xref: /netbsd-src/sys/netinet/ip_input.c (revision 5c46dd73a9bcb28b2994504ea090f64066b17a77)
1 /*	$NetBSD: ip_input.c,v 1.286 2010/04/01 01:23:32 tls Exp $	*/
2 
3 /*
4  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the project nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  */
31 
32 /*-
33  * Copyright (c) 1998 The NetBSD Foundation, Inc.
34  * All rights reserved.
35  *
36  * This code is derived from software contributed to The NetBSD Foundation
37  * by Public Access Networks Corporation ("Panix").  It was developed under
38  * contract to Panix by Eric Haszlakiewicz and Thor Lancelot Simon.
39  *
40  * Redistribution and use in source and binary forms, with or without
41  * modification, are permitted provided that the following conditions
42  * are met:
43  * 1. Redistributions of source code must retain the above copyright
44  *    notice, this list of conditions and the following disclaimer.
45  * 2. Redistributions in binary form must reproduce the above copyright
46  *    notice, this list of conditions and the following disclaimer in the
47  *    documentation and/or other materials provided with the distribution.
48  *
49  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
50  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
51  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
52  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
53  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
54  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
55  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
56  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
57  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
58  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
59  * POSSIBILITY OF SUCH DAMAGE.
60  */
61 
62 /*
63  * Copyright (c) 1982, 1986, 1988, 1993
64  *	The Regents of the University of California.  All rights reserved.
65  *
66  * Redistribution and use in source and binary forms, with or without
67  * modification, are permitted provided that the following conditions
68  * are met:
69  * 1. Redistributions of source code must retain the above copyright
70  *    notice, this list of conditions and the following disclaimer.
71  * 2. Redistributions in binary form must reproduce the above copyright
72  *    notice, this list of conditions and the following disclaimer in the
73  *    documentation and/or other materials provided with the distribution.
74  * 3. Neither the name of the University nor the names of its contributors
75  *    may be used to endorse or promote products derived from this software
76  *    without specific prior written permission.
77  *
78  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
79  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
80  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
81  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
82  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
83  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
84  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
85  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
86  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
87  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
88  * SUCH DAMAGE.
89  *
90  *	@(#)ip_input.c	8.2 (Berkeley) 1/4/94
91  */
92 
93 #include <sys/cdefs.h>
94 __KERNEL_RCSID(0, "$NetBSD: ip_input.c,v 1.286 2010/04/01 01:23:32 tls Exp $");
95 
96 #include "opt_inet.h"
97 #include "opt_compat_netbsd.h"
98 #include "opt_gateway.h"
99 #include "opt_pfil_hooks.h"
100 #include "opt_ipsec.h"
101 #include "opt_mrouting.h"
102 #include "opt_mbuftrace.h"
103 #include "opt_inet_csum.h"
104 
105 #include <sys/param.h>
106 #include <sys/systm.h>
107 #include <sys/malloc.h>
108 #include <sys/mbuf.h>
109 #include <sys/domain.h>
110 #include <sys/protosw.h>
111 #include <sys/socket.h>
112 #include <sys/socketvar.h>
113 #include <sys/errno.h>
114 #include <sys/time.h>
115 #include <sys/kernel.h>
116 #include <sys/pool.h>
117 #include <sys/sysctl.h>
118 #include <sys/kauth.h>
119 
120 #include <net/if.h>
121 #include <net/if_dl.h>
122 #include <net/route.h>
123 #include <net/pfil.h>
124 
125 #include <netinet/in.h>
126 #include <netinet/in_systm.h>
127 #include <netinet/ip.h>
128 #include <netinet/in_pcb.h>
129 #include <netinet/in_proto.h>
130 #include <netinet/in_var.h>
131 #include <netinet/ip_var.h>
132 #include <netinet/ip_private.h>
133 #include <netinet/ip_icmp.h>
134 /* just for gif_ttl */
135 #include <netinet/in_gif.h>
136 #include "gif.h"
137 #include <net/if_gre.h>
138 #include "gre.h"
139 
140 #ifdef MROUTING
141 #include <netinet/ip_mroute.h>
142 #endif
143 
144 #ifdef IPSEC
145 #include <netinet6/ipsec.h>
146 #include <netinet6/ipsec_private.h>
147 #include <netkey/key.h>
148 #endif
149 #ifdef FAST_IPSEC
150 #include <netipsec/ipsec.h>
151 #include <netipsec/key.h>
152 #endif	/* FAST_IPSEC*/
153 
154 #ifndef	IPFORWARDING
155 #ifdef GATEWAY
156 #define	IPFORWARDING	1	/* forward IP packets not for us */
157 #else /* GATEWAY */
158 #define	IPFORWARDING	0	/* don't forward IP packets not for us */
159 #endif /* GATEWAY */
160 #endif /* IPFORWARDING */
161 #ifndef	IPSENDREDIRECTS
162 #define	IPSENDREDIRECTS	1
163 #endif
164 #ifndef IPFORWSRCRT
165 #define	IPFORWSRCRT	1	/* forward source-routed packets */
166 #endif
167 #ifndef IPALLOWSRCRT
168 #define	IPALLOWSRCRT	1	/* allow source-routed packets */
169 #endif
170 #ifndef IPMTUDISC
171 #define IPMTUDISC	1
172 #endif
173 #ifndef IPMTUDISCTIMEOUT
174 #define IPMTUDISCTIMEOUT (10 * 60)	/* as per RFC 1191 */
175 #endif
176 
177 #ifdef COMPAT_50
178 #include <compat/sys/time.h>
179 #include <compat/sys/socket.h>
180 #endif
181 
182 /*
183  * Note: DIRECTED_BROADCAST is handled this way so that previous
184  * configuration using this option will Just Work.
185  */
186 #ifndef IPDIRECTEDBCAST
187 #ifdef DIRECTED_BROADCAST
188 #define IPDIRECTEDBCAST	1
189 #else
190 #define	IPDIRECTEDBCAST	0
191 #endif /* DIRECTED_BROADCAST */
192 #endif /* IPDIRECTEDBCAST */
193 int	ipforwarding = IPFORWARDING;
194 int	ipsendredirects = IPSENDREDIRECTS;
195 int	ip_defttl = IPDEFTTL;
196 int	ip_forwsrcrt = IPFORWSRCRT;
197 int	ip_directedbcast = IPDIRECTEDBCAST;
198 int	ip_allowsrcrt = IPALLOWSRCRT;
199 int	ip_mtudisc = IPMTUDISC;
200 int	ip_mtudisc_timeout = IPMTUDISCTIMEOUT;
201 #ifdef DIAGNOSTIC
202 int	ipprintfs = 0;
203 #endif
204 
205 int	ip_do_randomid = 0;
206 
207 /*
208  * XXX - Setting ip_checkinterface mostly implements the receive side of
209  * the Strong ES model described in RFC 1122, but since the routing table
210  * and transmit implementation do not implement the Strong ES model,
211  * setting this to 1 results in an odd hybrid.
212  *
213  * XXX - ip_checkinterface currently must be disabled if you use ipnat
214  * to translate the destination address to another local interface.
215  *
216  * XXX - ip_checkinterface must be disabled if you add IP aliases
217  * to the loopback interface instead of the interface where the
218  * packets for those addresses are received.
219  */
220 int	ip_checkinterface = 0;
221 
222 
223 struct rttimer_queue *ip_mtudisc_timeout_q = NULL;
224 
225 int	ipqmaxlen = IFQ_MAXLEN;
226 u_long	in_ifaddrhash;				/* size of hash table - 1 */
227 int	in_ifaddrentries;			/* total number of addrs */
228 struct in_ifaddrhead in_ifaddrhead;
229 struct	in_ifaddrhashhead *in_ifaddrhashtbl;
230 u_long	in_multihash;				/* size of hash table - 1 */
231 int	in_multientries;			/* total number of addrs */
232 struct	in_multihashhead *in_multihashtbl;
233 struct	ifqueue ipintrq;
234 
235 uint16_t ip_id;
236 
237 percpu_t *ipstat_percpu;
238 
239 #ifdef PFIL_HOOKS
240 struct pfil_head inet_pfil_hook;
241 #endif
242 
243 /*
244  * Cached copy of nmbclusters. If nbclusters is different,
245  * recalculate IP parameters derived from nmbclusters.
246  */
247 static int	ip_nmbclusters;			/* copy of nmbclusters */
248 static void	ip_nmbclusters_changed(void);	/* recalc limits */
249 
250 #define CHECK_NMBCLUSTER_PARAMS()				\
251 do {								\
252 	if (__predict_false(ip_nmbclusters != nmbclusters))	\
253 		ip_nmbclusters_changed();			\
254 } while (/*CONSTCOND*/0)
255 
256 /* IP datagram reassembly queues (hashed) */
257 #define IPREASS_NHASH_LOG2      6
258 #define IPREASS_NHASH           (1 << IPREASS_NHASH_LOG2)
259 #define IPREASS_HMASK           (IPREASS_NHASH - 1)
260 #define IPREASS_HASH(x,y) \
261 	(((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK)
262 struct ipqhead ipq[IPREASS_NHASH];
263 int	ipq_locked;
264 static int	ip_nfragpackets;	/* packets in reass queue */
265 static int	ip_nfrags;		/* total fragments in reass queues */
266 
267 int	ip_maxfragpackets = 200;	/* limit on packets. XXX sysctl */
268 int	ip_maxfrags;		        /* limit on fragments. XXX sysctl */
269 
270 
271 /*
272  * Additive-Increase/Multiplicative-Decrease (AIMD) strategy for
273  * IP reassembly queue buffer managment.
274  *
275  * We keep a count of total IP fragments (NB: not fragmented packets!)
276  * awaiting reassembly (ip_nfrags) and a limit (ip_maxfrags) on fragments.
277  * If ip_nfrags exceeds ip_maxfrags the limit, we drop half the
278  * total fragments in  reassembly queues.This AIMD policy avoids
279  * repeatedly deleting single packets under heavy fragmentation load
280  * (e.g., from lossy NFS peers).
281  */
282 static u_int	ip_reass_ttl_decr(u_int ticks);
283 static void	ip_reass_drophalf(void);
284 
285 
286 static inline int ipq_lock_try(void);
287 static inline void ipq_unlock(void);
288 
289 static inline int
290 ipq_lock_try(void)
291 {
292 	int s;
293 
294 	/*
295 	 * Use splvm() -- we're blocking things that would cause
296 	 * mbuf allocation.
297 	 */
298 	s = splvm();
299 	if (ipq_locked) {
300 		splx(s);
301 		return (0);
302 	}
303 	ipq_locked = 1;
304 	splx(s);
305 	return (1);
306 }
307 
308 static inline void
309 ipq_unlock(void)
310 {
311 	int s;
312 
313 	s = splvm();
314 	ipq_locked = 0;
315 	splx(s);
316 }
317 
318 #ifdef DIAGNOSTIC
319 #define	IPQ_LOCK()							\
320 do {									\
321 	if (ipq_lock_try() == 0) {					\
322 		printf("%s:%d: ipq already locked\n", __FILE__, __LINE__); \
323 		panic("ipq_lock");					\
324 	}								\
325 } while (/*CONSTCOND*/ 0)
326 #define	IPQ_LOCK_CHECK()						\
327 do {									\
328 	if (ipq_locked == 0) {						\
329 		printf("%s:%d: ipq lock not held\n", __FILE__, __LINE__); \
330 		panic("ipq lock check");				\
331 	}								\
332 } while (/*CONSTCOND*/ 0)
333 #else
334 #define	IPQ_LOCK()		(void) ipq_lock_try()
335 #define	IPQ_LOCK_CHECK()	/* nothing */
336 #endif
337 
338 #define	IPQ_UNLOCK()		ipq_unlock()
339 
340 struct pool inmulti_pool;
341 struct pool ipqent_pool;
342 
343 #ifdef INET_CSUM_COUNTERS
344 #include <sys/device.h>
345 
346 struct evcnt ip_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
347     NULL, "inet", "hwcsum bad");
348 struct evcnt ip_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
349     NULL, "inet", "hwcsum ok");
350 struct evcnt ip_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
351     NULL, "inet", "swcsum");
352 
353 #define	INET_CSUM_COUNTER_INCR(ev)	(ev)->ev_count++
354 
355 EVCNT_ATTACH_STATIC(ip_hwcsum_bad);
356 EVCNT_ATTACH_STATIC(ip_hwcsum_ok);
357 EVCNT_ATTACH_STATIC(ip_swcsum);
358 
359 #else
360 
361 #define	INET_CSUM_COUNTER_INCR(ev)	/* nothing */
362 
363 #endif /* INET_CSUM_COUNTERS */
364 
365 /*
366  * We need to save the IP options in case a protocol wants to respond
367  * to an incoming packet over the same route if the packet got here
368  * using IP source routing.  This allows connection establishment and
369  * maintenance when the remote end is on a network that is not known
370  * to us.
