ip_reass.c revision 149635
1/*-
2 * Copyright (c) 1982, 1986, 1988, 1993
3 *	The Regents of the University of California.  All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 *    notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 *    notice, this list of conditions and the following disclaimer in the
12 *    documentation and/or other materials provided with the distribution.
13 * 4. Neither the name of the University nor the names of its contributors
14 *    may be used to endorse or promote products derived from this software
15 *    without specific prior written permission.
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27 * SUCH DAMAGE.
28 *
29 *	@(#)ip_input.c	8.2 (Berkeley) 1/4/94
30 * $FreeBSD: head/sys/netinet/ip_input.c 149635 2005-08-30 16:35:27Z andre $
31 */
32
33#include "opt_bootp.h"
34#include "opt_ipfw.h"
35#include "opt_ipstealth.h"
36#include "opt_ipsec.h"
37#include "opt_mac.h"
38#include "opt_carp.h"
39
40#include <sys/param.h>
41#include <sys/systm.h>
42#include <sys/callout.h>
43#include <sys/mac.h>
44#include <sys/mbuf.h>
45#include <sys/malloc.h>
46#include <sys/domain.h>
47#include <sys/protosw.h>
48#include <sys/socket.h>
49#include <sys/time.h>
50#include <sys/kernel.h>
51#include <sys/syslog.h>
52#include <sys/sysctl.h>
53
54#include <net/pfil.h>
55#include <net/if.h>
56#include <net/if_types.h>
57#include <net/if_var.h>
58#include <net/if_dl.h>
59#include <net/route.h>
60#include <net/netisr.h>
61
62#include <netinet/in.h>
63#include <netinet/in_systm.h>
64#include <netinet/in_var.h>
65#include <netinet/ip.h>
66#include <netinet/in_pcb.h>
67#include <netinet/ip_var.h>
68#include <netinet/ip_icmp.h>
69#include <machine/in_cksum.h>
70#ifdef DEV_CARP
71#include <netinet/ip_carp.h>
72#endif
73
74#include <sys/socketvar.h>
75
76/* XXX: Temporary until ipfw_ether and ipfw_bridge are converted. */
77#include <netinet/ip_fw.h>
78#include <netinet/ip_dummynet.h>
79
80#ifdef IPSEC
81#include <netinet6/ipsec.h>
82#include <netkey/key.h>
83#endif
84
85#ifdef FAST_IPSEC
86#include <netipsec/ipsec.h>
87#include <netipsec/key.h>
88#endif
89
90int rsvp_on = 0;
91
92int	ipforwarding = 0;
93SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW,
94    &ipforwarding, 0, "Enable IP forwarding between interfaces");
95
96static int	ipsendredirects = 1; /* XXX */
97SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW,
98    &ipsendredirects, 0, "Enable sending IP redirects");
99
100int	ip_defttl = IPDEFTTL;
101SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW,
102    &ip_defttl, 0, "Maximum TTL on IP packets");
103
104static int	ip_dosourceroute = 0;
105SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW,
106    &ip_dosourceroute, 0, "Enable forwarding source routed IP packets");
107
108static int	ip_acceptsourceroute = 0;
109SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute,
110    CTLFLAG_RW, &ip_acceptsourceroute, 0,
111    "Enable accepting source routed IP packets");
112
113int		ip_doopts = 1;	/* 0 = ignore, 1 = process, 2 = reject */
114SYSCTL_INT(_net_inet_ip, OID_AUTO, process_options, CTLFLAG_RW,
115    &ip_doopts, 0, "Enable IP options processing ([LS]SRR, RR, TS)");
116
117static int	ip_keepfaith = 0;
118SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW,
119	&ip_keepfaith,	0,
120	"Enable packet capture for FAITH IPv4->IPv6 translater daemon");
121
122static int    nipq = 0;         /* total # of reass queues */
123static int    maxnipq;
124SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragpackets, CTLFLAG_RW,
125	&maxnipq, 0,
126	"Maximum number of IPv4 fragment reassembly queue entries");
127
128static int    maxfragsperpacket;
129SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW,
130	&maxfragsperpacket, 0,
131	"Maximum number of IPv4 fragments allowed per packet");
132
133static int	ip_sendsourcequench = 0;
134SYSCTL_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW,
135	&ip_sendsourcequench, 0,
136	"Enable the transmission of source quench packets");
137
138int	ip_do_randomid = 0;
139SYSCTL_INT(_net_inet_ip, OID_AUTO, random_id, CTLFLAG_RW,
140	&ip_do_randomid, 0,
141	"Assign random ip_id values");
142
143/*
144 * XXX - Setting ip_checkinterface mostly implements the receive side of
145 * the Strong ES model described in RFC 1122, but since the routing table
146 * and transmit implementation do not implement the Strong ES model,
147 * setting this to 1 results in an odd hybrid.
148 *
149 * XXX - ip_checkinterface currently must be disabled if you use ipnat
150 * to translate the destination address to another local interface.
151 *
152 * XXX - ip_checkinterface must be disabled if you add IP aliases
153 * to the loopback interface instead of the interface where the
154 * packets for those addresses are received.
155 */
156static int	ip_checkinterface = 0;
157SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW,
158    &ip_checkinterface, 0, "Verify packet arrives on correct interface");
159
160#ifdef DIAGNOSTIC
161static int	ipprintfs = 0;
162#endif
163
164struct pfil_head inet_pfil_hook;	/* Packet filter hooks */
165
166static struct	ifqueue ipintrq;
167static int	ipqmaxlen = IFQ_MAXLEN;
168
169extern	struct domain inetdomain;
170extern	struct protosw inetsw[];
171u_char	ip_protox[IPPROTO_MAX];
172struct	in_ifaddrhead in_ifaddrhead; 		/* first inet address */
173struct	in_ifaddrhashhead *in_ifaddrhashtbl;	/* inet addr hash table  */
174u_long 	in_ifaddrhmask;				/* mask for hash table */
175
176SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW,
177    &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue");
178SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD,
179    &ipintrq.ifq_drops, 0, "Number of packets dropped from the IP input queue");
180
181struct ipstat ipstat;
182SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW,
183    &ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)");
184
185/* Packet reassembly stuff */
186#define IPREASS_NHASH_LOG2      6
187#define IPREASS_NHASH           (1 << IPREASS_NHASH_LOG2)
188#define IPREASS_HMASK           (IPREASS_NHASH - 1)
189#define IPREASS_HASH(x,y) \
190	(((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK)
191
192static TAILQ_HEAD(ipqhead, ipq) ipq[IPREASS_NHASH];
193struct mtx ipqlock;
194struct callout ipport_tick_callout;
195
196#define	IPQ_LOCK()	mtx_lock(&ipqlock)
197#define	IPQ_UNLOCK()	mtx_unlock(&ipqlock)
198#define	IPQ_LOCK_INIT()	mtx_init(&ipqlock, "ipqlock", NULL, MTX_DEF)
199#define	IPQ_LOCK_ASSERT()	mtx_assert(&ipqlock, MA_OWNED)
200
201#ifdef IPCTL_DEFMTU
202SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
203    &ip_mtu, 0, "Default MTU");
204#endif
205
206#ifdef IPSTEALTH
207int	ipstealth = 0;
208SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW,
209    &ipstealth, 0, "");
210#endif
211
212/*
213 * ipfw_ether and ipfw_bridge hooks.
214 * XXX: Temporary until those are converted to pfil_hooks as well.
215 */
216ip_fw_chk_t *ip_fw_chk_ptr = NULL;
217ip_dn_io_t *ip_dn_io_ptr = NULL;
218int fw_enable = 1;
219int fw_one_pass = 1;
220
221/*
222 * XXX this is ugly.  IP options source routing magic.
223 */
224struct ipoptrt {
225	struct	in_addr dst;			/* final destination */
226	char	nop;				/* one NOP to align */
227	char	srcopt[IPOPT_OFFSET + 1];	/* OPTVAL, OLEN and OFFSET */
228	struct	in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)];
229};
230
231struct ipopt_tag {
232	struct	m_tag tag;
233	int	ip_nhops;
234	struct	ipoptrt ip_srcrt;
235};
236
237static void	save_rte(struct mbuf *, u_char *, struct in_addr);
238static int	ip_dooptions(struct mbuf *m, int);
239static void	ip_forward(struct mbuf *m, int srcrt);
240static void	ip_freef(struct ipqhead *, struct ipq *);
241
242/*
243 * IP initialization: fill in IP protocol switch table.
244 * All protocols not implemented in kernel go to raw IP protocol handler.
245 */
246void
247ip_init()
248{
249	register struct protosw *pr;
250	register int i;
251
252	TAILQ_INIT(&in_ifaddrhead);
253	in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask);
254	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
255	if (pr == NULL)
256		panic("ip_init: PF_INET not found");
257
258	/* Initialize the entire ip_protox[] array to IPPROTO_RAW. */
259	for (i = 0; i < IPPROTO_MAX; i++)
260		ip_protox[i] = pr - inetsw;
261	/*
262	 * Cycle through IP protocols and put them into the appropriate place
263	 * in ip_protox[].
