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);
211 udp6_input(struct mbuf **mp, int *offp, int proto)
213 struct mbuf *m = *mp;
218 struct inpcbinfo *pcbinfo;
223 struct sockaddr_in6 fromsa[2];
224 struct m_tag *fwd_tag;
230 ifp = m->m_pkthdr.rcvif;
232 if (m->m_len < off + sizeof(struct udphdr)) {
233 m = m_pullup(m, off + sizeof(struct udphdr));
235 IP6STAT_INC(ip6s_exthdrtoolong);
237 return (IPPROTO_DONE);
240 ip6 = mtod(m, struct ip6_hdr *);
241 uh = (struct udphdr *)((caddr_t)ip6 + off);
243 UDPSTAT_INC(udps_ipackets);
246 * Destination port of 0 is illegal, based on RFC768.
248 if (uh->uh_dport == 0)
251 plen = ntohs(ip6->ip6_plen) - off + sizeof(*ip6);
252 ulen = ntohs((u_short)uh->uh_ulen);
255 cscov_partial = (nxt == IPPROTO_UDPLITE) ? 1 : 0;
256 if (nxt == IPPROTO_UDPLITE) {
257 /* Zero means checksum over the complete packet. */
262 if ((ulen < sizeof(struct udphdr)) || (ulen > plen)) {
263 /* XXX: What is the right UDPLite MIB counter? */
266 if (uh->uh_sum == 0) {
267 /* XXX: What is the right UDPLite MIB counter? */
271 if ((ulen < sizeof(struct udphdr)) || (plen != ulen)) {
272 UDPSTAT_INC(udps_badlen);
275 if (uh->uh_sum == 0) {
276 UDPSTAT_INC(udps_nosum);
278 * dport 0 was rejected earlier so this is OK even if
279 * zero_checksum_port is 0 (which is its default value).
281 if (ntohs(uh->uh_dport) == V_zero_checksum_port)
288 if ((m->m_pkthdr.csum_flags & CSUM_DATA_VALID_IPV6) &&
290 if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR)
291 uh_sum = m->m_pkthdr.csum_data;
293 uh_sum = in6_cksum_pseudo(ip6, ulen, nxt,
294 m->m_pkthdr.csum_data);
297 uh_sum = in6_cksum_partial(m, nxt, off, plen, ulen);
300 UDPSTAT_INC(udps_badsum);
306 * Construct sockaddr format source address.
308 init_sin6(&fromsa[0], m, 0);
309 fromsa[0].sin6_port = uh->uh_sport;
310 init_sin6(&fromsa[1], m, 1);
311 fromsa[1].sin6_port = uh->uh_dport;
313 pcbinfo = udp_get_inpcbinfo(nxt);
314 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
316 struct inpcbhead *pcblist;
317 struct ip6_moptions *imo;
320 * In the event that laddr should be set to the link-local
321 * address (this happens in RIPng), the multicast address
322 * specified in the received packet will not match laddr. To
323 * handle this situation, matching is relaxed if the
324 * receiving interface is the same as one specified in the
325 * socket and if the destination multicast address matches
326 * one of the multicast groups specified in the socket.
330 * KAME note: traditionally we dropped udpiphdr from mbuf
331 * here. We need udphdr for IPsec processing so we do that
334 pcblist = udp_get_pcblist(nxt);
336 CK_LIST_FOREACH(inp, pcblist, inp_list) {
337 if ((inp->inp_vflag & INP_IPV6) == 0)
339 if (inp->inp_lport != uh->uh_dport)
341 if (inp->inp_fport != 0 &&
342 inp->inp_fport != uh->uh_sport)
344 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
345 if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr,
349 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
350 if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr,
352 inp->inp_fport != uh->uh_sport)
358 if (__predict_false(inp->inp_flags2 & INP_FREED)) {
364 * XXXRW: Because we weren't holding either the inpcb
365 * or the hash lock when we checked for a match
366 * before, we should probably recheck now that the
367 * inpcb lock is (supposed to be) held.
