xref: /dflybsd-src/sys/netinet/ip_input.c (revision c2576c10ab1201aa47eb10096e226be7b357d01d)
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) 2003, 2004 Jeffrey M. Hsu.  All rights reserved.
36  *
37  * License terms: all terms for the DragonFly license above plus the following:
38  *
39  * 4. All advertising materials mentioning features or use of this software
40  *    must display the following acknowledgement:
41  *
42  *	This product includes software developed by Jeffrey M. Hsu
43  *	for the DragonFly Project.
44  *
45  *    This requirement may be waived with permission from Jeffrey Hsu.
46  *    This requirement will sunset and may be removed on July 8 2005,
47  *    after which the standard DragonFly license (as shown above) will
48  *    apply.
49  */
50 
51 /*
52  * Copyright (c) 1982, 1986, 1988, 1993
53  *	The Regents of the University of California.  All rights reserved.
54  *
55  * Redistribution and use in source and binary forms, with or without
56  * modification, are permitted provided that the following conditions
57  * are met:
58  * 1. Redistributions of source code must retain the above copyright
59  *    notice, this list of conditions and the following disclaimer.
60  * 2. Redistributions in binary form must reproduce the above copyright
61  *    notice, this list of conditions and the following disclaimer in the
62  *    documentation and/or other materials provided with the distribution.
63  * 3. All advertising materials mentioning features or use of this software
64  *    must display the following acknowledgement:
65  *	This product includes software developed by the University of
66  *	California, Berkeley and its contributors.
67  * 4. Neither the name of the University nor the names of its contributors
68  *    may be used to endorse or promote products derived from this software
69  *    without specific prior written permission.
70  *
71  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
72  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
73  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
74  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
75  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
76  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
77  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
78  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
79  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
80  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
81  * SUCH DAMAGE.
82  *
83  *	@(#)ip_input.c	8.2 (Berkeley) 1/4/94
84  * $FreeBSD: src/sys/netinet/ip_input.c,v 1.130.2.52 2003/03/07 07:01:28 silby Exp $
85  * $DragonFly: src/sys/netinet/ip_input.c,v 1.46 2005/02/11 22:25:57 joerg Exp $
86  */
87 
88 #define	_IP_VHL
89 
90 #include "opt_bootp.h"
91 #include "opt_ipfw.h"
92 #include "opt_ipdn.h"
93 #include "opt_ipdivert.h"
94 #include "opt_ipfilter.h"
95 #include "opt_ipstealth.h"
96 #include "opt_ipsec.h"
97 #include "opt_random_ip_id.h"
98 
99 #include <sys/param.h>
100 #include <sys/systm.h>
101 #include <sys/mbuf.h>
102 #include <sys/malloc.h>
103 #include <sys/mpipe.h>
104 #include <sys/domain.h>
105 #include <sys/protosw.h>
106 #include <sys/socket.h>
107 #include <sys/time.h>
108 #include <sys/globaldata.h>
109 #include <sys/thread.h>
110 #include <sys/kernel.h>
111 #include <sys/syslog.h>
112 #include <sys/sysctl.h>
113 #include <sys/in_cksum.h>
114 
115 #include <sys/thread2.h>
116 #include <sys/msgport2.h>
117 
118 #include <machine/stdarg.h>
119 
120 #include <net/if.h>
121 #include <net/if_types.h>
122 #include <net/if_var.h>
123 #include <net/if_dl.h>
124 #include <net/pfil.h>
125 #include <net/route.h>
126 #include <net/netisr.h>
127 #include <net/intrq.h>
128 
129 #include <netinet/in.h>
130 #include <netinet/in_systm.h>
131 #include <netinet/in_var.h>
132 #include <netinet/ip.h>
133 #include <netinet/in_pcb.h>
134 #include <netinet/ip_var.h>
135 #include <netinet/ip_icmp.h>
136 
137 
138 #include <sys/socketvar.h>
139 
140 #include <net/ipfw/ip_fw.h>
141 #include <net/dummynet/ip_dummynet.h>
142 
143 #ifdef IPSEC
144 #include <netinet6/ipsec.h>
145 #include <netproto/key/key.h>
146 #endif
147 
148 #ifdef FAST_IPSEC
149 #include <netproto/ipsec/ipsec.h>
150 #include <netproto/ipsec/key.h>
151 #endif
152 
153 int rsvp_on = 0;
154 static int ip_rsvp_on;
155 struct socket *ip_rsvpd;
156 
157 int ipforwarding = 0;
158 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW,
159     &ipforwarding, 0, "Enable IP forwarding between interfaces");
160 
161 static int ipsendredirects = 1; /* XXX */
162 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW,
163     &ipsendredirects, 0, "Enable sending IP redirects");
164 
165 int ip_defttl = IPDEFTTL;
166 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW,
167     &ip_defttl, 0, "Maximum TTL on IP packets");
168 
169 static int ip_dosourceroute = 0;
170 SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW,
171     &ip_dosourceroute, 0, "Enable forwarding source routed IP packets");
172 
173 static int ip_acceptsourceroute = 0;
174 SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute,
175     CTLFLAG_RW, &ip_acceptsourceroute, 0,
176     "Enable accepting source routed IP packets");
177 
178 static int ip_keepfaith = 0;
179 SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW,
180     &ip_keepfaith, 0,
181     "Enable packet capture for FAITH IPv4->IPv6 translater daemon");
182 
183 static int nipq = 0;	/* total # of reass queues */
184 static int maxnipq;
185 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragpackets, CTLFLAG_RW,
186     &maxnipq, 0,
187     "Maximum number of IPv4 fragment reassembly queue entries");
188 
189 static int maxfragsperpacket;
190 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW,
191     &maxfragsperpacket, 0,
192     "Maximum number of IPv4 fragments allowed per packet");
193 
194 static int ip_sendsourcequench = 0;
195 SYSCTL_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW,
196     &ip_sendsourcequench, 0,
197     "Enable the transmission of source quench packets");
198 
199 /*
200  * XXX - Setting ip_checkinterface mostly implements the receive side of
201  * the Strong ES model described in RFC 1122, but since the routing table
202  * and transmit implementation do not implement the Strong ES model,
203  * setting this to 1 results in an odd hybrid.
204  *
205  * XXX - ip_checkinterface currently must be disabled if you use ipnat
206  * to translate the destination address to another local interface.
207  *
208  * XXX - ip_checkinterface must be disabled if you add IP aliases
209  * to the loopback interface instead of the interface where the
210  * packets for those addresses are received.
211  */
212 static int ip_checkinterface = 0;
213 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW,
214     &ip_checkinterface, 0, "Verify packet arrives on correct interface");
215 
216 #ifdef DIAGNOSTIC
217 static int ipprintfs = 0;
218 #endif
219 
220 static struct ifqueue ipintrq;
221 static int ipqmaxlen = IFQ_MAXLEN;
222 
223 extern	struct domain inetdomain;
224 extern	struct protosw inetsw[];
225 u_char	ip_protox[IPPROTO_MAX];
226 struct	in_ifaddrhead in_ifaddrhead;		/* first inet address */
227 struct	in_ifaddrhashhead *in_ifaddrhashtbl;	/* inet addr hash table */
228 u_long	in_ifaddrhmask;				/* mask for hash table */
229 
230 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW,
231     &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue");
232 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD,
233     &ipintrq.ifq_drops, 0, "Number of packets dropped from the IP input queue");
234 
235 struct ip_stats ipstats_ary[MAXCPU];
236 #ifdef SMP
237 static int
238 sysctl_ipstats(SYSCTL_HANDLER_ARGS)
239 {
240 	int cpu, error = 0;
241 
242 	for (cpu = 0; cpu < ncpus; ++cpu) {
243 		if ((error = SYSCTL_OUT(req, &ipstats_ary[cpu],
244 					sizeof(struct ip_stats))))
245 			break;
246 		if ((error = SYSCTL_IN(req, &ipstats_ary[cpu],
247 				       sizeof(struct ip_stats))))
248 			break;
249 	}
250 
251 	return (error);
252 }
253 SYSCTL_PROC(_net_inet_ip, IPCTL_STATS, stats, (CTLTYPE_OPAQUE | CTLFLAG_RW),
254     0, 0, sysctl_ipstats, "S,ip_stats", "IP statistics");
255 #else
256 SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW,
257     &ipstat, ip_stats, "IP statistics");
258 #endif
259 
260 /* Packet reassembly stuff */
261 #define	IPREASS_NHASH_LOG2	6
262 #define	IPREASS_NHASH		(1 << IPREASS_NHASH_LOG2)
263 #define	IPREASS_HMASK		(IPREASS_NHASH - 1)
264 #define	IPREASS_HASH(x,y)						\
265     (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK)
266 
267 static struct ipq ipq[IPREASS_NHASH];
268 const  int    ipintrq_present = 1;
269 
270 #ifdef IPCTL_DEFMTU
271 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
272     &ip_mtu, 0, "Default MTU");
273 #endif
274 
275 #ifdef IPSTEALTH
276 static int ipstealth = 0;
277 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW, &ipstealth, 0, "");
278 #else
279 static const int ipstealth = 0;
280 #endif
281 
282 
283 /* Firewall hooks */
284 ip_fw_chk_t *ip_fw_chk_ptr;
285 int fw_enable = 1;
286 int fw_one_pass = 1;
287 
288 /* Dummynet hooks */
289 ip_dn_io_t *ip_dn_io_ptr;
290 
291 struct pfil_head inet_pfil_hook;
292 
293 /*
294  * XXX this is ugly -- the following two global variables are
295  * used to store packet state while it travels through the stack.
296  * Note that the code even makes assumptions on the size and
297  * alignment of fields inside struct ip_srcrt so e.g. adding some
298  * fields will break the code. This needs to be fixed.
299  *
300  * We need to save the IP options in case a protocol wants to respond
301  * to an incoming packet over the same route if the packet got here
302  * using IP source routing.  This allows connection establishment and
303  * maintenance when the remote end is on a network that is not known
304  * to us.
