xref: /freebsd/sys/netinet/ip_input.c (revision ef5d438ed4bc17ad7ece3e40fe4d1f9baf3aadf7)
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 1.37 1996/02/24 00:17:34 phk 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 #include <sys/sysctl.h>
49 
50 #include <net/if.h>
51 #include <net/route.h>
52 #include <net/netisr.h>
53 
54 #include <netinet/in.h>
55 #include <netinet/in_systm.h>
56 #include <netinet/in_var.h>
57 #include <netinet/ip.h>
58 #include <netinet/in_pcb.h>
59 #include <netinet/in_var.h>
60 #include <netinet/ip_var.h>
61 #include <netinet/ip_icmp.h>
62 
63 #include <netinet/ip_fw.h>
64 
65 #include <sys/socketvar.h>
66 int rsvp_on = 0;
67 static int ip_rsvp_on;
68 struct socket *ip_rsvpd;
69 
70 static int	ipforwarding = 0;
71 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW,
72 	&ipforwarding, 0, "");
73 
74 static int	ipsendredirects = 1; /* XXX */
75 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW,
76 	&ipsendredirects, 0, "");
77 
78 int	ip_defttl = IPDEFTTL;
79 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW,
80 	&ip_defttl, 0, "");
81 
82 static int	ip_dosourceroute = 0;
83 SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW,
84 	&ip_dosourceroute, 0, "");
85 #ifdef DIAGNOSTIC
86 static int	ipprintfs = 0;
87 #endif
88 
89 extern	struct domain inetdomain;
90 extern	struct protosw inetsw[];
91 u_char	ip_protox[IPPROTO_MAX];
92 static int	ipqmaxlen = IFQ_MAXLEN;
93 struct	in_ifaddr *in_ifaddr;			/* first inet address */
94 struct	ifqueue ipintrq;
95 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RD,
96 	&ipintrq.ifq_maxlen, 0, "");
97 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD,
98 	&ipintrq.ifq_drops, 0, "");
99 
100 struct ipstat ipstat;
101 static struct ipq ipq;
102 
103 #ifdef IPCTL_DEFMTU
104 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
105 	&ip_mtu, 0, "");
106 #endif
107 
108 /*
109  * The dummy IP-firewall function, and the pointer we access it through
110  */
111 static int
112 dummy_ip_fw_chk(m, ip, rif, dir)
113 	struct mbuf *m;
114 	struct ip *ip;
115 	struct ifnet *rif;
116 	int dir;
117 {
118 	return 1;
119 }
120 
121 int (*ip_fw_chk_ptr)(struct mbuf *, struct ip *, struct ifnet *, int dir) =
122 	dummy_ip_fw_chk;
123 
124 int (*ip_fw_ctl_ptr)(int, struct mbuf **);
125 
126 /*
127  * We need to save the IP options in case a protocol wants to respond
128  * to an incoming packet over the same route if the packet got here
129  * using IP source routing.  This allows connection establishment and
130  * maintenance when the remote end is on a network that is not known
131  * to us.
132  */
133 static int	ip_nhops = 0;
134 static	struct ip_srcrt {
135 	struct	in_addr dst;			/* final destination */
136 	char	nop;				/* one NOP to align */
137 	char	srcopt[IPOPT_OFFSET + 1];	/* OPTVAL, OLEN and OFFSET */
138 	struct	in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)];
139 } ip_srcrt;
140 
141 static void save_rte __P((u_char *, struct in_addr));
142 static void	 ip_deq __P((struct ipasfrag *));
143 static int	 ip_dooptions __P((struct mbuf *));
144 static void	 ip_enq __P((struct ipasfrag *, struct ipasfrag *));
145 static void	 ip_forward __P((struct mbuf *, int));
146 static void	 ip_freef __P((struct ipq *));
147 static struct ip *
148 	 ip_reass __P((struct ipasfrag *, struct ipq *));
149 static struct in_ifaddr *
150 	 ip_rtaddr __P((struct in_addr));
151 static void	 ipintr __P((void));
152 /*
153  * IP initialization: fill in IP protocol switch table.
154  * All protocols not implemented in kernel go to raw IP protocol handler.
