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.");
136 /* netinet/udp_usrreqs.c */
137 pr_abort_t udp_abort;
138 pr_disconnect_t udp_disconnect;
142 * UDP protocol implementation.
143 * Per RFC 768, August, 1980.
146 static void udp6_detach(struct socket *so);
149 udp6_append(struct inpcb *inp, struct mbuf *n, int off,
150 struct sockaddr_in6 *fromsa)
153 struct mbuf *opts = NULL, *tmp_opts;
157 INP_LOCK_ASSERT(inp);
160 * Engage the tunneling protocol.
163 if (up->u_tun_func != NULL) {
166 filtered = (*up->u_tun_func)(n, off, inp,
167 (struct sockaddr *)&fromsa[0], up->u_tun_ctx);
170 return (in_pcbrele_rlocked(inp));
172 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
173 /* Check AH/ESP integrity. */
174 if (IPSEC_ENABLED(ipv6)) {
175 if (IPSEC_CHECK_POLICY(ipv6, n, inp) != 0) {
182 if (mac_inpcb_check_deliver(inp, n) != 0) {
188 if (inp->inp_flags & INP_CONTROLOPTS ||
189 inp->inp_socket->so_options & SO_TIMESTAMP)
190 ip6_savecontrol(inp, n, &opts);
191 if ((inp->inp_vflag & INP_IPV6) && (inp->inp_flags2 & INP_ORIGDSTADDR)) {
192 tmp_opts = sbcreatecontrol(&fromsa[1],
193 sizeof(struct sockaddr_in6), IPV6_ORIGDSTADDR,
194 IPPROTO_IPV6, M_NOWAIT);
197 tmp_opts->m_next = opts;
203 m_adj(n, off + sizeof(struct udphdr));
205 so = inp->inp_socket;
206 SOCKBUF_LOCK(&so->so_rcv);
207 if (sbappendaddr_locked(&so->so_rcv, (struct sockaddr *)&fromsa[0], n,
209 soroverflow_locked(so);
213 UDPSTAT_INC(udps_fullsock);
215 sorwakeup_locked(so);
219 struct udp6_multi_match_ctx {
225 udp6_multi_match(const struct inpcb *inp, void *v)
227 struct udp6_multi_match_ctx *ctx = v;
229 if ((inp->inp_vflag & INP_IPV6) == 0)
231 if (inp->inp_lport != ctx->uh->uh_dport)
233 if (inp->inp_fport != 0 && inp->inp_fport != ctx->uh->uh_sport)
235 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
236 !IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, &ctx->ip6->ip6_dst))
238 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) &&
239 (!IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &ctx->ip6->ip6_src) ||
240 inp->inp_fport != ctx->uh->uh_sport))
247 udp6_multi_input(struct mbuf *m, int off, int proto,
248 struct sockaddr_in6 *fromsa)
250 struct udp6_multi_match_ctx ctx;
251 struct inpcb_iterator inpi = INP_ITERATOR(udp_get_inpcbinfo(proto),
252 INPLOOKUP_RLOCKPCB, udp6_multi_match, &ctx);
254 struct ip6_moptions *imo;
259 * In the event that laddr should be set to the link-local
260 * address (this happens in RIPng), the multicast address
261 * specified in the received packet will not match laddr. To
262 * handle this situation, matching is relaxed if the
263 * receiving interface is the same as one specified in the
264 * socket and if the destination multicast address matches
265 * one of the multicast groups specified in the socket.
269 * KAME note: traditionally we dropped udpiphdr from mbuf
270 * here. We need udphdr for IPsec processing so we do that
273 ctx.ip6 = mtod(m, struct ip6_hdr *);
274 ctx.uh = (struct udphdr *)((char *)ctx.ip6 + off);
275 while ((inp = inp_next(&inpi)) != NULL) {
276 INP_RLOCK_ASSERT(inp);
278 * XXXRW: Because we weren't holding either the inpcb
279 * or the hash lock when we checked for a match
280 * before, we should probably recheck now that the
281 * inpcb lock is (supposed to be) held.
