2 * Copyright (c) 2004 Jeffrey M. Hsu. All rights reserved.
3 * Copyright (c) 2004 The DragonFly Project. All rights reserved.
5 * This code is derived from software contributed to The DragonFly Project
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. Neither the name of The DragonFly Project nor the names of its
17 * contributors may be used to endorse or promote products derived
18 * from this software without specific, prior written permission.
20 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
21 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
22 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
23 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
24 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
25 * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
26 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
27 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
28 * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
29 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
30 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
35 * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995
36 * The Regents of the University of California. All rights reserved.
38 * Redistribution and use in source and binary forms, with or without
39 * modification, are permitted provided that the following conditions
41 * 1. Redistributions of source code must retain the above copyright
42 * notice, this list of conditions and the following disclaimer.
43 * 2. Redistributions in binary form must reproduce the above copyright
44 * notice, this list of conditions and the following disclaimer in the
45 * documentation and/or other materials provided with the distribution.
46 * 4. Neither the name of the University nor the names of its contributors
47 * may be used to endorse or promote products derived from this software
48 * without specific prior written permission.
50 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
51 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
52 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
53 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
54 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
55 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
56 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
57 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
58 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
59 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
62 * @(#)udp_usrreq.c 8.6 (Berkeley) 5/23/95
63 * $FreeBSD: src/sys/netinet/udp_usrreq.c,v 1.64.2.18 2003/01/24 05:11:34 sam Exp $
66 #include "opt_ipsec.h"
67 #include "opt_inet6.h"
69 #include <sys/param.h>
70 #include <sys/systm.h>
71 #include <sys/kernel.h>
72 #include <sys/malloc.h>
74 #include <sys/domain.h>
77 #include <sys/protosw.h>
78 #include <sys/socket.h>
79 #include <sys/socketvar.h>
80 #include <sys/sysctl.h>
81 #include <sys/syslog.h>
82 #include <sys/in_cksum.h>
84 #include <sys/thread2.h>
85 #include <sys/socketvar2.h>
86 #include <sys/serialize.h>
88 #include <machine/stdarg.h>
91 #include <net/route.h>
92 #include <net/netmsg2.h>
93 #include <net/netisr2.h>
95 #include <netinet/in.h>
96 #include <netinet/in_systm.h>
97 #include <netinet/ip.h>
99 #include <netinet/ip6.h>
101 #include <netinet/in_pcb.h>
102 #include <netinet/in_var.h>
103 #include <netinet/ip_var.h>
105 #include <netinet6/ip6_var.h>
107 #include <netinet/ip_icmp.h>
108 #include <netinet/icmp_var.h>
109 #include <netinet/udp.h>
110 #include <netinet/udp_var.h>
113 #include <netproto/ipsec/ipsec.h>
117 #include <netinet6/ipsec.h>
121 * UDP protocol implementation.
122 * Per RFC 768, August, 1980.
125 static int udpcksum = 1;
127 static int udpcksum = 0; /* XXX */
129 SYSCTL_INT(_net_inet_udp, UDPCTL_CHECKSUM, checksum, CTLFLAG_RW,
130 &udpcksum, 0, "Enable checksumming of UDP packets");
133 SYSCTL_INT(_net_inet_udp, OID_AUTO, log_in_vain, CTLFLAG_RW,
134 &log_in_vain, 0, "Log all incoming UDP packets");
136 static int blackhole = 0;
137 SYSCTL_INT(_net_inet_udp, OID_AUTO, blackhole, CTLFLAG_RW,
138 &blackhole, 0, "Do not send port unreachables for refused connects");
140 static int strict_mcast_mship = 1;
141 SYSCTL_INT(_net_inet_udp, OID_AUTO, strict_mcast_mship, CTLFLAG_RW,
142 &strict_mcast_mship, 0, "Only send multicast to member sockets");
144 int udp_sosend_async = 1;
145 SYSCTL_INT(_net_inet_udp, OID_AUTO, sosend_async, CTLFLAG_RW,
146 &udp_sosend_async, 0, "UDP asynchronized pru_send");
148 int udp_sosend_prepend = 1;
149 SYSCTL_INT(_net_inet_udp, OID_AUTO, sosend_prepend, CTLFLAG_RW,
150 &udp_sosend_prepend, 0,
151 "Prepend enough space for proto and link header in pru_send");
153 struct inpcbinfo udbinfo;
155 static struct netisr_barrier *udbinfo_br;
156 static struct lwkt_serialize udbinfo_slize = LWKT_SERIALIZE_INITIALIZER;
159 #define UDBHASHSIZE 16
162 struct udpstat udpstat_percpu[MAXCPU] __cachealign;
164 static struct sockaddr_in udp_in = { sizeof udp_in, AF_INET };
167 struct sockaddr_in6 uin6_sin;
168 u_char uin6_init_done : 1;
170 { sizeof udp_in6.uin6_sin, AF_INET6 },
174 struct ip6_hdr uip6_ip6;
175 u_char uip6_init_done : 1;
179 static void udp_append (struct inpcb *last, struct ip *ip,
180 struct mbuf *n, int off);
182 static void ip_2_ip6_hdr (struct ip6_hdr *ip6, struct ip *ip);
185 static int udp_connect_oncpu(struct socket *so, struct thread *td,
186 struct sockaddr_in *sin, struct sockaddr_in *if_sin);
187 static int udp_output (struct inpcb *, struct mbuf *, struct sockaddr *,
188 struct thread *, int);
195 in_pcbinfo_init(&udbinfo);
196 udbinfo.hashbase = hashinit(UDBHASHSIZE, M_PCB, &udbinfo.hashmask);
197 udbinfo.porthashbase = hashinit(UDBHASHSIZE, M_PCB,
198 &udbinfo.porthashmask);
199 udbinfo.wildcardhashbase = hashinit(UDBHASHSIZE, M_PCB,
200 &udbinfo.wildcardhashmask);
201 udbinfo.ipi_size = sizeof(struct inpcb);
203 udbinfo_br = netisr_barrier_create();
206 * Initialize UDP statistics counters for each CPU.
