xref: /freebsd/sys/netinet/ip_input.c (revision 5bb3134a8c21cb87b30e135ef168483f0333dabb)
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 #include <netinet/ip_mroute.h>
86 
87 #include <netipsec/ipsec_support.h>
88 
89 #include <sys/socketvar.h>
90 
91 #include <security/mac/mac_framework.h>
92 
93 #ifdef CTASSERT
94 CTASSERT(sizeof(struct ip) == 20);
95 #endif
96 
97 /* IP reassembly functions are defined in ip_reass.c. */
98 extern void ipreass_init(void);
99 extern void ipreass_drain(void);
100 extern void ipreass_slowtimo(void);
101 #ifdef VIMAGE
102 extern void ipreass_destroy(void);
103 #endif
104 
105 VNET_DEFINE(int, rsvp_on);
106 
107 VNET_DEFINE(int, ipforwarding);
108 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_VNET | CTLFLAG_RW,
109     &VNET_NAME(ipforwarding), 0,
110     "Enable IP forwarding between interfaces");
111 
112 /*
113  * Respond with an ICMP host redirect when we forward a packet out of
114  * the same interface on which it was received.  See RFC 792.
115  */
116 VNET_DEFINE(int, ipsendredirects) = 1;
117 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_VNET | CTLFLAG_RW,
118     &VNET_NAME(ipsendredirects), 0,
119     "Enable sending IP redirects");
120 
121 VNET_DEFINE_STATIC(bool, ip_strong_es) = false;
122 #define	V_ip_strong_es	VNET(ip_strong_es)
123 SYSCTL_BOOL(_net_inet_ip, OID_AUTO, rfc1122_strong_es,
124     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip_strong_es), false,
125     "Packet's IP destination address must match address on arrival interface");
126 
127 VNET_DEFINE_STATIC(bool, ip_sav) = true;
128 #define	V_ip_sav	VNET(ip_sav)
129 SYSCTL_BOOL(_net_inet_ip, OID_AUTO, source_address_validation,
130     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip_sav), true,
131     "Drop incoming packets with source address that is a local address");
132 
133 VNET_DEFINE(pfil_head_t, inet_pfil_head);	/* Packet filter hooks */
134 
135 static struct netisr_handler ip_nh = {
136 	.nh_name = "ip",
137 	.nh_handler = ip_input,
138 	.nh_proto = NETISR_IP,
139 #ifdef	RSS
140 	.nh_m2cpuid = rss_soft_m2cpuid_v4,
141 	.nh_policy = NETISR_POLICY_CPU,
142 	.nh_dispatch = NETISR_DISPATCH_HYBRID,
143 #else
144 	.nh_policy = NETISR_POLICY_FLOW,
145 #endif
146 };
147 
148 #ifdef	RSS
149 /*
150  * Directly dispatched frames are currently assumed
151  * to have a flowid already calculated.
152  *
153  * It should likely have something that assert it
154  * actually has valid flow details.
155  */
156 static struct netisr_handler ip_direct_nh = {
157 	.nh_name = "ip_direct",
158 	.nh_handler = ip_direct_input,
159 	.nh_proto = NETISR_IP_DIRECT,
160 	.nh_m2cpuid = rss_soft_m2cpuid_v4,
161 	.nh_policy = NETISR_POLICY_CPU,
162 	.nh_dispatch = NETISR_DISPATCH_HYBRID,
163 };
164 #endif
165 
166 extern	struct domain inetdomain;
167 extern	struct protosw inetsw[];
168 u_char	ip_protox[IPPROTO_MAX];
169 VNET_DEFINE(struct in_ifaddrhead, in_ifaddrhead);  /* first inet address */
170 VNET_DEFINE(struct in_ifaddrhashhead *, in_ifaddrhashtbl); /* inet addr hash table  */
171 VNET_DEFINE(u_long, in_ifaddrhmask);		/* mask for hash table */
172 
173 /* Make sure it is safe to use hashinit(9) on CK_LIST. */
174 CTASSERT(sizeof(struct in_ifaddrhashhead) == sizeof(LIST_HEAD(, in_addr)));
175 
176 #ifdef IPCTL_DEFMTU
177 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
178     &ip_mtu, 0, "Default MTU");
179 #endif
180 
181 #ifdef IPSTEALTH
182 VNET_DEFINE(int, ipstealth);
183 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_VNET | CTLFLAG_RW,
184     &VNET_NAME(ipstealth), 0,
185     "IP stealth mode, no TTL decrementation on forwarding");
186 #endif
187 
188 /*
189  * IP statistics are stored in the "array" of counter(9)s.
190  */
191 VNET_PCPUSTAT_DEFINE(struct ipstat, ipstat);
192 VNET_PCPUSTAT_SYSINIT(ipstat);
193 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, IPCTL_STATS, stats, struct ipstat, ipstat,
194     "IP statistics (struct ipstat, netinet/ip_var.h)");
195 
196 #ifdef VIMAGE
197 VNET_PCPUSTAT_SYSUNINIT(ipstat);
198 #endif /* VIMAGE */
199 
200 /*
201  * Kernel module interface for updating ipstat.  The argument is an index
202  * into ipstat treated as an array.
