xref: /freebsd/sys/netinet/ip_input.c (revision 963f5dc7a30624e95d72fb7f87b8892651164e46)
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() may generate redirects these days.
564 	 * XXX the logic below falling through to normal processing
565 	 * if redirects are required should be revisited as well.
566 	 * ip_tryforward() does inbound and outbound packet firewall
567 	 * processing. If firewall has decided that destination becomes
568 	 * our local address, it sets M_FASTFWD_OURS flag. In this
569 	 * case skip another inbound firewall processing and update
570 	 * ip pointer.
571 	 */
572 	if (V_ipforwarding != 0
573 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
574 	    && (!IPSEC_ENABLED(ipv4) ||
575 	    IPSEC_CAPS(ipv4, m, IPSEC_CAP_OPERABLE) == 0)
576 #endif
577 	    ) {
578 		/*
579 		 * ip_dooptions() was run so we can ignore the source route (or
580 		 * any IP options case) case for redirects in ip_tryforward().
581 		 */
582 		if ((m = ip_tryforward(m)) == NULL)
583 			return;
584 		if (m->m_flags & M_FASTFWD_OURS) {
585 			m->m_flags &= ~M_FASTFWD_OURS;
586 			ip = mtod(m, struct ip *);
587 			goto ours;
588 		}
589 	}
590 
591 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
592 	/*
593 	 * Bypass packet filtering for packets previously handled by IPsec.
594 	 */
595 	if (IPSEC_ENABLED(ipv4) &&
596 	    IPSEC_CAPS(ipv4, m, IPSEC_CAP_BYPASS_FILTER) != 0)
597 			goto passin;
598 #endif
599 
600 	/*
601 	 * Run through list of hooks for input packets.
602 	 *
603 	 * NB: Beware of the destination address changing (e.g.
604 	 *     by NAT rewriting).  When this happens, tell
605 	 *     ip_forward to do the right thing.
606 	 */
607 
608 	/* Jump over all PFIL processing if hooks are not active. */
609 	if (!PFIL_HOOKED_IN(V_inet_pfil_head))
610 		goto passin;
611 
612 	odst = ip->ip_dst;
613 	if (pfil_run_hooks(V_inet_pfil_head, &m, ifp, PFIL_IN, NULL) !=
614 	    PFIL_PASS)
615 		return;
616 	if (m == NULL)			/* consumed by filter */
617 		return;
618 
619 	ip = mtod(m, struct ip *);
620 	dchg = (odst.s_addr != ip->ip_dst.s_addr);
621 
622 	if (m->m_flags & M_FASTFWD_OURS) {
623 		m->m_flags &= ~M_FASTFWD_OURS;
624 		goto ours;
625 	}
626 	if (m->m_flags & M_IP_NEXTHOP) {
627 		if (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) {
628 			/*
629 			 * Directly ship the packet on.  This allows
630 			 * forwarding packets originally destined to us
631 			 * to some other directly connected host.
632 			 */
633 			ip_forward(m, 1);
634 			return;
635 		}
636 	}
637 passin:
638 
639 	/*
640 	 * Process options and, if not destined for us,
641 	 * ship it on.  ip_dooptions returns 1 when an
642 	 * error was detected (causing an icmp message
643 	 * to be sent and the original packet to be freed).
644 	 */
645 	if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
646 		return;
647 
648         /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
649          * matter if it is destined to another node, or whether it is
650          * a multicast one, RSVP wants it! and prevents it from being forwarded
651          * anywhere else. Also checks if the rsvp daemon is running before
652 	 * grabbing the packet.
653          */
654 	if (ip->ip_p == IPPROTO_RSVP && V_rsvp_on)
655 		goto ours;
656 
657 	/*
658 	 * Check our list of addresses, to see if the packet is for us.
659 	 * If we don't have any addresses, assume any unicast packet
660 	 * we receive might be for us (and let the upper layers deal
661 	 * with it).