208 for (cpu = 0; cpu < ncpus; ++cpu)
209 bzero(&udpstat_percpu[cpu], sizeof(struct udpstat));
213 sysctl_udpstat(SYSCTL_HANDLER_ARGS)
217 for (cpu = 0; cpu < ncpus; ++cpu) {
218 if ((error = SYSCTL_OUT(req, &udpstat_percpu[cpu],
219 sizeof(struct udpstat))))
221 if ((error = SYSCTL_IN(req, &udpstat_percpu[cpu],
222 sizeof(struct udpstat))))
228 SYSCTL_PROC(_net_inet_udp, UDPCTL_STATS, stats, (CTLTYPE_OPAQUE | CTLFLAG_RW),
229 0, 0, sysctl_udpstat, "S,udpstat", "UDP statistics");
232 * Check multicast packets to make sure they are only sent to sockets with
233 * multicast memberships for the packet's destination address and arrival
234 * interface. Multicast packets to multicast-unaware sockets are also
237 * Returns 0 if the packet is acceptable, -1 if it is not.
240 check_multicast_membership(struct ip *ip, struct inpcb *inp, struct mbuf *m)
243 struct ip_moptions *mopt;
245 if (strict_mcast_mship == 0 ||
246 !IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
249 mopt = inp->inp_moptions;
252 for (mshipno = 0; mshipno < mopt->imo_num_memberships; ++mshipno) {
253 struct in_multi *maddr = mopt->imo_membership[mshipno];
255 if (ip->ip_dst.s_addr == maddr->inm_addr.s_addr &&
256 m->m_pkthdr.rcvif == maddr->inm_ifp) {
264 udp_input(struct mbuf **mp, int *offp, int proto)
271 struct mbuf *opts = NULL;
274 struct sockaddr *append_sa;
281 udp_stat.udps_ipackets++;
284 * Strip IP options, if any; should skip this,
285 * make available to user, and use on returned packets,
286 * but we don't yet have a way to check the checksum
287 * with options still present.
289 if (iphlen > sizeof(struct ip)) {
291 iphlen = sizeof(struct ip);
295 * IP and UDP headers are together in first mbuf.
296 * Already checked and pulled up in ip_demux().
298 KASSERT(m->m_len >= iphlen + sizeof(struct udphdr),
299 ("UDP header not in one mbuf"));
301 ip = mtod(m, struct ip *);
302 uh = (struct udphdr *)((caddr_t)ip + iphlen);
304 /* destination port of 0 is illegal, based on RFC768. */
305 if (uh->uh_dport == 0)
309 * Make mbuf data length reflect UDP length.
310 * If not enough data to reflect UDP length, drop.
312 len = ntohs((u_short)uh->uh_ulen);
313 if (ip->ip_len != len) {
314 if (len > ip->ip_len || len < sizeof(struct udphdr)) {
315 udp_stat.udps_badlen++;
318 m_adj(m, len - ip->ip_len);
319 /* ip->ip_len = len; */
322 * Save a copy of the IP header in case we want restore it
323 * for sending an ICMP error message in response.
328 * Checksum extended UDP header and data.
