xref: /freebsd/sys/netinet/ip_input.c (revision 33644623554bb0fc57ed3c7d874193a498679b22)
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  */
31 
32 #include <sys/cdefs.h>
33 __FBSDID("$FreeBSD$");
34 
35 #include "opt_bootp.h"
36 #include "opt_ipfw.h"
37 #include "opt_ipstealth.h"
38 #include "opt_ipsec.h"
39 #include "opt_mac.h"
40 #include "opt_carp.h"
41 
42 #include <sys/param.h>
43 #include <sys/systm.h>
44 #include <sys/callout.h>
45 #include <sys/mbuf.h>
46 #include <sys/malloc.h>
47 #include <sys/domain.h>
48 #include <sys/protosw.h>
49 #include <sys/socket.h>
50 #include <sys/time.h>
51 #include <sys/kernel.h>
52 #include <sys/syslog.h>
53 #include <sys/sysctl.h>
54 #include <sys/vimage.h>
55 
56 #include <net/pfil.h>
57 #include <net/if.h>
58 #include <net/if_types.h>
59 #include <net/if_var.h>
60 #include <net/if_dl.h>
61 #include <net/route.h>
62 #include <net/netisr.h>
63 
64 #include <netinet/in.h>
65 #include <netinet/in_systm.h>
66 #include <netinet/in_var.h>
67 #include <netinet/ip.h>
68 #include <netinet/in_pcb.h>
69 #include <netinet/ip_var.h>
70 #include <netinet/ip_icmp.h>
71 #include <netinet/ip_options.h>
72 #include <machine/in_cksum.h>
73 #ifdef DEV_CARP
74 #include <netinet/ip_carp.h>
75 #endif
76 #ifdef IPSEC
77 #include <netinet/ip_ipsec.h>
78 #endif /* IPSEC */
79 
80 #include <sys/socketvar.h>
81 
82 /* XXX: Temporary until ipfw_ether and ipfw_bridge are converted. */
83 #include <netinet/ip_fw.h>
84 #include <netinet/ip_dummynet.h>
85 
86 #include <security/mac/mac_framework.h>
87 
88 #ifdef CTASSERT
89 CTASSERT(sizeof(struct ip) == 20);
90 #endif
91 
92 #ifdef VIMAGE_GLOBALS
93 static int	ipsendredirects;
94 static int	ip_checkinterface;
95 static int	ip_keepfaith;
96 static int	ip_sendsourcequench;
97 int	ip_defttl;
98 int	ip_do_randomid;
99 int	ipforwarding;
100 struct	in_ifaddrhead in_ifaddrhead; 		/* first inet address */
101 struct	in_ifaddrhashhead *in_ifaddrhashtbl;	/* inet addr hash table  */
102 u_long 	in_ifaddrhmask;				/* mask for hash table */
103 struct ipstat ipstat;
104 static int ip_rsvp_on;
105 struct socket *ip_rsvpd;
106 int	rsvp_on;
107 static struct ipqhead ipq[IPREASS_NHASH];
108 static int	maxnipq;	/* Administrative limit on # reass queues. */
109 static int	maxfragsperpacket;
110 int	ipstealth;
111 static int	nipq;	/* Total # of reass queues */
112 #endif
113 
114 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, IPCTL_FORWARDING,
115     forwarding, CTLFLAG_RW, ipforwarding, 0,
116     "Enable IP forwarding between interfaces");
117 
118 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, IPCTL_SENDREDIRECTS,
119     redirect, CTLFLAG_RW, ipsendredirects, 0,
120     "Enable sending IP redirects");
121 
122 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, IPCTL_DEFTTL,
123     ttl, CTLFLAG_RW, ip_defttl, 0, "Maximum TTL on IP packets");
124 
125 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, IPCTL_KEEPFAITH,
126     keepfaith, CTLFLAG_RW, ip_keepfaith,	0,
127     "Enable packet capture for FAITH IPv4->IPv6 translater daemon");
128 
129 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO,
130     sendsourcequench, CTLFLAG_RW, ip_sendsourcequench, 0,
131     "Enable the transmission of source quench packets");
132 
133 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, random_id,
134     CTLFLAG_RW, ip_do_randomid, 0, "Assign random ip_id values");
135 
136 /*
137  * XXX - Setting ip_checkinterface mostly implements the receive side of
138  * the Strong ES model described in RFC 1122, but since the routing table
139  * and transmit implementation do not implement the Strong ES model,
140  * setting this to 1 results in an odd hybrid.
141  *
142  * XXX - ip_checkinterface currently must be disabled if you use ipnat
143  * to translate the destination address to another local interface.
144  *
145  * XXX - ip_checkinterface must be disabled if you add IP aliases
146  * to the loopback interface instead of the interface where the
147  * packets for those addresses are received.
148  */
149 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO,
150     check_interface, CTLFLAG_RW, ip_checkinterface, 0,
151     "Verify packet arrives on correct interface");
152 
153 struct pfil_head inet_pfil_hook;	/* Packet filter hooks */
154 
155 static struct	ifqueue ipintrq;
156 static int	ipqmaxlen = IFQ_MAXLEN;
157 
158 extern	struct domain inetdomain;
159 extern	struct protosw inetsw[];
160 u_char	ip_protox[IPPROTO_MAX];
161 
162 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW,
163     &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue");
164 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD,
165     &ipintrq.ifq_drops, 0,
166     "Number of packets dropped from the IP input queue");
167 
168 SYSCTL_V_STRUCT(V_NET, vnet_inet, _net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW,
169     ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)");
170 
171 static uma_zone_t ipq_zone;
172 static struct mtx ipqlock;
173 
174 #define	IPQ_LOCK()	mtx_lock(&ipqlock)
175 #define	IPQ_UNLOCK()	mtx_unlock(&ipqlock)
176 #define	IPQ_LOCK_INIT()	mtx_init(&ipqlock, "ipqlock", NULL, MTX_DEF)
177 #define	IPQ_LOCK_ASSERT()	mtx_assert(&ipqlock, MA_OWNED)
178 
179 static void	maxnipq_update(void);
180 static void	ipq_zone_change(void *);
181 
182 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, fragpackets,
183     CTLFLAG_RD, nipq, 0,
184     "Current number of IPv4 fragment reassembly queue entries");
185 
186 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, maxfragsperpacket,
187     CTLFLAG_RW, maxfragsperpacket, 0,
188     "Maximum number of IPv4 fragments allowed per packet");
189 
190 struct callout	ipport_tick_callout;
191 
192 #ifdef IPCTL_DEFMTU
193 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
194     &ip_mtu, 0, "Default MTU");
195 #endif
196 
197 #ifdef IPSTEALTH
198 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW,
199     ipstealth, 0, "IP stealth mode, no TTL decrementation on forwarding");
200 #endif
201 
202 /*
203  * ipfw_ether and ipfw_bridge hooks.
204  * XXX: Temporary until those are converted to pfil_hooks as well.
205  */
206 ip_fw_chk_t *ip_fw_chk_ptr = NULL;
207 ip_dn_io_t *ip_dn_io_ptr = NULL;
208 int fw_one_pass = 1;
209 
210 static void	ip_freef(struct ipqhead *, struct ipq *);
211 
212 /*
213  * IP initialization: fill in IP protocol switch table.
214  * All protocols not implemented in kernel go to raw IP protocol handler.
