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 * 3. All advertising materials mentioning features or use of this software
47 * must display the following acknowledgement:
48 * This product includes software developed by the University of
49 * California, Berkeley and its contributors.
50 * 4. Neither the name of the University nor the names of its contributors
51 * may be used to endorse or promote products derived from this software
52 * without specific prior written permission.
54 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
55 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
56 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
57 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
58 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
59 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
60 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
61 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
62 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
63 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
66 * @(#)udp_usrreq.c 8.6 (Berkeley) 5/23/95
67 * $FreeBSD: src/sys/netinet/udp_usrreq.c,v 1.64.2.18 2003/01/24 05:11:34 sam Exp $
70 #include "opt_ipsec.h"
71 #include "opt_inet6.h"
73 #include <sys/param.h>
74 #include <sys/systm.h>
75 #include <sys/kernel.h>
76 #include <sys/malloc.h>
78 #include <sys/domain.h>
81 #include <sys/protosw.h>
82 #include <sys/socket.h>
83 #include <sys/socketvar.h>
84 #include <sys/sysctl.h>
85 #include <sys/syslog.h>
86 #include <sys/in_cksum.h>
88 #include <sys/thread2.h>
89 #include <sys/socketvar2.h>
90 #include <sys/serialize.h>
92 #include <machine/stdarg.h>
95 #include <net/route.h>
96 #include <net/netmsg2.h>
97 #include <net/netisr2.h>
99 #include <netinet/in.h>
100 #include <netinet/in_systm.h>
101 #include <netinet/ip.h>
103 #include <netinet/ip6.h>
105 #include <netinet/in_pcb.h>
106 #include <netinet/in_var.h>
107 #include <netinet/ip_var.h>
109 #include <netinet6/ip6_var.h>
111 #include <netinet/ip_icmp.h>
112 #include <netinet/icmp_var.h>
113 #include <netinet/udp.h>
114 #include <netinet/udp_var.h>
117 #include <netproto/ipsec/ipsec.h>
121 #include <netinet6/ipsec.h>
125 * UDP protocol implementation.
126 * Per RFC 768, August, 1980.
129 static int udpcksum = 1;
131 static int udpcksum = 0; /* XXX */
133 SYSCTL_INT(_net_inet_udp, UDPCTL_CHECKSUM, checksum, CTLFLAG_RW,
134 &udpcksum, 0, "Enable checksumming of UDP packets");
137 SYSCTL_INT(_net_inet_udp, OID_AUTO, log_in_vain, CTLFLAG_RW,
138 &log_in_vain, 0, "Log all incoming UDP packets");
140 static int blackhole = 0;
141 SYSCTL_INT(_net_inet_udp, OID_AUTO, blackhole, CTLFLAG_RW,
142 &blackhole, 0, "Do not send port unreachables for refused connects");
144 static int strict_mcast_mship = 1;
145 SYSCTL_INT(_net_inet_udp, OID_AUTO, strict_mcast_mship, CTLFLAG_RW,
146 &strict_mcast_mship, 0, "Only send multicast to member sockets");
148 int udp_sosend_async = 1;
149 SYSCTL_INT(_net_inet_udp, OID_AUTO, sosend_async, CTLFLAG_RW,
150 &udp_sosend_async, 0, "UDP asynchronized pru_send");
152 int udp_sosend_prepend = 1;
153 SYSCTL_INT(_net_inet_udp, OID_AUTO, sosend_prepend, CTLFLAG_RW,
154 &udp_sosend_prepend, 0,
155 "Prepend enough space for proto and link header in pru_send");
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.ipi_size = sizeof(struct inpcb);
207 udbinfo_br = netisr_barrier_create();
210 * Initialize UDP statistics counters for each CPU.
212 for (cpu = 0; cpu < ncpus; ++cpu)
213 bzero(&udpstat_percpu[cpu], sizeof(struct udpstat));
217 sysctl_udpstat(SYSCTL_HANDLER_ARGS)
221 for (cpu = 0; cpu < ncpus; ++cpu) {
222 if ((error = SYSCTL_OUT(req, &udpstat_percpu[cpu],
223 sizeof(struct udpstat))))
225 if ((error = SYSCTL_IN(req, &udpstat_percpu[cpu],
226 sizeof(struct udpstat))))
232 SYSCTL_PROC(_net_inet_udp, UDPCTL_STATS, stats, (CTLTYPE_OPAQUE | CTLFLAG_RW),
233 0, 0, sysctl_udpstat, "S,udpstat", "UDP statistics");
236 * Check multicast packets to make sure they are only sent to sockets with
237 * multicast memberships for the packet's destination address and arrival
238 * interface. Multicast packets to multicast-unaware sockets are also
241 * Returns 0 if the packet is acceptable, -1 if it is not.
