xref: /freebsd/sys/netinet/ip_input.c (revision 48991a368427cadb9cdac39581d1676c29619c52)
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 1.27 1995/11/01 17:18:27 wollman 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 #include <sys/syslog.h>
48 
49 #include <vm/vm.h>
50 #include <sys/sysctl.h>
51 
52 #include <net/if.h>
53 #include <net/route.h>
54 #include <net/netisr.h>
55 
56 #include <netinet/in.h>
57 #include <netinet/in_systm.h>
58 #include <netinet/in_var.h>
59 #include <netinet/ip.h>
60 #include <netinet/in_pcb.h>
61 #include <netinet/in_var.h>
62 #include <netinet/ip_var.h>
63 #include <netinet/ip_icmp.h>
64 
65 #include <netinet/ip_fw.h>
66 
67 #include <sys/socketvar.h>
68 int rsvp_on = 0;
69 int ip_rsvp_on;
70 struct socket *ip_rsvpd;
71 
72 #ifndef	IPFORWARDING
73 #ifdef GATEWAY
74 #define	IPFORWARDING	1	/* forward IP packets not for us */
75 #else /* GATEWAY */
76 #define	IPFORWARDING	0	/* don't forward IP packets not for us */
77 #endif /* GATEWAY */
78 #endif /* IPFORWARDING */
79 #ifndef	IPSENDREDIRECTS
80 #define	IPSENDREDIRECTS	1
81 #endif
82 #ifndef	DIRECTED_BROADCAST
83 #define	DIRECTED_BROADCAST	0
84 #endif
85 
86 static int	ipforwarding = IPFORWARDING;
87 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW,
88 	&ipforwarding, 0, "");
89 
90 static int	ipsendredirects = IPSENDREDIRECTS;
91 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW,
92 	&ipsendredirects, 0, "");
93 
94 static int	ipdirbroadcast = DIRECTED_BROADCAST;
95 SYSCTL_INT(_net_inet_ip, IPCTL_DIRECTEDBROADCAST, directed_broadcast,
96 	CTLFLAG_RW, &ipdirbroadcast, 0, "");
97 
98 int	ip_defttl = IPDEFTTL;
99 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW,
100 	&ip_defttl, 0, "");
101 
102 static int	ip_dosourceroute = 0;
103 SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW,
104 	&ip_dosourceroute, 0, "");
105 #ifdef DIAGNOSTIC
106 static int	ipprintfs = 0;
107 #endif
108 
109 extern	struct domain inetdomain;
110 extern	struct protosw inetsw[];
111 u_char	ip_protox[IPPROTO_MAX];
112 static int	ipqmaxlen = IFQ_MAXLEN;
113 struct	in_ifaddr *in_ifaddr;			/* first inet address */
114 struct	ifqueue ipintrq;
115 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RD,
116 	&ipintrq.ifq_maxlen, 0, "");
117 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD,
118 	&ipintrq.ifq_drops, 0, "");
119 
120 struct ipstat ipstat;
121 struct ipq ipq;
122 
123 #ifdef IPCTL_DEFMTU
124 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
125 	&ip_mtu, 0, "");
126 #endif
127 
128 /*
129  * We need to save the IP options in case a protocol wants to respond
130  * to an incoming packet over the same route if the packet got here
131  * using IP source routing.  This allows connection establishment and
132  * maintenance when the remote end is on a network that is not known
133  * to us.
134  */
135 static int	ip_nhops = 0;
136 static	struct ip_srcrt {
137 	struct	in_addr dst;			/* final destination */
138 	char	nop;				/* one NOP to align */
139 	char	srcopt[IPOPT_OFFSET + 1];	/* OPTVAL, OLEN and OFFSET */
140 	struct	in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)];
141 } ip_srcrt;
142 
143 static void save_rte __P((u_char *, struct in_addr));
144 static void	 ip_deq __P((struct ipasfrag *));
145 static int	 ip_dooptions __P((struct mbuf *));
146 static void	 ip_enq __P((struct ipasfrag *, struct ipasfrag *));
147 static void	 ip_forward __P((struct mbuf *, int));
148 static void	 ip_freef __P((struct ipq *));
149 static struct ip *
150 	 ip_reass __P((struct ipasfrag *, struct ipq *));
151 static struct in_ifaddr *
152 	 ip_rtaddr __P((struct in_addr));
153 static void	 ipintr __P((void));
154 /*
155  * IP initialization: fill in IP protocol switch table.
156  * All protocols not implemented in kernel go to raw IP protocol handler.
157  */
158 void
159 ip_init()
160 {
161 	register struct protosw *pr;
162 	register int i;
163 
164 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
165 	if (pr == 0)
166 		panic("ip_init");
167 	for (i = 0; i < IPPROTO_MAX; i++)
168 		ip_protox[i] = pr - inetsw;
169 	for (pr = inetdomain.dom_protosw;
170 	    pr < inetdomain.dom_protoswNPROTOSW; pr++)
171 		if (pr->pr_domain->dom_family == PF_INET &&
172 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
173 			ip_protox[pr->pr_protocol] = pr - inetsw;
174 	ipq.next = ipq.prev = &ipq;
175 	ip_id = time.tv_sec & 0xffff;
176 	ipintrq.ifq_maxlen = ipqmaxlen;
177 }
178 
179 static struct	sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET };
180 struct	route ipforward_rt;
181 
182 /*
183  * Ip input routine.  Checksum and byte swap header.  If fragmented
184  * try to reassemble.  Process options.  Pass to next level.
