2 * SPDX-License-Identifier: BSD-3-Clause
4 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5 * Copyright (c) 2010-2011 Juniper Networks, Inc.
6 * Copyright (c) 2014 Kevin Lo
9 * Portions of this software were developed by Robert N. M. Watson under
10 * contract to Juniper Networks, Inc.
12 * Redistribution and use in source and binary forms, with or without
13 * modification, are permitted provided that the following conditions
15 * 1. Redistributions of source code must retain the above copyright
16 * notice, this list of conditions and the following disclaimer.
17 * 2. Redistributions in binary form must reproduce the above copyright
18 * notice, this list of conditions and the following disclaimer in the
19 * documentation and/or other materials provided with the distribution.
20 * 3. Neither the name of the project nor the names of its contributors
21 * may be used to endorse or promote products derived from this software
22 * without specific prior written permission.
24 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
36 * $KAME: udp6_usrreq.c,v 1.27 2001/05/21 05:45:10 jinmei Exp $
37 * $KAME: udp6_output.c,v 1.31 2001/05/21 16:39:15 jinmei Exp $
41 * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995
42 * The Regents of the University of California.
43 * All rights reserved.
45 * Redistribution and use in source and binary forms, with or without
46 * modification, are permitted provided that the following conditions
48 * 1. Redistributions of source code must retain the above copyright
49 * notice, this list of conditions and the following disclaimer.
50 * 2. Redistributions in binary form must reproduce the above copyright
51 * notice, this list of conditions and the following disclaimer in the
52 * documentation and/or other materials provided with the distribution.
53 * 3. Neither the name of the University nor the names of its contributors
54 * may be used to endorse or promote products derived from this software
55 * without specific prior written permission.
57 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
58 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
59 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
60 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
61 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
62 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
63 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
64 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
65 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
66 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
69 * @(#)udp_usrreq.c 8.6 (Berkeley) 5/23/95
72 #include <sys/cdefs.h>
73 __FBSDID("$FreeBSD$");
76 #include "opt_inet6.h"
77 #include "opt_ipsec.h"
80 #include <sys/param.h>
82 #include <sys/kernel.h>
87 #include <sys/protosw.h>
89 #include <sys/signalvar.h>
90 #include <sys/socket.h>
91 #include <sys/socketvar.h>
93 #include <sys/sysctl.h>
94 #include <sys/syslog.h>
95 #include <sys/systm.h>
98 #include <net/if_var.h>
99 #include <net/if_types.h>
100 #include <net/route.h>
101 #include <net/rss_config.h>
103 #include <netinet/in.h>
104 #include <netinet/in_kdtrace.h>
105 #include <netinet/in_pcb.h>
106 #include <netinet/in_systm.h>
107 #include <netinet/in_var.h>
108 #include <netinet/ip.h>
109 #include <netinet/ip6.h>
110 #include <netinet/icmp6.h>
111 #include <netinet/ip_var.h>
112 #include <netinet/udp.h>
113 #include <netinet/udp_var.h>
114 #include <netinet/udplite.h>
116 #include <netinet6/ip6protosw.h>
117 #include <netinet6/ip6_var.h>
118 #include <netinet6/in6_pcb.h>
119 #include <netinet6/in6_rss.h>
120 #include <netinet6/udp6_var.h>
121 #include <netinet6/scope6_var.h>
123 #include <netipsec/ipsec_support.h>
125 #include <security/mac/mac_framework.h>
128 * UDP protocol implementation.
129 * Per RFC 768, August, 1980.
132 extern struct protosw inetsw[];
133 static void udp6_detach(struct socket *so);
136 udp6_append(struct inpcb *inp, struct mbuf *n, int off,
137 struct sockaddr_in6 *fromsa)
140 struct mbuf *opts = NULL, *tmp_opts;
143 INP_LOCK_ASSERT(inp);
146 * Engage the tunneling protocol.
