xref: /freebsd/sys/netinet/ip_input.c (revision 6e8394b8baa7d5d9153ab90de6824bcd19b3b4e1)
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 1.119 1999/05/04 16:20:32 luigi Exp $
35  */
36 
37 #define	_IP_VHL
38 
39 #include "opt_bootp.h"
40 #include "opt_ipfw.h"
41 #include "opt_ipdn.h"
42 #include "opt_ipdivert.h"
43 #include "opt_ipfilter.h"
44 
45 #include <stddef.h>
46 
47 #include <sys/param.h>
48 #include <sys/systm.h>
49 #include <sys/mbuf.h>
50 #include <sys/malloc.h>
51 #include <sys/domain.h>
52 #include <sys/protosw.h>
53 #include <sys/socket.h>
54 #include <sys/time.h>
55 #include <sys/kernel.h>
56 #include <sys/syslog.h>
57 #include <sys/sysctl.h>
58 
59 #include <net/if.h>
60 #include <net/if_var.h>
61 #include <net/if_dl.h>
62 #include <net/route.h>
63 #include <net/netisr.h>
64 
65 #include <netinet/in.h>
66 #include <netinet/in_systm.h>
67 #include <netinet/in_var.h>
68 #include <netinet/ip.h>
69 #include <netinet/in_pcb.h>
70 #include <netinet/ip_var.h>
71 #include <netinet/ip_icmp.h>
72 #include <machine/in_cksum.h>
73 
74 #include <sys/socketvar.h>
75 
76 #include <netinet/ip_fw.h>
77 
78 #ifdef DUMMYNET
79 #include <netinet/ip_dummynet.h>
80 #endif
81 
82 int rsvp_on = 0;
83 static int ip_rsvp_on;
84 struct socket *ip_rsvpd;
85 
86 int	ipforwarding = 0;
87 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW,
88     &ipforwarding, 0, "Enable IP forwarding between interfaces");
89 
90 static int	ipsendredirects = 1; /* XXX */
91 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW,
92     &ipsendredirects, 0, "Enable sending IP redirects");
93 
94 int	ip_defttl = IPDEFTTL;
95 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW,
96     &ip_defttl, 0, "Maximum TTL on IP packets");
97 
98 static int	ip_dosourceroute = 0;
99 SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW,
100     &ip_dosourceroute, 0, "Enable forwarding source routed IP packets");
101 
102 static int	ip_acceptsourceroute = 0;
103 SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute,
104     CTLFLAG_RW, &ip_acceptsourceroute, 0,
105     "Enable accepting source routed IP packets");
106 #ifdef DIAGNOSTIC
107 static int	ipprintfs = 0;
108 #endif
109 
110 extern	struct domain inetdomain;
111 extern	struct protosw inetsw[];
112 u_char	ip_protox[IPPROTO_MAX];
113 static int	ipqmaxlen = IFQ_MAXLEN;
114 struct	in_ifaddrhead in_ifaddrhead; /* first inet address */
115 struct	ifqueue ipintrq;
116 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW,
117     &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue");
118 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD,
119     &ipintrq.ifq_drops, 0, "Number of packets dropped from the IP input queue");
120 
121 struct ipstat ipstat;
122 SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RD,
123     &ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)");
124 
125 /* Packet reassembly stuff */
126 #define IPREASS_NHASH_LOG2      6
127 #define IPREASS_NHASH           (1 << IPREASS_NHASH_LOG2)
128 #define IPREASS_HMASK           (IPREASS_NHASH - 1)
129 #define IPREASS_HASH(x,y) \
130 	(((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK)
131 
132 static struct ipq ipq[IPREASS_NHASH];
133 static int    nipq = 0;         /* total # of reass queues */
134 static int    maxnipq;
135 
136 #ifdef IPCTL_DEFMTU
137 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
138     &ip_mtu, 0, "Default MTU");
139 #endif
140 
141 #ifdef IPSTEALTH
142 static int	ipstealth = 0;
143 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW,
144     &ipstealth, 0, "");
145 #endif
146 
147 
148 /* Firewall hooks */
149 ip_fw_chk_t *ip_fw_chk_ptr;
150 ip_fw_ctl_t *ip_fw_ctl_ptr;
151 
152 #ifdef DUMMYNET
153 ip_dn_ctl_t *ip_dn_ctl_ptr;
154 #endif
155 
156 #if defined(IPFILTER_LKM) || defined(IPFILTER)
157 int iplattach __P((void));
158 int (*fr_checkp) __P((struct ip *, int, struct ifnet *, int, struct mbuf **)) = NULL;
159 #endif
160 
161 
162 /*
163  * We need to save the IP options in case a protocol wants to respond
164  * to an incoming packet over the same route if the packet got here
165  * using IP source routing.  This allows connection establishment and
166  * maintenance when the remote end is on a network that is not known
167  * to us.
168  */
169 static int	ip_nhops = 0;
170 static	struct ip_srcrt {
171 	struct	in_addr dst;			/* final destination */
172 	char	nop;				/* one NOP to align */
173 	char	srcopt[IPOPT_OFFSET + 1];	/* OPTVAL, OLEN and OFFSET */
174 	struct	in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)];
175 } ip_srcrt;
176 
177 #ifdef IPDIVERT
178 /*
179  * Shared variable between ip_input() and ip_reass() to communicate
180  * about which packets, once assembled from fragments, get diverted,
181  * and to which port.
182  */
183 static u_short	frag_divert_port;
184 #endif
185 
186 struct sockaddr_in *ip_fw_fwd_addr;
187 
188 static void save_rte __P((u_char *, struct in_addr));
189 static int	 ip_dooptions __P((struct mbuf *));
190 static void	 ip_forward __P((struct mbuf *, int));
191 static void	 ip_freef __P((struct ipq *));
192 static struct ip *
193 	 ip_reass __P((struct mbuf *, struct ipq *, struct ipq *));
194 static struct in_ifaddr *
195 	 ip_rtaddr __P((struct in_addr));
196 static void	ipintr __P((void));
197 /*
198  * IP initialization: fill in IP protocol switch table.
199  * All protocols not implemented in kernel go to raw IP protocol handler.
200  */
201 void
202 ip_init()
203 {
204 	register struct protosw *pr;
205 	register int i;
206 
207 	TAILQ_INIT(&in_ifaddrhead);
208 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
209 	if (pr == 0)
210 		panic("ip_init");
211 	for (i = 0; i < IPPROTO_MAX; i++)
212 		ip_protox[i] = pr - inetsw;
213 	for (pr = inetdomain.dom_protosw;
214 	    pr < inetdomain.dom_protoswNPROTOSW; pr++)
215 		if (pr->pr_domain->dom_family == PF_INET &&
216 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
217 			ip_protox[pr->pr_protocol] = pr - inetsw;
218 
219 	for (i = 0; i < IPREASS_NHASH; i++)
220 	    ipq[i].next = ipq[i].prev = &ipq[i];
221 
222 	maxnipq = nmbclusters/4;
223 
224 	ip_id = time_second & 0xffff;
225 	ipintrq.ifq_maxlen = ipqmaxlen;
226 #ifdef IPFILTER
227         iplattach();
228 #endif
229 
230 }
231 
232 static struct	sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET };
233 static struct	route ipforward_rt;
234 
235 /*
236  * Ip input routine.  Checksum and byte swap header.  If fragmented
237  * try to reassemble.  Process options.  Pass to next level.
