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 * 4. 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/ip_icmp.h>
108 #include <netinet/ip6.h>
109 #include <netinet/icmp_var.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>
124 #include <netipsec/ipsec.h>
125 #include <netipsec/ipsec6.h>
128 #include <security/mac/mac_framework.h>
131 * UDP protocol implementation.
132 * Per RFC 768, August, 1980.
135 extern struct protosw inetsw[];
136 static void udp6_detach(struct socket *so);
139 udp6_append(struct inpcb *inp, struct mbuf *n, int off,
140 struct sockaddr_in6 *fromsa)
146 INP_LOCK_ASSERT(inp);
149 * Engage the tunneling protocol.
152 if (up->u_tun_func != NULL) {
155 (*up->u_tun_func)(n, off, inp, (struct sockaddr *)fromsa,
158 return (in_pcbrele_rlocked(inp));
161 /* Check AH/ESP integrity. */
162 if (ipsec6_in_reject(n, inp)) {
168 if (mac_inpcb_check_deliver(inp, n) != 0) {
174 if (inp->inp_flags & INP_CONTROLOPTS ||
175 inp->inp_socket->so_options & SO_TIMESTAMP)
176 ip6_savecontrol(inp, n, &opts);
177 m_adj(n, off + sizeof(struct udphdr));
179 so = inp->inp_socket;
180 SOCKBUF_LOCK(&so->so_rcv);
181 if (sbappendaddr_locked(&so->so_rcv, (struct sockaddr *)fromsa, n,
183 SOCKBUF_UNLOCK(&so->so_rcv);
187 UDPSTAT_INC(udps_fullsock);
189 sorwakeup_locked(so);
194 udp6_input(struct mbuf **mp, int *offp, int proto)
196 struct mbuf *m = *mp;
201 struct inpcbinfo *pcbinfo;
206 struct sockaddr_in6 fromsa;
207 struct m_tag *fwd_tag;
211 ifp = m->m_pkthdr.rcvif;
212 ip6 = mtod(m, struct ip6_hdr *);
214 #ifndef PULLDOWN_TEST
215 IP6_EXTHDR_CHECK(m, off, sizeof(struct udphdr), IPPROTO_DONE);
216 ip6 = mtod(m, struct ip6_hdr *);
217 uh = (struct udphdr *)((caddr_t)ip6 + off);
219 IP6_EXTHDR_GET(uh, struct udphdr *, m, off, sizeof(*uh));
221 return (IPPROTO_DONE);
224 UDPSTAT_INC(udps_ipackets);
227 * Destination port of 0 is illegal, based on RFC768.
229 if (uh->uh_dport == 0)
232 plen = ntohs(ip6->ip6_plen) - off + sizeof(*ip6);
233 ulen = ntohs((u_short)uh->uh_ulen);
236 cscov_partial = (nxt == IPPROTO_UDPLITE) ? 1 : 0;
237 if (nxt == IPPROTO_UDPLITE) {
238 /* Zero means checksum over the complete packet. */
243 if ((ulen < sizeof(struct udphdr)) || (ulen > plen)) {
244 /* XXX: What is the right UDPLite MIB counter? */
247 if (uh->uh_sum == 0) {
248 /* XXX: What is the right UDPLite MIB counter? */
252 if ((ulen < sizeof(struct udphdr)) || (plen != ulen)) {
253 UDPSTAT_INC(udps_badlen);
256 if (uh->uh_sum == 0) {
257 UDPSTAT_INC(udps_nosum);
262 if ((m->m_pkthdr.csum_flags & CSUM_DATA_VALID_IPV6) &&
264 if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR)
265 uh_sum = m->m_pkthdr.csum_data;
267 uh_sum = in6_cksum_pseudo(ip6, ulen, nxt,
268 m->m_pkthdr.csum_data);
271 uh_sum = in6_cksum_partial(m, nxt, off, plen, ulen);
274 UDPSTAT_INC(udps_badsum);
279 * Construct sockaddr format source address.
