1 /* $OpenBSD: ip_output.c,v 1.402 2025/01/03 21:27:40 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, 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,int transportmode)548 ip_output_ipsec_pmtu_update(struct tdb *tdb, struct route *ro,
549 struct in_addr dst, int rtableid, int transportmode)
550 {
551 struct rtentry *rt = NULL;
552 int rt_mtucloned = 0;
553
554 /* Find a host route to store the mtu in */
555 if (ro != NULL)
556 rt = ro->ro_rt;
557 /* but don't add a PMTU route for transport mode SAs */
558 if (transportmode)
559 rt = NULL;
560 else if (rt == NULL || (rt->rt_flags & RTF_HOST) == 0) {
561 rt = icmp_mtudisc_clone(dst, rtableid, 1);
562 rt_mtucloned = 1;
563 }
564 DPRINTF("spi %08x mtu %d rt %p cloned %d",
565 ntohl(tdb->tdb_spi), tdb->tdb_mtu, rt, rt_mtucloned);
566 if (rt != NULL) {
567 atomic_store_int(&rt->rt_mtu, tdb->tdb_mtu);
568 if (ro != NULL && ro->ro_rt != NULL) {
569 rtfree(ro->ro_rt);
570 ro->ro_rt = rtalloc(&ro->ro_dstsa, RT_RESOLVE,
571 rtableid);
572 }
573 if (rt_mtucloned)
574 rtfree(rt);
575 }
576 }
577
578 int
ip_output_ipsec_send(struct tdb * tdb,struct mbuf * m,struct route * ro,int fwd)579 ip_output_ipsec_send(struct tdb *tdb, struct mbuf *m, struct route *ro, int fwd)
580 {
581 struct mbuf_list ml;
582 struct ifnet *encif = NULL;
583 struct ip *ip;
584 struct in_addr dst;
585 u_int len;
586 int error, rtableid, tso = 0;
587
588 #if NPF > 0
589 /*
590 * Packet filter
591 */
592 if ((encif = enc_getif(tdb->tdb_rdomain, tdb->tdb_tap)) == NULL ||
593 pf_test(AF_INET, fwd ? PF_FWD : PF_OUT, encif, &m) != PF_PASS) {
594 m_freem(m);
595 return EACCES;
596 }
597 if (m == NULL)
598 return 0;
599 /*
600 * PF_TAG_REROUTE handling or not...
601 * Packet is entering IPsec so the routing is
602 * already overruled by the IPsec policy.
603 * Until now the change was not reconsidered.
604 * What's the behaviour?
605 */
606 #endif
607
608 /* Check if we can chop the TCP packet */
609 ip = mtod(m, struct ip *);
610 if (ISSET(m->m_pkthdr.csum_flags, M_TCP_TSO) &&
611 m->m_pkthdr.ph_mss <= tdb->tdb_mtu) {
612 tso = 1;
613 len = m->m_pkthdr.ph_mss;
614 } else
615 len = ntohs(ip->ip_len);
616
617 /* Check if we are allowed to fragment */
618 dst = ip->ip_dst;
619 rtableid = m->m_pkthdr.ph_rtableid;
620 if (ip_mtudisc && (ip->ip_off & htons(IP_DF)) && tdb->tdb_mtu &&
621 len > tdb->tdb_mtu && tdb->tdb_mtutimeout > gettime()) {
622 int transportmode;
623
624 transportmode = (tdb->tdb_dst.sa.sa_family == AF_INET) &&
625 (tdb->tdb_dst.sin.sin_addr.s_addr == dst.s_addr);
626 ip_output_ipsec_pmtu_update(tdb, ro, dst, rtableid,
627 transportmode);
628 ipsec_adjust_mtu(m, tdb->tdb_mtu);
629 m_freem(m);
630 return EMSGSIZE;
631 }
632 /* propagate IP_DF for v4-over-v6 */
633 if (ip_mtudisc && ip->ip_off & htons(IP_DF))
634 SET(m->m_pkthdr.csum_flags, M_IPV6_DF_OUT);
635
636 /*
637 * Clear these -- they'll be set in the recursive invocation
638 * as needed.
639 */
640 m->m_flags &= ~(M_MCAST | M_BCAST);
641
642 if (tso) {
643 error = tcp_chopper(m, &ml, encif, len);
644 if (error)
645 goto done;
646 } else {
647 CLR(m->m_pkthdr.csum_flags, M_TCP_TSO);
648 in_proto_cksum_out(m, encif);
649 ml_init(&ml);
650 ml_enqueue(&ml, m);
651 }
652
653 KERNEL_LOCK();
654 while ((m = ml_dequeue(&ml)) != NULL) {
655 /* Callee frees mbuf */
656 error = ipsp_process_packet(m, tdb, AF_INET, 0);
657 if (error)
658 break;
659 }
660 KERNEL_UNLOCK();
661 done:
662 if (error) {
663 ml_purge(&ml);
664 ipsecstat_inc(ipsec_odrops);
665 tdbstat_inc(tdb, tdb_odrops);
666 }
667 if (!error && tso)
668 tcpstat_inc(tcps_outswtso);
669 if (ip_mtudisc && error == EMSGSIZE)
670 ip_output_ipsec_pmtu_update(tdb, ro, dst, rtableid, 0);
671 return error;
672 }
673 #endif /* IPSEC */
674
675 int
ip_fragment(struct mbuf * m0,struct mbuf_list * ml,struct ifnet * ifp,u_long mtu)676 ip_fragment(struct mbuf *m0, struct mbuf_list *ml, struct ifnet *ifp,
677 u_long mtu)
678 {
679 struct ip *ip;
680 int firstlen, hlen, tlen, len, off;
681 int error;
682
683 ml_init(ml);
684 ml_enqueue(ml, m0);
685
686 ip = mtod(m0, struct ip *);
687 hlen = ip->ip_hl << 2;
688 tlen = m0->m_pkthdr.len;
689 len = (mtu - hlen) &~ 7;
690 if (len < 8) {
691 error = EMSGSIZE;
692 goto bad;
693 }
694 firstlen = len;
695
696 /*
697 * If we are doing fragmentation, we can't defer TCP/UDP
698 * checksumming; compute the checksum and clear the flag.
699 */
700 in_proto_cksum_out(m0, NULL);
701
702 /*
703 * Loop through length of payload after first fragment,
704 * make new header and copy data of each part and link onto chain.
