xref: /freebsd/sys/netinet/ip_input.c (revision a3e8fd0b7f663db7eafff527d5c3ca3bcfa8a537)
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	ip_nfragpackets = 0;
123 static int	ip_maxfragpackets;	/* initialized in ip_init() */
124 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragpackets, CTLFLAG_RW,
125 	&ip_maxfragpackets, 0,
126 	"Maximum number of IPv4 fragment reassembly queue entries");
127 
128 /*
129  * XXX - Setting ip_checkinterface mostly implements the receive side of
130  * the Strong ES model described in RFC 1122, but since the routing table
131  * and transmit implementation do not implement the Strong ES model,
132  * setting this to 1 results in an odd hybrid.
133  *
134  * XXX - ip_checkinterface currently must be disabled if you use ipnat
135  * to translate the destination address to another local interface.
136  *
137  * XXX - ip_checkinterface must be disabled if you add IP aliases
138  * to the loopback interface instead of the interface where the
139  * packets for those addresses are received.
140  */
141 static int	ip_checkinterface = 1;
142 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW,
143     &ip_checkinterface, 0, "Verify packet arrives on correct interface");
144 
145 #ifdef DIAGNOSTIC
146 static int	ipprintfs = 0;
147 #endif
148 
149 static int	ipqmaxlen = IFQ_MAXLEN;
150 
151 extern	struct domain inetdomain;
152 extern	struct protosw inetsw[];
153 u_char	ip_protox[IPPROTO_MAX];
154 struct	in_ifaddrhead in_ifaddrhead; 		/* first inet address */
155 struct	in_ifaddrhashhead *in_ifaddrhashtbl;	/* inet addr hash table  */
156 u_long 	in_ifaddrhmask;				/* mask for hash table */
157 
158 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW,
159     &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue");
160 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD,
161     &ipintrq.ifq_drops, 0, "Number of packets dropped from the IP input queue");
162 
163 struct ipstat ipstat;
164 SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW,
165     &ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)");
166 
167 /* Packet reassembly stuff */
168 #define IPREASS_NHASH_LOG2      6
169 #define IPREASS_NHASH           (1 << IPREASS_NHASH_LOG2)
170 #define IPREASS_HMASK           (IPREASS_NHASH - 1)
171 #define IPREASS_HASH(x,y) \
172 	(((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK)
173 
174 static TAILQ_HEAD(ipqhead, ipq) ipq[IPREASS_NHASH];
175 static int    nipq = 0;         /* total # of reass queues */
176 static int    maxnipq;
177 
178 #ifdef IPCTL_DEFMTU
179 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
180     &ip_mtu, 0, "Default MTU");
181 #endif
182 
183 #ifdef IPSTEALTH
184 static int	ipstealth = 0;
185 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW,
186     &ipstealth, 0, "");
187 #endif
188 
189 
190 /* Firewall hooks */
191 ip_fw_chk_t *ip_fw_chk_ptr;
192 int fw_enable = 1 ;
193 
194 /* Dummynet hooks */
195 ip_dn_io_t *ip_dn_io_ptr;
196 
197 
198 /*
199  * XXX this is ugly -- the following two global variables are
200  * used to store packet state while it travels through the stack.
201  * Note that the code even makes assumptions on the size and
202  * alignment of fields inside struct ip_srcrt so e.g. adding some
203  * fields will break the code. This needs to be fixed.
204  *
205  * We need to save the IP options in case a protocol wants to respond
206  * to an incoming packet over the same route if the packet got here
207  * using IP source routing.  This allows connection establishment and
208  * maintenance when the remote end is on a network that is not known
209  * to us.
210  */
211 static int	ip_nhops = 0;
212 static	struct ip_srcrt {
213 	struct	in_addr dst;			/* final destination */
214 	char	nop;				/* one NOP to align */
215 	char	srcopt[IPOPT_OFFSET + 1];	/* OPTVAL, OLEN and OFFSET */
216 	struct	in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)];
217 } ip_srcrt;
218 
219 static void	save_rte(u_char *, struct in_addr);
220 static int	ip_dooptions(struct mbuf *m, int,
221 			struct sockaddr_in *next_hop);
222 static void	ip_forward(struct mbuf *m, int srcrt,
223 			struct sockaddr_in *next_hop);
224 static void	ip_freef(struct ipqhead *, struct ipq *);
225 static struct	mbuf *ip_reass(struct mbuf *, struct ipqhead *,
226 		struct ipq *, u_int32_t *, u_int16_t *);
227 static void	ipintr(void);
228 
229 /*
230  * IP initialization: fill in IP protocol switch table.
231  * All protocols not implemented in kernel go to raw IP protocol handler.
232  */
233 void
234 ip_init()
235 {
236 	register struct protosw *pr;
237 	register int i;
238 
239 	TAILQ_INIT(&in_ifaddrhead);
240 	in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask);
241 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
242 	if (pr == 0)
243 		panic("ip_init");
244 	for (i = 0; i < IPPROTO_MAX; i++)
245 		ip_protox[i] = pr - inetsw;
246 	for (pr = inetdomain.dom_protosw;
247 	    pr < inetdomain.dom_protoswNPROTOSW; pr++)
248 		if (pr->pr_domain->dom_family == PF_INET &&
249 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
250 			ip_protox[pr->pr_protocol] = pr - inetsw;
251 
252 	for (i = 0; i < IPREASS_NHASH; i++)
253 	    TAILQ_INIT(&ipq[i]);
254 
255 	maxnipq = nmbclusters / 4;
256 	ip_maxfragpackets = nmbclusters / 4;
257 
258 #ifndef RANDOM_IP_ID
259 	ip_id = time_second & 0xffff;
260 #endif
261 	ipintrq.ifq_maxlen = ipqmaxlen;
262 	mtx_init(&ipintrq.ifq_mtx, "ip_inq", NULL, MTX_DEF);
263 	ipintrq_present = 1;
264 
265 	register_netisr(NETISR_IP, ipintr);
266 }
267 
268 /*
269  * XXX watch out this one. It is perhaps used as a cache for
270  * the most recently used route ? it is cleared in in_addroute()
271  * when a new route is successfully created.
272  */
273 struct	route ipforward_rt;
274 
275 /*
276  * Ip input routine.  Checksum and byte swap header.  If fragmented
277  * try to reassemble.  Process options.  Pass to next level.
278  */
279 void
280 ip_input(struct mbuf *m)
281 {
282 	struct ip *ip;
283 	struct ipq *fp;
284 	struct in_ifaddr *ia = NULL;
285 	struct ifaddr *ifa;
286 	int    i, hlen, checkif;
287 	u_short sum;
288 	struct in_addr pkt_dst;
289 	u_int32_t divert_info = 0;		/* packet divert/tee info */
290 	struct ip_fw_args args;
291 #ifdef PFIL_HOOKS
292 	struct packet_filter_hook *pfh;
293 	struct mbuf *m0;
294 	int rv;
295 #endif /* PFIL_HOOKS */
296 #ifdef FAST_IPSEC
297 	struct m_tag *mtag;
298 	struct tdb_ident *tdbi;
299 	struct secpolicy *sp;
300 	int s, error;
301 #endif /* FAST_IPSEC */
302 
303 	args.eh = NULL;
304 	args.oif = NULL;
305 	args.rule = NULL;
306 	args.divert_rule = 0;			/* divert cookie */
307 	args.next_hop = NULL;
308 
309 	/* Grab info from MT_TAG mbufs prepended to the chain.	*/
310 	for (; m && m->m_type == MT_TAG; m = m->m_next) {
311 		switch(m->_m_tag_id) {
312 		default:
313 			printf("ip_input: unrecognised MT_TAG tag %d\n",
314 			    m->_m_tag_id);
315 			break;
316 
317 		case PACKET_TAG_DUMMYNET:
318 			args.rule = ((struct dn_pkt *)m)->rule;
319 			break;
320 
321 		case PACKET_TAG_DIVERT:
322 			args.divert_rule = (intptr_t)m->m_hdr.mh_data & 0xffff;
323 			break;
324 
325 		case PACKET_TAG_IPFORWARD:
326 			args.next_hop = (struct sockaddr_in *)m->m_hdr.mh_data;
327 			break;
328 		}
329 	}
330 
331 	KASSERT(m != NULL && (m->m_flags & M_PKTHDR) != 0,
332 	    ("ip_input: no HDR"));
333 
334 	if (args.rule) {	/* dummynet already filtered us */
335 		ip = mtod(m, struct ip *);
336 		hlen = ip->ip_hl << 2;
337 		goto iphack ;
338 	}
339 
340 	ipstat.ips_total++;
341 
342 	if (m->m_pkthdr.len < sizeof(struct ip))
343 		goto tooshort;
344 
345 	if (m->m_len < sizeof (struct ip) &&
346 	    (m = m_pullup(m, sizeof (struct ip))) == 0) {
347 		ipstat.ips_toosmall++;
348 		return;
349 	}
350 	ip = mtod(m, struct ip *);
351 
352 	if (ip->ip_v != IPVERSION) {
353 		ipstat.ips_badvers++;
354 		goto bad;
355 	}
356 
357 	hlen = ip->ip_hl << 2;
358 	if (hlen < sizeof(struct ip)) {	/* minimum header length */
359 		ipstat.ips_badhlen++;
360 		goto bad;
361 	}
362 	if (hlen > m->m_len) {
363 		if ((m = m_pullup(m, hlen)) == 0) {
364 			ipstat.ips_badhlen++;
365 			return;
366 		}
367 		ip = mtod(m, struct ip *);
368 	}
369 
370 	/* 127/8 must not appear on wire - RFC1122 */
371 	if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
372 	    (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
373 		if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
374 			ipstat.ips_badaddr++;
375 			goto bad;
376 		}
377 	}
378 
379 	if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
380 		sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
381 	} else {
382 		if (hlen == sizeof(struct ip)) {
383 			sum = in_cksum_hdr(ip);
384 		} else {
385 			sum = in_cksum(m, hlen);
386 		}
387 	}
388 	if (sum) {
389 		ipstat.ips_badsum++;
390 		goto bad;
391 	}
392 
393 	/*
394 	 * Convert fields to host representation.
