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 * 4. Neither the name of the University nor the names of its contributors
14 * may be used to endorse or promote products derived from this software
15 * without specific prior written permission.
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27 * SUCH DAMAGE.
28 *
29 * @(#)ip_output.c 8.3 (Berkeley) 1/21/94
30 * $FreeBSD: releng/5.4/sys/netinet/ip_output.c 145335 2005-04-20 19:11:07Z cvs2svn $
31 */
32
33 #include "opt_ipfw.h"
34 #include "opt_ipsec.h"
35 #include "opt_mac.h"
36 #include "opt_mbuf_stress_test.h"
37
38 #include <sys/param.h>
39 #include <sys/systm.h>
40 #include <sys/kernel.h>
41 #include <sys/mac.h>
42 #include <sys/malloc.h>
43 #include <sys/mbuf.h>
44 #include <sys/protosw.h>
45 #include <sys/socket.h>
46 #include <sys/socketvar.h>
47 #include <sys/sysctl.h>
48
49 #include <net/if.h>
50 #include <net/netisr.h>
51 #include <net/pfil.h>
52 #include <net/route.h>
53
54 #include <netinet/in.h>
55 #include <netinet/in_systm.h>
56 #include <netinet/ip.h>
57 #include <netinet/in_pcb.h>
58 #include <netinet/in_var.h>
59 #include <netinet/ip_var.h>
60
61 #include <machine/in_cksum.h>
62
63 static MALLOC_DEFINE(M_IPMOPTS, "ip_moptions", "internet multicast options");
64
65 #ifdef IPSEC
66 #include <netinet6/ipsec.h>
67 #include <netkey/key.h>
68 #ifdef IPSEC_DEBUG
69 #include <netkey/key_debug.h>
70 #else
71 #define KEYDEBUG(lev,arg)
72 #endif
73 #endif /*IPSEC*/
74
75 #ifdef FAST_IPSEC
76 #include <netipsec/ipsec.h>
77 #include <netipsec/xform.h>
78 #include <netipsec/key.h>
79 #endif /*FAST_IPSEC*/
80
81 #define print_ip(x, a, y) printf("%s %d.%d.%d.%d%s",\
82 x, (ntohl(a.s_addr)>>24)&0xFF,\
83 (ntohl(a.s_addr)>>16)&0xFF,\
84 (ntohl(a.s_addr)>>8)&0xFF,\
85 (ntohl(a.s_addr))&0xFF, y);
86
87 u_short ip_id;
88
89 #ifdef MBUF_STRESS_TEST
90 int mbuf_frag_size = 0;
91 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW,
92 &mbuf_frag_size, 0, "Fragment outgoing mbufs to this size");
93 #endif
94
95 static struct mbuf *ip_insertoptions(struct mbuf *, struct mbuf *, int *);
96 static struct ifnet *ip_multicast_if(struct in_addr *, int *);
97 static void ip_mloopback
98 (struct ifnet *, struct mbuf *, struct sockaddr_in *, int);
99 static int ip_getmoptions(struct inpcb *, struct sockopt *);
100 static int ip_pcbopts(struct inpcb *, int, struct mbuf *);
101 static int ip_setmoptions(struct inpcb *, struct sockopt *);
102
103 int ip_optcopy(struct ip *, struct ip *);
104
105
106 extern struct protosw inetsw[];
107
108 /*
109 * IP output. The packet in mbuf chain m contains a skeletal IP
110 * header (with len, off, ttl, proto, tos, src, dst).
111 * The mbuf chain containing the packet will be freed.
112 * The mbuf opt, if present, will not be freed.
113 * In the IP forwarding case, the packet will arrive with options already
114 * inserted, so must have a NULL opt pointer.
115 */
116 int
117 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro,
118 int flags, struct ip_moptions *imo, struct inpcb *inp)
119 {
120 struct ip *ip;
121 struct ifnet *ifp = NULL; /* keep compiler happy */
122 struct mbuf *m0;
123 int hlen = sizeof (struct ip);
124 int len, error = 0;
125 struct sockaddr_in *dst = NULL; /* keep compiler happy */
126 struct in_ifaddr *ia = NULL;
127 int isbroadcast, sw_csum;
128 struct route iproute;
129 struct in_addr odst;
130 #ifdef IPFIREWALL_FORWARD
131 struct m_tag *fwd_tag = NULL;
132 #endif
133 #ifdef IPSEC
134 struct secpolicy *sp = NULL;
135 #endif
136 #ifdef FAST_IPSEC
137 struct secpolicy *sp = NULL;
138 struct tdb_ident *tdbi;
139 struct m_tag *mtag;
140 int s;
141 #endif /* FAST_IPSEC */
142
143 M_ASSERTPKTHDR(m);
144
145 if (ro == NULL) {
146 ro = &iproute;
147 bzero(ro, sizeof (*ro));
148 }
149
150 if (inp != NULL)
151 INP_LOCK_ASSERT(inp);
152
153 if (opt) {
154 len = 0;
155 m = ip_insertoptions(m, opt, &len);
156 if (len != 0)
157 hlen = len;
158 }
159 ip = mtod(m, struct ip *);
160
161 /*
162 * Fill in IP header. If we are not allowing fragmentation,
163 * then the ip_id field is meaningless, but we don't set it
164 * to zero. Doing so causes various problems when devices along
165 * the path (routers, load balancers, firewalls, etc.) illegally
166 * disable DF on our packet. Note that a 16-bit counter
167 * will wrap around in less than 10 seconds at 100 Mbit/s on a
168 * medium with MTU 1500. See Steven M. Bellovin, "A Technique
169 * for Counting NATted Hosts", Proc. IMW'02, available at
170 * <http://www.research.att.com/~smb/papers/fnat.pdf>.
171 */
172 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
173 ip->ip_v = IPVERSION;
174 ip->ip_hl = hlen >> 2;
175 ip->ip_id = ip_newid();
176 ipstat.ips_localout++;
177 } else {
178 hlen = ip->ip_hl << 2;
179 }
180
181 dst = (struct sockaddr_in *)&ro->ro_dst;
182 again:
183 /*
184 * If there is a cached route,
185 * check that it is to the same destination
186 * and is still up. If not, free it and try again.
187 * The address family should also be checked in case of sharing the
188 * cache with IPv6.
189 */
190 if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
191 dst->sin_family != AF_INET ||
192 dst->sin_addr.s_addr != ip->ip_dst.s_addr)) {
193 RTFREE(ro->ro_rt);
194 ro->ro_rt = (struct rtentry *)0;
195 }
196 #ifdef IPFIREWALL_FORWARD
197 if (ro->ro_rt == NULL && fwd_tag == NULL) {
198 #else
199 if (ro->ro_rt == NULL) {
200 #endif
201 bzero(dst, sizeof(*dst));
202 dst->sin_family = AF_INET;
203 dst->sin_len = sizeof(*dst);
204 dst->sin_addr = ip->ip_dst;
205 }
206 /*
207 * If routing to interface only,
208 * short circuit routing lookup.
209 */
210 if (flags & IP_ROUTETOIF) {
211 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL &&
212 (ia = ifatoia(ifa_ifwithnet(sintosa(dst)))) == NULL) {
213 ipstat.ips_noroute++;
214 error = ENETUNREACH;
215 goto bad;
216 }
217 ifp = ia->ia_ifp;
218 ip->ip_ttl = 1;
219 isbroadcast = in_broadcast(dst->sin_addr, ifp);
220 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
221 imo != NULL && imo->imo_multicast_ifp != NULL) {
222 /*
223 * Bypass the normal routing lookup for multicast
224 * packets if the interface is specified.
225 */
226 ifp = imo->imo_multicast_ifp;
227 IFP_TO_IA(ifp, ia);
228 isbroadcast = 0; /* fool gcc */
229 } else {
230 /*
231 * We want to do any cloning requested by the link layer,
232 * as this is probably required in all cases for correct
233 * operation (as it is for ARP).
234 */
235 if (ro->ro_rt == NULL)
236 rtalloc_ign(ro, 0);
237 if (ro->ro_rt == NULL) {
238 ipstat.ips_noroute++;
239 error = EHOSTUNREACH;
240 goto bad;
241 }
242 ia = ifatoia(ro->ro_rt->rt_ifa);
243 ifp = ro->ro_rt->rt_ifp;
244 ro->ro_rt->rt_rmx.rmx_pksent++;
245 if (ro->ro_rt->rt_flags & RTF_GATEWAY)
246 dst = (struct sockaddr_in *)ro->ro_rt->rt_gateway;
247 if (ro->ro_rt->rt_flags & RTF_HOST)
248 isbroadcast = (ro->ro_rt->rt_flags & RTF_BROADCAST);
249 else
250 isbroadcast = in_broadcast(dst->sin_addr, ifp);
251 }
252 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
253 struct in_multi *inm;
254
255 m->m_flags |= M_MCAST;
256 /*
257 * IP destination address is multicast. Make sure "dst"
258 * still points to the address in "ro". (It may have been
259 * changed to point to a gateway address, above.)
