xref: /freebsd/sys/netinet/ip_input.c (revision c6e88cbf92e43c8461d5434e5b02303b51a7099b)
1 /*
2  * Copyright (c) 1982, 1986, 1988, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. All advertising materials mentioning features or use of this software
14  *    must display the following acknowledgement:
15  *	This product includes software developed by the University of
16  *	California, Berkeley and its contributors.
17  * 4. Neither the name of the University nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  *
33  *	@(#)ip_input.c	8.2 (Berkeley) 1/4/94
34  * $FreeBSD$
35  */
36 
37 #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_divert.h>
82 #include <netinet/ip_dummynet.h>
83 
84 #ifdef IPSEC
85 #include <netinet6/ipsec.h>
86 #include <netkey/key.h>
87 #endif
88 
89 #ifdef FAST_IPSEC
90 #include <netipsec/ipsec.h>
91 #include <netipsec/key.h>
92 #endif
93 
94 int rsvp_on = 0;
95 
96 int	ipforwarding = 0;
97 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW,
98     &ipforwarding, 0, "Enable IP forwarding between interfaces");
99 
100 static int	ipsendredirects = 1; /* XXX */
101 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW,
102     &ipsendredirects, 0, "Enable sending IP redirects");
103 
104 int	ip_defttl = IPDEFTTL;
105 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW,
106     &ip_defttl, 0, "Maximum TTL on IP packets");
107 
108 static int	ip_dosourceroute = 0;
109 SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW,
110     &ip_dosourceroute, 0, "Enable forwarding source routed IP packets");
111 
112 static int	ip_acceptsourceroute = 0;
113 SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute,
114     CTLFLAG_RW, &ip_acceptsourceroute, 0,
115     "Enable accepting source routed IP packets");
116 
117 static int	ip_keepfaith = 0;
118 SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW,
119 	&ip_keepfaith,	0,
120 	"Enable packet capture for FAITH IPv4->IPv6 translater daemon");
121 
122 static int    nipq = 0;         /* total # of reass queues */
123 static int    maxnipq;
124 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragpackets, CTLFLAG_RW,
125 	&maxnipq, 0,
126 	"Maximum number of IPv4 fragment reassembly queue entries");
127 
128 static int    maxfragsperpacket;
129 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW,
130 	&maxfragsperpacket, 0,
131 	"Maximum number of IPv4 fragments allowed per packet");
132 
133 static int	ip_sendsourcequench = 0;
134 SYSCTL_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW,
135 	&ip_sendsourcequench, 0,
136 	"Enable the transmission of source quench packets");
137 
138 /*
139  * XXX - Setting ip_checkinterface mostly implements the receive side of
140  * the Strong ES model described in RFC 1122, but since the routing table
141  * and transmit implementation do not implement the Strong ES model,
142  * setting this to 1 results in an odd hybrid.
143  *
144  * XXX - ip_checkinterface currently must be disabled if you use ipnat
145  * to translate the destination address to another local interface.
146  *
147  * XXX - ip_checkinterface must be disabled if you add IP aliases
148  * to the loopback interface instead of the interface where the
149  * packets for those addresses are received.
150  */
151 static int	ip_checkinterface = 1;
152 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW,
153     &ip_checkinterface, 0, "Verify packet arrives on correct interface");
154 
155 #ifdef DIAGNOSTIC
156 static int	ipprintfs = 0;
157 #endif
158 #ifdef PFIL_HOOKS
159 struct pfil_head inet_pfil_hook;
160 #endif
161 
162 static struct	ifqueue ipintrq;
163 static int	ipqmaxlen = IFQ_MAXLEN;
164 
165 extern	struct domain inetdomain;
166 extern	struct protosw inetsw[];
167 u_char	ip_protox[IPPROTO_MAX];
168 struct	in_ifaddrhead in_ifaddrhead; 		/* first inet address */
169 struct	in_ifaddrhashhead *in_ifaddrhashtbl;	/* inet addr hash table  */
170 u_long 	in_ifaddrhmask;				/* mask for hash table */
171 
172 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW,
173     &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue");
174 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD,
175     &ipintrq.ifq_drops, 0, "Number of packets dropped from the IP input queue");
176 
177 struct ipstat ipstat;
178 SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW,
179     &ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)");
180 
181 /* Packet reassembly stuff */
182 #define IPREASS_NHASH_LOG2      6
183 #define IPREASS_NHASH           (1 << IPREASS_NHASH_LOG2)
184 #define IPREASS_HMASK           (IPREASS_NHASH - 1)
185 #define IPREASS_HASH(x,y) \
186 	(((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK)
187 
188 static TAILQ_HEAD(ipqhead, ipq) ipq[IPREASS_NHASH];
189 struct mtx ipqlock;
190 
191 #define	IPQ_LOCK()	mtx_lock(&ipqlock)
192 #define	IPQ_UNLOCK()	mtx_unlock(&ipqlock)
193 #define	IPQ_LOCK_INIT()	mtx_init(&ipqlock, "ipqlock", NULL, MTX_DEF)
194 #define	IPQ_LOCK_ASSERT()	mtx_assert(&ipqlock, MA_OWNED)
195 
196 #ifdef IPCTL_DEFMTU
197 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
198     &ip_mtu, 0, "Default MTU");
199 #endif
200 
201 #ifdef IPSTEALTH
202 int	ipstealth = 0;
203 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW,
204     &ipstealth, 0, "");
205 #endif
206 
207 
208 /* Firewall hooks */
209 ip_fw_chk_t *ip_fw_chk_ptr;
210 int fw_enable = 1 ;
211 int fw_one_pass = 1;
212 
213 /* Dummynet hooks */
214 ip_dn_io_t *ip_dn_io_ptr;
215 
216 /*
217  * XXX this is ugly -- the following two global variables are
218  * used to store packet state while it travels through the stack.
219  * Note that the code even makes assumptions on the size and
220  * alignment of fields inside struct ip_srcrt so e.g. adding some
221  * fields will break the code. This needs to be fixed.
222  *
223  * We need to save the IP options in case a protocol wants to respond
224  * to an incoming packet over the same route if the packet got here
225  * using IP source routing.  This allows connection establishment and
226  * maintenance when the remote end is on a network that is not known
227  * to us.
228  */
229 static int	ip_nhops = 0;
230 static	struct ip_srcrt {
231 	struct	in_addr dst;			/* final destination */
232 	char	nop;				/* one NOP to align */
233 	char	srcopt[IPOPT_OFFSET + 1];	/* OPTVAL, OLEN and OFFSET */
234 	struct	in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)];
235 } ip_srcrt;
236 
237 static void	save_rte(u_char *, struct in_addr);
238 static int	ip_dooptions(struct mbuf *m, int,
239 			struct sockaddr_in *next_hop);
240 static void	ip_forward(struct mbuf *m, int srcrt,
241 			struct sockaddr_in *next_hop);
242 static void	ip_freef(struct ipqhead *, struct ipq *);
243 static struct	mbuf *ip_reass(struct mbuf *, struct ipqhead *, struct ipq *);
244 
245 /*
246  * IP initialization: fill in IP protocol switch table.
247  * All protocols not implemented in kernel go to raw IP protocol handler.
248  */
249 void
250 ip_init()
251 {
252 	register struct protosw *pr;
253 	register int i;
254 
255 	TAILQ_INIT(&in_ifaddrhead);
256 	in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask);
257 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
258 	if (pr == 0)
259 		panic("ip_init");
260 	for (i = 0; i < IPPROTO_MAX; i++)
261 		ip_protox[i] = pr - inetsw;
262 	for (pr = inetdomain.dom_protosw;
263 	    pr < inetdomain.dom_protoswNPROTOSW; pr++)
264 		if (pr->pr_domain->dom_family == PF_INET &&
265 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
266 			ip_protox[pr->pr_protocol] = pr - inetsw;
267 
268 #ifdef PFIL_HOOKS
269 	inet_pfil_hook.ph_type = PFIL_TYPE_AF;
270 	inet_pfil_hook.ph_af = AF_INET;
271 	if ((i = pfil_head_register(&inet_pfil_hook)) != 0)
272 		printf("%s: WARNING: unable to register pfil hook, "
273 			"error %d\n", __func__, i);
274 #endif /* PFIL_HOOKS */
275 
276 	IPQ_LOCK_INIT();
277 	for (i = 0; i < IPREASS_NHASH; i++)
278 	    TAILQ_INIT(&ipq[i]);
279 
280 	maxnipq = nmbclusters / 32;
281 	maxfragsperpacket = 16;
282 
283 #ifndef RANDOM_IP_ID
284 	ip_id = time_second & 0xffff;
285 #endif
286 	ipintrq.ifq_maxlen = ipqmaxlen;
287 	mtx_init(&ipintrq.ifq_mtx, "ip_inq", NULL, MTX_DEF);
288 	netisr_register(NETISR_IP, ip_input, &ipintrq, NETISR_MPSAFE);
289 }
290 
291 /*
292  * Ip input routine.  Checksum and byte swap header.  If fragmented
293  * try to reassemble.  Process options.  Pass to next level.
