ip_input.c revision 1.154 1 /* $NetBSD: ip_input.c,v 1.154 2002/06/30 22:40:34 thorpej Exp $ */
2
3 /*
4 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5 * All rights reserved.
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 * 3. Neither the name of the project nor the names of its contributors
16 * may be used to endorse or promote products derived from this software
17 * without specific prior written permission.
18 *
19 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 * SUCH DAMAGE.
30 */
31
32 /*-
33 * Copyright (c) 1998 The NetBSD Foundation, Inc.
34 * All rights reserved.
35 *
36 * This code is derived from software contributed to The NetBSD Foundation
37 * by Public Access Networks Corporation ("Panix"). It was developed under
38 * contract to Panix by Eric Haszlakiewicz and Thor Lancelot Simon.
39 *
40 * Redistribution and use in source and binary forms, with or without
41 * modification, are permitted provided that the following conditions
42 * are met:
43 * 1. Redistributions of source code must retain the above copyright
44 * notice, this list of conditions and the following disclaimer.
45 * 2. Redistributions in binary form must reproduce the above copyright
46 * notice, this list of conditions and the following disclaimer in the
47 * documentation and/or other materials provided with the distribution.
48 * 3. All advertising materials mentioning features or use of this software
49 * must display the following acknowledgement:
50 * This product includes software developed by the NetBSD
51 * Foundation, Inc. and its contributors.
52 * 4. Neither the name of The NetBSD Foundation nor the names of its
53 * contributors may be used to endorse or promote products derived
54 * from this software without specific prior written permission.
55 *
56 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
57 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
58 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
59 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
60 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
61 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
62 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
63 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
64 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
65 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
66 * POSSIBILITY OF SUCH DAMAGE.
67 */
68
69 /*
70 * Copyright (c) 1982, 1986, 1988, 1993
71 * The Regents of the University of California. All rights reserved.
72 *
73 * Redistribution and use in source and binary forms, with or without
74 * modification, are permitted provided that the following conditions
75 * are met:
76 * 1. Redistributions of source code must retain the above copyright
77 * notice, this list of conditions and the following disclaimer.
78 * 2. Redistributions in binary form must reproduce the above copyright
79 * notice, this list of conditions and the following disclaimer in the
80 * documentation and/or other materials provided with the distribution.
81 * 3. All advertising materials mentioning features or use of this software
82 * must display the following acknowledgement:
83 * This product includes software developed by the University of
84 * California, Berkeley and its contributors.
85 * 4. Neither the name of the University nor the names of its contributors
86 * may be used to endorse or promote products derived from this software
87 * without specific prior written permission.
88 *
89 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
90 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
91 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
92 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
93 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
94 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
95 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
96 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
97 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
98 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
99 * SUCH DAMAGE.
100 *
101 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94
102 */
103
104 #include <sys/cdefs.h>
105 __KERNEL_RCSID(0, "$NetBSD: ip_input.c,v 1.154 2002/06/30 22:40:34 thorpej Exp $");
106
107 #include "opt_gateway.h"
108 #include "opt_pfil_hooks.h"
109 #include "opt_ipsec.h"
110 #include "opt_mrouting.h"
111 #include "opt_inet_csum.h"
112
113 #include <sys/param.h>
114 #include <sys/systm.h>
115 #include <sys/malloc.h>
116 #include <sys/mbuf.h>
117 #include <sys/domain.h>
118 #include <sys/protosw.h>
119 #include <sys/socket.h>
120 #include <sys/socketvar.h>
121 #include <sys/errno.h>
122 #include <sys/time.h>
123 #include <sys/kernel.h>
124 #include <sys/pool.h>
125 #include <sys/sysctl.h>
126
127 #include <net/if.h>
128 #include <net/if_dl.h>
129 #include <net/route.h>
130 #include <net/pfil.h>
131
132 #include <netinet/in.h>
133 #include <netinet/in_systm.h>
134 #include <netinet/ip.h>
135 #include <netinet/in_pcb.h>
136 #include <netinet/in_var.h>
137 #include <netinet/ip_var.h>
138 #include <netinet/ip_icmp.h>
139 /* just for gif_ttl */
140 #include <netinet/in_gif.h>
141 #include "gif.h"
142 #include <net/if_gre.h>
143 #include "gre.h"
144
145 #ifdef MROUTING
146 #include <netinet/ip_mroute.h>
147 #endif
148
149 #ifdef IPSEC
150 #include <netinet6/ipsec.h>
151 #include <netkey/key.h>
152 #endif
153
154 #ifndef IPFORWARDING
155 #ifdef GATEWAY
156 #define IPFORWARDING 1 /* forward IP packets not for us */
157 #else /* GATEWAY */
158 #define IPFORWARDING 0 /* don't forward IP packets not for us */
159 #endif /* GATEWAY */
160 #endif /* IPFORWARDING */
161 #ifndef IPSENDREDIRECTS
162 #define IPSENDREDIRECTS 1
163 #endif
164 #ifndef IPFORWSRCRT
165 #define IPFORWSRCRT 1 /* forward source-routed packets */
166 #endif
167 #ifndef IPALLOWSRCRT
168 #define IPALLOWSRCRT 1 /* allow source-routed packets */
169 #endif
170 #ifndef IPMTUDISC
171 #define IPMTUDISC 1
172 #endif
173 #ifndef IPMTUDISCTIMEOUT
174 #define IPMTUDISCTIMEOUT (10 * 60) /* as per RFC 1191 */
175 #endif
176
177 /*
178 * Note: DIRECTED_BROADCAST is handled this way so that previous
179 * configuration using this option will Just Work.
180 */
181 #ifndef IPDIRECTEDBCAST
182 #ifdef DIRECTED_BROADCAST
183 #define IPDIRECTEDBCAST 1
184 #else
185 #define IPDIRECTEDBCAST 0
186 #endif /* DIRECTED_BROADCAST */
187 #endif /* IPDIRECTEDBCAST */
188 int ipforwarding = IPFORWARDING;
189 int ipsendredirects = IPSENDREDIRECTS;
190 int ip_defttl = IPDEFTTL;
191 int ip_forwsrcrt = IPFORWSRCRT;
192 int ip_directedbcast = IPDIRECTEDBCAST;
193 int ip_allowsrcrt = IPALLOWSRCRT;
194 int ip_mtudisc = IPMTUDISC;
195 u_int ip_mtudisc_timeout = IPMTUDISCTIMEOUT;
196 #ifdef DIAGNOSTIC
197 int ipprintfs = 0;
198 #endif
199
200 struct rttimer_queue *ip_mtudisc_timeout_q = NULL;
201
202 extern struct domain inetdomain;
203 int ipqmaxlen = IFQ_MAXLEN;
204 u_long in_ifaddrhash; /* size of hash table - 1 */
205 int in_ifaddrentries; /* total number of addrs */
206 struct in_ifaddrhead in_ifaddr;
207 struct in_ifaddrhashhead *in_ifaddrhashtbl;
208 struct ifqueue ipintrq;
209 struct ipstat ipstat;
210 u_int16_t ip_id;
211
212 #ifdef PFIL_HOOKS
213 struct pfil_head inet_pfil_hook;
214 #endif
215
216 struct ipqhead ipq;
217 int ipq_locked;
218 int ip_nfragpackets = 0;
219 int ip_maxfragpackets = 200;
220
221 static __inline int ipq_lock_try __P((void));
222 static __inline void ipq_unlock __P((void));
223
224 static __inline int
225 ipq_lock_try()
226 {
227 int s;
228
229 /*
230 * Use splvm() -- we're blocking things that would cause
231 * mbuf allocation.
