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 $
68 * $DragonFly: src/sys/netinet/udp_usrreq.c,v 1.47 2008/11/11 10:46:58 sephe Exp $
71 #include "opt_ipsec.h"
72 #include "opt_inet6.h"
74 #include <sys/param.h>
75 #include <sys/systm.h>
76 #include <sys/kernel.h>
77 #include <sys/malloc.h>
79 #include <sys/domain.h>
82 #include <sys/protosw.h>
83 #include <sys/socket.h>
84 #include <sys/socketvar.h>
85 #include <sys/sysctl.h>
86 #include <sys/syslog.h>
87 #include <sys/in_cksum.h>
89 #include <sys/thread2.h>
90 #include <sys/socketvar2.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 struct inpcbinfo udbinfo;
149 static struct netisr_barrier *udbinfo_br;
152 #define UDBHASHSIZE 16
155 struct udpstat udpstat; /* from udp_var.h */
156 SYSCTL_STRUCT(_net_inet_udp, UDPCTL_STATS, stats, CTLFLAG_RW,
157 &udpstat, udpstat, "UDP statistics (struct udpstat, netinet/udp_var.h)");
159 static struct sockaddr_in udp_in = { sizeof udp_in, AF_INET };
162 struct sockaddr_in6 uin6_sin;
163 u_char uin6_init_done : 1;
165 { sizeof udp_in6.uin6_sin, AF_INET6 },
169 struct ip6_hdr uip6_ip6;
170 u_char uip6_init_done : 1;
174 static void udp_append (struct inpcb *last, struct ip *ip,
175 struct mbuf *n, int off);
177 static void ip_2_ip6_hdr (struct ip6_hdr *ip6, struct ip *ip);
180 static int udp_connect_oncpu(struct socket *so, struct thread *td,
181 struct sockaddr_in *sin, struct sockaddr_in *if_sin);
182 static int udp_output (struct inpcb *, struct mbuf *, struct sockaddr *,
183 struct mbuf *, struct thread *);
184 static void udbinfo_barrier_set(void);
185 static void udbinfo_barrier_rem(void);
190 in_pcbinfo_init(&udbinfo);
191 udbinfo.hashbase = hashinit(UDBHASHSIZE, M_PCB, &udbinfo.hashmask);
192 udbinfo.porthashbase = hashinit(UDBHASHSIZE, M_PCB,
193 &udbinfo.porthashmask);
194 udbinfo.wildcardhashbase = hashinit(UDBHASHSIZE, M_PCB,
195 &udbinfo.wildcardhashmask);
196 udbinfo.ipi_size = sizeof(struct inpcb);
198 udbinfo_br = netisr_barrier_create();
202 * Check multicast packets to make sure they are only sent to sockets with
203 * multicast memberships for the packet's destination address and arrival
204 * interface. Multicast packets to multicast-unaware sockets are also
207 * Returns 0 if the packet is acceptable, -1 if it is not.
210 check_multicast_membership(struct ip *ip, struct inpcb *inp, struct mbuf *m)
213 struct ip_moptions *mopt;
215 if (strict_mcast_mship == 0 ||
216 !IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
219 mopt = inp->inp_moptions;
222 for (mshipno = 0; mshipno <= mopt->imo_num_memberships; ++mshipno) {
223 struct in_multi *maddr = mopt->imo_membership[mshipno];
225 if (ip->ip_dst.s_addr == maddr->inm_addr.s_addr &&
226 m->m_pkthdr.rcvif == maddr->inm_ifp) {
234 udp_input(struct mbuf **mp, int *offp, int proto)
241 struct mbuf *opts = NULL;
244 struct sockaddr *append_sa;
251 udpstat.udps_ipackets++;
254 * Strip IP options, if any; should skip this,
255 * make available to user, and use on returned packets,
256 * but we don't yet have a way to check the checksum
257 * with options still present.
259 if (iphlen > sizeof(struct ip)) {
261 iphlen = sizeof(struct ip);
265 * IP and UDP headers are together in first mbuf.
266 * Already checked and pulled up in ip_demux().