203  */
204 void
205 kmod_ipstat_inc(int statnum)
206 {
207 
208 	counter_u64_add(VNET(ipstat)[statnum], 1);
209 }
210 
211 void
212 kmod_ipstat_dec(int statnum)
213 {
214 
215 	counter_u64_add(VNET(ipstat)[statnum], -1);
216 }
217 
218 static int
219 sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS)
220 {
221 	int error, qlimit;
222 
223 	netisr_getqlimit(&ip_nh, &qlimit);
224 	error = sysctl_handle_int(oidp, &qlimit, 0, req);
225 	if (error || !req->newptr)
226 		return (error);
227 	if (qlimit < 1)
228 		return (EINVAL);
229 	return (netisr_setqlimit(&ip_nh, qlimit));
230 }
231 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen,
232     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 0,
233     sysctl_netinet_intr_queue_maxlen, "I",
234     "Maximum size of the IP input queue");
235 
236 static int
237 sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS)
238 {
239 	u_int64_t qdrops_long;
240 	int error, qdrops;
241 
242 	netisr_getqdrops(&ip_nh, &qdrops_long);
243 	qdrops = qdrops_long;
244 	error = sysctl_handle_int(oidp, &qdrops, 0, req);
245 	if (error || !req->newptr)
246 		return (error);
247 	if (qdrops != 0)
248 		return (EINVAL);
249 	netisr_clearqdrops(&ip_nh);
250 	return (0);
251 }
252 
253 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops,
254     CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE,
255     0, 0, sysctl_netinet_intr_queue_drops, "I",
256     "Number of packets dropped from the IP input queue");
257 
258 #ifdef	RSS
259 static int
260 sysctl_netinet_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS)
261 {
262 	int error, qlimit;
263 
264 	netisr_getqlimit(&ip_direct_nh, &qlimit);
265 	error = sysctl_handle_int(oidp, &qlimit, 0, req);
266 	if (error || !req->newptr)
267 		return (error);
268 	if (qlimit < 1)
269 		return (EINVAL);
270 	return (netisr_setqlimit(&ip_direct_nh, qlimit));
271 }
272 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRDQMAXLEN, intr_direct_queue_maxlen,
273     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
274     0, 0, sysctl_netinet_intr_direct_queue_maxlen,
275     "I", "Maximum size of the IP direct input queue");
276 
277 static int
278 sysctl_netinet_intr_direct_queue_drops(SYSCTL_HANDLER_ARGS)
279 {
280 	u_int64_t qdrops_long;
281 	int error, qdrops;
282 
283 	netisr_getqdrops(&ip_direct_nh, &qdrops_long);
284 	qdrops = qdrops_long;
285 	error = sysctl_handle_int(oidp, &qdrops, 0, req);
286 	if (error || !req->newptr)
287 		return (error);
288 	if (qdrops != 0)
289 		return (EINVAL);
290 	netisr_clearqdrops(&ip_direct_nh);
291 	return (0);
292 }
293 
294 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRDQDROPS, intr_direct_queue_drops,
295     CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, 0, 0,
296     sysctl_netinet_intr_direct_queue_drops, "I",
297     "Number of packets dropped from the IP direct input queue");
298 #endif	/* RSS */
299 
300 /*
301  * IP initialization: fill in IP protocol switch table.
302  * All protocols not implemented in kernel go to raw IP protocol handler.
303  */
304 void
305 ip_init(void)
306 {
307 	struct pfil_head_args args;
308 	struct protosw *pr;
309 	int i;
310 
311 	CK_STAILQ_INIT(&V_in_ifaddrhead);
312 	V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask);
313 
314 	/* Initialize IP reassembly queue. */
315 	ipreass_init();
316 
317 	/* Initialize packet filter hooks. */
318 	args.pa_version = PFIL_VERSION;
319 	args.pa_flags = PFIL_IN | PFIL_OUT;
320 	args.pa_type = PFIL_TYPE_IP4;
321 	args.pa_headname = PFIL_INET_NAME;
322 	V_inet_pfil_head = pfil_head_register(&args);
323 
324 	if (hhook_head_register(HHOOK_TYPE_IPSEC_IN, AF_INET,
325 	    &V_ipsec_hhh_in[HHOOK_IPSEC_INET],
326 	    HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
327 		printf("%s: WARNING: unable to register input helper hook\n",
328 		    __func__);
329 	if (hhook_head_register(HHOOK_TYPE_IPSEC_OUT, AF_INET,
330 	    &V_ipsec_hhh_out[HHOOK_IPSEC_INET],
331 	    HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
332 		printf("%s: WARNING: unable to register output helper hook\n",
333 		    __func__);
334 
335 	/* Skip initialization of globals for non-default instances. */
336 #ifdef VIMAGE
337 	if (!IS_DEFAULT_VNET(curvnet)) {
338 		netisr_register_vnet(&ip_nh);
339 #ifdef	RSS
340 		netisr_register_vnet(&ip_direct_nh);
341 #endif
342 		return;
343 	}
344 #endif
345 
346 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
347 	if (pr == NULL)
348 		panic("ip_init: PF_INET not found");
349 
350 	/* Initialize the entire ip_protox[] array to IPPROTO_RAW. */
351 	for (i = 0; i < IPPROTO_MAX; i++)
352 		ip_protox[i] = pr - inetsw;
353 	/*
354 	 * Cycle through IP protocols and put them into the appropriate place
355 	 * in ip_protox[].
356 	 */
357 	for (pr = inetdomain.dom_protosw;
358 	    pr < inetdomain.dom_protoswNPROTOSW; pr++)
359 		if (pr->pr_domain->dom_family == PF_INET &&
360 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
361 			/* Be careful to only index valid IP protocols. */
362 			if (pr->pr_protocol < IPPROTO_MAX)
363 				ip_protox[pr->pr_protocol] = pr - inetsw;
364 		}
365 
366 	netisr_register(&ip_nh);
367 #ifdef	RSS
368 	netisr_register(&ip_direct_nh);
369 #endif
370 }
371 
372 #ifdef VIMAGE
373 static void
374 ip_destroy(void *unused __unused)
375 {
376 	int error;
377 
378 #ifdef	RSS
379 	netisr_unregister_vnet(&ip_direct_nh);
380 #endif
381 	netisr_unregister_vnet(&ip_nh);
382 
383 	pfil_head_unregister(V_inet_pfil_head);
384 	error = hhook_head_deregister(V_ipsec_hhh_in[HHOOK_IPSEC_INET]);
385 	if (error != 0) {
386 		printf("%s: WARNING: unable to deregister input helper hook "
387 		    "type HHOOK_TYPE_IPSEC_IN, id HHOOK_IPSEC_INET: "
388 		    "error %d returned\n", __func__, error);
389 	}
390 	error = hhook_head_deregister(V_ipsec_hhh_out[HHOOK_IPSEC_INET]);
391 	if (error != 0) {
392 		printf("%s: WARNING: unable to deregister output helper hook "
393 		    "type HHOOK_TYPE_IPSEC_OUT, id HHOOK_IPSEC_INET: "
394 		    "error %d returned\n", __func__, error);
395 	}
396 
397 	/* Remove the IPv4 addresses from all interfaces. */
398 	in_ifscrub_all();
399 
400 	/* Make sure the IPv4 routes are gone as well. */
401 	rib_flush_routes_family(AF_INET);
402 
403 	/* Destroy IP reassembly queue. */
404 	ipreass_destroy();
405 
406 	/* Cleanup in_ifaddr hash table; should be empty. */
407 	hashdestroy(V_in_ifaddrhashtbl, M_IFADDR, V_in_ifaddrhmask);
408 }
409 
410 VNET_SYSUNINIT(ip, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip_destroy, NULL);
411 #endif
412 
413 #ifdef	RSS
414 /*
415  * IP direct input routine.
