xref: /freebsd/sys/netinet/ip_input.c (revision 8cc087a1eee9ec1ca9f7ac1e63ad51bdb5a682eb)
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 VNET_DEFINE(int, rsvp_on);
105 
106 VNET_DEFINE(int, ipforwarding);
107 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_VNET | CTLFLAG_RW,
108     &VNET_NAME(ipforwarding), 0,
109     "Enable IP forwarding between interfaces");
110 
111 /*
112  * Respond with an ICMP host redirect when we forward a packet out of
113  * the same interface on which it was received.  See RFC 792.
114  */
115 VNET_DEFINE(int, ipsendredirects) = 1;
116 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_VNET | CTLFLAG_RW,
117     &VNET_NAME(ipsendredirects), 0,
118     "Enable sending IP redirects");
119 
120 VNET_DEFINE_STATIC(bool, ip_strong_es) = false;
121 #define	V_ip_strong_es	VNET(ip_strong_es)
122 SYSCTL_BOOL(_net_inet_ip, OID_AUTO, rfc1122_strong_es,
123     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip_strong_es), false,
124     "Packet's IP destination address must match address on arrival interface");
125 
126 VNET_DEFINE_STATIC(bool, ip_sav) = true;
127 #define	V_ip_sav	VNET(ip_sav)
128 SYSCTL_BOOL(_net_inet_ip, OID_AUTO, source_address_validation,
129     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip_sav), true,
130     "Drop incoming packets with source address that is a local address");
131 
132 VNET_DEFINE(pfil_head_t, inet_pfil_head);	/* Packet filter hooks */
133 
134 static struct netisr_handler ip_nh = {
135 	.nh_name = "ip",
136 	.nh_handler = ip_input,
137 	.nh_proto = NETISR_IP,
138 #ifdef	RSS
139 	.nh_m2cpuid = rss_soft_m2cpuid_v4,
140 	.nh_policy = NETISR_POLICY_CPU,
141 	.nh_dispatch = NETISR_DISPATCH_HYBRID,
142 #else
143 	.nh_policy = NETISR_POLICY_FLOW,
144 #endif
145 };
146 
147 #ifdef	RSS
148 /*
149  * Directly dispatched frames are currently assumed
150  * to have a flowid already calculated.
151  *
152  * It should likely have something that assert it
153  * actually has valid flow details.
154  */
155 static struct netisr_handler ip_direct_nh = {
156 	.nh_name = "ip_direct",
157 	.nh_handler = ip_direct_input,
158 	.nh_proto = NETISR_IP_DIRECT,
159 	.nh_m2cpuid = rss_soft_m2cpuid_v4,
160 	.nh_policy = NETISR_POLICY_CPU,
161 	.nh_dispatch = NETISR_DISPATCH_HYBRID,
162 };
163 #endif
164 
165 extern	struct domain inetdomain;
166 extern	struct protosw inetsw[];
167 u_char	ip_protox[IPPROTO_MAX];
168 VNET_DEFINE(struct in_ifaddrhead, in_ifaddrhead);  /* first inet address */
169 VNET_DEFINE(struct in_ifaddrhashhead *, in_ifaddrhashtbl); /* inet addr hash table  */
170 VNET_DEFINE(u_long, in_ifaddrhmask);		/* mask for hash table */
171 
172 /* Make sure it is safe to use hashinit(9) on CK_LIST. */
173 CTASSERT(sizeof(struct in_ifaddrhashhead) == sizeof(LIST_HEAD(, in_addr)));
174 
175 #ifdef IPCTL_DEFMTU
176 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
177     &ip_mtu, 0, "Default MTU");
178 #endif
179 
180 #ifdef IPSTEALTH
181 VNET_DEFINE(int, ipstealth);
182 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_VNET | CTLFLAG_RW,
183     &VNET_NAME(ipstealth), 0,
184     "IP stealth mode, no TTL decrementation on forwarding");
185 #endif
186 
187 /*
188  * IP statistics are stored in the "array" of counter(9)s.
189  */
190 VNET_PCPUSTAT_DEFINE(struct ipstat, ipstat);
191 VNET_PCPUSTAT_SYSINIT(ipstat);
192 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, IPCTL_STATS, stats, struct ipstat, ipstat,
193     "IP statistics (struct ipstat, netinet/ip_var.h)");
194 
195 #ifdef VIMAGE
196 VNET_PCPUSTAT_SYSUNINIT(ipstat);
197 #endif /* VIMAGE */
198 
199 /*
200  * Kernel module interface for updating ipstat.  The argument is an index
201  * into ipstat treated as an array.