662 	 */
663 	if (CK_STAILQ_EMPTY(&V_in_ifaddrhead) &&
664 	    (m->m_flags & (M_MCAST|M_BCAST)) == 0)
665 		goto ours;
666 
667 	/*
668 	 * Enable a consistency check between the destination address
669 	 * and the arrival interface for a unicast packet (the RFC 1122
670 	 * strong ES model) with a list of additional predicates:
671 	 * - if IP forwarding is disabled
672 	 * - the packet is not locally generated
673 	 * - the packet is not subject to 'ipfw fwd'
674 	 * - Interface is not running CARP. If the packet got here, we already
675 	 *   checked it with carp_iamatch() and carp_forus().
676 	 */
677 	strong_es = V_ip_strong_es && (V_ipforwarding == 0) &&
678 	    ((ifp->if_flags & IFF_LOOPBACK) == 0) &&
679 	    ifp->if_carp == NULL && (dchg == 0);
680 
681 	/*
682 	 * Check for exact addresses in the hash bucket.
683 	 */
684 	CK_LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
685 		if (IA_SIN(ia)->sin_addr.s_addr != ip->ip_dst.s_addr)
686 			continue;
687 
688 		/*
689 		 * net.inet.ip.rfc1122_strong_es: the address matches, verify
690 		 * that the packet arrived via the correct interface.
691 		 */
692 		if (__predict_false(strong_es && ia->ia_ifp != ifp)) {
693 			IPSTAT_INC(ips_badaddr);
694 			goto bad;
695 		}
696 
697 		/*
698 		 * net.inet.ip.source_address_validation: drop incoming
699 		 * packets that pretend to be ours.
700 		 */
701 		if (V_ip_sav && !(ifp->if_flags & IFF_LOOPBACK) &&
702 		    __predict_false(in_localip_fib(ip->ip_src, ifp->if_fib))) {
703 			IPSTAT_INC(ips_badaddr);
704 			goto bad;
705 		}
706 
707 		counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
708 		counter_u64_add(ia->ia_ifa.ifa_ibytes, m->m_pkthdr.len);
709 		goto ours;
710 	}
711 
712 	/*
713 	 * Check for broadcast addresses.
714 	 *
715 	 * Only accept broadcast packets that arrive via the matching
716 	 * interface.  Reception of forwarded directed broadcasts would
717 	 * be handled via ip_forward() and ether_output() with the loopback
718 	 * into the stack for SIMPLEX interfaces handled by ether_output().
719 	 */
720 	if (ifp->if_flags & IFF_BROADCAST) {
721 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
722 			if (ifa->ifa_addr->sa_family != AF_INET)
723 				continue;
724 			ia = ifatoia(ifa);
725 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
726 			    ip->ip_dst.s_addr) {
727 				counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
728 				counter_u64_add(ia->ia_ifa.ifa_ibytes,
729 				    m->m_pkthdr.len);
730 				goto ours;
731 			}
732 #ifdef BOOTP_COMPAT
733 			if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) {
734 				counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
735 				counter_u64_add(ia->ia_ifa.ifa_ibytes,
736 				    m->m_pkthdr.len);
737 				goto ours;
738 			}
739 #endif
740 		}
741 		ia = NULL;
742 	}
743 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
744 		MROUTER_RLOCK();
745 		/*
746 		 * RFC 3927 2.7: Do not forward multicast packets from
747 		 * IN_LINKLOCAL.
748 		 */
749 		if (V_ip_mrouter && !IN_LINKLOCAL(ntohl(ip->ip_src.s_addr))) {
750 			/*
751 			 * If we are acting as a multicast router, all
752 			 * incoming multicast packets are passed to the
753 			 * kernel-level multicast forwarding function.
754 			 * The packet is returned (relatively) intact; if
755 			 * ip_mforward() returns a non-zero value, the packet
756 			 * must be discarded, else it may be accepted below.
757 			 */
758 			if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) {
759 				MROUTER_RUNLOCK();
760 				IPSTAT_INC(ips_cantforward);
761 				m_freem(m);
762 				return;
763 			}
764 
765 			/*
766 			 * The process-level routing daemon needs to receive
767 			 * all multicast IGMP packets, whether or not this
768 			 * host belongs to their destination groups.