185  */
186 static void
187 ipintr(void)
188 {
189 	register struct ip *ip;
190 	register struct mbuf *m;
191 	register struct ipq *fp;
192 	register struct in_ifaddr *ia;
193 	int hlen, s;
194 
195 next:
196 	/*
197 	 * Get next datagram off input queue and get IP header
198 	 * in first mbuf.
199 	 */
200 	s = splimp();
201 	IF_DEQUEUE(&ipintrq, m);
202 	splx(s);
203 	if (m == 0)
204 		return;
205 #ifdef	DIAGNOSTIC
206 	if ((m->m_flags & M_PKTHDR) == 0)
207 		panic("ipintr no HDR");
208 #endif
209 	/*
210 	 * If no IP addresses have been set yet but the interfaces
211 	 * are receiving, can't do anything with incoming packets yet.
212 	 */
213 	if (in_ifaddr == NULL)
214 		goto bad;
215 	ipstat.ips_total++;
216 	if (m->m_len < sizeof (struct ip) &&
217 	    (m = m_pullup(m, sizeof (struct ip))) == 0) {
218 		ipstat.ips_toosmall++;
219 		goto next;
220 	}
221 	ip = mtod(m, struct ip *);
222 	if (ip->ip_v != IPVERSION) {
223 		ipstat.ips_badvers++;
224 		goto bad;
225 	}
226 	hlen = ip->ip_hl << 2;
227 	if (hlen < sizeof(struct ip)) {	/* minimum header length */
228 		ipstat.ips_badhlen++;
229 		goto bad;
230 	}
231 	if (hlen > m->m_len) {
232 		if ((m = m_pullup(m, hlen)) == 0) {
233 			ipstat.ips_badhlen++;
234 			goto next;
235 		}
236 		ip = mtod(m, struct ip *);
237 	}
238 	ip->ip_sum = in_cksum(m, hlen);
239 	if (ip->ip_sum) {
240 		ipstat.ips_badsum++;
241 		goto bad;
242 	}
243 
244 	/*
245 	 * Convert fields to host representation.
246 	 */
247 	NTOHS(ip->ip_len);
248 	if (ip->ip_len < hlen) {
249 		ipstat.ips_badlen++;
250 		goto bad;
251 	}
252 	NTOHS(ip->ip_id);
253 	NTOHS(ip->ip_off);
254 
255 	/*
256 	 * Check that the amount of data in the buffers
257 	 * is as at least much as the IP header would have us expect.
258 	 * Trim mbufs if longer than we expect.
259 	 * Drop packet if shorter than we expect.
260 	 */
261 	if (m->m_pkthdr.len < ip->ip_len) {
262 		ipstat.ips_tooshort++;
263 		goto bad;
264 	}
265 	if (m->m_pkthdr.len > ip->ip_len) {
266 		if (m->m_len == m->m_pkthdr.len) {
267 			m->m_len = ip->ip_len;
268 			m->m_pkthdr.len = ip->ip_len;
269 		} else
270 			m_adj(m, ip->ip_len - m->m_pkthdr.len);
271 	}
272 	/*
273 	 * IpHack's section.
274 	 * Right now when no processing on packet has done
275 	 * and it is still fresh out of network we do our black
276 	 * deals with it.
277 	 * - Firewall: deny/allow
278 	 * - Wrap: fake packet's addr/port <unimpl.>
279 	 * - Encapsulate: put it in another IP and send out. <unimp.>
280  	 */
281 
282         if (ip_fw_chk_ptr!=NULL)
283                if (!(*ip_fw_chk_ptr)(m,ip,m->m_pkthdr.rcvif,ip_fw_chain) ) {
284                        goto next;
285                }
286 
287 	/*
288 	 * Process options and, if not destined for us,
289 	 * ship it on.  ip_dooptions returns 1 when an
290 	 * error was detected (causing an icmp message
291 	 * to be sent and the original packet to be freed).
292 	 */
293 	ip_nhops = 0;		/* for source routed packets */
294 	if (hlen > sizeof (struct ip) && ip_dooptions(m))
295 		goto next;
296 
297         /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
298          * matter if it is destined to another node, or whether it is
299          * a multicast one, RSVP wants it! and prevents it from being forwarded
300          * anywhere else. Also checks if the rsvp daemon is running before
301 	 * grabbing the packet.
302          */
303 	if (rsvp_on && ip->ip_p==IPPROTO_RSVP)
304 		goto ours;
305 
306 	/*
307 	 * Check our list of addresses, to see if the packet is for us.
308 	 */
309 	for (ia = in_ifaddr; ia; ia = ia->ia_next) {
310 #define	satosin(sa)	((struct sockaddr_in *)(sa))
311 
312 		if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr)
313 			goto ours;
314 		if ((!ipdirbroadcast || ia->ia_ifp == m->m_pkthdr.rcvif) &&
315 		    (ia->ia_ifp->if_flags & IFF_BROADCAST)) {
316 			u_long t;
317 
318 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
319 			    ip->ip_dst.s_addr)
320 				goto ours;
321 			if (ip->ip_dst.s_addr == ia->ia_netbroadcast.s_addr)
322 				goto ours;
323 			/*
324 			 * Look for all-0's host part (old broadcast addr),
325 			 * either for subnet or net.
