xref: /openbsd-src/sys/netinet/ip_input.c (revision 6a13ef69787db04ae501a22e92fa10865b44fd7c)
1 /*	$OpenBSD: ip_input.c,v 1.291 2016/12/20 18:33:43 bluhm 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/sysctl.h>
46 #include <sys/pool.h>
47 #include <sys/task.h>
48 
49 #include <net/if.h>
50 #include <net/if_var.h>
51 #include <net/if_dl.h>
52 #include <net/route.h>
53 #include <net/netisr.h>
54 
55 #include <netinet/in.h>
56 #include <netinet/in_systm.h>
57 #include <netinet/if_ether.h>
58 #include <netinet/ip.h>
59 #include <netinet/in_pcb.h>
60 #include <netinet/in_var.h>
61 #include <netinet/ip_var.h>
62 #include <netinet/ip_icmp.h>
63 
64 #if NPF > 0
65 #include <net/pfvar.h>
66 #endif
67 
68 #ifdef MROUTING
69 #include <netinet/ip_mroute.h>
70 #endif
71 
72 #ifdef IPSEC
73 #include <netinet/ip_ipsp.h>
74 #endif /* IPSEC */
75 
76 #if NCARP > 0
77 #include <net/if_types.h>
78 #include <netinet/ip_carp.h>
79 #endif
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 
110 struct rttimer_queue *ip_mtudisc_timeout_q = NULL;
111 
112 /* Keep track of memory used for reassembly */
113 int	ip_maxqueue = 300;
114 int	ip_frags = 0;
115 
116 int *ipctl_vars[IPCTL_MAXID] = IPCTL_VARS;
117 
118 struct niqueue ipintrq = NIQUEUE_INITIALIZER(IFQ_MAXLEN, NETISR_IP);
119 
120 struct pool ipqent_pool;
121 struct pool ipq_pool;
122 
123 struct cpumem *ipcounters;
124 
125 int ip_sysctl_ipstat(void *, size_t *, void *);
126 
127 static struct mbuf_queue	ipsend_mq;
128 
129 void	ip_ours(struct mbuf *);
130 int	ip_dooptions(struct mbuf *, struct ifnet *);
131 int	in_ouraddr(struct mbuf *, struct ifnet *, struct rtentry **);
132 #ifdef IPSEC
133 int	ip_input_ipsec_fwd_check(struct mbuf *, int);
134 int	ip_input_ipsec_ours_check(struct mbuf *, int);
135 #endif /* IPSEC */
136 
137 static void ip_send_dispatch(void *);
138 static struct task ipsend_task = TASK_INITIALIZER(ip_send_dispatch, &ipsend_mq);
139 /*
140  * Used to save the IP options in case a protocol wants to respond
141  * to an incoming packet over the same route if the packet got here
142  * using IP source routing.  This allows connection establishment and
143  * maintenance when the remote end is on a network that is not known
144  * to us.
145  */
146 struct ip_srcrt {
147 	int		isr_nhops;		   /* number of hops */
148 	struct in_addr	isr_dst;		   /* final destination */
149 	char		isr_nop;		   /* one NOP to align */
150 	char		isr_hdr[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN & OFFSET */
151 	struct in_addr	isr_routes[MAX_IPOPTLEN/sizeof(struct in_addr)];
152 };
153 
154 void save_rte(struct mbuf *, u_char *, struct in_addr);
155 
156 /*
157  * IP initialization: fill in IP protocol switch table.
158  * All protocols not implemented in kernel go to raw IP protocol handler.
159  */
160 void
161 ip_init(void)
162 {
163 	struct protosw *pr;
164 	int i;
165 	const u_int16_t defbaddynamicports_tcp[] = DEFBADDYNAMICPORTS_TCP;
166 	const u_int16_t defbaddynamicports_udp[] = DEFBADDYNAMICPORTS_UDP;
167 	const u_int16_t defrootonlyports_tcp[] = DEFROOTONLYPORTS_TCP;
168 	const u_int16_t defrootonlyports_udp[] = DEFROOTONLYPORTS_UDP;
169 
170 	ipcounters = counters_alloc(ips_ncounters, M_COUNTERS);
171 
172 	pool_init(&ipqent_pool, sizeof(struct ipqent), 0,
173 	    IPL_SOFTNET, 0, "ipqe",  NULL);
174 	pool_init(&ipq_pool, sizeof(struct ipq), 0,
175 	    IPL_SOFTNET, 0, "ipq", NULL);
176 
177 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
178 	if (pr == NULL)
179 		panic("ip_init");
180 	for (i = 0; i < IPPROTO_MAX; i++)
181 		ip_protox[i] = pr - inetsw;
182 	for (pr = inetdomain.dom_protosw;
183 	    pr < inetdomain.dom_protoswNPROTOSW; pr++)
184 		if (pr->pr_domain->dom_family == PF_INET &&
185 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW &&
186 		    pr->pr_protocol < IPPROTO_MAX)
187 			ip_protox[pr->pr_protocol] = pr - inetsw;
188 	LIST_INIT(&ipq);
189 	if (ip_mtudisc != 0)
190 		ip_mtudisc_timeout_q =
191 		    rt_timer_queue_create(ip_mtudisc_timeout);
192 
193 	/* Fill in list of ports not to allocate dynamically. */
194 	memset(&baddynamicports, 0, sizeof(baddynamicports));
195 	for (i = 0; defbaddynamicports_tcp[i] != 0; i++)
196 		DP_SET(baddynamicports.tcp, defbaddynamicports_tcp[i]);
197 	for (i = 0; defbaddynamicports_udp[i] != 0; i++)
198 		DP_SET(baddynamicports.udp, defbaddynamicports_udp[i]);
199 
200 	/* Fill in list of ports only root can bind to. */
201 	memset(&rootonlyports, 0, sizeof(rootonlyports));
202 	for (i = 0; defrootonlyports_tcp[i] != 0; i++)
203 		DP_SET(rootonlyports.tcp, defrootonlyports_tcp[i]);
204 	for (i = 0; defrootonlyports_udp[i] != 0; i++)
205 		DP_SET(rootonlyports.udp, defrootonlyports_udp[i]);
206 
207 	strlcpy(ipsec_def_enc, IPSEC_DEFAULT_DEF_ENC, sizeof(ipsec_def_enc));
208 	strlcpy(ipsec_def_auth, IPSEC_DEFAULT_DEF_AUTH, sizeof(ipsec_def_auth));
209 	strlcpy(ipsec_def_comp, IPSEC_DEFAULT_DEF_COMP, sizeof(ipsec_def_comp));
210 
211 	mq_init(&ipsend_mq, 64, IPL_SOFTNET);
212 }
213 
214 void
215 ipintr(void)
216 {
217 	struct mbuf *m;
218 
219 	/*
220 	 * Get next datagram off input queue and get IP header
221 	 * in first mbuf.
222 	 */
223 	while ((m = niq_dequeue(&ipintrq)) != NULL) {
224 #ifdef	DIAGNOSTIC
225 		if ((m->m_flags & M_PKTHDR) == 0)
226 			panic("ipintr no HDR");
227 #endif
228 		ipv4_input(m);
229 	}
230 }
231 
232 /*
233  * IPv4 input routine.
234  *
235  * Checksum and byte swap header.  Process options. Forward or deliver.
236  */
237 void
238 ipv4_input(struct mbuf *m)
239 {
240 	struct ifnet	*ifp;
241 	struct rtentry	*rt = NULL;
242 	struct ip	*ip;
243 	int hlen, len;
244 #if defined(MROUTING) || defined(IPSEC)
245 	int rv;
246 #endif
247 	in_addr_t pfrdr = 0;
248 
249 	ifp = if_get(m->m_pkthdr.ph_ifidx);
250 	if (ifp == NULL)
251 		goto bad;
252 
253 	ipstat_inc(ips_total);
254 	if (m->m_len < sizeof (struct ip) &&
255 	    (m = m_pullup(m, sizeof (struct ip))) == NULL) {
256 		ipstat_inc(ips_toosmall);
257 		goto out;
258 	}
259 	ip = mtod(m, struct ip *);
260 	if (ip->ip_v != IPVERSION) {
261 		ipstat_inc(ips_badvers);
262 		goto bad;
263 	}
264 	hlen = ip->ip_hl << 2;
265 	if (hlen < sizeof(struct ip)) {	/* minimum header length */
266 		ipstat_inc(ips_badhlen);
267 		goto bad;
268 	}
269 	if (hlen > m->m_len) {
270 		if ((m = m_pullup(m, hlen)) == NULL) {
271 			ipstat_inc(ips_badhlen);
272 			goto out;
273 		}
274 		ip = mtod(m, struct ip *);
275 	}
276 
277 	/* 127/8 must not appear on wire - RFC1122 */
278 	if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
279 	    (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
280 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
281 			ipstat_inc(ips_badaddr);
282 			goto bad;
283 		}
284 	}
285 
286 	if ((m->m_pkthdr.csum_flags & M_IPV4_CSUM_IN_OK) == 0) {
287 		if (m->m_pkthdr.csum_flags & M_IPV4_CSUM_IN_BAD) {
288 			ipstat_inc(ips_badsum);
289 			goto bad;
290 		}
291 
292 		ipstat_inc(ips_inswcsum);
293 		if (in_cksum(m, hlen) != 0) {
294 			ipstat_inc(ips_badsum);
295 			goto bad;
296 		}
297 	}
298 
299 	/* Retrieve the packet length. */
300 	len = ntohs(ip->ip_len);
301 
302 	/*
303 	 * Convert fields to host representation.
