amd64 - add kvtop and add back ed(4) to AMD64_GENERIC
[dragonfly.git] / sys / netinet / udp_usrreq.c
blobac1c849b5b2b407751005f080fed6ffe320165b2
1 /*
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
6 * by Jeffrey M. Hsu.
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
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
31 * SUCH DAMAGE.
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
40 * are met:
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
64 * SUCH DAMAGE.
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>
78 #include <sys/mbuf.h>
79 #include <sys/domain.h>
80 #include <sys/proc.h>
81 #include <sys/priv.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/thread2.h>
88 #include <sys/in_cksum.h>
90 #include <machine/stdarg.h>
92 #include <vm/vm_zone.h>
94 #include <net/if.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>
101 #ifdef INET6
102 #include <netinet/ip6.h>
103 #endif
104 #include <netinet/in_pcb.h>
105 #include <netinet/in_var.h>
106 #include <netinet/ip_var.h>
107 #ifdef INET6
108 #include <netinet6/ip6_var.h>
109 #endif
110 #include <netinet/ip_icmp.h>
111 #include <netinet/icmp_var.h>
112 #include <netinet/udp.h>
113 #include <netinet/udp_var.h>
115 #ifdef FAST_IPSEC
116 #include <netproto/ipsec/ipsec.h>
117 #endif
119 #ifdef IPSEC
120 #include <netinet6/ipsec.h>
121 #endif
123 int udp_mpsafe_proto = 0;
124 TUNABLE_INT("net.inet.udp.mpsafe_proto", &udp_mpsafe_proto);
126 int udp_mpsafe_thread = NETMSG_SERVICE_ADAPTIVE;
127 TUNABLE_INT("net.inet.udp.mpsafe_thread", &udp_mpsafe_thread);
128 SYSCTL_INT(_net_inet_udp, OID_AUTO, mpsafe_thread, CTLFLAG_RW,
129 &udp_mpsafe_thread, 0,
130 "0:BGL, 1:Adaptive BGL, 2:No BGL(experimental)");
133 * UDP protocol implementation.
134 * Per RFC 768, August, 1980.
136 #ifndef COMPAT_42
137 static int udpcksum = 1;
138 #else
139 static int udpcksum = 0; /* XXX */
140 #endif
141 SYSCTL_INT(_net_inet_udp, UDPCTL_CHECKSUM, checksum, CTLFLAG_RW,
142 &udpcksum, 0, "");
144 int log_in_vain = 0;
145 SYSCTL_INT(_net_inet_udp, OID_AUTO, log_in_vain, CTLFLAG_RW,
146 &log_in_vain, 0, "Log all incoming UDP packets");
148 static int blackhole = 0;
149 SYSCTL_INT(_net_inet_udp, OID_AUTO, blackhole, CTLFLAG_RW,
150 &blackhole, 0, "Do not send port unreachables for refused connects");
152 static int strict_mcast_mship = 1;
153 SYSCTL_INT(_net_inet_udp, OID_AUTO, strict_mcast_mship, CTLFLAG_RW,
154 &strict_mcast_mship, 0, "Only send multicast to member sockets");
156 struct inpcbinfo udbinfo;
158 #ifndef UDBHASHSIZE
159 #define UDBHASHSIZE 16
160 #endif
162 struct udpstat udpstat; /* from udp_var.h */
163 SYSCTL_STRUCT(_net_inet_udp, UDPCTL_STATS, stats, CTLFLAG_RW,
164 &udpstat, udpstat, "UDP statistics (struct udpstat, netinet/udp_var.h)");
166 static struct sockaddr_in udp_in = { sizeof udp_in, AF_INET };
167 #ifdef INET6
168 struct udp_in6 {
169 struct sockaddr_in6 uin6_sin;
170 u_char uin6_init_done : 1;
171 } udp_in6 = {
172 { sizeof udp_in6.uin6_sin, AF_INET6 },
175 struct udp_ip6 {
176 struct ip6_hdr uip6_ip6;
177 u_char uip6_init_done : 1;
178 } udp_ip6;
179 #endif /* INET6 */
181 static void udp_append (struct inpcb *last, struct ip *ip,
182 struct mbuf *n, int off);
183 #ifdef INET6
184 static void ip_2_ip6_hdr (struct ip6_hdr *ip6, struct ip *ip);
185 #endif
187 static int udp_detach (struct socket *so);
188 static int udp_output (struct inpcb *, struct mbuf *, struct sockaddr *,
189 struct mbuf *, struct thread *);
191 void
192 udp_init(void)
194 in_pcbinfo_init(&udbinfo);
195 udbinfo.hashbase = hashinit(UDBHASHSIZE, M_PCB, &udbinfo.hashmask);
196 udbinfo.porthashbase = hashinit(UDBHASHSIZE, M_PCB,
197 &udbinfo.porthashmask);
198 udbinfo.wildcardhashbase = hashinit(UDBHASHSIZE, M_PCB,
199 &udbinfo.wildcardhashmask);
200 udbinfo.ipi_zone = zinit("udpcb", sizeof(struct inpcb), maxsockets,
201 ZONE_INTERRUPT, 0);
202 udp_thread_init();
206 * Check multicast packets to make sure they are only sent to sockets with
207 * multicast memberships for the packet's destination address and arrival
208 * interface. Multicast packets to multicast-unaware sockets are also
209 * disallowed.