238  */
239 void
240 ip_input(struct mbuf *m)
241 {
242 	struct ip *ip;
243 	struct ipq *fp;
244 	struct in_ifaddr *ia;
245 	int    i, hlen, mff;
246 	u_short sum;
247 #ifndef IPDIVERT /* dummy variable for the firewall code to play with */
248         u_short ip_divert_cookie = 0 ;
249 #endif
250 	struct ip_fw_chain *rule = NULL ;
251 
252 #if defined(IPFIREWALL) && defined(DUMMYNET)
253         /*
254          * dummynet packet are prepended a vestigial mbuf with
255          * m_type = MT_DUMMYNET and m_data pointing to the matching
256          * rule.
257          */
258         if (m->m_type == MT_DUMMYNET) {
259             rule = (struct ip_fw_chain *)(m->m_data) ;
260             m = m->m_next ;
261             ip = mtod(m, struct ip *);
262             hlen = IP_VHL_HL(ip->ip_vhl) << 2;
263             goto iphack ;
264         } else
265             rule = NULL ;
266 #endif
267 
268 #ifdef	DIAGNOSTIC
269 	if (m == NULL || (m->m_flags & M_PKTHDR) == 0)
270 		panic("ip_input no HDR");
271 #endif
272 	ipstat.ips_total++;
273 
274 	if (m->m_pkthdr.len < sizeof(struct ip))
275 		goto tooshort;
276 
277 	if (m->m_len < sizeof (struct ip) &&
278 	    (m = m_pullup(m, sizeof (struct ip))) == 0) {
279 		ipstat.ips_toosmall++;
280 		return;
281 	}
282 	ip = mtod(m, struct ip *);
283 
284 	if (IP_VHL_V(ip->ip_vhl) != IPVERSION) {
285 		ipstat.ips_badvers++;
286 		goto bad;
287 	}
288 
289 	hlen = IP_VHL_HL(ip->ip_vhl) << 2;
290 	if (hlen < sizeof(struct ip)) {	/* minimum header length */
291 		ipstat.ips_badhlen++;
292 		goto bad;
293 	}
294 	if (hlen > m->m_len) {
295 		if ((m = m_pullup(m, hlen)) == 0) {
296 			ipstat.ips_badhlen++;
297 			return;
298 		}
299 		ip = mtod(m, struct ip *);
300 	}
301 	if (hlen == sizeof(struct ip)) {
302 		sum = in_cksum_hdr(ip);
303 	} else {
304 		sum = in_cksum(m, hlen);
305 	}
306 	if (sum) {
307 		ipstat.ips_badsum++;
308 		goto bad;
309 	}
310 
311 	/*
312 	 * Convert fields to host representation.
313 	 */
314 	NTOHS(ip->ip_len);
315 	if (ip->ip_len < hlen) {
316 		ipstat.ips_badlen++;
317 		goto bad;
318 	}
319 	NTOHS(ip->ip_id);
320 	NTOHS(ip->ip_off);
321 
322 	/*
323 	 * Check that the amount of data in the buffers
324 	 * is as at least much as the IP header would have us expect.
325 	 * Trim mbufs if longer than we expect.
326 	 * Drop packet if shorter than we expect.
327 	 */
328 	if (m->m_pkthdr.len < ip->ip_len) {
329 tooshort:
330 		ipstat.ips_tooshort++;
331 		goto bad;
332 	}
333 	if (m->m_pkthdr.len > ip->ip_len) {
334 		if (m->m_len == m->m_pkthdr.len) {
335 			m->m_len = ip->ip_len;
336 			m->m_pkthdr.len = ip->ip_len;
337 		} else
338 			m_adj(m, ip->ip_len - m->m_pkthdr.len);
339 	}
340 	/*
341 	 * IpHack's section.
342 	 * Right now when no processing on packet has done
343 	 * and it is still fresh out of network we do our black
344 	 * deals with it.
345 	 * - Firewall: deny/allow/divert
346 	 * - Xlate: translate packet's addr/port (NAT).
347 	 * - Pipe: pass pkt through dummynet.
348 	 * - Wrap: fake packet's addr/port <unimpl.>
349 	 * - Encapsulate: put it in another IP and send out. <unimp.>
350  	 */
351 
352 #if defined(IPFIREWALL) && defined(DUMMYNET)
353 iphack:
354 #endif
355 #if defined(IPFILTER) || defined(IPFILTER_LKM)
356 	/*
357 	 * Check if we want to allow this packet to be processed.
358 	 * Consider it to be bad if not.
359 	 */
360 	if (fr_checkp) {
361 		struct	mbuf	*m1 = m;
362 
363 		if ((*fr_checkp)(ip, hlen, m->m_pkthdr.rcvif, 0, &m1) || !m1)
364 			return;
365 		ip = mtod(m = m1, struct ip *);
366 	}
367 #endif
368 	if (ip_fw_chk_ptr) {
369 #ifdef IPFIREWALL_FORWARD
370 		/*
371 		 * If we've been forwarded from the output side, then
372 		 * skip the firewall a second time
373 		 */
374 		if (ip_fw_fwd_addr)
375 			goto ours;
376 #endif	/* IPFIREWALL_FORWARD */
377 		i = (*ip_fw_chk_ptr)(&ip, hlen, NULL, &ip_divert_cookie,
378 					&m, &rule, &ip_fw_fwd_addr);
379 		/*
380 		 * see the comment in ip_output for the return values
381 		 * produced by the firewall.
382 		 */
383 		if (!m) /* packet discarded by firewall */
384 			return ;
385 		if (i == 0 && ip_fw_fwd_addr == NULL) /* common case */
386 			goto pass ;
387 #ifdef DUMMYNET
388                 if (i & 0x10000) {
389                         /* send packet to the appropriate pipe */
390                         dummynet_io(i&0xffff,DN_TO_IP_IN,m,NULL,NULL,0, rule);
391 			return ;
392 		}
393 #endif
394 #ifdef IPDIVERT
395 		if (i > 0 && i < 0x10000) {
396 			/* Divert packet */
397 			frag_divert_port = i & 0xffff ;
398 			goto ours;
399 		}
400 #endif
401 #ifdef IPFIREWALL_FORWARD
402 		if (i == 0 && ip_fw_fwd_addr != NULL)
403 			goto pass ;
404 #endif
405 		/*
406 		 * if we get here, the packet must be dropped
407 		 */
408 			m_freem(m);
409 			return;
410 	}
411 pass:
412 
413 	/*
414 	 * Process options and, if not destined for us,
415 	 * ship it on.  ip_dooptions returns 1 when an
416 	 * error was detected (causing an icmp message
417 	 * to be sent and the original packet to be freed).