395 	 */
396 	ip->ip_len = ntohs(ip->ip_len);
397 	if (ip->ip_len < hlen) {
398 		ipstat.ips_badlen++;
399 		goto bad;
400 	}
401 	ip->ip_off = ntohs(ip->ip_off);
402 
403 	/*
404 	 * Check that the amount of data in the buffers
405 	 * is as at least much as the IP header would have us expect.
406 	 * Trim mbufs if longer than we expect.
407 	 * Drop packet if shorter than we expect.
408 	 */
409 	if (m->m_pkthdr.len < ip->ip_len) {
410 tooshort:
411 		ipstat.ips_tooshort++;
412 		goto bad;
413 	}
414 	if (m->m_pkthdr.len > ip->ip_len) {
415 		if (m->m_len == m->m_pkthdr.len) {
416 			m->m_len = ip->ip_len;
417 			m->m_pkthdr.len = ip->ip_len;
418 		} else
419 			m_adj(m, ip->ip_len - m->m_pkthdr.len);
420 	}
421 
422 	/*
423 	 * IpHack's section.
424 	 * Right now when no processing on packet has done
425 	 * and it is still fresh out of network we do our black
426 	 * deals with it.
427 	 * - Firewall: deny/allow/divert
428 	 * - Xlate: translate packet's addr/port (NAT).
429 	 * - Pipe: pass pkt through dummynet.
430 	 * - Wrap: fake packet's addr/port <unimpl.>
431 	 * - Encapsulate: put it in another IP and send out. <unimp.>
432  	 */
433 
434 iphack:
435 
436 #ifdef PFIL_HOOKS
437 	/*
438 	 * Run through list of hooks for input packets.  If there are any
439 	 * filters which require that additional packets in the flow are
440 	 * not fast-forwarded, they must clear the M_CANFASTFWD flag.
441 	 * Note that filters must _never_ set this flag, as another filter
442 	 * in the list may have previously cleared it.
443 	 */
444 	m0 = m;
445 	pfh = pfil_hook_get(PFIL_IN, &inetsw[ip_protox[IPPROTO_IP]].pr_pfh);
446 	for (; pfh; pfh = TAILQ_NEXT(pfh, pfil_link))
447 		if (pfh->pfil_func) {
448 			rv = pfh->pfil_func(ip, hlen,
449 					    m->m_pkthdr.rcvif, 0, &m0);
450 			if (rv)
451 				return;
452 			m = m0;
453 			if (m == NULL)
454 				return;
455 			ip = mtod(m, struct ip *);
456 		}
457 #endif /* PFIL_HOOKS */
458 
459 	if (fw_enable && IPFW_LOADED) {
460 		/*
461 		 * If we've been forwarded from the output side, then
462 		 * skip the firewall a second time
463 		 */
464 		if (args.next_hop)
465 			goto ours;
466 
467 		args.m = m;
468 		i = ip_fw_chk_ptr(&args);
469 		m = args.m;
470 
471 		if ( (i & IP_FW_PORT_DENY_FLAG) || m == NULL) { /* drop */
472 			if (m)
473 				m_freem(m);
474 			return;
475 		}
476 		ip = mtod(m, struct ip *); /* just in case m changed */
477 		if (i == 0 && args.next_hop == NULL)	/* common case */
478 			goto pass;
479                 if (DUMMYNET_LOADED && (i & IP_FW_PORT_DYNT_FLAG) != 0) {
480 			/* Send packet to the appropriate pipe */
481 			ip_dn_io_ptr(m, i&0xffff, DN_TO_IP_IN, &args);
482 			return;
483 		}
484 #ifdef IPDIVERT
485 		if (i != 0 && (i & IP_FW_PORT_DYNT_FLAG) == 0) {
486 			/* Divert or tee packet */
487 			divert_info = i;
488 			goto ours;
489 		}
490 #endif
491 		if (i == 0 && args.next_hop != NULL)
492 			goto pass;
493 		/*
494 		 * if we get here, the packet must be dropped
495 		 */
496 		m_freem(m);
497 		return;
498 	}
499 pass:
500 
501 	/*
502 	 * Process options and, if not destined for us,
503 	 * ship it on.  ip_dooptions returns 1 when an
504 	 * error was detected (causing an icmp message
505 	 * to be sent and the original packet to be freed).
506 	 */
507 	ip_nhops = 0;		/* for source routed packets */
508 	if (hlen > sizeof (struct ip) && ip_dooptions(m, 0, args.next_hop))
509 		return;
510 
511         /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
512          * matter if it is destined to another node, or whether it is
513          * a multicast one, RSVP wants it! and prevents it from being forwarded
514          * anywhere else. Also checks if the rsvp daemon is running before
515 	 * grabbing the packet.
516          */
517 	if (rsvp_on && ip->ip_p==IPPROTO_RSVP)
518 		goto ours;
519 
520 	/*
521 	 * Check our list of addresses, to see if the packet is for us.
522 	 * If we don't have any addresses, assume any unicast packet
523 	 * we receive might be for us (and let the upper layers deal
524 	 * with it).
525 	 */
526 	if (TAILQ_EMPTY(&in_ifaddrhead) &&
527 	    (m->m_flags & (M_MCAST|M_BCAST)) == 0)
528 		goto ours;
529 
530 	/*
531 	 * Cache the destination address of the packet; this may be
532 	 * changed by use of 'ipfw fwd'.
533 	 */
534 	pkt_dst = args.next_hop ? args.next_hop->sin_addr : ip->ip_dst;
535 
536 	/*
537 	 * Enable a consistency check between the destination address
538 	 * and the arrival interface for a unicast packet (the RFC 1122
539 	 * strong ES model) if IP forwarding is disabled and the packet
540 	 * is not locally generated and the packet is not subject to
541 	 * 'ipfw fwd'.
542 	 *
543 	 * XXX - Checking also should be disabled if the destination
544 	 * address is ipnat'ed to a different interface.
545 	 *
546 	 * XXX - Checking is incompatible with IP aliases added
547 	 * to the loopback interface instead of the interface where
548 	 * the packets are received.
549 	 */
550 	checkif = ip_checkinterface && (ipforwarding == 0) &&
551 	    m->m_pkthdr.rcvif != NULL &&
552 	    ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) &&
553 	    (args.next_hop == NULL);
554 
555 	/*
556 	 * Check for exact addresses in the hash bucket.
557 	 */
558 	LIST_FOREACH(ia, INADDR_HASH(pkt_dst.s_addr), ia_hash) {
559 		/*
560 		 * If the address matches, verify that the packet
561 		 * arrived via the correct interface if checking is
562 		 * enabled.
563 		 */
564 		if (IA_SIN(ia)->sin_addr.s_addr == pkt_dst.s_addr &&
565 		    (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif))
566 			goto ours;
567 	}
568 	/*
569 	 * Check for broadcast addresses.
570 	 *
571 	 * Only accept broadcast packets that arrive via the matching
572 	 * interface.  Reception of forwarded directed broadcasts would
573 	 * be handled via ip_forward() and ether_output() with the loopback
574 	 * into the stack for SIMPLEX interfaces handled by ether_output().
575 	 */
576 	if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
577 	        TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
578 			if (ifa->ifa_addr->sa_family != AF_INET)
579 				continue;
580 			ia = ifatoia(ifa);
581 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
582 			    pkt_dst.s_addr)
583 				goto ours;
584 			if (ia->ia_netbroadcast.s_addr == pkt_dst.s_addr)
585 				goto ours;
586 #ifdef BOOTP_COMPAT
587 			if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY)
588 				goto ours;
589 #endif
590 		}
591 	}
592 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
593 		struct in_multi *inm;
594 		if (ip_mrouter) {
595 			/*
596 			 * If we are acting as a multicast router, all
597 			 * incoming multicast packets are passed to the
598 			 * kernel-level multicast forwarding function.
