2 * Copyright (c) 2004 Jeffrey M. Hsu. All rights reserved.
3 * Copyright (c) 2004 The DragonFly Project. All rights reserved.
5 * This code is derived from software contributed to The DragonFly Project
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. Neither the name of The DragonFly Project nor the names of its
17 * contributors may be used to endorse or promote products derived
18 * from this software without specific, prior written permission.
20 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
21 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
22 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
23 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
24 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
25 * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
26 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
27 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
28 * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
29 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
30 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
35 * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995
36 * The Regents of the University of California. All rights reserved.
38 * Redistribution and use in source and binary forms, with or without
39 * modification, are permitted provided that the following conditions
41 * 1. Redistributions of source code must retain the above copyright
42 * notice, this list of conditions and the following disclaimer.
43 * 2. Redistributions in binary form must reproduce the above copyright
44 * notice, this list of conditions and the following disclaimer in the
45 * documentation and/or other materials provided with the distribution.
46 * 4. Neither the name of the University nor the names of its contributors
47 * may be used to endorse or promote products derived from this software
48 * without specific prior written permission.
50 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
51 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
52 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
53 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
54 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
55 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
56 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
57 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
58 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
59 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
62 * @(#)udp_usrreq.c 8.6 (Berkeley) 5/23/95
63 * $FreeBSD: src/sys/netinet/udp_usrreq.c,v 1.64.2.18 2003/01/24 05:11:34 sam Exp $
66 #include "opt_ipsec.h"
67 #include "opt_inet6.h"
69 #include <sys/param.h>
70 #include <sys/systm.h>
71 #include <sys/kernel.h>
72 #include <sys/malloc.h>
74 #include <sys/domain.h>
77 #include <sys/protosw.h>
78 #include <sys/socket.h>
79 #include <sys/socketvar.h>
80 #include <sys/sysctl.h>
81 #include <sys/syslog.h>
82 #include <sys/in_cksum.h>
84 #include <sys/thread2.h>
85 #include <sys/socketvar2.h>
86 #include <sys/serialize.h>
88 #include <machine/stdarg.h>
91 #include <net/route.h>
92 #include <net/netmsg2.h>
93 #include <net/netisr2.h>
95 #include <netinet/in.h>
96 #include <netinet/in_systm.h>
97 #include <netinet/ip.h>
99 #include <netinet/ip6.h>
101 #include <netinet/in_pcb.h>
102 #include <netinet/in_var.h>
103 #include <netinet/ip_var.h>
105 #include <netinet6/ip6_var.h>
107 #include <netinet/ip_icmp.h>
108 #include <netinet/icmp_var.h>
109 #include <netinet/udp.h>
110 #include <netinet/udp_var.h>
113 #include <netproto/ipsec/ipsec.h>
117 #include <netinet6/ipsec.h>
121 * UDP protocol implementation.
122 * Per RFC 768, August, 1980.
125 static int udpcksum = 1;
127 static int udpcksum = 0; /* XXX */
129 SYSCTL_INT(_net_inet_udp, UDPCTL_CHECKSUM, checksum, CTLFLAG_RW,
130 &udpcksum, 0, "Enable checksumming of UDP packets");
133 SYSCTL_INT(_net_inet_udp, OID_AUTO, log_in_vain, CTLFLAG_RW,
134 &log_in_vain, 0, "Log all incoming UDP packets");
136 static int blackhole = 0;
137 SYSCTL_INT(_net_inet_udp, OID_AUTO, blackhole, CTLFLAG_RW,
138 &blackhole, 0, "Do not send port unreachables for refused connects");
140 static int strict_mcast_mship = 1;
141 SYSCTL_INT(_net_inet_udp, OID_AUTO, strict_mcast_mship, CTLFLAG_RW,
142 &strict_mcast_mship, 0, "Only send multicast to member sockets");
144 int udp_sosend_async = 1;