264	 */
265	for (pr = inetdomain.dom_protosw;
266	    pr < inetdomain.dom_protoswNPROTOSW; pr++)
267		if (pr->pr_domain->dom_family == PF_INET &&
268		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
269			/* Be careful to only index valid IP protocols. */
270			if (pr->pr_protocol < IPPROTO_MAX)
271				ip_protox[pr->pr_protocol] = pr - inetsw;
272		}
273
274	/* Initialize packet filter hooks. */
275	inet_pfil_hook.ph_type = PFIL_TYPE_AF;
276	inet_pfil_hook.ph_af = AF_INET;
277	if ((i = pfil_head_register(&inet_pfil_hook)) != 0)
278		printf("%s: WARNING: unable to register pfil hook, "
279			"error %d\n", __func__, i);
280
281	/* Initialize IP reassembly queue. */
282	IPQ_LOCK_INIT();
283	for (i = 0; i < IPREASS_NHASH; i++)
284	    TAILQ_INIT(&ipq[i]);
285	maxnipq = nmbclusters / 32;
286	maxfragsperpacket = 16;
287
288	/* Start ipport_tick. */
289	callout_init(&ipport_tick_callout, CALLOUT_MPSAFE);
290	ipport_tick(NULL);
291	EVENTHANDLER_REGISTER(shutdown_pre_sync, ip_fini, NULL,
292		SHUTDOWN_PRI_DEFAULT);
293
294	/* Initialize various other remaining things. */
295	ip_id = time_second & 0xffff;
296	ipintrq.ifq_maxlen = ipqmaxlen;
297	mtx_init(&ipintrq.ifq_mtx, "ip_inq", NULL, MTX_DEF);
298	netisr_register(NETISR_IP, ip_input, &ipintrq, NETISR_MPSAFE);
299}
300
301void ip_fini(xtp)
302	void *xtp;
303{
304	callout_stop(&ipport_tick_callout);
305}
306
307/*
308 * Ip input routine.  Checksum and byte swap header.  If fragmented
309 * try to reassemble.  Process options.  Pass to next level.
310 */
311void
312ip_input(struct mbuf *m)
313{
314	struct ip *ip = NULL;
315	struct in_ifaddr *ia = NULL;
316	struct ifaddr *ifa;
317	int    checkif, hlen = 0;
318	u_short sum;
319	int dchg = 0;				/* dest changed after fw */
320	struct in_addr odst;			/* original dst address */
321#ifdef FAST_IPSEC
322	struct m_tag *mtag;
323	struct tdb_ident *tdbi;
324	struct secpolicy *sp;
325	int s, error;
326#endif /* FAST_IPSEC */
327
328  	M_ASSERTPKTHDR(m);
329
330	if (m->m_flags & M_FASTFWD_OURS) {
331		/*
332		 * Firewall or NAT changed destination to local.
333		 * We expect ip_len and ip_off to be in host byte order.
334		 */
335		m->m_flags &= ~M_FASTFWD_OURS;
336		/* Set up some basics that will be used later. */
337		ip = mtod(m, struct ip *);
338		hlen = ip->ip_hl << 2;
339  		goto ours;
340  	}
341
342	ipstat.ips_total++;
343
344	if (m->m_pkthdr.len < sizeof(struct ip))
345		goto tooshort;
346
347	if (m->m_len < sizeof (struct ip) &&
348	    (m = m_pullup(m, sizeof (struct ip))) == NULL) {
349		ipstat.ips_toosmall++;
350		return;
351	}
352	ip = mtod(m, struct ip *);
353
354	if (ip->ip_v != IPVERSION) {
355		ipstat.ips_badvers++;
356		goto bad;
357	}
358
359	hlen = ip->ip_hl << 2;
360	if (hlen < sizeof(struct ip)) {	/* minimum header length */
361		ipstat.ips_badhlen++;
362		goto bad;
363	}
364	if (hlen > m->m_len) {
365		if ((m = m_pullup(m, hlen)) == NULL) {
366			ipstat.ips_badhlen++;
367			return;
368		}
369		ip = mtod(m, struct ip *);
370	}
371
372	/* 127/8 must not appear on wire - RFC1122 */
373	if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
374	    (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
375		if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
376			ipstat.ips_badaddr++;
377			goto bad;
378		}
379	}
380
381	if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
382		sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
383	} else {
384		if (hlen == sizeof(struct ip)) {
385			sum = in_cksum_hdr(ip);
386		} else {
387			sum = in_cksum(m, hlen);
388		}
389	}
390	if (sum) {
391		ipstat.ips_badsum++;
392		goto bad;
393	}
394
395#ifdef ALTQ
396	if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
397		/* packet is dropped by traffic conditioner */
398		return;
399#endif
400
401	/*
402	 * Convert fields to host representation.
403	 */
404	ip->ip_len = ntohs(ip->ip_len);
405	if (ip->ip_len < hlen) {
406		ipstat.ips_badlen++;
407		goto bad;
408	}
409	ip->ip_off = ntohs(ip->ip_off);
410
411	/*
412	 * Check that the amount of data in the buffers
413	 * is as at least much as the IP header would have us expect.
414	 * Trim mbufs if longer than we expect.
415	 * Drop packet if shorter than we expect.
416	 */
417	if (m->m_pkthdr.len < ip->ip_len) {
418tooshort:
419		ipstat.ips_tooshort++;
420		goto bad;
421	}
422	if (m->m_pkthdr.len > ip->ip_len) {
423		if (m->m_len == m->m_pkthdr.len) {
424			m->m_len = ip->ip_len;
425			m->m_pkthdr.len = ip->ip_len;
426		} else
427			m_adj(m, ip->ip_len - m->m_pkthdr.len);
428	}
429#if defined(IPSEC) && !defined(IPSEC_FILTERGIF)
430	/*
431	 * Bypass packet filtering for packets from a tunnel (gif).
432	 */
433	if (ipsec_getnhist(m))
434		goto passin;
435#endif
436#if defined(FAST_IPSEC) && !defined(IPSEC_FILTERGIF)
437	/*
438	 * Bypass packet filtering for packets from a tunnel (gif).
439	 */
440	if (m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL)
441		goto passin;
442#endif
443
444	/*
445	 * Run through list of hooks for input packets.
446	 *
447	 * NB: Beware of the destination address changing (e.g.
448	 *     by NAT rewriting).  When this happens, tell
449	 *     ip_forward to do the right thing.
450	 */
451
452	/* Jump over all PFIL processing if hooks are not active. */
453	if (inet_pfil_hook.ph_busy_count == -1)
454		goto passin;
455
456	odst = ip->ip_dst;
457	if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif,
458	    PFIL_IN, NULL) != 0)
459		return;
460	if (m == NULL)			/* consumed by filter */
461		return;
462
463	ip = mtod(m, struct ip *);
464	dchg = (odst.s_addr != ip->ip_dst.s_addr);
465
466#ifdef IPFIREWALL_FORWARD
467	if (m->m_flags & M_FASTFWD_OURS) {
468		m->m_flags &= ~M_FASTFWD_OURS;
469		goto ours;
470	}
471#ifndef IPFIREWALL_FORWARD_EXTENDED
472	dchg = (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL);
473#else
474	if ((dchg = (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL)) != 0) {
475		/*
476		 * Directly ship on the packet.  This allows to forward packets
477		 * that were destined for us to some other directly connected
478		 * host.
479		 */
480		ip_forward(m, dchg);
481		return;
482	}
483#endif /* IPFIREWALL_FORWARD_EXTENDED */
484#endif /* IPFIREWALL_FORWARD */
485
486passin:
487	/*
488	 * Process options and, if not destined for us,
489	 * ship it on.  ip_dooptions returns 1 when an
490	 * error was detected (causing an icmp message
491	 * to be sent and the original packet to be freed).
492	 */
493	if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
494		return;
495
496        /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
497         * matter if it is destined to another node, or whether it is
498         * a multicast one, RSVP wants it! and prevents it from being forwarded
499         * anywhere else. Also checks if the rsvp daemon is running before
500	 * grabbing the packet.
501         */
502	if (rsvp_on && ip->ip_p==IPPROTO_RSVP)
503		goto ours;
504
505	/*
506	 * Check our list of addresses, to see if the packet is for us.
507	 * If we don't have any addresses, assume any unicast packet
508	 * we receive might be for us (and let the upper layers deal
509	 * with it).
510	 */
511	if (TAILQ_EMPTY(&in_ifaddrhead) &&
512	    (m->m_flags & (M_MCAST|M_BCAST)) == 0)
513		goto ours;
514
515	/*
516	 * Enable a consistency check between the destination address
517	 * and the arrival interface for a unicast packet (the RFC 1122
518	 * strong ES model) if IP forwarding is disabled and the packet
519	 * is not locally generated and the packet is not subject to
520	 * 'ipfw fwd'.
521	 *
522	 * XXX - Checking also should be disabled if the destination
523	 * address is ipnat'ed to a different interface.
