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: head/sys/netinet/ip_input.c 122723 2003-11-15 01:45:56Z 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
93int rsvp_on = 0;
94
95int ipforwarding = 0;
96SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW,
97 &ipforwarding, 0, "Enable IP forwarding between interfaces");
98
99static int ipsendredirects = 1; /* XXX */
100SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW,
101 &ipsendredirects, 0, "Enable sending IP redirects");
102
103int ip_defttl = IPDEFTTL;
104SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW,
105 &ip_defttl, 0, "Maximum TTL on IP packets");
106
107static int ip_dosourceroute = 0;
108SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW,
109 &ip_dosourceroute, 0, "Enable forwarding source routed IP packets");
110
111static int ip_acceptsourceroute = 0;
112SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute,
113 CTLFLAG_RW, &ip_acceptsourceroute, 0,
114 "Enable accepting source routed IP packets");
115
116static int ip_keepfaith = 0;
117SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW,
118 &ip_keepfaith, 0,
119 "Enable packet capture for FAITH IPv4->IPv6 translater daemon");
120
121static int nipq = 0; /* total # of reass queues */
122static int maxnipq;
123SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragpackets, CTLFLAG_RW,
124 &maxnipq, 0,
125 "Maximum number of IPv4 fragment reassembly queue entries");
126
127static int maxfragsperpacket;
128SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW,
129 &maxfragsperpacket, 0,
130 "Maximum number of IPv4 fragments allowed per packet");
131
132static int ip_sendsourcequench = 0;
133SYSCTL_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 */
150static int ip_checkinterface = 1;
151SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW,
152 &ip_checkinterface, 0, "Verify packet arrives on correct interface");
153
154#ifdef DIAGNOSTIC
155static int ipprintfs = 0;
156#endif
157#ifdef PFIL_HOOKS
158struct pfil_head inet_pfil_hook;
159#endif
160
161static struct ifqueue ipintrq;
162static int ipqmaxlen = IFQ_MAXLEN;
163
164extern struct domain inetdomain;
165extern struct protosw inetsw[];
166u_char ip_protox[IPPROTO_MAX];
167struct in_ifaddrhead in_ifaddrhead; /* first inet address */
168struct in_ifaddrhashhead *in_ifaddrhashtbl; /* inet addr hash table */
169u_long in_ifaddrhmask; /* mask for hash table */
170
171SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW,
172 &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue");
173SYSCTL_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
176struct ipstat ipstat;
177SYSCTL_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
187static TAILQ_HEAD(ipqhead, ipq) ipq[IPREASS_NHASH];
188struct 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
196SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
197 &ip_mtu, 0, "Default MTU");
198#endif
199
200#ifdef IPSTEALTH
201int ipstealth = 0;
202SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW,
203 &ipstealth, 0, "");
204#endif
205
206
207/* Firewall hooks */
208ip_fw_chk_t *ip_fw_chk_ptr;
209int fw_enable = 1 ;
210int fw_one_pass = 1;
211
212/* Dummynet hooks */
213ip_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 */
228static int ip_nhops = 0;
229static 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
236static void save_rte(u_char *, struct in_addr);
237static int ip_dooptions(struct mbuf *m, int,
238 struct sockaddr_in *next_hop);
239static void ip_forward(struct mbuf *m, int srcrt,
240 struct sockaddr_in *next_hop);
241static void ip_freef(struct ipqhead *, struct ipq *);
242static 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 */
249void
250ip_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 */
295void
296ip_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) {
448tooshort:
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
486iphack:
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 }
546pass:
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
743ours:
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 }
819found:
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;
956DPRINTF(("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
979 (*inetsw[ip_protox[ip->ip_p]].pr_input)(
980 (struct mbuf *)&tag, hlen);
981 } else
982 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen);
983 NET_DROP_GIANT();
984 return;
985bad:
986 m_freem(m);
987}
988
989/*
990 * Take incoming datagram fragment and try to reassemble it into
991 * whole datagram. If a chain for reassembly of this datagram already
992 * exists, then it is given as fp; otherwise have to make a chain.
993 *
994 * When IPDIVERT enabled, keep additional state with each packet that
995 * tells us if we need to divert or tee the packet we're building.
996 * In particular, *divinfo includes the port and TEE flag,
997 * *divert_rule is the number of the matching rule.
998 */
999
1000static struct mbuf *
1001ip_reass(struct mbuf *m, struct ipqhead *head, struct ipq *fp,
1002 u_int32_t *divinfo, u_int16_t *divert_rule)
1003{
1004 struct ip *ip = mtod(m, struct ip *);
1005 register struct mbuf *p, *q, *nq;
1006 struct mbuf *t;
1007 int hlen = ip->ip_hl << 2;
1008 int i, next;
1009 u_int8_t ecn, ecn0;
1010
1011 IPQ_LOCK_ASSERT();
1012
1013 /*
1014 * Presence of header sizes in mbufs
1015 * would confuse code below.
1016 */
1017 m->m_data += hlen;
1018 m->m_len -= hlen;
1019
1020 /*
1021 * If first fragment to arrive, create a reassembly queue.
1022 */
1023 if (fp == NULL) {
1024 if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL)
1025 goto dropfrag;
1026 fp = mtod(t, struct ipq *);
1027#ifdef MAC
1028 if (mac_init_ipq(fp, M_NOWAIT) != 0) {
1029 m_free(t);
1030 goto dropfrag;
1031 }
1032 mac_create_ipq(m, fp);
1033#endif
1034 TAILQ_INSERT_HEAD(head, fp, ipq_list);
1035 nipq++;
1036 fp->ipq_nfrags = 1;
1037 fp->ipq_ttl = IPFRAGTTL;
1038 fp->ipq_p = ip->ip_p;
1039 fp->ipq_id = ip->ip_id;
1040 fp->ipq_src = ip->ip_src;
1041 fp->ipq_dst = ip->ip_dst;
1042 fp->ipq_frags = m;
1043 m->m_nextpkt = NULL;
1044#ifdef IPDIVERT
1045 fp->ipq_div_info = 0;
1046 fp->ipq_div_cookie = 0;
1047#endif
1048 goto inserted;
1049 } else {
1050 fp->ipq_nfrags++;
1051#ifdef MAC
1052 mac_update_ipq(m, fp);
1053#endif
1054 }
1055
1056#define GETIP(m) ((struct ip*)((m)->m_pkthdr.header))
1057
1058 /*
1059 * Handle ECN by comparing this segment with the first one;
1060 * if CE is set, do not lose CE.
1061 * drop if CE and not-ECT are mixed for the same packet.
1062 */
1063 ecn = ip->ip_tos & IPTOS_ECN_MASK;
1064 ecn0 = GETIP(fp->ipq_frags)->ip_tos & IPTOS_ECN_MASK;
1065 if (ecn == IPTOS_ECN_CE) {
1066 if (ecn0 == IPTOS_ECN_NOTECT)
1067 goto dropfrag;
1068 if (ecn0 != IPTOS_ECN_CE)
1069 GETIP(fp->ipq_frags)->ip_tos |= IPTOS_ECN_CE;
1070 }
1071 if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT)
1072 goto dropfrag;
1073
1074 /*
1075 * Find a segment which begins after this one does.
1076 */
1077 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt)
1078 if (GETIP(q)->ip_off > ip->ip_off)
1079 break;
1080
1081 /*
1082 * If there is a preceding segment, it may provide some of
1083 * our data already. If so, drop the data from the incoming
1084 * segment. If it provides all of our data, drop us, otherwise
1085 * stick new segment in the proper place.
1086 *
1087 * If some of the data is dropped from the the preceding
1088 * segment, then it's checksum is invalidated.
1089 */
1090 if (p) {
1091 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off;
1092 if (i > 0) {
1093 if (i >= ip->ip_len)
1094 goto dropfrag;
1095 m_adj(m, i);
1096 m->m_pkthdr.csum_flags = 0;
1097 ip->ip_off += i;
1098 ip->ip_len -= i;
1099 }
1100 m->m_nextpkt = p->m_nextpkt;
1101 p->m_nextpkt = m;
1102 } else {
1103 m->m_nextpkt = fp->ipq_frags;
1104 fp->ipq_frags = m;
1105 }
1106
1107 /*
1108 * While we overlap succeeding segments trim them or,
1109 * if they are completely covered, dequeue them.
1110 */
1111 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off;
1112 q = nq) {
1113 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off;
1114 if (i < GETIP(q)->ip_len) {
1115 GETIP(q)->ip_len -= i;
1116 GETIP(q)->ip_off += i;
1117 m_adj(q, i);
1118 q->m_pkthdr.csum_flags = 0;
1119 break;
1120 }
1121 nq = q->m_nextpkt;
1122 m->m_nextpkt = nq;
1123 ipstat.ips_fragdropped++;
1124 fp->ipq_nfrags--;
1125 m_freem(q);
1126 }
1127
1128inserted:
1129
1130#ifdef IPDIVERT
1131 /*
1132 * Transfer firewall instructions to the fragment structure.
1133 * Only trust info in the fragment at offset 0.
1134 */
1135 if (ip->ip_off == 0) {
1136 fp->ipq_div_info = *divinfo;
1137 fp->ipq_div_cookie = *divert_rule;
1138 }
1139 *divinfo = 0;
1140 *divert_rule = 0;
1141#endif
1142
1143 /*
1144 * Check for complete reassembly and perform frag per packet
1145 * limiting.
1146 *
1147 * Frag limiting is performed here so that the nth frag has
1148 * a chance to complete the packet before we drop the packet.
1149 * As a result, n+1 frags are actually allowed per packet, but
1150 * only n will ever be stored. (n = maxfragsperpacket.)
1151 *
1152 */
1153 next = 0;
1154 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
1155 if (GETIP(q)->ip_off != next) {
1156 if (fp->ipq_nfrags > maxfragsperpacket) {
1157 ipstat.ips_fragdropped += fp->ipq_nfrags;
1158 ip_freef(head, fp);
1159 }
1160 return (0);
1161 }
1162 next += GETIP(q)->ip_len;
1163 }
1164 /* Make sure the last packet didn't have the IP_MF flag */
1165 if (p->m_flags & M_FRAG) {
1166 if (fp->ipq_nfrags > maxfragsperpacket) {
1167 ipstat.ips_fragdropped += fp->ipq_nfrags;
1168 ip_freef(head, fp);
1169 }
1170 return (0);
1171 }
1172
1173 /*
1174 * Reassembly is complete. Make sure the packet is a sane size.
1175 */
1176 q = fp->ipq_frags;
1177 ip = GETIP(q);
1178 if (next + (ip->ip_hl << 2) > IP_MAXPACKET) {
1179 ipstat.ips_toolong++;
1180 ipstat.ips_fragdropped += fp->ipq_nfrags;
1181 ip_freef(head, fp);
1182 return (0);
1183 }
1184
1185 /*
1186 * Concatenate fragments.
1187 */
1188 m = q;
1189 t = m->m_next;
1190 m->m_next = 0;
1191 m_cat(m, t);
1192 nq = q->m_nextpkt;
1193 q->m_nextpkt = 0;
1194 for (q = nq; q != NULL; q = nq) {
1195 nq = q->m_nextpkt;
1196 q->m_nextpkt = NULL;
1197 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
1198 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
1199 m_cat(m, q);
1200 }
1201#ifdef MAC
1202 mac_create_datagram_from_ipq(fp, m);
1203 mac_destroy_ipq(fp);
1204#endif
1205
1206#ifdef IPDIVERT
1207 /*
1208 * Extract firewall instructions from the fragment structure.
