2 * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995
3 * The Regents of the University of California.
4 * Copyright (c) 2008 Robert N. M. Watson
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 * 4. Neither the name of the University 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 REGENTS 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 REGENTS 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 * @(#)udp_usrreq.c 8.6 (Berkeley) 5/23/95
39 #include <sys/cdefs.h>
40 __FBSDID("$FreeBSD$");
43 #include "opt_inet6.h"
44 #include "opt_ipsec.h"
47 #include <sys/param.h>
48 #include <sys/domain.h>
49 #include <sys/eventhandler.h>
51 #include <sys/kernel.h>
53 #include <sys/malloc.h>
57 #include <sys/protosw.h>
59 #include <sys/signalvar.h>
60 #include <sys/socket.h>
61 #include <sys/socketvar.h>
63 #include <sys/sysctl.h>
64 #include <sys/syslog.h>
65 #include <sys/systm.h>
70 #include <net/if_var.h>
71 #include <net/route.h>
72 #include <net/rss_config.h>
74 #include <netinet/in.h>
75 #include <netinet/in_kdtrace.h>
76 #include <netinet/in_pcb.h>
77 #include <netinet/in_systm.h>
78 #include <netinet/in_var.h>
79 #include <netinet/ip.h>
81 #include <netinet/ip6.h>
83 #include <netinet/ip_icmp.h>
84 #include <netinet/icmp_var.h>
85 #include <netinet/ip_var.h>
86 #include <netinet/ip_options.h>
88 #include <netinet6/ip6_var.h>
90 #include <netinet/udp.h>
91 #include <netinet/udp_var.h>
92 #include <netinet/udplite.h>
93 #include <netinet/in_rss.h>
95 #include <netipsec/ipsec_support.h>
97 #include <machine/in_cksum.h>
99 #include <security/mac/mac_framework.h>
102 * UDP and UDP-Lite protocols implementation.
103 * Per RFC 768, August, 1980.
104 * Per RFC 3828, July, 2004.
108 * BSD 4.2 defaulted the udp checksum to be off. Turning off udp checksums
109 * removes the only data integrity mechanism for packets and malformed
110 * packets that would otherwise be discarded due to bad checksums, and may
111 * cause problems (especially for NFS data blocks).
113 VNET_DEFINE(int, udp_cksum) = 1;
114 SYSCTL_INT(_net_inet_udp, UDPCTL_CHECKSUM, checksum, CTLFLAG_VNET | CTLFLAG_RW,
115 &VNET_NAME(udp_cksum), 0, "compute udp checksum");
117 int udp_log_in_vain = 0;
118 SYSCTL_INT(_net_inet_udp, OID_AUTO, log_in_vain, CTLFLAG_RW,
119 &udp_log_in_vain, 0, "Log all incoming UDP packets");
121 VNET_DEFINE(int, udp_blackhole) = 0;
122 SYSCTL_INT(_net_inet_udp, OID_AUTO, blackhole, CTLFLAG_VNET | CTLFLAG_RW,
123 &VNET_NAME(udp_blackhole), 0,
124 "Do not send port unreachables for refused connects");
126 static VNET_DEFINE(int, udp_require_l2_bcast) = 0;
127 #define V_udp_require_l2_bcast VNET(udp_require_l2_bcast)
128 SYSCTL_INT(_net_inet_udp, OID_AUTO, require_l2_bcast, CTLFLAG_VNET | CTLFLAG_RW,
129 &VNET_NAME(udp_require_l2_bcast), 0,
130 "Only treat packets sent to an L2 broadcast address as broadcast packets");
132 u_long udp_sendspace = 9216; /* really max datagram size */
133 SYSCTL_ULONG(_net_inet_udp, UDPCTL_MAXDGRAM, maxdgram, CTLFLAG_RW,
134 &udp_sendspace, 0, "Maximum outgoing UDP datagram size");
136 u_long udp_recvspace = 40 * (1024 +
138 sizeof(struct sockaddr_in6)
140 sizeof(struct sockaddr_in)
142 ); /* 40 1K datagrams */
144 SYSCTL_ULONG(_net_inet_udp, UDPCTL_RECVSPACE, recvspace, CTLFLAG_RW,
145 &udp_recvspace, 0, "Maximum space for incoming UDP datagrams");
147 VNET_DEFINE(struct inpcbhead, udb); /* from udp_var.h */
148 VNET_DEFINE(struct inpcbinfo, udbinfo);
149 VNET_DEFINE(struct inpcbhead, ulitecb);
150 VNET_DEFINE(struct inpcbinfo, ulitecbinfo);
151 static VNET_DEFINE(uma_zone_t, udpcb_zone);
152 #define V_udpcb_zone VNET(udpcb_zone)
155 #define UDBHASHSIZE 128
158 VNET_PCPUSTAT_DEFINE(struct udpstat, udpstat); /* from udp_var.h */
159 VNET_PCPUSTAT_SYSINIT(udpstat);
160 SYSCTL_VNET_PCPUSTAT(_net_inet_udp, UDPCTL_STATS, stats, struct udpstat,
161 udpstat, "UDP statistics (struct udpstat, netinet/udp_var.h)");
164 VNET_PCPUSTAT_SYSUNINIT(udpstat);
167 static void udp_detach(struct socket *so);
168 static int udp_output(struct inpcb *, struct mbuf *, struct sockaddr *,
169 struct mbuf *, struct thread *);
173 udp_zone_change(void *tag)
176 uma_zone_set_max(V_udbinfo.ipi_zone, maxsockets);
177 uma_zone_set_max(V_udpcb_zone, maxsockets);
181 udp_inpcb_init(void *mem, int size, int flags)
186 INP_LOCK_INIT(inp, "inp", "udpinp");
191 udplite_inpcb_init(void *mem, int size, int flags)
196 INP_LOCK_INIT(inp, "inp", "udpliteinp");
205 * For now default to 2-tuple UDP hashing - until the fragment
206 * reassembly code can also update the flowid.
208 * Once we can calculate the flowid that way and re-establish
209 * a 4-tuple, flip this to 4-tuple.
