xref: /freebsd/sys/netinet/ip_input.c (revision 902136e0fe112383ec64d2ef43a446063b5e6417)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1982, 1986, 1988, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the University nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  */
31 
32 #include "opt_bootp.h"
33 #include "opt_inet.h"
34 #include "opt_ipstealth.h"
35 #include "opt_ipsec.h"
36 #include "opt_route.h"
37 #include "opt_rss.h"
38 #include "opt_sctp.h"
39 
40 #include <sys/param.h>
41 #include <sys/systm.h>
42 #include <sys/hhook.h>
43 #include <sys/mbuf.h>
44 #include <sys/malloc.h>
45 #include <sys/domain.h>
46 #include <sys/protosw.h>
47 #include <sys/socket.h>
48 #include <sys/time.h>
49 #include <sys/kernel.h>
50 #include <sys/lock.h>
51 #include <sys/rmlock.h>
52 #include <sys/rwlock.h>
53 #include <sys/sdt.h>
54 #include <sys/syslog.h>
55 #include <sys/sysctl.h>
56 
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/if_private.h>
62 #include <net/pfil.h>
63 #include <net/route.h>
64 #include <net/route/nhop.h>
65 #include <net/netisr.h>
66 #include <net/rss_config.h>
67 #include <net/vnet.h>
68 
69 #include <netinet/in.h>
70 #include <netinet/in_kdtrace.h>
71 #include <netinet/in_systm.h>
72 #include <netinet/in_var.h>
73 #include <netinet/ip.h>
74 #include <netinet/in_fib.h>
75 #include <netinet/in_pcb.h>
76 #include <netinet/ip_var.h>
77 #include <netinet/ip_encap.h>
78 #include <netinet/ip_fw.h>
79 #include <netinet/ip_icmp.h>
80 #include <netinet/igmp_var.h>
81 #include <netinet/ip_options.h>
82 #include <machine/in_cksum.h>
83 #include <netinet/ip_carp.h>
84 #include <netinet/in_rss.h>
85 #include <netinet/ip_mroute.h>
86 #ifdef SCTP
87 #include <netinet/sctp_var.h>
88 #endif
89 
90 #include <netipsec/ipsec_support.h>
91 
92 #include <sys/socketvar.h>
93 
94 #include <security/mac/mac_framework.h>
95 
96 #ifdef CTASSERT
97 CTASSERT(sizeof(struct ip) == 20);
98 #endif
99 
100 /* IP reassembly functions are defined in ip_reass.c. */
101 extern void ipreass_init(void);
102 extern void ipreass_vnet_init(void);
103 #ifdef VIMAGE
104 extern void ipreass_destroy(void);
105 #endif
106 
107 VNET_DEFINE(int, rsvp_on);
108 
109 VNET_DEFINE(int, ipforwarding);
110 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_VNET | CTLFLAG_RW,
111     &VNET_NAME(ipforwarding), 0,
112     "Enable IP forwarding between interfaces");
113 
114 /*
115  * Respond with an ICMP host redirect when we forward a packet out of
116  * the same interface on which it was received.  See RFC 792.
117  */
118 VNET_DEFINE(int, ipsendredirects) = 1;
119 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_VNET | CTLFLAG_RW,
120     &VNET_NAME(ipsendredirects), 0,
121     "Enable sending IP redirects");
122 
123 VNET_DEFINE_STATIC(bool, ip_strong_es) = false;
124 #define	V_ip_strong_es	VNET(ip_strong_es)
125 SYSCTL_BOOL(_net_inet_ip, OID_AUTO, rfc1122_strong_es,
126     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip_strong_es), false,
127     "Packet's IP destination address must match address on arrival interface");
128 
129 VNET_DEFINE_STATIC(bool, ip_sav) = true;
130 #define	V_ip_sav	VNET(ip_sav)
131 SYSCTL_BOOL(_net_inet_ip, OID_AUTO, source_address_validation,
132     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip_sav), true,
133     "Drop incoming packets with source address that is a local address");
134 
135 /* Packet filter hooks */
136 VNET_DEFINE(pfil_head_t, inet_pfil_head);
137 VNET_DEFINE(pfil_head_t, inet_local_pfil_head);
138 
139 static struct netisr_handler ip_nh = {
140 	.nh_name = "ip",
141 	.nh_handler = ip_input,
142 	.nh_proto = NETISR_IP,
143 #ifdef	RSS
144 	.nh_m2cpuid = rss_soft_m2cpuid_v4,
145 	.nh_policy = NETISR_POLICY_CPU,
146 	.nh_dispatch = NETISR_DISPATCH_HYBRID,
147 #else
148 	.nh_policy = NETISR_POLICY_FLOW,
149 #endif
150 };
151 
152 #ifdef	RSS
153 /*
154  * Directly dispatched frames are currently assumed
155  * to have a flowid already calculated.
156  *
157  * It should likely have something that assert it
158  * actually has valid flow details.
159  */
160 static struct netisr_handler ip_direct_nh = {
161 	.nh_name = "ip_direct",
162 	.nh_handler = ip_direct_input,
163 	.nh_proto = NETISR_IP_DIRECT,
164 	.nh_m2cpuid = rss_soft_m2cpuid_v4,
165 	.nh_policy = NETISR_POLICY_CPU,
166 	.nh_dispatch = NETISR_DISPATCH_HYBRID,
167 };
168 #endif
169 
170 ipproto_input_t		*ip_protox[IPPROTO_MAX] = {
171 			    [0 ... IPPROTO_MAX - 1] = rip_input };
172 ipproto_ctlinput_t	*ip_ctlprotox[IPPROTO_MAX] = {
173 			    [0 ... IPPROTO_MAX - 1] = rip_ctlinput };
174 
175 VNET_DEFINE(struct in_ifaddrhead, in_ifaddrhead);  /* first inet address */
176 VNET_DEFINE(struct in_ifaddrhashhead *, in_ifaddrhashtbl); /* inet addr hash table  */
177 VNET_DEFINE(u_long, in_ifaddrhmask);		/* mask for hash table */
178 
179 /* Make sure it is safe to use hashinit(9) on CK_LIST. */
180 CTASSERT(sizeof(struct in_ifaddrhashhead) == sizeof(LIST_HEAD(, in_addr)));
181 
182 #ifdef IPCTL_DEFMTU
183 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
184     &ip_mtu, 0, "Default MTU");
185 #endif
186 
187 #ifdef IPSTEALTH
188 VNET_DEFINE(int, ipstealth);
189 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_VNET | CTLFLAG_RW,
190     &VNET_NAME(ipstealth), 0,
191     "IP stealth mode, no TTL decrementation on forwarding");
192 #endif
193 
194 /*
195  * IP statistics are stored in the "array" of counter(9)s.
