xref: /freebsd/sys/netinet/ip_input.c (revision b601c69bdbe8755d26570261d7fd4c02ee4eff74)
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 #include <net/intrq.h>
67 
68 #include <netinet/in.h>
69 #include <netinet/in_systm.h>
70 #include <netinet/in_var.h>
71 #include <netinet/ip.h>
72 #include <netinet/in_pcb.h>
73 #include <netinet/ip_var.h>
74 #include <netinet/ip_icmp.h>
75 #include <machine/in_cksum.h>
76 
77 #include <netinet/ipprotosw.h>
78 
79 #include <sys/socketvar.h>
80 
81 #include <netinet/ip_fw.h>
82 
83 #ifdef IPSEC
84 #include <netinet6/ipsec.h>
85 #include <netkey/key.h>
86 #endif
87 
88 #include "faith.h"
89 #if defined(NFAITH) && NFAITH > 0
90 #include <net/if_types.h>
91 #endif
92 
93 #ifdef DUMMYNET
94 #include <netinet/ip_dummynet.h>
95 #endif
96 
97 int rsvp_on = 0;
98 static int ip_rsvp_on;
99 struct socket *ip_rsvpd;
100 
101 int	ipforwarding = 0;
102 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW,
103     &ipforwarding, 0, "Enable IP forwarding between interfaces");
104 
105 static int	ipsendredirects = 1; /* XXX */
106 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW,
107     &ipsendredirects, 0, "Enable sending IP redirects");
108 
109 int	ip_defttl = IPDEFTTL;
110 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW,
111     &ip_defttl, 0, "Maximum TTL on IP packets");
112 
113 static int	ip_dosourceroute = 0;
114 SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW,
115     &ip_dosourceroute, 0, "Enable forwarding source routed IP packets");
116 
117 static int	ip_acceptsourceroute = 0;
118 SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute,
119     CTLFLAG_RW, &ip_acceptsourceroute, 0,
120     "Enable accepting source routed IP packets");
121 
122 static int	ip_keepfaith = 0;
123 SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW,
124 	&ip_keepfaith,	0,
125 	"Enable packet capture for FAITH IPv4->IPv6 translater daemon");
126 
127 #ifdef DIAGNOSTIC
128 static int	ipprintfs = 0;
129 #endif
130 
131 extern	struct domain inetdomain;
132 extern	struct ipprotosw inetsw[];
133 u_char	ip_protox[IPPROTO_MAX];
134 static int	ipqmaxlen = IFQ_MAXLEN;
135 struct	in_ifaddrhead in_ifaddrhead; /* first inet address */
136 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW,
137     &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue");
138 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD,
139     &ipintrq.ifq_drops, 0, "Number of packets dropped from the IP input queue");
140 
141 struct ipstat ipstat;
142 SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RD,
143     &ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)");
144 
145 /* Packet reassembly stuff */
146 #define IPREASS_NHASH_LOG2      6
147 #define IPREASS_NHASH           (1 << IPREASS_NHASH_LOG2)
148 #define IPREASS_HMASK           (IPREASS_NHASH - 1)
149 #define IPREASS_HASH(x,y) \
150 	(((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK)
151 
152 static struct ipq ipq[IPREASS_NHASH];
153 static int    nipq = 0;         /* total # of reass queues */
154 static int    maxnipq;
155 const  int    ipintrq_present = 1;
156 
157 #ifdef IPCTL_DEFMTU
158 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
159     &ip_mtu, 0, "Default MTU");
160 #endif
161 
162 #ifdef IPSTEALTH
163 static int	ipstealth = 0;
164 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW,
165     &ipstealth, 0, "");
166 #endif
167 
168 
169 /* Firewall hooks */
170 ip_fw_chk_t *ip_fw_chk_ptr;
171 ip_fw_ctl_t *ip_fw_ctl_ptr;
172 int fw_enable = 1 ;
173 
174 #ifdef DUMMYNET
175 ip_dn_ctl_t *ip_dn_ctl_ptr;
176 #endif
177 
178 int (*fr_checkp) __P((struct ip *, int, struct ifnet *, int, struct mbuf **)) = NULL;
179 
180 
181 /*
182  * We need to save the IP options in case a protocol wants to respond
183  * to an incoming packet over the same route if the packet got here
184  * using IP source routing.  This allows connection establishment and
185  * maintenance when the remote end is on a network that is not known
186  * to us.
187  */
188 static int	ip_nhops = 0;
189 static	struct ip_srcrt {
190 	struct	in_addr dst;			/* final destination */
191 	char	nop;				/* one NOP to align */
192 	char	srcopt[IPOPT_OFFSET + 1];	/* OPTVAL, OLEN and OFFSET */
193 	struct	in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)];
194 } ip_srcrt;
195 
196 struct sockaddr_in *ip_fw_fwd_addr;
197 
198 static void	save_rte __P((u_char *, struct in_addr));
199 static int	ip_dooptions __P((struct mbuf *));
200 static void	ip_forward __P((struct mbuf *, int));
201 static void	ip_freef __P((struct ipq *));
202 #ifdef IPDIVERT
203 static struct	mbuf *ip_reass __P((struct mbuf *,
204 			struct ipq *, struct ipq *, u_int32_t *, u_int16_t *));
205 #else
206 static struct	mbuf *ip_reass __P((struct mbuf *, struct ipq *, struct ipq *));
207 #endif
208 static struct	in_ifaddr *ip_rtaddr __P((struct in_addr));
209 static void	ipintr __P((void));
210 
211 /*
212  * IP initialization: fill in IP protocol switch table.
213  * All protocols not implemented in kernel go to raw IP protocol handler.
214  */
215 void
216 ip_init()
217 {
218 	register struct ipprotosw *pr;
219 	register int i;
220 
221 	TAILQ_INIT(&in_ifaddrhead);
222 	pr = (struct ipprotosw *)pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
223 	if (pr == 0)
224 		panic("ip_init");
225 	for (i = 0; i < IPPROTO_MAX; i++)
226 		ip_protox[i] = pr - inetsw;
227 	for (pr = (struct ipprotosw *)inetdomain.dom_protosw;
228 	    pr < (struct ipprotosw *)inetdomain.dom_protoswNPROTOSW; pr++)
229 		if (pr->pr_domain->dom_family == PF_INET &&
230 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
231 			ip_protox[pr->pr_protocol] = pr - inetsw;
232 
233 	for (i = 0; i < IPREASS_NHASH; i++)
234 	    ipq[i].next = ipq[i].prev = &ipq[i];
235 
236 	maxnipq = nmbclusters/4;
237 
238 	ip_id = time_second & 0xffff;
239 	ipintrq.ifq_maxlen = ipqmaxlen;
240 
241 	register_netisr(NETISR_IP, ipintr);
242 }
243 
244 static struct	sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET };
245 static struct	route ipforward_rt;
246 
247 /*
248  * Ip input routine.  Checksum and byte swap header.  If fragmented
249  * try to reassemble.  Process options.  Pass to next level.
250  */
251 void
252 ip_input(struct mbuf *m)
253 {
254 	struct ip *ip;
255 	struct ipq *fp;
256 	struct in_ifaddr *ia;
257 	int    i, hlen, mff;
258 	u_short sum;
259 	u_int16_t divert_cookie;		/* firewall cookie */
260 #ifdef IPDIVERT
261 	u_int32_t divert_info = 0;		/* packet divert/tee info */
262 #endif
263 	struct ip_fw_chain *rule = NULL;
264 
265 #ifdef IPDIVERT
266 	/* Get and reset firewall cookie */
267 	divert_cookie = ip_divert_cookie;
268 	ip_divert_cookie = 0;
269 #else
270 	divert_cookie = 0;
271 #endif
272 
273 #if defined(IPFIREWALL) && defined(DUMMYNET)
274         /*
275          * dummynet packet are prepended a vestigial mbuf with
276          * m_type = MT_DUMMYNET and m_data pointing to the matching
277          * rule.