281 init_sin6(&fromsa, m);
282 fromsa.sin6_port = uh->uh_sport;
284 pcbinfo = udp_get_inpcbinfo(nxt);
285 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
287 struct inpcbhead *pcblist;
288 struct ip6_moptions *imo;
290 INP_INFO_RLOCK(pcbinfo);
292 * In the event that laddr should be set to the link-local
293 * address (this happens in RIPng), the multicast address
294 * specified in the received packet will not match laddr. To
295 * handle this situation, matching is relaxed if the
296 * receiving interface is the same as one specified in the
297 * socket and if the destination multicast address matches
298 * one of the multicast groups specified in the socket.
302 * KAME note: traditionally we dropped udpiphdr from mbuf
303 * here. We need udphdr for IPsec processing so we do that
306 pcblist = udp_get_pcblist(nxt);
308 LIST_FOREACH(inp, pcblist, inp_list) {
309 if ((inp->inp_vflag & INP_IPV6) == 0)
311 if (inp->inp_lport != uh->uh_dport)
313 if (inp->inp_fport != 0 &&
314 inp->inp_fport != uh->uh_sport)
316 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
317 if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr,
321 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
322 if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr,
324 inp->inp_fport != uh->uh_sport)
329 * XXXRW: Because we weren't holding either the inpcb
330 * or the hash lock when we checked for a match
331 * before, we should probably recheck now that the
332 * inpcb lock is (supposed to be) held.
336 * Handle socket delivery policy for any-source
337 * and source-specific multicast. [RFC3678]
339 imo = inp->in6p_moptions;
340 if (imo && IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
341 struct sockaddr_in6 mcaddr;
346 bzero(&mcaddr, sizeof(struct sockaddr_in6));
347 mcaddr.sin6_len = sizeof(struct sockaddr_in6);
348 mcaddr.sin6_family = AF_INET6;
349 mcaddr.sin6_addr = ip6->ip6_dst;
351 blocked = im6o_mc_filter(imo, ifp,
352 (struct sockaddr *)&mcaddr,
353 (struct sockaddr *)&fromsa);
354 if (blocked != MCAST_PASS) {
355 if (blocked == MCAST_NOTGMEMBER)
356 IP6STAT_INC(ip6s_notmember);
357 if (blocked == MCAST_NOTSMEMBER ||
358 blocked == MCAST_MUTED)
359 UDPSTAT_INC(udps_filtermcast);
360 INP_RUNLOCK(inp); /* XXX */
369 if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) !=
372 UDP_PROBE(receive, NULL, last, ip6,
374 if (udp6_append(last, n, off, &fromsa))
381 * Don't look for additional matches if this one does
382 * not have either the SO_REUSEPORT or SO_REUSEADDR
383 * socket options set. This heuristic avoids
384 * searching through all pcbs in the common case of a
385 * non-shared port. It assumes that an application
386 * will never clear these options after setting them.
388 if ((last->inp_socket->so_options &
389 (SO_REUSEPORT|SO_REUSEADDR)) == 0)
395 * No matching pcb found; discard datagram. (No need
396 * to send an ICMP Port Unreachable for a broadcast
397 * or multicast datgram.)
399 UDPSTAT_INC(udps_noport);
400 UDPSTAT_INC(udps_noportmcast);
404 INP_INFO_RUNLOCK(pcbinfo);
405 UDP_PROBE(receive, NULL, last, ip6, last, uh);
406 if (udp6_append(last, m, off, &fromsa) == 0)
409 return (IPPROTO_DONE);
412 * Locate pcb for datagram.
416 * Grab info from PACKET_TAG_IPFORWARD tag prepended to the chain.
418 if ((m->m_flags & M_IP6_NEXTHOP) &&
419 (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
420 struct sockaddr_in6 *next_hop6;
422 next_hop6 = (struct sockaddr_in6 *)(fwd_tag + 1);
425 * Transparently forwarded. Pretend to be the destination.