149 if (up->u_tun_func != NULL) {
152 (*up->u_tun_func)(n, off, inp, (struct sockaddr *)&fromsa[0],
155 return (in_pcbrele_rlocked(inp));
157 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
158 /* Check AH/ESP integrity. */
159 if (IPSEC_ENABLED(ipv6)) {
160 if (IPSEC_CHECK_POLICY(ipv6, n, inp) != 0) {
167 if (mac_inpcb_check_deliver(inp, n) != 0) {
173 if (inp->inp_flags & INP_CONTROLOPTS ||
174 inp->inp_socket->so_options & SO_TIMESTAMP)
175 ip6_savecontrol(inp, n, &opts);
176 if ((inp->inp_vflag & INP_IPV6) && (inp->inp_flags2 & INP_ORIGDSTADDR)) {
177 tmp_opts = sbcreatecontrol((caddr_t)&fromsa[1],
178 sizeof(struct sockaddr_in6), IPV6_ORIGDSTADDR, IPPROTO_IPV6);
181 tmp_opts->m_next = opts;
188 m_adj(n, off + sizeof(struct udphdr));
190 so = inp->inp_socket;
191 SOCKBUF_LOCK(&so->so_rcv);
192 if (sbappendaddr_locked(&so->so_rcv, (struct sockaddr *)&fromsa[0], n,
194 SOCKBUF_UNLOCK(&so->so_rcv);
198 UDPSTAT_INC(udps_fullsock);
200 sorwakeup_locked(so);
205 udp6_input(struct mbuf **mp, int *offp, int proto)
207 struct mbuf *m = *mp;
212 struct inpcbinfo *pcbinfo;
217 struct sockaddr_in6 fromsa[2];
218 struct m_tag *fwd_tag;
224 ifp = m->m_pkthdr.rcvif;
226 #ifndef PULLDOWN_TEST
227 IP6_EXTHDR_CHECK(m, off, sizeof(struct udphdr), IPPROTO_DONE);
228 ip6 = mtod(m, struct ip6_hdr *);
229 uh = (struct udphdr *)((caddr_t)ip6 + off);
231 IP6_EXTHDR_GET(uh, struct udphdr *, m, off, sizeof(*uh));
233 return (IPPROTO_DONE);
234 ip6 = mtod(m, struct ip6_hdr *);
237 UDPSTAT_INC(udps_ipackets);
240 * Destination port of 0 is illegal, based on RFC768.
242 if (uh->uh_dport == 0)
245 plen = ntohs(ip6->ip6_plen) - off + sizeof(*ip6);
246 ulen = ntohs((u_short)uh->uh_ulen);
249 cscov_partial = (nxt == IPPROTO_UDPLITE) ? 1 : 0;
250 if (nxt == IPPROTO_UDPLITE) {
251 /* Zero means checksum over the complete packet. */
256 if ((ulen < sizeof(struct udphdr)) || (ulen > plen)) {
257 /* XXX: What is the right UDPLite MIB counter? */
260 if (uh->uh_sum == 0) {
261 /* XXX: What is the right UDPLite MIB counter? */
265 if ((ulen < sizeof(struct udphdr)) || (plen != ulen)) {
266 UDPSTAT_INC(udps_badlen);
269 if (uh->uh_sum == 0) {
270 UDPSTAT_INC(udps_nosum);
275 if ((m->m_pkthdr.csum_flags & CSUM_DATA_VALID_IPV6) &&
277 if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR)
278 uh_sum = m->m_pkthdr.csum_data;
280 uh_sum = in6_cksum_pseudo(ip6, ulen, nxt,
281 m->m_pkthdr.csum_data);
284 uh_sum = in6_cksum_partial(m, nxt, off, plen, ulen);
287 UDPSTAT_INC(udps_badsum);
292 * Construct sockaddr format source address.
294 init_sin6(&fromsa[0], m, 0);
295 fromsa[0].sin6_port = uh->uh_sport;
296 init_sin6(&fromsa[1], m, 1);
297 fromsa[1].sin6_port = uh->uh_dport;
299 pcbinfo = udp_get_inpcbinfo(nxt);
300 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
302 struct inpcbhead *pcblist;
303 struct ip6_moptions *imo;
306 * In the event that laddr should be set to the link-local
307 * address (this happens in RIPng), the multicast address
308 * specified in the received packet will not match laddr. To
309 * handle this situation, matching is relaxed if the
310 * receiving interface is the same as one specified in the
311 * socket and if the destination multicast address matches
312 * one of the multicast groups specified in the socket.
316 * KAME note: traditionally we dropped udpiphdr from mbuf
317 * here. We need udphdr for IPsec processing so we do that
320 pcblist = udp_get_pcblist(nxt);
322 CK_LIST_FOREACH(inp, pcblist, inp_list) {
323 if ((inp->inp_vflag & INP_IPV6) == 0)
325 if (inp->inp_lport != uh->uh_dport)
327 if (inp->inp_fport != 0 &&
328 inp->inp_fport != uh->uh_sport)
330 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
331 if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr,
335 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
336 if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr,
338 inp->inp_fport != uh->uh_sport)
343 * XXXRW: Because we weren't holding either the inpcb
344 * or the hash lock when we checked for a match
345 * before, we should probably recheck now that the
346 * inpcb lock is (supposed to be) held.
