xref: /freebsd/sys/netinet/ip_input.c (revision 500f4659d7c8947082dba040a1d58e7d228f8d44)
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  *	@(#)ip_input.c	8.2 (Berkeley) 1/4/94
32  */
33 
34 #include <sys/cdefs.h>
35 __FBSDID("$FreeBSD$");
36 
37 #include "opt_bootp.h"
38 #include "opt_ipstealth.h"
39 #include "opt_ipsec.h"
40 #include "opt_route.h"
41 #include "opt_rss.h"
42 
43 #include <sys/param.h>
44 #include <sys/systm.h>
45 #include <sys/hhook.h>
46 #include <sys/mbuf.h>
47 #include <sys/malloc.h>
48 #include <sys/domain.h>
49 #include <sys/protosw.h>
50 #include <sys/socket.h>
51 #include <sys/time.h>
52 #include <sys/kernel.h>
53 #include <sys/lock.h>
54 #include <sys/rmlock.h>
55 #include <sys/rwlock.h>
56 #include <sys/sdt.h>
57 #include <sys/syslog.h>
58 #include <sys/sysctl.h>
59 
60 #include <net/if.h>
61 #include <net/if_types.h>
62 #include <net/if_var.h>
63 #include <net/if_dl.h>
64 #include <net/pfil.h>
65 #include <net/route.h>
66 #include <net/route/nhop.h>
67 #include <net/netisr.h>
68 #include <net/rss_config.h>
69 #include <net/vnet.h>
70 
71 #include <netinet/in.h>
72 #include <netinet/in_kdtrace.h>
73 #include <netinet/in_systm.h>
74 #include <netinet/in_var.h>
75 #include <netinet/ip.h>
76 #include <netinet/in_fib.h>
77 #include <netinet/in_pcb.h>
78 #include <netinet/ip_var.h>
79 #include <netinet/ip_fw.h>
80 #include <netinet/ip_icmp.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 
86 #include <netipsec/ipsec_support.h>
87 
88 #include <sys/socketvar.h>
89 
90 #include <security/mac/mac_framework.h>
91 
92 #ifdef CTASSERT
93 CTASSERT(sizeof(struct ip) == 20);
94 #endif
95 
96 /* IP reassembly functions are defined in ip_reass.c. */
97 extern void ipreass_init(void);
98 extern void ipreass_drain(void);
99 extern void ipreass_slowtimo(void);
100 #ifdef VIMAGE
101 extern void ipreass_destroy(void);
102 #endif
103 
104 struct rmlock in_ifaddr_lock;
105 RM_SYSINIT(in_ifaddr_lock, &in_ifaddr_lock, "in_ifaddr_lock");
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 /*
124  * XXX - Setting ip_checkinterface mostly implements the receive side of
125  * the Strong ES model described in RFC 1122, but since the routing table
126  * and transmit implementation do not implement the Strong ES model,
127  * setting this to 1 results in an odd hybrid.
128  *
129  * XXX - ip_checkinterface currently must be disabled if you use ipnat
130  * to translate the destination address to another local interface.
131  *
132  * XXX - ip_checkinterface must be disabled if you add IP aliases
133  * to the loopback interface instead of the interface where the
134  * packets for those addresses are received.
135  */
136 VNET_DEFINE_STATIC(int, ip_checkinterface);
137 #define	V_ip_checkinterface	VNET(ip_checkinterface)
138 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_VNET | CTLFLAG_RW,
139     &VNET_NAME(ip_checkinterface), 0,
140     "Verify packet arrives on correct interface");
141 
142 VNET_DEFINE(pfil_head_t, inet_pfil_head);	/* Packet filter hooks */
143 
144 static struct netisr_handler ip_nh = {
145 	.nh_name = "ip",
146 	.nh_handler = ip_input,
147 	.nh_proto = NETISR_IP,
148 #ifdef	RSS
149 	.nh_m2cpuid = rss_soft_m2cpuid_v4,
150 	.nh_policy = NETISR_POLICY_CPU,
151 	.nh_dispatch = NETISR_DISPATCH_HYBRID,
152 #else
153 	.nh_policy = NETISR_POLICY_FLOW,
154 #endif
155 };
156 
157 #ifdef	RSS
158 /*
159  * Directly dispatched frames are currently assumed
160  * to have a flowid already calculated.
161  *
162  * It should likely have something that assert it
163  * actually has valid flow details.
