1 /* $NetBSD: ip_output.c,v 1.149.2.5 2007/03/31 14:15:43 bouyer Exp $ */
2
3 /*
4 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5 * All rights reserved.
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 * 3. Neither the name of the project nor the names of its contributors
16 * may be used to endorse or promote products derived from this software
17 * without specific prior written permission.
18 *
19 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 * SUCH DAMAGE.
30 */
31
32 /*-
33 * Copyright (c) 1998 The NetBSD Foundation, Inc.
34 * All rights reserved.
35 *
36 * This code is derived from software contributed to The NetBSD Foundation
37 * by Public Access Networks Corporation ("Panix"). It was developed under
38 * contract to Panix by Eric Haszlakiewicz and Thor Lancelot Simon.
39 *
40 * Redistribution and use in source and binary forms, with or without
41 * modification, are permitted provided that the following conditions
42 * are met:
43 * 1. Redistributions of source code must retain the above copyright
44 * notice, this list of conditions and the following disclaimer.
45 * 2. Redistributions in binary form must reproduce the above copyright
46 * notice, this list of conditions and the following disclaimer in the
47 * documentation and/or other materials provided with the distribution.
48 * 3. All advertising materials mentioning features or use of this software
49 * must display the following acknowledgement:
50 * This product includes software developed by the NetBSD
51 * Foundation, Inc. and its contributors.
52 * 4. Neither the name of The NetBSD Foundation nor the names of its
53 * contributors may be used to endorse or promote products derived
54 * from this software without specific prior written permission.
55 *
56 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
57 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
58 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
59 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
60 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
61 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
62 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
63 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
64 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
65 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
66 * POSSIBILITY OF SUCH DAMAGE.
67 */
68
69 /*
70 * Copyright (c) 1982, 1986, 1988, 1990, 1993
71 * The Regents of the University of California. All rights reserved.
72 *
73 * Redistribution and use in source and binary forms, with or without
74 * modification, are permitted provided that the following conditions
75 * are met:
76 * 1. Redistributions of source code must retain the above copyright
77 * notice, this list of conditions and the following disclaimer.
78 * 2. Redistributions in binary form must reproduce the above copyright
79 * notice, this list of conditions and the following disclaimer in the
80 * documentation and/or other materials provided with the distribution.
81 * 3. Neither the name of the University nor the names of its contributors
82 * may be used to endorse or promote products derived from this software
83 * without specific prior written permission.
84 *
85 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
86 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
87 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
88 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
89 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
90 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
91 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
92 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
93 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
94 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
95 * SUCH DAMAGE.
96 *
97 * @(#)ip_output.c 8.3 (Berkeley) 1/21/94
98 */
99
100 #include <sys/cdefs.h>
101 __KERNEL_RCSID(0, "$NetBSD: ip_output.c,v 1.149.2.5 2007/03/31 14:15:43 bouyer Exp $");
102
103 #include "opt_pfil_hooks.h"
104 #include "opt_inet.h"
105 #include "opt_ipsec.h"
106 #include "opt_mrouting.h"
107
108 #include <sys/param.h>
109 #include <sys/malloc.h>
110 #include <sys/mbuf.h>
111 #include <sys/errno.h>
112 #include <sys/protosw.h>
113 #include <sys/socket.h>
114 #include <sys/socketvar.h>
115 #ifdef FAST_IPSEC
116 #include <sys/domain.h>
117 #endif
118 #include <sys/systm.h>
119 #include <sys/proc.h>
120
121 #include <net/if.h>
122 #include <net/route.h>
123 #include <net/pfil.h>
124
125 #include <netinet/in.h>
126 #include <netinet/in_systm.h>
127 #include <netinet/ip.h>
128 #include <netinet/in_pcb.h>
129 #include <netinet/in_var.h>
130 #include <netinet/ip_var.h>
131
132 #ifdef MROUTING
133 #include <netinet/ip_mroute.h>
134 #endif
135
136 #include <machine/stdarg.h>
137
138 #ifdef IPSEC
139 #include <netinet6/ipsec.h>
140 #include <netkey/key.h>
141 #include <netkey/key_debug.h>
142 #ifdef IPSEC_NAT_T
143 #include <netinet/udp.h>
144 #endif
145 #endif /*IPSEC*/
146
147 #ifdef FAST_IPSEC
148 #include <netipsec/ipsec.h>
149 #include <netipsec/key.h>
150 #include <netipsec/xform.h>
151 #endif /* FAST_IPSEC*/
152
153 static struct mbuf *ip_insertoptions(struct mbuf *, struct mbuf *, int *);
154 static struct ifnet *ip_multicast_if(struct in_addr *, int *);
155 static void ip_mloopback(struct ifnet *, struct mbuf *, struct sockaddr_in *);
156 static int ip_getoptval(struct mbuf *, u_int8_t *, u_int);
157
158 #ifdef PFIL_HOOKS
159 extern struct pfil_head inet_pfil_hook; /* XXX */
160 #endif
161
162 int udp_do_loopback_cksum = 0;
163 int tcp_do_loopback_cksum = 0;
164 int ip_do_loopback_cksum = 0;
165
166 #define IN_NEED_CHECKSUM(ifp, csum_flags) \
167 (__predict_true(((ifp)->if_flags & IFF_LOOPBACK) == 0 || \
168 (((csum_flags) & M_CSUM_UDPv4) != 0 && udp_do_loopback_cksum) || \
169 (((csum_flags) & M_CSUM_TCPv4) != 0 && tcp_do_loopback_cksum) || \
170 (((csum_flags) & M_CSUM_IPv4) != 0 && ip_do_loopback_cksum)))
171
172 /*
173 * IP output. The packet in mbuf chain m contains a skeletal IP
174 * header (with len, off, ttl, proto, tos, src, dst).
175 * The mbuf chain containing the packet will be freed.
176 * The mbuf opt, if present, will not be freed.
177 */
178 int
179 ip_output(struct mbuf *m0, ...)
180 {
181 struct ip *ip;
182 struct ifnet *ifp;
183 struct mbuf *m = m0;
184 int hlen = sizeof (struct ip);
185 int len, error = 0;
186 struct route iproute;
187 struct sockaddr_in *dst;
188 struct in_ifaddr *ia;
189 struct mbuf *opt;
190 struct route *ro;
191 int flags, sw_csum;
192 int *mtu_p;
193 u_long mtu;
194 struct ip_moptions *imo;
195 struct socket *so;
196 va_list ap;
197 #ifdef IPSEC
198 struct secpolicy *sp = NULL;
199 #ifdef IPSEC_NAT_T
200 int natt_frag = 0;
201 #endif
202 #endif /*IPSEC*/
203 #ifdef FAST_IPSEC
204 struct inpcb *inp;
205 struct m_tag *mtag;
206 struct secpolicy *sp = NULL;
207 struct tdb_ident *tdbi;
208 int s;
209 #endif
210 u_int16_t ip_len;
211
212 len = 0;
213 va_start(ap, m0);
214 opt = va_arg(ap, struct mbuf *);
215 ro = va_arg(ap, struct route *);
216 flags = va_arg(ap, int);
217 imo = va_arg(ap, struct ip_moptions *);
218 so = va_arg(ap, struct socket *);
219 if (flags & IP_RETURNMTU)
220 mtu_p = va_arg(ap, int *);
221 else
222 mtu_p = NULL;
223 va_end(ap);
224
225 MCLAIM(m, &ip_tx_mowner);
226 #ifdef FAST_IPSEC
227 if (so != NULL && so->so_proto->pr_domain->dom_family == AF_INET)
228 inp = (struct inpcb *)so->so_pcb;
229 else
230 inp = NULL;
231 #endif /* FAST_IPSEC */
232
233 #ifdef DIAGNOSTIC
234 if ((m->m_flags & M_PKTHDR) == 0)
235 panic("ip_output no HDR");
236 #endif
237 if (opt) {
238 m = ip_insertoptions(m, opt, &len);
239 if (len >= sizeof(struct ip))
240 hlen = len;
241 }
242 ip = mtod(m, struct ip *);
243 /*
244 * Fill in IP header.
245 */
246 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
247 ip->ip_v = IPVERSION;
248 ip->ip_off = htons(0);
249 if ((m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0) {
250 ip->ip_id = ip_newid();
251 } else {
252
253 /*
254 * TSO capable interfaces (typically?) increment
255 * ip_id for each segment.