284 * Handle socket delivery policy for any-source
285 * and source-specific multicast. [RFC3678]
287 if ((imo = inp->in6p_moptions) != NULL) {
288 struct sockaddr_in6 mcaddr;
291 bzero(&mcaddr, sizeof(struct sockaddr_in6));
292 mcaddr.sin6_len = sizeof(struct sockaddr_in6);
293 mcaddr.sin6_family = AF_INET6;
294 mcaddr.sin6_addr = ctx.ip6->ip6_dst;
296 blocked = im6o_mc_filter(imo, m->m_pkthdr.rcvif,
297 (struct sockaddr *)&mcaddr,
298 (struct sockaddr *)&fromsa[0]);
299 if (blocked != MCAST_PASS) {
300 if (blocked == MCAST_NOTGMEMBER)
301 IP6STAT_INC(ip6s_notmember);
302 if (blocked == MCAST_NOTSMEMBER ||
303 blocked == MCAST_MUTED)
304 UDPSTAT_INC(udps_filtermcast);
308 if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) != NULL) {
309 if (proto == IPPROTO_UDPLITE)
310 UDPLITE_PROBE(receive, NULL, inp, ctx.ip6,
313 UDP_PROBE(receive, NULL, inp, ctx.ip6, inp,
315 if (udp6_append(inp, n, off, fromsa)) {
322 * Don't look for additional matches if this one does
323 * not have either the SO_REUSEPORT or SO_REUSEADDR
324 * socket options set. This heuristic avoids
325 * searching through all pcbs in the common case of a
326 * non-shared port. It assumes that an application
327 * will never clear these options after setting them.
329 if ((inp->inp_socket->so_options &
330 (SO_REUSEPORT|SO_REUSEPORT_LB|SO_REUSEADDR)) == 0) {
339 * No matching pcb found; discard datagram. (No need
340 * to send an ICMP Port Unreachable for a broadcast
341 * or multicast datgram.)
343 UDPSTAT_INC(udps_noport);
344 UDPSTAT_INC(udps_noportmcast);
347 return (IPPROTO_DONE);
351 udp6_input(struct mbuf **mp, int *offp, int proto)
353 struct mbuf *m = *mp;
357 struct inpcbinfo *pcbinfo;
362 struct sockaddr_in6 fromsa[2];
363 struct m_tag *fwd_tag;
369 if (m->m_len < off + sizeof(struct udphdr)) {
370 m = m_pullup(m, off + sizeof(struct udphdr));
372 IP6STAT_INC(ip6s_exthdrtoolong);
374 return (IPPROTO_DONE);
377 ip6 = mtod(m, struct ip6_hdr *);
378 uh = (struct udphdr *)((caddr_t)ip6 + off);
380 UDPSTAT_INC(udps_ipackets);
383 * Destination port of 0 is illegal, based on RFC768.
385 if (uh->uh_dport == 0)
388 plen = ntohs(ip6->ip6_plen) - off + sizeof(*ip6);
389 ulen = ntohs((u_short)uh->uh_ulen);
392 cscov_partial = (nxt == IPPROTO_UDPLITE) ? 1 : 0;
393 if (nxt == IPPROTO_UDPLITE) {
394 /* Zero means checksum over the complete packet. */
399 if ((ulen < sizeof(struct udphdr)) || (ulen > plen)) {
400 /* XXX: What is the right UDPLite MIB counter? */
403 if (uh->uh_sum == 0) {
404 /* XXX: What is the right UDPLite MIB counter? */
408 if ((ulen < sizeof(struct udphdr)) || (plen != ulen)) {
409 UDPSTAT_INC(udps_badlen);
412 if (uh->uh_sum == 0) {
413 UDPSTAT_INC(udps_nosum);
415 * dport 0 was rejected earlier so this is OK even if
416 * zero_checksum_port is 0 (which is its default value).
418 if (ntohs(uh->uh_dport) == V_zero_checksum_port)
425 if ((m->m_pkthdr.csum_flags & CSUM_DATA_VALID_IPV6) &&
427 if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR)
428 uh_sum = m->m_pkthdr.csum_data;
430 uh_sum = in6_cksum_pseudo(ip6, ulen, nxt,
431 m->m_pkthdr.csum_data);
434 uh_sum = in6_cksum_partial(m, nxt, off, plen, ulen);
437 UDPSTAT_INC(udps_badsum);
443 * Construct sockaddr format source address.