371  */
372 int	ip_nhops = 0;
373 static	struct ip_srcrt {
374 	struct	in_addr dst;			/* final destination */
375 	char	nop;				/* one NOP to align */
376 	char	srcopt[IPOPT_OFFSET + 1];	/* OPTVAL, OLEN and OFFSET */
377 	struct	in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)];
378 } ip_srcrt;
379 
380 static void save_rte(u_char *, struct in_addr);
381 
382 #ifdef MBUFTRACE
383 struct mowner ip_rx_mowner = MOWNER_INIT("internet", "rx");
384 struct mowner ip_tx_mowner = MOWNER_INIT("internet", "tx");
385 #endif
386 
387 static void sysctl_net_inet_ip_setup(struct sysctllog **);
388 
389 /*
390  * Compute IP limits derived from the value of nmbclusters.
391  */
392 static void
393 ip_nmbclusters_changed(void)
394 {
395 	ip_maxfrags = nmbclusters / 4;
396 	ip_nmbclusters =  nmbclusters;
397 }
398 
399 /*
400  * IP initialization: fill in IP protocol switch table.
401  * All protocols not implemented in kernel go to raw IP protocol handler.
402  */
403 void
404 ip_init(void)
405 {
406 	const struct protosw *pr;
407 	int i;
408 
409 	sysctl_net_inet_ip_setup(NULL);
410 
411 	pool_init(&inmulti_pool, sizeof(struct in_multi), 0, 0, 0, "inmltpl",
412 	    NULL, IPL_SOFTNET);
413 	pool_init(&ipqent_pool, sizeof(struct ipqent), 0, 0, 0, "ipqepl",
414 	    NULL, IPL_VM);
415 
416 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
417 	if (pr == 0)
418 		panic("ip_init");
419 	for (i = 0; i < IPPROTO_MAX; i++)
420 		ip_protox[i] = pr - inetsw;
421 	for (pr = inetdomain.dom_protosw;
422 	    pr < inetdomain.dom_protoswNPROTOSW; pr++)
423 		if (pr->pr_domain->dom_family == PF_INET &&
424 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
425 			ip_protox[pr->pr_protocol] = pr - inetsw;
426 
427 	for (i = 0; i < IPREASS_NHASH; i++)
428 	    	LIST_INIT(&ipq[i]);
429 
430 	ip_initid();
431 	ip_id = time_second & 0xfffff;
432 
433 	ipintrq.ifq_maxlen = ipqmaxlen;
434 	ip_nmbclusters_changed();
435 
436 	TAILQ_INIT(&in_ifaddrhead);
437 	in_ifaddrhashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, true,
438 	    &in_ifaddrhash);
439 	in_multihashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, true,
440 	    &in_multihash);
441 	ip_mtudisc_timeout_q = rt_timer_queue_create(ip_mtudisc_timeout);
442 #ifdef GATEWAY
443 	ipflow_init(ip_hashsize);
444 #endif
445 
446 #ifdef PFIL_HOOKS
447 	/* Register our Packet Filter hook. */
448 	inet_pfil_hook.ph_type = PFIL_TYPE_AF;
449 	inet_pfil_hook.ph_af   = AF_INET;
450 	i = pfil_head_register(&inet_pfil_hook);
451 	if (i != 0)
452 		printf("ip_init: WARNING: unable to register pfil hook, "
453 		    "error %d\n", i);
454 #endif /* PFIL_HOOKS */
455 
456 #ifdef MBUFTRACE
457 	MOWNER_ATTACH(&ip_tx_mowner);
458 	MOWNER_ATTACH(&ip_rx_mowner);
459 #endif /* MBUFTRACE */
460 
461 	ipstat_percpu = percpu_alloc(sizeof(uint64_t) * IP_NSTATS);
462 }
463 
464 struct	sockaddr_in ipaddr = {
465 	.sin_len = sizeof(ipaddr),
466 	.sin_family = AF_INET,
467 };
468 struct	route ipforward_rt;
469 
470 /*
471  * IP software interrupt routine
472  */
473 void
474 ipintr(void)
475 {
476 	int s;
477 	struct mbuf *m;
478 	struct ifqueue lcl_intrq;
479 
480 	memset(&lcl_intrq, 0, sizeof(lcl_intrq));
481 	ipintrq.ifq_maxlen = ipqmaxlen;
482 
483 	mutex_enter(softnet_lock);
484 	KERNEL_LOCK(1, NULL);
485 	if (!IF_IS_EMPTY(&ipintrq)) {
486 		s = splnet();
487 
488 		/* Take existing queue onto stack */
489 		lcl_intrq = ipintrq;
490 
491 		/* Zero out global queue, preserving maxlen and drops */
492 		ipintrq.ifq_head = NULL;
493 		ipintrq.ifq_tail = NULL;
494 		ipintrq.ifq_len = 0;
495 		ipintrq.ifq_maxlen = lcl_intrq.ifq_maxlen;
496 		ipintrq.ifq_drops = lcl_intrq.ifq_drops;
497 
498 		splx(s);
499 	}
500 	KERNEL_UNLOCK_ONE(NULL);
501 	while (!IF_IS_EMPTY(&lcl_intrq)) {
502 		IF_DEQUEUE(&lcl_intrq, m);
503 		if (m == NULL)
504 			break;
505 		ip_input(m);
506 	}
507 	mutex_exit(softnet_lock);
508 }
509 
510 /*
511  * Ip input routine.  Checksum and byte swap header.  If fragmented
512  * try to reassemble.  Process options.  Pass to next level.
513  */
514 void
515 ip_input(struct mbuf *m)
516 {
517 	struct ip *ip = NULL;
518 	struct ipq *fp;
519 	struct in_ifaddr *ia;
520 	struct ifaddr *ifa;
521 	struct ipqent *ipqe;
522 	int hlen = 0, mff, len;
523 	int downmatch;
524 	int checkif;
525 	int srcrt = 0;
526 	int s;
527 	u_int hash;
528 #ifdef FAST_IPSEC
529 	struct m_tag *mtag;
530 	struct tdb_ident *tdbi;
531 	struct secpolicy *sp;
532 	int error;
533 #endif /* FAST_IPSEC */
534 
535 	MCLAIM(m, &ip_rx_mowner);
536 #ifdef	DIAGNOSTIC
537 	if ((m->m_flags & M_PKTHDR) == 0)
538 		panic("ipintr no HDR");
539 #endif
540 
541 	/*
542 	 * If no IP addresses have been set yet but the interfaces
543 	 * are receiving, can't do anything with incoming packets yet.
544 	 */
545 	if (TAILQ_FIRST(&in_ifaddrhead) == 0)
546 		goto bad;
547 	IP_STATINC(IP_STAT_TOTAL);
548 	/*
549 	 * If the IP header is not aligned, slurp it up into a new
550 	 * mbuf with space for link headers, in the event we forward
551 	 * it.  Otherwise, if it is aligned, make sure the entire
552 	 * base IP header is in the first mbuf of the chain.
553 	 */
554 	if (IP_HDR_ALIGNED_P(mtod(m, void *)) == 0) {
555 		if ((m = m_copyup(m, sizeof(struct ip),
556 				  (max_linkhdr + 3) & ~3)) == NULL) {
557 			/* XXXJRT new stat, please */
558 			IP_STATINC(IP_STAT_TOOSMALL);
559 			return;
560 		}
561 	} else if (__predict_false(m->m_len < sizeof (struct ip))) {
562 		if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
563 			IP_STATINC(IP_STAT_TOOSMALL);
564 			return;
565 		}
566 	}
567 	ip = mtod(m, struct ip *);
568 	if (ip->ip_v != IPVERSION) {
569 		IP_STATINC(IP_STAT_BADVERS);
570 		goto bad;
571 	}
572 	hlen = ip->ip_hl << 2;
573 	if (hlen < sizeof(struct ip)) {	/* minimum header length */
574 		IP_STATINC(IP_STAT_BADHLEN);
575 		goto bad;
576 	}
577 	if (hlen > m->m_len) {
578 		if ((m = m_pullup(m, hlen)) == 0) {
579 			IP_STATINC(IP_STAT_BADHLEN);
580 			return;
581 		}
582 		ip = mtod(m, struct ip *);
583 	}
584 
585 	/*
586 	 * RFC1122: packets with a multicast source address are
587 	 * not allowed.
588 	 */
589 	if (IN_MULTICAST(ip->ip_src.s_addr)) {
590 		IP_STATINC(IP_STAT_BADADDR);
591 		goto bad;
592 	}
593 
594 	/* 127/8 must not appear on wire - RFC1122 */
595 	if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
596 	    (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
597 		if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
598 			IP_STATINC(IP_STAT_BADADDR);
599 			goto bad;
600 		}
601 	}
602 
603 	switch (m->m_pkthdr.csum_flags &
604 		((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_IPv4) |
605 		 M_CSUM_IPv4_BAD)) {
606 	case M_CSUM_IPv4|M_CSUM_IPv4_BAD:
607 		INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad);
608 		goto badcsum;
609 
610 	case M_CSUM_IPv4:
611 		/* Checksum was okay. */
612 		INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok);
613 		break;
614 
615 	default:
616 		/*
617 		 * Must compute it ourselves.  Maybe skip checksum on
618 		 * loopback interfaces.
619 		 */
620 		if (__predict_true(!(m->m_pkthdr.rcvif->if_flags &
621 				     IFF_LOOPBACK) || ip_do_loopback_cksum)) {
622 			INET_CSUM_COUNTER_INCR(&ip_swcsum);
623 			if (in_cksum(m, hlen) != 0)
624 				goto badcsum;
625 		}
626 		break;
627 	}
628 
629 	/* Retrieve the packet length. */
630 	len = ntohs(ip->ip_len);
631 
632 	/*
633 	 * Check for additional length bogosity
634 	 */
635 	if (len < hlen) {
636 		IP_STATINC(IP_STAT_BADLEN);
637 		goto bad;
638 	}
639 
640 	/*
641 	 * Check that the amount of data in the buffers
642 	 * is as at least much as the IP header would have us expect.
643 	 * Trim mbufs if longer than we expect.
644 	 * Drop packet if shorter than we expect.
645 	 */
646 	if (m->m_pkthdr.len < len) {
647 		IP_STATINC(IP_STAT_TOOSHORT);
648 		goto bad;
649 	}
650 	if (m->m_pkthdr.len > len) {
651 		if (m->m_len == m->m_pkthdr.len) {
652 			m->m_len = len;
653 			m->m_pkthdr.len = len;
654 		} else
655 			m_adj(m, len - m->m_pkthdr.len);
656 	}
657 
658 #if defined(IPSEC)
659 	/* ipflow (IP fast forwarding) is not compatible with IPsec. */
660 	m->m_flags &= ~M_CANFASTFWD;
661 #else
662 	/*
663 	 * Assume that we can create a fast-forward IP flow entry
664 	 * based on this packet.
665 	 */
666 	m->m_flags |= M_CANFASTFWD;
667 #endif
668 
669 #ifdef PFIL_HOOKS
670 	/*
671 	 * Run through list of hooks for input packets.  If there are any
672 	 * filters which require that additional packets in the flow are
673 	 * not fast-forwarded, they must clear the M_CANFASTFWD flag.
674 	 * Note that filters must _never_ set this flag, as another filter
675 	 * in the list may have previously cleared it.
676 	 */
677 	/*
678 	 * let ipfilter look at packet on the wire,
679 	 * not the decapsulated packet.
