xref: /dflybsd-src/sys/netinet/ip_input.c (revision c666f28aa7c82e205ee3709528b79a41e8cc5308)
1 /*
2  * Copyright (c) 2003, 2004 Jeffrey M. Hsu.  All rights reserved.
3  * Copyright (c) 2003, 2004 The DragonFly Project.  All rights reserved.
4  *
5  * This code is derived from software contributed to The DragonFly Project
6  * by Jeffrey M. Hsu.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of The DragonFly Project nor the names of its
17  *    contributors may be used to endorse or promote products derived
18  *    from this software without specific, prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
21  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
22  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
23  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE
24  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
25  * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
26  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
27  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
28  * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
29  * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
30  * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  */
33 
34 /*
35  * Copyright (c) 1982, 1986, 1988, 1993
36  *	The Regents of the University of California.  All rights reserved.
37  *
38  * Redistribution and use in source and binary forms, with or without
39  * modification, are permitted provided that the following conditions
40  * are met:
41  * 1. Redistributions of source code must retain the above copyright
42  *    notice, this list of conditions and the following disclaimer.
43  * 2. Redistributions in binary form must reproduce the above copyright
44  *    notice, this list of conditions and the following disclaimer in the
45  *    documentation and/or other materials provided with the distribution.
46  * 3. All advertising materials mentioning features or use of this software
47  *    must display the following acknowledgement:
48  *	This product includes software developed by the University of
49  *	California, Berkeley and its contributors.
50  * 4. Neither the name of the University nor the names of its contributors
51  *    may be used to endorse or promote products derived from this software
52  *    without specific prior written permission.
53  *
54  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
55  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
56  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
57  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
58  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
59  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
60  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
61  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
62  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
63  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
64  * SUCH DAMAGE.
65  *
66  *	@(#)ip_input.c	8.2 (Berkeley) 1/4/94
67  * $FreeBSD: src/sys/netinet/ip_input.c,v 1.130.2.52 2003/03/07 07:01:28 silby Exp $
68  */
69 
70 #define	_IP_VHL
71 
72 #include "opt_bootp.h"
73 #include "opt_ipdn.h"
74 #include "opt_ipdivert.h"
75 #include "opt_ipstealth.h"
76 #include "opt_ipsec.h"
77 #include "opt_rss.h"
78 
79 #include <sys/param.h>
80 #include <sys/systm.h>
81 #include <sys/mbuf.h>
82 #include <sys/malloc.h>
83 #include <sys/mpipe.h>
84 #include <sys/domain.h>
85 #include <sys/protosw.h>
86 #include <sys/socket.h>
87 #include <sys/time.h>
88 #include <sys/globaldata.h>
89 #include <sys/thread.h>
90 #include <sys/kernel.h>
91 #include <sys/syslog.h>
92 #include <sys/sysctl.h>
93 #include <sys/in_cksum.h>
94 #include <sys/lock.h>
95 
96 #include <sys/mplock2.h>
97 
98 #include <machine/stdarg.h>
99 
100 #include <net/if.h>
101 #include <net/if_types.h>
102 #include <net/if_var.h>
103 #include <net/if_dl.h>
104 #include <net/pfil.h>
105 #include <net/route.h>
106 #include <net/netisr.h>
107 
108 #include <netinet/in.h>
109 #include <netinet/in_systm.h>
110 #include <netinet/in_var.h>
111 #include <netinet/ip.h>
112 #include <netinet/in_pcb.h>
113 #include <netinet/ip_var.h>
114 #include <netinet/ip_icmp.h>
115 #include <netinet/ip_divert.h>
116 #include <netinet/ip_flow.h>
117 
118 #include <sys/thread2.h>
119 #include <sys/msgport2.h>
120 #include <net/netmsg2.h>
121 
122 #include <sys/socketvar.h>
123 
124 #include <net/ipfw/ip_fw.h>
125 #include <net/dummynet/ip_dummynet.h>
126 
127 #ifdef IPSEC
128 #include <netinet6/ipsec.h>
129 #include <netproto/key/key.h>
130 #endif
131 
132 #ifdef FAST_IPSEC
133 #include <netproto/ipsec/ipsec.h>
134 #include <netproto/ipsec/key.h>
135 #endif
136 
137 int rsvp_on = 0;
138 static int ip_rsvp_on;
139 struct socket *ip_rsvpd;
140 
141 int ipforwarding = 0;
142 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW,
143     &ipforwarding, 0, "Enable IP forwarding between interfaces");
144 
145 static int ipsendredirects = 1; /* XXX */
146 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW,
147     &ipsendredirects, 0, "Enable sending IP redirects");
148 
149 int ip_defttl = IPDEFTTL;
150 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW,
151     &ip_defttl, 0, "Maximum TTL on IP packets");
152 
153 static int ip_dosourceroute = 0;
154 SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW,
155     &ip_dosourceroute, 0, "Enable forwarding source routed IP packets");
156 
157 static int ip_acceptsourceroute = 0;
158 SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute,
159     CTLFLAG_RW, &ip_acceptsourceroute, 0,
160     "Enable accepting source routed IP packets");
161 
162 static int ip_keepfaith = 0;
163 SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW,
164     &ip_keepfaith, 0,
165     "Enable packet capture for FAITH IPv4->IPv6 translator daemon");
166 
167 static int nipq = 0;	/* total # of reass queues */
168 static int maxnipq;
169 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragpackets, CTLFLAG_RW,
170     &maxnipq, 0,
171     "Maximum number of IPv4 fragment reassembly queue entries");
172 
173 static int maxfragsperpacket;
174 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW,
175     &maxfragsperpacket, 0,
176     "Maximum number of IPv4 fragments allowed per packet");
177 
178 static int ip_sendsourcequench = 0;
179 SYSCTL_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW,
180     &ip_sendsourcequench, 0,
181     "Enable the transmission of source quench packets");
182 
183 int ip_do_randomid = 1;
184 SYSCTL_INT(_net_inet_ip, OID_AUTO, random_id, CTLFLAG_RW,
185     &ip_do_randomid, 0,
186     "Assign random ip_id values");
187 /*
188  * XXX - Setting ip_checkinterface mostly implements the receive side of
189  * the Strong ES model described in RFC 1122, but since the routing table
190  * and transmit implementation do not implement the Strong ES model,
191  * setting this to 1 results in an odd hybrid.
192  *
193  * XXX - ip_checkinterface currently must be disabled if you use ipnat
194  * to translate the destination address to another local interface.
195  *
196  * XXX - ip_checkinterface must be disabled if you add IP aliases
197  * to the loopback interface instead of the interface where the
198  * packets for those addresses are received.
199  */
200 static int ip_checkinterface = 0;
201 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW,
202     &ip_checkinterface, 0, "Verify packet arrives on correct interface");
203 
204 static u_long ip_hash_count = 0;
205 SYSCTL_ULONG(_net_inet_ip, OID_AUTO, hash_count, CTLFLAG_RD,
206     &ip_hash_count, 0, "Number of packets hashed by IP");
207 
208 #ifdef RSS_DEBUG
209 static u_long ip_rehash_count = 0;
210 SYSCTL_ULONG(_net_inet_ip, OID_AUTO, rehash_count, CTLFLAG_RD,
211     &ip_rehash_count, 0, "Number of packets rehashed by IP");
212 
213 static u_long ip_dispatch_fast = 0;
214 SYSCTL_ULONG(_net_inet_ip, OID_AUTO, dispatch_fast_count, CTLFLAG_RD,
215     &ip_dispatch_fast, 0, "Number of packets handled on current CPU");
216 
217 static u_long ip_dispatch_slow = 0;
218 SYSCTL_ULONG(_net_inet_ip, OID_AUTO, dispatch_slow_count, CTLFLAG_RD,
219     &ip_dispatch_slow, 0, "Number of packets messaged to another CPU");
220 #endif
221 
222 static struct lwkt_token ipq_token = LWKT_TOKEN_INITIALIZER(ipq_token);
223 
224 #ifdef DIAGNOSTIC
225 static int ipprintfs = 0;
226 #endif
227 
228 extern	struct domain inetdomain;
229 extern	struct protosw inetsw[];
230 u_char	ip_protox[IPPROTO_MAX];
231 struct	in_ifaddrhead in_ifaddrheads[MAXCPU];	/* first inet address */
232 struct	in_ifaddrhashhead *in_ifaddrhashtbls[MAXCPU];
233 						/* inet addr hash table */
234 u_long	in_ifaddrhmask;				/* mask for hash table */
235 
236 struct ip_stats ipstats_percpu[MAXCPU];
237 #ifdef SMP
238 static int
239 sysctl_ipstats(SYSCTL_HANDLER_ARGS)
240 {
241 	int cpu, error = 0;
242 
243 	for (cpu = 0; cpu < ncpus; ++cpu) {
244 		if ((error = SYSCTL_OUT(req, &ipstats_percpu[cpu],
245 					sizeof(struct ip_stats))))
246 			break;
247 		if ((error = SYSCTL_IN(req, &ipstats_percpu[cpu],
248 				       sizeof(struct ip_stats))))
249 			break;
250 	}
251 
252 	return (error);
253 }
254 SYSCTL_PROC(_net_inet_ip, IPCTL_STATS, stats, (CTLTYPE_OPAQUE | CTLFLAG_RW),
255     0, 0, sysctl_ipstats, "S,ip_stats", "IP statistics");
256 #else
257 SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW,
258     &ipstat, ip_stats, "IP statistics");
259 #endif
260 
261 /* Packet reassembly stuff */
262 #define	IPREASS_NHASH_LOG2	6
263 #define	IPREASS_NHASH		(1 << IPREASS_NHASH_LOG2)
264 #define	IPREASS_HMASK		(IPREASS_NHASH - 1)
265 #define	IPREASS_HASH(x,y)						\
266     (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK)
267 
268 static TAILQ_HEAD(ipqhead, ipq) ipq[IPREASS_NHASH];
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 struct mbuf *(*ip_divert_p)(struct mbuf *, int, int);
283 
284 struct pfil_head inet_pfil_hook;
285 
286 /*
287  * struct ip_srcrt_opt is used to store packet state while it travels
288  * through the stack.
289  *
290  * XXX Note that the code even makes assumptions on the size and
291  * alignment of fields inside struct ip_srcrt so e.g. adding some
292  * fields will break the code.  This needs to be fixed.
