xref: /freebsd/sys/netinet/ip_input.c (revision 7660b554bc59a07be0431c17e0e33815818baa69)
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_dummynet.h>
82 
83 #ifdef IPSEC
84 #include <netinet6/ipsec.h>
85 #include <netkey/key.h>
86 #endif
87 
88 #ifdef FAST_IPSEC
89 #include <netipsec/ipsec.h>
90 #include <netipsec/key.h>
91 #endif
92 
93 int rsvp_on = 0;
94 
95 int	ipforwarding = 0;
96 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW,
97     &ipforwarding, 0, "Enable IP forwarding between interfaces");
98 
99 static int	ipsendredirects = 1; /* XXX */
100 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW,
101     &ipsendredirects, 0, "Enable sending IP redirects");
102 
103 int	ip_defttl = IPDEFTTL;
104 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW,
105     &ip_defttl, 0, "Maximum TTL on IP packets");
106 
107 static int	ip_dosourceroute = 0;
108 SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW,
109     &ip_dosourceroute, 0, "Enable forwarding source routed IP packets");
110 
111 static int	ip_acceptsourceroute = 0;
112 SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute,
113     CTLFLAG_RW, &ip_acceptsourceroute, 0,
114     "Enable accepting source routed IP packets");
115 
116 static int	ip_keepfaith = 0;
117 SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW,
118 	&ip_keepfaith,	0,
119 	"Enable packet capture for FAITH IPv4->IPv6 translater daemon");
120 
121 static int    nipq = 0;         /* total # of reass queues */
122 static int    maxnipq;
123 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragpackets, CTLFLAG_RW,
124 	&maxnipq, 0,
125 	"Maximum number of IPv4 fragment reassembly queue entries");
126 
127 static int    maxfragsperpacket;
128 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW,
129 	&maxfragsperpacket, 0,
130 	"Maximum number of IPv4 fragments allowed per packet");
131 
132 static int	ip_sendsourcequench = 0;
133 SYSCTL_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW,
134 	&ip_sendsourcequench, 0,
135 	"Enable the transmission of source quench packets");
136 
137 /*
138  * XXX - Setting ip_checkinterface mostly implements the receive side of
139  * the Strong ES model described in RFC 1122, but since the routing table
140  * and transmit implementation do not implement the Strong ES model,
141  * setting this to 1 results in an odd hybrid.
142  *
143  * XXX - ip_checkinterface currently must be disabled if you use ipnat
144  * to translate the destination address to another local interface.
145  *
146  * XXX - ip_checkinterface must be disabled if you add IP aliases
147  * to the loopback interface instead of the interface where the
148  * packets for those addresses are received.
149  */
150 static int	ip_checkinterface = 1;
151 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW,
152     &ip_checkinterface, 0, "Verify packet arrives on correct interface");
153 
154 #ifdef DIAGNOSTIC
155 static int	ipprintfs = 0;
156 #endif
157 
158 static struct	ifqueue ipintrq;
159 static int	ipqmaxlen = IFQ_MAXLEN;
160 
161 extern	struct domain inetdomain;
162 extern	struct protosw inetsw[];
163 u_char	ip_protox[IPPROTO_MAX];
164 struct	in_ifaddrhead in_ifaddrhead; 		/* first inet address */
165 struct	in_ifaddrhashhead *in_ifaddrhashtbl;	/* inet addr hash table  */
166 u_long 	in_ifaddrhmask;				/* mask for hash table */
167 
168 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW,
169     &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue");
170 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD,
171     &ipintrq.ifq_drops, 0, "Number of packets dropped from the IP input queue");
172 
173 struct ipstat ipstat;
174 SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW,
175     &ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)");
176 
177 /* Packet reassembly stuff */
178 #define IPREASS_NHASH_LOG2      6
179 #define IPREASS_NHASH           (1 << IPREASS_NHASH_LOG2)
180 #define IPREASS_HMASK           (IPREASS_NHASH - 1)
181 #define IPREASS_HASH(x,y) \
182 	(((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK)
183 
184 static TAILQ_HEAD(ipqhead, ipq) ipq[IPREASS_NHASH];
185 struct mtx ipqlock;
186 
187 #define	IPQ_LOCK()	mtx_lock(&ipqlock)
188 #define	IPQ_UNLOCK()	mtx_unlock(&ipqlock)
189 #define	IPQ_LOCK_INIT()	mtx_init(&ipqlock, "ipqlock", NULL, MTX_DEF);
190 #define	IPQ_LOCK_ASSERT()	mtx_assert(&ipqlock, MA_OWNED);
191 
192 #ifdef IPCTL_DEFMTU
193 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
194     &ip_mtu, 0, "Default MTU");
195 #endif
196 
197 #ifdef IPSTEALTH
198 static int	ipstealth = 0;
199 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW,
200     &ipstealth, 0, "");
201 #endif
202 
203 
204 /* Firewall hooks */
205 ip_fw_chk_t *ip_fw_chk_ptr;
206 int fw_enable = 1 ;
207 int fw_one_pass = 1;
208 
209 /* Dummynet hooks */
210 ip_dn_io_t *ip_dn_io_ptr;
211 
212 
213 /*
214  * XXX this is ugly -- the following two global variables are
215  * used to store packet state while it travels through the stack.
216  * Note that the code even makes assumptions on the size and
217  * alignment of fields inside struct ip_srcrt so e.g. adding some
218  * fields will break the code. This needs to be fixed.
219  *
220  * We need to save the IP options in case a protocol wants to respond
221  * to an incoming packet over the same route if the packet got here
222  * using IP source routing.  This allows connection establishment and
223  * maintenance when the remote end is on a network that is not known
224  * to us.
225  */
226 static int	ip_nhops = 0;
227 static	struct ip_srcrt {
228 	struct	in_addr dst;			/* final destination */
229 	char	nop;				/* one NOP to align */
230 	char	srcopt[IPOPT_OFFSET + 1];	/* OPTVAL, OLEN and OFFSET */
231 	struct	in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)];
232 } ip_srcrt;
233 
234 static void	save_rte(u_char *, struct in_addr);
235 static int	ip_dooptions(struct mbuf *m, int,
236 			struct sockaddr_in *next_hop);
237 static void	ip_forward(struct mbuf *m, int srcrt,
238 			struct sockaddr_in *next_hop);
239 static void	ip_freef(struct ipqhead *, struct ipq *);
240 static struct	mbuf *ip_reass(struct mbuf *, struct ipqhead *,
241 		struct ipq *, u_int32_t *, u_int16_t *);
242 
243 /*
244  * IP initialization: fill in IP protocol switch table.
245  * All protocols not implemented in kernel go to raw IP protocol handler.
246  */
247 void
248 ip_init()
249 {
250 	register struct protosw *pr;
251 	register int i;
252 
253 	TAILQ_INIT(&in_ifaddrhead);
254 	in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask);
255 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
256 	if (pr == 0)
257 		panic("ip_init");
258 	for (i = 0; i < IPPROTO_MAX; i++)
259 		ip_protox[i] = pr - inetsw;
260 	for (pr = inetdomain.dom_protosw;
261 	    pr < inetdomain.dom_protoswNPROTOSW; pr++)
262 		if (pr->pr_domain->dom_family == PF_INET &&
263 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
264 			ip_protox[pr->pr_protocol] = pr - inetsw;
265 
266 	IPQ_LOCK_INIT();
267 	for (i = 0; i < IPREASS_NHASH; i++)
268 	    TAILQ_INIT(&ipq[i]);
269 
270 	maxnipq = nmbclusters / 32;
271 	maxfragsperpacket = 16;
272 
273 #ifndef RANDOM_IP_ID
274 	ip_id = time_second & 0xffff;
275 #endif
276 	ipintrq.ifq_maxlen = ipqmaxlen;
277 	mtx_init(&ipintrq.ifq_mtx, "ip_inq", NULL, MTX_DEF);
278 	netisr_register(NETISR_IP, ip_input, &ipintrq);
279 }
280 
281 /*
282  * XXX watch out this one. It is perhaps used as a cache for
283  * the most recently used route ? it is cleared in in_addroute()
284  * when a new route is successfully created.
285  */
286 struct	route ipforward_rt;
287 
288 /*
289  * Ip input routine.  Checksum and byte swap header.  If fragmented
290  * try to reassemble.  Process options.  Pass to next level.