326 			 */
327 			t = ntohl(ip->ip_dst.s_addr);
328 			if (t == ia->ia_subnet)
329 				goto ours;
330 			if (t == ia->ia_net)
331 				goto ours;
332 		}
333 	}
334 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
335 		struct in_multi *inm;
336 		if (ip_mrouter) {
337 			/*
338 			 * If we are acting as a multicast router, all
339 			 * incoming multicast packets are passed to the
340 			 * kernel-level multicast forwarding function.
341 			 * The packet is returned (relatively) intact; if
342 			 * ip_mforward() returns a non-zero value, the packet
343 			 * must be discarded, else it may be accepted below.
344 			 *
345 			 * (The IP ident field is put in the same byte order
346 			 * as expected when ip_mforward() is called from
347 			 * ip_output().)
348 			 */
349 			ip->ip_id = htons(ip->ip_id);
350 			if (ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) {
351 				ipstat.ips_cantforward++;
352 				m_freem(m);
353 				goto next;
354 			}
355 			ip->ip_id = ntohs(ip->ip_id);
356 
357 			/*
358 			 * The process-level routing demon needs to receive
359 			 * all multicast IGMP packets, whether or not this
360 			 * host belongs to their destination groups.
361 			 */
362 			if (ip->ip_p == IPPROTO_IGMP)
363 				goto ours;
364 			ipstat.ips_forward++;
365 		}
366 		/*
367 		 * See if we belong to the destination multicast group on the
368 		 * arrival interface.
369 		 */
370 		IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
371 		if (inm == NULL) {
372 			ipstat.ips_cantforward++;
373 			m_freem(m);
374 			goto next;
375 		}
376 		goto ours;
377 	}
378 	if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
379 		goto ours;
380 	if (ip->ip_dst.s_addr == INADDR_ANY)
381 		goto ours;
382 
383 	/*
384 	 * Not for us; forward if possible and desirable.
385 	 */
386 	if (ipforwarding == 0) {
387 		ipstat.ips_cantforward++;
388 		m_freem(m);
389 	} else
390 		ip_forward(m, 0);
391 	goto next;
392 
393 ours:
394 
395 		/*
396 		 * If packet came to us we count it...
397 		 * This way we count all incoming packets which has
398 		 * not been forwarded...
399 		 * Do not convert ip_len to host byte order when
400 		 * counting,ppl already made it for us before..
401 		 */
402 	if (ip_acct_cnt_ptr!=NULL)
403 		(*ip_acct_cnt_ptr)(ip,m->m_pkthdr.rcvif,ip_acct_chain,0);
404 
405 	/*
406 	 * If offset or IP_MF are set, must reassemble.
407 	 * Otherwise, nothing need be done.
408 	 * (We could look in the reassembly queue to see
409 	 * if the packet was previously fragmented,
410 	 * but it's not worth the time; just let them time out.)
411 	 */
412 	if (ip->ip_off &~ IP_DF) {
413 		if (m->m_flags & M_EXT) {		/* XXX */
414 			if ((m = m_pullup(m, sizeof (struct ip))) == 0) {
415 				ipstat.ips_toosmall++;
416 				goto next;
417 			}
418 			ip = mtod(m, struct ip *);
419 		}
420 		/*
421 		 * Look for queue of fragments
422 		 * of this datagram.
423 		 */
424 		for (fp = ipq.next; fp != &ipq; fp = fp->next)
425 			if (ip->ip_id == fp->ipq_id &&
426 			    ip->ip_src.s_addr == fp->ipq_src.s_addr &&
427 			    ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
428 			    ip->ip_p == fp->ipq_p)
429 				goto found;
430 		fp = 0;
431 found:
432 
433 		/*
434 		 * Adjust ip_len to not reflect header,
435 		 * set ip_mff if more fragments are expected,
436 		 * convert offset of this to bytes.
437 		 */
438 		ip->ip_len -= hlen;
439 		((struct ipasfrag *)ip)->ipf_mff &= ~1;
440 		if (ip->ip_off & IP_MF)
441 			((struct ipasfrag *)ip)->ipf_mff |= 1;
442 		ip->ip_off <<= 3;
443 
444 		/*
445 		 * If datagram marked as having more fragments
446 		 * or if this is not the first fragment,
447 		 * attempt reassembly; if it succeeds, proceed.
448 		 */
449 		if (((struct ipasfrag *)ip)->ipf_mff & 1 || ip->ip_off) {
450 			ipstat.ips_fragments++;
451 			ip = ip_reass((struct ipasfrag *)ip, fp);
452 			if (ip == 0)
453 				goto next;
454 			ipstat.ips_reassembled++;
455 			m = dtom(ip);
456 		} else
457 			if (fp)
458 				ip_freef(fp);
459 	} else
460 		ip->ip_len -= hlen;
461 
462 	/*
463 	 * Switch out to protocol's input routine.
464 	 */
465 	ipstat.ips_delivered++;
466 	(*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen);
467 	goto next;
468 bad:
469 	m_freem(m);
470 	goto next;
471 }
472 
473 NETISR_SET(NETISR_IP, ipintr);
474 
475 /*
476  * Take incoming datagram fragment and try to
477  * reassemble it into whole datagram.  If a chain for
478  * reassembly of this datagram already exists, then it
479  * is given as fp; otherwise have to make a chain.
