1 /* $OpenBSD: ip_output.c,v 1.403 2025/02/06 23:53:55 bluhm Exp $ */
2 /* $NetBSD: ip_output.c,v 1.28 1996/02/13 23:43:07 christos Exp $ */
3
4 /*
5 * Copyright (c) 1982, 1986, 1988, 1990, 1993
6 * The Regents of the University of California. All rights reserved.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. Neither the name of the University nor the names of its contributors
17 * may be used to endorse or promote products derived from this software
18 * without specific prior written permission.
19 *
20 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30 * SUCH DAMAGE.
31 *
32 * @(#)ip_output.c 8.3 (Berkeley) 1/21/94
33 */
34
35 #include "pf.h"
36
37 #include <sys/param.h>
38 #include <sys/systm.h>
39 #include <sys/mbuf.h>
40 #include <sys/protosw.h>
41 #include <sys/socket.h>
42 #include <sys/socketvar.h>
43 #include <sys/proc.h>
44 #include <sys/kernel.h>
45
46 #include <net/if.h>
47 #include <net/if_var.h>
48 #include <net/if_enc.h>
49 #include <net/route.h>
50
51 #include <netinet/in.h>
52 #include <netinet/ip.h>
53 #include <netinet/in_pcb.h>
54 #include <netinet/in_var.h>
55 #include <netinet/ip_var.h>
56 #include <netinet/ip_icmp.h>
57 #include <netinet/tcp.h>
58 #include <netinet/udp.h>
59 #include <netinet/tcp_timer.h>
60 #include <netinet/tcp_var.h>
61 #include <netinet/udp_var.h>
62
63 #if NPF > 0
64 #include <net/pfvar.h>
65 #endif
66
67 #ifdef IPSEC
68 #ifdef ENCDEBUG
69 #define DPRINTF(fmt, args...) \
70 do { \
71 if (encdebug) \
72 printf("%s: " fmt "\n", __func__, ## args); \
73 } while (0)
74 #else
75 #define DPRINTF(fmt, args...) \
76 do { } while (0)
77 #endif
78 #endif /* IPSEC */
79
80 int ip_pcbopts(struct mbuf **, struct mbuf *);
81 int ip_multicast_if(struct ip_mreqn *, u_int, unsigned int *);
82 int ip_setmoptions(int, struct ip_moptions **, struct mbuf *, u_int);
83 void ip_mloopback(struct ifnet *, struct mbuf *, struct sockaddr_in *);
84 static u_int16_t in_cksum_phdr(u_int32_t, u_int32_t, u_int32_t);
85 void in_delayed_cksum(struct mbuf *);
86
87 int ip_output_ipsec_lookup(struct mbuf *m, int hlen,
88 const struct ipsec_level *seclevel, struct tdb **, int ipsecflowinfo);
89 void ip_output_ipsec_pmtu_update(struct tdb *, struct route *, struct in_addr,
90 int);
91 int ip_output_ipsec_send(struct tdb *, struct mbuf *, struct route *, int);
92
93 /*
94 * IP output. The packet in mbuf chain m contains a skeletal IP
95 * header (with len, off, ttl, proto, tos, src, dst).
96 * The mbuf chain containing the packet will be freed.
97 * The mbuf opt, if present, will not be freed.
98 */
99 int
ip_output(struct mbuf * m,struct mbuf * opt,struct route * ro,int flags,struct ip_moptions * imo,const struct ipsec_level * seclevel,u_int32_t ipsecflowinfo)100 ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags,
101 struct ip_moptions *imo, const struct ipsec_level *seclevel,
102 u_int32_t ipsecflowinfo)
103 {
104 struct ip *ip;
105 struct ifnet *ifp = NULL;
106 struct mbuf_list ml;
107 int hlen = sizeof (struct ip);
108 int error = 0;
109 struct route iproute;
110 struct sockaddr_in *dst;
111 struct tdb *tdb = NULL;
112 u_long mtu;
113 #if NPF > 0
114 u_int orig_rtableid;
115 #endif
116
117 NET_ASSERT_LOCKED();
118
119 #ifdef DIAGNOSTIC
120 if ((m->m_flags & M_PKTHDR) == 0)
121 panic("ip_output no HDR");
122 #endif
123 if (opt)
124 m = ip_insertoptions(m, opt, &hlen);
125
126 ip = mtod(m, struct ip *);
127
128 /*
129 * Fill in IP header.
130 */
131 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) {
132 ip->ip_v = IPVERSION;
133 ip->ip_off &= htons(IP_DF);
134 ip->ip_id = htons(ip_randomid());
135 ip->ip_hl = hlen >> 2;
136 ipstat_inc(ips_localout);
137 } else {
138 hlen = ip->ip_hl << 2;
139 }
140
141 /*
142 * We should not send traffic to 0/8 say both Stevens and RFCs
143 * 5735 section 3 and 1122 sections 3.2.1.3 and 3.3.6.
144 */
145 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == 0) {
146 error = ENETUNREACH;
147 goto bad;
148 }
149
150 #if NPF > 0
151 orig_rtableid = m->m_pkthdr.ph_rtableid;
152 reroute:
153 #endif
154
155 /*
156 * Do a route lookup now in case we need the source address to
157 * do an SPD lookup in IPsec; for most packets, the source address
158 * is set at a higher level protocol. ICMPs and other packets
159 * though (e.g., traceroute) have a source address of zeroes.
160 */
161 if (ro == NULL) {
162 ro = &iproute;
163 ro->ro_rt = NULL;
164 }
165
166 /*
167 * If there is a cached route, check that it is to the same
168 * destination and is still up. If not, free it and try again.
169 */
170 route_cache(ro, &ip->ip_dst, &ip->ip_src, m->m_pkthdr.ph_rtableid);
171 dst = &ro->ro_dstsin;
172
173 if ((IN_MULTICAST(ip->ip_dst.s_addr) ||
174 (ip->ip_dst.s_addr == INADDR_BROADCAST)) &&
175 imo != NULL && (ifp = if_get(imo->imo_ifidx)) != NULL) {
176
177 mtu = ifp->if_mtu;
178 if (ip->ip_src.s_addr == INADDR_ANY) {
179 struct in_ifaddr *ia;
180
181 IFP_TO_IA(ifp, ia);
182 if (ia != NULL)
183 ip->ip_src = ia->ia_addr.sin_addr;
184 }
185 } else {
186 struct in_ifaddr *ia;
187
188 if (ro->ro_rt == NULL)
189 ro->ro_rt = rtalloc_mpath(&ro->ro_dstsa,
190 &ip->ip_src.s_addr, ro->ro_tableid);
191
192 if (ro->ro_rt == NULL) {
193 ipstat_inc(ips_noroute);
194 error = EHOSTUNREACH;
195 goto bad;
196 }
197
198 ia = ifatoia(ro->ro_rt->rt_ifa);
199 if (ISSET(ro->ro_rt->rt_flags, RTF_LOCAL))
200 ifp = if_get(rtable_loindex(m->m_pkthdr.ph_rtableid));
201 else
202 ifp = if_get(ro->ro_rt->rt_ifidx);
203 /*
204 * We aren't using rtisvalid() here because the UP/DOWN state
205 * machine is broken with some Ethernet drivers like em(4).
206 * As a result we might try to use an invalid cached route
207 * entry while an interface is being detached.
208 */
209 if (ifp == NULL) {
210 ipstat_inc(ips_noroute);
211 error = EHOSTUNREACH;
212 goto bad;
213 }
214 mtu = atomic_load_int(&ro->ro_rt->rt_mtu);
215 if (mtu == 0)
216 mtu = ifp->if_mtu;
217
218 if (ro->ro_rt->rt_flags & RTF_GATEWAY)
219 dst = satosin(ro->ro_rt->rt_gateway);
220
221 /* Set the source IP address */
222 if (ip->ip_src.s_addr == INADDR_ANY && ia)
223 ip->ip_src = ia->ia_addr.sin_addr;
224 }
225
226 #ifdef IPSEC
227 if (ipsec_in_use || seclevel != NULL) {
228 /* Do we have any pending SAs to apply ? */
229 error = ip_output_ipsec_lookup(m, hlen, seclevel, &tdb,
230 ipsecflowinfo);
231 if (error) {
232 /* Should silently drop packet */
233 if (error == -EINVAL)
234 error = 0;
235 goto bad;
236 }
237 if (tdb != NULL) {
238 /*
239 * If it needs TCP/UDP hardware-checksumming, do the
240 * computation now.
241 */
242 in_proto_cksum_out(m, NULL);
243 }
244 }
245 #endif /* IPSEC */
246
247 if (IN_MULTICAST(ip->ip_dst.s_addr) ||
248 (ip->ip_dst.s_addr == INADDR_BROADCAST)) {
249
250 m->m_flags |= (ip->ip_dst.s_addr == INADDR_BROADCAST) ?