294  */
295 void
296 ip_input(struct mbuf *m)
297 {
298 	struct ip *ip = NULL;
299 	struct ipq *fp;
300 	struct in_ifaddr *ia = NULL;
301 	struct ifaddr *ifa;
302 	int    i, checkif, hlen = 0;
303 	u_short sum;
304 	struct in_addr pkt_dst;
305 #ifdef IPDIVERT
306 	u_int32_t divert_info;			/* packet divert/tee info */
307 #endif
308 	struct ip_fw_args args;
309 	int dchg = 0;				/* dest changed after fw */
310 #ifdef PFIL_HOOKS
311 	struct in_addr odst;			/* original dst address */
312 #endif
313 #ifdef FAST_IPSEC
314 	struct m_tag *mtag;
315 	struct tdb_ident *tdbi;
316 	struct secpolicy *sp;
317 	int s, error;
318 #endif /* FAST_IPSEC */
319 
320 	args.eh = NULL;
321 	args.oif = NULL;
322 
323   	M_ASSERTPKTHDR(m);
324 
325 	args.next_hop = ip_claim_next_hop(m);
326 	args.rule = ip_dn_claim_rule(m);
327 
328 	if (m->m_flags & M_FASTFWD_OURS) {
329 		/* ip_fastforward firewall changed dest to local */
330 		m->m_flags &= ~M_FASTFWD_OURS;	/* for reflected mbufs */
331   		goto ours;
332   	}
333 
334   	if (args.rule) {	/* dummynet already filtered us */
335   		ip = mtod(m, struct ip *);
336   		hlen = ip->ip_hl << 2;
337 		goto iphack ;
338 	}
339 
340 	ipstat.ips_total++;
341 
342 	if (m->m_pkthdr.len < sizeof(struct ip))
343 		goto tooshort;
344 
345 	if (m->m_len < sizeof (struct ip) &&
346 	    (m = m_pullup(m, sizeof (struct ip))) == 0) {
347 		ipstat.ips_toosmall++;
348 		return;
349 	}
350 	ip = mtod(m, struct ip *);
351 
352 	if (ip->ip_v != IPVERSION) {
353 		ipstat.ips_badvers++;
354 		goto bad;
355 	}
356 
357 	hlen = ip->ip_hl << 2;
358 	if (hlen < sizeof(struct ip)) {	/* minimum header length */
359 		ipstat.ips_badhlen++;
360 		goto bad;
361 	}
362 	if (hlen > m->m_len) {
363 		if ((m = m_pullup(m, hlen)) == 0) {
364 			ipstat.ips_badhlen++;
365 			return;
366 		}
367 		ip = mtod(m, struct ip *);
368 	}
369 
370 	/* 127/8 must not appear on wire - RFC1122 */
371 	if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
372 	    (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
373 		if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
374 			ipstat.ips_badaddr++;
375 			goto bad;
376 		}
377 	}
378 
379 	if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
380 		sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
381 	} else {
382 		if (hlen == sizeof(struct ip)) {
383 			sum = in_cksum_hdr(ip);
384 		} else {
385 			sum = in_cksum(m, hlen);
386 		}
387 	}
388 	if (sum) {
389 		ipstat.ips_badsum++;
390 		goto bad;
391 	}
392 
393 	/*
394 	 * Convert fields to host representation.
395 	 */
396 	ip->ip_len = ntohs(ip->ip_len);
397 	if (ip->ip_len < hlen) {
398 		ipstat.ips_badlen++;
399 		goto bad;
400 	}
401 	ip->ip_off = ntohs(ip->ip_off);
402 
403 	/*
404 	 * Check that the amount of data in the buffers
405 	 * is as at least much as the IP header would have us expect.
406 	 * Trim mbufs if longer than we expect.
407 	 * Drop packet if shorter than we expect.
408 	 */
409 	if (m->m_pkthdr.len < ip->ip_len) {
410 tooshort:
411 		ipstat.ips_tooshort++;
412 		goto bad;
413 	}
414 	if (m->m_pkthdr.len > ip->ip_len) {
415 		if (m->m_len == m->m_pkthdr.len) {
416 			m->m_len = ip->ip_len;
417 			m->m_pkthdr.len = ip->ip_len;
418 		} else
419 			m_adj(m, ip->ip_len - m->m_pkthdr.len);
420 	}
421 #if defined(IPSEC) && !defined(IPSEC_FILTERGIF)
422 	/*
423 	 * Bypass packet filtering for packets from a tunnel (gif).
424 	 */
425 	if (ipsec_getnhist(m))
426 		goto pass;
427 #endif
428 #if defined(FAST_IPSEC) && !defined(IPSEC_FILTERGIF)
429 	/*
430 	 * Bypass packet filtering for packets from a tunnel (gif).
431 	 */
432 	if (m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL)
433 		goto pass;
434 #endif
435 
436 	/*
437 	 * IpHack's section.
438 	 * Right now when no processing on packet has done
439 	 * and it is still fresh out of network we do our black
440 	 * deals with it.
441 	 * - Firewall: deny/allow/divert
442 	 * - Xlate: translate packet's addr/port (NAT).
443 	 * - Pipe: pass pkt through dummynet.
444 	 * - Wrap: fake packet's addr/port <unimpl.>
445 	 * - Encapsulate: put it in another IP and send out. <unimp.>
446  	 */
447 
448 iphack:
449 
450 #ifdef PFIL_HOOKS
451 	/*
452 	 * Run through list of hooks for input packets.
453 	 *
454 	 * NB: Beware of the destination address changing (e.g.
455 	 *     by NAT rewriting).  When this happens, tell
456 	 *     ip_forward to do the right thing.
457 	 */
458 	odst = ip->ip_dst;
459 	if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif,
460 	    PFIL_IN) != 0)
461 		return;
462 	if (m == NULL)			/* consumed by filter */
463 		return;
464 	ip = mtod(m, struct ip *);
465 	dchg = (odst.s_addr != ip->ip_dst.s_addr);
466 #endif /* PFIL_HOOKS */
467 
468 	if (fw_enable && IPFW_LOADED) {
469 		/*
470 		 * If we've been forwarded from the output side, then
471 		 * skip the firewall a second time
472 		 */
473 		if (args.next_hop)
474 			goto ours;
475 
476 		args.m = m;
477 		i = ip_fw_chk_ptr(&args);
478 		m = args.m;
479 
480 		if ( (i & IP_FW_PORT_DENY_FLAG) || m == NULL) { /* drop */
481 			if (m)
482 				m_freem(m);
483 			return;
484 		}
485 		ip = mtod(m, struct ip *); /* just in case m changed */
486 		if (i == 0 && args.next_hop == NULL)	/* common case */
487 			goto pass;
488                 if (DUMMYNET_LOADED && (i & IP_FW_PORT_DYNT_FLAG) != 0) {
489 			/* Send packet to the appropriate pipe */
490 			ip_dn_io_ptr(m, i&0xffff, DN_TO_IP_IN, &args);
491 			return;
492 		}
493 #ifdef IPDIVERT
494 		if (i != 0 && (i & IP_FW_PORT_DYNT_FLAG) == 0) {
495 			/* Divert or tee packet */
496 			goto ours;
497 		}
498 #endif
499 		if (i == 0 && args.next_hop != NULL)
500 			goto pass;
501 		/*
502 		 * if we get here, the packet must be dropped
503 		 */
504 		m_freem(m);
505 		return;
506 	}
507 pass:
508 
509 	/*
510 	 * Process options and, if not destined for us,
511 	 * ship it on.  ip_dooptions returns 1 when an
512 	 * error was detected (causing an icmp message
513 	 * to be sent and the original packet to be freed).
514 	 */
515 	ip_nhops = 0;		/* for source routed packets */
516 	if (hlen > sizeof (struct ip) && ip_dooptions(m, 0, args.next_hop))
517 		return;
518 
519         /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
520          * matter if it is destined to another node, or whether it is
521          * a multicast one, RSVP wants it! and prevents it from being forwarded
522          * anywhere else. Also checks if the rsvp daemon is running before
523 	 * grabbing the packet.
524          */
525 	if (rsvp_on && ip->ip_p==IPPROTO_RSVP)
526 		goto ours;
527 
528 	/*
529 	 * Check our list of addresses, to see if the packet is for us.
530 	 * If we don't have any addresses, assume any unicast packet
531 	 * we receive might be for us (and let the upper layers deal
532 	 * with it).
533 	 */
534 	if (TAILQ_EMPTY(&in_ifaddrhead) &&
535 	    (m->m_flags & (M_MCAST|M_BCAST)) == 0)
536 		goto ours;
537 
538 	/*
539 	 * Cache the destination address of the packet; this may be
540 	 * changed by use of 'ipfw fwd'.
541 	 */
542 	pkt_dst = args.next_hop ? args.next_hop->sin_addr : ip->ip_dst;
543 
544 	/*
545 	 * Enable a consistency check between the destination address
546 	 * and the arrival interface for a unicast packet (the RFC 1122
547 	 * strong ES model) if IP forwarding is disabled and the packet
548 	 * is not locally generated and the packet is not subject to
549 	 * 'ipfw fwd'.
550 	 *
551 	 * XXX - Checking also should be disabled if the destination
552 	 * address is ipnat'ed to a different interface.
553 	 *
554 	 * XXX - Checking is incompatible with IP aliases added
555 	 * to the loopback interface instead of the interface where
556 	 * the packets are received.
557 	 */
558 	checkif = ip_checkinterface && (ipforwarding == 0) &&
559 	    m->m_pkthdr.rcvif != NULL &&
560 	    ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) &&
561 	    (args.next_hop == NULL) && (dchg == 0);
562 
563 	/*
564 	 * Check for exact addresses in the hash bucket.
565 	 */
566 	LIST_FOREACH(ia, INADDR_HASH(pkt_dst.s_addr), ia_hash) {
567 		/*
568 		 * If the address matches, verify that the packet
569 		 * arrived via the correct interface if checking is
570 		 * enabled.
