xref: /freebsd/sys/netinet/ip_input.c (revision f9218d3d4fd34f082473b3a021c6d4d109fb47cf)
1 /*
2  * Copyright (c) 1982, 1986, 1988, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. All advertising materials mentioning features or use of this software
14  *    must display the following acknowledgement:
15  *	This product includes software developed by the University of
16  *	California, Berkeley and its contributors.
17  * 4. Neither the name of the University nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  *
33  *	@(#)ip_input.c	8.2 (Berkeley) 1/4/94
34  * $FreeBSD$
35  */
36 
37 #include "opt_bootp.h"
38 #include "opt_ipfw.h"
39 #include "opt_ipdn.h"
40 #include "opt_ipdivert.h"
41 #include "opt_ipfilter.h"
42 #include "opt_ipstealth.h"
43 #include "opt_ipsec.h"
44 #include "opt_mac.h"
45 #include "opt_pfil_hooks.h"
46 #include "opt_random_ip_id.h"
47 
48 #include <sys/param.h>
49 #include <sys/systm.h>
50 #include <sys/mac.h>
51 #include <sys/mbuf.h>
52 #include <sys/malloc.h>
53 #include <sys/domain.h>
54 #include <sys/protosw.h>
55 #include <sys/socket.h>
56 #include <sys/time.h>
57 #include <sys/kernel.h>
58 #include <sys/syslog.h>
59 #include <sys/sysctl.h>
60 
61 #include <net/pfil.h>
62 #include <net/if.h>
63 #include <net/if_types.h>
64 #include <net/if_var.h>
65 #include <net/if_dl.h>
66 #include <net/route.h>
67 #include <net/netisr.h>
68 #include <net/intrq.h>
69 
70 #include <netinet/in.h>
71 #include <netinet/in_systm.h>
72 #include <netinet/in_var.h>
73 #include <netinet/ip.h>
74 #include <netinet/in_pcb.h>
75 #include <netinet/ip_var.h>
76 #include <netinet/ip_icmp.h>
77 #include <machine/in_cksum.h>
78 
79 #include <sys/socketvar.h>
80 
81 #include <netinet/ip_fw.h>
82 #include <netinet/ip_dummynet.h>
83 
84 #ifdef IPSEC
85 #include <netinet6/ipsec.h>
86 #include <netkey/key.h>
87 #endif
88 
89 #ifdef FAST_IPSEC
90 #include <netipsec/ipsec.h>
91 #include <netipsec/key.h>
92 #endif
93 
94 int rsvp_on = 0;
95 
96 int	ipforwarding = 0;
97 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW,
98     &ipforwarding, 0, "Enable IP forwarding between interfaces");
99 
100 static int	ipsendredirects = 1; /* XXX */
101 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW,
102     &ipsendredirects, 0, "Enable sending IP redirects");
103 
104 int	ip_defttl = IPDEFTTL;
105 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW,
106     &ip_defttl, 0, "Maximum TTL on IP packets");
107 
108 static int	ip_dosourceroute = 0;
109 SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW,
110     &ip_dosourceroute, 0, "Enable forwarding source routed IP packets");
111 
112 static int	ip_acceptsourceroute = 0;
113 SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute,
114     CTLFLAG_RW, &ip_acceptsourceroute, 0,
115     "Enable accepting source routed IP packets");
116 
117 static int	ip_keepfaith = 0;
118 SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW,
119 	&ip_keepfaith,	0,
120 	"Enable packet capture for FAITH IPv4->IPv6 translater daemon");
121 
122 static int    nipq = 0;         /* total # of reass queues */
123 static int    maxnipq;
124 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragpackets, CTLFLAG_RW,
125 	&maxnipq, 0,
126 	"Maximum number of IPv4 fragment reassembly queue entries");
127 
128 static int    maxfragsperpacket;
129 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW,
130 	&maxfragsperpacket, 0,
131 	"Maximum number of IPv4 fragments allowed per packet");
132 
133 static int	ip_sendsourcequench = 0;
134 SYSCTL_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW,
135 	&ip_sendsourcequench, 0,
136 	"Enable the transmission of source quench packets");
137 
138 /*
139  * XXX - Setting ip_checkinterface mostly implements the receive side of
140  * the Strong ES model described in RFC 1122, but since the routing table
141  * and transmit implementation do not implement the Strong ES model,
142  * setting this to 1 results in an odd hybrid.
143  *
144  * XXX - ip_checkinterface currently must be disabled if you use ipnat
145  * to translate the destination address to another local interface.
146  *
147  * XXX - ip_checkinterface must be disabled if you add IP aliases
148  * to the loopback interface instead of the interface where the
149  * packets for those addresses are received.
150  */
151 static int	ip_checkinterface = 1;
152 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW,
153     &ip_checkinterface, 0, "Verify packet arrives on correct interface");
154 
155 #ifdef DIAGNOSTIC
156 static int	ipprintfs = 0;
157 #endif
158 
159 static int	ipqmaxlen = IFQ_MAXLEN;
160 
161 extern	struct domain inetdomain;
162 extern	struct protosw inetsw[];
163 u_char	ip_protox[IPPROTO_MAX];
164 struct	in_ifaddrhead in_ifaddrhead; 		/* first inet address */
165 struct	in_ifaddrhashhead *in_ifaddrhashtbl;	/* inet addr hash table  */
166 u_long 	in_ifaddrhmask;				/* mask for hash table */
167 
168 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW,
169     &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue");
170 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD,
171     &ipintrq.ifq_drops, 0, "Number of packets dropped from the IP input queue");
172 
173 struct ipstat ipstat;
174 SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW,
175     &ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)");
176 
177 /* Packet reassembly stuff */
178 #define IPREASS_NHASH_LOG2      6
179 #define IPREASS_NHASH           (1 << IPREASS_NHASH_LOG2)
180 #define IPREASS_HMASK           (IPREASS_NHASH - 1)
181 #define IPREASS_HASH(x,y) \
182 	(((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK)
183 
184 static TAILQ_HEAD(ipqhead, ipq) ipq[IPREASS_NHASH];
185 
186 #ifdef IPCTL_DEFMTU
187 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
188     &ip_mtu, 0, "Default MTU");
189 #endif
190 
191 #ifdef IPSTEALTH
192 static int	ipstealth = 0;
193 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW,
194     &ipstealth, 0, "");
195 #endif
196 
197 
198 /* Firewall hooks */
199 ip_fw_chk_t *ip_fw_chk_ptr;
200 int fw_enable = 1 ;
201 int fw_one_pass = 1;
202 
203 /* Dummynet hooks */
204 ip_dn_io_t *ip_dn_io_ptr;
205 
206 
207 /*
208  * XXX this is ugly -- the following two global variables are
209  * used to store packet state while it travels through the stack.
210  * Note that the code even makes assumptions on the size and
211  * alignment of fields inside struct ip_srcrt so e.g. adding some
212  * fields will break the code. This needs to be fixed.
213  *
214  * We need to save the IP options in case a protocol wants to respond
215  * to an incoming packet over the same route if the packet got here
216  * using IP source routing.  This allows connection establishment and
217  * maintenance when the remote end is on a network that is not known
218  * to us.
219  */
220 static int	ip_nhops = 0;
221 static	struct ip_srcrt {
222 	struct	in_addr dst;			/* final destination */
223 	char	nop;				/* one NOP to align */
224 	char	srcopt[IPOPT_OFFSET + 1];	/* OPTVAL, OLEN and OFFSET */
225 	struct	in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)];
226 } ip_srcrt;
227 
228 static void	save_rte(u_char *, struct in_addr);
229 static int	ip_dooptions(struct mbuf *m, int,
230 			struct sockaddr_in *next_hop);
231 static void	ip_forward(struct mbuf *m, int srcrt,
232 			struct sockaddr_in *next_hop);
233 static void	ip_freef(struct ipqhead *, struct ipq *);
234 static struct	mbuf *ip_reass(struct mbuf *, struct ipqhead *,
235 		struct ipq *, u_int32_t *, u_int16_t *);
236 static void	ipintr(void);
237 
238 /*
239  * IP initialization: fill in IP protocol switch table.
240  * All protocols not implemented in kernel go to raw IP protocol handler.
241  */
242 void
243 ip_init()
244 {
245 	register struct protosw *pr;
246 	register int i;
247 
248 	TAILQ_INIT(&in_ifaddrhead);
249 	in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask);
250 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
251 	if (pr == 0)
252 		panic("ip_init");
253 	for (i = 0; i < IPPROTO_MAX; i++)
254 		ip_protox[i] = pr - inetsw;
255 	for (pr = inetdomain.dom_protosw;
256 	    pr < inetdomain.dom_protoswNPROTOSW; pr++)
257 		if (pr->pr_domain->dom_family == PF_INET &&
258 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
259 			ip_protox[pr->pr_protocol] = pr - inetsw;
260 
261 	for (i = 0; i < IPREASS_NHASH; i++)
262 	    TAILQ_INIT(&ipq[i]);
263 
264 	maxnipq = nmbclusters / 32;
265 	maxfragsperpacket = 16;
266 
267 #ifndef RANDOM_IP_ID
268 	ip_id = time_second & 0xffff;
269 #endif
270 	ipintrq.ifq_maxlen = ipqmaxlen;
271 	mtx_init(&ipintrq.ifq_mtx, "ip_inq", NULL, MTX_DEF);
272 	ipintrq_present = 1;
273 
274 	register_netisr(NETISR_IP, ipintr);
275 }
276 
277 /*
278  * XXX watch out this one. It is perhaps used as a cache for
279  * the most recently used route ? it is cleared in in_addroute()
280  * when a new route is successfully created.
281  */
282 struct	route ipforward_rt;
283 
284 /*
285  * Ip input routine.  Checksum and byte swap header.  If fragmented
286  * try to reassemble.  Process options.  Pass to next level.