196  */
197 VNET_PCPUSTAT_DEFINE(struct ipstat, ipstat);
198 VNET_PCPUSTAT_SYSINIT(ipstat);
199 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, IPCTL_STATS, stats, struct ipstat, ipstat,
200     "IP statistics (struct ipstat, netinet/ip_var.h)");
201 
202 #ifdef VIMAGE
203 VNET_PCPUSTAT_SYSUNINIT(ipstat);
204 #endif /* VIMAGE */
205 
206 /*
207  * Kernel module interface for updating ipstat.  The argument is an index
208  * into ipstat treated as an array.
209  */
210 void
211 kmod_ipstat_inc(int statnum)
212 {
213 
214 	counter_u64_add(VNET(ipstat)[statnum], 1);
215 }
216 
217 void
218 kmod_ipstat_dec(int statnum)
219 {
220 
221 	counter_u64_add(VNET(ipstat)[statnum], -1);
222 }
223 
224 static int
225 sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS)
226 {
227 	int error, qlimit;
228 
229 	netisr_getqlimit(&ip_nh, &qlimit);
230 	error = sysctl_handle_int(oidp, &qlimit, 0, req);
231 	if (error || !req->newptr)
232 		return (error);
233 	if (qlimit < 1)
234 		return (EINVAL);
235 	return (netisr_setqlimit(&ip_nh, qlimit));
236 }
237 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen,
238     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 0,
239     sysctl_netinet_intr_queue_maxlen, "I",
240     "Maximum size of the IP input queue");
241 
242 static int
243 sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS)
244 {
245 	u_int64_t qdrops_long;
246 	int error, qdrops;
247 
248 	netisr_getqdrops(&ip_nh, &qdrops_long);
249 	qdrops = qdrops_long;
250 	error = sysctl_handle_int(oidp, &qdrops, 0, req);
251 	if (error || !req->newptr)
252 		return (error);
253 	if (qdrops != 0)
254 		return (EINVAL);
255 	netisr_clearqdrops(&ip_nh);
256 	return (0);
257 }
258 
259 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops,
260     CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE,
261     0, 0, sysctl_netinet_intr_queue_drops, "I",
262     "Number of packets dropped from the IP input queue");
263 
264 #ifdef	RSS
265 static int
266 sysctl_netinet_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS)
267 {
268 	int error, qlimit;
269 
270 	netisr_getqlimit(&ip_direct_nh, &qlimit);
271 	error = sysctl_handle_int(oidp, &qlimit, 0, req);
272 	if (error || !req->newptr)
273 		return (error);
274 	if (qlimit < 1)
275 		return (EINVAL);
276 	return (netisr_setqlimit(&ip_direct_nh, qlimit));
277 }
278 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRDQMAXLEN, intr_direct_queue_maxlen,
279     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
280     0, 0, sysctl_netinet_intr_direct_queue_maxlen,
281     "I", "Maximum size of the IP direct input queue");
282 
283 static int
284 sysctl_netinet_intr_direct_queue_drops(SYSCTL_HANDLER_ARGS)
285 {
286 	u_int64_t qdrops_long;
287 	int error, qdrops;
288 
289 	netisr_getqdrops(&ip_direct_nh, &qdrops_long);
290 	qdrops = qdrops_long;
291 	error = sysctl_handle_int(oidp, &qdrops, 0, req);
292 	if (error || !req->newptr)
293 		return (error);
294 	if (qdrops != 0)
295 		return (EINVAL);
296 	netisr_clearqdrops(&ip_direct_nh);
297 	return (0);
298 }
299 
300 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRDQDROPS, intr_direct_queue_drops,
301     CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, 0, 0,
302     sysctl_netinet_intr_direct_queue_drops, "I",
303     "Number of packets dropped from the IP direct input queue");
304 #endif	/* RSS */
305 
306 /*
307  * IP initialization: fill in IP protocol switch table.
308  * All protocols not implemented in kernel go to raw IP protocol handler.
309  */
310 static void
311 ip_vnet_init(void *arg __unused)
312 {
313 	struct pfil_head_args args;
314 
315 	CK_STAILQ_INIT(&V_in_ifaddrhead);
316 	V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask);
317 
318 	/* Initialize IP reassembly queue. */
319 	ipreass_vnet_init();
320 
321 	/* Initialize packet filter hooks. */
322 	args.pa_version = PFIL_VERSION;
323 	args.pa_flags = PFIL_IN | PFIL_OUT;
324 	args.pa_type = PFIL_TYPE_IP4;
325 	args.pa_headname = PFIL_INET_NAME;
326 	V_inet_pfil_head = pfil_head_register(&args);
327 
328 	args.pa_flags = PFIL_OUT;
329 	args.pa_headname = PFIL_INET_LOCAL_NAME;
330 	V_inet_local_pfil_head = pfil_head_register(&args);
331 
332 	if (hhook_head_register(HHOOK_TYPE_IPSEC_IN, AF_INET,
333 	    &V_ipsec_hhh_in[HHOOK_IPSEC_INET],
334 	    HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
335 		printf("%s: WARNING: unable to register input helper hook\n",
336 		    __func__);
337 	if (hhook_head_register(HHOOK_TYPE_IPSEC_OUT, AF_INET,
338 	    &V_ipsec_hhh_out[HHOOK_IPSEC_INET],
339 	    HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
340 		printf("%s: WARNING: unable to register output helper hook\n",
341 		    __func__);
342 
343 #ifdef VIMAGE
344 	netisr_register_vnet(&ip_nh);
345 #ifdef	RSS
346 	netisr_register_vnet(&ip_direct_nh);
347 #endif
348 #endif
349 }
350 VNET_SYSINIT(ip_vnet_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_FOURTH,
351     ip_vnet_init, NULL);
352 
353 static void
354 ip_init(const void *unused __unused)
355 {
356 	struct ifnet *ifp;
357 
358 	ipreass_init();
359 
360 	/*
361 	 * Register statically compiled protocols, that are unlikely to
362 	 * ever become dynamic.