571 		 */
572 		if (IA_SIN(ia)->sin_addr.s_addr == pkt_dst.s_addr &&
573 		    (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif))
574 			goto ours;
575 	}
576 	/*
577 	 * Check for broadcast addresses.
578 	 *
579 	 * Only accept broadcast packets that arrive via the matching
580 	 * interface.  Reception of forwarded directed broadcasts would
581 	 * be handled via ip_forward() and ether_output() with the loopback
582 	 * into the stack for SIMPLEX interfaces handled by ether_output().
583 	 */
584 	if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
585 	        TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
586 			if (ifa->ifa_addr->sa_family != AF_INET)
587 				continue;
588 			ia = ifatoia(ifa);
589 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
590 			    pkt_dst.s_addr)
591 				goto ours;
592 			if (ia->ia_netbroadcast.s_addr == pkt_dst.s_addr)
593 				goto ours;
594 #ifdef BOOTP_COMPAT
595 			if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY)
596 				goto ours;
597 #endif
598 		}
599 	}
600 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
601 		struct in_multi *inm;
602 		if (ip_mrouter) {
603 			/*
604 			 * If we are acting as a multicast router, all
605 			 * incoming multicast packets are passed to the
606 			 * kernel-level multicast forwarding function.
607 			 * The packet is returned (relatively) intact; if
608 			 * ip_mforward() returns a non-zero value, the packet
609 			 * must be discarded, else it may be accepted below.
610 			 */
611 			if (ip_mforward &&
612 			    ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) {
613 				ipstat.ips_cantforward++;
614 				m_freem(m);
615 				return;
616 			}
617 
618 			/*
619 			 * The process-level routing daemon needs to receive
620 			 * all multicast IGMP packets, whether or not this
621 			 * host belongs to their destination groups.
622 			 */
623 			if (ip->ip_p == IPPROTO_IGMP)
624 				goto ours;
625 			ipstat.ips_forward++;
626 		}
627 		/*
628 		 * See if we belong to the destination multicast group on the
629 		 * arrival interface.
630 		 */
631 		IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
632 		if (inm == NULL) {
633 			ipstat.ips_notmember++;
634 			m_freem(m);
635 			return;
636 		}
637 		goto ours;
638 	}
639 	if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
640 		goto ours;
641 	if (ip->ip_dst.s_addr == INADDR_ANY)
642 		goto ours;
643 
644 	/*
645 	 * FAITH(Firewall Aided Internet Translator)
646 	 */
647 	if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
648 		if (ip_keepfaith) {
649 			if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP)
650 				goto ours;
651 		}
652 		m_freem(m);
653 		return;
654 	}
655 
656 	/*
657 	 * Not for us; forward if possible and desirable.
658 	 */
659 	if (ipforwarding == 0) {
660 		ipstat.ips_cantforward++;
661 		m_freem(m);
662 	} else {
663 #ifdef IPSEC
664 		/*
665 		 * Enforce inbound IPsec SPD.
666 		 */
667 		if (ipsec4_in_reject(m, NULL)) {
668 			ipsecstat.in_polvio++;
669 			goto bad;
670 		}
671 #endif /* IPSEC */
672 #ifdef FAST_IPSEC
673 		mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
674 		s = splnet();
675 		if (mtag != NULL) {
676 			tdbi = (struct tdb_ident *)(mtag + 1);
677 			sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
678 		} else {
679 			sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
680 						   IP_FORWARDING, &error);
681 		}
682 		if (sp == NULL) {	/* NB: can happen if error */
683 			splx(s);
684 			/*XXX error stat???*/
685 			DPRINTF(("ip_input: no SP for forwarding\n"));	/*XXX*/
686 			goto bad;
687 		}
688 
689 		/*
690 		 * Check security policy against packet attributes.
691 		 */
692 		error = ipsec_in_reject(sp, m);
693 		KEY_FREESP(&sp);
694 		splx(s);
695 		if (error) {
696 			ipstat.ips_cantforward++;
697 			goto bad;
698 		}
699 #endif /* FAST_IPSEC */
700 		ip_forward(m, dchg, args.next_hop);
701 	}
702 	return;
703 
704 ours:
705 #ifdef IPSTEALTH
706 	/*
707 	 * IPSTEALTH: Process non-routing options only
708 	 * if the packet is destined for us.
709 	 */
710 	if (ipstealth && hlen > sizeof (struct ip) &&
711 	    ip_dooptions(m, 1, args.next_hop))
712 		return;
713 #endif /* IPSTEALTH */
714 
715 	/* Count the packet in the ip address stats */
716 	if (ia != NULL) {
717 		ia->ia_ifa.if_ipackets++;
718 		ia->ia_ifa.if_ibytes += m->m_pkthdr.len;
719 	}
720 
721 	/*
722 	 * If offset or IP_MF are set, must reassemble.
723 	 * Otherwise, nothing need be done.
724 	 * (We could look in the reassembly queue to see
725 	 * if the packet was previously fragmented,
726 	 * but it's not worth the time; just let them time out.)
727 	 */
728 	if (ip->ip_off & (IP_MF | IP_OFFMASK)) {
729 
730 		/* If maxnipq is 0, never accept fragments. */
731 		if (maxnipq == 0) {
732                 	ipstat.ips_fragments++;
733 			ipstat.ips_fragdropped++;
734 			goto bad;
735 		}
736 
737 		sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
738 		IPQ_LOCK();
739 		/*
740 		 * Look for queue of fragments
741 		 * of this datagram.
742 		 */
743 		TAILQ_FOREACH(fp, &ipq[sum], ipq_list)
744 			if (ip->ip_id == fp->ipq_id &&
745 			    ip->ip_src.s_addr == fp->ipq_src.s_addr &&
746 			    ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
747 #ifdef MAC
748 			    mac_fragment_match(m, fp) &&
749 #endif
750 			    ip->ip_p == fp->ipq_p)
751 				goto found;
752 
753 		fp = NULL;
754 
755 		/*
756 		 * Enforce upper bound on number of fragmented packets
757 		 * for which we attempt reassembly;
758 		 * If maxnipq is -1, accept all fragments without limitation.
759 		 */
760 		if ((nipq > maxnipq) && (maxnipq > 0)) {
761 		    /*
762 		     * drop something from the tail of the current queue
763 		     * before proceeding further
764 		     */
765 		    struct ipq *q = TAILQ_LAST(&ipq[sum], ipqhead);
766 		    if (q == NULL) {   /* gak */
767 			for (i = 0; i < IPREASS_NHASH; i++) {
768 			    struct ipq *r = TAILQ_LAST(&ipq[i], ipqhead);
769 			    if (r) {
770 				ipstat.ips_fragtimeout += r->ipq_nfrags;
771 				ip_freef(&ipq[i], r);
772 				break;
773 			    }
774 			}
775 		    } else {
776 			ipstat.ips_fragtimeout += q->ipq_nfrags;
777 			ip_freef(&ipq[sum], q);
778 		    }
779 		}
780 found:
781 		/*
782 		 * Adjust ip_len to not reflect header,
783 		 * convert offset of this to bytes.
784 		 */
785 		ip->ip_len -= hlen;
786 		if (ip->ip_off & IP_MF) {
787 		        /*
788 		         * Make sure that fragments have a data length
789 			 * that's a non-zero multiple of 8 bytes.
790 		         */
791 			if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
792 				IPQ_UNLOCK();
793 				ipstat.ips_toosmall++; /* XXX */
794 				goto bad;
795 			}
796 			m->m_flags |= M_FRAG;
797 		} else
798 			m->m_flags &= ~M_FRAG;
799 		ip->ip_off <<= 3;
800 
801 		/*
802 		 * Attempt reassembly; if it succeeds, proceed.
803 		 * ip_reass() will return a different mbuf.
804 		 */
805 		ipstat.ips_fragments++;
806 		m->m_pkthdr.header = ip;
807 		m = ip_reass(m, &ipq[sum], fp);
808 		IPQ_UNLOCK();
809 		if (m == 0)
810 			return;
811 		ipstat.ips_reassembled++;
812 		ip = mtod(m, struct ip *);
813 		/* Get the header length of the reassembled packet */
814 		hlen = ip->ip_hl << 2;
815 #ifdef IPDIVERT
816 		/* Restore original checksum before diverting packet */
817 		if (divert_find_info(m) != 0) {
818 			ip->ip_len += hlen;
819 			ip->ip_len = htons(ip->ip_len);
820 			ip->ip_off = htons(ip->ip_off);
821 			ip->ip_sum = 0;
822 			if (hlen == sizeof(struct ip))
823 				ip->ip_sum = in_cksum_hdr(ip);
824 			else
825 				ip->ip_sum = in_cksum(m, hlen);
826 			ip->ip_off = ntohs(ip->ip_off);
827 			ip->ip_len = ntohs(ip->ip_len);
828 			ip->ip_len -= hlen;
829 		}
830 #endif
831 	} else
832 		ip->ip_len -= hlen;
833 
834 #ifdef IPDIVERT
835 	/*
836 	 * Divert or tee packet to the divert protocol if required.