244 check_multicast_membership(struct ip *ip, struct inpcb *inp, struct mbuf *m)
247 struct ip_moptions *mopt;
249 if (strict_mcast_mship == 0 ||
250 !IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
253 mopt = inp->inp_moptions;
256 for (mshipno = 0; mshipno < mopt->imo_num_memberships; ++mshipno) {
257 struct in_multi *maddr = mopt->imo_membership[mshipno];
259 if (ip->ip_dst.s_addr == maddr->inm_addr.s_addr &&
260 m->m_pkthdr.rcvif == maddr->inm_ifp) {
268 udp_input(struct mbuf **mp, int *offp, int proto)
275 struct mbuf *opts = NULL;
278 struct sockaddr *append_sa;
285 udp_stat.udps_ipackets++;
288 * Strip IP options, if any; should skip this,
289 * make available to user, and use on returned packets,
290 * but we don't yet have a way to check the checksum
291 * with options still present.
293 if (iphlen > sizeof(struct ip)) {
295 iphlen = sizeof(struct ip);
299 * IP and UDP headers are together in first mbuf.
300 * Already checked and pulled up in ip_demux().
302 KASSERT(m->m_len >= iphlen + sizeof(struct udphdr),
303 ("UDP header not in one mbuf"));
305 ip = mtod(m, struct ip *);
306 uh = (struct udphdr *)((caddr_t)ip + iphlen);
308 /* destination port of 0 is illegal, based on RFC768. */
309 if (uh->uh_dport == 0)
313 * Make mbuf data length reflect UDP length.
314 * If not enough data to reflect UDP length, drop.
316 len = ntohs((u_short)uh->uh_ulen);
317 if (ip->ip_len != len) {
318 if (len > ip->ip_len || len < sizeof(struct udphdr)) {
319 udp_stat.udps_badlen++;
322 m_adj(m, len - ip->ip_len);
323 /* ip->ip_len = len; */
326 * Save a copy of the IP header in case we want restore it
327 * for sending an ICMP error message in response.
332 * Checksum extended UDP header and data.
335 if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) {
336 if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR)
337 uh->uh_sum = m->m_pkthdr.csum_data;
339 uh->uh_sum = in_pseudo(ip->ip_src.s_addr,
340 ip->ip_dst.s_addr, htonl((u_short)len +
341 m->m_pkthdr.csum_data + IPPROTO_UDP));
342 uh->uh_sum ^= 0xffff;
346 bcopy(((struct ipovly *)ip)->ih_x1, b, 9);
347 bzero(((struct ipovly *)ip)->ih_x1, 9);
348 ((struct ipovly *)ip)->ih_len = uh->uh_ulen;
349 uh->uh_sum = in_cksum(m, len + sizeof(struct ip));
350 bcopy(b, ((struct ipovly *)ip)->ih_x1, 9);
353 udp_stat.udps_badsum++;
355 return(IPPROTO_DONE);
358 udp_stat.udps_nosum++;
360 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) ||
361 in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif)) {
365 * Deliver a multicast or broadcast datagram to *all* sockets
366 * for which the local and remote addresses and ports match
367 * those of the incoming datagram. This allows more than
368 * one process to receive multi/broadcasts on the same port.
369 * (This really ought to be done for unicast datagrams as
370 * well, but that would cause problems with existing
371 * applications that open both address-specific sockets and
372 * a wildcard socket listening to the same port -- they would
373 * end up receiving duplicates of every unicast datagram.
374 * Those applications open the multiple sockets to overcome an
375 * inadequacy of the UDP socket interface, but for backwards
376 * compatibility we avoid the problem here rather than
377 * fixing the interface. Maybe 4.5BSD will remedy this?)
381 * Construct sockaddr format source address.
383 udp_in.sin_port = uh->uh_sport;
384 udp_in.sin_addr = ip->ip_src;
386 * Locate pcb(s) for datagram.
387 * (Algorithm copied from raw_intr().)