305  */
306 static int ip_nhops = 0;
307 
308 static	struct ip_srcrt {
309 	struct	in_addr dst;			/* final destination */
310 	char	nop;				/* one NOP to align */
311 	char	srcopt[IPOPT_OFFSET + 1];	/* OPTVAL, OLEN and OFFSET */
312 	struct	in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)];
313 } ip_srcrt;
314 
315 static MALLOC_DEFINE(M_IPQ, "ipq", "IP Fragment Management");
316 static struct malloc_pipe ipq_mpipe;
317 
318 static void		save_rte (u_char *, struct in_addr);
319 static int		ip_dooptions (struct mbuf *m, int,
320 					struct sockaddr_in *next_hop);
321 static void		ip_forward (struct mbuf *m, int srcrt,
322 					struct sockaddr_in *next_hop);
323 static void		ip_freef (struct ipq *);
324 static int		ip_input_handler (struct netmsg *);
325 static struct mbuf	*ip_reass (struct mbuf *, struct ipq *,
326 					struct ipq *, u_int32_t *, u_int16_t *);
327 
328 /*
329  * IP initialization: fill in IP protocol switch table.
330  * All protocols not implemented in kernel go to raw IP protocol handler.
331  */
332 void
333 ip_init(void)
334 {
335 	struct protosw *pr;
336 	int i;
337 #ifdef SMP
338 	int cpu;
339 #endif
340 
341 	/*
342 	 * Make sure we can handle a reasonable number of fragments but
343 	 * cap it at 4000 (XXX).
344 	 */
345 	mpipe_init(&ipq_mpipe, M_IPQ, sizeof(struct ipq),
346 		    IFQ_MAXLEN, 4000, 0, NULL);
347 	TAILQ_INIT(&in_ifaddrhead);
348 	in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask);
349 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
350 	if (pr == NULL)
351 		panic("ip_init");
352 	for (i = 0; i < IPPROTO_MAX; i++)
353 		ip_protox[i] = pr - inetsw;
354 	for (pr = inetdomain.dom_protosw;
355 	     pr < inetdomain.dom_protoswNPROTOSW; pr++)
356 		if (pr->pr_domain->dom_family == PF_INET &&
357 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
358 			ip_protox[pr->pr_protocol] = pr - inetsw;
359 
360 	inet_pfil_hook.ph_type = PFIL_TYPE_AF;
361 	inet_pfil_hook.ph_af = AF_INET;
362 	if ((i = pfil_head_register(&inet_pfil_hook)) != 0) {
363 		printf("%s: WARNING: unable to register pfil hook, "
364 			"error %d\n", __func__, i);
365 	}
366 
367 	for (i = 0; i < IPREASS_NHASH; i++)
368 	    ipq[i].next = ipq[i].prev = &ipq[i];
369 
370 	maxnipq = nmbclusters / 32;
371 	maxfragsperpacket = 16;
372 
373 #ifndef RANDOM_IP_ID
374 	ip_id = time_second & 0xffff;
375 #endif
376 	ipintrq.ifq_maxlen = ipqmaxlen;
377 
378 	/*
379 	 * Initialize IP statistics.
380 	 *
381 	 * It is layed out as an array which is has one element for UP,
382 	 * and SMP_MAXCPU elements for SMP.  This allows us to retain
383 	 * the access mechanism from userland for both UP and SMP.
384 	 */
385 #ifdef SMP
386 	for (cpu = 0; cpu < ncpus; ++cpu) {
387 		bzero(&ipstats_ary[cpu], sizeof(struct ip_stats));
388 	}
389 #else
390 	bzero(&ipstat, sizeof(struct ip_stats));
391 #endif
392 
393 	netisr_register(NETISR_IP, ip_mport, ip_input_handler);
394 }
395 
396 /*
397  * XXX watch out this one. It is perhaps used as a cache for
398  * the most recently used route ? it is cleared in in_addroute()
399  * when a new route is successfully created.
400  */
401 struct route ipforward_rt;
402 
403 /* Do transport protocol processing. */
404 static void
405 transport_processing_oncpu(struct mbuf *m, int hlen, struct ip *ip,
406 			   struct sockaddr_in *nexthop)
407 {
408 	/*
409 	 * Switch out to protocol's input routine.
410 	 */
411 	if (nexthop && ip->ip_p == IPPROTO_TCP) {
412 		/* TCP needs IPFORWARD info if available */
413 		struct m_hdr tag;
414 
415 		tag.mh_type = MT_TAG;
416 		tag.mh_flags = PACKET_TAG_IPFORWARD;
417 		tag.mh_data = (caddr_t)nexthop;
418 		tag.mh_next = m;
419 
420 		(*inetsw[ip_protox[ip->ip_p]].pr_input)
421 		    ((struct mbuf *)&tag, hlen, ip->ip_p);
422 	} else {
423 		(*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen, ip->ip_p);
424 	}
425 }
426 
427 struct netmsg_transport_packet {
428 	struct lwkt_msg		nm_lmsg;
429 	struct mbuf		*nm_mbuf;
430 	int			nm_hlen;
431 	boolean_t		nm_hasnexthop;
432 	struct sockaddr_in	nm_nexthop;
433 };
434 
435 static int
436 transport_processing_handler(lwkt_msg_t lmsg)
437 {
438 	struct netmsg_transport_packet *msg = (void *)lmsg;
439 	struct sockaddr_in *nexthop;
440 	struct ip *ip;
441 
442 	ip = mtod(msg->nm_mbuf, struct ip *);
443 	nexthop = msg->nm_hasnexthop ? &msg->nm_nexthop : NULL;
444 	transport_processing_oncpu(msg->nm_mbuf, msg->nm_hlen, ip, nexthop);
445 	lwkt_replymsg(lmsg, 0);
446 	return(EASYNC);
447 }
448 
449 static int
450 ip_input_handler(struct netmsg *msg0)
451 {
452 	struct mbuf *m = ((struct netmsg_packet *)msg0)->nm_packet;
453 
454 	ip_input(m);
455 	lwkt_replymsg(&msg0->nm_lmsg, 0);
456 	return(EASYNC);
457 }
458 
459 /*
460  * IP input routine.  Checksum and byte swap header.  If fragmented
461  * try to reassemble.  Process options.  Pass to next level.
462  */
463 void
464 ip_input(struct mbuf *m)
465 {
466 	struct ip *ip;
467 	struct ipq *fp;
468 	struct in_ifaddr *ia = NULL;
469 	struct ifaddr *ifa;
470 	int i, hlen, checkif;
471 	u_short sum;
472 	struct in_addr pkt_dst;
473 	u_int32_t divert_info = 0;		/* packet divert/tee info */
474 	struct ip_fw_args args;
475 	boolean_t using_srcrt = FALSE;		/* forward (by PFIL_HOOKS) */
476 	boolean_t needredispatch = FALSE;
477 	struct in_addr odst;			/* original dst address(NAT) */
478 #ifdef FAST_IPSEC
479 	struct m_tag *mtag;
480 	struct tdb_ident *tdbi;
481 	struct secpolicy *sp;
482 	int s, error;
483 #endif
484 
485 	args.eh = NULL;
486 	args.oif = NULL;
487 	args.rule = NULL;
488 	args.divert_rule = 0;			/* divert cookie */
489 	args.next_hop = NULL;
490 
491 	/* Grab info from MT_TAG mbufs prepended to the chain. */
492 	while (m != NULL && m->m_type == MT_TAG) {
493 		switch(m->_m_tag_id) {
494 		case PACKET_TAG_DUMMYNET:
495 			args.rule = ((struct dn_pkt *)m)->rule;
496 			break;
497 		case PACKET_TAG_DIVERT:
498 			args.divert_rule = (int)m->m_hdr.mh_data & 0xffff;
499 			break;
500 		case PACKET_TAG_IPFORWARD:
501 			args.next_hop = (struct sockaddr_in *)m->m_hdr.mh_data;
502 			break;
503 		default:
504 			printf("ip_input: unrecognised MT_TAG tag %d\n",
505 			    m->_m_tag_id);
506 			break;
507 		}
508 		m = m->m_next;
509 	}
510 	KASSERT(m != NULL && (m->m_flags & M_PKTHDR), ("ip_input: no HDR"));
511 
512 	if (args.rule != NULL) {	/* dummynet already filtered us */
513 		ip = mtod(m, struct ip *);
514 		hlen = IP_VHL_HL(ip->ip_vhl) << 2;
515 		goto iphack;
516 	}
517 
518 	ipstat.ips_total++;
519 
520 	/* length checks already done in ip_demux() */
521 	KASSERT(m->m_len >= sizeof(ip), ("IP header not in one mbuf"));
522 
523 	ip = mtod(m, struct ip *);
524 
525 	if (IP_VHL_V(ip->ip_vhl) != IPVERSION) {
526 		ipstat.ips_badvers++;
527 		goto bad;
528 	}
529 
530 	hlen = IP_VHL_HL(ip->ip_vhl) << 2;
531 	/* length checks already done in ip_demux() */
532 	KASSERT(hlen >= sizeof(struct ip), ("IP header len too small"));
533 	KASSERT(m->m_len >= hlen, ("packet shorter than IP header length"));
534 
535 	/* 127/8 must not appear on wire - RFC1122 */
536 	if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
537 	    (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
538 		if (!(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK)) {
539 			ipstat.ips_badaddr++;
540 			goto bad;
541 		}
542 	}
543 
544 	if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
545 		sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
546 	} else {
547 		if (hlen == sizeof(struct ip)) {
548 			sum = in_cksum_hdr(ip);
549 		} else {
550 			sum = in_cksum(m, hlen);
551 		}
552 	}
553 	if (sum != 0) {
554 		ipstat.ips_badsum++;
555 		goto bad;
556 	}
557 
558 #ifdef ALTQ
559 	if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) {
560 		/* packet is dropped by traffic conditioner */
561 		return;
562 	}
563 #endif
564 	/*
565 	 * Convert fields to host representation.
566 	 */
567 	ip->ip_len = ntohs(ip->ip_len);
568 	if (ip->ip_len < hlen) {
569 		ipstat.ips_badlen++;
570 		goto bad;
571 	}
572 	ip->ip_off = ntohs(ip->ip_off);
573 
574 	/*
575 	 * Check that the amount of data in the buffers
576 	 * is as at least much as the IP header would have us expect.
577 	 * Trim mbufs if longer than we expect.
578 	 * Drop packet if shorter than we expect.