837 	 */
838 	divert_info = divert_find_info(m);
839 	if (divert_info != 0) {
840 		struct mbuf *clone;
841 
842 		/* Clone packet if we're doing a 'tee' */
843 		if ((divert_info & IP_FW_PORT_TEE_FLAG) != 0)
844 			clone = divert_clone(m);
845 		else
846 			clone = NULL;
847 
848 		/* Restore packet header fields to original values */
849 		ip->ip_len += hlen;
850 		ip->ip_len = htons(ip->ip_len);
851 		ip->ip_off = htons(ip->ip_off);
852 
853 		/* Deliver packet to divert input routine */
854 		divert_packet(m, 1);
855 		ipstat.ips_delivered++;
856 
857 		/* If 'tee', continue with original packet */
858 		if (clone == NULL)
859 			return;
860 		m = clone;
861 		ip = mtod(m, struct ip *);
862 		ip->ip_len += hlen;
863 		/*
864 		 * Jump backwards to complete processing of the
865 		 * packet.  We do not need to clear args.next_hop
866 		 * as that will not be used again and the cloned packet
867 		 * doesn't contain a divert packet tag so we won't
868 		 * re-entry this block.
869 		 */
870 		goto pass;
871 	}
872 #endif
873 
874 #ifdef IPSEC
875 	/*
876 	 * enforce IPsec policy checking if we are seeing last header.
877 	 * note that we do not visit this with protocols with pcb layer
878 	 * code - like udp/tcp/raw ip.
879 	 */
880 	if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0 &&
881 	    ipsec4_in_reject(m, NULL)) {
882 		ipsecstat.in_polvio++;
883 		goto bad;
884 	}
885 #endif
886 #if FAST_IPSEC
887 	/*
888 	 * enforce IPsec policy checking if we are seeing last header.
889 	 * note that we do not visit this with protocols with pcb layer
890 	 * code - like udp/tcp/raw ip.
891 	 */
892 	if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) {
893 		/*
894 		 * Check if the packet has already had IPsec processing
895 		 * done.  If so, then just pass it along.  This tag gets
896 		 * set during AH, ESP, etc. input handling, before the
897 		 * packet is returned to the ip input queue for delivery.
898 		 */
899 		mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
900 		s = splnet();
901 		if (mtag != NULL) {
902 			tdbi = (struct tdb_ident *)(mtag + 1);
903 			sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
904 		} else {
905 			sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
906 						   IP_FORWARDING, &error);
907 		}
908 		if (sp != NULL) {
909 			/*
910 			 * Check security policy against packet attributes.
911 			 */
912 			error = ipsec_in_reject(sp, m);
913 			KEY_FREESP(&sp);
914 		} else {
915 			/* XXX error stat??? */
916 			error = EINVAL;
917 DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/
918 			goto bad;
919 		}
920 		splx(s);
921 		if (error)
922 			goto bad;
923 	}
924 #endif /* FAST_IPSEC */
925 
926 	/*
927 	 * Switch out to protocol's input routine.
928 	 */
929 	ipstat.ips_delivered++;
930 	if (args.next_hop && ip->ip_p == IPPROTO_TCP) {
931 		/* attach next hop info for TCP */
932 		struct m_tag *mtag = m_tag_get(PACKET_TAG_IPFORWARD,
933 		    sizeof(struct sockaddr_in *), M_NOWAIT);
934 		if (mtag == NULL)
935 			goto bad;
936 		*(struct sockaddr_in **)(mtag+1) = args.next_hop;
937 		m_tag_prepend(m, mtag);
938 	}
939 	(*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen);
940 	return;
941 bad:
942 	m_freem(m);
943 }
944 
945 /*
946  * Take incoming datagram fragment and try to reassemble it into
947  * whole datagram.  If a chain for reassembly of this datagram already
948  * exists, then it is given as fp; otherwise have to make a chain.
949  *
950  * When IPDIVERT enabled, keep additional state with each packet that
951  * tells us if we need to divert or tee the packet we're building.
952  * In particular, *divinfo includes the port and TEE flag,
953  * *divert_rule is the number of the matching rule.
954  */
955 
956 static struct mbuf *
957 ip_reass(struct mbuf *m, struct ipqhead *head, struct ipq *fp)
958 {
959 	struct ip *ip = mtod(m, struct ip *);
960 	register struct mbuf *p, *q, *nq;
961 	struct mbuf *t;
962 	int hlen = ip->ip_hl << 2;
963 	int i, next;
964 	u_int8_t ecn, ecn0;
965 
966 	IPQ_LOCK_ASSERT();
967 
968 	/*
969 	 * Presence of header sizes in mbufs
970 	 * would confuse code below.
971 	 */
972 	m->m_data += hlen;
973 	m->m_len -= hlen;
974 
975 	/*
976 	 * If first fragment to arrive, create a reassembly queue.
977 	 */
978 	if (fp == NULL) {
979 		if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL)
980 			goto dropfrag;
981 		fp = mtod(t, struct ipq *);
982 #ifdef MAC
983 		if (mac_init_ipq(fp, M_NOWAIT) != 0) {
984 			m_free(t);
985 			goto dropfrag;
986 		}
987 		mac_create_ipq(m, fp);
988 #endif
989 		TAILQ_INSERT_HEAD(head, fp, ipq_list);
990 		nipq++;
991 		fp->ipq_nfrags = 1;
992 		fp->ipq_ttl = IPFRAGTTL;
993 		fp->ipq_p = ip->ip_p;
994 		fp->ipq_id = ip->ip_id;
995 		fp->ipq_src = ip->ip_src;
996 		fp->ipq_dst = ip->ip_dst;
997 		fp->ipq_frags = m;
998 		m->m_nextpkt = NULL;
999 		goto inserted;
1000 	} else {
1001 		fp->ipq_nfrags++;
1002 #ifdef MAC
1003 		mac_update_ipq(m, fp);
1004 #endif
1005 	}
1006 
1007 #define GETIP(m)	((struct ip*)((m)->m_pkthdr.header))
1008 
1009 	/*
1010 	 * Handle ECN by comparing this segment with the first one;
1011 	 * if CE is set, do not lose CE.
1012 	 * drop if CE and not-ECT are mixed for the same packet.
1013 	 */
1014 	ecn = ip->ip_tos & IPTOS_ECN_MASK;
1015 	ecn0 = GETIP(fp->ipq_frags)->ip_tos & IPTOS_ECN_MASK;
1016 	if (ecn == IPTOS_ECN_CE) {
1017 		if (ecn0 == IPTOS_ECN_NOTECT)
1018 			goto dropfrag;
1019 		if (ecn0 != IPTOS_ECN_CE)
1020 			GETIP(fp->ipq_frags)->ip_tos |= IPTOS_ECN_CE;
1021 	}
1022 	if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT)
1023 		goto dropfrag;
1024 
1025 	/*
1026 	 * Find a segment which begins after this one does.
1027 	 */
1028 	for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt)
1029 		if (GETIP(q)->ip_off > ip->ip_off)
1030 			break;
1031 
1032 	/*
1033 	 * If there is a preceding segment, it may provide some of
1034 	 * our data already.  If so, drop the data from the incoming
1035 	 * segment.  If it provides all of our data, drop us, otherwise
1036 	 * stick new segment in the proper place.
1037 	 *
1038 	 * If some of the data is dropped from the the preceding
1039 	 * segment, then it's checksum is invalidated.
1040 	 */
1041 	if (p) {
1042 		i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off;
1043 		if (i > 0) {
1044 			if (i >= ip->ip_len)
1045 				goto dropfrag;
1046 			m_adj(m, i);
1047 			m->m_pkthdr.csum_flags = 0;
1048 			ip->ip_off += i;
1049 			ip->ip_len -= i;
1050 		}
1051 		m->m_nextpkt = p->m_nextpkt;
1052 		p->m_nextpkt = m;
1053 	} else {
1054 		m->m_nextpkt = fp->ipq_frags;
1055 		fp->ipq_frags = m;
1056 	}
1057 
1058 	/*
1059 	 * While we overlap succeeding segments trim them or,
1060 	 * if they are completely covered, dequeue them.
1061 	 */
1062 	for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off;
1063 	     q = nq) {
1064 		i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off;
1065 		if (i < GETIP(q)->ip_len) {
1066 			GETIP(q)->ip_len -= i;
1067 			GETIP(q)->ip_off += i;
1068 			m_adj(q, i);
1069 			q->m_pkthdr.csum_flags = 0;
1070 			break;
1071 		}
1072 		nq = q->m_nextpkt;
1073 		m->m_nextpkt = nq;
1074 		ipstat.ips_fragdropped++;
1075 		fp->ipq_nfrags--;
1076 		m_freem(q);
1077 	}
1078 
1079 inserted:
1080 
1081 #ifdef IPDIVERT
1082 	if (ip->ip_off != 0) {
1083 		/*
1084 		 * Strip any divert information; only the info
1085 		 * on the first fragment is used/kept.
1086 		 */
1087 		struct m_tag *mtag = m_tag_find(m, PACKET_TAG_DIVERT, NULL);
1088 		if (mtag)
1089 			m_tag_delete(m, mtag);
1090 	}
1091 #endif
1092 
1093 	/*
1094 	 * Check for complete reassembly and perform frag per packet
1095 	 * limiting.
1096 	 *
1097 	 * Frag limiting is performed here so that the nth frag has
1098 	 * a chance to complete the packet before we drop the packet.
1099 	 * As a result, n+1 frags are actually allowed per packet, but
1100 	 * only n will ever be stored. (n = maxfragsperpacket.)
