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. 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;
158 struct inpcbportinfo udbportinfo;
160 static struct netisr_barrier *udbinfo_br;
161 static struct lwkt_serialize udbinfo_slize = LWKT_SERIALIZE_INITIALIZER;
164 #define UDBHASHSIZE 16
167 struct udpstat udpstat_percpu[MAXCPU] __cachealign;
171 struct sockaddr_in6 uin6_sin;
172 u_char uin6_init_done : 1;
175 struct ip6_hdr uip6_ip6;
176 u_char uip6_init_done : 1;
183 static void udp_append (struct inpcb *last, struct ip *ip,
184 struct mbuf *n, int off, struct sockaddr_in *udp_in,
185 struct udp_in6 *, struct udp_ip6 *);
187 static void ip_2_ip6_hdr (struct ip6_hdr *ip6, struct ip *ip);
190 static int udp_connect_oncpu(struct socket *so, struct thread *td,
191 struct sockaddr_in *sin, struct sockaddr_in *if_sin);
192 static int udp_output (struct inpcb *, struct mbuf *, struct sockaddr *,
193 struct thread *, int, boolean_t);
200 in_pcbinfo_init(&udbinfo);
201 in_pcbportinfo_init(&udbportinfo, UDBHASHSIZE, FALSE);
203 udbinfo.hashbase = hashinit(UDBHASHSIZE, M_PCB, &udbinfo.hashmask);
204 udbinfo.portinfo = &udbportinfo;
205 udbinfo.wildcardhashbase = hashinit(UDBHASHSIZE, M_PCB,
206 &udbinfo.wildcardhashmask);
207 udbinfo.localgrphashbase = hashinit(UDBHASHSIZE, M_PCB,
208 &udbinfo.localgrphashmask);
209 udbinfo.ipi_size = sizeof(struct inpcb);
211 udbinfo_br = netisr_barrier_create();
214 * Initialize UDP statistics counters for each CPU.
216 for (cpu = 0; cpu < ncpus; ++cpu)
217 bzero(&udpstat_percpu[cpu], sizeof(struct udpstat));
221 sysctl_udpstat(SYSCTL_HANDLER_ARGS)
225 for (cpu = 0; cpu < ncpus; ++cpu) {
226 if ((error = SYSCTL_OUT(req, &udpstat_percpu[cpu],
227 sizeof(struct udpstat))))
229 if ((error = SYSCTL_IN(req, &udpstat_percpu[cpu],
230 sizeof(struct udpstat))))
236 SYSCTL_PROC(_net_inet_udp, UDPCTL_STATS, stats, (CTLTYPE_OPAQUE | CTLFLAG_RW),
237 0, 0, sysctl_udpstat, "S,udpstat", "UDP statistics");
240 * Check multicast packets to make sure they are only sent to sockets with
241 * multicast memberships for the packet's destination address and arrival
242 * interface. Multicast packets to multicast-unaware sockets are also
245 * Returns 0 if the packet is acceptable, -1 if it is not.
248 check_multicast_membership(struct ip *ip, struct inpcb *inp, struct mbuf *m)
251 struct ip_moptions *mopt;
253 if (strict_mcast_mship == 0 ||
254 !IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
257 mopt = inp->inp_moptions;
260 for (mshipno = 0; mshipno < mopt->imo_num_memberships; ++mshipno) {
261 struct in_multi *maddr = mopt->imo_membership[mshipno];
263 if (ip->ip_dst.s_addr == maddr->inm_addr.s_addr &&
264 m->m_pkthdr.rcvif == maddr->inm_ifp) {
272 udp_input(struct mbuf **mp, int *offp, int proto)
274 struct sockaddr_in udp_in = { sizeof udp_in, AF_INET };
276 struct udp_in6 udp_in6 = {
277 { sizeof udp_in6.uin6_sin, AF_INET6 }, 0
279 struct udp_ip6 udp_ip6;
287 struct mbuf *opts = NULL;
290 struct sockaddr *append_sa;
297 udp_stat.udps_ipackets++;
300 * Strip IP options, if any; should skip this,
301 * make available to user, and use on returned packets,
302 * but we don't yet have a way to check the checksum
303 * with options still present.
305 if (iphlen > sizeof(struct ip)) {
307 iphlen = sizeof(struct ip);
311 * IP and UDP headers are together in first mbuf.
312 * Already checked and pulled up in ip_demux().