211 in_pcbinfo_init(&V_udbinfo, "udp", &V_udb, UDBHASHSIZE, UDBHASHSIZE,
212 "udp_inpcb", udp_inpcb_init, NULL, 0,
213 IPI_HASHFIELDS_2TUPLE);
214 V_udpcb_zone = uma_zcreate("udpcb", sizeof(struct udpcb),
215 NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0);
216 uma_zone_set_max(V_udpcb_zone, maxsockets);
217 uma_zone_set_warning(V_udpcb_zone, "kern.ipc.maxsockets limit reached");
218 EVENTHANDLER_REGISTER(maxsockets_change, udp_zone_change, NULL,
219 EVENTHANDLER_PRI_ANY);
226 in_pcbinfo_init(&V_ulitecbinfo, "udplite", &V_ulitecb, UDBHASHSIZE,
227 UDBHASHSIZE, "udplite_inpcb", udplite_inpcb_init, NULL,
228 0, IPI_HASHFIELDS_2TUPLE);
232 * Kernel module interface for updating udpstat. The argument is an index
233 * into udpstat treated as an array of u_long. While this encodes the
234 * general layout of udpstat into the caller, it doesn't encode its location,
235 * so that future changes to add, for example, per-CPU stats support won't
236 * cause binary compatibility problems for kernel modules.
239 kmod_udpstat_inc(int statnum)
242 counter_u64_add(VNET(udpstat)[statnum], 1);
246 udp_newudpcb(struct inpcb *inp)
250 up = uma_zalloc(V_udpcb_zone, M_NOWAIT | M_ZERO);
258 udp_discardcb(struct udpcb *up)
261 uma_zfree(V_udpcb_zone, up);
266 udp_destroy(void *unused __unused)
269 in_pcbinfo_destroy(&V_udbinfo);
270 uma_zdestroy(V_udpcb_zone);
272 VNET_SYSUNINIT(udp, SI_SUB_PROTO_DOMAIN, SI_ORDER_FOURTH, udp_destroy, NULL);
275 udplite_destroy(void *unused __unused)
278 in_pcbinfo_destroy(&V_ulitecbinfo);
280 VNET_SYSUNINIT(udplite, SI_SUB_PROTO_DOMAIN, SI_ORDER_FOURTH, udplite_destroy,
286 * Subroutine of udp_input(), which appends the provided mbuf chain to the
287 * passed pcb/socket. The caller must provide a sockaddr_in via udp_in that
288 * contains the source address. If the socket ends up being an IPv6 socket,
289 * udp_append() will convert to a sockaddr_in6 before passing the address
290 * into the socket code.
292 * In the normal case udp_append() will return 0, indicating that you
293 * must unlock the inp. However if a tunneling protocol is in place we increment
294 * the inpcb refcnt and unlock the inp, on return from the tunneling protocol we
295 * then decrement the reference count. If the inp_rele returns 1, indicating the
296 * inp is gone, we return that to the caller to tell them *not* to unlock
297 * the inp. In the case of multi-cast this will cause the distribution
298 * to stop (though most tunneling protocols known currently do *not* use
302 udp_append(struct inpcb *inp, struct ip *ip, struct mbuf *n, int off,
303 struct sockaddr_in *udp_in)
305 struct sockaddr *append_sa;
307 struct mbuf *opts = NULL;
309 struct sockaddr_in6 udp_in6;
313 INP_LOCK_ASSERT(inp);
316 * Engage the tunneling protocol.
319 if (up->u_tun_func != NULL) {
322 (*up->u_tun_func)(n, off, inp, (struct sockaddr *)udp_in,
325 return (in_pcbrele_rlocked(inp));
328 off += sizeof(struct udphdr);
330 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
331 /* Check AH/ESP integrity. */
332 if (IPSEC_ENABLED(ipv4) &&
333 IPSEC_CHECK_POLICY(ipv4, n, inp) != 0) {
337 if (up->u_flags & UF_ESPINUDP) {/* IPSec UDP encaps. */
338 if (IPSEC_ENABLED(ipv4) &&
339 UDPENCAP_INPUT(n, off, AF_INET) != 0)
340 return (0); /* Consumed. */
344 if (mac_inpcb_check_deliver(inp, n) != 0) {
349 if (inp->inp_flags & INP_CONTROLOPTS ||
350 inp->inp_socket->so_options & (SO_TIMESTAMP | SO_BINTIME)) {
352 if (inp->inp_vflag & INP_IPV6)
353 (void)ip6_savecontrol_v4(inp, n, &opts, NULL);
356 ip_savecontrol(inp, &opts, ip, n);
359 if (inp->inp_vflag & INP_IPV6) {
360 bzero(&udp_in6, sizeof(udp_in6));
361 udp_in6.sin6_len = sizeof(udp_in6);
362 udp_in6.sin6_family = AF_INET6;
363 in6_sin_2_v4mapsin6(udp_in, &udp_in6);
364 append_sa = (struct sockaddr *)&udp_in6;
367 append_sa = (struct sockaddr *)udp_in;
370 so = inp->inp_socket;
371 SOCKBUF_LOCK(&so->so_rcv);
372 if (sbappendaddr_locked(&so->so_rcv, append_sa, n, opts) == 0) {
373 SOCKBUF_UNLOCK(&so->so_rcv);
377 UDPSTAT_INC(udps_fullsock);
379 sorwakeup_locked(so);
384 udp_input(struct mbuf **mp, int *offp, int proto)
390 uint16_t len, ip_len;
391 struct inpcbinfo *pcbinfo;
393 struct sockaddr_in udp_in;
395 struct m_tag *fwd_tag;
396 int cscov_partial, iphlen;
400 ifp = m->m_pkthdr.rcvif;
402 UDPSTAT_INC(udps_ipackets);
405 * Strip IP options, if any; should skip this, make available to
406 * user, and use on returned packets, but we don't yet have a way to
407 * check the checksum with options still present.
409 if (iphlen > sizeof (struct ip)) {
411 iphlen = sizeof(struct ip);
415 * Get IP and UDP header together in first mbuf.
417 ip = mtod(m, struct ip *);
418 if (m->m_len < iphlen + sizeof(struct udphdr)) {
419 if ((m = m_pullup(m, iphlen + sizeof(struct udphdr))) == NULL) {
420 UDPSTAT_INC(udps_hdrops);
421 return (IPPROTO_DONE);
423 ip = mtod(m, struct ip *);
425 uh = (struct udphdr *)((caddr_t)ip + iphlen);
426 cscov_partial = (proto == IPPROTO_UDPLITE) ? 1 : 0;
429 * Destination port of 0 is illegal, based on RFC768.
