xref: /dflybsd-src/sys/netinet/ip_input.c (revision ea0c0e035b7a9a0487f98f059f693585ea607805)
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 1.56 2005/08/15 16:46:21 dillon 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 
98 #include <sys/param.h>
99 #include <sys/systm.h>
100 #include <sys/mbuf.h>
101 #include <sys/malloc.h>
102 #include <sys/mpipe.h>
103 #include <sys/domain.h>
104 #include <sys/protosw.h>
105 #include <sys/socket.h>
106 #include <sys/time.h>
107 #include <sys/globaldata.h>
108 #include <sys/thread.h>
109 #include <sys/kernel.h>
110 #include <sys/syslog.h>
111 #include <sys/sysctl.h>
112 #include <sys/in_cksum.h>
113 
114 #include <sys/thread2.h>
115 #include <sys/msgport2.h>
116 
117 #include <machine/stdarg.h>
118 
119 #include <net/if.h>
120 #include <net/if_types.h>
121 #include <net/if_var.h>
122 #include <net/if_dl.h>
123 #include <net/pfil.h>
124 #include <net/route.h>
125 #include <net/netisr.h>
126 #include <net/intrq.h>
127 
128 #include <netinet/in.h>
129 #include <netinet/in_systm.h>
130 #include <netinet/in_var.h>
131 #include <netinet/ip.h>
132 #include <netinet/in_pcb.h>
133 #include <netinet/ip_var.h>
134 #include <netinet/ip_icmp.h>
135 
136 
137 #include <sys/socketvar.h>
138 
139 #include <net/ipfw/ip_fw.h>
140 #include <net/dummynet/ip_dummynet.h>
141 
142 #ifdef IPSEC
143 #include <netinet6/ipsec.h>
144 #include <netproto/key/key.h>
145 #endif
146 
147 #ifdef FAST_IPSEC
148 #include <netproto/ipsec/ipsec.h>
149 #include <netproto/ipsec/key.h>
150 #endif
151 
152 int rsvp_on = 0;
153 static int ip_rsvp_on;
154 struct socket *ip_rsvpd;
155 
156 int ipforwarding = 0;
157 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW,
158     &ipforwarding, 0, "Enable IP forwarding between interfaces");
159 
160 static int ipsendredirects = 1; /* XXX */
161 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW,
162     &ipsendredirects, 0, "Enable sending IP redirects");
163 
164 int ip_defttl = IPDEFTTL;
165 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW,
166     &ip_defttl, 0, "Maximum TTL on IP packets");
167 
168 static int ip_dosourceroute = 0;
169 SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW,
170     &ip_dosourceroute, 0, "Enable forwarding source routed IP packets");
171 
172 static int ip_acceptsourceroute = 0;
173 SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute,
174     CTLFLAG_RW, &ip_acceptsourceroute, 0,
175     "Enable accepting source routed IP packets");
176 
177 static int ip_keepfaith = 0;
178 SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW,
179     &ip_keepfaith, 0,
180     "Enable packet capture for FAITH IPv4->IPv6 translater daemon");
181 
182 static int nipq = 0;	/* total # of reass queues */
183 static int maxnipq;
184 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragpackets, CTLFLAG_RW,
185     &maxnipq, 0,
186     "Maximum number of IPv4 fragment reassembly queue entries");
187 
188 static int maxfragsperpacket;
189 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW,
190     &maxfragsperpacket, 0,
191     "Maximum number of IPv4 fragments allowed per packet");
192 
193 static int ip_sendsourcequench = 0;
194 SYSCTL_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW,
195     &ip_sendsourcequench, 0,
196     "Enable the transmission of source quench packets");
197 
198 int ip_do_randomid = 0;
199 SYSCTL_INT(_net_inet_ip, OID_AUTO, random_id, CTLFLAG_RW,
200     &ip_do_randomid, 0,
201     "Assign random ip_id values");
202 /*
203  * XXX - Setting ip_checkinterface mostly implements the receive side of
204  * the Strong ES model described in RFC 1122, but since the routing table
205  * and transmit implementation do not implement the Strong ES model,
206  * setting this to 1 results in an odd hybrid.
207  *
208  * XXX - ip_checkinterface currently must be disabled if you use ipnat
209  * to translate the destination address to another local interface.
210  *
211  * XXX - ip_checkinterface must be disabled if you add IP aliases
212  * to the loopback interface instead of the interface where the
213  * packets for those addresses are received.
214  */
215 static int ip_checkinterface = 0;
216 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW,
217     &ip_checkinterface, 0, "Verify packet arrives on correct interface");
218 
219 #ifdef DIAGNOSTIC
220 static int ipprintfs = 0;
221 #endif
222 
223 static struct ifqueue ipintrq;
224 static int ipqmaxlen = IFQ_MAXLEN;
225 
226 extern	struct domain inetdomain;
227 extern	struct protosw inetsw[];
228 u_char	ip_protox[IPPROTO_MAX];
229 struct	in_ifaddrhead in_ifaddrhead;		/* first inet address */
230 struct	in_ifaddrhashhead *in_ifaddrhashtbl;	/* inet addr hash table */
231 u_long	in_ifaddrhmask;				/* mask for hash table */
232 
233 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW,
234     &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue");
235 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD,
236     &ipintrq.ifq_drops, 0, "Number of packets dropped from the IP input queue");
237 
238 struct ip_stats ipstats_percpu[MAXCPU];
239 #ifdef SMP
240 static int
241 sysctl_ipstats(SYSCTL_HANDLER_ARGS)
242 {
243 	int cpu, error = 0;
244 
245 	for (cpu = 0; cpu < ncpus; ++cpu) {
246 		if ((error = SYSCTL_OUT(req, &ipstats_percpu[cpu],
247 					sizeof(struct ip_stats))))
248 			break;
249 		if ((error = SYSCTL_IN(req, &ipstats_percpu[cpu],
250 				       sizeof(struct ip_stats))))
251 			break;
252 	}
253 
254 	return (error);
255 }
256 SYSCTL_PROC(_net_inet_ip, IPCTL_STATS, stats, (CTLTYPE_OPAQUE | CTLFLAG_RW),
257     0, 0, sysctl_ipstats, "S,ip_stats", "IP statistics");
258 #else
259 SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW,
260     &ipstat, ip_stats, "IP statistics");
261 #endif
262 
263 /* Packet reassembly stuff */
264 #define	IPREASS_NHASH_LOG2	6
265 #define	IPREASS_NHASH		(1 << IPREASS_NHASH_LOG2)
266 #define	IPREASS_HMASK		(IPREASS_NHASH - 1)
267 #define	IPREASS_HASH(x,y)						\
268     (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK)
269 
270 static struct ipq ipq[IPREASS_NHASH];
271 const  int    ipintrq_present = 1;
272 
273 #ifdef IPCTL_DEFMTU
274 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
275     &ip_mtu, 0, "Default MTU");
276 #endif
277 
278 #ifdef IPSTEALTH
279 static int ipstealth = 0;
280 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW, &ipstealth, 0, "");
281 #else
282 static const int ipstealth = 0;
283 #endif
284 
285 
286 /* Firewall hooks */
287 ip_fw_chk_t *ip_fw_chk_ptr;
288 int fw_enable = 1;
289 int fw_one_pass = 1;
290 
291 /* Dummynet hooks */
292 ip_dn_io_t *ip_dn_io_ptr;
293 
294 struct pfil_head inet_pfil_hook;
295 
296 /*
297  * XXX this is ugly -- the following two global variables are
298  * used to store packet state while it travels through the stack.
299  * Note that the code even makes assumptions on the size and
300  * alignment of fields inside struct ip_srcrt so e.g. adding some
301  * fields will break the code. This needs to be fixed.
302  *
303  * We need to save the IP options in case a protocol wants to respond
304  * to an incoming packet over the same route if the packet got here
305  * using IP source routing.  This allows connection establishment and
306  * maintenance when the remote end is on a network that is not known
307  * to us.
308  */
309 static int ip_nhops = 0;
310 
311 static	struct ip_srcrt {
312 	struct	in_addr dst;			/* final destination */
313 	char	nop;				/* one NOP to align */
314 	char	srcopt[IPOPT_OFFSET + 1];	/* OPTVAL, OLEN and OFFSET */
315 	struct	in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)];
316 } ip_srcrt;
317 
318 static MALLOC_DEFINE(M_IPQ, "ipq", "IP Fragment Management");
319 static struct malloc_pipe ipq_mpipe;
320 
321 static void		save_rte (u_char *, struct in_addr);
322 static int		ip_dooptions (struct mbuf *m, int,
323 					struct sockaddr_in *next_hop);
324 static void		ip_forward (struct mbuf *m, boolean_t using_srcrt,
325 					struct sockaddr_in *next_hop);
326 static void		ip_freef (struct ipq *);
327 static int		ip_input_handler (struct netmsg *);
328 static struct mbuf	*ip_reass (struct mbuf *, struct ipq *,
329 					struct ipq *, u_int32_t *);
330 
331 /*
332  * IP initialization: fill in IP protocol switch table.
333  * All protocols not implemented in kernel go to raw IP protocol handler.
334  */
335 void
336 ip_init(void)
337 {
338 	struct protosw *pr;
339 	int i;
340 #ifdef SMP
341 	int cpu;
342 #endif
343 
344 	/*
345 	 * Make sure we can handle a reasonable number of fragments but
346 	 * cap it at 4000 (XXX).
347 	 */
348 	mpipe_init(&ipq_mpipe, M_IPQ, sizeof(struct ipq),
349 		    IFQ_MAXLEN, 4000, 0, NULL);
350 	TAILQ_INIT(&in_ifaddrhead);
351 	in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask);
352 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
353 	if (pr == NULL)
354 		panic("ip_init");
355 	for (i = 0; i < IPPROTO_MAX; i++)
356 		ip_protox[i] = pr - inetsw;
357 	for (pr = inetdomain.dom_protosw;
358 	     pr < inetdomain.dom_protoswNPROTOSW; pr++)
359 		if (pr->pr_domain->dom_family == PF_INET &&
360 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
361 			ip_protox[pr->pr_protocol] = pr - inetsw;
362 
363 	inet_pfil_hook.ph_type = PFIL_TYPE_AF;
364 	inet_pfil_hook.ph_af = AF_INET;
365 	if ((i = pfil_head_register(&inet_pfil_hook)) != 0) {
366 		printf("%s: WARNING: unable to register pfil hook, "
367 			"error %d\n", __func__, i);
368 	}
369 
370 	for (i = 0; i < IPREASS_NHASH; i++)
371 	    ipq[i].next = ipq[i].prev = &ipq[i];
372 
373 	maxnipq = nmbclusters / 32;
374 	maxfragsperpacket = 16;
375 
376 	ip_id = time_second & 0xffff;
377 	ipintrq.ifq_maxlen = ipqmaxlen;
378 
379 	/*
380 	 * Initialize IP statistics counters for each CPU.