304 	 */
305 	if (len < hlen) {
306 		ipstat_inc(ips_badlen);
307 		goto bad;
308 	}
309 
310 	/*
311 	 * Check that the amount of data in the buffers
312 	 * is at least as much as the IP header would have us expect.
313 	 * Trim mbufs if longer than we expect.
314 	 * Drop packet if shorter than we expect.
315 	 */
316 	if (m->m_pkthdr.len < len) {
317 		ipstat_inc(ips_tooshort);
318 		goto bad;
319 	}
320 	if (m->m_pkthdr.len > len) {
321 		if (m->m_len == m->m_pkthdr.len) {
322 			m->m_len = len;
323 			m->m_pkthdr.len = len;
324 		} else
325 			m_adj(m, len - m->m_pkthdr.len);
326 	}
327 
328 #if NCARP > 0
329 	if (ifp->if_type == IFT_CARP && ip->ip_p != IPPROTO_ICMP &&
330 	    carp_lsdrop(m, AF_INET, &ip->ip_src.s_addr, &ip->ip_dst.s_addr))
331 		goto bad;
332 #endif
333 
334 #if NPF > 0
335 	/*
336 	 * Packet filter
337 	 */
338 	pfrdr = ip->ip_dst.s_addr;
339 	if (pf_test(AF_INET, PF_IN, ifp, &m) != PF_PASS)
340 		goto bad;
341 	if (m == NULL)
342 		goto out;
343 
344 	ip = mtod(m, struct ip *);
345 	hlen = ip->ip_hl << 2;
346 	pfrdr = (pfrdr != ip->ip_dst.s_addr);
347 #endif
348 
349 	/*
350 	 * Process options and, if not destined for us,
351 	 * ship it on.  ip_dooptions returns 1 when an
352 	 * error was detected (causing an icmp message
353 	 * to be sent and the original packet to be freed).
354 	 */
355 	if (hlen > sizeof (struct ip) && ip_dooptions(m, ifp)) {
356 	        goto out;
357 	}
358 
359 	if (ip->ip_dst.s_addr == INADDR_BROADCAST ||
360 	    ip->ip_dst.s_addr == INADDR_ANY) {
361 		ip_ours(m);
362 		goto out;
363 	}
364 
365 	if (in_ouraddr(m, ifp, &rt)) {
366 		ip_ours(m);
367 		goto out;
368 	}
369 
370 	if (IN_MULTICAST(ip->ip_dst.s_addr)) {
371 		/*
372 		 * Make sure M_MCAST is set.  It should theoretically
373 		 * already be there, but let's play safe because upper
374 		 * layers check for this flag.
375 		 */
376 		m->m_flags |= M_MCAST;
377 
378 #ifdef MROUTING
379 		if (ipmforwarding && ip_mrouter[ifp->if_rdomain]) {
380 			if (m->m_flags & M_EXT) {
381 				if ((m = m_pullup(m, hlen)) == NULL) {
382 					ipstat_inc(ips_toosmall);
383 					goto out;
384 				}
385 				ip = mtod(m, struct ip *);
386 			}
387 			/*
388 			 * If we are acting as a multicast router, all
389 			 * incoming multicast packets are passed to the
390 			 * kernel-level multicast forwarding function.
391 			 * The packet is returned (relatively) intact; if
392 			 * ip_mforward() returns a non-zero value, the packet
393 			 * must be discarded, else it may be accepted below.
394 			 *
395 			 * (The IP ident field is put in the same byte order
396 			 * as expected when ip_mforward() is called from
397 			 * ip_output().)
398 			 */
399 			KERNEL_LOCK();
400 			rv = ip_mforward(m, ifp);
401 			KERNEL_UNLOCK();
402 			if (rv != 0) {
403 				ipstat_inc(ips_cantforward);
404 				goto bad;
405 			}
406 
407 			/*
408 			 * The process-level routing daemon needs to receive
409 			 * all multicast IGMP packets, whether or not this
410 			 * host belongs to their destination groups.
411 			 */
412 			if (ip->ip_p == IPPROTO_IGMP) {
413 				ip_ours(m);
414 				goto out;
415 			}
416 			ipstat_inc(ips_forward);
417 		}
418 #endif
419 		/*
420 		 * See if we belong to the destination multicast group on the
421 		 * arrival interface.
422 		 */
423 		if (!in_hasmulti(&ip->ip_dst, ifp)) {
424 			ipstat_inc(ips_notmember);
425 			if (!IN_LOCAL_GROUP(ip->ip_dst.s_addr))
426 				ipstat_inc(ips_cantforward);
427 			goto bad;
428 		}
429 		ip_ours(m);
430 		goto out;
431 	}
432 
433 #if NCARP > 0
434 	if (ifp->if_type == IFT_CARP && ip->ip_p == IPPROTO_ICMP &&
435 	    carp_lsdrop(m, AF_INET, &ip->ip_src.s_addr, &ip->ip_dst.s_addr))
436 		goto bad;
437 #endif
438 	/*
439 	 * Not for us; forward if possible and desirable.
440 	 */
441 	if (ipforwarding == 0) {
442 		ipstat_inc(ips_cantforward);
443 		goto bad;
444 	}
445 #ifdef IPSEC
446 	if (ipsec_in_use) {
447 		KERNEL_LOCK();
448 		rv = ip_input_ipsec_fwd_check(m, hlen);
449 		KERNEL_UNLOCK();
450 		if (rv != 0) {
451 			ipstat_inc(ips_cantforward);
452 			goto bad;
453 		}
454 		/*
455 		 * Fall through, forward packet. Outbound IPsec policy
456 		 * checking will occur in ip_output().
457 		 */
458 	}
459 #endif /* IPSEC */
460 
461 	ip_forward(m, ifp, rt, pfrdr);
462 	if_put(ifp);
463 	return;
464 bad:
465 	m_freem(m);
466 out:
467 	rtfree(rt);
468 	if_put(ifp);
469 }
470 
471 /*
472  * IPv4 local-delivery routine.
473  *
474  * If fragmented try to reassemble.  Pass to next level.
475  */
476 void
477 ip_ours(struct mbuf *m)
478 {
479 	struct ip *ip = mtod(m, struct ip *);
480 	struct ipq *fp;
481 	struct ipqent *ipqe;
482 	int mff, hlen;
483 
484 	hlen = ip->ip_hl << 2;
485 
486 	/* pf might have modified stuff, might have to chksum */
487 	in_proto_cksum_out(m, NULL);
488 
489 	/*
490 	 * If offset or IP_MF are set, must reassemble.
491 	 * Otherwise, nothing need be done.
492 	 * (We could look in the reassembly queue to see
493 	 * if the packet was previously fragmented,
494 	 * but it's not worth the time; just let them time out.)
495 	 */
496 	if (ip->ip_off &~ htons(IP_DF | IP_RF)) {
497 		if (m->m_flags & M_EXT) {		/* XXX */
498 			if ((m = m_pullup(m, hlen)) == NULL) {
499 				ipstat_inc(ips_toosmall);
500 				return;
501 			}
502 			ip = mtod(m, struct ip *);
503 		}
504 
505 		/*
506 		 * Look for queue of fragments
507 		 * of this datagram.