480  */
481 static struct ip *
482 ip_reass(ip, fp)
483 	register struct ipasfrag *ip;
484 	register struct ipq *fp;
485 {
486 	register struct mbuf *m = dtom(ip);
487 	register struct ipasfrag *q;
488 	struct mbuf *t;
489 	int hlen = ip->ip_hl << 2;
490 	int i, next;
491 
492 	/*
493 	 * Presence of header sizes in mbufs
494 	 * would confuse code below.
495 	 */
496 	m->m_data += hlen;
497 	m->m_len -= hlen;
498 
499 	/*
500 	 * If first fragment to arrive, create a reassembly queue.
501 	 */
502 	if (fp == 0) {
503 		if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL)
504 			goto dropfrag;
505 		fp = mtod(t, struct ipq *);
506 		insque(fp, &ipq);
507 		fp->ipq_ttl = IPFRAGTTL;
508 		fp->ipq_p = ip->ip_p;
509 		fp->ipq_id = ip->ip_id;
510 		fp->ipq_next = fp->ipq_prev = (struct ipasfrag *)fp;
511 		fp->ipq_src = ((struct ip *)ip)->ip_src;
512 		fp->ipq_dst = ((struct ip *)ip)->ip_dst;
513 		q = (struct ipasfrag *)fp;
514 		goto insert;
515 	}
516 
517 	/*
518 	 * Find a segment which begins after this one does.
519 	 */
520 	for (q = fp->ipq_next; q != (struct ipasfrag *)fp; q = q->ipf_next)
521 		if (q->ip_off > ip->ip_off)
522 			break;
523 
524 	/*
525 	 * If there is a preceding segment, it may provide some of
526 	 * our data already.  If so, drop the data from the incoming
527 	 * segment.  If it provides all of our data, drop us.
528 	 */
529 	if (q->ipf_prev != (struct ipasfrag *)fp) {
530 		i = q->ipf_prev->ip_off + q->ipf_prev->ip_len - ip->ip_off;
531 		if (i > 0) {
532 			if (i >= ip->ip_len)
533 				goto dropfrag;
534 			m_adj(dtom(ip), i);
535 			ip->ip_off += i;
536 			ip->ip_len -= i;
537 		}
538 	}
539 
540 	/*
541 	 * While we overlap succeeding segments trim them or,
542 	 * if they are completely covered, dequeue them.
543 	 */
544 	while (q != (struct ipasfrag *)fp && ip->ip_off + ip->ip_len > q->ip_off) {
545 		i = (ip->ip_off + ip->ip_len) - q->ip_off;
546 		if (i < q->ip_len) {
547 			q->ip_len -= i;
548 			q->ip_off += i;
549 			m_adj(dtom(q), i);
550 			break;
551 		}
552 		q = q->ipf_next;
553 		m_freem(dtom(q->ipf_prev));
554 		ip_deq(q->ipf_prev);
555 	}
556 
557 insert:
558 	/*
559 	 * Stick new segment in its place;
560 	 * check for complete reassembly.
561 	 */
562 	ip_enq(ip, q->ipf_prev);
563 	next = 0;
564 	for (q = fp->ipq_next; q != (struct ipasfrag *)fp; q = q->ipf_next) {
565 		if (q->ip_off != next)
566 			return (0);
567 		next += q->ip_len;
568 	}
569 	if (q->ipf_prev->ipf_mff & 1)
570 		return (0);
571 
572 	/*
573 	 * Reassembly is complete; concatenate fragments.
574 	 */
575 	q = fp->ipq_next;
576 	m = dtom(q);
577 	t = m->m_next;
578 	m->m_next = 0;
579 	m_cat(m, t);
580 	q = q->ipf_next;
581 	while (q != (struct ipasfrag *)fp) {
582 		t = dtom(q);
583 		q = q->ipf_next;
584 		m_cat(m, t);
585 	}
586 
587 	/*
588 	 * Create header for new ip packet by
589 	 * modifying header of first packet;
590 	 * dequeue and discard fragment reassembly header.
591 	 * Make header visible.
592 	 */
593 	ip = fp->ipq_next;
594 	ip->ip_len = next;
595 	ip->ipf_mff &= ~1;
596 	((struct ip *)ip)->ip_src = fp->ipq_src;
597 	((struct ip *)ip)->ip_dst = fp->ipq_dst;
598 	remque(fp);
599 	(void) m_free(dtom(fp));
600 	m = dtom(ip);
601 	m->m_len += (ip->ip_hl << 2);
602 	m->m_data -= (ip->ip_hl << 2);
603 	/* some debugging cruft by sklower, below, will go away soon */
604 	if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */
605 		register int plen = 0;
606 		for (t = m; m; m = m->m_next)
607 			plen += m->m_len;
608 		t->m_pkthdr.len = plen;
609 	}
610 	return ((struct ip *)ip);
611 
612 dropfrag:
613 	ipstat.ips_fragdropped++;
614 	m_freem(m);
615 	return (0);
616 }
617 
618 /*
619  * Free a fragment reassembly header and all
620  * associated datagrams.
621  */
622 static void
623 ip_freef(fp)
624 	struct ipq *fp;
625 {
626 	register struct ipasfrag *q, *p;
627 
628 	for (q = fp->ipq_next; q != (struct ipasfrag *)fp; q = p) {
629 		p = q->ipf_next;
630 		ip_deq(q);
631 		m_freem(dtom(q));
632 	}
633 	remque(fp);
634 	(void) m_free(dtom(fp));
635 }
636 
637 /*
638  * Put an ip fragment on a reassembly chain.
639  * Like insque, but pointers in middle of structure.
