xref: /freebsd/sys/netinet/ip_input.c (revision 6580f5c38dd5b01aeeaed16b370f1a12423437f0)
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 <sys/cdefs.h>
33 #include "opt_bootp.h"
34 #include "opt_inet.h"
35 #include "opt_ipstealth.h"
36 #include "opt_ipsec.h"
37 #include "opt_route.h"
38 #include "opt_rss.h"
39 #include "opt_sctp.h"
40 
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/hhook.h>
44 #include <sys/mbuf.h>
45 #include <sys/malloc.h>
46 #include <sys/domain.h>
47 #include <sys/protosw.h>
48 #include <sys/socket.h>
49 #include <sys/time.h>
50 #include <sys/kernel.h>
51 #include <sys/lock.h>
52 #include <sys/rmlock.h>
53 #include <sys/rwlock.h>
54 #include <sys/sdt.h>
55 #include <sys/syslog.h>
56 #include <sys/sysctl.h>
57 
58 #include <net/if.h>
59 #include <net/if_types.h>
60 #include <net/if_var.h>
61 #include <net/if_dl.h>
62 #include <net/if_private.h>
63 #include <net/pfil.h>
64 #include <net/route.h>
65 #include <net/route/nhop.h>
66 #include <net/netisr.h>
67 #include <net/rss_config.h>
68 #include <net/vnet.h>
69 
70 #include <netinet/in.h>
71 #include <netinet/in_kdtrace.h>
72 #include <netinet/in_systm.h>
73 #include <netinet/in_var.h>
74 #include <netinet/ip.h>
75 #include <netinet/in_fib.h>
76 #include <netinet/in_pcb.h>
77 #include <netinet/ip_var.h>
78 #include <netinet/ip_encap.h>
79 #include <netinet/ip_fw.h>
80 #include <netinet/ip_icmp.h>
81 #include <netinet/igmp_var.h>
82 #include <netinet/ip_options.h>
83 #include <machine/in_cksum.h>
84 #include <netinet/ip_carp.h>
85 #include <netinet/in_rss.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 
357 	ipreass_init();
358 
359 	/*
360 	 * Register statically compiled protocols, that are unlikely to
361 	 * ever become dynamic.
362 	 */
363 	IPPROTO_REGISTER(IPPROTO_ICMP, icmp_input, NULL);
364 	IPPROTO_REGISTER(IPPROTO_IGMP, igmp_input, NULL);
365 	IPPROTO_REGISTER(IPPROTO_RSVP, rsvp_input, NULL);
366 	IPPROTO_REGISTER(IPPROTO_IPV4, encap4_input, NULL);
367 	IPPROTO_REGISTER(IPPROTO_MOBILE, encap4_input, NULL);
368 	IPPROTO_REGISTER(IPPROTO_ETHERIP, encap4_input, NULL);
369 	IPPROTO_REGISTER(IPPROTO_GRE, encap4_input, NULL);
370 	IPPROTO_REGISTER(IPPROTO_IPV6, encap4_input, NULL);
371 	IPPROTO_REGISTER(IPPROTO_PIM, encap4_input, NULL);
372 #ifdef SCTP	/* XXX: has a loadable & static version */
373 	IPPROTO_REGISTER(IPPROTO_SCTP, sctp_input, sctp_ctlinput);
374 #endif
375 
376 	netisr_register(&ip_nh);
377 #ifdef	RSS
378 	netisr_register(&ip_direct_nh);
379 #endif
380 }
381 SYSINIT(ip_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip_init, NULL);
382 
383 #ifdef VIMAGE
384 static void
385 ip_destroy(void *unused __unused)
386 {
387 	int error;
388 
389 #ifdef	RSS
390 	netisr_unregister_vnet(&ip_direct_nh);
391 #endif
392 	netisr_unregister_vnet(&ip_nh);
393 
394 	pfil_head_unregister(V_inet_pfil_head);
395 	error = hhook_head_deregister(V_ipsec_hhh_in[HHOOK_IPSEC_INET]);
396 	if (error != 0) {
397 		printf("%s: WARNING: unable to deregister input helper hook "
398 		    "type HHOOK_TYPE_IPSEC_IN, id HHOOK_IPSEC_INET: "
399 		    "error %d returned\n", __func__, error);
400 	}
401 	error = hhook_head_deregister(V_ipsec_hhh_out[HHOOK_IPSEC_INET]);
402 	if (error != 0) {
403 		printf("%s: WARNING: unable to deregister output helper hook "
404 		    "type HHOOK_TYPE_IPSEC_OUT, id HHOOK_IPSEC_INET: "
405 		    "error %d returned\n", __func__, error);
406 	}
407 
408 	/* Remove the IPv4 addresses from all interfaces. */
409 	in_ifscrub_all();
410 
411 	/* Make sure the IPv4 routes are gone as well. */
412 	rib_flush_routes_family(AF_INET);
413 
414 	/* Destroy IP reassembly queue. */
415 	ipreass_destroy();
416 
417 	/* Cleanup in_ifaddr hash table; should be empty. */
418 	hashdestroy(V_in_ifaddrhashtbl, M_IFADDR, V_in_ifaddrhmask);
419 }
420 
421 VNET_SYSUNINIT(ip, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip_destroy, NULL);
422 #endif
423 
424 #ifdef	RSS
425 /*
426  * IP direct input routine.