416  *
417  * This is called when reinjecting completed fragments where
418  * all of the previous checking and book-keeping has been done.
419  */
420 void
421 ip_direct_input(struct mbuf *m)
422 {
423 	struct ip *ip;
424 	int hlen;
425 
426 	ip = mtod(m, struct ip *);
427 	hlen = ip->ip_hl << 2;
428 
429 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
430 	if (IPSEC_ENABLED(ipv4)) {
431 		if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
432 			return;
433 	}
434 #endif /* IPSEC */
435 	IPSTAT_INC(ips_delivered);
436 	(*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
437 	return;
438 }
439 #endif
440 
441 /*
442  * Ip input routine.  Checksum and byte swap header.  If fragmented
443  * try to reassemble.  Process options.  Pass to next level.
444  */
445 void
446 ip_input(struct mbuf *m)
447 {
448 	MROUTER_RLOCK_TRACKER;
449 	struct ip *ip = NULL;
450 	struct in_ifaddr *ia = NULL;
451 	struct ifaddr *ifa;
452 	struct ifnet *ifp;
453 	int hlen = 0;
454 	uint16_t sum, ip_len;
455 	int dchg = 0;				/* dest changed after fw */
456 	struct in_addr odst;			/* original dst address */
457 	bool strong_es;
458 
459 	M_ASSERTPKTHDR(m);
460 	NET_EPOCH_ASSERT();
461 
462 	if (m->m_flags & M_FASTFWD_OURS) {
463 		m->m_flags &= ~M_FASTFWD_OURS;
464 		/* Set up some basics that will be used later. */
465 		ip = mtod(m, struct ip *);
466 		hlen = ip->ip_hl << 2;
467 		ip_len = ntohs(ip->ip_len);
468 		goto ours;
469 	}
470 
471 	IPSTAT_INC(ips_total);
472 
473 	if (__predict_false(m->m_pkthdr.len < sizeof(struct ip)))
474 		goto tooshort;
475 
476 	if (m->m_len < sizeof(struct ip)) {
477 		m = m_pullup(m, sizeof(struct ip));
478 		if (__predict_false(m == NULL)) {
479 			IPSTAT_INC(ips_toosmall);
480 			return;
481 		}
482 	}
483 	ip = mtod(m, struct ip *);
484 
485 	if (__predict_false(ip->ip_v != IPVERSION)) {
486 		IPSTAT_INC(ips_badvers);
487 		goto bad;
488 	}
489 
490 	hlen = ip->ip_hl << 2;
491 	if (__predict_false(hlen < sizeof(struct ip))) {	/* minimum header length */
492 		IPSTAT_INC(ips_badhlen);
493 		goto bad;
494 	}
495 	if (hlen > m->m_len) {
496 		m = m_pullup(m, hlen);
497 		if (__predict_false(m == NULL)) {
498 			IPSTAT_INC(ips_badhlen);
499 			return;
500 		}
501 		ip = mtod(m, struct ip *);
502 	}
503 
504 	IP_PROBE(receive, NULL, NULL, ip, m->m_pkthdr.rcvif, ip, NULL);
505 
506 	/* IN_LOOPBACK must not appear on the wire - RFC1122 */
507 	ifp = m->m_pkthdr.rcvif;
508 	if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) ||
509 	    IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) {
510 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
511 			IPSTAT_INC(ips_badaddr);
512 			goto bad;
513 		}
514 	}
515 
516 	if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
517 		sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
518 	} else {
519 		if (hlen == sizeof(struct ip)) {
520 			sum = in_cksum_hdr(ip);
521 		} else {
522 			sum = in_cksum(m, hlen);
523 		}
524 	}
525 	if (__predict_false(sum)) {
526 		IPSTAT_INC(ips_badsum);
527 		goto bad;
528 	}
529 
530 #ifdef ALTQ
531 	if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
532 		/* packet is dropped by traffic conditioner */
533 		return;
534 #endif
535 
536 	ip_len = ntohs(ip->ip_len);
537 	if (__predict_false(ip_len < hlen)) {
538 		IPSTAT_INC(ips_badlen);
539 		goto bad;
540 	}
541 
542 	/*
543 	 * Check that the amount of data in the buffers
544 	 * is as at least much as the IP header would have us expect.
545 	 * Trim mbufs if longer than we expect.
546 	 * Drop packet if shorter than we expect.
547 	 */
548 	if (__predict_false(m->m_pkthdr.len < ip_len)) {
549 tooshort:
550 		IPSTAT_INC(ips_tooshort);
551 		goto bad;
552 	}
553 	if (m->m_pkthdr.len > ip_len) {
554 		if (m->m_len == m->m_pkthdr.len) {
555 			m->m_len = ip_len;
556 			m->m_pkthdr.len = ip_len;
557 		} else
558 			m_adj(m, ip_len - m->m_pkthdr.len);
559 	}
560 
561 	/*
562 	 * Try to forward the packet, but if we fail continue.
563 	 * ip_tryforward() does not generate redirects, so fall
564 	 * through to normal processing if redirects are required.
565 	 * ip_tryforward() does inbound and outbound packet firewall
566 	 * processing. If firewall has decided that destination becomes
567 	 * our local address, it sets M_FASTFWD_OURS flag. In this
568 	 * case skip another inbound firewall processing and update
569 	 * ip pointer.
570 	 */
571 	if (V_ipforwarding != 0
572 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
573 	    && (!IPSEC_ENABLED(ipv4) ||
574 	    IPSEC_CAPS(ipv4, m, IPSEC_CAP_OPERABLE) == 0)
575 #endif
576 	    ) {
577 		if ((m = ip_tryforward(m)) == NULL)
578 			return;
579 		if (m->m_flags & M_FASTFWD_OURS) {
580 			m->m_flags &= ~M_FASTFWD_OURS;
581 			ip = mtod(m, struct ip *);
582 			goto ours;
583 		}
584 	}
585 
586 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
587 	/*
588 	 * Bypass packet filtering for packets previously handled by IPsec.
