xref: /dflybsd-src/sys/netinet/ip_input.c (revision 0bb7d8c82a64940013681cf515d16f3e62eb7e3c)
1 /*
2  * Copyright (c) 2003, 2004 Jeffrey M. Hsu.  All rights reserved.
3  * Copyright (c) 2003, 2004 The DragonFly Project.  All rights reserved.
4  *
5  * This code is derived from software contributed to The DragonFly Project
6  * by Jeffrey M. Hsu.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of The DragonFly Project nor the names of its
17  *    contributors may be used to endorse or promote products derived
18  *    from this software without specific, prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
21  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
22  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
23  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE
24  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
25  * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
26  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
27  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
28  * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
29  * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
30  * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  */
33 
34 /*
35  * Copyright (c) 1982, 1986, 1988, 1993
36  *	The Regents of the University of California.  All rights reserved.
37  *
38  * Redistribution and use in source and binary forms, with or without
39  * modification, are permitted provided that the following conditions
40  * are met:
41  * 1. Redistributions of source code must retain the above copyright
42  *    notice, this list of conditions and the following disclaimer.
43  * 2. Redistributions in binary form must reproduce the above copyright
44  *    notice, this list of conditions and the following disclaimer in the
45  *    documentation and/or other materials provided with the distribution.
46  * 3. All advertising materials mentioning features or use of this software
47  *    must display the following acknowledgement:
48  *	This product includes software developed by the University of
49  *	California, Berkeley and its contributors.
50  * 4. Neither the name of the University nor the names of its contributors
51  *    may be used to endorse or promote products derived from this software
52  *    without specific prior written permission.
53  *
54  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
55  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
56  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
57  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
58  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
59  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
60  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
61  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
62  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
63  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
64  * SUCH DAMAGE.
65  *
66  *	@(#)ip_input.c	8.2 (Berkeley) 1/4/94
67  * $FreeBSD: src/sys/netinet/ip_input.c,v 1.130.2.52 2003/03/07 07:01:28 silby Exp $
68  * $DragonFly: src/sys/netinet/ip_input.c,v 1.115 2008/10/28 07:09:26 sephe Exp $
69  */
70 
71 #define	_IP_VHL
72 
73 #include "opt_bootp.h"
74 #include "opt_ipfw.h"
75 #include "opt_ipdn.h"
76 #include "opt_ipdivert.h"
77 #include "opt_ipfilter.h"
78 #include "opt_ipstealth.h"
79 #include "opt_ipsec.h"
80 
81 #include <sys/param.h>
82 #include <sys/systm.h>
83 #include <sys/mbuf.h>
84 #include <sys/malloc.h>
85 #include <sys/mpipe.h>
86 #include <sys/domain.h>
87 #include <sys/protosw.h>
88 #include <sys/socket.h>
89 #include <sys/time.h>
90 #include <sys/globaldata.h>
91 #include <sys/thread.h>
92 #include <sys/kernel.h>
93 #include <sys/syslog.h>
94 #include <sys/sysctl.h>
95 #include <sys/in_cksum.h>
96 #include <sys/lock.h>
97 
98 #include <sys/mplock2.h>
99 
100 #include <machine/stdarg.h>
101 
102 #include <net/if.h>
103 #include <net/if_types.h>
104 #include <net/if_var.h>
105 #include <net/if_dl.h>
106 #include <net/pfil.h>
107 #include <net/route.h>
108 #include <net/netisr.h>
109 
110 #include <netinet/in.h>
111 #include <netinet/in_systm.h>
112 #include <netinet/in_var.h>
113 #include <netinet/ip.h>
114 #include <netinet/in_pcb.h>
115 #include <netinet/ip_var.h>
116 #include <netinet/ip_icmp.h>
117 #include <netinet/ip_divert.h>
118 #include <netinet/ip_flow.h>
119 
120 #include <sys/thread2.h>
121 #include <sys/msgport2.h>
122 #include <net/netmsg2.h>
123 
124 #include <sys/socketvar.h>
125 
126 #include <net/ipfw/ip_fw.h>
127 #include <net/dummynet/ip_dummynet.h>
128 
129 #ifdef IPSEC
130 #include <netinet6/ipsec.h>
131 #include <netproto/key/key.h>
132 #endif
133 
134 #ifdef FAST_IPSEC
135 #include <netproto/ipsec/ipsec.h>
136 #include <netproto/ipsec/key.h>
137 #endif
138 
139 int rsvp_on = 0;
140 static int ip_rsvp_on;
141 struct socket *ip_rsvpd;
142 
143 int ip_mpsafe = 1;
144 TUNABLE_INT("net.inet.ip.mpsafe", &ip_mpsafe);
145 
146 int ipforwarding = 0;
147 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW,
148     &ipforwarding, 0, "Enable IP forwarding between interfaces");
149 
150 static int ipsendredirects = 1; /* XXX */
151 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW,
152     &ipsendredirects, 0, "Enable sending IP redirects");
153 
154 int ip_defttl = IPDEFTTL;
155 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW,
156     &ip_defttl, 0, "Maximum TTL on IP packets");
157 
158 static int ip_dosourceroute = 0;
159 SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW,
160     &ip_dosourceroute, 0, "Enable forwarding source routed IP packets");
161 
162 static int ip_acceptsourceroute = 0;
163 SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute,
164     CTLFLAG_RW, &ip_acceptsourceroute, 0,
165     "Enable accepting source routed IP packets");
166 
167 static int ip_keepfaith = 0;
168 SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW,
169     &ip_keepfaith, 0,
170     "Enable packet capture for FAITH IPv4->IPv6 translator daemon");
171 
172 static int nipq = 0;	/* total # of reass queues */
173 static int maxnipq;
174 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragpackets, CTLFLAG_RW,
175     &maxnipq, 0,
176     "Maximum number of IPv4 fragment reassembly queue entries");
177 
178 static int maxfragsperpacket;
179 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW,
180     &maxfragsperpacket, 0,
181     "Maximum number of IPv4 fragments allowed per packet");
182 
183 static int ip_sendsourcequench = 0;
184 SYSCTL_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW,
185     &ip_sendsourcequench, 0,
186     "Enable the transmission of source quench packets");
187 
188 int ip_do_randomid = 1;
189 SYSCTL_INT(_net_inet_ip, OID_AUTO, random_id, CTLFLAG_RW,
190     &ip_do_randomid, 0,
191     "Assign random ip_id values");
192 /*
193  * XXX - Setting ip_checkinterface mostly implements the receive side of
194  * the Strong ES model described in RFC 1122, but since the routing table
195  * and transmit implementation do not implement the Strong ES model,
196  * setting this to 1 results in an odd hybrid.
197  *
198  * XXX - ip_checkinterface currently must be disabled if you use ipnat
199  * to translate the destination address to another local interface.
200  *
201  * XXX - ip_checkinterface must be disabled if you add IP aliases
202  * to the loopback interface instead of the interface where the
203  * packets for those addresses are received.
204  */
205 static int ip_checkinterface = 0;
206 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW,
207     &ip_checkinterface, 0, "Verify packet arrives on correct interface");
208 
209 static struct lwkt_token ipq_token = LWKT_TOKEN_MP_INITIALIZER(ipq_token);
210 
211 #ifdef DIAGNOSTIC
212 static int ipprintfs = 0;
213 #endif
214 
215 extern	int udp_mpsafe_proto;
216 extern	int tcp_mpsafe_proto;
217 
218 extern	struct domain inetdomain;
219 extern	struct protosw inetsw[];
220 u_char	ip_protox[IPPROTO_MAX];
221 struct	in_ifaddrhead in_ifaddrheads[MAXCPU];	/* first inet address */
222 struct	in_ifaddrhashhead *in_ifaddrhashtbls[MAXCPU];
223 						/* inet addr hash table */
224 u_long	in_ifaddrhmask;				/* mask for hash table */
225 
226 struct ip_stats ipstats_percpu[MAXCPU];
227 #ifdef SMP
228 static int
229 sysctl_ipstats(SYSCTL_HANDLER_ARGS)
230 {
231 	int cpu, error = 0;
232 
233 	for (cpu = 0; cpu < ncpus; ++cpu) {
234 		if ((error = SYSCTL_OUT(req, &ipstats_percpu[cpu],
235 					sizeof(struct ip_stats))))
236 			break;
237 		if ((error = SYSCTL_IN(req, &ipstats_percpu[cpu],
238 				       sizeof(struct ip_stats))))
239 			break;
240 	}
241 
242 	return (error);
243 }
244 SYSCTL_PROC(_net_inet_ip, IPCTL_STATS, stats, (CTLTYPE_OPAQUE | CTLFLAG_RW),
245     0, 0, sysctl_ipstats, "S,ip_stats", "IP statistics");
246 #else
247 SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW,
248     &ipstat, ip_stats, "IP statistics");
249 #endif
250 
251 /* Packet reassembly stuff */
252 #define	IPREASS_NHASH_LOG2	6
253 #define	IPREASS_NHASH		(1 << IPREASS_NHASH_LOG2)
254 #define	IPREASS_HMASK		(IPREASS_NHASH - 1)
255 #define	IPREASS_HASH(x,y)						\
256     (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK)
257 
258 static struct ipq ipq[IPREASS_NHASH];
259 
260 #ifdef IPCTL_DEFMTU
261 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
262     &ip_mtu, 0, "Default MTU");
263 #endif
264 
265 #ifdef IPSTEALTH
266 static int ipstealth = 0;
267 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW, &ipstealth, 0, "");
268 #else
269 static const int ipstealth = 0;
270 #endif
271 
272 struct mbuf *(*ip_divert_p)(struct mbuf *, int, int);
273 
274 struct pfil_head inet_pfil_hook;
275 
276 /*
277  * struct ip_srcrt_opt is used to store packet state while it travels
278  * through the stack.
279  *
280  * XXX Note that the code even makes assumptions on the size and
281  * alignment of fields inside struct ip_srcrt so e.g. adding some
282  * fields will break the code.  This needs to be fixed.
283  *
284  * We need to save the IP options in case a protocol wants to respond
285  * to an incoming packet over the same route if the packet got here
286  * using IP source routing.  This allows connection establishment and
287  * maintenance when the remote end is on a network that is not known
288  * to us.
