1 /*
2 * Copyright (c) 1982, 1986, 1988, 1993
3 * The Regents of the University of California. All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * 3. All advertising materials mentioning features or use of this software
14 * must display the following acknowledgement:
15 * This product includes software developed by the University of
16 * California, Berkeley and its contributors.
17 * 4. Neither the name of the University nor the names of its contributors
18 * may be used to endorse or promote products derived from this software
19 * without specific prior written permission.
20 *
21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31 * SUCH DAMAGE.
32 *
33 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94
34 * $FreeBSD$
35 */
36
37 #define _IP_VHL
38
39 #include "opt_bootp.h"
40 #include "opt_ipfw.h"
41 #include "opt_ipdn.h"
42 #include "opt_ipdivert.h"
43 #include "opt_ipfilter.h"
44 #include "opt_ipstealth.h"
45 #include "opt_ipsec.h"
46 #include "opt_random_ip_id.h"
47
48 #include <sys/param.h>
49 #include <sys/systm.h>
50 #include <sys/callout.h>
51 #include <sys/eventhandler.h>
52 #include <sys/mbuf.h>
53 #include <sys/malloc.h>
54 #include <sys/domain.h>
55 #include <sys/protosw.h>
56 #include <sys/socket.h>
57 #include <sys/time.h>
58 #include <sys/kernel.h>
59 #include <sys/syslog.h>
60 #include <sys/sysctl.h>
61
62 #include <net/if.h>
63 #include <net/if_types.h>
64 #include <net/if_var.h>
65 #include <net/if_dl.h>
66 #include <net/route.h>
67 #include <net/netisr.h>
68 #include <net/intrq.h>
69
70 #include <netinet/in.h>
71 #include <netinet/in_systm.h>
72 #include <netinet/in_var.h>
73 #include <netinet/ip.h>
74 #include <netinet/in_pcb.h>
75 #include <netinet/ip_var.h>
76 #include <netinet/ip_icmp.h>
77 #include <machine/in_cksum.h>
78
79 #include <netinet/ipprotosw.h>
80
81 #include <sys/socketvar.h>
82
83 #include <netinet/ip_fw.h>
84 #include <netinet/ip_dummynet.h>
85
86 #ifdef IPSEC
87 #include <netinet6/ipsec.h>
88 #include <netkey/key.h>
89 #endif
90
91 #ifdef FAST_IPSEC
92 #include <netipsec/ipsec.h>
93 #include <netipsec/key.h>
94 #endif
95
96 int rsvp_on = 0;
97 static int ip_rsvp_on;
98 struct socket *ip_rsvpd;
99
100 int ipforwarding = 0;
101 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW,
102 &ipforwarding, 0, "Enable IP forwarding between interfaces");
103
104 static int ipsendredirects = 1; /* XXX */
105 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW,
106 &ipsendredirects, 0, "Enable sending IP redirects");
107
108 int ip_defttl = IPDEFTTL;
109 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW,
110 &ip_defttl, 0, "Maximum TTL on IP packets");
111
112 static int ip_dosourceroute = 0;
113 SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW,
114 &ip_dosourceroute, 0, "Enable forwarding source routed IP packets");
115
116 static int ip_acceptsourceroute = 0;
117 SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute,
118 CTLFLAG_RW, &ip_acceptsourceroute, 0,
119 "Enable accepting source routed IP packets");
120
121 static int ip_keepfaith = 0;
122 SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW,
123 &ip_keepfaith, 0,
124 "Enable packet capture for FAITH IPv4->IPv6 translater daemon");
125
126 static int nipq = 0; /* total # of reass queues */
127 static int maxnipq;
128 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragpackets, CTLFLAG_RW,
129 &maxnipq, 0,
130 "Maximum number of IPv4 fragment reassembly queue entries");
131
132 static int maxfragsperpacket;
133 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW,
134 &maxfragsperpacket, 0,
135 "Maximum number of IPv4 fragments allowed per packet");
136
137 static int ip_sendsourcequench = 0;
138 SYSCTL_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW,
139 &ip_sendsourcequench, 0,
140 "Enable the transmission of source quench packets");
141
142 /*
143 * XXX - Setting ip_checkinterface mostly implements the receive side of
144 * the Strong ES model described in RFC 1122, but since the routing table
145 * and transmit implementation do not implement the Strong ES model,
146 * setting this to 1 results in an odd hybrid.
147 *
148 * XXX - ip_checkinterface currently must be disabled if you use ipnat
149 * to translate the destination address to another local interface.
150 *
151 * XXX - ip_checkinterface must be disabled if you add IP aliases
152 * to the loopback interface instead of the interface where the
153 * packets for those addresses are received.
154 */
155 static int ip_checkinterface = 0;
156 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW,
157 &ip_checkinterface, 0, "Verify packet arrives on correct interface");
158
159 #ifdef DIAGNOSTIC
160 static int ipprintfs = 0;
161 #endif
162
163 static int ipqmaxlen = IFQ_MAXLEN;
164
165 extern struct domain inetdomain;
166 extern struct ipprotosw inetsw[];
167 u_char ip_protox[IPPROTO_MAX];
168 struct in_ifaddrhead in_ifaddrhead; /* first inet address */
169 struct in_ifaddrhashhead *in_ifaddrhashtbl; /* inet addr hash table */
170 u_long in_ifaddrhmask; /* mask for hash table */
171 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW,
172 &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue");
173 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD,
174 &ipintrq.ifq_drops, 0, "Number of packets dropped from the IP input queue");
175
176 struct ipstat ipstat;
177 SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW,
178 &ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)");
179
180 /* Packet reassembly stuff */
181 #define IPREASS_NHASH_LOG2 6
182 #define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2)
183 #define IPREASS_HMASK (IPREASS_NHASH - 1)
184 #define IPREASS_HASH(x,y) \
185 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK)
186
187 static struct ipq ipq[IPREASS_NHASH];
188 struct callout ipport_tick_callout;
189 const int ipintrq_present = 1;
190
191 #ifdef IPCTL_DEFMTU
192 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
193 &ip_mtu, 0, "Default MTU");
194 #endif
195
196 #ifdef IPSTEALTH
197 static int ipstealth = 0;
198 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW,
199 &ipstealth, 0, "");
200 #endif
201
202
203 /* Firewall hooks */
204 ip_fw_chk_t *ip_fw_chk_ptr;
205 int fw_enable = 1 ;
206 int fw_one_pass = 1;
207
208 /* Dummynet hooks */
209 ip_dn_io_t *ip_dn_io_ptr;
210
211 int (*fr_checkp) __P((struct ip *, int, struct ifnet *, int, struct mbuf **)) = NULL;
212
213 /*
214 * XXX this is ugly -- the following two global variables are
215 * used to store packet state while it travels through the stack.
216 * Note that the code even makes assumptions on the size and
217 * alignment of fields inside struct ip_srcrt so e.g. adding some
218 * fields will break the code. This needs to be fixed.
219 *
220 * We need to save the IP options in case a protocol wants to respond
221 * to an incoming packet over the same route if the packet got here
222 * using IP source routing. This allows connection establishment and
223 * maintenance when the remote end is on a network that is not known
224 * to us.
225 */
226 static int ip_nhops = 0;
227 static struct ip_srcrt {
228 struct in_addr dst; /* final destination */
229 char nop; /* one NOP to align */
230 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */
231 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)];
232 } ip_srcrt;
233
234 static void save_rte(u_char *, struct in_addr);
235 static int ip_dooptions(struct mbuf *m, int,
236 struct sockaddr_in *next_hop);
237 static void ip_forward(struct mbuf *m, int srcrt,
238 struct sockaddr_in *next_hop);
239 static void ip_freef(struct ipq *);
240 static struct mbuf *ip_reass(struct mbuf *, struct ipq *,
241 struct ipq *, u_int32_t *, u_int16_t *);
242 static void ipintr(void);
243
244 /*
245 * IP initialization: fill in IP protocol switch table.
246 * All protocols not implemented in kernel go to raw IP protocol handler.
247 */
248 void
249 ip_init()
250 {
251 register struct ipprotosw *pr;
252 register int i;
253
254 TAILQ_INIT(&in_ifaddrhead);
255 in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask);
256 pr = (struct ipprotosw *)pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
257 if (pr == 0)
258 panic("ip_init");
259 for (i = 0; i < IPPROTO_MAX; i++)
260 ip_protox[i] = pr - inetsw;
261 for (pr = (struct ipprotosw *)inetdomain.dom_protosw;
262 pr < (struct ipprotosw *)inetdomain.dom_protoswNPROTOSW; pr++)
263 if (pr->pr_domain->dom_family == PF_INET &&
264 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
265 ip_protox[pr->pr_protocol] = pr - inetsw;
266
267 for (i = 0; i < IPREASS_NHASH; i++)
268 ipq[i].next = ipq[i].prev = &ipq[i];
269
270 maxnipq = nmbclusters / 32;
271 maxfragsperpacket = 16;
272
273 /* Start ipport_tick. */
274 callout_init(&ipport_tick_callout);
275 ipport_tick(NULL);
276 EVENTHANDLER_REGISTER(shutdown_pre_sync, ip_fini, NULL,
277 SHUTDOWN_PRI_DEFAULT);
278
279 #ifndef RANDOM_IP_ID
280 ip_id = time_second & 0xffff;
281 #endif
282 ipintrq.ifq_maxlen = ipqmaxlen;
283
284 register_netisr(NETISR_IP, ipintr);
285 }
286
287 void ip_fini(xtp)
288 void *xtp;
289 {
290 callout_stop(&ipport_tick_callout);
291 }
292
293
294 /*
295 * XXX watch out this one. It is perhaps used as a cache for
296 * the most recently used route ? it is cleared in in_addroute()
297 * when a new route is successfully created.