418 	 */
419 	ip_nhops = 0;		/* for source routed packets */
420 	if (hlen > sizeof (struct ip) && ip_dooptions(m)) {
421 #ifdef IPFIREWALL_FORWARD
422 		ip_fw_fwd_addr = NULL;
423 #endif
424 		return;
425 	}
426 
427         /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
428          * matter if it is destined to another node, or whether it is
429          * a multicast one, RSVP wants it! and prevents it from being forwarded
430          * anywhere else. Also checks if the rsvp daemon is running before
431 	 * grabbing the packet.
432          */
433 	if (rsvp_on && ip->ip_p==IPPROTO_RSVP)
434 		goto ours;
435 
436 	/*
437 	 * Check our list of addresses, to see if the packet is for us.
438 	 * If we don't have any addresses, assume any unicast packet
439 	 * we receive might be for us (and let the upper layers deal
440 	 * with it).
441 	 */
442 	if (TAILQ_EMPTY(&in_ifaddrhead) &&
443 	    (m->m_flags & (M_MCAST|M_BCAST)) == 0)
444 		goto ours;
445 
446 	for (ia = TAILQ_FIRST(&in_ifaddrhead); ia;
447 					ia = TAILQ_NEXT(ia, ia_link)) {
448 #define	satosin(sa)	((struct sockaddr_in *)(sa))
449 
450 #ifdef BOOTP_COMPAT
451 		if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY)
452 			goto ours;
453 #endif
454 #ifdef IPFIREWALL_FORWARD
455 		/*
456 		 * If the addr to forward to is one of ours, we pretend to
457 		 * be the destination for this packet.
458 		 */
459 		if (ip_fw_fwd_addr == NULL) {
460 			if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr)
461 				goto ours;
462 		} else if (IA_SIN(ia)->sin_addr.s_addr ==
463 					 ip_fw_fwd_addr->sin_addr.s_addr)
464 			goto ours;
465 #else
466 		if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr)
467 			goto ours;
468 #endif
469 		if (ia->ia_ifp && ia->ia_ifp->if_flags & IFF_BROADCAST) {
470 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
471 			    ip->ip_dst.s_addr)
472 				goto ours;
473 			if (ip->ip_dst.s_addr == ia->ia_netbroadcast.s_addr)
474 				goto ours;
475 		}
476 	}
477 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
478 		struct in_multi *inm;
479 		if (ip_mrouter) {
480 			/*
481 			 * If we are acting as a multicast router, all
482 			 * incoming multicast packets are passed to the
483 			 * kernel-level multicast forwarding function.
484 			 * The packet is returned (relatively) intact; if
485 			 * ip_mforward() returns a non-zero value, the packet
486 			 * must be discarded, else it may be accepted below.
487 			 *
488 			 * (The IP ident field is put in the same byte order
489 			 * as expected when ip_mforward() is called from
490 			 * ip_output().)
491 			 */
492 			ip->ip_id = htons(ip->ip_id);
493 			if (ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) {
494 				ipstat.ips_cantforward++;
495 				m_freem(m);
496 				return;
497 			}
498 			ip->ip_id = ntohs(ip->ip_id);
499 
500 			/*
501 			 * The process-level routing demon needs to receive
502 			 * all multicast IGMP packets, whether or not this
503 			 * host belongs to their destination groups.
504 			 */
505 			if (ip->ip_p == IPPROTO_IGMP)
506 				goto ours;
507 			ipstat.ips_forward++;
508 		}
509 		/*
510 		 * See if we belong to the destination multicast group on the
511 		 * arrival interface.
512 		 */
513 		IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
514 		if (inm == NULL) {
515 			ipstat.ips_notmember++;
516 			m_freem(m);
517 			return;
518 		}
519 		goto ours;
520 	}
521 	if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
522 		goto ours;
523 	if (ip->ip_dst.s_addr == INADDR_ANY)
524 		goto ours;
525 
526 	/*
527 	 * Not for us; forward if possible and desirable.
528 	 */
529 	if (ipforwarding == 0) {
530 		ipstat.ips_cantforward++;
531 		m_freem(m);
532 	} else
533 		ip_forward(m, 0);
534 #ifdef IPFIREWALL_FORWARD
535 	ip_fw_fwd_addr = NULL;
536 #endif
537 	return;
538 
539 ours:
540 
541 	/*
542 	 * If offset or IP_MF are set, must reassemble.
543 	 * Otherwise, nothing need be done.
544 	 * (We could look in the reassembly queue to see
545 	 * if the packet was previously fragmented,
546 	 * but it's not worth the time; just let them time out.)
547 	 */
548 	if (ip->ip_off & (IP_MF | IP_OFFMASK | IP_RF)) {
549 		if (m->m_flags & M_EXT) {		/* XXX */
550 			if ((m = m_pullup(m, hlen)) == 0) {
551 				ipstat.ips_toosmall++;
552 #ifdef IPDIVERT
553 				frag_divert_port = 0;
554 				ip_divert_cookie = 0;
555 #endif
556 #ifdef IPFIREWALL_FORWARD
557 				ip_fw_fwd_addr = NULL;
558 #endif
559 				return;
560 			}
561 			ip = mtod(m, struct ip *);
562 		}
563 		sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
564 		/*
565 		 * Look for queue of fragments
566 		 * of this datagram.
567 		 */
568 		for (fp = ipq[sum].next; fp != &ipq[sum]; fp = fp->next)
569 			if (ip->ip_id == fp->ipq_id &&
570 			    ip->ip_src.s_addr == fp->ipq_src.s_addr &&
571 			    ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
572 			    ip->ip_p == fp->ipq_p)
573 				goto found;
574 
575 		fp = 0;
576 
577 		/* check if there's a place for the new queue */
578 		if (nipq > maxnipq) {
579 		    /*
580 		     * drop something from the tail of the current queue
581 		     * before proceeding further
582 		     */
583 		    if (ipq[sum].prev == &ipq[sum]) {   /* gak */
584 			for (i = 0; i < IPREASS_NHASH; i++) {
585 			    if (ipq[i].prev != &ipq[i]) {
586 				ip_freef(ipq[i].prev);
587 				break;
588 			    }
589 			}
590 		    } else
591 			ip_freef(ipq[sum].prev);
592 		}
593 found:
594 		/*
595 		 * Adjust ip_len to not reflect header,
596 		 * set ip_mff if more fragments are expected,
597 		 * convert offset of this to bytes.
598 		 */
599 		ip->ip_len -= hlen;
600 		mff = (ip->ip_off & IP_MF) != 0;
601 		if (mff) {
602 		        /*
603 		         * Make sure that fragments have a data length
604 			 * that's a non-zero multiple of 8 bytes.
605 		         */
606 			if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
607 				ipstat.ips_toosmall++; /* XXX */
608 				goto bad;
609 			}
610 			m->m_flags |= M_FRAG;
611 		}
612 		ip->ip_off <<= 3;
613 
614 		/*
615 		 * If datagram marked as having more fragments
616 		 * or if this is not the first fragment,
617 		 * attempt reassembly; if it succeeds, proceed.