680 	 */
681 #ifdef IPSEC
682 	if (!ipsec_getnhist(m))
683 #elif defined(FAST_IPSEC)
684 	if (!ipsec_indone(m))
685 #else
686 	if (1)
687 #endif
688 	{
689 		struct in_addr odst;
690 
691 		odst = ip->ip_dst;
692 		if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif,
693 		    PFIL_IN) != 0)
694 			return;
695 		if (m == NULL)
696 			return;
697 		ip = mtod(m, struct ip *);
698 		hlen = ip->ip_hl << 2;
699 		/*
700 		 * XXX The setting of "srcrt" here is to prevent ip_forward()
701 		 * from generating ICMP redirects for packets that have
702 		 * been redirected by a hook back out on to the same LAN that
703 		 * they came from and is not an indication that the packet
704 		 * is being inffluenced by source routing options.  This
705 		 * allows things like
706 		 * "rdr tlp0 0/0 port 80 -> 1.1.1.200 3128 tcp"
707 		 * where tlp0 is both on the 1.1.1.0/24 network and is the
708 		 * default route for hosts on 1.1.1.0/24.  Of course this
709 		 * also requires a "map tlp0 ..." to complete the story.
710 		 * One might argue whether or not this kind of network config.
711 		 * should be supported in this manner...
712 		 */
713 		srcrt = (odst.s_addr != ip->ip_dst.s_addr);
714 	}
715 #endif /* PFIL_HOOKS */
716 
717 #ifdef ALTQ
718 	/* XXX Temporary until ALTQ is changed to use a pfil hook */
719 	if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) {
720 		/* packet dropped by traffic conditioner */
721 		return;
722 	}
723 #endif
724 
725 	/*
726 	 * Process options and, if not destined for us,
727 	 * ship it on.  ip_dooptions returns 1 when an
728 	 * error was detected (causing an icmp message
729 	 * to be sent and the original packet to be freed).
730 	 */
731 	ip_nhops = 0;		/* for source routed packets */
732 	if (hlen > sizeof (struct ip) && ip_dooptions(m))
733 		return;
734 
735 	/*
736 	 * Enable a consistency check between the destination address
737 	 * and the arrival interface for a unicast packet (the RFC 1122
738 	 * strong ES model) if IP forwarding is disabled and the packet
739 	 * is not locally generated.
740 	 *
741 	 * XXX - Checking also should be disabled if the destination
742 	 * address is ipnat'ed to a different interface.
743 	 *
744 	 * XXX - Checking is incompatible with IP aliases added
745 	 * to the loopback interface instead of the interface where
746 	 * the packets are received.
747 	 *
748 	 * XXX - We need to add a per ifaddr flag for this so that
749 	 * we get finer grain control.
750 	 */
751 	checkif = ip_checkinterface && (ipforwarding == 0) &&
752 	    (m->m_pkthdr.rcvif != NULL) &&
753 	    ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0);
754 
755 	/*
756 	 * Check our list of addresses, to see if the packet is for us.
757 	 *
758 	 * Traditional 4.4BSD did not consult IFF_UP at all.
759 	 * The behavior here is to treat addresses on !IFF_UP interface
760 	 * as not mine.
761 	 */
762 	downmatch = 0;
763 	LIST_FOREACH(ia, &IN_IFADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
764 		if (in_hosteq(ia->ia_addr.sin_addr, ip->ip_dst)) {
765 			if (checkif && ia->ia_ifp != m->m_pkthdr.rcvif)
766 				continue;
767 			if ((ia->ia_ifp->if_flags & IFF_UP) != 0)
768 				break;
769 			else
770 				downmatch++;
771 		}
772 	}
773 	if (ia != NULL)
774 		goto ours;
775 	if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
776 		IFADDR_FOREACH(ifa, m->m_pkthdr.rcvif) {
777 			if (ifa->ifa_addr->sa_family != AF_INET)
778 				continue;
779 			ia = ifatoia(ifa);
780 			if (in_hosteq(ip->ip_dst, ia->ia_broadaddr.sin_addr) ||
781 			    in_hosteq(ip->ip_dst, ia->ia_netbroadcast) ||
782 			    /*
783 			     * Look for all-0's host part (old broadcast addr),
784 			     * either for subnet or net.
785 			     */
786 			    ip->ip_dst.s_addr == ia->ia_subnet ||
787 			    ip->ip_dst.s_addr == ia->ia_net)
788 				goto ours;
789 			/*
790 			 * An interface with IP address zero accepts
791 			 * all packets that arrive on that interface.
792 			 */
793 			if (in_nullhost(ia->ia_addr.sin_addr))
794 				goto ours;
795 		}
796 	}
797 	if (IN_MULTICAST(ip->ip_dst.s_addr)) {
798 		struct in_multi *inm;
799 #ifdef MROUTING
800 		extern struct socket *ip_mrouter;
801 
802 		if (ip_mrouter) {
803 			/*
804 			 * If we are acting as a multicast router, all
805 			 * incoming multicast packets are passed to the
806 			 * kernel-level multicast forwarding function.
807 			 * The packet is returned (relatively) intact; if
808 			 * ip_mforward() returns a non-zero value, the packet
809 			 * must be discarded, else it may be accepted below.
810 			 *
811 			 * (The IP ident field is put in the same byte order
812 			 * as expected when ip_mforward() is called from
813 			 * ip_output().)
814 			 */
815 			if (ip_mforward(m, m->m_pkthdr.rcvif) != 0) {
816 				IP_STATINC(IP_STAT_CANTFORWARD);
817 				m_freem(m);
818 				return;
819 			}
820 
821 			/*
822 			 * The process-level routing demon needs to receive
823 			 * all multicast IGMP packets, whether or not this
824 			 * host belongs to their destination groups.
825 			 */
826 			if (ip->ip_p == IPPROTO_IGMP)
827 				goto ours;
828 			IP_STATINC(IP_STAT_CANTFORWARD);
829 		}
830 #endif
831 		/*
832 		 * See if we belong to the destination multicast group on the
833 		 * arrival interface.
834 		 */
835 		IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
836 		if (inm == NULL) {
837 			IP_STATINC(IP_STAT_CANTFORWARD);
838 			m_freem(m);
839 			return;
840 		}
841 		goto ours;
842 	}
843 	if (ip->ip_dst.s_addr == INADDR_BROADCAST ||
844 	    in_nullhost(ip->ip_dst))
845 		goto ours;
846 
847 	/*
848 	 * Not for us; forward if possible and desirable.
849 	 */
850 	if (ipforwarding == 0) {
851 		IP_STATINC(IP_STAT_CANTFORWARD);
852 		m_freem(m);
853 	} else {
854 		/*
855 		 * If ip_dst matched any of my address on !IFF_UP interface,
856 		 * and there's no IFF_UP interface that matches ip_dst,
857 		 * send icmp unreach.  Forwarding it will result in in-kernel
858 		 * forwarding loop till TTL goes to 0.
859 		 */
860 		if (downmatch) {
861 			icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
862 			IP_STATINC(IP_STAT_CANTFORWARD);
863 			return;
864 		}
865 #ifdef IPSEC
866 		if (ipsec4_in_reject(m, NULL)) {
867 			IPSEC_STATINC(IPSEC_STAT_IN_POLVIO);
868 			goto bad;
869 		}
870 #endif
871 #ifdef FAST_IPSEC
872 		mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
873 		s = splsoftnet();
874 		if (mtag != NULL) {
875 			tdbi = (struct tdb_ident *)(mtag + 1);
876 			sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
877 		} else {
878 			sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
879 						   IP_FORWARDING, &error);
880 		}
881 		if (sp == NULL) {	/* NB: can happen if error */
882 			splx(s);
883 			/*XXX error stat???*/
884 			DPRINTF(("ip_input: no SP for forwarding\n"));	/*XXX*/
885 			goto bad;
886 		}
887 
888 		/*
889 		 * Check security policy against packet attributes.
890 		 */
891 		error = ipsec_in_reject(sp, m);
892 		KEY_FREESP(&sp);
893 		splx(s);
894 		if (error) {
895 			IP_STATINC(IP_STAT_CANTFORWARD);
896 			goto bad;
897 		}
898 
899 		/*
900 		 * Peek at the outbound SP for this packet to determine if
901 		 * it's a Fast Forward candidate.
902 		 */
903 		mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL);
904 		if (mtag != NULL)
905 			m->m_flags &= ~M_CANFASTFWD;
906 		else {
907 			s = splsoftnet();
908 			sp = ipsec4_checkpolicy(m, IPSEC_DIR_OUTBOUND,
909 			    (IP_FORWARDING |
910 			     (ip_directedbcast ? IP_ALLOWBROADCAST : 0)),
911 			    &error, NULL);
912 			if (sp != NULL) {
913 				m->m_flags &= ~M_CANFASTFWD;
914 				KEY_FREESP(&sp);
915 			}
916 			splx(s);
917 		}
918 #endif	/* FAST_IPSEC */
919 
920 		ip_forward(m, srcrt);
921 	}
922 	return;
923 
924 ours:
925 	/*
926 	 * If offset or IP_MF are set, must reassemble.
927 	 * Otherwise, nothing need be done.
928 	 * (We could look in the reassembly queue to see
929 	 * if the packet was previously fragmented,
930 	 * but it's not worth the time; just let them time out.)
931 	 */
932 	if (ip->ip_off & ~htons(IP_DF|IP_RF)) {
933 		uint16_t off;
934 		/*
935 		 * Prevent TCP blind data attacks by not allowing non-initial
936 		 * fragments to start at less than 68 bytes (minimal fragment
937 		 * size) and making sure the first fragment is at least 68
938 		 * bytes.
939 		 */
940 		off = (ntohs(ip->ip_off) & IP_OFFMASK) << 3;
941 		if ((off > 0 ? off + hlen : len) < IP_MINFRAGSIZE - 1) {
942 			IP_STATINC(IP_STAT_BADFRAGS);
943 			goto bad;
944 		}
945 		/*
946 		 * Look for queue of fragments
947 		 * of this datagram.
948 		 */
949 		IPQ_LOCK();
950 		hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
951 		LIST_FOREACH(fp, &ipq[hash], ipq_q) {
952 			if (ip->ip_id == fp->ipq_id &&
953 			    in_hosteq(ip->ip_src, fp->ipq_src) &&
954 			    in_hosteq(ip->ip_dst, fp->ipq_dst) &&
955 			    ip->ip_p == fp->ipq_p) {
956 				/*
957 				 * Make sure the TOS is matches previous
958 				 * fragments.
959 				 */
960 				if (ip->ip_tos != fp->ipq_tos) {
961 					IP_STATINC(IP_STAT_BADFRAGS);
962 					IPQ_UNLOCK();
963 					goto bad;
964 				}
965 				goto found;
966 			}
967 		}
968 		fp = 0;
969 found:
970 
971 		/*
972 		 * Adjust ip_len to not reflect header,
973 		 * set ipqe_mff if more fragments are expected,
974 		 * convert offset of this to bytes.
975 		 */
976 		ip->ip_len = htons(ntohs(ip->ip_len) - hlen);
977 		mff = (ip->ip_off & htons(IP_MF)) != 0;
978 		if (mff) {
979 		        /*
980 		         * Make sure that fragments have a data length
981 			 * that's a non-zero multiple of 8 bytes.
982 		         */
983 			if (ntohs(ip->ip_len) == 0 ||
984 			    (ntohs(ip->ip_len) & 0x7) != 0) {
985 				IP_STATINC(IP_STAT_BADFRAGS);
986 				IPQ_UNLOCK();
987 				goto bad;
988 			}
989 		}
990 		ip->ip_off = htons((ntohs(ip->ip_off) & IP_OFFMASK) << 3);
991 
992 		/*
993 		 * If datagram marked as having more fragments
994 		 * or if this is not the first fragment,
995 		 * attempt reassembly; if it succeeds, proceed.