164  */
165 static struct netisr_handler ip_direct_nh = {
166 	.nh_name = "ip_direct",
167 	.nh_handler = ip_direct_input,
168 	.nh_proto = NETISR_IP_DIRECT,
169 	.nh_m2cpuid = rss_soft_m2cpuid_v4,
170 	.nh_policy = NETISR_POLICY_CPU,
171 	.nh_dispatch = NETISR_DISPATCH_HYBRID,
172 };
173 #endif
174 
175 extern	struct domain inetdomain;
176 extern	struct protosw inetsw[];
177 u_char	ip_protox[IPPROTO_MAX];
178 VNET_DEFINE(struct in_ifaddrhead, in_ifaddrhead);  /* first inet address */
179 VNET_DEFINE(struct in_ifaddrhashhead *, in_ifaddrhashtbl); /* inet addr hash table  */
180 VNET_DEFINE(u_long, in_ifaddrhmask);		/* mask for hash table */
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 void
311 ip_init(void)
312 {
313 	struct pfil_head_args args;
314 	struct protosw *pr;
315 	int i;
316 
317 	CK_STAILQ_INIT(&V_in_ifaddrhead);
318 	V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask);
319 
320 	/* Initialize IP reassembly queue. */
321 	ipreass_init();
322 
323 	/* Initialize packet filter hooks. */
324 	args.pa_version = PFIL_VERSION;
325 	args.pa_flags = PFIL_IN | PFIL_OUT;
326 	args.pa_type = PFIL_TYPE_IP4;
327 	args.pa_headname = PFIL_INET_NAME;
328 	V_inet_pfil_head = pfil_head_register(&args);
329 
330 	if (hhook_head_register(HHOOK_TYPE_IPSEC_IN, AF_INET,
331 	    &V_ipsec_hhh_in[HHOOK_IPSEC_INET],
332 	    HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
333 		printf("%s: WARNING: unable to register input helper hook\n",
334 		    __func__);
335 	if (hhook_head_register(HHOOK_TYPE_IPSEC_OUT, AF_INET,
336 	    &V_ipsec_hhh_out[HHOOK_IPSEC_INET],
337 	    HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
338 		printf("%s: WARNING: unable to register output helper hook\n",
339 		    __func__);
340 
341 	/* Skip initialization of globals for non-default instances. */
342 #ifdef VIMAGE
343 	if (!IS_DEFAULT_VNET(curvnet)) {
344 		netisr_register_vnet(&ip_nh);
345 #ifdef	RSS
346 		netisr_register_vnet(&ip_direct_nh);
347 #endif
348 		return;
349 	}
350 #endif
351 
352 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
353 	if (pr == NULL)
354 		panic("ip_init: PF_INET not found");
355 
356 	/* Initialize the entire ip_protox[] array to IPPROTO_RAW. */
357 	for (i = 0; i < IPPROTO_MAX; i++)
358 		ip_protox[i] = pr - inetsw;
359 	/*
360 	 * Cycle through IP protocols and put them into the appropriate place
361 	 * in ip_protox[].
362 	 */
363 	for (pr = inetdomain.dom_protosw;
364 	    pr < inetdomain.dom_protoswNPROTOSW; pr++)
365 		if (pr->pr_domain->dom_family == PF_INET &&
366 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
367 			/* Be careful to only index valid IP protocols. */
368 			if (pr->pr_protocol < IPPROTO_MAX)
369 				ip_protox[pr->pr_protocol] = pr - inetsw;
370 		}
371 
372 	netisr_register(&ip_nh);
373 #ifdef	RSS
374 	netisr_register(&ip_direct_nh);
375 #endif
376 }
377 
378 #ifdef VIMAGE
379 static void
380 ip_destroy(void *unused __unused)
381 {
382 	int error;
383 
384 #ifdef	RSS
385 	netisr_unregister_vnet(&ip_direct_nh);
386 #endif
387 	netisr_unregister_vnet(&ip_nh);
388 
389 	pfil_head_unregister(V_inet_pfil_head);
390 	error = hhook_head_deregister(V_ipsec_hhh_in[HHOOK_IPSEC_INET]);
391 	if (error != 0) {
392 		printf("%s: WARNING: unable to deregister input helper hook "
393 		    "type HHOOK_TYPE_IPSEC_IN, id HHOOK_IPSEC_INET: "
394 		    "error %d returned\n", __func__, error);
395 	}
396 	error = hhook_head_deregister(V_ipsec_hhh_out[HHOOK_IPSEC_INET]);
397 	if (error != 0) {
398 		printf("%s: WARNING: unable to deregister output helper hook "
399 		    "type HHOOK_TYPE_IPSEC_OUT, id HHOOK_IPSEC_INET: "
400 		    "error %d returned\n", __func__, error);
401 	}
402 
403 	/* Remove the IPv4 addresses from all interfaces. */
404 	in_ifscrub_all();
405 
406 	/* Make sure the IPv4 routes are gone as well. */
407 	rib_flush_routes_family(AF_INET);
408 
409 	/* Destroy IP reassembly queue. */
410 	ipreass_destroy();
411 
412 	/* Cleanup in_ifaddr hash table; should be empty. */
413 	hashdestroy(V_in_ifaddrhashtbl, M_IFADDR, V_in_ifaddrhmask);
414 }
415 
416 VNET_SYSUNINIT(ip, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip_destroy, NULL);
417 #endif
418 
419 #ifdef	RSS
420 /*
421  * IP direct input routine.
422  *
423  * This is called when reinjecting completed fragments where
424  * all of the previous checking and book-keeping has been done.
425  */
426 void
427 ip_direct_input(struct mbuf *m)
428 {
429 	struct ip *ip;
430 	int hlen;
431 
432 	ip = mtod(m, struct ip *);
433 	hlen = ip->ip_hl << 2;
434 
435 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
436 	if (IPSEC_ENABLED(ipv4)) {
437 		if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
438 			return;
439 	}
440 #endif /* IPSEC */
441 	IPSTAT_INC(ips_delivered);
442 	(*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
443 	return;
444 }
445 #endif
446 
447 /*
448  * Ip input routine.  Checksum and byte swap header.  If fragmented
449  * try to reassemble.  Process options.  Pass to next level.
