xref: /freebsd/sys/netinet/ip_input.c (revision afcf05e46aacbe58f9fd8338ab6ecf9b7d06d004)
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 
48 #include <sys/param.h>
49 #include <sys/systm.h>
50 #include <sys/mbuf.h>
51 #include <sys/malloc.h>
52 #include <sys/domain.h>
53 #include <sys/protosw.h>
54 #include <sys/socket.h>
55 #include <sys/time.h>
56 #include <sys/kernel.h>
57 #include <sys/syslog.h>
58 #include <sys/sysctl.h>
59 
60 #include <net/pfil.h>
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 
179 /*
180  * We need to save the IP options in case a protocol wants to respond
181  * to an incoming packet over the same route if the packet got here
182  * using IP source routing.  This allows connection establishment and
183  * maintenance when the remote end is on a network that is not known
184  * to us.
185  */
186 static int	ip_nhops = 0;
187 static	struct ip_srcrt {
188 	struct	in_addr dst;			/* final destination */
189 	char	nop;				/* one NOP to align */
190 	char	srcopt[IPOPT_OFFSET + 1];	/* OPTVAL, OLEN and OFFSET */
191 	struct	in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)];
192 } ip_srcrt;
193 
194 struct sockaddr_in *ip_fw_fwd_addr;
195 
196 static void	save_rte __P((u_char *, struct in_addr));
197 static int	ip_dooptions __P((struct mbuf *));
198 static void	ip_forward __P((struct mbuf *, int));
199 static void	ip_freef __P((struct ipq *));
200 #ifdef IPDIVERT
201 static struct	mbuf *ip_reass __P((struct mbuf *,
202 			struct ipq *, struct ipq *, u_int32_t *, u_int16_t *));
203 #else
204 static struct	mbuf *ip_reass __P((struct mbuf *, struct ipq *, struct ipq *));
205 #endif
206 static struct	in_ifaddr *ip_rtaddr __P((struct in_addr));
207 static void	ipintr __P((void));
208 
209 /*
210  * IP initialization: fill in IP protocol switch table.
211  * All protocols not implemented in kernel go to raw IP protocol handler.
212  */
213 void
214 ip_init()
215 {
216 	register struct ipprotosw *pr;
217 	register int i;
218 
219 	TAILQ_INIT(&in_ifaddrhead);
220 	pr = (struct ipprotosw *)pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
221 	if (pr == 0)
222 		panic("ip_init");
223 	for (i = 0; i < IPPROTO_MAX; i++)
224 		ip_protox[i] = pr - inetsw;
225 	for (pr = (struct ipprotosw *)inetdomain.dom_protosw;
226 	    pr < (struct ipprotosw *)inetdomain.dom_protoswNPROTOSW; pr++)
227 		if (pr->pr_domain->dom_family == PF_INET &&
228 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
229 			ip_protox[pr->pr_protocol] = pr - inetsw;
230 
231 	for (i = 0; i < IPREASS_NHASH; i++)
232 	    ipq[i].next = ipq[i].prev = &ipq[i];
233 
234 	maxnipq = nmbclusters/4;
235 
236 	ip_id = time_second & 0xffff;
237 	ipintrq.ifq_maxlen = ipqmaxlen;
238 	mtx_init(&ipintrq.ifq_mtx, "ip_inq", MTX_DEF);
239 
240 	register_netisr(NETISR_IP, ipintr);
241 }
242 
243 static struct	sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET };
244 static struct	route ipforward_rt;
245 
246 /*
247  * Ip input routine.  Checksum and byte swap header.  If fragmented
248  * try to reassemble.  Process options.  Pass to next level.
249  */
250 void
251 ip_input(struct mbuf *m)
252 {
253 	struct ip *ip;
254 	struct ipq *fp;
255 	struct in_ifaddr *ia = NULL;
256 	int    i, hlen;
257 	u_short sum;
258 	u_int16_t divert_cookie;		/* firewall cookie */
259 	struct in_addr pkt_dst;
260 #ifdef IPDIVERT
261 	u_int32_t divert_info = 0;		/* packet divert/tee info */
262 #endif
263 	struct ip_fw_chain *rule = NULL;
264 #ifdef PFIL_HOOKS
265 	struct packet_filter_hook *pfh;
266 	struct mbuf *m0;
267 	int rv;
268 #endif /* PFIL_HOOKS */
269 
270 #ifdef IPDIVERT
271 	/* Get and reset firewall cookie */
272 	divert_cookie = ip_divert_cookie;
273 	ip_divert_cookie = 0;
274 #else
275 	divert_cookie = 0;
276 #endif
277 
278 #if defined(IPFIREWALL) && defined(DUMMYNET)
279         /*
280          * dummynet packet are prepended a vestigial mbuf with
281          * m_type = MT_DUMMYNET and m_data pointing to the matching
282          * rule.
283          */
284         if (m->m_type == MT_DUMMYNET) {
285             rule = (struct ip_fw_chain *)(m->m_data) ;
286             m = m->m_next ;
287             ip = mtod(m, struct ip *);
288             hlen = IP_VHL_HL(ip->ip_vhl) << 2;
289             goto iphack ;
290         } else
291             rule = NULL ;
292 #endif
293 
294 #ifdef	DIAGNOSTIC
295 	if (m == NULL || (m->m_flags & M_PKTHDR) == 0)
296 		panic("ip_input no HDR");
297 #endif
298 	ipstat.ips_total++;
299 
300 	if (m->m_pkthdr.len < sizeof(struct ip))
301 		goto tooshort;
302 
303 	if (m->m_len < sizeof (struct ip) &&
304 	    (m = m_pullup(m, sizeof (struct ip))) == 0) {
305 		ipstat.ips_toosmall++;
306 		return;
307 	}
308 	ip = mtod(m, struct ip *);
309 
310 	if (IP_VHL_V(ip->ip_vhl) != IPVERSION) {
311 		ipstat.ips_badvers++;
312 		goto bad;
313 	}
314 
315 	hlen = IP_VHL_HL(ip->ip_vhl) << 2;
316 	if (hlen < sizeof(struct ip)) {	/* minimum header length */
317 		ipstat.ips_badhlen++;
318 		goto bad;
319 	}
320 	if (hlen > m->m_len) {
321 		if ((m = m_pullup(m, hlen)) == 0) {
322 			ipstat.ips_badhlen++;
323 			return;
324 		}
325 		ip = mtod(m, struct ip *);
326 	}
327 	if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
328 		sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
329 	} else {
330 		if (hlen == sizeof(struct ip)) {
331 			sum = in_cksum_hdr(ip);
332 		} else {
333 			sum = in_cksum(m, hlen);
334 		}
335 	}
336 	if (sum) {
337 		ipstat.ips_badsum++;
338 		goto bad;
339 	}
340 
341 	/*
342 	 * Convert fields to host representation.
343 	 */
344 	NTOHS(ip->ip_len);
345 	if (ip->ip_len < hlen) {
346 		ipstat.ips_badlen++;
347 		goto bad;
348 	}
349 	NTOHS(ip->ip_off);
350 
351 	/*
352 	 * Check that the amount of data in the buffers
353 	 * is as at least much as the IP header would have us expect.
