xref: /dflybsd-src/sys/netinet/ip_input.c (revision 5f39c7e70ca0960d1868c75a449064df712dbb10)
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. Neither the name of the University nor the names of its contributors
47  *    may be used to endorse or promote products derived from this software
48  *    without specific prior written permission.
49  *
50  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
51  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
52  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
53  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
54  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
55  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
56  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
57  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
58  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
59  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
60  * SUCH DAMAGE.
61  *
62  *	@(#)ip_input.c	8.2 (Berkeley) 1/4/94
63  * $FreeBSD: src/sys/netinet/ip_input.c,v 1.130.2.52 2003/03/07 07:01:28 silby Exp $
64  */
65 
66 #define	_IP_VHL
67 
68 #include "opt_bootp.h"
69 #include "opt_ipdn.h"
70 #include "opt_ipdivert.h"
71 #include "opt_ipstealth.h"
72 #include "opt_ipsec.h"
73 #include "opt_rss.h"
74 
75 #include <sys/param.h>
76 #include <sys/systm.h>
77 #include <sys/mbuf.h>
78 #include <sys/malloc.h>
79 #include <sys/mpipe.h>
80 #include <sys/domain.h>
81 #include <sys/protosw.h>
82 #include <sys/socket.h>
83 #include <sys/time.h>
84 #include <sys/globaldata.h>
85 #include <sys/thread.h>
86 #include <sys/kernel.h>
87 #include <sys/syslog.h>
88 #include <sys/sysctl.h>
89 #include <sys/in_cksum.h>
90 #include <sys/lock.h>
91 
92 #include <sys/mplock2.h>
93 
94 #include <machine/stdarg.h>
95 
96 #include <net/if.h>
97 #include <net/if_types.h>
98 #include <net/if_var.h>
99 #include <net/if_dl.h>
100 #include <net/pfil.h>
101 #include <net/route.h>
102 #include <net/netisr2.h>
103 
104 #include <netinet/in.h>
105 #include <netinet/in_systm.h>
106 #include <netinet/in_var.h>
107 #include <netinet/ip.h>
108 #include <netinet/in_pcb.h>
109 #include <netinet/ip_var.h>
110 #include <netinet/ip_icmp.h>
111 #include <netinet/ip_divert.h>
112 #include <netinet/ip_flow.h>
113 
114 #include <sys/thread2.h>
115 #include <sys/msgport2.h>
116 #include <net/netmsg2.h>
117 
118 #include <sys/socketvar.h>
119 
120 #include <net/ipfw/ip_fw.h>
121 #include <net/dummynet/ip_dummynet.h>
122 
123 #ifdef IPSEC
124 #include <netinet6/ipsec.h>
125 #include <netproto/key/key.h>
126 #endif
127 
128 #ifdef FAST_IPSEC
129 #include <netproto/ipsec/ipsec.h>
130 #include <netproto/ipsec/key.h>
131 #endif
132 
133 int rsvp_on = 0;
134 static int ip_rsvp_on;
135 struct socket *ip_rsvpd;
136 
137 int ipforwarding = 0;
138 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW,
139     &ipforwarding, 0, "Enable IP forwarding between interfaces");
140 
141 static int ipsendredirects = 1; /* XXX */
142 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW,
143     &ipsendredirects, 0, "Enable sending IP redirects");
144 
145 int ip_defttl = IPDEFTTL;
146 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW,
147     &ip_defttl, 0, "Maximum TTL on IP packets");
148 
149 static int ip_dosourceroute = 0;
150 SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW,
151     &ip_dosourceroute, 0, "Enable forwarding source routed IP packets");
152 
153 static int ip_acceptsourceroute = 0;
154 SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute,
155     CTLFLAG_RW, &ip_acceptsourceroute, 0,
156     "Enable accepting source routed IP packets");
157 
158 static int ip_keepfaith = 0;
159 SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW,
160     &ip_keepfaith, 0,
161     "Enable packet capture for FAITH IPv4->IPv6 translator daemon");
162 
163 static int nipq = 0;	/* total # of reass queues */
164 static int maxnipq;
165 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragpackets, CTLFLAG_RW,
166     &maxnipq, 0,
167     "Maximum number of IPv4 fragment reassembly queue entries");
168 
169 static int maxfragsperpacket;
170 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW,
171     &maxfragsperpacket, 0,
172     "Maximum number of IPv4 fragments allowed per packet");
173 
174 static int ip_sendsourcequench = 0;
175 SYSCTL_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW,
176     &ip_sendsourcequench, 0,
177     "Enable the transmission of source quench packets");
178 
179 int ip_do_randomid = 1;
180 SYSCTL_INT(_net_inet_ip, OID_AUTO, random_id, CTLFLAG_RW,
181     &ip_do_randomid, 0,
182     "Assign random ip_id values");
183 /*
184  * XXX - Setting ip_checkinterface mostly implements the receive side of
185  * the Strong ES model described in RFC 1122, but since the routing table
186  * and transmit implementation do not implement the Strong ES model,
187  * setting this to 1 results in an odd hybrid.
188  *
189  * XXX - ip_checkinterface currently must be disabled if you use ipnat
190  * to translate the destination address to another local interface.
191  *
192  * XXX - ip_checkinterface must be disabled if you add IP aliases
193  * to the loopback interface instead of the interface where the
194  * packets for those addresses are received.
195  */
196 static int ip_checkinterface = 0;
197 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW,
198     &ip_checkinterface, 0, "Verify packet arrives on correct interface");
199 
200 static u_long ip_hash_count = 0;
201 SYSCTL_ULONG(_net_inet_ip, OID_AUTO, hash_count, CTLFLAG_RD,
202     &ip_hash_count, 0, "Number of packets hashed by IP");
203 
204 #ifdef RSS_DEBUG
205 static u_long ip_rehash_count = 0;
206 SYSCTL_ULONG(_net_inet_ip, OID_AUTO, rehash_count, CTLFLAG_RD,
207     &ip_rehash_count, 0, "Number of packets rehashed by IP");
208 
209 static u_long ip_dispatch_fast = 0;
210 SYSCTL_ULONG(_net_inet_ip, OID_AUTO, dispatch_fast_count, CTLFLAG_RD,
211     &ip_dispatch_fast, 0, "Number of packets handled on current CPU");
212 
213 static u_long ip_dispatch_slow = 0;
214 SYSCTL_ULONG(_net_inet_ip, OID_AUTO, dispatch_slow_count, CTLFLAG_RD,
215     &ip_dispatch_slow, 0, "Number of packets messaged to another CPU");
216 #endif
217 
218 static struct lwkt_token ipq_token = LWKT_TOKEN_INITIALIZER(ipq_token);
219 
220 #ifdef DIAGNOSTIC
221 static int ipprintfs = 0;
222 #endif
223 
224 extern	struct domain inetdomain;
225 extern	struct protosw inetsw[];
226 u_char	ip_protox[IPPROTO_MAX];
227 struct	in_ifaddrhead in_ifaddrheads[MAXCPU];	/* first inet address */
228 struct	in_ifaddrhashhead *in_ifaddrhashtbls[MAXCPU];
229 						/* inet addr hash table */
230 u_long	in_ifaddrhmask;				/* mask for hash table */
231 
232 static struct mbuf *ipforward_mtemp[MAXCPU];
233 
234 struct ip_stats ipstats_percpu[MAXCPU] __cachealign;
235 
236 static int
237 sysctl_ipstats(SYSCTL_HANDLER_ARGS)
238 {
239 	int cpu, error = 0;
240 
241 	for (cpu = 0; cpu < ncpus; ++cpu) {
242 		if ((error = SYSCTL_OUT(req, &ipstats_percpu[cpu],
243 					sizeof(struct ip_stats))))
244 			break;
245 		if ((error = SYSCTL_IN(req, &ipstats_percpu[cpu],
246 				       sizeof(struct ip_stats))))
247 			break;
248 	}
249 
250 	return (error);
251 }
252 SYSCTL_PROC(_net_inet_ip, IPCTL_STATS, stats, (CTLTYPE_OPAQUE | CTLFLAG_RW),
253     0, 0, sysctl_ipstats, "S,ip_stats", "IP statistics");
254 
255 /* Packet reassembly stuff */
256 #define	IPREASS_NHASH_LOG2	6
257 #define	IPREASS_NHASH		(1 << IPREASS_NHASH_LOG2)
258 #define	IPREASS_HMASK		(IPREASS_NHASH - 1)
259 #define	IPREASS_HASH(x,y)						\
260     (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK)
261 
262 static TAILQ_HEAD(ipqhead, ipq) ipq[IPREASS_NHASH];
263 
264 #ifdef IPCTL_DEFMTU
265 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
266     &ip_mtu, 0, "Default MTU");
267 #endif
268 
269 #ifdef IPSTEALTH
270 static int ipstealth = 0;
271 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW, &ipstealth, 0, "");
272 #else
273 static const int ipstealth = 0;
274 #endif
275 
276 struct mbuf *(*ip_divert_p)(struct mbuf *, int, int);
277 
278 struct pfil_head inet_pfil_hook;
279 
280 /*
281  * struct ip_srcrt_opt is used to store packet state while it travels
282  * through the stack.
283  *
284  * XXX Note that the code even makes assumptions on the size and
285  * alignment of fields inside struct ip_srcrt so e.g. adding some
286  * fields will break the code.  This needs to be fixed.
287  *
288  * We need to save the IP options in case a protocol wants to respond
289  * to an incoming packet over the same route if the packet got here
290  * using IP source routing.  This allows connection establishment and
291  * maintenance when the remote end is on a network that is not known
292  * to us.
293  */
294 struct ip_srcrt {
295 	struct	in_addr dst;			/* final destination */
296 	char	nop;				/* one NOP to align */
297 	char	srcopt[IPOPT_OFFSET + 1];	/* OPTVAL, OLEN and OFFSET */
298 	struct	in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)];
299 };
300 
301 struct ip_srcrt_opt {
302 	int		ip_nhops;
303 	struct ip_srcrt	ip_srcrt;
304 };
305 
306 static MALLOC_DEFINE(M_IPQ, "ipq", "IP Fragment Management");
307 static struct malloc_pipe ipq_mpipe;
308 
309 static void		save_rte(struct mbuf *, u_char *, struct in_addr);
310 static int		ip_dooptions(struct mbuf *m, int, struct sockaddr_in *);
311 static void		ip_freef(struct ipqhead *, struct ipq *);
312 static void		ip_input_handler(netmsg_t);
313 
314 /*
315  * IP initialization: fill in IP protocol switch table.