251 M_BCAST : M_MCAST;
252
253 /*
254 * IP destination address is multicast. Make sure "dst"
255 * still points to the address in "ro". (It may have been
256 * changed to point to a gateway address, above.)
257 */
258 dst = &ro->ro_dstsin;
259
260 /*
261 * See if the caller provided any multicast options
262 */
263 if (imo != NULL)
264 ip->ip_ttl = imo->imo_ttl;
265 else
266 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL;
267
268 /*
269 * if we don't know the outgoing ifp yet, we can't generate
270 * output
271 */
272 if (!ifp) {
273 ipstat_inc(ips_noroute);
274 error = EHOSTUNREACH;
275 goto bad;
276 }
277
278 /*
279 * Confirm that the outgoing interface supports multicast,
280 * but only if the packet actually is going out on that
281 * interface (i.e., no IPsec is applied).
282 */
283 if ((((m->m_flags & M_MCAST) &&
284 (ifp->if_flags & IFF_MULTICAST) == 0) ||
285 ((m->m_flags & M_BCAST) &&
286 (ifp->if_flags & IFF_BROADCAST) == 0)) && (tdb == NULL)) {
287 ipstat_inc(ips_noroute);
288 error = ENETUNREACH;
289 goto bad;
290 }
291
292 /*
293 * If source address not specified yet, use address
294 * of outgoing interface.
295 */
296 if (ip->ip_src.s_addr == INADDR_ANY) {
297 struct in_ifaddr *ia;
298
299 IFP_TO_IA(ifp, ia);
300 if (ia != NULL)
301 ip->ip_src = ia->ia_addr.sin_addr;
302 }
303
304 if ((imo == NULL || imo->imo_loop) &&
305 in_hasmulti(&ip->ip_dst, ifp)) {
306 /*
307 * If we belong to the destination multicast group
308 * on the outgoing interface, and the caller did not
309 * forbid loopback, loop back a copy.
310 * Can't defer TCP/UDP checksumming, do the
311 * computation now.
312 */
313 in_proto_cksum_out(m, NULL);
314 ip_mloopback(ifp, m, dst);
315 }
316 #ifdef MROUTING
317 else {
318 /*
319 * If we are acting as a multicast router, perform
320 * multicast forwarding as if the packet had just
321 * arrived on the interface to which we are about
322 * to send. The multicast forwarding function
323 * recursively calls this function, using the
324 * IP_FORWARDING flag to prevent infinite recursion.
325 *
326 * Multicasts that are looped back by ip_mloopback(),
327 * above, will be forwarded by the ip_input() routine,
328 * if necessary.
329 */
330 if (ipmforwarding && ip_mrouter[ifp->if_rdomain] &&
331 (flags & IP_FORWARDING) == 0) {
332 int rv;
333
334 KERNEL_LOCK();
335 rv = ip_mforward(m, ifp, flags);
336 KERNEL_UNLOCK();
337 if (rv != 0)
338 goto bad;
339 }
340 }
341 #endif
342 /*
343 * Multicasts with a time-to-live of zero may be looped-
344 * back, above, but must not be transmitted on a network.
345 * Also, multicasts addressed to the loopback interface
346 * are not sent -- the above call to ip_mloopback() will
347 * loop back a copy if this host actually belongs to the
348 * destination group on the loopback interface.
349 */
350 if (ip->ip_ttl == 0 || (ifp->if_flags & IFF_LOOPBACK) != 0)
351 goto bad;
352
353 goto sendit;
354 }
355
356 /*
357 * Look for broadcast address and verify user is allowed to send
358 * such a packet; if the packet is going in an IPsec tunnel, skip
359 * this check.
360 */
361 if ((tdb == NULL) && ((dst->sin_addr.s_addr == INADDR_BROADCAST) ||
362 (ro && ro->ro_rt && ISSET(ro->ro_rt->rt_flags, RTF_BROADCAST)))) {
363 if ((ifp->if_flags & IFF_BROADCAST) == 0) {
364 error = EADDRNOTAVAIL;
365 goto bad;
366 }
367 if ((flags & IP_ALLOWBROADCAST) == 0) {
368 error = EACCES;
369 goto bad;
370 }
371
372 /* Don't allow broadcast messages to be fragmented */
373 if (ntohs(ip->ip_len) > ifp->if_mtu) {
374 error = EMSGSIZE;
375 goto bad;
376 }
377 m->m_flags |= M_BCAST;
378 } else
379 m->m_flags &= ~M_BCAST;
380
381 sendit:
382 /*
383 * If we're doing Path MTU discovery, we need to set DF unless
384 * the route's MTU is locked.
385 */
386 if ((flags & IP_MTUDISC) && ro && ro->ro_rt &&
387 (ro->ro_rt->rt_locks & RTV_MTU) == 0)
388 ip->ip_off |= htons(IP_DF);
389
390 #ifdef IPSEC
391 /*
392 * Check if the packet needs encapsulation.
393 */
394 if (tdb != NULL) {
395 /* Callee frees mbuf */
396 error = ip_output_ipsec_send(tdb, m, ro,
397 (flags & IP_FORWARDING) ? 1 : 0);
398 goto done;
399 }
400 #endif /* IPSEC */
401
402 /*
403 * Packet filter
404 */
405 #if NPF > 0
406 if (pf_test(AF_INET, (flags & IP_FORWARDING) ? PF_FWD : PF_OUT,
407 ifp, &m) != PF_PASS) {
408 error = EACCES;
409 goto bad;
410 }
411 if (m == NULL)
412 goto done;
413 ip = mtod(m, struct ip *);
414 hlen = ip->ip_hl << 2;
415 if ((m->m_pkthdr.pf.flags & (PF_TAG_REROUTE | PF_TAG_GENERATED)) ==
416 (PF_TAG_REROUTE | PF_TAG_GENERATED))
417 /* already rerun the route lookup, go on */
418 m->m_pkthdr.pf.flags &= ~(PF_TAG_GENERATED | PF_TAG_REROUTE);
419 else if (m->m_pkthdr.pf.flags & PF_TAG_REROUTE) {
420 /* tag as generated to skip over pf_test on rerun */
421 m->m_pkthdr.pf.flags |= PF_TAG_GENERATED;
422 if (ro == &iproute)
423 rtfree(ro->ro_rt);
424 ro = NULL;
425 if_put(ifp); /* drop reference since target changed */
426 ifp = NULL;
427 goto reroute;
428 }
429 #endif
430
431 #ifdef IPSEC
432 if (ISSET(flags, IP_FORWARDING) && ISSET(flags, IP_FORWARDING_IPSEC) &&
433 !ISSET(m->m_pkthdr.ph_tagsset, PACKET_TAG_IPSEC_IN_DONE)) {
434 error = EHOSTUNREACH;
435 goto bad;
436 }
437 #endif
438
439 /*
440 * If TSO or small enough for interface, can just send directly.
441 */
442 error = if_output_tso(ifp, &m, sintosa(dst), ro->ro_rt, mtu);
443 if (error || m == NULL)
444 goto done;
445
446 /*
447 * Too large for interface; fragment if possible.
448 * Must be able to put at least 8 bytes per fragment.
449 */
450 if (ip->ip_off & htons(IP_DF)) {
451 #ifdef IPSEC
452 if (ip_mtudisc)
453 ipsec_adjust_mtu(m, ifp->if_mtu);
454 #endif
455 error = EMSGSIZE;
456 #if NPF > 0
457 /* pf changed routing table, use orig rtable for path MTU */
458 if (ro->ro_tableid != orig_rtableid) {
459 rtfree(ro->ro_rt);
460 ro->ro_tableid = orig_rtableid;
461 ro->ro_rt = icmp_mtudisc_clone(
462 ro->ro_dstsin.sin_addr, ro->ro_tableid, 0);
463 }
464 #endif
465 /*
466 * This case can happen if the user changed the MTU
467 * of an interface after enabling IP on it. Because
468 * most netifs don't keep track of routes pointing to
469 * them, there is no way for one to update all its
470 * routes when the MTU is changed.
