xref: /dflybsd-src/sys/netinet/ip_input.c (revision 9b42cabedff59ea3b351e8eccb2da9bab5cefe7e)
1 /*
2  * Copyright (c) 2003, 2004 Jeffrey M. Hsu.  All rights reserved.
3  * Copyright (c) 2003, 2004 The DragonFly Project.  All rights reserved.
4  *
5  * This code is derived from software contributed to The DragonFly Project
6  * by Jeffrey M. Hsu.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of The DragonFly Project nor the names of its
17  *    contributors may be used to endorse or promote products derived
18  *    from this software without specific, prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
21  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
22  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
23  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE
24  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
25  * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
26  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
27  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
28  * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
29  * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
30  * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  */
33 
34 /*
35  * Copyright (c) 1982, 1986, 1988, 1993
36  *	The Regents of the University of California.  All rights reserved.
37  *
38  * Redistribution and use in source and binary forms, with or without
39  * modification, are permitted provided that the following conditions
40  * are met:
41  * 1. Redistributions of source code must retain the above copyright
42  *    notice, this list of conditions and the following disclaimer.
43  * 2. Redistributions in binary form must reproduce the above copyright
44  *    notice, this list of conditions and the following disclaimer in the
45  *    documentation and/or other materials provided with the distribution.
46  * 3. All advertising materials mentioning features or use of this software
47  *    must display the following acknowledgement:
48  *	This product includes software developed by the University of
49  *	California, Berkeley and its contributors.
50  * 4. Neither the name of the University nor the names of its contributors
51  *    may be used to endorse or promote products derived from this software
52  *    without specific prior written permission.
53  *
54  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
55  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
56  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
57  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
58  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
59  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
60  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
61  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
62  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
63  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
64  * SUCH DAMAGE.
65  *
66  *	@(#)ip_input.c	8.2 (Berkeley) 1/4/94
67  * $FreeBSD: src/sys/netinet/ip_input.c,v 1.130.2.52 2003/03/07 07:01:28 silby Exp $
68  * $DragonFly: src/sys/netinet/ip_input.c,v 1.82 2008/07/03 08:22:36 corecode Exp $
69  */
70 
71 #define	_IP_VHL
72 
73 #include "opt_bootp.h"
74 #include "opt_ipfw.h"
75 #include "opt_ipdn.h"
76 #include "opt_ipdivert.h"
77 #include "opt_ipfilter.h"
78 #include "opt_ipstealth.h"
79 #include "opt_ipsec.h"
80 
81 #include <sys/param.h>
82 #include <sys/systm.h>
83 #include <sys/mbuf.h>
84 #include <sys/malloc.h>
85 #include <sys/mpipe.h>
86 #include <sys/domain.h>
87 #include <sys/protosw.h>
88 #include <sys/socket.h>
89 #include <sys/time.h>
90 #include <sys/globaldata.h>
91 #include <sys/thread.h>
92 #include <sys/kernel.h>
93 #include <sys/syslog.h>
94 #include <sys/sysctl.h>
95 #include <sys/in_cksum.h>
96 
97 #include <machine/stdarg.h>
98 
99 #include <net/if.h>
100 #include <net/if_types.h>
101 #include <net/if_var.h>
102 #include <net/if_dl.h>
103 #include <net/pfil.h>
104 #include <net/route.h>
105 #include <net/netisr.h>
106 
107 #include <netinet/in.h>
108 #include <netinet/in_systm.h>
109 #include <netinet/in_var.h>
110 #include <netinet/ip.h>
111 #include <netinet/in_pcb.h>
112 #include <netinet/ip_var.h>
113 #include <netinet/ip_icmp.h>
114 
115 #include <sys/thread2.h>
116 #include <sys/msgport2.h>
117 #include <net/netmsg2.h>
118 
119 #include <sys/socketvar.h>
120 
121 #include <net/ipfw/ip_fw.h>
122 #include <net/dummynet/ip_dummynet.h>
123 
124 #ifdef IPSEC
125 #include <netinet6/ipsec.h>
126 #include <netproto/key/key.h>
127 #endif
128 
129 #ifdef FAST_IPSEC
130 #include <netproto/ipsec/ipsec.h>
131 #include <netproto/ipsec/key.h>
132 #endif
133 
134 int rsvp_on = 0;
135 static int ip_rsvp_on;
136 struct socket *ip_rsvpd;
137 
138 int ipforwarding = 0;
139 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW,
140     &ipforwarding, 0, "Enable IP forwarding between interfaces");
141 
142 static int ipsendredirects = 1; /* XXX */
143 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW,
144     &ipsendredirects, 0, "Enable sending IP redirects");
145 
146 int ip_defttl = IPDEFTTL;
147 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW,
148     &ip_defttl, 0, "Maximum TTL on IP packets");
149 
150 static int ip_dosourceroute = 0;
151 SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW,
152     &ip_dosourceroute, 0, "Enable forwarding source routed IP packets");
153 
154 static int ip_acceptsourceroute = 0;
155 SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute,
156     CTLFLAG_RW, &ip_acceptsourceroute, 0,
157     "Enable accepting source routed IP packets");
158 
159 static int ip_keepfaith = 0;
160 SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW,
161     &ip_keepfaith, 0,
162     "Enable packet capture for FAITH IPv4->IPv6 translator daemon");
163 
164 static int nipq = 0;	/* total # of reass queues */
165 static int maxnipq;
166 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragpackets, CTLFLAG_RW,
167     &maxnipq, 0,
168     "Maximum number of IPv4 fragment reassembly queue entries");
169 
170 static int maxfragsperpacket;
171 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW,
172     &maxfragsperpacket, 0,
173     "Maximum number of IPv4 fragments allowed per packet");
174 
175 static int ip_sendsourcequench = 0;
176 SYSCTL_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW,
177     &ip_sendsourcequench, 0,
178     "Enable the transmission of source quench packets");
179 
180 int ip_do_randomid = 1;
181 SYSCTL_INT(_net_inet_ip, OID_AUTO, random_id, CTLFLAG_RW,
182     &ip_do_randomid, 0,
183     "Assign random ip_id values");
184 /*
185  * XXX - Setting ip_checkinterface mostly implements the receive side of
186  * the Strong ES model described in RFC 1122, but since the routing table
187  * and transmit implementation do not implement the Strong ES model,
188  * setting this to 1 results in an odd hybrid.
189  *
190  * XXX - ip_checkinterface currently must be disabled if you use ipnat
191  * to translate the destination address to another local interface.
192  *
193  * XXX - ip_checkinterface must be disabled if you add IP aliases
194  * to the loopback interface instead of the interface where the
195  * packets for those addresses are received.
196  */
197 static int ip_checkinterface = 0;
198 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW,
199     &ip_checkinterface, 0, "Verify packet arrives on correct interface");
200 
201 #ifdef DIAGNOSTIC
202 static int ipprintfs = 0;
203 #endif
204 
205 extern	struct domain inetdomain;
206 extern	struct protosw inetsw[];
207 u_char	ip_protox[IPPROTO_MAX];
208 struct	in_ifaddrhead in_ifaddrheads[MAXCPU];	/* first inet address */
209 struct	in_ifaddrhashhead *in_ifaddrhashtbls[MAXCPU];
210 						/* inet addr hash table */
211 u_long	in_ifaddrhmask;				/* mask for hash table */
212 
213 struct ip_stats ipstats_percpu[MAXCPU];
214 #ifdef SMP
215 static int
216 sysctl_ipstats(SYSCTL_HANDLER_ARGS)
217 {
218 	int cpu, error = 0;
219 
220 	for (cpu = 0; cpu < ncpus; ++cpu) {
221 		if ((error = SYSCTL_OUT(req, &ipstats_percpu[cpu],
222 					sizeof(struct ip_stats))))
223 			break;
224 		if ((error = SYSCTL_IN(req, &ipstats_percpu[cpu],
225 				       sizeof(struct ip_stats))))
226 			break;
227 	}
228 
229 	return (error);
230 }
231 SYSCTL_PROC(_net_inet_ip, IPCTL_STATS, stats, (CTLTYPE_OPAQUE | CTLFLAG_RW),
232     0, 0, sysctl_ipstats, "S,ip_stats", "IP statistics");
233 #else
234 SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW,
235     &ipstat, ip_stats, "IP statistics");
236 #endif
237 
238 /* Packet reassembly stuff */
239 #define	IPREASS_NHASH_LOG2	6
240 #define	IPREASS_NHASH		(1 << IPREASS_NHASH_LOG2)
241 #define	IPREASS_HMASK		(IPREASS_NHASH - 1)
242 #define	IPREASS_HASH(x,y)						\
243     (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK)
244 
245 static struct ipq ipq[IPREASS_NHASH];
246 
247 #ifdef IPCTL_DEFMTU
248 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
249     &ip_mtu, 0, "Default MTU");
250 #endif
251 
252 #ifdef IPSTEALTH
253 static int ipstealth = 0;
254 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW, &ipstealth, 0, "");
255 #else
256 static const int ipstealth = 0;
257 #endif
258 
259 
260 /* Firewall hooks */
261 ip_fw_chk_t *ip_fw_chk_ptr;
262 ip_fw_dn_io_t *ip_fw_dn_io_ptr;
263 int fw_enable = 1;
264 int fw_one_pass = 1;
265 
266 struct pfil_head inet_pfil_hook;
267 
268 /*
269  * XXX this is ugly -- the following two global variables are
270  * used to store packet state while it travels through the stack.
271  * Note that the code even makes assumptions on the size and
272  * alignment of fields inside struct ip_srcrt so e.g. adding some
273  * fields will break the code. This needs to be fixed.
274  *
275  * We need to save the IP options in case a protocol wants to respond
276  * to an incoming packet over the same route if the packet got here
277  * using IP source routing.  This allows connection establishment and
278  * maintenance when the remote end is on a network that is not known
279  * to us.
280  */
281 static int ip_nhops = 0;
282 
283 static	struct ip_srcrt {
284 	struct	in_addr dst;			/* final destination */
285 	char	nop;				/* one NOP to align */
286 	char	srcopt[IPOPT_OFFSET + 1];	/* OPTVAL, OLEN and OFFSET */
287 	struct	in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)];
288 } ip_srcrt;
289 
290 static MALLOC_DEFINE(M_IPQ, "ipq", "IP Fragment Management");
291 static struct malloc_pipe ipq_mpipe;
292 
293 static void		save_rte (u_char *, struct in_addr);
294 static int		ip_dooptions (struct mbuf *m, int,
295 					struct sockaddr_in *next_hop);
296 static void		ip_freef (struct ipq *);
297 static void		ip_input_handler (struct netmsg *);
298 static struct mbuf	*ip_reass (struct mbuf *, struct ipq *,
299 					struct ipq *, u_int32_t *);
300 
301 /*
302  * IP initialization: fill in IP protocol switch table.