291  */
292 void
293 ip_input(struct mbuf *m)
294 {
295 	struct ip *ip;
296 	struct ipq *fp;
297 	struct in_ifaddr *ia = NULL;
298 	struct ifaddr *ifa;
299 	int    i, hlen, checkif;
300 	u_short sum;
301 	struct in_addr pkt_dst;
302 	u_int32_t divert_info = 0;		/* packet divert/tee info */
303 	struct ip_fw_args args;
304 #ifdef PFIL_HOOKS
305 	struct packet_filter_hook *pfh;
306 	struct mbuf *m0;
307 	int rv;
308 #endif /* PFIL_HOOKS */
309 #ifdef FAST_IPSEC
310 	struct m_tag *mtag;
311 	struct tdb_ident *tdbi;
312 	struct secpolicy *sp;
313 	int s, error;
314 #endif /* FAST_IPSEC */
315 
316 	args.eh = NULL;
317 	args.oif = NULL;
318 	args.rule = NULL;
319 	args.divert_rule = 0;			/* divert cookie */
320 	args.next_hop = NULL;
321 
322 	/* Grab info from MT_TAG mbufs prepended to the chain.	*/
323 	for (; m && m->m_type == MT_TAG; m = m->m_next) {
324 		switch(m->_m_tag_id) {
325 		default:
326 			printf("ip_input: unrecognised MT_TAG tag %d\n",
327 			    m->_m_tag_id);
328 			break;
329 
330 		case PACKET_TAG_DUMMYNET:
331 			args.rule = ((struct dn_pkt *)m)->rule;
332 			break;
333 
334 		case PACKET_TAG_DIVERT:
335 			args.divert_rule = (intptr_t)m->m_hdr.mh_data & 0xffff;
336 			break;
337 
338 		case PACKET_TAG_IPFORWARD:
339 			args.next_hop = (struct sockaddr_in *)m->m_hdr.mh_data;
340 			break;
341 		}
342 	}
343 
344 	M_ASSERTPKTHDR(m);
345 
346 	if (args.rule) {	/* dummynet already filtered us */
347 		ip = mtod(m, struct ip *);
348 		hlen = ip->ip_hl << 2;
349 		goto iphack ;
350 	}
351 
352 	ipstat.ips_total++;
353 
354 	if (m->m_pkthdr.len < sizeof(struct ip))
355 		goto tooshort;
356 
357 	if (m->m_len < sizeof (struct ip) &&
358 	    (m = m_pullup(m, sizeof (struct ip))) == 0) {
359 		ipstat.ips_toosmall++;
360 		return;
361 	}
362 	ip = mtod(m, struct ip *);
363 
364 	if (ip->ip_v != IPVERSION) {
365 		ipstat.ips_badvers++;
366 		goto bad;
367 	}
368 
369 	hlen = ip->ip_hl << 2;
370 	if (hlen < sizeof(struct ip)) {	/* minimum header length */
371 		ipstat.ips_badhlen++;
372 		goto bad;
373 	}
374 	if (hlen > m->m_len) {
375 		if ((m = m_pullup(m, hlen)) == 0) {
376 			ipstat.ips_badhlen++;
377 			return;
378 		}
379 		ip = mtod(m, struct ip *);
380 	}
381 
382 	/* 127/8 must not appear on wire - RFC1122 */
383 	if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
384 	    (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
385 		if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
386 			ipstat.ips_badaddr++;
387 			goto bad;
388 		}
389 	}
390 
391 	if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
392 		sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
393 	} else {
394 		if (hlen == sizeof(struct ip)) {
395 			sum = in_cksum_hdr(ip);
396 		} else {
397 			sum = in_cksum(m, hlen);
398 		}
399 	}
400 	if (sum) {
401 		ipstat.ips_badsum++;
402 		goto bad;
403 	}
404 
405 	/*
406 	 * Convert fields to host representation.
407 	 */
408 	ip->ip_len = ntohs(ip->ip_len);
409 	if (ip->ip_len < hlen) {
410 		ipstat.ips_badlen++;
411 		goto bad;
412 	}
413 	ip->ip_off = ntohs(ip->ip_off);
414 
415 	/*
416 	 * Check that the amount of data in the buffers
417 	 * is as at least much as the IP header would have us expect.
418 	 * Trim mbufs if longer than we expect.
419 	 * Drop packet if shorter than we expect.
420 	 */
421 	if (m->m_pkthdr.len < ip->ip_len) {
422 tooshort:
423 		ipstat.ips_tooshort++;
424 		goto bad;
425 	}
426 	if (m->m_pkthdr.len > ip->ip_len) {
427 		if (m->m_len == m->m_pkthdr.len) {
428 			m->m_len = ip->ip_len;
429 			m->m_pkthdr.len = ip->ip_len;
430 		} else
431 			m_adj(m, ip->ip_len - m->m_pkthdr.len);
432 	}
433 #if defined(IPSEC) && !defined(IPSEC_FILTERGIF)
434 	/*
435 	 * Bypass packet filtering for packets from a tunnel (gif).
436 	 */
437 	if (ipsec_gethist(m, NULL))
438 		goto pass;
439 #endif
440 #if defined(FAST_IPSEC) && !defined(IPSEC_FILTERGIF)
441 	/*
442 	 * Bypass packet filtering for packets from a tunnel (gif).
443 	 */
444 	if (m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL)
445 		goto pass;
446 #endif
447 
448 	/*
449 	 * IpHack's section.
450 	 * Right now when no processing on packet has done
451 	 * and it is still fresh out of network we do our black
452 	 * deals with it.
453 	 * - Firewall: deny/allow/divert
454 	 * - Xlate: translate packet's addr/port (NAT).
455 	 * - Pipe: pass pkt through dummynet.
456 	 * - Wrap: fake packet's addr/port <unimpl.>
457 	 * - Encapsulate: put it in another IP and send out. <unimp.>
458  	 */
459 
460 iphack:
461 
462 #ifdef PFIL_HOOKS
463 	/*
464 	 * Run through list of hooks for input packets.  If there are any
465 	 * filters which require that additional packets in the flow are
466 	 * not fast-forwarded, they must clear the M_CANFASTFWD flag.
467 	 * Note that filters must _never_ set this flag, as another filter
468 	 * in the list may have previously cleared it.
469 	 */
470 	m0 = m;
471 	pfh = pfil_hook_get(PFIL_IN, &inetsw[ip_protox[IPPROTO_IP]].pr_pfh);
472 	for (; pfh; pfh = TAILQ_NEXT(pfh, pfil_link))
473 		if (pfh->pfil_func) {
474 			rv = pfh->pfil_func(ip, hlen,
475 					    m->m_pkthdr.rcvif, 0, &m0);
476 			if (rv)
477 				return;
478 			m = m0;
479 			if (m == NULL)
480 				return;
481 			ip = mtod(m, struct ip *);
482 		}
483 #endif /* PFIL_HOOKS */
484 
485 	if (fw_enable && IPFW_LOADED) {
486 		/*
487 		 * If we've been forwarded from the output side, then
488 		 * skip the firewall a second time
489 		 */
490 		if (args.next_hop)
491 			goto ours;
492 
493 		args.m = m;
494 		i = ip_fw_chk_ptr(&args);
495 		m = args.m;
496 
497 		if ( (i & IP_FW_PORT_DENY_FLAG) || m == NULL) { /* drop */
498 			if (m)
499 				m_freem(m);
500 			return;
501 		}
502 		ip = mtod(m, struct ip *); /* just in case m changed */
503 		if (i == 0 && args.next_hop == NULL)	/* common case */
504 			goto pass;
505                 if (DUMMYNET_LOADED && (i & IP_FW_PORT_DYNT_FLAG) != 0) {
506 			/* Send packet to the appropriate pipe */
507 			ip_dn_io_ptr(m, i&0xffff, DN_TO_IP_IN, &args);
508 			return;
509 		}
510 #ifdef IPDIVERT
511 		if (i != 0 && (i & IP_FW_PORT_DYNT_FLAG) == 0) {
512 			/* Divert or tee packet */
513 			divert_info = i;
514 			goto ours;
515 		}
516 #endif
517 		if (i == 0 && args.next_hop != NULL)
518 			goto pass;
519 		/*
520 		 * if we get here, the packet must be dropped
521 		 */
522 		m_freem(m);
523 		return;
524 	}
525 pass:
526 
527 	/*
528 	 * Process options and, if not destined for us,
529 	 * ship it on.  ip_dooptions returns 1 when an
530 	 * error was detected (causing an icmp message
531 	 * to be sent and the original packet to be freed).
532 	 */
533 	ip_nhops = 0;		/* for source routed packets */
534 	if (hlen > sizeof (struct ip) && ip_dooptions(m, 0, args.next_hop))
535 		return;
536 
537         /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
538          * matter if it is destined to another node, or whether it is
539          * a multicast one, RSVP wants it! and prevents it from being forwarded
540          * anywhere else. Also checks if the rsvp daemon is running before
541 	 * grabbing the packet.
542          */
543 	if (rsvp_on && ip->ip_p==IPPROTO_RSVP)
544 		goto ours;
545 
546 	/*
547 	 * Check our list of addresses, to see if the packet is for us.
548 	 * If we don't have any addresses, assume any unicast packet
549 	 * we receive might be for us (and let the upper layers deal
550 	 * with it).
551 	 */
552 	if (TAILQ_EMPTY(&in_ifaddrhead) &&
553 	    (m->m_flags & (M_MCAST|M_BCAST)) == 0)
554 		goto ours;
555 
556 	/*
557 	 * Cache the destination address of the packet; this may be
558 	 * changed by use of 'ipfw fwd'.
559 	 */
560 	pkt_dst = args.next_hop ? args.next_hop->sin_addr : ip->ip_dst;
561 
562 	/*
563 	 * Enable a consistency check between the destination address
564 	 * and the arrival interface for a unicast packet (the RFC 1122
565 	 * strong ES model) if IP forwarding is disabled and the packet
566 	 * is not locally generated and the packet is not subject to
567 	 * 'ipfw fwd'.
568 	 *
569 	 * XXX - Checking also should be disabled if the destination
570 	 * address is ipnat'ed to a different interface.
571 	 *
572 	 * XXX - Checking is incompatible with IP aliases added
573 	 * to the loopback interface instead of the interface where
574 	 * the packets are received.
575 	 */
576 	checkif = ip_checkinterface && (ipforwarding == 0) &&
577 	    m->m_pkthdr.rcvif != NULL &&
578 	    ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) &&
579 	    (args.next_hop == NULL);
580 
581 	/*
582 	 * Check for exact addresses in the hash bucket.
583 	 */
584 	LIST_FOREACH(ia, INADDR_HASH(pkt_dst.s_addr), ia_hash) {
585 		/*
586 		 * If the address matches, verify that the packet
587 		 * arrived via the correct interface if checking is
588 		 * enabled.
589 		 */
590 		if (IA_SIN(ia)->sin_addr.s_addr == pkt_dst.s_addr &&
591 		    (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif))
592 			goto ours;
593 	}
594 	/*
595 	 * Check for broadcast addresses.