145 SYSCTL_INT(_net_inet_udp, OID_AUTO, sosend_async, CTLFLAG_RW,
146 &udp_sosend_async, 0, "UDP asynchronized pru_send");
148 int udp_sosend_prepend = 1;
149 SYSCTL_INT(_net_inet_udp, OID_AUTO, sosend_prepend, CTLFLAG_RW,
150 &udp_sosend_prepend, 0,
151 "Prepend enough space for proto and link header in pru_send");
153 static int udp_reuseport_ext = 1;
154 SYSCTL_INT(_net_inet_udp, OID_AUTO, reuseport_ext, CTLFLAG_RW,
155 &udp_reuseport_ext, 0, "SO_REUSEPORT extension");
157 struct inpcbinfo udbinfo;
159 static struct netisr_barrier *udbinfo_br;
160 static struct lwkt_serialize udbinfo_slize = LWKT_SERIALIZE_INITIALIZER;
163 #define UDBHASHSIZE 16
166 struct udpstat udpstat_percpu[MAXCPU] __cachealign;
168 static struct sockaddr_in udp_in = { sizeof udp_in, AF_INET };
171 struct sockaddr_in6 uin6_sin;
172 u_char uin6_init_done : 1;
174 { sizeof udp_in6.uin6_sin, AF_INET6 },
178 struct ip6_hdr uip6_ip6;
179 u_char uip6_init_done : 1;
183 static void udp_append (struct inpcb *last, struct ip *ip,
184 struct mbuf *n, int off);
186 static void ip_2_ip6_hdr (struct ip6_hdr *ip6, struct ip *ip);
189 static int udp_connect_oncpu(struct socket *so, struct thread *td,
190 struct sockaddr_in *sin, struct sockaddr_in *if_sin);
191 static int udp_output (struct inpcb *, struct mbuf *, struct sockaddr *,
192 struct thread *, int);
199 in_pcbinfo_init(&udbinfo);
200 udbinfo.hashbase = hashinit(UDBHASHSIZE, M_PCB, &udbinfo.hashmask);
201 udbinfo.porthashbase = hashinit(UDBHASHSIZE, M_PCB,
202 &udbinfo.porthashmask);
203 udbinfo.wildcardhashbase = hashinit(UDBHASHSIZE, M_PCB,
204 &udbinfo.wildcardhashmask);
205 udbinfo.localgrphashbase = hashinit(UDBHASHSIZE, M_PCB,
206 &udbinfo.localgrphashmask);
207 udbinfo.ipi_size = sizeof(struct inpcb);
209 udbinfo_br = netisr_barrier_create();
212 * Initialize UDP statistics counters for each CPU.
214 for (cpu = 0; cpu < ncpus; ++cpu)
215 bzero(&udpstat_percpu[cpu], sizeof(struct udpstat));
219 sysctl_udpstat(SYSCTL_HANDLER_ARGS)
223 for (cpu = 0; cpu < ncpus; ++cpu) {
224 if ((error = SYSCTL_OUT(req, &udpstat_percpu[cpu],
225 sizeof(struct udpstat))))
227 if ((error = SYSCTL_IN(req, &udpstat_percpu[cpu],
228 sizeof(struct udpstat))))
234 SYSCTL_PROC(_net_inet_udp, UDPCTL_STATS, stats, (CTLTYPE_OPAQUE | CTLFLAG_RW),
235 0, 0, sysctl_udpstat, "S,udpstat", "UDP statistics");
238 * Check multicast packets to make sure they are only sent to sockets with
239 * multicast memberships for the packet's destination address and arrival
240 * interface. Multicast packets to multicast-unaware sockets are also
243 * Returns 0 if the packet is acceptable, -1 if it is not.
246 check_multicast_membership(struct ip *ip, struct inpcb *inp, struct mbuf *m)
249 struct ip_moptions *mopt;
251 if (strict_mcast_mship == 0 ||
252 !IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
255 mopt = inp->inp_moptions;
258 for (mshipno = 0; mshipno < mopt->imo_num_memberships; ++mshipno) {
259 struct in_multi *maddr = mopt->imo_membership[mshipno];
261 if (ip->ip_dst.s_addr == maddr->inm_addr.s_addr &&
262 m->m_pkthdr.rcvif == maddr->inm_ifp) {
270 udp_input(struct mbuf **mp, int *offp, int proto)
277 struct mbuf *opts = NULL;
280 struct sockaddr *append_sa;
287 udp_stat.udps_ipackets++;
290 * Strip IP options, if any; should skip this,
291 * make available to user, and use on returned packets,
292 * but we don't yet have a way to check the checksum
293 * with options still present.
295 if (iphlen > sizeof(struct ip)) {
297 iphlen = sizeof(struct ip);
301 * IP and UDP headers are together in first mbuf.
302 * Already checked and pulled up in ip_demux().
304 KASSERT(m->m_len >= iphlen + sizeof(struct udphdr),
305 ("UDP header not in one mbuf"));
307 ip = mtod(m, struct ip *);
308 uh = (struct udphdr *)((caddr_t)ip + iphlen);
310 /* destination port of 0 is illegal, based on RFC768. */
311 if (uh->uh_dport == 0)
315 * Make mbuf data length reflect UDP length.
316 * If not enough data to reflect UDP length, drop.
318 len = ntohs((u_short)uh->uh_ulen);
319 if (ip->ip_len != len) {
320 if (len > ip->ip_len || len < sizeof(struct udphdr)) {
321 udp_stat.udps_badlen++;
324 m_adj(m, len - ip->ip_len);
325 /* ip->ip_len = len; */
328 * Save a copy of the IP header in case we want restore it
329 * for sending an ICMP error message in response.
334 * Checksum extended UDP header and data.