445 init_sin6(&fromsa[0], m, 0);
446 fromsa[0].sin6_port = uh->uh_sport;
447 init_sin6(&fromsa[1], m, 1);
448 fromsa[1].sin6_port = uh->uh_dport;
450 pcbinfo = udp_get_inpcbinfo(nxt);
451 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
453 return (udp6_multi_input(m, off, proto, fromsa));
457 * Locate pcb for datagram.
461 * Grab info from PACKET_TAG_IPFORWARD tag prepended to the chain.
463 if ((m->m_flags & M_IP6_NEXTHOP) &&
464 (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
465 struct sockaddr_in6 *next_hop6;
467 next_hop6 = (struct sockaddr_in6 *)(fwd_tag + 1);
470 * Transparently forwarded. Pretend to be the destination.
471 * Already got one like this?
473 inp = in6_pcblookup_mbuf(pcbinfo, &ip6->ip6_src,
474 uh->uh_sport, &ip6->ip6_dst, uh->uh_dport,
475 INPLOOKUP_RLOCKPCB, m->m_pkthdr.rcvif, m);
478 * It's new. Try to find the ambushing socket.
479 * Because we've rewritten the destination address,
480 * any hardware-generated hash is ignored.
482 inp = in6_pcblookup(pcbinfo, &ip6->ip6_src,
483 uh->uh_sport, &next_hop6->sin6_addr,
484 next_hop6->sin6_port ? htons(next_hop6->sin6_port) :
485 uh->uh_dport, INPLOOKUP_WILDCARD |
486 INPLOOKUP_RLOCKPCB, m->m_pkthdr.rcvif);
488 /* Remove the tag from the packet. We don't need it anymore. */
489 m_tag_delete(m, fwd_tag);
490 m->m_flags &= ~M_IP6_NEXTHOP;
492 inp = in6_pcblookup_mbuf(pcbinfo, &ip6->ip6_src,
493 uh->uh_sport, &ip6->ip6_dst, uh->uh_dport,
494 INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB,
495 m->m_pkthdr.rcvif, m);
497 if (V_udp_log_in_vain) {
498 char ip6bufs[INET6_ADDRSTRLEN];
499 char ip6bufd[INET6_ADDRSTRLEN];
502 "Connection attempt to UDP [%s]:%d from [%s]:%d\n",
503 ip6_sprintf(ip6bufd, &ip6->ip6_dst),
505 ip6_sprintf(ip6bufs, &ip6->ip6_src),
506 ntohs(uh->uh_sport));
508 if (nxt == IPPROTO_UDPLITE)
509 UDPLITE_PROBE(receive, NULL, NULL, ip6, NULL, uh);
511 UDP_PROBE(receive, NULL, NULL, ip6, NULL, uh);
512 UDPSTAT_INC(udps_noport);
513 if (m->m_flags & M_MCAST) {
514 printf("UDP6: M_MCAST is set in a unicast packet.\n");
515 UDPSTAT_INC(udps_noportmcast);
518 if (V_udp_blackhole && (V_udp_blackhole_local ||
519 !in6_localaddr(&ip6->ip6_src)))
521 icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0);
523 return (IPPROTO_DONE);
525 INP_RLOCK_ASSERT(inp);
528 if (up->u_rxcslen == 0 || up->u_rxcslen > ulen) {
532 return (IPPROTO_DONE);
535 if (nxt == IPPROTO_UDPLITE)
536 UDPLITE_PROBE(receive, NULL, inp, ip6, inp, uh);
538 UDP_PROBE(receive, NULL, inp, ip6, inp, uh);
539 if (udp6_append(inp, m, off, fromsa) == 0)
542 return (IPPROTO_DONE);
547 return (IPPROTO_DONE);
551 udp6_common_ctlinput(int cmd, struct sockaddr *sa, void *d,
552 struct inpcbinfo *pcbinfo)
558 struct ip6ctlparam *ip6cp = NULL;
559 const struct sockaddr_in6 *sa6_src = NULL;
561 struct inpcb *(*notify)(struct inpcb *, int) = udp_notify;
562 struct udp_portonly {
567 if (sa->sa_family != AF_INET6 ||
568 sa->sa_len != sizeof(struct sockaddr_in6))
571 if ((unsigned)cmd >= PRC_NCMDS)
573 if (PRC_IS_REDIRECT(cmd))
574 notify = in6_rtchange, d = NULL;
575 else if (cmd == PRC_HOSTDEAD)
577 else if (inet6ctlerrmap[cmd] == 0)
580 /* if the parameter is from icmp6, decode it. */
582 ip6cp = (struct ip6ctlparam *)d;
584 ip6 = ip6cp->ip6c_ip6;
585 off = ip6cp->ip6c_off;
586 cmdarg = ip6cp->ip6c_cmdarg;
587 sa6_src = ip6cp->ip6c_src;
597 * XXX: We assume that when IPV6 is non NULL,
598 * M and OFF are valid.