268 KASSERT(m->m_len >= iphlen + sizeof(struct udphdr),
269 ("UDP header not in one mbuf"));
271 ip = mtod(m, struct ip *);
272 uh = (struct udphdr *)((caddr_t)ip + iphlen);
274 /* destination port of 0 is illegal, based on RFC768. */
275 if (uh->uh_dport == 0)
279 * Make mbuf data length reflect UDP length.
280 * If not enough data to reflect UDP length, drop.
282 len = ntohs((u_short)uh->uh_ulen);
283 if (ip->ip_len != len) {
284 if (len > ip->ip_len || len < sizeof(struct udphdr)) {
285 udpstat.udps_badlen++;
288 m_adj(m, len - ip->ip_len);
289 /* ip->ip_len = len; */
292 * Save a copy of the IP header in case we want restore it
293 * for sending an ICMP error message in response.
298 * Checksum extended UDP header and data.
301 if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) {
302 if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR)
303 uh->uh_sum = m->m_pkthdr.csum_data;
305 uh->uh_sum = in_pseudo(ip->ip_src.s_addr,
306 ip->ip_dst.s_addr, htonl((u_short)len +
307 m->m_pkthdr.csum_data + IPPROTO_UDP));
308 uh->uh_sum ^= 0xffff;
312 bcopy(((struct ipovly *)ip)->ih_x1, b, 9);
313 bzero(((struct ipovly *)ip)->ih_x1, 9);
314 ((struct ipovly *)ip)->ih_len = uh->uh_ulen;
315 uh->uh_sum = in_cksum(m, len + sizeof(struct ip));
316 bcopy(b, ((struct ipovly *)ip)->ih_x1, 9);
319 udpstat.udps_badsum++;
321 return(IPPROTO_DONE);
324 udpstat.udps_nosum++;
326 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) ||
327 in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif)) {
331 * Deliver a multicast or broadcast datagram to *all* sockets
332 * for which the local and remote addresses and ports match
333 * those of the incoming datagram. This allows more than
334 * one process to receive multi/broadcasts on the same port.
335 * (This really ought to be done for unicast datagrams as
336 * well, but that would cause problems with existing
337 * applications that open both address-specific sockets and
338 * a wildcard socket listening to the same port -- they would
339 * end up receiving duplicates of every unicast datagram.
340 * Those applications open the multiple sockets to overcome an
341 * inadequacy of the UDP socket interface, but for backwards
342 * compatibility we avoid the problem here rather than
343 * fixing the interface. Maybe 4.5BSD will remedy this?)
347 * Construct sockaddr format source address.
349 udp_in.sin_port = uh->uh_sport;
350 udp_in.sin_addr = ip->ip_src;
352 * Locate pcb(s) for datagram.
353 * (Algorithm copied from raw_intr().)
357 udp_in6.uin6_init_done = udp_ip6.uip6_init_done = 0;
359 LIST_FOREACH(inp, &udbinfo.pcblisthead, inp_list) {
360 if (inp->inp_flags & INP_PLACEMARKER)
363 if (!(inp->inp_vflag & INP_IPV4))
366 if (inp->inp_lport != uh->uh_dport)
368 if (inp->inp_laddr.s_addr != INADDR_ANY) {
369 if (inp->inp_laddr.s_addr !=
373 if (inp->inp_faddr.s_addr != INADDR_ANY) {
374 if (inp->inp_faddr.s_addr !=
376 inp->inp_fport != uh->uh_sport)
380 if (check_multicast_membership(ip, inp, m) < 0)
387 /* check AH/ESP integrity. */
388 if (ipsec4_in_reject_so(m, last->inp_socket))
389 ipsecstat.in_polvio++;
390 /* do not inject data to pcb */
394 /* check AH/ESP integrity. */
395 if (ipsec4_in_reject(m, last))
398 #endif /*FAST_IPSEC*/
399 if ((n = m_copypacket(m, MB_DONTWAIT)) != NULL)
400 udp_append(last, ip, n,
402 sizeof(struct udphdr));
406 * Don't look for additional matches if this one does
407 * not have either the SO_REUSEPORT or SO_REUSEADDR
408 * socket options set. This heuristic avoids searching
409 * through all pcbs in the common case of a non-shared
410 * port. It * assumes that an application will never
411 * clear these options after setting them.