337 if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) {
338 if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR)
339 uh->uh_sum = m->m_pkthdr.csum_data;
341 uh->uh_sum = in_pseudo(ip->ip_src.s_addr,
342 ip->ip_dst.s_addr, htonl((u_short)len +
343 m->m_pkthdr.csum_data + IPPROTO_UDP));
344 uh->uh_sum ^= 0xffff;
348 bcopy(((struct ipovly *)ip)->ih_x1, b, 9);
349 bzero(((struct ipovly *)ip)->ih_x1, 9);
350 ((struct ipovly *)ip)->ih_len = uh->uh_ulen;
351 uh->uh_sum = in_cksum(m, len + sizeof(struct ip));
352 bcopy(b, ((struct ipovly *)ip)->ih_x1, 9);
355 udp_stat.udps_badsum++;
357 return(IPPROTO_DONE);
360 udp_stat.udps_nosum++;
362 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) ||
363 in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif)) {
367 * Deliver a multicast or broadcast datagram to *all* sockets
368 * for which the local and remote addresses and ports match
369 * those of the incoming datagram. This allows more than
370 * one process to receive multi/broadcasts on the same port.
371 * (This really ought to be done for unicast datagrams as
372 * well, but that would cause problems with existing
373 * applications that open both address-specific sockets and
374 * a wildcard socket listening to the same port -- they would
375 * end up receiving duplicates of every unicast datagram.
376 * Those applications open the multiple sockets to overcome an
377 * inadequacy of the UDP socket interface, but for backwards
378 * compatibility we avoid the problem here rather than
379 * fixing the interface. Maybe 4.5BSD will remedy this?)
383 * Construct sockaddr format source address.
385 udp_in.sin_port = uh->uh_sport;
386 udp_in.sin_addr = ip->ip_src;
388 * Locate pcb(s) for datagram.
389 * (Algorithm copied from raw_intr().)
393 udp_in6.uin6_init_done = udp_ip6.uip6_init_done = 0;
395 LIST_FOREACH(inp, &udbinfo.pcblisthead, inp_list) {
396 KKASSERT((inp->inp_flags & INP_PLACEMARKER) == 0);
398 if (!(inp->inp_vflag & INP_IPV4))
401 if (inp->inp_lport != uh->uh_dport)
403 if (inp->inp_laddr.s_addr != INADDR_ANY) {
404 if (inp->inp_laddr.s_addr !=
408 if (inp->inp_faddr.s_addr != INADDR_ANY) {
409 if (inp->inp_faddr.s_addr !=
411 inp->inp_fport != uh->uh_sport)
415 if (check_multicast_membership(ip, inp, m) < 0)
422 /* check AH/ESP integrity. */
423 if (ipsec4_in_reject_so(m, last->inp_socket))
424 ipsecstat.in_polvio++;
425 /* do not inject data to pcb */
429 /* check AH/ESP integrity. */
430 if (ipsec4_in_reject(m, last))
433 #endif /*FAST_IPSEC*/
434 if ((n = m_copypacket(m, MB_DONTWAIT)) != NULL)
435 udp_append(last, ip, n,
437 sizeof(struct udphdr));
441 * Don't look for additional matches if this one does
442 * not have either the SO_REUSEPORT or SO_REUSEADDR
443 * socket options set. This heuristic avoids searching
444 * through all pcbs in the common case of a non-shared
445 * port. It * assumes that an application will never
446 * clear these options after setting them.
448 if (!(last->inp_socket->so_options &
449 (SO_REUSEPORT | SO_REUSEADDR)))
455 * No matching pcb found; discard datagram.
456 * (No need to send an ICMP Port Unreachable
457 * for a broadcast or multicast datgram.)
459 udp_stat.udps_noportbcast++;
463 /* check AH/ESP integrity. */
464 if (ipsec4_in_reject_so(m, last->inp_socket)) {
465 ipsecstat.in_polvio++;
470 /* check AH/ESP integrity. */
471 if (ipsec4_in_reject(m, last))
473 #endif /*FAST_IPSEC*/
474 udp_append(last, ip, m, iphlen + sizeof(struct udphdr));
475 return(IPPROTO_DONE);
478 * Locate pcb for datagram.
480 inp = in_pcblookup_pkthash(&udbinfo, ip->ip_src, uh->uh_sport,
481 ip->ip_dst, uh->uh_dport, 1, m->m_pkthdr.rcvif,
482 udp_reuseport_ext ? m : NULL);
485 char buf[sizeof "aaa.bbb.ccc.ddd"];
487 strcpy(buf, inet_ntoa(ip->ip_dst));
489 "Connection attempt to UDP %s:%d from %s:%d\n",
490 buf, ntohs(uh->uh_dport), inet_ntoa(ip->ip_src),
491 ntohs(uh->uh_sport));
493 udp_stat.udps_noport++;
494 if (m->m_flags & (M_BCAST | M_MCAST)) {
495 udp_stat.udps_noportbcast++;
501 if (badport_bandlim(BANDLIM_ICMP_UNREACH) < 0)
505 ip->ip_len += iphlen;
506 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0);
507 return(IPPROTO_DONE);
510 if (ipsec4_in_reject_so(m, inp->inp_socket)) {
511 ipsecstat.in_polvio++;
516 if (ipsec4_in_reject(m, inp))
518 #endif /*FAST_IPSEC*/
520 * Check the minimum TTL for socket.