1209 */
1210 *divinfo = fp->ipq_div_info;
1211 *divert_rule = fp->ipq_div_cookie;
1212#endif
1213
1214 /*
1215 * Create header for new ip packet by
1216 * modifying header of first packet;
1217 * dequeue and discard fragment reassembly header.
1218 * Make header visible.
1219 */
1220 ip->ip_len = next;
1221 ip->ip_src = fp->ipq_src;
1222 ip->ip_dst = fp->ipq_dst;
1223 TAILQ_REMOVE(head, fp, ipq_list);
1224 nipq--;
1225 (void) m_free(dtom(fp));
1226 m->m_len += (ip->ip_hl << 2);
1227 m->m_data -= (ip->ip_hl << 2);
1228 /* some debugging cruft by sklower, below, will go away soon */
1229 if (m->m_flags & M_PKTHDR) /* XXX this should be done elsewhere */
1230 m_fixhdr(m);
1231 return (m);
1232
1233dropfrag:
1234#ifdef IPDIVERT
1235 *divinfo = 0;
1236 *divert_rule = 0;
1237#endif
1238 ipstat.ips_fragdropped++;
1239 if (fp != NULL)
1240 fp->ipq_nfrags--;
1241 m_freem(m);
1242 return (0);
1243
1244#undef GETIP
1245}
1246
1247/*
1248 * Free a fragment reassembly header and all
1249 * associated datagrams.
1250 */
1251static void
1252ip_freef(fhp, fp)
1253 struct ipqhead *fhp;
1254 struct ipq *fp;
1255{
1256 register struct mbuf *q;
1257
1258 IPQ_LOCK_ASSERT();
1259
1260 while (fp->ipq_frags) {
1261 q = fp->ipq_frags;
1262 fp->ipq_frags = q->m_nextpkt;
1263 m_freem(q);
1264 }
1265 TAILQ_REMOVE(fhp, fp, ipq_list);
1266 (void) m_free(dtom(fp));
1267 nipq--;
1268}
1269
1270/*
1271 * IP timer processing;
1272 * if a timer expires on a reassembly
1273 * queue, discard it.
1274 */
1275void
1276ip_slowtimo()
1277{
1278 register struct ipq *fp;
1279 int s = splnet();
1280 int i;
1281
1282 IPQ_LOCK();
1283 for (i = 0; i < IPREASS_NHASH; i++) {
1284 for(fp = TAILQ_FIRST(&ipq[i]); fp;) {
1285 struct ipq *fpp;
1286
1287 fpp = fp;
1288 fp = TAILQ_NEXT(fp, ipq_list);
1289 if(--fpp->ipq_ttl == 0) {
1290 ipstat.ips_fragtimeout += fpp->ipq_nfrags;
1291 ip_freef(&ipq[i], fpp);
1292 }
1293 }
1294 }
1295 /*
1296 * If we are over the maximum number of fragments
1297 * (due to the limit being lowered), drain off
1298 * enough to get down to the new limit.
1299 */
1300 if (maxnipq >= 0 && nipq > maxnipq) {
1301 for (i = 0; i < IPREASS_NHASH; i++) {
1302 while (nipq > maxnipq && !TAILQ_EMPTY(&ipq[i])) {
1303 ipstat.ips_fragdropped +=
1304 TAILQ_FIRST(&ipq[i])->ipq_nfrags;
1305 ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i]));
1306 }
1307 }
1308 }
1309 IPQ_UNLOCK();
1310 splx(s);
1311}
1312
1313/*
1314 * Drain off all datagram fragments.
1315 */
1316void
1317ip_drain()
1318{
1319 int i;
1320
1321 IPQ_LOCK();
1322 for (i = 0; i < IPREASS_NHASH; i++) {
1323 while(!TAILQ_EMPTY(&ipq[i])) {
1324 ipstat.ips_fragdropped +=
1325 TAILQ_FIRST(&ipq[i])->ipq_nfrags;
1326 ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i]));
1327 }
1328 }
1329 IPQ_UNLOCK();
1330 in_rtqdrain();
1331}
1332
1333/*
1334 * Do option processing on a datagram,
1335 * possibly discarding it if bad options are encountered,
1336 * or forwarding it if source-routed.
1337 * The pass argument is used when operating in the IPSTEALTH
1338 * mode to tell what options to process:
1339 * [LS]SRR (pass 0) or the others (pass 1).
1340 * The reason for as many as two passes is that when doing IPSTEALTH,
1341 * non-routing options should be processed only if the packet is for us.
1342 * Returns 1 if packet has been forwarded/freed,
1343 * 0 if the packet should be processed further.
1344 */
1345static int
1346ip_dooptions(struct mbuf *m, int pass, struct sockaddr_in *next_hop)
1347{
1348 struct ip *ip = mtod(m, struct ip *);
1349 u_char *cp;
1350 struct in_ifaddr *ia;
1351 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0;
1352 struct in_addr *sin, dst;
1353 n_time ntime;
1354 struct sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET };
1355
1356 dst = ip->ip_dst;
1357 cp = (u_char *)(ip + 1);
1358 cnt = (ip->ip_hl << 2) - sizeof (struct ip);
1359 for (; cnt > 0; cnt -= optlen, cp += optlen) {
1360 opt = cp[IPOPT_OPTVAL];
1361 if (opt == IPOPT_EOL)
1362 break;
1363 if (opt == IPOPT_NOP)
1364 optlen = 1;
1365 else {
1366 if (cnt < IPOPT_OLEN + sizeof(*cp)) {
1367 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1368 goto bad;
1369 }
1370 optlen = cp[IPOPT_OLEN];
1371 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
1372 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1373 goto bad;
1374 }
1375 }
1376 switch (opt) {
1377
1378 default:
1379 break;
1380
1381 /*
1382 * Source routing with record.
1383 * Find interface with current destination address.
1384 * If none on this machine then drop if strictly routed,
1385 * or do nothing if loosely routed.
1386 * Record interface address and bring up next address
1387 * component. If strictly routed make sure next
1388 * address is on directly accessible net.
1389 */
1390 case IPOPT_LSRR:
1391 case IPOPT_SSRR:
1392#ifdef IPSTEALTH
1393 if (ipstealth && pass > 0)
1394 break;
1395#endif
1396 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1397 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1398 goto bad;
1399 }
1400 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1401 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1402 goto bad;
1403 }
1404 ipaddr.sin_addr = ip->ip_dst;
1405 ia = (struct in_ifaddr *)
1406 ifa_ifwithaddr((struct sockaddr *)&ipaddr);
1407 if (ia == 0) {
1408 if (opt == IPOPT_SSRR) {
1409 type = ICMP_UNREACH;
1410 code = ICMP_UNREACH_SRCFAIL;
1411 goto bad;
1412 }
1413 if (!ip_dosourceroute)
1414 goto nosourcerouting;
1415 /*
1416 * Loose routing, and not at next destination
1417 * yet; nothing to do except forward.
1418 */
1419 break;
1420 }
1421 off--; /* 0 origin */
1422 if (off > optlen - (int)sizeof(struct in_addr)) {
1423 /*
1424 * End of source route. Should be for us.
1425 */
1426 if (!ip_acceptsourceroute)
1427 goto nosourcerouting;
1428 save_rte(cp, ip->ip_src);
1429 break;
1430 }
1431#ifdef IPSTEALTH
1432 if (ipstealth)
1433 goto dropit;
1434#endif
1435 if (!ip_dosourceroute) {
1436 if (ipforwarding) {
1437 char buf[16]; /* aaa.bbb.ccc.ddd\0 */
1438 /*
1439 * Acting as a router, so generate ICMP
1440 */
1441nosourcerouting:
1442 strcpy(buf, inet_ntoa(ip->ip_dst));
1443 log(LOG_WARNING,
1444 "attempted source route from %s to %s\n",
1445 inet_ntoa(ip->ip_src), buf);
1446 type = ICMP_UNREACH;
1447 code = ICMP_UNREACH_SRCFAIL;
1448 goto bad;
1449 } else {
1450 /*
1451 * Not acting as a router, so silently drop.
1452 */
1453#ifdef IPSTEALTH
1454dropit:
1455#endif
1456 ipstat.ips_cantforward++;
1457 m_freem(m);
1458 return (1);
1459 }
1460 }
1461
1462 /*
1463 * locate outgoing interface
1464 */
1465 (void)memcpy(&ipaddr.sin_addr, cp + off,
1466 sizeof(ipaddr.sin_addr));
1467
1468 if (opt == IPOPT_SSRR) {
1469#define INA struct in_ifaddr *
1470#define SA struct sockaddr *
1471 if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == 0)
1472 ia = (INA)ifa_ifwithnet((SA)&ipaddr);
1473 } else
1474 ia = ip_rtaddr(ipaddr.sin_addr);
1475 if (ia == 0) {
1476 type = ICMP_UNREACH;
1477 code = ICMP_UNREACH_SRCFAIL;
1478 goto bad;
1479 }
1480 ip->ip_dst = ipaddr.sin_addr;
1481 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
1482 sizeof(struct in_addr));
1483 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1484 /*
1485 * Let ip_intr's mcast routing check handle mcast pkts
1486 */
1487 forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr));
1488 break;
1489
1490 case IPOPT_RR:
1491#ifdef IPSTEALTH
1492 if (ipstealth && pass == 0)
1493 break;
1494#endif
1495 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1496 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1497 goto bad;
1498 }
1499 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1500 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1501 goto bad;
1502 }
1503 /*
1504 * If no space remains, ignore.
1505 */
1506 off--; /* 0 origin */
1507 if (off > optlen - (int)sizeof(struct in_addr))
1508 break;
1509 (void)memcpy(&ipaddr.sin_addr, &ip->ip_dst,
1510 sizeof(ipaddr.sin_addr));
1511 /*
1512 * locate outgoing interface; if we're the destination,
1513 * use the incoming interface (should be same).
