xref: /freebsd/sys/netinet/ip_input.c (revision 0c43d89a0d8e976ca494d4837f4c1f3734d2c300)
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  * $Id: ip_input.c,v 1.3 1994/08/02 07:48:38 davidg Exp $
35  */
36 
37 #include <sys/param.h>
38 #include <sys/systm.h>
39 #include <sys/malloc.h>
40 #include <sys/mbuf.h>
41 #include <sys/domain.h>
42 #include <sys/protosw.h>
43 #include <sys/socket.h>
44 #include <sys/errno.h>
45 #include <sys/time.h>
46 #include <sys/kernel.h>
47 
48 #include <net/if.h>
49 #include <net/route.h>
50 
51 #include <netinet/in.h>
52 #include <netinet/in_systm.h>
53 #include <netinet/ip.h>
54 #include <netinet/in_pcb.h>
55 #include <netinet/in_var.h>
56 #include <netinet/ip_var.h>
57 #include <netinet/ip_icmp.h>
58 
59 #ifndef	IPFORWARDING
60 #ifdef GATEWAY
61 #define	IPFORWARDING	1	/* forward IP packets not for us */
62 #else /* GATEWAY */
63 #define	IPFORWARDING	0	/* don't forward IP packets not for us */
64 #endif /* GATEWAY */
65 #endif /* IPFORWARDING */
66 #ifndef	IPSENDREDIRECTS
67 #define	IPSENDREDIRECTS	1
68 #endif
69 int	ipforwarding = IPFORWARDING;
70 int	ipsendredirects = IPSENDREDIRECTS;
71 int	ip_defttl = IPDEFTTL;
72 #ifdef DIAGNOSTIC
73 int	ipprintfs = 0;
74 #endif
75 
76 extern	struct domain inetdomain;
77 extern	struct protosw inetsw[];
78 u_char	ip_protox[IPPROTO_MAX];
79 int	ipqmaxlen = IFQ_MAXLEN;
80 struct	in_ifaddr *in_ifaddr;			/* first inet address */
81 struct	ifqueue ipintrq;
82 
83 struct ipstat ipstat;
84 struct ipq ipq;
85 
86 /*
87  * We need to save the IP options in case a protocol wants to respond
88  * to an incoming packet over the same route if the packet got here
89  * using IP source routing.  This allows connection establishment and
90  * maintenance when the remote end is on a network that is not known
91  * to us.
92  */
93 int	ip_nhops = 0;
94 static	struct ip_srcrt {
95 	struct	in_addr dst;			/* final destination */
96 	char	nop;				/* one NOP to align */
97 	char	srcopt[IPOPT_OFFSET + 1];	/* OPTVAL, OLEN and OFFSET */
98 	struct	in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)];
99 } ip_srcrt;
100 
101 #ifdef GATEWAY
102 extern	int if_index;
103 u_long	*ip_ifmatrix;
104 #endif
105 
106 static void save_rte __P((u_char *, struct in_addr));
107 /*
108  * IP initialization: fill in IP protocol switch table.
109  * All protocols not implemented in kernel go to raw IP protocol handler.
110  */
111 void
112 ip_init()
113 {
114 	register struct protosw *pr;
115 	register int i;
116 
117 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
118 	if (pr == 0)
119 		panic("ip_init");
120 	for (i = 0; i < IPPROTO_MAX; i++)
121 		ip_protox[i] = pr - inetsw;
122 	for (pr = inetdomain.dom_protosw;
123 	    pr < inetdomain.dom_protoswNPROTOSW; pr++)
124 		if (pr->pr_domain->dom_family == PF_INET &&
125 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
126 			ip_protox[pr->pr_protocol] = pr - inetsw;
127 	ipq.next = ipq.prev = &ipq;
128 	ip_id = time.tv_sec & 0xffff;
129 	ipintrq.ifq_maxlen = ipqmaxlen;
130 #ifdef GATEWAY
131 	i = (if_index + 1) * (if_index + 1) * sizeof (u_long);
132 	ip_ifmatrix = (u_long *) malloc(i, M_RTABLE, M_WAITOK);
133 	bzero((char *)ip_ifmatrix, i);
134 #endif
135 }
136 
137 struct	sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET };
138 struct	route ipforward_rt;
139 
140 /*
141  * Ip input routine.  Checksum and byte swap header.  If fragmented
142  * try to reassemble.  Process options.  Pass to next level.
143  */
144 void
145 ipintr()
146 {
147 	register struct ip *ip;
148 	register struct mbuf *m;
149 	register struct ipq *fp;
150 	register struct in_ifaddr *ia;
151 	int hlen, s;
152 
153 next:
154 	/*
155 	 * Get next datagram off input queue and get IP header
156 	 * in first mbuf.
157 	 */
158 	s = splimp();
159 	IF_DEQUEUE(&ipintrq, m);
160 	splx(s);
161 	if (m == 0)
162 		return;
163 #ifdef	DIAGNOSTIC
164 	if ((m->m_flags & M_PKTHDR) == 0)
165 		panic("ipintr no HDR");
166 #endif
167 	/*
168 	 * If no IP addresses have been set yet but the interfaces
169 	 * are receiving, can't do anything with incoming packets yet.