431 if (uh->uh_dport == 0)
435 * Construct sockaddr format source address. Stuff source address
436 * and datagram in user buffer.
438 bzero(&udp_in, sizeof(udp_in));
439 udp_in.sin_len = sizeof(udp_in);
440 udp_in.sin_family = AF_INET;
441 udp_in.sin_port = uh->uh_sport;
442 udp_in.sin_addr = ip->ip_src;
445 * Make mbuf data length reflect UDP length. If not enough data to
446 * reflect UDP length, drop.
448 len = ntohs((u_short)uh->uh_ulen);
449 ip_len = ntohs(ip->ip_len) - iphlen;
450 if (proto == IPPROTO_UDPLITE && (len == 0 || len == ip_len)) {
451 /* Zero means checksum over the complete packet. */
457 if (len > ip_len || len < sizeof(struct udphdr)) {
458 UDPSTAT_INC(udps_badlen);
461 if (proto == IPPROTO_UDP)
462 m_adj(m, len - ip_len);
466 * Save a copy of the IP header in case we want restore it for
467 * sending an ICMP error message in response.
469 if (!V_udp_blackhole)
472 memset(&save_ip, 0, sizeof(save_ip));
475 * Checksum extended UDP header and data.
480 if ((m->m_pkthdr.csum_flags & CSUM_DATA_VALID) &&
482 if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR)
483 uh_sum = m->m_pkthdr.csum_data;
485 uh_sum = in_pseudo(ip->ip_src.s_addr,
486 ip->ip_dst.s_addr, htonl((u_short)len +
487 m->m_pkthdr.csum_data + proto));
492 bcopy(((struct ipovly *)ip)->ih_x1, b, 9);
493 bzero(((struct ipovly *)ip)->ih_x1, 9);
494 ((struct ipovly *)ip)->ih_len = (proto == IPPROTO_UDP) ?
495 uh->uh_ulen : htons(ip_len);
496 uh_sum = in_cksum(m, len + sizeof (struct ip));
497 bcopy(b, ((struct ipovly *)ip)->ih_x1, 9);
500 UDPSTAT_INC(udps_badsum);
502 return (IPPROTO_DONE);
505 if (proto == IPPROTO_UDP) {
506 UDPSTAT_INC(udps_nosum);
508 /* UDPLite requires a checksum */
509 /* XXX: What is the right UDPLite MIB counter here? */
511 return (IPPROTO_DONE);
515 pcbinfo = udp_get_inpcbinfo(proto);
516 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) ||
517 ((!V_udp_require_l2_bcast || m->m_flags & M_BCAST) &&
518 in_broadcast(ip->ip_dst, ifp))) {
520 struct inpcbhead *pcblist;
521 struct ip_moptions *imo;
523 INP_INFO_RLOCK(pcbinfo);
524 pcblist = udp_get_pcblist(proto);
526 LIST_FOREACH(inp, pcblist, inp_list) {
527 if (inp->inp_lport != uh->uh_dport)
530 if ((inp->inp_vflag & INP_IPV4) == 0)
533 if (inp->inp_laddr.s_addr != INADDR_ANY &&
534 inp->inp_laddr.s_addr != ip->ip_dst.s_addr)
536 if (inp->inp_faddr.s_addr != INADDR_ANY &&
537 inp->inp_faddr.s_addr != ip->ip_src.s_addr)
539 if (inp->inp_fport != 0 &&
540 inp->inp_fport != uh->uh_sport)
546 * XXXRW: Because we weren't holding either the inpcb
547 * or the hash lock when we checked for a match
548 * before, we should probably recheck now that the
549 * inpcb lock is held.
553 * Handle socket delivery policy for any-source
554 * and source-specific multicast. [RFC3678]
556 imo = inp->inp_moptions;
557 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
558 struct sockaddr_in group;
564 bzero(&group, sizeof(struct sockaddr_in));
565 group.sin_len = sizeof(struct sockaddr_in);
566 group.sin_family = AF_INET;
567 group.sin_addr = ip->ip_dst;
569 blocked = imo_multi_filter(imo, ifp,
570 (struct sockaddr *)&group,
571 (struct sockaddr *)&udp_in);
572 if (blocked != MCAST_PASS) {
573 if (blocked == MCAST_NOTGMEMBER)
574 IPSTAT_INC(ips_notmember);
575 if (blocked == MCAST_NOTSMEMBER ||
576 blocked == MCAST_MUTED)
577 UDPSTAT_INC(udps_filtermcast);
585 if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) !=
587 UDP_PROBE(receive, NULL, last, ip,
589 if (udp_append(last, ip, n, iphlen,
598 * Don't look for additional matches if this one does
599 * not have either the SO_REUSEPORT or SO_REUSEADDR
600 * socket options set. This heuristic avoids
601 * searching through all pcbs in the common case of a
602 * non-shared port. It assumes that an application
603 * will never clear these options after setting them.
605 if ((last->inp_socket->so_options &
606 (SO_REUSEPORT|SO_REUSEADDR)) == 0)
612 * No matching pcb found; discard datagram. (No need
613 * to send an ICMP Port Unreachable for a broadcast
614 * or multicast datgram.)
616 UDPSTAT_INC(udps_noportbcast);
619 INP_INFO_RUNLOCK(pcbinfo);
622 UDP_PROBE(receive, NULL, last, ip, last, uh);
623 if (udp_append(last, ip, m, iphlen, &udp_in) == 0)
626 INP_INFO_RUNLOCK(pcbinfo);
627 return (IPPROTO_DONE);
631 * Locate pcb for datagram.
635 * Grab info from PACKET_TAG_IPFORWARD tag prepended to the chain.
637 if ((m->m_flags & M_IP_NEXTHOP) &&
638 (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
639 struct sockaddr_in *next_hop;
641 next_hop = (struct sockaddr_in *)(fwd_tag + 1);
644 * Transparently forwarded. Pretend to be the destination.
645 * Already got one like this?
647 inp = in_pcblookup_mbuf(pcbinfo, ip->ip_src, uh->uh_sport,
648 ip->ip_dst, uh->uh_dport, INPLOOKUP_RLOCKPCB, ifp, m);
651 * It's new. Try to find the ambushing socket.
652 * Because we've rewritten the destination address,
653 * any hardware-generated hash is ignored.