427  *
428  * This is called when reinjecting completed fragments where
429  * all of the previous checking and book-keeping has been done.
430  */
431 void
432 ip_direct_input(struct mbuf *m)
433 {
434 	struct ip *ip;
435 	int hlen;
436 
437 	ip = mtod(m, struct ip *);
438 	hlen = ip->ip_hl << 2;
439 
440 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
441 	if (IPSEC_ENABLED(ipv4)) {
442 		if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
443 			return;
444 	}
445 #endif /* IPSEC */
446 	IPSTAT_INC(ips_delivered);
447 	ip_protox[ip->ip_p](&m, &hlen, ip->ip_p);
448 }
449 #endif
450 
451 /*
452  * Ip input routine.  Checksum and byte swap header.  If fragmented
453  * try to reassemble.  Process options.  Pass to next level.
454  */
455 void
456 ip_input(struct mbuf *m)
457 {
458 	struct ip *ip = NULL;
459 	struct in_ifaddr *ia = NULL;
460 	struct ifaddr *ifa;
461 	struct ifnet *ifp;
462 	int hlen = 0;
463 	uint16_t sum, ip_len;
464 	int dchg = 0;				/* dest changed after fw */
465 	struct in_addr odst;			/* original dst address */
466 	bool strong_es;
467 
468 	M_ASSERTPKTHDR(m);
469 	NET_EPOCH_ASSERT();
470 
471 	if (m->m_flags & M_FASTFWD_OURS) {
472 		m->m_flags &= ~M_FASTFWD_OURS;
473 		/* Set up some basics that will be used later. */
474 		ip = mtod(m, struct ip *);
475 		hlen = ip->ip_hl << 2;
476 		ip_len = ntohs(ip->ip_len);
477 		goto ours;
478 	}
479 
480 	IPSTAT_INC(ips_total);
481 
482 	if (__predict_false(m->m_pkthdr.len < sizeof(struct ip)))
483 		goto tooshort;
484 
485 	if (m->m_len < sizeof(struct ip)) {
486 		m = m_pullup(m, sizeof(struct ip));
487 		if (__predict_false(m == NULL)) {
488 			IPSTAT_INC(ips_toosmall);
489 			return;
490 		}
491 	}
492 	ip = mtod(m, struct ip *);
493 
494 	if (__predict_false(ip->ip_v != IPVERSION)) {
495 		IPSTAT_INC(ips_badvers);
496 		goto bad;
497 	}
498 
499 	hlen = ip->ip_hl << 2;
500 	if (__predict_false(hlen < sizeof(struct ip))) {	/* minimum header length */
501 		IPSTAT_INC(ips_badhlen);
502 		goto bad;
503 	}
504 	if (hlen > m->m_len) {
505 		m = m_pullup(m, hlen);
506 		if (__predict_false(m == NULL)) {
507 			IPSTAT_INC(ips_badhlen);
508 			return;
509 		}
510 		ip = mtod(m, struct ip *);
511 	}
512 
513 	IP_PROBE(receive, NULL, NULL, ip, m->m_pkthdr.rcvif, ip, NULL);
514 
515 	/* IN_LOOPBACK must not appear on the wire - RFC1122 */
516 	ifp = m->m_pkthdr.rcvif;
517 	if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) ||
518 	    IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) {
519 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
520 			IPSTAT_INC(ips_badaddr);
521 			goto bad;
522 		}
523 	}
524 	/* The unspecified address can appear only as a src address - RFC1122 */
525 	if (__predict_false(ntohl(ip->ip_dst.s_addr) == INADDR_ANY)) {
526 		IPSTAT_INC(ips_badaddr);
527 		goto bad;
528 	}
529 
530 	if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
531 		sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
532 	} else {
533 		if (hlen == sizeof(struct ip)) {
534 			sum = in_cksum_hdr(ip);
535 		} else {
536 			sum = in_cksum(m, hlen);
537 		}
538 	}
539 	if (__predict_false(sum)) {
540 		IPSTAT_INC(ips_badsum);
541 		goto bad;
542 	}
543 
544 	ip_len = ntohs(ip->ip_len);
545 	if (__predict_false(ip_len < hlen)) {
546 		IPSTAT_INC(ips_badlen);
547 		goto bad;
548 	}
549 
550 	/*
551 	 * Check that the amount of data in the buffers
552 	 * is as at least much as the IP header would have us expect.
553 	 * Trim mbufs if longer than we expect.
554 	 * Drop packet if shorter than we expect.
555 	 */
556 	if (__predict_false(m->m_pkthdr.len < ip_len)) {
557 tooshort:
558 		IPSTAT_INC(ips_tooshort);
559 		goto bad;
560 	}
561 	if (m->m_pkthdr.len > ip_len) {
562 		if (m->m_len == m->m_pkthdr.len) {
563 			m->m_len = ip_len;
564 			m->m_pkthdr.len = ip_len;
565 		} else
566 			m_adj(m, ip_len - m->m_pkthdr.len);
567 	}
568 
569 	/*
570 	 * Try to forward the packet, but if we fail continue.
571 	 * ip_tryforward() may generate redirects these days.