381 	 *
382 	 */
383 #ifdef SMP
384 	for (cpu = 0; cpu < ncpus; ++cpu) {
385 		bzero(&ipstats_percpu[cpu], sizeof(struct ip_stats));
386 	}
387 #else
388 	bzero(&ipstat, sizeof(struct ip_stats));
389 #endif
390 
391 	netisr_register(NETISR_IP, ip_mport, ip_input_handler);
392 }
393 
394 /*
395  * XXX watch out this one. It is perhaps used as a cache for
396  * the most recently used route ? it is cleared in in_addroute()
397  * when a new route is successfully created.
398  */
399 struct route ipforward_rt;
400 
401 /* Do transport protocol processing. */
402 static void
403 transport_processing_oncpu(struct mbuf *m, int hlen, struct ip *ip,
404 			   struct sockaddr_in *nexthop)
405 {
406 	/*
407 	 * Switch out to protocol's input routine.
408 	 */
409 	if (nexthop && ip->ip_p == IPPROTO_TCP) {
410 		/* TCP needs IPFORWARD info if available */
411 		struct m_hdr tag;
412 
413 		tag.mh_type = MT_TAG;
414 		tag.mh_flags = PACKET_TAG_IPFORWARD;
415 		tag.mh_data = (caddr_t)nexthop;
416 		tag.mh_next = m;
417 
418 		(*inetsw[ip_protox[ip->ip_p]].pr_input)
419 		    ((struct mbuf *)&tag, hlen, ip->ip_p);
420 	} else {
421 		(*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen, ip->ip_p);
422 	}
423 }
424 
425 struct netmsg_transport_packet {
426 	struct lwkt_msg		nm_lmsg;
427 	struct mbuf		*nm_mbuf;
428 	int			nm_hlen;
429 	boolean_t		nm_hasnexthop;
430 	struct sockaddr_in	nm_nexthop;
431 };
432 
433 static int
434 transport_processing_handler(lwkt_msg_t lmsg)
435 {
436 	struct netmsg_transport_packet *msg = (void *)lmsg;
437 	struct sockaddr_in *nexthop;
438 	struct ip *ip;
439 
440 	ip = mtod(msg->nm_mbuf, struct ip *);
441 	nexthop = msg->nm_hasnexthop ? &msg->nm_nexthop : NULL;
442 	transport_processing_oncpu(msg->nm_mbuf, msg->nm_hlen, ip, nexthop);
443 	lwkt_replymsg(lmsg, 0);
444 	return(EASYNC);
445 }
446 
447 static int
448 ip_input_handler(struct netmsg *msg0)
449 {
450 	struct mbuf *m = ((struct netmsg_packet *)msg0)->nm_packet;
451 
452 	ip_input(m);
453 	lwkt_replymsg(&msg0->nm_lmsg, 0);
454 	return(EASYNC);
455 }
456 
457 /*
458  * IP input routine.  Checksum and byte swap header.  If fragmented
459  * try to reassemble.  Process options.  Pass to next level.
460  */
461 void
462 ip_input(struct mbuf *m)
463 {
464 	struct ip *ip;
465 	struct ipq *fp;
466 	struct in_ifaddr *ia = NULL;
467 	struct ifaddr *ifa;
468 	int i, hlen, checkif;
469 	u_short sum;
470 	struct in_addr pkt_dst;
471 	u_int32_t divert_info = 0;		/* packet divert/tee info */
472 	struct ip_fw_args args;
473 	boolean_t using_srcrt = FALSE;		/* forward (by PFIL_HOOKS) */
474 	boolean_t needredispatch = FALSE;
475 	struct in_addr odst;			/* original dst address(NAT) */
476 #if defined(FAST_IPSEC) || defined(IPDIVERT)
477 	struct m_tag *mtag;
478 #endif
479 #ifdef FAST_IPSEC
480 	struct tdb_ident *tdbi;
481 	struct secpolicy *sp;
482 	int error;
483 #endif
484 
485 	args.eh = NULL;
486 	args.oif = NULL;
487 	args.rule = NULL;
488 	args.next_hop = NULL;
489 
490 	/* Grab info from MT_TAG mbufs prepended to the chain. */
491 	while (m != NULL && m->m_type == MT_TAG) {
492 		switch(m->_m_tag_id) {
493 		case PACKET_TAG_DUMMYNET:
494 			args.rule = ((struct dn_pkt *)m)->rule;
495 			break;
496 		case PACKET_TAG_IPFORWARD:
497 			args.next_hop = (struct sockaddr_in *)m->m_hdr.mh_data;
498 			break;
499 		default:
500 			printf("ip_input: unrecognised MT_TAG tag %d\n",
501 			    m->_m_tag_id);
502 			break;
503 		}
504 		m = m->m_next;
505 	}
506 	KASSERT(m != NULL && (m->m_flags & M_PKTHDR), ("ip_input: no HDR"));
507 
508 	if (args.rule != NULL) {	/* dummynet already filtered us */
509 		ip = mtod(m, struct ip *);
510 		hlen = IP_VHL_HL(ip->ip_vhl) << 2;
511 		goto iphack;
512 	}
513 
514 	ipstat.ips_total++;
515 
516 	/* length checks already done in ip_demux() */
517 	KASSERT(m->m_len >= sizeof(ip), ("IP header not in one mbuf"));
518 
519 	ip = mtod(m, struct ip *);
520 
521 	if (IP_VHL_V(ip->ip_vhl) != IPVERSION) {
522 		ipstat.ips_badvers++;
523 		goto bad;
524 	}
525 
526 	hlen = IP_VHL_HL(ip->ip_vhl) << 2;
527 	/* length checks already done in ip_demux() */
528 	KASSERT(hlen >= sizeof(struct ip), ("IP header len too small"));
529 	KASSERT(m->m_len >= hlen, ("packet shorter than IP header length"));
530 
531 	/* 127/8 must not appear on wire - RFC1122 */
532 	if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
533 	    (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
534 		if (!(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK)) {
535 			ipstat.ips_badaddr++;
536 			goto bad;
537 		}
538 	}
539 
540 	if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
541 		sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
542 	} else {
543 		if (hlen == sizeof(struct ip)) {
544 			sum = in_cksum_hdr(ip);
545 		} else {
546 			sum = in_cksum(m, hlen);
547 		}
548 	}
549 	if (sum != 0) {
550 		ipstat.ips_badsum++;
551 		goto bad;
552 	}
553 
554 #ifdef ALTQ
555 	if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) {
556 		/* packet is dropped by traffic conditioner */
557 		return;
558 	}
559 #endif
560 	/*
561 	 * Convert fields to host representation.
562 	 */
563 	ip->ip_len = ntohs(ip->ip_len);
564 	if (ip->ip_len < hlen) {
565 		ipstat.ips_badlen++;
566 		goto bad;
567 	}
568 	ip->ip_off = ntohs(ip->ip_off);
569 
570 	/*
571 	 * Check that the amount of data in the buffers
572 	 * is as at least much as the IP header would have us expect.
573 	 * Trim mbufs if longer than we expect.
574 	 * Drop packet if shorter than we expect.
575 	 */
576 	if (m->m_pkthdr.len < ip->ip_len) {
577 		ipstat.ips_tooshort++;
578 		goto bad;
579 	}
580 	if (m->m_pkthdr.len > ip->ip_len) {
581 		if (m->m_len == m->m_pkthdr.len) {
582 			m->m_len = ip->ip_len;
583 			m->m_pkthdr.len = ip->ip_len;
584 		} else
585 			m_adj(m, ip->ip_len - m->m_pkthdr.len);
586 	}
587 #if defined(IPSEC) && !defined(IPSEC_FILTERGIF)
588 	/*
589 	 * Bypass packet filtering for packets from a tunnel (gif).
590 	 */
591 	if (ipsec_gethist(m, NULL))
592 		goto pass;
593 #endif
594 
595 	/*
596 	 * IpHack's section.
597 	 * Right now when no processing on packet has done
598 	 * and it is still fresh out of network we do our black
599 	 * deals with it.
600 	 * - Firewall: deny/allow/divert
601 	 * - Xlate: translate packet's addr/port (NAT).
602 	 * - Pipe: pass pkt through dummynet.
603 	 * - Wrap: fake packet's addr/port <unimpl.>
604 	 * - Encapsulate: put it in another IP and send out. <unimp.>
605 	 */
606 
607 iphack:
608 
609 	/*
610 	 * Run through list of hooks for input packets.
611 	 *
612 	 * NB: Beware of the destination address changing (e.g.
613 	 *     by NAT rewriting). When this happens, tell
614 	 *     ip_forward to do the right thing.