508 		 */
509 		LIST_FOREACH(fp, &ipq, ipq_q)
510 			if (ip->ip_id == fp->ipq_id &&
511 			    ip->ip_src.s_addr == fp->ipq_src.s_addr &&
512 			    ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
513 			    ip->ip_p == fp->ipq_p)
514 				goto found;
515 		fp = 0;
516 found:
517 
518 		/*
519 		 * Adjust ip_len to not reflect header,
520 		 * set ipqe_mff if more fragments are expected,
521 		 * convert offset of this to bytes.
522 		 */
523 		ip->ip_len = htons(ntohs(ip->ip_len) - hlen);
524 		mff = (ip->ip_off & htons(IP_MF)) != 0;
525 		if (mff) {
526 			/*
527 			 * Make sure that fragments have a data length
528 			 * that's a non-zero multiple of 8 bytes.
529 			 */
530 			if (ntohs(ip->ip_len) == 0 ||
531 			    (ntohs(ip->ip_len) & 0x7) != 0) {
532 				ipstat_inc(ips_badfrags);
533 				goto bad;
534 			}
535 		}
536 		ip->ip_off = htons(ntohs(ip->ip_off) << 3);
537 
538 		/*
539 		 * If datagram marked as having more fragments
540 		 * or if this is not the first fragment,
541 		 * attempt reassembly; if it succeeds, proceed.
542 		 */
543 		if (mff || ip->ip_off) {
544 			ipstat_inc(ips_fragments);
545 			if (ip_frags + 1 > ip_maxqueue) {
546 				ip_flush();
547 				ipstat_inc(ips_rcvmemdrop);
548 				goto bad;
549 			}
550 
551 			ipqe = pool_get(&ipqent_pool, PR_NOWAIT);
552 			if (ipqe == NULL) {
553 				ipstat_inc(ips_rcvmemdrop);
554 				goto bad;
555 			}
556 			ip_frags++;
557 			ipqe->ipqe_mff = mff;
558 			ipqe->ipqe_m = m;
559 			ipqe->ipqe_ip = ip;
560 			m = ip_reass(ipqe, fp);
561 			if (m == NULL) {
562 				return;
563 			}
564 			ipstat_inc(ips_reassembled);
565 			ip = mtod(m, struct ip *);
566 			hlen = ip->ip_hl << 2;
567 			ip->ip_len = htons(ntohs(ip->ip_len) + hlen);
568 		} else
569 			if (fp)
570 				ip_freef(fp);
571 	}
572 
573 #ifdef IPSEC
574 	if (ipsec_in_use) {
575 		if (ip_input_ipsec_ours_check(m, hlen) != 0) {
576 			ipstat_inc(ips_cantforward);
577 			goto bad;
578 		}
579 	}
580 	/* Otherwise, just fall through and deliver the packet */
581 #endif /* IPSEC */
582 
583 	/*
584 	 * Switch out to protocol's input routine.
585 	 */
586 	ipstat_inc(ips_delivered);
587 	(*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen, NULL, 0);
588 	return;
589 bad:
590 	m_freem(m);
591 }
592 
593 int
594 in_ouraddr(struct mbuf *m, struct ifnet *ifp, struct rtentry **prt)
595 {
596 	struct rtentry		*rt;
597 	struct ip		*ip;
598 	struct sockaddr_in	 sin;
599 	int			 match = 0;
600 
601 #if NPF > 0
602 	switch (pf_ouraddr(m)) {
603 	case 0:
604 		return (0);
605 	case 1:
606 		return (1);
607 	default:
608 		/* pf does not know it */
609 		break;
610 	}
611 #endif
612 
613 	ip = mtod(m, struct ip *);
614 
615 	memset(&sin, 0, sizeof(sin));
616 	sin.sin_len = sizeof(sin);
617 	sin.sin_family = AF_INET;
618 	sin.sin_addr = ip->ip_dst;
619 	rt = rtalloc_mpath(sintosa(&sin), &ip->ip_src.s_addr,
620 	    m->m_pkthdr.ph_rtableid);
621 	if (rtisvalid(rt)) {
622 		if (ISSET(rt->rt_flags, RTF_LOCAL))
623 			match = 1;
624 
625 		/*
626 		 * If directedbcast is enabled we only consider it local
627 		 * if it is received on the interface with that address.
628 		 */
629 		if (ISSET(rt->rt_flags, RTF_BROADCAST) &&
630 		    (!ip_directedbcast || rt->rt_ifidx == ifp->if_index)) {
631 			match = 1;
632 
633 			/* Make sure M_BCAST is set */
634 			m->m_flags |= M_BCAST;
635 		}
636 	}
637 	*prt = rt;
638 
639 	if (!match) {
640 		struct ifaddr *ifa;
641 
642 		/*
643 		 * No local address or broadcast address found, so check for
644 		 * ancient classful broadcast addresses.
645 		 * It must have been broadcast on the link layer, and for an
646 		 * address on the interface it was received on.
647 		 */
648 		if (!ISSET(m->m_flags, M_BCAST) ||
649 		    !IN_CLASSFULBROADCAST(ip->ip_dst.s_addr, ip->ip_dst.s_addr))
650 			return (0);
651 
652 		if (ifp->if_rdomain != rtable_l2(m->m_pkthdr.ph_rtableid))
653 			return (0);
654 		/*
655 		 * The check in the loop assumes you only rx a packet on an UP
656 		 * interface, and that M_BCAST will only be set on a BROADCAST
657 		 * interface.
658 		 */
659 		KERNEL_LOCK();
660 		TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
661 			if (ifa->ifa_addr->sa_family != AF_INET)
662 				continue;
663 
664 			if (IN_CLASSFULBROADCAST(ip->ip_dst.s_addr,
665 			    ifatoia(ifa)->ia_addr.sin_addr.s_addr)) {
666 			    	match = 1;
667 			    	break;
668 			}
669 		}
670 		KERNEL_UNLOCK();
671 	}
672 
673 	return (match);
674 }
675 
676 #ifdef IPSEC
677 int
678 ip_input_ipsec_fwd_check(struct mbuf *m, int hlen)
679 {
680 	struct tdb *tdb;
681 	struct tdb_ident *tdbi;
682 	struct m_tag *mtag;
683 	int error = 0;
684 
685 	/*
686 	 * IPsec policy check for forwarded packets. Look at
687 	 * inner-most IPsec SA used.
688 	 */
689 	mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
690 	if (mtag != NULL) {
691 		tdbi = (struct tdb_ident *)(mtag + 1);
692 		tdb = gettdb(tdbi->rdomain, tdbi->spi, &tdbi->dst, tdbi->proto);
693 	} else
694 		tdb = NULL;
695 	ipsp_spd_lookup(m, AF_INET, hlen, &error, IPSP_DIRECTION_IN, tdb, NULL,
696 	    0);
697 
698 	return error;
699 }
700 
701 int
702 ip_input_ipsec_ours_check(struct mbuf *m, int hlen)
703 {
704 	struct ip *ip = mtod(m, struct ip *);
705 	struct tdb *tdb;
706 	struct tdb_ident *tdbi;
707 	struct m_tag *mtag;
708 	int error = 0;
709 
710 	/*
711 	 * If it's a protected packet for us, skip the policy check.
712 	 * That's because we really only care about the properties of
713 	 * the protected packet, and not the intermediate versions.
714 	 * While this is not the most paranoid setting, it allows
715 	 * some flexibility in handling nested tunnels (in setting up
716 	 * the policies).
717 	 */
718 	if ((ip->ip_p == IPPROTO_ESP) || (ip->ip_p == IPPROTO_AH) ||
719 	    (ip->ip_p == IPPROTO_IPCOMP))
720 		return 0;
721 
722 	/*
723 	 * If the protected packet was tunneled, then we need to
724 	 * verify the protected packet's information, not the
725 	 * external headers. Thus, skip the policy lookup for the
726 	 * external packet, and keep the IPsec information linked on
727 	 * the packet header (the encapsulation routines know how
728 	 * to deal with that).
729 	 */
730 	if ((ip->ip_p == IPPROTO_IPIP) || (ip->ip_p == IPPROTO_IPV6))
731 		return 0;
732 
733 	/*
734 	 * If the protected packet is TCP or UDP, we'll do the
735 	 * policy check in the respective input routine, so we can
736 	 * check for bypass sockets.
737 	 */
738 	if ((ip->ip_p == IPPROTO_TCP) || (ip->ip_p == IPPROTO_UDP))
739 		return 0;
740 
741 	/*
742 	 * IPsec policy check for local-delivery packets. Look at the
743 	 * inner-most SA that protected the packet. This is in fact
744 	 * a bit too restrictive (it could end up causing packets to
745 	 * be dropped that semantically follow the policy, e.g., in
746 	 * certain SA-bundle configurations); but the alternative is
747 	 * very complicated (and requires keeping track of what
748 	 * kinds of tunneling headers have been seen in-between the
749 	 * IPsec headers), and I don't think we lose much functionality
750 	 * that's needed in the real world (who uses bundles anyway ?).