170 	 */
171 	if (in_ifaddr == NULL)
172 		goto bad;
173 	ipstat.ips_total++;
174 	if (m->m_len < sizeof (struct ip) &&
175 	    (m = m_pullup(m, sizeof (struct ip))) == 0) {
176 		ipstat.ips_toosmall++;
177 		goto next;
178 	}
179 	ip = mtod(m, struct ip *);
180 	if (ip->ip_v != IPVERSION) {
181 		ipstat.ips_badvers++;
182 		goto bad;
183 	}
184 	hlen = ip->ip_hl << 2;
185 	if (hlen < sizeof(struct ip)) {	/* minimum header length */
186 		ipstat.ips_badhlen++;
187 		goto bad;
188 	}
189 	if (hlen > m->m_len) {
190 		if ((m = m_pullup(m, hlen)) == 0) {
191 			ipstat.ips_badhlen++;
192 			goto next;
193 		}
194 		ip = mtod(m, struct ip *);
195 	}
196 	if (ip->ip_sum = in_cksum(m, hlen)) {
197 		ipstat.ips_badsum++;
198 		goto bad;
199 	}
200 
201 	/*
202 	 * Convert fields to host representation.
203 	 */
204 	NTOHS(ip->ip_len);
205 	if (ip->ip_len < hlen) {
206 		ipstat.ips_badlen++;
207 		goto bad;
208 	}
209 	NTOHS(ip->ip_id);
210 	NTOHS(ip->ip_off);
211 
212 	/*
213 	 * Check that the amount of data in the buffers
214 	 * is as at least much as the IP header would have us expect.
215 	 * Trim mbufs if longer than we expect.
216 	 * Drop packet if shorter than we expect.
217 	 */
218 	if (m->m_pkthdr.len < ip->ip_len) {
219 		ipstat.ips_tooshort++;
220 		goto bad;
221 	}
222 	if (m->m_pkthdr.len > ip->ip_len) {
223 		if (m->m_len == m->m_pkthdr.len) {
224 			m->m_len = ip->ip_len;
225 			m->m_pkthdr.len = ip->ip_len;
226 		} else
227 			m_adj(m, ip->ip_len - m->m_pkthdr.len);
228 	}
229 
230 	/*
231 	 * Process options and, if not destined for us,
232 	 * ship it on.  ip_dooptions returns 1 when an
233 	 * error was detected (causing an icmp message
234 	 * to be sent and the original packet to be freed).
235 	 */
236 	ip_nhops = 0;		/* for source routed packets */
237 	if (hlen > sizeof (struct ip) && ip_dooptions(m))
238 		goto next;
239 
240 	/*
241 	 * Check our list of addresses, to see if the packet is for us.
242 	 */
243 	for (ia = in_ifaddr; ia; ia = ia->ia_next) {
244 #define	satosin(sa)	((struct sockaddr_in *)(sa))
245 
246 		if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr)
247 			goto ours;
248 		if (
249 #ifdef	DIRECTED_BROADCAST
250 		    ia->ia_ifp == m->m_pkthdr.rcvif &&
251 #endif
252 		    (ia->ia_ifp->if_flags & IFF_BROADCAST)) {
253 			u_long t;
254 
255 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
256 			    ip->ip_dst.s_addr)
257 				goto ours;
258 			if (ip->ip_dst.s_addr == ia->ia_netbroadcast.s_addr)
259 				goto ours;
260 			/*
261 			 * Look for all-0's host part (old broadcast addr),
262 			 * either for subnet or net.
263 			 */
264 			t = ntohl(ip->ip_dst.s_addr);
265 			if (t == ia->ia_subnet)
266 				goto ours;
267 			if (t == ia->ia_net)
268 				goto ours;
269 		}
270 	}
271 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
272 		struct in_multi *inm;
273 #ifdef MROUTING
274 		extern struct socket *ip_mrouter;
275 
276 		if (ip_mrouter) {
277 			/*
278 			 * If we are acting as a multicast router, all
279 			 * incoming multicast packets are passed to the
280 			 * kernel-level multicast forwarding function.
281 			 * The packet is returned (relatively) intact; if
282 			 * ip_mforward() returns a non-zero value, the packet
283 			 * must be discarded, else it may be accepted below.
284 			 *
285 			 * (The IP ident field is put in the same byte order
286 			 * as expected when ip_mforward() is called from
287 			 * ip_output().)
288 			 */
289 			ip->ip_id = htons(ip->ip_id);
290 			if (ip_mforward(m, m->m_pkthdr.rcvif) != 0) {
291 				ipstat.ips_cantforward++;
292 				m_freem(m);
293 				goto next;
294 			}
295 			ip->ip_id = ntohs(ip->ip_id);
296 
297 			/*
298 			 * The process-level routing demon needs to receive
299 			 * all multicast IGMP packets, whether or not this
300 			 * host belongs to their destination groups.
301 			 */
302 			if (ip->ip_p == IPPROTO_IGMP)
303 				goto ours;
304 			ipstat.ips_forward++;
305 		}
306 #endif
307 		/*
308 		 * See if we belong to the destination multicast group on the
309 		 * arrival interface.