596 	 *
597 	 * Only accept broadcast packets that arrive via the matching
598 	 * interface.  Reception of forwarded directed broadcasts would
599 	 * be handled via ip_forward() and ether_output() with the loopback
600 	 * into the stack for SIMPLEX interfaces handled by ether_output().
601 	 */
602 	if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
603 	        TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
604 			if (ifa->ifa_addr->sa_family != AF_INET)
605 				continue;
606 			ia = ifatoia(ifa);
607 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
608 			    pkt_dst.s_addr)
609 				goto ours;
610 			if (ia->ia_netbroadcast.s_addr == pkt_dst.s_addr)
611 				goto ours;
612 #ifdef BOOTP_COMPAT
613 			if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY)
614 				goto ours;
615 #endif
616 		}
617 	}
618 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
619 		struct in_multi *inm;
620 		if (ip_mrouter) {
621 			/*
622 			 * If we are acting as a multicast router, all
623 			 * incoming multicast packets are passed to the
624 			 * kernel-level multicast forwarding function.
625 			 * The packet is returned (relatively) intact; if
626 			 * ip_mforward() returns a non-zero value, the packet
627 			 * must be discarded, else it may be accepted below.
628 			 */
629 			if (ip_mforward &&
630 			    ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) {
631 				ipstat.ips_cantforward++;
632 				m_freem(m);
633 				return;
634 			}
635 
636 			/*
637 			 * The process-level routing daemon needs to receive
638 			 * all multicast IGMP packets, whether or not this
639 			 * host belongs to their destination groups.
640 			 */
641 			if (ip->ip_p == IPPROTO_IGMP)
642 				goto ours;
643 			ipstat.ips_forward++;
644 		}
645 		/*
646 		 * See if we belong to the destination multicast group on the
647 		 * arrival interface.
648 		 */
649 		IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
650 		if (inm == NULL) {
651 			ipstat.ips_notmember++;
652 			m_freem(m);
653 			return;
654 		}
655 		goto ours;
656 	}
657 	if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
658 		goto ours;
659 	if (ip->ip_dst.s_addr == INADDR_ANY)
660 		goto ours;
661 
662 	/*
663 	 * FAITH(Firewall Aided Internet Translator)
664 	 */
665 	if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
666 		if (ip_keepfaith) {
667 			if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP)
668 				goto ours;
669 		}
670 		m_freem(m);
671 		return;
672 	}
673 
674 	/*
675 	 * Not for us; forward if possible and desirable.
676 	 */
677 	if (ipforwarding == 0) {
678 		ipstat.ips_cantforward++;
679 		m_freem(m);
680 	} else {
681 #ifdef IPSEC
682 		/*
683 		 * Enforce inbound IPsec SPD.
684 		 */
685 		if (ipsec4_in_reject(m, NULL)) {
686 			ipsecstat.in_polvio++;
687 			goto bad;
688 		}
689 #endif /* IPSEC */
690 #ifdef FAST_IPSEC
691 		mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
692 		s = splnet();
693 		if (mtag != NULL) {
694 			tdbi = (struct tdb_ident *)(mtag + 1);
695 			sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
696 		} else {
697 			sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
698 						   IP_FORWARDING, &error);
699 		}
700 		if (sp == NULL) {	/* NB: can happen if error */
701 			splx(s);
702 			/*XXX error stat???*/
703 			DPRINTF(("ip_input: no SP for forwarding\n"));	/*XXX*/
704 			goto bad;
705 		}
706 
707 		/*
708 		 * Check security policy against packet attributes.
709 		 */
710 		error = ipsec_in_reject(sp, m);
711 		KEY_FREESP(&sp);
712 		splx(s);
713 		if (error) {
714 			ipstat.ips_cantforward++;
715 			goto bad;
716 		}
717 #endif /* FAST_IPSEC */
718 		ip_forward(m, 0, args.next_hop);
719 	}
720 	return;
721 
722 ours:
723 #ifdef IPSTEALTH
724 	/*
725 	 * IPSTEALTH: Process non-routing options only
726 	 * if the packet is destined for us.
727 	 */
728 	if (ipstealth && hlen > sizeof (struct ip) &&
729 	    ip_dooptions(m, 1, args.next_hop))
730 		return;
731 #endif /* IPSTEALTH */
732 
733 	/* Count the packet in the ip address stats */
734 	if (ia != NULL) {
735 		ia->ia_ifa.if_ipackets++;
736 		ia->ia_ifa.if_ibytes += m->m_pkthdr.len;
737 	}
738 
739 	/*
740 	 * If offset or IP_MF are set, must reassemble.
741 	 * Otherwise, nothing need be done.
742 	 * (We could look in the reassembly queue to see
743 	 * if the packet was previously fragmented,
744 	 * but it's not worth the time; just let them time out.)
745 	 */
746 	if (ip->ip_off & (IP_MF | IP_OFFMASK)) {
747 
748 		/* If maxnipq is 0, never accept fragments. */
749 		if (maxnipq == 0) {
750                 	ipstat.ips_fragments++;
751 			ipstat.ips_fragdropped++;
752 			goto bad;
753 		}
754 
755 		sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
756 		IPQ_LOCK();
757 		/*
758 		 * Look for queue of fragments
759 		 * of this datagram.
760 		 */
761 		TAILQ_FOREACH(fp, &ipq[sum], ipq_list)
762 			if (ip->ip_id == fp->ipq_id &&
763 			    ip->ip_src.s_addr == fp->ipq_src.s_addr &&
764 			    ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
765 #ifdef MAC
766 			    mac_fragment_match(m, fp) &&
767 #endif
768 			    ip->ip_p == fp->ipq_p)
769 				goto found;
770 
771 		fp = NULL;
772 
773 		/*
774 		 * Enforce upper bound on number of fragmented packets
775 		 * for which we attempt reassembly;
776 		 * If maxnipq is -1, accept all fragments without limitation.
777 		 */
778 		if ((nipq > maxnipq) && (maxnipq > 0)) {
779 		    /*
780 		     * drop something from the tail of the current queue
781 		     * before proceeding further
782 		     */
783 		    struct ipq *q = TAILQ_LAST(&ipq[sum], ipqhead);
784 		    if (q == NULL) {   /* gak */
785 			for (i = 0; i < IPREASS_NHASH; i++) {
786 			    struct ipq *r = TAILQ_LAST(&ipq[i], ipqhead);
787 			    if (r) {
788 				ipstat.ips_fragtimeout += r->ipq_nfrags;
789 				ip_freef(&ipq[i], r);
790 				break;
791 			    }
792 			}
793 		    } else {
794 			ipstat.ips_fragtimeout += q->ipq_nfrags;
795 			ip_freef(&ipq[sum], q);
796 		    }
797 		}
798 found:
799 		/*
800 		 * Adjust ip_len to not reflect header,
801 		 * convert offset of this to bytes.
802 		 */
803 		ip->ip_len -= hlen;
804 		if (ip->ip_off & IP_MF) {
805 		        /*
806 		         * Make sure that fragments have a data length
807 			 * that's a non-zero multiple of 8 bytes.
808 		         */
809 			if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
810 				IPQ_UNLOCK();
811 				ipstat.ips_toosmall++; /* XXX */
812 				goto bad;
813 			}
814 			m->m_flags |= M_FRAG;
815 		} else
816 			m->m_flags &= ~M_FRAG;
817 		ip->ip_off <<= 3;
818 
819 		/*
820 		 * Attempt reassembly; if it succeeds, proceed.
821 		 * ip_reass() will return a different mbuf, and update
822 		 * the divert info in divert_info and args.divert_rule.
823 		 */
824 		ipstat.ips_fragments++;
825 		m->m_pkthdr.header = ip;
826 		m = ip_reass(m,
827 		    &ipq[sum], fp, &divert_info, &args.divert_rule);
828 		IPQ_UNLOCK();
829 		if (m == 0)
830 			return;
831 		ipstat.ips_reassembled++;
832 		ip = mtod(m, struct ip *);
833 		/* Get the header length of the reassembled packet */
834 		hlen = ip->ip_hl << 2;
835 #ifdef IPDIVERT
836 		/* Restore original checksum before diverting packet */
837 		if (divert_info != 0) {
838 			ip->ip_len += hlen;
839 			ip->ip_len = htons(ip->ip_len);
840 			ip->ip_off = htons(ip->ip_off);
841 			ip->ip_sum = 0;
842 			if (hlen == sizeof(struct ip))
843 				ip->ip_sum = in_cksum_hdr(ip);
844 			else
845 				ip->ip_sum = in_cksum(m, hlen);
846 			ip->ip_off = ntohs(ip->ip_off);
847 			ip->ip_len = ntohs(ip->ip_len);
848 			ip->ip_len -= hlen;
849 		}
850 #endif
851 	} else
852 		ip->ip_len -= hlen;
853 
854 #ifdef IPDIVERT
855 	/*
856 	 * Divert or tee packet to the divert protocol if required.
857 	 */
858 	if (divert_info != 0) {
859 		struct mbuf *clone = NULL;
860 
861 		/* Clone packet if we're doing a 'tee' */
862 		if ((divert_info & IP_FW_PORT_TEE_FLAG) != 0)
863 			clone = m_dup(m, M_DONTWAIT);
864 
865 		/* Restore packet header fields to original values */
866 		ip->ip_len += hlen;
867 		ip->ip_len = htons(ip->ip_len);
868 		ip->ip_off = htons(ip->ip_off);
869 
870 		/* Deliver packet to divert input routine */
871 		divert_packet(m, 1, divert_info & 0xffff, args.divert_rule);
872 		ipstat.ips_delivered++;
873 
874 		/* If 'tee', continue with original packet */
875 		if (clone == NULL)
876 			return;
877 		m = clone;
878 		ip = mtod(m, struct ip *);
879 		ip->ip_len += hlen;
880 		/*
881 		 * Jump backwards to complete processing of the
882 		 * packet. But first clear divert_info to avoid
883 		 * entering this block again.
