xref: /freebsd/sys/netinet/ip_input.c (revision 9ba2b298dfda78b6669281507b314a4c441cd706)
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_route.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/lock.h>
53 #include <sys/rwlock.h>
54 #include <sys/syslog.h>
55 #include <sys/sysctl.h>
56 #include <sys/vimage.h>
57 
58 #include <net/pfil.h>
59 #include <net/if.h>
60 #include <net/if_types.h>
61 #include <net/if_var.h>
62 #include <net/if_dl.h>
63 #include <net/route.h>
64 #include <net/netisr.h>
65 #include <net/vnet.h>
66 #include <net/flowtable.h>
67 
68 #include <netinet/in.h>
69 #include <netinet/in_systm.h>
70 #include <netinet/in_var.h>
71 #include <netinet/ip.h>
72 #include <netinet/in_pcb.h>
73 #include <netinet/ip_var.h>
74 #include <netinet/ip_icmp.h>
75 #include <netinet/ip_options.h>
76 #include <machine/in_cksum.h>
77 #include <netinet/vinet.h>
78 #ifdef DEV_CARP
79 #include <netinet/ip_carp.h>
80 #endif
81 #ifdef IPSEC
82 #include <netinet/ip_ipsec.h>
83 #endif /* IPSEC */
84 
85 #include <sys/socketvar.h>
86 
87 #include <security/mac/mac_framework.h>
88 
89 #ifdef CTASSERT
90 CTASSERT(sizeof(struct ip) == 20);
91 #endif
92 
93 #ifndef VIMAGE
94 #ifndef VIMAGE_GLOBALS
95 struct vnet_inet vnet_inet_0;
96 #endif
97 #endif
98 
99 #ifdef VIMAGE_GLOBALS
100 static int	ipsendredirects;
101 static int	ip_checkinterface;
102 static int	ip_keepfaith;
103 static int	ip_sendsourcequench;
104 int	ip_defttl;
105 int	ip_do_randomid;
106 int	ipforwarding;
107 struct	in_ifaddrhead in_ifaddrhead; 		/* first inet address */
108 struct	in_ifaddrhashhead *in_ifaddrhashtbl;	/* inet addr hash table  */
109 u_long 	in_ifaddrhmask;				/* mask for hash table */
110 struct ipstat ipstat;
111 static int ip_rsvp_on;
112 struct socket *ip_rsvpd;
113 int	rsvp_on;
114 static struct ipqhead ipq[IPREASS_NHASH];
115 static int	maxnipq;	/* Administrative limit on # reass queues. */
116 static int	maxfragsperpacket;
117 int	ipstealth;
118 static int	nipq;	/* Total # of reass queues */
119 #endif
120 
121 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, IPCTL_FORWARDING,
122     forwarding, CTLFLAG_RW, ipforwarding, 0,
123     "Enable IP forwarding between interfaces");
124 
125 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, IPCTL_SENDREDIRECTS,
126     redirect, CTLFLAG_RW, ipsendredirects, 0,
127     "Enable sending IP redirects");
128 
129 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, IPCTL_DEFTTL,
130     ttl, CTLFLAG_RW, ip_defttl, 0, "Maximum TTL on IP packets");
131 
132 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, IPCTL_KEEPFAITH,
133     keepfaith, CTLFLAG_RW, ip_keepfaith,	0,
134     "Enable packet capture for FAITH IPv4->IPv6 translater daemon");
135 
136 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO,
137     sendsourcequench, CTLFLAG_RW, ip_sendsourcequench, 0,
138     "Enable the transmission of source quench packets");
139 
140 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, random_id,
141     CTLFLAG_RW, ip_do_randomid, 0, "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  */
156 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO,
157     check_interface, CTLFLAG_RW, ip_checkinterface, 0,
158     "Verify packet arrives on correct interface");
159 
160 struct pfil_head inet_pfil_hook;	/* Packet filter hooks */
161 
162 static struct netisr_handler ip_nh = {
163 	.nh_name = "ip",
164 	.nh_handler = ip_input,
165 	.nh_proto = NETISR_IP,
166 	.nh_policy = NETISR_POLICY_FLOW,
167 };
168 
169 extern	struct domain inetdomain;
170 extern	struct protosw inetsw[];
171 u_char	ip_protox[IPPROTO_MAX];
172 
173 
174 SYSCTL_V_STRUCT(V_NET, vnet_inet, _net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW,
175     ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)");
176 
177 #ifdef VIMAGE_GLOBALS
178 static uma_zone_t ipq_zone;
179 #endif
180 static struct mtx ipqlock;
181 
182 #define	IPQ_LOCK()	mtx_lock(&ipqlock)
183 #define	IPQ_UNLOCK()	mtx_unlock(&ipqlock)
184 #define	IPQ_LOCK_INIT()	mtx_init(&ipqlock, "ipqlock", NULL, MTX_DEF)
185 #define	IPQ_LOCK_ASSERT()	mtx_assert(&ipqlock, MA_OWNED)
186 
187 static void	maxnipq_update(void);
188 static void	ipq_zone_change(void *);
189 
190 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, fragpackets,
191     CTLFLAG_RD, nipq, 0,
192     "Current number of IPv4 fragment reassembly queue entries");
193 
194 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, maxfragsperpacket,
195     CTLFLAG_RW, maxfragsperpacket, 0,
196     "Maximum number of IPv4 fragments allowed per packet");
197 
198 struct callout	ipport_tick_callout;
199 
200 #ifdef IPCTL_DEFMTU
201 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
202     &ip_mtu, 0, "Default MTU");
203 #endif
204 
205 #ifdef IPSTEALTH
206 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW,
207     ipstealth, 0, "IP stealth mode, no TTL decrementation on forwarding");
208 #endif
209 #ifdef FLOWTABLE
210 #ifdef VIMAGE_GLOBALS
211 static int ip_output_flowtable_size;
212 struct flowtable *ip_ft;
213 #endif
214 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_ip, OID_AUTO, output_flowtable_size,
215     CTLFLAG_RDTUN, ip_output_flowtable_size, 2048,
216     "number of entries in the per-cpu output flow caches");
217 #endif
218 
219 #ifdef VIMAGE_GLOBALS
220 int fw_one_pass;
221 #endif
222 
223 static void	ip_freef(struct ipqhead *, struct ipq *);
224 
225 #ifndef VIMAGE_GLOBALS
226 static void vnet_inet_register(void);
227 
228 static const vnet_modinfo_t vnet_inet_modinfo = {
229 	.vmi_id		= VNET_MOD_INET,
230 	.vmi_name	= "inet",
231 	.vmi_size	= sizeof(struct vnet_inet)
232 };
233 
234 static void vnet_inet_register()
235 {
236 
237 	vnet_mod_register(&vnet_inet_modinfo);
238 }
239 
240 SYSINIT(inet, SI_SUB_PROTO_BEGIN, SI_ORDER_FIRST, vnet_inet_register, 0);
241 #endif
242 
243 static int
244 sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS)
245 {
246 	int error, qlimit;
247 
248 	netisr_getqlimit(&ip_nh, &qlimit);
249 	error = sysctl_handle_int(oidp, &qlimit, 0, req);
250 	if (error || !req->newptr)
251 		return (error);
252 	if (qlimit < 1)
253 		return (EINVAL);
254 	return (netisr_setqlimit(&ip_nh, qlimit));
255 }
256 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen,
257     CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet_intr_queue_maxlen, "I",
258     "Maximum size of the IP input queue");
259 
260 static int
261 sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS)
262 {
263 	u_int64_t qdrops_long;
264 	int error, qdrops;
265 
266 	netisr_getqdrops(&ip_nh, &qdrops_long);
267 	qdrops = qdrops_long;
268 	error = sysctl_handle_int(oidp, &qdrops, 0, req);
269 	if (error || !req->newptr)
270 		return (error);
271 	if (qdrops != 0)
272 		return (EINVAL);
273 	netisr_clearqdrops(&ip_nh);
274 	return (0);
275 }
276 
277 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops,
278     CTLTYPE_INT|CTLFLAG_RD, 0, 0, sysctl_netinet_intr_queue_drops, "I",
279     "Number of packets dropped from the IP input queue");
280 
281 /*
282  * IP initialization: fill in IP protocol switch table.
