_dport = fport;
900 if (nxt == IPPROTO_UDPLITE) {
904 cscov = up->u_txcslen;
907 udp6->uh_ulen = htons(cscov);
909 * For UDP-Lite, checksum coverage length of zero means
910 * the entire UDPLite packet is covered by the checksum.
912 cscov_partial = (cscov == 0) ? 0 : 1;
913 } else if (plen <= 0xffff)
914 udp6->uh_ulen = htons((u_short)plen);
919 ip6 = mtod(m, struct ip6_hdr *);
920 ip6->ip6_flow = inp->inp_flow & IPV6_FLOWINFO_MASK;
921 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
922 ip6->ip6_vfc |= IPV6_VERSION;
923 ip6->ip6_plen = htons((u_short)plen);
925 ip6->ip6_hlim = in6_selecthlim(inp, NULL);
926 ip6->ip6_src = *laddr;
927 ip6->ip6_dst = *faddr;
930 mac_inpcb_create_mbuf(inp, m);
934 if ((udp6->uh_sum = in6_cksum_partial(m, nxt,
935 sizeof(struct ip6_hdr), plen, cscov)) == 0)
936 udp6->uh_sum = 0xffff;
938 udp6->uh_sum = in6_cksum_pseudo(ip6, plen, nxt, 0);
939 m->m_pkthdr.csum_flags = CSUM_UDP_IPV6;
940 m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
946 uint32_t hash_val, hash_type;
949 pr = inp->inp_socket->so_proto->pr_protocol;
951 * Calculate an appropriate RSS hash for UDP and
954 * The called function will take care of figuring out
955 * whether a 2-tuple or 4-tuple hash is required based
956 * on the currently configured scheme.
958 * Later later on connected socket values should be
959 * cached in the inpcb and reused, rather than constantly
962 * UDP Lite is a different protocol number and will
963 * likely end up being hashed as a 2-tuple until
964 * RSS / NICs grow UDP Lite protocol awareness.
966 if (rss_proto_software_hash_v6(faddr, laddr, fport,
967 inp->inp_lport, pr, &hash_val, &hash_type) == 0) {
968 m->m_pkthdr.flowid = hash_val;
969 M_HASHTYPE_SET(m, hash_type);
973 * Don't override with the inp cached flowid.
975 * Until the whole UDP path is vetted, it may actually
978 flags |= IP_NODEFAULTFLOWID;
982 UDPSTAT_INC(udps_opackets);
983 if (nxt == IPPROTO_UDPLITE)
984 UDPLITE_PROBE(send, NULL, inp, ip6, inp, udp6);
986 UDP_PROBE(send, NULL, inp, ip6, inp, udp6);
987 error = ip6_output(m, optp,
988 INP_WLOCKED(inp) ? &inp->inp_route6 : NULL, flags,
989 inp->in6p_moptions, NULL, inp);
994 ip6_clearpktopts(&opt, -1);
1003 ip6_clearpktopts(&opt, -1);
1012 udp6_abort(struct socket *so)
1015 struct inpcbinfo *pcbinfo;
1017 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1018 inp = sotoinpcb(so);
1019 KASSERT(inp != NULL, ("udp6_abort: inp == NULL"));
1023 if (inp->inp_vflag & INP_IPV4) {
1024 struct pr_usrreqs *pru;
1027 nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ?