1514 */
1515 if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == 0 &&
1516 (ia = ip_rtaddr(ipaddr.sin_addr)) == 0) {
1517 type = ICMP_UNREACH;
1518 code = ICMP_UNREACH_HOST;
1519 goto bad;
1520 }
1521 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
1522 sizeof(struct in_addr));
1523 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1524 break;
1525
1526 case IPOPT_TS:
1527#ifdef IPSTEALTH
1528 if (ipstealth && pass == 0)
1529 break;
1530#endif
1531 code = cp - (u_char *)ip;
1532 if (optlen < 4 || optlen > 40) {
1533 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1534 goto bad;
1535 }
1536 if ((off = cp[IPOPT_OFFSET]) < 5) {
1537 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1538 goto bad;
1539 }
1540 if (off > optlen - (int)sizeof(int32_t)) {
1541 cp[IPOPT_OFFSET + 1] += (1 << 4);
1542 if ((cp[IPOPT_OFFSET + 1] & 0xf0) == 0) {
1543 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1544 goto bad;
1545 }
1546 break;
1547 }
1548 off--; /* 0 origin */
1549 sin = (struct in_addr *)(cp + off);
1550 switch (cp[IPOPT_OFFSET + 1] & 0x0f) {
1551
1552 case IPOPT_TS_TSONLY:
1553 break;
1554
1555 case IPOPT_TS_TSANDADDR:
1556 if (off + sizeof(n_time) +
1557 sizeof(struct in_addr) > optlen) {
1558 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1559 goto bad;
1560 }
1561 ipaddr.sin_addr = dst;
1562 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr,
1563 m->m_pkthdr.rcvif);
1564 if (ia == 0)
1565 continue;
1566 (void)memcpy(sin, &IA_SIN(ia)->sin_addr,
1567 sizeof(struct in_addr));
1568 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1569 off += sizeof(struct in_addr);
1570 break;
1571
1572 case IPOPT_TS_PRESPEC:
1573 if (off + sizeof(n_time) +
1574 sizeof(struct in_addr) > optlen) {
1575 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1576 goto bad;
1577 }
1578 (void)memcpy(&ipaddr.sin_addr, sin,
1579 sizeof(struct in_addr));
1580 if (ifa_ifwithaddr((SA)&ipaddr) == 0)
1581 continue;
1582 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1583 off += sizeof(struct in_addr);
1584 break;
1585
1586 default:
1587 code = &cp[IPOPT_OFFSET + 1] - (u_char *)ip;
1588 goto bad;
1589 }
1590 ntime = iptime();
1591 (void)memcpy(cp + off, &ntime, sizeof(n_time));
1592 cp[IPOPT_OFFSET] += sizeof(n_time);
1593 }
1594 }
1595 if (forward && ipforwarding) {
1596 ip_forward(m, 1, next_hop);
1597 return (1);
1598 }
1599 return (0);
1600bad:
1601 icmp_error(m, type, code, 0, 0);
1602 ipstat.ips_badoptions++;
1603 return (1);
1604}
1605
1606/*
1607 * Given address of next destination (final or next hop),
1608 * return internet address info of interface to be used to get there.
1609 */
1610struct in_ifaddr *
1611ip_rtaddr(dst)
1612 struct in_addr dst;
1613{
1614 struct sockaddr_in *sin;
1615 struct in_ifaddr *ifa;
1616 struct route ro;
1617
1618 bzero(&ro, sizeof(ro));
1619 sin = (struct sockaddr_in *)&ro.ro_dst;
1620 sin->sin_family = AF_INET;
1621 sin->sin_len = sizeof(*sin);
1622 sin->sin_addr = dst;
1623 rtalloc_ign(&ro, (RTF_PRCLONING | RTF_CLONING));
1624
1625 if (ro.ro_rt == 0)
1626 return ((struct in_ifaddr *)0);
1627
1628 ifa = ifatoia(ro.ro_rt->rt_ifa);
1629 RTFREE(ro.ro_rt);
1630 return ifa;
1631}
1632
1633/*
1634 * Save incoming source route for use in replies,
1635 * to be picked up later by ip_srcroute if the receiver is interested.
1636 */
1637static void
1638save_rte(option, dst)
1639 u_char *option;
1640 struct in_addr dst;
1641{
1642 unsigned olen;
1643
1644 olen = option[IPOPT_OLEN];
1645#ifdef DIAGNOSTIC
1646 if (ipprintfs)
1647 printf("save_rte: olen %d\n", olen);
1648#endif
1649 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst)))
1650 return;
1651 bcopy(option, ip_srcrt.srcopt, olen);
1652 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1653 ip_srcrt.dst = dst;
1654}
1655
1656/*
1657 * Retrieve incoming source route for use in replies,
1658 * in the same form used by setsockopt.
1659 * The first hop is placed before the options, will be removed later.
1660 */
1661struct mbuf *
1662ip_srcroute()
1663{
1664 register struct in_addr *p, *q;
1665 register struct mbuf *m;
1666
1667 if (ip_nhops == 0)
1668 return ((struct mbuf *)0);
1669 m = m_get(M_DONTWAIT, MT_HEADER);
1670 if (m == 0)
1671 return ((struct mbuf *)0);
1672
1673#define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt))
1674
1675 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
1676 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) +
1677 OPTSIZ;
1678#ifdef DIAGNOSTIC
1679 if (ipprintfs)
1680 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len);
1681#endif
1682
1683 /*
1684 * First save first hop for return route
1685 */
1686 p = &ip_srcrt.route[ip_nhops - 1];
1687 *(mtod(m, struct in_addr *)) = *p--;
1688#ifdef DIAGNOSTIC
1689 if (ipprintfs)
1690 printf(" hops %lx", (u_long)ntohl(mtod(m, struct in_addr *)->s_addr));
1691#endif
1692
1693 /*
1694 * Copy option fields and padding (nop) to mbuf.
1695 */
1696 ip_srcrt.nop = IPOPT_NOP;
1697 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
1698 (void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr),
1699 &ip_srcrt.nop, OPTSIZ);
1700 q = (struct in_addr *)(mtod(m, caddr_t) +
1701 sizeof(struct in_addr) + OPTSIZ);
1702#undef OPTSIZ
1703 /*
1704 * Record return path as an IP source route,
1705 * reversing the path (pointers are now aligned).
1706 */
1707 while (p >= ip_srcrt.route) {
1708#ifdef DIAGNOSTIC
1709 if (ipprintfs)
1710 printf(" %lx", (u_long)ntohl(q->s_addr));
1711#endif
1712 *q++ = *p--;
1713 }
1714 /*
1715 * Last hop goes to final destination.
1716 */
1717 *q = ip_srcrt.dst;
1718#ifdef DIAGNOSTIC
1719 if (ipprintfs)
1720 printf(" %lx\n", (u_long)ntohl(q->s_addr));
1721#endif
1722 return (m);
1723}
1724
1725/*
1726 * Strip out IP options, at higher
1727 * level protocol in the kernel.
1728 * Second argument is buffer to which options
1729 * will be moved, and return value is their length.
1730 * XXX should be deleted; last arg currently ignored.
1731 */
1732void
1733ip_stripoptions(m, mopt)
1734 register struct mbuf *m;
1735 struct mbuf *mopt;
1736{
1737 register int i;
1738 struct ip *ip = mtod(m, struct ip *);
1739 register caddr_t opts;
1740 int olen;
1741
1742 olen = (ip->ip_hl << 2) - sizeof (struct ip);
1743 opts = (caddr_t)(ip + 1);
1744 i = m->m_len - (sizeof (struct ip) + olen);
1745 bcopy(opts + olen, opts, (unsigned)i);
1746 m->m_len -= olen;
1747 if (m->m_flags & M_PKTHDR)
1748 m->m_pkthdr.len -= olen;
1749 ip->ip_v = IPVERSION;
1750 ip->ip_hl = sizeof(struct ip) >> 2;
1751}
1752
1753u_char inetctlerrmap[PRC_NCMDS] = {
1754 0, 0, 0, 0,
1755 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
1756 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
1757 EMSGSIZE, EHOSTUNREACH, 0, 0,
1758 0, 0, EHOSTUNREACH, 0,
1759 ENOPROTOOPT, ECONNREFUSED
1760};
1761
1762/*
1763 * Forward a packet. If some error occurs return the sender
1764 * an icmp packet. Note we can't always generate a meaningful
1765 * icmp message because icmp doesn't have a large enough repertoire
1766 * of codes and types.
1767 *
1768 * If not forwarding, just drop the packet. This could be confusing
1769 * if ipforwarding was zero but some routing protocol was advancing
1770 * us as a gateway to somewhere. However, we must let the routing
1771 * protocol deal with that.
1772 *
1773 * The srcrt parameter indicates whether the packet is being forwarded
1774 * via a source route.
1775 */
1776static void
1777ip_forward(struct mbuf *m, int srcrt, struct sockaddr_in *next_hop)
1778{
1779 struct ip *ip = mtod(m, struct ip *);
1780 struct in_ifaddr *ia;
1781 int error, type = 0, code = 0;
1782 struct mbuf *mcopy;
1783 n_long dest;
1784 struct in_addr pkt_dst;
1785 struct ifnet *destifp;
1786#if defined(IPSEC) || defined(FAST_IPSEC)
1787 struct ifnet dummyifp;
1788#endif
1789
1790 /*
1791 * Cache the destination address of the packet; this may be
1792 * changed by use of 'ipfw fwd'.
1793 */
1794 pkt_dst = next_hop ? next_hop->sin_addr : ip->ip_dst;
1795
1796#ifdef DIAGNOSTIC
1797 if (ipprintfs)
1798 printf("forward: src %lx dst %lx ttl %x\n",
1799 (u_long)ip->ip_src.s_addr, (u_long)pkt_dst.s_addr,
1800 ip->ip_ttl);
1801#endif
1802
1803
1804 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(pkt_dst) == 0) {
1805 ipstat.ips_cantforward++;
1806 m_freem(m);
1807 return;
1808 }
1809#ifdef IPSTEALTH
1810 if (!ipstealth) {
1811#endif
1812 if (ip->ip_ttl <= IPTTLDEC) {
1813 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS,
1814 0, 0);
1815 return;
1816 }
1817#ifdef IPSTEALTH
1818 }
1819#endif
1820
1821 if ((ia = ip_rtaddr(pkt_dst)) == 0) {
1822 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
1823 return;
1824 }
1825
1826 /*
1827 * Save the IP header and at most 8 bytes of the payload,
1828 * in case we need to generate an ICMP message to the src.
1829 *
1830 * XXX this can be optimized a lot by saving the data in a local
1831 * buffer on the stack (72 bytes at most), and only allocating the
1832 * mbuf if really necessary. The vast majority of the packets
1833 * are forwarded without having to send an ICMP back (either
1834 * because unnecessary, or because rate limited), so we are
1835 * really we are wasting a lot of work here.
1836 *
1837 * We don't use m_copy() because it might return a reference
1838 * to a shared cluster. Both this function and ip_output()
1839 * assume exclusive access to the IP header in `m', so any
1840 * data in a cluster may change before we reach icmp_error().
1841 */
1842 MGET(mcopy, M_DONTWAIT, m->m_type);
1843 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_DONTWAIT)) {
1844 /*
1845 * It's probably ok if the pkthdr dup fails (because
1846 * the deep copy of the tag chain failed), but for now
1847 * be conservative and just discard the copy since
1848 * code below may some day want the tags.
1849 */
1850 m_free(mcopy);
1851 mcopy = NULL;
1852 }
1853 if (mcopy != NULL) {
1854 mcopy->m_len = imin((ip->ip_hl << 2) + 8,
1855 (int)ip->ip_len);
1856 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1857 }
1858
1859#ifdef IPSTEALTH
1860 if (!ipstealth) {
1861#endif
1862 ip->ip_ttl -= IPTTLDEC;
1863#ifdef IPSTEALTH
1864 }
1865#endif
1866
1867 /*
1868 * If forwarding packet using same interface that it came in on,
1869 * perhaps should send a redirect to sender to shortcut a hop.
1870 * Only send redirect if source is sending directly to us,
1871 * and if packet was not source routed (or has any options).
1872 * Also, don't send redirect if forwarding using a default route
1873 * or a route modified by a redirect.