572 	 * XXX the logic below falling through to normal processing
573 	 * if redirects are required should be revisited as well.
574 	 * ip_tryforward() does inbound and outbound packet firewall
575 	 * processing. If firewall has decided that destination becomes
576 	 * our local address, it sets M_FASTFWD_OURS flag. In this
577 	 * case skip another inbound firewall processing and update
578 	 * ip pointer.
579 	 */
580 	if (V_ipforwarding != 0
581 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
582 	    && (!IPSEC_ENABLED(ipv4) ||
583 	    IPSEC_CAPS(ipv4, m, IPSEC_CAP_OPERABLE) == 0)
584 #endif
585 	    ) {
586 		/*
587 		 * ip_dooptions() was run so we can ignore the source route (or
588 		 * any IP options case) case for redirects in ip_tryforward().
589 		 */
590 		if ((m = ip_tryforward(m)) == NULL)
591 			return;
592 		if (m->m_flags & M_FASTFWD_OURS) {
593 			m->m_flags &= ~M_FASTFWD_OURS;
594 			ip = mtod(m, struct ip *);
595 			goto ours;
596 		}
597 	}
598 
599 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
600 	/*
601 	 * Bypass packet filtering for packets previously handled by IPsec.
602 	 */
603 	if (IPSEC_ENABLED(ipv4) &&
604 	    IPSEC_CAPS(ipv4, m, IPSEC_CAP_BYPASS_FILTER) != 0)
605 			goto passin;
606 #endif
607 
608 	/*
609 	 * Run through list of hooks for input packets.
610 	 *
611 	 * NB: Beware of the destination address changing (e.g.
612 	 *     by NAT rewriting).  When this happens, tell
613 	 *     ip_forward to do the right thing.
614 	 */
615 
616 	/* Jump over all PFIL processing if hooks are not active. */
617 	if (!PFIL_HOOKED_IN(V_inet_pfil_head))
618 		goto passin;
619 
620 	odst = ip->ip_dst;
621 	if (pfil_mbuf_in(V_inet_pfil_head, &m, ifp, NULL) !=
622 	    PFIL_PASS)
623 		return;
624 
625 	ip = mtod(m, struct ip *);
626 	dchg = (odst.s_addr != ip->ip_dst.s_addr);
627 
628 	if (m->m_flags & M_FASTFWD_OURS) {
629 		m->m_flags &= ~M_FASTFWD_OURS;
630 		goto ours;
631 	}
632 	if (m->m_flags & M_IP_NEXTHOP) {
633 		if (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) {
634 			/*
635 			 * Directly ship the packet on.  This allows
636 			 * forwarding packets originally destined to us
637 			 * to some other directly connected host.
638 			 */
639 			ip_forward(m, 1);
640 			return;
641 		}
642 	}
643 passin:
644 
645 	/*
646 	 * Process options and, if not destined for us,
647 	 * ship it on.  ip_dooptions returns 1 when an
648 	 * error was detected (causing an icmp message
649 	 * to be sent and the original packet to be freed).
650 	 */
651 	if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
652 		return;
653 
654         /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
655          * matter if it is destined to another node, or whether it is
656          * a multicast one, RSVP wants it! and prevents it from being forwarded
657          * anywhere else. Also checks if the rsvp daemon is running before
658 	 * grabbing the packet.
659          */
660 	if (ip->ip_p == IPPROTO_RSVP && V_rsvp_on)
661 		goto ours;
662 
663 	/*
664 	 * Check our list of addresses, to see if the packet is for us.
665 	 * If we don't have any addresses, assume any unicast packet
666 	 * we receive might be for us (and let the upper layers deal
667 	 * with it).
668 	 */
669 	if (CK_STAILQ_EMPTY(&V_in_ifaddrhead) &&
670 	    (m->m_flags & (M_MCAST|M_BCAST)) == 0)
671 		goto ours;
672 
673 	/*
674 	 * Enable a consistency check between the destination address
675 	 * and the arrival interface for a unicast packet (the RFC 1122
676 	 * strong ES model) with a list of additional predicates:
677 	 * - if IP forwarding is disabled
678 	 * - the packet is not locally generated
679 	 * - the packet is not subject to 'ipfw fwd'
680 	 * - Interface is not running CARP. If the packet got here, we already
681 	 *   checked it with carp_iamatch() and carp_forus().
682 	 */
683 	strong_es = V_ip_strong_es && (V_ipforwarding == 0) &&
684 	    ((ifp->if_flags & IFF_LOOPBACK) == 0) &&
685 	    ifp->if_carp == NULL && (dchg == 0);
686 
687 	/*
688 	 * Check for exact addresses in the hash bucket.
689 	 */
690 	CK_LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
691 		if (IA_SIN(ia)->sin_addr.s_addr != ip->ip_dst.s_addr)
692 			continue;
693 
694 		/*
695 		 * net.inet.ip.rfc1122_strong_es: the address matches, verify
696 		 * that the packet arrived via the correct interface.
697 		 */
698 		if (__predict_false(strong_es && ia->ia_ifp != ifp)) {
699 			IPSTAT_INC(ips_badaddr);
700 			goto bad;
701 		}
702 
703 		/*
704 		 * net.inet.ip.source_address_validation: drop incoming
705 		 * packets that pretend to be ours.