202  */
203 void
204 kmod_ipstat_inc(int statnum)
205 {
206 
207 	counter_u64_add(VNET(ipstat)[statnum], 1);
208 }
209 
210 void
211 kmod_ipstat_dec(int statnum)
212 {
213 
214 	counter_u64_add(VNET(ipstat)[statnum], -1);
215 }
216 
217 static int
218 sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS)
219 {
220 	int error, qlimit;
221 
222 	netisr_getqlimit(&ip_nh, &qlimit);
223 	error = sysctl_handle_int(oidp, &qlimit, 0, req);
224 	if (error || !req->newptr)
225 		return (error);
226 	if (qlimit < 1)
227 		return (EINVAL);
228 	return (netisr_setqlimit(&ip_nh, qlimit));
229 }
230 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen,
231     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 0,
232     sysctl_netinet_intr_queue_maxlen, "I",
233     "Maximum size of the IP input queue");
234 
235 static int
236 sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS)
237 {
238 	u_int64_t qdrops_long;
239 	int error, qdrops;
240 
241 	netisr_getqdrops(&ip_nh, &qdrops_long);
242 	qdrops = qdrops_long;
243 	error = sysctl_handle_int(oidp, &qdrops, 0, req);
244 	if (error || !req->newptr)
245 		return (error);
246 	if (qdrops != 0)
247 		return (EINVAL);
248 	netisr_clearqdrops(&ip_nh);
249 	return (0);
250 }
251 
252 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops,
253     CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE,
254     0, 0, sysctl_netinet_intr_queue_drops, "I",
255     "Number of packets dropped from the IP input queue");
256 
257 #ifdef	RSS
258 static int
259 sysctl_netinet_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS)
260 {
261 	int error, qlimit;
262 
263 	netisr_getqlimit(&ip_direct_nh, &qlimit);
264 	error = sysctl_handle_int(oidp, &qlimit, 0, req);
265 	if (error || !req->newptr)
266 		return (error);
267 	if (qlimit < 1)
268 		return (EINVAL);
269 	return (netisr_setqlimit(&ip_direct_nh, qlimit));
270 }
271 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRDQMAXLEN, intr_direct_queue_maxlen,
272     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
273     0, 0, sysctl_netinet_intr_direct_queue_maxlen,
274     "I", "Maximum size of the IP direct input queue");
275 
276 static int
277 sysctl_netinet_intr_direct_queue_drops(SYSCTL_HANDLER_ARGS)
278 {
279 	u_int64_t qdrops_long;
280 	int error, qdrops;
281 
282 	netisr_getqdrops(&ip_direct_nh, &qdrops_long);
283 	qdrops = qdrops_long;
284 	error = sysctl_handle_int(oidp, &qdrops, 0, req);
285 	if (error || !req->newptr)
286 		return (error);
287 	if (qdrops != 0)
288 		return (EINVAL);
289 	netisr_clearqdrops(&ip_direct_nh);
290 	return (0);
291 }
292 
293 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRDQDROPS, intr_direct_queue_drops,
294     CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, 0, 0,
295     sysctl_netinet_intr_direct_queue_drops, "I",
296     "Number of packets dropped from the IP direct input queue");
297 #endif	/* RSS */
298 
299 /*
300  * IP initialization: fill in IP protocol switch table.
301  * All protocols not implemented in kernel go to raw IP protocol handler.
302  */
303 static void
304 ip_vnet_init(void *arg __unused)
305 {
306 	struct pfil_head_args args;
307 
308 	CK_STAILQ_INIT(&V_in_ifaddrhead);
309 	V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask);
310 
311 	/* Initialize IP reassembly queue. */
312 	ipreass_init();
313 
314 	/* Initialize packet filter hooks. */
315 	args.pa_version = PFIL_VERSION;
316 	args.pa_flags = PFIL_IN | PFIL_OUT;
317 	args.pa_type = PFIL_TYPE_IP4;
318 	args.pa_headname = PFIL_INET_NAME;
319 	V_inet_pfil_head = pfil_head_register(&args);
320 
321 	if (hhook_head_register(HHOOK_TYPE_IPSEC_IN, AF_INET,
322 	    &V_ipsec_hhh_in[HHOOK_IPSEC_INET],
323 	    HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
324 		printf("%s: WARNING: unable to register input helper hook\n",
325 		    __func__);
326 	if (hhook_head_register(HHOOK_TYPE_IPSEC_OUT, AF_INET,
327 	    &V_ipsec_hhh_out[HHOOK_IPSEC_INET],
328 	    HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
329 		printf("%s: WARNING: unable to register output helper hook\n",
330 		    __func__);
331 
332 #ifdef VIMAGE
333 	netisr_register_vnet(&ip_nh);
334 #ifdef	RSS
335 	netisr_register_vnet(&ip_direct_nh);
336 #endif
337 #endif
338 }
339 VNET_SYSINIT(ip_vnet_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_FOURTH,
340     ip_vnet_init, NULL);
341 
342 
343 static void
344 ip_init(const void *unused __unused)
345 {
346 	struct protosw *pr;
347 
348 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
349 	KASSERT(pr, ("%s: PF_INET not found", __func__));
350 
351 	/* Initialize the entire ip_protox[] array to IPPROTO_RAW. */
352 	for (int i = 0; i < IPPROTO_MAX; i++)
353 		ip_protox[i] = pr - inetsw;
354 	/*
355 	 * Cycle through IP protocols and put them into the appropriate place
356 	 * in ip_protox[].