316  * All protocols not implemented in kernel go to raw IP protocol handler.
317  */
318 void
319 ip_init(void)
320 {
321 	struct protosw *pr;
322 	int i;
323 	int cpu;
324 
325 	/*
326 	 * Make sure we can handle a reasonable number of fragments but
327 	 * cap it at 4000 (XXX).
328 	 */
329 	mpipe_init(&ipq_mpipe, M_IPQ, sizeof(struct ipq),
330 		    IFQ_MAXLEN, 4000, 0, NULL, NULL, NULL);
331 	for (i = 0; i < ncpus; ++i) {
332 		TAILQ_INIT(&in_ifaddrheads[i]);
333 		in_ifaddrhashtbls[i] =
334 			hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask);
335 	}
336 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
337 	if (pr == NULL)
338 		panic("ip_init");
339 	for (i = 0; i < IPPROTO_MAX; i++)
340 		ip_protox[i] = pr - inetsw;
341 	for (pr = inetdomain.dom_protosw;
342 	     pr < inetdomain.dom_protoswNPROTOSW; pr++) {
343 		if (pr->pr_domain->dom_family == PF_INET && pr->pr_protocol) {
344 			if (pr->pr_protocol != IPPROTO_RAW)
345 				ip_protox[pr->pr_protocol] = pr - inetsw;
346 		}
347 	}
348 
349 	inet_pfil_hook.ph_type = PFIL_TYPE_AF;
350 	inet_pfil_hook.ph_af = AF_INET;
351 	if ((i = pfil_head_register(&inet_pfil_hook)) != 0) {
352 		kprintf("%s: WARNING: unable to register pfil hook, "
353 			"error %d\n", __func__, i);
354 	}
355 
356 	for (i = 0; i < IPREASS_NHASH; i++)
357 		TAILQ_INIT(&ipq[i]);
358 
359 	maxnipq = nmbclusters / 32;
360 	maxfragsperpacket = 16;
361 
362 	ip_id = time_second & 0xffff;	/* time_second survives reboots */
363 
364 	for (cpu = 0; cpu < ncpus; ++cpu) {
365 		/*
366 		 * Initialize IP statistics counters for each CPU.
367 		 */
368 		bzero(&ipstats_percpu[cpu], sizeof(struct ip_stats));
369 
370 		/*
371 		 * Preallocate mbuf template for forwarding
372 		 */
373 		MGETHDR(ipforward_mtemp[cpu], MB_WAIT, MT_DATA);
374 	}
375 
376 	netisr_register(NETISR_IP, ip_input_handler, ip_hashfn_in);
377 	netisr_register_hashcheck(NETISR_IP, ip_hashcheck);
378 }
379 
380 /* Do transport protocol processing. */
381 static void
382 transport_processing_oncpu(struct mbuf *m, int hlen, struct ip *ip)
383 {
384 	const struct protosw *pr = &inetsw[ip_protox[ip->ip_p]];
385 
386 	/*
387 	 * Switch out to protocol's input routine.
388 	 */
389 	PR_GET_MPLOCK(pr);
390 	pr->pr_input(&m, &hlen, ip->ip_p);
391 	PR_REL_MPLOCK(pr);
392 }
393 
394 static void
395 transport_processing_handler(netmsg_t msg)
396 {
397 	struct netmsg_packet *pmsg = &msg->packet;
398 	struct ip *ip;
399 	int hlen;
400 
401 	ip = mtod(pmsg->nm_packet, struct ip *);
402 	hlen = pmsg->base.lmsg.u.ms_result;
403 
404 	transport_processing_oncpu(pmsg->nm_packet, hlen, ip);
405 	/* msg was embedded in the mbuf, do not reply! */
406 }
407 
408 static void
409 ip_input_handler(netmsg_t msg)
410 {
411 	ip_input(msg->packet.nm_packet);
412 	/* msg was embedded in the mbuf, do not reply! */
413 }
414 
415 /*
416  * IP input routine.  Checksum and byte swap header.  If fragmented
417  * try to reassemble.  Process options.  Pass to next level.
418  */
419 void
420 ip_input(struct mbuf *m)
421 {
422 	struct ip *ip;
423 	struct in_ifaddr *ia = NULL;
424 	struct in_ifaddr_container *iac;
425 	int hlen, checkif;
426 	u_short sum;
427 	struct in_addr pkt_dst;
428 	boolean_t using_srcrt = FALSE;		/* forward (by PFIL_HOOKS) */
429 	struct in_addr odst;			/* original dst address(NAT) */
430 	struct m_tag *mtag;
431 	struct sockaddr_in *next_hop = NULL;
432 	lwkt_port_t port;
433 #ifdef FAST_IPSEC
434 	struct tdb_ident *tdbi;
435 	struct secpolicy *sp;
436 	int error;
437 #endif
438 
439 	M_ASSERTPKTHDR(m);
440 
441 	/*
442 	 * This routine is called from numerous places which may not have
443 	 * characterized the packet.
444 	 */
445 	ip = mtod(m, struct ip *);
446 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) ||
447 	    ntohs(ip->ip_off) & (IP_MF | IP_OFFMASK)) {
448 		/*
449 		 * Force hash recalculation for fragments and multicast
450 		 * packets; hardware may not do it correctly.
451 		 * XXX add flag to indicate the hash is from hardware
452 		 */
453 		m->m_flags &= ~M_HASH;
454 	}
455 	if ((m->m_flags & M_HASH) == 0) {
456 		ip_hashfn(&m, 0, IP_MPORT_IN);
457 		if (m == NULL)
458 			return;
459 		KKASSERT(m->m_flags & M_HASH);
460 
461 		if (&curthread->td_msgport !=
462 		    netisr_hashport(m->m_pkthdr.hash)) {
463 			netisr_queue(NETISR_IP, m);
464 			/* Requeued to other netisr msgport; done */
465 			return;
466 		}
467 
468 		/* mbuf could have been changed */
469 		ip = mtod(m, struct ip *);
470 	}
471 
472 	/*
473 	 * Pull out certain tags
474 	 */
475 	if (m->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) {
476 		/* Next hop */
477 		mtag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL);
478 		KKASSERT(mtag != NULL);
479 		next_hop = m_tag_data(mtag);
480 	}
481 
482 	if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
483 		/* dummynet already filtered us */
484 		ip = mtod(m, struct ip *);
485 		hlen = IP_VHL_HL(ip->ip_vhl) << 2;
486 		goto iphack;
487 	}
488 
489 	ipstat.ips_total++;
490 
491 	/* length checks already done in ip_hashfn() */
492 	KASSERT(m->m_len >= sizeof(struct ip), ("IP header not in one mbuf"));
493 
494 	if (IP_VHL_V(ip->ip_vhl) != IPVERSION) {
495 		ipstat.ips_badvers++;
496 		goto bad;
497 	}
498 
499 	hlen = IP_VHL_HL(ip->ip_vhl) << 2;
500 	/* length checks already done in ip_hashfn() */
501 	KASSERT(hlen >= sizeof(struct ip), ("IP header len too small"));
502 	KASSERT(m->m_len >= hlen, ("complete IP header not in one mbuf"));
503 
504 	/* 127/8 must not appear on wire - RFC1122 */
505 	if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
506 	    (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
507 		if (!(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK)) {
508 			ipstat.ips_badaddr++;
509 			goto bad;
510 		}
511 	}
512 
513 	if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
514 		sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
515 	} else {
516 		if (hlen == sizeof(struct ip))
517 			sum = in_cksum_hdr(ip);
518 		else
519 			sum = in_cksum(m, hlen);
520 	}
521 	if (sum != 0) {
522 		ipstat.ips_badsum++;
523 		goto bad;
524 	}
525 
526 #ifdef ALTQ
527 	if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) {
528 		/* packet is dropped by traffic conditioner */
529 		return;
530 	}
531 #endif
532 	/*
533 	 * Convert fields to host representation.
534 	 */
535 	ip->ip_len = ntohs(ip->ip_len);
536 	ip->ip_off = ntohs(ip->ip_off);
537 
538 	/* length checks already done in ip_hashfn() */
539 	KASSERT(ip->ip_len >= hlen, ("total length less then header length"));
540 	KASSERT(m->m_pkthdr.len >= ip->ip_len, ("mbuf too short"));
541 
542 	/*
543 	 * Trim mbufs if longer than the IP header would have us expect.
544 	 */
545 	if (m->m_pkthdr.len > ip->ip_len) {
546 		if (m->m_len == m->m_pkthdr.len) {
547 			m->m_len = ip->ip_len;
548 			m->m_pkthdr.len = ip->ip_len;
549 		} else {
550 			m_adj(m, ip->ip_len - m->m_pkthdr.len);
551 		}
552 	}
553 #if defined(IPSEC) && !defined(IPSEC_FILTERGIF)
554 	/*
555 	 * Bypass packet filtering for packets from a tunnel (gif).
556 	 */
557 	if (ipsec_gethist(m, NULL))
558 		goto pass;
559 #endif
560 
561 	/*
562 	 * IpHack's section.
563 	 * Right now when no processing on packet has done
564 	 * and it is still fresh out of network we do our black
565 	 * deals with it.
566 	 * - Firewall: deny/allow/divert
567 	 * - Xlate: translate packet's addr/port (NAT).
568 	 * - Pipe: pass pkt through dummynet.
569 	 * - Wrap: fake packet's addr/port <unimpl.>
570 	 * - Encapsulate: put it in another IP and send out. <unimp.>
571 	 */
572 
573 iphack:
574 	/*
575 	 * If we've been forwarded from the output side, then
576 	 * skip the firewall a second time
577 	 */
578 	if (next_hop != NULL)
579 		goto ours;
580 
581 	/* No pfil hooks */
582 	if (!pfil_has_hooks(&inet_pfil_hook)) {
583 		if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
584 			/*
585 			 * Strip dummynet tags from stranded packets
586 			 */
587 			mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
588 			KKASSERT(mtag != NULL);
589 			m_tag_delete(m, mtag);
590 			m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED;
591 		}
592 		goto pass;
593 	}
594 
595 	/*
596 	 * Run through list of hooks for input packets.
597 	 *
598 	 * NOTE!  If the packet is rewritten pf/ipfw/whoever must
599 	 *	  clear M_HASH.
