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 epoch_tracker et;
218 struct sockaddr_in6 fromsa[2];
219 struct m_tag *fwd_tag;
223 ifp = m->m_pkthdr.rcvif;
225 #ifndef PULLDOWN_TEST
226 IP6_EXTHDR_CHECK(m, off, sizeof(struct udphdr), IPPROTO_DONE);
227 ip6 = mtod(m, struct ip6_hdr *);
228 uh = (struct udphdr *)((caddr_t)ip6 + off);
230 IP6_EXTHDR_GET(uh, struct udphdr *, m, off, sizeof(*uh));
232 return (IPPROTO_DONE);
233 ip6 = mtod(m, struct ip6_hdr *);
236 UDPSTAT_INC(udps_ipackets);
239 * Destination port of 0 is illegal, based on RFC768.
241 if (uh->uh_dport == 0)
244 plen = ntohs(ip6->ip6_plen) - off + sizeof(*ip6);
245 ulen = ntohs((u_short)uh->uh_ulen);
248 cscov_partial = (nxt == IPPROTO_UDPLITE) ? 1 : 0;
249 if (nxt == IPPROTO_UDPLITE) {
250 /* Zero means checksum over the complete packet. */
255 if ((ulen < sizeof(struct udphdr)) || (ulen > plen)) {
256 /* XXX: What is the right UDPLite MIB counter? */
259 if (uh->uh_sum == 0) {
260 /* XXX: What is the right UDPLite MIB counter? */
264 if ((ulen < sizeof(struct udphdr)) || (plen != ulen)) {
265 UDPSTAT_INC(udps_badlen);
268 if (uh->uh_sum == 0) {
269 UDPSTAT_INC(udps_nosum);
274 if ((m->m_pkthdr.csum_flags & CSUM_DATA_VALID_IPV6) &&
276 if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR)
277 uh_sum = m->m_pkthdr.csum_data;
279 uh_sum = in6_cksum_pseudo(ip6, ulen, nxt,
280 m->m_pkthdr.csum_data);
283 uh_sum = in6_cksum_partial(m, nxt, off, plen, ulen);
286 UDPSTAT_INC(udps_badsum);
291 * Construct sockaddr format source address.
293 init_sin6(&fromsa[0], m, 0);
294 fromsa[0].sin6_port = uh->uh_sport;
295 init_sin6(&fromsa[1], m, 1);
296 fromsa[1].sin6_port = uh->uh_dport;
298 pcbinfo = udp_get_inpcbinfo(nxt);
299 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
301 struct inpcbhead *pcblist;
302 struct ip6_moptions *imo;
304 INP_INFO_RLOCK_ET(pcbinfo, et);
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))
405 * Don't look for additional matches if this one does
406 * not have either the SO_REUSEPORT or SO_REUSEADDR
407 * socket options set. This heuristic avoids
408 * searching through all pcbs in the common case of a
409 * non-shared port. It assumes that an application
410 * will never clear these options after setting them.
412 if ((last->inp_socket->so_options &
413 (SO_REUSEPORT|SO_REUSEPORT_LB|SO_REUSEADDR)) == 0)
419 * No matching pcb found; discard datagram. (No need
420 * to send an ICMP Port Unreachable for a broadcast
421 * or multicast datgram.)
423 UDPSTAT_INC(udps_noport);
424 UDPSTAT_INC(udps_noportmcast);
428 if (__predict_true(last->inp_flags2 & INP_FREED) == 0) {
429 if (nxt == IPPROTO_UDPLITE)
430 UDPLITE_PROBE(receive, NULL, last, ip6, last, uh);
432 UDP_PROBE(receive, NULL, last, ip6, last, uh);
433 if (udp6_append(last, m, off, fromsa) == 0)
438 INP_INFO_RUNLOCK_ET(pcbinfo, et);
439 return (IPPROTO_DONE);
442 * Locate pcb for datagram.
446 * Grab info from PACKET_TAG_IPFORWARD tag prepended to the chain.
448 if ((m->m_flags & M_IP6_NEXTHOP) &&
449 (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
450 struct sockaddr_in6 *next_hop6;
452 next_hop6 = (struct sockaddr_in6 *)(fwd_tag + 1);
455 * Transparently forwarded. Pretend to be the destination.