256 * "allocate" enough ids here to increase the chance
257 * for them to be unique.
258 *
259 * note that the following calculation is not
260 * needed to be precise. wasting some ip_id is fine.
261 */
262
263 unsigned int segsz = m->m_pkthdr.segsz;
264 unsigned int datasz = ntohs(ip->ip_len) - hlen;
265 unsigned int num = howmany(datasz, segsz);
266
267 ip->ip_id = ip_newid_range(num);
268 }
269 ip->ip_hl = hlen >> 2;
270 ipstat.ips_localout++;
271 } else {
272 hlen = ip->ip_hl << 2;
273 }
274 /*
275 * Route packet.
276 */
277 bzero(&iproute, sizeof(iproute));
278 if (ro == NULL)
279 ro = &iproute;
280 dst = satosin(&ro->ro_dst);
281 /*
282 * If there is a cached route,
283 * check that it is to the same destination
284 * and is still up. If not, free it and try again.
285 * The address family should also be checked in case of sharing the
286 * cache with IPv6.
287 */
288 if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
289 dst->sin_family != AF_INET ||
290 !in_hosteq(dst->sin_addr, ip->ip_dst))) {
291 RTFREE(ro->ro_rt);
292 ro->ro_rt = (struct rtentry *)0;
293 }
294 if (ro->ro_rt == 0) {
295 bzero(dst, sizeof(*dst));
296 dst->sin_family = AF_INET;
297 dst->sin_len = sizeof(*dst);
298 dst->sin_addr = ip->ip_dst;
299 }
300 /*
301 * If routing to interface only,
302 * short circuit routing lookup.
303 */
304 if (flags & IP_ROUTETOIF) {
305 if ((ia = ifatoia(ifa_ifwithladdr(sintosa(dst)))) == 0) {
306 ipstat.ips_noroute++;
307 error = ENETUNREACH;
308 goto bad;
309 }
310 ifp = ia->ia_ifp;
311 mtu = ifp->if_mtu;
312 ip->ip_ttl = 1;
313 } else if ((IN_MULTICAST(ip->ip_dst.s_addr) ||
314 ip->ip_dst.s_addr == INADDR_BROADCAST) &&
315 imo != NULL && imo->imo_multicast_ifp != NULL) {
316 ifp = imo->imo_multicast_ifp;
317 mtu = ifp->if_mtu;
318 IFP_TO_IA(ifp, ia);
319 } else {
320 if (ro->ro_rt == 0)
321 rtalloc(ro);
322 if (ro->ro_rt == 0) {
323 ipstat.ips_noroute++;
324 error = EHOSTUNREACH;
325 goto bad;
326 }
327 ia = ifatoia(ro->ro_rt->rt_ifa);
328 ifp = ro->ro_rt->rt_ifp;
329 if ((mtu = ro->ro_rt->rt_rmx.rmx_mtu) == 0)
330 mtu = ifp->if_mtu;
331 ro->ro_rt->rt_use++;
332 if (ro->ro_rt->rt_flags & RTF_GATEWAY)
333 dst = satosin(ro->ro_rt->rt_gateway);
334 }
335 if (IN_MULTICAST(ip->ip_dst.s_addr) ||
336 (ip->ip_dst.s_addr == INADDR_BROADCAST)) {
337 struct in_multi *inm;
338
339 m->m_flags |= (ip->ip_dst.s_addr == INADDR_BROADCAST) ?
340 M_BCAST : M_MCAST;
341 /*
342 * IP destination address is multicast. Make sure "dst"
343 * still points to the address in "ro". (It may have been
344 * changed to point to a gateway address, above.)
345 */
346 dst = satosin(&ro->ro_dst);
347 /*
348 * See if the caller provided any multicast options
349 */
350 if (imo != NULL)
351 ip->ip_ttl = imo->imo_multicast_ttl;
352 else
353 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
354
355 /*
356 * if we don't know the outgoing ifp yet, we can't generate
357 * output
358 */
359 if (!ifp) {
360 ipstat.ips_noroute++;
361 error = ENETUNREACH;
362 goto bad;
363 }
364
365 /*
366 * If the packet is multicast or broadcast, confirm that
367 * the outgoing interface can transmit it.
368 */
369 if (((m->m_flags & M_MCAST) &&
370 (ifp->if_flags & IFF_MULTICAST) == 0) ||
371 ((m->m_flags & M_BCAST) &&
372 (ifp->if_flags & (IFF_BROADCAST|IFF_POINTOPOINT)) == 0)) {
373 ipstat.ips_noroute++;
374 error = ENETUNREACH;
375 goto bad;
376 }
377 /*
378 * If source address not specified yet, use an address
379 * of outgoing interface.
380 */
381 if (in_nullhost(ip->ip_src)) {
382 struct in_ifaddr *ia;
383
384 IFP_TO_IA(ifp, ia);
385 if (!ia) {
386 error = EADDRNOTAVAIL;
387 goto bad;
388 }
389 ip->ip_src = ia->ia_addr.sin_addr;
390 }
391
392 IN_LOOKUP_MULTI(ip->ip_dst, ifp, inm);
393 if (inm != NULL &&
394 (imo == NULL || imo->imo_multicast_loop)) {
395 /*
396 * If we belong to the destination multicast group
397 * on the outgoing interface, and the caller did not
398 * forbid loopback, loop back a copy.
399 */
400 ip_mloopback(ifp, m, dst);
401 }
402 #ifdef MROUTING
403 else {
404 /*
405 * If we are acting as a multicast router, perform
406 * multicast forwarding as if the packet had just
407 * arrived on the interface to which we are about
408 * to send. The multicast forwarding function
409 * recursively calls this function, using the
410 * IP_FORWARDING flag to prevent infinite recursion.
411 *
412 * Multicasts that are looped back by ip_mloopback(),
413 * above, will be forwarded by the ip_input() routine,
414 * if necessary.
415 */
416 extern struct socket *ip_mrouter;
417
418 if (ip_mrouter && (flags & IP_FORWARDING) == 0) {
419 if (ip_mforward(m, ifp) != 0) {
420 m_freem(m);
421 goto done;
422 }
423 }
424 }
425 #endif
426 /*
427 * Multicasts with a time-to-live of zero may be looped-
428 * back, above, but must not be transmitted on a network.
429 * Also, multicasts addressed to the loopback interface
430 * are not sent -- the above call to ip_mloopback() will
431 * loop back a copy if this host actually belongs to the
432 * destination group on the loopback interface.
433 */
434 if (ip->ip_ttl == 0 || (ifp->if_flags & IFF_LOOPBACK) != 0) {
435 m_freem(m);
436 goto done;
437 }
438
439 goto sendit;
440 }
441 #ifndef notdef
442 /*
443 * If source address not specified yet, use address
444 * of outgoing interface.
445 */
446 if (in_nullhost(ip->ip_src))
447 ip->ip_src = ia->ia_addr.sin_addr;
448 #endif
449
450 /*
451 * packets with Class-D address as source are not valid per
452 * RFC 1112
453 */
454 if (IN_MULTICAST(ip->ip_src.s_addr)) {
455 ipstat.ips_odropped++;
456 error = EADDRNOTAVAIL;
457 goto bad;
458 }
459
460 /*
461 * Look for broadcast address and
462 * and verify user is allowed to send
463 * such a packet.
464 */
465 if (in_broadcast(dst->sin_addr, ifp)) {
466 if ((ifp->if_flags & IFF_BROADCAST) == 0) {
467 error = EADDRNOTAVAIL;
468 goto bad;
469 }
470 if ((flags & IP_ALLOWBROADCAST) == 0) {
471 error = EACCES;
472 goto bad;
473 }
474 /* don't allow broadcast messages to be fragmented */
475 if (ntohs(ip->ip_len) > ifp->if_mtu) {
476 error = EMSGSIZE;
477 goto bad;
478 }
479 m->m_flags |= M_BCAST;
480 } else
481 m->m_flags &= ~M_BCAST;
482
483 sendit:
484 /*
485 * If we're doing Path MTU Discovery, we need to set DF unless
486 * the route's MTU is locked.
