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>
98 #include <netinet/in.h>
99 #include <netinet/in_systm.h>
100 #include <netinet/ip.h>
102 #include <netinet/ip6.h>
104 #include <netinet/in_pcb.h>
105 #include <netinet/in_var.h>
106 #include <netinet/ip_var.h>
108 #include <netinet6/ip6_var.h>
110 #include <netinet/ip_icmp.h>
111 #include <netinet/icmp_var.h>
112 #include <netinet/udp.h>
113 #include <netinet/udp_var.h>
116 #include <netproto/ipsec/ipsec.h>
120 #include <netinet6/ipsec.h>
124 * UDP protocol implementation.
125 * Per RFC 768, August, 1980.
128 static int udpcksum = 1;
130 static int udpcksum = 0; /* XXX */
132 SYSCTL_INT(_net_inet_udp, UDPCTL_CHECKSUM, checksum, CTLFLAG_RW,
133 &udpcksum, 0, "Enable checksumming of UDP packets");
136 SYSCTL_INT(_net_inet_udp, OID_AUTO, log_in_vain, CTLFLAG_RW,
137 &log_in_vain, 0, "Log all incoming UDP packets");
139 static int blackhole = 0;
140 SYSCTL_INT(_net_inet_udp, OID_AUTO, blackhole, CTLFLAG_RW,
141 &blackhole, 0, "Do not send port unreachables for refused connects");
143 static int strict_mcast_mship = 1;
144 SYSCTL_INT(_net_inet_udp, OID_AUTO, strict_mcast_mship, CTLFLAG_RW,
145 &strict_mcast_mship, 0, "Only send multicast to member sockets");
147 int udp_sosend_async = 1;
148 SYSCTL_INT(_net_inet_udp, OID_AUTO, sosend_async, CTLFLAG_RW,
149 &udp_sosend_async, 0, "UDP asynchronized pru_send");
151 int udp_sosend_prepend = 1;
152 SYSCTL_INT(_net_inet_udp, OID_AUTO, sosend_prepend, CTLFLAG_RW,
153 &udp_sosend_prepend, 0,
154 "Prepend enough space for proto and link header in pru_send");
156 struct inpcbinfo udbinfo;
158 static struct netisr_barrier *udbinfo_br;
159 static struct lwkt_serialize udbinfo_slize = LWKT_SERIALIZE_INITIALIZER;
162 #define UDBHASHSIZE 16
165 struct udpstat udpstat_percpu[MAXCPU] __cachealign;
167 static struct sockaddr_in udp_in = { sizeof udp_in, AF_INET };
170 struct sockaddr_in6 uin6_sin;
171 u_char uin6_init_done : 1;
173 { sizeof udp_in6.uin6_sin, AF_INET6 },
177 struct ip6_hdr uip6_ip6;
178 u_char uip6_init_done : 1;
182 static void udp_append (struct inpcb *last, struct ip *ip,
183 struct mbuf *n, int off);
185 static void ip_2_ip6_hdr (struct ip6_hdr *ip6, struct ip *ip);
188 static int udp_connect_oncpu(struct socket *so, struct thread *td,
189 struct sockaddr_in *sin, struct sockaddr_in *if_sin);
190 static int udp_output (struct inpcb *, struct mbuf *, struct sockaddr *,
191 struct thread *, int);
198 in_pcbinfo_init(&udbinfo);
199 udbinfo.hashbase = hashinit(UDBHASHSIZE, M_PCB, &udbinfo.hashmask);
200 udbinfo.porthashbase = hashinit(UDBHASHSIZE, M_PCB,
201 &udbinfo.porthashmask);
202 udbinfo.wildcardhashbase = hashinit(UDBHASHSIZE, M_PCB,
203 &udbinfo.wildcardhashmask);
204 udbinfo.ipi_size = sizeof(struct inpcb);
206 udbinfo_br = netisr_barrier_create();
209 * Initialize UDP statistics counters for each CPU.
211 for (cpu = 0; cpu < ncpus; ++cpu)
212 bzero(&udpstat_percpu[cpu], sizeof(struct udpstat));
216 sysctl_udpstat(SYSCTL_HANDLER_ARGS)
220 for (cpu = 0; cpu < ncpus; ++cpu) {
221 if ((error = SYSCTL_OUT(req, &udpstat_percpu[cpu],
222 sizeof(struct udpstat))))
224 if ((error = SYSCTL_IN(req, &udpstat_percpu[cpu],
225 sizeof(struct udpstat))))
231 SYSCTL_PROC(_net_inet_udp, UDPCTL_STATS, stats, (CTLTYPE_OPAQUE | CTLFLAG_RW),
232 0, 0, sysctl_udpstat, "S,udpstat", "UDP statistics");
235 * Check multicast packets to make sure they are only sent to sockets with
236 * multicast memberships for the packet's destination address and arrival
237 * interface. Multicast packets to multicast-unaware sockets are also
240 * Returns 0 if the packet is acceptable, -1 if it is not.