589 	 */
590 	if (IPSEC_ENABLED(ipv4) &&
591 	    IPSEC_CAPS(ipv4, m, IPSEC_CAP_BYPASS_FILTER) != 0)
592 			goto passin;
593 #endif
594 
595 	/*
596 	 * Run through list of hooks for input packets.
597 	 *
598 	 * NB: Beware of the destination address changing (e.g.
599 	 *     by NAT rewriting).  When this happens, tell
600 	 *     ip_forward to do the right thing.
601 	 */
602 
603 	/* Jump over all PFIL processing if hooks are not active. */
604 	if (!PFIL_HOOKED_IN(V_inet_pfil_head))
605 		goto passin;
606 
607 	odst = ip->ip_dst;
608 	if (pfil_run_hooks(V_inet_pfil_head, &m, ifp, PFIL_IN, NULL) !=
609 	    PFIL_PASS)
610 		return;
611 	if (m == NULL)			/* consumed by filter */
612 		return;
613 
614 	ip = mtod(m, struct ip *);
615 	dchg = (odst.s_addr != ip->ip_dst.s_addr);
616 
617 	if (m->m_flags & M_FASTFWD_OURS) {
618 		m->m_flags &= ~M_FASTFWD_OURS;
619 		goto ours;
620 	}
621 	if (m->m_flags & M_IP_NEXTHOP) {
622 		if (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) {
623 			/*
624 			 * Directly ship the packet on.  This allows
625 			 * forwarding packets originally destined to us
626 			 * to some other directly connected host.
627 			 */
628 			ip_forward(m, 1);
629 			return;
630 		}
631 	}
632 passin:
633 
634 	/*
635 	 * Process options and, if not destined for us,
636 	 * ship it on.  ip_dooptions returns 1 when an
637 	 * error was detected (causing an icmp message
638 	 * to be sent and the original packet to be freed).
639 	 */
640 	if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
641 		return;
642 
643         /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
644          * matter if it is destined to another node, or whether it is
645          * a multicast one, RSVP wants it! and prevents it from being forwarded
646          * anywhere else. Also checks if the rsvp daemon is running before
647 	 * grabbing the packet.
648          */
649 	if (ip->ip_p == IPPROTO_RSVP && V_rsvp_on)
650 		goto ours;
651 
652 	/*
653 	 * Check our list of addresses, to see if the packet is for us.
654 	 * If we don't have any addresses, assume any unicast packet
655 	 * we receive might be for us (and let the upper layers deal
656 	 * with it).
657 	 */
658 	if (CK_STAILQ_EMPTY(&V_in_ifaddrhead) &&
659 	    (m->m_flags & (M_MCAST|M_BCAST)) == 0)
660 		goto ours;
661 
662 	/*
663 	 * Enable a consistency check between the destination address
664 	 * and the arrival interface for a unicast packet (the RFC 1122
665 	 * strong ES model) with a list of additional predicates:
666 	 * - if IP forwarding is disabled
667 	 * - the packet is not locally generated
668 	 * - the packet is not subject to 'ipfw fwd'
669 	 * - Interface is not running CARP. If the packet got here, we already
670 	 *   checked it with carp_iamatch() and carp_forus().
671 	 */
672 	strong_es = V_ip_strong_es && (V_ipforwarding == 0) &&
673 	    ((ifp->if_flags & IFF_LOOPBACK) == 0) &&
674 	    ifp->if_carp == NULL && (dchg == 0);
675 
676 	/*
677 	 * Check for exact addresses in the hash bucket.
678 	 */
679 	CK_LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
680 		if (IA_SIN(ia)->sin_addr.s_addr != ip->ip_dst.s_addr)
681 			continue;
682 
683 		/*
684 		 * net.inet.ip.rfc1122_strong_es: the address matches, verify
685 		 * that the packet arrived via the correct interface.
686 		 */
687 		if (__predict_false(strong_es && ia->ia_ifp != ifp)) {
688 			IPSTAT_INC(ips_badaddr);
689 			goto bad;
690 		}
691 
692 		/*
693 		 * net.inet.ip.source_address_validation: drop incoming
694 		 * packets that pretend to be ours.
695 		 */
696 		if (V_ip_sav && !(ifp->if_flags & IFF_LOOPBACK) &&
697 		    __predict_false(in_localip_fib(ip->ip_src, ifp->if_fib))) {
698 			IPSTAT_INC(ips_badaddr);
699 			goto bad;
700 		}
701 
702 		counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
703 		counter_u64_add(ia->ia_ifa.ifa_ibytes, m->m_pkthdr.len);
704 		goto ours;
705 	}
706 
707 	/*
708 	 * Check for broadcast addresses.
709 	 *
710 	 * Only accept broadcast packets that arrive via the matching
711 	 * interface.  Reception of forwarded directed broadcasts would
712 	 * be handled via ip_forward() and ether_output() with the loopback
713 	 * into the stack for SIMPLEX interfaces handled by ether_output().
714 	 */
715 	if (ifp->if_flags & IFF_BROADCAST) {
716 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
717 			if (ifa->ifa_addr->sa_family != AF_INET)
718 				continue;
719 			ia = ifatoia(ifa);
720 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
721 			    ip->ip_dst.s_addr) {
722 				counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
723 				counter_u64_add(ia->ia_ifa.ifa_ibytes,
724 				    m->m_pkthdr.len);
725 				goto ours;
726 			}
727 #ifdef BOOTP_COMPAT
728 			if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) {
729 				counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
730 				counter_u64_add(ia->ia_ifa.ifa_ibytes,
731 				    m->m_pkthdr.len);
732 				goto ours;
733 			}
734 #endif
735 		}
736 		ia = NULL;
737 	}
738 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
739 		MROUTER_RLOCK();
740 		/*
741 		 * RFC 3927 2.7: Do not forward multicast packets from
742 		 * IN_LINKLOCAL.
743 		 */
744 		if (V_ip_mrouter && !IN_LINKLOCAL(ntohl(ip->ip_src.s_addr))) {
745 			/*
746 			 * If we are acting as a multicast router, all
747 			 * incoming multicast packets are passed to the
748 			 * kernel-level multicast forwarding function.
749 			 * The packet is returned (relatively) intact; if
750 			 * ip_mforward() returns a non-zero value, the packet
751 			 * must be discarded, else it may be accepted below.
