xref: /dflybsd-src/sys/netinet/ip_input.c (revision bc76a771df54af7e361532b257cecc26227736b4)
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: src/sys/netinet/ip_input.c,v 1.130.2.52 2003/03/07 07:01:28 silby Exp $
35  * $DragonFly: src/sys/netinet/ip_input.c,v 1.18 2004/04/13 00:14:01 hsu Exp $
36  */
37 
38 #define	_IP_VHL
39 
40 #include "opt_bootp.h"
41 #include "opt_ipfw.h"
42 #include "opt_ipdn.h"
43 #include "opt_ipdivert.h"
44 #include "opt_ipfilter.h"
45 #include "opt_ipstealth.h"
46 #include "opt_ipsec.h"
47 #include "opt_pfil_hooks.h"
48 #include "opt_random_ip_id.h"
49 
50 #include <sys/param.h>
51 #include <sys/systm.h>
52 #include <sys/mbuf.h>
53 #include <sys/malloc.h>
54 #include <sys/domain.h>
55 #include <sys/protosw.h>
56 #include <sys/socket.h>
57 #include <sys/time.h>
58 #include <sys/kernel.h>
59 #include <sys/syslog.h>
60 #include <sys/sysctl.h>
61 #include <sys/in_cksum.h>
62 
63 #include <sys/thread2.h>
64 #include <sys/msgport2.h>
65 
66 #include <net/if.h>
67 #include <net/if_types.h>
68 #include <net/if_var.h>
69 #include <net/if_dl.h>
70 #ifdef PFIL_HOOKS
71 #include <net/pfil.h>
72 #endif
73 #include <net/route.h>
74 #include <net/netisr.h>
75 #include <net/intrq.h>
76 
77 #include <netinet/in.h>
78 #include <netinet/in_systm.h>
79 #include <netinet/in_var.h>
80 #include <netinet/ip.h>
81 #include <netinet/in_pcb.h>
82 #include <netinet/ip_var.h>
83 #include <netinet/ip_icmp.h>
84 
85 #include <netinet/ipprotosw.h>
86 
87 #include <sys/socketvar.h>
88 
89 #include <net/ipfw/ip_fw.h>
90 #include <net/dummynet/ip_dummynet.h>
91 
92 #ifdef IPSEC
93 #include <netinet6/ipsec.h>
94 #include <netproto/key/key.h>
95 #endif
96 
97 #ifdef FAST_IPSEC
98 #include <netipsec/ipsec.h>
99 #include <netipsec/key.h>
100 #endif
101 
102 int rsvp_on = 0;
103 static int ip_rsvp_on;
104 struct socket *ip_rsvpd;
105 
106 int ipforwarding = 0;
107 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW,
108     &ipforwarding, 0, "Enable IP forwarding between interfaces");
109 
110 static int ipsendredirects = 1; /* XXX */
111 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW,
112     &ipsendredirects, 0, "Enable sending IP redirects");
113 
114 int ip_defttl = IPDEFTTL;
115 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW,
116     &ip_defttl, 0, "Maximum TTL on IP packets");
117 
118 static int ip_dosourceroute = 0;
119 SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW,
120     &ip_dosourceroute, 0, "Enable forwarding source routed IP packets");
121 
122 static int ip_acceptsourceroute = 0;
123 SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute,
124     CTLFLAG_RW, &ip_acceptsourceroute, 0,
125     "Enable accepting source routed IP packets");
126 
127 static int ip_keepfaith = 0;
128 SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW,
129     &ip_keepfaith, 0,
130     "Enable packet capture for FAITH IPv4->IPv6 translater daemon");
131 
132 static int nipq = 0;	/* total # of reass queues */
133 static int maxnipq;
134 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragpackets, CTLFLAG_RW,
135     &maxnipq, 0,
136     "Maximum number of IPv4 fragment reassembly queue entries");
137 
138 static int maxfragsperpacket;
139 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW,
140     &maxfragsperpacket, 0,
141     "Maximum number of IPv4 fragments allowed per packet");
142 
143 static int ip_sendsourcequench = 0;
144 SYSCTL_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW,
145     &ip_sendsourcequench, 0,
146     "Enable the transmission of source quench packets");
147 
148 /*
149  * XXX - Setting ip_checkinterface mostly implements the receive side of
150  * the Strong ES model described in RFC 1122, but since the routing table
151  * and transmit implementation do not implement the Strong ES model,
152  * setting this to 1 results in an odd hybrid.
153  *
154  * XXX - ip_checkinterface currently must be disabled if you use ipnat
155  * to translate the destination address to another local interface.
156  *
157  * XXX - ip_checkinterface must be disabled if you add IP aliases
158  * to the loopback interface instead of the interface where the
159  * packets for those addresses are received.
160  */
161 static int ip_checkinterface = 0;
162 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW,
163     &ip_checkinterface, 0, "Verify packet arrives on correct interface");
164 
165 #ifdef DIAGNOSTIC
166 static int ipprintfs = 0;
167 #endif
168 
169 static struct ifqueue ipintrq;
170 static int ipqmaxlen = IFQ_MAXLEN;
171 
172 extern	struct domain inetdomain;
173 extern	struct ipprotosw inetsw[];
174 u_char	ip_protox[IPPROTO_MAX];
175 struct	in_ifaddrhead in_ifaddrhead;		/* first inet address */
176 struct	in_ifaddrhashhead *in_ifaddrhashtbl;	/* inet addr hash table */
177 u_long	in_ifaddrhmask;				/* mask for hash table */
178 
179 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW,
180     &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue");
181 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD,
182     &ipintrq.ifq_drops, 0, "Number of packets dropped from the IP input queue");
183 
184 struct ipstat ipstat;
185 SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW,
186     &ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)");
187 
188 /* Packet reassembly stuff */
189 #define	IPREASS_NHASH_LOG2	6
190 #define	IPREASS_NHASH		(1 << IPREASS_NHASH_LOG2)
191 #define	IPREASS_HMASK		(IPREASS_NHASH - 1)
192 #define	IPREASS_HASH(x,y)						\
193     (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK)
194 
195 static struct ipq ipq[IPREASS_NHASH];
196 const  int    ipintrq_present = 1;
197 
198 #ifdef IPCTL_DEFMTU
199 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
200     &ip_mtu, 0, "Default MTU");
201 #endif
202 
203 #ifdef IPSTEALTH
204 static int ipstealth = 0;
205 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW, &ipstealth, 0, "");
206 #else
207 static const int ipstealth = 0;
208 #endif
209 
210 
211 /* Firewall hooks */
212 ip_fw_chk_t *ip_fw_chk_ptr;
213 int fw_enable = 1;
214 int fw_one_pass = 1;
215 
216 /* Dummynet hooks */
217 ip_dn_io_t *ip_dn_io_ptr;
218 
219 #ifdef PFIL_HOOKS
220 struct pfil_head inet_pfil_hook;
221 #endif
222 
223 /*
224  * XXX this is ugly -- the following two global variables are
225  * used to store packet state while it travels through the stack.
226  * Note that the code even makes assumptions on the size and
227  * alignment of fields inside struct ip_srcrt so e.g. adding some
228  * fields will break the code. This needs to be fixed.
229  *
230  * We need to save the IP options in case a protocol wants to respond
231  * to an incoming packet over the same route if the packet got here
232  * using IP source routing.  This allows connection establishment and
233  * maintenance when the remote end is on a network that is not known
234  * to us.
235  */
236 static int ip_nhops = 0;
237 
238 static	struct ip_srcrt {
239 	struct	in_addr dst;			/* final destination */
240 	char	nop;				/* one NOP to align */
241 	char	srcopt[IPOPT_OFFSET + 1];	/* OPTVAL, OLEN and OFFSET */
242 	struct	in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)];
243 } ip_srcrt;
244 
245 static void		save_rte (u_char *, struct in_addr);
246 static int		ip_dooptions (struct mbuf *m, int,
247 					struct sockaddr_in *next_hop);
248 static void		ip_forward (struct mbuf *m, int srcrt,
249 					struct sockaddr_in *next_hop);
250 static void		ip_freef (struct ipq *);
251 static void		ip_input_handler (struct netmsg *);
252 static struct mbuf	*ip_reass (struct mbuf *, struct ipq *,
253 					struct ipq *, u_int32_t *, u_int16_t *);
254 
255 /*
256  * IP initialization: fill in IP protocol switch table.
257  * All protocols not implemented in kernel go to raw IP protocol handler.
258  */
259 void
260 ip_init()
261 {
262 	struct ipprotosw *pr;
263 	int i;
264 
265 	TAILQ_INIT(&in_ifaddrhead);
266 	in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask);
267 	pr = (struct ipprotosw *)pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
268 	if (pr == NULL)
269 		panic("ip_init");
270 	for (i = 0; i < IPPROTO_MAX; i++)
271 		ip_protox[i] = pr - inetsw;
272 	for (pr = (struct ipprotosw *)inetdomain.dom_protosw;
273 	     pr < (struct ipprotosw *)inetdomain.dom_protoswNPROTOSW; pr++)
274 		if (pr->pr_domain->dom_family == PF_INET &&
275 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
276 			ip_protox[pr->pr_protocol] = pr - inetsw;
277 
278 #ifdef PFIL_HOOKS
279 	inet_pfil_hook.ph_type = PFIL_TYPE_AF;
280 	inet_pfil_hook.ph_af = AF_INET;
281 	if ((i = pfil_head_register(&inet_pfil_hook)) != 0)
282 		printf("%s: WARNING: unable to register pfil hook, "
283 			"error %d\n", __func__, i);
284 #endif
285 
286 	for (i = 0; i < IPREASS_NHASH; i++)
287 	    ipq[i].next = ipq[i].prev = &ipq[i];
288 
289 	maxnipq = nmbclusters / 32;
290 	maxfragsperpacket = 16;
291 
292 #ifndef RANDOM_IP_ID
293 	ip_id = time_second & 0xffff;
294 #endif
295 	ipintrq.ifq_maxlen = ipqmaxlen;
296 
297 	netisr_register(NETISR_IP, ip_mport, ip_input_handler);
298 }
299 
300 /*
301  * XXX watch out this one. It is perhaps used as a cache for
302  * the most recently used route ? it is cleared in in_addroute()
303  * when a new route is successfully created.