524	 *
525	 * XXX - Checking is incompatible with IP aliases added
526	 * to the loopback interface instead of the interface where
527	 * the packets are received.
528	 *
529	 * XXX - This is the case for carp vhost IPs as well so we
530	 * insert a workaround. If the packet got here, we already
531	 * checked with carp_iamatch() and carp_forus().
532	 */
533	checkif = ip_checkinterface && (ipforwarding == 0) &&
534	    m->m_pkthdr.rcvif != NULL &&
535	    ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) &&
536#ifdef DEV_CARP
537	    !m->m_pkthdr.rcvif->if_carp &&
538#endif
539	    (dchg == 0);
540
541	/*
542	 * Check for exact addresses in the hash bucket.
543	 */
544	LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
545		/*
546		 * If the address matches, verify that the packet
547		 * arrived via the correct interface if checking is
548		 * enabled.
549		 */
550		if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr &&
551		    (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif))
552			goto ours;
553	}
554	/*
555	 * Check for broadcast addresses.
556	 *
557	 * Only accept broadcast packets that arrive via the matching
558	 * interface.  Reception of forwarded directed broadcasts would
559	 * be handled via ip_forward() and ether_output() with the loopback
560	 * into the stack for SIMPLEX interfaces handled by ether_output().
561	 */
562	if (m->m_pkthdr.rcvif != NULL &&
563	    m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
564	        TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
565			if (ifa->ifa_addr->sa_family != AF_INET)
566				continue;
567			ia = ifatoia(ifa);
568			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
569			    ip->ip_dst.s_addr)
570				goto ours;
571			if (ia->ia_netbroadcast.s_addr == ip->ip_dst.s_addr)
572				goto ours;
573#ifdef BOOTP_COMPAT
574			if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY)
575				goto ours;
576#endif
577		}
578	}
579	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
580		struct in_multi *inm;
581		if (ip_mrouter) {
582			/*
583			 * If we are acting as a multicast router, all
584			 * incoming multicast packets are passed to the
585			 * kernel-level multicast forwarding function.
586			 * The packet is returned (relatively) intact; if
587			 * ip_mforward() returns a non-zero value, the packet
588			 * must be discarded, else it may be accepted below.
589			 */
590			if (ip_mforward &&
591			    ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) {
592				ipstat.ips_cantforward++;
593				m_freem(m);
594				return;
595			}
596
597			/*
598			 * The process-level routing daemon needs to receive
599			 * all multicast IGMP packets, whether or not this
600			 * host belongs to their destination groups.
601			 */
602			if (ip->ip_p == IPPROTO_IGMP)
603				goto ours;
604			ipstat.ips_forward++;
605		}
606		/*
607		 * See if we belong to the destination multicast group on the
608		 * arrival interface.
609		 */
610		IN_MULTI_LOCK();
611		IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
612		IN_MULTI_UNLOCK();
613		if (inm == NULL) {
614			ipstat.ips_notmember++;
615			m_freem(m);
616			return;
617		}
618		goto ours;
619	}
620	if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
621		goto ours;
622	if (ip->ip_dst.s_addr == INADDR_ANY)
623		goto ours;
624
625	/*
626	 * FAITH(Firewall Aided Internet Translator)
627	 */
628	if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
629		if (ip_keepfaith) {
630			if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP)
631				goto ours;
632		}
633		m_freem(m);
634		return;
635	}
636
637	/*
638	 * Not for us; forward if possible and desirable.
639	 */
640	if (ipforwarding == 0) {
641		ipstat.ips_cantforward++;
642		m_freem(m);
643	} else {
644#ifdef IPSEC
645		/*
646		 * Enforce inbound IPsec SPD.
647		 */
648		if (ipsec4_in_reject(m, NULL)) {
649			ipsecstat.in_polvio++;
650			goto bad;
651		}
652#endif /* IPSEC */
653#ifdef FAST_IPSEC
654		mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
655		s = splnet();
656		if (mtag != NULL) {
657			tdbi = (struct tdb_ident *)(mtag + 1);
658			sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
659		} else {
660			sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
661						   IP_FORWARDING, &error);
662		}
663		if (sp == NULL) {	/* NB: can happen if error */
664			splx(s);
665			/*XXX error stat???*/
666			DPRINTF(("ip_input: no SP for forwarding\n"));	/*XXX*/
667			goto bad;
668		}
669
670		/*
671		 * Check security policy against packet attributes.
672		 */
673		error = ipsec_in_reject(sp, m);
674		KEY_FREESP(&sp);
675		splx(s);
676		if (error) {
677			ipstat.ips_cantforward++;
678			goto bad;
679		}
680#endif /* FAST_IPSEC */
681		ip_forward(m, dchg);
682	}
683	return;
684
685ours:
686#ifdef IPSTEALTH
687	/*
688	 * IPSTEALTH: Process non-routing options only
689	 * if the packet is destined for us.
690	 */
691	if (ipstealth && hlen > sizeof (struct ip) &&
692	    ip_dooptions(m, 1))
693		return;
694#endif /* IPSTEALTH */
695
696	/* Count the packet in the ip address stats */
697	if (ia != NULL) {
698		ia->ia_ifa.if_ipackets++;
699		ia->ia_ifa.if_ibytes += m->m_pkthdr.len;
700	}
701
702	/*
703	 * Attempt reassembly; if it succeeds, proceed.
704	 * ip_reass() will return a different mbuf.
705	 */
706	if (ip->ip_off & (IP_MF | IP_OFFMASK)) {
707		m = ip_reass(m);
708		if (m == NULL)
709			return;
710		ip = mtod(m, struct ip *);
711		/* Get the header length of the reassembled packet */
712		hlen = ip->ip_hl << 2;
713	}
714
715	/*
716	 * Further protocols expect the packet length to be w/o the
717	 * IP header.
718	 */
719	ip->ip_len -= hlen;
720
721#ifdef IPSEC
722	/*
723	 * enforce IPsec policy checking if we are seeing last header.
724	 * note that we do not visit this with protocols with pcb layer
725	 * code - like udp/tcp/raw ip.
726	 */
727	if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0 &&
728	    ipsec4_in_reject(m, NULL)) {
729		ipsecstat.in_polvio++;
730		goto bad;
731	}
732#endif
733#if FAST_IPSEC
734	/*
735	 * enforce IPsec policy checking if we are seeing last header.
736	 * note that we do not visit this with protocols with pcb layer
737	 * code - like udp/tcp/raw ip.
738	 */
739	if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) {
740		/*
741		 * Check if the packet has already had IPsec processing
742		 * done.  If so, then just pass it along.  This tag gets
743		 * set during AH, ESP, etc. input handling, before the
744		 * packet is returned to the ip input queue for delivery.
745		 */
746		mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
747		s = splnet();
748		if (mtag != NULL) {
749			tdbi = (struct tdb_ident *)(mtag + 1);
750			sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
751		} else {
752			sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
753						   IP_FORWARDING, &error);
754		}
755		if (sp != NULL) {
756			/*
757			 * Check security policy against packet attributes.
758			 */
759			error = ipsec_in_reject(sp, m);
760			KEY_FREESP(&sp);
761		} else {
762			/* XXX error stat??? */
763			error = EINVAL;
764DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/
765			goto bad;
766		}
767		splx(s);
768		if (error)
769			goto bad;
770	}
771#endif /* FAST_IPSEC */
772
773	/*
774	 * Switch out to protocol's input routine.
775	 */
776	ipstat.ips_delivered++;
777
778	(*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen);
779	return;
780bad:
781	m_freem(m);
782}
783
784/*
785 * Take incoming datagram fragment and try to reassemble it into
786 * whole datagram.  If the argument is the first fragment or one
787 * in between the function will return NULL and store the mbuf
788 * in the fragment chain.  If the argument is the last fragment
789 * the packet will be reassembled and the pointer to the new
790 * mbuf returned for further processing.  Only m_tags attached
791 * to the first packet/fragment are preserved.
792 * The IP header is *NOT* adjusted out of iplen.
793 */
794
795struct mbuf *
796ip_reass(struct mbuf *m)
797{
798	struct ip *ip;
799	struct mbuf *p, *q, *nq, *t;
800	struct ipq *fp = NULL;
801	struct ipqhead *head;
802	int i, hlen, next;
803	u_int8_t ecn, ecn0;
804	u_short hash;
805
806	/* If maxnipq or maxfragsperpacket are 0, never accept fragments. */
807	if (maxnipq == 0 || maxfragsperpacket == 0) {
808		ipstat.ips_fragments++;
809		ipstat.ips_fragdropped++;
810		m_freem(m);
811		return (NULL);
812	}
813
814	ip = mtod(m, struct ip *);
815	hlen = ip->ip_hl << 2;
816
817	hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
818	head = &ipq[hash];
819	IPQ_LOCK();
820
821	/*
822	 * Look for queue of fragments
823	 * of this datagram.