371 * Handle socket delivery policy for any-source
372 * and source-specific multicast. [RFC3678]
374 imo = inp->in6p_moptions;
376 struct sockaddr_in6 mcaddr;
379 bzero(&mcaddr, sizeof(struct sockaddr_in6));
380 mcaddr.sin6_len = sizeof(struct sockaddr_in6);
381 mcaddr.sin6_family = AF_INET6;
382 mcaddr.sin6_addr = ip6->ip6_dst;
384 blocked = im6o_mc_filter(imo, ifp,
385 (struct sockaddr *)&mcaddr,
386 (struct sockaddr *)&fromsa[0]);
387 if (blocked != MCAST_PASS) {
388 if (blocked == MCAST_NOTGMEMBER)
389 IP6STAT_INC(ip6s_notmember);
390 if (blocked == MCAST_NOTSMEMBER ||
391 blocked == MCAST_MUTED)
392 UDPSTAT_INC(udps_filtermcast);
401 if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) !=
403 if (nxt == IPPROTO_UDPLITE)
404 UDPLITE_PROBE(receive, NULL,
405 last, ip6, last, uh);
407 UDP_PROBE(receive, NULL, last,
409 if (udp6_append(last, n, off,
415 /* Release PCB lock taken on previous pass. */
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 (nxt == IPPROTO_UDPLITE)
444 UDPLITE_PROBE(receive, NULL, last, ip6, last, uh);
446 UDP_PROBE(receive, NULL, last, ip6, last, uh);
447 if (udp6_append(last, m, off, fromsa) == 0)
450 return (IPPROTO_DONE);
453 * Locate pcb for datagram.
457 * Grab info from PACKET_TAG_IPFORWARD tag prepended to the chain.
459 if ((m->m_flags & M_IP6_NEXTHOP) &&
460 (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
461 struct sockaddr_in6 *next_hop6;
463 next_hop6 = (struct sockaddr_in6 *)(fwd_tag + 1);
466 * Transparently forwarded. Pretend to be the destination.
467 * Already got one like this?
469 inp = in6_pcblookup_mbuf(pcbinfo, &ip6->ip6_src,
470 uh->uh_sport, &ip6->ip6_dst, uh->uh_dport,
471 INPLOOKUP_RLOCKPCB, m->m_pkthdr.rcvif, m);
474 * It's new. Try to find the ambushing socket.
475 * Because we've rewritten the destination address,
476 * any hardware-generated hash is ignored.
478 inp = in6_pcblookup(pcbinfo, &ip6->ip6_src,
479 uh->uh_sport, &next_hop6->sin6_addr,
480 next_hop6->sin6_port ? htons(next_hop6->sin6_port) :
481 uh->uh_dport, INPLOOKUP_WILDCARD |
482 INPLOOKUP_RLOCKPCB, m->m_pkthdr.rcvif);
484 /* Remove the tag from the packet. We don't need it anymore. */
485 m_tag_delete(m, fwd_tag);
486 m->m_flags &= ~M_IP6_NEXTHOP;
488 inp = in6_pcblookup_mbuf(pcbinfo, &ip6->ip6_src,
489 uh->uh_sport, &ip6->ip6_dst, uh->uh_dport,
490 INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB,
491 m->m_pkthdr.rcvif, m);
493 if (V_udp_log_in_vain) {
494 char ip6bufs[INET6_ADDRSTRLEN];
495 char ip6bufd[INET6_ADDRSTRLEN];
498 "Connection attempt to UDP [%s]:%d from [%s]:%d\n",
499 ip6_sprintf(ip6bufd, &ip6->ip6_dst),
501 ip6_sprintf(ip6bufs, &ip6->ip6_src),
502 ntohs(uh->uh_sport));
504 if (nxt == IPPROTO_UDPLITE)
505 UDPLITE_PROBE(receive, NULL, NULL, ip6, NULL, uh);
507 UDP_PROBE(receive, NULL, NULL, ip6, NULL, uh);
508 UDPSTAT_INC(udps_noport);
509 if (m->m_flags & M_MCAST) {
510 printf("UDP6: M_MCAST is set in a unicast packet.\n");
511 UDPSTAT_INC(udps_noportmcast);
516 icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0);
518 return (IPPROTO_DONE);
520 INP_RLOCK_ASSERT(inp);
523 if (up->u_rxcslen == 0 || up->u_rxcslen > ulen) {
527 return (IPPROTO_DONE);
530 if (nxt == IPPROTO_UDPLITE)
531 UDPLITE_PROBE(receive, NULL, inp, ip6, inp, uh);
533 UDP_PROBE(receive, NULL, inp, ip6, inp, uh);
534 if (udp6_append(inp, m, off, fromsa) == 0)
537 return (IPPROTO_DONE);