232 */
233 s = splvm();
234 if (ipq_locked) {
235 splx(s);
236 return (0);
237 }
238 ipq_locked = 1;
239 splx(s);
240 return (1);
241 }
242
243 static __inline void
244 ipq_unlock()
245 {
246 int s;
247
248 s = splvm();
249 ipq_locked = 0;
250 splx(s);
251 }
252
253 #ifdef DIAGNOSTIC
254 #define IPQ_LOCK() \
255 do { \
256 if (ipq_lock_try() == 0) { \
257 printf("%s:%d: ipq already locked\n", __FILE__, __LINE__); \
258 panic("ipq_lock"); \
259 } \
260 } while (0)
261 #define IPQ_LOCK_CHECK() \
262 do { \
263 if (ipq_locked == 0) { \
264 printf("%s:%d: ipq lock not held\n", __FILE__, __LINE__); \
265 panic("ipq lock check"); \
266 } \
267 } while (0)
268 #else
269 #define IPQ_LOCK() (void) ipq_lock_try()
270 #define IPQ_LOCK_CHECK() /* nothing */
271 #endif
272
273 #define IPQ_UNLOCK() ipq_unlock()
274
275 struct pool ipqent_pool;
276
277 #ifdef INET_CSUM_COUNTERS
278 #include <sys/device.h>
279
280 struct evcnt ip_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
281 NULL, "inet", "hwcsum bad");
282 struct evcnt ip_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
283 NULL, "inet", "hwcsum ok");
284 struct evcnt ip_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
285 NULL, "inet", "swcsum");
286
287 #define INET_CSUM_COUNTER_INCR(ev) (ev)->ev_count++
288
289 #else
290
291 #define INET_CSUM_COUNTER_INCR(ev) /* nothing */
292
293 #endif /* INET_CSUM_COUNTERS */
294
295 /*
296 * We need to save the IP options in case a protocol wants to respond
297 * to an incoming packet over the same route if the packet got here
298 * using IP source routing. This allows connection establishment and
299 * maintenance when the remote end is on a network that is not known
300 * to us.
301 */
302 int ip_nhops = 0;
303 static struct ip_srcrt {
304 struct in_addr dst; /* final destination */
305 char nop; /* one NOP to align */
306 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */
307 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)];
308 } ip_srcrt;
309
310 static void save_rte __P((u_char *, struct in_addr));
311
312 /*
313 * IP initialization: fill in IP protocol switch table.
314 * All protocols not implemented in kernel go to raw IP protocol handler.
315 */
316 void
317 ip_init()
318 {
319 struct protosw *pr;
320 int i;
321
322 pool_init(&ipqent_pool, sizeof(struct ipqent), 0, 0, 0, "ipqepl",
323 NULL);
324
325 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
326 if (pr == 0)
327 panic("ip_init");
328 for (i = 0; i < IPPROTO_MAX; i++)
329 ip_protox[i] = pr - inetsw;
330 for (pr = inetdomain.dom_protosw;
331 pr < inetdomain.dom_protoswNPROTOSW; pr++)
332 if (pr->pr_domain->dom_family == PF_INET &&
333 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
334 ip_protox[pr->pr_protocol] = pr - inetsw;
335 LIST_INIT(&ipq);
336 ip_id = time.tv_sec & 0xffff;
337 ipintrq.ifq_maxlen = ipqmaxlen;
338 TAILQ_INIT(&in_ifaddr);
339 in_ifaddrhashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, M_IFADDR,
340 M_WAITOK, &in_ifaddrhash);
341 if (ip_mtudisc != 0)
342 ip_mtudisc_timeout_q =
343 rt_timer_queue_create(ip_mtudisc_timeout);
344 #ifdef GATEWAY
345 ipflow_init();
346 #endif
347
348 #ifdef PFIL_HOOKS
349 /* Register our Packet Filter hook. */
350 inet_pfil_hook.ph_type = PFIL_TYPE_AF;
351 inet_pfil_hook.ph_af = AF_INET;
352 i = pfil_head_register(&inet_pfil_hook);
353 if (i != 0)
354 printf("ip_init: WARNING: unable to register pfil hook, "
355 "error %d\n", i);
356 #endif /* PFIL_HOOKS */
357
358 #ifdef INET_CSUM_COUNTERS
359 evcnt_attach_static(&ip_hwcsum_bad);
360 evcnt_attach_static(&ip_hwcsum_ok);
361 evcnt_attach_static(&ip_swcsum);
362 #endif /* INET_CSUM_COUNTERS */
363 }
364
365 struct sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET };
366 struct route ipforward_rt;
367
368 /*
369 * IP software interrupt routine
370 */
371 void
372 ipintr()
373 {
374 int s;
375 struct mbuf *m;
376
377 while (1) {
378 s = splnet();
379 IF_DEQUEUE(&ipintrq, m);
380 splx(s);
381 if (m == 0)
382 return;
383 ip_input(m);
384 }
385 }
386
387 /*
388 * Ip input routine. Checksum and byte swap header. If fragmented
389 * try to reassemble. Process options. Pass to next level.
390 */
391 void
392 ip_input(struct mbuf *m)
393 {
394 struct ip *ip = NULL;
395 struct ipq *fp;
396 struct in_ifaddr *ia;
397 struct ifaddr *ifa;
398 struct ipqent *ipqe;
399 int hlen = 0, mff, len;
400 int downmatch;
401
402 #ifdef DIAGNOSTIC
403 if ((m->m_flags & M_PKTHDR) == 0)
404 panic("ipintr no HDR");
405 #endif
406 #ifdef IPSEC
407 /*
408 * should the inner packet be considered authentic?
409 * see comment in ah4_input().
410 */
411 if (m) {
412 m->m_flags &= ~M_AUTHIPHDR;
413 m->m_flags &= ~M_AUTHIPDGM;
414 }
415 #endif
416 /*
417 * If no IP addresses have been set yet but the interfaces
418 * are receiving, can't do anything with incoming packets yet.
419 */
420 if (TAILQ_FIRST(&in_ifaddr) == 0)
421 goto bad;
422 ipstat.ips_total++;
423 /*
424 * If the IP header is not aligned, slurp it up into a new
425 * mbuf with space for link headers, in the event we forward
426 * it. Otherwise, if it is aligned, make sure the entire
427 * base IP header is in the first mbuf of the chain.
428 */
429 if (IP_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
430 if ((m = m_copyup(m, sizeof(struct ip),
431 (max_linkhdr + 3) & ~3)) == NULL) {
432 /* XXXJRT new stat, please */
433 ipstat.ips_toosmall++;
434 return;
435 }
436 } else if (__predict_false(m->m_len < sizeof (struct ip))) {
437 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
438 ipstat.ips_toosmall++;
439 return;
440 }
441 }
442 ip = mtod(m, struct ip *);
443 if (ip->ip_v != IPVERSION) {
444 ipstat.ips_badvers++;
445 goto bad;
446 }
447 hlen = ip->ip_hl << 2;
448 if (hlen < sizeof(struct ip)) { /* minimum header length */
449 ipstat.ips_badhlen++;
450 goto bad;
451 }
452 if (hlen > m->m_len) {
453 if ((m = m_pullup(m, hlen)) == 0) {
454 ipstat.ips_badhlen++;
455 return;
456 }
457 ip = mtod(m, struct ip *);
458 }
459
460 /*
461 * RFC1122: packets with a multicast source address are
462 * not allowed.
463 */
464 if (IN_MULTICAST(ip->ip_src.s_addr)) {
465 ipstat.ips_badaddr++;
466 goto bad;
467 }
468
469 /* 127/8 must not appear on wire - RFC1122 */
470 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
471 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
472 if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
473 ipstat.ips_badaddr++;
474 goto bad;
475 }
476 }
477
478 switch (m->m_pkthdr.csum_flags &
479 ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_IPv4) |
480 M_CSUM_IPv4_BAD)) {
481 case M_CSUM_IPv4|M_CSUM_IPv4_BAD:
482 INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad);
483 goto badcsum;
484
485 case M_CSUM_IPv4:
486 /* Checksum was okay. */
487 INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok);
488 break;
489
490 default:
491 /* Must compute it ourselves. */
492 INET_CSUM_COUNTER_INCR(&ip_swcsum);
493 if (in_cksum(m, hlen) != 0)
494 goto bad;
495 break;
496 }
497
498 /* Retrieve the packet length. */
499 len = ntohs(ip->ip_len);
500
501 /*
502 * Check for additional length bogosity
503 */
504 if (len < hlen) {
505 ipstat.ips_badlen++;
506 goto bad;
507 }
508
509 /*
510 * Check that the amount of data in the buffers
511 * is as at least much as the IP header would have us expect.
512 * Trim mbufs if longer than we expect.
513 * Drop packet if shorter than we expect.