615 	 */
616 	if (pfil_has_hooks(&inet_pfil_hook)) {
617 		odst = ip->ip_dst;
618 		if (pfil_run_hooks(&inet_pfil_hook, &m,
619 		    m->m_pkthdr.rcvif, PFIL_IN)) {
620 			return;
621 		}
622 		if (m == NULL)			/* consumed by filter */
623 			return;
624 		ip = mtod(m, struct ip *);
625 		using_srcrt = (odst.s_addr != ip->ip_dst.s_addr);
626 	}
627 
628 	if (fw_enable && IPFW_LOADED) {
629 		/*
630 		 * If we've been forwarded from the output side, then
631 		 * skip the firewall a second time
632 		 */
633 		if (args.next_hop != NULL)
634 			goto ours;
635 
636 		args.m = m;
637 		i = ip_fw_chk_ptr(&args);
638 		m = args.m;
639 
640 		if ((i & IP_FW_PORT_DENY_FLAG) || m == NULL) {	/* drop */
641 			if (m != NULL)
642 				m_freem(m);
643 			return;
644 		}
645 		ip = mtod(m, struct ip *);	/* just in case m changed */
646 		if (i == 0 && args.next_hop == NULL)	/* common case */
647 			goto pass;
648 		if (DUMMYNET_LOADED && (i & IP_FW_PORT_DYNT_FLAG)) {
649 			/* Send packet to the appropriate pipe */
650 			ip_dn_io_ptr(m, i&0xffff, DN_TO_IP_IN, &args);
651 			return;
652 		}
653 #ifdef IPDIVERT
654 		if (i != 0 && !(i & IP_FW_PORT_DYNT_FLAG)) {
655 			/* Divert or tee packet */
656 			divert_info = i;
657 			goto ours;
658 		}
659 #endif
660 		if (i == 0 && args.next_hop != NULL)
661 			goto pass;
662 		/*
663 		 * if we get here, the packet must be dropped
664 		 */
665 		m_freem(m);
666 		return;
667 	}
668 pass:
669 
670 	/*
671 	 * Process options and, if not destined for us,
672 	 * ship it on.  ip_dooptions returns 1 when an
673 	 * error was detected (causing an icmp message
674 	 * to be sent and the original packet to be freed).
675 	 */
676 	ip_nhops = 0;		/* for source routed packets */
677 	if (hlen > sizeof(struct ip) && ip_dooptions(m, 0, args.next_hop))
678 		return;
679 
680 	/* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
681 	 * matter if it is destined to another node, or whether it is
682 	 * a multicast one, RSVP wants it! and prevents it from being forwarded
683 	 * anywhere else. Also checks if the rsvp daemon is running before
684 	 * grabbing the packet.
685 	 */
686 	if (rsvp_on && ip->ip_p == IPPROTO_RSVP)
687 		goto ours;
688 
689 	/*
690 	 * Check our list of addresses, to see if the packet is for us.
691 	 * If we don't have any addresses, assume any unicast packet
692 	 * we receive might be for us (and let the upper layers deal
693 	 * with it).
694 	 */
695 	if (TAILQ_EMPTY(&in_ifaddrhead) && !(m->m_flags & (M_MCAST | M_BCAST)))
696 		goto ours;
697 
698 	/*
699 	 * Cache the destination address of the packet; this may be
700 	 * changed by use of 'ipfw fwd'.
701 	 */
702 	pkt_dst = args.next_hop ? args.next_hop->sin_addr : ip->ip_dst;
703 
704 	/*
705 	 * Enable a consistency check between the destination address
706 	 * and the arrival interface for a unicast packet (the RFC 1122
707 	 * strong ES model) if IP forwarding is disabled and the packet
708 	 * is not locally generated and the packet is not subject to
709 	 * 'ipfw fwd'.
710 	 *
711 	 * XXX - Checking also should be disabled if the destination
712 	 * address is ipnat'ed to a different interface.
713 	 *
714 	 * XXX - Checking is incompatible with IP aliases added
715 	 * to the loopback interface instead of the interface where
716 	 * the packets are received.
717 	 */
718 	checkif = ip_checkinterface &&
719 		  !ipforwarding &&
720 		  m->m_pkthdr.rcvif != NULL &&
721 		  !(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) &&
722 		  (args.next_hop == NULL);
723 
724 	/*
725 	 * Check for exact addresses in the hash bucket.
726 	 */
727 	LIST_FOREACH(ia, INADDR_HASH(pkt_dst.s_addr), ia_hash) {
728 		/*
729 		 * If the address matches, verify that the packet
730 		 * arrived via the correct interface if checking is
731 		 * enabled.
732 		 */
733 		if (IA_SIN(ia)->sin_addr.s_addr == pkt_dst.s_addr &&
734 		    (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif))
735 			goto ours;
736 	}
737 	/*
738 	 * Check for broadcast addresses.
739 	 *
740 	 * Only accept broadcast packets that arrive via the matching
741 	 * interface.  Reception of forwarded directed broadcasts would
742 	 * be handled via ip_forward() and ether_output() with the loopback
743 	 * into the stack for SIMPLEX interfaces handled by ether_output().
744 	 */
745 	if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
746 		TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
747 			if (ifa->ifa_addr == NULL) /* shutdown/startup race */
748 				continue;
749 			if (ifa->ifa_addr->sa_family != AF_INET)
750 				continue;
751 			ia = ifatoia(ifa);
752 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
753 								pkt_dst.s_addr)
754 				goto ours;
755 			if (ia->ia_netbroadcast.s_addr == pkt_dst.s_addr)
756 				goto ours;
757 #ifdef BOOTP_COMPAT
758 			if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY)
759 				goto ours;
760 #endif
761 		}
762 	}
763 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
764 		struct in_multi *inm;
765 
766 		if (ip_mrouter != NULL) {
767 			/*
768 			 * If we are acting as a multicast router, all
769 			 * incoming multicast packets are passed to the
770 			 * kernel-level multicast forwarding function.
771 			 * The packet is returned (relatively) intact; if
772 			 * ip_mforward() returns a non-zero value, the packet
773 			 * must be discarded, else it may be accepted below.
774 			 */
775 			if (ip_mforward != NULL &&
776 			    ip_mforward(ip, m->m_pkthdr.rcvif, m, NULL) != 0) {
777 				ipstat.ips_cantforward++;
778 				m_freem(m);
779 				return;
780 			}
781 
782 			/*
783 			 * The process-level routing daemon needs to receive
784 			 * all multicast IGMP packets, whether or not this
785 			 * host belongs to their destination groups.
786 			 */
787 			if (ip->ip_p == IPPROTO_IGMP)
788 				goto ours;
789 			ipstat.ips_forward++;
790 		}
791 		/*
792 		 * See if we belong to the destination multicast group on the
793 		 * arrival interface.
794 		 */
795 		IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
796 		if (inm == NULL) {
797 			ipstat.ips_notmember++;
798 			m_freem(m);
799 			return;
800 		}
801 		goto ours;
802 	}
803 	if (ip->ip_dst.s_addr == INADDR_BROADCAST)
804 		goto ours;
805 	if (ip->ip_dst.s_addr == INADDR_ANY)
806 		goto ours;
807 
808 	/*
809 	 * FAITH(Firewall Aided Internet Translator)
810 	 */
811 	if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
812 		if (ip_keepfaith) {
813 			if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP)
814 				goto ours;
815 		}
816 		m_freem(m);
817 		return;
818 	}
819 
820 	/*
821 	 * Not for us; forward if possible and desirable.
822 	 */
823 	if (!ipforwarding) {
824 		ipstat.ips_cantforward++;
825 		m_freem(m);
826 	} else {
827 #ifdef IPSEC
828 		/*
829 		 * Enforce inbound IPsec SPD.
830 		 */
831 		if (ipsec4_in_reject(m, NULL)) {
832 			ipsecstat.in_polvio++;
833 			goto bad;
834 		}
835 #endif
836 #ifdef FAST_IPSEC
837 		mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
838 		crit_enter();
839 		if (mtag != NULL) {
840 			tdbi = (struct tdb_ident *)m_tag_data(mtag);
841 			sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
842 		} else {
843 			sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
844 						   IP_FORWARDING, &error);
845 		}
846 		if (sp == NULL) {	/* NB: can happen if error */
847 			crit_exit();
848 			/*XXX error stat???*/
849 			DPRINTF(("ip_input: no SP for forwarding\n"));	/*XXX*/
850 			goto bad;
851 		}
852 
853 		/*
854 		 * Check security policy against packet attributes.
855 		 */
856 		error = ipsec_in_reject(sp, m);
857 		KEY_FREESP(&sp);
858 		crit_exit();
859 		if (error) {
860 			ipstat.ips_cantforward++;
861 			goto bad;
862 		}
863 #endif
864 		ip_forward(m, using_srcrt, args.next_hop);
865 	}
866 	return;
867 
868 ours:
869 
870 	/*
871 	 * IPSTEALTH: Process non-routing options only
872 	 * if the packet is destined for us.
873 	 */
874 	if (ipstealth &&
875 	    hlen > sizeof(struct ip) &&
876 	    ip_dooptions(m, 1, args.next_hop))
877 		return;
878 
879 	/* Count the packet in the ip address stats */
880 	if (ia != NULL) {
881 		ia->ia_ifa.if_ipackets++;
882 		ia->ia_ifa.if_ibytes += m->m_pkthdr.len;
883 	}
884 
885 	/*
886 	 * If offset or IP_MF are set, must reassemble.
887 	 * Otherwise, nothing need be done.
888 	 * (We could look in the reassembly queue to see
889 	 * if the packet was previously fragmented,
890 	 * but it's not worth the time; just let them time out.)
891 	 */
892 	if (ip->ip_off & (IP_MF | IP_OFFMASK)) {
893 
894 		/* If maxnipq is 0, never accept fragments. */
895 		if (maxnipq == 0) {
896 			ipstat.ips_fragments++;
897 			ipstat.ips_fragdropped++;
898 			goto bad;
899 		}
900 
901 		sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
902 		/*
903 		 * Look for queue of fragments
904 		 * of this datagram.
905 		 */
906 		for (fp = ipq[sum].next; fp != &ipq[sum]; fp = fp->next)
907 			if (ip->ip_id == fp->ipq_id &&
908 			    ip->ip_src.s_addr == fp->ipq_src.s_addr &&
909 			    ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
910 			    ip->ip_p == fp->ipq_p)
911 				goto found;
912 
913 		fp = NULL;
914 
915 		/*
916 		 * Enforce upper bound on number of fragmented packets
917 		 * for which we attempt reassembly;
918 		 * If maxnipq is -1, accept all fragments without limitation.
919 		 */
920 		if ((nipq > maxnipq) && (maxnipq > 0)) {
921 			/*
922 			 * drop something from the tail of the current queue
923 			 * before proceeding further
924 			 */
925 			if (ipq[sum].prev == &ipq[sum]) {   /* gak */
926 				for (i = 0; i < IPREASS_NHASH; i++) {
927 					if (ipq[i].prev != &ipq[i]) {
928 						ipstat.ips_fragtimeout +=
929 						    ipq[i].prev->ipq_nfrags;
930 						ip_freef(ipq[i].prev);
931 						break;
932 					}
933 				}
934 			} else {
935 				ipstat.ips_fragtimeout +=
936 				    ipq[sum].prev->ipq_nfrags;
937 				ip_freef(ipq[sum].prev);
938 			}
939 		}
940 found:
941 		/*
942 		 * Adjust ip_len to not reflect header,
943 		 * convert offset of this to bytes.