310 		 */
311 		IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
312 		if (inm == NULL) {
313 			ipstat.ips_cantforward++;
314 			m_freem(m);
315 			goto next;
316 		}
317 		goto ours;
318 	}
319 	if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
320 		goto ours;
321 	if (ip->ip_dst.s_addr == INADDR_ANY)
322 		goto ours;
323 
324 	/*
325 	 * Not for us; forward if possible and desirable.
326 	 */
327 	if (ipforwarding == 0) {
328 		ipstat.ips_cantforward++;
329 		m_freem(m);
330 	} else
331 		ip_forward(m, 0);
332 	goto next;
333 
334 ours:
335 	/*
336 	 * If offset or IP_MF are set, must reassemble.
337 	 * Otherwise, nothing need be done.
338 	 * (We could look in the reassembly queue to see
339 	 * if the packet was previously fragmented,
340 	 * but it's not worth the time; just let them time out.)
341 	 */
342 	if (ip->ip_off &~ IP_DF) {
343 		if (m->m_flags & M_EXT) {		/* XXX */
344 			if ((m = m_pullup(m, sizeof (struct ip))) == 0) {
345 				ipstat.ips_toosmall++;
346 				goto next;
347 			}
348 			ip = mtod(m, struct ip *);
349 		}
350 		/*
351 		 * Look for queue of fragments
352 		 * of this datagram.
353 		 */
354 		for (fp = ipq.next; fp != &ipq; fp = fp->next)
355 			if (ip->ip_id == fp->ipq_id &&
356 			    ip->ip_src.s_addr == fp->ipq_src.s_addr &&
357 			    ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
358 			    ip->ip_p == fp->ipq_p)
359 				goto found;
360 		fp = 0;
361 found:
362 
363 		/*
364 		 * Adjust ip_len to not reflect header,
365 		 * set ip_mff if more fragments are expected,
366 		 * convert offset of this to bytes.
367 		 */
368 		ip->ip_len -= hlen;
369 		((struct ipasfrag *)ip)->ipf_mff &= ~1;
370 		if (ip->ip_off & IP_MF)
371 			((struct ipasfrag *)ip)->ipf_mff |= 1;
372 		ip->ip_off <<= 3;
373 
374 		/*
375 		 * If datagram marked as having more fragments
376 		 * or if this is not the first fragment,
377 		 * attempt reassembly; if it succeeds, proceed.
378 		 */
379 		if (((struct ipasfrag *)ip)->ipf_mff & 1 || ip->ip_off) {
380 			ipstat.ips_fragments++;
381 			ip = ip_reass((struct ipasfrag *)ip, fp);
382 			if (ip == 0)
383 				goto next;
384 			ipstat.ips_reassembled++;
385 			m = dtom(ip);
386 		} else
387 			if (fp)
388 				ip_freef(fp);
389 	} else
390 		ip->ip_len -= hlen;
391 
392 	/*
393 	 * Switch out to protocol's input routine.
394 	 */
395 	ipstat.ips_delivered++;
396 	(*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen);
397 	goto next;
398 bad:
399 	m_freem(m);
400 	goto next;
401 }
402 
403 /*
404  * Take incoming datagram fragment and try to
405  * reassemble it into whole datagram.  If a chain for
406  * reassembly of this datagram already exists, then it
407  * is given as fp; otherwise have to make a chain.
408  */
409 struct ip *
410 ip_reass(ip, fp)
411 	register struct ipasfrag *ip;
412 	register struct ipq *fp;
413 {
414 	register struct mbuf *m = dtom(ip);
415 	register struct ipasfrag *q;
416 	struct mbuf *t;
417 	int hlen = ip->ip_hl << 2;
418 	int i, next;
419 
420 	/*
421 	 * Presence of header sizes in mbufs
422 	 * would confuse code below.
423 	 */
424 	m->m_data += hlen;
425 	m->m_len -= hlen;
426 
427 	/*
428 	 * If first fragment to arrive, create a reassembly queue.
429 	 */
430 	if (fp == 0) {
431 		if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL)
432 			goto dropfrag;
433 		fp = mtod(t, struct ipq *);
434 		insque(fp, &ipq);
435 		fp->ipq_ttl = IPFRAGTTL;
436 		fp->ipq_p = ip->ip_p;
437 		fp->ipq_id = ip->ip_id;
438 		fp->ipq_next = fp->ipq_prev = (struct ipasfrag *)fp;
439 		fp->ipq_src = ((struct ip *)ip)->ip_src;
440 		fp->ipq_dst = ((struct ip *)ip)->ip_dst;
441 		q = (struct ipasfrag *)fp;
442 		goto insert;
443 	}
444 
445 	/*
446 	 * Find a segment which begins after this one does.
447 	 */
448 	for (q = fp->ipq_next; q != (struct ipasfrag *)fp; q = q->ipf_next)
449 		if (q->ip_off > ip->ip_off)
450 			break;
451 
452 	/*
453 	 * If there is a preceding segment, it may provide some of
454 	 * our data already.  If so, drop the data from the incoming
455 	 * segment.  If it provides all of our data, drop us.
