1 /*
2 * Copyright (c) 1982, 1986, 1988, 1990, 1993
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 * 3. All advertising materials mentioning features or use of this software
14 * must display the following acknowledgement:
15 * This product includes software developed by the University of
16 * California, Berkeley and its contributors.
17 * 4. Neither the name of the University nor the names of its contributors
18 * may be used to endorse or promote products derived from this software
19 * without specific prior written permission.
20 *
21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31 * SUCH DAMAGE.
32 *
33 * @(#)ip_output.c 8.3 (Berkeley) 1/21/94
34 * $FreeBSD$
35 */
36
37 #define _IP_VHL
38
39 #include "opt_ipfw.h"
40 #include "opt_ipdn.h"
41 #include "opt_ipdivert.h"
42 #include "opt_ipfilter.h"
43 #include "opt_ipsec.h"
44 #include "opt_random_ip_id.h"
45 #include "opt_mbuf_stress_test.h"
46
47 #include <sys/param.h>
48 #include <sys/systm.h>
49 #include <sys/kernel.h>
50 #include <sys/malloc.h>
51 #include <sys/mbuf.h>
52 #include <sys/protosw.h>
53 #include <sys/socket.h>
54 #include <sys/socketvar.h>
55 #include <sys/proc.h>
56 #include <sys/sysctl.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/ip.h>
64 #include <netinet/in_pcb.h>
65 #include <netinet/in_var.h>
66 #include <netinet/ip_var.h>
67
68 #include <machine/in_cksum.h>
69
70 static MALLOC_DEFINE(M_IPMOPTS, "ip_moptions", "internet multicast options");
71
72 #ifdef IPSEC
73 #include <netinet6/ipsec.h>
74 #include <netkey/key.h>
75 #ifdef IPSEC_DEBUG
76 #include <netkey/key_debug.h>
77 #else
78 #define KEYDEBUG(lev,arg)
79 #endif
80 #endif /*IPSEC*/
81
82 #ifdef FAST_IPSEC
83 #include <netipsec/ipsec.h>
84 #include <netipsec/xform.h>
85 #include <netipsec/key.h>
86 #endif /*FAST_IPSEC*/
87
88 #include <netinet/ip_fw.h>
89 #include <netinet/ip_dummynet.h>
90
91 #define print_ip(x, a, y) printf("%s %d.%d.%d.%d%s",\
92 x, (ntohl(a.s_addr)>>24)&0xFF,\
93 (ntohl(a.s_addr)>>16)&0xFF,\
94 (ntohl(a.s_addr)>>8)&0xFF,\
95 (ntohl(a.s_addr))&0xFF, y);
96
97 u_short ip_id;
98
99 #ifdef MBUF_STRESS_TEST
100 int mbuf_frag_size = 0;
101 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
102 &mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
103 #endif
104
105 static struct mbuf *ip_insertoptions(struct mbuf *, struct mbuf *, int *);
106 static struct ifnet *ip_multicast_if(struct in_addr *, int *);
107 static void ip_mloopback
108 (struct ifnet *, struct mbuf *, struct sockaddr_in *, int);
109 static int ip_getmoptions
110 (struct sockopt *, struct ip_moptions *);
111 static int ip_pcbopts(int, struct mbuf **, struct mbuf *);
112 static int ip_setmoptions
113 (struct sockopt *, struct ip_moptions **);
114
115 int ip_optcopy(struct ip *, struct ip *);
116 extern int (*fr_checkp) (struct ip *, int, struct ifnet *, int, struct mbuf **);
117
118
119 extern struct protosw inetsw[];
120
121 /*
122 * IP output. The packet in mbuf chain m contains a skeletal IP
123 * header (with len, off, ttl, proto, tos, src, dst).
124 * The mbuf chain containing the packet will be freed.
125 * The mbuf opt, if present, will not be freed.
126 */
127 int
128 ip_output(struct mbuf *m0, struct mbuf *opt, struct route *ro,
129 int flags, struct ip_moptions *imo, struct inpcb *inp)
130 {
131 struct ip *ip;
132 struct ifnet *ifp = NULL; /* keep compiler happy */
133 struct mbuf *m;
134 int hlen = sizeof (struct ip);
135 int len, off, error = 0;
136 struct sockaddr_in *dst = NULL; /* keep compiler happy */
137 struct in_ifaddr *ia = NULL;
138 int isbroadcast, sw_csum;
139 struct in_addr pkt_dst;
140 #ifdef IPSEC
141 struct route iproute;
142 struct secpolicy *sp = NULL;
143 struct socket *so = inp ? inp->inp_socket : NULL;
144 #endif
145 #ifdef FAST_IPSEC
146 struct route iproute;
147 struct m_tag *mtag;
148 struct secpolicy *sp = NULL;
149 struct tdb_ident *tdbi;
150 int s;
151 #endif /* FAST_IPSEC */
152 struct ip_fw_args args;
153 int src_was_INADDR_ANY = 0; /* as the name says... */
154
155 args.eh = NULL;
156 args.rule = NULL;
157 args.next_hop = NULL;
158 args.divert_rule = 0; /* divert cookie */
159
160 /* Grab info from MT_TAG mbufs prepended to the chain. */
161 for (; m0 && m0->m_type == MT_TAG; m0 = m0->m_next) {
162 switch(m0->_m_tag_id) {
163 default:
164 printf("ip_output: unrecognised MT_TAG tag %d\n",
165 m0->_m_tag_id);
166 break;
167
168 case PACKET_TAG_DUMMYNET:
169 /*
170 * the packet was already tagged, so part of the
171 * processing was already done, and we need to go down.
172 * Get parameters from the header.
173 */
174 args.rule = ((struct dn_pkt *)m0)->rule;
175 opt = NULL ;
176 ro = & ( ((struct dn_pkt *)m0)->ro ) ;
177 imo = NULL ;
178 dst = ((struct dn_pkt *)m0)->dn_dst ;
179 ifp = ((struct dn_pkt *)m0)->ifp ;
180 flags = ((struct dn_pkt *)m0)->flags ;
181 break;
182
183 case PACKET_TAG_DIVERT:
184 args.divert_rule = (int)m0->m_data & 0xffff;
185 break;
186
187 case PACKET_TAG_IPFORWARD:
188 args.next_hop = (struct sockaddr_in *)m0->m_data;
189 break;
190 }
191 }
192 m = m0;
193
194 KASSERT(!m || (m->m_flags & M_PKTHDR) != 0, ("ip_output: no HDR"));
195 #ifndef FAST_IPSEC
196 KASSERT(ro != NULL, ("ip_output: no route, proto %d",
197 mtod(m, struct ip *)->ip_p));
198 #endif
199
200 if (args.rule != NULL) { /* dummynet already saw us */
201 ip = mtod(m, struct ip *);
202 hlen = IP_VHL_HL(ip->ip_vhl) << 2 ;
203 if (ro->ro_rt)
204 ia = ifatoia(ro->ro_rt->rt_ifa);
205 goto sendit;
206 }
207
208 if (opt) {
209 len = 0;
210 m = ip_insertoptions(m, opt, &len);
211 if (len != 0)
212 hlen = len;
213 }
214 ip = mtod(m, struct ip *);
215 pkt_dst = args.next_hop ? args.next_hop->sin_addr : ip->ip_dst;
216
217 /*
218 * Fill in IP header.
219 */
220 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
221 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, hlen >> 2);
222 ip->ip_off &= IP_DF;
223 #ifdef RANDOM_IP_ID
224 ip->ip_id = ip_randomid();
225 #else
226 ip->ip_id = htons(ip_id++);
227 #endif
228 ipstat.ips_localout++;
229 } else {
230 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
231 }
232
233 #ifdef FAST_IPSEC
234 if (ro == NULL) {
235 ro = &iproute;
236 bzero(ro, sizeof (*ro));
237 }
238 #endif /* FAST_IPSEC */
239 dst = (struct sockaddr_in *)&ro->ro_dst;
240 /*
241 * If there is a cached route,
242 * check that it is to the same destination
243 * and is still up. If not, free it and try again.
244 * The address family should also be checked in case of sharing the
245 * cache with IPv6.
246 */
247 if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
248 dst->sin_family != AF_INET ||
249 dst->sin_addr.s_addr != pkt_dst.s_addr)) {
250 RTFREE(ro->ro_rt);
251 ro->ro_rt = (struct rtentry *)0;
252 }
253 if (ro->ro_rt == 0) {
254 bzero(dst, sizeof(*dst));
255 dst->sin_family = AF_INET;
256 dst->sin_len = sizeof(*dst);
257 dst->sin_addr = pkt_dst;
258 }
259 /*
260 * If routing to interface only,
261 * short circuit routing lookup.
262 */
263 if (flags & IP_ROUTETOIF) {
264 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == 0 &&
265 (ia = ifatoia(ifa_ifwithnet(sintosa(dst)))) == 0) {
266 ipstat.ips_noroute++;
267 error = ENETUNREACH;
268 goto bad;
269 }
270 ifp = ia->ia_ifp;
271 ip->ip_ttl = 1;
272 isbroadcast = in_broadcast(dst->sin_addr, ifp);
273 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
274 imo != NULL && imo->imo_multicast_ifp != NULL) {
275 /*
276 * Bypass the normal routing lookup for multicast
277 * packets if the interface is specified.