996 		 */
997 		if (mff || ip->ip_off != htons(0)) {
998 			IP_STATINC(IP_STAT_FRAGMENTS);
999 			s = splvm();
1000 			ipqe = pool_get(&ipqent_pool, PR_NOWAIT);
1001 			splx(s);
1002 			if (ipqe == NULL) {
1003 				IP_STATINC(IP_STAT_RCVMEMDROP);
1004 				IPQ_UNLOCK();
1005 				goto bad;
1006 			}
1007 			ipqe->ipqe_mff = mff;
1008 			ipqe->ipqe_m = m;
1009 			ipqe->ipqe_ip = ip;
1010 			m = ip_reass(ipqe, fp, &ipq[hash]);
1011 			if (m == 0) {
1012 				IPQ_UNLOCK();
1013 				return;
1014 			}
1015 			IP_STATINC(IP_STAT_REASSEMBLED);
1016 			ip = mtod(m, struct ip *);
1017 			hlen = ip->ip_hl << 2;
1018 			ip->ip_len = htons(ntohs(ip->ip_len) + hlen);
1019 		} else
1020 			if (fp)
1021 				ip_freef(fp);
1022 		IPQ_UNLOCK();
1023 	}
1024 
1025 #if defined(IPSEC)
1026 	/*
1027 	 * enforce IPsec policy checking if we are seeing last header.
1028 	 * note that we do not visit this with protocols with pcb layer
1029 	 * code - like udp/tcp/raw ip.
1030 	 */
1031 	if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0 &&
1032 	    ipsec4_in_reject(m, NULL)) {
1033 		IPSEC_STATINC(IPSEC_STAT_IN_POLVIO);
1034 		goto bad;
1035 	}
1036 #endif
1037 #ifdef FAST_IPSEC
1038 	/*
1039 	 * enforce IPsec policy checking if we are seeing last header.
1040 	 * note that we do not visit this with protocols with pcb layer
1041 	 * code - like udp/tcp/raw ip.
1042 	 */
1043 	if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) {
1044 		/*
1045 		 * Check if the packet has already had IPsec processing
1046 		 * done.  If so, then just pass it along.  This tag gets
1047 		 * set during AH, ESP, etc. input handling, before the
1048 		 * packet is returned to the ip input queue for delivery.
1049 		 */
1050 		mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
1051 		s = splsoftnet();
1052 		if (mtag != NULL) {
1053 			tdbi = (struct tdb_ident *)(mtag + 1);
1054 			sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
1055 		} else {
1056 			sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
1057 						   IP_FORWARDING, &error);
1058 		}
1059 		if (sp != NULL) {
1060 			/*
1061 			 * Check security policy against packet attributes.
1062 			 */
1063 			error = ipsec_in_reject(sp, m);
1064 			KEY_FREESP(&sp);
1065 		} else {
1066 			/* XXX error stat??? */
1067 			error = EINVAL;
1068 DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/
1069 		}
1070 		splx(s);
1071 		if (error)
1072 			goto bad;
1073 	}
1074 #endif /* FAST_IPSEC */
1075 
1076 	/*
1077 	 * Switch out to protocol's input routine.
1078 	 */
1079 #if IFA_STATS
1080 	if (ia && ip)
1081 		ia->ia_ifa.ifa_data.ifad_inbytes += ntohs(ip->ip_len);
1082 #endif
1083 	IP_STATINC(IP_STAT_DELIVERED);
1084     {
1085 	int off = hlen, nh = ip->ip_p;
1086 
1087 	(*inetsw[ip_protox[nh]].pr_input)(m, off, nh);
1088 	return;
1089     }
1090 bad:
1091 	m_freem(m);
1092 	return;
1093 
1094 badcsum:
1095 	IP_STATINC(IP_STAT_BADSUM);
1096 	m_freem(m);
1097 }
1098 
1099 /*
1100  * Take incoming datagram fragment and try to
1101  * reassemble it into whole datagram.  If a chain for
1102  * reassembly of this datagram already exists, then it
1103  * is given as fp; otherwise have to make a chain.
1104  */
1105 struct mbuf *
1106 ip_reass(struct ipqent *ipqe, struct ipq *fp, struct ipqhead *ipqhead)
1107 {
1108 	struct mbuf *m = ipqe->ipqe_m;
1109 	struct ipqent *nq, *p, *q;
1110 	struct ip *ip;
1111 	struct mbuf *t;
1112 	int hlen = ipqe->ipqe_ip->ip_hl << 2;
1113 	int i, next, s;
1114 
1115 	IPQ_LOCK_CHECK();
1116 
1117 	/*
1118 	 * Presence of header sizes in mbufs
1119 	 * would confuse code below.
1120 	 */
1121 	m->m_data += hlen;
1122 	m->m_len -= hlen;
1123 
1124 #ifdef	notyet
1125 	/* make sure fragment limit is up-to-date */
1126 	CHECK_NMBCLUSTER_PARAMS();
1127 
1128 	/* If we have too many fragments, drop the older half. */
1129 	if (ip_nfrags >= ip_maxfrags)
1130 		ip_reass_drophalf(void);
1131 #endif
1132 
1133 	/*
1134 	 * We are about to add a fragment; increment frag count.
1135 	 */
1136 	ip_nfrags++;
1137 
1138 	/*
1139 	 * If first fragment to arrive, create a reassembly queue.
1140 	 */
1141 	if (fp == 0) {
1142 		/*
1143 		 * Enforce upper bound on number of fragmented packets
1144 		 * for which we attempt reassembly;
1145 		 * If maxfrag is 0, never accept fragments.
1146 		 * If maxfrag is -1, accept all fragments without limitation.
1147 		 */
1148 		if (ip_maxfragpackets < 0)
1149 			;
1150 		else if (ip_nfragpackets >= ip_maxfragpackets)
1151 			goto dropfrag;
1152 		ip_nfragpackets++;
1153 		fp = malloc(sizeof (struct ipq), M_FTABLE, M_NOWAIT);
1154 		if (fp == NULL)
1155 			goto dropfrag;
1156 		LIST_INSERT_HEAD(ipqhead, fp, ipq_q);
1157 		fp->ipq_nfrags = 1;
1158 		fp->ipq_ttl = IPFRAGTTL;
1159 		fp->ipq_p = ipqe->ipqe_ip->ip_p;
1160 		fp->ipq_id = ipqe->ipqe_ip->ip_id;
1161 		fp->ipq_tos = ipqe->ipqe_ip->ip_tos;
1162 		TAILQ_INIT(&fp->ipq_fragq);
1163 		fp->ipq_src = ipqe->ipqe_ip->ip_src;
1164 		fp->ipq_dst = ipqe->ipqe_ip->ip_dst;
1165 		p = NULL;
1166 		goto insert;
1167 	} else {
1168 		fp->ipq_nfrags++;
1169 	}
1170 
1171 	/*
1172 	 * Find a segment which begins after this one does.
1173 	 */
1174 	for (p = NULL, q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL;
1175 	    p = q, q = TAILQ_NEXT(q, ipqe_q))
1176 		if (ntohs(q->ipqe_ip->ip_off) > ntohs(ipqe->ipqe_ip->ip_off))
1177 			break;
1178 
1179 	/*
1180 	 * If there is a preceding segment, it may provide some of
1181 	 * our data already.  If so, drop the data from the incoming
1182 	 * segment.  If it provides all of our data, drop us.
1183 	 */
1184 	if (p != NULL) {
1185 		i = ntohs(p->ipqe_ip->ip_off) + ntohs(p->ipqe_ip->ip_len) -
1186 		    ntohs(ipqe->ipqe_ip->ip_off);
1187 		if (i > 0) {
1188 			if (i >= ntohs(ipqe->ipqe_ip->ip_len))
1189 				goto dropfrag;
1190 			m_adj(ipqe->ipqe_m, i);
1191 			ipqe->ipqe_ip->ip_off =
1192 			    htons(ntohs(ipqe->ipqe_ip->ip_off) + i);
1193 			ipqe->ipqe_ip->ip_len =
1194 			    htons(ntohs(ipqe->ipqe_ip->ip_len) - i);
1195 		}
1196 	}
1197 
1198 	/*
1199 	 * While we overlap succeeding segments trim them or,
1200 	 * if they are completely covered, dequeue them.
1201 	 */
1202 	for (; q != NULL &&
1203 	    ntohs(ipqe->ipqe_ip->ip_off) + ntohs(ipqe->ipqe_ip->ip_len) >
1204 	    ntohs(q->ipqe_ip->ip_off); q = nq) {
1205 		i = (ntohs(ipqe->ipqe_ip->ip_off) +
1206 		    ntohs(ipqe->ipqe_ip->ip_len)) - ntohs(q->ipqe_ip->ip_off);
1207 		if (i < ntohs(q->ipqe_ip->ip_len)) {
1208 			q->ipqe_ip->ip_len =
1209 			    htons(ntohs(q->ipqe_ip->ip_len) - i);
1210 			q->ipqe_ip->ip_off =
1211 			    htons(ntohs(q->ipqe_ip->ip_off) + i);
1212 			m_adj(q->ipqe_m, i);
1213 			break;
1214 		}
1215 		nq = TAILQ_NEXT(q, ipqe_q);
1216 		m_freem(q->ipqe_m);
1217 		TAILQ_REMOVE(&fp->ipq_fragq, q, ipqe_q);
1218 		s = splvm();
1219 		pool_put(&ipqent_pool, q);
1220 		splx(s);
1221 		fp->ipq_nfrags--;
1222 		ip_nfrags--;
1223 	}
1224 
1225 insert:
1226 	/*
1227 	 * Stick new segment in its place;
1228 	 * check for complete reassembly.
1229 	 */
1230 	if (p == NULL) {
1231 		TAILQ_INSERT_HEAD(&fp->ipq_fragq, ipqe, ipqe_q);
1232 	} else {
1233 		TAILQ_INSERT_AFTER(&fp->ipq_fragq, p, ipqe, ipqe_q);
1234 	}
1235 	next = 0;
1236 	for (p = NULL, q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL;
1237 	    p = q, q = TAILQ_NEXT(q, ipqe_q)) {
1238 		if (ntohs(q->ipqe_ip->ip_off) != next)
1239 			return (0);
1240 		next += ntohs(q->ipqe_ip->ip_len);
1241 	}
1242 	if (p->ipqe_mff)
1243 		return (0);
1244 
1245 	/*
1246 	 * Reassembly is complete.  Check for a bogus message size and
1247 	 * concatenate fragments.
1248 	 */
1249 	q = TAILQ_FIRST(&fp->ipq_fragq);
1250 	ip = q->ipqe_ip;
1251 	if ((next + (ip->ip_hl << 2)) > IP_MAXPACKET) {
1252 		IP_STATINC(IP_STAT_TOOLONG);
1253 		ip_freef(fp);
1254 		return (0);
1255 	}
1256 	m = q->ipqe_m;
1257 	t = m->m_next;
1258 	m->m_next = 0;
1259 	m_cat(m, t);
1260 	nq = TAILQ_NEXT(q, ipqe_q);
1261 	s = splvm();
1262 	pool_put(&ipqent_pool, q);
1263 	splx(s);
1264 	for (q = nq; q != NULL; q = nq) {
1265 		t = q->ipqe_m;
1266 		nq = TAILQ_NEXT(q, ipqe_q);
1267 		s = splvm();
1268 		pool_put(&ipqent_pool, q);
1269 		splx(s);
1270 		m_cat(m, t);
1271 	}
1272 	ip_nfrags -= fp->ipq_nfrags;
1273 
1274 	/*
1275 	 * Create header for new ip packet by
1276 	 * modifying header of first packet;
1277 	 * dequeue and discard fragment reassembly header.
1278 	 * Make header visible.
1279 	 */
1280 	ip->ip_len = htons(next);
1281 	ip->ip_src = fp->ipq_src;
1282 	ip->ip_dst = fp->ipq_dst;
1283 	LIST_REMOVE(fp, ipq_q);
1284 	free(fp, M_FTABLE);
1285 	ip_nfragpackets--;
1286 	m->m_len += (ip->ip_hl << 2);
1287 	m->m_data -= (ip->ip_hl << 2);
1288 	/* some debugging cruft by sklower, below, will go away soon */
1289 	if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */
1290 		int plen = 0;
1291 		for (t = m; t; t = t->m_next)
1292 			plen += t->m_len;
1293 		m->m_pkthdr.len = plen;
1294 		m->m_pkthdr.csum_flags = 0;
1295 	}
1296 	return (m);
1297 
1298 dropfrag:
1299 	if (fp != 0)
1300 		fp->ipq_nfrags--;
1301 	ip_nfrags--;
1302 	IP_STATINC(IP_STAT_FRAGDROPPED);
1303 	m_freem(m);
1304 	s = splvm();
1305 	pool_put(&ipqent_pool, ipqe);
1306 	splx(s);
1307 	return (0);
1308 }
1309 
1310 /*
1311  * Free a fragment reassembly header and all
1312  * associated datagrams.