354 	 * Trim mbufs if longer than we expect.
355 	 * Drop packet if shorter than we expect.
356 	 */
357 	if (m->m_pkthdr.len < ip->ip_len) {
358 tooshort:
359 		ipstat.ips_tooshort++;
360 		goto bad;
361 	}
362 	if (m->m_pkthdr.len > ip->ip_len) {
363 		if (m->m_len == m->m_pkthdr.len) {
364 			m->m_len = ip->ip_len;
365 			m->m_pkthdr.len = ip->ip_len;
366 		} else
367 			m_adj(m, ip->ip_len - m->m_pkthdr.len);
368 	}
369 	/*
370 	 * IpHack's section.
371 	 * Right now when no processing on packet has done
372 	 * and it is still fresh out of network we do our black
373 	 * deals with it.
374 	 * - Firewall: deny/allow/divert
375 	 * - Xlate: translate packet's addr/port (NAT).
376 	 * - Pipe: pass pkt through dummynet.
377 	 * - Wrap: fake packet's addr/port <unimpl.>
378 	 * - Encapsulate: put it in another IP and send out. <unimp.>
379  	 */
380 
381 #if defined(IPFIREWALL) && defined(DUMMYNET)
382 iphack:
383 #endif
384 
385 #ifdef PFIL_HOOKS
386 	/*
387 	 * Run through list of hooks for input packets.  If there are any
388 	 * filters which require that additional packets in the flow are
389 	 * not fast-forwarded, they must clear the M_CANFASTFWD flag.
390 	 * Note that filters must _never_ set this flag, as another filter
391 	 * in the list may have previously cleared it.
392 	 */
393 	m0 = m;
394 	pfh = pfil_hook_get(PFIL_IN, &inetsw[ip_protox[IPPROTO_IP]].pr_pfh);
395 	for (; pfh; pfh = TAILQ_NEXT(pfh, pfil_link))
396 		if (pfh->pfil_func) {
397 			rv = pfh->pfil_func(ip, hlen,
398 					    m->m_pkthdr.rcvif, 0, &m0);
399 			if (rv)
400 				return;
401 			m = m0;
402 			if (m == NULL)
403 				return;
404 			ip = mtod(m, struct ip *);
405 		}
406 #endif /* PFIL_HOOKS */
407 
408 	if (fw_enable && ip_fw_chk_ptr) {
409 #ifdef IPFIREWALL_FORWARD
410 		/*
411 		 * If we've been forwarded from the output side, then
412 		 * skip the firewall a second time
413 		 */
414 		if (ip_fw_fwd_addr)
415 			goto ours;
416 #endif	/* IPFIREWALL_FORWARD */
417 		/*
418 		 * See the comment in ip_output for the return values
419 		 * produced by the firewall.
420 		 */
421 		i = (*ip_fw_chk_ptr)(&ip,
422 		    hlen, NULL, &divert_cookie, &m, &rule, &ip_fw_fwd_addr);
423 		if (i & IP_FW_PORT_DENY_FLAG) { /* XXX new interface-denied */
424 		    if (m)
425 			m_freem(m);
426 		    return ;
427 		}
428 		if (m == NULL) {	/* Packet discarded by firewall */
429 		    static int __debug=10;
430 		    if (__debug >0) {
431 			printf("firewall returns NULL, please update!\n");
432 			__debug-- ;
433 		    }
434 		    return;
435 		}
436 		if (i == 0 && ip_fw_fwd_addr == NULL)	/* common case */
437 			goto pass;
438 #ifdef DUMMYNET
439                 if ((i & IP_FW_PORT_DYNT_FLAG) != 0) {
440                         /* Send packet to the appropriate pipe */
441                         dummynet_io(i&0xffff,DN_TO_IP_IN,m,NULL,NULL,0, rule,
442 				    0);
443 			return;
444 		}
445 #endif
446 #ifdef IPDIVERT
447 		if (i != 0 && (i & IP_FW_PORT_DYNT_FLAG) == 0) {
448 			/* Divert or tee packet */
449 			divert_info = i;
450 			goto ours;
451 		}
452 #endif
453 #ifdef IPFIREWALL_FORWARD
454 		if (i == 0 && ip_fw_fwd_addr != NULL)
455 			goto pass;
456 #endif
457 		/*
458 		 * if we get here, the packet must be dropped
459 		 */
460 		m_freem(m);
461 		return;
462 	}
463 pass:
464 
465 	/*
466 	 * Process options and, if not destined for us,
467 	 * ship it on.  ip_dooptions returns 1 when an
468 	 * error was detected (causing an icmp message
469 	 * to be sent and the original packet to be freed).
470 	 */
471 	ip_nhops = 0;		/* for source routed packets */
472 	if (hlen > sizeof (struct ip) && ip_dooptions(m)) {
473 #ifdef IPFIREWALL_FORWARD
474 		ip_fw_fwd_addr = NULL;
475 #endif
476 		return;
477 	}
478 
479         /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
480          * matter if it is destined to another node, or whether it is
481          * a multicast one, RSVP wants it! and prevents it from being forwarded
482          * anywhere else. Also checks if the rsvp daemon is running before
483 	 * grabbing the packet.
484          */
485 	if (rsvp_on && ip->ip_p==IPPROTO_RSVP)
486 		goto ours;
487 
488 	/*
489 	 * Check our list of addresses, to see if the packet is for us.
490 	 * If we don't have any addresses, assume any unicast packet
491 	 * we receive might be for us (and let the upper layers deal
492 	 * with it).
493 	 */
494 	if (TAILQ_EMPTY(&in_ifaddrhead) &&
495 	    (m->m_flags & (M_MCAST|M_BCAST)) == 0)
496 		goto ours;
497 
498 	/*
499 	 * Cache the destination address of the packet; this may be
500 	 * changed by use of 'ipfw fwd'.
501 	 */
502 	pkt_dst = ip_fw_fwd_addr == NULL ?
503 	    ip->ip_dst : ip_fw_fwd_addr->sin_addr;
504 
505 	TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link) {
506 #define	satosin(sa)	((struct sockaddr_in *)(sa))
507 
508 #ifdef BOOTP_COMPAT
509 		if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY)
510 			goto ours;
511 #endif
512 		/*
513 		 * check that the packet is either arriving from the
514 		 * correct interface or is locally generated.
515 		 */
516 		if (ia->ia_ifp != m->m_pkthdr.rcvif &&
517 		     (m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0)
518 			continue;
519 
520 		if (IA_SIN(ia)->sin_addr.s_addr == pkt_dst.s_addr)
521 			goto ours;
522 
523 		if (ia->ia_ifp && ia->ia_ifp->if_flags & IFF_BROADCAST) {
524 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
525 			    pkt_dst.s_addr)
526 				goto ours;
527 			if (ia->ia_netbroadcast.s_addr == pkt_dst.s_addr)
528 				goto ours;
529 		}
530 	}
531 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
532 		struct in_multi *inm;
533 		if (ip_mrouter) {
534 			/*
535 			 * If we are acting as a multicast router, all
536 			 * incoming multicast packets are passed to the
537 			 * kernel-level multicast forwarding function.