522 if (ip->ip_ttl < inp->inp_ip_minttl)
526 * Construct sockaddr format source address.
527 * Stuff source address and datagram in user buffer.
529 udp_in.sin_port = uh->uh_sport;
530 udp_in.sin_addr = ip->ip_src;
531 if ((inp->inp_flags & INP_CONTROLOPTS) ||
532 (inp->inp_socket->so_options & SO_TIMESTAMP)) {
534 if (inp->inp_vflag & INP_IPV6) {
537 ip_2_ip6_hdr(&udp_ip6.uip6_ip6, ip);
538 savedflags = inp->inp_flags;
539 inp->inp_flags &= ~INP_UNMAPPABLEOPTS;
540 ip6_savecontrol(inp, &opts, &udp_ip6.uip6_ip6, m);
541 inp->inp_flags = savedflags;
544 ip_savecontrol(inp, &opts, ip, m);
546 m_adj(m, iphlen + sizeof(struct udphdr));
548 if (inp->inp_vflag & INP_IPV6) {
549 in6_sin_2_v4mapsin6(&udp_in, &udp_in6.uin6_sin);
550 append_sa = (struct sockaddr *)&udp_in6;
553 append_sa = (struct sockaddr *)&udp_in;
555 lwkt_gettoken(&inp->inp_socket->so_rcv.ssb_token);
556 if (ssb_appendaddr(&inp->inp_socket->so_rcv, append_sa, m, opts) == 0) {
557 udp_stat.udps_fullsock++;
558 lwkt_reltoken(&inp->inp_socket->so_rcv.ssb_token);
561 lwkt_reltoken(&inp->inp_socket->so_rcv.ssb_token);
562 sorwakeup(inp->inp_socket);
563 return(IPPROTO_DONE);
568 return(IPPROTO_DONE);
573 ip_2_ip6_hdr(struct ip6_hdr *ip6, struct ip *ip)
575 bzero(ip6, sizeof *ip6);
577 ip6->ip6_vfc = IPV6_VERSION;
578 ip6->ip6_plen = ip->ip_len;
579 ip6->ip6_nxt = ip->ip_p;
580 ip6->ip6_hlim = ip->ip_ttl;
581 ip6->ip6_src.s6_addr32[2] = ip6->ip6_dst.s6_addr32[2] =
583 ip6->ip6_src.s6_addr32[3] = ip->ip_src.s_addr;
584 ip6->ip6_dst.s6_addr32[3] = ip->ip_dst.s_addr;
589 * subroutine of udp_input(), mainly for source code readability.
590 * caller must properly init udp_ip6 and udp_in6 beforehand.
593 udp_append(struct inpcb *last, struct ip *ip, struct mbuf *n, int off)
595 struct sockaddr *append_sa;
596 struct mbuf *opts = NULL;
598 if (last->inp_flags & INP_CONTROLOPTS ||
599 last->inp_socket->so_options & SO_TIMESTAMP) {
601 if (last->inp_vflag & INP_IPV6) {
604 if (udp_ip6.uip6_init_done == 0) {
605 ip_2_ip6_hdr(&udp_ip6.uip6_ip6, ip);
606 udp_ip6.uip6_init_done = 1;
608 savedflags = last->inp_flags;
609 last->inp_flags &= ~INP_UNMAPPABLEOPTS;
610 ip6_savecontrol(last, &opts, &udp_ip6.uip6_ip6, n);
611 last->inp_flags = savedflags;
614 ip_savecontrol(last, &opts, ip, n);
617 if (last->inp_vflag & INP_IPV6) {
618 if (udp_in6.uin6_init_done == 0) {
619 in6_sin_2_v4mapsin6(&udp_in, &udp_in6.uin6_sin);
620 udp_in6.uin6_init_done = 1;
622 append_sa = (struct sockaddr *)&udp_in6.uin6_sin;
625 append_sa = (struct sockaddr *)&udp_in;
627 lwkt_gettoken(&last->inp_socket->so_rcv.ssb_token);
628 if (ssb_appendaddr(&last->inp_socket->so_rcv, append_sa, n, opts) == 0) {
632 udp_stat.udps_fullsock++;
634 sorwakeup(last->inp_socket);
636 lwkt_reltoken(&last->inp_socket->so_rcv.ssb_token);
640 * Notify a udp user of an asynchronous error;
641 * just wake up so that he can collect error status.
