ip_reass.c revision 142215
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 142215 2005-02-22 13:04:05Z glebius $
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	dchg = (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL);
472#endif /* IPFIREWALL_FORWARD */
473
474passin:
475	/*
476	 * Process options and, if not destined for us,
477	 * ship it on.  ip_dooptions returns 1 when an
478	 * error was detected (causing an icmp message
479	 * to be sent and the original packet to be freed).
480	 */
481	if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
482		return;
483
484        /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
485         * matter if it is destined to another node, or whether it is
486         * a multicast one, RSVP wants it! and prevents it from being forwarded
487         * anywhere else. Also checks if the rsvp daemon is running before
488	 * grabbing the packet.
489         */
490	if (rsvp_on && ip->ip_p==IPPROTO_RSVP)
491		goto ours;
492
493	/*
494	 * Check our list of addresses, to see if the packet is for us.
495	 * If we don't have any addresses, assume any unicast packet
496	 * we receive might be for us (and let the upper layers deal
497	 * with it).
498	 */
499	if (TAILQ_EMPTY(&in_ifaddrhead) &&
500	    (m->m_flags & (M_MCAST|M_BCAST)) == 0)
501		goto ours;
502
503	/*
504	 * Enable a consistency check between the destination address
505	 * and the arrival interface for a unicast packet (the RFC 1122
506	 * strong ES model) if IP forwarding is disabled and the packet
507	 * is not locally generated and the packet is not subject to
508	 * 'ipfw fwd'.
509	 *
510	 * XXX - Checking also should be disabled if the destination
511	 * address is ipnat'ed to a different interface.
512	 *
513	 * XXX - Checking is incompatible with IP aliases added
514	 * to the loopback interface instead of the interface where
515	 * the packets are received.
516	 *
517	 * XXX - This is the case for carp vhost IPs as well so we
518	 * insert a workaround. If the packet got here, we already
519	 * checked with carp_iamatch() and carp_forus().
520	 */
521	checkif = ip_checkinterface && (ipforwarding == 0) &&
522	    m->m_pkthdr.rcvif != NULL &&
523	    ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) &&
524#ifdef DEV_CARP
525	    !m->m_pkthdr.rcvif->if_carp &&
526#endif
527	    (dchg == 0);
528
529	/*
530	 * Check for exact addresses in the hash bucket.
531	 */
532	LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
533		/*
534		 * If the address matches, verify that the packet
535		 * arrived via the correct interface if checking is
536		 * enabled.
537		 */
538		if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr &&
539		    (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif))
540			goto ours;
541	}
542	/*
543	 * Check for broadcast addresses.
544	 *
545	 * Only accept broadcast packets that arrive via the matching
546	 * interface.  Reception of forwarded directed broadcasts would
547	 * be handled via ip_forward() and ether_output() with the loopback
548	 * into the stack for SIMPLEX interfaces handled by ether_output().
549	 */
550	if (m->m_pkthdr.rcvif != NULL &&
551	    m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
552	        TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
553			if (ifa->ifa_addr->sa_family != AF_INET)
554				continue;
555			ia = ifatoia(ifa);
556			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
557			    ip->ip_dst.s_addr)
558				goto ours;
559			if (ia->ia_netbroadcast.s_addr == ip->ip_dst.s_addr)
560				goto ours;
561#ifdef BOOTP_COMPAT
562			if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY)
563				goto ours;
564#endif
565		}
566	}
567	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
568		struct in_multi *inm;
569		if (ip_mrouter) {
570			/*
571			 * If we are acting as a multicast router, all
572			 * incoming multicast packets are passed to the
573			 * kernel-level multicast forwarding function.
574			 * The packet is returned (relatively) intact; if
575			 * ip_mforward() returns a non-zero value, the packet
576			 * must be discarded, else it may be accepted below.
577			 */
578			if (ip_mforward &&
579			    ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) {
580				ipstat.ips_cantforward++;
581				m_freem(m);
582				return;
583			}
584
585			/*
586			 * The process-level routing daemon needs to receive
587			 * all multicast IGMP packets, whether or not this
588			 * host belongs to their destination groups.
589			 */
590			if (ip->ip_p == IPPROTO_IGMP)
591				goto ours;
592			ipstat.ips_forward++;
593		}
594		/*
595		 * See if we belong to the destination multicast group on the
596		 * arrival interface.
597		 */
598		IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
599		if (inm == NULL) {
600			ipstat.ips_notmember++;
601			m_freem(m);
602			return;
603		}
604		goto ours;
605	}
606	if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
607		goto ours;
608	if (ip->ip_dst.s_addr == INADDR_ANY)
609		goto ours;
610
611	/*
612	 * FAITH(Firewall Aided Internet Translator)
613	 */
614	if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
615		if (ip_keepfaith) {
616			if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP)
617				goto ours;
618		}
619		m_freem(m);
620		return;
621	}
622
623	/*
624	 * Not for us; forward if possible and desirable.
625	 */
626	if (ipforwarding == 0) {
627		ipstat.ips_cantforward++;
628		m_freem(m);
629	} else {
630#ifdef IPSEC
631		/*
632		 * Enforce inbound IPsec SPD.
633		 */
634		if (ipsec4_in_reject(m, NULL)) {
635			ipsecstat.in_polvio++;
636			goto bad;
637		}
638#endif /* IPSEC */
639#ifdef FAST_IPSEC
640		mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
641		s = splnet();
642		if (mtag != NULL) {
643			tdbi = (struct tdb_ident *)(mtag + 1);
644			sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
645		} else {
646			sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
647						   IP_FORWARDING, &error);
648		}
649		if (sp == NULL) {	/* NB: can happen if error */
650			splx(s);
651			/*XXX error stat???*/
652			DPRINTF(("ip_input: no SP for forwarding\n"));	/*XXX*/
653			goto bad;
654		}
655
656		/*
657		 * Check security policy against packet attributes.
658		 */
659		error = ipsec_in_reject(sp, m);
660		KEY_FREESP(&sp);
661		splx(s);
662		if (error) {
663			ipstat.ips_cantforward++;
664			goto bad;
665		}
666#endif /* FAST_IPSEC */
667		ip_forward(m, dchg);
668	}
669	return;
670
671ours:
672#ifdef IPSTEALTH
673	/*
674	 * IPSTEALTH: Process non-routing options only
675	 * if the packet is destined for us.
676	 */
677	if (ipstealth && hlen > sizeof (struct ip) &&
678	    ip_dooptions(m, 1))
679		return;
680#endif /* IPSTEALTH */
681
682	/* Count the packet in the ip address stats */
683	if (ia != NULL) {
684		ia->ia_ifa.if_ipackets++;
685		ia->ia_ifa.if_ibytes += m->m_pkthdr.len;
686	}
687
688	/*
689	 * Attempt reassembly; if it succeeds, proceed.
690	 * ip_reass() will return a different mbuf.
691	 */
692	if (ip->ip_off & (IP_MF | IP_OFFMASK)) {
693		m = ip_reass(m);
694		if (m == NULL)
695			return;
696		ip = mtod(m, struct ip *);
697		/* Get the header length of the reassembled packet */
698		hlen = ip->ip_hl << 2;
699	}
700
701	/*
702	 * Further protocols expect the packet length to be w/o the
703	 * IP header.