471 */
472 if (rtisvalid(ro->ro_rt) &&
473 ISSET(ro->ro_rt->rt_flags, RTF_HOST) &&
474 !(ro->ro_rt->rt_locks & RTV_MTU)) {
475 u_int rtmtu;
476
477 rtmtu = atomic_load_int(&ro->ro_rt->rt_mtu);
478 if (rtmtu > ifp->if_mtu) {
479 atomic_cas_uint(&ro->ro_rt->rt_mtu, rtmtu,
480 ifp->if_mtu);
481 }
482 }
483 ipstat_inc(ips_cantfrag);
484 goto bad;
485 }
486
487 if ((error = ip_fragment(m, &ml, ifp, mtu)) ||
488 (error = if_output_ml(ifp, &ml, sintosa(dst), ro->ro_rt)))
489 goto done;
490 ipstat_inc(ips_fragmented);
491
492 done:
493 if (ro == &iproute)
494 rtfree(ro->ro_rt);
495 if_put(ifp);
496 #ifdef IPSEC
497 tdb_unref(tdb);
498 #endif /* IPSEC */
499 return (error);
500
501 bad:
502 m_freem(m);
503 goto done;
504 }
505
506 #ifdef IPSEC
507 int
ip_output_ipsec_lookup(struct mbuf * m,int hlen,const struct ipsec_level * seclevel,struct tdb ** tdbout,int ipsecflowinfo)508 ip_output_ipsec_lookup(struct mbuf *m, int hlen,
509 const struct ipsec_level *seclevel, struct tdb **tdbout, int ipsecflowinfo)
510 {
511 struct m_tag *mtag;
512 struct tdb_ident *tdbi;
513 struct tdb *tdb;
514 struct ipsec_ids *ids = NULL;
515 int error;
516
517 /* Do we have any pending SAs to apply ? */
518 if (ipsecflowinfo)
519 ids = ipsp_ids_lookup(ipsecflowinfo);
520 error = ipsp_spd_lookup(m, AF_INET, hlen, IPSP_DIRECTION_OUT,
521 NULL, seclevel, &tdb, ids);
522 ipsp_ids_free(ids);
523 if (error || tdb == NULL) {
524 *tdbout = NULL;
525 return error;
526 }
527 /* Loop detection */
528 for (mtag = m_tag_first(m); mtag != NULL; mtag = m_tag_next(m, mtag)) {
529 if (mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_DONE)
530 continue;
531 tdbi = (struct tdb_ident *)(mtag + 1);
532 if (tdbi->spi == tdb->tdb_spi &&
533 tdbi->proto == tdb->tdb_sproto &&
534 tdbi->rdomain == tdb->tdb_rdomain &&
535 !memcmp(&tdbi->dst, &tdb->tdb_dst,
536 sizeof(union sockaddr_union))) {
537 /* no IPsec needed */
538 tdb_unref(tdb);
539 *tdbout = NULL;
540 return 0;
541 }
542 }
543 *tdbout = tdb;
544 return 0;
545 }
546
547 void
ip_output_ipsec_pmtu_update(struct tdb * tdb,struct route * ro,struct in_addr dst,int rtableid)548 ip_output_ipsec_pmtu_update(struct tdb *tdb, struct route *ro,
549 struct in_addr dst, int rtableid)
550 {
551 struct rtentry *rt = NULL;
552 int rt_mtucloned = 0;
553 int transportmode = (tdb->tdb_dst.sa.sa_family == AF_INET) &&
554 (tdb->tdb_dst.sin.sin_addr.s_addr == dst.s_addr);
555
556 /* Find a host route to store the mtu in */
557 if (ro != NULL)
558 rt = ro->ro_rt;
559 /* but don't add a PMTU route for transport mode SAs */
560 if (transportmode)
561 rt = NULL;
562 else if (rt == NULL || (rt->rt_flags & RTF_HOST) == 0) {
563 rt = icmp_mtudisc_clone(dst, rtableid, 1);
564 rt_mtucloned = 1;
565 }
566 DPRINTF("spi %08x mtu %d rt %p cloned %d",
567 ntohl(tdb->tdb_spi), tdb->tdb_mtu, rt, rt_mtucloned);
568 if (rt != NULL) {
569 atomic_store_int(&rt->rt_mtu, tdb->tdb_mtu);
570 if (ro != NULL && ro->ro_rt != NULL) {
571 rtfree(ro->ro_rt);
572 ro->ro_rt = rtalloc(&ro->ro_dstsa, RT_RESOLVE,
573 rtableid);
574 }
575 if (rt_mtucloned)
576 rtfree(rt);
577 }
578 }
579
580 int
ip_output_ipsec_send(struct tdb * tdb,struct mbuf * m,struct route * ro,int fwd)581 ip_output_ipsec_send(struct tdb *tdb, struct mbuf *m, struct route *ro, int fwd)
582 {
583 struct mbuf_list ml;
584 struct ifnet *encif = NULL;
585 struct ip *ip;
586 struct in_addr dst;
587 u_int len;
588 int error, rtableid, tso = 0;
589
590 #if NPF > 0
591 /*
592 * Packet filter
593 */
594 if ((encif = enc_getif(tdb->tdb_rdomain, tdb->tdb_tap)) == NULL ||
595 pf_test(AF_INET, fwd ? PF_FWD : PF_OUT, encif, &m) != PF_PASS) {
596 m_freem(m);
597 return EACCES;
598 }
599 if (m == NULL)
600 return 0;
601 /*
602 * PF_TAG_REROUTE handling or not...
603 * Packet is entering IPsec so the routing is
604 * already overruled by the IPsec policy.
605 * Until now the change was not reconsidered.
606 * What's the behaviour?
607 */
608 #endif
609
610 /* Check if we can chop the TCP packet */
611 ip = mtod(m, struct ip *);
612 if (ISSET(m->m_pkthdr.csum_flags, M_TCP_TSO) &&
613 m->m_pkthdr.ph_mss <= tdb->tdb_mtu) {
614 tso = 1;
615 len = m->m_pkthdr.ph_mss;
616 } else
617 len = ntohs(ip->ip_len);
618
619 /* Check if we are allowed to fragment */
620 dst = ip->ip_dst;
621 rtableid = m->m_pkthdr.ph_rtableid;
622 if (ip_mtudisc && (ip->ip_off & htons(IP_DF)) && tdb->tdb_mtu &&
623 len > tdb->tdb_mtu && tdb->tdb_mtutimeout > gettime()) {
624 ip_output_ipsec_pmtu_update(tdb, ro, dst, rtableid);
625 ipsec_adjust_mtu(m, tdb->tdb_mtu);
626 m_freem(m);
627 return EMSGSIZE;
628 }
629 /* propagate IP_DF for v4-over-v6 */
630 if (ip_mtudisc && ip->ip_off & htons(IP_DF))
631 SET(m->m_pkthdr.csum_flags, M_IPV6_DF_OUT);
632
633 /*
634 * Clear these -- they'll be set in the recursive invocation
635 * as needed.
636 */
637 m->m_flags &= ~(M_MCAST | M_BCAST);
638
639 if (tso) {
640 error = tcp_chopper(m, &ml, encif, len);
641 if (error)
642 goto done;
643 } else {
644 CLR(m->m_pkthdr.csum_flags, M_TCP_TSO);
645 in_proto_cksum_out(m, encif);
646 ml_init(&ml);
647 ml_enqueue(&ml, m);
648 }
649
650 KERNEL_LOCK();
651 while ((m = ml_dequeue(&ml)) != NULL) {
652 /* Callee frees mbuf */
653 error = ipsp_process_packet(m, tdb, AF_INET, 0);
654 if (error)
655 break;
656 }
657 KERNEL_UNLOCK();
658 done:
659 if (error) {
660 ml_purge(&ml);
661 ipsecstat_inc(ipsec_odrops);
662 tdbstat_inc(tdb, tdb_odrops);
663 }
664 if (!error && tso)
665 tcpstat_inc(tcps_outswtso);
666 if (ip_mtudisc && error == EMSGSIZE)
667 ip_output_ipsec_pmtu_update(tdb, ro, dst, rtableid);
668 return error;
669 }
670 #endif /* IPSEC */
671
672 int
ip_fragment(struct mbuf * m0,struct mbuf_list * ml,struct ifnet * ifp,u_long mtu)673 ip_fragment(struct mbuf *m0, struct mbuf_list *ml, struct ifnet *ifp,
674 u_long mtu)
675 {
676 struct ip *ip;
677 int firstlen, hlen, tlen, len, off;
678 int error;
679
680 ml_init(ml);
681 ml_enqueue(ml, m0);
682
683 ip = mtod(m0, struct ip *);
684 hlen = ip->ip_hl << 2;
685 tlen = m0->m_pkthdr.len;
686 len = (mtu - hlen) &~ 7;
687 if (len < 8) {
688 error = EMSGSIZE;
689 goto bad;
690 }
691 firstlen = len;
692
693 /*
694 * If we are doing fragmentation, we can't defer TCP/UDP
695 * checksumming; compute the checksum and clear the flag.
696 */
697 in_proto_cksum_out(m0, NULL);
698
699 /*
700 * Loop through length of payload after first fragment,
701 * make new header and copy data of each part and link onto chain.