542 return (IPPROTO_DONE);
546 udp6_common_ctlinput(int cmd, struct sockaddr *sa, void *d,
547 struct inpcbinfo *pcbinfo)
553 struct ip6ctlparam *ip6cp = NULL;
554 const struct sockaddr_in6 *sa6_src = NULL;
556 struct inpcb *(*notify)(struct inpcb *, int) = udp_notify;
557 struct udp_portonly {
562 if (sa->sa_family != AF_INET6 ||
563 sa->sa_len != sizeof(struct sockaddr_in6))
566 if ((unsigned)cmd >= PRC_NCMDS)
568 if (PRC_IS_REDIRECT(cmd))
569 notify = in6_rtchange, d = NULL;
570 else if (cmd == PRC_HOSTDEAD)
572 else if (inet6ctlerrmap[cmd] == 0)
575 /* if the parameter is from icmp6, decode it. */
577 ip6cp = (struct ip6ctlparam *)d;
579 ip6 = ip6cp->ip6c_ip6;
580 off = ip6cp->ip6c_off;
581 cmdarg = ip6cp->ip6c_cmdarg;
582 sa6_src = ip6cp->ip6c_src;
592 * XXX: We assume that when IPV6 is non NULL,
593 * M and OFF are valid.
596 /* Check if we can safely examine src and dst ports. */
597 if (m->m_pkthdr.len < off + sizeof(*uhp))
600 bzero(&uh, sizeof(uh));
601 m_copydata(m, off, sizeof(*uhp), (caddr_t)&uh);
603 if (!PRC_IS_REDIRECT(cmd)) {
604 /* Check to see if its tunneled */
606 inp = in6_pcblookup_mbuf(pcbinfo, &ip6->ip6_dst,
607 uh.uh_dport, &ip6->ip6_src, uh.uh_sport,
608 INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB,
609 m->m_pkthdr.rcvif, m);
614 if (up->u_icmp_func) {
617 (*up->u_icmp_func)(cmd, (struct sockaddr *)ip6cp->ip6c_src,
626 (void)in6_pcbnotify(pcbinfo, sa, uh.uh_dport,
627 (struct sockaddr *)ip6cp->ip6c_src, uh.uh_sport, cmd,
630 (void)in6_pcbnotify(pcbinfo, sa, 0,
631 (const struct sockaddr *)sa6_src, 0, cmd, cmdarg, notify);
635 udp6_ctlinput(int cmd, struct sockaddr *sa, void *d)
638 return (udp6_common_ctlinput(cmd, sa, d, &V_udbinfo));
642 udplite6_ctlinput(int cmd, struct sockaddr *sa, void *d)
645 return (udp6_common_ctlinput(cmd, sa, d, &V_ulitecbinfo));
649 udp6_getcred(SYSCTL_HANDLER_ARGS)
652 struct sockaddr_in6 addrs[2];
653 struct epoch_tracker et;
657 error = priv_check(req->td, PRIV_NETINET_GETCRED);
661 if (req->newlen != sizeof(addrs))
663 if (req->oldlen != sizeof(struct xucred))
665 error = SYSCTL_IN(req, addrs, sizeof(addrs));
668 if ((error = sa6_embedscope(&addrs[0], V_ip6_use_defzone)) != 0 ||
669 (error = sa6_embedscope(&addrs[1], V_ip6_use_defzone)) != 0) {
673 inp = in6_pcblookup(&V_udbinfo, &addrs[1].sin6_addr,
674 addrs[1].sin6_port, &addrs[0].sin6_addr, addrs[0].sin6_port,
675 INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB, NULL);
678 INP_RLOCK_ASSERT(inp);
679 if (inp->inp_socket == NULL)
682 error = cr_canseesocket(req->td->td_ucred,
685 cru2x(inp->inp_cred, &xuc);
690 error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred));
694 SYSCTL_PROC(_net_inet6_udp6, OID_AUTO, getcred,
695 CTLTYPE_OPAQUE | CTLFLAG_RW | CTLFLAG_MPSAFE,
696 0, 0, udp6_getcred, "S,xucred",
697 "Get the xucred of a UDP6 connection");
700 udp6_output(struct socket *so, int flags_arg, struct mbuf *m,
701 struct sockaddr *addr6, struct mbuf *control, struct thread *td)
706 struct in6_addr *laddr, *faddr, in6a;
707 struct ip6_pktopts *optp, opt;
708 struct sockaddr_in6 *sin6, tmp;
709 struct epoch_tracker et;
710 int cscov_partial, error, flags, hlen, scope_ambiguous;
711 u_int32_t ulen, plen;
716 /* addr6 has been validated in udp6_send(). */
717 sin6 = (struct sockaddr_in6 *)addr6;
720 * In contrast to IPv4 we do not validate the max. packet length
721 * here due to IPv6 Jumbograms (RFC2675).