944 		 */
945 		ip->ip_len -= hlen;
946 		if (ip->ip_off & IP_MF) {
947 			/*
948 			 * Make sure that fragments have a data length
949 			 * that's a non-zero multiple of 8 bytes.
950 			 */
951 			if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
952 				ipstat.ips_toosmall++; /* XXX */
953 				goto bad;
954 			}
955 			m->m_flags |= M_FRAG;
956 		} else
957 			m->m_flags &= ~M_FRAG;
958 		ip->ip_off <<= 3;
959 
960 		/*
961 		 * Attempt reassembly; if it succeeds, proceed.
962 		 * ip_reass() will return a different mbuf, and update
963 		 * the divert info in divert_info.
964 		 */
965 		ipstat.ips_fragments++;
966 		m->m_pkthdr.header = ip;
967 		m = ip_reass(m, fp, &ipq[sum], &divert_info);
968 		if (m == NULL)
969 			return;
970 		ipstat.ips_reassembled++;
971 		needredispatch = TRUE;
972 		ip = mtod(m, struct ip *);
973 		/* Get the header length of the reassembled packet */
974 		hlen = IP_VHL_HL(ip->ip_vhl) << 2;
975 #ifdef IPDIVERT
976 		/* Restore original checksum before diverting packet */
977 		if (divert_info != 0) {
978 			ip->ip_len += hlen;
979 			ip->ip_len = htons(ip->ip_len);
980 			ip->ip_off = htons(ip->ip_off);
981 			ip->ip_sum = 0;
982 			if (hlen == sizeof(struct ip))
983 				ip->ip_sum = in_cksum_hdr(ip);
984 			else
985 				ip->ip_sum = in_cksum(m, hlen);
986 			ip->ip_off = ntohs(ip->ip_off);
987 			ip->ip_len = ntohs(ip->ip_len);
988 			ip->ip_len -= hlen;
989 		}
990 #endif
991 	} else {
992 		ip->ip_len -= hlen;
993 	}
994 
995 #ifdef IPDIVERT
996 	/*
997 	 * Divert or tee packet to the divert protocol if required.
998 	 */
999 	if (divert_info != 0) {
1000 		struct mbuf *clone = NULL;
1001 
1002 		/* Clone packet if we're doing a 'tee' */
1003 		if ((divert_info & IP_FW_PORT_TEE_FLAG) != 0)
1004 			clone = m_dup(m, MB_DONTWAIT);
1005 
1006 		/* Restore packet header fields to original values */
1007 		ip->ip_len += hlen;
1008 		ip->ip_len = htons(ip->ip_len);
1009 		ip->ip_off = htons(ip->ip_off);
1010 
1011 		/* Deliver packet to divert input routine */
1012 		divert_packet(m, 1, divert_info & 0xffff);
1013 		ipstat.ips_delivered++;
1014 
1015 		/* If 'tee', continue with original packet */
1016 		if (clone == NULL)
1017 			return;
1018 		m = clone;
1019 		ip = mtod(m, struct ip *);
1020 		ip->ip_len += hlen;
1021 		/*
1022 		 * Jump backwards to complete processing of the
1023 		 * packet. But first clear divert_info to avoid
1024 		 * entering this block again.
1025 		 * We do not need to clear args.divert_rule
1026 		 * or args.next_hop as they will not be used.
1027 		 *
1028 		 * XXX Better safe than sorry, remove the DIVERT tag.
1029 		 */
1030 		mtag = m_tag_find(m, PACKET_TAG_IPFW_DIVERT, NULL);
1031 		if (mtag != NULL)
1032 			m_tag_delete(m, mtag);
1033 
1034 		divert_info = 0;
1035 		goto pass;
1036 	}
1037 #endif
1038 
1039 #ifdef IPSEC
1040 	/*
1041 	 * enforce IPsec policy checking if we are seeing last header.
1042 	 * note that we do not visit this with protocols with pcb layer
1043 	 * code - like udp/tcp/raw ip.
1044 	 */
1045 	if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) &&
1046 	    ipsec4_in_reject(m, NULL)) {
1047 		ipsecstat.in_polvio++;
1048 		goto bad;
1049 	}
1050 #endif
1051 #if FAST_IPSEC
1052 	/*
1053 	 * enforce IPsec policy checking if we are seeing last header.
1054 	 * note that we do not visit this with protocols with pcb layer
1055 	 * code - like udp/tcp/raw ip.
1056 	 */
1057 	if (inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) {
1058 		/*
1059 		 * Check if the packet has already had IPsec processing
1060 		 * done.  If so, then just pass it along.  This tag gets
1061 		 * set during AH, ESP, etc. input handling, before the
1062 		 * packet is returned to the ip input queue for delivery.
1063 		 */
1064 		mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
1065 		crit_enter();
1066 		if (mtag != NULL) {
1067 			tdbi = (struct tdb_ident *)m_tag_data(mtag);
1068 			sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
1069 		} else {
1070 			sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
1071 						   IP_FORWARDING, &error);
1072 		}
1073 		if (sp != NULL) {
1074 			/*
1075 			 * Check security policy against packet attributes.
1076 			 */
1077 			error = ipsec_in_reject(sp, m);
1078 			KEY_FREESP(&sp);
1079 		} else {
1080 			/* XXX error stat??? */
1081 			error = EINVAL;
1082 DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/
1083 			goto bad;
1084 		}
1085 		crit_exit();
1086 		if (error)
1087 			goto bad;
1088 	}
1089 #endif /* FAST_IPSEC */
1090 
1091 	ipstat.ips_delivered++;
1092 	if (needredispatch) {
1093 		struct netmsg_transport_packet *msg;
1094 		lwkt_port_t port;
1095 
1096 		msg = malloc(sizeof(struct netmsg_transport_packet),
1097 				M_LWKTMSG, M_INTWAIT | M_NULLOK);
1098 		if (msg == NULL)
1099 			goto bad;
1100 
1101 		lwkt_initmsg(&msg->nm_lmsg, &netisr_afree_rport, 0,
1102 			lwkt_cmd_func(transport_processing_handler),
1103 			lwkt_cmd_op_none);
1104 		msg->nm_hlen = hlen;
1105 		msg->nm_hasnexthop = (args.next_hop != NULL);
1106 		if (msg->nm_hasnexthop)
1107 			msg->nm_nexthop = *args.next_hop;  /* structure copy */
1108 
1109 		ip->ip_off = htons(ip->ip_off);
1110 		ip->ip_len = htons(ip->ip_len);
1111 		port = ip_mport(&m);
1112 		if (port) {
1113 		    msg->nm_mbuf = m;
1114 		    ip = mtod(m, struct ip *);
1115 		    ip->ip_len = ntohs(ip->ip_len);
1116 		    ip->ip_off = ntohs(ip->ip_off);
1117 		    lwkt_sendmsg(port, &msg->nm_lmsg);
1118 		}
1119 	} else {
1120 		transport_processing_oncpu(m, hlen, ip, args.next_hop);
1121 	}
1122 	return;
1123 
1124 bad:
1125 	m_freem(m);
1126 }
1127 
1128 /*
1129  * Take incoming datagram fragment and try to reassemble it into
1130  * whole datagram.  If a chain for reassembly of this datagram already
1131  * exists, then it is given as fp; otherwise have to make a chain.
1132  *
1133  * When IPDIVERT enabled, keep additional state with each packet that
1134  * tells us if we need to divert or tee the packet we're building.
1135  * In particular, *divinfo includes the port and TEE flag.
1136  */
1137 
1138 static struct mbuf *
1139 ip_reass(struct mbuf *m, struct ipq *fp, struct ipq *where,
1140 	 u_int32_t *divinfo)
1141 {
1142 	struct ip *ip = mtod(m, struct ip *);
1143 	struct mbuf *p = NULL, *q, *nq;
1144 	struct mbuf *n;
1145 	int hlen = IP_VHL_HL(ip->ip_vhl) << 2;
1146 	int i, next;
1147 #ifdef IPDIVERT
1148 	struct m_tag *mtag;
1149 #endif
1150 
1151 	/*
1152 	 * Presence of header sizes in mbufs
1153 	 * would confuse code below.
1154 	 */
1155 	m->m_data += hlen;
1156 	m->m_len -= hlen;
1157 
1158 	/*
1159 	 * If first fragment to arrive, create a reassembly queue.
1160 	 */
1161 	if (fp == NULL) {
1162 		if ((fp = mpipe_alloc_nowait(&ipq_mpipe)) == NULL)
1163 			goto dropfrag;
1164 		insque(fp, where);
1165 		nipq++;
1166 		fp->ipq_nfrags = 1;
1167 		fp->ipq_ttl = IPFRAGTTL;
1168 		fp->ipq_p = ip->ip_p;
1169 		fp->ipq_id = ip->ip_id;
1170 		fp->ipq_src = ip->ip_src;
1171 		fp->ipq_dst = ip->ip_dst;
1172 		fp->ipq_frags = m;
1173 		m->m_nextpkt = NULL;
1174 #ifdef IPDIVERT
1175 		fp->ipq_div_info = 0;
1176 #endif
1177 		goto inserted;
1178 	} else {
1179 		fp->ipq_nfrags++;
1180 	}
1181 
1182 #define	GETIP(m)	((struct ip*)((m)->m_pkthdr.header))
1183 
1184 	/*
1185 	 * Find a segment which begins after this one does.
1186 	 */
1187 	for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt)
1188 		if (GETIP(q)->ip_off > ip->ip_off)
1189 			break;
1190 
1191 	/*
1192 	 * If there is a preceding segment, it may provide some of
1193 	 * our data already.  If so, drop the data from the incoming
1194 	 * segment.  If it provides all of our data, drop us, otherwise
1195 	 * stick new segment in the proper place.
1196 	 *
1197 	 * If some of the data is dropped from the the preceding
1198 	 * segment, then it's checksum is invalidated.