752 			 */
753 			if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) {
754 				MROUTER_RUNLOCK();
755 				IPSTAT_INC(ips_cantforward);
756 				m_freem(m);
757 				return;
758 			}
759 
760 			/*
761 			 * The process-level routing daemon needs to receive
762 			 * all multicast IGMP packets, whether or not this
763 			 * host belongs to their destination groups.
764 			 */
765 			if (ip->ip_p == IPPROTO_IGMP) {
766 				MROUTER_RUNLOCK();
767 				goto ours;
768 			}
769 			IPSTAT_INC(ips_forward);
770 		}
771 		MROUTER_RUNLOCK();
772 		/*
773 		 * Assume the packet is for us, to avoid prematurely taking
774 		 * a lock on the in_multi hash. Protocols must perform
775 		 * their own filtering and update statistics accordingly.
776 		 */
777 		goto ours;
778 	}
779 	if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
780 		goto ours;
781 	if (ip->ip_dst.s_addr == INADDR_ANY)
782 		goto ours;
783 	/* RFC 3927 2.7: Do not forward packets to or from IN_LINKLOCAL. */
784 	if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr)) ||
785 	    IN_LINKLOCAL(ntohl(ip->ip_src.s_addr))) {
786 		IPSTAT_INC(ips_cantforward);
787 		m_freem(m);
788 		return;
789 	}
790 
791 	/*
792 	 * Not for us; forward if possible and desirable.
793 	 */
794 	if (V_ipforwarding == 0) {
795 		IPSTAT_INC(ips_cantforward);
796 		m_freem(m);
797 	} else {
798 		ip_forward(m, dchg);
799 	}
800 	return;
801 
802 ours:
803 #ifdef IPSTEALTH
804 	/*
805 	 * IPSTEALTH: Process non-routing options only
806 	 * if the packet is destined for us.
807 	 */
808 	if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1))
809 		return;
810 #endif /* IPSTEALTH */
811 
812 	/*
813 	 * Attempt reassembly; if it succeeds, proceed.
814 	 * ip_reass() will return a different mbuf.
815 	 */
816 	if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) {
817 		/* XXXGL: shouldn't we save & set m_flags? */
818 		m = ip_reass(m);
819 		if (m == NULL)
820 			return;
821 		ip = mtod(m, struct ip *);
822 		/* Get the header length of the reassembled packet */
823 		hlen = ip->ip_hl << 2;
824 	}
825 
826 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
827 	if (IPSEC_ENABLED(ipv4)) {
828 		if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
829 			return;
830 	}
831 #endif /* IPSEC */
832 
833 	/*
834 	 * Switch out to protocol's input routine.
835 	 */
836 	IPSTAT_INC(ips_delivered);
837 
838 	(*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
839 	return;
840 bad:
841 	m_freem(m);
842 }
843 
844 /*
845  * IP timer processing;
846  * if a timer expires on a reassembly
847  * queue, discard it.
848  */
849 void
850 ip_slowtimo(void)
851 {
852 	VNET_ITERATOR_DECL(vnet_iter);
853 
854 	VNET_LIST_RLOCK_NOSLEEP();
855 	VNET_FOREACH(vnet_iter) {
856 		CURVNET_SET(vnet_iter);
857 		ipreass_slowtimo();
858 		CURVNET_RESTORE();
859 	}
860 	VNET_LIST_RUNLOCK_NOSLEEP();
861 }
862 
863 void
864 ip_drain(void)
865 {
866 	VNET_ITERATOR_DECL(vnet_iter);
867 
868 	VNET_LIST_RLOCK_NOSLEEP();
869 	VNET_FOREACH(vnet_iter) {
870 		CURVNET_SET(vnet_iter);
871 		ipreass_drain();
872 		CURVNET_RESTORE();
873 	}
874 	VNET_LIST_RUNLOCK_NOSLEEP();
875 }
876 
877 /*
878  * The protocol to be inserted into ip_protox[] must be already registered
879  * in inetsw[], either statically or through pf_proto_register().
880  */
881 int
882 ipproto_register(short ipproto)
883 {
884 	struct protosw *pr;
885 
886 	/* Sanity checks. */
887 	if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
888 		return (EPROTONOSUPPORT);
889 
890 	/*
891 	 * The protocol slot must not be occupied by another protocol
892 	 * already.  An index pointing to IPPROTO_RAW is unused.
893 	 */
894 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
895 	if (pr == NULL)
896 		return (EPFNOSUPPORT);
897 	if (ip_protox[ipproto] != pr - inetsw)	/* IPPROTO_RAW */
898 		return (EEXIST);
899 
900 	/* Find the protocol position in inetsw[] and set the index. */
901 	for (pr = inetdomain.dom_protosw;
902 	     pr < inetdomain.dom_protoswNPROTOSW; pr++) {
903 		if (pr->pr_domain->dom_family == PF_INET &&
904 		    pr->pr_protocol && pr->pr_protocol == ipproto) {
905 			ip_protox[pr->pr_protocol] = pr - inetsw;
906 			return (0);
907 		}
908 	}
909 	return (EPROTONOSUPPORT);
910 }
911 
912 int
913 ipproto_unregister(short ipproto)
914 {
915 	struct protosw *pr;
916 
917 	/* Sanity checks. */
918 	if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
919 		return (EPROTONOSUPPORT);
920 
921 	/* Check if the protocol was indeed registered. */
922 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
923 	if (pr == NULL)
924 		return (EPFNOSUPPORT);
925 	if (ip_protox[ipproto] == pr - inetsw)  /* IPPROTO_RAW */
926 		return (ENOENT);
927 
928 	/* Reset the protocol slot to IPPROTO_RAW. */
929 	ip_protox[ipproto] = pr - inetsw;
930 	return (0);
931 }
932 
933 u_char inetctlerrmap[PRC_NCMDS] = {
934 	0,		0,		0,		0,
935 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
936 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
937 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
938 	0,		0,		EHOSTUNREACH,	0,
939 	ENOPROTOOPT,	ECONNREFUSED
940 };
941 
942 /*
943  * Forward a packet.  If some error occurs return the sender
944  * an icmp packet.  Note we can't always generate a meaningful
945  * icmp message because icmp doesn't have a large enough repertoire
946  * of codes and types.
947  *
948  * If not forwarding, just drop the packet.  This could be confusing
949  * if ipforwarding was zero but some routing protocol was advancing
950  * us as a gateway to somewhere.  However, we must let the routing
951  * protocol deal with that.
952  *
953  * The srcrt parameter indicates whether the packet is being forwarded
954  * via a source route.