298 */
299 struct route ipforward_rt;
300 static struct sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET };
301
302 /*
303 * Ip input routine. Checksum and byte swap header. If fragmented
304 * try to reassemble. Process options. Pass to next level.
305 */
306 void
307 ip_input(struct mbuf *m)
308 {
309 struct ip *ip;
310 struct ipq *fp;
311 struct in_ifaddr *ia = NULL;
312 struct ifaddr *ifa;
313 int i, hlen, checkif;
314 u_short sum;
315 struct in_addr pkt_dst;
316 u_int32_t divert_info = 0; /* packet divert/tee info */
317 struct ip_fw_args args;
318 #ifdef FAST_IPSEC
319 struct m_tag *mtag;
320 struct tdb_ident *tdbi;
321 struct secpolicy *sp;
322 int s, error;
323 #endif /* FAST_IPSEC */
324
325 args.eh = NULL;
326 args.oif = NULL;
327 args.rule = NULL;
328 args.divert_rule = 0; /* divert cookie */
329 args.next_hop = NULL;
330
331 /* Grab info from MT_TAG mbufs prepended to the chain. */
332 for (; m && m->m_type == MT_TAG; m = m->m_next) {
333 switch(m->_m_tag_id) {
334 default:
335 printf("ip_input: unrecognised MT_TAG tag %d\n",
336 m->_m_tag_id);
337 break;
338
339 case PACKET_TAG_DUMMYNET:
340 args.rule = ((struct dn_pkt *)m)->rule;
341 break;
342
343 case PACKET_TAG_DIVERT:
344 args.divert_rule = (int)m->m_hdr.mh_data & 0xffff;
345 break;
346
347 case PACKET_TAG_IPFORWARD:
348 args.next_hop = (struct sockaddr_in *)m->m_hdr.mh_data;
349 break;
350 }
351 }
352
353 KASSERT(m != NULL && (m->m_flags & M_PKTHDR) != 0,
354 ("ip_input: no HDR"));
355
356 if (args.rule) { /* dummynet already filtered us */
357 ip = mtod(m, struct ip *);
358 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
359 goto iphack ;
360 }
361
362 ipstat.ips_total++;
363
364 if (m->m_pkthdr.len < sizeof(struct ip))
365 goto tooshort;
366
367 if (m->m_len < sizeof (struct ip) &&
368 (m = m_pullup(m, sizeof (struct ip))) == 0) {
369 ipstat.ips_toosmall++;
370 return;
371 }
372 ip = mtod(m, struct ip *);
373
374 if (IP_VHL_V(ip->ip_vhl) != IPVERSION) {
375 ipstat.ips_badvers++;
376 goto bad;
377 }
378
379 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
380 if (hlen < sizeof(struct ip)) { /* minimum header length */
381 ipstat.ips_badhlen++;
382 goto bad;
383 }
384 if (hlen > m->m_len) {
385 if ((m = m_pullup(m, hlen)) == 0) {
386 ipstat.ips_badhlen++;
387 return;
388 }
389 ip = mtod(m, struct ip *);
390 }
391
392 /* 127/8 must not appear on wire - RFC1122 */
393 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
394 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
395 if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
396 ipstat.ips_badaddr++;
397 goto bad;
398 }
399 }
400
401 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
402 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
403 } else {
404 if (hlen == sizeof(struct ip)) {
405 sum = in_cksum_hdr(ip);
406 } else {
407 sum = in_cksum(m, hlen);
408 }
409 }
410 if (sum) {
411 ipstat.ips_badsum++;
412 goto bad;
413 }
414
415 /*
416 * Convert fields to host representation.
417 */
418 ip->ip_len = ntohs(ip->ip_len);
419 if (ip->ip_len < hlen) {
420 ipstat.ips_badlen++;
421 goto bad;
422 }
423 ip->ip_off = ntohs(ip->ip_off);
424
425 /*
426 * Check that the amount of data in the buffers
427 * is as at least much as the IP header would have us expect.
428 * Trim mbufs if longer than we expect.
429 * Drop packet if shorter than we expect.
430 */
431 if (m->m_pkthdr.len < ip->ip_len) {
432 tooshort:
433 ipstat.ips_tooshort++;
434 goto bad;
435 }
436 if (m->m_pkthdr.len > ip->ip_len) {
437 if (m->m_len == m->m_pkthdr.len) {
438 m->m_len = ip->ip_len;
439 m->m_pkthdr.len = ip->ip_len;
440 } else
441 m_adj(m, ip->ip_len - m->m_pkthdr.len);
442 }
443 #if defined(IPSEC) && !defined(IPSEC_FILTERGIF)
444 /*
445 * Bypass packet filtering for packets from a tunnel (gif).
446 */
447 if (ipsec_gethist(m, NULL))
448 goto pass;
449 #endif
450 #if defined(FAST_IPSEC) && !defined(IPSEC_FILTERGIF)
451 /*
452 * Bypass packet filtering for packets from a tunnel (gif).
453 */
454 if (m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL)
455 goto pass;
456 #endif
457
458 /*
459 * IpHack's section.
460 * Right now when no processing on packet has done
461 * and it is still fresh out of network we do our black
462 * deals with it.
463 * - Firewall: deny/allow/divert
464 * - Xlate: translate packet's addr/port (NAT).
465 * - Pipe: pass pkt through dummynet.
466 * - Wrap: fake packet's addr/port <unimpl.>
467 * - Encapsulate: put it in another IP and send out. <unimp.>
468 */
469
470 iphack:
471 /*
472 * Check if we want to allow this packet to be processed.
473 * Consider it to be bad if not.
474 */
475 if (fr_checkp) {
476 struct mbuf *m1 = m;
477
478 if ((*fr_checkp)(ip, hlen, m->m_pkthdr.rcvif, 0, &m1) || !m1)
479 return;
480 ip = mtod(m = m1, struct ip *);
481 }
482 if (fw_enable && IPFW_LOADED) {
483 /*
484 * If we've been forwarded from the output side, then
485 * skip the firewall a second time
486 */
487 if (args.next_hop)
488 goto ours;
489
490 args.m = m;
491 i = ip_fw_chk_ptr(&args);
492 m = args.m;
493
494 if ( (i & IP_FW_PORT_DENY_FLAG) || m == NULL) { /* drop */
495 if (m)
496 m_freem(m);
497 return;
498 }
499 ip = mtod(m, struct ip *); /* just in case m changed */
500 if (i == 0 && args.next_hop == NULL) /* common case */
501 goto pass;
502 if (DUMMYNET_LOADED && (i & IP_FW_PORT_DYNT_FLAG) != 0) {
503 /* Send packet to the appropriate pipe */
504 ip_dn_io_ptr(m, i&0xffff, DN_TO_IP_IN, &args);
505 return;
506 }
507 #ifdef IPDIVERT
508 if (i != 0 && (i & IP_FW_PORT_DYNT_FLAG) == 0) {
509 /* Divert or tee packet */
510 divert_info = i;
511 goto ours;
512 }
513 #endif
514 if (i == 0 && args.next_hop != NULL)
515 goto pass;
516 /*
517 * if we get here, the packet must be dropped
518 */
519 m_freem(m);
520 return;
521 }
522 pass:
523
524 /*
525 * Process options and, if not destined for us,
526 * ship it on. ip_dooptions returns 1 when an
527 * error was detected (causing an icmp message
528 * to be sent and the original packet to be freed).
529 */
530 ip_nhops = 0; /* for source routed packets */
531 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0, args.next_hop))
532 return;
533
534 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
535 * matter if it is destined to another node, or whether it is
536 * a multicast one, RSVP wants it! and prevents it from being forwarded
537 * anywhere else. Also checks if the rsvp daemon is running before
538 * grabbing the packet.
539 */
540 if (rsvp_on && ip->ip_p==IPPROTO_RSVP)
541 goto ours;
542
543 /*
544 * Check our list of addresses, to see if the packet is for us.
545 * If we don't have any addresses, assume any unicast packet
546 * we receive might be for us (and let the upper layers deal
547 * with it).
548 */
549 if (TAILQ_EMPTY(&in_ifaddrhead) &&
550 (m->m_flags & (M_MCAST|M_BCAST)) == 0)
551 goto ours;
552
553 /*
554 * Cache the destination address of the packet; this may be
555 * changed by use of 'ipfw fwd'.