514 */
515 if (m->m_pkthdr.len < len) {
516 ipstat.ips_tooshort++;
517 goto bad;
518 }
519 if (m->m_pkthdr.len > len) {
520 if (m->m_len == m->m_pkthdr.len) {
521 m->m_len = len;
522 m->m_pkthdr.len = len;
523 } else
524 m_adj(m, len - m->m_pkthdr.len);
525 }
526
527 #ifdef IPSEC
528 /* ipflow (IP fast forwarding) is not compatible with IPsec. */
529 m->m_flags &= ~M_CANFASTFWD;
530 #else
531 /*
532 * Assume that we can create a fast-forward IP flow entry
533 * based on this packet.
534 */
535 m->m_flags |= M_CANFASTFWD;
536 #endif
537
538 #ifdef PFIL_HOOKS
539 /*
540 * Run through list of hooks for input packets. If there are any
541 * filters which require that additional packets in the flow are
542 * not fast-forwarded, they must clear the M_CANFASTFWD flag.
543 * Note that filters must _never_ set this flag, as another filter
544 * in the list may have previously cleared it.
545 */
546 /*
547 * let ipfilter look at packet on the wire,
548 * not the decapsulated packet.
549 */
550 #ifdef IPSEC
551 if (!ipsec_getnhist(m))
552 #else
553 if (1)
554 #endif
555 {
556 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif,
557 PFIL_IN) != 0)
558 return;
559 if (m == NULL)
560 return;
561 ip = mtod(m, struct ip *);
562 hlen = ip->ip_hl << 2;
563 }
564 #endif /* PFIL_HOOKS */
565
566 #ifdef ALTQ
567 /* XXX Temporary until ALTQ is changed to use a pfil hook */
568 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) {
569 /* packet dropped by traffic conditioner */
570 return;
571 }
572 #endif
573
574 /*
575 * Convert fields to host representation.
576 */
577 NTOHS(ip->ip_len);
578 NTOHS(ip->ip_off);
579
580 /*
581 * Process options and, if not destined for us,
582 * ship it on. ip_dooptions returns 1 when an
583 * error was detected (causing an icmp message
584 * to be sent and the original packet to be freed).
585 */
586 ip_nhops = 0; /* for source routed packets */
587 if (hlen > sizeof (struct ip) && ip_dooptions(m))
588 return;
589
590 /*
591 * Check our list of addresses, to see if the packet is for us.
592 *
593 * Traditional 4.4BSD did not consult IFF_UP at all.
594 * The behavior here is to treat addresses on !IFF_UP interface
595 * as not mine.
596 */
597 downmatch = 0;
598 LIST_FOREACH(ia, &IN_IFADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
599 if (in_hosteq(ia->ia_addr.sin_addr, ip->ip_dst)) {
600 if ((ia->ia_ifp->if_flags & IFF_UP) != 0)
601 break;
602 else
603 downmatch++;
604 }
605 }
606 if (ia != NULL)
607 goto ours;
608 if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
609 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrlist, ifa_list) {
610 if (ifa->ifa_addr->sa_family != AF_INET)
611 continue;
612 ia = ifatoia(ifa);
613 if (in_hosteq(ip->ip_dst, ia->ia_broadaddr.sin_addr) ||
614 in_hosteq(ip->ip_dst, ia->ia_netbroadcast) ||
615 /*
616 * Look for all-0's host part (old broadcast addr),
617 * either for subnet or net.
618 */
619 ip->ip_dst.s_addr == ia->ia_subnet ||
620 ip->ip_dst.s_addr == ia->ia_net)
621 goto ours;
622 /*
623 * An interface with IP address zero accepts
624 * all packets that arrive on that interface.
625 */
626 if (in_nullhost(ia->ia_addr.sin_addr))
627 goto ours;
628 }
629 }
630 if (IN_MULTICAST(ip->ip_dst.s_addr)) {
631 struct in_multi *inm;
632 #ifdef MROUTING
633 extern struct socket *ip_mrouter;
634
635 if (M_READONLY(m)) {
636 if ((m = m_pullup(m, hlen)) == 0) {
637 ipstat.ips_toosmall++;
638 return;
639 }
640 ip = mtod(m, struct ip *);
641 }
642
643 if (ip_mrouter) {
644 /*
645 * If we are acting as a multicast router, all
646 * incoming multicast packets are passed to the
647 * kernel-level multicast forwarding function.
648 * The packet is returned (relatively) intact; if
649 * ip_mforward() returns a non-zero value, the packet
650 * must be discarded, else it may be accepted below.
651 *
652 * (The IP ident field is put in the same byte order
653 * as expected when ip_mforward() is called from
654 * ip_output().)
655 */
656 if (ip_mforward(m, m->m_pkthdr.rcvif) != 0) {
657 ipstat.ips_cantforward++;
658 m_freem(m);
659 return;
660 }
661
662 /*
663 * The process-level routing demon needs to receive
664 * all multicast IGMP packets, whether or not this
665 * host belongs to their destination groups.
666 */
667 if (ip->ip_p == IPPROTO_IGMP)
668 goto ours;
669 ipstat.ips_forward++;
670 }
671 #endif
672 /*
673 * See if we belong to the destination multicast group on the
674 * arrival interface.
675 */
676 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
677 if (inm == NULL) {
678 ipstat.ips_cantforward++;
679 m_freem(m);
680 return;
681 }
682 goto ours;
683 }
684 if (ip->ip_dst.s_addr == INADDR_BROADCAST ||
685 in_nullhost(ip->ip_dst))
686 goto ours;
687
688 /*
689 * Not for us; forward if possible and desirable.
690 */
691 if (ipforwarding == 0) {
692 ipstat.ips_cantforward++;
693 m_freem(m);
694 } else {
695 /*
696 * If ip_dst matched any of my address on !IFF_UP interface,
697 * and there's no IFF_UP interface that matches ip_dst,
698 * send icmp unreach. Forwarding it will result in in-kernel
699 * forwarding loop till TTL goes to 0.
700 */
701 if (downmatch) {
702 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
703 ipstat.ips_cantforward++;
704 return;
705 }
706 #ifdef IPSEC
707 if (ipsec4_in_reject(m, NULL)) {
708 ipsecstat.in_polvio++;
709 goto bad;
710 }
711 #endif
712
713 ip_forward(m, 0);
714 }
715 return;
716
717 ours:
718 /*
719 * If offset or IP_MF are set, must reassemble.
720 * Otherwise, nothing need be done.
721 * (We could look in the reassembly queue to see
722 * if the packet was previously fragmented,
723 * but it's not worth the time; just let them time out.)
724 */
725 if (ip->ip_off & ~(IP_DF|IP_RF)) {
726 /*
727 * Look for queue of fragments
728 * of this datagram.
729 */
730 IPQ_LOCK();
731 LIST_FOREACH(fp, &ipq, ipq_q)
732 if (ip->ip_id == fp->ipq_id &&
733 in_hosteq(ip->ip_src, fp->ipq_src) &&
734 in_hosteq(ip->ip_dst, fp->ipq_dst) &&
735 ip->ip_p == fp->ipq_p)
736 goto found;
737 fp = 0;
738 found:
739
740 /*
741 * Adjust ip_len to not reflect header,
742 * set ipqe_mff if more fragments are expected,
743 * convert offset of this to bytes.
744 */
745 ip->ip_len -= hlen;
746 mff = (ip->ip_off & IP_MF) != 0;
747 if (mff) {
748 /*
749 * Make sure that fragments have a data length
750 * that's a non-zero multiple of 8 bytes.
751 */
752 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
753 ipstat.ips_badfrags++;
754 IPQ_UNLOCK();
755 goto bad;
756 }
757 }
758 ip->ip_off <<= 3;
759
760 /*
761 * If datagram marked as having more fragments
762 * or if this is not the first fragment,
763 * attempt reassembly; if it succeeds, proceed.