487 */
488 if ((flags & IP_MTUDISC) != 0 && ro->ro_rt != NULL &&
489 (ro->ro_rt->rt_rmx.rmx_locks & RTV_MTU) == 0)
490 ip->ip_off |= htons(IP_DF);
491
492 /* Remember the current ip_len */
493 ip_len = ntohs(ip->ip_len);
494
495 #ifdef IPSEC
496 /* get SP for this packet */
497 if (so == NULL)
498 sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND,
499 flags, &error);
500 else {
501 if (IPSEC_PCB_SKIP_IPSEC(sotoinpcb_hdr(so)->inph_sp,
502 IPSEC_DIR_OUTBOUND))
503 goto skip_ipsec;
504 sp = ipsec4_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error);
505 }
506
507 if (sp == NULL) {
508 ipsecstat.out_inval++;
509 goto bad;
510 }
511
512 error = 0;
513
514 /* check policy */
515 switch (sp->policy) {
516 case IPSEC_POLICY_DISCARD:
517 /*
518 * This packet is just discarded.
519 */
520 ipsecstat.out_polvio++;
521 goto bad;
522
523 case IPSEC_POLICY_BYPASS:
524 case IPSEC_POLICY_NONE:
525 /* no need to do IPsec. */
526 goto skip_ipsec;
527
528 case IPSEC_POLICY_IPSEC:
529 if (sp->req == NULL) {
530 /* XXX should be panic ? */
531 printf("ip_output: No IPsec request specified.\n");
532 error = EINVAL;
533 goto bad;
534 }
535 break;
536
537 case IPSEC_POLICY_ENTRUST:
538 default:
539 printf("ip_output: Invalid policy found. %d\n", sp->policy);
540 }
541
542 #ifdef IPSEC_NAT_T
543 /*
544 * NAT-T ESP fragmentation: don't do IPSec processing now,
545 * we'll do it on each fragmented packet.
546 */
547 if (sp->req->sav &&
548 ((sp->req->sav->natt_type & UDP_ENCAP_ESPINUDP) ||
549 (sp->req->sav->natt_type & UDP_ENCAP_ESPINUDP_NON_IKE))) {
550 if (ntohs(ip->ip_len) > sp->req->sav->esp_frag) {
551 natt_frag = 1;
552 mtu = sp->req->sav->esp_frag;
553 goto skip_ipsec;
554 }
555 }
556 #endif /* IPSEC_NAT_T */
557
558 /*
559 * ipsec4_output() expects ip_len and ip_off in network
560 * order. They have been set to network order above.
561 */
562
563 {
564 struct ipsec_output_state state;
565 bzero(&state, sizeof(state));
566 state.m = m;
567 if (flags & IP_ROUTETOIF) {
568 state.ro = &iproute;
569 bzero(&iproute, sizeof(iproute));
570 } else
571 state.ro = ro;
572 state.dst = (struct sockaddr *)dst;
573
574 /*
575 * We can't defer the checksum of payload data if
576 * we're about to encrypt/authenticate it.
577 *
578 * XXX When we support crypto offloading functions of
579 * XXX network interfaces, we need to reconsider this,
580 * XXX since it's likely that they'll support checksumming,
581 * XXX as well.
582 */
583 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
584 in_delayed_cksum(m);
585 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
586 }
587
588 error = ipsec4_output(&state, sp, flags);
589
590 m = state.m;
591 if (flags & IP_ROUTETOIF) {
592 /*
593 * if we have tunnel mode SA, we may need to ignore
594 * IP_ROUTETOIF.
595 */
596 if (state.ro != &iproute || state.ro->ro_rt != NULL) {
597 flags &= ~IP_ROUTETOIF;
598 ro = state.ro;
599 }
600 } else
601 ro = state.ro;
602 dst = (struct sockaddr_in *)state.dst;
603 if (error) {
604 /* mbuf is already reclaimed in ipsec4_output. */
605 m0 = NULL;
606 switch (error) {
607 case EHOSTUNREACH:
608 case ENETUNREACH:
609 case EMSGSIZE:
610 case ENOBUFS:
611 case ENOMEM:
612 break;
613 default:
614 printf("ip4_output (ipsec): error code %d\n", error);
615 /*fall through*/
616 case ENOENT:
617 /* don't show these error codes to the user */
618 error = 0;
619 break;
620 }
621 goto bad;
622 }
623
624 /* be sure to update variables that are affected by ipsec4_output() */
625 ip = mtod(m, struct ip *);
626 hlen = ip->ip_hl << 2;
627 ip_len = ntohs(ip->ip_len);
628
629 if (ro->ro_rt == NULL) {
630 if ((flags & IP_ROUTETOIF) == 0) {
631 printf("ip_output: "
632 "can't update route after IPsec processing\n");
633 error = EHOSTUNREACH; /*XXX*/
634 goto bad;
635 }
636 } else {
637 /* nobody uses ia beyond here */
638 if (state.encap) {
639 ifp = ro->ro_rt->rt_ifp;
640 if ((mtu = ro->ro_rt->rt_rmx.rmx_mtu) == 0)
641 mtu = ifp->if_mtu;
642 }
643 }
644 }
645 skip_ipsec:
646 #endif /*IPSEC*/
647 #ifdef FAST_IPSEC
648 /*
649 * Check the security policy (SP) for the packet and, if
650 * required, do IPsec-related processing. There are two
651 * cases here; the first time a packet is sent through
652 * it will be untagged and handled by ipsec4_checkpolicy.
653 * If the packet is resubmitted to ip_output (e.g. after
654 * AH, ESP, etc. processing), there will be a tag to bypass
655 * the lookup and related policy checking.
656 */
657 mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL);
658 s = splsoftnet();
659 if (mtag != NULL) {
660 tdbi = (struct tdb_ident *)(mtag + 1);
661 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_OUTBOUND);
662 if (sp == NULL)
663 error = -EINVAL; /* force silent drop */
664 m_tag_delete(m, mtag);
665 } else {
666 if (inp != NULL &&
667 IPSEC_PCB_SKIP_IPSEC(inp->inp_sp, IPSEC_DIR_OUTBOUND))
668 goto spd_done;
669 sp = ipsec4_checkpolicy(m, IPSEC_DIR_OUTBOUND, flags,
670 &error, inp);
671 }
672 /*
673 * There are four return cases:
674 * sp != NULL apply IPsec policy
675 * sp == NULL, error == 0 no IPsec handling needed
676 * sp == NULL, error == -EINVAL discard packet w/o error
677 * sp == NULL, error != 0 discard packet, report error
678 */
679 if (sp != NULL) {
680 /* Loop detection, check if ipsec processing already done */
681 IPSEC_ASSERT(sp->req != NULL, ("ip_output: no ipsec request"));
682 for (mtag = m_tag_first(m); mtag != NULL;
683 mtag = m_tag_next(m, mtag)) {
684 #ifdef MTAG_ABI_COMPAT
685 if (mtag->m_tag_cookie != MTAG_ABI_COMPAT)
686 continue;
687 #endif
688 if (mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_DONE &&
689 mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED)
690 continue;
691 /*
692 * Check if policy has an SA associated with it.
693 * This can happen when an SP has yet to acquire
694 * an SA; e.g. on first reference. If it occurs,
695 * then we let ipsec4_process_packet do its thing.
696 */
697 if (sp->req->sav == NULL)
698 break;
699 tdbi = (struct tdb_ident *)(mtag + 1);
700 if (tdbi->spi == sp->req->sav->spi &&
701 tdbi->proto == sp->req->sav->sah->saidx.proto &&
702 bcmp(&tdbi->dst, &sp->req->sav->sah->saidx.dst,
703 sizeof (union sockaddr_union)) == 0) {
704 /*
705 * No IPsec processing is needed, free
706 * reference to SP.
707 *
708 * NB: null pointer to avoid free at
709 * done: below.
710 */
711 KEY_FREESP(&sp), sp = NULL;
712 splx(s);
713 goto spd_done;
714 }
715 }
716
717 /*
718 * Do delayed checksums now because we send before
719 * this is done in the normal processing path.
720 */
721 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
722 in_delayed_cksum(m);
723 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
724 }
725
726 #ifdef __FreeBSD__
727 ip->ip_len = htons(ip->ip_len);
728 ip->ip_off = htons(ip->ip_off);
729 #endif
730
731 /* NB: callee frees mbuf */
732 error = ipsec4_process_packet(m, sp->req, flags, 0);
733 /*
734 * Preserve KAME behaviour: ENOENT can be returned
735 * when an SA acquire is in progress. Don't propagate
736 * this to user-level; it confuses applications.