601 /* Check if we can safely examine src and dst ports. */
602 if (m->m_pkthdr.len < off + sizeof(*uhp))
605 bzero(&uh, sizeof(uh));
606 m_copydata(m, off, sizeof(*uhp), (caddr_t)&uh);
608 if (!PRC_IS_REDIRECT(cmd)) {
609 /* Check to see if its tunneled */
611 inp = in6_pcblookup_mbuf(pcbinfo, &ip6->ip6_dst,
612 uh.uh_dport, &ip6->ip6_src, uh.uh_sport,
613 INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB,
614 m->m_pkthdr.rcvif, m);
619 if (up->u_icmp_func) {
622 (*up->u_icmp_func)(cmd, (struct sockaddr *)ip6cp->ip6c_src,
631 (void)in6_pcbnotify(pcbinfo, sa, uh.uh_dport,
632 (struct sockaddr *)ip6cp->ip6c_src, uh.uh_sport, cmd,
635 (void)in6_pcbnotify(pcbinfo, sa, 0,
636 (const struct sockaddr *)sa6_src, 0, cmd, cmdarg, notify);
640 udp6_ctlinput(int cmd, struct sockaddr *sa, void *d)
643 return (udp6_common_ctlinput(cmd, sa, d, &V_udbinfo));
647 udplite6_ctlinput(int cmd, struct sockaddr *sa, void *d)
650 return (udp6_common_ctlinput(cmd, sa, d, &V_ulitecbinfo));
654 udp6_getcred(SYSCTL_HANDLER_ARGS)
657 struct sockaddr_in6 addrs[2];
658 struct epoch_tracker et;
662 error = priv_check(req->td, PRIV_NETINET_GETCRED);
666 if (req->newlen != sizeof(addrs))
668 if (req->oldlen != sizeof(struct xucred))
670 error = SYSCTL_IN(req, addrs, sizeof(addrs));
673 if ((error = sa6_embedscope(&addrs[0], V_ip6_use_defzone)) != 0 ||
674 (error = sa6_embedscope(&addrs[1], V_ip6_use_defzone)) != 0) {
678 inp = in6_pcblookup(&V_udbinfo, &addrs[1].sin6_addr,
679 addrs[1].sin6_port, &addrs[0].sin6_addr, addrs[0].sin6_port,
680 INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB, NULL);
683 INP_RLOCK_ASSERT(inp);
684 if (inp->inp_socket == NULL)
687 error = cr_canseesocket(req->td->td_ucred,
690 cru2x(inp->inp_cred, &xuc);
695 error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred));
699 SYSCTL_PROC(_net_inet6_udp6, OID_AUTO, getcred,
700 CTLTYPE_OPAQUE | CTLFLAG_RW | CTLFLAG_MPSAFE,
701 0, 0, udp6_getcred, "S,xucred",
702 "Get the xucred of a UDP6 connection");
705 udp6_output(struct socket *so, int flags_arg, struct mbuf *m,
706 struct sockaddr *addr6, struct mbuf *control, struct thread *td)
711 struct in6_addr *laddr, *faddr, in6a;
712 struct ip6_pktopts *optp, opt;
713 struct sockaddr_in6 *sin6, tmp;
714 struct epoch_tracker et;
715 int cscov_partial, error, flags, hlen, scope_ambiguous;
716 u_int32_t ulen, plen;
721 /* addr6 has been validated in udp6_send(). */
722 sin6 = (struct sockaddr_in6 *)addr6;
725 * In contrast to to IPv4 we do not validate the max. packet length
726 * here due to IPv6 Jumbograms (RFC2675).