1313  */
1314 void
1315 ip_freef(struct ipq *fp)
1316 {
1317 	struct ipqent *q, *p;
1318 	u_int nfrags = 0;
1319 	int s;
1320 
1321 	IPQ_LOCK_CHECK();
1322 
1323 	for (q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL; q = p) {
1324 		p = TAILQ_NEXT(q, ipqe_q);
1325 		m_freem(q->ipqe_m);
1326 		nfrags++;
1327 		TAILQ_REMOVE(&fp->ipq_fragq, q, ipqe_q);
1328 		s = splvm();
1329 		pool_put(&ipqent_pool, q);
1330 		splx(s);
1331 	}
1332 
1333 	if (nfrags != fp->ipq_nfrags)
1334 	    printf("ip_freef: nfrags %d != %d\n", fp->ipq_nfrags, nfrags);
1335 	ip_nfrags -= nfrags;
1336 	LIST_REMOVE(fp, ipq_q);
1337 	free(fp, M_FTABLE);
1338 	ip_nfragpackets--;
1339 }
1340 
1341 /*
1342  * IP reassembly TTL machinery for  multiplicative drop.
1343  */
1344 static u_int	fragttl_histo[(IPFRAGTTL+1)];
1345 
1346 
1347 /*
1348  * Decrement TTL of all reasembly queue entries by `ticks'.
1349  * Count number of distinct fragments (as opposed to partial, fragmented
1350  * datagrams) in the reassembly queue.  While we  traverse the entire
1351  * reassembly queue, compute and return the median TTL over all fragments.
1352  */
1353 static u_int
1354 ip_reass_ttl_decr(u_int ticks)
1355 {
1356 	u_int nfrags, median, dropfraction, keepfraction;
1357 	struct ipq *fp, *nfp;
1358 	int i;
1359 
1360 	nfrags = 0;
1361 	memset(fragttl_histo, 0, sizeof fragttl_histo);
1362 
1363 	for (i = 0; i < IPREASS_NHASH; i++) {
1364 		for (fp = LIST_FIRST(&ipq[i]); fp != NULL; fp = nfp) {
1365 			fp->ipq_ttl = ((fp->ipq_ttl  <= ticks) ?
1366 				       0 : fp->ipq_ttl - ticks);
1367 			nfp = LIST_NEXT(fp, ipq_q);
1368 			if (fp->ipq_ttl == 0) {
1369 				IP_STATINC(IP_STAT_FRAGTIMEOUT);
1370 				ip_freef(fp);
1371 			} else {
1372 				nfrags += fp->ipq_nfrags;
1373 				fragttl_histo[fp->ipq_ttl] += fp->ipq_nfrags;
1374 			}
1375 		}
1376 	}
1377 
1378 	KASSERT(ip_nfrags == nfrags);
1379 
1380 	/* Find median (or other drop fraction) in histogram. */
1381 	dropfraction = (ip_nfrags / 2);
1382 	keepfraction = ip_nfrags - dropfraction;
1383 	for (i = IPFRAGTTL, median = 0; i >= 0; i--) {
1384 		median +=  fragttl_histo[i];
1385 		if (median >= keepfraction)
1386 			break;
1387 	}
1388 
1389 	/* Return TTL of median (or other fraction). */
1390 	return (u_int)i;
1391 }
1392 
1393 void
1394 ip_reass_drophalf(void)
1395 {
1396 
1397 	u_int median_ticks;
1398 	/*
1399 	 * Compute median TTL of all fragments, and count frags
1400 	 * with that TTL or lower (roughly half of all fragments).
1401 	 */
1402 	median_ticks = ip_reass_ttl_decr(0);
1403 
1404 	/* Drop half. */
1405 	median_ticks = ip_reass_ttl_decr(median_ticks);
1406 
1407 }
1408 
1409 /*
1410  * IP timer processing;
1411  * if a timer expires on a reassembly
1412  * queue, discard it.
1413  */
1414 void
1415 ip_slowtimo(void)
1416 {
1417 	static u_int dropscanidx = 0;
1418 	u_int i;
1419 	u_int median_ttl;
1420 
1421 	mutex_enter(softnet_lock);
1422 	KERNEL_LOCK(1, NULL);
1423 
1424 	IPQ_LOCK();
1425 
1426 	/* Age TTL of all fragments by 1 tick .*/
1427 	median_ttl = ip_reass_ttl_decr(1);
1428 
1429 	/* make sure fragment limit is up-to-date */
1430 	CHECK_NMBCLUSTER_PARAMS();
1431 
1432 	/* If we have too many fragments, drop the older half. */
1433 	if (ip_nfrags > ip_maxfrags)
1434 		ip_reass_ttl_decr(median_ttl);
1435 
1436 	/*
1437 	 * If we are over the maximum number of fragmented packets
1438 	 * (due to the limit being lowered), drain off
1439 	 * enough to get down to the new limit. Start draining
1440 	 * from the reassembly hashqueue most recently drained.
1441 	 */
1442 	if (ip_maxfragpackets < 0)
1443 		;
1444 	else {
1445 		int wrapped = 0;
1446 
1447 		i = dropscanidx;
1448 		while (ip_nfragpackets > ip_maxfragpackets && wrapped == 0) {
1449 			while (LIST_FIRST(&ipq[i]) != NULL)
1450 				ip_freef(LIST_FIRST(&ipq[i]));
1451 			if (++i >= IPREASS_NHASH) {
1452 				i = 0;
1453 			}
1454 			/*
1455 			 * Dont scan forever even if fragment counters are
1456 			 * wrong: stop after scanning entire reassembly queue.
1457 			 */
1458 			if (i == dropscanidx)
1459 			    wrapped = 1;
1460 		}
1461 		dropscanidx = i;
1462 	}
1463 	IPQ_UNLOCK();
1464 
1465 	KERNEL_UNLOCK_ONE(NULL);
1466 	mutex_exit(softnet_lock);
1467 }
1468 
1469 /*
1470  * Drain off all datagram fragments.  Don't acquire softnet_lock as
1471  * can be called from hardware interrupt context.
1472  */
1473 void
1474 ip_drain(void)
1475 {
1476 
1477 	KERNEL_LOCK(1, NULL);
1478 
1479 	/*
1480 	 * We may be called from a device's interrupt context.  If
1481 	 * the ipq is already busy, just bail out now.
1482 	 */
1483 	if (ipq_lock_try() != 0) {
1484 		/*
1485 		 * Drop half the total fragments now. If more mbufs are
1486 		 * needed, we will be called again soon.
1487 		 */
1488 		ip_reass_drophalf();
1489 		IPQ_UNLOCK();
1490 	}
1491 
1492 	KERNEL_UNLOCK_ONE(NULL);
1493 }
1494 
1495 /*
1496  * Do option processing on a datagram,
1497  * possibly discarding it if bad options are encountered,
1498  * or forwarding it if source-routed.
1499  * Returns 1 if packet has been forwarded/freed,
1500  * 0 if the packet should be processed further.
1501  */
1502 int
1503 ip_dooptions(struct mbuf *m)
1504 {
1505 	struct ip *ip = mtod(m, struct ip *);
1506 	u_char *cp, *cp0;
1507 	struct ip_timestamp *ipt;
1508 	struct in_ifaddr *ia;
1509 	int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0;
1510 	struct in_addr dst;
1511 	n_time ntime;
1512 
1513 	dst = ip->ip_dst;
1514 	cp = (u_char *)(ip + 1);
1515 	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
1516 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1517 		opt = cp[IPOPT_OPTVAL];
1518 		if (opt == IPOPT_EOL)
1519 			break;
1520 		if (opt == IPOPT_NOP)
1521 			optlen = 1;
1522 		else {
1523 			if (cnt < IPOPT_OLEN + sizeof(*cp)) {
1524 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1525 				goto bad;
1526 			}
1527 			optlen = cp[IPOPT_OLEN];
1528 			if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
1529 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1530 				goto bad;
1531 			}
1532 		}
1533 		switch (opt) {
1534 
1535 		default:
1536 			break;
1537 
1538 		/*
1539 		 * Source routing with record.
1540 		 * Find interface with current destination address.
1541 		 * If none on this machine then drop if strictly routed,
1542 		 * or do nothing if loosely routed.
1543 		 * Record interface address and bring up next address
1544 		 * component.  If strictly routed make sure next
1545 		 * address is on directly accessible net.
1546 		 */
1547 		case IPOPT_LSRR:
1548 		case IPOPT_SSRR:
1549 			if (ip_allowsrcrt == 0) {
1550 				type = ICMP_UNREACH;
1551 				code = ICMP_UNREACH_NET_PROHIB;
1552 				goto bad;
1553 			}
1554 			if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1555 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1556 				goto bad;
1557 			}
1558 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1559 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1560 				goto bad;
1561 			}
1562 			ipaddr.sin_addr = ip->ip_dst;
1563 			ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr)));
1564 			if (ia == 0) {
1565 				if (opt == IPOPT_SSRR) {
1566 					type = ICMP_UNREACH;
1567 					code = ICMP_UNREACH_SRCFAIL;
1568 					goto bad;
1569 				}
1570 				/*
1571 				 * Loose routing, and not at next destination
1572 				 * yet; nothing to do except forward.
1573 				 */
1574 				break;
1575 			}
1576 			off--;			/* 0 origin */
1577 			if ((off + sizeof(struct in_addr)) > optlen) {
1578 				/*
1579 				 * End of source route.  Should be for us.
1580 				 */
1581 				save_rte(cp, ip->ip_src);
1582 				break;
1583 			}
1584 			/*
1585 			 * locate outgoing interface
1586 			 */
1587 			memcpy((void *)&ipaddr.sin_addr, (void *)(cp + off),
1588 			    sizeof(ipaddr.sin_addr));
1589 			if (opt == IPOPT_SSRR)
1590 				ia = ifatoia(ifa_ifwithladdr(sintosa(&ipaddr)));
1591 			else
1592 				ia = ip_rtaddr(ipaddr.sin_addr);
1593 			if (ia == 0) {
1594 				type = ICMP_UNREACH;
1595 				code = ICMP_UNREACH_SRCFAIL;
1596 				goto bad;
1597 			}
1598 			ip->ip_dst = ipaddr.sin_addr;
1599 			bcopy((void *)&ia->ia_addr.sin_addr,
1600 			    (void *)(cp + off), sizeof(struct in_addr));
1601 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1602 			/*
1603 			 * Let ip_intr's mcast routing check handle mcast pkts
1604 			 */
1605 			forward = !IN_MULTICAST(ip->ip_dst.s_addr);
1606 			break;
1607 
1608 		case IPOPT_RR:
1609 			if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1610 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1611 				goto bad;
1612 			}
1613 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1614 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1615 				goto bad;
1616 			}
1617 			/*
1618 			 * If no space remains, ignore.
1619 			 */
1620 			off--;			/* 0 origin */
1621 			if ((off + sizeof(struct in_addr)) > optlen)
1622 				break;
1623 			memcpy((void *)&ipaddr.sin_addr, (void *)(&ip->ip_dst),
1624 			    sizeof(ipaddr.sin_addr));
1625 			/*
1626 			 * locate outgoing interface; if we're the destination,
1627 			 * use the incoming interface (should be same).