426 * Already got one like this?
428 inp = in6_pcblookup_mbuf(pcbinfo, &ip6->ip6_src,
429 uh->uh_sport, &ip6->ip6_dst, uh->uh_dport,
430 INPLOOKUP_RLOCKPCB, m->m_pkthdr.rcvif, m);
433 * It's new. Try to find the ambushing socket.
434 * Because we've rewritten the destination address,
435 * any hardware-generated hash is ignored.
437 inp = in6_pcblookup(pcbinfo, &ip6->ip6_src,
438 uh->uh_sport, &next_hop6->sin6_addr,
439 next_hop6->sin6_port ? htons(next_hop6->sin6_port) :
440 uh->uh_dport, INPLOOKUP_WILDCARD |
441 INPLOOKUP_RLOCKPCB, m->m_pkthdr.rcvif);
443 /* Remove the tag from the packet. We don't need it anymore. */
444 m_tag_delete(m, fwd_tag);
445 m->m_flags &= ~M_IP6_NEXTHOP;
447 inp = in6_pcblookup_mbuf(pcbinfo, &ip6->ip6_src,
448 uh->uh_sport, &ip6->ip6_dst, uh->uh_dport,
449 INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB,
450 m->m_pkthdr.rcvif, m);
452 if (udp_log_in_vain) {
453 char ip6bufs[INET6_ADDRSTRLEN];
454 char ip6bufd[INET6_ADDRSTRLEN];
457 "Connection attempt to UDP [%s]:%d from [%s]:%d\n",
458 ip6_sprintf(ip6bufd, &ip6->ip6_dst),
460 ip6_sprintf(ip6bufs, &ip6->ip6_src),
461 ntohs(uh->uh_sport));
463 UDPSTAT_INC(udps_noport);
464 if (m->m_flags & M_MCAST) {
465 printf("UDP6: M_MCAST is set in a unicast packet.\n");
466 UDPSTAT_INC(udps_noportmcast);
471 if (badport_bandlim(BANDLIM_ICMP6_UNREACH) < 0)
473 icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0);
474 return (IPPROTO_DONE);
476 INP_RLOCK_ASSERT(inp);
479 if (up->u_rxcslen == 0 || up->u_rxcslen > ulen) {
482 return (IPPROTO_DONE);
485 UDP_PROBE(receive, NULL, inp, ip6, inp, uh);
486 if (udp6_append(inp, m, off, &fromsa) == 0)
488 return (IPPROTO_DONE);
491 INP_INFO_RUNLOCK(pcbinfo);
495 return (IPPROTO_DONE);
499 udp6_common_ctlinput(int cmd, struct sockaddr *sa, void *d,
500 struct inpcbinfo *pcbinfo)
506 struct ip6ctlparam *ip6cp = NULL;
507 const struct sockaddr_in6 *sa6_src = NULL;
509 struct inpcb *(*notify)(struct inpcb *, int) = udp_notify;
510 struct udp_portonly {
515 if (sa->sa_family != AF_INET6 ||
516 sa->sa_len != sizeof(struct sockaddr_in6))
519 if ((unsigned)cmd >= PRC_NCMDS)
521 if (PRC_IS_REDIRECT(cmd))
522 notify = in6_rtchange, d = NULL;
523 else if (cmd == PRC_HOSTDEAD)
525 else if (inet6ctlerrmap[cmd] == 0)
528 /* if the parameter is from icmp6, decode it. */
530 ip6cp = (struct ip6ctlparam *)d;
532 ip6 = ip6cp->ip6c_ip6;
533 off = ip6cp->ip6c_off;
534 cmdarg = ip6cp->ip6c_cmdarg;
535 sa6_src = ip6cp->ip6c_src;
545 * XXX: We assume that when IPV6 is non NULL,
546 * M and OFF are valid.