726 /* Protect *addr6 from overwrites. */
731 * Application should provide a proper zone ID or the use of
732 * default zone IDs should be enabled. Unfortunately, some
733 * applications do not behave as it should, so we need a
734 * workaround. Even if an appropriate ID is not determined,
735 * we'll see if we can determine the outgoing interface. If we
736 * can, determine the zone ID based on the interface below.
738 if (sin6->sin6_scope_id == 0 && !V_ip6_use_defzone)
740 if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0) {
749 KASSERT(inp != NULL, ("%s: inp == NULL", __func__));
751 * In the following cases we want a write lock on the inp for either
752 * local operations or for possible route cache updates in the IPv6
754 * - on connected sockets (sin6 is NULL) for route cache updates,
755 * - when we are not bound to an address and source port (it is
756 * in6_pcbsetport() which will require the write lock).
758 * We check the inp fields before actually locking the inp, so
759 * here exists a race, and we may WLOCK the inp and end with already
760 * bound one by other thread. This is fine.
762 if (sin6 == NULL || (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
763 inp->inp_lport == 0))
768 nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ?
769 IPPROTO_UDP : IPPROTO_UDPLITE;
772 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
776 hasv4addr = (inp->inp_vflag & INP_IPV4);
778 hasv4addr = IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)
781 struct pr_usrreqs *pru;
784 * XXXRW: We release UDP-layer locks before calling
785 * udp_send() in order to avoid recursion. However,
786 * this does mean there is a short window where inp's
787 * fields are unstable. Could this lead to a
788 * potential race in which the factors causing us to
789 * select the UDPv4 output routine are invalidated?
793 in6_sin6_2_sin_in_sock((struct sockaddr *)sin6);
794 pru = inetsw[ip_protox[nxt]].pr_usrreqs;
795 /* addr will just be freed in sendit(). */
796 return ((*pru->pru_send)(so, flags_arg | PRUS_IPV6, m,
797 (struct sockaddr *)sin6, control, td));
801 if (sin6 && IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
803 * Given this is either an IPv6-only socket or no INET is
804 * supported we will fail the send if the given destination
805 * address is a v4mapped address.
807 * XXXGL: do we leak m and control?
815 if ((error = ip6_setpktopts(control, &opt,
816 inp->in6p_outputopts, td->td_ucred, nxt)) != 0) {
821 optp = inp->in6p_outputopts;
825 * Since we saw no essential reason for calling in_pcbconnect,
826 * we get rid of such kind of logic, and call in6_selectsrc
827 * and in6_pcbsetport in order to fill in the local address
828 * and the local port.
830 if (sin6->sin6_port == 0) {
831 error = EADDRNOTAVAIL;
835 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
836 /* how about ::ffff:0.0.0.0 case? */
842 * Given we handle the v4mapped case in the INET block above
843 * assert here that it must not happen anymore.
845 KASSERT(!IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr),
846 ("%s: sin6(%p)->sin6_addr is v4mapped which we "
847 "should have handled.", __func__, sin6));
849 /* This only requires read-locking. */
850 error = in6_selectsrc_socket(sin6, optp, inp,
851 td->td_ucred, scope_ambiguous, &in6a, NULL);
856 if (inp->inp_lport == 0) {
857 struct inpcbinfo *pcbinfo;
859 INP_WLOCK_ASSERT(inp);
861 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
862 INP_HASH_WLOCK(pcbinfo);
863 error = in6_pcbsetport(laddr, inp, td->td_ucred);
864 INP_HASH_WUNLOCK(pcbinfo);
866 /* Undo an address bind that may have occurred. */
867 inp->in6p_laddr = in6addr_any;
871 faddr = &sin6->sin6_addr;
872 fport = sin6->sin6_port; /* allow 0 port */
875 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
879 laddr = &inp->in6p_laddr;
880 faddr = &inp->in6p_faddr;
881 fport = inp->inp_fport;
884 ulen = m->m_pkthdr.len;
885 plen = sizeof(struct udphdr) + ulen;
886 hlen = sizeof(struct ip6_hdr);
889 * Calculate data length and get a mbuf
890 * for UDP and IP6 headers.