314 KASSERT(m->m_len >= iphlen + sizeof(struct udphdr),
315 ("UDP header not in one mbuf"));
317 ip = mtod(m, struct ip *);
318 uh = (struct udphdr *)((caddr_t)ip + iphlen);
320 /* destination port of 0 is illegal, based on RFC768. */
321 if (uh->uh_dport == 0)
325 * Make mbuf data length reflect UDP length.
326 * If not enough data to reflect UDP length, drop.
328 len = ntohs((u_short)uh->uh_ulen);
329 if (ip->ip_len != len) {
330 if (len > ip->ip_len || len < sizeof(struct udphdr)) {
331 udp_stat.udps_badlen++;
334 m_adj(m, len - ip->ip_len);
335 /* ip->ip_len = len; */
338 * Save a copy of the IP header in case we want restore it
339 * for sending an ICMP error message in response.
344 * Checksum extended UDP header and data.
347 if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) {
348 if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR)
349 uh->uh_sum = m->m_pkthdr.csum_data;
351 uh->uh_sum = in_pseudo(ip->ip_src.s_addr,
352 ip->ip_dst.s_addr, htonl((u_short)len +
353 m->m_pkthdr.csum_data + IPPROTO_UDP));
354 uh->uh_sum ^= 0xffff;
358 bcopy(((struct ipovly *)ip)->ih_x1, b, 9);
359 bzero(((struct ipovly *)ip)->ih_x1, 9);
360 ((struct ipovly *)ip)->ih_len = uh->uh_ulen;
361 uh->uh_sum = in_cksum(m, len + sizeof(struct ip));
362 bcopy(b, ((struct ipovly *)ip)->ih_x1, 9);
365 udp_stat.udps_badsum++;
367 return(IPPROTO_DONE);
370 udp_stat.udps_nosum++;
372 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) ||
373 in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif)) {
377 * Deliver a multicast or broadcast datagram to *all* sockets
378 * for which the local and remote addresses and ports match
379 * those of the incoming datagram. This allows more than
380 * one process to receive multi/broadcasts on the same port.
381 * (This really ought to be done for unicast datagrams as
382 * well, but that would cause problems with existing
383 * applications that open both address-specific sockets and
384 * a wildcard socket listening to the same port -- they would
385 * end up receiving duplicates of every unicast datagram.
386 * Those applications open the multiple sockets to overcome an
387 * inadequacy of the UDP socket interface, but for backwards
388 * compatibility we avoid the problem here rather than
389 * fixing the interface. Maybe 4.5BSD will remedy this?)
393 * Construct sockaddr format source address.
395 udp_in.sin_port = uh->uh_sport;
396 udp_in.sin_addr = ip->ip_src;
398 * Locate pcb(s) for datagram.
399 * (Algorithm copied from raw_intr().)
403 udp_in6.uin6_init_done = udp_ip6.uip6_init_done = 0;
405 LIST_FOREACH(inp, &udbinfo.pcblisthead, inp_list) {
406 KKASSERT((inp->inp_flags & INP_PLACEMARKER) == 0);
408 if (!(inp->inp_vflag & INP_IPV4))
411 if (inp->inp_lport != uh->uh_dport)
413 if (inp->inp_laddr.s_addr != INADDR_ANY) {
414 if (inp->inp_laddr.s_addr !=
418 if (inp->inp_faddr.s_addr != INADDR_ANY) {
419 if (inp->inp_faddr.s_addr !=
421 inp->inp_fport != uh->uh_sport)
425 if (check_multicast_membership(ip, inp, m) < 0)
432 /* check AH/ESP integrity. */
433 if (ipsec4_in_reject_so(m, last->inp_socket))
434 ipsecstat.in_polvio++;
435 /* do not inject data to pcb */
439 /* check AH/ESP integrity. */
440 if (ipsec4_in_reject(m, last))
443 #endif /*FAST_IPSEC*/
444 if ((n = m_copypacket(m, MB_DONTWAIT)) != NULL)
445 udp_append(last, ip, n,
446 iphlen + sizeof(struct udphdr),
457 * Don't look for additional matches if this one does
458 * not have either the SO_REUSEPORT or SO_REUSEADDR
459 * socket options set. This heuristic avoids searching
460 * through all pcbs in the common case of a non-shared
461 * port. It * assumes that an application will never
462 * clear these options after setting them.
464 if (!(last->inp_socket->so_options &
465 (SO_REUSEPORT | SO_REUSEADDR)))
471 * No matching pcb found; discard datagram.