243 check_multicast_membership(struct ip *ip, struct inpcb *inp, struct mbuf *m)
246 struct ip_moptions *mopt;
248 if (strict_mcast_mship == 0 ||
249 !IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
252 mopt = inp->inp_moptions;
255 for (mshipno = 0; mshipno < mopt->imo_num_memberships; ++mshipno) {
256 struct in_multi *maddr = mopt->imo_membership[mshipno];
258 if (ip->ip_dst.s_addr == maddr->inm_addr.s_addr &&
259 m->m_pkthdr.rcvif == maddr->inm_ifp) {
267 udp_input(struct mbuf **mp, int *offp, int proto)
274 struct mbuf *opts = NULL;
277 struct sockaddr *append_sa;
284 udp_stat.udps_ipackets++;
287 * Strip IP options, if any; should skip this,
288 * make available to user, and use on returned packets,
289 * but we don't yet have a way to check the checksum
290 * with options still present.
292 if (iphlen > sizeof(struct ip)) {
294 iphlen = sizeof(struct ip);
298 * IP and UDP headers are together in first mbuf.
299 * Already checked and pulled up in ip_demux().
301 KASSERT(m->m_len >= iphlen + sizeof(struct udphdr),
302 ("UDP header not in one mbuf"));
304 ip = mtod(m, struct ip *);
305 uh = (struct udphdr *)((caddr_t)ip + iphlen);
307 /* destination port of 0 is illegal, based on RFC768. */
308 if (uh->uh_dport == 0)
312 * Make mbuf data length reflect UDP length.
313 * If not enough data to reflect UDP length, drop.
315 len = ntohs((u_short)uh->uh_ulen);
316 if (ip->ip_len != len) {
317 if (len > ip->ip_len || len < sizeof(struct udphdr)) {
318 udp_stat.udps_badlen++;
321 m_adj(m, len - ip->ip_len);
322 /* ip->ip_len = len; */
325 * Save a copy of the IP header in case we want restore it
326 * for sending an ICMP error message in response.
331 * Checksum extended UDP header and data.
334 if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) {
335 if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR)
336 uh->uh_sum = m->m_pkthdr.csum_data;
338 uh->uh_sum = in_pseudo(ip->ip_src.s_addr,
339 ip->ip_dst.s_addr, htonl((u_short)len +
340 m->m_pkthdr.csum_data + IPPROTO_UDP));
341 uh->uh_sum ^= 0xffff;
345 bcopy(((struct ipovly *)ip)->ih_x1, b, 9);
346 bzero(((struct ipovly *)ip)->ih_x1, 9);
347 ((struct ipovly *)ip)->ih_len = uh->uh_ulen;
348 uh->uh_sum = in_cksum(m, len + sizeof(struct ip));
349 bcopy(b, ((struct ipovly *)ip)->ih_x1, 9);
352 udp_stat.udps_badsum++;
354 return(IPPROTO_DONE);
357 udp_stat.udps_nosum++;
359 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) ||
360 in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif)) {
364 * Deliver a multicast or broadcast datagram to *all* sockets
365 * for which the local and remote addresses and ports match
366 * those of the incoming datagram. This allows more than
367 * one process to receive multi/broadcasts on the same port.
368 * (This really ought to be done for unicast datagrams as
369 * well, but that would cause problems with existing
370 * applications that open both address-specific sockets and
371 * a wildcard socket listening to the same port -- they would
372 * end up receiving duplicates of every unicast datagram.
373 * Those applications open the multiple sockets to overcome an
374 * inadequacy of the UDP socket interface, but for backwards
375 * compatibility we avoid the problem here rather than
376 * fixing the interface. Maybe 4.5BSD will remedy this?)
380 * Construct sockaddr format source address.
382 udp_in.sin_port = uh->uh_sport;
383 udp_in.sin_addr = ip->ip_src;
385 * Locate pcb(s) for datagram.
386 * (Algorithm copied from raw_intr().)
