1 /*-
2 * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995
3 * The Regents of the University of California. All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * 4. Neither the name of the University nor the names of its contributors
14 * may be used to endorse or promote products derived from this software
15 * without specific prior written permission.
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27 * SUCH DAMAGE.
28 *
29 * @(#)udp_usrreq.c 8.6 (Berkeley) 5/23/95
30 * $FreeBSD: releng/6.1/sys/netinet/udp_usrreq.c 155694 2006-02-14 21:40:21Z rwatson $
31 */
32
33 #include "opt_ipfw.h"
34 #include "opt_ipsec.h"
35 #include "opt_inet6.h"
36 #include "opt_mac.h"
37
38 #include <sys/param.h>
39 #include <sys/systm.h>
40 #include <sys/domain.h>
41 #include <sys/jail.h>
42 #include <sys/kernel.h>
43 #include <sys/lock.h>
44 #include <sys/mac.h>
45 #include <sys/malloc.h>
46 #include <sys/mbuf.h>
47 #include <sys/proc.h>
48 #include <sys/protosw.h>
49 #include <sys/signalvar.h>
50 #include <sys/socket.h>
51 #include <sys/socketvar.h>
52 #include <sys/sx.h>
53 #include <sys/sysctl.h>
54 #include <sys/syslog.h>
55
56 #include <vm/uma.h>
57
58 #include <net/if.h>
59 #include <net/route.h>
60
61 #include <netinet/in.h>
62 #include <netinet/in_systm.h>
63 #include <netinet/in_pcb.h>
64 #include <netinet/in_var.h>
65 #include <netinet/ip.h>
66 #ifdef INET6
67 #include <netinet/ip6.h>
68 #endif
69 #include <netinet/ip_icmp.h>
70 #include <netinet/icmp_var.h>
71 #include <netinet/ip_var.h>
72 #ifdef INET6
73 #include <netinet6/ip6_var.h>
74 #endif
75 #include <netinet/udp.h>
76 #include <netinet/udp_var.h>
77
78 #ifdef FAST_IPSEC
79 #include <netipsec/ipsec.h>
80 #endif /*FAST_IPSEC*/
81
82 #ifdef IPSEC
83 #include <netinet6/ipsec.h>
84 #endif /*IPSEC*/
85
86 #include <machine/in_cksum.h>
87
88 /*
89 * UDP protocol implementation.
90 * Per RFC 768, August, 1980.
91 */
92 #ifndef COMPAT_42
93 static int udpcksum = 1;
94 #else
95 static int udpcksum = 0; /* XXX */
96 #endif
97 SYSCTL_INT(_net_inet_udp, UDPCTL_CHECKSUM, checksum, CTLFLAG_RW,
98 &udpcksum, 0, "");
99
100 int log_in_vain = 0;
101 SYSCTL_INT(_net_inet_udp, OID_AUTO, log_in_vain, CTLFLAG_RW,
102 &log_in_vain, 0, "Log all incoming UDP packets");
103
104 static int blackhole = 0;
105 SYSCTL_INT(_net_inet_udp, OID_AUTO, blackhole, CTLFLAG_RW,
106 &blackhole, 0, "Do not send port unreachables for refused connects");
107
108 static int strict_mcast_mship = 0;
109 SYSCTL_INT(_net_inet_udp, OID_AUTO, strict_mcast_mship, CTLFLAG_RW,
110 &strict_mcast_mship, 0, "Only send multicast to member sockets");
111
112 struct inpcbhead udb; /* from udp_var.h */
113 #define udb6 udb /* for KAME src sync over BSD*'s */
114 struct inpcbinfo udbinfo;
115
116 #ifndef UDBHASHSIZE
117 #define UDBHASHSIZE 16
118 #endif
119
120 struct udpstat udpstat; /* from udp_var.h */
121 SYSCTL_STRUCT(_net_inet_udp, UDPCTL_STATS, stats, CTLFLAG_RW,
122 &udpstat, udpstat, "UDP statistics (struct udpstat, netinet/udp_var.h)");
123
124 static void udp_append(struct inpcb *last, struct ip *ip, struct mbuf *n,
125 int off, struct sockaddr_in *udp_in);
126
127 static int udp_detach(struct socket *so);
128 static int udp_output(struct inpcb *, struct mbuf *, struct sockaddr *,
129 struct mbuf *, struct thread *);
130
131 void
132 udp_init()
133 {
134 INP_INFO_LOCK_INIT(&udbinfo, "udp");
135 LIST_INIT(&udb);
136 udbinfo.listhead = &udb;
137 udbinfo.hashbase = hashinit(UDBHASHSIZE, M_PCB, &udbinfo.hashmask);
138 udbinfo.porthashbase = hashinit(UDBHASHSIZE, M_PCB,
139 &udbinfo.porthashmask);
140 udbinfo.ipi_zone = uma_zcreate("udpcb", sizeof(struct inpcb), NULL,
141 NULL, NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
142 uma_zone_set_max(udbinfo.ipi_zone, maxsockets);
143 }
144
145 void
146 udp_input(m, off)
147 register struct mbuf *m;
148 int off;
149 {
150 int iphlen = off;
151 register struct ip *ip;
152 register struct udphdr *uh;
153 register struct inpcb *inp;
154 int len;
155 struct ip save_ip;
156 struct sockaddr_in udp_in;
157 #ifdef IPFIREWALL_FORWARD
158 struct m_tag *fwd_tag;
159 #endif
160
161 udpstat.udps_ipackets++;
162
163 /*
164 * Strip IP options, if any; should skip this,
165 * make available to user, and use on returned packets,
166 * but we don't yet have a way to check the checksum
167 * with options still present.