456 * Already got one like this?
458 inp = in6_pcblookup_mbuf(pcbinfo, &ip6->ip6_src,
459 uh->uh_sport, &ip6->ip6_dst, uh->uh_dport,
460 INPLOOKUP_RLOCKPCB, m->m_pkthdr.rcvif, m);
463 * It's new. Try to find the ambushing socket.
464 * Because we've rewritten the destination address,
465 * any hardware-generated hash is ignored.
467 inp = in6_pcblookup(pcbinfo, &ip6->ip6_src,
468 uh->uh_sport, &next_hop6->sin6_addr,
469 next_hop6->sin6_port ? htons(next_hop6->sin6_port) :
470 uh->uh_dport, INPLOOKUP_WILDCARD |
471 INPLOOKUP_RLOCKPCB, m->m_pkthdr.rcvif);
473 /* Remove the tag from the packet. We don't need it anymore. */
474 m_tag_delete(m, fwd_tag);
475 m->m_flags &= ~M_IP6_NEXTHOP;
477 inp = in6_pcblookup_mbuf(pcbinfo, &ip6->ip6_src,
478 uh->uh_sport, &ip6->ip6_dst, uh->uh_dport,
479 INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB,
480 m->m_pkthdr.rcvif, m);
482 if (udp_log_in_vain) {
483 char ip6bufs[INET6_ADDRSTRLEN];
484 char ip6bufd[INET6_ADDRSTRLEN];
487 "Connection attempt to UDP [%s]:%d from [%s]:%d\n",
488 ip6_sprintf(ip6bufd, &ip6->ip6_dst),
490 ip6_sprintf(ip6bufs, &ip6->ip6_src),
491 ntohs(uh->uh_sport));
493 if (nxt == IPPROTO_UDPLITE)
494 UDPLITE_PROBE(receive, NULL, NULL, ip6, NULL, uh);
496 UDP_PROBE(receive, NULL, NULL, ip6, NULL, uh);
497 UDPSTAT_INC(udps_noport);
498 if (m->m_flags & M_MCAST) {
499 printf("UDP6: M_MCAST is set in a unicast packet.\n");
500 UDPSTAT_INC(udps_noportmcast);
505 icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0);
506 return (IPPROTO_DONE);
508 INP_RLOCK_ASSERT(inp);
511 if (up->u_rxcslen == 0 || up->u_rxcslen > ulen) {
514 return (IPPROTO_DONE);
517 if (nxt == IPPROTO_UDPLITE)
518 UDPLITE_PROBE(receive, NULL, inp, ip6, inp, uh);
520 UDP_PROBE(receive, NULL, inp, ip6, inp, uh);
521 if (udp6_append(inp, m, off, fromsa) == 0)
523 return (IPPROTO_DONE);
526 INP_INFO_RUNLOCK_ET(pcbinfo, et);
530 return (IPPROTO_DONE);
534 udp6_common_ctlinput(int cmd, struct sockaddr *sa, void *d,
535 struct inpcbinfo *pcbinfo)
541 struct ip6ctlparam *ip6cp = NULL;
542 const struct sockaddr_in6 *sa6_src = NULL;
544 struct inpcb *(*notify)(struct inpcb *, int) = udp_notify;
545 struct udp_portonly {
550 if (sa->sa_family != AF_INET6 ||
551 sa->sa_len != sizeof(struct sockaddr_in6))
554 if ((unsigned)cmd >= PRC_NCMDS)
556 if (PRC_IS_REDIRECT(cmd))
557 notify = in6_rtchange, d = NULL;
558 else if (cmd == PRC_HOSTDEAD)
560 else if (inet6ctlerrmap[cmd] == 0)
563 /* if the parameter is from icmp6, decode it. */
565 ip6cp = (struct ip6ctlparam *)d;
567 ip6 = ip6cp->ip6c_ip6;
568 off = ip6cp->ip6c_off;
569 cmdarg = ip6cp->ip6c_cmdarg;
570 sa6_src = ip6cp->ip6c_src;
580 * XXX: We assume that when IPV6 is non NULL,
581 * M and OFF are valid.