618 		 */
619 		if (mff || ip->ip_off) {
620 			ipstat.ips_fragments++;
621 			m->m_pkthdr.header = ip;
622 			ip = ip_reass(m, fp, &ipq[sum]);
623 			if (ip == 0) {
624 #ifdef	IPFIREWALL_FORWARD
625 				ip_fw_fwd_addr = NULL;
626 #endif
627 				return;
628 			}
629 			/* Get the length of the reassembled packets header */
630 			hlen = IP_VHL_HL(ip->ip_vhl) << 2;
631 			ipstat.ips_reassembled++;
632 			m = dtom(ip);
633 #ifdef IPDIVERT
634 			if (frag_divert_port) {
635 				ip->ip_len += hlen;
636 				HTONS(ip->ip_len);
637 				HTONS(ip->ip_off);
638 				HTONS(ip->ip_id);
639 				ip->ip_sum = 0;
640 				ip->ip_sum = in_cksum_hdr(ip);
641 				NTOHS(ip->ip_id);
642 				NTOHS(ip->ip_off);
643 				NTOHS(ip->ip_len);
644 				ip->ip_len -= hlen;
645 			}
646 #endif
647 		} else
648 			if (fp)
649 				ip_freef(fp);
650 	} else
651 		ip->ip_len -= hlen;
652 
653 #ifdef IPDIVERT
654 	/*
655 	 * Divert reassembled packets to the divert protocol if required
656 	 *  If divert port is null then cookie should be too,
657 	 * so we shouldn't need to clear them here. Assume ip_divert does so.
658 	 */
659 	if (frag_divert_port) {
660 		ipstat.ips_delivered++;
661 		ip_divert_port = frag_divert_port;
662 		frag_divert_port = 0;
663 		(*inetsw[ip_protox[IPPROTO_DIVERT]].pr_input)(m, hlen);
664 		return;
665 	}
666 
667 	/* Don't let packets divert themselves */
668 	if (ip->ip_p == IPPROTO_DIVERT) {
669 		ipstat.ips_noproto++;
670 		goto bad;
671 	}
672 
673 #endif
674 
675 	/*
676 	 * Switch out to protocol's input routine.
677 	 */
678 	ipstat.ips_delivered++;
679 	(*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen);
680 #ifdef	IPFIREWALL_FORWARD
681 	ip_fw_fwd_addr = NULL;	/* tcp needed it */
682 #endif
683 	return;
684 bad:
685 #ifdef	IPFIREWALL_FORWARD
686 	ip_fw_fwd_addr = NULL;
687 #endif
688 	m_freem(m);
689 }
690 
691 /*
692  * IP software interrupt routine - to go away sometime soon
693  */
694 static void
695 ipintr(void)
696 {
697 	int s;
698 	struct mbuf *m;
699 
700 	while(1) {
701 		s = splimp();
702 		IF_DEQUEUE(&ipintrq, m);
703 		splx(s);
704 		if (m == 0)
705 			return;
706 		ip_input(m);
707 	}
708 }
709 
710 NETISR_SET(NETISR_IP, ipintr);
711 
712 /*
713  * Take incoming datagram fragment and try to
714  * reassemble it into whole datagram.  If a chain for
715  * reassembly of this datagram already exists, then it
716  * is given as fp; otherwise have to make a chain.
717  */
718 static struct ip *
719 ip_reass(m, fp, where)
720 	register struct mbuf *m;
721 	register struct ipq *fp;
722 	struct   ipq    *where;
723 {
724 	struct ip *ip = mtod(m, struct ip *);
725 	register struct mbuf *p = 0, *q, *nq;
726 	struct mbuf *t;
727 	int hlen = IP_VHL_HL(ip->ip_vhl) << 2;
728 	int i, next;
729 
730 	/*
731 	 * Presence of header sizes in mbufs
732 	 * would confuse code below.
733 	 */
734 	m->m_data += hlen;
735 	m->m_len -= hlen;
736 
737 	/*
738 	 * If first fragment to arrive, create a reassembly queue.
739 	 */
740 	if (fp == 0) {
741 		if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL)
742 			goto dropfrag;
743 		fp = mtod(t, struct ipq *);
744 		insque(fp, where);
745 		nipq++;
746 		fp->ipq_ttl = IPFRAGTTL;
747 		fp->ipq_p = ip->ip_p;
748 		fp->ipq_id = ip->ip_id;
749 		fp->ipq_src = ip->ip_src;
750 		fp->ipq_dst = ip->ip_dst;
751 		fp->ipq_frags = m;
752 		m->m_nextpkt = NULL;
753 #ifdef IPDIVERT
754 		fp->ipq_divert = 0;
755 		fp->ipq_div_cookie = 0;
756 #endif
757 		goto inserted;
758 	}
759 
760 #define GETIP(m)	((struct ip*)((m)->m_pkthdr.header))
761 
762 	/*
763 	 * Find a segment which begins after this one does.
764 	 */
765 	for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt)
766 		if (GETIP(q)->ip_off > ip->ip_off)
767 			break;
768 
769 	/*
770 	 * If there is a preceding segment, it may provide some of
771 	 * our data already.  If so, drop the data from the incoming
772 	 * segment.  If it provides all of our data, drop us, otherwise
773 	 * stick new segment in the proper place.
774 	 */
775 	if (p) {
776 		i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off;
777 		if (i > 0) {
778 			if (i >= ip->ip_len)
779 				goto dropfrag;
780 			m_adj(dtom(ip), i);
781 			ip->ip_off += i;
782 			ip->ip_len -= i;
783 		}
784 		m->m_nextpkt = p->m_nextpkt;
785 		p->m_nextpkt = m;
786 	} else {
787 		m->m_nextpkt = fp->ipq_frags;
788 		fp->ipq_frags = m;
789 	}
790 
791 	/*
792 	 * While we overlap succeeding segments trim them or,
793 	 * if they are completely covered, dequeue them.
794 	 */
795 	for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off;
796 	     q = nq) {
797 		i = (ip->ip_off + ip->ip_len) -
798 		    GETIP(q)->ip_off;
799 		if (i < GETIP(q)->ip_len) {
800 			GETIP(q)->ip_len -= i;
801 			GETIP(q)->ip_off += i;
802 			m_adj(q, i);
803 			break;
804 		}
805 		nq = q->m_nextpkt;
806 		m->m_nextpkt = nq;
807 		m_freem(q);
808 	}
809 
810 inserted:
811 
812 #ifdef IPDIVERT
813 	/*
814 	 * Any fragment diverting causes the whole packet to divert
815 	 */
816 	if (frag_divert_port) {
817 		fp->ipq_divert = frag_divert_port;
818 		fp->ipq_div_cookie = ip_divert_cookie;
819 	}
820 	frag_divert_port = 0;
821 	ip_divert_cookie = 0;
822 #endif
823 
824 	/*
825 	 * Check for complete reassembly.
826 	 */
827 	next = 0;
828 	for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
829 		if (GETIP(q)->ip_off != next)
830 			return (0);
831 		next += GETIP(q)->ip_len;
832 	}
833 	/* Make sure the last packet didn't have the IP_MF flag */
834 	if (p->m_flags & M_FRAG)
835 		return (0);
836 
837 	/*
838 	 * Reassembly is complete.  Make sure the packet is a sane size.