579 	 */
580 	if (m->m_pkthdr.len < ip->ip_len) {
581 		ipstat.ips_tooshort++;
582 		goto bad;
583 	}
584 	if (m->m_pkthdr.len > ip->ip_len) {
585 		if (m->m_len == m->m_pkthdr.len) {
586 			m->m_len = ip->ip_len;
587 			m->m_pkthdr.len = ip->ip_len;
588 		} else
589 			m_adj(m, ip->ip_len - m->m_pkthdr.len);
590 	}
591 #if defined(IPSEC) && !defined(IPSEC_FILTERGIF)
592 	/*
593 	 * Bypass packet filtering for packets from a tunnel (gif).
594 	 */
595 	if (ipsec_gethist(m, NULL))
596 		goto pass;
597 #endif
598 
599 	/*
600 	 * IpHack's section.
601 	 * Right now when no processing on packet has done
602 	 * and it is still fresh out of network we do our black
603 	 * deals with it.
604 	 * - Firewall: deny/allow/divert
605 	 * - Xlate: translate packet's addr/port (NAT).
606 	 * - Pipe: pass pkt through dummynet.
607 	 * - Wrap: fake packet's addr/port <unimpl.>
608 	 * - Encapsulate: put it in another IP and send out. <unimp.>
609 	 */
610 
611 iphack:
612 
613 	/*
614 	 * Run through list of hooks for input packets.
615 	 *
616 	 * NB: Beware of the destination address changing (e.g.
617 	 *     by NAT rewriting). When this happens, tell
618 	 *     ip_forward to do the right thing.
619 	 */
620 	if (pfil_has_hooks(&inet_pfil_hook)) {
621 		odst = ip->ip_dst;
622 		if (pfil_run_hooks(&inet_pfil_hook, &m,
623 		    m->m_pkthdr.rcvif, PFIL_IN)) {
624 			return;
625 		}
626 		if (m == NULL)			/* consumed by filter */
627 			return;
628 		ip = mtod(m, struct ip *);
629 		using_srcrt = (odst.s_addr != ip->ip_dst.s_addr);
630 	}
631 
632 	if (fw_enable && IPFW_LOADED) {
633 		/*
634 		 * If we've been forwarded from the output side, then
635 		 * skip the firewall a second time
636 		 */
637 		if (args.next_hop != NULL)
638 			goto ours;
639 
640 		args.m = m;
641 		i = ip_fw_chk_ptr(&args);
642 		m = args.m;
643 
644 		if ((i & IP_FW_PORT_DENY_FLAG) || m == NULL) {	/* drop */
645 			if (m != NULL)
646 				m_freem(m);
647 			return;
648 		}
649 		ip = mtod(m, struct ip *);	/* just in case m changed */
650 		if (i == 0 && args.next_hop == NULL)	/* common case */
651 			goto pass;
652 		if (DUMMYNET_LOADED && (i & IP_FW_PORT_DYNT_FLAG)) {
653 			/* Send packet to the appropriate pipe */
654 			ip_dn_io_ptr(m, i&0xffff, DN_TO_IP_IN, &args);
655 			return;
656 		}
657 #ifdef IPDIVERT
658 		if (i != 0 && !(i & IP_FW_PORT_DYNT_FLAG)) {
659 			/* Divert or tee packet */
660 			divert_info = i;
661 			goto ours;
662 		}
663 #endif
664 		if (i == 0 && args.next_hop != NULL)
665 			goto pass;
666 		/*
667 		 * if we get here, the packet must be dropped
668 		 */
669 		m_freem(m);
670 		return;
671 	}
672 pass:
673 
674 	/*
675 	 * Process options and, if not destined for us,
676 	 * ship it on.  ip_dooptions returns 1 when an
677 	 * error was detected (causing an icmp message
678 	 * to be sent and the original packet to be freed).
679 	 */
680 	ip_nhops = 0;		/* for source routed packets */
681 	if (hlen > sizeof(struct ip) && ip_dooptions(m, 0, args.next_hop))
682 		return;
683 
684 	/* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
685 	 * matter if it is destined to another node, or whether it is
686 	 * a multicast one, RSVP wants it! and prevents it from being forwarded
687 	 * anywhere else. Also checks if the rsvp daemon is running before
688 	 * grabbing the packet.
689 	 */
690 	if (rsvp_on && ip->ip_p == IPPROTO_RSVP)
691 		goto ours;
692 
693 	/*
694 	 * Check our list of addresses, to see if the packet is for us.
695 	 * If we don't have any addresses, assume any unicast packet
696 	 * we receive might be for us (and let the upper layers deal
697 	 * with it).
698 	 */
699 	if (TAILQ_EMPTY(&in_ifaddrhead) && !(m->m_flags & (M_MCAST | M_BCAST)))
700 		goto ours;
701 
702 	/*
703 	 * Cache the destination address of the packet; this may be
704 	 * changed by use of 'ipfw fwd'.
705 	 */
706 	pkt_dst = args.next_hop ? args.next_hop->sin_addr : ip->ip_dst;
707 
708 	/*
709 	 * Enable a consistency check between the destination address
710 	 * and the arrival interface for a unicast packet (the RFC 1122
711 	 * strong ES model) if IP forwarding is disabled and the packet
712 	 * is not locally generated and the packet is not subject to
713 	 * 'ipfw fwd'.
714 	 *
715 	 * XXX - Checking also should be disabled if the destination
716 	 * address is ipnat'ed to a different interface.
717 	 *
718 	 * XXX - Checking is incompatible with IP aliases added
719 	 * to the loopback interface instead of the interface where
720 	 * the packets are received.
721 	 */
722 	checkif = ip_checkinterface &&
723 		  !ipforwarding &&
724 		  m->m_pkthdr.rcvif != NULL &&
725 		  !(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) &&
726 		  (args.next_hop == NULL);
727 
728 	/*
729 	 * Check for exact addresses in the hash bucket.
730 	 */
731 	LIST_FOREACH(ia, INADDR_HASH(pkt_dst.s_addr), ia_hash) {
732 		/*
733 		 * If the address matches, verify that the packet
734 		 * arrived via the correct interface if checking is
735 		 * enabled.
736 		 */
737 		if (IA_SIN(ia)->sin_addr.s_addr == pkt_dst.s_addr &&
738 		    (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif))
739 			goto ours;
740 	}
741 	/*
742 	 * Check for broadcast addresses.
743 	 *
744 	 * Only accept broadcast packets that arrive via the matching
745 	 * interface.  Reception of forwarded directed broadcasts would
746 	 * be handled via ip_forward() and ether_output() with the loopback
747 	 * into the stack for SIMPLEX interfaces handled by ether_output().
748 	 */
749 	if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
750 		TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
751 			if (ifa->ifa_addr == NULL) /* shutdown/startup race */
752 				continue;
753 			if (ifa->ifa_addr->sa_family != AF_INET)
754 				continue;
755 			ia = ifatoia(ifa);
756 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
757 								pkt_dst.s_addr)
758 				goto ours;
759 			if (ia->ia_netbroadcast.s_addr == pkt_dst.s_addr)
760 				goto ours;
761 #ifdef BOOTP_COMPAT
762 			if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY)
763 				goto ours;
764 #endif
765 		}
766 	}
767 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
768 		struct in_multi *inm;
769 
770 		if (ip_mrouter != NULL) {
771 			/*
772 			 * If we are acting as a multicast router, all
773 			 * incoming multicast packets are passed to the
774 			 * kernel-level multicast forwarding function.
775 			 * The packet is returned (relatively) intact; if
776 			 * ip_mforward() returns a non-zero value, the packet
777 			 * must be discarded, else it may be accepted below.
778 			 */
779 			if (ip_mforward != NULL &&
780 			    ip_mforward(ip, m->m_pkthdr.rcvif, m, NULL) != 0) {
781 				ipstat.ips_cantforward++;
782 				m_freem(m);
783 				return;
784 			}
785 
786 			/*
787 			 * The process-level routing daemon needs to receive
788 			 * all multicast IGMP packets, whether or not this
789 			 * host belongs to their destination groups.
790 			 */
791 			if (ip->ip_p == IPPROTO_IGMP)
792 				goto ours;
793 			ipstat.ips_forward++;
794 		}
795 		/*
796 		 * See if we belong to the destination multicast group on the
797 		 * arrival interface.
798 		 */
799 		IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
800 		if (inm == NULL) {
801 			ipstat.ips_notmember++;
802 			m_freem(m);
803 			return;
804 		}
805 		goto ours;
806 	}
807 	if (ip->ip_dst.s_addr == INADDR_BROADCAST)
808 		goto ours;
809 	if (ip->ip_dst.s_addr == INADDR_ANY)
810 		goto ours;
811 
812 	/*
813 	 * FAITH(Firewall Aided Internet Translator)
814 	 */
815 	if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
816 		if (ip_keepfaith) {
817 			if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP)
818 				goto ours;
819 		}
820 		m_freem(m);
821 		return;
822 	}
823 
824 	/*
825 	 * Not for us; forward if possible and desirable.
826 	 */
827 	if (!ipforwarding) {
828 		ipstat.ips_cantforward++;
829 		m_freem(m);
830 	} else {
831 #ifdef IPSEC
832 		/*
833 		 * Enforce inbound IPsec SPD.
834 		 */
835 		if (ipsec4_in_reject(m, NULL)) {
836 			ipsecstat.in_polvio++;
837 			goto bad;
838 		}
839 #endif
840 #ifdef FAST_IPSEC
841 		mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
842 		s = splnet();
843 		if (mtag != NULL) {
844 			tdbi = (struct tdb_ident *)(mtag + 1);
845 			sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
846 		} else {
847 			sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
848 						   IP_FORWARDING, &error);
849 		}
850 		if (sp == NULL) {	/* NB: can happen if error */
851 			splx(s);
852 			/*XXX error stat???*/
853 			DPRINTF(("ip_input: no SP for forwarding\n"));	/*XXX*/
854 			goto bad;
855 		}
856 
857 		/*
858 		 * Check security policy against packet attributes.
859 		 */
860 		error = ipsec_in_reject(sp, m);
861 		KEY_FREESP(&sp);
862 		splx(s);
863 		if (error) {
864 			ipstat.ips_cantforward++;
865 			goto bad;
866 		}
867 #endif
868 		ip_forward(m, using_srcrt, args.next_hop);
869 	}
870 	return;
871 
872 ours:
873 
874 	/*
875 	 * IPSTEALTH: Process non-routing options only
876 	 * if the packet is destined for us.