293  *
294  * We need to save the IP options in case a protocol wants to respond
295  * to an incoming packet over the same route if the packet got here
296  * using IP source routing.  This allows connection establishment and
297  * maintenance when the remote end is on a network that is not known
298  * to us.
299  */
300 struct ip_srcrt {
301 	struct	in_addr dst;			/* final destination */
302 	char	nop;				/* one NOP to align */
303 	char	srcopt[IPOPT_OFFSET + 1];	/* OPTVAL, OLEN and OFFSET */
304 	struct	in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)];
305 };
306 
307 struct ip_srcrt_opt {
308 	int		ip_nhops;
309 	struct ip_srcrt	ip_srcrt;
310 };
311 
312 static MALLOC_DEFINE(M_IPQ, "ipq", "IP Fragment Management");
313 static struct malloc_pipe ipq_mpipe;
314 
315 static void		save_rte(struct mbuf *, u_char *, struct in_addr);
316 static int		ip_dooptions(struct mbuf *m, int, struct sockaddr_in *);
317 static void		ip_freef(struct ipqhead *, struct ipq *);
318 static void		ip_input_handler(netmsg_t);
319 
320 /*
321  * IP initialization: fill in IP protocol switch table.
322  * All protocols not implemented in kernel go to raw IP protocol handler.
323  */
324 void
325 ip_init(void)
326 {
327 	struct protosw *pr;
328 	int i;
329 #ifdef SMP
330 	int cpu;
331 #endif
332 
333 	/*
334 	 * Make sure we can handle a reasonable number of fragments but
335 	 * cap it at 4000 (XXX).
336 	 */
337 	mpipe_init(&ipq_mpipe, M_IPQ, sizeof(struct ipq),
338 		    IFQ_MAXLEN, 4000, 0, NULL, NULL, NULL);
339 	for (i = 0; i < ncpus; ++i) {
340 		TAILQ_INIT(&in_ifaddrheads[i]);
341 		in_ifaddrhashtbls[i] =
342 			hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask);
343 	}
344 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
345 	if (pr == NULL)
346 		panic("ip_init");
347 	for (i = 0; i < IPPROTO_MAX; i++)
348 		ip_protox[i] = pr - inetsw;
349 	for (pr = inetdomain.dom_protosw;
350 	     pr < inetdomain.dom_protoswNPROTOSW; pr++) {
351 		if (pr->pr_domain->dom_family == PF_INET && pr->pr_protocol) {
352 			if (pr->pr_protocol != IPPROTO_RAW)
353 				ip_protox[pr->pr_protocol] = pr - inetsw;
354 		}
355 	}
356 
357 	inet_pfil_hook.ph_type = PFIL_TYPE_AF;
358 	inet_pfil_hook.ph_af = AF_INET;
359 	if ((i = pfil_head_register(&inet_pfil_hook)) != 0) {
360 		kprintf("%s: WARNING: unable to register pfil hook, "
361 			"error %d\n", __func__, i);
362 	}
363 
364 	for (i = 0; i < IPREASS_NHASH; i++)
365 		TAILQ_INIT(&ipq[i]);
366 
367 	maxnipq = nmbclusters / 32;
368 	maxfragsperpacket = 16;
369 
370 	ip_id = time_second & 0xffff;
371 
372 	/*
373 	 * Initialize IP statistics counters for each CPU.
374 	 *
375 	 */
376 #ifdef SMP
377 	for (cpu = 0; cpu < ncpus; ++cpu) {
378 		bzero(&ipstats_percpu[cpu], sizeof(struct ip_stats));
379 	}
380 #else
381 	bzero(&ipstat, sizeof(struct ip_stats));
382 #endif
383 
384 	netisr_register(NETISR_IP, ip_input_handler, ip_cpufn_in);
385 	netisr_register_hashcheck(NETISR_IP, ip_hashcheck);
386 }
387 
388 /* Do transport protocol processing. */
389 static void
390 transport_processing_oncpu(struct mbuf *m, int hlen, struct ip *ip)
391 {
392 	const struct protosw *pr = &inetsw[ip_protox[ip->ip_p]];
393 
394 	/*
395 	 * Switch out to protocol's input routine.
396 	 */
397 	PR_GET_MPLOCK(pr);
398 	pr->pr_input(&m, &hlen, ip->ip_p);
399 	PR_REL_MPLOCK(pr);
400 }
401 
402 static void
403 transport_processing_handler(netmsg_t msg)
404 {
405 	struct netmsg_packet *pmsg = &msg->packet;
406 	struct ip *ip;
407 	int hlen;
408 
409 	ip = mtod(pmsg->nm_packet, struct ip *);
410 	hlen = pmsg->base.lmsg.u.ms_result;
411 
412 	transport_processing_oncpu(pmsg->nm_packet, hlen, ip);
413 	/* msg was embedded in the mbuf, do not reply! */
414 }
415 
416 static void
417 ip_input_handler(netmsg_t msg)
418 {
419 	ip_input(msg->packet.nm_packet);
420 	/* msg was embedded in the mbuf, do not reply! */
421 }
422 
423 /*
424  * IP input routine.  Checksum and byte swap header.  If fragmented
425  * try to reassemble.  Process options.  Pass to next level.
426  */
427 void
428 ip_input(struct mbuf *m)
429 {
430 	struct ip *ip;
431 	struct in_ifaddr *ia = NULL;
432 	struct in_ifaddr_container *iac;
433 	int hlen, checkif;
434 	u_short sum;
435 	struct in_addr pkt_dst;
436 	boolean_t using_srcrt = FALSE;		/* forward (by PFIL_HOOKS) */
437 	struct in_addr odst;			/* original dst address(NAT) */
438 	struct m_tag *mtag;
439 	struct sockaddr_in *next_hop = NULL;
440 	lwkt_port_t port;
441 #ifdef FAST_IPSEC
442 	struct tdb_ident *tdbi;
443 	struct secpolicy *sp;
444 	int error;
445 #endif
446 
447 	M_ASSERTPKTHDR(m);
448 
449 	/*
450 	 * This routine is called from numerous places which may not have
451 	 * characterized the packet.
452 	 */
453 	if ((m->m_flags & M_HASH) == 0) {
454 		atomic_add_long(&ip_hash_count, 1);
455 		ip_cpufn(&m, 0, IP_MPORT_IN);
456 		if (m == NULL)
457 			return;
458 		KKASSERT(m->m_flags & M_HASH);
459 	}
460 	ip = mtod(m, struct ip *);
461 
462 	/*
463 	 * Pull out certain tags
464 	 */
465 	if (m->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) {
466 		/* Next hop */
467 		mtag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL);
468 		KKASSERT(mtag != NULL);
469 		next_hop = m_tag_data(mtag);
470 	}
471 
472 	if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
473 		/* dummynet already filtered us */
474 		ip = mtod(m, struct ip *);
475 		hlen = IP_VHL_HL(ip->ip_vhl) << 2;
476 		goto iphack;
477 	}
478 
479 	ipstat.ips_total++;
480 
481 	/* length checks already done in ip_cpufn() */
482 	KASSERT(m->m_len >= sizeof(struct ip), ("IP header not in one mbuf"));
483 
484 	if (IP_VHL_V(ip->ip_vhl) != IPVERSION) {
485 		ipstat.ips_badvers++;
486 		goto bad;
487 	}
488 
489 	hlen = IP_VHL_HL(ip->ip_vhl) << 2;
490 	/* length checks already done in ip_cpufn() */
491 	KASSERT(hlen >= sizeof(struct ip), ("IP header len too small"));
492 	KASSERT(m->m_len >= hlen, ("complete IP header not in one mbuf"));
493 
494 	/* 127/8 must not appear on wire - RFC1122 */
495 	if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
496 	    (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
497 		if (!(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK)) {
498 			ipstat.ips_badaddr++;
499 			goto bad;
500 		}
501 	}
502 
503 	if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
504 		sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
505 	} else {
506 		if (hlen == sizeof(struct ip))
507 			sum = in_cksum_hdr(ip);
508 		else
509 			sum = in_cksum(m, hlen);
510 	}
511 	if (sum != 0) {
512 		ipstat.ips_badsum++;
513 		goto bad;
514 	}
515 
516 #ifdef ALTQ
517 	if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) {
518 		/* packet is dropped by traffic conditioner */
519 		return;
520 	}
521 #endif
522 	/*
523 	 * Convert fields to host representation.
524 	 */
525 	ip->ip_len = ntohs(ip->ip_len);
526 	ip->ip_off = ntohs(ip->ip_off);
527 
528 	/* length checks already done in ip_cpufn() */
529 	KASSERT(ip->ip_len >= hlen, ("total length less then header length"));
530 	KASSERT(m->m_pkthdr.len >= ip->ip_len, ("mbuf too short"));
531 
532 	/*
533 	 * Trim mbufs if longer than the IP header would have us expect.
534 	 */
535 	if (m->m_pkthdr.len > ip->ip_len) {
536 		if (m->m_len == m->m_pkthdr.len) {
537 			m->m_len = ip->ip_len;
538 			m->m_pkthdr.len = ip->ip_len;
539 		} else {
540 			m_adj(m, ip->ip_len - m->m_pkthdr.len);
541 		}
542 	}
543 #if defined(IPSEC) && !defined(IPSEC_FILTERGIF)
544 	/*
545 	 * Bypass packet filtering for packets from a tunnel (gif).
546 	 */
547 	if (ipsec_gethist(m, NULL))
548 		goto pass;
549 #endif
550 
551 	/*
552 	 * IpHack's section.
553 	 * Right now when no processing on packet has done
554 	 * and it is still fresh out of network we do our black
555 	 * deals with it.
556 	 * - Firewall: deny/allow/divert
557 	 * - Xlate: translate packet's addr/port (NAT).
558 	 * - Pipe: pass pkt through dummynet.
559 	 * - Wrap: fake packet's addr/port <unimpl.>
560 	 * - Encapsulate: put it in another IP and send out. <unimp.>
561 	 */
562 
563 iphack:
564 	/*
565 	 * If we've been forwarded from the output side, then
566 	 * skip the firewall a second time
567 	 */
568 	if (next_hop != NULL)
569 		goto ours;
570 
571 	/* No pfil hooks */
572 	if (!pfil_has_hooks(&inet_pfil_hook)) {
573 		if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
574 			/*
575 			 * Strip dummynet tags from stranded packets
576 			 */
577 			mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
578 			KKASSERT(mtag != NULL);
579 			m_tag_delete(m, mtag);
580 			m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED;
581 		}
582 		goto pass;
583 	}
584 
585 	/*
586 	 * Run through list of hooks for input packets.
