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: releng/5.2/sys/netinet/ip_input.c 122996 2003-11-26 20:31:13Z andre $
35 */
36
37 #include "opt_bootp.h"
38 #include "opt_ipfw.h"
39 #include "opt_ipdn.h"
40 #include "opt_ipdivert.h"
41 #include "opt_ipfilter.h"
42 #include "opt_ipstealth.h"
43 #include "opt_ipsec.h"
44 #include "opt_mac.h"
45 #include "opt_pfil_hooks.h"
46 #include "opt_random_ip_id.h"
47
48 #include <sys/param.h>
49 #include <sys/systm.h>
50 #include <sys/mac.h>
51 #include <sys/mbuf.h>
52 #include <sys/malloc.h>
53 #include <sys/domain.h>
54 #include <sys/protosw.h>
55 #include <sys/socket.h>
56 #include <sys/time.h>
57 #include <sys/kernel.h>
58 #include <sys/syslog.h>
59 #include <sys/sysctl.h>
60
61 #include <net/pfil.h>
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
69 #include <netinet/in.h>
70 #include <netinet/in_systm.h>
71 #include <netinet/in_var.h>
72 #include <netinet/ip.h>
73 #include <netinet/in_pcb.h>
74 #include <netinet/ip_var.h>
75 #include <netinet/ip_icmp.h>
76 #include <machine/in_cksum.h>
77
78 #include <sys/socketvar.h>
79
80 #include <netinet/ip_fw.h>
81 #include <netinet/ip_dummynet.h>
82
83 #ifdef IPSEC
84 #include <netinet6/ipsec.h>
85 #include <netkey/key.h>
86 #endif
87
88 #ifdef FAST_IPSEC
89 #include <netipsec/ipsec.h>
90 #include <netipsec/key.h>
91 #endif
92
93 int rsvp_on = 0;
94
95 int ipforwarding = 0;
96 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW,
97 &ipforwarding, 0, "Enable IP forwarding between interfaces");
98
99 static int ipsendredirects = 1; /* XXX */
100 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW,
101 &ipsendredirects, 0, "Enable sending IP redirects");
102
103 int ip_defttl = IPDEFTTL;
104 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW,
105 &ip_defttl, 0, "Maximum TTL on IP packets");
106
107 static int ip_dosourceroute = 0;
108 SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW,
109 &ip_dosourceroute, 0, "Enable forwarding source routed IP packets");
110
111 static int ip_acceptsourceroute = 0;
112 SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute,
113 CTLFLAG_RW, &ip_acceptsourceroute, 0,
114 "Enable accepting source routed IP packets");
115
116 static int ip_keepfaith = 0;
117 SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW,
118 &ip_keepfaith, 0,
119 "Enable packet capture for FAITH IPv4->IPv6 translater daemon");
120
121 static int nipq = 0; /* total # of reass queues */
122 static int maxnipq;
123 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragpackets, CTLFLAG_RW,
124 &maxnipq, 0,
125 "Maximum number of IPv4 fragment reassembly queue entries");
126
127 static int maxfragsperpacket;
128 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW,
129 &maxfragsperpacket, 0,
130 "Maximum number of IPv4 fragments allowed per packet");
131
132 static int ip_sendsourcequench = 0;
133 SYSCTL_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW,
134 &ip_sendsourcequench, 0,
135 "Enable the transmission of source quench packets");
136
137 /*
138 * XXX - Setting ip_checkinterface mostly implements the receive side of
139 * the Strong ES model described in RFC 1122, but since the routing table
140 * and transmit implementation do not implement the Strong ES model,
141 * setting this to 1 results in an odd hybrid.
142 *
143 * XXX - ip_checkinterface currently must be disabled if you use ipnat
144 * to translate the destination address to another local interface.
145 *
146 * XXX - ip_checkinterface must be disabled if you add IP aliases
147 * to the loopback interface instead of the interface where the
148 * packets for those addresses are received.
149 */
150 static int ip_checkinterface = 1;
151 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW,
152 &ip_checkinterface, 0, "Verify packet arrives on correct interface");
153
154 #ifdef DIAGNOSTIC
155 static int ipprintfs = 0;
156 #endif
157 #ifdef PFIL_HOOKS
158 struct pfil_head inet_pfil_hook;
159 #endif
160
161 static struct ifqueue ipintrq;
162 static int ipqmaxlen = IFQ_MAXLEN;
163
164 extern struct domain inetdomain;
165 extern struct protosw inetsw[];
166 u_char ip_protox[IPPROTO_MAX];
167 struct in_ifaddrhead in_ifaddrhead; /* first inet address */
168 struct in_ifaddrhashhead *in_ifaddrhashtbl; /* inet addr hash table */
169 u_long in_ifaddrhmask; /* mask for hash table */
170
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 TAILQ_HEAD(ipqhead, ipq) ipq[IPREASS_NHASH];
188 struct mtx ipqlock;
189
190 #define IPQ_LOCK() mtx_lock(&ipqlock)
191 #define IPQ_UNLOCK() mtx_unlock(&ipqlock)
192 #define IPQ_LOCK_INIT() mtx_init(&ipqlock, "ipqlock", NULL, MTX_DEF)
193 #define IPQ_LOCK_ASSERT() mtx_assert(&ipqlock, MA_OWNED)
194
195 #ifdef IPCTL_DEFMTU
196 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
197 &ip_mtu, 0, "Default MTU");
198 #endif
199
200 #ifdef IPSTEALTH
201 int ipstealth = 0;
202 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW,
203 &ipstealth, 0, "");
204 #endif
205
206
207 /* Firewall hooks */
208 ip_fw_chk_t *ip_fw_chk_ptr;
209 int fw_enable = 1 ;
210 int fw_one_pass = 1;
211
212 /* Dummynet hooks */
213 ip_dn_io_t *ip_dn_io_ptr;
214
215 /*
216 * XXX this is ugly -- the following two global variables are
217 * used to store packet state while it travels through the stack.
218 * Note that the code even makes assumptions on the size and
219 * alignment of fields inside struct ip_srcrt so e.g. adding some
220 * fields will break the code. This needs to be fixed.
221 *
222 * We need to save the IP options in case a protocol wants to respond
223 * to an incoming packet over the same route if the packet got here
224 * using IP source routing. This allows connection establishment and
225 * maintenance when the remote end is on a network that is not known
226 * to us.
227 */
228 static int ip_nhops = 0;
229 static struct ip_srcrt {
230 struct in_addr dst; /* final destination */
231 char nop; /* one NOP to align */
232 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */
233 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)];
234 } ip_srcrt;
235
236 static void save_rte(u_char *, struct in_addr);
237 static int ip_dooptions(struct mbuf *m, int,
238 struct sockaddr_in *next_hop);
239 static void ip_forward(struct mbuf *m, int srcrt,
240 struct sockaddr_in *next_hop);
241 static void ip_freef(struct ipqhead *, struct ipq *);
242 static struct mbuf *ip_reass(struct mbuf *, struct ipqhead *,
243 struct ipq *, u_int32_t *, u_int16_t *);
244
245 /*
246 * IP initialization: fill in IP protocol switch table.
247 * All protocols not implemented in kernel go to raw IP protocol handler.
248 */
249 void
250 ip_init()
251 {
252 register struct protosw *pr;
253 register int i;
254
255 TAILQ_INIT(&in_ifaddrhead);
256 in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask);
257 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
258 if (pr == 0)
259 panic("ip_init");
260 for (i = 0; i < IPPROTO_MAX; i++)
261 ip_protox[i] = pr - inetsw;
262 for (pr = inetdomain.dom_protosw;
263 pr < inetdomain.dom_protoswNPROTOSW; pr++)
264 if (pr->pr_domain->dom_family == PF_INET &&
265 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
266 ip_protox[pr->pr_protocol] = pr - inetsw;
267
268 #ifdef PFIL_HOOKS
269 inet_pfil_hook.ph_type = PFIL_TYPE_AF;
270 inet_pfil_hook.ph_af = AF_INET;
271 if ((i = pfil_head_register(&inet_pfil_hook)) != 0)
272 printf("%s: WARNING: unable to register pfil hook, "
273 "error %d\n", __func__, i);
274 #endif /* PFIL_HOOKS */
275
276 IPQ_LOCK_INIT();
277 for (i = 0; i < IPREASS_NHASH; i++)
278 TAILQ_INIT(&ipq[i]);
279
280 maxnipq = nmbclusters / 32;
281 maxfragsperpacket = 16;
282
283 #ifndef RANDOM_IP_ID
284 ip_id = time_second & 0xffff;
285 #endif
286 ipintrq.ifq_maxlen = ipqmaxlen;
287 mtx_init(&ipintrq.ifq_mtx, "ip_inq", NULL, MTX_DEF);
288 netisr_register(NETISR_IP, ip_input, &ipintrq, NETISR_MPSAFE);
289 }
290
291 /*
292 * Ip input routine. Checksum and byte swap header. If fragmented
293 * try to reassemble. Process options. Pass to next level.