877 	 */
878 	if (ipstealth &&
879 	    hlen > sizeof(struct ip) &&
880 	    ip_dooptions(m, 1, args.next_hop))
881 		return;
882 
883 	/* Count the packet in the ip address stats */
884 	if (ia != NULL) {
885 		ia->ia_ifa.if_ipackets++;
886 		ia->ia_ifa.if_ibytes += m->m_pkthdr.len;
887 	}
888 
889 	/*
890 	 * If offset or IP_MF are set, must reassemble.
891 	 * Otherwise, nothing need be done.
892 	 * (We could look in the reassembly queue to see
893 	 * if the packet was previously fragmented,
894 	 * but it's not worth the time; just let them time out.)
895 	 */
896 	if (ip->ip_off & (IP_MF | IP_OFFMASK)) {
897 
898 		/* If maxnipq is 0, never accept fragments. */
899 		if (maxnipq == 0) {
900 			ipstat.ips_fragments++;
901 			ipstat.ips_fragdropped++;
902 			goto bad;
903 		}
904 
905 		sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
906 		/*
907 		 * Look for queue of fragments
908 		 * of this datagram.
909 		 */
910 		for (fp = ipq[sum].next; fp != &ipq[sum]; fp = fp->next)
911 			if (ip->ip_id == fp->ipq_id &&
912 			    ip->ip_src.s_addr == fp->ipq_src.s_addr &&
913 			    ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
914 			    ip->ip_p == fp->ipq_p)
915 				goto found;
916 
917 		fp = NULL;
918 
919 		/*
920 		 * Enforce upper bound on number of fragmented packets
921 		 * for which we attempt reassembly;
922 		 * If maxnipq is -1, accept all fragments without limitation.
923 		 */
924 		if ((nipq > maxnipq) && (maxnipq > 0)) {
925 			/*
926 			 * drop something from the tail of the current queue
927 			 * before proceeding further
928 			 */
929 			if (ipq[sum].prev == &ipq[sum]) {   /* gak */
930 				for (i = 0; i < IPREASS_NHASH; i++) {
931 					if (ipq[i].prev != &ipq[i]) {
932 						ipstat.ips_fragtimeout +=
933 						    ipq[i].prev->ipq_nfrags;
934 						ip_freef(ipq[i].prev);
935 						break;
936 					}
937 				}
938 			} else {
939 				ipstat.ips_fragtimeout +=
940 				    ipq[sum].prev->ipq_nfrags;
941 				ip_freef(ipq[sum].prev);
942 			}
943 		}
944 found:
945 		/*
946 		 * Adjust ip_len to not reflect header,
947 		 * convert offset of this to bytes.
948 		 */
949 		ip->ip_len -= hlen;
950 		if (ip->ip_off & IP_MF) {
951 			/*
952 			 * Make sure that fragments have a data length
953 			 * that's a non-zero multiple of 8 bytes.
954 			 */
955 			if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
956 				ipstat.ips_toosmall++; /* XXX */
957 				goto bad;
958 			}
959 			m->m_flags |= M_FRAG;
960 		} else
961 			m->m_flags &= ~M_FRAG;
962 		ip->ip_off <<= 3;
963 
964 		/*
965 		 * Attempt reassembly; if it succeeds, proceed.
966 		 * ip_reass() will return a different mbuf, and update
967 		 * the divert info in divert_info and args.divert_rule.
968 		 */
969 		ipstat.ips_fragments++;
970 		m->m_pkthdr.header = ip;
971 		m = ip_reass(m, fp, &ipq[sum], &divert_info, &args.divert_rule);
972 		if (m == NULL)
973 			return;
974 		ipstat.ips_reassembled++;
975 		needredispatch = TRUE;
976 		ip = mtod(m, struct ip *);
977 		/* Get the header length of the reassembled packet */
978 		hlen = IP_VHL_HL(ip->ip_vhl) << 2;
979 #ifdef IPDIVERT
980 		/* Restore original checksum before diverting packet */
981 		if (divert_info != 0) {
982 			ip->ip_len += hlen;
983 			ip->ip_len = htons(ip->ip_len);
984 			ip->ip_off = htons(ip->ip_off);
985 			ip->ip_sum = 0;
986 			if (hlen == sizeof(struct ip))
987 				ip->ip_sum = in_cksum_hdr(ip);
988 			else
989 				ip->ip_sum = in_cksum(m, hlen);
990 			ip->ip_off = ntohs(ip->ip_off);
991 			ip->ip_len = ntohs(ip->ip_len);
992 			ip->ip_len -= hlen;
993 		}
994 #endif
995 	} else {
996 		ip->ip_len -= hlen;
997 	}
998 
999 #ifdef IPDIVERT
1000 	/*
1001 	 * Divert or tee packet to the divert protocol if required.
1002 	 */
1003 	if (divert_info != 0) {
1004 		struct mbuf *clone = NULL;
1005 
1006 		/* Clone packet if we're doing a 'tee' */
1007 		if ((divert_info & IP_FW_PORT_TEE_FLAG) != 0)
1008 			clone = m_dup(m, MB_DONTWAIT);
1009 
1010 		/* Restore packet header fields to original values */
1011 		ip->ip_len += hlen;
1012 		ip->ip_len = htons(ip->ip_len);
1013 		ip->ip_off = htons(ip->ip_off);
1014 
1015 		/* Deliver packet to divert input routine */
1016 		divert_packet(m, 1, divert_info & 0xffff, args.divert_rule);
1017 		ipstat.ips_delivered++;
1018 
1019 		/* If 'tee', continue with original packet */
1020 		if (clone == NULL)
1021 			return;
1022 		m = clone;
1023 		ip = mtod(m, struct ip *);
1024 		ip->ip_len += hlen;
1025 		/*
1026 		 * Jump backwards to complete processing of the
1027 		 * packet. But first clear divert_info to avoid
1028 		 * entering this block again.
1029 		 * We do not need to clear args.divert_rule
1030 		 * or args.next_hop as they will not be used.
1031 		 */
1032 		divert_info = 0;
1033 		goto pass;
1034 	}
1035 #endif
1036 
1037 #ifdef IPSEC
1038 	/*
1039 	 * enforce IPsec policy checking if we are seeing last header.
1040 	 * note that we do not visit this with protocols with pcb layer
1041 	 * code - like udp/tcp/raw ip.
1042 	 */
1043 	if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) &&
1044 	    ipsec4_in_reject(m, NULL)) {
1045 		ipsecstat.in_polvio++;
1046 		goto bad;
1047 	}
1048 #endif
1049 #if FAST_IPSEC
1050 	/*
1051 	 * enforce IPsec policy checking if we are seeing last header.
1052 	 * note that we do not visit this with protocols with pcb layer
1053 	 * code - like udp/tcp/raw ip.
1054 	 */
1055 	if (inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) {
1056 		/*
1057 		 * Check if the packet has already had IPsec processing
1058 		 * done.  If so, then just pass it along.  This tag gets
1059 		 * set during AH, ESP, etc. input handling, before the
1060 		 * packet is returned to the ip input queue for delivery.
1061 		 */
1062 		mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
1063 		s = splnet();
1064 		if (mtag != NULL) {
1065 			tdbi = (struct tdb_ident *)(mtag + 1);
1066 			sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
1067 		} else {
1068 			sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
1069 						   IP_FORWARDING, &error);
1070 		}
1071 		if (sp != NULL) {
1072 			/*
1073 			 * Check security policy against packet attributes.
1074 			 */
1075 			error = ipsec_in_reject(sp, m);
1076 			KEY_FREESP(&sp);
1077 		} else {
1078 			/* XXX error stat??? */
1079 			error = EINVAL;
1080 DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/
1081 			goto bad;
1082 		}
1083 		splx(s);
1084 		if (error)
1085 			goto bad;
1086 	}
1087 #endif /* FAST_IPSEC */
1088 
1089 	ipstat.ips_delivered++;
1090 	if (needredispatch) {
1091 		struct netmsg_transport_packet *msg;
1092 		lwkt_port_t port;
1093 
1094 		msg = malloc(sizeof(struct netmsg_transport_packet),
1095 				M_LWKTMSG, M_INTWAIT | M_NULLOK);
1096 		if (msg == NULL)
1097 			goto bad;
1098 
1099 		lwkt_initmsg(&msg->nm_lmsg, &netisr_afree_rport, 0,
1100 			lwkt_cmd_func(transport_processing_handler),
1101 			lwkt_cmd_op_none);
1102 		msg->nm_hlen = hlen;
1103 		msg->nm_hasnexthop = (args.next_hop != NULL);
1104 		if (msg->nm_hasnexthop)
1105 			msg->nm_nexthop = *args.next_hop;  /* structure copy */
1106 
1107 		ip->ip_off = htons(ip->ip_off);
1108 		ip->ip_len = htons(ip->ip_len);
1109 		port = ip_mport(&m);
1110 		if (port) {
1111 		    msg->nm_mbuf = m;
1112 		    ip = mtod(m, struct ip *);
1113 		    ip->ip_len = ntohs(ip->ip_len);
1114 		    ip->ip_off = ntohs(ip->ip_off);
1115 		    lwkt_sendmsg(port, &msg->nm_lmsg);
1116 		}
1117 	} else {
1118 		transport_processing_oncpu(m, hlen, ip, args.next_hop);
1119 	}
1120 	return;
1121 
1122 bad:
1123 	m_freem(m);
1124 }
1125 
1126 /*
1127  * Take incoming datagram fragment and try to reassemble it into
1128  * whole datagram.  If a chain for reassembly of this datagram already
1129  * exists, then it is given as fp; otherwise have to make a chain.
1130  *
1131  * When IPDIVERT enabled, keep additional state with each packet that
1132  * tells us if we need to divert or tee the packet we're building.
1133  * In particular, *divinfo includes the port and TEE flag,
1134  * *divert_rule is the number of the matching rule.