304  */
305 struct route ipforward_rt;
306 static struct sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET };
307 
308 /* Do transport protocol processing. */
309 static void
310 transport_processing_oncpu(struct mbuf *m, int hlen, struct ip *ip,
311 			   struct sockaddr_in *nexthop)
312 {
313 	/*
314 	 * Switch out to protocol's input routine.
315 	 */
316 	if (nexthop && ip->ip_p == IPPROTO_TCP) {
317 		/* TCP needs IPFORWARD info if available */
318 		struct m_hdr tag;
319 
320 		tag.mh_type = MT_TAG;
321 		tag.mh_flags = PACKET_TAG_IPFORWARD;
322 		tag.mh_data = (caddr_t)nexthop;
323 		tag.mh_next = m;
324 
325 		(*inetsw[ip_protox[ip->ip_p]].pr_input)
326 		    ((struct mbuf *)&tag, hlen, ip->ip_p);
327 	} else {
328 		(*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen, ip->ip_p);
329 	}
330 }
331 
332 struct netmsg_transport_packet {
333 	struct lwkt_msg		nm_lmsg;
334 	netisr_fn_t		nm_handler;
335 	struct mbuf		*nm_mbuf;
336 	int			nm_hlen;
337 	boolean_t		nm_hasnexthop;
338 	struct sockaddr_in	nm_nexthop;
339 };
340 
341 static void
342 transport_processing_handler(struct netmsg *msg0)
343 {
344 	struct netmsg_transport_packet *msg =
345 	    (struct netmsg_transport_packet *)msg0;
346 	struct sockaddr_in *nexthop;
347 	struct ip *ip;
348 
349 	ip = mtod(msg->nm_mbuf, struct ip *);
350 	nexthop = msg->nm_hasnexthop ? &msg->nm_nexthop : NULL;
351 	transport_processing_oncpu(msg->nm_mbuf, msg->nm_hlen, ip, nexthop);
352 	lwkt_replymsg(&msg0->nm_lmsg, 0);
353 }
354 
355 static void
356 ip_input_handler(struct netmsg *msg0)
357 {
358 	struct mbuf *m = ((struct netmsg_packet *)msg0)->nm_packet;
359 
360 	ip_input(m);
361 	lwkt_replymsg(&msg0->nm_lmsg, 0);
362 }
363 
364 /*
365  * Ip input routine.  Checksum and byte swap header.  If fragmented
366  * try to reassemble.  Process options.  Pass to next level.
367  */
368 void
369 ip_input(struct mbuf *m)
370 {
371 	struct ip *ip;
372 	struct ipq *fp;
373 	struct in_ifaddr *ia = NULL;
374 	struct ifaddr *ifa;
375 	int i, hlen, checkif;
376 	u_short sum;
377 	struct in_addr pkt_dst;
378 	u_int32_t divert_info = 0;		/* packet divert/tee info */
379 	struct ip_fw_args args;
380 	boolean_t using_srcrt = FALSE;		/* forward (by PFIL_HOOKS) */
381 	boolean_t needredispatch = FALSE;
382 #ifdef PFIL_HOOKS
383 	struct in_addr odst;			/* original dst address(NAT) */
384 #endif
385 #ifdef FAST_IPSEC
386 	struct m_tag *mtag;
387 	struct tdb_ident *tdbi;
388 	struct secpolicy *sp;
389 	int s, error;
390 #endif
391 
392 	args.eh = NULL;
393 	args.oif = NULL;
394 	args.rule = NULL;
395 	args.divert_rule = 0;			/* divert cookie */
396 	args.next_hop = NULL;
397 
398 	/* Grab info from MT_TAG mbufs prepended to the chain. */
399 	for (; m && m->m_type == MT_TAG; m = m->m_next) {
400 		switch(m->_m_tag_id) {
401 		default:
402 			printf("ip_input: unrecognised MT_TAG tag %d\n",
403 			    m->_m_tag_id);
404 			break;
405 
406 		case PACKET_TAG_DUMMYNET:
407 			args.rule = ((struct dn_pkt *)m)->rule;
408 			break;
409 
410 		case PACKET_TAG_DIVERT:
411 			args.divert_rule = (int)m->m_hdr.mh_data & 0xffff;
412 			break;
413 
414 		case PACKET_TAG_IPFORWARD:
415 			args.next_hop = (struct sockaddr_in *)m->m_hdr.mh_data;
416 			break;
417 		}
418 	}
419 
420 	KASSERT(m != NULL && (m->m_flags & M_PKTHDR) != 0,
421 	    ("ip_input: no HDR"));
422 
423 	if (args.rule) {	/* dummynet already filtered us */
424 		ip = mtod(m, struct ip *);
425 		hlen = IP_VHL_HL(ip->ip_vhl) << 2;
426 		goto iphack;
427 	}
428 
429 	ipstat.ips_total++;
430 
431 	/* length checks already done in ip_demux() */
432 	KASSERT(m->m_len >= sizeof(ip), ("IP header not in one mbuf"));
433 
434 	ip = mtod(m, struct ip *);
435 
436 	if (IP_VHL_V(ip->ip_vhl) != IPVERSION) {
437 		ipstat.ips_badvers++;
438 		goto bad;
439 	}
440 
441 	hlen = IP_VHL_HL(ip->ip_vhl) << 2;
442 	/* length checks already done in ip_demux() */
443 	KASSERT(hlen >= sizeof(struct ip), ("IP header len too small"));
444 	KASSERT(m->m_len >= hlen, ("packet shorter than IP header length"));
445 
446 	/* 127/8 must not appear on wire - RFC1122 */
447 	if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
448 	    (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
449 		if (!(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK)) {
450 			ipstat.ips_badaddr++;
451 			goto bad;
452 		}
453 	}
454 
455 	if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
456 		sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
457 	} else {
458 		if (hlen == sizeof(struct ip)) {
459 			sum = in_cksum_hdr(ip);
460 		} else {
461 			sum = in_cksum(m, hlen);
462 		}
463 	}
464 	if (sum) {
465 		ipstat.ips_badsum++;
466 		goto bad;
467 	}
468 
469 	/*
470 	 * Convert fields to host representation.
471 	 */
472 	ip->ip_len = ntohs(ip->ip_len);
473 	if (ip->ip_len < hlen) {
474 		ipstat.ips_badlen++;
475 		goto bad;
476 	}
477 	ip->ip_off = ntohs(ip->ip_off);
478 
479 	/*
480 	 * Check that the amount of data in the buffers
481 	 * is as at least much as the IP header would have us expect.
482 	 * Trim mbufs if longer than we expect.
483 	 * Drop packet if shorter than we expect.
484 	 */
485 	if (m->m_pkthdr.len < ip->ip_len) {
486 		ipstat.ips_tooshort++;
487 		goto bad;
488 	}
489 	if (m->m_pkthdr.len > ip->ip_len) {
490 		if (m->m_len == m->m_pkthdr.len) {
491 			m->m_len = ip->ip_len;
492 			m->m_pkthdr.len = ip->ip_len;
493 		} else
494 			m_adj(m, ip->ip_len - m->m_pkthdr.len);
495 	}
496 #if defined(IPSEC) && !defined(IPSEC_FILTERGIF)
497 	/*
498 	 * Bypass packet filtering for packets from a tunnel (gif).
499 	 */
500 	if (ipsec_gethist(m, NULL))
501 		goto pass;
502 #endif
503 
504 	/*
505 	 * IpHack's section.
506 	 * Right now when no processing on packet has done
507 	 * and it is still fresh out of network we do our black
508 	 * deals with it.
509 	 * - Firewall: deny/allow/divert
510 	 * - Xlate: translate packet's addr/port (NAT).
511 	 * - Pipe: pass pkt through dummynet.
512 	 * - Wrap: fake packet's addr/port <unimpl.>
513 	 * - Encapsulate: put it in another IP and send out. <unimp.>
514 	 */
515 
516 iphack:
517 
518 #ifdef PFIL_HOOKS
519 	/*
520 	 * Run through list of hooks for input packets.
521 	 *
522 	 * NB: Beware of the destination address changing (e.g.
523 	 *     by NAT rewriting). When this happens, tell
524 	 *     ip_forward to do the right thing.
525 	 */
526 	odst = ip->ip_dst;
527 	if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif, PFIL_IN))
528 		return;
529 	if (m == NULL)			/* consumed by filter */
530 		return;
531 	ip = mtod(m, struct ip *);
532 	using_srcrt = (odst.s_addr != ip->ip_dst.s_addr);
533 #endif
534 
535 	if (fw_enable && IPFW_LOADED) {
536 		/*
537 		 * If we've been forwarded from the output side, then
538 		 * skip the firewall a second time
539 		 */
540 		if (args.next_hop)
541 			goto ours;
542 
543 		args.m = m;
544 		i = ip_fw_chk_ptr(&args);
545 		m = args.m;
546 
547 		if ( (i & IP_FW_PORT_DENY_FLAG) || m == NULL) { /* drop */
548 			if (m)
549 				m_freem(m);
550 			return;
551 		}
552 		ip = mtod(m, struct ip *); /* just in case m changed */
553 		if (i == 0 && args.next_hop == NULL)	/* common case */
554 			goto pass;
555 		if (DUMMYNET_LOADED && (i & IP_FW_PORT_DYNT_FLAG)) {
556 			/* Send packet to the appropriate pipe */
557 			ip_dn_io_ptr(m, i&0xffff, DN_TO_IP_IN, &args);
558 			return;
559 		}
560 #ifdef IPDIVERT
561 		if (i != 0 && !(i & IP_FW_PORT_DYNT_FLAG)) {
562 			/* Divert or tee packet */
563 			divert_info = i;
564 			goto ours;
565 		}
566 #endif
567 		if (i == 0 && args.next_hop != NULL)
568 			goto pass;
569 		/*
570 		 * if we get here, the packet must be dropped
571 		 */
572 		m_freem(m);
573 		return;
574 	}
575 pass:
576 
577 	/*
578 	 * Process options and, if not destined for us,
579 	 * ship it on.  ip_dooptions returns 1 when an
580 	 * error was detected (causing an icmp message
581 	 * to be sent and the original packet to be freed).