168 */
169 if (iphlen > sizeof (struct ip)) {
170 ip_stripoptions(m, (struct mbuf *)0);
171 iphlen = sizeof(struct ip);
172 }
173
174 /*
175 * Get IP and UDP header together in first mbuf.
176 */
177 ip = mtod(m, struct ip *);
178 if (m->m_len < iphlen + sizeof(struct udphdr)) {
179 if ((m = m_pullup(m, iphlen + sizeof(struct udphdr))) == 0) {
180 udpstat.udps_hdrops++;
181 return;
182 }
183 ip = mtod(m, struct ip *);
184 }
185 uh = (struct udphdr *)((caddr_t)ip + iphlen);
186
187 /* destination port of 0 is illegal, based on RFC768. */
188 if (uh->uh_dport == 0)
189 goto badunlocked;
190
191 /*
192 * Construct sockaddr format source address.
193 * Stuff source address and datagram in user buffer.
194 */
195 bzero(&udp_in, sizeof(udp_in));
196 udp_in.sin_len = sizeof(udp_in);
197 udp_in.sin_family = AF_INET;
198 udp_in.sin_port = uh->uh_sport;
199 udp_in.sin_addr = ip->ip_src;
200
201 /*
202 * Make mbuf data length reflect UDP length.
203 * If not enough data to reflect UDP length, drop.
204 */
205 len = ntohs((u_short)uh->uh_ulen);
206 if (ip->ip_len != len) {
207 if (len > ip->ip_len || len < sizeof(struct udphdr)) {
208 udpstat.udps_badlen++;
209 goto badunlocked;
210 }
211 m_adj(m, len - ip->ip_len);
212 /* ip->ip_len = len; */
213 }
214 /*
215 * Save a copy of the IP header in case we want restore it
216 * for sending an ICMP error message in response.
217 */
218 if (!blackhole)
219 save_ip = *ip;
220
221 /*
222 * Checksum extended UDP header and data.
223 */
224 if (uh->uh_sum) {
225 if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) {
226 if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR)
227 uh->uh_sum = m->m_pkthdr.csum_data;
228 else
229 uh->uh_sum = in_pseudo(ip->ip_src.s_addr,
230 ip->ip_dst.s_addr, htonl((u_short)len +
231 m->m_pkthdr.csum_data + IPPROTO_UDP));
232 uh->uh_sum ^= 0xffff;
233 } else {
234 char b[9];
235 bcopy(((struct ipovly *)ip)->ih_x1, b, 9);
236 bzero(((struct ipovly *)ip)->ih_x1, 9);
237 ((struct ipovly *)ip)->ih_len = uh->uh_ulen;
238 uh->uh_sum = in_cksum(m, len + sizeof (struct ip));
239 bcopy(b, ((struct ipovly *)ip)->ih_x1, 9);
240 }
241 if (uh->uh_sum) {
242 udpstat.udps_badsum++;
243 m_freem(m);
244 return;
245 }
246 } else
247 udpstat.udps_nosum++;
248
249 #ifdef IPFIREWALL_FORWARD
250 /* Grab info from PACKET_TAG_IPFORWARD tag prepended to the chain. */
251 fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL);
252
253 if (fwd_tag != NULL) {
254 struct sockaddr_in *next_hop;
255
256 /* Do the hack. */
257 next_hop = (struct sockaddr_in *)(fwd_tag + 1);
258 ip->ip_dst = next_hop->sin_addr;
259 uh->uh_dport = ntohs(next_hop->sin_port);
260 /* Remove the tag from the packet. We don't need it anymore. */
261 m_tag_delete(m, fwd_tag);
262 }
263 #endif
264
265 INP_INFO_RLOCK(&udbinfo);
266
267 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) ||
268 in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif)) {
269 struct inpcb *last;
270 /*
271 * Deliver a multicast or broadcast datagram to *all* sockets
272 * for which the local and remote addresses and ports match
273 * those of the incoming datagram. This allows more than
274 * one process to receive multi/broadcasts on the same port.
275 * (This really ought to be done for unicast datagrams as
276 * well, but that would cause problems with existing
277 * applications that open both address-specific sockets and
278 * a wildcard socket listening to the same port -- they would
279 * end up receiving duplicates of every unicast datagram.
280 * Those applications open the multiple sockets to overcome an
281 * inadequacy of the UDP socket interface, but for backwards
282 * compatibility we avoid the problem here rather than
283 * fixing the interface. Maybe 4.5BSD will remedy this?)
284 */
285
286 /*
287 * Locate pcb(s) for datagram.
288 * (Algorithm copied from raw_intr().)