769 			 */
770 			if (ip->ip_p == IPPROTO_IGMP) {
771 				MROUTER_RUNLOCK();
772 				goto ours;
773 			}
774 			IPSTAT_INC(ips_forward);
775 		}
776 		MROUTER_RUNLOCK();
777 		/*
778 		 * Assume the packet is for us, to avoid prematurely taking
779 		 * a lock on the in_multi hash. Protocols must perform
780 		 * their own filtering and update statistics accordingly.
781 		 */
782 		goto ours;
783 	}
784 	if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
785 		goto ours;
786 	if (ip->ip_dst.s_addr == INADDR_ANY)
787 		goto ours;
788 	/* RFC 3927 2.7: Do not forward packets to or from IN_LINKLOCAL. */
789 	if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr)) ||
790 	    IN_LINKLOCAL(ntohl(ip->ip_src.s_addr))) {
791 		IPSTAT_INC(ips_cantforward);
792 		m_freem(m);
793 		return;
794 	}
795 
796 	/*
797 	 * Not for us; forward if possible and desirable.
798 	 */
799 	if (V_ipforwarding == 0) {
800 		IPSTAT_INC(ips_cantforward);
801 		m_freem(m);
802 	} else {
803 		ip_forward(m, dchg);
804 	}
805 	return;
806 
807 ours:
808 #ifdef IPSTEALTH
809 	/*
810 	 * IPSTEALTH: Process non-routing options only
811 	 * if the packet is destined for us.
812 	 */
813 	if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1))
814 		return;
815 #endif /* IPSTEALTH */
816 
817 	/*
818 	 * Attempt reassembly; if it succeeds, proceed.
819 	 * ip_reass() will return a different mbuf.
820 	 */
821 	if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) {
822 		/* XXXGL: shouldn't we save & set m_flags? */
823 		m = ip_reass(m);
824 		if (m == NULL)
825 			return;
826 		ip = mtod(m, struct ip *);
827 		/* Get the header length of the reassembled packet */
828 		hlen = ip->ip_hl << 2;
829 	}
830 
831 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
832 	if (IPSEC_ENABLED(ipv4)) {
833 		if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
834 			return;
835 	}
836 #endif /* IPSEC */
837 
838 	/*
839 	 * Switch out to protocol's input routine.
840 	 */
841 	IPSTAT_INC(ips_delivered);
842 
843 	(*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
844 	return;
845 bad:
846 	m_freem(m);
847 }
848 
849 /*
850  * IP timer processing;
851  * if a timer expires on a reassembly
852  * queue, discard it.
853  */
854 void
855 ip_slowtimo(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_slowtimo();
863 		CURVNET_RESTORE();
864 	}
865 	VNET_LIST_RUNLOCK_NOSLEEP();
866 }
867 
868 void
869 ip_drain(void)
870 {
871 	VNET_ITERATOR_DECL(vnet_iter);
872 
873 	VNET_LIST_RLOCK_NOSLEEP();
874 	VNET_FOREACH(vnet_iter) {
875 		CURVNET_SET(vnet_iter);
876 		ipreass_drain();
877 		CURVNET_RESTORE();
878 	}
879 	VNET_LIST_RUNLOCK_NOSLEEP();
880 }
881 
882 /*
883  * The protocol to be inserted into ip_protox[] must be already registered
884  * in inetsw[], either statically or through pf_proto_register().
885  */
886 int
887 ipproto_register(short ipproto)
888 {
889 	struct protosw *pr;
890 
891 	/* Sanity checks. */
892 	if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
893 		return (EPROTONOSUPPORT);
894 
895 	/*
896 	 * The protocol slot must not be occupied by another protocol
897 	 * already.  An index pointing to IPPROTO_RAW is unused.