278 */
279 ifp = imo->imo_multicast_ifp;
280 IFP_TO_IA(ifp, ia);
281 isbroadcast = 0; /* fool gcc */
282 } else {
283 /*
284 * If this is the case, we probably don't want to allocate
285 * a protocol-cloned route since we didn't get one from the
286 * ULP. This lets TCP do its thing, while not burdening
287 * forwarding or ICMP with the overhead of cloning a route.
288 * Of course, we still want to do any cloning requested by
289 * the link layer, as this is probably required in all cases
290 * for correct operation (as it is for ARP).
291 */
292 if (ro->ro_rt == 0)
293 rtalloc_ign(ro, RTF_PRCLONING);
294 if (ro->ro_rt == 0) {
295 ipstat.ips_noroute++;
296 error = EHOSTUNREACH;
297 goto bad;
298 }
299 ia = ifatoia(ro->ro_rt->rt_ifa);
300 ifp = ro->ro_rt->rt_ifp;
301 ro->ro_rt->rt_use++;
302 if (ro->ro_rt->rt_flags & RTF_GATEWAY)
303 dst = (struct sockaddr_in *)ro->ro_rt->rt_gateway;
304 if (ro->ro_rt->rt_flags & RTF_HOST)
305 isbroadcast = (ro->ro_rt->rt_flags & RTF_BROADCAST);
306 else
307 isbroadcast = in_broadcast(dst->sin_addr, ifp);
308 }
309 if (IN_MULTICAST(ntohl(pkt_dst.s_addr))) {
310 struct in_multi *inm;
311
312 m->m_flags |= M_MCAST;
313 /*
314 * IP destination address is multicast. Make sure "dst"
315 * still points to the address in "ro". (It may have been
316 * changed to point to a gateway address, above.)
317 */
318 dst = (struct sockaddr_in *)&ro->ro_dst;
319 /*
320 * See if the caller provided any multicast options
321 */
322 if (imo != NULL) {
323 ip->ip_ttl = imo->imo_multicast_ttl;
324 if (imo->imo_multicast_vif != -1)
325 ip->ip_src.s_addr =
326 ip_mcast_src ?
327 ip_mcast_src(imo->imo_multicast_vif) :
328 INADDR_ANY;
329 } else
330 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
331 /*
332 * Confirm that the outgoing interface supports multicast.
333 */
334 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
335 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
336 ipstat.ips_noroute++;
337 error = ENETUNREACH;
338 goto bad;
339 }
340 }
341 /*
342 * If source address not specified yet, use address
343 * of outgoing interface.
344 */
345 if (ip->ip_src.s_addr == INADDR_ANY) {
346 /* Interface may have no addresses. */
347 if (ia != NULL)
348 ip->ip_src = IA_SIN(ia)->sin_addr;
349 }
350
351 IN_LOOKUP_MULTI(pkt_dst, ifp, inm);
352 if (inm != NULL &&
353 (imo == NULL || imo->imo_multicast_loop)) {
354 /*
355 * If we belong to the destination multicast group
356 * on the outgoing interface, and the caller did not
357 * forbid loopback, loop back a copy.
358 */
359 ip_mloopback(ifp, m, dst, hlen);
360 }
361 else {
362 /*
363 * If we are acting as a multicast router, perform
364 * multicast forwarding as if the packet had just
365 * arrived on the interface to which we are about
366 * to send. The multicast forwarding function
367 * recursively calls this function, using the
368 * IP_FORWARDING flag to prevent infinite recursion.
369 *
370 * Multicasts that are looped back by ip_mloopback(),
371 * above, will be forwarded by the ip_input() routine,
372 * if necessary.
373 */
374 if (ip_mrouter && (flags & IP_FORWARDING) == 0) {
375 /*
376 * If rsvp daemon is not running, do not
377 * set ip_moptions. This ensures that the packet
378 * is multicast and not just sent down one link
379 * as prescribed by rsvpd.
380 */
381 if (!rsvp_on)
382 imo = NULL;
383 if (ip_mforward &&
384 ip_mforward(ip, ifp, m, imo) != 0) {
385 m_freem(m);
386 goto done;
387 }
388 }
389 }
390
391 /*
392 * Multicasts with a time-to-live of zero may be looped-
393 * back, above, but must not be transmitted on a network.
394 * Also, multicasts addressed to the loopback interface
395 * are not sent -- the above call to ip_mloopback() will
396 * loop back a copy if this host actually belongs to the
397 * destination group on the loopback interface.
398 */
399 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
400 m_freem(m);
401 goto done;
402 }
403
404 goto sendit;
405 }
406 #ifndef notdef
407 /*
408 * If the source address is not specified yet, use the address
409 * of the outoing interface. In case, keep note we did that, so
410 * if the the firewall changes the next-hop causing the output
411 * interface to change, we can fix that.
412 */
413 if (ip->ip_src.s_addr == INADDR_ANY) {
414 /* Interface may have no addresses. */
415 if (ia != NULL) {
416 ip->ip_src = IA_SIN(ia)->sin_addr;
417 src_was_INADDR_ANY = 1;
418 }
419 }
420 #endif /* notdef */
421 /*
422 * Verify that we have any chance at all of being able to queue
423 * the packet or packet fragments
424 */
425 if ((ifp->if_snd.ifq_len + ip->ip_len / ifp->if_mtu + 1) >=
426 ifp->if_snd.ifq_maxlen) {
427 error = ENOBUFS;
428 ipstat.ips_odropped++;
429 goto bad;
430 }
431
432 /*
433 * Look for broadcast address and
434 * verify user is allowed to send
435 * such a packet.
436 */
437 if (isbroadcast) {
438 if ((ifp->if_flags & IFF_BROADCAST) == 0) {
439 error = EADDRNOTAVAIL;
440 goto bad;
441 }
442 if ((flags & IP_ALLOWBROADCAST) == 0) {
443 error = EACCES;
444 goto bad;
445 }
446 /* don't allow broadcast messages to be fragmented */
447 if (ip->ip_len > ifp->if_mtu) {
448 error = EMSGSIZE;
449 goto bad;
450 }
451 if (flags & IP_SENDONES)
452 ip->ip_dst.s_addr = INADDR_BROADCAST;
453 m->m_flags |= M_BCAST;
454 } else {
455 m->m_flags &= ~M_BCAST;
456 }
457
458 sendit:
459 #ifdef IPSEC
460 /* get SP for this packet */
461 if (so == NULL)
462 sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, flags, &error);
463 else
464 sp = ipsec4_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error);
465
466 if (sp == NULL) {
467 ipsecstat.out_inval++;
468 goto bad;
469 }
470
471 error = 0;
472
473 /* check policy */
474 switch (sp->policy) {
475 case IPSEC_POLICY_DISCARD:
476 /*
477 * This packet is just discarded.
478 */
479 ipsecstat.out_polvio++;
480 goto bad;
481
482 case IPSEC_POLICY_BYPASS:
483 case IPSEC_POLICY_NONE:
484 case IPSEC_POLICY_TCP:
485 /* no need to do IPsec. */
486 goto skip_ipsec;
487
488 case IPSEC_POLICY_IPSEC:
489 if (sp->req == NULL) {
490 /* acquire a policy */
491 error = key_spdacquire(sp);
492 goto bad;
493 }
494 break;
495
496 case IPSEC_POLICY_ENTRUST:
497 default:
498 printf("ip_output: Invalid policy found. %d\n", sp->policy);
499 }
500 {
501 struct ipsec_output_state state;
502 bzero(&state, sizeof(state));
503 state.m = m;
504 if (flags & IP_ROUTETOIF) {
505 state.ro = &iproute;
506 bzero(&iproute, sizeof(iproute));
507 } else
508 state.ro = ro;
509 state.dst = (struct sockaddr *)dst;
510
511 ip->ip_sum = 0;
512
513 /*
514 * XXX
515 * delayed checksums are not currently compatible with IPsec
516 */
517 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
518 in_delayed_cksum(m);
519 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
520 }
521
522 ip->ip_len = htons(ip->ip_len);
523 ip->ip_off = htons(ip->ip_off);
524
525 error = ipsec4_output(&state, sp, flags);
526
527 m = state.m;
528 if (flags & IP_ROUTETOIF) {
529 /*
530 * if we have tunnel mode SA, we may need to ignore
531 * IP_ROUTETOIF.
532 */
533 if (state.ro != &iproute || state.ro->ro_rt != NULL) {
534 flags &= ~IP_ROUTETOIF;
535 ro = state.ro;
536 }
537 } else
538 ro = state.ro;
539 dst = (struct sockaddr_in *)state.dst;
540 if (error) {
541 /* mbuf is already reclaimed in ipsec4_output. */
542 m0 = NULL;
543 switch (error) {
544 case EHOSTUNREACH:
545 case ENETUNREACH:
546 case EMSGSIZE:
547 case ENOBUFS:
548 case ENOMEM:
549 break;
550 default:
551 printf("ip4_output (ipsec): error code %d\n", error);
552 /*fall through*/
553 case ENOENT:
554 /* don't show these error codes to the user */
555 error = 0;
556 break;
557 }
558 goto bad;
559 }
560 }
561
562 /* be sure to update variables that are affected by ipsec4_output() */
563 ip = mtod(m, struct ip *);
564 #ifdef _IP_VHL
565 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
566 #else
567 hlen = ip->ip_hl << 2;
568 #endif
569 if (ro->ro_rt == NULL) {
570 if ((flags & IP_ROUTETOIF) == 0) {
571 printf("ip_output: "
572 "can't update route after IPsec processing\n");
573 error = EHOSTUNREACH; /*XXX*/
574 goto bad;
575 }
576 } else {
577 ia = ifatoia(ro->ro_rt->rt_ifa);
578 ifp = ro->ro_rt->rt_ifp;
579 }
580
581 /* make it flipped, again. */
582 ip->ip_len = ntohs(ip->ip_len);
583 ip->ip_off = ntohs(ip->ip_off);
584 skip_ipsec:
585 #endif /*IPSEC*/
586 #ifdef FAST_IPSEC
587 /*
588 * Check the security policy (SP) for the packet and, if
589 * required, do IPsec-related processing. There are two
590 * cases here; the first time a packet is sent through
591 * it will be untagged and handled by ipsec4_checkpolicy.