391 udp_in6.uin6_init_done = udp_ip6.uip6_init_done = 0;
393 LIST_FOREACH(inp, &udbinfo.pcblisthead, inp_list) {
394 KKASSERT((inp->inp_flags & INP_PLACEMARKER) == 0);
396 if (!(inp->inp_vflag & INP_IPV4))
399 if (inp->inp_lport != uh->uh_dport)
401 if (inp->inp_laddr.s_addr != INADDR_ANY) {
402 if (inp->inp_laddr.s_addr !=
406 if (inp->inp_faddr.s_addr != INADDR_ANY) {
407 if (inp->inp_faddr.s_addr !=
409 inp->inp_fport != uh->uh_sport)
413 if (check_multicast_membership(ip, inp, m) < 0)
420 /* check AH/ESP integrity. */
421 if (ipsec4_in_reject_so(m, last->inp_socket))
422 ipsecstat.in_polvio++;
423 /* do not inject data to pcb */
427 /* check AH/ESP integrity. */
428 if (ipsec4_in_reject(m, last))
431 #endif /*FAST_IPSEC*/
432 if ((n = m_copypacket(m, MB_DONTWAIT)) != NULL)
433 udp_append(last, ip, n,
435 sizeof(struct udphdr));
439 * Don't look for additional matches if this one does
440 * not have either the SO_REUSEPORT or SO_REUSEADDR
441 * socket options set. This heuristic avoids searching
442 * through all pcbs in the common case of a non-shared
443 * port. It * assumes that an application will never
444 * clear these options after setting them.
446 if (!(last->inp_socket->so_options &
447 (SO_REUSEPORT | SO_REUSEADDR)))
453 * No matching pcb found; discard datagram.
454 * (No need to send an ICMP Port Unreachable
455 * for a broadcast or multicast datgram.)
457 udp_stat.udps_noportbcast++;
461 /* check AH/ESP integrity. */
462 if (ipsec4_in_reject_so(m, last->inp_socket)) {
463 ipsecstat.in_polvio++;
468 /* check AH/ESP integrity. */
469 if (ipsec4_in_reject(m, last))
471 #endif /*FAST_IPSEC*/
472 udp_append(last, ip, m, iphlen + sizeof(struct udphdr));
473 return(IPPROTO_DONE);
476 * Locate pcb for datagram.
478 inp = in_pcblookup_hash(&udbinfo, ip->ip_src, uh->uh_sport,
479 ip->ip_dst, uh->uh_dport, 1, m->m_pkthdr.rcvif);
482 char buf[sizeof "aaa.bbb.ccc.ddd"];
484 strcpy(buf, inet_ntoa(ip->ip_dst));
486 "Connection attempt to UDP %s:%d from %s:%d\n",
487 buf, ntohs(uh->uh_dport), inet_ntoa(ip->ip_src),
488 ntohs(uh->uh_sport));
490 udp_stat.udps_noport++;
491 if (m->m_flags & (M_BCAST | M_MCAST)) {
492 udp_stat.udps_noportbcast++;
498 if (badport_bandlim(BANDLIM_ICMP_UNREACH) < 0)
502 ip->ip_len += iphlen;
503 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0);
504 return(IPPROTO_DONE);
507 if (ipsec4_in_reject_so(m, inp->inp_socket)) {
508 ipsecstat.in_polvio++;
513 if (ipsec4_in_reject(m, inp))
515 #endif /*FAST_IPSEC*/
517 * Check the minimum TTL for socket.
519 if (ip->ip_ttl < inp->inp_ip_minttl)
523 * Construct sockaddr format source address.
524 * Stuff source address and datagram in user buffer.
526 udp_in.sin_port = uh->uh_sport;
527 udp_in.sin_addr = ip->ip_src;
528 if ((inp->inp_flags & INP_CONTROLOPTS) ||
529 (inp->inp_socket->so_options & SO_TIMESTAMP)) {
531 if (inp->inp_vflag & INP_IPV6) {
534 ip_2_ip6_hdr(&udp_ip6.uip6_ip6, ip);
535 savedflags = inp->inp_flags;
536 inp->inp_flags &= ~INP_UNMAPPABLEOPTS;
537 ip6_savecontrol(inp, &opts, &udp_ip6.uip6_ip6, m);
538 inp->inp_flags = savedflags;
541 ip_savecontrol(inp, &opts, ip, m);
543 m_adj(m, iphlen + sizeof(struct udphdr));
545 if (inp->inp_vflag & INP_IPV6) {
546 in6_sin_2_v4mapsin6(&udp_in, &udp_in6.uin6_sin);
547 append_sa = (struct sockaddr *)&udp_in6;
550 append_sa = (struct sockaddr *)&udp_in;
552 lwkt_gettoken(&inp->inp_socket->so_rcv.ssb_token);
553 if (ssb_appendaddr(&inp->inp_socket->so_rcv, append_sa, m, opts) == 0) {
554 udp_stat.udps_fullsock++;
555 lwkt_reltoken(&inp->inp_socket->so_rcv.ssb_token);
558 lwkt_reltoken(&inp->inp_socket->so_rcv.ssb_token);
559 sorwakeup(inp->inp_socket);
560 return(IPPROTO_DONE);
565 return(IPPROTO_DONE);
570 ip_2_ip6_hdr(struct ip6_hdr *ip6, struct ip *ip)
572 bzero(ip6, sizeof *ip6);
574 ip6->ip6_vfc = IPV6_VERSION;
575 ip6->ip6_plen = ip->ip_len;
576 ip6->ip6_nxt = ip->ip_p;
577 ip6->ip6_hlim = ip->ip_ttl;
578 ip6->ip6_src.s6_addr32[2] = ip6->ip6_dst.s6_addr32[2] =
580 ip6->ip6_src.s6_addr32[3] = ip->ip_src.s_addr;
581 ip6->ip6_dst.s6_addr32[3] = ip->ip_dst.s_addr;
586 * subroutine of udp_input(), mainly for source code readability.