450  */
451 void
452 ip_input(struct mbuf *m)
453 {
454 	struct rm_priotracker in_ifa_tracker;
455 	struct ip *ip = NULL;
456 	struct in_ifaddr *ia = NULL;
457 	struct ifaddr *ifa;
458 	struct ifnet *ifp;
459 	int    checkif, hlen = 0;
460 	uint16_t sum, ip_len;
461 	int dchg = 0;				/* dest changed after fw */
462 	struct in_addr odst;			/* original dst address */
463 
464 	M_ASSERTPKTHDR(m);
465 	NET_EPOCH_ASSERT();
466 
467 	if (m->m_flags & M_FASTFWD_OURS) {
468 		m->m_flags &= ~M_FASTFWD_OURS;
469 		/* Set up some basics that will be used later. */
470 		ip = mtod(m, struct ip *);
471 		hlen = ip->ip_hl << 2;
472 		ip_len = ntohs(ip->ip_len);
473 		goto ours;
474 	}
475 
476 	IPSTAT_INC(ips_total);
477 
478 	if (m->m_pkthdr.len < sizeof(struct ip))
479 		goto tooshort;
480 
481 	if (m->m_len < sizeof (struct ip) &&
482 	    (m = m_pullup(m, sizeof (struct ip))) == NULL) {
483 		IPSTAT_INC(ips_toosmall);
484 		return;
485 	}
486 	ip = mtod(m, struct ip *);
487 
488 	if (ip->ip_v != IPVERSION) {
489 		IPSTAT_INC(ips_badvers);
490 		goto bad;
491 	}
492 
493 	hlen = ip->ip_hl << 2;
494 	if (hlen < sizeof(struct ip)) {	/* minimum header length */
495 		IPSTAT_INC(ips_badhlen);
496 		goto bad;
497 	}
498 	if (hlen > m->m_len) {
499 		if ((m = m_pullup(m, hlen)) == NULL) {
500 			IPSTAT_INC(ips_badhlen);
501 			return;
502 		}
503 		ip = mtod(m, struct ip *);
504 	}
505 
506 	IP_PROBE(receive, NULL, NULL, ip, m->m_pkthdr.rcvif, ip, NULL);
507 
508 	/* IN_LOOPBACK must not appear on the wire - RFC1122 */
509 	ifp = m->m_pkthdr.rcvif;
510 	if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) ||
511 	    IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) {
512 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
513 			IPSTAT_INC(ips_badaddr);
514 			goto bad;
515 		}
516 	}
517 
518 	if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
519 		sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
520 	} else {
521 		if (hlen == sizeof(struct ip)) {
522 			sum = in_cksum_hdr(ip);
523 		} else {
524 			sum = in_cksum(m, hlen);
525 		}
526 	}
527 	if (sum) {
528 		IPSTAT_INC(ips_badsum);
529 		goto bad;
530 	}
531 
532 #ifdef ALTQ
533 	if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
534 		/* packet is dropped by traffic conditioner */
535 		return;
536 #endif
537 
538 	ip_len = ntohs(ip->ip_len);
539 	if (ip_len < hlen) {
540 		IPSTAT_INC(ips_badlen);
541 		goto bad;
542 	}
543 
544 	/*
545 	 * Check that the amount of data in the buffers
546 	 * is as at least much as the IP header would have us expect.
547 	 * Trim mbufs if longer than we expect.
548 	 * Drop packet if shorter than we expect.
549 	 */
550 	if (m->m_pkthdr.len < ip_len) {
551 tooshort:
552 		IPSTAT_INC(ips_tooshort);
553 		goto bad;
554 	}
555 	if (m->m_pkthdr.len > ip_len) {
556 		if (m->m_len == m->m_pkthdr.len) {
557 			m->m_len = ip_len;
558 			m->m_pkthdr.len = ip_len;
559 		} else
560 			m_adj(m, ip_len - m->m_pkthdr.len);
561 	}
562 
563 	/*
564 	 * Try to forward the packet, but if we fail continue.
565 	 * ip_tryforward() does not generate redirects, so fall
566 	 * through to normal processing if redirects are required.
567 	 * ip_tryforward() does inbound and outbound packet firewall
568 	 * processing. If firewall has decided that destination becomes
569 	 * our local address, it sets M_FASTFWD_OURS flag. In this
570 	 * case skip another inbound firewall processing and update
571 	 * ip pointer.
572 	 */
573 	if (V_ipforwarding != 0
574 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
575 	    && (!IPSEC_ENABLED(ipv4) ||
576 	    IPSEC_CAPS(ipv4, m, IPSEC_CAP_OPERABLE) == 0)
577 #endif
578 	    ) {
579 		if ((m = ip_tryforward(m)) == NULL)
580 			return;
581 		if (m->m_flags & M_FASTFWD_OURS) {
582 			m->m_flags &= ~M_FASTFWD_OURS;
583 			ip = mtod(m, struct ip *);
584 			goto ours;
585 		}
586 	}
587 
588 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
589 	/*
590 	 * Bypass packet filtering for packets previously handled by IPsec.
591 	 */
592 	if (IPSEC_ENABLED(ipv4) &&
593 	    IPSEC_CAPS(ipv4, m, IPSEC_CAP_BYPASS_FILTER) != 0)
594 			goto passin;
595 #endif
596 
597 	/*
598 	 * Run through list of hooks for input packets.
599 	 *
600 	 * NB: Beware of the destination address changing (e.g.
601 	 *     by NAT rewriting).  When this happens, tell
602 	 *     ip_forward to do the right thing.
603 	 */
604 
605 	/* Jump over all PFIL processing if hooks are not active. */
606 	if (!PFIL_HOOKED_IN(V_inet_pfil_head))
607 		goto passin;
608 
609 	odst = ip->ip_dst;
610 	if (pfil_run_hooks(V_inet_pfil_head, &m, ifp, PFIL_IN, NULL) !=
611 	    PFIL_PASS)
612 		return;
613 	if (m == NULL)			/* consumed by filter */
614 		return;
615 
616 	ip = mtod(m, struct ip *);
617 	dchg = (odst.s_addr != ip->ip_dst.s_addr);
618 	ifp = m->m_pkthdr.rcvif;
619 
620 	if (m->m_flags & M_FASTFWD_OURS) {
621 		m->m_flags &= ~M_FASTFWD_OURS;
622 		goto ours;
623 	}
624 	if (m->m_flags & M_IP_NEXTHOP) {
625 		if (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) {
626 			/*
627 			 * Directly ship the packet on.  This allows
628 			 * forwarding packets originally destined to us
629 			 * to some other directly connected host.
630 			 */
631 			ip_forward(m, 1);
632 			return;
633 		}
634 	}
635 passin:
636 
637 	/*
638 	 * Process options and, if not destined for us,
639 	 * ship it on.  ip_dooptions returns 1 when an
640 	 * error was detected (causing an icmp message
641 	 * to be sent and the original packet to be freed).