331 if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) {
332 if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR)
333 uh->uh_sum = m->m_pkthdr.csum_data;
335 uh->uh_sum = in_pseudo(ip->ip_src.s_addr,
336 ip->ip_dst.s_addr, htonl((u_short)len +
337 m->m_pkthdr.csum_data + IPPROTO_UDP));
338 uh->uh_sum ^= 0xffff;
342 bcopy(((struct ipovly *)ip)->ih_x1, b, 9);
343 bzero(((struct ipovly *)ip)->ih_x1, 9);
344 ((struct ipovly *)ip)->ih_len = uh->uh_ulen;
345 uh->uh_sum = in_cksum(m, len + sizeof(struct ip));
346 bcopy(b, ((struct ipovly *)ip)->ih_x1, 9);
349 udp_stat.udps_badsum++;
351 return(IPPROTO_DONE);
354 udp_stat.udps_nosum++;
356 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) ||
357 in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif)) {
361 * Deliver a multicast or broadcast datagram to *all* sockets
362 * for which the local and remote addresses and ports match
363 * those of the incoming datagram. This allows more than
364 * one process to receive multi/broadcasts on the same port.
365 * (This really ought to be done for unicast datagrams as
366 * well, but that would cause problems with existing
367 * applications that open both address-specific sockets and
368 * a wildcard socket listening to the same port -- they would
369 * end up receiving duplicates of every unicast datagram.
370 * Those applications open the multiple sockets to overcome an
371 * inadequacy of the UDP socket interface, but for backwards
372 * compatibility we avoid the problem here rather than
373 * fixing the interface. Maybe 4.5BSD will remedy this?)
377 * Construct sockaddr format source address.
379 udp_in.sin_port = uh->uh_sport;
380 udp_in.sin_addr = ip->ip_src;
382 * Locate pcb(s) for datagram.
383 * (Algorithm copied from raw_intr().)
387 udp_in6.uin6_init_done = udp_ip6.uip6_init_done = 0;
389 LIST_FOREACH(inp, &udbinfo.pcblisthead, inp_list) {
390 KKASSERT((inp->inp_flags & INP_PLACEMARKER) == 0);
392 if (!(inp->inp_vflag & INP_IPV4))
395 if (inp->inp_lport != uh->uh_dport)
397 if (inp->inp_laddr.s_addr != INADDR_ANY) {
398 if (inp->inp_laddr.s_addr !=
402 if (inp->inp_faddr.s_addr != INADDR_ANY) {
403 if (inp->inp_faddr.s_addr !=
405 inp->inp_fport != uh->uh_sport)
409 if (check_multicast_membership(ip, inp, m) < 0)
416 /* check AH/ESP integrity. */
417 if (ipsec4_in_reject_so(m, last->inp_socket))
418 ipsecstat.in_polvio++;
419 /* do not inject data to pcb */
423 /* check AH/ESP integrity. */
424 if (ipsec4_in_reject(m, last))
427 #endif /*FAST_IPSEC*/
428 if ((n = m_copypacket(m, MB_DONTWAIT)) != NULL)
429 udp_append(last, ip, n,
431 sizeof(struct udphdr));
435 * Don't look for additional matches if this one does
436 * not have either the SO_REUSEPORT or SO_REUSEADDR
437 * socket options set. This heuristic avoids searching
438 * through all pcbs in the common case of a non-shared
439 * port. It * assumes that an application will never
440 * clear these options after setting them.
442 if (!(last->inp_socket->so_options &
443 (SO_REUSEPORT | SO_REUSEADDR)))
449 * No matching pcb found; discard datagram.
450 * (No need to send an ICMP Port Unreachable
451 * for a broadcast or multicast datgram.)
453 udp_stat.udps_noportbcast++;
457 /* check AH/ESP integrity. */
458 if (ipsec4_in_reject_so(m, last->inp_socket)) {
459 ipsecstat.in_polvio++;
464 /* check AH/ESP integrity. */
465 if (ipsec4_in_reject(m, last))
467 #endif /*FAST_IPSEC*/
468 udp_append(last, ip, m, iphlen + sizeof(struct udphdr));
469 return(IPPROTO_DONE);
472 * Locate pcb for datagram.
474 inp = in_pcblookup_hash(&udbinfo, ip->ip_src, uh->uh_sport,
475 ip->ip_dst, uh->uh_dport, 1, m->m_pkthdr.rcvif);
478 char buf[sizeof "aaa.bbb.ccc.ddd"];
480 strcpy(buf, inet_ntoa(ip->ip_dst));
482 "Connection attempt to UDP %s:%d from %s:%d\n",
483 buf, ntohs(uh->uh_dport), inet_ntoa(ip->ip_src),
484 ntohs(uh->uh_sport));
486 udp_stat.udps_noport++;
487 if (m->m_flags & (M_BCAST | M_MCAST)) {
488 udp_stat.udps_noportbcast++;
494 if (badport_bandlim(BANDLIM_ICMP_UNREACH) < 0)
498 ip->ip_len += iphlen;
499 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0);
500 return(IPPROTO_DONE);
503 if (ipsec4_in_reject_so(m, inp->inp_socket)) {
504 ipsecstat.in_polvio++;
509 if (ipsec4_in_reject(m, inp))
511 #endif /*FAST_IPSEC*/
513 * Check the minimum TTL for socket.