363 	 */
364 	IPPROTO_REGISTER(IPPROTO_ICMP, icmp_input, NULL);
365 	IPPROTO_REGISTER(IPPROTO_IGMP, igmp_input, NULL);
366 	IPPROTO_REGISTER(IPPROTO_RSVP, rsvp_input, NULL);
367 	IPPROTO_REGISTER(IPPROTO_IPV4, encap4_input, NULL);
368 	IPPROTO_REGISTER(IPPROTO_MOBILE, encap4_input, NULL);
369 	IPPROTO_REGISTER(IPPROTO_ETHERIP, encap4_input, NULL);
370 	IPPROTO_REGISTER(IPPROTO_GRE, encap4_input, NULL);
371 	IPPROTO_REGISTER(IPPROTO_IPV6, encap4_input, NULL);
372 	IPPROTO_REGISTER(IPPROTO_PIM, encap4_input, NULL);
373 #ifdef SCTP	/* XXX: has a loadable & static version */
374 	IPPROTO_REGISTER(IPPROTO_SCTP, sctp_input, sctp_ctlinput);
375 #endif
376 
377 	netisr_register(&ip_nh);
378 #ifdef	RSS
379 	netisr_register(&ip_direct_nh);
380 #endif
381 	/*
382 	 * XXXGL: we use SYSINIT() here, but go over V_ifnet.  It was the same
383 	 * way before dom_ifattach removal.  This worked because when any
384 	 * non-default vnet is created, there are no interfaces inside.
385 	 * Eventually this needs to be fixed.
386 	 */
387 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link)
388 		in_ifattach(NULL, ifp);
389 }
390 SYSINIT(ip_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip_init, NULL);
391 
392 #ifdef VIMAGE
393 static void
394 ip_destroy(void *unused __unused)
395 {
396 	int error;
397 
398 #ifdef	RSS
399 	netisr_unregister_vnet(&ip_direct_nh);
400 #endif
401 	netisr_unregister_vnet(&ip_nh);
402 
403 	pfil_head_unregister(V_inet_pfil_head);
404 	error = hhook_head_deregister(V_ipsec_hhh_in[HHOOK_IPSEC_INET]);
405 	if (error != 0) {
406 		printf("%s: WARNING: unable to deregister input helper hook "
407 		    "type HHOOK_TYPE_IPSEC_IN, id HHOOK_IPSEC_INET: "
408 		    "error %d returned\n", __func__, error);
409 	}
410 	error = hhook_head_deregister(V_ipsec_hhh_out[HHOOK_IPSEC_INET]);
411 	if (error != 0) {
412 		printf("%s: WARNING: unable to deregister output helper hook "
413 		    "type HHOOK_TYPE_IPSEC_OUT, id HHOOK_IPSEC_INET: "
414 		    "error %d returned\n", __func__, error);
415 	}
416 
417 	/* Remove the IPv4 addresses from all interfaces. */
418 	in_ifscrub_all();
419 
420 	/* Make sure the IPv4 routes are gone as well. */
421 	rib_flush_routes_family(AF_INET);
422 
423 	/* Destroy IP reassembly queue. */
424 	ipreass_destroy();
425 
426 	/* Cleanup in_ifaddr hash table; should be empty. */
427 	hashdestroy(V_in_ifaddrhashtbl, M_IFADDR, V_in_ifaddrhmask);
428 }
429 
430 VNET_SYSUNINIT(ip, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip_destroy, NULL);
431 #endif
432 
433 #ifdef	RSS
434 /*
435  * IP direct input routine.
436  *
437  * This is called when reinjecting completed fragments where
438  * all of the previous checking and book-keeping has been done.
439  */
440 void
441 ip_direct_input(struct mbuf *m)
442 {
443 	struct ip *ip;
444 	int hlen;
445 
446 	ip = mtod(m, struct ip *);
447 	hlen = ip->ip_hl << 2;
448 
449 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
450 	if (IPSEC_ENABLED(ipv4)) {
451 		if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
452 			return;
453 	}
454 #endif /* IPSEC */
455 	IPSTAT_INC(ips_delivered);
456 	ip_protox[ip->ip_p](&m, &hlen, ip->ip_p);
457 }
458 #endif
459 
460 /*
461  * Ip input routine.  Checksum and byte swap header.  If fragmented
462  * try to reassemble.  Process options.  Pass to next level.
463  */
464 void
465 ip_input(struct mbuf *m)
466 {
467 	struct ip *ip = NULL;
468 	struct in_ifaddr *ia = NULL;
469 	struct ifaddr *ifa;
470 	struct ifnet *ifp;
471 	int hlen = 0;
472 	uint16_t sum, ip_len;
473 	int dchg = 0;				/* dest changed after fw */
474 	struct in_addr odst;			/* original dst address */
475 	bool strong_es;
476 
477 	M_ASSERTPKTHDR(m);
478 	NET_EPOCH_ASSERT();
479 
480 	if (m->m_flags & M_FASTFWD_OURS) {
481 		m->m_flags &= ~M_FASTFWD_OURS;
482 		/* Set up some basics that will be used later. */
483 		ip = mtod(m, struct ip *);
484 		hlen = ip->ip_hl << 2;
485 		ip_len = ntohs(ip->ip_len);
486 		goto ours;
487 	}
488 
489 	IPSTAT_INC(ips_total);
490 
491 	if (__predict_false(m->m_pkthdr.len < sizeof(struct ip)))
492 		goto tooshort;
493 
494 	if (m->m_len < sizeof(struct ip)) {
495 		m = m_pullup(m, sizeof(struct ip));
496 		if (__predict_false(m == NULL)) {
497 			IPSTAT_INC(ips_toosmall);
498 			return;
499 		}
500 	}
501 	ip = mtod(m, struct ip *);
502 
503 	if (__predict_false(ip->ip_v != IPVERSION)) {
504 		IPSTAT_INC(ips_badvers);
505 		goto bad;
506 	}
507 
508 	hlen = ip->ip_hl << 2;
509 	if (__predict_false(hlen < sizeof(struct ip))) {	/* minimum header length */
510 		IPSTAT_INC(ips_badhlen);
511 		goto bad;
512 	}
513 	if (hlen > m->m_len) {
514 		m = m_pullup(m, hlen);
515 		if (__predict_false(m == NULL)) {
516 			IPSTAT_INC(ips_badhlen);
517 			return;
518 		}
519 		ip = mtod(m, struct ip *);
520 	}
521 
522 	IP_PROBE(receive, NULL, NULL, ip, m->m_pkthdr.rcvif, ip, NULL);
523 
524 	/* IN_LOOPBACK must not appear on the wire - RFC1122 */
525 	ifp = m->m_pkthdr.rcvif;
526 	if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) ||
527 	    IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) {
528 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
529 			IPSTAT_INC(ips_badaddr);
530 			goto bad;
531 		}
532 	}
533 
534 	if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
535 		sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
536 	} else if (m->m_pkthdr.csum_flags & CSUM_IP) {
537 		/*
538 		 * Packet from local host that offloaded checksum computation.
539 		 * Checksum not required since the packet wasn't on the wire.