278          */
279         if (m->m_type == MT_DUMMYNET) {
280             rule = (struct ip_fw_chain *)(m->m_data) ;
281             m = m->m_next ;
282             ip = mtod(m, struct ip *);
283             hlen = IP_VHL_HL(ip->ip_vhl) << 2;
284             goto iphack ;
285         } else
286             rule = NULL ;
287 #endif
288 
289 #ifdef	DIAGNOSTIC
290 	if (m == NULL || (m->m_flags & M_PKTHDR) == 0)
291 		panic("ip_input no HDR");
292 #endif
293 	ipstat.ips_total++;
294 
295 	if (m->m_pkthdr.len < sizeof(struct ip))
296 		goto tooshort;
297 
298 	if (m->m_len < sizeof (struct ip) &&
299 	    (m = m_pullup(m, sizeof (struct ip))) == 0) {
300 		ipstat.ips_toosmall++;
301 		return;
302 	}
303 	ip = mtod(m, struct ip *);
304 
305 	if (IP_VHL_V(ip->ip_vhl) != IPVERSION) {
306 		ipstat.ips_badvers++;
307 		goto bad;
308 	}
309 
310 	hlen = IP_VHL_HL(ip->ip_vhl) << 2;
311 	if (hlen < sizeof(struct ip)) {	/* minimum header length */
312 		ipstat.ips_badhlen++;
313 		goto bad;
314 	}
315 	if (hlen > m->m_len) {
316 		if ((m = m_pullup(m, hlen)) == 0) {
317 			ipstat.ips_badhlen++;
318 			return;
319 		}
320 		ip = mtod(m, struct ip *);
321 	}
322 	if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
323 		sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
324 	} else {
325 		if (hlen == sizeof(struct ip)) {
326 			sum = in_cksum_hdr(ip);
327 		} else {
328 			sum = in_cksum(m, hlen);
329 		}
330 	}
331 	if (sum) {
332 		ipstat.ips_badsum++;
333 		goto bad;
334 	}
335 
336 	/*
337 	 * Convert fields to host representation.
338 	 */
339 	NTOHS(ip->ip_len);
340 	if (ip->ip_len < hlen) {
341 		ipstat.ips_badlen++;
342 		goto bad;
343 	}
344 	NTOHS(ip->ip_id);
345 	NTOHS(ip->ip_off);
346 
347 	/*
348 	 * Check that the amount of data in the buffers
349 	 * is as at least much as the IP header would have us expect.
350 	 * Trim mbufs if longer than we expect.
351 	 * Drop packet if shorter than we expect.
352 	 */
353 	if (m->m_pkthdr.len < ip->ip_len) {
354 tooshort:
355 		ipstat.ips_tooshort++;
356 		goto bad;
357 	}
358 	if (m->m_pkthdr.len > ip->ip_len) {
359 		if (m->m_len == m->m_pkthdr.len) {
360 			m->m_len = ip->ip_len;
361 			m->m_pkthdr.len = ip->ip_len;
362 		} else
363 			m_adj(m, ip->ip_len - m->m_pkthdr.len);
364 	}
365 	/*
366 	 * IpHack's section.
367 	 * Right now when no processing on packet has done
368 	 * and it is still fresh out of network we do our black
369 	 * deals with it.
370 	 * - Firewall: deny/allow/divert
371 	 * - Xlate: translate packet's addr/port (NAT).
372 	 * - Pipe: pass pkt through dummynet.
373 	 * - Wrap: fake packet's addr/port <unimpl.>
374 	 * - Encapsulate: put it in another IP and send out. <unimp.>
375  	 */
376 
377 #if defined(IPFIREWALL) && defined(DUMMYNET)
378 iphack:
379 #endif
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 	if (fw_enable && ip_fw_chk_ptr) {
392 #ifdef IPFIREWALL_FORWARD
393 		/*
394 		 * If we've been forwarded from the output side, then
395 		 * skip the firewall a second time
396 		 */
397 		if (ip_fw_fwd_addr)
398 			goto ours;
399 #endif	/* IPFIREWALL_FORWARD */
400 		/*
401 		 * See the comment in ip_output for the return values
402 		 * produced by the firewall.
403 		 */
404 		i = (*ip_fw_chk_ptr)(&ip,
405 		    hlen, NULL, &divert_cookie, &m, &rule, &ip_fw_fwd_addr);
406 		if (m == NULL)		/* Packet discarded by firewall */
407 			return;
408 		if (i == 0 && ip_fw_fwd_addr == NULL)	/* common case */
409 			goto pass;
410 #ifdef DUMMYNET
411                 if ((i & IP_FW_PORT_DYNT_FLAG) != 0) {
412                         /* Send packet to the appropriate pipe */
413                         dummynet_io(i&0xffff,DN_TO_IP_IN,m,NULL,NULL,0, rule,
414 				    0);
415 			return;
416 		}
417 #endif
418 #ifdef IPDIVERT
419 		if (i != 0 && (i & IP_FW_PORT_DYNT_FLAG) == 0) {
420 			/* Divert or tee packet */
421 			divert_info = i;
422 			goto ours;
423 		}
424 #endif
425 #ifdef IPFIREWALL_FORWARD
426 		if (i == 0 && ip_fw_fwd_addr != NULL)
427 			goto pass;
428 #endif
429 		/*
430 		 * if we get here, the packet must be dropped
431 		 */
432 		m_freem(m);
433 		return;
434 	}
435 pass:
436 
437 	/*
438 	 * Process options and, if not destined for us,
439 	 * ship it on.  ip_dooptions returns 1 when an
440 	 * error was detected (causing an icmp message
441 	 * to be sent and the original packet to be freed).
442 	 */
443 	ip_nhops = 0;		/* for source routed packets */
444 	if (hlen > sizeof (struct ip) && ip_dooptions(m)) {
445 #ifdef IPFIREWALL_FORWARD
446 		ip_fw_fwd_addr = NULL;
447 #endif
448 		return;
449 	}
450 
451         /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
452          * matter if it is destined to another node, or whether it is
453          * a multicast one, RSVP wants it! and prevents it from being forwarded
454          * anywhere else. Also checks if the rsvp daemon is running before
455 	 * grabbing the packet.
456          */
457 	if (rsvp_on && ip->ip_p==IPPROTO_RSVP)
458 		goto ours;
459 
460 	/*
461 	 * Check our list of addresses, to see if the packet is for us.
462 	 * If we don't have any addresses, assume any unicast packet
463 	 * we receive might be for us (and let the upper layers deal
464 	 * with it).
465 	 */
466 	if (TAILQ_EMPTY(&in_ifaddrhead) &&
467 	    (m->m_flags & (M_MCAST|M_BCAST)) == 0)
468 		goto ours;
469 
470 	for (ia = TAILQ_FIRST(&in_ifaddrhead); ia;
471 					ia = TAILQ_NEXT(ia, ia_link)) {
472 #define	satosin(sa)	((struct sockaddr_in *)(sa))
473 
474 #ifdef BOOTP_COMPAT
475 		if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY)
476 			goto ours;
477 #endif
478 #ifdef IPFIREWALL_FORWARD
479 		/*
480 		 * If the addr to forward to is one of ours, we pretend to
481 		 * be the destination for this packet.