1874 */
1875 dest = 0;
1876 if (ipsendredirects && ia->ia_ifp == m->m_pkthdr.rcvif) {
1877 struct sockaddr_in *sin;
1878 struct route ro;
1879 struct rtentry *rt;
1880
1881 bzero(&ro, sizeof(ro));
1882 sin = (struct sockaddr_in *)&ro.ro_dst;
1883 sin->sin_family = AF_INET;
1884 sin->sin_len = sizeof(*sin);
1885 sin->sin_addr = pkt_dst;
1886 rtalloc_ign(&ro, (RTF_PRCLONING | RTF_CLONING));
1887
1888 rt = ro.ro_rt;
1889
1890 if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1891 satosin(rt_key(rt))->sin_addr.s_addr != 0 &&
1892 ipsendredirects && !srcrt && !next_hop) {
1893#define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa))
1894 u_long src = ntohl(ip->ip_src.s_addr);
1895
1896 if (RTA(rt) &&
1897 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1898 if (rt->rt_flags & RTF_GATEWAY)
1899 dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
1900 else
1901 dest = pkt_dst.s_addr;
1902 /* Router requirements says to only send host redirects */
1903 type = ICMP_REDIRECT;
1904 code = ICMP_REDIRECT_HOST;
1905#ifdef DIAGNOSTIC
1906 if (ipprintfs)
1907 printf("redirect (%d) to %lx\n", code, (u_long)dest);
1908#endif
1909 }
1910 }
1911 if (rt)
1912 RTFREE(rt);
1913 }
1914
1915 {
1916 struct m_hdr tag;
1917
1918 if (next_hop) {
1919 /* Pass IPFORWARD info if available */
1920
1921 tag.mh_type = MT_TAG;
1922 tag.mh_flags = PACKET_TAG_IPFORWARD;
1923 tag.mh_data = (caddr_t)next_hop;
1924 tag.mh_next = m;
1925 m = (struct mbuf *)&tag;
1926 }
1927 error = ip_output(m, (struct mbuf *)0, NULL, IP_FORWARDING, 0, NULL);
1928 }
1929 if (error)
1930 ipstat.ips_cantforward++;
1931 else {
1932 ipstat.ips_forward++;
1933 if (type)
1934 ipstat.ips_redirectsent++;
1935 else {
1936 if (mcopy)
1937 m_freem(mcopy);
1938 return;
1939 }
1940 }
1941 if (mcopy == NULL)
1942 return;
1943 destifp = NULL;
1944
1945 switch (error) {
1946
1947 case 0: /* forwarded, but need redirect */
1948 /* type, code set above */
1949 break;
1950
1951 case ENETUNREACH: /* shouldn't happen, checked above */
1952 case EHOSTUNREACH:
1953 case ENETDOWN:
1954 case EHOSTDOWN:
1955 default:
1956 type = ICMP_UNREACH;
1957 code = ICMP_UNREACH_HOST;
1958 break;
1959
1960 case EMSGSIZE:
1961 type = ICMP_UNREACH;
1962 code = ICMP_UNREACH_NEEDFRAG;
1963#if defined(IPSEC) || defined(FAST_IPSEC)
1964 /*
1965 * If the packet is routed over IPsec tunnel, tell the
1966 * originator the tunnel MTU.
1967 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
1968 * XXX quickhack!!!
1969 */
1970 {
1971 struct secpolicy *sp = NULL;
1972 int ipsecerror;
1973 int ipsechdr;
1974 struct route *ro;
1975
1976#ifdef IPSEC
1977 sp = ipsec4_getpolicybyaddr(mcopy,
1978 IPSEC_DIR_OUTBOUND,
1979 IP_FORWARDING,
1980 &ipsecerror);
1981#else /* FAST_IPSEC */
1982 sp = ipsec_getpolicybyaddr(mcopy,
1983 IPSEC_DIR_OUTBOUND,
1984 IP_FORWARDING,
1985 &ipsecerror);
1986#endif
1987 if (sp != NULL) {
1988 /* count IPsec header size */
1989 ipsechdr = ipsec4_hdrsiz(mcopy,
1990 IPSEC_DIR_OUTBOUND,
1991 NULL);
1992
1993 /*
1994 * find the correct route for outer IPv4
1995 * header, compute tunnel MTU.
1996 *
1997 * XXX BUG ALERT
1998 * The "dummyifp" code relies upon the fact
1999 * that icmp_error() touches only ifp->if_mtu.
2000 */
2001 /*XXX*/
2002 destifp = NULL;
2003 if (sp->req != NULL
2004 && sp->req->sav != NULL
2005 && sp->req->sav->sah != NULL) {
2006 ro = &sp->req->sav->sah->sa_route;
2007 if (ro->ro_rt && ro->ro_rt->rt_ifp) {
2008 dummyifp.if_mtu =
2009 ro->ro_rt->rt_ifp->if_mtu;
2010 dummyifp.if_mtu -= ipsechdr;
2011 destifp = &dummyifp;
2012 }
2013 }
2014
2015#ifdef IPSEC
2016 key_freesp(sp);
2017#else /* FAST_IPSEC */
2018 KEY_FREESP(&sp);
2019#endif
2020 ipstat.ips_cantfrag++;
2021 break;
2022 } else
2023#endif /*IPSEC || FAST_IPSEC*/
2024 destifp = ia->ia_ifp;
2025#if defined(IPSEC) || defined(FAST_IPSEC)
2026 }
2027#endif /*IPSEC || FAST_IPSEC*/
2028 ipstat.ips_cantfrag++;
2029 break;
2030
2031 case ENOBUFS:
2032 /*
2033 * A router should not generate ICMP_SOURCEQUENCH as
2034 * required in RFC1812 Requirements for IP Version 4 Routers.
2035 * Source quench could be a big problem under DoS attacks,
2036 * or if the underlying interface is rate-limited.
2037 * Those who need source quench packets may re-enable them
2038 * via the net.inet.ip.sendsourcequench sysctl.
2039 */
2040 if (ip_sendsourcequench == 0) {
2041 m_freem(mcopy);
2042 return;
2043 } else {
2044 type = ICMP_SOURCEQUENCH;
2045 code = 0;
2046 }
2047 break;
2048
2049 case EACCES: /* ipfw denied packet */
2050 m_freem(mcopy);
2051 return;
2052 }
2053 icmp_error(mcopy, type, code, dest, destifp);
2054}
2055
2056void
2057ip_savecontrol(inp, mp, ip, m)
2058 register struct inpcb *inp;
2059 register struct mbuf **mp;
2060 register struct ip *ip;
2061 register struct mbuf *m;
2062{
2063 if (inp->inp_socket->so_options & SO_TIMESTAMP) {
2064 struct timeval tv;
2065
2066 microtime(&tv);
2067 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
2068 SCM_TIMESTAMP, SOL_SOCKET);
2069 if (*mp)
2070 mp = &(*mp)->m_next;
2071 }
2072 if (inp->inp_flags & INP_RECVDSTADDR) {
2073 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
2074 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
2075 if (*mp)
2076 mp = &(*mp)->m_next;
2077 }
2078 if (inp->inp_flags & INP_RECVTTL) {
2079 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl,
2080 sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
2081 if (*mp)
2082 mp = &(*mp)->m_next;
2083 }
2084#ifdef notyet
2085 /* XXX
2086 * Moving these out of udp_input() made them even more broken
2087 * than they already were.
2088 */
2089 /* options were tossed already */
2090 if (inp->inp_flags & INP_RECVOPTS) {
2091 *mp = sbcreatecontrol((caddr_t) opts_deleted_above,
2092 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
2093 if (*mp)
2094 mp = &(*mp)->m_next;
2095 }
2096 /* ip_srcroute doesn't do what we want here, need to fix */
2097 if (inp->inp_flags & INP_RECVRETOPTS) {
2098 *mp = sbcreatecontrol((caddr_t) ip_srcroute(),
2099 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
2100 if (*mp)
2101 mp = &(*mp)->m_next;
2102 }
2103#endif
2104 if (inp->inp_flags & INP_RECVIF) {
2105 struct ifnet *ifp;
2106 struct sdlbuf {
2107 struct sockaddr_dl sdl;
2108 u_char pad[32];
2109 } sdlbuf;
2110 struct sockaddr_dl *sdp;
2111 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
2112
2113 if (((ifp = m->m_pkthdr.rcvif))
2114 && ( ifp->if_index && (ifp->if_index <= if_index))) {
2115 sdp = (struct sockaddr_dl *)
2116 (ifaddr_byindex(ifp->if_index)->ifa_addr);
2117 /*
2118 * Change our mind and don't try copy.
2119 */
2120 if ((sdp->sdl_family != AF_LINK)
2121 || (sdp->sdl_len > sizeof(sdlbuf))) {
2122 goto makedummy;
2123 }
2124 bcopy(sdp, sdl2, sdp->sdl_len);
2125 } else {
2126makedummy:
2127 sdl2->sdl_len
2128 = offsetof(struct sockaddr_dl, sdl_data[0]);
2129 sdl2->sdl_family = AF_LINK;
2130 sdl2->sdl_index = 0;
2131 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
2132 }
2133 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len,
2134 IP_RECVIF, IPPROTO_IP);
2135 if (*mp)
2136 mp = &(*mp)->m_next;
2137 }
2138}
2139
2140/*
2141 * XXX these routines are called from the upper part of the kernel.
2142 * They need to be locked when we remove Giant.
2143 *
2144 * They could also be moved to ip_mroute.c, since all the RSVP
2145 * handling is done there already.
2146 */
2147static int ip_rsvp_on;
2148struct socket *ip_rsvpd;
2149int
2150ip_rsvp_init(struct socket *so)
2151{
2152 if (so->so_type != SOCK_RAW ||
2153 so->so_proto->pr_protocol != IPPROTO_RSVP)
2154 return EOPNOTSUPP;
2155
2156 if (ip_rsvpd != NULL)
2157 return EADDRINUSE;
2158
2159 ip_rsvpd = so;
2160 /*
2161 * This may seem silly, but we need to be sure we don't over-increment
2162 * the RSVP counter, in case something slips up.
2163 */
2164 if (!ip_rsvp_on) {
2165 ip_rsvp_on = 1;
2166 rsvp_on++;
2167 }
2168
2169 return 0;
2170}
2171
2172int
2173ip_rsvp_done(void)
2174{
2175 ip_rsvpd = NULL;
2176 /*
2177 * This may seem silly, but we need to be sure we don't over-decrement
2178 * the RSVP counter, in case something slips up.
2179 */
2180 if (ip_rsvp_on) {
2181 ip_rsvp_on = 0;
2182 rsvp_on--;
2183 }
2184 return 0;
2185}
2186
2187void
2188rsvp_input(struct mbuf *m, int off) /* XXX must fixup manually */
2189{
2190 if (rsvp_input_p) { /* call the real one if loaded */
2191 rsvp_input_p(m, off);
2192 return;
2193 }
2194
2195 /* Can still get packets with rsvp_on = 0 if there is a local member
2196 * of the group to which the RSVP packet is addressed. But in this
2197 * case we want to throw the packet away.