1199 	 */
1200 	if (p) {
1201 		i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off;
1202 		if (i > 0) {
1203 			if (i >= ip->ip_len)
1204 				goto dropfrag;
1205 			m_adj(m, i);
1206 			m->m_pkthdr.csum_flags = 0;
1207 			ip->ip_off += i;
1208 			ip->ip_len -= i;
1209 		}
1210 		m->m_nextpkt = p->m_nextpkt;
1211 		p->m_nextpkt = m;
1212 	} else {
1213 		m->m_nextpkt = fp->ipq_frags;
1214 		fp->ipq_frags = m;
1215 	}
1216 
1217 	/*
1218 	 * While we overlap succeeding segments trim them or,
1219 	 * if they are completely covered, dequeue them.
1220 	 */
1221 	for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off;
1222 	     q = nq) {
1223 		i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off;
1224 		if (i < GETIP(q)->ip_len) {
1225 			GETIP(q)->ip_len -= i;
1226 			GETIP(q)->ip_off += i;
1227 			m_adj(q, i);
1228 			q->m_pkthdr.csum_flags = 0;
1229 			break;
1230 		}
1231 		nq = q->m_nextpkt;
1232 		m->m_nextpkt = nq;
1233 		ipstat.ips_fragdropped++;
1234 		fp->ipq_nfrags--;
1235 		m_freem(q);
1236 	}
1237 
1238 inserted:
1239 
1240 #ifdef IPDIVERT
1241 	/*
1242 	 * Transfer firewall instructions to the fragment structure.
1243 	 * Only trust info in the fragment at offset 0.
1244 	 */
1245 	if (ip->ip_off == 0) {
1246 		fp->ipq_div_info = *divinfo;
1247 	} else {
1248 		mtag = m_tag_find(m, PACKET_TAG_IPFW_DIVERT, NULL);
1249 		if (mtag != NULL)
1250 			m_tag_delete(m, mtag);
1251 	}
1252 	*divinfo = 0;
1253 #endif
1254 
1255 	/*
1256 	 * Check for complete reassembly and perform frag per packet
1257 	 * limiting.
1258 	 *
1259 	 * Frag limiting is performed here so that the nth frag has
1260 	 * a chance to complete the packet before we drop the packet.
1261 	 * As a result, n+1 frags are actually allowed per packet, but
1262 	 * only n will ever be stored. (n = maxfragsperpacket.)
1263 	 *
1264 	 */
1265 	next = 0;
1266 	for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
1267 		if (GETIP(q)->ip_off != next) {
1268 			if (fp->ipq_nfrags > maxfragsperpacket) {
1269 				ipstat.ips_fragdropped += fp->ipq_nfrags;
1270 				ip_freef(fp);
1271 			}
1272 			return (NULL);
1273 		}
1274 		next += GETIP(q)->ip_len;
1275 	}
1276 	/* Make sure the last packet didn't have the IP_MF flag */
1277 	if (p->m_flags & M_FRAG) {
1278 		if (fp->ipq_nfrags > maxfragsperpacket) {
1279 			ipstat.ips_fragdropped += fp->ipq_nfrags;
1280 			ip_freef(fp);
1281 		}
1282 		return (NULL);
1283 	}
1284 
1285 	/*
1286 	 * Reassembly is complete.  Make sure the packet is a sane size.
1287 	 */
1288 	q = fp->ipq_frags;
1289 	ip = GETIP(q);
1290 	if (next + (IP_VHL_HL(ip->ip_vhl) << 2) > IP_MAXPACKET) {
1291 		ipstat.ips_toolong++;
1292 		ipstat.ips_fragdropped += fp->ipq_nfrags;
1293 		ip_freef(fp);
1294 		return (NULL);
1295 	}
1296 
1297 	/*
1298 	 * Concatenate fragments.
1299 	 */
1300 	m = q;
1301 	n = m->m_next;
1302 	m->m_next = NULL;
1303 	m_cat(m, n);
1304 	nq = q->m_nextpkt;
1305 	q->m_nextpkt = NULL;
1306 	for (q = nq; q != NULL; q = nq) {
1307 		nq = q->m_nextpkt;
1308 		q->m_nextpkt = NULL;
1309 		m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
1310 		m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
1311 		m_cat(m, q);
1312 	}
1313 
1314 #ifdef IPDIVERT
1315 	/*
1316 	 * Extract firewall instructions from the fragment structure.
1317 	 */
1318 	*divinfo = fp->ipq_div_info;
1319 #endif
1320 
1321 	/*
1322 	 * Create header for new ip packet by
1323 	 * modifying header of first packet;
1324 	 * dequeue and discard fragment reassembly header.
1325 	 * Make header visible.
1326 	 */
1327 	ip->ip_len = next;
1328 	ip->ip_src = fp->ipq_src;
1329 	ip->ip_dst = fp->ipq_dst;
1330 	remque(fp);
1331 	nipq--;
1332 	mpipe_free(&ipq_mpipe, fp);
1333 	m->m_len += (IP_VHL_HL(ip->ip_vhl) << 2);
1334 	m->m_data -= (IP_VHL_HL(ip->ip_vhl) << 2);
1335 	/* some debugging cruft by sklower, below, will go away soon */
1336 	if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */
1337 		int plen = 0;
1338 
1339 		for (n = m; n; n = n->m_next)
1340 			plen += n->m_len;
1341 		m->m_pkthdr.len = plen;
1342 	}
1343 	return (m);
1344 
1345 dropfrag:
1346 #ifdef IPDIVERT
1347 	*divinfo = 0;
1348 #endif
1349 	ipstat.ips_fragdropped++;
1350 	if (fp != NULL)
1351 		fp->ipq_nfrags--;
1352 	m_freem(m);
1353 	return (NULL);
1354 
1355 #undef GETIP
1356 }
1357 
1358 /*
1359  * Free a fragment reassembly header and all
1360  * associated datagrams.
1361  */
1362 static void
1363 ip_freef(struct ipq *fp)
1364 {
1365 	struct mbuf *q;
1366 
1367 	while (fp->ipq_frags) {
1368 		q = fp->ipq_frags;
1369 		fp->ipq_frags = q->m_nextpkt;
1370 		m_freem(q);
1371 	}
1372 	remque(fp);
1373 	mpipe_free(&ipq_mpipe, fp);
1374 	nipq--;
1375 }
1376 
1377 /*
1378  * IP timer processing;
1379  * if a timer expires on a reassembly
1380  * queue, discard it.
1381  */
1382 void
1383 ip_slowtimo(void)
1384 {
1385 	struct ipq *fp;
1386 	int i;
1387 
1388 	crit_enter();
1389 	for (i = 0; i < IPREASS_NHASH; i++) {
1390 		fp = ipq[i].next;
1391 		if (fp == NULL)
1392 			continue;
1393 		while (fp != &ipq[i]) {
1394 			--fp->ipq_ttl;
1395 			fp = fp->next;
1396 			if (fp->prev->ipq_ttl == 0) {
1397 				ipstat.ips_fragtimeout += fp->prev->ipq_nfrags;
1398 				ip_freef(fp->prev);
1399 			}
1400 		}
1401 	}
1402 	/*
1403 	 * If we are over the maximum number of fragments
1404 	 * (due to the limit being lowered), drain off
1405 	 * enough to get down to the new limit.
1406 	 */
1407 	if (maxnipq >= 0 && nipq > maxnipq) {
1408 		for (i = 0; i < IPREASS_NHASH; i++) {
1409 			while (nipq > maxnipq &&
1410 				(ipq[i].next != &ipq[i])) {
1411 				ipstat.ips_fragdropped +=
1412 				    ipq[i].next->ipq_nfrags;
1413 				ip_freef(ipq[i].next);
1414 			}
1415 		}
1416 	}
1417 	ipflow_slowtimo();
1418 	crit_exit();
1419 }
1420 
1421 /*
1422  * Drain off all datagram fragments.
1423  */
1424 void
1425 ip_drain(void)
1426 {
1427 	int i;
1428 
1429 	for (i = 0; i < IPREASS_NHASH; i++) {
1430 		while (ipq[i].next != &ipq[i]) {
1431 			ipstat.ips_fragdropped += ipq[i].next->ipq_nfrags;
1432 			ip_freef(ipq[i].next);
1433 		}
1434 	}
1435 	in_rtqdrain();
1436 }
1437 
1438 /*
1439  * Do option processing on a datagram,
1440  * possibly discarding it if bad options are encountered,
1441  * or forwarding it if source-routed.
1442  * The pass argument is used when operating in the IPSTEALTH
1443  * mode to tell what options to process:
1444  * [LS]SRR (pass 0) or the others (pass 1).
1445  * The reason for as many as two passes is that when doing IPSTEALTH,
1446  * non-routing options should be processed only if the packet is for us.
1447  * Returns 1 if packet has been forwarded/freed,
1448  * 0 if the packet should be processed further.
1449  */
1450 static int
1451 ip_dooptions(struct mbuf *m, int pass, struct sockaddr_in *next_hop)
1452 {
1453 	struct sockaddr_in ipaddr = { sizeof ipaddr, AF_INET };
1454 	struct ip *ip = mtod(m, struct ip *);
1455 	u_char *cp;
1456 	struct in_ifaddr *ia;
1457 	int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB;
1458 	boolean_t forward = FALSE;
1459 	struct in_addr *sin, dst;
1460 	n_time ntime;
1461 
1462 	dst = ip->ip_dst;
1463 	cp = (u_char *)(ip + 1);
1464 	cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
1465 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1466 		opt = cp[IPOPT_OPTVAL];
1467 		if (opt == IPOPT_EOL)
1468 			break;
1469 		if (opt == IPOPT_NOP)
1470 			optlen = 1;
1471 		else {
1472 			if (cnt < IPOPT_OLEN + sizeof(*cp)) {
1473 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1474 				goto bad;
1475 			}
1476 			optlen = cp[IPOPT_OLEN];
1477 			if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
1478 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1479 				goto bad;
1480 			}
1481 		}
1482 		switch (opt) {
1483 
1484 		default:
1485 			break;
1486 
1487 		/*
1488 		 * Source routing with record.
1489 		 * Find interface with current destination address.
1490 		 * If none on this machine then drop if strictly routed,
1491 		 * or do nothing if loosely routed.
1492 		 * Record interface address and bring up next address
1493 		 * component.  If strictly routed make sure next
1494 		 * address is on directly accessible net.