294 */
295 void
296 ip_input(struct mbuf *m)
297 {
298 struct ip *ip = NULL;
299 struct ipq *fp;
300 struct in_ifaddr *ia = NULL;
301 struct ifaddr *ifa;
302 int i, checkif, hlen = 0;
303 int ours = 0;
304 u_short sum;
305 struct in_addr pkt_dst;
306 u_int32_t divert_info = 0; /* packet divert/tee info */
307 struct ip_fw_args args;
308 int dchg = 0; /* dest changed after fw */
309 #ifdef PFIL_HOOKS
310 struct in_addr odst; /* original dst address */
311 #endif
312 #ifdef FAST_IPSEC
313 struct m_tag *mtag;
314 struct tdb_ident *tdbi;
315 struct secpolicy *sp;
316 int s, error;
317 #endif /* FAST_IPSEC */
318
319 args.eh = NULL;
320 args.oif = NULL;
321 args.rule = NULL;
322 args.divert_rule = 0; /* divert cookie */
323 args.next_hop = NULL;
324
325 /*
326 * Grab info from MT_TAG mbufs prepended to the chain.
327 *
328 * XXX: This is ugly. These pseudo mbuf prepend tags should really
329 * be real m_tags. Before these have always been allocated on the
330 * callers stack, so we didn't have to free them. Now with
331 * ip_fastforward they are true mbufs and we have to free them
332 * otherwise we have a leak. Must rewrite ipfw to use m_tags.
333 */
334 for (; m && m->m_type == MT_TAG;) {
335 struct mbuf *m0;
336
337 switch(m->_m_tag_id) {
338 default:
339 printf("ip_input: unrecognised MT_TAG tag %d\n",
340 m->_m_tag_id);
341 break;
342
343 case PACKET_TAG_DUMMYNET:
344 args.rule = ((struct dn_pkt *)m)->rule;
345 break;
346
347 case PACKET_TAG_DIVERT:
348 args.divert_rule = (intptr_t)m->m_hdr.mh_data & 0xffff;
349 break;
350
351 case PACKET_TAG_IPFORWARD:
352 args.next_hop = (struct sockaddr_in *)m->m_hdr.mh_data;
353 break;
354
355 case PACKET_TAG_IPFASTFWD_OURS:
356 ours = 1;
357 break;
358 }
359
360 m0 = m;
361 m = m->m_next;
362 /* XXX: This is set by ip_fastforward */
363 if (m0->m_nextpkt == (struct mbuf *)1)
364 m_free(m0);
365 }
366
367 M_ASSERTPKTHDR(m);
368
369 if (ours) /* ip_fastforward firewall changed dest to local */
370 goto ours;
371
372 if (args.rule) { /* dummynet already filtered us */
373 ip = mtod(m, struct ip *);
374 hlen = ip->ip_hl << 2;
375 goto iphack ;
376 }
377
378 ipstat.ips_total++;
379
380 if (m->m_pkthdr.len < sizeof(struct ip))
381 goto tooshort;
382
383 if (m->m_len < sizeof (struct ip) &&
384 (m = m_pullup(m, sizeof (struct ip))) == 0) {
385 ipstat.ips_toosmall++;
386 return;
387 }
388 ip = mtod(m, struct ip *);
389
390 if (ip->ip_v != IPVERSION) {
391 ipstat.ips_badvers++;
392 goto bad;
393 }
394
395 hlen = ip->ip_hl << 2;
396 if (hlen < sizeof(struct ip)) { /* minimum header length */
397 ipstat.ips_badhlen++;
398 goto bad;
399 }
400 if (hlen > m->m_len) {
401 if ((m = m_pullup(m, hlen)) == 0) {
402 ipstat.ips_badhlen++;
403 return;
404 }
405 ip = mtod(m, struct ip *);
406 }
407
408 /* 127/8 must not appear on wire - RFC1122 */
409 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
410 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
411 if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
412 ipstat.ips_badaddr++;
413 goto bad;
414 }
415 }
416
417 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
418 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
419 } else {
420 if (hlen == sizeof(struct ip)) {
421 sum = in_cksum_hdr(ip);
422 } else {
423 sum = in_cksum(m, hlen);
424 }
425 }
426 if (sum) {
427 ipstat.ips_badsum++;
428 goto bad;
429 }
430
431 /*
432 * Convert fields to host representation.
433 */
434 ip->ip_len = ntohs(ip->ip_len);
435 if (ip->ip_len < hlen) {
436 ipstat.ips_badlen++;
437 goto bad;
438 }
439 ip->ip_off = ntohs(ip->ip_off);
440
441 /*
442 * Check that the amount of data in the buffers
443 * is as at least much as the IP header would have us expect.
444 * Trim mbufs if longer than we expect.
445 * Drop packet if shorter than we expect.
446 */
447 if (m->m_pkthdr.len < ip->ip_len) {
448 tooshort:
449 ipstat.ips_tooshort++;
450 goto bad;
451 }
452 if (m->m_pkthdr.len > ip->ip_len) {
453 if (m->m_len == m->m_pkthdr.len) {
454 m->m_len = ip->ip_len;
455 m->m_pkthdr.len = ip->ip_len;
456 } else
457 m_adj(m, ip->ip_len - m->m_pkthdr.len);
458 }
459 #if defined(IPSEC) && !defined(IPSEC_FILTERGIF)
460 /*
461 * Bypass packet filtering for packets from a tunnel (gif).
462 */
463 if (ipsec_getnhist(m))
464 goto pass;
465 #endif
466 #if defined(FAST_IPSEC) && !defined(IPSEC_FILTERGIF)
467 /*
468 * Bypass packet filtering for packets from a tunnel (gif).
469 */
470 if (m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL)
471 goto pass;
472 #endif
473
474 /*
475 * IpHack's section.
476 * Right now when no processing on packet has done
477 * and it is still fresh out of network we do our black
478 * deals with it.
479 * - Firewall: deny/allow/divert
480 * - Xlate: translate packet's addr/port (NAT).
481 * - Pipe: pass pkt through dummynet.
482 * - Wrap: fake packet's addr/port <unimpl.>
483 * - Encapsulate: put it in another IP and send out. <unimp.>
484 */
485
486 iphack:
487
488 #ifdef PFIL_HOOKS
489 /*
490 * Run through list of hooks for input packets.
491 *
492 * NB: Beware of the destination address changing (e.g.
493 * by NAT rewriting). When this happens, tell
494 * ip_forward to do the right thing.
495 */
496 odst = ip->ip_dst;
497 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif,
498 PFIL_IN) != 0)
499 return;
500 if (m == NULL) /* consumed by filter */
501 return;
502 ip = mtod(m, struct ip *);
503 dchg = (odst.s_addr != ip->ip_dst.s_addr);
504 #endif /* PFIL_HOOKS */
505
506 if (fw_enable && IPFW_LOADED) {
507 /*
508 * If we've been forwarded from the output side, then
509 * skip the firewall a second time
510 */
511 if (args.next_hop)
512 goto ours;
513
514 args.m = m;
515 i = ip_fw_chk_ptr(&args);
516 m = args.m;
517
518 if ( (i & IP_FW_PORT_DENY_FLAG) || m == NULL) { /* drop */
519 if (m)
520 m_freem(m);
521 return;
522 }
523 ip = mtod(m, struct ip *); /* just in case m changed */
524 if (i == 0 && args.next_hop == NULL) /* common case */
525 goto pass;
526 if (DUMMYNET_LOADED && (i & IP_FW_PORT_DYNT_FLAG) != 0) {
527 /* Send packet to the appropriate pipe */
528 ip_dn_io_ptr(m, i&0xffff, DN_TO_IP_IN, &args);
529 return;
530 }
531 #ifdef IPDIVERT
532 if (i != 0 && (i & IP_FW_PORT_DYNT_FLAG) == 0) {
533 /* Divert or tee packet */
534 divert_info = i;
535 goto ours;
536 }
537 #endif
538 if (i == 0 && args.next_hop != NULL)
539 goto pass;
540 /*
541 * if we get here, the packet must be dropped
542 */
543 m_freem(m);
544 return;
545 }
546 pass:
547
548 /*
549 * Process options and, if not destined for us,
550 * ship it on. ip_dooptions returns 1 when an
551 * error was detected (causing an icmp message
552 * to be sent and the original packet to be freed).
553 */
554 ip_nhops = 0; /* for source routed packets */
555 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0, args.next_hop))
556 return;
557
558 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
559 * matter if it is destined to another node, or whether it is
560 * a multicast one, RSVP wants it! and prevents it from being forwarded
561 * anywhere else. Also checks if the rsvp daemon is running before
562 * grabbing the packet.
563 */
564 if (rsvp_on && ip->ip_p==IPPROTO_RSVP)
565 goto ours;
566
567 /*
568 * Check our list of addresses, to see if the packet is for us.
569 * If we don't have any addresses, assume any unicast packet
570 * we receive might be for us (and let the upper layers deal
571 * with it).
572 */
573 if (TAILQ_EMPTY(&in_ifaddrhead) &&
574 (m->m_flags & (M_MCAST|M_BCAST)) == 0)
575 goto ours;
576
577 /*
578 * Cache the destination address of the packet; this may be
579 * changed by use of 'ipfw fwd'.