706 		 */
707 		if (V_ip_sav && !(ifp->if_flags & IFF_LOOPBACK) &&
708 		    __predict_false(in_localip_fib(ip->ip_src, ifp->if_fib))) {
709 			IPSTAT_INC(ips_badaddr);
710 			goto bad;
711 		}
712 
713 		counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
714 		counter_u64_add(ia->ia_ifa.ifa_ibytes, m->m_pkthdr.len);
715 		goto ours;
716 	}
717 
718 	/*
719 	 * Check for broadcast addresses.
720 	 *
721 	 * Only accept broadcast packets that arrive via the matching
722 	 * interface.  Reception of forwarded directed broadcasts would
723 	 * be handled via ip_forward() and ether_output() with the loopback
724 	 * into the stack for SIMPLEX interfaces handled by ether_output().
725 	 */
726 	if (ifp->if_flags & IFF_BROADCAST) {
727 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
728 			if (ifa->ifa_addr->sa_family != AF_INET)
729 				continue;
730 			ia = ifatoia(ifa);
731 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
732 			    ip->ip_dst.s_addr) {
733 				counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
734 				counter_u64_add(ia->ia_ifa.ifa_ibytes,
735 				    m->m_pkthdr.len);
736 				goto ours;
737 			}
738 #ifdef BOOTP_COMPAT
739 			if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) {
740 				counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
741 				counter_u64_add(ia->ia_ifa.ifa_ibytes,
742 				    m->m_pkthdr.len);
743 				goto ours;
744 			}
745 #endif
746 		}
747 		ia = NULL;
748 	}
749 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
750 		/*
751 		 * RFC 3927 2.7: Do not forward multicast packets from
752 		 * IN_LINKLOCAL.
753 		 */
754 		if (V_ip_mrouter && !IN_LINKLOCAL(ntohl(ip->ip_src.s_addr))) {
755 			/*
756 			 * If we are acting as a multicast router, all
757 			 * incoming multicast packets are passed to the
758 			 * kernel-level multicast forwarding function.
759 			 * The packet is returned (relatively) intact; if
760 			 * ip_mforward() returns a non-zero value, the packet
761 			 * must be discarded, else it may be accepted below.
762 			 */
763 			if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) {
764 				IPSTAT_INC(ips_cantforward);
765 				m_freem(m);
766 				return;
767 			}
768 
769 			/*
770 			 * The process-level routing daemon needs to receive
771 			 * all multicast IGMP packets, whether or not this
772 			 * host belongs to their destination groups.
773 			 */
774 			if (ip->ip_p == IPPROTO_IGMP) {
775 				goto ours;
776 			}
777 			IPSTAT_INC(ips_forward);
778 		}
779 		/*
780 		 * Assume the packet is for us, to avoid prematurely taking
781 		 * a lock on the in_multi hash. Protocols must perform
782 		 * their own filtering and update statistics accordingly.
783 		 */
784 		goto ours;
785 	}
786 	if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
787 		goto ours;
788 	if (ip->ip_dst.s_addr == INADDR_ANY)
789 		goto ours;
790 	/* RFC 3927 2.7: Do not forward packets to or from IN_LINKLOCAL. */
791 	if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr)) ||
792 	    IN_LINKLOCAL(ntohl(ip->ip_src.s_addr))) {
793 		IPSTAT_INC(ips_cantforward);
794 		m_freem(m);
795 		return;
796 	}
797 
798 	/*
799 	 * Not for us; forward if possible and desirable.
800 	 */
801 	if (V_ipforwarding == 0) {
802 		IPSTAT_INC(ips_cantforward);
803 		m_freem(m);
804 	} else {
805 		ip_forward(m, dchg);
806 	}
807 	return;
808 
809 ours:
810 #ifdef IPSTEALTH
811 	/*
812 	 * IPSTEALTH: Process non-routing options only
813 	 * if the packet is destined for us.
814 	 */
815 	if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1))
816 		return;
817 #endif /* IPSTEALTH */
818 
819 	/*
820 	 * We are going to ship the packet to the local protocol stack. Call the
821 	 * filter again for this 'output' action, allowing redirect-like rules
822 	 * to adjust the source address.
823 	 */
824 	if (PFIL_HOOKED_OUT(V_inet_local_pfil_head)) {
825 		if (pfil_mbuf_out(V_inet_local_pfil_head, &m, V_loif, NULL) !=
826 		    PFIL_PASS)
827 			return;
828 		ip = mtod(m, struct ip *);
829 	}
830 
831 	/*
832 	 * Attempt reassembly; if it succeeds, proceed.
833 	 * ip_reass() will return a different mbuf.
834 	 */
835 	if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) {
836 		/* XXXGL: shouldn't we save & set m_flags? */
837 		m = ip_reass(m);
838 		if (m == NULL)
839 			return;
840 		ip = mtod(m, struct ip *);
841 		/* Get the header length of the reassembled packet */
842 		hlen = ip->ip_hl << 2;
843 	}
844 
845 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
846 	if (IPSEC_ENABLED(ipv4)) {
847 		if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
848 			return;
849 	}
850 #endif /* IPSEC */
851 
852 	/*
853 	 * Switch out to protocol's input routine.