584 /* Check if we can safely examine src and dst ports. */
585 if (m->m_pkthdr.len < off + sizeof(*uhp))
588 bzero(&uh, sizeof(uh));
589 m_copydata(m, off, sizeof(*uhp), (caddr_t)&uh);
591 if (!PRC_IS_REDIRECT(cmd)) {
592 /* Check to see if its tunneled */
594 inp = in6_pcblookup_mbuf(pcbinfo, &ip6->ip6_dst,
595 uh.uh_dport, &ip6->ip6_src, uh.uh_sport,
596 INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB,
597 m->m_pkthdr.rcvif, m);
602 if (up->u_icmp_func) {
605 (*up->u_icmp_func)(cmd, (struct sockaddr *)ip6cp->ip6c_src,
614 (void)in6_pcbnotify(pcbinfo, sa, uh.uh_dport,
615 (struct sockaddr *)ip6cp->ip6c_src, uh.uh_sport, cmd,
618 (void)in6_pcbnotify(pcbinfo, sa, 0,
619 (const struct sockaddr *)sa6_src, 0, cmd, cmdarg, notify);
623 udp6_ctlinput(int cmd, struct sockaddr *sa, void *d)
626 return (udp6_common_ctlinput(cmd, sa, d, &V_udbinfo));
630 udplite6_ctlinput(int cmd, struct sockaddr *sa, void *d)
633 return (udp6_common_ctlinput(cmd, sa, d, &V_ulitecbinfo));
637 udp6_getcred(SYSCTL_HANDLER_ARGS)
640 struct sockaddr_in6 addrs[2];
644 error = priv_check(req->td, PRIV_NETINET_GETCRED);
648 if (req->newlen != sizeof(addrs))
650 if (req->oldlen != sizeof(struct xucred))
652 error = SYSCTL_IN(req, addrs, sizeof(addrs));
655 if ((error = sa6_embedscope(&addrs[0], V_ip6_use_defzone)) != 0 ||
656 (error = sa6_embedscope(&addrs[1], V_ip6_use_defzone)) != 0) {
659 inp = in6_pcblookup(&V_udbinfo, &addrs[1].sin6_addr,
660 addrs[1].sin6_port, &addrs[0].sin6_addr, addrs[0].sin6_port,
661 INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB, NULL);
663 INP_RLOCK_ASSERT(inp);
664 if (inp->inp_socket == NULL)
667 error = cr_canseesocket(req->td->td_ucred,
670 cru2x(inp->inp_cred, &xuc);
675 error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred));
679 SYSCTL_PROC(_net_inet6_udp6, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW, 0,
680 0, udp6_getcred, "S,xucred", "Get the xucred of a UDP6 connection");
684 #define UH_UNLOCKED 0
686 udp6_output(struct socket *so, int flags_arg, struct mbuf *m,
687 struct sockaddr *addr6, struct mbuf *control, struct thread *td)
689 struct inpcbinfo *pcbinfo;
693 struct in6_addr *laddr, *faddr, in6a;
694 struct ip6_pktopts *optp, opt;
695 struct sockaddr_in6 *sin6, tmp;
696 struct epoch_tracker et;
697 int cscov_partial, error, flags, hlen, scope_ambiguous;
698 u_int32_t ulen, plen;
701 uint8_t nxt, unlock_inp, unlock_udbinfo;
703 /* addr6 has been validated in udp6_send(). */
704 sin6 = (struct sockaddr_in6 *)addr6;
707 * In contrast to to IPv4 we do not validate the max. packet length
708 * here due to IPv6 Jumbograms (RFC2675).
713 /* Protect *addr6 from overwrites. */
718 * Application should provide a proper zone ID or the use of
719 * default zone IDs should be enabled. Unfortunately, some
720 * applications do not behave as it should, so we need a
721 * workaround. Even if an appropriate ID is not determined,
722 * we'll see if we can determine the outgoing interface. If we
723 * can, determine the zone ID based on the interface below.
725 if (sin6->sin6_scope_id == 0 && !V_ip6_use_defzone)
727 if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0) {
736 KASSERT(inp != NULL, ("%s: inp == NULL", __func__));
738 * In the following cases we want a write lock on the inp for either
739 * local operations or for possible route cache updates in the IPv6
741 * - on connected sockets (sin6 is NULL) for route cache updates,
742 * - when we are not bound to an address and source port (it is
743 * in6_pcbsetport() which will require the write lock).