640  */
641 static void
642 ip_enq(p, prev)
643 	register struct ipasfrag *p, *prev;
644 {
645 
646 	p->ipf_prev = prev;
647 	p->ipf_next = prev->ipf_next;
648 	prev->ipf_next->ipf_prev = p;
649 	prev->ipf_next = p;
650 }
651 
652 /*
653  * To ip_enq as remque is to insque.
654  */
655 static void
656 ip_deq(p)
657 	register struct ipasfrag *p;
658 {
659 
660 	p->ipf_prev->ipf_next = p->ipf_next;
661 	p->ipf_next->ipf_prev = p->ipf_prev;
662 }
663 
664 /*
665  * IP timer processing;
666  * if a timer expires on a reassembly
667  * queue, discard it.
668  */
669 void
670 ip_slowtimo()
671 {
672 	register struct ipq *fp;
673 	int s = splnet();
674 
675 	fp = ipq.next;
676 	if (fp == 0) {
677 		splx(s);
678 		return;
679 	}
680 	while (fp != &ipq) {
681 		--fp->ipq_ttl;
682 		fp = fp->next;
683 		if (fp->prev->ipq_ttl == 0) {
684 			ipstat.ips_fragtimeout++;
685 			ip_freef(fp->prev);
686 		}
687 	}
688 	splx(s);
689 }
690 
691 /*
692  * Drain off all datagram fragments.
693  */
694 void
695 ip_drain()
696 {
697 
698 	while (ipq.next != &ipq) {
699 		ipstat.ips_fragdropped++;
700 		ip_freef(ipq.next);
701 	}
702 }
703 
704 /*
705  * Do option processing on a datagram,
706  * possibly discarding it if bad options are encountered,
707  * or forwarding it if source-routed.
708  * Returns 1 if packet has been forwarded/freed,
709  * 0 if the packet should be processed further.
710  */
711 static int
712 ip_dooptions(m)
713 	struct mbuf *m;
714 {
715 	register struct ip *ip = mtod(m, struct ip *);
716 	register u_char *cp;
717 	register struct ip_timestamp *ipt;
718 	register struct in_ifaddr *ia;
719 	int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0;
720 	struct in_addr *sin, dst;
721 	n_time ntime;
722 
723 	dst = ip->ip_dst;
724 	cp = (u_char *)(ip + 1);
725 	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
726 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
727 		opt = cp[IPOPT_OPTVAL];
728 		if (opt == IPOPT_EOL)
729 			break;
730 		if (opt == IPOPT_NOP)
731 			optlen = 1;
732 		else {
733 			optlen = cp[IPOPT_OLEN];
734 			if (optlen <= 0 || optlen > cnt) {
735 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
736 				goto bad;
737 			}
738 		}
739 		switch (opt) {
740 
741 		default:
742 			break;
743 
744 		/*
745 		 * Source routing with record.
746 		 * Find interface with current destination address.
747 		 * If none on this machine then drop if strictly routed,
748 		 * or do nothing if loosely routed.
749 		 * Record interface address and bring up next address
750 		 * component.  If strictly routed make sure next
751 		 * address is on directly accessible net.
752 		 */
753 		case IPOPT_LSRR:
754 		case IPOPT_SSRR:
755 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
756 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
757 				goto bad;
758 			}
759 			ipaddr.sin_addr = ip->ip_dst;
760 			ia = (struct in_ifaddr *)
761 				ifa_ifwithaddr((struct sockaddr *)&ipaddr);
762 			if (ia == 0) {
763 				if (opt == IPOPT_SSRR) {
764 					type = ICMP_UNREACH;
765 					code = ICMP_UNREACH_SRCFAIL;
766 					goto bad;
767 				}
768 				/*
769 				 * Loose routing, and not at next destination
770 				 * yet; nothing to do except forward.
771 				 */
772 				break;
773 			}
774 			off--;			/* 0 origin */
775 			if (off > optlen - sizeof(struct in_addr)) {
776 				/*
777 				 * End of source route.  Should be for us.
778 				 */
779 				save_rte(cp, ip->ip_src);
780 				break;
781 			}
782 
783 			if (!ip_dosourceroute) {
784 				char buf[4*sizeof "123"];
785 				strcpy(buf, inet_ntoa(ip->ip_dst));
786 
787 				log(LOG_WARNING,
788 				    "attempted source route from %s to %s\n",
789 				    inet_ntoa(ip->ip_src), buf);
790 				type = ICMP_UNREACH;
791 				code = ICMP_UNREACH_SRCFAIL;
792 				goto bad;
793 			}
794 
795 			/*
796 			 * locate outgoing interface
797 			 */
798 			(void)memcpy(&ipaddr.sin_addr, cp + off,
799 			    sizeof(ipaddr.sin_addr));
800 
801 			if (opt == IPOPT_SSRR) {
802 #define	INA	struct in_ifaddr *
803 #define	SA	struct sockaddr *
804 			    if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == 0)
805 				ia = (INA)ifa_ifwithnet((SA)&ipaddr);
806 			} else
807 				ia = ip_rtaddr(ipaddr.sin_addr);
808 			if (ia == 0) {
809 				type = ICMP_UNREACH;
810 				code = ICMP_UNREACH_SRCFAIL;
811 				goto bad;
812 			}
813 			ip->ip_dst = ipaddr.sin_addr;
814 			(void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
815 			    sizeof(struct in_addr));
816 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
817 			/*
818 			 * Let ip_intr's mcast routing check handle mcast pkts
819 			 */
820 			forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr));
821 			break;
822 
823 		case IPOPT_RR:
824 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
825 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
826 				goto bad;
827 			}
828 			/*
829 			 * If no space remains, ignore.