472 * (No need to send an ICMP Port Unreachable
473 * for a broadcast or multicast datgram.)
475 udp_stat.udps_noportbcast++;
479 /* check AH/ESP integrity. */
480 if (ipsec4_in_reject_so(m, last->inp_socket)) {
481 ipsecstat.in_polvio++;
486 /* check AH/ESP integrity. */
487 if (ipsec4_in_reject(m, last))
489 #endif /*FAST_IPSEC*/
490 udp_append(last, ip, m, iphlen + sizeof(struct udphdr),
498 return(IPPROTO_DONE);
501 * Locate pcb for datagram.
503 inp = in_pcblookup_pkthash(&udbinfo, ip->ip_src, uh->uh_sport,
504 ip->ip_dst, uh->uh_dport, 1, m->m_pkthdr.rcvif,
505 udp_reuseport_ext ? m : NULL);
508 char buf[sizeof "aaa.bbb.ccc.ddd"];
510 strcpy(buf, inet_ntoa(ip->ip_dst));
512 "Connection attempt to UDP %s:%d from %s:%d\n",
513 buf, ntohs(uh->uh_dport), inet_ntoa(ip->ip_src),
514 ntohs(uh->uh_sport));
516 udp_stat.udps_noport++;
517 if (m->m_flags & (M_BCAST | M_MCAST)) {
518 udp_stat.udps_noportbcast++;
524 if (badport_bandlim(BANDLIM_ICMP_UNREACH) < 0)
528 ip->ip_len += iphlen;
529 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0);
530 return(IPPROTO_DONE);
533 if (ipsec4_in_reject_so(m, inp->inp_socket)) {
534 ipsecstat.in_polvio++;
539 if (ipsec4_in_reject(m, inp))
541 #endif /*FAST_IPSEC*/
543 * Check the minimum TTL for socket.
545 if (ip->ip_ttl < inp->inp_ip_minttl)
549 * Construct sockaddr format source address.
550 * Stuff source address and datagram in user buffer.
552 udp_in.sin_port = uh->uh_sport;
553 udp_in.sin_addr = ip->ip_src;
554 if ((inp->inp_flags & INP_CONTROLOPTS) ||
555 (inp->inp_socket->so_options & SO_TIMESTAMP)) {
557 if (inp->inp_vflag & INP_IPV6) {
560 ip_2_ip6_hdr(&udp_ip6.uip6_ip6, ip);
561 savedflags = inp->inp_flags;
562 inp->inp_flags &= ~INP_UNMAPPABLEOPTS;
563 ip6_savecontrol(inp, &opts, &udp_ip6.uip6_ip6, m);
564 inp->inp_flags = savedflags;
567 ip_savecontrol(inp, &opts, ip, m);
569 m_adj(m, iphlen + sizeof(struct udphdr));
571 if (inp->inp_vflag & INP_IPV6) {
572 in6_sin_2_v4mapsin6(&udp_in, &udp_in6.uin6_sin);
573 append_sa = (struct sockaddr *)&udp_in6;
576 append_sa = (struct sockaddr *)&udp_in;
578 lwkt_gettoken(&inp->inp_socket->so_rcv.ssb_token);
579 if (ssb_appendaddr(&inp->inp_socket->so_rcv, append_sa, m, opts) == 0) {
580 udp_stat.udps_fullsock++;
581 lwkt_reltoken(&inp->inp_socket->so_rcv.ssb_token);
584 lwkt_reltoken(&inp->inp_socket->so_rcv.ssb_token);
585 sorwakeup(inp->inp_socket);
586 return(IPPROTO_DONE);
591 return(IPPROTO_DONE);
596 ip_2_ip6_hdr(struct ip6_hdr *ip6, struct ip *ip)
598 bzero(ip6, sizeof *ip6);
600 ip6->ip6_vfc = IPV6_VERSION;
601 ip6->ip6_plen = ip->ip_len;
602 ip6->ip6_nxt = ip->ip_p;
603 ip6->ip6_hlim = ip->ip_ttl;
604 ip6->ip6_src.s6_addr32[2] = ip6->ip6_dst.s6_addr32[2] =
606 ip6->ip6_src.s6_addr32[3] = ip->ip_src.s_addr;
607 ip6->ip6_dst.s6_addr32[3] = ip->ip_dst.s_addr;
612 * subroutine of udp_input(), mainly for source code readability.