705 */
706 for (off = hlen + firstlen; off < tlen; off += len) {
707 struct mbuf *m;
708 struct ip *mhip;
709 int mhlen;
710
711 MGETHDR(m, M_DONTWAIT, MT_HEADER);
712 if (m == NULL) {
713 error = ENOBUFS;
714 goto bad;
715 }
716 ml_enqueue(ml, m);
717 if ((error = m_dup_pkthdr(m, m0, M_DONTWAIT)) != 0)
718 goto bad;
719 m->m_data += max_linkhdr;
720 mhip = mtod(m, struct ip *);
721 *mhip = *ip;
722 if (hlen > sizeof(struct ip)) {
723 mhlen = ip_optcopy(ip, mhip) + sizeof(struct ip);
724 mhip->ip_hl = mhlen >> 2;
725 } else
726 mhlen = sizeof(struct ip);
727 m->m_len = mhlen;
728
729 mhip->ip_off = ((off - hlen) >> 3) +
730 (ntohs(ip->ip_off) & ~IP_MF);
731 if (ip->ip_off & htons(IP_MF))
732 mhip->ip_off |= IP_MF;
733 if (off + len >= tlen)
734 len = tlen - off;
735 else
736 mhip->ip_off |= IP_MF;
737 mhip->ip_off = htons(mhip->ip_off);
738
739 m->m_pkthdr.len = mhlen + len;
740 mhip->ip_len = htons(m->m_pkthdr.len);
741 m->m_next = m_copym(m0, off, len, M_NOWAIT);
742 if (m->m_next == NULL) {
743 error = ENOBUFS;
744 goto bad;
745 }
746
747 in_hdr_cksum_out(m, ifp);
748 }
749
750 /*
751 * Update first fragment by trimming what's been copied out
752 * and updating header, then send each fragment (in order).
753 */
754 if (hlen + firstlen < tlen) {
755 m_adj(m0, hlen + firstlen - tlen);
756 ip->ip_off |= htons(IP_MF);
757 }
758 ip->ip_len = htons(m0->m_pkthdr.len);
759
760 in_hdr_cksum_out(m0, ifp);
761
762 ipstat_add(ips_ofragments, ml_len(ml));
763 return (0);
764
765 bad:
766 ipstat_inc(ips_odropped);
767 ml_purge(ml);
768 return (error);
769 }
770
771 /*
772 * Insert IP options into preformed packet.
773 * Adjust IP destination as required for IP source routing,
774 * as indicated by a non-zero in_addr at the start of the options.
775 */
776 struct mbuf *
ip_insertoptions(struct mbuf * m,struct mbuf * opt,int * phlen)777 ip_insertoptions(struct mbuf *m, struct mbuf *opt, int *phlen)
778 {
779 struct ipoption *p = mtod(opt, struct ipoption *);
780 struct mbuf *n;
781 struct ip *ip = mtod(m, struct ip *);
782 unsigned int optlen;
783
784 optlen = opt->m_len - sizeof(p->ipopt_dst);
785 if (optlen + ntohs(ip->ip_len) > IP_MAXPACKET)
786 return (m); /* XXX should fail */
787
788 /* check if options will fit to IP header */
789 if ((optlen + sizeof(struct ip)) > (0x0f << 2)) {
790 *phlen = sizeof(struct ip);
791 return (m);
792 }
793
794 if (p->ipopt_dst.s_addr)
795 ip->ip_dst = p->ipopt_dst;
796 if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) {
797 MGETHDR(n, M_DONTWAIT, MT_HEADER);
798 if (n == NULL)
799 return (m);
800 M_MOVE_HDR(n, m);
801 n->m_pkthdr.len += optlen;
802 m->m_len -= sizeof(struct ip);
803 m->m_data += sizeof(struct ip);
804 n->m_next = m;
805 m = n;
806 m->m_len = optlen + sizeof(struct ip);
807 m->m_data += max_linkhdr;
808 memcpy(mtod(m, caddr_t), ip, sizeof(struct ip));
809 } else {
810 m->m_data -= optlen;
811 m->m_len += optlen;
812 m->m_pkthdr.len += optlen;
813 memmove(mtod(m, caddr_t), (caddr_t)ip, sizeof(struct ip));
814 }
815 ip = mtod(m, struct ip *);
816 memcpy(ip + 1, p->ipopt_list, optlen);
817 *phlen = sizeof(struct ip) + optlen;
818 ip->ip_len = htons(ntohs(ip->ip_len) + optlen);
819 return (m);
820 }
821
822 /*
823 * Copy options from ip to jp,
824 * omitting those not copied during fragmentation.
825 */
826 int
ip_optcopy(struct ip * ip,struct ip * jp)827 ip_optcopy(struct ip *ip, struct ip *jp)
828 {
829 u_char *cp, *dp;
830 int opt, optlen, cnt;
831
832 cp = (u_char *)(ip + 1);
833 dp = (u_char *)(jp + 1);
834 cnt = (ip->ip_hl << 2) - sizeof (struct ip);
835 for (; cnt > 0; cnt -= optlen, cp += optlen) {
836 opt = cp[0];
837 if (opt == IPOPT_EOL)
838 break;
839 if (opt == IPOPT_NOP) {
840 /* Preserve for IP mcast tunnel's LSRR alignment. */
841 *dp++ = IPOPT_NOP;
842 optlen = 1;
843 continue;
844 }
845 #ifdef DIAGNOSTIC
846 if (cnt < IPOPT_OLEN + sizeof(*cp))
847 panic("malformed IPv4 option passed to ip_optcopy");
848 #endif
849 optlen = cp[IPOPT_OLEN];
850 #ifdef DIAGNOSTIC
851 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt)
852 panic("malformed IPv4 option passed to ip_optcopy");
853 #endif
854 /* bogus lengths should have been caught by ip_dooptions */
855 if (optlen > cnt)
856 optlen = cnt;
857 if (IPOPT_COPIED(opt)) {
858 memcpy(dp, cp, optlen);
859 dp += optlen;
860 }
861 }
862 for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++)
863 *dp++ = IPOPT_EOL;
864 return (optlen);
865 }
866
867 /*
868 * IP socket option processing.