764 */
765 if (mff || ip->ip_off) {
766 ipstat.ips_fragments++;
767 ipqe = pool_get(&ipqent_pool, PR_NOWAIT);
768 if (ipqe == NULL) {
769 ipstat.ips_rcvmemdrop++;
770 IPQ_UNLOCK();
771 goto bad;
772 }
773 ipqe->ipqe_mff = mff;
774 ipqe->ipqe_m = m;
775 ipqe->ipqe_ip = ip;
776 m = ip_reass(ipqe, fp);
777 if (m == 0) {
778 IPQ_UNLOCK();
779 return;
780 }
781 ipstat.ips_reassembled++;
782 ip = mtod(m, struct ip *);
783 hlen = ip->ip_hl << 2;
784 ip->ip_len += hlen;
785 } else
786 if (fp)
787 ip_freef(fp);
788 IPQ_UNLOCK();
789 }
790
791 #ifdef IPSEC
792 /*
793 * enforce IPsec policy checking if we are seeing last header.
794 * note that we do not visit this with protocols with pcb layer
795 * code - like udp/tcp/raw ip.
796 */
797 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0 &&
798 ipsec4_in_reject(m, NULL)) {
799 ipsecstat.in_polvio++;
800 goto bad;
801 }
802 #endif
803
804 /*
805 * Switch out to protocol's input routine.
806 */
807 #if IFA_STATS
808 if (ia && ip)
809 ia->ia_ifa.ifa_data.ifad_inbytes += ip->ip_len;
810 #endif
811 ipstat.ips_delivered++;
812 {
813 int off = hlen, nh = ip->ip_p;
814
815 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh);
816 return;
817 }
818 bad:
819 m_freem(m);
820 return;
821
822 badcsum:
823 ipstat.ips_badsum++;
824 m_freem(m);
825 }
826
827 /*
828 * Take incoming datagram fragment and try to
829 * reassemble it into whole datagram. If a chain for
830 * reassembly of this datagram already exists, then it
831 * is given as fp; otherwise have to make a chain.
832 */
833 struct mbuf *
834 ip_reass(ipqe, fp)
835 struct ipqent *ipqe;
836 struct ipq *fp;
837 {
838 struct mbuf *m = ipqe->ipqe_m;
839 struct ipqent *nq, *p, *q;
840 struct ip *ip;
841 struct mbuf *t;
842 int hlen = ipqe->ipqe_ip->ip_hl << 2;
843 int i, next;
844
845 IPQ_LOCK_CHECK();
846
847 /*
848 * Presence of header sizes in mbufs
849 * would confuse code below.
850 */
851 m->m_data += hlen;
852 m->m_len -= hlen;
853
854 /*
855 * If first fragment to arrive, create a reassembly queue.
856 */
857 if (fp == 0) {
858 /*
859 * Enforce upper bound on number of fragmented packets
860 * for which we attempt reassembly;
861 * If maxfrag is 0, never accept fragments.
862 * If maxfrag is -1, accept all fragments without limitation.
863 */
864 if (ip_maxfragpackets < 0)
865 ;
866 else if (ip_nfragpackets >= ip_maxfragpackets)
867 goto dropfrag;
868 ip_nfragpackets++;
869 MALLOC(fp, struct ipq *, sizeof (struct ipq),
870 M_FTABLE, M_NOWAIT);
871 if (fp == NULL)
872 goto dropfrag;
873 LIST_INSERT_HEAD(&ipq, fp, ipq_q);
874 fp->ipq_ttl = IPFRAGTTL;
875 fp->ipq_p = ipqe->ipqe_ip->ip_p;
876 fp->ipq_id = ipqe->ipqe_ip->ip_id;
877 TAILQ_INIT(&fp->ipq_fragq);
878 fp->ipq_src = ipqe->ipqe_ip->ip_src;
879 fp->ipq_dst = ipqe->ipqe_ip->ip_dst;
880 p = NULL;
881 goto insert;
882 }
883
884 /*
885 * Find a segment which begins after this one does.
886 */
887 for (p = NULL, q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL;
888 p = q, q = TAILQ_NEXT(q, ipqe_q))
889 if (q->ipqe_ip->ip_off > ipqe->ipqe_ip->ip_off)
890 break;
891
892 /*
893 * If there is a preceding segment, it may provide some of
894 * our data already. If so, drop the data from the incoming
895 * segment. If it provides all of our data, drop us.
896 */
897 if (p != NULL) {
898 i = p->ipqe_ip->ip_off + p->ipqe_ip->ip_len -
899 ipqe->ipqe_ip->ip_off;
900 if (i > 0) {
901 if (i >= ipqe->ipqe_ip->ip_len)
902 goto dropfrag;
903 m_adj(ipqe->ipqe_m, i);
904 ipqe->ipqe_ip->ip_off += i;
905 ipqe->ipqe_ip->ip_len -= i;
906 }
907 }
908
909 /*
910 * While we overlap succeeding segments trim them or,
911 * if they are completely covered, dequeue them.
912 */
913 for (; q != NULL && ipqe->ipqe_ip->ip_off + ipqe->ipqe_ip->ip_len >
914 q->ipqe_ip->ip_off; q = nq) {
915 i = (ipqe->ipqe_ip->ip_off + ipqe->ipqe_ip->ip_len) -
916 q->ipqe_ip->ip_off;
917 if (i < q->ipqe_ip->ip_len) {
918 q->ipqe_ip->ip_len -= i;
919 q->ipqe_ip->ip_off += i;
920 m_adj(q->ipqe_m, i);
921 break;
922 }
923 nq = TAILQ_NEXT(q, ipqe_q);
924 m_freem(q->ipqe_m);
925 TAILQ_REMOVE(&fp->ipq_fragq, q, ipqe_q);
926 pool_put(&ipqent_pool, q);
927 }
928
929 insert:
930 /*
931 * Stick new segment in its place;
932 * check for complete reassembly.
933 */
934 if (p == NULL) {
935 TAILQ_INSERT_HEAD(&fp->ipq_fragq, ipqe, ipqe_q);
936 } else {
937 TAILQ_INSERT_AFTER(&fp->ipq_fragq, p, ipqe, ipqe_q);
938 }
939 next = 0;
940 for (p = NULL, q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL;
941 p = q, q = TAILQ_NEXT(q, ipqe_q)) {
942 if (q->ipqe_ip->ip_off != next)
943 return (0);
944 next += q->ipqe_ip->ip_len;
945 }
946 if (p->ipqe_mff)
947 return (0);
948
949 /*
950 * Reassembly is complete. Check for a bogus message size and
951 * concatenate fragments.
952 */
953 q = TAILQ_FIRST(&fp->ipq_fragq);
954 ip = q->ipqe_ip;
955 if ((next + (ip->ip_hl << 2)) > IP_MAXPACKET) {
956 ipstat.ips_toolong++;
957 ip_freef(fp);
958 return (0);
959 }
960 m = q->ipqe_m;
961 t = m->m_next;
962 m->m_next = 0;
963 m_cat(m, t);
964 nq = TAILQ_NEXT(q, ipqe_q);
965 pool_put(&ipqent_pool, q);
966 for (q = nq; q != NULL; q = nq) {
967 t = q->ipqe_m;
968 nq = TAILQ_NEXT(q, ipqe_q);
969 pool_put(&ipqent_pool, q);
970 m_cat(m, t);
971 }
972
973 /*
974 * Create header for new ip packet by
975 * modifying header of first packet;
976 * dequeue and discard fragment reassembly header.
977 * Make header visible.
978 */
979 ip->ip_len = next;
980 ip->ip_src = fp->ipq_src;
981 ip->ip_dst = fp->ipq_dst;
982 LIST_REMOVE(fp, ipq_q);
983 FREE(fp, M_FTABLE);
984 ip_nfragpackets--;
985 m->m_len += (ip->ip_hl << 2);
986 m->m_data -= (ip->ip_hl << 2);
987 /* some debugging cruft by sklower, below, will go away soon */
988 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */
989 int plen = 0;
990 for (t = m; t; t = t->m_next)
991 plen += t->m_len;
992 m->m_pkthdr.len = plen;
993 }
994 return (m);
995
996 dropfrag:
997 ipstat.ips_fragdropped++;
998 m_freem(m);
999 pool_put(&ipqent_pool, ipqe);
1000 return (0);
1001 }
1002
1003 /*
1004 * Free a fragment reassembly header and all
1005 * associated datagrams.