390 udp_in6.uin6_init_done = udp_ip6.uip6_init_done = 0;
392 LIST_FOREACH(inp, &udbinfo.pcblisthead, inp_list) {
393 KKASSERT((inp->inp_flags & INP_PLACEMARKER) == 0);
395 if (!(inp->inp_vflag & INP_IPV4))
398 if (inp->inp_lport != uh->uh_dport)
400 if (inp->inp_laddr.s_addr != INADDR_ANY) {
401 if (inp->inp_laddr.s_addr !=
405 if (inp->inp_faddr.s_addr != INADDR_ANY) {
406 if (inp->inp_faddr.s_addr !=
408 inp->inp_fport != uh->uh_sport)
412 if (check_multicast_membership(ip, inp, m) < 0)
419 /* check AH/ESP integrity. */
420 if (ipsec4_in_reject_so(m, last->inp_socket))
421 ipsecstat.in_polvio++;
422 /* do not inject data to pcb */
426 /* check AH/ESP integrity. */
427 if (ipsec4_in_reject(m, last))
430 #endif /*FAST_IPSEC*/
431 if ((n = m_copypacket(m, MB_DONTWAIT)) != NULL)
432 udp_append(last, ip, n,
434 sizeof(struct udphdr));
438 * Don't look for additional matches if this one does
439 * not have either the SO_REUSEPORT or SO_REUSEADDR
440 * socket options set. This heuristic avoids searching
441 * through all pcbs in the common case of a non-shared
442 * port. It * assumes that an application will never
443 * clear these options after setting them.
445 if (!(last->inp_socket->so_options &
446 (SO_REUSEPORT | SO_REUSEADDR)))
452 * No matching pcb found; discard datagram.
453 * (No need to send an ICMP Port Unreachable
454 * for a broadcast or multicast datgram.)
456 udp_stat.udps_noportbcast++;
460 /* check AH/ESP integrity. */
461 if (ipsec4_in_reject_so(m, last->inp_socket)) {
462 ipsecstat.in_polvio++;
467 /* check AH/ESP integrity. */
468 if (ipsec4_in_reject(m, last))
470 #endif /*FAST_IPSEC*/
471 udp_append(last, ip, m, iphlen + sizeof(struct udphdr));
472 return(IPPROTO_DONE);
475 * Locate pcb for datagram.
477 inp = in_pcblookup_hash(&udbinfo, ip->ip_src, uh->uh_sport,
478 ip->ip_dst, uh->uh_dport, 1, m->m_pkthdr.rcvif);
481 char buf[sizeof "aaa.bbb.ccc.ddd"];
483 strcpy(buf, inet_ntoa(ip->ip_dst));
485 "Connection attempt to UDP %s:%d from %s:%d\n",
486 buf, ntohs(uh->uh_dport), inet_ntoa(ip->ip_src),
487 ntohs(uh->uh_sport));
489 udp_stat.udps_noport++;
490 if (m->m_flags & (M_BCAST | M_MCAST)) {
491 udp_stat.udps_noportbcast++;
497 if (badport_bandlim(BANDLIM_ICMP_UNREACH) < 0)
501 ip->ip_len += iphlen;
502 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0);
503 return(IPPROTO_DONE);
506 if (ipsec4_in_reject_so(m, inp->inp_socket)) {
507 ipsecstat.in_polvio++;
512 if (ipsec4_in_reject(m, inp))
514 #endif /*FAST_IPSEC*/
516 * Check the minimum TTL for socket.
518 if (ip->ip_ttl < inp->inp_ip_minttl)
522 * Construct sockaddr format source address.
523 * Stuff source address and datagram in user buffer.
525 udp_in.sin_port = uh->uh_sport;
526 udp_in.sin_addr = ip->ip_src;
527 if ((inp->inp_flags & INP_CONTROLOPTS) ||
528 (inp->inp_socket->so_options & SO_TIMESTAMP)) {
530 if (inp->inp_vflag & INP_IPV6) {
533 ip_2_ip6_hdr(&udp_ip6.uip6_ip6, ip);
534 savedflags = inp->inp_flags;
535 inp->inp_flags &= ~INP_UNMAPPABLEOPTS;
536 ip6_savecontrol(inp, &opts, &udp_ip6.uip6_ip6, m);
537 inp->inp_flags = savedflags;
540 ip_savecontrol(inp, &opts, ip, m);
542 m_adj(m, iphlen + sizeof(struct udphdr));
544 if (inp->inp_vflag & INP_IPV6) {
545 in6_sin_2_v4mapsin6(&udp_in, &udp_in6.uin6_sin);
546 append_sa = (struct sockaddr *)&udp_in6;
549 append_sa = (struct sockaddr *)&udp_in;
551 lwkt_gettoken(&inp->inp_socket->so_rcv.ssb_token);
552 if (ssb_appendaddr(&inp->inp_socket->so_rcv, append_sa, m, opts) == 0) {
553 udp_stat.udps_fullsock++;
554 lwkt_reltoken(&inp->inp_socket->so_rcv.ssb_token);
557 lwkt_reltoken(&inp->inp_socket->so_rcv.ssb_token);
558 sorwakeup(inp->inp_socket);
559 return(IPPROTO_DONE);
564 return(IPPROTO_DONE);
569 ip_2_ip6_hdr(struct ip6_hdr *ip6, struct ip *ip)
571 bzero(ip6, sizeof *ip6);
573 ip6->ip6_vfc = IPV6_VERSION;
574 ip6->ip6_plen = ip->ip_len;
575 ip6->ip6_nxt = ip->ip_p;
576 ip6->ip6_hlim = ip->ip_ttl;
577 ip6->ip6_src.s6_addr32[2] = ip6->ip6_dst.s6_addr32[2] =
579 ip6->ip6_src.s6_addr32[3] = ip->ip_src.s_addr;
580 ip6->ip6_dst.s6_addr32[3] = ip->ip_dst.s_addr;
585 * subroutine of udp_input(), mainly for source code readability.