704	 */
705	ip->ip_len -= hlen;
706
707#ifdef IPSEC
708	/*
709	 * enforce IPsec policy checking if we are seeing last header.
710	 * note that we do not visit this with protocols with pcb layer
711	 * code - like udp/tcp/raw ip.
712	 */
713	if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0 &&
714	    ipsec4_in_reject(m, NULL)) {
715		ipsecstat.in_polvio++;
716		goto bad;
717	}
718#endif
719#if FAST_IPSEC
720	/*
721	 * enforce IPsec policy checking if we are seeing last header.
722	 * note that we do not visit this with protocols with pcb layer
723	 * code - like udp/tcp/raw ip.
724	 */
725	if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) {
726		/*
727		 * Check if the packet has already had IPsec processing
728		 * done.  If so, then just pass it along.  This tag gets
729		 * set during AH, ESP, etc. input handling, before the
730		 * packet is returned to the ip input queue for delivery.
731		 */
732		mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
733		s = splnet();
734		if (mtag != NULL) {
735			tdbi = (struct tdb_ident *)(mtag + 1);
736			sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
737		} else {
738			sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
739						   IP_FORWARDING, &error);
740		}
741		if (sp != NULL) {
742			/*
743			 * Check security policy against packet attributes.
744			 */
745			error = ipsec_in_reject(sp, m);
746			KEY_FREESP(&sp);
747		} else {
748			/* XXX error stat??? */
749			error = EINVAL;
750DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/
751			goto bad;
752		}
753		splx(s);
754		if (error)
755			goto bad;
756	}
757#endif /* FAST_IPSEC */
758
759	/*
760	 * Switch out to protocol's input routine.
761	 */
762	ipstat.ips_delivered++;
763
764	(*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen);
765	return;
766bad:
767	m_freem(m);
768}
769
770/*
771 * Take incoming datagram fragment and try to reassemble it into
772 * whole datagram.  If the argument is the first fragment or one
773 * in between the function will return NULL and store the mbuf
774 * in the fragment chain.  If the argument is the last fragment
775 * the packet will be reassembled and the pointer to the new
776 * mbuf returned for further processing.  Only m_tags attached
777 * to the first packet/fragment are preserved.
778 * The IP header is *NOT* adjusted out of iplen.
779 */
780
781struct mbuf *
782ip_reass(struct mbuf *m)
783{
784	struct ip *ip;
785	struct mbuf *p, *q, *nq, *t;
786	struct ipq *fp = NULL;
787	struct ipqhead *head;
788	int i, hlen, next;
789	u_int8_t ecn, ecn0;
790	u_short hash;
791
792	/* If maxnipq is 0, never accept fragments. */
793	if (maxnipq == 0) {
794		ipstat.ips_fragments++;
795		ipstat.ips_fragdropped++;
796		m_freem(m);
797		return (NULL);
798	}
799
800	ip = mtod(m, struct ip *);
801	hlen = ip->ip_hl << 2;
802
803	hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
804	head = &ipq[hash];
805	IPQ_LOCK();
806
807	/*
808	 * Look for queue of fragments
809	 * of this datagram.
810	 */
811	TAILQ_FOREACH(fp, head, ipq_list)
812		if (ip->ip_id == fp->ipq_id &&
813		    ip->ip_src.s_addr == fp->ipq_src.s_addr &&
814		    ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
815#ifdef MAC
816		    mac_fragment_match(m, fp) &&
817#endif
818		    ip->ip_p == fp->ipq_p)
819			goto found;
820
821	fp = NULL;
822
823	/*
824	 * Enforce upper bound on number of fragmented packets
825	 * for which we attempt reassembly;
826	 * If maxnipq is -1, accept all fragments without limitation.
827	 */
828	if ((nipq > maxnipq) && (maxnipq > 0)) {
829		/*
830		 * drop something from the tail of the current queue
831		 * before proceeding further
832		 */
833		struct ipq *q = TAILQ_LAST(head, ipqhead);
834		if (q == NULL) {   /* gak */
835			for (i = 0; i < IPREASS_NHASH; i++) {
836				struct ipq *r = TAILQ_LAST(&ipq[i], ipqhead);
837				if (r) {
838					ipstat.ips_fragtimeout += r->ipq_nfrags;
839					ip_freef(&ipq[i], r);
840					break;
841				}
842			}
843		} else {
844			ipstat.ips_fragtimeout += q->ipq_nfrags;
845			ip_freef(head, q);
846		}
847	}
848
849found:
850	/*
851	 * Adjust ip_len to not reflect header,
852	 * convert offset of this to bytes.
853	 */
854	ip->ip_len -= hlen;
855	if (ip->ip_off & IP_MF) {
856		/*
857		 * Make sure that fragments have a data length
858		 * that's a non-zero multiple of 8 bytes.
859		 */
860		if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
861			ipstat.ips_toosmall++; /* XXX */
862			goto dropfrag;
863		}
864		m->m_flags |= M_FRAG;
865	} else
866		m->m_flags &= ~M_FRAG;
867	ip->ip_off <<= 3;
868
869
870	/*
871	 * Attempt reassembly; if it succeeds, proceed.
872	 * ip_reass() will return a different mbuf.
873	 */
874	ipstat.ips_fragments++;
875	m->m_pkthdr.header = ip;
876
877	/* Previous ip_reass() started here. */
878	/*
879	 * Presence of header sizes in mbufs
880	 * would confuse code below.
881	 */
882	m->m_data += hlen;
883	m->m_len -= hlen;
884
885	/*
886	 * If first fragment to arrive, create a reassembly queue.
887	 */
888	if (fp == NULL) {
889		if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL)
890			goto dropfrag;
891		fp = mtod(t, struct ipq *);
892#ifdef MAC
893		if (mac_init_ipq(fp, M_NOWAIT) != 0) {
894			m_free(t);
895			goto dropfrag;
896		}
897		mac_create_ipq(m, fp);
898#endif
899		TAILQ_INSERT_HEAD(head, fp, ipq_list);
900		nipq++;
901		fp->ipq_nfrags = 1;
902		fp->ipq_ttl = IPFRAGTTL;
903		fp->ipq_p = ip->ip_p;
904		fp->ipq_id = ip->ip_id;
905		fp->ipq_src = ip->ip_src;
906		fp->ipq_dst = ip->ip_dst;
907		fp->ipq_frags = m;
908		m->m_nextpkt = NULL;
909		goto inserted;
910	} else {
911		fp->ipq_nfrags++;
912#ifdef MAC
913		mac_update_ipq(m, fp);
914#endif
915	}
916
917#define GETIP(m)	((struct ip*)((m)->m_pkthdr.header))
918
919	/*
920	 * Handle ECN by comparing this segment with the first one;
921	 * if CE is set, do not lose CE.
922	 * drop if CE and not-ECT are mixed for the same packet.