2198 */
2199
2200 if (!rsvp_on) {
2201 m_freem(m);
2202 return;
2203 }
2204
2205 if (ip_rsvpd != NULL) {
2206 rip_input(m, off);
2207 return;
2208 }
2209 /* Drop the packet */
2210 m_freem(m);
2211}
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: head/sys/netinet/ip_input.c 122723 2003-11-15 01:45:56Z 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
93int rsvp_on = 0;
94
95int ipforwarding = 0;
96SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW,
97 &ipforwarding, 0, "Enable IP forwarding between interfaces");
98
99static int ipsendredirects = 1; /* XXX */
100SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW,
101 &ipsendredirects, 0, "Enable sending IP redirects");
102
103int ip_defttl = IPDEFTTL;
104SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW,
105 &ip_defttl, 0, "Maximum TTL on IP packets");
106
107static int ip_dosourceroute = 0;
108SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW,
109 &ip_dosourceroute, 0, "Enable forwarding source routed IP packets");
110
111static int ip_acceptsourceroute = 0;
112SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute,
113 CTLFLAG_RW, &ip_acceptsourceroute, 0,
114 "Enable accepting source routed IP packets");
115
116static int ip_keepfaith = 0;
117SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW,
118 &ip_keepfaith, 0,
119 "Enable packet capture for FAITH IPv4->IPv6 translater daemon");
120
121static int nipq = 0; /* total # of reass queues */
122static int maxnipq;
123SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragpackets, CTLFLAG_RW,
124 &maxnipq, 0,
125 "Maximum number of IPv4 fragment reassembly queue entries");
126
127static int maxfragsperpacket;
128SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW,
129 &maxfragsperpacket, 0,
130 "Maximum number of IPv4 fragments allowed per packet");
131
132static int ip_sendsourcequench = 0;
133SYSCTL_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 */
150static int ip_checkinterface = 1;
151SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW,
152 &ip_checkinterface, 0, "Verify packet arrives on correct interface");
153
154#ifdef DIAGNOSTIC
155static int ipprintfs = 0;
156#endif
157#ifdef PFIL_HOOKS
158struct pfil_head inet_pfil_hook;
159#endif
160
161static struct ifqueue ipintrq;
162static int ipqmaxlen = IFQ_MAXLEN;
163
164extern struct domain inetdomain;
165extern struct protosw inetsw[];
166u_char ip_protox[IPPROTO_MAX];
167struct in_ifaddrhead in_ifaddrhead; /* first inet address */
168struct in_ifaddrhashhead *in_ifaddrhashtbl; /* inet addr hash table */
169u_long in_ifaddrhmask; /* mask for hash table */
170
171SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW,
172 &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue");
173SYSCTL_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
176struct ipstat ipstat;
177SYSCTL_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
187static TAILQ_HEAD(ipqhead, ipq) ipq[IPREASS_NHASH];
188struct 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
196SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
197 &ip_mtu, 0, "Default MTU");
198#endif
199
200#ifdef IPSTEALTH
201int ipstealth = 0;
202SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW,
203 &ipstealth, 0, "");
204#endif
205
206
207/* Firewall hooks */
208ip_fw_chk_t *ip_fw_chk_ptr;
209int fw_enable = 1 ;
210int fw_one_pass = 1;
211
212/* Dummynet hooks */
213ip_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 */
228static int ip_nhops = 0;
229static 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
236static void save_rte(u_char *, struct in_addr);
237static int ip_dooptions(struct mbuf *m, int,
238 struct sockaddr_in *next_hop);
239static void ip_forward(struct mbuf *m, int srcrt,
240 struct sockaddr_in *next_hop);
241static void ip_freef(struct ipqhead *, struct ipq *);
242static 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 */
249void
250ip_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 */
295void
296ip_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) {
448tooshort:
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
486iphack:
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 }
546pass:
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
743ours:
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 }
819found:
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;
956DPRINTF(("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
979 (*inetsw[ip_protox[ip->ip_p]].pr_input)(
980 (struct mbuf *)&tag, hlen);
981 } else
982 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen);
983 NET_DROP_GIANT();
984 return;
985bad:
986 m_freem(m);
987}
988
989/*
990 * Take incoming datagram fragment and try to reassemble it into
991 * whole datagram. If a chain for reassembly of this datagram already
992 * exists, then it is given as fp; otherwise have to make a chain.
993 *
994 * When IPDIVERT enabled, keep additional state with each packet that
995 * tells us if we need to divert or tee the packet we're building.
996 * In particular, *divinfo includes the port and TEE flag,
997 * *divert_rule is the number of the matching rule.
998 */
999
1000static struct mbuf *
1001ip_reass(struct mbuf *m, struct ipqhead *head, struct ipq *fp,
1002 u_int32_t *divinfo, u_int16_t *divert_rule)
1003{
1004 struct ip *ip = mtod(m, struct ip *);
1005 register struct mbuf *p, *q, *nq;
1006 struct mbuf *t;
1007 int hlen = ip->ip_hl << 2;
1008 int i, next;
1009 u_int8_t ecn, ecn0;
1010
1011 IPQ_LOCK_ASSERT();
1012
1013 /*
1014 * Presence of header sizes in mbufs
1015 * would confuse code below.
1016 */
1017 m->m_data += hlen;
1018 m->m_len -= hlen;
1019
1020 /*
1021 * If first fragment to arrive, create a reassembly queue.
1022 */
1023 if (fp == NULL) {
1024 if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL)
1025 goto dropfrag;
1026 fp = mtod(t, struct ipq *);
1027#ifdef MAC
1028 if (mac_init_ipq(fp, M_NOWAIT) != 0) {
1029 m_free(t);
1030 goto dropfrag;
1031 }
1032 mac_create_ipq(m, fp);
1033#endif
1034 TAILQ_INSERT_HEAD(head, fp, ipq_list);
1035 nipq++;
1036 fp->ipq_nfrags = 1;
1037 fp->ipq_ttl = IPFRAGTTL;
1038 fp->ipq_p = ip->ip_p;
1039 fp->ipq_id = ip->ip_id;
1040 fp->ipq_src = ip->ip_src;
1041 fp->ipq_dst = ip->ip_dst;
1042 fp->ipq_frags = m;
1043 m->m_nextpkt = NULL;
1044#ifdef IPDIVERT
1045 fp->ipq_div_info = 0;
1046 fp->ipq_div_cookie = 0;
1047#endif
1048 goto inserted;
1049 } else {
1050 fp->ipq_nfrags++;
1051#ifdef MAC
1052 mac_update_ipq(m, fp);
1053#endif
1054 }
1055
1056#define GETIP(m) ((struct ip*)((m)->m_pkthdr.header))
1057
1058 /*
1059 * Handle ECN by comparing this segment with the first one;
1060 * if CE is set, do not lose CE.
1061 * drop if CE and not-ECT are mixed for the same packet.
1062 */
1063 ecn = ip->ip_tos & IPTOS_ECN_MASK;
1064 ecn0 = GETIP(fp->ipq_frags)->ip_tos & IPTOS_ECN_MASK;
1065 if (ecn == IPTOS_ECN_CE) {
1066 if (ecn0 == IPTOS_ECN_NOTECT)
1067 goto dropfrag;
1068 if (ecn0 != IPTOS_ECN_CE)
1069 GETIP(fp->ipq_frags)->ip_tos |= IPTOS_ECN_CE;
1070 }
1071 if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT)
1072 goto dropfrag;
1073
1074 /*
1075 * Find a segment which begins after this one does.
1076 */
1077 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt)
1078 if (GETIP(q)->ip_off > ip->ip_off)
1079 break;
1080
1081 /*
1082 * If there is a preceding segment, it may provide some of
1083 * our data already. If so, drop the data from the incoming
1084 * segment. If it provides all of our data, drop us, otherwise
1085 * stick new segment in the proper place.
1086 *
1087 * If some of the data is dropped from the the preceding
1088 * segment, then it's checksum is invalidated.
1089 */
1090 if (p) {
1091 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off;
1092 if (i > 0) {
1093 if (i >= ip->ip_len)
1094 goto dropfrag;
1095 m_adj(m, i);
1096 m->m_pkthdr.csum_flags = 0;
1097 ip->ip_off += i;
1098 ip->ip_len -= i;
1099 }
1100 m->m_nextpkt = p->m_nextpkt;
1101 p->m_nextpkt = m;
1102 } else {
1103 m->m_nextpkt = fp->ipq_frags;
1104 fp->ipq_frags = m;
1105 }
1106
1107 /*
1108 * While we overlap succeeding segments trim them or,
1109 * if they are completely covered, dequeue them.
1110 */
1111 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off;
1112 q = nq) {
1113 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off;
1114 if (i < GETIP(q)->ip_len) {
1115 GETIP(q)->ip_len -= i;
1116 GETIP(q)->ip_off += i;
1117 m_adj(q, i);
1118 q->m_pkthdr.csum_flags = 0;
1119 break;
1120 }
1121 nq = q->m_nextpkt;
1122 m->m_nextpkt = nq;
1123 ipstat.ips_fragdropped++;
1124 fp->ipq_nfrags--;
1125 m_freem(q);
1126 }
1127
1128inserted:
1129
1130#ifdef IPDIVERT
1131 /*
1132 * Transfer firewall instructions to the fragment structure.
1133 * Only trust info in the fragment at offset 0.
1134 */
1135 if (ip->ip_off == 0) {
1136 fp->ipq_div_info = *divinfo;
1137 fp->ipq_div_cookie = *divert_rule;
1138 }
1139 *divinfo = 0;
1140 *divert_rule = 0;
1141#endif
1142
1143 /*
1144 * Check for complete reassembly and perform frag per packet
1145 * limiting.
1146 *
1147 * Frag limiting is performed here so that the nth frag has
1148 * a chance to complete the packet before we drop the packet.
1149 * As a result, n+1 frags are actually allowed per packet, but
1150 * only n will ever be stored. (n = maxfragsperpacket.)
1151 *
1152 */
1153 next = 0;
1154 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
1155 if (GETIP(q)->ip_off != next) {
1156 if (fp->ipq_nfrags > maxfragsperpacket) {
1157 ipstat.ips_fragdropped += fp->ipq_nfrags;
1158 ip_freef(head, fp);
1159 }
1160 return (0);
1161 }
1162 next += GETIP(q)->ip_len;
1163 }
1164 /* Make sure the last packet didn't have the IP_MF flag */
1165 if (p->m_flags & M_FRAG) {
1166 if (fp->ipq_nfrags > maxfragsperpacket) {
1167 ipstat.ips_fragdropped += fp->ipq_nfrags;
1168 ip_freef(head, fp);
1169 }
1170 return (0);
1171 }
1172
1173 /*
1174 * Reassembly is complete. Make sure the packet is a sane size.
1175 */
1176 q = fp->ipq_frags;
1177 ip = GETIP(q);
1178 if (next + (ip->ip_hl << 2) > IP_MAXPACKET) {
1179 ipstat.ips_toolong++;
1180 ipstat.ips_fragdropped += fp->ipq_nfrags;
1181 ip_freef(head, fp);
1182 return (0);
1183 }
1184
1185 /*
1186 * Concatenate fragments.
1187 */
1188 m = q;
1189 t = m->m_next;
1190 m->m_next = 0;
1191 m_cat(m, t);
1192 nq = q->m_nextpkt;
1193 q->m_nextpkt = 0;
1194 for (q = nq; q != NULL; q = nq) {
1195 nq = q->m_nextpkt;
1196 q->m_nextpkt = NULL;
1197 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
1198 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
1199 m_cat(m, q);
1200 }
1201#ifdef MAC
1202 mac_create_datagram_from_ipq(fp, m);
1203 mac_destroy_ipq(fp);
1204#endif
1205
1206#ifdef IPDIVERT
1207 /*
1208 * Extract firewall instructions from the fragment structure.