898 	 */
899 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
900 	if (pr == NULL)
901 		return (EPFNOSUPPORT);
902 	if (ip_protox[ipproto] != pr - inetsw)	/* IPPROTO_RAW */
903 		return (EEXIST);
904 
905 	/* Find the protocol position in inetsw[] and set the index. */
906 	for (pr = inetdomain.dom_protosw;
907 	     pr < inetdomain.dom_protoswNPROTOSW; pr++) {
908 		if (pr->pr_domain->dom_family == PF_INET &&
909 		    pr->pr_protocol && pr->pr_protocol == ipproto) {
910 			ip_protox[pr->pr_protocol] = pr - inetsw;
911 			return (0);
912 		}
913 	}
914 	return (EPROTONOSUPPORT);
915 }
916 
917 int
918 ipproto_unregister(short ipproto)
919 {
920 	struct protosw *pr;
921 
922 	/* Sanity checks. */
923 	if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
924 		return (EPROTONOSUPPORT);
925 
926 	/* Check if the protocol was indeed registered. */
927 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
928 	if (pr == NULL)
929 		return (EPFNOSUPPORT);
930 	if (ip_protox[ipproto] == pr - inetsw)  /* IPPROTO_RAW */
931 		return (ENOENT);
932 
933 	/* Reset the protocol slot to IPPROTO_RAW. */
934 	ip_protox[ipproto] = pr - inetsw;
935 	return (0);
936 }
937 
938 u_char inetctlerrmap[PRC_NCMDS] = {
939 	0,		0,		0,		0,
940 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
941 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
942 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
943 	0,		0,		EHOSTUNREACH,	0,
944 	ENOPROTOOPT,	ECONNREFUSED
945 };
946 
947 /*
948  * Forward a packet.  If some error occurs return the sender
949  * an icmp packet.  Note we can't always generate a meaningful
950  * icmp message because icmp doesn't have a large enough repertoire
951  * of codes and types.
952  *
953  * If not forwarding, just drop the packet.  This could be confusing
954  * if ipforwarding was zero but some routing protocol was advancing
955  * us as a gateway to somewhere.  However, we must let the routing
956  * protocol deal with that.
957  *
958  * The srcrt parameter indicates whether the packet is being forwarded
959  * via a source route.
960  */
961 void
962 ip_forward(struct mbuf *m, int srcrt)
963 {
964 	struct ip *ip = mtod(m, struct ip *);
965 	struct in_ifaddr *ia;
966 	struct mbuf *mcopy;
967 	struct sockaddr_in *sin;
968 	struct in_addr dest;
969 	struct route ro;
970 	uint32_t flowid;
971 	int error, type = 0, code = 0, mtu = 0;
972 
973 	NET_EPOCH_ASSERT();
974 
975 	if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
976 		IPSTAT_INC(ips_cantforward);
977 		m_freem(m);
978 		return;
979 	}
980 	if (
981 #ifdef IPSTEALTH
982 	    V_ipstealth == 0 &&
983 #endif
984 	    ip->ip_ttl <= IPTTLDEC) {
985 		icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 0, 0);
986 		return;
987 	}
988 
989 	bzero(&ro, sizeof(ro));
990 	sin = (struct sockaddr_in *)&ro.ro_dst;
991 	sin->sin_family = AF_INET;
992 	sin->sin_len = sizeof(*sin);
993 	sin->sin_addr = ip->ip_dst;
994 	flowid = m->m_pkthdr.flowid;
995 	ro.ro_nh = fib4_lookup(M_GETFIB(m), ip->ip_dst, 0, NHR_REF, flowid);
996 	if (ro.ro_nh != NULL) {
997 		ia = ifatoia(ro.ro_nh->nh_ifa);
998 	} else
999 		ia = NULL;
1000 	/*
1001 	 * Save the IP header and at most 8 bytes of the payload,
1002 	 * in case we need to generate an ICMP message to the src.
1003 	 *
1004 	 * XXX this can be optimized a lot by saving the data in a local
1005 	 * buffer on the stack (72 bytes at most), and only allocating the
1006 	 * mbuf if really necessary. The vast majority of the packets
1007 	 * are forwarded without having to send an ICMP back (either
1008 	 * because unnecessary, or because rate limited), so we are
1009 	 * really we are wasting a lot of work here.
1010 	 *
1011 	 * We don't use m_copym() because it might return a reference
1012 	 * to a shared cluster. Both this function and ip_output()
1013 	 * assume exclusive access to the IP header in `m', so any
1014 	 * data in a cluster may change before we reach icmp_error().
1015 	 */
1016 	mcopy = m_gethdr(M_NOWAIT, m->m_type);
1017 	if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_NOWAIT)) {
1018 		/*
1019 		 * It's probably ok if the pkthdr dup fails (because
1020 		 * the deep copy of the tag chain failed), but for now
1021 		 * be conservative and just discard the copy since
1022 		 * code below may some day want the tags.