824	 */
825	TAILQ_FOREACH(fp, head, ipq_list)
826		if (ip->ip_id == fp->ipq_id &&
827		    ip->ip_src.s_addr == fp->ipq_src.s_addr &&
828		    ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
829#ifdef MAC
830		    mac_fragment_match(m, fp) &&
831#endif
832		    ip->ip_p == fp->ipq_p)
833			goto found;
834
835	fp = NULL;
836
837	/*
838	 * Enforce upper bound on number of fragmented packets
839	 * for which we attempt reassembly;
840	 * If maxnipq is -1, accept all fragments without limitation.
841	 */
842	if ((nipq > maxnipq) && (maxnipq > 0)) {
843		/*
844		 * drop something from the tail of the current queue
845		 * before proceeding further
846		 */
847		struct ipq *q = TAILQ_LAST(head, ipqhead);
848		if (q == NULL) {   /* gak */
849			for (i = 0; i < IPREASS_NHASH; i++) {
850				struct ipq *r = TAILQ_LAST(&ipq[i], ipqhead);
851				if (r) {
852					ipstat.ips_fragtimeout += r->ipq_nfrags;
853					ip_freef(&ipq[i], r);
854					break;
855				}
856			}
857		} else {
858			ipstat.ips_fragtimeout += q->ipq_nfrags;
859			ip_freef(head, q);
860		}
861	}
862
863found:
864	/*
865	 * Adjust ip_len to not reflect header,
866	 * convert offset of this to bytes.
867	 */
868	ip->ip_len -= hlen;
869	if (ip->ip_off & IP_MF) {
870		/*
871		 * Make sure that fragments have a data length
872		 * that's a non-zero multiple of 8 bytes.
873		 */
874		if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
875			ipstat.ips_toosmall++; /* XXX */
876			goto dropfrag;
877		}
878		m->m_flags |= M_FRAG;
879	} else
880		m->m_flags &= ~M_FRAG;
881	ip->ip_off <<= 3;
882
883
884	/*
885	 * Attempt reassembly; if it succeeds, proceed.
886	 * ip_reass() will return a different mbuf.
887	 */
888	ipstat.ips_fragments++;
889	m->m_pkthdr.header = ip;
890
891	/* Previous ip_reass() started here. */
892	/*
893	 * Presence of header sizes in mbufs
894	 * would confuse code below.
895	 */
896	m->m_data += hlen;
897	m->m_len -= hlen;
898
899	/*
900	 * If first fragment to arrive, create a reassembly queue.
901	 */
902	if (fp == NULL) {
903		if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL)
904			goto dropfrag;
905		fp = mtod(t, struct ipq *);
906#ifdef MAC
907		if (mac_init_ipq(fp, M_NOWAIT) != 0) {
908			m_free(t);
909			goto dropfrag;
910		}
911		mac_create_ipq(m, fp);
912#endif
913		TAILQ_INSERT_HEAD(head, fp, ipq_list);
914		nipq++;
915		fp->ipq_nfrags = 1;
916		fp->ipq_ttl = IPFRAGTTL;
917		fp->ipq_p = ip->ip_p;
918		fp->ipq_id = ip->ip_id;
919		fp->ipq_src = ip->ip_src;
920		fp->ipq_dst = ip->ip_dst;
921		fp->ipq_frags = m;
922		m->m_nextpkt = NULL;
923		goto done;
924	} else {
925		fp->ipq_nfrags++;
926#ifdef MAC
927		mac_update_ipq(m, fp);
928#endif
929	}
930
931#define GETIP(m)	((struct ip*)((m)->m_pkthdr.header))
932
933	/*
934	 * Handle ECN by comparing this segment with the first one;
935	 * if CE is set, do not lose CE.
936	 * drop if CE and not-ECT are mixed for the same packet.
937	 */
938	ecn = ip->ip_tos & IPTOS_ECN_MASK;
939	ecn0 = GETIP(fp->ipq_frags)->ip_tos & IPTOS_ECN_MASK;
940	if (ecn == IPTOS_ECN_CE) {
941		if (ecn0 == IPTOS_ECN_NOTECT)
942			goto dropfrag;
943		if (ecn0 != IPTOS_ECN_CE)
944			GETIP(fp->ipq_frags)->ip_tos |= IPTOS_ECN_CE;
945	}
946	if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT)
947		goto dropfrag;
948
949	/*
950	 * Find a segment which begins after this one does.
951	 */
952	for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt)
953		if (GETIP(q)->ip_off > ip->ip_off)
954			break;
955
956	/*
957	 * If there is a preceding segment, it may provide some of
958	 * our data already.  If so, drop the data from the incoming
959	 * segment.  If it provides all of our data, drop us, otherwise
960	 * stick new segment in the proper place.
961	 *
962	 * If some of the data is dropped from the the preceding
963	 * segment, then it's checksum is invalidated.
964	 */
965	if (p) {
966		i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off;
967		if (i > 0) {
968			if (i >= ip->ip_len)
969				goto dropfrag;
970			m_adj(m, i);
971			m->m_pkthdr.csum_flags = 0;
972			ip->ip_off += i;
973			ip->ip_len -= i;
974		}
975		m->m_nextpkt = p->m_nextpkt;
976		p->m_nextpkt = m;
977	} else {
978		m->m_nextpkt = fp->ipq_frags;
979		fp->ipq_frags = m;
980	}
981
982	/*
983	 * While we overlap succeeding segments trim them or,
984	 * if they are completely covered, dequeue them.
985	 */
986	for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off;
987	     q = nq) {
988		i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off;
989		if (i < GETIP(q)->ip_len) {
990			GETIP(q)->ip_len -= i;
991			GETIP(q)->ip_off += i;
992			m_adj(q, i);
993			q->m_pkthdr.csum_flags = 0;
994			break;
995		}
996		nq = q->m_nextpkt;
997		m->m_nextpkt = nq;
998		ipstat.ips_fragdropped++;
999		fp->ipq_nfrags--;
1000		m_freem(q);
1001	}
1002
1003	/*
1004	 * Check for complete reassembly and perform frag per packet
1005	 * limiting.
1006	 *
1007	 * Frag limiting is performed here so that the nth frag has
1008	 * a chance to complete the packet before we drop the packet.
1009	 * As a result, n+1 frags are actually allowed per packet, but
1010	 * only n will ever be stored. (n = maxfragsperpacket.)
1011	 *
1012	 */
1013	next = 0;
1014	for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
1015		if (GETIP(q)->ip_off != next) {
1016			if (fp->ipq_nfrags > maxfragsperpacket) {
1017				ipstat.ips_fragdropped += fp->ipq_nfrags;
1018				ip_freef(head, fp);
1019			}
1020			goto done;
1021		}
1022		next += GETIP(q)->ip_len;
1023	}
1024	/* Make sure the last packet didn't have the IP_MF flag */
1025	if (p->m_flags & M_FRAG) {
1026		if (fp->ipq_nfrags > maxfragsperpacket) {
1027			ipstat.ips_fragdropped += fp->ipq_nfrags;
1028			ip_freef(head, fp);
1029		}
1030		goto done;
1031	}
1032
1033	/*
1034	 * Reassembly is complete.  Make sure the packet is a sane size.
1035	 */
1036	q = fp->ipq_frags;
1037	ip = GETIP(q);
1038	if (next + (ip->ip_hl << 2) > IP_MAXPACKET) {
1039		ipstat.ips_toolong++;
1040		ipstat.ips_fragdropped += fp->ipq_nfrags;
1041		ip_freef(head, fp);
1042		goto done;
1043	}
1044
1045	/*
1046	 * Concatenate fragments.
1047	 */
1048	m = q;
1049	t = m->m_next;
1050	m->m_next = NULL;
1051	m_cat(m, t);
1052	nq = q->m_nextpkt;
1053	q->m_nextpkt = NULL;
1054	for (q = nq; q != NULL; q = nq) {
1055		nq = q->m_nextpkt;
1056		q->m_nextpkt = NULL;
1057		m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
1058		m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
1059		m_cat(m, q);
1060	}
1061#ifdef MAC
1062	mac_create_datagram_from_ipq(fp, m);
1063	mac_destroy_ipq(fp);
1064#endif
1065
1066	/*
1067	 * Create header for new ip packet by modifying header of first
1068	 * packet;  dequeue and discard fragment reassembly header.
1069	 * Make header visible.
1070	 */
1071	ip->ip_len = (ip->ip_hl << 2) + next;
1072	ip->ip_src = fp->ipq_src;
1073	ip->ip_dst = fp->ipq_dst;
1074	TAILQ_REMOVE(head, fp, ipq_list);
1075	nipq--;
1076	(void) m_free(dtom(fp));
1077	m->m_len += (ip->ip_hl << 2);
1078	m->m_data -= (ip->ip_hl << 2);
1079	/* some debugging cruft by sklower, below, will go away soon */
1080	if (m->m_flags & M_PKTHDR)	/* XXX this should be done elsewhere */
1081		m_fixhdr(m);
1082	ipstat.ips_reassembled++;
1083	IPQ_UNLOCK();
1084	return (m);
1085
1086dropfrag:
1087	ipstat.ips_fragdropped++;
1088	if (fp != NULL)
1089		fp->ipq_nfrags--;
1090	m_freem(m);
1091done:
1092	IPQ_UNLOCK();
1093	return (NULL);
1094
1095#undef GETIP
1096}
1097
1098/*
1099 * Free a fragment reassembly header and all
1100 * associated datagrams.