655 inp = in_pcblookup(pcbinfo, ip->ip_src,
656 uh->uh_sport, next_hop->sin_addr,
657 next_hop->sin_port ? htons(next_hop->sin_port) :
658 uh->uh_dport, INPLOOKUP_WILDCARD |
659 INPLOOKUP_RLOCKPCB, ifp);
661 /* Remove the tag from the packet. We don't need it anymore. */
662 m_tag_delete(m, fwd_tag);
663 m->m_flags &= ~M_IP_NEXTHOP;
665 inp = in_pcblookup_mbuf(pcbinfo, ip->ip_src, uh->uh_sport,
666 ip->ip_dst, uh->uh_dport, INPLOOKUP_WILDCARD |
667 INPLOOKUP_RLOCKPCB, ifp, m);
669 if (udp_log_in_vain) {
670 char buf[4*sizeof "123"];
672 strcpy(buf, inet_ntoa(ip->ip_dst));
674 "Connection attempt to UDP %s:%d from %s:%d\n",
675 buf, ntohs(uh->uh_dport), inet_ntoa(ip->ip_src),
676 ntohs(uh->uh_sport));
678 UDPSTAT_INC(udps_noport);
679 if (m->m_flags & (M_BCAST | M_MCAST)) {
680 UDPSTAT_INC(udps_noportbcast);
685 if (badport_bandlim(BANDLIM_ICMP_UNREACH) < 0)
688 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0);
689 return (IPPROTO_DONE);
693 * Check the minimum TTL for socket.
695 INP_RLOCK_ASSERT(inp);
696 if (inp->inp_ip_minttl && inp->inp_ip_minttl > ip->ip_ttl) {
699 return (IPPROTO_DONE);
705 if (up->u_rxcslen == 0 || up->u_rxcslen > len) {
708 return (IPPROTO_DONE);
712 UDP_PROBE(receive, NULL, inp, ip, inp, uh);
713 if (udp_append(inp, ip, m, iphlen, &udp_in) == 0)
715 return (IPPROTO_DONE);
719 return (IPPROTO_DONE);
724 * Notify a udp user of an asynchronous error; just wake up so that they can
725 * collect error status.
728 udp_notify(struct inpcb *inp, int errno)
732 * While udp_ctlinput() always calls udp_notify() with a read lock
733 * when invoking it directly, in_pcbnotifyall() currently uses write
734 * locks due to sharing code with TCP. For now, accept either a read
735 * or a write lock, but a read lock is sufficient.
737 INP_LOCK_ASSERT(inp);
738 if ((errno == EHOSTUNREACH || errno == ENETUNREACH ||
739 errno == EHOSTDOWN) && inp->inp_route.ro_rt) {
740 RTFREE(inp->inp_route.ro_rt);
741 inp->inp_route.ro_rt = (struct rtentry *)NULL;
744 inp->inp_socket->so_error = errno;
745 sorwakeup(inp->inp_socket);
746 sowwakeup(inp->inp_socket);
752 udp_common_ctlinput(int cmd, struct sockaddr *sa, void *vip,
753 struct inpcbinfo *pcbinfo)
757 struct in_addr faddr;
760 faddr = ((struct sockaddr_in *)sa)->sin_addr;
761 if (sa->sa_family != AF_INET || faddr.s_addr == INADDR_ANY)
764 if (PRC_IS_REDIRECT(cmd)) {
765 /* signal EHOSTDOWN, as it flushes the cached route */
766 in_pcbnotifyall(&V_udbinfo, faddr, EHOSTDOWN, udp_notify);
771 * Hostdead is ugly because it goes linearly through all PCBs.
773 * XXX: We never get this from ICMP, otherwise it makes an excellent
774 * DoS attack on machines with many connections.
776 if (cmd == PRC_HOSTDEAD)
778 else if ((unsigned)cmd >= PRC_NCMDS || inetctlerrmap[cmd] == 0)
781 uh = (struct udphdr *)((caddr_t)ip + (ip->ip_hl << 2));
782 inp = in_pcblookup(pcbinfo, faddr, uh->uh_dport,
783 ip->ip_src, uh->uh_sport, INPLOOKUP_RLOCKPCB, NULL);
785 INP_RLOCK_ASSERT(inp);
786 if (inp->inp_socket != NULL) {
787 udp_notify(inp, inetctlerrmap[cmd]);
791 inp = in_pcblookup(pcbinfo, faddr, uh->uh_dport,
792 ip->ip_src, uh->uh_sport,
793 INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB, NULL);
798 if (up->u_icmp_func != NULL) {
800 (*up->u_icmp_func)(cmd, sa, vip, up->u_tun_ctx);
807 in_pcbnotifyall(pcbinfo, faddr, inetctlerrmap[cmd],
811 udp_ctlinput(int cmd, struct sockaddr *sa, void *vip)
814 return (udp_common_ctlinput(cmd, sa, vip, &V_udbinfo));
818 udplite_ctlinput(int cmd, struct sockaddr *sa, void *vip)
821 return (udp_common_ctlinput(cmd, sa, vip, &V_ulitecbinfo));
826 udp_pcblist(SYSCTL_HANDLER_ARGS)
829 struct inpcb *inp, **inp_list;
834 * The process of preparing the PCB list is too time-consuming and
835 * resource-intensive to repeat twice on every request.
837 if (req->oldptr == 0) {
838 n = V_udbinfo.ipi_count;
839 n += imax(n / 8, 10);
840 req->oldidx = 2 * (sizeof xig) + n * sizeof(struct xinpcb);
844 if (req->newptr != 0)
848 * OK, now we're committed to doing something.