303  * All protocols not implemented in kernel go to raw IP protocol handler.
304  */
305 void
306 ip_init(void)
307 {
308 	struct protosw *pr;
309 	int i;
310 #ifdef SMP
311 	int cpu;
312 #endif
313 
314 	/*
315 	 * Make sure we can handle a reasonable number of fragments but
316 	 * cap it at 4000 (XXX).
317 	 */
318 	mpipe_init(&ipq_mpipe, M_IPQ, sizeof(struct ipq),
319 		    IFQ_MAXLEN, 4000, 0, NULL);
320 	for (i = 0; i < ncpus; ++i) {
321 		TAILQ_INIT(&in_ifaddrheads[i]);
322 		in_ifaddrhashtbls[i] =
323 			hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask);
324 	}
325 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
326 	if (pr == NULL)
327 		panic("ip_init");
328 	for (i = 0; i < IPPROTO_MAX; i++)
329 		ip_protox[i] = pr - inetsw;
330 	for (pr = inetdomain.dom_protosw;
331 	     pr < inetdomain.dom_protoswNPROTOSW; pr++)
332 		if (pr->pr_domain->dom_family == PF_INET &&
333 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
334 			ip_protox[pr->pr_protocol] = pr - inetsw;
335 
336 	inet_pfil_hook.ph_type = PFIL_TYPE_AF;
337 	inet_pfil_hook.ph_af = AF_INET;
338 	if ((i = pfil_head_register(&inet_pfil_hook)) != 0) {
339 		kprintf("%s: WARNING: unable to register pfil hook, "
340 			"error %d\n", __func__, i);
341 	}
342 
343 	for (i = 0; i < IPREASS_NHASH; i++)
344 	    ipq[i].next = ipq[i].prev = &ipq[i];
345 
346 	maxnipq = nmbclusters / 32;
347 	maxfragsperpacket = 16;
348 
349 	ip_id = time_second & 0xffff;
350 
351 	/*
352 	 * Initialize IP statistics counters for each CPU.
353 	 *
354 	 */
355 #ifdef SMP
356 	for (cpu = 0; cpu < ncpus; ++cpu) {
357 		bzero(&ipstats_percpu[cpu], sizeof(struct ip_stats));
358 	}
359 #else
360 	bzero(&ipstat, sizeof(struct ip_stats));
361 #endif
362 
363 	netisr_register(NETISR_IP, ip_mport_in, ip_input_handler);
364 }
365 
366 /*
367  * XXX watch out this one. It is perhaps used as a cache for
368  * the most recently used route ? it is cleared in in_addroute()
369  * when a new route is successfully created.
370  */
371 struct route ipforward_rt[MAXCPU];
372 
373 /* Do transport protocol processing. */
374 static void
375 transport_processing_oncpu(struct mbuf *m, int hlen, struct ip *ip,
376 			   struct sockaddr_in *nexthop)
377 {
378 	/*
379 	 * Switch out to protocol's input routine.
380 	 */
381 	if (nexthop && ip->ip_p == IPPROTO_TCP) {
382 		/* TCP needs IPFORWARD info if available */
383 		struct m_hdr tag;
384 
385 		tag.mh_type = MT_TAG;
386 		tag.mh_flags = PACKET_TAG_IPFORWARD;
387 		tag.mh_data = (caddr_t)nexthop;
388 		tag.mh_next = m;
389 
390 		(*inetsw[ip_protox[ip->ip_p]].pr_input)
391 		    ((struct mbuf *)&tag, hlen, ip->ip_p);
392 	} else {
393 		(*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen, ip->ip_p);
394 	}
395 }
396 
397 struct netmsg_transport_packet {
398 	struct netmsg		nm_netmsg;
399 	struct mbuf		*nm_mbuf;
400 	int			nm_hlen;
401 	boolean_t		nm_hasnexthop;
402 	struct sockaddr_in	nm_nexthop;
403 };
404 
405 static void
406 transport_processing_handler(netmsg_t netmsg)
407 {
408 	struct netmsg_transport_packet *msg = (void *)netmsg;
409 	struct sockaddr_in *nexthop;
410 	struct ip *ip;
411 
412 	ip = mtod(msg->nm_mbuf, struct ip *);
413 	nexthop = msg->nm_hasnexthop ? &msg->nm_nexthop : NULL;
414 	transport_processing_oncpu(msg->nm_mbuf, msg->nm_hlen, ip, nexthop);
415 	lwkt_replymsg(&msg->nm_netmsg.nm_lmsg, 0);
416 }
417 
418 static void
419 ip_input_handler(struct netmsg *msg0)
420 {
421 	struct mbuf *m = ((struct netmsg_packet *)msg0)->nm_packet;
422 
423 	ip_input(m);
424 	/* msg0 was embedded in the mbuf, do not reply! */
425 }
426 
427 /*
428  * IP input routine.  Checksum and byte swap header.  If fragmented
429  * try to reassemble.  Process options.  Pass to next level.
430  */
431 void
432 ip_input(struct mbuf *m)
433 {
434 	struct ip *ip;
435 	struct ipq *fp;
436 	struct in_ifaddr *ia = NULL;
437 	struct in_ifaddr_container *iac;
438 	int i, hlen, checkif;
439 	u_short sum;
440 	struct in_addr pkt_dst;
441 	u_int32_t divert_info = 0;		/* packet divert/tee info */
442 	struct ip_fw_args args;
443 	boolean_t using_srcrt = FALSE;		/* forward (by PFIL_HOOKS) */
444 	boolean_t needredispatch = FALSE;
445 	struct in_addr odst;			/* original dst address(NAT) */
446 	struct m_tag *mtag;
447 #ifdef FAST_IPSEC
448 	struct tdb_ident *tdbi;
449 	struct secpolicy *sp;
450 	int error;
451 #endif
452 
453 	args.eh = NULL;
454 	args.oif = NULL;
455 	args.rule = NULL;
456 	args.next_hop = NULL;
457 
458 	/* Grab info from MT_TAG mbufs prepended to the chain. */
459 	while (m != NULL && m->m_type == MT_TAG) {
460 		switch(m->_m_tag_id) {
461 		case PACKET_TAG_IPFORWARD:
462 			args.next_hop = (struct sockaddr_in *)m->m_hdr.mh_data;
463 			break;
464 		default:
465 			kprintf("ip_input: unrecognised MT_TAG tag %d\n",
466 			    m->_m_tag_id);
467 			break;
468 		}
469 		m = m->m_next;
470 	}
471 	M_ASSERTPKTHDR(m);
472 
473 	/* Extract info from dummynet tag */
474 	mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
475 	if (mtag != NULL) {
476 		args.rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv;
477 
478 		m_tag_delete(m, mtag);
479 		mtag = NULL;
480 	}
481 
482 	if (args.rule != NULL) {	/* dummynet already filtered us */
483 		ip = mtod(m, struct ip *);
484 		hlen = IP_VHL_HL(ip->ip_vhl) << 2;
485 		goto iphack;
486 	}
487 
488 	ipstat.ips_total++;
489 
490 	/* length checks already done in ip_demux() */
491 	KASSERT(m->m_len >= sizeof(ip), ("IP header not in one mbuf"));
492 
493 	ip = mtod(m, struct ip *);
494 
495 	if (IP_VHL_V(ip->ip_vhl) != IPVERSION) {
496 		ipstat.ips_badvers++;
497 		goto bad;
498 	}
499 
500 	hlen = IP_VHL_HL(ip->ip_vhl) << 2;
501 	/* length checks already done in ip_demux() */
502 	KASSERT(hlen >= sizeof(struct ip), ("IP header len too small"));
503 	KASSERT(m->m_len >= hlen, ("packet shorter than IP header length"));
504 
505 	/* 127/8 must not appear on wire - RFC1122 */
506 	if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
507 	    (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
508 		if (!(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK)) {
509 			ipstat.ips_badaddr++;
510 			goto bad;
511 		}
512 	}
513 
514 	if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
515 		sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
516 	} else {
517 		if (hlen == sizeof(struct ip)) {
518 			sum = in_cksum_hdr(ip);
519 		} else {
520 			sum = in_cksum(m, hlen);
521 		}
522 	}
523 	if (sum != 0) {
524 		ipstat.ips_badsum++;
525 		goto bad;
526 	}
527 
528 #ifdef ALTQ
529 	if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) {
530 		/* packet is dropped by traffic conditioner */
531 		return;
532 	}
533 #endif
534 	/*
535 	 * Convert fields to host representation.
536 	 */
537 	ip->ip_len = ntohs(ip->ip_len);
538 	if (ip->ip_len < hlen) {
539 		ipstat.ips_badlen++;
540 		goto bad;
541 	}
542 	ip->ip_off = ntohs(ip->ip_off);
543 
544 	/*
545 	 * Check that the amount of data in the buffers
546 	 * is as at least much as the IP header would have us expect.
547 	 * Trim mbufs if longer than we expect.
548 	 * Drop packet if shorter than we expect.
549 	 */
550 	if (m->m_pkthdr.len < ip->ip_len) {
551 		ipstat.ips_tooshort++;
552 		goto bad;
553 	}
554 	if (m->m_pkthdr.len > ip->ip_len) {
555 		if (m->m_len == m->m_pkthdr.len) {
556 			m->m_len = ip->ip_len;
557 			m->m_pkthdr.len = ip->ip_len;
558 		} else
559 			m_adj(m, ip->ip_len - m->m_pkthdr.len);
560 	}
561 #if defined(IPSEC) && !defined(IPSEC_FILTERGIF)
562 	/*
563 	 * Bypass packet filtering for packets from a tunnel (gif).
564 	 */
565 	if (ipsec_gethist(m, NULL))
566 		goto pass;
567 #endif
568 
569 	/*
570 	 * IpHack's section.
571 	 * Right now when no processing on packet has done
572 	 * and it is still fresh out of network we do our black
573 	 * deals with it.
574 	 * - Firewall: deny/allow/divert
575 	 * - Xlate: translate packet's addr/port (NAT).
576 	 * - Pipe: pass pkt through dummynet.
577 	 * - Wrap: fake packet's addr/port <unimpl.>
578 	 * - Encapsulate: put it in another IP and send out. <unimp.>
579 	 */
580 
581 iphack:
582 
583 	/*
584 	 * Run through list of hooks for input packets.
585 	 *
586 	 * NB: Beware of the destination address changing (e.g.
587 	 *     by NAT rewriting). When this happens, tell
588 	 *     ip_forward to do the right thing.