456 	 */
457 	if (q->ipf_prev != (struct ipasfrag *)fp) {
458 		i = q->ipf_prev->ip_off + q->ipf_prev->ip_len - ip->ip_off;
459 		if (i > 0) {
460 			if (i >= ip->ip_len)
461 				goto dropfrag;
462 			m_adj(dtom(ip), i);
463 			ip->ip_off += i;
464 			ip->ip_len -= i;
465 		}
466 	}
467 
468 	/*
469 	 * While we overlap succeeding segments trim them or,
470 	 * if they are completely covered, dequeue them.
471 	 */
472 	while (q != (struct ipasfrag *)fp && ip->ip_off + ip->ip_len > q->ip_off) {
473 		i = (ip->ip_off + ip->ip_len) - q->ip_off;
474 		if (i < q->ip_len) {
475 			q->ip_len -= i;
476 			q->ip_off += i;
477 			m_adj(dtom(q), i);
478 			break;
479 		}
480 		q = q->ipf_next;
481 		m_freem(dtom(q->ipf_prev));
482 		ip_deq(q->ipf_prev);
483 	}
484 
485 insert:
486 	/*
487 	 * Stick new segment in its place;
488 	 * check for complete reassembly.
489 	 */
490 	ip_enq(ip, q->ipf_prev);
491 	next = 0;
492 	for (q = fp->ipq_next; q != (struct ipasfrag *)fp; q = q->ipf_next) {
493 		if (q->ip_off != next)
494 			return (0);
495 		next += q->ip_len;
496 	}
497 	if (q->ipf_prev->ipf_mff & 1)
498 		return (0);
499 
500 	/*
501 	 * Reassembly is complete; concatenate fragments.
502 	 */
503 	q = fp->ipq_next;
504 	m = dtom(q);
505 	t = m->m_next;
506 	m->m_next = 0;
507 	m_cat(m, t);
508 	q = q->ipf_next;
509 	while (q != (struct ipasfrag *)fp) {
510 		t = dtom(q);
511 		q = q->ipf_next;
512 		m_cat(m, t);
513 	}
514 
515 	/*
516 	 * Create header for new ip packet by
517 	 * modifying header of first packet;
518 	 * dequeue and discard fragment reassembly header.
519 	 * Make header visible.
520 	 */
521 	ip = fp->ipq_next;
522 	ip->ip_len = next;
523 	ip->ipf_mff &= ~1;
524 	((struct ip *)ip)->ip_src = fp->ipq_src;
525 	((struct ip *)ip)->ip_dst = fp->ipq_dst;
526 	remque(fp);
527 	(void) m_free(dtom(fp));
528 	m = dtom(ip);
529 	m->m_len += (ip->ip_hl << 2);
530 	m->m_data -= (ip->ip_hl << 2);
531 	/* some debugging cruft by sklower, below, will go away soon */
532 	if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */
533 		register int plen = 0;
534 		for (t = m; m; m = m->m_next)
535 			plen += m->m_len;
536 		t->m_pkthdr.len = plen;
537 	}
538 	return ((struct ip *)ip);
539 
540 dropfrag:
541 	ipstat.ips_fragdropped++;
542 	m_freem(m);
543 	return (0);
544 }
545 
546 /*
547  * Free a fragment reassembly header and all
548  * associated datagrams.
549  */
550 void
551 ip_freef(fp)
552 	struct ipq *fp;
553 {
554 	register struct ipasfrag *q, *p;
555 
556 	for (q = fp->ipq_next; q != (struct ipasfrag *)fp; q = p) {
557 		p = q->ipf_next;
558 		ip_deq(q);
559 		m_freem(dtom(q));
560 	}
561 	remque(fp);
562 	(void) m_free(dtom(fp));
563 }
564 
565 /*
566  * Put an ip fragment on a reassembly chain.
567  * Like insque, but pointers in middle of structure.
568  */
569 void
570 ip_enq(p, prev)
571 	register struct ipasfrag *p, *prev;
572 {
573 
574 	p->ipf_prev = prev;
575 	p->ipf_next = prev->ipf_next;
576 	prev->ipf_next->ipf_prev = p;
577 	prev->ipf_next = p;
578 }
579 
580 /*
581  * To ip_enq as remque is to insque.
582  */
583 void
584 ip_deq(p)
585 	register struct ipasfrag *p;
586 {
587 
588 	p->ipf_prev->ipf_next = p->ipf_next;
589 	p->ipf_next->ipf_prev = p->ipf_prev;
590 }
591 
592 /*
593  * IP timer processing;
594  * if a timer expires on a reassembly
595  * queue, discard it.
596  */
597 void
598 ip_slowtimo()
599 {
600 	register struct ipq *fp;
601 	int s = splnet();
602 
603 	fp = ipq.next;
604 	if (fp == 0) {
605 		splx(s);
606 		return;
607 	}
608 	while (fp != &ipq) {
609 		--fp->ipq_ttl;
610 		fp = fp->next;
611 		if (fp->prev->ipq_ttl == 0) {
612 			ipstat.ips_fragtimeout++;
613 			ip_freef(fp->prev);
614 		}
615 	}
616 	splx(s);
617 }
618 
619 /*
620  * Drain off all datagram fragments.