586 * caller must properly init udp_ip6 and udp_in6 beforehand.
589 udp_append(struct inpcb *last, struct ip *ip, struct mbuf *n, int off)
591 struct sockaddr *append_sa;
592 struct mbuf *opts = NULL;
594 if (last->inp_flags & INP_CONTROLOPTS ||
595 last->inp_socket->so_options & SO_TIMESTAMP) {
597 if (last->inp_vflag & INP_IPV6) {
600 if (udp_ip6.uip6_init_done == 0) {
601 ip_2_ip6_hdr(&udp_ip6.uip6_ip6, ip);
602 udp_ip6.uip6_init_done = 1;
604 savedflags = last->inp_flags;
605 last->inp_flags &= ~INP_UNMAPPABLEOPTS;
606 ip6_savecontrol(last, &opts, &udp_ip6.uip6_ip6, n);
607 last->inp_flags = savedflags;
610 ip_savecontrol(last, &opts, ip, n);
613 if (last->inp_vflag & INP_IPV6) {
614 if (udp_in6.uin6_init_done == 0) {
615 in6_sin_2_v4mapsin6(&udp_in, &udp_in6.uin6_sin);
616 udp_in6.uin6_init_done = 1;
618 append_sa = (struct sockaddr *)&udp_in6.uin6_sin;
621 append_sa = (struct sockaddr *)&udp_in;
623 lwkt_gettoken(&last->inp_socket->so_rcv.ssb_token);
624 if (ssb_appendaddr(&last->inp_socket->so_rcv, append_sa, n, opts) == 0) {
628 udp_stat.udps_fullsock++;
630 sorwakeup(last->inp_socket);
632 lwkt_reltoken(&last->inp_socket->so_rcv.ssb_token);
636 * Notify a udp user of an asynchronous error;
637 * just wake up so that he can collect error status.
640 udp_notify(struct inpcb *inp, int error)
642 inp->inp_socket->so_error = error;
643 sorwakeup(inp->inp_socket);
644 sowwakeup(inp->inp_socket);
647 struct netmsg_udp_notify {
648 struct netmsg_base base;
649 void (*nm_notify)(struct inpcb *, int);
650 struct in_addr nm_faddr;
655 udp_notifyall_oncpu(netmsg_t msg)
657 struct netmsg_udp_notify *nm = (struct netmsg_udp_notify *)msg;
662 in_pcbnotifyall(&udbinfo.pcblisthead, nm->nm_faddr,
663 nm->nm_arg, nm->nm_notify);
664 lwkt_replymsg(&nm->base.lmsg, 0);
667 /* XXX currently udp only runs on cpu 0 */
668 nextcpu = mycpuid + 1;
669 if (nextcpu < ncpus2)
670 lwkt_forwardmsg(netisr_portfn(nextcpu), &nm->base.lmsg);
672 lwkt_replymsg(&nmsg->base.lmsg, 0);
677 udp_rtchange(struct inpcb *inp, int err)
679 /* XXX Nuke this, once UDP inpcbs are CPU localized */
680 if (inp->inp_route.ro_rt && inp->inp_route.ro_rt->rt_cpuid == mycpuid) {
681 rtfree(inp->inp_route.ro_rt);
682 inp->inp_route.ro_rt = NULL;
684 * A new route can be allocated the next time
685 * output is attempted.