854 	 */
855 	IPSTAT_INC(ips_delivered);
856 
857 	ip_protox[ip->ip_p](&m, &hlen, ip->ip_p);
858 	return;
859 bad:
860 	m_freem(m);
861 }
862 
863 int
864 ipproto_register(uint8_t proto, ipproto_input_t input, ipproto_ctlinput_t ctl)
865 {
866 
867 	MPASS(proto > 0);
868 
869 	/*
870 	 * The protocol slot must not be occupied by another protocol
871 	 * already.  An index pointing to rip_input() is unused.
872 	 */
873 	if (ip_protox[proto] == rip_input) {
874 		ip_protox[proto] = input;
875 		ip_ctlprotox[proto] = ctl;
876 		return (0);
877 	} else
878 		return (EEXIST);
879 }
880 
881 int
882 ipproto_unregister(uint8_t proto)
883 {
884 
885 	MPASS(proto > 0);
886 
887 	if (ip_protox[proto] != rip_input) {
888 		ip_protox[proto] = rip_input;
889 		ip_ctlprotox[proto] = rip_ctlinput;
890 		return (0);
891 	} else
892 		return (ENOENT);
893 }
894 
895 /*
896  * Forward a packet.  If some error occurs return the sender
897  * an icmp packet.  Note we can't always generate a meaningful
898  * icmp message because icmp doesn't have a large enough repertoire
899  * of codes and types.
900  *
901  * If not forwarding, just drop the packet.  This could be confusing
902  * if ipforwarding was zero but some routing protocol was advancing
903  * us as a gateway to somewhere.  However, we must let the routing
904  * protocol deal with that.
905  *
906  * The srcrt parameter indicates whether the packet is being forwarded
907  * via a source route.
908  */
909 void
910 ip_forward(struct mbuf *m, int srcrt)
911 {
912 	struct ip *ip = mtod(m, struct ip *);
913 	struct in_ifaddr *ia;
914 	struct mbuf *mcopy;
915 	struct sockaddr_in *sin;
916 	struct in_addr dest;
917 	struct route ro;
918 	uint32_t flowid;
919 	int error, type = 0, code = 0, mtu = 0;
920 
921 	NET_EPOCH_ASSERT();
922 
923 	if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
924 		IPSTAT_INC(ips_cantforward);
925 		m_freem(m);
926 		return;
927 	}
928 	if (
929 #ifdef IPSTEALTH
930 	    V_ipstealth == 0 &&
931 #endif
932 	    ip->ip_ttl <= IPTTLDEC) {
933 		icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 0, 0);
934 		return;
935 	}
936 
937 	bzero(&ro, sizeof(ro));
938 	sin = (struct sockaddr_in *)&ro.ro_dst;
939 	sin->sin_family = AF_INET;
940 	sin->sin_len = sizeof(*sin);
941 	sin->sin_addr = ip->ip_dst;
942 	flowid = m->m_pkthdr.flowid;
943 	ro.ro_nh = fib4_lookup(M_GETFIB(m), ip->ip_dst, 0, NHR_REF, flowid);
944 	if (ro.ro_nh != NULL) {
945 		ia = ifatoia(ro.ro_nh->nh_ifa);
946 	} else
947 		ia = NULL;
948 	/*
949 	 * Save the IP header and at most 8 bytes of the payload,
950 	 * in case we need to generate an ICMP message to the src.
951 	 *
952 	 * XXX this can be optimized a lot by saving the data in a local
953 	 * buffer on the stack (72 bytes at most), and only allocating the
954 	 * mbuf if really necessary. The vast majority of the packets
955 	 * are forwarded without having to send an ICMP back (either
956 	 * because unnecessary, or because rate limited), so we are
957 	 * really we are wasting a lot of work here.
958 	 *
959 	 * We don't use m_copym() because it might return a reference
960 	 * to a shared cluster. Both this function and ip_output()
961 	 * assume exclusive access to the IP header in `m', so any
962 	 * data in a cluster may change before we reach icmp_error().
963 	 */
964 	mcopy = m_gethdr(M_NOWAIT, m->m_type);
965 	if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_NOWAIT)) {
966 		/*
967 		 * It's probably ok if the pkthdr dup fails (because
968 		 * the deep copy of the tag chain failed), but for now
969 		 * be conservative and just discard the copy since
970 		 * code below may some day want the tags.
971 		 */
972 		m_free(mcopy);
973 		mcopy = NULL;
974 	}
975 	if (mcopy != NULL) {
976 		mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy));
977 		mcopy->m_pkthdr.len = mcopy->m_len;
978 		m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
979 	}
980 #ifdef IPSTEALTH
981 	if (V_ipstealth == 0)
982 #endif
983 		ip->ip_ttl -= IPTTLDEC;
984 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
985 	if (IPSEC_ENABLED(ipv4)) {
986 		if ((error = IPSEC_FORWARD(ipv4, m)) != 0) {
987 			/* mbuf consumed by IPsec */
988 			RO_NHFREE(&ro);
989 			m_freem(mcopy);
990 			if (error != EINPROGRESS)
991 				IPSTAT_INC(ips_cantforward);
992 			return;
993 		}
994 		/* No IPsec processing required */
995 	}
996 #endif /* IPSEC */
997 	/*
998 	 * If forwarding packet using same interface that it came in on,
999 	 * perhaps should send a redirect to sender to shortcut a hop.