644 udp_notify(struct inpcb *inp, int error)
646 inp->inp_socket->so_error = error;
647 sorwakeup(inp->inp_socket);
648 sowwakeup(inp->inp_socket);
651 struct netmsg_udp_notify {
652 struct netmsg_base base;
653 void (*nm_notify)(struct inpcb *, int);
654 struct in_addr nm_faddr;
659 udp_notifyall_oncpu(netmsg_t msg)
661 struct netmsg_udp_notify *nm = (struct netmsg_udp_notify *)msg;
666 in_pcbnotifyall(&udbinfo.pcblisthead, nm->nm_faddr,
667 nm->nm_arg, nm->nm_notify);
668 lwkt_replymsg(&nm->base.lmsg, 0);
671 /* XXX currently udp only runs on cpu 0 */
672 nextcpu = mycpuid + 1;
673 if (nextcpu < ncpus2)
674 lwkt_forwardmsg(netisr_cpuport(nextcpu), &nm->base.lmsg);
676 lwkt_replymsg(&nmsg->base.lmsg, 0);
681 udp_rtchange(struct inpcb *inp, int err)
683 /* XXX Nuke this, once UDP inpcbs are CPU localized */
684 if (inp->inp_route.ro_rt && inp->inp_route.ro_rt->rt_cpuid == mycpuid) {
685 rtfree(inp->inp_route.ro_rt);
686 inp->inp_route.ro_rt = NULL;
688 * A new route can be allocated the next time
689 * output is attempted.
695 udp_ctlinput(netmsg_t msg)
697 struct sockaddr *sa = msg->ctlinput.nm_arg;
698 struct ip *ip = msg->ctlinput.nm_extra;
699 int cmd = msg->ctlinput.nm_cmd;
701 void (*notify) (struct inpcb *, int) = udp_notify;
702 struct in_addr faddr;
705 KKASSERT(&curthread->td_msgport == netisr_cpuport(0));
707 faddr = ((struct sockaddr_in *)sa)->sin_addr;
708 if (sa->sa_family != AF_INET || faddr.s_addr == INADDR_ANY)
711 if (PRC_IS_REDIRECT(cmd)) {
713 notify = udp_rtchange;
714 } else if (cmd == PRC_HOSTDEAD) {
716 } else if ((unsigned)cmd >= PRC_NCMDS || inetctlerrmap[cmd] == 0) {
721 uh = (struct udphdr *)((caddr_t)ip + (ip->ip_hl << 2));
722 inp = in_pcblookup_hash(&udbinfo, faddr, uh->uh_dport,
723 ip->ip_src, uh->uh_sport, 0, NULL);
724 if (inp != NULL && inp->inp_socket != NULL)
725 (*notify)(inp, inetctlerrmap[cmd]);
726 } else if (PRC_IS_REDIRECT(cmd)) {
727 struct netmsg_udp_notify *nm;
729 KKASSERT(&curthread->td_msgport == netisr_cpuport(0));
730 nm = kmalloc(sizeof(*nm), M_LWKTMSG, M_INTWAIT);
731 netmsg_init(&nm->base, NULL, &netisr_afree_rport,
732 0, udp_notifyall_oncpu);
733 nm->nm_faddr = faddr;
734 nm->nm_arg = inetctlerrmap[cmd];
735 nm->nm_notify = notify;
736 lwkt_sendmsg(netisr_cpuport(0), &nm->base.lmsg);
739 * XXX We should forward msg upon PRC_HOSTHEAD and ip == NULL,
740 * once UDP inpcbs are CPU localized
742 KKASSERT(&curthread->td_msgport == netisr_cpuport(0));
743 in_pcbnotifyall(&udbinfo.pcblisthead, faddr, inetctlerrmap[cmd],
747 lwkt_replymsg(&msg->lmsg, 0);
751 udp_pcblist(SYSCTL_HANDLER_ARGS)
757 error = in_pcblist_global_nomarker(oidp, arg1, arg2, req, &xi, &nxi);
761 KKASSERT(xi == NULL);
765 KKASSERT(xi == NULL);
769 for (i = 0; i < nxi; ++i) {
770 error = SYSCTL_OUT(req, &xi[i], sizeof(xi[i]));
778 SYSCTL_PROC(_net_inet_udp, UDPCTL_PCBLIST, pcblist, CTLFLAG_RD, &udbinfo, 0,
779 udp_pcblist, "S,xinpcb", "List of active UDP sockets");
782 udp_getcred(SYSCTL_HANDLER_ARGS)
784 struct sockaddr_in addrs[2];
785 struct ucred cred0, *cred = NULL;
789 error = priv_check(req->td, PRIV_ROOT);
792 error = SYSCTL_IN(req, addrs, sizeof addrs);
797 inp = in_pcblookup_hash(&udbinfo, addrs[1].sin_addr, addrs[1].sin_port,
798 addrs[0].sin_addr, addrs[0].sin_port, 1, NULL);
799 if (inp == NULL || inp->inp_socket == NULL) {
802 if (inp->inp_socket->so_cred != NULL) {
803 cred0 = *(inp->inp_socket->so_cred);
812 return SYSCTL_OUT(req, cred, sizeof(struct ucred));
815 SYSCTL_PROC(_net_inet_udp, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW,
816 0, 0, udp_getcred, "S,ucred", "Get the ucred of a UDP connection");
819 udp_output(struct inpcb *inp, struct mbuf *m, struct sockaddr *dstaddr,
820 struct thread *td, int flags)
823 int len = m->m_pkthdr.len;
824 struct sockaddr_in *sin; /* really is initialized before use */
825 int error = 0, lport_any = 0;
827 if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) {
832 if (inp->inp_lport == 0) { /* unbound socket */
833 error = in_pcbbind(inp, NULL, td);
837 udbinfo_barrier_set();
838 in_pcbinswildcardhash(inp);
839 udbinfo_barrier_rem();
843 if (dstaddr != NULL) { /* destination address specified */
844 if (inp->inp_faddr.s_addr != INADDR_ANY) {
845 /* already connected */
849 sin = (struct sockaddr_in *)dstaddr;
850 if (!prison_remote_ip(td, (struct sockaddr *)&sin)) {
851 error = EAFNOSUPPORT; /* IPv6 only jail */
855 if (inp->inp_faddr.s_addr == INADDR_ANY) {
856 /* no destination specified and not already connected */
864 * Calculate data length and get a mbuf
865 * for UDP and IP headers.