691 udp_ctlinput(netmsg_t msg)
693 struct sockaddr *sa = msg->ctlinput.nm_arg;
694 struct ip *ip = msg->ctlinput.nm_extra;
695 int cmd = msg->ctlinput.nm_cmd;
697 void (*notify) (struct inpcb *, int) = udp_notify;
698 struct in_addr faddr;
701 KKASSERT(&curthread->td_msgport == netisr_portfn(0));
703 faddr = ((struct sockaddr_in *)sa)->sin_addr;
704 if (sa->sa_family != AF_INET || faddr.s_addr == INADDR_ANY)
707 if (PRC_IS_REDIRECT(cmd)) {
709 notify = udp_rtchange;
710 } else if (cmd == PRC_HOSTDEAD) {
712 } else if ((unsigned)cmd >= PRC_NCMDS || inetctlerrmap[cmd] == 0) {
717 uh = (struct udphdr *)((caddr_t)ip + (ip->ip_hl << 2));
718 inp = in_pcblookup_hash(&udbinfo, faddr, uh->uh_dport,
719 ip->ip_src, uh->uh_sport, 0, NULL);
720 if (inp != NULL && inp->inp_socket != NULL)
721 (*notify)(inp, inetctlerrmap[cmd]);
722 } else if (PRC_IS_REDIRECT(cmd)) {
723 struct netmsg_udp_notify *nm;
725 KKASSERT(&curthread->td_msgport == netisr_portfn(0));
726 nm = kmalloc(sizeof(*nm), M_LWKTMSG, M_INTWAIT);
727 netmsg_init(&nm->base, NULL, &netisr_afree_rport,
728 0, udp_notifyall_oncpu);
729 nm->nm_faddr = faddr;
730 nm->nm_arg = inetctlerrmap[cmd];
731 nm->nm_notify = notify;
732 lwkt_sendmsg(netisr_portfn(0), &nm->base.lmsg);
735 * XXX We should forward msg upon PRC_HOSTHEAD and ip == NULL,
736 * once UDP inpcbs are CPU localized
738 KKASSERT(&curthread->td_msgport == netisr_portfn(0));
739 in_pcbnotifyall(&udbinfo.pcblisthead, faddr, inetctlerrmap[cmd],
743 lwkt_replymsg(&msg->lmsg, 0);
747 udp_pcblist(SYSCTL_HANDLER_ARGS)
753 error = in_pcblist_global_nomarker(oidp, arg1, arg2, req, &xi, &nxi);
757 KKASSERT(xi == NULL);
761 KKASSERT(xi == NULL);
765 for (i = 0; i < nxi; ++i) {
766 error = SYSCTL_OUT(req, &xi[i], sizeof(xi[i]));
774 SYSCTL_PROC(_net_inet_udp, UDPCTL_PCBLIST, pcblist, CTLFLAG_RD, &udbinfo, 0,
775 udp_pcblist, "S,xinpcb", "List of active UDP sockets");
778 udp_getcred(SYSCTL_HANDLER_ARGS)
780 struct sockaddr_in addrs[2];
781 struct ucred cred0, *cred = NULL;
785 error = priv_check(req->td, PRIV_ROOT);
788 error = SYSCTL_IN(req, addrs, sizeof addrs);
793 inp = in_pcblookup_hash(&udbinfo, addrs[1].sin_addr, addrs[1].sin_port,
794 addrs[0].sin_addr, addrs[0].sin_port, 1, NULL);
795 if (inp == NULL || inp->inp_socket == NULL) {
798 if (inp->inp_socket->so_cred != NULL) {
799 cred0 = *(inp->inp_socket->so_cred);
808 return SYSCTL_OUT(req, cred, sizeof(struct ucred));
811 SYSCTL_PROC(_net_inet_udp, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW,
812 0, 0, udp_getcred, "S,ucred", "Get the ucred of a UDP connection");
815 udp_output(struct inpcb *inp, struct mbuf *m, struct sockaddr *dstaddr,
816 struct thread *td, int flags)
819 int len = m->m_pkthdr.len;
820 struct sockaddr_in *sin; /* really is initialized before use */
821 int error = 0, lport_any = 0;
823 if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) {
828 if (inp->inp_lport == 0) { /* unbound socket */
829 error = in_pcbbind(inp, NULL, td);
833 udbinfo_barrier_set();
834 in_pcbinswildcardhash(inp);
835 udbinfo_barrier_rem();
839 if (dstaddr != NULL) { /* destination address specified */
840 if (inp->inp_faddr.s_addr != INADDR_ANY) {
841 /* already connected */
845 sin = (struct sockaddr_in *)dstaddr;
846 if (!prison_remote_ip(td, (struct sockaddr *)&sin)) {
847 error = EAFNOSUPPORT; /* IPv6 only jail */
851 if (inp->inp_faddr.s_addr == INADDR_ANY) {
852 /* no destination specified and not already connected */
860 * Calculate data length and get a mbuf
861 * for UDP and IP headers.