582 	 */
583 	ip_nhops = 0;		/* for source routed packets */
584 	if (hlen > sizeof(struct ip) && ip_dooptions(m, 0, args.next_hop))
585 		return;
586 
587 	/* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
588 	 * matter if it is destined to another node, or whether it is
589 	 * a multicast one, RSVP wants it! and prevents it from being forwarded
590 	 * anywhere else. Also checks if the rsvp daemon is running before
591 	 * grabbing the packet.
592 	 */
593 	if (rsvp_on && ip->ip_p == IPPROTO_RSVP)
594 		goto ours;
595 
596 	/*
597 	 * Check our list of addresses, to see if the packet is for us.
598 	 * If we don't have any addresses, assume any unicast packet
599 	 * we receive might be for us (and let the upper layers deal
600 	 * with it).
601 	 */
602 	if (TAILQ_EMPTY(&in_ifaddrhead) && !(m->m_flags & (M_MCAST | M_BCAST)))
603 		goto ours;
604 
605 	/*
606 	 * Cache the destination address of the packet; this may be
607 	 * changed by use of 'ipfw fwd'.
608 	 */
609 	pkt_dst = args.next_hop ? args.next_hop->sin_addr : ip->ip_dst;
610 
611 	/*
612 	 * Enable a consistency check between the destination address
613 	 * and the arrival interface for a unicast packet (the RFC 1122
614 	 * strong ES model) if IP forwarding is disabled and the packet
615 	 * is not locally generated and the packet is not subject to
616 	 * 'ipfw fwd'.
617 	 *
618 	 * XXX - Checking also should be disabled if the destination
619 	 * address is ipnat'ed to a different interface.
620 	 *
621 	 * XXX - Checking is incompatible with IP aliases added
622 	 * to the loopback interface instead of the interface where
623 	 * the packets are received.
624 	 */
625 	checkif = ip_checkinterface &&
626 		  !ipforwarding &&
627 		  m->m_pkthdr.rcvif != NULL &&
628 		  !(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) &&
629 		  (args.next_hop == NULL);
630 
631 	/*
632 	 * Check for exact addresses in the hash bucket.
633 	 */
634 	LIST_FOREACH(ia, INADDR_HASH(pkt_dst.s_addr), ia_hash) {
635 		/*
636 		 * If the address matches, verify that the packet
637 		 * arrived via the correct interface if checking is
638 		 * enabled.
639 		 */
640 		if (IA_SIN(ia)->sin_addr.s_addr == pkt_dst.s_addr &&
641 		    (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif))
642 			goto ours;
643 	}
644 	/*
645 	 * Check for broadcast addresses.
646 	 *
647 	 * Only accept broadcast packets that arrive via the matching
648 	 * interface.  Reception of forwarded directed broadcasts would
649 	 * be handled via ip_forward() and ether_output() with the loopback
650 	 * into the stack for SIMPLEX interfaces handled by ether_output().
651 	 */
652 	if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
653 		TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
654 			if (ifa->ifa_addr->sa_family != AF_INET)
655 				continue;
656 			ia = ifatoia(ifa);
657 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
658 								pkt_dst.s_addr)
659 				goto ours;
660 			if (ia->ia_netbroadcast.s_addr == pkt_dst.s_addr)
661 				goto ours;
662 #ifdef BOOTP_COMPAT
663 			if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY)
664 				goto ours;
665 #endif
666 		}
667 	}
668 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
669 		struct in_multi *inm;
670 		if (ip_mrouter) {
671 			/*
672 			 * If we are acting as a multicast router, all
673 			 * incoming multicast packets are passed to the
674 			 * kernel-level multicast forwarding function.
675 			 * The packet is returned (relatively) intact; if
676 			 * ip_mforward() returns a non-zero value, the packet
677 			 * must be discarded, else it may be accepted below.
678 			 */
679 			if (ip_mforward &&
680 			    ip_mforward(ip, m->m_pkthdr.rcvif, m, NULL) != 0) {
681 				ipstat.ips_cantforward++;
682 				m_freem(m);
683 				return;
684 			}
685 
686 			/*
687 			 * The process-level routing daemon needs to receive
688 			 * all multicast IGMP packets, whether or not this
689 			 * host belongs to their destination groups.
690 			 */
691 			if (ip->ip_p == IPPROTO_IGMP)
692 				goto ours;
693 			ipstat.ips_forward++;
694 		}
695 		/*
696 		 * See if we belong to the destination multicast group on the
697 		 * arrival interface.
698 		 */
699 		IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
700 		if (inm == NULL) {
701 			ipstat.ips_notmember++;
702 			m_freem(m);
703 			return;
704 		}
705 		goto ours;
706 	}
707 	if (ip->ip_dst.s_addr == INADDR_BROADCAST)
708 		goto ours;
709 	if (ip->ip_dst.s_addr == INADDR_ANY)
710 		goto ours;
711 
712 	/*
713 	 * FAITH(Firewall Aided Internet Translator)
714 	 */
715 	if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
716 		if (ip_keepfaith) {
717 			if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP)
718 				goto ours;
719 		}
720 		m_freem(m);
721 		return;
722 	}
723 
724 	/*
725 	 * Not for us; forward if possible and desirable.
726 	 */
727 	if (!ipforwarding) {
728 		ipstat.ips_cantforward++;
729 		m_freem(m);
730 	} else {
731 #ifdef IPSEC
732 		/*
733 		 * Enforce inbound IPsec SPD.
734 		 */
735 		if (ipsec4_in_reject(m, NULL)) {
736 			ipsecstat.in_polvio++;
737 			goto bad;
738 		}
739 #endif
740 #ifdef FAST_IPSEC
741 		mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
742 		s = splnet();
743 		if (mtag != NULL) {
744 			tdbi = (struct tdb_ident *)(mtag + 1);
745 			sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
746 		} else {
747 			sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
748 						   IP_FORWARDING, &error);
749 		}
750 		if (sp == NULL) {	/* NB: can happen if error */
751 			splx(s);
752 			/*XXX error stat???*/
753 			DPRINTF(("ip_input: no SP for forwarding\n"));	/*XXX*/
754 			goto bad;
755 		}
756 
757 		/*
758 		 * Check security policy against packet attributes.
759 		 */
760 		error = ipsec_in_reject(sp, m);
761 		KEY_FREESP(&sp);
762 		splx(s);
763 		if (error) {
764 			ipstat.ips_cantforward++;
765 			goto bad;
766 		}
767 #endif
768 		ip_forward(m, using_srcrt, args.next_hop);
769 	}
770 	return;
771 
772 ours:
773 
774 	/*
775 	 * IPSTEALTH: Process non-routing options only
776 	 * if the packet is destined for us.
777 	 */
778 	if (ipstealth &&
779 	    hlen > sizeof(struct ip) &&
780 	    ip_dooptions(m, 1, args.next_hop))
781 		return;
782 
783 	/* Count the packet in the ip address stats */
784 	if (ia != NULL) {
785 		ia->ia_ifa.if_ipackets++;
786 		ia->ia_ifa.if_ibytes += m->m_pkthdr.len;
787 	}
788 
789 	/*
790 	 * If offset or IP_MF are set, must reassemble.
791 	 * Otherwise, nothing need be done.
792 	 * (We could look in the reassembly queue to see
793 	 * if the packet was previously fragmented,
794 	 * but it's not worth the time; just let them time out.)
795 	 */
796 	if (ip->ip_off & (IP_MF | IP_OFFMASK)) {
797 
798 		/* If maxnipq is 0, never accept fragments. */
799 		if (maxnipq == 0) {
800 			ipstat.ips_fragments++;
801 			ipstat.ips_fragdropped++;
802 			goto bad;
803 		}
804 
805 		sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
806 		/*
807 		 * Look for queue of fragments
808 		 * of this datagram.
809 		 */
810 		for (fp = ipq[sum].next; fp != &ipq[sum]; fp = fp->next)
811 			if (ip->ip_id == fp->ipq_id &&
812 			    ip->ip_src.s_addr == fp->ipq_src.s_addr &&
813 			    ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
814 			    ip->ip_p == fp->ipq_p)
815 				goto found;
816 
817 		fp = NULL;
818 
819 		/*
820 		 * Enforce upper bound on number of fragmented packets
821 		 * for which we attempt reassembly;
822 		 * If maxnipq is -1, accept all fragments without limitation.
823 		 */
824 		if ((nipq > maxnipq) && (maxnipq > 0)) {
825 			/*
826 			 * drop something from the tail of the current queue
827 			 * before proceeding further
828 			 */
829 			if (ipq[sum].prev == &ipq[sum]) {   /* gak */
830 				for (i = 0; i < IPREASS_NHASH; i++) {
831 					if (ipq[i].prev != &ipq[i]) {
832 						ipstat.ips_fragtimeout +=
833 						    ipq[i].prev->ipq_nfrags;
834 						ip_freef(ipq[i].prev);
835 						break;
836 					}
837 				}
838 			} else {
839 				ipstat.ips_fragtimeout +=
840 				    ipq[sum].prev->ipq_nfrags;
841 				ip_freef(ipq[sum].prev);
842 			}
843 		}
844 found:
845 		/*
846 		 * Adjust ip_len to not reflect header,
847 		 * convert offset of this to bytes.
848 		 */
849 		ip->ip_len -= hlen;
850 		if (ip->ip_off & IP_MF) {
851 			/*
852 			 * Make sure that fragments have a data length
853 			 * that's a non-zero multiple of 8 bytes.