287  */
288 void
289 ip_input(struct mbuf *m)
290 {
291 	struct ip *ip;
292 	struct ipq *fp;
293 	struct in_ifaddr *ia = NULL;
294 	struct ifaddr *ifa;
295 	int    i, hlen, checkif;
296 	u_short sum;
297 	struct in_addr pkt_dst;
298 	u_int32_t divert_info = 0;		/* packet divert/tee info */
299 	struct ip_fw_args args;
300 #ifdef PFIL_HOOKS
301 	struct packet_filter_hook *pfh;
302 	struct mbuf *m0;
303 	int rv;
304 #endif /* PFIL_HOOKS */
305 #ifdef FAST_IPSEC
306 	struct m_tag *mtag;
307 	struct tdb_ident *tdbi;
308 	struct secpolicy *sp;
309 	int s, error;
310 #endif /* FAST_IPSEC */
311 
312 	args.eh = NULL;
313 	args.oif = NULL;
314 	args.rule = NULL;
315 	args.divert_rule = 0;			/* divert cookie */
316 	args.next_hop = NULL;
317 
318 	/* Grab info from MT_TAG mbufs prepended to the chain.	*/
319 	for (; m && m->m_type == MT_TAG; m = m->m_next) {
320 		switch(m->_m_tag_id) {
321 		default:
322 			printf("ip_input: unrecognised MT_TAG tag %d\n",
323 			    m->_m_tag_id);
324 			break;
325 
326 		case PACKET_TAG_DUMMYNET:
327 			args.rule = ((struct dn_pkt *)m)->rule;
328 			break;
329 
330 		case PACKET_TAG_DIVERT:
331 			args.divert_rule = (intptr_t)m->m_hdr.mh_data & 0xffff;
332 			break;
333 
334 		case PACKET_TAG_IPFORWARD:
335 			args.next_hop = (struct sockaddr_in *)m->m_hdr.mh_data;
336 			break;
337 		}
338 	}
339 
340 	KASSERT(m != NULL && (m->m_flags & M_PKTHDR) != 0,
341 	    ("ip_input: no HDR"));
342 
343 	if (args.rule) {	/* dummynet already filtered us */
344 		ip = mtod(m, struct ip *);
345 		hlen = ip->ip_hl << 2;
346 		goto iphack ;
347 	}
348 
349 	ipstat.ips_total++;
350 
351 	if (m->m_pkthdr.len < sizeof(struct ip))
352 		goto tooshort;
353 
354 	if (m->m_len < sizeof (struct ip) &&
355 	    (m = m_pullup(m, sizeof (struct ip))) == 0) {
356 		ipstat.ips_toosmall++;
357 		return;
358 	}
359 	ip = mtod(m, struct ip *);
360 
361 	if (ip->ip_v != IPVERSION) {
362 		ipstat.ips_badvers++;
363 		goto bad;
364 	}
365 
366 	hlen = ip->ip_hl << 2;
367 	if (hlen < sizeof(struct ip)) {	/* minimum header length */
368 		ipstat.ips_badhlen++;
369 		goto bad;
370 	}
371 	if (hlen > m->m_len) {
372 		if ((m = m_pullup(m, hlen)) == 0) {
373 			ipstat.ips_badhlen++;
374 			return;
375 		}
376 		ip = mtod(m, struct ip *);
377 	}
378 
379 	/* 127/8 must not appear on wire - RFC1122 */
380 	if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
381 	    (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
382 		if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
383 			ipstat.ips_badaddr++;
384 			goto bad;
385 		}
386 	}
387 
388 	if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
389 		sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
390 	} else {
391 		if (hlen == sizeof(struct ip)) {
392 			sum = in_cksum_hdr(ip);
393 		} else {
394 			sum = in_cksum(m, hlen);
395 		}
396 	}
397 	if (sum) {
398 		ipstat.ips_badsum++;
399 		goto bad;
400 	}
401 
402 	/*
403 	 * Convert fields to host representation.
404 	 */
405 	ip->ip_len = ntohs(ip->ip_len);
406 	if (ip->ip_len < hlen) {
407 		ipstat.ips_badlen++;
408 		goto bad;
409 	}
410 	ip->ip_off = ntohs(ip->ip_off);
411 
412 	/*
413 	 * Check that the amount of data in the buffers
414 	 * is as at least much as the IP header would have us expect.
415 	 * Trim mbufs if longer than we expect.
416 	 * Drop packet if shorter than we expect.
417 	 */
418 	if (m->m_pkthdr.len < ip->ip_len) {
419 tooshort:
420 		ipstat.ips_tooshort++;
421 		goto bad;
422 	}
423 	if (m->m_pkthdr.len > ip->ip_len) {
424 		if (m->m_len == m->m_pkthdr.len) {
425 			m->m_len = ip->ip_len;
426 			m->m_pkthdr.len = ip->ip_len;
427 		} else
428 			m_adj(m, ip->ip_len - m->m_pkthdr.len);
429 	}
430 #if defined(IPSEC) && !defined(IPSEC_FILTERGIF)
431 	/*
432 	 * Bypass packet filtering for packets from a tunnel (gif).
433 	 */
434 	if (ipsec_gethist(m, NULL))
435 		goto pass;
436 #endif
437 
438 	/*
439 	 * IpHack's section.
440 	 * Right now when no processing on packet has done
441 	 * and it is still fresh out of network we do our black
442 	 * deals with it.
443 	 * - Firewall: deny/allow/divert
444 	 * - Xlate: translate packet's addr/port (NAT).
445 	 * - Pipe: pass pkt through dummynet.
446 	 * - Wrap: fake packet's addr/port <unimpl.>
447 	 * - Encapsulate: put it in another IP and send out. <unimp.>
448  	 */
449 
450 iphack:
451 
452 #ifdef PFIL_HOOKS
453 	/*
454 	 * Run through list of hooks for input packets.  If there are any
455 	 * filters which require that additional packets in the flow are
456 	 * not fast-forwarded, they must clear the M_CANFASTFWD flag.
457 	 * Note that filters must _never_ set this flag, as another filter
458 	 * in the list may have previously cleared it.
459 	 */
460 	m0 = m;
461 	pfh = pfil_hook_get(PFIL_IN, &inetsw[ip_protox[IPPROTO_IP]].pr_pfh);
462 	for (; pfh; pfh = TAILQ_NEXT(pfh, pfil_link))
463 		if (pfh->pfil_func) {
464 			rv = pfh->pfil_func(ip, hlen,
465 					    m->m_pkthdr.rcvif, 0, &m0);
466 			if (rv)
467 				return;
468 			m = m0;
469 			if (m == NULL)
470 				return;
471 			ip = mtod(m, struct ip *);
472 		}
473 #endif /* PFIL_HOOKS */
474 
475 	if (fw_enable && IPFW_LOADED) {
476 		/*
477 		 * If we've been forwarded from the output side, then
478 		 * skip the firewall a second time
479 		 */
480 		if (args.next_hop)
481 			goto ours;
482 
483 		args.m = m;
484 		i = ip_fw_chk_ptr(&args);
485 		m = args.m;
486 
487 		if ( (i & IP_FW_PORT_DENY_FLAG) || m == NULL) { /* drop */
488 			if (m)
489 				m_freem(m);
490 			return;
491 		}
492 		ip = mtod(m, struct ip *); /* just in case m changed */
493 		if (i == 0 && args.next_hop == NULL)	/* common case */
494 			goto pass;
495                 if (DUMMYNET_LOADED && (i & IP_FW_PORT_DYNT_FLAG) != 0) {
496 			/* Send packet to the appropriate pipe */
497 			ip_dn_io_ptr(m, i&0xffff, DN_TO_IP_IN, &args);
498 			return;
499 		}
500 #ifdef IPDIVERT
501 		if (i != 0 && (i & IP_FW_PORT_DYNT_FLAG) == 0) {
502 			/* Divert or tee packet */
503 			divert_info = i;
504 			goto ours;
505 		}
506 #endif
507 		if (i == 0 && args.next_hop != NULL)
508 			goto pass;
509 		/*
510 		 * if we get here, the packet must be dropped
511 		 */
512 		m_freem(m);
513 		return;
514 	}
515 pass:
516 
517 	/*
518 	 * Process options and, if not destined for us,
519 	 * ship it on.  ip_dooptions returns 1 when an
520 	 * error was detected (causing an icmp message
521 	 * to be sent and the original packet to be freed).
522 	 */
523 	ip_nhops = 0;		/* for source routed packets */
524 	if (hlen > sizeof (struct ip) && ip_dooptions(m, 0, args.next_hop))
525 		return;
526 
527         /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
528          * matter if it is destined to another node, or whether it is
529          * a multicast one, RSVP wants it! and prevents it from being forwarded
530          * anywhere else. Also checks if the rsvp daemon is running before
531 	 * grabbing the packet.
532          */
533 	if (rsvp_on && ip->ip_p==IPPROTO_RSVP)
534 		goto ours;
535 
536 	/*
537 	 * Check our list of addresses, to see if the packet is for us.
538 	 * If we don't have any addresses, assume any unicast packet
539 	 * we receive might be for us (and let the upper layers deal
540 	 * with it).
541 	 */
542 	if (TAILQ_EMPTY(&in_ifaddrhead) &&
543 	    (m->m_flags & (M_MCAST|M_BCAST)) == 0)
544 		goto ours;
545 
546 	/*
547 	 * Cache the destination address of the packet; this may be
548 	 * changed by use of 'ipfw fwd'.
549 	 */
550 	pkt_dst = args.next_hop ? args.next_hop->sin_addr : ip->ip_dst;
551 
552 	/*
553 	 * Enable a consistency check between the destination address
554 	 * and the arrival interface for a unicast packet (the RFC 1122
555 	 * strong ES model) if IP forwarding is disabled and the packet
556 	 * is not locally generated and the packet is not subject to
557 	 * 'ipfw fwd'.