283  * All protocols not implemented in kernel go to raw IP protocol handler.
284  */
285 void
286 ip_init(void)
287 {
288 	INIT_VNET_INET(curvnet);
289 	struct protosw *pr;
290 	int i;
291 
292 	V_ipsendredirects = 1; /* XXX */
293 	V_ip_checkinterface = 0;
294 	V_ip_keepfaith = 0;
295 	V_ip_sendsourcequench = 0;
296 	V_rsvp_on = 0;
297 	V_ip_defttl = IPDEFTTL;
298 	V_ip_do_randomid = 0;
299 	V_ip_id = time_second & 0xffff;
300 	V_ipforwarding = 0;
301 	V_ipstealth = 0;
302 	V_nipq = 0;	/* Total # of reass queues */
303 
304 	V_ipport_lowfirstauto = IPPORT_RESERVED - 1;	/* 1023 */
305 	V_ipport_lowlastauto = IPPORT_RESERVEDSTART;	/* 600 */
306 	V_ipport_firstauto = IPPORT_EPHEMERALFIRST;	/* 10000 */
307 	V_ipport_lastauto = IPPORT_EPHEMERALLAST;	/* 65535 */
308 	V_ipport_hifirstauto = IPPORT_HIFIRSTAUTO;	/* 49152 */
309 	V_ipport_hilastauto = IPPORT_HILASTAUTO;	/* 65535 */
310 	V_ipport_reservedhigh = IPPORT_RESERVED - 1;	/* 1023 */
311 	V_ipport_reservedlow = 0;
312 	V_ipport_randomized = 1;	/* user controlled via sysctl */
313 	V_ipport_randomcps = 10;	/* user controlled via sysctl */
314 	V_ipport_randomtime = 45;	/* user controlled via sysctl */
315 	V_ipport_stoprandom = 0;	/* toggled by ipport_tick */
316 
317 	V_fw_one_pass = 1;
318 
319 #ifdef NOTYET
320 	/* XXX global static but not instantiated in this file */
321 	V_ipfastforward_active = 0;
322 	V_subnetsarelocal = 0;
323 	V_sameprefixcarponly = 0;
324 #endif
325 
326 	TAILQ_INIT(&V_in_ifaddrhead);
327 	V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask);
328 
329 	/* Initialize IP reassembly queue. */
330 	for (i = 0; i < IPREASS_NHASH; i++)
331 		TAILQ_INIT(&V_ipq[i]);
332 	V_maxnipq = nmbclusters / 32;
333 	V_maxfragsperpacket = 16;
334 	V_ipq_zone = uma_zcreate("ipq", sizeof(struct ipq), NULL, NULL, NULL,
335 	    NULL, UMA_ALIGN_PTR, 0);
336 	maxnipq_update();
337 
338 #ifdef FLOWTABLE
339 	V_ip_output_flowtable_size = 2048;
340 	TUNABLE_INT_FETCH("net.inet.ip.output_flowtable_size",
341 	    &V_ip_output_flowtable_size);
342 	V_ip_ft = flowtable_alloc(V_ip_output_flowtable_size, FL_PCPU);
343 #endif
344 
345 	/* Skip initialization of globals for non-default instances. */
346 	if (!IS_DEFAULT_VNET(curvnet))
347 		return;
348 
349 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
350 	if (pr == NULL)
351 		panic("ip_init: PF_INET not found");
352 
353 	/* Initialize the entire ip_protox[] array to IPPROTO_RAW. */
354 	for (i = 0; i < IPPROTO_MAX; i++)
355 		ip_protox[i] = pr - inetsw;
356 	/*
357 	 * Cycle through IP protocols and put them into the appropriate place
358 	 * in ip_protox[].
359 	 */
360 	for (pr = inetdomain.dom_protosw;
361 	    pr < inetdomain.dom_protoswNPROTOSW; pr++)
362 		if (pr->pr_domain->dom_family == PF_INET &&
363 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
364 			/* Be careful to only index valid IP protocols. */
365 			if (pr->pr_protocol < IPPROTO_MAX)
366 				ip_protox[pr->pr_protocol] = pr - inetsw;
367 		}
368 
369 	/* Initialize packet filter hooks. */
370 	inet_pfil_hook.ph_type = PFIL_TYPE_AF;
371 	inet_pfil_hook.ph_af = AF_INET;
372 	if ((i = pfil_head_register(&inet_pfil_hook)) != 0)
373 		printf("%s: WARNING: unable to register pfil hook, "
374 			"error %d\n", __func__, i);
375 
376 	/* Start ipport_tick. */
377 	callout_init(&ipport_tick_callout, CALLOUT_MPSAFE);
378 	callout_reset(&ipport_tick_callout, 1, ipport_tick, NULL);
379 	EVENTHANDLER_REGISTER(shutdown_pre_sync, ip_fini, NULL,
380 		SHUTDOWN_PRI_DEFAULT);
381 	EVENTHANDLER_REGISTER(nmbclusters_change, ipq_zone_change,
382 		NULL, EVENTHANDLER_PRI_ANY);
383 
384 	/* Initialize various other remaining things. */
385 	IPQ_LOCK_INIT();
386 	netisr_register(&ip_nh);
387 }
388 
389 void
390 ip_fini(void *xtp)
391 {
392 
393 	callout_stop(&ipport_tick_callout);
394 }
395 
396 /*
397  * Ip input routine.  Checksum and byte swap header.  If fragmented
398  * try to reassemble.  Process options.  Pass to next level.
399  */
400 void
401 ip_input(struct mbuf *m)
402 {
403 	INIT_VNET_INET(curvnet);
404 	struct ip *ip = NULL;
405 	struct in_ifaddr *ia = NULL;
406 	struct ifaddr *ifa;
407 	struct ifnet *ifp;
408 	int    checkif, hlen = 0;
409 	u_short sum;
410 	int dchg = 0;				/* dest changed after fw */
411 	struct in_addr odst;			/* original dst address */
412 
413 	M_ASSERTPKTHDR(m);
414 
415 	if (m->m_flags & M_FASTFWD_OURS) {
416 		/*
417 		 * Firewall or NAT changed destination to local.
418 		 * We expect ip_len and ip_off to be in host byte order.
419 		 */
420 		m->m_flags &= ~M_FASTFWD_OURS;
421 		/* Set up some basics that will be used later. */
422 		ip = mtod(m, struct ip *);
423 		hlen = ip->ip_hl << 2;
424 		goto ours;
425 	}
426 
427 	IPSTAT_INC(ips_total);
428 
429 	if (m->m_pkthdr.len < sizeof(struct ip))
430 		goto tooshort;
431 
432 	if (m->m_len < sizeof (struct ip) &&
433 	    (m = m_pullup(m, sizeof (struct ip))) == NULL) {
434 		IPSTAT_INC(ips_toosmall);
435 		return;
436 	}
437 	ip = mtod(m, struct ip *);
438 
439 	if (ip->ip_v != IPVERSION) {
440 		IPSTAT_INC(ips_badvers);
441 		goto bad;
442 	}
443 
444 	hlen = ip->ip_hl << 2;
445 	if (hlen < sizeof(struct ip)) {	/* minimum header length */
446 		IPSTAT_INC(ips_badhlen);
447 		goto bad;
448 	}
449 	if (hlen > m->m_len) {
450 		if ((m = m_pullup(m, hlen)) == NULL) {
451 			IPSTAT_INC(ips_badhlen);
452 			return;
453 		}
454 		ip = mtod(m, struct ip *);
455 	}
456 
457 	/* 127/8 must not appear on wire - RFC1122 */
458 	ifp = m->m_pkthdr.rcvif;
459 	if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
460 	    (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
461 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
462 			IPSTAT_INC(ips_badaddr);
463 			goto bad;
464 		}
465 	}
466 
467 	if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
468 		sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
469 	} else {
470 		if (hlen == sizeof(struct ip)) {
471 			sum = in_cksum_hdr(ip);
472 		} else {
473 			sum = in_cksum(m, hlen);
474 		}
475 	}
476 	if (sum) {
477 		IPSTAT_INC(ips_badsum);
478 		goto bad;
479 	}
480 
481 #ifdef ALTQ
482 	if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
483 		/* packet is dropped by traffic conditioner */
484 		return;
485 #endif
486 
487 	/*
488 	 * Convert fields to host representation.