854 			 */
855 			if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
856 				ipstat.ips_toosmall++; /* XXX */
857 				goto bad;
858 			}
859 			m->m_flags |= M_FRAG;
860 		} else
861 			m->m_flags &= ~M_FRAG;
862 		ip->ip_off <<= 3;
863 
864 		/*
865 		 * Attempt reassembly; if it succeeds, proceed.
866 		 * ip_reass() will return a different mbuf, and update
867 		 * the divert info in divert_info and args.divert_rule.
868 		 */
869 		ipstat.ips_fragments++;
870 		m->m_pkthdr.header = ip;
871 		m = ip_reass(m, fp, &ipq[sum], &divert_info, &args.divert_rule);
872 		if (m == NULL)
873 			return;
874 		ipstat.ips_reassembled++;
875 		needredispatch = TRUE;
876 		ip = mtod(m, struct ip *);
877 		/* Get the header length of the reassembled packet */
878 		hlen = IP_VHL_HL(ip->ip_vhl) << 2;
879 #ifdef IPDIVERT
880 		/* Restore original checksum before diverting packet */
881 		if (divert_info != 0) {
882 			ip->ip_len += hlen;
883 			ip->ip_len = htons(ip->ip_len);
884 			ip->ip_off = htons(ip->ip_off);
885 			ip->ip_sum = 0;
886 			if (hlen == sizeof(struct ip))
887 				ip->ip_sum = in_cksum_hdr(ip);
888 			else
889 				ip->ip_sum = in_cksum(m, hlen);
890 			ip->ip_off = ntohs(ip->ip_off);
891 			ip->ip_len = ntohs(ip->ip_len);
892 			ip->ip_len -= hlen;
893 		}
894 #endif
895 	} else {
896 		ip->ip_len -= hlen;
897 	}
898 
899 #ifdef IPDIVERT
900 	/*
901 	 * Divert or tee packet to the divert protocol if required.
902 	 */
903 	if (divert_info != 0) {
904 		struct mbuf *clone = NULL;
905 
906 		/* Clone packet if we're doing a 'tee' */
907 		if ((divert_info & IP_FW_PORT_TEE_FLAG) != 0)
908 			clone = m_dup(m, M_DONTWAIT);
909 
910 		/* Restore packet header fields to original values */
911 		ip->ip_len += hlen;
912 		ip->ip_len = htons(ip->ip_len);
913 		ip->ip_off = htons(ip->ip_off);
914 
915 		/* Deliver packet to divert input routine */
916 		divert_packet(m, 1, divert_info & 0xffff, args.divert_rule);
917 		ipstat.ips_delivered++;
918 
919 		/* If 'tee', continue with original packet */
920 		if (clone == NULL)
921 			return;
922 		m = clone;
923 		ip = mtod(m, struct ip *);
924 		ip->ip_len += hlen;
925 		/*
926 		 * Jump backwards to complete processing of the
927 		 * packet. But first clear divert_info to avoid
928 		 * entering this block again.
929 		 * We do not need to clear args.divert_rule
930 		 * or args.next_hop as they will not be used.
931 		 */
932 		divert_info = 0;
933 		goto pass;
934 	}
935 #endif
936 
937 #ifdef IPSEC
938 	/*
939 	 * enforce IPsec policy checking if we are seeing last header.
940 	 * note that we do not visit this with protocols with pcb layer
941 	 * code - like udp/tcp/raw ip.
942 	 */
943 	if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) &&
944 	    ipsec4_in_reject(m, NULL)) {
945 		ipsecstat.in_polvio++;
946 		goto bad;
947 	}
948 #endif
949 #if FAST_IPSEC
950 	/*
951 	 * enforce IPsec policy checking if we are seeing last header.
952 	 * note that we do not visit this with protocols with pcb layer
953 	 * code - like udp/tcp/raw ip.
954 	 */
955 	if (inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) {
956 		/*
957 		 * Check if the packet has already had IPsec processing
958 		 * done.  If so, then just pass it along.  This tag gets
959 		 * set during AH, ESP, etc. input handling, before the
960 		 * packet is returned to the ip input queue for delivery.
961 		 */
962 		mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
963 		s = splnet();
964 		if (mtag != NULL) {
965 			tdbi = (struct tdb_ident *)(mtag + 1);
966 			sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
967 		} else {
968 			sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
969 						   IP_FORWARDING, &error);
970 		}
971 		if (sp != NULL) {
972 			/*
973 			 * Check security policy against packet attributes.
974 			 */
975 			error = ipsec_in_reject(sp, m);
976 			KEY_FREESP(&sp);
977 		} else {
978 			/* XXX error stat??? */
979 			error = EINVAL;
980 DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/
981 			goto bad;
982 		}
983 		splx(s);
984 		if (error)
985 			goto bad;
986 	}
987 #endif /* FAST_IPSEC */
988 
989 	ipstat.ips_delivered++;
990 	if (needredispatch) {
991 		struct netmsg_transport_packet *msg;
992 		lwkt_port_t port;
993 
994 		msg = malloc(sizeof(struct netmsg_transport_packet),
995 		    M_LWKTMSG, M_NOWAIT);
996 		if (!msg)
997 			goto bad;
998 		lwkt_initmsg_rp(&msg->nm_lmsg, &netisr_afree_rport,
999 		    CMD_NETMSG_ONCPU);
1000 		msg->nm_handler = transport_processing_handler;
1001 		msg->nm_mbuf = m;
1002 		msg->nm_hlen = hlen;
1003 		msg->nm_hasnexthop = (args.next_hop != NULL);
1004 		if (msg->nm_hasnexthop)
1005 			msg->nm_nexthop = *args.next_hop;  /* structure copy */
1006 
1007 		ip->ip_off = htons(ip->ip_off);
1008 		ip->ip_len = htons(ip->ip_len);
1009 		port = ip_mport(m);
1010 		ip->ip_len = ntohs(ip->ip_len);
1011 		ip->ip_off = ntohs(ip->ip_off);
1012 
1013 		lwkt_sendmsg(port, &msg->nm_lmsg);
1014 	} else {
1015 		transport_processing_oncpu(m, hlen, ip, args.next_hop);
1016 	}
1017 	return;
1018 
1019 bad:
1020 	m_freem(m);
1021 }
1022 
1023 /*
1024  * Take incoming datagram fragment and try to reassemble it into
1025  * whole datagram.  If a chain for reassembly of this datagram already
1026  * exists, then it is given as fp; otherwise have to make a chain.
1027  *
1028  * When IPDIVERT enabled, keep additional state with each packet that
1029  * tells us if we need to divert or tee the packet we're building.
1030  * In particular, *divinfo includes the port and TEE flag,
1031  * *divert_rule is the number of the matching rule.
1032  */
1033 
1034 static struct mbuf *
1035 ip_reass(struct mbuf *m, struct ipq *fp, struct ipq *where,
1036 	 u_int32_t *divinfo, u_int16_t *divert_rule)
1037 {
1038 	struct ip *ip = mtod(m, struct ip *);
1039 	struct mbuf *p = NULL, *q, *nq;
1040 	struct mbuf *t;
1041 	int hlen = IP_VHL_HL(ip->ip_vhl) << 2;
1042 	int i, next;
1043 
1044 	/*
1045 	 * Presence of header sizes in mbufs
1046 	 * would confuse code below.
1047 	 */
1048 	m->m_data += hlen;
1049 	m->m_len -= hlen;
1050 
1051 	/*
1052 	 * If first fragment to arrive, create a reassembly queue.
1053 	 */
1054 	if (fp == NULL) {
1055 		if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL)
1056 			goto dropfrag;
1057 		fp = mtod(t, struct ipq *);
1058 		insque(fp, where);
1059 		nipq++;
1060 		fp->ipq_nfrags = 1;
1061 		fp->ipq_ttl = IPFRAGTTL;
1062 		fp->ipq_p = ip->ip_p;
1063 		fp->ipq_id = ip->ip_id;
1064 		fp->ipq_src = ip->ip_src;
1065 		fp->ipq_dst = ip->ip_dst;
1066 		fp->ipq_frags = m;
1067 		m->m_nextpkt = NULL;
1068 #ifdef IPDIVERT
1069 		fp->ipq_div_info = 0;
1070 		fp->ipq_div_cookie = 0;
1071 #endif
1072 		goto inserted;
1073 	} else {
1074 		fp->ipq_nfrags++;
1075 	}
1076 
1077 #define	GETIP(m)	((struct ip*)((m)->m_pkthdr.header))
1078 
1079 	/*
1080 	 * Find a segment which begins after this one does.
1081 	 */
1082 	for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt)
1083 		if (GETIP(q)->ip_off > ip->ip_off)
1084 			break;
1085 
1086 	/*
1087 	 * If there is a preceding segment, it may provide some of
1088 	 * our data already.  If so, drop the data from the incoming
1089 	 * segment.  If it provides all of our data, drop us, otherwise
1090 	 * stick new segment in the proper place.
1091 	 *
1092 	 * If some of the data is dropped from the the preceding
1093 	 * segment, then it's checksum is invalidated.
1094 	 */
1095 	if (p) {
1096 		i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off;
1097 		if (i > 0) {
1098 			if (i >= ip->ip_len)
1099 				goto dropfrag;
1100 			m_adj(m, i);
1101 			m->m_pkthdr.csum_flags = 0;
1102 			ip->ip_off += i;
1103 			ip->ip_len -= i;
1104 		}
1105 		m->m_nextpkt = p->m_nextpkt;
1106 		p->m_nextpkt = m;
1107 	} else {
1108 		m->m_nextpkt = fp->ipq_frags;
1109 		fp->ipq_frags = m;
1110 	}
1111 
1112 	/*
1113 	 * While we overlap succeeding segments trim them or,
1114 	 * if they are completely covered, dequeue them.