587 * caller must properly init udp_ip6 and udp_in6 beforehand.
590 udp_append(struct inpcb *last, struct ip *ip, struct mbuf *n, int off)
592 struct sockaddr *append_sa;
593 struct mbuf *opts = NULL;
595 if (last->inp_flags & INP_CONTROLOPTS ||
596 last->inp_socket->so_options & SO_TIMESTAMP) {
598 if (last->inp_vflag & INP_IPV6) {
601 if (udp_ip6.uip6_init_done == 0) {
602 ip_2_ip6_hdr(&udp_ip6.uip6_ip6, ip);
603 udp_ip6.uip6_init_done = 1;
605 savedflags = last->inp_flags;
606 last->inp_flags &= ~INP_UNMAPPABLEOPTS;
607 ip6_savecontrol(last, &opts, &udp_ip6.uip6_ip6, n);
608 last->inp_flags = savedflags;
611 ip_savecontrol(last, &opts, ip, n);
614 if (last->inp_vflag & INP_IPV6) {
615 if (udp_in6.uin6_init_done == 0) {
616 in6_sin_2_v4mapsin6(&udp_in, &udp_in6.uin6_sin);
617 udp_in6.uin6_init_done = 1;
619 append_sa = (struct sockaddr *)&udp_in6.uin6_sin;
622 append_sa = (struct sockaddr *)&udp_in;
624 lwkt_gettoken(&last->inp_socket->so_rcv.ssb_token);
625 if (ssb_appendaddr(&last->inp_socket->so_rcv, append_sa, n, opts) == 0) {
629 udp_stat.udps_fullsock++;
631 sorwakeup(last->inp_socket);
633 lwkt_reltoken(&last->inp_socket->so_rcv.ssb_token);
637 * Notify a udp user of an asynchronous error;
638 * just wake up so that he can collect error status.
641 udp_notify(struct inpcb *inp, int error)
643 inp->inp_socket->so_error = error;
644 sorwakeup(inp->inp_socket);
645 sowwakeup(inp->inp_socket);
648 struct netmsg_udp_notify {
649 struct netmsg_base base;
650 void (*nm_notify)(struct inpcb *, int);
651 struct in_addr nm_faddr;
656 udp_notifyall_oncpu(netmsg_t msg)
658 struct netmsg_udp_notify *nm = (struct netmsg_udp_notify *)msg;
663 in_pcbnotifyall(&udbinfo.pcblisthead, nm->nm_faddr,
664 nm->nm_arg, nm->nm_notify);
665 lwkt_replymsg(&nm->base.lmsg, 0);
668 /* XXX currently udp only runs on cpu 0 */
669 nextcpu = mycpuid + 1;
670 if (nextcpu < ncpus2)
671 lwkt_forwardmsg(netisr_cpuport(nextcpu), &nm->base.lmsg);
673 lwkt_replymsg(&nmsg->base.lmsg, 0);
678 udp_rtchange(struct inpcb *inp, int err)
680 /* XXX Nuke this, once UDP inpcbs are CPU localized */
681 if (inp->inp_route.ro_rt && inp->inp_route.ro_rt->rt_cpuid == mycpuid) {
682 rtfree(inp->inp_route.ro_rt);
683 inp->inp_route.ro_rt = NULL;
685 * A new route can be allocated the next time
686 * output is attempted.