642 	 */
643 	if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
644 		return;
645 
646         /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
647          * matter if it is destined to another node, or whether it is
648          * a multicast one, RSVP wants it! and prevents it from being forwarded
649          * anywhere else. Also checks if the rsvp daemon is running before
650 	 * grabbing the packet.
651          */
652 	if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP)
653 		goto ours;
654 
655 	/*
656 	 * Check our list of addresses, to see if the packet is for us.
657 	 * If we don't have any addresses, assume any unicast packet
658 	 * we receive might be for us (and let the upper layers deal
659 	 * with it).
660 	 */
661 	if (CK_STAILQ_EMPTY(&V_in_ifaddrhead) &&
662 	    (m->m_flags & (M_MCAST|M_BCAST)) == 0)
663 		goto ours;
664 
665 	/*
666 	 * Enable a consistency check between the destination address
667 	 * and the arrival interface for a unicast packet (the RFC 1122
668 	 * strong ES model) if IP forwarding is disabled and the packet
669 	 * is not locally generated and the packet is not subject to
670 	 * 'ipfw fwd'.
671 	 *
672 	 * XXX - Checking also should be disabled if the destination
673 	 * address is ipnat'ed to a different interface.
674 	 *
675 	 * XXX - Checking is incompatible with IP aliases added
676 	 * to the loopback interface instead of the interface where
677 	 * the packets are received.
678 	 *
679 	 * XXX - This is the case for carp vhost IPs as well so we
680 	 * insert a workaround. If the packet got here, we already
681 	 * checked with carp_iamatch() and carp_forus().
682 	 */
683 	checkif = V_ip_checkinterface && (V_ipforwarding == 0) &&
684 	    ifp != NULL && ((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 	IN_IFADDR_RLOCK(&in_ifa_tracker);
691 	LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
692 		/*
693 		 * If the address matches, verify that the packet
694 		 * arrived via the correct interface if checking is
695 		 * enabled.
696 		 */
697 		if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr &&
698 		    (!checkif || ia->ia_ifp == ifp)) {
699 			counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
700 			counter_u64_add(ia->ia_ifa.ifa_ibytes,
701 			    m->m_pkthdr.len);
702 			IN_IFADDR_RUNLOCK(&in_ifa_tracker);
703 			goto ours;
704 		}
705 	}
706 	IN_IFADDR_RUNLOCK(&in_ifa_tracker);
707 
708 	/*
709 	 * Check for broadcast addresses.
710 	 *
711 	 * Only accept broadcast packets that arrive via the matching
712 	 * interface.  Reception of forwarded directed broadcasts would
713 	 * be handled via ip_forward() and ether_output() with the loopback
714 	 * into the stack for SIMPLEX interfaces handled by ether_output().
715 	 */
716 	if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
717 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
718 			if (ifa->ifa_addr->sa_family != AF_INET)
719 				continue;
720 			ia = ifatoia(ifa);
721 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
722 			    ip->ip_dst.s_addr) {
723 				counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
724 				counter_u64_add(ia->ia_ifa.ifa_ibytes,
725 				    m->m_pkthdr.len);
726 				goto ours;
727 			}
728 #ifdef BOOTP_COMPAT
729 			if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) {
730 				counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
731 				counter_u64_add(ia->ia_ifa.ifa_ibytes,
732 				    m->m_pkthdr.len);
733 				goto ours;
734 			}
735 #endif
736 		}
737 		ia = NULL;
738 	}
739 	/* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */
740 	if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) {
741 		IPSTAT_INC(ips_cantforward);
742 		m_freem(m);
743 		return;
744 	}
745 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
746 		if (V_ip_mrouter) {
747 			/*
748 			 * If we are acting as a multicast router, all
749 			 * incoming multicast packets are passed to the
750 			 * kernel-level multicast forwarding function.
751 			 * The packet is returned (relatively) intact; if
752 			 * ip_mforward() returns a non-zero value, the packet
753 			 * must be discarded, else it may be accepted below.
754 			 */
755 			if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) {
756 				IPSTAT_INC(ips_cantforward);
757 				m_freem(m);
758 				return;
759 			}
760 
761 			/*
762 			 * The process-level routing daemon needs to receive
763 			 * all multicast IGMP packets, whether or not this
764 			 * host belongs to their destination groups.
765 			 */
766 			if (ip->ip_p == IPPROTO_IGMP)
767 				goto ours;
768 			IPSTAT_INC(ips_forward);
769 		}
770 		/*
771 		 * Assume the packet is for us, to avoid prematurely taking
772 		 * a lock on the in_multi hash. Protocols must perform
773 		 * their own filtering and update statistics accordingly.
774 		 */
775 		goto ours;
776 	}
777 	if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
778 		goto ours;
779 	if (ip->ip_dst.s_addr == INADDR_ANY)
780 		goto ours;
781 
782 	/*
783 	 * Not for us; forward if possible and desirable.
784 	 */
785 	if (V_ipforwarding == 0) {
786 		IPSTAT_INC(ips_cantforward);
787 		m_freem(m);
788 	} else {
789 		ip_forward(m, dchg);
790 	}
791 	return;
792 
793 ours:
794 #ifdef IPSTEALTH
795 	/*
796 	 * IPSTEALTH: Process non-routing options only
797 	 * if the packet is destined for us.
798 	 */
799 	if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1))
800 		return;
801 #endif /* IPSTEALTH */
802 
803 	/*
804 	 * Attempt reassembly; if it succeeds, proceed.
805 	 * ip_reass() will return a different mbuf.
