1 /*	$OpenBSD: ip_input.c,v 1.122.2.1 2005/06/14 01:49:24 brad Exp $	*/
2 /*	$NetBSD: ip_input.c,v 1.30 1996/03/16 23:53:58 christos Exp $	*/
3 
4 /*
5  * Copyright (c) 1982, 1986, 1988, 1993
6  *	The Regents of the University of California.  All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the University nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  *
32  *	@(#)ip_input.c	8.2 (Berkeley) 1/4/94
33  */
34 
35 #include "pf.h"
36 
37 #include <sys/param.h>
38 #include <sys/systm.h>
39 #include <sys/mbuf.h>
40 #include <sys/domain.h>
41 #include <sys/protosw.h>
42 #include <sys/socket.h>
43 #include <sys/syslog.h>
44 #include <sys/sysctl.h>
45 
46 #include <net/if.h>
47 #include <net/if_dl.h>
48 #include <net/route.h>
49 
50 #include <netinet/in.h>
51 #include <netinet/in_systm.h>
52 #include <netinet/if_ether.h>
53 #include <netinet/ip.h>
54 #include <netinet/in_pcb.h>
55 #include <netinet/in_var.h>
56 #include <netinet/ip_var.h>
57 #include <netinet/ip_icmp.h>
58 
59 #if NPF > 0
60 #include <net/pfvar.h>
61 #endif
62 
63 #ifdef IPSEC
64 #include <netinet/ip_ipsp.h>
65 #endif /* IPSEC */
66 
67 #ifndef	IPFORWARDING
68 #ifdef GATEWAY
69 #define	IPFORWARDING	1	/* forward IP packets not for us */
70 #else /* GATEWAY */
71 #define	IPFORWARDING	0	/* don't forward IP packets not for us */
72 #endif /* GATEWAY */
73 #endif /* IPFORWARDING */
74 #ifndef	IPSENDREDIRECTS
75 #define	IPSENDREDIRECTS	1
76 #endif
77 
78 #ifndef IPMTUDISC
79 #define IPMTUDISC	1
80 #endif
81 #ifndef IPMTUDISCTIMEOUT
82 #define IPMTUDISCTIMEOUT (10 * 60)	/* as per RFC 1191 */
83 #endif
84 
85 struct ipqhead ipq;
86 
87 int encdebug = 0;
88 int ipsec_keep_invalid = IPSEC_DEFAULT_EMBRYONIC_SA_TIMEOUT;
89 int ipsec_require_pfs = IPSEC_DEFAULT_PFS;
90 int ipsec_soft_allocations = IPSEC_DEFAULT_SOFT_ALLOCATIONS;
91 int ipsec_exp_allocations = IPSEC_DEFAULT_EXP_ALLOCATIONS;
92 int ipsec_soft_bytes = IPSEC_DEFAULT_SOFT_BYTES;
93 int ipsec_exp_bytes = IPSEC_DEFAULT_EXP_BYTES;
94 int ipsec_soft_timeout = IPSEC_DEFAULT_SOFT_TIMEOUT;
95 int ipsec_exp_timeout = IPSEC_DEFAULT_EXP_TIMEOUT;
96 int ipsec_soft_first_use = IPSEC_DEFAULT_SOFT_FIRST_USE;
97 int ipsec_exp_first_use = IPSEC_DEFAULT_EXP_FIRST_USE;
98 int ipsec_expire_acquire = IPSEC_DEFAULT_EXPIRE_ACQUIRE;
99 char ipsec_def_enc[20];
100 char ipsec_def_auth[20];
101 char ipsec_def_comp[20];
102 
103 /*
104  * Note: DIRECTED_BROADCAST is handled this way so that previous
105  * configuration using this option will Just Work.
106  */
107 #ifndef IPDIRECTEDBCAST
108 #ifdef DIRECTED_BROADCAST
109 #define IPDIRECTEDBCAST	1
110 #else
111 #define	IPDIRECTEDBCAST	0
112 #endif /* DIRECTED_BROADCAST */
113 #endif /* IPDIRECTEDBCAST */
114 int	ipforwarding = IPFORWARDING;
115 int	ipsendredirects = IPSENDREDIRECTS;
116 int	ip_dosourceroute = 0;	/* no src-routing unless sysctl'd to enable */
117 int	ip_defttl = IPDEFTTL;
118 int	ip_mtudisc = IPMTUDISC;
119 u_int	ip_mtudisc_timeout = IPMTUDISCTIMEOUT;
120 int	ip_directedbcast = IPDIRECTEDBCAST;
121 #ifdef DIAGNOSTIC
122 int	ipprintfs = 0;
123 #endif
124 
125 struct rttimer_queue *ip_mtudisc_timeout_q = NULL;
126 
127 int	ipsec_auth_default_level = IPSEC_AUTH_LEVEL_DEFAULT;
128 int	ipsec_esp_trans_default_level = IPSEC_ESP_TRANS_LEVEL_DEFAULT;
129 int	ipsec_esp_network_default_level = IPSEC_ESP_NETWORK_LEVEL_DEFAULT;
130 int	ipsec_ipcomp_default_level = IPSEC_IPCOMP_LEVEL_DEFAULT;
131 
132 /* Keep track of memory used for reassembly */
133 int	ip_maxqueue = 300;
134 int	ip_frags = 0;
135 
136 /* from in_pcb.c */
137 extern int ipport_firstauto;
138 extern int ipport_lastauto;
139 extern int ipport_hifirstauto;
140 extern int ipport_hilastauto;
141 extern struct baddynamicports baddynamicports;
142 
143 int *ipctl_vars[IPCTL_MAXID] = IPCTL_VARS;
144 
145 extern	struct domain inetdomain;
146 extern	struct protosw inetsw[];
147 u_char	ip_protox[IPPROTO_MAX];
148 int	ipqmaxlen = IFQ_MAXLEN;
149 struct	in_ifaddrhead in_ifaddr;
150 struct	ifqueue ipintrq;
151 
152 int	ipq_locked;
153 static __inline int ipq_lock_try(void);
154 static __inline void ipq_unlock(void);
155 
156 struct pool ipqent_pool;
157 
158 struct ipstat ipstat;
159 
160 static __inline int
ipq_lock_try()161 ipq_lock_try()
162 {
163 	int s;
164 
165 	s = splimp();
166 	if (ipq_locked) {
167 		splx(s);
168 		return (0);
169 	}
170 	ipq_locked = 1;
171 	splx(s);
172 	return (1);
173 }
174 
175 #define ipq_lock() ipq_lock_try()
176 
177 static __inline void
ipq_unlock()178 ipq_unlock()
179 {
180 	int s;
181 
182 	s = splimp();
183 	ipq_locked = 0;
184 	splx(s);
185 }
186 
187 char *
inet_ntoa(ina)188 inet_ntoa(ina)
189 	struct in_addr ina;
190 {
191 	static char buf[4*sizeof "123"];
192 	unsigned char *ucp = (unsigned char *)&ina;
193 
194 	snprintf(buf, sizeof buf, "%d.%d.%d.%d",
195 	    ucp[0] & 0xff, ucp[1] & 0xff,
196 	    ucp[2] & 0xff, ucp[3] & 0xff);
197 	return (buf);
198 }
199 
200 /*
201  * We need to save the IP options in case a protocol wants to respond
202  * to an incoming packet over the same route if the packet got here
203  * using IP source routing.  This allows connection establishment and
204  * maintenance when the remote end is on a network that is not known
205  * to us.
206  */
207 int	ip_nhops = 0;
208 static	struct ip_srcrt {
209 	struct	in_addr dst;			/* final destination */
210 	char	nop;				/* one NOP to align */
211 	char	srcopt[IPOPT_OFFSET + 1];	/* OPTVAL, OLEN and OFFSET */
212 	struct	in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)];
213 } ip_srcrt;
214 
215 static void save_rte(u_char *, struct in_addr);
216 static int ip_weadvertise(u_int32_t);
217 
218 /*
219  * IP initialization: fill in IP protocol switch table.