1101 */
1102static void
1103ip_freef(fhp, fp)
1104	struct ipqhead *fhp;
1105	struct ipq *fp;
1106{
1107	register struct mbuf *q;
1108
1109	IPQ_LOCK_ASSERT();
1110
1111	while (fp->ipq_frags) {
1112		q = fp->ipq_frags;
1113		fp->ipq_frags = q->m_nextpkt;
1114		m_freem(q);
1115	}
1116	TAILQ_REMOVE(fhp, fp, ipq_list);
1117	(void) m_free(dtom(fp));
1118	nipq--;
1119}
1120
1121/*
1122 * IP timer processing;
1123 * if a timer expires on a reassembly
1124 * queue, discard it.
1125 */
1126void
1127ip_slowtimo()
1128{
1129	register struct ipq *fp;
1130	int i;
1131
1132	IPQ_LOCK();
1133	for (i = 0; i < IPREASS_NHASH; i++) {
1134		for(fp = TAILQ_FIRST(&ipq[i]); fp;) {
1135			struct ipq *fpp;
1136
1137			fpp = fp;
1138			fp = TAILQ_NEXT(fp, ipq_list);
1139			if(--fpp->ipq_ttl == 0) {
1140				ipstat.ips_fragtimeout += fpp->ipq_nfrags;
1141				ip_freef(&ipq[i], fpp);
1142			}
1143		}
1144	}
1145	/*
1146	 * If we are over the maximum number of fragments
1147	 * (due to the limit being lowered), drain off
1148	 * enough to get down to the new limit.
1149	 */
1150	if (maxnipq >= 0 && nipq > maxnipq) {
1151		for (i = 0; i < IPREASS_NHASH; i++) {
1152			while (nipq > maxnipq && !TAILQ_EMPTY(&ipq[i])) {
1153				ipstat.ips_fragdropped +=
1154				    TAILQ_FIRST(&ipq[i])->ipq_nfrags;
1155				ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i]));
1156			}
1157		}
1158	}
1159	IPQ_UNLOCK();
1160}
1161
1162/*
1163 * Drain off all datagram fragments.
1164 */
1165void
1166ip_drain()
1167{
1168	int     i;
1169
1170	IPQ_LOCK();
1171	for (i = 0; i < IPREASS_NHASH; i++) {
1172		while(!TAILQ_EMPTY(&ipq[i])) {
1173			ipstat.ips_fragdropped +=
1174			    TAILQ_FIRST(&ipq[i])->ipq_nfrags;
1175			ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i]));
1176		}
1177	}
1178	IPQ_UNLOCK();
1179	in_rtqdrain();
1180}
1181
1182/*
1183 * The protocol to be inserted into ip_protox[] must be already registered
1184 * in inetsw[], either statically or through pf_proto_register().
1185 */
1186int
1187ipproto_register(u_char ipproto)
1188{
1189	struct protosw *pr;
1190
1191	/* Sanity checks. */
1192	if (ipproto == 0)
1193		return (EPROTONOSUPPORT);
1194
1195	/*
1196	 * The protocol slot must not be occupied by another protocol
1197	 * already.  An index pointing to IPPROTO_RAW is unused.
1198	 */
1199	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
1200	if (pr == NULL)
1201		return (EPFNOSUPPORT);
1202	if (ip_protox[ipproto] != pr - inetsw)	/* IPPROTO_RAW */
1203		return (EEXIST);
1204
1205	/* Find the protocol position in inetsw[] and set the index. */
1206	for (pr = inetdomain.dom_protosw;
1207	     pr < inetdomain.dom_protoswNPROTOSW; pr++) {
1208		if (pr->pr_domain->dom_family == PF_INET &&
1209		    pr->pr_protocol && pr->pr_protocol == ipproto) {
1210			/* Be careful to only index valid IP protocols. */
1211			if (pr->pr_protocol < IPPROTO_MAX) {
1212				ip_protox[pr->pr_protocol] = pr - inetsw;
1213				return (0);
1214			} else
1215				return (EINVAL);
1216		}
1217	}
1218	return (EPROTONOSUPPORT);
1219}
1220
1221int
1222ipproto_unregister(u_char ipproto)
1223{
1224	struct protosw *pr;
1225
1226	/* Sanity checks. */
1227	if (ipproto == 0)
1228		return (EPROTONOSUPPORT);
1229
1230	/* Check if the protocol was indeed registered. */
1231	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
1232	if (pr == NULL)
1233		return (EPFNOSUPPORT);
1234	if (ip_protox[ipproto] == pr - inetsw)  /* IPPROTO_RAW */
1235		return (ENOENT);
1236
1237	/* Reset the protocol slot to IPPROTO_RAW. */
1238	ip_protox[ipproto] = pr - inetsw;
1239	return (0);
1240}
1241
1242
1243/*
1244 * Do option processing on a datagram,
1245 * possibly discarding it if bad options are encountered,
1246 * or forwarding it if source-routed.
1247 * The pass argument is used when operating in the IPSTEALTH
1248 * mode to tell what options to process:
1249 * [LS]SRR (pass 0) or the others (pass 1).
1250 * The reason for as many as two passes is that when doing IPSTEALTH,
1251 * non-routing options should be processed only if the packet is for us.
1252 * Returns 1 if packet has been forwarded/freed,
1253 * 0 if the packet should be processed further.
1254 */
1255static int
1256ip_dooptions(struct mbuf *m, int pass)
1257{
1258	struct ip *ip = mtod(m, struct ip *);
1259	u_char *cp;
1260	struct in_ifaddr *ia;
1261	int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0;
1262	struct in_addr *sin, dst;
1263	n_time ntime;
1264	struct	sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET };
1265
1266	/* ignore or reject packets with IP options */
1267	if (ip_doopts == 0)
1268		return 0;
1269	else if (ip_doopts == 2) {
1270		type = ICMP_UNREACH;
1271		code = ICMP_UNREACH_FILTER_PROHIB;
1272		goto bad;
1273	}
1274
1275	dst = ip->ip_dst;
1276	cp = (u_char *)(ip + 1);
1277	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
1278	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1279		opt = cp[IPOPT_OPTVAL];
1280		if (opt == IPOPT_EOL)
1281			break;
1282		if (opt == IPOPT_NOP)
1283			optlen = 1;
1284		else {
1285			if (cnt < IPOPT_OLEN + sizeof(*cp)) {
1286				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1287				goto bad;
1288			}
1289			optlen = cp[IPOPT_OLEN];
1290			if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
1291				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1292				goto bad;
1293			}
1294		}
1295		switch (opt) {
1296
1297		default:
1298			break;
1299
1300		/*
1301		 * Source routing with record.
1302		 * Find interface with current destination address.
1303		 * If none on this machine then drop if strictly routed,
1304		 * or do nothing if loosely routed.
1305		 * Record interface address and bring up next address
1306		 * component.  If strictly routed make sure next
1307		 * address is on directly accessible net.
1308		 */
1309		case IPOPT_LSRR:
1310		case IPOPT_SSRR:
1311#ifdef IPSTEALTH
1312			if (ipstealth && pass > 0)
1313				break;
1314#endif
1315			if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1316				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1317				goto bad;
1318			}
1319			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1320				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1321				goto bad;
1322			}
1323			ipaddr.sin_addr = ip->ip_dst;
1324			ia = (struct in_ifaddr *)
1325				ifa_ifwithaddr((struct sockaddr *)&ipaddr);
1326			if (ia == NULL) {
1327				if (opt == IPOPT_SSRR) {
1328					type = ICMP_UNREACH;
1329					code = ICMP_UNREACH_SRCFAIL;
1330					goto bad;
1331				}
1332				if (!ip_dosourceroute)
1333					goto nosourcerouting;
1334				/*
1335				 * Loose routing, and not at next destination
1336				 * yet; nothing to do except forward.
1337				 */
1338				break;
1339			}
1340			off--;			/* 0 origin */
1341			if (off > optlen - (int)sizeof(struct in_addr)) {
1342				/*
1343				 * End of source route.  Should be for us.
1344				 */
1345				if (!ip_acceptsourceroute)
1346					goto nosourcerouting;
1347				save_rte(m, cp, ip->ip_src);
1348				break;
1349			}
1350#ifdef IPSTEALTH
1351			if (ipstealth)
1352				goto dropit;
1353#endif
1354			if (!ip_dosourceroute) {
1355				if (ipforwarding) {
1356					char buf[16]; /* aaa.bbb.ccc.ddd\0 */
1357					/*
1358					 * Acting as a router, so generate ICMP
1359					 */
1360nosourcerouting:
1361					strcpy(buf, inet_ntoa(ip->ip_dst));
1362					log(LOG_WARNING,
1363					    "attempted source route from %s to %s\n",
1364					    inet_ntoa(ip->ip_src), buf);
1365					type = ICMP_UNREACH;
1366					code = ICMP_UNREACH_SRCFAIL;
1367					goto bad;
1368				} else {
1369					/*
1370					 * Not acting as a router, so silently drop.