538 			 * The packet is returned (relatively) intact; if
539 			 * ip_mforward() returns a non-zero value, the packet
540 			 * must be discarded, else it may be accepted below.
541 			 */
542 			if (ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) {
543 				ipstat.ips_cantforward++;
544 				m_freem(m);
545 				return;
546 			}
547 
548 			/*
549 			 * The process-level routing demon needs to receive
550 			 * all multicast IGMP packets, whether or not this
551 			 * host belongs to their destination groups.
552 			 */
553 			if (ip->ip_p == IPPROTO_IGMP)
554 				goto ours;
555 			ipstat.ips_forward++;
556 		}
557 		/*
558 		 * See if we belong to the destination multicast group on the
559 		 * arrival interface.
560 		 */
561 		IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
562 		if (inm == NULL) {
563 			ipstat.ips_notmember++;
564 			m_freem(m);
565 			return;
566 		}
567 		goto ours;
568 	}
569 	if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
570 		goto ours;
571 	if (ip->ip_dst.s_addr == INADDR_ANY)
572 		goto ours;
573 
574 #if defined(NFAITH) && 0 < NFAITH
575 	/*
576 	 * FAITH(Firewall Aided Internet Translator)
577 	 */
578 	if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
579 		if (ip_keepfaith) {
580 			if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP)
581 				goto ours;
582 		}
583 		m_freem(m);
584 		return;
585 	}
586 #endif
587 	/*
588 	 * Not for us; forward if possible and desirable.
589 	 */
590 	if (ipforwarding == 0) {
591 		ipstat.ips_cantforward++;
592 		m_freem(m);
593 	} else
594 		ip_forward(m, 0);
595 #ifdef IPFIREWALL_FORWARD
596 	ip_fw_fwd_addr = NULL;
597 #endif
598 	return;
599 
600 ours:
601 	/* Count the packet in the ip address stats */
602 	if (ia != NULL) {
603 		ia->ia_ifa.if_ipackets++;
604 		ia->ia_ifa.if_ibytes += m->m_pkthdr.len;
605 	}
606 
607 	/*
608 	 * If offset or IP_MF are set, must reassemble.
609 	 * Otherwise, nothing need be done.
610 	 * (We could look in the reassembly queue to see
611 	 * if the packet was previously fragmented,
612 	 * but it's not worth the time; just let them time out.)
613 	 */
614 	if (ip->ip_off & (IP_MF | IP_OFFMASK)) {
615 
616 		sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
617 		/*
618 		 * Look for queue of fragments
619 		 * of this datagram.
620 		 */
621 		for (fp = ipq[sum].next; fp != &ipq[sum]; fp = fp->next)
622 			if (ip->ip_id == fp->ipq_id &&
623 			    ip->ip_src.s_addr == fp->ipq_src.s_addr &&
624 			    ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
625 			    ip->ip_p == fp->ipq_p)
626 				goto found;
627 
628 		fp = 0;
629 
630 		/* check if there's a place for the new queue */
631 		if (nipq > maxnipq) {
632 		    /*
633 		     * drop something from the tail of the current queue
634 		     * before proceeding further
635 		     */
636 		    if (ipq[sum].prev == &ipq[sum]) {   /* gak */
637 			for (i = 0; i < IPREASS_NHASH; i++) {
638 			    if (ipq[i].prev != &ipq[i]) {
639 				ip_freef(ipq[i].prev);
640 				break;
641 			    }
642 			}
643 		    } else
644 			ip_freef(ipq[sum].prev);
645 		}
646 found:
647 		/*
648 		 * Adjust ip_len to not reflect header,
649 		 * convert offset of this to bytes.
650 		 */
651 		ip->ip_len -= hlen;
652 		if (ip->ip_off & IP_MF) {
653 		        /*
654 		         * Make sure that fragments have a data length
655 			 * that's a non-zero multiple of 8 bytes.
656 		         */
657 			if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
658 				ipstat.ips_toosmall++; /* XXX */
659 				goto bad;
660 			}
661 			m->m_flags |= M_FRAG;
662 		}
663 		ip->ip_off <<= 3;
664 
665 		/*
666 		 * Attempt reassembly; if it succeeds, proceed.
667 		 */
668 		ipstat.ips_fragments++;
669 		m->m_pkthdr.header = ip;
670 #ifdef IPDIVERT
671 		m = ip_reass(m,
672 		    fp, &ipq[sum], &divert_info, &divert_cookie);
673 #else
674 		m = ip_reass(m, fp, &ipq[sum]);
675 #endif
676 		if (m == 0) {
677 #ifdef IPFIREWALL_FORWARD
678 			ip_fw_fwd_addr = NULL;
679 #endif
680 			return;
681 		}
682 		ipstat.ips_reassembled++;
683 		ip = mtod(m, struct ip *);
684 		/* Get the header length of the reassembled packet */
685 		hlen = IP_VHL_HL(ip->ip_vhl) << 2;
686 #ifdef IPDIVERT
687 		/* Restore original checksum before diverting packet */
688 		if (divert_info != 0) {
689 			ip->ip_len += hlen;
690 			HTONS(ip->ip_len);
691 			HTONS(ip->ip_off);
692 			ip->ip_sum = 0;
693 			if (hlen == sizeof(struct ip))
694 				ip->ip_sum = in_cksum_hdr(ip);
695 			else
696 				ip->ip_sum = in_cksum(m, hlen);
697 			NTOHS(ip->ip_off);
698 			NTOHS(ip->ip_len);
699 			ip->ip_len -= hlen;
700 		}
701 #endif
702 	} else
703 		ip->ip_len -= hlen;
704 
705 #ifdef IPDIVERT
706 	/*
707 	 * Divert or tee packet to the divert protocol if required.
708 	 *
709 	 * If divert_info is zero then cookie should be too, so we shouldn't
710 	 * need to clear them here.  Assume divert_packet() does so also.
711 	 */
712 	if (divert_info != 0) {
713 		struct mbuf *clone = NULL;
714 
715 		/* Clone packet if we're doing a 'tee' */
716 		if ((divert_info & IP_FW_PORT_TEE_FLAG) != 0)
717 			clone = m_dup(m, M_DONTWAIT);
718 
719 		/* Restore packet header fields to original values */
720 		ip->ip_len += hlen;
721 		HTONS(ip->ip_len);
722 		HTONS(ip->ip_off);
723 
724 		/* Deliver packet to divert input routine */
725 		ip_divert_cookie = divert_cookie;
726 		divert_packet(m, 1, divert_info & 0xffff);
727 		ipstat.ips_delivered++;
728 
729 		/* If 'tee', continue with original packet */
730 		if (clone == NULL)
731 			return;
732 		m = clone;
733 		ip = mtod(m, struct ip *);
734 	}
735 #endif
736 
737 	/*
738 	 * Switch out to protocol's input routine.