955  */
956 void
957 ip_forward(struct mbuf *m, int srcrt)
958 {
959 	struct ip *ip = mtod(m, struct ip *);
960 	struct in_ifaddr *ia;
961 	struct mbuf *mcopy;
962 	struct sockaddr_in *sin;
963 	struct in_addr dest;
964 	struct route ro;
965 	uint32_t flowid;
966 	int error, type = 0, code = 0, mtu = 0;
967 
968 	NET_EPOCH_ASSERT();
969 
970 	if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
971 		IPSTAT_INC(ips_cantforward);
972 		m_freem(m);
973 		return;
974 	}
975 	if (
976 #ifdef IPSTEALTH
977 	    V_ipstealth == 0 &&
978 #endif
979 	    ip->ip_ttl <= IPTTLDEC) {
980 		icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 0, 0);
981 		return;
982 	}
983 
984 	bzero(&ro, sizeof(ro));
985 	sin = (struct sockaddr_in *)&ro.ro_dst;
986 	sin->sin_family = AF_INET;
987 	sin->sin_len = sizeof(*sin);
988 	sin->sin_addr = ip->ip_dst;
989 	flowid = m->m_pkthdr.flowid;
990 	ro.ro_nh = fib4_lookup(M_GETFIB(m), ip->ip_dst, 0, NHR_REF, flowid);
991 	if (ro.ro_nh != NULL) {
992 		ia = ifatoia(ro.ro_nh->nh_ifa);
993 	} else
994 		ia = NULL;
995 	/*
996 	 * Save the IP header and at most 8 bytes of the payload,
997 	 * in case we need to generate an ICMP message to the src.
998 	 *
999 	 * XXX this can be optimized a lot by saving the data in a local
1000 	 * buffer on the stack (72 bytes at most), and only allocating the
1001 	 * mbuf if really necessary. The vast majority of the packets
1002 	 * are forwarded without having to send an ICMP back (either
1003 	 * because unnecessary, or because rate limited), so we are
1004 	 * really we are wasting a lot of work here.
1005 	 *
1006 	 * We don't use m_copym() because it might return a reference
1007 	 * to a shared cluster. Both this function and ip_output()
1008 	 * assume exclusive access to the IP header in `m', so any
1009 	 * data in a cluster may change before we reach icmp_error().
1010 	 */
1011 	mcopy = m_gethdr(M_NOWAIT, m->m_type);
1012 	if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_NOWAIT)) {
1013 		/*
1014 		 * It's probably ok if the pkthdr dup fails (because
1015 		 * the deep copy of the tag chain failed), but for now
1016 		 * be conservative and just discard the copy since
1017 		 * code below may some day want the tags.
1018 		 */
1019 		m_free(mcopy);
1020 		mcopy = NULL;
1021 	}
1022 	if (mcopy != NULL) {
1023 		mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy));
1024 		mcopy->m_pkthdr.len = mcopy->m_len;
1025 		m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1026 	}
1027 #ifdef IPSTEALTH
1028 	if (V_ipstealth == 0)
1029 #endif
1030 		ip->ip_ttl -= IPTTLDEC;
1031 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1032 	if (IPSEC_ENABLED(ipv4)) {
1033 		if ((error = IPSEC_FORWARD(ipv4, m)) != 0) {
1034 			/* mbuf consumed by IPsec */
1035 			RO_NHFREE(&ro);
1036 			m_freem(mcopy);
1037 			if (error != EINPROGRESS)
1038 				IPSTAT_INC(ips_cantforward);
1039 			return;
1040 		}
1041 		/* No IPsec processing required */
1042 	}
1043 #endif /* IPSEC */
1044 	/*
1045 	 * If forwarding packet using same interface that it came in on,
1046 	 * perhaps should send a redirect to sender to shortcut a hop.
1047 	 * Only send redirect if source is sending directly to us,
1048 	 * and if packet was not source routed (or has any options).
1049 	 * Also, don't send redirect if forwarding using a default route
1050 	 * or a route modified by a redirect.
1051 	 */
1052 	dest.s_addr = 0;
1053 	if (!srcrt && V_ipsendredirects &&
1054 	    ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) {
1055 		struct nhop_object *nh;
1056 
1057 		nh = ro.ro_nh;
1058 
1059 		if (nh != NULL && ((nh->nh_flags & (NHF_REDIRECT|NHF_DEFAULT)) == 0)) {
1060 			struct in_ifaddr *nh_ia = (struct in_ifaddr *)(nh->nh_ifa);
1061 			u_long src = ntohl(ip->ip_src.s_addr);
1062 
1063 			if (nh_ia != NULL &&
1064 			    (src & nh_ia->ia_subnetmask) == nh_ia->ia_subnet) {
1065 				/* Router requirements says to only send host redirects */
1066 				type = ICMP_REDIRECT;
1067 				code = ICMP_REDIRECT_HOST;
1068 				if (nh->nh_flags & NHF_GATEWAY) {
1069 				    if (nh->gw_sa.sa_family == AF_INET)
1070 					dest.s_addr = nh->gw4_sa.sin_addr.s_addr;
1071 				    else /* Do not redirect in case gw is AF_INET6 */
1072 					type = 0;
1073 				} else
1074 					dest.s_addr = ip->ip_dst.s_addr;
1075 			}
1076 		}
1077 	}
1078 
1079 	error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
1080 
1081 	if (error == EMSGSIZE && ro.ro_nh)
1082 		mtu = ro.ro_nh->nh_mtu;
1083 	RO_NHFREE(&ro);
1084 
1085 	if (error)
1086 		IPSTAT_INC(ips_cantforward);
1087 	else {
1088 		IPSTAT_INC(ips_forward);
1089 		if (type)
1090 			IPSTAT_INC(ips_redirectsent);
1091 		else {
1092 			if (mcopy)
1093 				m_freem(mcopy);
1094 			return;
1095 		}
1096 	}
1097 	if (mcopy == NULL)
1098 		return;
1099 
1100 	switch (error) {
1101 	case 0:				/* forwarded, but need redirect */
1102 		/* type, code set above */
1103 		break;
1104 
1105 	case ENETUNREACH:
1106 	case EHOSTUNREACH:
1107 	case ENETDOWN:
1108 	case EHOSTDOWN:
1109 	default:
1110 		type = ICMP_UNREACH;
1111 		code = ICMP_UNREACH_HOST;
1112 		break;
1113 
1114 	case EMSGSIZE:
1115 		type = ICMP_UNREACH;
1116 		code = ICMP_UNREACH_NEEDFRAG;
1117 		/*
1118 		 * If the MTU was set before make sure we are below the
1119 		 * interface MTU.