587 	 *
588 	 * NOTE!  If the packet is rewritten pf/ipfw/whoever must
589 	 *	  clear M_HASH.
590 	 */
591 	odst = ip->ip_dst;
592 	if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif, PFIL_IN))
593 		return;
594 	if (m == NULL)	/* consumed by filter */
595 		return;
596 	ip = mtod(m, struct ip *);
597 	hlen = IP_VHL_HL(ip->ip_vhl) << 2;
598 	using_srcrt = (odst.s_addr != ip->ip_dst.s_addr);
599 
600 	if (m->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) {
601 		mtag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL);
602 		KKASSERT(mtag != NULL);
603 		next_hop = m_tag_data(mtag);
604 	}
605 	if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
606 		ip_dn_queue(m);
607 		return;
608 	}
609 	if (m->m_pkthdr.fw_flags & FW_MBUF_REDISPATCH) {
610 		m->m_pkthdr.fw_flags &= ~FW_MBUF_REDISPATCH;
611 	}
612 pass:
613 	/*
614 	 * Process options and, if not destined for us,
615 	 * ship it on.  ip_dooptions returns 1 when an
616 	 * error was detected (causing an icmp message
617 	 * to be sent and the original packet to be freed).
618 	 */
619 	if (hlen > sizeof(struct ip) && ip_dooptions(m, 0, next_hop))
620 		return;
621 
622 	/* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
623 	 * matter if it is destined to another node, or whether it is
624 	 * a multicast one, RSVP wants it! and prevents it from being forwarded
625 	 * anywhere else. Also checks if the rsvp daemon is running before
626 	 * grabbing the packet.
627 	 */
628 	if (rsvp_on && ip->ip_p == IPPROTO_RSVP)
629 		goto ours;
630 
631 	/*
632 	 * Check our list of addresses, to see if the packet is for us.
633 	 * If we don't have any addresses, assume any unicast packet
634 	 * we receive might be for us (and let the upper layers deal
635 	 * with it).
636 	 */
637 	if (TAILQ_EMPTY(&in_ifaddrheads[mycpuid]) &&
638 	    !(m->m_flags & (M_MCAST | M_BCAST)))
639 		goto ours;
640 
641 	/*
642 	 * Cache the destination address of the packet; this may be
643 	 * changed by use of 'ipfw fwd'.
644 	 */
645 	pkt_dst = next_hop ? next_hop->sin_addr : ip->ip_dst;
646 
647 	/*
648 	 * Enable a consistency check between the destination address
649 	 * and the arrival interface for a unicast packet (the RFC 1122
650 	 * strong ES model) if IP forwarding is disabled and the packet
651 	 * is not locally generated and the packet is not subject to
652 	 * 'ipfw fwd'.
653 	 *
654 	 * XXX - Checking also should be disabled if the destination
655 	 * address is ipnat'ed to a different interface.
656 	 *
657 	 * XXX - Checking is incompatible with IP aliases added
658 	 * to the loopback interface instead of the interface where
659 	 * the packets are received.
660 	 */
661 	checkif = ip_checkinterface &&
662 		  !ipforwarding &&
663 		  m->m_pkthdr.rcvif != NULL &&
664 		  !(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) &&
665 		  next_hop == NULL;
666 
667 	/*
668 	 * Check for exact addresses in the hash bucket.
669 	 */
670 	LIST_FOREACH(iac, INADDR_HASH(pkt_dst.s_addr), ia_hash) {
671 		ia = iac->ia;
672 
673 		/*
674 		 * If the address matches, verify that the packet
675 		 * arrived via the correct interface if checking is
676 		 * enabled.
677 		 */
678 		if (IA_SIN(ia)->sin_addr.s_addr == pkt_dst.s_addr &&
679 		    (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif))
680 			goto ours;
681 	}
682 	ia = NULL;
683 
684 	/*
685 	 * Check for broadcast addresses.
686 	 *
687 	 * Only accept broadcast packets that arrive via the matching
688 	 * interface.  Reception of forwarded directed broadcasts would
689 	 * be handled via ip_forward() and ether_output() with the loopback
690 	 * into the stack for SIMPLEX interfaces handled by ether_output().
691 	 */
692 	if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
693 		struct ifaddr_container *ifac;
694 
695 		TAILQ_FOREACH(ifac, &m->m_pkthdr.rcvif->if_addrheads[mycpuid],
696 			      ifa_link) {
697 			struct ifaddr *ifa = ifac->ifa;
698 
699 			if (ifa->ifa_addr == NULL) /* shutdown/startup race */
700 				continue;
701 			if (ifa->ifa_addr->sa_family != AF_INET)
702 				continue;
703 			ia = ifatoia(ifa);
704 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
705 								pkt_dst.s_addr)
706 				goto ours;
707 			if (ia->ia_netbroadcast.s_addr == pkt_dst.s_addr)
708 				goto ours;
709 #ifdef BOOTP_COMPAT
710 			if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY)
711 				goto ours;
712 #endif
713 		}
714 	}
715 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
716 		struct in_multi *inm;
717 
718 		/* XXX Multicast is not MPSAFE yet */
719 		get_mplock();
720 
721 		if (ip_mrouter != NULL) {
722 			/*
723 			 * If we are acting as a multicast router, all
724 			 * incoming multicast packets are passed to the
725 			 * kernel-level multicast forwarding function.
726 			 * The packet is returned (relatively) intact; if
727 			 * ip_mforward() returns a non-zero value, the packet
728 			 * must be discarded, else it may be accepted below.
729 			 */
730 			if (ip_mforward != NULL &&
731 			    ip_mforward(ip, m->m_pkthdr.rcvif, m, NULL) != 0) {
732 				rel_mplock();
733 				ipstat.ips_cantforward++;
734 				m_freem(m);
735 				return;
736 			}
737 
738 			/*
739 			 * The process-level routing daemon needs to receive
740 			 * all multicast IGMP packets, whether or not this
741 			 * host belongs to their destination groups.
742 			 */
743 			if (ip->ip_p == IPPROTO_IGMP) {
744 				rel_mplock();
745 				goto ours;
746 			}
747 			ipstat.ips_forward++;
748 		}
749 		/*
750 		 * See if we belong to the destination multicast group on the
751 		 * arrival interface.
752 		 */
753 		IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
754 		if (inm == NULL) {
755 			rel_mplock();
756 			ipstat.ips_notmember++;
757 			m_freem(m);
758 			return;
759 		}
760 
761 		rel_mplock();
762 		goto ours;
763 	}
764 	if (ip->ip_dst.s_addr == INADDR_BROADCAST)
765 		goto ours;
766 	if (ip->ip_dst.s_addr == INADDR_ANY)
767 		goto ours;
768 
769 	/*
770 	 * FAITH(Firewall Aided Internet Translator)
771 	 */
772 	if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
773 		if (ip_keepfaith) {
774 			if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP)
775 				goto ours;
776 		}
777 		m_freem(m);
778 		return;
779 	}
780 
781 	/*
782 	 * Not for us; forward if possible and desirable.
783 	 */
784 	if (!ipforwarding) {
785 		ipstat.ips_cantforward++;
786 		m_freem(m);
787 	} else {
788 #ifdef IPSEC
789 		/*
790 		 * Enforce inbound IPsec SPD.
791 		 */
792 		if (ipsec4_in_reject(m, NULL)) {
793 			ipsecstat.in_polvio++;
794 			goto bad;
795 		}
796 #endif
797 #ifdef FAST_IPSEC
798 		mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
799 		crit_enter();
800 		if (mtag != NULL) {
801 			tdbi = (struct tdb_ident *)m_tag_data(mtag);
802 			sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
803 		} else {
804 			sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
805 						   IP_FORWARDING, &error);
806 		}
807 		if (sp == NULL) {	/* NB: can happen if error */
808 			crit_exit();
809 			/*XXX error stat???*/
810 			DPRINTF(("ip_input: no SP for forwarding\n"));	/*XXX*/
811 			goto bad;
812 		}
813 
814 		/*
815 		 * Check security policy against packet attributes.
816 		 */
817 		error = ipsec_in_reject(sp, m);
818 		KEY_FREESP(&sp);
819 		crit_exit();
820 		if (error) {
821 			ipstat.ips_cantforward++;
822 			goto bad;
823 		}
824 #endif
825 		ip_forward(m, using_srcrt, next_hop);
826 	}
827 	return;
828 
829 ours:
830 
831 	/*
832 	 * IPSTEALTH: Process non-routing options only
833 	 * if the packet is destined for us.
834 	 */
835 	if (ipstealth &&
836 	    hlen > sizeof(struct ip) &&
837 	    ip_dooptions(m, 1, next_hop))
838 		return;
839 
840 	/* Count the packet in the ip address stats */
841 	if (ia != NULL) {
842 		ia->ia_ifa.if_ipackets++;
843 		ia->ia_ifa.if_ibytes += m->m_pkthdr.len;
844 	}
845 
846 	/*
847 	 * If offset or IP_MF are set, must reassemble.
848 	 * Otherwise, nothing need be done.
849 	 * (We could look in the reassembly queue to see
850 	 * if the packet was previously fragmented,
851 	 * but it's not worth the time; just let them time out.)
852 	 */
853 	if (ip->ip_off & (IP_MF | IP_OFFMASK)) {
854 		/*
855 		 * Attempt reassembly; if it succeeds, proceed.  ip_reass()
856 		 * will return a different mbuf.
857 		 *
858 		 * NOTE: ip_reass() returns m with M_HASH cleared to force
859 		 *	 us to recharacterize the packet.
860 		 */
861 		m = ip_reass(m);
862 		if (m == NULL)
863 			return;
864 		ip = mtod(m, struct ip *);
865 
866 		/* Get the header length of the reassembled packet */
867 		hlen = IP_VHL_HL(ip->ip_vhl) << 2;
868 	} else {
869 		ip->ip_len -= hlen;
870 	}
871 
872 #ifdef IPSEC
873 	/*
874 	 * enforce IPsec policy checking if we are seeing last header.
875 	 * note that we do not visit this with protocols with pcb layer
876 	 * code - like udp/tcp/raw ip.
877 	 */
878 	if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) &&
879 	    ipsec4_in_reject(m, NULL)) {
880 		ipsecstat.in_polvio++;
881 		goto bad;
882 	}
883 #endif
884 #if FAST_IPSEC
885 	/*
886 	 * enforce IPsec policy checking if we are seeing last header.