731 /* Protect *addr6 from overwrites. */
736 * Application should provide a proper zone ID or the use of
737 * default zone IDs should be enabled. Unfortunately, some
738 * applications do not behave as it should, so we need a
739 * workaround. Even if an appropriate ID is not determined,
740 * we'll see if we can determine the outgoing interface. If we
741 * can, determine the zone ID based on the interface below.
743 if (sin6->sin6_scope_id == 0 && !V_ip6_use_defzone)
745 if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0) {
754 KASSERT(inp != NULL, ("%s: inp == NULL", __func__));
756 * In the following cases we want a write lock on the inp for either
757 * local operations or for possible route cache updates in the IPv6
759 * - on connected sockets (sin6 is NULL) for route cache updates,
760 * - when we are not bound to an address and source port (it is
761 * in6_pcbsetport() which will require the write lock).
763 * We check the inp fields before actually locking the inp, so
764 * here exists a race, and we may WLOCK the inp and end with already
765 * bound one by other thread. This is fine.
767 if (sin6 == NULL || (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
768 inp->inp_lport == 0))
773 nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ?
774 IPPROTO_UDP : IPPROTO_UDPLITE;
777 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
781 hasv4addr = (inp->inp_vflag & INP_IPV4);
783 hasv4addr = IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)
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 /* addr will just be freed in sendit(). */
798 return (udp_send(so, flags_arg | PRUS_IPV6, m,
799 (struct sockaddr *)sin6, control, td));
803 if (sin6 && IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
805 * Given this is either an IPv6-only socket or no INET is
806 * supported we will fail the send if the given destination
807 * address is a v4mapped address.
809 * XXXGL: do we leak m and control?
817 if ((error = ip6_setpktopts(control, &opt,
818 inp->in6p_outputopts, td->td_ucred, nxt)) != 0) {
823 optp = inp->in6p_outputopts;
827 * Since we saw no essential reason for calling in_pcbconnect,
828 * we get rid of such kind of logic, and call in6_selectsrc
829 * and in6_pcbsetport in order to fill in the local address
830 * and the local port.
832 if (sin6->sin6_port == 0) {
833 error = EADDRNOTAVAIL;
837 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
838 /* how about ::ffff:0.0.0.0 case? */
844 * Given we handle the v4mapped case in the INET block above
845 * assert here that it must not happen anymore.
847 KASSERT(!IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr),
848 ("%s: sin6(%p)->sin6_addr is v4mapped which we "
849 "should have handled.", __func__, sin6));
851 /* This only requires read-locking. */
852 error = in6_selectsrc_socket(sin6, optp, inp,
853 td->td_ucred, scope_ambiguous, &in6a, NULL);
858 if (inp->inp_lport == 0) {
859 struct inpcbinfo *pcbinfo;
861 INP_WLOCK_ASSERT(inp);
863 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
864 INP_HASH_WLOCK(pcbinfo);
865 error = in6_pcbsetport(laddr, inp, td->td_ucred);
866 INP_HASH_WUNLOCK(pcbinfo);
868 /* Undo an address bind that may have occurred. */
869 inp->in6p_laddr = in6addr_any;
873 faddr = &sin6->sin6_addr;
874 fport = sin6->sin6_port; /* allow 0 port */
877 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
881 laddr = &inp->in6p_laddr;
882 faddr = &inp->in6p_faddr;
883 fport = inp->inp_fport;
886 ulen = m->m_pkthdr.len;
887 plen = sizeof(struct udphdr) + ulen;
888 hlen = sizeof(struct ip6_hdr);
891 * Calculate data length and get a mbuf
892 * for UDP and IP6 headers.