1023 		 */
1024 		m_free(mcopy);
1025 		mcopy = NULL;
1026 	}
1027 	if (mcopy != NULL) {
1028 		mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy));
1029 		mcopy->m_pkthdr.len = mcopy->m_len;
1030 		m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1031 	}
1032 #ifdef IPSTEALTH
1033 	if (V_ipstealth == 0)
1034 #endif
1035 		ip->ip_ttl -= IPTTLDEC;
1036 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1037 	if (IPSEC_ENABLED(ipv4)) {
1038 		if ((error = IPSEC_FORWARD(ipv4, m)) != 0) {
1039 			/* mbuf consumed by IPsec */
1040 			RO_NHFREE(&ro);
1041 			m_freem(mcopy);
1042 			if (error != EINPROGRESS)
1043 				IPSTAT_INC(ips_cantforward);
1044 			return;
1045 		}
1046 		/* No IPsec processing required */
1047 	}
1048 #endif /* IPSEC */
1049 	/*
1050 	 * If forwarding packet using same interface that it came in on,
1051 	 * perhaps should send a redirect to sender to shortcut a hop.
1052 	 * Only send redirect if source is sending directly to us,
1053 	 * and if packet was not source routed (or has any options).
1054 	 * Also, don't send redirect if forwarding using a default route
1055 	 * or a route modified by a redirect.
1056 	 */
1057 	dest.s_addr = 0;
1058 	if (!srcrt && V_ipsendredirects &&
1059 	    ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) {
1060 		struct nhop_object *nh;
1061 
1062 		nh = ro.ro_nh;
1063 
1064 		if (nh != NULL && ((nh->nh_flags & (NHF_REDIRECT|NHF_DEFAULT)) == 0)) {
1065 			struct in_ifaddr *nh_ia = (struct in_ifaddr *)(nh->nh_ifa);
1066 			u_long src = ntohl(ip->ip_src.s_addr);
1067 
1068 			if (nh_ia != NULL &&
1069 			    (src & nh_ia->ia_subnetmask) == nh_ia->ia_subnet) {
1070 				/* Router requirements says to only send host redirects */
1071 				type = ICMP_REDIRECT;
1072 				code = ICMP_REDIRECT_HOST;
1073 				if (nh->nh_flags & NHF_GATEWAY) {
1074 				    if (nh->gw_sa.sa_family == AF_INET)
1075 					dest.s_addr = nh->gw4_sa.sin_addr.s_addr;
1076 				    else /* Do not redirect in case gw is AF_INET6 */
1077 					type = 0;
1078 				} else
1079 					dest.s_addr = ip->ip_dst.s_addr;
1080 			}
1081 		}
1082 	}
1083 
1084 	error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
1085 
1086 	if (error == EMSGSIZE && ro.ro_nh)
1087 		mtu = ro.ro_nh->nh_mtu;
1088 	RO_NHFREE(&ro);
1089 
1090 	if (error)
1091 		IPSTAT_INC(ips_cantforward);
1092 	else {
1093 		IPSTAT_INC(ips_forward);
1094 		if (type)
1095 			IPSTAT_INC(ips_redirectsent);
1096 		else {
1097 			if (mcopy)
1098 				m_freem(mcopy);
1099 			return;
1100 		}
1101 	}
1102 	if (mcopy == NULL)
1103 		return;
1104 
1105 	switch (error) {
1106 	case 0:				/* forwarded, but need redirect */
1107 		/* type, code set above */
1108 		break;
1109 
1110 	case ENETUNREACH:
1111 	case EHOSTUNREACH:
1112 	case ENETDOWN:
1113 	case EHOSTDOWN:
1114 	default:
1115 		type = ICMP_UNREACH;
1116 		code = ICMP_UNREACH_HOST;
1117 		break;
1118 
1119 	case EMSGSIZE:
1120 		type = ICMP_UNREACH;
1121 		code = ICMP_UNREACH_NEEDFRAG;
1122 		/*
1123 		 * If the MTU was set before make sure we are below the
1124 		 * interface MTU.