839 	 */
840 	q = fp->ipq_frags;
841 	ip = GETIP(q);
842 	if (next + (IP_VHL_HL(ip->ip_vhl) << 2) > IP_MAXPACKET) {
843 		ipstat.ips_toolong++;
844 		ip_freef(fp);
845 		return (0);
846 	}
847 
848 	/*
849 	 * Concatenate fragments.
850 	 */
851 	m = q;
852 	t = m->m_next;
853 	m->m_next = 0;
854 	m_cat(m, t);
855 	nq = q->m_nextpkt;
856 	q->m_nextpkt = 0;
857 	for (q = nq; q != NULL; q = nq) {
858 		nq = q->m_nextpkt;
859 		q->m_nextpkt = NULL;
860 		m_cat(m, q);
861 	}
862 
863 #ifdef IPDIVERT
864 	/*
865 	 * extract divert port for packet, if any
866 	 */
867 	frag_divert_port = fp->ipq_divert;
868 	ip_divert_cookie = fp->ipq_div_cookie;
869 #endif
870 
871 	/*
872 	 * Create header for new ip packet by
873 	 * modifying header of first packet;
874 	 * dequeue and discard fragment reassembly header.
875 	 * Make header visible.
876 	 */
877 	ip->ip_len = next;
878 	ip->ip_src = fp->ipq_src;
879 	ip->ip_dst = fp->ipq_dst;
880 	remque(fp);
881 	nipq--;
882 	(void) m_free(dtom(fp));
883 	m->m_len += (IP_VHL_HL(ip->ip_vhl) << 2);
884 	m->m_data -= (IP_VHL_HL(ip->ip_vhl) << 2);
885 	/* some debugging cruft by sklower, below, will go away soon */
886 	if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */
887 		register int plen = 0;
888 		for (t = m; m; m = m->m_next)
889 			plen += m->m_len;
890 		t->m_pkthdr.len = plen;
891 	}
892 	return (ip);
893 
894 dropfrag:
895 #ifdef IPDIVERT
896 	frag_divert_port = 0;
897 	ip_divert_cookie = 0;
898 #endif
899 	ipstat.ips_fragdropped++;
900 	m_freem(m);
901 	return (0);
902 
903 #undef GETIP
904 }
905 
906 /*
907  * Free a fragment reassembly header and all
908  * associated datagrams.
909  */
910 static void
911 ip_freef(fp)
912 	struct ipq *fp;
913 {
914 	register struct mbuf *q;
915 
916 	while (fp->ipq_frags) {
917 		q = fp->ipq_frags;
918 		fp->ipq_frags = q->m_nextpkt;
919 		m_freem(q);
920 	}
921 	remque(fp);
922 	(void) m_free(dtom(fp));
923 	nipq--;
924 }
925 
926 /*
927  * IP timer processing;
928  * if a timer expires on a reassembly
929  * queue, discard it.
930  */
931 void
932 ip_slowtimo()
933 {
934 	register struct ipq *fp;
935 	int s = splnet();
936 	int i;
937 
938 	for (i = 0; i < IPREASS_NHASH; i++) {
939 		fp = ipq[i].next;
940 		if (fp == 0)
941 			continue;
942 		while (fp != &ipq[i]) {
943 			--fp->ipq_ttl;
944 			fp = fp->next;
945 			if (fp->prev->ipq_ttl == 0) {
946 				ipstat.ips_fragtimeout++;
947 				ip_freef(fp->prev);
948 			}
949 		}
950 	}
951 	ipflow_slowtimo();
952 	splx(s);
953 }
954 
955 /*
956  * Drain off all datagram fragments.
957  */
958 void
959 ip_drain()
960 {
961 	int     i;
962 
963 	for (i = 0; i < IPREASS_NHASH; i++) {
964 		while (ipq[i].next != &ipq[i]) {
965 			ipstat.ips_fragdropped++;
966 			ip_freef(ipq[i].next);
967 		}
968 	}
969 	in_rtqdrain();
970 }
971 
972 /*
973  * Do option processing on a datagram,
974  * possibly discarding it if bad options are encountered,
975  * or forwarding it if source-routed.
976  * Returns 1 if packet has been forwarded/freed,
977  * 0 if the packet should be processed further.
978  */
979 static int
980 ip_dooptions(m)
981 	struct mbuf *m;
982 {
983 	register struct ip *ip = mtod(m, struct ip *);
984 	register u_char *cp;
985 	register struct ip_timestamp *ipt;
986 	register struct in_ifaddr *ia;
987 	int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0;
988 	struct in_addr *sin, dst;
989 	n_time ntime;
990 
991 	dst = ip->ip_dst;
992 	cp = (u_char *)(ip + 1);
993 	cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip);
994 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
995 		opt = cp[IPOPT_OPTVAL];
996 		if (opt == IPOPT_EOL)
997 			break;
998 		if (opt == IPOPT_NOP)
999 			optlen = 1;
1000 		else {
1001 			optlen = cp[IPOPT_OLEN];
1002 			if (optlen <= 0 || optlen > cnt) {
1003 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1004 				goto bad;
1005 			}
1006 		}
1007 		switch (opt) {
1008 
1009 		default:
1010 			break;
1011 
1012 		/*
1013 		 * Source routing with record.
1014 		 * Find interface with current destination address.
1015 		 * If none on this machine then drop if strictly routed,
1016 		 * or do nothing if loosely routed.
1017 		 * Record interface address and bring up next address
1018 		 * component.  If strictly routed make sure next
1019 		 * address is on directly accessible net.
1020 		 */
1021 		case IPOPT_LSRR:
1022 		case IPOPT_SSRR:
1023 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1024 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1025 				goto bad;
1026 			}
1027 			ipaddr.sin_addr = ip->ip_dst;
1028 			ia = (struct in_ifaddr *)
1029 				ifa_ifwithaddr((struct sockaddr *)&ipaddr);
1030 			if (ia == 0) {
1031 				if (opt == IPOPT_SSRR) {
1032 					type = ICMP_UNREACH;
1033 					code = ICMP_UNREACH_SRCFAIL;
1034 					goto bad;
1035 				}
1036 				if (!ip_dosourceroute)
1037 					goto nosourcerouting;
1038 				/*
1039 				 * Loose routing, and not at next destination
1040 				 * yet; nothing to do except forward.
1041 				 */
1042 				break;
1043 			}
1044 			off--;			/* 0 origin */
1045 			if (off > optlen - sizeof(struct in_addr)) {
1046 				/*
1047 				 * End of source route.  Should be for us.
1048 				 */
1049 				if (!ip_acceptsourceroute)
1050 					goto nosourcerouting;
1051 				save_rte(cp, ip->ip_src);
1052 				break;
1053 			}
1054 
1055 			if (!ip_dosourceroute) {
1056 				if (ipforwarding) {
1057 					char buf[16]; /* aaa.bbb.ccc.ddd\0 */
1058 					/*
1059 					 * Acting as a router, so generate ICMP
1060 					 */
1061 nosourcerouting:
1062 					strcpy(buf, inet_ntoa(ip->ip_dst));
1063 					log(LOG_WARNING,
1064 					    "attempted source route from %s to %s\n",
1065 					    inet_ntoa(ip->ip_src), buf);
1066 					type = ICMP_UNREACH;
1067 					code = ICMP_UNREACH_SRCFAIL;
1068 					goto bad;
1069 				} else {
1070 					/*
1071 					 * Not acting as a router, so silently drop.