599 			 * The packet is returned (relatively) intact; if
600 			 * ip_mforward() returns a non-zero value, the packet
601 			 * must be discarded, else it may be accepted below.
602 			 */
603 			if (ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) {
604 				ipstat.ips_cantforward++;
605 				m_freem(m);
606 				return;
607 			}
608 
609 			/*
610 			 * The process-level routing daemon needs to receive
611 			 * all multicast IGMP packets, whether or not this
612 			 * host belongs to their destination groups.
613 			 */
614 			if (ip->ip_p == IPPROTO_IGMP)
615 				goto ours;
616 			ipstat.ips_forward++;
617 		}
618 		/*
619 		 * See if we belong to the destination multicast group on the
620 		 * arrival interface.
621 		 */
622 		IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
623 		if (inm == NULL) {
624 			ipstat.ips_notmember++;
625 			m_freem(m);
626 			return;
627 		}
628 		goto ours;
629 	}
630 	if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
631 		goto ours;
632 	if (ip->ip_dst.s_addr == INADDR_ANY)
633 		goto ours;
634 
635 	/*
636 	 * FAITH(Firewall Aided Internet Translator)
637 	 */
638 	if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
639 		if (ip_keepfaith) {
640 			if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP)
641 				goto ours;
642 		}
643 		m_freem(m);
644 		return;
645 	}
646 
647 	/*
648 	 * Not for us; forward if possible and desirable.
649 	 */
650 	if (ipforwarding == 0) {
651 		ipstat.ips_cantforward++;
652 		m_freem(m);
653 	} else {
654 #ifdef IPSEC
655 		/*
656 		 * Enforce inbound IPsec SPD.
657 		 */
658 		if (ipsec4_in_reject(m, NULL)) {
659 			ipsecstat.in_polvio++;
660 			goto bad;
661 		}
662 #endif /* IPSEC */
663 #ifdef FAST_IPSEC
664 		mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
665 		s = splnet();
666 		if (mtag != NULL) {
667 			tdbi = (struct tdb_ident *)(mtag + 1);
668 			sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
669 		} else {
670 			sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
671 						   IP_FORWARDING, &error);
672 		}
673 		if (sp == NULL) {	/* NB: can happen if error */
674 			splx(s);
675 			/*XXX error stat???*/
676 			DPRINTF(("ip_input: no SP for forwarding\n"));	/*XXX*/
677 			goto bad;
678 		}
679 
680 		/*
681 		 * Check security policy against packet attributes.
682 		 */
683 		error = ipsec_in_reject(sp, m);
684 		KEY_FREESP(&sp);
685 		splx(s);
686 		if (error) {
687 			ipstat.ips_cantforward++;
688 			goto bad;
689 		}
690 #endif /* FAST_IPSEC */
691 		ip_forward(m, 0, args.next_hop);
692 	}
693 	return;
694 
695 ours:
696 #ifdef IPSTEALTH
697 	/*
698 	 * IPSTEALTH: Process non-routing options only
699 	 * if the packet is destined for us.
700 	 */
701 	if (ipstealth && hlen > sizeof (struct ip) &&
702 	    ip_dooptions(m, 1, args.next_hop))
703 		return;
704 #endif /* IPSTEALTH */
705 
706 	/* Count the packet in the ip address stats */
707 	if (ia != NULL) {
708 		ia->ia_ifa.if_ipackets++;
709 		ia->ia_ifa.if_ibytes += m->m_pkthdr.len;
710 	}
711 
712 	/*
713 	 * If offset or IP_MF are set, must reassemble.
714 	 * Otherwise, nothing need be done.
715 	 * (We could look in the reassembly queue to see
716 	 * if the packet was previously fragmented,
717 	 * but it's not worth the time; just let them time out.)
718 	 */
719 	if (ip->ip_off & (IP_MF | IP_OFFMASK)) {
720 
721 		sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
722 		/*
723 		 * Look for queue of fragments
724 		 * of this datagram.
725 		 */
726 		TAILQ_FOREACH(fp, &ipq[sum], ipq_list)
727 			if (ip->ip_id == fp->ipq_id &&
728 			    ip->ip_src.s_addr == fp->ipq_src.s_addr &&
729 			    ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
730 #ifdef MAC
731 			    mac_fragment_match(m, fp) &&
732 #endif
733 			    ip->ip_p == fp->ipq_p)
734 				goto found;
735 
736 		fp = 0;
737 
738 		/* check if there's a place for the new queue */
739 		if (nipq > maxnipq) {
740 		    /*
741 		     * drop something from the tail of the current queue
742 		     * before proceeding further
743 		     */
744 		    struct ipq *q = TAILQ_LAST(&ipq[sum], ipqhead);
745 		    if (q == NULL) {   /* gak */
746 			for (i = 0; i < IPREASS_NHASH; i++) {
747 			    struct ipq *r = TAILQ_LAST(&ipq[i], ipqhead);
748 			    if (r) {
749 				ip_freef(&ipq[i], r);
750 				break;
751 			    }
752 			}
753 		    } else
754 			ip_freef(&ipq[sum], q);
755 		}
756 found:
757 		/*
758 		 * Adjust ip_len to not reflect header,
759 		 * convert offset of this to bytes.
760 		 */
761 		ip->ip_len -= hlen;
762 		if (ip->ip_off & IP_MF) {
763 		        /*
764 		         * Make sure that fragments have a data length
765 			 * that's a non-zero multiple of 8 bytes.
766 		         */
767 			if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
768 				ipstat.ips_toosmall++; /* XXX */
769 				goto bad;
770 			}
771 			m->m_flags |= M_FRAG;
772 		} else
773 			m->m_flags &= ~M_FRAG;
774 		ip->ip_off <<= 3;
775 
776 		/*
777 		 * Attempt reassembly; if it succeeds, proceed.
778 		 * ip_reass() will return a different mbuf, and update
779 		 * the divert info in divert_info and args.divert_rule.
780 		 */
781 		ipstat.ips_fragments++;
782 		m->m_pkthdr.header = ip;
783 		m = ip_reass(m,
784 		    &ipq[sum], fp, &divert_info, &args.divert_rule);
785 		if (m == 0)
786 			return;
787 		ipstat.ips_reassembled++;
788 		ip = mtod(m, struct ip *);
789 		/* Get the header length of the reassembled packet */
790 		hlen = ip->ip_hl << 2;
791 #ifdef IPDIVERT
792 		/* Restore original checksum before diverting packet */
793 		if (divert_info != 0) {
794 			ip->ip_len += hlen;
795 			ip->ip_len = htons(ip->ip_len);
796 			ip->ip_off = htons(ip->ip_off);
797 			ip->ip_sum = 0;
798 			if (hlen == sizeof(struct ip))
799 				ip->ip_sum = in_cksum_hdr(ip);
800 			else
801 				ip->ip_sum = in_cksum(m, hlen);
802 			ip->ip_off = ntohs(ip->ip_off);
803 			ip->ip_len = ntohs(ip->ip_len);
804 			ip->ip_len -= hlen;
805 		}
806 #endif
807 	} else
808 		ip->ip_len -= hlen;
809 
810 #ifdef IPDIVERT
811 	/*
812 	 * Divert or tee packet to the divert protocol if required.
813 	 */
814 	if (divert_info != 0) {
815 		struct mbuf *clone = NULL;
816 
817 		/* Clone packet if we're doing a 'tee' */
818 		if ((divert_info & IP_FW_PORT_TEE_FLAG) != 0)
819 			clone = m_dup(m, M_DONTWAIT);
820 
821 		/* Restore packet header fields to original values */
822 		ip->ip_len += hlen;
823 		ip->ip_len = htons(ip->ip_len);
824 		ip->ip_off = htons(ip->ip_off);
825 
826 		/* Deliver packet to divert input routine */
827 		divert_packet(m, 1, divert_info & 0xffff, args.divert_rule);
828 		ipstat.ips_delivered++;
829 
830 		/* If 'tee', continue with original packet */
831 		if (clone == NULL)
832 			return;
833 		m = clone;
834 		ip = mtod(m, struct ip *);
835 		ip->ip_len += hlen;
836 		/*
837 		 * Jump backwards to complete processing of the
838 		 * packet. But first clear divert_info to avoid
839 		 * entering this block again.
840 		 * We do not need to clear args.divert_rule
841 		 * or args.next_hop as they will not be used.
842 		 */
843 		divert_info = 0;
844 		goto pass;
845 	}
846 #endif
847 
848 #ifdef IPSEC
849 	/*
850 	 * enforce IPsec policy checking if we are seeing last header.
851 	 * note that we do not visit this with protocols with pcb layer
852 	 * code - like udp/tcp/raw ip.
853 	 */
854 	if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0 &&
855 	    ipsec4_in_reject(m, NULL)) {
856 		ipsecstat.in_polvio++;
857 		goto bad;
858 	}
859 #endif
860 #if FAST_IPSEC
861 	/*
862 	 * enforce IPsec policy checking if we are seeing last header.