589 	 */
590 	if (pfil_has_hooks(&inet_pfil_hook)) {
591 		odst = ip->ip_dst;
592 		if (pfil_run_hooks(&inet_pfil_hook, &m,
593 		    m->m_pkthdr.rcvif, PFIL_IN)) {
594 			return;
595 		}
596 		if (m == NULL)			/* consumed by filter */
597 			return;
598 		ip = mtod(m, struct ip *);
599 		using_srcrt = (odst.s_addr != ip->ip_dst.s_addr);
600 	}
601 
602 	if (fw_enable && IPFW_LOADED) {
603 		/*
604 		 * If we've been forwarded from the output side, then
605 		 * skip the firewall a second time
606 		 */
607 		if (args.next_hop != NULL)
608 			goto ours;
609 
610 		args.m = m;
611 		i = ip_fw_chk_ptr(&args);
612 		m = args.m;
613 
614 		if ((i & IP_FW_PORT_DENY_FLAG) || m == NULL) {	/* drop */
615 			if (m != NULL)
616 				m_freem(m);
617 			return;
618 		}
619 		ip = mtod(m, struct ip *);	/* just in case m changed */
620 		if (i == 0 && args.next_hop == NULL)	/* common case */
621 			goto pass;
622 		if (i & IP_FW_PORT_DYNT_FLAG) {
623 			/* Send packet to the appropriate pipe */
624 			ip_fw_dn_io_ptr(m, i&0xffff, DN_TO_IP_IN, &args);
625 			return;
626 		}
627 #ifdef IPDIVERT
628 		if (i != 0 && !(i & IP_FW_PORT_DYNT_FLAG)) {
629 			/* Divert or tee packet */
630 			divert_info = i;
631 			goto ours;
632 		}
633 #endif
634 		if (i == 0 && args.next_hop != NULL)
635 			goto pass;
636 		/*
637 		 * if we get here, the packet must be dropped
638 		 */
639 		m_freem(m);
640 		return;
641 	}
642 pass:
643 
644 	/*
645 	 * Process options and, if not destined for us,
646 	 * ship it on.  ip_dooptions returns 1 when an
647 	 * error was detected (causing an icmp message
648 	 * to be sent and the original packet to be freed).
649 	 */
650 	ip_nhops = 0;		/* for source routed packets */
651 	if (hlen > sizeof(struct ip) && ip_dooptions(m, 0, args.next_hop))
652 		return;
653 
654 	/* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
655 	 * matter if it is destined to another node, or whether it is
656 	 * a multicast one, RSVP wants it! and prevents it from being forwarded
657 	 * anywhere else. Also checks if the rsvp daemon is running before
658 	 * grabbing the packet.
659 	 */
660 	if (rsvp_on && ip->ip_p == IPPROTO_RSVP)
661 		goto ours;
662 
663 	/*
664 	 * Check our list of addresses, to see if the packet is for us.
665 	 * If we don't have any addresses, assume any unicast packet
666 	 * we receive might be for us (and let the upper layers deal
667 	 * with it).
668 	 */
669 	if (TAILQ_EMPTY(&in_ifaddrheads[mycpuid]) &&
670 	    !(m->m_flags & (M_MCAST | M_BCAST)))
671 		goto ours;
672 
673 	/*
674 	 * Cache the destination address of the packet; this may be
675 	 * changed by use of 'ipfw fwd'.
676 	 */
677 	pkt_dst = args.next_hop ? args.next_hop->sin_addr : ip->ip_dst;
678 
679 	/*
680 	 * Enable a consistency check between the destination address
681 	 * and the arrival interface for a unicast packet (the RFC 1122
682 	 * strong ES model) if IP forwarding is disabled and the packet
683 	 * is not locally generated and the packet is not subject to
684 	 * 'ipfw fwd'.
685 	 *
686 	 * XXX - Checking also should be disabled if the destination
687 	 * address is ipnat'ed to a different interface.
688 	 *
689 	 * XXX - Checking is incompatible with IP aliases added
690 	 * to the loopback interface instead of the interface where
691 	 * the packets are received.
692 	 */
693 	checkif = ip_checkinterface &&
694 		  !ipforwarding &&
695 		  m->m_pkthdr.rcvif != NULL &&
696 		  !(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) &&
697 		  (args.next_hop == NULL);
698 
699 	/*
700 	 * Check for exact addresses in the hash bucket.
701 	 */
702 	LIST_FOREACH(iac, INADDR_HASH(pkt_dst.s_addr), ia_hash) {
703 		ia = iac->ia;
704 
705 		/*
706 		 * If the address matches, verify that the packet
707 		 * arrived via the correct interface if checking is
708 		 * enabled.
709 		 */
710 		if (IA_SIN(ia)->sin_addr.s_addr == pkt_dst.s_addr &&
711 		    (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif))
712 			goto ours;
713 	}
714 	ia = NULL;
715 
716 	/*
717 	 * Check for broadcast addresses.
718 	 *
719 	 * Only accept broadcast packets that arrive via the matching
720 	 * interface.  Reception of forwarded directed broadcasts would
721 	 * be handled via ip_forward() and ether_output() with the loopback
722 	 * into the stack for SIMPLEX interfaces handled by ether_output().
723 	 */
724 	if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
725 		struct ifaddr_container *ifac;
726 
727 		TAILQ_FOREACH(ifac, &m->m_pkthdr.rcvif->if_addrheads[mycpuid],
728 			      ifa_link) {
729 			struct ifaddr *ifa = ifac->ifa;
730 
731 			if (ifa->ifa_addr == NULL) /* shutdown/startup race */
732 				continue;
733 			if (ifa->ifa_addr->sa_family != AF_INET)
734 				continue;
735 			ia = ifatoia(ifa);
736 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
737 								pkt_dst.s_addr)
738 				goto ours;
739 			if (ia->ia_netbroadcast.s_addr == pkt_dst.s_addr)
740 				goto ours;
741 #ifdef BOOTP_COMPAT
742 			if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY)
743 				goto ours;
744 #endif
745 		}
746 	}
747 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
748 		struct in_multi *inm;
749 
750 		if (ip_mrouter != NULL) {
751 			/*
752 			 * If we are acting as a multicast router, all
753 			 * incoming multicast packets are passed to the
754 			 * kernel-level multicast forwarding function.
755 			 * The packet is returned (relatively) intact; if
756 			 * ip_mforward() returns a non-zero value, the packet
757 			 * must be discarded, else it may be accepted below.
758 			 */
759 			if (ip_mforward != NULL &&
760 			    ip_mforward(ip, m->m_pkthdr.rcvif, m, NULL) != 0) {
761 				ipstat.ips_cantforward++;
762 				m_freem(m);
763 				return;
764 			}
765 
766 			/*
767 			 * The process-level routing daemon needs to receive
768 			 * all multicast IGMP packets, whether or not this
769 			 * host belongs to their destination groups.
770 			 */
771 			if (ip->ip_p == IPPROTO_IGMP)
772 				goto ours;
773 			ipstat.ips_forward++;
774 		}
775 		/*
776 		 * See if we belong to the destination multicast group on the
777 		 * arrival interface.
778 		 */
779 		IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
780 		if (inm == NULL) {
781 			ipstat.ips_notmember++;
782 			m_freem(m);
783 			return;
784 		}
785 		goto ours;
786 	}
787 	if (ip->ip_dst.s_addr == INADDR_BROADCAST)
788 		goto ours;
789 	if (ip->ip_dst.s_addr == INADDR_ANY)
790 		goto ours;
791 
792 	/*
793 	 * FAITH(Firewall Aided Internet Translator)
794 	 */
795 	if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
796 		if (ip_keepfaith) {
797 			if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP)
798 				goto ours;
799 		}
800 		m_freem(m);
801 		return;
802 	}
803 
804 	/*
805 	 * Not for us; forward if possible and desirable.
806 	 */
807 	if (!ipforwarding) {
808 		ipstat.ips_cantforward++;
809 		m_freem(m);
810 	} else {
811 #ifdef IPSEC
812 		/*
813 		 * Enforce inbound IPsec SPD.
814 		 */
815 		if (ipsec4_in_reject(m, NULL)) {
816 			ipsecstat.in_polvio++;
817 			goto bad;
818 		}
819 #endif
820 #ifdef FAST_IPSEC
821 		mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
822 		crit_enter();
823 		if (mtag != NULL) {
824 			tdbi = (struct tdb_ident *)m_tag_data(mtag);
825 			sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
826 		} else {
827 			sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
828 						   IP_FORWARDING, &error);
829 		}
830 		if (sp == NULL) {	/* NB: can happen if error */
831 			crit_exit();
832 			/*XXX error stat???*/
833 			DPRINTF(("ip_input: no SP for forwarding\n"));	/*XXX*/
834 			goto bad;
835 		}
836 
837 		/*
838 		 * Check security policy against packet attributes.
839 		 */
840 		error = ipsec_in_reject(sp, m);
841 		KEY_FREESP(&sp);
842 		crit_exit();
843 		if (error) {
844 			ipstat.ips_cantforward++;
845 			goto bad;
846 		}
847 #endif
848 		ip_forward(m, using_srcrt, args.next_hop);
849 	}
850 	return;
851 
852 ours:
853 
854 	/*
855 	 * IPSTEALTH: Process non-routing options only
856 	 * if the packet is destined for us.
857 	 */
858 	if (ipstealth &&
859 	    hlen > sizeof(struct ip) &&
860 	    ip_dooptions(m, 1, args.next_hop))
861 		return;
862 
863 	/* Count the packet in the ip address stats */
864 	if (ia != NULL) {
865 		ia->ia_ifa.if_ipackets++;
866 		ia->ia_ifa.if_ibytes += m->m_pkthdr.len;
867 	}
868 
869 	/*
870 	 * If offset or IP_MF are set, must reassemble.
871 	 * Otherwise, nothing need be done.
872 	 * (We could look in the reassembly queue to see
873 	 * if the packet was previously fragmented,
874 	 * but it's not worth the time; just let them time out.)
875 	 */
876 	if (ip->ip_off & (IP_MF | IP_OFFMASK)) {
877 
878 		/* If maxnipq is 0, never accept fragments. */
879 		if (maxnipq == 0) {
880 			ipstat.ips_fragments++;
881 			ipstat.ips_fragdropped++;
882 			goto bad;
883 		}
884 
885 		sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
886 		/*
887 		 * Look for queue of fragments
888 		 * of this datagram.
889 		 */
890 		for (fp = ipq[sum].next; fp != &ipq[sum]; fp = fp->next)
891 			if (ip->ip_id == fp->ipq_id &&
892 			    ip->ip_src.s_addr == fp->ipq_src.s_addr &&
893 			    ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
894 			    ip->ip_p == fp->ipq_p)
895 				goto found;
896 
897 		fp = NULL;
898 
899 		/*
900 		 * Enforce upper bound on number of fragmented packets
901 		 * for which we attempt reassembly;
902 		 * If maxnipq is -1, accept all fragments without limitation.
