ip_input.c revision 111275
1/*
2 * Copyright (c) 1982, 1986, 1988, 1993
3 *	The Regents of the University of California.  All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 *    notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 *    notice, this list of conditions and the following disclaimer in the
12 *    documentation and/or other materials provided with the distribution.
13 * 3. All advertising materials mentioning features or use of this software
14 *    must display the following acknowledgement:
15 *	This product includes software developed by the University of
16 *	California, Berkeley and its contributors.
17 * 4. Neither the name of the University nor the names of its contributors
18 *    may be used to endorse or promote products derived from this software
19 *    without specific prior written permission.
20 *
21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31 * SUCH DAMAGE.
32 *
33 *	@(#)ip_input.c	8.2 (Berkeley) 1/4/94
34 * $FreeBSD: head/sys/netinet/ip_input.c 111275 2003-02-23 00:47:06Z sam $
35 */
36
37#include "opt_bootp.h"
38#include "opt_ipfw.h"
39#include "opt_ipdn.h"
40#include "opt_ipdivert.h"
41#include "opt_ipfilter.h"
42#include "opt_ipstealth.h"
43#include "opt_ipsec.h"
44#include "opt_mac.h"
45#include "opt_pfil_hooks.h"
46#include "opt_random_ip_id.h"
47
48#include <sys/param.h>
49#include <sys/systm.h>
50#include <sys/mac.h>
51#include <sys/mbuf.h>
52#include <sys/malloc.h>
53#include <sys/domain.h>
54#include <sys/protosw.h>
55#include <sys/socket.h>
56#include <sys/time.h>
57#include <sys/kernel.h>
58#include <sys/syslog.h>
59#include <sys/sysctl.h>
60
61#include <net/pfil.h>
62#include <net/if.h>
63#include <net/if_types.h>
64#include <net/if_var.h>
65#include <net/if_dl.h>
66#include <net/route.h>
67#include <net/netisr.h>
68#include <net/intrq.h>
69
70#include <netinet/in.h>
71#include <netinet/in_systm.h>
72#include <netinet/in_var.h>
73#include <netinet/ip.h>
74#include <netinet/in_pcb.h>
75#include <netinet/ip_var.h>
76#include <netinet/ip_icmp.h>
77#include <machine/in_cksum.h>
78
79#include <sys/socketvar.h>
80
81#include <netinet/ip_fw.h>
82#include <netinet/ip_dummynet.h>
83
84#ifdef IPSEC
85#include <netinet6/ipsec.h>
86#include <netkey/key.h>
87#endif
88
89#ifdef FAST_IPSEC
90#include <netipsec/ipsec.h>
91#include <netipsec/key.h>
92#endif
93
94int rsvp_on = 0;
95
96int	ipforwarding = 0;
97SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW,
98    &ipforwarding, 0, "Enable IP forwarding between interfaces");
99
100static int	ipsendredirects = 1; /* XXX */
101SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW,
102    &ipsendredirects, 0, "Enable sending IP redirects");
103
104int	ip_defttl = IPDEFTTL;
105SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW,
106    &ip_defttl, 0, "Maximum TTL on IP packets");
107
108static int	ip_dosourceroute = 0;
109SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW,
110    &ip_dosourceroute, 0, "Enable forwarding source routed IP packets");
111
112static int	ip_acceptsourceroute = 0;
113SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute,
114    CTLFLAG_RW, &ip_acceptsourceroute, 0,
115    "Enable accepting source routed IP packets");
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
138/*
139 * XXX - Setting ip_checkinterface mostly implements the receive side of
140 * the Strong ES model described in RFC 1122, but since the routing table
141 * and transmit implementation do not implement the Strong ES model,
142 * setting this to 1 results in an odd hybrid.
143 *
144 * XXX - ip_checkinterface currently must be disabled if you use ipnat
145 * to translate the destination address to another local interface.
146 *
147 * XXX - ip_checkinterface must be disabled if you add IP aliases
148 * to the loopback interface instead of the interface where the
149 * packets for those addresses are received.
150 */
151static int	ip_checkinterface = 1;
152SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW,
153    &ip_checkinterface, 0, "Verify packet arrives on correct interface");
154
155#ifdef DIAGNOSTIC
156static int	ipprintfs = 0;
157#endif
158
159static int	ipqmaxlen = IFQ_MAXLEN;
160
161extern	struct domain inetdomain;
162extern	struct protosw inetsw[];
163u_char	ip_protox[IPPROTO_MAX];
164struct	in_ifaddrhead in_ifaddrhead; 		/* first inet address */
165struct	in_ifaddrhashhead *in_ifaddrhashtbl;	/* inet addr hash table  */
166u_long 	in_ifaddrhmask;				/* mask for hash table */
167
168SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW,
169    &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue");
170SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD,
171    &ipintrq.ifq_drops, 0, "Number of packets dropped from the IP input queue");
172
173struct ipstat ipstat;
174SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW,
175    &ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)");
176
177/* Packet reassembly stuff */
178#define IPREASS_NHASH_LOG2      6
179#define IPREASS_NHASH           (1 << IPREASS_NHASH_LOG2)
180#define IPREASS_HMASK           (IPREASS_NHASH - 1)
181#define IPREASS_HASH(x,y) \
182	(((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK)
183
184static TAILQ_HEAD(ipqhead, ipq) ipq[IPREASS_NHASH];
185
186#ifdef IPCTL_DEFMTU
187SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
188    &ip_mtu, 0, "Default MTU");
189#endif
190
191#ifdef IPSTEALTH
192static int	ipstealth = 0;
193SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW,
194    &ipstealth, 0, "");
195#endif
196
197
198/* Firewall hooks */
199ip_fw_chk_t *ip_fw_chk_ptr;
200int fw_enable = 1 ;
201int fw_one_pass = 1;
202
203/* Dummynet hooks */
204ip_dn_io_t *ip_dn_io_ptr;
205
206
207/*
208 * XXX this is ugly -- the following two global variables are
209 * used to store packet state while it travels through the stack.
210 * Note that the code even makes assumptions on the size and
211 * alignment of fields inside struct ip_srcrt so e.g. adding some
212 * fields will break the code. This needs to be fixed.
213 *
214 * We need to save the IP options in case a protocol wants to respond
215 * to an incoming packet over the same route if the packet got here
216 * using IP source routing.  This allows connection establishment and
217 * maintenance when the remote end is on a network that is not known
218 * to us.
219 */
220static int	ip_nhops = 0;
221static	struct ip_srcrt {
222	struct	in_addr dst;			/* final destination */
223	char	nop;				/* one NOP to align */
224	char	srcopt[IPOPT_OFFSET + 1];	/* OPTVAL, OLEN and OFFSET */
225	struct	in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)];
226} ip_srcrt;
227
228static void	save_rte(u_char *, struct in_addr);
229static int	ip_dooptions(struct mbuf *m, int,
230			struct sockaddr_in *next_hop);
231static void	ip_forward(struct mbuf *m, int srcrt,
232			struct sockaddr_in *next_hop);
233static void	ip_freef(struct ipqhead *, struct ipq *);
234static struct	mbuf *ip_reass(struct mbuf *, struct ipqhead *,
235		struct ipq *, u_int32_t *, u_int16_t *);
236static void	ipintr(void);
237
238/*
239 * IP initialization: fill in IP protocol switch table.
240 * All protocols not implemented in kernel go to raw IP protocol handler.
241 */
242void
243ip_init()
244{
245	register struct protosw *pr;
246	register int i;
247
248	TAILQ_INIT(&in_ifaddrhead);
249	in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask);
250	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
251	if (pr == 0)
252		panic("ip_init");
253	for (i = 0; i < IPPROTO_MAX; i++)
254		ip_protox[i] = pr - inetsw;
255	for (pr = inetdomain.dom_protosw;
256	    pr < inetdomain.dom_protoswNPROTOSW; pr++)
257		if (pr->pr_domain->dom_family == PF_INET &&
258		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
259			ip_protox[pr->pr_protocol] = pr - inetsw;
260
261	for (i = 0; i < IPREASS_NHASH; i++)
262	    TAILQ_INIT(&ipq[i]);
263
264	maxnipq = nmbclusters / 32;
265	maxfragsperpacket = 16;
266
267#ifndef RANDOM_IP_ID
268	ip_id = time_second & 0xffff;
269#endif
270	ipintrq.ifq_maxlen = ipqmaxlen;
271	mtx_init(&ipintrq.ifq_mtx, "ip_inq", NULL, MTX_DEF);
272	ipintrq_present = 1;
273
274	register_netisr(NETISR_IP, ipintr);
275}
276
277/*
278 * XXX watch out this one. It is perhaps used as a cache for
279 * the most recently used route ? it is cleared in in_addroute()
280 * when a new route is successfully created.
281 */
282struct	route ipforward_rt;
283
284/*
285 * Ip input routine.  Checksum and byte swap header.  If fragmented
286 * try to reassemble.  Process options.  Pass to next level.