413 if (!(last->inp_socket->so_options &
414 (SO_REUSEPORT | SO_REUSEADDR)))
420 * No matching pcb found; discard datagram.
421 * (No need to send an ICMP Port Unreachable
422 * for a broadcast or multicast datgram.)
424 udpstat.udps_noportbcast++;
428 /* check AH/ESP integrity. */
429 if (ipsec4_in_reject_so(m, last->inp_socket)) {
430 ipsecstat.in_polvio++;
435 /* check AH/ESP integrity. */
436 if (ipsec4_in_reject(m, last))
438 #endif /*FAST_IPSEC*/
439 udp_append(last, ip, m, iphlen + sizeof(struct udphdr));
440 return(IPPROTO_DONE);
443 * Locate pcb for datagram.
445 inp = in_pcblookup_hash(&udbinfo, ip->ip_src, uh->uh_sport,
446 ip->ip_dst, uh->uh_dport, 1, m->m_pkthdr.rcvif);
449 char buf[sizeof "aaa.bbb.ccc.ddd"];
451 strcpy(buf, inet_ntoa(ip->ip_dst));
453 "Connection attempt to UDP %s:%d from %s:%d\n",
454 buf, ntohs(uh->uh_dport), inet_ntoa(ip->ip_src),
455 ntohs(uh->uh_sport));
457 udpstat.udps_noport++;
458 if (m->m_flags & (M_BCAST | M_MCAST)) {
459 udpstat.udps_noportbcast++;
465 if (badport_bandlim(BANDLIM_ICMP_UNREACH) < 0)
469 ip->ip_len += iphlen;
470 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0);
471 return(IPPROTO_DONE);
474 if (ipsec4_in_reject_so(m, inp->inp_socket)) {
475 ipsecstat.in_polvio++;
480 if (ipsec4_in_reject(m, inp))
482 #endif /*FAST_IPSEC*/
484 * Check the minimum TTL for socket.
486 if (ip->ip_ttl < inp->inp_ip_minttl)
490 * Construct sockaddr format source address.
491 * Stuff source address and datagram in user buffer.
493 udp_in.sin_port = uh->uh_sport;
494 udp_in.sin_addr = ip->ip_src;
495 if ((inp->inp_flags & INP_CONTROLOPTS) ||
496 (inp->inp_socket->so_options & SO_TIMESTAMP)) {
498 if (inp->inp_vflag & INP_IPV6) {
501 ip_2_ip6_hdr(&udp_ip6.uip6_ip6, ip);
502 savedflags = inp->inp_flags;
503 inp->inp_flags &= ~INP_UNMAPPABLEOPTS;
504 ip6_savecontrol(inp, &opts, &udp_ip6.uip6_ip6, m);
505 inp->inp_flags = savedflags;
508 ip_savecontrol(inp, &opts, ip, m);
510 m_adj(m, iphlen + sizeof(struct udphdr));
512 if (inp->inp_vflag & INP_IPV6) {
513 in6_sin_2_v4mapsin6(&udp_in, &udp_in6.uin6_sin);
514 append_sa = (struct sockaddr *)&udp_in6;
517 append_sa = (struct sockaddr *)&udp_in;
519 lwkt_gettoken(&inp->inp_socket->so_rcv.ssb_token);
520 if (ssb_appendaddr(&inp->inp_socket->so_rcv, append_sa, m, opts) == 0) {
521 udpstat.udps_fullsock++;
522 lwkt_reltoken(&inp->inp_socket->so_rcv.ssb_token);
525 lwkt_reltoken(&inp->inp_socket->so_rcv.ssb_token);
526 sorwakeup(inp->inp_socket);
527 return(IPPROTO_DONE);
532 return(IPPROTO_DONE);
537 ip_2_ip6_hdr(struct ip6_hdr *ip6, struct ip *ip)
539 bzero(ip6, sizeof *ip6);
541 ip6->ip6_vfc = IPV6_VERSION;
542 ip6->ip6_plen = ip->ip_len;
543 ip6->ip6_nxt = ip->ip_p;
544 ip6->ip6_hlim = ip->ip_ttl;
545 ip6->ip6_src.s6_addr32[2] = ip6->ip6_dst.s6_addr32[2] =
547 ip6->ip6_src.s6_addr32[3] = ip->ip_src.s_addr;
548 ip6->ip6_dst.s6_addr32[3] = ip->ip_dst.s_addr;
553 * subroutine of udp_input(), mainly for source code readability.