621  */
622 void
623 ip_drain()
624 {
625 
626 	while (ipq.next != &ipq) {
627 		ipstat.ips_fragdropped++;
628 		ip_freef(ipq.next);
629 	}
630 }
631 
632 /*
633  * Do option processing on a datagram,
634  * possibly discarding it if bad options are encountered,
635  * or forwarding it if source-routed.
636  * Returns 1 if packet has been forwarded/freed,
637  * 0 if the packet should be processed further.
638  */
639 int
640 ip_dooptions(m)
641 	struct mbuf *m;
642 {
643 	register struct ip *ip = mtod(m, struct ip *);
644 	register u_char *cp;
645 	register struct ip_timestamp *ipt;
646 	register struct in_ifaddr *ia;
647 	int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0;
648 	struct in_addr *sin, dst;
649 	n_time ntime;
650 
651 	dst = ip->ip_dst;
652 	cp = (u_char *)(ip + 1);
653 	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
654 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
655 		opt = cp[IPOPT_OPTVAL];
656 		if (opt == IPOPT_EOL)
657 			break;
658 		if (opt == IPOPT_NOP)
659 			optlen = 1;
660 		else {
661 			optlen = cp[IPOPT_OLEN];
662 			if (optlen <= 0 || optlen > cnt) {
663 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
664 				goto bad;
665 			}
666 		}
667 		switch (opt) {
668 
669 		default:
670 			break;
671 
672 		/*
673 		 * Source routing with record.
674 		 * Find interface with current destination address.
675 		 * If none on this machine then drop if strictly routed,
676 		 * or do nothing if loosely routed.
677 		 * Record interface address and bring up next address
678 		 * component.  If strictly routed make sure next
679 		 * address is on directly accessible net.
680 		 */
681 		case IPOPT_LSRR:
682 		case IPOPT_SSRR:
683 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
684 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
685 				goto bad;
686 			}
687 			ipaddr.sin_addr = ip->ip_dst;
688 			ia = (struct in_ifaddr *)
689 				ifa_ifwithaddr((struct sockaddr *)&ipaddr);
690 			if (ia == 0) {
691 				if (opt == IPOPT_SSRR) {
692 					type = ICMP_UNREACH;
693 					code = ICMP_UNREACH_SRCFAIL;
694 					goto bad;
695 				}
696 				/*
697 				 * Loose routing, and not at next destination
698 				 * yet; nothing to do except forward.
699 				 */
700 				break;
701 			}
702 			off--;			/* 0 origin */
703 			if (off > optlen - sizeof(struct in_addr)) {
704 				/*
705 				 * End of source route.  Should be for us.
706 				 */
707 				save_rte(cp, ip->ip_src);
708 				break;
709 			}
710 			/*
711 			 * locate outgoing interface
712 			 */
713 			bcopy((caddr_t)(cp + off), (caddr_t)&ipaddr.sin_addr,
714 			    sizeof(ipaddr.sin_addr));
715 			if (opt == IPOPT_SSRR) {
716 #define	INA	struct in_ifaddr *
717 #define	SA	struct sockaddr *
718 			    if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == 0)
719 				ia = (INA)ifa_ifwithnet((SA)&ipaddr);
720 			} else
721 				ia = ip_rtaddr(ipaddr.sin_addr);
722 			if (ia == 0) {
723 				type = ICMP_UNREACH;
724 				code = ICMP_UNREACH_SRCFAIL;
725 				goto bad;
726 			}
727 			ip->ip_dst = ipaddr.sin_addr;
728 			bcopy((caddr_t)&(IA_SIN(ia)->sin_addr),
729 			    (caddr_t)(cp + off), sizeof(struct in_addr));
730 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
731 			/*
732 			 * Let ip_intr's mcast routing check handle mcast pkts
733 			 */
734 			forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr));
735 			break;
736 
737 		case IPOPT_RR:
738 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
739 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
740 				goto bad;
741 			}
742 			/*
743 			 * If no space remains, ignore.
744 			 */
745 			off--;			/* 0 origin */
746 			if (off > optlen - sizeof(struct in_addr))
747 				break;
748 			bcopy((caddr_t)(&ip->ip_dst), (caddr_t)&ipaddr.sin_addr,
749 			    sizeof(ipaddr.sin_addr));
750 			/*
751 			 * locate outgoing interface; if we're the destination,
752 			 * use the incoming interface (should be same).