1120 		 * If the MTU wasn't set before use the interface mtu or
1121 		 * fall back to the next smaller mtu step compared to the
1122 		 * current packet size.
1123 		 */
1124 		if (mtu != 0) {
1125 			if (ia != NULL)
1126 				mtu = min(mtu, ia->ia_ifp->if_mtu);
1127 		} else {
1128 			if (ia != NULL)
1129 				mtu = ia->ia_ifp->if_mtu;
1130 			else
1131 				mtu = ip_next_mtu(ntohs(ip->ip_len), 0);
1132 		}
1133 		IPSTAT_INC(ips_cantfrag);
1134 		break;
1135 
1136 	case ENOBUFS:
1137 	case EACCES:			/* ipfw denied packet */
1138 		m_freem(mcopy);
1139 		return;
1140 	}
1141 	icmp_error(mcopy, type, code, dest.s_addr, mtu);
1142 }
1143 
1144 #define	CHECK_SO_CT(sp, ct) \
1145     (((sp->so_options & SO_TIMESTAMP) && (sp->so_ts_clock == ct)) ? 1 : 0)
1146 
1147 void
1148 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1149     struct mbuf *m)
1150 {
1151 	bool stamped;
1152 
1153 	stamped = false;
1154 	if ((inp->inp_socket->so_options & SO_BINTIME) ||
1155 	    CHECK_SO_CT(inp->inp_socket, SO_TS_BINTIME)) {
1156 		struct bintime boottimebin, bt;
1157 		struct timespec ts1;
1158 
1159 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1160 		    M_TSTMP)) {
1161 			mbuf_tstmp2timespec(m, &ts1);
1162 			timespec2bintime(&ts1, &bt);
1163 			getboottimebin(&boottimebin);
1164 			bintime_add(&bt, &boottimebin);
1165 		} else {
1166 			bintime(&bt);
1167 		}
1168 		*mp = sbcreatecontrol((caddr_t)&bt, sizeof(bt),
1169 		    SCM_BINTIME, SOL_SOCKET);
1170 		if (*mp != NULL) {
1171 			mp = &(*mp)->m_next;
1172 			stamped = true;
1173 		}
1174 	}
1175 	if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME_MICRO)) {
1176 		struct bintime boottimebin, bt1;
1177 		struct timespec ts1;
1178 		struct timeval tv;
1179 
1180 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1181 		    M_TSTMP)) {
1182 			mbuf_tstmp2timespec(m, &ts1);
1183 			timespec2bintime(&ts1, &bt1);
1184 			getboottimebin(&boottimebin);
1185 			bintime_add(&bt1, &boottimebin);
1186 			bintime2timeval(&bt1, &tv);
1187 		} else {
1188 			microtime(&tv);
1189 		}
1190 		*mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv),
1191 		    SCM_TIMESTAMP, SOL_SOCKET);
1192 		if (*mp != NULL) {
1193 			mp = &(*mp)->m_next;
1194 			stamped = true;
1195 		}
1196 	} else if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME)) {
1197 		struct bintime boottimebin;
1198 		struct timespec ts, ts1;
1199 
1200 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1201 		    M_TSTMP)) {
1202 			mbuf_tstmp2timespec(m, &ts);
1203 			getboottimebin(&boottimebin);
1204 			bintime2timespec(&boottimebin, &ts1);
1205 			timespecadd(&ts, &ts1, &ts);
1206 		} else {
1207 			nanotime(&ts);
1208 		}
1209 		*mp = sbcreatecontrol((caddr_t)&ts, sizeof(ts),
1210 		    SCM_REALTIME, SOL_SOCKET);
1211 		if (*mp != NULL) {
1212 			mp = &(*mp)->m_next;
1213 			stamped = true;
1214 		}
1215 	} else if (CHECK_SO_CT(inp->inp_socket, SO_TS_MONOTONIC)) {
1216 		struct timespec ts;
1217 
1218 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1219 		    M_TSTMP))
1220 			mbuf_tstmp2timespec(m, &ts);
1221 		else
1222 			nanouptime(&ts);
1223 		*mp = sbcreatecontrol((caddr_t)&ts, sizeof(ts),
1224 		    SCM_MONOTONIC, SOL_SOCKET);
1225 		if (*mp != NULL) {
1226 			mp = &(*mp)->m_next;
1227 			stamped = true;
1228 		}
1229 	}
1230 	if (stamped && (m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1231 	    M_TSTMP)) {
1232 		struct sock_timestamp_info sti;
1233 
1234 		bzero(&sti, sizeof(sti));
1235 		sti.st_info_flags = ST_INFO_HW;
1236 		if ((m->m_flags & M_TSTMP_HPREC) != 0)
1237 			sti.st_info_flags |= ST_INFO_HW_HPREC;
1238 		*mp = sbcreatecontrol((caddr_t)&sti, sizeof(sti), SCM_TIME_INFO,
1239 		    SOL_SOCKET);
1240 		if (*mp != NULL)
1241 			mp = &(*mp)->m_next;
1242 	}
1243 	if (inp->inp_flags & INP_RECVDSTADDR) {
1244 		*mp = sbcreatecontrol((caddr_t)&ip->ip_dst,
1245 		    sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1246 		if (*mp)
1247 			mp = &(*mp)->m_next;
1248 	}
1249 	if (inp->inp_flags & INP_RECVTTL) {
1250 		*mp = sbcreatecontrol((caddr_t)&ip->ip_ttl,
1251 		    sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
1252 		if (*mp)
1253 			mp = &(*mp)->m_next;
1254 	}
1255 #ifdef notyet
1256 	/* XXX
1257 	 * Moving these out of udp_input() made them even more broken
1258 	 * than they already were.