751 	 */
752 	mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
753 	if (mtag) {
754 		tdbi = (struct tdb_ident *)(mtag + 1);
755 		tdb = gettdb(tdbi->rdomain, tdbi->spi, &tdbi->dst,
756 		    tdbi->proto);
757 	} else
758 		tdb = NULL;
759 	ipsp_spd_lookup(m, AF_INET, hlen, &error, IPSP_DIRECTION_IN,
760 	    tdb, NULL, 0);
761 
762 	return error;
763 }
764 #endif /* IPSEC */
765 
766 /*
767  * Take incoming datagram fragment and try to
768  * reassemble it into whole datagram.  If a chain for
769  * reassembly of this datagram already exists, then it
770  * is given as fp; otherwise have to make a chain.
771  */
772 struct mbuf *
773 ip_reass(struct ipqent *ipqe, struct ipq *fp)
774 {
775 	struct mbuf *m = ipqe->ipqe_m;
776 	struct ipqent *nq, *p, *q;
777 	struct ip *ip;
778 	struct mbuf *t;
779 	int hlen = ipqe->ipqe_ip->ip_hl << 2;
780 	int i, next;
781 	u_int8_t ecn, ecn0;
782 
783 	/*
784 	 * Presence of header sizes in mbufs
785 	 * would confuse code below.
786 	 */
787 	m->m_data += hlen;
788 	m->m_len -= hlen;
789 
790 	/*
791 	 * If first fragment to arrive, create a reassembly queue.
792 	 */
793 	if (fp == NULL) {
794 		fp = pool_get(&ipq_pool, PR_NOWAIT);
795 		if (fp == NULL)
796 			goto dropfrag;
797 		LIST_INSERT_HEAD(&ipq, fp, ipq_q);
798 		fp->ipq_ttl = IPFRAGTTL;
799 		fp->ipq_p = ipqe->ipqe_ip->ip_p;
800 		fp->ipq_id = ipqe->ipqe_ip->ip_id;
801 		LIST_INIT(&fp->ipq_fragq);
802 		fp->ipq_src = ipqe->ipqe_ip->ip_src;
803 		fp->ipq_dst = ipqe->ipqe_ip->ip_dst;
804 		p = NULL;
805 		goto insert;
806 	}
807 
808 	/*
809 	 * Handle ECN by comparing this segment with the first one;
810 	 * if CE is set, do not lose CE.
811 	 * drop if CE and not-ECT are mixed for the same packet.
812 	 */
813 	ecn = ipqe->ipqe_ip->ip_tos & IPTOS_ECN_MASK;
814 	ecn0 = LIST_FIRST(&fp->ipq_fragq)->ipqe_ip->ip_tos & IPTOS_ECN_MASK;
815 	if (ecn == IPTOS_ECN_CE) {
816 		if (ecn0 == IPTOS_ECN_NOTECT)
817 			goto dropfrag;
818 		if (ecn0 != IPTOS_ECN_CE)
819 			LIST_FIRST(&fp->ipq_fragq)->ipqe_ip->ip_tos |= IPTOS_ECN_CE;
820 	}
821 	if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT)
822 		goto dropfrag;
823 
824 	/*
825 	 * Find a segment which begins after this one does.
826 	 */
827 	for (p = NULL, q = LIST_FIRST(&fp->ipq_fragq); q != NULL;
828 	    p = q, q = LIST_NEXT(q, ipqe_q))
829 		if (ntohs(q->ipqe_ip->ip_off) > ntohs(ipqe->ipqe_ip->ip_off))
830 			break;
831 
832 	/*
833 	 * If there is a preceding segment, it may provide some of
834 	 * our data already.  If so, drop the data from the incoming
835 	 * segment.  If it provides all of our data, drop us.
836 	 */
837 	if (p != NULL) {
838 		i = ntohs(p->ipqe_ip->ip_off) + ntohs(p->ipqe_ip->ip_len) -
839 		    ntohs(ipqe->ipqe_ip->ip_off);
840 		if (i > 0) {
841 			if (i >= ntohs(ipqe->ipqe_ip->ip_len))
842 				goto dropfrag;
843 			m_adj(ipqe->ipqe_m, i);
844 			ipqe->ipqe_ip->ip_off =
845 			    htons(ntohs(ipqe->ipqe_ip->ip_off) + i);
846 			ipqe->ipqe_ip->ip_len =
847 			    htons(ntohs(ipqe->ipqe_ip->ip_len) - i);
848 		}
849 	}
850 
851 	/*
852 	 * While we overlap succeeding segments trim them or,
853 	 * if they are completely covered, dequeue them.
854 	 */
855 	for (; q != NULL &&
856 	    ntohs(ipqe->ipqe_ip->ip_off) + ntohs(ipqe->ipqe_ip->ip_len) >
857 	    ntohs(q->ipqe_ip->ip_off); q = nq) {
858 		i = (ntohs(ipqe->ipqe_ip->ip_off) +
859 		    ntohs(ipqe->ipqe_ip->ip_len)) - ntohs(q->ipqe_ip->ip_off);
860 		if (i < ntohs(q->ipqe_ip->ip_len)) {
861 			q->ipqe_ip->ip_len =
862 			    htons(ntohs(q->ipqe_ip->ip_len) - i);
863 			q->ipqe_ip->ip_off =
864 			    htons(ntohs(q->ipqe_ip->ip_off) + i);
865 			m_adj(q->ipqe_m, i);
866 			break;
867 		}
868 		nq = LIST_NEXT(q, ipqe_q);
869 		m_freem(q->ipqe_m);
870 		LIST_REMOVE(q, ipqe_q);
871 		pool_put(&ipqent_pool, q);
872 		ip_frags--;
873 	}
874 
875 insert:
876 	/*
877 	 * Stick new segment in its place;
878 	 * check for complete reassembly.
879 	 */
880 	if (p == NULL) {
881 		LIST_INSERT_HEAD(&fp->ipq_fragq, ipqe, ipqe_q);
882 	} else {
883 		LIST_INSERT_AFTER(p, ipqe, ipqe_q);
884 	}
885 	next = 0;
886 	for (p = NULL, q = LIST_FIRST(&fp->ipq_fragq); q != NULL;
887 	    p = q, q = LIST_NEXT(q, ipqe_q)) {
888 		if (ntohs(q->ipqe_ip->ip_off) != next)
889 			return (0);
890 		next += ntohs(q->ipqe_ip->ip_len);
891 	}
892 	if (p->ipqe_mff)
893 		return (0);
894 
895 	/*
896 	 * Reassembly is complete.  Check for a bogus message size and
897 	 * concatenate fragments.
898 	 */
899 	q = LIST_FIRST(&fp->ipq_fragq);
900 	ip = q->ipqe_ip;
901 	if ((next + (ip->ip_hl << 2)) > IP_MAXPACKET) {
902 		ipstat_inc(ips_toolong);
903 		ip_freef(fp);
904 		return (0);
905 	}
906 	m = q->ipqe_m;
907 	t = m->m_next;
908 	m->m_next = 0;
909 	m_cat(m, t);
910 	nq = LIST_NEXT(q, ipqe_q);
911 	pool_put(&ipqent_pool, q);
912 	ip_frags--;
913 	for (q = nq; q != NULL; q = nq) {
914 		t = q->ipqe_m;
915 		nq = LIST_NEXT(q, ipqe_q);
916 		pool_put(&ipqent_pool, q);
917 		ip_frags--;
918 		m_cat(m, t);
919 	}
920 
921 	/*
922 	 * Create header for new ip packet by
923 	 * modifying header of first packet;
924 	 * dequeue and discard fragment reassembly header.
925 	 * Make header visible.