887 	 * note that we do not visit this with protocols with pcb layer
888 	 * code - like udp/tcp/raw ip.
889 	 */
890 	if (inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) {
891 		/*
892 		 * Check if the packet has already had IPsec processing
893 		 * done.  If so, then just pass it along.  This tag gets
894 		 * set during AH, ESP, etc. input handling, before the
895 		 * packet is returned to the ip input queue for delivery.
896 		 */
897 		mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
898 		crit_enter();
899 		if (mtag != NULL) {
900 			tdbi = (struct tdb_ident *)m_tag_data(mtag);
901 			sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
902 		} else {
903 			sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
904 						   IP_FORWARDING, &error);
905 		}
906 		if (sp != NULL) {
907 			/*
908 			 * Check security policy against packet attributes.
909 			 */
910 			error = ipsec_in_reject(sp, m);
911 			KEY_FREESP(&sp);
912 		} else {
913 			/* XXX error stat??? */
914 			error = EINVAL;
915 DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/
916 			goto bad;
917 		}
918 		crit_exit();
919 		if (error)
920 			goto bad;
921 	}
922 #endif /* FAST_IPSEC */
923 
924 	/*
925 	 * We must forward the packet to the correct protocol thread if
926 	 * we are not already in it.
927 	 *
928 	 * NOTE: ip_len is now in host form.  ip_len is not adjusted
929 	 *	 further for protocol processing, instead we pass hlen
930 	 *	 to the protosw and let it deal with it.
931 	 */
932 	ipstat.ips_delivered++;
933 
934 	if ((m->m_flags & M_HASH) == 0) {
935 #ifdef RSS_DEBUG
936 		atomic_add_long(&ip_rehash_count, 1);
937 #endif
938 		ip->ip_len = htons(ip->ip_len + hlen);
939 		ip->ip_off = htons(ip->ip_off);
940 
941 		ip_cpufn(&m, 0, IP_MPORT_IN);
942 		if (m == NULL)
943 			return;
944 
945 		ip = mtod(m, struct ip *);
946 		ip->ip_len = ntohs(ip->ip_len) - hlen;
947 		ip->ip_off = ntohs(ip->ip_off);
948 		KKASSERT(m->m_flags & M_HASH);
949 	}
950 	port = netisr_portfn(m->m_pkthdr.hash);
951 
952 	if (port != &curthread->td_msgport) {
953 		struct netmsg_packet *pmsg;
954 
955 #ifdef RSS_DEBUG
956 		atomic_add_long(&ip_dispatch_slow, 1);
957 #endif
958 
959 		pmsg = &m->m_hdr.mh_netmsg;
960 		netmsg_init(&pmsg->base, NULL, &netisr_apanic_rport,
961 			    0, transport_processing_handler);
962 		pmsg->nm_packet = m;
963 		pmsg->base.lmsg.u.ms_result = hlen;
964 		lwkt_sendmsg(port, &pmsg->base.lmsg);
965 	} else {
966 #ifdef RSS_DEBUG
967 		atomic_add_long(&ip_dispatch_fast, 1);
968 #endif
969 		transport_processing_oncpu(m, hlen, ip);
970 	}
971 	return;
972 
973 bad:
974 	m_freem(m);
975 }
976 
977 /*
978  * Take incoming datagram fragment and try to reassemble it into
979  * whole datagram.  If a chain for reassembly of this datagram already
980  * exists, then it is given as fp; otherwise have to make a chain.
981  */
982 struct mbuf *
983 ip_reass(struct mbuf *m)
984 {
985 	struct ip *ip = mtod(m, struct ip *);
986 	struct mbuf *p = NULL, *q, *nq;
987 	struct mbuf *n;
988 	struct ipq *fp = NULL;
989 	struct ipqhead *head;
990 	int hlen = IP_VHL_HL(ip->ip_vhl) << 2;
991 	int i, next;
992 	u_short sum;
993 
994 	/* If maxnipq is 0, never accept fragments. */
995 	if (maxnipq == 0) {
996 		ipstat.ips_fragments++;
997 		ipstat.ips_fragdropped++;
998 		m_freem(m);
999 		return NULL;
1000 	}
1001 
1002 	sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
1003 	/*
1004 	 * Look for queue of fragments of this datagram.
1005 	 */
1006 	lwkt_gettoken(&ipq_token);
1007 	head = &ipq[sum];
1008 	TAILQ_FOREACH(fp, head, ipq_list) {
1009 		if (ip->ip_id == fp->ipq_id &&
1010 		    ip->ip_src.s_addr == fp->ipq_src.s_addr &&
1011 		    ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
1012 		    ip->ip_p == fp->ipq_p)
1013 			goto found;
1014 	}
1015 
1016 	fp = NULL;
1017 
1018 	/*
1019 	 * Enforce upper bound on number of fragmented packets
1020 	 * for which we attempt reassembly;
1021 	 * If maxnipq is -1, accept all fragments without limitation.
1022 	 */
1023 	if (nipq > maxnipq && maxnipq > 0) {
1024 		/*
1025 		 * drop something from the tail of the current queue
1026 		 * before proceeding further
1027 		 */
1028 		struct ipq *q = TAILQ_LAST(head, ipqhead);
1029 		if (q == NULL) {
1030 			/*
1031 			 * The current queue is empty,
1032 			 * so drop from one of the others.
1033 			 */
1034 			for (i = 0; i < IPREASS_NHASH; i++) {
1035 				struct ipq *r = TAILQ_LAST(&ipq[i], ipqhead);
1036 				if (r) {
1037 					ipstat.ips_fragtimeout += r->ipq_nfrags;
1038 					ip_freef(&ipq[i], r);
1039 					break;
1040 				}
1041 			}
1042 		} else {
1043 			ipstat.ips_fragtimeout += q->ipq_nfrags;
1044 			ip_freef(head, q);
1045 		}
1046 	}
1047 found:
1048 	/*
1049 	 * Adjust ip_len to not reflect header,
1050 	 * convert offset of this to bytes.
1051 	 */
1052 	ip->ip_len -= hlen;
1053 	if (ip->ip_off & IP_MF) {
1054 		/*
1055 		 * Make sure that fragments have a data length
1056 		 * that's a non-zero multiple of 8 bytes.
1057 		 */
1058 		if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
1059 			ipstat.ips_toosmall++; /* XXX */
1060 			m_freem(m);
1061 			goto done;
1062 		}
1063 		m->m_flags |= M_FRAG;
1064 	} else {
1065 		m->m_flags &= ~M_FRAG;
1066 	}
1067 	ip->ip_off <<= 3;
1068 
1069 	ipstat.ips_fragments++;
1070 	m->m_pkthdr.header = ip;
1071 
1072 	/*
1073 	 * If the hardware has not done csum over this fragment
1074 	 * then csum_data is not valid at all.
1075 	 */
1076 	if ((m->m_pkthdr.csum_flags & (CSUM_FRAG_NOT_CHECKED | CSUM_DATA_VALID))
1077 	    == (CSUM_FRAG_NOT_CHECKED | CSUM_DATA_VALID)) {
1078 		m->m_pkthdr.csum_data = 0;
1079 		m->m_pkthdr.csum_flags &= ~(CSUM_DATA_VALID | CSUM_PSEUDO_HDR);
1080 	}
1081 
1082 	/*
1083 	 * Presence of header sizes in mbufs
1084 	 * would confuse code below.
1085 	 */
1086 	m->m_data += hlen;
1087 	m->m_len -= hlen;
1088 
1089 	/*
1090 	 * If first fragment to arrive, create a reassembly queue.
1091 	 */
1092 	if (fp == NULL) {
1093 		if ((fp = mpipe_alloc_nowait(&ipq_mpipe)) == NULL)
1094 			goto dropfrag;
1095 		TAILQ_INSERT_HEAD(head, fp, ipq_list);
1096 		nipq++;
1097 		fp->ipq_nfrags = 1;
1098 		fp->ipq_ttl = IPFRAGTTL;
1099 		fp->ipq_p = ip->ip_p;
1100 		fp->ipq_id = ip->ip_id;
1101 		fp->ipq_src = ip->ip_src;
1102 		fp->ipq_dst = ip->ip_dst;
1103 		fp->ipq_frags = m;
1104 		m->m_nextpkt = NULL;
1105 		goto inserted;
1106 	} else {
1107 		fp->ipq_nfrags++;
1108 	}
1109 
1110 #define	GETIP(m)	((struct ip*)((m)->m_pkthdr.header))
1111 
1112 	/*
1113 	 * Find a segment which begins after this one does.
1114 	 */
1115 	for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
1116 		if (GETIP(q)->ip_off > ip->ip_off)
1117 			break;
1118 	}
1119 
1120 	/*
1121 	 * If there is a preceding segment, it may provide some of
1122 	 * our data already.  If so, drop the data from the incoming
1123 	 * segment.  If it provides all of our data, drop us, otherwise
1124 	 * stick new segment in the proper place.
1125 	 *
1126 	 * If some of the data is dropped from the the preceding
1127 	 * segment, then it's checksum is invalidated.
1128 	 */
1129 	if (p) {
1130 		i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off;
1131 		if (i > 0) {
1132 			if (i >= ip->ip_len)
1133 				goto dropfrag;
1134 			m_adj(m, i);
1135 			m->m_pkthdr.csum_flags = 0;
1136 			ip->ip_off += i;
1137 			ip->ip_len -= i;
1138 		}
1139 		m->m_nextpkt = p->m_nextpkt;
1140 		p->m_nextpkt = m;
1141 	} else {
1142 		m->m_nextpkt = fp->ipq_frags;
1143 		fp->ipq_frags = m;
1144 	}
1145 
1146 	/*
1147 	 * While we overlap succeeding segments trim them or,
1148 	 * if they are completely covered, dequeue them.
1149 	 */
1150 	for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off;
1151 	     q = nq) {
1152 		i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off;
1153 		if (i < GETIP(q)->ip_len) {
1154 			GETIP(q)->ip_len -= i;
1155 			GETIP(q)->ip_off += i;
1156 			m_adj(q, i);
1157 			q->m_pkthdr.csum_flags = 0;
1158 			break;
1159 		}
1160 		nq = q->m_nextpkt;
1161 		m->m_nextpkt = nq;
1162 		ipstat.ips_fragdropped++;
1163 		fp->ipq_nfrags--;
1164 		q->m_nextpkt = NULL;
1165 		m_freem(q);
1166 	}
1167 
1168 inserted:
1169 	/*
1170 	 * Check for complete reassembly and perform frag per packet
1171 	 * limiting.