869 */
870 int
ip_ctloutput(int op,struct socket * so,int level,int optname,struct mbuf * m)871 ip_ctloutput(int op, struct socket *so, int level, int optname,
872 struct mbuf *m)
873 {
874 struct inpcb *inp = sotoinpcb(so);
875 int optval = 0;
876 struct proc *p = curproc; /* XXX */
877 int error = 0;
878 u_int rtableid, rtid = 0;
879
880 if (level != IPPROTO_IP)
881 return (EINVAL);
882
883 rtableid = p->p_p->ps_rtableid;
884
885 switch (op) {
886 case PRCO_SETOPT:
887 switch (optname) {
888 case IP_OPTIONS:
889 return (ip_pcbopts(&inp->inp_options, m));
890
891 case IP_TOS:
892 case IP_TTL:
893 case IP_MINTTL:
894 case IP_RECVOPTS:
895 case IP_RECVRETOPTS:
896 case IP_RECVDSTADDR:
897 case IP_RECVIF:
898 case IP_RECVTTL:
899 case IP_RECVDSTPORT:
900 case IP_RECVRTABLE:
901 case IP_IPSECFLOWINFO:
902 if (m == NULL || m->m_len != sizeof(int))
903 error = EINVAL;
904 else {
905 optval = *mtod(m, int *);
906 switch (optname) {
907
908 case IP_TOS:
909 inp->inp_ip.ip_tos = optval;
910 break;
911
912 case IP_TTL:
913 if (optval > 0 && optval <= MAXTTL)
914 inp->inp_ip.ip_ttl = optval;
915 else if (optval == -1)
916 inp->inp_ip.ip_ttl = ip_defttl;
917 else
918 error = EINVAL;
919 break;
920
921 case IP_MINTTL:
922 if (optval >= 0 && optval <= MAXTTL)
923 inp->inp_ip_minttl = optval;
924 else
925 error = EINVAL;
926 break;
927 #define OPTSET(bit) \
928 if (optval) \
929 inp->inp_flags |= bit; \
930 else \
931 inp->inp_flags &= ~bit;
932
933 case IP_RECVOPTS:
934 OPTSET(INP_RECVOPTS);
935 break;
936
937 case IP_RECVRETOPTS:
938 OPTSET(INP_RECVRETOPTS);
939 break;
940
941 case IP_RECVDSTADDR:
942 OPTSET(INP_RECVDSTADDR);
943 break;
944 case IP_RECVIF:
945 OPTSET(INP_RECVIF);
946 break;
947 case IP_RECVTTL:
948 OPTSET(INP_RECVTTL);
949 break;
950 case IP_RECVDSTPORT:
951 OPTSET(INP_RECVDSTPORT);
952 break;
953 case IP_RECVRTABLE:
954 OPTSET(INP_RECVRTABLE);
955 break;
956 case IP_IPSECFLOWINFO:
957 OPTSET(INP_IPSECFLOWINFO);
958 break;
959 }
960 }
961 break;
962 #undef OPTSET
963
964 case IP_MULTICAST_IF:
965 case IP_MULTICAST_TTL:
966 case IP_MULTICAST_LOOP:
967 case IP_ADD_MEMBERSHIP:
968 case IP_DROP_MEMBERSHIP:
969 error = ip_setmoptions(optname, &inp->inp_moptions, m,
970 inp->inp_rtableid);
971 break;
972
973 case IP_PORTRANGE:
974 if (m == NULL || m->m_len != sizeof(int))
975 error = EINVAL;
976 else {
977 optval = *mtod(m, int *);
978
979 switch (optval) {
980
981 case IP_PORTRANGE_DEFAULT:
982 inp->inp_flags &= ~(INP_LOWPORT);
983 inp->inp_flags &= ~(INP_HIGHPORT);
984 break;
985
986 case IP_PORTRANGE_HIGH:
987 inp->inp_flags &= ~(INP_LOWPORT);
988 inp->inp_flags |= INP_HIGHPORT;
989 break;
990
991 case IP_PORTRANGE_LOW:
992 inp->inp_flags &= ~(INP_HIGHPORT);
993 inp->inp_flags |= INP_LOWPORT;
994 break;
995
996 default:
997
998 error = EINVAL;
999 break;
1000 }
1001 }
1002 break;
1003 case IP_AUTH_LEVEL:
1004 case IP_ESP_TRANS_LEVEL:
1005 case IP_ESP_NETWORK_LEVEL:
1006 case IP_IPCOMP_LEVEL:
1007 #ifndef IPSEC
1008 error = EOPNOTSUPP;
1009 #else
1010 if (m == NULL || m->m_len != sizeof(int)) {
1011 error = EINVAL;
1012 break;
1013 }
1014 optval = *mtod(m, int *);
1015
1016 if (optval < IPSEC_LEVEL_BYPASS ||
1017 optval > IPSEC_LEVEL_UNIQUE) {
1018 error = EINVAL;
1019 break;
1020 }
1021
1022 switch (optname) {
1023 case IP_AUTH_LEVEL:
1024 if (optval < IPSEC_AUTH_LEVEL_DEFAULT &&
1025 suser(p)) {
1026 error = EACCES;
1027 break;
1028 }
1029 inp->inp_seclevel.sl_auth = optval;
1030 break;
1031
1032 case IP_ESP_TRANS_LEVEL:
1033 if (optval < IPSEC_ESP_TRANS_LEVEL_DEFAULT &&
1034 suser(p)) {
1035 error = EACCES;
1036 break;
1037 }
1038 inp->inp_seclevel.sl_esp_trans = optval;
1039 break;
1040
1041 case IP_ESP_NETWORK_LEVEL:
1042 if (optval < IPSEC_ESP_NETWORK_LEVEL_DEFAULT &&
1043 suser(p)) {
1044 error = EACCES;
1045 break;
1046 }
1047 inp->inp_seclevel.sl_esp_network = optval;
1048 break;
1049 case IP_IPCOMP_LEVEL:
1050 if (optval < IPSEC_IPCOMP_LEVEL_DEFAULT &&
1051 suser(p)) {
1052 error = EACCES;
1053 break;
1054 }
1055 inp->inp_seclevel.sl_ipcomp = optval;
1056 break;
1057 }
1058 #endif
1059 break;
1060
1061 case IP_IPSEC_LOCAL_ID:
1062 case IP_IPSEC_REMOTE_ID:
1063 error = EOPNOTSUPP;
1064 break;
1065 case SO_RTABLE:
1066 if (m == NULL || m->m_len < sizeof(u_int)) {
1067 error = EINVAL;
1068 break;
1069 }
1070 rtid = *mtod(m, u_int *);