863 M_PREPEND(m, sizeof(struct udpiphdr), MB_DONTWAIT);
870 * Fill in mbuf with extended UDP header
871 * and addresses and length put into network format.
873 ui = mtod(m, struct udpiphdr *);
874 bzero(ui->ui_x1, sizeof ui->ui_x1); /* XXX still needed? */
875 ui->ui_pr = IPPROTO_UDP;
878 * Set destination address.
880 if (dstaddr != NULL) { /* use specified destination */
881 ui->ui_dst = sin->sin_addr;
882 ui->ui_dport = sin->sin_port;
883 } else { /* use connected destination */
884 ui->ui_dst = inp->inp_faddr;
885 ui->ui_dport = inp->inp_fport;
889 * Set source address.
891 if (inp->inp_laddr.s_addr == INADDR_ANY) {
892 struct sockaddr_in *if_sin;
894 if (dstaddr == NULL) {
896 * connect() had (or should have) failed because
897 * the interface had no IP address, but the
898 * application proceeded to call send() anyways.
904 /* Look up outgoing interface. */
905 if ((error = in_pcbladdr(inp, dstaddr, &if_sin, td)))
907 ui->ui_src = if_sin->sin_addr; /* use address of interface */
908 } else if (!IN_MULTICAST(ntohl(inp->inp_laddr.s_addr))) {
909 ui->ui_src = inp->inp_laddr; /* use non-null bound address */
911 /* Bound to multicast address; let ip_output choose */
912 ui->ui_src.s_addr = INADDR_ANY;
914 ui->ui_sport = inp->inp_lport;
915 KASSERT(inp->inp_lport != 0, ("inp lport should have been bound"));
917 ui->ui_ulen = htons((u_short)len + sizeof(struct udphdr));
920 * Set up checksum and output datagram.
923 ui->ui_sum = in_pseudo(ui->ui_src.s_addr, ui->ui_dst.s_addr,
924 htons((u_short)len + sizeof(struct udphdr) + IPPROTO_UDP));
925 m->m_pkthdr.csum_flags = CSUM_UDP;
926 m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
927 m->m_pkthdr.csum_thlen = sizeof(struct udphdr);
931 ((struct ip *)ui)->ip_len = sizeof(struct udpiphdr) + len;
932 ((struct ip *)ui)->ip_ttl = inp->inp_ip_ttl; /* XXX */
933 ((struct ip *)ui)->ip_tos = inp->inp_ip_tos; /* XXX */
934 udp_stat.udps_opackets++;
936 error = ip_output(m, inp->inp_options, &inp->inp_route,
937 (inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST)) |
938 flags | IP_DEBUGROUTE,
939 inp->inp_moptions, inp);
942 * If this is the first data gram sent on an unbound and unconnected
943 * UDP socket, lport will be changed in this function. If target
944 * CPU after this lport changing is no longer the current CPU, then
945 * free the route entry allocated on the current CPU.
948 if (udp_addrcpu(inp->inp_faddr.s_addr, inp->inp_fport,
949 inp->inp_laddr.s_addr, inp->inp_lport) != mycpuid) {
951 struct route *ro = &inp->inp_route;
953 if (ro->ro_rt != NULL)
955 bzero(ro, sizeof(*ro));
957 panic("UDP activity should only be in netisr0");
968 u_long udp_sendspace = 9216; /* really max datagram size */
969 /* 40 1K datagrams */
970 SYSCTL_INT(_net_inet_udp, UDPCTL_MAXDGRAM, maxdgram, CTLFLAG_RW,
971 &udp_sendspace, 0, "Maximum outgoing UDP datagram size");
973 u_long udp_recvspace = 40 * (1024 +
975 sizeof(struct sockaddr_in6)
977 sizeof(struct sockaddr_in)
980 SYSCTL_INT(_net_inet_udp, UDPCTL_RECVSPACE, recvspace, CTLFLAG_RW,
981 &udp_recvspace, 0, "Maximum incoming UDP datagram size");
984 * NOTE: (so) is referenced from soabort*() and netmsg_pru_abort()
985 * will sofree() it when we return.
