xref: /freebsd/sys/netinet/ip_input.c (revision 0b381bf1fd8fbb2df974c318d58643ecfeec44b0)
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 #include "opt_pfil_hooks.h"
47 #include "opt_random_ip_id.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/pfil.h>
62 #include <net/if.h>
63 #include <net/if_var.h>
64 #include <net/if_dl.h>
65 #include <net/route.h>
66 #include <net/netisr.h>
67 #include <net/intrq.h>
68 
69 #include <netinet/in.h>
70 #include <netinet/in_systm.h>
71 #include <netinet/in_var.h>
72 #include <netinet/ip.h>
73 #include <netinet/in_pcb.h>
74 #include <netinet/ip_var.h>
75 #include <netinet/ip_icmp.h>
76 #include <machine/in_cksum.h>
77 
78 #include <netinet/ipprotosw.h>
79 
80 #include <sys/socketvar.h>
81 
82 #include <netinet/ip_fw.h>
83 
84 #ifdef IPSEC
85 #include <netinet6/ipsec.h>
86 #include <netkey/key.h>
87 #endif
88 
89 #include "faith.h"
90 #if defined(NFAITH) && NFAITH > 0
91 #include <net/if_types.h>
92 #endif
93 
94 #ifdef DUMMYNET
95 #include <netinet/ip_dummynet.h>
96 #endif
97 
98 int rsvp_on = 0;
99 static int ip_rsvp_on;
100 struct socket *ip_rsvpd;
101 
102 int	ipforwarding = 0;
103 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW,
104     &ipforwarding, 0, "Enable IP forwarding between interfaces");
105 
106 static int	ipsendredirects = 1; /* XXX */
107 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW,
108     &ipsendredirects, 0, "Enable sending IP redirects");
109 
110 int	ip_defttl = IPDEFTTL;
111 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW,
112     &ip_defttl, 0, "Maximum TTL on IP packets");
113 
114 static int	ip_dosourceroute = 0;
115 SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW,
116     &ip_dosourceroute, 0, "Enable forwarding source routed IP packets");
117 
118 static int	ip_acceptsourceroute = 0;
119 SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute,
120     CTLFLAG_RW, &ip_acceptsourceroute, 0,
121     "Enable accepting source routed IP packets");
122 
123 static int	ip_keepfaith = 0;
124 SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW,
125 	&ip_keepfaith,	0,
126 	"Enable packet capture for FAITH IPv4->IPv6 translater daemon");
127 
128 /*
129  * XXX - Setting ip_checkinterface mostly implements the receive side of
130  * the Strong ES model described in RFC 1122, but since the routing table
131  * and transmit implementation do not implement the Strong ES model,
132  * setting this to 1 results in an odd hybrid.
133  *
134  * XXX - ip_checkinterface currently must be disabled if you use ipnat
135  * to translate the destination address to another local interface.
136  *
137  * XXX - ip_checkinterface must be disabled if you add IP aliases
138  * to the loopback interface instead of the interface where the
139  * packets for those addresses are received.
140  */
141 static int	ip_checkinterface = 1;
142 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW,
143     &ip_checkinterface, 0, "Verify packet arrives on correct interface");
144 
145 #ifdef DIAGNOSTIC
146 static int	ipprintfs = 0;
147 #endif
148 
149 extern	struct domain inetdomain;
150 extern	struct ipprotosw inetsw[];
151 u_char	ip_protox[IPPROTO_MAX];
152 static int	ipqmaxlen = IFQ_MAXLEN;
153 struct	in_ifaddrhead in_ifaddrhead; /* first inet address */
154 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW,
155     &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue");
156 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD,
157     &ipintrq.ifq_drops, 0, "Number of packets dropped from the IP input queue");
158 
159 struct ipstat ipstat;
160 SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RD,
161     &ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)");
162 
163 /* Packet reassembly stuff */
164 #define IPREASS_NHASH_LOG2      6
165 #define IPREASS_NHASH           (1 << IPREASS_NHASH_LOG2)
166 #define IPREASS_HMASK           (IPREASS_NHASH - 1)
167 #define IPREASS_HASH(x,y) \
168 	(((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK)
169 
170 static TAILQ_HEAD(ipqhead, ipq) ipq[IPREASS_NHASH];
171 static int    nipq = 0;         /* total # of reass queues */
172 static int    maxnipq;
173 const  int    ipintrq_present = 1;
174 
175 #ifdef IPCTL_DEFMTU
176 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
177     &ip_mtu, 0, "Default MTU");
178 #endif
179 
180 #ifdef IPSTEALTH
181 static int	ipstealth = 0;
182 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW,
183     &ipstealth, 0, "");
184 #endif
185 
186 
187 /* Firewall hooks */
188 ip_fw_chk_t *ip_fw_chk_ptr;
189 ip_fw_ctl_t *ip_fw_ctl_ptr;
190 int fw_enable = 1 ;
191 
192 #ifdef DUMMYNET
193 ip_dn_ctl_t *ip_dn_ctl_ptr;
194 #endif
195 
196 
197 /*
198  * We need to save the IP options in case a protocol wants to respond
199  * to an incoming packet over the same route if the packet got here
200  * using IP source routing.  This allows connection establishment and
201  * maintenance when the remote end is on a network that is not known
202  * to us.
203  */
204 static int	ip_nhops = 0;
205 static	struct ip_srcrt {
206 	struct	in_addr dst;			/* final destination */
207 	char	nop;				/* one NOP to align */
208 	char	srcopt[IPOPT_OFFSET + 1];	/* OPTVAL, OLEN and OFFSET */
209 	struct	in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)];
210 } ip_srcrt;
211 
212 struct sockaddr_in *ip_fw_fwd_addr;
213 
214 static void	save_rte __P((u_char *, struct in_addr));
215 static int	ip_dooptions __P((struct mbuf *));
216 static void	ip_forward __P((struct mbuf *, int));
217 static void	ip_freef __P((struct ipqhead *, struct ipq *));
218 #ifdef IPDIVERT
219 static struct	mbuf *ip_reass __P((struct mbuf *, struct ipqhead *, struct ipq *, u_int32_t *, u_int16_t *));
220 #else
221 static struct	mbuf *ip_reass __P((struct mbuf *, struct ipqhead *, struct ipq *));
222 #endif
223 static struct	in_ifaddr *ip_rtaddr __P((struct in_addr));
224 static void	ipintr __P((void));
225 
226 /*
227  * IP initialization: fill in IP protocol switch table.
228  * All protocols not implemented in kernel go to raw IP protocol handler.
229  */
230 void
231 ip_init()
232 {
233 	register struct ipprotosw *pr;
234 	register int i;
235 
236 	TAILQ_INIT(&in_ifaddrhead);
237 	pr = (struct ipprotosw *)pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
238 	if (pr == 0)
239 		panic("ip_init");
240 	for (i = 0; i < IPPROTO_MAX; i++)
241 		ip_protox[i] = pr - inetsw;
242 	for (pr = (struct ipprotosw *)inetdomain.dom_protosw;
243 	    pr < (struct ipprotosw *)inetdomain.dom_protoswNPROTOSW; pr++)
244 		if (pr->pr_domain->dom_family == PF_INET &&
245 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
246 			ip_protox[pr->pr_protocol] = pr - inetsw;
247 
248 	for (i = 0; i < IPREASS_NHASH; i++)
249 	    TAILQ_INIT(&ipq[i]);
250 
251 	maxnipq = nmbclusters/4;
252 
253 #ifndef RANDOM_IP_ID
254 	ip_id = time_second & 0xffff;
255 #endif
256 	ipintrq.ifq_maxlen = ipqmaxlen;
257 	mtx_init(&ipintrq.ifq_mtx, "ip_inq", MTX_DEF);
258 
259 	register_netisr(NETISR_IP, ipintr);
260 }
261 
262 static struct	sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET };
263 struct	route ipforward_rt;
264 
265 /*
266  * Ip input routine.  Checksum and byte swap header.  If fragmented
267  * try to reassemble.  Process options.  Pass to next level.