211 * Returns 0 if the packet is acceptable, -1 if it is not.
213 static __inline int
214 check_multicast_membership(struct ip *ip, struct inpcb *inp, struct mbuf *m)
216 int mshipno;
217 struct ip_moptions *mopt;
219 if (strict_mcast_mship == 0 ||
220 !IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
221 return (0);
223 mopt = inp->inp_moptions;
224 if (mopt == NULL)
225 return (-1);
226 for (mshipno = 0; mshipno <= mopt->imo_num_memberships; ++mshipno) {
227 struct in_multi *maddr = mopt->imo_membership[mshipno];
229 if (ip->ip_dst.s_addr == maddr->inm_addr.s_addr &&
230 m->m_pkthdr.rcvif == maddr->inm_ifp) {
231 return (0);
234 return (-1);
237 void
238 udp_input(struct mbuf *m, ...)
240 int iphlen;
241 struct ip *ip;
242 struct udphdr *uh;
243 struct inpcb *inp;
244 struct mbuf *opts = NULL;
245 int len, off, proto;
246 struct ip save_ip;
247 struct sockaddr *append_sa;
248 __va_list ap;
250 __va_start(ap, m);
251 off = __va_arg(ap, int);
252 proto = __va_arg(ap, int);
253 __va_end(ap);
255 iphlen = off;
256 udpstat.udps_ipackets++;
259 * Strip IP options, if any; should skip this,
260 * make available to user, and use on returned packets,
261 * but we don't yet have a way to check the checksum
262 * with options still present.
264 if (iphlen > sizeof(struct ip)) {
265 ip_stripoptions(m);
266 iphlen = sizeof(struct ip);
270 * IP and UDP headers are together in first mbuf.
271 * Already checked and pulled up in ip_demux().
273 KASSERT(m->m_len >= iphlen + sizeof(struct udphdr),
274 ("UDP header not in one mbuf"));
276 ip = mtod(m, struct ip *);
277 uh = (struct udphdr *)((caddr_t)ip + iphlen);
279 /* destination port of 0 is illegal, based on RFC768. */
280 if (uh->uh_dport == 0)
281 goto bad;
284 * Make mbuf data length reflect UDP length.
285 * If not enough data to reflect UDP length, drop.
287 len = ntohs((u_short)uh->uh_ulen);
288 if (ip->ip_len != len) {
289 if (len > ip->ip_len || len < sizeof(struct udphdr)) {
290 udpstat.udps_badlen++;
291 goto bad;
293 m_adj(m, len - ip->ip_len);
294 /* ip->ip_len = len; */
297 * Save a copy of the IP header in case we want restore it
298 * for sending an ICMP error message in response.
300 save_ip = *ip;
303 * Checksum extended UDP header and data.
305 if (uh->uh_sum) {
306 if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) {
307 if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR)
308 uh->uh_sum = m->m_pkthdr.csum_data;
309 else
310 uh->uh_sum = in_pseudo(ip->ip_src.s_addr,
311 ip->ip_dst.s_addr, htonl((u_short)len +
312 m->m_pkthdr.csum_data + IPPROTO_UDP));
313 uh->uh_sum ^= 0xffff;
314 } else {
315 char b[9];
317 bcopy(((struct ipovly *)ip)->ih_x1, b, 9);
318 bzero(((struct ipovly *)ip)->ih_x1, 9);
319 ((struct ipovly *)ip)->ih_len = uh->uh_ulen;
320 uh->uh_sum = in_cksum(m, len + sizeof(struct ip));
321 bcopy(b, ((struct ipovly *)ip)->ih_x1, 9);
323 if (uh->uh_sum) {
324 udpstat.udps_badsum++;
325 m_freem(m);
326 return;
328 } else
329 udpstat.udps_nosum++;
331 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) ||
332 in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif)) {
333 struct inpcb *last;
336 * Deliver a multicast or broadcast datagram to *all* sockets
337 * for which the local and remote addresses and ports match
338 * those of the incoming datagram. This allows more than
339 * one process to receive multi/broadcasts on the same port.