737 *
738 * XXX this will go away when the SADB is redone.
739 */
740 if (error == ENOENT)
741 error = 0;
742 splx(s);
743 goto done;
744 } else {
745 splx(s);
746
747 if (error != 0) {
748 /*
749 * Hack: -EINVAL is used to signal that a packet
750 * should be silently discarded. This is typically
751 * because we asked key management for an SA and
752 * it was delayed (e.g. kicked up to IKE).
753 */
754 if (error == -EINVAL)
755 error = 0;
756 goto bad;
757 } else {
758 /* No IPsec processing for this packet. */
759 }
760 #ifdef notyet
761 /*
762 * If deferred crypto processing is needed, check that
763 * the interface supports it.
764 */
765 mtag = m_tag_find(m, PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED, NULL);
766 if (mtag != NULL && (ifp->if_capenable & IFCAP_IPSEC) == 0) {
767 /* notify IPsec to do its own crypto */
768 ipsp_skipcrypto_unmark((struct tdb_ident *)(mtag + 1));
769 error = EHOSTUNREACH;
770 goto bad;
771 }
772 #endif
773 }
774 spd_done:
775 #endif /* FAST_IPSEC */
776
777 #ifdef PFIL_HOOKS
778 /*
779 * Run through list of hooks for output packets.
780 */
781 if ((error = pfil_run_hooks(&inet_pfil_hook, &m, ifp, PFIL_OUT)) != 0)
782 goto done;
783 if (m == NULL)
784 goto done;
785
786 ip = mtod(m, struct ip *);
787 hlen = ip->ip_hl << 2;
788 ip_len = ntohs(ip->ip_len);
789 #endif /* PFIL_HOOKS */
790
791 m->m_pkthdr.csum_data |= hlen << 16;
792
793 #if IFA_STATS
794 /*
795 * search for the source address structure to
796 * maintain output statistics.
797 */
798 INADDR_TO_IA(ip->ip_src, ia);
799 #endif
800
801 /* Maybe skip checksums on loopback interfaces. */
802 if (IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)) {
803 m->m_pkthdr.csum_flags |= M_CSUM_IPv4;
804 }
805 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx;
806 /*
807 * If small enough for mtu of path, or if using TCP segmentation
808 * offload, can just send directly.
809 */
810 if (ip_len <= mtu ||
811 (m->m_pkthdr.csum_flags & M_CSUM_TSOv4) != 0) {
812 #if IFA_STATS
813 if (ia)
814 ia->ia_ifa.ifa_data.ifad_outbytes += ip_len;
815 #endif
816 /*
817 * Always initialize the sum to 0! Some HW assisted
818 * checksumming requires this.
819 */
820 ip->ip_sum = 0;
821
822 if ((m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0) {
823 /*
824 * Perform any checksums that the hardware can't do
825 * for us.
826 *
827 * XXX Does any hardware require the {th,uh}_sum
828 * XXX fields to be 0?
829 */
830 if (sw_csum & M_CSUM_IPv4) {
831 KASSERT(IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4));
832 ip->ip_sum = in_cksum(m, hlen);
833 m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4;
834 }
835 if (sw_csum & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
836 if (IN_NEED_CHECKSUM(ifp,
837 sw_csum & (M_CSUM_TCPv4|M_CSUM_UDPv4))) {
838 in_delayed_cksum(m);
839 }
840 m->m_pkthdr.csum_flags &=
841 ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
842 }
843 }
844
845 #ifdef IPSEC
846 /* clean ipsec history once it goes out of the node */
847 ipsec_delaux(m);
848 #endif
849 error = (*ifp->if_output)(ifp, m, sintosa(dst), ro->ro_rt);
850 goto done;
851 }
852
853 /*
854 * We can't use HW checksumming if we're about to
855 * to fragment the packet.
856 *
857 * XXX Some hardware can do this.
858 */
859 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
860 if (IN_NEED_CHECKSUM(ifp,
861 m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4))) {
862 in_delayed_cksum(m);
863 }
864 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
865 }
866
867 /*
868 * Too large for interface; fragment if possible.
869 * Must be able to put at least 8 bytes per fragment.
870 */
871 if (ntohs(ip->ip_off) & IP_DF) {
872 if (flags & IP_RETURNMTU)
873 *mtu_p = mtu;
874 error = EMSGSIZE;
875 ipstat.ips_cantfrag++;
876 goto bad;
877 }
878
879 error = ip_fragment(m, ifp, mtu);
880 if (error) {
881 m = NULL;
882 goto bad;
883 }
884
885 for (; m; m = m0) {
886 m0 = m->m_nextpkt;
887 m->m_nextpkt = 0;
888 if (error == 0) {
889 #if IFA_STATS
890 if (ia)
891 ia->ia_ifa.ifa_data.ifad_outbytes +=
892 ntohs(ip->ip_len);
893 #endif
894 #ifdef IPSEC
895 /* clean ipsec history once it goes out of the node */
896 ipsec_delaux(m);
897
898 #ifdef IPSEC_NAT_T
899 /*
900 * If we get there, the packet has not been handeld by
901 * IPSec whereas it should have. Now that it has been
902 * fragmented, re-inject it in ip_output so that IPsec
903 * processing can occur.
904 */
905 if (natt_frag) {
906 error = ip_output(m, opt,
907 ro, flags, imo, so, mtu_p);
908 } else
909 #endif /* IPSEC_NAT_T */
910 #endif /* IPSEC */
911 {
912 KASSERT((m->m_pkthdr.csum_flags &
913 (M_CSUM_UDPv4 | M_CSUM_TCPv4)) == 0);
914 error = (*ifp->if_output)(ifp, m, sintosa(dst),
915 ro->ro_rt);
916 }
917 } else
918 m_freem(m);
919 }
920
921 if (error == 0)
922 ipstat.ips_fragmented++;
923 done:
924 if (iproute.ro_rt != NULL)
925 RTFREE(iproute.ro_rt);
926
927 #ifdef IPSEC
928 if (sp != NULL) {
929 KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
930 printf("DP ip_output call free SP:%p\n", sp));
931 key_freesp(sp);
932 }
933 #endif /* IPSEC */
934 #ifdef FAST_IPSEC
935 if (sp != NULL)
936 KEY_FREESP(&sp);
937 #endif /* FAST_IPSEC */
938
939 return (error);
940 bad:
941 m_freem(m);
942 goto done;
943 }
944
945 int
946 ip_fragment(struct mbuf *m, struct ifnet *ifp, u_long mtu)
947 {
948 struct ip *ip, *mhip;
949 struct mbuf *m0;
950 int len, hlen, off;
951 int mhlen, firstlen;
952 struct mbuf **mnext;
953 int sw_csum = m->m_pkthdr.csum_flags;
954 int fragments = 0;
955 int s;
956 int error = 0;
957
958 ip = mtod(m, struct ip *);
959 hlen = ip->ip_hl << 2;
960 if (ifp != NULL)
961 sw_csum &= ~ifp->if_csum_flags_tx;
962
963 len = (mtu - hlen) &~ 7;
964 if (len < 8) {
965 m_freem(m);
966 return (EMSGSIZE);
967 }
968
969 firstlen = len;
970 mnext = &m->m_nextpkt;
971
972 /*
973 * Loop through length of segment after first fragment,
974 * make new header and copy data of each part and link onto chain.