554 * caller must properly init udp_ip6 and udp_in6 beforehand.
557 udp_append(struct inpcb *last, struct ip *ip, struct mbuf *n, int off)
559 struct sockaddr *append_sa;
560 struct mbuf *opts = NULL;
562 if (last->inp_flags & INP_CONTROLOPTS ||
563 last->inp_socket->so_options & SO_TIMESTAMP) {
565 if (last->inp_vflag & INP_IPV6) {
568 if (udp_ip6.uip6_init_done == 0) {
569 ip_2_ip6_hdr(&udp_ip6.uip6_ip6, ip);
570 udp_ip6.uip6_init_done = 1;
572 savedflags = last->inp_flags;
573 last->inp_flags &= ~INP_UNMAPPABLEOPTS;
574 ip6_savecontrol(last, &opts, &udp_ip6.uip6_ip6, n);
575 last->inp_flags = savedflags;
578 ip_savecontrol(last, &opts, ip, n);
581 if (last->inp_vflag & INP_IPV6) {
582 if (udp_in6.uin6_init_done == 0) {
583 in6_sin_2_v4mapsin6(&udp_in, &udp_in6.uin6_sin);
584 udp_in6.uin6_init_done = 1;
586 append_sa = (struct sockaddr *)&udp_in6.uin6_sin;
589 append_sa = (struct sockaddr *)&udp_in;
591 lwkt_gettoken(&last->inp_socket->so_rcv.ssb_token);
592 if (ssb_appendaddr(&last->inp_socket->so_rcv, append_sa, n, opts) == 0) {
596 udpstat.udps_fullsock++;
598 sorwakeup(last->inp_socket);
600 lwkt_reltoken(&last->inp_socket->so_rcv.ssb_token);
604 * Notify a udp user of an asynchronous error;
605 * just wake up so that he can collect error status.
608 udp_notify(struct inpcb *inp, int error)
610 inp->inp_socket->so_error = error;
611 sorwakeup(inp->inp_socket);
612 sowwakeup(inp->inp_socket);
615 struct netmsg_udp_notify {
616 struct netmsg_base base;
617 void (*nm_notify)(struct inpcb *, int);
618 struct in_addr nm_faddr;
623 udp_notifyall_oncpu(netmsg_t msg)
625 struct netmsg_udp_notify *nm = (struct netmsg_udp_notify *)msg;
630 in_pcbnotifyall(&udbinfo.pcblisthead, nm->nm_faddr,
631 nm->nm_arg, nm->nm_notify);
632 lwkt_replymsg(&nm->base.lmsg, 0);
635 /* XXX currently udp only runs on cpu 0 */
636 nextcpu = mycpuid + 1;
637 if (nextcpu < ncpus2)
638 lwkt_forwardmsg(cpu_portfn(nextcpu), &nm->base.lmsg);
640 lwkt_replymsg(&nmsg->base.lmsg, 0);
645 udp_rtchange(struct inpcb *inp, int err)
648 /* XXX Nuke this, once UDP inpcbs are CPU localized */
649 if (inp->inp_route.ro_rt && inp->inp_route.ro_rt->rt_cpuid == mycpuid) {
650 rtfree(inp->inp_route.ro_rt);
651 inp->inp_route.ro_rt = NULL;
653 * A new route can be allocated the next time
654 * output is attempted.