357 	 */
358 	for (pr = inetdomain.dom_protosw;
359 	    pr < inetdomain.dom_protoswNPROTOSW; pr++)
360 		if (pr->pr_domain->dom_family == PF_INET &&
361 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
362 			/* Be careful to only index valid IP protocols. */
363 			if (pr->pr_protocol < IPPROTO_MAX)
364 				ip_protox[pr->pr_protocol] = pr - inetsw;
365 		}
366 
367 	netisr_register(&ip_nh);
368 #ifdef	RSS
369 	netisr_register(&ip_direct_nh);
370 #endif
371 }
372 SYSINIT(ip_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip_init, NULL);
373 
374 #ifdef VIMAGE
375 static void
376 ip_destroy(void *unused __unused)
377 {
378 	int error;
379 
380 #ifdef	RSS
381 	netisr_unregister_vnet(&ip_direct_nh);
382 #endif
383 	netisr_unregister_vnet(&ip_nh);
384 
385 	pfil_head_unregister(V_inet_pfil_head);
386 	error = hhook_head_deregister(V_ipsec_hhh_in[HHOOK_IPSEC_INET]);
387 	if (error != 0) {
388 		printf("%s: WARNING: unable to deregister input helper hook "
389 		    "type HHOOK_TYPE_IPSEC_IN, id HHOOK_IPSEC_INET: "
390 		    "error %d returned\n", __func__, error);
391 	}
392 	error = hhook_head_deregister(V_ipsec_hhh_out[HHOOK_IPSEC_INET]);
393 	if (error != 0) {
394 		printf("%s: WARNING: unable to deregister output helper hook "
395 		    "type HHOOK_TYPE_IPSEC_OUT, id HHOOK_IPSEC_INET: "
396 		    "error %d returned\n", __func__, error);
397 	}
398 
399 	/* Remove the IPv4 addresses from all interfaces. */
400 	in_ifscrub_all();
401 
402 	/* Make sure the IPv4 routes are gone as well. */
403 	rib_flush_routes_family(AF_INET);
404 
405 	/* Destroy IP reassembly queue. */
406 	ipreass_destroy();
407 
408 	/* Cleanup in_ifaddr hash table; should be empty. */
409 	hashdestroy(V_in_ifaddrhashtbl, M_IFADDR, V_in_ifaddrhmask);
410 }
411 
412 VNET_SYSUNINIT(ip, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip_destroy, NULL);
413 #endif
414 
415 #ifdef	RSS
416 /*
417  * IP direct input routine.
418  *
419  * This is called when reinjecting completed fragments where
420  * all of the previous checking and book-keeping has been done.
421  */
422 void
423 ip_direct_input(struct mbuf *m)
424 {
425 	struct ip *ip;
426 	int hlen;
427 
428 	ip = mtod(m, struct ip *);
429 	hlen = ip->ip_hl << 2;
430 
431 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
432 	if (IPSEC_ENABLED(ipv4)) {
433 		if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
434 			return;
435 	}
436 #endif /* IPSEC */
437 	IPSTAT_INC(ips_delivered);
438 	(*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
439 	return;
440 }
441 #endif
442 
443 /*
444  * Ip input routine.  Checksum and byte swap header.  If fragmented
445  * try to reassemble.  Process options.  Pass to next level.
446  */
447 void
448 ip_input(struct mbuf *m)
449 {
450 	struct ip *ip = NULL;
451 	struct in_ifaddr *ia = NULL;
452 	struct ifaddr *ifa;
453 	struct ifnet *ifp;
454 	int hlen = 0;
455 	uint16_t sum, ip_len;
456 	int dchg = 0;				/* dest changed after fw */
457 	struct in_addr odst;			/* original dst address */
458 	bool strong_es;
459 
460 	M_ASSERTPKTHDR(m);
461 	NET_EPOCH_ASSERT();
462 
463 	if (m->m_flags & M_FASTFWD_OURS) {
464 		m->m_flags &= ~M_FASTFWD_OURS;
465 		/* Set up some basics that will be used later. */
466 		ip = mtod(m, struct ip *);
467 		hlen = ip->ip_hl << 2;
468 		ip_len = ntohs(ip->ip_len);
469 		goto ours;
470 	}
471 
472 	IPSTAT_INC(ips_total);
473 
474 	if (__predict_false(m->m_pkthdr.len < sizeof(struct ip)))
475 		goto tooshort;
476 
477 	if (m->m_len < sizeof(struct ip)) {
478 		m = m_pullup(m, sizeof(struct ip));
479 		if (__predict_false(m == NULL)) {
480 			IPSTAT_INC(ips_toosmall);
481 			return;
482 		}
483 	}
484 	ip = mtod(m, struct ip *);
485 
486 	if (__predict_false(ip->ip_v != IPVERSION)) {
487 		IPSTAT_INC(ips_badvers);
488 		goto bad;
489 	}
490 
491 	hlen = ip->ip_hl << 2;
492 	if (__predict_false(hlen < sizeof(struct ip))) {	/* minimum header length */
493 		IPSTAT_INC(ips_badhlen);
494 		goto bad;
495 	}
496 	if (hlen > m->m_len) {
497 		m = m_pullup(m, hlen);
498 		if (__predict_false(m == NULL)) {
499 			IPSTAT_INC(ips_badhlen);
500 			return;
501 		}
502 		ip = mtod(m, struct ip *);
503 	}
504 
505 	IP_PROBE(receive, NULL, NULL, ip, m->m_pkthdr.rcvif, ip, NULL);
506 
507 	/* IN_LOOPBACK must not appear on the wire - RFC1122 */
508 	ifp = m->m_pkthdr.rcvif;
509 	if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) ||
510 	    IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) {
511 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
512 			IPSTAT_INC(ips_badaddr);
513 			goto bad;
514 		}
515 	}
516 
517 	if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
518 		sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
519 	} else {
520 		if (hlen == sizeof(struct ip)) {
521 			sum = in_cksum_hdr(ip);
522 		} else {
523 			sum = in_cksum(m, hlen);
524 		}
525 	}
526 	if (__predict_false(sum)) {
527 		IPSTAT_INC(ips_badsum);
528 		goto bad;
529 	}
530 
531 #ifdef ALTQ
532 	if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
533 		/* packet is dropped by traffic conditioner */
534 		return;
535 #endif
536 
537 	ip_len = ntohs(ip->ip_len);
538 	if (__predict_false(ip_len < hlen)) {
539 		IPSTAT_INC(ips_badlen);
540 		goto bad;
541 	}
542 
543 	/*
544 	 * Check that the amount of data in the buffers
545 	 * is as at least much as the IP header would have us expect.
