xref: /openbsd-src/sys/netinet/ip_input.c (revision a28daedfc357b214be5c701aa8ba8adb29a7f1c2)
1 /*	$OpenBSD: ip_input.c,v 1.161 2008/12/24 07:41:59 dlg 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 #include "carp.h"
37 
38 #include <sys/param.h>
39 #include <sys/systm.h>
40 #include <sys/mbuf.h>
41 #include <sys/domain.h>
42 #include <sys/protosw.h>
43 #include <sys/socket.h>
44 #include <sys/socketvar.h>
45 #include <sys/syslog.h>
46 #include <sys/sysctl.h>
47 #include <sys/pool.h>
48 
49 #include <net/if.h>
50 #include <net/if_dl.h>
51 #include <net/route.h>
52 
53 #include <netinet/in.h>
54 #include <netinet/in_systm.h>
55 #include <netinet/if_ether.h>
56 #include <netinet/ip.h>
57 #include <netinet/in_pcb.h>
58 #include <netinet/in_var.h>
59 #include <netinet/ip_var.h>
60 #include <netinet/ip_icmp.h>
61 
62 #if NPF > 0
63 #include <net/pfvar.h>
64 #endif
65 
66 #ifdef MROUTING
67 #include <netinet/ip_mroute.h>
68 #endif
69 
70 #ifdef IPSEC
71 #include <netinet/ip_ipsp.h>
72 #endif /* IPSEC */
73 
74 #if NCARP > 0
75 #include <net/if_types.h>
76 #include <netinet/ip_carp.h>
77 #endif
78 
79 #define IPMTUDISCTIMEOUT (10 * 60)	/* as per RFC 1191 */
80 
81 struct ipqhead ipq;
82 
83 int encdebug = 0;
84 int ipsec_keep_invalid = IPSEC_DEFAULT_EMBRYONIC_SA_TIMEOUT;
85 int ipsec_require_pfs = IPSEC_DEFAULT_PFS;
86 int ipsec_soft_allocations = IPSEC_DEFAULT_SOFT_ALLOCATIONS;
87 int ipsec_exp_allocations = IPSEC_DEFAULT_EXP_ALLOCATIONS;
88 int ipsec_soft_bytes = IPSEC_DEFAULT_SOFT_BYTES;
89 int ipsec_exp_bytes = IPSEC_DEFAULT_EXP_BYTES;
90 int ipsec_soft_timeout = IPSEC_DEFAULT_SOFT_TIMEOUT;
91 int ipsec_exp_timeout = IPSEC_DEFAULT_EXP_TIMEOUT;
92 int ipsec_soft_first_use = IPSEC_DEFAULT_SOFT_FIRST_USE;
93 int ipsec_exp_first_use = IPSEC_DEFAULT_EXP_FIRST_USE;
94 int ipsec_expire_acquire = IPSEC_DEFAULT_EXPIRE_ACQUIRE;
95 char ipsec_def_enc[20];
96 char ipsec_def_auth[20];
97 char ipsec_def_comp[20];
98 
99 /* values controllable via sysctl */
100 int	ipforwarding = 0;
101 int	ipmforwarding = 0;
102 int	ipmultipath = 0;
103 int	ipsendredirects = 1;
104 int	ip_dosourceroute = 0;
105 int	ip_defttl = IPDEFTTL;
106 int	ip_mtudisc = 1;
107 u_int	ip_mtudisc_timeout = IPMTUDISCTIMEOUT;
108 int	ip_directedbcast = 0;
109 #ifdef DIAGNOSTIC
110 int	ipprintfs = 0;
111 #endif
112 
113 struct rttimer_queue *ip_mtudisc_timeout_q = NULL;
114 
115 int	ipsec_auth_default_level = IPSEC_AUTH_LEVEL_DEFAULT;
116 int	ipsec_esp_trans_default_level = IPSEC_ESP_TRANS_LEVEL_DEFAULT;
117 int	ipsec_esp_network_default_level = IPSEC_ESP_NETWORK_LEVEL_DEFAULT;
118 int	ipsec_ipcomp_default_level = IPSEC_IPCOMP_LEVEL_DEFAULT;
119 
120 /* Keep track of memory used for reassembly */
121 int	ip_maxqueue = 300;
122 int	ip_frags = 0;
123 
124 /* from in_pcb.c */
125 extern int ipport_firstauto;
126 extern int ipport_lastauto;
127 extern int ipport_hifirstauto;
128 extern int ipport_hilastauto;
129 extern struct baddynamicports baddynamicports;
130 extern int la_hold_total;
131 
132 int *ipctl_vars[IPCTL_MAXID] = IPCTL_VARS;
133 
134 extern	struct domain inetdomain;
135 extern	struct protosw inetsw[];
136 u_char	ip_protox[IPPROTO_MAX];
137 int	ipqmaxlen = IFQ_MAXLEN;
138 struct	in_ifaddrhead in_ifaddr;
139 struct	ifqueue ipintrq;
140 
141 struct pool ipqent_pool;
142 struct pool ipq_pool;
143 
144 struct ipstat ipstat;
145 
146 char *
147 inet_ntoa(ina)
148 	struct in_addr ina;
149 {
150 	static char buf[4*sizeof "123"];
151 	unsigned char *ucp = (unsigned char *)&ina;
152 
153 	snprintf(buf, sizeof buf, "%d.%d.%d.%d",
154 	    ucp[0] & 0xff, ucp[1] & 0xff,
155 	    ucp[2] & 0xff, ucp[3] & 0xff);
156 	return (buf);
157 }
158 
159 /*
160  * We need to save the IP options in case a protocol wants to respond
161  * to an incoming packet over the same route if the packet got here
162  * using IP source routing.  This allows connection establishment and
163  * maintenance when the remote end is on a network that is not known
164  * to us.
165  */
166 int	ip_nhops = 0;
167 static	struct ip_srcrt {
168 	struct	in_addr dst;			/* final destination */
169 	char	nop;				/* one NOP to align */
170 	char	srcopt[IPOPT_OFFSET + 1];	/* OPTVAL, OLEN and OFFSET */
171 	struct	in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)];
172 } ip_srcrt;
173 
174 void save_rte(u_char *, struct in_addr);
175 int ip_weadvertise(u_int32_t);
176 
177 /*
178  * IP initialization: fill in IP protocol switch table.
179  * All protocols not implemented in kernel go to raw IP protocol handler.
180  */
181 void
182 ip_init()
183 {
184 	struct protosw *pr;
185 	int i;
186 	const u_int16_t defbaddynamicports_tcp[] = DEFBADDYNAMICPORTS_TCP;
187 	const u_int16_t defbaddynamicports_udp[] = DEFBADDYNAMICPORTS_UDP;
188 
189 	pool_init(&ipqent_pool, sizeof(struct ipqent), 0, 0, 0, "ipqepl",
190 	    NULL);
191 	pool_init(&ipq_pool, sizeof(struct ipq), 0, 0, 0, "ipqpl",
192 	    NULL);
193 
194 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
195 	if (pr == 0)
196 		panic("ip_init");
197 	for (i = 0; i < IPPROTO_MAX; i++)
198 		ip_protox[i] = pr - inetsw;
199 	for (pr = inetdomain.dom_protosw;
200 	    pr < inetdomain.dom_protoswNPROTOSW; pr++)
201 		if (pr->pr_domain->dom_family == PF_INET &&
202 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
203 			ip_protox[pr->pr_protocol] = pr - inetsw;
204 	LIST_INIT(&ipq);
205 	ipintrq.ifq_maxlen = ipqmaxlen;
206 	TAILQ_INIT(&in_ifaddr);
207 	if (ip_mtudisc != 0)
208 		ip_mtudisc_timeout_q =
209 		    rt_timer_queue_create(ip_mtudisc_timeout);
210 
211 	/* Fill in list of ports not to allocate dynamically. */
212 	bzero((void *)&baddynamicports, sizeof(baddynamicports));
213 	for (i = 0; defbaddynamicports_tcp[i] != 0; i++)
214 		DP_SET(baddynamicports.tcp, defbaddynamicports_tcp[i]);
215 	for (i = 0; defbaddynamicports_udp[i] != 0; i++)
216 		DP_SET(baddynamicports.udp, defbaddynamicports_udp[i]);
217 
218 	strlcpy(ipsec_def_enc, IPSEC_DEFAULT_DEF_ENC, sizeof(ipsec_def_enc));
219 	strlcpy(ipsec_def_auth, IPSEC_DEFAULT_DEF_AUTH, sizeof(ipsec_def_auth));
220 	strlcpy(ipsec_def_comp, IPSEC_DEFAULT_DEF_COMP, sizeof(ipsec_def_comp));
221 }
222 
223 struct	sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET };
224 struct	route ipforward_rt;
225 int	ipforward_rtableid;
226 
227 void
228 ipintr()
229 {
230 	struct mbuf *m;
231 	int s;
232 
233 	for (;;) {
234 		/*
235 		 * Get next datagram off input queue and get IP header
236 		 * in first mbuf.
