xref: /freebsd/sys/netinet/ip_input.c (revision 2b944ee2b959e9b29fd72dcbf87aad8ad5537bc4)
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 #define	_IP_VHL
38 
39 #include "opt_bootp.h"
40 #include "opt_ipfw.h"
41 #include "opt_ipdn.h"
42 #include "opt_ipdivert.h"
43 #include "opt_ipfilter.h"
44 #include "opt_ipstealth.h"
45 #include "opt_ipsec.h"
46 
47 #include <stddef.h>
48 
49 #include <sys/param.h>
50 #include <sys/systm.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/if.h>
62 #include <net/if_var.h>
63 #include <net/if_dl.h>
64 #include <net/route.h>
65 #include <net/netisr.h>
66 
67 #include <netinet/in.h>
68 #include <netinet/in_systm.h>
69 #include <netinet/in_var.h>
70 #include <netinet/ip.h>
71 #include <netinet/in_pcb.h>
72 #include <netinet/ip_var.h>
73 #include <netinet/ip_icmp.h>
74 #include <machine/in_cksum.h>
75 
76 #include <netinet/ipprotosw.h>
77 
78 #include <sys/socketvar.h>
79 
80 #include <netinet/ip_fw.h>
81 
82 #ifdef IPSEC
83 #include <netinet6/ipsec.h>
84 #include <netkey/key.h>
85 #ifdef IPSEC_DEBUG
86 #include <netkey/key_debug.h>
87 #else
88 #define KEYDEBUG(lev,arg)
89 #endif
90 #endif
91 
92 #include "faith.h"
93 #if defined(NFAITH) && NFAITH > 0
94 #include <net/if_types.h>
95 #endif
96 
97 #ifdef DUMMYNET
98 #include <netinet/ip_dummynet.h>
99 #endif
100 
101 int rsvp_on = 0;
102 static int ip_rsvp_on;
103 struct socket *ip_rsvpd;
104 
105 int	ipforwarding = 0;
106 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW,
107     &ipforwarding, 0, "Enable IP forwarding between interfaces");
108 
109 static int	ipsendredirects = 1; /* XXX */
110 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW,
111     &ipsendredirects, 0, "Enable sending IP redirects");
112 
113 int	ip_defttl = IPDEFTTL;
114 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW,
115     &ip_defttl, 0, "Maximum TTL on IP packets");
116 
117 static int	ip_dosourceroute = 0;
118 SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW,
119     &ip_dosourceroute, 0, "Enable forwarding source routed IP packets");
120 
121 static int	ip_acceptsourceroute = 0;
122 SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute,
123     CTLFLAG_RW, &ip_acceptsourceroute, 0,
124     "Enable accepting source routed IP packets");
125 
126 static int	ip_keepfaith = 0;
127 SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW,
128 	&ip_keepfaith,	0,
129 	"Enable packet capture for FAITH IPv4->IPv6 translater daemon");
130 
131 #ifdef DIAGNOSTIC
132 static int	ipprintfs = 0;
133 #endif
134 
135 extern	struct domain inetdomain;
136 extern	struct ipprotosw inetsw[];
137 u_char	ip_protox[IPPROTO_MAX];
138 static int	ipqmaxlen = IFQ_MAXLEN;
139 struct	in_ifaddrhead in_ifaddrhead; /* first inet address */
140 struct	ifqueue ipintrq;
141 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW,
142     &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue");
143 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD,
144     &ipintrq.ifq_drops, 0, "Number of packets dropped from the IP input queue");
145 
146 struct ipstat ipstat;
147 SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RD,
148     &ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)");
149 
150 /* Packet reassembly stuff */
151 #define IPREASS_NHASH_LOG2      6
152 #define IPREASS_NHASH           (1 << IPREASS_NHASH_LOG2)
153 #define IPREASS_HMASK           (IPREASS_NHASH - 1)
154 #define IPREASS_HASH(x,y) \
155 	(((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK)
156 
157 static struct ipq ipq[IPREASS_NHASH];
158 static int    nipq = 0;         /* total # of reass queues */
159 static int    maxnipq;
160 
161 #ifdef IPCTL_DEFMTU
162 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
163     &ip_mtu, 0, "Default MTU");
164 #endif
165 
166 #ifdef IPSTEALTH
167 static int	ipstealth = 0;
168 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW,
169     &ipstealth, 0, "");
170 #endif
171 
172 
173 /* Firewall hooks */
174 ip_fw_chk_t *ip_fw_chk_ptr;
175 ip_fw_ctl_t *ip_fw_ctl_ptr;
176 
177 #ifdef DUMMYNET
178 ip_dn_ctl_t *ip_dn_ctl_ptr;
179 #endif
180 
181 #if defined(IPFILTER_LKM) || defined(IPFILTER)
182 int (*fr_checkp) __P((struct ip *, int, struct ifnet *, int, struct mbuf **)) = NULL;
183 #endif
184 
185 
186 /*
187  * We need to save the IP options in case a protocol wants to respond
188  * to an incoming packet over the same route if the packet got here
189  * using IP source routing.  This allows connection establishment and
190  * maintenance when the remote end is on a network that is not known
191  * to us.
192  */
193 static int	ip_nhops = 0;
194 static	struct ip_srcrt {
195 	struct	in_addr dst;			/* final destination */
196 	char	nop;				/* one NOP to align */
197 	char	srcopt[IPOPT_OFFSET + 1];	/* OPTVAL, OLEN and OFFSET */
198 	struct	in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)];
199 } ip_srcrt;
200 
201 struct sockaddr_in *ip_fw_fwd_addr;
202 
203 static void	save_rte __P((u_char *, struct in_addr));
204 static int	ip_dooptions __P((struct mbuf *));
205 static void	ip_forward __P((struct mbuf *, int));
206 static void	ip_freef __P((struct ipq *));
207 #ifdef IPDIVERT
208 static struct	mbuf *ip_reass __P((struct mbuf *,
209 			struct ipq *, struct ipq *, u_int32_t *, u_int16_t *));
210 #else
211 static struct	mbuf *ip_reass __P((struct mbuf *, struct ipq *, struct ipq *));
212 #endif
213 static struct	in_ifaddr *ip_rtaddr __P((struct in_addr));
214 static void	ipintr __P((void));
215 
216 /*
217  * IP initialization: fill in IP protocol switch table.
218  * All protocols not implemented in kernel go to raw IP protocol handler.
219  */
220 void
221 ip_init()
222 {
223 	register struct ipprotosw *pr;
224 	register int i;
225 
226 	TAILQ_INIT(&in_ifaddrhead);
227 	pr = (struct ipprotosw *)pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
228 	if (pr == 0)
229 		panic("ip_init");
230 	for (i = 0; i < IPPROTO_MAX; i++)
231 		ip_protox[i] = pr - inetsw;
232 	for (pr = (struct ipprotosw *)inetdomain.dom_protosw;
233 	    pr < (struct ipprotosw *)inetdomain.dom_protoswNPROTOSW; pr++)
234 		if (pr->pr_domain->dom_family == PF_INET &&
235 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
236 			ip_protox[pr->pr_protocol] = pr - inetsw;
237 
238 	for (i = 0; i < IPREASS_NHASH; i++)
239 	    ipq[i].next = ipq[i].prev = &ipq[i];
240 
241 	maxnipq = nmbclusters/4;
242 
243 	ip_id = time_second & 0xffff;
244 	ipintrq.ifq_maxlen = ipqmaxlen;
245 }
246 
247 static struct	sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET };
248 static struct	route ipforward_rt;
249 
250 /*
251  * Ip input routine.  Checksum and byte swap header.  If fragmented
252  * try to reassemble.  Process options.  Pass to next level.