613 * caller must properly init udp_ip6 and udp_in6 beforehand.
616 udp_append(struct inpcb *last, struct ip *ip, struct mbuf *n, int off,
617 struct sockaddr_in *udp_in,
618 struct udp_in6 *udp_in6, struct udp_ip6 *udp_ip6)
620 struct sockaddr *append_sa;
621 struct mbuf *opts = NULL;
623 if (last->inp_flags & INP_CONTROLOPTS ||
624 last->inp_socket->so_options & SO_TIMESTAMP) {
626 if (last->inp_vflag & INP_IPV6) {
629 if (udp_ip6->uip6_init_done == 0) {
630 ip_2_ip6_hdr(&udp_ip6->uip6_ip6, ip);
631 udp_ip6->uip6_init_done = 1;
633 savedflags = last->inp_flags;
634 last->inp_flags &= ~INP_UNMAPPABLEOPTS;
635 ip6_savecontrol(last, &opts, &udp_ip6->uip6_ip6, n);
636 last->inp_flags = savedflags;
639 ip_savecontrol(last, &opts, ip, n);
642 if (last->inp_vflag & INP_IPV6) {
643 if (udp_in6->uin6_init_done == 0) {
644 in6_sin_2_v4mapsin6(udp_in, &udp_in6->uin6_sin);
645 udp_in6->uin6_init_done = 1;
647 append_sa = (struct sockaddr *)&udp_in6->uin6_sin;
650 append_sa = (struct sockaddr *)udp_in;
652 lwkt_gettoken(&last->inp_socket->so_rcv.ssb_token);
653 if (ssb_appendaddr(&last->inp_socket->so_rcv, append_sa, n, opts) == 0) {
657 udp_stat.udps_fullsock++;
659 sorwakeup(last->inp_socket);
661 lwkt_reltoken(&last->inp_socket->so_rcv.ssb_token);
665 * Notify a udp user of an asynchronous error;
666 * just wake up so that he can collect error status.
669 udp_notify(struct inpcb *inp, int error)
671 inp->inp_socket->so_error = error;
672 sorwakeup(inp->inp_socket);
673 sowwakeup(inp->inp_socket);
676 struct netmsg_udp_notify {
677 struct netmsg_base base;
678 void (*nm_notify)(struct inpcb *, int);
679 struct in_addr nm_faddr;
684 udp_notifyall_oncpu(netmsg_t msg)
686 struct netmsg_udp_notify *nm = (struct netmsg_udp_notify *)msg;
691 in_pcbnotifyall(&udbinfo.pcblisthead, nm->nm_faddr,
692 nm->nm_arg, nm->nm_notify);
693 lwkt_replymsg(&nm->base.lmsg, 0);
696 /* XXX currently udp only runs on cpu 0 */
697 nextcpu = mycpuid + 1;
698 if (nextcpu < ncpus2)
699 lwkt_forwardmsg(netisr_cpuport(nextcpu), &nm->base.lmsg);
701 lwkt_replymsg(&nmsg->base.lmsg, 0);
706 udp_rtchange(struct inpcb *inp, int err)
708 /* XXX Nuke this, once UDP inpcbs are CPU localized */
709 if (inp->inp_route.ro_rt && inp->inp_route.ro_rt->rt_cpuid == mycpuid) {
710 rtfree(inp->inp_route.ro_rt);
711 inp->inp_route.ro_rt = NULL;
713 * A new route can be allocated the next time
714 * output is attempted.