540 		 */
541 		sum = 0;
542 	} else {
543 		if (hlen == sizeof(struct ip)) {
544 			sum = in_cksum_hdr(ip);
545 		} else {
546 			sum = in_cksum(m, hlen);
547 		}
548 	}
549 	if (__predict_false(sum)) {
550 		IPSTAT_INC(ips_badsum);
551 		goto bad;
552 	}
553 
554 	ip_len = ntohs(ip->ip_len);
555 	if (__predict_false(ip_len < hlen)) {
556 		IPSTAT_INC(ips_badlen);
557 		goto bad;
558 	}
559 
560 	/*
561 	 * Check that the amount of data in the buffers
562 	 * is as at least much as the IP header would have us expect.
563 	 * Trim mbufs if longer than we expect.
564 	 * Drop packet if shorter than we expect.
565 	 */
566 	if (__predict_false(m->m_pkthdr.len < ip_len)) {
567 tooshort:
568 		IPSTAT_INC(ips_tooshort);
569 		goto bad;
570 	}
571 	if (m->m_pkthdr.len > ip_len) {
572 		if (m->m_len == m->m_pkthdr.len) {
573 			m->m_len = ip_len;
574 			m->m_pkthdr.len = ip_len;
575 		} else
576 			m_adj(m, ip_len - m->m_pkthdr.len);
577 	}
578 
579 	/*
580 	 * Try to forward the packet, but if we fail continue.
581 	 * ip_tryforward() may generate redirects these days.
582 	 * XXX the logic below falling through to normal processing
583 	 * if redirects are required should be revisited as well.
584 	 * ip_tryforward() does inbound and outbound packet firewall
585 	 * processing. If firewall has decided that destination becomes
586 	 * our local address, it sets M_FASTFWD_OURS flag. In this
587 	 * case skip another inbound firewall processing and update
588 	 * ip pointer.
589 	 */
590 	if (V_ipforwarding != 0
591 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
592 	    && (!IPSEC_ENABLED(ipv4) ||
593 	    IPSEC_CAPS(ipv4, m, IPSEC_CAP_OPERABLE) == 0)
594 #endif
595 	    ) {
596 		/*
597 		 * ip_dooptions() was run so we can ignore the source route (or
598 		 * any IP options case) case for redirects in ip_tryforward().
599 		 */
600 		if ((m = ip_tryforward(m)) == NULL)
601 			return;
602 		if (m->m_flags & M_FASTFWD_OURS) {
603 			m->m_flags &= ~M_FASTFWD_OURS;
604 			ip = mtod(m, struct ip *);
605 			goto ours;
606 		}
607 	}
608 
609 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
610 	/*
611 	 * Bypass packet filtering for packets previously handled by IPsec.
612 	 */
613 	if (IPSEC_ENABLED(ipv4) &&
614 	    IPSEC_CAPS(ipv4, m, IPSEC_CAP_BYPASS_FILTER) != 0)
615 			goto passin;
616 #endif
617 
618 	/*
619 	 * Run through list of hooks for input packets.
620 	 *
621 	 * NB: Beware of the destination address changing (e.g.
622 	 *     by NAT rewriting).  When this happens, tell
623 	 *     ip_forward to do the right thing.
624 	 */
625 
626 	/* Jump over all PFIL processing if hooks are not active. */
627 	if (!PFIL_HOOKED_IN(V_inet_pfil_head))
628 		goto passin;
629 
630 	odst = ip->ip_dst;
631 	if (pfil_mbuf_in(V_inet_pfil_head, &m, ifp, NULL) !=
632 	    PFIL_PASS)
633 		return;
634 
635 	ip = mtod(m, struct ip *);
636 	dchg = (odst.s_addr != ip->ip_dst.s_addr);
637 
638 	if (m->m_flags & M_FASTFWD_OURS) {
639 		m->m_flags &= ~M_FASTFWD_OURS;
640 		goto ours;
641 	}
642 	if (m->m_flags & M_IP_NEXTHOP) {
643 		if (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) {
644 			/*
645 			 * Directly ship the packet on.  This allows
646 			 * forwarding packets originally destined to us
647 			 * to some other directly connected host.
648 			 */
649 			ip_forward(m, 1);
650 			return;
651 		}
652 	}
653 passin:
654 	/*
655 	 * The unspecified address can appear only as a src address - RFC1122.
656 	 *
657 	 * The check is deferred to here to give firewalls a chance to block
658 	 * (and log) such packets.  ip_tryforward() will not process such
659 	 * packets.
660 	 */
661 	if (__predict_false(ntohl(ip->ip_dst.s_addr) == INADDR_ANY)) {
662 		IPSTAT_INC(ips_badaddr);
663 		goto bad;
664 	}
665 
666 	/*
667 	 * Process options and, if not destined for us,
668 	 * ship it on.  ip_dooptions returns 1 when an
669 	 * error was detected (causing an icmp message
670 	 * to be sent and the original packet to be freed).
671 	 */
672 	if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
673 		return;
674 
675         /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
676          * matter if it is destined to another node, or whether it is
677          * a multicast one, RSVP wants it! and prevents it from being forwarded
678          * anywhere else. Also checks if the rsvp daemon is running before
679 	 * grabbing the packet.
680          */
681 	if (ip->ip_p == IPPROTO_RSVP && V_rsvp_on)
682 		goto ours;
683 
684 	/*
685 	 * Check our list of addresses, to see if the packet is for us.
686 	 * If we don't have any addresses, assume any unicast packet
687 	 * we receive might be for us (and let the upper layers deal
688 	 * with it).
689 	 */
690 	if (CK_STAILQ_EMPTY(&V_in_ifaddrhead) &&
691 	    (m->m_flags & (M_MCAST|M_BCAST)) == 0)
692 		goto ours;
693 
694 	/*
695 	 * Enable a consistency check between the destination address
696 	 * and the arrival interface for a unicast packet (the RFC 1122
697 	 * strong ES model) with a list of additional predicates:
698 	 * - if IP forwarding is disabled
699 	 * - the packet is not locally generated
700 	 * - the packet is not subject to 'ipfw fwd'
701 	 * - Interface is not running CARP. If the packet got here, we already
702 	 *   checked it with carp_iamatch() and carp_forus().
703 	 */
704 	strong_es = V_ip_strong_es && (V_ipforwarding == 0) &&
705 	    ((ifp->if_flags & IFF_LOOPBACK) == 0) &&
706 	    ifp->if_carp == NULL && (dchg == 0);
707 
708 	/*
709 	 * Check for exact addresses in the hash bucket.
710 	 */
711 	CK_LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
712 		if (IA_SIN(ia)->sin_addr.s_addr != ip->ip_dst.s_addr)
713 			continue;
714 
715 		/*
716 		 * net.inet.ip.rfc1122_strong_es: the address matches, verify
717 		 * that the packet arrived via the correct interface.