1071 if (inp->inp_rtableid == rtid)
1072 break;
1073 /* needs privileges to switch when already set */
1074 if (rtableid != rtid && rtableid != 0 &&
1075 (error = suser(p)) != 0)
1076 break;
1077 error = in_pcbset_rtableid(inp, rtid);
1078 break;
1079 case IP_PIPEX:
1080 if (m != NULL && m->m_len == sizeof(int))
1081 inp->inp_pipex = *mtod(m, int *);
1082 else
1083 error = EINVAL;
1084 break;
1085
1086 default:
1087 error = ENOPROTOOPT;
1088 break;
1089 }
1090 break;
1091
1092 case PRCO_GETOPT:
1093 switch (optname) {
1094 case IP_OPTIONS:
1095 case IP_RETOPTS:
1096 if (inp->inp_options) {
1097 m->m_len = inp->inp_options->m_len;
1098 memcpy(mtod(m, caddr_t),
1099 mtod(inp->inp_options, caddr_t), m->m_len);
1100 } else
1101 m->m_len = 0;
1102 break;
1103
1104 case IP_TOS:
1105 case IP_TTL:
1106 case IP_MINTTL:
1107 case IP_RECVOPTS:
1108 case IP_RECVRETOPTS:
1109 case IP_RECVDSTADDR:
1110 case IP_RECVIF:
1111 case IP_RECVTTL:
1112 case IP_RECVDSTPORT:
1113 case IP_RECVRTABLE:
1114 case IP_IPSECFLOWINFO:
1115 case IP_IPDEFTTL:
1116 m->m_len = sizeof(int);
1117 switch (optname) {
1118
1119 case IP_TOS:
1120 optval = inp->inp_ip.ip_tos;
1121 break;
1122
1123 case IP_TTL:
1124 optval = inp->inp_ip.ip_ttl;
1125 break;
1126
1127 case IP_MINTTL:
1128 optval = inp->inp_ip_minttl;
1129 break;
1130
1131 case IP_IPDEFTTL:
1132 optval = ip_defttl;
1133 break;
1134
1135 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0)
1136
1137 case IP_RECVOPTS:
1138 optval = OPTBIT(INP_RECVOPTS);
1139 break;
1140
1141 case IP_RECVRETOPTS:
1142 optval = OPTBIT(INP_RECVRETOPTS);
1143 break;
1144
1145 case IP_RECVDSTADDR:
1146 optval = OPTBIT(INP_RECVDSTADDR);
1147 break;
1148 case IP_RECVIF:
1149 optval = OPTBIT(INP_RECVIF);
1150 break;
1151 case IP_RECVTTL:
1152 optval = OPTBIT(INP_RECVTTL);
1153 break;
1154 case IP_RECVDSTPORT:
1155 optval = OPTBIT(INP_RECVDSTPORT);
1156 break;
1157 case IP_RECVRTABLE:
1158 optval = OPTBIT(INP_RECVRTABLE);
1159 break;
1160 case IP_IPSECFLOWINFO:
1161 optval = OPTBIT(INP_IPSECFLOWINFO);
1162 break;
1163 }
1164 *mtod(m, int *) = optval;
1165 break;
1166
1167 case IP_MULTICAST_IF:
1168 case IP_MULTICAST_TTL:
1169 case IP_MULTICAST_LOOP:
1170 case IP_ADD_MEMBERSHIP:
1171 case IP_DROP_MEMBERSHIP:
1172 error = ip_getmoptions(optname, inp->inp_moptions, m);
1173 break;
1174
1175 case IP_PORTRANGE:
1176 m->m_len = sizeof(int);
1177
1178 if (inp->inp_flags & INP_HIGHPORT)
1179 optval = IP_PORTRANGE_HIGH;
1180 else if (inp->inp_flags & INP_LOWPORT)
1181 optval = IP_PORTRANGE_LOW;
1182 else
1183 optval = 0;
1184
1185 *mtod(m, int *) = optval;
1186 break;
1187
1188 case IP_AUTH_LEVEL:
1189 case IP_ESP_TRANS_LEVEL:
1190 case IP_ESP_NETWORK_LEVEL:
1191 case IP_IPCOMP_LEVEL:
1192 #ifndef IPSEC
1193 m->m_len = sizeof(int);
1194 *mtod(m, int *) = IPSEC_LEVEL_NONE;
1195 #else
1196 m->m_len = sizeof(int);
1197 switch (optname) {
1198 case IP_AUTH_LEVEL:
1199 optval = inp->inp_seclevel.sl_auth;
1200 break;
1201
1202 case IP_ESP_TRANS_LEVEL:
1203 optval = inp->inp_seclevel.sl_esp_trans;
1204 break;
1205
1206 case IP_ESP_NETWORK_LEVEL:
1207 optval = inp->inp_seclevel.sl_esp_network;
1208 break;
1209 case IP_IPCOMP_LEVEL:
1210 optval = inp->inp_seclevel.sl_ipcomp;
1211 break;
1212 }
1213 *mtod(m, int *) = optval;
1214 #endif
1215 break;
1216 case IP_IPSEC_LOCAL_ID:
1217 case IP_IPSEC_REMOTE_ID:
1218 error = EOPNOTSUPP;
1219 break;
1220 case SO_RTABLE:
1221 m->m_len = sizeof(u_int);
1222 *mtod(m, u_int *) = inp->inp_rtableid;
1223 break;
1224 case IP_PIPEX:
1225 m->m_len = sizeof(int);
1226 *mtod(m, int *) = inp->inp_pipex;
1227 break;
1228 default:
1229 error = ENOPROTOOPT;
1230 break;
1231 }
1232 break;
1233 }
1234 return (error);
1235 }
1236
1237 /*
1238 * Set up IP options in pcb for insertion in output packets.
1239 * Store in mbuf with pointer in pcbopt, adding pseudo-option
1240 * with destination address if source routed.
1241 */
1242 int
ip_pcbopts(struct mbuf ** pcbopt,struct mbuf * m)1243 ip_pcbopts(struct mbuf **pcbopt, struct mbuf *m)
1244 {
1245 struct mbuf *n;
1246 struct ipoption *p;
1247 int cnt, off, optlen;
1248 u_char *cp;
1249 u_char opt;
1250
1251 /* turn off any old options */
1252 m_freem(*pcbopt);
1253 *pcbopt = NULL;
1254 if (m == NULL || m->m_len == 0) {
1255 /*
1256 * Only turning off any previous options.