1125 		 * If the MTU wasn't set before use the interface mtu or
1126 		 * fall back to the next smaller mtu step compared to the
1127 		 * current packet size.
1128 		 */
1129 		if (mtu != 0) {
1130 			if (ia != NULL)
1131 				mtu = min(mtu, ia->ia_ifp->if_mtu);
1132 		} else {
1133 			if (ia != NULL)
1134 				mtu = ia->ia_ifp->if_mtu;
1135 			else
1136 				mtu = ip_next_mtu(ntohs(ip->ip_len), 0);
1137 		}
1138 		IPSTAT_INC(ips_cantfrag);
1139 		break;
1140 
1141 	case ENOBUFS:
1142 	case EACCES:			/* ipfw denied packet */
1143 		m_freem(mcopy);
1144 		return;
1145 	}
1146 	icmp_error(mcopy, type, code, dest.s_addr, mtu);
1147 }
1148 
1149 #define	CHECK_SO_CT(sp, ct) \
1150     (((sp->so_options & SO_TIMESTAMP) && (sp->so_ts_clock == ct)) ? 1 : 0)
1151 
1152 void
1153 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1154     struct mbuf *m)
1155 {
1156 	bool stamped;
1157 
1158 	stamped = false;
1159 	if ((inp->inp_socket->so_options & SO_BINTIME) ||
1160 	    CHECK_SO_CT(inp->inp_socket, SO_TS_BINTIME)) {
1161 		struct bintime boottimebin, bt;
1162 		struct timespec ts1;
1163 
1164 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1165 		    M_TSTMP)) {
1166 			mbuf_tstmp2timespec(m, &ts1);
1167 			timespec2bintime(&ts1, &bt);
1168 			getboottimebin(&boottimebin);
1169 			bintime_add(&bt, &boottimebin);
1170 		} else {
1171 			bintime(&bt);
1172 		}
1173 		*mp = sbcreatecontrol((caddr_t)&bt, sizeof(bt),
1174 		    SCM_BINTIME, SOL_SOCKET);
1175 		if (*mp != NULL) {
1176 			mp = &(*mp)->m_next;
1177 			stamped = true;
1178 		}
1179 	}
1180 	if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME_MICRO)) {
1181 		struct bintime boottimebin, bt1;
1182 		struct timespec ts1;
1183 		struct timeval tv;
1184 
1185 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1186 		    M_TSTMP)) {
1187 			mbuf_tstmp2timespec(m, &ts1);
1188 			timespec2bintime(&ts1, &bt1);
1189 			getboottimebin(&boottimebin);
1190 			bintime_add(&bt1, &boottimebin);
1191 			bintime2timeval(&bt1, &tv);
1192 		} else {
1193 			microtime(&tv);
1194 		}
1195 		*mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv),
1196 		    SCM_TIMESTAMP, SOL_SOCKET);
1197 		if (*mp != NULL) {
1198 			mp = &(*mp)->m_next;
1199 			stamped = true;
1200 		}
1201 	} else if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME)) {
1202 		struct bintime boottimebin;
1203 		struct timespec ts, ts1;
1204 
1205 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1206 		    M_TSTMP)) {
1207 			mbuf_tstmp2timespec(m, &ts);
1208 			getboottimebin(&boottimebin);
1209 			bintime2timespec(&boottimebin, &ts1);
1210 			timespecadd(&ts, &ts1, &ts);
1211 		} else {
1212 			nanotime(&ts);
1213 		}
1214 		*mp = sbcreatecontrol((caddr_t)&ts, sizeof(ts),
1215 		    SCM_REALTIME, SOL_SOCKET);
1216 		if (*mp != NULL) {
1217 			mp = &(*mp)->m_next;
1218 			stamped = true;
1219 		}
1220 	} else if (CHECK_SO_CT(inp->inp_socket, SO_TS_MONOTONIC)) {
1221 		struct timespec ts;
1222 
1223 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1224 		    M_TSTMP))
1225 			mbuf_tstmp2timespec(m, &ts);
1226 		else
1227 			nanouptime(&ts);
1228 		*mp = sbcreatecontrol((caddr_t)&ts, sizeof(ts),