867 M_PREPEND(m, sizeof(struct udpiphdr), MB_DONTWAIT);
874 * Fill in mbuf with extended UDP header
875 * and addresses and length put into network format.
877 ui = mtod(m, struct udpiphdr *);
878 bzero(ui->ui_x1, sizeof ui->ui_x1); /* XXX still needed? */
879 ui->ui_pr = IPPROTO_UDP;
882 * Set destination address.
884 if (dstaddr != NULL) { /* use specified destination */
885 ui->ui_dst = sin->sin_addr;
886 ui->ui_dport = sin->sin_port;
887 } else { /* use connected destination */
888 ui->ui_dst = inp->inp_faddr;
889 ui->ui_dport = inp->inp_fport;
893 * Set source address.
895 if (inp->inp_laddr.s_addr == INADDR_ANY ||
896 IN_MULTICAST(ntohl(inp->inp_laddr.s_addr))) {
897 struct sockaddr_in *if_sin;
899 if (dstaddr == NULL) {
901 * connect() had (or should have) failed because
902 * the interface had no IP address, but the
903 * application proceeded to call send() anyways.
909 /* Look up outgoing interface. */
910 error = in_pcbladdr_find(inp, dstaddr, &if_sin, td, 1);
913 ui->ui_src = if_sin->sin_addr; /* use address of interface */
915 ui->ui_src = inp->inp_laddr; /* use non-null bound address */
917 ui->ui_sport = inp->inp_lport;
918 KASSERT(inp->inp_lport != 0, ("inp lport should have been bound"));
920 ui->ui_ulen = htons((u_short)len + sizeof(struct udphdr));
923 * Set up checksum and output datagram.
926 ui->ui_sum = in_pseudo(ui->ui_src.s_addr, ui->ui_dst.s_addr,
927 htons((u_short)len + sizeof(struct udphdr) + IPPROTO_UDP));
928 m->m_pkthdr.csum_flags = CSUM_UDP;
929 m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
930 m->m_pkthdr.csum_thlen = sizeof(struct udphdr);
934 ((struct ip *)ui)->ip_len = sizeof(struct udpiphdr) + len;
935 ((struct ip *)ui)->ip_ttl = inp->inp_ip_ttl; /* XXX */
936 ((struct ip *)ui)->ip_tos = inp->inp_ip_tos; /* XXX */
937 udp_stat.udps_opackets++;
939 error = ip_output(m, inp->inp_options, &inp->inp_route,
940 (inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST)) |
941 flags | IP_DEBUGROUTE,
942 inp->inp_moptions, inp);
945 * If this is the first data gram sent on an unbound and unconnected
946 * UDP socket, lport will be changed in this function. If target
947 * CPU after this lport changing is no longer the current CPU, then
948 * free the route entry allocated on the current CPU.
951 if (udp_addrcpu(inp->inp_faddr.s_addr, inp->inp_fport,
952 inp->inp_laddr.s_addr, inp->inp_lport) != mycpuid) {
954 struct route *ro = &inp->inp_route;
956 if (ro->ro_rt != NULL)
958 bzero(ro, sizeof(*ro));
960 panic("UDP activity should only be in netisr0");
971 u_long udp_sendspace = 9216; /* really max datagram size */
972 /* 40 1K datagrams */
973 SYSCTL_INT(_net_inet_udp, UDPCTL_MAXDGRAM, maxdgram, CTLFLAG_RW,
974 &udp_sendspace, 0, "Maximum outgoing UDP datagram size");
976 u_long udp_recvspace = 40 * (1024 +
978 sizeof(struct sockaddr_in6)
980 sizeof(struct sockaddr_in)
983 SYSCTL_INT(_net_inet_udp, UDPCTL_RECVSPACE, recvspace, CTLFLAG_RW,
984 &udp_recvspace, 0, "Maximum incoming UDP datagram size");
987 * NOTE: (so) is referenced from soabort*() and netmsg_pru_abort()
988 * will sofree() it when we return.