894 M_PREPEND(m, hlen + sizeof(struct udphdr), M_NOWAIT);
901 * Stuff checksum and output datagram.
903 cscov = cscov_partial = 0;
904 udp6 = (struct udphdr *)(mtod(m, caddr_t) + hlen);
905 udp6->uh_sport = inp->inp_lport; /* lport is always set in the PCB */
906 udp6->uh_dport = fport;
907 if (nxt == IPPROTO_UDPLITE) {
911 cscov = up->u_txcslen;
914 udp6->uh_ulen = htons(cscov);
916 * For UDP-Lite, checksum coverage length of zero means
917 * the entire UDPLite packet is covered by the checksum.
919 cscov_partial = (cscov == 0) ? 0 : 1;
920 } else if (plen <= 0xffff)
921 udp6->uh_ulen = htons((u_short)plen);
926 ip6 = mtod(m, struct ip6_hdr *);
927 ip6->ip6_flow = inp->inp_flow & IPV6_FLOWINFO_MASK;
928 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
929 ip6->ip6_vfc |= IPV6_VERSION;
930 ip6->ip6_plen = htons((u_short)plen);
932 ip6->ip6_hlim = in6_selecthlim(inp, NULL);
933 ip6->ip6_src = *laddr;
934 ip6->ip6_dst = *faddr;
937 mac_inpcb_create_mbuf(inp, m);
941 if ((udp6->uh_sum = in6_cksum_partial(m, nxt,
942 sizeof(struct ip6_hdr), plen, cscov)) == 0)
943 udp6->uh_sum = 0xffff;
945 udp6->uh_sum = in6_cksum_pseudo(ip6, plen, nxt, 0);
946 m->m_pkthdr.csum_flags = CSUM_UDP_IPV6;
947 m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
951 #if defined(ROUTE_MPATH) || defined(RSS)
952 if (CALC_FLOWID_OUTBOUND_SENDTO) {
953 uint32_t hash_type, hash_val;
956 pr = inp->inp_socket->so_proto->pr_protocol;
958 hash_val = fib6_calc_packet_hash(laddr, faddr,
959 inp->inp_lport, fport, pr, &hash_type);
960 m->m_pkthdr.flowid = hash_val;
961 M_HASHTYPE_SET(m, hash_type);
963 /* do not use inp flowid */
964 flags |= IP_NODEFAULTFLOWID;
967 UDPSTAT_INC(udps_opackets);
968 if (nxt == IPPROTO_UDPLITE)
969 UDPLITE_PROBE(send, NULL, inp, ip6, inp, udp6);
971 UDP_PROBE(send, NULL, inp, ip6, inp, udp6);
972 error = ip6_output(m, optp,
973 INP_WLOCKED(inp) ? &inp->inp_route6 : NULL, flags,
974 inp->in6p_moptions, NULL, inp);
979 ip6_clearpktopts(&opt, -1);
988 ip6_clearpktopts(&opt, -1);
997 udp6_abort(struct socket *so)
1000 struct inpcbinfo *pcbinfo;
1002 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1003 inp = sotoinpcb(so);
1004 KASSERT(inp != NULL, ("udp6_abort: inp == NULL"));
1008 if (inp->inp_vflag & INP_IPV4) {
1015 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1016 INP_HASH_WLOCK(pcbinfo);
1017 in6_pcbdisconnect(inp);
1018 inp->in6p_laddr = in6addr_any;
1019 INP_HASH_WUNLOCK(pcbinfo);
1020 soisdisconnected(so);
1026 udp6_attach(struct socket *so, int proto, struct thread *td)
1029 struct inpcbinfo *pcbinfo;
1032 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1033 inp = sotoinpcb(so);
1034 KASSERT(inp == NULL, ("udp6_attach: inp != NULL"));
1036 if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
1037 error = soreserve(so, udp_sendspace, udp_recvspace);
1041 error = in_pcballoc(so, pcbinfo);
1044 inp = (struct inpcb *)so->so_pcb;
1045 inp->in6p_cksum = -1; /* just to be sure */
1048 * IPv4 TTL initialization is necessary for an IPv6 socket as well,
1049 * because the socket may be bound to an IPv6 wildcard address,
1050 * which may match an IPv4-mapped IPv6 address.