546 	 * Trim mbufs if longer than we expect.
547 	 * Drop packet if shorter than we expect.
548 	 */
549 	if (__predict_false(m->m_pkthdr.len < ip_len)) {
550 tooshort:
551 		IPSTAT_INC(ips_tooshort);
552 		goto bad;
553 	}
554 	if (m->m_pkthdr.len > ip_len) {
555 		if (m->m_len == m->m_pkthdr.len) {
556 			m->m_len = ip_len;
557 			m->m_pkthdr.len = ip_len;
558 		} else
559 			m_adj(m, ip_len - m->m_pkthdr.len);
560 	}
561 
562 	/*
563 	 * Try to forward the packet, but if we fail continue.
564 	 * ip_tryforward() may generate redirects these days.
565 	 * XXX the logic below falling through to normal processing
566 	 * if redirects are required should be revisited as well.
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 		/*
580 		 * ip_dooptions() was run so we can ignore the source route (or
581 		 * any IP options case) case for redirects in ip_tryforward().
582 		 */
583 		if ((m = ip_tryforward(m)) == NULL)
584 			return;
585 		if (m->m_flags & M_FASTFWD_OURS) {
586 			m->m_flags &= ~M_FASTFWD_OURS;
587 			ip = mtod(m, struct ip *);
588 			goto ours;
589 		}
590 	}
591 
592 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
593 	/*
594 	 * Bypass packet filtering for packets previously handled by IPsec.
595 	 */
596 	if (IPSEC_ENABLED(ipv4) &&
597 	    IPSEC_CAPS(ipv4, m, IPSEC_CAP_BYPASS_FILTER) != 0)
598 			goto passin;
599 #endif
600 
601 	/*
602 	 * Run through list of hooks for input packets.
603 	 *
604 	 * NB: Beware of the destination address changing (e.g.
605 	 *     by NAT rewriting).  When this happens, tell
606 	 *     ip_forward to do the right thing.
607 	 */
608 
609 	/* Jump over all PFIL processing if hooks are not active. */
610 	if (!PFIL_HOOKED_IN(V_inet_pfil_head))
611 		goto passin;
612 
613 	odst = ip->ip_dst;
614 	if (pfil_run_hooks(V_inet_pfil_head, &m, ifp, PFIL_IN, NULL) !=
615 	    PFIL_PASS)
616 		return;
617 	if (m == NULL)			/* consumed by filter */
618 		return;
619 
620 	ip = mtod(m, struct ip *);
621 	dchg = (odst.s_addr != ip->ip_dst.s_addr);
622 
623 	if (m->m_flags & M_FASTFWD_OURS) {
624 		m->m_flags &= ~M_FASTFWD_OURS;
625 		goto ours;
626 	}
627 	if (m->m_flags & M_IP_NEXTHOP) {
628 		if (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) {
629 			/*
630 			 * Directly ship the packet on.  This allows
631 			 * forwarding packets originally destined to us
632 			 * to some other directly connected host.
633 			 */
634 			ip_forward(m, 1);
635 			return;
636 		}
637 	}
638 passin:
639 
640 	/*
641 	 * Process options and, if not destined for us,
642 	 * ship it on.  ip_dooptions returns 1 when an
643 	 * error was detected (causing an icmp message
644 	 * to be sent and the original packet to be freed).
645 	 */
646 	if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
647 		return;
648 
649         /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
650          * matter if it is destined to another node, or whether it is
651          * a multicast one, RSVP wants it! and prevents it from being forwarded
652          * anywhere else. Also checks if the rsvp daemon is running before
653 	 * grabbing the packet.
654          */
655 	if (ip->ip_p == IPPROTO_RSVP && V_rsvp_on)
656 		goto ours;
657 
658 	/*
659 	 * Check our list of addresses, to see if the packet is for us.
660 	 * If we don't have any addresses, assume any unicast packet
661 	 * we receive might be for us (and let the upper layers deal
662 	 * with it).
663 	 */
664 	if (CK_STAILQ_EMPTY(&V_in_ifaddrhead) &&
665 	    (m->m_flags & (M_MCAST|M_BCAST)) == 0)
666 		goto ours;
667 
668 	/*
669 	 * Enable a consistency check between the destination address
670 	 * and the arrival interface for a unicast packet (the RFC 1122
671 	 * strong ES model) with a list of additional predicates:
672 	 * - if IP forwarding is disabled
673 	 * - the packet is not locally generated
674 	 * - the packet is not subject to 'ipfw fwd'
675 	 * - Interface is not running CARP. If the packet got here, we already
676 	 *   checked it with carp_iamatch() and carp_forus().