237 		 */
238 		s = splnet();
239 		IF_DEQUEUE(&ipintrq, m);
240 		splx(s);
241 		if (m == NULL)
242 			return;
243 #ifdef	DIAGNOSTIC
244 		if ((m->m_flags & M_PKTHDR) == 0)
245 			panic("ipintr no HDR");
246 #endif
247 		ipv4_input(m);
248 	}
249 }
250 
251 /*
252  * Ip input routine.  Checksum and byte swap header.  If fragmented
253  * try to reassemble.  Process options.  Pass to next level.
254  */
255 void
256 ipv4_input(m)
257 	struct mbuf *m;
258 {
259 	struct ip *ip;
260 	struct ipq *fp;
261 	struct in_ifaddr *ia;
262 	struct ipqent *ipqe;
263 	int hlen, mff, len;
264 	in_addr_t pfrdr = 0;
265 #ifdef IPSEC
266 	int error, s;
267 	struct tdb *tdb;
268 	struct tdb_ident *tdbi;
269 	struct m_tag *mtag;
270 #endif /* IPSEC */
271 
272 	/*
273 	 * If no IP addresses have been set yet but the interfaces
274 	 * are receiving, can't do anything with incoming packets yet.
275 	 */
276 	if (TAILQ_EMPTY(&in_ifaddr))
277 		goto bad;
278 	ipstat.ips_total++;
279 	if (m->m_len < sizeof (struct ip) &&
280 	    (m = m_pullup(m, sizeof (struct ip))) == NULL) {
281 		ipstat.ips_toosmall++;
282 		return;
283 	}
284 	ip = mtod(m, struct ip *);
285 	if (ip->ip_v != IPVERSION) {
286 		ipstat.ips_badvers++;
287 		goto bad;
288 	}
289 	hlen = ip->ip_hl << 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)) == NULL) {
296 			ipstat.ips_badhlen++;
297 			return;
298 		}
299 		ip = mtod(m, struct ip *);
300 	}
301 
302 	/* 127/8 must not appear on wire - RFC1122 */
303 	if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
304 	    (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
305 		if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
306 			ipstat.ips_badaddr++;
307 			goto bad;
308 		}
309 	}
310 
311 	if ((m->m_pkthdr.csum_flags & M_IPV4_CSUM_IN_OK) == 0) {
312 		if (m->m_pkthdr.csum_flags & M_IPV4_CSUM_IN_BAD) {
313 			ipstat.ips_inhwcsum++;
314 			ipstat.ips_badsum++;
315 			goto bad;
316 		}
317 
318 		if (in_cksum(m, hlen) != 0) {
319 			ipstat.ips_badsum++;
320 			goto bad;
321 		}
322 	} else {
323 		m->m_pkthdr.csum_flags &= ~M_IPV4_CSUM_IN_OK;
324 		ipstat.ips_inhwcsum++;
325 	}
326 
327 	/* Retrieve the packet length. */
328 	len = ntohs(ip->ip_len);
329 
330 	/*
331 	 * Convert fields to host representation.
332 	 */
333 	if (len < hlen) {
334 		ipstat.ips_badlen++;
335 		goto bad;
336 	}
337 
338 	/*
339 	 * Check that the amount of data in the buffers
340 	 * is at least as much as the IP header would have us expect.
341 	 * Trim mbufs if longer than we expect.
342 	 * Drop packet if shorter than we expect.
343 	 */
344 	if (m->m_pkthdr.len < len) {
345 		ipstat.ips_tooshort++;
346 		goto bad;
347 	}
348 	if (m->m_pkthdr.len > len) {
349 		if (m->m_len == m->m_pkthdr.len) {
350 			m->m_len = len;
351 			m->m_pkthdr.len = len;
352 		} else
353 			m_adj(m, len - m->m_pkthdr.len);
354 	}
355 
356 #if NCARP > 0
357 	if (m->m_pkthdr.rcvif->if_type == IFT_CARP &&
358 	    ip->ip_p != IPPROTO_ICMP && carp_lsdrop(m, AF_INET,
359 	    &ip->ip_src.s_addr, &ip->ip_dst.s_addr))
360 		goto bad;
361 #endif
362 
363 #if NPF > 0
364 	/*
365 	 * Packet filter
366 	 */
367 	pfrdr = ip->ip_dst.s_addr;
368 	if (pf_test(PF_IN, m->m_pkthdr.rcvif, &m, NULL) != PF_PASS)
369 		goto bad;
370 	if (m == NULL)
371 		return;
372 
373 	ip = mtod(m, struct ip *);
374 	hlen = ip->ip_hl << 2;
375 	pfrdr = (pfrdr != ip->ip_dst.s_addr);
376 #endif
377 
378 	/*
379 	 * Process options and, if not destined for us,
380 	 * ship it on.  ip_dooptions returns 1 when an
381 	 * error was detected (causing an icmp message
382 	 * to be sent and the original packet to be freed).
383 	 */
384 	ip_nhops = 0;		/* for source routed packets */
385 	if (hlen > sizeof (struct ip) && ip_dooptions(m)) {
386 	        return;
387 	}
388 
389 	/*
390 	 * Check our list of addresses, to see if the packet is for us.
391 	 */
392 	if ((ia = in_iawithaddr(ip->ip_dst, m)) != NULL &&
393 	    (ia->ia_ifp->if_flags & IFF_UP))
394 		goto ours;
395 
396 	if (m->m_pkthdr.pf.flags & PF_TAG_DIVERTED)
397 		goto ours;
398 
399 	if (IN_MULTICAST(ip->ip_dst.s_addr)) {
400 		struct in_multi *inm;
401 #ifdef MROUTING
402 		extern struct socket *ip_mrouter;
403 
404 		if (m->m_flags & M_EXT) {
405 			if ((m = m_pullup(m, hlen)) == NULL) {
406 				ipstat.ips_toosmall++;
407 				return;
408 			}
409 			ip = mtod(m, struct ip *);
410 		}
411 		if (ipmforwarding && ip_mrouter) {
412 			/*
413 			 * If we are acting as a multicast router, all
414 			 * incoming multicast packets are passed to the
415 			 * kernel-level multicast forwarding function.
416 			 * The packet is returned (relatively) intact; if
417 			 * ip_mforward() returns a non-zero value, the packet
418 			 * must be discarded, else it may be accepted below.
419 			 *
420 			 * (The IP ident field is put in the same byte order
421 			 * as expected when ip_mforward() is called from
422 			 * ip_output().)
423 			 */
424 			if (ip_mforward(m, m->m_pkthdr.rcvif) != 0) {
425 				ipstat.ips_cantforward++;
426 				m_freem(m);
427 				return;
428 			}
429 
430 			/*
431 			 * The process-level routing daemon needs to receive
432 			 * all multicast IGMP packets, whether or not this
433 			 * host belongs to their destination groups.
434 			 */
435 			if (ip->ip_p == IPPROTO_IGMP)
436 				goto ours;
437 			ipstat.ips_forward++;
438 		}
439 #endif
440 		/*
441 		 * See if we belong to the destination multicast group on the
442 		 * arrival interface.