268  */
269 void
270 ip_input(struct mbuf *m)
271 {
272 	struct ip *ip;
273 	struct ipq *fp;
274 	struct in_ifaddr *ia = NULL;
275 	int    i, hlen, checkif;
276 	u_short sum;
277 	u_int16_t divert_cookie;		/* firewall cookie */
278 	struct in_addr pkt_dst;
279 #ifdef IPDIVERT
280 	u_int32_t divert_info = 0;		/* packet divert/tee info */
281 #endif
282 	struct ip_fw_chain *rule = NULL;
283 #ifdef PFIL_HOOKS
284 	struct packet_filter_hook *pfh;
285 	struct mbuf *m0;
286 	int rv;
287 #endif /* PFIL_HOOKS */
288 
289 #ifdef IPDIVERT
290 	/* Get and reset firewall cookie */
291 	divert_cookie = ip_divert_cookie;
292 	ip_divert_cookie = 0;
293 #else
294 	divert_cookie = 0;
295 #endif
296 
297 #if defined(IPFIREWALL) && defined(DUMMYNET)
298         /*
299          * dummynet packet are prepended a vestigial mbuf with
300          * m_type = MT_DUMMYNET and m_data pointing to the matching
301          * rule.
302          */
303         if (m->m_type == MT_DUMMYNET) {
304             rule = (struct ip_fw_chain *)(m->m_data) ;
305             m = m->m_next ;
306             ip = mtod(m, struct ip *);
307             hlen = IP_VHL_HL(ip->ip_vhl) << 2;
308             goto iphack ;
309         } else
310             rule = NULL ;
311 #endif
312 
313 #ifdef	DIAGNOSTIC
314 	if (m == NULL || (m->m_flags & M_PKTHDR) == 0)
315 		panic("ip_input no HDR");
316 #endif
317 	ipstat.ips_total++;
318 
319 	if (m->m_pkthdr.len < sizeof(struct ip))
320 		goto tooshort;
321 
322 	if (m->m_len < sizeof (struct ip) &&
323 	    (m = m_pullup(m, sizeof (struct ip))) == 0) {
324 		ipstat.ips_toosmall++;
325 		return;
326 	}
327 	ip = mtod(m, struct ip *);
328 
329 	if (IP_VHL_V(ip->ip_vhl) != IPVERSION) {
330 		ipstat.ips_badvers++;
331 		goto bad;
332 	}
333 
334 	hlen = IP_VHL_HL(ip->ip_vhl) << 2;
335 	if (hlen < sizeof(struct ip)) {	/* minimum header length */
336 		ipstat.ips_badhlen++;
337 		goto bad;
338 	}
339 	if (hlen > m->m_len) {
340 		if ((m = m_pullup(m, hlen)) == 0) {
341 			ipstat.ips_badhlen++;
342 			return;
343 		}
344 		ip = mtod(m, struct ip *);
345 	}
346 	if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
347 		sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
348 	} else {
349 		if (hlen == sizeof(struct ip)) {
350 			sum = in_cksum_hdr(ip);
351 		} else {
352 			sum = in_cksum(m, hlen);
353 		}
354 	}
355 	if (sum) {
356 		ipstat.ips_badsum++;
357 		goto bad;
358 	}
359 
360 	/*
361 	 * Convert fields to host representation.
362 	 */
363 	NTOHS(ip->ip_len);
364 	if (ip->ip_len < hlen) {
365 		ipstat.ips_badlen++;
366 		goto bad;
367 	}
368 	NTOHS(ip->ip_off);
369 
370 	/*
371 	 * Check that the amount of data in the buffers
372 	 * is as at least much as the IP header would have us expect.
373 	 * Trim mbufs if longer than we expect.
374 	 * Drop packet if shorter than we expect.
375 	 */
376 	if (m->m_pkthdr.len < ip->ip_len) {
377 tooshort:
378 		ipstat.ips_tooshort++;
379 		goto bad;
380 	}
381 	if (m->m_pkthdr.len > ip->ip_len) {
382 		if (m->m_len == m->m_pkthdr.len) {
383 			m->m_len = ip->ip_len;
384 			m->m_pkthdr.len = ip->ip_len;
385 		} else
386 			m_adj(m, ip->ip_len - m->m_pkthdr.len);
387 	}
388 
389 	/*
390 	 * Don't accept packets with a loopback destination address
391 	 * unless they arrived via the loopback interface.
392 	 */
393 	if ((ntohl(ip->ip_dst.s_addr) & IN_CLASSA_NET) ==
394 	    (IN_LOOPBACKNET << IN_CLASSA_NSHIFT) &&
395 	    (m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
396 		goto bad;
397 	}
398 
399 	/*
400 	 * IpHack's section.
401 	 * Right now when no processing on packet has done
402 	 * and it is still fresh out of network we do our black
403 	 * deals with it.
404 	 * - Firewall: deny/allow/divert
405 	 * - Xlate: translate packet's addr/port (NAT).
406 	 * - Pipe: pass pkt through dummynet.
407 	 * - Wrap: fake packet's addr/port <unimpl.>
408 	 * - Encapsulate: put it in another IP and send out. <unimp.>
409  	 */
410 
411 #if defined(IPFIREWALL) && defined(DUMMYNET)
412 iphack:
413 #endif
414 
415 #ifdef PFIL_HOOKS
416 	/*
417 	 * Run through list of hooks for input packets.  If there are any
418 	 * filters which require that additional packets in the flow are
419 	 * not fast-forwarded, they must clear the M_CANFASTFWD flag.
420 	 * Note that filters must _never_ set this flag, as another filter
421 	 * in the list may have previously cleared it.
422 	 */
423 	m0 = m;
424 	pfh = pfil_hook_get(PFIL_IN, &inetsw[ip_protox[IPPROTO_IP]].pr_pfh);
425 	for (; pfh; pfh = TAILQ_NEXT(pfh, pfil_link))
426 		if (pfh->pfil_func) {
427 			rv = pfh->pfil_func(ip, hlen,
428 					    m->m_pkthdr.rcvif, 0, &m0);
429 			if (rv)
430 				return;
431 			m = m0;
432 			if (m == NULL)
433 				return;
434 			ip = mtod(m, struct ip *);
435 		}
436 #endif /* PFIL_HOOKS */
437 
438 	if (fw_enable && ip_fw_chk_ptr) {
439 #ifdef IPFIREWALL_FORWARD
440 		/*
441 		 * If we've been forwarded from the output side, then
442 		 * skip the firewall a second time
443 		 */
444 		if (ip_fw_fwd_addr)
445 			goto ours;
446 #endif	/* IPFIREWALL_FORWARD */
447 		/*
448 		 * See the comment in ip_output for the return values
449 		 * produced by the firewall.
450 		 */
451 		i = (*ip_fw_chk_ptr)(&ip,
452 		    hlen, NULL, &divert_cookie, &m, &rule, &ip_fw_fwd_addr);
453 		if (i & IP_FW_PORT_DENY_FLAG) { /* XXX new interface-denied */
454 		    if (m)
455 			m_freem(m);
456 		    return ;
457 		}
458 		if (m == NULL) {	/* Packet discarded by firewall */
459 		    static int __debug=10;
460 		    if (__debug >0) {
461 			printf("firewall returns NULL, please update!\n");
462 			__debug-- ;
463 		    }
464 		    return;
465 		}
466 		if (i == 0 && ip_fw_fwd_addr == NULL)	/* common case */
467 			goto pass;
468 #ifdef DUMMYNET
469                 if ((i & IP_FW_PORT_DYNT_FLAG) != 0) {
470                         /* Send packet to the appropriate pipe */
471                         dummynet_io(i&0xffff,DN_TO_IP_IN,m,NULL,NULL,0, rule,
472 				    0);
473 			return;
474 		}
475 #endif
476 #ifdef IPDIVERT
477 		if (i != 0 && (i & IP_FW_PORT_DYNT_FLAG) == 0) {
478 			/* Divert or tee packet */
479 			divert_info = i;
480 			goto ours;
481 		}
482 #endif
483 #ifdef IPFIREWALL_FORWARD
484 		if (i == 0 && ip_fw_fwd_addr != NULL)
485 			goto pass;
486 #endif
487 		/*
488 		 * if we get here, the packet must be dropped
489 		 */
490 		m_freem(m);
491 		return;
492 	}
493 pass:
494 
495 	/*
496 	 * Process options and, if not destined for us,
497 	 * ship it on.  ip_dooptions returns 1 when an
498 	 * error was detected (causing an icmp message
499 	 * to be sent and the original packet to be freed).