739 	 */
740 	ipstat.ips_delivered++;
741     {
742 	int off = hlen, nh = ip->ip_p;
743 
744 	(*inetsw[ip_protox[ip->ip_p]].pr_input)(m, off, nh);
745 #ifdef	IPFIREWALL_FORWARD
746 	ip_fw_fwd_addr = NULL;	/* tcp needed it */
747 #endif
748 	return;
749     }
750 bad:
751 #ifdef	IPFIREWALL_FORWARD
752 	ip_fw_fwd_addr = NULL;
753 #endif
754 	m_freem(m);
755 }
756 
757 /*
758  * IP software interrupt routine - to go away sometime soon
759  */
760 static void
761 ipintr(void)
762 {
763 	struct mbuf *m;
764 
765 	while (1) {
766 		IF_DEQUEUE(&ipintrq, m);
767 		if (m == 0)
768 			return;
769 		ip_input(m);
770 	}
771 }
772 
773 /*
774  * Take incoming datagram fragment and try to reassemble it into
775  * whole datagram.  If a chain for reassembly of this datagram already
776  * exists, then it is given as fp; otherwise have to make a chain.
777  *
778  * When IPDIVERT enabled, keep additional state with each packet that
779  * tells us if we need to divert or tee the packet we're building.
780  */
781 
782 static struct mbuf *
783 #ifdef IPDIVERT
784 ip_reass(m, fp, where, divinfo, divcookie)
785 #else
786 ip_reass(m, fp, where)
787 #endif
788 	register struct mbuf *m;
789 	register struct ipq *fp;
790 	struct   ipq    *where;
791 #ifdef IPDIVERT
792 	u_int32_t *divinfo;
793 	u_int16_t *divcookie;
794 #endif
795 {
796 	struct ip *ip = mtod(m, struct ip *);
797 	register struct mbuf *p, *q, *nq;
798 	struct mbuf *t;
799 	int hlen = IP_VHL_HL(ip->ip_vhl) << 2;
800 	int i, next;
801 
802 	/*
803 	 * Presence of header sizes in mbufs
804 	 * would confuse code below.
805 	 */
806 	m->m_data += hlen;
807 	m->m_len -= hlen;
808 
809 	/*
810 	 * If first fragment to arrive, create a reassembly queue.
811 	 */
812 	if (fp == 0) {
813 		if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL)
814 			goto dropfrag;
815 		fp = mtod(t, struct ipq *);
816 		insque(fp, where);
817 		nipq++;
818 		fp->ipq_ttl = IPFRAGTTL;
819 		fp->ipq_p = ip->ip_p;
820 		fp->ipq_id = ip->ip_id;
821 		fp->ipq_src = ip->ip_src;
822 		fp->ipq_dst = ip->ip_dst;
823 		fp->ipq_frags = m;
824 		m->m_nextpkt = NULL;
825 #ifdef IPDIVERT
826 		fp->ipq_div_info = 0;
827 		fp->ipq_div_cookie = 0;
828 #endif
829 		goto inserted;
830 	}
831 
832 #define GETIP(m)	((struct ip*)((m)->m_pkthdr.header))
833 
834 	/*
835 	 * Find a segment which begins after this one does.
836 	 */
837 	for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt)
838 		if (GETIP(q)->ip_off > ip->ip_off)
839 			break;
840 
841 	/*
842 	 * If there is a preceding segment, it may provide some of
843 	 * our data already.  If so, drop the data from the incoming
844 	 * segment.  If it provides all of our data, drop us, otherwise
845 	 * stick new segment in the proper place.
846 	 *
847 	 * If some of the data is dropped from the the preceding
848 	 * segment, then it's checksum is invalidated.
849 	 */
850 	if (p) {
851 		i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off;
852 		if (i > 0) {
853 			if (i >= ip->ip_len)
854 				goto dropfrag;
855 			m_adj(m, i);
856 			m->m_pkthdr.csum_flags = 0;
857 			ip->ip_off += i;
858 			ip->ip_len -= i;
859 		}
860 		m->m_nextpkt = p->m_nextpkt;
861 		p->m_nextpkt = m;
862 	} else {
863 		m->m_nextpkt = fp->ipq_frags;
864 		fp->ipq_frags = m;
865 	}
866 
867 	/*
868 	 * While we overlap succeeding segments trim them or,
869 	 * if they are completely covered, dequeue them.
870 	 */
871 	for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off;
872 	     q = nq) {
873 		i = (ip->ip_off + ip->ip_len) -
874 		    GETIP(q)->ip_off;
875 		if (i < GETIP(q)->ip_len) {
876 			GETIP(q)->ip_len -= i;
877 			GETIP(q)->ip_off += i;
878 			m_adj(q, i);
879 			q->m_pkthdr.csum_flags = 0;
880 			break;
881 		}
882 		nq = q->m_nextpkt;
883 		m->m_nextpkt = nq;
884 		m_freem(q);
885 	}
886 
887 inserted:
888 
889 #ifdef IPDIVERT
890 	/*
891 	 * Transfer firewall instructions to the fragment structure.
892 	 * Any fragment diverting causes the whole packet to divert.
893 	 */
894 	fp->ipq_div_info = *divinfo;
895 	fp->ipq_div_cookie = *divcookie;
896 	*divinfo = 0;
897 	*divcookie = 0;
898 #endif
899 
900 	/*
901 	 * Check for complete reassembly.
902 	 */
903 	next = 0;
904 	for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
905 		if (GETIP(q)->ip_off != next)
906 			return (0);
907 		next += GETIP(q)->ip_len;
908 	}
909 	/* Make sure the last packet didn't have the IP_MF flag */
910 	if (p->m_flags & M_FRAG)
911 		return (0);
912 
913 	/*
914 	 * Reassembly is complete.  Make sure the packet is a sane size.
915 	 */
916 	q = fp->ipq_frags;
917 	ip = GETIP(q);
918 	if (next + (IP_VHL_HL(ip->ip_vhl) << 2) > IP_MAXPACKET) {
919 		ipstat.ips_toolong++;
920 		ip_freef(fp);
921 		return (0);
922 	}
923 
924 	/*
925 	 * Concatenate fragments.
926 	 */
927 	m = q;
928 	t = m->m_next;
929 	m->m_next = 0;
930 	m_cat(m, t);
931 	nq = q->m_nextpkt;
932 	q->m_nextpkt = 0;
933 	for (q = nq; q != NULL; q = nq) {
934 		nq = q->m_nextpkt;
935 		q->m_nextpkt = NULL;
936 		m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
937 		m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
938 		m_cat(m, q);
939 	}
940 
941 #ifdef IPDIVERT
942 	/*
943 	 * Extract firewall instructions from the fragment structure.
944 	 */
945 	*divinfo = fp->ipq_div_info;
946 	*divcookie = fp->ipq_div_cookie;
947 #endif
948 
949 	/*
950 	 * Create header for new ip packet by
951 	 * modifying header of first packet;
952 	 * dequeue and discard fragment reassembly header.
953 	 * Make header visible.