1072 					 */
1073 					ipstat.ips_cantforward++;
1074 					m_freem(m);
1075 					return (1);
1076 				}
1077 			}
1078 
1079 			/*
1080 			 * locate outgoing interface
1081 			 */
1082 			(void)memcpy(&ipaddr.sin_addr, cp + off,
1083 			    sizeof(ipaddr.sin_addr));
1084 
1085 			if (opt == IPOPT_SSRR) {
1086 #define	INA	struct in_ifaddr *
1087 #define	SA	struct sockaddr *
1088 			    if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == 0)
1089 				ia = (INA)ifa_ifwithnet((SA)&ipaddr);
1090 			} else
1091 				ia = ip_rtaddr(ipaddr.sin_addr);
1092 			if (ia == 0) {
1093 				type = ICMP_UNREACH;
1094 				code = ICMP_UNREACH_SRCFAIL;
1095 				goto bad;
1096 			}
1097 			ip->ip_dst = ipaddr.sin_addr;
1098 			(void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
1099 			    sizeof(struct in_addr));
1100 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1101 			/*
1102 			 * Let ip_intr's mcast routing check handle mcast pkts
1103 			 */
1104 			forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr));
1105 			break;
1106 
1107 		case IPOPT_RR:
1108 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1109 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1110 				goto bad;
1111 			}
1112 			/*
1113 			 * If no space remains, ignore.
1114 			 */
1115 			off--;			/* 0 origin */
1116 			if (off > optlen - sizeof(struct in_addr))
1117 				break;
1118 			(void)memcpy(&ipaddr.sin_addr, &ip->ip_dst,
1119 			    sizeof(ipaddr.sin_addr));
1120 			/*
1121 			 * locate outgoing interface; if we're the destination,
1122 			 * use the incoming interface (should be same).
1123 			 */
1124 			if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == 0 &&
1125 			    (ia = ip_rtaddr(ipaddr.sin_addr)) == 0) {
1126 				type = ICMP_UNREACH;
1127 				code = ICMP_UNREACH_HOST;
1128 				goto bad;
1129 			}
1130 			(void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
1131 			    sizeof(struct in_addr));
1132 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1133 			break;
1134 
1135 		case IPOPT_TS:
1136 			code = cp - (u_char *)ip;
1137 			ipt = (struct ip_timestamp *)cp;
1138 			if (ipt->ipt_len < 5)
1139 				goto bad;
1140 			if (ipt->ipt_ptr > ipt->ipt_len - sizeof(int32_t)) {
1141 				if (++ipt->ipt_oflw == 0)
1142 					goto bad;
1143 				break;
1144 			}
1145 			sin = (struct in_addr *)(cp + ipt->ipt_ptr - 1);
1146 			switch (ipt->ipt_flg) {
1147 
1148 			case IPOPT_TS_TSONLY:
1149 				break;
1150 
1151 			case IPOPT_TS_TSANDADDR:
1152 				if (ipt->ipt_ptr - 1 + sizeof(n_time) +
1153 				    sizeof(struct in_addr) > ipt->ipt_len)
1154 					goto bad;
1155 				ipaddr.sin_addr = dst;
1156 				ia = (INA)ifaof_ifpforaddr((SA)&ipaddr,
1157 							    m->m_pkthdr.rcvif);
1158 				if (ia == 0)
1159 					continue;
1160 				(void)memcpy(sin, &IA_SIN(ia)->sin_addr,
1161 				    sizeof(struct in_addr));
1162 				ipt->ipt_ptr += sizeof(struct in_addr);
1163 				break;
1164 
1165 			case IPOPT_TS_PRESPEC:
1166 				if (ipt->ipt_ptr - 1 + sizeof(n_time) +
1167 				    sizeof(struct in_addr) > ipt->ipt_len)
1168 					goto bad;
1169 				(void)memcpy(&ipaddr.sin_addr, sin,
1170 				    sizeof(struct in_addr));
1171 				if (ifa_ifwithaddr((SA)&ipaddr) == 0)
1172 					continue;
1173 				ipt->ipt_ptr += sizeof(struct in_addr);
1174 				break;
1175 
1176 			default:
1177 				goto bad;
1178 			}
1179 			ntime = iptime();
1180 			(void)memcpy(cp + ipt->ipt_ptr - 1, &ntime,
1181 			    sizeof(n_time));
1182 			ipt->ipt_ptr += sizeof(n_time);
1183 		}
1184 	}
1185 	if (forward && ipforwarding) {
1186 		ip_forward(m, 1);
1187 		return (1);
1188 	}
1189 	return (0);
1190 bad:
1191 	ip->ip_len -= IP_VHL_HL(ip->ip_vhl) << 2;   /* XXX icmp_error adds in hdr length */
1192 	icmp_error(m, type, code, 0, 0);
1193 	ipstat.ips_badoptions++;
1194 	return (1);
1195 }
1196 
1197 /*
1198  * Given address of next destination (final or next hop),
1199  * return internet address info of interface to be used to get there.
1200  */
1201 static struct in_ifaddr *
1202 ip_rtaddr(dst)
1203 	 struct in_addr dst;
1204 {
1205 	register struct sockaddr_in *sin;
1206 
1207 	sin = (struct sockaddr_in *) &ipforward_rt.ro_dst;
1208 
1209 	if (ipforward_rt.ro_rt == 0 || dst.s_addr != sin->sin_addr.s_addr) {
1210 		if (ipforward_rt.ro_rt) {
1211 			RTFREE(ipforward_rt.ro_rt);
1212 			ipforward_rt.ro_rt = 0;
1213 		}
1214 		sin->sin_family = AF_INET;
1215 		sin->sin_len = sizeof(*sin);
1216 		sin->sin_addr = dst;
1217 
1218 		rtalloc_ign(&ipforward_rt, RTF_PRCLONING);
1219 	}
1220 	if (ipforward_rt.ro_rt == 0)
1221 		return ((struct in_ifaddr *)0);
1222 	return ((struct in_ifaddr *) ipforward_rt.ro_rt->rt_ifa);
1223 }
1224 
1225 /*
1226  * Save incoming source route for use in replies,
1227  * to be picked up later by ip_srcroute if the receiver is interested.
1228  */
1229 void
1230 save_rte(option, dst)
1231 	u_char *option;
1232 	struct in_addr dst;
1233 {
1234 	unsigned olen;
1235 
1236 	olen = option[IPOPT_OLEN];
1237 #ifdef DIAGNOSTIC
1238 	if (ipprintfs)
1239 		printf("save_rte: olen %d\n", olen);
1240 #endif
1241 	if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst)))
1242 		return;
1243 	bcopy(option, ip_srcrt.srcopt, olen);
1244 	ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1245 	ip_srcrt.dst = dst;
1246 }
1247 
1248 /*
1249  * Retrieve incoming source route for use in replies,
1250  * in the same form used by setsockopt.