287 */
288void
289ip_input(struct mbuf *m)
290{
291	struct ip *ip;
292	struct ipq *fp;
293	struct in_ifaddr *ia = NULL;
294	struct ifaddr *ifa;
295	int    i, hlen, checkif;
296	u_short sum;
297	struct in_addr pkt_dst;
298	u_int32_t divert_info = 0;		/* packet divert/tee info */
299	struct ip_fw_args args;
300#ifdef PFIL_HOOKS
301	struct packet_filter_hook *pfh;
302	struct mbuf *m0;
303	int rv;
304#endif /* PFIL_HOOKS */
305#ifdef FAST_IPSEC
306	struct m_tag *mtag;
307	struct tdb_ident *tdbi;
308	struct secpolicy *sp;
309	int s, error;
310#endif /* FAST_IPSEC */
311
312	args.eh = NULL;
313	args.oif = NULL;
314	args.rule = NULL;
315	args.divert_rule = 0;			/* divert cookie */
316	args.next_hop = NULL;
317
318	/* Grab info from MT_TAG mbufs prepended to the chain.	*/
319	for (; m && m->m_type == MT_TAG; m = m->m_next) {
320		switch(m->_m_tag_id) {
321		default:
322			printf("ip_input: unrecognised MT_TAG tag %d\n",
323			    m->_m_tag_id);
324			break;
325
326		case PACKET_TAG_DUMMYNET:
327			args.rule = ((struct dn_pkt *)m)->rule;
328			break;
329
330		case PACKET_TAG_DIVERT:
331			args.divert_rule = (intptr_t)m->m_hdr.mh_data & 0xffff;
332			break;
333
334		case PACKET_TAG_IPFORWARD:
335			args.next_hop = (struct sockaddr_in *)m->m_hdr.mh_data;
336			break;
337		}
338	}
339
340	KASSERT(m != NULL && (m->m_flags & M_PKTHDR) != 0,
341	    ("ip_input: no HDR"));
342
343	if (args.rule) {	/* dummynet already filtered us */
344		ip = mtod(m, struct ip *);
345		hlen = ip->ip_hl << 2;
346		goto iphack ;
347	}
348
349	ipstat.ips_total++;
350
351	if (m->m_pkthdr.len < sizeof(struct ip))
352		goto tooshort;
353
354	if (m->m_len < sizeof (struct ip) &&
355	    (m = m_pullup(m, sizeof (struct ip))) == 0) {
356		ipstat.ips_toosmall++;
357		return;
358	}
359	ip = mtod(m, struct ip *);
360
361	if (ip->ip_v != IPVERSION) {
362		ipstat.ips_badvers++;
363		goto bad;
364	}
365
366	hlen = ip->ip_hl << 2;
367	if (hlen < sizeof(struct ip)) {	/* minimum header length */
368		ipstat.ips_badhlen++;
369		goto bad;
370	}
371	if (hlen > m->m_len) {
372		if ((m = m_pullup(m, hlen)) == 0) {
373			ipstat.ips_badhlen++;
374			return;
375		}
376		ip = mtod(m, struct ip *);
377	}
378
379	/* 127/8 must not appear on wire - RFC1122 */
380	if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
381	    (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
382		if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
383			ipstat.ips_badaddr++;
384			goto bad;
385		}
386	}
387
388	if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
389		sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
390	} else {
391		if (hlen == sizeof(struct ip)) {
392			sum = in_cksum_hdr(ip);
393		} else {
394			sum = in_cksum(m, hlen);
395		}
396	}
397	if (sum) {
398		ipstat.ips_badsum++;
399		goto bad;
400	}
401
402	/*
403	 * Convert fields to host representation.
404	 */
405	ip->ip_len = ntohs(ip->ip_len);
406	if (ip->ip_len < hlen) {
407		ipstat.ips_badlen++;
408		goto bad;
409	}
410	ip->ip_off = ntohs(ip->ip_off);
411
412	/*
413	 * Check that the amount of data in the buffers
414	 * is as at least much as the IP header would have us expect.
415	 * Trim mbufs if longer than we expect.
416	 * Drop packet if shorter than we expect.
417	 */
418	if (m->m_pkthdr.len < ip->ip_len) {
419tooshort:
420		ipstat.ips_tooshort++;
421		goto bad;
422	}
423	if (m->m_pkthdr.len > ip->ip_len) {
424		if (m->m_len == m->m_pkthdr.len) {
425			m->m_len = ip->ip_len;
426			m->m_pkthdr.len = ip->ip_len;
427		} else
428			m_adj(m, ip->ip_len - m->m_pkthdr.len);
429	}
430#if defined(IPSEC) && !defined(IPSEC_FILTERGIF)
431	/*
432	 * Bypass packet filtering for packets from a tunnel (gif).
433	 */
434	if (ipsec_gethist(m, NULL))
435		goto pass;
436#endif
437
438	/*
439	 * IpHack's section.
440	 * Right now when no processing on packet has done
441	 * and it is still fresh out of network we do our black
442	 * deals with it.
443	 * - Firewall: deny/allow/divert
444	 * - Xlate: translate packet's addr/port (NAT).
445	 * - Pipe: pass pkt through dummynet.
446	 * - Wrap: fake packet's addr/port <unimpl.>
447	 * - Encapsulate: put it in another IP and send out. <unimp.>
448 	 */
449
450iphack:
451
452#ifdef PFIL_HOOKS
453	/*
454	 * Run through list of hooks for input packets.  If there are any
455	 * filters which require that additional packets in the flow are
456	 * not fast-forwarded, they must clear the M_CANFASTFWD flag.
457	 * Note that filters must _never_ set this flag, as another filter
458	 * in the list may have previously cleared it.
459	 */
460	m0 = m;
461	pfh = pfil_hook_get(PFIL_IN, &inetsw[ip_protox[IPPROTO_IP]].pr_pfh);
462	for (; pfh; pfh = TAILQ_NEXT(pfh, pfil_link))
463		if (pfh->pfil_func) {
464			rv = pfh->pfil_func(ip, hlen,
465					    m->m_pkthdr.rcvif, 0, &m0);
466			if (rv)
467				return;
468			m = m0;
469			if (m == NULL)
470				return;
471			ip = mtod(m, struct ip *);
472		}
473#endif /* PFIL_HOOKS */
474
475	if (fw_enable && IPFW_LOADED) {
476		/*
477		 * If we've been forwarded from the output side, then
478		 * skip the firewall a second time
479		 */
480		if (args.next_hop)
481			goto ours;
482
483		args.m = m;
484		i = ip_fw_chk_ptr(&args);
485		m = args.m;
486
487		if ( (i & IP_FW_PORT_DENY_FLAG) || m == NULL) { /* drop */
488			if (m)
489				m_freem(m);
490			return;
491		}
492		ip = mtod(m, struct ip *); /* just in case m changed */
493		if (i == 0 && args.next_hop == NULL)	/* common case */
494			goto pass;
495                if (DUMMYNET_LOADED && (i & IP_FW_PORT_DYNT_FLAG) != 0) {
496			/* Send packet to the appropriate pipe */
497			ip_dn_io_ptr(m, i&0xffff, DN_TO_IP_IN, &args);
498			return;
499		}
500#ifdef IPDIVERT
501		if (i != 0 && (i & IP_FW_PORT_DYNT_FLAG) == 0) {
502			/* Divert or tee packet */
503			divert_info = i;
504			goto ours;
505		}
506#endif
507		if (i == 0 && args.next_hop != NULL)
508			goto pass;
509		/*
510		 * if we get here, the packet must be dropped
511		 */
512		m_freem(m);
513		return;
514	}
515pass:
516
517	/*
518	 * Process options and, if not destined for us,
519	 * ship it on.  ip_dooptions returns 1 when an
520	 * error was detected (causing an icmp message
521	 * to be sent and the original packet to be freed).
522	 */
523	ip_nhops = 0;		/* for source routed packets */
524	if (hlen > sizeof (struct ip) && ip_dooptions(m, 0, args.next_hop))
525		return;
526
527        /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
528         * matter if it is destined to another node, or whether it is
529         * a multicast one, RSVP wants it! and prevents it from being forwarded
530         * anywhere else. Also checks if the rsvp daemon is running before
531	 * grabbing the packet.
532         */
533	if (rsvp_on && ip->ip_p==IPPROTO_RSVP)
534		goto ours;
535
536	/*
537	 * Check our list of addresses, to see if the packet is for us.
538	 * If we don't have any addresses, assume any unicast packet
539	 * we receive might be for us (and let the upper layers deal
540	 * with it).
541	 */
542	if (TAILQ_EMPTY(&in_ifaddrhead) &&
543	    (m->m_flags & (M_MCAST|M_BCAST)) == 0)
544		goto ours;
545
546	/*
547	 * Cache the destination address of the packet; this may be
548	 * changed by use of 'ipfw fwd'.
549	 */
550	pkt_dst = args.next_hop ? args.next_hop->sin_addr : ip->ip_dst;
551
552	/*
553	 * Enable a consistency check between the destination address
554	 * and the arrival interface for a unicast packet (the RFC 1122
555	 * strong ES model) if IP forwarding is disabled and the packet
556	 * is not locally generated and the packet is not subject to
557	 * 'ipfw fwd'.