991 udp_abort(netmsg_t msg)
993 struct socket *so = msg->abort.base.nm_so;
997 KKASSERT(&curthread->td_msgport == netisr_cpuport(0));
1001 soisdisconnected(so);
1003 udbinfo_barrier_set();
1005 udbinfo_barrier_rem();
1010 lwkt_replymsg(&msg->abort.base.lmsg, error);
1014 udp_attach(netmsg_t msg)
1016 struct socket *so = msg->attach.base.nm_so;
1017 struct pru_attach_info *ai = msg->attach.nm_ai;
1021 KKASSERT(&curthread->td_msgport == netisr_cpuport(0));
1028 error = soreserve(so, udp_sendspace, udp_recvspace, ai->sb_rlimit);
1032 udbinfo_barrier_set();
1033 error = in_pcballoc(so, &udbinfo);
1034 udbinfo_barrier_rem();
1040 * Set default port for protocol processing prior to bind/connect.
1042 sosetport(so, netisr_cpuport(0));
1044 inp = (struct inpcb *)so->so_pcb;
1045 inp->inp_vflag |= INP_IPV4;
1046 inp->inp_ip_ttl = ip_defttl;
1049 lwkt_replymsg(&msg->attach.base.lmsg, error);
1053 udp_bind(netmsg_t msg)
1055 struct socket *so = msg->bind.base.nm_so;
1056 struct sockaddr *nam = msg->bind.nm_nam;
1057 struct thread *td = msg->bind.nm_td;
1058 struct sockaddr_in *sin = (struct sockaddr_in *)nam;
1064 error = in_pcbbind(inp, nam, td);
1066 if (sin->sin_addr.s_addr != INADDR_ANY)
1067 inp->inp_flags |= INP_WASBOUND_NOTANY;
1069 udbinfo_barrier_set();
1070 in_pcbinswildcardhash(inp);
1071 udbinfo_barrier_rem();
1076 lwkt_replymsg(&msg->bind.base.lmsg, error);
1080 udp_connect(netmsg_t msg)
1082 struct socket *so = msg->connect.base.nm_so;
1083 struct sockaddr *nam = msg->connect.nm_nam;
1084 struct thread *td = msg->connect.nm_td;
1086 struct sockaddr_in *sin = (struct sockaddr_in *)nam;
1087 struct sockaddr_in *if_sin;
1091 KKASSERT(&curthread->td_msgport == netisr_cpuport(0));
1099 if (msg->connect.nm_reconnect & NMSG_RECONNECT_RECONNECT) {
1100 panic("UDP does not support RECONNECT");
1102 msg->connect.nm_reconnect &= ~NMSG_RECONNECT_RECONNECT;
1103 in_pcblink(inp, &udbinfo);
1107 if (inp->inp_faddr.s_addr != INADDR_ANY) {
1114 * Bind if we have to
1116 if (td->td_proc && td->td_proc->p_ucred->cr_prison != NULL &&
1117 inp->inp_laddr.s_addr == INADDR_ANY) {
1118 error = in_pcbbind(inp, NULL, td);
1124 * Calculate the correct protocol processing thread. The connect
1125 * operation must run there.
1127 error = in_pcbladdr(inp, nam, &if_sin, td);
1130 if (!prison_remote_ip(td, nam)) {
1131 error = EAFNOSUPPORT; /* IPv6 only jail */
1135 port = udp_addrport(sin->sin_addr.s_addr, sin->sin_port,
1136 inp->inp_laddr.s_addr, inp->inp_lport);
1137 if (port != &curthread->td_msgport) {
1139 struct route *ro = &inp->inp_route;
1142 * in_pcbladdr() may have allocated a route entry for us
1143 * on the current CPU, but we need a route entry on the
1144 * inpcb's owner CPU, so free it here.