340 * (This really ought to be done for unicast datagrams as
341 * well, but that would cause problems with existing
342 * applications that open both address-specific sockets and
343 * a wildcard socket listening to the same port -- they would
344 * end up receiving duplicates of every unicast datagram.
345 * Those applications open the multiple sockets to overcome an
346 * inadequacy of the UDP socket interface, but for backwards
347 * compatibility we avoid the problem here rather than
348 * fixing the interface. Maybe 4.5BSD will remedy this?)
352 * Construct sockaddr format source address.
354 udp_in.sin_port = uh->uh_sport;
355 udp_in.sin_addr = ip->ip_src;
357 * Locate pcb(s) for datagram.
358 * (Algorithm copied from raw_intr().)
360 last = NULL;
361 #ifdef INET6
362 udp_in6.uin6_init_done = udp_ip6.uip6_init_done = 0;
363 #endif
364 LIST_FOREACH(inp, &udbinfo.pcblisthead, inp_list) {
365 if (inp->inp_flags & INP_PLACEMARKER)
366 continue;
367 #ifdef INET6
368 if (!(inp->inp_vflag & INP_IPV4))
369 continue;
370 #endif
371 if (inp->inp_lport != uh->uh_dport)
372 continue;
373 if (inp->inp_laddr.s_addr != INADDR_ANY) {
374 if (inp->inp_laddr.s_addr !=
375 ip->ip_dst.s_addr)
376 continue;
378 if (inp->inp_faddr.s_addr != INADDR_ANY) {
379 if (inp->inp_faddr.s_addr !=
380 ip->ip_src.s_addr ||
381 inp->inp_fport != uh->uh_sport)
382 continue;
385 if (check_multicast_membership(ip, inp, m) < 0)
386 continue;
388 if (last != NULL) {
389 struct mbuf *n;
391 #ifdef IPSEC
392 /* check AH/ESP integrity. */
393 if (ipsec4_in_reject_so(m, last->inp_socket))
394 ipsecstat.in_polvio++;
395 /* do not inject data to pcb */
396 else
397 #endif /*IPSEC*/
398 #ifdef FAST_IPSEC
399 /* check AH/ESP integrity. */
400 if (ipsec4_in_reject(m, last))
402 else
403 #endif /*FAST_IPSEC*/
404 if ((n = m_copypacket(m, MB_DONTWAIT)) != NULL)
405 udp_append(last, ip, n,
406 iphlen +
407 sizeof(struct udphdr));
409 last = inp;
411 * Don't look for additional matches if this one does
412 * not have either the SO_REUSEPORT or SO_REUSEADDR
413 * socket options set. This heuristic avoids searching
414 * through all pcbs in the common case of a non-shared
415 * port. It * assumes that an application will never
416 * clear these options after setting them.
418 if (!(last->inp_socket->so_options &
419 (SO_REUSEPORT | SO_REUSEADDR)))
420 break;
423 if (last == NULL) {
425 * No matching pcb found; discard datagram.
426 * (No need to send an ICMP Port Unreachable
427 * for a broadcast or multicast datgram.)
429 udpstat.udps_noportbcast++;
430 goto bad;
432 #ifdef IPSEC
433 /* check AH/ESP integrity. */
434 if (ipsec4_in_reject_so(m, last->inp_socket)) {
435 ipsecstat.in_polvio++;
436 goto bad;
438 #endif /*IPSEC*/
439 #ifdef FAST_IPSEC
440 /* check AH/ESP integrity. */
441 if (ipsec4_in_reject(m, last))
442 goto bad;
443 #endif /*FAST_IPSEC*/
444 udp_append(last, ip, m, iphlen + sizeof(struct udphdr));
445 return;
448 * Locate pcb for datagram.