718 		 */
719 		if (__predict_false(strong_es && ia->ia_ifp != ifp)) {
720 			IPSTAT_INC(ips_badaddr);
721 			goto bad;
722 		}
723 
724 		/*
725 		 * net.inet.ip.source_address_validation: drop incoming
726 		 * packets that pretend to be ours.
727 		 */
728 		if (V_ip_sav && !(ifp->if_flags & IFF_LOOPBACK) &&
729 		    __predict_false(in_localip_fib(ip->ip_src, ifp->if_fib))) {
730 			IPSTAT_INC(ips_badaddr);
731 			goto bad;
732 		}
733 
734 		counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
735 		counter_u64_add(ia->ia_ifa.ifa_ibytes, m->m_pkthdr.len);
736 		goto ours;
737 	}
738 
739 	/*
740 	 * Check for broadcast addresses.
741 	 *
742 	 * Only accept broadcast packets that arrive via the matching
743 	 * interface.  Reception of forwarded directed broadcasts would
744 	 * be handled via ip_forward() and ether_output() with the loopback
745 	 * into the stack for SIMPLEX interfaces handled by ether_output().
746 	 */
747 	if (ifp->if_flags & IFF_BROADCAST) {
748 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
749 			if (ifa->ifa_addr->sa_family != AF_INET)
750 				continue;
751 			ia = ifatoia(ifa);
752 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
753 			    ip->ip_dst.s_addr) {
754 				counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
755 				counter_u64_add(ia->ia_ifa.ifa_ibytes,
756 				    m->m_pkthdr.len);
757 				goto ours;
758 			}
759 #ifdef BOOTP_COMPAT
760 			if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) {
761 				counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
762 				counter_u64_add(ia->ia_ifa.ifa_ibytes,
763 				    m->m_pkthdr.len);
764 				goto ours;
765 			}
766 #endif
767 		}
768 		ia = NULL;
769 	}
770 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
771 		/*
772 		 * RFC 3927 2.7: Do not forward multicast packets from
773 		 * IN_LINKLOCAL.
774 		 */
775 		if (V_ip_mrouter && !IN_LINKLOCAL(ntohl(ip->ip_src.s_addr))) {
776 			/*
777 			 * If we are acting as a multicast router, all
778 			 * incoming multicast packets are passed to the
779 			 * kernel-level multicast forwarding function.
780 			 * The packet is returned (relatively) intact; if
781 			 * ip_mforward() returns a non-zero value, the packet
782 			 * must be discarded, else it may be accepted below.
783 			 */
784 			if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) {
785 				IPSTAT_INC(ips_cantforward);
786 				m_freem(m);
787 				return;
788 			}
789 
790 			/*
791 			 * The process-level routing daemon needs to receive
792 			 * all multicast IGMP packets, whether or not this
793 			 * host belongs to their destination groups.
794 			 */
795 			if (ip->ip_p == IPPROTO_IGMP) {
796 				goto ours;
797 			}
798 			IPSTAT_INC(ips_forward);
799 		}
800 		/*
801 		 * Assume the packet is for us, to avoid prematurely taking
802 		 * a lock on the in_multi hash. Protocols must perform
803 		 * their own filtering and update statistics accordingly.
804 		 */
805 		goto ours;
806 	}
807 	if (in_broadcast(ip->ip_dst))
808 		goto ours;
809 	/* RFC 3927 2.7: Do not forward packets to or from IN_LINKLOCAL. */
810 	if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr)) ||
811 	    IN_LINKLOCAL(ntohl(ip->ip_src.s_addr))) {
812 		IPSTAT_INC(ips_cantforward);
813 		m_freem(m);
814 		return;
815 	}
816 
817 	/*
818 	 * Not for us; forward if possible and desirable.
819 	 */
820 	if (V_ipforwarding == 0) {
821 		IPSTAT_INC(ips_cantforward);
822 		m_freem(m);
823 	} else {
824 		ip_forward(m, dchg);
825 	}
826 	return;
827 
828 ours:
829 #ifdef IPSTEALTH
830 	/*
831 	 * IPSTEALTH: Process non-routing options only
832 	 * if the packet is destined for us.
833 	 */
834 	if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1))
835 		return;
836 #endif /* IPSTEALTH */
837 
838 	/*
839 	 * We are going to ship the packet to the local protocol stack. Call the
840 	 * filter again for this 'output' action, allowing redirect-like rules
841 	 * to adjust the source address.
842 	 */
843 	if (PFIL_HOOKED_OUT(V_inet_local_pfil_head)) {
844 		if (pfil_mbuf_out(V_inet_local_pfil_head, &m, V_loif, NULL) !=
845 		    PFIL_PASS)
846 			return;
847 		ip = mtod(m, struct ip *);
848 	}
849 
850 	/*
851 	 * Attempt reassembly; if it succeeds, proceed.
852 	 * ip_reass() will return a different mbuf.
853 	 */
854 	if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) {
855 		/* XXXGL: shouldn't we save & set m_flags? */
856 		m = ip_reass(m);
857 		if (m == NULL)
858 			return;
859 		ip = mtod(m, struct ip *);
860 		/* Get the header length of the reassembled packet */
861 		hlen = ip->ip_hl << 2;
862 	}
863 
864 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
865 	if (IPSEC_ENABLED(ipv4)) {
866 		if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
867 			return;
868 	}
869 #endif /* IPSEC */
870 
871 	/*
872 	 * Switch out to protocol's input routine.
873 	 */
874 	IPSTAT_INC(ips_delivered);
875 
876 	ip_protox[ip->ip_p](&m, &hlen, ip->ip_p);
877 	return;
878 bad:
879 	m_freem(m);
880 }
881 
882 int
883 ipproto_register(uint8_t proto, ipproto_input_t input, ipproto_ctlinput_t ctl)
884 {
885 
886 	MPASS(proto > 0);
887 
888 	/*
889 	 * The protocol slot must not be occupied by another protocol
890 	 * already.  An index pointing to rip_input() is unused.
891 	 */
892 	if (ip_protox[proto] == rip_input) {
893 		ip_protox[proto] = input;
894 		ip_ctlprotox[proto] = ctl;
895 		return (0);
896 	} else
897 		return (EEXIST);
898 }
899 
900 int
901 ipproto_unregister(uint8_t proto)
902 {
903 
904 	MPASS(proto > 0);
905 
906 	if (ip_protox[proto] != rip_input) {
907 		ip_protox[proto] = rip_input;
908 		ip_ctlprotox[proto] = rip_ctlinput;
909 		return (0);
910 	} else
911 		return (ENOENT);
912 }
913 
914 /*
915  * Forward a packet.  If some error occurs return the sender
916  * an icmp packet.  Note we can't always generate a meaningful
917  * icmp message because icmp doesn't have a large enough repertoire
918  * of codes and types.