482 		 */
483 		if (ip_fw_fwd_addr == NULL) {
484 			if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr)
485 				goto ours;
486 		} else if (IA_SIN(ia)->sin_addr.s_addr ==
487 					 ip_fw_fwd_addr->sin_addr.s_addr)
488 			goto ours;
489 #else
490 		if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr)
491 			goto ours;
492 #endif
493 		if (ia->ia_ifp && ia->ia_ifp->if_flags & IFF_BROADCAST) {
494 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
495 			    ip->ip_dst.s_addr)
496 				goto ours;
497 			if (ip->ip_dst.s_addr == ia->ia_netbroadcast.s_addr)
498 				goto ours;
499 		}
500 	}
501 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
502 		struct in_multi *inm;
503 		if (ip_mrouter) {
504 			/*
505 			 * If we are acting as a multicast router, all
506 			 * incoming multicast packets are passed to the
507 			 * kernel-level multicast forwarding function.
508 			 * The packet is returned (relatively) intact; if
509 			 * ip_mforward() returns a non-zero value, the packet
510 			 * must be discarded, else it may be accepted below.
511 			 *
512 			 * (The IP ident field is put in the same byte order
513 			 * as expected when ip_mforward() is called from
514 			 * ip_output().)
515 			 */
516 			ip->ip_id = htons(ip->ip_id);
517 			if (ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) {
518 				ipstat.ips_cantforward++;
519 				m_freem(m);
520 				return;
521 			}
522 			ip->ip_id = ntohs(ip->ip_id);
523 
524 			/*
525 			 * The process-level routing demon needs to receive
526 			 * all multicast IGMP packets, whether or not this
527 			 * host belongs to their destination groups.
528 			 */
529 			if (ip->ip_p == IPPROTO_IGMP)
530 				goto ours;
531 			ipstat.ips_forward++;
532 		}
533 		/*
534 		 * See if we belong to the destination multicast group on the
535 		 * arrival interface.
536 		 */
537 		IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
538 		if (inm == NULL) {
539 			ipstat.ips_notmember++;
540 			m_freem(m);
541 			return;
542 		}
543 		goto ours;
544 	}
545 	if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
546 		goto ours;
547 	if (ip->ip_dst.s_addr == INADDR_ANY)
548 		goto ours;
549 
550 #if defined(NFAITH) && 0 < NFAITH
551 	/*
552 	 * FAITH(Firewall Aided Internet Translator)
553 	 */
554 	if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
555 		if (ip_keepfaith) {
556 			if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP)
557 				goto ours;
558 		}
559 		m_freem(m);
560 		return;
561 	}
562 #endif
563 	/*
564 	 * Not for us; forward if possible and desirable.
565 	 */
566 	if (ipforwarding == 0) {
567 		ipstat.ips_cantforward++;
568 		m_freem(m);
569 	} else
570 		ip_forward(m, 0);
571 #ifdef IPFIREWALL_FORWARD
572 	ip_fw_fwd_addr = NULL;
573 #endif
574 	return;
575 
576 ours:
577 
578 	/*
579 	 * If offset or IP_MF are set, must reassemble.
580 	 * Otherwise, nothing need be done.
581 	 * (We could look in the reassembly queue to see
582 	 * if the packet was previously fragmented,
583 	 * but it's not worth the time; just let them time out.)
584 	 */
585 	if (ip->ip_off & (IP_MF | IP_OFFMASK | IP_RF)) {
586 
587 #if 0	/*
588 	 * Reassembly should be able to treat a mbuf cluster, for later
589 	 * operation of contiguous protocol headers on the cluster. (KAME)
590 	 */
591 		if (m->m_flags & M_EXT) {		/* XXX */
592 			if ((m = m_pullup(m, hlen)) == 0) {
593 				ipstat.ips_toosmall++;
594 #ifdef IPFIREWALL_FORWARD
595 				ip_fw_fwd_addr = NULL;
596 #endif
597 				return;
598 			}
599 			ip = mtod(m, struct ip *);
600 		}
601 #endif
602 		sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
603 		/*
604 		 * Look for queue of fragments
605 		 * of this datagram.
606 		 */
607 		for (fp = ipq[sum].next; fp != &ipq[sum]; fp = fp->next)
608 			if (ip->ip_id == fp->ipq_id &&
609 			    ip->ip_src.s_addr == fp->ipq_src.s_addr &&
610 			    ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
611 			    ip->ip_p == fp->ipq_p)
612 				goto found;
613 
614 		fp = 0;
615 
616 		/* check if there's a place for the new queue */
617 		if (nipq > maxnipq) {
618 		    /*
619 		     * drop something from the tail of the current queue
620 		     * before proceeding further
621 		     */
622 		    if (ipq[sum].prev == &ipq[sum]) {   /* gak */
623 			for (i = 0; i < IPREASS_NHASH; i++) {
624 			    if (ipq[i].prev != &ipq[i]) {
625 				ip_freef(ipq[i].prev);
626 				break;
627 			    }
628 			}
629 		    } else
630 			ip_freef(ipq[sum].prev);
631 		}
632 found:
633 		/*
634 		 * Adjust ip_len to not reflect header,
635 		 * set ip_mff if more fragments are expected,
636 		 * convert offset of this to bytes.
637 		 */
638 		ip->ip_len -= hlen;
639 		mff = (ip->ip_off & IP_MF) != 0;
640 		if (mff) {
641 		        /*
642 		         * Make sure that fragments have a data length
643 			 * that's a non-zero multiple of 8 bytes.
644 		         */
645 			if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
646 				ipstat.ips_toosmall++; /* XXX */
647 				goto bad;
648 			}
649 			m->m_flags |= M_FRAG;
650 		}
651 		ip->ip_off <<= 3;
652 
653 		/*
654 		 * If datagram marked as having more fragments
655 		 * or if this is not the first fragment,
656 		 * attempt reassembly; if it succeeds, proceed.
657 		 */
658 		if (mff || ip->ip_off) {
659 			ipstat.ips_fragments++;
660 			m->m_pkthdr.header = ip;
661 #ifdef IPDIVERT
662 			m = ip_reass(m,
663 			    fp, &ipq[sum], &divert_info, &divert_cookie);
664 #else
665 			m = ip_reass(m, fp, &ipq[sum]);
666 #endif
667 			if (m == 0) {
668 #ifdef IPFIREWALL_FORWARD
669 				ip_fw_fwd_addr = NULL;
670 #endif
671 				return;
672 			}
673 			/* Get the length of the reassembled packets header */
674 			hlen = IP_VHL_HL(ip->ip_vhl) << 2;
675 			ipstat.ips_reassembled++;
676 			ip = mtod(m, struct ip *);
677 #ifdef IPDIVERT
678 			/* Restore original checksum before diverting packet */
679 			if (divert_info != 0) {
680 				ip->ip_len += hlen;
681 				HTONS(ip->ip_len);
682 				HTONS(ip->ip_off);
683 				HTONS(ip->ip_id);
684 				ip->ip_sum = 0;
685 				ip->ip_sum = in_cksum_hdr(ip);
686 				NTOHS(ip->ip_id);
687 				NTOHS(ip->ip_off);
688 				NTOHS(ip->ip_len);
689 				ip->ip_len -= hlen;
690 			}
691 #endif
692 		} else
693 			if (fp)
694 				ip_freef(fp);
695 	} else
696 		ip->ip_len -= hlen;
697 
698 #ifdef IPDIVERT
699 	/*
700 	 * Divert or tee packet to the divert protocol if required.