1006 */
1007 void
1008 ip_freef(fp)
1009 struct ipq *fp;
1010 {
1011 struct ipqent *q, *p;
1012
1013 IPQ_LOCK_CHECK();
1014
1015 for (q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL; q = p) {
1016 p = TAILQ_NEXT(q, ipqe_q);
1017 m_freem(q->ipqe_m);
1018 TAILQ_REMOVE(&fp->ipq_fragq, q, ipqe_q);
1019 pool_put(&ipqent_pool, q);
1020 }
1021 LIST_REMOVE(fp, ipq_q);
1022 FREE(fp, M_FTABLE);
1023 ip_nfragpackets--;
1024 }
1025
1026 /*
1027 * IP timer processing;
1028 * if a timer expires on a reassembly
1029 * queue, discard it.
1030 */
1031 void
1032 ip_slowtimo()
1033 {
1034 struct ipq *fp, *nfp;
1035 int s = splsoftnet();
1036
1037 IPQ_LOCK();
1038 for (fp = LIST_FIRST(&ipq); fp != NULL; fp = nfp) {
1039 nfp = LIST_NEXT(fp, ipq_q);
1040 if (--fp->ipq_ttl == 0) {
1041 ipstat.ips_fragtimeout++;
1042 ip_freef(fp);
1043 }
1044 }
1045 /*
1046 * If we are over the maximum number of fragments
1047 * (due to the limit being lowered), drain off
1048 * enough to get down to the new limit.
1049 */
1050 if (ip_maxfragpackets < 0)
1051 ;
1052 else {
1053 while (ip_nfragpackets > ip_maxfragpackets && LIST_FIRST(&ipq))
1054 ip_freef(LIST_FIRST(&ipq));
1055 }
1056 IPQ_UNLOCK();
1057 #ifdef GATEWAY
1058 ipflow_slowtimo();
1059 #endif
1060 splx(s);
1061 }
1062
1063 /*
1064 * Drain off all datagram fragments.
1065 */
1066 void
1067 ip_drain()
1068 {
1069
1070 /*
1071 * We may be called from a device's interrupt context. If
1072 * the ipq is already busy, just bail out now.
1073 */
1074 if (ipq_lock_try() == 0)
1075 return;
1076
1077 while (LIST_FIRST(&ipq) != NULL) {
1078 ipstat.ips_fragdropped++;
1079 ip_freef(LIST_FIRST(&ipq));
1080 }
1081
1082 IPQ_UNLOCK();
1083 }
1084
1085 /*
1086 * Do option processing on a datagram,
1087 * possibly discarding it if bad options are encountered,
1088 * or forwarding it if source-routed.
1089 * Returns 1 if packet has been forwarded/freed,
1090 * 0 if the packet should be processed further.
1091 */
1092 int
1093 ip_dooptions(m)
1094 struct mbuf *m;
1095 {
1096 struct ip *ip = mtod(m, struct ip *);
1097 u_char *cp, *cp0;
1098 struct ip_timestamp *ipt;
1099 struct in_ifaddr *ia;
1100 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0;
1101 struct in_addr dst;
1102 n_time ntime;
1103
1104 dst = ip->ip_dst;
1105 cp = (u_char *)(ip + 1);
1106 cnt = (ip->ip_hl << 2) - sizeof (struct ip);
1107 for (; cnt > 0; cnt -= optlen, cp += optlen) {
1108 opt = cp[IPOPT_OPTVAL];
1109 if (opt == IPOPT_EOL)
1110 break;
1111 if (opt == IPOPT_NOP)
1112 optlen = 1;
1113 else {
1114 if (cnt < IPOPT_OLEN + sizeof(*cp)) {
1115 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1116 goto bad;
1117 }
1118 optlen = cp[IPOPT_OLEN];
1119 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
1120 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1121 goto bad;
1122 }
1123 }
1124 switch (opt) {
1125
1126 default:
1127 break;
1128
1129 /*
1130 * Source routing with record.
1131 * Find interface with current destination address.
1132 * If none on this machine then drop if strictly routed,
1133 * or do nothing if loosely routed.
1134 * Record interface address and bring up next address
1135 * component. If strictly routed make sure next
1136 * address is on directly accessible net.
1137 */
1138 case IPOPT_LSRR:
1139 case IPOPT_SSRR:
1140 if (ip_allowsrcrt == 0) {
1141 type = ICMP_UNREACH;
1142 code = ICMP_UNREACH_NET_PROHIB;
1143 goto bad;
1144 }
1145 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1146 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1147 goto bad;
1148 }
1149 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1150 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1151 goto bad;
1152 }
1153 ipaddr.sin_addr = ip->ip_dst;
1154 ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr)));
1155 if (ia == 0) {
1156 if (opt == IPOPT_SSRR) {
1157 type = ICMP_UNREACH;
1158 code = ICMP_UNREACH_SRCFAIL;
1159 goto bad;
1160 }
1161 /*
1162 * Loose routing, and not at next destination
1163 * yet; nothing to do except forward.
1164 */
1165 break;
1166 }
1167 off--; /* 0 origin */
1168 if ((off + sizeof(struct in_addr)) > optlen) {
1169 /*
1170 * End of source route. Should be for us.
1171 */
1172 save_rte(cp, ip->ip_src);
1173 break;
1174 }
1175 /*
1176 * locate outgoing interface
1177 */
1178 bcopy((caddr_t)(cp + off), (caddr_t)&ipaddr.sin_addr,
1179 sizeof(ipaddr.sin_addr));
1180 if (opt == IPOPT_SSRR)
1181 ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr)));
1182 else
1183 ia = ip_rtaddr(ipaddr.sin_addr);
1184 if (ia == 0) {
1185 type = ICMP_UNREACH;
1186 code = ICMP_UNREACH_SRCFAIL;
1187 goto bad;
1188 }
1189 ip->ip_dst = ipaddr.sin_addr;
1190 bcopy((caddr_t)&ia->ia_addr.sin_addr,
1191 (caddr_t)(cp + off), sizeof(struct in_addr));
1192 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1193 /*
1194 * Let ip_intr's mcast routing check handle mcast pkts
1195 */
1196 forward = !IN_MULTICAST(ip->ip_dst.s_addr);
1197 break;
1198
1199 case IPOPT_RR:
1200 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1201 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1202 goto bad;
1203 }
1204 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1205 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1206 goto bad;
1207 }
1208 /*
1209 * If no space remains, ignore.
1210 */
1211 off--; /* 0 origin */
1212 if ((off + sizeof(struct in_addr)) > optlen)
1213 break;
1214 bcopy((caddr_t)(&ip->ip_dst), (caddr_t)&ipaddr.sin_addr,
1215 sizeof(ipaddr.sin_addr));
1216 /*
1217 * locate outgoing interface; if we're the destination,
1218 * use the incoming interface (should be same).