884 		 * We do not need to clear args.divert_rule
885 		 * or args.next_hop as they will not be used.
886 		 */
887 		divert_info = 0;
888 		goto pass;
889 	}
890 #endif
891 
892 #ifdef IPSEC
893 	/*
894 	 * enforce IPsec policy checking if we are seeing last header.
895 	 * note that we do not visit this with protocols with pcb layer
896 	 * code - like udp/tcp/raw ip.
897 	 */
898 	if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0 &&
899 	    ipsec4_in_reject(m, NULL)) {
900 		ipsecstat.in_polvio++;
901 		goto bad;
902 	}
903 #endif
904 #if FAST_IPSEC
905 	/*
906 	 * enforce IPsec policy checking if we are seeing last header.
907 	 * note that we do not visit this with protocols with pcb layer
908 	 * code - like udp/tcp/raw ip.
909 	 */
910 	if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) {
911 		/*
912 		 * Check if the packet has already had IPsec processing
913 		 * done.  If so, then just pass it along.  This tag gets
914 		 * set during AH, ESP, etc. input handling, before the
915 		 * packet is returned to the ip input queue for delivery.
916 		 */
917 		mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
918 		s = splnet();
919 		if (mtag != NULL) {
920 			tdbi = (struct tdb_ident *)(mtag + 1);
921 			sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
922 		} else {
923 			sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
924 						   IP_FORWARDING, &error);
925 		}
926 		if (sp != NULL) {
927 			/*
928 			 * Check security policy against packet attributes.
929 			 */
930 			error = ipsec_in_reject(sp, m);
931 			KEY_FREESP(&sp);
932 		} else {
933 			/* XXX error stat??? */
934 			error = EINVAL;
935 DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/
936 			goto bad;
937 		}
938 		splx(s);
939 		if (error)
940 			goto bad;
941 	}
942 #endif /* FAST_IPSEC */
943 
944 	/*
945 	 * Switch out to protocol's input routine.
946 	 */
947 	ipstat.ips_delivered++;
948 	if (args.next_hop && ip->ip_p == IPPROTO_TCP) {
949 		/* TCP needs IPFORWARD info if available */
950 		struct m_hdr tag;
951 
952 		tag.mh_type = MT_TAG;
953 		tag.mh_flags = PACKET_TAG_IPFORWARD;
954 		tag.mh_data = (caddr_t)args.next_hop;
955 		tag.mh_next = m;
956 
957 		(*inetsw[ip_protox[ip->ip_p]].pr_input)(
958 			(struct mbuf *)&tag, hlen);
959 	} else
960 		(*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen);
961 	return;
962 bad:
963 	m_freem(m);
964 }
965 
966 /*
967  * Take incoming datagram fragment and try to reassemble it into
968  * whole datagram.  If a chain for reassembly of this datagram already
969  * exists, then it is given as fp; otherwise have to make a chain.
970  *
971  * When IPDIVERT enabled, keep additional state with each packet that
972  * tells us if we need to divert or tee the packet we're building.
973  * In particular, *divinfo includes the port and TEE flag,
974  * *divert_rule is the number of the matching rule.
975  */
976 
977 static struct mbuf *
978 ip_reass(struct mbuf *m, struct ipqhead *head, struct ipq *fp,
979 	u_int32_t *divinfo, u_int16_t *divert_rule)
980 {
981 	struct ip *ip = mtod(m, struct ip *);
982 	register struct mbuf *p, *q, *nq;
983 	struct mbuf *t;
984 	int hlen = ip->ip_hl << 2;
985 	int i, next;
986 
987 	IPQ_LOCK_ASSERT();
988 
989 	/*
990 	 * Presence of header sizes in mbufs
991 	 * would confuse code below.
992 	 */
993 	m->m_data += hlen;
994 	m->m_len -= hlen;
995 
996 	/*
997 	 * If first fragment to arrive, create a reassembly queue.
998 	 */
999 	if (fp == NULL) {
1000 		if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL)
1001 			goto dropfrag;
1002 		fp = mtod(t, struct ipq *);
1003 #ifdef MAC
1004 		if (mac_init_ipq(fp, M_NOWAIT) != 0) {
1005 			m_free(t);
1006 			goto dropfrag;
1007 		}
1008 		mac_create_ipq(m, fp);
1009 #endif
1010 		TAILQ_INSERT_HEAD(head, fp, ipq_list);
1011 		nipq++;
1012 		fp->ipq_nfrags = 1;
1013 		fp->ipq_ttl = IPFRAGTTL;
1014 		fp->ipq_p = ip->ip_p;
1015 		fp->ipq_id = ip->ip_id;
1016 		fp->ipq_src = ip->ip_src;
1017 		fp->ipq_dst = ip->ip_dst;
1018 		fp->ipq_frags = m;
1019 		m->m_nextpkt = NULL;
1020 #ifdef IPDIVERT
1021 		fp->ipq_div_info = 0;
1022 		fp->ipq_div_cookie = 0;
1023 #endif
1024 		goto inserted;
1025 	} else {
1026 		fp->ipq_nfrags++;
1027 #ifdef MAC
1028 		mac_update_ipq(m, fp);
1029 #endif
1030 	}
1031 
1032 #define GETIP(m)	((struct ip*)((m)->m_pkthdr.header))
1033 
1034 	/*
1035 	 * Find a segment which begins after this one does.
1036 	 */
1037 	for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt)
1038 		if (GETIP(q)->ip_off > ip->ip_off)
1039 			break;
1040 
1041 	/*
1042 	 * If there is a preceding segment, it may provide some of
1043 	 * our data already.  If so, drop the data from the incoming
1044 	 * segment.  If it provides all of our data, drop us, otherwise
1045 	 * stick new segment in the proper place.
1046 	 *
1047 	 * If some of the data is dropped from the the preceding
1048 	 * segment, then it's checksum is invalidated.
1049 	 */
1050 	if (p) {
1051 		i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off;
1052 		if (i > 0) {
1053 			if (i >= ip->ip_len)
1054 				goto dropfrag;
1055 			m_adj(m, i);
1056 			m->m_pkthdr.csum_flags = 0;
1057 			ip->ip_off += i;
1058 			ip->ip_len -= i;
1059 		}
1060 		m->m_nextpkt = p->m_nextpkt;
1061 		p->m_nextpkt = m;
1062 	} else {
1063 		m->m_nextpkt = fp->ipq_frags;
1064 		fp->ipq_frags = m;
1065 	}
1066 
1067 	/*
1068 	 * While we overlap succeeding segments trim them or,
1069 	 * if they are completely covered, dequeue them.
1070 	 */
1071 	for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off;
1072 	     q = nq) {
1073 		i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off;
1074 		if (i < GETIP(q)->ip_len) {
1075 			GETIP(q)->ip_len -= i;
1076 			GETIP(q)->ip_off += i;
1077 			m_adj(q, i);
1078 			q->m_pkthdr.csum_flags = 0;
1079 			break;
1080 		}
1081 		nq = q->m_nextpkt;
1082 		m->m_nextpkt = nq;
1083 		ipstat.ips_fragdropped++;
1084 		fp->ipq_nfrags--;
1085 		m_freem(q);
1086 	}
1087 
1088 inserted:
1089 
1090 #ifdef IPDIVERT
1091 	/*
1092 	 * Transfer firewall instructions to the fragment structure.
1093 	 * Only trust info in the fragment at offset 0.
1094 	 */
1095 	if (ip->ip_off == 0) {
1096 		fp->ipq_div_info = *divinfo;
1097 		fp->ipq_div_cookie = *divert_rule;
1098 	}
1099 	*divinfo = 0;
1100 	*divert_rule = 0;
1101 #endif
1102 
1103 	/*
1104 	 * Check for complete reassembly and perform frag per packet
1105 	 * limiting.
1106 	 *
1107 	 * Frag limiting is performed here so that the nth frag has
1108 	 * a chance to complete the packet before we drop the packet.
1109 	 * As a result, n+1 frags are actually allowed per packet, but
1110 	 * only n will ever be stored. (n = maxfragsperpacket.)
1111 	 *
1112 	 */
1113 	next = 0;
1114 	for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
1115 		if (GETIP(q)->ip_off != next) {
1116 			if (fp->ipq_nfrags > maxfragsperpacket) {
1117 				ipstat.ips_fragdropped += fp->ipq_nfrags;
1118 				ip_freef(head, fp);
1119 			}
1120 			return (0);
1121 		}
1122 		next += GETIP(q)->ip_len;
1123 	}
1124 	/* Make sure the last packet didn't have the IP_MF flag */
1125 	if (p->m_flags & M_FRAG) {
1126 		if (fp->ipq_nfrags > maxfragsperpacket) {
1127 			ipstat.ips_fragdropped += fp->ipq_nfrags;
1128 			ip_freef(head, fp);
1129 		}
1130 		return (0);
1131 	}
1132 
1133 	/*
1134 	 * Reassembly is complete.  Make sure the packet is a sane size.
1135 	 */
1136 	q = fp->ipq_frags;
1137 	ip = GETIP(q);
1138 	if (next + (ip->ip_hl << 2) > IP_MAXPACKET) {
1139 		ipstat.ips_toolong++;
1140 		ipstat.ips_fragdropped += fp->ipq_nfrags;
1141 		ip_freef(head, fp);
1142 		return (0);
1143 	}
1144 
1145 	/*
1146 	 * Concatenate fragments.