1000 	 * Only send redirect if source is sending directly to us,
1001 	 * and if packet was not source routed (or has any options).
1002 	 * Also, don't send redirect if forwarding using a default route
1003 	 * or a route modified by a redirect.
1004 	 */
1005 	dest.s_addr = 0;
1006 	if (!srcrt && V_ipsendredirects &&
1007 	    ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) {
1008 		struct nhop_object *nh;
1009 
1010 		nh = ro.ro_nh;
1011 
1012 		if (nh != NULL && ((nh->nh_flags & (NHF_REDIRECT|NHF_DEFAULT)) == 0)) {
1013 			struct in_ifaddr *nh_ia = (struct in_ifaddr *)(nh->nh_ifa);
1014 			u_long src = ntohl(ip->ip_src.s_addr);
1015 
1016 			if (nh_ia != NULL &&
1017 			    (src & nh_ia->ia_subnetmask) == nh_ia->ia_subnet) {
1018 				/* Router requirements says to only send host redirects */
1019 				type = ICMP_REDIRECT;
1020 				code = ICMP_REDIRECT_HOST;
1021 				if (nh->nh_flags & NHF_GATEWAY) {
1022 				    if (nh->gw_sa.sa_family == AF_INET)
1023 					dest.s_addr = nh->gw4_sa.sin_addr.s_addr;
1024 				    else /* Do not redirect in case gw is AF_INET6 */
1025 					type = 0;
1026 				} else
1027 					dest.s_addr = ip->ip_dst.s_addr;
1028 			}
1029 		}
1030 	}
1031 
1032 	error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
1033 
1034 	if (error == EMSGSIZE && ro.ro_nh)
1035 		mtu = ro.ro_nh->nh_mtu;
1036 	RO_NHFREE(&ro);
1037 
1038 	if (error)
1039 		IPSTAT_INC(ips_cantforward);
1040 	else {
1041 		IPSTAT_INC(ips_forward);
1042 		if (type)
1043 			IPSTAT_INC(ips_redirectsent);
1044 		else {
1045 			if (mcopy)
1046 				m_freem(mcopy);
1047 			return;
1048 		}
1049 	}
1050 	if (mcopy == NULL)
1051 		return;
1052 
1053 	switch (error) {
1054 	case 0:				/* forwarded, but need redirect */
1055 		/* type, code set above */
1056 		break;
1057 
1058 	case ENETUNREACH:
1059 	case EHOSTUNREACH:
1060 	case ENETDOWN:
1061 	case EHOSTDOWN:
1062 	default:
1063 		type = ICMP_UNREACH;
1064 		code = ICMP_UNREACH_HOST;
1065 		break;
1066 
1067 	case EMSGSIZE:
1068 		type = ICMP_UNREACH;
1069 		code = ICMP_UNREACH_NEEDFRAG;
1070 		/*
1071 		 * If the MTU was set before make sure we are below the
1072 		 * interface MTU.
1073 		 * If the MTU wasn't set before use the interface mtu or
1074 		 * fall back to the next smaller mtu step compared to the
1075 		 * current packet size.
1076 		 */
1077 		if (mtu != 0) {
1078 			if (ia != NULL)
1079 				mtu = min(mtu, ia->ia_ifp->if_mtu);
1080 		} else {
1081 			if (ia != NULL)
1082 				mtu = ia->ia_ifp->if_mtu;
1083 			else
1084 				mtu = ip_next_mtu(ntohs(ip->ip_len), 0);
1085 		}
1086 		IPSTAT_INC(ips_cantfrag);
1087 		break;
1088 
1089 	case ENOBUFS:
1090 	case EACCES:			/* ipfw denied packet */
1091 		m_freem(mcopy);
1092 		return;
1093 	}
1094 	icmp_error(mcopy, type, code, dest.s_addr, mtu);
1095 }
1096 
1097 #define	CHECK_SO_CT(sp, ct) \
1098     (((sp->so_options & SO_TIMESTAMP) && (sp->so_ts_clock == ct)) ? 1 : 0)
1099 
1100 void
1101 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1102     struct mbuf *m)
1103 {
1104 	bool stamped;
1105 
1106 	stamped = false;
1107 	if ((inp->inp_socket->so_options & SO_BINTIME) ||
1108 	    CHECK_SO_CT(inp->inp_socket, SO_TS_BINTIME)) {
1109 		struct bintime boottimebin, bt;
1110 		struct timespec ts1;
1111 
1112 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1113 		    M_TSTMP)) {
1114 			mbuf_tstmp2timespec(m, &ts1);
1115 			timespec2bintime(&ts1, &bt);
1116 			getboottimebin(&boottimebin);
1117 			bintime_add(&bt, &boottimebin);
1118 		} else {
1119 			bintime(&bt);
1120 		}
1121 		*mp = sbcreatecontrol(&bt, sizeof(bt), SCM_BINTIME,
1122 		    SOL_SOCKET, M_NOWAIT);
1123 		if (*mp != NULL) {
1124 			mp = &(*mp)->m_next;
1125 			stamped = true;
1126 		}
1127 	}