988 udp_abort(netmsg_t msg)
990 struct socket *so = msg->abort.base.nm_so;
994 KKASSERT(&curthread->td_msgport == netisr_portfn(0));
998 soisdisconnected(so);
1000 udbinfo_barrier_set();
1002 udbinfo_barrier_rem();
1007 lwkt_replymsg(&msg->abort.base.lmsg, error);
1011 udp_attach(netmsg_t msg)
1013 struct socket *so = msg->attach.base.nm_so;
1014 struct pru_attach_info *ai = msg->attach.nm_ai;
1018 KKASSERT(&curthread->td_msgport == netisr_portfn(0));
1025 error = soreserve(so, udp_sendspace, udp_recvspace, ai->sb_rlimit);
1029 udbinfo_barrier_set();
1030 error = in_pcballoc(so, &udbinfo);
1031 udbinfo_barrier_rem();
1037 * Set default port for protocol processing prior to bind/connect.
1039 sosetport(so, netisr_portfn(0));
1041 inp = (struct inpcb *)so->so_pcb;
1042 inp->inp_vflag |= INP_IPV4;
1043 inp->inp_ip_ttl = ip_defttl;
1046 lwkt_replymsg(&msg->attach.base.lmsg, error);
1050 udp_bind(netmsg_t msg)
1052 struct socket *so = msg->bind.base.nm_so;
1053 struct sockaddr *nam = msg->bind.nm_nam;
1054 struct thread *td = msg->bind.nm_td;
1055 struct sockaddr_in *sin = (struct sockaddr_in *)nam;
1061 error = in_pcbbind(inp, nam, td);
1063 if (sin->sin_addr.s_addr != INADDR_ANY)
1064 inp->inp_flags |= INP_WASBOUND_NOTANY;
1066 udbinfo_barrier_set();
1067 in_pcbinswildcardhash(inp);
1068 udbinfo_barrier_rem();
1073 lwkt_replymsg(&msg->bind.base.lmsg, error);
1077 udp_connect(netmsg_t msg)
1079 struct socket *so = msg->connect.base.nm_so;
1080 struct sockaddr *nam = msg->connect.nm_nam;
1081 struct thread *td = msg->connect.nm_td;
1083 struct sockaddr_in *sin = (struct sockaddr_in *)nam;
1084 struct sockaddr_in *if_sin;
1088 KKASSERT(&curthread->td_msgport == netisr_portfn(0));
1096 if (msg->connect.nm_reconnect & NMSG_RECONNECT_RECONNECT) {
1097 panic("UDP does not support RECONNECT");
1099 msg->connect.nm_reconnect &= ~NMSG_RECONNECT_RECONNECT;
1100 in_pcblink(inp, &udbinfo);
1104 if (inp->inp_faddr.s_addr != INADDR_ANY) {
1111 * Bind if we have to
1113 if (td->td_proc && td->td_proc->p_ucred->cr_prison != NULL &&
1114 inp->inp_laddr.s_addr == INADDR_ANY) {
1115 error = in_pcbbind(inp, NULL, td);
1121 * Calculate the correct protocol processing thread. The connect
1122 * operation must run there.
1124 error = in_pcbladdr(inp, nam, &if_sin, td);
1127 if (!prison_remote_ip(td, nam)) {
1128 error = EAFNOSUPPORT; /* IPv6 only jail */
1132 port = udp_addrport(sin->sin_addr.s_addr, sin->sin_port,
1133 inp->inp_laddr.s_addr, inp->inp_lport);
1134 if (port != &curthread->td_msgport) {
1136 struct route *ro = &inp->inp_route;
1139 * in_pcbladdr() may have allocated a route entry for us
1140 * on the current CPU, but we need a route entry on the
1141 * inpcb's owner CPU, so free it here.
1143 if (ro->ro_rt != NULL)
1145 bzero(ro, sizeof(*ro));
1148 * We are moving the protocol processing port the socket
1149 * is on, we have to unlink here and re-link on the
1152 in_pcbunlink(so->so_pcb, &udbinfo);
1153 /* in_pcbunlink(so->so_pcb, &udbinfo[mycpu->gd_cpuid]); */
1154 sosetport(so, port);
1155 msg->connect.nm_reconnect |= NMSG_RECONNECT_RECONNECT;
1156 msg->connect.base.nm_dispatch = udp_connect;
1158 lwkt_forwardmsg(port, &msg->connect.base.lmsg);
1159 /* msg invalid now */
1162 panic("UDP activity should only be in netisr0");
1165 KKASSERT(port == &curthread->td_msgport);
1166 error = udp_connect_oncpu(so, td, sin, if_sin);
1168 KKASSERT(msg->connect.nm_m == NULL);
1169 lwkt_replymsg(&msg->connect.base.lmsg, error);
1173 udp_connect_oncpu(struct socket *so, struct thread *td,
1174 struct sockaddr_in *sin, struct sockaddr_in *if_sin)
1179 udbinfo_barrier_set();
1182 if (inp->inp_flags & INP_WILDCARD)
1183 in_pcbremwildcardhash(inp);
1184 error = in_pcbconnect(inp, (struct sockaddr *)sin, td);
1188 * No more errors can occur, finish adjusting the socket
1189 * and change the processing port to reflect the connected
1190 * socket. Once set we can no longer safely mess with the
1194 } else if (error == EAFNOSUPPORT) { /* connection dissolved */
1196 * Follow traditional BSD behavior and retain
1197 * the local port binding. But, fix the old misbehavior
1198 * of overwriting any previously bound local address.