215  */
216 void
217 ip_init(void)
218 {
219 	INIT_VNET_INET(curvnet);
220 	struct protosw *pr;
221 	int i;
222 
223 	V_ipsendredirects = 1; /* XXX */
224 	V_ip_checkinterface = 0;
225 	V_ip_keepfaith = 0;
226 	V_ip_sendsourcequench = 0;
227 	V_rsvp_on = 0;
228 	V_ip_defttl = IPDEFTTL;
229 	V_ip_do_randomid = 0;
230 	V_ipforwarding = 0;
231 	V_ipstealth = 0;
232 	V_nipq = 0;	/* Total # of reass queues */
233 
234 	V_ipport_lowfirstauto = IPPORT_RESERVED - 1;	/* 1023 */
235 	V_ipport_lowlastauto = IPPORT_RESERVEDSTART;	/* 600 */
236 	V_ipport_firstauto = IPPORT_EPHEMERALFIRST;	/* 10000 */
237 	V_ipport_lastauto = IPPORT_EPHEMERALLAST;	/* 65535 */
238 	V_ipport_hifirstauto = IPPORT_HIFIRSTAUTO;	/* 49152 */
239 	V_ipport_hilastauto = IPPORT_HILASTAUTO;	/* 65535 */
240 	V_ipport_reservedhigh = IPPORT_RESERVED - 1;	/* 1023 */
241 	V_ipport_reservedlow = 0;
242 	V_ipport_randomized = 1;	/* user controlled via sysctl */
243 	V_ipport_randomcps = 10;	/* user controlled via sysctl */
244 	V_ipport_randomtime = 45;	/* user controlled via sysctl */
245 	V_ipport_stoprandom = 0;	/* toggled by ipport_tick */
246 
247 #ifdef NOTYET
248 	/* XXX global static but not instantiated in this file */
249 	V_ipfastforward_active = 0;
250 	V_subnetsarelocal = 0;
251 	V_sameprefixcarponly = 0;
252 #endif
253 
254 	TAILQ_INIT(&V_in_ifaddrhead);
255 	V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask);
256 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
257 	if (pr == NULL)
258 		panic("ip_init: PF_INET not found");
259 
260 	/* Initialize the entire ip_protox[] array to IPPROTO_RAW. */
261 	for (i = 0; i < IPPROTO_MAX; i++)
262 		ip_protox[i] = pr - inetsw;
263 	/*
264 	 * Cycle through IP protocols and put them into the appropriate place
265 	 * in ip_protox[].
266 	 */
267 	for (pr = inetdomain.dom_protosw;
268 	    pr < inetdomain.dom_protoswNPROTOSW; pr++)
269 		if (pr->pr_domain->dom_family == PF_INET &&
270 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
271 			/* Be careful to only index valid IP protocols. */
272 			if (pr->pr_protocol < IPPROTO_MAX)
273 				ip_protox[pr->pr_protocol] = pr - inetsw;
274 		}
275 
276 	/* Initialize packet filter hooks. */
277 	inet_pfil_hook.ph_type = PFIL_TYPE_AF;
278 	inet_pfil_hook.ph_af = AF_INET;
279 	if ((i = pfil_head_register(&inet_pfil_hook)) != 0)
280 		printf("%s: WARNING: unable to register pfil hook, "
281 			"error %d\n", __func__, i);
282 
283 	/* Initialize IP reassembly queue. */
284 	IPQ_LOCK_INIT();
285 	for (i = 0; i < IPREASS_NHASH; i++)
286 	    TAILQ_INIT(&V_ipq[i]);
287 	V_maxnipq = nmbclusters / 32;
288 	V_maxfragsperpacket = 16;
289 	V_ipq_zone = uma_zcreate("ipq", sizeof(struct ipq), NULL, NULL, NULL,
290 	    NULL, UMA_ALIGN_PTR, 0);
291 	maxnipq_update();
292 
293 	/* Start ipport_tick. */
294 	callout_init(&ipport_tick_callout, CALLOUT_MPSAFE);
295 	ipport_tick(NULL);
296 	EVENTHANDLER_REGISTER(shutdown_pre_sync, ip_fini, NULL,
297 		SHUTDOWN_PRI_DEFAULT);
298 	EVENTHANDLER_REGISTER(nmbclusters_change, ipq_zone_change,
299 		NULL, EVENTHANDLER_PRI_ANY);
300 
301 	/* Initialize various other remaining things. */
302 	V_ip_id = time_second & 0xffff;
303 	ipintrq.ifq_maxlen = ipqmaxlen;
304 	mtx_init(&ipintrq.ifq_mtx, "ip_inq", NULL, MTX_DEF);
305 	netisr_register(NETISR_IP, ip_input, &ipintrq, 0);
306 }
307 
308 void
309 ip_fini(void *xtp)
310 {
311 
312 	callout_stop(&ipport_tick_callout);
313 }
314 
315 /*
316  * Ip input routine.  Checksum and byte swap header.  If fragmented
317  * try to reassemble.  Process options.  Pass to next level.
318  */
319 void
320 ip_input(struct mbuf *m)
321 {
322 	INIT_VNET_INET(curvnet);
323 	struct ip *ip = NULL;
324 	struct in_ifaddr *ia = NULL;
325 	struct ifaddr *ifa;
326 	int    checkif, hlen = 0;
327 	u_short sum;
328 	int dchg = 0;				/* dest changed after fw */
329 	struct in_addr odst;			/* original dst address */
330 
331 	M_ASSERTPKTHDR(m);
332 
333 	if (m->m_flags & M_FASTFWD_OURS) {
334 		/*
335 		 * Firewall or NAT changed destination to local.
336 		 * We expect ip_len and ip_off to be in host byte order.
337 		 */
338 		m->m_flags &= ~M_FASTFWD_OURS;
339 		/* Set up some basics that will be used later. */
340 		ip = mtod(m, struct ip *);
341 		hlen = ip->ip_hl << 2;
342 		goto ours;
343 	}
344 
345 	V_ipstat.ips_total++;
346 
347 	if (m->m_pkthdr.len < sizeof(struct ip))
348 		goto tooshort;
349 
350 	if (m->m_len < sizeof (struct ip) &&
351 	    (m = m_pullup(m, sizeof (struct ip))) == NULL) {
352 		V_ipstat.ips_toosmall++;
353 		return;
354 	}
355 	ip = mtod(m, struct ip *);
356 
357 	if (ip->ip_v != IPVERSION) {
358 		V_ipstat.ips_badvers++;
359 		goto bad;
360 	}
361 
362 	hlen = ip->ip_hl << 2;
363 	if (hlen < sizeof(struct ip)) {	/* minimum header length */
364 		V_ipstat.ips_badhlen++;
365 		goto bad;
366 	}
367 	if (hlen > m->m_len) {
368 		if ((m = m_pullup(m, hlen)) == NULL) {
369 			V_ipstat.ips_badhlen++;
370 			return;
371 		}
372 		ip = mtod(m, struct ip *);
373 	}
374 
375 	/* 127/8 must not appear on wire - RFC1122 */
376 	if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
377 	    (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
378 		if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
379 			V_ipstat.ips_badaddr++;
380 			goto bad;
381 		}
382 	}
383 
384 	if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
385 		sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
386 	} else {
387 		if (hlen == sizeof(struct ip)) {
388 			sum = in_cksum_hdr(ip);
389 		} else {
390 			sum = in_cksum(m, hlen);
391 		}
392 	}
393 	if (sum) {
394 		V_ipstat.ips_badsum++;
395 		goto bad;
396 	}
397 
398 #ifdef ALTQ
399 	if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
400 		/* packet is dropped by traffic conditioner */
401 		return;
402 #endif
403 
404 	/*
405 	 * Convert fields to host representation.
406 	 */
407 	ip->ip_len = ntohs(ip->ip_len);
408 	if (ip->ip_len < hlen) {
409 		V_ipstat.ips_badlen++;
410 		goto bad;
411 	}
412 	ip->ip_off = ntohs(ip->ip_off);
413 
414 	/*
415 	 * Check that the amount of data in the buffers
416 	 * is as at least much as the IP header would have us expect.