1128 	if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME_MICRO)) {
1129 		struct bintime boottimebin, bt1;
1130 		struct timespec ts1;
1131 		struct timeval tv;
1132 
1133 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1134 		    M_TSTMP)) {
1135 			mbuf_tstmp2timespec(m, &ts1);
1136 			timespec2bintime(&ts1, &bt1);
1137 			getboottimebin(&boottimebin);
1138 			bintime_add(&bt1, &boottimebin);
1139 			bintime2timeval(&bt1, &tv);
1140 		} else {
1141 			microtime(&tv);
1142 		}
1143 		*mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv), SCM_TIMESTAMP,
1144 		    SOL_SOCKET, M_NOWAIT);
1145 		if (*mp != NULL) {
1146 			mp = &(*mp)->m_next;
1147 			stamped = true;
1148 		}
1149 	} else if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME)) {
1150 		struct bintime boottimebin;
1151 		struct timespec ts, ts1;
1152 
1153 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1154 		    M_TSTMP)) {
1155 			mbuf_tstmp2timespec(m, &ts);
1156 			getboottimebin(&boottimebin);
1157 			bintime2timespec(&boottimebin, &ts1);
1158 			timespecadd(&ts, &ts1, &ts);
1159 		} else {
1160 			nanotime(&ts);
1161 		}
1162 		*mp = sbcreatecontrol(&ts, sizeof(ts), SCM_REALTIME,
1163 		    SOL_SOCKET, M_NOWAIT);
1164 		if (*mp != NULL) {
1165 			mp = &(*mp)->m_next;
1166 			stamped = true;
1167 		}
1168 	} else if (CHECK_SO_CT(inp->inp_socket, SO_TS_MONOTONIC)) {
1169 		struct timespec ts;
1170 
1171 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1172 		    M_TSTMP))
1173 			mbuf_tstmp2timespec(m, &ts);
1174 		else
1175 			nanouptime(&ts);
1176 		*mp = sbcreatecontrol(&ts, sizeof(ts), SCM_MONOTONIC,
1177 		    SOL_SOCKET, M_NOWAIT);
1178 		if (*mp != NULL) {
1179 			mp = &(*mp)->m_next;
1180 			stamped = true;
1181 		}
1182 	}
1183 	if (stamped && (m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1184 	    M_TSTMP)) {
1185 		struct sock_timestamp_info sti;
1186 
1187 		bzero(&sti, sizeof(sti));
1188 		sti.st_info_flags = ST_INFO_HW;
1189 		if ((m->m_flags & M_TSTMP_HPREC) != 0)
1190 			sti.st_info_flags |= ST_INFO_HW_HPREC;
1191 		*mp = sbcreatecontrol(&sti, sizeof(sti), SCM_TIME_INFO,
1192 		    SOL_SOCKET, M_NOWAIT);
1193 		if (*mp != NULL)
1194 			mp = &(*mp)->m_next;
1195 	}
1196 	if (inp->inp_flags & INP_RECVDSTADDR) {
1197 		*mp = sbcreatecontrol(&ip->ip_dst, sizeof(struct in_addr),
1198 		    IP_RECVDSTADDR, IPPROTO_IP, M_NOWAIT);
1199 		if (*mp)
1200 			mp = &(*mp)->m_next;
1201 	}
1202 	if (inp->inp_flags & INP_RECVTTL) {
1203 		*mp = sbcreatecontrol(&ip->ip_ttl, sizeof(u_char), IP_RECVTTL,
1204 		    IPPROTO_IP, M_NOWAIT);
1205 		if (*mp)
1206 			mp = &(*mp)->m_next;
1207 	}
1208 #ifdef notyet
1209 	/* XXX
1210 	 * Moving these out of udp_input() made them even more broken
1211 	 * than they already were.
1212 	 */
1213 	/* options were tossed already */
1214 	if (inp->inp_flags & INP_RECVOPTS) {
1215 		*mp = sbcreatecontrol(opts_deleted_above,
1216 		    sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP, M_NOWAIT);
1217 		if (*mp)
1218 			mp = &(*mp)->m_next;
1219 	}
1220 	/* ip_srcroute doesn't do what we want here, need to fix */
1221 	if (inp->inp_flags & INP_RECVRETOPTS) {
1222 		*mp = sbcreatecontrol(ip_srcroute(m), sizeof(struct in_addr),
1223 		    IP_RECVRETOPTS, IPPROTO_IP, M_NOWAIT);
1224 		if (*mp)
1225 			mp = &(*mp)->m_next;
1226 	}
1227 #endif
1228 	if (inp->inp_flags & INP_RECVIF) {
1229 		struct ifnet *ifp;
1230 		struct sdlbuf {
1231 			struct sockaddr_dl sdl;
1232 			u_char	pad[32];
1233 		} sdlbuf;
1234 		struct sockaddr_dl *sdp;
1235 		struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1236 
1237 		if ((ifp = m->m_pkthdr.rcvif)) {
1238 			sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1239 			/*
1240 			 * Change our mind and don't try copy.