863 	 * note that we do not visit this with protocols with pcb layer
864 	 * code - like udp/tcp/raw ip.
865 	 */
866 	if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) {
867 		/*
868 		 * Check if the packet has already had IPsec processing
869 		 * done.  If so, then just pass it along.  This tag gets
870 		 * set during AH, ESP, etc. input handling, before the
871 		 * packet is returned to the ip input queue for delivery.
872 		 */
873 		mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
874 		s = splnet();
875 		if (mtag != NULL) {
876 			tdbi = (struct tdb_ident *)(mtag + 1);
877 			sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
878 		} else {
879 			sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
880 						   IP_FORWARDING, &error);
881 		}
882 		if (sp != NULL) {
883 			/*
884 			 * Check security policy against packet attributes.
885 			 */
886 			error = ipsec_in_reject(sp, m);
887 			KEY_FREESP(&sp);
888 		} else {
889 			/* XXX error stat??? */
890 			error = EINVAL;
891 DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/
892 			goto bad;
893 		}
894 		splx(s);
895 		if (error)
896 			goto bad;
897 	}
898 #endif /* FAST_IPSEC */
899 
900 	/*
901 	 * Switch out to protocol's input routine.
902 	 */
903 	ipstat.ips_delivered++;
904 	if (args.next_hop && ip->ip_p == IPPROTO_TCP) {
905 		/* TCP needs IPFORWARD info if available */
906 		struct m_hdr tag;
907 
908 		tag.mh_type = MT_TAG;
909 		tag.mh_flags = PACKET_TAG_IPFORWARD;
910 		tag.mh_data = (caddr_t)args.next_hop;
911 		tag.mh_next = m;
912 
913 		(*inetsw[ip_protox[ip->ip_p]].pr_input)(
914 			(struct mbuf *)&tag, hlen);
915 	} else
916 		(*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen);
917 	return;
918 bad:
919 	m_freem(m);
920 }
921 
922 /*
923  * IP software interrupt routine - to go away sometime soon
924  */
925 static void
926 ipintr(void)
927 {
928 	struct mbuf *m;
929 
930 	while (1) {
931 		IF_DEQUEUE(&ipintrq, m);
932 		if (m == 0)
933 			return;
934 		ip_input(m);
935 	}
936 }
937 
938 /*
939  * Take incoming datagram fragment and try to reassemble it into
940  * whole datagram.  If a chain for reassembly of this datagram already
941  * exists, then it is given as fp; otherwise have to make a chain.
942  *
943  * When IPDIVERT enabled, keep additional state with each packet that
944  * tells us if we need to divert or tee the packet we're building.
945  * In particular, *divinfo includes the port and TEE flag,
946  * *divert_rule is the number of the matching rule.
947  */
948 
949 static struct mbuf *
950 ip_reass(struct mbuf *m, struct ipqhead *head, struct ipq *fp,
951 	u_int32_t *divinfo, u_int16_t *divert_rule)
952 {
953 	struct ip *ip = mtod(m, struct ip *);
954 	register struct mbuf *p, *q, *nq;
955 	struct mbuf *t;
956 	int hlen = ip->ip_hl << 2;
957 	int i, next;
958 
959 	/*
960 	 * Presence of header sizes in mbufs
961 	 * would confuse code below.
962 	 */
963 	m->m_data += hlen;
964 	m->m_len -= hlen;
965 
966 	/*
967 	 * If first fragment to arrive, create a reassembly queue.
968 	 */
969 	if (fp == 0) {
970 		/*
971 		 * Enforce upper bound on number of fragmented packets
972 		 * for which we attempt reassembly;
973 		 * If maxfrag is 0, never accept fragments.
974 		 * If maxfrag is -1, accept all fragments without limitation.
975 		 */
976 		if ((ip_maxfragpackets >= 0) && (ip_nfragpackets >= ip_maxfragpackets))
977 			goto dropfrag;
978 		ip_nfragpackets++;
979 		if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL)
980 			goto dropfrag;
981 		fp = mtod(t, struct ipq *);
982 #ifdef MAC
983 		mac_init_ipq(fp);
984 		mac_create_ipq(m, fp);
985 #endif
986 		TAILQ_INSERT_HEAD(head, fp, ipq_list);
987 		nipq++;
988 		fp->ipq_ttl = IPFRAGTTL;
989 		fp->ipq_p = ip->ip_p;
990 		fp->ipq_id = ip->ip_id;
991 		fp->ipq_src = ip->ip_src;
992 		fp->ipq_dst = ip->ip_dst;
993 		fp->ipq_frags = m;
994 		m->m_nextpkt = NULL;
995 #ifdef IPDIVERT
996 		fp->ipq_div_info = 0;
997 		fp->ipq_div_cookie = 0;
998 #endif
999 		goto inserted;
1000 	} else {
1001 #ifdef MAC
1002 		mac_update_ipq(m, fp);
1003 #endif
1004 	}
1005 
1006 #define GETIP(m)	((struct ip*)((m)->m_pkthdr.header))
1007 
1008 	/*
1009 	 * Find a segment which begins after this one does.
1010 	 */
1011 	for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt)
1012 		if (GETIP(q)->ip_off > ip->ip_off)
1013 			break;
1014 
1015 	/*
1016 	 * If there is a preceding segment, it may provide some of
1017 	 * our data already.  If so, drop the data from the incoming
1018 	 * segment.  If it provides all of our data, drop us, otherwise
1019 	 * stick new segment in the proper place.
1020 	 *
1021 	 * If some of the data is dropped from the the preceding
1022 	 * segment, then it's checksum is invalidated.
1023 	 */
1024 	if (p) {
1025 		i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off;
1026 		if (i > 0) {
1027 			if (i >= ip->ip_len)
1028 				goto dropfrag;
1029 			m_adj(m, i);
1030 			m->m_pkthdr.csum_flags = 0;
1031 			ip->ip_off += i;
1032 			ip->ip_len -= i;
1033 		}
1034 		m->m_nextpkt = p->m_nextpkt;
1035 		p->m_nextpkt = m;
1036 	} else {
1037 		m->m_nextpkt = fp->ipq_frags;
1038 		fp->ipq_frags = m;
1039 	}
1040 
1041 	/*
1042 	 * While we overlap succeeding segments trim them or,
1043 	 * if they are completely covered, dequeue them.
1044 	 */
1045 	for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off;
1046 	     q = nq) {
1047 		i = (ip->ip_off + ip->ip_len) -
1048 		    GETIP(q)->ip_off;
1049 		if (i < GETIP(q)->ip_len) {
1050 			GETIP(q)->ip_len -= i;
1051 			GETIP(q)->ip_off += i;
1052 			m_adj(q, i);
1053 			q->m_pkthdr.csum_flags = 0;
1054 			break;
1055 		}
1056 		nq = q->m_nextpkt;
1057 		m->m_nextpkt = nq;
1058 		m_freem(q);
1059 	}
1060 
1061 inserted:
1062 
1063 #ifdef IPDIVERT
1064 	/*
1065 	 * Transfer firewall instructions to the fragment structure.
1066 	 * Only trust info in the fragment at offset 0.
1067 	 */
1068 	if (ip->ip_off == 0) {
1069 		fp->ipq_div_info = *divinfo;
1070 		fp->ipq_div_cookie = *divert_rule;
1071 	}
1072 	*divinfo = 0;
1073 	*divert_rule = 0;
1074 #endif
1075 
1076 	/*
1077 	 * Check for complete reassembly.
1078 	 */
1079 	next = 0;
1080 	for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
1081 		if (GETIP(q)->ip_off != next)
1082 			return (0);
1083 		next += GETIP(q)->ip_len;
1084 	}
1085 	/* Make sure the last packet didn't have the IP_MF flag */
1086 	if (p->m_flags & M_FRAG)
1087 		return (0);
1088 
1089 	/*
1090 	 * Reassembly is complete.  Make sure the packet is a sane size.
1091 	 */
1092 	q = fp->ipq_frags;
1093 	ip = GETIP(q);
1094 	if (next + (ip->ip_hl << 2) > IP_MAXPACKET) {
1095 		ipstat.ips_toolong++;
1096 		ip_freef(head, fp);
1097 		return (0);
1098 	}
1099 
1100 	/*
1101 	 * Concatenate fragments.
1102 	 */
1103 	m = q;
1104 	t = m->m_next;
1105 	m->m_next = 0;
1106 	m_cat(m, t);
1107 	nq = q->m_nextpkt;
1108 	q->m_nextpkt = 0;
1109 	for (q = nq; q != NULL; q = nq) {
1110 		nq = q->m_nextpkt;
1111 		q->m_nextpkt = NULL;
1112 		m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
1113 		m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
1114 		m_cat(m, q);
1115 	}
1116 #ifdef MAC
1117 	mac_create_datagram_from_ipq(fp, m);
1118 	mac_destroy_ipq(fp);
1119 #endif
1120 
1121 #ifdef IPDIVERT
1122 	/*
1123 	 * Extract firewall instructions from the fragment structure.