720 udp_ctlinput(netmsg_t msg)
722 struct sockaddr *sa = msg->ctlinput.nm_arg;
723 struct ip *ip = msg->ctlinput.nm_extra;
724 int cmd = msg->ctlinput.nm_cmd;
726 void (*notify) (struct inpcb *, int) = udp_notify;
727 struct in_addr faddr;
730 KKASSERT(&curthread->td_msgport == netisr_cpuport(0));
732 faddr = ((struct sockaddr_in *)sa)->sin_addr;
733 if (sa->sa_family != AF_INET || faddr.s_addr == INADDR_ANY)
736 if (PRC_IS_REDIRECT(cmd)) {
738 notify = udp_rtchange;
739 } else if (cmd == PRC_HOSTDEAD) {
741 } else if ((unsigned)cmd >= PRC_NCMDS || inetctlerrmap[cmd] == 0) {
746 uh = (struct udphdr *)((caddr_t)ip + (ip->ip_hl << 2));
747 inp = in_pcblookup_hash(&udbinfo, faddr, uh->uh_dport,
748 ip->ip_src, uh->uh_sport, 0, NULL);
749 if (inp != NULL && inp->inp_socket != NULL)
750 (*notify)(inp, inetctlerrmap[cmd]);
751 } else if (PRC_IS_REDIRECT(cmd)) {
752 struct netmsg_udp_notify *nm;
754 KKASSERT(&curthread->td_msgport == netisr_cpuport(0));
755 nm = kmalloc(sizeof(*nm), M_LWKTMSG, M_INTWAIT);
756 netmsg_init(&nm->base, NULL, &netisr_afree_rport,
757 0, udp_notifyall_oncpu);
758 nm->nm_faddr = faddr;
759 nm->nm_arg = inetctlerrmap[cmd];
760 nm->nm_notify = notify;
761 lwkt_sendmsg(netisr_cpuport(0), &nm->base.lmsg);
764 * XXX We should forward msg upon PRC_HOSTHEAD and ip == NULL,
765 * once UDP inpcbs are CPU localized
767 KKASSERT(&curthread->td_msgport == netisr_cpuport(0));
768 in_pcbnotifyall(&udbinfo.pcblisthead, faddr, inetctlerrmap[cmd],
772 lwkt_replymsg(&msg->lmsg, 0);
776 udp_pcblist(SYSCTL_HANDLER_ARGS)
782 error = in_pcblist_global_nomarker(oidp, arg1, arg2, req, &xi, &nxi);
786 KKASSERT(xi == NULL);
790 KKASSERT(xi == NULL);
794 for (i = 0; i < nxi; ++i) {
795 error = SYSCTL_OUT(req, &xi[i], sizeof(xi[i]));
803 SYSCTL_PROC(_net_inet_udp, UDPCTL_PCBLIST, pcblist, CTLFLAG_RD, &udbinfo, 0,
804 udp_pcblist, "S,xinpcb", "List of active UDP sockets");
807 udp_getcred(SYSCTL_HANDLER_ARGS)
809 struct sockaddr_in addrs[2];
810 struct ucred cred0, *cred = NULL;
814 error = priv_check(req->td, PRIV_ROOT);
817 error = SYSCTL_IN(req, addrs, sizeof addrs);
822 inp = in_pcblookup_hash(&udbinfo, addrs[1].sin_addr, addrs[1].sin_port,
823 addrs[0].sin_addr, addrs[0].sin_port, 1, NULL);
824 if (inp == NULL || inp->inp_socket == NULL) {
827 if (inp->inp_socket->so_cred != NULL) {
828 cred0 = *(inp->inp_socket->so_cred);
837 return SYSCTL_OUT(req, cred, sizeof(struct ucred));
840 SYSCTL_PROC(_net_inet_udp, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW,
841 0, 0, udp_getcred, "S,ucred", "Get the ucred of a UDP connection");
844 udp_output(struct inpcb *inp, struct mbuf *m, struct sockaddr *dstaddr,
845 struct thread *td, int flags, boolean_t held_td)
848 int len = m->m_pkthdr.len;
849 struct sockaddr_in *sin; /* really is initialized before use */
850 int error = 0, lport_any = 0;
852 if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) {
857 if (inp->inp_lport == 0) { /* unbound socket */
858 error = in_pcbbind(inp, NULL, td);
862 udbinfo_barrier_set();
863 in_pcbinswildcardhash(inp);
864 udbinfo_barrier_rem();
868 if (dstaddr != NULL) { /* destination address specified */
869 if (inp->inp_faddr.s_addr != INADDR_ANY) {
870 /* already connected */
874 sin = (struct sockaddr_in *)dstaddr;
875 if (!prison_remote_ip(td, (struct sockaddr *)&sin)) {
876 error = EAFNOSUPPORT; /* IPv6 only jail */
880 if (inp->inp_faddr.s_addr == INADDR_ANY) {
881 /* no destination specified and not already connected */
889 * Calculate data length and get a mbuf
890 * for UDP and IP headers.
892 M_PREPEND(m, sizeof(struct udpiphdr), MB_DONTWAIT);
899 * Fill in mbuf with extended UDP header
900 * and addresses and length put into network format.
902 ui = mtod(m, struct udpiphdr *);
903 bzero(ui->ui_x1, sizeof ui->ui_x1); /* XXX still needed? */
904 ui->ui_pr = IPPROTO_UDP;
907 * Set destination address.