558 	 *
559 	 * XXX - Checking also should be disabled if the destination
560 	 * address is ipnat'ed to a different interface.
561 	 *
562 	 * XXX - Checking is incompatible with IP aliases added
563 	 * to the loopback interface instead of the interface where
564 	 * the packets are received.
565 	 */
566 	checkif = ip_checkinterface && (ipforwarding == 0) &&
567 	    m->m_pkthdr.rcvif != NULL &&
568 	    ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) &&
569 	    (args.next_hop == NULL);
570 
571 	/*
572 	 * Check for exact addresses in the hash bucket.
573 	 */
574 	LIST_FOREACH(ia, INADDR_HASH(pkt_dst.s_addr), ia_hash) {
575 		/*
576 		 * If the address matches, verify that the packet
577 		 * arrived via the correct interface if checking is
578 		 * enabled.
579 		 */
580 		if (IA_SIN(ia)->sin_addr.s_addr == pkt_dst.s_addr &&
581 		    (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif))
582 			goto ours;
583 	}
584 	/*
585 	 * Check for broadcast addresses.
586 	 *
587 	 * Only accept broadcast packets that arrive via the matching
588 	 * interface.  Reception of forwarded directed broadcasts would
589 	 * be handled via ip_forward() and ether_output() with the loopback
590 	 * into the stack for SIMPLEX interfaces handled by ether_output().
591 	 */
592 	if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
593 	        TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
594 			if (ifa->ifa_addr->sa_family != AF_INET)
595 				continue;
596 			ia = ifatoia(ifa);
597 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
598 			    pkt_dst.s_addr)
599 				goto ours;
600 			if (ia->ia_netbroadcast.s_addr == pkt_dst.s_addr)
601 				goto ours;
602 #ifdef BOOTP_COMPAT
603 			if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY)
604 				goto ours;
605 #endif
606 		}
607 	}
608 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
609 		struct in_multi *inm;
610 		if (ip_mrouter) {
611 			/*
612 			 * If we are acting as a multicast router, all
613 			 * incoming multicast packets are passed to the
614 			 * kernel-level multicast forwarding function.
615 			 * The packet is returned (relatively) intact; if
616 			 * ip_mforward() returns a non-zero value, the packet
617 			 * must be discarded, else it may be accepted below.
618 			 */
619 			if (ip_mforward &&
620 			    ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) {
621 				ipstat.ips_cantforward++;
622 				m_freem(m);
623 				return;
624 			}
625 
626 			/*
627 			 * The process-level routing daemon needs to receive
628 			 * all multicast IGMP packets, whether or not this
629 			 * host belongs to their destination groups.
630 			 */
631 			if (ip->ip_p == IPPROTO_IGMP)
632 				goto ours;
633 			ipstat.ips_forward++;
634 		}
635 		/*
636 		 * See if we belong to the destination multicast group on the
637 		 * arrival interface.
638 		 */
639 		IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
640 		if (inm == NULL) {
641 			ipstat.ips_notmember++;
642 			m_freem(m);
643 			return;
644 		}
645 		goto ours;
646 	}
647 	if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
648 		goto ours;
649 	if (ip->ip_dst.s_addr == INADDR_ANY)
650 		goto ours;
651 
652 	/*
653 	 * FAITH(Firewall Aided Internet Translator)
654 	 */
655 	if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
656 		if (ip_keepfaith) {
657 			if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP)
658 				goto ours;
659 		}
660 		m_freem(m);
661 		return;
662 	}
663 
664 	/*
665 	 * Not for us; forward if possible and desirable.
666 	 */
667 	if (ipforwarding == 0) {
668 		ipstat.ips_cantforward++;
669 		m_freem(m);
670 	} else {
671 #ifdef IPSEC
672 		/*
673 		 * Enforce inbound IPsec SPD.
674 		 */
675 		if (ipsec4_in_reject(m, NULL)) {
676 			ipsecstat.in_polvio++;
677 			goto bad;
678 		}
679 #endif /* IPSEC */
680 #ifdef FAST_IPSEC
681 		mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
682 		s = splnet();
683 		if (mtag != NULL) {
684 			tdbi = (struct tdb_ident *)(mtag + 1);
685 			sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
686 		} else {
687 			sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
688 						   IP_FORWARDING, &error);
689 		}
690 		if (sp == NULL) {	/* NB: can happen if error */
691 			splx(s);
692 			/*XXX error stat???*/
693 			DPRINTF(("ip_input: no SP for forwarding\n"));	/*XXX*/
694 			goto bad;
695 		}
696 
697 		/*
698 		 * Check security policy against packet attributes.
699 		 */
700 		error = ipsec_in_reject(sp, m);
701 		KEY_FREESP(&sp);
702 		splx(s);
703 		if (error) {
704 			ipstat.ips_cantforward++;
705 			goto bad;
706 		}
707 #endif /* FAST_IPSEC */
708 		ip_forward(m, 0, args.next_hop);
709 	}
710 	return;
711 
712 ours:
713 #ifdef IPSTEALTH
714 	/*
715 	 * IPSTEALTH: Process non-routing options only
716 	 * if the packet is destined for us.
717 	 */
718 	if (ipstealth && hlen > sizeof (struct ip) &&
719 	    ip_dooptions(m, 1, args.next_hop))
720 		return;
721 #endif /* IPSTEALTH */
722 
723 	/* Count the packet in the ip address stats */
724 	if (ia != NULL) {
725 		ia->ia_ifa.if_ipackets++;
726 		ia->ia_ifa.if_ibytes += m->m_pkthdr.len;
727 	}
728 
729 	/*
730 	 * If offset or IP_MF are set, must reassemble.
731 	 * Otherwise, nothing need be done.
732 	 * (We could look in the reassembly queue to see
733 	 * if the packet was previously fragmented,
734 	 * but it's not worth the time; just let them time out.)
735 	 */
736 	if (ip->ip_off & (IP_MF | IP_OFFMASK)) {
737 
738 		/* If maxnipq is 0, never accept fragments. */
739 		if (maxnipq == 0) {
740                 	ipstat.ips_fragments++;
741 			ipstat.ips_fragdropped++;
742 			goto bad;
743 		}
744 
745 		sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
746 		/*
747 		 * Look for queue of fragments
748 		 * of this datagram.
749 		 */
750 		TAILQ_FOREACH(fp, &ipq[sum], ipq_list)
751 			if (ip->ip_id == fp->ipq_id &&
752 			    ip->ip_src.s_addr == fp->ipq_src.s_addr &&
753 			    ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
754 #ifdef MAC
755 			    mac_fragment_match(m, fp) &&
756 #endif
757 			    ip->ip_p == fp->ipq_p)
758 				goto found;
759 
760 		fp = 0;
761 
762 		/*
763 		 * Enforce upper bound on number of fragmented packets
764 		 * for which we attempt reassembly;
765 		 * If maxnipq is -1, accept all fragments without limitation.
766 		 */
767 		if ((nipq > maxnipq) && (maxnipq > 0)) {
768 		    /*
769 		     * drop something from the tail of the current queue
770 		     * before proceeding further
771 		     */
772 		    struct ipq *q = TAILQ_LAST(&ipq[sum], ipqhead);
773 		    if (q == NULL) {   /* gak */
774 			for (i = 0; i < IPREASS_NHASH; i++) {
775 			    struct ipq *r = TAILQ_LAST(&ipq[i], ipqhead);
776 			    if (r) {
777 				ipstat.ips_fragtimeout += r->ipq_nfrags;
778 				ip_freef(&ipq[i], r);
779 				break;
780 			    }
781 			}
782 		    } else {
783 			ipstat.ips_fragtimeout += q->ipq_nfrags;
784 			ip_freef(&ipq[sum], q);
785 		    }
786 		}
787 found:
788 		/*
789 		 * Adjust ip_len to not reflect header,
790 		 * convert offset of this to bytes.
791 		 */
792 		ip->ip_len -= hlen;
793 		if (ip->ip_off & IP_MF) {
794 		        /*
795 		         * Make sure that fragments have a data length
796 			 * that's a non-zero multiple of 8 bytes.
797 		         */
798 			if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
799 				ipstat.ips_toosmall++; /* XXX */
800 				goto bad;
801 			}
802 			m->m_flags |= M_FRAG;
803 		} else
804 			m->m_flags &= ~M_FRAG;
805 		ip->ip_off <<= 3;
806 
807 		/*
808 		 * Attempt reassembly; if it succeeds, proceed.
809 		 * ip_reass() will return a different mbuf, and update
810 		 * the divert info in divert_info and args.divert_rule.
811 		 */
812 		ipstat.ips_fragments++;
813 		m->m_pkthdr.header = ip;
814 		m = ip_reass(m,
815 		    &ipq[sum], fp, &divert_info, &args.divert_rule);
816 		if (m == 0)
817 			return;
818 		ipstat.ips_reassembled++;
819 		ip = mtod(m, struct ip *);
820 		/* Get the header length of the reassembled packet */
821 		hlen = ip->ip_hl << 2;
822 #ifdef IPDIVERT
823 		/* Restore original checksum before diverting packet */
824 		if (divert_info != 0) {
825 			ip->ip_len += hlen;
826 			ip->ip_len = htons(ip->ip_len);
827 			ip->ip_off = htons(ip->ip_off);
828 			ip->ip_sum = 0;
829 			if (hlen == sizeof(struct ip))
830 				ip->ip_sum = in_cksum_hdr(ip);
831 			else
832 				ip->ip_sum = in_cksum(m, hlen);
833 			ip->ip_off = ntohs(ip->ip_off);
834 			ip->ip_len = ntohs(ip->ip_len);
835 			ip->ip_len -= hlen;
836 		}
837 #endif
838 	} else
839 		ip->ip_len -= hlen;
840 
841 #ifdef IPDIVERT
842 	/*
843 	 * Divert or tee packet to the divert protocol if required.