253  */
254 void
255 ip_input(struct mbuf *m)
256 {
257 	struct ip *ip;
258 	struct ipq *fp;
259 	struct in_ifaddr *ia;
260 	int    i, hlen, mff;
261 	u_short sum;
262 	u_int16_t divert_cookie;		/* firewall cookie */
263 #ifdef IPDIVERT
264 	u_int32_t divert_info = 0;		/* packet divert/tee info */
265 #endif
266 	struct ip_fw_chain *rule = NULL;
267 
268 #ifdef IPDIVERT
269 	/* Get and reset firewall cookie */
270 	divert_cookie = ip_divert_cookie;
271 	ip_divert_cookie = 0;
272 #else
273 	divert_cookie = 0;
274 #endif
275 
276 #if defined(IPFIREWALL) && defined(DUMMYNET)
277         /*
278          * dummynet packet are prepended a vestigial mbuf with
279          * m_type = MT_DUMMYNET and m_data pointing to the matching
280          * rule.
281          */
282         if (m->m_type == MT_DUMMYNET) {
283             rule = (struct ip_fw_chain *)(m->m_data) ;
284             m = m->m_next ;
285             ip = mtod(m, struct ip *);
286             hlen = IP_VHL_HL(ip->ip_vhl) << 2;
287             goto iphack ;
288         } else
289             rule = NULL ;
290 #endif
291 
292 #ifdef	DIAGNOSTIC
293 	if (m == NULL || (m->m_flags & M_PKTHDR) == 0)
294 		panic("ip_input no HDR");
295 #endif
296 	ipstat.ips_total++;
297 
298 	if (m->m_pkthdr.len < sizeof(struct ip))
299 		goto tooshort;
300 
301 	if (m->m_len < sizeof (struct ip) &&
302 	    (m = m_pullup(m, sizeof (struct ip))) == 0) {
303 		ipstat.ips_toosmall++;
304 		return;
305 	}
306 	ip = mtod(m, struct ip *);
307 
308 	if (IP_VHL_V(ip->ip_vhl) != IPVERSION) {
309 		ipstat.ips_badvers++;
310 		goto bad;
311 	}
312 
313 	hlen = IP_VHL_HL(ip->ip_vhl) << 2;
314 	if (hlen < sizeof(struct ip)) {	/* minimum header length */
315 		ipstat.ips_badhlen++;
316 		goto bad;
317 	}
318 	if (hlen > m->m_len) {
319 		if ((m = m_pullup(m, hlen)) == 0) {
320 			ipstat.ips_badhlen++;
321 			return;
322 		}
323 		ip = mtod(m, struct ip *);
324 	}
325 	if (hlen == sizeof(struct ip)) {
326 		sum = in_cksum_hdr(ip);
327 	} else {
328 		sum = in_cksum(m, hlen);
329 	}
330 	if (sum) {
331 		ipstat.ips_badsum++;
332 		goto bad;
333 	}
334 
335 	/*
336 	 * Convert fields to host representation.
337 	 */
338 	NTOHS(ip->ip_len);
339 	if (ip->ip_len < hlen) {
340 		ipstat.ips_badlen++;
341 		goto bad;
342 	}
343 	NTOHS(ip->ip_id);
344 	NTOHS(ip->ip_off);
345 
346 	/*
347 	 * Check that the amount of data in the buffers
348 	 * is as at least much as the IP header would have us expect.
349 	 * Trim mbufs if longer than we expect.
350 	 * Drop packet if shorter than we expect.
351 	 */
352 	if (m->m_pkthdr.len < ip->ip_len) {
353 tooshort:
354 		ipstat.ips_tooshort++;
355 		goto bad;
356 	}
357 	if (m->m_pkthdr.len > ip->ip_len) {
358 		if (m->m_len == m->m_pkthdr.len) {
359 			m->m_len = ip->ip_len;
360 			m->m_pkthdr.len = ip->ip_len;
361 		} else
362 			m_adj(m, ip->ip_len - m->m_pkthdr.len);
363 	}
364 	/*
365 	 * IpHack's section.
366 	 * Right now when no processing on packet has done
367 	 * and it is still fresh out of network we do our black
368 	 * deals with it.
369 	 * - Firewall: deny/allow/divert
370 	 * - Xlate: translate packet's addr/port (NAT).
371 	 * - Pipe: pass pkt through dummynet.
372 	 * - Wrap: fake packet's addr/port <unimpl.>
373 	 * - Encapsulate: put it in another IP and send out. <unimp.>
374  	 */
375 
376 #if defined(IPFIREWALL) && defined(DUMMYNET)
377 iphack:
378 #endif
379 #if defined(IPFILTER) || defined(IPFILTER_LKM)
380 	/*
381 	 * Check if we want to allow this packet to be processed.
382 	 * Consider it to be bad if not.
383 	 */
384 	if (fr_checkp) {
385 		struct	mbuf	*m1 = m;
386 
387 		if ((*fr_checkp)(ip, hlen, m->m_pkthdr.rcvif, 0, &m1) || !m1)
388 			return;
389 		ip = mtod(m = m1, struct ip *);
390 	}
391 #endif
392 	if (ip_fw_chk_ptr) {
393 #ifdef IPFIREWALL_FORWARD
394 		/*
395 		 * If we've been forwarded from the output side, then
396 		 * skip the firewall a second time
397 		 */
398 		if (ip_fw_fwd_addr)
399 			goto ours;
400 #endif	/* IPFIREWALL_FORWARD */
401 		/*
402 		 * See the comment in ip_output for the return values
403 		 * produced by the firewall.
404 		 */
405 		i = (*ip_fw_chk_ptr)(&ip,
406 		    hlen, NULL, &divert_cookie, &m, &rule, &ip_fw_fwd_addr);
407 		if (m == NULL)		/* Packet discarded by firewall */
408 			return;
409 		if (i == 0 && ip_fw_fwd_addr == NULL)	/* common case */
410 			goto pass;
411 #ifdef DUMMYNET
412                 if ((i & IP_FW_PORT_DYNT_FLAG) != 0) {
413                         /* Send packet to the appropriate pipe */
414                         dummynet_io(i&0xffff,DN_TO_IP_IN,m,NULL,NULL,0, rule,
415 				    0);
416 			return;
417 		}
418 #endif
419 #ifdef IPDIVERT
420 		if (i != 0 && (i & IP_FW_PORT_DYNT_FLAG) == 0) {
421 			/* Divert or tee packet */
422 			divert_info = i;
423 			goto ours;
424 		}
425 #endif
426 #ifdef IPFIREWALL_FORWARD
427 		if (i == 0 && ip_fw_fwd_addr != NULL)
428 			goto pass;
429 #endif
430 		/*
431 		 * if we get here, the packet must be dropped
432 		 */
433 		m_freem(m);
434 		return;
435 	}
436 pass:
437 
438 	/*
439 	 * Process options and, if not destined for us,
440 	 * ship it on.  ip_dooptions returns 1 when an
441 	 * error was detected (causing an icmp message
442 	 * to be sent and the original packet to be freed).
443 	 */
444 	ip_nhops = 0;		/* for source routed packets */
445 	if (hlen > sizeof (struct ip) && ip_dooptions(m)) {
446 #ifdef IPFIREWALL_FORWARD
447 		ip_fw_fwd_addr = NULL;
448 #endif
449 		return;
450 	}
451 
452         /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
453          * matter if it is destined to another node, or whether it is
454          * a multicast one, RSVP wants it! and prevents it from being forwarded
455          * anywhere else. Also checks if the rsvp daemon is running before
456 	 * grabbing the packet.