1135  */
1136 
1137 static struct mbuf *
1138 ip_reass(struct mbuf *m, struct ipq *fp, struct ipq *where,
1139 	 u_int32_t *divinfo, u_int16_t *divert_rule)
1140 {
1141 	struct ip *ip = mtod(m, struct ip *);
1142 	struct mbuf *p = NULL, *q, *nq;
1143 	struct mbuf *n;
1144 	int hlen = IP_VHL_HL(ip->ip_vhl) << 2;
1145 	int i, next;
1146 
1147 	/*
1148 	 * Presence of header sizes in mbufs
1149 	 * would confuse code below.
1150 	 */
1151 	m->m_data += hlen;
1152 	m->m_len -= hlen;
1153 
1154 	/*
1155 	 * If first fragment to arrive, create a reassembly queue.
1156 	 */
1157 	if (fp == NULL) {
1158 		if ((fp = mpipe_alloc_nowait(&ipq_mpipe)) == NULL)
1159 			goto dropfrag;
1160 		insque(fp, where);
1161 		nipq++;
1162 		fp->ipq_nfrags = 1;
1163 		fp->ipq_ttl = IPFRAGTTL;
1164 		fp->ipq_p = ip->ip_p;
1165 		fp->ipq_id = ip->ip_id;
1166 		fp->ipq_src = ip->ip_src;
1167 		fp->ipq_dst = ip->ip_dst;
1168 		fp->ipq_frags = m;
1169 		m->m_nextpkt = NULL;
1170 #ifdef IPDIVERT
1171 		fp->ipq_div_info = 0;
1172 		fp->ipq_div_cookie = 0;
1173 #endif
1174 		goto inserted;
1175 	} else {
1176 		fp->ipq_nfrags++;
1177 	}
1178 
1179 #define	GETIP(m)	((struct ip*)((m)->m_pkthdr.header))
1180 
1181 	/*
1182 	 * Find a segment which begins after this one does.
1183 	 */
1184 	for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt)
1185 		if (GETIP(q)->ip_off > ip->ip_off)
1186 			break;
1187 
1188 	/*
1189 	 * If there is a preceding segment, it may provide some of
1190 	 * our data already.  If so, drop the data from the incoming
1191 	 * segment.  If it provides all of our data, drop us, otherwise
1192 	 * stick new segment in the proper place.
1193 	 *
1194 	 * If some of the data is dropped from the the preceding
1195 	 * segment, then it's checksum is invalidated.
1196 	 */
1197 	if (p) {
1198 		i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off;
1199 		if (i > 0) {
1200 			if (i >= ip->ip_len)
1201 				goto dropfrag;
1202 			m_adj(m, i);
1203 			m->m_pkthdr.csum_flags = 0;
1204 			ip->ip_off += i;
1205 			ip->ip_len -= i;
1206 		}
1207 		m->m_nextpkt = p->m_nextpkt;
1208 		p->m_nextpkt = m;
1209 	} else {
1210 		m->m_nextpkt = fp->ipq_frags;
1211 		fp->ipq_frags = m;
1212 	}
1213 
1214 	/*
1215 	 * While we overlap succeeding segments trim them or,
1216 	 * if they are completely covered, dequeue them.
1217 	 */
1218 	for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off;
1219 	     q = nq) {
1220 		i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off;
1221 		if (i < GETIP(q)->ip_len) {
1222 			GETIP(q)->ip_len -= i;
1223 			GETIP(q)->ip_off += i;
1224 			m_adj(q, i);
1225 			q->m_pkthdr.csum_flags = 0;
1226 			break;
1227 		}
1228 		nq = q->m_nextpkt;
1229 		m->m_nextpkt = nq;
1230 		ipstat.ips_fragdropped++;
1231 		fp->ipq_nfrags--;
1232 		m_freem(q);
1233 	}
1234 
1235 inserted:
1236 
1237 #ifdef IPDIVERT
1238 	/*
1239 	 * Transfer firewall instructions to the fragment structure.
1240 	 * Only trust info in the fragment at offset 0.
1241 	 */
1242 	if (ip->ip_off == 0) {
1243 		fp->ipq_div_info = *divinfo;
1244 		fp->ipq_div_cookie = *divert_rule;
1245 	}
1246 	*divinfo = 0;
1247 	*divert_rule = 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 #ifdef IPDIVERT
1310 	/*
1311 	 * Extract firewall instructions from the fragment structure.
1312 	 */
1313 	*divinfo = fp->ipq_div_info;
1314 	*divert_rule = fp->ipq_div_cookie;
1315 #endif
1316 
1317 	/*
1318 	 * Create header for new ip packet by
1319 	 * modifying header of first packet;
1320 	 * dequeue and discard fragment reassembly header.
1321 	 * Make header visible.
1322 	 */
1323 	ip->ip_len = next;
1324 	ip->ip_src = fp->ipq_src;
1325 	ip->ip_dst = fp->ipq_dst;
1326 	remque(fp);
1327 	nipq--;
1328 	mpipe_free(&ipq_mpipe, fp);
1329 	m->m_len += (IP_VHL_HL(ip->ip_vhl) << 2);
1330 	m->m_data -= (IP_VHL_HL(ip->ip_vhl) << 2);
1331 	/* some debugging cruft by sklower, below, will go away soon */
1332 	if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */
1333 		int plen = 0;
1334 
1335 		for (n = m; n; n = n->m_next)
1336 			plen += n->m_len;
1337 		m->m_pkthdr.len = plen;
1338 	}
1339 	return (m);
1340 
1341 dropfrag:
1342 #ifdef IPDIVERT
1343 	*divinfo = 0;
1344 	*divert_rule = 0;
1345 #endif
1346 	ipstat.ips_fragdropped++;
1347 	if (fp != NULL)
1348 		fp->ipq_nfrags--;
1349 	m_freem(m);
1350 	return (NULL);
1351 
1352 #undef GETIP
1353 }
1354 
1355 /*
1356  * Free a fragment reassembly header and all
1357  * associated datagrams.
1358  */
1359 static void
1360 ip_freef(struct ipq *fp)
1361 {
1362 	struct mbuf *q;
1363 
1364 	while (fp->ipq_frags) {
1365 		q = fp->ipq_frags;
1366 		fp->ipq_frags = q->m_nextpkt;
1367 		m_freem(q);
1368 	}
1369 	remque(fp);
1370 	mpipe_free(&ipq_mpipe, fp);
1371 	nipq--;
1372 }
1373 
1374 /*
1375  * IP timer processing;
1376  * if a timer expires on a reassembly
1377  * queue, discard it.
1378  */
1379 void
1380 ip_slowtimo(void)
1381 {
1382 	struct ipq *fp;
1383 	int s = splnet();
1384 	int i;
1385 
1386 	for (i = 0; i < IPREASS_NHASH; i++) {
1387 		fp = ipq[i].next;
1388 		if (fp == NULL)
1389 			continue;
1390 		while (fp != &ipq[i]) {
1391 			--fp->ipq_ttl;
1392 			fp = fp->next;
1393 			if (fp->prev->ipq_ttl == 0) {
1394 				ipstat.ips_fragtimeout += fp->prev->ipq_nfrags;
1395 				ip_freef(fp->prev);
1396 			}
1397 		}
1398 	}
1399 	/*
1400 	 * If we are over the maximum number of fragments
1401 	 * (due to the limit being lowered), drain off
1402 	 * enough to get down to the new limit.
1403 	 */
1404 	if (maxnipq >= 0 && nipq > maxnipq) {
1405 		for (i = 0; i < IPREASS_NHASH; i++) {
1406 			while (nipq > maxnipq &&
1407 				(ipq[i].next != &ipq[i])) {
1408 				ipstat.ips_fragdropped +=
1409 				    ipq[i].next->ipq_nfrags;
1410 				ip_freef(ipq[i].next);
1411 			}
1412 		}
1413 	}
1414 	ipflow_slowtimo();
1415 	splx(s);
1416 }
1417 
1418 /*
1419  * Drain off all datagram fragments.
1420  */
1421 void
1422 ip_drain(void)
1423 {
1424 	int i;
1425 
1426 	for (i = 0; i < IPREASS_NHASH; i++) {
1427 		while (ipq[i].next != &ipq[i]) {
1428 			ipstat.ips_fragdropped += ipq[i].next->ipq_nfrags;
1429 			ip_freef(ipq[i].next);
1430 		}
1431 	}
1432 	in_rtqdrain();
1433 }
1434 
1435 /*
1436  * Do option processing on a datagram,
1437  * possibly discarding it if bad options are encountered,
1438  * or forwarding it if source-routed.
1439  * The pass argument is used when operating in the IPSTEALTH
1440  * mode to tell what options to process:
1441  * [LS]SRR (pass 0) or the others (pass 1).
1442  * The reason for as many as two passes is that when doing IPSTEALTH,
1443  * non-routing options should be processed only if the packet is for us.
1444  * Returns 1 if packet has been forwarded/freed,
1445  * 0 if the packet should be processed further.
1446  */
1447 static int
1448 ip_dooptions(struct mbuf *m, int pass, struct sockaddr_in *next_hop)
1449 {
1450 	struct sockaddr_in ipaddr = { sizeof ipaddr, AF_INET };
1451 	struct ip *ip = mtod(m, struct ip *);
1452 	u_char *cp;
1453 	struct in_ifaddr *ia;
1454 	int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB;
1455 	boolean_t forward = FALSE;
1456 	struct in_addr *sin, dst;
1457 	n_time ntime;
1458 
1459 	dst = ip->ip_dst;
1460 	cp = (u_char *)(ip + 1);
1461 	cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
1462 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1463 		opt = cp[IPOPT_OPTVAL];
1464 		if (opt == IPOPT_EOL)
1465 			break;
1466 		if (opt == IPOPT_NOP)
1467 			optlen = 1;
1468 		else {
1469 			if (cnt < IPOPT_OLEN + sizeof(*cp)) {
1470 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1471 				goto bad;
1472 			}
1473 			optlen = cp[IPOPT_OLEN];
1474 			if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
1475 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1476 				goto bad;
1477 			}
1478 		}
1479 		switch (opt) {
1480 
1481 		default:
1482 			break;
1483 
1484 		/*
1485 		 * Source routing with record.
1486 		 * Find interface with current destination address.
1487 		 * If none on this machine then drop if strictly routed,
1488 		 * or do nothing if loosely routed.
1489 		 * Record interface address and bring up next address
1490 		 * component.  If strictly routed make sure next
1491 		 * address is on directly accessible net.