260 */
261 dst = (struct sockaddr_in *)&ro->ro_dst;
262 /*
263 * See if the caller provided any multicast options
264 */
265 if (imo != NULL) {
266 ip->ip_ttl = imo->imo_multicast_ttl;
267 if (imo->imo_multicast_vif != -1)
268 ip->ip_src.s_addr =
269 ip_mcast_src ?
270 ip_mcast_src(imo->imo_multicast_vif) :
271 INADDR_ANY;
272 } else
273 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
274 /*
275 * Confirm that the outgoing interface supports multicast.
276 */
277 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) {
278 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
279 ipstat.ips_noroute++;
280 error = ENETUNREACH;
281 goto bad;
282 }
283 }
284 /*
285 * If source address not specified yet, use address
286 * of outgoing interface.
287 */
288 if (ip->ip_src.s_addr == INADDR_ANY) {
289 /* Interface may have no addresses. */
290 if (ia != NULL)
291 ip->ip_src = IA_SIN(ia)->sin_addr;
292 }
293
294 IN_LOOKUP_MULTI(ip->ip_dst, ifp, inm);
295 if (inm != NULL &&
296 (imo == NULL || imo->imo_multicast_loop)) {
297 /*
298 * If we belong to the destination multicast group
299 * on the outgoing interface, and the caller did not
300 * forbid loopback, loop back a copy.
301 */
302 ip_mloopback(ifp, m, dst, hlen);
303 }
304 else {
305 /*
306 * If we are acting as a multicast router, perform
307 * multicast forwarding as if the packet had just
308 * arrived on the interface to which we are about
309 * to send. The multicast forwarding function
310 * recursively calls this function, using the
311 * IP_FORWARDING flag to prevent infinite recursion.
312 *
313 * Multicasts that are looped back by ip_mloopback(),
314 * above, will be forwarded by the ip_input() routine,
315 * if necessary.
316 */
317 if (ip_mrouter && (flags & IP_FORWARDING) == 0) {
318 /*
319 * If rsvp daemon is not running, do not
320 * set ip_moptions. This ensures that the packet
321 * is multicast and not just sent down one link
322 * as prescribed by rsvpd.
323 */
324 if (!rsvp_on)
325 imo = NULL;
326 if (ip_mforward &&
327 ip_mforward(ip, ifp, m, imo) != 0) {
328 m_freem(m);
329 goto done;
330 }
331 }
332 }
333
334 /*
335 * Multicasts with a time-to-live of zero may be looped-
336 * back, above, but must not be transmitted on a network.
337 * Also, multicasts addressed to the loopback interface
338 * are not sent -- the above call to ip_mloopback() will
339 * loop back a copy if this host actually belongs to the
340 * destination group on the loopback interface.
341 */
342 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) {
343 m_freem(m);
344 goto done;
345 }
346
347 goto sendit;
348 }
349 #ifndef notdef
350 /*
351 * If the source address is not specified yet, use the address
352 * of the outoing interface.
353 */
354 if (ip->ip_src.s_addr == INADDR_ANY) {
355 /* Interface may have no addresses. */
356 if (ia != NULL) {
357 ip->ip_src = IA_SIN(ia)->sin_addr;
358 }
359 }
360 #endif /* notdef */
361 #ifdef ALTQ
362 /*
363 * disable packet drop hack.
364 * packetdrop should be done by queueing.
365 */
366 #else /* !ALTQ */
367 /*
368 * Verify that we have any chance at all of being able to queue
369 * the packet or packet fragments
370 */
371 if ((ifp->if_snd.ifq_len + ip->ip_len / ifp->if_mtu + 1) >=
372 ifp->if_snd.ifq_maxlen) {
373 error = ENOBUFS;
374 ipstat.ips_odropped++;
375 goto bad;
376 }
377 #endif /* !ALTQ */
378
379 /*
380 * Look for broadcast address and
381 * verify user is allowed to send
382 * such a packet.
383 */
384 if (isbroadcast) {
385 if ((ifp->if_flags & IFF_BROADCAST) == 0) {
386 error = EADDRNOTAVAIL;
387 goto bad;
388 }
389 if ((flags & IP_ALLOWBROADCAST) == 0) {
390 error = EACCES;
391 goto bad;
392 }
393 /* don't allow broadcast messages to be fragmented */
394 if (ip->ip_len > ifp->if_mtu) {
395 error = EMSGSIZE;
396 goto bad;
397 }
398 if (flags & IP_SENDONES)
399 ip->ip_dst.s_addr = INADDR_BROADCAST;
400 m->m_flags |= M_BCAST;
401 } else {
402 m->m_flags &= ~M_BCAST;
403 }
404
405 sendit:
406 #ifdef IPSEC
407 /* get SP for this packet */
408 if (inp == NULL)
409 sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND,
410 flags, &error);
411 else
412 sp = ipsec4_getpolicybypcb(m, IPSEC_DIR_OUTBOUND, inp, &error);
413
414 if (sp == NULL) {
415 ipsecstat.out_inval++;
416 goto bad;
417 }
418
419 error = 0;
420
421 /* check policy */
422 switch (sp->policy) {
423 case IPSEC_POLICY_DISCARD:
424 /*
425 * This packet is just discarded.
426 */
427 ipsecstat.out_polvio++;
428 goto bad;
429
430 case IPSEC_POLICY_BYPASS:
431 case IPSEC_POLICY_NONE:
432 case IPSEC_POLICY_TCP:
433 /* no need to do IPsec. */
434 goto skip_ipsec;
435
436 case IPSEC_POLICY_IPSEC:
437 if (sp->req == NULL) {
438 /* acquire a policy */
439 error = key_spdacquire(sp);
440 goto bad;
441 }
442 break;
443
444 case IPSEC_POLICY_ENTRUST:
445 default:
446 printf("ip_output: Invalid policy found. %d\n", sp->policy);
447 }
448 {
449 struct ipsec_output_state state;
450 bzero(&state, sizeof(state));
451 state.m = m;
452 if (flags & IP_ROUTETOIF) {
453 state.ro = &iproute;
454 bzero(&iproute, sizeof(iproute));
455 } else
456 state.ro = ro;
457 state.dst = (struct sockaddr *)dst;
458
459 ip->ip_sum = 0;
460
461 /*
462 * XXX
463 * delayed checksums are not currently compatible with IPsec
464 */
465 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
466 in_delayed_cksum(m);
467 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
468 }
469
470 ip->ip_len = htons(ip->ip_len);
471 ip->ip_off = htons(ip->ip_off);
472
473 error = ipsec4_output(&state, sp, flags);
474
475 m = state.m;
476 if (flags & IP_ROUTETOIF) {
477 /*
478 * if we have tunnel mode SA, we may need to ignore
479 * IP_ROUTETOIF.
480 */
481 if (state.ro != &iproute || state.ro->ro_rt != NULL) {
482 flags &= ~IP_ROUTETOIF;
483 ro = state.ro;
484 }
485 } else
486 ro = state.ro;
487 dst = (struct sockaddr_in *)state.dst;
488 if (error) {
489 /* mbuf is already reclaimed in ipsec4_output. */
490 m = NULL;
491 switch (error) {
492 case EHOSTUNREACH:
493 case ENETUNREACH:
494 case EMSGSIZE:
495 case ENOBUFS:
496 case ENOMEM:
497 break;
498 default:
499 printf("ip4_output (ipsec): error code %d\n", error);
500 /*fall through*/
501 case ENOENT:
502 /* don't show these error codes to the user */
503 error = 0;
504 break;
505 }
506 goto bad;
507 }
508
509 /* be sure to update variables that are affected by ipsec4_output() */
510 ip = mtod(m, struct ip *);
511 hlen = ip->ip_hl << 2;
512 if (ro->ro_rt == NULL) {
513 if ((flags & IP_ROUTETOIF) == 0) {
514 printf("ip_output: "
515 "can't update route after IPsec processing\n");
516 error = EHOSTUNREACH; /*XXX*/
517 goto bad;
518 }
519 } else {
520 if (state.encap) {
521 ia = ifatoia(ro->ro_rt->rt_ifa);
522 ifp = ro->ro_rt->rt_ifp;
523 }
524 }
525 }
526
527 /* make it flipped, again. */
528 ip->ip_len = ntohs(ip->ip_len);
529 ip->ip_off = ntohs(ip->ip_off);
530 skip_ipsec:
531 #endif /*IPSEC*/
532 #ifdef FAST_IPSEC
533 /*
534 * Check the security policy (SP) for the packet and, if
535 * required, do IPsec-related processing. There are two
536 * cases here; the first time a packet is sent through
537 * it will be untagged and handled by ipsec4_checkpolicy.