1628 			 */
1629 			if ((ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))))
1630 			    == NULL &&
1631 			    (ia = ip_rtaddr(ipaddr.sin_addr)) == NULL) {
1632 				type = ICMP_UNREACH;
1633 				code = ICMP_UNREACH_HOST;
1634 				goto bad;
1635 			}
1636 			bcopy((void *)&ia->ia_addr.sin_addr,
1637 			    (void *)(cp + off), sizeof(struct in_addr));
1638 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1639 			break;
1640 
1641 		case IPOPT_TS:
1642 			code = cp - (u_char *)ip;
1643 			ipt = (struct ip_timestamp *)cp;
1644 			if (ipt->ipt_len < 4 || ipt->ipt_len > 40) {
1645 				code = (u_char *)&ipt->ipt_len - (u_char *)ip;
1646 				goto bad;
1647 			}
1648 			if (ipt->ipt_ptr < 5) {
1649 				code = (u_char *)&ipt->ipt_ptr - (u_char *)ip;
1650 				goto bad;
1651 			}
1652 			if (ipt->ipt_ptr > ipt->ipt_len - sizeof (int32_t)) {
1653 				if (++ipt->ipt_oflw == 0) {
1654 					code = (u_char *)&ipt->ipt_ptr -
1655 					    (u_char *)ip;
1656 					goto bad;
1657 				}
1658 				break;
1659 			}
1660 			cp0 = (cp + ipt->ipt_ptr - 1);
1661 			switch (ipt->ipt_flg) {
1662 
1663 			case IPOPT_TS_TSONLY:
1664 				break;
1665 
1666 			case IPOPT_TS_TSANDADDR:
1667 				if (ipt->ipt_ptr - 1 + sizeof(n_time) +
1668 				    sizeof(struct in_addr) > ipt->ipt_len) {
1669 					code = (u_char *)&ipt->ipt_ptr -
1670 					    (u_char *)ip;
1671 					goto bad;
1672 				}
1673 				ipaddr.sin_addr = dst;
1674 				ia = ifatoia(ifaof_ifpforaddr(sintosa(&ipaddr),
1675 				    m->m_pkthdr.rcvif));
1676 				if (ia == 0)
1677 					continue;
1678 				bcopy(&ia->ia_addr.sin_addr,
1679 				    cp0, sizeof(struct in_addr));
1680 				ipt->ipt_ptr += sizeof(struct in_addr);
1681 				break;
1682 
1683 			case IPOPT_TS_PRESPEC:
1684 				if (ipt->ipt_ptr - 1 + sizeof(n_time) +
1685 				    sizeof(struct in_addr) > ipt->ipt_len) {
1686 					code = (u_char *)&ipt->ipt_ptr -
1687 					    (u_char *)ip;
1688 					goto bad;
1689 				}
1690 				memcpy(&ipaddr.sin_addr, cp0,
1691 				    sizeof(struct in_addr));
1692 				if (ifatoia(ifa_ifwithaddr(sintosa(&ipaddr)))
1693 				    == NULL)
1694 					continue;
1695 				ipt->ipt_ptr += sizeof(struct in_addr);
1696 				break;
1697 
1698 			default:
1699 				/* XXX can't take &ipt->ipt_flg */
1700 				code = (u_char *)&ipt->ipt_ptr -
1701 				    (u_char *)ip + 1;
1702 				goto bad;
1703 			}
1704 			ntime = iptime();
1705 			cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */
1706 			memmove((char *)cp + ipt->ipt_ptr - 1, cp0,
1707 			    sizeof(n_time));
1708 			ipt->ipt_ptr += sizeof(n_time);
1709 		}
1710 	}
1711 	if (forward) {
1712 		if (ip_forwsrcrt == 0) {
1713 			type = ICMP_UNREACH;
1714 			code = ICMP_UNREACH_SRCFAIL;
1715 			goto bad;
1716 		}
1717 		ip_forward(m, 1);
1718 		return (1);
1719 	}
1720 	return (0);
1721 bad:
1722 	icmp_error(m, type, code, 0, 0);
1723 	IP_STATINC(IP_STAT_BADOPTIONS);
1724 	return (1);
1725 }
1726 
1727 /*
1728  * Given address of next destination (final or next hop),
1729  * return internet address info of interface to be used to get there.
1730  */
1731 struct in_ifaddr *
1732 ip_rtaddr(struct in_addr dst)
1733 {
1734 	struct rtentry *rt;
1735 	union {
1736 		struct sockaddr		dst;
1737 		struct sockaddr_in	dst4;
1738 	} u;
1739 
1740 	sockaddr_in_init(&u.dst4, &dst, 0);
1741 
1742 	if ((rt = rtcache_lookup(&ipforward_rt, &u.dst)) == NULL)
1743 		return NULL;
1744 
1745 	return ifatoia(rt->rt_ifa);
1746 }
1747 
1748 /*
1749  * Save incoming source route for use in replies,
1750  * to be picked up later by ip_srcroute if the receiver is interested.
1751  */
1752 void
1753 save_rte(u_char *option, struct in_addr dst)
1754 {
1755 	unsigned olen;
1756 
1757 	olen = option[IPOPT_OLEN];
1758 #ifdef DIAGNOSTIC
1759 	if (ipprintfs)
1760 		printf("save_rte: olen %d\n", olen);
1761 #endif /* 0 */
1762 	if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst)))
1763 		return;
1764 	memcpy((void *)ip_srcrt.srcopt, (void *)option, olen);
1765 	ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1766 	ip_srcrt.dst = dst;
1767 }
1768 
1769 /*
1770  * Retrieve incoming source route for use in replies,
1771  * in the same form used by setsockopt.
1772  * The first hop is placed before the options, will be removed later.
1773  */
1774 struct mbuf *
1775 ip_srcroute(void)
1776 {
1777 	struct in_addr *p, *q;
1778 	struct mbuf *m;
1779 
1780 	if (ip_nhops == 0)
1781 		return NULL;
1782 	m = m_get(M_DONTWAIT, MT_SOOPTS);
1783 	if (m == 0)
1784 		return NULL;
1785 
1786 	MCLAIM(m, &inetdomain.dom_mowner);
1787 #define OPTSIZ	(sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt))
1788 
1789 	/* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
1790 	m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) +
1791 	    OPTSIZ;
1792 #ifdef DIAGNOSTIC
1793 	if (ipprintfs)
1794 		printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len);
1795 #endif
1796 
1797 	/*
1798 	 * First save first hop for return route
1799 	 */
1800 	p = &ip_srcrt.route[ip_nhops - 1];
1801 	*(mtod(m, struct in_addr *)) = *p--;
1802 #ifdef DIAGNOSTIC
1803 	if (ipprintfs)
1804 		printf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr));
1805 #endif
1806 
1807 	/*
1808 	 * Copy option fields and padding (nop) to mbuf.
1809 	 */
1810 	ip_srcrt.nop = IPOPT_NOP;
1811 	ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
1812 	memmove(mtod(m, char *) + sizeof(struct in_addr), &ip_srcrt.nop,
1813 	    OPTSIZ);
1814 	q = (struct in_addr *)(mtod(m, char *) +
1815 	    sizeof(struct in_addr) + OPTSIZ);
1816 #undef OPTSIZ
1817 	/*
1818 	 * Record return path as an IP source route,
1819 	 * reversing the path (pointers are now aligned).
1820 	 */
1821 	while (p >= ip_srcrt.route) {
1822 #ifdef DIAGNOSTIC
1823 		if (ipprintfs)
1824 			printf(" %x", ntohl(q->s_addr));
1825 #endif
1826 		*q++ = *p--;
1827 	}
1828 	/*
1829 	 * Last hop goes to final destination.
1830 	 */
1831 	*q = ip_srcrt.dst;
1832 #ifdef DIAGNOSTIC
1833 	if (ipprintfs)
1834 		printf(" %x\n", ntohl(q->s_addr));
1835 #endif
1836 	return (m);
1837 }
1838 
1839 const int inetctlerrmap[PRC_NCMDS] = {
1840 	[PRC_MSGSIZE] = EMSGSIZE,
1841 	[PRC_HOSTDEAD] = EHOSTDOWN,
1842 	[PRC_HOSTUNREACH] = EHOSTUNREACH,
1843 	[PRC_UNREACH_NET] = EHOSTUNREACH,
1844 	[PRC_UNREACH_HOST] = EHOSTUNREACH,
1845 	[PRC_UNREACH_PROTOCOL] = ECONNREFUSED,
1846 	[PRC_UNREACH_PORT] = ECONNREFUSED,
1847 	[PRC_UNREACH_SRCFAIL] = EHOSTUNREACH,
1848 	[PRC_PARAMPROB] = ENOPROTOOPT,
1849 };
1850 
1851 /*
1852  * Forward a packet.  If some error occurs return the sender
1853  * an icmp packet.  Note we can't always generate a meaningful
1854  * icmp message because icmp doesn't have a large enough repertoire
1855  * of codes and types.
1856  *
1857  * If not forwarding, just drop the packet.  This could be confusing
1858  * if ipforwarding was zero but some routing protocol was advancing
1859  * us as a gateway to somewhere.  However, we must let the routing
1860  * protocol deal with that.
1861  *
1862  * The srcrt parameter indicates whether the packet is being forwarded
1863  * via a source route.
1864  */
1865 void
1866 ip_forward(struct mbuf *m, int srcrt)
1867 {
1868 	struct ip *ip = mtod(m, struct ip *);
1869 	struct rtentry *rt;
1870 	int error, type = 0, code = 0, destmtu = 0;
1871 	struct mbuf *mcopy;
1872 	n_long dest;
1873 	union {
1874 		struct sockaddr		dst;
1875 		struct sockaddr_in	dst4;
1876 	} u;
1877 
1878 	/*
1879 	 * We are now in the output path.
1880 	 */
1881 	MCLAIM(m, &ip_tx_mowner);
1882 
1883 	/*
1884 	 * Clear any in-bound checksum flags for this packet.
1885 	 */
1886 	m->m_pkthdr.csum_flags = 0;
1887 
1888 	dest = 0;
1889 #ifdef DIAGNOSTIC
1890 	if (ipprintfs) {
1891 		printf("forward: src %s ", inet_ntoa(ip->ip_src));
1892 		printf("dst %s ttl %x\n", inet_ntoa(ip->ip_dst), ip->ip_ttl);
1893 	}
1894 #endif
1895 	if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
1896 		IP_STATINC(IP_STAT_CANTFORWARD);
1897 		m_freem(m);
1898 		return;
1899 	}
1900 	if (ip->ip_ttl <= IPTTLDEC) {
1901 		icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0);
1902 		return;
1903 	}
1904 
1905 	sockaddr_in_init(&u.dst4, &ip->ip_dst, 0);
1906 	if ((rt = rtcache_lookup(&ipforward_rt, &u.dst)) == NULL) {
1907 		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_NET, dest, 0);
1908 		return;
1909 	}
1910 
1911 	/*
1912 	 * Save at most 68 bytes of the packet in case
1913 	 * we need to generate an ICMP message to the src.
1914 	 * Pullup to avoid sharing mbuf cluster between m and mcopy.
1915 	 */
1916 	mcopy = m_copym(m, 0, imin(ntohs(ip->ip_len), 68), M_DONTWAIT);
1917 	if (mcopy)
1918 		mcopy = m_pullup(mcopy, ip->ip_hl << 2);
1919 
1920 	ip->ip_ttl -= IPTTLDEC;
1921 
1922 	/*
1923 	 * If forwarding packet using same interface that it came in on,
1924 	 * perhaps should send a redirect to sender to shortcut a hop.
1925 	 * Only send redirect if source is sending directly to us,
1926 	 * and if packet was not source routed (or has any options).
1927 	 * Also, don't send redirect if forwarding using a default route
1928 	 * or a route modified by a redirect.
1929 	 */
1930 	if (rt->rt_ifp == m->m_pkthdr.rcvif &&
1931 	    (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1932 	    !in_nullhost(satocsin(rt_getkey(rt))->sin_addr) &&
1933 	    ipsendredirects && !srcrt) {
1934 		if (rt->rt_ifa &&
1935 		    (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) ==
1936 		    ifatoia(rt->rt_ifa)->ia_subnet) {
1937 			if (rt->rt_flags & RTF_GATEWAY)
1938 				dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
1939 			else
1940 				dest = ip->ip_dst.s_addr;
1941 			/*
1942 			 * Router requirements says to only send host
1943 			 * redirects.