417 	 * Trim mbufs if longer than we expect.
418 	 * Drop packet if shorter than we expect.
419 	 */
420 	if (m->m_pkthdr.len < ip->ip_len) {
421 tooshort:
422 		V_ipstat.ips_tooshort++;
423 		goto bad;
424 	}
425 	if (m->m_pkthdr.len > ip->ip_len) {
426 		if (m->m_len == m->m_pkthdr.len) {
427 			m->m_len = ip->ip_len;
428 			m->m_pkthdr.len = ip->ip_len;
429 		} else
430 			m_adj(m, ip->ip_len - m->m_pkthdr.len);
431 	}
432 #ifdef IPSEC
433 	/*
434 	 * Bypass packet filtering for packets from a tunnel (gif).
435 	 */
436 	if (ip_ipsec_filtertunnel(m))
437 		goto passin;
438 #endif /* IPSEC */
439 
440 	/*
441 	 * Run through list of hooks for input packets.
442 	 *
443 	 * NB: Beware of the destination address changing (e.g.
444 	 *     by NAT rewriting).  When this happens, tell
445 	 *     ip_forward to do the right thing.
446 	 */
447 
448 	/* Jump over all PFIL processing if hooks are not active. */
449 	if (!PFIL_HOOKED(&inet_pfil_hook))
450 		goto passin;
451 
452 	odst = ip->ip_dst;
453 	if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif,
454 	    PFIL_IN, NULL) != 0)
455 		return;
456 	if (m == NULL)			/* consumed by filter */
457 		return;
458 
459 	ip = mtod(m, struct ip *);
460 	dchg = (odst.s_addr != ip->ip_dst.s_addr);
461 
462 #ifdef IPFIREWALL_FORWARD
463 	if (m->m_flags & M_FASTFWD_OURS) {
464 		m->m_flags &= ~M_FASTFWD_OURS;
465 		goto ours;
466 	}
467 	if ((dchg = (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL)) != 0) {
468 		/*
469 		 * Directly ship on the packet.  This allows to forward packets
470 		 * that were destined for us to some other directly connected
471 		 * host.
472 		 */
473 		ip_forward(m, dchg);
474 		return;
475 	}
476 #endif /* IPFIREWALL_FORWARD */
477 
478 passin:
479 	/*
480 	 * Process options and, if not destined for us,
481 	 * ship it on.  ip_dooptions returns 1 when an
482 	 * error was detected (causing an icmp message
483 	 * to be sent and the original packet to be freed).
484 	 */
485 	if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
486 		return;
487 
488         /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
489          * matter if it is destined to another node, or whether it is
490          * a multicast one, RSVP wants it! and prevents it from being forwarded
491          * anywhere else. Also checks if the rsvp daemon is running before
492 	 * grabbing the packet.
493          */
494 	if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP)
495 		goto ours;
496 
497 	/*
498 	 * Check our list of addresses, to see if the packet is for us.
499 	 * If we don't have any addresses, assume any unicast packet
500 	 * we receive might be for us (and let the upper layers deal
501 	 * with it).
502 	 */
503 	if (TAILQ_EMPTY(&V_in_ifaddrhead) &&
504 	    (m->m_flags & (M_MCAST|M_BCAST)) == 0)
505 		goto ours;
506 
507 	/*
508 	 * Enable a consistency check between the destination address
509 	 * and the arrival interface for a unicast packet (the RFC 1122
510 	 * strong ES model) if IP forwarding is disabled and the packet
511 	 * is not locally generated and the packet is not subject to
512 	 * 'ipfw fwd'.
513 	 *
514 	 * XXX - Checking also should be disabled if the destination
515 	 * address is ipnat'ed to a different interface.
516 	 *
517 	 * XXX - Checking is incompatible with IP aliases added
518 	 * to the loopback interface instead of the interface where
519 	 * the packets are received.
520 	 *
521 	 * XXX - This is the case for carp vhost IPs as well so we
522 	 * insert a workaround. If the packet got here, we already
523 	 * checked with carp_iamatch() and carp_forus().
524 	 */
525 	checkif = V_ip_checkinterface && (V_ipforwarding == 0) &&
526 	    m->m_pkthdr.rcvif != NULL &&
527 	    ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) &&
528 #ifdef DEV_CARP
529 	    !m->m_pkthdr.rcvif->if_carp &&
530 #endif
531 	    (dchg == 0);
532 
533 	/*
534 	 * Check for exact addresses in the hash bucket.
535 	 */
536 	LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
537 		/*
538 		 * If the address matches, verify that the packet
539 		 * arrived via the correct interface if checking is
540 		 * enabled.
541 		 */
542 		if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr &&
543 		    (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif))
544 			goto ours;
545 	}
546 	/*
547 	 * Check for broadcast addresses.
548 	 *
549 	 * Only accept broadcast packets that arrive via the matching
550 	 * interface.  Reception of forwarded directed broadcasts would
551 	 * be handled via ip_forward() and ether_output() with the loopback
552 	 * into the stack for SIMPLEX interfaces handled by ether_output().
553 	 */
554 	if (m->m_pkthdr.rcvif != NULL &&
555 	    m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
556 	        TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
557 			if (ifa->ifa_addr->sa_family != AF_INET)
558 				continue;
559 			ia = ifatoia(ifa);
560 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
561 			    ip->ip_dst.s_addr)
562 				goto ours;
563 			if (ia->ia_netbroadcast.s_addr == ip->ip_dst.s_addr)
564 				goto ours;
565 #ifdef BOOTP_COMPAT
566 			if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY)
567 				goto ours;
568 #endif
569 		}
570 	}
571 	/* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */
572 	if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) {
573 		V_ipstat.ips_cantforward++;
574 		m_freem(m);
575 		return;
576 	}
577 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
578 		struct in_multi *inm;
579 		if (V_ip_mrouter) {
580 			/*
581 			 * If we are acting as a multicast router, all
582 			 * incoming multicast packets are passed to the
583 			 * kernel-level multicast forwarding function.
584 			 * The packet is returned (relatively) intact; if
585 			 * ip_mforward() returns a non-zero value, the packet
586 			 * must be discarded, else it may be accepted below.
587 			 */
588 			if (ip_mforward &&
589 			    ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) {
590 				V_ipstat.ips_cantforward++;
591 				m_freem(m);
592 				return;
593 			}
594 
595 			/*
596 			 * The process-level routing daemon needs to receive
597 			 * all multicast IGMP packets, whether or not this
598 			 * host belongs to their destination groups.
599 			 */
600 			if (ip->ip_p == IPPROTO_IGMP)
601 				goto ours;
602 			V_ipstat.ips_forward++;
603 		}
604 		/*
605 		 * See if we belong to the destination multicast group on the
606 		 * arrival interface.
607 		 */
608 		IN_MULTI_LOCK();
609 		IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
610 		IN_MULTI_UNLOCK();
611 		if (inm == NULL) {
612 			V_ipstat.ips_notmember++;
613 			m_freem(m);
614 			return;
615 		}
616 		goto ours;
617 	}
618 	if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
619 		goto ours;
620 	if (ip->ip_dst.s_addr == INADDR_ANY)
621 		goto ours;
622 
623 	/*
624 	 * FAITH(Firewall Aided Internet Translator)
625 	 */
626 	if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
627 		if (V_ip_keepfaith) {
628 			if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP)
629 				goto ours;
630 		}
631 		m_freem(m);
632 		return;
633 	}
634 
635 	/*
636 	 * Not for us; forward if possible and desirable.