538 * If the packet is resubmitted to ip_output (e.g. after
539 * AH, ESP, etc. processing), there will be a tag to bypass
540 * the lookup and related policy checking.
541 */
542 mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL);
543 s = splnet();
544 if (mtag != NULL) {
545 tdbi = (struct tdb_ident *)(mtag + 1);
546 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_OUTBOUND);
547 if (sp == NULL)
548 error = -EINVAL; /* force silent drop */
549 m_tag_delete(m, mtag);
550 } else {
551 sp = ipsec4_checkpolicy(m, IPSEC_DIR_OUTBOUND, flags,
552 &error, inp);
553 }
554 /*
555 * There are four return cases:
556 * sp != NULL apply IPsec policy
557 * sp == NULL, error == 0 no IPsec handling needed
558 * sp == NULL, error == -EINVAL discard packet w/o error
559 * sp == NULL, error != 0 discard packet, report error
560 */
561 if (sp != NULL) {
562 /* Loop detection, check if ipsec processing already done */
563 KASSERT(sp->req != NULL, ("ip_output: no ipsec request"));
564 for (mtag = m_tag_first(m); mtag != NULL;
565 mtag = m_tag_next(m, mtag)) {
566 if (mtag->m_tag_cookie != MTAG_ABI_COMPAT)
567 continue;
568 if (mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_DONE &&
569 mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED)
570 continue;
571 /*
572 * Check if policy has an SA associated with it.
573 * This can happen when an SP has yet to acquire
574 * an SA; e.g. on first reference. If it occurs,
575 * then we let ipsec4_process_packet do its thing.
576 */
577 if (sp->req->sav == NULL)
578 break;
579 tdbi = (struct tdb_ident *)(mtag + 1);
580 if (tdbi->spi == sp->req->sav->spi &&
581 tdbi->proto == sp->req->sav->sah->saidx.proto &&
582 bcmp(&tdbi->dst, &sp->req->sav->sah->saidx.dst,
583 sizeof (union sockaddr_union)) == 0) {
584 /*
585 * No IPsec processing is needed, free
586 * reference to SP.
587 *
588 * NB: null pointer to avoid free at
589 * done: below.
590 */
591 KEY_FREESP(&sp), sp = NULL;
592 splx(s);
593 goto spd_done;
594 }
595 }
596
597 /*
598 * Do delayed checksums now because we send before
599 * this is done in the normal processing path.
600 */
601 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
602 in_delayed_cksum(m);
603 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
604 }
605
606 ip->ip_len = htons(ip->ip_len);
607 ip->ip_off = htons(ip->ip_off);
608
609 /* NB: callee frees mbuf */
610 error = ipsec4_process_packet(m, sp->req, flags, 0);
611 /*
612 * Preserve KAME behaviour: ENOENT can be returned
613 * when an SA acquire is in progress. Don't propagate
614 * this to user-level; it confuses applications.
615 *
616 * XXX this will go away when the SADB is redone.
617 */
618 if (error == ENOENT)
619 error = 0;
620 splx(s);
621 goto done;
622 } else {
623 splx(s);
624
625 if (error != 0) {
626 /*
627 * Hack: -EINVAL is used to signal that a packet
628 * should be silently discarded. This is typically
629 * because we asked key management for an SA and
630 * it was delayed (e.g. kicked up to IKE).
631 */
632 if (error == -EINVAL)
633 error = 0;
634 goto bad;
635 } else {
636 /* No IPsec processing for this packet. */
637 }
638 #ifdef notyet
639 /*
640 * If deferred crypto processing is needed, check that
641 * the interface supports it.
642 */
643 mtag = m_tag_find(m, PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED, NULL);
644 if (mtag != NULL && (ifp->if_capenable & IFCAP_IPSEC) == 0) {
645 /* notify IPsec to do its own crypto */
646 ipsp_skipcrypto_unmark((struct tdb_ident *)(mtag + 1));
647 error = EHOSTUNREACH;
648 goto bad;
649 }
650 #endif
651 }
652 spd_done:
653 #endif /* FAST_IPSEC */
654
655 /* Jump over all PFIL processing if hooks are not active. */
656 if (inet_pfil_hook.ph_busy_count == -1)
657 goto passout;
658
659 /* Run through list of hooks for output packets. */
660 odst.s_addr = ip->ip_dst.s_addr;
661 error = pfil_run_hooks(&inet_pfil_hook, &m, ifp, PFIL_OUT, inp);
662 if (error != 0 || m == NULL)
663 goto done;
664
665 ip = mtod(m, struct ip *);
666
667 /* See if destination IP address was changed by packet filter. */
668 if (odst.s_addr != ip->ip_dst.s_addr) {
669 m->m_flags |= M_SKIP_FIREWALL;
670 if (in_localip(ip->ip_dst)) {
671 m->m_flags |= M_FASTFWD_OURS;
672 if (m->m_pkthdr.rcvif == NULL)
673 m->m_pkthdr.rcvif = loif;
674 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
675 m->m_pkthdr.csum_flags |=
676 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
677 m->m_pkthdr.csum_data = 0xffff;
678 }
679 m->m_pkthdr.csum_flags |=
680 CSUM_IP_CHECKED | CSUM_IP_VALID;
681
682 error = netisr_queue(NETISR_IP, m);
683 goto done;
684 } else
685 goto again;
686 }
687
688 #ifdef IPFIREWALL_FORWARD
689 /* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */
690 if (m->m_flags & M_FASTFWD_OURS) {
691 if (m->m_pkthdr.rcvif == NULL)
692 m->m_pkthdr.rcvif = loif;
693 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
694 m->m_pkthdr.csum_flags |=
695 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
696 m->m_pkthdr.csum_data = 0xffff;
697 }
698 m->m_pkthdr.csum_flags |=
699 CSUM_IP_CHECKED | CSUM_IP_VALID;
700
701 error = netisr_queue(NETISR_IP, m);
702 goto done;
703 }
704 /* Or forward to some other address? */
705 fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL);
706 if (fwd_tag) {
707 #ifndef IPFIREWALL_FORWARD_EXTENDED
708 if (!in_localip(ip->ip_src) && !in_localaddr(ip->ip_dst)) {
709 #endif
710 dst = (struct sockaddr_in *)&ro->ro_dst;
711 bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in));
712 m->m_flags |= M_SKIP_FIREWALL;
713 m_tag_delete(m, fwd_tag);
714 goto again;
715 #ifndef IPFIREWALL_FORWARD_EXTENDED
716 } else {
717 m_tag_delete(m, fwd_tag);
718 /* Continue. */
719 }
720 #endif
721 }
722 #endif /* IPFIREWALL_FORWARD */
723
724 passout:
725 /* 127/8 must not appear on wire - RFC1122. */
726 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
727 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
728 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
729 ipstat.ips_badaddr++;
730 error = EADDRNOTAVAIL;
731 goto bad;
732 }
733 }
734
735 m->m_pkthdr.csum_flags |= CSUM_IP;
736 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_hwassist;
737 if (sw_csum & CSUM_DELAY_DATA) {
738 in_delayed_cksum(m);
739 sw_csum &= ~CSUM_DELAY_DATA;
740 }
741 m->m_pkthdr.csum_flags &= ifp->if_hwassist;
742
743 /*
744 * If small enough for interface, or the interface will take
745 * care of the fragmentation for us, can just send directly.