923	 */
924	ecn = ip->ip_tos & IPTOS_ECN_MASK;
925	ecn0 = GETIP(fp->ipq_frags)->ip_tos & IPTOS_ECN_MASK;
926	if (ecn == IPTOS_ECN_CE) {
927		if (ecn0 == IPTOS_ECN_NOTECT)
928			goto dropfrag;
929		if (ecn0 != IPTOS_ECN_CE)
930			GETIP(fp->ipq_frags)->ip_tos |= IPTOS_ECN_CE;
931	}
932	if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT)
933		goto dropfrag;
934
935	/*
936	 * Find a segment which begins after this one does.
937	 */
938	for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt)
939		if (GETIP(q)->ip_off > ip->ip_off)
940			break;
941
942	/*
943	 * If there is a preceding segment, it may provide some of
944	 * our data already.  If so, drop the data from the incoming
945	 * segment.  If it provides all of our data, drop us, otherwise
946	 * stick new segment in the proper place.
947	 *
948	 * If some of the data is dropped from the the preceding
949	 * segment, then it's checksum is invalidated.
950	 */
951	if (p) {
952		i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off;
953		if (i > 0) {
954			if (i >= ip->ip_len)
955				goto dropfrag;
956			m_adj(m, i);
957			m->m_pkthdr.csum_flags = 0;
958			ip->ip_off += i;
959			ip->ip_len -= i;
960		}
961		m->m_nextpkt = p->m_nextpkt;
962		p->m_nextpkt = m;
963	} else {
964		m->m_nextpkt = fp->ipq_frags;
965		fp->ipq_frags = m;
966	}
967
968	/*
969	 * While we overlap succeeding segments trim them or,
970	 * if they are completely covered, dequeue them.
971	 */
972	for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off;
973	     q = nq) {
974		i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off;
975		if (i < GETIP(q)->ip_len) {
976			GETIP(q)->ip_len -= i;
977			GETIP(q)->ip_off += i;
978			m_adj(q, i);
979			q->m_pkthdr.csum_flags = 0;
980			break;
981		}
982		nq = q->m_nextpkt;
983		m->m_nextpkt = nq;
984		ipstat.ips_fragdropped++;
985		fp->ipq_nfrags--;
986		m_freem(q);
987	}
988
989inserted:
990
991	/*
992	 * Check for complete reassembly and perform frag per packet
993	 * limiting.
994	 *
995	 * Frag limiting is performed here so that the nth frag has
996	 * a chance to complete the packet before we drop the packet.
997	 * As a result, n+1 frags are actually allowed per packet, but
998	 * only n will ever be stored. (n = maxfragsperpacket.)
999	 *
1000	 */
1001	next = 0;
1002	for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
1003		if (GETIP(q)->ip_off != next) {
1004			if (fp->ipq_nfrags > maxfragsperpacket) {
1005				ipstat.ips_fragdropped += fp->ipq_nfrags;
1006				ip_freef(head, fp);
1007			}
1008			goto done;
1009		}
1010		next += GETIP(q)->ip_len;
1011	}
1012	/* Make sure the last packet didn't have the IP_MF flag */
1013	if (p->m_flags & M_FRAG) {
1014		if (fp->ipq_nfrags > maxfragsperpacket) {
1015			ipstat.ips_fragdropped += fp->ipq_nfrags;
1016			ip_freef(head, fp);
1017		}
1018		goto done;
1019	}
1020
1021	/*
1022	 * Reassembly is complete.  Make sure the packet is a sane size.
1023	 */
1024	q = fp->ipq_frags;
1025	ip = GETIP(q);
1026	if (next + (ip->ip_hl << 2) > IP_MAXPACKET) {
1027		ipstat.ips_toolong++;
1028		ipstat.ips_fragdropped += fp->ipq_nfrags;
1029		ip_freef(head, fp);
1030		goto done;
1031	}
1032
1033	/*
1034	 * Concatenate fragments.
1035	 */
1036	m = q;
1037	t = m->m_next;
1038	m->m_next = NULL;
1039	m_cat(m, t);
1040	nq = q->m_nextpkt;
1041	q->m_nextpkt = NULL;
1042	for (q = nq; q != NULL; q = nq) {
1043		nq = q->m_nextpkt;
1044		q->m_nextpkt = NULL;
1045		m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
1046		m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
1047		m_cat(m, q);
1048	}
1049#ifdef MAC
1050	mac_create_datagram_from_ipq(fp, m);
1051	mac_destroy_ipq(fp);
1052#endif
1053
1054	/*
1055	 * Create header for new ip packet by modifying header of first
1056	 * packet;  dequeue and discard fragment reassembly header.
1057	 * Make header visible.
1058	 */
1059	ip->ip_len = (ip->ip_hl << 2) + next;
1060	ip->ip_src = fp->ipq_src;
1061	ip->ip_dst = fp->ipq_dst;
1062	TAILQ_REMOVE(head, fp, ipq_list);
1063	nipq--;
1064	(void) m_free(dtom(fp));
1065	m->m_len += (ip->ip_hl << 2);
1066	m->m_data -= (ip->ip_hl << 2);
1067	/* some debugging cruft by sklower, below, will go away soon */
1068	if (m->m_flags & M_PKTHDR)	/* XXX this should be done elsewhere */
1069		m_fixhdr(m);
1070	ipstat.ips_reassembled++;
1071	IPQ_UNLOCK();
1072	return (m);
1073
1074dropfrag:
1075	ipstat.ips_fragdropped++;
1076	if (fp != NULL)
1077		fp->ipq_nfrags--;
1078	m_freem(m);
1079done:
1080	IPQ_UNLOCK();
1081	return (NULL);
1082
1083#undef GETIP
1084}
1085
1086/*
1087 * Free a fragment reassembly header and all
1088 * associated datagrams.
1089 */
1090static void
1091ip_freef(fhp, fp)
1092	struct ipqhead *fhp;
1093	struct ipq *fp;
1094{
1095	register struct mbuf *q;
1096
1097	IPQ_LOCK_ASSERT();
1098
1099	while (fp->ipq_frags) {
1100		q = fp->ipq_frags;
1101		fp->ipq_frags = q->m_nextpkt;
1102		m_freem(q);
1103	}
1104	TAILQ_REMOVE(fhp, fp, ipq_list);
1105	(void) m_free(dtom(fp));
1106	nipq--;
1107}
1108
1109/*
1110 * IP timer processing;
1111 * if a timer expires on a reassembly
1112 * queue, discard it.
1113 */
1114void
1115ip_slowtimo()
1116{
1117	register struct ipq *fp;
1118	int s = splnet();
1119	int i;
1120
1121	IPQ_LOCK();
1122	for (i = 0; i < IPREASS_NHASH; i++) {
1123		for(fp = TAILQ_FIRST(&ipq[i]); fp;) {
1124			struct ipq *fpp;
1125
1126			fpp = fp;
1127			fp = TAILQ_NEXT(fp, ipq_list);
1128			if(--fpp->ipq_ttl == 0) {
1129				ipstat.ips_fragtimeout += fpp->ipq_nfrags;
1130				ip_freef(&ipq[i], fpp);
1131			}
1132		}
1133	}
1134	/*
1135	 * If we are over the maximum number of fragments
1136	 * (due to the limit being lowered), drain off
1137	 * enough to get down to the new limit.