592 * If the packet is resubmitted to ip_output (e.g. after
593 * AH, ESP, etc. processing), there will be a tag to bypass
594 * the lookup and related policy checking.
595 */
596 mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL);
597 s = splnet();
598 if (mtag != NULL) {
599 tdbi = (struct tdb_ident *)(mtag + 1);
600 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_OUTBOUND);
601 if (sp == NULL)
602 error = -EINVAL; /* force silent drop */
603 m_tag_delete(m, mtag);
604 } else {
605 sp = ipsec4_checkpolicy(m, IPSEC_DIR_OUTBOUND, flags,
606 &error, inp);
607 }
608 /*
609 * There are four return cases:
610 * sp != NULL apply IPsec policy
611 * sp == NULL, error == 0 no IPsec handling needed
612 * sp == NULL, error == -EINVAL discard packet w/o error
613 * sp == NULL, error != 0 discard packet, report error
614 */
615 if (sp != NULL) {
616 /* Loop detection, check if ipsec processing already done */
617 KASSERT(sp->req != NULL, ("ip_output: no ipsec request"));
618 for (mtag = m_tag_first(m); mtag != NULL;
619 mtag = m_tag_next(m, mtag)) {
620 if (mtag->m_tag_cookie != MTAG_ABI_COMPAT)
621 continue;
622 if (mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_DONE &&
623 mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED)
624 continue;
625 /*
626 * Check if policy has an SA associated with it.
627 * This can happen when an SP has yet to acquire
628 * an SA; e.g. on first reference. If it occurs,
629 * then we let ipsec4_process_packet do its thing.
630 */
631 if (sp->req->sav == NULL)
632 break;
633 tdbi = (struct tdb_ident *)(mtag + 1);
634 if (tdbi->spi == sp->req->sav->spi &&
635 tdbi->proto == sp->req->sav->sah->saidx.proto &&
636 bcmp(&tdbi->dst, &sp->req->sav->sah->saidx.dst,
637 sizeof (union sockaddr_union)) == 0) {
638 /*
639 * No IPsec processing is needed, free
640 * reference to SP.
641 *
642 * NB: null pointer to avoid free at
643 * done: below.
644 */
645 KEY_FREESP(&sp), sp = NULL;
646 splx(s);
647 goto spd_done;
648 }
649 }
650
651 /*
652 * Do delayed checksums now because we send before
653 * this is done in the normal processing path.
654 */
655 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
656 in_delayed_cksum(m);
657 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
658 }
659
660 ip->ip_len = htons(ip->ip_len);
661 ip->ip_off = htons(ip->ip_off);
662
663 /* NB: callee frees mbuf */
664 error = ipsec4_process_packet(m, sp->req, flags, 0);
665 /*
666 * Preserve KAME behaviour: ENOENT can be returned
667 * when an SA acquire is in progress. Don't propagate
668 * this to user-level; it confuses applications.
669 *
670 * XXX this will go away when the SADB is redone.
671 */
672 if (error == ENOENT)
673 error = 0;
674 splx(s);
675 goto done;
676 } else {
677 splx(s);
678
679 if (error != 0) {
680 /*
681 * Hack: -EINVAL is used to signal that a packet
682 * should be silently discarded. This is typically
683 * because we asked key management for an SA and
684 * it was delayed (e.g. kicked up to IKE).
685 */
686 if (error == -EINVAL)
687 error = 0;
688 goto bad;
689 } else {
690 /* No IPsec processing for this packet. */
691 }
692 #ifdef notyet
693 /*
694 * If deferred crypto processing is needed, check that
695 * the interface supports it.
696 */
697 mtag = m_tag_find(m, PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED, NULL);
698 if (mtag != NULL && (ifp->if_capenable & IFCAP_IPSEC) == 0) {
699 /* notify IPsec to do its own crypto */
700 ipsp_skipcrypto_unmark((struct tdb_ident *)(mtag + 1));
701 error = EHOSTUNREACH;
702 goto bad;
703 }
704 #endif
705 }
706 spd_done:
707 #endif /* FAST_IPSEC */
708 /*
709 * IpHack's section.
710 * - Xlate: translate packet's addr/port (NAT).
711 * - Firewall: deny/allow/etc.
712 * - Wrap: fake packet's addr/port <unimpl.>
713 * - Encapsulate: put it in another IP and send out. <unimp.>
714 */
715 if (fr_checkp) {
716 struct mbuf *m1 = m;
717
718 if ((error = (*fr_checkp)(ip, hlen, ifp, 1, &m1)) || !m1)
719 goto done;
720 ip = mtod(m = m1, struct ip *);
721 }
722
723 /*
724 * Check with the firewall...
725 * but not if we are already being fwd'd from a firewall.
726 */
727 if (fw_enable && IPFW_LOADED && !args.next_hop) {
728 struct sockaddr_in *old = dst;
729
730 args.m = m;
731 args.next_hop = dst;
732 args.oif = ifp;
733 off = ip_fw_chk_ptr(&args);
734 m = args.m;
735 dst = args.next_hop;
736
737 /*
738 * On return we must do the following:
739 * m == NULL -> drop the pkt (old interface, deprecated)
740 * (off & IP_FW_PORT_DENY_FLAG) -> drop the pkt (new interface)
741 * 1<=off<= 0xffff -> DIVERT
742 * (off & IP_FW_PORT_DYNT_FLAG) -> send to a DUMMYNET pipe
743 * (off & IP_FW_PORT_TEE_FLAG) -> TEE the packet
744 * dst != old -> IPFIREWALL_FORWARD
745 * off==0, dst==old -> accept
746 * If some of the above modules are not compiled in, then
747 * we should't have to check the corresponding condition
748 * (because the ipfw control socket should not accept
749 * unsupported rules), but better play safe and drop
750 * packets in case of doubt.
751 */
752 if ( (off & IP_FW_PORT_DENY_FLAG) || m == NULL) {
753 if (m)
754 m_freem(m);
755 error = EACCES;
756 goto done;
757 }
758 ip = mtod(m, struct ip *);
759 if (off == 0 && dst == old) /* common case */
760 goto pass;
761 if (DUMMYNET_LOADED && (off & IP_FW_PORT_DYNT_FLAG) != 0) {
762 /*
763 * pass the pkt to dummynet. Need to include
764 * pipe number, m, ifp, ro, dst because these are
765 * not recomputed in the next pass.
766 * All other parameters have been already used and
767 * so they are not needed anymore.
768 * XXX note: if the ifp or ro entry are deleted
769 * while a pkt is in dummynet, we are in trouble!
770 */
771 args.ro = ro;
772 args.dst = dst;
773 args.flags = flags;
774
775 error = ip_dn_io_ptr(m, off & 0xffff, DN_TO_IP_OUT,
776 &args);
777 goto done;
778 }
779 #ifdef IPDIVERT
780 if (off != 0 && (off & IP_FW_PORT_DYNT_FLAG) == 0) {
781 struct mbuf *clone = NULL;
782
783 /* Clone packet if we're doing a 'tee' */
784 if ((off & IP_FW_PORT_TEE_FLAG) != 0)
785 clone = m_dup(m, M_DONTWAIT);
786
787 /*
788 * XXX
789 * delayed checksums are not currently compatible
790 * with divert sockets.
791 */
792 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
793 in_delayed_cksum(m);
794 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
795 }
796
797 /* Restore packet header fields to original values */
798 ip->ip_len = htons(ip->ip_len);
799 ip->ip_off = htons(ip->ip_off);
800
801 /* Deliver packet to divert input routine */
802 divert_packet(m, 0, off & 0xffff, args.divert_rule);
803
804 /* If 'tee', continue with original packet */
805 if (clone != NULL) {
806 m = clone;
807 ip = mtod(m, struct ip *);
808 goto pass;
809 }
810 goto done;
811 }
812 #endif
813
814 /* IPFIREWALL_FORWARD */
815 /*
816 * Check dst to make sure it is directly reachable on the
817 * interface we previously thought it was.
818 * If it isn't (which may be likely in some situations) we have
819 * to re-route it (ie, find a route for the next-hop and the
820 * associated interface) and set them here. This is nested
821 * forwarding which in most cases is undesirable, except where
822 * such control is nigh impossible. So we do it here.
823 * And I'm babbling.
824 */
825 if (off == 0 && old != dst) { /* FORWARD, dst has changed */
826 #if 0
827 /*
828 * XXX To improve readability, this block should be
829 * changed into a function call as below:
830 */
831 error = ip_ipforward(&m, &dst, &ifp);
832 if (error)
833 goto bad;
834 if (m == NULL) /* ip_input consumed the mbuf */
835 goto done;
836 #else
837 struct in_ifaddr *ia;
838
839 /*
840 * XXX sro_fwd below is static, and a pointer
841 * to it gets passed to routines downstream.
842 * This could have surprisingly bad results in
843 * practice, because its content is overwritten
844 * by subsequent packets.