850 INP_INFO_RLOCK(&V_udbinfo);
851 gencnt = V_udbinfo.ipi_gencnt;
852 n = V_udbinfo.ipi_count;
853 INP_INFO_RUNLOCK(&V_udbinfo);
855 error = sysctl_wire_old_buffer(req, 2 * (sizeof xig)
856 + n * sizeof(struct xinpcb));
860 xig.xig_len = sizeof xig;
862 xig.xig_gen = gencnt;
863 xig.xig_sogen = so_gencnt;
864 error = SYSCTL_OUT(req, &xig, sizeof xig);
868 inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK);
869 if (inp_list == NULL)
872 INP_INFO_RLOCK(&V_udbinfo);
873 for (inp = LIST_FIRST(V_udbinfo.ipi_listhead), i = 0; inp && i < n;
874 inp = LIST_NEXT(inp, inp_list)) {
876 if (inp->inp_gencnt <= gencnt &&
877 cr_canseeinpcb(req->td->td_ucred, inp) == 0) {
883 INP_INFO_RUNLOCK(&V_udbinfo);
887 for (i = 0; i < n; i++) {
890 if (inp->inp_gencnt <= gencnt) {
893 bzero(&xi, sizeof(xi));
894 xi.xi_len = sizeof xi;
895 /* XXX should avoid extra copy */
896 bcopy(inp, &xi.xi_inp, sizeof *inp);
898 sotoxsocket(inp->inp_socket, &xi.xi_socket);
899 xi.xi_inp.inp_gencnt = inp->inp_gencnt;
901 error = SYSCTL_OUT(req, &xi, sizeof xi);
905 INP_INFO_WLOCK(&V_udbinfo);
906 for (i = 0; i < n; i++) {
909 if (!in_pcbrele_rlocked(inp))
912 INP_INFO_WUNLOCK(&V_udbinfo);
916 * Give the user an updated idea of our state. If the
917 * generation differs from what we told her before, she knows
918 * that something happened while we were processing this
919 * request, and it might be necessary to retry.
921 INP_INFO_RLOCK(&V_udbinfo);
922 xig.xig_gen = V_udbinfo.ipi_gencnt;
923 xig.xig_sogen = so_gencnt;
924 xig.xig_count = V_udbinfo.ipi_count;
925 INP_INFO_RUNLOCK(&V_udbinfo);
926 error = SYSCTL_OUT(req, &xig, sizeof xig);
928 free(inp_list, M_TEMP);
932 SYSCTL_PROC(_net_inet_udp, UDPCTL_PCBLIST, pcblist,
933 CTLTYPE_OPAQUE | CTLFLAG_RD, NULL, 0,
934 udp_pcblist, "S,xinpcb", "List of active UDP sockets");
938 udp_getcred(SYSCTL_HANDLER_ARGS)
941 struct sockaddr_in addrs[2];
945 error = priv_check(req->td, PRIV_NETINET_GETCRED);
948 error = SYSCTL_IN(req, addrs, sizeof(addrs));
951 inp = in_pcblookup(&V_udbinfo, addrs[1].sin_addr, addrs[1].sin_port,
952 addrs[0].sin_addr, addrs[0].sin_port,
953 INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB, NULL);
955 INP_RLOCK_ASSERT(inp);
956 if (inp->inp_socket == NULL)
959 error = cr_canseeinpcb(req->td->td_ucred, inp);
961 cru2x(inp->inp_cred, &xuc);
966 error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred));
970 SYSCTL_PROC(_net_inet_udp, OID_AUTO, getcred,
971 CTLTYPE_OPAQUE|CTLFLAG_RW|CTLFLAG_PRISON, 0, 0,
972 udp_getcred, "S,xucred", "Get the xucred of a UDP connection");
976 udp_ctloutput(struct socket *so, struct sockopt *sopt)
980 int isudplite, error, optval;
983 isudplite = (so->so_proto->pr_protocol == IPPROTO_UDPLITE) ? 1 : 0;
985 KASSERT(inp != NULL, ("%s: inp == NULL", __func__));
987 if (sopt->sopt_level != so->so_proto->pr_protocol) {
989 if (INP_CHECK_SOCKAF(so, AF_INET6)) {
991 error = ip6_ctloutput(so, sopt);
994 #if defined(INET) && defined(INET6)
1000 error = ip_ctloutput(so, sopt);
1006 switch (sopt->sopt_dir) {
1008 switch (sopt->sopt_name) {
1009 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1012 if (!IPSEC_ENABLED(ipv4)) {
1014 return (ENOPROTOOPT);
1016 error = UDPENCAP_PCBCTL(inp, sopt);
1020 case UDPLITE_SEND_CSCOV:
1021 case UDPLITE_RECV_CSCOV:
1024 error = ENOPROTOOPT;
1028 error = sooptcopyin(sopt, &optval, sizeof(optval),
1032 inp = sotoinpcb(so);
1033 KASSERT(inp != NULL, ("%s: inp == NULL", __func__));
1035 up = intoudpcb(inp);
1036 KASSERT(up != NULL, ("%s: up == NULL", __func__));
1037 if ((optval != 0 && optval < 8) || (optval > 65535)) {
1042 if (sopt->sopt_name == UDPLITE_SEND_CSCOV)
1043 up->u_txcslen = optval;
1045 up->u_rxcslen = optval;
1050 error = ENOPROTOOPT;
1055 switch (sopt->sopt_name) {
1056 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1059 if (!IPSEC_ENABLED(ipv4)) {
1061 return (ENOPROTOOPT);
1063 error = UDPENCAP_PCBCTL(inp, sopt);
1067 case UDPLITE_SEND_CSCOV:
1068 case UDPLITE_RECV_CSCOV:
1071 error = ENOPROTOOPT;
1074 up = intoudpcb(inp);
1075 KASSERT(up != NULL, ("%s: up == NULL", __func__));
1076 if (sopt->sopt_name == UDPLITE_SEND_CSCOV)
1077 optval = up->u_txcslen;
1079 optval = up->u_rxcslen;
1081 error = sooptcopyout(sopt, &optval, sizeof(optval));
1085 error = ENOPROTOOPT;
1094 #define UH_WLOCKED 2
1095 #define UH_RLOCKED 1
1096 #define UH_UNLOCKED 0
1098 udp_output(struct inpcb *inp, struct mbuf *m, struct sockaddr *addr,
1099 struct mbuf *control, struct thread *td)
1101 struct udpiphdr *ui;
1102 int len = m->m_pkthdr.len;
1103 struct in_addr faddr, laddr;
1105 struct inpcbinfo *pcbinfo;
1106 struct sockaddr_in *sin, src;
1107 int cscov_partial = 0;
1110 u_short fport, lport;
1111 int unlock_udbinfo, unlock_inp;
1115 uint32_t flowid = 0;
1116 uint8_t flowtype = M_HASHTYPE_NONE;
1119 * udp_output() may need to temporarily bind or connect the current
1120 * inpcb. As such, we don't know up front whether we will need the
1121 * pcbinfo lock or not. Do any work to decide what is needed up
1122 * front before acquiring any locks.