903 		 */
904 		if ((nipq > maxnipq) && (maxnipq > 0)) {
905 			/*
906 			 * drop something from the tail of the current queue
907 			 * before proceeding further
908 			 */
909 			if (ipq[sum].prev == &ipq[sum]) {   /* gak */
910 				for (i = 0; i < IPREASS_NHASH; i++) {
911 					if (ipq[i].prev != &ipq[i]) {
912 						ipstat.ips_fragtimeout +=
913 						    ipq[i].prev->ipq_nfrags;
914 						ip_freef(ipq[i].prev);
915 						break;
916 					}
917 				}
918 			} else {
919 				ipstat.ips_fragtimeout +=
920 				    ipq[sum].prev->ipq_nfrags;
921 				ip_freef(ipq[sum].prev);
922 			}
923 		}
924 found:
925 		/*
926 		 * Adjust ip_len to not reflect header,
927 		 * convert offset of this to bytes.
928 		 */
929 		ip->ip_len -= hlen;
930 		if (ip->ip_off & IP_MF) {
931 			/*
932 			 * Make sure that fragments have a data length
933 			 * that's a non-zero multiple of 8 bytes.
934 			 */
935 			if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
936 				ipstat.ips_toosmall++; /* XXX */
937 				goto bad;
938 			}
939 			m->m_flags |= M_FRAG;
940 		} else
941 			m->m_flags &= ~M_FRAG;
942 		ip->ip_off <<= 3;
943 
944 		/*
945 		 * Attempt reassembly; if it succeeds, proceed.
946 		 * ip_reass() will return a different mbuf, and update
947 		 * the divert info in divert_info.
948 		 */
949 		ipstat.ips_fragments++;
950 		m->m_pkthdr.header = ip;
951 		m = ip_reass(m, fp, &ipq[sum], &divert_info);
952 		if (m == NULL)
953 			return;
954 		ipstat.ips_reassembled++;
955 		needredispatch = TRUE;
956 		ip = mtod(m, struct ip *);
957 		/* Get the header length of the reassembled packet */
958 		hlen = IP_VHL_HL(ip->ip_vhl) << 2;
959 #ifdef IPDIVERT
960 		/* Restore original checksum before diverting packet */
961 		if (divert_info != 0) {
962 			ip->ip_len += hlen;
963 			ip->ip_len = htons(ip->ip_len);
964 			ip->ip_off = htons(ip->ip_off);
965 			ip->ip_sum = 0;
966 			if (hlen == sizeof(struct ip))
967 				ip->ip_sum = in_cksum_hdr(ip);
968 			else
969 				ip->ip_sum = in_cksum(m, hlen);
970 			ip->ip_off = ntohs(ip->ip_off);
971 			ip->ip_len = ntohs(ip->ip_len);
972 			ip->ip_len -= hlen;
973 		}
974 #endif
975 	} else {
976 		ip->ip_len -= hlen;
977 	}
978 
979 #ifdef IPDIVERT
980 	/*
981 	 * Divert or tee packet to the divert protocol if required.
982 	 */
983 	if (divert_info != 0) {
984 		struct mbuf *clone = NULL;
985 
986 		/* Clone packet if we're doing a 'tee' */
987 		if ((divert_info & IP_FW_PORT_TEE_FLAG) != 0)
988 			clone = m_dup(m, MB_DONTWAIT);
989 
990 		/* Restore packet header fields to original values */
991 		ip->ip_len += hlen;
992 		ip->ip_len = htons(ip->ip_len);
993 		ip->ip_off = htons(ip->ip_off);
994 
995 		/* Deliver packet to divert input routine */
996 		divert_packet(m, 1, divert_info & 0xffff);
997 		ipstat.ips_delivered++;
998 
999 		/* If 'tee', continue with original packet */
1000 		if (clone == NULL)
1001 			return;
1002 		m = clone;
1003 		ip = mtod(m, struct ip *);
1004 		ip->ip_len += hlen;
1005 		/*
1006 		 * Jump backwards to complete processing of the
1007 		 * packet. But first clear divert_info to avoid
1008 		 * entering this block again.
1009 		 * We do not need to clear args.divert_rule
1010 		 * or args.next_hop as they will not be used.
1011 		 *
1012 		 * XXX Better safe than sorry, remove the DIVERT tag.
1013 		 */
1014 		mtag = m_tag_find(m, PACKET_TAG_IPFW_DIVERT, NULL);
1015 		if (mtag != NULL)
1016 			m_tag_delete(m, mtag);
1017 
1018 		divert_info = 0;
1019 		goto pass;
1020 	}
1021 #endif
1022 
1023 #ifdef IPSEC
1024 	/*
1025 	 * enforce IPsec policy checking if we are seeing last header.
1026 	 * note that we do not visit this with protocols with pcb layer
1027 	 * code - like udp/tcp/raw ip.
1028 	 */
1029 	if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) &&
1030 	    ipsec4_in_reject(m, NULL)) {
1031 		ipsecstat.in_polvio++;
1032 		goto bad;
1033 	}
1034 #endif
1035 #if FAST_IPSEC
1036 	/*
1037 	 * enforce IPsec policy checking if we are seeing last header.
1038 	 * note that we do not visit this with protocols with pcb layer
1039 	 * code - like udp/tcp/raw ip.
1040 	 */
1041 	if (inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) {
1042 		/*
1043 		 * Check if the packet has already had IPsec processing
1044 		 * done.  If so, then just pass it along.  This tag gets
1045 		 * set during AH, ESP, etc. input handling, before the
1046 		 * packet is returned to the ip input queue for delivery.
1047 		 */
1048 		mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
1049 		crit_enter();
1050 		if (mtag != NULL) {
1051 			tdbi = (struct tdb_ident *)m_tag_data(mtag);
1052 			sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
1053 		} else {
1054 			sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
1055 						   IP_FORWARDING, &error);
1056 		}
1057 		if (sp != NULL) {
1058 			/*
1059 			 * Check security policy against packet attributes.
1060 			 */
1061 			error = ipsec_in_reject(sp, m);
1062 			KEY_FREESP(&sp);
1063 		} else {
1064 			/* XXX error stat??? */
1065 			error = EINVAL;
1066 DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/
1067 			goto bad;
1068 		}
1069 		crit_exit();
1070 		if (error)
1071 			goto bad;
1072 	}
1073 #endif /* FAST_IPSEC */
1074 
1075 	ipstat.ips_delivered++;
1076 	if (needredispatch) {
1077 		struct netmsg_transport_packet *msg;
1078 		lwkt_port_t port;
1079 
1080 		ip->ip_off = htons(ip->ip_off);
1081 		ip->ip_len = htons(ip->ip_len);
1082 		port = ip_mport_in(&m);
1083 		if (port == NULL)
1084 			return;
1085 
1086 		msg = kmalloc(sizeof(struct netmsg_transport_packet), M_LWKTMSG,
1087 			     M_INTWAIT | M_NULLOK);
1088 		if (msg == NULL)
1089 			goto bad;
1090 
1091 		netmsg_init(&msg->nm_netmsg, &netisr_afree_rport, 0,
1092 			    transport_processing_handler);
1093 		msg->nm_hlen = hlen;
1094 		msg->nm_hasnexthop = (args.next_hop != NULL);
1095 		if (msg->nm_hasnexthop)
1096 			msg->nm_nexthop = *args.next_hop;  /* structure copy */
1097 
1098 		msg->nm_mbuf = m;
1099 		ip = mtod(m, struct ip *);
1100 		ip->ip_len = ntohs(ip->ip_len);
1101 		ip->ip_off = ntohs(ip->ip_off);
1102 		lwkt_sendmsg(port, &msg->nm_netmsg.nm_lmsg);
1103 	} else {
1104 		transport_processing_oncpu(m, hlen, ip, args.next_hop);
1105 	}
1106 	return;
1107 
1108 bad:
1109 	m_freem(m);
1110 }
1111 
1112 /*
1113  * Take incoming datagram fragment and try to reassemble it into
1114  * whole datagram.  If a chain for reassembly of this datagram already
1115  * exists, then it is given as fp; otherwise have to make a chain.
1116  *
1117  * When IPDIVERT enabled, keep additional state with each packet that
1118  * tells us if we need to divert or tee the packet we're building.
1119  * In particular, *divinfo includes the port and TEE flag.
1120  */
1121 
1122 static struct mbuf *
1123 ip_reass(struct mbuf *m, struct ipq *fp, struct ipq *where,
1124 	 u_int32_t *divinfo)
1125 {
1126 	struct ip *ip = mtod(m, struct ip *);
1127 	struct mbuf *p = NULL, *q, *nq;
1128 	struct mbuf *n;
1129 	int hlen = IP_VHL_HL(ip->ip_vhl) << 2;
1130 	int i, next;
1131 #ifdef IPDIVERT
1132 	struct m_tag *mtag;
1133 #endif
1134 
1135 	/*
1136 	 * If the hardware has not done csum over this fragment
1137 	 * then csum_data is not valid at all.
1138 	 */
1139 	if ((m->m_pkthdr.csum_flags & (CSUM_FRAG_NOT_CHECKED | CSUM_DATA_VALID))
1140 	    == (CSUM_FRAG_NOT_CHECKED | CSUM_DATA_VALID)) {
1141 		m->m_pkthdr.csum_data = 0;
1142 		m->m_pkthdr.csum_flags &= ~(CSUM_DATA_VALID | CSUM_PSEUDO_HDR);
1143 	}
1144 
1145 	/*
1146 	 * Presence of header sizes in mbufs
1147 	 * would confuse code below.
1148 	 */
1149 	m->m_data += hlen;
1150 	m->m_len -= hlen;
1151 
1152 	/*
1153 	 * If first fragment to arrive, create a reassembly queue.
1154 	 */
1155 	if (fp == NULL) {
1156 		if ((fp = mpipe_alloc_nowait(&ipq_mpipe)) == NULL)
1157 			goto dropfrag;
1158 		insque(fp, where);
1159 		nipq++;
1160 		fp->ipq_nfrags = 1;
1161 		fp->ipq_ttl = IPFRAGTTL;
1162 		fp->ipq_p = ip->ip_p;
1163 		fp->ipq_id = ip->ip_id;
1164 		fp->ipq_src = ip->ip_src;
1165 		fp->ipq_dst = ip->ip_dst;
1166 		fp->ipq_frags = m;
1167 		m->m_nextpkt = NULL;
1168 #ifdef IPDIVERT
1169 		fp->ipq_div_info = 0;
1170 #endif
1171 		goto inserted;
1172 	} else {
1173 		fp->ipq_nfrags++;
1174 	}
1175 
1176 #define	GETIP(m)	((struct ip*)((m)->m_pkthdr.header))
1177 
1178 	/*
1179 	 * Find a segment which begins after this one does.