701 	 *
702 	 * If divert_info is zero then cookie should be too, so we shouldn't
703 	 * need to clear them here.  Assume divert_packet() does so also.
704 	 */
705 	if (divert_info != 0) {
706 		struct mbuf *clone = NULL;
707 
708 		/* Clone packet if we're doing a 'tee' */
709 		if ((divert_info & IP_FW_PORT_TEE_FLAG) != 0)
710 			clone = m_dup(m, M_DONTWAIT);
711 
712 		/* Restore packet header fields to original values */
713 		ip->ip_len += hlen;
714 		HTONS(ip->ip_len);
715 		HTONS(ip->ip_off);
716 		HTONS(ip->ip_id);
717 
718 		/* Deliver packet to divert input routine */
719 		ip_divert_cookie = divert_cookie;
720 		divert_packet(m, 1, divert_info & 0xffff);
721 		ipstat.ips_delivered++;
722 
723 		/* If 'tee', continue with original packet */
724 		if (clone == NULL)
725 			return;
726 		m = clone;
727 		ip = mtod(m, struct ip *);
728 	}
729 #endif
730 
731 	/*
732 	 * Switch out to protocol's input routine.
733 	 */
734 	ipstat.ips_delivered++;
735     {
736 	int off = hlen, nh = ip->ip_p;
737 
738 	(*inetsw[ip_protox[ip->ip_p]].pr_input)(m, off, nh);
739 #ifdef	IPFIREWALL_FORWARD
740 	ip_fw_fwd_addr = NULL;	/* tcp needed it */
741 #endif
742 	return;
743     }
744 bad:
745 #ifdef	IPFIREWALL_FORWARD
746 	ip_fw_fwd_addr = NULL;
747 #endif
748 	m_freem(m);
749 }
750 
751 /*
752  * IP software interrupt routine - to go away sometime soon
753  */
754 static void
755 ipintr(void)
756 {
757 	int s;
758 	struct mbuf *m;
759 
760 	while(1) {
761 		s = splimp();
762 		IF_DEQUEUE(&ipintrq, m);
763 		splx(s);
764 		if (m == 0)
765 			return;
766 		ip_input(m);
767 	}
768 }
769 
770 /*
771  * Take incoming datagram fragment and try to reassemble it into
772  * whole datagram.  If a chain for reassembly of this datagram already
773  * exists, then it is given as fp; otherwise have to make a chain.
774  *
775  * When IPDIVERT enabled, keep additional state with each packet that
776  * tells us if we need to divert or tee the packet we're building.
777  */
778 
779 static struct mbuf *
780 #ifdef IPDIVERT
781 ip_reass(m, fp, where, divinfo, divcookie)
782 #else
783 ip_reass(m, fp, where)
784 #endif
785 	register struct mbuf *m;
786 	register struct ipq *fp;
787 	struct   ipq    *where;
788 #ifdef IPDIVERT
789 	u_int32_t *divinfo;
790 	u_int16_t *divcookie;
791 #endif
792 {
793 	struct ip *ip = mtod(m, struct ip *);
794 	register struct mbuf *p = 0, *q, *nq;
795 	struct mbuf *t;
796 	int hlen = IP_VHL_HL(ip->ip_vhl) << 2;
797 	int i, next;
798 
799 	/*
800 	 * Presence of header sizes in mbufs
801 	 * would confuse code below.
802 	 */
803 	m->m_data += hlen;
804 	m->m_len -= hlen;
805 
806 	/*
807 	 * If first fragment to arrive, create a reassembly queue.
808 	 */
809 	if (fp == 0) {
810 		if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL)
811 			goto dropfrag;
812 		fp = mtod(t, struct ipq *);
813 		insque(fp, where);
814 		nipq++;
815 		fp->ipq_ttl = IPFRAGTTL;
816 		fp->ipq_p = ip->ip_p;
817 		fp->ipq_id = ip->ip_id;
818 		fp->ipq_src = ip->ip_src;
819 		fp->ipq_dst = ip->ip_dst;
820 		fp->ipq_frags = m;
821 		m->m_nextpkt = NULL;
822 #ifdef IPDIVERT
823 		fp->ipq_div_info = 0;
824 		fp->ipq_div_cookie = 0;
825 #endif
826 		goto inserted;
827 	}
828 
829 #define GETIP(m)	((struct ip*)((m)->m_pkthdr.header))
830 
831 	/*
832 	 * Find a segment which begins after this one does.
833 	 */
834 	for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt)
835 		if (GETIP(q)->ip_off > ip->ip_off)
836 			break;
837 
838 	/*
839 	 * If there is a preceding segment, it may provide some of
840 	 * our data already.  If so, drop the data from the incoming
841 	 * segment.  If it provides all of our data, drop us, otherwise
842 	 * stick new segment in the proper place.
843 	 *
844 	 * If some of the data is dropped from the the preceding
845 	 * segment, then it's checksum is invalidated.
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 			m->m_pkthdr.csum_flags = 0;
854 			ip->ip_off += i;
855 			ip->ip_len -= i;
856 		}
857 		m->m_nextpkt = p->m_nextpkt;
858 		p->m_nextpkt = m;
859 	} else {
860 		m->m_nextpkt = fp->ipq_frags;
861 		fp->ipq_frags = m;
862 	}
863 
864 	/*
865 	 * While we overlap succeeding segments trim them or,
866 	 * if they are completely covered, dequeue them.
867 	 */
868 	for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off;
869 	     q = nq) {
870 		i = (ip->ip_off + ip->ip_len) -
871 		    GETIP(q)->ip_off;
872 		if (i < GETIP(q)->ip_len) {
873 			GETIP(q)->ip_len -= i;
874 			GETIP(q)->ip_off += i;
875 			m_adj(q, i);
876 			q->m_pkthdr.csum_flags = 0;
877 			break;
878 		}
879 		nq = q->m_nextpkt;
880 		m->m_nextpkt = nq;
881 		m_freem(q);
882 	}
883 
884 inserted:
885 
886 #ifdef IPDIVERT
887 	/*
888 	 * Transfer firewall instructions to the fragment structure.
889 	 * Any fragment diverting causes the whole packet to divert.
890 	 */
891 	fp->ipq_div_info = *divinfo;
892 	fp->ipq_div_cookie = *divcookie;
893 	*divinfo = 0;
894 	*divcookie = 0;
895 #endif
896 
897 	/*
898 	 * Check for complete reassembly.
899 	 */
900 	next = 0;
901 	for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
902 		if (GETIP(q)->ip_off != next)
903 			return (0);
904 		next += GETIP(q)->ip_len;
905 	}
906 	/* Make sure the last packet didn't have the IP_MF flag */
907 	if (p->m_flags & M_FRAG)
908 		return (0);
909 
910 	/*
911 	 * Reassembly is complete.  Make sure the packet is a sane size.
912 	 */
913 	q = fp->ipq_frags;
914 	ip = GETIP(q);
915 	if (next + (IP_VHL_HL(ip->ip_vhl) << 2) > IP_MAXPACKET) {
916 		ipstat.ips_toolong++;
917 		ip_freef(fp);
918 		return (0);
919 	}
920 
921 	/*
922 	 * Concatenate fragments.
923 	 */
924 	m = q;
925 	t = m->m_next;
926 	m->m_next = 0;
927 	m_cat(m, t);
928 	nq = q->m_nextpkt;
929 	q->m_nextpkt = 0;
930 	for (q = nq; q != NULL; q = nq) {
931 		nq = q->m_nextpkt;
932 		q->m_nextpkt = NULL;
933 		m_cat(m, q);
934 		m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
935 		m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
936 	}
937 
938 #ifdef IPDIVERT
939 	/*
940 	 * Extract firewall instructions from the fragment structure.