220  * All protocols not implemented in kernel go to raw IP protocol handler.
221  */
222 void
ip_init()223 ip_init()
224 {
225 	struct protosw *pr;
226 	int i;
227 	const u_int16_t defbaddynamicports_tcp[] = DEFBADDYNAMICPORTS_TCP;
228 	const u_int16_t defbaddynamicports_udp[] = DEFBADDYNAMICPORTS_UDP;
229 
230 	pool_init(&ipqent_pool, sizeof(struct ipqent), 0, 0, 0, "ipqepl",
231 	    NULL);
232 
233 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
234 	if (pr == 0)
235 		panic("ip_init");
236 	for (i = 0; i < IPPROTO_MAX; i++)
237 		ip_protox[i] = pr - inetsw;
238 	for (pr = inetdomain.dom_protosw;
239 	    pr < inetdomain.dom_protoswNPROTOSW; pr++)
240 		if (pr->pr_domain->dom_family == PF_INET &&
241 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
242 			ip_protox[pr->pr_protocol] = pr - inetsw;
243 	LIST_INIT(&ipq);
244 	ipintrq.ifq_maxlen = ipqmaxlen;
245 	TAILQ_INIT(&in_ifaddr);
246 	if (ip_mtudisc != 0)
247 		ip_mtudisc_timeout_q =
248 		    rt_timer_queue_create(ip_mtudisc_timeout);
249 
250 	/* Fill in list of ports not to allocate dynamically. */
251 	bzero((void *)&baddynamicports, sizeof(baddynamicports));
252 	for (i = 0; defbaddynamicports_tcp[i] != 0; i++)
253 		DP_SET(baddynamicports.tcp, defbaddynamicports_tcp[i]);
254 	for (i = 0; defbaddynamicports_udp[i] != 0; i++)
255 		DP_SET(baddynamicports.udp, defbaddynamicports_udp[i]);
256 
257 	strlcpy(ipsec_def_enc, IPSEC_DEFAULT_DEF_ENC, sizeof(ipsec_def_enc));
258 	strlcpy(ipsec_def_auth, IPSEC_DEFAULT_DEF_AUTH, sizeof(ipsec_def_auth));
259 	strlcpy(ipsec_def_comp, IPSEC_DEFAULT_DEF_COMP, sizeof(ipsec_def_comp));
260 }
261 
262 struct	sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET };
263 struct	route ipforward_rt;
264 
265 void
ipintr()266 ipintr()
267 {
268 	struct mbuf *m;
269 	int s;
270 
271 	while (1) {
272 		/*
273 		 * Get next datagram off input queue and get IP header
274 		 * in first mbuf.
275 		 */
276 		s = splimp();
277 		IF_DEQUEUE(&ipintrq, m);
278 		splx(s);
279 		if (m == 0)
280 			return;
281 #ifdef	DIAGNOSTIC
282 		if ((m->m_flags & M_PKTHDR) == 0)
283 			panic("ipintr no HDR");
284 #endif
285 		ipv4_input(m);
286 	}
287 }
288 
289 /*
290  * Ip input routine.  Checksum and byte swap header.  If fragmented
291  * try to reassemble.  Process options.  Pass to next level.
292  */
293 void
ipv4_input(m)294 ipv4_input(m)
295 	struct mbuf *m;
296 {
297 	struct ip *ip;
298 	struct ipq *fp;
299 	struct in_ifaddr *ia;
300 	struct ipqent *ipqe;
301 	int hlen, mff, len;
302 	in_addr_t pfrdr = 0;
303 #ifdef IPSEC
304 	int error, s;
305 	struct tdb *tdb;
306 	struct tdb_ident *tdbi;
307 	struct m_tag *mtag;
308 #endif /* IPSEC */
309 
310 	/*
311 	 * If no IP addresses have been set yet but the interfaces
312 	 * are receiving, can't do anything with incoming packets yet.
313 	 */
314 	if (in_ifaddr.tqh_first == 0)
315 		goto bad;
316 	ipstat.ips_total++;
317 	if (m->m_len < sizeof (struct ip) &&
318 	    (m = m_pullup(m, sizeof (struct ip))) == NULL) {
319 		ipstat.ips_toosmall++;
320 		return;
321 	}
322 	ip = mtod(m, struct ip *);
323 	if (ip->ip_v != IPVERSION) {
324 		ipstat.ips_badvers++;
325 		goto bad;
326 	}
327 	hlen = ip->ip_hl << 2;
328 	if (hlen < sizeof(struct ip)) {	/* minimum header length */
329 		ipstat.ips_badhlen++;
330 		goto bad;
331 	}
332 	if (hlen > m->m_len) {
333 		if ((m = m_pullup(m, hlen)) == NULL) {
334 			ipstat.ips_badhlen++;
335 			return;
336 		}
337 		ip = mtod(m, struct ip *);
338 	}
339 
340 	/* 127/8 must not appear on wire - RFC1122 */
341 	if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
342 	    (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
343 		if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
344 			ipstat.ips_badaddr++;
345 			goto bad;
346 		}
347 	}
348 
349 	if ((m->m_pkthdr.csum & M_IPV4_CSUM_IN_OK) == 0) {
350 		if (m->m_pkthdr.csum & M_IPV4_CSUM_IN_BAD) {
351 			ipstat.ips_inhwcsum++;
352 			ipstat.ips_badsum++;
353 			goto bad;
354 		}
355 
356 		if (in_cksum(m, hlen) != 0) {
357 			ipstat.ips_badsum++;
358 			goto bad;
359 		}
360 	} else {
361 		m->m_pkthdr.csum &= ~M_IPV4_CSUM_IN_OK;
362 		ipstat.ips_inhwcsum++;
363 	}
364 
365 	/* Retrieve the packet length. */
366 	len = ntohs(ip->ip_len);
367 
368 	/*
369 	 * Convert fields to host representation.
370 	 */
371 	if (len < hlen) {
372 		ipstat.ips_badlen++;
373 		goto bad;
374 	}
375 
376 	/*
377 	 * Check that the amount of data in the buffers
378 	 * is at least as much as the IP header would have us expect.
379 	 * Trim mbufs if longer than we expect.
380 	 * Drop packet if shorter than we expect.
381 	 */
382 	if (m->m_pkthdr.len < len) {
383 		ipstat.ips_tooshort++;
384 		goto bad;
385 	}
386 	if (m->m_pkthdr.len > len) {
387 		if (m->m_len == m->m_pkthdr.len) {
388 			m->m_len = len;
389 			m->m_pkthdr.len = len;
390 		} else
391 			m_adj(m, len - m->m_pkthdr.len);
392 	}
393 
394 #if NPF > 0
395 	/*
396 	 * Packet filter
397 	 */
398 	pfrdr = ip->ip_dst.s_addr;
399 	if (pf_test(PF_IN, m->m_pkthdr.rcvif, &m) != PF_PASS)
400 		goto bad;
401 	if (m == NULL)
402 		return;
403 
404 	ip = mtod(m, struct ip *);
405 	hlen = ip->ip_hl << 2;
406 	pfrdr = (pfrdr != ip->ip_dst.s_addr);
407 #endif
408 
409 	/*
410 	 * Process options and, if not destined for us,
411 	 * ship it on.  ip_dooptions returns 1 when an
412 	 * error was detected (causing an icmp message
413 	 * to be sent and the original packet to be freed).
414 	 */
415 	ip_nhops = 0;		/* for source routed packets */
416 	if (hlen > sizeof (struct ip) && ip_dooptions(m)) {
417 	        return;
418 	}
419 
420 	/*
421 	 * Check our list of addresses, to see if the packet is for us.