1124 	 */
1125 	*divinfo = fp->ipq_div_info;
1126 	*divert_rule = fp->ipq_div_cookie;
1127 #endif
1128 
1129 	/*
1130 	 * Create header for new ip packet by
1131 	 * modifying header of first packet;
1132 	 * dequeue and discard fragment reassembly header.
1133 	 * Make header visible.
1134 	 */
1135 	ip->ip_len = next;
1136 	ip->ip_src = fp->ipq_src;
1137 	ip->ip_dst = fp->ipq_dst;
1138 	TAILQ_REMOVE(head, fp, ipq_list);
1139 	nipq--;
1140 	(void) m_free(dtom(fp));
1141 	ip_nfragpackets--;
1142 	m->m_len += (ip->ip_hl << 2);
1143 	m->m_data -= (ip->ip_hl << 2);
1144 	/* some debugging cruft by sklower, below, will go away soon */
1145 	if (m->m_flags & M_PKTHDR)	/* XXX this should be done elsewhere */
1146 		m_fixhdr(m);
1147 	return (m);
1148 
1149 dropfrag:
1150 #ifdef IPDIVERT
1151 	*divinfo = 0;
1152 	*divert_rule = 0;
1153 #endif
1154 	ipstat.ips_fragdropped++;
1155 	m_freem(m);
1156 	return (0);
1157 
1158 #undef GETIP
1159 }
1160 
1161 /*
1162  * Free a fragment reassembly header and all
1163  * associated datagrams.
1164  */
1165 static void
1166 ip_freef(fhp, fp)
1167 	struct ipqhead *fhp;
1168 	struct ipq *fp;
1169 {
1170 	register struct mbuf *q;
1171 
1172 	while (fp->ipq_frags) {
1173 		q = fp->ipq_frags;
1174 		fp->ipq_frags = q->m_nextpkt;
1175 		m_freem(q);
1176 	}
1177 	TAILQ_REMOVE(fhp, fp, ipq_list);
1178 	(void) m_free(dtom(fp));
1179 	ip_nfragpackets--;
1180 	nipq--;
1181 }
1182 
1183 /*
1184  * IP timer processing;
1185  * if a timer expires on a reassembly
1186  * queue, discard it.
1187  */
1188 void
1189 ip_slowtimo()
1190 {
1191 	register struct ipq *fp;
1192 	int s = splnet();
1193 	int i;
1194 
1195 	for (i = 0; i < IPREASS_NHASH; i++) {
1196 		for(fp = TAILQ_FIRST(&ipq[i]); fp;) {
1197 			struct ipq *fpp;
1198 
1199 			fpp = fp;
1200 			fp = TAILQ_NEXT(fp, ipq_list);
1201 			if(--fpp->ipq_ttl == 0) {
1202 				ipstat.ips_fragtimeout++;
1203 				ip_freef(&ipq[i], fpp);
1204 			}
1205 		}
1206 	}
1207 	/*
1208 	 * If we are over the maximum number of fragments
1209 	 * (due to the limit being lowered), drain off
1210 	 * enough to get down to the new limit.
1211 	 */
1212 	for (i = 0; i < IPREASS_NHASH; i++) {
1213 		if (ip_maxfragpackets >= 0) {
1214 			while (ip_nfragpackets > ip_maxfragpackets &&
1215 				!TAILQ_EMPTY(&ipq[i])) {
1216 				ipstat.ips_fragdropped++;
1217 				ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i]));
1218 			}
1219 		}
1220 	}
1221 	ipflow_slowtimo();
1222 	splx(s);
1223 }
1224 
1225 /*
1226  * Drain off all datagram fragments.
1227  */
1228 void
1229 ip_drain()
1230 {
1231 	int     i;
1232 
1233 	for (i = 0; i < IPREASS_NHASH; i++) {
1234 		while(!TAILQ_EMPTY(&ipq[i])) {
1235 			ipstat.ips_fragdropped++;
1236 			ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i]));
1237 		}
1238 	}
1239 	in_rtqdrain();
1240 }
1241 
1242 /*
1243  * Do option processing on a datagram,
1244  * possibly discarding it if bad options are encountered,
1245  * or forwarding it if source-routed.
1246  * The pass argument is used when operating in the IPSTEALTH
1247  * mode to tell what options to process:
1248  * [LS]SRR (pass 0) or the others (pass 1).
1249  * The reason for as many as two passes is that when doing IPSTEALTH,
1250  * non-routing options should be processed only if the packet is for us.
1251  * Returns 1 if packet has been forwarded/freed,
1252  * 0 if the packet should be processed further.
1253  */
1254 static int
1255 ip_dooptions(struct mbuf *m, int pass, struct sockaddr_in *next_hop)
1256 {
1257 	struct ip *ip = mtod(m, struct ip *);
1258 	u_char *cp;
1259 	struct in_ifaddr *ia;
1260 	int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0;
1261 	struct in_addr *sin, dst;
1262 	n_time ntime;
1263 	struct	sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET };
1264 
1265 	dst = ip->ip_dst;
1266 	cp = (u_char *)(ip + 1);
1267 	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
1268 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1269 		opt = cp[IPOPT_OPTVAL];
1270 		if (opt == IPOPT_EOL)
1271 			break;
1272 		if (opt == IPOPT_NOP)
1273 			optlen = 1;
1274 		else {
1275 			if (cnt < IPOPT_OLEN + sizeof(*cp)) {
1276 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1277 				goto bad;
1278 			}
1279 			optlen = cp[IPOPT_OLEN];
1280 			if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
1281 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1282 				goto bad;
1283 			}
1284 		}
1285 		switch (opt) {
1286 
1287 		default:
1288 			break;
1289 
1290 		/*
1291 		 * Source routing with record.
1292 		 * Find interface with current destination address.
1293 		 * If none on this machine then drop if strictly routed,
1294 		 * or do nothing if loosely routed.
1295 		 * Record interface address and bring up next address
1296 		 * component.  If strictly routed make sure next
1297 		 * address is on directly accessible net.
1298 		 */
1299 		case IPOPT_LSRR:
1300 		case IPOPT_SSRR:
1301 #ifdef IPSTEALTH
1302 			if (ipstealth && pass > 0)
1303 				break;
1304 #endif
1305 			if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1306 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1307 				goto bad;
1308 			}
1309 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1310 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1311 				goto bad;
1312 			}
1313 			ipaddr.sin_addr = ip->ip_dst;
1314 			ia = (struct in_ifaddr *)
1315 				ifa_ifwithaddr((struct sockaddr *)&ipaddr);
1316 			if (ia == 0) {
1317 				if (opt == IPOPT_SSRR) {
1318 					type = ICMP_UNREACH;
1319 					code = ICMP_UNREACH_SRCFAIL;
1320 					goto bad;
1321 				}
1322 				if (!ip_dosourceroute)
1323 					goto nosourcerouting;
1324 				/*
1325 				 * Loose routing, and not at next destination
1326 				 * yet; nothing to do except forward.
1327 				 */
1328 				break;
1329 			}
1330 			off--;			/* 0 origin */
1331 			if (off > optlen - (int)sizeof(struct in_addr)) {
1332 				/*
1333 				 * End of source route.  Should be for us.
1334 				 */
1335 				if (!ip_acceptsourceroute)
1336 					goto nosourcerouting;
1337 				save_rte(cp, ip->ip_src);
1338 				break;
1339 			}
1340 #ifdef IPSTEALTH
1341 			if (ipstealth)
1342 				goto dropit;
1343 #endif
1344 			if (!ip_dosourceroute) {
1345 				if (ipforwarding) {
1346 					char buf[16]; /* aaa.bbb.ccc.ddd\0 */
1347 					/*
1348 					 * Acting as a router, so generate ICMP
1349 					 */
1350 nosourcerouting:
1351 					strcpy(buf, inet_ntoa(ip->ip_dst));
1352 					log(LOG_WARNING,
1353 					    "attempted source route from %s to %s\n",
1354 					    inet_ntoa(ip->ip_src), buf);
1355 					type = ICMP_UNREACH;
1356 					code = ICMP_UNREACH_SRCFAIL;
1357 					goto bad;
1358 				} else {
1359 					/*
1360 					 * Not acting as a router, so silently drop.