746 */
747 if (ip->ip_len <= ifp->if_mtu || (ifp->if_hwassist & CSUM_FRAGMENT &&
748 ((ip->ip_off & IP_DF) == 0))) {
749 ip->ip_len = htons(ip->ip_len);
750 ip->ip_off = htons(ip->ip_off);
751 ip->ip_sum = 0;
752 if (sw_csum & CSUM_DELAY_IP)
753 ip->ip_sum = in_cksum(m, hlen);
754
755 /* Record statistics for this interface address. */
756 if (!(flags & IP_FORWARDING) && ia) {
757 ia->ia_ifa.if_opackets++;
758 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
759 }
760
761 #ifdef IPSEC
762 /* clean ipsec history once it goes out of the node */
763 ipsec_delaux(m);
764 #endif
765
766 #ifdef MBUF_STRESS_TEST
767 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size)
768 m = m_fragment(m, M_DONTWAIT, mbuf_frag_size);
769 #endif
770 error = (*ifp->if_output)(ifp, m,
771 (struct sockaddr *)dst, ro->ro_rt);
772 goto done;
773 }
774
775 if (ip->ip_off & IP_DF) {
776 error = EMSGSIZE;
777 /*
778 * This case can happen if the user changed the MTU
779 * of an interface after enabling IP on it. Because
780 * most netifs don't keep track of routes pointing to
781 * them, there is no way for one to update all its
782 * routes when the MTU is changed.
783 */
784 if ((ro->ro_rt->rt_flags & (RTF_UP | RTF_HOST)) &&
785 (ro->ro_rt->rt_rmx.rmx_mtu > ifp->if_mtu)) {
786 ro->ro_rt->rt_rmx.rmx_mtu = ifp->if_mtu;
787 }
788 ipstat.ips_cantfrag++;
789 goto bad;
790 }
791
792 /*
793 * Too large for interface; fragment if possible. If successful,
794 * on return, m will point to a list of packets to be sent.
795 */
796 error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist, sw_csum);
797 if (error)
798 goto bad;
799 for (; m; m = m0) {
800 m0 = m->m_nextpkt;
801 m->m_nextpkt = 0;
802 #ifdef IPSEC
803 /* clean ipsec history once it goes out of the node */
804 ipsec_delaux(m);
805 #endif
806 if (error == 0) {
807 /* Record statistics for this interface address. */
808 if (ia != NULL) {
809 ia->ia_ifa.if_opackets++;
810 ia->ia_ifa.if_obytes += m->m_pkthdr.len;
811 }
812
813 error = (*ifp->if_output)(ifp, m,
814 (struct sockaddr *)dst, ro->ro_rt);
815 } else
816 m_freem(m);
817 }
818
819 if (error == 0)
820 ipstat.ips_fragmented++;
821
822 done:
823 if (ro == &iproute && ro->ro_rt) {
824 RTFREE(ro->ro_rt);
825 }
826 #ifdef IPSEC
827 if (sp != NULL) {
828 KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
829 printf("DP ip_output call free SP:%p\n", sp));
830 key_freesp(sp);
831 }
832 #endif
833 #ifdef FAST_IPSEC
834 if (sp != NULL)
835 KEY_FREESP(&sp);
836 #endif
837 return (error);
838 bad:
839 m_freem(m);
840 goto done;
841 }
842
843 /*
844 * Create a chain of fragments which fit the given mtu. m_frag points to the
845 * mbuf to be fragmented; on return it points to the chain with the fragments.
846 * Return 0 if no error. If error, m_frag may contain a partially built
847 * chain of fragments that should be freed by the caller.
848 *
849 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist)
850 * sw_csum contains the delayed checksums flags (e.g., CSUM_DELAY_IP).
851 */
852 int
853 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu,
854 u_long if_hwassist_flags, int sw_csum)
855 {
856 int error = 0;
857 int hlen = ip->ip_hl << 2;
858 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */
859 int off;
860 struct mbuf *m0 = *m_frag; /* the original packet */
861 int firstlen;
862 struct mbuf **mnext;
863 int nfrags;
864
865 if (ip->ip_off & IP_DF) { /* Fragmentation not allowed */
866 ipstat.ips_cantfrag++;
867 return EMSGSIZE;
868 }
869
870 /*
871 * Must be able to put at least 8 bytes per fragment.
872 */
873 if (len < 8)
874 return EMSGSIZE;
875
876 /*
877 * If the interface will not calculate checksums on
878 * fragmented packets, then do it here.
879 */
880 if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA &&
881 (if_hwassist_flags & CSUM_IP_FRAGS) == 0) {
882 in_delayed_cksum(m0);
883 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
884 }
885
886 if (len > PAGE_SIZE) {
887 /*
888 * Fragment large datagrams such that each segment
889 * contains a multiple of PAGE_SIZE amount of data,
890 * plus headers. This enables a receiver to perform
891 * page-flipping zero-copy optimizations.
892 *
893 * XXX When does this help given that sender and receiver
894 * could have different page sizes, and also mtu could
895 * be less than the receiver's page size ?
896 */
897 int newlen;
898 struct mbuf *m;
899
900 for (m = m0, off = 0; m && (off+m->m_len) <= mtu; m = m->m_next)
901 off += m->m_len;
902
903 /*
904 * firstlen (off - hlen) must be aligned on an
905 * 8-byte boundary
906 */
907 if (off < hlen)
908 goto smart_frag_failure;
909 off = ((off - hlen) & ~7) + hlen;
910 newlen = (~PAGE_MASK) & mtu;
911 if ((newlen + sizeof (struct ip)) > mtu) {
912 /* we failed, go back the default */
913 smart_frag_failure:
914 newlen = len;
915 off = hlen + len;
916 }
917 len = newlen;
918
919 } else {
920 off = hlen + len;
921 }
922
923 firstlen = off - hlen;
924 mnext = &m0->m_nextpkt; /* pointer to next packet */
925
926 /*
927 * Loop through length of segment after first fragment,
928 * make new header and copy data of each part and link onto chain.
929 * Here, m0 is the original packet, m is the fragment being created.
930 * The fragments are linked off the m_nextpkt of the original
931 * packet, which after processing serves as the first fragment.
932 */
933 for (nfrags = 1; off < ip->ip_len; off += len, nfrags++) {
934 struct ip *mhip; /* ip header on the fragment */
935 struct mbuf *m;
936 int mhlen = sizeof (struct ip);
937
938 MGETHDR(m, M_DONTWAIT, MT_HEADER);
939 if (m == NULL) {
940 error = ENOBUFS;
941 ipstat.ips_odropped++;
942 goto done;
943 }
944 m->m_flags |= (m0->m_flags & M_MCAST) | M_FRAG;
945 /*
946 * In the first mbuf, leave room for the link header, then
947 * copy the original IP header including options. The payload
948 * goes into an additional mbuf chain returned by m_copy().
949 */
950 m->m_data += max_linkhdr;
951 mhip = mtod(m, struct ip *);
952 *mhip = *ip;
953 if (hlen > sizeof (struct ip)) {
954 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip);
955 mhip->ip_v = IPVERSION;
956 mhip->ip_hl = mhlen >> 2;
957 }
958 m->m_len = mhlen;
959 /* XXX do we need to add ip->ip_off below ? */
960 mhip->ip_off = ((off - hlen) >> 3) + ip->ip_off;
961 if (off + len >= ip->ip_len) { /* last fragment */
962 len = ip->ip_len - off;
963 m->m_flags |= M_LASTFRAG;
964 } else
965 mhip->ip_off |= IP_MF;
966 mhip->ip_len = htons((u_short)(len + mhlen));
967 m->m_next = m_copy(m0, off, len);
968 if (m->m_next == NULL) { /* copy failed */
969 m_free(m);
970 error = ENOBUFS; /* ??? */
971 ipstat.ips_odropped++;
972 goto done;
973 }
974 m->m_pkthdr.len = mhlen + len;
975 m->m_pkthdr.rcvif = (struct ifnet *)0;
976 #ifdef MAC
977 mac_create_fragment(m0, m);
978 #endif
979 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags;
980 mhip->ip_off = htons(mhip->ip_off);
981 mhip->ip_sum = 0;
982 if (sw_csum & CSUM_DELAY_IP)
983 mhip->ip_sum = in_cksum(m, mhlen);
984 *mnext = m;
985 mnext = &m->m_nextpkt;
986 }
987 ipstat.ips_ofragments += nfrags;
988
989 /* set first marker for fragment chain */
990 m0->m_flags |= M_FIRSTFRAG | M_FRAG;
991 m0->m_pkthdr.csum_data = nfrags;
992
993 /*
994 * Update first fragment by trimming what's been copied out
995 * and updating header.