600 	 */
601 	odst = ip->ip_dst;
602 	if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif, PFIL_IN))
603 		return;
604 	if (m == NULL)	/* consumed by filter */
605 		return;
606 	ip = mtod(m, struct ip *);
607 	hlen = IP_VHL_HL(ip->ip_vhl) << 2;
608 	using_srcrt = (odst.s_addr != ip->ip_dst.s_addr);
609 
610 	if (m->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) {
611 		mtag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL);
612 		KKASSERT(mtag != NULL);
613 		next_hop = m_tag_data(mtag);
614 	}
615 	if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
616 		ip_dn_queue(m);
617 		return;
618 	}
619 	if (m->m_pkthdr.fw_flags & FW_MBUF_REDISPATCH) {
620 		m->m_pkthdr.fw_flags &= ~FW_MBUF_REDISPATCH;
621 	}
622 pass:
623 	/*
624 	 * Process options and, if not destined for us,
625 	 * ship it on.  ip_dooptions returns 1 when an
626 	 * error was detected (causing an icmp message
627 	 * to be sent and the original packet to be freed).
628 	 */
629 	if (hlen > sizeof(struct ip) && ip_dooptions(m, 0, next_hop))
630 		return;
631 
632 	/* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
633 	 * matter if it is destined to another node, or whether it is
634 	 * a multicast one, RSVP wants it! and prevents it from being forwarded
635 	 * anywhere else. Also checks if the rsvp daemon is running before
636 	 * grabbing the packet.
637 	 */
638 	if (rsvp_on && ip->ip_p == IPPROTO_RSVP)
639 		goto ours;
640 
641 	/*
642 	 * Check our list of addresses, to see if the packet is for us.
643 	 * If we don't have any addresses, assume any unicast packet
644 	 * we receive might be for us (and let the upper layers deal
645 	 * with it).
646 	 */
647 	if (TAILQ_EMPTY(&in_ifaddrheads[mycpuid]) &&
648 	    !(m->m_flags & (M_MCAST | M_BCAST)))
649 		goto ours;
650 
651 	/*
652 	 * Cache the destination address of the packet; this may be
653 	 * changed by use of 'ipfw fwd'.
654 	 */
655 	pkt_dst = next_hop ? next_hop->sin_addr : ip->ip_dst;
656 
657 	/*
658 	 * Enable a consistency check between the destination address
659 	 * and the arrival interface for a unicast packet (the RFC 1122
660 	 * strong ES model) if IP forwarding is disabled and the packet
661 	 * is not locally generated and the packet is not subject to
662 	 * 'ipfw fwd'.
663 	 *
664 	 * XXX - Checking also should be disabled if the destination
665 	 * address is ipnat'ed to a different interface.
666 	 *
667 	 * XXX - Checking is incompatible with IP aliases added
668 	 * to the loopback interface instead of the interface where
669 	 * the packets are received.
670 	 */
671 	checkif = ip_checkinterface &&
672 		  !ipforwarding &&
673 		  m->m_pkthdr.rcvif != NULL &&
674 		  !(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) &&
675 		  next_hop == NULL;
676 
677 	/*
678 	 * Check for exact addresses in the hash bucket.
679 	 */
680 	LIST_FOREACH(iac, INADDR_HASH(pkt_dst.s_addr), ia_hash) {
681 		ia = iac->ia;
682 
683 		/*
684 		 * If the address matches, verify that the packet
685 		 * arrived via the correct interface if checking is
686 		 * enabled.
687 		 */
688 		if (IA_SIN(ia)->sin_addr.s_addr == pkt_dst.s_addr &&
689 		    (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif))
690 			goto ours;
691 	}
692 	ia = NULL;
693 
694 	/*
695 	 * Check for broadcast addresses.
696 	 *
697 	 * Only accept broadcast packets that arrive via the matching
698 	 * interface.  Reception of forwarded directed broadcasts would
699 	 * be handled via ip_forward() and ether_output() with the loopback
700 	 * into the stack for SIMPLEX interfaces handled by ether_output().
701 	 */
702 	if (m->m_pkthdr.rcvif != NULL &&
703 	    m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
704 		struct ifaddr_container *ifac;
705 
706 		TAILQ_FOREACH(ifac, &m->m_pkthdr.rcvif->if_addrheads[mycpuid],
707 			      ifa_link) {
708 			struct ifaddr *ifa = ifac->ifa;
709 
710 			if (ifa->ifa_addr == NULL) /* shutdown/startup race */
711 				continue;
712 			if (ifa->ifa_addr->sa_family != AF_INET)
713 				continue;
714 			ia = ifatoia(ifa);
715 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
716 								pkt_dst.s_addr)
717 				goto ours;
718 			if (ia->ia_netbroadcast.s_addr == pkt_dst.s_addr)
719 				goto ours;
720 #ifdef BOOTP_COMPAT
721 			if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY)
722 				goto ours;
723 #endif
724 		}
725 	}
726 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
727 		struct in_multi *inm;
728 
729 		if (ip_mrouter != NULL) {
730 			/* XXX Multicast routing is not MPSAFE yet */
731 			get_mplock();
732 
733 			/*
734 			 * If we are acting as a multicast router, all
735 			 * incoming multicast packets are passed to the
736 			 * kernel-level multicast forwarding function.
737 			 * The packet is returned (relatively) intact; if
738 			 * ip_mforward() returns a non-zero value, the packet
739 			 * must be discarded, else it may be accepted below.
740 			 */
741 			if (ip_mforward != NULL &&
742 			    ip_mforward(ip, m->m_pkthdr.rcvif, m, NULL) != 0) {
743 				rel_mplock();
744 				ipstat.ips_cantforward++;
745 				m_freem(m);
746 				return;
747 			}
748 
749 			rel_mplock();
750 
751 			/*
752 			 * The process-level routing daemon needs to receive
753 			 * all multicast IGMP packets, whether or not this
754 			 * host belongs to their destination groups.
755 			 */
756 			if (ip->ip_p == IPPROTO_IGMP)
757 				goto ours;
758 			ipstat.ips_forward++;
759 		}
760 		/*
761 		 * See if we belong to the destination multicast group on the
762 		 * arrival interface.
763 		 */
764 		inm = IN_LOOKUP_MULTI(&ip->ip_dst, m->m_pkthdr.rcvif);
765 		if (inm == NULL) {
766 			ipstat.ips_notmember++;
767 			m_freem(m);
768 			return;
769 		}
770 		goto ours;
771 	}
772 	if (ip->ip_dst.s_addr == INADDR_BROADCAST)
773 		goto ours;
774 	if (ip->ip_dst.s_addr == INADDR_ANY)
775 		goto ours;
776 
777 	/*
778 	 * FAITH(Firewall Aided Internet Translator)
779 	 */
780 	if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
781 		if (ip_keepfaith) {
782 			if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP)
783 				goto ours;
784 		}
785 		m_freem(m);
786 		return;
787 	}
788 
789 	/*
790 	 * Not for us; forward if possible and desirable.
791 	 */
792 	if (!ipforwarding) {
793 		ipstat.ips_cantforward++;
794 		m_freem(m);
795 	} else {
796 #ifdef IPSEC
797 		/*
798 		 * Enforce inbound IPsec SPD.
799 		 */
800 		if (ipsec4_in_reject(m, NULL)) {
801 			ipsecstat.in_polvio++;
802 			goto bad;
803 		}
804 #endif
805 #ifdef FAST_IPSEC
806 		mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
807 		crit_enter();
808 		if (mtag != NULL) {
809 			tdbi = (struct tdb_ident *)m_tag_data(mtag);
810 			sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
811 		} else {
812 			sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
813 						   IP_FORWARDING, &error);
814 		}
815 		if (sp == NULL) {	/* NB: can happen if error */
816 			crit_exit();
817 			/*XXX error stat???*/
818 			DPRINTF(("ip_input: no SP for forwarding\n"));	/*XXX*/
819 			goto bad;
820 		}
821 
822 		/*
823 		 * Check security policy against packet attributes.
824 		 */
825 		error = ipsec_in_reject(sp, m);
826 		KEY_FREESP(&sp);
827 		crit_exit();
828 		if (error) {
829 			ipstat.ips_cantforward++;
830 			goto bad;
831 		}
832 #endif
833 		ip_forward(m, using_srcrt, next_hop);
834 	}
835 	return;
836 
837 ours:
838 
839 	/*
840 	 * IPSTEALTH: Process non-routing options only
841 	 * if the packet is destined for us.
842 	 */
843 	if (ipstealth &&
844 	    hlen > sizeof(struct ip) &&
845 	    ip_dooptions(m, 1, next_hop))
846 		return;
847 
848 	/* Count the packet in the ip address stats */
849 	if (ia != NULL) {
850 		IFA_STAT_INC(&ia->ia_ifa, ipackets, 1);
851 		IFA_STAT_INC(&ia->ia_ifa, ibytes, m->m_pkthdr.len);
852 	}
853 
854 	/*
855 	 * If offset or IP_MF are set, must reassemble.
856 	 * Otherwise, nothing need be done.
857 	 * (We could look in the reassembly queue to see
858 	 * if the packet was previously fragmented,
859 	 * but it's not worth the time; just let them time out.)
860 	 */
861 	if (ip->ip_off & (IP_MF | IP_OFFMASK)) {
862 		/*
863 		 * Attempt reassembly; if it succeeds, proceed.  ip_reass()
864 		 * will return a different mbuf.
865 		 *
866 		 * NOTE: ip_reass() returns m with M_HASH cleared to force
867 		 *	 us to recharacterize the packet.
868 		 */
869 		m = ip_reass(m);
870 		if (m == NULL)
871 			return;
872 		ip = mtod(m, struct ip *);
873 
874 		/* Get the header length of the reassembled packet */
875 		hlen = IP_VHL_HL(ip->ip_vhl) << 2;
876 	} else {
877 		ip->ip_len -= hlen;
878 	}
879 
880 #ifdef IPSEC
881 	/*
882 	 * enforce IPsec policy checking if we are seeing last header.
883 	 * note that we do not visit this with protocols with pcb layer
884 	 * code - like udp/tcp/raw ip.
885 	 */
886 	if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) &&
887 	    ipsec4_in_reject(m, NULL)) {
888 		ipsecstat.in_polvio++;
889 		goto bad;
890 	}
891 #endif
892 #if FAST_IPSEC
893 	/*
894 	 * enforce IPsec policy checking if we are seeing last header.
895 	 * note that we do not visit this with protocols with pcb layer
896 	 * code - like udp/tcp/raw ip.