926 	 */
927 	ip->ip_len = htons(next);
928 	ip->ip_src = fp->ipq_src;
929 	ip->ip_dst = fp->ipq_dst;
930 	LIST_REMOVE(fp, ipq_q);
931 	pool_put(&ipq_pool, fp);
932 	m->m_len += (ip->ip_hl << 2);
933 	m->m_data -= (ip->ip_hl << 2);
934 	/* some debugging cruft by sklower, below, will go away soon */
935 	if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */
936 		int plen = 0;
937 		for (t = m; t; t = t->m_next)
938 			plen += t->m_len;
939 		m->m_pkthdr.len = plen;
940 	}
941 	return (m);
942 
943 dropfrag:
944 	ipstat_inc(ips_fragdropped);
945 	m_freem(m);
946 	pool_put(&ipqent_pool, ipqe);
947 	ip_frags--;
948 	return (0);
949 }
950 
951 /*
952  * Free a fragment reassembly header and all
953  * associated datagrams.
954  */
955 void
956 ip_freef(struct ipq *fp)
957 {
958 	struct ipqent *q, *p;
959 
960 	for (q = LIST_FIRST(&fp->ipq_fragq); q != NULL; q = p) {
961 		p = LIST_NEXT(q, ipqe_q);
962 		m_freem(q->ipqe_m);
963 		LIST_REMOVE(q, ipqe_q);
964 		pool_put(&ipqent_pool, q);
965 		ip_frags--;
966 	}
967 	LIST_REMOVE(fp, ipq_q);
968 	pool_put(&ipq_pool, fp);
969 }
970 
971 /*
972  * IP timer processing;
973  * if a timer expires on a reassembly queue, discard it.
974  * clear the forwarding cache, there might be a better route.
975  */
976 void
977 ip_slowtimo(void)
978 {
979 	struct ipq *fp, *nfp;
980 
981 	splsoftassert(IPL_SOFTNET);
982 
983 	for (fp = LIST_FIRST(&ipq); fp != NULL; fp = nfp) {
984 		nfp = LIST_NEXT(fp, ipq_q);
985 		if (--fp->ipq_ttl == 0) {
986 			ipstat_inc(ips_fragtimeout);
987 			ip_freef(fp);
988 		}
989 	}
990 }
991 
992 /*
993  * Drain off all datagram fragments.
994  */
995 void
996 ip_drain(void)
997 {
998 	while (!LIST_EMPTY(&ipq)) {
999 		ipstat_inc(ips_fragdropped);
1000 		ip_freef(LIST_FIRST(&ipq));
1001 	}
1002 }
1003 
1004 /*
1005  * Flush a bunch of datagram fragments, till we are down to 75%.
1006  */
1007 void
1008 ip_flush(void)
1009 {
1010 	int max = 50;
1011 
1012 	/* ipq already locked */
1013 	while (!LIST_EMPTY(&ipq) && ip_frags > ip_maxqueue * 3 / 4 && --max) {
1014 		ipstat_inc(ips_fragdropped);
1015 		ip_freef(LIST_FIRST(&ipq));
1016 	}
1017 }
1018 
1019 /*
1020  * Do option processing on a datagram,
1021  * possibly discarding it if bad options are encountered,
1022  * or forwarding it if source-routed.
1023  * Returns 1 if packet has been forwarded/freed,
1024  * 0 if the packet should be processed further.
1025  */
1026 int
1027 ip_dooptions(struct mbuf *m, struct ifnet *ifp)
1028 {
1029 	struct ip *ip = mtod(m, struct ip *);
1030 	unsigned int rtableid = m->m_pkthdr.ph_rtableid;
1031 	struct rtentry *rt;
1032 	struct sockaddr_in ipaddr;
1033 	u_char *cp;
1034 	struct ip_timestamp ipt;
1035 	struct in_ifaddr *ia;
1036 	int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0;
1037 	struct in_addr sin, dst;
1038 	u_int32_t ntime;
1039 
1040 	dst = ip->ip_dst;
1041 	cp = (u_char *)(ip + 1);
1042 	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
1043 
1044 	KERNEL_LOCK();
1045 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1046 		opt = cp[IPOPT_OPTVAL];
1047 		if (opt == IPOPT_EOL)
1048 			break;
1049 		if (opt == IPOPT_NOP)
1050 			optlen = 1;
1051 		else {
1052 			if (cnt < IPOPT_OLEN + sizeof(*cp)) {
1053 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1054 				goto bad;
1055 			}
1056 			optlen = cp[IPOPT_OLEN];
1057 			if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
1058 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1059 				goto bad;
1060 			}
1061 		}
1062 
1063 		switch (opt) {
1064 
1065 		default:
1066 			break;
1067 
1068 		/*
1069 		 * Source routing with record.
1070 		 * Find interface with current destination address.
1071 		 * If none on this machine then drop if strictly routed,
1072 		 * or do nothing if loosely routed.
1073 		 * Record interface address and bring up next address
1074 		 * component.  If strictly routed make sure next
1075 		 * address is on directly accessible net.
1076 		 */
1077 		case IPOPT_LSRR:
1078 		case IPOPT_SSRR:
1079 			if (!ip_dosourceroute) {
1080 				type = ICMP_UNREACH;
1081 				code = ICMP_UNREACH_SRCFAIL;
1082 				goto bad;
1083 			}
1084 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1085 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1086 				goto bad;
1087 			}
1088 			memset(&ipaddr, 0, sizeof(ipaddr));
1089 			ipaddr.sin_family = AF_INET;
1090 			ipaddr.sin_len = sizeof(ipaddr);
1091 			ipaddr.sin_addr = ip->ip_dst;
1092 			ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr),
1093 			    m->m_pkthdr.ph_rtableid));
1094 			if (ia == NULL) {
1095 				if (opt == IPOPT_SSRR) {
1096 					type = ICMP_UNREACH;
1097 					code = ICMP_UNREACH_SRCFAIL;
1098 					goto bad;
1099 				}
1100 				/*
1101 				 * Loose routing, and not at next destination
1102 				 * yet; nothing to do except forward.
1103 				 */
1104 				break;
1105 			}
1106 			off--;			/* 0 origin */
1107 			if ((off + sizeof(struct in_addr)) > optlen) {
1108 				/*
1109 				 * End of source route.  Should be for us.
1110 				 */
1111 				save_rte(m, cp, ip->ip_src);
1112 				break;
1113 			}
1114 
1115 			/*
1116 			 * locate outgoing interface
1117 			 */
1118 			memset(&ipaddr, 0, sizeof(ipaddr));
1119 			ipaddr.sin_family = AF_INET;
1120 			ipaddr.sin_len = sizeof(ipaddr);
1121 			memcpy(&ipaddr.sin_addr, cp + off,
1122 			    sizeof(ipaddr.sin_addr));
1123 			/* keep packet in the virtual instance */
1124 			rt = rtalloc(sintosa(&ipaddr), RT_RESOLVE, rtableid);
1125 			if (!rtisvalid(rt) || ((opt == IPOPT_SSRR) &&
1126 			    ISSET(rt->rt_flags, RTF_GATEWAY))) {
1127 				type = ICMP_UNREACH;
1128 				code = ICMP_UNREACH_SRCFAIL;
1129 				rtfree(rt);
1130 				goto bad;
1131 			}
1132 			ia = ifatoia(rt->rt_ifa);
1133 			memcpy(cp + off, &ia->ia_addr.sin_addr,
1134 			    sizeof(struct in_addr));
1135 			rtfree(rt);
1136 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1137 			ip->ip_dst = ipaddr.sin_addr;
1138 			/*
1139 			 * Let ip_intr's mcast routing check handle mcast pkts
1140 			 */
1141 			forward = !IN_MULTICAST(ip->ip_dst.s_addr);
1142 			break;
1143 
1144 		case IPOPT_RR:
1145 			if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1146 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1147 				goto bad;
1148 			}
1149 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1150 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1151 				goto bad;
1152 			}
1153 
1154 			/*
1155 			 * If no space remains, ignore.
1156 			 */
1157 			off--;			/* 0 origin */
1158 			if ((off + sizeof(struct in_addr)) > optlen)
1159 				break;
1160 			memset(&ipaddr, 0, sizeof(ipaddr));
1161 			ipaddr.sin_family = AF_INET;
1162 			ipaddr.sin_len = sizeof(ipaddr);
1163 			ipaddr.sin_addr = ip->ip_dst;
1164 			/*
1165 			 * locate outgoing interface; if we're the destination,
1166 			 * use the incoming interface (should be same).
1167 			 * Again keep the packet inside the virtual instance.