289 */
290 last = NULL;
291 LIST_FOREACH(inp, &udb, inp_list) {
292 if (inp->inp_lport != uh->uh_dport)
293 continue;
294 #ifdef INET6
295 if ((inp->inp_vflag & INP_IPV4) == 0)
296 continue;
297 #endif
298 if (inp->inp_laddr.s_addr != INADDR_ANY) {
299 if (inp->inp_laddr.s_addr != ip->ip_dst.s_addr)
300 continue;
301 }
302 if (inp->inp_faddr.s_addr != INADDR_ANY) {
303 if (inp->inp_faddr.s_addr !=
304 ip->ip_src.s_addr ||
305 inp->inp_fport != uh->uh_sport)
306 continue;
307 }
308 INP_LOCK(inp);
309
310 /*
311 * Check multicast packets to make sure they are only
312 * sent to sockets with multicast memberships for the
313 * packet's destination address and arrival interface
314 */
315 #define MSHIP(_inp, n) ((_inp)->inp_moptions->imo_membership[(n)])
316 #define NMSHIPS(_inp) ((_inp)->inp_moptions->imo_num_memberships)
317 if (strict_mcast_mship && inp->inp_moptions != NULL) {
318 int mship, foundmship = 0;
319
320 for (mship = 0; mship < NMSHIPS(inp); mship++) {
321 if (MSHIP(inp, mship)->inm_addr.s_addr
322 == ip->ip_dst.s_addr &&
323 MSHIP(inp, mship)->inm_ifp
324 == m->m_pkthdr.rcvif) {
325 foundmship = 1;
326 break;
327 }
328 }
329 if (foundmship == 0) {
330 INP_UNLOCK(inp);
331 continue;
332 }
333 }
334 #undef NMSHIPS
335 #undef MSHIP
336 if (last != NULL) {
337 struct mbuf *n;
338
339 n = m_copy(m, 0, M_COPYALL);
340 if (n != NULL)
341 udp_append(last, ip, n,
342 iphlen +
343 sizeof(struct udphdr),
344 &udp_in);
345 INP_UNLOCK(last);
346 }
347 last = inp;
348 /*
349 * Don't look for additional matches if this one does
350 * not have either the SO_REUSEPORT or SO_REUSEADDR
351 * socket options set. This heuristic avoids searching
352 * through all pcbs in the common case of a non-shared
353 * port. It * assumes that an application will never
354 * clear these options after setting them.
355 */
356 if ((last->inp_socket->so_options&(SO_REUSEPORT|SO_REUSEADDR)) == 0)
357 break;
358 }
359
360 if (last == NULL) {
361 /*
362 * No matching pcb found; discard datagram.
363 * (No need to send an ICMP Port Unreachable
364 * for a broadcast or multicast datgram.)
365 */
366 udpstat.udps_noportbcast++;
367 goto badheadlocked;
368 }
369 udp_append(last, ip, m, iphlen + sizeof(struct udphdr),
370 &udp_in);
371 INP_UNLOCK(last);
372 INP_INFO_RUNLOCK(&udbinfo);
373 return;
374 }
375 /*
376 * Locate pcb for datagram.
377 */
378 inp = in_pcblookup_hash(&udbinfo, ip->ip_src, uh->uh_sport,
379 ip->ip_dst, uh->uh_dport, 1, m->m_pkthdr.rcvif);
380 if (inp == NULL) {
381 if (log_in_vain) {
382 char buf[4*sizeof "123"];
383
384 strcpy(buf, inet_ntoa(ip->ip_dst));
385 log(LOG_INFO,
386 "Connection attempt to UDP %s:%d from %s:%d\n",
387 buf, ntohs(uh->uh_dport), inet_ntoa(ip->ip_src),
388 ntohs(uh->uh_sport));
389 }
390 udpstat.udps_noport++;
391 if (m->m_flags & (M_BCAST | M_MCAST)) {
392 udpstat.udps_noportbcast++;
393 goto badheadlocked;
394 }
395 if (blackhole)
396 goto badheadlocked;
397 if (badport_bandlim(BANDLIM_ICMP_UNREACH) < 0)
398 goto badheadlocked;
399 *ip = save_ip;
400 ip->ip_len += iphlen;
401 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0);
402 INP_INFO_RUNLOCK(&udbinfo);
403 return;
404 }
405 INP_LOCK(inp);
406 /* Check the minimum TTL for socket. */
407 if (inp->inp_ip_minttl && inp->inp_ip_minttl > ip->ip_ttl)
408 goto badheadlocked;
409 udp_append(inp, ip, m, iphlen + sizeof(struct udphdr), &udp_in);
410 INP_UNLOCK(inp);
411 INP_INFO_RUNLOCK(&udbinfo);
412 return;
413
414 badheadlocked:
415 if (inp)
416 INP_UNLOCK(inp);
417 INP_INFO_RUNLOCK(&udbinfo);
418 badunlocked:
419 m_freem(m);
420 return;
421 }
422
423 /*
424 * Subroutine of udp_input(), which appends the provided mbuf chain to the
425 * passed pcb/socket. The caller must provide a sockaddr_in via udp_in that
426 * contains the source address. If the socket ends up being an IPv6 socket,
427 * udp_append() will convert to a sockaddr_in6 before passing the address
428 * into the socket code.