954 	 */
955 	ip->ip_len = next;
956 	ip->ip_src = fp->ipq_src;
957 	ip->ip_dst = fp->ipq_dst;
958 	remque(fp);
959 	nipq--;
960 	(void) m_free(dtom(fp));
961 	m->m_len += (IP_VHL_HL(ip->ip_vhl) << 2);
962 	m->m_data -= (IP_VHL_HL(ip->ip_vhl) << 2);
963 	/* some debugging cruft by sklower, below, will go away soon */
964 	if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */
965 		register int plen = 0;
966 		for (t = m; t; t = t->m_next)
967 			plen += t->m_len;
968 		m->m_pkthdr.len = plen;
969 	}
970 	return (m);
971 
972 dropfrag:
973 #ifdef IPDIVERT
974 	*divinfo = 0;
975 	*divcookie = 0;
976 #endif
977 	ipstat.ips_fragdropped++;
978 	m_freem(m);
979 	return (0);
980 
981 #undef GETIP
982 }
983 
984 /*
985  * Free a fragment reassembly header and all
986  * associated datagrams.
987  */
988 static void
989 ip_freef(fp)
990 	struct ipq *fp;
991 {
992 	register struct mbuf *q;
993 
994 	while (fp->ipq_frags) {
995 		q = fp->ipq_frags;
996 		fp->ipq_frags = q->m_nextpkt;
997 		m_freem(q);
998 	}
999 	remque(fp);
1000 	(void) m_free(dtom(fp));
1001 	nipq--;
1002 }
1003 
1004 /*
1005  * IP timer processing;
1006  * if a timer expires on a reassembly
1007  * queue, discard it.
1008  */
1009 void
1010 ip_slowtimo()
1011 {
1012 	register struct ipq *fp;
1013 	int s = splnet();
1014 	int i;
1015 
1016 	for (i = 0; i < IPREASS_NHASH; i++) {
1017 		fp = ipq[i].next;
1018 		if (fp == 0)
1019 			continue;
1020 		while (fp != &ipq[i]) {
1021 			--fp->ipq_ttl;
1022 			fp = fp->next;
1023 			if (fp->prev->ipq_ttl == 0) {
1024 				ipstat.ips_fragtimeout++;
1025 				ip_freef(fp->prev);
1026 			}
1027 		}
1028 	}
1029 	ipflow_slowtimo();
1030 	splx(s);
1031 }
1032 
1033 /*
1034  * Drain off all datagram fragments.
1035  */
1036 void
1037 ip_drain()
1038 {
1039 	int     i;
1040 
1041 	for (i = 0; i < IPREASS_NHASH; i++) {
1042 		while (ipq[i].next != &ipq[i]) {
1043 			ipstat.ips_fragdropped++;
1044 			ip_freef(ipq[i].next);
1045 		}
1046 	}
1047 	in_rtqdrain();
1048 }
1049 
1050 /*
1051  * Do option processing on a datagram,
1052  * possibly discarding it if bad options are encountered,
1053  * or forwarding it if source-routed.
1054  * Returns 1 if packet has been forwarded/freed,
1055  * 0 if the packet should be processed further.
1056  */
1057 static int
1058 ip_dooptions(m)
1059 	struct mbuf *m;
1060 {
1061 	register struct ip *ip = mtod(m, struct ip *);
1062 	register u_char *cp;
1063 	register struct ip_timestamp *ipt;
1064 	register struct in_ifaddr *ia;
1065 	int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0;
1066 	struct in_addr *sin, dst;
1067 	n_time ntime;
1068 
1069 	dst = ip->ip_dst;
1070 	cp = (u_char *)(ip + 1);
1071 	cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip);
1072 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1073 		opt = cp[IPOPT_OPTVAL];
1074 		if (opt == IPOPT_EOL)
1075 			break;
1076 		if (opt == IPOPT_NOP)
1077 			optlen = 1;
1078 		else {
1079 			if (cnt < IPOPT_OLEN + sizeof(*cp)) {
1080 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1081 				goto bad;
1082 			}
1083 			optlen = cp[IPOPT_OLEN];
1084 			if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
1085 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1086 				goto bad;
1087 			}
1088 		}
1089 		switch (opt) {
1090 
1091 		default:
1092 			break;
1093 
1094 		/*
1095 		 * Source routing with record.
1096 		 * Find interface with current destination address.
1097 		 * If none on this machine then drop if strictly routed,
1098 		 * or do nothing if loosely routed.
1099 		 * Record interface address and bring up next address
1100 		 * component.  If strictly routed make sure next
1101 		 * address is on directly accessible net.
1102 		 */
1103 		case IPOPT_LSRR:
1104 		case IPOPT_SSRR:
1105 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1106 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1107 				goto bad;
1108 			}
1109 			ipaddr.sin_addr = ip->ip_dst;
1110 			ia = (struct in_ifaddr *)
1111 				ifa_ifwithaddr((struct sockaddr *)&ipaddr);
1112 			if (ia == 0) {
1113 				if (opt == IPOPT_SSRR) {
1114 					type = ICMP_UNREACH;
1115 					code = ICMP_UNREACH_SRCFAIL;
1116 					goto bad;
1117 				}
1118 				if (!ip_dosourceroute)
1119 					goto nosourcerouting;
1120 				/*
1121 				 * Loose routing, and not at next destination
1122 				 * yet; nothing to do except forward.
1123 				 */
1124 				break;
1125 			}
1126 			off--;			/* 0 origin */
1127 			if (off > optlen - (int)sizeof(struct in_addr)) {
1128 				/*
1129 				 * End of source route.  Should be for us.
1130 				 */
1131 				if (!ip_acceptsourceroute)
1132 					goto nosourcerouting;
1133 				save_rte(cp, ip->ip_src);
1134 				break;
1135 			}
1136 
1137 			if (!ip_dosourceroute) {
1138 				if (ipforwarding) {
1139 					char buf[16]; /* aaa.bbb.ccc.ddd\0 */
1140 					/*
1141 					 * Acting as a router, so generate ICMP
1142 					 */
1143 nosourcerouting:
1144 					strcpy(buf, inet_ntoa(ip->ip_dst));
1145 					log(LOG_WARNING,
1146 					    "attempted source route from %s to %s\n",
1147 					    inet_ntoa(ip->ip_src), buf);
1148 					type = ICMP_UNREACH;
1149 					code = ICMP_UNREACH_SRCFAIL;
1150 					goto bad;
1151 				} else {
1152 					/*
1153 					 * Not acting as a router, so silently drop.
1154 					 */
1155 					ipstat.ips_cantforward++;
1156 					m_freem(m);
1157 					return (1);
1158 				}
1159 			}
1160 
1161 			/*
1162 			 * locate outgoing interface
1163 			 */
1164 			(void)memcpy(&ipaddr.sin_addr, cp + off,
1165 			    sizeof(ipaddr.sin_addr));
1166 
1167 			if (opt == IPOPT_SSRR) {
1168 #define	INA	struct in_ifaddr *
1169 #define	SA	struct sockaddr *
1170 			    if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == 0)
1171 				ia = (INA)ifa_ifwithnet((SA)&ipaddr);
1172 			} else
1173 				ia = ip_rtaddr(ipaddr.sin_addr);
1174 			if (ia == 0) {
1175 				type = ICMP_UNREACH;
1176 				code = ICMP_UNREACH_SRCFAIL;
1177 				goto bad;
1178 			}
1179 			ip->ip_dst = ipaddr.sin_addr;
1180 			(void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
1181 			    sizeof(struct in_addr));
1182 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1183 			/*
1184 			 * Let ip_intr's mcast routing check handle mcast pkts
1185 			 */
1186 			forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr));
1187 			break;
1188 
1189 		case IPOPT_RR:
1190 			if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1191 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1192 				goto bad;
1193 			}
1194 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1195 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1196 				goto bad;
1197 			}
1198 			/*
1199 			 * If no space remains, ignore.