549 /* Check if we can safely examine src and dst ports. */
550 if (m->m_pkthdr.len < off + sizeof(*uhp))
553 bzero(&uh, sizeof(uh));
554 m_copydata(m, off, sizeof(*uhp), (caddr_t)&uh);
556 if (!PRC_IS_REDIRECT(cmd)) {
557 /* Check to see if its tunneled */
559 inp = in6_pcblookup_mbuf(pcbinfo, &ip6->ip6_dst,
560 uh.uh_dport, &ip6->ip6_src, uh.uh_sport,
561 INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB,
562 m->m_pkthdr.rcvif, m);
567 if (up->u_icmp_func) {
570 (*up->u_icmp_func)(cmd, (struct sockaddr *)ip6cp->ip6c_src,
579 (void)in6_pcbnotify(pcbinfo, sa, uh.uh_dport,
580 (struct sockaddr *)ip6cp->ip6c_src, uh.uh_sport, cmd,
583 (void)in6_pcbnotify(pcbinfo, sa, 0,
584 (const struct sockaddr *)sa6_src, 0, cmd, cmdarg, notify);
588 udp6_ctlinput(int cmd, struct sockaddr *sa, void *d)
591 return (udp6_common_ctlinput(cmd, sa, d, &V_udbinfo));
595 udplite6_ctlinput(int cmd, struct sockaddr *sa, void *d)
598 return (udp6_common_ctlinput(cmd, sa, d, &V_ulitecbinfo));
602 udp6_getcred(SYSCTL_HANDLER_ARGS)
605 struct sockaddr_in6 addrs[2];
609 error = priv_check(req->td, PRIV_NETINET_GETCRED);
613 if (req->newlen != sizeof(addrs))
615 if (req->oldlen != sizeof(struct xucred))
617 error = SYSCTL_IN(req, addrs, sizeof(addrs));
620 if ((error = sa6_embedscope(&addrs[0], V_ip6_use_defzone)) != 0 ||
621 (error = sa6_embedscope(&addrs[1], V_ip6_use_defzone)) != 0) {
624 inp = in6_pcblookup(&V_udbinfo, &addrs[1].sin6_addr,
625 addrs[1].sin6_port, &addrs[0].sin6_addr, addrs[0].sin6_port,
626 INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB, NULL);
628 INP_RLOCK_ASSERT(inp);
629 if (inp->inp_socket == NULL)
632 error = cr_canseesocket(req->td->td_ucred,
635 cru2x(inp->inp_cred, &xuc);
640 error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred));
644 SYSCTL_PROC(_net_inet6_udp6, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW, 0,
645 0, udp6_getcred, "S,xucred", "Get the xucred of a UDP6 connection");
648 udp6_output(struct inpcb *inp, struct mbuf *m, struct sockaddr *addr6,
649 struct mbuf *control, struct thread *td)
651 u_int32_t ulen = m->m_pkthdr.len;
652 u_int32_t plen = sizeof(struct udphdr) + ulen;
655 struct in6_addr *laddr, *faddr, in6a;
656 struct sockaddr_in6 *sin6 = NULL;
657 int cscov_partial = 0;
658 int scope_ambiguous = 0;
663 struct ip6_pktopts *optp, opt;
664 int af = AF_INET6, hlen = sizeof(struct ip6_hdr);
666 struct sockaddr_in6 tmp;
668 INP_WLOCK_ASSERT(inp);
669 INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo);
672 /* addr6 has been validated in udp6_send(). */
673 sin6 = (struct sockaddr_in6 *)addr6;
675 /* protect *sin6 from overwrites */
680 * Application should provide a proper zone ID or the use of
681 * default zone IDs should be enabled. Unfortunately, some
682 * applications do not behave as it should, so we need a
683 * workaround. Even if an appropriate ID is not determined,
684 * we'll see if we can determine the outgoing interface. If we
685 * can, determine the zone ID based on the interface below.