1115 	 */
1116 	for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off;
1117 	     q = nq) {
1118 		i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off;
1119 		if (i < GETIP(q)->ip_len) {
1120 			GETIP(q)->ip_len -= i;
1121 			GETIP(q)->ip_off += i;
1122 			m_adj(q, i);
1123 			q->m_pkthdr.csum_flags = 0;
1124 			break;
1125 		}
1126 		nq = q->m_nextpkt;
1127 		m->m_nextpkt = nq;
1128 		ipstat.ips_fragdropped++;
1129 		fp->ipq_nfrags--;
1130 		m_freem(q);
1131 	}
1132 
1133 inserted:
1134 
1135 #ifdef IPDIVERT
1136 	/*
1137 	 * Transfer firewall instructions to the fragment structure.
1138 	 * Only trust info in the fragment at offset 0.
1139 	 */
1140 	if (ip->ip_off == 0) {
1141 		fp->ipq_div_info = *divinfo;
1142 		fp->ipq_div_cookie = *divert_rule;
1143 	}
1144 	*divinfo = 0;
1145 	*divert_rule = 0;
1146 #endif
1147 
1148 	/*
1149 	 * Check for complete reassembly and perform frag per packet
1150 	 * limiting.
1151 	 *
1152 	 * Frag limiting is performed here so that the nth frag has
1153 	 * a chance to complete the packet before we drop the packet.
1154 	 * As a result, n+1 frags are actually allowed per packet, but
1155 	 * only n will ever be stored. (n = maxfragsperpacket.)
1156 	 *
1157 	 */
1158 	next = 0;
1159 	for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
1160 		if (GETIP(q)->ip_off != next) {
1161 			if (fp->ipq_nfrags > maxfragsperpacket) {
1162 				ipstat.ips_fragdropped += fp->ipq_nfrags;
1163 				ip_freef(fp);
1164 			}
1165 			return (NULL);
1166 		}
1167 		next += GETIP(q)->ip_len;
1168 	}
1169 	/* Make sure the last packet didn't have the IP_MF flag */
1170 	if (p->m_flags & M_FRAG) {
1171 		if (fp->ipq_nfrags > maxfragsperpacket) {
1172 			ipstat.ips_fragdropped += fp->ipq_nfrags;
1173 			ip_freef(fp);
1174 		}
1175 		return (NULL);
1176 	}
1177 
1178 	/*
1179 	 * Reassembly is complete.  Make sure the packet is a sane size.
1180 	 */
1181 	q = fp->ipq_frags;
1182 	ip = GETIP(q);
1183 	if (next + (IP_VHL_HL(ip->ip_vhl) << 2) > IP_MAXPACKET) {
1184 		ipstat.ips_toolong++;
1185 		ipstat.ips_fragdropped += fp->ipq_nfrags;
1186 		ip_freef(fp);
1187 		return (NULL);
1188 	}
1189 
1190 	/*
1191 	 * Concatenate fragments.
1192 	 */
1193 	m = q;
1194 	t = m->m_next;
1195 	m->m_next = NULL;
1196 	m_cat(m, t);
1197 	nq = q->m_nextpkt;
1198 	q->m_nextpkt = NULL;
1199 	for (q = nq; q != NULL; q = nq) {
1200 		nq = q->m_nextpkt;
1201 		q->m_nextpkt = NULL;
1202 		m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
1203 		m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
1204 		m_cat(m, q);
1205 	}
1206 
1207 #ifdef IPDIVERT
1208 	/*
1209 	 * Extract firewall instructions from the fragment structure.
1210 	 */
1211 	*divinfo = fp->ipq_div_info;
1212 	*divert_rule = fp->ipq_div_cookie;
1213 #endif
1214 
1215 	/*
1216 	 * Create header for new ip packet by
1217 	 * modifying header of first packet;
1218 	 * dequeue and discard fragment reassembly header.
1219 	 * Make header visible.
1220 	 */
1221 	ip->ip_len = next;
1222 	ip->ip_src = fp->ipq_src;
1223 	ip->ip_dst = fp->ipq_dst;
1224 	remque(fp);
1225 	nipq--;
1226 	(void) m_free(dtom(fp));
1227 	m->m_len += (IP_VHL_HL(ip->ip_vhl) << 2);
1228 	m->m_data -= (IP_VHL_HL(ip->ip_vhl) << 2);
1229 	/* some debugging cruft by sklower, below, will go away soon */
1230 	if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */
1231 		int plen = 0;
1232 
1233 		for (t = m; t; t = t->m_next)
1234 			plen += t->m_len;
1235 		m->m_pkthdr.len = plen;
1236 	}
1237 	return (m);
1238 
1239 dropfrag:
1240 #ifdef IPDIVERT
1241 	*divinfo = 0;
1242 	*divert_rule = 0;
1243 #endif
1244 	ipstat.ips_fragdropped++;
1245 	if (fp != NULL)
1246 		fp->ipq_nfrags--;
1247 	m_freem(m);
1248 	return (NULL);
1249 
1250 #undef GETIP
1251 }
1252 
1253 /*
1254  * Free a fragment reassembly header and all
1255  * associated datagrams.
1256  */
1257 static void
1258 ip_freef(struct ipq *fp)
1259 {
1260 	struct mbuf *q;
1261 
1262 	while (fp->ipq_frags) {
1263 		q = fp->ipq_frags;
1264 		fp->ipq_frags = q->m_nextpkt;
1265 		m_freem(q);
1266 	}
1267 	remque(fp);
1268 	(void) m_free(dtom(fp));
1269 	nipq--;
1270 }
1271 
1272 /*
1273  * IP timer processing;
1274  * if a timer expires on a reassembly
1275  * queue, discard it.
1276  */
1277 void
1278 ip_slowtimo()
1279 {
1280 	struct ipq *fp;
1281 	int s = splnet();
1282 	int i;
1283 
1284 	for (i = 0; i < IPREASS_NHASH; i++) {
1285 		fp = ipq[i].next;
1286 		if (fp == NULL)
1287 			continue;
1288 		while (fp != &ipq[i]) {
1289 			--fp->ipq_ttl;
1290 			fp = fp->next;
1291 			if (fp->prev->ipq_ttl == 0) {
1292 				ipstat.ips_fragtimeout += fp->prev->ipq_nfrags;
1293 				ip_freef(fp->prev);
1294 			}
1295 		}
1296 	}
1297 	/*
1298 	 * If we are over the maximum number of fragments
1299 	 * (due to the limit being lowered), drain off
1300 	 * enough to get down to the new limit.
1301 	 */
1302 	if (maxnipq >= 0 && nipq > maxnipq) {
1303 		for (i = 0; i < IPREASS_NHASH; i++) {
1304 			while (nipq > maxnipq &&
1305 				(ipq[i].next != &ipq[i])) {
1306 				ipstat.ips_fragdropped +=
1307 				    ipq[i].next->ipq_nfrags;
1308 				ip_freef(ipq[i].next);
1309 			}
1310 		}
1311 	}
1312 	ipflow_slowtimo();
1313 	splx(s);
1314 }
1315 
1316 /*
1317  * Drain off all datagram fragments.
1318  */
1319 void
1320 ip_drain()
1321 {
1322 	int i;
1323 
1324 	for (i = 0; i < IPREASS_NHASH; i++) {
1325 		while (ipq[i].next != &ipq[i]) {
1326 			ipstat.ips_fragdropped += ipq[i].next->ipq_nfrags;
1327 			ip_freef(ipq[i].next);
1328 		}
1329 	}
1330 	in_rtqdrain();
1331 }
1332 
1333 /*
1334  * Do option processing on a datagram,
1335  * possibly discarding it if bad options are encountered,
1336  * or forwarding it if source-routed.
1337  * The pass argument is used when operating in the IPSTEALTH
1338  * mode to tell what options to process:
1339  * [LS]SRR (pass 0) or the others (pass 1).
1340  * The reason for as many as two passes is that when doing IPSTEALTH,
1341  * non-routing options should be processed only if the packet is for us.
1342  * Returns 1 if packet has been forwarded/freed,
1343  * 0 if the packet should be processed further.
1344  */
1345 static int
1346 ip_dooptions(struct mbuf *m, int pass, struct sockaddr_in *next_hop)
1347 {
1348 	struct ip *ip = mtod(m, struct ip *);
1349 	u_char *cp;
1350 	struct in_ifaddr *ia;
1351 	int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB;
1352 	boolean_t forward = FALSE;
1353 	struct in_addr *sin, dst;
1354 	n_time ntime;
1355 
1356 	dst = ip->ip_dst;
1357 	cp = (u_char *)(ip + 1);
1358 	cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
1359 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1360 		opt = cp[IPOPT_OPTVAL];
1361 		if (opt == IPOPT_EOL)
1362 			break;
1363 		if (opt == IPOPT_NOP)
1364 			optlen = 1;
1365 		else {
1366 			if (cnt < IPOPT_OLEN + sizeof(*cp)) {
1367 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1368 				goto bad;
1369 			}
1370 			optlen = cp[IPOPT_OLEN];
1371 			if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
1372 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1373 				goto bad;
1374 			}
1375 		}
1376 		switch (opt) {
1377 
1378 		default:
1379 			break;
1380 
1381 		/*
1382 		 * Source routing with record.
1383 		 * Find interface with current destination address.
1384 		 * If none on this machine then drop if strictly routed,
1385 		 * or do nothing if loosely routed.
1386 		 * Record interface address and bring up next address
1387 		 * component.  If strictly routed make sure next
1388 		 * address is on directly accessible net.
1389 		 */
1390 		case IPOPT_LSRR:
1391 		case IPOPT_SSRR:
1392 			if (ipstealth && pass > 0)
1393 				break;
1394 			if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1395 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1396 				goto bad;
1397 			}
1398 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1399 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1400 				goto bad;
1401 			}
1402 			ipaddr.sin_addr = ip->ip_dst;
1403 			ia = (struct in_ifaddr *)
1404 				ifa_ifwithaddr((struct sockaddr *)&ipaddr);
1405 			if (ia == NULL) {
1406 				if (opt == IPOPT_SSRR) {
1407 					type = ICMP_UNREACH;
1408 					code = ICMP_UNREACH_SRCFAIL;
1409 					goto bad;
1410 				}
1411 				if (!ip_dosourceroute)
1412 					goto nosourcerouting;
1413 				/*
1414 				 * Loose routing, and not at next destination
1415 				 * yet; nothing to do except forward.