996 */
997 m_adj(m0, hlen + firstlen - ip->ip_len);
998 m0->m_pkthdr.len = hlen + firstlen;
999 ip->ip_len = htons((u_short)m0->m_pkthdr.len);
1000 ip->ip_off |= IP_MF;
1001 ip->ip_off = htons(ip->ip_off);
1002 ip->ip_sum = 0;
1003 if (sw_csum & CSUM_DELAY_IP)
1004 ip->ip_sum = in_cksum(m0, hlen);
1005
1006 done:
1007 *m_frag = m0;
1008 return error;
1009 }
1010
1011 void
1012 in_delayed_cksum(struct mbuf *m)
1013 {
1014 struct ip *ip;
1015 u_short csum, offset;
1016
1017 ip = mtod(m, struct ip *);
1018 offset = ip->ip_hl << 2 ;
1019 csum = in_cksum_skip(m, ip->ip_len, offset);
1020 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0)
1021 csum = 0xffff;
1022 offset += m->m_pkthdr.csum_data; /* checksum offset */
1023
1024 if (offset + sizeof(u_short) > m->m_len) {
1025 printf("delayed m_pullup, m->len: %d off: %d p: %d\n",
1026 m->m_len, offset, ip->ip_p);
1027 /*
1028 * XXX
1029 * this shouldn't happen, but if it does, the
1030 * correct behavior may be to insert the checksum
1031 * in the existing chain instead of rearranging it.
1032 */
1033 m = m_pullup(m, offset + sizeof(u_short));
1034 }
1035 *(u_short *)(m->m_data + offset) = csum;
1036 }
1037
1038 /*
1039 * Insert IP options into preformed packet.
1040 * Adjust IP destination as required for IP source routing,
1041 * as indicated by a non-zero in_addr at the start of the options.
1042 *
1043 * XXX This routine assumes that the packet has no options in place.
1044 */
1045 static struct mbuf *
1046 ip_insertoptions(m, opt, phlen)
1047 register struct mbuf *m;
1048 struct mbuf *opt;
1049 int *phlen;
1050 {
1051 register struct ipoption *p = mtod(opt, struct ipoption *);
1052 struct mbuf *n;
1053 register struct ip *ip = mtod(m, struct ip *);
1054 unsigned optlen;
1055
1056 optlen = opt->m_len - sizeof(p->ipopt_dst);
1057 if (optlen + ip->ip_len > IP_MAXPACKET) {
1058 *phlen = 0;
1059 return (m); /* XXX should fail */
1060 }
1061 if (p->ipopt_dst.s_addr)
1062 ip->ip_dst = p->ipopt_dst;
1063 if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) {
1064 MGETHDR(n, M_DONTWAIT, MT_HEADER);
1065 if (n == NULL) {
1066 *phlen = 0;
1067 return (m);
1068 }
1069 M_MOVE_PKTHDR(n, m);
1070 n->m_pkthdr.rcvif = (struct ifnet *)0;
1071 #ifdef MAC
1072 mac_create_mbuf_from_mbuf(m, n);
1073 #endif
1074 n->m_pkthdr.len += optlen;
1075 m->m_len -= sizeof(struct ip);
1076 m->m_data += sizeof(struct ip);
1077 n->m_next = m;
1078 m = n;
1079 m->m_len = optlen + sizeof(struct ip);
1080 m->m_data += max_linkhdr;
1081 bcopy(ip, mtod(m, void *), sizeof(struct ip));
1082 } else {
1083 m->m_data -= optlen;
1084 m->m_len += optlen;
1085 m->m_pkthdr.len += optlen;
1086 bcopy(ip, mtod(m, void *), sizeof(struct ip));
1087 }
1088 ip = mtod(m, struct ip *);
1089 bcopy(p->ipopt_list, ip + 1, optlen);
1090 *phlen = sizeof(struct ip) + optlen;
1091 ip->ip_v = IPVERSION;
1092 ip->ip_hl = *phlen >> 2;
1093 ip->ip_len += optlen;
1094 return (m);
1095 }
1096
1097 /*
1098 * Copy options from ip to jp,
1099 * omitting those not copied during fragmentation.
1100 */
1101 int
1102 ip_optcopy(ip, jp)
1103 struct ip *ip, *jp;
1104 {
1105 register u_char *cp, *dp;
1106 int opt, optlen, cnt;
1107
1108 cp = (u_char *)(ip + 1);
1109 dp = (u_char *)(jp + 1);
1110 cnt = (ip->ip_hl << 2) - sizeof (struct ip);
1111 for (; cnt > 0; cnt -= optlen, cp += optlen) {
1112 opt = cp[0];
1113 if (opt == IPOPT_EOL)
1114 break;
1115 if (opt == IPOPT_NOP) {
1116 /* Preserve for IP mcast tunnel's LSRR alignment. */
1117 *dp++ = IPOPT_NOP;
1118 optlen = 1;
1119 continue;
1120 }
1121
1122 KASSERT(cnt >= IPOPT_OLEN + sizeof(*cp),
1123 ("ip_optcopy: malformed ipv4 option"));
1124 optlen = cp[IPOPT_OLEN];
1125 KASSERT(optlen >= IPOPT_OLEN + sizeof(*cp) && optlen <= cnt,
1126 ("ip_optcopy: malformed ipv4 option"));
1127
1128 /* bogus lengths should have been caught by ip_dooptions */
1129 if (optlen > cnt)
1130 optlen = cnt;
1131 if (IPOPT_COPIED(opt)) {
1132 bcopy(cp, dp, optlen);
1133 dp += optlen;
1134 }
1135 }
1136 for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++)
1137 *dp++ = IPOPT_EOL;
1138 return (optlen);
1139 }
1140
1141 /*
1142 * IP socket option processing.
1143 */
1144 int
1145 ip_ctloutput(so, sopt)
1146 struct socket *so;
1147 struct sockopt *sopt;
1148 {
1149 struct inpcb *inp = sotoinpcb(so);
1150 int error, optval;
1151
1152 error = optval = 0;
1153 if (sopt->sopt_level != IPPROTO_IP) {
1154 return (EINVAL);
1155 }
1156
1157 switch (sopt->sopt_dir) {
1158 case SOPT_SET:
1159 switch (sopt->sopt_name) {
1160 case IP_OPTIONS:
1161 #ifdef notyet
1162 case IP_RETOPTS:
1163 #endif
1164 {
1165 struct mbuf *m;
1166 if (sopt->sopt_valsize > MLEN) {
1167 error = EMSGSIZE;
1168 break;
1169 }
1170 MGET(m, sopt->sopt_td ? M_TRYWAIT : M_DONTWAIT, MT_HEADER);
1171 if (m == NULL) {
1172 error = ENOBUFS;
1173 break;
1174 }
1175 m->m_len = sopt->sopt_valsize;
1176 error = sooptcopyin(sopt, mtod(m, char *), m->m_len,
1177 m->m_len);
1178 INP_LOCK(inp);
1179 error = ip_pcbopts(inp, sopt->sopt_name, m);
1180 INP_UNLOCK(inp);
1181 return (error);
1182 }
1183
1184 case IP_TOS:
1185 case IP_TTL:
1186 case IP_RECVOPTS:
1187 case IP_RECVRETOPTS:
1188 case IP_RECVDSTADDR:
1189 case IP_RECVTTL:
1190 case IP_RECVIF:
1191 case IP_FAITH:
1192 case IP_ONESBCAST:
1193 error = sooptcopyin(sopt, &optval, sizeof optval,
1194 sizeof optval);
1195 if (error)
1196 break;
1197
1198 switch (sopt->sopt_name) {
1199 case IP_TOS:
1200 inp->inp_ip_tos = optval;
1201 break;
1202
1203 case IP_TTL:
1204 inp->inp_ip_ttl = optval;
1205 break;
1206 #define OPTSET(bit) do { \
1207 INP_LOCK(inp); \
1208 if (optval) \
1209 inp->inp_flags |= bit; \
1210 else \
1211 inp->inp_flags &= ~bit; \
1212 INP_UNLOCK(inp); \
1213 } while (0)
1214
1215 case IP_RECVOPTS:
1216 OPTSET(INP_RECVOPTS);
1217 break;
1218
1219 case IP_RECVRETOPTS:
1220 OPTSET(INP_RECVRETOPTS);
1221 break;
1222
1223 case IP_RECVDSTADDR:
1224 OPTSET(INP_RECVDSTADDR);
1225 break;
1226
1227 case IP_RECVTTL:
1228 OPTSET(INP_RECVTTL);
1229 break;
1230
1231 case IP_RECVIF:
1232 OPTSET(INP_RECVIF);
1233 break;
1234
1235 case IP_FAITH:
1236 OPTSET(INP_FAITH);
1237 break;
1238
1239 case IP_ONESBCAST:
1240 OPTSET(INP_ONESBCAST);
1241 break;
1242 }
1243 break;
1244 #undef OPTSET