1101 	 *
1102 	 */
1103 	next = 0;
1104 	for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
1105 		if (GETIP(q)->ip_off != next) {
1106 			if (fp->ipq_nfrags > maxfragsperpacket) {
1107 				ipstat.ips_fragdropped += fp->ipq_nfrags;
1108 				ip_freef(head, fp);
1109 			}
1110 			return (0);
1111 		}
1112 		next += GETIP(q)->ip_len;
1113 	}
1114 	/* Make sure the last packet didn't have the IP_MF flag */
1115 	if (p->m_flags & M_FRAG) {
1116 		if (fp->ipq_nfrags > maxfragsperpacket) {
1117 			ipstat.ips_fragdropped += fp->ipq_nfrags;
1118 			ip_freef(head, fp);
1119 		}
1120 		return (0);
1121 	}
1122 
1123 	/*
1124 	 * Reassembly is complete.  Make sure the packet is a sane size.
1125 	 */
1126 	q = fp->ipq_frags;
1127 	ip = GETIP(q);
1128 	if (next + (ip->ip_hl << 2) > IP_MAXPACKET) {
1129 		ipstat.ips_toolong++;
1130 		ipstat.ips_fragdropped += fp->ipq_nfrags;
1131 		ip_freef(head, fp);
1132 		return (0);
1133 	}
1134 
1135 	/*
1136 	 * Concatenate fragments.
1137 	 */
1138 	m = q;
1139 	t = m->m_next;
1140 	m->m_next = 0;
1141 	m_cat(m, t);
1142 	nq = q->m_nextpkt;
1143 	q->m_nextpkt = 0;
1144 	for (q = nq; q != NULL; q = nq) {
1145 		nq = q->m_nextpkt;
1146 		q->m_nextpkt = NULL;
1147 		m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
1148 		m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
1149 		m_cat(m, q);
1150 	}
1151 #ifdef MAC
1152 	mac_create_datagram_from_ipq(fp, m);
1153 	mac_destroy_ipq(fp);
1154 #endif
1155 
1156 	/*
1157 	 * Create header for new ip packet by
1158 	 * modifying header of first packet;
1159 	 * dequeue and discard fragment reassembly header.
1160 	 * Make header visible.
1161 	 */
1162 	ip->ip_len = next;
1163 	ip->ip_src = fp->ipq_src;
1164 	ip->ip_dst = fp->ipq_dst;
1165 	TAILQ_REMOVE(head, fp, ipq_list);
1166 	nipq--;
1167 	(void) m_free(dtom(fp));
1168 	m->m_len += (ip->ip_hl << 2);
1169 	m->m_data -= (ip->ip_hl << 2);
1170 	/* some debugging cruft by sklower, below, will go away soon */
1171 	if (m->m_flags & M_PKTHDR)	/* XXX this should be done elsewhere */
1172 		m_fixhdr(m);
1173 	return (m);
1174 
1175 dropfrag:
1176 	ipstat.ips_fragdropped++;
1177 	if (fp != NULL)
1178 		fp->ipq_nfrags--;
1179 	m_freem(m);
1180 	return (0);
1181 
1182 #undef GETIP
1183 }
1184 
1185 /*
1186  * Free a fragment reassembly header and all
1187  * associated datagrams.
1188  */
1189 static void
1190 ip_freef(fhp, fp)
1191 	struct ipqhead *fhp;
1192 	struct ipq *fp;
1193 {
1194 	register struct mbuf *q;
1195 
1196 	IPQ_LOCK_ASSERT();
1197 
1198 	while (fp->ipq_frags) {
1199 		q = fp->ipq_frags;
1200 		fp->ipq_frags = q->m_nextpkt;
1201 		m_freem(q);
1202 	}
1203 	TAILQ_REMOVE(fhp, fp, ipq_list);
1204 	(void) m_free(dtom(fp));
1205 	nipq--;
1206 }
1207 
1208 /*
1209  * IP timer processing;
1210  * if a timer expires on a reassembly
1211  * queue, discard it.
1212  */
1213 void
1214 ip_slowtimo()
1215 {
1216 	register struct ipq *fp;
1217 	int s = splnet();
1218 	int i;
1219 
1220 	IPQ_LOCK();
1221 	for (i = 0; i < IPREASS_NHASH; i++) {
1222 		for(fp = TAILQ_FIRST(&ipq[i]); fp;) {
1223 			struct ipq *fpp;
1224 
1225 			fpp = fp;
1226 			fp = TAILQ_NEXT(fp, ipq_list);
1227 			if(--fpp->ipq_ttl == 0) {
1228 				ipstat.ips_fragtimeout += fpp->ipq_nfrags;
1229 				ip_freef(&ipq[i], fpp);
1230 			}
1231 		}
1232 	}
1233 	/*
1234 	 * If we are over the maximum number of fragments
1235 	 * (due to the limit being lowered), drain off
1236 	 * enough to get down to the new limit.
1237 	 */
1238 	if (maxnipq >= 0 && nipq > maxnipq) {
1239 		for (i = 0; i < IPREASS_NHASH; i++) {
1240 			while (nipq > maxnipq && !TAILQ_EMPTY(&ipq[i])) {
1241 				ipstat.ips_fragdropped +=
1242 				    TAILQ_FIRST(&ipq[i])->ipq_nfrags;
1243 				ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i]));
1244 			}
1245 		}
1246 	}
1247 	IPQ_UNLOCK();
1248 	splx(s);
1249 }
1250 
1251 /*
1252  * Drain off all datagram fragments.
1253  */
1254 void
1255 ip_drain()
1256 {
1257 	int     i;
1258 
1259 	IPQ_LOCK();
1260 	for (i = 0; i < IPREASS_NHASH; i++) {
1261 		while(!TAILQ_EMPTY(&ipq[i])) {
1262 			ipstat.ips_fragdropped +=
1263 			    TAILQ_FIRST(&ipq[i])->ipq_nfrags;
1264 			ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i]));
1265 		}
1266 	}
1267 	IPQ_UNLOCK();
1268 	in_rtqdrain();
1269 }
1270 
1271 /*
1272  * Do option processing on a datagram,
1273  * possibly discarding it if bad options are encountered,
1274  * or forwarding it if source-routed.
1275  * The pass argument is used when operating in the IPSTEALTH
1276  * mode to tell what options to process:
1277  * [LS]SRR (pass 0) or the others (pass 1).
1278  * The reason for as many as two passes is that when doing IPSTEALTH,
1279  * non-routing options should be processed only if the packet is for us.
1280  * Returns 1 if packet has been forwarded/freed,
1281  * 0 if the packet should be processed further.
1282  */
1283 static int
1284 ip_dooptions(struct mbuf *m, int pass, struct sockaddr_in *next_hop)
1285 {
1286 	struct ip *ip = mtod(m, struct ip *);
1287 	u_char *cp;
1288 	struct in_ifaddr *ia;
1289 	int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0;
1290 	struct in_addr *sin, dst;
1291 	n_time ntime;
1292 	struct	sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET };
1293 
1294 	dst = ip->ip_dst;
1295 	cp = (u_char *)(ip + 1);
1296 	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
1297 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1298 		opt = cp[IPOPT_OPTVAL];
1299 		if (opt == IPOPT_EOL)
1300 			break;
1301 		if (opt == IPOPT_NOP)
1302 			optlen = 1;
1303 		else {
1304 			if (cnt < IPOPT_OLEN + sizeof(*cp)) {
1305 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1306 				goto bad;
1307 			}
1308 			optlen = cp[IPOPT_OLEN];
1309 			if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
1310 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1311 				goto bad;
1312 			}
1313 		}
1314 		switch (opt) {
1315 
1316 		default:
1317 			break;
1318 
1319 		/*
1320 		 * Source routing with record.
1321 		 * Find interface with current destination address.
1322 		 * If none on this machine then drop if strictly routed,
1323 		 * or do nothing if loosely routed.
1324 		 * Record interface address and bring up next address
1325 		 * component.  If strictly routed make sure next
1326 		 * address is on directly accessible net.
1327 		 */
1328 		case IPOPT_LSRR:
1329 		case IPOPT_SSRR:
1330 #ifdef IPSTEALTH
1331 			if (ipstealth && pass > 0)
1332 				break;
1333 #endif
1334 			if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1335 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1336 				goto bad;
1337 			}
1338 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1339 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1340 				goto bad;
1341 			}
1342 			ipaddr.sin_addr = ip->ip_dst;
1343 			ia = (struct in_ifaddr *)
1344 				ifa_ifwithaddr((struct sockaddr *)&ipaddr);
1345 			if (ia == 0) {
1346 				if (opt == IPOPT_SSRR) {
1347 					type = ICMP_UNREACH;
1348 					code = ICMP_UNREACH_SRCFAIL;
1349 					goto bad;
1350 				}
1351 				if (!ip_dosourceroute)
1352 					goto nosourcerouting;
1353 				/*
1354 				 * Loose routing, and not at next destination
1355 				 * yet; nothing to do except forward.
1356 				 */
1357 				break;
1358 			}
1359 			off--;			/* 0 origin */
1360 			if (off > optlen - (int)sizeof(struct in_addr)) {
1361 				/*
1362 				 * End of source route.  Should be for us.
1363 				 */
1364 				if (!ip_acceptsourceroute)
1365 					goto nosourcerouting;
1366 				save_rte(cp, ip->ip_src);
1367 				break;
1368 			}
1369 #ifdef IPSTEALTH
1370 			if (ipstealth)
1371 				goto dropit;
1372 #endif
1373 			if (!ip_dosourceroute) {
1374 				if (ipforwarding) {
1375 					char buf[16]; /* aaa.bbb.ccc.ddd\0 */
1376 					/*
1377 					 * Acting as a router, so generate ICMP
1378 					 */
1379 nosourcerouting:
1380 					strcpy(buf, inet_ntoa(ip->ip_dst));
1381 					log(LOG_WARNING,
1382 					    "attempted source route from %s to %s\n",
1383 					    inet_ntoa(ip->ip_src), buf);
1384 					type = ICMP_UNREACH;
1385 					code = ICMP_UNREACH_SRCFAIL;
1386 					goto bad;
1387 				} else {
1388 					/*
1389 					 * Not acting as a router, so silently drop.