429 */
430 static void
431 udp_append(last, ip, n, off, udp_in)
432 struct inpcb *last;
433 struct ip *ip;
434 struct mbuf *n;
435 int off;
436 struct sockaddr_in *udp_in;
437 {
438 struct sockaddr *append_sa;
439 struct socket *so;
440 struct mbuf *opts = 0;
441 #ifdef INET6
442 struct sockaddr_in6 udp_in6;
443 #endif
444
445 INP_LOCK_ASSERT(last);
446
447 #if defined(IPSEC) || defined(FAST_IPSEC)
448 /* check AH/ESP integrity. */
449 if (ipsec4_in_reject(n, last)) {
450 #ifdef IPSEC
451 ipsecstat.in_polvio++;
452 #endif /*IPSEC*/
453 m_freem(n);
454 return;
455 }
456 #endif /*IPSEC || FAST_IPSEC*/
457 #ifdef MAC
458 if (mac_check_inpcb_deliver(last, n) != 0) {
459 m_freem(n);
460 return;
461 }
462 #endif
463 if (last->inp_flags & INP_CONTROLOPTS ||
464 last->inp_socket->so_options & (SO_TIMESTAMP | SO_BINTIME)) {
465 #ifdef INET6
466 if (last->inp_vflag & INP_IPV6) {
467 int savedflags;
468
469 savedflags = last->inp_flags;
470 last->inp_flags &= ~INP_UNMAPPABLEOPTS;
471 ip6_savecontrol(last, n, &opts);
472 last->inp_flags = savedflags;
473 } else
474 #endif
475 ip_savecontrol(last, &opts, ip, n);
476 }
477 #ifdef INET6
478 if (last->inp_vflag & INP_IPV6) {
479 bzero(&udp_in6, sizeof(udp_in6));
480 udp_in6.sin6_len = sizeof(udp_in6);
481 udp_in6.sin6_family = AF_INET6;
482 in6_sin_2_v4mapsin6(udp_in, &udp_in6);
483 append_sa = (struct sockaddr *)&udp_in6;
484 } else
485 #endif
486 append_sa = (struct sockaddr *)udp_in;
487 m_adj(n, off);
488
489 so = last->inp_socket;
490 SOCKBUF_LOCK(&so->so_rcv);
491 if (sbappendaddr_locked(&so->so_rcv, append_sa, n, opts) == 0) {
492 m_freem(n);
493 if (opts)
494 m_freem(opts);
495 udpstat.udps_fullsock++;
496 SOCKBUF_UNLOCK(&so->so_rcv);
497 } else
498 sorwakeup_locked(so);
499 }
500
501 /*
502 * Notify a udp user of an asynchronous error;
503 * just wake up so that he can collect error status.
504 */
505 struct inpcb *
506 udp_notify(inp, errno)
507 register struct inpcb *inp;
508 int errno;
509 {
510 inp->inp_socket->so_error = errno;
511 sorwakeup(inp->inp_socket);
512 sowwakeup(inp->inp_socket);
513 return inp;
514 }
515
516 void
517 udp_ctlinput(cmd, sa, vip)
518 int cmd;
519 struct sockaddr *sa;
520 void *vip;
521 {
522 struct ip *ip = vip;
523 struct udphdr *uh;
524 struct inpcb *(*notify)(struct inpcb *, int) = udp_notify;
525 struct in_addr faddr;
526 struct inpcb *inp;
527
528 faddr = ((struct sockaddr_in *)sa)->sin_addr;
529 if (sa->sa_family != AF_INET || faddr.s_addr == INADDR_ANY)
530 return;
531
532 /*
533 * Redirects don't need to be handled up here.
534 */
535 if (PRC_IS_REDIRECT(cmd))
536 return;
537 /*
538 * Hostdead is ugly because it goes linearly through all PCBs.
539 * XXX: We never get this from ICMP, otherwise it makes an
540 * excellent DoS attack on machines with many connections.
541 */
542 if (cmd == PRC_HOSTDEAD)
543 ip = 0;
544 else if ((unsigned)cmd >= PRC_NCMDS || inetctlerrmap[cmd] == 0)
545 return;
546 if (ip) {
547 uh = (struct udphdr *)((caddr_t)ip + (ip->ip_hl << 2));
548 INP_INFO_RLOCK(&udbinfo);
549 inp = in_pcblookup_hash(&udbinfo, faddr, uh->uh_dport,
550 ip->ip_src, uh->uh_sport, 0, NULL);
551 if (inp != NULL) {
552 INP_LOCK(inp);
553 if (inp->inp_socket != NULL) {
554 (*notify)(inp, inetctlerrmap[cmd]);
555 }
556 INP_UNLOCK(inp);
557 }
558 INP_INFO_RUNLOCK(&udbinfo);
559 } else
560 in_pcbnotifyall(&udbinfo, faddr, inetctlerrmap[cmd], notify);
561 }
562
563 static int
564 udp_pcblist(SYSCTL_HANDLER_ARGS)
565 {
566 int error, i, n;
567 struct inpcb *inp, **inp_list;
568 inp_gen_t gencnt;
569 struct xinpgen xig;
570
571 /*
572 * The process of preparing the TCB list is too time-consuming and
573 * resource-intensive to repeat twice on every request.