289  */
290 struct ip_srcrt {
291 	struct	in_addr dst;			/* final destination */
292 	char	nop;				/* one NOP to align */
293 	char	srcopt[IPOPT_OFFSET + 1];	/* OPTVAL, OLEN and OFFSET */
294 	struct	in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)];
295 };
296 
297 struct ip_srcrt_opt {
298 	int		ip_nhops;
299 	struct ip_srcrt	ip_srcrt;
300 };
301 
302 static MALLOC_DEFINE(M_IPQ, "ipq", "IP Fragment Management");
303 static struct malloc_pipe ipq_mpipe;
304 
305 static void		save_rte(struct mbuf *, u_char *, struct in_addr);
306 static int		ip_dooptions(struct mbuf *m, int, struct sockaddr_in *);
307 static void		ip_freef(struct ipq *);
308 static void		ip_input_handler(struct netmsg *);
309 
310 /*
311  * IP initialization: fill in IP protocol switch table.
312  * All protocols not implemented in kernel go to raw IP protocol handler.
313  */
314 void
315 ip_init(void)
316 {
317 	struct protosw *pr;
318 	uint32_t flags;
319 	int i;
320 #ifdef SMP
321 	int cpu;
322 #endif
323 
324 	/*
325 	 * Make sure we can handle a reasonable number of fragments but
326 	 * cap it at 4000 (XXX).
327 	 */
328 	mpipe_init(&ipq_mpipe, M_IPQ, sizeof(struct ipq),
329 		    IFQ_MAXLEN, 4000, 0, NULL);
330 	for (i = 0; i < ncpus; ++i) {
331 		TAILQ_INIT(&in_ifaddrheads[i]);
332 		in_ifaddrhashtbls[i] =
333 			hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask);
334 	}
335 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
336 	if (pr == NULL)
337 		panic("ip_init");
338 	for (i = 0; i < IPPROTO_MAX; i++)
339 		ip_protox[i] = pr - inetsw;
340 	for (pr = inetdomain.dom_protosw;
341 	     pr < inetdomain.dom_protoswNPROTOSW; pr++) {
342 		if (pr->pr_domain->dom_family == PF_INET && pr->pr_protocol) {
343 			if (pr->pr_protocol != IPPROTO_RAW)
344 				ip_protox[pr->pr_protocol] = pr - inetsw;
345 
346 			/* XXX */
347 			switch (pr->pr_protocol) {
348 			case IPPROTO_TCP:
349 				if (tcp_mpsafe_proto)
350 					pr->pr_flags |= PR_MPSAFE;
351 				break;
352 
353 			case IPPROTO_UDP:
354 				if (udp_mpsafe_proto)
355 					pr->pr_flags |= PR_MPSAFE;
356 				break;
357 			}
358 		}
359 	}
360 
361 	inet_pfil_hook.ph_type = PFIL_TYPE_AF;
362 	inet_pfil_hook.ph_af = AF_INET;
363 	if ((i = pfil_head_register(&inet_pfil_hook)) != 0) {
364 		kprintf("%s: WARNING: unable to register pfil hook, "
365 			"error %d\n", __func__, i);
366 	}
367 
368 	for (i = 0; i < IPREASS_NHASH; i++)
369 		ipq[i].next = ipq[i].prev = &ipq[i];
370 
371 	maxnipq = nmbclusters / 32;
372 	maxfragsperpacket = 16;
373 
374 	ip_id = time_second & 0xffff;
375 
376 	/*
377 	 * Initialize IP statistics counters for each CPU.
378 	 *
379 	 */
380 #ifdef SMP
381 	for (cpu = 0; cpu < ncpus; ++cpu) {
382 		bzero(&ipstats_percpu[cpu], sizeof(struct ip_stats));
383 	}
384 #else
385 	bzero(&ipstat, sizeof(struct ip_stats));
386 #endif
387 
388 #if defined(IPSEC) || defined(FAST_IPSEC)
389 	/* XXX IPSEC is not MPSAFE yet */
390 	flags = NETISR_FLAG_NOTMPSAFE;
391 #else
392 	if (ip_mpsafe) {
393 		kprintf("ip: MPSAFE\n");
394 		flags = NETISR_FLAG_MPSAFE;
395 	} else {
396 		flags = NETISR_FLAG_NOTMPSAFE;
397 	}
398 #endif
399 	netisr_register(NETISR_IP, ip_mport_in, ip_mport_pktinfo,
400 			ip_input_handler, flags);
401 }
402 
403 /* Do transport protocol processing. */
404 static void
405 transport_processing_oncpu(struct mbuf *m, int hlen, struct ip *ip)
406 {
407 	const struct protosw *pr = &inetsw[ip_protox[ip->ip_p]];
408 
409 	/*
410 	 * Switch out to protocol's input routine.
411 	 */
412 	PR_GET_MPLOCK(pr);
413 	pr->pr_input(m, hlen, ip->ip_p);
414 	PR_REL_MPLOCK(pr);
415 }
416 
417 static void
418 transport_processing_handler(netmsg_t netmsg)
419 {
420 	struct netmsg_packet *pmsg = (struct netmsg_packet *)netmsg;
421 	struct ip *ip;
422 	int hlen;
423 
424 	ip = mtod(pmsg->nm_packet, struct ip *);
425 	hlen = pmsg->nm_netmsg.nm_lmsg.u.ms_result;
426 
427 	transport_processing_oncpu(pmsg->nm_packet, hlen, ip);
428 	/* netmsg was embedded in the mbuf, do not reply! */
429 }
430 
431 static void
432 ip_input_handler(struct netmsg *msg0)
433 {
434 	struct mbuf *m = ((struct netmsg_packet *)msg0)->nm_packet;
435 
436 	ip_input(m);
437 	/* msg0 was embedded in the mbuf, do not reply! */
438 }
439 
440 /*
441  * IP input routine.  Checksum and byte swap header.  If fragmented
442  * try to reassemble.  Process options.  Pass to next level.
443  */
444 void
445 ip_input(struct mbuf *m)
446 {
447 	struct ip *ip;
448 	struct in_ifaddr *ia = NULL;
449 	struct in_ifaddr_container *iac;
450 	int hlen, checkif;
451 	u_short sum;
452 	struct in_addr pkt_dst;
453 	boolean_t using_srcrt = FALSE;		/* forward (by PFIL_HOOKS) */
454 	struct in_addr odst;			/* original dst address(NAT) */
455 	struct m_tag *mtag;
456 	struct sockaddr_in *next_hop = NULL;
457 	lwkt_port_t port;
458 #ifdef FAST_IPSEC
459 	struct tdb_ident *tdbi;
460 	struct secpolicy *sp;
461 	int error;
462 #endif
463 
464 	M_ASSERTPKTHDR(m);
465 
466 	/*
467 	 * This does necessary pullups and figures out the protocol
468 	 * port.  If the packet is really badly formed it will blow
469 	 * it away and return NULL.
470 	 *
471 	 * We do not necessarily make use of the port (forwarding,
472 	 * defragmentation, etc).
473 	 */
474 	port = ip_mport(&m, IP_MPORT_IN);
475 	if (port == NULL)
476 		return;
477 	ip = mtod(m, struct ip *);
478 
479 	/*
480 	 * Pull out certain tags
481 	 */
482 	if (m->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) {
483 		/* Next hop */
484 		mtag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL);
485 		KKASSERT(mtag != NULL);
486 		next_hop = m_tag_data(mtag);
487 	}
488 
489 	if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
490 		/* dummynet already filtered us */
491 		ip = mtod(m, struct ip *);
492 		hlen = IP_VHL_HL(ip->ip_vhl) << 2;
493 		goto iphack;
494 	}
495 
496 	ipstat.ips_total++;
497 
498 	/* length checks already done in ip_mport() */
499 	KASSERT(m->m_len >= sizeof(struct ip), ("IP header not in one mbuf"));
500 
501 	if (IP_VHL_V(ip->ip_vhl) != IPVERSION) {
502 		ipstat.ips_badvers++;
503 		goto bad;
504 	}
505 
506 	hlen = IP_VHL_HL(ip->ip_vhl) << 2;
507 	/* length checks already done in ip_mport() */
508 	KASSERT(hlen >= sizeof(struct ip), ("IP header len too small"));
509 	KASSERT(m->m_len >= hlen, ("complete IP header not in one mbuf"));
510 
511 	/* 127/8 must not appear on wire - RFC1122 */
512 	if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
513 	    (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
514 		if (!(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK)) {
515 			ipstat.ips_badaddr++;
516 			goto bad;
517 		}
518 	}
519 
520 	if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
521 		sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
522 	} else {
523 		if (hlen == sizeof(struct ip))
524 			sum = in_cksum_hdr(ip);
525 		else
526 			sum = in_cksum(m, hlen);
527 	}
528 	if (sum != 0) {
529 		ipstat.ips_badsum++;
530 		goto bad;
531 	}
532 
533 #ifdef ALTQ
534 	if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) {
535 		/* packet is dropped by traffic conditioner */
536 		return;
537 	}
538 #endif
539 	/*
540 	 * Convert fields to host representation.
541 	 */
542 	ip->ip_len = ntohs(ip->ip_len);
543 	ip->ip_off = ntohs(ip->ip_off);
544 
545 	/* length checks already done in ip_mport() */
546 	KASSERT(ip->ip_len >= hlen, ("total length less then header length"));
547 	KASSERT(m->m_pkthdr.len >= ip->ip_len, ("mbuf too short"));
548 
549 	/*
550 	 * Trim mbufs if longer than the IP header would have us expect.
551 	 */
552 	if (m->m_pkthdr.len > ip->ip_len) {
553 		if (m->m_len == m->m_pkthdr.len) {
554 			m->m_len = ip->ip_len;
555 			m->m_pkthdr.len = ip->ip_len;
556 		} else {
557 			m_adj(m, ip->ip_len - m->m_pkthdr.len);
558 		}
559 	}
560 #if defined(IPSEC) && !defined(IPSEC_FILTERGIF)
561 	/*
562 	 * Bypass packet filtering for packets from a tunnel (gif).
563 	 */
564 	if (ipsec_gethist(m, NULL))
565 		goto pass;
566 #endif
567 
568 	/*
569 	 * IpHack's section.
570 	 * Right now when no processing on packet has done
571 	 * and it is still fresh out of network we do our black
572 	 * deals with it.