580 */
581 pkt_dst = args.next_hop ? args.next_hop->sin_addr : ip->ip_dst;
582
583 /*
584 * Enable a consistency check between the destination address
585 * and the arrival interface for a unicast packet (the RFC 1122
586 * strong ES model) if IP forwarding is disabled and the packet
587 * is not locally generated and the packet is not subject to
588 * 'ipfw fwd'.
589 *
590 * XXX - Checking also should be disabled if the destination
591 * address is ipnat'ed to a different interface.
592 *
593 * XXX - Checking is incompatible with IP aliases added
594 * to the loopback interface instead of the interface where
595 * the packets are received.
596 */
597 checkif = ip_checkinterface && (ipforwarding == 0) &&
598 m->m_pkthdr.rcvif != NULL &&
599 ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) &&
600 (args.next_hop == NULL);
601
602 /*
603 * Check for exact addresses in the hash bucket.
604 */
605 LIST_FOREACH(ia, INADDR_HASH(pkt_dst.s_addr), ia_hash) {
606 /*
607 * If the address matches, verify that the packet
608 * arrived via the correct interface if checking is
609 * enabled.
610 */
611 if (IA_SIN(ia)->sin_addr.s_addr == pkt_dst.s_addr &&
612 (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif))
613 goto ours;
614 }
615 /*
616 * Check for broadcast addresses.
617 *
618 * Only accept broadcast packets that arrive via the matching
619 * interface. Reception of forwarded directed broadcasts would
620 * be handled via ip_forward() and ether_output() with the loopback
621 * into the stack for SIMPLEX interfaces handled by ether_output().
622 */
623 if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
624 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
625 if (ifa->ifa_addr->sa_family != AF_INET)
626 continue;
627 ia = ifatoia(ifa);
628 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
629 pkt_dst.s_addr)
630 goto ours;
631 if (ia->ia_netbroadcast.s_addr == pkt_dst.s_addr)
632 goto ours;
633 #ifdef BOOTP_COMPAT
634 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY)
635 goto ours;
636 #endif
637 }
638 }
639 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
640 struct in_multi *inm;
641 if (ip_mrouter) {
642 /*
643 * If we are acting as a multicast router, all
644 * incoming multicast packets are passed to the
645 * kernel-level multicast forwarding function.
646 * The packet is returned (relatively) intact; if
647 * ip_mforward() returns a non-zero value, the packet
648 * must be discarded, else it may be accepted below.
649 */
650 if (ip_mforward &&
651 ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) {
652 ipstat.ips_cantforward++;
653 m_freem(m);
654 return;
655 }
656
657 /*
658 * The process-level routing daemon needs to receive
659 * all multicast IGMP packets, whether or not this
660 * host belongs to their destination groups.
661 */
662 if (ip->ip_p == IPPROTO_IGMP)
663 goto ours;
664 ipstat.ips_forward++;
665 }
666 /*
667 * See if we belong to the destination multicast group on the
668 * arrival interface.
669 */
670 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
671 if (inm == NULL) {
672 ipstat.ips_notmember++;
673 m_freem(m);
674 return;
675 }
676 goto ours;
677 }
678 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
679 goto ours;
680 if (ip->ip_dst.s_addr == INADDR_ANY)
681 goto ours;
682
683 /*
684 * FAITH(Firewall Aided Internet Translator)
685 */
686 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
687 if (ip_keepfaith) {
688 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP)
689 goto ours;
690 }
691 m_freem(m);
692 return;
693 }
694
695 /*
696 * Not for us; forward if possible and desirable.
697 */
698 if (ipforwarding == 0) {
699 ipstat.ips_cantforward++;
700 m_freem(m);
701 } else {
702 #ifdef IPSEC
703 /*
704 * Enforce inbound IPsec SPD.
705 */
706 if (ipsec4_in_reject(m, NULL)) {
707 ipsecstat.in_polvio++;
708 goto bad;
709 }
710 #endif /* IPSEC */
711 #ifdef FAST_IPSEC
712 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
713 s = splnet();
714 if (mtag != NULL) {
715 tdbi = (struct tdb_ident *)(mtag + 1);
716 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
717 } else {
718 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
719 IP_FORWARDING, &error);
720 }
721 if (sp == NULL) { /* NB: can happen if error */
722 splx(s);
723 /*XXX error stat???*/
724 DPRINTF(("ip_input: no SP for forwarding\n")); /*XXX*/
725 goto bad;
726 }
727
728 /*
729 * Check security policy against packet attributes.
730 */
731 error = ipsec_in_reject(sp, m);
732 KEY_FREESP(&sp);
733 splx(s);
734 if (error) {
735 ipstat.ips_cantforward++;
736 goto bad;
737 }
738 #endif /* FAST_IPSEC */
739 ip_forward(m, dchg, args.next_hop);
740 }
741 return;
742
743 ours:
744 #ifdef IPSTEALTH
745 /*
746 * IPSTEALTH: Process non-routing options only
747 * if the packet is destined for us.
748 */
749 if (ipstealth && hlen > sizeof (struct ip) &&
750 ip_dooptions(m, 1, args.next_hop))
751 return;
752 #endif /* IPSTEALTH */
753
754 /* Count the packet in the ip address stats */
755 if (ia != NULL) {
756 ia->ia_ifa.if_ipackets++;
757 ia->ia_ifa.if_ibytes += m->m_pkthdr.len;
758 }
759
760 /*
761 * If offset or IP_MF are set, must reassemble.
762 * Otherwise, nothing need be done.
763 * (We could look in the reassembly queue to see
764 * if the packet was previously fragmented,
765 * but it's not worth the time; just let them time out.)
766 */
767 if (ip->ip_off & (IP_MF | IP_OFFMASK)) {
768
769 /* If maxnipq is 0, never accept fragments. */
770 if (maxnipq == 0) {
771 ipstat.ips_fragments++;
772 ipstat.ips_fragdropped++;
773 goto bad;
774 }
775
776 sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
777 IPQ_LOCK();
778 /*
779 * Look for queue of fragments
780 * of this datagram.
781 */
782 TAILQ_FOREACH(fp, &ipq[sum], ipq_list)
783 if (ip->ip_id == fp->ipq_id &&
784 ip->ip_src.s_addr == fp->ipq_src.s_addr &&
785 ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
786 #ifdef MAC
787 mac_fragment_match(m, fp) &&
788 #endif
789 ip->ip_p == fp->ipq_p)
790 goto found;
791
792 fp = NULL;
793
794 /*
795 * Enforce upper bound on number of fragmented packets
796 * for which we attempt reassembly;
797 * If maxnipq is -1, accept all fragments without limitation.
798 */
799 if ((nipq > maxnipq) && (maxnipq > 0)) {
800 /*
801 * drop something from the tail of the current queue
802 * before proceeding further
803 */
804 struct ipq *q = TAILQ_LAST(&ipq[sum], ipqhead);
805 if (q == NULL) { /* gak */
806 for (i = 0; i < IPREASS_NHASH; i++) {
807 struct ipq *r = TAILQ_LAST(&ipq[i], ipqhead);
808 if (r) {
809 ipstat.ips_fragtimeout += r->ipq_nfrags;
810 ip_freef(&ipq[i], r);
811 break;
812 }
813 }
814 } else {
815 ipstat.ips_fragtimeout += q->ipq_nfrags;
816 ip_freef(&ipq[sum], q);
817 }
818 }
819 found:
820 /*
821 * Adjust ip_len to not reflect header,
822 * convert offset of this to bytes.
823 */
824 ip->ip_len -= hlen;
825 if (ip->ip_off & IP_MF) {
826 /*
827 * Make sure that fragments have a data length
828 * that's a non-zero multiple of 8 bytes.
829 */
830 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
831 IPQ_UNLOCK();
832 ipstat.ips_toosmall++; /* XXX */
833 goto bad;
834 }
835 m->m_flags |= M_FRAG;
836 } else
837 m->m_flags &= ~M_FRAG;
838 ip->ip_off <<= 3;
839
840 /*
841 * Attempt reassembly; if it succeeds, proceed.
842 * ip_reass() will return a different mbuf, and update
843 * the divert info in divert_info and args.divert_rule.
844 */
845 ipstat.ips_fragments++;
846 m->m_pkthdr.header = ip;
847 m = ip_reass(m,
848 &ipq[sum], fp, &divert_info, &args.divert_rule);
849 IPQ_UNLOCK();
850 if (m == 0)
851 return;
852 ipstat.ips_reassembled++;
853 ip = mtod(m, struct ip *);
854 /* Get the header length of the reassembled packet */
855 hlen = ip->ip_hl << 2;
856 #ifdef IPDIVERT
857 /* Restore original checksum before diverting packet */
858 if (divert_info != 0) {
859 ip->ip_len += hlen;
860 ip->ip_len = htons(ip->ip_len);
861 ip->ip_off = htons(ip->ip_off);
862 ip->ip_sum = 0;
863 if (hlen == sizeof(struct ip))
864 ip->ip_sum = in_cksum_hdr(ip);
865 else
866 ip->ip_sum = in_cksum(m, hlen);
867 ip->ip_off = ntohs(ip->ip_off);
868 ip->ip_len = ntohs(ip->ip_len);
869 ip->ip_len -= hlen;
870 }
871 #endif
872 } else
873 ip->ip_len -= hlen;
874
875 #ifdef IPDIVERT
876 /*
877 * Divert or tee packet to the divert protocol if required.