1371					 */
1372#ifdef IPSTEALTH
1373dropit:
1374#endif
1375					ipstat.ips_cantforward++;
1376					m_freem(m);
1377					return (1);
1378				}
1379			}
1380
1381			/*
1382			 * locate outgoing interface
1383			 */
1384			(void)memcpy(&ipaddr.sin_addr, cp + off,
1385			    sizeof(ipaddr.sin_addr));
1386
1387			if (opt == IPOPT_SSRR) {
1388#define	INA	struct in_ifaddr *
1389#define	SA	struct sockaddr *
1390			    if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == NULL)
1391				ia = (INA)ifa_ifwithnet((SA)&ipaddr);
1392			} else
1393				ia = ip_rtaddr(ipaddr.sin_addr);
1394			if (ia == NULL) {
1395				type = ICMP_UNREACH;
1396				code = ICMP_UNREACH_SRCFAIL;
1397				goto bad;
1398			}
1399			ip->ip_dst = ipaddr.sin_addr;
1400			(void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
1401			    sizeof(struct in_addr));
1402			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1403			/*
1404			 * Let ip_intr's mcast routing check handle mcast pkts
1405			 */
1406			forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr));
1407			break;
1408
1409		case IPOPT_RR:
1410#ifdef IPSTEALTH
1411			if (ipstealth && pass == 0)
1412				break;
1413#endif
1414			if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1415				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1416				goto bad;
1417			}
1418			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1419				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1420				goto bad;
1421			}
1422			/*
1423			 * If no space remains, ignore.
1424			 */
1425			off--;			/* 0 origin */
1426			if (off > optlen - (int)sizeof(struct in_addr))
1427				break;
1428			(void)memcpy(&ipaddr.sin_addr, &ip->ip_dst,
1429			    sizeof(ipaddr.sin_addr));
1430			/*
1431			 * locate outgoing interface; if we're the destination,
1432			 * use the incoming interface (should be same).
1433			 */
1434			if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == NULL &&
1435			    (ia = ip_rtaddr(ipaddr.sin_addr)) == NULL) {
1436				type = ICMP_UNREACH;
1437				code = ICMP_UNREACH_HOST;
1438				goto bad;
1439			}
1440			(void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
1441			    sizeof(struct in_addr));
1442			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1443			break;
1444
1445		case IPOPT_TS:
1446#ifdef IPSTEALTH
1447			if (ipstealth && pass == 0)
1448				break;
1449#endif
1450			code = cp - (u_char *)ip;
1451			if (optlen < 4 || optlen > 40) {
1452				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1453				goto bad;
1454			}
1455			if ((off = cp[IPOPT_OFFSET]) < 5) {
1456				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1457				goto bad;
1458			}
1459			if (off > optlen - (int)sizeof(int32_t)) {
1460				cp[IPOPT_OFFSET + 1] += (1 << 4);
1461				if ((cp[IPOPT_OFFSET + 1] & 0xf0) == 0) {
1462					code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1463					goto bad;
1464				}
1465				break;
1466			}
1467			off--;				/* 0 origin */
1468			sin = (struct in_addr *)(cp + off);
1469			switch (cp[IPOPT_OFFSET + 1] & 0x0f) {
1470
1471			case IPOPT_TS_TSONLY:
1472				break;
1473
1474			case IPOPT_TS_TSANDADDR:
1475				if (off + sizeof(n_time) +
1476				    sizeof(struct in_addr) > optlen) {
1477					code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1478					goto bad;
1479				}
1480				ipaddr.sin_addr = dst;
1481				ia = (INA)ifaof_ifpforaddr((SA)&ipaddr,
1482							    m->m_pkthdr.rcvif);
1483				if (ia == NULL)
1484					continue;
1485				(void)memcpy(sin, &IA_SIN(ia)->sin_addr,
1486				    sizeof(struct in_addr));
1487				cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1488				off += sizeof(struct in_addr);
1489				break;
1490
1491			case IPOPT_TS_PRESPEC:
1492				if (off + sizeof(n_time) +
1493				    sizeof(struct in_addr) > optlen) {
1494					code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1495					goto bad;
1496				}
1497				(void)memcpy(&ipaddr.sin_addr, sin,
1498				    sizeof(struct in_addr));
1499				if (ifa_ifwithaddr((SA)&ipaddr) == NULL)
1500					continue;
1501				cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1502				off += sizeof(struct in_addr);
1503				break;
1504
1505			default:
1506				code = &cp[IPOPT_OFFSET + 1] - (u_char *)ip;
1507				goto bad;
1508			}
1509			ntime = iptime();
1510			(void)memcpy(cp + off, &ntime, sizeof(n_time));
1511			cp[IPOPT_OFFSET] += sizeof(n_time);
1512		}
1513	}
1514	if (forward && ipforwarding) {
1515		ip_forward(m, 1);
1516		return (1);
1517	}
1518	return (0);
1519bad:
1520	icmp_error(m, type, code, 0, 0);
1521	ipstat.ips_badoptions++;
1522	return (1);
1523}
1524
1525/*
1526 * Given address of next destination (final or next hop),
1527 * return internet address info of interface to be used to get there.
1528 */
1529struct in_ifaddr *
1530ip_rtaddr(dst)
1531	struct in_addr dst;
1532{
1533	struct route sro;
1534	struct sockaddr_in *sin;
1535	struct in_ifaddr *ifa;
1536
1537	bzero(&sro, sizeof(sro));
1538	sin = (struct sockaddr_in *)&sro.ro_dst;
1539	sin->sin_family = AF_INET;
1540	sin->sin_len = sizeof(*sin);
1541	sin->sin_addr = dst;
1542	rtalloc_ign(&sro, RTF_CLONING);
1543
1544	if (sro.ro_rt == NULL)
1545		return (NULL);
1546
1547	ifa = ifatoia(sro.ro_rt->rt_ifa);
1548	RTFREE(sro.ro_rt);
1549	return (ifa);
1550}
1551
1552/*
1553 * Save incoming source route for use in replies,
1554 * to be picked up later by ip_srcroute if the receiver is interested.
1555 */
1556static void
1557save_rte(m, option, dst)
1558	struct mbuf *m;
1559	u_char *option;
1560	struct in_addr dst;
1561{
1562	unsigned olen;
1563	struct ipopt_tag *opts;
1564
1565	opts = (struct ipopt_tag *)m_tag_get(PACKET_TAG_IPOPTIONS,
1566					sizeof(struct ipopt_tag), M_NOWAIT);
1567	if (opts == NULL)
1568		return;
1569
1570	olen = option[IPOPT_OLEN];
1571#ifdef DIAGNOSTIC
1572	if (ipprintfs)
1573		printf("save_rte: olen %d\n", olen);
1574#endif
1575	if (olen > sizeof(opts->ip_srcrt) - (1 + sizeof(dst))) {
1576		m_tag_free((struct m_tag *)opts);
1577		return;
1578	}
1579	bcopy(option, opts->ip_srcrt.srcopt, olen);
1580	opts->ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1581	opts->ip_srcrt.dst = dst;
1582	m_tag_prepend(m, (struct m_tag *)opts);
1583}
1584
1585/*
1586 * Retrieve incoming source route for use in replies,
1587 * in the same form used by setsockopt.
1588 * The first hop is placed before the options, will be removed later.
1589 */
1590struct mbuf *
1591ip_srcroute(m0)
1592	struct mbuf *m0;
1593{
1594	register struct in_addr *p, *q;
1595	register struct mbuf *m;
1596	struct ipopt_tag *opts;
1597
1598	opts = (struct ipopt_tag *)m_tag_find(m0, PACKET_TAG_IPOPTIONS, NULL);
1599	if (opts == NULL)
1600		return (NULL);
1601
1602	if (opts->ip_nhops == 0)
1603		return (NULL);
1604	m = m_get(M_DONTWAIT, MT_DATA);
1605	if (m == NULL)
1606		return (NULL);
1607
1608#define OPTSIZ	(sizeof(opts->ip_srcrt.nop) + sizeof(opts->ip_srcrt.srcopt))
1609
1610	/* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
1611	m->m_len = opts->ip_nhops * sizeof(struct in_addr) +
1612	    sizeof(struct in_addr) + OPTSIZ;
1613#ifdef DIAGNOSTIC
1614	if (ipprintfs)
1615		printf("ip_srcroute: nhops %d mlen %d", opts->ip_nhops, m->m_len);
1616#endif
1617
1618	/*
1619	 * First save first hop for return route
1620	 */
1621	p = &(opts->ip_srcrt.route[opts->ip_nhops - 1]);
1622	*(mtod(m, struct in_addr *)) = *p--;
1623#ifdef DIAGNOSTIC
1624	if (ipprintfs)
1625		printf(" hops %lx", (u_long)ntohl(mtod(m, struct in_addr *)->s_addr));
1626#endif
1627
1628	/*
1629	 * Copy option fields and padding (nop) to mbuf.