897 	 */
898 	if (inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) {
899 		/*
900 		 * Check if the packet has already had IPsec processing
901 		 * done.  If so, then just pass it along.  This tag gets
902 		 * set during AH, ESP, etc. input handling, before the
903 		 * packet is returned to the ip input queue for delivery.
904 		 */
905 		mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
906 		crit_enter();
907 		if (mtag != NULL) {
908 			tdbi = (struct tdb_ident *)m_tag_data(mtag);
909 			sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
910 		} else {
911 			sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
912 						   IP_FORWARDING, &error);
913 		}
914 		if (sp != NULL) {
915 			/*
916 			 * Check security policy against packet attributes.
917 			 */
918 			error = ipsec_in_reject(sp, m);
919 			KEY_FREESP(&sp);
920 		} else {
921 			/* XXX error stat??? */
922 			error = EINVAL;
923 DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/
924 			crit_exit();
925 			goto bad;
926 		}
927 		crit_exit();
928 		if (error)
929 			goto bad;
930 	}
931 #endif /* FAST_IPSEC */
932 
933 	/*
934 	 * We must forward the packet to the correct protocol thread if
935 	 * we are not already in it.
936 	 *
937 	 * NOTE: ip_len is now in host form.  ip_len is not adjusted
938 	 *	 further for protocol processing, instead we pass hlen
939 	 *	 to the protosw and let it deal with it.
940 	 */
941 	ipstat.ips_delivered++;
942 
943 	if ((m->m_flags & M_HASH) == 0) {
944 #ifdef RSS_DEBUG
945 		atomic_add_long(&ip_rehash_count, 1);
946 #endif
947 		ip->ip_len = htons(ip->ip_len + hlen);
948 		ip->ip_off = htons(ip->ip_off);
949 
950 		ip_hashfn(&m, 0, IP_MPORT_IN);
951 		if (m == NULL)
952 			return;
953 
954 		ip = mtod(m, struct ip *);
955 		ip->ip_len = ntohs(ip->ip_len) - hlen;
956 		ip->ip_off = ntohs(ip->ip_off);
957 		KKASSERT(m->m_flags & M_HASH);
958 	}
959 	port = netisr_hashport(m->m_pkthdr.hash);
960 
961 	if (port != &curthread->td_msgport) {
962 		struct netmsg_packet *pmsg;
963 
964 #ifdef RSS_DEBUG
965 		atomic_add_long(&ip_dispatch_slow, 1);
966 #endif
967 
968 		pmsg = &m->m_hdr.mh_netmsg;
969 		netmsg_init(&pmsg->base, NULL, &netisr_apanic_rport,
970 			    0, transport_processing_handler);
971 		pmsg->nm_packet = m;
972 		pmsg->base.lmsg.u.ms_result = hlen;
973 		lwkt_sendmsg(port, &pmsg->base.lmsg);
974 	} else {
975 #ifdef RSS_DEBUG
976 		atomic_add_long(&ip_dispatch_fast, 1);
977 #endif
978 		transport_processing_oncpu(m, hlen, ip);
979 	}
980 	return;
981 
982 bad:
983 	m_freem(m);
984 }
985 
986 /*
987  * Take incoming datagram fragment and try to reassemble it into
988  * whole datagram.  If a chain for reassembly of this datagram already
989  * exists, then it is given as fp; otherwise have to make a chain.
990  */
991 struct mbuf *
992 ip_reass(struct mbuf *m)
993 {
994 	struct ip *ip = mtod(m, struct ip *);
995 	struct mbuf *p = NULL, *q, *nq;
996 	struct mbuf *n;
997 	struct ipq *fp = NULL;
998 	struct ipqhead *head;
999 	int hlen = IP_VHL_HL(ip->ip_vhl) << 2;
1000 	int i, next;
1001 	u_short sum;
1002 
1003 	/* If maxnipq is 0, never accept fragments. */
1004 	if (maxnipq == 0) {
1005 		ipstat.ips_fragments++;
1006 		ipstat.ips_fragdropped++;
1007 		m_freem(m);
1008 		return NULL;
1009 	}
1010 
1011 	sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
1012 	/*
1013 	 * Look for queue of fragments of this datagram.
1014 	 */
1015 	lwkt_gettoken(&ipq_token);
1016 	head = &ipq[sum];
1017 	TAILQ_FOREACH(fp, head, ipq_list) {
1018 		if (ip->ip_id == fp->ipq_id &&
1019 		    ip->ip_src.s_addr == fp->ipq_src.s_addr &&
1020 		    ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
1021 		    ip->ip_p == fp->ipq_p)
1022 			goto found;
1023 	}
1024 
1025 	fp = NULL;
1026 
1027 	/*
1028 	 * Enforce upper bound on number of fragmented packets
1029 	 * for which we attempt reassembly;
1030 	 * If maxnipq is -1, accept all fragments without limitation.
1031 	 */
1032 	if (nipq > maxnipq && maxnipq > 0) {
1033 		/*
1034 		 * drop something from the tail of the current queue
1035 		 * before proceeding further
1036 		 */
1037 		struct ipq *q = TAILQ_LAST(head, ipqhead);
1038 		if (q == NULL) {
1039 			/*
1040 			 * The current queue is empty,
1041 			 * so drop from one of the others.
1042 			 */
1043 			for (i = 0; i < IPREASS_NHASH; i++) {
1044 				struct ipq *r = TAILQ_LAST(&ipq[i], ipqhead);
1045 				if (r) {
1046 					ipstat.ips_fragtimeout += r->ipq_nfrags;
1047 					ip_freef(&ipq[i], r);
1048 					break;
1049 				}
1050 			}
1051 		} else {
1052 			ipstat.ips_fragtimeout += q->ipq_nfrags;
1053 			ip_freef(head, q);
1054 		}
1055 	}
1056 found:
1057 	/*
1058 	 * Adjust ip_len to not reflect header,
1059 	 * convert offset of this to bytes.
1060 	 */
1061 	ip->ip_len -= hlen;
1062 	if (ip->ip_off & IP_MF) {
1063 		/*
1064 		 * Make sure that fragments have a data length
1065 		 * that's a non-zero multiple of 8 bytes.
1066 		 */
1067 		if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
1068 			ipstat.ips_toosmall++; /* XXX */
1069 			m_freem(m);
1070 			goto done;
1071 		}
1072 		m->m_flags |= M_FRAG;
1073 	} else {
1074 		m->m_flags &= ~M_FRAG;
1075 	}
1076 	ip->ip_off <<= 3;
1077 
1078 	ipstat.ips_fragments++;
1079 	m->m_pkthdr.header = ip;
1080 
1081 	/*
1082 	 * If the hardware has not done csum over this fragment
1083 	 * then csum_data is not valid at all.
1084 	 */
1085 	if ((m->m_pkthdr.csum_flags & (CSUM_FRAG_NOT_CHECKED | CSUM_DATA_VALID))
1086 	    == (CSUM_FRAG_NOT_CHECKED | CSUM_DATA_VALID)) {
1087 		m->m_pkthdr.csum_data = 0;
1088 		m->m_pkthdr.csum_flags &= ~(CSUM_DATA_VALID | CSUM_PSEUDO_HDR);
1089 	}
1090 
1091 	/*
1092 	 * Presence of header sizes in mbufs
1093 	 * would confuse code below.
1094 	 */
1095 	m->m_data += hlen;
1096 	m->m_len -= hlen;
1097 
1098 	/*
1099 	 * If first fragment to arrive, create a reassembly queue.
1100 	 */
1101 	if (fp == NULL) {
1102 		if ((fp = mpipe_alloc_nowait(&ipq_mpipe)) == NULL)
1103 			goto dropfrag;
1104 		TAILQ_INSERT_HEAD(head, fp, ipq_list);
1105 		nipq++;
1106 		fp->ipq_nfrags = 1;
1107 		fp->ipq_ttl = IPFRAGTTL;
1108 		fp->ipq_p = ip->ip_p;
1109 		fp->ipq_id = ip->ip_id;
1110 		fp->ipq_src = ip->ip_src;
1111 		fp->ipq_dst = ip->ip_dst;
1112 		fp->ipq_frags = m;
1113 		m->m_nextpkt = NULL;
1114 		goto inserted;
1115 	} else {
1116 		fp->ipq_nfrags++;
1117 	}
1118 
1119 #define	GETIP(m)	((struct ip*)((m)->m_pkthdr.header))
1120 
1121 	/*
1122 	 * Find a segment which begins after this one does.
1123 	 */
1124 	for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
1125 		if (GETIP(q)->ip_off > ip->ip_off)
1126 			break;
1127 	}
1128 
1129 	/*
1130 	 * If there is a preceding segment, it may provide some of
1131 	 * our data already.  If so, drop the data from the incoming
1132 	 * segment.  If it provides all of our data, drop us, otherwise
1133 	 * stick new segment in the proper place.
1134 	 *
1135 	 * If some of the data is dropped from the the preceding
1136 	 * segment, then it's checksum is invalidated.
1137 	 */
1138 	if (p) {
1139 		i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off;
1140 		if (i > 0) {
1141 			if (i >= ip->ip_len)
1142 				goto dropfrag;
1143 			m_adj(m, i);
1144 			m->m_pkthdr.csum_flags = 0;
1145 			ip->ip_off += i;
1146 			ip->ip_len -= i;
1147 		}
1148 		m->m_nextpkt = p->m_nextpkt;
1149 		p->m_nextpkt = m;
1150 	} else {
1151 		m->m_nextpkt = fp->ipq_frags;
1152 		fp->ipq_frags = m;
1153 	}
1154 
1155 	/*
1156 	 * While we overlap succeeding segments trim them or,
1157 	 * if they are completely covered, dequeue them.
1158 	 */
1159 	for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off;
1160 	     q = nq) {
1161 		i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off;
1162 		if (i < GETIP(q)->ip_len) {
1163 			GETIP(q)->ip_len -= i;
1164 			GETIP(q)->ip_off += i;
1165 			m_adj(q, i);
1166 			q->m_pkthdr.csum_flags = 0;
1167 			break;
1168 		}
1169 		nq = q->m_nextpkt;
1170 		m->m_nextpkt = nq;
1171 		ipstat.ips_fragdropped++;
1172 		fp->ipq_nfrags--;
1173 		q->m_nextpkt = NULL;
1174 		m_freem(q);
1175 	}
1176 
1177 inserted:
1178 	/*
1179 	 * Check for complete reassembly and perform frag per packet
1180 	 * limiting.
1181 	 *
1182 	 * Frag limiting is performed here so that the nth frag has
1183 	 * a chance to complete the packet before we drop the packet.