637 	 */
638 	if (V_ipforwarding == 0) {
639 		V_ipstat.ips_cantforward++;
640 		m_freem(m);
641 	} else {
642 #ifdef IPSEC
643 		if (ip_ipsec_fwd(m))
644 			goto bad;
645 #endif /* IPSEC */
646 		ip_forward(m, dchg);
647 	}
648 	return;
649 
650 ours:
651 #ifdef IPSTEALTH
652 	/*
653 	 * IPSTEALTH: Process non-routing options only
654 	 * if the packet is destined for us.
655 	 */
656 	if (V_ipstealth && hlen > sizeof (struct ip) &&
657 	    ip_dooptions(m, 1))
658 		return;
659 #endif /* IPSTEALTH */
660 
661 	/* Count the packet in the ip address stats */
662 	if (ia != NULL) {
663 		ia->ia_ifa.if_ipackets++;
664 		ia->ia_ifa.if_ibytes += m->m_pkthdr.len;
665 	}
666 
667 	/*
668 	 * Attempt reassembly; if it succeeds, proceed.
669 	 * ip_reass() will return a different mbuf.
670 	 */
671 	if (ip->ip_off & (IP_MF | IP_OFFMASK)) {
672 		m = ip_reass(m);
673 		if (m == NULL)
674 			return;
675 		ip = mtod(m, struct ip *);
676 		/* Get the header length of the reassembled packet */
677 		hlen = ip->ip_hl << 2;
678 	}
679 
680 	/*
681 	 * Further protocols expect the packet length to be w/o the
682 	 * IP header.
683 	 */
684 	ip->ip_len -= hlen;
685 
686 #ifdef IPSEC
687 	/*
688 	 * enforce IPsec policy checking if we are seeing last header.
689 	 * note that we do not visit this with protocols with pcb layer
690 	 * code - like udp/tcp/raw ip.
691 	 */
692 	if (ip_ipsec_input(m))
693 		goto bad;
694 #endif /* IPSEC */
695 
696 	/*
697 	 * Switch out to protocol's input routine.
698 	 */
699 	V_ipstat.ips_delivered++;
700 
701 	(*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen);
702 	return;
703 bad:
704 	m_freem(m);
705 }
706 
707 /*
708  * After maxnipq has been updated, propagate the change to UMA.  The UMA zone
709  * max has slightly different semantics than the sysctl, for historical
710  * reasons.
711  */
712 static void
713 maxnipq_update(void)
714 {
715 	INIT_VNET_INET(curvnet);
716 
717 	/*
718 	 * -1 for unlimited allocation.
719 	 */
720 	if (V_maxnipq < 0)
721 		uma_zone_set_max(V_ipq_zone, 0);
722 	/*
723 	 * Positive number for specific bound.
724 	 */
725 	if (V_maxnipq > 0)
726 		uma_zone_set_max(V_ipq_zone, V_maxnipq);
727 	/*
728 	 * Zero specifies no further fragment queue allocation -- set the
729 	 * bound very low, but rely on implementation elsewhere to actually
730 	 * prevent allocation and reclaim current queues.
731 	 */
732 	if (V_maxnipq == 0)
733 		uma_zone_set_max(V_ipq_zone, 1);
734 }
735 
736 static void
737 ipq_zone_change(void *tag)
738 {
739 	INIT_VNET_INET(curvnet);
740 
741 	if (V_maxnipq > 0 && V_maxnipq < (nmbclusters / 32)) {
742 		V_maxnipq = nmbclusters / 32;
743 		maxnipq_update();
744 	}
745 }
746 
747 static int
748 sysctl_maxnipq(SYSCTL_HANDLER_ARGS)
749 {
750 	INIT_VNET_INET(curvnet);
751 	int error, i;
752 
753 	i = V_maxnipq;
754 	error = sysctl_handle_int(oidp, &i, 0, req);
755 	if (error || !req->newptr)
756 		return (error);
757 
758 	/*
759 	 * XXXRW: Might be a good idea to sanity check the argument and place
760 	 * an extreme upper bound.
761 	 */
762 	if (i < -1)
763 		return (EINVAL);
764 	V_maxnipq = i;
765 	maxnipq_update();
766 	return (0);
767 }
768 
769 SYSCTL_PROC(_net_inet_ip, OID_AUTO, maxfragpackets, CTLTYPE_INT|CTLFLAG_RW,
770     NULL, 0, sysctl_maxnipq, "I",
771     "Maximum number of IPv4 fragment reassembly queue entries");
772 
773 /*
774  * Take incoming datagram fragment and try to reassemble it into
775  * whole datagram.  If the argument is the first fragment or one
776  * in between the function will return NULL and store the mbuf
777  * in the fragment chain.  If the argument is the last fragment
778  * the packet will be reassembled and the pointer to the new
779  * mbuf returned for further processing.  Only m_tags attached
780  * to the first packet/fragment are preserved.
781  * The IP header is *NOT* adjusted out of iplen.
782  */
783 struct mbuf *
784 ip_reass(struct mbuf *m)
785 {
786 	INIT_VNET_INET(curvnet);
787 	struct ip *ip;
788 	struct mbuf *p, *q, *nq, *t;
789 	struct ipq *fp = NULL;
790 	struct ipqhead *head;
791 	int i, hlen, next;
792 	u_int8_t ecn, ecn0;
793 	u_short hash;
794 
795 	/* If maxnipq or maxfragsperpacket are 0, never accept fragments. */
796 	if (V_maxnipq == 0 || V_maxfragsperpacket == 0) {
797 		V_ipstat.ips_fragments++;
798 		V_ipstat.ips_fragdropped++;
799 		m_freem(m);
800 		return (NULL);
801 	}
802 
803 	ip = mtod(m, struct ip *);
804 	hlen = ip->ip_hl << 2;
805 
806 	hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
807 	head = &V_ipq[hash];
808 	IPQ_LOCK();
809 
810 	/*
811 	 * Look for queue of fragments
812 	 * of this datagram.
813 	 */
814 	TAILQ_FOREACH(fp, head, ipq_list)
815 		if (ip->ip_id == fp->ipq_id &&
816 		    ip->ip_src.s_addr == fp->ipq_src.s_addr &&
817 		    ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
818 #ifdef MAC
819 		    mac_ipq_match(m, fp) &&
820 #endif
821 		    ip->ip_p == fp->ipq_p)
822 			goto found;
823 
824 	fp = NULL;
825 
826 	/*
827 	 * Attempt to trim the number of allocated fragment queues if it
828 	 * exceeds the administrative limit.
829 	 */
830 	if ((V_nipq > V_maxnipq) && (V_maxnipq > 0)) {
831 		/*
832 		 * drop something from the tail of the current queue
833 		 * before proceeding further
834 		 */
835 		struct ipq *q = TAILQ_LAST(head, ipqhead);
836 		if (q == NULL) {   /* gak */
837 			for (i = 0; i < IPREASS_NHASH; i++) {
838 				struct ipq *r = TAILQ_LAST(&V_ipq[i], ipqhead);
839 				if (r) {
840 					V_ipstat.ips_fragtimeout +=
841 					    r->ipq_nfrags;
842 					ip_freef(&V_ipq[i], r);
843 					break;
844 				}
845 			}
846 		} else {
847 			V_ipstat.ips_fragtimeout += q->ipq_nfrags;
848 			ip_freef(head, q);
849 		}
850 	}
851 
852 found:
853 	/*
854 	 * Adjust ip_len to not reflect header,
855 	 * convert offset of this to bytes.
856 	 */
857 	ip->ip_len -= hlen;
858 	if (ip->ip_off & IP_MF) {
859 		/*
860 		 * Make sure that fragments have a data length
861 		 * that's a non-zero multiple of 8 bytes.
862 		 */
863 		if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
864 			V_ipstat.ips_toosmall++; /* XXX */
865 			goto dropfrag;
866 		}
867 		m->m_flags |= M_FRAG;
868 	} else
869 		m->m_flags &= ~M_FRAG;
870 	ip->ip_off <<= 3;
871 
872 
873 	/*
874 	 * Attempt reassembly; if it succeeds, proceed.