702 */
703 for (off = hlen + firstlen; off < tlen; off += len) {
704 struct mbuf *m;
705 struct ip *mhip;
706 int mhlen;
707
708 MGETHDR(m, M_DONTWAIT, MT_HEADER);
709 if (m == NULL) {
710 error = ENOBUFS;
711 goto bad;
712 }
713 ml_enqueue(ml, m);
714 if ((error = m_dup_pkthdr(m, m0, M_DONTWAIT)) != 0)
715 goto bad;
716 m->m_data += max_linkhdr;
717 mhip = mtod(m, struct ip *);
718 *mhip = *ip;
719 if (hlen > sizeof(struct ip)) {
720 mhlen = ip_optcopy(ip, mhip) + sizeof(struct ip);
721 mhip->ip_hl = mhlen >> 2;
722 } else
723 mhlen = sizeof(struct ip);
724 m->m_len = mhlen;
725
726 mhip->ip_off = ((off - hlen) >> 3) +
727 (ntohs(ip->ip_off) & ~IP_MF);
728 if (ip->ip_off & htons(IP_MF))
729 mhip->ip_off |= IP_MF;
730 if (off + len >= tlen)
731 len = tlen - off;
732 else
733 mhip->ip_off |= IP_MF;
734 mhip->ip_off = htons(mhip->ip_off);
735
736 m->m_pkthdr.len = mhlen + len;
737 mhip->ip_len = htons(m->m_pkthdr.len);
738 m->m_next = m_copym(m0, off, len, M_NOWAIT);
739 if (m->m_next == NULL) {
740 error = ENOBUFS;
741 goto bad;
742 }
743
744 in_hdr_cksum_out(m, ifp);
745 }
746
747 /*
748 * Update first fragment by trimming what's been copied out
749 * and updating header, then send each fragment (in order).
750 */
751 if (hlen + firstlen < tlen) {
752 m_adj(m0, hlen + firstlen - tlen);
753 ip->ip_off |= htons(IP_MF);
754 }
755 ip->ip_len = htons(m0->m_pkthdr.len);
756
757 in_hdr_cksum_out(m0, ifp);
758
759 ipstat_add(ips_ofragments, ml_len(ml));
760 return (0);
761
762 bad:
763 ipstat_inc(ips_odropped);
764 ml_purge(ml);
765 return (error);
766 }
767
768 /*
769 * Insert IP options into preformed packet.
770 * Adjust IP destination as required for IP source routing,
771 * as indicated by a non-zero in_addr at the start of the options.
772 */
773 struct mbuf *
ip_insertoptions(struct mbuf * m,struct mbuf * opt,int * phlen)774 ip_insertoptions(struct mbuf *m, struct mbuf *opt, int *phlen)
775 {
776 struct ipoption *p = mtod(opt, struct ipoption *);
777 struct mbuf *n;
778 struct ip *ip = mtod(m, struct ip *);
779 unsigned int optlen;
780
781 optlen = opt->m_len - sizeof(p->ipopt_dst);
782 if (optlen + ntohs(ip->ip_len) > IP_MAXPACKET)
783 return (m); /* XXX should fail */
784
785 /* check if options will fit to IP header */
786 if ((optlen + sizeof(struct ip)) > (0x0f << 2)) {
787 *phlen = sizeof(struct ip);
788 return (m);
789 }
790
791 if (p->ipopt_dst.s_addr)
792 ip->ip_dst = p->ipopt_dst;
793 if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) {
794 MGETHDR(n, M_DONTWAIT, MT_HEADER);
795 if (n == NULL)
796 return (m);
797 M_MOVE_HDR(n, m);
798 n->m_pkthdr.len += optlen;
799 m->m_len -= sizeof(struct ip);
800 m->m_data += sizeof(struct ip);
801 n->m_next = m;
802 m = n;
803 m->m_len = optlen + sizeof(struct ip);
804 m->m_data += max_linkhdr;
805 memcpy(mtod(m, caddr_t), ip, sizeof(struct ip));
806 } else {
807 m->m_data -= optlen;
808 m->m_len += optlen;
809 m->m_pkthdr.len += optlen;
810 memmove(mtod(m, caddr_t), (caddr_t)ip, sizeof(struct ip));
811 }
812 ip = mtod(m, struct ip *);
813 memcpy(ip + 1, p->ipopt_list, optlen);
814 *phlen = sizeof(struct ip) + optlen;
815 ip->ip_len = htons(ntohs(ip->ip_len) + optlen);
816 return (m);
817 }
818
819 /*
820 * Copy options from ip to jp,
821 * omitting those not copied during fragmentation.
822 */
823 int
ip_optcopy(struct ip * ip,struct ip * jp)824 ip_optcopy(struct ip *ip, struct ip *jp)
825 {
826 u_char *cp, *dp;
827 int opt, optlen, cnt;
828
829 cp = (u_char *)(ip + 1);
830 dp = (u_char *)(jp + 1);
831 cnt = (ip->ip_hl << 2) - sizeof (struct ip);
832 for (; cnt > 0; cnt -= optlen, cp += optlen) {
833 opt = cp[0];
834 if (opt == IPOPT_EOL)
835 break;
836 if (opt == IPOPT_NOP) {
837 /* Preserve for IP mcast tunnel's LSRR alignment. */
838 *dp++ = IPOPT_NOP;
839 optlen = 1;
840 continue;
841 }
842 #ifdef DIAGNOSTIC
843 if (cnt < IPOPT_OLEN + sizeof(*cp))
844 panic("malformed IPv4 option passed to ip_optcopy");
845 #endif
846 optlen = cp[IPOPT_OLEN];
847 #ifdef DIAGNOSTIC
848 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt)
849 panic("malformed IPv4 option passed to ip_optcopy");
850 #endif
851 /* bogus lengths should have been caught by ip_dooptions */
852 if (optlen > cnt)
853 optlen = cnt;
854 if (IPOPT_COPIED(opt)) {
855 memcpy(dp, cp, optlen);
856 dp += optlen;
857 }
858 }
859 for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++)
860 *dp++ = IPOPT_EOL;
861 return (optlen);
862 }
863
864 /*
865 * IP socket option processing.
866 */
867 int
ip_ctloutput(int op,struct socket * so,int level,int optname,struct mbuf * m)868 ip_ctloutput(int op, struct socket *so, int level, int optname,
869 struct mbuf *m)
870 {
871 struct inpcb *inp = sotoinpcb(so);
872 int optval = 0;
873 struct proc *p = curproc; /* XXX */
874 int error = 0;
875 u_int rtableid, rtid = 0;
876
877 if (level != IPPROTO_IP)
878 return (EINVAL);
879
880 rtableid = p->p_p->ps_rtableid;
881
882 switch (op) {
883 case PRCO_SETOPT:
884 switch (optname) {
885 case IP_OPTIONS:
886 return (ip_pcbopts(&inp->inp_options, m));
887
888 case IP_TOS:
889 case IP_TTL:
890 case IP_MINTTL:
891 case IP_RECVOPTS:
892 case IP_RECVRETOPTS:
893 case IP_RECVDSTADDR:
894 case IP_RECVIF:
895 case IP_RECVTTL:
896 case IP_RECVDSTPORT:
897 case IP_RECVRTABLE:
898 case IP_IPSECFLOWINFO:
899 if (m == NULL || m->m_len != sizeof(int))
900 error = EINVAL;
901 else {
902 optval = *mtod(m, int *);
903 switch (optname) {
904
905 case IP_TOS:
906 inp->inp_ip.ip_tos = optval;
907 break;
908
909 case IP_TTL:
910 if (optval > 0 && optval <= MAXTTL)
911 inp->inp_ip.ip_ttl = optval;
912 else if (optval == -1)
913 inp->inp_ip.ip_ttl = ip_defttl;
914 else
915 error = EINVAL;
916 break;
917
918 case IP_MINTTL:
919 if (optval >= 0 && optval <= MAXTTL)
920 inp->inp_ip_minttl = optval;
921 else
922 error = EINVAL;
923 break;
924 #define OPTSET(bit) \
925 if (optval) \
926 inp->inp_flags |= bit; \
927 else \
928 inp->inp_flags &= ~bit;
929
930 case IP_RECVOPTS:
931 OPTSET(INP_RECVOPTS);
932 break;
933
934 case IP_RECVRETOPTS:
935 OPTSET(INP_RECVRETOPTS);
936 break;
937
938 case IP_RECVDSTADDR:
939 OPTSET(INP_RECVDSTADDR);
940 break;
941 case IP_RECVIF:
942 OPTSET(INP_RECVIF);
943 break;
944 case IP_RECVTTL:
945 OPTSET(INP_RECVTTL);
946 break;
947 case IP_RECVDSTPORT:
948 OPTSET(INP_RECVDSTPORT);
949 break;
950 case IP_RECVRTABLE:
951 OPTSET(INP_RECVRTABLE);
952 break;
953 case IP_IPSECFLOWINFO:
954 OPTSET(INP_IPSECFLOWINFO);
955 break;
956 }
957 }
958 break;
959 #undef OPTSET
960
961 case IP_MULTICAST_IF:
962 case IP_MULTICAST_TTL:
963 case IP_MULTICAST_LOOP:
964 case IP_ADD_MEMBERSHIP:
965 case IP_DROP_MEMBERSHIP:
966 error = ip_setmoptions(optname, &inp->inp_moptions, m,
967 inp->inp_rtableid);
968 break;
969
970 case IP_PORTRANGE:
971 if (m == NULL || m->m_len != sizeof(int))
972 error = EINVAL;
973 else {
974 optval = *mtod(m, int *);
975
976 switch (optval) {
977
978 case IP_PORTRANGE_DEFAULT:
979 inp->inp_flags &= ~(INP_LOWPORT);
980 inp->inp_flags &= ~(INP_HIGHPORT);
981 break;
982
983 case IP_PORTRANGE_HIGH:
984 inp->inp_flags &= ~(INP_LOWPORT);
985 inp->inp_flags |= INP_HIGHPORT;
986 break;
987
988 case IP_PORTRANGE_LOW:
989 inp->inp_flags &= ~(INP_HIGHPORT);
990 inp->inp_flags |= INP_LOWPORT;
991 break;
992
993 default:
994
995 error = EINVAL;
996 break;
997 }
998 }
999 break;
1000 case IP_AUTH_LEVEL:
1001 case IP_ESP_TRANS_LEVEL:
1002 case IP_ESP_NETWORK_LEVEL:
1003 case IP_IPCOMP_LEVEL:
1004 #ifndef IPSEC
1005 error = EOPNOTSUPP;
1006 #else
1007 if (m == NULL || m->m_len != sizeof(int)) {
1008 error = EINVAL;
1009 break;
1010 }
1011 optval = *mtod(m, int *);
1012
1013 if (optval < IPSEC_LEVEL_BYPASS ||
1014 optval > IPSEC_LEVEL_UNIQUE) {
1015 error = EINVAL;
1016 break;
1017 }
1018
1019 switch (optname) {
1020 case IP_AUTH_LEVEL:
1021 if (optval < IPSEC_AUTH_LEVEL_DEFAULT &&
1022 suser(p)) {
1023 error = EACCES;
1024 break;
1025 }
1026 inp->inp_seclevel.sl_auth = optval;
1027 break;
1028
1029 case IP_ESP_TRANS_LEVEL:
1030 if (optval < IPSEC_ESP_TRANS_LEVEL_DEFAULT &&
1031 suser(p)) {
1032 error = EACCES;
1033 break;
1034 }
1035 inp->inp_seclevel.sl_esp_trans = optval;
1036 break;
1037
1038 case IP_ESP_NETWORK_LEVEL:
1039 if (optval < IPSEC_ESP_NETWORK_LEVEL_DEFAULT &&
1040 suser(p)) {
1041 error = EACCES;
1042 break;
1043 }
1044 inp->inp_seclevel.sl_esp_network = optval;
1045 break;
1046 case IP_IPCOMP_LEVEL:
1047 if (optval < IPSEC_IPCOMP_LEVEL_DEFAULT &&
1048 suser(p)) {
1049 error = EACCES;
1050 break;
1051 }
1052 inp->inp_seclevel.sl_ipcomp = optval;
1053 break;
1054 }
1055 #endif
1056 break;
1057
1058 case IP_IPSEC_LOCAL_ID:
1059 case IP_IPSEC_REMOTE_ID:
1060 error = EOPNOTSUPP;
1061 break;
1062 case SO_RTABLE:
1063 if (m == NULL || m->m_len < sizeof(u_int)) {
1064 error = EINVAL;
1065 break;
1066 }
1067 rtid = *mtod(m, u_int *);
1068 if (inp->inp_rtableid == rtid)
1069 break;
1070 /* needs privileges to switch when already set */
1071 if (rtableid != rtid && rtableid != 0 &&
1072 (error = suser(p)) != 0)
1073 break;
1074 error = in_pcbset_rtableid(inp, rtid);
1075 break;
1076 case IP_PIPEX:
1077 if (m != NULL && m->m_len == sizeof(int))
1078 inp->inp_pipex = *mtod(m, int *);
1079 else
1080 error = EINVAL;
1081 break;
1082
1083 default:
1084 error = ENOPROTOOPT;
1085 break;
1086 }
1087 break;
1088
1089 case PRCO_GETOPT:
1090 switch (optname) {
1091 case IP_OPTIONS:
1092 case IP_RETOPTS:
1093 if (inp->inp_options) {
1094 m->m_len = inp->inp_options->m_len;
1095 memcpy(mtod(m, caddr_t),
1096 mtod(inp->inp_options, caddr_t), m->m_len);
1097 } else
1098 m->m_len = 0;
1099 break;
1100
1101 case IP_TOS:
1102 case IP_TTL:
1103 case IP_MINTTL:
1104 case IP_RECVOPTS:
1105 case IP_RECVRETOPTS:
1106 case IP_RECVDSTADDR:
1107 case IP_RECVIF:
1108 case IP_RECVTTL:
1109 case IP_RECVDSTPORT:
1110 case IP_RECVRTABLE:
1111 case IP_IPSECFLOWINFO:
1112 case IP_IPDEFTTL:
1113 m->m_len = sizeof(int);
1114 switch (optname) {
1115
1116 case IP_TOS:
1117 optval = inp->inp_ip.ip_tos;
1118 break;
1119
1120 case IP_TTL:
1121 optval = inp->inp_ip.ip_ttl;
1122 break;
1123
1124 case IP_MINTTL:
1125 optval = inp->inp_ip_minttl;
1126 break;
1127
1128 case IP_IPDEFTTL:
1129 optval = ip_defttl;
1130 break;
1131
1132 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0)
1133
1134 case IP_RECVOPTS:
1135 optval = OPTBIT(INP_RECVOPTS);
1136 break;
1137
1138 case IP_RECVRETOPTS:
1139 optval = OPTBIT(INP_RECVRETOPTS);
1140 break;
1141
1142 case IP_RECVDSTADDR:
1143 optval = OPTBIT(INP_RECVDSTADDR);
1144 break;
1145 case IP_RECVIF:
1146 optval = OPTBIT(INP_RECVIF);
1147 break;
1148 case IP_RECVTTL:
1149 optval = OPTBIT(INP_RECVTTL);
1150 break;
1151 case IP_RECVDSTPORT:
1152 optval = OPTBIT(INP_RECVDSTPORT);
1153 break;
1154 case IP_RECVRTABLE:
1155 optval = OPTBIT(INP_RECVRTABLE);
1156 break;
1157 case IP_IPSECFLOWINFO:
1158 optval = OPTBIT(INP_IPSECFLOWINFO);
1159 break;
1160 }
1161 *mtod(m, int *) = optval;
1162 break;
1163
1164 case IP_MULTICAST_IF:
1165 case IP_MULTICAST_TTL:
1166 case IP_MULTICAST_LOOP:
1167 case IP_ADD_MEMBERSHIP:
1168 case IP_DROP_MEMBERSHIP:
1169 error = ip_getmoptions(optname, inp->inp_moptions, m);
1170 break;
1171
1172 case IP_PORTRANGE:
1173 m->m_len = sizeof(int);
1174
1175 if (inp->inp_flags & INP_HIGHPORT)
1176 optval = IP_PORTRANGE_HIGH;
1177 else if (inp->inp_flags & INP_LOWPORT)
1178 optval = IP_PORTRANGE_LOW;
1179 else
1180 optval = 0;
1181
1182 *mtod(m, int *) = optval;
1183 break;
1184
1185 case IP_AUTH_LEVEL:
1186 case IP_ESP_TRANS_LEVEL:
1187 case IP_ESP_NETWORK_LEVEL:
1188 case IP_IPCOMP_LEVEL:
1189 #ifndef IPSEC
1190 m->m_len = sizeof(int);
1191 *mtod(m, int *) = IPSEC_LEVEL_NONE;
1192 #else
1193 m->m_len = sizeof(int);
1194 switch (optname) {
1195 case IP_AUTH_LEVEL:
1196 optval = inp->inp_seclevel.sl_auth;
1197 break;
1198
1199 case IP_ESP_TRANS_LEVEL:
1200 optval = inp->inp_seclevel.sl_esp_trans;
1201 break;
1202
1203 case IP_ESP_NETWORK_LEVEL:
1204 optval = inp->inp_seclevel.sl_esp_network;
1205 break;
1206 case IP_IPCOMP_LEVEL:
1207 optval = inp->inp_seclevel.sl_ipcomp;
1208 break;
1209 }
1210 *mtod(m, int *) = optval;
1211 #endif
1212 break;
1213 case IP_IPSEC_LOCAL_ID:
1214 case IP_IPSEC_REMOTE_ID:
1215 error = EOPNOTSUPP;
1216 break;
1217 case SO_RTABLE:
1218 m->m_len = sizeof(u_int);
1219 *mtod(m, u_int *) = inp->inp_rtableid;
1220 break;
1221 case IP_PIPEX:
1222 m->m_len = sizeof(int);
1223 *mtod(m, int *) = inp->inp_pipex;
1224 break;
1225 default:
1226 error = ENOPROTOOPT;
1227 break;
1228 }
1229 break;
1230 }
1231 return (error);
1232 }
1233
1234 /*
1235 * Set up IP options in pcb for insertion in output packets.
1236 * Store in mbuf with pointer in pcbopt, adding pseudo-option
1237 * with destination address if source routed.