556 */
557 pkt_dst = args.next_hop ? args.next_hop->sin_addr : ip->ip_dst;
558
559 /*
560 * Enable a consistency check between the destination address
561 * and the arrival interface for a unicast packet (the RFC 1122
562 * strong ES model) if IP forwarding is disabled and the packet
563 * is not locally generated and the packet is not subject to
564 * 'ipfw fwd'.
565 *
566 * XXX - Checking also should be disabled if the destination
567 * address is ipnat'ed to a different interface.
568 *
569 * XXX - Checking is incompatible with IP aliases added
570 * to the loopback interface instead of the interface where
571 * the packets are received.
572 */
573 checkif = ip_checkinterface && (ipforwarding == 0) &&
574 m->m_pkthdr.rcvif != NULL &&
575 ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) &&
576 (args.next_hop == NULL);
577
578 /*
579 * Check for exact addresses in the hash bucket.
580 */
581 LIST_FOREACH(ia, INADDR_HASH(pkt_dst.s_addr), ia_hash) {
582 /*
583 * If the address matches, verify that the packet
584 * arrived via the correct interface if checking is
585 * enabled.
586 */
587 if (IA_SIN(ia)->sin_addr.s_addr == pkt_dst.s_addr &&
588 (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif))
589 goto ours;
590 }
591 /*
592 * Check for broadcast addresses.
593 *
594 * Only accept broadcast packets that arrive via the matching
595 * interface. Reception of forwarded directed broadcasts would
596 * be handled via ip_forward() and ether_output() with the loopback
597 * into the stack for SIMPLEX interfaces handled by ether_output().
598 */
599 if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
600 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
601 if (ifa->ifa_addr->sa_family != AF_INET)
602 continue;
603 ia = ifatoia(ifa);
604 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
605 pkt_dst.s_addr)
606 goto ours;
607 if (ia->ia_netbroadcast.s_addr == pkt_dst.s_addr)
608 goto ours;
609 #ifdef BOOTP_COMPAT
610 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY)
611 goto ours;
612 #endif
613 }
614 }
615 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
616 struct in_multi *inm;
617 if (ip_mrouter) {
618 /*
619 * If we are acting as a multicast router, all
620 * incoming multicast packets are passed to the
621 * kernel-level multicast forwarding function.
622 * The packet is returned (relatively) intact; if
623 * ip_mforward() returns a non-zero value, the packet
624 * must be discarded, else it may be accepted below.
625 */
626 if (ip_mforward &&
627 ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) {
628 ipstat.ips_cantforward++;
629 m_freem(m);
630 return;
631 }
632
633 /*
634 * The process-level routing daemon needs to receive
635 * all multicast IGMP packets, whether or not this
636 * host belongs to their destination groups.
637 */
638 if (ip->ip_p == IPPROTO_IGMP)
639 goto ours;
640 ipstat.ips_forward++;
641 }
642 /*
643 * See if we belong to the destination multicast group on the
644 * arrival interface.
645 */
646 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
647 if (inm == NULL) {
648 ipstat.ips_notmember++;
649 m_freem(m);
650 return;
651 }
652 goto ours;
653 }
654 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
655 goto ours;
656 if (ip->ip_dst.s_addr == INADDR_ANY)
657 goto ours;
658
659 /*
660 * FAITH(Firewall Aided Internet Translator)
661 */
662 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
663 if (ip_keepfaith) {
664 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP)
665 goto ours;
666 }
667 m_freem(m);
668 return;
669 }
670
671 /*
672 * Not for us; forward if possible and desirable.
673 */
674 if (ipforwarding == 0) {
675 ipstat.ips_cantforward++;
676 m_freem(m);
677 } else {
678 #ifdef IPSEC
679 /*
680 * Enforce inbound IPsec SPD.
681 */
682 if (ipsec4_in_reject(m, NULL)) {
683 ipsecstat.in_polvio++;
684 goto bad;
685 }
686 #endif /* IPSEC */
687 #ifdef FAST_IPSEC
688 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
689 s = splnet();
690 if (mtag != NULL) {
691 tdbi = (struct tdb_ident *)(mtag + 1);
692 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
693 } else {
694 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
695 IP_FORWARDING, &error);
696 }
697 if (sp == NULL) { /* NB: can happen if error */
698 splx(s);
699 /*XXX error stat???*/
700 DPRINTF(("ip_input: no SP for forwarding\n")); /*XXX*/
701 goto bad;
702 }
703
704 /*
705 * Check security policy against packet attributes.
706 */
707 error = ipsec_in_reject(sp, m);
708 KEY_FREESP(&sp);
709 splx(s);
710 if (error) {
711 ipstat.ips_cantforward++;
712 goto bad;
713 }
714 #endif /* FAST_IPSEC */
715 ip_forward(m, 0, args.next_hop);
716 }
717 return;
718
719 ours:
720 #ifdef IPSTEALTH
721 /*
722 * IPSTEALTH: Process non-routing options only
723 * if the packet is destined for us.
724 */
725 if (ipstealth && hlen > sizeof (struct ip) &&
726 ip_dooptions(m, 1, args.next_hop))
727 return;
728 #endif /* IPSTEALTH */
729
730 /* Count the packet in the ip address stats */
731 if (ia != NULL) {
732 ia->ia_ifa.if_ipackets++;
733 ia->ia_ifa.if_ibytes += m->m_pkthdr.len;
734 }
735
736 /*
737 * If offset or IP_MF are set, must reassemble.
738 * Otherwise, nothing need be done.
739 * (We could look in the reassembly queue to see
740 * if the packet was previously fragmented,
741 * but it's not worth the time; just let them time out.)
742 */
743 if (ip->ip_off & (IP_MF | IP_OFFMASK)) {
744
745 /* If maxnipq is 0, never accept fragments. */
746 if (maxnipq == 0) {
747 ipstat.ips_fragments++;
748 ipstat.ips_fragdropped++;
749 goto bad;
750 }
751
752 sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
753 /*
754 * Look for queue of fragments
755 * of this datagram.
756 */
757 for (fp = ipq[sum].next; fp != &ipq[sum]; fp = fp->next)
758 if (ip->ip_id == fp->ipq_id &&
759 ip->ip_src.s_addr == fp->ipq_src.s_addr &&
760 ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
761 ip->ip_p == fp->ipq_p)
762 goto found;
763
764 fp = 0;
765
766 /*
767 * Enforce upper bound on number of fragmented packets
768 * for which we attempt reassembly;
769 * If maxnipq is -1, accept all fragments without limitation.
770 */
771 if ((nipq > maxnipq) && (maxnipq > 0)) {
772 /*
773 * drop something from the tail of the current queue
774 * before proceeding further
775 */
776 if (ipq[sum].prev == &ipq[sum]) { /* gak */
777 for (i = 0; i < IPREASS_NHASH; i++) {
778 if (ipq[i].prev != &ipq[i]) {
779 ipstat.ips_fragtimeout +=
780 ipq[i].prev->ipq_nfrags;
781 ip_freef(ipq[i].prev);
782 break;
783 }
784 }
785 } else {
786 ipstat.ips_fragtimeout += ipq[sum].prev->ipq_nfrags;
787 ip_freef(ipq[sum].prev);
788 }
789 }
790 found:
791 /*
792 * Adjust ip_len to not reflect header,
793 * convert offset of this to bytes.
794 */
795 ip->ip_len -= hlen;
796 if (ip->ip_off & IP_MF) {
797 /*
798 * Make sure that fragments have a data length
799 * that's a non-zero multiple of 8 bytes.
800 */
801 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
802 ipstat.ips_toosmall++; /* XXX */
803 goto bad;
804 }
805 m->m_flags |= M_FRAG;
806 } else
807 m->m_flags &= ~M_FRAG;
808 ip->ip_off <<= 3;
809
810 /*
811 * Attempt reassembly; if it succeeds, proceed.
812 * ip_reass() will return a different mbuf, and update
813 * the divert info in divert_info and args.divert_rule.
814 */
815 ipstat.ips_fragments++;
816 m->m_pkthdr.header = ip;
817 m = ip_reass(m,
818 fp, &ipq[sum], &divert_info, &args.divert_rule);
819 if (m == 0)
820 return;
821 ipstat.ips_reassembled++;
822 ip = mtod(m, struct ip *);
823 /* Get the header length of the reassembled packet */
824 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
825 #ifdef IPDIVERT
826 /* Restore original checksum before diverting packet */
827 if (divert_info != 0) {
828 ip->ip_len += hlen;
829 ip->ip_len = htons(ip->ip_len);
830 ip->ip_off = htons(ip->ip_off);
831 ip->ip_sum = 0;
832 if (hlen == sizeof(struct ip))
833 ip->ip_sum = in_cksum_hdr(ip);
834 else
835 ip->ip_sum = in_cksum(m, hlen);
836 ip->ip_off = ntohs(ip->ip_off);
837 ip->ip_len = ntohs(ip->ip_len);
838 ip->ip_len -= hlen;
839 }
840 #endif
841 } else
842 ip->ip_len -= hlen;
843
844 #ifdef IPDIVERT
845 /*
846 * Divert or tee packet to the divert protocol if required.