875 	 * ip_reass() will return a different mbuf.
876 	 */
877 	V_ipstat.ips_fragments++;
878 	m->m_pkthdr.header = ip;
879 
880 	/* Previous ip_reass() started here. */
881 	/*
882 	 * Presence of header sizes in mbufs
883 	 * would confuse code below.
884 	 */
885 	m->m_data += hlen;
886 	m->m_len -= hlen;
887 
888 	/*
889 	 * If first fragment to arrive, create a reassembly queue.
890 	 */
891 	if (fp == NULL) {
892 		fp = uma_zalloc(V_ipq_zone, M_NOWAIT);
893 		if (fp == NULL)
894 			goto dropfrag;
895 #ifdef MAC
896 		if (mac_ipq_init(fp, M_NOWAIT) != 0) {
897 			uma_zfree(V_ipq_zone, fp);
898 			fp = NULL;
899 			goto dropfrag;
900 		}
901 		mac_ipq_create(m, fp);
902 #endif
903 		TAILQ_INSERT_HEAD(head, fp, ipq_list);
904 		V_nipq++;
905 		fp->ipq_nfrags = 1;
906 		fp->ipq_ttl = IPFRAGTTL;
907 		fp->ipq_p = ip->ip_p;
908 		fp->ipq_id = ip->ip_id;
909 		fp->ipq_src = ip->ip_src;
910 		fp->ipq_dst = ip->ip_dst;
911 		fp->ipq_frags = m;
912 		m->m_nextpkt = NULL;
913 		goto done;
914 	} else {
915 		fp->ipq_nfrags++;
916 #ifdef MAC
917 		mac_ipq_update(m, fp);
918 #endif
919 	}
920 
921 #define GETIP(m)	((struct ip*)((m)->m_pkthdr.header))
922 
923 	/*
924 	 * Handle ECN by comparing this segment with the first one;
925 	 * if CE is set, do not lose CE.
926 	 * drop if CE and not-ECT are mixed for the same packet.
927 	 */
928 	ecn = ip->ip_tos & IPTOS_ECN_MASK;
929 	ecn0 = GETIP(fp->ipq_frags)->ip_tos & IPTOS_ECN_MASK;
930 	if (ecn == IPTOS_ECN_CE) {
931 		if (ecn0 == IPTOS_ECN_NOTECT)
932 			goto dropfrag;
933 		if (ecn0 != IPTOS_ECN_CE)
934 			GETIP(fp->ipq_frags)->ip_tos |= IPTOS_ECN_CE;
935 	}
936 	if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT)
937 		goto dropfrag;
938 
939 	/*
940 	 * Find a segment which begins after this one does.
941 	 */
942 	for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt)
943 		if (GETIP(q)->ip_off > ip->ip_off)
944 			break;
945 
946 	/*
947 	 * If there is a preceding segment, it may provide some of
948 	 * our data already.  If so, drop the data from the incoming
949 	 * segment.  If it provides all of our data, drop us, otherwise
950 	 * stick new segment in the proper place.
951 	 *
952 	 * If some of the data is dropped from the the preceding
953 	 * segment, then it's checksum is invalidated.
954 	 */
955 	if (p) {
956 		i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off;
957 		if (i > 0) {
958 			if (i >= ip->ip_len)
959 				goto dropfrag;
960 			m_adj(m, i);
961 			m->m_pkthdr.csum_flags = 0;
962 			ip->ip_off += i;
963 			ip->ip_len -= i;
964 		}
965 		m->m_nextpkt = p->m_nextpkt;
966 		p->m_nextpkt = m;
967 	} else {
968 		m->m_nextpkt = fp->ipq_frags;
969 		fp->ipq_frags = m;
970 	}
971 
972 	/*
973 	 * While we overlap succeeding segments trim them or,
974 	 * if they are completely covered, dequeue them.
975 	 */
976 	for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off;
977 	     q = nq) {
978 		i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off;
979 		if (i < GETIP(q)->ip_len) {
980 			GETIP(q)->ip_len -= i;
981 			GETIP(q)->ip_off += i;
982 			m_adj(q, i);
983 			q->m_pkthdr.csum_flags = 0;
984 			break;
985 		}
986 		nq = q->m_nextpkt;
987 		m->m_nextpkt = nq;
988 		V_ipstat.ips_fragdropped++;
989 		fp->ipq_nfrags--;
990 		m_freem(q);
991 	}
992 
993 	/*
994 	 * Check for complete reassembly and perform frag per packet
995 	 * limiting.
996 	 *
997 	 * Frag limiting is performed here so that the nth frag has
998 	 * a chance to complete the packet before we drop the packet.
999 	 * As a result, n+1 frags are actually allowed per packet, but
1000 	 * only n will ever be stored. (n = maxfragsperpacket.)
1001 	 *
1002 	 */
1003 	next = 0;
1004 	for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
1005 		if (GETIP(q)->ip_off != next) {
1006 			if (fp->ipq_nfrags > V_maxfragsperpacket) {
1007 				V_ipstat.ips_fragdropped += fp->ipq_nfrags;
1008 				ip_freef(head, fp);
1009 			}
1010 			goto done;
1011 		}
1012 		next += GETIP(q)->ip_len;
1013 	}
1014 	/* Make sure the last packet didn't have the IP_MF flag */
1015 	if (p->m_flags & M_FRAG) {
1016 		if (fp->ipq_nfrags > V_maxfragsperpacket) {
1017 			V_ipstat.ips_fragdropped += fp->ipq_nfrags;
1018 			ip_freef(head, fp);
1019 		}
1020 		goto done;
1021 	}
1022 
1023 	/*
1024 	 * Reassembly is complete.  Make sure the packet is a sane size.
1025 	 */
1026 	q = fp->ipq_frags;
1027 	ip = GETIP(q);
1028 	if (next + (ip->ip_hl << 2) > IP_MAXPACKET) {
1029 		V_ipstat.ips_toolong++;
1030 		V_ipstat.ips_fragdropped += fp->ipq_nfrags;
1031 		ip_freef(head, fp);
1032 		goto done;
1033 	}
1034 
1035 	/*
1036 	 * Concatenate fragments.
1037 	 */
1038 	m = q;
1039 	t = m->m_next;
1040 	m->m_next = NULL;
1041 	m_cat(m, t);
1042 	nq = q->m_nextpkt;
1043 	q->m_nextpkt = NULL;
1044 	for (q = nq; q != NULL; q = nq) {
1045 		nq = q->m_nextpkt;
1046 		q->m_nextpkt = NULL;
1047 		m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
1048 		m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
1049 		m_cat(m, q);
1050 	}
1051 	/*
1052 	 * In order to do checksumming faster we do 'end-around carry' here
1053 	 * (and not in for{} loop), though it implies we are not going to
1054 	 * reassemble more than 64k fragments.
1055 	 */
1056 	m->m_pkthdr.csum_data =
1057 	    (m->m_pkthdr.csum_data & 0xffff) + (m->m_pkthdr.csum_data >> 16);
1058 #ifdef MAC
1059 	mac_ipq_reassemble(fp, m);
1060 	mac_ipq_destroy(fp);
1061 #endif
1062 
1063 	/*
1064 	 * Create header for new ip packet by modifying header of first
1065 	 * packet;  dequeue and discard fragment reassembly header.
1066 	 * Make header visible.