574 */
575 if (req->oldptr == 0) {
576 n = udbinfo.ipi_count;
577 req->oldidx = 2 * (sizeof xig)
578 + (n + n/8) * sizeof(struct xinpcb);
579 return 0;
580 }
581
582 if (req->newptr != 0)
583 return EPERM;
584
585 /*
586 * OK, now we're committed to doing something.
587 */
588 INP_INFO_RLOCK(&udbinfo);
589 gencnt = udbinfo.ipi_gencnt;
590 n = udbinfo.ipi_count;
591 INP_INFO_RUNLOCK(&udbinfo);
592
593 error = sysctl_wire_old_buffer(req, 2 * (sizeof xig)
594 + n * sizeof(struct xinpcb));
595 if (error != 0)
596 return (error);
597
598 xig.xig_len = sizeof xig;
599 xig.xig_count = n;
600 xig.xig_gen = gencnt;
601 xig.xig_sogen = so_gencnt;
602 error = SYSCTL_OUT(req, &xig, sizeof xig);
603 if (error)
604 return error;
605
606 inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK);
607 if (inp_list == 0)
608 return ENOMEM;
609
610 INP_INFO_RLOCK(&udbinfo);
611 for (inp = LIST_FIRST(udbinfo.listhead), i = 0; inp && i < n;
612 inp = LIST_NEXT(inp, inp_list)) {
613 INP_LOCK(inp);
614 if (inp->inp_gencnt <= gencnt &&
615 cr_canseesocket(req->td->td_ucred, inp->inp_socket) == 0)
616 inp_list[i++] = inp;
617 INP_UNLOCK(inp);
618 }
619 INP_INFO_RUNLOCK(&udbinfo);
620 n = i;
621
622 error = 0;
623 for (i = 0; i < n; i++) {
624 inp = inp_list[i];
625 if (inp->inp_gencnt <= gencnt) {
626 struct xinpcb xi;
627 bzero(&xi, sizeof(xi));
628 xi.xi_len = sizeof xi;
629 /* XXX should avoid extra copy */
630 bcopy(inp, &xi.xi_inp, sizeof *inp);
631 if (inp->inp_socket)
632 sotoxsocket(inp->inp_socket, &xi.xi_socket);
633 xi.xi_inp.inp_gencnt = inp->inp_gencnt;
634 error = SYSCTL_OUT(req, &xi, sizeof xi);
635 }
636 }
637 if (!error) {
638 /*
639 * Give the user an updated idea of our state.
640 * If the generation differs from what we told
641 * her before, she knows that something happened
642 * while we were processing this request, and it
643 * might be necessary to retry.
644 */
645 INP_INFO_RLOCK(&udbinfo);
646 xig.xig_gen = udbinfo.ipi_gencnt;
647 xig.xig_sogen = so_gencnt;
648 xig.xig_count = udbinfo.ipi_count;
649 INP_INFO_RUNLOCK(&udbinfo);
650 error = SYSCTL_OUT(req, &xig, sizeof xig);
651 }
652 free(inp_list, M_TEMP);
653 return error;
654 }
655
656 SYSCTL_PROC(_net_inet_udp, UDPCTL_PCBLIST, pcblist, CTLFLAG_RD, 0, 0,
657 udp_pcblist, "S,xinpcb", "List of active UDP sockets");
658
659 static int
660 udp_getcred(SYSCTL_HANDLER_ARGS)
661 {
662 struct xucred xuc;
663 struct sockaddr_in addrs[2];
664 struct inpcb *inp;
665 int error;
666
667 error = suser_cred(req->td->td_ucred, SUSER_ALLOWJAIL);
668 if (error)
669 return (error);
670 error = SYSCTL_IN(req, addrs, sizeof(addrs));
671 if (error)
672 return (error);
673 INP_INFO_RLOCK(&udbinfo);
674 inp = in_pcblookup_hash(&udbinfo, addrs[1].sin_addr, addrs[1].sin_port,
675 addrs[0].sin_addr, addrs[0].sin_port, 1, NULL);
676 if (inp == NULL || inp->inp_socket == NULL) {
677 error = ENOENT;
678 goto out;
679 }
680 error = cr_canseesocket(req->td->td_ucred, inp->inp_socket);
681 if (error)
682 goto out;
683 cru2x(inp->inp_socket->so_cred, &xuc);
684 out:
685 INP_INFO_RUNLOCK(&udbinfo);
686 if (error == 0)
687 error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred));
688 return (error);
689 }
690
691 SYSCTL_PROC(_net_inet_udp, OID_AUTO, getcred,
692 CTLTYPE_OPAQUE|CTLFLAG_RW|CTLFLAG_PRISON, 0, 0,
693 udp_getcred, "S,xucred", "Get the xucred of a UDP connection");
694
695 static int
696 udp_output(inp, m, addr, control, td)
697 register struct inpcb *inp;
698 struct mbuf *m;
699 struct sockaddr *addr;
700 struct mbuf *control;
701 struct thread *td;
702 {
703 register struct udpiphdr *ui;
704 register int len = m->m_pkthdr.len;
705 struct in_addr faddr, laddr;
706 struct cmsghdr *cm;
707 struct sockaddr_in *sin, src;
708 int error = 0;
709 int ipflags;
710 u_short fport, lport;
711 int unlock_udbinfo;
712
713 /*
714 * udp_output() may need to temporarily bind or connect the current
715 * inpcb. As such, we don't know up front what inpcb locks we will
716 * need. Do any work to decide what is needed up front before
717 * acquiring locks.