847 */
848 if (divert_info != 0) {
849 struct mbuf *clone = NULL;
850
851 /* Clone packet if we're doing a 'tee' */
852 if ((divert_info & IP_FW_PORT_TEE_FLAG) != 0)
853 clone = m_dup(m, M_DONTWAIT);
854
855 /* Restore packet header fields to original values */
856 ip->ip_len += hlen;
857 ip->ip_len = htons(ip->ip_len);
858 ip->ip_off = htons(ip->ip_off);
859
860 /* Deliver packet to divert input routine */
861 divert_packet(m, 1, divert_info & 0xffff, args.divert_rule);
862 ipstat.ips_delivered++;
863
864 /* If 'tee', continue with original packet */
865 if (clone == NULL)
866 return;
867 m = clone;
868 ip = mtod(m, struct ip *);
869 ip->ip_len += hlen;
870 /*
871 * Jump backwards to complete processing of the
872 * packet. But first clear divert_info to avoid
873 * entering this block again.
874 * We do not need to clear args.divert_rule
875 * or args.next_hop as they will not be used.
876 */
877 divert_info = 0;
878 goto pass;
879 }
880 #endif
881
882 #ifdef IPSEC
883 /*
884 * enforce IPsec policy checking if we are seeing last header.
885 * note that we do not visit this with protocols with pcb layer
886 * code - like udp/tcp/raw ip.
887 */
888 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0 &&
889 ipsec4_in_reject(m, NULL)) {
890 ipsecstat.in_polvio++;
891 goto bad;
892 }
893 #endif
894 #if FAST_IPSEC
895 /*
896 * enforce IPsec policy checking if we are seeing last header.
897 * note that we do not visit this with protocols with pcb layer
898 * code - like udp/tcp/raw ip.
899 */
900 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) {
901 /*
902 * Check if the packet has already had IPsec processing
903 * done. If so, then just pass it along. This tag gets
904 * set during AH, ESP, etc. input handling, before the
905 * packet is returned to the ip input queue for delivery.
906 */
907 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
908 s = splnet();
909 if (mtag != NULL) {
910 tdbi = (struct tdb_ident *)(mtag + 1);
911 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
912 } else {
913 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
914 IP_FORWARDING, &error);
915 }
916 if (sp != NULL) {
917 /*
918 * Check security policy against packet attributes.
919 */
920 error = ipsec_in_reject(sp, m);
921 KEY_FREESP(&sp);
922 } else {
923 /* XXX error stat??? */
924 error = EINVAL;
925 DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/
926 goto bad;
927 }
928 splx(s);
929 if (error)
930 goto bad;
931 }
932 #endif /* FAST_IPSEC */
933
934 /*
935 * Switch out to protocol's input routine.
936 */
937 ipstat.ips_delivered++;
938 if (args.next_hop && ip->ip_p == IPPROTO_TCP) {
939 /* TCP needs IPFORWARD info if available */
940 struct m_hdr tag;
941
942 tag.mh_type = MT_TAG;
943 tag.mh_flags = PACKET_TAG_IPFORWARD;
944 tag.mh_data = (caddr_t)args.next_hop;
945 tag.mh_next = m;
946
947 (*inetsw[ip_protox[ip->ip_p]].pr_input)(
948 (struct mbuf *)&tag, hlen, ip->ip_p);
949 } else
950 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen, ip->ip_p);
951 return;
952 bad:
953 m_freem(m);
954 }
955
956 /*
957 * IP software interrupt routine - to go away sometime soon
958 */
959 static void
960 ipintr(void)
961 {
962 int s;
963 struct mbuf *m;
964
965 while(1) {
966 s = splimp();
967 IF_DEQUEUE(&ipintrq, m);
968 splx(s);
969 if (m == 0)
970 return;
971 ip_input(m);
972 }
973 }
974
975 /*
976 * Take incoming datagram fragment and try to reassemble it into
977 * whole datagram. If a chain for reassembly of this datagram already
978 * exists, then it is given as fp; otherwise have to make a chain.
979 *
980 * When IPDIVERT enabled, keep additional state with each packet that
981 * tells us if we need to divert or tee the packet we're building.
982 * In particular, *divinfo includes the port and TEE flag,
983 * *divert_rule is the number of the matching rule.
984 */
985
986 static struct mbuf *
987 ip_reass(struct mbuf *m, struct ipq *fp, struct ipq *where,
988 u_int32_t *divinfo, u_int16_t *divert_rule)
989 {
990 struct ip *ip = mtod(m, struct ip *);
991 register struct mbuf *p = 0, *q, *nq;
992 struct mbuf *t;
993 int hlen = IP_VHL_HL(ip->ip_vhl) << 2;
994 int i, next;
995
996 /*
997 * Presence of header sizes in mbufs
998 * would confuse code below.
999 */
1000 m->m_data += hlen;
1001 m->m_len -= hlen;
1002
1003 /*
1004 * If first fragment to arrive, create a reassembly queue.
1005 */
1006 if (fp == 0) {
1007 if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL)
1008 goto dropfrag;
1009 fp = mtod(t, struct ipq *);
1010 insque(fp, where);
1011 nipq++;
1012 fp->ipq_nfrags = 1;
1013 fp->ipq_ttl = IPFRAGTTL;
1014 fp->ipq_p = ip->ip_p;
1015 fp->ipq_id = ip->ip_id;
1016 fp->ipq_src = ip->ip_src;
1017 fp->ipq_dst = ip->ip_dst;
1018 fp->ipq_frags = m;
1019 m->m_nextpkt = NULL;
1020 #ifdef IPDIVERT
1021 fp->ipq_div_info = 0;
1022 fp->ipq_div_cookie = 0;
1023 #endif
1024 goto inserted;
1025 } else {
1026 fp->ipq_nfrags++;
1027 }
1028
1029 #define GETIP(m) ((struct ip*)((m)->m_pkthdr.header))
1030
1031 /*
1032 * Find a segment which begins after this one does.
1033 */
1034 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt)
1035 if (GETIP(q)->ip_off > ip->ip_off)
1036 break;
1037
1038 /*
1039 * If there is a preceding segment, it may provide some of
1040 * our data already. If so, drop the data from the incoming
1041 * segment. If it provides all of our data, drop us, otherwise
1042 * stick new segment in the proper place.
1043 *
1044 * If some of the data is dropped from the the preceding
1045 * segment, then it's checksum is invalidated.
1046 */
1047 if (p) {
1048 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off;
1049 if (i > 0) {
1050 if (i >= ip->ip_len)
1051 goto dropfrag;
1052 m_adj(m, i);
1053 m->m_pkthdr.csum_flags = 0;
1054 ip->ip_off += i;
1055 ip->ip_len -= i;
1056 }
1057 m->m_nextpkt = p->m_nextpkt;
1058 p->m_nextpkt = m;
1059 } else {
1060 m->m_nextpkt = fp->ipq_frags;
1061 fp->ipq_frags = m;
1062 }
1063
1064 /*
1065 * While we overlap succeeding segments trim them or,
1066 * if they are completely covered, dequeue them.
1067 */
1068 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off;
1069 q = nq) {
1070 i = (ip->ip_off + ip->ip_len) -
1071 GETIP(q)->ip_off;
1072 if (i < GETIP(q)->ip_len) {
1073 GETIP(q)->ip_len -= i;
1074 GETIP(q)->ip_off += i;
1075 m_adj(q, i);
1076 q->m_pkthdr.csum_flags = 0;
1077 break;
1078 }
1079 nq = q->m_nextpkt;
1080 m->m_nextpkt = nq;
1081 ipstat.ips_fragdropped++;
1082 fp->ipq_nfrags--;
1083 m_freem(q);
1084 }
1085
1086 inserted:
1087
1088 #ifdef IPDIVERT
1089 /*
1090 * Transfer firewall instructions to the fragment structure.
1091 * Only trust info in the fragment at offset 0.
1092 */
1093 if (ip->ip_off == 0) {
1094 fp->ipq_div_info = *divinfo;
1095 fp->ipq_div_cookie = *divert_rule;
1096 }
1097 *divinfo = 0;
1098 *divert_rule = 0;
1099 #endif
1100
1101 /*
1102 * Check for complete reassembly and perform frag per packet
1103 * limiting.
1104 *
1105 * Frag limiting is performed here so that the nth frag has
1106 * a chance to complete the packet before we drop the packet.
1107 * As a result, n+1 frags are actually allowed per packet, but
1108 * only n will ever be stored. (n = maxfragsperpacket.)
1109 *
1110 */
1111 next = 0;
1112 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
1113 if (GETIP(q)->ip_off != next) {
1114 if (fp->ipq_nfrags > maxfragsperpacket) {
1115 ipstat.ips_fragdropped += fp->ipq_nfrags;
1116 ip_freef(fp);
1117 }
1118 return (0);
1119 }
1120 next += GETIP(q)->ip_len;
1121 }
1122 /* Make sure the last packet didn't have the IP_MF flag */
1123 if (p->m_flags & M_FRAG) {
1124 if (fp->ipq_nfrags > maxfragsperpacket) {
1125 ipstat.ips_fragdropped += fp->ipq_nfrags;
1126 ip_freef(fp);
1127 }
1128 return (0);
1129 }
1130
1131 /*
1132 * Reassembly is complete. Make sure the packet is a sane size.