1416 				 */
1417 				break;
1418 			}
1419 			off--;			/* 0 origin */
1420 			if (off > optlen - (int)sizeof(struct in_addr)) {
1421 				/*
1422 				 * End of source route.  Should be for us.
1423 				 */
1424 				if (!ip_acceptsourceroute)
1425 					goto nosourcerouting;
1426 				save_rte(cp, ip->ip_src);
1427 				break;
1428 			}
1429 			if (ipstealth)
1430 				goto dropit;
1431 			if (!ip_dosourceroute) {
1432 				if (ipforwarding) {
1433 					char buf[16]; /* aaa.bbb.ccc.ddd\0 */
1434 					/*
1435 					 * Acting as a router, so generate ICMP
1436 					 */
1437 nosourcerouting:
1438 					strcpy(buf, inet_ntoa(ip->ip_dst));
1439 					log(LOG_WARNING,
1440 					    "attempted source route from %s to %s\n",
1441 					    inet_ntoa(ip->ip_src), buf);
1442 					type = ICMP_UNREACH;
1443 					code = ICMP_UNREACH_SRCFAIL;
1444 					goto bad;
1445 				} else {
1446 					/*
1447 					 * Not acting as a router,
1448 					 * so silently drop.
1449 					 */
1450 dropit:
1451 					ipstat.ips_cantforward++;
1452 					m_freem(m);
1453 					return (1);
1454 				}
1455 			}
1456 
1457 			/*
1458 			 * locate outgoing interface
1459 			 */
1460 			(void)memcpy(&ipaddr.sin_addr, cp + off,
1461 			    sizeof(ipaddr.sin_addr));
1462 
1463 			if (opt == IPOPT_SSRR) {
1464 #define	INA	struct in_ifaddr *
1465 #define	SA	struct sockaddr *
1466 				if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr))
1467 									== NULL)
1468 					ia = (INA)ifa_ifwithnet((SA)&ipaddr);
1469 			} else
1470 				ia = ip_rtaddr(ipaddr.sin_addr, &ipforward_rt);
1471 			if (ia == NULL) {
1472 				type = ICMP_UNREACH;
1473 				code = ICMP_UNREACH_SRCFAIL;
1474 				goto bad;
1475 			}
1476 			ip->ip_dst = ipaddr.sin_addr;
1477 			(void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
1478 			    sizeof(struct in_addr));
1479 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1480 			/*
1481 			 * Let ip_intr's mcast routing check handle mcast pkts
1482 			 */
1483 			forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr));
1484 			break;
1485 
1486 		case IPOPT_RR:
1487 			if (ipstealth && pass == 0)
1488 				break;
1489 			if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1490 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1491 				goto bad;
1492 			}
1493 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1494 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1495 				goto bad;
1496 			}
1497 			/*
1498 			 * If no space remains, ignore.
1499 			 */
1500 			off--;			/* 0 origin */
1501 			if (off > optlen - (int)sizeof(struct in_addr))
1502 				break;
1503 			(void)memcpy(&ipaddr.sin_addr, &ip->ip_dst,
1504 			    sizeof(ipaddr.sin_addr));
1505 			/*
1506 			 * locate outgoing interface; if we're the destination,
1507 			 * use the incoming interface (should be same).
1508 			 */
1509 			if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == NULL &&
1510 			    (ia = ip_rtaddr(ipaddr.sin_addr, &ipforward_rt))
1511 								     == NULL) {
1512 				type = ICMP_UNREACH;
1513 				code = ICMP_UNREACH_HOST;
1514 				goto bad;
1515 			}
1516 			(void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
1517 			    sizeof(struct in_addr));
1518 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1519 			break;
1520 
1521 		case IPOPT_TS:
1522 			if (ipstealth && pass == 0)
1523 				break;
1524 			code = cp - (u_char *)ip;
1525 			if (optlen < 4 || optlen > 40) {
1526 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1527 				goto bad;
1528 			}
1529 			if ((off = cp[IPOPT_OFFSET]) < 5) {
1530 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1531 				goto bad;
1532 			}
1533 			if (off > optlen - (int)sizeof(int32_t)) {
1534 				cp[IPOPT_OFFSET + 1] += (1 << 4);
1535 				if ((cp[IPOPT_OFFSET + 1] & 0xf0) == 0) {
1536 					code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1537 					goto bad;
1538 				}
1539 				break;
1540 			}
1541 			off--;				/* 0 origin */
1542 			sin = (struct in_addr *)(cp + off);
1543 			switch (cp[IPOPT_OFFSET + 1] & 0x0f) {
1544 
1545 			case IPOPT_TS_TSONLY:
1546 				break;
1547 
1548 			case IPOPT_TS_TSANDADDR:
1549 				if (off + sizeof(n_time) +
1550 				    sizeof(struct in_addr) > optlen) {
1551 					code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1552 					goto bad;
1553 				}
1554 				ipaddr.sin_addr = dst;
1555 				ia = (INA)ifaof_ifpforaddr((SA)&ipaddr,
1556 							    m->m_pkthdr.rcvif);
1557 				if (ia == NULL)
1558 					continue;
1559 				(void)memcpy(sin, &IA_SIN(ia)->sin_addr,
1560 				    sizeof(struct in_addr));
1561 				cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1562 				off += sizeof(struct in_addr);
1563 				break;
1564 
1565 			case IPOPT_TS_PRESPEC:
1566 				if (off + sizeof(n_time) +
1567 				    sizeof(struct in_addr) > optlen) {
1568 					code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1569 					goto bad;
1570 				}
1571 				(void)memcpy(&ipaddr.sin_addr, sin,
1572 				    sizeof(struct in_addr));
1573 				if (ifa_ifwithaddr((SA)&ipaddr) == NULL)
1574 					continue;
1575 				cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1576 				off += sizeof(struct in_addr);
1577 				break;
1578 
1579 			default:
1580 				code = &cp[IPOPT_OFFSET + 1] - (u_char *)ip;
1581 				goto bad;
1582 			}
1583 			ntime = iptime();
1584 			(void)memcpy(cp + off, &ntime, sizeof(n_time));
1585 			cp[IPOPT_OFFSET] += sizeof(n_time);
1586 		}
1587 	}
1588 	if (forward && ipforwarding) {
1589 		ip_forward(m, 1, next_hop);
1590 		return (1);
1591 	}
1592 	return (0);
1593 bad:
1594 	icmp_error(m, type, code, 0, NULL);
1595 	ipstat.ips_badoptions++;
1596 	return (1);
1597 }
1598 
1599 /*
1600  * Given address of next destination (final or next hop),
1601  * return internet address info of interface to be used to get there.
1602  */
1603 struct in_ifaddr *
1604 ip_rtaddr(struct in_addr dst, struct route *rt)
1605 {
1606 	struct sockaddr_in *sin;
1607 
1608 	sin = (struct sockaddr_in *)&rt->ro_dst;
1609 
1610 	if (rt->ro_rt == NULL || dst.s_addr != sin->sin_addr.s_addr) {
1611 		if (rt->ro_rt) {
1612 			RTFREE(rt->ro_rt);
1613 			rt->ro_rt = NULL;
1614 		}
1615 		sin->sin_family = AF_INET;
1616 		sin->sin_len = sizeof(*sin);
1617 		sin->sin_addr = dst;
1618 		rtalloc_ign(rt, RTF_PRCLONING);
1619 	}
1620 
1621 	if (rt->ro_rt == NULL)
1622 		return (NULL);
1623 
1624 	return (ifatoia(rt->ro_rt->rt_ifa));
1625 }
1626 
1627 /*
1628  * Save incoming source route for use in replies,
1629  * to be picked up later by ip_srcroute if the receiver is interested.
1630  */
1631 void
1632 save_rte(u_char *option, struct in_addr dst)
1633 {
1634 	unsigned olen;
1635 
1636 	olen = option[IPOPT_OLEN];
1637 #ifdef DIAGNOSTIC
1638 	if (ipprintfs)
1639 		printf("save_rte: olen %d\n", olen);
1640 #endif
1641 	if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst)))
1642 		return;
1643 	bcopy(option, ip_srcrt.srcopt, olen);
1644 	ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1645 	ip_srcrt.dst = dst;
1646 }
1647 
1648 /*
1649  * Retrieve incoming source route for use in replies,
1650  * in the same form used by setsockopt.
1651  * The first hop is placed before the options, will be removed later.
1652  */
1653 struct mbuf *
1654 ip_srcroute()
1655 {
1656 	struct in_addr *p, *q;
1657 	struct mbuf *m;
1658 
1659 	if (ip_nhops == 0)
1660 		return (NULL);
1661 	m = m_get(M_DONTWAIT, MT_HEADER);
1662 	if (m == NULL)
1663 		return (NULL);
1664 
1665 #define	OPTSIZ	(sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt))
1666 
1667 	/* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
1668 	m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) +
1669 	    OPTSIZ;
1670 #ifdef DIAGNOSTIC
1671 	if (ipprintfs)
1672 		printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len);
1673 #endif
1674 
1675 	/*
1676 	 * First save first hop for return route
1677 	 */
1678 	p = &ip_srcrt.route[ip_nhops - 1];
1679 	*(mtod(m, struct in_addr *)) = *p--;
1680 #ifdef DIAGNOSTIC
1681 	if (ipprintfs)
1682 		printf(" hops %lx", ntohl(mtod(m, struct in_addr *)->s_addr));
1683 #endif
1684 
1685 	/*
1686 	 * Copy option fields and padding (nop) to mbuf.
1687 	 */
1688 	ip_srcrt.nop = IPOPT_NOP;
1689 	ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
1690 	(void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr), &ip_srcrt.nop,
1691 	    OPTSIZ);
1692 	q = (struct in_addr *)(mtod(m, caddr_t) +
1693 	    sizeof(struct in_addr) + OPTSIZ);
1694 #undef OPTSIZ
1695 	/*
1696 	 * Record return path as an IP source route,
1697 	 * reversing the path (pointers are now aligned).