457          */
458 	if (rsvp_on && ip->ip_p==IPPROTO_RSVP)
459 		goto ours;
460 
461 	/*
462 	 * Check our list of addresses, to see if the packet is for us.
463 	 * If we don't have any addresses, assume any unicast packet
464 	 * we receive might be for us (and let the upper layers deal
465 	 * with it).
466 	 */
467 	if (TAILQ_EMPTY(&in_ifaddrhead) &&
468 	    (m->m_flags & (M_MCAST|M_BCAST)) == 0)
469 		goto ours;
470 
471 	for (ia = TAILQ_FIRST(&in_ifaddrhead); ia;
472 					ia = TAILQ_NEXT(ia, ia_link)) {
473 #define	satosin(sa)	((struct sockaddr_in *)(sa))
474 
475 #ifdef BOOTP_COMPAT
476 		if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY)
477 			goto ours;
478 #endif
479 #ifdef IPFIREWALL_FORWARD
480 		/*
481 		 * If the addr to forward to is one of ours, we pretend to
482 		 * be the destination for this packet.
483 		 */
484 		if (ip_fw_fwd_addr == NULL) {
485 			if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr)
486 				goto ours;
487 		} else if (IA_SIN(ia)->sin_addr.s_addr ==
488 					 ip_fw_fwd_addr->sin_addr.s_addr)
489 			goto ours;
490 #else
491 		if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr)
492 			goto ours;
493 #endif
494 		if (ia->ia_ifp && ia->ia_ifp->if_flags & IFF_BROADCAST) {
495 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
496 			    ip->ip_dst.s_addr)
497 				goto ours;
498 			if (ip->ip_dst.s_addr == ia->ia_netbroadcast.s_addr)
499 				goto ours;
500 		}
501 	}
502 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
503 		struct in_multi *inm;
504 		if (ip_mrouter) {
505 			/*
506 			 * If we are acting as a multicast router, all
507 			 * incoming multicast packets are passed to the
508 			 * kernel-level multicast forwarding function.
509 			 * The packet is returned (relatively) intact; if
510 			 * ip_mforward() returns a non-zero value, the packet
511 			 * must be discarded, else it may be accepted below.
512 			 *
513 			 * (The IP ident field is put in the same byte order
514 			 * as expected when ip_mforward() is called from
515 			 * ip_output().)
516 			 */
517 			ip->ip_id = htons(ip->ip_id);
518 			if (ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) {
519 				ipstat.ips_cantforward++;
520 				m_freem(m);
521 				return;
522 			}
523 			ip->ip_id = ntohs(ip->ip_id);
524 
525 			/*
526 			 * The process-level routing demon needs to receive
527 			 * all multicast IGMP packets, whether or not this
528 			 * host belongs to their destination groups.
529 			 */
530 			if (ip->ip_p == IPPROTO_IGMP)
531 				goto ours;
532 			ipstat.ips_forward++;
533 		}
534 		/*
535 		 * See if we belong to the destination multicast group on the
536 		 * arrival interface.
537 		 */
538 		IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
539 		if (inm == NULL) {
540 			ipstat.ips_notmember++;
541 			m_freem(m);
542 			return;
543 		}
544 		goto ours;
545 	}
546 	if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
547 		goto ours;
548 	if (ip->ip_dst.s_addr == INADDR_ANY)
549 		goto ours;
550 
551 #if defined(NFAITH) && 0 < NFAITH
552 	/*
553 	 * FAITH(Firewall Aided Internet Translator)
554 	 */
555 	if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
556 		if (ip_keepfaith) {
557 			if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP)
558 				goto ours;
559 		}
560 		m_freem(m);
561 		return;
562 	}
563 #endif
564 	/*
565 	 * Not for us; forward if possible and desirable.
566 	 */
567 	if (ipforwarding == 0) {
568 		ipstat.ips_cantforward++;
569 		m_freem(m);
570 	} else
571 		ip_forward(m, 0);
572 #ifdef IPFIREWALL_FORWARD
573 	ip_fw_fwd_addr = NULL;
574 #endif
575 	return;
576 
577 ours:
578 
579 	/*
580 	 * If offset or IP_MF are set, must reassemble.
581 	 * Otherwise, nothing need be done.
582 	 * (We could look in the reassembly queue to see
583 	 * if the packet was previously fragmented,
584 	 * but it's not worth the time; just let them time out.)
585 	 */
586 	if (ip->ip_off & (IP_MF | IP_OFFMASK | IP_RF)) {
587 
588 #if 0	/*
589 	 * Reassembly should be able to treat a mbuf cluster, for later
590 	 * operation of contiguous protocol headers on the cluster. (KAME)
591 	 */
592 		if (m->m_flags & M_EXT) {		/* XXX */
593 			if ((m = m_pullup(m, hlen)) == 0) {
594 				ipstat.ips_toosmall++;
595 #ifdef IPFIREWALL_FORWARD
596 				ip_fw_fwd_addr = NULL;
597 #endif
598 				return;
599 			}
600 			ip = mtod(m, struct ip *);
601 		}
602 #endif
603 		sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
604 		/*
605 		 * Look for queue of fragments
606 		 * of this datagram.
607 		 */
608 		for (fp = ipq[sum].next; fp != &ipq[sum]; fp = fp->next)
609 			if (ip->ip_id == fp->ipq_id &&
610 			    ip->ip_src.s_addr == fp->ipq_src.s_addr &&
611 			    ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
612 			    ip->ip_p == fp->ipq_p)
613 				goto found;
614 
615 		fp = 0;
616 
617 		/* check if there's a place for the new queue */
618 		if (nipq > maxnipq) {
619 		    /*
620 		     * drop something from the tail of the current queue
621 		     * before proceeding further
622 		     */
623 		    if (ipq[sum].prev == &ipq[sum]) {   /* gak */
624 			for (i = 0; i < IPREASS_NHASH; i++) {
625 			    if (ipq[i].prev != &ipq[i]) {
626 				ip_freef(ipq[i].prev);
627 				break;
628 			    }
629 			}
630 		    } else
631 			ip_freef(ipq[sum].prev);
632 		}
633 found:
634 		/*
635 		 * Adjust ip_len to not reflect header,
636 		 * set ip_mff if more fragments are expected,
637 		 * convert offset of this to bytes.
638 		 */
639 		ip->ip_len -= hlen;
640 		mff = (ip->ip_off & IP_MF) != 0;
641 		if (mff) {
642 		        /*
643 		         * Make sure that fragments have a data length
644 			 * that's a non-zero multiple of 8 bytes.
645 		         */
646 			if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
647 				ipstat.ips_toosmall++; /* XXX */
648 				goto bad;
649 			}
650 			m->m_flags |= M_FRAG;
651 		}
652 		ip->ip_off <<= 3;
653 
654 		/*
655 		 * If datagram marked as having more fragments
656 		 * or if this is not the first fragment,
657 		 * attempt reassembly; if it succeeds, proceed.
658 		 */
659 		if (mff || ip->ip_off) {
660 			ipstat.ips_fragments++;
661 			m->m_pkthdr.header = ip;
662 #ifdef IPDIVERT
663 			m = ip_reass(m,
664 			    fp, &ipq[sum], &divert_info, &divert_cookie);
665 #else
666 			m = ip_reass(m, fp, &ipq[sum]);
667 #endif
668 			if (m == 0) {
669 #ifdef IPFIREWALL_FORWARD
670 				ip_fw_fwd_addr = NULL;
671 #endif
672 				return;
673 			}
674 			/* Get the length of the reassembled packets header */
675 			hlen = IP_VHL_HL(ip->ip_vhl) << 2;
676 			ipstat.ips_reassembled++;
677 			ip = mtod(m, struct ip *);
678 #ifdef IPDIVERT
679 			/* Restore original checksum before diverting packet */
680 			if (divert_info != 0) {
681 				ip->ip_len += hlen;
682 				HTONS(ip->ip_len);
683 				HTONS(ip->ip_off);
684 				HTONS(ip->ip_id);
685 				ip->ip_sum = 0;
686 				ip->ip_sum = in_cksum_hdr(ip);
687 				NTOHS(ip->ip_id);
688 				NTOHS(ip->ip_off);
689 				NTOHS(ip->ip_len);
690 				ip->ip_len -= hlen;
691 			}
692 #endif
693 		} else
694 			if (fp)
695 				ip_freef(fp);
696 	} else
697 		ip->ip_len -= hlen;
698 
699 #ifdef IPDIVERT
700 	/*
701 	 * Divert or tee packet to the divert protocol if required.