919  *
920  * If not forwarding, just drop the packet.  This could be confusing
921  * if ipforwarding was zero but some routing protocol was advancing
922  * us as a gateway to somewhere.  However, we must let the routing
923  * protocol deal with that.
924  *
925  * The srcrt parameter indicates whether the packet is being forwarded
926  * via a source route.
927  */
928 void
929 ip_forward(struct mbuf *m, int srcrt)
930 {
931 	struct ip *ip = mtod(m, struct ip *);
932 	struct in_ifaddr *ia;
933 	struct mbuf *mcopy;
934 	struct sockaddr_in *sin;
935 	struct in_addr dest;
936 	struct route ro;
937 	uint32_t flowid;
938 	int error, type = 0, code = 0, mtu = 0;
939 
940 	NET_EPOCH_ASSERT();
941 
942 	if (m->m_flags & (M_BCAST|M_MCAST) || !in_canforward(ip->ip_dst)) {
943 		IPSTAT_INC(ips_cantforward);
944 		m_freem(m);
945 		return;
946 	}
947 	if (
948 #ifdef IPSTEALTH
949 	    V_ipstealth == 0 &&
950 #endif
951 	    ip->ip_ttl <= IPTTLDEC) {
952 		icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 0, 0);
953 		return;
954 	}
955 
956 	bzero(&ro, sizeof(ro));
957 	sin = (struct sockaddr_in *)&ro.ro_dst;
958 	sin->sin_family = AF_INET;
959 	sin->sin_len = sizeof(*sin);
960 	sin->sin_addr = ip->ip_dst;
961 	flowid = m->m_pkthdr.flowid;
962 	ro.ro_nh = fib4_lookup(M_GETFIB(m), ip->ip_dst, 0, NHR_REF, flowid);
963 	if (ro.ro_nh != NULL) {
964 		if (ro.ro_nh->nh_flags & (NHF_BLACKHOLE | NHF_BROADCAST)) {
965 			IPSTAT_INC(ips_cantforward);
966 			m_freem(m);
967 			NH_FREE(ro.ro_nh);
968 			return;
969 		}
970 		if (ro.ro_nh->nh_flags & NHF_REJECT) {
971 			IPSTAT_INC(ips_cantforward);
972 			NH_FREE(ro.ro_nh);
973 			icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
974 			return;
975 		}
976 		ia = ifatoia(ro.ro_nh->nh_ifa);
977 	} else
978 		ia = NULL;
979 	/*
980 	 * Save the IP header and at most 8 bytes of the payload,
981 	 * in case we need to generate an ICMP message to the src.
982 	 *
983 	 * XXX this can be optimized a lot by saving the data in a local
984 	 * buffer on the stack (72 bytes at most), and only allocating the
985 	 * mbuf if really necessary. The vast majority of the packets
986 	 * are forwarded without having to send an ICMP back (either
987 	 * because unnecessary, or because rate limited), so we are
988 	 * really we are wasting a lot of work here.
989 	 *
990 	 * We don't use m_copym() because it might return a reference
991 	 * to a shared cluster. Both this function and ip_output()
992 	 * assume exclusive access to the IP header in `m', so any
993 	 * data in a cluster may change before we reach icmp_error().
994 	 */
995 	mcopy = m_gethdr(M_NOWAIT, m->m_type);
996 	if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_NOWAIT)) {
997 		/*
998 		 * It's probably ok if the pkthdr dup fails (because
999 		 * the deep copy of the tag chain failed), but for now
1000 		 * be conservative and just discard the copy since
1001 		 * code below may some day want the tags.
1002 		 */
1003 		m_free(mcopy);
1004 		mcopy = NULL;
1005 	}
1006 	if (mcopy != NULL) {
1007 		mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy));
1008 		mcopy->m_pkthdr.len = mcopy->m_len;
1009 		m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1010 	}
1011 #ifdef IPSTEALTH
1012 	if (V_ipstealth == 0)
1013 #endif
1014 		ip->ip_ttl -= IPTTLDEC;
1015 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1016 	if (IPSEC_ENABLED(ipv4)) {
1017 		if ((error = IPSEC_FORWARD(ipv4, m)) != 0) {
1018 			/* mbuf consumed by IPsec */
1019 			RO_NHFREE(&ro);
1020 			m_freem(mcopy);
1021 			if (error != EINPROGRESS)
1022 				IPSTAT_INC(ips_cantforward);
1023 			return;
1024 		}
1025 		/* No IPsec processing required */
1026 	}
1027 #endif /* IPSEC */
1028 	/*
1029 	 * If forwarding packet using same interface that it came in on,
1030 	 * perhaps should send a redirect to sender to shortcut a hop.
1031 	 * Only send redirect if source is sending directly to us,
1032 	 * and if packet was not source routed (or has any options).
1033 	 * Also, don't send redirect if forwarding using a default route
1034 	 * or a route modified by a redirect.
1035 	 */
1036 	dest.s_addr = 0;
1037 	if (!srcrt && V_ipsendredirects &&
1038 	    ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) {
1039 		struct nhop_object *nh;
1040 
1041 		nh = ro.ro_nh;
1042 
1043 		if (nh != NULL && ((nh->nh_flags & (NHF_REDIRECT|NHF_DEFAULT)) == 0)) {
1044 			struct in_ifaddr *nh_ia = (struct in_ifaddr *)(nh->nh_ifa);
1045 			u_long src = ntohl(ip->ip_src.s_addr);
1046 
1047 			if (nh_ia != NULL &&
1048 			    (src & nh_ia->ia_subnetmask) == nh_ia->ia_subnet) {
1049 				/* Router requirements says to only send host redirects */
1050 				type = ICMP_REDIRECT;
1051 				code = ICMP_REDIRECT_HOST;
1052 				if (nh->nh_flags & NHF_GATEWAY) {
1053 				    if (nh->gw_sa.sa_family == AF_INET)
1054 					dest.s_addr = nh->gw4_sa.sin_addr.s_addr;
1055 				    else /* Do not redirect in case gw is AF_INET6 */
1056 					type = 0;
1057 				} else
1058 					dest.s_addr = ip->ip_dst.s_addr;
1059 			}
1060 		}
1061 	}
1062 
1063 	error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
1064 
1065 	if (error == EMSGSIZE && ro.ro_nh)
1066 		mtu = ro.ro_nh->nh_mtu;
1067 	RO_NHFREE(&ro);
1068 
1069 	if (error)
1070 		IPSTAT_INC(ips_cantforward);
1071 	else {
1072 		IPSTAT_INC(ips_forward);
1073 		if (type)
1074 			IPSTAT_INC(ips_redirectsent);
1075 		else {
1076 			if (mcopy)
1077 				m_freem(mcopy);
1078 			return;
1079 		}
1080 	}
1081 	if (mcopy == NULL)
1082 		return;
1083 
1084 	switch (error) {
1085 	case 0:				/* forwarded, but need redirect */
1086 		/* type, code set above */
1087 		break;
1088 
1089 	case ENETUNREACH:
1090 	case EHOSTUNREACH:
1091 	case ENETDOWN:
1092 	case EHOSTDOWN:
1093 	default:
1094 		type = ICMP_UNREACH;
1095 		code = ICMP_UNREACH_HOST;
1096 		break;
1097 
1098 	case EMSGSIZE:
1099 		type = ICMP_UNREACH;
1100 		code = ICMP_UNREACH_NEEDFRAG;
1101 		/*
1102 		 * If the MTU was set before make sure we are below the
1103 		 * interface MTU.