1172 	 *
1173 	 * Frag limiting is performed here so that the nth frag has
1174 	 * a chance to complete the packet before we drop the packet.
1175 	 * As a result, n+1 frags are actually allowed per packet, but
1176 	 * only n will ever be stored. (n = maxfragsperpacket.)
1177 	 *
1178 	 */
1179 	next = 0;
1180 	for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
1181 		if (GETIP(q)->ip_off != next) {
1182 			if (fp->ipq_nfrags > maxfragsperpacket) {
1183 				ipstat.ips_fragdropped += fp->ipq_nfrags;
1184 				ip_freef(head, fp);
1185 			}
1186 			goto done;
1187 		}
1188 		next += GETIP(q)->ip_len;
1189 	}
1190 	/* Make sure the last packet didn't have the IP_MF flag */
1191 	if (p->m_flags & M_FRAG) {
1192 		if (fp->ipq_nfrags > maxfragsperpacket) {
1193 			ipstat.ips_fragdropped += fp->ipq_nfrags;
1194 			ip_freef(head, fp);
1195 		}
1196 		goto done;
1197 	}
1198 
1199 	/*
1200 	 * Reassembly is complete.  Make sure the packet is a sane size.
1201 	 */
1202 	q = fp->ipq_frags;
1203 	ip = GETIP(q);
1204 	if (next + (IP_VHL_HL(ip->ip_vhl) << 2) > IP_MAXPACKET) {
1205 		ipstat.ips_toolong++;
1206 		ipstat.ips_fragdropped += fp->ipq_nfrags;
1207 		ip_freef(head, fp);
1208 		goto done;
1209 	}
1210 
1211 	/*
1212 	 * Concatenate fragments.
1213 	 */
1214 	m = q;
1215 	n = m->m_next;
1216 	m->m_next = NULL;
1217 	m_cat(m, n);
1218 	nq = q->m_nextpkt;
1219 	q->m_nextpkt = NULL;
1220 	for (q = nq; q != NULL; q = nq) {
1221 		nq = q->m_nextpkt;
1222 		q->m_nextpkt = NULL;
1223 		m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
1224 		m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
1225 		m_cat(m, q);
1226 	}
1227 
1228 	/*
1229 	 * Clean up the 1's complement checksum.  Carry over 16 bits must
1230 	 * be added back.  This assumes no more then 65535 packet fragments
1231 	 * were reassembled.  A second carry can also occur (but not a third).
1232 	 */
1233 	m->m_pkthdr.csum_data = (m->m_pkthdr.csum_data & 0xffff) +
1234 				(m->m_pkthdr.csum_data >> 16);
1235 	if (m->m_pkthdr.csum_data > 0xFFFF)
1236 		m->m_pkthdr.csum_data -= 0xFFFF;
1237 
1238 	/*
1239 	 * Create header for new ip packet by
1240 	 * modifying header of first packet;
1241 	 * dequeue and discard fragment reassembly header.
1242 	 * Make header visible.
1243 	 */
1244 	ip->ip_len = next;
1245 	ip->ip_src = fp->ipq_src;
1246 	ip->ip_dst = fp->ipq_dst;
1247 	TAILQ_REMOVE(head, fp, ipq_list);
1248 	nipq--;
1249 	mpipe_free(&ipq_mpipe, fp);
1250 	m->m_len += (IP_VHL_HL(ip->ip_vhl) << 2);
1251 	m->m_data -= (IP_VHL_HL(ip->ip_vhl) << 2);
1252 	/* some debugging cruft by sklower, below, will go away soon */
1253 	if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */
1254 		int plen = 0;
1255 
1256 		for (n = m; n; n = n->m_next)
1257 			plen += n->m_len;
1258 		m->m_pkthdr.len = plen;
1259 	}
1260 
1261 	/*
1262 	 * Reassembly complete, return the next protocol.
1263 	 *
1264 	 * Be sure to clear M_HASH to force the packet
1265 	 * to be re-characterized.
1266 	 *
1267 	 * Clear M_FRAG, we are no longer a fragment.
1268 	 */
1269 	m->m_flags &= ~(M_HASH | M_FRAG);
1270 
1271 	ipstat.ips_reassembled++;
1272 	lwkt_reltoken(&ipq_token);
1273 	return (m);
1274 
1275 dropfrag:
1276 	ipstat.ips_fragdropped++;
1277 	if (fp != NULL)
1278 		fp->ipq_nfrags--;
1279 	m_freem(m);
1280 done:
1281 	lwkt_reltoken(&ipq_token);
1282 	return (NULL);
1283 
1284 #undef GETIP
1285 }
1286 
1287 /*
1288  * Free a fragment reassembly header and all
1289  * associated datagrams.
1290  *
1291  * Called with ipq_token held.
1292  */
1293 static void
1294 ip_freef(struct ipqhead *fhp, struct ipq *fp)
1295 {
1296 	struct mbuf *q;
1297 
1298 	/*
1299 	 * Remove first to protect against blocking
1300 	 */
1301 	TAILQ_REMOVE(fhp, fp, ipq_list);
1302 
1303 	/*
1304 	 * Clean out at our leisure
1305 	 */
1306 	while (fp->ipq_frags) {
1307 		q = fp->ipq_frags;
1308 		fp->ipq_frags = q->m_nextpkt;
1309 		q->m_nextpkt = NULL;
1310 		m_freem(q);
1311 	}
1312 	mpipe_free(&ipq_mpipe, fp);
1313 	nipq--;
1314 }
1315 
1316 /*
1317  * IP timer processing;
1318  * if a timer expires on a reassembly
1319  * queue, discard it.
1320  */
1321 void
1322 ip_slowtimo(void)
1323 {
1324 	struct ipq *fp, *fp_temp;
1325 	struct ipqhead *head;
1326 	int i;
1327 
1328 	lwkt_gettoken(&ipq_token);
1329 	for (i = 0; i < IPREASS_NHASH; i++) {
1330 		head = &ipq[i];
1331 		TAILQ_FOREACH_MUTABLE(fp, head, ipq_list, fp_temp) {
1332 			if (--fp->ipq_ttl == 0) {
1333 				ipstat.ips_fragtimeout += fp->ipq_nfrags;
1334 				ip_freef(head, fp);
1335 			}
1336 		}
1337 	}
1338 	/*
1339 	 * If we are over the maximum number of fragments
1340 	 * (due to the limit being lowered), drain off
1341 	 * enough to get down to the new limit.
1342 	 */
1343 	if (maxnipq >= 0 && nipq > maxnipq) {
1344 		for (i = 0; i < IPREASS_NHASH; i++) {
1345 			head = &ipq[i];
1346 			while (nipq > maxnipq && !TAILQ_EMPTY(head)) {
1347 				ipstat.ips_fragdropped +=
1348 				    TAILQ_FIRST(head)->ipq_nfrags;
1349 				ip_freef(head, TAILQ_FIRST(head));
1350 			}
1351 		}
1352 	}
1353 	lwkt_reltoken(&ipq_token);
1354 	ipflow_slowtimo();
1355 }
1356 
1357 /*
1358  * Drain off all datagram fragments.
1359  */
1360 void
1361 ip_drain(void)
1362 {
1363 	struct ipqhead *head;
1364 	int i;
1365 
1366 	lwkt_gettoken(&ipq_token);
1367 	for (i = 0; i < IPREASS_NHASH; i++) {
1368 		head = &ipq[i];
1369 		while (!TAILQ_EMPTY(head)) {
1370 			ipstat.ips_fragdropped += TAILQ_FIRST(head)->ipq_nfrags;
1371 			ip_freef(head, TAILQ_FIRST(head));
1372 		}
1373 	}
1374 	lwkt_reltoken(&ipq_token);
1375 	in_rtqdrain();
1376 }
1377 
1378 /*
1379  * Do option processing on a datagram,
1380  * possibly discarding it if bad options are encountered,
1381  * or forwarding it if source-routed.
1382  * The pass argument is used when operating in the IPSTEALTH
1383  * mode to tell what options to process:
1384  * [LS]SRR (pass 0) or the others (pass 1).
1385  * The reason for as many as two passes is that when doing IPSTEALTH,
1386  * non-routing options should be processed only if the packet is for us.
1387  * Returns 1 if packet has been forwarded/freed,
1388  * 0 if the packet should be processed further.
1389  */
1390 static int
1391 ip_dooptions(struct mbuf *m, int pass, struct sockaddr_in *next_hop)
1392 {
1393 	struct sockaddr_in ipaddr = { sizeof ipaddr, AF_INET };
1394 	struct ip *ip = mtod(m, struct ip *);
1395 	u_char *cp;
1396 	struct in_ifaddr *ia;
1397 	int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB;
1398 	boolean_t forward = FALSE;
1399 	struct in_addr *sin, dst;
1400 	n_time ntime;
1401 
1402 	dst = ip->ip_dst;
1403 	cp = (u_char *)(ip + 1);
1404 	cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
1405 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1406 		opt = cp[IPOPT_OPTVAL];
1407 		if (opt == IPOPT_EOL)
1408 			break;
1409 		if (opt == IPOPT_NOP)
1410 			optlen = 1;
1411 		else {
1412 			if (cnt < IPOPT_OLEN + sizeof(*cp)) {
1413 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1414 				goto bad;
1415 			}
1416 			optlen = cp[IPOPT_OLEN];
1417 			if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
1418 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1419 				goto bad;
1420 			}
1421 		}
1422 		switch (opt) {
1423 
1424 		default:
1425 			break;
1426 
1427 		/*
1428 		 * Source routing with record.
1429 		 * Find interface with current destination address.
1430 		 * If none on this machine then drop if strictly routed,
1431 		 * or do nothing if loosely routed.
1432 		 * Record interface address and bring up next address
1433 		 * component.  If strictly routed make sure next
1434 		 * address is on directly accessible net.