489 	 */
490 	ip->ip_len = ntohs(ip->ip_len);
491 	if (ip->ip_len < hlen) {
492 		IPSTAT_INC(ips_badlen);
493 		goto bad;
494 	}
495 	ip->ip_off = ntohs(ip->ip_off);
496 
497 	/*
498 	 * Check that the amount of data in the buffers
499 	 * is as at least much as the IP header would have us expect.
500 	 * Trim mbufs if longer than we expect.
501 	 * Drop packet if shorter than we expect.
502 	 */
503 	if (m->m_pkthdr.len < ip->ip_len) {
504 tooshort:
505 		IPSTAT_INC(ips_tooshort);
506 		goto bad;
507 	}
508 	if (m->m_pkthdr.len > ip->ip_len) {
509 		if (m->m_len == m->m_pkthdr.len) {
510 			m->m_len = ip->ip_len;
511 			m->m_pkthdr.len = ip->ip_len;
512 		} else
513 			m_adj(m, ip->ip_len - m->m_pkthdr.len);
514 	}
515 #ifdef IPSEC
516 	/*
517 	 * Bypass packet filtering for packets from a tunnel (gif).
518 	 */
519 	if (ip_ipsec_filtertunnel(m))
520 		goto passin;
521 #endif /* IPSEC */
522 
523 	/*
524 	 * Run through list of hooks for input packets.
525 	 *
526 	 * NB: Beware of the destination address changing (e.g.
527 	 *     by NAT rewriting).  When this happens, tell
528 	 *     ip_forward to do the right thing.
529 	 */
530 
531 	/* Jump over all PFIL processing if hooks are not active. */
532 	if (!PFIL_HOOKED(&inet_pfil_hook))
533 		goto passin;
534 
535 	odst = ip->ip_dst;
536 	if (pfil_run_hooks(&inet_pfil_hook, &m, ifp, PFIL_IN, NULL) != 0)
537 		return;
538 	if (m == NULL)			/* consumed by filter */
539 		return;
540 
541 	ip = mtod(m, struct ip *);
542 	dchg = (odst.s_addr != ip->ip_dst.s_addr);
543 	ifp = m->m_pkthdr.rcvif;
544 
545 #ifdef IPFIREWALL_FORWARD
546 	if (m->m_flags & M_FASTFWD_OURS) {
547 		m->m_flags &= ~M_FASTFWD_OURS;
548 		goto ours;
549 	}
550 	if ((dchg = (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL)) != 0) {
551 		/*
552 		 * Directly ship on the packet.  This allows to forward packets
553 		 * that were destined for us to some other directly connected
554 		 * host.
555 		 */
556 		ip_forward(m, dchg);
557 		return;
558 	}
559 #endif /* IPFIREWALL_FORWARD */
560 
561 passin:
562 	/*
563 	 * Process options and, if not destined for us,
564 	 * ship it on.  ip_dooptions returns 1 when an
565 	 * error was detected (causing an icmp message
566 	 * to be sent and the original packet to be freed).
567 	 */
568 	if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
569 		return;
570 
571         /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
572          * matter if it is destined to another node, or whether it is
573          * a multicast one, RSVP wants it! and prevents it from being forwarded
574          * anywhere else. Also checks if the rsvp daemon is running before
575 	 * grabbing the packet.
576          */
577 	if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP)
578 		goto ours;
579 
580 	/*
581 	 * Check our list of addresses, to see if the packet is for us.
582 	 * If we don't have any addresses, assume any unicast packet
583 	 * we receive might be for us (and let the upper layers deal
584 	 * with it).
585 	 */
586 	if (TAILQ_EMPTY(&V_in_ifaddrhead) &&
587 	    (m->m_flags & (M_MCAST|M_BCAST)) == 0)
588 		goto ours;
589 
590 	/*
591 	 * Enable a consistency check between the destination address
592 	 * and the arrival interface for a unicast packet (the RFC 1122
593 	 * strong ES model) if IP forwarding is disabled and the packet
594 	 * is not locally generated and the packet is not subject to
595 	 * 'ipfw fwd'.
596 	 *
597 	 * XXX - Checking also should be disabled if the destination
598 	 * address is ipnat'ed to a different interface.
599 	 *
600 	 * XXX - Checking is incompatible with IP aliases added
601 	 * to the loopback interface instead of the interface where
602 	 * the packets are received.
603 	 *
604 	 * XXX - This is the case for carp vhost IPs as well so we
605 	 * insert a workaround. If the packet got here, we already
606 	 * checked with carp_iamatch() and carp_forus().
607 	 */
608 	checkif = V_ip_checkinterface && (V_ipforwarding == 0) &&
609 	    ifp != NULL && ((ifp->if_flags & IFF_LOOPBACK) == 0) &&
610 #ifdef DEV_CARP
611 	    !ifp->if_carp &&
612 #endif
613 	    (dchg == 0);
614 
615 	/*
616 	 * Check for exact addresses in the hash bucket.
617 	 */
618 	LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
619 		/*
620 		 * If the address matches, verify that the packet
621 		 * arrived via the correct interface if checking is
622 		 * enabled.
623 		 */
624 		if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr &&
625 		    (!checkif || ia->ia_ifp == ifp)) {
626 			ifa_ref(&ia->ia_ifa);
627 			goto ours;
628 		}
629 	}
630 	/*
631 	 * Check for broadcast addresses.
632 	 *
633 	 * Only accept broadcast packets that arrive via the matching
634 	 * interface.  Reception of forwarded directed broadcasts would
635 	 * be handled via ip_forward() and ether_output() with the loopback
636 	 * into the stack for SIMPLEX interfaces handled by ether_output().
637 	 */
638 	if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
639 		IF_ADDR_LOCK(ifp);
640 	        TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
641 			if (ifa->ifa_addr->sa_family != AF_INET)
642 				continue;
643 			ia = ifatoia(ifa);
644 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
645 			    ip->ip_dst.s_addr) {
646 				ifa_ref(ifa);
647 				IF_ADDR_UNLOCK(ifp);
648 				goto ours;
649 			}
650 			if (ia->ia_netbroadcast.s_addr == ip->ip_dst.s_addr) {
651 				ifa_ref(ifa);
652 				IF_ADDR_UNLOCK(ifp);
653 				goto ours;
654 			}
655 #ifdef BOOTP_COMPAT
656 			if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) {
657 				ifa_ref(ifa);
658 				IF_ADDR_UNLOCK(ifp);
659 				goto ours;
660 			}
661 #endif
662 		}
663 		IF_ADDR_UNLOCK(ifp);
664 	}
665 	/* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */
666 	if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) {
667 		IPSTAT_INC(ips_cantforward);
668 		m_freem(m);
669 		return;
670 	}
671 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
672 		if (V_ip_mrouter) {
673 			/*
674 			 * If we are acting as a multicast router, all
675 			 * incoming multicast packets are passed to the
676 			 * kernel-level multicast forwarding function.
677 			 * The packet is returned (relatively) intact; if
678 			 * ip_mforward() returns a non-zero value, the packet
679 			 * must be discarded, else it may be accepted below.
680 			 */
681 			if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) {
682 				IPSTAT_INC(ips_cantforward);
683 				m_freem(m);
684 				return;
685 			}
686 
687 			/*
688 			 * The process-level routing daemon needs to receive
689 			 * all multicast IGMP packets, whether or not this
690 			 * host belongs to their destination groups.
691 			 */
692 			if (ip->ip_p == IPPROTO_IGMP)
693 				goto ours;
694 			IPSTAT_INC(ips_forward);
695 		}
696 		/*
697 		 * Assume the packet is for us, to avoid prematurely taking
698 		 * a lock on the in_multi hash. Protocols must perform
699 		 * their own filtering and update statistics accordingly.