1698 	 */
1699 	while (p >= ip_srcrt.route) {
1700 #ifdef DIAGNOSTIC
1701 		if (ipprintfs)
1702 			printf(" %lx", ntohl(q->s_addr));
1703 #endif
1704 		*q++ = *p--;
1705 	}
1706 	/*
1707 	 * Last hop goes to final destination.
1708 	 */
1709 	*q = ip_srcrt.dst;
1710 #ifdef DIAGNOSTIC
1711 	if (ipprintfs)
1712 		printf(" %lx\n", ntohl(q->s_addr));
1713 #endif
1714 	return (m);
1715 }
1716 
1717 /*
1718  * Strip out IP options.
1719  */
1720 void
1721 ip_stripoptions(struct mbuf *m)
1722 {
1723 	int datalen;
1724 	struct ip *ip = mtod(m, struct ip *);
1725 	caddr_t opts;
1726 	int optlen;
1727 
1728 	optlen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
1729 	opts = (caddr_t)(ip + 1);
1730 	datalen = m->m_len - (sizeof(struct ip) + optlen);
1731 	bcopy(opts + optlen, opts, datalen);
1732 	m->m_len -= optlen;
1733 	if (m->m_flags & M_PKTHDR)
1734 		m->m_pkthdr.len -= optlen;
1735 	ip->ip_vhl = IP_MAKE_VHL(IPVERSION, sizeof(struct ip) >> 2);
1736 }
1737 
1738 u_char inetctlerrmap[PRC_NCMDS] = {
1739 	0,		0,		0,		0,
1740 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1741 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1742 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1743 	0,		0,		0,		0,
1744 	ENOPROTOOPT,	ECONNREFUSED
1745 };
1746 
1747 /*
1748  * Forward a packet.  If some error occurs return the sender
1749  * an icmp packet.  Note we can't always generate a meaningful
1750  * icmp message because icmp doesn't have a large enough repertoire
1751  * of codes and types.
1752  *
1753  * If not forwarding, just drop the packet.  This could be confusing
1754  * if ipforwarding was zero but some routing protocol was advancing
1755  * us as a gateway to somewhere.  However, we must let the routing
1756  * protocol deal with that.
1757  *
1758  * The using_srcrt parameter indicates whether the packet is being forwarded
1759  * via a source route.
1760  */
1761 static void
1762 ip_forward(struct mbuf *m, int using_srcrt, struct sockaddr_in *next_hop)
1763 {
1764 	struct ip *ip = mtod(m, struct ip *);
1765 	struct sockaddr_in *sin;
1766 	struct rtentry *rt;
1767 	int error, type = 0, code = 0;
1768 	struct mbuf *mcopy;
1769 	n_long dest;
1770 	struct in_addr pkt_dst;
1771 	struct ifnet *destifp;
1772 	struct m_hdr tag;
1773 #if defined(IPSEC) || defined(FAST_IPSEC)
1774 	struct ifnet dummyifp;
1775 #endif
1776 
1777 	dest = 0;
1778 	/*
1779 	 * Cache the destination address of the packet; this may be
1780 	 * changed by use of 'ipfw fwd'.
1781 	 */
1782 	pkt_dst = next_hop ? next_hop->sin_addr : ip->ip_dst;
1783 
1784 #ifdef DIAGNOSTIC
1785 	if (ipprintfs)
1786 		printf("forward: src %lx dst %lx ttl %x\n",
1787 		       ip->ip_src.s_addr, pkt_dst.s_addr, ip->ip_ttl);
1788 #endif
1789 
1790 	if (m->m_flags & (M_BCAST | M_MCAST) || !in_canforward(pkt_dst)) {
1791 		ipstat.ips_cantforward++;
1792 		m_freem(m);
1793 		return;
1794 	}
1795 	if (!ipstealth && ip->ip_ttl <= IPTTLDEC) {
1796 		icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, NULL);
1797 		return;
1798 	}
1799 
1800 	sin = (struct sockaddr_in *)&ipforward_rt.ro_dst;
1801 	if ((rt = ipforward_rt.ro_rt) == NULL ||
1802 	    pkt_dst.s_addr != sin->sin_addr.s_addr) {
1803 		if (ipforward_rt.ro_rt) {
1804 			RTFREE(ipforward_rt.ro_rt);
1805 			ipforward_rt.ro_rt = NULL;
1806 		}
1807 		sin->sin_family = AF_INET;
1808 		sin->sin_len = sizeof(*sin);
1809 		sin->sin_addr = pkt_dst;
1810 
1811 		rtalloc_ign(&ipforward_rt, RTF_PRCLONING);
1812 		if (ipforward_rt.ro_rt == NULL) {
1813 			icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest,
1814 				   NULL);
1815 			return;
1816 		}
1817 		rt = ipforward_rt.ro_rt;
1818 	}
1819 
1820 	/*
1821 	 * Save the IP header and at most 8 bytes of the payload,
1822 	 * in case we need to generate an ICMP message to the src.
1823 	 *
1824 	 * XXX this can be optimized a lot by saving the data in a local
1825 	 * buffer on the stack (72 bytes at most), and only allocating the
1826 	 * mbuf if really necessary. The vast majority of the packets
1827 	 * are forwarded without having to send an ICMP back (either
1828 	 * because unnecessary, or because rate limited), so we are
1829 	 * really we are wasting a lot of work here.
1830 	 *
1831 	 * We don't use m_copy() because it might return a reference
1832 	 * to a shared cluster. Both this function and ip_output()
1833 	 * assume exclusive access to the IP header in `m', so any
1834 	 * data in a cluster may change before we reach icmp_error().
1835 	 */
1836 	MGET(mcopy, M_DONTWAIT, m->m_type);
1837 	if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_DONTWAIT)) {
1838 		/*
1839 		 * It's probably ok if the pkthdr dup fails (because
1840 		 * the deep copy of the tag chain failed), but for now
1841 		 * be conservative and just discard the copy since
1842 		 * code below may some day want the tags.
1843 		 */
1844 		m_free(mcopy);
1845 		mcopy = NULL;
1846 	}
1847 	if (mcopy != NULL) {
1848 		mcopy->m_len = imin((IP_VHL_HL(ip->ip_vhl) << 2) + 8,
1849 		    (int)ip->ip_len);
1850 		m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1851 	}
1852 
1853 	if (!ipstealth)
1854 		ip->ip_ttl -= IPTTLDEC;
1855 
1856 	/*
1857 	 * If forwarding packet using same interface that it came in on,
1858 	 * perhaps should send a redirect to sender to shortcut a hop.
1859 	 * Only send redirect if source is sending directly to us,
1860 	 * and if packet was not source routed (or has any options).
1861 	 * Also, don't send redirect if forwarding using a default route
1862 	 * or a route modified by a redirect.
1863 	 */
1864 	if (rt->rt_ifp == m->m_pkthdr.rcvif &&
1865 	    !(rt->rt_flags & (RTF_DYNAMIC | RTF_MODIFIED)) &&
1866 	    satosin(rt_key(rt))->sin_addr.s_addr != INADDR_ANY &&
1867 	    ipsendredirects && !using_srcrt && next_hop != NULL) {
1868 		u_long src = ntohl(ip->ip_src.s_addr);
1869 
1870 #define	RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa))
1871 		if (RTA(rt) != NULL &&
1872 		    (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1873 			if (rt->rt_flags & RTF_GATEWAY)
1874 				dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
1875 			else
1876 				dest = pkt_dst.s_addr;
1877 			/*
1878 			 * Router requirements says to only send
1879 			 * host redirects.
1880 			 */
1881 			type = ICMP_REDIRECT;
1882 			code = ICMP_REDIRECT_HOST;
1883 #ifdef DIAGNOSTIC
1884 			if (ipprintfs)
1885 				printf("redirect (%d) to %lx\n", code, dest);
1886 #endif
1887 		}
1888 	}
1889 
1890 	if (next_hop) {
1891 		/* Pass IPFORWARD info if available */
1892 
1893 		tag.mh_type = MT_TAG;
1894 		tag.mh_flags = PACKET_TAG_IPFORWARD;
1895 		tag.mh_data = (caddr_t)next_hop;
1896 		tag.mh_next = m;
1897 		m = (struct mbuf *)&tag;
1898 	}
1899 
1900 	error = ip_output(m, NULL, &ipforward_rt, IP_FORWARDING, NULL, NULL);
1901 
1902 	if (error)
1903 		ipstat.ips_cantforward++;
1904 	else {
1905 		ipstat.ips_forward++;
1906 		if (type)
1907 			ipstat.ips_redirectsent++;
1908 		else {
1909 			if (mcopy) {
1910 				ipflow_create(&ipforward_rt, mcopy);
1911 				m_freem(mcopy);
1912 			}
1913 			return;
1914 		}
1915 	}
1916 	if (mcopy == NULL)
1917 		return;
1918 	destifp = NULL;
1919 
1920 	switch (error) {
1921 
1922 	case 0:				/* forwarded, but need redirect */
1923 		/* type, code set above */
1924 		break;
1925 
1926 	case ENETUNREACH:		/* shouldn't happen, checked above */
1927 	case EHOSTUNREACH:
1928 	case ENETDOWN:
1929 	case EHOSTDOWN:
1930 	default:
1931 		type = ICMP_UNREACH;
1932 		code = ICMP_UNREACH_HOST;
1933 		break;
1934 
1935 	case EMSGSIZE:
1936 		type = ICMP_UNREACH;
1937 		code = ICMP_UNREACH_NEEDFRAG;
1938 #ifdef IPSEC
1939 		/*
1940 		 * If the packet is routed over IPsec tunnel, tell the
1941 		 * originator the tunnel MTU.
1942 		 *	tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
1943 		 * XXX quickhack!!!