443 		 */
444 		IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
445 		if (inm == NULL) {
446 			ipstat.ips_notmember++;
447 			if (!IN_LOCAL_GROUP(ip->ip_dst.s_addr))
448 				ipstat.ips_cantforward++;
449 			m_freem(m);
450 			return;
451 		}
452 		goto ours;
453 	}
454 	if (ip->ip_dst.s_addr == INADDR_BROADCAST ||
455 	    ip->ip_dst.s_addr == INADDR_ANY)
456 		goto ours;
457 
458 #if NCARP > 0
459 	if (m->m_pkthdr.rcvif->if_type == IFT_CARP &&
460 	    ip->ip_p == IPPROTO_ICMP && carp_lsdrop(m, AF_INET,
461 	    &ip->ip_src.s_addr, &ip->ip_dst.s_addr))
462 		goto bad;
463 #endif
464 	/*
465 	 * Not for us; forward if possible and desirable.
466 	 */
467 	if (ipforwarding == 0) {
468 		ipstat.ips_cantforward++;
469 		m_freem(m);
470 		return;
471 	}
472 #ifdef IPSEC
473 	if (ipsec_in_use) {
474 	        /*
475 		 * IPsec policy check for forwarded packets. Look at
476 		 * inner-most IPsec SA used.
477 		 */
478 		mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
479                 s = splnet();
480 		if (mtag != NULL) {
481 			tdbi = (struct tdb_ident *)(mtag + 1);
482 			tdb = gettdb(tdbi->spi, &tdbi->dst, tdbi->proto);
483 		} else
484 			tdb = NULL;
485 	        ipsp_spd_lookup(m, AF_INET, hlen, &error,
486 		    IPSP_DIRECTION_IN, tdb, NULL);
487                 splx(s);
488 
489 		/* Error or otherwise drop-packet indication */
490 		if (error) {
491 			ipstat.ips_cantforward++;
492 			m_freem(m);
493 			return;
494 		}
495 
496 		/*
497 		 * Fall through, forward packet. Outbound IPsec policy
498 		 * checking will occur in ip_output().
499 		 */
500 	}
501 #endif /* IPSEC */
502 
503 	ip_forward(m, pfrdr);
504 	return;
505 
506 ours:
507 	/*
508 	 * If offset or IP_MF are set, must reassemble.
509 	 * Otherwise, nothing need be done.
510 	 * (We could look in the reassembly queue to see
511 	 * if the packet was previously fragmented,
512 	 * but it's not worth the time; just let them time out.)
513 	 */
514 	if (ip->ip_off &~ htons(IP_DF | IP_RF)) {
515 		if (m->m_flags & M_EXT) {		/* XXX */
516 			if ((m = m_pullup(m, hlen)) == NULL) {
517 				ipstat.ips_toosmall++;
518 				return;
519 			}
520 			ip = mtod(m, struct ip *);
521 		}
522 
523 		/*
524 		 * Look for queue of fragments
525 		 * of this datagram.
526 		 */
527 		LIST_FOREACH(fp, &ipq, ipq_q)
528 			if (ip->ip_id == fp->ipq_id &&
529 			    ip->ip_src.s_addr == fp->ipq_src.s_addr &&
530 			    ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
531 			    ip->ip_p == fp->ipq_p)
532 				goto found;
533 		fp = 0;
534 found:
535 
536 		/*
537 		 * Adjust ip_len to not reflect header,
538 		 * set ipqe_mff if more fragments are expected,
539 		 * convert offset of this to bytes.
540 		 */
541 		ip->ip_len = htons(ntohs(ip->ip_len) - hlen);
542 		mff = (ip->ip_off & htons(IP_MF)) != 0;
543 		if (mff) {
544 			/*
545 			 * Make sure that fragments have a data length
546 			 * that's a non-zero multiple of 8 bytes.
547 			 */
548 			if (ntohs(ip->ip_len) == 0 ||
549 			    (ntohs(ip->ip_len) & 0x7) != 0) {
550 				ipstat.ips_badfrags++;
551 				goto bad;
552 			}
553 		}
554 		ip->ip_off = htons(ntohs(ip->ip_off) << 3);
555 
556 		/*
557 		 * If datagram marked as having more fragments
558 		 * or if this is not the first fragment,
559 		 * attempt reassembly; if it succeeds, proceed.
560 		 */
561 		if (mff || ip->ip_off) {
562 			ipstat.ips_fragments++;
563 			if (ip_frags + 1 > ip_maxqueue) {
564 				ip_flush();
565 				ipstat.ips_rcvmemdrop++;
566 				goto bad;
567 			}
568 
569 			ipqe = pool_get(&ipqent_pool, PR_NOWAIT);
570 			if (ipqe == NULL) {
571 				ipstat.ips_rcvmemdrop++;
572 				goto bad;
573 			}
574 			ip_frags++;
575 			ipqe->ipqe_mff = mff;
576 			ipqe->ipqe_m = m;
577 			ipqe->ipqe_ip = ip;
578 			m = ip_reass(ipqe, fp);
579 			if (m == 0) {
580 				return;
581 			}
582 			ipstat.ips_reassembled++;
583 			ip = mtod(m, struct ip *);
584 			hlen = ip->ip_hl << 2;
585 			ip->ip_len = htons(ntohs(ip->ip_len) + hlen);
586 		} else
587 			if (fp)
588 				ip_freef(fp);
589 	}
590 
591 #ifdef IPSEC
592 	if (!ipsec_in_use)
593 		goto skipipsec;
594 
595         /*
596          * If it's a protected packet for us, skip the policy check.
597          * That's because we really only care about the properties of
598          * the protected packet, and not the intermediate versions.
599          * While this is not the most paranoid setting, it allows
600          * some flexibility in handling nested tunnels (in setting up
601 	 * the policies).
602          */
603         if ((ip->ip_p == IPPROTO_ESP) || (ip->ip_p == IPPROTO_AH) ||
604 	    (ip->ip_p == IPPROTO_IPCOMP))
605           goto skipipsec;
606 
607 	/*
608 	 * If the protected packet was tunneled, then we need to
609 	 * verify the protected packet's information, not the
610 	 * external headers. Thus, skip the policy lookup for the
611 	 * external packet, and keep the IPsec information linked on
612 	 * the packet header (the encapsulation routines know how
613 	 * to deal with that).
614 	 */
615 	if ((ip->ip_p == IPPROTO_IPIP) || (ip->ip_p == IPPROTO_IPV6))
616 	  goto skipipsec;
617 
618 	/*
619 	 * If the protected packet is TCP or UDP, we'll do the
620 	 * policy check in the respective input routine, so we can
621 	 * check for bypass sockets.
622 	 */
623 	if ((ip->ip_p == IPPROTO_TCP) || (ip->ip_p == IPPROTO_UDP))
624 	  goto skipipsec;
625 
626 	/*
627 	 * IPsec policy check for local-delivery packets. Look at the
628 	 * inner-most SA that protected the packet. This is in fact
629 	 * a bit too restrictive (it could end up causing packets to
630 	 * be dropped that semantically follow the policy, e.g., in
631 	 * certain SA-bundle configurations); but the alternative is
632 	 * very complicated (and requires keeping track of what
633 	 * kinds of tunneling headers have been seen in-between the
634 	 * IPsec headers), and I don't think we lose much functionality
635 	 * that's needed in the real world (who uses bundles anyway ?).
636 	 */
637 	mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
638         s = splnet();
639 	if (mtag) {
640 		tdbi = (struct tdb_ident *)(mtag + 1);
641 	        tdb = gettdb(tdbi->spi, &tdbi->dst, tdbi->proto);
642 	} else
643 		tdb = NULL;
644 	ipsp_spd_lookup(m, AF_INET, hlen, &error, IPSP_DIRECTION_IN,
645 	    tdb, NULL);
646         splx(s);
647 
648 	/* Error or otherwise drop-packet indication. */
649 	if (error) {
650 	        ipstat.ips_cantforward++;
651 		m_freem(m);
652 		return;
653 	}
654 
655  skipipsec:
656 	/* Otherwise, just fall through and deliver the packet */
657 #endif /* IPSEC */
658 
659 	/*
660 	 * Switch out to protocol's input routine.