700 		 */
701 		goto ours;
702 	}
703 	if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
704 		goto ours;
705 	if (ip->ip_dst.s_addr == INADDR_ANY)
706 		goto ours;
707 
708 	/*
709 	 * FAITH(Firewall Aided Internet Translator)
710 	 */
711 	if (ifp && ifp->if_type == IFT_FAITH) {
712 		if (V_ip_keepfaith) {
713 			if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP)
714 				goto ours;
715 		}
716 		m_freem(m);
717 		return;
718 	}
719 
720 	/*
721 	 * Not for us; forward if possible and desirable.
722 	 */
723 	if (V_ipforwarding == 0) {
724 		IPSTAT_INC(ips_cantforward);
725 		m_freem(m);
726 	} else {
727 #ifdef IPSEC
728 		if (ip_ipsec_fwd(m))
729 			goto bad;
730 #endif /* IPSEC */
731 		ip_forward(m, dchg);
732 	}
733 	return;
734 
735 ours:
736 #ifdef IPSTEALTH
737 	/*
738 	 * IPSTEALTH: Process non-routing options only
739 	 * if the packet is destined for us.
740 	 */
741 	if (V_ipstealth && hlen > sizeof (struct ip) &&
742 	    ip_dooptions(m, 1))
743 		return;
744 #endif /* IPSTEALTH */
745 
746 	/* Count the packet in the ip address stats */
747 	if (ia != NULL) {
748 		ia->ia_ifa.if_ipackets++;
749 		ia->ia_ifa.if_ibytes += m->m_pkthdr.len;
750 		ifa_free(&ia->ia_ifa);
751 	}
752 
753 	/*
754 	 * Attempt reassembly; if it succeeds, proceed.
755 	 * ip_reass() will return a different mbuf.
756 	 */
757 	if (ip->ip_off & (IP_MF | IP_OFFMASK)) {
758 		m = ip_reass(m);
759 		if (m == NULL)
760 			return;
761 		ip = mtod(m, struct ip *);
762 		/* Get the header length of the reassembled packet */
763 		hlen = ip->ip_hl << 2;
764 	}
765 
766 	/*
767 	 * Further protocols expect the packet length to be w/o the
768 	 * IP header.
769 	 */
770 	ip->ip_len -= hlen;
771 
772 #ifdef IPSEC
773 	/*
774 	 * enforce IPsec policy checking if we are seeing last header.
775 	 * note that we do not visit this with protocols with pcb layer
776 	 * code - like udp/tcp/raw ip.
777 	 */
778 	if (ip_ipsec_input(m))
779 		goto bad;
780 #endif /* IPSEC */
781 
782 	/*
783 	 * Switch out to protocol's input routine.
784 	 */
785 	IPSTAT_INC(ips_delivered);
786 
787 	(*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen);
788 	return;
789 bad:
790 	m_freem(m);
791 }
792 
793 /*
794  * After maxnipq has been updated, propagate the change to UMA.  The UMA zone
795  * max has slightly different semantics than the sysctl, for historical
796  * reasons.
797  */
798 static void
799 maxnipq_update(void)
800 {
801 	INIT_VNET_INET(curvnet);
802 
803 	/*
804 	 * -1 for unlimited allocation.
805 	 */
806 	if (V_maxnipq < 0)
807 		uma_zone_set_max(V_ipq_zone, 0);
808 	/*
809 	 * Positive number for specific bound.
810 	 */
811 	if (V_maxnipq > 0)
812 		uma_zone_set_max(V_ipq_zone, V_maxnipq);
813 	/*
814 	 * Zero specifies no further fragment queue allocation -- set the
815 	 * bound very low, but rely on implementation elsewhere to actually
816 	 * prevent allocation and reclaim current queues.
817 	 */
818 	if (V_maxnipq == 0)
819 		uma_zone_set_max(V_ipq_zone, 1);
820 }
821 
822 static void
823 ipq_zone_change(void *tag)
824 {
825 	INIT_VNET_INET(curvnet);
826 
827 	if (V_maxnipq > 0 && V_maxnipq < (nmbclusters / 32)) {
828 		V_maxnipq = nmbclusters / 32;
829 		maxnipq_update();
830 	}
831 }
832 
833 static int
834 sysctl_maxnipq(SYSCTL_HANDLER_ARGS)
835 {
836 	INIT_VNET_INET(curvnet);
837 	int error, i;
838 
839 	i = V_maxnipq;
840 	error = sysctl_handle_int(oidp, &i, 0, req);
841 	if (error || !req->newptr)
842 		return (error);
843 
844 	/*
845 	 * XXXRW: Might be a good idea to sanity check the argument and place
846 	 * an extreme upper bound.
847 	 */
848 	if (i < -1)
849 		return (EINVAL);
850 	V_maxnipq = i;
851 	maxnipq_update();
852 	return (0);
853 }
854 
855 SYSCTL_PROC(_net_inet_ip, OID_AUTO, maxfragpackets, CTLTYPE_INT|CTLFLAG_RW,
856     NULL, 0, sysctl_maxnipq, "I",
857     "Maximum number of IPv4 fragment reassembly queue entries");
858 
859 /*
860  * Take incoming datagram fragment and try to reassemble it into
861  * whole datagram.  If the argument is the first fragment or one
862  * in between the function will return NULL and store the mbuf
863  * in the fragment chain.  If the argument is the last fragment
864  * the packet will be reassembled and the pointer to the new
865  * mbuf returned for further processing.  Only m_tags attached
866  * to the first packet/fragment are preserved.
867  * The IP header is *NOT* adjusted out of iplen.
868  */
869 struct mbuf *
870 ip_reass(struct mbuf *m)
871 {
872 	INIT_VNET_INET(curvnet);
873 	struct ip *ip;
874 	struct mbuf *p, *q, *nq, *t;
875 	struct ipq *fp = NULL;
876 	struct ipqhead *head;
877 	int i, hlen, next;
878 	u_int8_t ecn, ecn0;
879 	u_short hash;
880 
881 	/* If maxnipq or maxfragsperpacket are 0, never accept fragments. */
882 	if (V_maxnipq == 0 || V_maxfragsperpacket == 0) {
883 		IPSTAT_INC(ips_fragments);
884 		IPSTAT_INC(ips_fragdropped);
885 		m_freem(m);
886 		return (NULL);
887 	}
888 
889 	ip = mtod(m, struct ip *);
890 	hlen = ip->ip_hl << 2;
891 
892 	hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
893 	head = &V_ipq[hash];
894 	IPQ_LOCK();
895 
896 	/*
897 	 * Look for queue of fragments
898 	 * of this datagram.
899 	 */
900 	TAILQ_FOREACH(fp, head, ipq_list)
901 		if (ip->ip_id == fp->ipq_id &&
902 		    ip->ip_src.s_addr == fp->ipq_src.s_addr &&
903 		    ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
904 #ifdef MAC
905 		    mac_ipq_match(m, fp) &&
906 #endif
907 		    ip->ip_p == fp->ipq_p)
908 			goto found;
909 
910 	fp = NULL;
911 
912 	/*
913 	 * Attempt to trim the number of allocated fragment queues if it
914 	 * exceeds the administrative limit.
915 	 */
916 	if ((V_nipq > V_maxnipq) && (V_maxnipq > 0)) {
917 		/*
918 		 * drop something from the tail of the current queue
919 		 * before proceeding further
920 		 */
921 		struct ipq *q = TAILQ_LAST(head, ipqhead);
922 		if (q == NULL) {   /* gak */
923 			for (i = 0; i < IPREASS_NHASH; i++) {
924 				struct ipq *r = TAILQ_LAST(&V_ipq[i], ipqhead);
925 				if (r) {
926 					IPSTAT_ADD(ips_fragtimeout,
927 					    r->ipq_nfrags);
928 					ip_freef(&V_ipq[i], r);
929 					break;
930 				}
931 			}
932 		} else {
933 			IPSTAT_ADD(ips_fragtimeout, q->ipq_nfrags);
934 			ip_freef(head, q);
935 		}
936 	}
937 
938 found:
939 	/*
940 	 * Adjust ip_len to not reflect header,
941 	 * convert offset of this to bytes.