702 	 *
703 	 * If divert_info is zero then cookie should be too, so we shouldn't
704 	 * need to clear them here.  Assume divert_packet() does so also.
705 	 */
706 	if (divert_info != 0) {
707 		struct mbuf *clone = NULL;
708 
709 		/* Clone packet if we're doing a 'tee' */
710 		if ((divert_info & IP_FW_PORT_TEE_FLAG) != 0)
711 			clone = m_dup(m, M_DONTWAIT);
712 
713 		/* Restore packet header fields to original values */
714 		ip->ip_len += hlen;
715 		HTONS(ip->ip_len);
716 		HTONS(ip->ip_off);
717 		HTONS(ip->ip_id);
718 
719 		/* Deliver packet to divert input routine */
720 		ip_divert_cookie = divert_cookie;
721 		divert_packet(m, 1, divert_info & 0xffff);
722 		ipstat.ips_delivered++;
723 
724 		/* If 'tee', continue with original packet */
725 		if (clone == NULL)
726 			return;
727 		m = clone;
728 		ip = mtod(m, struct ip *);
729 	}
730 #endif
731 
732 	/*
733 	 * Switch out to protocol's input routine.
734 	 */
735 	ipstat.ips_delivered++;
736     {
737 	int off = hlen, nh = ip->ip_p;
738 
739 	(*inetsw[ip_protox[ip->ip_p]].pr_input)(m, off, nh);
740 #ifdef	IPFIREWALL_FORWARD
741 	ip_fw_fwd_addr = NULL;	/* tcp needed it */
742 #endif
743 	return;
744     }
745 bad:
746 #ifdef	IPFIREWALL_FORWARD
747 	ip_fw_fwd_addr = NULL;
748 #endif
749 	m_freem(m);
750 }
751 
752 /*
753  * IP software interrupt routine - to go away sometime soon
754  */
755 static void
756 ipintr(void)
757 {
758 	int s;
759 	struct mbuf *m;
760 
761 	while(1) {
762 		s = splimp();
763 		IF_DEQUEUE(&ipintrq, m);
764 		splx(s);
765 		if (m == 0)
766 			return;
767 		ip_input(m);
768 	}
769 }
770 
771 NETISR_SET(NETISR_IP, ipintr);
772 
773 /*
774  * Take incoming datagram fragment and try to reassemble it into
775  * whole datagram.  If a chain for reassembly of this datagram already
776  * exists, then it is given as fp; otherwise have to make a chain.
777  *
778  * When IPDIVERT enabled, keep additional state with each packet that
779  * tells us if we need to divert or tee the packet we're building.
780  */
781 
782 static struct mbuf *
783 #ifdef IPDIVERT
784 ip_reass(m, fp, where, divinfo, divcookie)
785 #else
786 ip_reass(m, fp, where)
787 #endif
788 	register struct mbuf *m;
789 	register struct ipq *fp;
790 	struct   ipq    *where;
791 #ifdef IPDIVERT
792 	u_int32_t *divinfo;
793 	u_int16_t *divcookie;
794 #endif
795 {
796 	struct ip *ip = mtod(m, struct ip *);
797 	register struct mbuf *p = 0, *q, *nq;
798 	struct mbuf *t;
799 	int hlen = IP_VHL_HL(ip->ip_vhl) << 2;
800 	int i, next;
801 
802 	/*
803 	 * Presence of header sizes in mbufs
804 	 * would confuse code below.
805 	 */
806 	m->m_data += hlen;
807 	m->m_len -= hlen;
808 
809 	/*
810 	 * If first fragment to arrive, create a reassembly queue.
811 	 */
812 	if (fp == 0) {
813 		if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL)
814 			goto dropfrag;
815 		fp = mtod(t, struct ipq *);
816 		insque(fp, where);
817 		nipq++;
818 		fp->ipq_ttl = IPFRAGTTL;
819 		fp->ipq_p = ip->ip_p;
820 		fp->ipq_id = ip->ip_id;
821 		fp->ipq_src = ip->ip_src;
822 		fp->ipq_dst = ip->ip_dst;
823 		fp->ipq_frags = m;
824 		m->m_nextpkt = NULL;
825 #ifdef IPDIVERT
826 		fp->ipq_div_info = 0;
827 		fp->ipq_div_cookie = 0;
828 #endif
829 		goto inserted;
830 	}
831 
832 #define GETIP(m)	((struct ip*)((m)->m_pkthdr.header))
833 
834 	/*
835 	 * Find a segment which begins after this one does.
836 	 */
837 	for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt)
838 		if (GETIP(q)->ip_off > ip->ip_off)
839 			break;
840 
841 	/*
842 	 * If there is a preceding segment, it may provide some of
843 	 * our data already.  If so, drop the data from the incoming
844 	 * segment.  If it provides all of our data, drop us, otherwise
845 	 * stick new segment in the proper place.
846 	 */
847 	if (p) {
848 		i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off;
849 		if (i > 0) {
850 			if (i >= ip->ip_len)
851 				goto dropfrag;
852 			m_adj(m, i);
853 			ip->ip_off += i;
854 			ip->ip_len -= i;
855 		}
856 		m->m_nextpkt = p->m_nextpkt;
857 		p->m_nextpkt = m;
858 	} else {
859 		m->m_nextpkt = fp->ipq_frags;
860 		fp->ipq_frags = m;
861 	}
862 
863 	/*
864 	 * While we overlap succeeding segments trim them or,
865 	 * if they are completely covered, dequeue them.
866 	 */
867 	for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off;
868 	     q = nq) {
869 		i = (ip->ip_off + ip->ip_len) -
870 		    GETIP(q)->ip_off;
871 		if (i < GETIP(q)->ip_len) {
872 			GETIP(q)->ip_len -= i;
873 			GETIP(q)->ip_off += i;
874 			m_adj(q, i);
875 			break;
876 		}
877 		nq = q->m_nextpkt;
878 		m->m_nextpkt = nq;
879 		m_freem(q);
880 	}
881 
882 inserted:
883 
884 #ifdef IPDIVERT
885 	/*
886 	 * Transfer firewall instructions to the fragment structure.
887 	 * Any fragment diverting causes the whole packet to divert.
888 	 */
889 	fp->ipq_div_info = *divinfo;
890 	fp->ipq_div_cookie = *divcookie;
891 	*divinfo = 0;
892 	*divcookie = 0;
893 #endif
894 
895 	/*
896 	 * Check for complete reassembly.
897 	 */
898 	next = 0;
899 	for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
900 		if (GETIP(q)->ip_off != next)
901 			return (0);
902 		next += GETIP(q)->ip_len;
903 	}
904 	/* Make sure the last packet didn't have the IP_MF flag */
905 	if (p->m_flags & M_FRAG)
906 		return (0);
907 
908 	/*
909 	 * Reassembly is complete.  Make sure the packet is a sane size.