1168 			 */
1169 			rt = rtalloc(sintosa(&ipaddr), RT_RESOLVE, rtableid);
1170 			if (!rtisvalid(rt)) {
1171 				type = ICMP_UNREACH;
1172 				code = ICMP_UNREACH_HOST;
1173 				rtfree(rt);
1174 				goto bad;
1175 			}
1176 			ia = ifatoia(rt->rt_ifa);
1177 			memcpy(cp + off, &ia->ia_addr.sin_addr,
1178 			    sizeof(struct in_addr));
1179 			rtfree(rt);
1180 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1181 			break;
1182 
1183 		case IPOPT_TS:
1184 			code = cp - (u_char *)ip;
1185 			if (optlen < sizeof(struct ip_timestamp))
1186 				goto bad;
1187 			memcpy(&ipt, cp, sizeof(struct ip_timestamp));
1188 			if (ipt.ipt_ptr < 5 || ipt.ipt_len < 5)
1189 				goto bad;
1190 			if (ipt.ipt_ptr - 1 + sizeof(u_int32_t) > ipt.ipt_len) {
1191 				if (++ipt.ipt_oflw == 0)
1192 					goto bad;
1193 				break;
1194 			}
1195 			memcpy(&sin, cp + ipt.ipt_ptr - 1, sizeof sin);
1196 			switch (ipt.ipt_flg) {
1197 
1198 			case IPOPT_TS_TSONLY:
1199 				break;
1200 
1201 			case IPOPT_TS_TSANDADDR:
1202 				if (ipt.ipt_ptr - 1 + sizeof(u_int32_t) +
1203 				    sizeof(struct in_addr) > ipt.ipt_len)
1204 					goto bad;
1205 				memset(&ipaddr, 0, sizeof(ipaddr));
1206 				ipaddr.sin_family = AF_INET;
1207 				ipaddr.sin_len = sizeof(ipaddr);
1208 				ipaddr.sin_addr = dst;
1209 				ia = ifatoia(ifaof_ifpforaddr(sintosa(&ipaddr),
1210 				    ifp));
1211 				if (ia == NULL)
1212 					continue;
1213 				memcpy(&sin, &ia->ia_addr.sin_addr,
1214 				    sizeof(struct in_addr));
1215 				ipt.ipt_ptr += sizeof(struct in_addr);
1216 				break;
1217 
1218 			case IPOPT_TS_PRESPEC:
1219 				if (ipt.ipt_ptr - 1 + sizeof(u_int32_t) +
1220 				    sizeof(struct in_addr) > ipt.ipt_len)
1221 					goto bad;
1222 				memset(&ipaddr, 0, sizeof(ipaddr));
1223 				ipaddr.sin_family = AF_INET;
1224 				ipaddr.sin_len = sizeof(ipaddr);
1225 				ipaddr.sin_addr = sin;
1226 				if (ifa_ifwithaddr(sintosa(&ipaddr),
1227 				    m->m_pkthdr.ph_rtableid) == NULL)
1228 					continue;
1229 				ipt.ipt_ptr += sizeof(struct in_addr);
1230 				break;
1231 
1232 			default:
1233 				/* XXX can't take &ipt->ipt_flg */
1234 				code = (u_char *)&ipt.ipt_ptr -
1235 				    (u_char *)ip + 1;
1236 				goto bad;
1237 			}
1238 			ntime = iptime();
1239 			memcpy(cp + ipt.ipt_ptr - 1, &ntime, sizeof(u_int32_t));
1240 			ipt.ipt_ptr += sizeof(u_int32_t);
1241 		}
1242 	}
1243 	KERNEL_UNLOCK();
1244 	if (forward && ipforwarding) {
1245 		ip_forward(m, ifp, NULL, 1);
1246 		return (1);
1247 	}
1248 	return (0);
1249 bad:
1250 	KERNEL_UNLOCK();
1251 	icmp_error(m, type, code, 0, 0);
1252 	ipstat_inc(ips_badoptions);
1253 	return (1);
1254 }
1255 
1256 /*
1257  * Save incoming source route for use in replies,
1258  * to be picked up later by ip_srcroute if the receiver is interested.
1259  */
1260 void
1261 save_rte(struct mbuf *m, u_char *option, struct in_addr dst)
1262 {
1263 	struct ip_srcrt *isr;
1264 	struct m_tag *mtag;
1265 	unsigned olen;
1266 
1267 	olen = option[IPOPT_OLEN];
1268 	if (olen > sizeof(isr->isr_hdr) + sizeof(isr->isr_routes))
1269 		return;
1270 
1271 	mtag = m_tag_get(PACKET_TAG_SRCROUTE, sizeof(*isr), M_NOWAIT);
1272 	if (mtag == NULL)
1273 		return;
1274 	isr = (struct ip_srcrt *)(mtag + 1);
1275 
1276 	memcpy(isr->isr_hdr, option, olen);
1277 	isr->isr_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1278 	isr->isr_dst = dst;
1279 	m_tag_prepend(m, mtag);
1280 }
1281 
1282 /*
1283  * Retrieve incoming source route for use in replies,
1284  * in the same form used by setsockopt.
1285  * The first hop is placed before the options, will be removed later.
1286  */
1287 struct mbuf *
1288 ip_srcroute(struct mbuf *m0)
1289 {
1290 	struct in_addr *p, *q;
1291 	struct mbuf *m;
1292 	struct ip_srcrt *isr;
1293 	struct m_tag *mtag;
1294 
1295 	if (!ip_dosourceroute)
1296 		return (NULL);
1297 
1298 	mtag = m_tag_find(m0, PACKET_TAG_SRCROUTE, NULL);
1299 	if (mtag == NULL)
1300 		return (NULL);
1301 	isr = (struct ip_srcrt *)(mtag + 1);
1302 
1303 	if (isr->isr_nhops == 0)
1304 		return (NULL);
1305 	m = m_get(M_DONTWAIT, MT_SOOPTS);
1306 	if (m == NULL)
1307 		return (NULL);
1308 
1309 #define OPTSIZ	(sizeof(isr->isr_nop) + sizeof(isr->isr_hdr))
1310 
1311 	/* length is (nhops+1)*sizeof(addr) + sizeof(nop + header) */
1312 	m->m_len = (isr->isr_nhops + 1) * sizeof(struct in_addr) + OPTSIZ;
1313 
1314 	/*
1315 	 * First save first hop for return route
1316 	 */
1317 	p = &(isr->isr_routes[isr->isr_nhops - 1]);
1318 	*(mtod(m, struct in_addr *)) = *p--;
1319 
1320 	/*
1321 	 * Copy option fields and padding (nop) to mbuf.
1322 	 */
1323 	isr->isr_nop = IPOPT_NOP;
1324 	isr->isr_hdr[IPOPT_OFFSET] = IPOPT_MINOFF;
1325 	memcpy(mtod(m, caddr_t) + sizeof(struct in_addr), &isr->isr_nop,
1326 	    OPTSIZ);
1327 	q = (struct in_addr *)(mtod(m, caddr_t) +
1328 	    sizeof(struct in_addr) + OPTSIZ);
1329 #undef OPTSIZ
1330 	/*
1331 	 * Record return path as an IP source route,
1332 	 * reversing the path (pointers are now aligned).
1333 	 */
1334 	while (p >= isr->isr_routes) {
1335 		*q++ = *p--;
1336 	}
1337 	/*
1338 	 * Last hop goes to final destination.
1339 	 */
1340 	*q = isr->isr_dst;
1341 	m_tag_delete(m0, (struct m_tag *)isr);
1342 	return (m);
1343 }
1344 
1345 /*
1346  * Strip out IP options, at higher level protocol in the kernel.
1347  */
1348 void
1349 ip_stripoptions(struct mbuf *m)
1350 {
1351 	int i;
1352 	struct ip *ip = mtod(m, struct ip *);
1353 	caddr_t opts;
1354 	int olen;
1355 
1356 	olen = (ip->ip_hl<<2) - sizeof (struct ip);
1357 	opts = (caddr_t)(ip + 1);
1358 	i = m->m_len - (sizeof (struct ip) + olen);
1359 	memmove(opts, opts  + olen, i);
1360 	m->m_len -= olen;
1361 	if (m->m_flags & M_PKTHDR)
1362 		m->m_pkthdr.len -= olen;
1363 	ip->ip_hl = sizeof(struct ip) >> 2;
1364 	ip->ip_len = htons(ntohs(ip->ip_len) - olen);
1365 }
1366 
1367 int inetctlerrmap[PRC_NCMDS] = {
1368 	0,		0,		0,		0,
1369 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1370 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1371 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1372 	0,		0,		0,		0,
1373 	ENOPROTOOPT
1374 };
1375 
1376 /*
1377  * Forward a packet.  If some error occurs return the sender
1378  * an icmp packet.  Note we can't always generate a meaningful
1379  * icmp message because icmp doesn't have a large enough repertoire
1380  * of codes and types.