844 	 */
845 	if (divert_info != 0) {
846 		struct mbuf *clone = NULL;
847 
848 		/* Clone packet if we're doing a 'tee' */
849 		if ((divert_info & IP_FW_PORT_TEE_FLAG) != 0)
850 			clone = m_dup(m, M_DONTWAIT);
851 
852 		/* Restore packet header fields to original values */
853 		ip->ip_len += hlen;
854 		ip->ip_len = htons(ip->ip_len);
855 		ip->ip_off = htons(ip->ip_off);
856 
857 		/* Deliver packet to divert input routine */
858 		divert_packet(m, 1, divert_info & 0xffff, args.divert_rule);
859 		ipstat.ips_delivered++;
860 
861 		/* If 'tee', continue with original packet */
862 		if (clone == NULL)
863 			return;
864 		m = clone;
865 		ip = mtod(m, struct ip *);
866 		ip->ip_len += hlen;
867 		/*
868 		 * Jump backwards to complete processing of the
869 		 * packet. But first clear divert_info to avoid
870 		 * entering this block again.
871 		 * We do not need to clear args.divert_rule
872 		 * or args.next_hop as they will not be used.
873 		 */
874 		divert_info = 0;
875 		goto pass;
876 	}
877 #endif
878 
879 #ifdef IPSEC
880 	/*
881 	 * enforce IPsec policy checking if we are seeing last header.
882 	 * note that we do not visit this with protocols with pcb layer
883 	 * code - like udp/tcp/raw ip.
884 	 */
885 	if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0 &&
886 	    ipsec4_in_reject(m, NULL)) {
887 		ipsecstat.in_polvio++;
888 		goto bad;
889 	}
890 #endif
891 #if FAST_IPSEC
892 	/*
893 	 * enforce IPsec policy checking if we are seeing last header.
894 	 * note that we do not visit this with protocols with pcb layer
895 	 * code - like udp/tcp/raw ip.
896 	 */
897 	if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) {
898 		/*
899 		 * Check if the packet has already had IPsec processing
900 		 * done.  If so, then just pass it along.  This tag gets
901 		 * set during AH, ESP, etc. input handling, before the
902 		 * packet is returned to the ip input queue for delivery.
903 		 */
904 		mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
905 		s = splnet();
906 		if (mtag != NULL) {
907 			tdbi = (struct tdb_ident *)(mtag + 1);
908 			sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
909 		} else {
910 			sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
911 						   IP_FORWARDING, &error);
912 		}
913 		if (sp != NULL) {
914 			/*
915 			 * Check security policy against packet attributes.
916 			 */
917 			error = ipsec_in_reject(sp, m);
918 			KEY_FREESP(&sp);
919 		} else {
920 			/* XXX error stat??? */
921 			error = EINVAL;
922 DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/
923 			goto bad;
924 		}
925 		splx(s);
926 		if (error)
927 			goto bad;
928 	}
929 #endif /* FAST_IPSEC */
930 
931 	/*
932 	 * Switch out to protocol's input routine.
933 	 */
934 	ipstat.ips_delivered++;
935 	if (args.next_hop && ip->ip_p == IPPROTO_TCP) {
936 		/* TCP needs IPFORWARD info if available */
937 		struct m_hdr tag;
938 
939 		tag.mh_type = MT_TAG;
940 		tag.mh_flags = PACKET_TAG_IPFORWARD;
941 		tag.mh_data = (caddr_t)args.next_hop;
942 		tag.mh_next = m;
943 
944 		(*inetsw[ip_protox[ip->ip_p]].pr_input)(
945 			(struct mbuf *)&tag, hlen);
946 	} else
947 		(*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen);
948 	return;
949 bad:
950 	m_freem(m);
951 }
952 
953 /*
954  * IP software interrupt routine - to go away sometime soon
955  */
956 static void
957 ipintr(void)
958 {
959 	struct mbuf *m;
960 
961 	while (1) {
962 		IF_DEQUEUE(&ipintrq, m);
963 		if (m == 0)
964 			return;
965 		ip_input(m);
966 	}
967 }
968 
969 /*
970  * Take incoming datagram fragment and try to reassemble it into
971  * whole datagram.  If a chain for reassembly of this datagram already
972  * exists, then it is given as fp; otherwise have to make a chain.
973  *
974  * When IPDIVERT enabled, keep additional state with each packet that
975  * tells us if we need to divert or tee the packet we're building.
976  * In particular, *divinfo includes the port and TEE flag,
977  * *divert_rule is the number of the matching rule.
978  */
979 
980 static struct mbuf *
981 ip_reass(struct mbuf *m, struct ipqhead *head, struct ipq *fp,
982 	u_int32_t *divinfo, u_int16_t *divert_rule)
983 {
984 	struct ip *ip = mtod(m, struct ip *);
985 	register struct mbuf *p, *q, *nq;
986 	struct mbuf *t;
987 	int hlen = ip->ip_hl << 2;
988 	int i, next;
989 
990 	/*
991 	 * Presence of header sizes in mbufs
992 	 * would confuse code below.
993 	 */
994 	m->m_data += hlen;
995 	m->m_len -= hlen;
996 
997 	/*
998 	 * If first fragment to arrive, create a reassembly queue.
999 	 */
1000 	if (fp == 0) {
1001 		if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL)
1002 			goto dropfrag;
1003 		fp = mtod(t, struct ipq *);
1004 #ifdef MAC
1005 		mac_init_ipq(fp);
1006 		mac_create_ipq(m, fp);
1007 #endif
1008 		TAILQ_INSERT_HEAD(head, fp, ipq_list);
1009 		nipq++;
1010 		fp->ipq_nfrags = 1;
1011 		fp->ipq_ttl = IPFRAGTTL;
1012 		fp->ipq_p = ip->ip_p;
1013 		fp->ipq_id = ip->ip_id;
1014 		fp->ipq_src = ip->ip_src;
1015 		fp->ipq_dst = ip->ip_dst;
1016 		fp->ipq_frags = m;
1017 		m->m_nextpkt = NULL;
1018 #ifdef IPDIVERT
1019 		fp->ipq_div_info = 0;
1020 		fp->ipq_div_cookie = 0;
1021 #endif
1022 		goto inserted;
1023 	} else {
1024 		fp->ipq_nfrags++;
1025 #ifdef MAC
1026 		mac_update_ipq(m, fp);
1027 #endif
1028 	}
1029 
1030 #define GETIP(m)	((struct ip*)((m)->m_pkthdr.header))
1031 
1032 	/*
1033 	 * Find a segment which begins after this one does.
1034 	 */
1035 	for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt)
1036 		if (GETIP(q)->ip_off > ip->ip_off)
1037 			break;
1038 
1039 	/*
1040 	 * If there is a preceding segment, it may provide some of
1041 	 * our data already.  If so, drop the data from the incoming
1042 	 * segment.  If it provides all of our data, drop us, otherwise
1043 	 * stick new segment in the proper place.
1044 	 *
1045 	 * If some of the data is dropped from the the preceding
1046 	 * segment, then it's checksum is invalidated.
1047 	 */
1048 	if (p) {
1049 		i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off;
1050 		if (i > 0) {
1051 			if (i >= ip->ip_len)
1052 				goto dropfrag;
1053 			m_adj(m, i);
1054 			m->m_pkthdr.csum_flags = 0;
1055 			ip->ip_off += i;
1056 			ip->ip_len -= i;
1057 		}
1058 		m->m_nextpkt = p->m_nextpkt;
1059 		p->m_nextpkt = m;
1060 	} else {
1061 		m->m_nextpkt = fp->ipq_frags;
1062 		fp->ipq_frags = m;
1063 	}
1064 
1065 	/*
1066 	 * While we overlap succeeding segments trim them or,
1067 	 * if they are completely covered, dequeue them.
1068 	 */
1069 	for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off;
1070 	     q = nq) {
1071 		i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off;
1072 		if (i < GETIP(q)->ip_len) {
1073 			GETIP(q)->ip_len -= i;
1074 			GETIP(q)->ip_off += i;
1075 			m_adj(q, i);
1076 			q->m_pkthdr.csum_flags = 0;
1077 			break;
1078 		}
1079 		nq = q->m_nextpkt;
1080 		m->m_nextpkt = nq;
1081 		ipstat.ips_fragdropped++;
1082 		fp->ipq_nfrags--;
1083 		m_freem(q);
1084 	}
1085 
1086 inserted:
1087 
1088 #ifdef IPDIVERT
1089 	/*
1090 	 * Transfer firewall instructions to the fragment structure.
1091 	 * Only trust info in the fragment at offset 0.
1092 	 */
1093 	if (ip->ip_off == 0) {
1094 		fp->ipq_div_info = *divinfo;
1095 		fp->ipq_div_cookie = *divert_rule;
1096 	}
1097 	*divinfo = 0;
1098 	*divert_rule = 0;
1099 #endif
1100 
1101 	/*
1102 	 * Check for complete reassembly and perform frag per packet
1103 	 * limiting.
1104 	 *
1105 	 * Frag limiting is performed here so that the nth frag has
1106 	 * a chance to complete the packet before we drop the packet.
1107 	 * As a result, n+1 frags are actually allowed per packet, but
1108 	 * only n will ever be stored. (n = maxfragsperpacket.)