1180 	 */
1181 	for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt)
1182 		if (GETIP(q)->ip_off > ip->ip_off)
1183 			break;
1184 
1185 	/*
1186 	 * If there is a preceding segment, it may provide some of
1187 	 * our data already.  If so, drop the data from the incoming
1188 	 * segment.  If it provides all of our data, drop us, otherwise
1189 	 * stick new segment in the proper place.
1190 	 *
1191 	 * If some of the data is dropped from the the preceding
1192 	 * segment, then it's checksum is invalidated.
1193 	 */
1194 	if (p) {
1195 		i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off;
1196 		if (i > 0) {
1197 			if (i >= ip->ip_len)
1198 				goto dropfrag;
1199 			m_adj(m, i);
1200 			m->m_pkthdr.csum_flags = 0;
1201 			ip->ip_off += i;
1202 			ip->ip_len -= i;
1203 		}
1204 		m->m_nextpkt = p->m_nextpkt;
1205 		p->m_nextpkt = m;
1206 	} else {
1207 		m->m_nextpkt = fp->ipq_frags;
1208 		fp->ipq_frags = m;
1209 	}
1210 
1211 	/*
1212 	 * While we overlap succeeding segments trim them or,
1213 	 * if they are completely covered, dequeue them.
1214 	 */
1215 	for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off;
1216 	     q = nq) {
1217 		i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off;
1218 		if (i < GETIP(q)->ip_len) {
1219 			GETIP(q)->ip_len -= i;
1220 			GETIP(q)->ip_off += i;
1221 			m_adj(q, i);
1222 			q->m_pkthdr.csum_flags = 0;
1223 			break;
1224 		}
1225 		nq = q->m_nextpkt;
1226 		m->m_nextpkt = nq;
1227 		ipstat.ips_fragdropped++;
1228 		fp->ipq_nfrags--;
1229 		q->m_nextpkt = NULL;
1230 		m_freem(q);
1231 	}
1232 
1233 inserted:
1234 
1235 #ifdef IPDIVERT
1236 	/*
1237 	 * Transfer firewall instructions to the fragment structure.
1238 	 * Only trust info in the fragment at offset 0.
1239 	 */
1240 	if (ip->ip_off == 0) {
1241 		fp->ipq_div_info = *divinfo;
1242 	} else {
1243 		mtag = m_tag_find(m, PACKET_TAG_IPFW_DIVERT, NULL);
1244 		if (mtag != NULL)
1245 			m_tag_delete(m, mtag);
1246 	}
1247 	*divinfo = 0;
1248 #endif
1249 
1250 	/*
1251 	 * Check for complete reassembly and perform frag per packet
1252 	 * limiting.
1253 	 *
1254 	 * Frag limiting is performed here so that the nth frag has
1255 	 * a chance to complete the packet before we drop the packet.
1256 	 * As a result, n+1 frags are actually allowed per packet, but
1257 	 * only n will ever be stored. (n = maxfragsperpacket.)
1258 	 *
1259 	 */
1260 	next = 0;
1261 	for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
1262 		if (GETIP(q)->ip_off != next) {
1263 			if (fp->ipq_nfrags > maxfragsperpacket) {
1264 				ipstat.ips_fragdropped += fp->ipq_nfrags;
1265 				ip_freef(fp);
1266 			}
1267 			return (NULL);
1268 		}
1269 		next += GETIP(q)->ip_len;
1270 	}
1271 	/* Make sure the last packet didn't have the IP_MF flag */
1272 	if (p->m_flags & M_FRAG) {
1273 		if (fp->ipq_nfrags > maxfragsperpacket) {
1274 			ipstat.ips_fragdropped += fp->ipq_nfrags;
1275 			ip_freef(fp);
1276 		}
1277 		return (NULL);
1278 	}
1279 
1280 	/*
1281 	 * Reassembly is complete.  Make sure the packet is a sane size.
1282 	 */
1283 	q = fp->ipq_frags;
1284 	ip = GETIP(q);
1285 	if (next + (IP_VHL_HL(ip->ip_vhl) << 2) > IP_MAXPACKET) {
1286 		ipstat.ips_toolong++;
1287 		ipstat.ips_fragdropped += fp->ipq_nfrags;
1288 		ip_freef(fp);
1289 		return (NULL);
1290 	}
1291 
1292 	/*
1293 	 * Concatenate fragments.
1294 	 */
1295 	m = q;
1296 	n = m->m_next;
1297 	m->m_next = NULL;
1298 	m_cat(m, n);
1299 	nq = q->m_nextpkt;
1300 	q->m_nextpkt = NULL;
1301 	for (q = nq; q != NULL; q = nq) {
1302 		nq = q->m_nextpkt;
1303 		q->m_nextpkt = NULL;
1304 		m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
1305 		m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
1306 		m_cat(m, q);
1307 	}
1308 
1309 	/*
1310 	 * Clean up the 1's complement checksum.  Carry over 16 bits must
1311 	 * be added back.  This assumes no more then 65535 packet fragments
1312 	 * were reassembled.  A second carry can also occur (but not a third).
1313 	 */
1314 	m->m_pkthdr.csum_data = (m->m_pkthdr.csum_data & 0xffff) +
1315 				(m->m_pkthdr.csum_data >> 16);
1316 	if (m->m_pkthdr.csum_data > 0xFFFF)
1317 		m->m_pkthdr.csum_data -= 0xFFFF;
1318 
1319 
1320 #ifdef IPDIVERT
1321 	/*
1322 	 * Extract firewall instructions from the fragment structure.
1323 	 */
1324 	*divinfo = fp->ipq_div_info;
1325 #endif
1326 
1327 	/*
1328 	 * Create header for new ip packet by
1329 	 * modifying header of first packet;
1330 	 * dequeue and discard fragment reassembly header.
1331 	 * Make header visible.
1332 	 */
1333 	ip->ip_len = next;
1334 	ip->ip_src = fp->ipq_src;
1335 	ip->ip_dst = fp->ipq_dst;
1336 	remque(fp);
1337 	nipq--;
1338 	mpipe_free(&ipq_mpipe, fp);
1339 	m->m_len += (IP_VHL_HL(ip->ip_vhl) << 2);
1340 	m->m_data -= (IP_VHL_HL(ip->ip_vhl) << 2);
1341 	/* some debugging cruft by sklower, below, will go away soon */
1342 	if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */
1343 		int plen = 0;
1344 
1345 		for (n = m; n; n = n->m_next)
1346 			plen += n->m_len;
1347 		m->m_pkthdr.len = plen;
1348 	}
1349 	return (m);
1350 
1351 dropfrag:
1352 #ifdef IPDIVERT
1353 	*divinfo = 0;
1354 #endif
1355 	ipstat.ips_fragdropped++;
1356 	if (fp != NULL)
1357 		fp->ipq_nfrags--;
1358 	m_freem(m);
1359 	return (NULL);
1360 
1361 #undef GETIP
1362 }
1363 
1364 /*
1365  * Free a fragment reassembly header and all
1366  * associated datagrams.
1367  */
1368 static void
1369 ip_freef(struct ipq *fp)
1370 {
1371 	struct mbuf *q;
1372 
1373 	while (fp->ipq_frags) {
1374 		q = fp->ipq_frags;
1375 		fp->ipq_frags = q->m_nextpkt;
1376 		q->m_nextpkt = NULL;
1377 		m_freem(q);
1378 	}
1379 	remque(fp);
1380 	mpipe_free(&ipq_mpipe, fp);
1381 	nipq--;
1382 }
1383 
1384 /*
1385  * IP timer processing;
1386  * if a timer expires on a reassembly
1387  * queue, discard it.
1388  */
1389 void
1390 ip_slowtimo(void)
1391 {
1392 	struct ipq *fp;
1393 	int i;
1394 
1395 	crit_enter();
1396 	for (i = 0; i < IPREASS_NHASH; i++) {
1397 		fp = ipq[i].next;
1398 		if (fp == NULL)
1399 			continue;
1400 		while (fp != &ipq[i]) {
1401 			--fp->ipq_ttl;
1402 			fp = fp->next;
1403 			if (fp->prev->ipq_ttl == 0) {
1404 				ipstat.ips_fragtimeout += fp->prev->ipq_nfrags;
1405 				ip_freef(fp->prev);
1406 			}
1407 		}
1408 	}
1409 	/*
1410 	 * If we are over the maximum number of fragments
1411 	 * (due to the limit being lowered), drain off
1412 	 * enough to get down to the new limit.
1413 	 */
1414 	if (maxnipq >= 0 && nipq > maxnipq) {
1415 		for (i = 0; i < IPREASS_NHASH; i++) {
1416 			while (nipq > maxnipq &&
1417 				(ipq[i].next != &ipq[i])) {
1418 				ipstat.ips_fragdropped +=
1419 				    ipq[i].next->ipq_nfrags;
1420 				ip_freef(ipq[i].next);
1421 			}
1422 		}
1423 	}
1424 	ipflow_slowtimo();
1425 	crit_exit();
1426 }
1427 
1428 /*
1429  * Drain off all datagram fragments.
1430  */
1431 void
1432 ip_drain(void)
1433 {
1434 	int i;
1435 
1436 	for (i = 0; i < IPREASS_NHASH; i++) {
1437 		while (ipq[i].next != &ipq[i]) {
1438 			ipstat.ips_fragdropped += ipq[i].next->ipq_nfrags;
1439 			ip_freef(ipq[i].next);
1440 		}
1441 	}
1442 	in_rtqdrain();
1443 }
1444 
1445 /*
1446  * Do option processing on a datagram,
1447  * possibly discarding it if bad options are encountered,
1448  * or forwarding it if source-routed.
1449  * The pass argument is used when operating in the IPSTEALTH
1450  * mode to tell what options to process:
1451  * [LS]SRR (pass 0) or the others (pass 1).
1452  * The reason for as many as two passes is that when doing IPSTEALTH,
1453  * non-routing options should be processed only if the packet is for us.
1454  * Returns 1 if packet has been forwarded/freed,
1455  * 0 if the packet should be processed further.