500 	 */
501 	ip_nhops = 0;		/* for source routed packets */
502 	if (hlen > sizeof (struct ip) && ip_dooptions(m)) {
503 #ifdef IPFIREWALL_FORWARD
504 		ip_fw_fwd_addr = NULL;
505 #endif
506 		return;
507 	}
508 
509         /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
510          * matter if it is destined to another node, or whether it is
511          * a multicast one, RSVP wants it! and prevents it from being forwarded
512          * anywhere else. Also checks if the rsvp daemon is running before
513 	 * grabbing the packet.
514          */
515 	if (rsvp_on && ip->ip_p==IPPROTO_RSVP)
516 		goto ours;
517 
518 	/*
519 	 * Check our list of addresses, to see if the packet is for us.
520 	 * If we don't have any addresses, assume any unicast packet
521 	 * we receive might be for us (and let the upper layers deal
522 	 * with it).
523 	 */
524 	if (TAILQ_EMPTY(&in_ifaddrhead) &&
525 	    (m->m_flags & (M_MCAST|M_BCAST)) == 0)
526 		goto ours;
527 
528 	/*
529 	 * Cache the destination address of the packet; this may be
530 	 * changed by use of 'ipfw fwd'.
531 	 */
532 	pkt_dst = ip_fw_fwd_addr == NULL ?
533 	    ip->ip_dst : ip_fw_fwd_addr->sin_addr;
534 
535 	/*
536 	 * Enable a consistency check between the destination address
537 	 * and the arrival interface for a unicast packet (the RFC 1122
538 	 * strong ES model) if IP forwarding is disabled and the packet
539 	 * is not locally generated and the packet is not subject to
540 	 * 'ipfw fwd'.
541 	 *
542          * XXX - Checking also should be disabled if the destination
543 	 * address is ipnat'ed to a different interface.
544 	 *
545 	 * XXX - Checking is incompatible with IP aliases added
546 	 * to the loopback interface instead of the interface where
547 	 * the packets are received.
548 	 */
549 	checkif = ip_checkinterface && (ipforwarding == 0) &&
550 	    ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) &&
551 	    (ip_fw_fwd_addr == NULL);
552 
553 	TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link) {
554 #define	satosin(sa)	((struct sockaddr_in *)(sa))
555 
556 #ifdef BOOTP_COMPAT
557 		if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY)
558 			goto ours;
559 #endif
560 		/*
561 		 * If the address matches, verify that the packet
562 		 * arrived via the correct interface if checking is
563 		 * enabled.
564 		 */
565 		if (IA_SIN(ia)->sin_addr.s_addr == pkt_dst.s_addr &&
566 		    (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif))
567 			goto ours;
568 		/*
569 		 * Only accept broadcast packets that arrive via the
570 		 * matching interface.  Reception of forwarded directed
571 		 * broadcasts would be handled via ip_forward() and
572 		 * ether_output() with the loopback into the stack for
573 		 * SIMPLEX interfaces handled by ether_output().
574 		 */
575 		if (ia->ia_ifp == m->m_pkthdr.rcvif &&
576 		    ia->ia_ifp && ia->ia_ifp->if_flags & IFF_BROADCAST) {
577 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
578 			    pkt_dst.s_addr)
579 				goto ours;
580 			if (ia->ia_netbroadcast.s_addr == pkt_dst.s_addr)
581 				goto ours;
582 		}
583 	}
584 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
585 		struct in_multi *inm;
586 		if (ip_mrouter) {
587 			/*
588 			 * If we are acting as a multicast router, all
589 			 * incoming multicast packets are passed to the
590 			 * kernel-level multicast forwarding function.
591 			 * The packet is returned (relatively) intact; if
592 			 * ip_mforward() returns a non-zero value, the packet
593 			 * must be discarded, else it may be accepted below.
594 			 */
595 			if (ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) {
596 				ipstat.ips_cantforward++;
597 				m_freem(m);
598 				return;
599 			}
600 
601 			/*
602 			 * The process-level routing demon needs to receive
603 			 * all multicast IGMP packets, whether or not this
604 			 * host belongs to their destination groups.
605 			 */
606 			if (ip->ip_p == IPPROTO_IGMP)
607 				goto ours;
608 			ipstat.ips_forward++;
609 		}
610 		/*
611 		 * See if we belong to the destination multicast group on the
612 		 * arrival interface.
613 		 */
614 		IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
615 		if (inm == NULL) {
616 			ipstat.ips_notmember++;
617 			m_freem(m);
618 			return;
619 		}
620 		goto ours;
621 	}
622 	if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
623 		goto ours;
624 	if (ip->ip_dst.s_addr == INADDR_ANY)
625 		goto ours;
626 
627 #if defined(NFAITH) && 0 < NFAITH
628 	/*
629 	 * FAITH(Firewall Aided Internet Translator)
630 	 */
631 	if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
632 		if (ip_keepfaith) {
633 			if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP)
634 				goto ours;
635 		}
636 		m_freem(m);
637 		return;
638 	}
639 #endif
640 	/*
641 	 * Not for us; forward if possible and desirable.
642 	 */
643 	if (ipforwarding == 0) {
644 		ipstat.ips_cantforward++;
645 		m_freem(m);
646 	} else
647 		ip_forward(m, 0);
648 #ifdef IPFIREWALL_FORWARD
649 	ip_fw_fwd_addr = NULL;
650 #endif
651 	return;
652 
653 ours:
654 	/* Count the packet in the ip address stats */
655 	if (ia != NULL) {
656 		ia->ia_ifa.if_ipackets++;
657 		ia->ia_ifa.if_ibytes += m->m_pkthdr.len;
658 	}
659 
660 	/*
661 	 * If offset or IP_MF are set, must reassemble.
662 	 * Otherwise, nothing need be done.
663 	 * (We could look in the reassembly queue to see
664 	 * if the packet was previously fragmented,
665 	 * but it's not worth the time; just let them time out.)
666 	 */
667 	if (ip->ip_off & (IP_MF | IP_OFFMASK)) {
668 
669 		sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
670 		/*
671 		 * Look for queue of fragments
672 		 * of this datagram.
673 		 */
674 		TAILQ_FOREACH(fp, &ipq[sum], ipq_list)
675 			if (ip->ip_id == fp->ipq_id &&
676 			    ip->ip_src.s_addr == fp->ipq_src.s_addr &&
677 			    ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
678 			    ip->ip_p == fp->ipq_p)
679 				goto found;
680 
681 		fp = 0;
682 
683 		/* check if there's a place for the new queue */
684 		if (nipq > maxnipq) {
685 		    /*
686 		     * drop something from the tail of the current queue
687 		     * before proceeding further
688 		     */
689 		    struct ipq *q = TAILQ_LAST(&ipq[sum], ipqhead);
690 		    if (q == NULL) {   /* gak */
691 			for (i = 0; i < IPREASS_NHASH; i++) {
692 			    struct ipq *r = TAILQ_LAST(&ipq[i], ipqhead);
693 			    if (r) {
694 				ip_freef(&ipq[i], r);
695 				break;
696 			    }
697 			}
698 		    } else
699 			ip_freef(&ipq[sum], q);
700 		}
701 found:
702 		/*
703 		 * Adjust ip_len to not reflect header,
704 		 * convert offset of this to bytes.
705 		 */
706 		ip->ip_len -= hlen;
707 		if (ip->ip_off & IP_MF) {
708 		        /*
709 		         * Make sure that fragments have a data length
710 			 * that's a non-zero multiple of 8 bytes.