1492 		 */
1493 		case IPOPT_LSRR:
1494 		case IPOPT_SSRR:
1495 			if (ipstealth && pass > 0)
1496 				break;
1497 			if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1498 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1499 				goto bad;
1500 			}
1501 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1502 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1503 				goto bad;
1504 			}
1505 			ipaddr.sin_addr = ip->ip_dst;
1506 			ia = (struct in_ifaddr *)
1507 				ifa_ifwithaddr((struct sockaddr *)&ipaddr);
1508 			if (ia == NULL) {
1509 				if (opt == IPOPT_SSRR) {
1510 					type = ICMP_UNREACH;
1511 					code = ICMP_UNREACH_SRCFAIL;
1512 					goto bad;
1513 				}
1514 				if (!ip_dosourceroute)
1515 					goto nosourcerouting;
1516 				/*
1517 				 * Loose routing, and not at next destination
1518 				 * yet; nothing to do except forward.
1519 				 */
1520 				break;
1521 			}
1522 			off--;			/* 0 origin */
1523 			if (off > optlen - (int)sizeof(struct in_addr)) {
1524 				/*
1525 				 * End of source route.  Should be for us.
1526 				 */
1527 				if (!ip_acceptsourceroute)
1528 					goto nosourcerouting;
1529 				save_rte(cp, ip->ip_src);
1530 				break;
1531 			}
1532 			if (ipstealth)
1533 				goto dropit;
1534 			if (!ip_dosourceroute) {
1535 				if (ipforwarding) {
1536 					char buf[sizeof "aaa.bbb.ccc.ddd"];
1537 
1538 					/*
1539 					 * Acting as a router, so generate ICMP
1540 					 */
1541 nosourcerouting:
1542 					strcpy(buf, inet_ntoa(ip->ip_dst));
1543 					log(LOG_WARNING,
1544 					    "attempted source route from %s to %s\n",
1545 					    inet_ntoa(ip->ip_src), buf);
1546 					type = ICMP_UNREACH;
1547 					code = ICMP_UNREACH_SRCFAIL;
1548 					goto bad;
1549 				} else {
1550 					/*
1551 					 * Not acting as a router,
1552 					 * so silently drop.
1553 					 */
1554 dropit:
1555 					ipstat.ips_cantforward++;
1556 					m_freem(m);
1557 					return (1);
1558 				}
1559 			}
1560 
1561 			/*
1562 			 * locate outgoing interface
1563 			 */
1564 			memcpy(&ipaddr.sin_addr, cp + off,
1565 			    sizeof ipaddr.sin_addr);
1566 
1567 			if (opt == IPOPT_SSRR) {
1568 #define	INA	struct in_ifaddr *
1569 #define	SA	struct sockaddr *
1570 				if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr))
1571 									== NULL)
1572 					ia = (INA)ifa_ifwithnet((SA)&ipaddr);
1573 			} else
1574 				ia = ip_rtaddr(ipaddr.sin_addr, &ipforward_rt);
1575 			if (ia == NULL) {
1576 				type = ICMP_UNREACH;
1577 				code = ICMP_UNREACH_SRCFAIL;
1578 				goto bad;
1579 			}
1580 			ip->ip_dst = ipaddr.sin_addr;
1581 			memcpy(cp + off, &IA_SIN(ia)->sin_addr,
1582 			    sizeof(struct in_addr));
1583 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1584 			/*
1585 			 * Let ip_intr's mcast routing check handle mcast pkts
1586 			 */
1587 			forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr));
1588 			break;
1589 
1590 		case IPOPT_RR:
1591 			if (ipstealth && pass == 0)
1592 				break;
1593 			if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1594 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1595 				goto bad;
1596 			}
1597 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1598 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1599 				goto bad;
1600 			}
1601 			/*
1602 			 * If no space remains, ignore.
1603 			 */
1604 			off--;			/* 0 origin */
1605 			if (off > optlen - (int)sizeof(struct in_addr))
1606 				break;
1607 			memcpy(&ipaddr.sin_addr, &ip->ip_dst,
1608 			    sizeof ipaddr.sin_addr);
1609 			/*
1610 			 * locate outgoing interface; if we're the destination,
1611 			 * use the incoming interface (should be same).
1612 			 */
1613 			if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == NULL &&
1614 			    (ia = ip_rtaddr(ipaddr.sin_addr, &ipforward_rt))
1615 								     == NULL) {
1616 				type = ICMP_UNREACH;
1617 				code = ICMP_UNREACH_HOST;
1618 				goto bad;
1619 			}
1620 			memcpy(cp + off, &IA_SIN(ia)->sin_addr,
1621 			    sizeof(struct in_addr));
1622 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1623 			break;
1624 
1625 		case IPOPT_TS:
1626 			if (ipstealth && pass == 0)
1627 				break;
1628 			code = cp - (u_char *)ip;
1629 			if (optlen < 4 || optlen > 40) {
1630 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1631 				goto bad;
1632 			}
1633 			if ((off = cp[IPOPT_OFFSET]) < 5) {
1634 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1635 				goto bad;
1636 			}
1637 			if (off > optlen - (int)sizeof(int32_t)) {
1638 				cp[IPOPT_OFFSET + 1] += (1 << 4);
1639 				if ((cp[IPOPT_OFFSET + 1] & 0xf0) == 0) {
1640 					code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1641 					goto bad;
1642 				}
1643 				break;
1644 			}
1645 			off--;				/* 0 origin */
1646 			sin = (struct in_addr *)(cp + off);
1647 			switch (cp[IPOPT_OFFSET + 1] & 0x0f) {
1648 
1649 			case IPOPT_TS_TSONLY:
1650 				break;
1651 
1652 			case IPOPT_TS_TSANDADDR:
1653 				if (off + sizeof(n_time) +
1654 				    sizeof(struct in_addr) > optlen) {
1655 					code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1656 					goto bad;
1657 				}
1658 				ipaddr.sin_addr = dst;
1659 				ia = (INA)ifaof_ifpforaddr((SA)&ipaddr,
1660 							    m->m_pkthdr.rcvif);
1661 				if (ia == NULL)
1662 					continue;
1663 				memcpy(sin, &IA_SIN(ia)->sin_addr,
1664 				    sizeof(struct in_addr));
1665 				cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1666 				off += sizeof(struct in_addr);
1667 				break;
1668 
1669 			case IPOPT_TS_PRESPEC:
1670 				if (off + sizeof(n_time) +
1671 				    sizeof(struct in_addr) > optlen) {
1672 					code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1673 					goto bad;
1674 				}
1675 				memcpy(&ipaddr.sin_addr, sin,
1676 				    sizeof(struct in_addr));
1677 				if (ifa_ifwithaddr((SA)&ipaddr) == NULL)
1678 					continue;
1679 				cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1680 				off += sizeof(struct in_addr);
1681 				break;
1682 
1683 			default:
1684 				code = &cp[IPOPT_OFFSET + 1] - (u_char *)ip;
1685 				goto bad;
1686 			}
1687 			ntime = iptime();
1688 			memcpy(cp + off, &ntime, sizeof(n_time));
1689 			cp[IPOPT_OFFSET] += sizeof(n_time);
1690 		}
1691 	}
1692 	if (forward && ipforwarding) {
1693 		ip_forward(m, 1, next_hop);
1694 		return (1);
1695 	}
1696 	return (0);
1697 bad:
1698 	icmp_error(m, type, code, 0, NULL);
1699 	ipstat.ips_badoptions++;
1700 	return (1);
1701 }
1702 
1703 /*
1704  * Given address of next destination (final or next hop),
1705  * return internet address info of interface to be used to get there.
1706  */
1707 struct in_ifaddr *
1708 ip_rtaddr(struct in_addr dst, struct route *rt)
1709 {
1710 	struct sockaddr_in *sin;
1711 
1712 	sin = (struct sockaddr_in *)&rt->ro_dst;
1713 
1714 	if (rt->ro_rt == NULL || dst.s_addr != sin->sin_addr.s_addr) {
1715 		if (rt->ro_rt != NULL) {
1716 			RTFREE(rt->ro_rt);
1717 			rt->ro_rt = NULL;
1718 		}
1719 		sin->sin_family = AF_INET;
1720 		sin->sin_len = sizeof *sin;
1721 		sin->sin_addr = dst;
1722 		rtalloc_ign(rt, RTF_PRCLONING);
1723 	}
1724 
1725 	if (rt->ro_rt == NULL)
1726 		return (NULL);
1727 
1728 	return (ifatoia(rt->ro_rt->rt_ifa));
1729 }
1730 
1731 /*
1732  * Save incoming source route for use in replies,
1733  * to be picked up later by ip_srcroute if the receiver is interested.
1734  */
1735 void
1736 save_rte(u_char *option, struct in_addr dst)
1737 {
1738 	unsigned olen;
1739 
1740 	olen = option[IPOPT_OLEN];
1741 #ifdef DIAGNOSTIC
1742 	if (ipprintfs)
1743 		printf("save_rte: olen %d\n", olen);
1744 #endif
1745 	if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst)))
1746 		return;
1747 	bcopy(option, ip_srcrt.srcopt, olen);
1748 	ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1749 	ip_srcrt.dst = dst;
1750 }
1751 
1752 /*
1753  * Retrieve incoming source route for use in replies,
1754  * in the same form used by setsockopt.
1755  * The first hop is placed before the options, will be removed later.
1756  */
1757 struct mbuf *
1758 ip_srcroute(void)
1759 {
1760 	struct in_addr *p, *q;
1761 	struct mbuf *m;
1762 
1763 	if (ip_nhops == 0)
1764 		return (NULL);
1765 	m = m_get(MB_DONTWAIT, MT_HEADER);
1766 	if (m == NULL)
1767 		return (NULL);
1768 
1769 #define	OPTSIZ	(sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt))
1770 
1771 	/* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
1772 	m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) +
1773 	    OPTSIZ;
1774 #ifdef DIAGNOSTIC
1775 	if (ipprintfs)
1776 		printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len);
1777 #endif
1778 
1779 	/*
1780 	 * First save first hop for return route
1781 	 */
1782 	p = &ip_srcrt.route[ip_nhops - 1];
1783 	*(mtod(m, struct in_addr *)) = *p--;
1784 #ifdef DIAGNOSTIC
1785 	if (ipprintfs)
1786 		printf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr));
1787 #endif
1788 
1789 	/*
1790 	 * Copy option fields and padding (nop) to mbuf.