1200 			 */
1201 			off--;			/* 0 origin */
1202 			if (off > optlen - (int)sizeof(struct in_addr))
1203 				break;
1204 			(void)memcpy(&ipaddr.sin_addr, &ip->ip_dst,
1205 			    sizeof(ipaddr.sin_addr));
1206 			/*
1207 			 * locate outgoing interface; if we're the destination,
1208 			 * use the incoming interface (should be same).
1209 			 */
1210 			if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == 0 &&
1211 			    (ia = ip_rtaddr(ipaddr.sin_addr)) == 0) {
1212 				type = ICMP_UNREACH;
1213 				code = ICMP_UNREACH_HOST;
1214 				goto bad;
1215 			}
1216 			(void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
1217 			    sizeof(struct in_addr));
1218 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1219 			break;
1220 
1221 		case IPOPT_TS:
1222 			code = cp - (u_char *)ip;
1223 			ipt = (struct ip_timestamp *)cp;
1224 			if (ipt->ipt_len < 5)
1225 				goto bad;
1226 			if (ipt->ipt_ptr >
1227 			    ipt->ipt_len - (int)sizeof(int32_t)) {
1228 				if (++ipt->ipt_oflw == 0)
1229 					goto bad;
1230 				break;
1231 			}
1232 			sin = (struct in_addr *)(cp + ipt->ipt_ptr - 1);
1233 			switch (ipt->ipt_flg) {
1234 
1235 			case IPOPT_TS_TSONLY:
1236 				break;
1237 
1238 			case IPOPT_TS_TSANDADDR:
1239 				if (ipt->ipt_ptr - 1 + sizeof(n_time) +
1240 				    sizeof(struct in_addr) > ipt->ipt_len)
1241 					goto bad;
1242 				ipaddr.sin_addr = dst;
1243 				ia = (INA)ifaof_ifpforaddr((SA)&ipaddr,
1244 							    m->m_pkthdr.rcvif);
1245 				if (ia == 0)
1246 					continue;
1247 				(void)memcpy(sin, &IA_SIN(ia)->sin_addr,
1248 				    sizeof(struct in_addr));
1249 				ipt->ipt_ptr += sizeof(struct in_addr);
1250 				break;
1251 
1252 			case IPOPT_TS_PRESPEC:
1253 				if (ipt->ipt_ptr - 1 + sizeof(n_time) +
1254 				    sizeof(struct in_addr) > ipt->ipt_len)
1255 					goto bad;
1256 				(void)memcpy(&ipaddr.sin_addr, sin,
1257 				    sizeof(struct in_addr));
1258 				if (ifa_ifwithaddr((SA)&ipaddr) == 0)
1259 					continue;
1260 				ipt->ipt_ptr += sizeof(struct in_addr);
1261 				break;
1262 
1263 			default:
1264 				goto bad;
1265 			}
1266 			ntime = iptime();
1267 			(void)memcpy(cp + ipt->ipt_ptr - 1, &ntime,
1268 			    sizeof(n_time));
1269 			ipt->ipt_ptr += sizeof(n_time);
1270 		}
1271 	}
1272 	if (forward && ipforwarding) {
1273 		ip_forward(m, 1);
1274 		return (1);
1275 	}
1276 	return (0);
1277 bad:
1278 	icmp_error(m, type, code, 0, 0);
1279 	ipstat.ips_badoptions++;
1280 	return (1);
1281 }
1282 
1283 /*
1284  * Given address of next destination (final or next hop),
1285  * return internet address info of interface to be used to get there.
1286  */
1287 static struct in_ifaddr *
1288 ip_rtaddr(dst)
1289 	 struct in_addr dst;
1290 {
1291 	register struct sockaddr_in *sin;
1292 
1293 	sin = (struct sockaddr_in *) &ipforward_rt.ro_dst;
1294 
1295 	if (ipforward_rt.ro_rt == 0 || dst.s_addr != sin->sin_addr.s_addr) {
1296 		if (ipforward_rt.ro_rt) {
1297 			RTFREE(ipforward_rt.ro_rt);
1298 			ipforward_rt.ro_rt = 0;
1299 		}
1300 		sin->sin_family = AF_INET;
1301 		sin->sin_len = sizeof(*sin);
1302 		sin->sin_addr = dst;
1303 
1304 		rtalloc_ign(&ipforward_rt, RTF_PRCLONING);
1305 	}
1306 	if (ipforward_rt.ro_rt == 0)
1307 		return ((struct in_ifaddr *)0);
1308 	return ((struct in_ifaddr *) ipforward_rt.ro_rt->rt_ifa);
1309 }
1310 
1311 /*
1312  * Save incoming source route for use in replies,
1313  * to be picked up later by ip_srcroute if the receiver is interested.
1314  */
1315 void
1316 save_rte(option, dst)
1317 	u_char *option;
1318 	struct in_addr dst;
1319 {
1320 	unsigned olen;
1321 
1322 	olen = option[IPOPT_OLEN];
1323 #ifdef DIAGNOSTIC
1324 	if (ipprintfs)
1325 		printf("save_rte: olen %d\n", olen);
1326 #endif
1327 	if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst)))
1328 		return;
1329 	bcopy(option, ip_srcrt.srcopt, olen);
1330 	ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1331 	ip_srcrt.dst = dst;
1332 }
1333 
1334 /*
1335  * Retrieve incoming source route for use in replies,
1336  * in the same form used by setsockopt.
1337  * The first hop is placed before the options, will be removed later.
1338  */
1339 struct mbuf *
1340 ip_srcroute()
1341 {
1342 	register struct in_addr *p, *q;
1343 	register struct mbuf *m;
1344 
1345 	if (ip_nhops == 0)
1346 		return ((struct mbuf *)0);
1347 	m = m_get(M_DONTWAIT, MT_HEADER);
1348 	if (m == 0)
1349 		return ((struct mbuf *)0);
1350 
1351 #define OPTSIZ	(sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt))
1352 
1353 	/* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
1354 	m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) +
1355 	    OPTSIZ;
1356 #ifdef DIAGNOSTIC
1357 	if (ipprintfs)
1358 		printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len);
1359 #endif
1360 
1361 	/*
1362 	 * First save first hop for return route
1363 	 */
1364 	p = &ip_srcrt.route[ip_nhops - 1];
1365 	*(mtod(m, struct in_addr *)) = *p--;
1366 #ifdef DIAGNOSTIC
1367 	if (ipprintfs)
1368 		printf(" hops %lx", (u_long)ntohl(mtod(m, struct in_addr *)->s_addr));
1369 #endif
1370 
1371 	/*
1372 	 * Copy option fields and padding (nop) to mbuf.