753 			 */
754 			if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == 0 &&
755 			    (ia = ip_rtaddr(ipaddr.sin_addr)) == 0) {
756 				type = ICMP_UNREACH;
757 				code = ICMP_UNREACH_HOST;
758 				goto bad;
759 			}
760 			bcopy((caddr_t)&(IA_SIN(ia)->sin_addr),
761 			    (caddr_t)(cp + off), sizeof(struct in_addr));
762 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
763 			break;
764 
765 		case IPOPT_TS:
766 			code = cp - (u_char *)ip;
767 			ipt = (struct ip_timestamp *)cp;
768 			if (ipt->ipt_len < 5)
769 				goto bad;
770 			if (ipt->ipt_ptr > ipt->ipt_len - sizeof (long)) {
771 				if (++ipt->ipt_oflw == 0)
772 					goto bad;
773 				break;
774 			}
775 			sin = (struct in_addr *)(cp + ipt->ipt_ptr - 1);
776 			switch (ipt->ipt_flg) {
777 
778 			case IPOPT_TS_TSONLY:
779 				break;
780 
781 			case IPOPT_TS_TSANDADDR:
782 				if (ipt->ipt_ptr + sizeof(n_time) +
783 				    sizeof(struct in_addr) > ipt->ipt_len)
784 					goto bad;
785 				ipaddr.sin_addr = dst;
786 				ia = (INA)ifaof_ifpforaddr((SA)&ipaddr,
787 							    m->m_pkthdr.rcvif);
788 				if (ia == 0)
789 					continue;
790 				bcopy((caddr_t)&IA_SIN(ia)->sin_addr,
791 				    (caddr_t)sin, sizeof(struct in_addr));
792 				ipt->ipt_ptr += sizeof(struct in_addr);
793 				break;
794 
795 			case IPOPT_TS_PRESPEC:
796 				if (ipt->ipt_ptr + sizeof(n_time) +
797 				    sizeof(struct in_addr) > ipt->ipt_len)
798 					goto bad;
799 				bcopy((caddr_t)sin, (caddr_t)&ipaddr.sin_addr,
800 				    sizeof(struct in_addr));
801 				if (ifa_ifwithaddr((SA)&ipaddr) == 0)
802 					continue;
803 				ipt->ipt_ptr += sizeof(struct in_addr);
804 				break;
805 
806 			default:
807 				goto bad;
808 			}
809 			ntime = iptime();
810 			bcopy((caddr_t)&ntime, (caddr_t)cp + ipt->ipt_ptr - 1,
811 			    sizeof(n_time));
812 			ipt->ipt_ptr += sizeof(n_time);
813 		}
814 	}
815 	if (forward) {
816 		ip_forward(m, 1);
817 		return (1);
818 	}
819 	return (0);
820 bad:
821 	ip->ip_len -= ip->ip_hl << 2;   /* XXX icmp_error adds in hdr length */
822 	icmp_error(m, type, code, 0, 0);
823 	ipstat.ips_badoptions++;
824 	return (1);
825 }
826 
827 /*
828  * Given address of next destination (final or next hop),
829  * return internet address info of interface to be used to get there.
830  */
831 struct in_ifaddr *
832 ip_rtaddr(dst)
833 	 struct in_addr dst;
834 {
835 	register struct sockaddr_in *sin;
836 
837 	sin = (struct sockaddr_in *) &ipforward_rt.ro_dst;
838 
839 	if (ipforward_rt.ro_rt == 0 || dst.s_addr != sin->sin_addr.s_addr) {
840 		if (ipforward_rt.ro_rt) {
841 			RTFREE(ipforward_rt.ro_rt);
842 			ipforward_rt.ro_rt = 0;
843 		}
844 		sin->sin_family = AF_INET;
845 		sin->sin_len = sizeof(*sin);
846 		sin->sin_addr = dst;
847 
848 		rtalloc(&ipforward_rt);
849 	}
850 	if (ipforward_rt.ro_rt == 0)
851 		return ((struct in_ifaddr *)0);
852 	return ((struct in_ifaddr *) ipforward_rt.ro_rt->rt_ifa);
853 }
854 
855 /*
856  * Save incoming source route for use in replies,
857  * to be picked up later by ip_srcroute if the receiver is interested.
858  */
859 void
860 save_rte(option, dst)
861 	u_char *option;
862 	struct in_addr dst;
863 {
864 	unsigned olen;
865 
866 	olen = option[IPOPT_OLEN];
867 #ifdef DIAGNOSTIC
868 	if (ipprintfs)
869 		printf("save_rte: olen %d\n", olen);
870 #endif
871 	if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst)))
872 		return;
873 	bcopy((caddr_t)option, (caddr_t)ip_srcrt.srcopt, olen);
874 	ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
875 	ip_srcrt.dst = dst;
876 }
877 
878 /*
879  * Retrieve incoming source route for use in replies,
880  * in the same form used by setsockopt.
881  * The first hop is placed before the options, will be removed later.
882  */
883 struct mbuf *
884 ip_srcroute()
885 {
886 	register struct in_addr *p, *q;
887 	register struct mbuf *m;
888 
889 	if (ip_nhops == 0)
890 		return ((struct mbuf *)0);
891 	m = m_get(M_DONTWAIT, MT_SOOPTS);
892 	if (m == 0)
893 		return ((struct mbuf *)0);
894 
895 #define OPTSIZ	(sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt))
896 
897 	/* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
898 	m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) +
899 	    OPTSIZ;
900 #ifdef DIAGNOSTIC
901 	if (ipprintfs)
902 		printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len);
903 #endif
904 
905 	/*
906 	 * First save first hop for return route
907 	 */
908 	p = &ip_srcrt.route[ip_nhops - 1];
909 	*(mtod(m, struct in_addr *)) = *p--;
910 #ifdef DIAGNOSTIC
911 	if (ipprintfs)
912 		printf(" hops %lx", ntohl(mtod(m, struct in_addr *)->s_addr));
913 #endif
914 
915 	/*
916 	 * Copy option fields and padding (nop) to mbuf.