661 	 */
662 	ipstat.ips_delivered++;
663 	(*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen, NULL, 0);
664 	return;
665 bad:
666 	m_freem(m);
667 }
668 
669 struct in_ifaddr *
670 in_iawithaddr(ina, m)
671 	struct in_addr ina;
672 	struct mbuf *m;
673 {
674 	struct in_ifaddr *ia;
675 
676 	TAILQ_FOREACH(ia, &in_ifaddr, ia_list) {
677 		if ((ina.s_addr == ia->ia_addr.sin_addr.s_addr) ||
678 		    ((ia->ia_ifp->if_flags & (IFF_LOOPBACK|IFF_LINK1)) ==
679 			(IFF_LOOPBACK|IFF_LINK1) &&
680 		     ia->ia_subnet == (ina.s_addr & ia->ia_subnetmask)))
681 			return ia;
682 		if (((ip_directedbcast == 0) || (m && ip_directedbcast &&
683 		    ia->ia_ifp == m->m_pkthdr.rcvif)) &&
684 		    (ia->ia_ifp->if_flags & IFF_BROADCAST)) {
685 			if (ina.s_addr == ia->ia_broadaddr.sin_addr.s_addr ||
686 			    ina.s_addr == ia->ia_netbroadcast.s_addr ||
687 			    /*
688 			     * Look for all-0's host part (old broadcast addr),
689 			     * either for subnet or net.
690 			     */
691 			    ina.s_addr == ia->ia_subnet ||
692 			    ina.s_addr == ia->ia_net) {
693 				/* Make sure M_BCAST is set */
694 				if (m)
695 					m->m_flags |= M_BCAST;
696 				return ia;
697 			    }
698 		}
699 	}
700 
701 	return NULL;
702 }
703 
704 /*
705  * Take incoming datagram fragment and try to
706  * reassemble it into whole datagram.  If a chain for
707  * reassembly of this datagram already exists, then it
708  * is given as fp; otherwise have to make a chain.
709  */
710 struct mbuf *
711 ip_reass(ipqe, fp)
712 	struct ipqent *ipqe;
713 	struct ipq *fp;
714 {
715 	struct mbuf *m = ipqe->ipqe_m;
716 	struct ipqent *nq, *p, *q;
717 	struct ip *ip;
718 	struct mbuf *t;
719 	int hlen = ipqe->ipqe_ip->ip_hl << 2;
720 	int i, next;
721 	u_int8_t ecn, ecn0;
722 
723 	/*
724 	 * Presence of header sizes in mbufs
725 	 * would confuse code below.
726 	 */
727 	m->m_data += hlen;
728 	m->m_len -= hlen;
729 
730 	/*
731 	 * If first fragment to arrive, create a reassembly queue.
732 	 */
733 	if (fp == NULL) {
734 		fp = pool_get(&ipq_pool, PR_NOWAIT);
735 		if (fp == NULL)
736 			goto dropfrag;
737 		LIST_INSERT_HEAD(&ipq, fp, ipq_q);
738 		fp->ipq_ttl = IPFRAGTTL;
739 		fp->ipq_p = ipqe->ipqe_ip->ip_p;
740 		fp->ipq_id = ipqe->ipqe_ip->ip_id;
741 		LIST_INIT(&fp->ipq_fragq);
742 		fp->ipq_src = ipqe->ipqe_ip->ip_src;
743 		fp->ipq_dst = ipqe->ipqe_ip->ip_dst;
744 		p = NULL;
745 		goto insert;
746 	}
747 
748 	/*
749 	 * Handle ECN by comparing this segment with the first one;
750 	 * if CE is set, do not lose CE.
751 	 * drop if CE and not-ECT are mixed for the same packet.
752 	 */
753 	ecn = ipqe->ipqe_ip->ip_tos & IPTOS_ECN_MASK;
754 	ecn0 = LIST_FIRST(&fp->ipq_fragq)->ipqe_ip->ip_tos & IPTOS_ECN_MASK;
755 	if (ecn == IPTOS_ECN_CE) {
756 		if (ecn0 == IPTOS_ECN_NOTECT)
757 			goto dropfrag;
758 		if (ecn0 != IPTOS_ECN_CE)
759 			LIST_FIRST(&fp->ipq_fragq)->ipqe_ip->ip_tos |= IPTOS_ECN_CE;
760 	}
761 	if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT)
762 		goto dropfrag;
763 
764 	/*
765 	 * Find a segment which begins after this one does.
766 	 */
767 	for (p = NULL, q = LIST_FIRST(&fp->ipq_fragq);
768 	    q != LIST_END(&fp->ipq_fragq); p = q, q = LIST_NEXT(q, ipqe_q))
769 		if (ntohs(q->ipqe_ip->ip_off) > ntohs(ipqe->ipqe_ip->ip_off))
770 			break;
771 
772 	/*
773 	 * If there is a preceding segment, it may provide some of
774 	 * our data already.  If so, drop the data from the incoming
775 	 * segment.  If it provides all of our data, drop us.
776 	 */
777 	if (p != NULL) {
778 		i = ntohs(p->ipqe_ip->ip_off) + ntohs(p->ipqe_ip->ip_len) -
779 		    ntohs(ipqe->ipqe_ip->ip_off);
780 		if (i > 0) {
781 			if (i >= ntohs(ipqe->ipqe_ip->ip_len))
782 				goto dropfrag;
783 			m_adj(ipqe->ipqe_m, i);
784 			ipqe->ipqe_ip->ip_off =
785 			    htons(ntohs(ipqe->ipqe_ip->ip_off) + i);
786 			ipqe->ipqe_ip->ip_len =
787 			    htons(ntohs(ipqe->ipqe_ip->ip_len) - i);
788 		}
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 &&
796 	    ntohs(ipqe->ipqe_ip->ip_off) + ntohs(ipqe->ipqe_ip->ip_len) >
797 	    ntohs(q->ipqe_ip->ip_off); q = nq) {
798 		i = (ntohs(ipqe->ipqe_ip->ip_off) +
799 		    ntohs(ipqe->ipqe_ip->ip_len)) - ntohs(q->ipqe_ip->ip_off);
800 		if (i < ntohs(q->ipqe_ip->ip_len)) {
801 			q->ipqe_ip->ip_len =
802 			    htons(ntohs(q->ipqe_ip->ip_len) - i);
803 			q->ipqe_ip->ip_off =
804 			    htons(ntohs(q->ipqe_ip->ip_off) + i);
805 			m_adj(q->ipqe_m, i);
806 			break;
807 		}
808 		nq = LIST_NEXT(q, ipqe_q);
809 		m_freem(q->ipqe_m);
810 		LIST_REMOVE(q, ipqe_q);
811 		pool_put(&ipqent_pool, q);
812 		ip_frags--;
813 	}
814 
815 insert:
816 	/*
817 	 * Stick new segment in its place;
818 	 * check for complete reassembly.
819 	 */
820 	if (p == NULL) {
821 		LIST_INSERT_HEAD(&fp->ipq_fragq, ipqe, ipqe_q);
822 	} else {
823 		LIST_INSERT_AFTER(p, ipqe, ipqe_q);
824 	}
825 	next = 0;
826 	for (p = NULL, q = LIST_FIRST(&fp->ipq_fragq);
827 	    q != LIST_END(&fp->ipq_fragq); p = q, q = LIST_NEXT(q, ipqe_q)) {
828 		if (ntohs(q->ipqe_ip->ip_off) != next)
829 			return (0);
830 		next += ntohs(q->ipqe_ip->ip_len);
831 	}
832 	if (p->ipqe_mff)
833 		return (0);
834 
835 	/*
836 	 * Reassembly is complete.  Check for a bogus message size and
837 	 * concatenate fragments.
838 	 */
839 	q = LIST_FIRST(&fp->ipq_fragq);
840 	ip = q->ipqe_ip;
841 	if ((next + (ip->ip_hl << 2)) > IP_MAXPACKET) {
842 		ipstat.ips_toolong++;
843 		ip_freef(fp);
844 		return (0);
845 	}
846 	m = q->ipqe_m;
847 	t = m->m_next;
848 	m->m_next = 0;
849 	m_cat(m, t);
850 	nq = LIST_NEXT(q, ipqe_q);
851 	pool_put(&ipqent_pool, q);
852 	ip_frags--;
853 	for (q = nq; q != NULL; q = nq) {
854 		t = q->ipqe_m;
855 		nq = LIST_NEXT(q, ipqe_q);
856 		pool_put(&ipqent_pool, q);
857 		ip_frags--;
858 		m_cat(m, t);
859 	}
860 
861 	/*
862 	 * Create header for new ip packet by
863 	 * modifying header of first packet;
864 	 * dequeue and discard fragment reassembly header.
865 	 * Make header visible.