155  */
156 void
157 ip_init()
158 {
159 	register struct protosw *pr;
160 	register int i;
161 
162 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
163 	if (pr == 0)
164 		panic("ip_init");
165 	for (i = 0; i < IPPROTO_MAX; i++)
166 		ip_protox[i] = pr - inetsw;
167 	for (pr = inetdomain.dom_protosw;
168 	    pr < inetdomain.dom_protoswNPROTOSW; pr++)
169 		if (pr->pr_domain->dom_family == PF_INET &&
170 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
171 			ip_protox[pr->pr_protocol] = pr - inetsw;
172 	ipq.next = ipq.prev = &ipq;
173 	ip_id = time.tv_sec & 0xffff;
174 	ipintrq.ifq_maxlen = ipqmaxlen;
175 #ifdef IPFIREWALL
176 	ip_fw_init();
177 #endif
178 }
179 
180 static struct	sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET };
181 static struct	route ipforward_rt;
182 
183 /*
184  * Ip input routine.  Checksum and byte swap header.  If fragmented
185  * try to reassemble.  Process options.  Pass to next level.
186  */
187 void
188 ip_input(struct mbuf *m)
189 {
190 	register struct ip *ip;
191 	register struct ipq *fp;
192 	register struct in_ifaddr *ia;
193 	int hlen;
194 
195 #ifdef	DIAGNOSTIC
196 	if ((m->m_flags & M_PKTHDR) == 0)
197 		panic("ipintr no HDR");
198 #endif
199 	/*
200 	 * If no IP addresses have been set yet but the interfaces
201 	 * are receiving, can't do anything with incoming packets yet.
202 	 */
203 	if (in_ifaddr == NULL)
204 		goto bad;
205 	ipstat.ips_total++;
206 	if (m->m_len < sizeof (struct ip) &&
207 	    (m = m_pullup(m, sizeof (struct ip))) == 0) {
208 		ipstat.ips_toosmall++;
209 		return;
210 	}
211 	ip = mtod(m, struct ip *);
212 	if (ip->ip_v != IPVERSION) {
213 		ipstat.ips_badvers++;
214 		goto bad;
215 	}
216 	hlen = ip->ip_hl << 2;
217 	if (hlen < sizeof(struct ip)) {	/* minimum header length */
218 		ipstat.ips_badhlen++;
219 		goto bad;
220 	}
221 	if (hlen > m->m_len) {
222 		if ((m = m_pullup(m, hlen)) == 0) {
223 			ipstat.ips_badhlen++;
224 			return;
225 		}
226 		ip = mtod(m, struct ip *);
227 	}
228 	ip->ip_sum = in_cksum(m, hlen);
229 	if (ip->ip_sum) {
230 		ipstat.ips_badsum++;
231 		goto bad;
232 	}
233 
234 	/*
235 	 * Convert fields to host representation.
236 	 */
237 	NTOHS(ip->ip_len);
238 	if (ip->ip_len < hlen) {
239 		ipstat.ips_badlen++;
240 		goto bad;
241 	}
242 	NTOHS(ip->ip_id);
243 	NTOHS(ip->ip_off);
244 
245 	/*
246 	 * Check that the amount of data in the buffers
247 	 * is as at least much as the IP header would have us expect.
248 	 * Trim mbufs if longer than we expect.
249 	 * Drop packet if shorter than we expect.
250 	 */
251 	if (m->m_pkthdr.len < ip->ip_len) {
252 		ipstat.ips_tooshort++;
253 		goto bad;
254 	}
255 	if (m->m_pkthdr.len > ip->ip_len) {
256 		if (m->m_len == m->m_pkthdr.len) {
257 			m->m_len = ip->ip_len;
258 			m->m_pkthdr.len = ip->ip_len;
259 		} else
260 			m_adj(m, ip->ip_len - m->m_pkthdr.len);
261 	}
262 	/*
263 	 * IpHack's section.
264 	 * Right now when no processing on packet has done
265 	 * and it is still fresh out of network we do our black
266 	 * deals with it.
267 	 * - Firewall: deny/allow
268 	 * - Wrap: fake packet's addr/port <unimpl.>
269 	 * - Encapsulate: put it in another IP and send out. <unimp.>
270  	 */
271 
272 	if (!(*ip_fw_chk_ptr)(m,ip,m->m_pkthdr.rcvif,0))
273 	       return;
274 
275 	/*
276 	 * Process options and, if not destined for us,
277 	 * ship it on.  ip_dooptions returns 1 when an
278 	 * error was detected (causing an icmp message
279 	 * to be sent and the original packet to be freed).