711 		         */
712 			if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
713 				ipstat.ips_toosmall++; /* XXX */
714 				goto bad;
715 			}
716 			m->m_flags |= M_FRAG;
717 		}
718 		ip->ip_off <<= 3;
719 
720 		/*
721 		 * Attempt reassembly; if it succeeds, proceed.
722 		 */
723 		ipstat.ips_fragments++;
724 		m->m_pkthdr.header = ip;
725 #ifdef IPDIVERT
726 		m = ip_reass(m,
727 		    &ipq[sum], fp, &divert_info, &divert_cookie);
728 #else
729 		m = ip_reass(m, &ipq[sum], fp);
730 #endif
731 		if (m == 0) {
732 #ifdef IPFIREWALL_FORWARD
733 			ip_fw_fwd_addr = NULL;
734 #endif
735 			return;
736 		}
737 		ipstat.ips_reassembled++;
738 		ip = mtod(m, struct ip *);
739 		/* Get the header length of the reassembled packet */
740 		hlen = IP_VHL_HL(ip->ip_vhl) << 2;
741 #ifdef IPDIVERT
742 		/* Restore original checksum before diverting packet */
743 		if (divert_info != 0) {
744 			ip->ip_len += hlen;
745 			HTONS(ip->ip_len);
746 			HTONS(ip->ip_off);
747 			ip->ip_sum = 0;
748 			if (hlen == sizeof(struct ip))
749 				ip->ip_sum = in_cksum_hdr(ip);
750 			else
751 				ip->ip_sum = in_cksum(m, hlen);
752 			NTOHS(ip->ip_off);
753 			NTOHS(ip->ip_len);
754 			ip->ip_len -= hlen;
755 		}
756 #endif
757 	} else
758 		ip->ip_len -= hlen;
759 
760 #ifdef IPDIVERT
761 	/*
762 	 * Divert or tee packet to the divert protocol if required.
763 	 *
764 	 * If divert_info is zero then cookie should be too, so we shouldn't
765 	 * need to clear them here.  Assume divert_packet() does so also.
766 	 */
767 	if (divert_info != 0) {
768 		struct mbuf *clone = NULL;
769 
770 		/* Clone packet if we're doing a 'tee' */
771 		if ((divert_info & IP_FW_PORT_TEE_FLAG) != 0)
772 			clone = m_dup(m, M_DONTWAIT);
773 
774 		/* Restore packet header fields to original values */
775 		ip->ip_len += hlen;
776 		HTONS(ip->ip_len);
777 		HTONS(ip->ip_off);
778 
779 		/* Deliver packet to divert input routine */
780 		ip_divert_cookie = divert_cookie;
781 		divert_packet(m, 1, divert_info & 0xffff);
782 		ipstat.ips_delivered++;
783 
784 		/* If 'tee', continue with original packet */
785 		if (clone == NULL)
786 			return;
787 		m = clone;
788 		ip = mtod(m, struct ip *);
789 	}
790 #endif
791 
792 	/*
793 	 * Switch out to protocol's input routine.
794 	 */
795 	ipstat.ips_delivered++;
796     {
797 	int off = hlen, nh = ip->ip_p;
798 
799 	(*inetsw[ip_protox[ip->ip_p]].pr_input)(m, off, nh);
800 #ifdef	IPFIREWALL_FORWARD
801 	ip_fw_fwd_addr = NULL;	/* tcp needed it */
802 #endif
803 	return;
804     }
805 bad:
806 #ifdef	IPFIREWALL_FORWARD
807 	ip_fw_fwd_addr = NULL;
808 #endif
809 	m_freem(m);
810 }
811 
812 /*
813  * IP software interrupt routine - to go away sometime soon
814  */
815 static void
816 ipintr(void)
817 {
818 	struct mbuf *m;
819 
820 	while (1) {
821 		IF_DEQUEUE(&ipintrq, m);
822 		if (m == 0)
823 			return;
824 		ip_input(m);
825 	}
826 }
827 
828 /*
829  * Take incoming datagram fragment and try to reassemble it into
830  * whole datagram.  If a chain for reassembly of this datagram already
831  * exists, then it is given as fp; otherwise have to make a chain.
832  *
833  * When IPDIVERT enabled, keep additional state with each packet that
834  * tells us if we need to divert or tee the packet we're building.
835  */
836 
837 static struct mbuf *
838 #ifdef IPDIVERT
839 ip_reass(m, head, fp, divinfo, divcookie)
840 #else
841 ip_reass(m, head, fp)
842 #endif
843 	struct mbuf *m;
844 	struct ipqhead *head;
845 	struct ipq *fp;
846 #ifdef IPDIVERT
847 	u_int32_t *divinfo;
848 	u_int16_t *divcookie;
849 #endif
850 {
851 	struct ip *ip = mtod(m, struct ip *);
852 	register struct mbuf *p, *q, *nq;
853 	struct mbuf *t;
854 	int hlen = IP_VHL_HL(ip->ip_vhl) << 2;
855 	int i, next;
856 
857 	/*
858 	 * Presence of header sizes in mbufs
859 	 * would confuse code below.
860 	 */
861 	m->m_data += hlen;
862 	m->m_len -= hlen;
863 
864 	/*
865 	 * If first fragment to arrive, create a reassembly queue.
866 	 */
867 	if (fp == 0) {
868 		if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL)
869 			goto dropfrag;
870 		fp = mtod(t, struct ipq *);
871 		TAILQ_INSERT_HEAD(head, fp, ipq_list);
872 		nipq++;
873 		fp->ipq_ttl = IPFRAGTTL;
874 		fp->ipq_p = ip->ip_p;
875 		fp->ipq_id = ip->ip_id;
876 		fp->ipq_src = ip->ip_src;
877 		fp->ipq_dst = ip->ip_dst;
878 		fp->ipq_frags = m;
879 		m->m_nextpkt = NULL;
880 #ifdef IPDIVERT
881 		fp->ipq_div_info = 0;
882 		fp->ipq_div_cookie = 0;
883 #endif
884 		goto inserted;
885 	}
886 
887 #define GETIP(m)	((struct ip*)((m)->m_pkthdr.header))
888 
889 	/*
890 	 * Find a segment which begins after this one does.
891 	 */
892 	for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt)
893 		if (GETIP(q)->ip_off > ip->ip_off)
894 			break;
895 
896 	/*
897 	 * If there is a preceding segment, it may provide some of
898 	 * our data already.  If so, drop the data from the incoming
899 	 * segment.  If it provides all of our data, drop us, otherwise
900 	 * stick new segment in the proper place.
901 	 *
902 	 * If some of the data is dropped from the the preceding
903 	 * segment, then it's checksum is invalidated.
904 	 */
905 	if (p) {
906 		i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off;
907 		if (i > 0) {
908 			if (i >= ip->ip_len)
909 				goto dropfrag;
910 			m_adj(m, i);
911 			m->m_pkthdr.csum_flags = 0;
912 			ip->ip_off += i;
913 			ip->ip_len -= i;
914 		}
915 		m->m_nextpkt = p->m_nextpkt;
916 		p->m_nextpkt = m;
917 	} else {
918 		m->m_nextpkt = fp->ipq_frags;
919 		fp->ipq_frags = m;
920 	}
921 
922 	/*
923 	 * While we overlap succeeding segments trim them or,
924 	 * if they are completely covered, dequeue them.
925 	 */
926 	for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off;
927 	     q = nq) {
928 		i = (ip->ip_off + ip->ip_len) -
929 		    GETIP(q)->ip_off;
930 		if (i < GETIP(q)->ip_len) {
931 			GETIP(q)->ip_len -= i;
932 			GETIP(q)->ip_off += i;
933 			m_adj(q, i);
934 			q->m_pkthdr.csum_flags = 0;
935 			break;
936 		}
937 		nq = q->m_nextpkt;
938 		m->m_nextpkt = nq;
939 		m_freem(q);
940 	}
941 
942 inserted:
943 
944 #ifdef IPDIVERT
945 	/*
946 	 * Transfer firewall instructions to the fragment structure.