910 	 */
911 	q = fp->ipq_frags;
912 	ip = GETIP(q);
913 	if (next + (IP_VHL_HL(ip->ip_vhl) << 2) > IP_MAXPACKET) {
914 		ipstat.ips_toolong++;
915 		ip_freef(fp);
916 		return (0);
917 	}
918 
919 	/*
920 	 * Concatenate fragments.
921 	 */
922 	m = q;
923 	t = m->m_next;
924 	m->m_next = 0;
925 	m_cat(m, t);
926 	nq = q->m_nextpkt;
927 	q->m_nextpkt = 0;
928 	for (q = nq; q != NULL; q = nq) {
929 		nq = q->m_nextpkt;
930 		q->m_nextpkt = NULL;
931 		m_cat(m, q);
932 	}
933 
934 #ifdef IPDIVERT
935 	/*
936 	 * Extract firewall instructions from the fragment structure.
937 	 */
938 	*divinfo = fp->ipq_div_info;
939 	*divcookie = fp->ipq_div_cookie;
940 #endif
941 
942 	/*
943 	 * Create header for new ip packet by
944 	 * modifying header of first packet;
945 	 * dequeue and discard fragment reassembly header.
946 	 * Make header visible.
947 	 */
948 	ip->ip_len = next;
949 	ip->ip_src = fp->ipq_src;
950 	ip->ip_dst = fp->ipq_dst;
951 	remque(fp);
952 	nipq--;
953 	(void) m_free(dtom(fp));
954 	m->m_len += (IP_VHL_HL(ip->ip_vhl) << 2);
955 	m->m_data -= (IP_VHL_HL(ip->ip_vhl) << 2);
956 	/* some debugging cruft by sklower, below, will go away soon */
957 	if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */
958 		register int plen = 0;
959 		for (t = m; t; t = t->m_next)
960 			plen += t->m_len;
961 		m->m_pkthdr.len = plen;
962 	}
963 	return (m);
964 
965 dropfrag:
966 #ifdef IPDIVERT
967 	*divinfo = 0;
968 	*divcookie = 0;
969 #endif
970 	ipstat.ips_fragdropped++;
971 	m_freem(m);
972 	return (0);
973 
974 #undef GETIP
975 }
976 
977 /*
978  * Free a fragment reassembly header and all
979  * associated datagrams.
980  */
981 static void
982 ip_freef(fp)
983 	struct ipq *fp;
984 {
985 	register struct mbuf *q;
986 
987 	while (fp->ipq_frags) {
988 		q = fp->ipq_frags;
989 		fp->ipq_frags = q->m_nextpkt;
990 		m_freem(q);
991 	}
992 	remque(fp);
993 	(void) m_free(dtom(fp));
994 	nipq--;
995 }
996 
997 /*
998  * IP timer processing;
999  * if a timer expires on a reassembly
1000  * queue, discard it.
1001  */
1002 void
1003 ip_slowtimo()
1004 {
1005 	register struct ipq *fp;
1006 	int s = splnet();
1007 	int i;
1008 
1009 	for (i = 0; i < IPREASS_NHASH; i++) {
1010 		fp = ipq[i].next;
1011 		if (fp == 0)
1012 			continue;
1013 		while (fp != &ipq[i]) {
1014 			--fp->ipq_ttl;
1015 			fp = fp->next;
1016 			if (fp->prev->ipq_ttl == 0) {
1017 				ipstat.ips_fragtimeout++;
1018 				ip_freef(fp->prev);
1019 			}
1020 		}
1021 	}
1022 	ipflow_slowtimo();
1023 	splx(s);
1024 }
1025 
1026 /*
1027  * Drain off all datagram fragments.
1028  */
1029 void
1030 ip_drain()
1031 {
1032 	int     i;
1033 
1034 	for (i = 0; i < IPREASS_NHASH; i++) {
1035 		while (ipq[i].next != &ipq[i]) {
1036 			ipstat.ips_fragdropped++;
1037 			ip_freef(ipq[i].next);
1038 		}
1039 	}
1040 	in_rtqdrain();
1041 }
1042 
1043 /*
1044  * Do option processing on a datagram,
1045  * possibly discarding it if bad options are encountered,
1046  * or forwarding it if source-routed.
1047  * Returns 1 if packet has been forwarded/freed,
1048  * 0 if the packet should be processed further.
1049  */
1050 static int
1051 ip_dooptions(m)
1052 	struct mbuf *m;
1053 {
1054 	register struct ip *ip = mtod(m, struct ip *);
1055 	register u_char *cp;
1056 	register struct ip_timestamp *ipt;
1057 	register struct in_ifaddr *ia;
1058 	int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0;
1059 	struct in_addr *sin, dst;
1060 	n_time ntime;
1061 
1062 	dst = ip->ip_dst;
1063 	cp = (u_char *)(ip + 1);
1064 	cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip);
1065 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1066 		opt = cp[IPOPT_OPTVAL];
1067 		if (opt == IPOPT_EOL)
1068 			break;
1069 		if (opt == IPOPT_NOP)
1070 			optlen = 1;
1071 		else {
1072 			optlen = cp[IPOPT_OLEN];
1073 			if (optlen <= 0 || optlen > cnt) {
1074 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1075 				goto bad;
1076 			}
1077 		}
1078 		switch (opt) {
1079 
1080 		default:
1081 			break;
1082 
1083 		/*
1084 		 * Source routing with record.
1085 		 * Find interface with current destination address.
1086 		 * If none on this machine then drop if strictly routed,
1087 		 * or do nothing if loosely routed.
1088 		 * Record interface address and bring up next address
1089 		 * component.  If strictly routed make sure next
1090 		 * address is on directly accessible net.
1091 		 */
1092 		case IPOPT_LSRR:
1093 		case IPOPT_SSRR:
1094 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1095 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1096 				goto bad;
1097 			}
1098 			ipaddr.sin_addr = ip->ip_dst;
1099 			ia = (struct in_ifaddr *)
1100 				ifa_ifwithaddr((struct sockaddr *)&ipaddr);
1101 			if (ia == 0) {
1102 				if (opt == IPOPT_SSRR) {
1103 					type = ICMP_UNREACH;
1104 					code = ICMP_UNREACH_SRCFAIL;
1105 					goto bad;
1106 				}
1107 				if (!ip_dosourceroute)
1108 					goto nosourcerouting;
1109 				/*
1110 				 * Loose routing, and not at next destination
1111 				 * yet; nothing to do except forward.
1112 				 */
1113 				break;
1114 			}
1115 			off--;			/* 0 origin */
1116 			if (off > optlen - sizeof(struct in_addr)) {
1117 				/*
1118 				 * End of source route.  Should be for us.
1119 				 */
1120 				if (!ip_acceptsourceroute)
1121 					goto nosourcerouting;
1122 				save_rte(cp, ip->ip_src);
1123 				break;
1124 			}
1125 
1126 			if (!ip_dosourceroute) {
1127 				if (ipforwarding) {
1128 					char buf[16]; /* aaa.bbb.ccc.ddd\0 */
1129 					/*
1130 					 * Acting as a router, so generate ICMP
1131 					 */
1132 nosourcerouting:
1133 					strcpy(buf, inet_ntoa(ip->ip_dst));
1134 					log(LOG_WARNING,
1135 					    "attempted source route from %s to %s\n",
1136 					    inet_ntoa(ip->ip_src), buf);
1137 					type = ICMP_UNREACH;
1138 					code = ICMP_UNREACH_SRCFAIL;
1139 					goto bad;
1140 				} else {
1141 					/*
1142 					 * Not acting as a router, so silently drop.