1630	 */
1631	opts->ip_srcrt.nop = IPOPT_NOP;
1632	opts->ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
1633	(void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr),
1634	    &(opts->ip_srcrt.nop), OPTSIZ);
1635	q = (struct in_addr *)(mtod(m, caddr_t) +
1636	    sizeof(struct in_addr) + OPTSIZ);
1637#undef OPTSIZ
1638	/*
1639	 * Record return path as an IP source route,
1640	 * reversing the path (pointers are now aligned).
1641	 */
1642	while (p >= opts->ip_srcrt.route) {
1643#ifdef DIAGNOSTIC
1644		if (ipprintfs)
1645			printf(" %lx", (u_long)ntohl(q->s_addr));
1646#endif
1647		*q++ = *p--;
1648	}
1649	/*
1650	 * Last hop goes to final destination.
1651	 */
1652	*q = opts->ip_srcrt.dst;
1653#ifdef DIAGNOSTIC
1654	if (ipprintfs)
1655		printf(" %lx\n", (u_long)ntohl(q->s_addr));
1656#endif
1657	m_tag_delete(m0, (struct m_tag *)opts);
1658	return (m);
1659}
1660
1661/*
1662 * Strip out IP options, at higher
1663 * level protocol in the kernel.
1664 * Second argument is buffer to which options
1665 * will be moved, and return value is their length.
1666 * XXX should be deleted; last arg currently ignored.
1667 */
1668void
1669ip_stripoptions(m, mopt)
1670	register struct mbuf *m;
1671	struct mbuf *mopt;
1672{
1673	register int i;
1674	struct ip *ip = mtod(m, struct ip *);
1675	register caddr_t opts;
1676	int olen;
1677
1678	olen = (ip->ip_hl << 2) - sizeof (struct ip);
1679	opts = (caddr_t)(ip + 1);
1680	i = m->m_len - (sizeof (struct ip) + olen);
1681	bcopy(opts + olen, opts, (unsigned)i);
1682	m->m_len -= olen;
1683	if (m->m_flags & M_PKTHDR)
1684		m->m_pkthdr.len -= olen;
1685	ip->ip_v = IPVERSION;
1686	ip->ip_hl = sizeof(struct ip) >> 2;
1687}
1688
1689u_char inetctlerrmap[PRC_NCMDS] = {
1690	0,		0,		0,		0,
1691	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1692	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1693	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1694	0,		0,		EHOSTUNREACH,	0,
1695	ENOPROTOOPT,	ECONNREFUSED
1696};
1697
1698/*
1699 * Forward a packet.  If some error occurs return the sender
1700 * an icmp packet.  Note we can't always generate a meaningful
1701 * icmp message because icmp doesn't have a large enough repertoire
1702 * of codes and types.
1703 *
1704 * If not forwarding, just drop the packet.  This could be confusing
1705 * if ipforwarding was zero but some routing protocol was advancing
1706 * us as a gateway to somewhere.  However, we must let the routing
1707 * protocol deal with that.
1708 *
1709 * The srcrt parameter indicates whether the packet is being forwarded
1710 * via a source route.
1711 */
1712void
1713ip_forward(struct mbuf *m, int srcrt)
1714{
1715	struct ip *ip = mtod(m, struct ip *);
1716	struct in_ifaddr *ia = NULL;
1717	struct mbuf *mcopy;
1718	struct in_addr dest;
1719	int error, type = 0, code = 0, mtu = 0;
1720
1721#ifdef DIAGNOSTIC
1722	if (ipprintfs)
1723		printf("forward: src %lx dst %lx ttl %x\n",
1724		    (u_long)ip->ip_src.s_addr, (u_long)ip->ip_dst.s_addr,
1725		    ip->ip_ttl);
1726#endif
1727
1728
1729	if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
1730		ipstat.ips_cantforward++;
1731		m_freem(m);
1732		return;
1733	}
1734#ifdef IPSTEALTH
1735	if (!ipstealth) {
1736#endif
1737		if (ip->ip_ttl <= IPTTLDEC) {
1738			icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS,
1739			    0, 0);
1740			return;
1741		}
1742#ifdef IPSTEALTH
1743	}
1744#endif
1745
1746	if (!srcrt && (ia = ip_rtaddr(ip->ip_dst)) == NULL) {
1747		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
1748		return;
1749	}
1750
1751	/*
1752	 * Save the IP header and at most 8 bytes of the payload,
1753	 * in case we need to generate an ICMP message to the src.
1754	 *
1755	 * XXX this can be optimized a lot by saving the data in a local
1756	 * buffer on the stack (72 bytes at most), and only allocating the
1757	 * mbuf if really necessary. The vast majority of the packets
1758	 * are forwarded without having to send an ICMP back (either
1759	 * because unnecessary, or because rate limited), so we are
1760	 * really we are wasting a lot of work here.
1761	 *
1762	 * We don't use m_copy() because it might return a reference
1763	 * to a shared cluster. Both this function and ip_output()
1764	 * assume exclusive access to the IP header in `m', so any
1765	 * data in a cluster may change before we reach icmp_error().
1766	 */
1767	MGET(mcopy, M_DONTWAIT, m->m_type);
1768	if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_DONTWAIT)) {
1769		/*
1770		 * It's probably ok if the pkthdr dup fails (because
1771		 * the deep copy of the tag chain failed), but for now
1772		 * be conservative and just discard the copy since
1773		 * code below may some day want the tags.
1774		 */
1775		m_free(mcopy);
1776		mcopy = NULL;
1777	}
1778	if (mcopy != NULL) {
1779		mcopy->m_len = imin((ip->ip_hl << 2) + 8,
1780		    (int)ip->ip_len);
1781		mcopy->m_pkthdr.len = mcopy->m_len;
1782		m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1783	}
1784
1785#ifdef IPSTEALTH
1786	if (!ipstealth) {
1787#endif
1788		ip->ip_ttl -= IPTTLDEC;
1789#ifdef IPSTEALTH
1790	}
1791#endif
1792
1793	/*
1794	 * If forwarding packet using same interface that it came in on,
1795	 * perhaps should send a redirect to sender to shortcut a hop.
1796	 * Only send redirect if source is sending directly to us,
1797	 * and if packet was not source routed (or has any options).
1798	 * Also, don't send redirect if forwarding using a default route
1799	 * or a route modified by a redirect.
1800	 */
1801	dest.s_addr = 0;
1802	if (!srcrt && ipsendredirects && ia->ia_ifp == m->m_pkthdr.rcvif) {
1803		struct sockaddr_in *sin;
1804		struct route ro;
1805		struct rtentry *rt;
1806
1807		bzero(&ro, sizeof(ro));
1808		sin = (struct sockaddr_in *)&ro.ro_dst;
1809		sin->sin_family = AF_INET;
1810		sin->sin_len = sizeof(*sin);
1811		sin->sin_addr = ip->ip_dst;
1812		rtalloc_ign(&ro, RTF_CLONING);
1813
1814		rt = ro.ro_rt;
1815
1816		if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1817		    satosin(rt_key(rt))->sin_addr.s_addr != 0) {
1818#define	RTA(rt)	((struct in_ifaddr *)(rt->rt_ifa))
1819			u_long src = ntohl(ip->ip_src.s_addr);
1820
1821			if (RTA(rt) &&
1822			    (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1823				if (rt->rt_flags & RTF_GATEWAY)
1824					dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr;
1825				else
1826					dest.s_addr = ip->ip_dst.s_addr;
1827				/* Router requirements says to only send host redirects */
1828				type = ICMP_REDIRECT;
1829				code = ICMP_REDIRECT_HOST;
1830#ifdef DIAGNOSTIC
1831				if (ipprintfs)
1832					printf("redirect (%d) to %lx\n", code, (u_long)dest.s_addr);
1833#endif
1834			}
1835		}
1836		if (rt)
1837			RTFREE(rt);
1838	}
1839
1840	error = ip_output(m, NULL, NULL, IP_FORWARDING, NULL, NULL);
1841	if (error)
1842		ipstat.ips_cantforward++;
1843	else {
1844		ipstat.ips_forward++;
1845		if (type)
1846			ipstat.ips_redirectsent++;
1847		else {
1848			if (mcopy)
1849				m_freem(mcopy);
1850			return;
1851		}
1852	}
1853	if (mcopy == NULL)
1854		return;
1855
1856	switch (error) {
1857
1858	case 0:				/* forwarded, but need redirect */
1859		/* type, code set above */
1860		break;
1861
1862	case ENETUNREACH:		/* shouldn't happen, checked above */
1863	case EHOSTUNREACH:
1864	case ENETDOWN:
1865	case EHOSTDOWN:
1866	default:
1867		type = ICMP_UNREACH;
1868		code = ICMP_UNREACH_HOST;
1869		break;
1870
1871	case EMSGSIZE:
1872		type = ICMP_UNREACH;
1873		code = ICMP_UNREACH_NEEDFRAG;
1874#if defined(IPSEC) || defined(FAST_IPSEC)
1875		/*
1876		 * If the packet is routed over IPsec tunnel, tell the
1877		 * originator the tunnel MTU.