947 	 * Any fragment diverting causes the whole packet to divert.
948 	 */
949 	fp->ipq_div_info = *divinfo;
950 	fp->ipq_div_cookie = *divcookie;
951 	*divinfo = 0;
952 	*divcookie = 0;
953 #endif
954 
955 	/*
956 	 * Check for complete reassembly.
957 	 */
958 	next = 0;
959 	for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
960 		if (GETIP(q)->ip_off != next)
961 			return (0);
962 		next += GETIP(q)->ip_len;
963 	}
964 	/* Make sure the last packet didn't have the IP_MF flag */
965 	if (p->m_flags & M_FRAG)
966 		return (0);
967 
968 	/*
969 	 * Reassembly is complete.  Make sure the packet is a sane size.
970 	 */
971 	q = fp->ipq_frags;
972 	ip = GETIP(q);
973 	if (next + (IP_VHL_HL(ip->ip_vhl) << 2) > IP_MAXPACKET) {
974 		ipstat.ips_toolong++;
975 		ip_freef(head, fp);
976 		return (0);
977 	}
978 
979 	/*
980 	 * Concatenate fragments.
981 	 */
982 	m = q;
983 	t = m->m_next;
984 	m->m_next = 0;
985 	m_cat(m, t);
986 	nq = q->m_nextpkt;
987 	q->m_nextpkt = 0;
988 	for (q = nq; q != NULL; q = nq) {
989 		nq = q->m_nextpkt;
990 		q->m_nextpkt = NULL;
991 		m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
992 		m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
993 		m_cat(m, q);
994 	}
995 
996 #ifdef IPDIVERT
997 	/*
998 	 * Extract firewall instructions from the fragment structure.
999 	 */
1000 	*divinfo = fp->ipq_div_info;
1001 	*divcookie = fp->ipq_div_cookie;
1002 #endif
1003 
1004 	/*
1005 	 * Create header for new ip packet by
1006 	 * modifying header of first packet;
1007 	 * dequeue and discard fragment reassembly header.
1008 	 * Make header visible.
1009 	 */
1010 	ip->ip_len = next;
1011 	ip->ip_src = fp->ipq_src;
1012 	ip->ip_dst = fp->ipq_dst;
1013 	TAILQ_REMOVE(head, fp, ipq_list);
1014 	nipq--;
1015 	(void) m_free(dtom(fp));
1016 	m->m_len += (IP_VHL_HL(ip->ip_vhl) << 2);
1017 	m->m_data -= (IP_VHL_HL(ip->ip_vhl) << 2);
1018 	/* some debugging cruft by sklower, below, will go away soon */
1019 	if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */
1020 		register int plen = 0;
1021 		for (t = m; t; t = t->m_next)
1022 			plen += t->m_len;
1023 		m->m_pkthdr.len = plen;
1024 	}
1025 	return (m);
1026 
1027 dropfrag:
1028 #ifdef IPDIVERT
1029 	*divinfo = 0;
1030 	*divcookie = 0;
1031 #endif
1032 	ipstat.ips_fragdropped++;
1033 	m_freem(m);
1034 	return (0);
1035 
1036 #undef GETIP
1037 }
1038 
1039 /*
1040  * Free a fragment reassembly header and all
1041  * associated datagrams.
1042  */
1043 static void
1044 ip_freef(fhp, fp)
1045 	struct ipqhead *fhp;
1046 	struct ipq *fp;
1047 {
1048 	register struct mbuf *q;
1049 
1050 	while (fp->ipq_frags) {
1051 		q = fp->ipq_frags;
1052 		fp->ipq_frags = q->m_nextpkt;
1053 		m_freem(q);
1054 	}
1055 	TAILQ_REMOVE(fhp, fp, ipq_list);
1056 	(void) m_free(dtom(fp));
1057 	nipq--;
1058 }
1059 
1060 /*
1061  * IP timer processing;
1062  * if a timer expires on a reassembly
1063  * queue, discard it.
1064  */
1065 void
1066 ip_slowtimo()
1067 {
1068 	register struct ipq *fp;
1069 	int s = splnet();
1070 	int i;
1071 
1072 	for (i = 0; i < IPREASS_NHASH; i++) {
1073 		for(fp = TAILQ_FIRST(&ipq[i]); fp;) {
1074 			struct ipq *fpp;
1075 
1076 			fpp = fp;
1077 			fp = TAILQ_NEXT(fp, ipq_list);
1078 			if(--fpp->ipq_ttl == 0) {
1079 				ipstat.ips_fragtimeout++;
1080 				ip_freef(&ipq[i], fpp);
1081 			}
1082 		}
1083 	}
1084 	ipflow_slowtimo();
1085 	splx(s);
1086 }
1087 
1088 /*
1089  * Drain off all datagram fragments.
1090  */
1091 void
1092 ip_drain()
1093 {
1094 	int     i;
1095 
1096 	for (i = 0; i < IPREASS_NHASH; i++) {
1097 		while(!TAILQ_EMPTY(&ipq[i])) {
1098 			ipstat.ips_fragdropped++;
1099 			ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i]));
1100 		}
1101 	}
1102 	in_rtqdrain();
1103 }
1104 
1105 /*
1106  * Do option processing on a datagram,
1107  * possibly discarding it if bad options are encountered,
1108  * or forwarding it if source-routed.
1109  * Returns 1 if packet has been forwarded/freed,
1110  * 0 if the packet should be processed further.
1111  */
1112 static int
1113 ip_dooptions(m)
1114 	struct mbuf *m;
1115 {
1116 	register struct ip *ip = mtod(m, struct ip *);
1117 	register u_char *cp;
1118 	register struct ip_timestamp *ipt;
1119 	register struct in_ifaddr *ia;
1120 	int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0;
1121 	struct in_addr *sin, dst;
1122 	n_time ntime;
1123 
1124 	dst = ip->ip_dst;
1125 	cp = (u_char *)(ip + 1);
1126 	cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip);
1127 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1128 		opt = cp[IPOPT_OPTVAL];
1129 		if (opt == IPOPT_EOL)
1130 			break;
1131 		if (opt == IPOPT_NOP)
1132 			optlen = 1;
1133 		else {
1134 			if (cnt < IPOPT_OLEN + sizeof(*cp)) {
1135 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1136 				goto bad;
1137 			}
1138 			optlen = cp[IPOPT_OLEN];
1139 			if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
1140 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1141 				goto bad;
1142 			}
1143 		}
1144 		switch (opt) {
1145 
1146 		default:
1147 			break;
1148 
1149 		/*
1150 		 * Source routing with record.
1151 		 * Find interface with current destination address.
1152 		 * If none on this machine then drop if strictly routed,
1153 		 * or do nothing if loosely routed.
1154 		 * Record interface address and bring up next address
1155 		 * component.  If strictly routed make sure next
1156 		 * address is on directly accessible net.
1157 		 */
1158 		case IPOPT_LSRR:
1159 		case IPOPT_SSRR:
1160 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1161 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1162 				goto bad;
1163 			}
1164 			ipaddr.sin_addr = ip->ip_dst;
1165 			ia = (struct in_ifaddr *)
1166 				ifa_ifwithaddr((struct sockaddr *)&ipaddr);
1167 			if (ia == 0) {
1168 				if (opt == IPOPT_SSRR) {
1169 					type = ICMP_UNREACH;
1170 					code = ICMP_UNREACH_SRCFAIL;
1171 					goto bad;
1172 				}
1173 				if (!ip_dosourceroute)
1174 					goto nosourcerouting;
1175 				/*
1176 				 * Loose routing, and not at next destination
1177 				 * yet; nothing to do except forward.
1178 				 */
1179 				break;
1180 			}
1181 			off--;			/* 0 origin */
1182 			if (off > optlen - (int)sizeof(struct in_addr)) {
1183 				/*
1184 				 * End of source route.  Should be for us.