892 M_PREPEND(m, hlen + sizeof(struct udphdr), M_NOWAIT);
899 * Stuff checksum and output datagram.
901 cscov = cscov_partial = 0;
902 udp6 = (struct udphdr *)(mtod(m, caddr_t) + hlen);
903 udp6->uh_sport = inp->inp_lport; /* lport is always set in the PCB */
904 udp6->uh_dport = fport;
905 if (nxt == IPPROTO_UDPLITE) {
909 cscov = up->u_txcslen;
912 udp6->uh_ulen = htons(cscov);
914 * For UDP-Lite, checksum coverage length of zero means
915 * the entire UDPLite packet is covered by the checksum.
917 cscov_partial = (cscov == 0) ? 0 : 1;
918 } else if (plen <= 0xffff)
919 udp6->uh_ulen = htons((u_short)plen);
924 ip6 = mtod(m, struct ip6_hdr *);
925 ip6->ip6_flow = inp->inp_flow & IPV6_FLOWINFO_MASK;
926 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
927 ip6->ip6_vfc |= IPV6_VERSION;
928 ip6->ip6_plen = htons((u_short)plen);
930 ip6->ip6_hlim = in6_selecthlim(inp, NULL);
931 ip6->ip6_src = *laddr;
932 ip6->ip6_dst = *faddr;
935 mac_inpcb_create_mbuf(inp, m);
939 if ((udp6->uh_sum = in6_cksum_partial(m, nxt,
940 sizeof(struct ip6_hdr), plen, cscov)) == 0)
941 udp6->uh_sum = 0xffff;
943 udp6->uh_sum = in6_cksum_pseudo(ip6, plen, nxt, 0);
944 m->m_pkthdr.csum_flags = CSUM_UDP_IPV6;
945 m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
949 #if defined(ROUTE_MPATH) || defined(RSS)
950 if (CALC_FLOWID_OUTBOUND_SENDTO) {
951 uint32_t hash_type, hash_val;
954 pr = inp->inp_socket->so_proto->pr_protocol;
956 hash_val = fib6_calc_packet_hash(laddr, faddr,
957 inp->inp_lport, fport, pr, &hash_type);
958 m->m_pkthdr.flowid = hash_val;
959 M_HASHTYPE_SET(m, hash_type);
961 /* do not use inp flowid */
962 flags |= IP_NODEFAULTFLOWID;
965 UDPSTAT_INC(udps_opackets);
966 if (nxt == IPPROTO_UDPLITE)
967 UDPLITE_PROBE(send, NULL, inp, ip6, inp, udp6);
969 UDP_PROBE(send, NULL, inp, ip6, inp, udp6);
970 error = ip6_output(m, optp,
971 INP_WLOCKED(inp) ? &inp->inp_route6 : NULL, flags,
972 inp->in6p_moptions, NULL, inp);
977 ip6_clearpktopts(&opt, -1);
986 ip6_clearpktopts(&opt, -1);
995 udp6_abort(struct socket *so)
998 struct inpcbinfo *pcbinfo;
1000 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1001 inp = sotoinpcb(so);
1002 KASSERT(inp != NULL, ("udp6_abort: inp == NULL"));
1006 if (inp->inp_vflag & INP_IPV4) {
1007 struct pr_usrreqs *pru;
1010 nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ?