350 * Handle socket delivery policy for any-source
351 * and source-specific multicast. [RFC3678]
353 imo = inp->in6p_moptions;
354 if (imo && IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
355 struct sockaddr_in6 mcaddr;
359 if (__predict_false(inp->inp_flags2 & INP_FREED)) {
364 bzero(&mcaddr, sizeof(struct sockaddr_in6));
365 mcaddr.sin6_len = sizeof(struct sockaddr_in6);
366 mcaddr.sin6_family = AF_INET6;
367 mcaddr.sin6_addr = ip6->ip6_dst;
369 blocked = im6o_mc_filter(imo, ifp,
370 (struct sockaddr *)&mcaddr,
371 (struct sockaddr *)&fromsa[0]);
372 if (blocked != MCAST_PASS) {
373 if (blocked == MCAST_NOTGMEMBER)
374 IP6STAT_INC(ip6s_notmember);
375 if (blocked == MCAST_NOTSMEMBER ||
376 blocked == MCAST_MUTED)
377 UDPSTAT_INC(udps_filtermcast);
378 INP_RUNLOCK(inp); /* XXX */
387 if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) !=
390 if (__predict_true(last->inp_flags2 & INP_FREED) == 0) {
391 if (nxt == IPPROTO_UDPLITE)
392 UDPLITE_PROBE(receive, NULL, last,
395 UDP_PROBE(receive, NULL, last,
397 if (udp6_append(last, n, off, fromsa)) {
398 /* XXX-BZ do we leak m here? */
400 return (IPPROTO_DONE);
408 * Don't look for additional matches if this one does
409 * not have either the SO_REUSEPORT or SO_REUSEADDR
410 * socket options set. This heuristic avoids
411 * searching through all pcbs in the common case of a
412 * non-shared port. It assumes that an application
413 * will never clear these options after setting them.
415 if ((last->inp_socket->so_options &
416 (SO_REUSEPORT|SO_REUSEPORT_LB|SO_REUSEADDR)) == 0)
422 * No matching pcb found; discard datagram. (No need
423 * to send an ICMP Port Unreachable for a broadcast
424 * or multicast datgram.)
426 UDPSTAT_INC(udps_noport);
427 UDPSTAT_INC(udps_noportmcast);
431 if (__predict_true(last->inp_flags2 & INP_FREED) == 0) {
432 if (nxt == IPPROTO_UDPLITE)
433 UDPLITE_PROBE(receive, NULL, last, ip6, last, uh);
435 UDP_PROBE(receive, NULL, last, ip6, last, uh);
436 if (udp6_append(last, m, off, fromsa) == 0)
441 return (IPPROTO_DONE);
444 * Locate pcb for datagram.
448 * Grab info from PACKET_TAG_IPFORWARD tag prepended to the chain.
450 if ((m->m_flags & M_IP6_NEXTHOP) &&
451 (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
452 struct sockaddr_in6 *next_hop6;
454 next_hop6 = (struct sockaddr_in6 *)(fwd_tag + 1);
457 * Transparently forwarded. Pretend to be the destination.
458 * Already got one like this?
460 inp = in6_pcblookup_mbuf(pcbinfo, &ip6->ip6_src,
461 uh->uh_sport, &ip6->ip6_dst, uh->uh_dport,
462 INPLOOKUP_RLOCKPCB, m->m_pkthdr.rcvif, m);
465 * It's new. Try to find the ambushing socket.
466 * Because we've rewritten the destination address,
467 * any hardware-generated hash is ignored.