1185 				 */
1186 				if (!ip_acceptsourceroute)
1187 					goto nosourcerouting;
1188 				save_rte(cp, ip->ip_src);
1189 				break;
1190 			}
1191 
1192 			if (!ip_dosourceroute) {
1193 				if (ipforwarding) {
1194 					char buf[16]; /* aaa.bbb.ccc.ddd\0 */
1195 					/*
1196 					 * Acting as a router, so generate ICMP
1197 					 */
1198 nosourcerouting:
1199 					strcpy(buf, inet_ntoa(ip->ip_dst));
1200 					log(LOG_WARNING,
1201 					    "attempted source route from %s to %s\n",
1202 					    inet_ntoa(ip->ip_src), buf);
1203 					type = ICMP_UNREACH;
1204 					code = ICMP_UNREACH_SRCFAIL;
1205 					goto bad;
1206 				} else {
1207 					/*
1208 					 * Not acting as a router, so silently drop.
1209 					 */
1210 					ipstat.ips_cantforward++;
1211 					m_freem(m);
1212 					return (1);
1213 				}
1214 			}
1215 
1216 			/*
1217 			 * locate outgoing interface
1218 			 */
1219 			(void)memcpy(&ipaddr.sin_addr, cp + off,
1220 			    sizeof(ipaddr.sin_addr));
1221 
1222 			if (opt == IPOPT_SSRR) {
1223 #define	INA	struct in_ifaddr *
1224 #define	SA	struct sockaddr *
1225 			    if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == 0)
1226 				ia = (INA)ifa_ifwithnet((SA)&ipaddr);
1227 			} else
1228 				ia = ip_rtaddr(ipaddr.sin_addr);
1229 			if (ia == 0) {
1230 				type = ICMP_UNREACH;
1231 				code = ICMP_UNREACH_SRCFAIL;
1232 				goto bad;
1233 			}
1234 			ip->ip_dst = ipaddr.sin_addr;
1235 			(void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
1236 			    sizeof(struct in_addr));
1237 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1238 			/*
1239 			 * Let ip_intr's mcast routing check handle mcast pkts
1240 			 */
1241 			forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr));
1242 			break;
1243 
1244 		case IPOPT_RR:
1245 			if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1246 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1247 				goto bad;
1248 			}
1249 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1250 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1251 				goto bad;
1252 			}
1253 			/*
1254 			 * If no space remains, ignore.
1255 			 */
1256 			off--;			/* 0 origin */
1257 			if (off > optlen - (int)sizeof(struct in_addr))
1258 				break;
1259 			(void)memcpy(&ipaddr.sin_addr, &ip->ip_dst,
1260 			    sizeof(ipaddr.sin_addr));
1261 			/*
1262 			 * locate outgoing interface; if we're the destination,
1263 			 * use the incoming interface (should be same).
1264 			 */
1265 			if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == 0 &&
1266 			    (ia = ip_rtaddr(ipaddr.sin_addr)) == 0) {
1267 				type = ICMP_UNREACH;
1268 				code = ICMP_UNREACH_HOST;
1269 				goto bad;
1270 			}
1271 			(void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
1272 			    sizeof(struct in_addr));
1273 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1274 			break;
1275 
1276 		case IPOPT_TS:
1277 			code = cp - (u_char *)ip;
1278 			ipt = (struct ip_timestamp *)cp;
1279 			if (ipt->ipt_len < 5)
1280 				goto bad;
1281 			if (ipt->ipt_ptr >
1282 			    ipt->ipt_len - (int)sizeof(int32_t)) {
1283 				if (++ipt->ipt_oflw == 0)
1284 					goto bad;
1285 				break;
1286 			}
1287 			sin = (struct in_addr *)(cp + ipt->ipt_ptr - 1);
1288 			switch (ipt->ipt_flg) {
1289 
1290 			case IPOPT_TS_TSONLY:
1291 				break;
1292 
1293 			case IPOPT_TS_TSANDADDR:
1294 				if (ipt->ipt_ptr - 1 + sizeof(n_time) +
1295 				    sizeof(struct in_addr) > ipt->ipt_len)
1296 					goto bad;
1297 				ipaddr.sin_addr = dst;
1298 				ia = (INA)ifaof_ifpforaddr((SA)&ipaddr,
1299 							    m->m_pkthdr.rcvif);
1300 				if (ia == 0)
1301 					continue;
1302 				(void)memcpy(sin, &IA_SIN(ia)->sin_addr,
1303 				    sizeof(struct in_addr));
1304 				ipt->ipt_ptr += sizeof(struct in_addr);
1305 				break;
1306 
1307 			case IPOPT_TS_PRESPEC:
1308 				if (ipt->ipt_ptr - 1 + sizeof(n_time) +
1309 				    sizeof(struct in_addr) > ipt->ipt_len)
1310 					goto bad;
1311 				(void)memcpy(&ipaddr.sin_addr, sin,
1312 				    sizeof(struct in_addr));
1313 				if (ifa_ifwithaddr((SA)&ipaddr) == 0)
1314 					continue;
1315 				ipt->ipt_ptr += sizeof(struct in_addr);
1316 				break;
1317 
1318 			default:
1319 				goto bad;
1320 			}
1321 			ntime = iptime();
1322 			(void)memcpy(cp + ipt->ipt_ptr - 1, &ntime,
1323 			    sizeof(n_time));
1324 			ipt->ipt_ptr += sizeof(n_time);
1325 		}
1326 	}
1327 	if (forward && ipforwarding) {
1328 		ip_forward(m, 1);
1329 		return (1);
1330 	}
1331 	return (0);
1332 bad:
1333 	icmp_error(m, type, code, 0, 0);
1334 	ipstat.ips_badoptions++;
1335 	return (1);
1336 }
1337 
1338 /*
1339  * Given address of next destination (final or next hop),
1340  * return internet address info of interface to be used to get there.
1341  */
1342 static struct in_ifaddr *
1343 ip_rtaddr(dst)
1344 	 struct in_addr dst;
1345 {
1346 	register struct sockaddr_in *sin;
1347 
1348 	sin = (struct sockaddr_in *) &ipforward_rt.ro_dst;
1349 
1350 	if (ipforward_rt.ro_rt == 0 ||
1351 	    !(ipforward_rt.ro_rt->rt_flags & RTF_UP) ||
1352 	    dst.s_addr != sin->sin_addr.s_addr) {
1353 		if (ipforward_rt.ro_rt) {
1354 			RTFREE(ipforward_rt.ro_rt);
1355 			ipforward_rt.ro_rt = 0;
1356 		}
1357 		sin->sin_family = AF_INET;
1358 		sin->sin_len = sizeof(*sin);
1359 		sin->sin_addr = dst;
1360 
1361 		rtalloc_ign(&ipforward_rt, RTF_PRCLONING);
1362 	}
1363 	if (ipforward_rt.ro_rt == 0)
1364 		return ((struct in_ifaddr *)0);
1365 	return ((struct in_ifaddr *) ipforward_rt.ro_rt->rt_ifa);
1366 }
1367 
1368 /*
1369  * Save incoming source route for use in replies,
1370  * to be picked up later by ip_srcroute if the receiver is interested.
1371  */
1372 void
1373 save_rte(option, dst)
1374 	u_char *option;
1375 	struct in_addr dst;
1376 {
1377 	unsigned olen;
1378 
1379 	olen = option[IPOPT_OLEN];
1380 #ifdef DIAGNOSTIC
1381 	if (ipprintfs)
1382 		printf("save_rte: olen %d\n", olen);
1383 #endif
1384 	if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst)))
1385 		return;
1386 	bcopy(option, ip_srcrt.srcopt, olen);
1387 	ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1388 	ip_srcrt.dst = dst;
1389 }
1390 
1391 /*
1392  * Retrieve incoming source route for use in replies,
1393  * in the same form used by setsockopt.
1394  * The first hop is placed before the options, will be removed later.