718 */
719 if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) {
720 if (control)
721 m_freem(control);
722 m_freem(m);
723 return EMSGSIZE;
724 }
725
726 src.sin_addr.s_addr = INADDR_ANY;
727 if (control != NULL) {
728 /*
729 * XXX: Currently, we assume all the optional information
730 * is stored in a single mbuf.
731 */
732 if (control->m_next) {
733 m_freem(control);
734 m_freem(m);
735 return EINVAL;
736 }
737 for (; control->m_len > 0;
738 control->m_data += CMSG_ALIGN(cm->cmsg_len),
739 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
740 cm = mtod(control, struct cmsghdr *);
741 if (control->m_len < sizeof(*cm) || cm->cmsg_len == 0 ||
742 cm->cmsg_len > control->m_len) {
743 error = EINVAL;
744 break;
745 }
746 if (cm->cmsg_level != IPPROTO_IP)
747 continue;
748
749 switch (cm->cmsg_type) {
750 case IP_SENDSRCADDR:
751 if (cm->cmsg_len !=
752 CMSG_LEN(sizeof(struct in_addr))) {
753 error = EINVAL;
754 break;
755 }
756 bzero(&src, sizeof(src));
757 src.sin_family = AF_INET;
758 src.sin_len = sizeof(src);
759 src.sin_port = inp->inp_lport;
760 src.sin_addr = *(struct in_addr *)CMSG_DATA(cm);
761 break;
762 default:
763 error = ENOPROTOOPT;
764 break;
765 }
766 if (error)
767 break;
768 }
769 m_freem(control);
770 }
771 if (error) {
772 m_freem(m);
773 return error;
774 }
775
776 if (src.sin_addr.s_addr != INADDR_ANY ||
777 addr != NULL) {
778 INP_INFO_WLOCK(&udbinfo);
779 unlock_udbinfo = 1;
780 } else
781 unlock_udbinfo = 0;
782 INP_LOCK(inp);
783
784 #ifdef MAC
785 mac_create_mbuf_from_inpcb(inp, m);
786 #endif
787
788 laddr = inp->inp_laddr;
789 lport = inp->inp_lport;
790 if (src.sin_addr.s_addr != INADDR_ANY) {
791 if (lport == 0) {
792 error = EINVAL;
793 goto release;
794 }
795 error = in_pcbbind_setup(inp, (struct sockaddr *)&src,
796 &laddr.s_addr, &lport, td->td_ucred);
797 if (error)
798 goto release;
799 }
800
801 if (addr) {
802 sin = (struct sockaddr_in *)addr;
803 if (jailed(td->td_ucred))
804 prison_remote_ip(td->td_ucred, 0, &sin->sin_addr.s_addr);
805 if (inp->inp_faddr.s_addr != INADDR_ANY) {
806 error = EISCONN;
807 goto release;
808 }
809 error = in_pcbconnect_setup(inp, addr, &laddr.s_addr, &lport,
810 &faddr.s_addr, &fport, NULL, td->td_ucred);
811 if (error)
812 goto release;
813
814 /* Commit the local port if newly assigned. */
815 if (inp->inp_laddr.s_addr == INADDR_ANY &&
816 inp->inp_lport == 0) {
817 /*
818 * Remember addr if jailed, to prevent rebinding.
819 */
820 if (jailed(td->td_ucred))
821 inp->inp_laddr = laddr;
822 inp->inp_lport = lport;
823 if (in_pcbinshash(inp) != 0) {
824 inp->inp_lport = 0;
825 error = EAGAIN;
826 goto release;
827 }
828 inp->inp_flags |= INP_ANONPORT;
829 }
830 } else {
831 faddr = inp->inp_faddr;
832 fport = inp->inp_fport;
833 if (faddr.s_addr == INADDR_ANY) {
834 error = ENOTCONN;
835 goto release;
836 }
837 }
838
839 /*
840 * Calculate data length and get a mbuf for UDP, IP, and possible
841 * link-layer headers. Immediate slide the data pointer back forward
842 * since we won't use that space at this layer.
843 */
844 M_PREPEND(m, sizeof(struct udpiphdr) + max_linkhdr, M_DONTWAIT);
845 if (m == NULL) {
846 error = ENOBUFS;
847 goto release;
848 }
849 m->m_data += max_linkhdr;
850 m->m_len -= max_linkhdr;
851 m->m_pkthdr.len -= max_linkhdr;
852
853 /*
854 * Fill in mbuf with extended UDP header
855 * and addresses and length put into network format.