942 	 */
943 	ip->ip_len -= hlen;
944 	if (ip->ip_off & IP_MF) {
945 		/*
946 		 * Make sure that fragments have a data length
947 		 * that's a non-zero multiple of 8 bytes.
948 		 */
949 		if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
950 			IPSTAT_INC(ips_toosmall); /* XXX */
951 			goto dropfrag;
952 		}
953 		m->m_flags |= M_FRAG;
954 	} else
955 		m->m_flags &= ~M_FRAG;
956 	ip->ip_off <<= 3;
957 
958 
959 	/*
960 	 * Attempt reassembly; if it succeeds, proceed.
961 	 * ip_reass() will return a different mbuf.
962 	 */
963 	IPSTAT_INC(ips_fragments);
964 	m->m_pkthdr.header = ip;
965 
966 	/* Previous ip_reass() started here. */
967 	/*
968 	 * Presence of header sizes in mbufs
969 	 * would confuse code below.
970 	 */
971 	m->m_data += hlen;
972 	m->m_len -= hlen;
973 
974 	/*
975 	 * If first fragment to arrive, create a reassembly queue.
976 	 */
977 	if (fp == NULL) {
978 		fp = uma_zalloc(V_ipq_zone, M_NOWAIT);
979 		if (fp == NULL)
980 			goto dropfrag;
981 #ifdef MAC
982 		if (mac_ipq_init(fp, M_NOWAIT) != 0) {
983 			uma_zfree(V_ipq_zone, fp);
984 			fp = NULL;
985 			goto dropfrag;
986 		}
987 		mac_ipq_create(m, fp);
988 #endif
989 		TAILQ_INSERT_HEAD(head, fp, ipq_list);
990 		V_nipq++;
991 		fp->ipq_nfrags = 1;
992 		fp->ipq_ttl = IPFRAGTTL;
993 		fp->ipq_p = ip->ip_p;
994 		fp->ipq_id = ip->ip_id;
995 		fp->ipq_src = ip->ip_src;
996 		fp->ipq_dst = ip->ip_dst;
997 		fp->ipq_frags = m;
998 		m->m_nextpkt = NULL;
999 		goto done;
1000 	} else {
1001 		fp->ipq_nfrags++;
1002 #ifdef MAC
1003 		mac_ipq_update(m, fp);
1004 #endif
1005 	}
1006 
1007 #define GETIP(m)	((struct ip*)((m)->m_pkthdr.header))
1008 
1009 	/*
1010 	 * Handle ECN by comparing this segment with the first one;
1011 	 * if CE is set, do not lose CE.
1012 	 * drop if CE and not-ECT are mixed for the same packet.
1013 	 */
1014 	ecn = ip->ip_tos & IPTOS_ECN_MASK;
1015 	ecn0 = GETIP(fp->ipq_frags)->ip_tos & IPTOS_ECN_MASK;
1016 	if (ecn == IPTOS_ECN_CE) {
1017 		if (ecn0 == IPTOS_ECN_NOTECT)
1018 			goto dropfrag;
1019 		if (ecn0 != IPTOS_ECN_CE)
1020 			GETIP(fp->ipq_frags)->ip_tos |= IPTOS_ECN_CE;
1021 	}
1022 	if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT)
1023 		goto dropfrag;
1024 
1025 	/*
1026 	 * Find a segment which begins after this one does.
1027 	 */
1028 	for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt)
1029 		if (GETIP(q)->ip_off > ip->ip_off)
1030 			break;
1031 
1032 	/*
1033 	 * If there is a preceding segment, it may provide some of
1034 	 * our data already.  If so, drop the data from the incoming
1035 	 * segment.  If it provides all of our data, drop us, otherwise
1036 	 * stick new segment in the proper place.
1037 	 *
1038 	 * If some of the data is dropped from the the preceding
1039 	 * segment, then it's checksum is invalidated.
1040 	 */
1041 	if (p) {
1042 		i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off;
1043 		if (i > 0) {
1044 			if (i >= ip->ip_len)
1045 				goto dropfrag;
1046 			m_adj(m, i);
1047 			m->m_pkthdr.csum_flags = 0;
1048 			ip->ip_off += i;
1049 			ip->ip_len -= i;
1050 		}
1051 		m->m_nextpkt = p->m_nextpkt;
1052 		p->m_nextpkt = m;
1053 	} else {
1054 		m->m_nextpkt = fp->ipq_frags;
1055 		fp->ipq_frags = m;
1056 	}
1057 
1058 	/*
1059 	 * While we overlap succeeding segments trim them or,
1060 	 * if they are completely covered, dequeue them.
1061 	 */
1062 	for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off;
1063 	     q = nq) {
1064 		i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off;
1065 		if (i < GETIP(q)->ip_len) {
1066 			GETIP(q)->ip_len -= i;
1067 			GETIP(q)->ip_off += i;
1068 			m_adj(q, i);
1069 			q->m_pkthdr.csum_flags = 0;
1070 			break;
1071 		}
1072 		nq = q->m_nextpkt;
1073 		m->m_nextpkt = nq;
1074 		IPSTAT_INC(ips_fragdropped);
1075 		fp->ipq_nfrags--;
1076 		m_freem(q);
1077 	}
1078 
1079 	/*
1080 	 * Check for complete reassembly and perform frag per packet
1081 	 * limiting.
1082 	 *
1083 	 * Frag limiting is performed here so that the nth frag has
1084 	 * a chance to complete the packet before we drop the packet.
1085 	 * As a result, n+1 frags are actually allowed per packet, but
1086 	 * only n will ever be stored. (n = maxfragsperpacket.)
1087 	 *
1088 	 */
1089 	next = 0;
1090 	for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
1091 		if (GETIP(q)->ip_off != next) {
1092 			if (fp->ipq_nfrags > V_maxfragsperpacket) {
1093 				IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags);
1094 				ip_freef(head, fp);
1095 			}
1096 			goto done;
1097 		}
1098 		next += GETIP(q)->ip_len;
1099 	}
1100 	/* Make sure the last packet didn't have the IP_MF flag */
1101 	if (p->m_flags & M_FRAG) {
1102 		if (fp->ipq_nfrags > V_maxfragsperpacket) {
1103 			IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags);
1104 			ip_freef(head, fp);
1105 		}
1106 		goto done;
1107 	}
1108 
1109 	/*
1110 	 * Reassembly is complete.  Make sure the packet is a sane size.
1111 	 */
1112 	q = fp->ipq_frags;
1113 	ip = GETIP(q);
1114 	if (next + (ip->ip_hl << 2) > IP_MAXPACKET) {
1115 		IPSTAT_INC(ips_toolong);
1116 		IPSTAT_ADD(ips_fragdropped, fp->ipq_nfrags);
1117 		ip_freef(head, fp);
1118 		goto done;
1119 	}
1120 
1121 	/*
1122 	 * Concatenate fragments.
1123 	 */
1124 	m = q;
1125 	t = m->m_next;
1126 	m->m_next = NULL;
1127 	m_cat(m, t);
1128 	nq = q->m_nextpkt;
1129 	q->m_nextpkt = NULL;
1130 	for (q = nq; q != NULL; q = nq) {
1131 		nq = q->m_nextpkt;
1132 		q->m_nextpkt = NULL;
1133 		m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
1134 		m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
1135 		m_cat(m, q);
1136 	}
1137 	/*
1138 	 * In order to do checksumming faster we do 'end-around carry' here
1139 	 * (and not in for{} loop), though it implies we are not going to
1140 	 * reassemble more than 64k fragments.
1141 	 */
1142 	m->m_pkthdr.csum_data =
1143 	    (m->m_pkthdr.csum_data & 0xffff) + (m->m_pkthdr.csum_data >> 16);
1144 #ifdef MAC
1145 	mac_ipq_reassemble(fp, m);
1146 	mac_ipq_destroy(fp);
1147 #endif
1148 
1149 	/*
1150 	 * Create header for new ip packet by modifying header of first
1151 	 * packet;  dequeue and discard fragment reassembly header.