280 	 */
281 	ip_nhops = 0;		/* for source routed packets */
282 	if (hlen > sizeof (struct ip) && ip_dooptions(m))
283 		return;
284 
285         /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
286          * matter if it is destined to another node, or whether it is
287          * a multicast one, RSVP wants it! and prevents it from being forwarded
288          * anywhere else. Also checks if the rsvp daemon is running before
289 	 * grabbing the packet.
290          */
291 	if (rsvp_on && ip->ip_p==IPPROTO_RSVP)
292 		goto ours;
293 
294 	/*
295 	 * Check our list of addresses, to see if the packet is for us.
296 	 */
297 	for (ia = in_ifaddr; ia; ia = ia->ia_next) {
298 #define	satosin(sa)	((struct sockaddr_in *)(sa))
299 
300 		if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr)
301 			goto ours;
302 		if (ia->ia_ifp->if_flags & IFF_BROADCAST) {
303 #if 0
304 			u_long t;
305 #endif
306 
307 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
308 			    ip->ip_dst.s_addr)
309 				goto ours;
310 			if (ip->ip_dst.s_addr == ia->ia_netbroadcast.s_addr)
311 				goto ours;
312 #if 0 /* XXX - this should go away */
313 			/*
314 			 * Look for all-0's host part (old broadcast addr),
315 			 * either for subnet or net.
316 			 */
317 			t = ntohl(ip->ip_dst.s_addr);
318 			if (t == ia->ia_subnet)
319 				goto ours;
320 			if (t == ia->ia_net)
321 				goto ours;
322 #endif /* compatibility cruft */
323 		}
324 	}
325 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
326 		struct in_multi *inm;
327 		if (ip_mrouter) {
328 			/*
329 			 * If we are acting as a multicast router, all
330 			 * incoming multicast packets are passed to the
331 			 * kernel-level multicast forwarding function.
332 			 * The packet is returned (relatively) intact; if
333 			 * ip_mforward() returns a non-zero value, the packet
334 			 * must be discarded, else it may be accepted below.
335 			 *
336 			 * (The IP ident field is put in the same byte order
337 			 * as expected when ip_mforward() is called from
338 			 * ip_output().)
339 			 */
340 			ip->ip_id = htons(ip->ip_id);
341 			if (ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) {
342 				ipstat.ips_cantforward++;
343 				m_freem(m);
344 				return;
345 			}
346 			ip->ip_id = ntohs(ip->ip_id);
347 
348 			/*
349 			 * The process-level routing demon needs to receive
350 			 * all multicast IGMP packets, whether or not this
351 			 * host belongs to their destination groups.
352 			 */
353 			if (ip->ip_p == IPPROTO_IGMP)
354 				goto ours;
355 			ipstat.ips_forward++;
356 		}
357 		/*
358 		 * See if we belong to the destination multicast group on the
359 		 * arrival interface.
360 		 */
361 		IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
362 		if (inm == NULL) {
363 			ipstat.ips_cantforward++;
364 			m_freem(m);
365 			return;
366 		}
367 		goto ours;
368 	}
369 	if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
370 		goto ours;
371 	if (ip->ip_dst.s_addr == INADDR_ANY)
372 		goto ours;
373 
374 	/*
375 	 * Not for us; forward if possible and desirable.
376 	 */
377 	if (ipforwarding == 0) {
378 		ipstat.ips_cantforward++;
379 		m_freem(m);
380 	} else
381 		ip_forward(m, 0);
382 	return;
383 
384 ours:
385 
386 	/*
387 	 * If offset or IP_MF are set, must reassemble.
388 	 * Otherwise, nothing need be done.
389 	 * (We could look in the reassembly queue to see
390 	 * if the packet was previously fragmented,
391 	 * but it's not worth the time; just let them time out.)