1138	 */
1139	if (maxnipq >= 0 && nipq > maxnipq) {
1140		for (i = 0; i < IPREASS_NHASH; i++) {
1141			while (nipq > maxnipq && !TAILQ_EMPTY(&ipq[i])) {
1142				ipstat.ips_fragdropped +=
1143				    TAILQ_FIRST(&ipq[i])->ipq_nfrags;
1144				ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i]));
1145			}
1146		}
1147	}
1148	IPQ_UNLOCK();
1149	splx(s);
1150}
1151
1152/*
1153 * Drain off all datagram fragments.
1154 */
1155void
1156ip_drain()
1157{
1158	int     i;
1159
1160	IPQ_LOCK();
1161	for (i = 0; i < IPREASS_NHASH; i++) {
1162		while(!TAILQ_EMPTY(&ipq[i])) {
1163			ipstat.ips_fragdropped +=
1164			    TAILQ_FIRST(&ipq[i])->ipq_nfrags;
1165			ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i]));
1166		}
1167	}
1168	IPQ_UNLOCK();
1169	in_rtqdrain();
1170}
1171
1172/*
1173 * The protocol to be inserted into ip_protox[] must be already registered
1174 * in inetsw[], either statically or through pf_proto_register().
1175 */
1176int
1177ipproto_register(u_char ipproto)
1178{
1179	struct protosw *pr;
1180
1181	/* Sanity checks. */
1182	if (ipproto == 0)
1183		return (EPROTONOSUPPORT);
1184
1185	/*
1186	 * The protocol slot must not be occupied by another protocol
1187	 * already.  An index pointing to IPPROTO_RAW is unused.
1188	 */
1189	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
1190	if (pr == NULL)
1191		return (EPFNOSUPPORT);
1192	if (ip_protox[ipproto] != pr - inetsw)	/* IPPROTO_RAW */
1193		return (EEXIST);
1194
1195	/* Find the protocol position in inetsw[] and set the index. */
1196	for (pr = inetdomain.dom_protosw;
1197	     pr < inetdomain.dom_protoswNPROTOSW; pr++) {
1198		if (pr->pr_domain->dom_family == PF_INET &&
1199		    pr->pr_protocol && pr->pr_protocol == ipproto) {
1200			/* Be careful to only index valid IP protocols. */
1201			if (pr->pr_protocol <= IPPROTO_MAX) {
1202				ip_protox[pr->pr_protocol] = pr - inetsw;
1203				return (0);
1204			} else
1205				return (EINVAL);
1206		}
1207	}
1208	return (EPROTONOSUPPORT);
1209}
1210
1211int
1212ipproto_unregister(u_char ipproto)
1213{
1214	struct protosw *pr;
1215
1216	/* Sanity checks. */
1217	if (ipproto == 0)
1218		return (EPROTONOSUPPORT);
1219
1220	/* Check if the protocol was indeed registered. */
1221	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
1222	if (pr == NULL)
1223		return (EPFNOSUPPORT);
1224	if (ip_protox[ipproto] == pr - inetsw)  /* IPPROTO_RAW */
1225		return (ENOENT);
1226
1227	/* Reset the protocol slot to IPPROTO_RAW. */
1228	ip_protox[ipproto] = pr - inetsw;
1229	return (0);
1230}
1231
1232
1233/*
1234 * Do option processing on a datagram,
1235 * possibly discarding it if bad options are encountered,
1236 * or forwarding it if source-routed.
1237 * The pass argument is used when operating in the IPSTEALTH
1238 * mode to tell what options to process:
1239 * [LS]SRR (pass 0) or the others (pass 1).
1240 * The reason for as many as two passes is that when doing IPSTEALTH,
1241 * non-routing options should be processed only if the packet is for us.
1242 * Returns 1 if packet has been forwarded/freed,
1243 * 0 if the packet should be processed further.
1244 */
1245static int
1246ip_dooptions(struct mbuf *m, int pass)
1247{
1248	struct ip *ip = mtod(m, struct ip *);
1249	u_char *cp;
1250	struct in_ifaddr *ia;
1251	int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0;
1252	struct in_addr *sin, dst;
1253	n_time ntime;
1254	struct	sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET };
1255
1256	/* ignore or reject packets with IP options */
1257	if (ip_doopts == 0)
1258		return 0;
1259	else if (ip_doopts == 2) {
1260		type = ICMP_UNREACH;
1261		code = ICMP_UNREACH_FILTER_PROHIB;
1262		goto bad;
1263	}
1264
1265	dst = ip->ip_dst;
1266	cp = (u_char *)(ip + 1);
1267	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
1268	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1269		opt = cp[IPOPT_OPTVAL];
1270		if (opt == IPOPT_EOL)
1271			break;
1272		if (opt == IPOPT_NOP)
1273			optlen = 1;
1274		else {
1275			if (cnt < IPOPT_OLEN + sizeof(*cp)) {
1276				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1277				goto bad;
1278			}
1279			optlen = cp[IPOPT_OLEN];
1280			if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
1281				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1282				goto bad;
1283			}
1284		}
1285		switch (opt) {
1286
1287		default:
1288			break;
1289
1290		/*
1291		 * Source routing with record.
1292		 * Find interface with current destination address.
1293		 * If none on this machine then drop if strictly routed,
1294		 * or do nothing if loosely routed.
1295		 * Record interface address and bring up next address
1296		 * component.  If strictly routed make sure next
1297		 * address is on directly accessible net.
1298		 */
1299		case IPOPT_LSRR:
1300		case IPOPT_SSRR:
1301#ifdef IPSTEALTH
1302			if (ipstealth && pass > 0)
1303				break;
1304#endif
1305			if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1306				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1307				goto bad;
1308			}
1309			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1310				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1311				goto bad;
1312			}
1313			ipaddr.sin_addr = ip->ip_dst;
1314			ia = (struct in_ifaddr *)
1315				ifa_ifwithaddr((struct sockaddr *)&ipaddr);
1316			if (ia == NULL) {
1317				if (opt == IPOPT_SSRR) {
1318					type = ICMP_UNREACH;
1319					code = ICMP_UNREACH_SRCFAIL;
1320					goto bad;
1321				}
1322				if (!ip_dosourceroute)
1323					goto nosourcerouting;
1324				/*
1325				 * Loose routing, and not at next destination
1326				 * yet; nothing to do except forward.
1327				 */
1328				break;
1329			}
1330			off--;			/* 0 origin */
1331			if (off > optlen - (int)sizeof(struct in_addr)) {
1332				/*
1333				 * End of source route.  Should be for us.