1109 	 *
1110 	 */
1111 	next = 0;
1112 	for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
1113 		if (GETIP(q)->ip_off != next) {
1114 			if (fp->ipq_nfrags > maxfragsperpacket) {
1115 				ipstat.ips_fragdropped += fp->ipq_nfrags;
1116 				ip_freef(head, fp);
1117 			}
1118 			return (0);
1119 		}
1120 		next += GETIP(q)->ip_len;
1121 	}
1122 	/* Make sure the last packet didn't have the IP_MF flag */
1123 	if (p->m_flags & M_FRAG) {
1124 		if (fp->ipq_nfrags > maxfragsperpacket) {
1125 			ipstat.ips_fragdropped += fp->ipq_nfrags;
1126 			ip_freef(head, fp);
1127 		}
1128 		return (0);
1129 	}
1130 
1131 	/*
1132 	 * Reassembly is complete.  Make sure the packet is a sane size.
1133 	 */
1134 	q = fp->ipq_frags;
1135 	ip = GETIP(q);
1136 	if (next + (ip->ip_hl << 2) > IP_MAXPACKET) {
1137 		ipstat.ips_toolong++;
1138 		ipstat.ips_fragdropped += fp->ipq_nfrags;
1139 		ip_freef(head, fp);
1140 		return (0);
1141 	}
1142 
1143 	/*
1144 	 * Concatenate fragments.
1145 	 */
1146 	m = q;
1147 	t = m->m_next;
1148 	m->m_next = 0;
1149 	m_cat(m, t);
1150 	nq = q->m_nextpkt;
1151 	q->m_nextpkt = 0;
1152 	for (q = nq; q != NULL; q = nq) {
1153 		nq = q->m_nextpkt;
1154 		q->m_nextpkt = NULL;
1155 		m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
1156 		m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
1157 		m_cat(m, q);
1158 	}
1159 #ifdef MAC
1160 	mac_create_datagram_from_ipq(fp, m);
1161 	mac_destroy_ipq(fp);
1162 #endif
1163 
1164 #ifdef IPDIVERT
1165 	/*
1166 	 * Extract firewall instructions from the fragment structure.
1167 	 */
1168 	*divinfo = fp->ipq_div_info;
1169 	*divert_rule = fp->ipq_div_cookie;
1170 #endif
1171 
1172 	/*
1173 	 * Create header for new ip packet by
1174 	 * modifying header of first packet;
1175 	 * dequeue and discard fragment reassembly header.
1176 	 * Make header visible.
1177 	 */
1178 	ip->ip_len = next;
1179 	ip->ip_src = fp->ipq_src;
1180 	ip->ip_dst = fp->ipq_dst;
1181 	TAILQ_REMOVE(head, fp, ipq_list);
1182 	nipq--;
1183 	(void) m_free(dtom(fp));
1184 	m->m_len += (ip->ip_hl << 2);
1185 	m->m_data -= (ip->ip_hl << 2);
1186 	/* some debugging cruft by sklower, below, will go away soon */
1187 	if (m->m_flags & M_PKTHDR)	/* XXX this should be done elsewhere */
1188 		m_fixhdr(m);
1189 	return (m);
1190 
1191 dropfrag:
1192 #ifdef IPDIVERT
1193 	*divinfo = 0;
1194 	*divert_rule = 0;
1195 #endif
1196 	ipstat.ips_fragdropped++;
1197 	if (fp != 0)
1198 		fp->ipq_nfrags--;
1199 	m_freem(m);
1200 	return (0);
1201 
1202 #undef GETIP
1203 }
1204 
1205 /*
1206  * Free a fragment reassembly header and all
1207  * associated datagrams.
1208  */
1209 static void
1210 ip_freef(fhp, fp)
1211 	struct ipqhead *fhp;
1212 	struct ipq *fp;
1213 {
1214 	register struct mbuf *q;
1215 
1216 	while (fp->ipq_frags) {
1217 		q = fp->ipq_frags;
1218 		fp->ipq_frags = q->m_nextpkt;
1219 		m_freem(q);
1220 	}
1221 	TAILQ_REMOVE(fhp, fp, ipq_list);
1222 	(void) m_free(dtom(fp));
1223 	nipq--;
1224 }
1225 
1226 /*
1227  * IP timer processing;
1228  * if a timer expires on a reassembly
1229  * queue, discard it.
1230  */
1231 void
1232 ip_slowtimo()
1233 {
1234 	register struct ipq *fp;
1235 	int s = splnet();
1236 	int i;
1237 
1238 	for (i = 0; i < IPREASS_NHASH; i++) {
1239 		for(fp = TAILQ_FIRST(&ipq[i]); fp;) {
1240 			struct ipq *fpp;
1241 
1242 			fpp = fp;
1243 			fp = TAILQ_NEXT(fp, ipq_list);
1244 			if(--fpp->ipq_ttl == 0) {
1245 				ipstat.ips_fragtimeout += fpp->ipq_nfrags;
1246 				ip_freef(&ipq[i], fpp);
1247 			}
1248 		}
1249 	}
1250 	/*
1251 	 * If we are over the maximum number of fragments
1252 	 * (due to the limit being lowered), drain off
1253 	 * enough to get down to the new limit.
1254 	 */
1255 	if (maxnipq >= 0 && nipq > maxnipq) {
1256 		for (i = 0; i < IPREASS_NHASH; i++) {
1257 			while (nipq > maxnipq && !TAILQ_EMPTY(&ipq[i])) {
1258 				ipstat.ips_fragdropped +=
1259 				    TAILQ_FIRST(&ipq[i])->ipq_nfrags;
1260 				ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i]));
1261 			}
1262 		}
1263 	}
1264 	ipflow_slowtimo();
1265 	splx(s);
1266 }
1267 
1268 /*
1269  * Drain off all datagram fragments.
1270  */
1271 void
1272 ip_drain()
1273 {
1274 	int     i;
1275 
1276 	for (i = 0; i < IPREASS_NHASH; i++) {
1277 		while(!TAILQ_EMPTY(&ipq[i])) {
1278 			ipstat.ips_fragdropped +=
1279 			    TAILQ_FIRST(&ipq[i])->ipq_nfrags;
1280 			ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i]));
1281 		}
1282 	}
1283 	in_rtqdrain();
1284 }
1285 
1286 /*
1287  * Do option processing on a datagram,
1288  * possibly discarding it if bad options are encountered,
1289  * or forwarding it if source-routed.
1290  * The pass argument is used when operating in the IPSTEALTH
1291  * mode to tell what options to process:
1292  * [LS]SRR (pass 0) or the others (pass 1).
1293  * The reason for as many as two passes is that when doing IPSTEALTH,
1294  * non-routing options should be processed only if the packet is for us.
1295  * Returns 1 if packet has been forwarded/freed,
1296  * 0 if the packet should be processed further.
1297  */
1298 static int
1299 ip_dooptions(struct mbuf *m, int pass, struct sockaddr_in *next_hop)
1300 {
1301 	struct ip *ip = mtod(m, struct ip *);
1302 	u_char *cp;
1303 	struct in_ifaddr *ia;
1304 	int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0;
1305 	struct in_addr *sin, dst;
1306 	n_time ntime;
1307 	struct	sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET };
1308 
1309 	dst = ip->ip_dst;
1310 	cp = (u_char *)(ip + 1);
1311 	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
1312 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1313 		opt = cp[IPOPT_OPTVAL];
1314 		if (opt == IPOPT_EOL)
1315 			break;
1316 		if (opt == IPOPT_NOP)
1317 			optlen = 1;
1318 		else {
1319 			if (cnt < IPOPT_OLEN + sizeof(*cp)) {
1320 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1321 				goto bad;
1322 			}
1323 			optlen = cp[IPOPT_OLEN];
1324 			if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
1325 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1326 				goto bad;
1327 			}
1328 		}
1329 		switch (opt) {
1330 
1331 		default:
1332 			break;
1333 
1334 		/*
1335 		 * Source routing with record.
1336 		 * Find interface with current destination address.
1337 		 * If none on this machine then drop if strictly routed,
1338 		 * or do nothing if loosely routed.
1339 		 * Record interface address and bring up next address
1340 		 * component.  If strictly routed make sure next
1341 		 * address is on directly accessible net.
1342 		 */
1343 		case IPOPT_LSRR:
1344 		case IPOPT_SSRR:
1345 #ifdef IPSTEALTH
1346 			if (ipstealth && pass > 0)
1347 				break;
1348 #endif
1349 			if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1350 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1351 				goto bad;
1352 			}
1353 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1354 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1355 				goto bad;
1356 			}
1357 			ipaddr.sin_addr = ip->ip_dst;
1358 			ia = (struct in_ifaddr *)
1359 				ifa_ifwithaddr((struct sockaddr *)&ipaddr);
1360 			if (ia == 0) {
1361 				if (opt == IPOPT_SSRR) {
1362 					type = ICMP_UNREACH;
1363 					code = ICMP_UNREACH_SRCFAIL;
1364 					goto bad;
1365 				}
1366 				if (!ip_dosourceroute)
1367 					goto nosourcerouting;
1368 				/*
1369 				 * Loose routing, and not at next destination
1370 				 * yet; nothing to do except forward.
1371 				 */
1372 				break;
1373 			}
1374 			off--;			/* 0 origin */
1375 			if (off > optlen - (int)sizeof(struct in_addr)) {
1376 				/*
1377 				 * End of source route.  Should be for us.
1378 				 */
1379 				if (!ip_acceptsourceroute)
1380 					goto nosourcerouting;
1381 				save_rte(cp, ip->ip_src);
1382 				break;
1383 			}
1384 #ifdef IPSTEALTH
1385 			if (ipstealth)
1386 				goto dropit;
1387 #endif
1388 			if (!ip_dosourceroute) {
1389 				if (ipforwarding) {
1390 					char buf[16]; /* aaa.bbb.ccc.ddd\0 */
1391 					/*
1392 					 * Acting as a router, so generate ICMP
1393 					 */
1394 nosourcerouting:
1395 					strcpy(buf, inet_ntoa(ip->ip_dst));
1396 					log(LOG_WARNING,
1397 					    "attempted source route from %s to %s\n",
1398 					    inet_ntoa(ip->ip_src), buf);
1399 					type = ICMP_UNREACH;
1400 					code = ICMP_UNREACH_SRCFAIL;
1401 					goto bad;
1402 				} else {
1403 					/*
1404 					 * Not acting as a router, so silently drop.