1361 					 */
1362 #ifdef IPSTEALTH
1363 dropit:
1364 #endif
1365 					ipstat.ips_cantforward++;
1366 					m_freem(m);
1367 					return (1);
1368 				}
1369 			}
1370 
1371 			/*
1372 			 * locate outgoing interface
1373 			 */
1374 			(void)memcpy(&ipaddr.sin_addr, cp + off,
1375 			    sizeof(ipaddr.sin_addr));
1376 
1377 			if (opt == IPOPT_SSRR) {
1378 #define	INA	struct in_ifaddr *
1379 #define	SA	struct sockaddr *
1380 			    if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == 0)
1381 				ia = (INA)ifa_ifwithnet((SA)&ipaddr);
1382 			} else
1383 				ia = ip_rtaddr(ipaddr.sin_addr, &ipforward_rt);
1384 			if (ia == 0) {
1385 				type = ICMP_UNREACH;
1386 				code = ICMP_UNREACH_SRCFAIL;
1387 				goto bad;
1388 			}
1389 			ip->ip_dst = ipaddr.sin_addr;
1390 			(void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
1391 			    sizeof(struct in_addr));
1392 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1393 			/*
1394 			 * Let ip_intr's mcast routing check handle mcast pkts
1395 			 */
1396 			forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr));
1397 			break;
1398 
1399 		case IPOPT_RR:
1400 #ifdef IPSTEALTH
1401 			if (ipstealth && pass == 0)
1402 				break;
1403 #endif
1404 			if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1405 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1406 				goto bad;
1407 			}
1408 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1409 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1410 				goto bad;
1411 			}
1412 			/*
1413 			 * If no space remains, ignore.
1414 			 */
1415 			off--;			/* 0 origin */
1416 			if (off > optlen - (int)sizeof(struct in_addr))
1417 				break;
1418 			(void)memcpy(&ipaddr.sin_addr, &ip->ip_dst,
1419 			    sizeof(ipaddr.sin_addr));
1420 			/*
1421 			 * locate outgoing interface; if we're the destination,
1422 			 * use the incoming interface (should be same).
1423 			 */
1424 			if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == 0 &&
1425 			    (ia = ip_rtaddr(ipaddr.sin_addr,
1426 			    &ipforward_rt)) == 0) {
1427 				type = ICMP_UNREACH;
1428 				code = ICMP_UNREACH_HOST;
1429 				goto bad;
1430 			}
1431 			(void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
1432 			    sizeof(struct in_addr));
1433 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1434 			break;
1435 
1436 		case IPOPT_TS:
1437 #ifdef IPSTEALTH
1438 			if (ipstealth && pass == 0)
1439 				break;
1440 #endif
1441 			code = cp - (u_char *)ip;
1442 			if (optlen < 4 || optlen > 40) {
1443 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1444 				goto bad;
1445 			}
1446 			if ((off = cp[IPOPT_OFFSET]) < 5) {
1447 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1448 				goto bad;
1449 			}
1450 			if (off > optlen - (int)sizeof(int32_t)) {
1451 				cp[IPOPT_OFFSET + 1] += (1 << 4);
1452 				if ((cp[IPOPT_OFFSET + 1] & 0xf0) == 0) {
1453 					code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1454 					goto bad;
1455 				}
1456 				break;
1457 			}
1458 			off--;				/* 0 origin */
1459 			sin = (struct in_addr *)(cp + off);
1460 			switch (cp[IPOPT_OFFSET + 1] & 0x0f) {
1461 
1462 			case IPOPT_TS_TSONLY:
1463 				break;
1464 
1465 			case IPOPT_TS_TSANDADDR:
1466 				if (off + sizeof(n_time) +
1467 				    sizeof(struct in_addr) > optlen) {
1468 					code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1469 					goto bad;
1470 				}
1471 				ipaddr.sin_addr = dst;
1472 				ia = (INA)ifaof_ifpforaddr((SA)&ipaddr,
1473 							    m->m_pkthdr.rcvif);
1474 				if (ia == 0)
1475 					continue;
1476 				(void)memcpy(sin, &IA_SIN(ia)->sin_addr,
1477 				    sizeof(struct in_addr));
1478 				cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1479 				off += sizeof(struct in_addr);
1480 				break;
1481 
1482 			case IPOPT_TS_PRESPEC:
1483 				if (off + sizeof(n_time) +
1484 				    sizeof(struct in_addr) > optlen) {
1485 					code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1486 					goto bad;
1487 				}
1488 				(void)memcpy(&ipaddr.sin_addr, sin,
1489 				    sizeof(struct in_addr));
1490 				if (ifa_ifwithaddr((SA)&ipaddr) == 0)
1491 					continue;
1492 				cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1493 				off += sizeof(struct in_addr);
1494 				break;
1495 
1496 			default:
1497 				code = &cp[IPOPT_OFFSET + 1] - (u_char *)ip;
1498 				goto bad;
1499 			}
1500 			ntime = iptime();
1501 			(void)memcpy(cp + off, &ntime, sizeof(n_time));
1502 			cp[IPOPT_OFFSET] += sizeof(n_time);
1503 		}
1504 	}
1505 	if (forward && ipforwarding) {
1506 		ip_forward(m, 1, next_hop);
1507 		return (1);
1508 	}
1509 	return (0);
1510 bad:
1511 	icmp_error(m, type, code, 0, 0);
1512 	ipstat.ips_badoptions++;
1513 	return (1);
1514 }
1515 
1516 /*
1517  * Given address of next destination (final or next hop),
1518  * return internet address info of interface to be used to get there.
1519  */
1520 struct in_ifaddr *
1521 ip_rtaddr(dst, rt)
1522 	struct in_addr dst;
1523 	struct route *rt;
1524 {
1525 	register struct sockaddr_in *sin;
1526 
1527 	sin = (struct sockaddr_in *)&rt->ro_dst;
1528 
1529 	if (rt->ro_rt == 0 ||
1530 	    !(rt->ro_rt->rt_flags & RTF_UP) ||
1531 	    dst.s_addr != sin->sin_addr.s_addr) {
1532 		if (rt->ro_rt) {
1533 			RTFREE(rt->ro_rt);
1534 			rt->ro_rt = 0;
1535 		}
1536 		sin->sin_family = AF_INET;
1537 		sin->sin_len = sizeof(*sin);
1538 		sin->sin_addr = dst;
1539 
1540 		rtalloc_ign(rt, RTF_PRCLONING);
1541 	}
1542 	if (rt->ro_rt == 0)
1543 		return ((struct in_ifaddr *)0);
1544 	return (ifatoia(rt->ro_rt->rt_ifa));
1545 }
1546 
1547 /*
1548  * Save incoming source route for use in replies,
1549  * to be picked up later by ip_srcroute if the receiver is interested.
1550  */
1551 static void
1552 save_rte(option, dst)
1553 	u_char *option;
1554 	struct in_addr dst;
1555 {
1556 	unsigned olen;
1557 
1558 	olen = option[IPOPT_OLEN];
1559 #ifdef DIAGNOSTIC
1560 	if (ipprintfs)
1561 		printf("save_rte: olen %d\n", olen);
1562 #endif
1563 	if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst)))
1564 		return;
1565 	bcopy(option, ip_srcrt.srcopt, olen);
1566 	ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1567 	ip_srcrt.dst = dst;
1568 }
1569 
1570 /*
1571  * Retrieve incoming source route for use in replies,
1572  * in the same form used by setsockopt.
1573  * The first hop is placed before the options, will be removed later.
1574  */
1575 struct mbuf *
1576 ip_srcroute()
1577 {
1578 	register struct in_addr *p, *q;
1579 	register struct mbuf *m;
1580 
1581 	if (ip_nhops == 0)
1582 		return ((struct mbuf *)0);
1583 	m = m_get(M_DONTWAIT, MT_HEADER);
1584 	if (m == 0)
1585 		return ((struct mbuf *)0);
1586 
1587 #define OPTSIZ	(sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt))
1588 
1589 	/* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
1590 	m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) +
1591 	    OPTSIZ;
1592 #ifdef DIAGNOSTIC
1593 	if (ipprintfs)
1594 		printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len);
1595 #endif
1596 
1597 	/*
1598 	 * First save first hop for return route
1599 	 */
1600 	p = &ip_srcrt.route[ip_nhops - 1];
1601 	*(mtod(m, struct in_addr *)) = *p--;
1602 #ifdef DIAGNOSTIC
1603 	if (ipprintfs)
1604 		printf(" hops %lx", (u_long)ntohl(mtod(m, struct in_addr *)->s_addr));
1605 #endif
1606 
1607 	/*
1608 	 * Copy option fields and padding (nop) to mbuf.
1609 	 */
1610 	ip_srcrt.nop = IPOPT_NOP;
1611 	ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
1612 	(void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr),
1613 	    &ip_srcrt.nop, OPTSIZ);
1614 	q = (struct in_addr *)(mtod(m, caddr_t) +
1615 	    sizeof(struct in_addr) + OPTSIZ);
1616 #undef OPTSIZ
1617 	/*
1618 	 * Record return path as an IP source route,
1619 	 * reversing the path (pointers are now aligned).
1620 	 */
1621 	while (p >= ip_srcrt.route) {
1622 #ifdef DIAGNOSTIC
1623 		if (ipprintfs)
1624 			printf(" %lx", (u_long)ntohl(q->s_addr));
1625 #endif
1626 		*q++ = *p--;
1627 	}
1628 	/*
1629 	 * Last hop goes to final destination.