658 in_rtchange(inp, err);
663 udp_ctlinput(netmsg_t msg)
665 struct sockaddr *sa = msg->ctlinput.nm_arg;
666 struct ip *ip = msg->ctlinput.nm_extra;
667 int cmd = msg->ctlinput.nm_cmd;
669 void (*notify) (struct inpcb *, int) = udp_notify;
670 struct in_addr faddr;
673 KKASSERT(&curthread->td_msgport == cpu_portfn(0));
675 faddr = ((struct sockaddr_in *)sa)->sin_addr;
676 if (sa->sa_family != AF_INET || faddr.s_addr == INADDR_ANY)
679 if (PRC_IS_REDIRECT(cmd)) {
681 notify = udp_rtchange;
682 } else if (cmd == PRC_HOSTDEAD) {
684 } else if ((unsigned)cmd >= PRC_NCMDS || inetctlerrmap[cmd] == 0) {
689 uh = (struct udphdr *)((caddr_t)ip + (ip->ip_hl << 2));
690 inp = in_pcblookup_hash(&udbinfo, faddr, uh->uh_dport,
691 ip->ip_src, uh->uh_sport, 0, NULL);
692 if (inp != NULL && inp->inp_socket != NULL)
693 (*notify)(inp, inetctlerrmap[cmd]);
694 } else if (PRC_IS_REDIRECT(cmd)) {
695 struct netmsg_udp_notify *nm;
697 KKASSERT(&curthread->td_msgport == cpu_portfn(0));
698 nm = kmalloc(sizeof(*nm), M_LWKTMSG, M_INTWAIT);
699 netmsg_init(&nm->base, NULL, &netisr_afree_rport,
700 0, udp_notifyall_oncpu);
701 nm->nm_faddr = faddr;
702 nm->nm_arg = inetctlerrmap[cmd];
703 nm->nm_notify = notify;
704 lwkt_sendmsg(cpu_portfn(0), &nm->base.lmsg);
707 * XXX We should forward msg upon PRC_HOSTHEAD and ip == NULL,
708 * once UDP inpcbs are CPU localized
710 KKASSERT(&curthread->td_msgport == cpu_portfn(0));
711 in_pcbnotifyall(&udbinfo.pcblisthead, faddr, inetctlerrmap[cmd],
715 lwkt_replymsg(&msg->lmsg, 0);
718 SYSCTL_PROC(_net_inet_udp, UDPCTL_PCBLIST, pcblist, CTLFLAG_RD, &udbinfo, 0,
719 in_pcblist_global, "S,xinpcb", "List of active UDP sockets");
722 udp_getcred(SYSCTL_HANDLER_ARGS)
724 struct sockaddr_in addrs[2];
728 error = priv_check(req->td, PRIV_ROOT);
731 error = SYSCTL_IN(req, addrs, sizeof addrs);
734 inp = in_pcblookup_hash(&udbinfo, addrs[1].sin_addr, addrs[1].sin_port,
735 addrs[0].sin_addr, addrs[0].sin_port, 1, NULL);
736 if (inp == NULL || inp->inp_socket == NULL) {
740 error = SYSCTL_OUT(req, inp->inp_socket->so_cred, sizeof(struct ucred));
745 SYSCTL_PROC(_net_inet_udp, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW,
746 0, 0, udp_getcred, "S,ucred", "Get the ucred of a UDP connection");
749 udp_output(struct inpcb *inp, struct mbuf *m, struct sockaddr *dstaddr,
750 struct mbuf *control, struct thread *td)
753 int len = m->m_pkthdr.len;
754 struct sockaddr_in *sin; /* really is initialized before use */
755 int error = 0, lport_any = 0;
757 if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) {
762 if (inp->inp_lport == 0) { /* unbound socket */
763 error = in_pcbbind(inp, NULL, td);
767 udbinfo_barrier_set();
768 in_pcbinswildcardhash(inp);
769 udbinfo_barrier_rem();
773 if (dstaddr != NULL) { /* destination address specified */
774 if (inp->inp_faddr.s_addr != INADDR_ANY) {
775 /* already connected */
779 sin = (struct sockaddr_in *)dstaddr;
780 if (!prison_remote_ip(td, (struct sockaddr *)&sin)) {
781 error = EAFNOSUPPORT; /* IPv6 only jail */
785 if (inp->inp_faddr.s_addr == INADDR_ANY) {
786 /* no destination specified and not already connected */
794 * Calculate data length and get a mbuf
795 * for UDP and IP headers.