515 if (ip->ip_ttl < inp->inp_ip_minttl)
519 * Construct sockaddr format source address.
520 * Stuff source address and datagram in user buffer.
522 udp_in.sin_port = uh->uh_sport;
523 udp_in.sin_addr = ip->ip_src;
524 if ((inp->inp_flags & INP_CONTROLOPTS) ||
525 (inp->inp_socket->so_options & SO_TIMESTAMP)) {
527 if (inp->inp_vflag & INP_IPV6) {
530 ip_2_ip6_hdr(&udp_ip6.uip6_ip6, ip);
531 savedflags = inp->inp_flags;
532 inp->inp_flags &= ~INP_UNMAPPABLEOPTS;
533 ip6_savecontrol(inp, &opts, &udp_ip6.uip6_ip6, m);
534 inp->inp_flags = savedflags;
537 ip_savecontrol(inp, &opts, ip, m);
539 m_adj(m, iphlen + sizeof(struct udphdr));
541 if (inp->inp_vflag & INP_IPV6) {
542 in6_sin_2_v4mapsin6(&udp_in, &udp_in6.uin6_sin);
543 append_sa = (struct sockaddr *)&udp_in6;
546 append_sa = (struct sockaddr *)&udp_in;
548 lwkt_gettoken(&inp->inp_socket->so_rcv.ssb_token);
549 if (ssb_appendaddr(&inp->inp_socket->so_rcv, append_sa, m, opts) == 0) {
550 udp_stat.udps_fullsock++;
551 lwkt_reltoken(&inp->inp_socket->so_rcv.ssb_token);
554 lwkt_reltoken(&inp->inp_socket->so_rcv.ssb_token);
555 sorwakeup(inp->inp_socket);
556 return(IPPROTO_DONE);
561 return(IPPROTO_DONE);
566 ip_2_ip6_hdr(struct ip6_hdr *ip6, struct ip *ip)
568 bzero(ip6, sizeof *ip6);
570 ip6->ip6_vfc = IPV6_VERSION;
571 ip6->ip6_plen = ip->ip_len;
572 ip6->ip6_nxt = ip->ip_p;
573 ip6->ip6_hlim = ip->ip_ttl;
574 ip6->ip6_src.s6_addr32[2] = ip6->ip6_dst.s6_addr32[2] =
576 ip6->ip6_src.s6_addr32[3] = ip->ip_src.s_addr;
577 ip6->ip6_dst.s6_addr32[3] = ip->ip_dst.s_addr;
582 * subroutine of udp_input(), mainly for source code readability.
583 * caller must properly init udp_ip6 and udp_in6 beforehand.
586 udp_append(struct inpcb *last, struct ip *ip, struct mbuf *n, int off)
588 struct sockaddr *append_sa;
589 struct mbuf *opts = NULL;
591 if (last->inp_flags & INP_CONTROLOPTS ||
592 last->inp_socket->so_options & SO_TIMESTAMP) {
594 if (last->inp_vflag & INP_IPV6) {
597 if (udp_ip6.uip6_init_done == 0) {
598 ip_2_ip6_hdr(&udp_ip6.uip6_ip6, ip);
599 udp_ip6.uip6_init_done = 1;
601 savedflags = last->inp_flags;
602 last->inp_flags &= ~INP_UNMAPPABLEOPTS;
603 ip6_savecontrol(last, &opts, &udp_ip6.uip6_ip6, n);
604 last->inp_flags = savedflags;
607 ip_savecontrol(last, &opts, ip, n);
610 if (last->inp_vflag & INP_IPV6) {
611 if (udp_in6.uin6_init_done == 0) {
612 in6_sin_2_v4mapsin6(&udp_in, &udp_in6.uin6_sin);
613 udp_in6.uin6_init_done = 1;
615 append_sa = (struct sockaddr *)&udp_in6.uin6_sin;
618 append_sa = (struct sockaddr *)&udp_in;
620 lwkt_gettoken(&last->inp_socket->so_rcv.ssb_token);
621 if (ssb_appendaddr(&last->inp_socket->so_rcv, append_sa, n, opts) == 0) {
625 udp_stat.udps_fullsock++;
627 sorwakeup(last->inp_socket);
629 lwkt_reltoken(&last->inp_socket->so_rcv.ssb_token);
633 * Notify a udp user of an asynchronous error;
634 * just wake up so that he can collect error status.