1028 IPPROTO_UDP : IPPROTO_UDPLITE;
1030 pru = inetsw[ip_protox[nxt]].pr_usrreqs;
1031 (*pru->pru_abort)(so);
1036 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1037 INP_HASH_WLOCK(pcbinfo);
1038 in6_pcbdisconnect(inp);
1039 inp->in6p_laddr = in6addr_any;
1040 INP_HASH_WUNLOCK(pcbinfo);
1041 soisdisconnected(so);
1047 udp6_attach(struct socket *so, int proto, struct thread *td)
1050 struct inpcbinfo *pcbinfo;
1053 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1054 inp = sotoinpcb(so);
1055 KASSERT(inp == NULL, ("udp6_attach: inp != NULL"));
1057 if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
1058 error = soreserve(so, udp_sendspace, udp_recvspace);
1062 INP_INFO_WLOCK(pcbinfo);
1063 error = in_pcballoc(so, pcbinfo);
1065 INP_INFO_WUNLOCK(pcbinfo);
1068 inp = (struct inpcb *)so->so_pcb;
1069 inp->inp_vflag |= INP_IPV6;
1070 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0)
1071 inp->inp_vflag |= INP_IPV4;
1072 inp->in6p_hops = -1; /* use kernel default */
1073 inp->in6p_cksum = -1; /* just to be sure */
1076 * IPv4 TTL initialization is necessary for an IPv6 socket as well,
1077 * because the socket may be bound to an IPv6 wildcard address,
1078 * which may match an IPv4-mapped IPv6 address.
1080 inp->inp_ip_ttl = V_ip_defttl;
1082 error = udp_newudpcb(inp);
1086 INP_INFO_WUNLOCK(pcbinfo);
1090 INP_INFO_WUNLOCK(pcbinfo);
1095 udp6_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
1098 struct inpcbinfo *pcbinfo;
1102 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1103 inp = sotoinpcb(so);
1104 KASSERT(inp != NULL, ("udp6_bind: inp == NULL"));
1107 INP_HASH_WLOCK(pcbinfo);
1108 vflagsav = inp->inp_vflag;
1109 inp->inp_vflag &= ~INP_IPV4;
1110 inp->inp_vflag |= INP_IPV6;
1111 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
1112 struct sockaddr_in6 *sin6_p;
1114 sin6_p = (struct sockaddr_in6 *)nam;
1116 if (IN6_IS_ADDR_UNSPECIFIED(&sin6_p->sin6_addr))
1117 inp->inp_vflag |= INP_IPV4;
1119 else if (IN6_IS_ADDR_V4MAPPED(&sin6_p->sin6_addr)) {
1120 struct sockaddr_in sin;
1122 in6_sin6_2_sin(&sin, sin6_p);
1123 inp->inp_vflag |= INP_IPV4;
1124 inp->inp_vflag &= ~INP_IPV6;
1125 error = in_pcbbind(inp, (struct sockaddr *)&sin,
1132 error = in6_pcbbind(inp, nam, td->td_ucred);
1137 inp->inp_vflag = vflagsav;
1138 INP_HASH_WUNLOCK(pcbinfo);
1144 udp6_close(struct socket *so)
1147 struct inpcbinfo *pcbinfo;
1149 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1150 inp = sotoinpcb(so);
1151 KASSERT(inp != NULL, ("udp6_close: inp == NULL"));
1155 if (inp->inp_vflag & INP_IPV4) {
1156 struct pr_usrreqs *pru;
1159 nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ?
1160 IPPROTO_UDP : IPPROTO_UDPLITE;
1162 pru = inetsw[ip_protox[nxt]].pr_usrreqs;
1163 (*pru->pru_disconnect)(so);
1167 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1168 INP_HASH_WLOCK(pcbinfo);
1169 in6_pcbdisconnect(inp);
1170 inp->in6p_laddr = in6addr_any;
1171 INP_HASH_WUNLOCK(pcbinfo);
1172 soisdisconnected(so);
1178 udp6_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
1181 struct inpcbinfo *pcbinfo;
1182 struct sockaddr_in6 *sin6;
1186 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1187 inp = sotoinpcb(so);
1188 sin6 = (struct sockaddr_in6 *)nam;
1189 KASSERT(inp != NULL, ("udp6_connect: inp == NULL"));
1192 * XXXRW: Need to clarify locking of v4/v6 flags.