422 	 */
423 	if ((ia = in_iawithaddr(ip->ip_dst, m)) != NULL &&
424 	    (ia->ia_ifp->if_flags & IFF_UP))
425 		goto ours;
426 
427 	if (IN_MULTICAST(ip->ip_dst.s_addr)) {
428 		struct in_multi *inm;
429 #ifdef MROUTING
430 		extern struct socket *ip_mrouter;
431 
432 		if (m->m_flags & M_EXT) {
433 			if ((m = m_pullup(m, hlen)) == NULL) {
434 				ipstat.ips_toosmall++;
435 				return;
436 			}
437 			ip = mtod(m, struct ip *);
438 		}
439 
440 		if (ip_mrouter) {
441 			/*
442 			 * If we are acting as a multicast router, all
443 			 * incoming multicast packets are passed to the
444 			 * kernel-level multicast forwarding function.
445 			 * The packet is returned (relatively) intact; if
446 			 * ip_mforward() returns a non-zero value, the packet
447 			 * must be discarded, else it may be accepted below.
448 			 *
449 			 * (The IP ident field is put in the same byte order
450 			 * as expected when ip_mforward() is called from
451 			 * ip_output().)
452 			 */
453 			if (ip_mforward(m, m->m_pkthdr.rcvif) != 0) {
454 				ipstat.ips_cantforward++;
455 				m_freem(m);
456 				return;
457 			}
458 
459 			/*
460 			 * The process-level routing demon needs to receive
461 			 * all multicast IGMP packets, whether or not this
462 			 * host belongs to their destination groups.
463 			 */
464 			if (ip->ip_p == IPPROTO_IGMP)
465 				goto ours;
466 			ipstat.ips_forward++;
467 		}
468 #endif
469 		/*
470 		 * See if we belong to the destination multicast group on the
471 		 * arrival interface.
472 		 */
473 		IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
474 		if (inm == NULL) {
475 			ipstat.ips_cantforward++;
476 			m_freem(m);
477 			return;
478 		}
479 		goto ours;
480 	}
481 	if (ip->ip_dst.s_addr == INADDR_BROADCAST ||
482 	    ip->ip_dst.s_addr == INADDR_ANY)
483 		goto ours;
484 
485 	/*
486 	 * Not for us; forward if possible and desirable.
487 	 */
488 	if (ipforwarding == 0) {
489 		ipstat.ips_cantforward++;
490 		m_freem(m);
491 	} else {
492 #ifdef IPSEC
493 	        /*
494 		 * IPsec policy check for forwarded packets. Look at
495 		 * inner-most IPsec SA used.
496 		 */
497 		mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
498                 s = splnet();
499 		if (mtag != NULL) {
500 			tdbi = (struct tdb_ident *)(mtag + 1);
501 			tdb = gettdb(tdbi->spi, &tdbi->dst, tdbi->proto);
502 		} else
503 			tdb = NULL;
504 	        ipsp_spd_lookup(m, AF_INET, hlen, &error,
505 		    IPSP_DIRECTION_IN, tdb, NULL);
506                 splx(s);
507 
508 		/* Error or otherwise drop-packet indication */
509 		if (error) {
510 			ipstat.ips_cantforward++;
511 			m_freem(m);
512 			return;
513 		}
514 
515 		/*
516 		 * Fall through, forward packet. Outbound IPsec policy
517 		 * checking will occur in ip_output().
518 		 */
519 #endif /* IPSEC */
520 
521 		ip_forward(m, pfrdr);
522 	}
523 	return;
524 
525 ours:
526 	/*
527 	 * If offset or IP_MF are set, must reassemble.
528 	 * Otherwise, nothing need be done.
529 	 * (We could look in the reassembly queue to see
530 	 * if the packet was previously fragmented,
531 	 * but it's not worth the time; just let them time out.)
532 	 */
533 	if (ip->ip_off &~ htons(IP_DF | IP_RF)) {
534 		if (m->m_flags & M_EXT) {		/* XXX */
535 			if ((m = m_pullup(m, hlen)) == NULL) {
536 				ipstat.ips_toosmall++;
537 				return;
538 			}
539 			ip = mtod(m, struct ip *);
540 		}
541 
542 		/*
543 		 * Look for queue of fragments
544 		 * of this datagram.
545 		 */
546 		ipq_lock();
547 		for (fp = ipq.lh_first; fp != NULL; fp = fp->ipq_q.le_next)
548 			if (ip->ip_id == fp->ipq_id &&
549 			    ip->ip_src.s_addr == fp->ipq_src.s_addr &&
550 			    ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
551 			    ip->ip_p == fp->ipq_p)
552 				goto found;
553 		fp = 0;
554 found:
555 
556 		/*
557 		 * Adjust ip_len to not reflect header,
558 		 * set ipqe_mff if more fragments are expected,
559 		 * convert offset of this to bytes.
560 		 */
561 		ip->ip_len = htons(ntohs(ip->ip_len) - hlen);
562 		mff = (ip->ip_off & htons(IP_MF)) != 0;
563 		if (mff) {
564 			/*
565 			 * Make sure that fragments have a data length
566 			 * that's a non-zero multiple of 8 bytes.
567 			 */
568 			if (ntohs(ip->ip_len) == 0 ||
569 			    (ntohs(ip->ip_len) & 0x7) != 0) {
570 				ipstat.ips_badfrags++;
571 				ipq_unlock();
572 				goto bad;
573 			}
574 		}
575 		ip->ip_off = htons(ntohs(ip->ip_off) << 3);
576 
577 		/*
578 		 * If datagram marked as having more fragments
579 		 * or if this is not the first fragment,
580 		 * attempt reassembly; if it succeeds, proceed.
581 		 */
582 		if (mff || ip->ip_off) {
583 			ipstat.ips_fragments++;
584 			if (ip_frags + 1 > ip_maxqueue) {
585 				ip_flush();
586 				ipstat.ips_rcvmemdrop++;
587 				ipq_unlock();
588 				goto bad;
589 			}
590 
591 			ipqe = pool_get(&ipqent_pool, PR_NOWAIT);
592 			if (ipqe == NULL) {
593 				ipstat.ips_rcvmemdrop++;
594 				ipq_unlock();
595 				goto bad;
596 			}
597 			ip_frags++;
598 			ipqe->ipqe_mff = mff;
599 			ipqe->ipqe_m = m;
600 			ipqe->ipqe_ip = ip;
601 			m = ip_reass(ipqe, fp);
602 			if (m == 0) {
603 				ipq_unlock();
604 				return;
605 			}
606 			ipstat.ips_reassembled++;
607 			ip = mtod(m, struct ip *);
608 			hlen = ip->ip_hl << 2;
609 			ip->ip_len = htons(ntohs(ip->ip_len) + hlen);
610 		} else
611 			if (fp)
612 				ip_freef(fp);
613 		ipq_unlock();
614 	}
615 
616 #ifdef IPSEC
617         /*
618          * If it's a protected packet for us, skip the policy check.
619          * That's because we really only care about the properties of
620          * the protected packet, and not the intermediate versions.
621          * While this is not the most paranoid setting, it allows
622          * some flexibility in handling nested tunnels (in setting up
623 	 * the policies).
624          */
625         if ((ip->ip_p == IPPROTO_ESP) || (ip->ip_p == IPPROTO_AH) ||
626 	    (ip->ip_p == IPPROTO_IPCOMP))
627           goto skipipsec;
628 
629 	/*
630 	 * If the protected packet was tunneled, then we need to
631 	 * verify the protected packet's information, not the
632 	 * external headers. Thus, skip the policy lookup for the
633 	 * external packet, and keep the IPsec information linked on
634 	 * the packet header (the encapsulation routines know how
635 	 * to deal with that).
636 	 */
637 	if ((ip->ip_p == IPPROTO_IPIP) || (ip->ip_p == IPPROTO_IPV6))
638 	  goto skipipsec;
639 
640 	/*
641 	 * If the protected packet is TCP or UDP, we'll do the
642 	 * policy check in the respective input routine, so we can
643 	 * check for bypass sockets.