909 if (dstaddr != NULL) { /* use specified destination */
910 ui->ui_dst = sin->sin_addr;
911 ui->ui_dport = sin->sin_port;
912 } else { /* use connected destination */
913 ui->ui_dst = inp->inp_faddr;
914 ui->ui_dport = inp->inp_fport;
918 * Set source address.
920 if (inp->inp_laddr.s_addr == INADDR_ANY ||
921 IN_MULTICAST(ntohl(inp->inp_laddr.s_addr))) {
922 struct sockaddr_in *if_sin;
924 if (dstaddr == NULL) {
926 * connect() had (or should have) failed because
927 * the interface had no IP address, but the
928 * application proceeded to call send() anyways.
934 /* Look up outgoing interface. */
935 error = in_pcbladdr_find(inp, dstaddr, &if_sin, td, 1);
938 ui->ui_src = if_sin->sin_addr; /* use address of interface */
940 ui->ui_src = inp->inp_laddr; /* use non-null bound address */
942 ui->ui_sport = inp->inp_lport;
943 KASSERT(inp->inp_lport != 0, ("inp lport should have been bound"));
946 * Release the original thread, since it is no longer used
953 ui->ui_ulen = htons((u_short)len + sizeof(struct udphdr));
956 * Set up checksum and output datagram.
959 ui->ui_sum = in_pseudo(ui->ui_src.s_addr, ui->ui_dst.s_addr,
960 htons((u_short)len + sizeof(struct udphdr) + IPPROTO_UDP));
961 m->m_pkthdr.csum_flags = CSUM_UDP;
962 m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
963 m->m_pkthdr.csum_thlen = sizeof(struct udphdr);
967 ((struct ip *)ui)->ip_len = sizeof(struct udpiphdr) + len;
968 ((struct ip *)ui)->ip_ttl = inp->inp_ip_ttl; /* XXX */
969 ((struct ip *)ui)->ip_tos = inp->inp_ip_tos; /* XXX */
970 udp_stat.udps_opackets++;
972 error = ip_output(m, inp->inp_options, &inp->inp_route,
973 (inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST)) |
974 flags | IP_DEBUGROUTE,
975 inp->inp_moptions, inp);
978 * If this is the first data gram sent on an unbound and unconnected
979 * UDP socket, lport will be changed in this function. If target
980 * CPU after this lport changing is no longer the current CPU, then
981 * free the route entry allocated on the current CPU.
984 if (udp_addrcpu(inp->inp_faddr.s_addr, inp->inp_fport,
985 inp->inp_laddr.s_addr, inp->inp_lport) != mycpuid) {
987 struct route *ro = &inp->inp_route;
989 if (ro->ro_rt != NULL)
991 bzero(ro, sizeof(*ro));
993 panic("UDP activity should only be in netisr0");
1006 u_long udp_sendspace = 9216; /* really max datagram size */
1007 /* 40 1K datagrams */
1008 SYSCTL_INT(_net_inet_udp, UDPCTL_MAXDGRAM, maxdgram, CTLFLAG_RW,
1009 &udp_sendspace, 0, "Maximum outgoing UDP datagram size");
1011 u_long udp_recvspace = 40 * (1024 +
1013 sizeof(struct sockaddr_in6)
1015 sizeof(struct sockaddr_in)
1018 SYSCTL_INT(_net_inet_udp, UDPCTL_RECVSPACE, recvspace, CTLFLAG_RW,
1019 &udp_recvspace, 0, "Maximum incoming UDP datagram size");
1022 * NOTE: (so) is referenced from soabort*() and netmsg_pru_abort()
1023 * will sofree() it when we return.
1026 udp_abort(netmsg_t msg)
1028 struct socket *so = msg->abort.base.nm_so;
1032 KKASSERT(&curthread->td_msgport == netisr_cpuport(0));
1036 soisdisconnected(so);
1038 udbinfo_barrier_set();
1040 udbinfo_barrier_rem();
1045 lwkt_replymsg(&msg->abort.base.lmsg, error);
1049 udp_attach(netmsg_t msg)
1051 struct socket *so = msg->attach.base.nm_so;
1052 struct pru_attach_info *ai = msg->attach.nm_ai;
1056 KKASSERT(&curthread->td_msgport == netisr_cpuport(0));
1063 error = soreserve(so, udp_sendspace, udp_recvspace, ai->sb_rlimit);
1067 udbinfo_barrier_set();
1068 error = in_pcballoc(so, &udbinfo);
1069 udbinfo_barrier_rem();
1075 * Set default port for protocol processing prior to bind/connect.