687 if (sin6->sin6_scope_id == 0 && !V_ip6_use_defzone)
689 if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0)
693 nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ?
694 IPPROTO_UDP : IPPROTO_UDPLITE;
696 if ((error = ip6_setpktopts(control, &opt,
697 inp->in6p_outputopts, td->td_ucred, nxt)) != 0)
701 optp = inp->in6p_outputopts;
704 faddr = &sin6->sin6_addr;
707 * Since we saw no essential reason for calling in_pcbconnect,
708 * we get rid of such kind of logic, and call in6_selectsrc
709 * and in6_pcbsetport in order to fill in the local address
710 * and the local port.
712 if (sin6->sin6_port == 0) {
713 error = EADDRNOTAVAIL;
717 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
718 /* how about ::ffff:0.0.0.0 case? */
723 fport = sin6->sin6_port; /* allow 0 port */
725 if (IN6_IS_ADDR_V4MAPPED(faddr)) {
726 if ((inp->inp_flags & IN6P_IPV6_V6ONLY)) {
728 * I believe we should explicitly discard the
729 * packet when mapped addresses are disabled,
730 * rather than send the packet as an IPv6 one.
731 * If we chose the latter approach, the packet
732 * might be sent out on the wire based on the
733 * default route, the situation which we'd
734 * probably want to avoid.
735 * (20010421 jinmei@kame.net)
740 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
741 !IN6_IS_ADDR_V4MAPPED(&inp->in6p_laddr)) {
743 * when remote addr is an IPv4-mapped address,
744 * local addr should not be an IPv6 address,
745 * since you cannot determine how to map IPv6
746 * source address to IPv4.
755 if (!IN6_IS_ADDR_V4MAPPED(faddr)) {
756 error = in6_selectsrc_socket(sin6, optp, inp,
757 td->td_ucred, scope_ambiguous, &in6a, NULL);
762 laddr = &inp->in6p_laddr; /* XXX */
765 error = EADDRNOTAVAIL;
768 if (inp->inp_lport == 0 &&
769 (error = in6_pcbsetport(laddr, inp, td->td_ucred)) != 0) {
770 /* Undo an address bind that may have occurred. */
771 inp->in6p_laddr = in6addr_any;
775 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
779 if (IN6_IS_ADDR_V4MAPPED(&inp->in6p_faddr)) {
780 if ((inp->inp_flags & IN6P_IPV6_V6ONLY)) {
782 * XXX: this case would happen when the
783 * application sets the V6ONLY flag after
784 * connecting the foreign address.
785 * Such applications should be fixed,
788 log(LOG_INFO, "udp6_output: IPV6_V6ONLY "
789 "option was set for a connected socket\n");
795 laddr = &inp->in6p_laddr;
796 faddr = &inp->in6p_faddr;
797 fport = inp->inp_fport;
801 hlen = sizeof(struct ip);
804 * Calculate data length and get a mbuf
805 * for UDP and IP6 headers.
807 M_PREPEND(m, hlen + sizeof(struct udphdr), M_NOWAIT);
814 * Stuff checksum and output datagram.
816 udp6 = (struct udphdr *)(mtod(m, caddr_t) + hlen);
817 udp6->uh_sport = inp->inp_lport; /* lport is always set in the PCB */
818 udp6->uh_dport = fport;
819 if (nxt == IPPROTO_UDPLITE) {
823 cscov = up->u_txcslen;
826 udp6->uh_ulen = htons(cscov);
828 * For UDP-Lite, checksum coverage length of zero means
829 * the entire UDPLite packet is covered by the checksum.