1245
1246 case IP_MULTICAST_IF:
1247 case IP_MULTICAST_VIF:
1248 case IP_MULTICAST_TTL:
1249 case IP_MULTICAST_LOOP:
1250 case IP_ADD_MEMBERSHIP:
1251 case IP_DROP_MEMBERSHIP:
1252 error = ip_setmoptions(inp, sopt);
1253 break;
1254
1255 case IP_PORTRANGE:
1256 error = sooptcopyin(sopt, &optval, sizeof optval,
1257 sizeof optval);
1258 if (error)
1259 break;
1260
1261 INP_LOCK(inp);
1262 switch (optval) {
1263 case IP_PORTRANGE_DEFAULT:
1264 inp->inp_flags &= ~(INP_LOWPORT);
1265 inp->inp_flags &= ~(INP_HIGHPORT);
1266 break;
1267
1268 case IP_PORTRANGE_HIGH:
1269 inp->inp_flags &= ~(INP_LOWPORT);
1270 inp->inp_flags |= INP_HIGHPORT;
1271 break;
1272
1273 case IP_PORTRANGE_LOW:
1274 inp->inp_flags &= ~(INP_HIGHPORT);
1275 inp->inp_flags |= INP_LOWPORT;
1276 break;
1277
1278 default:
1279 error = EINVAL;
1280 break;
1281 }
1282 INP_UNLOCK(inp);
1283 break;
1284
1285 #if defined(IPSEC) || defined(FAST_IPSEC)
1286 case IP_IPSEC_POLICY:
1287 {
1288 caddr_t req;
1289 size_t len = 0;
1290 int priv;
1291 struct mbuf *m;
1292 int optname;
1293
1294 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1295 break;
1296 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1297 break;
1298 priv = (sopt->sopt_td != NULL &&
1299 suser(sopt->sopt_td) != 0) ? 0 : 1;
1300 req = mtod(m, caddr_t);
1301 len = m->m_len;
1302 optname = sopt->sopt_name;
1303 error = ipsec4_set_policy(inp, optname, req, len, priv);
1304 m_freem(m);
1305 break;
1306 }
1307 #endif /*IPSEC*/
1308
1309 default:
1310 error = ENOPROTOOPT;
1311 break;
1312 }
1313 break;
1314
1315 case SOPT_GET:
1316 switch (sopt->sopt_name) {
1317 case IP_OPTIONS:
1318 case IP_RETOPTS:
1319 if (inp->inp_options)
1320 error = sooptcopyout(sopt,
1321 mtod(inp->inp_options,
1322 char *),
1323 inp->inp_options->m_len);
1324 else
1325 sopt->sopt_valsize = 0;
1326 break;
1327
1328 case IP_TOS:
1329 case IP_TTL:
1330 case IP_RECVOPTS:
1331 case IP_RECVRETOPTS:
1332 case IP_RECVDSTADDR:
1333 case IP_RECVTTL:
1334 case IP_RECVIF:
1335 case IP_PORTRANGE:
1336 case IP_FAITH:
1337 case IP_ONESBCAST:
1338 switch (sopt->sopt_name) {
1339
1340 case IP_TOS:
1341 optval = inp->inp_ip_tos;
1342 break;
1343
1344 case IP_TTL:
1345 optval = inp->inp_ip_ttl;
1346 break;
1347
1348 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0)
1349
1350 case IP_RECVOPTS:
1351 optval = OPTBIT(INP_RECVOPTS);
1352 break;
1353
1354 case IP_RECVRETOPTS:
1355 optval = OPTBIT(INP_RECVRETOPTS);
1356 break;
1357
1358 case IP_RECVDSTADDR:
1359 optval = OPTBIT(INP_RECVDSTADDR);
1360 break;
1361
1362 case IP_RECVTTL:
1363 optval = OPTBIT(INP_RECVTTL);
1364 break;
1365
1366 case IP_RECVIF:
1367 optval = OPTBIT(INP_RECVIF);
1368 break;
1369
1370 case IP_PORTRANGE:
1371 if (inp->inp_flags & INP_HIGHPORT)
1372 optval = IP_PORTRANGE_HIGH;
1373 else if (inp->inp_flags & INP_LOWPORT)
1374 optval = IP_PORTRANGE_LOW;
1375 else
1376 optval = 0;
1377 break;
1378
1379 case IP_FAITH:
1380 optval = OPTBIT(INP_FAITH);
1381 break;
1382
1383 case IP_ONESBCAST:
1384 optval = OPTBIT(INP_ONESBCAST);
1385 break;
1386 }
1387 error = sooptcopyout(sopt, &optval, sizeof optval);
1388 break;
1389
1390 case IP_MULTICAST_IF:
1391 case IP_MULTICAST_VIF:
1392 case IP_MULTICAST_TTL:
1393 case IP_MULTICAST_LOOP:
1394 case IP_ADD_MEMBERSHIP:
1395 case IP_DROP_MEMBERSHIP:
1396 error = ip_getmoptions(inp, sopt);
1397 break;
1398
1399 #if defined(IPSEC) || defined(FAST_IPSEC)
1400 case IP_IPSEC_POLICY:
1401 {
1402 struct mbuf *m = NULL;
1403 caddr_t req = NULL;
1404 size_t len = 0;
1405
1406 if (m != 0) {
1407 req = mtod(m, caddr_t);
1408 len = m->m_len;
1409 }
1410 error = ipsec4_get_policy(sotoinpcb(so), req, len, &m);
1411 if (error == 0)
1412 error = soopt_mcopyout(sopt, m); /* XXX */
1413 if (error == 0)
1414 m_freem(m);
1415 break;
1416 }
1417 #endif /*IPSEC*/
1418
1419 default:
1420 error = ENOPROTOOPT;
1421 break;
1422 }
1423 break;
1424 }
1425 return (error);
1426 }
1427
1428 /*
1429 * Set up IP options in pcb for insertion in output packets.
1430 * Store in mbuf with pointer in pcbopt, adding pseudo-option
1431 * with destination address if source routed.
1432 */
1433 static int
1434 ip_pcbopts(struct inpcb *inp, int optname, struct mbuf *m)
1435 {
1436 register int cnt, optlen;
1437 register u_char *cp;
1438 struct mbuf **pcbopt;
1439 u_char opt;
1440
1441 INP_LOCK_ASSERT(inp);
1442
1443 pcbopt = &inp->inp_options;
1444
1445 /* turn off any old options */
1446 if (*pcbopt)
1447 (void)m_free(*pcbopt);
1448 *pcbopt = 0;
1449 if (m == NULL || m->m_len == 0) {
1450 /*
1451 * Only turning off any previous options.
1452 */
1453 if (m != NULL)
1454 (void)m_free(m);
1455 return (0);
1456 }
1457
1458 if (m->m_len % sizeof(int32_t))
1459 goto bad;
1460 /*
1461 * IP first-hop destination address will be stored before
1462 * actual options; move other options back
1463 * and clear it when none present.
1464 */
1465 if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN])
1466 goto bad;
1467 cnt = m->m_len;
1468 m->m_len += sizeof(struct in_addr);
1469 cp = mtod(m, u_char *) + sizeof(struct in_addr);
1470 bcopy(mtod(m, void *), cp, (unsigned)cnt);
1471 bzero(mtod(m, void *), sizeof(struct in_addr));
1472
1473 for (; cnt > 0; cnt -= optlen, cp += optlen) {
1474 opt = cp[IPOPT_OPTVAL];
1475 if (opt == IPOPT_EOL)
1476 break;
1477 if (opt == IPOPT_NOP)
1478 optlen = 1;
1479 else {
1480 if (cnt < IPOPT_OLEN + sizeof(*cp))
1481 goto bad;
1482 optlen = cp[IPOPT_OLEN];
1483 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt)
1484 goto bad;
1485 }
1486 switch (opt) {
1487
1488 default:
1489 break;
1490
1491 case IPOPT_LSRR:
1492 case IPOPT_SSRR:
1493 /*
1494 * user process specifies route as:
1495 * ->A->B->C->D
1496 * D must be our final destination (but we can't
1497 * check that since we may not have connected yet).
1498 * A is first hop destination, which doesn't appear in
1499 * actual IP option, but is stored before the options.
1500 */
1501 if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr))
1502 goto bad;
1503 m->m_len -= sizeof(struct in_addr);
1504 cnt -= sizeof(struct in_addr);
1505 optlen -= sizeof(struct in_addr);
1506 cp[IPOPT_OLEN] = optlen;
1507 /*
1508 * Move first hop before start of options.