1334				 */
1335				if (!ip_acceptsourceroute)
1336					goto nosourcerouting;
1337				save_rte(m, cp, ip->ip_src);
1338				break;
1339			}
1340#ifdef IPSTEALTH
1341			if (ipstealth)
1342				goto dropit;
1343#endif
1344			if (!ip_dosourceroute) {
1345				if (ipforwarding) {
1346					char buf[16]; /* aaa.bbb.ccc.ddd\0 */
1347					/*
1348					 * Acting as a router, so generate ICMP
1349					 */
1350nosourcerouting:
1351					strcpy(buf, inet_ntoa(ip->ip_dst));
1352					log(LOG_WARNING,
1353					    "attempted source route from %s to %s\n",
1354					    inet_ntoa(ip->ip_src), buf);
1355					type = ICMP_UNREACH;
1356					code = ICMP_UNREACH_SRCFAIL;
1357					goto bad;
1358				} else {
1359					/*
1360					 * Not acting as a router, so silently drop.
1361					 */
1362#ifdef IPSTEALTH
1363dropit:
1364#endif
1365					ipstat.ips_cantforward++;
1366					m_freem(m);
1367					return (1);
1368				}
1369			}
1370
1371			/*
1372			 * locate outgoing interface
1373			 */
1374			(void)memcpy(&ipaddr.sin_addr, cp + off,
1375			    sizeof(ipaddr.sin_addr));
1376
1377			if (opt == IPOPT_SSRR) {
1378#define	INA	struct in_ifaddr *
1379#define	SA	struct sockaddr *
1380			    if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == NULL)
1381				ia = (INA)ifa_ifwithnet((SA)&ipaddr);
1382			} else
1383				ia = ip_rtaddr(ipaddr.sin_addr);
1384			if (ia == NULL) {
1385				type = ICMP_UNREACH;
1386				code = ICMP_UNREACH_SRCFAIL;
1387				goto bad;
1388			}
1389			ip->ip_dst = ipaddr.sin_addr;
1390			(void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
1391			    sizeof(struct in_addr));
1392			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1393			/*
1394			 * Let ip_intr's mcast routing check handle mcast pkts
1395			 */
1396			forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr));
1397			break;
1398
1399		case IPOPT_RR:
1400#ifdef IPSTEALTH
1401			if (ipstealth && pass == 0)
1402				break;
1403#endif
1404			if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1405				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1406				goto bad;
1407			}
1408			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1409				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1410				goto bad;
1411			}
1412			/*
1413			 * If no space remains, ignore.
1414			 */
1415			off--;			/* 0 origin */
1416			if (off > optlen - (int)sizeof(struct in_addr))
1417				break;
1418			(void)memcpy(&ipaddr.sin_addr, &ip->ip_dst,
1419			    sizeof(ipaddr.sin_addr));
1420			/*
1421			 * locate outgoing interface; if we're the destination,
1422			 * use the incoming interface (should be same).
1423			 */
1424			if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == NULL &&
1425			    (ia = ip_rtaddr(ipaddr.sin_addr)) == NULL) {
1426				type = ICMP_UNREACH;
1427				code = ICMP_UNREACH_HOST;
1428				goto bad;
1429			}
1430			(void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
1431			    sizeof(struct in_addr));
1432			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1433			break;
1434
1435		case IPOPT_TS:
1436#ifdef IPSTEALTH
1437			if (ipstealth && pass == 0)
1438				break;
1439#endif
1440			code = cp - (u_char *)ip;
1441			if (optlen < 4 || optlen > 40) {
1442				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1443				goto bad;
1444			}
1445			if ((off = cp[IPOPT_OFFSET]) < 5) {
1446				code = &cp[IPOPT_OLEN] - (u_char *)ip;
1447				goto bad;
1448			}
1449			if (off > optlen - (int)sizeof(int32_t)) {
1450				cp[IPOPT_OFFSET + 1] += (1 << 4);
1451				if ((cp[IPOPT_OFFSET + 1] & 0xf0) == 0) {
1452					code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1453					goto bad;
1454				}
1455				break;
1456			}
1457			off--;				/* 0 origin */
1458			sin = (struct in_addr *)(cp + off);
1459			switch (cp[IPOPT_OFFSET + 1] & 0x0f) {
1460
1461			case IPOPT_TS_TSONLY:
1462				break;
1463
1464			case IPOPT_TS_TSANDADDR:
1465				if (off + sizeof(n_time) +
1466				    sizeof(struct in_addr) > optlen) {
1467					code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1468					goto bad;
1469				}
1470				ipaddr.sin_addr = dst;
1471				ia = (INA)ifaof_ifpforaddr((SA)&ipaddr,
1472							    m->m_pkthdr.rcvif);
1473				if (ia == NULL)
1474					continue;
1475				(void)memcpy(sin, &IA_SIN(ia)->sin_addr,
1476				    sizeof(struct in_addr));
1477				cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1478				off += sizeof(struct in_addr);
1479				break;
1480
1481			case IPOPT_TS_PRESPEC:
1482				if (off + sizeof(n_time) +
1483				    sizeof(struct in_addr) > optlen) {
1484					code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1485					goto bad;
1486				}
1487				(void)memcpy(&ipaddr.sin_addr, sin,
1488				    sizeof(struct in_addr));
1489				if (ifa_ifwithaddr((SA)&ipaddr) == NULL)
1490					continue;
1491				cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1492				off += sizeof(struct in_addr);
1493				break;
1494
1495			default:
1496				code = &cp[IPOPT_OFFSET + 1] - (u_char *)ip;
1497				goto bad;
1498			}
1499			ntime = iptime();
1500			(void)memcpy(cp + off, &ntime, sizeof(n_time));
1501			cp[IPOPT_OFFSET] += sizeof(n_time);
1502		}
1503	}
1504	if (forward && ipforwarding) {
1505		ip_forward(m, 1);
1506		return (1);
1507	}
1508	return (0);
1509bad:
1510	icmp_error(m, type, code, 0, 0);
1511	ipstat.ips_badoptions++;
1512	return (1);
1513}
1514
1515/*
1516 * Given address of next destination (final or next hop),
1517 * return internet address info of interface to be used to get there.
1518 */
1519struct in_ifaddr *
1520ip_rtaddr(dst)
1521	struct in_addr dst;
1522{
1523	struct route sro;
1524	struct sockaddr_in *sin;
1525	struct in_ifaddr *ifa;
1526
1527	bzero(&sro, sizeof(sro));
1528	sin = (struct sockaddr_in *)&sro.ro_dst;
1529	sin->sin_family = AF_INET;
1530	sin->sin_len = sizeof(*sin);
1531	sin->sin_addr = dst;
1532	rtalloc_ign(&sro, RTF_CLONING);
1533
1534	if (sro.ro_rt == NULL)
1535		return (NULL);
1536
1537	ifa = ifatoia(sro.ro_rt->rt_ifa);
1538	RTFREE(sro.ro_rt);
1539	return (ifa);
1540}
1541
1542/*
1543 * Save incoming source route for use in replies,
1544 * to be picked up later by ip_srcroute if the receiver is interested.