866 	 */
867 	ip->ip_len = htons(next);
868 	ip->ip_src = fp->ipq_src;
869 	ip->ip_dst = fp->ipq_dst;
870 	LIST_REMOVE(fp, ipq_q);
871 	pool_put(&ipq_pool, fp);
872 	m->m_len += (ip->ip_hl << 2);
873 	m->m_data -= (ip->ip_hl << 2);
874 	/* some debugging cruft by sklower, below, will go away soon */
875 	if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */
876 		int plen = 0;
877 		for (t = m; t; t = t->m_next)
878 			plen += t->m_len;
879 		m->m_pkthdr.len = plen;
880 	}
881 	return (m);
882 
883 dropfrag:
884 	ipstat.ips_fragdropped++;
885 	m_freem(m);
886 	pool_put(&ipqent_pool, ipqe);
887 	ip_frags--;
888 	return (0);
889 }
890 
891 /*
892  * Free a fragment reassembly header and all
893  * associated datagrams.
894  */
895 void
896 ip_freef(fp)
897 	struct ipq *fp;
898 {
899 	struct ipqent *q, *p;
900 
901 	for (q = LIST_FIRST(&fp->ipq_fragq); q != LIST_END(&fp->ipq_fragq);
902 	    q = p) {
903 		p = LIST_NEXT(q, ipqe_q);
904 		m_freem(q->ipqe_m);
905 		LIST_REMOVE(q, ipqe_q);
906 		pool_put(&ipqent_pool, q);
907 		ip_frags--;
908 	}
909 	LIST_REMOVE(fp, ipq_q);
910 	pool_put(&ipq_pool, fp);
911 }
912 
913 /*
914  * IP timer processing;
915  * if a timer expires on a reassembly queue, discard it.
916  * clear the forwarding cache, there might be a better route.
917  */
918 void
919 ip_slowtimo()
920 {
921 	struct ipq *fp, *nfp;
922 	int s = splsoftnet();
923 
924 	for (fp = LIST_FIRST(&ipq); fp != LIST_END(&ipq); fp = nfp) {
925 		nfp = LIST_NEXT(fp, ipq_q);
926 		if (--fp->ipq_ttl == 0) {
927 			ipstat.ips_fragtimeout++;
928 			ip_freef(fp);
929 		}
930 	}
931 	if (ipforward_rt.ro_rt) {
932 		RTFREE(ipforward_rt.ro_rt);
933 		ipforward_rt.ro_rt = 0;
934 	}
935 	splx(s);
936 }
937 
938 /*
939  * Drain off all datagram fragments.
940  */
941 void
942 ip_drain()
943 {
944 
945 	while (!LIST_EMPTY(&ipq)) {
946 		ipstat.ips_fragdropped++;
947 		ip_freef(LIST_FIRST(&ipq));
948 	}
949 }
950 
951 /*
952  * Flush a bunch of datagram fragments, till we are down to 75%.
953  */
954 void
955 ip_flush()
956 {
957 	int max = 50;
958 
959 	/* ipq already locked */
960 	while (!LIST_EMPTY(&ipq) && ip_frags > ip_maxqueue * 3 / 4 && --max) {
961 		ipstat.ips_fragdropped++;
962 		ip_freef(LIST_FIRST(&ipq));
963 	}
964 }
965 
966 /*
967  * Do option processing on a datagram,
968  * possibly discarding it if bad options are encountered,
969  * or forwarding it if source-routed.
970  * Returns 1 if packet has been forwarded/freed,
971  * 0 if the packet should be processed further.
972  */
973 int
974 ip_dooptions(m)
975 	struct mbuf *m;
976 {
977 	struct ip *ip = mtod(m, struct ip *);
978 	u_char *cp;
979 	struct ip_timestamp ipt;
980 	struct in_ifaddr *ia;
981 	int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0;
982 	struct in_addr sin, dst;
983 	n_time ntime;
984 
985 	dst = ip->ip_dst;
986 	cp = (u_char *)(ip + 1);
987 	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
988 
989 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
990 		opt = cp[IPOPT_OPTVAL];
991 		if (opt == IPOPT_EOL)
992 			break;
993 		if (opt == IPOPT_NOP)
994 			optlen = 1;
995 		else {
996 			if (cnt < IPOPT_OLEN + sizeof(*cp)) {
997 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
998 				goto bad;
999 			}
1000 			optlen = cp[IPOPT_OLEN];
1001 			if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
1002 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1003 				goto bad;
1004 			}
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 (!ip_dosourceroute) {
1024 				type = ICMP_UNREACH;
1025 				code = ICMP_UNREACH_SRCFAIL;
1026 				goto bad;
1027 			}
1028 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1029 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1030 				goto bad;
1031 			}
1032 			ipaddr.sin_addr = ip->ip_dst;
1033 			ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr)));
1034 			if (ia == 0) {
1035 				if (opt == IPOPT_SSRR) {
1036 					type = ICMP_UNREACH;
1037 					code = ICMP_UNREACH_SRCFAIL;
1038 					goto bad;
1039 				}
1040 				/*
1041 				 * Loose routing, and not at next destination
1042 				 * yet; nothing to do except forward.
1043 				 */
1044 				break;
1045 			}
1046 			off--;			/* 0 origin */
1047 			if ((off + sizeof(struct in_addr)) > optlen) {
1048 				/*
1049 				 * End of source route.  Should be for us.
1050 				 */
1051 				save_rte(cp, ip->ip_src);
1052 				break;
1053 			}
1054 
1055 			/*
1056 			 * locate outgoing interface
1057 			 */
1058 			bcopy((caddr_t)(cp + off), (caddr_t)&ipaddr.sin_addr,
1059 			    sizeof(ipaddr.sin_addr));
1060 			if (opt == IPOPT_SSRR) {
1061 #define	INA	struct in_ifaddr *
1062 #define	SA	struct sockaddr *
1063 			    if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == 0)
1064 				ia = (INA)ifa_ifwithnet((SA)&ipaddr);
1065 			} else
1066 				ia = ip_rtaddr(ipaddr.sin_addr);
1067 			if (ia == 0) {
1068 				type = ICMP_UNREACH;
1069 				code = ICMP_UNREACH_SRCFAIL;
1070 				goto bad;
1071 			}
1072 			ip->ip_dst = ipaddr.sin_addr;
1073 			bcopy((caddr_t)&ia->ia_addr.sin_addr,
1074 			    (caddr_t)(cp + off), sizeof(struct in_addr));
1075 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1076 			/*
1077 			 * Let ip_intr's mcast routing check handle mcast pkts
1078 			 */
1079 			forward = !IN_MULTICAST(ip->ip_dst.s_addr);
1080 			break;
1081 
1082 		case IPOPT_RR:
1083 			if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1084 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1085 				goto bad;
1086 			}
1087 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1088 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1089 				goto bad;
1090 			}
1091 
1092 			/*
1093 			 * If no space remains, ignore.
1094 			 */
1095 			off--;			/* 0 origin */
1096 			if ((off + sizeof(struct in_addr)) > optlen)
1097 				break;
1098 			bcopy((caddr_t)(&ip->ip_dst), (caddr_t)&ipaddr.sin_addr,
1099 			    sizeof(ipaddr.sin_addr));
1100 			/*
1101 			 * locate outgoing interface; if we're the destination,
1102 			 * use the incoming interface (should be same).