558	 *
559	 * XXX - Checking also should be disabled if the destination
560	 * address is ipnat'ed to a different interface.
561	 *
562	 * XXX - Checking is incompatible with IP aliases added
563	 * to the loopback interface instead of the interface where
564	 * the packets are received.
565	 */
566	checkif = ip_checkinterface && (ipforwarding == 0) &&
567	    m->m_pkthdr.rcvif != NULL &&
568	    ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) &&
569	    (args.next_hop == NULL);
570
571	/*
572	 * Check for exact addresses in the hash bucket.
573	 */
574	LIST_FOREACH(ia, INADDR_HASH(pkt_dst.s_addr), ia_hash) {
575		/*
576		 * If the address matches, verify that the packet
577		 * arrived via the correct interface if checking is
578		 * enabled.
579		 */
580		if (IA_SIN(ia)->sin_addr.s_addr == pkt_dst.s_addr &&
581		    (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif))
582			goto ours;
583	}
584	/*
585	 * Check for broadcast addresses.
586	 *
587	 * Only accept broadcast packets that arrive via the matching
588	 * interface.  Reception of forwarded directed broadcasts would
589	 * be handled via ip_forward() and ether_output() with the loopback
590	 * into the stack for SIMPLEX interfaces handled by ether_output().
591	 */
592	if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
593	        TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
594			if (ifa->ifa_addr->sa_family != AF_INET)
595				continue;
596			ia = ifatoia(ifa);
597			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
598			    pkt_dst.s_addr)
599				goto ours;
600			if (ia->ia_netbroadcast.s_addr == pkt_dst.s_addr)
601				goto ours;
602#ifdef BOOTP_COMPAT
603			if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY)
604				goto ours;
605#endif
606		}
607	}
608	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
609		struct in_multi *inm;
610		if (ip_mrouter) {
611			/*
612			 * If we are acting as a multicast router, all
613			 * incoming multicast packets are passed to the
614			 * kernel-level multicast forwarding function.
615			 * The packet is returned (relatively) intact; if
616			 * ip_mforward() returns a non-zero value, the packet
617			 * must be discarded, else it may be accepted below.
618			 */
619			if (ip_mforward &&
620			    ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) {
621				ipstat.ips_cantforward++;
622				m_freem(m);
623				return;
624			}
625
626			/*
627			 * The process-level routing daemon needs to receive
628			 * all multicast IGMP packets, whether or not this
629			 * host belongs to their destination groups.
630			 */
631			if (ip->ip_p == IPPROTO_IGMP)
632				goto ours;
633			ipstat.ips_forward++;
634		}
635		/*
636		 * See if we belong to the destination multicast group on the
637		 * arrival interface.
638		 */
639		IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
640		if (inm == NULL) {
641			ipstat.ips_notmember++;
642			m_freem(m);
643			return;
644		}
645		goto ours;
646	}
647	if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
648		goto ours;
649	if (ip->ip_dst.s_addr == INADDR_ANY)
650		goto ours;
651
652	/*
653	 * FAITH(Firewall Aided Internet Translator)
654	 */
655	if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
656		if (ip_keepfaith) {
657			if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP)
658				goto ours;
659		}
660		m_freem(m);
661		return;
662	}
663
664	/*
665	 * Not for us; forward if possible and desirable.
666	 */
667	if (ipforwarding == 0) {
668		ipstat.ips_cantforward++;
669		m_freem(m);
670	} else {
671#ifdef IPSEC
672		/*
673		 * Enforce inbound IPsec SPD.
674		 */
675		if (ipsec4_in_reject(m, NULL)) {
676			ipsecstat.in_polvio++;
677			goto bad;
678		}
679#endif /* IPSEC */
680#ifdef FAST_IPSEC
681		mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
682		s = splnet();
683		if (mtag != NULL) {
684			tdbi = (struct tdb_ident *)(mtag + 1);
685			sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
686		} else {
687			sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
688						   IP_FORWARDING, &error);
689		}
690		if (sp == NULL) {	/* NB: can happen if error */
691			splx(s);
692			/*XXX error stat???*/
693			DPRINTF(("ip_input: no SP for forwarding\n"));	/*XXX*/
694			goto bad;
695		}
696
697		/*
698		 * Check security policy against packet attributes.
699		 */
700		error = ipsec_in_reject(sp, m);
701		KEY_FREESP(&sp);
702		splx(s);
703		if (error) {
704			ipstat.ips_cantforward++;
705			goto bad;
706		}
707#endif /* FAST_IPSEC */
708		ip_forward(m, 0, args.next_hop);
709	}
710	return;
711
712ours:
713#ifdef IPSTEALTH
714	/*
715	 * IPSTEALTH: Process non-routing options only
716	 * if the packet is destined for us.
717	 */
718	if (ipstealth && hlen > sizeof (struct ip) &&
719	    ip_dooptions(m, 1, args.next_hop))
720		return;
721#endif /* IPSTEALTH */
722
723	/* Count the packet in the ip address stats */
724	if (ia != NULL) {
725		ia->ia_ifa.if_ipackets++;
726		ia->ia_ifa.if_ibytes += m->m_pkthdr.len;
727	}
728
729	/*
730	 * If offset or IP_MF are set, must reassemble.
731	 * Otherwise, nothing need be done.
732	 * (We could look in the reassembly queue to see
733	 * if the packet was previously fragmented,
734	 * but it's not worth the time; just let them time out.)
735	 */
736	if (ip->ip_off & (IP_MF | IP_OFFMASK)) {
737
738		/* If maxnipq is 0, never accept fragments. */
739		if (maxnipq == 0) {
740                	ipstat.ips_fragments++;
741			ipstat.ips_fragdropped++;
742			goto bad;
743		}
744
745		sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
746		/*
747		 * Look for queue of fragments
748		 * of this datagram.
749		 */
750		TAILQ_FOREACH(fp, &ipq[sum], ipq_list)
751			if (ip->ip_id == fp->ipq_id &&
752			    ip->ip_src.s_addr == fp->ipq_src.s_addr &&
753			    ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
754#ifdef MAC
755			    mac_fragment_match(m, fp) &&
756#endif
757			    ip->ip_p == fp->ipq_p)
758				goto found;
759
760		fp = 0;
761
762		/*
763		 * Enforce upper bound on number of fragmented packets
764		 * for which we attempt reassembly;
765		 * If maxnipq is -1, accept all fragments without limitation.
766		 */
767		if ((nipq > maxnipq) && (maxnipq > 0)) {
768		    /*
769		     * drop something from the tail of the current queue
770		     * before proceeding further
771		     */
772		    struct ipq *q = TAILQ_LAST(&ipq[sum], ipqhead);
773		    if (q == NULL) {   /* gak */
774			for (i = 0; i < IPREASS_NHASH; i++) {
775			    struct ipq *r = TAILQ_LAST(&ipq[i], ipqhead);
776			    if (r) {
777				ip_freef(&ipq[i], r);
778				ipstat.ips_fragtimeout++;
779				break;
780			    }
781			}
782		    } else {
783			ip_freef(&ipq[sum], q);
784			ipstat.ips_fragtimeout++;
785		    }
786		}
787found:
788		/*
789		 * Adjust ip_len to not reflect header,
790		 * convert offset of this to bytes.
791		 */
792		ip->ip_len -= hlen;
793		if (ip->ip_off & IP_MF) {
794		        /*
795		         * Make sure that fragments have a data length
796			 * that's a non-zero multiple of 8 bytes.
797		         */
798			if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
799				ipstat.ips_toosmall++; /* XXX */
800				goto bad;
801			}
802			m->m_flags |= M_FRAG;
803		} else
804			m->m_flags &= ~M_FRAG;
805		ip->ip_off <<= 3;
806
807		/*
808		 * Attempt reassembly; if it succeeds, proceed.
809		 * ip_reass() will return a different mbuf, and update
810		 * the divert info in divert_info and args.divert_rule.
811		 */
812		ipstat.ips_fragments++;
813		m->m_pkthdr.header = ip;
814		m = ip_reass(m,
815		    &ipq[sum], fp, &divert_info, &args.divert_rule);
816		if (m == 0)
817			return;
818		ipstat.ips_reassembled++;
819		ip = mtod(m, struct ip *);
820		/* Get the header length of the reassembled packet */
821		hlen = ip->ip_hl << 2;
822#ifdef IPDIVERT
823		/* Restore original checksum before diverting packet */
824		if (divert_info != 0) {
825			ip->ip_len += hlen;
826			ip->ip_len = htons(ip->ip_len);
827			ip->ip_off = htons(ip->ip_off);
828			ip->ip_sum = 0;
829			if (hlen == sizeof(struct ip))
830				ip->ip_sum = in_cksum_hdr(ip);
831			else
832				ip->ip_sum = in_cksum(m, hlen);
833			ip->ip_off = ntohs(ip->ip_off);
834			ip->ip_len = ntohs(ip->ip_len);
835			ip->ip_len -= hlen;
836		}
837#endif
838	} else
839		ip->ip_len -= hlen;
840
841#ifdef IPDIVERT
842	/*
843	 * Divert or tee packet to the divert protocol if required.