1545 */
1546static void
1547save_rte(m, option, dst)
1548	struct mbuf *m;
1549	u_char *option;
1550	struct in_addr dst;
1551{
1552	unsigned olen;
1553	struct ipopt_tag *opts;
1554
1555	opts = (struct ipopt_tag *)m_tag_get(PACKET_TAG_IPOPTIONS,
1556					sizeof(struct ipopt_tag), M_NOWAIT);
1557	if (opts == NULL)
1558		return;
1559
1560	olen = option[IPOPT_OLEN];
1561#ifdef DIAGNOSTIC
1562	if (ipprintfs)
1563		printf("save_rte: olen %d\n", olen);
1564#endif
1565	if (olen > sizeof(opts->ip_srcrt) - (1 + sizeof(dst)))
1566		return;
1567	bcopy(option, opts->ip_srcrt.srcopt, olen);
1568	opts->ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1569	opts->ip_srcrt.dst = dst;
1570	m_tag_prepend(m, (struct m_tag *)opts);
1571}
1572
1573/*
1574 * Retrieve incoming source route for use in replies,
1575 * in the same form used by setsockopt.
1576 * The first hop is placed before the options, will be removed later.
1577 */
1578struct mbuf *
1579ip_srcroute(m0)
1580	struct mbuf *m0;
1581{
1582	register struct in_addr *p, *q;
1583	register struct mbuf *m;
1584	struct ipopt_tag *opts;
1585
1586	opts = (struct ipopt_tag *)m_tag_find(m0, PACKET_TAG_IPOPTIONS, NULL);
1587	if (opts == NULL)
1588		return (NULL);
1589
1590	if (opts->ip_nhops == 0)
1591		return (NULL);
1592	m = m_get(M_DONTWAIT, MT_HEADER);
1593	if (m == NULL)
1594		return (NULL);
1595
1596#define OPTSIZ	(sizeof(opts->ip_srcrt.nop) + sizeof(opts->ip_srcrt.srcopt))
1597
1598	/* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
1599	m->m_len = opts->ip_nhops * sizeof(struct in_addr) +
1600	    sizeof(struct in_addr) + OPTSIZ;
1601#ifdef DIAGNOSTIC
1602	if (ipprintfs)
1603		printf("ip_srcroute: nhops %d mlen %d", opts->ip_nhops, m->m_len);
1604#endif
1605
1606	/*
1607	 * First save first hop for return route
1608	 */
1609	p = &(opts->ip_srcrt.route[opts->ip_nhops - 1]);
1610	*(mtod(m, struct in_addr *)) = *p--;
1611#ifdef DIAGNOSTIC
1612	if (ipprintfs)
1613		printf(" hops %lx", (u_long)ntohl(mtod(m, struct in_addr *)->s_addr));
1614#endif
1615
1616	/*
1617	 * Copy option fields and padding (nop) to mbuf.
1618	 */
1619	opts->ip_srcrt.nop = IPOPT_NOP;
1620	opts->ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
1621	(void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr),
1622	    &(opts->ip_srcrt.nop), OPTSIZ);
1623	q = (struct in_addr *)(mtod(m, caddr_t) +
1624	    sizeof(struct in_addr) + OPTSIZ);
1625#undef OPTSIZ
1626	/*
1627	 * Record return path as an IP source route,
1628	 * reversing the path (pointers are now aligned).
1629	 */
1630	while (p >= opts->ip_srcrt.route) {
1631#ifdef DIAGNOSTIC
1632		if (ipprintfs)
1633			printf(" %lx", (u_long)ntohl(q->s_addr));
1634#endif
1635		*q++ = *p--;
1636	}
1637	/*
1638	 * Last hop goes to final destination.
1639	 */
1640	*q = opts->ip_srcrt.dst;
1641#ifdef DIAGNOSTIC
1642	if (ipprintfs)
1643		printf(" %lx\n", (u_long)ntohl(q->s_addr));
1644#endif
1645	m_tag_delete(m0, (struct m_tag *)opts);
1646	return (m);
1647}
1648
1649/*
1650 * Strip out IP options, at higher
1651 * level protocol in the kernel.
1652 * Second argument is buffer to which options
1653 * will be moved, and return value is their length.
1654 * XXX should be deleted; last arg currently ignored.
1655 */
1656void
1657ip_stripoptions(m, mopt)
1658	register struct mbuf *m;
1659	struct mbuf *mopt;
1660{
1661	register int i;
1662	struct ip *ip = mtod(m, struct ip *);
1663	register caddr_t opts;
1664	int olen;
1665
1666	olen = (ip->ip_hl << 2) - sizeof (struct ip);
1667	opts = (caddr_t)(ip + 1);
1668	i = m->m_len - (sizeof (struct ip) + olen);
1669	bcopy(opts + olen, opts, (unsigned)i);
1670	m->m_len -= olen;
1671	if (m->m_flags & M_PKTHDR)
1672		m->m_pkthdr.len -= olen;
1673	ip->ip_v = IPVERSION;
1674	ip->ip_hl = sizeof(struct ip) >> 2;
1675}
1676
1677u_char inetctlerrmap[PRC_NCMDS] = {
1678	0,		0,		0,		0,
1679	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1680	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1681	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1682	0,		0,		EHOSTUNREACH,	0,
1683	ENOPROTOOPT,	ECONNREFUSED
1684};
1685
1686/*
1687 * Forward a packet.  If some error occurs return the sender
1688 * an icmp packet.  Note we can't always generate a meaningful
1689 * icmp message because icmp doesn't have a large enough repertoire
1690 * of codes and types.
1691 *
1692 * If not forwarding, just drop the packet.  This could be confusing
1693 * if ipforwarding was zero but some routing protocol was advancing
1694 * us as a gateway to somewhere.  However, we must let the routing
1695 * protocol deal with that.
1696 *
1697 * The srcrt parameter indicates whether the packet is being forwarded
1698 * via a source route.
1699 */
1700void
1701ip_forward(struct mbuf *m, int srcrt)
1702{
1703	struct ip *ip = mtod(m, struct ip *);
1704	struct in_ifaddr *ia = NULL;
1705	int error, type = 0, code = 0;
1706	struct mbuf *mcopy;
1707	struct in_addr dest;
1708	struct ifnet *destifp, dummyifp;
1709
1710#ifdef DIAGNOSTIC
1711	if (ipprintfs)
1712		printf("forward: src %lx dst %lx ttl %x\n",
1713		    (u_long)ip->ip_src.s_addr, (u_long)ip->ip_dst.s_addr,
1714		    ip->ip_ttl);
1715#endif
1716
1717
1718	if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
1719		ipstat.ips_cantforward++;
1720		m_freem(m);
1721		return;
1722	}
1723#ifdef IPSTEALTH
1724	if (!ipstealth) {
1725#endif
1726		if (ip->ip_ttl <= IPTTLDEC) {
1727			icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS,
1728			    0, 0);
1729			return;
1730		}
1731#ifdef IPSTEALTH
1732	}
1733#endif
1734
1735	if (!srcrt && (ia = ip_rtaddr(ip->ip_dst)) == NULL) {
1736		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
1737		return;
1738	}
1739
1740	/*
1741	 * Save the IP header and at most 8 bytes of the payload,
1742	 * in case we need to generate an ICMP message to the src.