1147 	 */
1148 	m = q;
1149 	t = m->m_next;
1150 	m->m_next = 0;
1151 	m_cat(m, t);
1152 	nq = q->m_nextpkt;
1153 	q->m_nextpkt = 0;
1154 	for (q = nq; q != NULL; q = nq) {
1155 		nq = q->m_nextpkt;
1156 		q->m_nextpkt = NULL;
1157 		m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
1158 		m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
1159 		m_cat(m, q);
1160 	}
1161 #ifdef MAC
1162 	mac_create_datagram_from_ipq(fp, m);
1163 	mac_destroy_ipq(fp);
1164 #endif
1165 
1166 #ifdef IPDIVERT
1167 	/*
1168 	 * Extract firewall instructions from the fragment structure.
1169 	 */
1170 	*divinfo = fp->ipq_div_info;
1171 	*divert_rule = fp->ipq_div_cookie;
1172 #endif
1173 
1174 	/*
1175 	 * Create header for new ip packet by
1176 	 * modifying header of first packet;
1177 	 * dequeue and discard fragment reassembly header.
1178 	 * Make header visible.
1179 	 */
1180 	ip->ip_len = next;
1181 	ip->ip_src = fp->ipq_src;
1182 	ip->ip_dst = fp->ipq_dst;
1183 	TAILQ_REMOVE(head, fp, ipq_list);
1184 	nipq--;
1185 	(void) m_free(dtom(fp));
1186 	m->m_len += (ip->ip_hl << 2);
1187 	m->m_data -= (ip->ip_hl << 2);
1188 	/* some debugging cruft by sklower, below, will go away soon */
1189 	if (m->m_flags & M_PKTHDR)	/* XXX this should be done elsewhere */
1190 		m_fixhdr(m);
1191 	return (m);
1192 
1193 dropfrag:
1194 #ifdef IPDIVERT
1195 	*divinfo = 0;
1196 	*divert_rule = 0;
1197 #endif
1198 	ipstat.ips_fragdropped++;
1199 	if (fp != NULL)
1200 		fp->ipq_nfrags--;
1201 	m_freem(m);
1202 	return (0);
1203 
1204 #undef GETIP
1205 }
1206 
1207 /*
1208  * Free a fragment reassembly header and all
1209  * associated datagrams.
1210  */
1211 static void
1212 ip_freef(fhp, fp)
1213 	struct ipqhead *fhp;
1214 	struct ipq *fp;
1215 {
1216 	register struct mbuf *q;
1217 
1218 	IPQ_LOCK_ASSERT();
1219 
1220 	while (fp->ipq_frags) {
1221 		q = fp->ipq_frags;
1222 		fp->ipq_frags = q->m_nextpkt;
1223 		m_freem(q);
1224 	}
1225 	TAILQ_REMOVE(fhp, fp, ipq_list);
1226 	(void) m_free(dtom(fp));
1227 	nipq--;
1228 }
1229 
1230 /*
1231  * IP timer processing;
1232  * if a timer expires on a reassembly
1233  * queue, discard it.
1234  */
1235 void
1236 ip_slowtimo()
1237 {
1238 	register struct ipq *fp;
1239 	int s = splnet();
1240 	int i;
1241 
1242 	IPQ_LOCK();
1243 	for (i = 0; i < IPREASS_NHASH; i++) {
1244 		for(fp = TAILQ_FIRST(&ipq[i]); fp;) {
1245 			struct ipq *fpp;
1246 
1247 			fpp = fp;
1248 			fp = TAILQ_NEXT(fp, ipq_list);
1249 			if(--fpp->ipq_ttl == 0) {
1250 				ipstat.ips_fragtimeout += fpp->ipq_nfrags;
1251 				ip_freef(&ipq[i], fpp);
1252 			}
1253 		}
1254 	}
1255 	/*
1256 	 * If we are over the maximum number of fragments
1257 	 * (due to the limit being lowered), drain off
1258 	 * enough to get down to the new limit.
1259 	 */
1260 	if (maxnipq >= 0 && nipq > maxnipq) {
1261 		for (i = 0; i < IPREASS_NHASH; i++) {
1262 			while (nipq > maxnipq && !TAILQ_EMPTY(&ipq[i])) {
1263 				ipstat.ips_fragdropped +=
1264 				    TAILQ_FIRST(&ipq[i])->ipq_nfrags;
1265 				ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i]));
1266 			}
1267 		}
1268 	}
1269 	IPQ_UNLOCK();
1270 	ipflow_slowtimo();
1271 	splx(s);
1272 }
1273 
1274 /*
1275  * Drain off all datagram fragments.
1276  */
1277 void
1278 ip_drain()
1279 {
1280 	int     i;
1281 
1282 	IPQ_LOCK();
1283 	for (i = 0; i < IPREASS_NHASH; i++) {
1284 		while(!TAILQ_EMPTY(&ipq[i])) {
1285 			ipstat.ips_fragdropped +=
1286 			    TAILQ_FIRST(&ipq[i])->ipq_nfrags;
1287 			ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i]));
1288 		}
1289 	}
1290 	IPQ_UNLOCK();
1291 	in_rtqdrain();
1292 }
1293 
1294 /*
1295  * Do option processing on a datagram,
1296  * possibly discarding it if bad options are encountered,
1297  * or forwarding it if source-routed.
1298  * The pass argument is used when operating in the IPSTEALTH
1299  * mode to tell what options to process:
1300  * [LS]SRR (pass 0) or the others (pass 1).
1301  * The reason for as many as two passes is that when doing IPSTEALTH,
1302  * non-routing options should be processed only if the packet is for us.
1303  * Returns 1 if packet has been forwarded/freed,
1304  * 0 if the packet should be processed further.
1305  */
1306 static int
1307 ip_dooptions(struct mbuf *m, int pass, struct sockaddr_in *next_hop)
1308 {
1309 	struct ip *ip = mtod(m, struct ip *);
1310 	u_char *cp;
1311 	struct in_ifaddr *ia;
1312 	int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0;
1313 	struct in_addr *sin, dst;
1314 	n_time ntime;
1315 	struct	sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET };
1316 
1317 	dst = ip->ip_dst;
1318 	cp = (u_char *)(ip + 1);
1319 	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
1320 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1321 		opt = cp[IPOPT_OPTVAL];
1322 		if (opt == IPOPT_EOL)
1323 			break;
1324 		if (opt == IPOPT_NOP)
1325 			optlen = 1;
1326 		else {
1327 			if (cnt < IPOPT_OLEN + sizeof(*cp)) {
1328 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1329 				goto bad;
1330 			}
1331 			optlen = cp[IPOPT_OLEN];
1332 			if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
1333 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1334 				goto bad;
1335 			}
1336 		}
1337 		switch (opt) {
1338 
1339 		default:
1340 			break;
1341 
1342 		/*
1343 		 * Source routing with record.
1344 		 * Find interface with current destination address.
1345 		 * If none on this machine then drop if strictly routed,
1346 		 * or do nothing if loosely routed.
1347 		 * Record interface address and bring up next address
1348 		 * component.  If strictly routed make sure next
1349 		 * address is on directly accessible net.
1350 		 */
1351 		case IPOPT_LSRR:
1352 		case IPOPT_SSRR:
1353 #ifdef IPSTEALTH
1354 			if (ipstealth && pass > 0)
1355 				break;
1356 #endif
1357 			if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1358 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1359 				goto bad;
1360 			}
1361 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1362 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1363 				goto bad;
1364 			}
1365 			ipaddr.sin_addr = ip->ip_dst;
1366 			ia = (struct in_ifaddr *)
1367 				ifa_ifwithaddr((struct sockaddr *)&ipaddr);
1368 			if (ia == 0) {
1369 				if (opt == IPOPT_SSRR) {
1370 					type = ICMP_UNREACH;
1371 					code = ICMP_UNREACH_SRCFAIL;
1372 					goto bad;
1373 				}
1374 				if (!ip_dosourceroute)
1375 					goto nosourcerouting;
1376 				/*
1377 				 * Loose routing, and not at next destination
1378 				 * yet; nothing to do except forward.
1379 				 */
1380 				break;
1381 			}
1382 			off--;			/* 0 origin */
1383 			if (off > optlen - (int)sizeof(struct in_addr)) {
1384 				/*
1385 				 * End of source route.  Should be for us.
1386 				 */
1387 				if (!ip_acceptsourceroute)
1388 					goto nosourcerouting;
1389 				save_rte(cp, ip->ip_src);
1390 				break;
1391 			}
1392 #ifdef IPSTEALTH
1393 			if (ipstealth)
1394 				goto dropit;
1395 #endif
1396 			if (!ip_dosourceroute) {
1397 				if (ipforwarding) {
1398 					char buf[16]; /* aaa.bbb.ccc.ddd\0 */
1399 					/*
1400 					 * Acting as a router, so generate ICMP
1401 					 */
1402 nosourcerouting:
1403 					strcpy(buf, inet_ntoa(ip->ip_dst));
1404 					log(LOG_WARNING,
1405 					    "attempted source route from %s to %s\n",
1406 					    inet_ntoa(ip->ip_src), buf);
1407 					type = ICMP_UNREACH;
1408 					code = ICMP_UNREACH_SRCFAIL;
1409 					goto bad;
1410 				} else {
1411 					/*
1412 					 * Not acting as a router, so silently drop.