1495 		 */
1496 		case IPOPT_LSRR:
1497 		case IPOPT_SSRR:
1498 			if (ipstealth && pass > 0)
1499 				break;
1500 			if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1501 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1502 				goto bad;
1503 			}
1504 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1505 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1506 				goto bad;
1507 			}
1508 			ipaddr.sin_addr = ip->ip_dst;
1509 			ia = (struct in_ifaddr *)
1510 				ifa_ifwithaddr((struct sockaddr *)&ipaddr);
1511 			if (ia == NULL) {
1512 				if (opt == IPOPT_SSRR) {
1513 					type = ICMP_UNREACH;
1514 					code = ICMP_UNREACH_SRCFAIL;
1515 					goto bad;
1516 				}
1517 				if (!ip_dosourceroute)
1518 					goto nosourcerouting;
1519 				/*
1520 				 * Loose routing, and not at next destination
1521 				 * yet; nothing to do except forward.
1522 				 */
1523 				break;
1524 			}
1525 			off--;			/* 0 origin */
1526 			if (off > optlen - (int)sizeof(struct in_addr)) {
1527 				/*
1528 				 * End of source route.  Should be for us.
1529 				 */
1530 				if (!ip_acceptsourceroute)
1531 					goto nosourcerouting;
1532 				save_rte(cp, ip->ip_src);
1533 				break;
1534 			}
1535 			if (ipstealth)
1536 				goto dropit;
1537 			if (!ip_dosourceroute) {
1538 				if (ipforwarding) {
1539 					char buf[sizeof "aaa.bbb.ccc.ddd"];
1540 
1541 					/*
1542 					 * Acting as a router, so generate ICMP
1543 					 */
1544 nosourcerouting:
1545 					strcpy(buf, inet_ntoa(ip->ip_dst));
1546 					log(LOG_WARNING,
1547 					    "attempted source route from %s to %s\n",
1548 					    inet_ntoa(ip->ip_src), buf);
1549 					type = ICMP_UNREACH;
1550 					code = ICMP_UNREACH_SRCFAIL;
1551 					goto bad;
1552 				} else {
1553 					/*
1554 					 * Not acting as a router,
1555 					 * so silently drop.
1556 					 */
1557 dropit:
1558 					ipstat.ips_cantforward++;
1559 					m_freem(m);
1560 					return (1);
1561 				}
1562 			}
1563 
1564 			/*
1565 			 * locate outgoing interface
1566 			 */
1567 			memcpy(&ipaddr.sin_addr, cp + off,
1568 			    sizeof ipaddr.sin_addr);
1569 
1570 			if (opt == IPOPT_SSRR) {
1571 #define	INA	struct in_ifaddr *
1572 #define	SA	struct sockaddr *
1573 				if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr))
1574 									== NULL)
1575 					ia = (INA)ifa_ifwithnet((SA)&ipaddr);
1576 			} else
1577 				ia = ip_rtaddr(ipaddr.sin_addr, &ipforward_rt);
1578 			if (ia == NULL) {
1579 				type = ICMP_UNREACH;
1580 				code = ICMP_UNREACH_SRCFAIL;
1581 				goto bad;
1582 			}
1583 			ip->ip_dst = ipaddr.sin_addr;
1584 			memcpy(cp + off, &IA_SIN(ia)->sin_addr,
1585 			    sizeof(struct in_addr));
1586 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1587 			/*
1588 			 * Let ip_intr's mcast routing check handle mcast pkts
1589 			 */
1590 			forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr));
1591 			break;
1592 
1593 		case IPOPT_RR:
1594 			if (ipstealth && pass == 0)
1595 				break;
1596 			if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1597 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1598 				goto bad;
1599 			}
1600 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1601 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1602 				goto bad;
1603 			}
1604 			/*
1605 			 * If no space remains, ignore.
1606 			 */
1607 			off--;			/* 0 origin */
1608 			if (off > optlen - (int)sizeof(struct in_addr))
1609 				break;
1610 			memcpy(&ipaddr.sin_addr, &ip->ip_dst,
1611 			    sizeof ipaddr.sin_addr);
1612 			/*
1613 			 * locate outgoing interface; if we're the destination,
1614 			 * use the incoming interface (should be same).
1615 			 */
1616 			if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == NULL &&
1617 			    (ia = ip_rtaddr(ipaddr.sin_addr, &ipforward_rt))
1618 								     == NULL) {
1619 				type = ICMP_UNREACH;
1620 				code = ICMP_UNREACH_HOST;
1621 				goto bad;
1622 			}
1623 			memcpy(cp + off, &IA_SIN(ia)->sin_addr,
1624 			    sizeof(struct in_addr));
1625 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1626 			break;
1627 
1628 		case IPOPT_TS:
1629 			if (ipstealth && pass == 0)
1630 				break;
1631 			code = cp - (u_char *)ip;
1632 			if (optlen < 4 || optlen > 40) {
1633 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1634 				goto bad;
1635 			}
1636 			if ((off = cp[IPOPT_OFFSET]) < 5) {
1637 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1638 				goto bad;
1639 			}
1640 			if (off > optlen - (int)sizeof(int32_t)) {
1641 				cp[IPOPT_OFFSET + 1] += (1 << 4);
1642 				if ((cp[IPOPT_OFFSET + 1] & 0xf0) == 0) {
1643 					code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1644 					goto bad;
1645 				}
1646 				break;
1647 			}
1648 			off--;				/* 0 origin */
1649 			sin = (struct in_addr *)(cp + off);
1650 			switch (cp[IPOPT_OFFSET + 1] & 0x0f) {
1651 
1652 			case IPOPT_TS_TSONLY:
1653 				break;
1654 
1655 			case IPOPT_TS_TSANDADDR:
1656 				if (off + sizeof(n_time) +
1657 				    sizeof(struct in_addr) > optlen) {
1658 					code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1659 					goto bad;
1660 				}
1661 				ipaddr.sin_addr = dst;
1662 				ia = (INA)ifaof_ifpforaddr((SA)&ipaddr,
1663 							    m->m_pkthdr.rcvif);
1664 				if (ia == NULL)
1665 					continue;
1666 				memcpy(sin, &IA_SIN(ia)->sin_addr,
1667 				    sizeof(struct in_addr));
1668 				cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1669 				off += sizeof(struct in_addr);
1670 				break;
1671 
1672 			case IPOPT_TS_PRESPEC:
1673 				if (off + sizeof(n_time) +
1674 				    sizeof(struct in_addr) > optlen) {
1675 					code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1676 					goto bad;
1677 				}
1678 				memcpy(&ipaddr.sin_addr, sin,
1679 				    sizeof(struct in_addr));
1680 				if (ifa_ifwithaddr((SA)&ipaddr) == NULL)
1681 					continue;
1682 				cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1683 				off += sizeof(struct in_addr);
1684 				break;
1685 
1686 			default:
1687 				code = &cp[IPOPT_OFFSET + 1] - (u_char *)ip;
1688 				goto bad;
1689 			}
1690 			ntime = iptime();
1691 			memcpy(cp + off, &ntime, sizeof(n_time));
1692 			cp[IPOPT_OFFSET] += sizeof(n_time);
1693 		}
1694 	}
1695 	if (forward && ipforwarding) {
1696 		ip_forward(m, TRUE, next_hop);
1697 		return (1);
1698 	}
1699 	return (0);
1700 bad:
1701 	icmp_error(m, type, code, 0, NULL);
1702 	ipstat.ips_badoptions++;
1703 	return (1);
1704 }
1705 
1706 /*
1707  * Given address of next destination (final or next hop),
1708  * return internet address info of interface to be used to get there.
1709  */
1710 struct in_ifaddr *
1711 ip_rtaddr(struct in_addr dst, struct route *ro)
1712 {
1713 	struct sockaddr_in *sin;
1714 
1715 	sin = (struct sockaddr_in *)&ro->ro_dst;
1716 
1717 	if (ro->ro_rt == NULL || dst.s_addr != sin->sin_addr.s_addr) {
1718 		if (ro->ro_rt != NULL) {
1719 			RTFREE(ro->ro_rt);
1720 			ro->ro_rt = NULL;
1721 		}
1722 		sin->sin_family = AF_INET;
1723 		sin->sin_len = sizeof *sin;
1724 		sin->sin_addr = dst;
1725 		rtalloc_ign(ro, RTF_PRCLONING);
1726 	}
1727 
1728 	if (ro->ro_rt == NULL)
1729 		return (NULL);
1730 
1731 	return (ifatoia(ro->ro_rt->rt_ifa));
1732 }
1733 
1734 /*
1735  * Save incoming source route for use in replies,
1736  * to be picked up later by ip_srcroute if the receiver is interested.
1737  */
1738 void
1739 save_rte(u_char *option, struct in_addr dst)
1740 {
1741 	unsigned olen;
1742 
1743 	olen = option[IPOPT_OLEN];
1744 #ifdef DIAGNOSTIC
1745 	if (ipprintfs)
1746 		printf("save_rte: olen %d\n", olen);
1747 #endif
1748 	if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst)))
1749 		return;
1750 	bcopy(option, ip_srcrt.srcopt, olen);
1751 	ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1752 	ip_srcrt.dst = dst;
1753 }
1754 
1755 /*
1756  * Retrieve incoming source route for use in replies,
1757  * in the same form used by setsockopt.
1758  * The first hop is placed before the options, will be removed later.
1759  */
1760 struct mbuf *
1761 ip_srcroute(void)
1762 {
1763 	struct in_addr *p, *q;
1764 	struct mbuf *m;
1765 
1766 	if (ip_nhops == 0)
1767 		return (NULL);
1768 	m = m_get(MB_DONTWAIT, MT_HEADER);
1769 	if (m == NULL)
1770 		return (NULL);
1771 
1772 #define	OPTSIZ	(sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt))
1773 
1774 	/* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
1775 	m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) +
1776 	    OPTSIZ;
1777 #ifdef DIAGNOSTIC
1778 	if (ipprintfs)
1779 		printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len);
1780 #endif
1781 
1782 	/*
1783 	 * First save first hop for return route
1784 	 */
1785 	p = &ip_srcrt.route[ip_nhops - 1];
1786 	*(mtod(m, struct in_addr *)) = *p--;
1787 #ifdef DIAGNOSTIC
1788 	if (ipprintfs)
1789 		printf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr));
1790 #endif
1791 
1792 	/*
1793 	 * Copy option fields and padding (nop) to mbuf.