1456  */
1457 static int
1458 ip_dooptions(struct mbuf *m, int pass, struct sockaddr_in *next_hop)
1459 {
1460 	struct sockaddr_in ipaddr = { sizeof ipaddr, AF_INET };
1461 	struct ip *ip = mtod(m, struct ip *);
1462 	u_char *cp;
1463 	struct in_ifaddr *ia;
1464 	int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB;
1465 	boolean_t forward = FALSE;
1466 	struct in_addr *sin, dst;
1467 	n_time ntime;
1468 
1469 	dst = ip->ip_dst;
1470 	cp = (u_char *)(ip + 1);
1471 	cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
1472 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1473 		opt = cp[IPOPT_OPTVAL];
1474 		if (opt == IPOPT_EOL)
1475 			break;
1476 		if (opt == IPOPT_NOP)
1477 			optlen = 1;
1478 		else {
1479 			if (cnt < IPOPT_OLEN + sizeof(*cp)) {
1480 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1481 				goto bad;
1482 			}
1483 			optlen = cp[IPOPT_OLEN];
1484 			if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
1485 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1486 				goto bad;
1487 			}
1488 		}
1489 		switch (opt) {
1490 
1491 		default:
1492 			break;
1493 
1494 		/*
1495 		 * Source routing with record.
1496 		 * Find interface with current destination address.
1497 		 * If none on this machine then drop if strictly routed,
1498 		 * or do nothing if loosely routed.
1499 		 * Record interface address and bring up next address
1500 		 * component.  If strictly routed make sure next
1501 		 * address is on directly accessible net.
1502 		 */
1503 		case IPOPT_LSRR:
1504 		case IPOPT_SSRR:
1505 			if (ipstealth && pass > 0)
1506 				break;
1507 			if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1508 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1509 				goto bad;
1510 			}
1511 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1512 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1513 				goto bad;
1514 			}
1515 			ipaddr.sin_addr = ip->ip_dst;
1516 			ia = (struct in_ifaddr *)
1517 				ifa_ifwithaddr((struct sockaddr *)&ipaddr);
1518 			if (ia == NULL) {
1519 				if (opt == IPOPT_SSRR) {
1520 					type = ICMP_UNREACH;
1521 					code = ICMP_UNREACH_SRCFAIL;
1522 					goto bad;
1523 				}
1524 				if (!ip_dosourceroute)
1525 					goto nosourcerouting;
1526 				/*
1527 				 * Loose routing, and not at next destination
1528 				 * yet; nothing to do except forward.
1529 				 */
1530 				break;
1531 			}
1532 			off--;			/* 0 origin */
1533 			if (off > optlen - (int)sizeof(struct in_addr)) {
1534 				/*
1535 				 * End of source route.  Should be for us.
1536 				 */
1537 				if (!ip_acceptsourceroute)
1538 					goto nosourcerouting;
1539 				save_rte(cp, ip->ip_src);
1540 				break;
1541 			}
1542 			if (ipstealth)
1543 				goto dropit;
1544 			if (!ip_dosourceroute) {
1545 				if (ipforwarding) {
1546 					char buf[sizeof "aaa.bbb.ccc.ddd"];
1547 
1548 					/*
1549 					 * Acting as a router, so generate ICMP
1550 					 */
1551 nosourcerouting:
1552 					strcpy(buf, inet_ntoa(ip->ip_dst));
1553 					log(LOG_WARNING,
1554 					    "attempted source route from %s to %s\n",
1555 					    inet_ntoa(ip->ip_src), buf);
1556 					type = ICMP_UNREACH;
1557 					code = ICMP_UNREACH_SRCFAIL;
1558 					goto bad;
1559 				} else {
1560 					/*
1561 					 * Not acting as a router,
1562 					 * so silently drop.
1563 					 */
1564 dropit:
1565 					ipstat.ips_cantforward++;
1566 					m_freem(m);
1567 					return (1);
1568 				}
1569 			}
1570 
1571 			/*
1572 			 * locate outgoing interface
1573 			 */
1574 			memcpy(&ipaddr.sin_addr, cp + off,
1575 			    sizeof ipaddr.sin_addr);
1576 
1577 			if (opt == IPOPT_SSRR) {
1578 #define	INA	struct in_ifaddr *
1579 #define	SA	struct sockaddr *
1580 				if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr))
1581 									== NULL)
1582 					ia = (INA)ifa_ifwithnet((SA)&ipaddr);
1583 			} else
1584 				ia = ip_rtaddr(ipaddr.sin_addr,
1585 					       &ipforward_rt[mycpuid]);
1586 			if (ia == NULL) {
1587 				type = ICMP_UNREACH;
1588 				code = ICMP_UNREACH_SRCFAIL;
1589 				goto bad;
1590 			}
1591 			ip->ip_dst = ipaddr.sin_addr;
1592 			memcpy(cp + off, &IA_SIN(ia)->sin_addr,
1593 			    sizeof(struct in_addr));
1594 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1595 			/*
1596 			 * Let ip_intr's mcast routing check handle mcast pkts
1597 			 */
1598 			forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr));
1599 			break;
1600 
1601 		case IPOPT_RR:
1602 			if (ipstealth && pass == 0)
1603 				break;
1604 			if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1605 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1606 				goto bad;
1607 			}
1608 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1609 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1610 				goto bad;
1611 			}
1612 			/*
1613 			 * If no space remains, ignore.
1614 			 */
1615 			off--;			/* 0 origin */
1616 			if (off > optlen - (int)sizeof(struct in_addr))
1617 				break;
1618 			memcpy(&ipaddr.sin_addr, &ip->ip_dst,
1619 			    sizeof ipaddr.sin_addr);
1620 			/*
1621 			 * locate outgoing interface; if we're the destination,
1622 			 * use the incoming interface (should be same).
1623 			 */
1624 			if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == NULL &&
1625 			    (ia = ip_rtaddr(ipaddr.sin_addr,
1626 			    		    &ipforward_rt[mycpuid]))
1627 								     == NULL) {
1628 				type = ICMP_UNREACH;
1629 				code = ICMP_UNREACH_HOST;
1630 				goto bad;
1631 			}
1632 			memcpy(cp + off, &IA_SIN(ia)->sin_addr,
1633 			    sizeof(struct in_addr));
1634 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1635 			break;
1636 
1637 		case IPOPT_TS:
1638 			if (ipstealth && pass == 0)
1639 				break;
1640 			code = cp - (u_char *)ip;
1641 			if (optlen < 4 || optlen > 40) {
1642 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1643 				goto bad;
1644 			}
1645 			if ((off = cp[IPOPT_OFFSET]) < 5) {
1646 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1647 				goto bad;
1648 			}
1649 			if (off > optlen - (int)sizeof(int32_t)) {
1650 				cp[IPOPT_OFFSET + 1] += (1 << 4);
1651 				if ((cp[IPOPT_OFFSET + 1] & 0xf0) == 0) {
1652 					code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1653 					goto bad;
1654 				}
1655 				break;
1656 			}
1657 			off--;				/* 0 origin */
1658 			sin = (struct in_addr *)(cp + off);
1659 			switch (cp[IPOPT_OFFSET + 1] & 0x0f) {
1660 
1661 			case IPOPT_TS_TSONLY:
1662 				break;
1663 
1664 			case IPOPT_TS_TSANDADDR:
1665 				if (off + sizeof(n_time) +
1666 				    sizeof(struct in_addr) > optlen) {
1667 					code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1668 					goto bad;
1669 				}
1670 				ipaddr.sin_addr = dst;
1671 				ia = (INA)ifaof_ifpforaddr((SA)&ipaddr,
1672 							    m->m_pkthdr.rcvif);
1673 				if (ia == NULL)
1674 					continue;
1675 				memcpy(sin, &IA_SIN(ia)->sin_addr,
1676 				    sizeof(struct in_addr));
1677 				cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1678 				off += sizeof(struct in_addr);
1679 				break;
1680 
1681 			case IPOPT_TS_PRESPEC:
1682 				if (off + sizeof(n_time) +
1683 				    sizeof(struct in_addr) > optlen) {
1684 					code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1685 					goto bad;
1686 				}
1687 				memcpy(&ipaddr.sin_addr, sin,
1688 				    sizeof(struct in_addr));
1689 				if (ifa_ifwithaddr((SA)&ipaddr) == NULL)
1690 					continue;
1691 				cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1692 				off += sizeof(struct in_addr);
1693 				break;
1694 
1695 			default:
1696 				code = &cp[IPOPT_OFFSET + 1] - (u_char *)ip;
1697 				goto bad;
1698 			}
1699 			ntime = iptime();
1700 			memcpy(cp + off, &ntime, sizeof(n_time));
1701 			cp[IPOPT_OFFSET] += sizeof(n_time);
1702 		}
1703 	}
1704 	if (forward && ipforwarding) {
1705 		ip_forward(m, TRUE, next_hop);
1706 		return (1);
1707 	}
1708 	return (0);
1709 bad:
1710 	icmp_error(m, type, code, 0, 0);
1711 	ipstat.ips_badoptions++;
1712 	return (1);
1713 }
1714 
1715 /*
1716  * Given address of next destination (final or next hop),
1717  * return internet address info of interface to be used to get there.
1718  */
1719 struct in_ifaddr *
1720 ip_rtaddr(struct in_addr dst, struct route *ro)
1721 {
1722 	struct sockaddr_in *sin;
1723 
1724 	sin = (struct sockaddr_in *)&ro->ro_dst;
1725 
1726 	if (ro->ro_rt == NULL || dst.s_addr != sin->sin_addr.s_addr) {
1727 		if (ro->ro_rt != NULL) {
1728 			RTFREE(ro->ro_rt);
1729 			ro->ro_rt = NULL;
1730 		}
1731 		sin->sin_family = AF_INET;
1732 		sin->sin_len = sizeof *sin;
1733 		sin->sin_addr = dst;
1734 		rtalloc_ign(ro, RTF_PRCLONING);
1735 	}
1736 
1737 	if (ro->ro_rt == NULL)
1738 		return (NULL);
1739 
1740 	return (ifatoia(ro->ro_rt->rt_ifa));
1741 }
1742 
1743 /*
1744  * Save incoming source route for use in replies,
1745  * to be picked up later by ip_srcroute if the receiver is interested.
1746  */
1747 void
1748 save_rte(u_char *option, struct in_addr dst)
1749 {
1750 	unsigned olen;
1751 
1752 	olen = option[IPOPT_OLEN];
1753 #ifdef DIAGNOSTIC
1754 	if (ipprintfs)
1755 		kprintf("save_rte: olen %d\n", olen);
1756 #endif
1757 	if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst)))
1758 		return;
1759 	bcopy(option, ip_srcrt.srcopt, olen);
1760 	ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1761 	ip_srcrt.dst = dst;
1762 }
1763 
1764 /*
1765  * Retrieve incoming source route for use in replies,
1766  * in the same form used by setsockopt.
1767  * The first hop is placed before the options, will be removed later.