1743	 *
1744	 * XXX this can be optimized a lot by saving the data in a local
1745	 * buffer on the stack (72 bytes at most), and only allocating the
1746	 * mbuf if really necessary. The vast majority of the packets
1747	 * are forwarded without having to send an ICMP back (either
1748	 * because unnecessary, or because rate limited), so we are
1749	 * really we are wasting a lot of work here.
1750	 *
1751	 * We don't use m_copy() because it might return a reference
1752	 * to a shared cluster. Both this function and ip_output()
1753	 * assume exclusive access to the IP header in `m', so any
1754	 * data in a cluster may change before we reach icmp_error().
1755	 */
1756	MGET(mcopy, M_DONTWAIT, m->m_type);
1757	if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_DONTWAIT)) {
1758		/*
1759		 * It's probably ok if the pkthdr dup fails (because
1760		 * the deep copy of the tag chain failed), but for now
1761		 * be conservative and just discard the copy since
1762		 * code below may some day want the tags.
1763		 */
1764		m_free(mcopy);
1765		mcopy = NULL;
1766	}
1767	if (mcopy != NULL) {
1768		mcopy->m_len = imin((ip->ip_hl << 2) + 8,
1769		    (int)ip->ip_len);
1770		mcopy->m_pkthdr.len = mcopy->m_len;
1771		m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1772	}
1773
1774#ifdef IPSTEALTH
1775	if (!ipstealth) {
1776#endif
1777		ip->ip_ttl -= IPTTLDEC;
1778#ifdef IPSTEALTH
1779	}
1780#endif
1781
1782	/*
1783	 * If forwarding packet using same interface that it came in on,
1784	 * perhaps should send a redirect to sender to shortcut a hop.
1785	 * Only send redirect if source is sending directly to us,
1786	 * and if packet was not source routed (or has any options).
1787	 * Also, don't send redirect if forwarding using a default route
1788	 * or a route modified by a redirect.
1789	 */
1790	dest.s_addr = 0;
1791	if (!srcrt && ipsendredirects && ia->ia_ifp == m->m_pkthdr.rcvif) {
1792		struct sockaddr_in *sin;
1793		struct route ro;
1794		struct rtentry *rt;
1795
1796		bzero(&ro, sizeof(ro));
1797		sin = (struct sockaddr_in *)&ro.ro_dst;
1798		sin->sin_family = AF_INET;
1799		sin->sin_len = sizeof(*sin);
1800		sin->sin_addr = ip->ip_dst;
1801		rtalloc_ign(&ro, RTF_CLONING);
1802
1803		rt = ro.ro_rt;
1804
1805		if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1806		    satosin(rt_key(rt))->sin_addr.s_addr != 0) {
1807#define	RTA(rt)	((struct in_ifaddr *)(rt->rt_ifa))
1808			u_long src = ntohl(ip->ip_src.s_addr);
1809
1810			if (RTA(rt) &&
1811			    (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1812				if (rt->rt_flags & RTF_GATEWAY)
1813					dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr;
1814				else
1815					dest.s_addr = ip->ip_dst.s_addr;
1816				/* Router requirements says to only send host redirects */
1817				type = ICMP_REDIRECT;
1818				code = ICMP_REDIRECT_HOST;
1819#ifdef DIAGNOSTIC
1820				if (ipprintfs)
1821					printf("redirect (%d) to %lx\n", code, (u_long)dest.s_addr);
1822#endif
1823			}
1824		}
1825		if (rt)
1826			RTFREE(rt);
1827	}
1828
1829	error = ip_output(m, NULL, NULL, IP_FORWARDING, NULL, NULL);
1830	if (error)
1831		ipstat.ips_cantforward++;
1832	else {
1833		ipstat.ips_forward++;
1834		if (type)
1835			ipstat.ips_redirectsent++;
1836		else {
1837			if (mcopy)
1838				m_freem(mcopy);
1839			return;
1840		}
1841	}
1842	if (mcopy == NULL)
1843		return;
1844	destifp = NULL;
1845
1846	switch (error) {
1847
1848	case 0:				/* forwarded, but need redirect */
1849		/* type, code set above */
1850		break;
1851
1852	case ENETUNREACH:		/* shouldn't happen, checked above */
1853	case EHOSTUNREACH:
1854	case ENETDOWN:
1855	case EHOSTDOWN:
1856	default:
1857		type = ICMP_UNREACH;
1858		code = ICMP_UNREACH_HOST;
1859		break;
1860
1861	case EMSGSIZE:
1862		type = ICMP_UNREACH;
1863		code = ICMP_UNREACH_NEEDFRAG;
1864#if defined(IPSEC) || defined(FAST_IPSEC)
1865		/*
1866		 * If the packet is routed over IPsec tunnel, tell the
1867		 * originator the tunnel MTU.
1868		 *	tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
1869		 * XXX quickhack!!!
1870		 */
1871		{
1872			struct secpolicy *sp = NULL;
1873			int ipsecerror;
1874			int ipsechdr;
1875			struct route *ro;
1876
1877#ifdef IPSEC
1878			sp = ipsec4_getpolicybyaddr(mcopy,
1879						    IPSEC_DIR_OUTBOUND,
1880						    IP_FORWARDING,
1881						    &ipsecerror);
1882#else /* FAST_IPSEC */
1883			sp = ipsec_getpolicybyaddr(mcopy,
1884						   IPSEC_DIR_OUTBOUND,
1885						   IP_FORWARDING,
1886						   &ipsecerror);
1887#endif
1888			if (sp != NULL) {
1889				/* count IPsec header size */
1890				ipsechdr = ipsec4_hdrsiz(mcopy,
1891							 IPSEC_DIR_OUTBOUND,
1892							 NULL);
1893
1894				/*
1895				 * find the correct route for outer IPv4
1896				 * header, compute tunnel MTU.
1897				 *
1898				 * XXX BUG ALERT
1899				 * The "dummyifp" code relies upon the fact
1900				 * that icmp_error() touches only ifp->if_mtu.
1901				 */
1902				/*XXX*/
1903				destifp = NULL;
1904				if (sp->req != NULL
1905				 && sp->req->sav != NULL
1906				 && sp->req->sav->sah != NULL) {
1907					ro = &sp->req->sav->sah->sa_route;
1908					if (ro->ro_rt && ro->ro_rt->rt_ifp) {
1909						dummyifp.if_mtu =
1910						    ro->ro_rt->rt_rmx.rmx_mtu ?
1911						    ro->ro_rt->rt_rmx.rmx_mtu :
1912						    ro->ro_rt->rt_ifp->if_mtu;
1913						dummyifp.if_mtu -= ipsechdr;
1914						destifp = &dummyifp;
1915					}
1916				}
1917
1918#ifdef IPSEC
1919				key_freesp(sp);
1920#else /* FAST_IPSEC */
1921				KEY_FREESP(&sp);
1922#endif
1923				ipstat.ips_cantfrag++;
1924				break;
1925			} else
1926#endif /*IPSEC || FAST_IPSEC*/
1927		/*
1928		 * When doing source routing 'ia' can be NULL.  Fall back
1929		 * to the minimum guaranteed routeable packet size and use
1930		 * the same hack as IPSEC to setup a dummyifp for icmp.