1104 		 * If the MTU wasn't set before use the interface mtu or
1105 		 * fall back to the next smaller mtu step compared to the
1106 		 * current packet size.
1107 		 */
1108 		if (mtu != 0) {
1109 			if (ia != NULL)
1110 				mtu = min(mtu, ia->ia_ifp->if_mtu);
1111 		} else {
1112 			if (ia != NULL)
1113 				mtu = ia->ia_ifp->if_mtu;
1114 			else
1115 				mtu = ip_next_mtu(ntohs(ip->ip_len), 0);
1116 		}
1117 		IPSTAT_INC(ips_cantfrag);
1118 		break;
1119 
1120 	case ENOBUFS:
1121 	case EACCES:			/* ipfw denied packet */
1122 		m_freem(mcopy);
1123 		return;
1124 	}
1125 	icmp_error(mcopy, type, code, dest.s_addr, mtu);
1126 }
1127 
1128 #define	CHECK_SO_CT(sp, ct) \
1129     (((sp->so_options & SO_TIMESTAMP) && (sp->so_ts_clock == ct)) ? 1 : 0)
1130 
1131 void
1132 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1133     struct mbuf *m)
1134 {
1135 	bool stamped;
1136 
1137 	stamped = false;
1138 	if ((inp->inp_socket->so_options & SO_BINTIME) ||
1139 	    CHECK_SO_CT(inp->inp_socket, SO_TS_BINTIME)) {
1140 		struct bintime boottimebin, bt;
1141 		struct timespec ts1;
1142 
1143 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1144 		    M_TSTMP)) {
1145 			mbuf_tstmp2timespec(m, &ts1);
1146 			timespec2bintime(&ts1, &bt);
1147 			getboottimebin(&boottimebin);
1148 			bintime_add(&bt, &boottimebin);
1149 		} else {
1150 			bintime(&bt);
1151 		}
1152 		*mp = sbcreatecontrol(&bt, sizeof(bt), SCM_BINTIME,
1153 		    SOL_SOCKET, M_NOWAIT);
1154 		if (*mp != NULL) {
1155 			mp = &(*mp)->m_next;
1156 			stamped = true;
1157 		}
1158 	}
1159 	if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME_MICRO)) {
1160 		struct bintime boottimebin, bt1;
1161 		struct timespec ts1;
1162 		struct timeval tv;
1163 
1164 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1165 		    M_TSTMP)) {
1166 			mbuf_tstmp2timespec(m, &ts1);
1167 			timespec2bintime(&ts1, &bt1);
1168 			getboottimebin(&boottimebin);
1169 			bintime_add(&bt1, &boottimebin);
1170 			bintime2timeval(&bt1, &tv);
1171 		} else {
1172 			microtime(&tv);
1173 		}
1174 		*mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv), SCM_TIMESTAMP,
1175 		    SOL_SOCKET, M_NOWAIT);
1176 		if (*mp != NULL) {
1177 			mp = &(*mp)->m_next;
1178 			stamped = true;
1179 		}
1180 	} else if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME)) {
1181 		struct bintime boottimebin;
1182 		struct timespec ts, ts1;
1183 
1184 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1185 		    M_TSTMP)) {
1186 			mbuf_tstmp2timespec(m, &ts);
1187 			getboottimebin(&boottimebin);
1188 			bintime2timespec(&boottimebin, &ts1);
1189 			timespecadd(&ts, &ts1, &ts);
1190 		} else {
1191 			nanotime(&ts);
1192 		}
1193 		*mp = sbcreatecontrol(&ts, sizeof(ts), SCM_REALTIME,
1194 		    SOL_SOCKET, M_NOWAIT);
1195 		if (*mp != NULL) {
1196 			mp = &(*mp)->m_next;
1197 			stamped = true;
1198 		}
1199 	} else if (CHECK_SO_CT(inp->inp_socket, SO_TS_MONOTONIC)) {
1200 		struct timespec ts;
1201 
1202 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1203 		    M_TSTMP))
1204 			mbuf_tstmp2timespec(m, &ts);
1205 		else
1206 			nanouptime(&ts);
1207 		*mp = sbcreatecontrol(&ts, sizeof(ts), SCM_MONOTONIC,
1208 		    SOL_SOCKET, M_NOWAIT);
1209 		if (*mp != NULL) {
1210 			mp = &(*mp)->m_next;
1211 			stamped = true;
1212 		}
1213 	}
1214 	if (stamped && (m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1215 	    M_TSTMP)) {
1216 		struct sock_timestamp_info sti;
1217 
1218 		bzero(&sti, sizeof(sti));
1219 		sti.st_info_flags = ST_INFO_HW;
1220 		if ((m->m_flags & M_TSTMP_HPREC) != 0)
1221 			sti.st_info_flags |= ST_INFO_HW_HPREC;
1222 		*mp = sbcreatecontrol(&sti, sizeof(sti), SCM_TIME_INFO,
1223 		    SOL_SOCKET, M_NOWAIT);
1224 		if (*mp != NULL)
1225 			mp = &(*mp)->m_next;
1226 	}
1227 	if (inp->inp_flags & INP_RECVDSTADDR) {
1228 		*mp = sbcreatecontrol(&ip->ip_dst, sizeof(struct in_addr),
1229 		    IP_RECVDSTADDR, IPPROTO_IP, M_NOWAIT);
1230 		if (*mp)
1231 			mp = &(*mp)->m_next;
1232 	}
1233 	if (inp->inp_flags & INP_RECVTTL) {
1234 		*mp = sbcreatecontrol(&ip->ip_ttl, sizeof(u_char), IP_RECVTTL,
1235 		    IPPROTO_IP, M_NOWAIT);
1236 		if (*mp)
1237 			mp = &(*mp)->m_next;
1238 	}
1239 #ifdef notyet
1240 	/* XXX
1241 	 * Moving these out of udp_input() made them even more broken
1242 	 * than they already were.