1011 IPPROTO_UDP : IPPROTO_UDPLITE;
1013 pru = inetsw[ip_protox[nxt]].pr_usrreqs;
1014 (*pru->pru_abort)(so);
1019 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1020 INP_HASH_WLOCK(pcbinfo);
1021 in6_pcbdisconnect(inp);
1022 inp->in6p_laddr = in6addr_any;
1023 INP_HASH_WUNLOCK(pcbinfo);
1024 soisdisconnected(so);
1030 udp6_attach(struct socket *so, int proto, struct thread *td)
1033 struct inpcbinfo *pcbinfo;
1036 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1037 inp = sotoinpcb(so);
1038 KASSERT(inp == NULL, ("udp6_attach: inp != NULL"));
1040 if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
1041 error = soreserve(so, udp_sendspace, udp_recvspace);
1045 INP_INFO_WLOCK(pcbinfo);
1046 error = in_pcballoc(so, pcbinfo);
1048 INP_INFO_WUNLOCK(pcbinfo);
1051 inp = (struct inpcb *)so->so_pcb;
1052 inp->inp_vflag |= INP_IPV6;
1053 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0)
1054 inp->inp_vflag |= INP_IPV4;
1055 inp->in6p_hops = -1; /* use kernel default */
1056 inp->in6p_cksum = -1; /* just to be sure */
1059 * IPv4 TTL initialization is necessary for an IPv6 socket as well,
1060 * because the socket may be bound to an IPv6 wildcard address,
1061 * which may match an IPv4-mapped IPv6 address.
1063 inp->inp_ip_ttl = V_ip_defttl;
1065 error = udp_newudpcb(inp);
1069 INP_INFO_WUNLOCK(pcbinfo);
1073 INP_INFO_WUNLOCK(pcbinfo);
1078 udp6_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
1081 struct inpcbinfo *pcbinfo;
1085 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1086 inp = sotoinpcb(so);
1087 KASSERT(inp != NULL, ("udp6_bind: inp == NULL"));
1089 if (nam->sa_family != AF_INET6)
1090 return (EAFNOSUPPORT);
1091 if (nam->sa_len != sizeof(struct sockaddr_in6))
1095 INP_HASH_WLOCK(pcbinfo);
1096 vflagsav = inp->inp_vflag;
1097 inp->inp_vflag &= ~INP_IPV4;
1098 inp->inp_vflag |= INP_IPV6;
1099 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
1100 struct sockaddr_in6 *sin6_p;
1102 sin6_p = (struct sockaddr_in6 *)nam;
1104 if (IN6_IS_ADDR_UNSPECIFIED(&sin6_p->sin6_addr))
1105 inp->inp_vflag |= INP_IPV4;
1107 else if (IN6_IS_ADDR_V4MAPPED(&sin6_p->sin6_addr)) {
1108 struct sockaddr_in sin;
1110 in6_sin6_2_sin(&sin, sin6_p);
1111 inp->inp_vflag |= INP_IPV4;
1112 inp->inp_vflag &= ~INP_IPV6;
1113 error = in_pcbbind(inp, (struct sockaddr *)&sin,
1120 error = in6_pcbbind(inp, nam, td->td_ucred);
1125 inp->inp_vflag = vflagsav;
1126 INP_HASH_WUNLOCK(pcbinfo);
1132 udp6_close(struct socket *so)
1135 struct inpcbinfo *pcbinfo;
1137 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1138 inp = sotoinpcb(so);
1139 KASSERT(inp != NULL, ("udp6_close: inp == NULL"));
1143 if (inp->inp_vflag & INP_IPV4) {
1144 struct pr_usrreqs *pru;
1147 nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ?
1148 IPPROTO_UDP : IPPROTO_UDPLITE;
1150 pru = inetsw[ip_protox[nxt]].pr_usrreqs;
1151 (*pru->pru_disconnect)(so);
1155 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1156 INP_HASH_WLOCK(pcbinfo);
1157 in6_pcbdisconnect(inp);
1158 inp->in6p_laddr = in6addr_any;
1159 INP_HASH_WUNLOCK(pcbinfo);
1160 soisdisconnected(so);
1166 udp6_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
1169 struct epoch_tracker et;
1172 struct inpcbinfo *pcbinfo;
1173 struct sockaddr_in6 *sin6;
1177 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1178 inp = sotoinpcb(so);
1179 KASSERT(inp != NULL, ("udp6_connect: inp == NULL"));
1181 sin6 = (struct sockaddr_in6 *)nam;
1182 if (sin6->sin6_family != AF_INET6)
1183 return (EAFNOSUPPORT);
1184 if (sin6->sin6_len != sizeof(*sin6))
1188 * XXXRW: Need to clarify locking of v4/v6 flags.