831 cscov_partial = (cscov == 0) ? 0 : 1;
832 } else if (plen <= 0xffff)
833 udp6->uh_ulen = htons((u_short)plen);
840 ip6 = mtod(m, struct ip6_hdr *);
841 ip6->ip6_flow = inp->inp_flow & IPV6_FLOWINFO_MASK;
842 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
843 ip6->ip6_vfc |= IPV6_VERSION;
844 ip6->ip6_plen = htons((u_short)plen);
846 ip6->ip6_hlim = in6_selecthlim(inp, NULL);
847 ip6->ip6_src = *laddr;
848 ip6->ip6_dst = *faddr;
851 if ((udp6->uh_sum = in6_cksum_partial(m, nxt,
852 sizeof(struct ip6_hdr), plen, cscov)) == 0)
853 udp6->uh_sum = 0xffff;
855 udp6->uh_sum = in6_cksum_pseudo(ip6, plen, nxt, 0);
856 m->m_pkthdr.csum_flags = CSUM_UDP_IPV6;
857 m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
862 uint32_t hash_val, hash_type;
865 pr = inp->inp_socket->so_proto->pr_protocol;
867 * Calculate an appropriate RSS hash for UDP and
870 * The called function will take care of figuring out
871 * whether a 2-tuple or 4-tuple hash is required based
872 * on the currently configured scheme.
874 * Later later on connected socket values should be
875 * cached in the inpcb and reused, rather than constantly
878 * UDP Lite is a different protocol number and will
879 * likely end up being hashed as a 2-tuple until
880 * RSS / NICs grow UDP Lite protocol awareness.
882 if (rss_proto_software_hash_v6(faddr, laddr, fport,
883 inp->inp_lport, pr, &hash_val, &hash_type) == 0) {
884 m->m_pkthdr.flowid = hash_val;
885 M_HASHTYPE_SET(m, hash_type);
892 * Don't override with the inp cached flowid.
894 * Until the whole UDP path is vetted, it may actually
897 flags |= IP_NODEFAULTFLOWID;
900 UDP_PROBE(send, NULL, inp, ip6, inp, udp6);
901 UDPSTAT_INC(udps_opackets);
902 error = ip6_output(m, optp, &inp->inp_route6, flags,
903 inp->in6p_moptions, NULL, inp);
906 error = EAFNOSUPPORT;
916 ip6_clearpktopts(&opt, -1);
923 udp6_abort(struct socket *so)
926 struct inpcbinfo *pcbinfo;
928 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
930 KASSERT(inp != NULL, ("udp6_abort: inp == NULL"));
934 if (inp->inp_vflag & INP_IPV4) {
935 struct pr_usrreqs *pru;
938 nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ?
939 IPPROTO_UDP : IPPROTO_UDPLITE;
941 pru = inetsw[ip_protox[nxt]].pr_usrreqs;
942 (*pru->pru_abort)(so);
947 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
948 INP_HASH_WLOCK(pcbinfo);
949 in6_pcbdisconnect(inp);
950 inp->in6p_laddr = in6addr_any;
951 INP_HASH_WUNLOCK(pcbinfo);
952 soisdisconnected(so);
958 udp6_attach(struct socket *so, int proto, struct thread *td)
961 struct inpcbinfo *pcbinfo;
964 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
966 KASSERT(inp == NULL, ("udp6_attach: inp != NULL"));
968 if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
969 error = soreserve(so, udp_sendspace, udp_recvspace);
973 INP_INFO_WLOCK(pcbinfo);
974 error = in_pcballoc(so, pcbinfo);
976 INP_INFO_WUNLOCK(pcbinfo);
979 inp = (struct inpcb *)so->so_pcb;
980 inp->inp_vflag |= INP_IPV6;
981 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0)
982 inp->inp_vflag |= INP_IPV4;
983 inp->in6p_hops = -1; /* use kernel default */
984 inp->in6p_cksum = -1; /* just to be sure */
987 * IPv4 TTL initialization is necessary for an IPv6 socket as well,
988 * because the socket may be bound to an IPv6 wildcard address,
989 * which may match an IPv4-mapped IPv6 address.