975 */
976 m0 = m;
977 mhlen = sizeof (struct ip);
978 for (off = hlen + len; off < ntohs(ip->ip_len); off += len) {
979 MGETHDR(m, M_DONTWAIT, MT_HEADER);
980 if (m == 0) {
981 error = ENOBUFS;
982 ipstat.ips_odropped++;
983 goto sendorfree;
984 }
985 MCLAIM(m, m0->m_owner);
986 *mnext = m;
987 mnext = &m->m_nextpkt;
988 m->m_data += max_linkhdr;
989 mhip = mtod(m, struct ip *);
990 *mhip = *ip;
991 /* we must inherit MCAST and BCAST flags */
992 m->m_flags |= m0->m_flags & (M_MCAST|M_BCAST);
993 if (hlen > sizeof (struct ip)) {
994 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
995 mhip->ip_hl = mhlen >> 2;
996 }
997 m->m_len = mhlen;
998 mhip->ip_off = ((off - hlen) >> 3) +
999 (ntohs(ip->ip_off) & ~IP_MF);
1000 if (ip->ip_off & htons(IP_MF))
1001 mhip->ip_off |= IP_MF;
1002 if (off + len >= ntohs(ip->ip_len))
1003 len = ntohs(ip->ip_len) - off;
1004 else
1005 mhip->ip_off |= IP_MF;
1006 HTONS(mhip->ip_off);
1007 mhip->ip_len = htons((u_int16_t)(len + mhlen));
1008 m->m_next = m_copy(m0, off, len);
1009 if (m->m_next == 0) {
1010 error = ENOBUFS; /* ??? */
1011 ipstat.ips_odropped++;
1012 goto sendorfree;
1013 }
1014 m->m_pkthdr.len = mhlen + len;
1015 m->m_pkthdr.rcvif = (struct ifnet *)0;
1016 mhip->ip_sum = 0;
1017 if (sw_csum & M_CSUM_IPv4) {
1018 mhip->ip_sum = in_cksum(m, mhlen);
1019 KASSERT((m->m_pkthdr.csum_flags & M_CSUM_IPv4) == 0);
1020 } else {
1021 m->m_pkthdr.csum_flags |= M_CSUM_IPv4;
1022 m->m_pkthdr.csum_data |= mhlen << 16;
1023 }
1024 ipstat.ips_ofragments++;
1025 fragments++;
1026 }
1027 /*
1028 * Update first fragment by trimming what's been copied out
1029 * and updating header, then send each fragment (in order).
1030 */
1031 m = m0;
1032 m_adj(m, hlen + firstlen - ntohs(ip->ip_len));
1033 m->m_pkthdr.len = hlen + firstlen;
1034 ip->ip_len = htons((u_int16_t)m->m_pkthdr.len);
1035 ip->ip_off |= htons(IP_MF);
1036 ip->ip_sum = 0;
1037 if (sw_csum & M_CSUM_IPv4) {
1038 ip->ip_sum = in_cksum(m, hlen);
1039 m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4;
1040 } else {
1041 KASSERT(m->m_pkthdr.csum_flags & M_CSUM_IPv4);
1042 KASSERT(M_CSUM_DATA_IPv4_IPHL(m->m_pkthdr.csum_data) >=
1043 sizeof(struct ip));
1044 }
1045 sendorfree:
1046 /*
1047 * If there is no room for all the fragments, don't queue
1048 * any of them.
1049 */
1050 if (ifp != NULL) {
1051 s = splnet();
1052 if (ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len < fragments &&
1053 error == 0) {
1054 error = ENOBUFS;
1055 ipstat.ips_odropped++;
1056 IFQ_INC_DROPS(&ifp->if_snd);
1057 }
1058 splx(s);
1059 }
1060 if (error) {
1061 for (m = m0; m; m = m0) {
1062 m0 = m->m_nextpkt;
1063 m->m_nextpkt = NULL;
1064 m_freem(m);
1065 }
1066 }
1067 return (error);
1068 }
1069
1070 /*
1071 * Process a delayed payload checksum calculation.
1072 */
1073 void
1074 in_delayed_cksum(struct mbuf *m)
1075 {
1076 struct ip *ip;
1077 u_int16_t csum, offset;
1078
1079 ip = mtod(m, struct ip *);
1080 offset = ip->ip_hl << 2;
1081 csum = in4_cksum(m, 0, offset, ntohs(ip->ip_len) - offset);
1082 if (csum == 0 && (m->m_pkthdr.csum_flags & M_CSUM_UDPv4) != 0)
1083 csum = 0xffff;
1084
1085 offset += M_CSUM_DATA_IPv4_OFFSET(m->m_pkthdr.csum_data);
1086
1087 if ((offset + sizeof(u_int16_t)) > m->m_len) {
1088 /* This happen when ip options were inserted
1089 printf("in_delayed_cksum: pullup len %d off %d proto %d\n",
1090 m->m_len, offset, ip->ip_p);
1091 */
1092 m_copyback(m, offset, sizeof(csum), (caddr_t) &csum);
1093 } else
1094 *(u_int16_t *)(mtod(m, caddr_t) + offset) = csum;
1095 }
1096
1097 /*
1098 * Determine the maximum length of the options to be inserted;
1099 * we would far rather allocate too much space rather than too little.
1100 */
1101
1102 u_int
1103 ip_optlen(struct inpcb *inp)
1104 {
1105 struct mbuf *m = inp->inp_options;
1106
1107 if (m && m->m_len > offsetof(struct ipoption, ipopt_dst))
1108 return (m->m_len - offsetof(struct ipoption, ipopt_dst));
1109 else
1110 return 0;
1111 }
1112
1113
1114 /*
1115 * Insert IP options into preformed packet.
1116 * Adjust IP destination as required for IP source routing,
1117 * as indicated by a non-zero in_addr at the start of the options.
1118 */
1119 static struct mbuf *
1120 ip_insertoptions(struct mbuf *m, struct mbuf *opt, int *phlen)
1121 {
1122 struct ipoption *p = mtod(opt, struct ipoption *);
1123 struct mbuf *n;
1124 struct ip *ip = mtod(m, struct ip *);
1125 unsigned optlen;
1126
1127 optlen = opt->m_len - sizeof(p->ipopt_dst);
1128 if (optlen + ntohs(ip->ip_len) > IP_MAXPACKET)
1129 return (m); /* XXX should fail */
1130 if (!in_nullhost(p->ipopt_dst))
1131 ip->ip_dst = p->ipopt_dst;
1132 if (M_READONLY(m) || M_LEADINGSPACE(m) < optlen) {
1133 MGETHDR(n, M_DONTWAIT, MT_HEADER);
1134 if (n == 0)
1135 return (m);
1136 MCLAIM(n, m->m_owner);
1137 M_COPY_PKTHDR(n, m);
1138 m_tag_delete_chain(m, NULL);
1139 m->m_flags &= ~M_PKTHDR;
1140 m->m_len -= sizeof(struct ip);
1141 m->m_data += sizeof(struct ip);
1142 n->m_next = m;
1143 m = n;
1144 m->m_len = optlen + sizeof(struct ip);
1145 m->m_data += max_linkhdr;
1146 bcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
1147 } else {
1148 m->m_data -= optlen;
1149 m->m_len += optlen;
1150 memmove(mtod(m, caddr_t), ip, sizeof(struct ip));
1151 }
1152 m->m_pkthdr.len += optlen;
1153 ip = mtod(m, struct ip *);
1154 bcopy((caddr_t)p->ipopt_list, (caddr_t)(ip + 1), (unsigned)optlen);
1155 *phlen = sizeof(struct ip) + optlen;
1156 ip->ip_len = htons(ntohs(ip->ip_len) + optlen);
1157 return (m);
1158 }
1159
1160 /*
1161 * Copy options from ip to jp,
1162 * omitting those not copied during fragmentation.
1163 */
1164 int
1165 ip_optcopy(struct ip *ip, struct ip *jp)
1166 {
1167 u_char *cp, *dp;
1168 int opt, optlen, cnt;
1169
1170 cp = (u_char *)(ip + 1);
1171 dp = (u_char *)(jp + 1);
1172 cnt = (ip->ip_hl << 2) - sizeof (struct ip);
1173 for (; cnt > 0; cnt -= optlen, cp += optlen) {
1174 opt = cp[0];
1175 if (opt == IPOPT_EOL)
1176 break;
1177 if (opt == IPOPT_NOP) {
1178 /* Preserve for IP mcast tunnel's LSRR alignment. */
1179 *dp++ = IPOPT_NOP;
1180 optlen = 1;
1181 continue;
1182 }
1183 #ifdef DIAGNOSTIC
1184 if (cnt < IPOPT_OLEN + sizeof(*cp))
1185 panic("malformed IPv4 option passed to ip_optcopy");
1186 #endif
1187 optlen = cp[IPOPT_OLEN];
1188 #ifdef DIAGNOSTIC
1189 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt)
1190 panic("malformed IPv4 option passed to ip_optcopy");
1191 #endif
1192 /* bogus lengths should have been caught by ip_dooptions */
1193 if (optlen > cnt)
1194 optlen = cnt;
1195 if (IPOPT_COPIED(opt)) {
1196 bcopy((caddr_t)cp, (caddr_t)dp, (unsigned)optlen);
1197 dp += optlen;
1198 }
1199 }
1200 for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++)
1201 *dp++ = IPOPT_EOL;
1202 return (optlen);
1203 }
1204
1205 /*
1206 * IP socket option processing.