1133 */
1134 q = fp->ipq_frags;
1135 ip = GETIP(q);
1136 if (next + (IP_VHL_HL(ip->ip_vhl) << 2) > IP_MAXPACKET) {
1137 ipstat.ips_toolong++;
1138 ipstat.ips_fragdropped += fp->ipq_nfrags;
1139 ip_freef(fp);
1140 return (0);
1141 }
1142
1143 /*
1144 * Concatenate fragments.
1145 */
1146 m = q;
1147 t = m->m_next;
1148 m->m_next = 0;
1149 m_cat(m, t);
1150 nq = q->m_nextpkt;
1151 q->m_nextpkt = 0;
1152 for (q = nq; q != NULL; q = nq) {
1153 nq = q->m_nextpkt;
1154 q->m_nextpkt = NULL;
1155 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
1156 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
1157 m_cat(m, q);
1158 }
1159
1160 #ifdef IPDIVERT
1161 /*
1162 * Extract firewall instructions from the fragment structure.
1163 */
1164 *divinfo = fp->ipq_div_info;
1165 *divert_rule = fp->ipq_div_cookie;
1166 #endif
1167
1168 /*
1169 * Create header for new ip packet by
1170 * modifying header of first packet;
1171 * dequeue and discard fragment reassembly header.
1172 * Make header visible.
1173 */
1174 ip->ip_len = next;
1175 ip->ip_src = fp->ipq_src;
1176 ip->ip_dst = fp->ipq_dst;
1177 remque(fp);
1178 nipq--;
1179 (void) m_free(dtom(fp));
1180 m->m_len += (IP_VHL_HL(ip->ip_vhl) << 2);
1181 m->m_data -= (IP_VHL_HL(ip->ip_vhl) << 2);
1182 /* some debugging cruft by sklower, below, will go away soon */
1183 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */
1184 register int plen = 0;
1185 for (t = m; t; t = t->m_next)
1186 plen += t->m_len;
1187 m->m_pkthdr.len = plen;
1188 }
1189 return (m);
1190
1191 dropfrag:
1192 #ifdef IPDIVERT
1193 *divinfo = 0;
1194 *divert_rule = 0;
1195 #endif
1196 ipstat.ips_fragdropped++;
1197 if (fp != 0)
1198 fp->ipq_nfrags--;
1199 m_freem(m);
1200 return (0);
1201
1202 #undef GETIP
1203 }
1204
1205 /*
1206 * Free a fragment reassembly header and all
1207 * associated datagrams.
1208 */
1209 static void
1210 ip_freef(fp)
1211 struct ipq *fp;
1212 {
1213 register struct mbuf *q;
1214
1215 while (fp->ipq_frags) {
1216 q = fp->ipq_frags;
1217 fp->ipq_frags = q->m_nextpkt;
1218 m_freem(q);
1219 }
1220 remque(fp);
1221 (void) m_free(dtom(fp));
1222 nipq--;
1223 }
1224
1225 /*
1226 * IP timer processing;
1227 * if a timer expires on a reassembly
1228 * queue, discard it.
1229 */
1230 void
1231 ip_slowtimo()
1232 {
1233 register struct ipq *fp;
1234 int s = splnet();
1235 int i;
1236
1237 for (i = 0; i < IPREASS_NHASH; i++) {
1238 fp = ipq[i].next;
1239 if (fp == 0)
1240 continue;
1241 while (fp != &ipq[i]) {
1242 --fp->ipq_ttl;
1243 fp = fp->next;
1244 if (fp->prev->ipq_ttl == 0) {
1245 ipstat.ips_fragtimeout += fp->prev->ipq_nfrags;
1246 ip_freef(fp->prev);
1247 }
1248 }
1249 }
1250 /*
1251 * If we are over the maximum number of fragments
1252 * (due to the limit being lowered), drain off
1253 * enough to get down to the new limit.
1254 */
1255 if (maxnipq >= 0 && nipq > maxnipq) {
1256 for (i = 0; i < IPREASS_NHASH; i++) {
1257 while (nipq > maxnipq &&
1258 (ipq[i].next != &ipq[i])) {
1259 ipstat.ips_fragdropped +=
1260 ipq[i].next->ipq_nfrags;
1261 ip_freef(ipq[i].next);
1262 }
1263 }
1264 }
1265 ipflow_slowtimo();
1266 splx(s);
1267 }
1268
1269 /*
1270 * Drain off all datagram fragments.
1271 */
1272 void
1273 ip_drain()
1274 {
1275 int i;
1276
1277 for (i = 0; i < IPREASS_NHASH; i++) {
1278 while (ipq[i].next != &ipq[i]) {
1279 ipstat.ips_fragdropped += ipq[i].next->ipq_nfrags;
1280 ip_freef(ipq[i].next);
1281 }
1282 }
1283 in_rtqdrain();
1284 }
1285
1286 /*
1287 * Do option processing on a datagram,
1288 * possibly discarding it if bad options are encountered,
1289 * or forwarding it if source-routed.
1290 * The pass argument is used when operating in the IPSTEALTH
1291 * mode to tell what options to process:
1292 * [LS]SRR (pass 0) or the others (pass 1).
1293 * The reason for as many as two passes is that when doing IPSTEALTH,
1294 * non-routing options should be processed only if the packet is for us.
1295 * Returns 1 if packet has been forwarded/freed,
1296 * 0 if the packet should be processed further.
1297 */
1298 static int
1299 ip_dooptions(struct mbuf *m, int pass, struct sockaddr_in *next_hop)
1300 {
1301 struct ip *ip = mtod(m, struct ip *);
1302 u_char *cp;
1303 struct in_ifaddr *ia;
1304 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0;
1305 struct in_addr *sin, dst;
1306 n_time ntime;
1307
1308 dst = ip->ip_dst;
1309 cp = (u_char *)(ip + 1);
1310 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip);
1311 for (; cnt > 0; cnt -= optlen, cp += optlen) {
1312 opt = cp[IPOPT_OPTVAL];
1313 if (opt == IPOPT_EOL)
1314 break;
1315 if (opt == IPOPT_NOP)
1316 optlen = 1;
1317 else {
1318 if (cnt < IPOPT_OLEN + sizeof(*cp)) {
1319 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1320 goto bad;
1321 }
1322 optlen = cp[IPOPT_OLEN];
1323 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
1324 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1325 goto bad;
1326 }
1327 }
1328 switch (opt) {
1329
1330 default:
1331 break;
1332
1333 /*
1334 * Source routing with record.
1335 * Find interface with current destination address.
1336 * If none on this machine then drop if strictly routed,
1337 * or do nothing if loosely routed.
1338 * Record interface address and bring up next address
1339 * component. If strictly routed make sure next
1340 * address is on directly accessible net.
1341 */
1342 case IPOPT_LSRR:
1343 case IPOPT_SSRR:
1344 #ifdef IPSTEALTH
1345 if (ipstealth && pass > 0)
1346 break;
1347 #endif
1348 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1349 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1350 goto bad;
1351 }
1352 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1353 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1354 goto bad;
1355 }
1356 ipaddr.sin_addr = ip->ip_dst;
1357 ia = (struct in_ifaddr *)
1358 ifa_ifwithaddr((struct sockaddr *)&ipaddr);
1359 if (ia == 0) {
1360 if (opt == IPOPT_SSRR) {
1361 type = ICMP_UNREACH;
1362 code = ICMP_UNREACH_SRCFAIL;
1363 goto bad;
1364 }
1365 if (!ip_dosourceroute)
1366 goto nosourcerouting;
1367 /*
1368 * Loose routing, and not at next destination
1369 * yet; nothing to do except forward.
1370 */
1371 break;
1372 }
1373 off--; /* 0 origin */
1374 if (off > optlen - (int)sizeof(struct in_addr)) {
1375 /*
1376 * End of source route. Should be for us.
1377 */
1378 if (!ip_acceptsourceroute)
1379 goto nosourcerouting;
1380 save_rte(cp, ip->ip_src);
1381 break;
1382 }
1383 #ifdef IPSTEALTH
1384 if (ipstealth)
1385 goto dropit;
1386 #endif
1387 if (!ip_dosourceroute) {
1388 if (ipforwarding) {
1389 char buf[16]; /* aaa.bbb.ccc.ddd\0 */
1390 /*
1391 * Acting as a router, so generate ICMP
1392 */
1393 nosourcerouting:
1394 strcpy(buf, inet_ntoa(ip->ip_dst));
1395 log(LOG_WARNING,
1396 "attempted source route from %s to %s\n",
1397 inet_ntoa(ip->ip_src), buf);
1398 type = ICMP_UNREACH;
1399 code = ICMP_UNREACH_SRCFAIL;
1400 goto bad;
1401 } else {
1402 /*
1403 * Not acting as a router, so silently drop.