806 	 */
807 	if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) {
808 		/* XXXGL: shouldn't we save & set m_flags? */
809 		m = ip_reass(m);
810 		if (m == NULL)
811 			return;
812 		ip = mtod(m, struct ip *);
813 		/* Get the header length of the reassembled packet */
814 		hlen = ip->ip_hl << 2;
815 	}
816 
817 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
818 	if (IPSEC_ENABLED(ipv4)) {
819 		if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
820 			return;
821 	}
822 #endif /* IPSEC */
823 
824 	/*
825 	 * Switch out to protocol's input routine.
826 	 */
827 	IPSTAT_INC(ips_delivered);
828 
829 	(*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
830 	return;
831 bad:
832 	m_freem(m);
833 }
834 
835 /*
836  * IP timer processing;
837  * if a timer expires on a reassembly
838  * queue, discard it.
839  */
840 void
841 ip_slowtimo(void)
842 {
843 	VNET_ITERATOR_DECL(vnet_iter);
844 
845 	VNET_LIST_RLOCK_NOSLEEP();
846 	VNET_FOREACH(vnet_iter) {
847 		CURVNET_SET(vnet_iter);
848 		ipreass_slowtimo();
849 		CURVNET_RESTORE();
850 	}
851 	VNET_LIST_RUNLOCK_NOSLEEP();
852 }
853 
854 void
855 ip_drain(void)
856 {
857 	VNET_ITERATOR_DECL(vnet_iter);
858 
859 	VNET_LIST_RLOCK_NOSLEEP();
860 	VNET_FOREACH(vnet_iter) {
861 		CURVNET_SET(vnet_iter);
862 		ipreass_drain();
863 		CURVNET_RESTORE();
864 	}
865 	VNET_LIST_RUNLOCK_NOSLEEP();
866 }
867 
868 /*
869  * The protocol to be inserted into ip_protox[] must be already registered
870  * in inetsw[], either statically or through pf_proto_register().
871  */
872 int
873 ipproto_register(short ipproto)
874 {
875 	struct protosw *pr;
876 
877 	/* Sanity checks. */
878 	if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
879 		return (EPROTONOSUPPORT);
880 
881 	/*
882 	 * The protocol slot must not be occupied by another protocol
883 	 * already.  An index pointing to IPPROTO_RAW is unused.
884 	 */
885 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
886 	if (pr == NULL)
887 		return (EPFNOSUPPORT);
888 	if (ip_protox[ipproto] != pr - inetsw)	/* IPPROTO_RAW */
889 		return (EEXIST);
890 
891 	/* Find the protocol position in inetsw[] and set the index. */
892 	for (pr = inetdomain.dom_protosw;
893 	     pr < inetdomain.dom_protoswNPROTOSW; pr++) {
894 		if (pr->pr_domain->dom_family == PF_INET &&
895 		    pr->pr_protocol && pr->pr_protocol == ipproto) {
896 			ip_protox[pr->pr_protocol] = pr - inetsw;
897 			return (0);
898 		}
899 	}
900 	return (EPROTONOSUPPORT);
901 }
902 
903 int
904 ipproto_unregister(short ipproto)
905 {
906 	struct protosw *pr;
907 
908 	/* Sanity checks. */
909 	if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
910 		return (EPROTONOSUPPORT);
911 
912 	/* Check if the protocol was indeed registered. */
913 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
914 	if (pr == NULL)
915 		return (EPFNOSUPPORT);
916 	if (ip_protox[ipproto] == pr - inetsw)  /* IPPROTO_RAW */
917 		return (ENOENT);
918 
919 	/* Reset the protocol slot to IPPROTO_RAW. */
920 	ip_protox[ipproto] = pr - inetsw;
921 	return (0);
922 }
923 
924 u_char inetctlerrmap[PRC_NCMDS] = {
925 	0,		0,		0,		0,
926 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
927 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
928 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
929 	0,		0,		EHOSTUNREACH,	0,
930 	ENOPROTOOPT,	ECONNREFUSED
931 };
932 
933 /*
934  * Forward a packet.  If some error occurs return the sender
935  * an icmp packet.  Note we can't always generate a meaningful
936  * icmp message because icmp doesn't have a large enough repertoire
937  * of codes and types.
938  *
939  * If not forwarding, just drop the packet.  This could be confusing
940  * if ipforwarding was zero but some routing protocol was advancing
941  * us as a gateway to somewhere.  However, we must let the routing
942  * protocol deal with that.
943  *
944  * The srcrt parameter indicates whether the packet is being forwarded
945  * via a source route.
946  */
947 void
948 ip_forward(struct mbuf *m, int srcrt)
949 {
950 	struct ip *ip = mtod(m, struct ip *);
951 	struct in_ifaddr *ia;
952 	struct mbuf *mcopy;
953 	struct sockaddr_in *sin;
954 	struct in_addr dest;
955 	struct route ro;
956 	uint32_t flowid;
957 	int error, type = 0, code = 0, mtu = 0;
958 
959 	NET_EPOCH_ASSERT();
960 
961 	if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
962 		IPSTAT_INC(ips_cantforward);
963 		m_freem(m);
964 		return;
965 	}
966 	if (
967 #ifdef IPSTEALTH
968 	    V_ipstealth == 0 &&
969 #endif
970 	    ip->ip_ttl <= IPTTLDEC) {
971 		icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 0, 0);
972 		return;
973 	}
974 
975 	bzero(&ro, sizeof(ro));
976 	sin = (struct sockaddr_in *)&ro.ro_dst;
977 	sin->sin_family = AF_INET;
978 	sin->sin_len = sizeof(*sin);
979 	sin->sin_addr = ip->ip_dst;
980 	flowid = m->m_pkthdr.flowid;
981 	ro.ro_nh = fib4_lookup(M_GETFIB(m), ip->ip_dst, 0, NHR_REF, flowid);
982 	if (ro.ro_nh != NULL) {
983 		ia = ifatoia(ro.ro_nh->nh_ifa);
984 	} else
985 		ia = NULL;
986 	/*
987 	 * Save the IP header and at most 8 bytes of the payload,
988 	 * in case we need to generate an ICMP message to the src.