573 	 * - Firewall: deny/allow/divert
574 	 * - Xlate: translate packet's addr/port (NAT).
575 	 * - Pipe: pass pkt through dummynet.
576 	 * - Wrap: fake packet's addr/port <unimpl.>
577 	 * - Encapsulate: put it in another IP and send out. <unimp.>
578 	 */
579 
580 iphack:
581 	/*
582 	 * If we've been forwarded from the output side, then
583 	 * skip the firewall a second time
584 	 */
585 	if (next_hop != NULL)
586 		goto ours;
587 
588 	/* No pfil hooks */
589 	if (!pfil_has_hooks(&inet_pfil_hook)) {
590 		if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
591 			/*
592 			 * Strip dummynet tags from stranded packets
593 			 */
594 			mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
595 			KKASSERT(mtag != NULL);
596 			m_tag_delete(m, mtag);
597 			m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED;
598 		}
599 		goto pass;
600 	}
601 
602 	/*
603 	 * Run through list of hooks for input packets.
604 	 *
605 	 * NB: Beware of the destination address changing (e.g.
606 	 *     by NAT rewriting). When this happens, tell
607 	 *     ip_forward to do the right thing.
608 	 */
609 	odst = ip->ip_dst;
610 	if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif, PFIL_IN))
611 		return;
612 	if (m == NULL)	/* consumed by filter */
613 		return;
614 	ip = mtod(m, struct ip *);
615 	hlen = IP_VHL_HL(ip->ip_vhl) << 2;
616 	using_srcrt = (odst.s_addr != ip->ip_dst.s_addr);
617 
618 	if (m->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) {
619 		mtag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL);
620 		KKASSERT(mtag != NULL);
621 		next_hop = m_tag_data(mtag);
622 	}
623 	if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
624 		ip_dn_queue(m);
625 		return;
626 	}
627 	if (m->m_pkthdr.fw_flags & FW_MBUF_REDISPATCH) {
628 		m->m_pkthdr.fw_flags &= ~FW_MBUF_REDISPATCH;
629 	}
630 pass:
631 	/*
632 	 * Process options and, if not destined for us,
633 	 * ship it on.  ip_dooptions returns 1 when an
634 	 * error was detected (causing an icmp message
635 	 * to be sent and the original packet to be freed).
636 	 */
637 	if (hlen > sizeof(struct ip) && ip_dooptions(m, 0, next_hop))
638 		return;
639 
640 	/* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
641 	 * matter if it is destined to another node, or whether it is
642 	 * a multicast one, RSVP wants it! and prevents it from being forwarded
643 	 * anywhere else. Also checks if the rsvp daemon is running before
644 	 * grabbing the packet.
645 	 */
646 	if (rsvp_on && ip->ip_p == IPPROTO_RSVP)
647 		goto ours;
648 
649 	/*
650 	 * Check our list of addresses, to see if the packet is for us.
651 	 * If we don't have any addresses, assume any unicast packet
652 	 * we receive might be for us (and let the upper layers deal
653 	 * with it).
654 	 */
655 	if (TAILQ_EMPTY(&in_ifaddrheads[mycpuid]) &&
656 	    !(m->m_flags & (M_MCAST | M_BCAST)))
657 		goto ours;
658 
659 	/*
660 	 * Cache the destination address of the packet; this may be
661 	 * changed by use of 'ipfw fwd'.
662 	 */
663 	pkt_dst = next_hop ? next_hop->sin_addr : ip->ip_dst;
664 
665 	/*
666 	 * Enable a consistency check between the destination address
667 	 * and the arrival interface for a unicast packet (the RFC 1122
668 	 * strong ES model) if IP forwarding is disabled and the packet
669 	 * is not locally generated and the packet is not subject to
670 	 * 'ipfw fwd'.
671 	 *
672 	 * XXX - Checking also should be disabled if the destination
673 	 * address is ipnat'ed to a different interface.
674 	 *
675 	 * XXX - Checking is incompatible with IP aliases added
676 	 * to the loopback interface instead of the interface where
677 	 * the packets are received.
678 	 */
679 	checkif = ip_checkinterface &&
680 		  !ipforwarding &&
681 		  m->m_pkthdr.rcvif != NULL &&
682 		  !(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) &&
683 		  next_hop == NULL;
684 
685 	/*
686 	 * Check for exact addresses in the hash bucket.
687 	 */
688 	LIST_FOREACH(iac, INADDR_HASH(pkt_dst.s_addr), ia_hash) {
689 		ia = iac->ia;
690 
691 		/*
692 		 * If the address matches, verify that the packet
693 		 * arrived via the correct interface if checking is
694 		 * enabled.
695 		 */
696 		if (IA_SIN(ia)->sin_addr.s_addr == pkt_dst.s_addr &&
697 		    (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif))
698 			goto ours;
699 	}
700 	ia = NULL;
701 
702 	/*
703 	 * Check for broadcast addresses.
704 	 *
705 	 * Only accept broadcast packets that arrive via the matching
706 	 * interface.  Reception of forwarded directed broadcasts would
707 	 * be handled via ip_forward() and ether_output() with the loopback
708 	 * into the stack for SIMPLEX interfaces handled by ether_output().
709 	 */
710 	if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
711 		struct ifaddr_container *ifac;
712 
713 		TAILQ_FOREACH(ifac, &m->m_pkthdr.rcvif->if_addrheads[mycpuid],
714 			      ifa_link) {
715 			struct ifaddr *ifa = ifac->ifa;
716 
717 			if (ifa->ifa_addr == NULL) /* shutdown/startup race */
718 				continue;
719 			if (ifa->ifa_addr->sa_family != AF_INET)
720 				continue;
721 			ia = ifatoia(ifa);
722 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
723 								pkt_dst.s_addr)
724 				goto ours;
725 			if (ia->ia_netbroadcast.s_addr == pkt_dst.s_addr)
726 				goto ours;
727 #ifdef BOOTP_COMPAT
728 			if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY)
729 				goto ours;
730 #endif
731 		}
732 	}
733 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
734 		struct in_multi *inm;
735 
736 		/* XXX Multicast is not MPSAFE yet */
737 		get_mplock();
738 
739 		if (ip_mrouter != NULL) {
740 			/*
741 			 * If we are acting as a multicast router, all
742 			 * incoming multicast packets are passed to the
743 			 * kernel-level multicast forwarding function.
744 			 * The packet is returned (relatively) intact; if
745 			 * ip_mforward() returns a non-zero value, the packet
746 			 * must be discarded, else it may be accepted below.
747 			 */
748 			if (ip_mforward != NULL &&
749 			    ip_mforward(ip, m->m_pkthdr.rcvif, m, NULL) != 0) {
750 				rel_mplock();
751 				ipstat.ips_cantforward++;
752 				m_freem(m);
753 				return;
754 			}
755 
756 			/*
757 			 * The process-level routing daemon needs to receive
758 			 * all multicast IGMP packets, whether or not this
759 			 * host belongs to their destination groups.
760 			 */
761 			if (ip->ip_p == IPPROTO_IGMP) {
762 				rel_mplock();
763 				goto ours;
764 			}
765 			ipstat.ips_forward++;
766 		}
767 		/*
768 		 * See if we belong to the destination multicast group on the
769 		 * arrival interface.
770 		 */
771 		IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
772 		if (inm == NULL) {
773 			rel_mplock();
774 			ipstat.ips_notmember++;
775 			m_freem(m);
776 			return;
777 		}
778 
779 		rel_mplock();
780 		goto ours;
781 	}
782 	if (ip->ip_dst.s_addr == INADDR_BROADCAST)
783 		goto ours;
784 	if (ip->ip_dst.s_addr == INADDR_ANY)
785 		goto ours;
786 
787 	/*
788 	 * FAITH(Firewall Aided Internet Translator)
789 	 */
790 	if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
791 		if (ip_keepfaith) {
792 			if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP)
793 				goto ours;
794 		}
795 		m_freem(m);
796 		return;
797 	}
798 
799 	/*
800 	 * Not for us; forward if possible and desirable.
801 	 */
802 	if (!ipforwarding) {
803 		ipstat.ips_cantforward++;
804 		m_freem(m);
805 	} else {
806 #ifdef IPSEC
807 		/*
808 		 * Enforce inbound IPsec SPD.
809 		 */
810 		if (ipsec4_in_reject(m, NULL)) {
811 			ipsecstat.in_polvio++;
812 			goto bad;
813 		}
814 #endif
815 #ifdef FAST_IPSEC
816 		mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
817 		crit_enter();
818 		if (mtag != NULL) {
819 			tdbi = (struct tdb_ident *)m_tag_data(mtag);
820 			sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
821 		} else {
822 			sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
823 						   IP_FORWARDING, &error);
824 		}
825 		if (sp == NULL) {	/* NB: can happen if error */
826 			crit_exit();
827 			/*XXX error stat???*/
828 			DPRINTF(("ip_input: no SP for forwarding\n"));	/*XXX*/
829 			goto bad;
830 		}
831 
832 		/*
833 		 * Check security policy against packet attributes.
834 		 */
835 		error = ipsec_in_reject(sp, m);
836 		KEY_FREESP(&sp);
837 		crit_exit();
838 		if (error) {
839 			ipstat.ips_cantforward++;
840 			goto bad;
841 		}
842 #endif
843 		ip_forward(m, using_srcrt, next_hop);
844 	}
845 	return;
846 
847 ours:
848 
849 	/*
850 	 * IPSTEALTH: Process non-routing options only
851 	 * if the packet is destined for us.
852 	 */
853 	if (ipstealth &&
854 	    hlen > sizeof(struct ip) &&
855 	    ip_dooptions(m, 1, next_hop))
856 		return;
857 
858 	/* Count the packet in the ip address stats */
859 	if (ia != NULL) {
860 		ia->ia_ifa.if_ipackets++;
861 		ia->ia_ifa.if_ibytes += m->m_pkthdr.len;
862 	}
863 
864 	/*
865 	 * If offset or IP_MF are set, must reassemble.
866 	 * Otherwise, nothing need be done.
867 	 * (We could look in the reassembly queue to see
868 	 * if the packet was previously fragmented,
869 	 * but it's not worth the time; just let them time out.)