830 			 */
831 			off--;			/* 0 origin */
832 			if (off > optlen - sizeof(struct in_addr))
833 				break;
834 			(void)memcpy(&ipaddr.sin_addr, &ip->ip_dst,
835 			    sizeof(ipaddr.sin_addr));
836 			/*
837 			 * locate outgoing interface; if we're the destination,
838 			 * use the incoming interface (should be same).
839 			 */
840 			if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == 0 &&
841 			    (ia = ip_rtaddr(ipaddr.sin_addr)) == 0) {
842 				type = ICMP_UNREACH;
843 				code = ICMP_UNREACH_HOST;
844 				goto bad;
845 			}
846 			(void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
847 			    sizeof(struct in_addr));
848 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
849 			break;
850 
851 		case IPOPT_TS:
852 			code = cp - (u_char *)ip;
853 			ipt = (struct ip_timestamp *)cp;
854 			if (ipt->ipt_len < 5)
855 				goto bad;
856 			if (ipt->ipt_ptr > ipt->ipt_len - sizeof (long)) {
857 				if (++ipt->ipt_oflw == 0)
858 					goto bad;
859 				break;
860 			}
861 			sin = (struct in_addr *)(cp + ipt->ipt_ptr - 1);
862 			switch (ipt->ipt_flg) {
863 
864 			case IPOPT_TS_TSONLY:
865 				break;
866 
867 			case IPOPT_TS_TSANDADDR:
868 				if (ipt->ipt_ptr + sizeof(n_time) +
869 				    sizeof(struct in_addr) > ipt->ipt_len)
870 					goto bad;
871 				ipaddr.sin_addr = dst;
872 				ia = (INA)ifaof_ifpforaddr((SA)&ipaddr,
873 							    m->m_pkthdr.rcvif);
874 				if (ia == 0)
875 					continue;
876 				(void)memcpy(sin, &IA_SIN(ia)->sin_addr,
877 				    sizeof(struct in_addr));
878 				ipt->ipt_ptr += sizeof(struct in_addr);
879 				break;
880 
881 			case IPOPT_TS_PRESPEC:
882 				if (ipt->ipt_ptr + sizeof(n_time) +
883 				    sizeof(struct in_addr) > ipt->ipt_len)
884 					goto bad;
885 				(void)memcpy(&ipaddr.sin_addr, sin,
886 				    sizeof(struct in_addr));
887 				if (ifa_ifwithaddr((SA)&ipaddr) == 0)
888 					continue;
889 				ipt->ipt_ptr += sizeof(struct in_addr);
890 				break;
891 
892 			default:
893 				goto bad;
894 			}
895 			ntime = iptime();
896 			(void)memcpy(cp + ipt->ipt_ptr - 1, &ntime,
897 			    sizeof(n_time));
898 			ipt->ipt_ptr += sizeof(n_time);
899 		}
900 	}
901 	if (forward) {
902 		ip_forward(m, 1);
903 		return (1);
904 	}
905 	return (0);
906 bad:
907 	ip->ip_len -= ip->ip_hl << 2;   /* XXX icmp_error adds in hdr length */
908 	icmp_error(m, type, code, 0, 0);
909 	ipstat.ips_badoptions++;
910 	return (1);
911 }
912 
913 /*
914  * Given address of next destination (final or next hop),
915  * return internet address info of interface to be used to get there.
916  */
917 static struct in_ifaddr *
918 ip_rtaddr(dst)
919 	 struct in_addr dst;
920 {
921 	register struct sockaddr_in *sin;
922 
923 	sin = (struct sockaddr_in *) &ipforward_rt.ro_dst;
924 
925 	if (ipforward_rt.ro_rt == 0 || dst.s_addr != sin->sin_addr.s_addr) {
926 		if (ipforward_rt.ro_rt) {
927 			RTFREE(ipforward_rt.ro_rt);
928 			ipforward_rt.ro_rt = 0;
929 		}
930 		sin->sin_family = AF_INET;
931 		sin->sin_len = sizeof(*sin);
932 		sin->sin_addr = dst;
933 
934 		rtalloc_ign(&ipforward_rt, RTF_PRCLONING);
935 	}
936 	if (ipforward_rt.ro_rt == 0)
937 		return ((struct in_ifaddr *)0);
938 	return ((struct in_ifaddr *) ipforward_rt.ro_rt->rt_ifa);
939 }
940 
941 /*
942  * Save incoming source route for use in replies,
943  * to be picked up later by ip_srcroute if the receiver is interested.
944  */
945 void
946 save_rte(option, dst)
947 	u_char *option;
948 	struct in_addr dst;
949 {
950 	unsigned olen;
951 
952 	olen = option[IPOPT_OLEN];
953 #ifdef DIAGNOSTIC
954 	if (ipprintfs)
955 		printf("save_rte: olen %d\n", olen);
956 #endif
957 	if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst)))
958 		return;
959 	(void)memcpy(ip_srcrt.srcopt, option, olen);
960 	ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
961 	ip_srcrt.dst = dst;
962 }
963 
964 /*
965  * Retrieve incoming source route for use in replies,
966  * in the same form used by setsockopt.
967  * The first hop is placed before the options, will be removed later.