878 */
879 if (divert_info != 0) {
880 struct mbuf *clone = NULL;
881
882 /* Clone packet if we're doing a 'tee' */
883 if ((divert_info & IP_FW_PORT_TEE_FLAG) != 0)
884 clone = m_dup(m, M_DONTWAIT);
885
886 /* Restore packet header fields to original values */
887 ip->ip_len += hlen;
888 ip->ip_len = htons(ip->ip_len);
889 ip->ip_off = htons(ip->ip_off);
890
891 /* Deliver packet to divert input routine */
892 divert_packet(m, 1, divert_info & 0xffff, args.divert_rule);
893 ipstat.ips_delivered++;
894
895 /* If 'tee', continue with original packet */
896 if (clone == NULL)
897 return;
898 m = clone;
899 ip = mtod(m, struct ip *);
900 ip->ip_len += hlen;
901 /*
902 * Jump backwards to complete processing of the
903 * packet. But first clear divert_info to avoid
904 * entering this block again.
905 * We do not need to clear args.divert_rule
906 * or args.next_hop as they will not be used.
907 */
908 divert_info = 0;
909 goto pass;
910 }
911 #endif
912
913 #ifdef IPSEC
914 /*
915 * enforce IPsec policy checking if we are seeing last header.
916 * note that we do not visit this with protocols with pcb layer
917 * code - like udp/tcp/raw ip.
918 */
919 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0 &&
920 ipsec4_in_reject(m, NULL)) {
921 ipsecstat.in_polvio++;
922 goto bad;
923 }
924 #endif
925 #if FAST_IPSEC
926 /*
927 * enforce IPsec policy checking if we are seeing last header.
928 * note that we do not visit this with protocols with pcb layer
929 * code - like udp/tcp/raw ip.
930 */
931 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) {
932 /*
933 * Check if the packet has already had IPsec processing
934 * done. If so, then just pass it along. This tag gets
935 * set during AH, ESP, etc. input handling, before the
936 * packet is returned to the ip input queue for delivery.
937 */
938 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
939 s = splnet();
940 if (mtag != NULL) {
941 tdbi = (struct tdb_ident *)(mtag + 1);
942 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
943 } else {
944 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
945 IP_FORWARDING, &error);
946 }
947 if (sp != NULL) {
948 /*
949 * Check security policy against packet attributes.
950 */
951 error = ipsec_in_reject(sp, m);
952 KEY_FREESP(&sp);
953 } else {
954 /* XXX error stat??? */
955 error = EINVAL;
956 DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/
957 goto bad;
958 }
959 splx(s);
960 if (error)
961 goto bad;
962 }
963 #endif /* FAST_IPSEC */
964
965 /*
966 * Switch out to protocol's input routine.
967 */
968 ipstat.ips_delivered++;
969 NET_PICKUP_GIANT();
970 if (args.next_hop && ip->ip_p == IPPROTO_TCP) {
971 /* TCP needs IPFORWARD info if available */
972 struct m_hdr tag;
973
974 tag.mh_type = MT_TAG;
975 tag.mh_flags = PACKET_TAG_IPFORWARD;
976 tag.mh_data = (caddr_t)args.next_hop;
977 tag.mh_next = m;
978 tag.mh_nextpkt = NULL;
979
980 (*inetsw[ip_protox[ip->ip_p]].pr_input)(
981 (struct mbuf *)&tag, hlen);
982 } else
983 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen);
984 NET_DROP_GIANT();
985 return;
986 bad:
987 m_freem(m);
988 }
989
990 /*
991 * Take incoming datagram fragment and try to reassemble it into
992 * whole datagram. If a chain for reassembly of this datagram already
993 * exists, then it is given as fp; otherwise have to make a chain.
994 *
995 * When IPDIVERT enabled, keep additional state with each packet that
996 * tells us if we need to divert or tee the packet we're building.
997 * In particular, *divinfo includes the port and TEE flag,
998 * *divert_rule is the number of the matching rule.
999 */
1000
1001 static struct mbuf *
1002 ip_reass(struct mbuf *m, struct ipqhead *head, struct ipq *fp,
1003 u_int32_t *divinfo, u_int16_t *divert_rule)
1004 {
1005 struct ip *ip = mtod(m, struct ip *);
1006 register struct mbuf *p, *q, *nq;
1007 struct mbuf *t;
1008 int hlen = ip->ip_hl << 2;
1009 int i, next;
1010 u_int8_t ecn, ecn0;
1011
1012 IPQ_LOCK_ASSERT();
1013
1014 /*
1015 * Presence of header sizes in mbufs
1016 * would confuse code below.
1017 */
1018 m->m_data += hlen;
1019 m->m_len -= hlen;
1020
1021 /*
1022 * If first fragment to arrive, create a reassembly queue.
1023 */
1024 if (fp == NULL) {
1025 if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL)
1026 goto dropfrag;
1027 fp = mtod(t, struct ipq *);
1028 #ifdef MAC
1029 if (mac_init_ipq(fp, M_NOWAIT) != 0) {
1030 m_free(t);
1031 goto dropfrag;
1032 }
1033 mac_create_ipq(m, fp);
1034 #endif
1035 TAILQ_INSERT_HEAD(head, fp, ipq_list);
1036 nipq++;
1037 fp->ipq_nfrags = 1;
1038 fp->ipq_ttl = IPFRAGTTL;
1039 fp->ipq_p = ip->ip_p;
1040 fp->ipq_id = ip->ip_id;
1041 fp->ipq_src = ip->ip_src;
1042 fp->ipq_dst = ip->ip_dst;
1043 fp->ipq_frags = m;
1044 m->m_nextpkt = NULL;
1045 #ifdef IPDIVERT
1046 fp->ipq_div_info = 0;
1047 fp->ipq_div_cookie = 0;
1048 #endif
1049 goto inserted;
1050 } else {
1051 fp->ipq_nfrags++;
1052 #ifdef MAC
1053 mac_update_ipq(m, fp);
1054 #endif
1055 }
1056
1057 #define GETIP(m) ((struct ip*)((m)->m_pkthdr.header))
1058
1059 /*
1060 * Handle ECN by comparing this segment with the first one;
1061 * if CE is set, do not lose CE.
1062 * drop if CE and not-ECT are mixed for the same packet.
1063 */
1064 ecn = ip->ip_tos & IPTOS_ECN_MASK;
1065 ecn0 = GETIP(fp->ipq_frags)->ip_tos & IPTOS_ECN_MASK;
1066 if (ecn == IPTOS_ECN_CE) {
1067 if (ecn0 == IPTOS_ECN_NOTECT)
1068 goto dropfrag;
1069 if (ecn0 != IPTOS_ECN_CE)
1070 GETIP(fp->ipq_frags)->ip_tos |= IPTOS_ECN_CE;
1071 }
1072 if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT)
1073 goto dropfrag;
1074
1075 /*
1076 * Find a segment which begins after this one does.
1077 */
1078 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt)
1079 if (GETIP(q)->ip_off > ip->ip_off)
1080 break;
1081
1082 /*
1083 * If there is a preceding segment, it may provide some of
1084 * our data already. If so, drop the data from the incoming
1085 * segment. If it provides all of our data, drop us, otherwise
1086 * stick new segment in the proper place.
1087 *
1088 * If some of the data is dropped from the the preceding
1089 * segment, then it's checksum is invalidated.
1090 */
1091 if (p) {
1092 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off;
1093 if (i > 0) {
1094 if (i >= ip->ip_len)
1095 goto dropfrag;
1096 m_adj(m, i);
1097 m->m_pkthdr.csum_flags = 0;
1098 ip->ip_off += i;
1099 ip->ip_len -= i;
1100 }
1101 m->m_nextpkt = p->m_nextpkt;
1102 p->m_nextpkt = m;
1103 } else {
1104 m->m_nextpkt = fp->ipq_frags;
1105 fp->ipq_frags = m;
1106 }
1107
1108 /*
1109 * While we overlap succeeding segments trim them or,
1110 * if they are completely covered, dequeue them.
1111 */
1112 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off;
1113 q = nq) {
1114 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off;
1115 if (i < GETIP(q)->ip_len) {
1116 GETIP(q)->ip_len -= i;
1117 GETIP(q)->ip_off += i;
1118 m_adj(q, i);
1119 q->m_pkthdr.csum_flags = 0;
1120 break;
1121 }
1122 nq = q->m_nextpkt;
1123 m->m_nextpkt = nq;
1124 ipstat.ips_fragdropped++;
1125 fp->ipq_nfrags--;
1126 m_freem(q);
1127 }
1128
1129 inserted:
1130
1131 #ifdef IPDIVERT
1132 /*
1133 * Transfer firewall instructions to the fragment structure.
1134 * Only trust info in the fragment at offset 0.