844	 */
845	if (divert_info != 0) {
846		struct mbuf *clone = NULL;
847
848		/* Clone packet if we're doing a 'tee' */
849		if ((divert_info & IP_FW_PORT_TEE_FLAG) != 0)
850			clone = m_dup(m, M_DONTWAIT);
851
852		/* Restore packet header fields to original values */
853		ip->ip_len += hlen;
854		ip->ip_len = htons(ip->ip_len);
855		ip->ip_off = htons(ip->ip_off);
856
857		/* Deliver packet to divert input routine */
858		divert_packet(m, 1, divert_info & 0xffff, args.divert_rule);
859		ipstat.ips_delivered++;
860
861		/* If 'tee', continue with original packet */
862		if (clone == NULL)
863			return;
864		m = clone;
865		ip = mtod(m, struct ip *);
866		ip->ip_len += hlen;
867		/*
868		 * Jump backwards to complete processing of the
869		 * packet. But first clear divert_info to avoid
870		 * entering this block again.
871		 * We do not need to clear args.divert_rule
872		 * or args.next_hop as they will not be used.
873		 */
874		divert_info = 0;
875		goto pass;
876	}
877#endif
878
879#ifdef IPSEC
880	/*
881	 * enforce IPsec policy checking if we are seeing last header.
882	 * note that we do not visit this with protocols with pcb layer
883	 * code - like udp/tcp/raw ip.
884	 */
885	if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0 &&
886	    ipsec4_in_reject(m, NULL)) {
887		ipsecstat.in_polvio++;
888		goto bad;
889	}
890#endif
891#if FAST_IPSEC
892	/*
893	 * enforce IPsec policy checking if we are seeing last header.
894	 * note that we do not visit this with protocols with pcb layer
895	 * code - like udp/tcp/raw ip.
896	 */
897	if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) {
898		/*
899		 * Check if the packet has already had IPsec processing
900		 * done.  If so, then just pass it along.  This tag gets
901		 * set during AH, ESP, etc. input handling, before the
902		 * packet is returned to the ip input queue for delivery.
903		 */
904		mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
905		s = splnet();
906		if (mtag != NULL) {
907			tdbi = (struct tdb_ident *)(mtag + 1);
908			sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
909		} else {
910			sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
911						   IP_FORWARDING, &error);
912		}
913		if (sp != NULL) {
914			/*
915			 * Check security policy against packet attributes.
916			 */
917			error = ipsec_in_reject(sp, m);
918			KEY_FREESP(&sp);
919		} else {
920			/* XXX error stat??? */
921			error = EINVAL;
922DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/
923			goto bad;
924		}
925		splx(s);
926		if (error)
927			goto bad;
928	}
929#endif /* FAST_IPSEC */
930
931	/*
932	 * Switch out to protocol's input routine.
933	 */
934	ipstat.ips_delivered++;
935	if (args.next_hop && ip->ip_p == IPPROTO_TCP) {
936		/* TCP needs IPFORWARD info if available */
937		struct m_hdr tag;
938
939		tag.mh_type = MT_TAG;
940		tag.mh_flags = PACKET_TAG_IPFORWARD;
941		tag.mh_data = (caddr_t)args.next_hop;
942		tag.mh_next = m;
943
944		(*inetsw[ip_protox[ip->ip_p]].pr_input)(
945			(struct mbuf *)&tag, hlen);
946	} else
947		(*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen);
948	return;
949bad:
950	m_freem(m);
951}
952
953/*
954 * IP software interrupt routine - to go away sometime soon
955 */
956static void
957ipintr(void)
958{
959	struct mbuf *m;
960
961	while (1) {
962		IF_DEQUEUE(&ipintrq, m);
963		if (m == 0)
964			return;
965		ip_input(m);
966	}
967}
968
969/*
970 * Take incoming datagram fragment and try to reassemble it into
971 * whole datagram.  If a chain for reassembly of this datagram already
972 * exists, then it is given as fp; otherwise have to make a chain.
973 *
974 * When IPDIVERT enabled, keep additional state with each packet that
975 * tells us if we need to divert or tee the packet we're building.
976 * In particular, *divinfo includes the port and TEE flag,
977 * *divert_rule is the number of the matching rule.
978 */
979
980static struct mbuf *
981ip_reass(struct mbuf *m, struct ipqhead *head, struct ipq *fp,
982	u_int32_t *divinfo, u_int16_t *divert_rule)
983{
984	struct ip *ip = mtod(m, struct ip *);
985	register struct mbuf *p, *q, *nq;
986	struct mbuf *t;
987	int hlen = ip->ip_hl << 2;
988	int i, next;
989
990	/*
991	 * Presence of header sizes in mbufs
992	 * would confuse code below.
993	 */
994	m->m_data += hlen;
995	m->m_len -= hlen;
996
997	/*
998	 * If first fragment to arrive, create a reassembly queue.
999	 */
1000	if (fp == 0) {
1001		if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL)
1002			goto dropfrag;
1003		fp = mtod(t, struct ipq *);
1004#ifdef MAC
1005		mac_init_ipq(fp);
1006		mac_create_ipq(m, fp);
1007#endif
1008		TAILQ_INSERT_HEAD(head, fp, ipq_list);
1009		nipq++;
1010		fp->ipq_nfrags = 1;
1011		fp->ipq_ttl = IPFRAGTTL;
1012		fp->ipq_p = ip->ip_p;
1013		fp->ipq_id = ip->ip_id;
1014		fp->ipq_src = ip->ip_src;
1015		fp->ipq_dst = ip->ip_dst;
1016		fp->ipq_frags = m;
1017		m->m_nextpkt = NULL;
1018#ifdef IPDIVERT
1019		fp->ipq_div_info = 0;
1020		fp->ipq_div_cookie = 0;
1021#endif
1022		goto inserted;
1023	} else {
1024		fp->ipq_nfrags++;
1025#ifdef MAC
1026		mac_update_ipq(m, fp);
1027#endif
1028	}
1029
1030#define GETIP(m)	((struct ip*)((m)->m_pkthdr.header))
1031
1032	/*
1033	 * Find a segment which begins after this one does.
1034	 */
1035	for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt)
1036		if (GETIP(q)->ip_off > ip->ip_off)
1037			break;
1038
1039	/*
1040	 * If there is a preceding segment, it may provide some of
1041	 * our data already.  If so, drop the data from the incoming
1042	 * segment.  If it provides all of our data, drop us, otherwise
1043	 * stick new segment in the proper place.
1044	 *
1045	 * If some of the data is dropped from the the preceding
1046	 * segment, then it's checksum is invalidated.
1047	 */
1048	if (p) {
1049		i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off;
1050		if (i > 0) {
1051			if (i >= ip->ip_len)
1052				goto dropfrag;
1053			m_adj(m, i);
1054			m->m_pkthdr.csum_flags = 0;
1055			ip->ip_off += i;
1056			ip->ip_len -= i;
1057		}
1058		m->m_nextpkt = p->m_nextpkt;
1059		p->m_nextpkt = m;
1060	} else {
1061		m->m_nextpkt = fp->ipq_frags;
1062		fp->ipq_frags = m;
1063	}
1064
1065	/*
1066	 * While we overlap succeeding segments trim them or,
1067	 * if they are completely covered, dequeue them.
1068	 */
1069	for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off;
1070	     q = nq) {
1071		i = (ip->ip_off + ip->ip_len) -
1072		    GETIP(q)->ip_off;
1073		if (i < GETIP(q)->ip_len) {
1074			GETIP(q)->ip_len -= i;
1075			GETIP(q)->ip_off += i;
1076			m_adj(q, i);
1077			q->m_pkthdr.csum_flags = 0;
1078			break;
1079		}
1080		nq = q->m_nextpkt;
1081		m->m_nextpkt = nq;
1082		fp->ipq_nfrags--;
1083		m_freem(q);
1084	}
1085
1086inserted:
1087
1088#ifdef IPDIVERT
1089	/*
1090	 * Transfer firewall instructions to the fragment structure.
1091	 * Only trust info in the fragment at offset 0.
1092	 */
1093	if (ip->ip_off == 0) {
1094		fp->ipq_div_info = *divinfo;
1095		fp->ipq_div_cookie = *divert_rule;
1096	}
1097	*divinfo = 0;
1098	*divert_rule = 0;
1099#endif
1100
1101	/*
1102	 * Check for complete reassembly and perform frag per packet
1103	 * limiting.
1104	 *
1105	 * Frag limiting is performed here so that the nth frag has
1106	 * a chance to complete the packet before we drop the packet.
1107	 * As a result, n+1 frags are actually allowed per packet, but
1108	 * only n will ever be stored. (n = maxfragsperpacket.)
1109	 *
1110	 */
1111	next = 0;
1112	for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
1113		if (GETIP(q)->ip_off != next) {
1114			if (fp->ipq_nfrags > maxfragsperpacket)
1115				ip_freef(head, fp);
1116			return (0);
1117		}
1118		next += GETIP(q)->ip_len;
1119	}
1120	/* Make sure the last packet didn't have the IP_MF flag */
1121	if (p->m_flags & M_FRAG) {
1122		if (fp->ipq_nfrags > maxfragsperpacket)
1123			ip_freef(head, fp);
1124		return (0);
1125	}
1126
1127	/*
1128	 * Reassembly is complete.  Make sure the packet is a sane size.