1259 	 */
1260 	/* options were tossed already */
1261 	if (inp->inp_flags & INP_RECVOPTS) {
1262 		*mp = sbcreatecontrol((caddr_t)opts_deleted_above,
1263 		    sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1264 		if (*mp)
1265 			mp = &(*mp)->m_next;
1266 	}
1267 	/* ip_srcroute doesn't do what we want here, need to fix */
1268 	if (inp->inp_flags & INP_RECVRETOPTS) {
1269 		*mp = sbcreatecontrol((caddr_t)ip_srcroute(m),
1270 		    sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1271 		if (*mp)
1272 			mp = &(*mp)->m_next;
1273 	}
1274 #endif
1275 	if (inp->inp_flags & INP_RECVIF) {
1276 		struct ifnet *ifp;
1277 		struct sdlbuf {
1278 			struct sockaddr_dl sdl;
1279 			u_char	pad[32];
1280 		} sdlbuf;
1281 		struct sockaddr_dl *sdp;
1282 		struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1283 
1284 		if ((ifp = m->m_pkthdr.rcvif) &&
1285 		    ifp->if_index && ifp->if_index <= V_if_index) {
1286 			sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1287 			/*
1288 			 * Change our mind and don't try copy.
1289 			 */
1290 			if (sdp->sdl_family != AF_LINK ||
1291 			    sdp->sdl_len > sizeof(sdlbuf)) {
1292 				goto makedummy;
1293 			}
1294 			bcopy(sdp, sdl2, sdp->sdl_len);
1295 		} else {
1296 makedummy:
1297 			sdl2->sdl_len =
1298 			    offsetof(struct sockaddr_dl, sdl_data[0]);
1299 			sdl2->sdl_family = AF_LINK;
1300 			sdl2->sdl_index = 0;
1301 			sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1302 		}
1303 		*mp = sbcreatecontrol((caddr_t)sdl2, sdl2->sdl_len,
1304 		    IP_RECVIF, IPPROTO_IP);
1305 		if (*mp)
1306 			mp = &(*mp)->m_next;
1307 	}
1308 	if (inp->inp_flags & INP_RECVTOS) {
1309 		*mp = sbcreatecontrol((caddr_t)&ip->ip_tos,
1310 		    sizeof(u_char), IP_RECVTOS, IPPROTO_IP);
1311 		if (*mp)
1312 			mp = &(*mp)->m_next;
1313 	}
1314 
1315 	if (inp->inp_flags2 & INP_RECVFLOWID) {
1316 		uint32_t flowid, flow_type;
1317 
1318 		flowid = m->m_pkthdr.flowid;
1319 		flow_type = M_HASHTYPE_GET(m);
1320 
1321 		/*
1322 		 * XXX should handle the failure of one or the
1323 		 * other - don't populate both?
1324 		 */
1325 		*mp = sbcreatecontrol((caddr_t) &flowid,
1326 		    sizeof(uint32_t), IP_FLOWID, IPPROTO_IP);
1327 		if (*mp)
1328 			mp = &(*mp)->m_next;
1329 		*mp = sbcreatecontrol((caddr_t) &flow_type,
1330 		    sizeof(uint32_t), IP_FLOWTYPE, IPPROTO_IP);
1331 		if (*mp)
1332 			mp = &(*mp)->m_next;
1333 	}
1334 
1335 #ifdef	RSS
1336 	if (inp->inp_flags2 & INP_RECVRSSBUCKETID) {
1337 		uint32_t flowid, flow_type;
1338 		uint32_t rss_bucketid;
1339 
1340 		flowid = m->m_pkthdr.flowid;
1341 		flow_type = M_HASHTYPE_GET(m);
1342 
1343 		if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) {
1344 			*mp = sbcreatecontrol((caddr_t) &rss_bucketid,
1345 			   sizeof(uint32_t), IP_RSSBUCKETID, IPPROTO_IP);
1346 			if (*mp)
1347 				mp = &(*mp)->m_next;
1348 		}
1349 	}
1350 #endif
1351 }
1352 
1353 /*
1354  * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
1355  * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
1356  * locking.  This code remains in ip_input.c as ip_mroute.c is optionally
1357  * compiled.
1358  */
1359 VNET_DEFINE_STATIC(int, ip_rsvp_on);
1360 VNET_DEFINE(struct socket *, ip_rsvpd);
1361 
1362 #define	V_ip_rsvp_on		VNET(ip_rsvp_on)
1363 
1364 int
1365 ip_rsvp_init(struct socket *so)
1366 {
1367 
1368 	if (so->so_type != SOCK_RAW ||
1369 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
1370 		return EOPNOTSUPP;
1371 
1372 	if (V_ip_rsvpd != NULL)
1373 		return EADDRINUSE;
1374 
1375 	V_ip_rsvpd = so;
1376 	/*
1377 	 * This may seem silly, but we need to be sure we don't over-increment
1378 	 * the RSVP counter, in case something slips up.
1379 	 */
1380 	if (!V_ip_rsvp_on) {
1381 		V_ip_rsvp_on = 1;
1382 		V_rsvp_on++;
1383 	}
1384 
1385 	return 0;
1386 }
1387 
1388 int
1389 ip_rsvp_done(void)
1390 {
1391 
1392 	V_ip_rsvpd = NULL;
1393 	/*
1394 	 * This may seem silly, but we need to be sure we don't over-decrement
1395 	 * the RSVP counter, in case something slips up.
1396 	 */
1397 	if (V_ip_rsvp_on) {
1398 		V_ip_rsvp_on = 0;
1399 		V_rsvp_on--;
1400 	}
1401 	return 0;
1402 }
1403 
1404 int
1405 rsvp_input(struct mbuf **mp, int *offp, int proto)
1406 {
1407 	struct mbuf *m;
1408 
1409 	m = *mp;
1410 	*mp = NULL;
1411 
1412 	if (rsvp_input_p) { /* call the real one if loaded */
1413 		*mp = m;
1414 		rsvp_input_p(mp, offp, proto);
1415 		return (IPPROTO_DONE);
1416 	}
1417 
1418 	/* Can still get packets with rsvp_on = 0 if there is a local member
1419 	 * of the group to which the RSVP packet is addressed.  But in this
1420 	 * case we want to throw the packet away.
1421 	 */
1422 
1423 	if (!V_rsvp_on) {
1424 		m_freem(m);
1425 		return (IPPROTO_DONE);
1426 	}
1427 
1428 	if (V_ip_rsvpd != NULL) {
1429 		*mp = m;
1430 		rip_input(mp, offp, proto);
1431 		return (IPPROTO_DONE);
1432 	}
1433 	/* Drop the packet */
1434 	m_freem(m);
1435 	return (IPPROTO_DONE);
1436 }
1437