1219 */
1220 if ((ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))))
1221 == NULL &&
1222 (ia = ip_rtaddr(ipaddr.sin_addr)) == NULL) {
1223 type = ICMP_UNREACH;
1224 code = ICMP_UNREACH_HOST;
1225 goto bad;
1226 }
1227 bcopy((caddr_t)&ia->ia_addr.sin_addr,
1228 (caddr_t)(cp + off), sizeof(struct in_addr));
1229 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1230 break;
1231
1232 case IPOPT_TS:
1233 code = cp - (u_char *)ip;
1234 ipt = (struct ip_timestamp *)cp;
1235 if (ipt->ipt_len < 4 || ipt->ipt_len > 40) {
1236 code = (u_char *)&ipt->ipt_len - (u_char *)ip;
1237 goto bad;
1238 }
1239 if (ipt->ipt_ptr < 5) {
1240 code = (u_char *)&ipt->ipt_ptr - (u_char *)ip;
1241 goto bad;
1242 }
1243 if (ipt->ipt_ptr > ipt->ipt_len - sizeof (int32_t)) {
1244 if (++ipt->ipt_oflw == 0) {
1245 code = (u_char *)&ipt->ipt_ptr -
1246 (u_char *)ip;
1247 goto bad;
1248 }
1249 break;
1250 }
1251 cp0 = (cp + ipt->ipt_ptr - 1);
1252 switch (ipt->ipt_flg) {
1253
1254 case IPOPT_TS_TSONLY:
1255 break;
1256
1257 case IPOPT_TS_TSANDADDR:
1258 if (ipt->ipt_ptr - 1 + sizeof(n_time) +
1259 sizeof(struct in_addr) > ipt->ipt_len) {
1260 code = (u_char *)&ipt->ipt_ptr -
1261 (u_char *)ip;
1262 goto bad;
1263 }
1264 ipaddr.sin_addr = dst;
1265 ia = ifatoia(ifaof_ifpforaddr(sintosa(&ipaddr),
1266 m->m_pkthdr.rcvif));
1267 if (ia == 0)
1268 continue;
1269 bcopy(&ia->ia_addr.sin_addr,
1270 cp0, sizeof(struct in_addr));
1271 ipt->ipt_ptr += sizeof(struct in_addr);
1272 break;
1273
1274 case IPOPT_TS_PRESPEC:
1275 if (ipt->ipt_ptr - 1 + sizeof(n_time) +
1276 sizeof(struct in_addr) > ipt->ipt_len) {
1277 code = (u_char *)&ipt->ipt_ptr -
1278 (u_char *)ip;
1279 goto bad;
1280 }
1281 bcopy(cp0, &ipaddr.sin_addr,
1282 sizeof(struct in_addr));
1283 if (ifatoia(ifa_ifwithaddr(sintosa(&ipaddr)))
1284 == NULL)
1285 continue;
1286 ipt->ipt_ptr += sizeof(struct in_addr);
1287 break;
1288
1289 default:
1290 /* XXX can't take &ipt->ipt_flg */
1291 code = (u_char *)&ipt->ipt_ptr -
1292 (u_char *)ip + 1;
1293 goto bad;
1294 }
1295 ntime = iptime();
1296 cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */
1297 bcopy(cp0, (caddr_t)cp + ipt->ipt_ptr - 1,
1298 sizeof(n_time));
1299 ipt->ipt_ptr += sizeof(n_time);
1300 }
1301 }
1302 if (forward) {
1303 if (ip_forwsrcrt == 0) {
1304 type = ICMP_UNREACH;
1305 code = ICMP_UNREACH_SRCFAIL;
1306 goto bad;
1307 }
1308 ip_forward(m, 1);
1309 return (1);
1310 }
1311 return (0);
1312 bad:
1313 icmp_error(m, type, code, 0, 0);
1314 ipstat.ips_badoptions++;
1315 return (1);
1316 }
1317
1318 /*
1319 * Given address of next destination (final or next hop),
1320 * return internet address info of interface to be used to get there.
1321 */
1322 struct in_ifaddr *
1323 ip_rtaddr(dst)
1324 struct in_addr dst;
1325 {
1326 struct sockaddr_in *sin;
1327
1328 sin = satosin(&ipforward_rt.ro_dst);
1329
1330 if (ipforward_rt.ro_rt == 0 || !in_hosteq(dst, sin->sin_addr)) {
1331 if (ipforward_rt.ro_rt) {
1332 RTFREE(ipforward_rt.ro_rt);
1333 ipforward_rt.ro_rt = 0;
1334 }
1335 sin->sin_family = AF_INET;
1336 sin->sin_len = sizeof(*sin);
1337 sin->sin_addr = dst;
1338
1339 rtalloc(&ipforward_rt);
1340 }
1341 if (ipforward_rt.ro_rt == 0)
1342 return ((struct in_ifaddr *)0);
1343 return (ifatoia(ipforward_rt.ro_rt->rt_ifa));
1344 }
1345
1346 /*
1347 * Save incoming source route for use in replies,
1348 * to be picked up later by ip_srcroute if the receiver is interested.
1349 */
1350 void
1351 save_rte(option, dst)
1352 u_char *option;
1353 struct in_addr dst;
1354 {
1355 unsigned olen;
1356
1357 olen = option[IPOPT_OLEN];
1358 #ifdef DIAGNOSTIC
1359 if (ipprintfs)
1360 printf("save_rte: olen %d\n", olen);
1361 #endif /* 0 */
1362 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst)))
1363 return;
1364 bcopy((caddr_t)option, (caddr_t)ip_srcrt.srcopt, olen);
1365 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1366 ip_srcrt.dst = dst;
1367 }
1368
1369 /*
1370 * Retrieve incoming source route for use in replies,
1371 * in the same form used by setsockopt.
1372 * The first hop is placed before the options, will be removed later.
1373 */
1374 struct mbuf *
1375 ip_srcroute()
1376 {
1377 struct in_addr *p, *q;
1378 struct mbuf *m;
1379
1380 if (ip_nhops == 0)
1381 return ((struct mbuf *)0);
1382 m = m_get(M_DONTWAIT, MT_SOOPTS);
1383 if (m == 0)
1384 return ((struct mbuf *)0);
1385
1386 #define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt))
1387
1388 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
1389 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) +
1390 OPTSIZ;
1391 #ifdef DIAGNOSTIC
1392 if (ipprintfs)
1393 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len);
1394 #endif
1395
1396 /*
1397 * First save first hop for return route
1398 */
1399 p = &ip_srcrt.route[ip_nhops - 1];
1400 *(mtod(m, struct in_addr *)) = *p--;
1401 #ifdef DIAGNOSTIC
1402 if (ipprintfs)
1403 printf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr));
1404 #endif
1405
1406 /*
1407 * Copy option fields and padding (nop) to mbuf.
1408 */
1409 ip_srcrt.nop = IPOPT_NOP;
1410 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
1411 bcopy((caddr_t)&ip_srcrt.nop,
1412 mtod(m, caddr_t) + sizeof(struct in_addr), OPTSIZ);
1413 q = (struct in_addr *)(mtod(m, caddr_t) +
1414 sizeof(struct in_addr) + OPTSIZ);
1415 #undef OPTSIZ
1416 /*
1417 * Record return path as an IP source route,
1418 * reversing the path (pointers are now aligned).
1419 */
1420 while (p >= ip_srcrt.route) {
1421 #ifdef DIAGNOSTIC
1422 if (ipprintfs)
1423 printf(" %x", ntohl(q->s_addr));
1424 #endif
1425 *q++ = *p--;
1426 }
1427 /*
1428 * Last hop goes to final destination.
1429 */
1430 *q = ip_srcrt.dst;
1431 #ifdef DIAGNOSTIC
1432 if (ipprintfs)
1433 printf(" %x\n", ntohl(q->s_addr));
1434 #endif
1435 return (m);
1436 }
1437
1438 /*
1439 * Strip out IP options, at higher
1440 * level protocol in the kernel.
1441 * Second argument is buffer to which options
1442 * will be moved, and return value is their length.
1443 * XXX should be deleted; last arg currently ignored.
1444 */
1445 void
1446 ip_stripoptions(m, mopt)
1447 struct mbuf *m;
1448 struct mbuf *mopt;
1449 {
1450 int i;
1451 struct ip *ip = mtod(m, struct ip *);
1452 caddr_t opts;
1453 int olen;
1454
1455 olen = (ip->ip_hl << 2) - sizeof (struct ip);
1456 opts = (caddr_t)(ip + 1);
1457 i = m->m_len - (sizeof (struct ip) + olen);
1458 bcopy(opts + olen, opts, (unsigned)i);
1459 m->m_len -= olen;
1460 if (m->m_flags & M_PKTHDR)
1461 m->m_pkthdr.len -= olen;
1462 ip->ip_len -= olen;
1463 ip->ip_hl = sizeof (struct ip) >> 2;
1464 }
1465
1466 const int inetctlerrmap[PRC_NCMDS] = {
1467 0, 0, 0, 0,
1468 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
1469 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
1470 EMSGSIZE, EHOSTUNREACH, 0, 0,
1471 0, 0, 0, 0,
1472 ENOPROTOOPT
1473 };
1474
1475 /*
1476 * Forward a packet. If some error occurs return the sender
1477 * an icmp packet. Note we can't always generate a meaningful
1478 * icmp message because icmp doesn't have a large enough repertoire
1479 * of codes and types.
1480 *
1481 * If not forwarding, just drop the packet. This could be confusing
1482 * if ipforwarding was zero but some routing protocol was advancing
1483 * us as a gateway to somewhere. However, we must let the routing
1484 * protocol deal with that.
1485 *
1486 * The srcrt parameter indicates whether the packet is being forwarded
1487 * via a source route.