1209 */
1210 *divinfo = fp->ipq_div_info;
1211 *divert_rule = fp->ipq_div_cookie;
1212#endif
1213
1214 /*
1215 * Create header for new ip packet by
1216 * modifying header of first packet;
1217 * dequeue and discard fragment reassembly header.
1218 * Make header visible.
1219 */
1220 ip->ip_len = next;
1221 ip->ip_src = fp->ipq_src;
1222 ip->ip_dst = fp->ipq_dst;
1223 TAILQ_REMOVE(head, fp, ipq_list);
1224 nipq--;
1225 (void) m_free(dtom(fp));
1226 m->m_len += (ip->ip_hl << 2);
1227 m->m_data -= (ip->ip_hl << 2);
1228 /* some debugging cruft by sklower, below, will go away soon */
1229 if (m->m_flags & M_PKTHDR) /* XXX this should be done elsewhere */
1230 m_fixhdr(m);
1231 return (m);
1232
1233dropfrag:
1234#ifdef IPDIVERT
1235 *divinfo = 0;
1236 *divert_rule = 0;
1237#endif
1238 ipstat.ips_fragdropped++;
1239 if (fp != NULL)
1240 fp->ipq_nfrags--;
1241 m_freem(m);
1242 return (0);
1243
1244#undef GETIP
1245}
1246
1247/*
1248 * Free a fragment reassembly header and all
1249 * associated datagrams.
1250 */
1251static void
1252ip_freef(fhp, fp)
1253 struct ipqhead *fhp;
1254 struct ipq *fp;
1255{
1256 register struct mbuf *q;
1257
1258 IPQ_LOCK_ASSERT();
1259
1260 while (fp->ipq_frags) {
1261 q = fp->ipq_frags;
1262 fp->ipq_frags = q->m_nextpkt;
1263 m_freem(q);
1264 }
1265 TAILQ_REMOVE(fhp, fp, ipq_list);
1266 (void) m_free(dtom(fp));
1267 nipq--;
1268}
1269
1270/*
1271 * IP timer processing;
1272 * if a timer expires on a reassembly
1273 * queue, discard it.
1274 */
1275void
1276ip_slowtimo()
1277{
1278 register struct ipq *fp;
1279 int s = splnet();
1280 int i;
1281
1282 IPQ_LOCK();
1283 for (i = 0; i < IPREASS_NHASH; i++) {
1284 for(fp = TAILQ_FIRST(&ipq[i]); fp;) {
1285 struct ipq *fpp;
1286
1287 fpp = fp;
1288 fp = TAILQ_NEXT(fp, ipq_list);
1289 if(--fpp->ipq_ttl == 0) {
1290 ipstat.ips_fragtimeout += fpp->ipq_nfrags;
1291 ip_freef(&ipq[i], fpp);
1292 }
1293 }
1294 }
1295 /*
1296 * If we are over the maximum number of fragments
1297 * (due to the limit being lowered), drain off
1298 * enough to get down to the new limit.
1299 */
1300 if (maxnipq >= 0 && nipq > maxnipq) {
1301 for (i = 0; i < IPREASS_NHASH; i++) {
1302 while (nipq > maxnipq && !TAILQ_EMPTY(&ipq[i])) {
1303 ipstat.ips_fragdropped +=
1304 TAILQ_FIRST(&ipq[i])->ipq_nfrags;
1305 ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i]));
1306 }
1307 }
1308 }
1309 IPQ_UNLOCK();
1310 splx(s);
1311}
1312
1313/*
1314 * Drain off all datagram fragments.
1315 */
1316void
1317ip_drain()
1318{
1319 int i;
1320
1321 IPQ_LOCK();
1322 for (i = 0; i < IPREASS_NHASH; i++) {
1323 while(!TAILQ_EMPTY(&ipq[i])) {
1324 ipstat.ips_fragdropped +=
1325 TAILQ_FIRST(&ipq[i])->ipq_nfrags;
1326 ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i]));
1327 }
1328 }
1329 IPQ_UNLOCK();
1330 in_rtqdrain();
1331}
1332
1333/*
1334 * Do option processing on a datagram,
1335 * possibly discarding it if bad options are encountered,
1336 * or forwarding it if source-routed.
1337 * The pass argument is used when operating in the IPSTEALTH
1338 * mode to tell what options to process:
1339 * [LS]SRR (pass 0) or the others (pass 1).
1340 * The reason for as many as two passes is that when doing IPSTEALTH,
1341 * non-routing options should be processed only if the packet is for us.
1342 * Returns 1 if packet has been forwarded/freed,
1343 * 0 if the packet should be processed further.
1344 */
1345static int
1346ip_dooptions(struct mbuf *m, int pass, struct sockaddr_in *next_hop)
1347{
1348 struct ip *ip = mtod(m, struct ip *);
1349 u_char *cp;
1350 struct in_ifaddr *ia;
1351 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0;
1352 struct in_addr *sin, dst;
1353 n_time ntime;
1354 struct sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET };
1355
1356 dst = ip->ip_dst;
1357 cp = (u_char *)(ip + 1);
1358 cnt = (ip->ip_hl << 2) - sizeof (struct ip);
1359 for (; cnt > 0; cnt -= optlen, cp += optlen) {
1360 opt = cp[IPOPT_OPTVAL];
1361 if (opt == IPOPT_EOL)
1362 break;
1363 if (opt == IPOPT_NOP)
1364 optlen = 1;
1365 else {
1366 if (cnt < IPOPT_OLEN + sizeof(*cp)) {
1367 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1368 goto bad;
1369 }
1370 optlen = cp[IPOPT_OLEN];
1371 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
1372 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1373 goto bad;
1374 }
1375 }
1376 switch (opt) {
1377
1378 default:
1379 break;
1380
1381 /*
1382 * Source routing with record.
1383 * Find interface with current destination address.
1384 * If none on this machine then drop if strictly routed,
1385 * or do nothing if loosely routed.
1386 * Record interface address and bring up next address
1387 * component. If strictly routed make sure next
1388 * address is on directly accessible net.
1389 */
1390 case IPOPT_LSRR:
1391 case IPOPT_SSRR:
1392#ifdef IPSTEALTH
1393 if (ipstealth && pass > 0)
1394 break;
1395#endif
1396 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1397 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1398 goto bad;
1399 }
1400 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1401 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1402 goto bad;
1403 }
1404 ipaddr.sin_addr = ip->ip_dst;
1405 ia = (struct in_ifaddr *)
1406 ifa_ifwithaddr((struct sockaddr *)&ipaddr);
1407 if (ia == 0) {
1408 if (opt == IPOPT_SSRR) {
1409 type = ICMP_UNREACH;
1410 code = ICMP_UNREACH_SRCFAIL;
1411 goto bad;
1412 }
1413 if (!ip_dosourceroute)
1414 goto nosourcerouting;
1415 /*
1416 * Loose routing, and not at next destination
1417 * yet; nothing to do except forward.
1418 */
1419 break;
1420 }
1421 off--; /* 0 origin */
1422 if (off > optlen - (int)sizeof(struct in_addr)) {
1423 /*
1424 * End of source route. Should be for us.
1425 */
1426 if (!ip_acceptsourceroute)
1427 goto nosourcerouting;
1428 save_rte(cp, ip->ip_src);
1429 break;
1430 }
1431#ifdef IPSTEALTH
1432 if (ipstealth)
1433 goto dropit;
1434#endif
1435 if (!ip_dosourceroute) {
1436 if (ipforwarding) {
1437 char buf[16]; /* aaa.bbb.ccc.ddd\0 */
1438 /*
1439 * Acting as a router, so generate ICMP
1440 */
1441nosourcerouting:
1442 strcpy(buf, inet_ntoa(ip->ip_dst));
1443 log(LOG_WARNING,
1444 "attempted source route from %s to %s\n",
1445 inet_ntoa(ip->ip_src), buf);
1446 type = ICMP_UNREACH;
1447 code = ICMP_UNREACH_SRCFAIL;
1448 goto bad;
1449 } else {
1450 /*
1451 * Not acting as a router, so silently drop.
1452 */
1453#ifdef IPSTEALTH
1454dropit:
1455#endif
1456 ipstat.ips_cantforward++;
1457 m_freem(m);
1458 return (1);
1459 }
1460 }
1461
1462 /*
1463 * locate outgoing interface
1464 */
1465 (void)memcpy(&ipaddr.sin_addr, cp + off,
1466 sizeof(ipaddr.sin_addr));
1467
1468 if (opt == IPOPT_SSRR) {
1469#define INA struct in_ifaddr *
1470#define SA struct sockaddr *
1471 if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == 0)
1472 ia = (INA)ifa_ifwithnet((SA)&ipaddr);
1473 } else
1474 ia = ip_rtaddr(ipaddr.sin_addr);
1475 if (ia == 0) {
1476 type = ICMP_UNREACH;
1477 code = ICMP_UNREACH_SRCFAIL;
1478 goto bad;
1479 }
1480 ip->ip_dst = ipaddr.sin_addr;
1481 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
1482 sizeof(struct in_addr));
1483 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1484 /*
1485 * Let ip_intr's mcast routing check handle mcast pkts
1486 */
1487 forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr));
1488 break;
1489
1490 case IPOPT_RR:
1491#ifdef IPSTEALTH
1492 if (ipstealth && pass == 0)
1493 break;
1494#endif
1495 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1496 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1497 goto bad;
1498 }
1499 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1500 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1501 goto bad;
1502 }
1503 /*
1504 * If no space remains, ignore.
1505 */
1506 off--; /* 0 origin */
1507 if (off > optlen - (int)sizeof(struct in_addr))
1508 break;
1509 (void)memcpy(&ipaddr.sin_addr, &ip->ip_dst,
1510 sizeof(ipaddr.sin_addr));
1511 /*
1512 * locate outgoing interface; if we're the destination,
1513 * use the incoming interface (should be same).