637 udp_notify(struct inpcb *inp, int error)
639 inp->inp_socket->so_error = error;
640 sorwakeup(inp->inp_socket);
641 sowwakeup(inp->inp_socket);
644 struct netmsg_udp_notify {
645 struct netmsg_base base;
646 void (*nm_notify)(struct inpcb *, int);
647 struct in_addr nm_faddr;
652 udp_notifyall_oncpu(netmsg_t msg)
654 struct netmsg_udp_notify *nm = (struct netmsg_udp_notify *)msg;
659 in_pcbnotifyall(&udbinfo.pcblisthead, nm->nm_faddr,
660 nm->nm_arg, nm->nm_notify);
661 lwkt_replymsg(&nm->base.lmsg, 0);
664 /* XXX currently udp only runs on cpu 0 */
665 nextcpu = mycpuid + 1;
666 if (nextcpu < ncpus2)
667 lwkt_forwardmsg(netisr_cpuport(nextcpu), &nm->base.lmsg);
669 lwkt_replymsg(&nmsg->base.lmsg, 0);
674 udp_rtchange(struct inpcb *inp, int err)
676 /* XXX Nuke this, once UDP inpcbs are CPU localized */
677 if (inp->inp_route.ro_rt && inp->inp_route.ro_rt->rt_cpuid == mycpuid) {
678 rtfree(inp->inp_route.ro_rt);
679 inp->inp_route.ro_rt = NULL;
681 * A new route can be allocated the next time
682 * output is attempted.
688 udp_ctlinput(netmsg_t msg)
690 struct sockaddr *sa = msg->ctlinput.nm_arg;
691 struct ip *ip = msg->ctlinput.nm_extra;
692 int cmd = msg->ctlinput.nm_cmd;
694 void (*notify) (struct inpcb *, int) = udp_notify;
695 struct in_addr faddr;
698 KKASSERT(&curthread->td_msgport == netisr_cpuport(0));
700 faddr = ((struct sockaddr_in *)sa)->sin_addr;
701 if (sa->sa_family != AF_INET || faddr.s_addr == INADDR_ANY)
704 if (PRC_IS_REDIRECT(cmd)) {
706 notify = udp_rtchange;
707 } else if (cmd == PRC_HOSTDEAD) {
709 } else if ((unsigned)cmd >= PRC_NCMDS || inetctlerrmap[cmd] == 0) {
714 uh = (struct udphdr *)((caddr_t)ip + (ip->ip_hl << 2));
715 inp = in_pcblookup_hash(&udbinfo, faddr, uh->uh_dport,
716 ip->ip_src, uh->uh_sport, 0, NULL);
717 if (inp != NULL && inp->inp_socket != NULL)
718 (*notify)(inp, inetctlerrmap[cmd]);
719 } else if (PRC_IS_REDIRECT(cmd)) {
720 struct netmsg_udp_notify *nm;
722 KKASSERT(&curthread->td_msgport == netisr_cpuport(0));
723 nm = kmalloc(sizeof(*nm), M_LWKTMSG, M_INTWAIT);
724 netmsg_init(&nm->base, NULL, &netisr_afree_rport,
725 0, udp_notifyall_oncpu);
726 nm->nm_faddr = faddr;
727 nm->nm_arg = inetctlerrmap[cmd];
728 nm->nm_notify = notify;
729 lwkt_sendmsg(netisr_cpuport(0), &nm->base.lmsg);
732 * XXX We should forward msg upon PRC_HOSTHEAD and ip == NULL,
733 * once UDP inpcbs are CPU localized
735 KKASSERT(&curthread->td_msgport == netisr_cpuport(0));
736 in_pcbnotifyall(&udbinfo.pcblisthead, faddr, inetctlerrmap[cmd],
740 lwkt_replymsg(&msg->lmsg, 0);
744 udp_pcblist(SYSCTL_HANDLER_ARGS)
750 error = in_pcblist_global_nomarker(oidp, arg1, arg2, req, &xi, &nxi);
754 KKASSERT(xi == NULL);
758 KKASSERT(xi == NULL);
762 for (i = 0; i < nxi; ++i) {
763 error = SYSCTL_OUT(req, &xi[i], sizeof(xi[i]));
771 SYSCTL_PROC(_net_inet_udp, UDPCTL_PCBLIST, pcblist, CTLFLAG_RD, &udbinfo, 0,
772 udp_pcblist, "S,xinpcb", "List of active UDP sockets");
775 udp_getcred(SYSCTL_HANDLER_ARGS)
777 struct sockaddr_in addrs[2];
778 struct ucred cred0, *cred = NULL;
782 error = priv_check(req->td, PRIV_ROOT);
785 error = SYSCTL_IN(req, addrs, sizeof addrs);
790 inp = in_pcblookup_hash(&udbinfo, addrs[1].sin_addr, addrs[1].sin_port,
791 addrs[0].sin_addr, addrs[0].sin_port, 1, NULL);
792 if (inp == NULL || inp->inp_socket == NULL) {
795 if (inp->inp_socket->so_cred != NULL) {
796 cred0 = *(inp->inp_socket->so_cred);
805 return SYSCTL_OUT(req, cred, sizeof(struct ucred));
808 SYSCTL_PROC(_net_inet_udp, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW,
809 0, 0, udp_getcred, "S,ucred", "Get the ucred of a UDP connection");
812 udp_output(struct inpcb *inp, struct mbuf *m, struct sockaddr *dstaddr,
813 struct thread *td, int flags)
816 int len = m->m_pkthdr.len;
817 struct sockaddr_in *sin; /* really is initialized before use */
818 int error = 0, lport_any = 0;
820 if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) {
825 if (inp->inp_lport == 0) { /* unbound socket */
826 error = in_pcbbind(inp, NULL, td);
830 udbinfo_barrier_set();
831 in_pcbinswildcardhash(inp);
832 udbinfo_barrier_rem();
836 if (dstaddr != NULL) { /* destination address specified */
837 if (inp->inp_faddr.s_addr != INADDR_ANY) {
838 /* already connected */
842 sin = (struct sockaddr_in *)dstaddr;
843 if (!prison_remote_ip(td, (struct sockaddr *)&sin)) {
844 error = EAFNOSUPPORT; /* IPv6 only jail */
848 if (inp->inp_faddr.s_addr == INADDR_ANY) {
849 /* no destination specified and not already connected */
857 * Calculate data length and get a mbuf
858 * for UDP and IP headers.