1077 sosetport(so, netisr_cpuport(0));
1079 inp = (struct inpcb *)so->so_pcb;
1080 inp->inp_vflag |= INP_IPV4;
1081 inp->inp_ip_ttl = ip_defttl;
1084 lwkt_replymsg(&msg->attach.base.lmsg, error);
1088 udp_bind(netmsg_t msg)
1090 struct socket *so = msg->bind.base.nm_so;
1091 struct sockaddr *nam = msg->bind.nm_nam;
1092 struct thread *td = msg->bind.nm_td;
1093 struct sockaddr_in *sin = (struct sockaddr_in *)nam;
1099 error = in_pcbbind(inp, nam, td);
1101 if (sin->sin_addr.s_addr != INADDR_ANY)
1102 inp->inp_flags |= INP_WASBOUND_NOTANY;
1104 udbinfo_barrier_set();
1105 in_pcbinswildcardhash(inp);
1106 udbinfo_barrier_rem();
1111 lwkt_replymsg(&msg->bind.base.lmsg, error);
1115 udp_connect(netmsg_t msg)
1117 struct socket *so = msg->connect.base.nm_so;
1118 struct sockaddr *nam = msg->connect.nm_nam;
1119 struct thread *td = msg->connect.nm_td;
1121 struct sockaddr_in *sin = (struct sockaddr_in *)nam;
1122 struct sockaddr_in *if_sin;
1126 KKASSERT(&curthread->td_msgport == netisr_cpuport(0));
1134 if (msg->connect.nm_flags & PRUC_RECONNECT) {
1135 panic("UDP does not support RECONNECT");
1137 msg->connect.nm_flags &= ~PRUC_RECONNECT;
1138 in_pcblink(inp, &udbinfo);
1142 if (inp->inp_faddr.s_addr != INADDR_ANY) {
1149 * Bind if we have to
1151 if (td->td_proc && td->td_proc->p_ucred->cr_prison != NULL &&
1152 inp->inp_laddr.s_addr == INADDR_ANY) {
1153 error = in_pcbbind(inp, NULL, td);
1159 * Calculate the correct protocol processing thread. The connect
1160 * operation must run there.
1162 error = in_pcbladdr(inp, nam, &if_sin, td);
1165 if (!prison_remote_ip(td, nam)) {
1166 error = EAFNOSUPPORT; /* IPv6 only jail */
1170 port = udp_addrport(sin->sin_addr.s_addr, sin->sin_port,
1171 inp->inp_laddr.s_addr, inp->inp_lport);
1172 if (port != &curthread->td_msgport) {
1174 struct route *ro = &inp->inp_route;
1177 * in_pcbladdr() may have allocated a route entry for us
1178 * on the current CPU, but we need a route entry on the
1179 * inpcb's owner CPU, so free it here.
1181 if (ro->ro_rt != NULL)
1183 bzero(ro, sizeof(*ro));
1186 * We are moving the protocol processing port the socket
1187 * is on, we have to unlink here and re-link on the
1190 in_pcbunlink(so->so_pcb, &udbinfo);
1191 /* in_pcbunlink(so->so_pcb, &udbinfo[mycpu->gd_cpuid]); */
1192 sosetport(so, port);
1193 msg->connect.nm_flags |= PRUC_RECONNECT;
1194 msg->connect.base.nm_dispatch = udp_connect;
1196 lwkt_forwardmsg(port, &msg->connect.base.lmsg);
1197 /* msg invalid now */
1200 panic("UDP activity should only be in netisr0");
1203 KKASSERT(port == &curthread->td_msgport);
1204 error = udp_connect_oncpu(so, td, sin, if_sin);
1206 KKASSERT(msg->connect.nm_m == NULL);
1207 lwkt_replymsg(&msg->connect.base.lmsg, error);
1211 udp_connect_oncpu(struct socket *so, struct thread *td,
1212 struct sockaddr_in *sin, struct sockaddr_in *if_sin)
1217 udbinfo_barrier_set();
1220 if (inp->inp_flags & INP_WILDCARD)
1221 in_pcbremwildcardhash(inp);
1222 error = in_pcbconnect(inp, (struct sockaddr *)sin, td);
1226 * No more errors can occur, finish adjusting the socket
1227 * and change the processing port to reflect the connected
1228 * socket. Once set we can no longer safely mess with the
1232 } else if (error == EAFNOSUPPORT) { /* connection dissolved */
1234 * Follow traditional BSD behavior and retain
1235 * the local port binding. But, fix the old misbehavior
1236 * of overwriting any previously bound local address.