991 inp->inp_ip_ttl = V_ip_defttl;
993 error = udp_newudpcb(inp);
997 INP_INFO_WUNLOCK(pcbinfo);
1001 INP_INFO_WUNLOCK(pcbinfo);
1006 udp6_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
1009 struct inpcbinfo *pcbinfo;
1012 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1013 inp = sotoinpcb(so);
1014 KASSERT(inp != NULL, ("udp6_bind: inp == NULL"));
1017 INP_HASH_WLOCK(pcbinfo);
1018 inp->inp_vflag &= ~INP_IPV4;
1019 inp->inp_vflag |= INP_IPV6;
1020 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
1021 struct sockaddr_in6 *sin6_p;
1023 sin6_p = (struct sockaddr_in6 *)nam;
1025 if (IN6_IS_ADDR_UNSPECIFIED(&sin6_p->sin6_addr))
1026 inp->inp_vflag |= INP_IPV4;
1028 else if (IN6_IS_ADDR_V4MAPPED(&sin6_p->sin6_addr)) {
1029 struct sockaddr_in sin;
1031 in6_sin6_2_sin(&sin, sin6_p);
1032 inp->inp_vflag |= INP_IPV4;
1033 inp->inp_vflag &= ~INP_IPV6;
1034 error = in_pcbbind(inp, (struct sockaddr *)&sin,
1041 error = in6_pcbbind(inp, nam, td->td_ucred);
1045 INP_HASH_WUNLOCK(pcbinfo);
1051 udp6_close(struct socket *so)
1054 struct inpcbinfo *pcbinfo;
1056 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1057 inp = sotoinpcb(so);
1058 KASSERT(inp != NULL, ("udp6_close: inp == NULL"));
1062 if (inp->inp_vflag & INP_IPV4) {
1063 struct pr_usrreqs *pru;
1066 nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ?
1067 IPPROTO_UDP : IPPROTO_UDPLITE;
1069 pru = inetsw[ip_protox[nxt]].pr_usrreqs;
1070 (*pru->pru_disconnect)(so);
1074 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1075 INP_HASH_WLOCK(pcbinfo);
1076 in6_pcbdisconnect(inp);
1077 inp->in6p_laddr = in6addr_any;
1078 INP_HASH_WUNLOCK(pcbinfo);
1079 soisdisconnected(so);
1085 udp6_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
1088 struct inpcbinfo *pcbinfo;
1089 struct sockaddr_in6 *sin6;
1092 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1093 inp = sotoinpcb(so);
1094 sin6 = (struct sockaddr_in6 *)nam;
1095 KASSERT(inp != NULL, ("udp6_connect: inp == NULL"));
1098 * XXXRW: Need to clarify locking of v4/v6 flags.
1102 if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
1103 struct sockaddr_in sin;
1105 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) {
1109 if (inp->inp_faddr.s_addr != INADDR_ANY) {
1113 in6_sin6_2_sin(&sin, sin6);
1114 inp->inp_vflag |= INP_IPV4;
1115 inp->inp_vflag &= ~INP_IPV6;
1116 error = prison_remote_ip4(td->td_ucred, &sin.sin_addr);
1119 INP_HASH_WLOCK(pcbinfo);
1120 error = in_pcbconnect(inp, (struct sockaddr *)&sin,
1122 INP_HASH_WUNLOCK(pcbinfo);
1128 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1132 inp->inp_vflag &= ~INP_IPV4;
1133 inp->inp_vflag |= INP_IPV6;
1134 error = prison_remote_ip6(td->td_ucred, &sin6->sin6_addr);
1137 INP_HASH_WLOCK(pcbinfo);
1138 error = in6_pcbconnect(inp, nam, td->td_ucred);
1139 INP_HASH_WUNLOCK(pcbinfo);
1148 udp6_detach(struct socket *so)
1151 struct inpcbinfo *pcbinfo;
1154 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1155 inp = sotoinpcb(so);
1156 KASSERT(inp != NULL, ("udp6_detach: inp == NULL"));
1158 INP_INFO_WLOCK(pcbinfo);
1160 up = intoudpcb(inp);
1161 KASSERT(up != NULL, ("%s: up == NULL", __func__));
1164 INP_INFO_WUNLOCK(pcbinfo);
1169 udp6_disconnect(struct socket *so)
1172 struct inpcbinfo *pcbinfo;
1175 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1176 inp = sotoinpcb(so);
1177 KASSERT(inp != NULL, ("udp6_disconnect: inp == NULL"));
1181 if (inp->inp_vflag & INP_IPV4) {
1182 struct pr_usrreqs *pru;
1185 nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ?