860 M_PREPEND(m, sizeof(struct udpiphdr), MB_DONTWAIT);
867 * Fill in mbuf with extended UDP header
868 * and addresses and length put into network format.
870 ui = mtod(m, struct udpiphdr *);
871 bzero(ui->ui_x1, sizeof ui->ui_x1); /* XXX still needed? */
872 ui->ui_pr = IPPROTO_UDP;
875 * Set destination address.
877 if (dstaddr != NULL) { /* use specified destination */
878 ui->ui_dst = sin->sin_addr;
879 ui->ui_dport = sin->sin_port;
880 } else { /* use connected destination */
881 ui->ui_dst = inp->inp_faddr;
882 ui->ui_dport = inp->inp_fport;
886 * Set source address.
888 if (inp->inp_laddr.s_addr == INADDR_ANY) {
889 struct sockaddr_in *if_sin;
891 if (dstaddr == NULL) {
893 * connect() had (or should have) failed because
894 * the interface had no IP address, but the
895 * application proceeded to call send() anyways.
901 /* Look up outgoing interface. */
902 if ((error = in_pcbladdr(inp, dstaddr, &if_sin, td)))
904 ui->ui_src = if_sin->sin_addr; /* use address of interface */
905 } else if (!IN_MULTICAST(ntohl(inp->inp_laddr.s_addr))) {
906 ui->ui_src = inp->inp_laddr; /* use non-null bound address */
908 /* Bound to multicast address; let ip_output choose */
909 ui->ui_src.s_addr = INADDR_ANY;
911 ui->ui_sport = inp->inp_lport;
912 KASSERT(inp->inp_lport != 0, ("inp lport should have been bound"));
914 ui->ui_ulen = htons((u_short)len + sizeof(struct udphdr));
917 * Set up checksum and output datagram.
920 ui->ui_sum = in_pseudo(ui->ui_src.s_addr, ui->ui_dst.s_addr,
921 htons((u_short)len + sizeof(struct udphdr) + IPPROTO_UDP));
922 m->m_pkthdr.csum_flags = CSUM_UDP;
923 m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
924 m->m_pkthdr.csum_thlen = sizeof(struct udphdr);
928 ((struct ip *)ui)->ip_len = sizeof(struct udpiphdr) + len;
929 ((struct ip *)ui)->ip_ttl = inp->inp_ip_ttl; /* XXX */
930 ((struct ip *)ui)->ip_tos = inp->inp_ip_tos; /* XXX */
931 udp_stat.udps_opackets++;
933 error = ip_output(m, inp->inp_options, &inp->inp_route,
934 (inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST)) |
935 flags | IP_DEBUGROUTE,
936 inp->inp_moptions, inp);
939 * If this is the first data gram sent on an unbound and unconnected
940 * UDP socket, lport will be changed in this function. If target
941 * CPU after this lport changing is no longer the current CPU, then
942 * free the route entry allocated on the current CPU.
945 if (udp_addrcpu(inp->inp_faddr.s_addr, inp->inp_fport,
946 inp->inp_laddr.s_addr, inp->inp_lport) != mycpuid) {
948 struct route *ro = &inp->inp_route;
950 if (ro->ro_rt != NULL)
952 bzero(ro, sizeof(*ro));
954 panic("UDP activity should only be in netisr0");
965 u_long udp_sendspace = 9216; /* really max datagram size */
966 /* 40 1K datagrams */
967 SYSCTL_INT(_net_inet_udp, UDPCTL_MAXDGRAM, maxdgram, CTLFLAG_RW,
968 &udp_sendspace, 0, "Maximum outgoing UDP datagram size");
970 u_long udp_recvspace = 40 * (1024 +
972 sizeof(struct sockaddr_in6)
974 sizeof(struct sockaddr_in)
977 SYSCTL_INT(_net_inet_udp, UDPCTL_RECVSPACE, recvspace, CTLFLAG_RW,
978 &udp_recvspace, 0, "Maximum incoming UDP datagram size");
981 * NOTE: (so) is referenced from soabort*() and netmsg_pru_abort()
982 * will sofree() it when we return.