1207 */
1208 int
1209 ip_ctloutput(int op, struct socket *so, int level, int optname,
1210 struct mbuf **mp)
1211 {
1212 struct inpcb *inp = sotoinpcb(so);
1213 struct mbuf *m = *mp;
1214 int optval = 0;
1215 int error = 0;
1216 #if defined(IPSEC) || defined(FAST_IPSEC)
1217 struct proc *p = curproc; /*XXX*/
1218 #endif
1219
1220 if (level != IPPROTO_IP) {
1221 error = EINVAL;
1222 if (op == PRCO_SETOPT && *mp)
1223 (void) m_free(*mp);
1224 } else switch (op) {
1225
1226 case PRCO_SETOPT:
1227 switch (optname) {
1228 case IP_OPTIONS:
1229 #ifdef notyet
1230 case IP_RETOPTS:
1231 return (ip_pcbopts(optname, &inp->inp_options, m));
1232 #else
1233 return (ip_pcbopts(&inp->inp_options, m));
1234 #endif
1235
1236 case IP_TOS:
1237 case IP_TTL:
1238 case IP_RECVOPTS:
1239 case IP_RECVRETOPTS:
1240 case IP_RECVDSTADDR:
1241 case IP_RECVIF:
1242 if (m == NULL || m->m_len != sizeof(int))
1243 error = EINVAL;
1244 else {
1245 optval = *mtod(m, int *);
1246 switch (optname) {
1247
1248 case IP_TOS:
1249 inp->inp_ip.ip_tos = optval;
1250 break;
1251
1252 case IP_TTL:
1253 inp->inp_ip.ip_ttl = optval;
1254 break;
1255 #define OPTSET(bit) \
1256 if (optval) \
1257 inp->inp_flags |= bit; \
1258 else \
1259 inp->inp_flags &= ~bit;
1260
1261 case IP_RECVOPTS:
1262 OPTSET(INP_RECVOPTS);
1263 break;
1264
1265 case IP_RECVRETOPTS:
1266 OPTSET(INP_RECVRETOPTS);
1267 break;
1268
1269 case IP_RECVDSTADDR:
1270 OPTSET(INP_RECVDSTADDR);
1271 break;
1272
1273 case IP_RECVIF:
1274 OPTSET(INP_RECVIF);
1275 break;
1276 }
1277 }
1278 break;
1279 #undef OPTSET
1280
1281 case IP_MULTICAST_IF:
1282 case IP_MULTICAST_TTL:
1283 case IP_MULTICAST_LOOP:
1284 case IP_ADD_MEMBERSHIP:
1285 case IP_DROP_MEMBERSHIP:
1286 error = ip_setmoptions(optname, &inp->inp_moptions, m);
1287 break;
1288
1289 case IP_PORTRANGE:
1290 if (m == 0 || m->m_len != sizeof(int))
1291 error = EINVAL;
1292 else {
1293 optval = *mtod(m, int *);
1294
1295 switch (optval) {
1296
1297 case IP_PORTRANGE_DEFAULT:
1298 case IP_PORTRANGE_HIGH:
1299 inp->inp_flags &= ~(INP_LOWPORT);
1300 break;
1301
1302 case IP_PORTRANGE_LOW:
1303 inp->inp_flags |= INP_LOWPORT;
1304 break;
1305
1306 default:
1307 error = EINVAL;
1308 break;
1309 }
1310 }
1311 break;
1312
1313 #if defined(IPSEC) || defined(FAST_IPSEC)
1314 case IP_IPSEC_POLICY:
1315 {
1316 caddr_t req = NULL;
1317 size_t len = 0;
1318 int priv = 0;
1319
1320 #ifdef __NetBSD__
1321 if (p == 0 || suser(p->p_ucred, &p->p_acflag))
1322 priv = 0;
1323 else
1324 priv = 1;
1325 #else
1326 priv = (in6p->in6p_socket->so_state & SS_PRIV);
1327 #endif
1328 if (m) {
1329 req = mtod(m, caddr_t);
1330 len = m->m_len;
1331 }
1332 error = ipsec4_set_policy(inp, optname, req, len, priv);
1333 break;
1334 }
1335 #endif /*IPSEC*/
1336
1337 default:
1338 error = ENOPROTOOPT;
1339 break;
1340 }
1341 if (m)
1342 (void)m_free(m);
1343 break;
1344
1345 case PRCO_GETOPT:
1346 switch (optname) {
1347 case IP_OPTIONS:
1348 case IP_RETOPTS:
1349 *mp = m = m_get(M_WAIT, MT_SOOPTS);
1350 MCLAIM(m, so->so_mowner);
1351 if (inp->inp_options) {
1352 m->m_len = inp->inp_options->m_len;
1353 bcopy(mtod(inp->inp_options, caddr_t),
1354 mtod(m, caddr_t), (unsigned)m->m_len);
1355 } else
1356 m->m_len = 0;
1357 break;
1358
1359 case IP_TOS:
1360 case IP_TTL:
1361 case IP_RECVOPTS:
1362 case IP_RECVRETOPTS:
1363 case IP_RECVDSTADDR:
1364 case IP_RECVIF:
1365 case IP_ERRORMTU:
1366 *mp = m = m_get(M_WAIT, MT_SOOPTS);
1367 MCLAIM(m, so->so_mowner);
1368 m->m_len = sizeof(int);
1369 switch (optname) {
1370
1371 case IP_TOS:
1372 optval = inp->inp_ip.ip_tos;
1373 break;
1374
1375 case IP_TTL:
1376 optval = inp->inp_ip.ip_ttl;
1377 break;
1378
1379 case IP_ERRORMTU:
1380 optval = inp->inp_errormtu;
1381 break;
1382
1383 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0)
1384
1385 case IP_RECVOPTS:
1386 optval = OPTBIT(INP_RECVOPTS);
1387 break;
1388
1389 case IP_RECVRETOPTS:
1390 optval = OPTBIT(INP_RECVRETOPTS);
1391 break;
1392
1393 case IP_RECVDSTADDR:
1394 optval = OPTBIT(INP_RECVDSTADDR);
1395 break;
1396
1397 case IP_RECVIF:
1398 optval = OPTBIT(INP_RECVIF);
1399 break;
1400 }
1401 *mtod(m, int *) = optval;
1402 break;
1403
1404 #if 0 /* defined(IPSEC) || defined(FAST_IPSEC) */
1405 /* XXX: code broken */
1406 case IP_IPSEC_POLICY:
1407 {
1408 caddr_t req = NULL;
1409 size_t len = 0;
1410
1411 if (m) {
1412 req = mtod(m, caddr_t);
1413 len = m->m_len;
1414 }
1415 error = ipsec4_get_policy(inp, req, len, mp);
1416 break;
1417 }
1418 #endif /*IPSEC*/
1419
1420 case IP_MULTICAST_IF:
1421 case IP_MULTICAST_TTL:
1422 case IP_MULTICAST_LOOP:
1423 case IP_ADD_MEMBERSHIP:
1424 case IP_DROP_MEMBERSHIP:
1425 error = ip_getmoptions(optname, inp->inp_moptions, mp);
1426 if (*mp)
1427 MCLAIM(*mp, so->so_mowner);
1428 break;
1429
1430 case IP_PORTRANGE:
1431 *mp = m = m_get(M_WAIT, MT_SOOPTS);
1432 MCLAIM(m, so->so_mowner);
1433 m->m_len = sizeof(int);
1434
1435 if (inp->inp_flags & INP_LOWPORT)
1436 optval = IP_PORTRANGE_LOW;
1437 else
1438 optval = IP_PORTRANGE_DEFAULT;
1439
1440 *mtod(m, int *) = optval;
1441 break;
1442
1443 default:
1444 error = ENOPROTOOPT;
1445 break;
1446 }
1447 break;
1448 }
1449 return (error);
1450 }
1451
1452 /*
1453 * Set up IP options in pcb for insertion in output packets.