1435 		 */
1436 		case IPOPT_LSRR:
1437 		case IPOPT_SSRR:
1438 			if (ipstealth && pass > 0)
1439 				break;
1440 			if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1441 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1442 				goto bad;
1443 			}
1444 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1445 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1446 				goto bad;
1447 			}
1448 			ipaddr.sin_addr = ip->ip_dst;
1449 			ia = (struct in_ifaddr *)
1450 				ifa_ifwithaddr((struct sockaddr *)&ipaddr);
1451 			if (ia == NULL) {
1452 				if (opt == IPOPT_SSRR) {
1453 					type = ICMP_UNREACH;
1454 					code = ICMP_UNREACH_SRCFAIL;
1455 					goto bad;
1456 				}
1457 				if (!ip_dosourceroute)
1458 					goto nosourcerouting;
1459 				/*
1460 				 * Loose routing, and not at next destination
1461 				 * yet; nothing to do except forward.
1462 				 */
1463 				break;
1464 			}
1465 			off--;			/* 0 origin */
1466 			if (off > optlen - (int)sizeof(struct in_addr)) {
1467 				/*
1468 				 * End of source route.  Should be for us.
1469 				 */
1470 				if (!ip_acceptsourceroute)
1471 					goto nosourcerouting;
1472 				save_rte(m, cp, ip->ip_src);
1473 				break;
1474 			}
1475 			if (ipstealth)
1476 				goto dropit;
1477 			if (!ip_dosourceroute) {
1478 				if (ipforwarding) {
1479 					char buf[sizeof "aaa.bbb.ccc.ddd"];
1480 
1481 					/*
1482 					 * Acting as a router, so generate ICMP
1483 					 */
1484 nosourcerouting:
1485 					strcpy(buf, inet_ntoa(ip->ip_dst));
1486 					log(LOG_WARNING,
1487 					    "attempted source route from %s to %s\n",
1488 					    inet_ntoa(ip->ip_src), buf);
1489 					type = ICMP_UNREACH;
1490 					code = ICMP_UNREACH_SRCFAIL;
1491 					goto bad;
1492 				} else {
1493 					/*
1494 					 * Not acting as a router,
1495 					 * so silently drop.
1496 					 */
1497 dropit:
1498 					ipstat.ips_cantforward++;
1499 					m_freem(m);
1500 					return (1);
1501 				}
1502 			}
1503 
1504 			/*
1505 			 * locate outgoing interface
1506 			 */
1507 			memcpy(&ipaddr.sin_addr, cp + off,
1508 			    sizeof ipaddr.sin_addr);
1509 
1510 			if (opt == IPOPT_SSRR) {
1511 #define	INA	struct in_ifaddr *
1512 #define	SA	struct sockaddr *
1513 				if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr))
1514 									== NULL)
1515 					ia = (INA)ifa_ifwithnet((SA)&ipaddr);
1516 			} else {
1517 				ia = ip_rtaddr(ipaddr.sin_addr, NULL);
1518 			}
1519 			if (ia == NULL) {
1520 				type = ICMP_UNREACH;
1521 				code = ICMP_UNREACH_SRCFAIL;
1522 				goto bad;
1523 			}
1524 			ip->ip_dst = ipaddr.sin_addr;
1525 			memcpy(cp + off, &IA_SIN(ia)->sin_addr,
1526 			    sizeof(struct in_addr));
1527 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1528 			/*
1529 			 * Let ip_intr's mcast routing check handle mcast pkts
1530 			 */
1531 			forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr));
1532 			break;
1533 
1534 		case IPOPT_RR:
1535 			if (ipstealth && pass == 0)
1536 				break;
1537 			if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1538 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1539 				goto bad;
1540 			}
1541 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1542 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1543 				goto bad;
1544 			}
1545 			/*
1546 			 * If no space remains, ignore.
1547 			 */
1548 			off--;			/* 0 origin */
1549 			if (off > optlen - (int)sizeof(struct in_addr))
1550 				break;
1551 			memcpy(&ipaddr.sin_addr, &ip->ip_dst,
1552 			    sizeof ipaddr.sin_addr);
1553 			/*
1554 			 * locate outgoing interface; if we're the destination,
1555 			 * use the incoming interface (should be same).
1556 			 */
1557 			if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == NULL &&
1558 			    (ia = ip_rtaddr(ipaddr.sin_addr, NULL)) == NULL) {
1559 				type = ICMP_UNREACH;
1560 				code = ICMP_UNREACH_HOST;
1561 				goto bad;
1562 			}
1563 			memcpy(cp + off, &IA_SIN(ia)->sin_addr,
1564 			    sizeof(struct in_addr));
1565 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1566 			break;
1567 
1568 		case IPOPT_TS:
1569 			if (ipstealth && pass == 0)
1570 				break;
1571 			code = cp - (u_char *)ip;
1572 			if (optlen < 4 || optlen > 40) {
1573 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1574 				goto bad;
1575 			}
1576 			if ((off = cp[IPOPT_OFFSET]) < 5) {
1577 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1578 				goto bad;
1579 			}
1580 			if (off > optlen - (int)sizeof(int32_t)) {
1581 				cp[IPOPT_OFFSET + 1] += (1 << 4);
1582 				if ((cp[IPOPT_OFFSET + 1] & 0xf0) == 0) {
1583 					code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1584 					goto bad;
1585 				}
1586 				break;
1587 			}
1588 			off--;				/* 0 origin */
1589 			sin = (struct in_addr *)(cp + off);
1590 			switch (cp[IPOPT_OFFSET + 1] & 0x0f) {
1591 
1592 			case IPOPT_TS_TSONLY:
1593 				break;
1594 
1595 			case IPOPT_TS_TSANDADDR:
1596 				if (off + sizeof(n_time) +
1597 				    sizeof(struct in_addr) > optlen) {
1598 					code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1599 					goto bad;
1600 				}
1601 				ipaddr.sin_addr = dst;
1602 				ia = (INA)ifaof_ifpforaddr((SA)&ipaddr,
1603 							    m->m_pkthdr.rcvif);
1604 				if (ia == NULL)
1605 					continue;
1606 				memcpy(sin, &IA_SIN(ia)->sin_addr,
1607 				    sizeof(struct in_addr));
1608 				cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1609 				off += sizeof(struct in_addr);
1610 				break;
1611 
1612 			case IPOPT_TS_PRESPEC:
1613 				if (off + sizeof(n_time) +
1614 				    sizeof(struct in_addr) > optlen) {
1615 					code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1616 					goto bad;
1617 				}
1618 				memcpy(&ipaddr.sin_addr, sin,
1619 				    sizeof(struct in_addr));
1620 				if (ifa_ifwithaddr((SA)&ipaddr) == NULL)
1621 					continue;
1622 				cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1623 				off += sizeof(struct in_addr);
1624 				break;
1625 
1626 			default:
1627 				code = &cp[IPOPT_OFFSET + 1] - (u_char *)ip;
1628 				goto bad;
1629 			}
1630 			ntime = iptime();
1631 			memcpy(cp + off, &ntime, sizeof(n_time));
1632 			cp[IPOPT_OFFSET] += sizeof(n_time);
1633 		}
1634 	}
1635 	if (forward && ipforwarding) {
1636 		ip_forward(m, TRUE, next_hop);
1637 		return (1);
1638 	}
1639 	return (0);
1640 bad:
1641 	icmp_error(m, type, code, 0, 0);
1642 	ipstat.ips_badoptions++;
1643 	return (1);
1644 }
1645 
1646 /*
1647  * Given address of next destination (final or next hop),
1648  * return internet address info of interface to be used to get there.
1649  */
1650 struct in_ifaddr *
1651 ip_rtaddr(struct in_addr dst, struct route *ro0)
1652 {
1653 	struct route sro, *ro;
1654 	struct sockaddr_in *sin;
1655 	struct in_ifaddr *ia;
1656 
1657 	if (ro0 != NULL) {
1658 		ro = ro0;
1659 	} else {
1660 		bzero(&sro, sizeof(sro));
1661 		ro = &sro;
1662 	}
1663 
1664 	sin = (struct sockaddr_in *)&ro->ro_dst;
1665 
1666 	if (ro->ro_rt == NULL || dst.s_addr != sin->sin_addr.s_addr) {
1667 		if (ro->ro_rt != NULL) {
1668 			RTFREE(ro->ro_rt);
1669 			ro->ro_rt = NULL;
1670 		}
1671 		sin->sin_family = AF_INET;
1672 		sin->sin_len = sizeof *sin;
1673 		sin->sin_addr = dst;
1674 		rtalloc_ign(ro, RTF_PRCLONING);
1675 	}
1676 
1677 	if (ro->ro_rt == NULL)
1678 		return (NULL);
1679 
1680 	ia = ifatoia(ro->ro_rt->rt_ifa);
1681 
1682 	if (ro == &sro)
1683 		RTFREE(ro->ro_rt);
1684 	return ia;
1685 }
1686 
1687 /*
1688  * Save incoming source route for use in replies,
1689  * to be picked up later by ip_srcroute if the receiver is interested.
1690  */
1691 static void
1692 save_rte(struct mbuf *m, u_char *option, struct in_addr dst)
1693 {
1694 	struct m_tag *mtag;
1695 	struct ip_srcrt_opt *opt;
1696 	unsigned olen;
1697 
1698 	mtag = m_tag_get(PACKET_TAG_IPSRCRT, sizeof(*opt), MB_DONTWAIT);
1699 	if (mtag == NULL)
1700 		return;
1701 	opt = m_tag_data(mtag);
1702 
1703 	olen = option[IPOPT_OLEN];
1704 #ifdef DIAGNOSTIC
1705 	if (ipprintfs)
1706 		kprintf("save_rte: olen %d\n", olen);
1707 #endif
1708 	if (olen > sizeof(opt->ip_srcrt) - (1 + sizeof(dst))) {
1709 		m_tag_free(mtag);
1710 		return;
1711 	}
1712 	bcopy(option, opt->ip_srcrt.srcopt, olen);
1713 	opt->ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1714 	opt->ip_srcrt.dst = dst;
1715 	m_tag_prepend(m, mtag);
1716 }
1717 
1718 /*
1719  * Retrieve incoming source route for use in replies,
1720  * in the same form used by setsockopt.
1721  * The first hop is placed before the options, will be removed later.