1381  *
1382  * If not forwarding, just drop the packet.  This could be confusing
1383  * if ipforwarding was zero but some routing protocol was advancing
1384  * us as a gateway to somewhere.  However, we must let the routing
1385  * protocol deal with that.
1386  *
1387  * The srcrt parameter indicates whether the packet is being forwarded
1388  * via a source route.
1389  */
1390 void
1391 ip_forward(struct mbuf *m, struct ifnet *ifp, struct rtentry *rt, int srcrt)
1392 {
1393 	struct mbuf mfake, *mcopy = NULL;
1394 	struct ip *ip = mtod(m, struct ip *);
1395 	struct sockaddr_in *sin;
1396 	struct route ro;
1397 	int error, type = 0, code = 0, destmtu = 0, fake = 0, len;
1398 	u_int32_t dest;
1399 
1400 	dest = 0;
1401 	if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
1402 		ipstat_inc(ips_cantforward);
1403 		m_freem(m);
1404 		goto freecopy;
1405 	}
1406 	if (ip->ip_ttl <= IPTTLDEC) {
1407 		icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0);
1408 		goto freecopy;
1409 	}
1410 
1411 	sin = satosin(&ro.ro_dst);
1412 	memset(sin, 0, sizeof(*sin));
1413 	sin->sin_family = AF_INET;
1414 	sin->sin_len = sizeof(*sin);
1415 	sin->sin_addr = ip->ip_dst;
1416 
1417 	if (!rtisvalid(rt)) {
1418 		rtfree(rt);
1419 		rt = rtalloc_mpath(sintosa(sin), &ip->ip_src.s_addr,
1420 		    m->m_pkthdr.ph_rtableid);
1421 		if (rt == NULL) {
1422 			icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0);
1423 			return;
1424 		}
1425 	}
1426 
1427 	/*
1428 	 * Save at most 68 bytes of the packet in case
1429 	 * we need to generate an ICMP message to the src.
1430 	 * The data is saved in the mbuf on the stack that
1431 	 * acts as a temporary storage not intended to be
1432 	 * passed down the IP stack or to the mfree.
1433 	 */
1434 	memset(&mfake.m_hdr, 0, sizeof(mfake.m_hdr));
1435 	mfake.m_type = m->m_type;
1436 	if (m_dup_pkthdr(&mfake, m, M_DONTWAIT) == 0) {
1437 		mfake.m_data = mfake.m_pktdat;
1438 		len = min(ntohs(ip->ip_len), 68);
1439 		m_copydata(m, 0, len, mfake.m_pktdat);
1440 		mfake.m_pkthdr.len = mfake.m_len = len;
1441 #if NPF > 0
1442 		pf_pkt_unlink_state_key(&mfake);
1443 #endif	/* NPF > 0 */
1444 		fake = 1;
1445 	}
1446 
1447 	ip->ip_ttl -= IPTTLDEC;
1448 
1449 	/*
1450 	 * If forwarding packet using same interface that it came in on,
1451 	 * perhaps should send a redirect to sender to shortcut a hop.
1452 	 * Only send redirect if source is sending directly to us,
1453 	 * and if packet was not source routed (or has any options).
1454 	 * Also, don't send redirect if forwarding using a default route
1455 	 * or a route modified by a redirect.
1456 	 * Don't send redirect if we advertise destination's arp address
1457 	 * as ours (proxy arp).
1458 	 */
1459 	if ((rt->rt_ifidx == ifp->if_index) &&
1460 	    (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1461 	    satosin(rt_key(rt))->sin_addr.s_addr != 0 &&
1462 	    ipsendredirects && !srcrt &&
1463 	    !arpproxy(satosin(rt_key(rt))->sin_addr, m->m_pkthdr.ph_rtableid)) {
1464 		if ((ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_netmask) ==
1465 		    ifatoia(rt->rt_ifa)->ia_net) {
1466 		    if (rt->rt_flags & RTF_GATEWAY)
1467 			dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
1468 		    else
1469 			dest = ip->ip_dst.s_addr;
1470 		    /* Router requirements says to only send host redirects */
1471 		    type = ICMP_REDIRECT;
1472 		    code = ICMP_REDIRECT_HOST;
1473 		}
1474 	}
1475 
1476 	ro.ro_rt = rt;
1477 	ro.ro_tableid = m->m_pkthdr.ph_rtableid;
1478 	error = ip_output(m, NULL, &ro,
1479 	    (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)),
1480 	    NULL, NULL, 0);
1481 	rt = ro.ro_rt;
1482 	if (error)
1483 		ipstat_inc(ips_cantforward);
1484 	else {
1485 		ipstat_inc(ips_forward);
1486 		if (type)
1487 			ipstat_inc(ips_redirectsent);
1488 		else
1489 			goto freecopy;
1490 	}
1491 	if (!fake)
1492 		goto freecopy;
1493 
1494 	switch (error) {
1495 
1496 	case 0:				/* forwarded, but need redirect */
1497 		/* type, code set above */
1498 		break;
1499 
1500 	case ENETUNREACH:		/* shouldn't happen, checked above */
1501 	case EHOSTUNREACH:
1502 	case ENETDOWN:
1503 	case EHOSTDOWN:
1504 	default:
1505 		type = ICMP_UNREACH;
1506 		code = ICMP_UNREACH_HOST;
1507 		break;
1508 
1509 	case EMSGSIZE:
1510 		type = ICMP_UNREACH;
1511 		code = ICMP_UNREACH_NEEDFRAG;
1512 
1513 #ifdef IPSEC
1514 		if (rt != NULL) {
1515 			if (rt->rt_rmx.rmx_mtu)
1516 				destmtu = rt->rt_rmx.rmx_mtu;
1517 			else {
1518 				struct ifnet *destifp;
1519 
1520 				destifp = if_get(rt->rt_ifidx);
1521 				if (destifp != NULL)
1522 					destmtu = destifp->if_mtu;
1523 				if_put(destifp);
1524 			}
1525 		}
1526 #endif /*IPSEC*/
1527 		ipstat_inc(ips_cantfrag);
1528 		break;
1529 
1530 	case EACCES:
1531 		/*
1532 		 * pf(4) blocked the packet. There is no need to send an ICMP
1533 		 * packet back since pf(4) takes care of it.
1534 		 */
1535 		goto freecopy;
1536 	case ENOBUFS:
1537 		/*
1538 		 * a router should not generate ICMP_SOURCEQUENCH as
1539 		 * required in RFC1812 Requirements for IP Version 4 Routers.
1540 		 * source quench could be a big problem under DoS attacks,
1541 		 * or the underlying interface is rate-limited.
1542 		 */
1543 		goto freecopy;
1544 	}
1545 
1546 	mcopy = m_copym(&mfake, 0, len, M_DONTWAIT);
1547 	if (mcopy)
1548 		icmp_error(mcopy, type, code, dest, destmtu);
1549 
1550 freecopy:
1551 	if (fake)
1552 		m_tag_delete_chain(&mfake);
1553 	rtfree(rt);
1554 }
1555 
1556 int
1557 ip_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp, void *newp,
1558     size_t newlen)
1559 {
1560 	int error;
1561 #ifdef MROUTING
1562 	extern int ip_mrtproto;
1563 	extern struct mrtstat mrtstat;
1564 #endif
1565 
1566 	NET_ASSERT_LOCKED();
1567 
1568 	/* Almost all sysctl names at this level are terminal. */
1569 	if (namelen != 1 && name[0] != IPCTL_IFQUEUE)
1570 		return (ENOTDIR);
1571 
1572 	switch (name[0]) {
1573 #ifdef notyet
1574 	case IPCTL_DEFMTU:
1575 		return (sysctl_int(oldp, oldlenp, newp, newlen, &ip_mtu));
1576 #endif
1577 	case IPCTL_SOURCEROUTE:
1578 		/*
1579 		 * Don't allow this to change in a secure environment.