1152 	 * Make header visible.
1153 	 */
1154 	ip->ip_len = (ip->ip_hl << 2) + next;
1155 	ip->ip_src = fp->ipq_src;
1156 	ip->ip_dst = fp->ipq_dst;
1157 	TAILQ_REMOVE(head, fp, ipq_list);
1158 	V_nipq--;
1159 	uma_zfree(V_ipq_zone, fp);
1160 	m->m_len += (ip->ip_hl << 2);
1161 	m->m_data -= (ip->ip_hl << 2);
1162 	/* some debugging cruft by sklower, below, will go away soon */
1163 	if (m->m_flags & M_PKTHDR)	/* XXX this should be done elsewhere */
1164 		m_fixhdr(m);
1165 	IPSTAT_INC(ips_reassembled);
1166 	IPQ_UNLOCK();
1167 	return (m);
1168 
1169 dropfrag:
1170 	IPSTAT_INC(ips_fragdropped);
1171 	if (fp != NULL)
1172 		fp->ipq_nfrags--;
1173 	m_freem(m);
1174 done:
1175 	IPQ_UNLOCK();
1176 	return (NULL);
1177 
1178 #undef GETIP
1179 }
1180 
1181 /*
1182  * Free a fragment reassembly header and all
1183  * associated datagrams.
1184  */
1185 static void
1186 ip_freef(struct ipqhead *fhp, struct ipq *fp)
1187 {
1188 	INIT_VNET_INET(curvnet);
1189 	struct mbuf *q;
1190 
1191 	IPQ_LOCK_ASSERT();
1192 
1193 	while (fp->ipq_frags) {
1194 		q = fp->ipq_frags;
1195 		fp->ipq_frags = q->m_nextpkt;
1196 		m_freem(q);
1197 	}
1198 	TAILQ_REMOVE(fhp, fp, ipq_list);
1199 	uma_zfree(V_ipq_zone, fp);
1200 	V_nipq--;
1201 }
1202 
1203 /*
1204  * IP timer processing;
1205  * if a timer expires on a reassembly
1206  * queue, discard it.
1207  */
1208 void
1209 ip_slowtimo(void)
1210 {
1211 	VNET_ITERATOR_DECL(vnet_iter);
1212 	struct ipq *fp;
1213 	int i;
1214 
1215 	IPQ_LOCK();
1216 	VNET_LIST_RLOCK();
1217 	VNET_FOREACH(vnet_iter) {
1218 		CURVNET_SET(vnet_iter);
1219 		INIT_VNET_INET(vnet_iter);
1220 		for (i = 0; i < IPREASS_NHASH; i++) {
1221 			for(fp = TAILQ_FIRST(&V_ipq[i]); fp;) {
1222 				struct ipq *fpp;
1223 
1224 				fpp = fp;
1225 				fp = TAILQ_NEXT(fp, ipq_list);
1226 				if(--fpp->ipq_ttl == 0) {
1227 					IPSTAT_ADD(ips_fragtimeout,
1228 					    fpp->ipq_nfrags);
1229 					ip_freef(&V_ipq[i], fpp);
1230 				}
1231 			}
1232 		}
1233 		/*
1234 		 * If we are over the maximum number of fragments
1235 		 * (due to the limit being lowered), drain off
1236 		 * enough to get down to the new limit.
1237 		 */
1238 		if (V_maxnipq >= 0 && V_nipq > V_maxnipq) {
1239 			for (i = 0; i < IPREASS_NHASH; i++) {
1240 				while (V_nipq > V_maxnipq &&
1241 				    !TAILQ_EMPTY(&V_ipq[i])) {
1242 					IPSTAT_ADD(ips_fragdropped,
1243 					    TAILQ_FIRST(&V_ipq[i])->ipq_nfrags);
1244 					ip_freef(&V_ipq[i],
1245 					    TAILQ_FIRST(&V_ipq[i]));
1246 				}
1247 			}
1248 		}
1249 		CURVNET_RESTORE();
1250 	}
1251 	VNET_LIST_RUNLOCK();
1252 	IPQ_UNLOCK();
1253 }
1254 
1255 /*
1256  * Drain off all datagram fragments.
1257  */
1258 void
1259 ip_drain(void)
1260 {
1261 	VNET_ITERATOR_DECL(vnet_iter);
1262 	int     i;
1263 
1264 	IPQ_LOCK();
1265 	VNET_LIST_RLOCK();
1266 	VNET_FOREACH(vnet_iter) {
1267 		CURVNET_SET(vnet_iter);
1268 		INIT_VNET_INET(vnet_iter);
1269 		for (i = 0; i < IPREASS_NHASH; i++) {
1270 			while(!TAILQ_EMPTY(&V_ipq[i])) {
1271 				IPSTAT_ADD(ips_fragdropped,
1272 				    TAILQ_FIRST(&V_ipq[i])->ipq_nfrags);
1273 				ip_freef(&V_ipq[i], TAILQ_FIRST(&V_ipq[i]));
1274 			}
1275 		}
1276 		CURVNET_RESTORE();
1277 	}
1278 	VNET_LIST_RUNLOCK();
1279 	IPQ_UNLOCK();
1280 	in_rtqdrain();
1281 }
1282 
1283 /*
1284  * The protocol to be inserted into ip_protox[] must be already registered
1285  * in inetsw[], either statically or through pf_proto_register().
1286  */
1287 int
1288 ipproto_register(u_char ipproto)
1289 {
1290 	struct protosw *pr;
1291 
1292 	/* Sanity checks. */
1293 	if (ipproto == 0)
1294 		return (EPROTONOSUPPORT);
1295 
1296 	/*
1297 	 * The protocol slot must not be occupied by another protocol
1298 	 * already.  An index pointing to IPPROTO_RAW is unused.
1299 	 */
1300 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
1301 	if (pr == NULL)
1302 		return (EPFNOSUPPORT);
1303 	if (ip_protox[ipproto] != pr - inetsw)	/* IPPROTO_RAW */
1304 		return (EEXIST);
1305 
1306 	/* Find the protocol position in inetsw[] and set the index. */
1307 	for (pr = inetdomain.dom_protosw;
1308 	     pr < inetdomain.dom_protoswNPROTOSW; pr++) {
1309 		if (pr->pr_domain->dom_family == PF_INET &&
1310 		    pr->pr_protocol && pr->pr_protocol == ipproto) {
1311 			/* Be careful to only index valid IP protocols. */
1312 			if (pr->pr_protocol < IPPROTO_MAX) {
1313 				ip_protox[pr->pr_protocol] = pr - inetsw;
1314 				return (0);
1315 			} else
1316 				return (EINVAL);
1317 		}
1318 	}
1319 	return (EPROTONOSUPPORT);
1320 }
1321 
1322 int
1323 ipproto_unregister(u_char ipproto)
1324 {
1325 	struct protosw *pr;
1326 
1327 	/* Sanity checks. */
1328 	if (ipproto == 0)
1329 		return (EPROTONOSUPPORT);
1330 
1331 	/* Check if the protocol was indeed registered. */
1332 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
1333 	if (pr == NULL)
1334 		return (EPFNOSUPPORT);
1335 	if (ip_protox[ipproto] == pr - inetsw)  /* IPPROTO_RAW */
1336 		return (ENOENT);
1337 
1338 	/* Reset the protocol slot to IPPROTO_RAW. */
1339 	ip_protox[ipproto] = pr - inetsw;
1340 	return (0);
1341 }
1342 
1343 /*
1344  * Given address of next destination (final or next hop), return (referenced)
1345  * internet address info of interface to be used to get there.