989 	 *
990 	 * XXX this can be optimized a lot by saving the data in a local
991 	 * buffer on the stack (72 bytes at most), and only allocating the
992 	 * mbuf if really necessary. The vast majority of the packets
993 	 * are forwarded without having to send an ICMP back (either
994 	 * because unnecessary, or because rate limited), so we are
995 	 * really we are wasting a lot of work here.
996 	 *
997 	 * We don't use m_copym() because it might return a reference
998 	 * to a shared cluster. Both this function and ip_output()
999 	 * assume exclusive access to the IP header in `m', so any
1000 	 * data in a cluster may change before we reach icmp_error().
1001 	 */
1002 	mcopy = m_gethdr(M_NOWAIT, m->m_type);
1003 	if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_NOWAIT)) {
1004 		/*
1005 		 * It's probably ok if the pkthdr dup fails (because
1006 		 * the deep copy of the tag chain failed), but for now
1007 		 * be conservative and just discard the copy since
1008 		 * code below may some day want the tags.
1009 		 */
1010 		m_free(mcopy);
1011 		mcopy = NULL;
1012 	}
1013 	if (mcopy != NULL) {
1014 		mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy));
1015 		mcopy->m_pkthdr.len = mcopy->m_len;
1016 		m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1017 	}
1018 #ifdef IPSTEALTH
1019 	if (V_ipstealth == 0)
1020 #endif
1021 		ip->ip_ttl -= IPTTLDEC;
1022 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1023 	if (IPSEC_ENABLED(ipv4)) {
1024 		if ((error = IPSEC_FORWARD(ipv4, m)) != 0) {
1025 			/* mbuf consumed by IPsec */
1026 			RO_NHFREE(&ro);
1027 			m_freem(mcopy);
1028 			if (error != EINPROGRESS)
1029 				IPSTAT_INC(ips_cantforward);
1030 			return;
1031 		}
1032 		/* No IPsec processing required */
1033 	}
1034 #endif /* IPSEC */
1035 	/*
1036 	 * If forwarding packet using same interface that it came in on,
1037 	 * perhaps should send a redirect to sender to shortcut a hop.
1038 	 * Only send redirect if source is sending directly to us,
1039 	 * and if packet was not source routed (or has any options).
1040 	 * Also, don't send redirect if forwarding using a default route
1041 	 * or a route modified by a redirect.
1042 	 */
1043 	dest.s_addr = 0;
1044 	if (!srcrt && V_ipsendredirects &&
1045 	    ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) {
1046 		struct nhop_object *nh;
1047 
1048 		nh = ro.ro_nh;
1049 
1050 		if (nh != NULL && ((nh->nh_flags & (NHF_REDIRECT|NHF_DEFAULT)) == 0)) {
1051 			struct in_ifaddr *nh_ia = (struct in_ifaddr *)(nh->nh_ifa);
1052 			u_long src = ntohl(ip->ip_src.s_addr);
1053 
1054 			if (nh_ia != NULL &&
1055 			    (src & nh_ia->ia_subnetmask) == nh_ia->ia_subnet) {
1056 				if (nh->nh_flags & NHF_GATEWAY)
1057 					dest.s_addr = nh->gw4_sa.sin_addr.s_addr;
1058 				else
1059 					dest.s_addr = ip->ip_dst.s_addr;
1060 				/* Router requirements says to only send host redirects */
1061 				type = ICMP_REDIRECT;
1062 				code = ICMP_REDIRECT_HOST;
1063 			}
1064 		}
1065 	}
1066 
1067 	error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
1068 
1069 	if (error == EMSGSIZE && ro.ro_nh)
1070 		mtu = ro.ro_nh->nh_mtu;
1071 	RO_NHFREE(&ro);
1072 
1073 	if (error)
1074 		IPSTAT_INC(ips_cantforward);
1075 	else {
1076 		IPSTAT_INC(ips_forward);
1077 		if (type)
1078 			IPSTAT_INC(ips_redirectsent);
1079 		else {
1080 			if (mcopy)
1081 				m_freem(mcopy);
1082 			return;
1083 		}
1084 	}
1085 	if (mcopy == NULL)
1086 		return;
1087 
1088 	switch (error) {
1089 	case 0:				/* forwarded, but need redirect */
1090 		/* type, code set above */
1091 		break;
1092 
1093 	case ENETUNREACH:
1094 	case EHOSTUNREACH:
1095 	case ENETDOWN:
1096 	case EHOSTDOWN:
1097 	default:
1098 		type = ICMP_UNREACH;
1099 		code = ICMP_UNREACH_HOST;
1100 		break;
1101 
1102 	case EMSGSIZE:
1103 		type = ICMP_UNREACH;
1104 		code = ICMP_UNREACH_NEEDFRAG;
1105 		/*
1106 		 * If the MTU was set before make sure we are below the
1107 		 * interface MTU.
1108 		 * If the MTU wasn't set before use the interface mtu or
1109 		 * fall back to the next smaller mtu step compared to the
1110 		 * current packet size.