450 inp = in_pcblookup_hash(&udbinfo, ip->ip_src, uh->uh_sport,
451 ip->ip_dst, uh->uh_dport, 1, m->m_pkthdr.rcvif);
452 if (inp == NULL) {
453 if (log_in_vain) {
454 char buf[sizeof "aaa.bbb.ccc.ddd"];
456 strcpy(buf, inet_ntoa(ip->ip_dst));
457 log(LOG_INFO,
458 "Connection attempt to UDP %s:%d from %s:%d\n",
459 buf, ntohs(uh->uh_dport), inet_ntoa(ip->ip_src),
460 ntohs(uh->uh_sport));
462 udpstat.udps_noport++;
463 if (m->m_flags & (M_BCAST | M_MCAST)) {
464 udpstat.udps_noportbcast++;
465 goto bad;
467 if (blackhole)
468 goto bad;
469 #ifdef ICMP_BANDLIM
470 if (badport_bandlim(BANDLIM_ICMP_UNREACH) < 0)
471 goto bad;
472 #endif
473 *ip = save_ip;
474 ip->ip_len += iphlen;
475 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0);
476 return;
478 #ifdef IPSEC
479 if (ipsec4_in_reject_so(m, inp->inp_socket)) {
480 ipsecstat.in_polvio++;
481 goto bad;
483 #endif /*IPSEC*/
484 #ifdef FAST_IPSEC
485 if (ipsec4_in_reject(m, inp))
486 goto bad;
487 #endif /*FAST_IPSEC*/
489 * Check the minimum TTL for socket.
491 if (ip->ip_ttl < inp->inp_ip_minttl)
492 goto bad;
495 * Construct sockaddr format source address.
496 * Stuff source address and datagram in user buffer.
498 udp_in.sin_port = uh->uh_sport;
499 udp_in.sin_addr = ip->ip_src;
500 if ((inp->inp_flags & INP_CONTROLOPTS) ||
501 (inp->inp_socket->so_options & SO_TIMESTAMP)) {
502 #ifdef INET6
503 if (inp->inp_vflag & INP_IPV6) {
504 int savedflags;
506 ip_2_ip6_hdr(&udp_ip6.uip6_ip6, ip);
507 savedflags = inp->inp_flags;
508 inp->inp_flags &= ~INP_UNMAPPABLEOPTS;
509 ip6_savecontrol(inp, &opts, &udp_ip6.uip6_ip6, m);
510 inp->inp_flags = savedflags;
511 } else
512 #endif
513 ip_savecontrol(inp, &opts, ip, m);
515 m_adj(m, iphlen + sizeof(struct udphdr));
516 #ifdef INET6
517 if (inp->inp_vflag & INP_IPV6) {
518 in6_sin_2_v4mapsin6(&udp_in, &udp_in6.uin6_sin);
519 append_sa = (struct sockaddr *)&udp_in6;
520 } else
521 #endif
522 append_sa = (struct sockaddr *)&udp_in;
523 if (ssb_appendaddr(&inp->inp_socket->so_rcv, append_sa, m, opts) == 0) {
524 udpstat.udps_fullsock++;
525 goto bad;
527 sorwakeup(inp->inp_socket);
528 return;
529 bad:
530 m_freem(m);
531 if (opts)
532 m_freem(opts);
533 return;
536 #ifdef INET6
537 static void
538 ip_2_ip6_hdr(struct ip6_hdr *ip6, struct ip *ip)
540 bzero(ip6, sizeof *ip6);
542 ip6->ip6_vfc = IPV6_VERSION;
543 ip6->ip6_plen = ip->ip_len;
544 ip6->ip6_nxt = ip->ip_p;
545 ip6->ip6_hlim = ip->ip_ttl;
546 ip6->ip6_src.s6_addr32[2] = ip6->ip6_dst.s6_addr32[2] =
547 IPV6_ADDR_INT32_SMP;
548 ip6->ip6_src.s6_addr32[3] = ip->ip_src.s_addr;
549 ip6->ip6_dst.s6_addr32[3] = ip->ip_dst.s_addr;
551 #endif
554 * subroutine of udp_input(), mainly for source code readability.
555 * caller must properly init udp_ip6 and udp_in6 beforehand.