1184 	 * As a result, n+1 frags are actually allowed per packet, but
1185 	 * only n will ever be stored. (n = maxfragsperpacket.)
1186 	 *
1187 	 */
1188 	next = 0;
1189 	for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
1190 		if (GETIP(q)->ip_off != next) {
1191 			if (fp->ipq_nfrags > maxfragsperpacket) {
1192 				ipstat.ips_fragdropped += fp->ipq_nfrags;
1193 				ip_freef(head, fp);
1194 			}
1195 			goto done;
1196 		}
1197 		next += GETIP(q)->ip_len;
1198 	}
1199 	/* Make sure the last packet didn't have the IP_MF flag */
1200 	if (p->m_flags & M_FRAG) {
1201 		if (fp->ipq_nfrags > maxfragsperpacket) {
1202 			ipstat.ips_fragdropped += fp->ipq_nfrags;
1203 			ip_freef(head, fp);
1204 		}
1205 		goto done;
1206 	}
1207 
1208 	/*
1209 	 * Reassembly is complete.  Make sure the packet is a sane size.
1210 	 */
1211 	q = fp->ipq_frags;
1212 	ip = GETIP(q);
1213 	if (next + (IP_VHL_HL(ip->ip_vhl) << 2) > IP_MAXPACKET) {
1214 		ipstat.ips_toolong++;
1215 		ipstat.ips_fragdropped += fp->ipq_nfrags;
1216 		ip_freef(head, fp);
1217 		goto done;
1218 	}
1219 
1220 	/*
1221 	 * Concatenate fragments.
1222 	 */
1223 	m = q;
1224 	n = m->m_next;
1225 	m->m_next = NULL;
1226 	m_cat(m, n);
1227 	nq = q->m_nextpkt;
1228 	q->m_nextpkt = NULL;
1229 	for (q = nq; q != NULL; q = nq) {
1230 		nq = q->m_nextpkt;
1231 		q->m_nextpkt = NULL;
1232 		m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
1233 		m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
1234 		m_cat(m, q);
1235 	}
1236 
1237 	/*
1238 	 * Clean up the 1's complement checksum.  Carry over 16 bits must
1239 	 * be added back.  This assumes no more then 65535 packet fragments
1240 	 * were reassembled.  A second carry can also occur (but not a third).
1241 	 */
1242 	m->m_pkthdr.csum_data = (m->m_pkthdr.csum_data & 0xffff) +
1243 				(m->m_pkthdr.csum_data >> 16);
1244 	if (m->m_pkthdr.csum_data > 0xFFFF)
1245 		m->m_pkthdr.csum_data -= 0xFFFF;
1246 
1247 	/*
1248 	 * Create header for new ip packet by
1249 	 * modifying header of first packet;
1250 	 * dequeue and discard fragment reassembly header.
1251 	 * Make header visible.
1252 	 */
1253 	ip->ip_len = next;
1254 	ip->ip_src = fp->ipq_src;
1255 	ip->ip_dst = fp->ipq_dst;
1256 	TAILQ_REMOVE(head, fp, ipq_list);
1257 	nipq--;
1258 	mpipe_free(&ipq_mpipe, fp);
1259 	m->m_len += (IP_VHL_HL(ip->ip_vhl) << 2);
1260 	m->m_data -= (IP_VHL_HL(ip->ip_vhl) << 2);
1261 	/* some debugging cruft by sklower, below, will go away soon */
1262 	if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */
1263 		int plen = 0;
1264 
1265 		for (n = m; n; n = n->m_next)
1266 			plen += n->m_len;
1267 		m->m_pkthdr.len = plen;
1268 	}
1269 
1270 	/*
1271 	 * Reassembly complete, return the next protocol.
1272 	 *
1273 	 * Be sure to clear M_HASH to force the packet
1274 	 * to be re-characterized.
1275 	 *
1276 	 * Clear M_FRAG, we are no longer a fragment.
1277 	 */
1278 	m->m_flags &= ~(M_HASH | M_FRAG);
1279 
1280 	ipstat.ips_reassembled++;
1281 	lwkt_reltoken(&ipq_token);
1282 	return (m);
1283 
1284 dropfrag:
1285 	ipstat.ips_fragdropped++;
1286 	if (fp != NULL)
1287 		fp->ipq_nfrags--;
1288 	m_freem(m);
1289 done:
1290 	lwkt_reltoken(&ipq_token);
1291 	return (NULL);
1292 
1293 #undef GETIP
1294 }
1295 
1296 /*
1297  * Free a fragment reassembly header and all
1298  * associated datagrams.
1299  *
1300  * Called with ipq_token held.
1301  */
1302 static void
1303 ip_freef(struct ipqhead *fhp, struct ipq *fp)
1304 {
1305 	struct mbuf *q;
1306 
1307 	/*
1308 	 * Remove first to protect against blocking
1309 	 */
1310 	TAILQ_REMOVE(fhp, fp, ipq_list);
1311 
1312 	/*
1313 	 * Clean out at our leisure
1314 	 */
1315 	while (fp->ipq_frags) {
1316 		q = fp->ipq_frags;
1317 		fp->ipq_frags = q->m_nextpkt;
1318 		q->m_nextpkt = NULL;
1319 		m_freem(q);
1320 	}
1321 	mpipe_free(&ipq_mpipe, fp);
1322 	nipq--;
1323 }
1324 
1325 /*
1326  * IP timer processing;
1327  * if a timer expires on a reassembly
1328  * queue, discard it.
1329  */
1330 void
1331 ip_slowtimo(void)
1332 {
1333 	struct ipq *fp, *fp_temp;
1334 	struct ipqhead *head;
1335 	int i;
1336 
1337 	lwkt_gettoken(&ipq_token);
1338 	for (i = 0; i < IPREASS_NHASH; i++) {
1339 		head = &ipq[i];
1340 		TAILQ_FOREACH_MUTABLE(fp, head, ipq_list, fp_temp) {
1341 			if (--fp->ipq_ttl == 0) {
1342 				ipstat.ips_fragtimeout += fp->ipq_nfrags;
1343 				ip_freef(head, fp);
1344 			}
1345 		}
1346 	}
1347 	/*
1348 	 * If we are over the maximum number of fragments
1349 	 * (due to the limit being lowered), drain off
1350 	 * enough to get down to the new limit.
1351 	 */
1352 	if (maxnipq >= 0 && nipq > maxnipq) {
1353 		for (i = 0; i < IPREASS_NHASH; i++) {
1354 			head = &ipq[i];
1355 			while (nipq > maxnipq && !TAILQ_EMPTY(head)) {
1356 				ipstat.ips_fragdropped +=
1357 				    TAILQ_FIRST(head)->ipq_nfrags;
1358 				ip_freef(head, TAILQ_FIRST(head));
1359 			}
1360 		}
1361 	}
1362 	lwkt_reltoken(&ipq_token);
1363 	ipflow_slowtimo();
1364 }
1365 
1366 /*
1367  * Drain off all datagram fragments.
1368  */
1369 void
1370 ip_drain(void)
1371 {
1372 	struct ipqhead *head;
1373 	int i;
1374 
1375 	lwkt_gettoken(&ipq_token);
1376 	for (i = 0; i < IPREASS_NHASH; i++) {
1377 		head = &ipq[i];
1378 		while (!TAILQ_EMPTY(head)) {
1379 			ipstat.ips_fragdropped += TAILQ_FIRST(head)->ipq_nfrags;
1380 			ip_freef(head, TAILQ_FIRST(head));
1381 		}
1382 	}
1383 	lwkt_reltoken(&ipq_token);
1384 	in_rtqdrain();
1385 }
1386 
1387 /*
1388  * Do option processing on a datagram,
1389  * possibly discarding it if bad options are encountered,
1390  * or forwarding it if source-routed.
1391  * The pass argument is used when operating in the IPSTEALTH
1392  * mode to tell what options to process:
1393  * [LS]SRR (pass 0) or the others (pass 1).
1394  * The reason for as many as two passes is that when doing IPSTEALTH,
1395  * non-routing options should be processed only if the packet is for us.
1396  * Returns 1 if packet has been forwarded/freed,
1397  * 0 if the packet should be processed further.
1398  */
1399 static int
1400 ip_dooptions(struct mbuf *m, int pass, struct sockaddr_in *next_hop)
1401 {
1402 	struct sockaddr_in ipaddr = { sizeof ipaddr, AF_INET };
1403 	struct ip *ip = mtod(m, struct ip *);
1404 	u_char *cp;
1405 	struct in_ifaddr *ia;
1406 	int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB;
1407 	boolean_t forward = FALSE;
1408 	struct in_addr *sin, dst;
1409 	n_time ntime;
1410 
1411 	dst = ip->ip_dst;
1412 	cp = (u_char *)(ip + 1);
1413 	cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
1414 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1415 		opt = cp[IPOPT_OPTVAL];
1416 		if (opt == IPOPT_EOL)
1417 			break;
1418 		if (opt == IPOPT_NOP)
1419 			optlen = 1;
1420 		else {
1421 			if (cnt < IPOPT_OLEN + sizeof(*cp)) {
1422 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1423 				goto bad;
1424 			}
1425 			optlen = cp[IPOPT_OLEN];
1426 			if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
1427 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1428 				goto bad;
1429 			}
1430 		}
1431 		switch (opt) {
1432 
1433 		default:
1434 			break;
1435 
1436 		/*
1437 		 * Source routing with record.
1438 		 * Find interface with current destination address.
1439 		 * If none on this machine then drop if strictly routed,
1440 		 * or do nothing if loosely routed.
1441 		 * Record interface address and bring up next address
1442 		 * component.  If strictly routed make sure next
1443 		 * address is on directly accessible net.
1444 		 */
1445 		case IPOPT_LSRR:
1446 		case IPOPT_SSRR:
1447 			if (ipstealth && pass > 0)
1448 				break;
1449 			if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1450 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1451 				goto bad;
1452 			}
1453 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1454 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1455 				goto bad;
1456 			}
1457 			ipaddr.sin_addr = ip->ip_dst;
1458 			ia = (struct in_ifaddr *)
1459 				ifa_ifwithaddr((struct sockaddr *)&ipaddr);
1460 			if (ia == NULL) {
1461 				if (opt == IPOPT_SSRR) {
1462 					type = ICMP_UNREACH;
1463 					code = ICMP_UNREACH_SRCFAIL;
1464 					goto bad;
1465 				}
1466 				if (!ip_dosourceroute)
1467 					goto nosourcerouting;
1468 				/*
1469 				 * Loose routing, and not at next destination
1470 				 * yet; nothing to do except forward.