692 udp_ctlinput(netmsg_t msg)
694 struct sockaddr *sa = msg->ctlinput.nm_arg;
695 struct ip *ip = msg->ctlinput.nm_extra;
696 int cmd = msg->ctlinput.nm_cmd;
698 void (*notify) (struct inpcb *, int) = udp_notify;
699 struct in_addr faddr;
702 KKASSERT(&curthread->td_msgport == netisr_cpuport(0));
704 faddr = ((struct sockaddr_in *)sa)->sin_addr;
705 if (sa->sa_family != AF_INET || faddr.s_addr == INADDR_ANY)
708 if (PRC_IS_REDIRECT(cmd)) {
710 notify = udp_rtchange;
711 } else if (cmd == PRC_HOSTDEAD) {
713 } else if ((unsigned)cmd >= PRC_NCMDS || inetctlerrmap[cmd] == 0) {
718 uh = (struct udphdr *)((caddr_t)ip + (ip->ip_hl << 2));
719 inp = in_pcblookup_hash(&udbinfo, faddr, uh->uh_dport,
720 ip->ip_src, uh->uh_sport, 0, NULL);
721 if (inp != NULL && inp->inp_socket != NULL)
722 (*notify)(inp, inetctlerrmap[cmd]);
723 } else if (PRC_IS_REDIRECT(cmd)) {
724 struct netmsg_udp_notify *nm;
726 KKASSERT(&curthread->td_msgport == netisr_cpuport(0));
727 nm = kmalloc(sizeof(*nm), M_LWKTMSG, M_INTWAIT);
728 netmsg_init(&nm->base, NULL, &netisr_afree_rport,
729 0, udp_notifyall_oncpu);
730 nm->nm_faddr = faddr;
731 nm->nm_arg = inetctlerrmap[cmd];
732 nm->nm_notify = notify;
733 lwkt_sendmsg(netisr_cpuport(0), &nm->base.lmsg);
736 * XXX We should forward msg upon PRC_HOSTHEAD and ip == NULL,
737 * once UDP inpcbs are CPU localized
739 KKASSERT(&curthread->td_msgport == netisr_cpuport(0));
740 in_pcbnotifyall(&udbinfo.pcblisthead, faddr, inetctlerrmap[cmd],
744 lwkt_replymsg(&msg->lmsg, 0);
748 udp_pcblist(SYSCTL_HANDLER_ARGS)
754 error = in_pcblist_global_nomarker(oidp, arg1, arg2, req, &xi, &nxi);
758 KKASSERT(xi == NULL);
762 KKASSERT(xi == NULL);
766 for (i = 0; i < nxi; ++i) {
767 error = SYSCTL_OUT(req, &xi[i], sizeof(xi[i]));
775 SYSCTL_PROC(_net_inet_udp, UDPCTL_PCBLIST, pcblist, CTLFLAG_RD, &udbinfo, 0,
776 udp_pcblist, "S,xinpcb", "List of active UDP sockets");
779 udp_getcred(SYSCTL_HANDLER_ARGS)
781 struct sockaddr_in addrs[2];
782 struct ucred cred0, *cred = NULL;
786 error = priv_check(req->td, PRIV_ROOT);
789 error = SYSCTL_IN(req, addrs, sizeof addrs);
794 inp = in_pcblookup_hash(&udbinfo, addrs[1].sin_addr, addrs[1].sin_port,
795 addrs[0].sin_addr, addrs[0].sin_port, 1, NULL);
796 if (inp == NULL || inp->inp_socket == NULL) {
799 if (inp->inp_socket->so_cred != NULL) {
800 cred0 = *(inp->inp_socket->so_cred);
809 return SYSCTL_OUT(req, cred, sizeof(struct ucred));
812 SYSCTL_PROC(_net_inet_udp, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW,
813 0, 0, udp_getcred, "S,ucred", "Get the ucred of a UDP connection");
816 udp_output(struct inpcb *inp, struct mbuf *m, struct sockaddr *dstaddr,
817 struct thread *td, int flags)
820 int len = m->m_pkthdr.len;
821 struct sockaddr_in *sin; /* really is initialized before use */
822 int error = 0, lport_any = 0;
824 if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) {
829 if (inp->inp_lport == 0) { /* unbound socket */
830 error = in_pcbbind(inp, NULL, td);
834 udbinfo_barrier_set();
835 in_pcbinswildcardhash(inp);
836 udbinfo_barrier_rem();
840 if (dstaddr != NULL) { /* destination address specified */
841 if (inp->inp_faddr.s_addr != INADDR_ANY) {
842 /* already connected */
846 sin = (struct sockaddr_in *)dstaddr;
847 if (!prison_remote_ip(td, (struct sockaddr *)&sin)) {
848 error = EAFNOSUPPORT; /* IPv6 only jail */
852 if (inp->inp_faddr.s_addr == INADDR_ANY) {
853 /* no destination specified and not already connected */
861 * Calculate data length and get a mbuf
862 * for UDP and IP headers.