1454 * Store in mbuf with pointer in pcbopt, adding pseudo-option
1455 * with destination address if source routed.
1456 */
1457 int
1458 #ifdef notyet
1459 ip_pcbopts(int optname, struct mbuf **pcbopt, struct mbuf *m)
1460 #else
1461 ip_pcbopts(struct mbuf **pcbopt, struct mbuf *m)
1462 #endif
1463 {
1464 int cnt, optlen;
1465 u_char *cp;
1466 u_char opt;
1467
1468 /* turn off any old options */
1469 if (*pcbopt)
1470 (void)m_free(*pcbopt);
1471 *pcbopt = 0;
1472 if (m == (struct mbuf *)0 || m->m_len == 0) {
1473 /*
1474 * Only turning off any previous options.
1475 */
1476 if (m)
1477 (void)m_free(m);
1478 return (0);
1479 }
1480
1481 #ifndef __vax__
1482 if (m->m_len % sizeof(int32_t))
1483 goto bad;
1484 #endif
1485 /*
1486 * IP first-hop destination address will be stored before
1487 * actual options; move other options back
1488 * and clear it when none present.
1489 */
1490 if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN])
1491 goto bad;
1492 cnt = m->m_len;
1493 m->m_len += sizeof(struct in_addr);
1494 cp = mtod(m, u_char *) + sizeof(struct in_addr);
1495 memmove(cp, mtod(m, caddr_t), (unsigned)cnt);
1496 bzero(mtod(m, caddr_t), sizeof(struct in_addr));
1497
1498 for (; cnt > 0; cnt -= optlen, cp += optlen) {
1499 opt = cp[IPOPT_OPTVAL];
1500 if (opt == IPOPT_EOL)
1501 break;
1502 if (opt == IPOPT_NOP)
1503 optlen = 1;
1504 else {
1505 if (cnt < IPOPT_OLEN + sizeof(*cp))
1506 goto bad;
1507 optlen = cp[IPOPT_OLEN];
1508 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt)
1509 goto bad;
1510 }
1511 switch (opt) {
1512
1513 default:
1514 break;
1515
1516 case IPOPT_LSRR:
1517 case IPOPT_SSRR:
1518 /*
1519 * user process specifies route as:
1520 * ->A->B->C->D
1521 * D must be our final destination (but we can't
1522 * check that since we may not have connected yet).
1523 * A is first hop destination, which doesn't appear in
1524 * actual IP option, but is stored before the options.
1525 */
1526 if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr))
1527 goto bad;
1528 m->m_len -= sizeof(struct in_addr);
1529 cnt -= sizeof(struct in_addr);
1530 optlen -= sizeof(struct in_addr);
1531 cp[IPOPT_OLEN] = optlen;
1532 /*
1533 * Move first hop before start of options.
1534 */
1535 bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t),
1536 sizeof(struct in_addr));
1537 /*
1538 * Then copy rest of options back
1539 * to close up the deleted entry.
1540 */
1541 (void)memmove(&cp[IPOPT_OFFSET+1],
1542 &cp[IPOPT_OFFSET+1] + sizeof(struct in_addr),
1543 (unsigned)cnt - (IPOPT_MINOFF - 1));
1544 break;
1545 }
1546 }
1547 if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr))
1548 goto bad;
1549 *pcbopt = m;
1550 return (0);
1551
1552 bad:
1553 (void)m_free(m);
1554 return (EINVAL);
1555 }
1556
1557 /*
1558 * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index.
1559 */
1560 static struct ifnet *
1561 ip_multicast_if(struct in_addr *a, int *ifindexp)
1562 {
1563 int ifindex;
1564 struct ifnet *ifp = NULL;
1565 struct in_ifaddr *ia;
1566
1567 if (ifindexp)
1568 *ifindexp = 0;
1569 if (ntohl(a->s_addr) >> 24 == 0) {
1570 ifindex = ntohl(a->s_addr) & 0xffffff;
1571 if (ifindex < 0 || if_indexlim <= ifindex)
1572 return NULL;
1573 ifp = ifindex2ifnet[ifindex];
1574 if (!ifp)
1575 return NULL;
1576 if (ifindexp)
1577 *ifindexp = ifindex;
1578 } else {
1579 LIST_FOREACH(ia, &IN_IFADDR_HASH(a->s_addr), ia_hash) {
1580 if (in_hosteq(ia->ia_addr.sin_addr, *a) &&
1581 (ia->ia_ifp->if_flags & IFF_MULTICAST) != 0) {
1582 ifp = ia->ia_ifp;
1583 break;
1584 }
1585 }
1586 }
1587 return ifp;
1588 }
1589
1590 static int
1591 ip_getoptval(struct mbuf *m, u_int8_t *val, u_int maxval)
1592 {
1593 u_int tval;
1594
1595 if (m == NULL)
1596 return EINVAL;
1597
1598 switch (m->m_len) {
1599 case sizeof(u_char):
1600 tval = *(mtod(m, u_char *));
1601 break;
1602 case sizeof(u_int):
1603 tval = *(mtod(m, u_int *));
1604 break;
1605 default:
1606 return EINVAL;
1607 }
1608
1609 if (tval > maxval)
1610 return EINVAL;
1611
1612 *val = tval;
1613 return 0;
1614 }
1615
1616 /*
1617 * Set the IP multicast options in response to user setsockopt().
1618 */
1619 int
1620 ip_setmoptions(int optname, struct ip_moptions **imop, struct mbuf *m)
1621 {
1622 int error = 0;
1623 int i;
1624 struct in_addr addr;
1625 struct ip_mreq *mreq;
1626 struct ifnet *ifp;
1627 struct ip_moptions *imo = *imop;
1628 struct route ro;
1629 struct sockaddr_in *dst;
1630 int ifindex;
1631
1632 if (imo == NULL) {
1633 /*
1634 * No multicast option buffer attached to the pcb;
1635 * allocate one and initialize to default values.
1636 */
1637 imo = (struct ip_moptions *)malloc(sizeof(*imo), M_IPMOPTS,
1638 M_WAITOK);
1639
1640 if (imo == NULL)
1641 return (ENOBUFS);
1642 *imop = imo;
1643 imo->imo_multicast_ifp = NULL;
1644 imo->imo_multicast_addr.s_addr = INADDR_ANY;
1645 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
1646 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP;
1647 imo->imo_num_memberships = 0;
1648 }
1649
1650 switch (optname) {
1651
1652 case IP_MULTICAST_IF:
1653 /*
1654 * Select the interface for outgoing multicast packets.
1655 */
1656 if (m == NULL || m->m_len != sizeof(struct in_addr)) {
1657 error = EINVAL;
1658 break;
1659 }
1660 addr = *(mtod(m, struct in_addr *));
1661 /*
1662 * INADDR_ANY is used to remove a previous selection.
1663 * When no interface is selected, a default one is
1664 * chosen every time a multicast packet is sent.
1665 */
1666 if (in_nullhost(addr)) {
1667 imo->imo_multicast_ifp = NULL;
1668 break;
1669 }
1670 /*
1671 * The selected interface is identified by its local
1672 * IP address. Find the interface and confirm that
1673 * it supports multicasting.
1674 */
1675 ifp = ip_multicast_if(&addr, &ifindex);
1676 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1677 error = EADDRNOTAVAIL;
1678 break;
1679 }
1680 imo->imo_multicast_ifp = ifp;
1681 if (ifindex)
1682 imo->imo_multicast_addr = addr;
1683 else
1684 imo->imo_multicast_addr.s_addr = INADDR_ANY;
1685 break;
1686
1687 case IP_MULTICAST_TTL:
1688 /*
1689 * Set the IP time-to-live for outgoing multicast packets.
1690 */
1691 error = ip_getoptval(m, &imo->imo_multicast_ttl, MAXTTL);
1692 break;
1693
1694 case IP_MULTICAST_LOOP:
1695 /*
1696 * Set the loopback flag for outgoing multicast packets.
1697 * Must be zero or one.
1698 */
1699 error = ip_getoptval(m, &imo->imo_multicast_loop, 1);
1700 break;
1701
1702 case IP_ADD_MEMBERSHIP:
1703 /*
1704 * Add a multicast group membership.