677 	 */
678 	strong_es = V_ip_strong_es && (V_ipforwarding == 0) &&
679 	    ((ifp->if_flags & IFF_LOOPBACK) == 0) &&
680 	    ifp->if_carp == NULL && (dchg == 0);
681 
682 	/*
683 	 * Check for exact addresses in the hash bucket.
684 	 */
685 	CK_LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
686 		if (IA_SIN(ia)->sin_addr.s_addr != ip->ip_dst.s_addr)
687 			continue;
688 
689 		/*
690 		 * net.inet.ip.rfc1122_strong_es: the address matches, verify
691 		 * that the packet arrived via the correct interface.
692 		 */
693 		if (__predict_false(strong_es && ia->ia_ifp != ifp)) {
694 			IPSTAT_INC(ips_badaddr);
695 			goto bad;
696 		}
697 
698 		/*
699 		 * net.inet.ip.source_address_validation: drop incoming
700 		 * packets that pretend to be ours.
701 		 */
702 		if (V_ip_sav && !(ifp->if_flags & IFF_LOOPBACK) &&
703 		    __predict_false(in_localip_fib(ip->ip_src, ifp->if_fib))) {
704 			IPSTAT_INC(ips_badaddr);
705 			goto bad;
706 		}
707 
708 		counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
709 		counter_u64_add(ia->ia_ifa.ifa_ibytes, m->m_pkthdr.len);
710 		goto ours;
711 	}
712 
713 	/*
714 	 * Check for broadcast addresses.
715 	 *
716 	 * Only accept broadcast packets that arrive via the matching
717 	 * interface.  Reception of forwarded directed broadcasts would
718 	 * be handled via ip_forward() and ether_output() with the loopback
719 	 * into the stack for SIMPLEX interfaces handled by ether_output().
720 	 */
721 	if (ifp->if_flags & IFF_BROADCAST) {
722 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
723 			if (ifa->ifa_addr->sa_family != AF_INET)
724 				continue;
725 			ia = ifatoia(ifa);
726 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
727 			    ip->ip_dst.s_addr) {
728 				counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
729 				counter_u64_add(ia->ia_ifa.ifa_ibytes,
730 				    m->m_pkthdr.len);
731 				goto ours;
732 			}
733 #ifdef BOOTP_COMPAT
734 			if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) {
735 				counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
736 				counter_u64_add(ia->ia_ifa.ifa_ibytes,
737 				    m->m_pkthdr.len);
738 				goto ours;
739 			}
740 #endif
741 		}
742 		ia = NULL;
743 	}
744 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
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 				IPSTAT_INC(ips_cantforward);
760 				m_freem(m);
761 				return;
762 			}
763 
764 			/*
765 			 * The process-level routing daemon needs to receive
766 			 * all multicast IGMP packets, whether or not this
767 			 * host belongs to their destination groups.
768 			 */
769 			if (ip->ip_p == IPPROTO_IGMP) {
770 				goto ours;
771 			}
772 			IPSTAT_INC(ips_forward);
773 		}
774 		/*
775 		 * Assume the packet is for us, to avoid prematurely taking
776 		 * a lock on the in_multi hash. Protocols must perform
777 		 * their own filtering and update statistics accordingly.
778 		 */
779 		goto ours;
780 	}
781 	if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
782 		goto ours;
783 	if (ip->ip_dst.s_addr == INADDR_ANY)
784 		goto ours;
785 	/* RFC 3927 2.7: Do not forward packets to or from IN_LINKLOCAL. */
786 	if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr)) ||
787 	    IN_LINKLOCAL(ntohl(ip->ip_src.s_addr))) {
788 		IPSTAT_INC(ips_cantforward);
789 		m_freem(m);
790 		return;
791 	}
792 
793 	/*
794 	 * Not for us; forward if possible and desirable.
795 	 */
796 	if (V_ipforwarding == 0) {
797 		IPSTAT_INC(ips_cantforward);
798 		m_freem(m);
799 	} else {
800 		ip_forward(m, dchg);
801 	}
802 	return;
803 
804 ours:
805 #ifdef IPSTEALTH
806 	/*
807 	 * IPSTEALTH: Process non-routing options only
808 	 * if the packet is destined for us.
809 	 */
810 	if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1))
811 		return;
812 #endif /* IPSTEALTH */
813 
814 	/*
815 	 * Attempt reassembly; if it succeeds, proceed.
816 	 * ip_reass() will return a different mbuf.
817 	 */
818 	if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) {
819 		/* XXXGL: shouldn't we save & set m_flags? */
820 		m = ip_reass(m);
821 		if (m == NULL)
822 			return;
823 		ip = mtod(m, struct ip *);
824 		/* Get the header length of the reassembled packet */
825 		hlen = ip->ip_hl << 2;
826 	}
827 
828 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
829 	if (IPSEC_ENABLED(ipv4)) {
830 		if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
831 			return;
832 	}
833 #endif /* IPSEC */
834 
835 	/*
836 	 * Switch out to protocol's input routine.