1129	 */
1130	q = fp->ipq_frags;
1131	ip = GETIP(q);
1132	if (next + (ip->ip_hl << 2) > IP_MAXPACKET) {
1133		ipstat.ips_toolong++;
1134		ip_freef(head, fp);
1135		return (0);
1136	}
1137
1138	/*
1139	 * Concatenate fragments.
1140	 */
1141	m = q;
1142	t = m->m_next;
1143	m->m_next = 0;
1144	m_cat(m, t);
1145	nq = q->m_nextpkt;
1146	q->m_nextpkt = 0;
1147	for (q = nq; q != NULL; q = nq) {
1148		nq = q->m_nextpkt;
1149		q->m_nextpkt = NULL;
1150		m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
1151		m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
1152		m_cat(m, q);
1153	}
1154#ifdef MAC
1155	mac_create_datagram_from_ipq(fp, m);
1156	mac_destroy_ipq(fp);
1157#endif
1158
1159#ifdef IPDIVERT
1160	/*
1161	 * Extract firewall instructions from the fragment structure.
1162	 */
1163	*divinfo = fp->ipq_div_info;
1164	*divert_rule = fp->ipq_div_cookie;
1165#endif
1166
1167	/*
1168	 * Create header for new ip packet by
1169	 * modifying header of first packet;
1170	 * dequeue and discard fragment reassembly header.
1171	 * Make header visible.
1172	 */
1173	ip->ip_len = next;
1174	ip->ip_src = fp->ipq_src;
1175	ip->ip_dst = fp->ipq_dst;
1176	TAILQ_REMOVE(head, fp, ipq_list);
1177	nipq--;
1178	(void) m_free(dtom(fp));
1179	m->m_len += (ip->ip_hl << 2);
1180	m->m_data -= (ip->ip_hl << 2);
1181	/* some debugging cruft by sklower, below, will go away soon */
1182	if (m->m_flags & M_PKTHDR)	/* XXX this should be done elsewhere */
1183		m_fixhdr(m);
1184	return (m);
1185
1186dropfrag:
1187#ifdef IPDIVERT
1188	*divinfo = 0;
1189	*divert_rule = 0;
1190#endif
1191	ipstat.ips_fragdropped++;
1192	if (fp != 0)
1193		fp->ipq_nfrags--;
1194	m_freem(m);
1195	return (0);
1196
1197#undef GETIP
1198}
1199
1200/*
1201 * Free a fragment reassembly header and all
1202 * associated datagrams.
1203 */
1204static void
1205ip_freef(fhp, fp)
1206	struct ipqhead *fhp;
1207	struct ipq *fp;
1208{
1209	register struct mbuf *q;
1210
1211	while (fp->ipq_frags) {
1212		q = fp->ipq_frags;
1213		fp->ipq_frags = q->m_nextpkt;
1214		m_freem(q);
1215	}
1216	TAILQ_REMOVE(fhp, fp, ipq_list);
1217	(void) m_free(dtom(fp));
1218	nipq--;
1219}
1220
1221/*
1222 * IP timer processing;
1223 * if a timer expires on a reassembly
1224 * queue, discard it.
1225 */
1226void
1227ip_slowtimo()
1228{
1229	register struct ipq *fp;
1230	int s = splnet();
1231	int i;
1232
1233	for (i = 0; i < IPREASS_NHASH; i++) {
1234		for(fp = TAILQ_FIRST(&ipq[i]); fp;) {
1235			struct ipq *fpp;
1236
1237			fpp = fp;
1238			fp = TAILQ_NEXT(fp, ipq_list);
1239			if(--fpp->ipq_ttl == 0) {
1240				ipstat.ips_fragtimeout++;
1241				ip_freef(&ipq[i], fpp);
1242			}
1243		}
1244	}
1245	/*
1246	 * If we are over the maximum number of fragments
1247	 * (due to the limit being lowered), drain off
1248	 * enough to get down to the new limit.
1249	 */
1250	if (maxnipq > 0 && nipq > maxnipq) {
1251		for (i = 0; i < IPREASS_NHASH; i++) {
1252			while (nipq > maxnipq &&
1253				!TAILQ_EMPTY(&ipq[i])) {
1254				ipstat.ips_fragdropped++;
1255				ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i]));
1256			}
1257		}
1258	}
1259	ipflow_slowtimo();
1260	splx(s);
1261}
1262
1263/*
1264 * Drain off all datagram fragments.
1265 */
1266void
1267ip_drain()
1268{
1269	int     i;
1270
1271	for (i = 0; i < IPREASS_NHASH; i++) {
1272		while(!TAILQ_EMPTY(&ipq[i])) {
1273			ipstat.ips_fragdropped++;
1274			ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i]));
1275		}
1276	}
1277	in_rtqdrain();
1278}
1279
1280/*
1281 * Do option processing on a datagram,
1282 * possibly discarding it if bad options are encountered,
1283 * or forwarding it if source-routed.
1284 * The pass argument is used when operating in the IPSTEALTH
1285 * mode to tell what options to process:
1286 * [LS]SRR (pass 0) or the others (pass 1).
1287 * The reason for as many as two passes is that when doing IPSTEALTH,
1288 * non-routing options should be processed only if the packet is for us.
1289 * Returns 1 if packet has been forwarded/freed,
1290 * 0 if the packet should be processed further.
1291 */
1292static int
1293ip_dooptions(struct mbuf *m, int pass, struct sockaddr_in *next_hop)
1294{
1295	struct ip *ip = mtod(m, struct ip *);
1296	u_char *cp;
1297	struct in_ifaddr *ia;
1298	int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0;
1299	struct in_addr *sin, dst;
1300	n_time ntime;
1301	struct	sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET };
1302
1303	dst = ip->ip_dst;
1304	cp = (u_char *)(ip + 1);
1305	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
1306	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1307		opt = cp[IPOPT_OPTVAL];
1308		if (opt == IPOPT_EOL)
1309			break;
1310		if (opt == IPOPT_NOP)
1311			optlen = 1;
1312		else {
1313			if (cnt < IPOPT_OLEN + sizeof(*cp)) {
1314				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1315				goto bad;
1316			}
1317			optlen = cp[IPOPT_OLEN];
1318			if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
1319				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1320				goto bad;
1321			}
1322		}
1323		switch (opt) {
1324
1325		default:
1326			break;
1327
1328		/*
1329		 * Source routing with record.
1330		 * Find interface with current destination address.
1331		 * If none on this machine then drop if strictly routed,
1332		 * or do nothing if loosely routed.
1333		 * Record interface address and bring up next address
1334		 * component.  If strictly routed make sure next
1335		 * address is on directly accessible net.
1336		 */
1337		case IPOPT_LSRR:
1338		case IPOPT_SSRR:
1339#ifdef IPSTEALTH
1340			if (ipstealth && pass > 0)
1341				break;
1342#endif
1343			if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1344				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1345				goto bad;
1346			}
1347			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1348				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1349				goto bad;
1350			}
1351			ipaddr.sin_addr = ip->ip_dst;
1352			ia = (struct in_ifaddr *)
1353				ifa_ifwithaddr((struct sockaddr *)&ipaddr);
1354			if (ia == 0) {
1355				if (opt == IPOPT_SSRR) {
1356					type = ICMP_UNREACH;
1357					code = ICMP_UNREACH_SRCFAIL;
1358					goto bad;
1359				}
1360				if (!ip_dosourceroute)
1361					goto nosourcerouting;
1362				/*
1363				 * Loose routing, and not at next destination
1364				 * yet; nothing to do except forward.
1365				 */
1366				break;
1367			}
1368			off--;			/* 0 origin */
1369			if (off > optlen - (int)sizeof(struct in_addr)) {
1370				/*
1371				 * End of source route.  Should be for us.
1372				 */
1373				if (!ip_acceptsourceroute)
1374					goto nosourcerouting;
1375				save_rte(cp, ip->ip_src);
1376				break;
1377			}
1378#ifdef IPSTEALTH
1379			if (ipstealth)
1380				goto dropit;
1381#endif
1382			if (!ip_dosourceroute) {
1383				if (ipforwarding) {
1384					char buf[16]; /* aaa.bbb.ccc.ddd\0 */
1385					/*
1386					 * Acting as a router, so generate ICMP
1387					 */
1388nosourcerouting:
1389					strcpy(buf, inet_ntoa(ip->ip_dst));
1390					log(LOG_WARNING,
1391					    "attempted source route from %s to %s\n",
1392					    inet_ntoa(ip->ip_src), buf);
1393					type = ICMP_UNREACH;
1394					code = ICMP_UNREACH_SRCFAIL;
1395					goto bad;
1396				} else {
1397					/*
1398					 * Not acting as a router, so silently drop.