1791 	 */
1792 	ip_srcrt.nop = IPOPT_NOP;
1793 	ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
1794 	memcpy(mtod(m, caddr_t) + sizeof(struct in_addr), &ip_srcrt.nop,
1795 	    OPTSIZ);
1796 	q = (struct in_addr *)(mtod(m, caddr_t) +
1797 	    sizeof(struct in_addr) + OPTSIZ);
1798 #undef OPTSIZ
1799 	/*
1800 	 * Record return path as an IP source route,
1801 	 * reversing the path (pointers are now aligned).
1802 	 */
1803 	while (p >= ip_srcrt.route) {
1804 #ifdef DIAGNOSTIC
1805 		if (ipprintfs)
1806 			printf(" %x", ntohl(q->s_addr));
1807 #endif
1808 		*q++ = *p--;
1809 	}
1810 	/*
1811 	 * Last hop goes to final destination.
1812 	 */
1813 	*q = ip_srcrt.dst;
1814 #ifdef DIAGNOSTIC
1815 	if (ipprintfs)
1816 		printf(" %x\n", ntohl(q->s_addr));
1817 #endif
1818 	return (m);
1819 }
1820 
1821 /*
1822  * Strip out IP options.
1823  */
1824 void
1825 ip_stripoptions(struct mbuf *m)
1826 {
1827 	int datalen;
1828 	struct ip *ip = mtod(m, struct ip *);
1829 	caddr_t opts;
1830 	int optlen;
1831 
1832 	optlen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
1833 	opts = (caddr_t)(ip + 1);
1834 	datalen = m->m_len - (sizeof(struct ip) + optlen);
1835 	bcopy(opts + optlen, opts, datalen);
1836 	m->m_len -= optlen;
1837 	if (m->m_flags & M_PKTHDR)
1838 		m->m_pkthdr.len -= optlen;
1839 	ip->ip_vhl = IP_MAKE_VHL(IPVERSION, sizeof(struct ip) >> 2);
1840 }
1841 
1842 u_char inetctlerrmap[PRC_NCMDS] = {
1843 	0,		0,		0,		0,
1844 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1845 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1846 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1847 	0,		0,		0,		0,
1848 	ENOPROTOOPT,	ECONNREFUSED
1849 };
1850 
1851 /*
1852  * Forward a packet.  If some error occurs return the sender
1853  * an icmp packet.  Note we can't always generate a meaningful
1854  * icmp message because icmp doesn't have a large enough repertoire
1855  * of codes and types.
1856  *
1857  * If not forwarding, just drop the packet.  This could be confusing
1858  * if ipforwarding was zero but some routing protocol was advancing
1859  * us as a gateway to somewhere.  However, we must let the routing
1860  * protocol deal with that.
1861  *
1862  * The using_srcrt parameter indicates whether the packet is being forwarded
1863  * via a source route.
1864  */
1865 static void
1866 ip_forward(struct mbuf *m, int using_srcrt, struct sockaddr_in *next_hop)
1867 {
1868 	struct ip *ip = mtod(m, struct ip *);
1869 	struct sockaddr_in *sin;
1870 	struct rtentry *rt;
1871 	int error, type = 0, code = 0;
1872 	struct mbuf *mcopy;
1873 	n_long dest;
1874 	struct in_addr pkt_dst;
1875 	struct ifnet *destifp;
1876 	struct m_hdr tag;
1877 #if defined(IPSEC) || defined(FAST_IPSEC)
1878 	struct ifnet dummyifp;
1879 #endif
1880 
1881 	dest = 0;
1882 	/*
1883 	 * Cache the destination address of the packet; this may be
1884 	 * changed by use of 'ipfw fwd'.
1885 	 */
1886 	pkt_dst = (next_hop != NULL) ? next_hop->sin_addr : ip->ip_dst;
1887 
1888 #ifdef DIAGNOSTIC
1889 	if (ipprintfs)
1890 		printf("forward: src %x dst %x ttl %x\n",
1891 		       ip->ip_src.s_addr, pkt_dst.s_addr, ip->ip_ttl);
1892 #endif
1893 
1894 	if (m->m_flags & (M_BCAST | M_MCAST) || !in_canforward(pkt_dst)) {
1895 		ipstat.ips_cantforward++;
1896 		m_freem(m);
1897 		return;
1898 	}
1899 	if (!ipstealth && ip->ip_ttl <= IPTTLDEC) {
1900 		icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, NULL);
1901 		return;
1902 	}
1903 
1904 	sin = (struct sockaddr_in *)&ipforward_rt.ro_dst;
1905 	if ((rt = ipforward_rt.ro_rt) == NULL ||
1906 	    pkt_dst.s_addr != sin->sin_addr.s_addr) {
1907 		if (ipforward_rt.ro_rt != NULL) {
1908 			RTFREE(ipforward_rt.ro_rt);
1909 			ipforward_rt.ro_rt = NULL;
1910 		}
1911 		sin->sin_family = AF_INET;
1912 		sin->sin_len = sizeof(*sin);
1913 		sin->sin_addr = pkt_dst;
1914 
1915 		rtalloc_ign(&ipforward_rt, RTF_PRCLONING);
1916 		if (ipforward_rt.ro_rt == NULL) {
1917 			icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest,
1918 				   NULL);
1919 			return;
1920 		}
1921 		rt = ipforward_rt.ro_rt;
1922 	}
1923 
1924 	/*
1925 	 * Save the IP header and at most 8 bytes of the payload,
1926 	 * in case we need to generate an ICMP message to the src.
1927 	 *
1928 	 * XXX this can be optimized a lot by saving the data in a local
1929 	 * buffer on the stack (72 bytes at most), and only allocating the
1930 	 * mbuf if really necessary. The vast majority of the packets
1931 	 * are forwarded without having to send an ICMP back (either
1932 	 * because unnecessary, or because rate limited), so we are
1933 	 * really we are wasting a lot of work here.
1934 	 *
1935 	 * We don't use m_copy() because it might return a reference
1936 	 * to a shared cluster. Both this function and ip_output()
1937 	 * assume exclusive access to the IP header in `m', so any
1938 	 * data in a cluster may change before we reach icmp_error().
1939 	 */
1940 	MGET(mcopy, MB_DONTWAIT, m->m_type);
1941 	if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, MB_DONTWAIT)) {
1942 		/*
1943 		 * It's probably ok if the pkthdr dup fails (because
1944 		 * the deep copy of the tag chain failed), but for now
1945 		 * be conservative and just discard the copy since
1946 		 * code below may some day want the tags.
1947 		 */
1948 		m_free(mcopy);
1949 		mcopy = NULL;
1950 	}
1951 	if (mcopy != NULL) {
1952 		mcopy->m_len = imin((IP_VHL_HL(ip->ip_vhl) << 2) + 8,
1953 		    (int)ip->ip_len);
1954 		mcopy->m_pkthdr.len = mcopy->m_len;
1955 		m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1956 	}
1957 
1958 	if (!ipstealth)
1959 		ip->ip_ttl -= IPTTLDEC;
1960 
1961 	/*
1962 	 * If forwarding packet using same interface that it came in on,
1963 	 * perhaps should send a redirect to sender to shortcut a hop.
1964 	 * Only send redirect if source is sending directly to us,
1965 	 * and if packet was not source routed (or has any options).
1966 	 * Also, don't send redirect if forwarding using a default route
1967 	 * or a route modified by a redirect.
1968 	 */
1969 	if (rt->rt_ifp == m->m_pkthdr.rcvif &&
1970 	    !(rt->rt_flags & (RTF_DYNAMIC | RTF_MODIFIED)) &&
1971 	    satosin(rt_key(rt))->sin_addr.s_addr != INADDR_ANY &&
1972 	    ipsendredirects && !using_srcrt && next_hop == NULL) {
1973 		u_long src = ntohl(ip->ip_src.s_addr);
1974 
1975 #define	RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa))
1976 		if (RTA(rt) != NULL &&
1977 		    (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1978 			if (rt->rt_flags & RTF_GATEWAY)
1979 				dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
1980 			else
1981 				dest = pkt_dst.s_addr;
1982 			/*
1983 			 * Router requirements says to only send
1984 			 * host redirects.
1985 			 */
1986 			type = ICMP_REDIRECT;
1987 			code = ICMP_REDIRECT_HOST;
1988 #ifdef DIAGNOSTIC
1989 			if (ipprintfs)
1990 				printf("redirect (%d) to %x\n", code, dest);
1991 #endif
1992 		}
1993 	}
1994 
1995 	if (next_hop != NULL) {
1996 		/* Pass IPFORWARD info if available */
1997 		tag.mh_type = MT_TAG;
1998 		tag.mh_flags = PACKET_TAG_IPFORWARD;
1999 		tag.mh_data = (caddr_t)next_hop;
2000 		tag.mh_next = m;
2001 		m = (struct mbuf *)&tag;
2002 	}
2003 
2004 	error = ip_output(m, NULL, &ipforward_rt, IP_FORWARDING, NULL, NULL);
2005 
2006 	if (error)
2007 		ipstat.ips_cantforward++;
2008 	else {
2009 		ipstat.ips_forward++;
2010 		if (type)
2011 			ipstat.ips_redirectsent++;
2012 		else {
2013 			if (mcopy) {
2014 				ipflow_create(&ipforward_rt, mcopy);
2015 				m_freem(mcopy);
2016 			}
2017 			return;
2018 		}
2019 	}
2020 	if (mcopy == NULL)
2021 		return;
2022 	destifp = NULL;
2023 
2024 	switch (error) {
2025 
2026 	case 0:				/* forwarded, but need redirect */
2027 		/* type, code set above */
2028 		break;
2029 
2030 	case ENETUNREACH:		/* shouldn't happen, checked above */
2031 	case EHOSTUNREACH:
2032 	case ENETDOWN:
2033 	case EHOSTDOWN:
2034 	default:
2035 		type = ICMP_UNREACH;
2036 		code = ICMP_UNREACH_HOST;
2037 		break;
2038 
2039 	case EMSGSIZE:
2040 		type = ICMP_UNREACH;
2041 		code = ICMP_UNREACH_NEEDFRAG;
2042 #ifdef IPSEC
2043 		/*
2044 		 * If the packet is routed over IPsec tunnel, tell the
2045 		 * originator the tunnel MTU.