1346  */
1347 struct in_ifaddr *
1348 ip_rtaddr(struct in_addr dst, u_int fibnum)
1349 {
1350 	struct route sro;
1351 	struct sockaddr_in *sin;
1352 	struct in_ifaddr *ifa;
1353 
1354 	bzero(&sro, sizeof(sro));
1355 	sin = (struct sockaddr_in *)&sro.ro_dst;
1356 	sin->sin_family = AF_INET;
1357 	sin->sin_len = sizeof(*sin);
1358 	sin->sin_addr = dst;
1359 	in_rtalloc_ign(&sro, 0, fibnum);
1360 
1361 	if (sro.ro_rt == NULL)
1362 		return (NULL);
1363 
1364 	ifa = ifatoia(sro.ro_rt->rt_ifa);
1365 	ifa_ref(&ifa->ia_ifa);
1366 	RTFREE(sro.ro_rt);
1367 	return (ifa);
1368 }
1369 
1370 u_char inetctlerrmap[PRC_NCMDS] = {
1371 	0,		0,		0,		0,
1372 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1373 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1374 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1375 	0,		0,		EHOSTUNREACH,	0,
1376 	ENOPROTOOPT,	ECONNREFUSED
1377 };
1378 
1379 /*
1380  * Forward a packet.  If some error occurs return the sender
1381  * an icmp packet.  Note we can't always generate a meaningful
1382  * icmp message because icmp doesn't have a large enough repertoire
1383  * of codes and types.
1384  *
1385  * If not forwarding, just drop the packet.  This could be confusing
1386  * if ipforwarding was zero but some routing protocol was advancing
1387  * us as a gateway to somewhere.  However, we must let the routing
1388  * protocol deal with that.
1389  *
1390  * The srcrt parameter indicates whether the packet is being forwarded
1391  * via a source route.
1392  */
1393 void
1394 ip_forward(struct mbuf *m, int srcrt)
1395 {
1396 	INIT_VNET_INET(curvnet);
1397 	struct ip *ip = mtod(m, struct ip *);
1398 	struct in_ifaddr *ia;
1399 	struct mbuf *mcopy;
1400 	struct in_addr dest;
1401 	struct route ro;
1402 	int error, type = 0, code = 0, mtu = 0;
1403 
1404 	if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
1405 		IPSTAT_INC(ips_cantforward);
1406 		m_freem(m);
1407 		return;
1408 	}
1409 #ifdef IPSTEALTH
1410 	if (!V_ipstealth) {
1411 #endif
1412 		if (ip->ip_ttl <= IPTTLDEC) {
1413 			icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS,
1414 			    0, 0);
1415 			return;
1416 		}
1417 #ifdef IPSTEALTH
1418 	}
1419 #endif
1420 
1421 	ia = ip_rtaddr(ip->ip_dst, M_GETFIB(m));
1422 #ifndef IPSEC
1423 	/*
1424 	 * 'ia' may be NULL if there is no route for this destination.
1425 	 * In case of IPsec, Don't discard it just yet, but pass it to
1426 	 * ip_output in case of outgoing IPsec policy.
1427 	 */
1428 	if (!srcrt && ia == NULL) {
1429 		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
1430 		return;
1431 	}
1432 #endif
1433 
1434 	/*
1435 	 * Save the IP header and at most 8 bytes of the payload,
1436 	 * in case we need to generate an ICMP message to the src.
1437 	 *
1438 	 * XXX this can be optimized a lot by saving the data in a local
1439 	 * buffer on the stack (72 bytes at most), and only allocating the
1440 	 * mbuf if really necessary. The vast majority of the packets
1441 	 * are forwarded without having to send an ICMP back (either
1442 	 * because unnecessary, or because rate limited), so we are
1443 	 * really we are wasting a lot of work here.
1444 	 *
1445 	 * We don't use m_copy() because it might return a reference
1446 	 * to a shared cluster. Both this function and ip_output()
1447 	 * assume exclusive access to the IP header in `m', so any
1448 	 * data in a cluster may change before we reach icmp_error().
1449 	 */
1450 	MGETHDR(mcopy, M_DONTWAIT, m->m_type);
1451 	if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_DONTWAIT)) {
1452 		/*
1453 		 * It's probably ok if the pkthdr dup fails (because
1454 		 * the deep copy of the tag chain failed), but for now
1455 		 * be conservative and just discard the copy since
1456 		 * code below may some day want the tags.
1457 		 */
1458 		m_free(mcopy);
1459 		mcopy = NULL;
1460 	}
1461 	if (mcopy != NULL) {
1462 		mcopy->m_len = min(ip->ip_len, M_TRAILINGSPACE(mcopy));
1463 		mcopy->m_pkthdr.len = mcopy->m_len;
1464 		m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1465 	}
1466 
1467 #ifdef IPSTEALTH
1468 	if (!V_ipstealth) {
1469 #endif
1470 		ip->ip_ttl -= IPTTLDEC;
1471 #ifdef IPSTEALTH
1472 	}
1473 #endif
1474 
1475 	/*
1476 	 * If forwarding packet using same interface that it came in on,
1477 	 * perhaps should send a redirect to sender to shortcut a hop.
1478 	 * Only send redirect if source is sending directly to us,
1479 	 * and if packet was not source routed (or has any options).
1480 	 * Also, don't send redirect if forwarding using a default route
1481 	 * or a route modified by a redirect.
1482 	 */
1483 	dest.s_addr = 0;
1484 	if (!srcrt && V_ipsendredirects &&
1485 	    ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) {
1486 		struct sockaddr_in *sin;
1487 		struct rtentry *rt;
1488 
1489 		bzero(&ro, sizeof(ro));
1490 		sin = (struct sockaddr_in *)&ro.ro_dst;
1491 		sin->sin_family = AF_INET;
1492 		sin->sin_len = sizeof(*sin);
1493 		sin->sin_addr = ip->ip_dst;
1494 		in_rtalloc_ign(&ro, 0, M_GETFIB(m));
1495 
1496 		rt = ro.ro_rt;
1497 
1498 		if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1499 		    satosin(rt_key(rt))->sin_addr.s_addr != 0) {
1500 #define	RTA(rt)	((struct in_ifaddr *)(rt->rt_ifa))
1501 			u_long src = ntohl(ip->ip_src.s_addr);
1502 
1503 			if (RTA(rt) &&
1504 			    (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1505 				if (rt->rt_flags & RTF_GATEWAY)
1506 					dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr;
1507 				else
1508 					dest.s_addr = ip->ip_dst.s_addr;
1509 				/* Router requirements says to only send host redirects */
1510 				type = ICMP_REDIRECT;
1511 				code = ICMP_REDIRECT_HOST;
1512 			}
1513 		}
1514 		if (rt)
1515 			RTFREE(rt);
1516 	}
1517 
1518 	/*
1519 	 * Try to cache the route MTU from ip_output so we can consider it for
1520 	 * the ICMP_UNREACH_NEEDFRAG "Next-Hop MTU" field described in RFC1191.
1521 	 */
1522 	bzero(&ro, sizeof(ro));
1523 
1524 	error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
1525 
1526 	if (error == EMSGSIZE && ro.ro_rt)
1527 		mtu = ro.ro_rt->rt_rmx.rmx_mtu;
1528 	if (ro.ro_rt)
1529 		RTFREE(ro.ro_rt);
1530 
1531 	if (error)
1532 		IPSTAT_INC(ips_cantforward);
1533 	else {
1534 		IPSTAT_INC(ips_forward);
1535 		if (type)
1536 			IPSTAT_INC(ips_redirectsent);
1537 		else {
1538 			if (mcopy)
1539 				m_freem(mcopy);
1540 			if (ia != NULL)
1541 				ifa_free(&ia->ia_ifa);
1542 			return;
1543 		}
1544 	}
1545 	if (mcopy == NULL) {
1546 		if (ia != NULL)
1547 			ifa_free(&ia->ia_ifa);
1548 		return;
1549 	}
1550 
1551 	switch (error) {
1552 
1553 	case 0:				/* forwarded, but need redirect */
1554 		/* type, code set above */
1555 		break;
1556 
1557 	case ENETUNREACH:
1558 	case EHOSTUNREACH:
1559 	case ENETDOWN:
1560 	case EHOSTDOWN:
1561 	default:
1562 		type = ICMP_UNREACH;
1563 		code = ICMP_UNREACH_HOST;
1564 		break;
1565 
1566 	case EMSGSIZE:
1567 		type = ICMP_UNREACH;
1568 		code = ICMP_UNREACH_NEEDFRAG;
1569 
1570 #ifdef IPSEC
1571 		/*
1572 		 * If IPsec is configured for this path,
1573 		 * override any possibly mtu value set by ip_output.