864 M_PREPEND(m, sizeof(struct udpiphdr), MB_DONTWAIT);
871 * Fill in mbuf with extended UDP header
872 * and addresses and length put into network format.
874 ui = mtod(m, struct udpiphdr *);
875 bzero(ui->ui_x1, sizeof ui->ui_x1); /* XXX still needed? */
876 ui->ui_pr = IPPROTO_UDP;
879 * Set destination address.
881 if (dstaddr != NULL) { /* use specified destination */
882 ui->ui_dst = sin->sin_addr;
883 ui->ui_dport = sin->sin_port;
884 } else { /* use connected destination */
885 ui->ui_dst = inp->inp_faddr;
886 ui->ui_dport = inp->inp_fport;
890 * Set source address.
892 if (inp->inp_laddr.s_addr == INADDR_ANY) {
893 struct sockaddr_in *if_sin;
895 if (dstaddr == NULL) {
897 * connect() had (or should have) failed because
898 * the interface had no IP address, but the
899 * application proceeded to call send() anyways.
905 /* Look up outgoing interface. */
906 if ((error = in_pcbladdr(inp, dstaddr, &if_sin, td)))
908 ui->ui_src = if_sin->sin_addr; /* use address of interface */
909 } else if (!IN_MULTICAST(ntohl(inp->inp_laddr.s_addr))) {
910 ui->ui_src = inp->inp_laddr; /* use non-null bound address */
912 /* Bound to multicast address; let ip_output choose */
913 ui->ui_src.s_addr = INADDR_ANY;
915 ui->ui_sport = inp->inp_lport;
916 KASSERT(inp->inp_lport != 0, ("inp lport should have been bound"));
918 ui->ui_ulen = htons((u_short)len + sizeof(struct udphdr));
921 * Set up checksum and output datagram.
924 ui->ui_sum = in_pseudo(ui->ui_src.s_addr, ui->ui_dst.s_addr,
925 htons((u_short)len + sizeof(struct udphdr) + IPPROTO_UDP));
926 m->m_pkthdr.csum_flags = CSUM_UDP;
927 m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
928 m->m_pkthdr.csum_thlen = sizeof(struct udphdr);
932 ((struct ip *)ui)->ip_len = sizeof(struct udpiphdr) + len;
933 ((struct ip *)ui)->ip_ttl = inp->inp_ip_ttl; /* XXX */
934 ((struct ip *)ui)->ip_tos = inp->inp_ip_tos; /* XXX */
935 udp_stat.udps_opackets++;
937 error = ip_output(m, inp->inp_options, &inp->inp_route,
938 (inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST)) |
939 flags | IP_DEBUGROUTE,
940 inp->inp_moptions, inp);
943 * If this is the first data gram sent on an unbound and unconnected
944 * UDP socket, lport will be changed in this function. If target
945 * CPU after this lport changing is no longer the current CPU, then
946 * free the route entry allocated on the current CPU.
949 if (udp_addrcpu(inp->inp_faddr.s_addr, inp->inp_fport,
950 inp->inp_laddr.s_addr, inp->inp_lport) != mycpuid) {
952 struct route *ro = &inp->inp_route;
954 if (ro->ro_rt != NULL)
956 bzero(ro, sizeof(*ro));
958 panic("UDP activity should only be in netisr0");
969 u_long udp_sendspace = 9216; /* really max datagram size */
970 /* 40 1K datagrams */
971 SYSCTL_INT(_net_inet_udp, UDPCTL_MAXDGRAM, maxdgram, CTLFLAG_RW,
972 &udp_sendspace, 0, "Maximum outgoing UDP datagram size");
974 u_long udp_recvspace = 40 * (1024 +
976 sizeof(struct sockaddr_in6)
978 sizeof(struct sockaddr_in)
981 SYSCTL_INT(_net_inet_udp, UDPCTL_RECVSPACE, recvspace, CTLFLAG_RW,
982 &udp_recvspace, 0, "Maximum incoming UDP datagram size");
985 * NOTE: (so) is referenced from soabort*() and netmsg_pru_abort()
986 * will sofree() it when we return.