870 	 */
871 	if (ip->ip_off & (IP_MF | IP_OFFMASK)) {
872 		/*
873 		 * Attempt reassembly; if it succeeds, proceed.
874 		 * ip_reass() will return a different mbuf.
875 		 */
876 		m = ip_reass(m);
877 		if (m == NULL)
878 			return;
879 		ip = mtod(m, struct ip *);
880 
881 		/* Get the header length of the reassembled packet */
882 		hlen = IP_VHL_HL(ip->ip_vhl) << 2;
883 	} else {
884 		ip->ip_len -= hlen;
885 	}
886 
887 #ifdef IPSEC
888 	/*
889 	 * enforce IPsec policy checking if we are seeing last header.
890 	 * note that we do not visit this with protocols with pcb layer
891 	 * code - like udp/tcp/raw ip.
892 	 */
893 	if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) &&
894 	    ipsec4_in_reject(m, NULL)) {
895 		ipsecstat.in_polvio++;
896 		goto bad;
897 	}
898 #endif
899 #if FAST_IPSEC
900 	/*
901 	 * enforce IPsec policy checking if we are seeing last header.
902 	 * note that we do not visit this with protocols with pcb layer
903 	 * code - like udp/tcp/raw ip.
904 	 */
905 	if (inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) {
906 		/*
907 		 * Check if the packet has already had IPsec processing
908 		 * done.  If so, then just pass it along.  This tag gets
909 		 * set during AH, ESP, etc. input handling, before the
910 		 * packet is returned to the ip input queue for delivery.
911 		 */
912 		mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
913 		crit_enter();
914 		if (mtag != NULL) {
915 			tdbi = (struct tdb_ident *)m_tag_data(mtag);
916 			sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
917 		} else {
918 			sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
919 						   IP_FORWARDING, &error);
920 		}
921 		if (sp != NULL) {
922 			/*
923 			 * Check security policy against packet attributes.
924 			 */
925 			error = ipsec_in_reject(sp, m);
926 			KEY_FREESP(&sp);
927 		} else {
928 			/* XXX error stat??? */
929 			error = EINVAL;
930 DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/
931 			goto bad;
932 		}
933 		crit_exit();
934 		if (error)
935 			goto bad;
936 	}
937 #endif /* FAST_IPSEC */
938 
939 	/*
940 	 * NOTE: ip_len is now in host form and adjusted down by hlen for
941 	 *	 protocol processing.
942 	 *
943 	 * We must forward the packet to the correct protocol thread if
944 	 * we are not already in it.
945 	 */
946 	ipstat.ips_delivered++;
947 
948 	if (port != &curthread->td_msgport) {
949 		struct netmsg_packet *pmsg;
950 
951 		pmsg = &m->m_hdr.mh_netmsg;
952 		netmsg_init(&pmsg->nm_netmsg, NULL, &netisr_apanic_rport,
953 			    MSGF_MPSAFE, transport_processing_handler);
954 		pmsg->nm_packet = m;
955 		pmsg->nm_netmsg.nm_lmsg.u.ms_result = hlen;
956 
957 		lwkt_sendmsg(port, &pmsg->nm_netmsg.nm_lmsg);
958 	} else {
959 		transport_processing_oncpu(m, hlen, ip);
960 	}
961 	return;
962 
963 bad:
964 	m_freem(m);
965 }
966 
967 /*
968  * Take incoming datagram fragment and try to reassemble it into
969  * whole datagram.  If a chain for reassembly of this datagram already
970  * exists, then it is given as fp; otherwise have to make a chain.
971  */
972 struct mbuf *
973 ip_reass(struct mbuf *m)
974 {
975 	struct ip *ip = mtod(m, struct ip *);
976 	struct mbuf *p = NULL, *q, *nq;
977 	struct mbuf *n;
978 	struct ipq *fp = NULL;
979 	int hlen = IP_VHL_HL(ip->ip_vhl) << 2;
980 	int i, next;
981 	u_short sum;
982 
983 	/* If maxnipq is 0, never accept fragments. */
984 	if (maxnipq == 0) {
985 		ipstat.ips_fragments++;
986 		ipstat.ips_fragdropped++;
987 		m_freem(m);
988 		return NULL;
989 	}
990 
991 	sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
992 	/*
993 	 * Look for queue of fragments of this datagram.
994 	 */
995 	lwkt_gettoken(&ipq_token);
996 	for (fp = ipq[sum].next; fp != &ipq[sum]; fp = fp->next) {
997 		if (ip->ip_id == fp->ipq_id &&
998 		    ip->ip_src.s_addr == fp->ipq_src.s_addr &&
999 		    ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
1000 		    ip->ip_p == fp->ipq_p)
1001 			goto found;
1002 	}
1003 
1004 	fp = NULL;
1005 
1006 	/*
1007 	 * Enforce upper bound on number of fragmented packets
1008 	 * for which we attempt reassembly;
1009 	 * If maxnipq is -1, accept all fragments without limitation.
1010 	 */
1011 	if (nipq > maxnipq && maxnipq > 0) {
1012 		/*
1013 		 * drop something from the tail of the current queue
1014 		 * before proceeding further
1015 		 */
1016 		if (ipq[sum].prev == &ipq[sum]) {   /* gak */
1017 			for (i = 0; i < IPREASS_NHASH; i++) {
1018 				if (ipq[i].prev != &ipq[i]) {
1019 					ipstat.ips_fragtimeout +=
1020 					    ipq[i].prev->ipq_nfrags;
1021 					ip_freef(ipq[i].prev);
1022 					break;
1023 				}
1024 			}
1025 		} else {
1026 			ipstat.ips_fragtimeout +=
1027 			    ipq[sum].prev->ipq_nfrags;
1028 			ip_freef(ipq[sum].prev);
1029 		}
1030 	}
1031 found:
1032 	/*
1033 	 * Adjust ip_len to not reflect header,
1034 	 * convert offset of this to bytes.
1035 	 */
1036 	ip->ip_len -= hlen;
1037 	if (ip->ip_off & IP_MF) {
1038 		/*
1039 		 * Make sure that fragments have a data length
1040 		 * that's a non-zero multiple of 8 bytes.
1041 		 */
1042 		if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
1043 			ipstat.ips_toosmall++; /* XXX */
1044 			m_freem(m);
1045 			goto done;
1046 		}
1047 		m->m_flags |= M_FRAG;
1048 	} else {
1049 		m->m_flags &= ~M_FRAG;
1050 	}
1051 	ip->ip_off <<= 3;
1052 
1053 	ipstat.ips_fragments++;
1054 	m->m_pkthdr.header = ip;
1055 
1056 	/*
1057 	 * If the hardware has not done csum over this fragment
1058 	 * then csum_data is not valid at all.
1059 	 */
1060 	if ((m->m_pkthdr.csum_flags & (CSUM_FRAG_NOT_CHECKED | CSUM_DATA_VALID))
1061 	    == (CSUM_FRAG_NOT_CHECKED | CSUM_DATA_VALID)) {
1062 		m->m_pkthdr.csum_data = 0;
1063 		m->m_pkthdr.csum_flags &= ~(CSUM_DATA_VALID | CSUM_PSEUDO_HDR);
1064 	}
1065 
1066 	/*
1067 	 * Presence of header sizes in mbufs
1068 	 * would confuse code below.
1069 	 */
1070 	m->m_data += hlen;
1071 	m->m_len -= hlen;
1072 
1073 	/*
1074 	 * If first fragment to arrive, create a reassembly queue.
1075 	 */
1076 	if (fp == NULL) {
1077 		if ((fp = mpipe_alloc_nowait(&ipq_mpipe)) == NULL)
1078 			goto dropfrag;
1079 		insque(fp, &ipq[sum]);
1080 		nipq++;
1081 		fp->ipq_nfrags = 1;
1082 		fp->ipq_ttl = IPFRAGTTL;
1083 		fp->ipq_p = ip->ip_p;
1084 		fp->ipq_id = ip->ip_id;
1085 		fp->ipq_src = ip->ip_src;
1086 		fp->ipq_dst = ip->ip_dst;
1087 		fp->ipq_frags = m;
1088 		m->m_nextpkt = NULL;
1089 		goto inserted;
1090 	} else {
1091 		fp->ipq_nfrags++;
1092 	}
1093 
1094 #define	GETIP(m)	((struct ip*)((m)->m_pkthdr.header))
1095 
1096 	/*
1097 	 * Find a segment which begins after this one does.
1098 	 */
1099 	for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
1100 		if (GETIP(q)->ip_off > ip->ip_off)
1101 			break;
1102 	}
1103 
1104 	/*
1105 	 * If there is a preceding segment, it may provide some of
1106 	 * our data already.  If so, drop the data from the incoming
1107 	 * segment.  If it provides all of our data, drop us, otherwise
1108 	 * stick new segment in the proper place.
1109 	 *
1110 	 * If some of the data is dropped from the the preceding
1111 	 * segment, then it's checksum is invalidated.
1112 	 */
1113 	if (p) {
1114 		i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off;
1115 		if (i > 0) {
1116 			if (i >= ip->ip_len)
1117 				goto dropfrag;
1118 			m_adj(m, i);
1119 			m->m_pkthdr.csum_flags = 0;
1120 			ip->ip_off += i;
1121 			ip->ip_len -= i;
1122 		}
1123 		m->m_nextpkt = p->m_nextpkt;
1124 		p->m_nextpkt = m;
1125 	} else {
1126 		m->m_nextpkt = fp->ipq_frags;
1127 		fp->ipq_frags = m;
1128 	}
1129 
1130 	/*
1131 	 * While we overlap succeeding segments trim them or,
1132 	 * if they are completely covered, dequeue them.
1133 	 */
1134 	for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off;
1135 	     q = nq) {
1136 		i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off;
1137 		if (i < GETIP(q)->ip_len) {
1138 			GETIP(q)->ip_len -= i;
1139 			GETIP(q)->ip_off += i;
1140 			m_adj(q, i);
1141 			q->m_pkthdr.csum_flags = 0;
1142 			break;
1143 		}
1144 		nq = q->m_nextpkt;
1145 		m->m_nextpkt = nq;
1146 		ipstat.ips_fragdropped++;
1147 		fp->ipq_nfrags--;
1148 		q->m_nextpkt = NULL;
1149 		m_freem(q);
1150 	}
1151 
1152 inserted:
1153 	/*
1154 	 * Check for complete reassembly and perform frag per packet
1155 	 * limiting.