1768  */
1769 struct mbuf *
1770 ip_srcroute(void)
1771 {
1772 	struct in_addr *p, *q;
1773 	struct mbuf *m;
1774 
1775 	if (ip_nhops == 0)
1776 		return (NULL);
1777 	m = m_get(MB_DONTWAIT, MT_HEADER);
1778 	if (m == NULL)
1779 		return (NULL);
1780 
1781 #define	OPTSIZ	(sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt))
1782 
1783 	/* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
1784 	m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) +
1785 	    OPTSIZ;
1786 #ifdef DIAGNOSTIC
1787 	if (ipprintfs)
1788 		kprintf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len);
1789 #endif
1790 
1791 	/*
1792 	 * First save first hop for return route
1793 	 */
1794 	p = &ip_srcrt.route[ip_nhops - 1];
1795 	*(mtod(m, struct in_addr *)) = *p--;
1796 #ifdef DIAGNOSTIC
1797 	if (ipprintfs)
1798 		kprintf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr));
1799 #endif
1800 
1801 	/*
1802 	 * Copy option fields and padding (nop) to mbuf.
1803 	 */
1804 	ip_srcrt.nop = IPOPT_NOP;
1805 	ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
1806 	memcpy(mtod(m, caddr_t) + sizeof(struct in_addr), &ip_srcrt.nop,
1807 	    OPTSIZ);
1808 	q = (struct in_addr *)(mtod(m, caddr_t) +
1809 	    sizeof(struct in_addr) + OPTSIZ);
1810 #undef OPTSIZ
1811 	/*
1812 	 * Record return path as an IP source route,
1813 	 * reversing the path (pointers are now aligned).
1814 	 */
1815 	while (p >= ip_srcrt.route) {
1816 #ifdef DIAGNOSTIC
1817 		if (ipprintfs)
1818 			kprintf(" %x", ntohl(q->s_addr));
1819 #endif
1820 		*q++ = *p--;
1821 	}
1822 	/*
1823 	 * Last hop goes to final destination.
1824 	 */
1825 	*q = ip_srcrt.dst;
1826 #ifdef DIAGNOSTIC
1827 	if (ipprintfs)
1828 		kprintf(" %x\n", ntohl(q->s_addr));
1829 #endif
1830 	return (m);
1831 }
1832 
1833 /*
1834  * Strip out IP options.
1835  */
1836 void
1837 ip_stripoptions(struct mbuf *m)
1838 {
1839 	int datalen;
1840 	struct ip *ip = mtod(m, struct ip *);
1841 	caddr_t opts;
1842 	int optlen;
1843 
1844 	optlen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
1845 	opts = (caddr_t)(ip + 1);
1846 	datalen = m->m_len - (sizeof(struct ip) + optlen);
1847 	bcopy(opts + optlen, opts, datalen);
1848 	m->m_len -= optlen;
1849 	if (m->m_flags & M_PKTHDR)
1850 		m->m_pkthdr.len -= optlen;
1851 	ip->ip_vhl = IP_MAKE_VHL(IPVERSION, sizeof(struct ip) >> 2);
1852 }
1853 
1854 u_char inetctlerrmap[PRC_NCMDS] = {
1855 	0,		0,		0,		0,
1856 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1857 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1858 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1859 	0,		0,		0,		0,
1860 	ENOPROTOOPT,	ECONNREFUSED
1861 };
1862 
1863 /*
1864  * Forward a packet.  If some error occurs return the sender
1865  * an icmp packet.  Note we can't always generate a meaningful
1866  * icmp message because icmp doesn't have a large enough repertoire
1867  * of codes and types.
1868  *
1869  * If not forwarding, just drop the packet.  This could be confusing
1870  * if ipforwarding was zero but some routing protocol was advancing
1871  * us as a gateway to somewhere.  However, we must let the routing
1872  * protocol deal with that.
1873  *
1874  * The using_srcrt parameter indicates whether the packet is being forwarded
1875  * via a source route.
1876  */
1877 void
1878 ip_forward(struct mbuf *m, boolean_t using_srcrt, struct sockaddr_in *next_hop)
1879 {
1880 	struct ip *ip = mtod(m, struct ip *);
1881 	struct sockaddr_in *ipforward_rtaddr;
1882 	struct rtentry *rt;
1883 	int error, type = 0, code = 0, destmtu = 0;
1884 	struct mbuf *mcopy;
1885 	n_long dest;
1886 	struct in_addr pkt_dst;
1887 	struct m_hdr tag;
1888 	struct route *cache_rt = &ipforward_rt[mycpuid];
1889 
1890 	dest = INADDR_ANY;
1891 	/*
1892 	 * Cache the destination address of the packet; this may be
1893 	 * changed by use of 'ipfw fwd'.
1894 	 */
1895 	pkt_dst = (next_hop != NULL) ? next_hop->sin_addr : ip->ip_dst;
1896 
1897 #ifdef DIAGNOSTIC
1898 	if (ipprintfs)
1899 		kprintf("forward: src %x dst %x ttl %x\n",
1900 		       ip->ip_src.s_addr, pkt_dst.s_addr, ip->ip_ttl);
1901 #endif
1902 
1903 	if (m->m_flags & (M_BCAST | M_MCAST) || !in_canforward(pkt_dst)) {
1904 		ipstat.ips_cantforward++;
1905 		m_freem(m);
1906 		return;
1907 	}
1908 	if (!ipstealth && ip->ip_ttl <= IPTTLDEC) {
1909 		icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0);
1910 		return;
1911 	}
1912 
1913 	ipforward_rtaddr = (struct sockaddr_in *) &cache_rt->ro_dst;
1914 	if (cache_rt->ro_rt == NULL ||
1915 	    ipforward_rtaddr->sin_addr.s_addr != pkt_dst.s_addr) {
1916 		if (cache_rt->ro_rt != NULL) {
1917 			RTFREE(cache_rt->ro_rt);
1918 			cache_rt->ro_rt = NULL;
1919 		}
1920 		ipforward_rtaddr->sin_family = AF_INET;
1921 		ipforward_rtaddr->sin_len = sizeof(struct sockaddr_in);
1922 		ipforward_rtaddr->sin_addr = pkt_dst;
1923 		rtalloc_ign(cache_rt, RTF_PRCLONING);
1924 		if (cache_rt->ro_rt == NULL) {
1925 			icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0);
1926 			return;
1927 		}
1928 	}
1929 	rt = cache_rt->ro_rt;
1930 
1931 	/*
1932 	 * Save the IP header and at most 8 bytes of the payload,
1933 	 * in case we need to generate an ICMP message to the src.
1934 	 *
1935 	 * XXX this can be optimized a lot by saving the data in a local
1936 	 * buffer on the stack (72 bytes at most), and only allocating the
1937 	 * mbuf if really necessary. The vast majority of the packets
1938 	 * are forwarded without having to send an ICMP back (either
1939 	 * because unnecessary, or because rate limited), so we are
1940 	 * really we are wasting a lot of work here.
1941 	 *
1942 	 * We don't use m_copy() because it might return a reference
1943 	 * to a shared cluster. Both this function and ip_output()
1944 	 * assume exclusive access to the IP header in `m', so any
1945 	 * data in a cluster may change before we reach icmp_error().
1946 	 */
1947 	MGETHDR(mcopy, MB_DONTWAIT, m->m_type);
1948 	if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, MB_DONTWAIT)) {
1949 		/*
1950 		 * It's probably ok if the pkthdr dup fails (because
1951 		 * the deep copy of the tag chain failed), but for now
1952 		 * be conservative and just discard the copy since
1953 		 * code below may some day want the tags.
1954 		 */
1955 		m_free(mcopy);
1956 		mcopy = NULL;
1957 	}
1958 	if (mcopy != NULL) {
1959 		mcopy->m_len = imin((IP_VHL_HL(ip->ip_vhl) << 2) + 8,
1960 		    (int)ip->ip_len);
1961 		mcopy->m_pkthdr.len = mcopy->m_len;
1962 		m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1963 	}
1964 
1965 	if (!ipstealth)
1966 		ip->ip_ttl -= IPTTLDEC;
1967 
1968 	/*
1969 	 * If forwarding packet using same interface that it came in on,
1970 	 * perhaps should send a redirect to sender to shortcut a hop.
1971 	 * Only send redirect if source is sending directly to us,
1972 	 * and if packet was not source routed (or has any options).
1973 	 * Also, don't send redirect if forwarding using a default route
1974 	 * or a route modified by a redirect.
1975 	 */
1976 	if (rt->rt_ifp == m->m_pkthdr.rcvif &&
1977 	    !(rt->rt_flags & (RTF_DYNAMIC | RTF_MODIFIED)) &&
1978 	    satosin(rt_key(rt))->sin_addr.s_addr != INADDR_ANY &&
1979 	    ipsendredirects && !using_srcrt && next_hop == NULL) {
1980 		u_long src = ntohl(ip->ip_src.s_addr);
1981 		struct in_ifaddr *rt_ifa = (struct in_ifaddr *)rt->rt_ifa;
1982 
1983 		if (rt_ifa != NULL &&
1984 		    (src & rt_ifa->ia_subnetmask) == rt_ifa->ia_subnet) {
1985 			if (rt->rt_flags & RTF_GATEWAY)
1986 				dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
1987 			else
1988 				dest = pkt_dst.s_addr;
1989 			/*
1990 			 * Router requirements says to only send
1991 			 * host redirects.
1992 			 */
1993 			type = ICMP_REDIRECT;
1994 			code = ICMP_REDIRECT_HOST;
1995 #ifdef DIAGNOSTIC
1996 			if (ipprintfs)
1997 				kprintf("redirect (%d) to %x\n", code, dest);
1998 #endif
1999 		}
2000 	}
2001 
2002 	if (next_hop != NULL) {
2003 		/* Pass IPFORWARD info if available */
2004 		tag.mh_type = MT_TAG;
2005 		tag.mh_flags = PACKET_TAG_IPFORWARD;
2006 		tag.mh_data = (caddr_t)next_hop;
2007 		tag.mh_next = m;
2008 		m = (struct mbuf *)&tag;
2009 	}
2010 
2011 	error = ip_output(m, NULL, cache_rt, IP_FORWARDING, NULL,
2012 			  NULL);
2013 	if (error == 0) {
2014 		ipstat.ips_forward++;
2015 		if (type == 0) {
2016 			if (mcopy) {
2017 				/* ipflow_create() will free mcopy */
2018 				ipflow_create(cache_rt, mcopy);
2019 			}
2020 			return;		/* most common case */
2021 		} else {
2022 			ipstat.ips_redirectsent++;
2023 		}
2024 	} else {
2025 		ipstat.ips_cantforward++;
2026 	}
2027 
2028 	if (mcopy == NULL)
2029 		return;
2030 
2031 	/*
2032 	 * Send ICMP message.