1722  */
1723 struct mbuf *
1724 ip_srcroute(struct mbuf *m0)
1725 {
1726 	struct in_addr *p, *q;
1727 	struct mbuf *m;
1728 	struct m_tag *mtag;
1729 	struct ip_srcrt_opt *opt;
1730 
1731 	if (m0 == NULL)
1732 		return NULL;
1733 
1734 	mtag = m_tag_find(m0, PACKET_TAG_IPSRCRT, NULL);
1735 	if (mtag == NULL)
1736 		return NULL;
1737 	opt = m_tag_data(mtag);
1738 
1739 	if (opt->ip_nhops == 0)
1740 		return (NULL);
1741 	m = m_get(MB_DONTWAIT, MT_HEADER);
1742 	if (m == NULL)
1743 		return (NULL);
1744 
1745 #define	OPTSIZ	(sizeof(opt->ip_srcrt.nop) + sizeof(opt->ip_srcrt.srcopt))
1746 
1747 	/* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
1748 	m->m_len = opt->ip_nhops * sizeof(struct in_addr) +
1749 		   sizeof(struct in_addr) + OPTSIZ;
1750 #ifdef DIAGNOSTIC
1751 	if (ipprintfs) {
1752 		kprintf("ip_srcroute: nhops %d mlen %d",
1753 			opt->ip_nhops, m->m_len);
1754 	}
1755 #endif
1756 
1757 	/*
1758 	 * First save first hop for return route
1759 	 */
1760 	p = &opt->ip_srcrt.route[opt->ip_nhops - 1];
1761 	*(mtod(m, struct in_addr *)) = *p--;
1762 #ifdef DIAGNOSTIC
1763 	if (ipprintfs)
1764 		kprintf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr));
1765 #endif
1766 
1767 	/*
1768 	 * Copy option fields and padding (nop) to mbuf.
1769 	 */
1770 	opt->ip_srcrt.nop = IPOPT_NOP;
1771 	opt->ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
1772 	memcpy(mtod(m, caddr_t) + sizeof(struct in_addr), &opt->ip_srcrt.nop,
1773 	    OPTSIZ);
1774 	q = (struct in_addr *)(mtod(m, caddr_t) +
1775 	    sizeof(struct in_addr) + OPTSIZ);
1776 #undef OPTSIZ
1777 	/*
1778 	 * Record return path as an IP source route,
1779 	 * reversing the path (pointers are now aligned).
1780 	 */
1781 	while (p >= opt->ip_srcrt.route) {
1782 #ifdef DIAGNOSTIC
1783 		if (ipprintfs)
1784 			kprintf(" %x", ntohl(q->s_addr));
1785 #endif
1786 		*q++ = *p--;
1787 	}
1788 	/*
1789 	 * Last hop goes to final destination.
1790 	 */
1791 	*q = opt->ip_srcrt.dst;
1792 	m_tag_delete(m0, mtag);
1793 #ifdef DIAGNOSTIC
1794 	if (ipprintfs)
1795 		kprintf(" %x\n", ntohl(q->s_addr));
1796 #endif
1797 	return (m);
1798 }
1799 
1800 /*
1801  * Strip out IP options.
1802  */
1803 void
1804 ip_stripoptions(struct mbuf *m)
1805 {
1806 	int datalen;
1807 	struct ip *ip = mtod(m, struct ip *);
1808 	caddr_t opts;
1809 	int optlen;
1810 
1811 	optlen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
1812 	opts = (caddr_t)(ip + 1);
1813 	datalen = m->m_len - (sizeof(struct ip) + optlen);
1814 	bcopy(opts + optlen, opts, datalen);
1815 	m->m_len -= optlen;
1816 	if (m->m_flags & M_PKTHDR)
1817 		m->m_pkthdr.len -= optlen;
1818 	ip->ip_vhl = IP_MAKE_VHL(IPVERSION, sizeof(struct ip) >> 2);
1819 }
1820 
1821 u_char inetctlerrmap[PRC_NCMDS] = {
1822 	0,		0,		0,		0,
1823 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1824 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1825 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1826 	0,		0,		0,		0,
1827 	ENOPROTOOPT,	ECONNREFUSED
1828 };
1829 
1830 /*
1831  * Forward a packet.  If some error occurs return the sender
1832  * an icmp packet.  Note we can't always generate a meaningful
1833  * icmp message because icmp doesn't have a large enough repertoire
1834  * of codes and types.
1835  *
1836  * If not forwarding, just drop the packet.  This could be confusing
1837  * if ipforwarding was zero but some routing protocol was advancing
1838  * us as a gateway to somewhere.  However, we must let the routing
1839  * protocol deal with that.
1840  *
1841  * The using_srcrt parameter indicates whether the packet is being forwarded
1842  * via a source route.
1843  */
1844 void
1845 ip_forward(struct mbuf *m, boolean_t using_srcrt, struct sockaddr_in *next_hop)
1846 {
1847 	struct ip *ip = mtod(m, struct ip *);
1848 	struct rtentry *rt;
1849 	struct route fwd_ro;
1850 	int error, type = 0, code = 0, destmtu = 0;
1851 	struct mbuf *mcopy;
1852 	n_long dest;
1853 	struct in_addr pkt_dst;
1854 
1855 	dest = INADDR_ANY;
1856 	/*
1857 	 * Cache the destination address of the packet; this may be
1858 	 * changed by use of 'ipfw fwd'.
1859 	 */
1860 	pkt_dst = (next_hop != NULL) ? next_hop->sin_addr : ip->ip_dst;
1861 
1862 #ifdef DIAGNOSTIC
1863 	if (ipprintfs)
1864 		kprintf("forward: src %x dst %x ttl %x\n",
1865 		       ip->ip_src.s_addr, pkt_dst.s_addr, ip->ip_ttl);
1866 #endif
1867 
1868 	if (m->m_flags & (M_BCAST | M_MCAST) || !in_canforward(pkt_dst)) {
1869 		ipstat.ips_cantforward++;
1870 		m_freem(m);
1871 		return;
1872 	}
1873 	if (!ipstealth && ip->ip_ttl <= IPTTLDEC) {
1874 		icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0);
1875 		return;
1876 	}
1877 
1878 	bzero(&fwd_ro, sizeof(fwd_ro));
1879 	ip_rtaddr(pkt_dst, &fwd_ro);
1880 	if (fwd_ro.ro_rt == NULL) {
1881 		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0);
1882 		return;
1883 	}
1884 	rt = fwd_ro.ro_rt;
1885 
1886 	/*
1887 	 * Save the IP header and at most 8 bytes of the payload,
1888 	 * in case we need to generate an ICMP message to the src.
1889 	 *
1890 	 * XXX this can be optimized a lot by saving the data in a local
1891 	 * buffer on the stack (72 bytes at most), and only allocating the
1892 	 * mbuf if really necessary. The vast majority of the packets
1893 	 * are forwarded without having to send an ICMP back (either
1894 	 * because unnecessary, or because rate limited), so we are
1895 	 * really we are wasting a lot of work here.
1896 	 *
1897 	 * We don't use m_copy() because it might return a reference
1898 	 * to a shared cluster. Both this function and ip_output()
1899 	 * assume exclusive access to the IP header in `m', so any
1900 	 * data in a cluster may change before we reach icmp_error().
1901 	 */
1902 	MGETHDR(mcopy, MB_DONTWAIT, m->m_type);
1903 	if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, MB_DONTWAIT)) {
1904 		/*
1905 		 * It's probably ok if the pkthdr dup fails (because
1906 		 * the deep copy of the tag chain failed), but for now
1907 		 * be conservative and just discard the copy since
1908 		 * code below may some day want the tags.
1909 		 */
1910 		m_free(mcopy);
1911 		mcopy = NULL;
1912 	}
1913 	if (mcopy != NULL) {
1914 		mcopy->m_len = imin((IP_VHL_HL(ip->ip_vhl) << 2) + 8,
1915 		    (int)ip->ip_len);
1916 		mcopy->m_pkthdr.len = mcopy->m_len;
1917 		m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1918 	}
1919 
1920 	if (!ipstealth)
1921 		ip->ip_ttl -= IPTTLDEC;
1922 
1923 	/*
1924 	 * If forwarding packet using same interface that it came in on,
1925 	 * perhaps should send a redirect to sender to shortcut a hop.
1926 	 * Only send redirect if source is sending directly to us,
1927 	 * and if packet was not source routed (or has any options).
1928 	 * Also, don't send redirect if forwarding using a default route
1929 	 * or a route modified by a redirect.
1930 	 */
1931 	if (rt->rt_ifp == m->m_pkthdr.rcvif &&
1932 	    !(rt->rt_flags & (RTF_DYNAMIC | RTF_MODIFIED)) &&
1933 	    satosin(rt_key(rt))->sin_addr.s_addr != INADDR_ANY &&
1934 	    ipsendredirects && !using_srcrt && next_hop == NULL) {
1935 		u_long src = ntohl(ip->ip_src.s_addr);
1936 		struct in_ifaddr *rt_ifa = (struct in_ifaddr *)rt->rt_ifa;
1937 
1938 		if (rt_ifa != NULL &&
1939 		    (src & rt_ifa->ia_subnetmask) == rt_ifa->ia_subnet) {
1940 			if (rt->rt_flags & RTF_GATEWAY)
1941 				dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
1942 			else
1943 				dest = pkt_dst.s_addr;
1944 			/*
1945 			 * Router requirements says to only send
1946 			 * host redirects.
1947 			 */
1948 			type = ICMP_REDIRECT;
1949 			code = ICMP_REDIRECT_HOST;
1950 #ifdef DIAGNOSTIC
1951 			if (ipprintfs)
1952 				kprintf("redirect (%d) to %x\n", code, dest);
1953 #endif
1954 		}
1955 	}
1956 
1957 	error = ip_output(m, NULL, &fwd_ro, IP_FORWARDING, NULL, NULL);
1958 	if (error == 0) {
1959 		ipstat.ips_forward++;
1960 		if (type == 0) {
1961 			if (mcopy) {
1962 				ipflow_create(&fwd_ro, mcopy);
1963 				m_freem(mcopy);
1964 			}
1965 			goto done;
1966 		} else {
1967 			ipstat.ips_redirectsent++;
1968 		}
1969 	} else {
1970 		ipstat.ips_cantforward++;
1971 	}
1972 
1973 	if (mcopy == NULL)
1974 		goto done;
1975 
1976 	/*
1977 	 * Send ICMP message.
1978 	 */
1979 
1980 	switch (error) {
1981 
1982 	case 0:				/* forwarded, but need redirect */
1983 		/* type, code set above */
1984 		break;
1985 
1986 	case ENETUNREACH:		/* shouldn't happen, checked above */
1987 	case EHOSTUNREACH:
1988 	case ENETDOWN:
1989 	case EHOSTDOWN:
1990 	default:
1991 		type = ICMP_UNREACH;
1992 		code = ICMP_UNREACH_HOST;
1993 		break;
1994 
1995 	case EMSGSIZE:
1996 		type = ICMP_UNREACH;
1997 		code = ICMP_UNREACH_NEEDFRAG;
1998 #ifdef IPSEC
1999 		/*
2000 		 * If the packet is routed over IPsec tunnel, tell the
2001 		 * originator the tunnel MTU.