469 inp = in6_pcblookup(pcbinfo, &ip6->ip6_src,
470 uh->uh_sport, &next_hop6->sin6_addr,
471 next_hop6->sin6_port ? htons(next_hop6->sin6_port) :
472 uh->uh_dport, INPLOOKUP_WILDCARD |
473 INPLOOKUP_RLOCKPCB, m->m_pkthdr.rcvif);
475 /* Remove the tag from the packet. We don't need it anymore. */
476 m_tag_delete(m, fwd_tag);
477 m->m_flags &= ~M_IP6_NEXTHOP;
479 inp = in6_pcblookup_mbuf(pcbinfo, &ip6->ip6_src,
480 uh->uh_sport, &ip6->ip6_dst, uh->uh_dport,
481 INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB,
482 m->m_pkthdr.rcvif, m);
484 if (udp_log_in_vain) {
485 char ip6bufs[INET6_ADDRSTRLEN];
486 char ip6bufd[INET6_ADDRSTRLEN];
489 "Connection attempt to UDP [%s]:%d from [%s]:%d\n",
490 ip6_sprintf(ip6bufd, &ip6->ip6_dst),
492 ip6_sprintf(ip6bufs, &ip6->ip6_src),
493 ntohs(uh->uh_sport));
495 if (nxt == IPPROTO_UDPLITE)
496 UDPLITE_PROBE(receive, NULL, NULL, ip6, NULL, uh);
498 UDP_PROBE(receive, NULL, NULL, ip6, NULL, uh);
499 UDPSTAT_INC(udps_noport);
500 if (m->m_flags & M_MCAST) {
501 printf("UDP6: M_MCAST is set in a unicast packet.\n");
502 UDPSTAT_INC(udps_noportmcast);
507 icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_NOPORT, 0);
509 return (IPPROTO_DONE);
511 INP_RLOCK_ASSERT(inp);
514 if (up->u_rxcslen == 0 || up->u_rxcslen > ulen) {
518 return (IPPROTO_DONE);
521 if (nxt == IPPROTO_UDPLITE)
522 UDPLITE_PROBE(receive, NULL, inp, ip6, inp, uh);
524 UDP_PROBE(receive, NULL, inp, ip6, inp, uh);
525 if (udp6_append(inp, m, off, fromsa) == 0)
528 return (IPPROTO_DONE);
534 return (IPPROTO_DONE);
538 udp6_common_ctlinput(int cmd, struct sockaddr *sa, void *d,
539 struct inpcbinfo *pcbinfo)
545 struct ip6ctlparam *ip6cp = NULL;
546 const struct sockaddr_in6 *sa6_src = NULL;
548 struct inpcb *(*notify)(struct inpcb *, int) = udp_notify;
549 struct udp_portonly {
554 if (sa->sa_family != AF_INET6 ||
555 sa->sa_len != sizeof(struct sockaddr_in6))
558 if ((unsigned)cmd >= PRC_NCMDS)
560 if (PRC_IS_REDIRECT(cmd))
561 notify = in6_rtchange, d = NULL;
562 else if (cmd == PRC_HOSTDEAD)
564 else if (inet6ctlerrmap[cmd] == 0)
567 /* if the parameter is from icmp6, decode it. */
569 ip6cp = (struct ip6ctlparam *)d;
571 ip6 = ip6cp->ip6c_ip6;
572 off = ip6cp->ip6c_off;
573 cmdarg = ip6cp->ip6c_cmdarg;
574 sa6_src = ip6cp->ip6c_src;
584 * XXX: We assume that when IPV6 is non NULL,
585 * M and OFF are valid.
588 /* Check if we can safely examine src and dst ports. */
589 if (m->m_pkthdr.len < off + sizeof(*uhp))
592 bzero(&uh, sizeof(uh));
593 m_copydata(m, off, sizeof(*uhp), (caddr_t)&uh);
595 if (!PRC_IS_REDIRECT(cmd)) {
596 /* Check to see if its tunneled */
598 inp = in6_pcblookup_mbuf(pcbinfo, &ip6->ip6_dst,
599 uh.uh_dport, &ip6->ip6_src, uh.uh_sport,
600 INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB,
601 m->m_pkthdr.rcvif, m);
606 if (up->u_icmp_func) {
609 (*up->u_icmp_func)(cmd, (struct sockaddr *)ip6cp->ip6c_src,
618 (void)in6_pcbnotify(pcbinfo, sa, uh.uh_dport,
619 (struct sockaddr *)ip6cp->ip6c_src, uh.uh_sport, cmd,
622 (void)in6_pcbnotify(pcbinfo, sa, 0,
623 (const struct sockaddr *)sa6_src, 0, cmd, cmdarg, notify);
627 udp6_ctlinput(int cmd, struct sockaddr *sa, void *d)
630 return (udp6_common_ctlinput(cmd, sa, d, &V_udbinfo));
634 udplite6_ctlinput(int cmd, struct sockaddr *sa, void *d)
637 return (udp6_common_ctlinput(cmd, sa, d, &V_ulitecbinfo));
641 udp6_getcred(SYSCTL_HANDLER_ARGS)
644 struct sockaddr_in6 addrs[2];
645 struct epoch_tracker et;
649 error = priv_check(req->td, PRIV_NETINET_GETCRED);
653 if (req->newlen != sizeof(addrs))
655 if (req->oldlen != sizeof(struct xucred))
657 error = SYSCTL_IN(req, addrs, sizeof(addrs));
660 if ((error = sa6_embedscope(&addrs[0], V_ip6_use_defzone)) != 0 ||
661 (error = sa6_embedscope(&addrs[1], V_ip6_use_defzone)) != 0) {
665 inp = in6_pcblookup(&V_udbinfo, &addrs[1].sin6_addr,
666 addrs[1].sin6_port, &addrs[0].sin6_addr, addrs[0].sin6_port,
667 INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB, NULL);