1395  */
1396 struct mbuf *
1397 ip_srcroute()
1398 {
1399 	register struct in_addr *p, *q;
1400 	register struct mbuf *m;
1401 
1402 	if (ip_nhops == 0)
1403 		return ((struct mbuf *)0);
1404 	m = m_get(M_DONTWAIT, MT_HEADER);
1405 	if (m == 0)
1406 		return ((struct mbuf *)0);
1407 
1408 #define OPTSIZ	(sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt))
1409 
1410 	/* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
1411 	m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) +
1412 	    OPTSIZ;
1413 #ifdef DIAGNOSTIC
1414 	if (ipprintfs)
1415 		printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len);
1416 #endif
1417 
1418 	/*
1419 	 * First save first hop for return route
1420 	 */
1421 	p = &ip_srcrt.route[ip_nhops - 1];
1422 	*(mtod(m, struct in_addr *)) = *p--;
1423 #ifdef DIAGNOSTIC
1424 	if (ipprintfs)
1425 		printf(" hops %lx", (u_long)ntohl(mtod(m, struct in_addr *)->s_addr));
1426 #endif
1427 
1428 	/*
1429 	 * Copy option fields and padding (nop) to mbuf.
1430 	 */
1431 	ip_srcrt.nop = IPOPT_NOP;
1432 	ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
1433 	(void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr),
1434 	    &ip_srcrt.nop, OPTSIZ);
1435 	q = (struct in_addr *)(mtod(m, caddr_t) +
1436 	    sizeof(struct in_addr) + OPTSIZ);
1437 #undef OPTSIZ
1438 	/*
1439 	 * Record return path as an IP source route,
1440 	 * reversing the path (pointers are now aligned).
1441 	 */
1442 	while (p >= ip_srcrt.route) {
1443 #ifdef DIAGNOSTIC
1444 		if (ipprintfs)
1445 			printf(" %lx", (u_long)ntohl(q->s_addr));
1446 #endif
1447 		*q++ = *p--;
1448 	}
1449 	/*
1450 	 * Last hop goes to final destination.
1451 	 */
1452 	*q = ip_srcrt.dst;
1453 #ifdef DIAGNOSTIC
1454 	if (ipprintfs)
1455 		printf(" %lx\n", (u_long)ntohl(q->s_addr));
1456 #endif
1457 	return (m);
1458 }
1459 
1460 /*
1461  * Strip out IP options, at higher
1462  * level protocol in the kernel.
1463  * Second argument is buffer to which options
1464  * will be moved, and return value is their length.
1465  * XXX should be deleted; last arg currently ignored.
1466  */
1467 void
1468 ip_stripoptions(m, mopt)
1469 	register struct mbuf *m;
1470 	struct mbuf *mopt;
1471 {
1472 	register int i;
1473 	struct ip *ip = mtod(m, struct ip *);
1474 	register caddr_t opts;
1475 	int olen;
1476 
1477 	olen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip);
1478 	opts = (caddr_t)(ip + 1);
1479 	i = m->m_len - (sizeof (struct ip) + olen);
1480 	bcopy(opts + olen, opts, (unsigned)i);
1481 	m->m_len -= olen;
1482 	if (m->m_flags & M_PKTHDR)
1483 		m->m_pkthdr.len -= olen;
1484 	ip->ip_vhl = IP_MAKE_VHL(IPVERSION, sizeof(struct ip) >> 2);
1485 }
1486 
1487 u_char inetctlerrmap[PRC_NCMDS] = {
1488 	0,		0,		0,		0,
1489 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1490 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1491 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1492 	0,		0,		0,		0,
1493 	ENOPROTOOPT,	ENETRESET
1494 };
1495 
1496 /*
1497  * Forward a packet.  If some error occurs return the sender
1498  * an icmp packet.  Note we can't always generate a meaningful
1499  * icmp message because icmp doesn't have a large enough repertoire
1500  * of codes and types.
1501  *
1502  * If not forwarding, just drop the packet.  This could be confusing
1503  * if ipforwarding was zero but some routing protocol was advancing
1504  * us as a gateway to somewhere.  However, we must let the routing
1505  * protocol deal with that.
1506  *
1507  * The srcrt parameter indicates whether the packet is being forwarded
1508  * via a source route.
1509  */
1510 static void
1511 ip_forward(m, srcrt)
1512 	struct mbuf *m;
1513 	int srcrt;
1514 {
1515 	register struct ip *ip = mtod(m, struct ip *);
1516 	register struct rtentry *rt;
1517 	int error, type = 0, code = 0;
1518 	struct mbuf *mcopy;
1519 	n_long dest;
1520 	struct ifnet *destifp;
1521 #ifdef IPSEC
1522 	struct ifnet dummyifp;
1523 #endif
1524 
1525 	dest = 0;
1526 #ifdef DIAGNOSTIC
1527 	if (ipprintfs)
1528 		printf("forward: src %lx dst %lx ttl %x\n",
1529 		    (u_long)ip->ip_src.s_addr, (u_long)ip->ip_dst.s_addr,
1530 		    ip->ip_ttl);
1531 #endif
1532 
1533 
1534 	if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
1535 		ipstat.ips_cantforward++;
1536 		m_freem(m);
1537 		return;
1538 	}
1539 #ifdef IPSTEALTH
1540 	if (!ipstealth) {
1541 #endif
1542 		if (ip->ip_ttl <= IPTTLDEC) {
1543 			icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS,
1544 			    dest, 0);
1545 			return;
1546 		}
1547 #ifdef IPSTEALTH
1548 	}
1549 #endif
1550 
1551 	if (ip_rtaddr(ip->ip_dst) == 0) {
1552 		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0);
1553 		return;
1554 	} else
1555 		rt = ipforward_rt.ro_rt;
1556 
1557 	/*
1558 	 * Save the IP header and at most 8 bytes of the payload,
1559 	 * in case we need to generate an ICMP message to the src.
1560 	 *
1561 	 * We don't use m_copy() because it might return a reference
1562 	 * to a shared cluster. Both this function and ip_output()
1563 	 * assume exclusive access to the IP header in `m', so any
1564 	 * data in a cluster may change before we reach icmp_error().
1565 	 */
1566 	MGET(mcopy, M_DONTWAIT, m->m_type);
1567 	if (mcopy != NULL) {
1568 		M_COPY_PKTHDR(mcopy, m);
1569 		mcopy->m_len = imin((IP_VHL_HL(ip->ip_vhl) << 2) + 8,
1570 		    (int)ip->ip_len);
1571 		m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1572 	}
1573 
1574 #ifdef IPSTEALTH
1575 	if (!ipstealth) {
1576 #endif
1577 		ip->ip_ttl -= IPTTLDEC;
1578 #ifdef IPSTEALTH
1579 	}
1580 #endif
1581 
1582 	/*
1583 	 * If forwarding packet using same interface that it came in on,
1584 	 * perhaps should send a redirect to sender to shortcut a hop.
1585 	 * Only send redirect if source is sending directly to us,
1586 	 * and if packet was not source routed (or has any options).
1587 	 * Also, don't send redirect if forwarding using a default route
1588 	 * or a route modified by a redirect.