1390 					 */
1391 #ifdef IPSTEALTH
1392 dropit:
1393 #endif
1394 					ipstat.ips_cantforward++;
1395 					m_freem(m);
1396 					return (1);
1397 				}
1398 			}
1399 
1400 			/*
1401 			 * locate outgoing interface
1402 			 */
1403 			(void)memcpy(&ipaddr.sin_addr, cp + off,
1404 			    sizeof(ipaddr.sin_addr));
1405 
1406 			if (opt == IPOPT_SSRR) {
1407 #define	INA	struct in_ifaddr *
1408 #define	SA	struct sockaddr *
1409 			    if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == 0)
1410 				ia = (INA)ifa_ifwithnet((SA)&ipaddr);
1411 			} else
1412 				ia = ip_rtaddr(ipaddr.sin_addr);
1413 			if (ia == 0) {
1414 				type = ICMP_UNREACH;
1415 				code = ICMP_UNREACH_SRCFAIL;
1416 				goto bad;
1417 			}
1418 			ip->ip_dst = ipaddr.sin_addr;
1419 			(void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
1420 			    sizeof(struct in_addr));
1421 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1422 			/*
1423 			 * Let ip_intr's mcast routing check handle mcast pkts
1424 			 */
1425 			forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr));
1426 			break;
1427 
1428 		case IPOPT_RR:
1429 #ifdef IPSTEALTH
1430 			if (ipstealth && pass == 0)
1431 				break;
1432 #endif
1433 			if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1434 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1435 				goto bad;
1436 			}
1437 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1438 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1439 				goto bad;
1440 			}
1441 			/*
1442 			 * If no space remains, ignore.
1443 			 */
1444 			off--;			/* 0 origin */
1445 			if (off > optlen - (int)sizeof(struct in_addr))
1446 				break;
1447 			(void)memcpy(&ipaddr.sin_addr, &ip->ip_dst,
1448 			    sizeof(ipaddr.sin_addr));
1449 			/*
1450 			 * locate outgoing interface; if we're the destination,
1451 			 * use the incoming interface (should be same).
1452 			 */
1453 			if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == 0 &&
1454 			    (ia = ip_rtaddr(ipaddr.sin_addr)) == 0) {
1455 				type = ICMP_UNREACH;
1456 				code = ICMP_UNREACH_HOST;
1457 				goto bad;
1458 			}
1459 			(void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
1460 			    sizeof(struct in_addr));
1461 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1462 			break;
1463 
1464 		case IPOPT_TS:
1465 #ifdef IPSTEALTH
1466 			if (ipstealth && pass == 0)
1467 				break;
1468 #endif
1469 			code = cp - (u_char *)ip;
1470 			if (optlen < 4 || optlen > 40) {
1471 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1472 				goto bad;
1473 			}
1474 			if ((off = cp[IPOPT_OFFSET]) < 5) {
1475 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1476 				goto bad;
1477 			}
1478 			if (off > optlen - (int)sizeof(int32_t)) {
1479 				cp[IPOPT_OFFSET + 1] += (1 << 4);
1480 				if ((cp[IPOPT_OFFSET + 1] & 0xf0) == 0) {
1481 					code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1482 					goto bad;
1483 				}
1484 				break;
1485 			}
1486 			off--;				/* 0 origin */
1487 			sin = (struct in_addr *)(cp + off);
1488 			switch (cp[IPOPT_OFFSET + 1] & 0x0f) {
1489 
1490 			case IPOPT_TS_TSONLY:
1491 				break;
1492 
1493 			case IPOPT_TS_TSANDADDR:
1494 				if (off + sizeof(n_time) +
1495 				    sizeof(struct in_addr) > optlen) {
1496 					code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1497 					goto bad;
1498 				}
1499 				ipaddr.sin_addr = dst;
1500 				ia = (INA)ifaof_ifpforaddr((SA)&ipaddr,
1501 							    m->m_pkthdr.rcvif);
1502 				if (ia == 0)
1503 					continue;
1504 				(void)memcpy(sin, &IA_SIN(ia)->sin_addr,
1505 				    sizeof(struct in_addr));
1506 				cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1507 				off += sizeof(struct in_addr);
1508 				break;
1509 
1510 			case IPOPT_TS_PRESPEC:
1511 				if (off + sizeof(n_time) +
1512 				    sizeof(struct in_addr) > optlen) {
1513 					code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1514 					goto bad;
1515 				}
1516 				(void)memcpy(&ipaddr.sin_addr, sin,
1517 				    sizeof(struct in_addr));
1518 				if (ifa_ifwithaddr((SA)&ipaddr) == 0)
1519 					continue;
1520 				cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1521 				off += sizeof(struct in_addr);
1522 				break;
1523 
1524 			default:
1525 				code = &cp[IPOPT_OFFSET + 1] - (u_char *)ip;
1526 				goto bad;
1527 			}
1528 			ntime = iptime();
1529 			(void)memcpy(cp + off, &ntime, sizeof(n_time));
1530 			cp[IPOPT_OFFSET] += sizeof(n_time);
1531 		}
1532 	}
1533 	if (forward && ipforwarding) {
1534 		ip_forward(m, 1, next_hop);
1535 		return (1);
1536 	}
1537 	return (0);
1538 bad:
1539 	icmp_error(m, type, code, 0, 0);
1540 	ipstat.ips_badoptions++;
1541 	return (1);
1542 }
1543 
1544 /*
1545  * Given address of next destination (final or next hop),
1546  * return internet address info of interface to be used to get there.
1547  */
1548 struct in_ifaddr *
1549 ip_rtaddr(dst)
1550 	struct in_addr dst;
1551 {
1552 	struct route sro;
1553 	struct sockaddr_in *sin;
1554 	struct in_ifaddr *ifa;
1555 
1556 	bzero(&sro, sizeof(sro));
1557 	sin = (struct sockaddr_in *)&sro.ro_dst;
1558 	sin->sin_family = AF_INET;
1559 	sin->sin_len = sizeof(*sin);
1560 	sin->sin_addr = dst;
1561 	rtalloc_ign(&sro, RTF_CLONING);
1562 
1563 	if (sro.ro_rt == NULL)
1564 		return ((struct in_ifaddr *)0);
1565 
1566 	ifa = ifatoia(sro.ro_rt->rt_ifa);
1567 	RTFREE(sro.ro_rt);
1568 	return ifa;
1569 }
1570 
1571 /*
1572  * Save incoming source route for use in replies,
1573  * to be picked up later by ip_srcroute if the receiver is interested.
1574  */
1575 static void
1576 save_rte(option, dst)
1577 	u_char *option;
1578 	struct in_addr dst;
1579 {
1580 	unsigned olen;
1581 
1582 	olen = option[IPOPT_OLEN];
1583 #ifdef DIAGNOSTIC
1584 	if (ipprintfs)
1585 		printf("save_rte: olen %d\n", olen);
1586 #endif
1587 	if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst)))
1588 		return;
1589 	bcopy(option, ip_srcrt.srcopt, olen);
1590 	ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1591 	ip_srcrt.dst = dst;
1592 }
1593 
1594 /*
1595  * Retrieve incoming source route for use in replies,
1596  * in the same form used by setsockopt.
1597  * The first hop is placed before the options, will be removed later.
1598  */
1599 struct mbuf *
1600 ip_srcroute()
1601 {
1602 	register struct in_addr *p, *q;
1603 	register struct mbuf *m;
1604 
1605 	if (ip_nhops == 0)
1606 		return ((struct mbuf *)0);
1607 	m = m_get(M_DONTWAIT, MT_HEADER);
1608 	if (m == 0)
1609 		return ((struct mbuf *)0);
1610 
1611 #define OPTSIZ	(sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt))
1612 
1613 	/* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
1614 	m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) +
1615 	    OPTSIZ;
1616 #ifdef DIAGNOSTIC
1617 	if (ipprintfs)
1618 		printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len);
1619 #endif
1620 
1621 	/*
1622 	 * First save first hop for return route
1623 	 */
1624 	p = &ip_srcrt.route[ip_nhops - 1];
1625 	*(mtod(m, struct in_addr *)) = *p--;
1626 #ifdef DIAGNOSTIC
1627 	if (ipprintfs)
1628 		printf(" hops %lx", (u_long)ntohl(mtod(m, struct in_addr *)->s_addr));
1629 #endif
1630 
1631 	/*
1632 	 * Copy option fields and padding (nop) to mbuf.
1633 	 */
1634 	ip_srcrt.nop = IPOPT_NOP;
1635 	ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
1636 	(void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr),
1637 	    &ip_srcrt.nop, OPTSIZ);
1638 	q = (struct in_addr *)(mtod(m, caddr_t) +
1639 	    sizeof(struct in_addr) + OPTSIZ);
1640 #undef OPTSIZ
1641 	/*
1642 	 * Record return path as an IP source route,
1643 	 * reversing the path (pointers are now aligned).
1644 	 */
1645 	while (p >= ip_srcrt.route) {
1646 #ifdef DIAGNOSTIC
1647 		if (ipprintfs)
1648 			printf(" %lx", (u_long)ntohl(q->s_addr));
1649 #endif
1650 		*q++ = *p--;
1651 	}
1652 	/*
1653 	 * Last hop goes to final destination.