670 INP_RLOCK_ASSERT(inp);
671 if (inp->inp_socket == NULL)
674 error = cr_canseesocket(req->td->td_ucred,
677 cru2x(inp->inp_cred, &xuc);
682 error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred));
686 SYSCTL_PROC(_net_inet6_udp6, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW, 0,
687 0, udp6_getcred, "S,xucred", "Get the xucred of a UDP6 connection");
690 udp6_output(struct socket *so, int flags_arg, struct mbuf *m,
691 struct sockaddr *addr6, struct mbuf *control, struct thread *td)
696 struct in6_addr *laddr, *faddr, in6a;
697 struct ip6_pktopts *optp, opt;
698 struct sockaddr_in6 *sin6, tmp;
699 struct epoch_tracker et;
700 int cscov_partial, error, flags, hlen, scope_ambiguous;
701 u_int32_t ulen, plen;
706 /* addr6 has been validated in udp6_send(). */
707 sin6 = (struct sockaddr_in6 *)addr6;
710 * In contrast to to IPv4 we do not validate the max. packet length
711 * here due to IPv6 Jumbograms (RFC2675).
716 /* Protect *addr6 from overwrites. */
721 * Application should provide a proper zone ID or the use of
722 * default zone IDs should be enabled. Unfortunately, some
723 * applications do not behave as it should, so we need a
724 * workaround. Even if an appropriate ID is not determined,
725 * we'll see if we can determine the outgoing interface. If we
726 * can, determine the zone ID based on the interface below.
728 if (sin6->sin6_scope_id == 0 && !V_ip6_use_defzone)
730 if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0) {
739 KASSERT(inp != NULL, ("%s: inp == NULL", __func__));
741 * In the following cases we want a write lock on the inp for either
742 * local operations or for possible route cache updates in the IPv6
744 * - on connected sockets (sin6 is NULL) for route cache updates,
745 * - when we are not bound to an address and source port (it is
746 * in6_pcbsetport() which will require the write lock).
748 * We check the inp fields before actually locking the inp, so
749 * here exists a race, and we may WLOCK the inp and end with already
750 * bound one by other thread. This is fine.
752 if (sin6 == NULL || (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
753 inp->inp_lport == 0))
758 nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ?
759 IPPROTO_UDP : IPPROTO_UDPLITE;
762 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
766 hasv4addr = (inp->inp_vflag & INP_IPV4);
768 hasv4addr = IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)
771 struct pr_usrreqs *pru;
774 * XXXRW: We release UDP-layer locks before calling
775 * udp_send() in order to avoid recursion. However,
776 * this does mean there is a short window where inp's
777 * fields are unstable. Could this lead to a
778 * potential race in which the factors causing us to
779 * select the UDPv4 output routine are invalidated?
783 in6_sin6_2_sin_in_sock((struct sockaddr *)sin6);
784 pru = inetsw[ip_protox[nxt]].pr_usrreqs;
785 /* addr will just be freed in sendit(). */
786 return ((*pru->pru_send)(so, flags_arg, m,
787 (struct sockaddr *)sin6, control, td));
791 if (sin6 && IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
793 * Given this is either an IPv6-only socket or no INET is
794 * supported we will fail the send if the given destination
795 * address is a v4mapped address.
797 * XXXGL: do we leak m and control?
804 if ((error = ip6_setpktopts(control, &opt,
805 inp->in6p_outputopts, td->td_ucred, nxt)) != 0) {
807 ip6_clearpktopts(&opt, -1);
815 optp = inp->in6p_outputopts;
821 * Since we saw no essential reason for calling in_pcbconnect,
822 * we get rid of such kind of logic, and call in6_selectsrc
823 * and in6_pcbsetport in order to fill in the local address
824 * and the local port.
826 if (sin6->sin6_port == 0) {
827 error = EADDRNOTAVAIL;
831 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
832 /* how about ::ffff:0.0.0.0 case? */
838 * Given we handle the v4mapped case in the INET block above
839 * assert here that it must not happen anymore.