1229 		    SCM_MONOTONIC, SOL_SOCKET);
1230 		if (*mp != NULL) {
1231 			mp = &(*mp)->m_next;
1232 			stamped = true;
1233 		}
1234 	}
1235 	if (stamped && (m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1236 	    M_TSTMP)) {
1237 		struct sock_timestamp_info sti;
1238 
1239 		bzero(&sti, sizeof(sti));
1240 		sti.st_info_flags = ST_INFO_HW;
1241 		if ((m->m_flags & M_TSTMP_HPREC) != 0)
1242 			sti.st_info_flags |= ST_INFO_HW_HPREC;
1243 		*mp = sbcreatecontrol((caddr_t)&sti, sizeof(sti), SCM_TIME_INFO,
1244 		    SOL_SOCKET);
1245 		if (*mp != NULL)
1246 			mp = &(*mp)->m_next;
1247 	}
1248 	if (inp->inp_flags & INP_RECVDSTADDR) {
1249 		*mp = sbcreatecontrol((caddr_t)&ip->ip_dst,
1250 		    sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1251 		if (*mp)
1252 			mp = &(*mp)->m_next;
1253 	}
1254 	if (inp->inp_flags & INP_RECVTTL) {
1255 		*mp = sbcreatecontrol((caddr_t)&ip->ip_ttl,
1256 		    sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
1257 		if (*mp)
1258 			mp = &(*mp)->m_next;
1259 	}
1260 #ifdef notyet
1261 	/* XXX
1262 	 * Moving these out of udp_input() made them even more broken
1263 	 * than they already were.
1264 	 */
1265 	/* options were tossed already */
1266 	if (inp->inp_flags & INP_RECVOPTS) {
1267 		*mp = sbcreatecontrol((caddr_t)opts_deleted_above,
1268 		    sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1269 		if (*mp)
1270 			mp = &(*mp)->m_next;
1271 	}
1272 	/* ip_srcroute doesn't do what we want here, need to fix */
1273 	if (inp->inp_flags & INP_RECVRETOPTS) {
1274 		*mp = sbcreatecontrol((caddr_t)ip_srcroute(m),
1275 		    sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1276 		if (*mp)
1277 			mp = &(*mp)->m_next;
1278 	}
1279 #endif
1280 	if (inp->inp_flags & INP_RECVIF) {
1281 		struct ifnet *ifp;
1282 		struct sdlbuf {
1283 			struct sockaddr_dl sdl;
1284 			u_char	pad[32];
1285 		} sdlbuf;
1286 		struct sockaddr_dl *sdp;
1287 		struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1288 
1289 		if ((ifp = m->m_pkthdr.rcvif)) {
1290 			sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1291 			/*
1292 			 * Change our mind and don't try copy.
1293 			 */
1294 			if (sdp->sdl_family != AF_LINK ||
1295 			    sdp->sdl_len > sizeof(sdlbuf)) {
1296 				goto makedummy;
1297 			}
1298 			bcopy(sdp, sdl2, sdp->sdl_len);
1299 		} else {
1300 makedummy:
1301 			sdl2->sdl_len =
1302 			    offsetof(struct sockaddr_dl, sdl_data[0]);
1303 			sdl2->sdl_family = AF_LINK;
1304 			sdl2->sdl_index = 0;
1305 			sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1306 		}
1307 		*mp = sbcreatecontrol((caddr_t)sdl2, sdl2->sdl_len,
1308 		    IP_RECVIF, IPPROTO_IP);
1309 		if (*mp)
1310 			mp = &(*mp)->m_next;
1311 	}
1312 	if (inp->inp_flags & INP_RECVTOS) {
1313 		*mp = sbcreatecontrol((caddr_t)&ip->ip_tos,
1314 		    sizeof(u_char), IP_RECVTOS, IPPROTO_IP);
1315 		if (*mp)
1316 			mp = &(*mp)->m_next;
1317 	}
1318 
1319 	if (inp->inp_flags2 & INP_RECVFLOWID) {
1320 		uint32_t flowid, flow_type;
1321 
1322 		flowid = m->m_pkthdr.flowid;
1323 		flow_type = M_HASHTYPE_GET(m);
1324 
1325 		/*
1326 		 * XXX should handle the failure of one or the
1327 		 * other - don't populate both?