1243 	 */
1244 	/* options were tossed already */
1245 	if (inp->inp_flags & INP_RECVOPTS) {
1246 		*mp = sbcreatecontrol(opts_deleted_above,
1247 		    sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP, M_NOWAIT);
1248 		if (*mp)
1249 			mp = &(*mp)->m_next;
1250 	}
1251 	/* ip_srcroute doesn't do what we want here, need to fix */
1252 	if (inp->inp_flags & INP_RECVRETOPTS) {
1253 		*mp = sbcreatecontrol(ip_srcroute(m), sizeof(struct in_addr),
1254 		    IP_RECVRETOPTS, IPPROTO_IP, M_NOWAIT);
1255 		if (*mp)
1256 			mp = &(*mp)->m_next;
1257 	}
1258 #endif
1259 	if (inp->inp_flags & INP_RECVIF) {
1260 		struct ifnet *ifp;
1261 		struct sdlbuf {
1262 			struct sockaddr_dl sdl;
1263 			u_char	pad[32];
1264 		} sdlbuf;
1265 		struct sockaddr_dl *sdp;
1266 		struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1267 
1268 		if ((ifp = m->m_pkthdr.rcvif)) {
1269 			sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1270 			/*
1271 			 * Change our mind and don't try copy.
1272 			 */
1273 			if (sdp->sdl_family != AF_LINK ||
1274 			    sdp->sdl_len > sizeof(sdlbuf)) {
1275 				goto makedummy;
1276 			}
1277 			bcopy(sdp, sdl2, sdp->sdl_len);
1278 		} else {
1279 makedummy:
1280 			sdl2->sdl_len =
1281 			    offsetof(struct sockaddr_dl, sdl_data[0]);
1282 			sdl2->sdl_family = AF_LINK;
1283 			sdl2->sdl_index = 0;
1284 			sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1285 		}
1286 		*mp = sbcreatecontrol(sdl2, sdl2->sdl_len, IP_RECVIF,
1287 		    IPPROTO_IP, M_NOWAIT);
1288 		if (*mp)
1289 			mp = &(*mp)->m_next;
1290 	}
1291 	if (inp->inp_flags & INP_RECVTOS) {
1292 		*mp = sbcreatecontrol(&ip->ip_tos, sizeof(u_char), IP_RECVTOS,
1293 		    IPPROTO_IP, M_NOWAIT);
1294 		if (*mp)
1295 			mp = &(*mp)->m_next;
1296 	}
1297 
1298 	if (inp->inp_flags2 & INP_RECVFLOWID) {
1299 		uint32_t flowid, flow_type;
1300 
1301 		flowid = m->m_pkthdr.flowid;
1302 		flow_type = M_HASHTYPE_GET(m);
1303 
1304 		/*
1305 		 * XXX should handle the failure of one or the
1306 		 * other - don't populate both?
1307 		 */
1308 		*mp = sbcreatecontrol(&flowid, sizeof(uint32_t), IP_FLOWID,
1309 		    IPPROTO_IP, M_NOWAIT);
1310 		if (*mp)
1311 			mp = &(*mp)->m_next;
1312 		*mp = sbcreatecontrol(&flow_type, sizeof(uint32_t),
1313 		    IP_FLOWTYPE, IPPROTO_IP, M_NOWAIT);
1314 		if (*mp)
1315 			mp = &(*mp)->m_next;
1316 	}
1317 
1318 #ifdef	RSS
1319 	if (inp->inp_flags2 & INP_RECVRSSBUCKETID) {
1320 		uint32_t flowid, flow_type;
1321 		uint32_t rss_bucketid;
1322 
1323 		flowid = m->m_pkthdr.flowid;
1324 		flow_type = M_HASHTYPE_GET(m);
1325 
1326 		if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) {
1327 			*mp = sbcreatecontrol(&rss_bucketid, sizeof(uint32_t),
1328 			    IP_RSSBUCKETID, IPPROTO_IP, M_NOWAIT);
1329 			if (*mp)
1330 				mp = &(*mp)->m_next;
1331 		}
1332 	}
1333 #endif
1334 }
1335 
1336 /*
1337  * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
1338  * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
1339  * locking.  This code remains in ip_input.c as ip_mroute.c is optionally
1340  * compiled.
1341  */
1342 VNET_DEFINE_STATIC(int, ip_rsvp_on);
1343 VNET_DEFINE(struct socket *, ip_rsvpd);
1344 
1345 #define	V_ip_rsvp_on		VNET(ip_rsvp_on)
1346 
1347 int
1348 ip_rsvp_init(struct socket *so)
1349 {
1350 
1351 	if (V_ip_rsvpd != NULL)
1352 		return EADDRINUSE;
1353 
1354 	V_ip_rsvpd = so;
1355 	/*
1356 	 * This may seem silly, but we need to be sure we don't over-increment
1357 	 * the RSVP counter, in case something slips up.
1358 	 */
1359 	if (!V_ip_rsvp_on) {
1360 		V_ip_rsvp_on = 1;
1361 		V_rsvp_on++;
1362 	}
1363 
1364 	return 0;
1365 }
1366 
1367 int
1368 ip_rsvp_done(void)
1369 {
1370 
1371 	V_ip_rsvpd = NULL;
1372 	/*
1373 	 * This may seem silly, but we need to be sure we don't over-decrement
1374 	 * the RSVP counter, in case something slips up.
1375 	 */
1376 	if (V_ip_rsvp_on) {
1377 		V_ip_rsvp_on = 0;
1378 		V_rsvp_on--;
1379 	}
1380 	return 0;
1381 }
1382 
1383 int
1384 rsvp_input(struct mbuf **mp, int *offp, int proto)
1385 {
1386 	struct mbuf *m;
1387 
1388 	m = *mp;
1389 	*mp = NULL;
1390 
1391 	if (rsvp_input_p) { /* call the real one if loaded */
1392 		*mp = m;
1393 		rsvp_input_p(mp, offp, proto);
1394 		return (IPPROTO_DONE);
1395 	}
1396 
1397 	/* Can still get packets with rsvp_on = 0 if there is a local member
1398 	 * of the group to which the RSVP packet is addressed.  But in this
1399 	 * case we want to throw the packet away.
1400 	 */
1401 
1402 	if (!V_rsvp_on) {
1403 		m_freem(m);
1404 		return (IPPROTO_DONE);
1405 	}
1406 
1407 	if (V_ip_rsvpd != NULL) {
1408 		*mp = m;
1409 		rip_input(mp, offp, proto);
1410 		return (IPPROTO_DONE);
1411 	}
1412 	/* Drop the packet */
1413 	m_freem(m);
1414 	return (IPPROTO_DONE);
1415 }
1416