557 static void
558 udp_append(struct inpcb *last, struct ip *ip, struct mbuf *n, int off)
560 struct sockaddr *append_sa;
561 struct mbuf *opts = NULL;
563 if (last->inp_flags & INP_CONTROLOPTS ||
564 last->inp_socket->so_options & SO_TIMESTAMP) {
565 #ifdef INET6
566 if (last->inp_vflag & INP_IPV6) {
567 int savedflags;
569 if (udp_ip6.uip6_init_done == 0) {
570 ip_2_ip6_hdr(&udp_ip6.uip6_ip6, ip);
571 udp_ip6.uip6_init_done = 1;
573 savedflags = last->inp_flags;
574 last->inp_flags &= ~INP_UNMAPPABLEOPTS;
575 ip6_savecontrol(last, &opts, &udp_ip6.uip6_ip6, n);
576 last->inp_flags = savedflags;
577 } else
578 #endif
579 ip_savecontrol(last, &opts, ip, n);
581 #ifdef INET6
582 if (last->inp_vflag & INP_IPV6) {
583 if (udp_in6.uin6_init_done == 0) {
584 in6_sin_2_v4mapsin6(&udp_in, &udp_in6.uin6_sin);
585 udp_in6.uin6_init_done = 1;
587 append_sa = (struct sockaddr *)&udp_in6.uin6_sin;
588 } else
589 #endif
590 append_sa = (struct sockaddr *)&udp_in;
591 m_adj(n, off);
592 if (ssb_appendaddr(&last->inp_socket->so_rcv, append_sa, n, opts) == 0) {
593 m_freem(n);
594 if (opts)
595 m_freem(opts);
596 udpstat.udps_fullsock++;
597 } else
598 sorwakeup(last->inp_socket);
602 * Notify a udp user of an asynchronous error;
603 * just wake up so that he can collect error status.
605 void
606 udp_notify(struct inpcb *inp, int error)
608 inp->inp_socket->so_error = error;
609 sorwakeup(inp->inp_socket);
610 sowwakeup(inp->inp_socket);
613 struct netmsg_udp_notify {
614 struct netmsg nm_nmsg;
615 void (*nm_notify)(struct inpcb *, int);
616 struct in_addr nm_faddr;
617 int nm_arg;
620 static void
621 udp_notifyall_oncpu(struct netmsg *netmsg)
623 struct netmsg_udp_notify *nmsg = (struct netmsg_udp_notify *)netmsg;
624 int nextcpu;
626 in_pcbnotifyall(&udbinfo.pcblisthead, nmsg->nm_faddr, nmsg->nm_arg,
627 nmsg->nm_notify);
629 nextcpu = mycpuid + 1;
630 if (nextcpu < ncpus2)
631 lwkt_forwardmsg(udp_cport(nextcpu), &netmsg->nm_lmsg);
632 else
633 lwkt_replymsg(&netmsg->nm_lmsg, 0);
636 static void
637 udp_rtchange(struct inpcb *inp, int err)
639 #ifdef SMP
640 /* XXX Nuke this, once UDP inpcbs are CPU localized */
641 if (inp->inp_route.ro_rt && inp->inp_route.ro_rt->rt_cpuid == mycpuid) {
642 rtfree(inp->inp_route.ro_rt);
643 inp->inp_route.ro_rt = NULL;
645 * A new route can be allocated the next time
646 * output is attempted.
649 #else
650 in_rtchange(inp, err);
651 #endif
654 void
655 udp_ctlinput(int cmd, struct sockaddr *sa, void *vip)
657 struct ip *ip = vip;
658 struct udphdr *uh;
659 void (*notify) (struct inpcb *, int) = udp_notify;
660 struct in_addr faddr;
661 struct inpcb *inp;
663 faddr = ((struct sockaddr_in *)sa)->sin_addr;
664 if (sa->sa_family != AF_INET || faddr.s_addr == INADDR_ANY)
665 return;
667 if (PRC_IS_REDIRECT(cmd)) {
668 ip = NULL;
669 notify = udp_rtchange;
670 } else if (cmd == PRC_HOSTDEAD)
671 ip = NULL;
672 else if ((unsigned)cmd >= PRC_NCMDS || inetctlerrmap[cmd] == 0)
673 return;
674 if (ip) {
675 crit_enter();
676 uh = (struct udphdr *)((caddr_t)ip + (ip->ip_hl << 2));
677 inp = in_pcblookup_hash(&udbinfo, faddr, uh->uh_dport,
678 ip->ip_src, uh->uh_sport, 0, NULL);
679 if (inp != NULL && inp->inp_socket != NULL)
680 (*notify)(inp, inetctlerrmap[cmd]);
681 crit_exit();
682 } else if (PRC_IS_REDIRECT(cmd)) {
683 struct netmsg_udp_notify nmsg;
685 KKASSERT(&curthread->td_msgport == cpu_portfn(0));
686 netmsg_init(&nmsg.nm_nmsg, &curthread->td_msgport, 0,