1944 			 */
1945 			type = ICMP_REDIRECT;
1946 			code = ICMP_REDIRECT_HOST;
1947 #ifdef DIAGNOSTIC
1948 			if (ipprintfs)
1949 				printf("redirect (%d) to %x\n", code,
1950 				    (u_int32_t)dest);
1951 #endif
1952 		}
1953 	}
1954 
1955 	error = ip_output(m, NULL, &ipforward_rt,
1956 	    (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)),
1957 	    (struct ip_moptions *)NULL, (struct socket *)NULL);
1958 
1959 	if (error)
1960 		IP_STATINC(IP_STAT_CANTFORWARD);
1961 	else {
1962 		uint64_t *ips = IP_STAT_GETREF();
1963 		ips[IP_STAT_FORWARD]++;
1964 		if (type) {
1965 			ips[IP_STAT_REDIRECTSENT]++;
1966 			IP_STAT_PUTREF();
1967 		} else {
1968 			IP_STAT_PUTREF();
1969 			if (mcopy) {
1970 #ifdef GATEWAY
1971 				if (mcopy->m_flags & M_CANFASTFWD)
1972 					ipflow_create(&ipforward_rt, mcopy);
1973 #endif
1974 				m_freem(mcopy);
1975 			}
1976 			return;
1977 		}
1978 	}
1979 	if (mcopy == NULL)
1980 		return;
1981 
1982 	switch (error) {
1983 
1984 	case 0:				/* forwarded, but need redirect */
1985 		/* type, code set above */
1986 		break;
1987 
1988 	case ENETUNREACH:		/* shouldn't happen, checked above */
1989 	case EHOSTUNREACH:
1990 	case ENETDOWN:
1991 	case EHOSTDOWN:
1992 	default:
1993 		type = ICMP_UNREACH;
1994 		code = ICMP_UNREACH_HOST;
1995 		break;
1996 
1997 	case EMSGSIZE:
1998 		type = ICMP_UNREACH;
1999 		code = ICMP_UNREACH_NEEDFRAG;
2000 
2001 		if ((rt = rtcache_validate(&ipforward_rt)) != NULL)
2002 			destmtu = rt->rt_ifp->if_mtu;
2003 
2004 #if defined(IPSEC) || defined(FAST_IPSEC)
2005 		{
2006 			/*
2007 			 * If the packet is routed over IPsec tunnel, tell the
2008 			 * originator the tunnel MTU.
2009 			 *	tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
2010 			 * XXX quickhack!!!
2011 			 */
2012 
2013 			struct secpolicy *sp;
2014 			int ipsecerror;
2015 			size_t ipsechdr;
2016 			struct route *ro;
2017 
2018 			sp = ipsec4_getpolicybyaddr(mcopy,
2019 			    IPSEC_DIR_OUTBOUND, IP_FORWARDING,
2020 			    &ipsecerror);
2021 
2022 			if (sp != NULL) {
2023 				/* count IPsec header size */
2024 				ipsechdr = ipsec4_hdrsiz(mcopy,
2025 				    IPSEC_DIR_OUTBOUND, NULL);
2026 
2027 				/*
2028 				 * find the correct route for outer IPv4
2029 				 * header, compute tunnel MTU.
2030 				 */
2031 
2032 				if (sp->req != NULL
2033 				 && sp->req->sav != NULL
2034 				 && sp->req->sav->sah != NULL) {
2035 					ro = &sp->req->sav->sah->sa_route;
2036 					rt = rtcache_validate(ro);
2037 					if (rt && rt->rt_ifp) {
2038 						destmtu =
2039 						    rt->rt_rmx.rmx_mtu ?
2040 						    rt->rt_rmx.rmx_mtu :
2041 						    rt->rt_ifp->if_mtu;
2042 						destmtu -= ipsechdr;
2043 					}
2044 				}
2045 
2046 #ifdef	IPSEC
2047 				key_freesp(sp);
2048 #else
2049 				KEY_FREESP(&sp);
2050 #endif
2051 			}
2052 		}
2053 #endif /*defined(IPSEC) || defined(FAST_IPSEC)*/
2054 		IP_STATINC(IP_STAT_CANTFRAG);
2055 		break;
2056 
2057 	case ENOBUFS:
2058 #if 1
2059 		/*
2060 		 * a router should not generate ICMP_SOURCEQUENCH as
2061 		 * required in RFC1812 Requirements for IP Version 4 Routers.
2062 		 * source quench could be a big problem under DoS attacks,
2063 		 * or if the underlying interface is rate-limited.
2064 		 */
2065 		if (mcopy)
2066 			m_freem(mcopy);
2067 		return;
2068 #else
2069 		type = ICMP_SOURCEQUENCH;
2070 		code = 0;
2071 		break;
2072 #endif
2073 	}
2074 	icmp_error(mcopy, type, code, dest, destmtu);
2075 }
2076 
2077 void
2078 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
2079     struct mbuf *m)
2080 {
2081 
2082 	if (inp->inp_socket->so_options & SO_TIMESTAMP
2083 #ifdef SO_OTIMESTAMP
2084 	    || inp->inp_socket->so_options & SO_OTIMESTAMP
2085 #endif
2086 	    ) {
2087 		struct timeval tv;
2088 
2089 		microtime(&tv);
2090 #ifdef SO_OTIMESTAMP
2091 		if (inp->inp_socket->so_options & SO_OTIMESTAMP) {
2092 			struct timeval50 tv50;
2093 			timeval_to_timeval50(&tv, &tv50);
2094 			*mp = sbcreatecontrol((void *) &tv50, sizeof(tv50),
2095 			    SCM_OTIMESTAMP, SOL_SOCKET);
2096 		} else
2097 #endif
2098 		*mp = sbcreatecontrol((void *) &tv, sizeof(tv),
2099 		    SCM_TIMESTAMP, SOL_SOCKET);
2100 		if (*mp)
2101 			mp = &(*mp)->m_next;
2102 	}
2103 	if (inp->inp_flags & INP_RECVDSTADDR) {
2104 		*mp = sbcreatecontrol((void *) &ip->ip_dst,
2105 		    sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
2106 		if (*mp)
2107 			mp = &(*mp)->m_next;
2108 	}
2109 #ifdef notyet
2110 	/*
2111 	 * XXX
2112 	 * Moving these out of udp_input() made them even more broken
2113 	 * than they already were.
2114 	 *	- fenner@parc.xerox.com
2115 	 */
2116 	/* options were tossed already */
2117 	if (inp->inp_flags & INP_RECVOPTS) {
2118 		*mp = sbcreatecontrol((void *) opts_deleted_above,
2119 		    sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
2120 		if (*mp)
2121 			mp = &(*mp)->m_next;
2122 	}
2123 	/* ip_srcroute doesn't do what we want here, need to fix */
2124 	if (inp->inp_flags & INP_RECVRETOPTS) {
2125 		*mp = sbcreatecontrol((void *) ip_srcroute(),
2126 		    sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
2127 		if (*mp)
2128 			mp = &(*mp)->m_next;
2129 	}
2130 #endif
2131 	if (inp->inp_flags & INP_RECVIF) {
2132 		struct sockaddr_dl sdl;
2133 
2134 		sockaddr_dl_init(&sdl, sizeof(sdl),
2135 		    (m->m_pkthdr.rcvif != NULL)
2136 		        ?  m->m_pkthdr.rcvif->if_index
2137 			: 0,
2138 			0, NULL, 0, NULL, 0);
2139 		*mp = sbcreatecontrol(&sdl, sdl.sdl_len, IP_RECVIF, IPPROTO_IP);
2140 		if (*mp)
2141 			mp = &(*mp)->m_next;
2142 	}
2143 	if (inp->inp_flags & INP_RECVTTL) {
2144 		*mp = sbcreatecontrol((void *) &ip->ip_ttl,
2145 		    sizeof(uint8_t), IP_RECVTTL, IPPROTO_IP);
2146 		if (*mp)
2147 			mp = &(*mp)->m_next;
2148 	}
2149 }
2150 
2151 /*
2152  * sysctl helper routine for net.inet.ip.forwsrcrt.
2153  */
2154 static int
2155 sysctl_net_inet_ip_forwsrcrt(SYSCTLFN_ARGS)
2156 {
2157 	int error, tmp;
2158 	struct sysctlnode node;
2159 
2160 	node = *rnode;
2161 	tmp = ip_forwsrcrt;
2162 	node.sysctl_data = &tmp;
2163 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
2164 	if (error || newp == NULL)
2165 		return (error);
2166 
2167 	error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_FORWSRCRT,
2168 	    0, NULL, NULL, NULL);
2169 	if (error)
2170 		return (error);
2171 
2172 	ip_forwsrcrt = tmp;
2173 
2174 	return (0);
2175 }
2176 
2177 /*
2178  * sysctl helper routine for net.inet.ip.mtudisctimeout.  checks the
2179  * range of the new value and tweaks timers if it changes.
2180  */
2181 static int
2182 sysctl_net_inet_ip_pmtudto(SYSCTLFN_ARGS)
2183 {
2184 	int error, tmp;
2185 	struct sysctlnode node;
2186 
2187 	node = *rnode;
2188 	tmp = ip_mtudisc_timeout;
2189 	node.sysctl_data = &tmp;
2190 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
2191 	if (error || newp == NULL)
2192 		return (error);
2193 	if (tmp < 0)
2194 		return (EINVAL);
2195 
2196 	mutex_enter(softnet_lock);
2197 
2198 	ip_mtudisc_timeout = tmp;
2199 	rt_timer_queue_change(ip_mtudisc_timeout_q, ip_mtudisc_timeout);
2200 
2201 	mutex_exit(softnet_lock);
2202 
2203 	return (0);
2204 }
2205 
2206 #ifdef GATEWAY
2207 /*
2208  * sysctl helper routine for net.inet.ip.maxflows.