1399					 */
1400#ifdef IPSTEALTH
1401dropit:
1402#endif
1403					ipstat.ips_cantforward++;
1404					m_freem(m);
1405					return (1);
1406				}
1407			}
1408
1409			/*
1410			 * locate outgoing interface
1411			 */
1412			(void)memcpy(&ipaddr.sin_addr, cp + off,
1413			    sizeof(ipaddr.sin_addr));
1414
1415			if (opt == IPOPT_SSRR) {
1416#define	INA	struct in_ifaddr *
1417#define	SA	struct sockaddr *
1418			    if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == 0)
1419				ia = (INA)ifa_ifwithnet((SA)&ipaddr);
1420			} else
1421				ia = ip_rtaddr(ipaddr.sin_addr, &ipforward_rt);
1422			if (ia == 0) {
1423				type = ICMP_UNREACH;
1424				code = ICMP_UNREACH_SRCFAIL;
1425				goto bad;
1426			}
1427			ip->ip_dst = ipaddr.sin_addr;
1428			(void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
1429			    sizeof(struct in_addr));
1430			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1431			/*
1432			 * Let ip_intr's mcast routing check handle mcast pkts
1433			 */
1434			forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr));
1435			break;
1436
1437		case IPOPT_RR:
1438#ifdef IPSTEALTH
1439			if (ipstealth && pass == 0)
1440				break;
1441#endif
1442			if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1443				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1444				goto bad;
1445			}
1446			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1447				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1448				goto bad;
1449			}
1450			/*
1451			 * If no space remains, ignore.
1452			 */
1453			off--;			/* 0 origin */
1454			if (off > optlen - (int)sizeof(struct in_addr))
1455				break;
1456			(void)memcpy(&ipaddr.sin_addr, &ip->ip_dst,
1457			    sizeof(ipaddr.sin_addr));
1458			/*
1459			 * locate outgoing interface; if we're the destination,
1460			 * use the incoming interface (should be same).
1461			 */
1462			if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == 0 &&
1463			    (ia = ip_rtaddr(ipaddr.sin_addr,
1464			    &ipforward_rt)) == 0) {
1465				type = ICMP_UNREACH;
1466				code = ICMP_UNREACH_HOST;
1467				goto bad;
1468			}
1469			(void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
1470			    sizeof(struct in_addr));
1471			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1472			break;
1473
1474		case IPOPT_TS:
1475#ifdef IPSTEALTH
1476			if (ipstealth && pass == 0)
1477				break;
1478#endif
1479			code = cp - (u_char *)ip;
1480			if (optlen < 4 || optlen > 40) {
1481				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1482				goto bad;
1483			}
1484			if ((off = cp[IPOPT_OFFSET]) < 5) {
1485				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1486				goto bad;
1487			}
1488			if (off > optlen - (int)sizeof(int32_t)) {
1489				cp[IPOPT_OFFSET + 1] += (1 << 4);
1490				if ((cp[IPOPT_OFFSET + 1] & 0xf0) == 0) {
1491					code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1492					goto bad;
1493				}
1494				break;
1495			}
1496			off--;				/* 0 origin */
1497			sin = (struct in_addr *)(cp + off);
1498			switch (cp[IPOPT_OFFSET + 1] & 0x0f) {
1499
1500			case IPOPT_TS_TSONLY:
1501				break;
1502
1503			case IPOPT_TS_TSANDADDR:
1504				if (off + sizeof(n_time) +
1505				    sizeof(struct in_addr) > optlen) {
1506					code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1507					goto bad;
1508				}
1509				ipaddr.sin_addr = dst;
1510				ia = (INA)ifaof_ifpforaddr((SA)&ipaddr,
1511							    m->m_pkthdr.rcvif);
1512				if (ia == 0)
1513					continue;
1514				(void)memcpy(sin, &IA_SIN(ia)->sin_addr,
1515				    sizeof(struct in_addr));
1516				cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1517				off += sizeof(struct in_addr);
1518				break;
1519
1520			case IPOPT_TS_PRESPEC:
1521				if (off + sizeof(n_time) +
1522				    sizeof(struct in_addr) > optlen) {
1523					code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1524					goto bad;
1525				}
1526				(void)memcpy(&ipaddr.sin_addr, sin,
1527				    sizeof(struct in_addr));
1528				if (ifa_ifwithaddr((SA)&ipaddr) == 0)
1529					continue;
1530				cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1531				off += sizeof(struct in_addr);
1532				break;
1533
1534			default:
1535				code = &cp[IPOPT_OFFSET + 1] - (u_char *)ip;
1536				goto bad;
1537			}
1538			ntime = iptime();
1539			(void)memcpy(cp + off, &ntime, sizeof(n_time));
1540			cp[IPOPT_OFFSET] += sizeof(n_time);
1541		}
1542	}
1543	if (forward && ipforwarding) {
1544		ip_forward(m, 1, next_hop);
1545		return (1);
1546	}
1547	return (0);
1548bad:
1549	icmp_error(m, type, code, 0, 0);
1550	ipstat.ips_badoptions++;
1551	return (1);
1552}
1553
1554/*
1555 * Given address of next destination (final or next hop),
1556 * return internet address info of interface to be used to get there.
1557 */
1558struct in_ifaddr *
1559ip_rtaddr(dst, rt)
1560	struct in_addr dst;
1561	struct route *rt;
1562{
1563	register struct sockaddr_in *sin;
1564
1565	sin = (struct sockaddr_in *)&rt->ro_dst;
1566
1567	if (rt->ro_rt == 0 ||
1568	    !(rt->ro_rt->rt_flags & RTF_UP) ||
1569	    dst.s_addr != sin->sin_addr.s_addr) {
1570		if (rt->ro_rt) {
1571			RTFREE(rt->ro_rt);
1572			rt->ro_rt = 0;
1573		}
1574		sin->sin_family = AF_INET;
1575		sin->sin_len = sizeof(*sin);
1576		sin->sin_addr = dst;
1577
1578		rtalloc_ign(rt, RTF_PRCLONING);
1579	}
1580	if (rt->ro_rt == 0)
1581		return ((struct in_ifaddr *)0);
1582	return (ifatoia(rt->ro_rt->rt_ifa));
1583}
1584
1585/*
1586 * Save incoming source route for use in replies,
1587 * to be picked up later by ip_srcroute if the receiver is interested.
1588 */
1589static void
1590save_rte(option, dst)
1591	u_char *option;
1592	struct in_addr dst;
1593{
1594	unsigned olen;
1595
1596	olen = option[IPOPT_OLEN];
1597#ifdef DIAGNOSTIC
1598	if (ipprintfs)
1599		printf("save_rte: olen %d\n", olen);
1600#endif
1601	if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst)))
1602		return;
1603	bcopy(option, ip_srcrt.srcopt, olen);
1604	ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1605	ip_srcrt.dst = dst;
1606}
1607
1608/*
1609 * Retrieve incoming source route for use in replies,
1610 * in the same form used by setsockopt.
1611 * The first hop is placed before the options, will be removed later.
1612 */
1613struct mbuf *
1614ip_srcroute()
1615{
1616	register struct in_addr *p, *q;
1617	register struct mbuf *m;
1618
1619	if (ip_nhops == 0)
1620		return ((struct mbuf *)0);
1621	m = m_get(M_DONTWAIT, MT_HEADER);
1622	if (m == 0)
1623		return ((struct mbuf *)0);
1624
1625#define OPTSIZ	(sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt))
1626
1627	/* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
1628	m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) +
1629	    OPTSIZ;
1630#ifdef DIAGNOSTIC
1631	if (ipprintfs)
1632		printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len);
1633#endif
1634
1635	/*
1636	 * First save first hop for return route
1637	 */
1638	p = &ip_srcrt.route[ip_nhops - 1];
1639	*(mtod(m, struct in_addr *)) = *p--;
1640#ifdef DIAGNOSTIC
1641	if (ipprintfs)
1642		printf(" hops %lx", (u_long)ntohl(mtod(m, struct in_addr *)->s_addr));
1643#endif
1644
1645	/*
1646	 * Copy option fields and padding (nop) to mbuf.
1647	 */
1648	ip_srcrt.nop = IPOPT_NOP;
1649	ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
1650	(void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr),
1651	    &ip_srcrt.nop, OPTSIZ);
1652	q = (struct in_addr *)(mtod(m, caddr_t) +
1653	    sizeof(struct in_addr) + OPTSIZ);
1654#undef OPTSIZ
1655	/*
1656	 * Record return path as an IP source route,
1657	 * reversing the path (pointers are now aligned).
1658	 */
1659	while (p >= ip_srcrt.route) {
1660#ifdef DIAGNOSTIC
1661		if (ipprintfs)
1662			printf(" %lx", (u_long)ntohl(q->s_addr));
1663#endif
1664		*q++ = *p--;
1665	}
1666	/*
1667	 * Last hop goes to final destination.
1668	 */
1669	*q = ip_srcrt.dst;
1670#ifdef DIAGNOSTIC
1671	if (ipprintfs)
1672		printf(" %lx\n", (u_long)ntohl(q->s_addr));
1673#endif
1674	return (m);
1675}
1676
1677/*
1678 * Strip out IP options, at higher
1679 * level protocol in the kernel.