1654 	 */
1655 	*q = ip_srcrt.dst;
1656 #ifdef DIAGNOSTIC
1657 	if (ipprintfs)
1658 		printf(" %lx\n", (u_long)ntohl(q->s_addr));
1659 #endif
1660 	return (m);
1661 }
1662 
1663 /*
1664  * Strip out IP options, at higher
1665  * level protocol in the kernel.
1666  * Second argument is buffer to which options
1667  * will be moved, and return value is their length.
1668  * XXX should be deleted; last arg currently ignored.
1669  */
1670 void
1671 ip_stripoptions(m, mopt)
1672 	register struct mbuf *m;
1673 	struct mbuf *mopt;
1674 {
1675 	register int i;
1676 	struct ip *ip = mtod(m, struct ip *);
1677 	register caddr_t opts;
1678 	int olen;
1679 
1680 	olen = (ip->ip_hl << 2) - sizeof (struct ip);
1681 	opts = (caddr_t)(ip + 1);
1682 	i = m->m_len - (sizeof (struct ip) + olen);
1683 	bcopy(opts + olen, opts, (unsigned)i);
1684 	m->m_len -= olen;
1685 	if (m->m_flags & M_PKTHDR)
1686 		m->m_pkthdr.len -= olen;
1687 	ip->ip_v = IPVERSION;
1688 	ip->ip_hl = sizeof(struct ip) >> 2;
1689 }
1690 
1691 u_char inetctlerrmap[PRC_NCMDS] = {
1692 	0,		0,		0,		0,
1693 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1694 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1695 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1696 	0,		0,		EHOSTUNREACH,	0,
1697 	ENOPROTOOPT,	ECONNREFUSED
1698 };
1699 
1700 /*
1701  * Forward a packet.  If some error occurs return the sender
1702  * an icmp packet.  Note we can't always generate a meaningful
1703  * icmp message because icmp doesn't have a large enough repertoire
1704  * of codes and types.
1705  *
1706  * If not forwarding, just drop the packet.  This could be confusing
1707  * if ipforwarding was zero but some routing protocol was advancing
1708  * us as a gateway to somewhere.  However, we must let the routing
1709  * protocol deal with that.
1710  *
1711  * The srcrt parameter indicates whether the packet is being forwarded
1712  * via a source route.
1713  */
1714 static void
1715 ip_forward(struct mbuf *m, int srcrt, struct sockaddr_in *next_hop)
1716 {
1717 	struct ip *ip = mtod(m, struct ip *);
1718 	struct in_ifaddr *ia;
1719 	int error, type = 0, code = 0;
1720 	struct mbuf *mcopy;
1721 	n_long dest;
1722 	struct in_addr pkt_dst;
1723 	struct ifnet *destifp;
1724 #if defined(IPSEC) || defined(FAST_IPSEC)
1725 	struct ifnet dummyifp;
1726 #endif
1727 
1728 	/*
1729 	 * Cache the destination address of the packet; this may be
1730 	 * changed by use of 'ipfw fwd'.
1731 	 */
1732 	pkt_dst = next_hop ? next_hop->sin_addr : ip->ip_dst;
1733 
1734 #ifdef DIAGNOSTIC
1735 	if (ipprintfs)
1736 		printf("forward: src %lx dst %lx ttl %x\n",
1737 		    (u_long)ip->ip_src.s_addr, (u_long)pkt_dst.s_addr,
1738 		    ip->ip_ttl);
1739 #endif
1740 
1741 
1742 	if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(pkt_dst) == 0) {
1743 		ipstat.ips_cantforward++;
1744 		m_freem(m);
1745 		return;
1746 	}
1747 #ifdef IPSTEALTH
1748 	if (!ipstealth) {
1749 #endif
1750 		if (ip->ip_ttl <= IPTTLDEC) {
1751 			icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS,
1752 			    0, 0);
1753 			return;
1754 		}
1755 #ifdef IPSTEALTH
1756 	}
1757 #endif
1758 
1759 	if ((ia = ip_rtaddr(pkt_dst)) == 0) {
1760 		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
1761 		return;
1762 	}
1763 
1764 	/*
1765 	 * Save the IP header and at most 8 bytes of the payload,
1766 	 * in case we need to generate an ICMP message to the src.
1767 	 *
1768 	 * XXX this can be optimized a lot by saving the data in a local
1769 	 * buffer on the stack (72 bytes at most), and only allocating the
1770 	 * mbuf if really necessary. The vast majority of the packets
1771 	 * are forwarded without having to send an ICMP back (either
1772 	 * because unnecessary, or because rate limited), so we are
1773 	 * really we are wasting a lot of work here.
1774 	 *
1775 	 * We don't use m_copy() because it might return a reference
1776 	 * to a shared cluster. Both this function and ip_output()
1777 	 * assume exclusive access to the IP header in `m', so any
1778 	 * data in a cluster may change before we reach icmp_error().
1779 	 */
1780 	MGET(mcopy, M_DONTWAIT, m->m_type);
1781 	if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_DONTWAIT)) {
1782 		/*
1783 		 * It's probably ok if the pkthdr dup fails (because
1784 		 * the deep copy of the tag chain failed), but for now
1785 		 * be conservative and just discard the copy since
1786 		 * code below may some day want the tags.
1787 		 */
1788 		m_free(mcopy);
1789 		mcopy = NULL;
1790 	}
1791 	if (mcopy != NULL) {
1792 		mcopy->m_len = imin((ip->ip_hl << 2) + 8,
1793 		    (int)ip->ip_len);
1794 		m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1795 	}
1796 
1797 #ifdef IPSTEALTH
1798 	if (!ipstealth) {
1799 #endif
1800 		ip->ip_ttl -= IPTTLDEC;
1801 #ifdef IPSTEALTH
1802 	}
1803 #endif
1804 
1805 	/*
1806 	 * If forwarding packet using same interface that it came in on,
1807 	 * perhaps should send a redirect to sender to shortcut a hop.
1808 	 * Only send redirect if source is sending directly to us,
1809 	 * and if packet was not source routed (or has any options).
1810 	 * Also, don't send redirect if forwarding using a default route
1811 	 * or a route modified by a redirect.
1812 	 */
1813 	dest = 0;
1814 	if (ipsendredirects && ia->ia_ifp == m->m_pkthdr.rcvif) {
1815 		struct sockaddr_in *sin;
1816 		struct route ro;
1817 		struct rtentry *rt;
1818 
1819 		bzero(&ro, sizeof(ro));
1820 		sin = (struct sockaddr_in *)&ro.ro_dst;
1821 		sin->sin_family = AF_INET;
1822 		sin->sin_len = sizeof(*sin);
1823 		sin->sin_addr = pkt_dst;
1824 		rtalloc_ign(&ro, RTF_CLONING);
1825 
1826 		rt = ro.ro_rt;
1827 
1828 		if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1829 		    satosin(rt_key(rt))->sin_addr.s_addr != 0 &&
1830 		    ipsendredirects && !srcrt && !next_hop) {
1831 #define	RTA(rt)	((struct in_ifaddr *)(rt->rt_ifa))
1832 			u_long src = ntohl(ip->ip_src.s_addr);
1833 
1834 			if (RTA(rt) &&
1835 			    (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1836 				if (rt->rt_flags & RTF_GATEWAY)
1837 					dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
1838 				else
1839 					dest = pkt_dst.s_addr;
1840 				/* Router requirements says to only send host redirects */
1841 				type = ICMP_REDIRECT;
1842 				code = ICMP_REDIRECT_HOST;
1843 #ifdef DIAGNOSTIC
1844 				if (ipprintfs)
1845 					printf("redirect (%d) to %lx\n", code, (u_long)dest);
1846 #endif
1847 			}
1848 		}
1849 		if (rt)
1850 			RTFREE(rt);
1851 	}
1852 
1853 	if (next_hop) {
1854 		struct m_tag *mtag = m_tag_get(PACKET_TAG_IPFORWARD,
1855 		    sizeof(struct sockaddr_in *), M_NOWAIT);
1856 		if (mtag == NULL) {
1857 			m_freem(m);
1858 			return;
1859 		}
1860 		*(struct sockaddr_in **)(mtag+1) = next_hop;
1861 		m_tag_prepend(m, mtag);
1862 	}
1863 	error = ip_output(m, (struct mbuf *)0, NULL, IP_FORWARDING, 0, NULL);
1864 	if (error)
1865 		ipstat.ips_cantforward++;
1866 	else {
1867 		ipstat.ips_forward++;
1868 		if (type)
1869 			ipstat.ips_redirectsent++;
1870 		else {
1871 			if (mcopy)
1872 				m_freem(mcopy);
1873 			return;
1874 		}
1875 	}
1876 	if (mcopy == NULL)
1877 		return;
1878 	destifp = NULL;
1879 
1880 	switch (error) {
1881 
1882 	case 0:				/* forwarded, but need redirect */
1883 		/* type, code set above */
1884 		break;
1885 
1886 	case ENETUNREACH:		/* shouldn't happen, checked above */
1887 	case EHOSTUNREACH:
1888 	case ENETDOWN:
1889 	case EHOSTDOWN:
1890 	default:
1891 		type = ICMP_UNREACH;
1892 		code = ICMP_UNREACH_HOST;
1893 		break;
1894 
1895 	case EMSGSIZE:
1896 		type = ICMP_UNREACH;
1897 		code = ICMP_UNREACH_NEEDFRAG;
1898 #if defined(IPSEC) || defined(FAST_IPSEC)
1899 		/*
1900 		 * If the packet is routed over IPsec tunnel, tell the
1901 		 * originator the tunnel MTU.