1257 */
1258 return (0);
1259 }
1260
1261 if (m->m_len % sizeof(int32_t) ||
1262 m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr))
1263 return (EINVAL);
1264
1265 /* Don't sleep because NET_LOCK() is hold. */
1266 if ((n = m_get(M_NOWAIT, MT_SOOPTS)) == NULL)
1267 return (ENOBUFS);
1268 p = mtod(n, struct ipoption *);
1269 memset(p, 0, sizeof (*p)); /* 0 = IPOPT_EOL, needed for padding */
1270 n->m_len = sizeof(struct in_addr);
1271
1272 off = 0;
1273 cnt = m->m_len;
1274 cp = mtod(m, u_char *);
1275
1276 while (cnt > 0) {
1277 opt = cp[IPOPT_OPTVAL];
1278
1279 if (opt == IPOPT_NOP || opt == IPOPT_EOL) {
1280 optlen = 1;
1281 } else {
1282 if (cnt < IPOPT_OLEN + sizeof(*cp))
1283 goto bad;
1284 optlen = cp[IPOPT_OLEN];
1285 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt)
1286 goto bad;
1287 }
1288 switch (opt) {
1289 default:
1290 memcpy(p->ipopt_list + off, cp, optlen);
1291 break;
1292
1293 case IPOPT_LSRR:
1294 case IPOPT_SSRR:
1295 /*
1296 * user process specifies route as:
1297 * ->A->B->C->D
1298 * D must be our final destination (but we can't
1299 * check that since we may not have connected yet).
1300 * A is first hop destination, which doesn't appear in
1301 * actual IP option, but is stored before the options.
1302 */
1303 if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr))
1304 goto bad;
1305
1306 /*
1307 * Optlen is smaller because first address is popped.
1308 * Cnt and cp will be adjusted a bit later to reflect
1309 * this.
1310 */
1311 optlen -= sizeof(struct in_addr);
1312 p->ipopt_list[off + IPOPT_OPTVAL] = opt;
1313 p->ipopt_list[off + IPOPT_OLEN] = optlen;
1314
1315 /*
1316 * Move first hop before start of options.
1317 */
1318 memcpy(&p->ipopt_dst, cp + IPOPT_OFFSET,
1319 sizeof(struct in_addr));
1320 cp += sizeof(struct in_addr);
1321 cnt -= sizeof(struct in_addr);
1322 /*
1323 * Then copy rest of options
1324 */
1325 memcpy(p->ipopt_list + off + IPOPT_OFFSET,
1326 cp + IPOPT_OFFSET, optlen - IPOPT_OFFSET);
1327 break;
1328 }
1329 off += optlen;
1330 cp += optlen;
1331 cnt -= optlen;
1332
1333 if (opt == IPOPT_EOL)
1334 break;
1335 }
1336 /* pad options to next word, since p was zeroed just adjust off */
1337 off = (off + sizeof(int32_t) - 1) & ~(sizeof(int32_t) - 1);
1338 n->m_len += off;
1339 if (n->m_len > sizeof(*p)) {
1340 bad:
1341 m_freem(n);
1342 return (EINVAL);
1343 }
1344
1345 *pcbopt = n;
1346 return (0);
1347 }
1348
1349 /*
1350 * Lookup the interface based on the information in the ip_mreqn struct.
1351 */
1352 int
ip_multicast_if(struct ip_mreqn * mreq,u_int rtableid,unsigned int * ifidx)1353 ip_multicast_if(struct ip_mreqn *mreq, u_int rtableid, unsigned int *ifidx)
1354 {
1355 struct sockaddr_in sin;
1356 struct rtentry *rt;
1357
1358 /*
1359 * In case userland provides the imr_ifindex use this as interface.
1360 * If no interface address was provided, use the interface of
1361 * the route to the given multicast address.
1362 */
1363 if (mreq->imr_ifindex != 0) {
1364 *ifidx = mreq->imr_ifindex;
1365 } else if (mreq->imr_address.s_addr == INADDR_ANY) {
1366 memset(&sin, 0, sizeof(sin));
1367 sin.sin_len = sizeof(sin);
1368 sin.sin_family = AF_INET;
1369 sin.sin_addr = mreq->imr_multiaddr;
1370 rt = rtalloc(sintosa(&sin), RT_RESOLVE, rtableid);
1371 if (!rtisvalid(rt)) {
1372 rtfree(rt);
1373 return EADDRNOTAVAIL;
1374 }
1375 *ifidx = rt->rt_ifidx;
1376 rtfree(rt);
1377 } else {
1378 memset(&sin, 0, sizeof(sin));
1379 sin.sin_len = sizeof(sin);
1380 sin.sin_family = AF_INET;
1381 sin.sin_addr = mreq->imr_address;
1382 rt = rtalloc(sintosa(&sin), 0, rtableid);
1383 if (!rtisvalid(rt) || !ISSET(rt->rt_flags, RTF_LOCAL)) {
1384 rtfree(rt);
1385 return EADDRNOTAVAIL;
1386 }
1387 *ifidx = rt->rt_ifidx;
1388 rtfree(rt);
1389 }
1390
1391 return 0;
1392 }
1393
1394 /*
1395 * Set the IP multicast options in response to user setsockopt().
1396 */
1397 int
ip_setmoptions(int optname,struct ip_moptions ** imop,struct mbuf * m,u_int rtableid)1398 ip_setmoptions(int optname, struct ip_moptions **imop, struct mbuf *m,
1399 u_int rtableid)
1400 {
1401 struct in_addr addr;
1402 struct in_ifaddr *ia;
1403 struct ip_mreqn mreqn;
1404 struct ifnet *ifp = NULL;
1405 struct ip_moptions *imo = *imop;
1406 struct in_multi **immp;
1407 struct sockaddr_in sin;
1408 unsigned int ifidx;
1409 int i, error = 0;
1410 u_char loop;
1411
1412 if (imo == NULL) {
1413 /*
1414 * No multicast option buffer attached to the pcb;
1415 * allocate one and initialize to default values.
1416 */
1417 imo = malloc(sizeof(*imo), M_IPMOPTS, M_WAITOK|M_ZERO);
1418 immp = mallocarray(IP_MIN_MEMBERSHIPS, sizeof(*immp), M_IPMOPTS,
1419 M_WAITOK|M_ZERO);
1420 *imop = imo;
1421 imo->imo_ifidx = 0;
1422 imo->imo_ttl = IP_DEFAULT_MULTICAST_TTL;
1423 imo->imo_loop = IP_DEFAULT_MULTICAST_LOOP;
1424 imo->imo_num_memberships = 0;
1425 imo->imo_max_memberships = IP_MIN_MEMBERSHIPS;
1426 imo->imo_membership = immp;
1427 }
1428
1429 switch (optname) {
1430
1431 case IP_MULTICAST_IF:
1432 /*
1433 * Select the interface for outgoing multicast packets.