1146 if (ro->ro_rt != NULL)
1148 bzero(ro, sizeof(*ro));
1151 * We are moving the protocol processing port the socket
1152 * is on, we have to unlink here and re-link on the
1155 in_pcbunlink(so->so_pcb, &udbinfo);
1156 /* in_pcbunlink(so->so_pcb, &udbinfo[mycpu->gd_cpuid]); */
1157 sosetport(so, port);
1158 msg->connect.nm_reconnect |= NMSG_RECONNECT_RECONNECT;
1159 msg->connect.base.nm_dispatch = udp_connect;
1161 lwkt_forwardmsg(port, &msg->connect.base.lmsg);
1162 /* msg invalid now */
1165 panic("UDP activity should only be in netisr0");
1168 KKASSERT(port == &curthread->td_msgport);
1169 error = udp_connect_oncpu(so, td, sin, if_sin);
1171 KKASSERT(msg->connect.nm_m == NULL);
1172 lwkt_replymsg(&msg->connect.base.lmsg, error);
1176 udp_connect_oncpu(struct socket *so, struct thread *td,
1177 struct sockaddr_in *sin, struct sockaddr_in *if_sin)
1182 udbinfo_barrier_set();
1185 if (inp->inp_flags & INP_WILDCARD)
1186 in_pcbremwildcardhash(inp);
1187 error = in_pcbconnect(inp, (struct sockaddr *)sin, td);
1191 * No more errors can occur, finish adjusting the socket
1192 * and change the processing port to reflect the connected
1193 * socket. Once set we can no longer safely mess with the
1197 } else if (error == EAFNOSUPPORT) { /* connection dissolved */
1199 * Follow traditional BSD behavior and retain
1200 * the local port binding. But, fix the old misbehavior
1201 * of overwriting any previously bound local address.
1203 if (!(inp->inp_flags & INP_WASBOUND_NOTANY))
1204 inp->inp_laddr.s_addr = INADDR_ANY;
1205 in_pcbinswildcardhash(inp);
1208 udbinfo_barrier_rem();
1213 udp_detach(netmsg_t msg)
1215 struct socket *so = msg->detach.base.nm_so;
1219 KKASSERT(&curthread->td_msgport == netisr_cpuport(0));
1223 udbinfo_barrier_set();
1225 udbinfo_barrier_rem();
1230 lwkt_replymsg(&msg->detach.base.lmsg, error);
1234 udp_disconnect(netmsg_t msg)
1236 struct socket *so = msg->disconnect.base.nm_so;
1241 KKASSERT(&curthread->td_msgport == netisr_cpuport(0));
1248 if (inp->inp_faddr.s_addr == INADDR_ANY) {
1255 udbinfo_barrier_set();
1256 in_pcbdisconnect(inp);
1257 udbinfo_barrier_rem();
1259 soclrstate(so, SS_ISCONNECTED); /* XXX */
1262 ro = &inp->inp_route;
1263 if (ro->ro_rt != NULL)
1265 bzero(ro, sizeof(*ro));
1268 lwkt_replymsg(&msg->disconnect.base.lmsg, error);
1272 udp_send(netmsg_t msg)
1274 struct socket *so = msg->send.base.nm_so;
1275 struct mbuf *m = msg->send.nm_m;
1276 struct sockaddr *addr = msg->send.nm_addr;
1277 int pru_flags = msg->send.nm_flags;
1281 KKASSERT(&curthread->td_msgport == netisr_cpuport(0));
1282 KKASSERT(msg->send.nm_control == NULL);
1286 struct thread *td = msg->send.nm_td;
1289 if (pru_flags & PRUS_DONTROUTE)
1290 flags |= SO_DONTROUTE;
1291 error = udp_output(inp, m, addr, td, flags);
1297 if (pru_flags & PRUS_FREEADDR)
1298 kfree(addr, M_SONAME);
1300 if ((pru_flags & PRUS_NOREPLY) == 0)
1301 lwkt_replymsg(&msg->send.base.lmsg, error);
1305 udp_shutdown(netmsg_t msg)
1307 struct socket *so = msg->shutdown.base.nm_so;
1311 KKASSERT(&curthread->td_msgport == netisr_cpuport(0));
1320 lwkt_replymsg(&msg->shutdown.base.lmsg, error);
1326 lwkt_serialize_enter(&udbinfo_slize);
1330 udbinfo_unlock(void)
1332 lwkt_serialize_exit(&udbinfo_slize);
1336 udbinfo_barrier_set(void)
1338 netisr_barrier_set(udbinfo_br);
1343 udbinfo_barrier_rem(void)
1346 netisr_barrier_rem(udbinfo_br);
1349 struct pr_usrreqs udp_usrreqs = {
1350 .pru_abort = udp_abort,
1351 .pru_accept = pr_generic_notsupp,
1352 .pru_attach = udp_attach,
1353 .pru_bind = udp_bind,
1354 .pru_connect = udp_connect,
1355 .pru_connect2 = pr_generic_notsupp,
1356 .pru_control = in_control_dispatch,
1357 .pru_detach = udp_detach,
1358 .pru_disconnect = udp_disconnect,
1359 .pru_listen = pr_generic_notsupp,
1360 .pru_peeraddr = in_setpeeraddr_dispatch,
1361 .pru_rcvd = pr_generic_notsupp,
1362 .pru_rcvoob = pr_generic_notsupp,
1363 .pru_send = udp_send,
1364 .pru_sense = pru_sense_null,
1365 .pru_shutdown = udp_shutdown,
1366 .pru_sockaddr = in_setsockaddr_dispatch,
1367 .pru_sosend = sosendudp,
1368 .pru_soreceive = soreceive