1156 	 *
1157 	 * Frag limiting is performed here so that the nth frag has
1158 	 * a chance to complete the packet before we drop the packet.
1159 	 * As a result, n+1 frags are actually allowed per packet, but
1160 	 * only n will ever be stored. (n = maxfragsperpacket.)
1161 	 *
1162 	 */
1163 	next = 0;
1164 	for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
1165 		if (GETIP(q)->ip_off != next) {
1166 			if (fp->ipq_nfrags > maxfragsperpacket) {
1167 				ipstat.ips_fragdropped += fp->ipq_nfrags;
1168 				ip_freef(fp);
1169 			}
1170 			goto done;
1171 		}
1172 		next += GETIP(q)->ip_len;
1173 	}
1174 	/* Make sure the last packet didn't have the IP_MF flag */
1175 	if (p->m_flags & M_FRAG) {
1176 		if (fp->ipq_nfrags > maxfragsperpacket) {
1177 			ipstat.ips_fragdropped += fp->ipq_nfrags;
1178 			ip_freef(fp);
1179 		}
1180 		goto done;
1181 	}
1182 
1183 	/*
1184 	 * Reassembly is complete.  Make sure the packet is a sane size.
1185 	 */
1186 	q = fp->ipq_frags;
1187 	ip = GETIP(q);
1188 	if (next + (IP_VHL_HL(ip->ip_vhl) << 2) > IP_MAXPACKET) {
1189 		ipstat.ips_toolong++;
1190 		ipstat.ips_fragdropped += fp->ipq_nfrags;
1191 		ip_freef(fp);
1192 		goto done;
1193 	}
1194 
1195 	/*
1196 	 * Concatenate fragments.
1197 	 */
1198 	m = q;
1199 	n = m->m_next;
1200 	m->m_next = NULL;
1201 	m_cat(m, n);
1202 	nq = q->m_nextpkt;
1203 	q->m_nextpkt = NULL;
1204 	for (q = nq; q != NULL; q = nq) {
1205 		nq = q->m_nextpkt;
1206 		q->m_nextpkt = NULL;
1207 		m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
1208 		m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
1209 		m_cat(m, q);
1210 	}
1211 
1212 	/*
1213 	 * Clean up the 1's complement checksum.  Carry over 16 bits must
1214 	 * be added back.  This assumes no more then 65535 packet fragments
1215 	 * were reassembled.  A second carry can also occur (but not a third).
1216 	 */
1217 	m->m_pkthdr.csum_data = (m->m_pkthdr.csum_data & 0xffff) +
1218 				(m->m_pkthdr.csum_data >> 16);
1219 	if (m->m_pkthdr.csum_data > 0xFFFF)
1220 		m->m_pkthdr.csum_data -= 0xFFFF;
1221 
1222 	/*
1223 	 * Create header for new ip packet by
1224 	 * modifying header of first packet;
1225 	 * dequeue and discard fragment reassembly header.
1226 	 * Make header visible.
1227 	 */
1228 	ip->ip_len = next;
1229 	ip->ip_src = fp->ipq_src;
1230 	ip->ip_dst = fp->ipq_dst;
1231 	remque(fp);
1232 	nipq--;
1233 	mpipe_free(&ipq_mpipe, fp);
1234 	m->m_len += (IP_VHL_HL(ip->ip_vhl) << 2);
1235 	m->m_data -= (IP_VHL_HL(ip->ip_vhl) << 2);
1236 	/* some debugging cruft by sklower, below, will go away soon */
1237 	if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */
1238 		int plen = 0;
1239 
1240 		for (n = m; n; n = n->m_next)
1241 			plen += n->m_len;
1242 		m->m_pkthdr.len = plen;
1243 	}
1244 
1245 	ipstat.ips_reassembled++;
1246 	lwkt_reltoken(&ipq_token);
1247 	return (m);
1248 
1249 dropfrag:
1250 	ipstat.ips_fragdropped++;
1251 	if (fp != NULL)
1252 		fp->ipq_nfrags--;
1253 	m_freem(m);
1254 done:
1255 	lwkt_reltoken(&ipq_token);
1256 	return (NULL);
1257 
1258 #undef GETIP
1259 }
1260 
1261 /*
1262  * Free a fragment reassembly header and all
1263  * associated datagrams.
1264  *
1265  * Called with ipq_token held.
1266  */
1267 static void
1268 ip_freef(struct ipq *fp)
1269 {
1270 	struct mbuf *q;
1271 
1272 	/*
1273 	 * Remove first to protect against blocking
1274 	 */
1275 	remque(fp);
1276 
1277 	/*
1278 	 * Clean out at our leisure
1279 	 */
1280 	while (fp->ipq_frags) {
1281 		q = fp->ipq_frags;
1282 		fp->ipq_frags = q->m_nextpkt;
1283 		q->m_nextpkt = NULL;
1284 		m_freem(q);
1285 	}
1286 	mpipe_free(&ipq_mpipe, fp);
1287 	nipq--;
1288 }
1289 
1290 /*
1291  * IP timer processing;
1292  * if a timer expires on a reassembly
1293  * queue, discard it.
1294  */
1295 void
1296 ip_slowtimo(void)
1297 {
1298 	struct ipq *fp;
1299 	int i;
1300 
1301 	lwkt_gettoken(&ipq_token);
1302 	for (i = 0; i < IPREASS_NHASH; i++) {
1303 		fp = ipq[i].next;
1304 		if (fp == NULL)
1305 			continue;
1306 		while (fp != &ipq[i]) {
1307 			--fp->ipq_ttl;
1308 			fp = fp->next;
1309 			if (fp->prev->ipq_ttl == 0) {
1310 				ipstat.ips_fragtimeout += fp->prev->ipq_nfrags;
1311 				ip_freef(fp->prev);
1312 			}
1313 		}
1314 	}
1315 	/*
1316 	 * If we are over the maximum number of fragments
1317 	 * (due to the limit being lowered), drain off
1318 	 * enough to get down to the new limit.
1319 	 */
1320 	if (maxnipq >= 0 && nipq > maxnipq) {
1321 		for (i = 0; i < IPREASS_NHASH; i++) {
1322 			while (nipq > maxnipq &&
1323 				(ipq[i].next != &ipq[i])) {
1324 				ipstat.ips_fragdropped +=
1325 				    ipq[i].next->ipq_nfrags;
1326 				ip_freef(ipq[i].next);
1327 			}
1328 		}
1329 	}
1330 	lwkt_reltoken(&ipq_token);
1331 	ipflow_slowtimo();
1332 }
1333 
1334 /*
1335  * Drain off all datagram fragments.
1336  */
1337 void
1338 ip_drain(void)
1339 {
1340 	int i;
1341 
1342 	lwkt_gettoken(&ipq_token);
1343 	for (i = 0; i < IPREASS_NHASH; i++) {
1344 		while (ipq[i].next != &ipq[i]) {
1345 			ipstat.ips_fragdropped += ipq[i].next->ipq_nfrags;
1346 			ip_freef(ipq[i].next);
1347 		}
1348 	}
1349 	lwkt_reltoken(&ipq_token);
1350 	in_rtqdrain();
1351 }
1352 
1353 /*
1354  * Do option processing on a datagram,
1355  * possibly discarding it if bad options are encountered,
1356  * or forwarding it if source-routed.
1357  * The pass argument is used when operating in the IPSTEALTH
1358  * mode to tell what options to process:
1359  * [LS]SRR (pass 0) or the others (pass 1).
1360  * The reason for as many as two passes is that when doing IPSTEALTH,
1361  * non-routing options should be processed only if the packet is for us.
1362  * Returns 1 if packet has been forwarded/freed,
1363  * 0 if the packet should be processed further.
1364  */
1365 static int
1366 ip_dooptions(struct mbuf *m, int pass, struct sockaddr_in *next_hop)
1367 {
1368 	struct sockaddr_in ipaddr = { sizeof ipaddr, AF_INET };
1369 	struct ip *ip = mtod(m, struct ip *);
1370 	u_char *cp;
1371 	struct in_ifaddr *ia;
1372 	int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB;
1373 	boolean_t forward = FALSE;
1374 	struct in_addr *sin, dst;
1375 	n_time ntime;
1376 
1377 	dst = ip->ip_dst;
1378 	cp = (u_char *)(ip + 1);
1379 	cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
1380 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1381 		opt = cp[IPOPT_OPTVAL];
1382 		if (opt == IPOPT_EOL)
1383 			break;
1384 		if (opt == IPOPT_NOP)
1385 			optlen = 1;
1386 		else {
1387 			if (cnt < IPOPT_OLEN + sizeof(*cp)) {
1388 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1389 				goto bad;
1390 			}
1391 			optlen = cp[IPOPT_OLEN];
1392 			if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
1393 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1394 				goto bad;
1395 			}
1396 		}
1397 		switch (opt) {
1398 
1399 		default:
1400 			break;
1401 
1402 		/*
1403 		 * Source routing with record.
1404 		 * Find interface with current destination address.
1405 		 * If none on this machine then drop if strictly routed,
1406 		 * or do nothing if loosely routed.
1407 		 * Record interface address and bring up next address
1408 		 * component.  If strictly routed make sure next
1409 		 * address is on directly accessible net.
1410 		 */
1411 		case IPOPT_LSRR:
1412 		case IPOPT_SSRR:
1413 			if (ipstealth && pass > 0)
1414 				break;
1415 			if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1416 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1417 				goto bad;
1418 			}
1419 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1420 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1421 				goto bad;
1422 			}
1423 			ipaddr.sin_addr = ip->ip_dst;
1424 			ia = (struct in_ifaddr *)
1425 				ifa_ifwithaddr((struct sockaddr *)&ipaddr);
1426 			if (ia == NULL) {
1427 				if (opt == IPOPT_SSRR) {
1428 					type = ICMP_UNREACH;
1429 					code = ICMP_UNREACH_SRCFAIL;
1430 					goto bad;
1431 				}
1432 				if (!ip_dosourceroute)
1433 					goto nosourcerouting;
1434 				/*
1435 				 * Loose routing, and not at next destination
1436 				 * yet; nothing to do except forward.