1404 */
1405 #ifdef IPSTEALTH
1406 dropit:
1407 #endif
1408 ipstat.ips_cantforward++;
1409 m_freem(m);
1410 return (1);
1411 }
1412 }
1413
1414 /*
1415 * locate outgoing interface
1416 */
1417 (void)memcpy(&ipaddr.sin_addr, cp + off,
1418 sizeof(ipaddr.sin_addr));
1419
1420 if (opt == IPOPT_SSRR) {
1421 #define INA struct in_ifaddr *
1422 #define SA struct sockaddr *
1423 if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == 0)
1424 ia = (INA)ifa_ifwithnet((SA)&ipaddr);
1425 } else
1426 ia = ip_rtaddr(ipaddr.sin_addr, &ipforward_rt);
1427 if (ia == 0) {
1428 type = ICMP_UNREACH;
1429 code = ICMP_UNREACH_SRCFAIL;
1430 goto bad;
1431 }
1432 ip->ip_dst = ipaddr.sin_addr;
1433 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
1434 sizeof(struct in_addr));
1435 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1436 /*
1437 * Let ip_intr's mcast routing check handle mcast pkts
1438 */
1439 forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr));
1440 break;
1441
1442 case IPOPT_RR:
1443 #ifdef IPSTEALTH
1444 if (ipstealth && pass == 0)
1445 break;
1446 #endif
1447 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1448 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1449 goto bad;
1450 }
1451 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1452 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1453 goto bad;
1454 }
1455 /*
1456 * If no space remains, ignore.
1457 */
1458 off--; /* 0 origin */
1459 if (off > optlen - (int)sizeof(struct in_addr))
1460 break;
1461 (void)memcpy(&ipaddr.sin_addr, &ip->ip_dst,
1462 sizeof(ipaddr.sin_addr));
1463 /*
1464 * locate outgoing interface; if we're the destination,
1465 * use the incoming interface (should be same).
1466 */
1467 if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == 0 &&
1468 (ia = ip_rtaddr(ipaddr.sin_addr,
1469 &ipforward_rt)) == 0) {
1470 type = ICMP_UNREACH;
1471 code = ICMP_UNREACH_HOST;
1472 goto bad;
1473 }
1474 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
1475 sizeof(struct in_addr));
1476 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1477 break;
1478
1479 case IPOPT_TS:
1480 #ifdef IPSTEALTH
1481 if (ipstealth && pass == 0)
1482 break;
1483 #endif
1484 code = cp - (u_char *)ip;
1485 if (optlen < 4 || optlen > 40) {
1486 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1487 goto bad;
1488 }
1489 if ((off = cp[IPOPT_OFFSET]) < 5) {
1490 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1491 goto bad;
1492 }
1493 if (off > optlen - (int)sizeof(int32_t)) {
1494 cp[IPOPT_OFFSET + 1] += (1 << 4);
1495 if ((cp[IPOPT_OFFSET + 1] & 0xf0) == 0) {
1496 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1497 goto bad;
1498 }
1499 break;
1500 }
1501 off--; /* 0 origin */
1502 sin = (struct in_addr *)(cp + off);
1503 switch (cp[IPOPT_OFFSET + 1] & 0x0f) {
1504
1505 case IPOPT_TS_TSONLY:
1506 break;
1507
1508 case IPOPT_TS_TSANDADDR:
1509 if (off + sizeof(n_time) +
1510 sizeof(struct in_addr) > optlen) {
1511 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1512 goto bad;
1513 }
1514 ipaddr.sin_addr = dst;
1515 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr,
1516 m->m_pkthdr.rcvif);
1517 if (ia == 0)
1518 continue;
1519 (void)memcpy(sin, &IA_SIN(ia)->sin_addr,
1520 sizeof(struct in_addr));
1521 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1522 off += sizeof(struct in_addr);
1523 break;
1524
1525 case IPOPT_TS_PRESPEC:
1526 if (off + sizeof(n_time) +
1527 sizeof(struct in_addr) > optlen) {
1528 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1529 goto bad;
1530 }
1531 (void)memcpy(&ipaddr.sin_addr, sin,
1532 sizeof(struct in_addr));
1533 if (ifa_ifwithaddr((SA)&ipaddr) == 0)
1534 continue;
1535 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1536 off += sizeof(struct in_addr);
1537 break;
1538
1539 default:
1540 code = &cp[IPOPT_OFFSET + 1] - (u_char *)ip;
1541 goto bad;
1542 }
1543 ntime = iptime();
1544 (void)memcpy(cp + off, &ntime, sizeof(n_time));
1545 cp[IPOPT_OFFSET] += sizeof(n_time);
1546 }
1547 }
1548 if (forward && ipforwarding) {
1549 ip_forward(m, 1, next_hop);
1550 return (1);
1551 }
1552 return (0);
1553 bad:
1554 icmp_error(m, type, code, 0, 0);
1555 ipstat.ips_badoptions++;
1556 return (1);
1557 }
1558
1559 /*
1560 * Given address of next destination (final or next hop),
1561 * return internet address info of interface to be used to get there.
1562 */
1563 struct in_ifaddr *
1564 ip_rtaddr(dst, rt)
1565 struct in_addr dst;
1566 struct route *rt;
1567 {
1568 register struct sockaddr_in *sin;
1569
1570 sin = (struct sockaddr_in *)&rt->ro_dst;
1571
1572 if (rt->ro_rt == 0 ||
1573 dst.s_addr != sin->sin_addr.s_addr) {
1574 if (rt->ro_rt) {
1575 RTFREE(rt->ro_rt);
1576 rt->ro_rt = 0;
1577 }
1578 sin->sin_family = AF_INET;
1579 sin->sin_len = sizeof(*sin);
1580 sin->sin_addr = dst;
1581
1582 rtalloc_ign(rt, RTF_PRCLONING);
1583 }
1584 if (rt->ro_rt == 0)
1585 return ((struct in_ifaddr *)0);
1586 return (ifatoia(rt->ro_rt->rt_ifa));
1587 }
1588
1589 /*
1590 * Save incoming source route for use in replies,
1591 * to be picked up later by ip_srcroute if the receiver is interested.
1592 */
1593 void
1594 save_rte(option, dst)
1595 u_char *option;
1596 struct in_addr dst;
1597 {
1598 unsigned olen;
1599
1600 olen = option[IPOPT_OLEN];
1601 #ifdef DIAGNOSTIC
1602 if (ipprintfs)
1603 printf("save_rte: olen %d\n", olen);
1604 #endif
1605 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst)))
1606 return;
1607 bcopy(option, ip_srcrt.srcopt, olen);
1608 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1609 ip_srcrt.dst = dst;
1610 }
1611
1612 /*
1613 * Retrieve incoming source route for use in replies,
1614 * in the same form used by setsockopt.
1615 * The first hop is placed before the options, will be removed later.
1616 */
1617 struct mbuf *
1618 ip_srcroute()
1619 {
1620 register struct in_addr *p, *q;
1621 register struct mbuf *m;
1622
1623 if (ip_nhops == 0)
1624 return ((struct mbuf *)0);
1625 m = m_get(M_DONTWAIT, MT_HEADER);
1626 if (m == 0)
1627 return ((struct mbuf *)0);
1628
1629 #define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt))
1630
1631 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
1632 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) +
1633 OPTSIZ;
1634 #ifdef DIAGNOSTIC
1635 if (ipprintfs)
1636 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len);
1637 #endif
1638
1639 /*
1640 * First save first hop for return route
1641 */
1642 p = &ip_srcrt.route[ip_nhops - 1];
1643 *(mtod(m, struct in_addr *)) = *p--;
1644 #ifdef DIAGNOSTIC
1645 if (ipprintfs)
1646 printf(" hops %lx", (u_long)ntohl(mtod(m, struct in_addr *)->s_addr));
1647 #endif
1648
1649 /*
1650 * Copy option fields and padding (nop) to mbuf.
1651 */
1652 ip_srcrt.nop = IPOPT_NOP;
1653 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
1654 (void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr),
1655 &ip_srcrt.nop, OPTSIZ);
1656 q = (struct in_addr *)(mtod(m, caddr_t) +
1657 sizeof(struct in_addr) + OPTSIZ);
1658 #undef OPTSIZ
1659 /*
1660 * Record return path as an IP source route,
1661 * reversing the path (pointers are now aligned).
1662 */
1663 while (p >= ip_srcrt.route) {
1664 #ifdef DIAGNOSTIC
1665 if (ipprintfs)
1666 printf(" %lx", (u_long)ntohl(q->s_addr));
1667 #endif
1668 *q++ = *p--;
1669 }
1670 /*
1671 * Last hop goes to final destination.
1672 */
1673 *q = ip_srcrt.dst;
1674 #ifdef DIAGNOSTIC
1675 if (ipprintfs)
1676 printf(" %lx\n", (u_long)ntohl(q->s_addr));
1677 #endif
1678 return (m);
1679 }
1680
1681 /*
1682 * Strip out IP options, at higher
1683 * level protocol in the kernel.
1684 * Second argument is buffer to which options
1685 * will be moved, and return value is their length.
1686 * XXX should be deleted; last arg currently ignored.