1413 					 */
1414 #ifdef IPSTEALTH
1415 dropit:
1416 #endif
1417 					ipstat.ips_cantforward++;
1418 					m_freem(m);
1419 					return (1);
1420 				}
1421 			}
1422 
1423 			/*
1424 			 * locate outgoing interface
1425 			 */
1426 			(void)memcpy(&ipaddr.sin_addr, cp + off,
1427 			    sizeof(ipaddr.sin_addr));
1428 
1429 			if (opt == IPOPT_SSRR) {
1430 #define	INA	struct in_ifaddr *
1431 #define	SA	struct sockaddr *
1432 			    if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == 0)
1433 				ia = (INA)ifa_ifwithnet((SA)&ipaddr);
1434 			} else
1435 				ia = ip_rtaddr(ipaddr.sin_addr, &ipforward_rt);
1436 			if (ia == 0) {
1437 				type = ICMP_UNREACH;
1438 				code = ICMP_UNREACH_SRCFAIL;
1439 				goto bad;
1440 			}
1441 			ip->ip_dst = ipaddr.sin_addr;
1442 			(void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
1443 			    sizeof(struct in_addr));
1444 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1445 			/*
1446 			 * Let ip_intr's mcast routing check handle mcast pkts
1447 			 */
1448 			forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr));
1449 			break;
1450 
1451 		case IPOPT_RR:
1452 #ifdef IPSTEALTH
1453 			if (ipstealth && pass == 0)
1454 				break;
1455 #endif
1456 			if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1457 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1458 				goto bad;
1459 			}
1460 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1461 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1462 				goto bad;
1463 			}
1464 			/*
1465 			 * If no space remains, ignore.
1466 			 */
1467 			off--;			/* 0 origin */
1468 			if (off > optlen - (int)sizeof(struct in_addr))
1469 				break;
1470 			(void)memcpy(&ipaddr.sin_addr, &ip->ip_dst,
1471 			    sizeof(ipaddr.sin_addr));
1472 			/*
1473 			 * locate outgoing interface; if we're the destination,
1474 			 * use the incoming interface (should be same).
1475 			 */
1476 			if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == 0 &&
1477 			    (ia = ip_rtaddr(ipaddr.sin_addr,
1478 			    &ipforward_rt)) == 0) {
1479 				type = ICMP_UNREACH;
1480 				code = ICMP_UNREACH_HOST;
1481 				goto bad;
1482 			}
1483 			(void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
1484 			    sizeof(struct in_addr));
1485 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1486 			break;
1487 
1488 		case IPOPT_TS:
1489 #ifdef IPSTEALTH
1490 			if (ipstealth && pass == 0)
1491 				break;
1492 #endif
1493 			code = cp - (u_char *)ip;
1494 			if (optlen < 4 || optlen > 40) {
1495 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1496 				goto bad;
1497 			}
1498 			if ((off = cp[IPOPT_OFFSET]) < 5) {
1499 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1500 				goto bad;
1501 			}
1502 			if (off > optlen - (int)sizeof(int32_t)) {
1503 				cp[IPOPT_OFFSET + 1] += (1 << 4);
1504 				if ((cp[IPOPT_OFFSET + 1] & 0xf0) == 0) {
1505 					code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1506 					goto bad;
1507 				}
1508 				break;
1509 			}
1510 			off--;				/* 0 origin */
1511 			sin = (struct in_addr *)(cp + off);
1512 			switch (cp[IPOPT_OFFSET + 1] & 0x0f) {
1513 
1514 			case IPOPT_TS_TSONLY:
1515 				break;
1516 
1517 			case IPOPT_TS_TSANDADDR:
1518 				if (off + sizeof(n_time) +
1519 				    sizeof(struct in_addr) > optlen) {
1520 					code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1521 					goto bad;
1522 				}
1523 				ipaddr.sin_addr = dst;
1524 				ia = (INA)ifaof_ifpforaddr((SA)&ipaddr,
1525 							    m->m_pkthdr.rcvif);
1526 				if (ia == 0)
1527 					continue;
1528 				(void)memcpy(sin, &IA_SIN(ia)->sin_addr,
1529 				    sizeof(struct in_addr));
1530 				cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1531 				off += sizeof(struct in_addr);
1532 				break;
1533 
1534 			case IPOPT_TS_PRESPEC:
1535 				if (off + sizeof(n_time) +
1536 				    sizeof(struct in_addr) > optlen) {
1537 					code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1538 					goto bad;
1539 				}
1540 				(void)memcpy(&ipaddr.sin_addr, sin,
1541 				    sizeof(struct in_addr));
1542 				if (ifa_ifwithaddr((SA)&ipaddr) == 0)
1543 					continue;
1544 				cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1545 				off += sizeof(struct in_addr);
1546 				break;
1547 
1548 			default:
1549 				code = &cp[IPOPT_OFFSET + 1] - (u_char *)ip;
1550 				goto bad;
1551 			}
1552 			ntime = iptime();
1553 			(void)memcpy(cp + off, &ntime, sizeof(n_time));
1554 			cp[IPOPT_OFFSET] += sizeof(n_time);
1555 		}
1556 	}
1557 	if (forward && ipforwarding) {
1558 		ip_forward(m, 1, next_hop);
1559 		return (1);
1560 	}
1561 	return (0);
1562 bad:
1563 	icmp_error(m, type, code, 0, 0);
1564 	ipstat.ips_badoptions++;
1565 	return (1);
1566 }
1567 
1568 /*
1569  * Given address of next destination (final or next hop),
1570  * return internet address info of interface to be used to get there.
1571  */
1572 struct in_ifaddr *
1573 ip_rtaddr(dst, rt)
1574 	struct in_addr dst;
1575 	struct route *rt;
1576 {
1577 	register struct sockaddr_in *sin;
1578 
1579 	sin = (struct sockaddr_in *)&rt->ro_dst;
1580 
1581 	if (rt->ro_rt == 0 ||
1582 	    !(rt->ro_rt->rt_flags & RTF_UP) ||
1583 	    dst.s_addr != sin->sin_addr.s_addr) {
1584 		if (rt->ro_rt) {
1585 			RTFREE(rt->ro_rt);
1586 			rt->ro_rt = 0;
1587 		}
1588 		sin->sin_family = AF_INET;
1589 		sin->sin_len = sizeof(*sin);
1590 		sin->sin_addr = dst;
1591 
1592 		rtalloc_ign(rt, RTF_PRCLONING);
1593 	}
1594 	if (rt->ro_rt == 0)
1595 		return ((struct in_ifaddr *)0);
1596 	return (ifatoia(rt->ro_rt->rt_ifa));
1597 }
1598 
1599 /*
1600  * Save incoming source route for use in replies,
1601  * to be picked up later by ip_srcroute if the receiver is interested.
1602  */
1603 static void
1604 save_rte(option, dst)
1605 	u_char *option;
1606 	struct in_addr dst;
1607 {
1608 	unsigned olen;
1609 
1610 	olen = option[IPOPT_OLEN];
1611 #ifdef DIAGNOSTIC
1612 	if (ipprintfs)
1613 		printf("save_rte: olen %d\n", olen);
1614 #endif
1615 	if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst)))
1616 		return;
1617 	bcopy(option, ip_srcrt.srcopt, olen);
1618 	ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1619 	ip_srcrt.dst = dst;
1620 }
1621 
1622 /*
1623  * Retrieve incoming source route for use in replies,
1624  * in the same form used by setsockopt.
1625  * The first hop is placed before the options, will be removed later.
1626  */
1627 struct mbuf *
1628 ip_srcroute()
1629 {
1630 	register struct in_addr *p, *q;
1631 	register struct mbuf *m;
1632 
1633 	if (ip_nhops == 0)
1634 		return ((struct mbuf *)0);
1635 	m = m_get(M_DONTWAIT, MT_HEADER);
1636 	if (m == 0)
1637 		return ((struct mbuf *)0);
1638 
1639 #define OPTSIZ	(sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt))
1640 
1641 	/* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
1642 	m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) +
1643 	    OPTSIZ;
1644 #ifdef DIAGNOSTIC
1645 	if (ipprintfs)
1646 		printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len);
1647 #endif
1648 
1649 	/*
1650 	 * First save first hop for return route
1651 	 */
1652 	p = &ip_srcrt.route[ip_nhops - 1];
1653 	*(mtod(m, struct in_addr *)) = *p--;
1654 #ifdef DIAGNOSTIC
1655 	if (ipprintfs)
1656 		printf(" hops %lx", (u_long)ntohl(mtod(m, struct in_addr *)->s_addr));
1657 #endif
1658 
1659 	/*
1660 	 * Copy option fields and padding (nop) to mbuf.
1661 	 */
1662 	ip_srcrt.nop = IPOPT_NOP;
1663 	ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
1664 	(void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr),
1665 	    &ip_srcrt.nop, OPTSIZ);
1666 	q = (struct in_addr *)(mtod(m, caddr_t) +
1667 	    sizeof(struct in_addr) + OPTSIZ);
1668 #undef OPTSIZ
1669 	/*
1670 	 * Record return path as an IP source route,
1671 	 * reversing the path (pointers are now aligned).
1672 	 */
1673 	while (p >= ip_srcrt.route) {
1674 #ifdef DIAGNOSTIC
1675 		if (ipprintfs)
1676 			printf(" %lx", (u_long)ntohl(q->s_addr));
1677 #endif
1678 		*q++ = *p--;
1679 	}
1680 	/*
1681 	 * Last hop goes to final destination.
1682 	 */
1683 	*q = ip_srcrt.dst;
1684 #ifdef DIAGNOSTIC
1685 	if (ipprintfs)
1686 		printf(" %lx\n", (u_long)ntohl(q->s_addr));
1687 #endif
1688 	return (m);
1689 }
1690 
1691 /*
1692  * Strip out IP options, at higher
1693  * level protocol in the kernel.
1694  * Second argument is buffer to which options
1695  * will be moved, and return value is their length.
1696  * XXX should be deleted; last arg currently ignored.