797 M_PREPEND(m, sizeof(struct udpiphdr), MB_DONTWAIT);
804 * Fill in mbuf with extended UDP header
805 * and addresses and length put into network format.
807 ui = mtod(m, struct udpiphdr *);
808 bzero(ui->ui_x1, sizeof ui->ui_x1); /* XXX still needed? */
809 ui->ui_pr = IPPROTO_UDP;
812 * Set destination address.
814 if (dstaddr != NULL) { /* use specified destination */
815 ui->ui_dst = sin->sin_addr;
816 ui->ui_dport = sin->sin_port;
817 } else { /* use connected destination */
818 ui->ui_dst = inp->inp_faddr;
819 ui->ui_dport = inp->inp_fport;
823 * Set source address.
825 if (inp->inp_laddr.s_addr == INADDR_ANY) {
826 struct sockaddr_in *if_sin;
828 if (dstaddr == NULL) {
830 * connect() had (or should have) failed because
831 * the interface had no IP address, but the
832 * application proceeded to call send() anyways.
838 /* Look up outgoing interface. */
839 if ((error = in_pcbladdr(inp, dstaddr, &if_sin, td)))
841 ui->ui_src = if_sin->sin_addr; /* use address of interface */
843 ui->ui_src = inp->inp_laddr; /* use non-null bound address */
845 ui->ui_sport = inp->inp_lport;
846 KASSERT(inp->inp_lport != 0, ("inp lport should have been bound"));
848 ui->ui_ulen = htons((u_short)len + sizeof(struct udphdr));
851 * Set up checksum and output datagram.
854 ui->ui_sum = in_pseudo(ui->ui_src.s_addr, ui->ui_dst.s_addr,
855 htons((u_short)len + sizeof(struct udphdr) + IPPROTO_UDP));
856 m->m_pkthdr.csum_flags = CSUM_UDP;
857 m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
861 ((struct ip *)ui)->ip_len = sizeof(struct udpiphdr) + len;
862 ((struct ip *)ui)->ip_ttl = inp->inp_ip_ttl; /* XXX */
863 ((struct ip *)ui)->ip_tos = inp->inp_ip_tos; /* XXX */
864 udpstat.udps_opackets++;
866 error = ip_output(m, inp->inp_options, &inp->inp_route,
867 (inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST)) |
869 inp->inp_moptions, inp);
872 * If this is the first data gram sent on an unbound and unconnected
873 * UDP socket, lport will be changed in this function. If target
874 * CPU after this lport changing is no longer the current CPU, then
875 * free the route entry allocated on the current CPU.
878 if (udp_addrcpu(inp->inp_faddr.s_addr, inp->inp_fport,
879 inp->inp_laddr.s_addr, inp->inp_lport) != mycpuid) {
881 struct route *ro = &inp->inp_route;
883 if (ro->ro_rt != NULL)
885 bzero(ro, sizeof(*ro));
887 panic("UDP activity should only be in netisr0");
898 u_long udp_sendspace = 9216; /* really max datagram size */
899 /* 40 1K datagrams */
900 SYSCTL_INT(_net_inet_udp, UDPCTL_MAXDGRAM, maxdgram, CTLFLAG_RW,
901 &udp_sendspace, 0, "Maximum outgoing UDP datagram size");
903 u_long udp_recvspace = 40 * (1024 +
905 sizeof(struct sockaddr_in6)
907 sizeof(struct sockaddr_in)
910 SYSCTL_INT(_net_inet_udp, UDPCTL_RECVSPACE, recvspace, CTLFLAG_RW,
911 &udp_recvspace, 0, "Maximum incoming UDP datagram size");
914 * NOTE: (so) is referenced from soabort*() and netmsg_pru_abort()
915 * will sofree() it when we return.