968  */
969 struct mbuf *
970 ip_srcroute()
971 {
972 	register struct in_addr *p, *q;
973 	register struct mbuf *m;
974 
975 	if (ip_nhops == 0)
976 		return ((struct mbuf *)0);
977 	m = m_get(M_DONTWAIT, MT_SOOPTS);
978 	if (m == 0)
979 		return ((struct mbuf *)0);
980 
981 #define OPTSIZ	(sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt))
982 
983 	/* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
984 	m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) +
985 	    OPTSIZ;
986 #ifdef DIAGNOSTIC
987 	if (ipprintfs)
988 		printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len);
989 #endif
990 
991 	/*
992 	 * First save first hop for return route
993 	 */
994 	p = &ip_srcrt.route[ip_nhops - 1];
995 	*(mtod(m, struct in_addr *)) = *p--;
996 #ifdef DIAGNOSTIC
997 	if (ipprintfs)
998 		printf(" hops %lx", ntohl(mtod(m, struct in_addr *)->s_addr));
999 #endif
1000 
1001 	/*
1002 	 * Copy option fields and padding (nop) to mbuf.
1003 	 */
1004 	ip_srcrt.nop = IPOPT_NOP;
1005 	ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
1006 	(void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr),
1007 	    &ip_srcrt.nop, OPTSIZ);
1008 	q = (struct in_addr *)(mtod(m, caddr_t) +
1009 	    sizeof(struct in_addr) + OPTSIZ);
1010 #undef OPTSIZ
1011 	/*
1012 	 * Record return path as an IP source route,
1013 	 * reversing the path (pointers are now aligned).
1014 	 */
1015 	while (p >= ip_srcrt.route) {
1016 #ifdef DIAGNOSTIC
1017 		if (ipprintfs)
1018 			printf(" %lx", ntohl(q->s_addr));
1019 #endif
1020 		*q++ = *p--;
1021 	}
1022 	/*
1023 	 * Last hop goes to final destination.
1024 	 */
1025 	*q = ip_srcrt.dst;
1026 #ifdef DIAGNOSTIC
1027 	if (ipprintfs)
1028 		printf(" %lx\n", ntohl(q->s_addr));
1029 #endif
1030 	return (m);
1031 }
1032 
1033 /*
1034  * Strip out IP options, at higher
1035  * level protocol in the kernel.
1036  * Second argument is buffer to which options
1037  * will be moved, and return value is their length.
1038  * XXX should be deleted; last arg currently ignored.
1039  */
1040 void
1041 ip_stripoptions(m, mopt)
1042 	register struct mbuf *m;
1043 	struct mbuf *mopt;
1044 {
1045 	register int i;
1046 	struct ip *ip = mtod(m, struct ip *);
1047 	register caddr_t opts;
1048 	int olen;
1049 
1050 	olen = (ip->ip_hl<<2) - sizeof (struct ip);
1051 	opts = (caddr_t)(ip + 1);
1052 	i = m->m_len - (sizeof (struct ip) + olen);
1053 	bcopy(opts + olen, opts, (unsigned)i);
1054 	m->m_len -= olen;
1055 	if (m->m_flags & M_PKTHDR)
1056 		m->m_pkthdr.len -= olen;
1057 	ip->ip_hl = sizeof(struct ip) >> 2;
1058 }
1059 
1060 u_char inetctlerrmap[PRC_NCMDS] = {
1061 	0,		0,		0,		0,
1062 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1063 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1064 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1065 	0,		0,		0,		0,
1066 	ENOPROTOOPT
1067 };
1068 
1069 /*
1070  * Forward a packet.  If some error occurs return the sender
1071  * an icmp packet.  Note we can't always generate a meaningful
1072  * icmp message because icmp doesn't have a large enough repertoire
1073  * of codes and types.
1074  *
1075  * If not forwarding, just drop the packet.  This could be confusing
1076  * if ipforwarding was zero but some routing protocol was advancing
1077  * us as a gateway to somewhere.  However, we must let the routing
1078  * protocol deal with that.
1079  *
1080  * The srcrt parameter indicates whether the packet is being forwarded
1081  * via a source route.
1082  */
1083 static void
1084 ip_forward(m, srcrt)
1085 	struct mbuf *m;
1086 	int srcrt;
1087 {
1088 	register struct ip *ip = mtod(m, struct ip *);
1089 	register struct sockaddr_in *sin;
1090 	register struct rtentry *rt;
1091 	int error, type = 0, code = 0;
1092 	struct mbuf *mcopy;
1093 	n_long dest;
1094 	struct ifnet *destifp;
1095 
1096 	dest = 0;
1097 #ifdef DIAGNOSTIC
1098 	if (ipprintfs)
1099 		printf("forward: src %lx dst %lx ttl %x\n",
1100 			ip->ip_src.s_addr, ip->ip_dst.s_addr, ip->ip_ttl);
1101 #endif
1102 
1103 
1104 	if (m->m_flags & M_BCAST || in_canforward(ip->ip_dst) == 0) {
1105 		ipstat.ips_cantforward++;
1106 		m_freem(m);
1107 		return;
1108 	}
1109 	HTONS(ip->ip_id);
1110 	if (ip->ip_ttl <= IPTTLDEC) {
1111 		icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0);
1112 		return;
1113 	}
1114 	ip->ip_ttl -= IPTTLDEC;
1115 
1116 	sin = (struct sockaddr_in *)&ipforward_rt.ro_dst;
1117 	if ((rt = ipforward_rt.ro_rt) == 0 ||
1118 	    ip->ip_dst.s_addr != sin->sin_addr.s_addr) {
1119 		if (ipforward_rt.ro_rt) {
1120 			RTFREE(ipforward_rt.ro_rt);
1121 			ipforward_rt.ro_rt = 0;
1122 		}
1123 		sin->sin_family = AF_INET;
1124 		sin->sin_len = sizeof(*sin);
1125 		sin->sin_addr = ip->ip_dst;
1126 
1127 		rtalloc_ign(&ipforward_rt, RTF_PRCLONING);
1128 		if (ipforward_rt.ro_rt == 0) {
1129 			icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0);
1130 			return;
1131 		}
1132 		rt = ipforward_rt.ro_rt;
1133 	}
1134 
1135 	/*
1136 	 * Save at most 64 bytes of the packet in case
1137 	 * we need to generate an ICMP message to the src.