1697  */
1698 void
1699 ip_stripoptions(m, mopt)
1700 	register struct mbuf *m;
1701 	struct mbuf *mopt;
1702 {
1703 	register int i;
1704 	struct ip *ip = mtod(m, struct ip *);
1705 	register caddr_t opts;
1706 	int olen;
1707 
1708 	olen = (ip->ip_hl << 2) - sizeof (struct ip);
1709 	opts = (caddr_t)(ip + 1);
1710 	i = m->m_len - (sizeof (struct ip) + olen);
1711 	bcopy(opts + olen, opts, (unsigned)i);
1712 	m->m_len -= olen;
1713 	if (m->m_flags & M_PKTHDR)
1714 		m->m_pkthdr.len -= olen;
1715 	ip->ip_v = IPVERSION;
1716 	ip->ip_hl = sizeof(struct ip) >> 2;
1717 }
1718 
1719 u_char inetctlerrmap[PRC_NCMDS] = {
1720 	0,		0,		0,		0,
1721 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1722 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1723 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1724 	0,		0,		EHOSTUNREACH,	0,
1725 	ENOPROTOOPT,	ECONNREFUSED
1726 };
1727 
1728 /*
1729  * Forward a packet.  If some error occurs return the sender
1730  * an icmp packet.  Note we can't always generate a meaningful
1731  * icmp message because icmp doesn't have a large enough repertoire
1732  * of codes and types.
1733  *
1734  * If not forwarding, just drop the packet.  This could be confusing
1735  * if ipforwarding was zero but some routing protocol was advancing
1736  * us as a gateway to somewhere.  However, we must let the routing
1737  * protocol deal with that.
1738  *
1739  * The srcrt parameter indicates whether the packet is being forwarded
1740  * via a source route.
1741  */
1742 static void
1743 ip_forward(struct mbuf *m, int srcrt, struct sockaddr_in *next_hop)
1744 {
1745 	struct ip *ip = mtod(m, struct ip *);
1746 	struct rtentry *rt;
1747 	int error, type = 0, code = 0;
1748 	struct mbuf *mcopy;
1749 	n_long dest;
1750 	struct in_addr pkt_dst;
1751 	struct ifnet *destifp;
1752 #if defined(IPSEC) || defined(FAST_IPSEC)
1753 	struct ifnet dummyifp;
1754 #endif
1755 
1756 	dest = 0;
1757 	/*
1758 	 * Cache the destination address of the packet; this may be
1759 	 * changed by use of 'ipfw fwd'.
1760 	 */
1761 	pkt_dst = next_hop ? next_hop->sin_addr : ip->ip_dst;
1762 
1763 #ifdef DIAGNOSTIC
1764 	if (ipprintfs)
1765 		printf("forward: src %lx dst %lx ttl %x\n",
1766 		    (u_long)ip->ip_src.s_addr, (u_long)pkt_dst.s_addr,
1767 		    ip->ip_ttl);
1768 #endif
1769 
1770 
1771 	if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(pkt_dst) == 0) {
1772 		ipstat.ips_cantforward++;
1773 		m_freem(m);
1774 		return;
1775 	}
1776 #ifdef IPSTEALTH
1777 	if (!ipstealth) {
1778 #endif
1779 		if (ip->ip_ttl <= IPTTLDEC) {
1780 			icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS,
1781 			    dest, 0);
1782 			return;
1783 		}
1784 #ifdef IPSTEALTH
1785 	}
1786 #endif
1787 
1788 	if (ip_rtaddr(pkt_dst, &ipforward_rt) == 0) {
1789 		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0);
1790 		return;
1791 	} else
1792 		rt = ipforward_rt.ro_rt;
1793 
1794 	/*
1795 	 * Save the IP header and at most 8 bytes of the payload,
1796 	 * in case we need to generate an ICMP message to the src.
1797 	 *
1798 	 * XXX this can be optimized a lot by saving the data in a local
1799 	 * buffer on the stack (72 bytes at most), and only allocating the
1800 	 * mbuf if really necessary. The vast majority of the packets
1801 	 * are forwarded without having to send an ICMP back (either
1802 	 * because unnecessary, or because rate limited), so we are
1803 	 * really we are wasting a lot of work here.
1804 	 *
1805 	 * We don't use m_copy() because it might return a reference
1806 	 * to a shared cluster. Both this function and ip_output()
1807 	 * assume exclusive access to the IP header in `m', so any
1808 	 * data in a cluster may change before we reach icmp_error().
1809 	 */
1810 	MGET(mcopy, M_DONTWAIT, m->m_type);
1811 	if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_DONTWAIT)) {
1812 		/*
1813 		 * It's probably ok if the pkthdr dup fails (because
1814 		 * the deep copy of the tag chain failed), but for now
1815 		 * be conservative and just discard the copy since
1816 		 * code below may some day want the tags.
1817 		 */
1818 		m_free(mcopy);
1819 		mcopy = NULL;
1820 	}
1821 	if (mcopy != NULL) {
1822 		mcopy->m_len = imin((ip->ip_hl << 2) + 8,
1823 		    (int)ip->ip_len);
1824 		m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1825 		/*
1826 		 * XXXMAC: Eventually, we may have an explict labeling
1827 		 * point here.
1828 		 */
1829 	}
1830 
1831 #ifdef IPSTEALTH
1832 	if (!ipstealth) {
1833 #endif
1834 		ip->ip_ttl -= IPTTLDEC;
1835 #ifdef IPSTEALTH
1836 	}
1837 #endif
1838 
1839 	/*
1840 	 * If forwarding packet using same interface that it came in on,
1841 	 * perhaps should send a redirect to sender to shortcut a hop.
1842 	 * Only send redirect if source is sending directly to us,
1843 	 * and if packet was not source routed (or has any options).
1844 	 * Also, don't send redirect if forwarding using a default route
1845 	 * or a route modified by a redirect.
1846 	 */
1847 	if (rt->rt_ifp == m->m_pkthdr.rcvif &&
1848 	    (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1849 	    satosin(rt_key(rt))->sin_addr.s_addr != 0 &&
1850 	    ipsendredirects && !srcrt && !next_hop) {
1851 #define	RTA(rt)	((struct in_ifaddr *)(rt->rt_ifa))
1852 		u_long src = ntohl(ip->ip_src.s_addr);
1853 
1854 		if (RTA(rt) &&
1855 		    (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1856 		    if (rt->rt_flags & RTF_GATEWAY)
1857 			dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
1858 		    else
1859 			dest = pkt_dst.s_addr;
1860 		    /* Router requirements says to only send host redirects */
1861 		    type = ICMP_REDIRECT;
1862 		    code = ICMP_REDIRECT_HOST;
1863 #ifdef DIAGNOSTIC
1864 		    if (ipprintfs)
1865 		        printf("redirect (%d) to %lx\n", code, (u_long)dest);
1866 #endif
1867 		}
1868 	}
1869 
1870     {
1871 	struct m_hdr tag;
1872 
1873 	if (next_hop) {
1874 		/* Pass IPFORWARD info if available */
1875 
1876 		tag.mh_type = MT_TAG;
1877 		tag.mh_flags = PACKET_TAG_IPFORWARD;
1878 		tag.mh_data = (caddr_t)next_hop;
1879 		tag.mh_next = m;
1880 		m = (struct mbuf *)&tag;
1881 	}
1882 	error = ip_output(m, (struct mbuf *)0, &ipforward_rt,
1883 			  IP_FORWARDING, 0, NULL);
1884     }
1885 	if (error)
1886 		ipstat.ips_cantforward++;
1887 	else {
1888 		ipstat.ips_forward++;
1889 		if (type)
1890 			ipstat.ips_redirectsent++;
1891 		else {
1892 			if (mcopy) {
1893 				ipflow_create(&ipforward_rt, mcopy);
1894 				m_freem(mcopy);
1895 			}
1896 			return;
1897 		}
1898 	}
1899 	if (mcopy == NULL)
1900 		return;
1901 	destifp = NULL;
1902 
1903 	switch (error) {
1904 
1905 	case 0:				/* forwarded, but need redirect */
1906 		/* type, code set above */
1907 		break;
1908 
1909 	case ENETUNREACH:		/* shouldn't happen, checked above */
1910 	case EHOSTUNREACH:
1911 	case ENETDOWN:
1912 	case EHOSTDOWN:
1913 	default:
1914 		type = ICMP_UNREACH;
1915 		code = ICMP_UNREACH_HOST;
1916 		break;
1917 
1918 	case EMSGSIZE:
1919 		type = ICMP_UNREACH;
1920 		code = ICMP_UNREACH_NEEDFRAG;
1921 #ifdef IPSEC
1922 		/*
1923 		 * If the packet is routed over IPsec tunnel, tell the
1924 		 * originator the tunnel MTU.
1925 		 *	tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
1926 		 * XXX quickhack!!!
1927 		 */
1928 		if (ipforward_rt.ro_rt) {
1929 			struct secpolicy *sp = NULL;
1930 			int ipsecerror;
1931 			int ipsechdr;
1932 			struct route *ro;
1933 
1934 			sp = ipsec4_getpolicybyaddr(mcopy,
1935 						    IPSEC_DIR_OUTBOUND,
1936 			                            IP_FORWARDING,
1937 			                            &ipsecerror);
1938 
1939 			if (sp == NULL)
1940 				destifp = ipforward_rt.ro_rt->rt_ifp;
1941 			else {
1942 				/* count IPsec header size */
1943 				ipsechdr = ipsec4_hdrsiz(mcopy,
1944 							 IPSEC_DIR_OUTBOUND,
1945 							 NULL);
1946 
1947 				/*
1948 				 * find the correct route for outer IPv4
1949 				 * header, compute tunnel MTU.