1434 */
1435 if (m == NULL) {
1436 error = EINVAL;
1437 break;
1438 }
1439 if (m->m_len == sizeof(struct in_addr)) {
1440 addr = *(mtod(m, struct in_addr *));
1441 } else if (m->m_len == sizeof(struct ip_mreq) ||
1442 m->m_len == sizeof(struct ip_mreqn)) {
1443 memset(&mreqn, 0, sizeof(mreqn));
1444 memcpy(&mreqn, mtod(m, void *), m->m_len);
1445
1446 /*
1447 * If an interface index is given use this
1448 * index to set the imo_ifidx but check first
1449 * that the interface actually exists.
1450 * In the other case just set the addr to
1451 * the imr_address and fall through to the
1452 * regular code.
1453 */
1454 if (mreqn.imr_ifindex != 0) {
1455 ifp = if_get(mreqn.imr_ifindex);
1456 if (ifp == NULL ||
1457 ifp->if_rdomain != rtable_l2(rtableid)) {
1458 error = EADDRNOTAVAIL;
1459 if_put(ifp);
1460 break;
1461 }
1462 imo->imo_ifidx = ifp->if_index;
1463 if_put(ifp);
1464 break;
1465 } else
1466 addr = mreqn.imr_address;
1467 } else {
1468 error = EINVAL;
1469 break;
1470 }
1471 /*
1472 * INADDR_ANY is used to remove a previous selection.
1473 * When no interface is selected, a default one is
1474 * chosen every time a multicast packet is sent.
1475 */
1476 if (addr.s_addr == INADDR_ANY) {
1477 imo->imo_ifidx = 0;
1478 break;
1479 }
1480 /*
1481 * The selected interface is identified by its local
1482 * IP address. Find the interface and confirm that
1483 * it supports multicasting.
1484 */
1485 memset(&sin, 0, sizeof(sin));
1486 sin.sin_len = sizeof(sin);
1487 sin.sin_family = AF_INET;
1488 sin.sin_addr = addr;
1489 ia = ifatoia(ifa_ifwithaddr(sintosa(&sin), rtableid));
1490 if (ia == NULL ||
1491 (ia->ia_ifp->if_flags & IFF_MULTICAST) == 0) {
1492 error = EADDRNOTAVAIL;
1493 break;
1494 }
1495 imo->imo_ifidx = ia->ia_ifp->if_index;
1496 break;
1497
1498 case IP_MULTICAST_TTL:
1499 /*
1500 * Set the IP time-to-live for outgoing multicast packets.
1501 */
1502 if (m == NULL || m->m_len != 1) {
1503 error = EINVAL;
1504 break;
1505 }
1506 imo->imo_ttl = *(mtod(m, u_char *));
1507 break;
1508
1509 case IP_MULTICAST_LOOP:
1510 /*
1511 * Set the loopback flag for outgoing multicast packets.
1512 * Must be zero or one.
1513 */
1514 if (m == NULL || m->m_len != 1 ||
1515 (loop = *(mtod(m, u_char *))) > 1) {
1516 error = EINVAL;
1517 break;
1518 }
1519 imo->imo_loop = loop;
1520 break;
1521
1522 case IP_ADD_MEMBERSHIP:
1523 /*
1524 * Add a multicast group membership.
1525 * Group must be a valid IP multicast address.
1526 */
1527 if (m == NULL || !(m->m_len == sizeof(struct ip_mreq) ||
1528 m->m_len == sizeof(struct ip_mreqn))) {
1529 error = EINVAL;
1530 break;
1531 }
1532 memset(&mreqn, 0, sizeof(mreqn));
1533 memcpy(&mreqn, mtod(m, void *), m->m_len);
1534 if (!IN_MULTICAST(mreqn.imr_multiaddr.s_addr)) {
1535 error = EINVAL;
1536 break;
1537 }
1538
1539 error = ip_multicast_if(&mreqn, rtableid, &ifidx);
1540 if (error)
1541 break;
1542
1543 /*
1544 * See if we found an interface, and confirm that it
1545 * supports multicast.
1546 */
1547 ifp = if_get(ifidx);
1548 if (ifp == NULL || ifp->if_rdomain != rtable_l2(rtableid) ||
1549 (ifp->if_flags & IFF_MULTICAST) == 0) {
1550 error = EADDRNOTAVAIL;
1551 if_put(ifp);
1552 break;
1553 }
1554
1555 /*
1556 * See if the membership already exists or if all the
1557 * membership slots are full.
1558 */
1559 for (i = 0; i < imo->imo_num_memberships; ++i) {
1560 if (imo->imo_membership[i]->inm_ifidx == ifidx &&
1561 imo->imo_membership[i]->inm_addr.s_addr
1562 == mreqn.imr_multiaddr.s_addr)
1563 break;
1564 }
1565 if (i < imo->imo_num_memberships) {
1566 error = EADDRINUSE;
1567 if_put(ifp);
1568 break;
1569 }
1570 if (imo->imo_num_memberships == imo->imo_max_memberships) {
1571 struct in_multi **nmships, **omships;
1572 size_t newmax;
1573 /*
1574 * Resize the vector to next power-of-two minus 1. If
1575 * the size would exceed the maximum then we know we've
1576 * really run out of entries. Otherwise, we reallocate
1577 * the vector.
1578 */
1579 nmships = NULL;
1580 omships = imo->imo_membership;
1581 newmax = ((imo->imo_max_memberships + 1) * 2) - 1;
1582 if (newmax <= IP_MAX_MEMBERSHIPS) {
1583 nmships = mallocarray(newmax, sizeof(*nmships),
1584 M_IPMOPTS, M_NOWAIT|M_ZERO);
1585 if (nmships != NULL) {
1586 memcpy(nmships, omships,
1587 sizeof(*omships) *
1588 imo->imo_max_memberships);
1589 free(omships, M_IPMOPTS,
1590 sizeof(*omships) *
1591 imo->imo_max_memberships);
1592 imo->imo_membership = nmships;
1593 imo->imo_max_memberships = newmax;
1594 }
1595 }
1596 if (nmships == NULL) {
1597 error = ENOBUFS;
1598 if_put(ifp);
1599 break;
1600 }
1601 }
1602 /*
1603 * Everything looks good; add a new record to the multicast
1604 * address list for the given interface.
1605 */
1606 if ((imo->imo_membership[i] =
1607 in_addmulti(&mreqn.imr_multiaddr, ifp)) == NULL) {
1608 error = ENOBUFS;
1609 if_put(ifp);
1610 break;
1611 }
1612 ++imo->imo_num_memberships;
1613 if_put(ifp);
1614 break;
1615
1616 case IP_DROP_MEMBERSHIP:
1617 /*
1618 * Drop a multicast group membership.