1405 					 */
1406 #ifdef IPSTEALTH
1407 dropit:
1408 #endif
1409 					ipstat.ips_cantforward++;
1410 					m_freem(m);
1411 					return (1);
1412 				}
1413 			}
1414 
1415 			/*
1416 			 * locate outgoing interface
1417 			 */
1418 			(void)memcpy(&ipaddr.sin_addr, cp + off,
1419 			    sizeof(ipaddr.sin_addr));
1420 
1421 			if (opt == IPOPT_SSRR) {
1422 #define	INA	struct in_ifaddr *
1423 #define	SA	struct sockaddr *
1424 			    if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == 0)
1425 				ia = (INA)ifa_ifwithnet((SA)&ipaddr);
1426 			} else
1427 				ia = ip_rtaddr(ipaddr.sin_addr, &ipforward_rt);
1428 			if (ia == 0) {
1429 				type = ICMP_UNREACH;
1430 				code = ICMP_UNREACH_SRCFAIL;
1431 				goto bad;
1432 			}
1433 			ip->ip_dst = ipaddr.sin_addr;
1434 			(void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
1435 			    sizeof(struct in_addr));
1436 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1437 			/*
1438 			 * Let ip_intr's mcast routing check handle mcast pkts
1439 			 */
1440 			forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr));
1441 			break;
1442 
1443 		case IPOPT_RR:
1444 #ifdef IPSTEALTH
1445 			if (ipstealth && pass == 0)
1446 				break;
1447 #endif
1448 			if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1449 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1450 				goto bad;
1451 			}
1452 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1453 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1454 				goto bad;
1455 			}
1456 			/*
1457 			 * If no space remains, ignore.
1458 			 */
1459 			off--;			/* 0 origin */
1460 			if (off > optlen - (int)sizeof(struct in_addr))
1461 				break;
1462 			(void)memcpy(&ipaddr.sin_addr, &ip->ip_dst,
1463 			    sizeof(ipaddr.sin_addr));
1464 			/*
1465 			 * locate outgoing interface; if we're the destination,
1466 			 * use the incoming interface (should be same).
1467 			 */
1468 			if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == 0 &&
1469 			    (ia = ip_rtaddr(ipaddr.sin_addr,
1470 			    &ipforward_rt)) == 0) {
1471 				type = ICMP_UNREACH;
1472 				code = ICMP_UNREACH_HOST;
1473 				goto bad;
1474 			}
1475 			(void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
1476 			    sizeof(struct in_addr));
1477 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1478 			break;
1479 
1480 		case IPOPT_TS:
1481 #ifdef IPSTEALTH
1482 			if (ipstealth && pass == 0)
1483 				break;
1484 #endif
1485 			code = cp - (u_char *)ip;
1486 			if (optlen < 4 || optlen > 40) {
1487 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1488 				goto bad;
1489 			}
1490 			if ((off = cp[IPOPT_OFFSET]) < 5) {
1491 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1492 				goto bad;
1493 			}
1494 			if (off > optlen - (int)sizeof(int32_t)) {
1495 				cp[IPOPT_OFFSET + 1] += (1 << 4);
1496 				if ((cp[IPOPT_OFFSET + 1] & 0xf0) == 0) {
1497 					code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1498 					goto bad;
1499 				}
1500 				break;
1501 			}
1502 			off--;				/* 0 origin */
1503 			sin = (struct in_addr *)(cp + off);
1504 			switch (cp[IPOPT_OFFSET + 1] & 0x0f) {
1505 
1506 			case IPOPT_TS_TSONLY:
1507 				break;
1508 
1509 			case IPOPT_TS_TSANDADDR:
1510 				if (off + sizeof(n_time) +
1511 				    sizeof(struct in_addr) > optlen) {
1512 					code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1513 					goto bad;
1514 				}
1515 				ipaddr.sin_addr = dst;
1516 				ia = (INA)ifaof_ifpforaddr((SA)&ipaddr,
1517 							    m->m_pkthdr.rcvif);
1518 				if (ia == 0)
1519 					continue;
1520 				(void)memcpy(sin, &IA_SIN(ia)->sin_addr,
1521 				    sizeof(struct in_addr));
1522 				cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1523 				off += sizeof(struct in_addr);
1524 				break;
1525 
1526 			case IPOPT_TS_PRESPEC:
1527 				if (off + sizeof(n_time) +
1528 				    sizeof(struct in_addr) > optlen) {
1529 					code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1530 					goto bad;
1531 				}
1532 				(void)memcpy(&ipaddr.sin_addr, sin,
1533 				    sizeof(struct in_addr));
1534 				if (ifa_ifwithaddr((SA)&ipaddr) == 0)
1535 					continue;
1536 				cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1537 				off += sizeof(struct in_addr);
1538 				break;
1539 
1540 			default:
1541 				code = &cp[IPOPT_OFFSET + 1] - (u_char *)ip;
1542 				goto bad;
1543 			}
1544 			ntime = iptime();
1545 			(void)memcpy(cp + off, &ntime, sizeof(n_time));
1546 			cp[IPOPT_OFFSET] += sizeof(n_time);
1547 		}
1548 	}
1549 	if (forward && ipforwarding) {
1550 		ip_forward(m, 1, next_hop);
1551 		return (1);
1552 	}
1553 	return (0);
1554 bad:
1555 	icmp_error(m, type, code, 0, 0);
1556 	ipstat.ips_badoptions++;
1557 	return (1);
1558 }
1559 
1560 /*
1561  * Given address of next destination (final or next hop),
1562  * return internet address info of interface to be used to get there.
1563  */
1564 struct in_ifaddr *
1565 ip_rtaddr(dst, rt)
1566 	struct in_addr dst;
1567 	struct route *rt;
1568 {
1569 	register struct sockaddr_in *sin;
1570 
1571 	sin = (struct sockaddr_in *)&rt->ro_dst;
1572 
1573 	if (rt->ro_rt == 0 ||
1574 	    !(rt->ro_rt->rt_flags & RTF_UP) ||
1575 	    dst.s_addr != sin->sin_addr.s_addr) {
1576 		if (rt->ro_rt) {
1577 			RTFREE(rt->ro_rt);
1578 			rt->ro_rt = 0;
1579 		}
1580 		sin->sin_family = AF_INET;
1581 		sin->sin_len = sizeof(*sin);
1582 		sin->sin_addr = dst;
1583 
1584 		rtalloc_ign(rt, RTF_PRCLONING);
1585 	}
1586 	if (rt->ro_rt == 0)
1587 		return ((struct in_ifaddr *)0);
1588 	return (ifatoia(rt->ro_rt->rt_ifa));
1589 }
1590 
1591 /*
1592  * Save incoming source route for use in replies,
1593  * to be picked up later by ip_srcroute if the receiver is interested.
1594  */
1595 static void
1596 save_rte(option, dst)
1597 	u_char *option;
1598 	struct in_addr dst;
1599 {
1600 	unsigned olen;
1601 
1602 	olen = option[IPOPT_OLEN];
1603 #ifdef DIAGNOSTIC
1604 	if (ipprintfs)
1605 		printf("save_rte: olen %d\n", olen);
1606 #endif
1607 	if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst)))
1608 		return;
1609 	bcopy(option, ip_srcrt.srcopt, olen);
1610 	ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1611 	ip_srcrt.dst = dst;
1612 }
1613 
1614 /*
1615  * Retrieve incoming source route for use in replies,
1616  * in the same form used by setsockopt.
1617  * The first hop is placed before the options, will be removed later.
1618  */
1619 struct mbuf *
1620 ip_srcroute()
1621 {
1622 	register struct in_addr *p, *q;
1623 	register struct mbuf *m;
1624 
1625 	if (ip_nhops == 0)
1626 		return ((struct mbuf *)0);
1627 	m = m_get(M_DONTWAIT, MT_HEADER);
1628 	if (m == 0)
1629 		return ((struct mbuf *)0);
1630 
1631 #define OPTSIZ	(sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt))
1632 
1633 	/* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
1634 	m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) +
1635 	    OPTSIZ;
1636 #ifdef DIAGNOSTIC
1637 	if (ipprintfs)
1638 		printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len);
1639 #endif
1640 
1641 	/*
1642 	 * First save first hop for return route
1643 	 */
1644 	p = &ip_srcrt.route[ip_nhops - 1];
1645 	*(mtod(m, struct in_addr *)) = *p--;
1646 #ifdef DIAGNOSTIC
1647 	if (ipprintfs)
1648 		printf(" hops %lx", (u_long)ntohl(mtod(m, struct in_addr *)->s_addr));
1649 #endif
1650 
1651 	/*
1652 	 * Copy option fields and padding (nop) to mbuf.
1653 	 */
1654 	ip_srcrt.nop = IPOPT_NOP;
1655 	ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
1656 	(void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr),
1657 	    &ip_srcrt.nop, OPTSIZ);
1658 	q = (struct in_addr *)(mtod(m, caddr_t) +
1659 	    sizeof(struct in_addr) + OPTSIZ);
1660 #undef OPTSIZ
1661 	/*
1662 	 * Record return path as an IP source route,
1663 	 * reversing the path (pointers are now aligned).
1664 	 */
1665 	while (p >= ip_srcrt.route) {
1666 #ifdef DIAGNOSTIC
1667 		if (ipprintfs)
1668 			printf(" %lx", (u_long)ntohl(q->s_addr));
1669 #endif
1670 		*q++ = *p--;
1671 	}
1672 	/*
1673 	 * Last hop goes to final destination.