845 */
846 /* There must be a better way to do this next line... */
847 static struct route sro_fwd;
848 struct route *ro_fwd = &sro_fwd;
849
850 #if 0
851 print_ip("IPFIREWALL_FORWARD: New dst ip: ",
852 dst->sin_addr, "\n");
853 #endif
854
855 /*
856 * We need to figure out if we have been forwarded
857 * to a local socket. If so, then we should somehow
858 * "loop back" to ip_input, and get directed to the
859 * PCB as if we had received this packet. This is
860 * because it may be dificult to identify the packets
861 * you want to forward until they are being output
862 * and have selected an interface. (e.g. locally
863 * initiated packets) If we used the loopback inteface,
864 * we would not be able to control what happens
865 * as the packet runs through ip_input() as
866 * it is done through a ISR.
867 */
868 LIST_FOREACH(ia,
869 INADDR_HASH(dst->sin_addr.s_addr), ia_hash) {
870 /*
871 * If the addr to forward to is one
872 * of ours, we pretend to
873 * be the destination for this packet.
874 */
875 if (IA_SIN(ia)->sin_addr.s_addr ==
876 dst->sin_addr.s_addr)
877 break;
878 }
879 if (ia) { /* tell ip_input "dont filter" */
880 struct m_hdr tag;
881
882 tag.mh_type = MT_TAG;
883 tag.mh_flags = PACKET_TAG_IPFORWARD;
884 tag.mh_data = (caddr_t)args.next_hop;
885 tag.mh_next = m;
886
887 if (m->m_pkthdr.rcvif == NULL)
888 m->m_pkthdr.rcvif = ifunit("lo0");
889 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
890 m->m_pkthdr.csum_flags |=
891 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
892 m->m_pkthdr.csum_data = 0xffff;
893 }
894 m->m_pkthdr.csum_flags |=
895 CSUM_IP_CHECKED | CSUM_IP_VALID;
896 ip->ip_len = htons(ip->ip_len);
897 ip->ip_off = htons(ip->ip_off);
898 ip_input((struct mbuf *)&tag);
899 goto done;
900 }
901 /* Some of the logic for this was
902 * nicked from above.
903 *
904 * This rewrites the cached route in a local PCB.
905 * Is this what we want to do?
906 */
907 bcopy(dst, &ro_fwd->ro_dst, sizeof(*dst));
908
909 ro_fwd->ro_rt = 0;
910 rtalloc_ign(ro_fwd, RTF_PRCLONING);
911
912 if (ro_fwd->ro_rt == 0) {
913 ipstat.ips_noroute++;
914 error = EHOSTUNREACH;
915 goto bad;
916 }
917
918 ia = ifatoia(ro_fwd->ro_rt->rt_ifa);
919 ifp = ro_fwd->ro_rt->rt_ifp;
920 ro_fwd->ro_rt->rt_use++;
921 if (ro_fwd->ro_rt->rt_flags & RTF_GATEWAY)
922 dst = (struct sockaddr_in *)
923 ro_fwd->ro_rt->rt_gateway;
924 if (ro_fwd->ro_rt->rt_flags & RTF_HOST)
925 isbroadcast =
926 (ro_fwd->ro_rt->rt_flags & RTF_BROADCAST);
927 else
928 isbroadcast = in_broadcast(dst->sin_addr, ifp);
929 if (ro->ro_rt)
930 RTFREE(ro->ro_rt);
931 ro->ro_rt = ro_fwd->ro_rt;
932 dst = (struct sockaddr_in *)&ro_fwd->ro_dst;
933
934 #endif /* ... block to be put into a function */
935 /*
936 * If we added a default src ip earlier,
937 * which would have been gotten from the-then
938 * interface, do it again, from the new one.
939 */
940 if (src_was_INADDR_ANY)
941 ip->ip_src = IA_SIN(ia)->sin_addr;
942 goto pass ;
943 }
944
945 /*
946 * if we get here, none of the above matches, and
947 * we have to drop the pkt
948 */
949 m_freem(m);
950 error = EACCES; /* not sure this is the right error msg */
951 goto done;
952 }
953
954 pass:
955 /* 127/8 must not appear on wire - RFC1122. */
956 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
957 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
958 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
959 ipstat.ips_badaddr++;
960 error = EADDRNOTAVAIL;
961 goto bad;
962 }
963 }
964
965 m->m_pkthdr.csum_flags |= CSUM_IP;
966 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_hwassist;
967 if (sw_csum & CSUM_DELAY_DATA) {
968 in_delayed_cksum(m);
969 sw_csum &= ~CSUM_DELAY_DATA;
970 }
971 m->m_pkthdr.csum_flags &= ifp->if_hwassist;
972
973 /*
974 * If small enough for interface, or the interface will take
975 * care of the fragmentation for us, can just send directly.
976 */
977 if (ip->ip_len <= ifp->if_mtu || ifp->if_hwassist & CSUM_FRAGMENT) {
978 ip->ip_len = htons(ip->ip_len);
979 ip->ip_off = htons(ip->ip_off);
980 ip->ip_sum = 0;
981 if (sw_csum & CSUM_DELAY_IP) {
982 if (ip->ip_vhl == IP_VHL_BORING) {
983 ip->ip_sum = in_cksum_hdr(ip);
984 } else {
985 ip->ip_sum = in_cksum(m, hlen);
986 }
987 }
988
989 /* Record statistics for this interface address. */
990 if (!(flags & IP_FORWARDING) && ia) {
991 ia->ia_ifa.if_opackets++;
992 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
993 }
994
995 #ifdef IPSEC
996 /* clean ipsec history once it goes out of the node */
997 ipsec_delaux(m);
998 #endif
999
1000 #ifdef MBUF_STRESS_TEST
1001 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
1002 m = m_fragment(m, M_DONTWAIT, mbuf_frag_size);
1003 #endif
1004 error = (*ifp->if_output)(ifp, m,
1005 (struct sockaddr *)dst, ro->ro_rt);
1006 goto done;
1007 }
1008
1009 if (ip->ip_off & IP_DF) {
1010 error = EMSGSIZE;
1011 /*
1012 * This case can happen if the user changed the MTU
1013 * of an interface after enabling IP on it. Because
1014 * most netifs don't keep track of routes pointing to
1015 * them, there is no way for one to update all its
1016 * routes when the MTU is changed.
1017 */
1018 if ((ro->ro_rt->rt_flags & (RTF_UP | RTF_HOST)) &&
1019 !(ro->ro_rt->rt_rmx.rmx_locks & RTV_MTU) &&
1020 (ro->ro_rt->rt_rmx.rmx_mtu > ifp->if_mtu)) {
1021 ro->ro_rt->rt_rmx.rmx_mtu = ifp->if_mtu;
1022 }
1023 ipstat.ips_cantfrag++;
1024 goto bad;
1025 }
1026
1027 /*
1028 * Too large for interface; fragment if possible. If successful,
1029 * on return, m will point to a list of packets to be sent.
1030 */
1031 error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist, sw_csum);
1032 if (error)
1033 goto bad;
1034 for (; m; m = m0) {
1035 m0 = m->m_nextpkt;
1036 m->m_nextpkt = 0;
1037 #ifdef IPSEC
1038 /* clean ipsec history once it goes out of the node */
1039 ipsec_delaux(m);
1040 #endif
1041 if (error == 0) {
1042 /* Record statistics for this interface address. */
1043 if (ia != NULL) {
1044 ia->ia_ifa.if_opackets++;
1045 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
1046 }
1047
1048 error = (*ifp->if_output)(ifp, m,
1049 (struct sockaddr *)dst, ro->ro_rt);
1050 } else
1051 m_freem(m);
1052 }
1053
1054 if (error == 0)
1055 ipstat.ips_fragmented++;
1056
1057 done:
1058 #ifdef IPSEC
1059 if (ro == &iproute && ro->ro_rt) {
1060 RTFREE(ro->ro_rt);
1061 ro->ro_rt = NULL;
1062 }
1063 if (sp != NULL) {
1064 KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
1065 printf("DP ip_output call free SP:%p\n", sp));
1066 key_freesp(sp);
1067 }
1068 #endif
1069 #ifdef FAST_IPSEC
1070 if (ro == &iproute && ro->ro_rt) {
1071 RTFREE(ro->ro_rt);
1072 ro->ro_rt = NULL;
1073 }
1074 if (sp != NULL)
1075 KEY_FREESP(&sp);
1076 #endif
1077 return (error);
1078 bad:
1079 m_freem(m);
1080 goto done;
1081 }
1082
1083 /*
1084 * Create a chain of fragments which fit the given mtu. m_frag points to the
1085 * mbuf to be fragmented; on return it points to the chain with the fragments.
1086 * Return 0 if no error. If error, m_frag may contain a partially built
1087 * chain of fragments that should be freed by the caller.
1088 *
1089 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
1090 * sw_csum contains the delayed checksums flags (e.g., CSUM_DELAY_IP).
1091 */
1092 int
1093 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
1094 u_long if_hwassist_flags, int sw_csum)
1095 {
1096 int error = 0;
1097 int hlen = IP_VHL_HL(ip->ip_vhl) << 2;
1098 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */
1099 int off;
1100 struct mbuf *m0 = *m_frag; /* the original packet */
1101 int firstlen;
1102 struct mbuf **mnext;
1103 int nfrags;
1104
1105 if (ip->ip_off & IP_DF) { /* Fragmentation not allowed */
1106 ipstat.ips_cantfrag++;
1107 return EMSGSIZE;
1108 }
1109
1110 /*
1111 * Must be able to put at least 8 bytes per fragment.