941 	 */
942 	*divinfo = fp->ipq_div_info;
943 	*divcookie = fp->ipq_div_cookie;
944 #endif
945 
946 	/*
947 	 * Create header for new ip packet by
948 	 * modifying header of first packet;
949 	 * dequeue and discard fragment reassembly header.
950 	 * Make header visible.
951 	 */
952 	ip->ip_len = next;
953 	ip->ip_src = fp->ipq_src;
954 	ip->ip_dst = fp->ipq_dst;
955 	remque(fp);
956 	nipq--;
957 	(void) m_free(dtom(fp));
958 	m->m_len += (IP_VHL_HL(ip->ip_vhl) << 2);
959 	m->m_data -= (IP_VHL_HL(ip->ip_vhl) << 2);
960 	/* some debugging cruft by sklower, below, will go away soon */
961 	if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */
962 		register int plen = 0;
963 		for (t = m; t; t = t->m_next)
964 			plen += t->m_len;
965 		m->m_pkthdr.len = plen;
966 	}
967 	return (m);
968 
969 dropfrag:
970 #ifdef IPDIVERT
971 	*divinfo = 0;
972 	*divcookie = 0;
973 #endif
974 	ipstat.ips_fragdropped++;
975 	m_freem(m);
976 	return (0);
977 
978 #undef GETIP
979 }
980 
981 /*
982  * Free a fragment reassembly header and all
983  * associated datagrams.
984  */
985 static void
986 ip_freef(fp)
987 	struct ipq *fp;
988 {
989 	register struct mbuf *q;
990 
991 	while (fp->ipq_frags) {
992 		q = fp->ipq_frags;
993 		fp->ipq_frags = q->m_nextpkt;
994 		m_freem(q);
995 	}
996 	remque(fp);
997 	(void) m_free(dtom(fp));
998 	nipq--;
999 }
1000 
1001 /*
1002  * IP timer processing;
1003  * if a timer expires on a reassembly
1004  * queue, discard it.
1005  */
1006 void
1007 ip_slowtimo()
1008 {
1009 	register struct ipq *fp;
1010 	int s = splnet();
1011 	int i;
1012 
1013 	for (i = 0; i < IPREASS_NHASH; i++) {
1014 		fp = ipq[i].next;
1015 		if (fp == 0)
1016 			continue;
1017 		while (fp != &ipq[i]) {
1018 			--fp->ipq_ttl;
1019 			fp = fp->next;
1020 			if (fp->prev->ipq_ttl == 0) {
1021 				ipstat.ips_fragtimeout++;
1022 				ip_freef(fp->prev);
1023 			}
1024 		}
1025 	}
1026 	ipflow_slowtimo();
1027 	splx(s);
1028 }
1029 
1030 /*
1031  * Drain off all datagram fragments.
1032  */
1033 void
1034 ip_drain()
1035 {
1036 	int     i;
1037 
1038 	for (i = 0; i < IPREASS_NHASH; i++) {
1039 		while (ipq[i].next != &ipq[i]) {
1040 			ipstat.ips_fragdropped++;
1041 			ip_freef(ipq[i].next);
1042 		}
1043 	}
1044 	in_rtqdrain();
1045 }
1046 
1047 /*
1048  * Do option processing on a datagram,
1049  * possibly discarding it if bad options are encountered,
1050  * or forwarding it if source-routed.
1051  * Returns 1 if packet has been forwarded/freed,
1052  * 0 if the packet should be processed further.
1053  */
1054 static int
1055 ip_dooptions(m)
1056 	struct mbuf *m;
1057 {
1058 	register struct ip *ip = mtod(m, struct ip *);
1059 	register u_char *cp;
1060 	register struct ip_timestamp *ipt;
1061 	register struct in_ifaddr *ia;
1062 	int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0;
1063 	struct in_addr *sin, dst;
1064 	n_time ntime;
1065 
1066 	dst = ip->ip_dst;
1067 	cp = (u_char *)(ip + 1);
1068 	cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip);
1069 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1070 		opt = cp[IPOPT_OPTVAL];
1071 		if (opt == IPOPT_EOL)
1072 			break;
1073 		if (opt == IPOPT_NOP)
1074 			optlen = 1;
1075 		else {
1076 			if (cnt < IPOPT_OLEN + sizeof(*cp)) {
1077 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1078 				goto bad;
1079 			}
1080 			optlen = cp[IPOPT_OLEN];
1081 			if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
1082 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1083 				goto bad;
1084 			}
1085 		}
1086 		switch (opt) {
1087 
1088 		default:
1089 			break;
1090 
1091 		/*
1092 		 * Source routing with record.
1093 		 * Find interface with current destination address.
1094 		 * If none on this machine then drop if strictly routed,
1095 		 * or do nothing if loosely routed.
1096 		 * Record interface address and bring up next address
1097 		 * component.  If strictly routed make sure next
1098 		 * address is on directly accessible net.
1099 		 */
1100 		case IPOPT_LSRR:
1101 		case IPOPT_SSRR:
1102 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1103 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1104 				goto bad;
1105 			}
1106 			ipaddr.sin_addr = ip->ip_dst;
1107 			ia = (struct in_ifaddr *)
1108 				ifa_ifwithaddr((struct sockaddr *)&ipaddr);
1109 			if (ia == 0) {
1110 				if (opt == IPOPT_SSRR) {
1111 					type = ICMP_UNREACH;
1112 					code = ICMP_UNREACH_SRCFAIL;
1113 					goto bad;
1114 				}
1115 				if (!ip_dosourceroute)
1116 					goto nosourcerouting;
1117 				/*
1118 				 * Loose routing, and not at next destination
1119 				 * yet; nothing to do except forward.
1120 				 */
1121 				break;
1122 			}
1123 			off--;			/* 0 origin */
1124 			if (off > optlen - (int)sizeof(struct in_addr)) {
1125 				/*
1126 				 * End of source route.  Should be for us.
1127 				 */
1128 				if (!ip_acceptsourceroute)
1129 					goto nosourcerouting;
1130 				save_rte(cp, ip->ip_src);
1131 				break;
1132 			}
1133 
1134 			if (!ip_dosourceroute) {
1135 				if (ipforwarding) {
1136 					char buf[16]; /* aaa.bbb.ccc.ddd\0 */
1137 					/*
1138 					 * Acting as a router, so generate ICMP
1139 					 */
1140 nosourcerouting:
1141 					strcpy(buf, inet_ntoa(ip->ip_dst));
1142 					log(LOG_WARNING,
1143 					    "attempted source route from %s to %s\n",
1144 					    inet_ntoa(ip->ip_src), buf);
1145 					type = ICMP_UNREACH;
1146 					code = ICMP_UNREACH_SRCFAIL;
1147 					goto bad;
1148 				} else {
1149 					/*
1150 					 * Not acting as a router, so silently drop.