1238 */
1239 int
ip_pcbopts(struct mbuf ** pcbopt,struct mbuf * m)1240 ip_pcbopts(struct mbuf **pcbopt, struct mbuf *m)
1241 {
1242 struct mbuf *n;
1243 struct ipoption *p;
1244 int cnt, off, optlen;
1245 u_char *cp;
1246 u_char opt;
1247
1248 /* turn off any old options */
1249 m_freem(*pcbopt);
1250 *pcbopt = NULL;
1251 if (m == NULL || m->m_len == 0) {
1252 /*
1253 * Only turning off any previous options.
1254 */
1255 return (0);
1256 }
1257
1258 if (m->m_len % sizeof(int32_t) ||
1259 m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr))
1260 return (EINVAL);
1261
1262 /* Don't sleep because NET_LOCK() is hold. */
1263 if ((n = m_get(M_NOWAIT, MT_SOOPTS)) == NULL)
1264 return (ENOBUFS);
1265 p = mtod(n, struct ipoption *);
1266 memset(p, 0, sizeof (*p)); /* 0 = IPOPT_EOL, needed for padding */
1267 n->m_len = sizeof(struct in_addr);
1268
1269 off = 0;
1270 cnt = m->m_len;
1271 cp = mtod(m, u_char *);
1272
1273 while (cnt > 0) {
1274 opt = cp[IPOPT_OPTVAL];
1275
1276 if (opt == IPOPT_NOP || opt == IPOPT_EOL) {
1277 optlen = 1;
1278 } else {
1279 if (cnt < IPOPT_OLEN + sizeof(*cp))
1280 goto bad;
1281 optlen = cp[IPOPT_OLEN];
1282 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt)
1283 goto bad;
1284 }
1285 switch (opt) {
1286 default:
1287 memcpy(p->ipopt_list + off, cp, optlen);
1288 break;
1289
1290 case IPOPT_LSRR:
1291 case IPOPT_SSRR:
1292 /*
1293 * user process specifies route as:
1294 * ->A->B->C->D
1295 * D must be our final destination (but we can't
1296 * check that since we may not have connected yet).
1297 * A is first hop destination, which doesn't appear in
1298 * actual IP option, but is stored before the options.
1299 */
1300 if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr))
1301 goto bad;
1302
1303 /*
1304 * Optlen is smaller because first address is popped.
1305 * Cnt and cp will be adjusted a bit later to reflect
1306 * this.
1307 */
1308 optlen -= sizeof(struct in_addr);
1309 p->ipopt_list[off + IPOPT_OPTVAL] = opt;
1310 p->ipopt_list[off + IPOPT_OLEN] = optlen;
1311
1312 /*
1313 * Move first hop before start of options.
1314 */
1315 memcpy(&p->ipopt_dst, cp + IPOPT_OFFSET,
1316 sizeof(struct in_addr));
1317 cp += sizeof(struct in_addr);
1318 cnt -= sizeof(struct in_addr);
1319 /*
1320 * Then copy rest of options
1321 */
1322 memcpy(p->ipopt_list + off + IPOPT_OFFSET,
1323 cp + IPOPT_OFFSET, optlen - IPOPT_OFFSET);
1324 break;
1325 }
1326 off += optlen;
1327 cp += optlen;
1328 cnt -= optlen;
1329
1330 if (opt == IPOPT_EOL)
1331 break;
1332 }
1333 /* pad options to next word, since p was zeroed just adjust off */
1334 off = (off + sizeof(int32_t) - 1) & ~(sizeof(int32_t) - 1);
1335 n->m_len += off;
1336 if (n->m_len > sizeof(*p)) {
1337 bad:
1338 m_freem(n);
1339 return (EINVAL);
1340 }
1341
1342 *pcbopt = n;
1343 return (0);
1344 }
1345
1346 /*
1347 * Lookup the interface based on the information in the ip_mreqn struct.
1348 */
1349 int
ip_multicast_if(struct ip_mreqn * mreq,u_int rtableid,unsigned int * ifidx)1350 ip_multicast_if(struct ip_mreqn *mreq, u_int rtableid, unsigned int *ifidx)
1351 {
1352 struct sockaddr_in sin;
1353 struct rtentry *rt;
1354
1355 /*
1356 * In case userland provides the imr_ifindex use this as interface.
1357 * If no interface address was provided, use the interface of
1358 * the route to the given multicast address.
1359 */
1360 if (mreq->imr_ifindex != 0) {
1361 *ifidx = mreq->imr_ifindex;
1362 } else if (mreq->imr_address.s_addr == INADDR_ANY) {
1363 memset(&sin, 0, sizeof(sin));
1364 sin.sin_len = sizeof(sin);
1365 sin.sin_family = AF_INET;
1366 sin.sin_addr = mreq->imr_multiaddr;
1367 rt = rtalloc(sintosa(&sin), RT_RESOLVE, rtableid);
1368 if (!rtisvalid(rt)) {
1369 rtfree(rt);
1370 return EADDRNOTAVAIL;
1371 }
1372 *ifidx = rt->rt_ifidx;
1373 rtfree(rt);
1374 } else {
1375 memset(&sin, 0, sizeof(sin));
1376 sin.sin_len = sizeof(sin);
1377 sin.sin_family = AF_INET;
1378 sin.sin_addr = mreq->imr_address;
1379 rt = rtalloc(sintosa(&sin), 0, rtableid);
1380 if (!rtisvalid(rt) || !ISSET(rt->rt_flags, RTF_LOCAL)) {
1381 rtfree(rt);
1382 return EADDRNOTAVAIL;
1383 }
1384 *ifidx = rt->rt_ifidx;
1385 rtfree(rt);
1386 }
1387
1388 return 0;
1389 }
1390
1391 /*
1392 * Set the IP multicast options in response to user setsockopt().
1393 */
1394 int
ip_setmoptions(int optname,struct ip_moptions ** imop,struct mbuf * m,u_int rtableid)1395 ip_setmoptions(int optname, struct ip_moptions **imop, struct mbuf *m,
1396 u_int rtableid)
1397 {
1398 struct in_addr addr;
1399 struct in_ifaddr *ia;
1400 struct ip_mreqn mreqn;
1401 struct ifnet *ifp = NULL;
1402 struct ip_moptions *imo = *imop;
1403 struct in_multi **immp;
1404 struct sockaddr_in sin;
1405 unsigned int ifidx;
1406 int i, error = 0;
1407 u_char loop;
1408
1409 if (imo == NULL) {
1410 /*
1411 * No multicast option buffer attached to the pcb;
1412 * allocate one and initialize to default values.
1413 */
1414 imo = malloc(sizeof(*imo), M_IPMOPTS, M_WAITOK|M_ZERO);
1415 immp = mallocarray(IP_MIN_MEMBERSHIPS, sizeof(*immp), M_IPMOPTS,
1416 M_WAITOK|M_ZERO);
1417 *imop = imo;
1418 imo->imo_ifidx = 0;
1419 imo->imo_ttl = IP_DEFAULT_MULTICAST_TTL;
1420 imo->imo_loop = IP_DEFAULT_MULTICAST_LOOP;
1421 imo->imo_num_memberships = 0;
1422 imo->imo_max_memberships = IP_MIN_MEMBERSHIPS;
1423 imo->imo_membership = immp;
1424 }
1425
1426 switch (optname) {
1427
1428 case IP_MULTICAST_IF:
1429 /*
1430 * Select the interface for outgoing multicast packets.
1431 */
1432 if (m == NULL) {
1433 error = EINVAL;
1434 break;
1435 }
1436 if (m->m_len == sizeof(struct in_addr)) {
1437 addr = *(mtod(m, struct in_addr *));
1438 } else if (m->m_len == sizeof(struct ip_mreq) ||
1439 m->m_len == sizeof(struct ip_mreqn)) {
1440 memset(&mreqn, 0, sizeof(mreqn));
1441 memcpy(&mreqn, mtod(m, void *), m->m_len);
1442
1443 /*
1444 * If an interface index is given use this
1445 * index to set the imo_ifidx but check first
1446 * that the interface actually exists.
1447 * In the other case just set the addr to
1448 * the imr_address and fall through to the
1449 * regular code.
1450 */
1451 if (mreqn.imr_ifindex != 0) {
1452 ifp = if_get(mreqn.imr_ifindex);
1453 if (ifp == NULL ||
1454 ifp->if_rdomain != rtable_l2(rtableid)) {
1455 error = EADDRNOTAVAIL;
1456 if_put(ifp);
1457 break;
1458 }
1459 imo->imo_ifidx = ifp->if_index;
1460 if_put(ifp);
1461 break;
1462 } else
1463 addr = mreqn.imr_address;
1464 } else {
1465 error = EINVAL;
1466 break;
1467 }
1468 /*
1469 * INADDR_ANY is used to remove a previous selection.
1470 * When no interface is selected, a default one is
1471 * chosen every time a multicast packet is sent.