1251  * The first hop is placed before the options, will be removed later.
1252  */
1253 struct mbuf *
1254 ip_srcroute()
1255 {
1256 	register struct in_addr *p, *q;
1257 	register struct mbuf *m;
1258 
1259 	if (ip_nhops == 0)
1260 		return ((struct mbuf *)0);
1261 	m = m_get(M_DONTWAIT, MT_HEADER);
1262 	if (m == 0)
1263 		return ((struct mbuf *)0);
1264 
1265 #define OPTSIZ	(sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt))
1266 
1267 	/* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
1268 	m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) +
1269 	    OPTSIZ;
1270 #ifdef DIAGNOSTIC
1271 	if (ipprintfs)
1272 		printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len);
1273 #endif
1274 
1275 	/*
1276 	 * First save first hop for return route
1277 	 */
1278 	p = &ip_srcrt.route[ip_nhops - 1];
1279 	*(mtod(m, struct in_addr *)) = *p--;
1280 #ifdef DIAGNOSTIC
1281 	if (ipprintfs)
1282 		printf(" hops %lx", (u_long)ntohl(mtod(m, struct in_addr *)->s_addr));
1283 #endif
1284 
1285 	/*
1286 	 * Copy option fields and padding (nop) to mbuf.
1287 	 */
1288 	ip_srcrt.nop = IPOPT_NOP;
1289 	ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
1290 	(void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr),
1291 	    &ip_srcrt.nop, OPTSIZ);
1292 	q = (struct in_addr *)(mtod(m, caddr_t) +
1293 	    sizeof(struct in_addr) + OPTSIZ);
1294 #undef OPTSIZ
1295 	/*
1296 	 * Record return path as an IP source route,
1297 	 * reversing the path (pointers are now aligned).
1298 	 */
1299 	while (p >= ip_srcrt.route) {
1300 #ifdef DIAGNOSTIC
1301 		if (ipprintfs)
1302 			printf(" %lx", (u_long)ntohl(q->s_addr));
1303 #endif
1304 		*q++ = *p--;
1305 	}
1306 	/*
1307 	 * Last hop goes to final destination.
1308 	 */
1309 	*q = ip_srcrt.dst;
1310 #ifdef DIAGNOSTIC
1311 	if (ipprintfs)
1312 		printf(" %lx\n", (u_long)ntohl(q->s_addr));
1313 #endif
1314 	return (m);
1315 }
1316 
1317 /*
1318  * Strip out IP options, at higher
1319  * level protocol in the kernel.
1320  * Second argument is buffer to which options
1321  * will be moved, and return value is their length.
1322  * XXX should be deleted; last arg currently ignored.
1323  */
1324 void
1325 ip_stripoptions(m, mopt)
1326 	register struct mbuf *m;
1327 	struct mbuf *mopt;
1328 {
1329 	register int i;
1330 	struct ip *ip = mtod(m, struct ip *);
1331 	register caddr_t opts;
1332 	int olen;
1333 
1334 	olen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip);
1335 	opts = (caddr_t)(ip + 1);
1336 	i = m->m_len - (sizeof (struct ip) + olen);
1337 	bcopy(opts + olen, opts, (unsigned)i);
1338 	m->m_len -= olen;
1339 	if (m->m_flags & M_PKTHDR)
1340 		m->m_pkthdr.len -= olen;
1341 	ip->ip_vhl = IP_MAKE_VHL(IPVERSION, sizeof(struct ip) >> 2);
1342 }
1343 
1344 u_char inetctlerrmap[PRC_NCMDS] = {
1345 	0,		0,		0,		0,
1346 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1347 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1348 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1349 	0,		0,		0,		0,
1350 	ENOPROTOOPT
1351 };
1352 
1353 /*
1354  * Forward a packet.  If some error occurs return the sender
1355  * an icmp packet.  Note we can't always generate a meaningful
1356  * icmp message because icmp doesn't have a large enough repertoire
1357  * of codes and types.
1358  *
1359  * If not forwarding, just drop the packet.  This could be confusing
1360  * if ipforwarding was zero but some routing protocol was advancing
1361  * us as a gateway to somewhere.  However, we must let the routing
1362  * protocol deal with that.
1363  *
1364  * The srcrt parameter indicates whether the packet is being forwarded
1365  * via a source route.
1366  */
1367 static void
1368 ip_forward(m, srcrt)
1369 	struct mbuf *m;
1370 	int srcrt;
1371 {
1372 	register struct ip *ip = mtod(m, struct ip *);
1373 	register struct sockaddr_in *sin;
1374 	register struct rtentry *rt;
1375 	int error, type = 0, code = 0;
1376 	struct mbuf *mcopy;
1377 	n_long dest;
1378 	struct ifnet *destifp;
1379 
1380 	dest = 0;
1381 #ifdef DIAGNOSTIC
1382 	if (ipprintfs)
1383 		printf("forward: src %lx dst %lx ttl %x\n",
1384 		    (u_long)ip->ip_src.s_addr, (u_long)ip->ip_dst.s_addr,
1385 		    ip->ip_ttl);
1386 #endif
1387 
1388 
1389 	if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
1390 		ipstat.ips_cantforward++;
1391 		m_freem(m);
1392 		return;
1393 	}
1394 	HTONS(ip->ip_id);
1395 #ifdef IPSTEALTH
1396 	if (!ipstealth) {
1397 #endif
1398 		if (ip->ip_ttl <= IPTTLDEC) {
1399 			icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS,
1400 			    dest, 0);
1401 			return;
1402 		}
1403 		ip->ip_ttl -= IPTTLDEC;
1404 #ifdef IPSTEALTH
1405 	}
1406 #endif
1407 
1408 	sin = (struct sockaddr_in *)&ipforward_rt.ro_dst;
1409 	if ((rt = ipforward_rt.ro_rt) == 0 ||
1410 	    ip->ip_dst.s_addr != sin->sin_addr.s_addr) {
1411 		if (ipforward_rt.ro_rt) {
1412 			RTFREE(ipforward_rt.ro_rt);
1413 			ipforward_rt.ro_rt = 0;
1414 		}
1415 		sin->sin_family = AF_INET;
1416 		sin->sin_len = sizeof(*sin);
1417 		sin->sin_addr = ip->ip_dst;
1418 
1419 		rtalloc_ign(&ipforward_rt, RTF_PRCLONING);
1420 		if (ipforward_rt.ro_rt == 0) {
1421 			icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0);
1422 			return;
1423 		}
1424 		rt = ipforward_rt.ro_rt;
1425 	}
1426 
1427 	/*
1428 	 * Save at most 64 bytes of the packet in case
1429 	 * we need to generate an ICMP message to the src.
1430 	 */
1431 	mcopy = m_copy(m, 0, imin((int)ip->ip_len, 64));
1432 
1433 	/*
1434 	 * If forwarding packet using same interface that it came in on,
1435 	 * perhaps should send a redirect to sender to shortcut a hop.
1436 	 * Only send redirect if source is sending directly to us,
1437 	 * and if packet was not source routed (or has any options).