745 if (sin6 == NULL || (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
746 inp->inp_lport == 0)) {
748 unlock_inp = UH_WLOCKED;
751 unlock_inp = UH_RLOCKED;
753 nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ?
754 IPPROTO_UDP : IPPROTO_UDPLITE;
757 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
761 hasv4addr = (inp->inp_vflag & INP_IPV4);
763 hasv4addr = IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)
766 struct pr_usrreqs *pru;
769 * XXXRW: We release UDP-layer locks before calling
770 * udp_send() in order to avoid recursion. However,
771 * this does mean there is a short window where inp's
772 * fields are unstable. Could this lead to a
773 * potential race in which the factors causing us to
774 * select the UDPv4 output routine are invalidated?
776 if (unlock_inp == UH_WLOCKED)
781 in6_sin6_2_sin_in_sock((struct sockaddr *)sin6);
782 pru = inetsw[ip_protox[nxt]].pr_usrreqs;
783 /* addr will just be freed in sendit(). */
784 return ((*pru->pru_send)(so, flags_arg, m,
785 (struct sockaddr *)sin6, control, td));
789 if (sin6 && IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
791 * Given this is either an IPv6-only socket or no INET is
792 * supported we will fail the send if the given destination
793 * address is a v4mapped address.
795 if (unlock_inp == UH_WLOCKED)
803 if ((error = ip6_setpktopts(control, &opt,
804 inp->in6p_outputopts, td->td_ucred, nxt)) != 0) {
805 if (unlock_inp == UH_WLOCKED)
809 ip6_clearpktopts(&opt, -1);
817 optp = inp->in6p_outputopts;
819 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
821 IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) && inp->inp_lport == 0) {
822 INP_HASH_WLOCK(pcbinfo);
823 unlock_udbinfo = UH_WLOCKED;
824 } else if (sin6 != NULL &&
825 (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) ||
826 IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) ||
827 inp->inp_lport == 0)) {
828 INP_HASH_RLOCK_ET(pcbinfo, et);
829 unlock_udbinfo = UH_RLOCKED;
831 unlock_udbinfo = UH_UNLOCKED;
836 * Since we saw no essential reason for calling in_pcbconnect,
837 * we get rid of such kind of logic, and call in6_selectsrc
838 * and in6_pcbsetport in order to fill in the local address
839 * and the local port.
841 if (sin6->sin6_port == 0) {
842 error = EADDRNOTAVAIL;
846 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
847 /* how about ::ffff:0.0.0.0 case? */
853 * Given we handle the v4mapped case in the INET block above
854 * assert here that it must not happen anymore.
856 KASSERT(!IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr),
857 ("%s: sin6(%p)->sin6_addr is v4mapped which we "
858 "should have handled.", __func__, sin6));
860 /* This only requires read-locking. */
861 error = in6_selectsrc_socket(sin6, optp, inp,
862 td->td_ucred, scope_ambiguous, &in6a, NULL);
867 if (inp->inp_lport == 0) {
869 INP_WLOCK_ASSERT(inp);
870 error = in6_pcbsetport(laddr, inp, td->td_ucred);
872 /* Undo an address bind that may have occurred. */
873 inp->in6p_laddr = in6addr_any;
877 faddr = &sin6->sin6_addr;
878 fport = sin6->sin6_port; /* allow 0 port */
881 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
885 laddr = &inp->in6p_laddr;
886 faddr = &inp->in6p_faddr;
887 fport = inp->inp_fport;
890 ulen = m->m_pkthdr.len;
891 plen = sizeof(struct udphdr) + ulen;
892 hlen = sizeof(struct ip6_hdr);
895 * Calculate data length and get a mbuf
896 * for UDP and IP6 headers.
898 M_PREPEND(m, hlen + sizeof(struct udphdr), M_NOWAIT);
905 * Stuff checksum and output datagram.
907 cscov = cscov_partial = 0;
908 udp6 = (struct udphdr *)(mtod(m, caddr_t) + hlen);
909 udp6->uh_sport = inp->inp_lport; /* lport is always set in the PCB */
910 udp6->uh_dport = fport;
911 if (nxt == IPPROTO_UDPLITE) {
915 cscov = up->u_txcslen;
918 udp6->uh_ulen = htons(cscov);
920 * For UDP-Lite, checksum coverage length of zero means
921 * the entire UDPLite packet is covered by the checksum.