1135 */
1136 if (ip->ip_off == 0) {
1137 fp->ipq_div_info = *divinfo;
1138 fp->ipq_div_cookie = *divert_rule;
1139 }
1140 *divinfo = 0;
1141 *divert_rule = 0;
1142 #endif
1143
1144 /*
1145 * Check for complete reassembly and perform frag per packet
1146 * limiting.
1147 *
1148 * Frag limiting is performed here so that the nth frag has
1149 * a chance to complete the packet before we drop the packet.
1150 * As a result, n+1 frags are actually allowed per packet, but
1151 * only n will ever be stored. (n = maxfragsperpacket.)
1152 *
1153 */
1154 next = 0;
1155 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
1156 if (GETIP(q)->ip_off != next) {
1157 if (fp->ipq_nfrags > maxfragsperpacket) {
1158 ipstat.ips_fragdropped += fp->ipq_nfrags;
1159 ip_freef(head, fp);
1160 }
1161 return (0);
1162 }
1163 next += GETIP(q)->ip_len;
1164 }
1165 /* Make sure the last packet didn't have the IP_MF flag */
1166 if (p->m_flags & M_FRAG) {
1167 if (fp->ipq_nfrags > maxfragsperpacket) {
1168 ipstat.ips_fragdropped += fp->ipq_nfrags;
1169 ip_freef(head, fp);
1170 }
1171 return (0);
1172 }
1173
1174 /*
1175 * Reassembly is complete. Make sure the packet is a sane size.
1176 */
1177 q = fp->ipq_frags;
1178 ip = GETIP(q);
1179 if (next + (ip->ip_hl << 2) > IP_MAXPACKET) {
1180 ipstat.ips_toolong++;
1181 ipstat.ips_fragdropped += fp->ipq_nfrags;
1182 ip_freef(head, fp);
1183 return (0);
1184 }
1185
1186 /*
1187 * Concatenate fragments.
1188 */
1189 m = q;
1190 t = m->m_next;
1191 m->m_next = 0;
1192 m_cat(m, t);
1193 nq = q->m_nextpkt;
1194 q->m_nextpkt = 0;
1195 for (q = nq; q != NULL; q = nq) {
1196 nq = q->m_nextpkt;
1197 q->m_nextpkt = NULL;
1198 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
1199 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
1200 m_cat(m, q);
1201 }
1202 #ifdef MAC
1203 mac_create_datagram_from_ipq(fp, m);
1204 mac_destroy_ipq(fp);
1205 #endif
1206
1207 #ifdef IPDIVERT
1208 /*
1209 * Extract firewall instructions from the fragment structure.
1210 */
1211 *divinfo = fp->ipq_div_info;
1212 *divert_rule = fp->ipq_div_cookie;
1213 #endif
1214
1215 /*
1216 * Create header for new ip packet by
1217 * modifying header of first packet;
1218 * dequeue and discard fragment reassembly header.
1219 * Make header visible.
1220 */
1221 ip->ip_len = next;
1222 ip->ip_src = fp->ipq_src;
1223 ip->ip_dst = fp->ipq_dst;
1224 TAILQ_REMOVE(head, fp, ipq_list);
1225 nipq--;
1226 (void) m_free(dtom(fp));
1227 m->m_len += (ip->ip_hl << 2);
1228 m->m_data -= (ip->ip_hl << 2);
1229 /* some debugging cruft by sklower, below, will go away soon */
1230 if (m->m_flags & M_PKTHDR) /* XXX this should be done elsewhere */
1231 m_fixhdr(m);
1232 return (m);
1233
1234 dropfrag:
1235 #ifdef IPDIVERT
1236 *divinfo = 0;
1237 *divert_rule = 0;
1238 #endif
1239 ipstat.ips_fragdropped++;
1240 if (fp != NULL)
1241 fp->ipq_nfrags--;
1242 m_freem(m);
1243 return (0);
1244
1245 #undef GETIP
1246 }
1247
1248 /*
1249 * Free a fragment reassembly header and all
1250 * associated datagrams.
1251 */
1252 static void
1253 ip_freef(fhp, fp)
1254 struct ipqhead *fhp;
1255 struct ipq *fp;
1256 {
1257 register struct mbuf *q;
1258
1259 IPQ_LOCK_ASSERT();
1260
1261 while (fp->ipq_frags) {
1262 q = fp->ipq_frags;
1263 fp->ipq_frags = q->m_nextpkt;
1264 m_freem(q);
1265 }
1266 TAILQ_REMOVE(fhp, fp, ipq_list);
1267 (void) m_free(dtom(fp));
1268 nipq--;
1269 }
1270
1271 /*
1272 * IP timer processing;
1273 * if a timer expires on a reassembly
1274 * queue, discard it.
1275 */
1276 void
1277 ip_slowtimo()
1278 {
1279 register struct ipq *fp;
1280 int s = splnet();
1281 int i;
1282
1283 IPQ_LOCK();
1284 for (i = 0; i < IPREASS_NHASH; i++) {
1285 for(fp = TAILQ_FIRST(&ipq[i]); fp;) {
1286 struct ipq *fpp;
1287
1288 fpp = fp;
1289 fp = TAILQ_NEXT(fp, ipq_list);
1290 if(--fpp->ipq_ttl == 0) {
1291 ipstat.ips_fragtimeout += fpp->ipq_nfrags;
1292 ip_freef(&ipq[i], fpp);
1293 }
1294 }
1295 }
1296 /*
1297 * If we are over the maximum number of fragments
1298 * (due to the limit being lowered), drain off
1299 * enough to get down to the new limit.
1300 */
1301 if (maxnipq >= 0 && nipq > maxnipq) {
1302 for (i = 0; i < IPREASS_NHASH; i++) {
1303 while (nipq > maxnipq && !TAILQ_EMPTY(&ipq[i])) {
1304 ipstat.ips_fragdropped +=
1305 TAILQ_FIRST(&ipq[i])->ipq_nfrags;
1306 ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i]));
1307 }
1308 }
1309 }
1310 IPQ_UNLOCK();
1311 splx(s);
1312 }
1313
1314 /*
1315 * Drain off all datagram fragments.
1316 */
1317 void
1318 ip_drain()
1319 {
1320 int i;
1321
1322 IPQ_LOCK();
1323 for (i = 0; i < IPREASS_NHASH; i++) {
1324 while(!TAILQ_EMPTY(&ipq[i])) {
1325 ipstat.ips_fragdropped +=
1326 TAILQ_FIRST(&ipq[i])->ipq_nfrags;
1327 ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i]));
1328 }
1329 }
1330 IPQ_UNLOCK();
1331 in_rtqdrain();
1332 }
1333
1334 /*
1335 * Do option processing on a datagram,
1336 * possibly discarding it if bad options are encountered,
1337 * or forwarding it if source-routed.
1338 * The pass argument is used when operating in the IPSTEALTH
1339 * mode to tell what options to process:
1340 * [LS]SRR (pass 0) or the others (pass 1).
1341 * The reason for as many as two passes is that when doing IPSTEALTH,
1342 * non-routing options should be processed only if the packet is for us.
1343 * Returns 1 if packet has been forwarded/freed,
1344 * 0 if the packet should be processed further.
1345 */
1346 static int
1347 ip_dooptions(struct mbuf *m, int pass, struct sockaddr_in *next_hop)
1348 {
1349 struct ip *ip = mtod(m, struct ip *);
1350 u_char *cp;
1351 struct in_ifaddr *ia;
1352 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0;
1353 struct in_addr *sin, dst;
1354 n_time ntime;
1355 struct sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET };
1356
1357 dst = ip->ip_dst;
1358 cp = (u_char *)(ip + 1);
1359 cnt = (ip->ip_hl << 2) - sizeof (struct ip);
1360 for (; cnt > 0; cnt -= optlen, cp += optlen) {
1361 opt = cp[IPOPT_OPTVAL];
1362 if (opt == IPOPT_EOL)
1363 break;
1364 if (opt == IPOPT_NOP)
1365 optlen = 1;
1366 else {
1367 if (cnt < IPOPT_OLEN + sizeof(*cp)) {
1368 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1369 goto bad;
1370 }
1371 optlen = cp[IPOPT_OLEN];
1372 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
1373 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1374 goto bad;
1375 }
1376 }
1377 switch (opt) {
1378
1379 default:
1380 break;
1381
1382 /*
1383 * Source routing with record.
1384 * Find interface with current destination address.
1385 * If none on this machine then drop if strictly routed,
1386 * or do nothing if loosely routed.
1387 * Record interface address and bring up next address
1388 * component. If strictly routed make sure next
1389 * address is on directly accessible net.
1390 */
1391 case IPOPT_LSRR:
1392 case IPOPT_SSRR:
1393 #ifdef IPSTEALTH
1394 if (ipstealth && pass > 0)
1395 break;
1396 #endif
1397 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1398 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1399 goto bad;
1400 }
1401 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1402 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1403 goto bad;
1404 }
1405 ipaddr.sin_addr = ip->ip_dst;
1406 ia = (struct in_ifaddr *)
1407 ifa_ifwithaddr((struct sockaddr *)&ipaddr);
1408 if (ia == 0) {
1409 if (opt == IPOPT_SSRR) {
1410 type = ICMP_UNREACH;
1411 code = ICMP_UNREACH_SRCFAIL;
1412 goto bad;
1413 }
1414 if (!ip_dosourceroute)
1415 goto nosourcerouting;
1416 /*
1417 * Loose routing, and not at next destination
1418 * yet; nothing to do except forward.