1619 * Group must be a valid IP multicast address.
1620 */
1621 if (m == NULL || !(m->m_len == sizeof(struct ip_mreq) ||
1622 m->m_len == sizeof(struct ip_mreqn))) {
1623 error = EINVAL;
1624 break;
1625 }
1626 memset(&mreqn, 0, sizeof(mreqn));
1627 memcpy(&mreqn, mtod(m, void *), m->m_len);
1628 if (!IN_MULTICAST(mreqn.imr_multiaddr.s_addr)) {
1629 error = EINVAL;
1630 break;
1631 }
1632
1633 /*
1634 * If an interface address was specified, get a pointer
1635 * to its ifnet structure.
1636 */
1637 error = ip_multicast_if(&mreqn, rtableid, &ifidx);
1638 if (error)
1639 break;
1640
1641 /*
1642 * Find the membership in the membership array.
1643 */
1644 for (i = 0; i < imo->imo_num_memberships; ++i) {
1645 if ((ifidx == 0 ||
1646 imo->imo_membership[i]->inm_ifidx == ifidx) &&
1647 imo->imo_membership[i]->inm_addr.s_addr ==
1648 mreqn.imr_multiaddr.s_addr)
1649 break;
1650 }
1651 if (i == imo->imo_num_memberships) {
1652 error = EADDRNOTAVAIL;
1653 break;
1654 }
1655 /*
1656 * Give up the multicast address record to which the
1657 * membership points.
1658 */
1659 in_delmulti(imo->imo_membership[i]);
1660 /*
1661 * Remove the gap in the membership array.
1662 */
1663 for (++i; i < imo->imo_num_memberships; ++i)
1664 imo->imo_membership[i-1] = imo->imo_membership[i];
1665 --imo->imo_num_memberships;
1666 break;
1667
1668 default:
1669 error = EOPNOTSUPP;
1670 break;
1671 }
1672
1673 /*
1674 * If all options have default values, no need to keep the data.
1675 */
1676 if (imo->imo_ifidx == 0 &&
1677 imo->imo_ttl == IP_DEFAULT_MULTICAST_TTL &&
1678 imo->imo_loop == IP_DEFAULT_MULTICAST_LOOP &&
1679 imo->imo_num_memberships == 0) {
1680 free(imo->imo_membership , M_IPMOPTS,
1681 imo->imo_max_memberships * sizeof(struct in_multi *));
1682 free(*imop, M_IPMOPTS, sizeof(**imop));
1683 *imop = NULL;
1684 }
1685
1686 return (error);
1687 }
1688
1689 /*
1690 * Return the IP multicast options in response to user getsockopt().
1691 */
1692 int
ip_getmoptions(int optname,struct ip_moptions * imo,struct mbuf * m)1693 ip_getmoptions(int optname, struct ip_moptions *imo, struct mbuf *m)
1694 {
1695 u_char *ttl;
1696 u_char *loop;
1697 struct in_addr *addr;
1698 struct in_ifaddr *ia;
1699 struct ifnet *ifp;
1700
1701 switch (optname) {
1702
1703 case IP_MULTICAST_IF:
1704 addr = mtod(m, struct in_addr *);
1705 m->m_len = sizeof(struct in_addr);
1706 if (imo == NULL || (ifp = if_get(imo->imo_ifidx)) == NULL)
1707 addr->s_addr = INADDR_ANY;
1708 else {
1709 IFP_TO_IA(ifp, ia);
1710 addr->s_addr = (ia == NULL) ? INADDR_ANY
1711 : ia->ia_addr.sin_addr.s_addr;
1712 if_put(ifp);
1713 }
1714 return (0);
1715
1716 case IP_MULTICAST_TTL:
1717 ttl = mtod(m, u_char *);
1718 m->m_len = 1;
1719 *ttl = (imo == NULL) ? IP_DEFAULT_MULTICAST_TTL
1720 : imo->imo_ttl;
1721 return (0);
1722
1723 case IP_MULTICAST_LOOP:
1724 loop = mtod(m, u_char *);
1725 m->m_len = 1;
1726 *loop = (imo == NULL) ? IP_DEFAULT_MULTICAST_LOOP
1727 : imo->imo_loop;
1728 return (0);
1729
1730 default:
1731 return (EOPNOTSUPP);
1732 }
1733 }
1734
1735 /*
1736 * Discard the IP multicast options.
1737 */
1738 void
ip_freemoptions(struct ip_moptions * imo)1739 ip_freemoptions(struct ip_moptions *imo)
1740 {
1741 int i;
1742
1743 if (imo != NULL) {
1744 for (i = 0; i < imo->imo_num_memberships; ++i)
1745 in_delmulti(imo->imo_membership[i]);
1746 free(imo->imo_membership, M_IPMOPTS,
1747 imo->imo_max_memberships * sizeof(struct in_multi *));
1748 free(imo, M_IPMOPTS, sizeof(*imo));
1749 }
1750 }
1751
1752 /*
1753 * Routine called from ip_output() to loop back a copy of an IP multicast
1754 * packet to the input queue of a specified interface.
1755 */
1756 void
ip_mloopback(struct ifnet * ifp,struct mbuf * m,struct sockaddr_in * dst)1757 ip_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in *dst)
1758 {
1759 struct mbuf *copym;
1760
1761 copym = m_dup_pkt(m, max_linkhdr, M_DONTWAIT);
1762 if (copym != NULL) {
1763 /*
1764 * We don't bother to fragment if the IP length is greater
1765 * than the interface's MTU. Can this possibly matter?
1766 */
1767 in_hdr_cksum_out(copym, NULL);
1768 if_input_local(ifp, copym, dst->sin_family);
1769 }
1770 }
1771
1772 void
in_hdr_cksum_out(struct mbuf * m,struct ifnet * ifp)1773 in_hdr_cksum_out(struct mbuf *m, struct ifnet *ifp)
1774 {
1775 struct ip *ip = mtod(m, struct ip *);
1776
1777 ip->ip_sum = 0;
1778 if (in_ifcap_cksum(m, ifp, IFCAP_CSUM_IPv4)) {
1779 SET(m->m_pkthdr.csum_flags, M_IPV4_CSUM_OUT);
1780 } else {
1781 ipstat_inc(ips_outswcsum);
1782 ip->ip_sum = in_cksum(m, ip->ip_hl << 2);
1783 CLR(m->m_pkthdr.csum_flags, M_IPV4_CSUM_OUT);
1784 }
1785 }
1786
1787 /*
1788 * Compute significant parts of the IPv4 checksum pseudo-header
1789 * for use in a delayed TCP/UDP checksum calculation.