2002 		 *	tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
2003 		 * XXX quickhack!!!
2004 		 */
2005 		if (fwd_ro.ro_rt != NULL) {
2006 			struct secpolicy *sp = NULL;
2007 			int ipsecerror;
2008 			int ipsechdr;
2009 			struct route *ro;
2010 
2011 			sp = ipsec4_getpolicybyaddr(mcopy,
2012 						    IPSEC_DIR_OUTBOUND,
2013 						    IP_FORWARDING,
2014 						    &ipsecerror);
2015 
2016 			if (sp == NULL)
2017 				destmtu = fwd_ro.ro_rt->rt_ifp->if_mtu;
2018 			else {
2019 				/* count IPsec header size */
2020 				ipsechdr = ipsec4_hdrsiz(mcopy,
2021 							 IPSEC_DIR_OUTBOUND,
2022 							 NULL);
2023 
2024 				/*
2025 				 * find the correct route for outer IPv4
2026 				 * header, compute tunnel MTU.
2027 				 *
2028 				 */
2029 				if (sp->req != NULL && sp->req->sav != NULL &&
2030 				    sp->req->sav->sah != NULL) {
2031 					ro = &sp->req->sav->sah->sa_route;
2032 					if (ro->ro_rt != NULL &&
2033 					    ro->ro_rt->rt_ifp != NULL) {
2034 						destmtu =
2035 						    ro->ro_rt->rt_ifp->if_mtu;
2036 						destmtu -= ipsechdr;
2037 					}
2038 				}
2039 
2040 				key_freesp(sp);
2041 			}
2042 		}
2043 #elif FAST_IPSEC
2044 		/*
2045 		 * If the packet is routed over IPsec tunnel, tell the
2046 		 * originator the tunnel MTU.
2047 		 *	tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
2048 		 * XXX quickhack!!!
2049 		 */
2050 		if (fwd_ro.ro_rt != NULL) {
2051 			struct secpolicy *sp = NULL;
2052 			int ipsecerror;
2053 			int ipsechdr;
2054 			struct route *ro;
2055 
2056 			sp = ipsec_getpolicybyaddr(mcopy,
2057 						   IPSEC_DIR_OUTBOUND,
2058 						   IP_FORWARDING,
2059 						   &ipsecerror);
2060 
2061 			if (sp == NULL)
2062 				destmtu = fwd_ro.ro_rt->rt_ifp->if_mtu;
2063 			else {
2064 				/* count IPsec header size */
2065 				ipsechdr = ipsec4_hdrsiz(mcopy,
2066 							 IPSEC_DIR_OUTBOUND,
2067 							 NULL);
2068 
2069 				/*
2070 				 * find the correct route for outer IPv4
2071 				 * header, compute tunnel MTU.
2072 				 */
2073 
2074 				if (sp->req != NULL &&
2075 				    sp->req->sav != NULL &&
2076 				    sp->req->sav->sah != NULL) {
2077 					ro = &sp->req->sav->sah->sa_route;
2078 					if (ro->ro_rt != NULL &&
2079 					    ro->ro_rt->rt_ifp != NULL) {
2080 						destmtu =
2081 						    ro->ro_rt->rt_ifp->if_mtu;
2082 						destmtu -= ipsechdr;
2083 					}
2084 				}
2085 
2086 				KEY_FREESP(&sp);
2087 			}
2088 		}
2089 #else /* !IPSEC && !FAST_IPSEC */
2090 		if (fwd_ro.ro_rt != NULL)
2091 			destmtu = fwd_ro.ro_rt->rt_ifp->if_mtu;
2092 #endif /*IPSEC*/
2093 		ipstat.ips_cantfrag++;
2094 		break;
2095 
2096 	case ENOBUFS:
2097 		/*
2098 		 * A router should not generate ICMP_SOURCEQUENCH as
2099 		 * required in RFC1812 Requirements for IP Version 4 Routers.
2100 		 * Source quench could be a big problem under DoS attacks,
2101 		 * or if the underlying interface is rate-limited.
2102 		 * Those who need source quench packets may re-enable them
2103 		 * via the net.inet.ip.sendsourcequench sysctl.
2104 		 */
2105 		if (!ip_sendsourcequench) {
2106 			m_freem(mcopy);
2107 			goto done;
2108 		} else {
2109 			type = ICMP_SOURCEQUENCH;
2110 			code = 0;
2111 		}
2112 		break;
2113 
2114 	case EACCES:			/* ipfw denied packet */
2115 		m_freem(mcopy);
2116 		goto done;
2117 	}
2118 	icmp_error(mcopy, type, code, dest, destmtu);
2119 done:
2120 	if (fwd_ro.ro_rt != NULL)
2121 		RTFREE(fwd_ro.ro_rt);
2122 }
2123 
2124 void
2125 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
2126 	       struct mbuf *m)
2127 {
2128 	if (inp->inp_socket->so_options & SO_TIMESTAMP) {
2129 		struct timeval tv;
2130 
2131 		microtime(&tv);
2132 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
2133 		    SCM_TIMESTAMP, SOL_SOCKET);
2134 		if (*mp)
2135 			mp = &(*mp)->m_next;
2136 	}
2137 	if (inp->inp_flags & INP_RECVDSTADDR) {
2138 		*mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
2139 		    sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
2140 		if (*mp)
2141 			mp = &(*mp)->m_next;
2142 	}
2143 	if (inp->inp_flags & INP_RECVTTL) {
2144 		*mp = sbcreatecontrol((caddr_t) &ip->ip_ttl,
2145 		    sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
2146 		if (*mp)
2147 			mp = &(*mp)->m_next;
2148 	}
2149 #ifdef notyet
2150 	/* XXX
2151 	 * Moving these out of udp_input() made them even more broken
2152 	 * than they already were.
2153 	 */
2154 	/* options were tossed already */
2155 	if (inp->inp_flags & INP_RECVOPTS) {
2156 		*mp = sbcreatecontrol((caddr_t) opts_deleted_above,
2157 		    sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
2158 		if (*mp)
2159 			mp = &(*mp)->m_next;
2160 	}
2161 	/* ip_srcroute doesn't do what we want here, need to fix */
2162 	if (inp->inp_flags & INP_RECVRETOPTS) {
2163 		*mp = sbcreatecontrol((caddr_t) ip_srcroute(m),
2164 		    sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
2165 		if (*mp)
2166 			mp = &(*mp)->m_next;
2167 	}
2168 #endif
2169 	if (inp->inp_flags & INP_RECVIF) {
2170 		struct ifnet *ifp;
2171 		struct sdlbuf {
2172 			struct sockaddr_dl sdl;
2173 			u_char	pad[32];
2174 		} sdlbuf;
2175 		struct sockaddr_dl *sdp;
2176 		struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
2177 
2178 		if (((ifp = m->m_pkthdr.rcvif)) &&
2179 		    ((ifp->if_index != 0) && (ifp->if_index <= if_index))) {
2180 			sdp = IF_LLSOCKADDR(ifp);
2181 			/*
2182 			 * Change our mind and don't try copy.
2183 			 */
2184 			if ((sdp->sdl_family != AF_LINK) ||
2185 			    (sdp->sdl_len > sizeof(sdlbuf))) {
2186 				goto makedummy;
2187 			}
2188 			bcopy(sdp, sdl2, sdp->sdl_len);
2189 		} else {
2190 makedummy:
2191 			sdl2->sdl_len =
2192 			    offsetof(struct sockaddr_dl, sdl_data[0]);
2193 			sdl2->sdl_family = AF_LINK;
2194 			sdl2->sdl_index = 0;
2195 			sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
2196 		}
2197 		*mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len,
2198 			IP_RECVIF, IPPROTO_IP);
2199 		if (*mp)
2200 			mp = &(*mp)->m_next;
2201 	}
2202 }
2203 
2204 /*
2205  * XXX these routines are called from the upper part of the kernel.
2206  *
2207  * They could also be moved to ip_mroute.c, since all the RSVP
2208  *  handling is done there already.
2209  */
2210 int
2211 ip_rsvp_init(struct socket *so)
2212 {
2213 	if (so->so_type != SOCK_RAW ||
2214 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
2215 		return EOPNOTSUPP;
2216 
2217 	if (ip_rsvpd != NULL)
2218 		return EADDRINUSE;
2219 
2220 	ip_rsvpd = so;
2221 	/*
2222 	 * This may seem silly, but we need to be sure we don't over-increment
2223 	 * the RSVP counter, in case something slips up.
2224 	 */
2225 	if (!ip_rsvp_on) {
2226 		ip_rsvp_on = 1;
2227 		rsvp_on++;
2228 	}
2229 
2230 	return 0;
2231 }
2232 
2233 int
2234 ip_rsvp_done(void)
2235 {
2236 	ip_rsvpd = NULL;
2237 	/*
2238 	 * This may seem silly, but we need to be sure we don't over-decrement
2239 	 * the RSVP counter, in case something slips up.
2240 	 */
2241 	if (ip_rsvp_on) {
2242 		ip_rsvp_on = 0;
2243 		rsvp_on--;
2244 	}
2245 	return 0;
2246 }
2247 
2248 int
2249 rsvp_input(struct mbuf **mp, int *offp, int proto)
2250 {
2251 	struct mbuf *m = *mp;
2252 
2253 	*mp = NULL;
2254 
2255 	if (rsvp_input_p) { /* call the real one if loaded */
2256 		*mp = m;
2257 		rsvp_input_p(mp, offp, proto);
2258 		return(IPPROTO_DONE);
2259 	}
2260 
2261 	/* Can still get packets with rsvp_on = 0 if there is a local member
2262 	 * of the group to which the RSVP packet is addressed.  But in this
2263 	 * case we want to throw the packet away.
2264 	 */
2265 
2266 	if (!rsvp_on) {
2267 		m_freem(m);
2268 		return(IPPROTO_DONE);
2269 	}
2270 
2271 	if (ip_rsvpd != NULL) {
2272 		*mp = m;
2273 		rip_input(mp, offp, proto);
2274 		return(IPPROTO_DONE);
2275 	}
2276 	/* Drop the packet */
2277 	m_freem(m);
2278 	return(IPPROTO_DONE);
2279 }
2280