1196 if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
1197 struct sockaddr_in sin;
1199 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) {
1203 if ((inp->inp_vflag & INP_IPV4) == 0) {
1204 error = EAFNOSUPPORT;
1207 if (inp->inp_faddr.s_addr != INADDR_ANY) {
1211 in6_sin6_2_sin(&sin, sin6);
1212 error = prison_remote_ip4(td->td_ucred, &sin.sin_addr);
1215 vflagsav = inp->inp_vflag;
1216 inp->inp_vflag |= INP_IPV4;
1217 inp->inp_vflag &= ~INP_IPV6;
1218 INP_HASH_WLOCK(pcbinfo);
1219 error = in_pcbconnect(inp, (struct sockaddr *)&sin,
1221 INP_HASH_WUNLOCK(pcbinfo);
1223 * If connect succeeds, mark socket as connected. If
1224 * connect fails and socket is unbound, reset inp_vflag
1229 else if (inp->inp_laddr.s_addr == INADDR_ANY &&
1230 inp->inp_lport == 0)
1231 inp->inp_vflag = vflagsav;
1234 if ((inp->inp_vflag & INP_IPV6) == 0) {
1235 error = EAFNOSUPPORT;
1240 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1244 error = prison_remote_ip6(td->td_ucred, &sin6->sin6_addr);
1247 vflagsav = inp->inp_vflag;
1248 inp->inp_vflag &= ~INP_IPV4;
1249 inp->inp_vflag |= INP_IPV6;
1250 INP_HASH_WLOCK(pcbinfo);
1251 error = in6_pcbconnect(inp, nam, td->td_ucred);
1252 INP_HASH_WUNLOCK(pcbinfo);
1254 * If connect succeeds, mark socket as connected. If
1255 * connect fails and socket is unbound, reset inp_vflag
1260 else if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
1261 inp->inp_lport == 0)
1262 inp->inp_vflag = vflagsav;
1269 udp6_detach(struct socket *so)
1272 struct inpcbinfo *pcbinfo;
1275 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1276 inp = sotoinpcb(so);
1277 KASSERT(inp != NULL, ("udp6_detach: inp == NULL"));
1279 INP_INFO_WLOCK(pcbinfo);
1281 up = intoudpcb(inp);
1282 KASSERT(up != NULL, ("%s: up == NULL", __func__));
1285 INP_INFO_WUNLOCK(pcbinfo);
1290 udp6_disconnect(struct socket *so)
1293 struct inpcbinfo *pcbinfo;
1295 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1296 inp = sotoinpcb(so);
1297 KASSERT(inp != NULL, ("udp6_disconnect: inp == NULL"));
1301 if (inp->inp_vflag & INP_IPV4) {
1302 struct pr_usrreqs *pru;
1305 nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ?
1306 IPPROTO_UDP : IPPROTO_UDPLITE;
1308 pru = inetsw[ip_protox[nxt]].pr_usrreqs;
1309 (void)(*pru->pru_disconnect)(so);
1314 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1319 INP_HASH_WLOCK(pcbinfo);
1320 in6_pcbdisconnect(inp);
1321 inp->in6p_laddr = in6addr_any;
1322 INP_HASH_WUNLOCK(pcbinfo);
1324 so->so_state &= ~SS_ISCONNECTED; /* XXX */
1331 udp6_send(struct socket *so, int flags, struct mbuf *m,
1332 struct sockaddr *addr, struct mbuf *control, struct thread *td)
1337 if (addr->sa_len != sizeof(struct sockaddr_in6)) {
1341 if (addr->sa_family != AF_INET6) {
1342 error = EAFNOSUPPORT;
1347 return (udp6_output(so, flags, m, addr, control, td));
1356 struct pr_usrreqs udp6_usrreqs = {
1357 .pru_abort = udp6_abort,
1358 .pru_attach = udp6_attach,
1359 .pru_bind = udp6_bind,
1360 .pru_connect = udp6_connect,
1361 .pru_control = in6_control,
1362 .pru_detach = udp6_detach,
1363 .pru_disconnect = udp6_disconnect,
1364 .pru_peeraddr = in6_mapped_peeraddr,
1365 .pru_send = udp6_send,
1366 .pru_shutdown = udp_shutdown,
1367 .pru_sockaddr = in6_mapped_sockaddr,
1368 .pru_soreceive = soreceive_dgram,
1369 .pru_sosend = sosend_dgram,
1370 .pru_sosetlabel = in_pcbsosetlabel,
1371 .pru_close = udp6_close