1509 */
1510 bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t),
1511 sizeof(struct in_addr));
1512 /*
1513 * Then copy rest of options back
1514 * to close up the deleted entry.
1515 */
1516 bcopy((&cp[IPOPT_OFFSET+1] + sizeof(struct in_addr)),
1517 &cp[IPOPT_OFFSET+1],
1518 (unsigned)cnt - (IPOPT_MINOFF - 1));
1519 break;
1520 }
1521 }
1522 if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr))
1523 goto bad;
1524 *pcbopt = m;
1525 return (0);
1526
1527 bad:
1528 (void)m_free(m);
1529 return (EINVAL);
1530 }
1531
1532 /*
1533 * XXX
1534 * The whole multicast option thing needs to be re-thought.
1535 * Several of these options are equally applicable to non-multicast
1536 * transmission, and one (IP_MULTICAST_TTL) totally duplicates a
1537 * standard option (IP_TTL).
1538 */
1539
1540 /*
1541 * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index.
1542 */
1543 static struct ifnet *
1544 ip_multicast_if(a, ifindexp)
1545 struct in_addr *a;
1546 int *ifindexp;
1547 {
1548 int ifindex;
1549 struct ifnet *ifp;
1550
1551 if (ifindexp)
1552 *ifindexp = 0;
1553 if (ntohl(a->s_addr) >> 24 == 0) {
1554 ifindex = ntohl(a->s_addr) & 0xffffff;
1555 if (ifindex < 0 || if_index < ifindex)
1556 return NULL;
1557 ifp = ifnet_byindex(ifindex);
1558 if (ifindexp)
1559 *ifindexp = ifindex;
1560 } else {
1561 INADDR_TO_IFP(*a, ifp);
1562 }
1563 return ifp;
1564 }
1565
1566 /*
1567 * Set the IP multicast options in response to user setsockopt().
1568 */
1569 static int
1570 ip_setmoptions(struct inpcb *inp, struct sockopt *sopt)
1571 {
1572 int error = 0;
1573 int i;
1574 struct in_addr addr;
1575 struct ip_mreq mreq;
1576 struct ifnet *ifp;
1577 struct ip_moptions *imo;
1578 struct route ro;
1579 struct sockaddr_in *dst;
1580 int ifindex;
1581 int s;
1582
1583 imo = inp->inp_moptions;
1584 if (imo == NULL) {
1585 /*
1586 * No multicast option buffer attached to the pcb;
1587 * allocate one and initialize to default values.
1588 */
1589 imo = (struct ip_moptions*)malloc(sizeof(*imo), M_IPMOPTS,
1590 M_WAITOK);
1591
1592 if (imo == NULL)
1593 return (ENOBUFS);
1594 inp->inp_moptions = imo;
1595 imo->imo_multicast_ifp = NULL;
1596 imo->imo_multicast_addr.s_addr = INADDR_ANY;
1597 imo->imo_multicast_vif = -1;
1598 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
1599 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP;
1600 imo->imo_num_memberships = 0;
1601 }
1602
1603 switch (sopt->sopt_name) {
1604 /* store an index number for the vif you wanna use in the send */
1605 case IP_MULTICAST_VIF:
1606 if (legal_vif_num == 0) {
1607 error = EOPNOTSUPP;
1608 break;
1609 }
1610 error = sooptcopyin(sopt, &i, sizeof i, sizeof i);
1611 if (error)
1612 break;
1613 if (!legal_vif_num(i) && (i != -1)) {
1614 error = EINVAL;
1615 break;
1616 }
1617 imo->imo_multicast_vif = i;
1618 break;
1619
1620 case IP_MULTICAST_IF:
1621 /*
1622 * Select the interface for outgoing multicast packets.
1623 */
1624 error = sooptcopyin(sopt, &addr, sizeof addr, sizeof addr);
1625 if (error)
1626 break;
1627 /*
1628 * INADDR_ANY is used to remove a previous selection.
1629 * When no interface is selected, a default one is
1630 * chosen every time a multicast packet is sent.
1631 */
1632 if (addr.s_addr == INADDR_ANY) {
1633 imo->imo_multicast_ifp = NULL;
1634 break;
1635 }
1636 /*
1637 * The selected interface is identified by its local
1638 * IP address. Find the interface and confirm that
1639 * it supports multicasting.
1640 */
1641 s = splimp();
1642 ifp = ip_multicast_if(&addr, &ifindex);
1643 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1644 splx(s);
1645 error = EADDRNOTAVAIL;
1646 break;
1647 }
1648 imo->imo_multicast_ifp = ifp;
1649 if (ifindex)
1650 imo->imo_multicast_addr = addr;
1651 else
1652 imo->imo_multicast_addr.s_addr = INADDR_ANY;
1653 splx(s);
1654 break;
1655
1656 case IP_MULTICAST_TTL:
1657 /*
1658 * Set the IP time-to-live for outgoing multicast packets.
1659 * The original multicast API required a char argument,
1660 * which is inconsistent with the rest of the socket API.
1661 * We allow either a char or an int.
1662 */
1663 if (sopt->sopt_valsize == 1) {
1664 u_char ttl;
1665 error = sooptcopyin(sopt, &ttl, 1, 1);
1666 if (error)
1667 break;
1668 imo->imo_multicast_ttl = ttl;
1669 } else {
1670 u_int ttl;
1671 error = sooptcopyin(sopt, &ttl, sizeof ttl,
1672 sizeof ttl);
1673 if (error)
1674 break;
1675 if (ttl > 255)
1676 error = EINVAL;
1677 else
1678 imo->imo_multicast_ttl = ttl;
1679 }
1680 break;
1681
1682 case IP_MULTICAST_LOOP:
1683 /*
1684 * Set the loopback flag for outgoing multicast packets.
1685 * Must be zero or one. The original multicast API required a
1686 * char argument, which is inconsistent with the rest
1687 * of the socket API. We allow either a char or an int.
1688 */
1689 if (sopt->sopt_valsize == 1) {
1690 u_char loop;
1691 error = sooptcopyin(sopt, &loop, 1, 1);
1692 if (error)
1693 break;
1694 imo->imo_multicast_loop = !!loop;
1695 } else {
1696 u_int loop;
1697 error = sooptcopyin(sopt, &loop, sizeof loop,
1698 sizeof loop);
1699 if (error)
1700 break;
1701 imo->imo_multicast_loop = !!loop;
1702 }
1703 break;
1704
1705 case IP_ADD_MEMBERSHIP:
1706 /*
1707 * Add a multicast group membership.
1708 * Group must be a valid IP multicast address.
1709 */
1710 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq);
1711 if (error)
1712 break;
1713
1714 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) {
1715 error = EINVAL;
1716 break;
1717 }
1718 s = splimp();
1719 /*
1720 * If no interface address was provided, use the interface of
1721 * the route to the given multicast address.
1722 */
1723 if (mreq.imr_interface.s_addr == INADDR_ANY) {
1724 bzero((caddr_t)&ro, sizeof(ro));
1725 dst = (struct sockaddr_in *)&ro.ro_dst;
1726 dst->sin_len = sizeof(*dst);
1727 dst->sin_family = AF_INET;
1728 dst->sin_addr = mreq.imr_multiaddr;
1729 rtalloc_ign(&ro, RTF_CLONING);
1730 if (ro.ro_rt == NULL) {
1731 error = EADDRNOTAVAIL;
1732 splx(s);
1733 break;
1734 }
1735 ifp = ro.ro_rt->rt_ifp;
1736 RTFREE(ro.ro_rt);
1737 }
1738 else {
1739 ifp = ip_multicast_if(&mreq.imr_interface, NULL);
1740 }
1741
1742 /*
1743 * See if we found an interface, and confirm that it
1744 * supports multicast.
1745 */
1746 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1747 error = EADDRNOTAVAIL;
1748 splx(s);
1749 break;
1750 }
1751 /*
1752 * See if the membership already exists or if all the
1753 * membership slots are full.
1754 */
1755 for (i = 0; i < imo->imo_num_memberships; ++i) {
1756 if (imo->imo_membership[i]->inm_ifp == ifp &&
1757 imo->imo_membership[i]->inm_addr.s_addr
1758 == mreq.imr_multiaddr.s_addr)
1759 break;
1760 }
1761 if (i < imo->imo_num_memberships) {
1762 error = EADDRINUSE;
1763 splx(s);
1764 break;
1765 }
1766 if (i == IP_MAX_MEMBERSHIPS) {
1767 error = ETOOMANYREFS;
1768 splx(s);
1769 break;
1770 }
1771 /*
1772 * Everything looks good; add a new record to the multicast
1773 * address list for the given interface.