687 udp_notifyall_oncpu);
688 nmsg.nm_faddr = faddr;
689 nmsg.nm_arg = inetctlerrmap[cmd];
690 nmsg.nm_notify = notify;
692 lwkt_domsg(udp_cport(0), &nmsg.nm_nmsg.nm_lmsg, 0);
693 } else {
695 * XXX We should forward msg upon PRC_HOSTHEAD and ip == NULL,
696 * once UDP inpcbs are CPU localized
698 KKASSERT(&curthread->td_msgport == udp_cport(0));
699 in_pcbnotifyall(&udbinfo.pcblisthead, faddr, inetctlerrmap[cmd],
700 notify);
704 SYSCTL_PROC(_net_inet_udp, UDPCTL_PCBLIST, pcblist, CTLFLAG_RD, &udbinfo, 0,
705 in_pcblist_global, "S,xinpcb", "List of active UDP sockets");
707 static int
708 udp_getcred(SYSCTL_HANDLER_ARGS)
710 struct sockaddr_in addrs[2];
711 struct inpcb *inp;
712 int error;
714 error = priv_check(req->td, PRIV_ROOT);
715 if (error)
716 return (error);
717 error = SYSCTL_IN(req, addrs, sizeof addrs);
718 if (error)
719 return (error);
720 crit_enter();
721 inp = in_pcblookup_hash(&udbinfo, addrs[1].sin_addr, addrs[1].sin_port,
722 addrs[0].sin_addr, addrs[0].sin_port, 1, NULL);
723 if (inp == NULL || inp->inp_socket == NULL) {
724 error = ENOENT;
725 goto out;
727 error = SYSCTL_OUT(req, inp->inp_socket->so_cred, sizeof(struct ucred));
728 out:
729 crit_exit();
730 return (error);
733 SYSCTL_PROC(_net_inet_udp, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW,
734 0, 0, udp_getcred, "S,ucred", "Get the ucred of a UDP connection");
736 static int
737 udp_output(struct inpcb *inp, struct mbuf *m, struct sockaddr *dstaddr,
738 struct mbuf *control, struct thread *td)
740 struct udpiphdr *ui;
741 int len = m->m_pkthdr.len;
742 struct sockaddr_in *sin; /* really is initialized before use */
743 int error = 0, lport_any = 0;
745 if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) {
746 error = EMSGSIZE;
747 goto release;
750 if (inp->inp_lport == 0) { /* unbound socket */
751 error = in_pcbbind(inp, NULL, td);
752 if (error)
753 goto release;
754 in_pcbinswildcardhash(inp);
755 lport_any = 1;
758 if (dstaddr != NULL) { /* destination address specified */
759 if (inp->inp_faddr.s_addr != INADDR_ANY) {
760 /* already connected */
761 error = EISCONN;
762 goto release;
764 sin = (struct sockaddr_in *)dstaddr;
765 if (!prison_remote_ip(td, (struct sockaddr *)&sin)) {
766 error = EAFNOSUPPORT; /* IPv6 only jail */
767 goto release;
769 } else {
770 if (inp->inp_faddr.s_addr == INADDR_ANY) {
771 /* no destination specified and not already connected */
772 error = ENOTCONN;
773 goto release;
775 sin = NULL;
779 * Calculate data length and get a mbuf
780 * for UDP and IP headers.
782 M_PREPEND(m, sizeof(struct udpiphdr), MB_DONTWAIT);
783 if (m == NULL) {
784 error = ENOBUFS;
785 goto release;
789 * Fill in mbuf with extended UDP header
790 * and addresses and length put into network format.
792 ui = mtod(m, struct udpiphdr *);
793 bzero(ui->ui_x1, sizeof ui->ui_x1); /* XXX still needed? */
794 ui->ui_pr = IPPROTO_UDP;
797 * Set destination address.
799 if (dstaddr != NULL) { /* use specified destination */
800 ui->ui_dst = sin->sin_addr;
801 ui->ui_dport = sin->sin_port;
802 } else { /* use connected destination */
803 ui->ui_dst = inp->inp_faddr;
804 ui->ui_dport = inp->inp_fport;
808 * Set source address.
810 if (inp->inp_laddr.s_addr == INADDR_ANY) {
811 struct sockaddr_in *if_sin;
813 if (dstaddr == NULL) {
815 * connect() had (or should have) failed because
816 * the interface had no IP address, but the
817 * application proceeded to call send() anyways.