1790 */
1791 static u_int16_t
in_cksum_phdr(u_int32_t src,u_int32_t dst,u_int32_t lenproto)1792 in_cksum_phdr(u_int32_t src, u_int32_t dst, u_int32_t lenproto)
1793 {
1794 u_int32_t sum;
1795
1796 sum = lenproto +
1797 (u_int16_t)(src >> 16) +
1798 (u_int16_t)(src /*& 0xffff*/) +
1799 (u_int16_t)(dst >> 16) +
1800 (u_int16_t)(dst /*& 0xffff*/);
1801
1802 sum = (u_int16_t)(sum >> 16) + (u_int16_t)(sum /*& 0xffff*/);
1803
1804 if (sum > 0xffff)
1805 sum -= 0xffff;
1806
1807 return (sum);
1808 }
1809
1810 /*
1811 * Process a delayed payload checksum calculation.
1812 */
1813 void
in_delayed_cksum(struct mbuf * m)1814 in_delayed_cksum(struct mbuf *m)
1815 {
1816 struct ip *ip;
1817 u_int16_t csum, offset;
1818
1819 ip = mtod(m, struct ip *);
1820 offset = ip->ip_hl << 2;
1821 csum = in4_cksum(m, 0, offset, m->m_pkthdr.len - offset);
1822 if (csum == 0 && ip->ip_p == IPPROTO_UDP)
1823 csum = 0xffff;
1824
1825 switch (ip->ip_p) {
1826 case IPPROTO_TCP:
1827 offset += offsetof(struct tcphdr, th_sum);
1828 break;
1829
1830 case IPPROTO_UDP:
1831 offset += offsetof(struct udphdr, uh_sum);
1832 break;
1833
1834 case IPPROTO_ICMP:
1835 offset += offsetof(struct icmp, icmp_cksum);
1836 break;
1837
1838 default:
1839 return;
1840 }
1841
1842 if ((offset + sizeof(u_int16_t)) > m->m_len)
1843 m_copyback(m, offset, sizeof(csum), &csum, M_NOWAIT);
1844 else
1845 *(u_int16_t *)(mtod(m, caddr_t) + offset) = csum;
1846 }
1847
1848 void
in_proto_cksum_out(struct mbuf * m,struct ifnet * ifp)1849 in_proto_cksum_out(struct mbuf *m, struct ifnet *ifp)
1850 {
1851 struct ip *ip = mtod(m, struct ip *);
1852
1853 /* some hw and in_delayed_cksum need the pseudo header cksum */
1854 if (m->m_pkthdr.csum_flags &
1855 (M_TCP_CSUM_OUT|M_UDP_CSUM_OUT|M_ICMP_CSUM_OUT)) {
1856 u_int16_t csum = 0, offset;
1857
1858 offset = ip->ip_hl << 2;
1859 if (ISSET(m->m_pkthdr.csum_flags, M_TCP_TSO) &&
1860 in_ifcap_cksum(m, ifp, IFCAP_TSOv4)) {
1861 csum = in_cksum_phdr(ip->ip_src.s_addr,
1862 ip->ip_dst.s_addr, htonl(ip->ip_p));
1863 } else if (ISSET(m->m_pkthdr.csum_flags,
1864 M_TCP_CSUM_OUT|M_UDP_CSUM_OUT)) {
1865 csum = in_cksum_phdr(ip->ip_src.s_addr,
1866 ip->ip_dst.s_addr, htonl(ntohs(ip->ip_len) -
1867 offset + ip->ip_p));
1868 }
1869 if (ip->ip_p == IPPROTO_TCP)
1870 offset += offsetof(struct tcphdr, th_sum);
1871 else if (ip->ip_p == IPPROTO_UDP)
1872 offset += offsetof(struct udphdr, uh_sum);
1873 else if (ip->ip_p == IPPROTO_ICMP)
1874 offset += offsetof(struct icmp, icmp_cksum);
1875 if ((offset + sizeof(u_int16_t)) > m->m_len)
1876 m_copyback(m, offset, sizeof(csum), &csum, M_NOWAIT);
1877 else
1878 *(u_int16_t *)(mtod(m, caddr_t) + offset) = csum;
1879 }
1880
1881 if (m->m_pkthdr.csum_flags & M_TCP_CSUM_OUT) {
1882 if (!in_ifcap_cksum(m, ifp, IFCAP_CSUM_TCPv4) ||
1883 ip->ip_hl != 5) {
1884 tcpstat_inc(tcps_outswcsum);
1885 in_delayed_cksum(m);
1886 m->m_pkthdr.csum_flags &= ~M_TCP_CSUM_OUT; /* Clear */
1887 }
1888 } else if (m->m_pkthdr.csum_flags & M_UDP_CSUM_OUT) {
1889 if (!in_ifcap_cksum(m, ifp, IFCAP_CSUM_UDPv4) ||
1890 ip->ip_hl != 5) {
1891 udpstat_inc(udps_outswcsum);
1892 in_delayed_cksum(m);
1893 m->m_pkthdr.csum_flags &= ~M_UDP_CSUM_OUT; /* Clear */
1894 }
1895 } else if (m->m_pkthdr.csum_flags & M_ICMP_CSUM_OUT) {
1896 in_delayed_cksum(m);
1897 m->m_pkthdr.csum_flags &= ~M_ICMP_CSUM_OUT; /* Clear */
1898 }
1899 }
1900
1901 int
in_ifcap_cksum(struct mbuf * m,struct ifnet * ifp,int ifcap)1902 in_ifcap_cksum(struct mbuf *m, struct ifnet *ifp, int ifcap)
1903 {
1904 if ((ifp == NULL) ||
1905 !ISSET(ifp->if_capabilities, ifcap) ||
1906 (ifp->if_bridgeidx != 0))
1907 return (0);
1908 /*
1909 * Simplex interface sends packet back without hardware cksum.
1910 * Keep this check in sync with the condition where ether_resolve()
1911 * calls if_input_local().
1912 */
1913 if (ISSET(m->m_flags, M_BCAST) &&
1914 ISSET(ifp->if_flags, IFF_SIMPLEX) &&
1915 !m->m_pkthdr.pf.routed)
1916 return (0);
1917 return (1);
1918 }
1919