1067 	 */
1068 	ip->ip_len = (ip->ip_hl << 2) + next;
1069 	ip->ip_src = fp->ipq_src;
1070 	ip->ip_dst = fp->ipq_dst;
1071 	TAILQ_REMOVE(head, fp, ipq_list);
1072 	V_nipq--;
1073 	uma_zfree(V_ipq_zone, fp);
1074 	m->m_len += (ip->ip_hl << 2);
1075 	m->m_data -= (ip->ip_hl << 2);
1076 	/* some debugging cruft by sklower, below, will go away soon */
1077 	if (m->m_flags & M_PKTHDR)	/* XXX this should be done elsewhere */
1078 		m_fixhdr(m);
1079 	V_ipstat.ips_reassembled++;
1080 	IPQ_UNLOCK();
1081 	return (m);
1082 
1083 dropfrag:
1084 	V_ipstat.ips_fragdropped++;
1085 	if (fp != NULL)
1086 		fp->ipq_nfrags--;
1087 	m_freem(m);
1088 done:
1089 	IPQ_UNLOCK();
1090 	return (NULL);
1091 
1092 #undef GETIP
1093 }
1094 
1095 /*
1096  * Free a fragment reassembly header and all
1097  * associated datagrams.
1098  */
1099 static void
1100 ip_freef(struct ipqhead *fhp, struct ipq *fp)
1101 {
1102 	INIT_VNET_INET(curvnet);
1103 	struct mbuf *q;
1104 
1105 	IPQ_LOCK_ASSERT();
1106 
1107 	while (fp->ipq_frags) {
1108 		q = fp->ipq_frags;
1109 		fp->ipq_frags = q->m_nextpkt;
1110 		m_freem(q);
1111 	}
1112 	TAILQ_REMOVE(fhp, fp, ipq_list);
1113 	uma_zfree(V_ipq_zone, fp);
1114 	V_nipq--;
1115 }
1116 
1117 /*
1118  * IP timer processing;
1119  * if a timer expires on a reassembly
1120  * queue, discard it.
1121  */
1122 void
1123 ip_slowtimo(void)
1124 {
1125 	VNET_ITERATOR_DECL(vnet_iter);
1126 	struct ipq *fp;
1127 	int i;
1128 
1129 	IPQ_LOCK();
1130 	VNET_LIST_RLOCK();
1131 	VNET_FOREACH(vnet_iter) {
1132 		CURVNET_SET(vnet_iter);
1133 		INIT_VNET_INET(vnet_iter);
1134 		for (i = 0; i < IPREASS_NHASH; i++) {
1135 			for(fp = TAILQ_FIRST(&V_ipq[i]); fp;) {
1136 				struct ipq *fpp;
1137 
1138 				fpp = fp;
1139 				fp = TAILQ_NEXT(fp, ipq_list);
1140 				if(--fpp->ipq_ttl == 0) {
1141 					V_ipstat.ips_fragtimeout +=
1142 					    fpp->ipq_nfrags;
1143 					ip_freef(&V_ipq[i], fpp);
1144 				}
1145 			}
1146 		}
1147 		/*
1148 		 * If we are over the maximum number of fragments
1149 		 * (due to the limit being lowered), drain off
1150 		 * enough to get down to the new limit.
1151 		 */
1152 		if (V_maxnipq >= 0 && V_nipq > V_maxnipq) {
1153 			for (i = 0; i < IPREASS_NHASH; i++) {
1154 				while (V_nipq > V_maxnipq &&
1155 				    !TAILQ_EMPTY(&V_ipq[i])) {
1156 					V_ipstat.ips_fragdropped +=
1157 					    TAILQ_FIRST(&V_ipq[i])->ipq_nfrags;
1158 					ip_freef(&V_ipq[i],
1159 					    TAILQ_FIRST(&V_ipq[i]));
1160 				}
1161 			}
1162 		}
1163 		CURVNET_RESTORE();
1164 	}
1165 	VNET_LIST_RUNLOCK();
1166 	IPQ_UNLOCK();
1167 }
1168 
1169 /*
1170  * Drain off all datagram fragments.
1171  */
1172 void
1173 ip_drain(void)
1174 {
1175 	VNET_ITERATOR_DECL(vnet_iter);
1176 	int     i;
1177 
1178 	IPQ_LOCK();
1179 	VNET_LIST_RLOCK();
1180 	VNET_FOREACH(vnet_iter) {
1181 		CURVNET_SET(vnet_iter);
1182 		INIT_VNET_INET(vnet_iter);
1183 		for (i = 0; i < IPREASS_NHASH; i++) {
1184 			while(!TAILQ_EMPTY(&V_ipq[i])) {
1185 				V_ipstat.ips_fragdropped +=
1186 				    TAILQ_FIRST(&V_ipq[i])->ipq_nfrags;
1187 				ip_freef(&V_ipq[i], TAILQ_FIRST(&V_ipq[i]));
1188 			}
1189 		}
1190 		CURVNET_RESTORE();
1191 	}
1192 	VNET_LIST_RUNLOCK();
1193 	IPQ_UNLOCK();
1194 	in_rtqdrain();
1195 }
1196 
1197 /*
1198  * The protocol to be inserted into ip_protox[] must be already registered
1199  * in inetsw[], either statically or through pf_proto_register().
1200  */
1201 int
1202 ipproto_register(u_char ipproto)
1203 {
1204 	struct protosw *pr;
1205 
1206 	/* Sanity checks. */
1207 	if (ipproto == 0)
1208 		return (EPROTONOSUPPORT);
1209 
1210 	/*
1211 	 * The protocol slot must not be occupied by another protocol
1212 	 * already.  An index pointing to IPPROTO_RAW is unused.
1213 	 */
1214 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
1215 	if (pr == NULL)
1216 		return (EPFNOSUPPORT);
1217 	if (ip_protox[ipproto] != pr - inetsw)	/* IPPROTO_RAW */
1218 		return (EEXIST);
1219 
1220 	/* Find the protocol position in inetsw[] and set the index. */
1221 	for (pr = inetdomain.dom_protosw;
1222 	     pr < inetdomain.dom_protoswNPROTOSW; pr++) {
1223 		if (pr->pr_domain->dom_family == PF_INET &&
1224 		    pr->pr_protocol && pr->pr_protocol == ipproto) {
1225 			/* Be careful to only index valid IP protocols. */
1226 			if (pr->pr_protocol < IPPROTO_MAX) {
1227 				ip_protox[pr->pr_protocol] = pr - inetsw;
1228 				return (0);
1229 			} else
1230 				return (EINVAL);
1231 		}
1232 	}
1233 	return (EPROTONOSUPPORT);
1234 }
1235 
1236 int
1237 ipproto_unregister(u_char ipproto)
1238 {
1239 	struct protosw *pr;
1240 
1241 	/* Sanity checks. */
1242 	if (ipproto == 0)
1243 		return (EPROTONOSUPPORT);
1244 
1245 	/* Check if the protocol was indeed registered. */
1246 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
1247 	if (pr == NULL)
1248 		return (EPFNOSUPPORT);
1249 	if (ip_protox[ipproto] == pr - inetsw)  /* IPPROTO_RAW */
1250 		return (ENOENT);
1251 
1252 	/* Reset the protocol slot to IPPROTO_RAW. */
1253 	ip_protox[ipproto] = pr - inetsw;
1254 	return (0);
1255 }
1256 
1257 /*
1258  * Given address of next destination (final or next hop),
1259  * return internet address info of interface to be used to get there.