819 error = ENOTCONN;
820 goto release;
823 /* Look up outgoing interface. */
824 if ((error = in_pcbladdr(inp, dstaddr, &if_sin, td)))
825 goto release;
826 ui->ui_src = if_sin->sin_addr; /* use address of interface */
827 } else {
828 ui->ui_src = inp->inp_laddr; /* use non-null bound address */
830 ui->ui_sport = inp->inp_lport;
831 KASSERT(inp->inp_lport != 0, ("inp lport should have been bound"));
833 ui->ui_ulen = htons((u_short)len + sizeof(struct udphdr));
836 * Set up checksum and output datagram.
838 if (udpcksum) {
839 ui->ui_sum = in_pseudo(ui->ui_src.s_addr, ui->ui_dst.s_addr,
840 htons((u_short)len + sizeof(struct udphdr) + IPPROTO_UDP));
841 m->m_pkthdr.csum_flags = CSUM_UDP;
842 m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
843 } else {
844 ui->ui_sum = 0;
846 ((struct ip *)ui)->ip_len = sizeof(struct udpiphdr) + len;
847 ((struct ip *)ui)->ip_ttl = inp->inp_ip_ttl; /* XXX */
848 ((struct ip *)ui)->ip_tos = inp->inp_ip_tos; /* XXX */
849 udpstat.udps_opackets++;
851 error = ip_output(m, inp->inp_options, &inp->inp_route,
852 (inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST)) |
853 IP_DEBUGROUTE,
854 inp->inp_moptions, inp);
857 * If this is the first data gram sent on an unbound and unconnected
858 * UDP socket, lport will be changed in this function. If target
859 * CPU after this lport changing is no longer the current CPU, then
860 * free the route entry allocated on the current CPU.
862 if (lport_any) {
863 if (udp_addrcpu(inp->inp_faddr.s_addr, inp->inp_fport,
864 inp->inp_laddr.s_addr, inp->inp_lport) != mycpuid) {
865 struct route *ro = &inp->inp_route;
867 if (ro->ro_rt != NULL)
868 RTFREE(ro->ro_rt);
869 bzero(ro, sizeof(*ro));
872 return (error);
874 release:
875 m_freem(m);
876 return (error);
879 u_long udp_sendspace = 9216; /* really max datagram size */
880 /* 40 1K datagrams */
881 SYSCTL_INT(_net_inet_udp, UDPCTL_MAXDGRAM, maxdgram, CTLFLAG_RW,
882 &udp_sendspace, 0, "Maximum outgoing UDP datagram size");
884 u_long udp_recvspace = 40 * (1024 +
885 #ifdef INET6
886 sizeof(struct sockaddr_in6)
887 #else
888 sizeof(struct sockaddr_in)
889 #endif
891 SYSCTL_INT(_net_inet_udp, UDPCTL_RECVSPACE, recvspace, CTLFLAG_RW,
892 &udp_recvspace, 0, "Maximum incoming UDP datagram size");
894 static int
895 udp_abort(struct socket *so)
897 struct inpcb *inp;
899 inp = so->so_pcb;
900 if (inp == NULL)
901 return EINVAL; /* ??? possible? panic instead? */
902 soisdisconnected(so);
903 crit_enter();
904 in_pcbdetach(inp);
905 crit_exit();
906 return 0;
909 static int
910 udp_attach(struct socket *so, int proto, struct pru_attach_info *ai)
912 struct inpcb *inp;
913 int error;
915 inp = so->so_pcb;
916 if (inp != NULL)
917 return EINVAL;
919 error = soreserve(so, udp_sendspace, udp_recvspace, ai->sb_rlimit);
920 if (error)
921 return error;
922 crit_enter();
923 error = in_pcballoc(so, &udbinfo);
924 crit_exit();
925 if (error)
926 return error;
928 inp = (struct inpcb *)so->so_pcb;
929 inp->inp_vflag |= INP_IPV4;
930 inp->inp_ip_ttl = ip_defttl;
931 return 0;
934 static int
935 udp_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
937 struct sockaddr_in *sin = (struct sockaddr_in *)nam;
938 struct inpcb *inp;
939 int error;
941 inp = so->so_pcb;
942 if (inp == NULL)
943 return EINVAL;
944 crit_enter();
945 error = in_pcbbind(inp, nam, td);
946 crit_exit();
947 if (error == 0) {
948 if (sin->sin_addr.s_addr != INADDR_ANY)
949 inp->inp_flags |= INP_WASBOUND_NOTANY;
950 in_pcbinswildcardhash(inp);
952 return error;
955 static int
956 udp_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
958 struct inpcb *inp;
959 int error;
961 inp = so->so_pcb;
962 if (inp == NULL)
963 return EINVAL;
964 if (inp->inp_faddr.s_addr != INADDR_ANY)
965 return EISCONN;
966 error = 0;
967 crit_enter();
968 if (td->td_proc && td->td_proc->p_ucred->cr_prison != NULL &&
969 inp->inp_laddr.s_addr == INADDR_ANY) {
970 error = in_pcbbind(inp, NULL, td);
972 if (error == 0) {
973 if (!prison_remote_ip(td, nam))
974 return(EAFNOSUPPORT); /* IPv6 only jail */
975 if (inp->inp_flags & INP_WILDCARD)
976 in_pcbremwildcardhash(inp);
977 error = in_pcbconnect(inp, nam, td);
979 crit_exit();
980 if (error == 0) {
981 soisconnected(so);
984 * Make sure that the new target CPU is same as current CPU,
985 * if it is not, then we will have to free the route entry
986 * allocated on the current CPU.