2046 		 *	tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
2047 		 * XXX quickhack!!!
2048 		 */
2049 		if (ipforward_rt.ro_rt != NULL) {
2050 			struct secpolicy *sp = NULL;
2051 			int ipsecerror;
2052 			int ipsechdr;
2053 			struct route *ro;
2054 
2055 			sp = ipsec4_getpolicybyaddr(mcopy,
2056 						    IPSEC_DIR_OUTBOUND,
2057 						    IP_FORWARDING,
2058 						    &ipsecerror);
2059 
2060 			if (sp == NULL)
2061 				destifp = ipforward_rt.ro_rt->rt_ifp;
2062 			else {
2063 				/* count IPsec header size */
2064 				ipsechdr = ipsec4_hdrsiz(mcopy,
2065 							 IPSEC_DIR_OUTBOUND,
2066 							 NULL);
2067 
2068 				/*
2069 				 * find the correct route for outer IPv4
2070 				 * header, compute tunnel MTU.
2071 				 *
2072 				 * XXX BUG ALERT
2073 				 * The "dummyifp" code relies upon the fact
2074 				 * that icmp_error() touches only ifp->if_mtu.
2075 				 */
2076 				/*XXX*/
2077 				destifp = NULL;
2078 				if (sp->req != NULL && sp->req->sav != NULL &&
2079 				    sp->req->sav->sah != NULL) {
2080 					ro = &sp->req->sav->sah->sa_route;
2081 					if (ro->ro_rt != NULL &&
2082 					    ro->ro_rt->rt_ifp != NULL) {
2083 						dummyifp.if_mtu =
2084 						    ro->ro_rt->rt_ifp->if_mtu;
2085 						dummyifp.if_mtu -= ipsechdr;
2086 						destifp = &dummyifp;
2087 					}
2088 				}
2089 
2090 				key_freesp(sp);
2091 			}
2092 		}
2093 #elif FAST_IPSEC
2094 		/*
2095 		 * If the packet is routed over IPsec tunnel, tell the
2096 		 * originator the tunnel MTU.
2097 		 *	tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
2098 		 * XXX quickhack!!!
2099 		 */
2100 		if (ipforward_rt.ro_rt != NULL) {
2101 			struct secpolicy *sp = NULL;
2102 			int ipsecerror;
2103 			int ipsechdr;
2104 			struct route *ro;
2105 
2106 			sp = ipsec_getpolicybyaddr(mcopy,
2107 						   IPSEC_DIR_OUTBOUND,
2108 						   IP_FORWARDING,
2109 						   &ipsecerror);
2110 
2111 			if (sp == NULL)
2112 				destifp = ipforward_rt.ro_rt->rt_ifp;
2113 			else {
2114 				/* count IPsec header size */
2115 				ipsechdr = ipsec4_hdrsiz(mcopy,
2116 							 IPSEC_DIR_OUTBOUND,
2117 							 NULL);
2118 
2119 				/*
2120 				 * find the correct route for outer IPv4
2121 				 * header, compute tunnel MTU.
2122 				 *
2123 				 * XXX BUG ALERT
2124 				 * The "dummyifp" code relies upon the fact
2125 				 * that icmp_error() touches only ifp->if_mtu.
2126 				 */
2127 				/*XXX*/
2128 				destifp = NULL;
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 						dummyifp.if_mtu =
2136 						    ro->ro_rt->rt_ifp->if_mtu;
2137 						dummyifp.if_mtu -= ipsechdr;
2138 						destifp = &dummyifp;
2139 					}
2140 				}
2141 
2142 				KEY_FREESP(&sp);
2143 			}
2144 		}
2145 #else /* !IPSEC && !FAST_IPSEC */
2146 		if (ipforward_rt.ro_rt != NULL)
2147 			destifp = ipforward_rt.ro_rt->rt_ifp;
2148 #endif /*IPSEC*/
2149 		ipstat.ips_cantfrag++;
2150 		break;
2151 
2152 	case ENOBUFS:
2153 		/*
2154 		 * A router should not generate ICMP_SOURCEQUENCH as
2155 		 * required in RFC1812 Requirements for IP Version 4 Routers.
2156 		 * Source quench could be a big problem under DoS attacks,
2157 		 * or if the underlying interface is rate-limited.
2158 		 * Those who need source quench packets may re-enable them
2159 		 * via the net.inet.ip.sendsourcequench sysctl.
2160 		 */
2161 		if (!ip_sendsourcequench) {
2162 			m_freem(mcopy);
2163 			return;
2164 		} else {
2165 			type = ICMP_SOURCEQUENCH;
2166 			code = 0;
2167 		}
2168 		break;
2169 
2170 	case EACCES:			/* ipfw denied packet */
2171 		m_freem(mcopy);
2172 		return;
2173 	}
2174 	icmp_error(mcopy, type, code, dest, destifp);
2175 }
2176 
2177 void
2178 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
2179 	       struct mbuf *m)
2180 {
2181 	if (inp->inp_socket->so_options & SO_TIMESTAMP) {
2182 		struct timeval tv;
2183 
2184 		microtime(&tv);
2185 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
2186 		    SCM_TIMESTAMP, SOL_SOCKET);
2187 		if (*mp)
2188 			mp = &(*mp)->m_next;
2189 	}
2190 	if (inp->inp_flags & INP_RECVDSTADDR) {
2191 		*mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
2192 		    sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
2193 		if (*mp)
2194 			mp = &(*mp)->m_next;
2195 	}
2196 #ifdef notyet
2197 	/* XXX
2198 	 * Moving these out of udp_input() made them even more broken
2199 	 * than they already were.
2200 	 */
2201 	/* options were tossed already */
2202 	if (inp->inp_flags & INP_RECVOPTS) {
2203 		*mp = sbcreatecontrol((caddr_t) opts_deleted_above,
2204 		    sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
2205 		if (*mp)
2206 			mp = &(*mp)->m_next;
2207 	}
2208 	/* ip_srcroute doesn't do what we want here, need to fix */
2209 	if (inp->inp_flags & INP_RECVRETOPTS) {
2210 		*mp = sbcreatecontrol((caddr_t) ip_srcroute(),
2211 		    sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
2212 		if (*mp)
2213 			mp = &(*mp)->m_next;
2214 	}
2215 #endif
2216 	if (inp->inp_flags & INP_RECVIF) {
2217 		struct ifnet *ifp;
2218 		struct sdlbuf {
2219 			struct sockaddr_dl sdl;
2220 			u_char	pad[32];
2221 		} sdlbuf;
2222 		struct sockaddr_dl *sdp;
2223 		struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
2224 
2225 		if (((ifp = m->m_pkthdr.rcvif)) &&
2226 		    ((ifp->if_index != 0) && (ifp->if_index <= if_index))) {
2227 			sdp = (struct sockaddr_dl *)
2228 			    ifnet_addrs[ifp->if_index - 1]->ifa_addr;
2229 			/*
2230 			 * Change our mind and don't try copy.
2231 			 */
2232 			if ((sdp->sdl_family != AF_LINK) ||
2233 			    (sdp->sdl_len > sizeof(sdlbuf))) {
2234 				goto makedummy;
2235 			}
2236 			bcopy(sdp, sdl2, sdp->sdl_len);
2237 		} else {
2238 makedummy:
2239 			sdl2->sdl_len =
2240 			    offsetof(struct sockaddr_dl, sdl_data[0]);
2241 			sdl2->sdl_family = AF_LINK;
2242 			sdl2->sdl_index = 0;
2243 			sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
2244 		}
2245 		*mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len,
2246 			IP_RECVIF, IPPROTO_IP);
2247 		if (*mp)
2248 			mp = &(*mp)->m_next;
2249 	}
2250 }
2251 
2252 /*
2253  * XXX these routines are called from the upper part of the kernel.
2254  *
2255  * They could also be moved to ip_mroute.c, since all the RSVP
2256  *  handling is done there already.
2257  */
2258 int
2259 ip_rsvp_init(struct socket *so)
2260 {
2261 	if (so->so_type != SOCK_RAW ||
2262 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
2263 		return EOPNOTSUPP;
2264 
2265 	if (ip_rsvpd != NULL)
2266 		return EADDRINUSE;
2267 
2268 	ip_rsvpd = so;
2269 	/*
2270 	 * This may seem silly, but we need to be sure we don't over-increment
2271 	 * the RSVP counter, in case something slips up.
2272 	 */
2273 	if (!ip_rsvp_on) {
2274 		ip_rsvp_on = 1;
2275 		rsvp_on++;
2276 	}
2277 
2278 	return 0;
2279 }
2280 
2281 int
2282 ip_rsvp_done(void)
2283 {
2284 	ip_rsvpd = NULL;
2285 	/*
2286 	 * This may seem silly, but we need to be sure we don't over-decrement
2287 	 * the RSVP counter, in case something slips up.
2288 	 */
2289 	if (ip_rsvp_on) {
2290 		ip_rsvp_on = 0;
2291 		rsvp_on--;
2292 	}
2293 	return 0;
2294 }
2295 
2296 void
2297 rsvp_input(struct mbuf *m, ...)	/* XXX must fixup manually */
2298 {
2299 	int off, proto;
2300 	__va_list ap;
2301 
2302 	__va_start(ap, m);
2303 	off = __va_arg(ap, int);
2304 	proto = __va_arg(ap, int);
2305 	__va_end(ap);
2306 
2307 	if (rsvp_input_p) { /* call the real one if loaded */
2308 		rsvp_input_p(m, off, proto);
2309 		return;
2310 	}
2311 
2312 	/* Can still get packets with rsvp_on = 0 if there is a local member
2313 	 * of the group to which the RSVP packet is addressed.  But in this
2314 	 * case we want to throw the packet away.
2315 	 */
2316 
2317 	if (!rsvp_on) {
2318 		m_freem(m);
2319 		return;
2320 	}
2321 
2322 	if (ip_rsvpd != NULL) {
2323 		rip_input(m, off, proto);
2324 		return;
2325 	}
2326 	/* Drop the packet */
2327 	m_freem(m);
2328 }
2329