1241 			 */
1242 			if (sdp->sdl_family != AF_LINK ||
1243 			    sdp->sdl_len > sizeof(sdlbuf)) {
1244 				goto makedummy;
1245 			}
1246 			bcopy(sdp, sdl2, sdp->sdl_len);
1247 		} else {
1248 makedummy:
1249 			sdl2->sdl_len =
1250 			    offsetof(struct sockaddr_dl, sdl_data[0]);
1251 			sdl2->sdl_family = AF_LINK;
1252 			sdl2->sdl_index = 0;
1253 			sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1254 		}
1255 		*mp = sbcreatecontrol(sdl2, sdl2->sdl_len, IP_RECVIF,
1256 		    IPPROTO_IP, M_NOWAIT);
1257 		if (*mp)
1258 			mp = &(*mp)->m_next;
1259 	}
1260 	if (inp->inp_flags & INP_RECVTOS) {
1261 		*mp = sbcreatecontrol(&ip->ip_tos, sizeof(u_char), IP_RECVTOS,
1262 		    IPPROTO_IP, M_NOWAIT);
1263 		if (*mp)
1264 			mp = &(*mp)->m_next;
1265 	}
1266 
1267 	if (inp->inp_flags2 & INP_RECVFLOWID) {
1268 		uint32_t flowid, flow_type;
1269 
1270 		flowid = m->m_pkthdr.flowid;
1271 		flow_type = M_HASHTYPE_GET(m);
1272 
1273 		/*
1274 		 * XXX should handle the failure of one or the
1275 		 * other - don't populate both?
1276 		 */
1277 		*mp = sbcreatecontrol(&flowid, sizeof(uint32_t), IP_FLOWID,
1278 		    IPPROTO_IP, M_NOWAIT);
1279 		if (*mp)
1280 			mp = &(*mp)->m_next;
1281 		*mp = sbcreatecontrol(&flow_type, sizeof(uint32_t),
1282 		    IP_FLOWTYPE, IPPROTO_IP, M_NOWAIT);
1283 		if (*mp)
1284 			mp = &(*mp)->m_next;
1285 	}
1286 
1287 #ifdef	RSS
1288 	if (inp->inp_flags2 & INP_RECVRSSBUCKETID) {
1289 		uint32_t flowid, flow_type;
1290 		uint32_t rss_bucketid;
1291 
1292 		flowid = m->m_pkthdr.flowid;
1293 		flow_type = M_HASHTYPE_GET(m);
1294 
1295 		if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) {
1296 			*mp = sbcreatecontrol(&rss_bucketid, sizeof(uint32_t),
1297 			    IP_RSSBUCKETID, IPPROTO_IP, M_NOWAIT);
1298 			if (*mp)
1299 				mp = &(*mp)->m_next;
1300 		}
1301 	}
1302 #endif
1303 }
1304 
1305 /*
1306  * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
1307  * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
1308  * locking.  This code remains in ip_input.c as ip_mroute.c is optionally
1309  * compiled.
1310  */
1311 VNET_DEFINE_STATIC(int, ip_rsvp_on);
1312 VNET_DEFINE(struct socket *, ip_rsvpd);
1313 
1314 #define	V_ip_rsvp_on		VNET(ip_rsvp_on)
1315 
1316 int
1317 ip_rsvp_init(struct socket *so)
1318 {
1319 
1320 	if (V_ip_rsvpd != NULL)
1321 		return EADDRINUSE;
1322 
1323 	V_ip_rsvpd = so;
1324 	/*
1325 	 * This may seem silly, but we need to be sure we don't over-increment
1326 	 * the RSVP counter, in case something slips up.
1327 	 */
1328 	if (!V_ip_rsvp_on) {
1329 		V_ip_rsvp_on = 1;
1330 		V_rsvp_on++;
1331 	}
1332 
1333 	return 0;
1334 }
1335 
1336 int
1337 ip_rsvp_done(void)
1338 {
1339 
1340 	V_ip_rsvpd = NULL;
1341 	/*
1342 	 * This may seem silly, but we need to be sure we don't over-decrement
1343 	 * the RSVP counter, in case something slips up.
1344 	 */
1345 	if (V_ip_rsvp_on) {
1346 		V_ip_rsvp_on = 0;
1347 		V_rsvp_on--;
1348 	}
1349 	return 0;
1350 }
1351 
1352 int
1353 rsvp_input(struct mbuf **mp, int *offp, int proto)
1354 {
1355 	struct mbuf *m;
1356 
1357 	m = *mp;
1358 	*mp = NULL;
1359 
1360 	if (rsvp_input_p) { /* call the real one if loaded */
1361 		*mp = m;
1362 		rsvp_input_p(mp, offp, proto);
1363 		return (IPPROTO_DONE);
1364 	}
1365 
1366 	/* Can still get packets with rsvp_on = 0 if there is a local member
1367 	 * of the group to which the RSVP packet is addressed.  But in this
1368 	 * case we want to throw the packet away.
1369 	 */
1370 
1371 	if (!V_rsvp_on) {
1372 		m_freem(m);
1373 		return (IPPROTO_DONE);
1374 	}
1375 
1376 	if (V_ip_rsvpd != NULL) {
1377 		*mp = m;
1378 		rip_input(mp, offp, proto);
1379 		return (IPPROTO_DONE);
1380 	}
1381 	/* Drop the packet */
1382 	m_freem(m);
1383 	return (IPPROTO_DONE);
1384 }
1385