917 	 */
918 	ip_srcrt.nop = IPOPT_NOP;
919 	ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
920 	bcopy((caddr_t)&ip_srcrt.nop,
921 	    mtod(m, caddr_t) + sizeof(struct in_addr), OPTSIZ);
922 	q = (struct in_addr *)(mtod(m, caddr_t) +
923 	    sizeof(struct in_addr) + OPTSIZ);
924 #undef OPTSIZ
925 	/*
926 	 * Record return path as an IP source route,
927 	 * reversing the path (pointers are now aligned).
928 	 */
929 	while (p >= ip_srcrt.route) {
930 #ifdef DIAGNOSTIC
931 		if (ipprintfs)
932 			printf(" %lx", ntohl(q->s_addr));
933 #endif
934 		*q++ = *p--;
935 	}
936 	/*
937 	 * Last hop goes to final destination.
938 	 */
939 	*q = ip_srcrt.dst;
940 #ifdef DIAGNOSTIC
941 	if (ipprintfs)
942 		printf(" %lx\n", ntohl(q->s_addr));
943 #endif
944 	return (m);
945 }
946 
947 /*
948  * Strip out IP options, at higher
949  * level protocol in the kernel.
950  * Second argument is buffer to which options
951  * will be moved, and return value is their length.
952  * XXX should be deleted; last arg currently ignored.
953  */
954 void
955 ip_stripoptions(m, mopt)
956 	register struct mbuf *m;
957 	struct mbuf *mopt;
958 {
959 	register int i;
960 	struct ip *ip = mtod(m, struct ip *);
961 	register caddr_t opts;
962 	int olen;
963 
964 	olen = (ip->ip_hl<<2) - sizeof (struct ip);
965 	opts = (caddr_t)(ip + 1);
966 	i = m->m_len - (sizeof (struct ip) + olen);
967 	bcopy(opts  + olen, opts, (unsigned)i);
968 	m->m_len -= olen;
969 	if (m->m_flags & M_PKTHDR)
970 		m->m_pkthdr.len -= olen;
971 	ip->ip_hl = sizeof(struct ip) >> 2;
972 }
973 
974 u_char inetctlerrmap[PRC_NCMDS] = {
975 	0,		0,		0,		0,
976 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
977 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
978 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
979 	0,		0,		0,		0,
980 	ENOPROTOOPT
981 };
982 
983 /*
984  * Forward a packet.  If some error occurs return the sender
985  * an icmp packet.  Note we can't always generate a meaningful
986  * icmp message because icmp doesn't have a large enough repertoire
987  * of codes and types.
988  *
989  * If not forwarding, just drop the packet.  This could be confusing
990  * if ipforwarding was zero but some routing protocol was advancing
991  * us as a gateway to somewhere.  However, we must let the routing
992  * protocol deal with that.
993  *
994  * The srcrt parameter indicates whether the packet is being forwarded
995  * via a source route.
996  */
997 void
998 ip_forward(m, srcrt)
999 	struct mbuf *m;
1000 	int srcrt;
1001 {
1002 	register struct ip *ip = mtod(m, struct ip *);
1003 	register struct sockaddr_in *sin;
1004 	register struct rtentry *rt;
1005 	int error, type = 0, code = 0;
1006 	struct mbuf *mcopy;
1007 	n_long dest;
1008 	struct ifnet *destifp;
1009 
1010 	dest = 0;
1011 #ifdef DIAGNOSTIC
1012 	if (ipprintfs)
1013 		printf("forward: src %x dst %x ttl %x\n", ip->ip_src,
1014 			ip->ip_dst, ip->ip_ttl);
1015 #endif
1016 	if (m->m_flags & M_BCAST || in_canforward(ip->ip_dst) == 0) {
1017 		ipstat.ips_cantforward++;
1018 		m_freem(m);
1019 		return;
1020 	}
1021 	HTONS(ip->ip_id);
1022 	if (ip->ip_ttl <= IPTTLDEC) {
1023 		icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0);
1024 		return;
1025 	}
1026 	ip->ip_ttl -= IPTTLDEC;
1027 
1028 	sin = (struct sockaddr_in *)&ipforward_rt.ro_dst;
1029 	if ((rt = ipforward_rt.ro_rt) == 0 ||
1030 	    ip->ip_dst.s_addr != sin->sin_addr.s_addr) {
1031 		if (ipforward_rt.ro_rt) {
1032 			RTFREE(ipforward_rt.ro_rt);
1033 			ipforward_rt.ro_rt = 0;
1034 		}
1035 		sin->sin_family = AF_INET;
1036 		sin->sin_len = sizeof(*sin);
1037 		sin->sin_addr = ip->ip_dst;
1038 
1039 		rtalloc(&ipforward_rt);
1040 		if (ipforward_rt.ro_rt == 0) {
1041 			icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0);
1042 			return;
1043 		}
1044 		rt = ipforward_rt.ro_rt;
1045 	}
1046 
1047 	/*
1048 	 * Save at most 64 bytes of the packet in case
1049 	 * we need to generate an ICMP message to the src.