1260  */
1261 struct in_ifaddr *
1262 ip_rtaddr(struct in_addr dst, u_int fibnum)
1263 {
1264 	struct route sro;
1265 	struct sockaddr_in *sin;
1266 	struct in_ifaddr *ifa;
1267 
1268 	bzero(&sro, sizeof(sro));
1269 	sin = (struct sockaddr_in *)&sro.ro_dst;
1270 	sin->sin_family = AF_INET;
1271 	sin->sin_len = sizeof(*sin);
1272 	sin->sin_addr = dst;
1273 	in_rtalloc_ign(&sro, RTF_CLONING, fibnum);
1274 
1275 	if (sro.ro_rt == NULL)
1276 		return (NULL);
1277 
1278 	ifa = ifatoia(sro.ro_rt->rt_ifa);
1279 	RTFREE(sro.ro_rt);
1280 	return (ifa);
1281 }
1282 
1283 u_char inetctlerrmap[PRC_NCMDS] = {
1284 	0,		0,		0,		0,
1285 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1286 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1287 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1288 	0,		0,		EHOSTUNREACH,	0,
1289 	ENOPROTOOPT,	ECONNREFUSED
1290 };
1291 
1292 /*
1293  * Forward a packet.  If some error occurs return the sender
1294  * an icmp packet.  Note we can't always generate a meaningful
1295  * icmp message because icmp doesn't have a large enough repertoire
1296  * of codes and types.
1297  *
1298  * If not forwarding, just drop the packet.  This could be confusing
1299  * if ipforwarding was zero but some routing protocol was advancing
1300  * us as a gateway to somewhere.  However, we must let the routing
1301  * protocol deal with that.
1302  *
1303  * The srcrt parameter indicates whether the packet is being forwarded
1304  * via a source route.
1305  */
1306 void
1307 ip_forward(struct mbuf *m, int srcrt)
1308 {
1309 	INIT_VNET_INET(curvnet);
1310 	struct ip *ip = mtod(m, struct ip *);
1311 	struct in_ifaddr *ia = NULL;
1312 	struct mbuf *mcopy;
1313 	struct in_addr dest;
1314 	struct route ro;
1315 	int error, type = 0, code = 0, mtu = 0;
1316 
1317 	if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
1318 		V_ipstat.ips_cantforward++;
1319 		m_freem(m);
1320 		return;
1321 	}
1322 #ifdef IPSTEALTH
1323 	if (!V_ipstealth) {
1324 #endif
1325 		if (ip->ip_ttl <= IPTTLDEC) {
1326 			icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS,
1327 			    0, 0);
1328 			return;
1329 		}
1330 #ifdef IPSTEALTH
1331 	}
1332 #endif
1333 
1334 	ia = ip_rtaddr(ip->ip_dst, M_GETFIB(m));
1335 	if (!srcrt && ia == NULL) {
1336 		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
1337 		return;
1338 	}
1339 
1340 	/*
1341 	 * Save the IP header and at most 8 bytes of the payload,
1342 	 * in case we need to generate an ICMP message to the src.
1343 	 *
1344 	 * XXX this can be optimized a lot by saving the data in a local
1345 	 * buffer on the stack (72 bytes at most), and only allocating the
1346 	 * mbuf if really necessary. The vast majority of the packets
1347 	 * are forwarded without having to send an ICMP back (either
1348 	 * because unnecessary, or because rate limited), so we are
1349 	 * really we are wasting a lot of work here.
1350 	 *
1351 	 * We don't use m_copy() because it might return a reference
1352 	 * to a shared cluster. Both this function and ip_output()
1353 	 * assume exclusive access to the IP header in `m', so any
1354 	 * data in a cluster may change before we reach icmp_error().
1355 	 */
1356 	MGETHDR(mcopy, M_DONTWAIT, m->m_type);
1357 	if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_DONTWAIT)) {
1358 		/*
1359 		 * It's probably ok if the pkthdr dup fails (because
1360 		 * the deep copy of the tag chain failed), but for now
1361 		 * be conservative and just discard the copy since
1362 		 * code below may some day want the tags.
1363 		 */
1364 		m_free(mcopy);
1365 		mcopy = NULL;
1366 	}
1367 	if (mcopy != NULL) {
1368 		mcopy->m_len = min(ip->ip_len, M_TRAILINGSPACE(mcopy));
1369 		mcopy->m_pkthdr.len = mcopy->m_len;
1370 		m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1371 	}
1372 
1373 #ifdef IPSTEALTH
1374 	if (!V_ipstealth) {
1375 #endif
1376 		ip->ip_ttl -= IPTTLDEC;
1377 #ifdef IPSTEALTH
1378 	}
1379 #endif
1380 
1381 	/*
1382 	 * If forwarding packet using same interface that it came in on,
1383 	 * perhaps should send a redirect to sender to shortcut a hop.
1384 	 * Only send redirect if source is sending directly to us,
1385 	 * and if packet was not source routed (or has any options).
1386 	 * Also, don't send redirect if forwarding using a default route
1387 	 * or a route modified by a redirect.
1388 	 */
1389 	dest.s_addr = 0;
1390 	if (!srcrt && V_ipsendredirects && ia->ia_ifp == m->m_pkthdr.rcvif) {
1391 		struct sockaddr_in *sin;
1392 		struct rtentry *rt;
1393 
1394 		bzero(&ro, sizeof(ro));
1395 		sin = (struct sockaddr_in *)&ro.ro_dst;
1396 		sin->sin_family = AF_INET;
1397 		sin->sin_len = sizeof(*sin);
1398 		sin->sin_addr = ip->ip_dst;
1399 		in_rtalloc_ign(&ro, RTF_CLONING, M_GETFIB(m));
1400 
1401 		rt = ro.ro_rt;
1402 
1403 		if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1404 		    satosin(rt_key(rt))->sin_addr.s_addr != 0) {
1405 #define	RTA(rt)	((struct in_ifaddr *)(rt->rt_ifa))
1406 			u_long src = ntohl(ip->ip_src.s_addr);
1407 
1408 			if (RTA(rt) &&
1409 			    (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1410 				if (rt->rt_flags & RTF_GATEWAY)
1411 					dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr;
1412 				else
1413 					dest.s_addr = ip->ip_dst.s_addr;
1414 				/* Router requirements says to only send host redirects */
1415 				type = ICMP_REDIRECT;
1416 				code = ICMP_REDIRECT_HOST;
1417 			}
1418 		}
1419 		if (rt)
1420 			RTFREE(rt);
1421 	}
1422 
1423 	/*
1424 	 * Try to cache the route MTU from ip_output so we can consider it for
1425 	 * the ICMP_UNREACH_NEEDFRAG "Next-Hop MTU" field described in RFC1191.
1426 	 */
1427 	bzero(&ro, sizeof(ro));
1428 
1429 	error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
1430 
1431 	if (error == EMSGSIZE && ro.ro_rt)
1432 		mtu = ro.ro_rt->rt_rmx.rmx_mtu;
1433 	if (ro.ro_rt)
1434 		RTFREE(ro.ro_rt);
1435 
1436 	if (error)
1437 		V_ipstat.ips_cantforward++;
1438 	else {
1439 		V_ipstat.ips_forward++;
1440 		if (type)
1441 			V_ipstat.ips_redirectsent++;
1442 		else {
1443 			if (mcopy)
1444 				m_freem(mcopy);
1445 			return;
1446 		}
1447 	}
1448 	if (mcopy == NULL)
1449 		return;
1450 
1451 	switch (error) {
1452 
1453 	case 0:				/* forwarded, but need redirect */
1454 		/* type, code set above */
1455 		break;
1456 
1457 	case ENETUNREACH:		/* shouldn't happen, checked above */
1458 	case EHOSTUNREACH:
1459 	case ENETDOWN:
1460 	case EHOSTDOWN:
1461 	default:
1462 		type = ICMP_UNREACH;
1463 		code = ICMP_UNREACH_HOST;
1464 		break;
1465 
1466 	case EMSGSIZE:
1467 		type = ICMP_UNREACH;
1468 		code = ICMP_UNREACH_NEEDFRAG;
1469 
1470 #ifdef IPSEC
1471 		/*
1472 		 * If IPsec is configured for this path,
1473 		 * override any possibly mtu value set by ip_output.