1373 	 */
1374 	ip_srcrt.nop = IPOPT_NOP;
1375 	ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
1376 	(void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr),
1377 	    &ip_srcrt.nop, OPTSIZ);
1378 	q = (struct in_addr *)(mtod(m, caddr_t) +
1379 	    sizeof(struct in_addr) + OPTSIZ);
1380 #undef OPTSIZ
1381 	/*
1382 	 * Record return path as an IP source route,
1383 	 * reversing the path (pointers are now aligned).
1384 	 */
1385 	while (p >= ip_srcrt.route) {
1386 #ifdef DIAGNOSTIC
1387 		if (ipprintfs)
1388 			printf(" %lx", (u_long)ntohl(q->s_addr));
1389 #endif
1390 		*q++ = *p--;
1391 	}
1392 	/*
1393 	 * Last hop goes to final destination.
1394 	 */
1395 	*q = ip_srcrt.dst;
1396 #ifdef DIAGNOSTIC
1397 	if (ipprintfs)
1398 		printf(" %lx\n", (u_long)ntohl(q->s_addr));
1399 #endif
1400 	return (m);
1401 }
1402 
1403 /*
1404  * Strip out IP options, at higher
1405  * level protocol in the kernel.
1406  * Second argument is buffer to which options
1407  * will be moved, and return value is their length.
1408  * XXX should be deleted; last arg currently ignored.
1409  */
1410 void
1411 ip_stripoptions(m, mopt)
1412 	register struct mbuf *m;
1413 	struct mbuf *mopt;
1414 {
1415 	register int i;
1416 	struct ip *ip = mtod(m, struct ip *);
1417 	register caddr_t opts;
1418 	int olen;
1419 
1420 	olen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip);
1421 	opts = (caddr_t)(ip + 1);
1422 	i = m->m_len - (sizeof (struct ip) + olen);
1423 	bcopy(opts + olen, opts, (unsigned)i);
1424 	m->m_len -= olen;
1425 	if (m->m_flags & M_PKTHDR)
1426 		m->m_pkthdr.len -= olen;
1427 	ip->ip_vhl = IP_MAKE_VHL(IPVERSION, sizeof(struct ip) >> 2);
1428 }
1429 
1430 u_char inetctlerrmap[PRC_NCMDS] = {
1431 	0,		0,		0,		0,
1432 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1433 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1434 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1435 	0,		0,		0,		0,
1436 	ENOPROTOOPT,	ENETRESET
1437 };
1438 
1439 /*
1440  * Forward a packet.  If some error occurs return the sender
1441  * an icmp packet.  Note we can't always generate a meaningful
1442  * icmp message because icmp doesn't have a large enough repertoire
1443  * of codes and types.
1444  *
1445  * If not forwarding, just drop the packet.  This could be confusing
1446  * if ipforwarding was zero but some routing protocol was advancing
1447  * us as a gateway to somewhere.  However, we must let the routing
1448  * protocol deal with that.
1449  *
1450  * The srcrt parameter indicates whether the packet is being forwarded
1451  * via a source route.
1452  */
1453 static void
1454 ip_forward(m, srcrt)
1455 	struct mbuf *m;
1456 	int srcrt;
1457 {
1458 	register struct ip *ip = mtod(m, struct ip *);
1459 	register struct sockaddr_in *sin;
1460 	register struct rtentry *rt;
1461 	int error, type = 0, code = 0;
1462 	struct mbuf *mcopy;
1463 	n_long dest;
1464 	struct ifnet *destifp;
1465 #ifdef IPSEC
1466 	struct ifnet dummyifp;
1467 #endif
1468 
1469 	dest = 0;
1470 #ifdef DIAGNOSTIC
1471 	if (ipprintfs)
1472 		printf("forward: src %lx dst %lx ttl %x\n",
1473 		    (u_long)ip->ip_src.s_addr, (u_long)ip->ip_dst.s_addr,
1474 		    ip->ip_ttl);
1475 #endif
1476 
1477 
1478 	if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
1479 		ipstat.ips_cantforward++;
1480 		m_freem(m);
1481 		return;
1482 	}
1483 #ifdef IPSTEALTH
1484 	if (!ipstealth) {
1485 #endif
1486 		if (ip->ip_ttl <= IPTTLDEC) {
1487 			icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS,
1488 			    dest, 0);
1489 			return;
1490 		}
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 	if (mcopy && (mcopy->m_flags & M_EXT))
1520 		m_copydata(mcopy, 0, sizeof(struct ip), mtod(mcopy, caddr_t));
1521 
1522 #ifdef IPSTEALTH
1523 	if (!ipstealth) {
1524 #endif
1525 		ip->ip_ttl -= IPTTLDEC;
1526 #ifdef IPSTEALTH
1527 	}
1528 #endif
1529 
1530 	/*
1531 	 * If forwarding packet using same interface that it came in on,
1532 	 * perhaps should send a redirect to sender to shortcut a hop.
1533 	 * Only send redirect if source is sending directly to us,
1534 	 * and if packet was not source routed (or has any options).
1535 	 * Also, don't send redirect if forwarding using a default route
1536 	 * or a route modified by a redirect.
1537 	 */
1538 #define	satosin(sa)	((struct sockaddr_in *)(sa))
1539 	if (rt->rt_ifp == m->m_pkthdr.rcvif &&
1540 	    (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1541 	    satosin(rt_key(rt))->sin_addr.s_addr != 0 &&
1542 	    ipsendredirects && !srcrt) {
1543 #define	RTA(rt)	((struct in_ifaddr *)(rt->rt_ifa))
1544 		u_long src = ntohl(ip->ip_src.s_addr);
1545 
1546 		if (RTA(rt) &&
1547 		    (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1548 		    if (rt->rt_flags & RTF_GATEWAY)
1549 			dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
1550 		    else
1551 			dest = ip->ip_dst.s_addr;
1552 		    /* Router requirements says to only send host redirects */
1553 		    type = ICMP_REDIRECT;
1554 		    code = ICMP_REDIRECT_HOST;
1555 #ifdef DIAGNOSTIC
1556 		    if (ipprintfs)
1557 		        printf("redirect (%d) to %lx\n", code, (u_long)dest);
1558 #endif
1559 		}
1560 	}
1561 
1562 	error = ip_output(m, (struct mbuf *)0, &ipforward_rt,
1563 			  IP_FORWARDING, 0);
1564 	if (error)
1565 		ipstat.ips_cantforward++;
1566 	else {
1567 		ipstat.ips_forward++;
1568 		if (type)
1569 			ipstat.ips_redirectsent++;
1570 		else {
1571 			if (mcopy) {
1572 				ipflow_create(&ipforward_rt, mcopy);
1573 				m_freem(mcopy);
1574 			}
1575 			return;
1576 		}
1577 	}
1578 	if (mcopy == NULL)
1579 		return;
1580 	destifp = NULL;
1581 
1582 	switch (error) {
1583 
1584 	case 0:				/* forwarded, but need redirect */
1585 		/* type, code set above */
1586 		break;
1587 
1588 	case ENETUNREACH:		/* shouldn't happen, checked above */
1589 	case EHOSTUNREACH:
1590 	case ENETDOWN:
1591 	case EHOSTDOWN:
1592 	default:
1593 		type = ICMP_UNREACH;
1594 		code = ICMP_UNREACH_HOST;
1595 		break;
1596 
1597 	case EMSGSIZE:
1598 		type = ICMP_UNREACH;
1599 		code = ICMP_UNREACH_NEEDFRAG;
1600 #ifndef IPSEC
1601 		if (ipforward_rt.ro_rt)
1602 			destifp = ipforward_rt.ro_rt->rt_ifp;
1603 #else
1604 		/*
1605 		 * If the packet is routed over IPsec tunnel, tell the
1606 		 * originator the tunnel MTU.