644 	 */
645 	if ((ip->ip_p == IPPROTO_TCP) || (ip->ip_p == IPPROTO_UDP))
646 	  goto skipipsec;
647 
648 	/*
649 	 * IPsec policy check for local-delivery packets. Look at the
650 	 * inner-most SA that protected the packet. This is in fact
651 	 * a bit too restrictive (it could end up causing packets to
652 	 * be dropped that semantically follow the policy, e.g., in
653 	 * certain SA-bundle configurations); but the alternative is
654 	 * very complicated (and requires keeping track of what
655 	 * kinds of tunneling headers have been seen in-between the
656 	 * IPsec headers), and I don't think we lose much functionality
657 	 * that's needed in the real world (who uses bundles anyway ?).
658 	 */
659 	mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
660         s = splnet();
661 	if (mtag) {
662 		tdbi = (struct tdb_ident *)(mtag + 1);
663 	        tdb = gettdb(tdbi->spi, &tdbi->dst, tdbi->proto);
664 	} else
665 		tdb = NULL;
666 	ipsp_spd_lookup(m, AF_INET, hlen, &error, IPSP_DIRECTION_IN,
667 	    tdb, NULL);
668         splx(s);
669 
670 	/* Error or otherwise drop-packet indication. */
671 	if (error) {
672 	        ipstat.ips_cantforward++;
673 		m_freem(m);
674 		return;
675 	}
676 
677  skipipsec:
678 	/* Otherwise, just fall through and deliver the packet */
679 #endif /* IPSEC */
680 
681 	/*
682 	 * Switch out to protocol's input routine.
683 	 */
684 	ipstat.ips_delivered++;
685 	(*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen, NULL, 0);
686 	return;
687 bad:
688 	m_freem(m);
689 }
690 
691 struct in_ifaddr *
in_iawithaddr(ina,m)692 in_iawithaddr(ina, m)
693 	struct in_addr ina;
694 	struct mbuf *m;
695 {
696 	struct in_ifaddr *ia;
697 
698 	TAILQ_FOREACH(ia, &in_ifaddr, ia_list) {
699 		if ((ina.s_addr == ia->ia_addr.sin_addr.s_addr) ||
700 		    ((ia->ia_ifp->if_flags & (IFF_LOOPBACK|IFF_LINK1)) ==
701 			(IFF_LOOPBACK|IFF_LINK1) &&
702 		     ia->ia_subnet == (ina.s_addr & ia->ia_subnetmask)))
703 			return ia;
704 		if (((ip_directedbcast == 0) || (m && ip_directedbcast &&
705 		    ia->ia_ifp == m->m_pkthdr.rcvif)) &&
706 		    (ia->ia_ifp->if_flags & IFF_BROADCAST)) {
707 			if (ina.s_addr == ia->ia_broadaddr.sin_addr.s_addr ||
708 			    ina.s_addr == ia->ia_netbroadcast.s_addr ||
709 			    /*
710 			     * Look for all-0's host part (old broadcast addr),
711 			     * either for subnet or net.
712 			     */
713 			    ina.s_addr == ia->ia_subnet ||
714 			    ina.s_addr == ia->ia_net) {
715 				/* Make sure M_BCAST is set */
716 				if (m)
717 					m->m_flags |= M_BCAST;
718 				return ia;
719 			    }
720 		}
721 	}
722 
723 	return NULL;
724 }
725 
726 /*
727  * Take incoming datagram fragment and try to
728  * reassemble it into whole datagram.  If a chain for
729  * reassembly of this datagram already exists, then it
730  * is given as fp; otherwise have to make a chain.
731  */
732 struct mbuf *
ip_reass(ipqe,fp)733 ip_reass(ipqe, fp)
734 	struct ipqent *ipqe;
735 	struct ipq *fp;
736 {
737 	struct mbuf *m = ipqe->ipqe_m;
738 	struct ipqent *nq, *p, *q;
739 	struct ip *ip;
740 	struct mbuf *t;
741 	int hlen = ipqe->ipqe_ip->ip_hl << 2;
742 	int i, next;
743 	u_int8_t ecn, ecn0;
744 
745 	/*
746 	 * Presence of header sizes in mbufs
747 	 * would confuse code below.
748 	 */
749 	m->m_data += hlen;
750 	m->m_len -= hlen;
751 
752 	/*
753 	 * If first fragment to arrive, create a reassembly queue.
754 	 */
755 	if (fp == 0) {
756 		MALLOC(fp, struct ipq *, sizeof (struct ipq),
757 		    M_FTABLE, M_NOWAIT);
758 		if (fp == NULL)
759 			goto dropfrag;
760 		LIST_INSERT_HEAD(&ipq, fp, ipq_q);
761 		fp->ipq_ttl = IPFRAGTTL;
762 		fp->ipq_p = ipqe->ipqe_ip->ip_p;
763 		fp->ipq_id = ipqe->ipqe_ip->ip_id;
764 		LIST_INIT(&fp->ipq_fragq);
765 		fp->ipq_src = ipqe->ipqe_ip->ip_src;
766 		fp->ipq_dst = ipqe->ipqe_ip->ip_dst;
767 		p = NULL;
768 		goto insert;
769 	}
770 
771 	/*
772 	 * Handle ECN by comparing this segment with the first one;
773 	 * if CE is set, do not lose CE.
774 	 * drop if CE and not-ECT are mixed for the same packet.
775 	 */
776 	ecn = ipqe->ipqe_ip->ip_tos & IPTOS_ECN_MASK;
777 	ecn0 = fp->ipq_fragq.lh_first->ipqe_ip->ip_tos & IPTOS_ECN_MASK;
778 	if (ecn == IPTOS_ECN_CE) {
779 		if (ecn0 == IPTOS_ECN_NOTECT)
780 			goto dropfrag;
781 		if (ecn0 != IPTOS_ECN_CE)
782 			fp->ipq_fragq.lh_first->ipqe_ip->ip_tos |= IPTOS_ECN_CE;
783 	}
784 	if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT)
785 		goto dropfrag;
786 
787 	/*
788 	 * Find a segment which begins after this one does.
789 	 */
790 	for (p = NULL, q = fp->ipq_fragq.lh_first; q != NULL;
791 	    p = q, q = q->ipqe_q.le_next)
792 		if (ntohs(q->ipqe_ip->ip_off) > ntohs(ipqe->ipqe_ip->ip_off))
793 			break;
794 
795 	/*
796 	 * If there is a preceding segment, it may provide some of
797 	 * our data already.  If so, drop the data from the incoming
798 	 * segment.  If it provides all of our data, drop us.
799 	 */
800 	if (p != NULL) {
801 		i = ntohs(p->ipqe_ip->ip_off) + ntohs(p->ipqe_ip->ip_len) -
802 		    ntohs(ipqe->ipqe_ip->ip_off);
803 		if (i > 0) {
804 			if (i >= ntohs(ipqe->ipqe_ip->ip_len))
805 				goto dropfrag;
806 			m_adj(ipqe->ipqe_m, i);
807 			ipqe->ipqe_ip->ip_off =
808 			    htons(ntohs(ipqe->ipqe_ip->ip_off) + i);
809 			ipqe->ipqe_ip->ip_len =
810 			    htons(ntohs(ipqe->ipqe_ip->ip_len) - i);
811 		}
812 	}
813 
814 	/*
815 	 * While we overlap succeeding segments trim them or,
816 	 * if they are completely covered, dequeue them.
817 	 */
818 	for (; q != NULL &&
819 	    ntohs(ipqe->ipqe_ip->ip_off) + ntohs(ipqe->ipqe_ip->ip_len) >
820 	    ntohs(q->ipqe_ip->ip_off); q = nq) {
821 		i = (ntohs(ipqe->ipqe_ip->ip_off) +
822 		    ntohs(ipqe->ipqe_ip->ip_len)) - ntohs(q->ipqe_ip->ip_off);
823 		if (i < ntohs(q->ipqe_ip->ip_len)) {
824 			q->ipqe_ip->ip_len =
825 			    htons(ntohs(q->ipqe_ip->ip_len) - i);
826 			q->ipqe_ip->ip_off =
827 			    htons(ntohs(q->ipqe_ip->ip_off) + i);
828 			m_adj(q->ipqe_m, i);
829 			break;
830 		}
831 		nq = q->ipqe_q.le_next;
832 		m_freem(q->ipqe_m);
833 		LIST_REMOVE(q, ipqe_q);
834 		pool_put(&ipqent_pool, q);
835 		ip_frags--;
836 	}
837 
838 insert:
839 	/*
840 	 * Stick new segment in its place;
841 	 * check for complete reassembly.