1630 	 */
1631 	*q = ip_srcrt.dst;
1632 #ifdef DIAGNOSTIC
1633 	if (ipprintfs)
1634 		printf(" %lx\n", (u_long)ntohl(q->s_addr));
1635 #endif
1636 	return (m);
1637 }
1638 
1639 /*
1640  * Strip out IP options, at higher
1641  * level protocol in the kernel.
1642  * Second argument is buffer to which options
1643  * will be moved, and return value is their length.
1644  * XXX should be deleted; last arg currently ignored.
1645  */
1646 void
1647 ip_stripoptions(m, mopt)
1648 	register struct mbuf *m;
1649 	struct mbuf *mopt;
1650 {
1651 	register int i;
1652 	struct ip *ip = mtod(m, struct ip *);
1653 	register caddr_t opts;
1654 	int olen;
1655 
1656 	olen = (ip->ip_hl << 2) - sizeof (struct ip);
1657 	opts = (caddr_t)(ip + 1);
1658 	i = m->m_len - (sizeof (struct ip) + olen);
1659 	bcopy(opts + olen, opts, (unsigned)i);
1660 	m->m_len -= olen;
1661 	if (m->m_flags & M_PKTHDR)
1662 		m->m_pkthdr.len -= olen;
1663 	ip->ip_v = IPVERSION;
1664 	ip->ip_hl = sizeof(struct ip) >> 2;
1665 }
1666 
1667 u_char inetctlerrmap[PRC_NCMDS] = {
1668 	0,		0,		0,		0,
1669 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1670 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1671 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1672 	0,		0,		0,		0,
1673 	ENOPROTOOPT,	ECONNREFUSED
1674 };
1675 
1676 /*
1677  * Forward a packet.  If some error occurs return the sender
1678  * an icmp packet.  Note we can't always generate a meaningful
1679  * icmp message because icmp doesn't have a large enough repertoire
1680  * of codes and types.
1681  *
1682  * If not forwarding, just drop the packet.  This could be confusing
1683  * if ipforwarding was zero but some routing protocol was advancing
1684  * us as a gateway to somewhere.  However, we must let the routing
1685  * protocol deal with that.
1686  *
1687  * The srcrt parameter indicates whether the packet is being forwarded
1688  * via a source route.
1689  */
1690 static void
1691 ip_forward(struct mbuf *m, int srcrt, struct sockaddr_in *next_hop)
1692 {
1693 	struct ip *ip = mtod(m, struct ip *);
1694 	struct rtentry *rt;
1695 	int error, type = 0, code = 0;
1696 	struct mbuf *mcopy;
1697 	n_long dest;
1698 	struct in_addr pkt_dst;
1699 	struct ifnet *destifp;
1700 #if defined(IPSEC) || defined(FAST_IPSEC)
1701 	struct ifnet dummyifp;
1702 #endif
1703 
1704 	dest = 0;
1705 	/*
1706 	 * Cache the destination address of the packet; this may be
1707 	 * changed by use of 'ipfw fwd'.
1708 	 */
1709 	pkt_dst = next_hop ? next_hop->sin_addr : ip->ip_dst;
1710 
1711 #ifdef DIAGNOSTIC
1712 	if (ipprintfs)
1713 		printf("forward: src %lx dst %lx ttl %x\n",
1714 		    (u_long)ip->ip_src.s_addr, (u_long)pkt_dst.s_addr,
1715 		    ip->ip_ttl);
1716 #endif
1717 
1718 
1719 	if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(pkt_dst) == 0) {
1720 		ipstat.ips_cantforward++;
1721 		m_freem(m);
1722 		return;
1723 	}
1724 #ifdef IPSTEALTH
1725 	if (!ipstealth) {
1726 #endif
1727 		if (ip->ip_ttl <= IPTTLDEC) {
1728 			icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS,
1729 			    dest, 0);
1730 			return;
1731 		}
1732 #ifdef IPSTEALTH
1733 	}
1734 #endif
1735 
1736 	if (ip_rtaddr(pkt_dst, &ipforward_rt) == 0) {
1737 		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0);
1738 		return;
1739 	} else
1740 		rt = ipforward_rt.ro_rt;
1741 
1742 	/*
1743 	 * Save the IP header and at most 8 bytes of the payload,
1744 	 * in case we need to generate an ICMP message to the src.
1745 	 *
1746 	 * XXX this can be optimized a lot by saving the data in a local
1747 	 * buffer on the stack (72 bytes at most), and only allocating the
1748 	 * mbuf if really necessary. The vast majority of the packets
1749 	 * are forwarded without having to send an ICMP back (either
1750 	 * because unnecessary, or because rate limited), so we are
1751 	 * really we are wasting a lot of work here.
1752 	 *
1753 	 * We don't use m_copy() because it might return a reference
1754 	 * to a shared cluster. Both this function and ip_output()
1755 	 * assume exclusive access to the IP header in `m', so any
1756 	 * data in a cluster may change before we reach icmp_error().
1757 	 */
1758 	MGET(mcopy, M_DONTWAIT, m->m_type);
1759 	if (mcopy != NULL) {
1760 		M_COPY_PKTHDR(mcopy, m);
1761 		mcopy->m_len = imin((ip->ip_hl << 2) + 8,
1762 		    (int)ip->ip_len);
1763 		m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1764 #ifdef MAC
1765 		/*
1766 		 * XXXMAC: This will eventually become an explicit
1767 		 * labeling point.
1768 		 */
1769 		mac_create_mbuf_from_mbuf(m, mcopy);
1770 #endif
1771 	}
1772 
1773 #ifdef IPSTEALTH
1774 	if (!ipstealth) {
1775 #endif
1776 		ip->ip_ttl -= IPTTLDEC;
1777 #ifdef IPSTEALTH
1778 	}
1779 #endif
1780 
1781 	/*
1782 	 * If forwarding packet using same interface that it came in on,
1783 	 * perhaps should send a redirect to sender to shortcut a hop.
1784 	 * Only send redirect if source is sending directly to us,
1785 	 * and if packet was not source routed (or has any options).
1786 	 * Also, don't send redirect if forwarding using a default route
1787 	 * or a route modified by a redirect.
1788 	 */
1789 	if (rt->rt_ifp == m->m_pkthdr.rcvif &&
1790 	    (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1791 	    satosin(rt_key(rt))->sin_addr.s_addr != 0 &&
1792 	    ipsendredirects && !srcrt && !next_hop) {
1793 #define	RTA(rt)	((struct in_ifaddr *)(rt->rt_ifa))
1794 		u_long src = ntohl(ip->ip_src.s_addr);
1795 
1796 		if (RTA(rt) &&
1797 		    (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1798 		    if (rt->rt_flags & RTF_GATEWAY)
1799 			dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
1800 		    else
1801 			dest = pkt_dst.s_addr;
1802 		    /* Router requirements says to only send host redirects */
1803 		    type = ICMP_REDIRECT;
1804 		    code = ICMP_REDIRECT_HOST;
1805 #ifdef DIAGNOSTIC
1806 		    if (ipprintfs)
1807 		        printf("redirect (%d) to %lx\n", code, (u_long)dest);
1808 #endif
1809 		}
1810 	}
1811 
1812     {
1813 	struct m_hdr tag;
1814 
1815 	if (next_hop) {
1816 		/* Pass IPFORWARD info if available */
1817 
1818 		tag.mh_type = MT_TAG;
1819 		tag.mh_flags = PACKET_TAG_IPFORWARD;
1820 		tag.mh_data = (caddr_t)next_hop;
1821 		tag.mh_next = m;
1822 		m = (struct mbuf *)&tag;
1823 	}
1824 	error = ip_output(m, (struct mbuf *)0, &ipforward_rt,
1825 			  IP_FORWARDING, 0, NULL);
1826     }
1827 	if (error)
1828 		ipstat.ips_cantforward++;
1829 	else {
1830 		ipstat.ips_forward++;
1831 		if (type)
1832 			ipstat.ips_redirectsent++;
1833 		else {
1834 			if (mcopy) {
1835 				ipflow_create(&ipforward_rt, mcopy);
1836 				m_freem(mcopy);
1837 			}
1838 			return;
1839 		}
1840 	}
1841 	if (mcopy == NULL)
1842 		return;
1843 	destifp = NULL;
1844 
1845 	switch (error) {
1846 
1847 	case 0:				/* forwarded, but need redirect */
1848 		/* type, code set above */
1849 		break;
1850 
1851 	case ENETUNREACH:		/* shouldn't happen, checked above */
1852 	case EHOSTUNREACH:
1853 	case ENETDOWN:
1854 	case EHOSTDOWN:
1855 	default:
1856 		type = ICMP_UNREACH;
1857 		code = ICMP_UNREACH_HOST;
1858 		break;
1859 
1860 	case EMSGSIZE:
1861 		type = ICMP_UNREACH;
1862 		code = ICMP_UNREACH_NEEDFRAG;
1863 #ifdef IPSEC
1864 		/*
1865 		 * If the packet is routed over IPsec tunnel, tell the
1866 		 * originator the tunnel MTU.