989 udp_abort(netmsg_t msg)
991 struct socket *so = msg->abort.base.nm_so;
995 KKASSERT(&curthread->td_msgport == netisr_cpuport(0));
999 soisdisconnected(so);
1001 udbinfo_barrier_set();
1003 udbinfo_barrier_rem();
1008 lwkt_replymsg(&msg->abort.base.lmsg, error);
1012 udp_attach(netmsg_t msg)
1014 struct socket *so = msg->attach.base.nm_so;
1015 struct pru_attach_info *ai = msg->attach.nm_ai;
1019 KKASSERT(&curthread->td_msgport == netisr_cpuport(0));
1026 error = soreserve(so, udp_sendspace, udp_recvspace, ai->sb_rlimit);
1030 udbinfo_barrier_set();
1031 error = in_pcballoc(so, &udbinfo);
1032 udbinfo_barrier_rem();
1038 * Set default port for protocol processing prior to bind/connect.
1040 sosetport(so, netisr_cpuport(0));
1042 inp = (struct inpcb *)so->so_pcb;
1043 inp->inp_vflag |= INP_IPV4;
1044 inp->inp_ip_ttl = ip_defttl;
1047 lwkt_replymsg(&msg->attach.base.lmsg, error);
1051 udp_bind(netmsg_t msg)
1053 struct socket *so = msg->bind.base.nm_so;
1054 struct sockaddr *nam = msg->bind.nm_nam;
1055 struct thread *td = msg->bind.nm_td;
1056 struct sockaddr_in *sin = (struct sockaddr_in *)nam;
1062 error = in_pcbbind(inp, nam, td);
1064 if (sin->sin_addr.s_addr != INADDR_ANY)
1065 inp->inp_flags |= INP_WASBOUND_NOTANY;
1067 udbinfo_barrier_set();
1068 in_pcbinswildcardhash(inp);
1069 udbinfo_barrier_rem();
1074 lwkt_replymsg(&msg->bind.base.lmsg, error);
1078 udp_connect(netmsg_t msg)
1080 struct socket *so = msg->connect.base.nm_so;
1081 struct sockaddr *nam = msg->connect.nm_nam;
1082 struct thread *td = msg->connect.nm_td;
1084 struct sockaddr_in *sin = (struct sockaddr_in *)nam;
1085 struct sockaddr_in *if_sin;
1089 KKASSERT(&curthread->td_msgport == netisr_cpuport(0));
1097 if (msg->connect.nm_reconnect & NMSG_RECONNECT_RECONNECT) {
1098 panic("UDP does not support RECONNECT");
1100 msg->connect.nm_reconnect &= ~NMSG_RECONNECT_RECONNECT;
1101 in_pcblink(inp, &udbinfo);
1105 if (inp->inp_faddr.s_addr != INADDR_ANY) {
1112 * Bind if we have to
1114 if (td->td_proc && td->td_proc->p_ucred->cr_prison != NULL &&
1115 inp->inp_laddr.s_addr == INADDR_ANY) {
1116 error = in_pcbbind(inp, NULL, td);
1122 * Calculate the correct protocol processing thread. The connect
1123 * operation must run there.
1125 error = in_pcbladdr(inp, nam, &if_sin, td);
1128 if (!prison_remote_ip(td, nam)) {
1129 error = EAFNOSUPPORT; /* IPv6 only jail */
1133 port = udp_addrport(sin->sin_addr.s_addr, sin->sin_port,
1134 inp->inp_laddr.s_addr, inp->inp_lport);
1135 if (port != &curthread->td_msgport) {
1137 struct route *ro = &inp->inp_route;
1140 * in_pcbladdr() may have allocated a route entry for us
1141 * on the current CPU, but we need a route entry on the
1142 * inpcb's owner CPU, so free it here.
1144 if (ro->ro_rt != NULL)
1146 bzero(ro, sizeof(*ro));
1149 * We are moving the protocol processing port the socket
1150 * is on, we have to unlink here and re-link on the
1153 in_pcbunlink(so->so_pcb, &udbinfo);
1154 /* in_pcbunlink(so->so_pcb, &udbinfo[mycpu->gd_cpuid]); */
1155 sosetport(so, port);
1156 msg->connect.nm_reconnect |= NMSG_RECONNECT_RECONNECT;
1157 msg->connect.base.nm_dispatch = udp_connect;
1159 lwkt_forwardmsg(port, &msg->connect.base.lmsg);
1160 /* msg invalid now */
1163 panic("UDP activity should only be in netisr0");
1166 KKASSERT(port == &curthread->td_msgport);
1167 error = udp_connect_oncpu(so, td, sin, if_sin);
1169 KKASSERT(msg->connect.nm_m == NULL);
1170 lwkt_replymsg(&msg->connect.base.lmsg, error);
1174 udp_connect_oncpu(struct socket *so, struct thread *td,
1175 struct sockaddr_in *sin, struct sockaddr_in *if_sin)
1180 udbinfo_barrier_set();
1183 if (inp->inp_flags & INP_WILDCARD)
1184 in_pcbremwildcardhash(inp);
1185 error = in_pcbconnect(inp, (struct sockaddr *)sin, td);
1189 * No more errors can occur, finish adjusting the socket
1190 * and change the processing port to reflect the connected
1191 * socket. Once set we can no longer safely mess with the
1195 } else if (error == EAFNOSUPPORT) { /* connection dissolved */
1197 * Follow traditional BSD behavior and retain
1198 * the local port binding. But, fix the old misbehavior
1199 * of overwriting any previously bound local address.