1437 				 */
1438 				break;
1439 			}
1440 			off--;			/* 0 origin */
1441 			if (off > optlen - (int)sizeof(struct in_addr)) {
1442 				/*
1443 				 * End of source route.  Should be for us.
1444 				 */
1445 				if (!ip_acceptsourceroute)
1446 					goto nosourcerouting;
1447 				save_rte(m, cp, ip->ip_src);
1448 				break;
1449 			}
1450 			if (ipstealth)
1451 				goto dropit;
1452 			if (!ip_dosourceroute) {
1453 				if (ipforwarding) {
1454 					char buf[sizeof "aaa.bbb.ccc.ddd"];
1455 
1456 					/*
1457 					 * Acting as a router, so generate ICMP
1458 					 */
1459 nosourcerouting:
1460 					strcpy(buf, inet_ntoa(ip->ip_dst));
1461 					log(LOG_WARNING,
1462 					    "attempted source route from %s to %s\n",
1463 					    inet_ntoa(ip->ip_src), buf);
1464 					type = ICMP_UNREACH;
1465 					code = ICMP_UNREACH_SRCFAIL;
1466 					goto bad;
1467 				} else {
1468 					/*
1469 					 * Not acting as a router,
1470 					 * so silently drop.
1471 					 */
1472 dropit:
1473 					ipstat.ips_cantforward++;
1474 					m_freem(m);
1475 					return (1);
1476 				}
1477 			}
1478 
1479 			/*
1480 			 * locate outgoing interface
1481 			 */
1482 			memcpy(&ipaddr.sin_addr, cp + off,
1483 			    sizeof ipaddr.sin_addr);
1484 
1485 			if (opt == IPOPT_SSRR) {
1486 #define	INA	struct in_ifaddr *
1487 #define	SA	struct sockaddr *
1488 				if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr))
1489 									== NULL)
1490 					ia = (INA)ifa_ifwithnet((SA)&ipaddr);
1491 			} else {
1492 				ia = ip_rtaddr(ipaddr.sin_addr, NULL);
1493 			}
1494 			if (ia == NULL) {
1495 				type = ICMP_UNREACH;
1496 				code = ICMP_UNREACH_SRCFAIL;
1497 				goto bad;
1498 			}
1499 			ip->ip_dst = ipaddr.sin_addr;
1500 			memcpy(cp + off, &IA_SIN(ia)->sin_addr,
1501 			    sizeof(struct in_addr));
1502 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1503 			/*
1504 			 * Let ip_intr's mcast routing check handle mcast pkts
1505 			 */
1506 			forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr));
1507 			break;
1508 
1509 		case IPOPT_RR:
1510 			if (ipstealth && pass == 0)
1511 				break;
1512 			if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1513 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1514 				goto bad;
1515 			}
1516 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1517 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1518 				goto bad;
1519 			}
1520 			/*
1521 			 * If no space remains, ignore.
1522 			 */
1523 			off--;			/* 0 origin */
1524 			if (off > optlen - (int)sizeof(struct in_addr))
1525 				break;
1526 			memcpy(&ipaddr.sin_addr, &ip->ip_dst,
1527 			    sizeof ipaddr.sin_addr);
1528 			/*
1529 			 * locate outgoing interface; if we're the destination,
1530 			 * use the incoming interface (should be same).
1531 			 */
1532 			if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == NULL &&
1533 			    (ia = ip_rtaddr(ipaddr.sin_addr, NULL)) == NULL) {
1534 				type = ICMP_UNREACH;
1535 				code = ICMP_UNREACH_HOST;
1536 				goto bad;
1537 			}
1538 			memcpy(cp + off, &IA_SIN(ia)->sin_addr,
1539 			    sizeof(struct in_addr));
1540 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1541 			break;
1542 
1543 		case IPOPT_TS:
1544 			if (ipstealth && pass == 0)
1545 				break;
1546 			code = cp - (u_char *)ip;
1547 			if (optlen < 4 || optlen > 40) {
1548 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1549 				goto bad;
1550 			}
1551 			if ((off = cp[IPOPT_OFFSET]) < 5) {
1552 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1553 				goto bad;
1554 			}
1555 			if (off > optlen - (int)sizeof(int32_t)) {
1556 				cp[IPOPT_OFFSET + 1] += (1 << 4);
1557 				if ((cp[IPOPT_OFFSET + 1] & 0xf0) == 0) {
1558 					code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1559 					goto bad;
1560 				}
1561 				break;
1562 			}
1563 			off--;				/* 0 origin */
1564 			sin = (struct in_addr *)(cp + off);
1565 			switch (cp[IPOPT_OFFSET + 1] & 0x0f) {
1566 
1567 			case IPOPT_TS_TSONLY:
1568 				break;
1569 
1570 			case IPOPT_TS_TSANDADDR:
1571 				if (off + sizeof(n_time) +
1572 				    sizeof(struct in_addr) > optlen) {
1573 					code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1574 					goto bad;
1575 				}
1576 				ipaddr.sin_addr = dst;
1577 				ia = (INA)ifaof_ifpforaddr((SA)&ipaddr,
1578 							    m->m_pkthdr.rcvif);
1579 				if (ia == NULL)
1580 					continue;
1581 				memcpy(sin, &IA_SIN(ia)->sin_addr,
1582 				    sizeof(struct in_addr));
1583 				cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1584 				off += sizeof(struct in_addr);
1585 				break;
1586 
1587 			case IPOPT_TS_PRESPEC:
1588 				if (off + sizeof(n_time) +
1589 				    sizeof(struct in_addr) > optlen) {
1590 					code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1591 					goto bad;
1592 				}
1593 				memcpy(&ipaddr.sin_addr, sin,
1594 				    sizeof(struct in_addr));
1595 				if (ifa_ifwithaddr((SA)&ipaddr) == NULL)
1596 					continue;
1597 				cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1598 				off += sizeof(struct in_addr);
1599 				break;
1600 
1601 			default:
1602 				code = &cp[IPOPT_OFFSET + 1] - (u_char *)ip;
1603 				goto bad;
1604 			}
1605 			ntime = iptime();
1606 			memcpy(cp + off, &ntime, sizeof(n_time));
1607 			cp[IPOPT_OFFSET] += sizeof(n_time);
1608 		}
1609 	}
1610 	if (forward && ipforwarding) {
1611 		ip_forward(m, TRUE, next_hop);
1612 		return (1);
1613 	}
1614 	return (0);
1615 bad:
1616 	icmp_error(m, type, code, 0, 0);
1617 	ipstat.ips_badoptions++;
1618 	return (1);
1619 }
1620 
1621 /*
1622  * Given address of next destination (final or next hop),
1623  * return internet address info of interface to be used to get there.
1624  */
1625 struct in_ifaddr *
1626 ip_rtaddr(struct in_addr dst, struct route *ro0)
1627 {
1628 	struct route sro, *ro;
1629 	struct sockaddr_in *sin;
1630 	struct in_ifaddr *ia;
1631 
1632 	if (ro0 != NULL) {
1633 		ro = ro0;
1634 	} else {
1635 		bzero(&sro, sizeof(sro));
1636 		ro = &sro;
1637 	}
1638 
1639 	sin = (struct sockaddr_in *)&ro->ro_dst;
1640 
1641 	if (ro->ro_rt == NULL || dst.s_addr != sin->sin_addr.s_addr) {
1642 		if (ro->ro_rt != NULL) {
1643 			RTFREE(ro->ro_rt);
1644 			ro->ro_rt = NULL;
1645 		}
1646 		sin->sin_family = AF_INET;
1647 		sin->sin_len = sizeof *sin;
1648 		sin->sin_addr = dst;
1649 		rtalloc_ign(ro, RTF_PRCLONING);
1650 	}
1651 
1652 	if (ro->ro_rt == NULL)
1653 		return (NULL);
1654 
1655 	ia = ifatoia(ro->ro_rt->rt_ifa);
1656 
1657 	if (ro == &sro)
1658 		RTFREE(ro->ro_rt);
1659 	return ia;
1660 }
1661 
1662 /*
1663  * Save incoming source route for use in replies,
1664  * to be picked up later by ip_srcroute if the receiver is interested.
1665  */
1666 static void
1667 save_rte(struct mbuf *m, u_char *option, struct in_addr dst)
1668 {
1669 	struct m_tag *mtag;
1670 	struct ip_srcrt_opt *opt;
1671 	unsigned olen;
1672 
1673 	mtag = m_tag_get(PACKET_TAG_IPSRCRT, sizeof(*opt), MB_DONTWAIT);
1674 	if (mtag == NULL)
1675 		return;
1676 	opt = m_tag_data(mtag);
1677 
1678 	olen = option[IPOPT_OLEN];
1679 #ifdef DIAGNOSTIC
1680 	if (ipprintfs)
1681 		kprintf("save_rte: olen %d\n", olen);
1682 #endif
1683 	if (olen > sizeof(opt->ip_srcrt) - (1 + sizeof(dst))) {
1684 		m_tag_free(mtag);
1685 		return;
1686 	}
1687 	bcopy(option, opt->ip_srcrt.srcopt, olen);
1688 	opt->ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1689 	opt->ip_srcrt.dst = dst;
1690 	m_tag_prepend(m, mtag);
1691 }
1692 
1693 /*
1694  * Retrieve incoming source route for use in replies,
1695  * in the same form used by setsockopt.
1696  * The first hop is placed before the options, will be removed later.