1471 				 */
1472 				break;
1473 			}
1474 			off--;			/* 0 origin */
1475 			if (off > optlen - (int)sizeof(struct in_addr)) {
1476 				/*
1477 				 * End of source route.  Should be for us.
1478 				 */
1479 				if (!ip_acceptsourceroute)
1480 					goto nosourcerouting;
1481 				save_rte(m, cp, ip->ip_src);
1482 				break;
1483 			}
1484 			if (ipstealth)
1485 				goto dropit;
1486 			if (!ip_dosourceroute) {
1487 				if (ipforwarding) {
1488 					char buf[sizeof "aaa.bbb.ccc.ddd"];
1489 
1490 					/*
1491 					 * Acting as a router, so generate ICMP
1492 					 */
1493 nosourcerouting:
1494 					strcpy(buf, inet_ntoa(ip->ip_dst));
1495 					log(LOG_WARNING,
1496 					    "attempted source route from %s to %s\n",
1497 					    inet_ntoa(ip->ip_src), buf);
1498 					type = ICMP_UNREACH;
1499 					code = ICMP_UNREACH_SRCFAIL;
1500 					goto bad;
1501 				} else {
1502 					/*
1503 					 * Not acting as a router,
1504 					 * so silently drop.
1505 					 */
1506 dropit:
1507 					ipstat.ips_cantforward++;
1508 					m_freem(m);
1509 					return (1);
1510 				}
1511 			}
1512 
1513 			/*
1514 			 * locate outgoing interface
1515 			 */
1516 			memcpy(&ipaddr.sin_addr, cp + off,
1517 			    sizeof ipaddr.sin_addr);
1518 
1519 			if (opt == IPOPT_SSRR) {
1520 #define	INA	struct in_ifaddr *
1521 #define	SA	struct sockaddr *
1522 				if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr))
1523 									== NULL)
1524 					ia = (INA)ifa_ifwithnet((SA)&ipaddr);
1525 			} else {
1526 				ia = ip_rtaddr(ipaddr.sin_addr, NULL);
1527 			}
1528 			if (ia == NULL) {
1529 				type = ICMP_UNREACH;
1530 				code = ICMP_UNREACH_SRCFAIL;
1531 				goto bad;
1532 			}
1533 			ip->ip_dst = ipaddr.sin_addr;
1534 			memcpy(cp + off, &IA_SIN(ia)->sin_addr,
1535 			    sizeof(struct in_addr));
1536 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1537 			/*
1538 			 * Let ip_intr's mcast routing check handle mcast pkts
1539 			 */
1540 			forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr));
1541 			break;
1542 
1543 		case IPOPT_RR:
1544 			if (ipstealth && pass == 0)
1545 				break;
1546 			if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1547 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1548 				goto bad;
1549 			}
1550 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1551 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1552 				goto bad;
1553 			}
1554 			/*
1555 			 * If no space remains, ignore.
1556 			 */
1557 			off--;			/* 0 origin */
1558 			if (off > optlen - (int)sizeof(struct in_addr))
1559 				break;
1560 			memcpy(&ipaddr.sin_addr, &ip->ip_dst,
1561 			    sizeof ipaddr.sin_addr);
1562 			/*
1563 			 * locate outgoing interface; if we're the destination,
1564 			 * use the incoming interface (should be same).
1565 			 */
1566 			if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == NULL &&
1567 			    (ia = ip_rtaddr(ipaddr.sin_addr, NULL)) == NULL) {
1568 				type = ICMP_UNREACH;
1569 				code = ICMP_UNREACH_HOST;
1570 				goto bad;
1571 			}
1572 			memcpy(cp + off, &IA_SIN(ia)->sin_addr,
1573 			    sizeof(struct in_addr));
1574 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1575 			break;
1576 
1577 		case IPOPT_TS:
1578 			if (ipstealth && pass == 0)
1579 				break;
1580 			code = cp - (u_char *)ip;
1581 			if (optlen < 4 || optlen > 40) {
1582 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1583 				goto bad;
1584 			}
1585 			if ((off = cp[IPOPT_OFFSET]) < 5) {
1586 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1587 				goto bad;
1588 			}
1589 			if (off > optlen - (int)sizeof(int32_t)) {
1590 				cp[IPOPT_OFFSET + 1] += (1 << 4);
1591 				if ((cp[IPOPT_OFFSET + 1] & 0xf0) == 0) {
1592 					code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1593 					goto bad;
1594 				}
1595 				break;
1596 			}
1597 			off--;				/* 0 origin */
1598 			sin = (struct in_addr *)(cp + off);
1599 			switch (cp[IPOPT_OFFSET + 1] & 0x0f) {
1600 
1601 			case IPOPT_TS_TSONLY:
1602 				break;
1603 
1604 			case IPOPT_TS_TSANDADDR:
1605 				if (off + sizeof(n_time) +
1606 				    sizeof(struct in_addr) > optlen) {
1607 					code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1608 					goto bad;
1609 				}
1610 				ipaddr.sin_addr = dst;
1611 				ia = (INA)ifaof_ifpforaddr((SA)&ipaddr,
1612 							    m->m_pkthdr.rcvif);
1613 				if (ia == NULL)
1614 					continue;
1615 				memcpy(sin, &IA_SIN(ia)->sin_addr,
1616 				    sizeof(struct in_addr));
1617 				cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1618 				off += sizeof(struct in_addr);
1619 				break;
1620 
1621 			case IPOPT_TS_PRESPEC:
1622 				if (off + sizeof(n_time) +
1623 				    sizeof(struct in_addr) > optlen) {
1624 					code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1625 					goto bad;
1626 				}
1627 				memcpy(&ipaddr.sin_addr, sin,
1628 				    sizeof(struct in_addr));
1629 				if (ifa_ifwithaddr((SA)&ipaddr) == NULL)
1630 					continue;
1631 				cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1632 				off += sizeof(struct in_addr);
1633 				break;
1634 
1635 			default:
1636 				code = &cp[IPOPT_OFFSET + 1] - (u_char *)ip;
1637 				goto bad;
1638 			}
1639 			ntime = iptime();
1640 			memcpy(cp + off, &ntime, sizeof(n_time));
1641 			cp[IPOPT_OFFSET] += sizeof(n_time);
1642 		}
1643 	}
1644 	if (forward && ipforwarding) {
1645 		ip_forward(m, TRUE, next_hop);
1646 		return (1);
1647 	}
1648 	return (0);
1649 bad:
1650 	icmp_error(m, type, code, 0, 0);
1651 	ipstat.ips_badoptions++;
1652 	return (1);
1653 }
1654 
1655 /*
1656  * Given address of next destination (final or next hop),
1657  * return internet address info of interface to be used to get there.
1658  */
1659 struct in_ifaddr *
1660 ip_rtaddr(struct in_addr dst, struct route *ro0)
1661 {
1662 	struct route sro, *ro;
1663 	struct sockaddr_in *sin;
1664 	struct in_ifaddr *ia;
1665 
1666 	if (ro0 != NULL) {
1667 		ro = ro0;
1668 	} else {
1669 		bzero(&sro, sizeof(sro));
1670 		ro = &sro;
1671 	}
1672 
1673 	sin = (struct sockaddr_in *)&ro->ro_dst;
1674 
1675 	if (ro->ro_rt == NULL || dst.s_addr != sin->sin_addr.s_addr) {
1676 		if (ro->ro_rt != NULL) {
1677 			RTFREE(ro->ro_rt);
1678 			ro->ro_rt = NULL;
1679 		}
1680 		sin->sin_family = AF_INET;
1681 		sin->sin_len = sizeof *sin;
1682 		sin->sin_addr = dst;
1683 		rtalloc_ign(ro, RTF_PRCLONING);
1684 	}
1685 
1686 	if (ro->ro_rt == NULL)
1687 		return (NULL);
1688 
1689 	ia = ifatoia(ro->ro_rt->rt_ifa);
1690 
1691 	if (ro == &sro)
1692 		RTFREE(ro->ro_rt);
1693 	return ia;
1694 }
1695 
1696 /*
1697  * Save incoming source route for use in replies,
1698  * to be picked up later by ip_srcroute if the receiver is interested.
1699  */
1700 static void
1701 save_rte(struct mbuf *m, u_char *option, struct in_addr dst)
1702 {
1703 	struct m_tag *mtag;
1704 	struct ip_srcrt_opt *opt;
1705 	unsigned olen;
1706 
1707 	mtag = m_tag_get(PACKET_TAG_IPSRCRT, sizeof(*opt), MB_DONTWAIT);
1708 	if (mtag == NULL)
1709 		return;
1710 	opt = m_tag_data(mtag);
1711 
1712 	olen = option[IPOPT_OLEN];
1713 #ifdef DIAGNOSTIC
1714 	if (ipprintfs)
1715 		kprintf("save_rte: olen %d\n", olen);
1716 #endif
1717 	if (olen > sizeof(opt->ip_srcrt) - (1 + sizeof(dst))) {
1718 		m_tag_free(mtag);
1719 		return;
1720 	}
1721 	bcopy(option, opt->ip_srcrt.srcopt, olen);
1722 	opt->ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1723 	opt->ip_srcrt.dst = dst;
1724 	m_tag_prepend(m, mtag);
1725 }
1726 
1727 /*
1728  * Retrieve incoming source route for use in replies,
1729  * in the same form used by setsockopt.
1730  * The first hop is placed before the options, will be removed later.
1731  */
1732 struct mbuf *
1733 ip_srcroute(struct mbuf *m0)
1734 {
1735 	struct in_addr *p, *q;
1736 	struct mbuf *m;
1737 	struct m_tag *mtag;
1738 	struct ip_srcrt_opt *opt;
1739 
1740 	if (m0 == NULL)
1741 		return NULL;
1742 
1743 	mtag = m_tag_find(m0, PACKET_TAG_IPSRCRT, NULL);
1744 	if (mtag == NULL)
1745 		return NULL;
1746 	opt = m_tag_data(mtag);
1747 
1748 	if (opt->ip_nhops == 0)
1749 		return (NULL);
1750 	m = m_get(MB_DONTWAIT, MT_HEADER);
1751 	if (m == NULL)
1752 		return (NULL);
1753 
1754 #define	OPTSIZ	(sizeof(opt->ip_srcrt.nop) + sizeof(opt->ip_srcrt.srcopt))
1755 
1756 	/* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
1757 	m->m_len = opt->ip_nhops * sizeof(struct in_addr) +
1758 		   sizeof(struct in_addr) + OPTSIZ;
1759 #ifdef DIAGNOSTIC
1760 	if (ipprintfs) {
1761 		kprintf("ip_srcroute: nhops %d mlen %d",
1762 			opt->ip_nhops, m->m_len);
1763 	}
1764 #endif
1765 
1766 	/*
1767 	 * First save first hop for return route
1768 	 */
1769 	p = &opt->ip_srcrt.route[opt->ip_nhops - 1];
1770 	*(mtod(m, struct in_addr *)) = *p--;
1771 #ifdef DIAGNOSTIC
1772 	if (ipprintfs)
1773 		kprintf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr));
1774 #endif
1775 
1776 	/*
1777 	 * Copy option fields and padding (nop) to mbuf.