1238 if (!(inp->inp_flags & INP_WASBOUND_NOTANY))
1239 inp->inp_laddr.s_addr = INADDR_ANY;
1240 in_pcbinswildcardhash(inp);
1243 udbinfo_barrier_rem();
1248 udp_detach(netmsg_t msg)
1250 struct socket *so = msg->detach.base.nm_so;
1254 KKASSERT(&curthread->td_msgport == netisr_cpuport(0));
1258 udbinfo_barrier_set();
1260 udbinfo_barrier_rem();
1265 lwkt_replymsg(&msg->detach.base.lmsg, error);
1269 udp_disconnect(netmsg_t msg)
1271 struct socket *so = msg->disconnect.base.nm_so;
1276 KKASSERT(&curthread->td_msgport == netisr_cpuport(0));
1283 if (inp->inp_faddr.s_addr == INADDR_ANY) {
1290 udbinfo_barrier_set();
1291 in_pcbdisconnect(inp);
1292 udbinfo_barrier_rem();
1294 soclrstate(so, SS_ISCONNECTED); /* XXX */
1297 ro = &inp->inp_route;
1298 if (ro->ro_rt != NULL)
1300 bzero(ro, sizeof(*ro));
1303 lwkt_replymsg(&msg->disconnect.base.lmsg, error);
1307 udp_send(netmsg_t msg)
1309 struct socket *so = msg->send.base.nm_so;
1310 struct mbuf *m = msg->send.nm_m;
1311 struct sockaddr *addr = msg->send.nm_addr;
1312 int pru_flags = msg->send.nm_flags;
1316 KKASSERT(&curthread->td_msgport == netisr_cpuport(0));
1317 KKASSERT(msg->send.nm_control == NULL);
1321 struct thread *td = msg->send.nm_td;
1324 if (pru_flags & PRUS_DONTROUTE)
1325 flags |= SO_DONTROUTE;
1326 error = udp_output(inp, m, addr, td, flags,
1327 (pru_flags & PRUS_HELDTD) ? TRUE : FALSE);
1329 if (pru_flags & PRUS_HELDTD)
1330 lwkt_rele(msg->send.nm_td);
1335 if (pru_flags & PRUS_FREEADDR)
1336 kfree(addr, M_SONAME);
1338 if ((pru_flags & PRUS_NOREPLY) == 0)
1339 lwkt_replymsg(&msg->send.base.lmsg, error);
1343 udp_shutdown(netmsg_t msg)
1345 struct socket *so = msg->shutdown.base.nm_so;
1349 KKASSERT(&curthread->td_msgport == netisr_cpuport(0));
1358 lwkt_replymsg(&msg->shutdown.base.lmsg, error);
1364 lwkt_serialize_enter(&udbinfo_slize);
1368 udbinfo_unlock(void)
1370 lwkt_serialize_exit(&udbinfo_slize);
1374 udbinfo_barrier_set(void)
1376 netisr_barrier_set(udbinfo_br);
1381 udbinfo_barrier_rem(void)
1384 netisr_barrier_rem(udbinfo_br);
1387 struct pr_usrreqs udp_usrreqs = {
1388 .pru_abort = udp_abort,
1389 .pru_accept = pr_generic_notsupp,
1390 .pru_attach = udp_attach,
1391 .pru_bind = udp_bind,
1392 .pru_connect = udp_connect,
1393 .pru_connect2 = pr_generic_notsupp,
1394 .pru_control = in_control_dispatch,
1395 .pru_detach = udp_detach,
1396 .pru_disconnect = udp_disconnect,
1397 .pru_listen = pr_generic_notsupp,
1398 .pru_peeraddr = in_setpeeraddr_dispatch,
1399 .pru_rcvd = pr_generic_notsupp,
1400 .pru_rcvoob = pr_generic_notsupp,
1401 .pru_send = udp_send,
1402 .pru_sense = pru_sense_null,
1403 .pru_shutdown = udp_shutdown,
1404 .pru_sockaddr = in_setsockaddr_dispatch,
1405 .pru_sosend = sosendudp,
1406 .pru_soreceive = soreceive