842 	 */
843 	if (p == NULL) {
844 		LIST_INSERT_HEAD(&fp->ipq_fragq, ipqe, ipqe_q);
845 	} else {
846 		LIST_INSERT_AFTER(p, ipqe, ipqe_q);
847 	}
848 	next = 0;
849 	for (p = NULL, q = fp->ipq_fragq.lh_first; q != NULL;
850 	    p = q, q = q->ipqe_q.le_next) {
851 		if (ntohs(q->ipqe_ip->ip_off) != next)
852 			return (0);
853 		next += ntohs(q->ipqe_ip->ip_len);
854 	}
855 	if (p->ipqe_mff)
856 		return (0);
857 
858 	/*
859 	 * Reassembly is complete.  Check for a bogus message size and
860 	 * concatenate fragments.
861 	 */
862 	q = fp->ipq_fragq.lh_first;
863 	ip = q->ipqe_ip;
864 	if ((next + (ip->ip_hl << 2)) > IP_MAXPACKET) {
865 		ipstat.ips_toolong++;
866 		ip_freef(fp);
867 		return (0);
868 	}
869 	m = q->ipqe_m;
870 	t = m->m_next;
871 	m->m_next = 0;
872 	m_cat(m, t);
873 	nq = q->ipqe_q.le_next;
874 	pool_put(&ipqent_pool, q);
875 	ip_frags--;
876 	for (q = nq; q != NULL; q = nq) {
877 		t = q->ipqe_m;
878 		nq = q->ipqe_q.le_next;
879 		pool_put(&ipqent_pool, q);
880 		ip_frags--;
881 		m_cat(m, t);
882 	}
883 
884 	/*
885 	 * Create header for new ip packet by
886 	 * modifying header of first packet;
887 	 * dequeue and discard fragment reassembly header.
888 	 * Make header visible.
889 	 */
890 	ip->ip_len = htons(next);
891 	ip->ip_src = fp->ipq_src;
892 	ip->ip_dst = fp->ipq_dst;
893 	LIST_REMOVE(fp, ipq_q);
894 	FREE(fp, M_FTABLE);
895 	m->m_len += (ip->ip_hl << 2);
896 	m->m_data -= (ip->ip_hl << 2);
897 	/* some debugging cruft by sklower, below, will go away soon */
898 	if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */
899 		int plen = 0;
900 		for (t = m; t; t = t->m_next)
901 			plen += t->m_len;
902 		m->m_pkthdr.len = plen;
903 	}
904 	return (m);
905 
906 dropfrag:
907 	ipstat.ips_fragdropped++;
908 	m_freem(m);
909 	pool_put(&ipqent_pool, ipqe);
910 	ip_frags--;
911 	return (0);
912 }
913 
914 /*
915  * Free a fragment reassembly header and all
916  * associated datagrams.
917  */
918 void
ip_freef(fp)919 ip_freef(fp)
920 	struct ipq *fp;
921 {
922 	struct ipqent *q, *p;
923 
924 	for (q = fp->ipq_fragq.lh_first; q != NULL; q = p) {
925 		p = q->ipqe_q.le_next;
926 		m_freem(q->ipqe_m);
927 		LIST_REMOVE(q, ipqe_q);
928 		pool_put(&ipqent_pool, q);
929 		ip_frags--;
930 	}
931 	LIST_REMOVE(fp, ipq_q);
932 	FREE(fp, M_FTABLE);
933 }
934 
935 /*
936  * IP timer processing;
937  * if a timer expires on a reassembly
938  * queue, discard it.
939  */
940 void
ip_slowtimo()941 ip_slowtimo()
942 {
943 	struct ipq *fp, *nfp;
944 	int s = splsoftnet();
945 
946 	ipq_lock();
947 	for (fp = ipq.lh_first; fp != NULL; fp = nfp) {
948 		nfp = fp->ipq_q.le_next;
949 		if (--fp->ipq_ttl == 0) {
950 			ipstat.ips_fragtimeout++;
951 			ip_freef(fp);
952 		}
953 	}
954 	ipq_unlock();
955 	splx(s);
956 }
957 
958 /*
959  * Drain off all datagram fragments.
960  */
961 void
ip_drain()962 ip_drain()
963 {
964 
965 	if (ipq_lock_try() == 0)
966 		return;
967 	while (ipq.lh_first != NULL) {
968 		ipstat.ips_fragdropped++;
969 		ip_freef(ipq.lh_first);
970 	}
971 	ipq_unlock();
972 }
973 
974 /*
975  * Flush a bunch of datagram fragments, till we are down to 75%.
976  */
977 void
ip_flush()978 ip_flush()
979 {
980 	int max = 50;
981 
982 	/* ipq already locked */
983 	while (ipq.lh_first != NULL && ip_frags > ip_maxqueue * 3 / 4 && --max) {
984 		ipstat.ips_fragdropped++;
985 		ip_freef(ipq.lh_first);
986 	}
987 }
988 
989 /*
990  * Do option processing on a datagram,
991  * possibly discarding it if bad options are encountered,
992  * or forwarding it if source-routed.
993  * Returns 1 if packet has been forwarded/freed,
994  * 0 if the packet should be processed further.
995  */
996 int
ip_dooptions(m)997 ip_dooptions(m)
998 	struct mbuf *m;
999 {
1000 	struct ip *ip = mtod(m, struct ip *);
1001 	u_char *cp;
1002 	struct ip_timestamp ipt;
1003 	struct in_ifaddr *ia;
1004 	int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0;
1005 	struct in_addr sin, dst;
1006 	n_time ntime;
1007 
1008 	dst = ip->ip_dst;
1009 	cp = (u_char *)(ip + 1);
1010 	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
1011 
1012 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1013 		opt = cp[IPOPT_OPTVAL];
1014 		if (opt == IPOPT_EOL)
1015 			break;
1016 		if (opt == IPOPT_NOP)
1017 			optlen = 1;
1018 		else {
1019 			if (cnt < IPOPT_OLEN + sizeof(*cp)) {
1020 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1021 				goto bad;
1022 			}
1023 			optlen = cp[IPOPT_OLEN];
1024 			if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
1025 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1026 				goto bad;
1027 			}
1028 		}
1029 
1030 		switch (opt) {
1031 
1032 		default:
1033 			break;
1034 
1035 		/*
1036 		 * Source routing with record.
1037 		 * Find interface with current destination address.
1038 		 * If none on this machine then drop if strictly routed,
1039 		 * or do nothing if loosely routed.
1040 		 * Record interface address and bring up next address
1041 		 * component.  If strictly routed make sure next
1042 		 * address is on directly accessible net.
1043 		 */
1044 		case IPOPT_LSRR:
1045 		case IPOPT_SSRR:
1046 			if (!ip_dosourceroute) {
1047 				char buf[4*sizeof "123"];
1048 
1049 				strlcpy(buf, inet_ntoa(ip->ip_dst),
1050 				    sizeof buf);
1051 				log(LOG_WARNING,
1052 				    "attempted source route from %s to %s\n",
1053 				    inet_ntoa(ip->ip_src), buf);
1054 				type = ICMP_UNREACH;
1055 				code = ICMP_UNREACH_SRCFAIL;
1056 				goto bad;
1057 			}
1058 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1059 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1060 				goto bad;
1061 			}
1062 			ipaddr.sin_addr = ip->ip_dst;
1063 			ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr)));
1064 			if (ia == 0) {
1065 				if (opt == IPOPT_SSRR) {
1066 					type = ICMP_UNREACH;
1067 					code = ICMP_UNREACH_SRCFAIL;
1068 					goto bad;
1069 				}
1070 				/*
1071 				 * Loose routing, and not at next destination
1072 				 * yet; nothing to do except forward.