1687 */
1688 void
1689 ip_stripoptions(m, mopt)
1690 register struct mbuf *m;
1691 struct mbuf *mopt;
1692 {
1693 register int i;
1694 struct ip *ip = mtod(m, struct ip *);
1695 register caddr_t opts;
1696 int olen;
1697
1698 olen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip);
1699 opts = (caddr_t)(ip + 1);
1700 i = m->m_len - (sizeof (struct ip) + olen);
1701 bcopy(opts + olen, opts, (unsigned)i);
1702 m->m_len -= olen;
1703 if (m->m_flags & M_PKTHDR)
1704 m->m_pkthdr.len -= olen;
1705 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, sizeof(struct ip) >> 2);
1706 }
1707
1708 u_char inetctlerrmap[PRC_NCMDS] = {
1709 0, 0, 0, 0,
1710 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
1711 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
1712 EMSGSIZE, EHOSTUNREACH, 0, 0,
1713 0, 0, EHOSTUNREACH, 0,
1714 ENOPROTOOPT, ECONNREFUSED
1715 };
1716
1717 /*
1718 * Forward a packet. If some error occurs return the sender
1719 * an icmp packet. Note we can't always generate a meaningful
1720 * icmp message because icmp doesn't have a large enough repertoire
1721 * of codes and types.
1722 *
1723 * If not forwarding, just drop the packet. This could be confusing
1724 * if ipforwarding was zero but some routing protocol was advancing
1725 * us as a gateway to somewhere. However, we must let the routing
1726 * protocol deal with that.
1727 *
1728 * The srcrt parameter indicates whether the packet is being forwarded
1729 * via a source route.
1730 */
1731 static void
1732 ip_forward(struct mbuf *m, int srcrt, struct sockaddr_in *next_hop)
1733 {
1734 struct ip *ip = mtod(m, struct ip *);
1735 struct sockaddr_in *sin;
1736 struct rtentry *rt;
1737 int error, type = 0, code = 0;
1738 struct mbuf *mcopy;
1739 n_long dest;
1740 struct in_addr pkt_dst;
1741 struct ifnet *destifp;
1742 #if defined(IPSEC) || defined(FAST_IPSEC)
1743 struct ifnet dummyifp;
1744 #endif
1745
1746 dest = 0;
1747 /*
1748 * Cache the destination address of the packet; this may be
1749 * changed by use of 'ipfw fwd'.
1750 */
1751 pkt_dst = next_hop ? next_hop->sin_addr : ip->ip_dst;
1752
1753 #ifdef DIAGNOSTIC
1754 if (ipprintfs)
1755 printf("forward: src %lx dst %lx ttl %x\n",
1756 (u_long)ip->ip_src.s_addr, (u_long)pkt_dst.s_addr,
1757 ip->ip_ttl);
1758 #endif
1759
1760
1761 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(pkt_dst) == 0) {
1762 ipstat.ips_cantforward++;
1763 m_freem(m);
1764 return;
1765 }
1766 #ifdef IPSTEALTH
1767 if (!ipstealth) {
1768 #endif
1769 if (ip->ip_ttl <= IPTTLDEC) {
1770 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS,
1771 dest, 0);
1772 return;
1773 }
1774 #ifdef IPSTEALTH
1775 }
1776 #endif
1777
1778 sin = (struct sockaddr_in *)&ipforward_rt.ro_dst;
1779 if ((rt = ipforward_rt.ro_rt) == 0 ||
1780 pkt_dst.s_addr != sin->sin_addr.s_addr) {
1781 if (ipforward_rt.ro_rt) {
1782 RTFREE(ipforward_rt.ro_rt);
1783 ipforward_rt.ro_rt = 0;
1784 }
1785 sin->sin_family = AF_INET;
1786 sin->sin_len = sizeof(*sin);
1787 sin->sin_addr = pkt_dst;
1788
1789 rtalloc_ign(&ipforward_rt, RTF_PRCLONING);
1790 if (ipforward_rt.ro_rt == 0) {
1791 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0);
1792 return;
1793 }
1794 rt = ipforward_rt.ro_rt;
1795 }
1796
1797 /*
1798 * Save the IP header and at most 8 bytes of the payload,
1799 * in case we need to generate an ICMP message to the src.
1800 *
1801 * XXX this can be optimized a lot by saving the data in a local
1802 * buffer on the stack (72 bytes at most), and only allocating the
1803 * mbuf if really necessary. The vast majority of the packets
1804 * are forwarded without having to send an ICMP back (either
1805 * because unnecessary, or because rate limited), so we are
1806 * really we are wasting a lot of work here.
1807 *
1808 * We don't use m_copy() because it might return a reference
1809 * to a shared cluster. Both this function and ip_output()
1810 * assume exclusive access to the IP header in `m', so any
1811 * data in a cluster may change before we reach icmp_error().
1812 */
1813 MGET(mcopy, M_DONTWAIT, m->m_type);
1814 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_DONTWAIT)) {
1815 /*
1816 * It's probably ok if the pkthdr dup fails (because
1817 * the deep copy of the tag chain failed), but for now
1818 * be conservative and just discard the copy since
1819 * code below may some day want the tags.
1820 */
1821 m_free(mcopy);
1822 mcopy = NULL;
1823 }
1824 if (mcopy != NULL) {
1825 mcopy->m_len = imin((IP_VHL_HL(ip->ip_vhl) << 2) + 8,
1826 (int)ip->ip_len);
1827 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1828 }
1829
1830 #ifdef IPSTEALTH
1831 if (!ipstealth) {
1832 #endif
1833 ip->ip_ttl -= IPTTLDEC;
1834 #ifdef IPSTEALTH
1835 }
1836 #endif
1837
1838 /*
1839 * If forwarding packet using same interface that it came in on,
1840 * perhaps should send a redirect to sender to shortcut a hop.
1841 * Only send redirect if source is sending directly to us,
1842 * and if packet was not source routed (or has any options).
1843 * Also, don't send redirect if forwarding using a default route
1844 * or a route modified by a redirect.
1845 */
1846 if (rt->rt_ifp == m->m_pkthdr.rcvif &&
1847 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1848 satosin(rt_key(rt))->sin_addr.s_addr != 0 &&
1849 ipsendredirects && !srcrt && !next_hop) {
1850 #define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa))
1851 u_long src = ntohl(ip->ip_src.s_addr);
1852
1853 if (RTA(rt) &&
1854 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1855 if (rt->rt_flags & RTF_GATEWAY)
1856 dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
1857 else
1858 dest = pkt_dst.s_addr;
1859 /* Router requirements says to only send host redirects */
1860 type = ICMP_REDIRECT;
1861 code = ICMP_REDIRECT_HOST;
1862 #ifdef DIAGNOSTIC
1863 if (ipprintfs)
1864 printf("redirect (%d) to %lx\n", code, (u_long)dest);
1865 #endif
1866 }
1867 }
1868
1869 {
1870 struct m_hdr tag;
1871
1872 if (next_hop) {
1873 /* Pass IPFORWARD info if available */
1874
1875 tag.mh_type = MT_TAG;
1876 tag.mh_flags = PACKET_TAG_IPFORWARD;
1877 tag.mh_data = (caddr_t)next_hop;
1878 tag.mh_next = m;
1879 m = (struct mbuf *)&tag;
1880 }
1881 error = ip_output(m, (struct mbuf *)0, &ipforward_rt,
1882 IP_FORWARDING, 0, NULL);
1883 }
1884 if (error)
1885 ipstat.ips_cantforward++;
1886 else {
1887 ipstat.ips_forward++;
1888 if (type)
1889 ipstat.ips_redirectsent++;
1890 else {
1891 if (mcopy) {
1892 ipflow_create(&ipforward_rt, mcopy);
1893 m_freem(mcopy);
1894 }
1895 return;
1896 }
1897 }
1898 if (mcopy == NULL)
1899 return;
1900 destifp = NULL;
1901
1902 switch (error) {
1903
1904 case 0: /* forwarded, but need redirect */
1905 /* type, code set above */
1906 break;
1907
1908 case ENETUNREACH: /* shouldn't happen, checked above */
1909 case EHOSTUNREACH:
1910 case ENETDOWN:
1911 case EHOSTDOWN:
1912 default:
1913 type = ICMP_UNREACH;
1914 code = ICMP_UNREACH_HOST;
1915 break;
1916
1917 case EMSGSIZE:
1918 type = ICMP_UNREACH;
1919 code = ICMP_UNREACH_NEEDFRAG;
1920 #ifdef IPSEC
1921 /*
1922 * If the packet is routed over IPsec tunnel, tell the
1923 * originator the tunnel MTU.
1924 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
1925 * XXX quickhack!!!
1926 */
1927 if (ipforward_rt.ro_rt) {
1928 struct secpolicy *sp = NULL;
1929 int ipsecerror;
1930 int ipsechdr;
1931 struct route *ro;
1932
1933 sp = ipsec4_getpolicybyaddr(mcopy,
1934 IPSEC_DIR_OUTBOUND,
1935 IP_FORWARDING,
1936 &ipsecerror);
1937
1938 if (sp == NULL)
1939 destifp = ipforward_rt.ro_rt->rt_ifp;
1940 else {
1941 /* count IPsec header size */
1942 ipsechdr = ipsec4_hdrsiz(mcopy,
1943 IPSEC_DIR_OUTBOUND,
1944 NULL);
1945
1946 /*
1947 * find the correct route for outer IPv4
1948 * header, compute tunnel MTU.