923 cscov_partial = (cscov == 0) ? 0 : 1;
924 } else if (plen <= 0xffff)
925 udp6->uh_ulen = htons((u_short)plen);
930 ip6 = mtod(m, struct ip6_hdr *);
931 ip6->ip6_flow = inp->inp_flow & IPV6_FLOWINFO_MASK;
932 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
933 ip6->ip6_vfc |= IPV6_VERSION;
934 ip6->ip6_plen = htons((u_short)plen);
936 ip6->ip6_hlim = in6_selecthlim(inp, NULL);
937 ip6->ip6_src = *laddr;
938 ip6->ip6_dst = *faddr;
941 mac_inpcb_create_mbuf(inp, m);
945 if ((udp6->uh_sum = in6_cksum_partial(m, nxt,
946 sizeof(struct ip6_hdr), plen, cscov)) == 0)
947 udp6->uh_sum = 0xffff;
949 udp6->uh_sum = in6_cksum_pseudo(ip6, plen, nxt, 0);
950 m->m_pkthdr.csum_flags = CSUM_UDP_IPV6;
951 m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
957 uint32_t hash_val, hash_type;
960 pr = inp->inp_socket->so_proto->pr_protocol;
962 * Calculate an appropriate RSS hash for UDP and
965 * The called function will take care of figuring out
966 * whether a 2-tuple or 4-tuple hash is required based
967 * on the currently configured scheme.
969 * Later later on connected socket values should be
970 * cached in the inpcb and reused, rather than constantly
973 * UDP Lite is a different protocol number and will
974 * likely end up being hashed as a 2-tuple until
975 * RSS / NICs grow UDP Lite protocol awareness.
977 if (rss_proto_software_hash_v6(faddr, laddr, fport,
978 inp->inp_lport, pr, &hash_val, &hash_type) == 0) {
979 m->m_pkthdr.flowid = hash_val;
980 M_HASHTYPE_SET(m, hash_type);
984 * Don't override with the inp cached flowid.
986 * Until the whole UDP path is vetted, it may actually
989 flags |= IP_NODEFAULTFLOWID;
993 UDPSTAT_INC(udps_opackets);
994 if (unlock_udbinfo == UH_WLOCKED)
995 INP_HASH_WUNLOCK(pcbinfo);
996 else if (unlock_udbinfo == UH_RLOCKED)
997 INP_HASH_RUNLOCK_ET(pcbinfo, et);
998 if (nxt == IPPROTO_UDPLITE)
999 UDPLITE_PROBE(send, NULL, inp, ip6, inp, udp6);
1001 UDP_PROBE(send, NULL, inp, ip6, inp, udp6);
1002 error = ip6_output(m, optp,
1003 (unlock_inp == UH_WLOCKED) ? &inp->inp_route6 : NULL, flags,
1004 inp->in6p_moptions, NULL, inp);
1005 if (unlock_inp == UH_WLOCKED)
1011 ip6_clearpktopts(&opt, -1);
1017 if (unlock_udbinfo == UH_WLOCKED) {
1018 KASSERT(unlock_inp == UH_WLOCKED, ("%s: excl udbinfo lock, "
1019 "non-excl inp lock: pcbinfo %p %#x inp %p %#x",
1020 __func__, pcbinfo, unlock_udbinfo, inp, unlock_inp));
1021 INP_HASH_WUNLOCK(pcbinfo);
1023 } else if (unlock_udbinfo == UH_RLOCKED) {
1024 KASSERT(unlock_inp == UH_RLOCKED, ("%s: non-excl udbinfo lock, "
1025 "excl inp lock: pcbinfo %p %#x inp %p %#x",
1026 __func__, pcbinfo, unlock_udbinfo, inp, unlock_inp));
1027 INP_HASH_RUNLOCK_ET(pcbinfo, et);
1029 } else if (unlock_inp == UH_WLOCKED)
1034 ip6_clearpktopts(&opt, -1);
1043 udp6_abort(struct socket *so)
1046 struct inpcbinfo *pcbinfo;
1048 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1049 inp = sotoinpcb(so);
1050 KASSERT(inp != NULL, ("udp6_abort: inp == NULL"));
1054 if (inp->inp_vflag & INP_IPV4) {
1055 struct pr_usrreqs *pru;
1058 nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ?