2209  */
2210 static int
2211 sysctl_net_inet_ip_maxflows(SYSCTLFN_ARGS)
2212 {
2213 	int error;
2214 
2215 	error = sysctl_lookup(SYSCTLFN_CALL(rnode));
2216 	if (error || newp == NULL)
2217 		return (error);
2218 
2219 	mutex_enter(softnet_lock);
2220 	KERNEL_LOCK(1, NULL);
2221 
2222 	ipflow_prune();
2223 
2224 	KERNEL_UNLOCK_ONE(NULL);
2225 	mutex_exit(softnet_lock);
2226 
2227 	return (0);
2228 }
2229 
2230 static int
2231 sysctl_net_inet_ip_hashsize(SYSCTLFN_ARGS)
2232 {
2233 	int error, tmp;
2234 	struct sysctlnode node;
2235 
2236 	node = *rnode;
2237 	tmp = ip_hashsize;
2238 	node.sysctl_data = &tmp;
2239 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
2240 	if (error || newp == NULL)
2241 		return (error);
2242 
2243 	if ((tmp & (tmp - 1)) == 0 && tmp != 0) {
2244 		/*
2245 		 * Can only fail due to malloc()
2246 		 */
2247 		mutex_enter(softnet_lock);
2248 		KERNEL_LOCK(1, NULL);
2249 
2250 		error = ipflow_invalidate_all(tmp);
2251 
2252 		KERNEL_UNLOCK_ONE(NULL);
2253 		mutex_exit(softnet_lock);
2254 
2255 	} else {
2256 		/*
2257 		 * EINVAL if not a power of 2
2258 	         */
2259 		error = EINVAL;
2260 	}
2261 
2262 	return error;
2263 }
2264 #endif /* GATEWAY */
2265 
2266 static int
2267 sysctl_net_inet_ip_stats(SYSCTLFN_ARGS)
2268 {
2269 
2270 	return (NETSTAT_SYSCTL(ipstat_percpu, IP_NSTATS));
2271 }
2272 
2273 static void
2274 sysctl_net_inet_ip_setup(struct sysctllog **clog)
2275 {
2276 	extern int subnetsarelocal, hostzeroisbroadcast;
2277 
2278 	sysctl_createv(clog, 0, NULL, NULL,
2279 		       CTLFLAG_PERMANENT,
2280 		       CTLTYPE_NODE, "net", NULL,
2281 		       NULL, 0, NULL, 0,
2282 		       CTL_NET, CTL_EOL);
2283 	sysctl_createv(clog, 0, NULL, NULL,
2284 		       CTLFLAG_PERMANENT,
2285 		       CTLTYPE_NODE, "inet",
2286 		       SYSCTL_DESCR("PF_INET related settings"),
2287 		       NULL, 0, NULL, 0,
2288 		       CTL_NET, PF_INET, CTL_EOL);
2289 	sysctl_createv(clog, 0, NULL, NULL,
2290 		       CTLFLAG_PERMANENT,
2291 		       CTLTYPE_NODE, "ip",
2292 		       SYSCTL_DESCR("IPv4 related settings"),
2293 		       NULL, 0, NULL, 0,
2294 		       CTL_NET, PF_INET, IPPROTO_IP, CTL_EOL);
2295 
2296 	sysctl_createv(clog, 0, NULL, NULL,
2297 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2298 		       CTLTYPE_INT, "forwarding",
2299 		       SYSCTL_DESCR("Enable forwarding of INET datagrams"),
2300 		       NULL, 0, &ipforwarding, 0,
2301 		       CTL_NET, PF_INET, IPPROTO_IP,
2302 		       IPCTL_FORWARDING, CTL_EOL);
2303 	sysctl_createv(clog, 0, NULL, NULL,
2304 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2305 		       CTLTYPE_INT, "redirect",
2306 		       SYSCTL_DESCR("Enable sending of ICMP redirect messages"),
2307 		       NULL, 0, &ipsendredirects, 0,
2308 		       CTL_NET, PF_INET, IPPROTO_IP,
2309 		       IPCTL_SENDREDIRECTS, CTL_EOL);
2310 	sysctl_createv(clog, 0, NULL, NULL,
2311 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2312 		       CTLTYPE_INT, "ttl",
2313 		       SYSCTL_DESCR("Default TTL for an INET datagram"),
2314 		       NULL, 0, &ip_defttl, 0,
2315 		       CTL_NET, PF_INET, IPPROTO_IP,
2316 		       IPCTL_DEFTTL, CTL_EOL);
2317 #ifdef IPCTL_DEFMTU
2318 	sysctl_createv(clog, 0, NULL, NULL,
2319 		       CTLFLAG_PERMANENT /* |CTLFLAG_READWRITE? */,
2320 		       CTLTYPE_INT, "mtu",
2321 		       SYSCTL_DESCR("Default MTA for an INET route"),
2322 		       NULL, 0, &ip_mtu, 0,
2323 		       CTL_NET, PF_INET, IPPROTO_IP,
2324 		       IPCTL_DEFMTU, CTL_EOL);
2325 #endif /* IPCTL_DEFMTU */
2326 	sysctl_createv(clog, 0, NULL, NULL,
2327 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2328 		       CTLTYPE_INT, "forwsrcrt",
2329 		       SYSCTL_DESCR("Enable forwarding of source-routed "
2330 				    "datagrams"),
2331 		       sysctl_net_inet_ip_forwsrcrt, 0, &ip_forwsrcrt, 0,
2332 		       CTL_NET, PF_INET, IPPROTO_IP,
2333 		       IPCTL_FORWSRCRT, CTL_EOL);
2334 	sysctl_createv(clog, 0, NULL, NULL,
2335 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2336 		       CTLTYPE_INT, "directed-broadcast",
2337 		       SYSCTL_DESCR("Enable forwarding of broadcast datagrams"),
2338 		       NULL, 0, &ip_directedbcast, 0,
2339 		       CTL_NET, PF_INET, IPPROTO_IP,
2340 		       IPCTL_DIRECTEDBCAST, CTL_EOL);
2341 	sysctl_createv(clog, 0, NULL, NULL,
2342 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2343 		       CTLTYPE_INT, "allowsrcrt",
2344 		       SYSCTL_DESCR("Accept source-routed datagrams"),
2345 		       NULL, 0, &ip_allowsrcrt, 0,
2346 		       CTL_NET, PF_INET, IPPROTO_IP,
2347 		       IPCTL_ALLOWSRCRT, CTL_EOL);
2348 	sysctl_createv(clog, 0, NULL, NULL,
2349 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2350 		       CTLTYPE_INT, "subnetsarelocal",
2351 		       SYSCTL_DESCR("Whether logical subnets are considered "
2352 				    "local"),
2353 		       NULL, 0, &subnetsarelocal, 0,
2354 		       CTL_NET, PF_INET, IPPROTO_IP,
2355 		       IPCTL_SUBNETSARELOCAL, CTL_EOL);
2356 	sysctl_createv(clog, 0, NULL, NULL,
2357 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2358 		       CTLTYPE_INT, "mtudisc",
2359 		       SYSCTL_DESCR("Use RFC1191 Path MTU Discovery"),
2360 		       NULL, 0, &ip_mtudisc, 0,
2361 		       CTL_NET, PF_INET, IPPROTO_IP,
2362 		       IPCTL_MTUDISC, CTL_EOL);
2363 	sysctl_createv(clog, 0, NULL, NULL,
2364 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2365 		       CTLTYPE_INT, "anonportmin",
2366 		       SYSCTL_DESCR("Lowest ephemeral port number to assign"),
2367 		       sysctl_net_inet_ip_ports, 0, &anonportmin, 0,
2368 		       CTL_NET, PF_INET, IPPROTO_IP,
2369 		       IPCTL_ANONPORTMIN, CTL_EOL);
2370 	sysctl_createv(clog, 0, NULL, NULL,
2371 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2372 		       CTLTYPE_INT, "anonportmax",
2373 		       SYSCTL_DESCR("Highest ephemeral port number to assign"),
2374 		       sysctl_net_inet_ip_ports, 0, &anonportmax, 0,
2375 		       CTL_NET, PF_INET, IPPROTO_IP,
2376 		       IPCTL_ANONPORTMAX, CTL_EOL);
2377 	sysctl_createv(clog, 0, NULL, NULL,
2378 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2379 		       CTLTYPE_INT, "mtudisctimeout",
2380 		       SYSCTL_DESCR("Lifetime of a Path MTU Discovered route"),
2381 		       sysctl_net_inet_ip_pmtudto, 0, &ip_mtudisc_timeout, 0,
2382 		       CTL_NET, PF_INET, IPPROTO_IP,
2383 		       IPCTL_MTUDISCTIMEOUT, CTL_EOL);
2384 #ifdef GATEWAY
2385 	sysctl_createv(clog, 0, NULL, NULL,
2386 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2387 		       CTLTYPE_INT, "maxflows",
2388 		       SYSCTL_DESCR("Number of flows for fast forwarding"),
2389 		       sysctl_net_inet_ip_maxflows, 0, &ip_maxflows, 0,
2390 		       CTL_NET, PF_INET, IPPROTO_IP,
2391 		       IPCTL_MAXFLOWS, CTL_EOL);
2392 	sysctl_createv(clog, 0, NULL, NULL,
2393 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2394 			CTLTYPE_INT, "hashsize",
2395 			SYSCTL_DESCR("Size of hash table for fast forwarding (IPv4)"),
2396 			sysctl_net_inet_ip_hashsize, 0, &ip_hashsize, 0,
2397 			CTL_NET, PF_INET, IPPROTO_IP,
2398 			CTL_CREATE, CTL_EOL);
2399 #endif /* GATEWAY */
2400 	sysctl_createv(clog, 0, NULL, NULL,
2401 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2402 		       CTLTYPE_INT, "hostzerobroadcast",
2403 		       SYSCTL_DESCR("All zeroes address is broadcast address"),
2404 		       NULL, 0, &hostzeroisbroadcast, 0,
2405 		       CTL_NET, PF_INET, IPPROTO_IP,
2406 		       IPCTL_HOSTZEROBROADCAST, CTL_EOL);
2407 #if NGIF > 0
2408 	sysctl_createv(clog, 0, NULL, NULL,
2409 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2410 		       CTLTYPE_INT, "gifttl",
2411 		       SYSCTL_DESCR("Default TTL for a gif tunnel datagram"),
2412 		       NULL, 0, &ip_gif_ttl, 0,
2413 		       CTL_NET, PF_INET, IPPROTO_IP,
2414 		       IPCTL_GIF_TTL, CTL_EOL);
2415 #endif /* NGIF */
2416 #ifndef IPNOPRIVPORTS
2417 	sysctl_createv(clog, 0, NULL, NULL,
2418 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2419 		       CTLTYPE_INT, "lowportmin",
2420 		       SYSCTL_DESCR("Lowest privileged ephemeral port number "
2421 				    "to assign"),
2422 		       sysctl_net_inet_ip_ports, 0, &lowportmin, 0,
2423 		       CTL_NET, PF_INET, IPPROTO_IP,
2424 		       IPCTL_LOWPORTMIN, CTL_EOL);
2425 	sysctl_createv(clog, 0, NULL, NULL,
2426 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2427 		       CTLTYPE_INT, "lowportmax",
2428 		       SYSCTL_DESCR("Highest privileged ephemeral port number "
2429 				    "to assign"),
2430 		       sysctl_net_inet_ip_ports, 0, &lowportmax, 0,
2431 		       CTL_NET, PF_INET, IPPROTO_IP,
2432 		       IPCTL_LOWPORTMAX, CTL_EOL);
2433 #endif /* IPNOPRIVPORTS */
2434 	sysctl_createv(clog, 0, NULL, NULL,
2435 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2436 		       CTLTYPE_INT, "maxfragpackets",
2437 		       SYSCTL_DESCR("Maximum number of fragments to retain for "
2438 				    "possible reassembly"),
2439 		       NULL, 0, &ip_maxfragpackets, 0,
2440 		       CTL_NET, PF_INET, IPPROTO_IP,
2441 		       IPCTL_MAXFRAGPACKETS, CTL_EOL);
2442 #if NGRE > 0
2443 	sysctl_createv(clog, 0, NULL, NULL,
2444 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2445 		       CTLTYPE_INT, "grettl",
2446 		       SYSCTL_DESCR("Default TTL for a gre tunnel datagram"),
2447 		       NULL, 0, &ip_gre_ttl, 0,
2448 		       CTL_NET, PF_INET, IPPROTO_IP,
2449 		       IPCTL_GRE_TTL, CTL_EOL);
2450 #endif /* NGRE */
2451 	sysctl_createv(clog, 0, NULL, NULL,
2452 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2453 		       CTLTYPE_INT, "checkinterface",
2454 		       SYSCTL_DESCR("Enable receive side of Strong ES model "
2455 				    "from RFC1122"),
2456 		       NULL, 0, &ip_checkinterface, 0,
2457 		       CTL_NET, PF_INET, IPPROTO_IP,
2458 		       IPCTL_CHECKINTERFACE, CTL_EOL);
2459 	sysctl_createv(clog, 0, NULL, NULL,
2460 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2461 		       CTLTYPE_INT, "random_id",
2462 		       SYSCTL_DESCR("Assign random ip_id values"),
2463 		       NULL, 0, &ip_do_randomid, 0,
2464 		       CTL_NET, PF_INET, IPPROTO_IP,
2465 		       IPCTL_RANDOMID, CTL_EOL);
2466 	sysctl_createv(clog, 0, NULL, NULL,
2467 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2468 		       CTLTYPE_INT, "do_loopback_cksum",
2469 		       SYSCTL_DESCR("Perform IP checksum on loopback"),
2470 		       NULL, 0, &ip_do_loopback_cksum, 0,
2471 		       CTL_NET, PF_INET, IPPROTO_IP,
2472 		       IPCTL_LOOPBACKCKSUM, CTL_EOL);
2473 	sysctl_createv(clog, 0, NULL, NULL,
2474 		       CTLFLAG_PERMANENT,
2475 		       CTLTYPE_STRUCT, "stats",
2476 		       SYSCTL_DESCR("IP statistics"),
2477 		       sysctl_net_inet_ip_stats, 0, NULL, 0,
2478 		       CTL_NET, PF_INET, IPPROTO_IP, IPCTL_STATS,
2479 		       CTL_EOL);
2480 }
2481 
2482 void
2483 ip_statinc(u_int stat)
2484 {
2485 
2486 	KASSERT(stat < IP_NSTATS);
2487 	IP_STATINC(stat);
2488 }
2489