1201 if (!(inp->inp_flags & INP_WASBOUND_NOTANY))
1202 inp->inp_laddr.s_addr = INADDR_ANY;
1203 in_pcbinswildcardhash(inp);
1206 udbinfo_barrier_rem();
1211 udp_detach(netmsg_t msg)
1213 struct socket *so = msg->detach.base.nm_so;
1217 KKASSERT(&curthread->td_msgport == netisr_cpuport(0));
1221 udbinfo_barrier_set();
1223 udbinfo_barrier_rem();
1228 lwkt_replymsg(&msg->detach.base.lmsg, error);
1232 udp_disconnect(netmsg_t msg)
1234 struct socket *so = msg->disconnect.base.nm_so;
1239 KKASSERT(&curthread->td_msgport == netisr_cpuport(0));
1246 if (inp->inp_faddr.s_addr == INADDR_ANY) {
1253 udbinfo_barrier_set();
1254 in_pcbdisconnect(inp);
1255 udbinfo_barrier_rem();
1257 soclrstate(so, SS_ISCONNECTED); /* XXX */
1260 ro = &inp->inp_route;
1261 if (ro->ro_rt != NULL)
1263 bzero(ro, sizeof(*ro));
1266 lwkt_replymsg(&msg->disconnect.base.lmsg, error);
1270 udp_send(netmsg_t msg)
1272 struct socket *so = msg->send.base.nm_so;
1273 struct mbuf *m = msg->send.nm_m;
1274 struct sockaddr *addr = msg->send.nm_addr;
1275 int pru_flags = msg->send.nm_flags;
1279 KKASSERT(&curthread->td_msgport == netisr_cpuport(0));
1280 KKASSERT(msg->send.nm_control == NULL);
1284 struct thread *td = msg->send.nm_td;
1287 if (pru_flags & PRUS_DONTROUTE)
1288 flags |= SO_DONTROUTE;
1289 error = udp_output(inp, m, addr, td, flags);
1295 if (pru_flags & PRUS_FREEADDR)
1296 kfree(addr, M_SONAME);
1298 if ((pru_flags & PRUS_NOREPLY) == 0)
1299 lwkt_replymsg(&msg->send.base.lmsg, error);
1303 udp_shutdown(netmsg_t msg)
1305 struct socket *so = msg->shutdown.base.nm_so;
1309 KKASSERT(&curthread->td_msgport == netisr_cpuport(0));
1318 lwkt_replymsg(&msg->shutdown.base.lmsg, error);
1324 lwkt_serialize_enter(&udbinfo_slize);
1328 udbinfo_unlock(void)
1330 lwkt_serialize_exit(&udbinfo_slize);
1334 udbinfo_barrier_set(void)
1336 netisr_barrier_set(udbinfo_br);
1341 udbinfo_barrier_rem(void)
1344 netisr_barrier_rem(udbinfo_br);
1347 struct pr_usrreqs udp_usrreqs = {
1348 .pru_abort = udp_abort,
1349 .pru_accept = pr_generic_notsupp,
1350 .pru_attach = udp_attach,
1351 .pru_bind = udp_bind,
1352 .pru_connect = udp_connect,
1353 .pru_connect2 = pr_generic_notsupp,
1354 .pru_control = in_control_dispatch,
1355 .pru_detach = udp_detach,
1356 .pru_disconnect = udp_disconnect,
1357 .pru_listen = pr_generic_notsupp,
1358 .pru_peeraddr = in_setpeeraddr_dispatch,
1359 .pru_rcvd = pr_generic_notsupp,
1360 .pru_rcvoob = pr_generic_notsupp,
1361 .pru_send = udp_send,
1362 .pru_sense = pru_sense_null,
1363 .pru_shutdown = udp_shutdown,
1364 .pru_sockaddr = in_setsockaddr_dispatch,
1365 .pru_sosend = sosendudp,
1366 .pru_soreceive = soreceive