1111 		 */
1112 		if (mtu != 0) {
1113 			if (ia != NULL)
1114 				mtu = min(mtu, ia->ia_ifp->if_mtu);
1115 		} else {
1116 			if (ia != NULL)
1117 				mtu = ia->ia_ifp->if_mtu;
1118 			else
1119 				mtu = ip_next_mtu(ntohs(ip->ip_len), 0);
1120 		}
1121 		IPSTAT_INC(ips_cantfrag);
1122 		break;
1123 
1124 	case ENOBUFS:
1125 	case EACCES:			/* ipfw denied packet */
1126 		m_freem(mcopy);
1127 		return;
1128 	}
1129 	icmp_error(mcopy, type, code, dest.s_addr, mtu);
1130 }
1131 
1132 #define	CHECK_SO_CT(sp, ct) \
1133     (((sp->so_options & SO_TIMESTAMP) && (sp->so_ts_clock == ct)) ? 1 : 0)
1134 
1135 void
1136 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1137     struct mbuf *m)
1138 {
1139 	bool stamped;
1140 
1141 	stamped = false;
1142 	if ((inp->inp_socket->so_options & SO_BINTIME) ||
1143 	    CHECK_SO_CT(inp->inp_socket, SO_TS_BINTIME)) {
1144 		struct bintime boottimebin, bt;
1145 		struct timespec ts1;
1146 
1147 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1148 		    M_TSTMP)) {
1149 			mbuf_tstmp2timespec(m, &ts1);
1150 			timespec2bintime(&ts1, &bt);
1151 			getboottimebin(&boottimebin);
1152 			bintime_add(&bt, &boottimebin);
1153 		} else {
1154 			bintime(&bt);
1155 		}
1156 		*mp = sbcreatecontrol((caddr_t)&bt, sizeof(bt),
1157 		    SCM_BINTIME, SOL_SOCKET);
1158 		if (*mp != NULL) {
1159 			mp = &(*mp)->m_next;
1160 			stamped = true;
1161 		}
1162 	}
1163 	if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME_MICRO)) {
1164 		struct bintime boottimebin, bt1;
1165 		struct timespec ts1;
1166 		struct timeval tv;
1167 
1168 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1169 		    M_TSTMP)) {
1170 			mbuf_tstmp2timespec(m, &ts1);
1171 			timespec2bintime(&ts1, &bt1);
1172 			getboottimebin(&boottimebin);
1173 			bintime_add(&bt1, &boottimebin);
1174 			bintime2timeval(&bt1, &tv);
1175 		} else {
1176 			microtime(&tv);
1177 		}
1178 		*mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv),
1179 		    SCM_TIMESTAMP, SOL_SOCKET);
1180 		if (*mp != NULL) {
1181 			mp = &(*mp)->m_next;
1182 			stamped = true;
1183 		}
1184 	} else if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME)) {
1185 		struct bintime boottimebin;
1186 		struct timespec ts, ts1;
1187 
1188 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1189 		    M_TSTMP)) {
1190 			mbuf_tstmp2timespec(m, &ts);
1191 			getboottimebin(&boottimebin);
1192 			bintime2timespec(&boottimebin, &ts1);
1193 			timespecadd(&ts, &ts1, &ts);
1194 		} else {
1195 			nanotime(&ts);
1196 		}
1197 		*mp = sbcreatecontrol((caddr_t)&ts, sizeof(ts),
1198 		    SCM_REALTIME, SOL_SOCKET);
1199 		if (*mp != NULL) {
1200 			mp = &(*mp)->m_next;
1201 			stamped = true;
1202 		}
1203 	} else if (CHECK_SO_CT(inp->inp_socket, SO_TS_MONOTONIC)) {
1204 		struct timespec ts;
1205 
1206 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1207 		    M_TSTMP))
1208 			mbuf_tstmp2timespec(m, &ts);
1209 		else
1210 			nanouptime(&ts);
1211 		*mp = sbcreatecontrol((caddr_t)&ts, sizeof(ts),
1212 		    SCM_MONOTONIC, SOL_SOCKET);
1213 		if (*mp != NULL) {
1214 			mp = &(*mp)->m_next;
1215 			stamped = true;
1216 		}
1217 	}
1218 	if (stamped && (m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1219 	    M_TSTMP)) {
1220 		struct sock_timestamp_info sti;
1221 
1222 		bzero(&sti, sizeof(sti));
1223 		sti.st_info_flags = ST_INFO_HW;
1224 		if ((m->m_flags & M_TSTMP_HPREC) != 0)
1225 			sti.st_info_flags |= ST_INFO_HW_HPREC;
1226 		*mp = sbcreatecontrol((caddr_t)&sti, sizeof(sti), SCM_TIME_INFO,
1227 		    SOL_SOCKET);
1228 		if (*mp != NULL)
1229 			mp = &(*mp)->m_next;
1230 	}
1231 	if (inp->inp_flags & INP_RECVDSTADDR) {
1232 		*mp = sbcreatecontrol((caddr_t)&ip->ip_dst,
1233 		    sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1234 		if (*mp)
1235 			mp = &(*mp)->m_next;
1236 	}
1237 	if (inp->inp_flags & INP_RECVTTL) {
1238 		*mp = sbcreatecontrol((caddr_t)&ip->ip_ttl,
1239 		    sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
1240 		if (*mp)
1241 			mp = &(*mp)->m_next;
1242 	}
1243 #ifdef notyet
1244 	/* XXX
1245 	 * Moving these out of udp_input() made them even more broken
1246 	 * than they already were.
1247 	 */
1248 	/* options were tossed already */
1249 	if (inp->inp_flags & INP_RECVOPTS) {
1250 		*mp = sbcreatecontrol((caddr_t)opts_deleted_above,
1251 		    sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1252 		if (*mp)
1253 			mp = &(*mp)->m_next;
1254 	}
1255 	/* ip_srcroute doesn't do what we want here, need to fix */
1256 	if (inp->inp_flags & INP_RECVRETOPTS) {
1257 		*mp = sbcreatecontrol((caddr_t)ip_srcroute(m),
1258 		    sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1259 		if (*mp)
1260 			mp = &(*mp)->m_next;
1261 	}
1262 #endif
1263 	if (inp->inp_flags & INP_RECVIF) {
1264 		struct ifnet *ifp;
1265 		struct sdlbuf {
1266 			struct sockaddr_dl sdl;
1267 			u_char	pad[32];
1268 		} sdlbuf;
1269 		struct sockaddr_dl *sdp;
1270 		struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1271 
1272 		if ((ifp = m->m_pkthdr.rcvif) &&
1273 		    ifp->if_index && ifp->if_index <= V_if_index) {
1274 			sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1275 			/*
1276 			 * Change our mind and don't try copy.