1514 */
1515 if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == 0 &&
1516 (ia = ip_rtaddr(ipaddr.sin_addr)) == 0) {
1517 type = ICMP_UNREACH;
1518 code = ICMP_UNREACH_HOST;
1519 goto bad;
1520 }
1521 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
1522 sizeof(struct in_addr));
1523 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1524 break;
1525
1526 case IPOPT_TS:
1527#ifdef IPSTEALTH
1528 if (ipstealth && pass == 0)
1529 break;
1530#endif
1531 code = cp - (u_char *)ip;
1532 if (optlen < 4 || optlen > 40) {
1533 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1534 goto bad;
1535 }
1536 if ((off = cp[IPOPT_OFFSET]) < 5) {
1537 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1538 goto bad;
1539 }
1540 if (off > optlen - (int)sizeof(int32_t)) {
1541 cp[IPOPT_OFFSET + 1] += (1 << 4);
1542 if ((cp[IPOPT_OFFSET + 1] & 0xf0) == 0) {
1543 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1544 goto bad;
1545 }
1546 break;
1547 }
1548 off--; /* 0 origin */
1549 sin = (struct in_addr *)(cp + off);
1550 switch (cp[IPOPT_OFFSET + 1] & 0x0f) {
1551
1552 case IPOPT_TS_TSONLY:
1553 break;
1554
1555 case IPOPT_TS_TSANDADDR:
1556 if (off + sizeof(n_time) +
1557 sizeof(struct in_addr) > optlen) {
1558 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1559 goto bad;
1560 }
1561 ipaddr.sin_addr = dst;
1562 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr,
1563 m->m_pkthdr.rcvif);
1564 if (ia == 0)
1565 continue;
1566 (void)memcpy(sin, &IA_SIN(ia)->sin_addr,
1567 sizeof(struct in_addr));
1568 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1569 off += sizeof(struct in_addr);
1570 break;
1571
1572 case IPOPT_TS_PRESPEC:
1573 if (off + sizeof(n_time) +
1574 sizeof(struct in_addr) > optlen) {
1575 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1576 goto bad;
1577 }
1578 (void)memcpy(&ipaddr.sin_addr, sin,
1579 sizeof(struct in_addr));
1580 if (ifa_ifwithaddr((SA)&ipaddr) == 0)
1581 continue;
1582 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1583 off += sizeof(struct in_addr);
1584 break;
1585
1586 default:
1587 code = &cp[IPOPT_OFFSET + 1] - (u_char *)ip;
1588 goto bad;
1589 }
1590 ntime = iptime();
1591 (void)memcpy(cp + off, &ntime, sizeof(n_time));
1592 cp[IPOPT_OFFSET] += sizeof(n_time);
1593 }
1594 }
1595 if (forward && ipforwarding) {
1596 ip_forward(m, 1, next_hop);
1597 return (1);
1598 }
1599 return (0);
1600bad:
1601 icmp_error(m, type, code, 0, 0);
1602 ipstat.ips_badoptions++;
1603 return (1);
1604}
1605
1606/*
1607 * Given address of next destination (final or next hop),
1608 * return internet address info of interface to be used to get there.
1609 */
1610struct in_ifaddr *
1611ip_rtaddr(dst)
1612 struct in_addr dst;
1613{
1614 struct sockaddr_in *sin;
1615 struct in_ifaddr *ifa;
1616 struct route ro;
1617
1618 bzero(&ro, sizeof(ro));
1619 sin = (struct sockaddr_in *)&ro.ro_dst;
1620 sin->sin_family = AF_INET;
1621 sin->sin_len = sizeof(*sin);
1622 sin->sin_addr = dst;
1623 rtalloc_ign(&ro, (RTF_PRCLONING | RTF_CLONING));
1624
1625 if (ro.ro_rt == 0)
1626 return ((struct in_ifaddr *)0);
1627
1628 ifa = ifatoia(ro.ro_rt->rt_ifa);
1629 RTFREE(ro.ro_rt);
1630 return ifa;
1631}
1632
1633/*
1634 * Save incoming source route for use in replies,
1635 * to be picked up later by ip_srcroute if the receiver is interested.
1636 */
1637static void
1638save_rte(option, dst)
1639 u_char *option;
1640 struct in_addr dst;
1641{
1642 unsigned olen;
1643
1644 olen = option[IPOPT_OLEN];
1645#ifdef DIAGNOSTIC
1646 if (ipprintfs)
1647 printf("save_rte: olen %d\n", olen);
1648#endif
1649 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst)))
1650 return;
1651 bcopy(option, ip_srcrt.srcopt, olen);
1652 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1653 ip_srcrt.dst = dst;
1654}
1655
1656/*
1657 * Retrieve incoming source route for use in replies,
1658 * in the same form used by setsockopt.
1659 * The first hop is placed before the options, will be removed later.
1660 */
1661struct mbuf *
1662ip_srcroute()
1663{
1664 register struct in_addr *p, *q;
1665 register struct mbuf *m;
1666
1667 if (ip_nhops == 0)
1668 return ((struct mbuf *)0);
1669 m = m_get(M_DONTWAIT, MT_HEADER);
1670 if (m == 0)
1671 return ((struct mbuf *)0);
1672
1673#define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt))
1674
1675 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
1676 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) +
1677 OPTSIZ;
1678#ifdef DIAGNOSTIC
1679 if (ipprintfs)
1680 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len);
1681#endif
1682
1683 /*
1684 * First save first hop for return route
1685 */
1686 p = &ip_srcrt.route[ip_nhops - 1];
1687 *(mtod(m, struct in_addr *)) = *p--;
1688#ifdef DIAGNOSTIC
1689 if (ipprintfs)
1690 printf(" hops %lx", (u_long)ntohl(mtod(m, struct in_addr *)->s_addr));
1691#endif
1692
1693 /*
1694 * Copy option fields and padding (nop) to mbuf.
1695 */
1696 ip_srcrt.nop = IPOPT_NOP;
1697 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
1698 (void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr),
1699 &ip_srcrt.nop, OPTSIZ);
1700 q = (struct in_addr *)(mtod(m, caddr_t) +
1701 sizeof(struct in_addr) + OPTSIZ);
1702#undef OPTSIZ
1703 /*
1704 * Record return path as an IP source route,
1705 * reversing the path (pointers are now aligned).
1706 */
1707 while (p >= ip_srcrt.route) {
1708#ifdef DIAGNOSTIC
1709 if (ipprintfs)
1710 printf(" %lx", (u_long)ntohl(q->s_addr));
1711#endif
1712 *q++ = *p--;
1713 }
1714 /*
1715 * Last hop goes to final destination.
1716 */
1717 *q = ip_srcrt.dst;
1718#ifdef DIAGNOSTIC
1719 if (ipprintfs)
1720 printf(" %lx\n", (u_long)ntohl(q->s_addr));
1721#endif
1722 return (m);
1723}
1724
1725/*
1726 * Strip out IP options, at higher
1727 * level protocol in the kernel.
1728 * Second argument is buffer to which options
1729 * will be moved, and return value is their length.
1730 * XXX should be deleted; last arg currently ignored.
1731 */
1732void
1733ip_stripoptions(m, mopt)
1734 register struct mbuf *m;
1735 struct mbuf *mopt;
1736{
1737 register int i;
1738 struct ip *ip = mtod(m, struct ip *);
1739 register caddr_t opts;
1740 int olen;
1741
1742 olen = (ip->ip_hl << 2) - sizeof (struct ip);
1743 opts = (caddr_t)(ip + 1);
1744 i = m->m_len - (sizeof (struct ip) + olen);
1745 bcopy(opts + olen, opts, (unsigned)i);
1746 m->m_len -= olen;
1747 if (m->m_flags & M_PKTHDR)
1748 m->m_pkthdr.len -= olen;
1749 ip->ip_v = IPVERSION;
1750 ip->ip_hl = sizeof(struct ip) >> 2;
1751}
1752
1753u_char inetctlerrmap[PRC_NCMDS] = {
1754 0, 0, 0, 0,
1755 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
1756 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
1757 EMSGSIZE, EHOSTUNREACH, 0, 0,
1758 0, 0, EHOSTUNREACH, 0,
1759 ENOPROTOOPT, ECONNREFUSED
1760};
1761
1762/*
1763 * Forward a packet. If some error occurs return the sender
1764 * an icmp packet. Note we can't always generate a meaningful
1765 * icmp message because icmp doesn't have a large enough repertoire
1766 * of codes and types.
1767 *
1768 * If not forwarding, just drop the packet. This could be confusing
1769 * if ipforwarding was zero but some routing protocol was advancing
1770 * us as a gateway to somewhere. However, we must let the routing
1771 * protocol deal with that.
1772 *
1773 * The srcrt parameter indicates whether the packet is being forwarded
1774 * via a source route.
1775 */
1776static void
1777ip_forward(struct mbuf *m, int srcrt, struct sockaddr_in *next_hop)
1778{
1779 struct ip *ip = mtod(m, struct ip *);
1780 struct in_ifaddr *ia;
1781 int error, type = 0, code = 0;
1782 struct mbuf *mcopy;
1783 n_long dest;
1784 struct in_addr pkt_dst;
1785 struct ifnet *destifp;
1786#if defined(IPSEC) || defined(FAST_IPSEC)
1787 struct ifnet dummyifp;
1788#endif
1789
1790 /*
1791 * Cache the destination address of the packet; this may be
1792 * changed by use of 'ipfw fwd'.
1793 */
1794 pkt_dst = next_hop ? next_hop->sin_addr : ip->ip_dst;
1795
1796#ifdef DIAGNOSTIC
1797 if (ipprintfs)
1798 printf("forward: src %lx dst %lx ttl %x\n",
1799 (u_long)ip->ip_src.s_addr, (u_long)pkt_dst.s_addr,
1800 ip->ip_ttl);
1801#endif
1802
1803
1804 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(pkt_dst) == 0) {
1805 ipstat.ips_cantforward++;
1806 m_freem(m);
1807 return;
1808 }
1809#ifdef IPSTEALTH
1810 if (!ipstealth) {
1811#endif
1812 if (ip->ip_ttl <= IPTTLDEC) {
1813 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS,
1814 0, 0);
1815 return;
1816 }
1817#ifdef IPSTEALTH
1818 }
1819#endif
1820
1821 if ((ia = ip_rtaddr(pkt_dst)) == 0) {
1822 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
1823 return;
1824 }
1825
1826 /*
1827 * Save the IP header and at most 8 bytes of the payload,
1828 * in case we need to generate an ICMP message to the src.
1829 *
1830 * XXX this can be optimized a lot by saving the data in a local
1831 * buffer on the stack (72 bytes at most), and only allocating the
1832 * mbuf if really necessary. The vast majority of the packets
1833 * are forwarded without having to send an ICMP back (either
1834 * because unnecessary, or because rate limited), so we are
1835 * really we are wasting a lot of work here.
1836 *
1837 * We don't use m_copy() because it might return a reference
1838 * to a shared cluster. Both this function and ip_output()
1839 * assume exclusive access to the IP header in `m', so any
1840 * data in a cluster may change before we reach icmp_error().
1841 */
1842 MGET(mcopy, M_DONTWAIT, m->m_type);
1843 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_DONTWAIT)) {
1844 /*
1845 * It's probably ok if the pkthdr dup fails (because
1846 * the deep copy of the tag chain failed), but for now
1847 * be conservative and just discard the copy since
1848 * code below may some day want the tags.
1849 */
1850 m_free(mcopy);
1851 mcopy = NULL;
1852 }
1853 if (mcopy != NULL) {
1854 mcopy->m_len = imin((ip->ip_hl << 2) + 8,
1855 (int)ip->ip_len);
1856 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1857 }
1858
1859#ifdef IPSTEALTH
1860 if (!ipstealth) {
1861#endif
1862 ip->ip_ttl -= IPTTLDEC;
1863#ifdef IPSTEALTH
1864 }
1865#endif
1866
1867 /*
1868 * If forwarding packet using same interface that it came in on,
1869 * perhaps should send a redirect to sender to shortcut a hop.
1870 * Only send redirect if source is sending directly to us,
1871 * and if packet was not source routed (or has any options).
1872 * Also, don't send redirect if forwarding using a default route
1873 * or a route modified by a redirect.