1705 * Group must be a valid IP multicast address.
1706 */
1707 if (m == NULL || m->m_len != sizeof(struct ip_mreq)) {
1708 error = EINVAL;
1709 break;
1710 }
1711 mreq = mtod(m, struct ip_mreq *);
1712 if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) {
1713 error = EINVAL;
1714 break;
1715 }
1716 /*
1717 * If no interface address was provided, use the interface of
1718 * the route to the given multicast address.
1719 */
1720 if (in_nullhost(mreq->imr_interface)) {
1721 bzero((caddr_t)&ro, sizeof(ro));
1722 ro.ro_rt = NULL;
1723 dst = satosin(&ro.ro_dst);
1724 dst->sin_len = sizeof(*dst);
1725 dst->sin_family = AF_INET;
1726 dst->sin_addr = mreq->imr_multiaddr;
1727 rtalloc(&ro);
1728 if (ro.ro_rt == NULL) {
1729 error = EADDRNOTAVAIL;
1730 break;
1731 }
1732 ifp = ro.ro_rt->rt_ifp;
1733 rtfree(ro.ro_rt);
1734 } else {
1735 ifp = ip_multicast_if(&mreq->imr_interface, NULL);
1736 }
1737 /*
1738 * See if we found an interface, and confirm that it
1739 * supports multicast.
1740 */
1741 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1742 error = EADDRNOTAVAIL;
1743 break;
1744 }
1745 /*
1746 * See if the membership already exists or if all the
1747 * membership slots are full.
1748 */
1749 for (i = 0; i < imo->imo_num_memberships; ++i) {
1750 if (imo->imo_membership[i]->inm_ifp == ifp &&
1751 in_hosteq(imo->imo_membership[i]->inm_addr,
1752 mreq->imr_multiaddr))
1753 break;
1754 }
1755 if (i < imo->imo_num_memberships) {
1756 error = EADDRINUSE;
1757 break;
1758 }
1759 if (i == IP_MAX_MEMBERSHIPS) {
1760 error = ETOOMANYREFS;
1761 break;
1762 }
1763 /*
1764 * Everything looks good; add a new record to the multicast
1765 * address list for the given interface.
1766 */
1767 if ((imo->imo_membership[i] =
1768 in_addmulti(&mreq->imr_multiaddr, ifp)) == NULL) {
1769 error = ENOBUFS;
1770 break;
1771 }
1772 ++imo->imo_num_memberships;
1773 break;
1774
1775 case IP_DROP_MEMBERSHIP:
1776 /*
1777 * Drop a multicast group membership.
1778 * Group must be a valid IP multicast address.
1779 */
1780 if (m == NULL || m->m_len != sizeof(struct ip_mreq)) {
1781 error = EINVAL;
1782 break;
1783 }
1784 mreq = mtod(m, struct ip_mreq *);
1785 if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) {
1786 error = EINVAL;
1787 break;
1788 }
1789 /*
1790 * If an interface address was specified, get a pointer
1791 * to its ifnet structure.
1792 */
1793 if (in_nullhost(mreq->imr_interface))
1794 ifp = NULL;
1795 else {
1796 ifp = ip_multicast_if(&mreq->imr_interface, NULL);
1797 if (ifp == NULL) {
1798 error = EADDRNOTAVAIL;
1799 break;
1800 }
1801 }
1802 /*
1803 * Find the membership in the membership array.
1804 */
1805 for (i = 0; i < imo->imo_num_memberships; ++i) {
1806 if ((ifp == NULL ||
1807 imo->imo_membership[i]->inm_ifp == ifp) &&
1808 in_hosteq(imo->imo_membership[i]->inm_addr,
1809 mreq->imr_multiaddr))
1810 break;
1811 }
1812 if (i == imo->imo_num_memberships) {
1813 error = EADDRNOTAVAIL;
1814 break;
1815 }
1816 /*
1817 * Give up the multicast address record to which the
1818 * membership points.
1819 */
1820 in_delmulti(imo->imo_membership[i]);
1821 /*
1822 * Remove the gap in the membership array.
1823 */
1824 for (++i; i < imo->imo_num_memberships; ++i)
1825 imo->imo_membership[i-1] = imo->imo_membership[i];
1826 --imo->imo_num_memberships;
1827 break;
1828
1829 default:
1830 error = EOPNOTSUPP;
1831 break;
1832 }
1833
1834 /*
1835 * If all options have default values, no need to keep the mbuf.
1836 */
1837 if (imo->imo_multicast_ifp == NULL &&
1838 imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL &&
1839 imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP &&
1840 imo->imo_num_memberships == 0) {
1841 free(*imop, M_IPMOPTS);
1842 *imop = NULL;
1843 }
1844
1845 return (error);
1846 }
1847
1848 /*
1849 * Return the IP multicast options in response to user getsockopt().
1850 */
1851 int
1852 ip_getmoptions(int optname, struct ip_moptions *imo, struct mbuf **mp)
1853 {
1854 u_char *ttl;
1855 u_char *loop;
1856 struct in_addr *addr;
1857 struct in_ifaddr *ia;
1858
1859 *mp = m_get(M_WAIT, MT_SOOPTS);
1860
1861 switch (optname) {
1862
1863 case IP_MULTICAST_IF:
1864 addr = mtod(*mp, struct in_addr *);
1865 (*mp)->m_len = sizeof(struct in_addr);
1866 if (imo == NULL || imo->imo_multicast_ifp == NULL)
1867 *addr = zeroin_addr;
1868 else if (imo->imo_multicast_addr.s_addr) {
1869 /* return the value user has set */
1870 *addr = imo->imo_multicast_addr;
1871 } else {
1872 IFP_TO_IA(imo->imo_multicast_ifp, ia);
1873 *addr = ia ? ia->ia_addr.sin_addr : zeroin_addr;
1874 }
1875 return (0);
1876
1877 case IP_MULTICAST_TTL:
1878 ttl = mtod(*mp, u_char *);
1879 (*mp)->m_len = 1;
1880 *ttl = imo ? imo->imo_multicast_ttl
1881 : IP_DEFAULT_MULTICAST_TTL;
1882 return (0);
1883
1884 case IP_MULTICAST_LOOP:
1885 loop = mtod(*mp, u_char *);
1886 (*mp)->m_len = 1;
1887 *loop = imo ? imo->imo_multicast_loop
1888 : IP_DEFAULT_MULTICAST_LOOP;
1889 return (0);
1890
1891 default:
1892 return (EOPNOTSUPP);
1893 }
1894 }
1895
1896 /*
1897 * Discard the IP multicast options.
1898 */
1899 void
1900 ip_freemoptions(struct ip_moptions *imo)
1901 {
1902 int i;
1903
1904 if (imo != NULL) {
1905 for (i = 0; i < imo->imo_num_memberships; ++i)
1906 in_delmulti(imo->imo_membership[i]);
1907 free(imo, M_IPMOPTS);
1908 }
1909 }
1910
1911 /*
1912 * Routine called from ip_output() to loop back a copy of an IP multicast
1913 * packet to the input queue of a specified interface. Note that this
1914 * calls the output routine of the loopback "driver", but with an interface
1915 * pointer that might NOT be lo0ifp -- easier than replicating that code here.
1916 */
1917 static void
1918 ip_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in *dst)
1919 {
1920 struct ip *ip;
1921 struct mbuf *copym;
1922
1923 copym = m_copy(m, 0, M_COPYALL);
1924 if (copym != NULL
1925 && (copym->m_flags & M_EXT || copym->m_len < sizeof(struct ip)))
1926 copym = m_pullup(copym, sizeof(struct ip));
1927 if (copym != NULL) {
1928 /*
1929 * We don't bother to fragment if the IP length is greater
1930 * than the interface's MTU. Can this possibly matter?
1931 */
1932 ip = mtod(copym, struct ip *);
1933
1934 if (copym->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
1935 in_delayed_cksum(copym);
1936 copym->m_pkthdr.csum_flags &=
1937 ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
1938 }
1939
1940 ip->ip_sum = 0;
1941 ip->ip_sum = in_cksum(copym, ip->ip_hl << 2);
1942 (void) looutput(ifp, copym, sintosa(dst), NULL);
1943 }
1944 }
Cache object: 4007455cba1060b40fe33569517a5078
|