841 KASSERT(!IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr),
842 ("%s: sin6(%p)->sin6_addr is v4mapped which we "
843 "should have handled.", __func__, sin6));
845 /* This only requires read-locking. */
846 error = in6_selectsrc_socket(sin6, optp, inp,
847 td->td_ucred, scope_ambiguous, &in6a, NULL);
852 if (inp->inp_lport == 0) {
853 struct inpcbinfo *pcbinfo;
855 INP_WLOCK_ASSERT(inp);
857 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
858 INP_HASH_WLOCK(pcbinfo);
859 error = in6_pcbsetport(laddr, inp, td->td_ucred);
860 INP_HASH_WUNLOCK(pcbinfo);
862 /* Undo an address bind that may have occurred. */
863 inp->in6p_laddr = in6addr_any;
867 faddr = &sin6->sin6_addr;
868 fport = sin6->sin6_port; /* allow 0 port */
871 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
875 laddr = &inp->in6p_laddr;
876 faddr = &inp->in6p_faddr;
877 fport = inp->inp_fport;
880 ulen = m->m_pkthdr.len;
881 plen = sizeof(struct udphdr) + ulen;
882 hlen = sizeof(struct ip6_hdr);
885 * Calculate data length and get a mbuf
886 * for UDP and IP6 headers.
888 M_PREPEND(m, hlen + sizeof(struct udphdr), M_NOWAIT);
895 * Stuff checksum and output datagram.
897 cscov = cscov_partial = 0;
898 udp6 = (struct udphdr *)(mtod(m, caddr_t) + hlen);
899 udp6->uh_sport = inp->inp_lport; /* lport is always set in the PCB */
900 udp6->uh_dport = fport;
901 if (nxt == IPPROTO_UDPLITE) {
905 cscov = up->u_txcslen;
908 udp6->uh_ulen = htons(cscov);
910 * For UDP-Lite, checksum coverage length of zero means
911 * the entire UDPLite packet is covered by the checksum.
913 cscov_partial = (cscov == 0) ? 0 : 1;
914 } else if (plen <= 0xffff)
915 udp6->uh_ulen = htons((u_short)plen);
920 ip6 = mtod(m, struct ip6_hdr *);
921 ip6->ip6_flow = inp->inp_flow & IPV6_FLOWINFO_MASK;
922 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
923 ip6->ip6_vfc |= IPV6_VERSION;
924 ip6->ip6_plen = htons((u_short)plen);
926 ip6->ip6_hlim = in6_selecthlim(inp, NULL);
927 ip6->ip6_src = *laddr;
928 ip6->ip6_dst = *faddr;
931 mac_inpcb_create_mbuf(inp, m);
935 if ((udp6->uh_sum = in6_cksum_partial(m, nxt,
936 sizeof(struct ip6_hdr), plen, cscov)) == 0)
937 udp6->uh_sum = 0xffff;
939 udp6->uh_sum = in6_cksum_pseudo(ip6, plen, nxt, 0);
940 m->m_pkthdr.csum_flags = CSUM_UDP_IPV6;
941 m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
947 uint32_t hash_val, hash_type;
950 pr = inp->inp_socket->so_proto->pr_protocol;
952 * Calculate an appropriate RSS hash for UDP and
955 * The called function will take care of figuring out
956 * whether a 2-tuple or 4-tuple hash is required based
957 * on the currently configured scheme.
959 * Later later on connected socket values should be
960 * cached in the inpcb and reused, rather than constantly
963 * UDP Lite is a different protocol number and will
964 * likely end up being hashed as a 2-tuple until
965 * RSS / NICs grow UDP Lite protocol awareness.
967 if (rss_proto_software_hash_v6(faddr, laddr, fport,
968 inp->inp_lport, pr, &hash_val, &hash_type) == 0) {
969 m->m_pkthdr.flowid = hash_val;
970 M_HASHTYPE_SET(m, hash_type);
974 * Don't override with the inp cached flowid.
976 * Until the whole UDP path is vetted, it may actually
979 flags |= IP_NODEFAULTFLOWID;
983 UDPSTAT_INC(udps_opackets);
984 if (nxt == IPPROTO_UDPLITE)
985 UDPLITE_PROBE(send, NULL, inp, ip6, inp, udp6);
987 UDP_PROBE(send, NULL, inp, ip6, inp, udp6);
988 error = ip6_output(m, optp,
989 INP_WLOCKED(inp) ? &inp->inp_route6 : NULL, flags,
990 inp->in6p_moptions, NULL, inp);
995 ip6_clearpktopts(&opt, -1);
1004 ip6_clearpktopts(&opt, -1);
1013 udp6_abort(struct socket *so)
1016 struct inpcbinfo *pcbinfo;
1018 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1019 inp = sotoinpcb(so);
1020 KASSERT(inp != NULL, ("udp6_abort: inp == NULL"));
1024 if (inp->inp_vflag & INP_IPV4) {
1025 struct pr_usrreqs *pru;
1028 nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ?