988 if (udp_addrcpu(inp->inp_faddr.s_addr, inp->inp_fport,
989 inp->inp_laddr.s_addr, inp->inp_lport) != mycpuid) {
990 struct route *ro = &inp->inp_route;
992 if (ro->ro_rt != NULL)
993 RTFREE(ro->ro_rt);
994 bzero(ro, sizeof(*ro));
996 } else if (error == EAFNOSUPPORT) { /* connection dissolved */
998 * Follow traditional BSD behavior and retain
999 * the local port binding. But, fix the old misbehavior
1000 * of overwriting any previously bound local address.
1002 if (!(inp->inp_flags & INP_WASBOUND_NOTANY))
1003 inp->inp_laddr.s_addr = INADDR_ANY;
1004 in_pcbinswildcardhash(inp);
1006 return error;
1009 static int
1010 udp_detach(struct socket *so)
1012 struct inpcb *inp;
1014 inp = so->so_pcb;
1015 if (inp == NULL)
1016 return EINVAL;
1017 crit_enter();
1018 in_pcbdetach(inp);
1019 crit_exit();
1020 return 0;
1023 static int
1024 udp_disconnect(struct socket *so)
1026 struct route *ro;
1027 struct inpcb *inp;
1029 inp = so->so_pcb;
1030 if (inp == NULL)
1031 return EINVAL;
1032 if (inp->inp_faddr.s_addr == INADDR_ANY)
1033 return ENOTCONN;
1035 crit_enter();
1036 in_pcbdisconnect(inp);
1037 crit_exit();
1038 so->so_state &= ~SS_ISCONNECTED; /* XXX */
1040 ro = &inp->inp_route;
1041 if (ro->ro_rt != NULL)
1042 RTFREE(ro->ro_rt);
1043 bzero(ro, sizeof(*ro));
1045 return 0;
1048 static int
1049 udp_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *addr,
1050 struct mbuf *control, struct thread *td)
1052 struct inpcb *inp;
1054 inp = so->so_pcb;
1055 if (inp == NULL) {
1056 m_freem(m);
1057 return EINVAL;
1059 return udp_output(inp, m, addr, control, td);
1063 udp_shutdown(struct socket *so)
1065 struct inpcb *inp;
1067 inp = so->so_pcb;
1068 if (inp == NULL)
1069 return EINVAL;
1070 socantsendmore(so);
1071 return 0;
1074 struct pr_usrreqs udp_usrreqs = {
1075 .pru_abort = udp_abort,
1076 .pru_accept = pru_accept_notsupp,
1077 .pru_attach = udp_attach,
1078 .pru_bind = udp_bind,
1079 .pru_connect = udp_connect,
1080 .pru_connect2 = pru_connect2_notsupp,
1081 .pru_control = in_control,
1082 .pru_detach = udp_detach,
1083 .pru_disconnect = udp_disconnect,
1084 .pru_listen = pru_listen_notsupp,
1085 .pru_peeraddr = in_setpeeraddr,
1086 .pru_rcvd = pru_rcvd_notsupp,
1087 .pru_rcvoob = pru_rcvoob_notsupp,
1088 .pru_send = udp_send,
1089 .pru_sense = pru_sense_null,
1090 .pru_shutdown = udp_shutdown,
1091 .pru_sockaddr = in_setsockaddr,
1092 .pru_sosend = sosendudp,
1093 .pru_soreceive = soreceive,
1094 .pru_sopoll = sopoll