837 	 */
838 	IPSTAT_INC(ips_delivered);
839 
840 	(*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
841 	return;
842 bad:
843 	m_freem(m);
844 }
845 
846 /*
847  * IP timer processing;
848  * if a timer expires on a reassembly
849  * queue, discard it.
850  */
851 void
852 ip_slowtimo(void)
853 {
854 	VNET_ITERATOR_DECL(vnet_iter);
855 
856 	VNET_LIST_RLOCK_NOSLEEP();
857 	VNET_FOREACH(vnet_iter) {
858 		CURVNET_SET(vnet_iter);
859 		ipreass_slowtimo();
860 		CURVNET_RESTORE();
861 	}
862 	VNET_LIST_RUNLOCK_NOSLEEP();
863 }
864 
865 void
866 ip_drain(void)
867 {
868 	VNET_ITERATOR_DECL(vnet_iter);
869 
870 	VNET_LIST_RLOCK_NOSLEEP();
871 	VNET_FOREACH(vnet_iter) {
872 		CURVNET_SET(vnet_iter);
873 		ipreass_drain();
874 		CURVNET_RESTORE();
875 	}
876 	VNET_LIST_RUNLOCK_NOSLEEP();
877 }
878 
879 /*
880  * The protocol to be inserted into ip_protox[] must be already registered
881  * in inetsw[], either statically or through pf_proto_register().
882  */
883 int
884 ipproto_register(short ipproto)
885 {
886 	struct protosw *pr;
887 
888 	/* Sanity checks. */
889 	if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
890 		return (EPROTONOSUPPORT);
891 
892 	/*
893 	 * The protocol slot must not be occupied by another protocol
894 	 * already.  An index pointing to IPPROTO_RAW is unused.
895 	 */
896 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
897 	if (pr == NULL)
898 		return (EPFNOSUPPORT);
899 	if (ip_protox[ipproto] != pr - inetsw)	/* IPPROTO_RAW */
900 		return (EEXIST);
901 
902 	/* Find the protocol position in inetsw[] and set the index. */
903 	for (pr = inetdomain.dom_protosw;
904 	     pr < inetdomain.dom_protoswNPROTOSW; pr++) {
905 		if (pr->pr_domain->dom_family == PF_INET &&
906 		    pr->pr_protocol && pr->pr_protocol == ipproto) {
907 			ip_protox[pr->pr_protocol] = pr - inetsw;
908 			return (0);
909 		}
910 	}
911 	return (EPROTONOSUPPORT);
912 }
913 
914 int
915 ipproto_unregister(short ipproto)
916 {
917 	struct protosw *pr;
918 
919 	/* Sanity checks. */
920 	if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
921 		return (EPROTONOSUPPORT);
922 
923 	/* Check if the protocol was indeed registered. */
924 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
925 	if (pr == NULL)
926 		return (EPFNOSUPPORT);
927 	if (ip_protox[ipproto] == pr - inetsw)  /* IPPROTO_RAW */
928 		return (ENOENT);
929 
930 	/* Reset the protocol slot to IPPROTO_RAW. */
931 	ip_protox[ipproto] = pr - inetsw;
932 	return (0);
933 }
934 
935 u_char inetctlerrmap[PRC_NCMDS] = {
936 	0,		0,		0,		0,
937 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
938 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
939 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
940 	0,		0,		EHOSTUNREACH,	0,
941 	ENOPROTOOPT,	ECONNREFUSED
942 };
943 
944 /*
945  * Forward a packet.  If some error occurs return the sender
946  * an icmp packet.  Note we can't always generate a meaningful
947  * icmp message because icmp doesn't have a large enough repertoire
948  * of codes and types.
949  *
950  * If not forwarding, just drop the packet.  This could be confusing
951  * if ipforwarding was zero but some routing protocol was advancing
952  * us as a gateway to somewhere.  However, we must let the routing
953  * protocol deal with that.
954  *
955  * The srcrt parameter indicates whether the packet is being forwarded
956  * via a source route.
957  */
958 void
959 ip_forward(struct mbuf *m, int srcrt)
960 {
961 	struct ip *ip = mtod(m, struct ip *);
962 	struct in_ifaddr *ia;
963 	struct mbuf *mcopy;
964 	struct sockaddr_in *sin;
965 	struct in_addr dest;
966 	struct route ro;
967 	uint32_t flowid;
968 	int error, type = 0, code = 0, mtu = 0;
969 
970 	NET_EPOCH_ASSERT();
971 
972 	if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
973 		IPSTAT_INC(ips_cantforward);
974 		m_freem(m);
975 		return;
976 	}
977 	if (
978 #ifdef IPSTEALTH
979 	    V_ipstealth == 0 &&
980 #endif
981 	    ip->ip_ttl <= IPTTLDEC) {
982 		icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 0, 0);
983 		return;
984 	}
985 
986 	bzero(&ro, sizeof(ro));
987 	sin = (struct sockaddr_in *)&ro.ro_dst;
988 	sin->sin_family = AF_INET;
989 	sin->sin_len = sizeof(*sin);
990 	sin->sin_addr = ip->ip_dst;
991 	flowid = m->m_pkthdr.flowid;
992 	ro.ro_nh = fib4_lookup(M_GETFIB(m), ip->ip_dst, 0, NHR_REF, flowid);
993 	if (ro.ro_nh != NULL) {
994 		ia = ifatoia(ro.ro_nh->nh_ifa);
995 	} else
996 		ia = NULL;
997 	/*
998 	 * Save the IP header and at most 8 bytes of the payload,
999 	 * in case we need to generate an ICMP message to the src.