1029 IPPROTO_UDP : IPPROTO_UDPLITE;
1031 pru = inetsw[ip_protox[nxt]].pr_usrreqs;
1032 (*pru->pru_abort)(so);
1037 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1038 INP_HASH_WLOCK(pcbinfo);
1039 in6_pcbdisconnect(inp);
1040 inp->in6p_laddr = in6addr_any;
1041 INP_HASH_WUNLOCK(pcbinfo);
1042 soisdisconnected(so);
1048 udp6_attach(struct socket *so, int proto, struct thread *td)
1051 struct inpcbinfo *pcbinfo;
1054 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1055 inp = sotoinpcb(so);
1056 KASSERT(inp == NULL, ("udp6_attach: inp != NULL"));
1058 if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
1059 error = soreserve(so, udp_sendspace, udp_recvspace);
1063 INP_INFO_WLOCK(pcbinfo);
1064 error = in_pcballoc(so, pcbinfo);
1066 INP_INFO_WUNLOCK(pcbinfo);
1069 inp = (struct inpcb *)so->so_pcb;
1070 inp->inp_vflag |= INP_IPV6;
1071 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0)
1072 inp->inp_vflag |= INP_IPV4;
1073 inp->in6p_hops = -1; /* use kernel default */
1074 inp->in6p_cksum = -1; /* just to be sure */
1077 * IPv4 TTL initialization is necessary for an IPv6 socket as well,
1078 * because the socket may be bound to an IPv6 wildcard address,
1079 * which may match an IPv4-mapped IPv6 address.
1081 inp->inp_ip_ttl = V_ip_defttl;
1083 error = udp_newudpcb(inp);
1087 INP_INFO_WUNLOCK(pcbinfo);
1091 INP_INFO_WUNLOCK(pcbinfo);
1096 udp6_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
1099 struct inpcbinfo *pcbinfo;
1103 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1104 inp = sotoinpcb(so);
1105 KASSERT(inp != NULL, ("udp6_bind: inp == NULL"));
1108 INP_HASH_WLOCK(pcbinfo);
1109 vflagsav = inp->inp_vflag;
1110 inp->inp_vflag &= ~INP_IPV4;
1111 inp->inp_vflag |= INP_IPV6;
1112 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
1113 struct sockaddr_in6 *sin6_p;
1115 sin6_p = (struct sockaddr_in6 *)nam;
1117 if (IN6_IS_ADDR_UNSPECIFIED(&sin6_p->sin6_addr))
1118 inp->inp_vflag |= INP_IPV4;
1120 else if (IN6_IS_ADDR_V4MAPPED(&sin6_p->sin6_addr)) {
1121 struct sockaddr_in sin;
1123 in6_sin6_2_sin(&sin, sin6_p);
1124 inp->inp_vflag |= INP_IPV4;
1125 inp->inp_vflag &= ~INP_IPV6;
1126 error = in_pcbbind(inp, (struct sockaddr *)&sin,
1133 error = in6_pcbbind(inp, nam, td->td_ucred);
1138 inp->inp_vflag = vflagsav;
1139 INP_HASH_WUNLOCK(pcbinfo);
1145 udp6_close(struct socket *so)
1148 struct inpcbinfo *pcbinfo;
1150 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1151 inp = sotoinpcb(so);
1152 KASSERT(inp != NULL, ("udp6_close: inp == NULL"));
1156 if (inp->inp_vflag & INP_IPV4) {
1157 struct pr_usrreqs *pru;
1160 nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ?
1161 IPPROTO_UDP : IPPROTO_UDPLITE;
1163 pru = inetsw[ip_protox[nxt]].pr_usrreqs;
1164 (*pru->pru_disconnect)(so);
1168 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1169 INP_HASH_WLOCK(pcbinfo);
1170 in6_pcbdisconnect(inp);
1171 inp->in6p_laddr = in6addr_any;
1172 INP_HASH_WUNLOCK(pcbinfo);
1173 soisdisconnected(so);
1179 udp6_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
1182 struct inpcbinfo *pcbinfo;
1183 struct sockaddr_in6 *sin6;
1187 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1188 inp = sotoinpcb(so);
1189 sin6 = (struct sockaddr_in6 *)nam;
1190 KASSERT(inp != NULL, ("udp6_connect: inp == NULL"));
1193 * XXXRW: Need to clarify locking of v4/v6 flags.