1697  */
1698 struct mbuf *
1699 ip_srcroute(struct mbuf *m0)
1700 {
1701 	struct in_addr *p, *q;
1702 	struct mbuf *m;
1703 	struct m_tag *mtag;
1704 	struct ip_srcrt_opt *opt;
1705 
1706 	if (m0 == NULL)
1707 		return NULL;
1708 
1709 	mtag = m_tag_find(m0, PACKET_TAG_IPSRCRT, NULL);
1710 	if (mtag == NULL)
1711 		return NULL;
1712 	opt = m_tag_data(mtag);
1713 
1714 	if (opt->ip_nhops == 0)
1715 		return (NULL);
1716 	m = m_get(MB_DONTWAIT, MT_HEADER);
1717 	if (m == NULL)
1718 		return (NULL);
1719 
1720 #define	OPTSIZ	(sizeof(opt->ip_srcrt.nop) + sizeof(opt->ip_srcrt.srcopt))
1721 
1722 	/* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
1723 	m->m_len = opt->ip_nhops * sizeof(struct in_addr) +
1724 		   sizeof(struct in_addr) + OPTSIZ;
1725 #ifdef DIAGNOSTIC
1726 	if (ipprintfs) {
1727 		kprintf("ip_srcroute: nhops %d mlen %d",
1728 			opt->ip_nhops, m->m_len);
1729 	}
1730 #endif
1731 
1732 	/*
1733 	 * First save first hop for return route
1734 	 */
1735 	p = &opt->ip_srcrt.route[opt->ip_nhops - 1];
1736 	*(mtod(m, struct in_addr *)) = *p--;
1737 #ifdef DIAGNOSTIC
1738 	if (ipprintfs)
1739 		kprintf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr));
1740 #endif
1741 
1742 	/*
1743 	 * Copy option fields and padding (nop) to mbuf.
1744 	 */
1745 	opt->ip_srcrt.nop = IPOPT_NOP;
1746 	opt->ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
1747 	memcpy(mtod(m, caddr_t) + sizeof(struct in_addr), &opt->ip_srcrt.nop,
1748 	    OPTSIZ);
1749 	q = (struct in_addr *)(mtod(m, caddr_t) +
1750 	    sizeof(struct in_addr) + OPTSIZ);
1751 #undef OPTSIZ
1752 	/*
1753 	 * Record return path as an IP source route,
1754 	 * reversing the path (pointers are now aligned).
1755 	 */
1756 	while (p >= opt->ip_srcrt.route) {
1757 #ifdef DIAGNOSTIC
1758 		if (ipprintfs)
1759 			kprintf(" %x", ntohl(q->s_addr));
1760 #endif
1761 		*q++ = *p--;
1762 	}
1763 	/*
1764 	 * Last hop goes to final destination.
1765 	 */
1766 	*q = opt->ip_srcrt.dst;
1767 	m_tag_delete(m0, mtag);
1768 #ifdef DIAGNOSTIC
1769 	if (ipprintfs)
1770 		kprintf(" %x\n", ntohl(q->s_addr));
1771 #endif
1772 	return (m);
1773 }
1774 
1775 /*
1776  * Strip out IP options.
1777  */
1778 void
1779 ip_stripoptions(struct mbuf *m)
1780 {
1781 	int datalen;
1782 	struct ip *ip = mtod(m, struct ip *);
1783 	caddr_t opts;
1784 	int optlen;
1785 
1786 	optlen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
1787 	opts = (caddr_t)(ip + 1);
1788 	datalen = m->m_len - (sizeof(struct ip) + optlen);
1789 	bcopy(opts + optlen, opts, datalen);
1790 	m->m_len -= optlen;
1791 	if (m->m_flags & M_PKTHDR)
1792 		m->m_pkthdr.len -= optlen;
1793 	ip->ip_vhl = IP_MAKE_VHL(IPVERSION, sizeof(struct ip) >> 2);
1794 }
1795 
1796 u_char inetctlerrmap[PRC_NCMDS] = {
1797 	0,		0,		0,		0,
1798 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1799 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1800 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1801 	0,		0,		0,		0,
1802 	ENOPROTOOPT,	ECONNREFUSED
1803 };
1804 
1805 /*
1806  * Forward a packet.  If some error occurs return the sender
1807  * an icmp packet.  Note we can't always generate a meaningful
1808  * icmp message because icmp doesn't have a large enough repertoire
1809  * of codes and types.
1810  *
1811  * If not forwarding, just drop the packet.  This could be confusing
1812  * if ipforwarding was zero but some routing protocol was advancing
1813  * us as a gateway to somewhere.  However, we must let the routing
1814  * protocol deal with that.
1815  *
1816  * The using_srcrt parameter indicates whether the packet is being forwarded
1817  * via a source route.
1818  */
1819 void
1820 ip_forward(struct mbuf *m, boolean_t using_srcrt, struct sockaddr_in *next_hop)
1821 {
1822 	struct ip *ip = mtod(m, struct ip *);
1823 	struct rtentry *rt;
1824 	struct route fwd_ro;
1825 	int error, type = 0, code = 0, destmtu = 0;
1826 	struct mbuf *mcopy;
1827 	n_long dest;
1828 	struct in_addr pkt_dst;
1829 
1830 	dest = INADDR_ANY;
1831 	/*
1832 	 * Cache the destination address of the packet; this may be
1833 	 * changed by use of 'ipfw fwd'.
1834 	 */
1835 	pkt_dst = (next_hop != NULL) ? next_hop->sin_addr : ip->ip_dst;
1836 
1837 #ifdef DIAGNOSTIC
1838 	if (ipprintfs)
1839 		kprintf("forward: src %x dst %x ttl %x\n",
1840 		       ip->ip_src.s_addr, pkt_dst.s_addr, ip->ip_ttl);
1841 #endif
1842 
1843 	if (m->m_flags & (M_BCAST | M_MCAST) || !in_canforward(pkt_dst)) {
1844 		ipstat.ips_cantforward++;
1845 		m_freem(m);
1846 		return;
1847 	}
1848 	if (!ipstealth && ip->ip_ttl <= IPTTLDEC) {
1849 		icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0);
1850 		return;
1851 	}
1852 
1853 	bzero(&fwd_ro, sizeof(fwd_ro));
1854 	ip_rtaddr(pkt_dst, &fwd_ro);
1855 	if (fwd_ro.ro_rt == NULL) {
1856 		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0);
1857 		return;
1858 	}
1859 	rt = fwd_ro.ro_rt;
1860 
1861 	/*
1862 	 * Save the IP header and at most 8 bytes of the payload,
1863 	 * in case we need to generate an ICMP message to the src.
1864 	 *
1865 	 * XXX this can be optimized a lot by saving the data in a local
1866 	 * buffer on the stack (72 bytes at most), and only allocating the
1867 	 * mbuf if really necessary. The vast majority of the packets
1868 	 * are forwarded without having to send an ICMP back (either
1869 	 * because unnecessary, or because rate limited), so we are
1870 	 * really we are wasting a lot of work here.
1871 	 *
1872 	 * We don't use m_copy() because it might return a reference
1873 	 * to a shared cluster. Both this function and ip_output()
1874 	 * assume exclusive access to the IP header in `m', so any
1875 	 * data in a cluster may change before we reach icmp_error().
1876 	 */
1877 	MGETHDR(mcopy, MB_DONTWAIT, m->m_type);
1878 	if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, MB_DONTWAIT)) {
1879 		/*
1880 		 * It's probably ok if the pkthdr dup fails (because
1881 		 * the deep copy of the tag chain failed), but for now
1882 		 * be conservative and just discard the copy since
1883 		 * code below may some day want the tags.
1884 		 */
1885 		m_free(mcopy);
1886 		mcopy = NULL;
1887 	}
1888 	if (mcopy != NULL) {
1889 		mcopy->m_len = imin((IP_VHL_HL(ip->ip_vhl) << 2) + 8,
1890 		    (int)ip->ip_len);
1891 		mcopy->m_pkthdr.len = mcopy->m_len;
1892 		m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1893 	}
1894 
1895 	if (!ipstealth)
1896 		ip->ip_ttl -= IPTTLDEC;
1897 
1898 	/*
1899 	 * If forwarding packet using same interface that it came in on,
1900 	 * perhaps should send a redirect to sender to shortcut a hop.
1901 	 * Only send redirect if source is sending directly to us,
1902 	 * and if packet was not source routed (or has any options).
1903 	 * Also, don't send redirect if forwarding using a default route
1904 	 * or a route modified by a redirect.
1905 	 */
1906 	if (rt->rt_ifp == m->m_pkthdr.rcvif &&
1907 	    !(rt->rt_flags & (RTF_DYNAMIC | RTF_MODIFIED)) &&
1908 	    satosin(rt_key(rt))->sin_addr.s_addr != INADDR_ANY &&
1909 	    ipsendredirects && !using_srcrt && next_hop == NULL) {
1910 		u_long src = ntohl(ip->ip_src.s_addr);
1911 		struct in_ifaddr *rt_ifa = (struct in_ifaddr *)rt->rt_ifa;
1912 
1913 		if (rt_ifa != NULL &&
1914 		    (src & rt_ifa->ia_subnetmask) == rt_ifa->ia_subnet) {
1915 			if (rt->rt_flags & RTF_GATEWAY)
1916 				dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
1917 			else
1918 				dest = pkt_dst.s_addr;
1919 			/*
1920 			 * Router requirements says to only send
1921 			 * host redirects.
1922 			 */
1923 			type = ICMP_REDIRECT;
1924 			code = ICMP_REDIRECT_HOST;
1925 #ifdef DIAGNOSTIC
1926 			if (ipprintfs)
1927 				kprintf("redirect (%d) to %x\n", code, dest);
1928 #endif
1929 		}
1930 	}
1931 
1932 	error = ip_output(m, NULL, &fwd_ro, IP_FORWARDING, NULL, NULL);
1933 	if (error == 0) {
1934 		ipstat.ips_forward++;
1935 		if (type == 0) {
1936 			if (mcopy) {
1937 				ipflow_create(&fwd_ro, mcopy);
1938 				m_freem(mcopy);
1939 			}
1940 			goto done;
1941 		} else {
1942 			ipstat.ips_redirectsent++;
1943 		}
1944 	} else {
1945 		ipstat.ips_cantforward++;
1946 	}
1947 
1948 	if (mcopy == NULL)
1949 		goto done;
1950 
1951 	/*
1952 	 * Send ICMP message.
1953 	 */
1954 
1955 	switch (error) {
1956 
1957 	case 0:				/* forwarded, but need redirect */
1958 		/* type, code set above */
1959 		break;
1960 
1961 	case ENETUNREACH:		/* shouldn't happen, checked above */
1962 	case EHOSTUNREACH:
1963 	case ENETDOWN:
1964 	case EHOSTDOWN:
1965 	default:
1966 		type = ICMP_UNREACH;
1967 		code = ICMP_UNREACH_HOST;
1968 		break;
1969 
1970 	case EMSGSIZE:
1971 		type = ICMP_UNREACH;
1972 		code = ICMP_UNREACH_NEEDFRAG;
1973 #ifdef IPSEC
1974 		/*
1975 		 * If the packet is routed over IPsec tunnel, tell the
1976 		 * originator the tunnel MTU.