1124 if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) {
1132 sin = (struct sockaddr_in *)addr;
1134 (inp->inp_laddr.s_addr == INADDR_ANY && inp->inp_lport == 0)) {
1136 unlock_inp = UH_WLOCKED;
1139 unlock_inp = UH_RLOCKED;
1141 tos = inp->inp_ip_tos;
1142 if (control != NULL) {
1144 * XXX: Currently, we assume all the optional information is
1145 * stored in a single mbuf.
1147 if (control->m_next) {
1148 if (unlock_inp == UH_WLOCKED)
1156 for (; control->m_len > 0;
1157 control->m_data += CMSG_ALIGN(cm->cmsg_len),
1158 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
1159 cm = mtod(control, struct cmsghdr *);
1160 if (control->m_len < sizeof(*cm) || cm->cmsg_len == 0
1161 || cm->cmsg_len > control->m_len) {
1165 if (cm->cmsg_level != IPPROTO_IP)
1168 switch (cm->cmsg_type) {
1169 case IP_SENDSRCADDR:
1171 CMSG_LEN(sizeof(struct in_addr))) {
1175 bzero(&src, sizeof(src));
1176 src.sin_family = AF_INET;
1177 src.sin_len = sizeof(src);
1178 src.sin_port = inp->inp_lport;
1180 *(struct in_addr *)CMSG_DATA(cm);
1184 if (cm->cmsg_len != CMSG_LEN(sizeof(u_char))) {
1188 tos = *(u_char *)CMSG_DATA(cm);
1192 if (cm->cmsg_len != CMSG_LEN(sizeof(uint32_t))) {
1196 flowid = *(uint32_t *) CMSG_DATA(cm);
1200 if (cm->cmsg_len != CMSG_LEN(sizeof(uint32_t))) {
1204 flowtype = *(uint32_t *) CMSG_DATA(cm);
1208 case IP_RSSBUCKETID:
1209 if (cm->cmsg_len != CMSG_LEN(sizeof(uint32_t))) {
1213 /* This is just a placeholder for now */
1217 error = ENOPROTOOPT;
1226 if (unlock_inp == UH_WLOCKED)
1235 * Depending on whether or not the application has bound or connected
1236 * the socket, we may have to do varying levels of work. The optimal
1237 * case is for a connected UDP socket, as a global lock isn't
1240 * In order to decide which we need, we require stability of the
1241 * inpcb binding, which we ensure by acquiring a read lock on the
1242 * inpcb. This doesn't strictly follow the lock order, so we play
1243 * the trylock and retry game; note that we may end up with more
1244 * conservative locks than required the second time around, so later
1245 * assertions have to accept that. Further analysis of the number of
1246 * misses under contention is required.
1248 * XXXRW: Check that hash locking update here is correct.
1250 pr = inp->inp_socket->so_proto->pr_protocol;
1251 pcbinfo = udp_get_inpcbinfo(pr);
1252 sin = (struct sockaddr_in *)addr;
1254 (inp->inp_laddr.s_addr == INADDR_ANY && inp->inp_lport == 0)) {
1255 INP_HASH_WLOCK(pcbinfo);
1256 unlock_udbinfo = UH_WLOCKED;
1257 } else if ((sin != NULL && (
1258 (sin->sin_addr.s_addr == INADDR_ANY) ||
1259 (sin->sin_addr.s_addr == INADDR_BROADCAST) ||
1260 (inp->inp_laddr.s_addr == INADDR_ANY) ||
1261 (inp->inp_lport == 0))) ||
1262 (src.sin_family == AF_INET)) {
1263 INP_HASH_RLOCK(pcbinfo);
1264 unlock_udbinfo = UH_RLOCKED;
1266 unlock_udbinfo = UH_UNLOCKED;
1269 * If the IP_SENDSRCADDR control message was specified, override the
1270 * source address for this datagram. Its use is invalidated if the
1271 * address thus specified is incomplete or clobbers other inpcbs.
1273 laddr = inp->inp_laddr;
1274 lport = inp->inp_lport;
1275 if (src.sin_family == AF_INET) {
1276 INP_HASH_LOCK_ASSERT(pcbinfo);
1278 (laddr.s_addr == INADDR_ANY &&
1279 src.sin_addr.s_addr == INADDR_ANY)) {
1283 error = in_pcbbind_setup(inp, (struct sockaddr *)&src,
1284 &laddr.s_addr, &lport, td->td_ucred);
1290 * If a UDP socket has been connected, then a local address/port will
1291 * have been selected and bound.
1293 * If a UDP socket has not been connected to, then an explicit
1294 * destination address must be used, in which case a local
1295 * address/port may not have been selected and bound.
1298 INP_LOCK_ASSERT(inp);
1299 if (inp->inp_faddr.s_addr != INADDR_ANY) {
1305 * Jail may rewrite the destination address, so let it do
1306 * that before we use it.
1308 error = prison_remote_ip4(td->td_ucred, &sin->sin_addr);
1313 * If a local address or port hasn't yet been selected, or if
1314 * the destination address needs to be rewritten due to using
1315 * a special INADDR_ constant, invoke in_pcbconnect_setup()
1316 * to do the heavy lifting. Once a port is selected, we
1317 * commit the binding back to the socket; we also commit the
1318 * binding of the address if in jail.
1320 * If we already have a valid binding and we're not
1321 * requesting a destination address rewrite, use a fast path.
1323 if (inp->inp_laddr.s_addr == INADDR_ANY ||
1324 inp->inp_lport == 0 ||
1325 sin->sin_addr.s_addr == INADDR_ANY ||
1326 sin->sin_addr.s_addr == INADDR_BROADCAST) {
1327 INP_HASH_LOCK_ASSERT(pcbinfo);
1328 error = in_pcbconnect_setup(inp, addr, &laddr.s_addr,
1329 &lport, &faddr.s_addr, &fport, NULL,
1335 * XXXRW: Why not commit the port if the address is
1338 /* Commit the local port if newly assigned. */
1339 if (inp->inp_laddr.s_addr == INADDR_ANY &&
1340 inp->inp_lport == 0) {
1341 INP_WLOCK_ASSERT(inp);
1342 INP_HASH_WLOCK_ASSERT(pcbinfo);
1344 * Remember addr if jailed, to prevent
1347 if (prison_flag(td->td_ucred, PR_IP4))
1348 inp->inp_laddr = laddr;
1349 inp->inp_lport = lport;
1350 if (in_pcbinshash(inp) != 0) {
1355 inp->inp_flags |= INP_ANONPORT;
1358 faddr = sin->sin_addr;
1359 fport = sin->sin_port;
1362 INP_LOCK_ASSERT(inp);
1363 faddr = inp->inp_faddr;
1364 fport = inp->inp_fport;
1365 if (faddr.s_addr == INADDR_ANY) {
1372 * Calculate data length and get a mbuf for UDP, IP, and possible
1373 * link-layer headers. Immediate slide the data pointer back forward
1374 * since we won't use that space at this layer.