1151 					 */
1152 					ipstat.ips_cantforward++;
1153 					m_freem(m);
1154 					return (1);
1155 				}
1156 			}
1157 
1158 			/*
1159 			 * locate outgoing interface
1160 			 */
1161 			(void)memcpy(&ipaddr.sin_addr, cp + off,
1162 			    sizeof(ipaddr.sin_addr));
1163 
1164 			if (opt == IPOPT_SSRR) {
1165 #define	INA	struct in_ifaddr *
1166 #define	SA	struct sockaddr *
1167 			    if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == 0)
1168 				ia = (INA)ifa_ifwithnet((SA)&ipaddr);
1169 			} else
1170 				ia = ip_rtaddr(ipaddr.sin_addr);
1171 			if (ia == 0) {
1172 				type = ICMP_UNREACH;
1173 				code = ICMP_UNREACH_SRCFAIL;
1174 				goto bad;
1175 			}
1176 			ip->ip_dst = ipaddr.sin_addr;
1177 			(void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
1178 			    sizeof(struct in_addr));
1179 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1180 			/*
1181 			 * Let ip_intr's mcast routing check handle mcast pkts
1182 			 */
1183 			forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr));
1184 			break;
1185 
1186 		case IPOPT_RR:
1187 			if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1188 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1189 				goto bad;
1190 			}
1191 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1192 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1193 				goto bad;
1194 			}
1195 			/*
1196 			 * If no space remains, ignore.
1197 			 */
1198 			off--;			/* 0 origin */
1199 			if (off > optlen - (int)sizeof(struct in_addr))
1200 				break;
1201 			(void)memcpy(&ipaddr.sin_addr, &ip->ip_dst,
1202 			    sizeof(ipaddr.sin_addr));
1203 			/*
1204 			 * locate outgoing interface; if we're the destination,
1205 			 * use the incoming interface (should be same).
1206 			 */
1207 			if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == 0 &&
1208 			    (ia = ip_rtaddr(ipaddr.sin_addr)) == 0) {
1209 				type = ICMP_UNREACH;
1210 				code = ICMP_UNREACH_HOST;
1211 				goto bad;
1212 			}
1213 			(void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
1214 			    sizeof(struct in_addr));
1215 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1216 			break;
1217 
1218 		case IPOPT_TS:
1219 			code = cp - (u_char *)ip;
1220 			ipt = (struct ip_timestamp *)cp;
1221 			if (ipt->ipt_len < 5)
1222 				goto bad;
1223 			if (ipt->ipt_ptr >
1224 			    ipt->ipt_len - (int)sizeof(int32_t)) {
1225 				if (++ipt->ipt_oflw == 0)
1226 					goto bad;
1227 				break;
1228 			}
1229 			sin = (struct in_addr *)(cp + ipt->ipt_ptr - 1);
1230 			switch (ipt->ipt_flg) {
1231 
1232 			case IPOPT_TS_TSONLY:
1233 				break;
1234 
1235 			case IPOPT_TS_TSANDADDR:
1236 				if (ipt->ipt_ptr - 1 + sizeof(n_time) +
1237 				    sizeof(struct in_addr) > ipt->ipt_len)
1238 					goto bad;
1239 				ipaddr.sin_addr = dst;
1240 				ia = (INA)ifaof_ifpforaddr((SA)&ipaddr,
1241 							    m->m_pkthdr.rcvif);
1242 				if (ia == 0)
1243 					continue;
1244 				(void)memcpy(sin, &IA_SIN(ia)->sin_addr,
1245 				    sizeof(struct in_addr));
1246 				ipt->ipt_ptr += sizeof(struct in_addr);
1247 				break;
1248 
1249 			case IPOPT_TS_PRESPEC:
1250 				if (ipt->ipt_ptr - 1 + sizeof(n_time) +
1251 				    sizeof(struct in_addr) > ipt->ipt_len)
1252 					goto bad;
1253 				(void)memcpy(&ipaddr.sin_addr, sin,
1254 				    sizeof(struct in_addr));
1255 				if (ifa_ifwithaddr((SA)&ipaddr) == 0)
1256 					continue;
1257 				ipt->ipt_ptr += sizeof(struct in_addr);
1258 				break;
1259 
1260 			default:
1261 				goto bad;
1262 			}
1263 			ntime = iptime();
1264 			(void)memcpy(cp + ipt->ipt_ptr - 1, &ntime,
1265 			    sizeof(n_time));
1266 			ipt->ipt_ptr += sizeof(n_time);
1267 		}
1268 	}
1269 	if (forward && ipforwarding) {
1270 		ip_forward(m, 1);
1271 		return (1);
1272 	}
1273 	return (0);
1274 bad:
1275 	ip->ip_len -= IP_VHL_HL(ip->ip_vhl) << 2;   /* XXX icmp_error adds in hdr length */
1276 	icmp_error(m, type, code, 0, 0);
1277 	ipstat.ips_badoptions++;
1278 	return (1);
1279 }
1280 
1281 /*
1282  * Given address of next destination (final or next hop),
1283  * return internet address info of interface to be used to get there.
1284  */
1285 static struct in_ifaddr *
1286 ip_rtaddr(dst)
1287 	 struct in_addr dst;
1288 {
1289 	register struct sockaddr_in *sin;
1290 
1291 	sin = (struct sockaddr_in *) &ipforward_rt.ro_dst;
1292 
1293 	if (ipforward_rt.ro_rt == 0 || dst.s_addr != sin->sin_addr.s_addr) {
1294 		if (ipforward_rt.ro_rt) {
1295 			RTFREE(ipforward_rt.ro_rt);
1296 			ipforward_rt.ro_rt = 0;
1297 		}
1298 		sin->sin_family = AF_INET;
1299 		sin->sin_len = sizeof(*sin);
1300 		sin->sin_addr = dst;
1301 
1302 		rtalloc_ign(&ipforward_rt, RTF_PRCLONING);
1303 	}
1304 	if (ipforward_rt.ro_rt == 0)
1305 		return ((struct in_ifaddr *)0);
1306 	return ((struct in_ifaddr *) ipforward_rt.ro_rt->rt_ifa);
1307 }
1308 
1309 /*
1310  * Save incoming source route for use in replies,
1311  * to be picked up later by ip_srcroute if the receiver is interested.
1312  */
1313 void
1314 save_rte(option, dst)
1315 	u_char *option;
1316 	struct in_addr dst;
1317 {
1318 	unsigned olen;
1319 
1320 	olen = option[IPOPT_OLEN];
1321 #ifdef DIAGNOSTIC
1322 	if (ipprintfs)
1323 		printf("save_rte: olen %d\n", olen);
1324 #endif
1325 	if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst)))
1326 		return;
1327 	bcopy(option, ip_srcrt.srcopt, olen);
1328 	ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1329 	ip_srcrt.dst = dst;
1330 }
1331 
1332 /*
1333  * Retrieve incoming source route for use in replies,
1334  * in the same form used by setsockopt.
1335  * The first hop is placed before the options, will be removed later.
1336  */
1337 struct mbuf *
1338 ip_srcroute()
1339 {
1340 	register struct in_addr *p, *q;
1341 	register struct mbuf *m;
1342 
1343 	if (ip_nhops == 0)
1344 		return ((struct mbuf *)0);
1345 	m = m_get(M_DONTWAIT, MT_HEADER);
1346 	if (m == 0)
1347 		return ((struct mbuf *)0);
1348 
1349 #define OPTSIZ	(sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt))
1350 
1351 	/* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
1352 	m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) +
1353 	    OPTSIZ;
1354 #ifdef DIAGNOSTIC
1355 	if (ipprintfs)
1356 		printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len);
1357 #endif
1358 
1359 	/*
1360 	 * First save first hop for return route
1361 	 */
1362 	p = &ip_srcrt.route[ip_nhops - 1];
1363 	*(mtod(m, struct in_addr *)) = *p--;
1364 #ifdef DIAGNOSTIC
1365 	if (ipprintfs)
1366 		printf(" hops %lx", (u_long)ntohl(mtod(m, struct in_addr *)->s_addr));
1367 #endif
1368 
1369 	/*
1370 	 * Copy option fields and padding (nop) to mbuf.