918 udp_abort(netmsg_t msg)
920 struct socket *so = msg->abort.base.nm_so;
924 KKASSERT(&curthread->td_msgport == cpu_portfn(0));
928 soisdisconnected(so);
930 udbinfo_barrier_set();
932 udbinfo_barrier_rem();
937 lwkt_replymsg(&msg->abort.base.lmsg, error);
941 udp_attach(netmsg_t msg)
943 struct socket *so = msg->attach.base.nm_so;
944 struct pru_attach_info *ai = msg->attach.nm_ai;
948 KKASSERT(&curthread->td_msgport == cpu_portfn(0));
955 error = soreserve(so, udp_sendspace, udp_recvspace, ai->sb_rlimit);
959 udbinfo_barrier_set();
960 error = in_pcballoc(so, &udbinfo);
961 udbinfo_barrier_rem();
967 * Set default port for protocol processing prior to bind/connect.
969 sosetport(so, cpu_portfn(0));
971 inp = (struct inpcb *)so->so_pcb;
972 inp->inp_vflag |= INP_IPV4;
973 inp->inp_ip_ttl = ip_defttl;
976 lwkt_replymsg(&msg->attach.base.lmsg, error);
980 udp_bind(netmsg_t msg)
982 struct socket *so = msg->bind.base.nm_so;
983 struct sockaddr *nam = msg->bind.nm_nam;
984 struct thread *td = msg->bind.nm_td;
985 struct sockaddr_in *sin = (struct sockaddr_in *)nam;
991 error = in_pcbbind(inp, nam, td);
993 if (sin->sin_addr.s_addr != INADDR_ANY)
994 inp->inp_flags |= INP_WASBOUND_NOTANY;
996 udbinfo_barrier_set();
997 in_pcbinswildcardhash(inp);
998 udbinfo_barrier_rem();
1003 lwkt_replymsg(&msg->bind.base.lmsg, error);
1007 udp_connect(netmsg_t msg)
1009 struct socket *so = msg->connect.base.nm_so;
1010 struct sockaddr *nam = msg->connect.nm_nam;
1011 struct thread *td = msg->connect.nm_td;
1013 struct sockaddr_in *sin = (struct sockaddr_in *)nam;
1014 struct sockaddr_in *if_sin;
1018 KKASSERT(&curthread->td_msgport == cpu_portfn(0));
1026 if (msg->connect.nm_reconnect & NMSG_RECONNECT_RECONNECT) {
1027 panic("UDP does not support RECONNECT\n");
1029 msg->connect.nm_reconnect &= ~NMSG_RECONNECT_RECONNECT;
1030 in_pcblink(inp, &udbinfo);
1034 if (inp->inp_faddr.s_addr != INADDR_ANY) {
1041 * Bind if we have to
1043 if (td->td_proc && td->td_proc->p_ucred->cr_prison != NULL &&
1044 inp->inp_laddr.s_addr == INADDR_ANY) {
1045 error = in_pcbbind(inp, NULL, td);
1051 * Calculate the correct protocol processing thread. The connect
1052 * operation must run there.
1054 error = in_pcbladdr(inp, nam, &if_sin, td);
1057 if (!prison_remote_ip(td, nam)) {
1058 error = EAFNOSUPPORT; /* IPv6 only jail */
1062 port = udp_addrport(sin->sin_addr.s_addr, sin->sin_port,
1063 inp->inp_laddr.s_addr, inp->inp_lport);
1065 if (port != &curthread->td_msgport) {
1067 struct route *ro = &inp->inp_route;
1070 * in_pcbladdr() may have allocated a route entry for us
1071 * on the current CPU, but we need a route entry on the
1072 * inpcb's owner CPU, so free it here.