1059 IPPROTO_UDP : IPPROTO_UDPLITE;
1061 pru = inetsw[ip_protox[nxt]].pr_usrreqs;
1062 (*pru->pru_abort)(so);
1067 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1068 INP_HASH_WLOCK(pcbinfo);
1069 in6_pcbdisconnect(inp);
1070 inp->in6p_laddr = in6addr_any;
1071 INP_HASH_WUNLOCK(pcbinfo);
1072 soisdisconnected(so);
1078 udp6_attach(struct socket *so, int proto, struct thread *td)
1081 struct inpcbinfo *pcbinfo;
1084 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1085 inp = sotoinpcb(so);
1086 KASSERT(inp == NULL, ("udp6_attach: inp != NULL"));
1088 if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
1089 error = soreserve(so, udp_sendspace, udp_recvspace);
1093 INP_INFO_WLOCK(pcbinfo);
1094 error = in_pcballoc(so, pcbinfo);
1096 INP_INFO_WUNLOCK(pcbinfo);
1099 inp = (struct inpcb *)so->so_pcb;
1100 inp->inp_vflag |= INP_IPV6;
1101 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0)
1102 inp->inp_vflag |= INP_IPV4;
1103 inp->in6p_hops = -1; /* use kernel default */
1104 inp->in6p_cksum = -1; /* just to be sure */
1107 * IPv4 TTL initialization is necessary for an IPv6 socket as well,
1108 * because the socket may be bound to an IPv6 wildcard address,
1109 * which may match an IPv4-mapped IPv6 address.
1111 inp->inp_ip_ttl = V_ip_defttl;
1113 error = udp_newudpcb(inp);
1117 INP_INFO_WUNLOCK(pcbinfo);
1121 INP_INFO_WUNLOCK(pcbinfo);
1126 udp6_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
1129 struct inpcbinfo *pcbinfo;
1132 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1133 inp = sotoinpcb(so);
1134 KASSERT(inp != NULL, ("udp6_bind: inp == NULL"));
1137 INP_HASH_WLOCK(pcbinfo);
1138 inp->inp_vflag &= ~INP_IPV4;
1139 inp->inp_vflag |= INP_IPV6;
1140 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
1141 struct sockaddr_in6 *sin6_p;
1143 sin6_p = (struct sockaddr_in6 *)nam;
1145 if (IN6_IS_ADDR_UNSPECIFIED(&sin6_p->sin6_addr))
1146 inp->inp_vflag |= INP_IPV4;
1148 else if (IN6_IS_ADDR_V4MAPPED(&sin6_p->sin6_addr)) {
1149 struct sockaddr_in sin;
1151 in6_sin6_2_sin(&sin, sin6_p);
1152 inp->inp_vflag |= INP_IPV4;
1153 inp->inp_vflag &= ~INP_IPV6;
1154 error = in_pcbbind(inp, (struct sockaddr *)&sin,
1161 error = in6_pcbbind(inp, nam, td->td_ucred);
1165 INP_HASH_WUNLOCK(pcbinfo);
1171 udp6_close(struct socket *so)
1174 struct inpcbinfo *pcbinfo;
1176 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1177 inp = sotoinpcb(so);
1178 KASSERT(inp != NULL, ("udp6_close: inp == NULL"));
1182 if (inp->inp_vflag & INP_IPV4) {
1183 struct pr_usrreqs *pru;
1186 nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ?
1187 IPPROTO_UDP : IPPROTO_UDPLITE;
1189 pru = inetsw[ip_protox[nxt]].pr_usrreqs;
1190 (*pru->pru_disconnect)(so);
1194 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1195 INP_HASH_WLOCK(pcbinfo);
1196 in6_pcbdisconnect(inp);
1197 inp->in6p_laddr = in6addr_any;
1198 INP_HASH_WUNLOCK(pcbinfo);
1199 soisdisconnected(so);
1205 udp6_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
1208 struct inpcbinfo *pcbinfo;
1209 struct sockaddr_in6 *sin6;
1212 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1213 inp = sotoinpcb(so);
1214 sin6 = (struct sockaddr_in6 *)nam;
1215 KASSERT(inp != NULL, ("udp6_connect: inp == NULL"));
1218 * XXXRW: Need to clarify locking of v4/v6 flags.