1674 	 */
1675 	*q = ip_srcrt.dst;
1676 #ifdef DIAGNOSTIC
1677 	if (ipprintfs)
1678 		printf(" %lx\n", (u_long)ntohl(q->s_addr));
1679 #endif
1680 	return (m);
1681 }
1682 
1683 /*
1684  * Strip out IP options, at higher
1685  * level protocol in the kernel.
1686  * Second argument is buffer to which options
1687  * will be moved, and return value is their length.
1688  * XXX should be deleted; last arg currently ignored.
1689  */
1690 void
1691 ip_stripoptions(m, mopt)
1692 	register struct mbuf *m;
1693 	struct mbuf *mopt;
1694 {
1695 	register int i;
1696 	struct ip *ip = mtod(m, struct ip *);
1697 	register caddr_t opts;
1698 	int olen;
1699 
1700 	olen = (ip->ip_hl << 2) - sizeof (struct ip);
1701 	opts = (caddr_t)(ip + 1);
1702 	i = m->m_len - (sizeof (struct ip) + olen);
1703 	bcopy(opts + olen, opts, (unsigned)i);
1704 	m->m_len -= olen;
1705 	if (m->m_flags & M_PKTHDR)
1706 		m->m_pkthdr.len -= olen;
1707 	ip->ip_v = IPVERSION;
1708 	ip->ip_hl = sizeof(struct ip) >> 2;
1709 }
1710 
1711 u_char inetctlerrmap[PRC_NCMDS] = {
1712 	0,		0,		0,		0,
1713 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1714 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1715 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1716 	0,		0,		0,		0,
1717 	ENOPROTOOPT,	ECONNREFUSED
1718 };
1719 
1720 /*
1721  * Forward a packet.  If some error occurs return the sender
1722  * an icmp packet.  Note we can't always generate a meaningful
1723  * icmp message because icmp doesn't have a large enough repertoire
1724  * of codes and types.
1725  *
1726  * If not forwarding, just drop the packet.  This could be confusing
1727  * if ipforwarding was zero but some routing protocol was advancing
1728  * us as a gateway to somewhere.  However, we must let the routing
1729  * protocol deal with that.
1730  *
1731  * The srcrt parameter indicates whether the packet is being forwarded
1732  * via a source route.
1733  */
1734 static void
1735 ip_forward(struct mbuf *m, int srcrt, struct sockaddr_in *next_hop)
1736 {
1737 	struct ip *ip = mtod(m, struct ip *);
1738 	struct rtentry *rt;
1739 	int error, type = 0, code = 0;
1740 	struct mbuf *mcopy;
1741 	n_long dest;
1742 	struct in_addr pkt_dst;
1743 	struct ifnet *destifp;
1744 #if defined(IPSEC) || defined(FAST_IPSEC)
1745 	struct ifnet dummyifp;
1746 #endif
1747 
1748 	dest = 0;
1749 	/*
1750 	 * Cache the destination address of the packet; this may be
1751 	 * changed by use of 'ipfw fwd'.
1752 	 */
1753 	pkt_dst = next_hop ? next_hop->sin_addr : ip->ip_dst;
1754 
1755 #ifdef DIAGNOSTIC
1756 	if (ipprintfs)
1757 		printf("forward: src %lx dst %lx ttl %x\n",
1758 		    (u_long)ip->ip_src.s_addr, (u_long)pkt_dst.s_addr,
1759 		    ip->ip_ttl);
1760 #endif
1761 
1762 
1763 	if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(pkt_dst) == 0) {
1764 		ipstat.ips_cantforward++;
1765 		m_freem(m);
1766 		return;
1767 	}
1768 #ifdef IPSTEALTH
1769 	if (!ipstealth) {
1770 #endif
1771 		if (ip->ip_ttl <= IPTTLDEC) {
1772 			icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS,
1773 			    dest, 0);
1774 			return;
1775 		}
1776 #ifdef IPSTEALTH
1777 	}
1778 #endif
1779 
1780 	if (ip_rtaddr(pkt_dst, &ipforward_rt) == 0) {
1781 		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0);
1782 		return;
1783 	} else
1784 		rt = ipforward_rt.ro_rt;
1785 
1786 	/*
1787 	 * Save the IP header and at most 8 bytes of the payload,
1788 	 * in case we need to generate an ICMP message to the src.
1789 	 *
1790 	 * XXX this can be optimized a lot by saving the data in a local
1791 	 * buffer on the stack (72 bytes at most), and only allocating the
1792 	 * mbuf if really necessary. The vast majority of the packets
1793 	 * are forwarded without having to send an ICMP back (either
1794 	 * because unnecessary, or because rate limited), so we are
1795 	 * really we are wasting a lot of work here.
1796 	 *
1797 	 * We don't use m_copy() because it might return a reference
1798 	 * to a shared cluster. Both this function and ip_output()
1799 	 * assume exclusive access to the IP header in `m', so any
1800 	 * data in a cluster may change before we reach icmp_error().
1801 	 */
1802 	MGET(mcopy, M_DONTWAIT, m->m_type);
1803 	if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_DONTWAIT)) {
1804 		/*
1805 		 * It's probably ok if the pkthdr dup fails (because
1806 		 * the deep copy of the tag chain failed), but for now
1807 		 * be conservative and just discard the copy since
1808 		 * code below may some day want the tags.
1809 		 */
1810 		m_free(mcopy);
1811 		mcopy = NULL;
1812 	}
1813 	if (mcopy != NULL) {
1814 		mcopy->m_len = imin((ip->ip_hl << 2) + 8,
1815 		    (int)ip->ip_len);
1816 		m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1817 #ifdef MAC
1818 		/*
1819 		 * XXXMAC: This will eventually become an explicit
1820 		 * labeling point.
1821 		 */
1822 		mac_create_mbuf_from_mbuf(m, mcopy);
1823 #endif
1824 	}
1825 
1826 #ifdef IPSTEALTH
1827 	if (!ipstealth) {
1828 #endif
1829 		ip->ip_ttl -= IPTTLDEC;
1830 #ifdef IPSTEALTH
1831 	}
1832 #endif
1833 
1834 	/*
1835 	 * If forwarding packet using same interface that it came in on,
1836 	 * perhaps should send a redirect to sender to shortcut a hop.
1837 	 * Only send redirect if source is sending directly to us,
1838 	 * and if packet was not source routed (or has any options).
1839 	 * Also, don't send redirect if forwarding using a default route
1840 	 * or a route modified by a redirect.
1841 	 */
1842 	if (rt->rt_ifp == m->m_pkthdr.rcvif &&
1843 	    (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1844 	    satosin(rt_key(rt))->sin_addr.s_addr != 0 &&
1845 	    ipsendredirects && !srcrt && !next_hop) {
1846 #define	RTA(rt)	((struct in_ifaddr *)(rt->rt_ifa))
1847 		u_long src = ntohl(ip->ip_src.s_addr);
1848 
1849 		if (RTA(rt) &&
1850 		    (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1851 		    if (rt->rt_flags & RTF_GATEWAY)
1852 			dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
1853 		    else
1854 			dest = pkt_dst.s_addr;
1855 		    /* Router requirements says to only send host redirects */
1856 		    type = ICMP_REDIRECT;
1857 		    code = ICMP_REDIRECT_HOST;
1858 #ifdef DIAGNOSTIC
1859 		    if (ipprintfs)
1860 		        printf("redirect (%d) to %lx\n", code, (u_long)dest);
1861 #endif
1862 		}
1863 	}
1864 
1865     {
1866 	struct m_hdr tag;
1867 
1868 	if (next_hop) {
1869 		/* Pass IPFORWARD info if available */
1870 
1871 		tag.mh_type = MT_TAG;
1872 		tag.mh_flags = PACKET_TAG_IPFORWARD;
1873 		tag.mh_data = (caddr_t)next_hop;
1874 		tag.mh_next = m;
1875 		m = (struct mbuf *)&tag;
1876 	}
1877 	error = ip_output(m, (struct mbuf *)0, &ipforward_rt,
1878 			  IP_FORWARDING, 0, NULL);
1879     }
1880 	if (error)
1881 		ipstat.ips_cantforward++;
1882 	else {
1883 		ipstat.ips_forward++;
1884 		if (type)
1885 			ipstat.ips_redirectsent++;
1886 		else {
1887 			if (mcopy) {
1888 				ipflow_create(&ipforward_rt, mcopy);
1889 				m_freem(mcopy);
1890 			}
1891 			return;
1892 		}
1893 	}
1894 	if (mcopy == NULL)
1895 		return;
1896 	destifp = NULL;
1897 
1898 	switch (error) {
1899 
1900 	case 0:				/* forwarded, but need redirect */
1901 		/* type, code set above */
1902 		break;
1903 
1904 	case ENETUNREACH:		/* shouldn't happen, checked above */
1905 	case EHOSTUNREACH:
1906 	case ENETDOWN:
1907 	case EHOSTDOWN:
1908 	default:
1909 		type = ICMP_UNREACH;
1910 		code = ICMP_UNREACH_HOST;
1911 		break;
1912 
1913 	case EMSGSIZE:
1914 		type = ICMP_UNREACH;
1915 		code = ICMP_UNREACH_NEEDFRAG;
1916 #ifdef IPSEC
1917 		/*
1918 		 * If the packet is routed over IPsec tunnel, tell the
1919 		 * originator the tunnel MTU.
1920 		 *	tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
1921 		 * XXX quickhack!!!
1922 		 */
1923 		if (ipforward_rt.ro_rt) {
1924 			struct secpolicy *sp = NULL;
1925 			int ipsecerror;
1926 			int ipsechdr;
1927 			struct route *ro;
1928 
1929 			sp = ipsec4_getpolicybyaddr(mcopy,
1930 						    IPSEC_DIR_OUTBOUND,
1931 			                            IP_FORWARDING,
1932 			                            &ipsecerror);
1933 
1934 			if (sp == NULL)
1935 				destifp = ipforward_rt.ro_rt->rt_ifp;
1936 			else {
1937 				/* count IPsec header size */
1938 				ipsechdr = ipsec4_hdrsiz(mcopy,
1939 							 IPSEC_DIR_OUTBOUND,
1940 							 NULL);
1941 
1942 				/*
1943 				 * find the correct route for outer IPv4
1944 				 * header, compute tunnel MTU.