1419 */
1420 break;
1421 }
1422 off--; /* 0 origin */
1423 if (off > optlen - (int)sizeof(struct in_addr)) {
1424 /*
1425 * End of source route. Should be for us.
1426 */
1427 if (!ip_acceptsourceroute)
1428 goto nosourcerouting;
1429 save_rte(cp, ip->ip_src);
1430 break;
1431 }
1432 #ifdef IPSTEALTH
1433 if (ipstealth)
1434 goto dropit;
1435 #endif
1436 if (!ip_dosourceroute) {
1437 if (ipforwarding) {
1438 char buf[16]; /* aaa.bbb.ccc.ddd\0 */
1439 /*
1440 * Acting as a router, so generate ICMP
1441 */
1442 nosourcerouting:
1443 strcpy(buf, inet_ntoa(ip->ip_dst));
1444 log(LOG_WARNING,
1445 "attempted source route from %s to %s\n",
1446 inet_ntoa(ip->ip_src), buf);
1447 type = ICMP_UNREACH;
1448 code = ICMP_UNREACH_SRCFAIL;
1449 goto bad;
1450 } else {
1451 /*
1452 * Not acting as a router, so silently drop.
1453 */
1454 #ifdef IPSTEALTH
1455 dropit:
1456 #endif
1457 ipstat.ips_cantforward++;
1458 m_freem(m);
1459 return (1);
1460 }
1461 }
1462
1463 /*
1464 * locate outgoing interface
1465 */
1466 (void)memcpy(&ipaddr.sin_addr, cp + off,
1467 sizeof(ipaddr.sin_addr));
1468
1469 if (opt == IPOPT_SSRR) {
1470 #define INA struct in_ifaddr *
1471 #define SA struct sockaddr *
1472 if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == 0)
1473 ia = (INA)ifa_ifwithnet((SA)&ipaddr);
1474 } else
1475 ia = ip_rtaddr(ipaddr.sin_addr);
1476 if (ia == 0) {
1477 type = ICMP_UNREACH;
1478 code = ICMP_UNREACH_SRCFAIL;
1479 goto bad;
1480 }
1481 ip->ip_dst = ipaddr.sin_addr;
1482 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
1483 sizeof(struct in_addr));
1484 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1485 /*
1486 * Let ip_intr's mcast routing check handle mcast pkts
1487 */
1488 forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr));
1489 break;
1490
1491 case IPOPT_RR:
1492 #ifdef IPSTEALTH
1493 if (ipstealth && pass == 0)
1494 break;
1495 #endif
1496 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1497 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1498 goto bad;
1499 }
1500 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1501 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1502 goto bad;
1503 }
1504 /*
1505 * If no space remains, ignore.
1506 */
1507 off--; /* 0 origin */
1508 if (off > optlen - (int)sizeof(struct in_addr))
1509 break;
1510 (void)memcpy(&ipaddr.sin_addr, &ip->ip_dst,
1511 sizeof(ipaddr.sin_addr));
1512 /*
1513 * locate outgoing interface; if we're the destination,
1514 * use the incoming interface (should be same).
1515 */
1516 if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == 0 &&
1517 (ia = ip_rtaddr(ipaddr.sin_addr)) == 0) {
1518 type = ICMP_UNREACH;
1519 code = ICMP_UNREACH_HOST;
1520 goto bad;
1521 }
1522 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
1523 sizeof(struct in_addr));
1524 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1525 break;
1526
1527 case IPOPT_TS:
1528 #ifdef IPSTEALTH
1529 if (ipstealth && pass == 0)
1530 break;
1531 #endif
1532 code = cp - (u_char *)ip;
1533 if (optlen < 4 || optlen > 40) {
1534 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1535 goto bad;
1536 }
1537 if ((off = cp[IPOPT_OFFSET]) < 5) {
1538 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1539 goto bad;
1540 }
1541 if (off > optlen - (int)sizeof(int32_t)) {
1542 cp[IPOPT_OFFSET + 1] += (1 << 4);
1543 if ((cp[IPOPT_OFFSET + 1] & 0xf0) == 0) {
1544 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1545 goto bad;
1546 }
1547 break;
1548 }
1549 off--; /* 0 origin */
1550 sin = (struct in_addr *)(cp + off);
1551 switch (cp[IPOPT_OFFSET + 1] & 0x0f) {
1552
1553 case IPOPT_TS_TSONLY:
1554 break;
1555
1556 case IPOPT_TS_TSANDADDR:
1557 if (off + sizeof(n_time) +
1558 sizeof(struct in_addr) > optlen) {
1559 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1560 goto bad;
1561 }
1562 ipaddr.sin_addr = dst;
1563 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr,
1564 m->m_pkthdr.rcvif);
1565 if (ia == 0)
1566 continue;
1567 (void)memcpy(sin, &IA_SIN(ia)->sin_addr,
1568 sizeof(struct in_addr));
1569 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1570 off += sizeof(struct in_addr);
1571 break;
1572
1573 case IPOPT_TS_PRESPEC:
1574 if (off + sizeof(n_time) +
1575 sizeof(struct in_addr) > optlen) {
1576 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1577 goto bad;
1578 }
1579 (void)memcpy(&ipaddr.sin_addr, sin,
1580 sizeof(struct in_addr));
1581 if (ifa_ifwithaddr((SA)&ipaddr) == 0)
1582 continue;
1583 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1584 off += sizeof(struct in_addr);
1585 break;
1586
1587 default:
1588 code = &cp[IPOPT_OFFSET + 1] - (u_char *)ip;
1589 goto bad;
1590 }
1591 ntime = iptime();
1592 (void)memcpy(cp + off, &ntime, sizeof(n_time));
1593 cp[IPOPT_OFFSET] += sizeof(n_time);
1594 }
1595 }
1596 if (forward && ipforwarding) {
1597 ip_forward(m, 1, next_hop);
1598 return (1);
1599 }
1600 return (0);
1601 bad:
1602 icmp_error(m, type, code, 0, 0);
1603 ipstat.ips_badoptions++;
1604 return (1);
1605 }
1606
1607 /*
1608 * Given address of next destination (final or next hop),
1609 * return internet address info of interface to be used to get there.
1610 */
1611 struct in_ifaddr *
1612 ip_rtaddr(dst)
1613 struct in_addr dst;
1614 {
1615 struct route sro;
1616 struct sockaddr_in *sin;
1617 struct in_ifaddr *ifa;
1618
1619 bzero(&sro, sizeof(sro));
1620 sin = (struct sockaddr_in *)&sro.ro_dst;
1621 sin->sin_family = AF_INET;
1622 sin->sin_len = sizeof(*sin);
1623 sin->sin_addr = dst;
1624 rtalloc_ign(&sro, RTF_CLONING);
1625
1626 if (sro.ro_rt == NULL)
1627 return ((struct in_ifaddr *)0);
1628
1629 ifa = ifatoia(sro.ro_rt->rt_ifa);
1630 RTFREE(sro.ro_rt);
1631 return ifa;
1632 }
1633
1634 /*
1635 * Save incoming source route for use in replies,
1636 * to be picked up later by ip_srcroute if the receiver is interested.
1637 */
1638 static void
1639 save_rte(option, dst)
1640 u_char *option;
1641 struct in_addr dst;
1642 {
1643 unsigned olen;
1644
1645 olen = option[IPOPT_OLEN];
1646 #ifdef DIAGNOSTIC
1647 if (ipprintfs)
1648 printf("save_rte: olen %d\n", olen);
1649 #endif
1650 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst)))
1651 return;
1652 bcopy(option, ip_srcrt.srcopt, olen);
1653 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1654 ip_srcrt.dst = dst;
1655 }
1656
1657 /*
1658 * Retrieve incoming source route for use in replies,
1659 * in the same form used by setsockopt.
1660 * The first hop is placed before the options, will be removed later.
1661 */
1662 struct mbuf *
1663 ip_srcroute()
1664 {
1665 register struct in_addr *p, *q;
1666 register struct mbuf *m;
1667
1668 if (ip_nhops == 0)
1669 return ((struct mbuf *)0);
1670 m = m_get(M_DONTWAIT, MT_HEADER);
1671 if (m == 0)
1672 return ((struct mbuf *)0);
1673
1674 #define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt))
1675
1676 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
1677 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) +
1678 OPTSIZ;
1679 #ifdef DIAGNOSTIC
1680 if (ipprintfs)
1681 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len);
1682 #endif
1683
1684 /*
1685 * First save first hop for return route
1686 */
1687 p = &ip_srcrt.route[ip_nhops - 1];
1688 *(mtod(m, struct in_addr *)) = *p--;
1689 #ifdef DIAGNOSTIC
1690 if (ipprintfs)
1691 printf(" hops %lx", (u_long)ntohl(mtod(m, struct in_addr *)->s_addr));
1692 #endif
1693
1694 /*
1695 * Copy option fields and padding (nop) to mbuf.