1488 */
1489 void
1490 ip_forward(m, srcrt)
1491 struct mbuf *m;
1492 int srcrt;
1493 {
1494 struct ip *ip = mtod(m, struct ip *);
1495 struct sockaddr_in *sin;
1496 struct rtentry *rt;
1497 int error, type = 0, code = 0;
1498 struct mbuf *mcopy;
1499 n_long dest;
1500 struct ifnet *destifp;
1501 #ifdef IPSEC
1502 struct ifnet dummyifp;
1503 #endif
1504
1505 /*
1506 * Clear any in-bound checksum flags for this packet.
1507 */
1508 m->m_pkthdr.csum_flags = 0;
1509
1510 dest = 0;
1511 #ifdef DIAGNOSTIC
1512 if (ipprintfs)
1513 printf("forward: src %2.2x dst %2.2x ttl %x\n",
1514 ntohl(ip->ip_src.s_addr),
1515 ntohl(ip->ip_dst.s_addr), ip->ip_ttl);
1516 #endif
1517 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
1518 ipstat.ips_cantforward++;
1519 m_freem(m);
1520 return;
1521 }
1522 if (ip->ip_ttl <= IPTTLDEC) {
1523 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0);
1524 return;
1525 }
1526 ip->ip_ttl -= IPTTLDEC;
1527
1528 sin = satosin(&ipforward_rt.ro_dst);
1529 if ((rt = ipforward_rt.ro_rt) == 0 ||
1530 !in_hosteq(ip->ip_dst, sin->sin_addr)) {
1531 if (ipforward_rt.ro_rt) {
1532 RTFREE(ipforward_rt.ro_rt);
1533 ipforward_rt.ro_rt = 0;
1534 }
1535 sin->sin_family = AF_INET;
1536 sin->sin_len = sizeof(struct sockaddr_in);
1537 sin->sin_addr = ip->ip_dst;
1538
1539 rtalloc(&ipforward_rt);
1540 if (ipforward_rt.ro_rt == 0) {
1541 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0);
1542 return;
1543 }
1544 rt = ipforward_rt.ro_rt;
1545 }
1546
1547 /*
1548 * Save at most 68 bytes of the packet in case
1549 * we need to generate an ICMP message to the src.
1550 * Pullup to avoid sharing mbuf cluster between m and mcopy.
1551 */
1552 mcopy = m_copym(m, 0, imin((int)ip->ip_len, 68), M_DONTWAIT);
1553 if (mcopy)
1554 mcopy = m_pullup(mcopy, ip->ip_hl << 2);
1555
1556 /*
1557 * If forwarding packet using same interface that it came in on,
1558 * perhaps should send a redirect to sender to shortcut a hop.
1559 * Only send redirect if source is sending directly to us,
1560 * and if packet was not source routed (or has any options).
1561 * Also, don't send redirect if forwarding using a default route
1562 * or a route modified by a redirect.
1563 */
1564 if (rt->rt_ifp == m->m_pkthdr.rcvif &&
1565 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1566 !in_nullhost(satosin(rt_key(rt))->sin_addr) &&
1567 ipsendredirects && !srcrt) {
1568 if (rt->rt_ifa &&
1569 (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) ==
1570 ifatoia(rt->rt_ifa)->ia_subnet) {
1571 if (rt->rt_flags & RTF_GATEWAY)
1572 dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
1573 else
1574 dest = ip->ip_dst.s_addr;
1575 /*
1576 * Router requirements says to only send host
1577 * redirects.
1578 */
1579 type = ICMP_REDIRECT;
1580 code = ICMP_REDIRECT_HOST;
1581 #ifdef DIAGNOSTIC
1582 if (ipprintfs)
1583 printf("redirect (%d) to %x\n", code,
1584 (u_int32_t)dest);
1585 #endif
1586 }
1587 }
1588
1589 #ifdef IPSEC
1590 /* Don't lookup socket in forwarding case */
1591 (void)ipsec_setsocket(m, NULL);
1592 #endif
1593 error = ip_output(m, (struct mbuf *)0, &ipforward_rt,
1594 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 0);
1595 if (error)
1596 ipstat.ips_cantforward++;
1597 else {
1598 ipstat.ips_forward++;
1599 if (type)
1600 ipstat.ips_redirectsent++;
1601 else {
1602 if (mcopy) {
1603 #ifdef GATEWAY
1604 if (mcopy->m_flags & M_CANFASTFWD)
1605 ipflow_create(&ipforward_rt, mcopy);
1606 #endif
1607 m_freem(mcopy);
1608 }
1609 return;
1610 }
1611 }
1612 if (mcopy == NULL)
1613 return;
1614 destifp = NULL;
1615
1616 switch (error) {
1617
1618 case 0: /* forwarded, but need redirect */
1619 /* type, code set above */
1620 break;
1621
1622 case ENETUNREACH: /* shouldn't happen, checked above */
1623 case EHOSTUNREACH:
1624 case ENETDOWN:
1625 case EHOSTDOWN:
1626 default:
1627 type = ICMP_UNREACH;
1628 code = ICMP_UNREACH_HOST;
1629 break;
1630
1631 case EMSGSIZE:
1632 type = ICMP_UNREACH;
1633 code = ICMP_UNREACH_NEEDFRAG;
1634 #ifndef IPSEC
1635 if (ipforward_rt.ro_rt)
1636 destifp = ipforward_rt.ro_rt->rt_ifp;
1637 #else
1638 /*
1639 * If the packet is routed over IPsec tunnel, tell the
1640 * originator the tunnel MTU.
1641 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
1642 * XXX quickhack!!!
1643 */
1644 if (ipforward_rt.ro_rt) {
1645 struct secpolicy *sp;
1646 int ipsecerror;
1647 size_t ipsechdr;
1648 struct route *ro;
1649
1650 sp = ipsec4_getpolicybyaddr(mcopy,
1651 IPSEC_DIR_OUTBOUND,
1652 IP_FORWARDING,
1653 &ipsecerror);
1654
1655 if (sp == NULL)
1656 destifp = ipforward_rt.ro_rt->rt_ifp;
1657 else {
1658 /* count IPsec header size */
1659 ipsechdr = ipsec4_hdrsiz(mcopy,
1660 IPSEC_DIR_OUTBOUND,
1661 NULL);
1662
1663 /*
1664 * find the correct route for outer IPv4
1665 * header, compute tunnel MTU.
1666 *
1667 * XXX BUG ALERT
1668 * The "dummyifp" code relies upon the fact
1669 * that icmp_error() touches only ifp->if_mtu.
1670 */
1671 /*XXX*/
1672 destifp = NULL;
1673 if (sp->req != NULL
1674 && sp->req->sav != NULL
1675 && sp->req->sav->sah != NULL) {
1676 ro = &sp->req->sav->sah->sa_route;
1677 if (ro->ro_rt && ro->ro_rt->rt_ifp) {
1678 dummyifp.if_mtu =
1679 ro->ro_rt->rt_rmx.rmx_mtu ?
1680 ro->ro_rt->rt_rmx.rmx_mtu :
1681 ro->ro_rt->rt_ifp->if_mtu;
1682 dummyifp.if_mtu -= ipsechdr;
1683 destifp = &dummyifp;
1684 }
1685 }
1686
1687 key_freesp(sp);
1688 }
1689 }
1690 #endif /*IPSEC*/
1691 ipstat.ips_cantfrag++;
1692 break;
1693
1694 case ENOBUFS:
1695 #if 1
1696 /*
1697 * a router should not generate ICMP_SOURCEQUENCH as
1698 * required in RFC1812 Requirements for IP Version 4 Routers.
1699 * source quench could be a big problem under DoS attacks,
1700 * or if the underlying interface is rate-limited.
1701 */
1702 if (mcopy)
1703 m_freem(mcopy);
1704 return;
1705 #else
1706 type = ICMP_SOURCEQUENCH;
1707 code = 0;
1708 break;
1709 #endif
1710 }
1711 icmp_error(mcopy, type, code, dest, destifp);
1712 }
1713
1714 void
1715 ip_savecontrol(inp, mp, ip, m)
1716 struct inpcb *inp;
1717 struct mbuf **mp;
1718 struct ip *ip;
1719 struct mbuf *m;
1720 {
1721
1722 if (inp->inp_socket->so_options & SO_TIMESTAMP) {
1723 struct timeval tv;
1724
1725 microtime(&tv);
1726 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1727 SCM_TIMESTAMP, SOL_SOCKET);
1728 if (*mp)
1729 mp = &(*mp)->m_next;
1730 }
1731 if (inp->inp_flags & INP_RECVDSTADDR) {
1732 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
1733 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1734 if (*mp)
1735 mp = &(*mp)->m_next;
1736 }
1737 #ifdef notyet
1738 /*
1739 * XXX
1740 * Moving these out of udp_input() made them even more broken
1741 * than they already were.