1589 	 */
1590 #define	satosin(sa)	((struct sockaddr_in *)(sa))
1591 	if (rt->rt_ifp == m->m_pkthdr.rcvif &&
1592 	    (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1593 	    satosin(rt_key(rt))->sin_addr.s_addr != 0 &&
1594 	    ipsendredirects && !srcrt) {
1595 #define	RTA(rt)	((struct in_ifaddr *)(rt->rt_ifa))
1596 		u_long src = ntohl(ip->ip_src.s_addr);
1597 
1598 		if (RTA(rt) &&
1599 		    (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1600 		    if (rt->rt_flags & RTF_GATEWAY)
1601 			dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
1602 		    else
1603 			dest = ip->ip_dst.s_addr;
1604 		    /* Router requirements says to only send host redirects */
1605 		    type = ICMP_REDIRECT;
1606 		    code = ICMP_REDIRECT_HOST;
1607 #ifdef DIAGNOSTIC
1608 		    if (ipprintfs)
1609 		        printf("redirect (%d) to %lx\n", code, (u_long)dest);
1610 #endif
1611 		}
1612 	}
1613 
1614 	error = ip_output(m, (struct mbuf *)0, &ipforward_rt,
1615 			  IP_FORWARDING, 0);
1616 	if (error)
1617 		ipstat.ips_cantforward++;
1618 	else {
1619 		ipstat.ips_forward++;
1620 		if (type)
1621 			ipstat.ips_redirectsent++;
1622 		else {
1623 			if (mcopy) {
1624 				ipflow_create(&ipforward_rt, mcopy);
1625 				m_freem(mcopy);
1626 			}
1627 			return;
1628 		}
1629 	}
1630 	if (mcopy == NULL)
1631 		return;
1632 	destifp = NULL;
1633 
1634 	switch (error) {
1635 
1636 	case 0:				/* forwarded, but need redirect */
1637 		/* type, code set above */
1638 		break;
1639 
1640 	case ENETUNREACH:		/* shouldn't happen, checked above */
1641 	case EHOSTUNREACH:
1642 	case ENETDOWN:
1643 	case EHOSTDOWN:
1644 	default:
1645 		type = ICMP_UNREACH;
1646 		code = ICMP_UNREACH_HOST;
1647 		break;
1648 
1649 	case EMSGSIZE:
1650 		type = ICMP_UNREACH;
1651 		code = ICMP_UNREACH_NEEDFRAG;
1652 #ifndef IPSEC
1653 		if (ipforward_rt.ro_rt)
1654 			destifp = ipforward_rt.ro_rt->rt_ifp;
1655 #else
1656 		/*
1657 		 * If the packet is routed over IPsec tunnel, tell the
1658 		 * originator the tunnel MTU.
1659 		 *	tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
1660 		 * XXX quickhack!!!
1661 		 */
1662 		if (ipforward_rt.ro_rt) {
1663 			struct secpolicy *sp = NULL;
1664 			int ipsecerror;
1665 			int ipsechdr;
1666 			struct route *ro;
1667 
1668 			sp = ipsec4_getpolicybyaddr(mcopy,
1669 						    IPSEC_DIR_OUTBOUND,
1670 			                            IP_FORWARDING,
1671 			                            &ipsecerror);
1672 
1673 			if (sp == NULL)
1674 				destifp = ipforward_rt.ro_rt->rt_ifp;
1675 			else {
1676 				/* count IPsec header size */
1677 				ipsechdr = ipsec4_hdrsiz(mcopy,
1678 							 IPSEC_DIR_OUTBOUND,
1679 							 NULL);
1680 
1681 				/*
1682 				 * find the correct route for outer IPv4
1683 				 * header, compute tunnel MTU.
1684 				 *
1685 				 * XXX BUG ALERT
1686 				 * The "dummyifp" code relies upon the fact
1687 				 * that icmp_error() touches only ifp->if_mtu.
1688 				 */
1689 				/*XXX*/
1690 				destifp = NULL;
1691 				if (sp->req != NULL
1692 				 && sp->req->sav != NULL
1693 				 && sp->req->sav->sah != NULL) {
1694 					ro = &sp->req->sav->sah->sa_route;
1695 					if (ro->ro_rt && ro->ro_rt->rt_ifp) {
1696 						dummyifp.if_mtu =
1697 						    ro->ro_rt->rt_ifp->if_mtu;
1698 						dummyifp.if_mtu -= ipsechdr;
1699 						destifp = &dummyifp;
1700 					}
1701 				}
1702 
1703 				key_freesp(sp);
1704 			}
1705 		}
1706 #endif /*IPSEC*/
1707 		ipstat.ips_cantfrag++;
1708 		break;
1709 
1710 	case ENOBUFS:
1711 		type = ICMP_SOURCEQUENCH;
1712 		code = 0;
1713 		break;
1714 
1715 	case EACCES:			/* ipfw denied packet */
1716 		m_freem(mcopy);
1717 		return;
1718 	}
1719 	icmp_error(mcopy, type, code, dest, destifp);
1720 }
1721 
1722 void
1723 ip_savecontrol(inp, mp, ip, m)
1724 	register struct inpcb *inp;
1725 	register struct mbuf **mp;
1726 	register struct ip *ip;
1727 	register struct mbuf *m;
1728 {
1729 	if (inp->inp_socket->so_options & SO_TIMESTAMP) {
1730 		struct timeval tv;
1731 
1732 		microtime(&tv);
1733 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1734 			SCM_TIMESTAMP, SOL_SOCKET);
1735 		if (*mp)
1736 			mp = &(*mp)->m_next;
1737 	}
1738 	if (inp->inp_flags & INP_RECVDSTADDR) {
1739 		*mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
1740 		    sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1741 		if (*mp)
1742 			mp = &(*mp)->m_next;
1743 	}
1744 #ifdef notyet
1745 	/* XXX
1746 	 * Moving these out of udp_input() made them even more broken
1747 	 * than they already were.
1748 	 */
1749 	/* options were tossed already */
1750 	if (inp->inp_flags & INP_RECVOPTS) {
1751 		*mp = sbcreatecontrol((caddr_t) opts_deleted_above,
1752 		    sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1753 		if (*mp)
1754 			mp = &(*mp)->m_next;
1755 	}
1756 	/* ip_srcroute doesn't do what we want here, need to fix */
1757 	if (inp->inp_flags & INP_RECVRETOPTS) {
1758 		*mp = sbcreatecontrol((caddr_t) ip_srcroute(),
1759 		    sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1760 		if (*mp)
1761 			mp = &(*mp)->m_next;
1762 	}
1763 #endif
1764 	if (inp->inp_flags & INP_RECVIF) {
1765 		struct ifnet *ifp;
1766 		struct sdlbuf {
1767 			struct sockaddr_dl sdl;
1768 			u_char	pad[32];
1769 		} sdlbuf;
1770 		struct sockaddr_dl *sdp;
1771 		struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1772 
1773 		if (((ifp = m->m_pkthdr.rcvif))
1774 		&& ( ifp->if_index && (ifp->if_index <= if_index))) {
1775 			sdp = (struct sockaddr_dl *)(ifnet_addrs
1776 					[ifp->if_index - 1]->ifa_addr);
1777 			/*
1778 			 * Change our mind and don't try copy.
1779 			 */
1780 			if ((sdp->sdl_family != AF_LINK)
1781 			|| (sdp->sdl_len > sizeof(sdlbuf))) {
1782 				goto makedummy;
1783 			}
1784 			bcopy(sdp, sdl2, sdp->sdl_len);
1785 		} else {
1786 makedummy:
1787 			sdl2->sdl_len
1788 				= offsetof(struct sockaddr_dl, sdl_data[0]);
1789 			sdl2->sdl_family = AF_LINK;
1790 			sdl2->sdl_index = 0;
1791 			sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1792 		}
1793 		*mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len,
1794 			IP_RECVIF, IPPROTO_IP);
1795 		if (*mp)
1796 			mp = &(*mp)->m_next;
1797 	}
1798 }
1799 
1800 int
1801 ip_rsvp_init(struct socket *so)
1802 {
1803 	if (so->so_type != SOCK_RAW ||
1804 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
1805 	  return EOPNOTSUPP;
1806 
1807 	if (ip_rsvpd != NULL)
1808 	  return EADDRINUSE;
1809 
1810 	ip_rsvpd = so;
1811 	/*
1812 	 * This may seem silly, but we need to be sure we don't over-increment
1813 	 * the RSVP counter, in case something slips up.
1814 	 */
1815 	if (!ip_rsvp_on) {
1816 		ip_rsvp_on = 1;
1817 		rsvp_on++;
1818 	}
1819 
1820 	return 0;
1821 }
1822 
1823 int
1824 ip_rsvp_done(void)
1825 {
1826 	ip_rsvpd = NULL;
1827 	/*
1828 	 * This may seem silly, but we need to be sure we don't over-decrement
1829 	 * the RSVP counter, in case something slips up.
1830 	 */
1831 	if (ip_rsvp_on) {
1832 		ip_rsvp_on = 0;
1833 		rsvp_on--;
1834 	}
1835 	return 0;
1836 }
1837