1945 				 *
1946 				 * XXX BUG ALERT
1947 				 * The "dummyifp" code relies upon the fact
1948 				 * that icmp_error() touches only ifp->if_mtu.
1949 				 */
1950 				/*XXX*/
1951 				destifp = NULL;
1952 				if (sp->req != NULL
1953 				 && sp->req->sav != NULL
1954 				 && sp->req->sav->sah != NULL) {
1955 					ro = &sp->req->sav->sah->sa_route;
1956 					if (ro->ro_rt && ro->ro_rt->rt_ifp) {
1957 						dummyifp.if_mtu =
1958 						    ro->ro_rt->rt_ifp->if_mtu;
1959 						dummyifp.if_mtu -= ipsechdr;
1960 						destifp = &dummyifp;
1961 					}
1962 				}
1963 
1964 				key_freesp(sp);
1965 			}
1966 		}
1967 #elif FAST_IPSEC
1968 		/*
1969 		 * If the packet is routed over IPsec tunnel, tell the
1970 		 * originator the tunnel MTU.
1971 		 *	tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
1972 		 * XXX quickhack!!!
1973 		 */
1974 		if (ipforward_rt.ro_rt) {
1975 			struct secpolicy *sp = NULL;
1976 			int ipsecerror;
1977 			int ipsechdr;
1978 			struct route *ro;
1979 
1980 			sp = ipsec_getpolicybyaddr(mcopy,
1981 						   IPSEC_DIR_OUTBOUND,
1982 			                           IP_FORWARDING,
1983 			                           &ipsecerror);
1984 
1985 			if (sp == NULL)
1986 				destifp = ipforward_rt.ro_rt->rt_ifp;
1987 			else {
1988 				/* count IPsec header size */
1989 				ipsechdr = ipsec4_hdrsiz(mcopy,
1990 							 IPSEC_DIR_OUTBOUND,
1991 							 NULL);
1992 
1993 				/*
1994 				 * find the correct route for outer IPv4
1995 				 * header, compute tunnel MTU.
1996 				 *
1997 				 * XXX BUG ALERT
1998 				 * The "dummyifp" code relies upon the fact
1999 				 * that icmp_error() touches only ifp->if_mtu.
2000 				 */
2001 				/*XXX*/
2002 				destifp = NULL;
2003 				if (sp->req != NULL
2004 				 && sp->req->sav != NULL
2005 				 && sp->req->sav->sah != NULL) {
2006 					ro = &sp->req->sav->sah->sa_route;
2007 					if (ro->ro_rt && ro->ro_rt->rt_ifp) {
2008 						dummyifp.if_mtu =
2009 						    ro->ro_rt->rt_ifp->if_mtu;
2010 						dummyifp.if_mtu -= ipsechdr;
2011 						destifp = &dummyifp;
2012 					}
2013 				}
2014 
2015 				KEY_FREESP(&sp);
2016 			}
2017 		}
2018 #else /* !IPSEC && !FAST_IPSEC */
2019 		if (ipforward_rt.ro_rt)
2020 			destifp = ipforward_rt.ro_rt->rt_ifp;
2021 #endif /*IPSEC*/
2022 		ipstat.ips_cantfrag++;
2023 		break;
2024 
2025 	case ENOBUFS:
2026 		/*
2027 		 * A router should not generate ICMP_SOURCEQUENCH as
2028 		 * required in RFC1812 Requirements for IP Version 4 Routers.
2029 		 * Source quench could be a big problem under DoS attacks,
2030 		 * or if the underlying interface is rate-limited.
2031 		 * Those who need source quench packets may re-enable them
2032 		 * via the net.inet.ip.sendsourcequench sysctl.
2033 		 */
2034 		if (ip_sendsourcequench == 0) {
2035 			m_freem(mcopy);
2036 			return;
2037 		} else {
2038 			type = ICMP_SOURCEQUENCH;
2039 			code = 0;
2040 		}
2041 		break;
2042 
2043 	case EACCES:			/* ipfw denied packet */
2044 		m_freem(mcopy);
2045 		return;
2046 	}
2047 	icmp_error(mcopy, type, code, dest, destifp);
2048 }
2049 
2050 void
2051 ip_savecontrol(inp, mp, ip, m)
2052 	register struct inpcb *inp;
2053 	register struct mbuf **mp;
2054 	register struct ip *ip;
2055 	register struct mbuf *m;
2056 {
2057 	if (inp->inp_socket->so_options & SO_TIMESTAMP) {
2058 		struct timeval tv;
2059 
2060 		microtime(&tv);
2061 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
2062 			SCM_TIMESTAMP, SOL_SOCKET);
2063 		if (*mp)
2064 			mp = &(*mp)->m_next;
2065 	}
2066 	if (inp->inp_flags & INP_RECVDSTADDR) {
2067 		*mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
2068 		    sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
2069 		if (*mp)
2070 			mp = &(*mp)->m_next;
2071 	}
2072 #ifdef notyet
2073 	/* XXX
2074 	 * Moving these out of udp_input() made them even more broken
2075 	 * than they already were.
2076 	 */
2077 	/* options were tossed already */
2078 	if (inp->inp_flags & INP_RECVOPTS) {
2079 		*mp = sbcreatecontrol((caddr_t) opts_deleted_above,
2080 		    sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
2081 		if (*mp)
2082 			mp = &(*mp)->m_next;
2083 	}
2084 	/* ip_srcroute doesn't do what we want here, need to fix */
2085 	if (inp->inp_flags & INP_RECVRETOPTS) {
2086 		*mp = sbcreatecontrol((caddr_t) ip_srcroute(),
2087 		    sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
2088 		if (*mp)
2089 			mp = &(*mp)->m_next;
2090 	}
2091 #endif
2092 	if (inp->inp_flags & INP_RECVIF) {
2093 		struct ifnet *ifp;
2094 		struct sdlbuf {
2095 			struct sockaddr_dl sdl;
2096 			u_char	pad[32];
2097 		} sdlbuf;
2098 		struct sockaddr_dl *sdp;
2099 		struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
2100 
2101 		if (((ifp = m->m_pkthdr.rcvif))
2102 		&& ( ifp->if_index && (ifp->if_index <= if_index))) {
2103 			sdp = (struct sockaddr_dl *)
2104 			    (ifaddr_byindex(ifp->if_index)->ifa_addr);
2105 			/*
2106 			 * Change our mind and don't try copy.
2107 			 */
2108 			if ((sdp->sdl_family != AF_LINK)
2109 			|| (sdp->sdl_len > sizeof(sdlbuf))) {
2110 				goto makedummy;
2111 			}
2112 			bcopy(sdp, sdl2, sdp->sdl_len);
2113 		} else {
2114 makedummy:
2115 			sdl2->sdl_len
2116 				= offsetof(struct sockaddr_dl, sdl_data[0]);
2117 			sdl2->sdl_family = AF_LINK;
2118 			sdl2->sdl_index = 0;
2119 			sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
2120 		}
2121 		*mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len,
2122 			IP_RECVIF, IPPROTO_IP);
2123 		if (*mp)
2124 			mp = &(*mp)->m_next;
2125 	}
2126 }
2127 
2128 /*
2129  * XXX these routines are called from the upper part of the kernel.
2130  * They need to be locked when we remove Giant.
2131  *
2132  * They could also be moved to ip_mroute.c, since all the RSVP
2133  *  handling is done there already.
2134  */
2135 static int ip_rsvp_on;
2136 struct socket *ip_rsvpd;
2137 int
2138 ip_rsvp_init(struct socket *so)
2139 {
2140 	if (so->so_type != SOCK_RAW ||
2141 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
2142 		return EOPNOTSUPP;
2143 
2144 	if (ip_rsvpd != NULL)
2145 		return EADDRINUSE;
2146 
2147 	ip_rsvpd = so;
2148 	/*
2149 	 * This may seem silly, but we need to be sure we don't over-increment
2150 	 * the RSVP counter, in case something slips up.
2151 	 */
2152 	if (!ip_rsvp_on) {
2153 		ip_rsvp_on = 1;
2154 		rsvp_on++;
2155 	}
2156 
2157 	return 0;
2158 }
2159 
2160 int
2161 ip_rsvp_done(void)
2162 {
2163 	ip_rsvpd = NULL;
2164 	/*
2165 	 * This may seem silly, but we need to be sure we don't over-decrement
2166 	 * the RSVP counter, in case something slips up.
2167 	 */
2168 	if (ip_rsvp_on) {
2169 		ip_rsvp_on = 0;
2170 		rsvp_on--;
2171 	}
2172 	return 0;
2173 }
2174 
2175 void
2176 rsvp_input(struct mbuf *m, int off)	/* XXX must fixup manually */
2177 {
2178 	if (rsvp_input_p) { /* call the real one if loaded */
2179 		rsvp_input_p(m, off);
2180 		return;
2181 	}
2182 
2183 	/* Can still get packets with rsvp_on = 0 if there is a local member
2184 	 * of the group to which the RSVP packet is addressed.  But in this
2185 	 * case we want to throw the packet away.
2186 	 */
2187 
2188 	if (!rsvp_on) {
2189 		m_freem(m);
2190 		return;
2191 	}
2192 
2193 	if (ip_rsvpd != NULL) {
2194 		rip_input(m, off);
2195 		return;
2196 	}
2197 	/* Drop the packet */
2198 	m_freem(m);
2199 }
2200