1949 *
1950 * XXX BUG ALERT
1951 * The "dummyifp" code relies upon the fact
1952 * that icmp_error() touches only ifp->if_mtu.
1953 */
1954 /*XXX*/
1955 destifp = NULL;
1956 if (sp->req != NULL
1957 && sp->req->sav != NULL
1958 && sp->req->sav->sah != NULL) {
1959 ro = &sp->req->sav->sah->sa_route;
1960 if (ro->ro_rt && ro->ro_rt->rt_ifp) {
1961 dummyifp.if_mtu =
1962 ro->ro_rt->rt_ifp->if_mtu;
1963 dummyifp.if_mtu -= ipsechdr;
1964 destifp = &dummyifp;
1965 }
1966 }
1967
1968 key_freesp(sp);
1969 }
1970 }
1971 #elif FAST_IPSEC
1972 /*
1973 * If the packet is routed over IPsec tunnel, tell the
1974 * originator the tunnel MTU.
1975 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
1976 * XXX quickhack!!!
1977 */
1978 if (ipforward_rt.ro_rt) {
1979 struct secpolicy *sp = NULL;
1980 int ipsecerror;
1981 int ipsechdr;
1982 struct route *ro;
1983
1984 sp = ipsec_getpolicybyaddr(mcopy,
1985 IPSEC_DIR_OUTBOUND,
1986 IP_FORWARDING,
1987 &ipsecerror);
1988
1989 if (sp == NULL)
1990 destifp = ipforward_rt.ro_rt->rt_ifp;
1991 else {
1992 /* count IPsec header size */
1993 ipsechdr = ipsec4_hdrsiz(mcopy,
1994 IPSEC_DIR_OUTBOUND,
1995 NULL);
1996
1997 /*
1998 * find the correct route for outer IPv4
1999 * header, compute tunnel MTU.
2000 *
2001 * XXX BUG ALERT
2002 * The "dummyifp" code relies upon the fact
2003 * that icmp_error() touches only ifp->if_mtu.
2004 */
2005 /*XXX*/
2006 destifp = NULL;
2007 if (sp->req != NULL
2008 && sp->req->sav != NULL
2009 && sp->req->sav->sah != NULL) {
2010 ro = &sp->req->sav->sah->sa_route;
2011 if (ro->ro_rt && ro->ro_rt->rt_ifp) {
2012 dummyifp.if_mtu =
2013 ro->ro_rt->rt_ifp->if_mtu;
2014 dummyifp.if_mtu -= ipsechdr;
2015 destifp = &dummyifp;
2016 }
2017 }
2018
2019 KEY_FREESP(&sp);
2020 }
2021 }
2022 #else /* !IPSEC && !FAST_IPSEC */
2023 if (ipforward_rt.ro_rt)
2024 destifp = ipforward_rt.ro_rt->rt_ifp;
2025 #endif /*IPSEC*/
2026 ipstat.ips_cantfrag++;
2027 break;
2028
2029 case ENOBUFS:
2030 /*
2031 * A router should not generate ICMP_SOURCEQUENCH as
2032 * required in RFC1812 Requirements for IP Version 4 Routers.
2033 * Source quench could be a big problem under DoS attacks,
2034 * or if the underlying interface is rate-limited.
2035 * Those who need source quench packets may re-enable them
2036 * via the net.inet.ip.sendsourcequench sysctl.
2037 */
2038 if (ip_sendsourcequench == 0) {
2039 m_freem(mcopy);
2040 return;
2041 } else {
2042 type = ICMP_SOURCEQUENCH;
2043 code = 0;
2044 }
2045 break;
2046
2047 case EACCES: /* ipfw denied packet */
2048 m_freem(mcopy);
2049 return;
2050 }
2051 icmp_error(mcopy, type, code, dest, destifp);
2052 }
2053
2054 void
2055 ip_savecontrol(inp, mp, ip, m)
2056 register struct inpcb *inp;
2057 register struct mbuf **mp;
2058 register struct ip *ip;
2059 register struct mbuf *m;
2060 {
2061 if (inp->inp_socket->so_options & SO_TIMESTAMP) {
2062 struct timeval tv;
2063
2064 microtime(&tv);
2065 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
2066 SCM_TIMESTAMP, SOL_SOCKET);
2067 if (*mp)
2068 mp = &(*mp)->m_next;
2069 }
2070 if (inp->inp_flags & INP_RECVDSTADDR) {
2071 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
2072 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
2073 if (*mp)
2074 mp = &(*mp)->m_next;
2075 }
2076 #ifdef notyet
2077 /* XXX
2078 * Moving these out of udp_input() made them even more broken
2079 * than they already were.
2080 */
2081 /* options were tossed already */
2082 if (inp->inp_flags & INP_RECVOPTS) {
2083 *mp = sbcreatecontrol((caddr_t) opts_deleted_above,
2084 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
2085 if (*mp)
2086 mp = &(*mp)->m_next;
2087 }
2088 /* ip_srcroute doesn't do what we want here, need to fix */
2089 if (inp->inp_flags & INP_RECVRETOPTS) {
2090 *mp = sbcreatecontrol((caddr_t) ip_srcroute(),
2091 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
2092 if (*mp)
2093 mp = &(*mp)->m_next;
2094 }
2095 #endif
2096 if (inp->inp_flags & INP_RECVIF) {
2097 struct ifnet *ifp;
2098 struct sdlbuf {
2099 struct sockaddr_dl sdl;
2100 u_char pad[32];
2101 } sdlbuf;
2102 struct sockaddr_dl *sdp;
2103 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
2104
2105 if (((ifp = m->m_pkthdr.rcvif))
2106 && ( ifp->if_index && (ifp->if_index <= if_index))) {
2107 sdp = (struct sockaddr_dl *)(ifnet_addrs
2108 [ifp->if_index - 1]->ifa_addr);
2109 /*
2110 * Change our mind and don't try copy.
2111 */
2112 if ((sdp->sdl_family != AF_LINK)
2113 || (sdp->sdl_len > sizeof(sdlbuf))) {
2114 goto makedummy;
2115 }
2116 bcopy(sdp, sdl2, sdp->sdl_len);
2117 } else {
2118 makedummy:
2119 sdl2->sdl_len
2120 = offsetof(struct sockaddr_dl, sdl_data[0]);
2121 sdl2->sdl_family = AF_LINK;
2122 sdl2->sdl_index = 0;
2123 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
2124 }
2125 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len,
2126 IP_RECVIF, IPPROTO_IP);
2127 if (*mp)
2128 mp = &(*mp)->m_next;
2129 }
2130 }
2131
2132 /*
2133 * XXX these routines are called from the upper part of the kernel.
2134 *
2135 * They could also be moved to ip_mroute.c, since all the RSVP
2136 * handling is done there already.
2137 */
2138 int
2139 ip_rsvp_init(struct socket *so)
2140 {
2141 if (so->so_type != SOCK_RAW ||
2142 so->so_proto->pr_protocol != IPPROTO_RSVP)
2143 return EOPNOTSUPP;
2144
2145 if (ip_rsvpd != NULL)
2146 return EADDRINUSE;
2147
2148 ip_rsvpd = so;
2149 /*
2150 * This may seem silly, but we need to be sure we don't over-increment
2151 * the RSVP counter, in case something slips up.
2152 */
2153 if (!ip_rsvp_on) {
2154 ip_rsvp_on = 1;
2155 rsvp_on++;
2156 }
2157
2158 return 0;
2159 }
2160
2161 int
2162 ip_rsvp_done(void)
2163 {
2164 ip_rsvpd = NULL;
2165 /*
2166 * This may seem silly, but we need to be sure we don't over-decrement
2167 * the RSVP counter, in case something slips up.
2168 */
2169 if (ip_rsvp_on) {
2170 ip_rsvp_on = 0;
2171 rsvp_on--;
2172 }
2173 return 0;
2174 }
2175
2176 void
2177 rsvp_input(struct mbuf *m, int off, int proto) /* XXX must fixup manually */
2178 {
2179 if (rsvp_input_p) { /* call the real one if loaded */
2180 rsvp_input_p(m, off, proto);
2181 return;
2182 }
2183
2184 /* Can still get packets with rsvp_on = 0 if there is a local member
2185 * of the group to which the RSVP packet is addressed. But in this
2186 * case we want to throw the packet away.
2187 */
2188
2189 if (!rsvp_on) {
2190 m_freem(m);
2191 return;
2192 }
2193
2194 if (ip_rsvpd != NULL) {
2195 rip_input(m, off, proto);
2196 return;
2197 }
2198 /* Drop the packet */
2199 m_freem(m);
2200 }
2201
Cache object: 2e13bef08b5ab038f962bc935a45ed3a
|