1073 				 */
1074 				break;
1075 			}
1076 			off--;			/* 0 origin */
1077 			if ((off + sizeof(struct in_addr)) > optlen) {
1078 				/*
1079 				 * End of source route.  Should be for us.
1080 				 */
1081 				save_rte(cp, ip->ip_src);
1082 				break;
1083 			}
1084 
1085 			/*
1086 			 * locate outgoing interface
1087 			 */
1088 			bcopy((caddr_t)(cp + off), (caddr_t)&ipaddr.sin_addr,
1089 			    sizeof(ipaddr.sin_addr));
1090 			if (opt == IPOPT_SSRR) {
1091 #define	INA	struct in_ifaddr *
1092 #define	SA	struct sockaddr *
1093 			    if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == 0)
1094 				ia = (INA)ifa_ifwithnet((SA)&ipaddr);
1095 			} else
1096 				ia = ip_rtaddr(ipaddr.sin_addr);
1097 			if (ia == 0) {
1098 				type = ICMP_UNREACH;
1099 				code = ICMP_UNREACH_SRCFAIL;
1100 				goto bad;
1101 			}
1102 			ip->ip_dst = ipaddr.sin_addr;
1103 			bcopy((caddr_t)&ia->ia_addr.sin_addr,
1104 			    (caddr_t)(cp + off), sizeof(struct in_addr));
1105 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1106 			/*
1107 			 * Let ip_intr's mcast routing check handle mcast pkts
1108 			 */
1109 			forward = !IN_MULTICAST(ip->ip_dst.s_addr);
1110 			break;
1111 
1112 		case IPOPT_RR:
1113 			if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1114 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1115 				goto bad;
1116 			}
1117 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1118 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1119 				goto bad;
1120 			}
1121 
1122 			/*
1123 			 * If no space remains, ignore.
1124 			 */
1125 			off--;			/* 0 origin */
1126 			if ((off + sizeof(struct in_addr)) > optlen)
1127 				break;
1128 			bcopy((caddr_t)(&ip->ip_dst), (caddr_t)&ipaddr.sin_addr,
1129 			    sizeof(ipaddr.sin_addr));
1130 			/*
1131 			 * locate outgoing interface; if we're the destination,
1132 			 * use the incoming interface (should be same).
1133 			 */
1134 			if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == 0 &&
1135 			    (ia = ip_rtaddr(ipaddr.sin_addr)) == 0) {
1136 				type = ICMP_UNREACH;
1137 				code = ICMP_UNREACH_HOST;
1138 				goto bad;
1139 			}
1140 			bcopy((caddr_t)&ia->ia_addr.sin_addr,
1141 			    (caddr_t)(cp + off), sizeof(struct in_addr));
1142 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1143 			break;
1144 
1145 		case IPOPT_TS:
1146 			code = cp - (u_char *)ip;
1147 			if (optlen < sizeof(struct ip_timestamp))
1148 				goto bad;
1149 			bcopy(cp, &ipt, sizeof(struct ip_timestamp));
1150 			if (ipt.ipt_ptr < 5 || ipt.ipt_len < 5)
1151 				goto bad;
1152 			if (ipt.ipt_ptr - 1 + sizeof(n_time) > ipt.ipt_len) {
1153 				if (++ipt.ipt_oflw == 0)
1154 					goto bad;
1155 				break;
1156 			}
1157 			bcopy(cp + ipt.ipt_ptr - 1, &sin, sizeof sin);
1158 			switch (ipt.ipt_flg) {
1159 
1160 			case IPOPT_TS_TSONLY:
1161 				break;
1162 
1163 			case IPOPT_TS_TSANDADDR:
1164 				if (ipt.ipt_ptr - 1 + sizeof(n_time) +
1165 				    sizeof(struct in_addr) > ipt.ipt_len)
1166 					goto bad;
1167 				ipaddr.sin_addr = dst;
1168 				ia = (INA)ifaof_ifpforaddr((SA)&ipaddr,
1169 							    m->m_pkthdr.rcvif);
1170 				if (ia == 0)
1171 					continue;
1172 				bcopy((caddr_t)&ia->ia_addr.sin_addr,
1173 				    (caddr_t)&sin, sizeof(struct in_addr));
1174 				ipt.ipt_ptr += sizeof(struct in_addr);
1175 				break;
1176 
1177 			case IPOPT_TS_PRESPEC:
1178 				if (ipt.ipt_ptr - 1 + sizeof(n_time) +
1179 				    sizeof(struct in_addr) > ipt.ipt_len)
1180 					goto bad;
1181 				bcopy((caddr_t)&sin, (caddr_t)&ipaddr.sin_addr,
1182 				    sizeof(struct in_addr));
1183 				if (ifa_ifwithaddr((SA)&ipaddr) == 0)
1184 					continue;
1185 				ipt.ipt_ptr += sizeof(struct in_addr);
1186 				break;
1187 
1188 			default:
1189 				/* XXX can't take &ipt->ipt_flg */
1190 				code = (u_char *)&ipt.ipt_ptr -
1191 				    (u_char *)ip + 1;
1192 				goto bad;
1193 			}
1194 			ntime = iptime();
1195 			bcopy((caddr_t)&ntime, (caddr_t)cp + ipt.ipt_ptr - 1,
1196 			    sizeof(n_time));
1197 			ipt.ipt_ptr += sizeof(n_time);
1198 		}
1199 	}
1200 	if (forward && ipforwarding) {
1201 		ip_forward(m, 1);
1202 		return (1);
1203 	}
1204 	return (0);
1205 bad:
1206 	icmp_error(m, type, code, 0, 0);
1207 	ipstat.ips_badoptions++;
1208 	return (1);
1209 }
1210 
1211 /*
1212  * Given address of next destination (final or next hop),
1213  * return internet address info of interface to be used to get there.
1214  */
1215 struct in_ifaddr *
ip_rtaddr(dst)1216 ip_rtaddr(dst)
1217 	 struct in_addr dst;
1218 {
1219 	struct sockaddr_in *sin;
1220 
1221 	sin = satosin(&ipforward_rt.ro_dst);
1222 
1223 	if (ipforward_rt.ro_rt == 0 || dst.s_addr != sin->sin_addr.s_addr) {
1224 		if (ipforward_rt.ro_rt) {
1225 			RTFREE(ipforward_rt.ro_rt);
1226 			ipforward_rt.ro_rt = 0;
1227 		}
1228 		sin->sin_family = AF_INET;
1229 		sin->sin_len = sizeof(*sin);
1230 		sin->sin_addr = dst;
1231 
1232 		rtalloc(&ipforward_rt);
1233 	}
1234 	if (ipforward_rt.ro_rt == 0)
1235 		return ((struct in_ifaddr *)0);
1236 	return (ifatoia(ipforward_rt.ro_rt->rt_ifa));
1237 }
1238 
1239 /*
1240  * Save incoming source route for use in replies,
1241  * to be picked up later by ip_srcroute if the receiver is interested.
1242  */
1243 void
save_rte(option,dst)1244 save_rte(option, dst)
1245 	u_char *option;
1246 	struct in_addr dst;
1247 {
1248 	unsigned olen;
1249 
1250 	olen = option[IPOPT_OLEN];
1251 #ifdef DIAGNOSTIC
1252 	if (ipprintfs)
1253 		printf("save_rte: olen %d\n", olen);
1254 #endif /* 0 */
1255 	if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst)))
1256 		return;
1257 	bcopy((caddr_t)option, (caddr_t)ip_srcrt.srcopt, olen);
1258 	ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1259 	ip_srcrt.dst = dst;
1260 }
1261 
1262 /*
1263  * Check whether we do proxy ARP for this address and we point to ourselves.
1264  * Code shamelessly copied from arplookup().