985 udp_abort(netmsg_t msg)
987 struct socket *so = msg->abort.base.nm_so;
991 KKASSERT(&curthread->td_msgport == netisr_cpuport(0));
995 soisdisconnected(so);
997 udbinfo_barrier_set();
999 udbinfo_barrier_rem();
1004 lwkt_replymsg(&msg->abort.base.lmsg, error);
1008 udp_attach(netmsg_t msg)
1010 struct socket *so = msg->attach.base.nm_so;
1011 struct pru_attach_info *ai = msg->attach.nm_ai;
1015 KKASSERT(&curthread->td_msgport == netisr_cpuport(0));
1022 error = soreserve(so, udp_sendspace, udp_recvspace, ai->sb_rlimit);
1026 udbinfo_barrier_set();
1027 error = in_pcballoc(so, &udbinfo);
1028 udbinfo_barrier_rem();
1034 * Set default port for protocol processing prior to bind/connect.
1036 sosetport(so, netisr_cpuport(0));
1038 inp = (struct inpcb *)so->so_pcb;
1039 inp->inp_vflag |= INP_IPV4;
1040 inp->inp_ip_ttl = ip_defttl;
1043 lwkt_replymsg(&msg->attach.base.lmsg, error);
1047 udp_bind(netmsg_t msg)
1049 struct socket *so = msg->bind.base.nm_so;
1050 struct sockaddr *nam = msg->bind.nm_nam;
1051 struct thread *td = msg->bind.nm_td;
1052 struct sockaddr_in *sin = (struct sockaddr_in *)nam;
1058 error = in_pcbbind(inp, nam, td);
1060 if (sin->sin_addr.s_addr != INADDR_ANY)
1061 inp->inp_flags |= INP_WASBOUND_NOTANY;
1063 udbinfo_barrier_set();
1064 in_pcbinswildcardhash(inp);
1065 udbinfo_barrier_rem();
1070 lwkt_replymsg(&msg->bind.base.lmsg, error);
1074 udp_connect(netmsg_t msg)
1076 struct socket *so = msg->connect.base.nm_so;
1077 struct sockaddr *nam = msg->connect.nm_nam;
1078 struct thread *td = msg->connect.nm_td;
1080 struct sockaddr_in *sin = (struct sockaddr_in *)nam;
1081 struct sockaddr_in *if_sin;
1085 KKASSERT(&curthread->td_msgport == netisr_cpuport(0));
1093 if (msg->connect.nm_reconnect & NMSG_RECONNECT_RECONNECT) {
1094 panic("UDP does not support RECONNECT");
1096 msg->connect.nm_reconnect &= ~NMSG_RECONNECT_RECONNECT;
1097 in_pcblink(inp, &udbinfo);
1101 if (inp->inp_faddr.s_addr != INADDR_ANY) {
1108 * Bind if we have to
1110 if (td->td_proc && td->td_proc->p_ucred->cr_prison != NULL &&
1111 inp->inp_laddr.s_addr == INADDR_ANY) {
1112 error = in_pcbbind(inp, NULL, td);
1118 * Calculate the correct protocol processing thread. The connect
1119 * operation must run there.
1121 error = in_pcbladdr(inp, nam, &if_sin, td);
1124 if (!prison_remote_ip(td, nam)) {
1125 error = EAFNOSUPPORT; /* IPv6 only jail */
1129 port = udp_addrport(sin->sin_addr.s_addr, sin->sin_port,
1130 inp->inp_laddr.s_addr, inp->inp_lport);
1131 if (port != &curthread->td_msgport) {
1133 struct route *ro = &inp->inp_route;
1136 * in_pcbladdr() may have allocated a route entry for us
1137 * on the current CPU, but we need a route entry on the
1138 * inpcb's owner CPU, so free it here.