1050 	 */
1051 	mcopy = m_copy(m, 0, imin((int)ip->ip_len, 64));
1052 
1053 #ifdef GATEWAY
1054 	ip_ifmatrix[rt->rt_ifp->if_index +
1055 	     if_index * m->m_pkthdr.rcvif->if_index]++;
1056 #endif
1057 	/*
1058 	 * If forwarding packet using same interface that it came in on,
1059 	 * perhaps should send a redirect to sender to shortcut a hop.
1060 	 * Only send redirect if source is sending directly to us,
1061 	 * and if packet was not source routed (or has any options).
1062 	 * Also, don't send redirect if forwarding using a default route
1063 	 * or a route modified by a redirect.
1064 	 */
1065 #define	satosin(sa)	((struct sockaddr_in *)(sa))
1066 	if (rt->rt_ifp == m->m_pkthdr.rcvif &&
1067 	    (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1068 	    satosin(rt_key(rt))->sin_addr.s_addr != 0 &&
1069 	    ipsendredirects && !srcrt) {
1070 #define	RTA(rt)	((struct in_ifaddr *)(rt->rt_ifa))
1071 		u_long src = ntohl(ip->ip_src.s_addr);
1072 
1073 		if (RTA(rt) &&
1074 		    (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1075 		    if (rt->rt_flags & RTF_GATEWAY)
1076 			dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
1077 		    else
1078 			dest = ip->ip_dst.s_addr;
1079 		    /* Router requirements says to only send host redirects */
1080 		    type = ICMP_REDIRECT;
1081 		    code = ICMP_REDIRECT_HOST;
1082 #ifdef DIAGNOSTIC
1083 		    if (ipprintfs)
1084 		        printf("redirect (%d) to %lx\n", code, (u_long)dest);
1085 #endif
1086 		}
1087 	}
1088 
1089 	error = ip_output(m, (struct mbuf *)0, &ipforward_rt, IP_FORWARDING
1090 #ifdef DIRECTED_BROADCAST
1091 			    | IP_ALLOWBROADCAST
1092 #endif
1093 						, 0);
1094 	if (error)
1095 		ipstat.ips_cantforward++;
1096 	else {
1097 		ipstat.ips_forward++;
1098 		if (type)
1099 			ipstat.ips_redirectsent++;
1100 		else {
1101 			if (mcopy)
1102 				m_freem(mcopy);
1103 			return;
1104 		}
1105 	}
1106 	if (mcopy == NULL)
1107 		return;
1108 	destifp = NULL;
1109 
1110 	switch (error) {
1111 
1112 	case 0:				/* forwarded, but need redirect */
1113 		/* type, code set above */
1114 		break;
1115 
1116 	case ENETUNREACH:		/* shouldn't happen, checked above */
1117 	case EHOSTUNREACH:
1118 	case ENETDOWN:
1119 	case EHOSTDOWN:
1120 	default:
1121 		type = ICMP_UNREACH;
1122 		code = ICMP_UNREACH_HOST;
1123 		break;
1124 
1125 	case EMSGSIZE:
1126 		type = ICMP_UNREACH;
1127 		code = ICMP_UNREACH_NEEDFRAG;
1128 		if (ipforward_rt.ro_rt)
1129 			destifp = ipforward_rt.ro_rt->rt_ifp;
1130 		ipstat.ips_cantfrag++;
1131 		break;
1132 
1133 	case ENOBUFS:
1134 		type = ICMP_SOURCEQUENCH;
1135 		code = 0;
1136 		break;
1137 	}
1138 	icmp_error(mcopy, type, code, dest, destifp);
1139 }
1140 
1141 int
1142 ip_sysctl(name, namelen, oldp, oldlenp, newp, newlen)
1143 	int *name;
1144 	u_int namelen;
1145 	void *oldp;
1146 	size_t *oldlenp;
1147 	void *newp;
1148 	size_t newlen;
1149 {
1150 	/* All sysctl names at this level are terminal. */
1151 	if (namelen != 1)
1152 		return (ENOTDIR);
1153 
1154 	switch (name[0]) {
1155 	case IPCTL_FORWARDING:
1156 		return (sysctl_int(oldp, oldlenp, newp, newlen, &ipforwarding));
1157 	case IPCTL_SENDREDIRECTS:
1158 		return (sysctl_int(oldp, oldlenp, newp, newlen,
1159 			&ipsendredirects));
1160 	case IPCTL_DEFTTL:
1161 		return (sysctl_int(oldp, oldlenp, newp, newlen, &ip_defttl));
1162 #ifdef notyet
1163 	case IPCTL_DEFMTU:
1164 		return (sysctl_int(oldp, oldlenp, newp, newlen, &ip_mtu));
1165 #endif
1166 	default:
1167 		return (EOPNOTSUPP);
1168 	}
1169 	/* NOTREACHED */
1170 }
1171