1902 		 *	tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
1903 		 * XXX quickhack!!!
1904 		 */
1905 		{
1906 			struct secpolicy *sp = NULL;
1907 			int ipsecerror;
1908 			int ipsechdr;
1909 			struct route *ro;
1910 
1911 #ifdef IPSEC
1912 			sp = ipsec4_getpolicybyaddr(mcopy,
1913 						    IPSEC_DIR_OUTBOUND,
1914 						    IP_FORWARDING,
1915 						    &ipsecerror);
1916 #else /* FAST_IPSEC */
1917 			sp = ipsec_getpolicybyaddr(mcopy,
1918 						   IPSEC_DIR_OUTBOUND,
1919 						   IP_FORWARDING,
1920 						   &ipsecerror);
1921 #endif
1922 			if (sp != NULL) {
1923 				/* count IPsec header size */
1924 				ipsechdr = ipsec4_hdrsiz(mcopy,
1925 							 IPSEC_DIR_OUTBOUND,
1926 							 NULL);
1927 
1928 				/*
1929 				 * find the correct route for outer IPv4
1930 				 * header, compute tunnel MTU.
1931 				 *
1932 				 * XXX BUG ALERT
1933 				 * The "dummyifp" code relies upon the fact
1934 				 * that icmp_error() touches only ifp->if_mtu.
1935 				 */
1936 				/*XXX*/
1937 				destifp = NULL;
1938 				if (sp->req != NULL
1939 				 && sp->req->sav != NULL
1940 				 && sp->req->sav->sah != NULL) {
1941 					ro = &sp->req->sav->sah->sa_route;
1942 					if (ro->ro_rt && ro->ro_rt->rt_ifp) {
1943 						dummyifp.if_mtu =
1944 						    ro->ro_rt->rt_ifp->if_mtu;
1945 						dummyifp.if_mtu -= ipsechdr;
1946 						destifp = &dummyifp;
1947 					}
1948 				}
1949 
1950 #ifdef IPSEC
1951 				key_freesp(sp);
1952 #else /* FAST_IPSEC */
1953 				KEY_FREESP(&sp);
1954 #endif
1955 				ipstat.ips_cantfrag++;
1956 				break;
1957 			} else
1958 #endif /*IPSEC || FAST_IPSEC*/
1959 		destifp = ia->ia_ifp;
1960 #if defined(IPSEC) || defined(FAST_IPSEC)
1961 		}
1962 #endif /*IPSEC || FAST_IPSEC*/
1963 		ipstat.ips_cantfrag++;
1964 		break;
1965 
1966 	case ENOBUFS:
1967 		/*
1968 		 * A router should not generate ICMP_SOURCEQUENCH as
1969 		 * required in RFC1812 Requirements for IP Version 4 Routers.
1970 		 * Source quench could be a big problem under DoS attacks,
1971 		 * or if the underlying interface is rate-limited.
1972 		 * Those who need source quench packets may re-enable them
1973 		 * via the net.inet.ip.sendsourcequench sysctl.
1974 		 */
1975 		if (ip_sendsourcequench == 0) {
1976 			m_freem(mcopy);
1977 			return;
1978 		} else {
1979 			type = ICMP_SOURCEQUENCH;
1980 			code = 0;
1981 		}
1982 		break;
1983 
1984 	case EACCES:			/* ipfw denied packet */
1985 		m_freem(mcopy);
1986 		return;
1987 	}
1988 	icmp_error(mcopy, type, code, dest, destifp);
1989 }
1990 
1991 void
1992 ip_savecontrol(inp, mp, ip, m)
1993 	register struct inpcb *inp;
1994 	register struct mbuf **mp;
1995 	register struct ip *ip;
1996 	register struct mbuf *m;
1997 {
1998 	if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) {
1999 		struct bintime bt;
2000 
2001 		bintime(&bt);
2002 		if (inp->inp_socket->so_options & SO_BINTIME) {
2003 			*mp = sbcreatecontrol((caddr_t) &bt, sizeof(bt),
2004 			SCM_BINTIME, SOL_SOCKET);
2005 			if (*mp)
2006 				mp = &(*mp)->m_next;
2007 		}
2008 		if (inp->inp_socket->so_options & SO_TIMESTAMP) {
2009 			struct timeval tv;
2010 
2011 			bintime2timeval(&bt, &tv);
2012 			*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
2013 				SCM_TIMESTAMP, SOL_SOCKET);
2014 			if (*mp)
2015 				mp = &(*mp)->m_next;
2016 		}
2017 	}
2018 	if (inp->inp_flags & INP_RECVDSTADDR) {
2019 		*mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
2020 		    sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
2021 		if (*mp)
2022 			mp = &(*mp)->m_next;
2023 	}
2024 	if (inp->inp_flags & INP_RECVTTL) {
2025 		*mp = sbcreatecontrol((caddr_t) &ip->ip_ttl,
2026 		    sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
2027 		if (*mp)
2028 			mp = &(*mp)->m_next;
2029 	}
2030 #ifdef notyet
2031 	/* XXX
2032 	 * Moving these out of udp_input() made them even more broken
2033 	 * than they already were.
2034 	 */
2035 	/* options were tossed already */
2036 	if (inp->inp_flags & INP_RECVOPTS) {
2037 		*mp = sbcreatecontrol((caddr_t) opts_deleted_above,
2038 		    sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
2039 		if (*mp)
2040 			mp = &(*mp)->m_next;
2041 	}
2042 	/* ip_srcroute doesn't do what we want here, need to fix */
2043 	if (inp->inp_flags & INP_RECVRETOPTS) {
2044 		*mp = sbcreatecontrol((caddr_t) ip_srcroute(),
2045 		    sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
2046 		if (*mp)
2047 			mp = &(*mp)->m_next;
2048 	}
2049 #endif
2050 	if (inp->inp_flags & INP_RECVIF) {
2051 		struct ifnet *ifp;
2052 		struct sdlbuf {
2053 			struct sockaddr_dl sdl;
2054 			u_char	pad[32];
2055 		} sdlbuf;
2056 		struct sockaddr_dl *sdp;
2057 		struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
2058 
2059 		if (((ifp = m->m_pkthdr.rcvif))
2060 		&& ( ifp->if_index && (ifp->if_index <= if_index))) {
2061 			sdp = (struct sockaddr_dl *)
2062 			    (ifaddr_byindex(ifp->if_index)->ifa_addr);
2063 			/*
2064 			 * Change our mind and don't try copy.
2065 			 */
2066 			if ((sdp->sdl_family != AF_LINK)
2067 			|| (sdp->sdl_len > sizeof(sdlbuf))) {
2068 				goto makedummy;
2069 			}
2070 			bcopy(sdp, sdl2, sdp->sdl_len);
2071 		} else {
2072 makedummy:
2073 			sdl2->sdl_len
2074 				= offsetof(struct sockaddr_dl, sdl_data[0]);
2075 			sdl2->sdl_family = AF_LINK;
2076 			sdl2->sdl_index = 0;
2077 			sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
2078 		}
2079 		*mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len,
2080 			IP_RECVIF, IPPROTO_IP);
2081 		if (*mp)
2082 			mp = &(*mp)->m_next;
2083 	}
2084 }
2085 
2086 /*
2087  * XXX these routines are called from the upper part of the kernel.
2088  * They need to be locked when we remove Giant.
2089  *
2090  * They could also be moved to ip_mroute.c, since all the RSVP
2091  *  handling is done there already.
2092  */
2093 static int ip_rsvp_on;
2094 struct socket *ip_rsvpd;
2095 int
2096 ip_rsvp_init(struct socket *so)
2097 {
2098 	if (so->so_type != SOCK_RAW ||
2099 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
2100 		return EOPNOTSUPP;
2101 
2102 	if (ip_rsvpd != NULL)
2103 		return EADDRINUSE;
2104 
2105 	ip_rsvpd = so;
2106 	/*
2107 	 * This may seem silly, but we need to be sure we don't over-increment
2108 	 * the RSVP counter, in case something slips up.
2109 	 */
2110 	if (!ip_rsvp_on) {
2111 		ip_rsvp_on = 1;
2112 		rsvp_on++;
2113 	}
2114 
2115 	return 0;
2116 }
2117 
2118 int
2119 ip_rsvp_done(void)
2120 {
2121 	ip_rsvpd = NULL;
2122 	/*
2123 	 * This may seem silly, but we need to be sure we don't over-decrement
2124 	 * the RSVP counter, in case something slips up.
2125 	 */
2126 	if (ip_rsvp_on) {
2127 		ip_rsvp_on = 0;
2128 		rsvp_on--;
2129 	}
2130 	return 0;
2131 }
2132 
2133 void
2134 rsvp_input(struct mbuf *m, int off)	/* XXX must fixup manually */
2135 {
2136 	if (rsvp_input_p) { /* call the real one if loaded */
2137 		rsvp_input_p(m, off);
2138 		return;
2139 	}
2140 
2141 	/* Can still get packets with rsvp_on = 0 if there is a local member
2142 	 * of the group to which the RSVP packet is addressed.  But in this
2143 	 * case we want to throw the packet away.
2144 	 */
2145 
2146 	if (!rsvp_on) {
2147 		m_freem(m);
2148 		return;
2149 	}
2150 
2151 	if (ip_rsvpd != NULL) {
2152 		rip_input(m, off);
2153 		return;
2154 	}
2155 	/* Drop the packet */
2156 	m_freem(m);
2157 }
2158