1140 if (ro->ro_rt != NULL)
1142 bzero(ro, sizeof(*ro));
1145 * We are moving the protocol processing port the socket
1146 * is on, we have to unlink here and re-link on the
1149 in_pcbunlink(so->so_pcb, &udbinfo);
1150 /* in_pcbunlink(so->so_pcb, &udbinfo[mycpu->gd_cpuid]); */
1151 sosetport(so, port);
1152 msg->connect.nm_reconnect |= NMSG_RECONNECT_RECONNECT;
1153 msg->connect.base.nm_dispatch = udp_connect;
1155 lwkt_forwardmsg(port, &msg->connect.base.lmsg);
1156 /* msg invalid now */
1159 panic("UDP activity should only be in netisr0");
1162 KKASSERT(port == &curthread->td_msgport);
1163 error = udp_connect_oncpu(so, td, sin, if_sin);
1165 KKASSERT(msg->connect.nm_m == NULL);
1166 lwkt_replymsg(&msg->connect.base.lmsg, error);
1170 udp_connect_oncpu(struct socket *so, struct thread *td,
1171 struct sockaddr_in *sin, struct sockaddr_in *if_sin)
1176 udbinfo_barrier_set();
1179 if (inp->inp_flags & INP_WILDCARD)
1180 in_pcbremwildcardhash(inp);
1181 error = in_pcbconnect(inp, (struct sockaddr *)sin, td);
1185 * No more errors can occur, finish adjusting the socket
1186 * and change the processing port to reflect the connected
1187 * socket. Once set we can no longer safely mess with the
1191 } else if (error == EAFNOSUPPORT) { /* connection dissolved */
1193 * Follow traditional BSD behavior and retain
1194 * the local port binding. But, fix the old misbehavior
1195 * of overwriting any previously bound local address.
1197 if (!(inp->inp_flags & INP_WASBOUND_NOTANY))
1198 inp->inp_laddr.s_addr = INADDR_ANY;
1199 in_pcbinswildcardhash(inp);
1202 udbinfo_barrier_rem();
1207 udp_detach(netmsg_t msg)
1209 struct socket *so = msg->detach.base.nm_so;
1213 KKASSERT(&curthread->td_msgport == netisr_cpuport(0));
1217 udbinfo_barrier_set();
1219 udbinfo_barrier_rem();
1224 lwkt_replymsg(&msg->detach.base.lmsg, error);
1228 udp_disconnect(netmsg_t msg)
1230 struct socket *so = msg->disconnect.base.nm_so;
1235 KKASSERT(&curthread->td_msgport == netisr_cpuport(0));
1242 if (inp->inp_faddr.s_addr == INADDR_ANY) {
1249 udbinfo_barrier_set();
1250 in_pcbdisconnect(inp);
1251 udbinfo_barrier_rem();
1253 soclrstate(so, SS_ISCONNECTED); /* XXX */
1256 ro = &inp->inp_route;
1257 if (ro->ro_rt != NULL)
1259 bzero(ro, sizeof(*ro));
1262 lwkt_replymsg(&msg->disconnect.base.lmsg, error);
1266 udp_send(netmsg_t msg)
1268 struct socket *so = msg->send.base.nm_so;
1269 struct mbuf *m = msg->send.nm_m;
1270 struct sockaddr *addr = msg->send.nm_addr;
1271 int pru_flags = msg->send.nm_flags;
1275 KKASSERT(&curthread->td_msgport == netisr_cpuport(0));
1276 KKASSERT(msg->send.nm_control == NULL);
1280 struct thread *td = msg->send.nm_td;
1283 if (pru_flags & PRUS_DONTROUTE)
1284 flags |= SO_DONTROUTE;
1285 error = udp_output(inp, m, addr, td, flags);
1291 if (pru_flags & PRUS_FREEADDR)
1292 kfree(addr, M_SONAME);
1294 if ((pru_flags & PRUS_NOREPLY) == 0)
1295 lwkt_replymsg(&msg->send.base.lmsg, error);
1299 udp_shutdown(netmsg_t msg)
1301 struct socket *so = msg->shutdown.base.nm_so;
1305 KKASSERT(&curthread->td_msgport == netisr_cpuport(0));
1314 lwkt_replymsg(&msg->shutdown.base.lmsg, error);
1320 lwkt_serialize_enter(&udbinfo_slize);
1324 udbinfo_unlock(void)
1326 lwkt_serialize_exit(&udbinfo_slize);
1330 udbinfo_barrier_set(void)
1332 netisr_barrier_set(udbinfo_br);
1337 udbinfo_barrier_rem(void)
1340 netisr_barrier_rem(udbinfo_br);
1343 struct pr_usrreqs udp_usrreqs = {
1344 .pru_abort = udp_abort,
1345 .pru_accept = pr_generic_notsupp,
1346 .pru_attach = udp_attach,
1347 .pru_bind = udp_bind,
1348 .pru_connect = udp_connect,
1349 .pru_connect2 = pr_generic_notsupp,
1350 .pru_control = in_control_dispatch,
1351 .pru_detach = udp_detach,
1352 .pru_disconnect = udp_disconnect,
1353 .pru_listen = pr_generic_notsupp,
1354 .pru_peeraddr = in_setpeeraddr_dispatch,
1355 .pru_rcvd = pr_generic_notsupp,
1356 .pru_rcvoob = pr_generic_notsupp,
1357 .pru_send = udp_send,
1358 .pru_sense = pru_sense_null,
1359 .pru_shutdown = udp_shutdown,
1360 .pru_sockaddr = in_setsockaddr_dispatch,
1361 .pru_sosend = sosendudp,
1362 .pru_soreceive = soreceive