1376 M_PREPEND(m, sizeof(struct udpiphdr) + max_linkhdr, M_NOWAIT);
1381 m->m_data += max_linkhdr;
1382 m->m_len -= max_linkhdr;
1383 m->m_pkthdr.len -= max_linkhdr;
1386 * Fill in mbuf with extended UDP header and addresses and length put
1387 * into network format.
1389 ui = mtod(m, struct udpiphdr *);
1390 bzero(ui->ui_x1, sizeof(ui->ui_x1)); /* XXX still needed? */
1394 ui->ui_sport = lport;
1395 ui->ui_dport = fport;
1396 ui->ui_ulen = htons((u_short)len + sizeof(struct udphdr));
1397 if (pr == IPPROTO_UDPLITE) {
1401 up = intoudpcb(inp);
1402 cscov = up->u_txcslen;
1403 plen = (u_short)len + sizeof(struct udphdr);
1406 ui->ui_len = htons(plen);
1407 ui->ui_ulen = htons(cscov);
1409 * For UDP-Lite, checksum coverage length of zero means
1410 * the entire UDPLite packet is covered by the checksum.
1412 cscov_partial = (cscov == 0) ? 0 : 1;
1414 ui->ui_v = IPVERSION << 4;
1417 * Set the Don't Fragment bit in the IP header.
1419 if (inp->inp_flags & INP_DONTFRAG) {
1422 ip = (struct ip *)&ui->ui_i;
1423 ip->ip_off |= htons(IP_DF);
1427 if (inp->inp_socket->so_options & SO_DONTROUTE)
1428 ipflags |= IP_ROUTETOIF;
1429 if (inp->inp_socket->so_options & SO_BROADCAST)
1430 ipflags |= IP_ALLOWBROADCAST;
1431 if (inp->inp_flags & INP_ONESBCAST)
1432 ipflags |= IP_SENDONES;
1435 mac_inpcb_create_mbuf(inp, m);
1439 * Set up checksum and output datagram.
1442 if (pr == IPPROTO_UDPLITE) {
1443 if (inp->inp_flags & INP_ONESBCAST)
1444 faddr.s_addr = INADDR_BROADCAST;
1445 if (cscov_partial) {
1446 if ((ui->ui_sum = in_cksum(m, sizeof(struct ip) + cscov)) == 0)
1447 ui->ui_sum = 0xffff;
1449 if ((ui->ui_sum = in_cksum(m, sizeof(struct udpiphdr) + len)) == 0)
1450 ui->ui_sum = 0xffff;
1452 } else if (V_udp_cksum) {
1453 if (inp->inp_flags & INP_ONESBCAST)
1454 faddr.s_addr = INADDR_BROADCAST;
1455 ui->ui_sum = in_pseudo(ui->ui_src.s_addr, faddr.s_addr,
1456 htons((u_short)len + sizeof(struct udphdr) + pr));
1457 m->m_pkthdr.csum_flags = CSUM_UDP;
1458 m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
1460 ((struct ip *)ui)->ip_len = htons(sizeof(struct udpiphdr) + len);
1461 ((struct ip *)ui)->ip_ttl = inp->inp_ip_ttl; /* XXX */
1462 ((struct ip *)ui)->ip_tos = tos; /* XXX */
1463 UDPSTAT_INC(udps_opackets);
1466 * Setup flowid / RSS information for outbound socket.
1468 * Once the UDP code decides to set a flowid some other way,
1469 * this allows the flowid to be overridden by userland.
1471 if (flowtype != M_HASHTYPE_NONE) {
1472 m->m_pkthdr.flowid = flowid;
1473 M_HASHTYPE_SET(m, flowtype);
1476 uint32_t hash_val, hash_type;
1478 * Calculate an appropriate RSS hash for UDP and
1481 * The called function will take care of figuring out
1482 * whether a 2-tuple or 4-tuple hash is required based
1483 * on the currently configured scheme.
1485 * Later later on connected socket values should be
1486 * cached in the inpcb and reused, rather than constantly
1487 * re-calculating it.
1489 * UDP Lite is a different protocol number and will
1490 * likely end up being hashed as a 2-tuple until
1491 * RSS / NICs grow UDP Lite protocol awareness.
1493 if (rss_proto_software_hash_v4(faddr, laddr, fport, lport,
1494 pr, &hash_val, &hash_type) == 0) {
1495 m->m_pkthdr.flowid = hash_val;
1496 M_HASHTYPE_SET(m, hash_type);
1503 * Don't override with the inp cached flowid value.
1505 * Depending upon the kind of send being done, the inp
1506 * flowid/flowtype values may actually not be appropriate
1507 * for this particular socket send.
1509 * We should either leave the flowid at zero (which is what is
1510 * currently done) or set it to some software generated
1511 * hash value based on the packet contents.