1867 		 *	tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
1868 		 * XXX quickhack!!!
1869 		 */
1870 		if (ipforward_rt.ro_rt) {
1871 			struct secpolicy *sp = NULL;
1872 			int ipsecerror;
1873 			int ipsechdr;
1874 			struct route *ro;
1875 
1876 			sp = ipsec4_getpolicybyaddr(mcopy,
1877 						    IPSEC_DIR_OUTBOUND,
1878 			                            IP_FORWARDING,
1879 			                            &ipsecerror);
1880 
1881 			if (sp == NULL)
1882 				destifp = ipforward_rt.ro_rt->rt_ifp;
1883 			else {
1884 				/* count IPsec header size */
1885 				ipsechdr = ipsec4_hdrsiz(mcopy,
1886 							 IPSEC_DIR_OUTBOUND,
1887 							 NULL);
1888 
1889 				/*
1890 				 * find the correct route for outer IPv4
1891 				 * header, compute tunnel MTU.
1892 				 *
1893 				 * XXX BUG ALERT
1894 				 * The "dummyifp" code relies upon the fact
1895 				 * that icmp_error() touches only ifp->if_mtu.
1896 				 */
1897 				/*XXX*/
1898 				destifp = NULL;
1899 				if (sp->req != NULL
1900 				 && sp->req->sav != NULL
1901 				 && sp->req->sav->sah != NULL) {
1902 					ro = &sp->req->sav->sah->sa_route;
1903 					if (ro->ro_rt && ro->ro_rt->rt_ifp) {
1904 						dummyifp.if_mtu =
1905 						    ro->ro_rt->rt_ifp->if_mtu;
1906 						dummyifp.if_mtu -= ipsechdr;
1907 						destifp = &dummyifp;
1908 					}
1909 				}
1910 
1911 				key_freesp(sp);
1912 			}
1913 		}
1914 #elif FAST_IPSEC
1915 		/*
1916 		 * If the packet is routed over IPsec tunnel, tell the
1917 		 * originator the tunnel MTU.
1918 		 *	tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
1919 		 * XXX quickhack!!!
1920 		 */
1921 		if (ipforward_rt.ro_rt) {
1922 			struct secpolicy *sp = NULL;
1923 			int ipsecerror;
1924 			int ipsechdr;
1925 			struct route *ro;
1926 
1927 			sp = ipsec_getpolicybyaddr(mcopy,
1928 						   IPSEC_DIR_OUTBOUND,
1929 			                           IP_FORWARDING,
1930 			                           &ipsecerror);
1931 
1932 			if (sp == NULL)
1933 				destifp = ipforward_rt.ro_rt->rt_ifp;
1934 			else {
1935 				/* count IPsec header size */
1936 				ipsechdr = ipsec4_hdrsiz(mcopy,
1937 							 IPSEC_DIR_OUTBOUND,
1938 							 NULL);
1939 
1940 				/*
1941 				 * find the correct route for outer IPv4
1942 				 * header, compute tunnel MTU.
1943 				 *
1944 				 * XXX BUG ALERT
1945 				 * The "dummyifp" code relies upon the fact
1946 				 * that icmp_error() touches only ifp->if_mtu.
1947 				 */
1948 				/*XXX*/
1949 				destifp = NULL;
1950 				if (sp->req != NULL
1951 				 && sp->req->sav != NULL
1952 				 && sp->req->sav->sah != NULL) {
1953 					ro = &sp->req->sav->sah->sa_route;
1954 					if (ro->ro_rt && ro->ro_rt->rt_ifp) {
1955 						dummyifp.if_mtu =
1956 						    ro->ro_rt->rt_ifp->if_mtu;
1957 						dummyifp.if_mtu -= ipsechdr;
1958 						destifp = &dummyifp;
1959 					}
1960 				}
1961 
1962 				KEY_FREESP(&sp);
1963 			}
1964 		}
1965 #else /* !IPSEC && !FAST_IPSEC */
1966 		if (ipforward_rt.ro_rt)
1967 			destifp = ipforward_rt.ro_rt->rt_ifp;
1968 #endif /*IPSEC*/
1969 		ipstat.ips_cantfrag++;
1970 		break;
1971 
1972 	case ENOBUFS:
1973 		type = ICMP_SOURCEQUENCH;
1974 		code = 0;
1975 		break;
1976 
1977 	case EACCES:			/* ipfw denied packet */
1978 		m_freem(mcopy);
1979 		return;
1980 	}
1981 	icmp_error(mcopy, type, code, dest, destifp);
1982 }
1983 
1984 void
1985 ip_savecontrol(inp, mp, ip, m)
1986 	register struct inpcb *inp;
1987 	register struct mbuf **mp;
1988 	register struct ip *ip;
1989 	register struct mbuf *m;
1990 {
1991 	if (inp->inp_socket->so_options & SO_TIMESTAMP) {
1992 		struct timeval tv;
1993 
1994 		microtime(&tv);
1995 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1996 			SCM_TIMESTAMP, SOL_SOCKET);
1997 		if (*mp)
1998 			mp = &(*mp)->m_next;
1999 	}
2000 	if (inp->inp_flags & INP_RECVDSTADDR) {
2001 		*mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
2002 		    sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
2003 		if (*mp)
2004 			mp = &(*mp)->m_next;
2005 	}
2006 #ifdef notyet
2007 	/* XXX
2008 	 * Moving these out of udp_input() made them even more broken
2009 	 * than they already were.
2010 	 */
2011 	/* options were tossed already */
2012 	if (inp->inp_flags & INP_RECVOPTS) {
2013 		*mp = sbcreatecontrol((caddr_t) opts_deleted_above,
2014 		    sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
2015 		if (*mp)
2016 			mp = &(*mp)->m_next;
2017 	}
2018 	/* ip_srcroute doesn't do what we want here, need to fix */
2019 	if (inp->inp_flags & INP_RECVRETOPTS) {
2020 		*mp = sbcreatecontrol((caddr_t) ip_srcroute(),
2021 		    sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
2022 		if (*mp)
2023 			mp = &(*mp)->m_next;
2024 	}
2025 #endif
2026 	if (inp->inp_flags & INP_RECVIF) {
2027 		struct ifnet *ifp;
2028 		struct sdlbuf {
2029 			struct sockaddr_dl sdl;
2030 			u_char	pad[32];
2031 		} sdlbuf;
2032 		struct sockaddr_dl *sdp;
2033 		struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
2034 
2035 		if (((ifp = m->m_pkthdr.rcvif))
2036 		&& ( ifp->if_index && (ifp->if_index <= if_index))) {
2037 			sdp = (struct sockaddr_dl *)
2038 			    (ifaddr_byindex(ifp->if_index)->ifa_addr);
2039 			/*
2040 			 * Change our mind and don't try copy.
2041 			 */
2042 			if ((sdp->sdl_family != AF_LINK)
2043 			|| (sdp->sdl_len > sizeof(sdlbuf))) {
2044 				goto makedummy;
2045 			}
2046 			bcopy(sdp, sdl2, sdp->sdl_len);
2047 		} else {
2048 makedummy:
2049 			sdl2->sdl_len
2050 				= offsetof(struct sockaddr_dl, sdl_data[0]);
2051 			sdl2->sdl_family = AF_LINK;
2052 			sdl2->sdl_index = 0;
2053 			sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
2054 		}
2055 		*mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len,
2056 			IP_RECVIF, IPPROTO_IP);
2057 		if (*mp)
2058 			mp = &(*mp)->m_next;
2059 	}
2060 }
2061 
2062 /*
2063  * XXX these routines are called from the upper part of the kernel.
2064  * They need to be locked when we remove Giant.
2065  *
2066  * They could also be moved to ip_mroute.c, since all the RSVP
2067  *  handling is done there already.
2068  */
2069 static int ip_rsvp_on;
2070 struct socket *ip_rsvpd;
2071 int
2072 ip_rsvp_init(struct socket *so)
2073 {
2074 	if (so->so_type != SOCK_RAW ||
2075 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
2076 	  return EOPNOTSUPP;
2077 
2078 	if (ip_rsvpd != NULL)
2079 	  return EADDRINUSE;
2080 
2081 	ip_rsvpd = so;
2082 	/*
2083 	 * This may seem silly, but we need to be sure we don't over-increment
2084 	 * the RSVP counter, in case something slips up.
2085 	 */
2086 	if (!ip_rsvp_on) {
2087 		ip_rsvp_on = 1;
2088 		rsvp_on++;
2089 	}
2090 
2091 	return 0;
2092 }
2093 
2094 int
2095 ip_rsvp_done(void)
2096 {
2097 	ip_rsvpd = NULL;
2098 	/*
2099 	 * This may seem silly, but we need to be sure we don't over-decrement
2100 	 * the RSVP counter, in case something slips up.
2101 	 */
2102 	if (ip_rsvp_on) {
2103 		ip_rsvp_on = 0;
2104 		rsvp_on--;
2105 	}
2106 	return 0;
2107 }
2108