2033 	 */
2034 
2035 	switch (error) {
2036 
2037 	case 0:				/* forwarded, but need redirect */
2038 		/* type, code set above */
2039 		break;
2040 
2041 	case ENETUNREACH:		/* shouldn't happen, checked above */
2042 	case EHOSTUNREACH:
2043 	case ENETDOWN:
2044 	case EHOSTDOWN:
2045 	default:
2046 		type = ICMP_UNREACH;
2047 		code = ICMP_UNREACH_HOST;
2048 		break;
2049 
2050 	case EMSGSIZE:
2051 		type = ICMP_UNREACH;
2052 		code = ICMP_UNREACH_NEEDFRAG;
2053 #ifdef IPSEC
2054 		/*
2055 		 * If the packet is routed over IPsec tunnel, tell the
2056 		 * originator the tunnel MTU.
2057 		 *	tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
2058 		 * XXX quickhack!!!
2059 		 */
2060 		if (cache_rt->ro_rt != NULL) {
2061 			struct secpolicy *sp = NULL;
2062 			int ipsecerror;
2063 			int ipsechdr;
2064 			struct route *ro;
2065 
2066 			sp = ipsec4_getpolicybyaddr(mcopy,
2067 						    IPSEC_DIR_OUTBOUND,
2068 						    IP_FORWARDING,
2069 						    &ipsecerror);
2070 
2071 			if (sp == NULL)
2072 				destmtu = cache_rt->ro_rt->rt_ifp->if_mtu;
2073 			else {
2074 				/* count IPsec header size */
2075 				ipsechdr = ipsec4_hdrsiz(mcopy,
2076 							 IPSEC_DIR_OUTBOUND,
2077 							 NULL);
2078 
2079 				/*
2080 				 * find the correct route for outer IPv4
2081 				 * header, compute tunnel MTU.
2082 				 *
2083 				 */
2084 				if (sp->req != NULL && sp->req->sav != NULL &&
2085 				    sp->req->sav->sah != NULL) {
2086 					ro = &sp->req->sav->sah->sa_route;
2087 					if (ro->ro_rt != NULL &&
2088 					    ro->ro_rt->rt_ifp != NULL) {
2089 						destmtu =
2090 						    ro->ro_rt->rt_ifp->if_mtu;
2091 						destmtu -= ipsechdr;
2092 					}
2093 				}
2094 
2095 				key_freesp(sp);
2096 			}
2097 		}
2098 #elif FAST_IPSEC
2099 		/*
2100 		 * If the packet is routed over IPsec tunnel, tell the
2101 		 * originator the tunnel MTU.
2102 		 *	tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
2103 		 * XXX quickhack!!!
2104 		 */
2105 		if (cache_rt->ro_rt != NULL) {
2106 			struct secpolicy *sp = NULL;
2107 			int ipsecerror;
2108 			int ipsechdr;
2109 			struct route *ro;
2110 
2111 			sp = ipsec_getpolicybyaddr(mcopy,
2112 						   IPSEC_DIR_OUTBOUND,
2113 						   IP_FORWARDING,
2114 						   &ipsecerror);
2115 
2116 			if (sp == NULL)
2117 				destmtu = cache_rt->ro_rt->rt_ifp->if_mtu;
2118 			else {
2119 				/* count IPsec header size */
2120 				ipsechdr = ipsec4_hdrsiz(mcopy,
2121 							 IPSEC_DIR_OUTBOUND,
2122 							 NULL);
2123 
2124 				/*
2125 				 * find the correct route for outer IPv4
2126 				 * header, compute tunnel MTU.
2127 				 */
2128 
2129 				if (sp->req != NULL &&
2130 				    sp->req->sav != NULL &&
2131 				    sp->req->sav->sah != NULL) {
2132 					ro = &sp->req->sav->sah->sa_route;
2133 					if (ro->ro_rt != NULL &&
2134 					    ro->ro_rt->rt_ifp != NULL) {
2135 						destmtu =
2136 						    ro->ro_rt->rt_ifp->if_mtu;
2137 						destmtu -= ipsechdr;
2138 					}
2139 				}
2140 
2141 				KEY_FREESP(&sp);
2142 			}
2143 		}
2144 #else /* !IPSEC && !FAST_IPSEC */
2145 		if (cache_rt->ro_rt != NULL)
2146 			destmtu = cache_rt->ro_rt->rt_ifp->if_mtu;
2147 #endif /*IPSEC*/
2148 		ipstat.ips_cantfrag++;
2149 		break;
2150 
2151 	case ENOBUFS:
2152 		/*
2153 		 * A router should not generate ICMP_SOURCEQUENCH as
2154 		 * required in RFC1812 Requirements for IP Version 4 Routers.
2155 		 * Source quench could be a big problem under DoS attacks,
2156 		 * or if the underlying interface is rate-limited.
2157 		 * Those who need source quench packets may re-enable them
2158 		 * via the net.inet.ip.sendsourcequench sysctl.
2159 		 */
2160 		if (!ip_sendsourcequench) {
2161 			m_freem(mcopy);
2162 			return;
2163 		} else {
2164 			type = ICMP_SOURCEQUENCH;
2165 			code = 0;
2166 		}
2167 		break;
2168 
2169 	case EACCES:			/* ipfw denied packet */
2170 		m_freem(mcopy);
2171 		return;
2172 	}
2173 	icmp_error(mcopy, type, code, dest, destmtu);
2174 }
2175 
2176 void
2177 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
2178 	       struct mbuf *m)
2179 {
2180 	if (inp->inp_socket->so_options & SO_TIMESTAMP) {
2181 		struct timeval tv;
2182 
2183 		microtime(&tv);
2184 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
2185 		    SCM_TIMESTAMP, SOL_SOCKET);
2186 		if (*mp)
2187 			mp = &(*mp)->m_next;
2188 	}
2189 	if (inp->inp_flags & INP_RECVDSTADDR) {
2190 		*mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
2191 		    sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
2192 		if (*mp)
2193 			mp = &(*mp)->m_next;
2194 	}
2195 	if (inp->inp_flags & INP_RECVTTL) {
2196 		*mp = sbcreatecontrol((caddr_t) &ip->ip_ttl,
2197 		    sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
2198 		if (*mp)
2199 			mp = &(*mp)->m_next;
2200 	}
2201 #ifdef notyet
2202 	/* XXX
2203 	 * Moving these out of udp_input() made them even more broken
2204 	 * than they already were.
2205 	 */
2206 	/* options were tossed already */
2207 	if (inp->inp_flags & INP_RECVOPTS) {
2208 		*mp = sbcreatecontrol((caddr_t) opts_deleted_above,
2209 		    sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
2210 		if (*mp)
2211 			mp = &(*mp)->m_next;
2212 	}
2213 	/* ip_srcroute doesn't do what we want here, need to fix */
2214 	if (inp->inp_flags & INP_RECVRETOPTS) {
2215 		*mp = sbcreatecontrol((caddr_t) ip_srcroute(),
2216 		    sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
2217 		if (*mp)
2218 			mp = &(*mp)->m_next;
2219 	}
2220 #endif
2221 	if (inp->inp_flags & INP_RECVIF) {
2222 		struct ifnet *ifp;
2223 		struct sdlbuf {
2224 			struct sockaddr_dl sdl;
2225 			u_char	pad[32];
2226 		} sdlbuf;
2227 		struct sockaddr_dl *sdp;
2228 		struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
2229 
2230 		if (((ifp = m->m_pkthdr.rcvif)) &&
2231 		    ((ifp->if_index != 0) && (ifp->if_index <= if_index))) {
2232 			sdp = IF_LLSOCKADDR(ifp);
2233 			/*
2234 			 * Change our mind and don't try copy.
2235 			 */
2236 			if ((sdp->sdl_family != AF_LINK) ||
2237 			    (sdp->sdl_len > sizeof(sdlbuf))) {
2238 				goto makedummy;
2239 			}
2240 			bcopy(sdp, sdl2, sdp->sdl_len);
2241 		} else {
2242 makedummy:
2243 			sdl2->sdl_len =
2244 			    offsetof(struct sockaddr_dl, sdl_data[0]);
2245 			sdl2->sdl_family = AF_LINK;
2246 			sdl2->sdl_index = 0;
2247 			sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
2248 		}
2249 		*mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len,
2250 			IP_RECVIF, IPPROTO_IP);
2251 		if (*mp)
2252 			mp = &(*mp)->m_next;
2253 	}
2254 }
2255 
2256 /*
2257  * XXX these routines are called from the upper part of the kernel.
2258  *
2259  * They could also be moved to ip_mroute.c, since all the RSVP
2260  *  handling is done there already.
2261  */
2262 int
2263 ip_rsvp_init(struct socket *so)
2264 {
2265 	if (so->so_type != SOCK_RAW ||
2266 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
2267 		return EOPNOTSUPP;
2268 
2269 	if (ip_rsvpd != NULL)
2270 		return EADDRINUSE;
2271 
2272 	ip_rsvpd = so;
2273 	/*
2274 	 * This may seem silly, but we need to be sure we don't over-increment
2275 	 * the RSVP counter, in case something slips up.
2276 	 */
2277 	if (!ip_rsvp_on) {
2278 		ip_rsvp_on = 1;
2279 		rsvp_on++;
2280 	}
2281 
2282 	return 0;
2283 }
2284 
2285 int
2286 ip_rsvp_done(void)
2287 {
2288 	ip_rsvpd = NULL;
2289 	/*
2290 	 * This may seem silly, but we need to be sure we don't over-decrement
2291 	 * the RSVP counter, in case something slips up.
2292 	 */
2293 	if (ip_rsvp_on) {
2294 		ip_rsvp_on = 0;
2295 		rsvp_on--;
2296 	}
2297 	return 0;
2298 }
2299 
2300 void
2301 rsvp_input(struct mbuf *m, ...)	/* XXX must fixup manually */
2302 {
2303 	int off, proto;
2304 	__va_list ap;
2305 
2306 	__va_start(ap, m);
2307 	off = __va_arg(ap, int);
2308 	proto = __va_arg(ap, int);
2309 	__va_end(ap);
2310 
2311 	if (rsvp_input_p) { /* call the real one if loaded */
2312 		rsvp_input_p(m, off, proto);
2313 		return;
2314 	}
2315 
2316 	/* Can still get packets with rsvp_on = 0 if there is a local member
2317 	 * of the group to which the RSVP packet is addressed.  But in this
2318 	 * case we want to throw the packet away.
2319 	 */
2320 
2321 	if (!rsvp_on) {
2322 		m_freem(m);
2323 		return;
2324 	}
2325 
2326 	if (ip_rsvpd != NULL) {
2327 		rip_input(m, off, proto);
2328 		return;
2329 	}
2330 	/* Drop the packet */
2331 	m_freem(m);
2332 }
2333