1052 inp->inp_ip_ttl = V_ip_defttl;
1054 error = udp_newudpcb(inp);
1065 udp6_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
1068 struct inpcbinfo *pcbinfo;
1072 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1073 inp = sotoinpcb(so);
1074 KASSERT(inp != NULL, ("udp6_bind: inp == NULL"));
1076 if (nam->sa_family != AF_INET6)
1077 return (EAFNOSUPPORT);
1078 if (nam->sa_len != sizeof(struct sockaddr_in6))
1082 INP_HASH_WLOCK(pcbinfo);
1083 vflagsav = inp->inp_vflag;
1084 inp->inp_vflag &= ~INP_IPV4;
1085 inp->inp_vflag |= INP_IPV6;
1086 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
1087 struct sockaddr_in6 *sin6_p;
1089 sin6_p = (struct sockaddr_in6 *)nam;
1091 if (IN6_IS_ADDR_UNSPECIFIED(&sin6_p->sin6_addr))
1092 inp->inp_vflag |= INP_IPV4;
1094 else if (IN6_IS_ADDR_V4MAPPED(&sin6_p->sin6_addr)) {
1095 struct sockaddr_in sin;
1097 in6_sin6_2_sin(&sin, sin6_p);
1098 inp->inp_vflag |= INP_IPV4;
1099 inp->inp_vflag &= ~INP_IPV6;
1100 error = in_pcbbind(inp, (struct sockaddr *)&sin,
1107 error = in6_pcbbind(inp, nam, td->td_ucred);
1112 inp->inp_vflag = vflagsav;
1113 INP_HASH_WUNLOCK(pcbinfo);
1119 udp6_close(struct socket *so)
1122 struct inpcbinfo *pcbinfo;
1124 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1125 inp = sotoinpcb(so);
1126 KASSERT(inp != NULL, ("udp6_close: inp == NULL"));
1130 if (inp->inp_vflag & INP_IPV4) {
1132 (void)udp_disconnect(so);
1136 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1137 INP_HASH_WLOCK(pcbinfo);
1138 in6_pcbdisconnect(inp);
1139 inp->in6p_laddr = in6addr_any;
1140 INP_HASH_WUNLOCK(pcbinfo);
1141 soisdisconnected(so);
1147 udp6_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
1150 struct epoch_tracker et;
1153 struct inpcbinfo *pcbinfo;
1154 struct sockaddr_in6 *sin6;
1158 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1159 inp = sotoinpcb(so);
1160 KASSERT(inp != NULL, ("udp6_connect: inp == NULL"));
1162 sin6 = (struct sockaddr_in6 *)nam;
1163 if (sin6->sin6_family != AF_INET6)
1164 return (EAFNOSUPPORT);
1165 if (sin6->sin6_len != sizeof(*sin6))
1169 * XXXRW: Need to clarify locking of v4/v6 flags.