1874 */
1875 dest = 0;
1876 if (ipsendredirects && ia->ia_ifp == m->m_pkthdr.rcvif) {
1877 struct sockaddr_in *sin;
1878 struct route ro;
1879 struct rtentry *rt;
1880
1881 bzero(&ro, sizeof(ro));
1882 sin = (struct sockaddr_in *)&ro.ro_dst;
1883 sin->sin_family = AF_INET;
1884 sin->sin_len = sizeof(*sin);
1885 sin->sin_addr = pkt_dst;
1886 rtalloc_ign(&ro, (RTF_PRCLONING | RTF_CLONING));
1887
1888 rt = ro.ro_rt;
1889
1890 if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1891 satosin(rt_key(rt))->sin_addr.s_addr != 0 &&
1892 ipsendredirects && !srcrt && !next_hop) {
1893#define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa))
1894 u_long src = ntohl(ip->ip_src.s_addr);
1895
1896 if (RTA(rt) &&
1897 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1898 if (rt->rt_flags & RTF_GATEWAY)
1899 dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
1900 else
1901 dest = pkt_dst.s_addr;
1902 /* Router requirements says to only send host redirects */
1903 type = ICMP_REDIRECT;
1904 code = ICMP_REDIRECT_HOST;
1905#ifdef DIAGNOSTIC
1906 if (ipprintfs)
1907 printf("redirect (%d) to %lx\n", code, (u_long)dest);
1908#endif
1909 }
1910 }
1911 if (rt)
1912 RTFREE(rt);
1913 }
1914
1915 {
1916 struct m_hdr tag;
1917
1918 if (next_hop) {
1919 /* Pass IPFORWARD info if available */
1920
1921 tag.mh_type = MT_TAG;
1922 tag.mh_flags = PACKET_TAG_IPFORWARD;
1923 tag.mh_data = (caddr_t)next_hop;
1924 tag.mh_next = m;
1925 m = (struct mbuf *)&tag;
1926 }
1927 error = ip_output(m, (struct mbuf *)0, NULL, IP_FORWARDING, 0, NULL);
1928 }
1929 if (error)
1930 ipstat.ips_cantforward++;
1931 else {
1932 ipstat.ips_forward++;
1933 if (type)
1934 ipstat.ips_redirectsent++;
1935 else {
1936 if (mcopy)
1937 m_freem(mcopy);
1938 return;
1939 }
1940 }
1941 if (mcopy == NULL)
1942 return;
1943 destifp = NULL;
1944
1945 switch (error) {
1946
1947 case 0: /* forwarded, but need redirect */
1948 /* type, code set above */
1949 break;
1950
1951 case ENETUNREACH: /* shouldn't happen, checked above */
1952 case EHOSTUNREACH:
1953 case ENETDOWN:
1954 case EHOSTDOWN:
1955 default:
1956 type = ICMP_UNREACH;
1957 code = ICMP_UNREACH_HOST;
1958 break;
1959
1960 case EMSGSIZE:
1961 type = ICMP_UNREACH;
1962 code = ICMP_UNREACH_NEEDFRAG;
1963#if defined(IPSEC) || defined(FAST_IPSEC)
1964 /*
1965 * If the packet is routed over IPsec tunnel, tell the
1966 * originator the tunnel MTU.
1967 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
1968 * XXX quickhack!!!
1969 */
1970 {
1971 struct secpolicy *sp = NULL;
1972 int ipsecerror;
1973 int ipsechdr;
1974 struct route *ro;
1975
1976#ifdef IPSEC
1977 sp = ipsec4_getpolicybyaddr(mcopy,
1978 IPSEC_DIR_OUTBOUND,
1979 IP_FORWARDING,
1980 &ipsecerror);
1981#else /* FAST_IPSEC */
1982 sp = ipsec_getpolicybyaddr(mcopy,
1983 IPSEC_DIR_OUTBOUND,
1984 IP_FORWARDING,
1985 &ipsecerror);
1986#endif
1987 if (sp != NULL) {
1988 /* count IPsec header size */
1989 ipsechdr = ipsec4_hdrsiz(mcopy,
1990 IPSEC_DIR_OUTBOUND,
1991 NULL);
1992
1993 /*
1994 * find the correct route for outer IPv4
1995 * header, compute tunnel MTU.
1996 *
1997 * XXX BUG ALERT
1998 * The "dummyifp" code relies upon the fact
1999 * that icmp_error() touches only ifp->if_mtu.
2000 */
2001 /*XXX*/
2002 destifp = NULL;
2003 if (sp->req != NULL
2004 && sp->req->sav != NULL
2005 && sp->req->sav->sah != NULL) {
2006 ro = &sp->req->sav->sah->sa_route;
2007 if (ro->ro_rt && ro->ro_rt->rt_ifp) {
2008 dummyifp.if_mtu =
2009 ro->ro_rt->rt_ifp->if_mtu;
2010 dummyifp.if_mtu -= ipsechdr;
2011 destifp = &dummyifp;
2012 }
2013 }
2014
2015#ifdef IPSEC
2016 key_freesp(sp);
2017#else /* FAST_IPSEC */
2018 KEY_FREESP(&sp);
2019#endif
2020 ipstat.ips_cantfrag++;
2021 break;
2022 } else
2023#endif /*IPSEC || FAST_IPSEC*/
2024 destifp = ia->ia_ifp;
2025#if defined(IPSEC) || defined(FAST_IPSEC)
2026 }
2027#endif /*IPSEC || FAST_IPSEC*/
2028 ipstat.ips_cantfrag++;
2029 break;
2030
2031 case ENOBUFS:
2032 /*
2033 * A router should not generate ICMP_SOURCEQUENCH as
2034 * required in RFC1812 Requirements for IP Version 4 Routers.
2035 * Source quench could be a big problem under DoS attacks,
2036 * or if the underlying interface is rate-limited.
2037 * Those who need source quench packets may re-enable them
2038 * via the net.inet.ip.sendsourcequench sysctl.
2039 */
2040 if (ip_sendsourcequench == 0) {
2041 m_freem(mcopy);
2042 return;
2043 } else {
2044 type = ICMP_SOURCEQUENCH;
2045 code = 0;
2046 }
2047 break;
2048
2049 case EACCES: /* ipfw denied packet */
2050 m_freem(mcopy);
2051 return;
2052 }
2053 icmp_error(mcopy, type, code, dest, destifp);
2054}
2055
2056void
2057ip_savecontrol(inp, mp, ip, m)
2058 register struct inpcb *inp;
2059 register struct mbuf **mp;
2060 register struct ip *ip;
2061 register struct mbuf *m;
2062{
2063 if (inp->inp_socket->so_options & SO_TIMESTAMP) {
2064 struct timeval tv;
2065
2066 microtime(&tv);
2067 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
2068 SCM_TIMESTAMP, SOL_SOCKET);
2069 if (*mp)
2070 mp = &(*mp)->m_next;
2071 }
2072 if (inp->inp_flags & INP_RECVDSTADDR) {
2073 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
2074 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
2075 if (*mp)
2076 mp = &(*mp)->m_next;
2077 }
2078 if (inp->inp_flags & INP_RECVTTL) {
2079 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl,
2080 sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
2081 if (*mp)
2082 mp = &(*mp)->m_next;
2083 }
2084#ifdef notyet
2085 /* XXX
2086 * Moving these out of udp_input() made them even more broken
2087 * than they already were.
2088 */
2089 /* options were tossed already */
2090 if (inp->inp_flags & INP_RECVOPTS) {
2091 *mp = sbcreatecontrol((caddr_t) opts_deleted_above,
2092 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
2093 if (*mp)
2094 mp = &(*mp)->m_next;
2095 }
2096 /* ip_srcroute doesn't do what we want here, need to fix */
2097 if (inp->inp_flags & INP_RECVRETOPTS) {
2098 *mp = sbcreatecontrol((caddr_t) ip_srcroute(),
2099 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
2100 if (*mp)
2101 mp = &(*mp)->m_next;
2102 }
2103#endif
2104 if (inp->inp_flags & INP_RECVIF) {
2105 struct ifnet *ifp;
2106 struct sdlbuf {
2107 struct sockaddr_dl sdl;
2108 u_char pad[32];
2109 } sdlbuf;
2110 struct sockaddr_dl *sdp;
2111 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
2112
2113 if (((ifp = m->m_pkthdr.rcvif))
2114 && ( ifp->if_index && (ifp->if_index <= if_index))) {
2115 sdp = (struct sockaddr_dl *)
2116 (ifaddr_byindex(ifp->if_index)->ifa_addr);
2117 /*
2118 * Change our mind and don't try copy.
2119 */
2120 if ((sdp->sdl_family != AF_LINK)
2121 || (sdp->sdl_len > sizeof(sdlbuf))) {
2122 goto makedummy;
2123 }
2124 bcopy(sdp, sdl2, sdp->sdl_len);
2125 } else {
2126makedummy:
2127 sdl2->sdl_len
2128 = offsetof(struct sockaddr_dl, sdl_data[0]);
2129 sdl2->sdl_family = AF_LINK;
2130 sdl2->sdl_index = 0;
2131 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
2132 }
2133 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len,
2134 IP_RECVIF, IPPROTO_IP);
2135 if (*mp)
2136 mp = &(*mp)->m_next;
2137 }
2138}
2139
2140/*
2141 * XXX these routines are called from the upper part of the kernel.
2142 * They need to be locked when we remove Giant.
2143 *
2144 * They could also be moved to ip_mroute.c, since all the RSVP
2145 * handling is done there already.
2146 */
2147static int ip_rsvp_on;
2148struct socket *ip_rsvpd;
2149int
2150ip_rsvp_init(struct socket *so)
2151{
2152 if (so->so_type != SOCK_RAW ||
2153 so->so_proto->pr_protocol != IPPROTO_RSVP)
2154 return EOPNOTSUPP;
2155
2156 if (ip_rsvpd != NULL)
2157 return EADDRINUSE;
2158
2159 ip_rsvpd = so;
2160 /*
2161 * This may seem silly, but we need to be sure we don't over-increment
2162 * the RSVP counter, in case something slips up.
2163 */
2164 if (!ip_rsvp_on) {
2165 ip_rsvp_on = 1;
2166 rsvp_on++;
2167 }
2168
2169 return 0;
2170}
2171
2172int
2173ip_rsvp_done(void)
2174{
2175 ip_rsvpd = NULL;
2176 /*
2177 * This may seem silly, but we need to be sure we don't over-decrement
2178 * the RSVP counter, in case something slips up.
2179 */
2180 if (ip_rsvp_on) {
2181 ip_rsvp_on = 0;
2182 rsvp_on--;
2183 }
2184 return 0;
2185}
2186
2187void
2188rsvp_input(struct mbuf *m, int off) /* XXX must fixup manually */
2189{
2190 if (rsvp_input_p) { /* call the real one if loaded */
2191 rsvp_input_p(m, off);
2192 return;
2193 }
2194
2195 /* Can still get packets with rsvp_on = 0 if there is a local member
2196 * of the group to which the RSVP packet is addressed. But in this
2197 * case we want to throw the packet away.
2198 */
2199
2200 if (!rsvp_on) {
2201 m_freem(m);
2202 return;
2203 }
2204
2205 if (ip_rsvpd != NULL) {
2206 rip_input(m, off);
2207 return;
2208 }
2209 /* Drop the packet */
2210 m_freem(m);
2211}