1878		 *	tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
1879		 * XXX quickhack!!!
1880		 */
1881		{
1882			struct secpolicy *sp = NULL;
1883			int ipsecerror;
1884			int ipsechdr;
1885			struct route *ro;
1886
1887#ifdef IPSEC
1888			sp = ipsec4_getpolicybyaddr(mcopy,
1889						    IPSEC_DIR_OUTBOUND,
1890						    IP_FORWARDING,
1891						    &ipsecerror);
1892#else /* FAST_IPSEC */
1893			sp = ipsec_getpolicybyaddr(mcopy,
1894						   IPSEC_DIR_OUTBOUND,
1895						   IP_FORWARDING,
1896						   &ipsecerror);
1897#endif
1898			if (sp != NULL) {
1899				/* count IPsec header size */
1900				ipsechdr = ipsec4_hdrsiz(mcopy,
1901							 IPSEC_DIR_OUTBOUND,
1902							 NULL);
1903
1904				/*
1905				 * find the correct route for outer IPv4
1906				 * header, compute tunnel MTU.
1907				 */
1908				if (sp->req != NULL
1909				 && sp->req->sav != NULL
1910				 && sp->req->sav->sah != NULL) {
1911					ro = &sp->req->sav->sah->sa_route;
1912					if (ro->ro_rt && ro->ro_rt->rt_ifp) {
1913						mtu =
1914						    ro->ro_rt->rt_rmx.rmx_mtu ?
1915						    ro->ro_rt->rt_rmx.rmx_mtu :
1916						    ro->ro_rt->rt_ifp->if_mtu;
1917						mtu -= ipsechdr;
1918					}
1919				}
1920
1921#ifdef IPSEC
1922				key_freesp(sp);
1923#else /* FAST_IPSEC */
1924				KEY_FREESP(&sp);
1925#endif
1926				ipstat.ips_cantfrag++;
1927				break;
1928			} else
1929#endif /*IPSEC || FAST_IPSEC*/
1930		/*
1931		 * When doing source routing 'ia' can be NULL.  Fall back
1932		 * to the minimum guaranteed routeable packet size and use
1933		 * the same hack as IPSEC to setup a dummyifp for icmp.
1934		 */
1935		if (ia == NULL)
1936			mtu = IP_MSS;
1937		else
1938			mtu = ia->ia_ifp->if_mtu;
1939#if defined(IPSEC) || defined(FAST_IPSEC)
1940		}
1941#endif /*IPSEC || FAST_IPSEC*/
1942		ipstat.ips_cantfrag++;
1943		break;
1944
1945	case ENOBUFS:
1946		/*
1947		 * A router should not generate ICMP_SOURCEQUENCH as
1948		 * required in RFC1812 Requirements for IP Version 4 Routers.
1949		 * Source quench could be a big problem under DoS attacks,
1950		 * or if the underlying interface is rate-limited.
1951		 * Those who need source quench packets may re-enable them
1952		 * via the net.inet.ip.sendsourcequench sysctl.
1953		 */
1954		if (ip_sendsourcequench == 0) {
1955			m_freem(mcopy);
1956			return;
1957		} else {
1958			type = ICMP_SOURCEQUENCH;
1959			code = 0;
1960		}
1961		break;
1962
1963	case EACCES:			/* ipfw denied packet */
1964		m_freem(mcopy);
1965		return;
1966	}
1967	icmp_error(mcopy, type, code, dest.s_addr, mtu);
1968}
1969
1970void
1971ip_savecontrol(inp, mp, ip, m)
1972	register struct inpcb *inp;
1973	register struct mbuf **mp;
1974	register struct ip *ip;
1975	register struct mbuf *m;
1976{
1977	if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) {
1978		struct bintime bt;
1979
1980		bintime(&bt);
1981		if (inp->inp_socket->so_options & SO_BINTIME) {
1982			*mp = sbcreatecontrol((caddr_t) &bt, sizeof(bt),
1983			SCM_BINTIME, SOL_SOCKET);
1984			if (*mp)
1985				mp = &(*mp)->m_next;
1986		}
1987		if (inp->inp_socket->so_options & SO_TIMESTAMP) {
1988			struct timeval tv;
1989
1990			bintime2timeval(&bt, &tv);
1991			*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1992				SCM_TIMESTAMP, SOL_SOCKET);
1993			if (*mp)
1994				mp = &(*mp)->m_next;
1995		}
1996	}
1997	if (inp->inp_flags & INP_RECVDSTADDR) {
1998		*mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
1999		    sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
2000		if (*mp)
2001			mp = &(*mp)->m_next;
2002	}
2003	if (inp->inp_flags & INP_RECVTTL) {
2004		*mp = sbcreatecontrol((caddr_t) &ip->ip_ttl,
2005		    sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
2006		if (*mp)
2007			mp = &(*mp)->m_next;
2008	}
2009#ifdef notyet
2010	/* XXX
2011	 * Moving these out of udp_input() made them even more broken
2012	 * than they already were.
2013	 */
2014	/* options were tossed already */
2015	if (inp->inp_flags & INP_RECVOPTS) {
2016		*mp = sbcreatecontrol((caddr_t) opts_deleted_above,
2017		    sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
2018		if (*mp)
2019			mp = &(*mp)->m_next;
2020	}
2021	/* ip_srcroute doesn't do what we want here, need to fix */
2022	if (inp->inp_flags & INP_RECVRETOPTS) {
2023		*mp = sbcreatecontrol((caddr_t) ip_srcroute(m),
2024		    sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
2025		if (*mp)
2026			mp = &(*mp)->m_next;
2027	}
2028#endif
2029	if (inp->inp_flags & INP_RECVIF) {
2030		struct ifnet *ifp;
2031		struct sdlbuf {
2032			struct sockaddr_dl sdl;
2033			u_char	pad[32];
2034		} sdlbuf;
2035		struct sockaddr_dl *sdp;
2036		struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
2037
2038		if (((ifp = m->m_pkthdr.rcvif))
2039		&& ( ifp->if_index && (ifp->if_index <= if_index))) {
2040			sdp = (struct sockaddr_dl *)
2041			    (ifaddr_byindex(ifp->if_index)->ifa_addr);
2042			/*
2043			 * Change our mind and don't try copy.
2044			 */
2045			if ((sdp->sdl_family != AF_LINK)
2046			|| (sdp->sdl_len > sizeof(sdlbuf))) {
2047				goto makedummy;
2048			}
2049			bcopy(sdp, sdl2, sdp->sdl_len);
2050		} else {
2051makedummy:
2052			sdl2->sdl_len
2053				= offsetof(struct sockaddr_dl, sdl_data[0]);
2054			sdl2->sdl_family = AF_LINK;
2055			sdl2->sdl_index = 0;
2056			sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
2057		}
2058		*mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len,
2059			IP_RECVIF, IPPROTO_IP);
2060		if (*mp)
2061			mp = &(*mp)->m_next;
2062	}
2063}
2064
2065/*
2066 * XXX these routines are called from the upper part of the kernel.
2067 * They need to be locked when we remove Giant.
2068 *
2069 * They could also be moved to ip_mroute.c, since all the RSVP
2070 *  handling is done there already.
2071 */
2072static int ip_rsvp_on;
2073struct socket *ip_rsvpd;
2074int
2075ip_rsvp_init(struct socket *so)
2076{
2077	if (so->so_type != SOCK_RAW ||
2078	    so->so_proto->pr_protocol != IPPROTO_RSVP)
2079		return EOPNOTSUPP;
2080
2081	if (ip_rsvpd != NULL)
2082		return EADDRINUSE;
2083
2084	ip_rsvpd = so;
2085	/*
2086	 * This may seem silly, but we need to be sure we don't over-increment
2087	 * the RSVP counter, in case something slips up.
2088	 */
2089	if (!ip_rsvp_on) {
2090		ip_rsvp_on = 1;
2091		rsvp_on++;
2092	}
2093
2094	return 0;
2095}
2096
2097int
2098ip_rsvp_done(void)
2099{
2100	ip_rsvpd = NULL;
2101	/*
2102	 * This may seem silly, but we need to be sure we don't over-decrement
2103	 * the RSVP counter, in case something slips up.
2104	 */
2105	if (ip_rsvp_on) {
2106		ip_rsvp_on = 0;
2107		rsvp_on--;
2108	}
2109	return 0;
2110}
2111
2112void
2113rsvp_input(struct mbuf *m, int off)	/* XXX must fixup manually */
2114{
2115	if (rsvp_input_p) { /* call the real one if loaded */
2116		rsvp_input_p(m, off);
2117		return;
2118	}
2119
2120	/* Can still get packets with rsvp_on = 0 if there is a local member
2121	 * of the group to which the RSVP packet is addressed.  But in this
2122	 * case we want to throw the packet away.
2123	 */
2124
2125	if (!rsvp_on) {
2126		m_freem(m);
2127		return;
2128	}
2129
2130	if (ip_rsvpd != NULL) {
2131		rip_input(m, off);
2132		return;
2133	}
2134	/* Drop the packet */
2135	m_freem(m);
2136}
2137