1794 	 */
1795 	ip_srcrt.nop = IPOPT_NOP;
1796 	ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
1797 	memcpy(mtod(m, caddr_t) + sizeof(struct in_addr), &ip_srcrt.nop,
1798 	    OPTSIZ);
1799 	q = (struct in_addr *)(mtod(m, caddr_t) +
1800 	    sizeof(struct in_addr) + OPTSIZ);
1801 #undef OPTSIZ
1802 	/*
1803 	 * Record return path as an IP source route,
1804 	 * reversing the path (pointers are now aligned).
1805 	 */
1806 	while (p >= ip_srcrt.route) {
1807 #ifdef DIAGNOSTIC
1808 		if (ipprintfs)
1809 			printf(" %x", ntohl(q->s_addr));
1810 #endif
1811 		*q++ = *p--;
1812 	}
1813 	/*
1814 	 * Last hop goes to final destination.
1815 	 */
1816 	*q = ip_srcrt.dst;
1817 #ifdef DIAGNOSTIC
1818 	if (ipprintfs)
1819 		printf(" %x\n", ntohl(q->s_addr));
1820 #endif
1821 	return (m);
1822 }
1823 
1824 /*
1825  * Strip out IP options.
1826  */
1827 void
1828 ip_stripoptions(struct mbuf *m)
1829 {
1830 	int datalen;
1831 	struct ip *ip = mtod(m, struct ip *);
1832 	caddr_t opts;
1833 	int optlen;
1834 
1835 	optlen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
1836 	opts = (caddr_t)(ip + 1);
1837 	datalen = m->m_len - (sizeof(struct ip) + optlen);
1838 	bcopy(opts + optlen, opts, datalen);
1839 	m->m_len -= optlen;
1840 	if (m->m_flags & M_PKTHDR)
1841 		m->m_pkthdr.len -= optlen;
1842 	ip->ip_vhl = IP_MAKE_VHL(IPVERSION, sizeof(struct ip) >> 2);
1843 }
1844 
1845 u_char inetctlerrmap[PRC_NCMDS] = {
1846 	0,		0,		0,		0,
1847 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1848 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1849 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1850 	0,		0,		0,		0,
1851 	ENOPROTOOPT,	ECONNREFUSED
1852 };
1853 
1854 /*
1855  * Forward a packet.  If some error occurs return the sender
1856  * an icmp packet.  Note we can't always generate a meaningful
1857  * icmp message because icmp doesn't have a large enough repertoire
1858  * of codes and types.
1859  *
1860  * If not forwarding, just drop the packet.  This could be confusing
1861  * if ipforwarding was zero but some routing protocol was advancing
1862  * us as a gateway to somewhere.  However, we must let the routing
1863  * protocol deal with that.
1864  *
1865  * The using_srcrt parameter indicates whether the packet is being forwarded
1866  * via a source route.
1867  */
1868 static void
1869 ip_forward(struct mbuf *m, boolean_t using_srcrt, struct sockaddr_in *next_hop)
1870 {
1871 	struct ip *ip = mtod(m, struct ip *);
1872 	struct sockaddr_in *ipforward_rtaddr;
1873 	struct rtentry *rt;
1874 	int error, type = 0, code = 0;
1875 	struct mbuf *mcopy;
1876 	n_long dest;
1877 	struct in_addr pkt_dst;
1878 	struct ifnet *destifp;
1879 	struct m_hdr tag;
1880 #if defined(IPSEC) || defined(FAST_IPSEC)
1881 	struct ifnet dummyifp;
1882 #endif
1883 
1884 	dest = INADDR_ANY;
1885 	/*
1886 	 * Cache the destination address of the packet; this may be
1887 	 * changed by use of 'ipfw fwd'.
1888 	 */
1889 	pkt_dst = (next_hop != NULL) ? next_hop->sin_addr : ip->ip_dst;
1890 
1891 #ifdef DIAGNOSTIC
1892 	if (ipprintfs)
1893 		printf("forward: src %x dst %x ttl %x\n",
1894 		       ip->ip_src.s_addr, pkt_dst.s_addr, ip->ip_ttl);
1895 #endif
1896 
1897 	if (m->m_flags & (M_BCAST | M_MCAST) || !in_canforward(pkt_dst)) {
1898 		ipstat.ips_cantforward++;
1899 		m_freem(m);
1900 		return;
1901 	}
1902 	if (!ipstealth && ip->ip_ttl <= IPTTLDEC) {
1903 		icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, NULL);
1904 		return;
1905 	}
1906 
1907 	ipforward_rtaddr = (struct sockaddr_in *) &ipforward_rt.ro_dst;
1908 	if (ipforward_rt.ro_rt == NULL ||
1909 	    ipforward_rtaddr->sin_addr.s_addr != pkt_dst.s_addr) {
1910 		if (ipforward_rt.ro_rt != NULL) {
1911 			RTFREE(ipforward_rt.ro_rt);
1912 			ipforward_rt.ro_rt = NULL;
1913 		}
1914 		ipforward_rtaddr->sin_family = AF_INET;
1915 		ipforward_rtaddr->sin_len = sizeof(struct sockaddr_in);
1916 		ipforward_rtaddr->sin_addr = pkt_dst;
1917 		rtalloc_ign(&ipforward_rt, RTF_PRCLONING);
1918 		if (ipforward_rt.ro_rt == NULL) {
1919 			icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest,
1920 				   NULL);
1921 			return;
1922 		}
1923 	}
1924 	rt = ipforward_rt.ro_rt;
1925 
1926 	/*
1927 	 * Save the IP header and at most 8 bytes of the payload,
1928 	 * in case we need to generate an ICMP message to the src.
1929 	 *
1930 	 * XXX this can be optimized a lot by saving the data in a local
1931 	 * buffer on the stack (72 bytes at most), and only allocating the
1932 	 * mbuf if really necessary. The vast majority of the packets
1933 	 * are forwarded without having to send an ICMP back (either
1934 	 * because unnecessary, or because rate limited), so we are
1935 	 * really we are wasting a lot of work here.
1936 	 *
1937 	 * We don't use m_copy() because it might return a reference
1938 	 * to a shared cluster. Both this function and ip_output()
1939 	 * assume exclusive access to the IP header in `m', so any
1940 	 * data in a cluster may change before we reach icmp_error().
1941 	 */
1942 	MGETHDR(mcopy, MB_DONTWAIT, m->m_type);
1943 	if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, MB_DONTWAIT)) {
1944 		/*
1945 		 * It's probably ok if the pkthdr dup fails (because
1946 		 * the deep copy of the tag chain failed), but for now
1947 		 * be conservative and just discard the copy since
1948 		 * code below may some day want the tags.
1949 		 */
1950 		m_free(mcopy);
1951 		mcopy = NULL;
1952 	}
1953 	if (mcopy != NULL) {
1954 		mcopy->m_len = imin((IP_VHL_HL(ip->ip_vhl) << 2) + 8,
1955 		    (int)ip->ip_len);
1956 		mcopy->m_pkthdr.len = mcopy->m_len;
1957 		m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1958 	}
1959 
1960 	if (!ipstealth)
1961 		ip->ip_ttl -= IPTTLDEC;
1962 
1963 	/*
1964 	 * If forwarding packet using same interface that it came in on,
1965 	 * perhaps should send a redirect to sender to shortcut a hop.
1966 	 * Only send redirect if source is sending directly to us,
1967 	 * and if packet was not source routed (or has any options).
1968 	 * Also, don't send redirect if forwarding using a default route
1969 	 * or a route modified by a redirect.
1970 	 */
1971 	if (rt->rt_ifp == m->m_pkthdr.rcvif &&
1972 	    !(rt->rt_flags & (RTF_DYNAMIC | RTF_MODIFIED)) &&
1973 	    satosin(rt_key(rt))->sin_addr.s_addr != INADDR_ANY &&
1974 	    ipsendredirects && !using_srcrt && next_hop == NULL) {
1975 		u_long src = ntohl(ip->ip_src.s_addr);
1976 		struct in_ifaddr *rt_ifa = (struct in_ifaddr *)rt->rt_ifa;
1977 
1978 		if (rt_ifa != NULL &&
1979 		    (src & rt_ifa->ia_subnetmask) == rt_ifa->ia_subnet) {
1980 			if (rt->rt_flags & RTF_GATEWAY)
1981 				dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
1982 			else
1983 				dest = pkt_dst.s_addr;
1984 			/*
1985 			 * Router requirements says to only send
1986 			 * host redirects.
1987 			 */
1988 			type = ICMP_REDIRECT;
1989 			code = ICMP_REDIRECT_HOST;
1990 #ifdef DIAGNOSTIC
1991 			if (ipprintfs)
1992 				printf("redirect (%d) to %x\n", code, dest);
1993 #endif
1994 		}
1995 	}
1996 
1997 	if (next_hop != NULL) {
1998 		/* Pass IPFORWARD info if available */
1999 		tag.mh_type = MT_TAG;
2000 		tag.mh_flags = PACKET_TAG_IPFORWARD;
2001 		tag.mh_data = (caddr_t)next_hop;
2002 		tag.mh_next = m;
2003 		m = (struct mbuf *)&tag;
2004 	}
2005 
2006 	error = ip_output(m, NULL, &ipforward_rt, IP_FORWARDING, NULL, NULL);
2007 	if (error == 0) {
2008 		ipstat.ips_forward++;
2009 		if (type == 0) {
2010 			if (mcopy) {
2011 				ipflow_create(&ipforward_rt, mcopy);
2012 				m_freem(mcopy);
2013 			}
2014 			return;		/* most common case */
2015 		} else {
2016 			ipstat.ips_redirectsent++;
2017 		}
2018 	} else {
2019 		ipstat.ips_cantforward++;
2020 	}
2021 
2022 	if (mcopy == NULL)
2023 		return;
2024 
2025 	/*
2026 	 * Send ICMP message.
2027 	 */
2028 	destifp = NULL;
2029 
2030 	switch (error) {
2031 
2032 	case 0:				/* forwarded, but need redirect */
2033 		/* type, code set above */
2034 		break;
2035 
2036 	case ENETUNREACH:		/* shouldn't happen, checked above */
2037 	case EHOSTUNREACH:
2038 	case ENETDOWN:
2039 	case EHOSTDOWN:
2040 	default:
2041 		type = ICMP_UNREACH;
2042 		code = ICMP_UNREACH_HOST;
2043 		break;
2044 
2045 	case EMSGSIZE:
2046 		type = ICMP_UNREACH;
2047 		code = ICMP_UNREACH_NEEDFRAG;
2048 #ifdef IPSEC
2049 		/*
2050 		 * If the packet is routed over IPsec tunnel, tell the
2051 		 * originator the tunnel MTU.