856 */
857 ui = mtod(m, struct udpiphdr *);
858 bzero(ui->ui_x1, sizeof(ui->ui_x1)); /* XXX still needed? */
859 ui->ui_pr = IPPROTO_UDP;
860 ui->ui_src = laddr;
861 ui->ui_dst = faddr;
862 ui->ui_sport = lport;
863 ui->ui_dport = fport;
864 ui->ui_ulen = htons((u_short)len + sizeof(struct udphdr));
865
866 /*
867 * Set the Don't Fragment bit in the IP header.
868 */
869 if (inp->inp_flags & INP_DONTFRAG) {
870 struct ip *ip;
871 ip = (struct ip *)&ui->ui_i;
872 ip->ip_off |= IP_DF;
873 }
874
875 ipflags = 0;
876 if (inp->inp_socket->so_options & SO_DONTROUTE)
877 ipflags |= IP_ROUTETOIF;
878 if (inp->inp_socket->so_options & SO_BROADCAST)
879 ipflags |= IP_ALLOWBROADCAST;
880 if (inp->inp_vflag & INP_ONESBCAST)
881 ipflags |= IP_SENDONES;
882
883 /*
884 * Set up checksum and output datagram.
885 */
886 if (udpcksum) {
887 if (inp->inp_vflag & INP_ONESBCAST)
888 faddr.s_addr = INADDR_BROADCAST;
889 ui->ui_sum = in_pseudo(ui->ui_src.s_addr, faddr.s_addr,
890 htons((u_short)len + sizeof(struct udphdr) + IPPROTO_UDP));
891 m->m_pkthdr.csum_flags = CSUM_UDP;
892 m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
893 } else {
894 ui->ui_sum = 0;
895 }
896 ((struct ip *)ui)->ip_len = sizeof (struct udpiphdr) + len;
897 ((struct ip *)ui)->ip_ttl = inp->inp_ip_ttl; /* XXX */
898 ((struct ip *)ui)->ip_tos = inp->inp_ip_tos; /* XXX */
899 udpstat.udps_opackets++;
900
901 if (unlock_udbinfo)
902 INP_INFO_WUNLOCK(&udbinfo);
903 error = ip_output(m, inp->inp_options, NULL, ipflags,
904 inp->inp_moptions, inp);
905 INP_UNLOCK(inp);
906 return (error);
907
908 release:
909 INP_UNLOCK(inp);
910 if (unlock_udbinfo)
911 INP_INFO_WUNLOCK(&udbinfo);
912 m_freem(m);
913 return (error);
914 }
915
916 u_long udp_sendspace = 9216; /* really max datagram size */
917 /* 40 1K datagrams */
918 SYSCTL_INT(_net_inet_udp, UDPCTL_MAXDGRAM, maxdgram, CTLFLAG_RW,
919 &udp_sendspace, 0, "Maximum outgoing UDP datagram size");
920
921 u_long udp_recvspace = 40 * (1024 +
922 #ifdef INET6
923 sizeof(struct sockaddr_in6)
924 #else
925 sizeof(struct sockaddr_in)
926 #endif
927 );
928 SYSCTL_INT(_net_inet_udp, UDPCTL_RECVSPACE, recvspace, CTLFLAG_RW,
929 &udp_recvspace, 0, "Maximum space for incoming UDP datagrams");
930
931 static int
932 udp_abort(struct socket *so)
933 {
934 struct inpcb *inp;
935
936 INP_INFO_WLOCK(&udbinfo);
937 inp = sotoinpcb(so);
938 if (inp == 0) {
939 INP_INFO_WUNLOCK(&udbinfo);
940 return EINVAL; /* ??? possible? panic instead? */
941 }
942 INP_LOCK(inp);
943 soisdisconnected(so);
944 in_pcbdetach(inp);
945 INP_INFO_WUNLOCK(&udbinfo);
946 return 0;
947 }
948
949 static int
950 udp_attach(struct socket *so, int proto, struct thread *td)
951 {
952 struct inpcb *inp;
953 int error;
954
955 INP_INFO_WLOCK(&udbinfo);
956 inp = sotoinpcb(so);
957 if (inp != 0) {
958 INP_INFO_WUNLOCK(&udbinfo);
959 return EINVAL;
960 }
961 error = soreserve(so, udp_sendspace, udp_recvspace);
962 if (error) {
963 INP_INFO_WUNLOCK(&udbinfo);
964 return error;
965 }
966 error = in_pcballoc(so, &udbinfo, "udpinp");
967 if (error) {
968 INP_INFO_WUNLOCK(&udbinfo);
969 return error;
970 }
971
972 inp = (struct inpcb *)so->so_pcb;
973 INP_LOCK(inp);
974 INP_INFO_WUNLOCK(&udbinfo);
975 inp->inp_vflag |= INP_IPV4;
976 inp->inp_ip_ttl = ip_defttl;
977 INP_UNLOCK(inp);
978 return 0;
979 }
980
981 static int
982 udp_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
983 {
984 struct inpcb *inp;
985 int error;
986
987 INP_INFO_WLOCK(&udbinfo);
988 inp = sotoinpcb(so);
989 if (inp == 0) {
990 INP_INFO_WUNLOCK(&udbinfo);
991 return EINVAL;
992 }