1143 					 */
1144 					ipstat.ips_cantforward++;
1145 					m_freem(m);
1146 					return (1);
1147 				}
1148 			}
1149 
1150 			/*
1151 			 * locate outgoing interface
1152 			 */
1153 			(void)memcpy(&ipaddr.sin_addr, cp + off,
1154 			    sizeof(ipaddr.sin_addr));
1155 
1156 			if (opt == IPOPT_SSRR) {
1157 #define	INA	struct in_ifaddr *
1158 #define	SA	struct sockaddr *
1159 			    if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == 0)
1160 				ia = (INA)ifa_ifwithnet((SA)&ipaddr);
1161 			} else
1162 				ia = ip_rtaddr(ipaddr.sin_addr);
1163 			if (ia == 0) {
1164 				type = ICMP_UNREACH;
1165 				code = ICMP_UNREACH_SRCFAIL;
1166 				goto bad;
1167 			}
1168 			ip->ip_dst = ipaddr.sin_addr;
1169 			(void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
1170 			    sizeof(struct in_addr));
1171 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1172 			/*
1173 			 * Let ip_intr's mcast routing check handle mcast pkts
1174 			 */
1175 			forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr));
1176 			break;
1177 
1178 		case IPOPT_RR:
1179 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1180 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1181 				goto bad;
1182 			}
1183 			/*
1184 			 * If no space remains, ignore.
1185 			 */
1186 			off--;			/* 0 origin */
1187 			if (off > optlen - sizeof(struct in_addr))
1188 				break;
1189 			(void)memcpy(&ipaddr.sin_addr, &ip->ip_dst,
1190 			    sizeof(ipaddr.sin_addr));
1191 			/*
1192 			 * locate outgoing interface; if we're the destination,
1193 			 * use the incoming interface (should be same).
1194 			 */
1195 			if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == 0 &&
1196 			    (ia = ip_rtaddr(ipaddr.sin_addr)) == 0) {
1197 				type = ICMP_UNREACH;
1198 				code = ICMP_UNREACH_HOST;
1199 				goto bad;
1200 			}
1201 			(void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
1202 			    sizeof(struct in_addr));
1203 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1204 			break;
1205 
1206 		case IPOPT_TS:
1207 			code = cp - (u_char *)ip;
1208 			ipt = (struct ip_timestamp *)cp;
1209 			if (ipt->ipt_len < 5)
1210 				goto bad;
1211 			if (ipt->ipt_ptr > ipt->ipt_len - sizeof(int32_t)) {
1212 				if (++ipt->ipt_oflw == 0)
1213 					goto bad;
1214 				break;
1215 			}
1216 			sin = (struct in_addr *)(cp + ipt->ipt_ptr - 1);
1217 			switch (ipt->ipt_flg) {
1218 
1219 			case IPOPT_TS_TSONLY:
1220 				break;
1221 
1222 			case IPOPT_TS_TSANDADDR:
1223 				if (ipt->ipt_ptr - 1 + sizeof(n_time) +
1224 				    sizeof(struct in_addr) > ipt->ipt_len)
1225 					goto bad;
1226 				ipaddr.sin_addr = dst;
1227 				ia = (INA)ifaof_ifpforaddr((SA)&ipaddr,
1228 							    m->m_pkthdr.rcvif);
1229 				if (ia == 0)
1230 					continue;
1231 				(void)memcpy(sin, &IA_SIN(ia)->sin_addr,
1232 				    sizeof(struct in_addr));
1233 				ipt->ipt_ptr += sizeof(struct in_addr);
1234 				break;
1235 
1236 			case IPOPT_TS_PRESPEC:
1237 				if (ipt->ipt_ptr - 1 + sizeof(n_time) +
1238 				    sizeof(struct in_addr) > ipt->ipt_len)
1239 					goto bad;
1240 				(void)memcpy(&ipaddr.sin_addr, sin,
1241 				    sizeof(struct in_addr));
1242 				if (ifa_ifwithaddr((SA)&ipaddr) == 0)
1243 					continue;
1244 				ipt->ipt_ptr += sizeof(struct in_addr);
1245 				break;
1246 
1247 			default:
1248 				goto bad;
1249 			}
1250 			ntime = iptime();
1251 			(void)memcpy(cp + ipt->ipt_ptr - 1, &ntime,
1252 			    sizeof(n_time));
1253 			ipt->ipt_ptr += sizeof(n_time);
1254 		}
1255 	}
1256 	if (forward && ipforwarding) {
1257 		ip_forward(m, 1);
1258 		return (1);
1259 	}
1260 	return (0);
1261 bad:
1262 	ip->ip_len -= IP_VHL_HL(ip->ip_vhl) << 2;   /* XXX icmp_error adds in hdr length */
1263 	icmp_error(m, type, code, 0, 0);
1264 	ipstat.ips_badoptions++;
1265 	return (1);
1266 }
1267 
1268 /*
1269  * Given address of next destination (final or next hop),
1270  * return internet address info of interface to be used to get there.
1271  */
1272 static struct in_ifaddr *
1273 ip_rtaddr(dst)
1274 	 struct in_addr dst;
1275 {
1276 	register struct sockaddr_in *sin;
1277 
1278 	sin = (struct sockaddr_in *) &ipforward_rt.ro_dst;
1279 
1280 	if (ipforward_rt.ro_rt == 0 || dst.s_addr != sin->sin_addr.s_addr) {
1281 		if (ipforward_rt.ro_rt) {
1282 			RTFREE(ipforward_rt.ro_rt);
1283 			ipforward_rt.ro_rt = 0;
1284 		}
1285 		sin->sin_family = AF_INET;
1286 		sin->sin_len = sizeof(*sin);
1287 		sin->sin_addr = dst;
1288 
1289 		rtalloc_ign(&ipforward_rt, RTF_PRCLONING);
1290 	}
1291 	if (ipforward_rt.ro_rt == 0)
1292 		return ((struct in_ifaddr *)0);
1293 	return ((struct in_ifaddr *) ipforward_rt.ro_rt->rt_ifa);
1294 }
1295 
1296 /*
1297  * Save incoming source route for use in replies,
1298  * to be picked up later by ip_srcroute if the receiver is interested.
1299  */
1300 void
1301 save_rte(option, dst)
1302 	u_char *option;
1303 	struct in_addr dst;
1304 {
1305 	unsigned olen;
1306 
1307 	olen = option[IPOPT_OLEN];
1308 #ifdef DIAGNOSTIC
1309 	if (ipprintfs)
1310 		printf("save_rte: olen %d\n", olen);
1311 #endif
1312 	if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst)))
1313 		return;
1314 	bcopy(option, ip_srcrt.srcopt, olen);
1315 	ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1316 	ip_srcrt.dst = dst;
1317 }
1318 
1319 /*
1320  * Retrieve incoming source route for use in replies,
1321  * in the same form used by setsockopt.
1322  * The first hop is placed before the options, will be removed later.
1323  */
1324 struct mbuf *
1325 ip_srcroute()
1326 {
1327 	register struct in_addr *p, *q;
1328 	register struct mbuf *m;
1329 
1330 	if (ip_nhops == 0)
1331 		return ((struct mbuf *)0);
1332 	m = m_get(M_DONTWAIT, MT_HEADER);
1333 	if (m == 0)
1334 		return ((struct mbuf *)0);
1335 
1336 #define OPTSIZ	(sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt))
1337 
1338 	/* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
1339 	m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) +
1340 	    OPTSIZ;
1341 #ifdef DIAGNOSTIC
1342 	if (ipprintfs)
1343 		printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len);
1344 #endif
1345 
1346 	/*
1347 	 * First save first hop for return route
1348 	 */
1349 	p = &ip_srcrt.route[ip_nhops - 1];
1350 	*(mtod(m, struct in_addr *)) = *p--;
1351 #ifdef DIAGNOSTIC
1352 	if (ipprintfs)
1353 		printf(" hops %lx", (u_long)ntohl(mtod(m, struct in_addr *)->s_addr));
1354 #endif
1355 
1356 	/*
1357 	 * Copy option fields and padding (nop) to mbuf.