1200 if (!(inp->inp_flags & INP_WASBOUND_NOTANY))
1201 inp->inp_laddr.s_addr = INADDR_ANY;
1202 in_pcbinswildcardhash(inp);
1205 udbinfo_barrier_rem();
1210 udp_detach(netmsg_t msg)
1212 struct socket *so = msg->detach.base.nm_so;
1216 KKASSERT(&curthread->td_msgport == netisr_portfn(0));
1220 udbinfo_barrier_set();
1222 udbinfo_barrier_rem();
1227 lwkt_replymsg(&msg->detach.base.lmsg, error);
1231 udp_disconnect(netmsg_t msg)
1233 struct socket *so = msg->disconnect.base.nm_so;
1238 KKASSERT(&curthread->td_msgport == netisr_portfn(0));
1245 if (inp->inp_faddr.s_addr == INADDR_ANY) {
1252 udbinfo_barrier_set();
1253 in_pcbdisconnect(inp);
1254 udbinfo_barrier_rem();
1256 soclrstate(so, SS_ISCONNECTED); /* XXX */
1259 ro = &inp->inp_route;
1260 if (ro->ro_rt != NULL)
1262 bzero(ro, sizeof(*ro));
1265 lwkt_replymsg(&msg->disconnect.base.lmsg, error);
1269 udp_send(netmsg_t msg)
1271 struct socket *so = msg->send.base.nm_so;
1272 struct mbuf *m = msg->send.nm_m;
1273 struct sockaddr *addr = msg->send.nm_addr;
1274 int pru_flags = msg->send.nm_flags;
1278 KKASSERT(&curthread->td_msgport == netisr_portfn(0));
1279 KKASSERT(msg->send.nm_control == NULL);
1283 struct thread *td = msg->send.nm_td;
1286 if (pru_flags & PRUS_DONTROUTE)
1287 flags |= SO_DONTROUTE;
1288 error = udp_output(inp, m, addr, td, flags);
1294 if (pru_flags & PRUS_FREEADDR)
1295 kfree(addr, M_SONAME);
1297 if ((pru_flags & PRUS_NOREPLY) == 0)
1298 lwkt_replymsg(&msg->send.base.lmsg, error);
1302 udp_shutdown(netmsg_t msg)
1304 struct socket *so = msg->shutdown.base.nm_so;
1308 KKASSERT(&curthread->td_msgport == netisr_portfn(0));
1317 lwkt_replymsg(&msg->shutdown.base.lmsg, error);
1323 lwkt_serialize_enter(&udbinfo_slize);
1327 udbinfo_unlock(void)
1329 lwkt_serialize_exit(&udbinfo_slize);
1333 udbinfo_barrier_set(void)
1335 netisr_barrier_set(udbinfo_br);
1340 udbinfo_barrier_rem(void)
1343 netisr_barrier_rem(udbinfo_br);
1346 struct pr_usrreqs udp_usrreqs = {
1347 .pru_abort = udp_abort,
1348 .pru_accept = pr_generic_notsupp,
1349 .pru_attach = udp_attach,
1350 .pru_bind = udp_bind,
1351 .pru_connect = udp_connect,
1352 .pru_connect2 = pr_generic_notsupp,
1353 .pru_control = in_control_dispatch,
1354 .pru_detach = udp_detach,
1355 .pru_disconnect = udp_disconnect,
1356 .pru_listen = pr_generic_notsupp,
1357 .pru_peeraddr = in_setpeeraddr_dispatch,
1358 .pru_rcvd = pr_generic_notsupp,
1359 .pru_rcvoob = pr_generic_notsupp,
1360 .pru_send = udp_send,
1361 .pru_sense = pru_sense_null,
1362 .pru_shutdown = udp_shutdown,
1363 .pru_sockaddr = in_setsockaddr_dispatch,
1364 .pru_sosend = sosendudp,
1365 .pru_soreceive = soreceive