1580 		 */
1581 		if (newp && securelevel > 0)
1582 			return (EPERM);
1583 		return (sysctl_int(oldp, oldlenp, newp, newlen,
1584 		    &ip_dosourceroute));
1585 	case IPCTL_MTUDISC:
1586 		error = sysctl_int(oldp, oldlenp, newp, newlen,
1587 		    &ip_mtudisc);
1588 		if (ip_mtudisc != 0 && ip_mtudisc_timeout_q == NULL) {
1589 			ip_mtudisc_timeout_q =
1590 			    rt_timer_queue_create(ip_mtudisc_timeout);
1591 		} else if (ip_mtudisc == 0 && ip_mtudisc_timeout_q != NULL) {
1592 			rt_timer_queue_destroy(ip_mtudisc_timeout_q);
1593 			ip_mtudisc_timeout_q = NULL;
1594 		}
1595 		return error;
1596 	case IPCTL_MTUDISCTIMEOUT:
1597 		error = sysctl_int(oldp, oldlenp, newp, newlen,
1598 		   &ip_mtudisc_timeout);
1599 		if (ip_mtudisc_timeout_q != NULL)
1600 			rt_timer_queue_change(ip_mtudisc_timeout_q,
1601 					      ip_mtudisc_timeout);
1602 		return (error);
1603 	case IPCTL_IPSEC_ENC_ALGORITHM:
1604 	        return (sysctl_tstring(oldp, oldlenp, newp, newlen,
1605 				       ipsec_def_enc, sizeof(ipsec_def_enc)));
1606 	case IPCTL_IPSEC_AUTH_ALGORITHM:
1607 	        return (sysctl_tstring(oldp, oldlenp, newp, newlen,
1608 				       ipsec_def_auth,
1609 				       sizeof(ipsec_def_auth)));
1610 	case IPCTL_IPSEC_IPCOMP_ALGORITHM:
1611 	        return (sysctl_tstring(oldp, oldlenp, newp, newlen,
1612 				       ipsec_def_comp,
1613 				       sizeof(ipsec_def_comp)));
1614 	case IPCTL_IFQUEUE:
1615 	        return (sysctl_niq(name + 1, namelen - 1,
1616 		    oldp, oldlenp, newp, newlen, &ipintrq));
1617 	case IPCTL_STATS:
1618 		return (ip_sysctl_ipstat(oldp, oldlenp, newp));
1619 #ifdef MROUTING
1620 	case IPCTL_MRTSTATS:
1621 		return (sysctl_rdstruct(oldp, oldlenp, newp,
1622 		    &mrtstat, sizeof(mrtstat)));
1623 	case IPCTL_MRTPROTO:
1624 		return (sysctl_rdint(oldp, oldlenp, newp, ip_mrtproto));
1625 	case IPCTL_MRTMFC:
1626 		if (newp)
1627 			return (EPERM);
1628 		return mrt_sysctl_mfc(oldp, oldlenp);
1629 	case IPCTL_MRTVIF:
1630 		if (newp)
1631 			return (EPERM);
1632 		return mrt_sysctl_vif(oldp, oldlenp);
1633 #else
1634 	case IPCTL_MRTPROTO:
1635 	case IPCTL_MRTSTATS:
1636 	case IPCTL_MRTMFC:
1637 	case IPCTL_MRTVIF:
1638 		return (EOPNOTSUPP);
1639 #endif
1640 	default:
1641 		if (name[0] < IPCTL_MAXID)
1642 			return (sysctl_int_arr(ipctl_vars, name, namelen,
1643 			    oldp, oldlenp, newp, newlen));
1644 		return (EOPNOTSUPP);
1645 	}
1646 	/* NOTREACHED */
1647 }
1648 
1649 int
1650 ip_sysctl_ipstat(void *oldp, size_t *oldlenp, void *newp)
1651 {
1652 	uint64_t counters[ips_ncounters];
1653 	struct ipstat ipstat;
1654 	u_long *words = (u_long *)&ipstat;
1655 	int i;
1656 
1657 	KASSERT(sizeof(ipstat) == (nitems(counters) * sizeof(u_long)));
1658 
1659 	counters_read(ipcounters, counters, nitems(counters));
1660 
1661 	for (i = 0; i < nitems(counters); i++)
1662 		words[i] = (u_long)counters[i];
1663 
1664 	return (sysctl_rdstruct(oldp, oldlenp, newp, &ipstat, sizeof(ipstat)));
1665 }
1666 
1667 void
1668 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1669     struct mbuf *m)
1670 {
1671 #ifdef SO_TIMESTAMP
1672 	if (inp->inp_socket->so_options & SO_TIMESTAMP) {
1673 		struct timeval tv;
1674 
1675 		microtime(&tv);
1676 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1677 		    SCM_TIMESTAMP, SOL_SOCKET);
1678 		if (*mp)
1679 			mp = &(*mp)->m_next;
1680 	}
1681 #endif
1682 	if (inp->inp_flags & INP_RECVDSTADDR) {
1683 		*mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
1684 		    sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1685 		if (*mp)
1686 			mp = &(*mp)->m_next;
1687 	}
1688 #ifdef notyet
1689 	/* this code is broken and will probably never be fixed. */
1690 	/* options were tossed already */
1691 	if (inp->inp_flags & INP_RECVOPTS) {
1692 		*mp = sbcreatecontrol((caddr_t) opts_deleted_above,
1693 		    sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1694 		if (*mp)
1695 			mp = &(*mp)->m_next;
1696 	}
1697 	/* ip_srcroute doesn't do what we want here, need to fix */
1698 	if (inp->inp_flags & INP_RECVRETOPTS) {
1699 		*mp = sbcreatecontrol((caddr_t) ip_srcroute(m),
1700 		    sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1701 		if (*mp)
1702 			mp = &(*mp)->m_next;
1703 	}
1704 #endif
1705 	if (inp->inp_flags & INP_RECVIF) {
1706 		struct sockaddr_dl sdl;
1707 		struct ifnet *ifp;
1708 
1709 		ifp = if_get(m->m_pkthdr.ph_ifidx);
1710 		if (ifp == NULL || ifp->if_sadl == NULL) {
1711 			memset(&sdl, 0, sizeof(sdl));
1712 			sdl.sdl_len = offsetof(struct sockaddr_dl, sdl_data[0]);
1713 			sdl.sdl_family = AF_LINK;
1714 			sdl.sdl_index = ifp != NULL ? ifp->if_index : 0;
1715 			sdl.sdl_nlen = sdl.sdl_alen = sdl.sdl_slen = 0;
1716 			*mp = sbcreatecontrol((caddr_t) &sdl, sdl.sdl_len,
1717 			    IP_RECVIF, IPPROTO_IP);
1718 		} else {
1719 			*mp = sbcreatecontrol((caddr_t) ifp->if_sadl,
1720 			    ifp->if_sadl->sdl_len, IP_RECVIF, IPPROTO_IP);
1721 		}
1722 		if (*mp)
1723 			mp = &(*mp)->m_next;
1724 		if_put(ifp);
1725 	}
1726 	if (inp->inp_flags & INP_RECVTTL) {
1727 		*mp = sbcreatecontrol((caddr_t) &ip->ip_ttl,
1728 		    sizeof(u_int8_t), IP_RECVTTL, IPPROTO_IP);
1729 		if (*mp)
1730 			mp = &(*mp)->m_next;
1731 	}
1732 	if (inp->inp_flags & INP_RECVRTABLE) {
1733 		u_int rtableid = inp->inp_rtableid;
1734 #if NPF > 0
1735 		struct pf_divert *divert;
1736 
1737 		if (m && m->m_pkthdr.pf.flags & PF_TAG_DIVERTED &&
1738 		    (divert = pf_find_divert(m)) != NULL)
1739 			rtableid = divert->rdomain;
1740 #endif
1741 
1742 		*mp = sbcreatecontrol((caddr_t) &rtableid,
1743 		    sizeof(u_int), IP_RECVRTABLE, IPPROTO_IP);
1744 		if (*mp)
1745 			mp = &(*mp)->m_next;
1746 	}
1747 }
1748 
1749 void
1750 ip_send_dispatch(void *xmq)
1751 {
1752 	struct mbuf_queue *mq = xmq;
1753 	struct mbuf *m;
1754 	struct mbuf_list ml;
1755 	int s;
1756 
1757 	mq_delist(mq, &ml);
1758 	if (ml_empty(&ml))
1759 		return;
1760 
1761 	KERNEL_LOCK();
1762 	NET_LOCK(s);
1763 	while ((m = ml_dequeue(&ml)) != NULL) {
1764 		ip_output(m, NULL, NULL, 0, NULL, NULL, 0);
1765 	}
1766 	NET_UNLOCK(s);
1767 	KERNEL_UNLOCK();
1768 }
1769 
1770 void
1771 ip_send(struct mbuf *m)
1772 {
1773 	mq_enqueue(&ipsend_mq, m);
1774 	task_add(softnettq, &ipsend_task);
1775 }
1776