1103 			 */
1104 			if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == 0 &&
1105 			    (ia = ip_rtaddr(ipaddr.sin_addr)) == 0) {
1106 				type = ICMP_UNREACH;
1107 				code = ICMP_UNREACH_HOST;
1108 				goto bad;
1109 			}
1110 			bcopy((caddr_t)&ia->ia_addr.sin_addr,
1111 			    (caddr_t)(cp + off), sizeof(struct in_addr));
1112 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1113 			break;
1114 
1115 		case IPOPT_TS:
1116 			code = cp - (u_char *)ip;
1117 			if (optlen < sizeof(struct ip_timestamp))
1118 				goto bad;
1119 			bcopy(cp, &ipt, sizeof(struct ip_timestamp));
1120 			if (ipt.ipt_ptr < 5 || ipt.ipt_len < 5)
1121 				goto bad;
1122 			if (ipt.ipt_ptr - 1 + sizeof(n_time) > ipt.ipt_len) {
1123 				if (++ipt.ipt_oflw == 0)
1124 					goto bad;
1125 				break;
1126 			}
1127 			bcopy(cp + ipt.ipt_ptr - 1, &sin, sizeof sin);
1128 			switch (ipt.ipt_flg) {
1129 
1130 			case IPOPT_TS_TSONLY:
1131 				break;
1132 
1133 			case IPOPT_TS_TSANDADDR:
1134 				if (ipt.ipt_ptr - 1 + sizeof(n_time) +
1135 				    sizeof(struct in_addr) > ipt.ipt_len)
1136 					goto bad;
1137 				ipaddr.sin_addr = dst;
1138 				ia = (INA)ifaof_ifpforaddr((SA)&ipaddr,
1139 							    m->m_pkthdr.rcvif);
1140 				if (ia == 0)
1141 					continue;
1142 				bcopy((caddr_t)&ia->ia_addr.sin_addr,
1143 				    (caddr_t)&sin, sizeof(struct in_addr));
1144 				ipt.ipt_ptr += sizeof(struct in_addr);
1145 				break;
1146 
1147 			case IPOPT_TS_PRESPEC:
1148 				if (ipt.ipt_ptr - 1 + sizeof(n_time) +
1149 				    sizeof(struct in_addr) > ipt.ipt_len)
1150 					goto bad;
1151 				bcopy((caddr_t)&sin, (caddr_t)&ipaddr.sin_addr,
1152 				    sizeof(struct in_addr));
1153 				if (ifa_ifwithaddr((SA)&ipaddr) == 0)
1154 					continue;
1155 				ipt.ipt_ptr += sizeof(struct in_addr);
1156 				break;
1157 
1158 			default:
1159 				/* XXX can't take &ipt->ipt_flg */
1160 				code = (u_char *)&ipt.ipt_ptr -
1161 				    (u_char *)ip + 1;
1162 				goto bad;
1163 			}
1164 			ntime = iptime();
1165 			bcopy((caddr_t)&ntime, (caddr_t)cp + ipt.ipt_ptr - 1,
1166 			    sizeof(n_time));
1167 			ipt.ipt_ptr += sizeof(n_time);
1168 		}
1169 	}
1170 	if (forward && ipforwarding) {
1171 		ip_forward(m, 1);
1172 		return (1);
1173 	}
1174 	return (0);
1175 bad:
1176 	icmp_error(m, type, code, 0, 0);
1177 	ipstat.ips_badoptions++;
1178 	return (1);
1179 }
1180 
1181 /*
1182  * Given address of next destination (final or next hop),
1183  * return internet address info of interface to be used to get there.
1184  */
1185 struct in_ifaddr *
1186 ip_rtaddr(dst)
1187 	 struct in_addr dst;
1188 {
1189 	struct sockaddr_in *sin;
1190 
1191 	sin = satosin(&ipforward_rt.ro_dst);
1192 
1193 	if (ipforward_rt.ro_rt == 0 || dst.s_addr != sin->sin_addr.s_addr) {
1194 		if (ipforward_rt.ro_rt) {
1195 			RTFREE(ipforward_rt.ro_rt);
1196 			ipforward_rt.ro_rt = 0;
1197 		}
1198 		sin->sin_family = AF_INET;
1199 		sin->sin_len = sizeof(*sin);
1200 		sin->sin_addr = dst;
1201 
1202 		rtalloc(&ipforward_rt);
1203 	}
1204 	if (ipforward_rt.ro_rt == 0)
1205 		return ((struct in_ifaddr *)0);
1206 	return (ifatoia(ipforward_rt.ro_rt->rt_ifa));
1207 }
1208 
1209 /*
1210  * Save incoming source route for use in replies,
1211  * to be picked up later by ip_srcroute if the receiver is interested.
1212  */
1213 void
1214 save_rte(option, dst)
1215 	u_char *option;
1216 	struct in_addr dst;
1217 {
1218 	unsigned olen;
1219 
1220 	olen = option[IPOPT_OLEN];
1221 #ifdef DIAGNOSTIC
1222 	if (ipprintfs)
1223 		printf("save_rte: olen %d\n", olen);
1224 #endif /* 0 */
1225 	if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst)))
1226 		return;
1227 	bcopy((caddr_t)option, (caddr_t)ip_srcrt.srcopt, olen);
1228 	ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1229 	ip_srcrt.dst = dst;
1230 }
1231 
1232 /*
1233  * Check whether we do proxy ARP for this address and we point to ourselves.
1234  * Code shamelessly copied from arplookup().
1235  */
1236 int
1237 ip_weadvertise(addr)
1238 	u_int32_t addr;
1239 {
1240 	struct rtentry *rt;
1241 	struct ifnet *ifp;
1242 	struct ifaddr *ifa;
1243 	struct sockaddr_inarp sin;
1244 
1245 	sin.sin_len = sizeof(sin);
1246 	sin.sin_family = AF_INET;
1247 	sin.sin_addr.s_addr = addr;
1248 	sin.sin_other = SIN_PROXY;
1249 	rt = rtalloc1(sintosa(&sin), 0, 0);	/* XXX other tables? */
1250 	if (rt == 0)
1251 		return 0;
1252 
1253 	if ((rt->rt_flags & RTF_GATEWAY) || (rt->rt_flags & RTF_LLINFO) == 0 ||
1254 	    rt->rt_gateway->sa_family != AF_LINK) {
1255 		RTFREE(rt);
1256 		return 0;
1257 	}
1258 
1259 	TAILQ_FOREACH(ifp, &ifnet, if_list)
1260 		TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1261 			if (ifa->ifa_addr->sa_family != rt->rt_gateway->sa_family)
1262 				continue;
1263 
1264 			if (!bcmp(LLADDR((struct sockaddr_dl *)ifa->ifa_addr),
1265 			    LLADDR((struct sockaddr_dl *)rt->rt_gateway),
1266 			    ETHER_ADDR_LEN)) {
1267 				RTFREE(rt);
1268 				return 1;
1269 			}
1270 		}
1271 
1272 	RTFREE(rt);
1273 	return 0;
1274 }
1275 
1276 /*
1277  * Retrieve incoming source route for use in replies,
1278  * in the same form used by setsockopt.
1279  * The first hop is placed before the options, will be removed later.
1280  */
1281 struct mbuf *
1282 ip_srcroute()
1283 {
1284 	struct in_addr *p, *q;
1285 	struct mbuf *m;
1286 
1287 	if (ip_nhops == 0)
1288 		return ((struct mbuf *)0);
1289 	m = m_get(M_DONTWAIT, MT_SOOPTS);
1290 	if (m == 0)
1291 		return ((struct mbuf *)0);
1292 
1293 #define OPTSIZ	(sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt))
1294 
1295 	/* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
1296 	m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) +
1297 	    OPTSIZ;
1298 #ifdef DIAGNOSTIC
1299 	if (ipprintfs)
1300 		printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len);
1301 #endif
1302 
1303 	/*
1304 	 * First save first hop for return route
1305 	 */
1306 	p = &ip_srcrt.route[ip_nhops - 1];
1307 	*(mtod(m, struct in_addr *)) = *p--;
1308 #ifdef DIAGNOSTIC
1309 	if (ipprintfs)
1310 		printf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr));
1311 #endif
1312 
1313 	/*
1314 	 * Copy option fields and padding (nop) to mbuf.
1315 	 */
1316 	ip_srcrt.nop = IPOPT_NOP;
1317 	ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
1318 	bcopy((caddr_t)&ip_srcrt.nop,
1319 	    mtod(m, caddr_t) + sizeof(struct in_addr), OPTSIZ);
1320 	q = (struct in_addr *)(mtod(m, caddr_t) +
1321 	    sizeof(struct in_addr) + OPTSIZ);
1322 #undef OPTSIZ
1323 	/*
1324 	 * Record return path as an IP source route,
1325 	 * reversing the path (pointers are now aligned).
1326 	 */
1327 	while (p >= ip_srcrt.route) {
1328 #ifdef DIAGNOSTIC
1329 		if (ipprintfs)
1330 			printf(" %x", ntohl(q->s_addr));
1331 #endif
1332 		*q++ = *p--;
1333 	}
1334 	/*
1335 	 * Last hop goes to final destination.