392 	 */
393 	if (ip->ip_off &~ IP_DF) {
394 		if (m->m_flags & M_EXT) {		/* XXX */
395 			if ((m = m_pullup(m, sizeof (struct ip))) == 0) {
396 				ipstat.ips_toosmall++;
397 				return;
398 			}
399 			ip = mtod(m, struct ip *);
400 		}
401 		/*
402 		 * Look for queue of fragments
403 		 * of this datagram.
404 		 */
405 		for (fp = ipq.next; fp != &ipq; fp = fp->next)
406 			if (ip->ip_id == fp->ipq_id &&
407 			    ip->ip_src.s_addr == fp->ipq_src.s_addr &&
408 			    ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
409 			    ip->ip_p == fp->ipq_p)
410 				goto found;
411 		fp = 0;
412 found:
413 
414 		/*
415 		 * Adjust ip_len to not reflect header,
416 		 * set ip_mff if more fragments are expected,
417 		 * convert offset of this to bytes.
418 		 */
419 		ip->ip_len -= hlen;
420 		((struct ipasfrag *)ip)->ipf_mff &= ~1;
421 		if (ip->ip_off & IP_MF)
422 			((struct ipasfrag *)ip)->ipf_mff |= 1;
423 		ip->ip_off <<= 3;
424 
425 		/*
426 		 * If datagram marked as having more fragments
427 		 * or if this is not the first fragment,
428 		 * attempt reassembly; if it succeeds, proceed.
429 		 */
430 		if (((struct ipasfrag *)ip)->ipf_mff & 1 || ip->ip_off) {
431 			ipstat.ips_fragments++;
432 			ip = ip_reass((struct ipasfrag *)ip, fp);
433 			if (ip == 0)
434 				return;
435 			ipstat.ips_reassembled++;
436 			m = dtom(ip);
437 		} else
438 			if (fp)
439 				ip_freef(fp);
440 	} else
441 		ip->ip_len -= hlen;
442 
443 	/*
444 	 * Switch out to protocol's input routine.
445 	 */
446 	ipstat.ips_delivered++;
447 	(*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen);
448 	return;
449 bad:
450 	m_freem(m);
451 }
452 
453 /*
454  * IP software interrupt routine - to go away sometime soon
455  */
456 static void
457 ipintr(void)
458 {
459 	int s;
460 	struct mbuf *m;
461 
462 	while(1) {
463 		s = splimp();
464 		IF_DEQUEUE(&ipintrq, m);
465 		splx(s);
466 		if (m == 0)
467 			return;
468 		ip_input(m);
469 	}
470 }
471 
472 NETISR_SET(NETISR_IP, ipintr);
473 
474 /*
475  * Take incoming datagram fragment and try to
476  * reassemble it into whole datagram.  If a chain for
477  * reassembly of this datagram already exists, then it
478  * is given as fp; otherwise have to make a chain.
479  */
480 static struct ip *
481 ip_reass(ip, fp)
482 	register struct ipasfrag *ip;
483 	register struct ipq *fp;
484 {
485 	register struct mbuf *m = dtom(ip);
486 	register struct ipasfrag *q;
487 	struct mbuf *t;
488 	int hlen = ip->ip_hl << 2;
489 	int i, next;
490 
491 	/*
492 	 * Presence of header sizes in mbufs
493 	 * would confuse code below.
494 	 */
495 	m->m_data += hlen;
496 	m->m_len -= hlen;
497 
498 	/*
499 	 * If first fragment to arrive, create a reassembly queue.
500 	 */
501 	if (fp == 0) {
502 		if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL)
503 			goto dropfrag;
504 		fp = mtod(t, struct ipq *);
505 		insque(fp, &ipq);
506 		fp->ipq_ttl = IPFRAGTTL;
507 		fp->ipq_p = ip->ip_p;
508 		fp->ipq_id = ip->ip_id;
509 		fp->ipq_next = fp->ipq_prev = (struct ipasfrag *)fp;
510 		fp->ipq_src = ((struct ip *)ip)->ip_src;
511 		fp->ipq_dst = ((struct ip *)ip)->ip_dst;
512 		q = (struct ipasfrag *)fp;
513 		goto insert;
514 	}
515 
516 	/*
517 	 * Find a segment which begins after this one does.