1977 		 *	tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
1978 		 * XXX quickhack!!!
1979 		 */
1980 		if (fwd_ro.ro_rt != NULL) {
1981 			struct secpolicy *sp = NULL;
1982 			int ipsecerror;
1983 			int ipsechdr;
1984 			struct route *ro;
1985 
1986 			sp = ipsec4_getpolicybyaddr(mcopy,
1987 						    IPSEC_DIR_OUTBOUND,
1988 						    IP_FORWARDING,
1989 						    &ipsecerror);
1990 
1991 			if (sp == NULL)
1992 				destmtu = fwd_ro.ro_rt->rt_ifp->if_mtu;
1993 			else {
1994 				/* count IPsec header size */
1995 				ipsechdr = ipsec4_hdrsiz(mcopy,
1996 							 IPSEC_DIR_OUTBOUND,
1997 							 NULL);
1998 
1999 				/*
2000 				 * find the correct route for outer IPv4
2001 				 * header, compute tunnel MTU.
2002 				 *
2003 				 */
2004 				if (sp->req != NULL && sp->req->sav != NULL &&
2005 				    sp->req->sav->sah != NULL) {
2006 					ro = &sp->req->sav->sah->sa_route;
2007 					if (ro->ro_rt != NULL &&
2008 					    ro->ro_rt->rt_ifp != NULL) {
2009 						destmtu =
2010 						    ro->ro_rt->rt_ifp->if_mtu;
2011 						destmtu -= ipsechdr;
2012 					}
2013 				}
2014 
2015 				key_freesp(sp);
2016 			}
2017 		}
2018 #elif FAST_IPSEC
2019 		/*
2020 		 * If the packet is routed over IPsec tunnel, tell the
2021 		 * originator the tunnel MTU.
2022 		 *	tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
2023 		 * XXX quickhack!!!
2024 		 */
2025 		if (fwd_ro.ro_rt != NULL) {
2026 			struct secpolicy *sp = NULL;
2027 			int ipsecerror;
2028 			int ipsechdr;
2029 			struct route *ro;
2030 
2031 			sp = ipsec_getpolicybyaddr(mcopy,
2032 						   IPSEC_DIR_OUTBOUND,
2033 						   IP_FORWARDING,
2034 						   &ipsecerror);
2035 
2036 			if (sp == NULL)
2037 				destmtu = fwd_ro.ro_rt->rt_ifp->if_mtu;
2038 			else {
2039 				/* count IPsec header size */
2040 				ipsechdr = ipsec4_hdrsiz(mcopy,
2041 							 IPSEC_DIR_OUTBOUND,
2042 							 NULL);
2043 
2044 				/*
2045 				 * find the correct route for outer IPv4
2046 				 * header, compute tunnel MTU.
2047 				 */
2048 
2049 				if (sp->req != NULL &&
2050 				    sp->req->sav != NULL &&
2051 				    sp->req->sav->sah != NULL) {
2052 					ro = &sp->req->sav->sah->sa_route;
2053 					if (ro->ro_rt != NULL &&
2054 					    ro->ro_rt->rt_ifp != NULL) {
2055 						destmtu =
2056 						    ro->ro_rt->rt_ifp->if_mtu;
2057 						destmtu -= ipsechdr;
2058 					}
2059 				}
2060 
2061 				KEY_FREESP(&sp);
2062 			}
2063 		}
2064 #else /* !IPSEC && !FAST_IPSEC */
2065 		if (fwd_ro.ro_rt != NULL)
2066 			destmtu = fwd_ro.ro_rt->rt_ifp->if_mtu;
2067 #endif /*IPSEC*/
2068 		ipstat.ips_cantfrag++;
2069 		break;
2070 
2071 	case ENOBUFS:
2072 		/*
2073 		 * A router should not generate ICMP_SOURCEQUENCH as
2074 		 * required in RFC1812 Requirements for IP Version 4 Routers.
2075 		 * Source quench could be a big problem under DoS attacks,
2076 		 * or if the underlying interface is rate-limited.
2077 		 * Those who need source quench packets may re-enable them
2078 		 * via the net.inet.ip.sendsourcequench sysctl.
2079 		 */
2080 		if (!ip_sendsourcequench) {
2081 			m_freem(mcopy);
2082 			goto done;
2083 		} else {
2084 			type = ICMP_SOURCEQUENCH;
2085 			code = 0;
2086 		}
2087 		break;
2088 
2089 	case EACCES:			/* ipfw denied packet */
2090 		m_freem(mcopy);
2091 		goto done;
2092 	}
2093 	icmp_error(mcopy, type, code, dest, destmtu);
2094 done:
2095 	if (fwd_ro.ro_rt != NULL)
2096 		RTFREE(fwd_ro.ro_rt);
2097 }
2098 
2099 void
2100 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
2101 	       struct mbuf *m)
2102 {
2103 	if (inp->inp_socket->so_options & SO_TIMESTAMP) {
2104 		struct timeval tv;
2105 
2106 		microtime(&tv);
2107 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
2108 		    SCM_TIMESTAMP, SOL_SOCKET);
2109 		if (*mp)
2110 			mp = &(*mp)->m_next;
2111 	}
2112 	if (inp->inp_flags & INP_RECVDSTADDR) {
2113 		*mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
2114 		    sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
2115 		if (*mp)
2116 			mp = &(*mp)->m_next;
2117 	}
2118 	if (inp->inp_flags & INP_RECVTTL) {
2119 		*mp = sbcreatecontrol((caddr_t) &ip->ip_ttl,
2120 		    sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
2121 		if (*mp)
2122 			mp = &(*mp)->m_next;
2123 	}
2124 #ifdef notyet
2125 	/* XXX
2126 	 * Moving these out of udp_input() made them even more broken
2127 	 * than they already were.
2128 	 */
2129 	/* options were tossed already */
2130 	if (inp->inp_flags & INP_RECVOPTS) {
2131 		*mp = sbcreatecontrol((caddr_t) opts_deleted_above,
2132 		    sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
2133 		if (*mp)
2134 			mp = &(*mp)->m_next;
2135 	}
2136 	/* ip_srcroute doesn't do what we want here, need to fix */
2137 	if (inp->inp_flags & INP_RECVRETOPTS) {
2138 		*mp = sbcreatecontrol((caddr_t) ip_srcroute(m),
2139 		    sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
2140 		if (*mp)
2141 			mp = &(*mp)->m_next;
2142 	}
2143 #endif
2144 	if (inp->inp_flags & INP_RECVIF) {
2145 		struct ifnet *ifp;
2146 		struct sdlbuf {
2147 			struct sockaddr_dl sdl;
2148 			u_char	pad[32];
2149 		} sdlbuf;
2150 		struct sockaddr_dl *sdp;
2151 		struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
2152 
2153 		if (((ifp = m->m_pkthdr.rcvif)) &&
2154 		    ((ifp->if_index != 0) && (ifp->if_index <= if_index))) {
2155 			sdp = IF_LLSOCKADDR(ifp);
2156 			/*
2157 			 * Change our mind and don't try copy.
2158 			 */
2159 			if ((sdp->sdl_family != AF_LINK) ||
2160 			    (sdp->sdl_len > sizeof(sdlbuf))) {
2161 				goto makedummy;
2162 			}
2163 			bcopy(sdp, sdl2, sdp->sdl_len);
2164 		} else {
2165 makedummy:
2166 			sdl2->sdl_len =
2167 			    offsetof(struct sockaddr_dl, sdl_data[0]);
2168 			sdl2->sdl_family = AF_LINK;
2169 			sdl2->sdl_index = 0;
2170 			sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
2171 		}
2172 		*mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len,
2173 			IP_RECVIF, IPPROTO_IP);
2174 		if (*mp)
2175 			mp = &(*mp)->m_next;
2176 	}
2177 }
2178 
2179 /*
2180  * XXX these routines are called from the upper part of the kernel.
2181  *
2182  * They could also be moved to ip_mroute.c, since all the RSVP
2183  *  handling is done there already.
2184  */
2185 int
2186 ip_rsvp_init(struct socket *so)
2187 {
2188 	if (so->so_type != SOCK_RAW ||
2189 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
2190 		return EOPNOTSUPP;
2191 
2192 	if (ip_rsvpd != NULL)
2193 		return EADDRINUSE;
2194 
2195 	ip_rsvpd = so;
2196 	/*
2197 	 * This may seem silly, but we need to be sure we don't over-increment
2198 	 * the RSVP counter, in case something slips up.
2199 	 */
2200 	if (!ip_rsvp_on) {
2201 		ip_rsvp_on = 1;
2202 		rsvp_on++;
2203 	}
2204 
2205 	return 0;
2206 }
2207 
2208 int
2209 ip_rsvp_done(void)
2210 {
2211 	ip_rsvpd = NULL;
2212 	/*
2213 	 * This may seem silly, but we need to be sure we don't over-decrement
2214 	 * the RSVP counter, in case something slips up.
2215 	 */
2216 	if (ip_rsvp_on) {
2217 		ip_rsvp_on = 0;
2218 		rsvp_on--;
2219 	}
2220 	return 0;
2221 }
2222 
2223 void
2224 rsvp_input(struct mbuf *m, ...)	/* XXX must fixup manually */
2225 {
2226 	int off, proto;
2227 	__va_list ap;
2228 
2229 	__va_start(ap, m);
2230 	off = __va_arg(ap, int);
2231 	proto = __va_arg(ap, int);
2232 	__va_end(ap);
2233 
2234 	if (rsvp_input_p) { /* call the real one if loaded */
2235 		rsvp_input_p(m, off, proto);
2236 		return;
2237 	}
2238 
2239 	/* Can still get packets with rsvp_on = 0 if there is a local member
2240 	 * of the group to which the RSVP packet is addressed.  But in this
2241 	 * case we want to throw the packet away.
2242 	 */
2243 
2244 	if (!rsvp_on) {
2245 		m_freem(m);
2246 		return;
2247 	}
2248 
2249 	if (ip_rsvpd != NULL) {
2250 		rip_input(m, off, proto);
2251 		return;
2252 	}
2253 	/* Drop the packet */
2254 	m_freem(m);
2255 }
2256