1680 * Second argument is buffer to which options
1681 * will be moved, and return value is their length.
1682 * XXX should be deleted; last arg currently ignored.
1683 */
1684void
1685ip_stripoptions(m, mopt)
1686	register struct mbuf *m;
1687	struct mbuf *mopt;
1688{
1689	register int i;
1690	struct ip *ip = mtod(m, struct ip *);
1691	register caddr_t opts;
1692	int olen;
1693
1694	olen = (ip->ip_hl << 2) - sizeof (struct ip);
1695	opts = (caddr_t)(ip + 1);
1696	i = m->m_len - (sizeof (struct ip) + olen);
1697	bcopy(opts + olen, opts, (unsigned)i);
1698	m->m_len -= olen;
1699	if (m->m_flags & M_PKTHDR)
1700		m->m_pkthdr.len -= olen;
1701	ip->ip_v = IPVERSION;
1702	ip->ip_hl = sizeof(struct ip) >> 2;
1703}
1704
1705u_char inetctlerrmap[PRC_NCMDS] = {
1706	0,		0,		0,		0,
1707	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1708	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1709	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1710	0,		0,		0,		0,
1711	ENOPROTOOPT,	ECONNREFUSED
1712};
1713
1714/*
1715 * Forward a packet.  If some error occurs return the sender
1716 * an icmp packet.  Note we can't always generate a meaningful
1717 * icmp message because icmp doesn't have a large enough repertoire
1718 * of codes and types.
1719 *
1720 * If not forwarding, just drop the packet.  This could be confusing
1721 * if ipforwarding was zero but some routing protocol was advancing
1722 * us as a gateway to somewhere.  However, we must let the routing
1723 * protocol deal with that.
1724 *
1725 * The srcrt parameter indicates whether the packet is being forwarded
1726 * via a source route.
1727 */
1728static void
1729ip_forward(struct mbuf *m, int srcrt, struct sockaddr_in *next_hop)
1730{
1731	struct ip *ip = mtod(m, struct ip *);
1732	struct rtentry *rt;
1733	int error, type = 0, code = 0;
1734	struct mbuf *mcopy;
1735	n_long dest;
1736	struct in_addr pkt_dst;
1737	struct ifnet *destifp;
1738#if defined(IPSEC) || defined(FAST_IPSEC)
1739	struct ifnet dummyifp;
1740#endif
1741
1742	dest = 0;
1743	/*
1744	 * Cache the destination address of the packet; this may be
1745	 * changed by use of 'ipfw fwd'.
1746	 */
1747	pkt_dst = next_hop ? next_hop->sin_addr : ip->ip_dst;
1748
1749#ifdef DIAGNOSTIC
1750	if (ipprintfs)
1751		printf("forward: src %lx dst %lx ttl %x\n",
1752		    (u_long)ip->ip_src.s_addr, (u_long)pkt_dst.s_addr,
1753		    ip->ip_ttl);
1754#endif
1755
1756
1757	if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(pkt_dst) == 0) {
1758		ipstat.ips_cantforward++;
1759		m_freem(m);
1760		return;
1761	}
1762#ifdef IPSTEALTH
1763	if (!ipstealth) {
1764#endif
1765		if (ip->ip_ttl <= IPTTLDEC) {
1766			icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS,
1767			    dest, 0);
1768			return;
1769		}
1770#ifdef IPSTEALTH
1771	}
1772#endif
1773
1774	if (ip_rtaddr(pkt_dst, &ipforward_rt) == 0) {
1775		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0);
1776		return;
1777	} else
1778		rt = ipforward_rt.ro_rt;
1779
1780	/*
1781	 * Save the IP header and at most 8 bytes of the payload,
1782	 * in case we need to generate an ICMP message to the src.
1783	 *
1784	 * XXX this can be optimized a lot by saving the data in a local
1785	 * buffer on the stack (72 bytes at most), and only allocating the
1786	 * mbuf if really necessary. The vast majority of the packets
1787	 * are forwarded without having to send an ICMP back (either
1788	 * because unnecessary, or because rate limited), so we are
1789	 * really we are wasting a lot of work here.
1790	 *
1791	 * We don't use m_copy() because it might return a reference
1792	 * to a shared cluster. Both this function and ip_output()
1793	 * assume exclusive access to the IP header in `m', so any
1794	 * data in a cluster may change before we reach icmp_error().
1795	 */
1796	MGET(mcopy, M_DONTWAIT, m->m_type);
1797	if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_DONTWAIT)) {
1798		/*
1799		 * It's probably ok if the pkthdr dup fails (because
1800		 * the deep copy of the tag chain failed), but for now
1801		 * be conservative and just discard the copy since
1802		 * code below may some day want the tags.
1803		 */
1804		m_free(mcopy);
1805		mcopy = NULL;
1806	}
1807	if (mcopy != NULL) {
1808		mcopy->m_len = imin((ip->ip_hl << 2) + 8,
1809		    (int)ip->ip_len);
1810		m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1811#ifdef MAC
1812		/*
1813		 * XXXMAC: This will eventually become an explicit
1814		 * labeling point.
1815		 */
1816		mac_create_mbuf_from_mbuf(m, mcopy);
1817#endif
1818	}
1819
1820#ifdef IPSTEALTH
1821	if (!ipstealth) {
1822#endif
1823		ip->ip_ttl -= IPTTLDEC;
1824#ifdef IPSTEALTH
1825	}
1826#endif
1827
1828	/*
1829	 * If forwarding packet using same interface that it came in on,
1830	 * perhaps should send a redirect to sender to shortcut a hop.
1831	 * Only send redirect if source is sending directly to us,
1832	 * and if packet was not source routed (or has any options).
1833	 * Also, don't send redirect if forwarding using a default route
1834	 * or a route modified by a redirect.
1835	 */
1836	if (rt->rt_ifp == m->m_pkthdr.rcvif &&
1837	    (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1838	    satosin(rt_key(rt))->sin_addr.s_addr != 0 &&
1839	    ipsendredirects && !srcrt && !next_hop) {
1840#define	RTA(rt)	((struct in_ifaddr *)(rt->rt_ifa))
1841		u_long src = ntohl(ip->ip_src.s_addr);
1842
1843		if (RTA(rt) &&
1844		    (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1845		    if (rt->rt_flags & RTF_GATEWAY)
1846			dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
1847		    else
1848			dest = pkt_dst.s_addr;
1849		    /* Router requirements says to only send host redirects */
1850		    type = ICMP_REDIRECT;
1851		    code = ICMP_REDIRECT_HOST;
1852#ifdef DIAGNOSTIC
1853		    if (ipprintfs)
1854		        printf("redirect (%d) to %lx\n", code, (u_long)dest);
1855#endif
1856		}
1857	}
1858
1859    {
1860	struct m_hdr tag;
1861
1862	if (next_hop) {
1863		/* Pass IPFORWARD info if available */
1864
1865		tag.mh_type = MT_TAG;
1866		tag.mh_flags = PACKET_TAG_IPFORWARD;
1867		tag.mh_data = (caddr_t)next_hop;
1868		tag.mh_next = m;
1869		m = (struct mbuf *)&tag;
1870	}
1871	error = ip_output(m, (struct mbuf *)0, &ipforward_rt,
1872			  IP_FORWARDING, 0, NULL);
1873    }
1874	if (error)
1875		ipstat.ips_cantforward++;
1876	else {
1877		ipstat.ips_forward++;
1878		if (type)
1879			ipstat.ips_redirectsent++;
1880		else {
1881			if (mcopy) {
1882				ipflow_create(&ipforward_rt, mcopy);
1883				m_freem(mcopy);
1884			}
1885			return;
1886		}
1887	}
1888	if (mcopy == NULL)
1889		return;
1890	destifp = NULL;
1891
1892	switch (error) {
1893
1894	case 0:				/* forwarded, but need redirect */
1895		/* type, code set above */
1896		break;
1897
1898	case ENETUNREACH:		/* shouldn't happen, checked above */
1899	case EHOSTUNREACH:
1900	case ENETDOWN:
1901	case EHOSTDOWN:
1902	default:
1903		type = ICMP_UNREACH;
1904		code = ICMP_UNREACH_HOST;
1905		break;
1906
1907	case EMSGSIZE:
1908		type = ICMP_UNREACH;
1909		code = ICMP_UNREACH_NEEDFRAG;
1910#ifdef IPSEC
1911		/*
1912		 * If the packet is routed over IPsec tunnel, tell the
1913		 * originator the tunnel MTU.
1914		 *	tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
1915		 * XXX quickhack!!!
1916		 */
1917		if (ipforward_rt.ro_rt) {
1918			struct secpolicy *sp = NULL;
1919			int ipsecerror;
1920			int ipsechdr;
1921			struct route *ro;
1922
1923			sp = ipsec4_getpolicybyaddr(mcopy,
1924						    IPSEC_DIR_OUTBOUND,
1925			                            IP_FORWARDING,
1926			                            &ipsecerror);
1927
1928			if (sp == NULL)
1929				destifp = ipforward_rt.ro_rt->rt_ifp;
1930			else {
1931				/* count IPsec header size */
1932				ipsechdr = ipsec4_hdrsiz(mcopy,
1933							 IPSEC_DIR_OUTBOUND,
1934							 NULL);
1935
1936				/*
1937				 * find the correct route for outer IPv4
1938				 * header, compute tunnel MTU.