1000 	 *
1001 	 * XXX this can be optimized a lot by saving the data in a local
1002 	 * buffer on the stack (72 bytes at most), and only allocating the
1003 	 * mbuf if really necessary. The vast majority of the packets
1004 	 * are forwarded without having to send an ICMP back (either
1005 	 * because unnecessary, or because rate limited), so we are
1006 	 * really we are wasting a lot of work here.
1007 	 *
1008 	 * We don't use m_copym() because it might return a reference
1009 	 * to a shared cluster. Both this function and ip_output()
1010 	 * assume exclusive access to the IP header in `m', so any
1011 	 * data in a cluster may change before we reach icmp_error().
1012 	 */
1013 	mcopy = m_gethdr(M_NOWAIT, m->m_type);
1014 	if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_NOWAIT)) {
1015 		/*
1016 		 * It's probably ok if the pkthdr dup fails (because
1017 		 * the deep copy of the tag chain failed), but for now
1018 		 * be conservative and just discard the copy since
1019 		 * code below may some day want the tags.
1020 		 */
1021 		m_free(mcopy);
1022 		mcopy = NULL;
1023 	}
1024 	if (mcopy != NULL) {
1025 		mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy));
1026 		mcopy->m_pkthdr.len = mcopy->m_len;
1027 		m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1028 	}
1029 #ifdef IPSTEALTH
1030 	if (V_ipstealth == 0)
1031 #endif
1032 		ip->ip_ttl -= IPTTLDEC;
1033 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1034 	if (IPSEC_ENABLED(ipv4)) {
1035 		if ((error = IPSEC_FORWARD(ipv4, m)) != 0) {
1036 			/* mbuf consumed by IPsec */
1037 			RO_NHFREE(&ro);
1038 			m_freem(mcopy);
1039 			if (error != EINPROGRESS)
1040 				IPSTAT_INC(ips_cantforward);
1041 			return;
1042 		}
1043 		/* No IPsec processing required */
1044 	}
1045 #endif /* IPSEC */
1046 	/*
1047 	 * If forwarding packet using same interface that it came in on,
1048 	 * perhaps should send a redirect to sender to shortcut a hop.
1049 	 * Only send redirect if source is sending directly to us,
1050 	 * and if packet was not source routed (or has any options).
1051 	 * Also, don't send redirect if forwarding using a default route
1052 	 * or a route modified by a redirect.
1053 	 */
1054 	dest.s_addr = 0;
1055 	if (!srcrt && V_ipsendredirects &&
1056 	    ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) {
1057 		struct nhop_object *nh;
1058 
1059 		nh = ro.ro_nh;
1060 
1061 		if (nh != NULL && ((nh->nh_flags & (NHF_REDIRECT|NHF_DEFAULT)) == 0)) {
1062 			struct in_ifaddr *nh_ia = (struct in_ifaddr *)(nh->nh_ifa);
1063 			u_long src = ntohl(ip->ip_src.s_addr);
1064 
1065 			if (nh_ia != NULL &&
1066 			    (src & nh_ia->ia_subnetmask) == nh_ia->ia_subnet) {
1067 				/* Router requirements says to only send host redirects */
1068 				type = ICMP_REDIRECT;
1069 				code = ICMP_REDIRECT_HOST;
1070 				if (nh->nh_flags & NHF_GATEWAY) {
1071 				    if (nh->gw_sa.sa_family == AF_INET)
1072 					dest.s_addr = nh->gw4_sa.sin_addr.s_addr;
1073 				    else /* Do not redirect in case gw is AF_INET6 */
1074 					type = 0;
1075 				} else
1076 					dest.s_addr = ip->ip_dst.s_addr;
1077 			}
1078 		}
1079 	}
1080 
1081 	error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
1082 
1083 	if (error == EMSGSIZE && ro.ro_nh)
1084 		mtu = ro.ro_nh->nh_mtu;
1085 	RO_NHFREE(&ro);
1086 
1087 	if (error)
1088 		IPSTAT_INC(ips_cantforward);
1089 	else {
1090 		IPSTAT_INC(ips_forward);
1091 		if (type)
1092 			IPSTAT_INC(ips_redirectsent);
1093 		else {
1094 			if (mcopy)
1095 				m_freem(mcopy);
1096 			return;
1097 		}
1098 	}
1099 	if (mcopy == NULL)
1100 		return;
1101 
1102 	switch (error) {
1103 	case 0:				/* forwarded, but need redirect */
1104 		/* type, code set above */
1105 		break;
1106 
1107 	case ENETUNREACH:
1108 	case EHOSTUNREACH:
1109 	case ENETDOWN:
1110 	case EHOSTDOWN:
1111 	default:
1112 		type = ICMP_UNREACH;
1113 		code = ICMP_UNREACH_HOST;
1114 		break;
1115 
1116 	case EMSGSIZE:
1117 		type = ICMP_UNREACH;
1118 		code = ICMP_UNREACH_NEEDFRAG;
1119 		/*
1120 		 * If the MTU was set before make sure we are below the
1121 		 * interface MTU.
1122 		 * If the MTU wasn't set before use the interface mtu or
1123 		 * fall back to the next smaller mtu step compared to the
1124 		 * current packet size.