1074 if (ro->ro_rt != NULL)
1076 bzero(ro, sizeof(*ro));
1079 * We are moving the protocol processing port the socket
1080 * is on, we have to unlink here and re-link on the
1083 in_pcbunlink(so->so_pcb, &udbinfo);
1084 /* in_pcbunlink(so->so_pcb, &udbinfo[mycpu->gd_cpuid]); */
1085 sosetport(so, port);
1086 msg->connect.nm_reconnect |= NMSG_RECONNECT_RECONNECT;
1087 msg->connect.base.nm_dispatch = udp_connect;
1089 lwkt_forwardmsg(port, &msg->connect.base.lmsg);
1090 /* msg invalid now */
1093 panic("UDP activity should only be in netisr0");
1097 KKASSERT(port == &curthread->td_msgport);
1098 error = udp_connect_oncpu(so, td, sin, if_sin);
1100 KKASSERT(msg->connect.nm_m == NULL);
1101 lwkt_replymsg(&msg->connect.base.lmsg, error);
1105 udp_connect_oncpu(struct socket *so, struct thread *td,
1106 struct sockaddr_in *sin, struct sockaddr_in *if_sin)
1111 udbinfo_barrier_set();
1114 if (inp->inp_flags & INP_WILDCARD)
1115 in_pcbremwildcardhash(inp);
1116 error = in_pcbconnect(inp, (struct sockaddr *)sin, td);
1120 * No more errors can occur, finish adjusting the socket
1121 * and change the processing port to reflect the connected
1122 * socket. Once set we can no longer safely mess with the
1126 } else if (error == EAFNOSUPPORT) { /* connection dissolved */
1128 * Follow traditional BSD behavior and retain
1129 * the local port binding. But, fix the old misbehavior
1130 * of overwriting any previously bound local address.
1132 if (!(inp->inp_flags & INP_WASBOUND_NOTANY))
1133 inp->inp_laddr.s_addr = INADDR_ANY;
1134 in_pcbinswildcardhash(inp);
1137 udbinfo_barrier_rem();
1142 udp_detach(netmsg_t msg)
1144 struct socket *so = msg->detach.base.nm_so;
1148 KKASSERT(&curthread->td_msgport == cpu_portfn(0));
1152 udbinfo_barrier_set();
1154 udbinfo_barrier_rem();
1159 lwkt_replymsg(&msg->detach.base.lmsg, error);
1163 udp_disconnect(netmsg_t msg)
1165 struct socket *so = msg->disconnect.base.nm_so;
1170 KKASSERT(&curthread->td_msgport == cpu_portfn(0));
1177 if (inp->inp_faddr.s_addr == INADDR_ANY) {
1184 udbinfo_barrier_set();
1185 in_pcbdisconnect(inp);
1186 udbinfo_barrier_rem();
1188 soclrstate(so, SS_ISCONNECTED); /* XXX */
1191 ro = &inp->inp_route;
1192 if (ro->ro_rt != NULL)
1194 bzero(ro, sizeof(*ro));
1197 lwkt_replymsg(&msg->disconnect.base.lmsg, error);
1201 udp_send(netmsg_t msg)
1203 struct socket *so = msg->send.base.nm_so;
1207 KKASSERT(&curthread->td_msgport == cpu_portfn(0));
1211 error = udp_output(inp,
1214 msg->send.nm_control,
1217 m_freem(msg->send.nm_m);
1220 msg->send.nm_m = NULL;
1221 lwkt_replymsg(&msg->send.base.lmsg, error);
1225 udp_shutdown(netmsg_t msg)
1227 struct socket *so = msg->shutdown.base.nm_so;
1231 KKASSERT(&curthread->td_msgport == cpu_portfn(0));
1240 lwkt_replymsg(&msg->shutdown.base.lmsg, error);
1244 udbinfo_barrier_set(void)
1246 netisr_barrier_set(udbinfo_br);
1250 udbinfo_barrier_rem(void)
1252 netisr_barrier_rem(udbinfo_br);
1255 struct pr_usrreqs udp_usrreqs = {
1256 .pru_abort = udp_abort,
1257 .pru_accept = pr_generic_notsupp,
1258 .pru_attach = udp_attach,
1259 .pru_bind = udp_bind,
1260 .pru_connect = udp_connect,
1261 .pru_connect2 = pr_generic_notsupp,
1262 .pru_control = in_control_dispatch,
1263 .pru_detach = udp_detach,
1264 .pru_disconnect = udp_disconnect,
1265 .pru_listen = pr_generic_notsupp,
1266 .pru_peeraddr = in_setpeeraddr_dispatch,
1267 .pru_rcvd = pr_generic_notsupp,
1268 .pru_rcvoob = pr_generic_notsupp,
1269 .pru_send = udp_send,
1270 .pru_sense = pru_sense_null,
1271 .pru_shutdown = udp_shutdown,
1272 .pru_sockaddr = in_setsockaddr_dispatch,
1273 .pru_sosend = sosendudp,
1274 .pru_soreceive = soreceive