1574 		 */
1575 		mtu = ip_ipsec_mtu(m, mtu);
1576 #endif /* IPSEC */
1577 		/*
1578 		 * If the MTU was set before make sure we are below the
1579 		 * interface MTU.
1580 		 * If the MTU wasn't set before use the interface mtu or
1581 		 * fall back to the next smaller mtu step compared to the
1582 		 * current packet size.
1583 		 */
1584 		if (mtu != 0) {
1585 			if (ia != NULL)
1586 				mtu = min(mtu, ia->ia_ifp->if_mtu);
1587 		} else {
1588 			if (ia != NULL)
1589 				mtu = ia->ia_ifp->if_mtu;
1590 			else
1591 				mtu = ip_next_mtu(ip->ip_len, 0);
1592 		}
1593 		IPSTAT_INC(ips_cantfrag);
1594 		break;
1595 
1596 	case ENOBUFS:
1597 		/*
1598 		 * A router should not generate ICMP_SOURCEQUENCH as
1599 		 * required in RFC1812 Requirements for IP Version 4 Routers.
1600 		 * Source quench could be a big problem under DoS attacks,
1601 		 * or if the underlying interface is rate-limited.
1602 		 * Those who need source quench packets may re-enable them
1603 		 * via the net.inet.ip.sendsourcequench sysctl.
1604 		 */
1605 		if (V_ip_sendsourcequench == 0) {
1606 			m_freem(mcopy);
1607 			if (ia != NULL)
1608 				ifa_free(&ia->ia_ifa);
1609 			return;
1610 		} else {
1611 			type = ICMP_SOURCEQUENCH;
1612 			code = 0;
1613 		}
1614 		break;
1615 
1616 	case EACCES:			/* ipfw denied packet */
1617 		m_freem(mcopy);
1618 		if (ia != NULL)
1619 			ifa_free(&ia->ia_ifa);
1620 		return;
1621 	}
1622 	if (ia != NULL)
1623 		ifa_free(&ia->ia_ifa);
1624 	icmp_error(mcopy, type, code, dest.s_addr, mtu);
1625 }
1626 
1627 void
1628 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1629     struct mbuf *m)
1630 {
1631 	INIT_VNET_NET(inp->inp_vnet);
1632 
1633 	if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) {
1634 		struct bintime bt;
1635 
1636 		bintime(&bt);
1637 		if (inp->inp_socket->so_options & SO_BINTIME) {
1638 			*mp = sbcreatecontrol((caddr_t) &bt, sizeof(bt),
1639 			SCM_BINTIME, SOL_SOCKET);
1640 			if (*mp)
1641 				mp = &(*mp)->m_next;
1642 		}
1643 		if (inp->inp_socket->so_options & SO_TIMESTAMP) {
1644 			struct timeval tv;
1645 
1646 			bintime2timeval(&bt, &tv);
1647 			*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1648 				SCM_TIMESTAMP, SOL_SOCKET);
1649 			if (*mp)
1650 				mp = &(*mp)->m_next;
1651 		}
1652 	}
1653 	if (inp->inp_flags & INP_RECVDSTADDR) {
1654 		*mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
1655 		    sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1656 		if (*mp)
1657 			mp = &(*mp)->m_next;
1658 	}
1659 	if (inp->inp_flags & INP_RECVTTL) {
1660 		*mp = sbcreatecontrol((caddr_t) &ip->ip_ttl,
1661 		    sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
1662 		if (*mp)
1663 			mp = &(*mp)->m_next;
1664 	}
1665 #ifdef notyet
1666 	/* XXX
1667 	 * Moving these out of udp_input() made them even more broken
1668 	 * than they already were.
1669 	 */
1670 	/* options were tossed already */
1671 	if (inp->inp_flags & INP_RECVOPTS) {
1672 		*mp = sbcreatecontrol((caddr_t) opts_deleted_above,
1673 		    sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1674 		if (*mp)
1675 			mp = &(*mp)->m_next;
1676 	}
1677 	/* ip_srcroute doesn't do what we want here, need to fix */
1678 	if (inp->inp_flags & INP_RECVRETOPTS) {
1679 		*mp = sbcreatecontrol((caddr_t) ip_srcroute(m),
1680 		    sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1681 		if (*mp)
1682 			mp = &(*mp)->m_next;
1683 	}
1684 #endif
1685 	if (inp->inp_flags & INP_RECVIF) {
1686 		struct ifnet *ifp;
1687 		struct sdlbuf {
1688 			struct sockaddr_dl sdl;
1689 			u_char	pad[32];
1690 		} sdlbuf;
1691 		struct sockaddr_dl *sdp;
1692 		struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1693 
1694 		if (((ifp = m->m_pkthdr.rcvif))
1695 		&& ( ifp->if_index && (ifp->if_index <= V_if_index))) {
1696 			sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1697 			/*
1698 			 * Change our mind and don't try copy.
1699 			 */
1700 			if ((sdp->sdl_family != AF_LINK)
1701 			|| (sdp->sdl_len > sizeof(sdlbuf))) {
1702 				goto makedummy;
1703 			}
1704 			bcopy(sdp, sdl2, sdp->sdl_len);
1705 		} else {
1706 makedummy:
1707 			sdl2->sdl_len
1708 				= offsetof(struct sockaddr_dl, sdl_data[0]);
1709 			sdl2->sdl_family = AF_LINK;
1710 			sdl2->sdl_index = 0;
1711 			sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1712 		}
1713 		*mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len,
1714 			IP_RECVIF, IPPROTO_IP);
1715 		if (*mp)
1716 			mp = &(*mp)->m_next;
1717 	}
1718 }
1719 
1720 /*
1721  * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
1722  * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
1723  * locking.  This code remains in ip_input.c as ip_mroute.c is optionally
1724  * compiled.
1725  */
1726 int
1727 ip_rsvp_init(struct socket *so)
1728 {
1729 	INIT_VNET_INET(so->so_vnet);
1730 
1731 	if (so->so_type != SOCK_RAW ||
1732 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
1733 		return EOPNOTSUPP;
1734 
1735 	if (V_ip_rsvpd != NULL)
1736 		return EADDRINUSE;
1737 
1738 	V_ip_rsvpd = so;
1739 	/*
1740 	 * This may seem silly, but we need to be sure we don't over-increment
1741 	 * the RSVP counter, in case something slips up.
1742 	 */
1743 	if (!V_ip_rsvp_on) {
1744 		V_ip_rsvp_on = 1;
1745 		V_rsvp_on++;
1746 	}
1747 
1748 	return 0;
1749 }
1750 
1751 int
1752 ip_rsvp_done(void)
1753 {
1754 	INIT_VNET_INET(curvnet);
1755 
1756 	V_ip_rsvpd = NULL;
1757 	/*
1758 	 * This may seem silly, but we need to be sure we don't over-decrement
1759 	 * the RSVP counter, in case something slips up.
1760 	 */
1761 	if (V_ip_rsvp_on) {
1762 		V_ip_rsvp_on = 0;
1763 		V_rsvp_on--;
1764 	}
1765 	return 0;
1766 }
1767 
1768 void
1769 rsvp_input(struct mbuf *m, int off)	/* XXX must fixup manually */
1770 {
1771 	INIT_VNET_INET(curvnet);
1772 
1773 	if (rsvp_input_p) { /* call the real one if loaded */
1774 		rsvp_input_p(m, off);
1775 		return;
1776 	}
1777 
1778 	/* Can still get packets with rsvp_on = 0 if there is a local member
1779 	 * of the group to which the RSVP packet is addressed.  But in this
1780 	 * case we want to throw the packet away.
1781 	 */
1782 
1783 	if (!V_rsvp_on) {
1784 		m_freem(m);
1785 		return;
1786 	}
1787 
1788 	if (V_ip_rsvpd != NULL) {
1789 		rip_input(m, off);
1790 		return;
1791 	}
1792 	/* Drop the packet */
1793 	m_freem(m);
1794 }
1795