1358 	 */
1359 	ip_srcrt.nop = IPOPT_NOP;
1360 	ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
1361 	(void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr),
1362 	    &ip_srcrt.nop, OPTSIZ);
1363 	q = (struct in_addr *)(mtod(m, caddr_t) +
1364 	    sizeof(struct in_addr) + OPTSIZ);
1365 #undef OPTSIZ
1366 	/*
1367 	 * Record return path as an IP source route,
1368 	 * reversing the path (pointers are now aligned).
1369 	 */
1370 	while (p >= ip_srcrt.route) {
1371 #ifdef DIAGNOSTIC
1372 		if (ipprintfs)
1373 			printf(" %lx", (u_long)ntohl(q->s_addr));
1374 #endif
1375 		*q++ = *p--;
1376 	}
1377 	/*
1378 	 * Last hop goes to final destination.
1379 	 */
1380 	*q = ip_srcrt.dst;
1381 #ifdef DIAGNOSTIC
1382 	if (ipprintfs)
1383 		printf(" %lx\n", (u_long)ntohl(q->s_addr));
1384 #endif
1385 	return (m);
1386 }
1387 
1388 /*
1389  * Strip out IP options, at higher
1390  * level protocol in the kernel.
1391  * Second argument is buffer to which options
1392  * will be moved, and return value is their length.
1393  * XXX should be deleted; last arg currently ignored.
1394  */
1395 void
1396 ip_stripoptions(m, mopt)
1397 	register struct mbuf *m;
1398 	struct mbuf *mopt;
1399 {
1400 	register int i;
1401 	struct ip *ip = mtod(m, struct ip *);
1402 	register caddr_t opts;
1403 	int olen;
1404 
1405 	olen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip);
1406 	opts = (caddr_t)(ip + 1);
1407 	i = m->m_len - (sizeof (struct ip) + olen);
1408 	bcopy(opts + olen, opts, (unsigned)i);
1409 	m->m_len -= olen;
1410 	if (m->m_flags & M_PKTHDR)
1411 		m->m_pkthdr.len -= olen;
1412 	ip->ip_vhl = IP_MAKE_VHL(IPVERSION, sizeof(struct ip) >> 2);
1413 }
1414 
1415 u_char inetctlerrmap[PRC_NCMDS] = {
1416 	0,		0,		0,		0,
1417 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1418 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1419 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1420 	0,		0,		0,		0,
1421 	ENOPROTOOPT
1422 };
1423 
1424 /*
1425  * Forward a packet.  If some error occurs return the sender
1426  * an icmp packet.  Note we can't always generate a meaningful
1427  * icmp message because icmp doesn't have a large enough repertoire
1428  * of codes and types.
1429  *
1430  * If not forwarding, just drop the packet.  This could be confusing
1431  * if ipforwarding was zero but some routing protocol was advancing
1432  * us as a gateway to somewhere.  However, we must let the routing
1433  * protocol deal with that.
1434  *
1435  * The srcrt parameter indicates whether the packet is being forwarded
1436  * via a source route.
1437  */
1438 static void
1439 ip_forward(m, srcrt)
1440 	struct mbuf *m;
1441 	int srcrt;
1442 {
1443 	register struct ip *ip = mtod(m, struct ip *);
1444 	register struct sockaddr_in *sin;
1445 	register struct rtentry *rt;
1446 	int error, type = 0, code = 0;
1447 	struct mbuf *mcopy;
1448 	n_long dest;
1449 	struct ifnet *destifp;
1450 #ifdef IPSEC
1451 	struct ifnet dummyifp;
1452 #endif
1453 
1454 	dest = 0;
1455 #ifdef DIAGNOSTIC
1456 	if (ipprintfs)
1457 		printf("forward: src %lx dst %lx ttl %x\n",
1458 		    (u_long)ip->ip_src.s_addr, (u_long)ip->ip_dst.s_addr,
1459 		    ip->ip_ttl);
1460 #endif
1461 
1462 
1463 	if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
1464 		ipstat.ips_cantforward++;
1465 		m_freem(m);
1466 		return;
1467 	}
1468 	HTONS(ip->ip_id);
1469 #ifdef IPSTEALTH
1470 	if (!ipstealth) {
1471 #endif
1472 		if (ip->ip_ttl <= IPTTLDEC) {
1473 			icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS,
1474 			    dest, 0);
1475 			return;
1476 		}
1477 		ip->ip_ttl -= IPTTLDEC;
1478 #ifdef IPSTEALTH
1479 	}
1480 #endif
1481 
1482 	sin = (struct sockaddr_in *)&ipforward_rt.ro_dst;
1483 	if ((rt = ipforward_rt.ro_rt) == 0 ||
1484 	    ip->ip_dst.s_addr != sin->sin_addr.s_addr) {
1485 		if (ipforward_rt.ro_rt) {
1486 			RTFREE(ipforward_rt.ro_rt);
1487 			ipforward_rt.ro_rt = 0;
1488 		}
1489 		sin->sin_family = AF_INET;
1490 		sin->sin_len = sizeof(*sin);
1491 		sin->sin_addr = ip->ip_dst;
1492 
1493 		rtalloc_ign(&ipforward_rt, RTF_PRCLONING);
1494 		if (ipforward_rt.ro_rt == 0) {
1495 			icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0);
1496 			return;
1497 		}
1498 		rt = ipforward_rt.ro_rt;
1499 	}
1500 
1501 	/*
1502 	 * Save at most 64 bytes of the packet in case
1503 	 * we need to generate an ICMP message to the src.
1504 	 */
1505 	mcopy = m_copy(m, 0, imin((int)ip->ip_len, 64));
1506 
1507 	/*
1508 	 * If forwarding packet using same interface that it came in on,
1509 	 * perhaps should send a redirect to sender to shortcut a hop.
1510 	 * Only send redirect if source is sending directly to us,
1511 	 * and if packet was not source routed (or has any options).
1512 	 * Also, don't send redirect if forwarding using a default route
1513 	 * or a route modified by a redirect.
1514 	 */
1515 #define	satosin(sa)	((struct sockaddr_in *)(sa))
1516 	if (rt->rt_ifp == m->m_pkthdr.rcvif &&
1517 	    (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1518 	    satosin(rt_key(rt))->sin_addr.s_addr != 0 &&
1519 	    ipsendredirects && !srcrt) {
1520 #define	RTA(rt)	((struct in_ifaddr *)(rt->rt_ifa))
1521 		u_long src = ntohl(ip->ip_src.s_addr);
1522 
1523 		if (RTA(rt) &&
1524 		    (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1525 		    if (rt->rt_flags & RTF_GATEWAY)
1526 			dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
1527 		    else
1528 			dest = ip->ip_dst.s_addr;
1529 		    /* Router requirements says to only send host redirects */
1530 		    type = ICMP_REDIRECT;
1531 		    code = ICMP_REDIRECT_HOST;
1532 #ifdef DIAGNOSTIC
1533 		    if (ipprintfs)
1534 		        printf("redirect (%d) to %lx\n", code, (u_long)dest);
1535 #endif
1536 		}
1537 	}
1538 
1539 	error = ip_output(m, (struct mbuf *)0, &ipforward_rt,
1540 			  IP_FORWARDING, 0);
1541 	if (error)
1542 		ipstat.ips_cantforward++;
1543 	else {
1544 		ipstat.ips_forward++;
1545 		if (type)
1546 			ipstat.ips_redirectsent++;
1547 		else {
1548 			if (mcopy) {
1549 				ipflow_create(&ipforward_rt, mcopy);
1550 				m_freem(mcopy);
1551 			}
1552 			return;
1553 		}
1554 	}
1555 	if (mcopy == NULL)
1556 		return;
1557 	destifp = NULL;
1558 
1559 	switch (error) {
1560 
1561 	case 0:				/* forwarded, but need redirect */
1562 		/* type, code set above */
1563 		break;
1564 
1565 	case ENETUNREACH:		/* shouldn't happen, checked above */
1566 	case EHOSTUNREACH:
1567 	case ENETDOWN:
1568 	case EHOSTDOWN:
1569 	default:
1570 		type = ICMP_UNREACH;
1571 		code = ICMP_UNREACH_HOST;
1572 		break;
1573 
1574 	case EMSGSIZE:
1575 		type = ICMP_UNREACH;
1576 		code = ICMP_UNREACH_NEEDFRAG;
1577 #ifndef IPSEC
1578 		if (ipforward_rt.ro_rt)
1579 			destifp = ipforward_rt.ro_rt->rt_ifp;
1580 #else
1581 		/*
1582 		 * If the packet is routed over IPsec tunnel, tell the
1583 		 * originator the tunnel MTU.