1186 IPPROTO_UDP : IPPROTO_UDPLITE;
1188 pru = inetsw[ip_protox[nxt]].pr_usrreqs;
1189 (void)(*pru->pru_disconnect)(so);
1194 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1199 INP_HASH_WLOCK(pcbinfo);
1200 in6_pcbdisconnect(inp);
1201 inp->in6p_laddr = in6addr_any;
1202 INP_HASH_WUNLOCK(pcbinfo);
1204 so->so_state &= ~SS_ISCONNECTED; /* XXX */
1212 udp6_send(struct socket *so, int flags, struct mbuf *m,
1213 struct sockaddr *addr, struct mbuf *control, struct thread *td)
1216 struct inpcbinfo *pcbinfo;
1219 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1220 inp = sotoinpcb(so);
1221 KASSERT(inp != NULL, ("udp6_send: inp == NULL"));
1225 if (addr->sa_len != sizeof(struct sockaddr_in6)) {
1229 if (addr->sa_family != AF_INET6) {
1230 error = EAFNOSUPPORT;
1236 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
1238 struct sockaddr_in6 *sin6 = NULL;
1241 hasv4addr = (inp->inp_vflag & INP_IPV4);
1243 sin6 = (struct sockaddr_in6 *)addr;
1244 hasv4addr = IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)
1248 struct pr_usrreqs *pru;
1251 nxt = (inp->inp_socket->so_proto->pr_protocol ==
1252 IPPROTO_UDP) ? IPPROTO_UDP : IPPROTO_UDPLITE;
1254 * XXXRW: We release UDP-layer locks before calling
1255 * udp_send() in order to avoid recursion. However,
1256 * this does mean there is a short window where inp's
1257 * fields are unstable. Could this lead to a
1258 * potential race in which the factors causing us to
1259 * select the UDPv4 output routine are invalidated?
1263 in6_sin6_2_sin_in_sock(addr);
1264 pru = inetsw[ip_protox[nxt]].pr_usrreqs;
1265 /* addr will just be freed in sendit(). */
1266 return ((*pru->pru_send)(so, flags, m, addr, control,
1272 mac_inpcb_create_mbuf(inp, m);
1274 INP_HASH_WLOCK(pcbinfo);
1275 error = udp6_output(inp, m, addr, control, td);
1276 INP_HASH_WUNLOCK(pcbinfo);
1286 struct pr_usrreqs udp6_usrreqs = {
1287 .pru_abort = udp6_abort,
1288 .pru_attach = udp6_attach,
1289 .pru_bind = udp6_bind,
1290 .pru_connect = udp6_connect,
1291 .pru_control = in6_control,
1292 .pru_detach = udp6_detach,
1293 .pru_disconnect = udp6_disconnect,
1294 .pru_peeraddr = in6_mapped_peeraddr,
1295 .pru_send = udp6_send,
1296 .pru_shutdown = udp_shutdown,
1297 .pru_sockaddr = in6_mapped_sockaddr,
1298 .pru_soreceive = soreceive_dgram,
1299 .pru_sosend = sosend_dgram,
1300 .pru_sosetlabel = in_pcbsosetlabel,
1301 .pru_close = udp6_close