1778 	 */
1779 	opt->ip_srcrt.nop = IPOPT_NOP;
1780 	opt->ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
1781 	memcpy(mtod(m, caddr_t) + sizeof(struct in_addr), &opt->ip_srcrt.nop,
1782 	    OPTSIZ);
1783 	q = (struct in_addr *)(mtod(m, caddr_t) +
1784 	    sizeof(struct in_addr) + OPTSIZ);
1785 #undef OPTSIZ
1786 	/*
1787 	 * Record return path as an IP source route,
1788 	 * reversing the path (pointers are now aligned).
1789 	 */
1790 	while (p >= opt->ip_srcrt.route) {
1791 #ifdef DIAGNOSTIC
1792 		if (ipprintfs)
1793 			kprintf(" %x", ntohl(q->s_addr));
1794 #endif
1795 		*q++ = *p--;
1796 	}
1797 	/*
1798 	 * Last hop goes to final destination.
1799 	 */
1800 	*q = opt->ip_srcrt.dst;
1801 	m_tag_delete(m0, mtag);
1802 #ifdef DIAGNOSTIC
1803 	if (ipprintfs)
1804 		kprintf(" %x\n", ntohl(q->s_addr));
1805 #endif
1806 	return (m);
1807 }
1808 
1809 /*
1810  * Strip out IP options.
1811  */
1812 void
1813 ip_stripoptions(struct mbuf *m)
1814 {
1815 	int datalen;
1816 	struct ip *ip = mtod(m, struct ip *);
1817 	caddr_t opts;
1818 	int optlen;
1819 
1820 	optlen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
1821 	opts = (caddr_t)(ip + 1);
1822 	datalen = m->m_len - (sizeof(struct ip) + optlen);
1823 	bcopy(opts + optlen, opts, datalen);
1824 	m->m_len -= optlen;
1825 	if (m->m_flags & M_PKTHDR)
1826 		m->m_pkthdr.len -= optlen;
1827 	ip->ip_vhl = IP_MAKE_VHL(IPVERSION, sizeof(struct ip) >> 2);
1828 }
1829 
1830 u_char inetctlerrmap[PRC_NCMDS] = {
1831 	0,		0,		0,		0,
1832 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1833 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1834 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1835 	0,		0,		0,		0,
1836 	ENOPROTOOPT,	ECONNREFUSED
1837 };
1838 
1839 /*
1840  * Forward a packet.  If some error occurs return the sender
1841  * an icmp packet.  Note we can't always generate a meaningful
1842  * icmp message because icmp doesn't have a large enough repertoire
1843  * of codes and types.
1844  *
1845  * If not forwarding, just drop the packet.  This could be confusing
1846  * if ipforwarding was zero but some routing protocol was advancing
1847  * us as a gateway to somewhere.  However, we must let the routing
1848  * protocol deal with that.
1849  *
1850  * The using_srcrt parameter indicates whether the packet is being forwarded
1851  * via a source route.
1852  */
1853 void
1854 ip_forward(struct mbuf *m, boolean_t using_srcrt, struct sockaddr_in *next_hop)
1855 {
1856 	struct ip *ip = mtod(m, struct ip *);
1857 	struct rtentry *rt;
1858 	struct route fwd_ro;
1859 	int error, type = 0, code = 0, destmtu = 0;
1860 	struct mbuf *mcopy, *mtemp = NULL;
1861 	n_long dest;
1862 	struct in_addr pkt_dst;
1863 
1864 	dest = INADDR_ANY;
1865 	/*
1866 	 * Cache the destination address of the packet; this may be
1867 	 * changed by use of 'ipfw fwd'.
1868 	 */
1869 	pkt_dst = (next_hop != NULL) ? next_hop->sin_addr : ip->ip_dst;
1870 
1871 #ifdef DIAGNOSTIC
1872 	if (ipprintfs)
1873 		kprintf("forward: src %x dst %x ttl %x\n",
1874 		       ip->ip_src.s_addr, pkt_dst.s_addr, ip->ip_ttl);
1875 #endif
1876 
1877 	if (m->m_flags & (M_BCAST | M_MCAST) || !in_canforward(pkt_dst)) {
1878 		ipstat.ips_cantforward++;
1879 		m_freem(m);
1880 		return;
1881 	}
1882 	if (!ipstealth && ip->ip_ttl <= IPTTLDEC) {
1883 		icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0);
1884 		return;
1885 	}
1886 
1887 	bzero(&fwd_ro, sizeof(fwd_ro));
1888 	ip_rtaddr(pkt_dst, &fwd_ro);
1889 	if (fwd_ro.ro_rt == NULL) {
1890 		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0);
1891 		return;
1892 	}
1893 	rt = fwd_ro.ro_rt;
1894 
1895 	if (curthread->td_type == TD_TYPE_NETISR) {
1896 		/*
1897 		 * Save the IP header and at most 8 bytes of the payload,
1898 		 * in case we need to generate an ICMP message to the src.
1899 		 */
1900 		mtemp = ipforward_mtemp[mycpuid];
1901 		KASSERT((mtemp->m_flags & M_EXT) == 0 &&
1902 		    mtemp->m_data == mtemp->m_pktdat &&
1903 		    m_tag_first(mtemp) == NULL,
1904 		    ("ip_forward invalid mtemp1"));
1905 
1906 		if (!m_dup_pkthdr(mtemp, m, MB_DONTWAIT)) {
1907 			/*
1908 			 * It's probably ok if the pkthdr dup fails (because
1909 			 * the deep copy of the tag chain failed), but for now
1910 			 * be conservative and just discard the copy since
1911 			 * code below may some day want the tags.
1912 			 */
1913 			mtemp = NULL;
1914 		} else {
1915 			mtemp->m_type = m->m_type;
1916 			mtemp->m_len = imin((IP_VHL_HL(ip->ip_vhl) << 2) + 8,
1917 			    (int)ip->ip_len);
1918 			mtemp->m_pkthdr.len = mtemp->m_len;
1919 			m_copydata(m, 0, mtemp->m_len, mtod(mtemp, caddr_t));
1920 		}
1921 	}
1922 
1923 	if (!ipstealth)
1924 		ip->ip_ttl -= IPTTLDEC;
1925 
1926 	/*
1927 	 * If forwarding packet using same interface that it came in on,
1928 	 * perhaps should send a redirect to sender to shortcut a hop.
1929 	 * Only send redirect if source is sending directly to us,
1930 	 * and if packet was not source routed (or has any options).
1931 	 * Also, don't send redirect if forwarding using a default route
1932 	 * or a route modified by a redirect.
1933 	 */
1934 	if (rt->rt_ifp == m->m_pkthdr.rcvif &&
1935 	    !(rt->rt_flags & (RTF_DYNAMIC | RTF_MODIFIED)) &&
1936 	    satosin(rt_key(rt))->sin_addr.s_addr != INADDR_ANY &&
1937 	    ipsendredirects && !using_srcrt && next_hop == NULL) {
1938 		u_long src = ntohl(ip->ip_src.s_addr);
1939 		struct in_ifaddr *rt_ifa = (struct in_ifaddr *)rt->rt_ifa;
1940 
1941 		if (rt_ifa != NULL &&
1942 		    (src & rt_ifa->ia_subnetmask) == rt_ifa->ia_subnet) {
1943 			if (rt->rt_flags & RTF_GATEWAY)
1944 				dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
1945 			else
1946 				dest = pkt_dst.s_addr;
1947 			/*
1948 			 * Router requirements says to only send
1949 			 * host redirects.
1950 			 */
1951 			type = ICMP_REDIRECT;
1952 			code = ICMP_REDIRECT_HOST;
1953 #ifdef DIAGNOSTIC
1954 			if (ipprintfs)
1955 				kprintf("redirect (%d) to %x\n", code, dest);
1956 #endif
1957 		}
1958 	}
1959 
1960 	error = ip_output(m, NULL, &fwd_ro, IP_FORWARDING, NULL, NULL);
1961 	if (error == 0) {
1962 		ipstat.ips_forward++;
1963 		if (type == 0) {
1964 			if (mtemp)
1965 				ipflow_create(&fwd_ro, mtemp);
1966 			goto done;
1967 		} else {
1968 			ipstat.ips_redirectsent++;
1969 		}
1970 	} else {
1971 		ipstat.ips_cantforward++;
1972 	}
1973 
1974 	if (mtemp == NULL)
1975 		goto done;
1976 
1977 	/*
1978 	 * Errors that do not require generating ICMP message
1979 	 */
1980 	switch (error) {
1981 	case ENOBUFS:
1982 		/*
1983 		 * A router should not generate ICMP_SOURCEQUENCH as
1984 		 * required in RFC1812 Requirements for IP Version 4 Routers.
1985 		 * Source quench could be a big problem under DoS attacks,
1986 		 * or if the underlying interface is rate-limited.
1987 		 * Those who need source quench packets may re-enable them
1988 		 * via the net.inet.ip.sendsourcequench sysctl.
1989 		 */
1990 		if (!ip_sendsourcequench)
1991 			goto done;
1992 		break;
1993 
1994 	case EACCES:			/* ipfw denied packet */
1995 		goto done;
1996 	}
1997 
1998 	KASSERT((mtemp->m_flags & M_EXT) == 0 &&
1999 	    mtemp->m_data == mtemp->m_pktdat,
2000 	    ("ip_forward invalid mtemp2"));
2001 	mcopy = m_copym(mtemp, 0, mtemp->m_len, MB_DONTWAIT);
2002 	if (mcopy == NULL)
2003 		goto done;
2004 
2005 	/*
2006 	 * Send ICMP message.