1774 */
1775 if ((imo->imo_membership[i] =
1776 in_addmulti(&mreq.imr_multiaddr, ifp)) == NULL) {
1777 error = ENOBUFS;
1778 splx(s);
1779 break;
1780 }
1781 ++imo->imo_num_memberships;
1782 splx(s);
1783 break;
1784
1785 case IP_DROP_MEMBERSHIP:
1786 /*
1787 * Drop a multicast group membership.
1788 * Group must be a valid IP multicast address.
1789 */
1790 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq);
1791 if (error)
1792 break;
1793
1794 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) {
1795 error = EINVAL;
1796 break;
1797 }
1798
1799 s = splimp();
1800 /*
1801 * If an interface address was specified, get a pointer
1802 * to its ifnet structure.
1803 */
1804 if (mreq.imr_interface.s_addr == INADDR_ANY)
1805 ifp = NULL;
1806 else {
1807 ifp = ip_multicast_if(&mreq.imr_interface, NULL);
1808 if (ifp == NULL) {
1809 error = EADDRNOTAVAIL;
1810 splx(s);
1811 break;
1812 }
1813 }
1814 /*
1815 * Find the membership in the membership array.
1816 */
1817 for (i = 0; i < imo->imo_num_memberships; ++i) {
1818 if ((ifp == NULL ||
1819 imo->imo_membership[i]->inm_ifp == ifp) &&
1820 imo->imo_membership[i]->inm_addr.s_addr ==
1821 mreq.imr_multiaddr.s_addr)
1822 break;
1823 }
1824 if (i == imo->imo_num_memberships) {
1825 error = EADDRNOTAVAIL;
1826 splx(s);
1827 break;
1828 }
1829 /*
1830 * Give up the multicast address record to which the
1831 * membership points.
1832 */
1833 in_delmulti(imo->imo_membership[i]);
1834 /*
1835 * Remove the gap in the membership array.
1836 */
1837 for (++i; i < imo->imo_num_memberships; ++i)
1838 imo->imo_membership[i-1] = imo->imo_membership[i];
1839 --imo->imo_num_memberships;
1840 splx(s);
1841 break;
1842
1843 default:
1844 error = EOPNOTSUPP;
1845 break;
1846 }
1847
1848 /*
1849 * If all options have default values, no need to keep the mbuf.
1850 */
1851 if (imo->imo_multicast_ifp == NULL &&
1852 imo->imo_multicast_vif == -1 &&
1853 imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL &&
1854 imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP &&
1855 imo->imo_num_memberships == 0) {
1856 free(inp->inp_moptions, M_IPMOPTS);
1857 inp->inp_moptions = NULL;
1858 }
1859
1860 return (error);
1861 }
1862
1863 /*
1864 * Return the IP multicast options in response to user getsockopt().
1865 */
1866 static int
1867 ip_getmoptions(struct inpcb *inp, struct sockopt *sopt)
1868 {
1869 struct ip_moptions *imo;
1870 struct in_addr addr;
1871 struct in_ifaddr *ia;
1872 int error, optval;
1873 u_char coptval;
1874
1875 INP_LOCK(inp);
1876 imo = inp->inp_moptions;
1877
1878 error = 0;
1879 switch (sopt->sopt_name) {
1880 case IP_MULTICAST_VIF:
1881 if (imo != NULL)
1882 optval = imo->imo_multicast_vif;
1883 else
1884 optval = -1;
1885 INP_UNLOCK(inp);
1886 error = sooptcopyout(sopt, &optval, sizeof optval);
1887 break;
1888
1889 case IP_MULTICAST_IF:
1890 if (imo == NULL || imo->imo_multicast_ifp == NULL)
1891 addr.s_addr = INADDR_ANY;
1892 else if (imo->imo_multicast_addr.s_addr) {
1893 /* return the value user has set */
1894 addr = imo->imo_multicast_addr;
1895 } else {
1896 IFP_TO_IA(imo->imo_multicast_ifp, ia);
1897 addr.s_addr = (ia == NULL) ? INADDR_ANY
1898 : IA_SIN(ia)->sin_addr.s_addr;
1899 }
1900 INP_UNLOCK(inp);
1901 error = sooptcopyout(sopt, &addr, sizeof addr);
1902 break;
1903
1904 case IP_MULTICAST_TTL:
1905 if (imo == 0)
1906 optval = coptval = IP_DEFAULT_MULTICAST_TTL;
1907 else
1908 optval = coptval = imo->imo_multicast_ttl;
1909 INP_UNLOCK(inp);
1910 if (sopt->sopt_valsize == 1)
1911 error = sooptcopyout(sopt, &coptval, 1);
1912 else
1913 error = sooptcopyout(sopt, &optval, sizeof optval);
1914 break;
1915
1916 case IP_MULTICAST_LOOP:
1917 if (imo == 0)
1918 optval = coptval = IP_DEFAULT_MULTICAST_LOOP;
1919 else
1920 optval = coptval = imo->imo_multicast_loop;
1921 INP_UNLOCK(inp);
1922 if (sopt->sopt_valsize == 1)
1923 error = sooptcopyout(sopt, &coptval, 1);
1924 else
1925 error = sooptcopyout(sopt, &optval, sizeof optval);
1926 break;
1927
1928 default:
1929 INP_UNLOCK(inp);
1930 error = ENOPROTOOPT;
1931 break;
1932 }
1933 INP_UNLOCK_ASSERT(inp);
1934
1935 return (error);
1936 }
1937
1938 /*
1939 * Discard the IP multicast options.
1940 */
1941 void
1942 ip_freemoptions(imo)
1943 register struct ip_moptions *imo;
1944 {
1945 register int i;
1946
1947 if (imo != NULL) {
1948 for (i = 0; i < imo->imo_num_memberships; ++i)
1949 in_delmulti(imo->imo_membership[i]);
1950 free(imo, M_IPMOPTS);
1951 }
1952 }
1953
1954 /*
1955 * Routine called from ip_output() to loop back a copy of an IP multicast
1956 * packet to the input queue of a specified interface. Note that this
1957 * calls the output routine of the loopback "driver", but with an interface
1958 * pointer that might NOT be a loopback interface -- evil, but easier than
1959 * replicating that code here.
1960 */
1961 static void
1962 ip_mloopback(ifp, m, dst, hlen)
1963 struct ifnet *ifp;
1964 register struct mbuf *m;
1965 register struct sockaddr_in *dst;
1966 int hlen;
1967 {
1968 register struct ip *ip;
1969 struct mbuf *copym;
1970
1971 copym = m_copy(m, 0, M_COPYALL);
1972 if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen))
1973 copym = m_pullup(copym, hlen);
1974 if (copym != NULL) {
1975 /* If needed, compute the checksum and mark it as valid. */
1976 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1977 in_delayed_cksum(copym);
1978 copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1979 copym->m_pkthdr.csum_flags |=
1980 CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
1981 copym->m_pkthdr.csum_data = 0xffff;
1982 }
1983 /*
1984 * We don't bother to fragment if the IP length is greater
1985 * than the interface's MTU. Can this possibly matter?
1986 */
1987 ip = mtod(copym, struct ip *);
1988 ip->ip_len = htons(ip->ip_len);
1989 ip->ip_off = htons(ip->ip_off);
1990 ip->ip_sum = 0;
1991 ip->ip_sum = in_cksum(copym, hlen);
1992 /*
1993 * NB:
1994 * It's not clear whether there are any lingering
1995 * reentrancy problems in other areas which might
1996 * be exposed by using ip_input directly (in
1997 * particular, everything which modifies the packet
1998 * in-place). Yet another option is using the
1999 * protosw directly to deliver the looped back
2000 * packet. For the moment, we'll err on the side
2001 * of safety by using if_simloop().
2002 */
2003 #if 1 /* XXX */
2004 if (dst->sin_family != AF_INET) {
2005 printf("ip_mloopback: bad address family %d\n",
2006 dst->sin_family);
2007 dst->sin_family = AF_INET;
2008 }
2009 #endif
2010
2011 #ifdef notdef
2012 copym->m_pkthdr.rcvif = ifp;
2013 ip_input(copym);
2014 #else
2015 if_simloop(ifp, copym, dst->sin_family, 0);
2016 #endif
2017 }
2018 }
Cache object: 46bf6f8e0cc478b9efcf0f49bc6c98f9
|