1265  */
1266 static int
ip_weadvertise(addr)1267 ip_weadvertise(addr)
1268 	u_int32_t addr;
1269 {
1270 	struct rtentry *rt;
1271 	struct ifnet *ifp;
1272 	struct ifaddr *ifa;
1273 	struct sockaddr_inarp sin;
1274 
1275 	sin.sin_len = sizeof(sin);
1276 	sin.sin_family = AF_INET;
1277 	sin.sin_addr.s_addr = addr;
1278 	sin.sin_other = SIN_PROXY;
1279 	rt = rtalloc1(sintosa(&sin), 0);
1280 	if (rt == 0)
1281 		return 0;
1282 
1283 	if ((rt->rt_flags & RTF_GATEWAY) || (rt->rt_flags & RTF_LLINFO) == 0 ||
1284 	    rt->rt_gateway->sa_family != AF_LINK) {
1285 		RTFREE(rt);
1286 		return 0;
1287 	}
1288 
1289 	for (ifp = ifnet.tqh_first; ifp != 0; ifp = ifp->if_list.tqe_next)
1290 		for (ifa = ifp->if_addrlist.tqh_first; ifa != 0;
1291 		    ifa = ifa->ifa_list.tqe_next) {
1292 			if (ifa->ifa_addr->sa_family != rt->rt_gateway->sa_family)
1293 				continue;
1294 
1295 			if (!bcmp(LLADDR((struct sockaddr_dl *)ifa->ifa_addr),
1296 			    LLADDR((struct sockaddr_dl *)rt->rt_gateway),
1297 			    ETHER_ADDR_LEN)) {
1298 				RTFREE(rt);
1299 				return 1;
1300 			}
1301 		}
1302 
1303 	RTFREE(rt);
1304 	return 0;
1305 }
1306 
1307 /*
1308  * Retrieve incoming source route for use in replies,
1309  * in the same form used by setsockopt.
1310  * The first hop is placed before the options, will be removed later.
1311  */
1312 struct mbuf *
ip_srcroute()1313 ip_srcroute()
1314 {
1315 	struct in_addr *p, *q;
1316 	struct mbuf *m;
1317 
1318 	if (ip_nhops == 0)
1319 		return ((struct mbuf *)0);
1320 	m = m_get(M_DONTWAIT, MT_SOOPTS);
1321 	if (m == 0)
1322 		return ((struct mbuf *)0);
1323 
1324 #define OPTSIZ	(sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt))
1325 
1326 	/* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
1327 	m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) +
1328 	    OPTSIZ;
1329 #ifdef DIAGNOSTIC
1330 	if (ipprintfs)
1331 		printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len);
1332 #endif
1333 
1334 	/*
1335 	 * First save first hop for return route
1336 	 */
1337 	p = &ip_srcrt.route[ip_nhops - 1];
1338 	*(mtod(m, struct in_addr *)) = *p--;
1339 #ifdef DIAGNOSTIC
1340 	if (ipprintfs)
1341 		printf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr));
1342 #endif
1343 
1344 	/*
1345 	 * Copy option fields and padding (nop) to mbuf.
1346 	 */
1347 	ip_srcrt.nop = IPOPT_NOP;
1348 	ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
1349 	bcopy((caddr_t)&ip_srcrt.nop,
1350 	    mtod(m, caddr_t) + sizeof(struct in_addr), OPTSIZ);
1351 	q = (struct in_addr *)(mtod(m, caddr_t) +
1352 	    sizeof(struct in_addr) + OPTSIZ);
1353 #undef OPTSIZ
1354 	/*
1355 	 * Record return path as an IP source route,
1356 	 * reversing the path (pointers are now aligned).
1357 	 */
1358 	while (p >= ip_srcrt.route) {
1359 #ifdef DIAGNOSTIC
1360 		if (ipprintfs)
1361 			printf(" %x", ntohl(q->s_addr));
1362 #endif
1363 		*q++ = *p--;
1364 	}
1365 	/*
1366 	 * Last hop goes to final destination.
1367 	 */
1368 	*q = ip_srcrt.dst;
1369 #ifdef DIAGNOSTIC
1370 	if (ipprintfs)
1371 		printf(" %x\n", ntohl(q->s_addr));
1372 #endif
1373 	return (m);
1374 }
1375 
1376 /*
1377  * Strip out IP options, at higher
1378  * level protocol in the kernel.
1379  * Second argument is buffer to which options
1380  * will be moved, and return value is their length.
1381  * XXX should be deleted; last arg currently ignored.
1382  */
1383 void
ip_stripoptions(m,mopt)1384 ip_stripoptions(m, mopt)
1385 	struct mbuf *m;
1386 	struct mbuf *mopt;
1387 {
1388 	int i;
1389 	struct ip *ip = mtod(m, struct ip *);
1390 	caddr_t opts;
1391 	int olen;
1392 
1393 	olen = (ip->ip_hl<<2) - sizeof (struct ip);
1394 	opts = (caddr_t)(ip + 1);
1395 	i = m->m_len - (sizeof (struct ip) + olen);
1396 	bcopy(opts  + olen, opts, (unsigned)i);
1397 	m->m_len -= olen;
1398 	if (m->m_flags & M_PKTHDR)
1399 		m->m_pkthdr.len -= olen;
1400 	ip->ip_hl = sizeof(struct ip) >> 2;
1401 }
1402 
1403 int inetctlerrmap[PRC_NCMDS] = {
1404 	0,		0,		0,		0,
1405 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1406 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1407 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1408 	0,		0,		0,		0,
1409 	ENOPROTOOPT
1410 };
1411 
1412 /*
1413  * Forward a packet.  If some error occurs return the sender
1414  * an icmp packet.  Note we can't always generate a meaningful
1415  * icmp message because icmp doesn't have a large enough repertoire
1416  * of codes and types.
1417  *
1418  * If not forwarding, just drop the packet.  This could be confusing
1419  * if ipforwarding was zero but some routing protocol was advancing
1420  * us as a gateway to somewhere.  However, we must let the routing
1421  * protocol deal with that.
1422  *
1423  * The srcrt parameter indicates whether the packet is being forwarded
1424  * via a source route.
1425  */
1426 void
ip_forward(m,srcrt)1427 ip_forward(m, srcrt)
1428 	struct mbuf *m;
1429 	int srcrt;
1430 {
1431 	struct ip *ip = mtod(m, struct ip *);
1432 	struct sockaddr_in *sin;
1433 	struct rtentry *rt;
1434 	int error, type = 0, code = 0;
1435 	struct mbuf *mcopy;
1436 	n_long dest;
1437 	struct ifnet *destifp;
1438 #ifdef IPSEC
1439 	struct ifnet dummyifp;
1440 #endif
1441 
1442 	dest = 0;
1443 #ifdef DIAGNOSTIC
1444 	if (ipprintfs)
1445 		printf("forward: src %x dst %x ttl %x\n", ip->ip_src.s_addr,
1446 		    ip->ip_dst.s_addr, ip->ip_ttl);
1447 #endif
1448 	if (m->m_flags & M_BCAST || in_canforward(ip->ip_dst) == 0) {
1449 		ipstat.ips_cantforward++;
1450 		m_freem(m);
1451 		return;
1452 	}
1453 	if (ip->ip_ttl <= IPTTLDEC) {
1454 		icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0);
1455 		return;
1456 	}
1457 
1458 	sin = satosin(&ipforward_rt.ro_dst);
1459 	if ((rt = ipforward_rt.ro_rt) == 0 ||
1460 	    ip->ip_dst.s_addr != sin->sin_addr.s_addr) {
1461 		if (ipforward_rt.ro_rt) {
1462 			RTFREE(ipforward_rt.ro_rt);
1463 			ipforward_rt.ro_rt = 0;
1464 		}
1465 		sin->sin_family = AF_INET;
1466 		sin->sin_len = sizeof(*sin);
1467 		sin->sin_addr = ip->ip_dst;
1468 
1469 		rtalloc(&ipforward_rt);
1470 		if (ipforward_rt.ro_rt == 0) {
1471 			icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_NET, dest, 0);
1472 			return;
1473 		}
1474 		rt = ipforward_rt.ro_rt;
1475 	}
1476 
1477 	/*
1478 	 * Save at most 68 bytes of the packet in case
1479 	 * we need to generate an ICMP message to the src.