993 INP_LOCK(inp);
994 error = in_pcbbind(inp, nam, td->td_ucred);
995 INP_UNLOCK(inp);
996 INP_INFO_WUNLOCK(&udbinfo);
997 return error;
998 }
999
1000 static int
1001 udp_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
1002 {
1003 struct inpcb *inp;
1004 int error;
1005 struct sockaddr_in *sin;
1006
1007 INP_INFO_WLOCK(&udbinfo);
1008 inp = sotoinpcb(so);
1009 if (inp == 0) {
1010 INP_INFO_WUNLOCK(&udbinfo);
1011 return EINVAL;
1012 }
1013 INP_LOCK(inp);
1014 if (inp->inp_faddr.s_addr != INADDR_ANY) {
1015 INP_UNLOCK(inp);
1016 INP_INFO_WUNLOCK(&udbinfo);
1017 return EISCONN;
1018 }
1019 sin = (struct sockaddr_in *)nam;
1020 if (jailed(td->td_ucred))
1021 prison_remote_ip(td->td_ucred, 0, &sin->sin_addr.s_addr);
1022 error = in_pcbconnect(inp, nam, td->td_ucred);
1023 if (error == 0)
1024 soisconnected(so);
1025 INP_UNLOCK(inp);
1026 INP_INFO_WUNLOCK(&udbinfo);
1027 return error;
1028 }
1029
1030 static int
1031 udp_detach(struct socket *so)
1032 {
1033 struct inpcb *inp;
1034
1035 INP_INFO_WLOCK(&udbinfo);
1036 inp = sotoinpcb(so);
1037 if (inp == 0) {
1038 INP_INFO_WUNLOCK(&udbinfo);
1039 return EINVAL;
1040 }
1041 INP_LOCK(inp);
1042 in_pcbdetach(inp);
1043 INP_INFO_WUNLOCK(&udbinfo);
1044 return 0;
1045 }
1046
1047 static int
1048 udp_disconnect(struct socket *so)
1049 {
1050 struct inpcb *inp;
1051
1052 INP_INFO_WLOCK(&udbinfo);
1053 inp = sotoinpcb(so);
1054 if (inp == 0) {
1055 INP_INFO_WUNLOCK(&udbinfo);
1056 return EINVAL;
1057 }
1058 INP_LOCK(inp);
1059 if (inp->inp_faddr.s_addr == INADDR_ANY) {
1060 INP_INFO_WUNLOCK(&udbinfo);
1061 INP_UNLOCK(inp);
1062 return ENOTCONN;
1063 }
1064
1065 in_pcbdisconnect(inp);
1066 inp->inp_laddr.s_addr = INADDR_ANY;
1067 INP_UNLOCK(inp);
1068 INP_INFO_WUNLOCK(&udbinfo);
1069 so->so_state &= ~SS_ISCONNECTED; /* XXX */
1070 return 0;
1071 }
1072
1073 static int
1074 udp_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *addr,
1075 struct mbuf *control, struct thread *td)
1076 {
1077 struct inpcb *inp;
1078
1079 inp = sotoinpcb(so);
1080 return udp_output(inp, m, addr, control, td);
1081 }
1082
1083 int
1084 udp_shutdown(struct socket *so)
1085 {
1086 struct inpcb *inp;
1087
1088 INP_INFO_RLOCK(&udbinfo);
1089 inp = sotoinpcb(so);
1090 if (inp == 0) {
1091 INP_INFO_RUNLOCK(&udbinfo);
1092 return EINVAL;
1093 }
1094 INP_LOCK(inp);
1095 INP_INFO_RUNLOCK(&udbinfo);
1096 socantsendmore(so);
1097 INP_UNLOCK(inp);
1098 return 0;
1099 }
1100
1101 /*
1102 * This is the wrapper function for in_setsockaddr. We just pass down
1103 * the pcbinfo for in_setsockaddr to lock. We don't want to do the locking
1104 * here because in_setsockaddr will call malloc and might block.
1105 */
1106 static int
1107 udp_sockaddr(struct socket *so, struct sockaddr **nam)
1108 {
1109 return (in_setsockaddr(so, nam, &udbinfo));
1110 }
1111
1112 /*
1113 * This is the wrapper function for in_setpeeraddr. We just pass down
1114 * the pcbinfo for in_setpeeraddr to lock.
1115 */
1116 static int
1117 udp_peeraddr(struct socket *so, struct sockaddr **nam)
1118 {
1119 return (in_setpeeraddr(so, nam, &udbinfo));
1120 }
1121
1122 struct pr_usrreqs udp_usrreqs = {
1123 .pru_abort = udp_abort,
1124 .pru_attach = udp_attach,
1125 .pru_bind = udp_bind,
1126 .pru_connect = udp_connect,
1127 .pru_control = in_control,
1128 .pru_detach = udp_detach,
1129 .pru_disconnect = udp_disconnect,
1130 .pru_peeraddr = udp_peeraddr,
1131 .pru_send = udp_send,
1132 .pru_shutdown = udp_shutdown,
1133 .pru_sockaddr = udp_sockaddr,
1134 .pru_sosetlabel = in_pcbsosetlabel
1135 };
Cache object: 6e75679e3fccd9d2eb98bc7014137e6c
|