1138 	 */
1139 	mcopy = m_copy(m, 0, imin((int)ip->ip_len, 64));
1140 
1141 	/*
1142 	 * If forwarding packet using same interface that it came in on,
1143 	 * perhaps should send a redirect to sender to shortcut a hop.
1144 	 * Only send redirect if source is sending directly to us,
1145 	 * and if packet was not source routed (or has any options).
1146 	 * Also, don't send redirect if forwarding using a default route
1147 	 * or a route modified by a redirect.
1148 	 */
1149 #define	satosin(sa)	((struct sockaddr_in *)(sa))
1150 	if (rt->rt_ifp == m->m_pkthdr.rcvif &&
1151 	    (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1152 	    satosin(rt_key(rt))->sin_addr.s_addr != 0 &&
1153 	    ipsendredirects && !srcrt) {
1154 #define	RTA(rt)	((struct in_ifaddr *)(rt->rt_ifa))
1155 		u_long src = ntohl(ip->ip_src.s_addr);
1156 
1157 		if (RTA(rt) &&
1158 		    (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1159 		    if (rt->rt_flags & RTF_GATEWAY)
1160 			dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
1161 		    else
1162 			dest = ip->ip_dst.s_addr;
1163 		    /* Router requirements says to only send host redirects */
1164 		    type = ICMP_REDIRECT;
1165 		    code = ICMP_REDIRECT_HOST;
1166 #ifdef DIAGNOSTIC
1167 		    if (ipprintfs)
1168 		        printf("redirect (%d) to %lx\n", code, (u_long)dest);
1169 #endif
1170 		}
1171 	}
1172 
1173 	error = ip_output(m, (struct mbuf *)0, &ipforward_rt, IP_FORWARDING
1174 #ifdef DIRECTED_BROADCAST
1175 			    | IP_ALLOWBROADCAST
1176 #endif
1177 						, 0);
1178 	if (error)
1179 		ipstat.ips_cantforward++;
1180 	else {
1181 		ipstat.ips_forward++;
1182 		if (type)
1183 			ipstat.ips_redirectsent++;
1184 		else {
1185 			if (mcopy)
1186 				m_freem(mcopy);
1187 			return;
1188 		}
1189 	}
1190 	if (mcopy == NULL)
1191 		return;
1192 	destifp = NULL;
1193 
1194 	switch (error) {
1195 
1196 	case 0:				/* forwarded, but need redirect */
1197 		/* type, code set above */
1198 		break;
1199 
1200 	case ENETUNREACH:		/* shouldn't happen, checked above */
1201 	case EHOSTUNREACH:
1202 	case ENETDOWN:
1203 	case EHOSTDOWN:
1204 	default:
1205 		type = ICMP_UNREACH;
1206 		code = ICMP_UNREACH_HOST;
1207 		break;
1208 
1209 	case EMSGSIZE:
1210 		type = ICMP_UNREACH;
1211 		code = ICMP_UNREACH_NEEDFRAG;
1212 		if (ipforward_rt.ro_rt)
1213 			destifp = ipforward_rt.ro_rt->rt_ifp;
1214 		ipstat.ips_cantfrag++;
1215 		break;
1216 
1217 	case ENOBUFS:
1218 		type = ICMP_SOURCEQUENCH;
1219 		code = 0;
1220 		break;
1221 	}
1222 	icmp_error(mcopy, type, code, dest, destifp);
1223 }
1224 
1225 int
1226 ip_rsvp_init(struct socket *so)
1227 {
1228 	if (so->so_type != SOCK_RAW ||
1229 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
1230 	  return EOPNOTSUPP;
1231 
1232 	if (ip_rsvpd != NULL)
1233 	  return EADDRINUSE;
1234 
1235 	ip_rsvpd = so;
1236 	/*
1237 	 * This may seem silly, but we need to be sure we don't over-increment
1238 	 * the RSVP counter, in case something slips up.
1239 	 */
1240 	if (!ip_rsvp_on) {
1241 		ip_rsvp_on = 1;
1242 		rsvp_on++;
1243 	}
1244 
1245 	return 0;
1246 }
1247 
1248 int
1249 ip_rsvp_done(void)
1250 {
1251 	ip_rsvpd = NULL;
1252 	/*
1253 	 * This may seem silly, but we need to be sure we don't over-decrement
1254 	 * the RSVP counter, in case something slips up.
1255 	 */
1256 	if (ip_rsvp_on) {
1257 		ip_rsvp_on = 0;
1258 		rsvp_on--;
1259 	}
1260 	return 0;
1261 }
1262