1112 */
1113 if (len < 8)
1114 return EMSGSIZE;
1115
1116 /*
1117 * If the interface will not calculate checksums on
1118 * fragmented packets, then do it here.
1119 */
1120 if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA &&
1121 (if_hwassist_flags & CSUM_IP_FRAGS) == 0) {
1122 in_delayed_cksum(m0);
1123 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1124 }
1125
1126 if (len > PAGE_SIZE) {
1127 /*
1128 * Fragment large datagrams such that each segment
1129 * contains a multiple of PAGE_SIZE amount of data,
1130 * plus headers. This enables a receiver to perform
1131 * page-flipping zero-copy optimizations.
1132 *
1133 * XXX When does this help given that sender and receiver
1134 * could have different page sizes, and also mtu could
1135 * be less than the receiver's page size ?
1136 */
1137 int newlen;
1138 struct mbuf *m;
1139
1140 for (m = m0, off = 0; m && (off+m->m_len) <= mtu; m = m->m_next)
1141 off += m->m_len;
1142
1143 /*
1144 * firstlen (off - hlen) must be aligned on an
1145 * 8-byte boundary
1146 */
1147 if (off < hlen)
1148 goto smart_frag_failure;
1149 off = ((off - hlen) & ~7) + hlen;
1150 newlen = (~PAGE_MASK) & mtu;
1151 if ((newlen + sizeof (struct ip)) > mtu) {
1152 /* we failed, go back the default */
1153 smart_frag_failure:
1154 newlen = len;
1155 off = hlen + len;
1156 }
1157 len = newlen;
1158
1159 } else {
1160 off = hlen + len;
1161 }
1162
1163 firstlen = off - hlen;
1164 mnext = &m0->m_nextpkt; /* pointer to next packet */
1165
1166 /*
1167 * Loop through length of segment after first fragment,
1168 * make new header and copy data of each part and link onto chain.
1169 * Here, m0 is the original packet, m is the fragment being created.
1170 * The fragments are linked off the m_nextpkt of the original
1171 * packet, which after processing serves as the first fragment.
1172 */
1173 for (nfrags = 1; off < ip->ip_len; off += len, nfrags++) {
1174 struct ip *mhip; /* ip header on the fragment */
1175 struct mbuf *m;
1176 int mhlen = sizeof (struct ip);
1177
1178 MGETHDR(m, M_DONTWAIT, MT_HEADER);
1179 if (m == 0) {
1180 error = ENOBUFS;
1181 ipstat.ips_odropped++;
1182 goto done;
1183 }
1184 m->m_flags |= (m0->m_flags & M_MCAST) | M_FRAG;
1185 /*
1186 * In the first mbuf, leave room for the link header, then
1187 * copy the original IP header including options. The payload
1188 * goes into an additional mbuf chain returned by m_copy().
1189 */
1190 m->m_data += max_linkhdr;
1191 mhip = mtod(m, struct ip *);
1192 *mhip = *ip;
1193 if (hlen > sizeof (struct ip)) {
1194 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
1195 mhip->ip_vhl = IP_MAKE_VHL(IPVERSION, mhlen >> 2);
1196 }
1197 m->m_len = mhlen;
1198 /* XXX do we need to add ip->ip_off below ? */
1199 mhip->ip_off = ((off - hlen) >> 3) + ip->ip_off;
1200 if (off + len >= ip->ip_len) { /* last fragment */
1201 len = ip->ip_len - off;
1202 m->m_flags |= M_LASTFRAG;
1203 } else
1204 mhip->ip_off |= IP_MF;
1205 mhip->ip_len = htons((u_short)(len + mhlen));
1206 m->m_next = m_copy(m0, off, len);
1207 if (m->m_next == 0) { /* copy failed */
1208 m_free(m);
1209 error = ENOBUFS; /* ??? */
1210 ipstat.ips_odropped++;
1211 goto done;
1212 }
1213 m->m_pkthdr.len = mhlen + len;
1214 m->m_pkthdr.rcvif = (struct ifnet *)0;
1215 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags;
1216 mhip->ip_off = htons(mhip->ip_off);
1217 mhip->ip_sum = 0;
1218 if (sw_csum & CSUM_DELAY_IP)
1219 mhip->ip_sum = in_cksum(m, mhlen);
1220 *mnext = m;
1221 mnext = &m->m_nextpkt;
1222 }
1223 ipstat.ips_ofragments += nfrags;
1224
1225 /* set first marker for fragment chain */
1226 m0->m_flags |= M_FIRSTFRAG | M_FRAG;
1227 m0->m_pkthdr.csum_data = nfrags;
1228
1229 /*
1230 * Update first fragment by trimming what's been copied out
1231 * and updating header.
1232 */
1233 m_adj(m0, hlen + firstlen - ip->ip_len);
1234 m0->m_pkthdr.len = hlen + firstlen;
1235 ip->ip_len = htons((u_short)m0->m_pkthdr.len);
1236 ip->ip_off |= IP_MF;
1237 ip->ip_off = htons(ip->ip_off);
1238 ip->ip_sum = 0;
1239 if (sw_csum & CSUM_DELAY_IP)
1240 ip->ip_sum = in_cksum(m0, hlen);
1241
1242 done:
1243 *m_frag = m0;
1244 return error;
1245 }
1246
1247 void
1248 in_delayed_cksum(struct mbuf *m)
1249 {
1250 struct ip *ip;
1251 u_short csum, offset;
1252
1253 ip = mtod(m, struct ip *);
1254 offset = IP_VHL_HL(ip->ip_vhl) << 2 ;
1255 csum = in_cksum_skip(m, ip->ip_len, offset);
1256 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0)
1257 csum = 0xffff;
1258 offset += m->m_pkthdr.csum_data; /* checksum offset */
1259
1260 if (offset + sizeof(u_short) > m->m_len) {
1261 printf("delayed m_pullup, m->len: %d off: %d p: %d\n",
1262 m->m_len, offset, ip->ip_p);
1263 /*
1264 * XXX
1265 * this shouldn't happen, but if it does, the
1266 * correct behavior may be to insert the checksum
1267 * in the existing chain instead of rearranging it.
1268 */
1269 m = m_pullup(m, offset + sizeof(u_short));
1270 }
1271 *(u_short *)(m->m_data + offset) = csum;
1272 }
1273
1274 /*
1275 * Insert IP options into preformed packet.
1276 * Adjust IP destination as required for IP source routing,
1277 * as indicated by a non-zero in_addr at the start of the options.
1278 *
1279 * XXX This routine assumes that the packet has no options in place.
1280 */
1281 static struct mbuf *
1282 ip_insertoptions(m, opt, phlen)
1283 register struct mbuf *m;
1284 struct mbuf *opt;
1285 int *phlen;
1286 {
1287 register struct ipoption *p = mtod(opt, struct ipoption *);
1288 struct mbuf *n;
1289 register struct ip *ip = mtod(m, struct ip *);
1290 unsigned optlen;
1291
1292 optlen = opt->m_len - sizeof(p->ipopt_dst);
1293 if (optlen + (u_short)ip->ip_len > IP_MAXPACKET) {
1294 *phlen = 0;
1295 return (m); /* XXX should fail */
1296 }
1297 if (p->ipopt_dst.s_addr)
1298 ip->ip_dst = p->ipopt_dst;
1299 if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) {
1300 MGETHDR(n, M_DONTWAIT, MT_HEADER);
1301 if (n == 0) {
1302 *phlen = 0;
1303 return (m);
1304 }
1305 n->m_pkthdr.rcvif = (struct ifnet *)0;
1306 n->m_pkthdr.len = m->m_pkthdr.len + optlen;
1307 m->m_len -= sizeof(struct ip);
1308 m->m_data += sizeof(struct ip);
1309 n->m_next = m;
1310 m = n;
1311 m->m_len = optlen + sizeof(struct ip);
1312 m->m_data += max_linkhdr;
1313 (void)memcpy(mtod(m, void *), ip, sizeof(struct ip));
1314 } else {
1315 m->m_data -= optlen;
1316 m->m_len += optlen;
1317 m->m_pkthdr.len += optlen;
1318 ovbcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
1319 }
1320 ip = mtod(m, struct ip *);
1321 bcopy(p->ipopt_list, ip + 1, optlen);
1322 *phlen = sizeof(struct ip) + optlen;
1323 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, *phlen >> 2);
1324 ip->ip_len += optlen;
1325 return (m);
1326 }
1327
1328 /*
1329 * Copy options from ip to jp,
1330 * omitting those not copied during fragmentation.
1331 */
1332 int
1333 ip_optcopy(ip, jp)
1334 struct ip *ip, *jp;
1335 {
1336 register u_char *cp, *dp;
1337 int opt, optlen, cnt;
1338
1339 cp = (u_char *)(ip + 1);
1340 dp = (u_char *)(jp + 1);
1341 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip);
1342 for (; cnt > 0; cnt -= optlen, cp += optlen) {
1343 opt = cp[0];
1344 if (opt == IPOPT_EOL)
1345 break;
1346 if (opt == IPOPT_NOP) {
1347 /* Preserve for IP mcast tunnel's LSRR alignment. */
1348 *dp++ = IPOPT_NOP;
1349 optlen = 1;
1350 continue;
1351 }
1352
1353 KASSERT(cnt >= IPOPT_OLEN + sizeof(*cp),
1354 ("ip_optcopy: malformed ipv4 option"));
1355 optlen = cp[IPOPT_OLEN];
1356 KASSERT(optlen >= IPOPT_OLEN + sizeof(*cp) && optlen <= cnt,
1357 ("ip_optcopy: malformed ipv4 option"));
1358
1359 /* bogus lengths should have been caught by ip_dooptions */
1360 if (optlen > cnt)
1361 optlen = cnt;
1362 if (IPOPT_COPIED(opt)) {
1363 bcopy(cp, dp, optlen);
1364 dp += optlen;
1365 }
1366 }
1367 for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++)
1368 *dp++ = IPOPT_EOL;
1369 return (optlen);
1370 }
1371
1372 /*
1373 * IP socket option processing.