1607 		 *	tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
1608 		 * XXX quickhack!!!
1609 		 */
1610 		if (ipforward_rt.ro_rt) {
1611 			struct secpolicy *sp = NULL;
1612 			int ipsecerror;
1613 			int ipsechdr;
1614 			struct route *ro;
1615 
1616 			sp = ipsec4_getpolicybyaddr(mcopy,
1617 						    IPSEC_DIR_OUTBOUND,
1618 			                            IP_FORWARDING,
1619 			                            &ipsecerror);
1620 
1621 			if (sp == NULL)
1622 				destifp = ipforward_rt.ro_rt->rt_ifp;
1623 			else {
1624 				/* count IPsec header size */
1625 				ipsechdr = ipsec4_hdrsiz(mcopy,
1626 							 IPSEC_DIR_OUTBOUND,
1627 							 NULL);
1628 
1629 				/*
1630 				 * find the correct route for outer IPv4
1631 				 * header, compute tunnel MTU.
1632 				 *
1633 				 * XXX BUG ALERT
1634 				 * The "dummyifp" code relies upon the fact
1635 				 * that icmp_error() touches only ifp->if_mtu.
1636 				 */
1637 				/*XXX*/
1638 				destifp = NULL;
1639 				if (sp->req != NULL
1640 				 && sp->req->sav != NULL
1641 				 && sp->req->sav->sah != NULL) {
1642 					ro = &sp->req->sav->sah->sa_route;
1643 					if (ro->ro_rt && ro->ro_rt->rt_ifp) {
1644 						dummyifp.if_mtu =
1645 						    ro->ro_rt->rt_ifp->if_mtu;
1646 						dummyifp.if_mtu -= ipsechdr;
1647 						destifp = &dummyifp;
1648 					}
1649 				}
1650 
1651 				key_freesp(sp);
1652 			}
1653 		}
1654 #endif /*IPSEC*/
1655 		ipstat.ips_cantfrag++;
1656 		break;
1657 
1658 	case ENOBUFS:
1659 		type = ICMP_SOURCEQUENCH;
1660 		code = 0;
1661 		break;
1662 
1663 	case EACCES:			/* ipfw denied packet */
1664 		m_freem(mcopy);
1665 		return;
1666 	}
1667 	if (mcopy->m_flags & M_EXT)
1668 		m_copyback(mcopy, 0, sizeof(struct ip), mtod(mcopy, caddr_t));
1669 	icmp_error(mcopy, type, code, dest, destifp);
1670 }
1671 
1672 void
1673 ip_savecontrol(inp, mp, ip, m)
1674 	register struct inpcb *inp;
1675 	register struct mbuf **mp;
1676 	register struct ip *ip;
1677 	register struct mbuf *m;
1678 {
1679 	if (inp->inp_socket->so_options & SO_TIMESTAMP) {
1680 		struct timeval tv;
1681 
1682 		microtime(&tv);
1683 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1684 			SCM_TIMESTAMP, SOL_SOCKET);
1685 		if (*mp)
1686 			mp = &(*mp)->m_next;
1687 	}
1688 	if (inp->inp_flags & INP_RECVDSTADDR) {
1689 		*mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
1690 		    sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1691 		if (*mp)
1692 			mp = &(*mp)->m_next;
1693 	}
1694 #ifdef notyet
1695 	/* XXX
1696 	 * Moving these out of udp_input() made them even more broken
1697 	 * than they already were.
1698 	 */
1699 	/* options were tossed already */
1700 	if (inp->inp_flags & INP_RECVOPTS) {
1701 		*mp = sbcreatecontrol((caddr_t) opts_deleted_above,
1702 		    sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1703 		if (*mp)
1704 			mp = &(*mp)->m_next;
1705 	}
1706 	/* ip_srcroute doesn't do what we want here, need to fix */
1707 	if (inp->inp_flags & INP_RECVRETOPTS) {
1708 		*mp = sbcreatecontrol((caddr_t) ip_srcroute(),
1709 		    sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1710 		if (*mp)
1711 			mp = &(*mp)->m_next;
1712 	}
1713 #endif
1714 	if (inp->inp_flags & INP_RECVIF) {
1715 		struct ifnet *ifp;
1716 		struct sdlbuf {
1717 			struct sockaddr_dl sdl;
1718 			u_char	pad[32];
1719 		} sdlbuf;
1720 		struct sockaddr_dl *sdp;
1721 		struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1722 
1723 		if (((ifp = m->m_pkthdr.rcvif))
1724 		&& ( ifp->if_index && (ifp->if_index <= if_index))) {
1725 			sdp = (struct sockaddr_dl *)(ifnet_addrs
1726 					[ifp->if_index - 1]->ifa_addr);
1727 			/*
1728 			 * Change our mind and don't try copy.
1729 			 */
1730 			if ((sdp->sdl_family != AF_LINK)
1731 			|| (sdp->sdl_len > sizeof(sdlbuf))) {
1732 				goto makedummy;
1733 			}
1734 			bcopy(sdp, sdl2, sdp->sdl_len);
1735 		} else {
1736 makedummy:
1737 			sdl2->sdl_len
1738 				= offsetof(struct sockaddr_dl, sdl_data[0]);
1739 			sdl2->sdl_family = AF_LINK;
1740 			sdl2->sdl_index = 0;
1741 			sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1742 		}
1743 		*mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len,
1744 			IP_RECVIF, IPPROTO_IP);
1745 		if (*mp)
1746 			mp = &(*mp)->m_next;
1747 	}
1748 }
1749 
1750 int
1751 ip_rsvp_init(struct socket *so)
1752 {
1753 	if (so->so_type != SOCK_RAW ||
1754 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
1755 	  return EOPNOTSUPP;
1756 
1757 	if (ip_rsvpd != NULL)
1758 	  return EADDRINUSE;
1759 
1760 	ip_rsvpd = so;
1761 	/*
1762 	 * This may seem silly, but we need to be sure we don't over-increment
1763 	 * the RSVP counter, in case something slips up.
1764 	 */
1765 	if (!ip_rsvp_on) {
1766 		ip_rsvp_on = 1;
1767 		rsvp_on++;
1768 	}
1769 
1770 	return 0;
1771 }
1772 
1773 int
1774 ip_rsvp_done(void)
1775 {
1776 	ip_rsvpd = NULL;
1777 	/*
1778 	 * This may seem silly, but we need to be sure we don't over-decrement
1779 	 * the RSVP counter, in case something slips up.
1780 	 */
1781 	if (ip_rsvp_on) {
1782 		ip_rsvp_on = 0;
1783 		rsvp_on--;
1784 	}
1785 	return 0;
1786 }
1787