1950 				 *
1951 				 * XXX BUG ALERT
1952 				 * The "dummyifp" code relies upon the fact
1953 				 * that icmp_error() touches only ifp->if_mtu.
1954 				 */
1955 				/*XXX*/
1956 				destifp = NULL;
1957 				if (sp->req != NULL
1958 				 && sp->req->sav != NULL
1959 				 && sp->req->sav->sah != NULL) {
1960 					ro = &sp->req->sav->sah->sa_route;
1961 					if (ro->ro_rt && ro->ro_rt->rt_ifp) {
1962 						dummyifp.if_mtu =
1963 						    ro->ro_rt->rt_ifp->if_mtu;
1964 						dummyifp.if_mtu -= ipsechdr;
1965 						destifp = &dummyifp;
1966 					}
1967 				}
1968 
1969 				key_freesp(sp);
1970 			}
1971 		}
1972 #elif FAST_IPSEC
1973 		/*
1974 		 * If the packet is routed over IPsec tunnel, tell the
1975 		 * originator the tunnel MTU.
1976 		 *	tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
1977 		 * XXX quickhack!!!
1978 		 */
1979 		if (ipforward_rt.ro_rt) {
1980 			struct secpolicy *sp = NULL;
1981 			int ipsecerror;
1982 			int ipsechdr;
1983 			struct route *ro;
1984 
1985 			sp = ipsec_getpolicybyaddr(mcopy,
1986 						   IPSEC_DIR_OUTBOUND,
1987 			                           IP_FORWARDING,
1988 			                           &ipsecerror);
1989 
1990 			if (sp == NULL)
1991 				destifp = ipforward_rt.ro_rt->rt_ifp;
1992 			else {
1993 				/* count IPsec header size */
1994 				ipsechdr = ipsec4_hdrsiz(mcopy,
1995 							 IPSEC_DIR_OUTBOUND,
1996 							 NULL);
1997 
1998 				/*
1999 				 * find the correct route for outer IPv4
2000 				 * header, compute tunnel MTU.
2001 				 *
2002 				 * XXX BUG ALERT
2003 				 * The "dummyifp" code relies upon the fact
2004 				 * that icmp_error() touches only ifp->if_mtu.
2005 				 */
2006 				/*XXX*/
2007 				destifp = NULL;
2008 				if (sp->req != NULL
2009 				 && sp->req->sav != NULL
2010 				 && sp->req->sav->sah != NULL) {
2011 					ro = &sp->req->sav->sah->sa_route;
2012 					if (ro->ro_rt && ro->ro_rt->rt_ifp) {
2013 						dummyifp.if_mtu =
2014 						    ro->ro_rt->rt_ifp->if_mtu;
2015 						dummyifp.if_mtu -= ipsechdr;
2016 						destifp = &dummyifp;
2017 					}
2018 				}
2019 
2020 				KEY_FREESP(&sp);
2021 			}
2022 		}
2023 #else /* !IPSEC && !FAST_IPSEC */
2024 		if (ipforward_rt.ro_rt)
2025 			destifp = ipforward_rt.ro_rt->rt_ifp;
2026 #endif /*IPSEC*/
2027 		ipstat.ips_cantfrag++;
2028 		break;
2029 
2030 	case ENOBUFS:
2031 		/*
2032 		 * A router should not generate ICMP_SOURCEQUENCH as
2033 		 * required in RFC1812 Requirements for IP Version 4 Routers.
2034 		 * Source quench could be a big problem under DoS attacks,
2035 		 * or if the underlying interface is rate-limited.
2036 		 * Those who need source quench packets may re-enable them
2037 		 * via the net.inet.ip.sendsourcequench sysctl.
2038 		 */
2039 		if (ip_sendsourcequench == 0) {
2040 			m_freem(mcopy);
2041 			return;
2042 		} else {
2043 			type = ICMP_SOURCEQUENCH;
2044 			code = 0;
2045 		}
2046 		break;
2047 
2048 	case EACCES:			/* ipfw denied packet */
2049 		m_freem(mcopy);
2050 		return;
2051 	}
2052 	icmp_error(mcopy, type, code, dest, destifp);
2053 }
2054 
2055 void
2056 ip_savecontrol(inp, mp, ip, m)
2057 	register struct inpcb *inp;
2058 	register struct mbuf **mp;
2059 	register struct ip *ip;
2060 	register struct mbuf *m;
2061 {
2062 	if (inp->inp_socket->so_options & SO_TIMESTAMP) {
2063 		struct timeval tv;
2064 
2065 		microtime(&tv);
2066 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
2067 			SCM_TIMESTAMP, SOL_SOCKET);
2068 		if (*mp)
2069 			mp = &(*mp)->m_next;
2070 	}
2071 	if (inp->inp_flags & INP_RECVDSTADDR) {
2072 		*mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
2073 		    sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
2074 		if (*mp)
2075 			mp = &(*mp)->m_next;
2076 	}
2077 	if (inp->inp_flags & INP_RECVTTL) {
2078 		*mp = sbcreatecontrol((caddr_t) &ip->ip_ttl,
2079 		    sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
2080 		if (*mp)
2081 			mp = &(*mp)->m_next;
2082 	}
2083 #ifdef notyet
2084 	/* XXX
2085 	 * Moving these out of udp_input() made them even more broken
2086 	 * than they already were.
2087 	 */
2088 	/* options were tossed already */
2089 	if (inp->inp_flags & INP_RECVOPTS) {
2090 		*mp = sbcreatecontrol((caddr_t) opts_deleted_above,
2091 		    sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
2092 		if (*mp)
2093 			mp = &(*mp)->m_next;
2094 	}
2095 	/* ip_srcroute doesn't do what we want here, need to fix */
2096 	if (inp->inp_flags & INP_RECVRETOPTS) {
2097 		*mp = sbcreatecontrol((caddr_t) ip_srcroute(),
2098 		    sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
2099 		if (*mp)
2100 			mp = &(*mp)->m_next;
2101 	}
2102 #endif
2103 	if (inp->inp_flags & INP_RECVIF) {
2104 		struct ifnet *ifp;
2105 		struct sdlbuf {
2106 			struct sockaddr_dl sdl;
2107 			u_char	pad[32];
2108 		} sdlbuf;
2109 		struct sockaddr_dl *sdp;
2110 		struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
2111 
2112 		if (((ifp = m->m_pkthdr.rcvif))
2113 		&& ( ifp->if_index && (ifp->if_index <= if_index))) {
2114 			sdp = (struct sockaddr_dl *)
2115 			    (ifaddr_byindex(ifp->if_index)->ifa_addr);
2116 			/*
2117 			 * Change our mind and don't try copy.
2118 			 */
2119 			if ((sdp->sdl_family != AF_LINK)
2120 			|| (sdp->sdl_len > sizeof(sdlbuf))) {
2121 				goto makedummy;
2122 			}
2123 			bcopy(sdp, sdl2, sdp->sdl_len);
2124 		} else {
2125 makedummy:
2126 			sdl2->sdl_len
2127 				= offsetof(struct sockaddr_dl, sdl_data[0]);
2128 			sdl2->sdl_family = AF_LINK;
2129 			sdl2->sdl_index = 0;
2130 			sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
2131 		}
2132 		*mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len,
2133 			IP_RECVIF, IPPROTO_IP);
2134 		if (*mp)
2135 			mp = &(*mp)->m_next;
2136 	}
2137 }
2138 
2139 /*
2140  * XXX these routines are called from the upper part of the kernel.
2141  * They need to be locked when we remove Giant.
2142  *
2143  * They could also be moved to ip_mroute.c, since all the RSVP
2144  *  handling is done there already.
2145  */
2146 static int ip_rsvp_on;
2147 struct socket *ip_rsvpd;
2148 int
2149 ip_rsvp_init(struct socket *so)
2150 {
2151 	if (so->so_type != SOCK_RAW ||
2152 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
2153 		return EOPNOTSUPP;
2154 
2155 	if (ip_rsvpd != NULL)
2156 		return EADDRINUSE;
2157 
2158 	ip_rsvpd = so;
2159 	/*
2160 	 * This may seem silly, but we need to be sure we don't over-increment
2161 	 * the RSVP counter, in case something slips up.
2162 	 */
2163 	if (!ip_rsvp_on) {
2164 		ip_rsvp_on = 1;
2165 		rsvp_on++;
2166 	}
2167 
2168 	return 0;
2169 }
2170 
2171 int
2172 ip_rsvp_done(void)
2173 {
2174 	ip_rsvpd = NULL;
2175 	/*
2176 	 * This may seem silly, but we need to be sure we don't over-decrement
2177 	 * the RSVP counter, in case something slips up.
2178 	 */
2179 	if (ip_rsvp_on) {
2180 		ip_rsvp_on = 0;
2181 		rsvp_on--;
2182 	}
2183 	return 0;
2184 }
2185 
2186 void
2187 rsvp_input(struct mbuf *m, int off)	/* XXX must fixup manually */
2188 {
2189 	if (rsvp_input_p) { /* call the real one if loaded */
2190 		rsvp_input_p(m, off);
2191 		return;
2192 	}
2193 
2194 	/* Can still get packets with rsvp_on = 0 if there is a local member
2195 	 * of the group to which the RSVP packet is addressed.  But in this
2196 	 * case we want to throw the packet away.
2197 	 */
2198 
2199 	if (!rsvp_on) {
2200 		m_freem(m);
2201 		return;
2202 	}
2203 
2204 	if (ip_rsvpd != NULL) {
2205 		rip_input(m, off);
2206 		return;
2207 	}
2208 	/* Drop the packet */
2209 	m_freem(m);
2210 }
2211