1931		 */
1932		if (ia == NULL) {
1933			dummyifp.if_mtu = IP_MSS;
1934			destifp = &dummyifp;
1935		} else
1936			destifp = ia->ia_ifp;
1937#if defined(IPSEC) || defined(FAST_IPSEC)
1938		}
1939#endif /*IPSEC || FAST_IPSEC*/
1940		ipstat.ips_cantfrag++;
1941		break;
1942
1943	case ENOBUFS:
1944		/*
1945		 * A router should not generate ICMP_SOURCEQUENCH as
1946		 * required in RFC1812 Requirements for IP Version 4 Routers.
1947		 * Source quench could be a big problem under DoS attacks,
1948		 * or if the underlying interface is rate-limited.
1949		 * Those who need source quench packets may re-enable them
1950		 * via the net.inet.ip.sendsourcequench sysctl.
1951		 */
1952		if (ip_sendsourcequench == 0) {
1953			m_freem(mcopy);
1954			return;
1955		} else {
1956			type = ICMP_SOURCEQUENCH;
1957			code = 0;
1958		}
1959		break;
1960
1961	case EACCES:			/* ipfw denied packet */
1962		m_freem(mcopy);
1963		return;
1964	}
1965	icmp_error(mcopy, type, code, dest.s_addr, destifp);
1966}
1967
1968void
1969ip_savecontrol(inp, mp, ip, m)
1970	register struct inpcb *inp;
1971	register struct mbuf **mp;
1972	register struct ip *ip;
1973	register struct mbuf *m;
1974{
1975	if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) {
1976		struct bintime bt;
1977
1978		bintime(&bt);
1979		if (inp->inp_socket->so_options & SO_BINTIME) {
1980			*mp = sbcreatecontrol((caddr_t) &bt, sizeof(bt),
1981			SCM_BINTIME, SOL_SOCKET);
1982			if (*mp)
1983				mp = &(*mp)->m_next;
1984		}
1985		if (inp->inp_socket->so_options & SO_TIMESTAMP) {
1986			struct timeval tv;
1987
1988			bintime2timeval(&bt, &tv);
1989			*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1990				SCM_TIMESTAMP, SOL_SOCKET);
1991			if (*mp)
1992				mp = &(*mp)->m_next;
1993		}
1994	}
1995	if (inp->inp_flags & INP_RECVDSTADDR) {
1996		*mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
1997		    sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1998		if (*mp)
1999			mp = &(*mp)->m_next;
2000	}
2001	if (inp->inp_flags & INP_RECVTTL) {
2002		*mp = sbcreatecontrol((caddr_t) &ip->ip_ttl,
2003		    sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
2004		if (*mp)
2005			mp = &(*mp)->m_next;
2006	}
2007#ifdef notyet
2008	/* XXX
2009	 * Moving these out of udp_input() made them even more broken
2010	 * than they already were.
2011	 */
2012	/* options were tossed already */
2013	if (inp->inp_flags & INP_RECVOPTS) {
2014		*mp = sbcreatecontrol((caddr_t) opts_deleted_above,
2015		    sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
2016		if (*mp)
2017			mp = &(*mp)->m_next;
2018	}
2019	/* ip_srcroute doesn't do what we want here, need to fix */
2020	if (inp->inp_flags & INP_RECVRETOPTS) {
2021		*mp = sbcreatecontrol((caddr_t) ip_srcroute(m),
2022		    sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
2023		if (*mp)
2024			mp = &(*mp)->m_next;
2025	}
2026#endif
2027	if (inp->inp_flags & INP_RECVIF) {
2028		struct ifnet *ifp;
2029		struct sdlbuf {
2030			struct sockaddr_dl sdl;
2031			u_char	pad[32];
2032		} sdlbuf;
2033		struct sockaddr_dl *sdp;
2034		struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
2035
2036		if (((ifp = m->m_pkthdr.rcvif))
2037		&& ( ifp->if_index && (ifp->if_index <= if_index))) {
2038			sdp = (struct sockaddr_dl *)
2039			    (ifaddr_byindex(ifp->if_index)->ifa_addr);
2040			/*
2041			 * Change our mind and don't try copy.
2042			 */
2043			if ((sdp->sdl_family != AF_LINK)
2044			|| (sdp->sdl_len > sizeof(sdlbuf))) {
2045				goto makedummy;
2046			}
2047			bcopy(sdp, sdl2, sdp->sdl_len);
2048		} else {
2049makedummy:
2050			sdl2->sdl_len
2051				= offsetof(struct sockaddr_dl, sdl_data[0]);
2052			sdl2->sdl_family = AF_LINK;
2053			sdl2->sdl_index = 0;
2054			sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
2055		}
2056		*mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len,
2057			IP_RECVIF, IPPROTO_IP);
2058		if (*mp)
2059			mp = &(*mp)->m_next;
2060	}
2061}
2062
2063/*
2064 * XXX these routines are called from the upper part of the kernel.
2065 * They need to be locked when we remove Giant.
2066 *
2067 * They could also be moved to ip_mroute.c, since all the RSVP
2068 *  handling is done there already.
2069 */
2070static int ip_rsvp_on;
2071struct socket *ip_rsvpd;
2072int
2073ip_rsvp_init(struct socket *so)
2074{
2075	if (so->so_type != SOCK_RAW ||
2076	    so->so_proto->pr_protocol != IPPROTO_RSVP)
2077		return EOPNOTSUPP;
2078
2079	if (ip_rsvpd != NULL)
2080		return EADDRINUSE;
2081
2082	ip_rsvpd = so;
2083	/*
2084	 * This may seem silly, but we need to be sure we don't over-increment
2085	 * the RSVP counter, in case something slips up.
2086	 */
2087	if (!ip_rsvp_on) {
2088		ip_rsvp_on = 1;
2089		rsvp_on++;
2090	}
2091
2092	return 0;
2093}
2094
2095int
2096ip_rsvp_done(void)
2097{
2098	ip_rsvpd = NULL;
2099	/*
2100	 * This may seem silly, but we need to be sure we don't over-decrement
2101	 * the RSVP counter, in case something slips up.
2102	 */
2103	if (ip_rsvp_on) {
2104		ip_rsvp_on = 0;
2105		rsvp_on--;
2106	}
2107	return 0;
2108}
2109
2110void
2111rsvp_input(struct mbuf *m, int off)	/* XXX must fixup manually */
2112{
2113	if (rsvp_input_p) { /* call the real one if loaded */
2114		rsvp_input_p(m, off);
2115		return;
2116	}
2117
2118	/* Can still get packets with rsvp_on = 0 if there is a local member
2119	 * of the group to which the RSVP packet is addressed.  But in this
2120	 * case we want to throw the packet away.
2121	 */
2122
2123	if (!rsvp_on) {
2124		m_freem(m);
2125		return;
2126	}
2127
2128	if (ip_rsvpd != NULL) {
2129		rip_input(m, off);
2130		return;
2131	}
2132	/* Drop the packet */
2133	m_freem(m);
2134}
2135