2052 		 *	tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
2053 		 * XXX quickhack!!!
2054 		 */
2055 		if (ipforward_rt.ro_rt != NULL) {
2056 			struct secpolicy *sp = NULL;
2057 			int ipsecerror;
2058 			int ipsechdr;
2059 			struct route *ro;
2060 
2061 			sp = ipsec4_getpolicybyaddr(mcopy,
2062 						    IPSEC_DIR_OUTBOUND,
2063 						    IP_FORWARDING,
2064 						    &ipsecerror);
2065 
2066 			if (sp == NULL)
2067 				destifp = ipforward_rt.ro_rt->rt_ifp;
2068 			else {
2069 				/* count IPsec header size */
2070 				ipsechdr = ipsec4_hdrsiz(mcopy,
2071 							 IPSEC_DIR_OUTBOUND,
2072 							 NULL);
2073 
2074 				/*
2075 				 * find the correct route for outer IPv4
2076 				 * header, compute tunnel MTU.
2077 				 *
2078 				 * XXX BUG ALERT
2079 				 * The "dummyifp" code relies upon the fact
2080 				 * that icmp_error() touches only ifp->if_mtu.
2081 				 */
2082 				/*XXX*/
2083 				destifp = NULL;
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 						dummyifp.if_mtu =
2090 						    ro->ro_rt->rt_ifp->if_mtu;
2091 						dummyifp.if_mtu -= ipsechdr;
2092 						destifp = &dummyifp;
2093 					}
2094 				}
2095 
2096 				key_freesp(sp);
2097 			}
2098 		}
2099 #elif FAST_IPSEC
2100 		/*
2101 		 * If the packet is routed over IPsec tunnel, tell the
2102 		 * originator the tunnel MTU.
2103 		 *	tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
2104 		 * XXX quickhack!!!
2105 		 */
2106 		if (ipforward_rt.ro_rt != NULL) {
2107 			struct secpolicy *sp = NULL;
2108 			int ipsecerror;
2109 			int ipsechdr;
2110 			struct route *ro;
2111 
2112 			sp = ipsec_getpolicybyaddr(mcopy,
2113 						   IPSEC_DIR_OUTBOUND,
2114 						   IP_FORWARDING,
2115 						   &ipsecerror);
2116 
2117 			if (sp == NULL)
2118 				destifp = ipforward_rt.ro_rt->rt_ifp;
2119 			else {
2120 				/* count IPsec header size */
2121 				ipsechdr = ipsec4_hdrsiz(mcopy,
2122 							 IPSEC_DIR_OUTBOUND,
2123 							 NULL);
2124 
2125 				/*
2126 				 * find the correct route for outer IPv4
2127 				 * header, compute tunnel MTU.
2128 				 *
2129 				 * XXX BUG ALERT
2130 				 * The "dummyifp" code relies upon the fact
2131 				 * that icmp_error() touches only ifp->if_mtu.
2132 				 */
2133 				/*XXX*/
2134 				destifp = NULL;
2135 				if (sp->req != NULL &&
2136 				    sp->req->sav != NULL &&
2137 				    sp->req->sav->sah != NULL) {
2138 					ro = &sp->req->sav->sah->sa_route;
2139 					if (ro->ro_rt != NULL &&
2140 					    ro->ro_rt->rt_ifp != NULL) {
2141 						dummyifp.if_mtu =
2142 						    ro->ro_rt->rt_ifp->if_mtu;
2143 						dummyifp.if_mtu -= ipsechdr;
2144 						destifp = &dummyifp;
2145 					}
2146 				}
2147 
2148 				KEY_FREESP(&sp);
2149 			}
2150 		}
2151 #else /* !IPSEC && !FAST_IPSEC */
2152 		if (ipforward_rt.ro_rt != NULL)
2153 			destifp = ipforward_rt.ro_rt->rt_ifp;
2154 #endif /*IPSEC*/
2155 		ipstat.ips_cantfrag++;
2156 		break;
2157 
2158 	case ENOBUFS:
2159 		/*
2160 		 * A router should not generate ICMP_SOURCEQUENCH as
2161 		 * required in RFC1812 Requirements for IP Version 4 Routers.
2162 		 * Source quench could be a big problem under DoS attacks,
2163 		 * or if the underlying interface is rate-limited.
2164 		 * Those who need source quench packets may re-enable them
2165 		 * via the net.inet.ip.sendsourcequench sysctl.
2166 		 */
2167 		if (!ip_sendsourcequench) {
2168 			m_freem(mcopy);
2169 			return;
2170 		} else {
2171 			type = ICMP_SOURCEQUENCH;
2172 			code = 0;
2173 		}
2174 		break;
2175 
2176 	case EACCES:			/* ipfw denied packet */
2177 		m_freem(mcopy);
2178 		return;
2179 	}
2180 	icmp_error(mcopy, type, code, dest, destifp);
2181 }
2182 
2183 void
2184 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
2185 	       struct mbuf *m)
2186 {
2187 	if (inp->inp_socket->so_options & SO_TIMESTAMP) {
2188 		struct timeval tv;
2189 
2190 		microtime(&tv);
2191 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
2192 		    SCM_TIMESTAMP, SOL_SOCKET);
2193 		if (*mp)
2194 			mp = &(*mp)->m_next;
2195 	}
2196 	if (inp->inp_flags & INP_RECVDSTADDR) {
2197 		*mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
2198 		    sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
2199 		if (*mp)
2200 			mp = &(*mp)->m_next;
2201 	}
2202 #ifdef notyet
2203 	/* XXX
2204 	 * Moving these out of udp_input() made them even more broken
2205 	 * than they already were.
2206 	 */
2207 	/* options were tossed already */
2208 	if (inp->inp_flags & INP_RECVOPTS) {
2209 		*mp = sbcreatecontrol((caddr_t) opts_deleted_above,
2210 		    sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
2211 		if (*mp)
2212 			mp = &(*mp)->m_next;
2213 	}
2214 	/* ip_srcroute doesn't do what we want here, need to fix */
2215 	if (inp->inp_flags & INP_RECVRETOPTS) {
2216 		*mp = sbcreatecontrol((caddr_t) ip_srcroute(),
2217 		    sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
2218 		if (*mp)
2219 			mp = &(*mp)->m_next;
2220 	}
2221 #endif
2222 	if (inp->inp_flags & INP_RECVIF) {
2223 		struct ifnet *ifp;
2224 		struct sdlbuf {
2225 			struct sockaddr_dl sdl;
2226 			u_char	pad[32];
2227 		} sdlbuf;
2228 		struct sockaddr_dl *sdp;
2229 		struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
2230 
2231 		if (((ifp = m->m_pkthdr.rcvif)) &&
2232 		    ((ifp->if_index != 0) && (ifp->if_index <= if_index))) {
2233 			sdp = IF_LLSOCKADDR(ifp);
2234 			/*
2235 			 * Change our mind and don't try copy.
2236 			 */
2237 			if ((sdp->sdl_family != AF_LINK) ||
2238 			    (sdp->sdl_len > sizeof(sdlbuf))) {
2239 				goto makedummy;
2240 			}
2241 			bcopy(sdp, sdl2, sdp->sdl_len);
2242 		} else {
2243 makedummy:
2244 			sdl2->sdl_len =
2245 			    offsetof(struct sockaddr_dl, sdl_data[0]);
2246 			sdl2->sdl_family = AF_LINK;
2247 			sdl2->sdl_index = 0;
2248 			sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
2249 		}
2250 		*mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len,
2251 			IP_RECVIF, IPPROTO_IP);
2252 		if (*mp)
2253 			mp = &(*mp)->m_next;
2254 	}
2255 }
2256 
2257 /*
2258  * XXX these routines are called from the upper part of the kernel.
2259  *
2260  * They could also be moved to ip_mroute.c, since all the RSVP
2261  *  handling is done there already.
2262  */
2263 int
2264 ip_rsvp_init(struct socket *so)
2265 {
2266 	if (so->so_type != SOCK_RAW ||
2267 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
2268 		return EOPNOTSUPP;
2269 
2270 	if (ip_rsvpd != NULL)
2271 		return EADDRINUSE;
2272 
2273 	ip_rsvpd = so;
2274 	/*
2275 	 * This may seem silly, but we need to be sure we don't over-increment
2276 	 * the RSVP counter, in case something slips up.
2277 	 */
2278 	if (!ip_rsvp_on) {
2279 		ip_rsvp_on = 1;
2280 		rsvp_on++;
2281 	}
2282 
2283 	return 0;
2284 }
2285 
2286 int
2287 ip_rsvp_done(void)
2288 {
2289 	ip_rsvpd = NULL;
2290 	/*
2291 	 * This may seem silly, but we need to be sure we don't over-decrement
2292 	 * the RSVP counter, in case something slips up.
2293 	 */
2294 	if (ip_rsvp_on) {
2295 		ip_rsvp_on = 0;
2296 		rsvp_on--;
2297 	}
2298 	return 0;
2299 }
2300 
2301 void
2302 rsvp_input(struct mbuf *m, ...)	/* XXX must fixup manually */
2303 {
2304 	int off, proto;
2305 	__va_list ap;
2306 
2307 	__va_start(ap, m);
2308 	off = __va_arg(ap, int);
2309 	proto = __va_arg(ap, int);
2310 	__va_end(ap);
2311 
2312 	if (rsvp_input_p) { /* call the real one if loaded */
2313 		rsvp_input_p(m, off, proto);
2314 		return;
2315 	}
2316 
2317 	/* Can still get packets with rsvp_on = 0 if there is a local member
2318 	 * of the group to which the RSVP packet is addressed.  But in this
2319 	 * case we want to throw the packet away.
2320 	 */
2321 
2322 	if (!rsvp_on) {
2323 		m_freem(m);
2324 		return;
2325 	}
2326 
2327 	if (ip_rsvpd != NULL) {
2328 		rip_input(m, off, proto);
2329 		return;
2330 	}
2331 	/* Drop the packet */
2332 	m_freem(m);
2333 }
2334