1222 if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
1223 struct sockaddr_in sin;
1225 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) {
1229 if ((inp->inp_vflag & INP_IPV4) == 0) {
1230 error = EAFNOSUPPORT;
1233 if (inp->inp_faddr.s_addr != INADDR_ANY) {
1237 in6_sin6_2_sin(&sin, sin6);
1238 inp->inp_vflag |= INP_IPV4;
1239 inp->inp_vflag &= ~INP_IPV6;
1240 error = prison_remote_ip4(td->td_ucred, &sin.sin_addr);
1243 INP_HASH_WLOCK(pcbinfo);
1244 error = in_pcbconnect(inp, (struct sockaddr *)&sin,
1246 INP_HASH_WUNLOCK(pcbinfo);
1251 if ((inp->inp_vflag & INP_IPV6) == 0) {
1252 error = EAFNOSUPPORT;
1257 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1261 inp->inp_vflag &= ~INP_IPV4;
1262 inp->inp_vflag |= INP_IPV6;
1263 error = prison_remote_ip6(td->td_ucred, &sin6->sin6_addr);
1266 INP_HASH_WLOCK(pcbinfo);
1267 error = in6_pcbconnect(inp, nam, td->td_ucred);
1268 INP_HASH_WUNLOCK(pcbinfo);
1277 udp6_detach(struct socket *so)
1280 struct inpcbinfo *pcbinfo;
1283 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1284 inp = sotoinpcb(so);
1285 KASSERT(inp != NULL, ("udp6_detach: inp == NULL"));
1287 INP_INFO_WLOCK(pcbinfo);
1289 up = intoudpcb(inp);
1290 KASSERT(up != NULL, ("%s: up == NULL", __func__));
1293 INP_INFO_WUNLOCK(pcbinfo);
1298 udp6_disconnect(struct socket *so)
1301 struct inpcbinfo *pcbinfo;
1303 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1304 inp = sotoinpcb(so);
1305 KASSERT(inp != NULL, ("udp6_disconnect: inp == NULL"));
1309 if (inp->inp_vflag & INP_IPV4) {
1310 struct pr_usrreqs *pru;
1313 nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ?
1314 IPPROTO_UDP : IPPROTO_UDPLITE;
1316 pru = inetsw[ip_protox[nxt]].pr_usrreqs;
1317 (void)(*pru->pru_disconnect)(so);
1322 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1327 INP_HASH_WLOCK(pcbinfo);
1328 in6_pcbdisconnect(inp);
1329 inp->in6p_laddr = in6addr_any;
1330 INP_HASH_WUNLOCK(pcbinfo);
1332 so->so_state &= ~SS_ISCONNECTED; /* XXX */
1339 udp6_send(struct socket *so, int flags, struct mbuf *m,
1340 struct sockaddr *addr, struct mbuf *control, struct thread *td)
1345 if (addr->sa_len != sizeof(struct sockaddr_in6)) {
1349 if (addr->sa_family != AF_INET6) {
1350 error = EAFNOSUPPORT;
1355 return (udp6_output(so, flags, m, addr, control, td));
1364 struct pr_usrreqs udp6_usrreqs = {
1365 .pru_abort = udp6_abort,
1366 .pru_attach = udp6_attach,
1367 .pru_bind = udp6_bind,
1368 .pru_connect = udp6_connect,
1369 .pru_control = in6_control,
1370 .pru_detach = udp6_detach,
1371 .pru_disconnect = udp6_disconnect,
1372 .pru_peeraddr = in6_mapped_peeraddr,
1373 .pru_send = udp6_send,
1374 .pru_shutdown = udp_shutdown,
1375 .pru_sockaddr = in6_mapped_sockaddr,
1376 .pru_soreceive = soreceive_dgram,
1377 .pru_sosend = sosend_dgram,
1378 .pru_sosetlabel = in_pcbsosetlabel,
1379 .pru_close = udp6_close