1513 ipflags |= IP_NODEFAULTFLOWID;
1516 if (unlock_udbinfo == UH_WLOCKED)
1517 INP_HASH_WUNLOCK(pcbinfo);
1518 else if (unlock_udbinfo == UH_RLOCKED)
1519 INP_HASH_RUNLOCK(pcbinfo);
1520 UDP_PROBE(send, NULL, inp, &ui->ui_i, inp, &ui->ui_u);
1521 error = ip_output(m, inp->inp_options,
1522 (unlock_inp == UH_WLOCKED ? &inp->inp_route : NULL), ipflags,
1523 inp->inp_moptions, inp);
1524 if (unlock_inp == UH_WLOCKED)
1531 if (unlock_udbinfo == UH_WLOCKED) {
1532 KASSERT(unlock_inp == UH_WLOCKED,
1533 ("%s: excl udbinfo lock, shared inp lock", __func__));
1534 INP_HASH_WUNLOCK(pcbinfo);
1536 } else if (unlock_udbinfo == UH_RLOCKED) {
1537 KASSERT(unlock_inp == UH_RLOCKED,
1538 ("%s: shared udbinfo lock, excl inp lock", __func__));
1539 INP_HASH_RUNLOCK(pcbinfo);
1541 } else if (unlock_inp == UH_WLOCKED)
1550 udp_abort(struct socket *so)
1553 struct inpcbinfo *pcbinfo;
1555 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1556 inp = sotoinpcb(so);
1557 KASSERT(inp != NULL, ("udp_abort: inp == NULL"));
1559 if (inp->inp_faddr.s_addr != INADDR_ANY) {
1560 INP_HASH_WLOCK(pcbinfo);
1561 in_pcbdisconnect(inp);
1562 inp->inp_laddr.s_addr = INADDR_ANY;
1563 INP_HASH_WUNLOCK(pcbinfo);
1564 soisdisconnected(so);
1570 udp_attach(struct socket *so, int proto, struct thread *td)
1573 struct inpcbinfo *pcbinfo;
1576 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1577 inp = sotoinpcb(so);
1578 KASSERT(inp == NULL, ("udp_attach: inp != NULL"));
1579 error = soreserve(so, udp_sendspace, udp_recvspace);
1582 INP_INFO_WLOCK(pcbinfo);
1583 error = in_pcballoc(so, pcbinfo);
1585 INP_INFO_WUNLOCK(pcbinfo);
1589 inp = sotoinpcb(so);
1590 inp->inp_vflag |= INP_IPV4;
1591 inp->inp_ip_ttl = V_ip_defttl;
1593 error = udp_newudpcb(inp);
1597 INP_INFO_WUNLOCK(pcbinfo);
1602 INP_INFO_WUNLOCK(pcbinfo);
1608 udp_set_kernel_tunneling(struct socket *so, udp_tun_func_t f, udp_tun_icmp_t i, void *ctx)
1613 KASSERT(so->so_type == SOCK_DGRAM,
1614 ("udp_set_kernel_tunneling: !dgram"));
1615 inp = sotoinpcb(so);
1616 KASSERT(inp != NULL, ("udp_set_kernel_tunneling: inp == NULL"));
1618 up = intoudpcb(inp);
1619 if ((up->u_tun_func != NULL) ||
1620 (up->u_icmp_func != NULL)) {
1625 up->u_icmp_func = i;
1626 up->u_tun_ctx = ctx;
1633 udp_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
1636 struct inpcbinfo *pcbinfo;
1639 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1640 inp = sotoinpcb(so);
1641 KASSERT(inp != NULL, ("udp_bind: inp == NULL"));
1643 INP_HASH_WLOCK(pcbinfo);
1644 error = in_pcbbind(inp, nam, td->td_ucred);
1645 INP_HASH_WUNLOCK(pcbinfo);
1651 udp_close(struct socket *so)
1654 struct inpcbinfo *pcbinfo;
1656 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1657 inp = sotoinpcb(so);
1658 KASSERT(inp != NULL, ("udp_close: inp == NULL"));
1660 if (inp->inp_faddr.s_addr != INADDR_ANY) {
1661 INP_HASH_WLOCK(pcbinfo);
1662 in_pcbdisconnect(inp);
1663 inp->inp_laddr.s_addr = INADDR_ANY;
1664 INP_HASH_WUNLOCK(pcbinfo);
1665 soisdisconnected(so);
1671 udp_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
1674 struct inpcbinfo *pcbinfo;
1675 struct sockaddr_in *sin;
1678 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1679 inp = sotoinpcb(so);
1680 KASSERT(inp != NULL, ("udp_connect: inp == NULL"));
1682 if (inp->inp_faddr.s_addr != INADDR_ANY) {
1686 sin = (struct sockaddr_in *)nam;
1687 error = prison_remote_ip4(td->td_ucred, &sin->sin_addr);
1692 INP_HASH_WLOCK(pcbinfo);
1693 error = in_pcbconnect(inp, nam, td->td_ucred);
1694 INP_HASH_WUNLOCK(pcbinfo);
1702 udp_detach(struct socket *so)
1705 struct inpcbinfo *pcbinfo;
1708 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1709 inp = sotoinpcb(so);
1710 KASSERT(inp != NULL, ("udp_detach: inp == NULL"));
1711 KASSERT(inp->inp_faddr.s_addr == INADDR_ANY,
1712 ("udp_detach: not disconnected"));
1713 INP_INFO_WLOCK(pcbinfo);
1715 up = intoudpcb(inp);
1716 KASSERT(up != NULL, ("%s: up == NULL", __func__));
1717 inp->inp_ppcb = NULL;
1720 INP_INFO_WUNLOCK(pcbinfo);
1725 udp_disconnect(struct socket *so)
1728 struct inpcbinfo *pcbinfo;
1730 pcbinfo = udp_get_inpcbinfo(so->so_proto->pr_protocol);
1731 inp = sotoinpcb(so);
1732 KASSERT(inp != NULL, ("udp_disconnect: inp == NULL"));
1734 if (inp->inp_faddr.s_addr == INADDR_ANY) {
1738 INP_HASH_WLOCK(pcbinfo);
1739 in_pcbdisconnect(inp);
1740 inp->inp_laddr.s_addr = INADDR_ANY;
1741 INP_HASH_WUNLOCK(pcbinfo);
1743 so->so_state &= ~SS_ISCONNECTED; /* XXX */
1750 udp_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *addr,
1751 struct mbuf *control, struct thread *td)
1755 inp = sotoinpcb(so);
1756 KASSERT(inp != NULL, ("udp_send: inp == NULL"));
1757 return (udp_output(inp, m, addr, control, td));
1762 udp_shutdown(struct socket *so)
1766 inp = sotoinpcb(so);
1767 KASSERT(inp != NULL, ("udp_shutdown: inp == NULL"));
1775 struct pr_usrreqs udp_usrreqs = {
1776 .pru_abort = udp_abort,
1777 .pru_attach = udp_attach,
1778 .pru_bind = udp_bind,
1779 .pru_connect = udp_connect,
1780 .pru_control = in_control,
1781 .pru_detach = udp_detach,
1782 .pru_disconnect = udp_disconnect,
1783 .pru_peeraddr = in_getpeeraddr,
1784 .pru_send = udp_send,
1785 .pru_soreceive = soreceive_dgram,
1786 .pru_sosend = sosend_dgram,
1787 .pru_shutdown = udp_shutdown,
1788 .pru_sockaddr = in_getsockaddr,
1789 .pru_sosetlabel = in_pcbsosetlabel,
1790 .pru_close = udp_close,