518 	 */
519 	for (q = fp->ipq_next; q != (struct ipasfrag *)fp; q = q->ipf_next)
520 		if (q->ip_off > ip->ip_off)
521 			break;
522 
523 	/*
524 	 * If there is a preceding segment, it may provide some of
525 	 * our data already.  If so, drop the data from the incoming
526 	 * segment.  If it provides all of our data, drop us.
527 	 */
528 	if (q->ipf_prev != (struct ipasfrag *)fp) {
529 		i = q->ipf_prev->ip_off + q->ipf_prev->ip_len - ip->ip_off;
530 		if (i > 0) {
531 			if (i >= ip->ip_len)
532 				goto dropfrag;
533 			m_adj(dtom(ip), i);
534 			ip->ip_off += i;
535 			ip->ip_len -= i;
536 		}
537 	}
538 
539 	/*
540 	 * While we overlap succeeding segments trim them or,
541 	 * if they are completely covered, dequeue them.
542 	 */
543 	while (q != (struct ipasfrag *)fp && ip->ip_off + ip->ip_len > q->ip_off) {
544 		i = (ip->ip_off + ip->ip_len) - q->ip_off;
545 		if (i < q->ip_len) {
546 			q->ip_len -= i;
547 			q->ip_off += i;
548 			m_adj(dtom(q), i);
549 			break;
550 		}
551 		q = q->ipf_next;
552 		m_freem(dtom(q->ipf_prev));
553 		ip_deq(q->ipf_prev);
554 	}
555 
556 insert:
557 	/*
558 	 * Stick new segment in its place;
559 	 * check for complete reassembly.
560 	 */
561 	ip_enq(ip, q->ipf_prev);
562 	next = 0;
563 	for (q = fp->ipq_next; q != (struct ipasfrag *)fp; q = q->ipf_next) {
564 		if (q->ip_off != next)
565 			return (0);
566 		next += q->ip_len;
567 	}
568 	if (q->ipf_prev->ipf_mff & 1)
569 		return (0);
570 
571 	/*
572 	 * Reassembly is complete; concatenate fragments.
573 	 */
574 	q = fp->ipq_next;
575 	m = dtom(q);
576 	t = m->m_next;
577 	m->m_next = 0;
578 	m_cat(m, t);
579 	q = q->ipf_next;
580 	while (q != (struct ipasfrag *)fp) {
581 		t = dtom(q);
582 		q = q->ipf_next;
583 		m_cat(m, t);
584 	}
585 
586 	/*
587 	 * Create header for new ip packet by
588 	 * modifying header of first packet;
589 	 * dequeue and discard fragment reassembly header.
590 	 * Make header visible.
591 	 */
592 	ip = fp->ipq_next;
593 	ip->ip_len = next;
594 	ip->ipf_mff &= ~1;
595 	((struct ip *)ip)->ip_src = fp->ipq_src;
596 	((struct ip *)ip)->ip_dst = fp->ipq_dst;
597 	remque(fp);
598 	(void) m_free(dtom(fp));
599 	m = dtom(ip);
600 	m->m_len += (ip->ip_hl << 2);
601 	m->m_data -= (ip->ip_hl << 2);
602 	/* some debugging cruft by sklower, below, will go away soon */
603 	if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */
604 		register int plen = 0;
605 		for (t = m; m; m = m->m_next)
606 			plen += m->m_len;
607 		t->m_pkthdr.len = plen;
608 	}
609 	return ((struct ip *)ip);
610 
611 dropfrag:
612 	ipstat.ips_fragdropped++;
613 	m_freem(m);
614 	return (0);
615 }
616 
617 /*
618  * Free a fragment reassembly header and all
619  * associated datagrams.
620  */
621 static void
622 ip_freef(fp)
623 	struct ipq *fp;
624 {
625 	register struct ipasfrag *q, *p;
626 
627 	for (q = fp->ipq_next; q != (struct ipasfrag *)fp; q = p) {
628 		p = q->ipf_next;
629 		ip_deq(q);
630 		m_freem(dtom(q));
631 	}
632 	remque(fp);
633 	(void) m_free(dtom(fp));
634 }
635 
636 /*
637  * Put an ip fragment on a reassembly chain.
638  * Like insque, but pointers in middle of structure.