1438 	 * Also, don't send redirect if forwarding using a default route
1439 	 * or a route modified by a redirect.
1440 	 */
1441 #define	satosin(sa)	((struct sockaddr_in *)(sa))
1442 	if (rt->rt_ifp == m->m_pkthdr.rcvif &&
1443 	    (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1444 	    satosin(rt_key(rt))->sin_addr.s_addr != 0 &&
1445 	    ipsendredirects && !srcrt) {
1446 #define	RTA(rt)	((struct in_ifaddr *)(rt->rt_ifa))
1447 		u_long src = ntohl(ip->ip_src.s_addr);
1448 
1449 		if (RTA(rt) &&
1450 		    (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1451 		    if (rt->rt_flags & RTF_GATEWAY)
1452 			dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
1453 		    else
1454 			dest = ip->ip_dst.s_addr;
1455 		    /* Router requirements says to only send host redirects */
1456 		    type = ICMP_REDIRECT;
1457 		    code = ICMP_REDIRECT_HOST;
1458 #ifdef DIAGNOSTIC
1459 		    if (ipprintfs)
1460 		        printf("redirect (%d) to %lx\n", code, (u_long)dest);
1461 #endif
1462 		}
1463 	}
1464 
1465 	error = ip_output(m, (struct mbuf *)0, &ipforward_rt,
1466 			  IP_FORWARDING, 0);
1467 	if (error)
1468 		ipstat.ips_cantforward++;
1469 	else {
1470 		ipstat.ips_forward++;
1471 		if (type)
1472 			ipstat.ips_redirectsent++;
1473 		else {
1474 			if (mcopy) {
1475 				ipflow_create(&ipforward_rt, mcopy);
1476 				m_freem(mcopy);
1477 			}
1478 			return;
1479 		}
1480 	}
1481 	if (mcopy == NULL)
1482 		return;
1483 	destifp = NULL;
1484 
1485 	switch (error) {
1486 
1487 	case 0:				/* forwarded, but need redirect */
1488 		/* type, code set above */
1489 		break;
1490 
1491 	case ENETUNREACH:		/* shouldn't happen, checked above */
1492 	case EHOSTUNREACH:
1493 	case ENETDOWN:
1494 	case EHOSTDOWN:
1495 	default:
1496 		type = ICMP_UNREACH;
1497 		code = ICMP_UNREACH_HOST;
1498 		break;
1499 
1500 	case EMSGSIZE:
1501 		type = ICMP_UNREACH;
1502 		code = ICMP_UNREACH_NEEDFRAG;
1503 		if (ipforward_rt.ro_rt)
1504 			destifp = ipforward_rt.ro_rt->rt_ifp;
1505 		ipstat.ips_cantfrag++;
1506 		break;
1507 
1508 	case ENOBUFS:
1509 		type = ICMP_SOURCEQUENCH;
1510 		code = 0;
1511 		break;
1512 	}
1513 	icmp_error(mcopy, type, code, dest, destifp);
1514 }
1515 
1516 void
1517 ip_savecontrol(inp, mp, ip, m)
1518 	register struct inpcb *inp;
1519 	register struct mbuf **mp;
1520 	register struct ip *ip;
1521 	register struct mbuf *m;
1522 {
1523 	if (inp->inp_socket->so_options & SO_TIMESTAMP) {
1524 		struct timeval tv;
1525 
1526 		microtime(&tv);
1527 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1528 			SCM_TIMESTAMP, SOL_SOCKET);
1529 		if (*mp)
1530 			mp = &(*mp)->m_next;
1531 	}
1532 	if (inp->inp_flags & INP_RECVDSTADDR) {
1533 		*mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
1534 		    sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1535 		if (*mp)
1536 			mp = &(*mp)->m_next;
1537 	}
1538 #ifdef notyet
1539 	/* XXX
1540 	 * Moving these out of udp_input() made them even more broken
1541 	 * than they already were.
1542 	 */
1543 	/* options were tossed already */
1544 	if (inp->inp_flags & INP_RECVOPTS) {
1545 		*mp = sbcreatecontrol((caddr_t) opts_deleted_above,
1546 		    sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1547 		if (*mp)
1548 			mp = &(*mp)->m_next;
1549 	}
1550 	/* ip_srcroute doesn't do what we want here, need to fix */
1551 	if (inp->inp_flags & INP_RECVRETOPTS) {
1552 		*mp = sbcreatecontrol((caddr_t) ip_srcroute(),
1553 		    sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1554 		if (*mp)
1555 			mp = &(*mp)->m_next;
1556 	}
1557 #endif
1558 	if (inp->inp_flags & INP_RECVIF) {
1559 		struct ifnet *ifp;
1560 		struct sdlbuf {
1561 			struct sockaddr_dl sdl;
1562 			u_char	pad[32];
1563 		} sdlbuf;
1564 		struct sockaddr_dl *sdp;
1565 		struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1566 
1567 		if (((ifp = m->m_pkthdr.rcvif))
1568 		&& ( ifp->if_index && (ifp->if_index <= if_index))) {
1569 			sdp = (struct sockaddr_dl *)(ifnet_addrs
1570 					[ifp->if_index - 1]->ifa_addr);
1571 			/*
1572 			 * Change our mind and don't try copy.
1573 			 */
1574 			if ((sdp->sdl_family != AF_LINK)
1575 			|| (sdp->sdl_len > sizeof(sdlbuf))) {
1576 				goto makedummy;
1577 			}
1578 			bcopy(sdp, sdl2, sdp->sdl_len);
1579 		} else {
1580 makedummy:
1581 			sdl2->sdl_len
1582 				= offsetof(struct sockaddr_dl, sdl_data[0]);
1583 			sdl2->sdl_family = AF_LINK;
1584 			sdl2->sdl_index = 0;
1585 			sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1586 		}
1587 		*mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len,
1588 			IP_RECVIF, IPPROTO_IP);
1589 		if (*mp)
1590 			mp = &(*mp)->m_next;
1591 	}
1592 }
1593 
1594 int
1595 ip_rsvp_init(struct socket *so)
1596 {
1597 	if (so->so_type != SOCK_RAW ||
1598 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
1599 	  return EOPNOTSUPP;
1600 
1601 	if (ip_rsvpd != NULL)
1602 	  return EADDRINUSE;
1603 
1604 	ip_rsvpd = so;
1605 	/*
1606 	 * This may seem silly, but we need to be sure we don't over-increment
1607 	 * the RSVP counter, in case something slips up.
1608 	 */
1609 	if (!ip_rsvp_on) {
1610 		ip_rsvp_on = 1;
1611 		rsvp_on++;
1612 	}
1613 
1614 	return 0;
1615 }
1616 
1617 int
1618 ip_rsvp_done(void)
1619 {
1620 	ip_rsvpd = NULL;
1621 	/*
1622 	 * This may seem silly, but we need to be sure we don't over-decrement
1623 	 * the RSVP counter, in case something slips up.
1624 	 */
1625 	if (ip_rsvp_on) {
1626 		ip_rsvp_on = 0;
1627 		rsvp_on--;
1628 	}
1629 	return 0;
1630 }
1631