1277 			 */
1278 			if (sdp->sdl_family != AF_LINK ||
1279 			    sdp->sdl_len > sizeof(sdlbuf)) {
1280 				goto makedummy;
1281 			}
1282 			bcopy(sdp, sdl2, sdp->sdl_len);
1283 		} else {
1284 makedummy:
1285 			sdl2->sdl_len =
1286 			    offsetof(struct sockaddr_dl, sdl_data[0]);
1287 			sdl2->sdl_family = AF_LINK;
1288 			sdl2->sdl_index = 0;
1289 			sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1290 		}
1291 		*mp = sbcreatecontrol((caddr_t)sdl2, sdl2->sdl_len,
1292 		    IP_RECVIF, IPPROTO_IP);
1293 		if (*mp)
1294 			mp = &(*mp)->m_next;
1295 	}
1296 	if (inp->inp_flags & INP_RECVTOS) {
1297 		*mp = sbcreatecontrol((caddr_t)&ip->ip_tos,
1298 		    sizeof(u_char), IP_RECVTOS, IPPROTO_IP);
1299 		if (*mp)
1300 			mp = &(*mp)->m_next;
1301 	}
1302 
1303 	if (inp->inp_flags2 & INP_RECVFLOWID) {
1304 		uint32_t flowid, flow_type;
1305 
1306 		flowid = m->m_pkthdr.flowid;
1307 		flow_type = M_HASHTYPE_GET(m);
1308 
1309 		/*
1310 		 * XXX should handle the failure of one or the
1311 		 * other - don't populate both?
1312 		 */
1313 		*mp = sbcreatecontrol((caddr_t) &flowid,
1314 		    sizeof(uint32_t), IP_FLOWID, IPPROTO_IP);
1315 		if (*mp)
1316 			mp = &(*mp)->m_next;
1317 		*mp = sbcreatecontrol((caddr_t) &flow_type,
1318 		    sizeof(uint32_t), IP_FLOWTYPE, IPPROTO_IP);
1319 		if (*mp)
1320 			mp = &(*mp)->m_next;
1321 	}
1322 
1323 #ifdef	RSS
1324 	if (inp->inp_flags2 & INP_RECVRSSBUCKETID) {
1325 		uint32_t flowid, flow_type;
1326 		uint32_t rss_bucketid;
1327 
1328 		flowid = m->m_pkthdr.flowid;
1329 		flow_type = M_HASHTYPE_GET(m);
1330 
1331 		if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) {
1332 			*mp = sbcreatecontrol((caddr_t) &rss_bucketid,
1333 			   sizeof(uint32_t), IP_RSSBUCKETID, IPPROTO_IP);
1334 			if (*mp)
1335 				mp = &(*mp)->m_next;
1336 		}
1337 	}
1338 #endif
1339 }
1340 
1341 /*
1342  * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
1343  * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
1344  * locking.  This code remains in ip_input.c as ip_mroute.c is optionally
1345  * compiled.
1346  */
1347 VNET_DEFINE_STATIC(int, ip_rsvp_on);
1348 VNET_DEFINE(struct socket *, ip_rsvpd);
1349 
1350 #define	V_ip_rsvp_on		VNET(ip_rsvp_on)
1351 
1352 int
1353 ip_rsvp_init(struct socket *so)
1354 {
1355 
1356 	if (so->so_type != SOCK_RAW ||
1357 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
1358 		return EOPNOTSUPP;
1359 
1360 	if (V_ip_rsvpd != NULL)
1361 		return EADDRINUSE;
1362 
1363 	V_ip_rsvpd = so;
1364 	/*
1365 	 * This may seem silly, but we need to be sure we don't over-increment
1366 	 * the RSVP counter, in case something slips up.
1367 	 */
1368 	if (!V_ip_rsvp_on) {
1369 		V_ip_rsvp_on = 1;
1370 		V_rsvp_on++;
1371 	}
1372 
1373 	return 0;
1374 }
1375 
1376 int
1377 ip_rsvp_done(void)
1378 {
1379 
1380 	V_ip_rsvpd = NULL;
1381 	/*
1382 	 * This may seem silly, but we need to be sure we don't over-decrement
1383 	 * the RSVP counter, in case something slips up.
1384 	 */
1385 	if (V_ip_rsvp_on) {
1386 		V_ip_rsvp_on = 0;
1387 		V_rsvp_on--;
1388 	}
1389 	return 0;
1390 }
1391 
1392 int
1393 rsvp_input(struct mbuf **mp, int *offp, int proto)
1394 {
1395 	struct mbuf *m;
1396 
1397 	m = *mp;
1398 	*mp = NULL;
1399 
1400 	if (rsvp_input_p) { /* call the real one if loaded */
1401 		*mp = m;
1402 		rsvp_input_p(mp, offp, proto);
1403 		return (IPPROTO_DONE);
1404 	}
1405 
1406 	/* Can still get packets with rsvp_on = 0 if there is a local member
1407 	 * of the group to which the RSVP packet is addressed.  But in this
1408 	 * case we want to throw the packet away.
1409 	 */
1410 
1411 	if (!V_rsvp_on) {
1412 		m_freem(m);
1413 		return (IPPROTO_DONE);
1414 	}
1415 
1416 	if (V_ip_rsvpd != NULL) {
1417 		*mp = m;
1418 		rip_input(mp, offp, proto);
1419 		return (IPPROTO_DONE);
1420 	}
1421 	/* Drop the packet */
1422 	m_freem(m);
1423 	return (IPPROTO_DONE);
1424 }
1425