1173 if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
1174 struct sockaddr_in sin;
1176 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) {
1180 if ((inp->inp_vflag & INP_IPV4) == 0) {
1181 error = EAFNOSUPPORT;
1184 if (inp->inp_faddr.s_addr != INADDR_ANY) {
1188 in6_sin6_2_sin(&sin, sin6);
1189 error = prison_remote_ip4(td->td_ucred, &sin.sin_addr);
1192 vflagsav = inp->inp_vflag;
1193 inp->inp_vflag |= INP_IPV4;
1194 inp->inp_vflag &= ~INP_IPV6;
1195 NET_EPOCH_ENTER(et);
1196 INP_HASH_WLOCK(pcbinfo);
1197 error = in_pcbconnect(inp, (struct sockaddr *)&sin,
1198 td->td_ucred, true);
1199 INP_HASH_WUNLOCK(pcbinfo);
1202 * If connect succeeds, mark socket as connected. If
1203 * connect fails and socket is unbound, reset inp_vflag
1208 else if (inp->inp_laddr.s_addr == INADDR_ANY &&
1209 inp->inp_lport == 0)
1210 inp->inp_vflag = vflagsav;
1213 if ((inp->inp_vflag & INP_IPV6) == 0) {
1214 error = EAFNOSUPPORT;
1219 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1223 error = prison_remote_ip6(td->td_ucred, &sin6->sin6_addr);
1226 vflagsav = inp->inp_vflag;
1227 inp->inp_vflag &= ~INP_IPV4;
1228 inp->inp_vflag |= INP_IPV6;
1229 INP_HASH_WLOCK(pcbinfo);
1230 error = in6_pcbconnect(inp, nam, td->td_ucred);
1231 INP_HASH_WUNLOCK(pcbinfo);
1233 * If connect succeeds, mark socket as connected. If
1234 * connect fails and socket is unbound, reset inp_vflag
1239 else if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
1240 inp->inp_lport == 0)
1241 inp->inp_vflag = vflagsav;
1248 udp6_detach(struct socket *so)
1253 inp = sotoinpcb(so);
1254 KASSERT(inp != NULL, ("udp6_detach: inp == NULL"));
1257 up = intoudpcb(inp);
1258 KASSERT(up != NULL, ("%s: up == NULL", __func__));
1265 udp6_disconnect(struct socket *so)
1268 struct inpcbinfo *pcbinfo;
1270 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1271 inp = sotoinpcb(so);
1272 KASSERT(inp != NULL, ("udp6_disconnect: inp == NULL"));
1276 if (inp->inp_vflag & INP_IPV4) {
1278 (void)udp_disconnect(so);
1283 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1288 INP_HASH_WLOCK(pcbinfo);
1289 in6_pcbdisconnect(inp);
1290 inp->in6p_laddr = in6addr_any;
1291 INP_HASH_WUNLOCK(pcbinfo);
1293 so->so_state &= ~SS_ISCONNECTED; /* XXX */
1300 udp6_send(struct socket *so, int flags, struct mbuf *m,
1301 struct sockaddr *addr, struct mbuf *control, struct thread *td)
1306 if (addr->sa_len != sizeof(struct sockaddr_in6)) {
1310 if (addr->sa_family != AF_INET6) {
1311 error = EAFNOSUPPORT;
1316 return (udp6_output(so, flags, m, addr, control, td));
1325 #define UDP6_PROTOSW \
1326 .pr_type = SOCK_DGRAM, \
1327 .pr_flags = PR_ATOMIC|PR_ADDR|PR_CAPATTACH, \
1328 .pr_ctloutput = ip6_ctloutput, \
1329 .pr_abort = udp6_abort, \
1330 .pr_attach = udp6_attach, \
1331 .pr_bind = udp6_bind, \
1332 .pr_connect = udp6_connect, \
1333 .pr_control = in6_control, \
1334 .pr_detach = udp6_detach, \
1335 .pr_disconnect = udp6_disconnect, \
1336 .pr_peeraddr = in6_mapped_peeraddr, \
1337 .pr_send = udp6_send, \
1338 .pr_shutdown = udp_shutdown, \
1339 .pr_sockaddr = in6_mapped_sockaddr, \
1340 .pr_soreceive = soreceive_dgram, \
1341 .pr_sosend = sosend_dgram, \
1342 .pr_sosetlabel = in_pcbsosetlabel, \
1343 .pr_close = udp6_close
1345 struct protosw udp6_protosw = {
1346 .pr_protocol = IPPROTO_UDP,
1350 struct protosw udplite6_protosw = {
1351 .pr_protocol = IPPROTO_UDPLITE,
1356 udp6_init(void *arg __unused)
1359 IP6PROTO_REGISTER(IPPROTO_UDP, udp6_input, udp6_ctlinput);
1360 IP6PROTO_REGISTER(IPPROTO_UDPLITE, udp6_input, udplite6_ctlinput);
1362 SYSINIT(udp6_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, udp6_init, NULL);