1374 */
1375 int
1376 ip_ctloutput(so, sopt)
1377 struct socket *so;
1378 struct sockopt *sopt;
1379 {
1380 struct inpcb *inp = sotoinpcb(so);
1381 int error, optval;
1382
1383 error = optval = 0;
1384 if (sopt->sopt_level != IPPROTO_IP) {
1385 return (EINVAL);
1386 }
1387
1388 switch (sopt->sopt_dir) {
1389 case SOPT_SET:
1390 switch (sopt->sopt_name) {
1391 case IP_OPTIONS:
1392 #ifdef notyet
1393 case IP_RETOPTS:
1394 #endif
1395 {
1396 struct mbuf *m;
1397 if (sopt->sopt_valsize > MLEN) {
1398 error = EMSGSIZE;
1399 break;
1400 }
1401 MGET(m, sopt->sopt_p ? M_WAIT : M_DONTWAIT, MT_HEADER);
1402 if (m == 0) {
1403 error = ENOBUFS;
1404 break;
1405 }
1406 m->m_len = sopt->sopt_valsize;
1407 error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
1408 m->m_len);
1409
1410 return (ip_pcbopts(sopt->sopt_name, &inp->inp_options,
1411 m));
1412 }
1413
1414 case IP_TOS:
1415 case IP_TTL:
1416 case IP_RECVOPTS:
1417 case IP_RECVRETOPTS:
1418 case IP_RECVDSTADDR:
1419 case IP_RECVIF:
1420 case IP_FAITH:
1421 case IP_ONESBCAST:
1422 error = sooptcopyin(sopt, &optval, sizeof optval,
1423 sizeof optval);
1424 if (error)
1425 break;
1426
1427 switch (sopt->sopt_name) {
1428 case IP_TOS:
1429 inp->inp_ip_tos = optval;
1430 break;
1431
1432 case IP_TTL:
1433 inp->inp_ip_ttl = optval;
1434 break;
1435 #define OPTSET(bit) \
1436 if (optval) \
1437 inp->inp_flags |= bit; \
1438 else \
1439 inp->inp_flags &= ~bit;
1440
1441 case IP_RECVOPTS:
1442 OPTSET(INP_RECVOPTS);
1443 break;
1444
1445 case IP_RECVRETOPTS:
1446 OPTSET(INP_RECVRETOPTS);
1447 break;
1448
1449 case IP_RECVDSTADDR:
1450 OPTSET(INP_RECVDSTADDR);
1451 break;
1452
1453 case IP_RECVIF:
1454 OPTSET(INP_RECVIF);
1455 break;
1456
1457 case IP_FAITH:
1458 OPTSET(INP_FAITH);
1459 break;
1460
1461 case IP_ONESBCAST:
1462 OPTSET(INP_ONESBCAST);
1463 break;
1464 }
1465 break;
1466 #undef OPTSET
1467
1468 case IP_MULTICAST_IF:
1469 case IP_MULTICAST_VIF:
1470 case IP_MULTICAST_TTL:
1471 case IP_MULTICAST_LOOP:
1472 case IP_ADD_MEMBERSHIP:
1473 case IP_DROP_MEMBERSHIP:
1474 error = ip_setmoptions(sopt, &inp->inp_moptions);
1475 break;
1476
1477 case IP_PORTRANGE:
1478 error = sooptcopyin(sopt, &optval, sizeof optval,
1479 sizeof optval);
1480 if (error)
1481 break;
1482
1483 switch (optval) {
1484 case IP_PORTRANGE_DEFAULT:
1485 inp->inp_flags &= ~(INP_LOWPORT);
1486 inp->inp_flags &= ~(INP_HIGHPORT);
1487 break;
1488
1489 case IP_PORTRANGE_HIGH:
1490 inp->inp_flags &= ~(INP_LOWPORT);
1491 inp->inp_flags |= INP_HIGHPORT;
1492 break;
1493
1494 case IP_PORTRANGE_LOW:
1495 inp->inp_flags &= ~(INP_HIGHPORT);
1496 inp->inp_flags |= INP_LOWPORT;
1497 break;
1498
1499 default:
1500 error = EINVAL;
1501 break;
1502 }
1503 break;
1504
1505 #if defined(IPSEC) || defined(FAST_IPSEC)
1506 case IP_IPSEC_POLICY:
1507 {
1508 caddr_t req;
1509 size_t len = 0;
1510 int priv;
1511 struct mbuf *m;
1512 int optname;
1513
1514 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1515 break;
1516 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1517 break;
1518 priv = (sopt->sopt_p != NULL &&
1519 suser(sopt->sopt_p) != 0) ? 0 : 1;
1520 req = mtod(m, caddr_t);
1521 len = m->m_len;
1522 optname = sopt->sopt_name;
1523 error = ipsec4_set_policy(inp, optname, req, len, priv);
1524 m_freem(m);
1525 break;
1526 }
1527 #endif /*IPSEC*/
1528
1529 default:
1530 error = ENOPROTOOPT;
1531 break;
1532 }
1533 break;
1534
1535 case SOPT_GET:
1536 switch (sopt->sopt_name) {
1537 case IP_OPTIONS:
1538 case IP_RETOPTS:
1539 if (inp->inp_options)
1540 error = sooptcopyout(sopt,
1541 mtod(inp->inp_options,
1542 char *),
1543 inp->inp_options->m_len);
1544 else
1545 sopt->sopt_valsize = 0;
1546 break;
1547
1548 case IP_TOS:
1549 case IP_TTL:
1550 case IP_RECVOPTS:
1551 case IP_RECVRETOPTS:
1552 case IP_RECVDSTADDR:
1553 case IP_RECVIF:
1554 case IP_PORTRANGE:
1555 case IP_FAITH:
1556 case IP_ONESBCAST:
1557 switch (sopt->sopt_name) {
1558
1559 case IP_TOS:
1560 optval = inp->inp_ip_tos;
1561 break;
1562
1563 case IP_TTL:
1564 optval = inp->inp_ip_ttl;
1565 break;
1566
1567 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0)
1568
1569 case IP_RECVOPTS:
1570 optval = OPTBIT(INP_RECVOPTS);
1571 break;
1572
1573 case IP_RECVRETOPTS:
1574 optval = OPTBIT(INP_RECVRETOPTS);
1575 break;
1576
1577 case IP_RECVDSTADDR:
1578 optval = OPTBIT(INP_RECVDSTADDR);
1579 break;
1580
1581 case IP_RECVIF:
1582 optval = OPTBIT(INP_RECVIF);
1583 break;
1584
1585 case IP_PORTRANGE:
1586 if (inp->inp_flags & INP_HIGHPORT)
1587 optval = IP_PORTRANGE_HIGH;
1588 else if (inp->inp_flags & INP_LOWPORT)
1589 optval = IP_PORTRANGE_LOW;
1590 else
1591 optval = 0;
1592 break;
1593
1594 case IP_FAITH:
1595 optval = OPTBIT(INP_FAITH);
1596 break;
1597
1598 case IP_ONESBCAST:
1599 optval = OPTBIT(INP_ONESBCAST);
1600 break;
1601 }
1602 error = sooptcopyout(sopt, &optval, sizeof optval);
1603 break;
1604
1605 case IP_MULTICAST_IF:
1606 case IP_MULTICAST_VIF:
1607 case IP_MULTICAST_TTL:
1608 case IP_MULTICAST_LOOP:
1609 case IP_ADD_MEMBERSHIP:
1610 case IP_DROP_MEMBERSHIP:
1611 error = ip_getmoptions(sopt, inp->inp_moptions);
1612 break;
1613
1614 #if defined(IPSEC) || defined(FAST_IPSEC)
1615 case IP_IPSEC_POLICY:
1616 {
1617 struct mbuf *m = NULL;
1618 caddr_t req = NULL;
1619 size_t len = 0;
1620
1621 if (m != 0) {
1622 req = mtod(m, caddr_t);
1623 len = m->m_len;
1624 }
1625 error = ipsec4_get_policy(sotoinpcb(so), req, len, &m);
1626 if (error == 0)
1627 error = soopt_mcopyout(sopt, m); /* XXX */
1628 if (error == 0)
1629 m_freem(m);
1630 break;
1631 }
1632 #endif /*IPSEC*/
1633
1634 default:
1635 error = ENOPROTOOPT;
1636 break;
1637 }
1638 break;
1639 }
1640 return (error);
1641 }
1642
1643 /*
1644 * Set up IP options in pcb for insertion in output packets.
1645 * Store in mbuf with pointer in pcbopt, adding pseudo-option
1646 * with destination address if source routed.