1474 		 */
1475 		mtu = ip_ipsec_mtu(m, mtu);
1476 #endif /* IPSEC */
1477 		/*
1478 		 * If the MTU was set before make sure we are below the
1479 		 * interface MTU.
1480 		 * If the MTU wasn't set before use the interface mtu or
1481 		 * fall back to the next smaller mtu step compared to the
1482 		 * current packet size.
1483 		 */
1484 		if (mtu != 0) {
1485 			if (ia != NULL)
1486 				mtu = min(mtu, ia->ia_ifp->if_mtu);
1487 		} else {
1488 			if (ia != NULL)
1489 				mtu = ia->ia_ifp->if_mtu;
1490 			else
1491 				mtu = ip_next_mtu(ip->ip_len, 0);
1492 		}
1493 		V_ipstat.ips_cantfrag++;
1494 		break;
1495 
1496 	case ENOBUFS:
1497 		/*
1498 		 * A router should not generate ICMP_SOURCEQUENCH as
1499 		 * required in RFC1812 Requirements for IP Version 4 Routers.
1500 		 * Source quench could be a big problem under DoS attacks,
1501 		 * or if the underlying interface is rate-limited.
1502 		 * Those who need source quench packets may re-enable them
1503 		 * via the net.inet.ip.sendsourcequench sysctl.
1504 		 */
1505 		if (V_ip_sendsourcequench == 0) {
1506 			m_freem(mcopy);
1507 			return;
1508 		} else {
1509 			type = ICMP_SOURCEQUENCH;
1510 			code = 0;
1511 		}
1512 		break;
1513 
1514 	case EACCES:			/* ipfw denied packet */
1515 		m_freem(mcopy);
1516 		return;
1517 	}
1518 	icmp_error(mcopy, type, code, dest.s_addr, mtu);
1519 }
1520 
1521 void
1522 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1523     struct mbuf *m)
1524 {
1525 	INIT_VNET_NET(inp->inp_vnet);
1526 
1527 	if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) {
1528 		struct bintime bt;
1529 
1530 		bintime(&bt);
1531 		if (inp->inp_socket->so_options & SO_BINTIME) {
1532 			*mp = sbcreatecontrol((caddr_t) &bt, sizeof(bt),
1533 			SCM_BINTIME, SOL_SOCKET);
1534 			if (*mp)
1535 				mp = &(*mp)->m_next;
1536 		}
1537 		if (inp->inp_socket->so_options & SO_TIMESTAMP) {
1538 			struct timeval tv;
1539 
1540 			bintime2timeval(&bt, &tv);
1541 			*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1542 				SCM_TIMESTAMP, SOL_SOCKET);
1543 			if (*mp)
1544 				mp = &(*mp)->m_next;
1545 		}
1546 	}
1547 	if (inp->inp_flags & INP_RECVDSTADDR) {
1548 		*mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
1549 		    sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1550 		if (*mp)
1551 			mp = &(*mp)->m_next;
1552 	}
1553 	if (inp->inp_flags & INP_RECVTTL) {
1554 		*mp = sbcreatecontrol((caddr_t) &ip->ip_ttl,
1555 		    sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
1556 		if (*mp)
1557 			mp = &(*mp)->m_next;
1558 	}
1559 #ifdef notyet
1560 	/* XXX
1561 	 * Moving these out of udp_input() made them even more broken
1562 	 * than they already were.
1563 	 */
1564 	/* options were tossed already */
1565 	if (inp->inp_flags & INP_RECVOPTS) {
1566 		*mp = sbcreatecontrol((caddr_t) opts_deleted_above,
1567 		    sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1568 		if (*mp)
1569 			mp = &(*mp)->m_next;
1570 	}
1571 	/* ip_srcroute doesn't do what we want here, need to fix */
1572 	if (inp->inp_flags & INP_RECVRETOPTS) {
1573 		*mp = sbcreatecontrol((caddr_t) ip_srcroute(m),
1574 		    sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1575 		if (*mp)
1576 			mp = &(*mp)->m_next;
1577 	}
1578 #endif
1579 	if (inp->inp_flags & INP_RECVIF) {
1580 		struct ifnet *ifp;
1581 		struct sdlbuf {
1582 			struct sockaddr_dl sdl;
1583 			u_char	pad[32];
1584 		} sdlbuf;
1585 		struct sockaddr_dl *sdp;
1586 		struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1587 
1588 		if (((ifp = m->m_pkthdr.rcvif))
1589 		&& ( ifp->if_index && (ifp->if_index <= V_if_index))) {
1590 			sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1591 			/*
1592 			 * Change our mind and don't try copy.
1593 			 */
1594 			if ((sdp->sdl_family != AF_LINK)
1595 			|| (sdp->sdl_len > sizeof(sdlbuf))) {
1596 				goto makedummy;
1597 			}
1598 			bcopy(sdp, sdl2, sdp->sdl_len);
1599 		} else {
1600 makedummy:
1601 			sdl2->sdl_len
1602 				= offsetof(struct sockaddr_dl, sdl_data[0]);
1603 			sdl2->sdl_family = AF_LINK;
1604 			sdl2->sdl_index = 0;
1605 			sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1606 		}
1607 		*mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len,
1608 			IP_RECVIF, IPPROTO_IP);
1609 		if (*mp)
1610 			mp = &(*mp)->m_next;
1611 	}
1612 }
1613 
1614 /*
1615  * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
1616  * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
1617  * locking.  This code remains in ip_input.c as ip_mroute.c is optionally
1618  * compiled.
1619  */
1620 int
1621 ip_rsvp_init(struct socket *so)
1622 {
1623 	INIT_VNET_INET(so->so_vnet);
1624 
1625 	if (so->so_type != SOCK_RAW ||
1626 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
1627 		return EOPNOTSUPP;
1628 
1629 	if (V_ip_rsvpd != NULL)
1630 		return EADDRINUSE;
1631 
1632 	V_ip_rsvpd = so;
1633 	/*
1634 	 * This may seem silly, but we need to be sure we don't over-increment
1635 	 * the RSVP counter, in case something slips up.
1636 	 */
1637 	if (!V_ip_rsvp_on) {
1638 		V_ip_rsvp_on = 1;
1639 		V_rsvp_on++;
1640 	}
1641 
1642 	return 0;
1643 }
1644 
1645 int
1646 ip_rsvp_done(void)
1647 {
1648 	INIT_VNET_INET(curvnet);
1649 
1650 	V_ip_rsvpd = NULL;
1651 	/*
1652 	 * This may seem silly, but we need to be sure we don't over-decrement
1653 	 * the RSVP counter, in case something slips up.
1654 	 */
1655 	if (V_ip_rsvp_on) {
1656 		V_ip_rsvp_on = 0;
1657 		V_rsvp_on--;
1658 	}
1659 	return 0;
1660 }
1661 
1662 void
1663 rsvp_input(struct mbuf *m, int off)	/* XXX must fixup manually */
1664 {
1665 	INIT_VNET_INET(curvnet);
1666 
1667 	if (rsvp_input_p) { /* call the real one if loaded */
1668 		rsvp_input_p(m, off);
1669 		return;
1670 	}
1671 
1672 	/* Can still get packets with rsvp_on = 0 if there is a local member
1673 	 * of the group to which the RSVP packet is addressed.  But in this
1674 	 * case we want to throw the packet away.
1675 	 */
1676 
1677 	if (!V_rsvp_on) {
1678 		m_freem(m);
1679 		return;
1680 	}
1681 
1682 	if (V_ip_rsvpd != NULL) {
1683 		rip_input(m, off);
1684 		return;
1685 	}
1686 	/* Drop the packet */
1687 	m_freem(m);
1688 }
1689