1742 * - fenner (at) parc.xerox.com
1743 */
1744 /* options were tossed already */
1745 if (inp->inp_flags & INP_RECVOPTS) {
1746 *mp = sbcreatecontrol((caddr_t) opts_deleted_above,
1747 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1748 if (*mp)
1749 mp = &(*mp)->m_next;
1750 }
1751 /* ip_srcroute doesn't do what we want here, need to fix */
1752 if (inp->inp_flags & INP_RECVRETOPTS) {
1753 *mp = sbcreatecontrol((caddr_t) ip_srcroute(),
1754 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1755 if (*mp)
1756 mp = &(*mp)->m_next;
1757 }
1758 #endif
1759 if (inp->inp_flags & INP_RECVIF) {
1760 struct sockaddr_dl sdl;
1761
1762 sdl.sdl_len = offsetof(struct sockaddr_dl, sdl_data[0]);
1763 sdl.sdl_family = AF_LINK;
1764 sdl.sdl_index = m->m_pkthdr.rcvif ?
1765 m->m_pkthdr.rcvif->if_index : 0;
1766 sdl.sdl_nlen = sdl.sdl_alen = sdl.sdl_slen = 0;
1767 *mp = sbcreatecontrol((caddr_t) &sdl, sdl.sdl_len,
1768 IP_RECVIF, IPPROTO_IP);
1769 if (*mp)
1770 mp = &(*mp)->m_next;
1771 }
1772 }
1773
1774 int
1775 ip_sysctl(name, namelen, oldp, oldlenp, newp, newlen)
1776 int *name;
1777 u_int namelen;
1778 void *oldp;
1779 size_t *oldlenp;
1780 void *newp;
1781 size_t newlen;
1782 {
1783 extern int subnetsarelocal, hostzeroisbroadcast;
1784
1785 int error, old;
1786
1787 /* All sysctl names at this level are terminal. */
1788 if (namelen != 1)
1789 return (ENOTDIR);
1790
1791 switch (name[0]) {
1792 case IPCTL_FORWARDING:
1793 return (sysctl_int(oldp, oldlenp, newp, newlen, &ipforwarding));
1794 case IPCTL_SENDREDIRECTS:
1795 return (sysctl_int(oldp, oldlenp, newp, newlen,
1796 &ipsendredirects));
1797 case IPCTL_DEFTTL:
1798 return (sysctl_int(oldp, oldlenp, newp, newlen, &ip_defttl));
1799 #ifdef notyet
1800 case IPCTL_DEFMTU:
1801 return (sysctl_int(oldp, oldlenp, newp, newlen, &ip_mtu));
1802 #endif
1803 case IPCTL_FORWSRCRT:
1804 /* Don't allow this to change in a secure environment. */
1805 if (securelevel > 0)
1806 return (sysctl_rdint(oldp, oldlenp, newp,
1807 ip_forwsrcrt));
1808 else
1809 return (sysctl_int(oldp, oldlenp, newp, newlen,
1810 &ip_forwsrcrt));
1811 case IPCTL_DIRECTEDBCAST:
1812 return (sysctl_int(oldp, oldlenp, newp, newlen,
1813 &ip_directedbcast));
1814 case IPCTL_ALLOWSRCRT:
1815 return (sysctl_int(oldp, oldlenp, newp, newlen,
1816 &ip_allowsrcrt));
1817 case IPCTL_SUBNETSARELOCAL:
1818 return (sysctl_int(oldp, oldlenp, newp, newlen,
1819 &subnetsarelocal));
1820 case IPCTL_MTUDISC:
1821 error = sysctl_int(oldp, oldlenp, newp, newlen,
1822 &ip_mtudisc);
1823 if (ip_mtudisc != 0 && ip_mtudisc_timeout_q == NULL) {
1824 ip_mtudisc_timeout_q =
1825 rt_timer_queue_create(ip_mtudisc_timeout);
1826 } else if (ip_mtudisc == 0 && ip_mtudisc_timeout_q != NULL) {
1827 rt_timer_queue_destroy(ip_mtudisc_timeout_q, TRUE);
1828 ip_mtudisc_timeout_q = NULL;
1829 }
1830 return error;
1831 case IPCTL_ANONPORTMIN:
1832 old = anonportmin;
1833 error = sysctl_int(oldp, oldlenp, newp, newlen, &anonportmin);
1834 if (anonportmin >= anonportmax || anonportmin < 0
1835 || anonportmin > 65535
1836 #ifndef IPNOPRIVPORTS
1837 || anonportmin < IPPORT_RESERVED
1838 #endif
1839 ) {
1840 anonportmin = old;
1841 return (EINVAL);
1842 }
1843 return (error);
1844 case IPCTL_ANONPORTMAX:
1845 old = anonportmax;
1846 error = sysctl_int(oldp, oldlenp, newp, newlen, &anonportmax);
1847 if (anonportmin >= anonportmax || anonportmax < 0
1848 || anonportmax > 65535
1849 #ifndef IPNOPRIVPORTS
1850 || anonportmax < IPPORT_RESERVED
1851 #endif
1852 ) {
1853 anonportmax = old;
1854 return (EINVAL);
1855 }
1856 return (error);
1857 case IPCTL_MTUDISCTIMEOUT:
1858 error = sysctl_int(oldp, oldlenp, newp, newlen,
1859 &ip_mtudisc_timeout);
1860 if (ip_mtudisc_timeout_q != NULL)
1861 rt_timer_queue_change(ip_mtudisc_timeout_q,
1862 ip_mtudisc_timeout);
1863 return (error);
1864 #ifdef GATEWAY
1865 case IPCTL_MAXFLOWS:
1866 {
1867 int s;
1868
1869 error = sysctl_int(oldp, oldlenp, newp, newlen,
1870 &ip_maxflows);
1871 s = splsoftnet();
1872 ipflow_reap(0);
1873 splx(s);
1874 return (error);
1875 }
1876 #endif
1877 case IPCTL_HOSTZEROBROADCAST:
1878 return (sysctl_int(oldp, oldlenp, newp, newlen,
1879 &hostzeroisbroadcast));
1880 #if NGIF > 0
1881 case IPCTL_GIF_TTL:
1882 return(sysctl_int(oldp, oldlenp, newp, newlen,
1883 &ip_gif_ttl));
1884 #endif
1885
1886 #if NGRE > 0
1887 case IPCTL_GRE_TTL:
1888 return(sysctl_int(oldp, oldlenp, newp, newlen,
1889 &ip_gre_ttl));
1890 #endif
1891
1892 #ifndef IPNOPRIVPORTS
1893 case IPCTL_LOWPORTMIN:
1894 old = lowportmin;
1895 error = sysctl_int(oldp, oldlenp, newp, newlen, &lowportmin);
1896 if (lowportmin >= lowportmax
1897 || lowportmin > IPPORT_RESERVEDMAX
1898 || lowportmin < IPPORT_RESERVEDMIN
1899 ) {
1900 lowportmin = old;
1901 return (EINVAL);
1902 }
1903 return (error);
1904 case IPCTL_LOWPORTMAX:
1905 old = lowportmax;
1906 error = sysctl_int(oldp, oldlenp, newp, newlen, &lowportmax);
1907 if (lowportmin >= lowportmax
1908 || lowportmax > IPPORT_RESERVEDMAX
1909 || lowportmax < IPPORT_RESERVEDMIN
1910 ) {
1911 lowportmax = old;
1912 return (EINVAL);
1913 }
1914 return (error);
1915 #endif
1916
1917 case IPCTL_MAXFRAGPACKETS:
1918 return (sysctl_int(oldp, oldlenp, newp, newlen,
1919 &ip_maxfragpackets));
1920
1921 default:
1922 return (EOPNOTSUPP);
1923 }
1924 /* NOTREACHED */
1925 }
1926