1584 		 *	tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
1585 		 * XXX quickhack!!!
1586 		 */
1587 		if (ipforward_rt.ro_rt) {
1588 			struct secpolicy *sp = NULL;
1589 			int ipsecerror;
1590 			int ipsechdr;
1591 			struct route *ro;
1592 
1593 			sp = ipsec4_getpolicybyaddr(mcopy,
1594 						    IPSEC_DIR_OUTBOUND,
1595 			                            IP_FORWARDING,
1596 			                            &ipsecerror);
1597 
1598 			if (sp == NULL)
1599 				destifp = ipforward_rt.ro_rt->rt_ifp;
1600 			else {
1601 				/* count IPsec header size */
1602 				ipsechdr = ipsec4_hdrsiz(mcopy,
1603 							 IPSEC_DIR_OUTBOUND,
1604 							 NULL);
1605 
1606 				/*
1607 				 * find the correct route for outer IPv4
1608 				 * header, compute tunnel MTU.
1609 				 *
1610 				 * XXX BUG ALERT
1611 				 * The "dummyifp" code relies upon the fact
1612 				 * that icmp_error() touches only ifp->if_mtu.
1613 				 */
1614 				/*XXX*/
1615 				destifp = NULL;
1616 				if (sp->req != NULL
1617 				 && sp->req->sav != NULL
1618 				 && sp->req->sav->sah != NULL) {
1619 					ro = &sp->req->sav->sah->sa_route;
1620 					if (ro->ro_rt && ro->ro_rt->rt_ifp) {
1621 						dummyifp.if_mtu =
1622 						    ro->ro_rt->rt_ifp->if_mtu;
1623 						dummyifp.if_mtu -= ipsechdr;
1624 						destifp = &dummyifp;
1625 					}
1626 				}
1627 
1628 				key_freesp(sp);
1629 			}
1630 		}
1631 #endif /*IPSEC*/
1632 		ipstat.ips_cantfrag++;
1633 		break;
1634 
1635 	case ENOBUFS:
1636 		type = ICMP_SOURCEQUENCH;
1637 		code = 0;
1638 		break;
1639 	}
1640 	icmp_error(mcopy, type, code, dest, destifp);
1641 }
1642 
1643 void
1644 ip_savecontrol(inp, mp, ip, m)
1645 	register struct inpcb *inp;
1646 	register struct mbuf **mp;
1647 	register struct ip *ip;
1648 	register struct mbuf *m;
1649 {
1650 	if (inp->inp_socket->so_options & SO_TIMESTAMP) {
1651 		struct timeval tv;
1652 
1653 		microtime(&tv);
1654 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1655 			SCM_TIMESTAMP, SOL_SOCKET);
1656 		if (*mp)
1657 			mp = &(*mp)->m_next;
1658 	}
1659 	if (inp->inp_flags & INP_RECVDSTADDR) {
1660 		*mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
1661 		    sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1662 		if (*mp)
1663 			mp = &(*mp)->m_next;
1664 	}
1665 #ifdef notyet
1666 	/* XXX
1667 	 * Moving these out of udp_input() made them even more broken
1668 	 * than they already were.
1669 	 */
1670 	/* options were tossed already */
1671 	if (inp->inp_flags & INP_RECVOPTS) {
1672 		*mp = sbcreatecontrol((caddr_t) opts_deleted_above,
1673 		    sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1674 		if (*mp)
1675 			mp = &(*mp)->m_next;
1676 	}
1677 	/* ip_srcroute doesn't do what we want here, need to fix */
1678 	if (inp->inp_flags & INP_RECVRETOPTS) {
1679 		*mp = sbcreatecontrol((caddr_t) ip_srcroute(),
1680 		    sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1681 		if (*mp)
1682 			mp = &(*mp)->m_next;
1683 	}
1684 #endif
1685 	if (inp->inp_flags & INP_RECVIF) {
1686 		struct ifnet *ifp;
1687 		struct sdlbuf {
1688 			struct sockaddr_dl sdl;
1689 			u_char	pad[32];
1690 		} sdlbuf;
1691 		struct sockaddr_dl *sdp;
1692 		struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1693 
1694 		if (((ifp = m->m_pkthdr.rcvif))
1695 		&& ( ifp->if_index && (ifp->if_index <= if_index))) {
1696 			sdp = (struct sockaddr_dl *)(ifnet_addrs
1697 					[ifp->if_index - 1]->ifa_addr);
1698 			/*
1699 			 * Change our mind and don't try copy.
1700 			 */
1701 			if ((sdp->sdl_family != AF_LINK)
1702 			|| (sdp->sdl_len > sizeof(sdlbuf))) {
1703 				goto makedummy;
1704 			}
1705 			bcopy(sdp, sdl2, sdp->sdl_len);
1706 		} else {
1707 makedummy:
1708 			sdl2->sdl_len
1709 				= offsetof(struct sockaddr_dl, sdl_data[0]);
1710 			sdl2->sdl_family = AF_LINK;
1711 			sdl2->sdl_index = 0;
1712 			sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1713 		}
1714 		*mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len,
1715 			IP_RECVIF, IPPROTO_IP);
1716 		if (*mp)
1717 			mp = &(*mp)->m_next;
1718 	}
1719 }
1720 
1721 int
1722 ip_rsvp_init(struct socket *so)
1723 {
1724 	if (so->so_type != SOCK_RAW ||
1725 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
1726 	  return EOPNOTSUPP;
1727 
1728 	if (ip_rsvpd != NULL)
1729 	  return EADDRINUSE;
1730 
1731 	ip_rsvpd = so;
1732 	/*
1733 	 * This may seem silly, but we need to be sure we don't over-increment
1734 	 * the RSVP counter, in case something slips up.
1735 	 */
1736 	if (!ip_rsvp_on) {
1737 		ip_rsvp_on = 1;
1738 		rsvp_on++;
1739 	}
1740 
1741 	return 0;
1742 }
1743 
1744 int
1745 ip_rsvp_done(void)
1746 {
1747 	ip_rsvpd = NULL;
1748 	/*
1749 	 * This may seem silly, but we need to be sure we don't over-decrement
1750 	 * the RSVP counter, in case something slips up.
1751 	 */
1752 	if (ip_rsvp_on) {
1753 		ip_rsvp_on = 0;
1754 		rsvp_on--;
1755 	}
1756 	return 0;
1757 }
1758