1371 	 */
1372 	ip_srcrt.nop = IPOPT_NOP;
1373 	ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
1374 	(void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr),
1375 	    &ip_srcrt.nop, OPTSIZ);
1376 	q = (struct in_addr *)(mtod(m, caddr_t) +
1377 	    sizeof(struct in_addr) + OPTSIZ);
1378 #undef OPTSIZ
1379 	/*
1380 	 * Record return path as an IP source route,
1381 	 * reversing the path (pointers are now aligned).
1382 	 */
1383 	while (p >= ip_srcrt.route) {
1384 #ifdef DIAGNOSTIC
1385 		if (ipprintfs)
1386 			printf(" %lx", (u_long)ntohl(q->s_addr));
1387 #endif
1388 		*q++ = *p--;
1389 	}
1390 	/*
1391 	 * Last hop goes to final destination.
1392 	 */
1393 	*q = ip_srcrt.dst;
1394 #ifdef DIAGNOSTIC
1395 	if (ipprintfs)
1396 		printf(" %lx\n", (u_long)ntohl(q->s_addr));
1397 #endif
1398 	return (m);
1399 }
1400 
1401 /*
1402  * Strip out IP options, at higher
1403  * level protocol in the kernel.
1404  * Second argument is buffer to which options
1405  * will be moved, and return value is their length.
1406  * XXX should be deleted; last arg currently ignored.
1407  */
1408 void
1409 ip_stripoptions(m, mopt)
1410 	register struct mbuf *m;
1411 	struct mbuf *mopt;
1412 {
1413 	register int i;
1414 	struct ip *ip = mtod(m, struct ip *);
1415 	register caddr_t opts;
1416 	int olen;
1417 
1418 	olen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip);
1419 	opts = (caddr_t)(ip + 1);
1420 	i = m->m_len - (sizeof (struct ip) + olen);
1421 	bcopy(opts + olen, opts, (unsigned)i);
1422 	m->m_len -= olen;
1423 	if (m->m_flags & M_PKTHDR)
1424 		m->m_pkthdr.len -= olen;
1425 	ip->ip_vhl = IP_MAKE_VHL(IPVERSION, sizeof(struct ip) >> 2);
1426 }
1427 
1428 u_char inetctlerrmap[PRC_NCMDS] = {
1429 	0,		0,		0,		0,
1430 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1431 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1432 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1433 	0,		0,		0,		0,
1434 	ENOPROTOOPT
1435 };
1436 
1437 /*
1438  * Forward a packet.  If some error occurs return the sender
1439  * an icmp packet.  Note we can't always generate a meaningful
1440  * icmp message because icmp doesn't have a large enough repertoire
1441  * of codes and types.
1442  *
1443  * If not forwarding, just drop the packet.  This could be confusing
1444  * if ipforwarding was zero but some routing protocol was advancing
1445  * us as a gateway to somewhere.  However, we must let the routing
1446  * protocol deal with that.
1447  *
1448  * The srcrt parameter indicates whether the packet is being forwarded
1449  * via a source route.
1450  */
1451 static void
1452 ip_forward(m, srcrt)
1453 	struct mbuf *m;
1454 	int srcrt;
1455 {
1456 	register struct ip *ip = mtod(m, struct ip *);
1457 	register struct sockaddr_in *sin;
1458 	register struct rtentry *rt;
1459 	int error, type = 0, code = 0;
1460 	struct mbuf *mcopy;
1461 	n_long dest;
1462 	struct ifnet *destifp;
1463 #ifdef IPSEC
1464 	struct ifnet dummyifp;
1465 #endif
1466 
1467 	dest = 0;
1468 #ifdef DIAGNOSTIC
1469 	if (ipprintfs)
1470 		printf("forward: src %lx dst %lx ttl %x\n",
1471 		    (u_long)ip->ip_src.s_addr, (u_long)ip->ip_dst.s_addr,
1472 		    ip->ip_ttl);
1473 #endif
1474 
1475 
1476 	if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
1477 		ipstat.ips_cantforward++;
1478 		m_freem(m);
1479 		return;
1480 	}
1481 	HTONS(ip->ip_id);
1482 #ifdef IPSTEALTH
1483 	if (!ipstealth) {
1484 #endif
1485 		if (ip->ip_ttl <= IPTTLDEC) {
1486 			icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS,
1487 			    dest, 0);
1488 			return;
1489 		}
1490 		ip->ip_ttl -= IPTTLDEC;
1491 #ifdef IPSTEALTH
1492 	}
1493 #endif
1494 
1495 	sin = (struct sockaddr_in *)&ipforward_rt.ro_dst;
1496 	if ((rt = ipforward_rt.ro_rt) == 0 ||
1497 	    ip->ip_dst.s_addr != sin->sin_addr.s_addr) {
1498 		if (ipforward_rt.ro_rt) {
1499 			RTFREE(ipforward_rt.ro_rt);
1500 			ipforward_rt.ro_rt = 0;
1501 		}
1502 		sin->sin_family = AF_INET;
1503 		sin->sin_len = sizeof(*sin);
1504 		sin->sin_addr = ip->ip_dst;
1505 
1506 		rtalloc_ign(&ipforward_rt, RTF_PRCLONING);
1507 		if (ipforward_rt.ro_rt == 0) {
1508 			icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0);
1509 			return;
1510 		}
1511 		rt = ipforward_rt.ro_rt;
1512 	}
1513 
1514 	/*
1515 	 * Save at most 64 bytes of the packet in case
1516 	 * we need to generate an ICMP message to the src.
1517 	 */
1518 	mcopy = m_copy(m, 0, imin((int)ip->ip_len, 64));
1519 
1520 	/*
1521 	 * If forwarding packet using same interface that it came in on,
1522 	 * perhaps should send a redirect to sender to shortcut a hop.
1523 	 * Only send redirect if source is sending directly to us,
1524 	 * and if packet was not source routed (or has any options).
1525 	 * Also, don't send redirect if forwarding using a default route
1526 	 * or a route modified by a redirect.
1527 	 */
1528 #define	satosin(sa)	((struct sockaddr_in *)(sa))
1529 	if (rt->rt_ifp == m->m_pkthdr.rcvif &&
1530 	    (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1531 	    satosin(rt_key(rt))->sin_addr.s_addr != 0 &&
1532 	    ipsendredirects && !srcrt) {
1533 #define	RTA(rt)	((struct in_ifaddr *)(rt->rt_ifa))
1534 		u_long src = ntohl(ip->ip_src.s_addr);
1535 
1536 		if (RTA(rt) &&
1537 		    (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1538 		    if (rt->rt_flags & RTF_GATEWAY)
1539 			dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
1540 		    else
1541 			dest = ip->ip_dst.s_addr;
1542 		    /* Router requirements says to only send host redirects */
1543 		    type = ICMP_REDIRECT;
1544 		    code = ICMP_REDIRECT_HOST;
1545 #ifdef DIAGNOSTIC
1546 		    if (ipprintfs)
1547 		        printf("redirect (%d) to %lx\n", code, (u_long)dest);
1548 #endif
1549 		}
1550 	}
1551 
1552 	error = ip_output(m, (struct mbuf *)0, &ipforward_rt,
1553 			  IP_FORWARDING, 0);
1554 	if (error)
1555 		ipstat.ips_cantforward++;
1556 	else {
1557 		ipstat.ips_forward++;
1558 		if (type)
1559 			ipstat.ips_redirectsent++;
1560 		else {
1561 			if (mcopy) {
1562 				ipflow_create(&ipforward_rt, mcopy);
1563 				m_freem(mcopy);
1564 			}
1565 			return;
1566 		}
1567 	}
1568 	if (mcopy == NULL)
1569 		return;
1570 	destifp = NULL;
1571 
1572 	switch (error) {
1573 
1574 	case 0:				/* forwarded, but need redirect */
1575 		/* type, code set above */
1576 		break;
1577 
1578 	case ENETUNREACH:		/* shouldn't happen, checked above */
1579 	case EHOSTUNREACH:
1580 	case ENETDOWN:
1581 	case EHOSTDOWN:
1582 	default:
1583 		type = ICMP_UNREACH;
1584 		code = ICMP_UNREACH_HOST;
1585 		break;
1586 
1587 	case EMSGSIZE:
1588 		type = ICMP_UNREACH;
1589 		code = ICMP_UNREACH_NEEDFRAG;
1590 #ifndef IPSEC
1591 		if (ipforward_rt.ro_rt)
1592 			destifp = ipforward_rt.ro_rt->rt_ifp;
1593 #else
1594 		/*
1595 		 * If the packet is routed over IPsec tunnel, tell the
1596 		 * originator the tunnel MTU.