1696 */
1697 ip_srcrt.nop = IPOPT_NOP;
1698 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
1699 (void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr),
1700 &ip_srcrt.nop, OPTSIZ);
1701 q = (struct in_addr *)(mtod(m, caddr_t) +
1702 sizeof(struct in_addr) + OPTSIZ);
1703 #undef OPTSIZ
1704 /*
1705 * Record return path as an IP source route,
1706 * reversing the path (pointers are now aligned).
1707 */
1708 while (p >= ip_srcrt.route) {
1709 #ifdef DIAGNOSTIC
1710 if (ipprintfs)
1711 printf(" %lx", (u_long)ntohl(q->s_addr));
1712 #endif
1713 *q++ = *p--;
1714 }
1715 /*
1716 * Last hop goes to final destination.
1717 */
1718 *q = ip_srcrt.dst;
1719 #ifdef DIAGNOSTIC
1720 if (ipprintfs)
1721 printf(" %lx\n", (u_long)ntohl(q->s_addr));
1722 #endif
1723 return (m);
1724 }
1725
1726 /*
1727 * Strip out IP options, at higher
1728 * level protocol in the kernel.
1729 * Second argument is buffer to which options
1730 * will be moved, and return value is their length.
1731 * XXX should be deleted; last arg currently ignored.
1732 */
1733 void
1734 ip_stripoptions(m, mopt)
1735 register struct mbuf *m;
1736 struct mbuf *mopt;
1737 {
1738 register int i;
1739 struct ip *ip = mtod(m, struct ip *);
1740 register caddr_t opts;
1741 int olen;
1742
1743 olen = (ip->ip_hl << 2) - sizeof (struct ip);
1744 opts = (caddr_t)(ip + 1);
1745 i = m->m_len - (sizeof (struct ip) + olen);
1746 bcopy(opts + olen, opts, (unsigned)i);
1747 m->m_len -= olen;
1748 if (m->m_flags & M_PKTHDR)
1749 m->m_pkthdr.len -= olen;
1750 ip->ip_v = IPVERSION;
1751 ip->ip_hl = sizeof(struct ip) >> 2;
1752 }
1753
1754 u_char inetctlerrmap[PRC_NCMDS] = {
1755 0, 0, 0, 0,
1756 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
1757 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
1758 EMSGSIZE, EHOSTUNREACH, 0, 0,
1759 0, 0, EHOSTUNREACH, 0,
1760 ENOPROTOOPT, ECONNREFUSED
1761 };
1762
1763 /*
1764 * Forward a packet. If some error occurs return the sender
1765 * an icmp packet. Note we can't always generate a meaningful
1766 * icmp message because icmp doesn't have a large enough repertoire
1767 * of codes and types.
1768 *
1769 * If not forwarding, just drop the packet. This could be confusing
1770 * if ipforwarding was zero but some routing protocol was advancing
1771 * us as a gateway to somewhere. However, we must let the routing
1772 * protocol deal with that.
1773 *
1774 * The srcrt parameter indicates whether the packet is being forwarded
1775 * via a source route.
1776 */
1777 static void
1778 ip_forward(struct mbuf *m, int srcrt, struct sockaddr_in *next_hop)
1779 {
1780 struct ip *ip = mtod(m, struct ip *);
1781 struct in_ifaddr *ia;
1782 int error, type = 0, code = 0;
1783 struct mbuf *mcopy;
1784 n_long dest;
1785 struct in_addr pkt_dst;
1786 struct ifnet *destifp;
1787 #if defined(IPSEC) || defined(FAST_IPSEC)
1788 struct ifnet dummyifp;
1789 #endif
1790
1791 /*
1792 * Cache the destination address of the packet; this may be
1793 * changed by use of 'ipfw fwd'.
1794 */
1795 pkt_dst = next_hop ? next_hop->sin_addr : ip->ip_dst;
1796
1797 #ifdef DIAGNOSTIC
1798 if (ipprintfs)
1799 printf("forward: src %lx dst %lx ttl %x\n",
1800 (u_long)ip->ip_src.s_addr, (u_long)pkt_dst.s_addr,
1801 ip->ip_ttl);
1802 #endif
1803
1804
1805 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(pkt_dst) == 0) {
1806 ipstat.ips_cantforward++;
1807 m_freem(m);
1808 return;
1809 }
1810 #ifdef IPSTEALTH
1811 if (!ipstealth) {
1812 #endif
1813 if (ip->ip_ttl <= IPTTLDEC) {
1814 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS,
1815 0, 0);
1816 return;
1817 }
1818 #ifdef IPSTEALTH
1819 }
1820 #endif
1821
1822 if ((ia = ip_rtaddr(pkt_dst)) == 0) {
1823 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
1824 return;
1825 }
1826
1827 /*
1828 * Save the IP header and at most 8 bytes of the payload,
1829 * in case we need to generate an ICMP message to the src.
1830 *
1831 * XXX this can be optimized a lot by saving the data in a local
1832 * buffer on the stack (72 bytes at most), and only allocating the
1833 * mbuf if really necessary. The vast majority of the packets
1834 * are forwarded without having to send an ICMP back (either
1835 * because unnecessary, or because rate limited), so we are
1836 * really we are wasting a lot of work here.
1837 *
1838 * We don't use m_copy() because it might return a reference
1839 * to a shared cluster. Both this function and ip_output()
1840 * assume exclusive access to the IP header in `m', so any
1841 * data in a cluster may change before we reach icmp_error().
1842 */
1843 MGET(mcopy, M_DONTWAIT, m->m_type);
1844 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_DONTWAIT)) {
1845 /*
1846 * It's probably ok if the pkthdr dup fails (because
1847 * the deep copy of the tag chain failed), but for now
1848 * be conservative and just discard the copy since
1849 * code below may some day want the tags.
1850 */
1851 m_free(mcopy);
1852 mcopy = NULL;
1853 }
1854 if (mcopy != NULL) {
1855 mcopy->m_len = imin((ip->ip_hl << 2) + 8,
1856 (int)ip->ip_len);
1857 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1858 }
1859
1860 #ifdef IPSTEALTH
1861 if (!ipstealth) {
1862 #endif
1863 ip->ip_ttl -= IPTTLDEC;
1864 #ifdef IPSTEALTH
1865 }
1866 #endif
1867
1868 /*
1869 * If forwarding packet using same interface that it came in on,
1870 * perhaps should send a redirect to sender to shortcut a hop.
1871 * Only send redirect if source is sending directly to us,
1872 * and if packet was not source routed (or has any options).
1873 * Also, don't send redirect if forwarding using a default route
1874 * or a route modified by a redirect.
1875 */
1876 dest = 0;
1877 if (ipsendredirects && ia->ia_ifp == m->m_pkthdr.rcvif) {
1878 struct sockaddr_in *sin;
1879 struct route ro;
1880 struct rtentry *rt;
1881
1882 bzero(&ro, sizeof(ro));
1883 sin = (struct sockaddr_in *)&ro.ro_dst;
1884 sin->sin_family = AF_INET;
1885 sin->sin_len = sizeof(*sin);
1886 sin->sin_addr = pkt_dst;
1887 rtalloc_ign(&ro, RTF_CLONING);
1888
1889 rt = ro.ro_rt;
1890
1891 if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1892 satosin(rt_key(rt))->sin_addr.s_addr != 0 &&
1893 ipsendredirects && !srcrt && !next_hop) {
1894 #define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa))
1895 u_long src = ntohl(ip->ip_src.s_addr);
1896
1897 if (RTA(rt) &&
1898 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1899 if (rt->rt_flags & RTF_GATEWAY)
1900 dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
1901 else
1902 dest = pkt_dst.s_addr;
1903 /* Router requirements says to only send host redirects */
1904 type = ICMP_REDIRECT;
1905 code = ICMP_REDIRECT_HOST;
1906 #ifdef DIAGNOSTIC
1907 if (ipprintfs)
1908 printf("redirect (%d) to %lx\n", code, (u_long)dest);
1909 #endif
1910 }
1911 }
1912 if (rt)
1913 RTFREE(rt);
1914 }
1915
1916 {
1917 struct m_hdr tag;
1918
1919 if (next_hop) {
1920 /* Pass IPFORWARD info if available */
1921
1922 tag.mh_type = MT_TAG;
1923 tag.mh_flags = PACKET_TAG_IPFORWARD;
1924 tag.mh_data = (caddr_t)next_hop;
1925 tag.mh_next = m;
1926 tag.mh_nextpkt = NULL;
1927 m = (struct mbuf *)&tag;
1928 }
1929 error = ip_output(m, (struct mbuf *)0, NULL, IP_FORWARDING, 0, NULL);
1930 }
1931 if (error)
1932 ipstat.ips_cantforward++;
1933 else {
1934 ipstat.ips_forward++;
1935 if (type)
1936 ipstat.ips_redirectsent++;
1937 else {
1938 if (mcopy)
1939 m_freem(mcopy);
1940 return;
1941 }
1942 }
1943 if (mcopy == NULL)
1944 return;
1945 destifp = NULL;
1946
1947 switch (error) {
1948
1949 case 0: /* forwarded, but need redirect */
1950 /* type, code set above */
1951 break;
1952
1953 case ENETUNREACH: /* shouldn't happen, checked above */
1954 case EHOSTUNREACH:
1955 case ENETDOWN:
1956 case EHOSTDOWN:
1957 default:
1958 type = ICMP_UNREACH;
1959 code = ICMP_UNREACH_HOST;
1960 break;
1961
1962 case EMSGSIZE:
1963 type = ICMP_UNREACH;
1964 code = ICMP_UNREACH_NEEDFRAG;
1965 #if defined(IPSEC) || defined(FAST_IPSEC)
1966 /*
1967 * If the packet is routed over IPsec tunnel, tell the
1968 * originator the tunnel MTU.