1939				 *
1940				 * XXX BUG ALERT
1941				 * The "dummyifp" code relies upon the fact
1942				 * that icmp_error() touches only ifp->if_mtu.
1943				 */
1944				/*XXX*/
1945				destifp = NULL;
1946				if (sp->req != NULL
1947				 && sp->req->sav != NULL
1948				 && sp->req->sav->sah != NULL) {
1949					ro = &sp->req->sav->sah->sa_route;
1950					if (ro->ro_rt && ro->ro_rt->rt_ifp) {
1951						dummyifp.if_mtu =
1952						    ro->ro_rt->rt_ifp->if_mtu;
1953						dummyifp.if_mtu -= ipsechdr;
1954						destifp = &dummyifp;
1955					}
1956				}
1957
1958				key_freesp(sp);
1959			}
1960		}
1961#elif FAST_IPSEC
1962		/*
1963		 * If the packet is routed over IPsec tunnel, tell the
1964		 * originator the tunnel MTU.
1965		 *	tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
1966		 * XXX quickhack!!!
1967		 */
1968		if (ipforward_rt.ro_rt) {
1969			struct secpolicy *sp = NULL;
1970			int ipsecerror;
1971			int ipsechdr;
1972			struct route *ro;
1973
1974			sp = ipsec_getpolicybyaddr(mcopy,
1975						   IPSEC_DIR_OUTBOUND,
1976			                           IP_FORWARDING,
1977			                           &ipsecerror);
1978
1979			if (sp == NULL)
1980				destifp = ipforward_rt.ro_rt->rt_ifp;
1981			else {
1982				/* count IPsec header size */
1983				ipsechdr = ipsec4_hdrsiz(mcopy,
1984							 IPSEC_DIR_OUTBOUND,
1985							 NULL);
1986
1987				/*
1988				 * find the correct route for outer IPv4
1989				 * header, compute tunnel MTU.
1990				 *
1991				 * XXX BUG ALERT
1992				 * The "dummyifp" code relies upon the fact
1993				 * that icmp_error() touches only ifp->if_mtu.
1994				 */
1995				/*XXX*/
1996				destifp = NULL;
1997				if (sp->req != NULL
1998				 && sp->req->sav != NULL
1999				 && sp->req->sav->sah != NULL) {
2000					ro = &sp->req->sav->sah->sa_route;
2001					if (ro->ro_rt && ro->ro_rt->rt_ifp) {
2002						dummyifp.if_mtu =
2003						    ro->ro_rt->rt_ifp->if_mtu;
2004						dummyifp.if_mtu -= ipsechdr;
2005						destifp = &dummyifp;
2006					}
2007				}
2008
2009				KEY_FREESP(&sp);
2010			}
2011		}
2012#else /* !IPSEC && !FAST_IPSEC */
2013		if (ipforward_rt.ro_rt)
2014			destifp = ipforward_rt.ro_rt->rt_ifp;
2015#endif /*IPSEC*/
2016		ipstat.ips_cantfrag++;
2017		break;
2018
2019	case ENOBUFS:
2020		/*
2021		 * A router should not generate ICMP_SOURCEQUENCH as
2022		 * required in RFC1812 Requirements for IP Version 4 Routers.
2023		 * Source quench could be a big problem under DoS attacks,
2024		 * or if the underlying interface is rate-limited.
2025		 * Those who need source quench packets may re-enable them
2026		 * via the net.inet.ip.sendsourcequench sysctl.
2027		 */
2028		if (ip_sendsourcequench == 0) {
2029			m_freem(mcopy);
2030			return;
2031		} else {
2032			type = ICMP_SOURCEQUENCH;
2033			code = 0;
2034		}
2035		break;
2036
2037	case EACCES:			/* ipfw denied packet */
2038		m_freem(mcopy);
2039		return;
2040	}
2041	icmp_error(mcopy, type, code, dest, destifp);
2042}
2043
2044void
2045ip_savecontrol(inp, mp, ip, m)
2046	register struct inpcb *inp;
2047	register struct mbuf **mp;
2048	register struct ip *ip;
2049	register struct mbuf *m;
2050{
2051	if (inp->inp_socket->so_options & SO_TIMESTAMP) {
2052		struct timeval tv;
2053
2054		microtime(&tv);
2055		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
2056			SCM_TIMESTAMP, SOL_SOCKET);
2057		if (*mp)
2058			mp = &(*mp)->m_next;
2059	}
2060	if (inp->inp_flags & INP_RECVDSTADDR) {
2061		*mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
2062		    sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
2063		if (*mp)
2064			mp = &(*mp)->m_next;
2065	}
2066#ifdef notyet
2067	/* XXX
2068	 * Moving these out of udp_input() made them even more broken
2069	 * than they already were.
2070	 */
2071	/* options were tossed already */
2072	if (inp->inp_flags & INP_RECVOPTS) {
2073		*mp = sbcreatecontrol((caddr_t) opts_deleted_above,
2074		    sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
2075		if (*mp)
2076			mp = &(*mp)->m_next;
2077	}
2078	/* ip_srcroute doesn't do what we want here, need to fix */
2079	if (inp->inp_flags & INP_RECVRETOPTS) {
2080		*mp = sbcreatecontrol((caddr_t) ip_srcroute(),
2081		    sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
2082		if (*mp)
2083			mp = &(*mp)->m_next;
2084	}
2085#endif
2086	if (inp->inp_flags & INP_RECVIF) {
2087		struct ifnet *ifp;
2088		struct sdlbuf {
2089			struct sockaddr_dl sdl;
2090			u_char	pad[32];
2091		} sdlbuf;
2092		struct sockaddr_dl *sdp;
2093		struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
2094
2095		if (((ifp = m->m_pkthdr.rcvif))
2096		&& ( ifp->if_index && (ifp->if_index <= if_index))) {
2097			sdp = (struct sockaddr_dl *)
2098			    (ifaddr_byindex(ifp->if_index)->ifa_addr);
2099			/*
2100			 * Change our mind and don't try copy.
2101			 */
2102			if ((sdp->sdl_family != AF_LINK)
2103			|| (sdp->sdl_len > sizeof(sdlbuf))) {
2104				goto makedummy;
2105			}
2106			bcopy(sdp, sdl2, sdp->sdl_len);
2107		} else {
2108makedummy:
2109			sdl2->sdl_len
2110				= offsetof(struct sockaddr_dl, sdl_data[0]);
2111			sdl2->sdl_family = AF_LINK;
2112			sdl2->sdl_index = 0;
2113			sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
2114		}
2115		*mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len,
2116			IP_RECVIF, IPPROTO_IP);
2117		if (*mp)
2118			mp = &(*mp)->m_next;
2119	}
2120}
2121
2122/*
2123 * XXX these routines are called from the upper part of the kernel.
2124 * They need to be locked when we remove Giant.
2125 *
2126 * They could also be moved to ip_mroute.c, since all the RSVP
2127 *  handling is done there already.
2128 */
2129static int ip_rsvp_on;
2130struct socket *ip_rsvpd;
2131int
2132ip_rsvp_init(struct socket *so)
2133{
2134	if (so->so_type != SOCK_RAW ||
2135	    so->so_proto->pr_protocol != IPPROTO_RSVP)
2136		return EOPNOTSUPP;
2137
2138	if (ip_rsvpd != NULL)
2139		return EADDRINUSE;
2140
2141	ip_rsvpd = so;
2142	/*
2143	 * This may seem silly, but we need to be sure we don't over-increment
2144	 * the RSVP counter, in case something slips up.
2145	 */
2146	if (!ip_rsvp_on) {
2147		ip_rsvp_on = 1;
2148		rsvp_on++;
2149	}
2150
2151	return 0;
2152}
2153
2154int
2155ip_rsvp_done(void)
2156{
2157	ip_rsvpd = NULL;
2158	/*
2159	 * This may seem silly, but we need to be sure we don't over-decrement
2160	 * the RSVP counter, in case something slips up.
2161	 */
2162	if (ip_rsvp_on) {
2163		ip_rsvp_on = 0;
2164		rsvp_on--;
2165	}
2166	return 0;
2167}
2168
2169void
2170rsvp_input(struct mbuf *m, int off)	/* XXX must fixup manually */
2171{
2172	if (rsvp_input_p) { /* call the real one if loaded */
2173		rsvp_input_p(m, off);
2174		return;
2175	}
2176
2177	/* Can still get packets with rsvp_on = 0 if there is a local member
2178	 * of the group to which the RSVP packet is addressed.  But in this
2179	 * case we want to throw the packet away.
2180	 */
2181
2182	if (!rsvp_on) {
2183		m_freem(m);
2184		return;
2185	}
2186
2187	if (ip_rsvpd != NULL) {
2188		rip_input(m, off);
2189		return;
2190	}
2191	/* Drop the packet */
2192	m_freem(m);
2193}
2194