1472 */
1473 if (addr.s_addr == INADDR_ANY) {
1474 imo->imo_ifidx = 0;
1475 break;
1476 }
1477 /*
1478 * The selected interface is identified by its local
1479 * IP address. Find the interface and confirm that
1480 * it supports multicasting.
1481 */
1482 memset(&sin, 0, sizeof(sin));
1483 sin.sin_len = sizeof(sin);
1484 sin.sin_family = AF_INET;
1485 sin.sin_addr = addr;
1486 ia = ifatoia(ifa_ifwithaddr(sintosa(&sin), rtableid));
1487 if (ia == NULL ||
1488 (ia->ia_ifp->if_flags & IFF_MULTICAST) == 0) {
1489 error = EADDRNOTAVAIL;
1490 break;
1491 }
1492 imo->imo_ifidx = ia->ia_ifp->if_index;
1493 break;
1494
1495 case IP_MULTICAST_TTL:
1496 /*
1497 * Set the IP time-to-live for outgoing multicast packets.
1498 */
1499 if (m == NULL || m->m_len != 1) {
1500 error = EINVAL;
1501 break;
1502 }
1503 imo->imo_ttl = *(mtod(m, u_char *));
1504 break;
1505
1506 case IP_MULTICAST_LOOP:
1507 /*
1508 * Set the loopback flag for outgoing multicast packets.
1509 * Must be zero or one.
1510 */
1511 if (m == NULL || m->m_len != 1 ||
1512 (loop = *(mtod(m, u_char *))) > 1) {
1513 error = EINVAL;
1514 break;
1515 }
1516 imo->imo_loop = loop;
1517 break;
1518
1519 case IP_ADD_MEMBERSHIP:
1520 /*
1521 * Add a multicast group membership.
1522 * Group must be a valid IP multicast address.
1523 */
1524 if (m == NULL || !(m->m_len == sizeof(struct ip_mreq) ||
1525 m->m_len == sizeof(struct ip_mreqn))) {
1526 error = EINVAL;
1527 break;
1528 }
1529 memset(&mreqn, 0, sizeof(mreqn));
1530 memcpy(&mreqn, mtod(m, void *), m->m_len);
1531 if (!IN_MULTICAST(mreqn.imr_multiaddr.s_addr)) {
1532 error = EINVAL;
1533 break;
1534 }
1535
1536 error = ip_multicast_if(&mreqn, rtableid, &ifidx);
1537 if (error)
1538 break;
1539
1540 /*
1541 * See if we found an interface, and confirm that it
1542 * supports multicast.
1543 */
1544 ifp = if_get(ifidx);
1545 if (ifp == NULL || ifp->if_rdomain != rtable_l2(rtableid) ||
1546 (ifp->if_flags & IFF_MULTICAST) == 0) {
1547 error = EADDRNOTAVAIL;
1548 if_put(ifp);
1549 break;
1550 }
1551
1552 /*
1553 * See if the membership already exists or if all the
1554 * membership slots are full.
1555 */
1556 for (i = 0; i < imo->imo_num_memberships; ++i) {
1557 if (imo->imo_membership[i]->inm_ifidx == ifidx &&
1558 imo->imo_membership[i]->inm_addr.s_addr
1559 == mreqn.imr_multiaddr.s_addr)
1560 break;
1561 }
1562 if (i < imo->imo_num_memberships) {
1563 error = EADDRINUSE;
1564 if_put(ifp);
1565 break;
1566 }
1567 if (imo->imo_num_memberships == imo->imo_max_memberships) {
1568 struct in_multi **nmships, **omships;
1569 size_t newmax;
1570 /*
1571 * Resize the vector to next power-of-two minus 1. If
1572 * the size would exceed the maximum then we know we've
1573 * really run out of entries. Otherwise, we reallocate
1574 * the vector.
1575 */
1576 nmships = NULL;
1577 omships = imo->imo_membership;
1578 newmax = ((imo->imo_max_memberships + 1) * 2) - 1;
1579 if (newmax <= IP_MAX_MEMBERSHIPS) {
1580 nmships = mallocarray(newmax, sizeof(*nmships),
1581 M_IPMOPTS, M_NOWAIT|M_ZERO);
1582 if (nmships != NULL) {
1583 memcpy(nmships, omships,
1584 sizeof(*omships) *
1585 imo->imo_max_memberships);
1586 free(omships, M_IPMOPTS,
1587 sizeof(*omships) *
1588 imo->imo_max_memberships);
1589 imo->imo_membership = nmships;
1590 imo->imo_max_memberships = newmax;
1591 }
1592 }
1593 if (nmships == NULL) {
1594 error = ENOBUFS;
1595 if_put(ifp);
1596 break;
1597 }
1598 }
1599 /*
1600 * Everything looks good; add a new record to the multicast
1601 * address list for the given interface.
1602 */
1603 if ((imo->imo_membership[i] =
1604 in_addmulti(&mreqn.imr_multiaddr, ifp)) == NULL) {
1605 error = ENOBUFS;
1606 if_put(ifp);
1607 break;
1608 }
1609 ++imo->imo_num_memberships;
1610 if_put(ifp);
1611 break;
1612
1613 case IP_DROP_MEMBERSHIP:
1614 /*
1615 * Drop a multicast group membership.
1616 * Group must be a valid IP multicast address.
1617 */
1618 if (m == NULL || !(m->m_len == sizeof(struct ip_mreq) ||
1619 m->m_len == sizeof(struct ip_mreqn))) {
1620 error = EINVAL;
1621 break;
1622 }
1623 memset(&mreqn, 0, sizeof(mreqn));
1624 memcpy(&mreqn, mtod(m, void *), m->m_len);
1625 if (!IN_MULTICAST(mreqn.imr_multiaddr.s_addr)) {
1626 error = EINVAL;
1627 break;
1628 }
1629
1630 /*
1631 * If an interface address was specified, get a pointer
1632 * to its ifnet structure.
1633 */
1634 error = ip_multicast_if(&mreqn, rtableid, &ifidx);
1635 if (error)
1636 break;
1637
1638 /*
1639 * Find the membership in the membership array.
1640 */
1641 for (i = 0; i < imo->imo_num_memberships; ++i) {
1642 if ((ifidx == 0 ||
1643 imo->imo_membership[i]->inm_ifidx == ifidx) &&
1644 imo->imo_membership[i]->inm_addr.s_addr ==
1645 mreqn.imr_multiaddr.s_addr)
1646 break;
1647 }
1648 if (i == imo->imo_num_memberships) {
1649 error = EADDRNOTAVAIL;
1650 break;
1651 }
1652 /*
1653 * Give up the multicast address record to which the
1654 * membership points.
1655 */
1656 in_delmulti(imo->imo_membership[i]);
1657 /*
1658 * Remove the gap in the membership array.
1659 */
1660 for (++i; i < imo->imo_num_memberships; ++i)
1661 imo->imo_membership[i-1] = imo->imo_membership[i];
1662 --imo->imo_num_memberships;
1663 break;
1664
1665 default:
1666 error = EOPNOTSUPP;
1667 break;
1668 }
1669
1670 /*
1671 * If all options have default values, no need to keep the data.
1672 */
1673 if (imo->imo_ifidx == 0 &&
1674 imo->imo_ttl == IP_DEFAULT_MULTICAST_TTL &&
1675 imo->imo_loop == IP_DEFAULT_MULTICAST_LOOP &&
1676 imo->imo_num_memberships == 0) {
1677 free(imo->imo_membership , M_IPMOPTS,
1678 imo->imo_max_memberships * sizeof(struct in_multi *));
1679 free(*imop, M_IPMOPTS, sizeof(**imop));
1680 *imop = NULL;
1681 }
1682
1683 return (error);
1684 }
1685
1686 /*
1687 * Return the IP multicast options in response to user getsockopt().
1688 */
1689 int
ip_getmoptions(int optname,struct ip_moptions * imo,struct mbuf * m)1690 ip_getmoptions(int optname, struct ip_moptions *imo, struct mbuf *m)
1691 {
1692 u_char *ttl;
1693 u_char *loop;
1694 struct in_addr *addr;
1695 struct in_ifaddr *ia;
1696 struct ifnet *ifp;
1697
1698 switch (optname) {
1699
1700 case IP_MULTICAST_IF:
1701 addr = mtod(m, struct in_addr *);
1702 m->m_len = sizeof(struct in_addr);
1703 if (imo == NULL || (ifp = if_get(imo->imo_ifidx)) == NULL)
1704 addr->s_addr = INADDR_ANY;
1705 else {
1706 IFP_TO_IA(ifp, ia);
1707 addr->s_addr = (ia == NULL) ? INADDR_ANY
1708 : ia->ia_addr.sin_addr.s_addr;
1709 if_put(ifp);
1710 }
1711 return (0);
1712
1713 case IP_MULTICAST_TTL:
1714 ttl = mtod(m, u_char *);
1715 m->m_len = 1;
1716 *ttl = (imo == NULL) ? IP_DEFAULT_MULTICAST_TTL
1717 : imo->imo_ttl;
1718 return (0);
1719
1720 case IP_MULTICAST_LOOP:
1721 loop = mtod(m, u_char *);
1722 m->m_len = 1;
1723 *loop = (imo == NULL) ? IP_DEFAULT_MULTICAST_LOOP
1724 : imo->imo_loop;
1725 return (0);
1726
1727 default:
1728 return (EOPNOTSUPP);
1729 }
1730 }
1731
1732 /*
1733 * Discard the IP multicast options.