1647 */
1648 static int
1649 ip_pcbopts(optname, pcbopt, m)
1650 int optname;
1651 struct mbuf **pcbopt;
1652 register struct mbuf *m;
1653 {
1654 register int cnt, optlen;
1655 register u_char *cp;
1656 u_char opt;
1657
1658 /* turn off any old options */
1659 if (*pcbopt)
1660 (void)m_free(*pcbopt);
1661 *pcbopt = 0;
1662 if (m == (struct mbuf *)0 || m->m_len == 0) {
1663 /*
1664 * Only turning off any previous options.
1665 */
1666 if (m)
1667 (void)m_free(m);
1668 return (0);
1669 }
1670
1671 if (m->m_len % sizeof(int32_t))
1672 goto bad;
1673 /*
1674 * IP first-hop destination address will be stored before
1675 * actual options; move other options back
1676 * and clear it when none present.
1677 */
1678 if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN])
1679 goto bad;
1680 cnt = m->m_len;
1681 m->m_len += sizeof(struct in_addr);
1682 cp = mtod(m, u_char *) + sizeof(struct in_addr);
1683 ovbcopy(mtod(m, caddr_t), (caddr_t)cp, (unsigned)cnt);
1684 bzero(mtod(m, caddr_t), sizeof(struct in_addr));
1685
1686 for (; cnt > 0; cnt -= optlen, cp += optlen) {
1687 opt = cp[IPOPT_OPTVAL];
1688 if (opt == IPOPT_EOL)
1689 break;
1690 if (opt == IPOPT_NOP)
1691 optlen = 1;
1692 else {
1693 if (cnt < IPOPT_OLEN + sizeof(*cp))
1694 goto bad;
1695 optlen = cp[IPOPT_OLEN];
1696 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt)
1697 goto bad;
1698 }
1699 switch (opt) {
1700
1701 default:
1702 break;
1703
1704 case IPOPT_LSRR:
1705 case IPOPT_SSRR:
1706 /*
1707 * user process specifies route as:
1708 * ->A->B->C->D
1709 * D must be our final destination (but we can't
1710 * check that since we may not have connected yet).
1711 * A is first hop destination, which doesn't appear in
1712 * actual IP option, but is stored before the options.
1713 */
1714 if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr))
1715 goto bad;
1716 m->m_len -= sizeof(struct in_addr);
1717 cnt -= sizeof(struct in_addr);
1718 optlen -= sizeof(struct in_addr);
1719 cp[IPOPT_OLEN] = optlen;
1720 /*
1721 * Move first hop before start of options.
1722 */
1723 bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t),
1724 sizeof(struct in_addr));
1725 /*
1726 * Then copy rest of options back
1727 * to close up the deleted entry.
1728 */
1729 ovbcopy((caddr_t)(&cp[IPOPT_OFFSET+1] +
1730 sizeof(struct in_addr)),
1731 (caddr_t)&cp[IPOPT_OFFSET+1],
1732 (unsigned)cnt - (IPOPT_MINOFF - 1));
1733 break;
1734 }
1735 }
1736 if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr))
1737 goto bad;
1738 *pcbopt = m;
1739 return (0);
1740
1741 bad:
1742 (void)m_free(m);
1743 return (EINVAL);
1744 }
1745
1746 /*
1747 * XXX
1748 * The whole multicast option thing needs to be re-thought.
1749 * Several of these options are equally applicable to non-multicast
1750 * transmission, and one (IP_MULTICAST_TTL) totally duplicates a
1751 * standard option (IP_TTL).
1752 */
1753
1754 /*
1755 * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index.
1756 */
1757 static struct ifnet *
1758 ip_multicast_if(a, ifindexp)
1759 struct in_addr *a;
1760 int *ifindexp;
1761 {
1762 int ifindex;
1763 struct ifnet *ifp;
1764
1765 if (ifindexp)
1766 *ifindexp = 0;
1767 if (ntohl(a->s_addr) >> 24 == 0) {
1768 ifindex = ntohl(a->s_addr) & 0xffffff;
1769 if (ifindex < 0 || if_index < ifindex)
1770 return NULL;
1771 ifp = ifindex2ifnet[ifindex];
1772 if (ifindexp)
1773 *ifindexp = ifindex;
1774 } else {
1775 INADDR_TO_IFP(*a, ifp);
1776 }
1777 return ifp;
1778 }
1779
1780 /*
1781 * Set the IP multicast options in response to user setsockopt().
1782 */
1783 static int
1784 ip_setmoptions(sopt, imop)
1785 struct sockopt *sopt;
1786 struct ip_moptions **imop;
1787 {
1788 int error = 0;
1789 int i;
1790 struct in_addr addr;
1791 struct ip_mreq mreq;
1792 struct ifnet *ifp;
1793 struct ip_moptions *imo = *imop;
1794 struct route ro;
1795 struct sockaddr_in *dst;
1796 int ifindex;
1797 int s;
1798
1799 if (imo == NULL) {
1800 /*
1801 * No multicast option buffer attached to the pcb;
1802 * allocate one and initialize to default values.
1803 */
1804 imo = (struct ip_moptions*)malloc(sizeof(*imo), M_IPMOPTS,
1805 M_WAITOK);
1806
1807 if (imo == NULL)
1808 return (ENOBUFS);
1809 *imop = imo;
1810 imo->imo_multicast_ifp = NULL;
1811 imo->imo_multicast_addr.s_addr = INADDR_ANY;
1812 imo->imo_multicast_vif = -1;
1813 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
1814 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP;
1815 imo->imo_num_memberships = 0;
1816 }
1817
1818 switch (sopt->sopt_name) {
1819 /* store an index number for the vif you wanna use in the send */
1820 case IP_MULTICAST_VIF:
1821 if (legal_vif_num == 0) {
1822 error = EOPNOTSUPP;
1823 break;
1824 }
1825 error = sooptcopyin(sopt, &i, sizeof i, sizeof i);
1826 if (error)
1827 break;
1828 if (!legal_vif_num(i) && (i != -1)) {
1829 error = EINVAL;
1830 break;
1831 }
1832 imo->imo_multicast_vif = i;
1833 break;
1834
1835 case IP_MULTICAST_IF:
1836 /*
1837 * Select the interface for outgoing multicast packets.
1838 */
1839 error = sooptcopyin(sopt, &addr, sizeof addr, sizeof addr);
1840 if (error)
1841 break;
1842 /*
1843 * INADDR_ANY is used to remove a previous selection.
1844 * When no interface is selected, a default one is
1845 * chosen every time a multicast packet is sent.
1846 */
1847 if (addr.s_addr == INADDR_ANY) {
1848 imo->imo_multicast_ifp = NULL;
1849 break;
1850 }
1851 /*
1852 * The selected interface is identified by its local
1853 * IP address. Find the interface and confirm that
1854 * it supports multicasting.
1855 */
1856 s = splimp();
1857 ifp = ip_multicast_if(&addr, &ifindex);
1858 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1859 splx(s);
1860 error = EADDRNOTAVAIL;
1861 break;
1862 }
1863 imo->imo_multicast_ifp = ifp;
1864 if (ifindex)
1865 imo->imo_multicast_addr = addr;
1866 else
1867 imo->imo_multicast_addr.s_addr = INADDR_ANY;
1868 splx(s);
1869 break;
1870
1871 case IP_MULTICAST_TTL:
1872 /*
1873 * Set the IP time-to-live for outgoing multicast packets.
1874 * The original multicast API required a char argument,
1875 * which is inconsistent with the rest of the socket API.
1876 * We allow either a char or an int.
1877 */
1878 if (sopt->sopt_valsize == 1) {
1879 u_char ttl;
1880 error = sooptcopyin(sopt, &ttl, 1, 1);
1881 if (error)
1882 break;
1883 imo->imo_multicast_ttl = ttl;
1884 } else {
1885 u_int ttl;
1886 error = sooptcopyin(sopt, &ttl, sizeof ttl,
1887 sizeof ttl);
1888 if (error)
1889 break;
1890 if (ttl > 255)
1891 error = EINVAL;
1892 else
1893 imo->imo_multicast_ttl = ttl;
1894 }
1895 break;
1896
1897 case IP_MULTICAST_LOOP:
1898 /*
1899 * Set the loopback flag for outgoing multicast packets.
1900 * Must be zero or one. The original multicast API required a
1901 * char argument, which is inconsistent with the rest
1902 * of the socket API. We allow either a char or an int.
1903 */
1904 if (sopt->sopt_valsize == 1) {
1905 u_char loop;
1906 error = sooptcopyin(sopt, &loop, 1, 1);
1907 if (error)
1908 break;
1909 imo->imo_multicast_loop = !!loop;
1910 } else {
1911 u_int loop;
1912 error = sooptcopyin(sopt, &loop, sizeof loop,
1913 sizeof loop);
1914 if (error)
1915 break;
1916 imo->imo_multicast_loop = !!loop;
1917 }
1918 break;
1919
1920 case IP_ADD_MEMBERSHIP:
1921 /*
1922 * Add a multicast group membership.
1923 * Group must be a valid IP multicast address.
1924 */
1925 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq);
1926 if (error)
1927 break;
1928
1929 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) {
1930 error = EINVAL;
1931 break;
1932 }
1933 s = splimp();
1934 /*
1935 * If no interface address was provided, use the interface of
1936 * the route to the given multicast address.