1944 		 */
1945 		if (ipforward_rt.ro_rt) {
1946 			struct secpolicy *sp = NULL;
1947 			int ipsecerror;
1948 			int ipsechdr;
1949 			struct route *ro;
1950 
1951 			sp = ipsec4_getpolicybyaddr(mcopy,
1952 						    IPSEC_DIR_OUTBOUND,
1953 						    IP_FORWARDING,
1954 						    &ipsecerror);
1955 
1956 			if (sp == NULL)
1957 				destifp = ipforward_rt.ro_rt->rt_ifp;
1958 			else {
1959 				/* count IPsec header size */
1960 				ipsechdr = ipsec4_hdrsiz(mcopy,
1961 							 IPSEC_DIR_OUTBOUND,
1962 							 NULL);
1963 
1964 				/*
1965 				 * find the correct route for outer IPv4
1966 				 * header, compute tunnel MTU.
1967 				 *
1968 				 * XXX BUG ALERT
1969 				 * The "dummyifp" code relies upon the fact
1970 				 * that icmp_error() touches only ifp->if_mtu.
1971 				 */
1972 				/*XXX*/
1973 				destifp = NULL;
1974 				if (sp->req != NULL
1975 				 && sp->req->sav != NULL
1976 				 && sp->req->sav->sah != NULL) {
1977 					ro = &sp->req->sav->sah->sa_route;
1978 					if (ro->ro_rt && ro->ro_rt->rt_ifp) {
1979 						dummyifp.if_mtu =
1980 						    ro->ro_rt->rt_ifp->if_mtu;
1981 						dummyifp.if_mtu -= ipsechdr;
1982 						destifp = &dummyifp;
1983 					}
1984 				}
1985 
1986 				key_freesp(sp);
1987 			}
1988 		}
1989 #elif FAST_IPSEC
1990 		/*
1991 		 * If the packet is routed over IPsec tunnel, tell the
1992 		 * originator the tunnel MTU.
1993 		 *	tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
1994 		 * XXX quickhack!!!
1995 		 */
1996 		if (ipforward_rt.ro_rt) {
1997 			struct secpolicy *sp = NULL;
1998 			int ipsecerror;
1999 			int ipsechdr;
2000 			struct route *ro;
2001 
2002 			sp = ipsec_getpolicybyaddr(mcopy,
2003 						   IPSEC_DIR_OUTBOUND,
2004 						   IP_FORWARDING,
2005 						   &ipsecerror);
2006 
2007 			if (sp == NULL)
2008 				destifp = ipforward_rt.ro_rt->rt_ifp;
2009 			else {
2010 				/* count IPsec header size */
2011 				ipsechdr = ipsec4_hdrsiz(mcopy,
2012 							 IPSEC_DIR_OUTBOUND,
2013 							 NULL);
2014 
2015 				/*
2016 				 * find the correct route for outer IPv4
2017 				 * header, compute tunnel MTU.
2018 				 *
2019 				 * XXX BUG ALERT
2020 				 * The "dummyifp" code relies upon the fact
2021 				 * that icmp_error() touches only ifp->if_mtu.
2022 				 */
2023 				/*XXX*/
2024 				destifp = NULL;
2025 				if (sp->req != NULL
2026 				 && sp->req->sav != NULL
2027 				 && sp->req->sav->sah != NULL) {
2028 					ro = &sp->req->sav->sah->sa_route;
2029 					if (ro->ro_rt && ro->ro_rt->rt_ifp) {
2030 						dummyifp.if_mtu =
2031 						    ro->ro_rt->rt_ifp->if_mtu;
2032 						dummyifp.if_mtu -= ipsechdr;
2033 						destifp = &dummyifp;
2034 					}
2035 				}
2036 
2037 				KEY_FREESP(&sp);
2038 			}
2039 		}
2040 #else /* !IPSEC && !FAST_IPSEC */
2041 		if (ipforward_rt.ro_rt)
2042 			destifp = ipforward_rt.ro_rt->rt_ifp;
2043 #endif /*IPSEC*/
2044 		ipstat.ips_cantfrag++;
2045 		break;
2046 
2047 	case ENOBUFS:
2048 		/*
2049 		 * A router should not generate ICMP_SOURCEQUENCH as
2050 		 * required in RFC1812 Requirements for IP Version 4 Routers.
2051 		 * Source quench could be a big problem under DoS attacks,
2052 		 * or if the underlying interface is rate-limited.
2053 		 * Those who need source quench packets may re-enable them
2054 		 * via the net.inet.ip.sendsourcequench sysctl.
2055 		 */
2056 		if (!ip_sendsourcequench) {
2057 			m_freem(mcopy);
2058 			return;
2059 		} else {
2060 			type = ICMP_SOURCEQUENCH;
2061 			code = 0;
2062 		}
2063 		break;
2064 
2065 	case EACCES:			/* ipfw denied packet */
2066 		m_freem(mcopy);
2067 		return;
2068 	}
2069 	icmp_error(mcopy, type, code, dest, destifp);
2070 }
2071 
2072 void
2073 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
2074 	       struct mbuf *m)
2075 {
2076 	if (inp->inp_socket->so_options & SO_TIMESTAMP) {
2077 		struct timeval tv;
2078 
2079 		microtime(&tv);
2080 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
2081 		    SCM_TIMESTAMP, SOL_SOCKET);
2082 		if (*mp)
2083 			mp = &(*mp)->m_next;
2084 	}
2085 	if (inp->inp_flags & INP_RECVDSTADDR) {
2086 		*mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
2087 		    sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
2088 		if (*mp)
2089 			mp = &(*mp)->m_next;
2090 	}
2091 #ifdef notyet
2092 	/* XXX
2093 	 * Moving these out of udp_input() made them even more broken
2094 	 * than they already were.
2095 	 */
2096 	/* options were tossed already */
2097 	if (inp->inp_flags & INP_RECVOPTS) {
2098 		*mp = sbcreatecontrol((caddr_t) opts_deleted_above,
2099 		    sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
2100 		if (*mp)
2101 			mp = &(*mp)->m_next;
2102 	}
2103 	/* ip_srcroute doesn't do what we want here, need to fix */
2104 	if (inp->inp_flags & INP_RECVRETOPTS) {
2105 		*mp = sbcreatecontrol((caddr_t) ip_srcroute(),
2106 		    sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
2107 		if (*mp)
2108 			mp = &(*mp)->m_next;
2109 	}
2110 #endif
2111 	if (inp->inp_flags & INP_RECVIF) {
2112 		struct ifnet *ifp;
2113 		struct sdlbuf {
2114 			struct sockaddr_dl sdl;
2115 			u_char	pad[32];
2116 		} sdlbuf;
2117 		struct sockaddr_dl *sdp;
2118 		struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
2119 
2120 		if (((ifp = m->m_pkthdr.rcvif)) &&
2121 		    ((ifp->if_index != 0) && (ifp->if_index <= if_index))) {
2122 			sdp = (struct sockaddr_dl *)
2123 			    ifnet_addrs[ifp->if_index - 1]->ifa_addr;
2124 			/*
2125 			 * Change our mind and don't try copy.
2126 			 */
2127 			if ((sdp->sdl_family != AF_LINK) ||
2128 			    (sdp->sdl_len > sizeof(sdlbuf))) {
2129 				goto makedummy;
2130 			}
2131 			bcopy(sdp, sdl2, sdp->sdl_len);
2132 		} else {
2133 makedummy:
2134 			sdl2->sdl_len =
2135 			    offsetof(struct sockaddr_dl, sdl_data[0]);
2136 			sdl2->sdl_family = AF_LINK;
2137 			sdl2->sdl_index = 0;
2138 			sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
2139 		}
2140 		*mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len,
2141 			IP_RECVIF, IPPROTO_IP);
2142 		if (*mp)
2143 			mp = &(*mp)->m_next;
2144 	}
2145 }
2146 
2147 /*
2148  * XXX these routines are called from the upper part of the kernel.
2149  *
2150  * They could also be moved to ip_mroute.c, since all the RSVP
2151  *  handling is done there already.
2152  */
2153 int
2154 ip_rsvp_init(struct socket *so)
2155 {
2156 	if (so->so_type != SOCK_RAW ||
2157 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
2158 		return EOPNOTSUPP;
2159 
2160 	if (ip_rsvpd != NULL)
2161 		return EADDRINUSE;
2162 
2163 	ip_rsvpd = so;
2164 	/*
2165 	 * This may seem silly, but we need to be sure we don't over-increment
2166 	 * the RSVP counter, in case something slips up.
2167 	 */
2168 	if (!ip_rsvp_on) {
2169 		ip_rsvp_on = 1;
2170 		rsvp_on++;
2171 	}
2172 
2173 	return 0;
2174 }
2175 
2176 int
2177 ip_rsvp_done(void)
2178 {
2179 	ip_rsvpd = NULL;
2180 	/*
2181 	 * This may seem silly, but we need to be sure we don't over-decrement
2182 	 * the RSVP counter, in case something slips up.
2183 	 */
2184 	if (ip_rsvp_on) {
2185 		ip_rsvp_on = 0;
2186 		rsvp_on--;
2187 	}
2188 	return 0;
2189 }
2190 
2191 void
2192 rsvp_input(struct mbuf *m, int off, int proto)	/* XXX must fixup manually */
2193 {
2194 	if (rsvp_input_p) { /* call the real one if loaded */
2195 		rsvp_input_p(m, off, proto);
2196 		return;
2197 	}
2198 
2199 	/* Can still get packets with rsvp_on = 0 if there is a local member
2200 	 * of the group to which the RSVP packet is addressed.  But in this
2201 	 * case we want to throw the packet away.
2202 	 */
2203 
2204 	if (!rsvp_on) {
2205 		m_freem(m);
2206 		return;
2207 	}
2208 
2209 	if (ip_rsvpd != NULL) {
2210 		rip_input(m, off, proto);
2211 		return;
2212 	}
2213 	/* Drop the packet */
2214 	m_freem(m);
2215 }
2216