1336 	 */
1337 	*q = ip_srcrt.dst;
1338 #ifdef DIAGNOSTIC
1339 	if (ipprintfs)
1340 		printf(" %x\n", ntohl(q->s_addr));
1341 #endif
1342 	return (m);
1343 }
1344 
1345 /*
1346  * Strip out IP options, at higher
1347  * level protocol in the kernel.
1348  * Second argument is buffer to which options
1349  * will be moved, and return value is their length.
1350  * XXX should be deleted; last arg currently ignored.
1351  */
1352 void
1353 ip_stripoptions(m, mopt)
1354 	struct mbuf *m;
1355 	struct mbuf *mopt;
1356 {
1357 	int i;
1358 	struct ip *ip = mtod(m, struct ip *);
1359 	caddr_t opts;
1360 	int olen;
1361 
1362 	olen = (ip->ip_hl<<2) - sizeof (struct ip);
1363 	opts = (caddr_t)(ip + 1);
1364 	i = m->m_len - (sizeof (struct ip) + olen);
1365 	bcopy(opts  + olen, opts, (unsigned)i);
1366 	m->m_len -= olen;
1367 	if (m->m_flags & M_PKTHDR)
1368 		m->m_pkthdr.len -= olen;
1369 	ip->ip_hl = sizeof(struct ip) >> 2;
1370 }
1371 
1372 int inetctlerrmap[PRC_NCMDS] = {
1373 	0,		0,		0,		0,
1374 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1375 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1376 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1377 	0,		0,		0,		0,
1378 	ENOPROTOOPT
1379 };
1380 
1381 /*
1382  * Forward a packet.  If some error occurs return the sender
1383  * an icmp packet.  Note we can't always generate a meaningful
1384  * icmp message because icmp doesn't have a large enough repertoire
1385  * of codes and types.
1386  *
1387  * If not forwarding, just drop the packet.  This could be confusing
1388  * if ipforwarding was zero but some routing protocol was advancing
1389  * us as a gateway to somewhere.  However, we must let the routing
1390  * protocol deal with that.
1391  *
1392  * The srcrt parameter indicates whether the packet is being forwarded
1393  * via a source route.
1394  */
1395 void
1396 ip_forward(m, srcrt)
1397 	struct mbuf *m;
1398 	int srcrt;
1399 {
1400 	struct ip *ip = mtod(m, struct ip *);
1401 	struct sockaddr_in *sin;
1402 	struct rtentry *rt;
1403 	int error, type = 0, code = 0, destmtu = 0, rtableid = 0;
1404 	struct mbuf *mcopy;
1405 	n_long dest;
1406 
1407 	dest = 0;
1408 #ifdef DIAGNOSTIC
1409 	if (ipprintfs)
1410 		printf("forward: src %x dst %x ttl %x\n", ip->ip_src.s_addr,
1411 		    ip->ip_dst.s_addr, ip->ip_ttl);
1412 #endif
1413 	if (m->m_flags & M_BCAST || in_canforward(ip->ip_dst) == 0) {
1414 		ipstat.ips_cantforward++;
1415 		m_freem(m);
1416 		return;
1417 	}
1418 	if (ip->ip_ttl <= IPTTLDEC) {
1419 		icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0);
1420 		return;
1421 	}
1422 
1423 #if NPF > 0
1424 	rtableid = m->m_pkthdr.pf.rtableid;
1425 #endif
1426 
1427 	sin = satosin(&ipforward_rt.ro_dst);
1428 	if ((rt = ipforward_rt.ro_rt) == 0 ||
1429 	    ip->ip_dst.s_addr != sin->sin_addr.s_addr ||
1430 	    rtableid != ipforward_rtableid) {
1431 		if (ipforward_rt.ro_rt) {
1432 			RTFREE(ipforward_rt.ro_rt);
1433 			ipforward_rt.ro_rt = 0;
1434 		}
1435 		sin->sin_family = AF_INET;
1436 		sin->sin_len = sizeof(*sin);
1437 		sin->sin_addr = ip->ip_dst;
1438 
1439 		rtalloc_mpath(&ipforward_rt, &ip->ip_src.s_addr, rtableid);
1440 		if (ipforward_rt.ro_rt == 0) {
1441 			icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0);
1442 			return;
1443 		}
1444 		ipforward_rtableid = rtableid;
1445 		rt = ipforward_rt.ro_rt;
1446 	}
1447 
1448 	/*
1449 	 * Save at most 68 bytes of the packet in case
1450 	 * we need to generate an ICMP message to the src.
1451 	 * Pullup to avoid sharing mbuf cluster between m and mcopy.
1452 	 */
1453 	mcopy = m_copym(m, 0, min(ntohs(ip->ip_len), 68), M_DONTWAIT);
1454 	if (mcopy)
1455 		mcopy = m_pullup(mcopy, min(ntohs(ip->ip_len), 68));
1456 
1457 	ip->ip_ttl -= IPTTLDEC;
1458 
1459 	/*
1460 	 * If forwarding packet using same interface that it came in on,
1461 	 * perhaps should send a redirect to sender to shortcut a hop.
1462 	 * Only send redirect if source is sending directly to us,
1463 	 * and if packet was not source routed (or has any options).
1464 	 * Also, don't send redirect if forwarding using a default route
1465 	 * or a route modified by a redirect.
1466 	 * Don't send redirect if we advertise destination's arp address
1467 	 * as ours (proxy arp).
1468 	 */
1469 	if (rt->rt_ifp == m->m_pkthdr.rcvif &&
1470 	    (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1471 	    satosin(rt_key(rt))->sin_addr.s_addr != 0 &&
1472 	    ipsendredirects && !srcrt &&
1473 	    !ip_weadvertise(satosin(rt_key(rt))->sin_addr.s_addr)) {
1474 		if (rt->rt_ifa &&
1475 		    (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) ==
1476 		    ifatoia(rt->rt_ifa)->ia_subnet) {
1477 		    if (rt->rt_flags & RTF_GATEWAY)
1478 			dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
1479 		    else
1480 			dest = ip->ip_dst.s_addr;
1481 		    /* Router requirements says to only send host redirects */
1482 		    type = ICMP_REDIRECT;
1483 		    code = ICMP_REDIRECT_HOST;
1484 #ifdef DIAGNOSTIC
1485 		    if (ipprintfs)
1486 			printf("redirect (%d) to %x\n", code, (u_int32_t)dest);
1487 #endif
1488 		}
1489 	}
1490 
1491 	error = ip_output(m, (struct mbuf *)NULL, &ipforward_rt,
1492 	    (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)),
1493 	    (void *)NULL, (void *)NULL);
1494 	if (error)
1495 		ipstat.ips_cantforward++;
1496 	else {
1497 		ipstat.ips_forward++;
1498 		if (type)
1499 			ipstat.ips_redirectsent++;
1500 		else
1501 			goto freecopy;
1502 	}
1503 	if (mcopy == NULL)
1504 		goto freert;
1505 
1506 	switch (error) {
1507 
1508 	case 0:				/* forwarded, but need redirect */
1509 		/* type, code set above */
1510 		break;
1511 
1512 	case ENETUNREACH:		/* shouldn't happen, checked above */
1513 	case EHOSTUNREACH:
1514 	case ENETDOWN:
1515 	case EHOSTDOWN:
1516 	default:
1517 		type = ICMP_UNREACH;
1518 		code = ICMP_UNREACH_HOST;
1519 		break;
1520 
1521 	case EMSGSIZE:
1522 		type = ICMP_UNREACH;
1523 		code = ICMP_UNREACH_NEEDFRAG;
1524 
1525 #ifdef IPSEC
1526 		if (ipforward_rt.ro_rt) {
1527 			struct rtentry *rt = ipforward_rt.ro_rt;
1528 
1529 			if (rt->rt_rmx.rmx_mtu)
1530 				destmtu = rt->rt_rmx.rmx_mtu;
1531 			else
1532 				destmtu = ipforward_rt.ro_rt->rt_ifp->if_mtu;
1533 		}
1534 #endif /*IPSEC*/
1535 		ipstat.ips_cantfrag++;
1536 		break;
1537 
1538 	case ENOBUFS:
1539 #if 1
1540 		/*
1541 		 * a router should not generate ICMP_SOURCEQUENCH as
1542 		 * required in RFC1812 Requirements for IP Version 4 Routers.