1969 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
1970 * XXX quickhack!!!
1971 */
1972 {
1973 struct secpolicy *sp = NULL;
1974 int ipsecerror;
1975 int ipsechdr;
1976 struct route *ro;
1977
1978 #ifdef IPSEC
1979 sp = ipsec4_getpolicybyaddr(mcopy,
1980 IPSEC_DIR_OUTBOUND,
1981 IP_FORWARDING,
1982 &ipsecerror);
1983 #else /* FAST_IPSEC */
1984 sp = ipsec_getpolicybyaddr(mcopy,
1985 IPSEC_DIR_OUTBOUND,
1986 IP_FORWARDING,
1987 &ipsecerror);
1988 #endif
1989 if (sp != NULL) {
1990 /* count IPsec header size */
1991 ipsechdr = ipsec4_hdrsiz(mcopy,
1992 IPSEC_DIR_OUTBOUND,
1993 NULL);
1994
1995 /*
1996 * find the correct route for outer IPv4
1997 * header, compute tunnel MTU.
1998 *
1999 * XXX BUG ALERT
2000 * The "dummyifp" code relies upon the fact
2001 * that icmp_error() touches only ifp->if_mtu.
2002 */
2003 /*XXX*/
2004 destifp = NULL;
2005 if (sp->req != NULL
2006 && sp->req->sav != NULL
2007 && sp->req->sav->sah != NULL) {
2008 ro = &sp->req->sav->sah->sa_route;
2009 if (ro->ro_rt && ro->ro_rt->rt_ifp) {
2010 dummyifp.if_mtu =
2011 ro->ro_rt->rt_ifp->if_mtu;
2012 dummyifp.if_mtu -= ipsechdr;
2013 destifp = &dummyifp;
2014 }
2015 }
2016
2017 #ifdef IPSEC
2018 key_freesp(sp);
2019 #else /* FAST_IPSEC */
2020 KEY_FREESP(&sp);
2021 #endif
2022 ipstat.ips_cantfrag++;
2023 break;
2024 } else
2025 #endif /*IPSEC || FAST_IPSEC*/
2026 destifp = ia->ia_ifp;
2027 #if defined(IPSEC) || defined(FAST_IPSEC)
2028 }
2029 #endif /*IPSEC || FAST_IPSEC*/
2030 ipstat.ips_cantfrag++;
2031 break;
2032
2033 case ENOBUFS:
2034 /*
2035 * A router should not generate ICMP_SOURCEQUENCH as
2036 * required in RFC1812 Requirements for IP Version 4 Routers.
2037 * Source quench could be a big problem under DoS attacks,
2038 * or if the underlying interface is rate-limited.
2039 * Those who need source quench packets may re-enable them
2040 * via the net.inet.ip.sendsourcequench sysctl.
2041 */
2042 if (ip_sendsourcequench == 0) {
2043 m_freem(mcopy);
2044 return;
2045 } else {
2046 type = ICMP_SOURCEQUENCH;
2047 code = 0;
2048 }
2049 break;
2050
2051 case EACCES: /* ipfw denied packet */
2052 m_freem(mcopy);
2053 return;
2054 }
2055 icmp_error(mcopy, type, code, dest, destifp);
2056 }
2057
2058 void
2059 ip_savecontrol(inp, mp, ip, m)
2060 register struct inpcb *inp;
2061 register struct mbuf **mp;
2062 register struct ip *ip;
2063 register struct mbuf *m;
2064 {
2065 if (inp->inp_socket->so_options & SO_TIMESTAMP) {
2066 struct timeval tv;
2067
2068 microtime(&tv);
2069 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
2070 SCM_TIMESTAMP, SOL_SOCKET);
2071 if (*mp)
2072 mp = &(*mp)->m_next;
2073 }
2074 if (inp->inp_flags & INP_RECVDSTADDR) {
2075 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
2076 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
2077 if (*mp)
2078 mp = &(*mp)->m_next;
2079 }
2080 if (inp->inp_flags & INP_RECVTTL) {
2081 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl,
2082 sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
2083 if (*mp)
2084 mp = &(*mp)->m_next;
2085 }
2086 #ifdef notyet
2087 /* XXX
2088 * Moving these out of udp_input() made them even more broken
2089 * than they already were.
2090 */
2091 /* options were tossed already */
2092 if (inp->inp_flags & INP_RECVOPTS) {
2093 *mp = sbcreatecontrol((caddr_t) opts_deleted_above,
2094 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
2095 if (*mp)
2096 mp = &(*mp)->m_next;
2097 }
2098 /* ip_srcroute doesn't do what we want here, need to fix */
2099 if (inp->inp_flags & INP_RECVRETOPTS) {
2100 *mp = sbcreatecontrol((caddr_t) ip_srcroute(),
2101 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
2102 if (*mp)
2103 mp = &(*mp)->m_next;
2104 }
2105 #endif
2106 if (inp->inp_flags & INP_RECVIF) {
2107 struct ifnet *ifp;
2108 struct sdlbuf {
2109 struct sockaddr_dl sdl;
2110 u_char pad[32];
2111 } sdlbuf;
2112 struct sockaddr_dl *sdp;
2113 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
2114
2115 if (((ifp = m->m_pkthdr.rcvif))
2116 && ( ifp->if_index && (ifp->if_index <= if_index))) {
2117 sdp = (struct sockaddr_dl *)
2118 (ifaddr_byindex(ifp->if_index)->ifa_addr);
2119 /*
2120 * Change our mind and don't try copy.
2121 */
2122 if ((sdp->sdl_family != AF_LINK)
2123 || (sdp->sdl_len > sizeof(sdlbuf))) {
2124 goto makedummy;
2125 }
2126 bcopy(sdp, sdl2, sdp->sdl_len);
2127 } else {
2128 makedummy:
2129 sdl2->sdl_len
2130 = offsetof(struct sockaddr_dl, sdl_data[0]);
2131 sdl2->sdl_family = AF_LINK;
2132 sdl2->sdl_index = 0;
2133 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
2134 }
2135 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len,
2136 IP_RECVIF, IPPROTO_IP);
2137 if (*mp)
2138 mp = &(*mp)->m_next;
2139 }
2140 }
2141
2142 /*
2143 * XXX these routines are called from the upper part of the kernel.
2144 * They need to be locked when we remove Giant.
2145 *
2146 * They could also be moved to ip_mroute.c, since all the RSVP
2147 * handling is done there already.
2148 */
2149 static int ip_rsvp_on;
2150 struct socket *ip_rsvpd;
2151 int
2152 ip_rsvp_init(struct socket *so)
2153 {
2154 if (so->so_type != SOCK_RAW ||
2155 so->so_proto->pr_protocol != IPPROTO_RSVP)
2156 return EOPNOTSUPP;
2157
2158 if (ip_rsvpd != NULL)
2159 return EADDRINUSE;
2160
2161 ip_rsvpd = so;
2162 /*
2163 * This may seem silly, but we need to be sure we don't over-increment
2164 * the RSVP counter, in case something slips up.
2165 */
2166 if (!ip_rsvp_on) {
2167 ip_rsvp_on = 1;
2168 rsvp_on++;
2169 }
2170
2171 return 0;
2172 }
2173
2174 int
2175 ip_rsvp_done(void)
2176 {
2177 ip_rsvpd = NULL;
2178 /*
2179 * This may seem silly, but we need to be sure we don't over-decrement
2180 * the RSVP counter, in case something slips up.
2181 */
2182 if (ip_rsvp_on) {
2183 ip_rsvp_on = 0;
2184 rsvp_on--;
2185 }
2186 return 0;
2187 }
2188
2189 void
2190 rsvp_input(struct mbuf *m, int off) /* XXX must fixup manually */
2191 {
2192 if (rsvp_input_p) { /* call the real one if loaded */
2193 rsvp_input_p(m, off);
2194 return;
2195 }
2196
2197 /* Can still get packets with rsvp_on = 0 if there is a local member
2198 * of the group to which the RSVP packet is addressed. But in this
2199 * case we want to throw the packet away.
2200 */
2201
2202 if (!rsvp_on) {
2203 m_freem(m);
2204 return;
2205 }
2206
2207 if (ip_rsvpd != NULL) {
2208 rip_input(m, off);
2209 return;
2210 }
2211 /* Drop the packet */
2212 m_freem(m);
2213 }
Cache object: 876997b07aef21087f4eda557aaaf0be
|