639  */
640 static void
641 ip_enq(p, prev)
642 	register struct ipasfrag *p, *prev;
643 {
644 
645 	p->ipf_prev = prev;
646 	p->ipf_next = prev->ipf_next;
647 	prev->ipf_next->ipf_prev = p;
648 	prev->ipf_next = p;
649 }
650 
651 /*
652  * To ip_enq as remque is to insque.
653  */
654 static void
655 ip_deq(p)
656 	register struct ipasfrag *p;
657 {
658 
659 	p->ipf_prev->ipf_next = p->ipf_next;
660 	p->ipf_next->ipf_prev = p->ipf_prev;
661 }
662 
663 /*
664  * IP timer processing;
665  * if a timer expires on a reassembly
666  * queue, discard it.
667  */
668 void
669 ip_slowtimo()
670 {
671 	register struct ipq *fp;
672 	int s = splnet();
673 
674 	fp = ipq.next;
675 	if (fp == 0) {
676 		splx(s);
677 		return;
678 	}
679 	while (fp != &ipq) {
680 		--fp->ipq_ttl;
681 		fp = fp->next;
682 		if (fp->prev->ipq_ttl == 0) {
683 			ipstat.ips_fragtimeout++;
684 			ip_freef(fp->prev);
685 		}
686 	}
687 	splx(s);
688 }
689 
690 /*
691  * Drain off all datagram fragments.
692  */
693 void
694 ip_drain()
695 {
696 	while (ipq.next != &ipq) {
697 		ipstat.ips_fragdropped++;
698 		ip_freef(ipq.next);
699 	}
700 
701 	in_rtqdrain();
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,
1174 			  IP_FORWARDING, 0);
1175 	if (error)
1176 		ipstat.ips_cantforward++;
1177 	else {
1178 		ipstat.ips_forward++;
1179 		if (type)
1180 			ipstat.ips_redirectsent++;
1181 		else {
1182 			if (mcopy)
1183 				m_freem(mcopy);
1184 			return;
1185 		}
1186 	}
1187 	if (mcopy == NULL)
1188 		return;
1189 	destifp = NULL;
1190 
1191 	switch (error) {
1192 
1193 	case 0:				/* forwarded, but need redirect */
1194 		/* type, code set above */
1195 		break;
1196 
1197 	case ENETUNREACH:		/* shouldn't happen, checked above */
1198 	case EHOSTUNREACH:
1199 	case ENETDOWN:
1200 	case EHOSTDOWN:
1201 	default:
1202 		type = ICMP_UNREACH;
1203 		code = ICMP_UNREACH_HOST;
1204 		break;
1205 
1206 	case EMSGSIZE:
1207 		type = ICMP_UNREACH;
1208 		code = ICMP_UNREACH_NEEDFRAG;
1209 		if (ipforward_rt.ro_rt)
1210 			destifp = ipforward_rt.ro_rt->rt_ifp;
1211 		ipstat.ips_cantfrag++;
1212 		break;
1213 
1214 	case ENOBUFS:
1215 		type = ICMP_SOURCEQUENCH;
1216 		code = 0;
1217 		break;
1218 	}
1219 	icmp_error(mcopy, type, code, dest, destifp);
1220 }
1221 
1222 int
1223 ip_rsvp_init(struct socket *so)
1224 {
1225 	if (so->so_type != SOCK_RAW ||
1226 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
1227 	  return EOPNOTSUPP;
1228 
1229 	if (ip_rsvpd != NULL)
1230 	  return EADDRINUSE;
1231 
1232 	ip_rsvpd = so;
1233 	/*
1234 	 * This may seem silly, but we need to be sure we don't over-increment
1235 	 * the RSVP counter, in case something slips up.
1236 	 */
1237 	if (!ip_rsvp_on) {
1238 		ip_rsvp_on = 1;
1239 		rsvp_on++;
1240 	}
1241 
1242 	return 0;
1243 }
1244 
1245 int
1246 ip_rsvp_done(void)
1247 {
1248 	ip_rsvpd = NULL;
1249 	/*
1250 	 * This may seem silly, but we need to be sure we don't over-decrement
1251 	 * the RSVP counter, in case something slips up.
1252 	 */
1253 	if (ip_rsvp_on) {
1254 		ip_rsvp_on = 0;
1255 		rsvp_on--;
1256 	}
1257 	return 0;
1258 }
1259