1937 */
1938 if (mreq.imr_interface.s_addr == INADDR_ANY) {
1939 bzero((caddr_t)&ro, sizeof(ro));
1940 dst = (struct sockaddr_in *)&ro.ro_dst;
1941 dst->sin_len = sizeof(*dst);
1942 dst->sin_family = AF_INET;
1943 dst->sin_addr = mreq.imr_multiaddr;
1944 rtalloc(&ro);
1945 if (ro.ro_rt == NULL) {
1946 error = EADDRNOTAVAIL;
1947 splx(s);
1948 break;
1949 }
1950 ifp = ro.ro_rt->rt_ifp;
1951 rtfree(ro.ro_rt);
1952 }
1953 else {
1954 ifp = ip_multicast_if(&mreq.imr_interface, NULL);
1955 }
1956
1957 /*
1958 * See if we found an interface, and confirm that it
1959 * supports multicast.
1960 */
1961 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1962 error = EADDRNOTAVAIL;
1963 splx(s);
1964 break;
1965 }
1966 /*
1967 * See if the membership already exists or if all the
1968 * membership slots are full.
1969 */
1970 for (i = 0; i < imo->imo_num_memberships; ++i) {
1971 if (imo->imo_membership[i]->inm_ifp == ifp &&
1972 imo->imo_membership[i]->inm_addr.s_addr
1973 == mreq.imr_multiaddr.s_addr)
1974 break;
1975 }
1976 if (i < imo->imo_num_memberships) {
1977 error = EADDRINUSE;
1978 splx(s);
1979 break;
1980 }
1981 if (i == IP_MAX_MEMBERSHIPS) {
1982 error = ETOOMANYREFS;
1983 splx(s);
1984 break;
1985 }
1986 /*
1987 * Everything looks good; add a new record to the multicast
1988 * address list for the given interface.
1989 */
1990 if ((imo->imo_membership[i] =
1991 in_addmulti(&mreq.imr_multiaddr, ifp)) == NULL) {
1992 error = ENOBUFS;
1993 splx(s);
1994 break;
1995 }
1996 ++imo->imo_num_memberships;
1997 splx(s);
1998 break;
1999
2000 case IP_DROP_MEMBERSHIP:
2001 /*
2002 * Drop a multicast group membership.
2003 * Group must be a valid IP multicast address.
2004 */
2005 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq);
2006 if (error)
2007 break;
2008
2009 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) {
2010 error = EINVAL;
2011 break;
2012 }
2013
2014 s = splimp();
2015 /*
2016 * If an interface address was specified, get a pointer
2017 * to its ifnet structure.
2018 */
2019 if (mreq.imr_interface.s_addr == INADDR_ANY)
2020 ifp = NULL;
2021 else {
2022 ifp = ip_multicast_if(&mreq.imr_interface, NULL);
2023 if (ifp == NULL) {
2024 error = EADDRNOTAVAIL;
2025 splx(s);
2026 break;
2027 }
2028 }
2029 /*
2030 * Find the membership in the membership array.
2031 */
2032 for (i = 0; i < imo->imo_num_memberships; ++i) {
2033 if ((ifp == NULL ||
2034 imo->imo_membership[i]->inm_ifp == ifp) &&
2035 imo->imo_membership[i]->inm_addr.s_addr ==
2036 mreq.imr_multiaddr.s_addr)
2037 break;
2038 }
2039 if (i == imo->imo_num_memberships) {
2040 error = EADDRNOTAVAIL;
2041 splx(s);
2042 break;
2043 }
2044 /*
2045 * Give up the multicast address record to which the
2046 * membership points.
2047 */
2048 in_delmulti(imo->imo_membership[i]);
2049 /*
2050 * Remove the gap in the membership array.
2051 */
2052 for (++i; i < imo->imo_num_memberships; ++i)
2053 imo->imo_membership[i-1] = imo->imo_membership[i];
2054 --imo->imo_num_memberships;
2055 splx(s);
2056 break;
2057
2058 default:
2059 error = EOPNOTSUPP;
2060 break;
2061 }
2062
2063 /*
2064 * If all options have default values, no need to keep the mbuf.
2065 */
2066 if (imo->imo_multicast_ifp == NULL &&
2067 imo->imo_multicast_vif == -1 &&
2068 imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL &&
2069 imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP &&
2070 imo->imo_num_memberships == 0) {
2071 free(*imop, M_IPMOPTS);
2072 *imop = NULL;
2073 }
2074
2075 return (error);
2076 }
2077
2078 /*
2079 * Return the IP multicast options in response to user getsockopt().
2080 */
2081 static int
2082 ip_getmoptions(sopt, imo)
2083 struct sockopt *sopt;
2084 register struct ip_moptions *imo;
2085 {
2086 struct in_addr addr;
2087 struct in_ifaddr *ia;
2088 int error, optval;
2089 u_char coptval;
2090
2091 error = 0;
2092 switch (sopt->sopt_name) {
2093 case IP_MULTICAST_VIF:
2094 if (imo != NULL)
2095 optval = imo->imo_multicast_vif;
2096 else
2097 optval = -1;
2098 error = sooptcopyout(sopt, &optval, sizeof optval);
2099 break;
2100
2101 case IP_MULTICAST_IF:
2102 if (imo == NULL || imo->imo_multicast_ifp == NULL)
2103 addr.s_addr = INADDR_ANY;
2104 else if (imo->imo_multicast_addr.s_addr) {
2105 /* return the value user has set */
2106 addr = imo->imo_multicast_addr;
2107 } else {
2108 IFP_TO_IA(imo->imo_multicast_ifp, ia);
2109 addr.s_addr = (ia == NULL) ? INADDR_ANY
2110 : IA_SIN(ia)->sin_addr.s_addr;
2111 }
2112 error = sooptcopyout(sopt, &addr, sizeof addr);
2113 break;
2114
2115 case IP_MULTICAST_TTL:
2116 if (imo == 0)
2117 optval = coptval = IP_DEFAULT_MULTICAST_TTL;
2118 else
2119 optval = coptval = imo->imo_multicast_ttl;
2120 if (sopt->sopt_valsize == 1)
2121 error = sooptcopyout(sopt, &coptval, 1);
2122 else
2123 error = sooptcopyout(sopt, &optval, sizeof optval);
2124 break;
2125
2126 case IP_MULTICAST_LOOP:
2127 if (imo == 0)
2128 optval = coptval = IP_DEFAULT_MULTICAST_LOOP;
2129 else
2130 optval = coptval = imo->imo_multicast_loop;
2131 if (sopt->sopt_valsize == 1)
2132 error = sooptcopyout(sopt, &coptval, 1);
2133 else
2134 error = sooptcopyout(sopt, &optval, sizeof optval);
2135 break;
2136
2137 default:
2138 error = ENOPROTOOPT;
2139 break;
2140 }
2141 return (error);
2142 }
2143
2144 /*
2145 * Discard the IP multicast options.
2146 */
2147 void
2148 ip_freemoptions(imo)
2149 register struct ip_moptions *imo;
2150 {
2151 register int i;
2152
2153 if (imo != NULL) {
2154 for (i = 0; i < imo->imo_num_memberships; ++i)
2155 in_delmulti(imo->imo_membership[i]);
2156 free(imo, M_IPMOPTS);
2157 }
2158 }
2159
2160 /*
2161 * Routine called from ip_output() to loop back a copy of an IP multicast
2162 * packet to the input queue of a specified interface. Note that this
2163 * calls the output routine of the loopback "driver", but with an interface
2164 * pointer that might NOT be a loopback interface -- evil, but easier than
2165 * replicating that code here.
2166 */
2167 static void
2168 ip_mloopback(ifp, m, dst, hlen)
2169 struct ifnet *ifp;
2170 register struct mbuf *m;
2171 register struct sockaddr_in *dst;
2172 int hlen;
2173 {
2174 register struct ip *ip;
2175 struct mbuf *copym;
2176
2177 copym = m_copy(m, 0, M_COPYALL);
2178 if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen))
2179 copym = m_pullup(copym, hlen);
2180 if (copym != NULL) {
2181 /* If needed, compute the checksum and mark it as valid. */
2182 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
2183 in_delayed_cksum(copym);
2184 copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
2185 copym->m_pkthdr.csum_flags |=
2186 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
2187 copym->m_pkthdr.csum_data = 0xffff;
2188 }
2189 /*
2190 * We don't bother to fragment if the IP length is greater
2191 * than the interface's MTU. Can this possibly matter?
2192 */
2193 ip = mtod(copym, struct ip *);
2194 ip->ip_len = htons(ip->ip_len);
2195 ip->ip_off = htons(ip->ip_off);
2196 ip->ip_sum = 0;
2197 if (ip->ip_vhl == IP_VHL_BORING) {
2198 ip->ip_sum = in_cksum_hdr(ip);
2199 } else {
2200 ip->ip_sum = in_cksum(copym, hlen);
2201 }
2202 /*
2203 * NB:
2204 * It's not clear whether there are any lingering
2205 * reentrancy problems in other areas which might
2206 * be exposed by using ip_input directly (in
2207 * particular, everything which modifies the packet
2208 * in-place). Yet another option is using the
2209 * protosw directly to deliver the looped back
2210 * packet. For the moment, we'll err on the side
2211 * of safety by using if_simloop().
2212 */
2213 #if 1 /* XXX */
2214 if (dst->sin_family != AF_INET) {
2215 printf("ip_mloopback: bad address family %d\n",
2216 dst->sin_family);
2217 dst->sin_family = AF_INET;
2218 }
2219 #endif
2220
2221 #ifdef notdef
2222 copym->m_pkthdr.rcvif = ifp;
2223 ip_input(copym);
2224 #else
2225 if_simloop(ifp, copym, dst->sin_family, 0);
2226 #endif
2227 }
2228 }
Cache object: 21e8deb08c99050afc99ecd3ac7abee8
|