2007 	 */
2008 	switch (error) {
2009 	case 0:				/* forwarded, but need redirect */
2010 		/* type, code set above */
2011 		break;
2012 
2013 	case ENETUNREACH:		/* shouldn't happen, checked above */
2014 	case EHOSTUNREACH:
2015 	case ENETDOWN:
2016 	case EHOSTDOWN:
2017 	default:
2018 		type = ICMP_UNREACH;
2019 		code = ICMP_UNREACH_HOST;
2020 		break;
2021 
2022 	case EMSGSIZE:
2023 		type = ICMP_UNREACH;
2024 		code = ICMP_UNREACH_NEEDFRAG;
2025 #ifdef IPSEC
2026 		/*
2027 		 * If the packet is routed over IPsec tunnel, tell the
2028 		 * originator the tunnel MTU.
2029 		 *	tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
2030 		 * XXX quickhack!!!
2031 		 */
2032 		if (fwd_ro.ro_rt != NULL) {
2033 			struct secpolicy *sp = NULL;
2034 			int ipsecerror;
2035 			int ipsechdr;
2036 			struct route *ro;
2037 
2038 			sp = ipsec4_getpolicybyaddr(mcopy,
2039 						    IPSEC_DIR_OUTBOUND,
2040 						    IP_FORWARDING,
2041 						    &ipsecerror);
2042 
2043 			if (sp == NULL)
2044 				destmtu = fwd_ro.ro_rt->rt_ifp->if_mtu;
2045 			else {
2046 				/* count IPsec header size */
2047 				ipsechdr = ipsec4_hdrsiz(mcopy,
2048 							 IPSEC_DIR_OUTBOUND,
2049 							 NULL);
2050 
2051 				/*
2052 				 * find the correct route for outer IPv4
2053 				 * header, compute tunnel MTU.
2054 				 *
2055 				 */
2056 				if (sp->req != NULL && sp->req->sav != NULL &&
2057 				    sp->req->sav->sah != NULL) {
2058 					ro = &sp->req->sav->sah->sa_route;
2059 					if (ro->ro_rt != NULL &&
2060 					    ro->ro_rt->rt_ifp != NULL) {
2061 						destmtu =
2062 						    ro->ro_rt->rt_ifp->if_mtu;
2063 						destmtu -= ipsechdr;
2064 					}
2065 				}
2066 
2067 				key_freesp(sp);
2068 			}
2069 		}
2070 #elif FAST_IPSEC
2071 		/*
2072 		 * If the packet is routed over IPsec tunnel, tell the
2073 		 * originator the tunnel MTU.
2074 		 *	tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
2075 		 * XXX quickhack!!!
2076 		 */
2077 		if (fwd_ro.ro_rt != NULL) {
2078 			struct secpolicy *sp = NULL;
2079 			int ipsecerror;
2080 			int ipsechdr;
2081 			struct route *ro;
2082 
2083 			sp = ipsec_getpolicybyaddr(mcopy,
2084 						   IPSEC_DIR_OUTBOUND,
2085 						   IP_FORWARDING,
2086 						   &ipsecerror);
2087 
2088 			if (sp == NULL)
2089 				destmtu = fwd_ro.ro_rt->rt_ifp->if_mtu;
2090 			else {
2091 				/* count IPsec header size */
2092 				ipsechdr = ipsec4_hdrsiz(mcopy,
2093 							 IPSEC_DIR_OUTBOUND,
2094 							 NULL);
2095 
2096 				/*
2097 				 * find the correct route for outer IPv4
2098 				 * header, compute tunnel MTU.
2099 				 */
2100 
2101 				if (sp->req != NULL &&
2102 				    sp->req->sav != NULL &&
2103 				    sp->req->sav->sah != NULL) {
2104 					ro = &sp->req->sav->sah->sa_route;
2105 					if (ro->ro_rt != NULL &&
2106 					    ro->ro_rt->rt_ifp != NULL) {
2107 						destmtu =
2108 						    ro->ro_rt->rt_ifp->if_mtu;
2109 						destmtu -= ipsechdr;
2110 					}
2111 				}
2112 
2113 				KEY_FREESP(&sp);
2114 			}
2115 		}
2116 #else /* !IPSEC && !FAST_IPSEC */
2117 		if (fwd_ro.ro_rt != NULL)
2118 			destmtu = fwd_ro.ro_rt->rt_ifp->if_mtu;
2119 #endif /*IPSEC*/
2120 		ipstat.ips_cantfrag++;
2121 		break;
2122 
2123 	case ENOBUFS:
2124 		type = ICMP_SOURCEQUENCH;
2125 		code = 0;
2126 		break;
2127 
2128 	case EACCES:			/* ipfw denied packet */
2129 		panic("ip_forward EACCES should not reach");
2130 	}
2131 	icmp_error(mcopy, type, code, dest, destmtu);
2132 done:
2133 	if (mtemp != NULL)
2134 		m_tag_delete_chain(mtemp);
2135 	if (fwd_ro.ro_rt != NULL)
2136 		RTFREE(fwd_ro.ro_rt);
2137 }
2138 
2139 void
2140 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
2141 	       struct mbuf *m)
2142 {
2143 	if (inp->inp_socket->so_options & SO_TIMESTAMP) {
2144 		struct timeval tv;
2145 
2146 		microtime(&tv);
2147 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
2148 		    SCM_TIMESTAMP, SOL_SOCKET);
2149 		if (*mp)
2150 			mp = &(*mp)->m_next;
2151 	}
2152 	if (inp->inp_flags & INP_RECVDSTADDR) {
2153 		*mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
2154 		    sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
2155 		if (*mp)
2156 			mp = &(*mp)->m_next;
2157 	}
2158 	if (inp->inp_flags & INP_RECVTTL) {
2159 		*mp = sbcreatecontrol((caddr_t) &ip->ip_ttl,
2160 		    sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
2161 		if (*mp)
2162 			mp = &(*mp)->m_next;
2163 	}
2164 #ifdef notyet
2165 	/* XXX
2166 	 * Moving these out of udp_input() made them even more broken
2167 	 * than they already were.
2168 	 */
2169 	/* options were tossed already */
2170 	if (inp->inp_flags & INP_RECVOPTS) {
2171 		*mp = sbcreatecontrol((caddr_t) opts_deleted_above,
2172 		    sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
2173 		if (*mp)
2174 			mp = &(*mp)->m_next;
2175 	}
2176 	/* ip_srcroute doesn't do what we want here, need to fix */
2177 	if (inp->inp_flags & INP_RECVRETOPTS) {
2178 		*mp = sbcreatecontrol((caddr_t) ip_srcroute(m),
2179 		    sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
2180 		if (*mp)
2181 			mp = &(*mp)->m_next;
2182 	}
2183 #endif
2184 	if (inp->inp_flags & INP_RECVIF) {
2185 		struct ifnet *ifp;
2186 		struct sdlbuf {
2187 			struct sockaddr_dl sdl;
2188 			u_char	pad[32];
2189 		} sdlbuf;
2190 		struct sockaddr_dl *sdp;
2191 		struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
2192 
2193 		if (((ifp = m->m_pkthdr.rcvif)) &&
2194 		    ((ifp->if_index != 0) && (ifp->if_index <= if_index))) {
2195 			sdp = IF_LLSOCKADDR(ifp);
2196 			/*
2197 			 * Change our mind and don't try copy.
2198 			 */
2199 			if ((sdp->sdl_family != AF_LINK) ||
2200 			    (sdp->sdl_len > sizeof(sdlbuf))) {
2201 				goto makedummy;
2202 			}
2203 			bcopy(sdp, sdl2, sdp->sdl_len);
2204 		} else {
2205 makedummy:
2206 			sdl2->sdl_len =
2207 			    offsetof(struct sockaddr_dl, sdl_data[0]);
2208 			sdl2->sdl_family = AF_LINK;
2209 			sdl2->sdl_index = 0;
2210 			sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
2211 		}
2212 		*mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len,
2213 			IP_RECVIF, IPPROTO_IP);
2214 		if (*mp)
2215 			mp = &(*mp)->m_next;
2216 	}
2217 }
2218 
2219 /*
2220  * XXX these routines are called from the upper part of the kernel.
2221  *
2222  * They could also be moved to ip_mroute.c, since all the RSVP
2223  *  handling is done there already.
2224  */
2225 int
2226 ip_rsvp_init(struct socket *so)
2227 {
2228 	if (so->so_type != SOCK_RAW ||
2229 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
2230 		return EOPNOTSUPP;
2231 
2232 	if (ip_rsvpd != NULL)
2233 		return EADDRINUSE;
2234 
2235 	ip_rsvpd = so;
2236 	/*
2237 	 * This may seem silly, but we need to be sure we don't over-increment
2238 	 * the RSVP counter, in case something slips up.
2239 	 */
2240 	if (!ip_rsvp_on) {
2241 		ip_rsvp_on = 1;
2242 		rsvp_on++;
2243 	}
2244 
2245 	return 0;
2246 }
2247 
2248 int
2249 ip_rsvp_done(void)
2250 {
2251 	ip_rsvpd = NULL;
2252 	/*
2253 	 * This may seem silly, but we need to be sure we don't over-decrement
2254 	 * the RSVP counter, in case something slips up.
2255 	 */
2256 	if (ip_rsvp_on) {
2257 		ip_rsvp_on = 0;
2258 		rsvp_on--;
2259 	}
2260 	return 0;
2261 }
2262 
2263 int
2264 rsvp_input(struct mbuf **mp, int *offp, int proto)
2265 {
2266 	struct mbuf *m = *mp;
2267 
2268 	*mp = NULL;
2269 
2270 	if (rsvp_input_p) { /* call the real one if loaded */
2271 		*mp = m;
2272 		rsvp_input_p(mp, offp, proto);
2273 		return(IPPROTO_DONE);
2274 	}
2275 
2276 	/* Can still get packets with rsvp_on = 0 if there is a local member
2277 	 * of the group to which the RSVP packet is addressed.  But in this
2278 	 * case we want to throw the packet away.
2279 	 */
2280 
2281 	if (!rsvp_on) {
2282 		m_freem(m);
2283 		return(IPPROTO_DONE);
2284 	}
2285 
2286 	if (ip_rsvpd != NULL) {
2287 		*mp = m;
2288 		rip_input(mp, offp, proto);
2289 		return(IPPROTO_DONE);
2290 	}
2291 	/* Drop the packet */
2292 	m_freem(m);
2293 	return(IPPROTO_DONE);
2294 }
2295