1597 		 *	tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
1598 		 * XXX quickhack!!!
1599 		 */
1600 		if (ipforward_rt.ro_rt) {
1601 			struct secpolicy *sp = NULL;
1602 			int ipsecerror;
1603 			int ipsechdr;
1604 			struct route *ro;
1605 
1606 			sp = ipsec4_getpolicybyaddr(mcopy,
1607 						    IPSEC_DIR_OUTBOUND,
1608 			                            IP_FORWARDING,
1609 			                            &ipsecerror);
1610 
1611 			if (sp == NULL)
1612 				destifp = ipforward_rt.ro_rt->rt_ifp;
1613 			else {
1614 				/* count IPsec header size */
1615 				ipsechdr = ipsec4_hdrsiz(mcopy,
1616 							 IPSEC_DIR_OUTBOUND,
1617 							 NULL);
1618 
1619 				/*
1620 				 * find the correct route for outer IPv4
1621 				 * header, compute tunnel MTU.
1622 				 *
1623 				 * XXX BUG ALERT
1624 				 * The "dummyifp" code relies upon the fact
1625 				 * that icmp_error() touches only ifp->if_mtu.
1626 				 */
1627 				/*XXX*/
1628 				destifp = NULL;
1629 				if (sp->req != NULL
1630 				 && sp->req->sav != NULL
1631 				 && sp->req->sav->sah != NULL) {
1632 					ro = &sp->req->sav->sah->sa_route;
1633 					if (ro->ro_rt && ro->ro_rt->rt_ifp) {
1634 						dummyifp.if_mtu =
1635 						    ro->ro_rt->rt_ifp->if_mtu;
1636 						dummyifp.if_mtu -= ipsechdr;
1637 						destifp = &dummyifp;
1638 					}
1639 				}
1640 
1641 				key_freesp(sp);
1642 			}
1643 		}
1644 #endif /*IPSEC*/
1645 		ipstat.ips_cantfrag++;
1646 		break;
1647 
1648 	case ENOBUFS:
1649 		type = ICMP_SOURCEQUENCH;
1650 		code = 0;
1651 		break;
1652 
1653 	case EACCES:			/* ipfw denied packet */
1654 		m_freem(mcopy);
1655 		return;
1656 	}
1657 	icmp_error(mcopy, type, code, dest, destifp);
1658 }
1659 
1660 void
1661 ip_savecontrol(inp, mp, ip, m)
1662 	register struct inpcb *inp;
1663 	register struct mbuf **mp;
1664 	register struct ip *ip;
1665 	register struct mbuf *m;
1666 {
1667 	if (inp->inp_socket->so_options & SO_TIMESTAMP) {
1668 		struct timeval tv;
1669 
1670 		microtime(&tv);
1671 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1672 			SCM_TIMESTAMP, SOL_SOCKET);
1673 		if (*mp)
1674 			mp = &(*mp)->m_next;
1675 	}
1676 	if (inp->inp_flags & INP_RECVDSTADDR) {
1677 		*mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
1678 		    sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1679 		if (*mp)
1680 			mp = &(*mp)->m_next;
1681 	}
1682 #ifdef notyet
1683 	/* XXX
1684 	 * Moving these out of udp_input() made them even more broken
1685 	 * than they already were.
1686 	 */
1687 	/* options were tossed already */
1688 	if (inp->inp_flags & INP_RECVOPTS) {
1689 		*mp = sbcreatecontrol((caddr_t) opts_deleted_above,
1690 		    sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1691 		if (*mp)
1692 			mp = &(*mp)->m_next;
1693 	}
1694 	/* ip_srcroute doesn't do what we want here, need to fix */
1695 	if (inp->inp_flags & INP_RECVRETOPTS) {
1696 		*mp = sbcreatecontrol((caddr_t) ip_srcroute(),
1697 		    sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1698 		if (*mp)
1699 			mp = &(*mp)->m_next;
1700 	}
1701 #endif
1702 	if (inp->inp_flags & INP_RECVIF) {
1703 		struct ifnet *ifp;
1704 		struct sdlbuf {
1705 			struct sockaddr_dl sdl;
1706 			u_char	pad[32];
1707 		} sdlbuf;
1708 		struct sockaddr_dl *sdp;
1709 		struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1710 
1711 		if (((ifp = m->m_pkthdr.rcvif))
1712 		&& ( ifp->if_index && (ifp->if_index <= if_index))) {
1713 			sdp = (struct sockaddr_dl *)(ifnet_addrs
1714 					[ifp->if_index - 1]->ifa_addr);
1715 			/*
1716 			 * Change our mind and don't try copy.
1717 			 */
1718 			if ((sdp->sdl_family != AF_LINK)
1719 			|| (sdp->sdl_len > sizeof(sdlbuf))) {
1720 				goto makedummy;
1721 			}
1722 			bcopy(sdp, sdl2, sdp->sdl_len);
1723 		} else {
1724 makedummy:
1725 			sdl2->sdl_len
1726 				= offsetof(struct sockaddr_dl, sdl_data[0]);
1727 			sdl2->sdl_family = AF_LINK;
1728 			sdl2->sdl_index = 0;
1729 			sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1730 		}
1731 		*mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len,
1732 			IP_RECVIF, IPPROTO_IP);
1733 		if (*mp)
1734 			mp = &(*mp)->m_next;
1735 	}
1736 }
1737 
1738 int
1739 ip_rsvp_init(struct socket *so)
1740 {
1741 	if (so->so_type != SOCK_RAW ||
1742 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
1743 	  return EOPNOTSUPP;
1744 
1745 	if (ip_rsvpd != NULL)
1746 	  return EADDRINUSE;
1747 
1748 	ip_rsvpd = so;
1749 	/*
1750 	 * This may seem silly, but we need to be sure we don't over-increment
1751 	 * the RSVP counter, in case something slips up.
1752 	 */
1753 	if (!ip_rsvp_on) {
1754 		ip_rsvp_on = 1;
1755 		rsvp_on++;
1756 	}
1757 
1758 	return 0;
1759 }
1760 
1761 int
1762 ip_rsvp_done(void)
1763 {
1764 	ip_rsvpd = NULL;
1765 	/*
1766 	 * This may seem silly, but we need to be sure we don't over-decrement
1767 	 * the RSVP counter, in case something slips up.
1768 	 */
1769 	if (ip_rsvp_on) {
1770 		ip_rsvp_on = 0;
1771 		rsvp_on--;
1772 	}
1773 	return 0;
1774 }
1775