1328 		 */
1329 		*mp = sbcreatecontrol((caddr_t) &flowid,
1330 		    sizeof(uint32_t), IP_FLOWID, IPPROTO_IP);
1331 		if (*mp)
1332 			mp = &(*mp)->m_next;
1333 		*mp = sbcreatecontrol((caddr_t) &flow_type,
1334 		    sizeof(uint32_t), IP_FLOWTYPE, IPPROTO_IP);
1335 		if (*mp)
1336 			mp = &(*mp)->m_next;
1337 	}
1338 
1339 #ifdef	RSS
1340 	if (inp->inp_flags2 & INP_RECVRSSBUCKETID) {
1341 		uint32_t flowid, flow_type;
1342 		uint32_t rss_bucketid;
1343 
1344 		flowid = m->m_pkthdr.flowid;
1345 		flow_type = M_HASHTYPE_GET(m);
1346 
1347 		if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) {
1348 			*mp = sbcreatecontrol((caddr_t) &rss_bucketid,
1349 			   sizeof(uint32_t), IP_RSSBUCKETID, IPPROTO_IP);
1350 			if (*mp)
1351 				mp = &(*mp)->m_next;
1352 		}
1353 	}
1354 #endif
1355 }
1356 
1357 /*
1358  * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
1359  * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
1360  * locking.  This code remains in ip_input.c as ip_mroute.c is optionally
1361  * compiled.
1362  */
1363 VNET_DEFINE_STATIC(int, ip_rsvp_on);
1364 VNET_DEFINE(struct socket *, ip_rsvpd);
1365 
1366 #define	V_ip_rsvp_on		VNET(ip_rsvp_on)
1367 
1368 int
1369 ip_rsvp_init(struct socket *so)
1370 {
1371 
1372 	if (so->so_type != SOCK_RAW ||
1373 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
1374 		return EOPNOTSUPP;
1375 
1376 	if (V_ip_rsvpd != NULL)
1377 		return EADDRINUSE;
1378 
1379 	V_ip_rsvpd = so;
1380 	/*
1381 	 * This may seem silly, but we need to be sure we don't over-increment
1382 	 * the RSVP counter, in case something slips up.
1383 	 */
1384 	if (!V_ip_rsvp_on) {
1385 		V_ip_rsvp_on = 1;
1386 		V_rsvp_on++;
1387 	}
1388 
1389 	return 0;
1390 }
1391 
1392 int
1393 ip_rsvp_done(void)
1394 {
1395 
1396 	V_ip_rsvpd = NULL;
1397 	/*
1398 	 * This may seem silly, but we need to be sure we don't over-decrement
1399 	 * the RSVP counter, in case something slips up.
1400 	 */
1401 	if (V_ip_rsvp_on) {
1402 		V_ip_rsvp_on = 0;
1403 		V_rsvp_on--;
1404 	}
1405 	return 0;
1406 }
1407 
1408 int
1409 rsvp_input(struct mbuf **mp, int *offp, int proto)
1410 {
1411 	struct mbuf *m;
1412 
1413 	m = *mp;
1414 	*mp = NULL;
1415 
1416 	if (rsvp_input_p) { /* call the real one if loaded */
1417 		*mp = m;
1418 		rsvp_input_p(mp, offp, proto);
1419 		return (IPPROTO_DONE);
1420 	}
1421 
1422 	/* Can still get packets with rsvp_on = 0 if there is a local member
1423 	 * of the group to which the RSVP packet is addressed.  But in this
1424 	 * case we want to throw the packet away.
1425 	 */
1426 
1427 	if (!V_rsvp_on) {
1428 		m_freem(m);
1429 		return (IPPROTO_DONE);
1430 	}
1431 
1432 	if (V_ip_rsvpd != NULL) {
1433 		*mp = m;
1434 		rip_input(mp, offp, proto);
1435 		return (IPPROTO_DONE);
1436 	}
1437 	/* Drop the packet */
1438 	m_freem(m);
1439 	return (IPPROTO_DONE);
1440 }
1441