1192 if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
1193 struct sockaddr_in sin;
1195 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) {
1199 if ((inp->inp_vflag & INP_IPV4) == 0) {
1200 error = EAFNOSUPPORT;
1203 if (inp->inp_faddr.s_addr != INADDR_ANY) {
1207 in6_sin6_2_sin(&sin, sin6);
1208 error = prison_remote_ip4(td->td_ucred, &sin.sin_addr);
1211 vflagsav = inp->inp_vflag;
1212 inp->inp_vflag |= INP_IPV4;
1213 inp->inp_vflag &= ~INP_IPV6;
1214 NET_EPOCH_ENTER(et);
1215 INP_HASH_WLOCK(pcbinfo);
1216 error = in_pcbconnect(inp, (struct sockaddr *)&sin,
1218 INP_HASH_WUNLOCK(pcbinfo);
1221 * If connect succeeds, mark socket as connected. If
1222 * connect fails and socket is unbound, reset inp_vflag
1227 else if (inp->inp_laddr.s_addr == INADDR_ANY &&
1228 inp->inp_lport == 0)
1229 inp->inp_vflag = vflagsav;
1232 if ((inp->inp_vflag & INP_IPV6) == 0) {
1233 error = EAFNOSUPPORT;
1238 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1242 error = prison_remote_ip6(td->td_ucred, &sin6->sin6_addr);
1245 vflagsav = inp->inp_vflag;
1246 inp->inp_vflag &= ~INP_IPV4;
1247 inp->inp_vflag |= INP_IPV6;
1248 INP_HASH_WLOCK(pcbinfo);
1249 error = in6_pcbconnect(inp, nam, td->td_ucred);
1250 INP_HASH_WUNLOCK(pcbinfo);
1252 * If connect succeeds, mark socket as connected. If
1253 * connect fails and socket is unbound, reset inp_vflag
1258 else if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
1259 inp->inp_lport == 0)
1260 inp->inp_vflag = vflagsav;
1267 udp6_detach(struct socket *so)
1270 struct inpcbinfo *pcbinfo;
1273 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1274 inp = sotoinpcb(so);
1275 KASSERT(inp != NULL, ("udp6_detach: inp == NULL"));
1277 INP_INFO_WLOCK(pcbinfo);
1279 up = intoudpcb(inp);
1280 KASSERT(up != NULL, ("%s: up == NULL", __func__));
1283 INP_INFO_WUNLOCK(pcbinfo);
1288 udp6_disconnect(struct socket *so)
1291 struct inpcbinfo *pcbinfo;
1293 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1294 inp = sotoinpcb(so);
1295 KASSERT(inp != NULL, ("udp6_disconnect: inp == NULL"));
1299 if (inp->inp_vflag & INP_IPV4) {
1300 struct pr_usrreqs *pru;
1303 nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ?
1304 IPPROTO_UDP : IPPROTO_UDPLITE;
1306 pru = inetsw[ip_protox[nxt]].pr_usrreqs;
1307 (void)(*pru->pru_disconnect)(so);
1312 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1317 INP_HASH_WLOCK(pcbinfo);
1318 in6_pcbdisconnect(inp);
1319 inp->in6p_laddr = in6addr_any;
1320 INP_HASH_WUNLOCK(pcbinfo);
1322 so->so_state &= ~SS_ISCONNECTED; /* XXX */
1329 udp6_send(struct socket *so, int flags, struct mbuf *m,
1330 struct sockaddr *addr, struct mbuf *control, struct thread *td)
1335 if (addr->sa_len != sizeof(struct sockaddr_in6)) {
1339 if (addr->sa_family != AF_INET6) {
1340 error = EAFNOSUPPORT;
1345 return (udp6_output(so, flags, m, addr, control, td));
1354 struct pr_usrreqs udp6_usrreqs = {
1355 .pru_abort = udp6_abort,
1356 .pru_attach = udp6_attach,
1357 .pru_bind = udp6_bind,
1358 .pru_connect = udp6_connect,
1359 .pru_control = in6_control,
1360 .pru_detach = udp6_detach,
1361 .pru_disconnect = udp6_disconnect,
1362 .pru_peeraddr = in6_mapped_peeraddr,
1363 .pru_send = udp6_send,
1364 .pru_shutdown = udp_shutdown,
1365 .pru_sockaddr = in6_mapped_sockaddr,
1366 .pru_soreceive = soreceive_dgram,
1367 .pru_sosend = sosend_dgram,
1368 .pru_sosetlabel = in_pcbsosetlabel,
1369 .pru_close = udp6_close