1734 */
1735 void
ip_freemoptions(struct ip_moptions * imo)1736 ip_freemoptions(struct ip_moptions *imo)
1737 {
1738 int i;
1739
1740 if (imo != NULL) {
1741 for (i = 0; i < imo->imo_num_memberships; ++i)
1742 in_delmulti(imo->imo_membership[i]);
1743 free(imo->imo_membership, M_IPMOPTS,
1744 imo->imo_max_memberships * sizeof(struct in_multi *));
1745 free(imo, M_IPMOPTS, sizeof(*imo));
1746 }
1747 }
1748
1749 /*
1750 * Routine called from ip_output() to loop back a copy of an IP multicast
1751 * packet to the input queue of a specified interface.
1752 */
1753 void
ip_mloopback(struct ifnet * ifp,struct mbuf * m,struct sockaddr_in * dst)1754 ip_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in *dst)
1755 {
1756 struct mbuf *copym;
1757
1758 copym = m_dup_pkt(m, max_linkhdr, M_DONTWAIT);
1759 if (copym != NULL) {
1760 /*
1761 * We don't bother to fragment if the IP length is greater
1762 * than the interface's MTU. Can this possibly matter?
1763 */
1764 in_hdr_cksum_out(copym, NULL);
1765 if_input_local(ifp, copym, dst->sin_family);
1766 }
1767 }
1768
1769 void
in_hdr_cksum_out(struct mbuf * m,struct ifnet * ifp)1770 in_hdr_cksum_out(struct mbuf *m, struct ifnet *ifp)
1771 {
1772 struct ip *ip = mtod(m, struct ip *);
1773
1774 ip->ip_sum = 0;
1775 if (in_ifcap_cksum(m, ifp, IFCAP_CSUM_IPv4)) {
1776 SET(m->m_pkthdr.csum_flags, M_IPV4_CSUM_OUT);
1777 } else {
1778 ipstat_inc(ips_outswcsum);
1779 ip->ip_sum = in_cksum(m, ip->ip_hl << 2);
1780 CLR(m->m_pkthdr.csum_flags, M_IPV4_CSUM_OUT);
1781 }
1782 }
1783
1784 /*
1785 * Compute significant parts of the IPv4 checksum pseudo-header
1786 * for use in a delayed TCP/UDP checksum calculation.
1787 */
1788 static u_int16_t
in_cksum_phdr(u_int32_t src,u_int32_t dst,u_int32_t lenproto)1789 in_cksum_phdr(u_int32_t src, u_int32_t dst, u_int32_t lenproto)
1790 {
1791 u_int32_t sum;
1792
1793 sum = lenproto +
1794 (u_int16_t)(src >> 16) +
1795 (u_int16_t)(src /*& 0xffff*/) +
1796 (u_int16_t)(dst >> 16) +
1797 (u_int16_t)(dst /*& 0xffff*/);
1798
1799 sum = (u_int16_t)(sum >> 16) + (u_int16_t)(sum /*& 0xffff*/);
1800
1801 if (sum > 0xffff)
1802 sum -= 0xffff;
1803
1804 return (sum);
1805 }
1806
1807 /*
1808 * Process a delayed payload checksum calculation.
1809 */
1810 void
in_delayed_cksum(struct mbuf * m)1811 in_delayed_cksum(struct mbuf *m)
1812 {
1813 struct ip *ip;
1814 u_int16_t csum, offset;
1815
1816 ip = mtod(m, struct ip *);
1817 offset = ip->ip_hl << 2;
1818 csum = in4_cksum(m, 0, offset, m->m_pkthdr.len - offset);
1819 if (csum == 0 && ip->ip_p == IPPROTO_UDP)
1820 csum = 0xffff;
1821
1822 switch (ip->ip_p) {
1823 case IPPROTO_TCP:
1824 offset += offsetof(struct tcphdr, th_sum);
1825 break;
1826
1827 case IPPROTO_UDP:
1828 offset += offsetof(struct udphdr, uh_sum);
1829 break;
1830
1831 case IPPROTO_ICMP:
1832 offset += offsetof(struct icmp, icmp_cksum);
1833 break;
1834
1835 default:
1836 return;
1837 }
1838
1839 if ((offset + sizeof(u_int16_t)) > m->m_len)
1840 m_copyback(m, offset, sizeof(csum), &csum, M_NOWAIT);
1841 else
1842 *(u_int16_t *)(mtod(m, caddr_t) + offset) = csum;
1843 }
1844
1845 void
in_proto_cksum_out(struct mbuf * m,struct ifnet * ifp)1846 in_proto_cksum_out(struct mbuf *m, struct ifnet *ifp)
1847 {
1848 struct ip *ip = mtod(m, struct ip *);
1849
1850 /* some hw and in_delayed_cksum need the pseudo header cksum */
1851 if (m->m_pkthdr.csum_flags &
1852 (M_TCP_CSUM_OUT|M_UDP_CSUM_OUT|M_ICMP_CSUM_OUT)) {
1853 u_int16_t csum = 0, offset;
1854
1855 offset = ip->ip_hl << 2;
1856 if (ISSET(m->m_pkthdr.csum_flags, M_TCP_TSO) &&
1857 in_ifcap_cksum(m, ifp, IFCAP_TSOv4)) {
1858 csum = in_cksum_phdr(ip->ip_src.s_addr,
1859 ip->ip_dst.s_addr, htonl(ip->ip_p));
1860 } else if (ISSET(m->m_pkthdr.csum_flags,
1861 M_TCP_CSUM_OUT|M_UDP_CSUM_OUT)) {
1862 csum = in_cksum_phdr(ip->ip_src.s_addr,
1863 ip->ip_dst.s_addr, htonl(ntohs(ip->ip_len) -
1864 offset + ip->ip_p));
1865 }
1866 if (ip->ip_p == IPPROTO_TCP)
1867 offset += offsetof(struct tcphdr, th_sum);
1868 else if (ip->ip_p == IPPROTO_UDP)
1869 offset += offsetof(struct udphdr, uh_sum);
1870 else if (ip->ip_p == IPPROTO_ICMP)
1871 offset += offsetof(struct icmp, icmp_cksum);
1872 if ((offset + sizeof(u_int16_t)) > m->m_len)
1873 m_copyback(m, offset, sizeof(csum), &csum, M_NOWAIT);
1874 else
1875 *(u_int16_t *)(mtod(m, caddr_t) + offset) = csum;
1876 }
1877
1878 if (m->m_pkthdr.csum_flags & M_TCP_CSUM_OUT) {
1879 if (!in_ifcap_cksum(m, ifp, IFCAP_CSUM_TCPv4) ||
1880 ip->ip_hl != 5) {
1881 tcpstat_inc(tcps_outswcsum);
1882 in_delayed_cksum(m);
1883 m->m_pkthdr.csum_flags &= ~M_TCP_CSUM_OUT; /* Clear */
1884 }
1885 } else if (m->m_pkthdr.csum_flags & M_UDP_CSUM_OUT) {
1886 if (!in_ifcap_cksum(m, ifp, IFCAP_CSUM_UDPv4) ||
1887 ip->ip_hl != 5) {
1888 udpstat_inc(udps_outswcsum);
1889 in_delayed_cksum(m);
1890 m->m_pkthdr.csum_flags &= ~M_UDP_CSUM_OUT; /* Clear */
1891 }
1892 } else if (m->m_pkthdr.csum_flags & M_ICMP_CSUM_OUT) {
1893 in_delayed_cksum(m);
1894 m->m_pkthdr.csum_flags &= ~M_ICMP_CSUM_OUT; /* Clear */
1895 }
1896 }
1897
1898 int
in_ifcap_cksum(struct mbuf * m,struct ifnet * ifp,int ifcap)1899 in_ifcap_cksum(struct mbuf *m, struct ifnet *ifp, int ifcap)
1900 {
1901 if ((ifp == NULL) ||
1902 !ISSET(ifp->if_capabilities, ifcap) ||
1903 (ifp->if_bridgeidx != 0))
1904 return (0);
1905 /*
1906 * Simplex interface sends packet back without hardware cksum.
1907 * Keep this check in sync with the condition where ether_resolve()
1908 * calls if_input_local().
1909 */
1910 if (ISSET(m->m_flags, M_BCAST) &&
1911 ISSET(ifp->if_flags, IFF_SIMPLEX) &&
1912 !m->m_pkthdr.pf.routed)
1913 return (0);
1914 return (1);
1915 }
1916