1543 		 * source quench could be a big problem under DoS attacks,
1544 		 * or the underlying interface is rate-limited.
1545 		 */
1546 		goto freecopy;
1547 #else
1548 		type = ICMP_SOURCEQUENCH;
1549 		code = 0;
1550 		break;
1551 #endif
1552 	}
1553 
1554 	icmp_error(mcopy, type, code, dest, destmtu);
1555 	goto freert;
1556 
1557  freecopy:
1558 	if (mcopy)
1559 		m_free(mcopy);
1560  freert:
1561 #ifndef SMALL_KERNEL
1562 	if (ipmultipath && ipforward_rt.ro_rt &&
1563 	    (ipforward_rt.ro_rt->rt_flags & RTF_MPATH)) {
1564 		RTFREE(ipforward_rt.ro_rt);
1565 		ipforward_rt.ro_rt = 0;
1566 	}
1567 #endif
1568 	return;
1569 }
1570 
1571 int
1572 ip_sysctl(name, namelen, oldp, oldlenp, newp, newlen)
1573 	int *name;
1574 	u_int namelen;
1575 	void *oldp;
1576 	size_t *oldlenp;
1577 	void *newp;
1578 	size_t newlen;
1579 {
1580 	int error;
1581 #ifdef MROUTING
1582 	extern int ip_mrtproto;
1583 	extern struct mrtstat mrtstat;
1584 #endif
1585 
1586 	/* Almost all sysctl names at this level are terminal. */
1587 	if (namelen != 1 && name[0] != IPCTL_IFQUEUE)
1588 		return (ENOTDIR);
1589 
1590 	switch (name[0]) {
1591 #ifdef notyet
1592 	case IPCTL_DEFMTU:
1593 		return (sysctl_int(oldp, oldlenp, newp, newlen, &ip_mtu));
1594 #endif
1595 	case IPCTL_SOURCEROUTE:
1596 		/*
1597 		 * Don't allow this to change in a secure environment.
1598 		 */
1599 		if (newp && securelevel > 0)
1600 			return (EPERM);
1601 		return (sysctl_int(oldp, oldlenp, newp, newlen,
1602 		    &ip_dosourceroute));
1603 	case IPCTL_MTUDISC:
1604 		error = sysctl_int(oldp, oldlenp, newp, newlen,
1605 		    &ip_mtudisc);
1606 		if (ip_mtudisc != 0 && ip_mtudisc_timeout_q == NULL) {
1607 			ip_mtudisc_timeout_q =
1608 			    rt_timer_queue_create(ip_mtudisc_timeout);
1609 		} else if (ip_mtudisc == 0 && ip_mtudisc_timeout_q != NULL) {
1610 			rt_timer_queue_destroy(ip_mtudisc_timeout_q, TRUE);
1611 			Free(ip_mtudisc_timeout_q);
1612 			ip_mtudisc_timeout_q = NULL;
1613 		}
1614 		return error;
1615 	case IPCTL_MTUDISCTIMEOUT:
1616 		error = sysctl_int(oldp, oldlenp, newp, newlen,
1617 		   &ip_mtudisc_timeout);
1618 		if (ip_mtudisc_timeout_q != NULL)
1619 			rt_timer_queue_change(ip_mtudisc_timeout_q,
1620 					      ip_mtudisc_timeout);
1621 		return (error);
1622 	case IPCTL_IPSEC_ENC_ALGORITHM:
1623 	        return (sysctl_tstring(oldp, oldlenp, newp, newlen,
1624 				       ipsec_def_enc, sizeof(ipsec_def_enc)));
1625 	case IPCTL_IPSEC_AUTH_ALGORITHM:
1626 	        return (sysctl_tstring(oldp, oldlenp, newp, newlen,
1627 				       ipsec_def_auth,
1628 				       sizeof(ipsec_def_auth)));
1629 	case IPCTL_IPSEC_IPCOMP_ALGORITHM:
1630 	        return (sysctl_tstring(oldp, oldlenp, newp, newlen,
1631 				       ipsec_def_comp,
1632 				       sizeof(ipsec_def_comp)));
1633 	case IPCTL_IFQUEUE:
1634 	        return (sysctl_ifq(name + 1, namelen - 1,
1635 		    oldp, oldlenp, newp, newlen, &ipintrq));
1636 	case IPCTL_STATS:
1637 		if (newp != NULL)
1638 			return (EPERM);
1639 		return (sysctl_struct(oldp, oldlenp, newp, newlen,
1640 		    &ipstat, sizeof(ipstat)));
1641 	case IPCTL_MRTSTATS:
1642 #ifdef MROUTING
1643 		if (newp != NULL)
1644 			return (EPERM);
1645 		return (sysctl_struct(oldp, oldlenp, newp, newlen,
1646 		    &mrtstat, sizeof(mrtstat)));
1647 #else
1648 		return (EOPNOTSUPP);
1649 #endif
1650 	case IPCTL_MRTPROTO:
1651 #ifdef MROUTING
1652 		return (sysctl_rdint(oldp, oldlenp, newp, ip_mrtproto));
1653 #else
1654 		return (EOPNOTSUPP);
1655 #endif
1656 	default:
1657 		if (name[0] < IPCTL_MAXID)
1658 			return (sysctl_int_arr(ipctl_vars, name, namelen,
1659 			    oldp, oldlenp, newp, newlen));
1660 		return (EOPNOTSUPP);
1661 	}
1662 	/* NOTREACHED */
1663 }
1664 
1665 void
1666 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1667     struct mbuf *m)
1668 {
1669 #ifdef SO_TIMESTAMP
1670 	if (inp->inp_socket->so_options & SO_TIMESTAMP) {
1671 		struct timeval tv;
1672 
1673 		microtime(&tv);
1674 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1675 		    SCM_TIMESTAMP, SOL_SOCKET);
1676 		if (*mp)
1677 			mp = &(*mp)->m_next;
1678 	}
1679 #endif
1680 	if (inp->inp_flags & INP_RECVDSTADDR) {
1681 		*mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
1682 		    sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1683 		if (*mp)
1684 			mp = &(*mp)->m_next;
1685 	}
1686 #ifdef notyet
1687 	/* this code is broken and will probably never be fixed. */
1688 	/* options were tossed already */
1689 	if (inp->inp_flags & INP_RECVOPTS) {
1690 		*mp = sbcreatecontrol((caddr_t) opts_deleted_above,
1691 		    sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1692 		if (*mp)
1693 			mp = &(*mp)->m_next;
1694 	}
1695 	/* ip_srcroute doesn't do what we want here, need to fix */
1696 	if (inp->inp_flags & INP_RECVRETOPTS) {
1697 		*mp = sbcreatecontrol((caddr_t) ip_srcroute(),
1698 		    sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1699 		if (*mp)
1700 			mp = &(*mp)->m_next;
1701 	}
1702 #endif
1703 	if (inp->inp_flags & INP_RECVIF) {
1704 		struct sockaddr_dl sdl;
1705 		struct ifnet *ifp;
1706 
1707 		if ((ifp = m->m_pkthdr.rcvif) == NULL ||
1708 		    ifp->if_sadl == NULL) {
1709 			bzero(&sdl, sizeof(sdl));
1710 			sdl.sdl_len = offsetof(struct sockaddr_dl, sdl_data[0]);
1711 			sdl.sdl_family = AF_LINK;
1712 			sdl.sdl_index = ifp != NULL ? ifp->if_index : 0;
1713 			sdl.sdl_nlen = sdl.sdl_alen = sdl.sdl_slen = 0;
1714 			*mp = sbcreatecontrol((caddr_t) &sdl, sdl.sdl_len,
1715 			    IP_RECVIF, IPPROTO_IP);
1716 		} else {
1717 			*mp = sbcreatecontrol((caddr_t) ifp->if_sadl,
1718 			    ifp->if_sadl->sdl_len, IP_RECVIF, IPPROTO_IP);
1719 		}
1720 		if (*mp)
1721 			mp = &(*mp)->m_next;
1722 	}
1723 	if (inp->inp_flags & INP_RECVTTL) {
1724 		*mp = sbcreatecontrol((caddr_t) &ip->ip_ttl,
1725 		    sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
1726 		if (*mp)
1727 			mp = &(*mp)->m_next;
1728 	}
1729 }
1730 
1731