1197 if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
1198 struct sockaddr_in sin;
1200 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) {
1204 if ((inp->inp_vflag & INP_IPV4) == 0) {
1205 error = EAFNOSUPPORT;
1208 if (inp->inp_faddr.s_addr != INADDR_ANY) {
1212 in6_sin6_2_sin(&sin, sin6);
1213 error = prison_remote_ip4(td->td_ucred, &sin.sin_addr);
1216 vflagsav = inp->inp_vflag;
1217 inp->inp_vflag |= INP_IPV4;
1218 inp->inp_vflag &= ~INP_IPV6;
1219 INP_HASH_WLOCK(pcbinfo);
1220 error = in_pcbconnect(inp, (struct sockaddr *)&sin,
1222 INP_HASH_WUNLOCK(pcbinfo);
1224 * If connect succeeds, mark socket as connected. If
1225 * connect fails and socket is unbound, reset inp_vflag
1230 else if (inp->inp_laddr.s_addr == INADDR_ANY &&
1231 inp->inp_lport == 0)
1232 inp->inp_vflag = vflagsav;
1235 if ((inp->inp_vflag & INP_IPV6) == 0) {
1236 error = EAFNOSUPPORT;
1241 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1245 error = prison_remote_ip6(td->td_ucred, &sin6->sin6_addr);
1248 vflagsav = inp->inp_vflag;
1249 inp->inp_vflag &= ~INP_IPV4;
1250 inp->inp_vflag |= INP_IPV6;
1251 INP_HASH_WLOCK(pcbinfo);
1252 error = in6_pcbconnect(inp, nam, td->td_ucred);
1253 INP_HASH_WUNLOCK(pcbinfo);
1255 * If connect succeeds, mark socket as connected. If
1256 * connect fails and socket is unbound, reset inp_vflag
1261 else if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
1262 inp->inp_lport == 0)
1263 inp->inp_vflag = vflagsav;
1270 udp6_detach(struct socket *so)
1273 struct inpcbinfo *pcbinfo;
1276 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1277 inp = sotoinpcb(so);
1278 KASSERT(inp != NULL, ("udp6_detach: inp == NULL"));
1280 INP_INFO_WLOCK(pcbinfo);
1282 up = intoudpcb(inp);
1283 KASSERT(up != NULL, ("%s: up == NULL", __func__));
1286 INP_INFO_WUNLOCK(pcbinfo);
1291 udp6_disconnect(struct socket *so)
1294 struct inpcbinfo *pcbinfo;
1296 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1297 inp = sotoinpcb(so);
1298 KASSERT(inp != NULL, ("udp6_disconnect: inp == NULL"));
1302 if (inp->inp_vflag & INP_IPV4) {
1303 struct pr_usrreqs *pru;
1306 nxt = (inp->inp_socket->so_proto->pr_protocol == IPPROTO_UDP) ?
1307 IPPROTO_UDP : IPPROTO_UDPLITE;
1309 pru = inetsw[ip_protox[nxt]].pr_usrreqs;
1310 (void)(*pru->pru_disconnect)(so);
1315 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1320 INP_HASH_WLOCK(pcbinfo);
1321 in6_pcbdisconnect(inp);
1322 inp->in6p_laddr = in6addr_any;
1323 INP_HASH_WUNLOCK(pcbinfo);
1325 so->so_state &= ~SS_ISCONNECTED; /* XXX */
1332 udp6_send(struct socket *so, int flags, struct mbuf *m,
1333 struct sockaddr *addr, struct mbuf *control, struct thread *td)
1338 if (addr->sa_len != sizeof(struct sockaddr_in6)) {
1342 if (addr->sa_family != AF_INET6) {
1343 error = EAFNOSUPPORT;
1348 return (udp6_output(so, flags, m, addr, control, td));
1357 struct pr_usrreqs udp6_usrreqs = {
1358 .pru_abort = udp6_abort,
1359 .pru_attach = udp6_attach,
1360 .pru_bind = udp6_bind,
1361 .pru_connect = udp6_connect,
1362 .pru_control = in6_control,
1363 .pru_detach = udp6_detach,
1364 .pru_disconnect = udp6_disconnect,
1365 .pru_peeraddr = in6_mapped_peeraddr,
1366 .pru_send = udp6_send,
1367 .pru_shutdown = udp_shutdown,
1368 .pru_sockaddr = in6_mapped_sockaddr,
1369 .pru_soreceive = soreceive_dgram,
1370 .pru_sosend = sosend_dgram,
1371 .pru_sosetlabel = in_pcbsosetlabel,
1372 .pru_close = udp6_close