1480 	 * Pullup to avoid sharing mbuf cluster between m and mcopy.
1481 	 */
1482 	mcopy = m_copym(m, 0, imin(ntohs(ip->ip_len), 68), M_DONTWAIT);
1483 	if (mcopy)
1484 		mcopy = m_pullup(mcopy, ip->ip_hl << 2);
1485 
1486 	ip->ip_ttl -= IPTTLDEC;
1487 
1488 	/*
1489 	 * If forwarding packet using same interface that it came in on,
1490 	 * perhaps should send a redirect to sender to shortcut a hop.
1491 	 * Only send redirect if source is sending directly to us,
1492 	 * and if packet was not source routed (or has any options).
1493 	 * Also, don't send redirect if forwarding using a default route
1494 	 * or a route modified by a redirect.
1495 	 * Don't send redirect if we advertise destination's arp address
1496 	 * as ours (proxy arp).
1497 	 */
1498 	if (rt->rt_ifp == m->m_pkthdr.rcvif &&
1499 	    (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1500 	    satosin(rt_key(rt))->sin_addr.s_addr != 0 &&
1501 	    ipsendredirects && !srcrt &&
1502 	    !ip_weadvertise(satosin(rt_key(rt))->sin_addr.s_addr)) {
1503 		if (rt->rt_ifa &&
1504 		    (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) ==
1505 		    ifatoia(rt->rt_ifa)->ia_subnet) {
1506 		    if (rt->rt_flags & RTF_GATEWAY)
1507 			dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
1508 		    else
1509 			dest = ip->ip_dst.s_addr;
1510 		    /* Router requirements says to only send host redirects */
1511 		    type = ICMP_REDIRECT;
1512 		    code = ICMP_REDIRECT_HOST;
1513 #ifdef DIAGNOSTIC
1514 		    if (ipprintfs)
1515 			printf("redirect (%d) to %x\n", code, (u_int32_t)dest);
1516 #endif
1517 		}
1518 	}
1519 
1520 	error = ip_output(m, (struct mbuf *)0, &ipforward_rt,
1521 	    (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)),
1522 	    0, (void *)NULL, (void *)NULL);
1523 	if (error)
1524 		ipstat.ips_cantforward++;
1525 	else {
1526 		ipstat.ips_forward++;
1527 		if (type)
1528 			ipstat.ips_redirectsent++;
1529 		else {
1530 			if (mcopy)
1531 				m_freem(mcopy);
1532 			return;
1533 		}
1534 	}
1535 	if (mcopy == NULL)
1536 		return;
1537 	destifp = NULL;
1538 
1539 	switch (error) {
1540 
1541 	case 0:				/* forwarded, but need redirect */
1542 		/* type, code set above */
1543 		break;
1544 
1545 	case ENETUNREACH:		/* shouldn't happen, checked above */
1546 	case EHOSTUNREACH:
1547 	case ENETDOWN:
1548 	case EHOSTDOWN:
1549 	default:
1550 		type = ICMP_UNREACH;
1551 		code = ICMP_UNREACH_HOST;
1552 		break;
1553 
1554 	case EMSGSIZE:
1555 		type = ICMP_UNREACH;
1556 		code = ICMP_UNREACH_NEEDFRAG;
1557 
1558 #ifdef IPSEC
1559 		if (ipforward_rt.ro_rt) {
1560 			struct rtentry *rt = ipforward_rt.ro_rt;
1561 			destifp = ipforward_rt.ro_rt->rt_ifp;
1562 			/*
1563 			 * XXX BUG ALERT
1564 			 * The "dummyifp" code relies upon the fact
1565 			 * that icmp_error() touches only ifp->if_mtu.
1566 			 */
1567 			if (rt->rt_rmx.rmx_mtu) {
1568 				dummyifp.if_mtu = rt->rt_rmx.rmx_mtu;
1569 				destifp = &dummyifp;
1570 			}
1571 		}
1572 #endif /*IPSEC*/
1573 		ipstat.ips_cantfrag++;
1574 		break;
1575 
1576 	case ENOBUFS:
1577 #if 1
1578 		/*
1579 		 * a router should not generate ICMP_SOURCEQUENCH as
1580 		 * required in RFC1812 Requirements for IP Version 4 Routers.
1581 		 * source quench could be a big problem under DoS attacks,
1582 		 * or the underlying interface is rate-limited.
1583 		 */
1584 		if (mcopy)
1585 			m_freem(mcopy);
1586 		return;
1587 #else
1588 		type = ICMP_SOURCEQUENCH;
1589 		code = 0;
1590 		break;
1591 #endif
1592 	}
1593 
1594 	icmp_error(mcopy, type, code, dest, destifp);
1595 }
1596 
1597 int
ip_sysctl(name,namelen,oldp,oldlenp,newp,newlen)1598 ip_sysctl(name, namelen, oldp, oldlenp, newp, newlen)
1599 	int *name;
1600 	u_int namelen;
1601 	void *oldp;
1602 	size_t *oldlenp;
1603 	void *newp;
1604 	size_t newlen;
1605 {
1606 	int error;
1607 
1608 	/* All sysctl names at this level are terminal. */
1609 	if (namelen != 1)
1610 		return (ENOTDIR);
1611 
1612 	switch (name[0]) {
1613 #ifdef notyet
1614 	case IPCTL_DEFMTU:
1615 		return (sysctl_int(oldp, oldlenp, newp, newlen, &ip_mtu));
1616 #endif
1617 	case IPCTL_SOURCEROUTE:
1618 		/*
1619 		 * Don't allow this to change in a secure environment.
1620 		 */
1621 		if (newp && securelevel > 0)
1622 			return (EPERM);
1623 		return (sysctl_int(oldp, oldlenp, newp, newlen,
1624 		    &ip_dosourceroute));
1625 	case IPCTL_MTUDISC:
1626 		error = sysctl_int(oldp, oldlenp, newp, newlen,
1627 		    &ip_mtudisc);
1628 		if (ip_mtudisc != 0 && ip_mtudisc_timeout_q == NULL) {
1629 			ip_mtudisc_timeout_q =
1630 			    rt_timer_queue_create(ip_mtudisc_timeout);
1631 		} else if (ip_mtudisc == 0 && ip_mtudisc_timeout_q != NULL) {
1632 			rt_timer_queue_destroy(ip_mtudisc_timeout_q, TRUE);
1633 			Free(ip_mtudisc_timeout_q);
1634 			ip_mtudisc_timeout_q = NULL;
1635 		}
1636 		return error;
1637 	case IPCTL_MTUDISCTIMEOUT:
1638 		error = sysctl_int(oldp, oldlenp, newp, newlen,
1639 		   &ip_mtudisc_timeout);
1640 		if (ip_mtudisc_timeout_q != NULL)
1641 			rt_timer_queue_change(ip_mtudisc_timeout_q,
1642 					      ip_mtudisc_timeout);
1643 		return (error);
1644 	case IPCTL_IPSEC_ENC_ALGORITHM:
1645 	        return (sysctl_tstring(oldp, oldlenp, newp, newlen,
1646 				       ipsec_def_enc, sizeof(ipsec_def_enc)));
1647 	case IPCTL_IPSEC_AUTH_ALGORITHM:
1648 	        return (sysctl_tstring(oldp, oldlenp, newp, newlen,
1649 				       ipsec_def_auth,
1650 				       sizeof(ipsec_def_auth)));
1651 	case IPCTL_IPSEC_IPCOMP_ALGORITHM:
1652 	        return (sysctl_tstring(oldp, oldlenp, newp, newlen,
1653 				       ipsec_def_comp,
1654 				       sizeof(ipsec_def_comp)));
1655 	default:
1656 		if (name[0] < IPCTL_MAXID)
1657 			return (sysctl_int_arr(ipctl_vars, name, namelen,
1658 			    oldp, oldlenp, newp, newlen));
1659 		return (EOPNOTSUPP);
1660 	}
1661 	/* NOTREACHED */
1662 }
1663