1125 		 */
1126 		if (mtu != 0) {
1127 			if (ia != NULL)
1128 				mtu = min(mtu, ia->ia_ifp->if_mtu);
1129 		} else {
1130 			if (ia != NULL)
1131 				mtu = ia->ia_ifp->if_mtu;
1132 			else
1133 				mtu = ip_next_mtu(ntohs(ip->ip_len), 0);
1134 		}
1135 		IPSTAT_INC(ips_cantfrag);
1136 		break;
1137 
1138 	case ENOBUFS:
1139 	case EACCES:			/* ipfw denied packet */
1140 		m_freem(mcopy);
1141 		return;
1142 	}
1143 	icmp_error(mcopy, type, code, dest.s_addr, mtu);
1144 }
1145 
1146 #define	CHECK_SO_CT(sp, ct) \
1147     (((sp->so_options & SO_TIMESTAMP) && (sp->so_ts_clock == ct)) ? 1 : 0)
1148 
1149 void
1150 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1151     struct mbuf *m)
1152 {
1153 	bool stamped;
1154 
1155 	stamped = false;
1156 	if ((inp->inp_socket->so_options & SO_BINTIME) ||
1157 	    CHECK_SO_CT(inp->inp_socket, SO_TS_BINTIME)) {
1158 		struct bintime boottimebin, bt;
1159 		struct timespec ts1;
1160 
1161 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1162 		    M_TSTMP)) {
1163 			mbuf_tstmp2timespec(m, &ts1);
1164 			timespec2bintime(&ts1, &bt);
1165 			getboottimebin(&boottimebin);
1166 			bintime_add(&bt, &boottimebin);
1167 		} else {
1168 			bintime(&bt);
1169 		}
1170 		*mp = sbcreatecontrol(&bt, sizeof(bt), SCM_BINTIME,
1171 		    SOL_SOCKET, M_NOWAIT);
1172 		if (*mp != NULL) {
1173 			mp = &(*mp)->m_next;
1174 			stamped = true;
1175 		}
1176 	}
1177 	if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME_MICRO)) {
1178 		struct bintime boottimebin, bt1;
1179 		struct timespec ts1;
1180 		struct timeval tv;
1181 
1182 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1183 		    M_TSTMP)) {
1184 			mbuf_tstmp2timespec(m, &ts1);
1185 			timespec2bintime(&ts1, &bt1);
1186 			getboottimebin(&boottimebin);
1187 			bintime_add(&bt1, &boottimebin);
1188 			bintime2timeval(&bt1, &tv);
1189 		} else {
1190 			microtime(&tv);
1191 		}
1192 		*mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv), SCM_TIMESTAMP,
1193 		    SOL_SOCKET, M_NOWAIT);
1194 		if (*mp != NULL) {
1195 			mp = &(*mp)->m_next;
1196 			stamped = true;
1197 		}
1198 	} else if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME)) {
1199 		struct bintime boottimebin;
1200 		struct timespec ts, ts1;
1201 
1202 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1203 		    M_TSTMP)) {
1204 			mbuf_tstmp2timespec(m, &ts);
1205 			getboottimebin(&boottimebin);
1206 			bintime2timespec(&boottimebin, &ts1);
1207 			timespecadd(&ts, &ts1, &ts);
1208 		} else {
1209 			nanotime(&ts);
1210 		}
1211 		*mp = sbcreatecontrol(&ts, sizeof(ts), SCM_REALTIME,
1212 		    SOL_SOCKET, M_NOWAIT);
1213 		if (*mp != NULL) {
1214 			mp = &(*mp)->m_next;
1215 			stamped = true;
1216 		}
1217 	} else if (CHECK_SO_CT(inp->inp_socket, SO_TS_MONOTONIC)) {
1218 		struct timespec ts;
1219 
1220 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1221 		    M_TSTMP))
1222 			mbuf_tstmp2timespec(m, &ts);
1223 		else
1224 			nanouptime(&ts);
1225 		*mp = sbcreatecontrol(&ts, sizeof(ts), SCM_MONOTONIC,
1226 		    SOL_SOCKET, M_NOWAIT);
1227 		if (*mp != NULL) {
1228 			mp = &(*mp)->m_next;
1229 			stamped = true;
1230 		}
1231 	}
1232 	if (stamped && (m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1233 	    M_TSTMP)) {
1234 		struct sock_timestamp_info sti;
1235 
1236 		bzero(&sti, sizeof(sti));
1237 		sti.st_info_flags = ST_INFO_HW;
1238 		if ((m->m_flags & M_TSTMP_HPREC) != 0)
1239 			sti.st_info_flags |= ST_INFO_HW_HPREC;
1240 		*mp = sbcreatecontrol(&sti, sizeof(sti), SCM_TIME_INFO,
1241 		    SOL_SOCKET, M_NOWAIT);
1242 		if (*mp != NULL)
1243 			mp = &(*mp)->m_next;
1244 	}
1245 	if (inp->inp_flags & INP_RECVDSTADDR) {
1246 		*mp = sbcreatecontrol(&ip->ip_dst, sizeof(struct in_addr),
1247 		    IP_RECVDSTADDR, IPPROTO_IP, M_NOWAIT);
1248 		if (*mp)
1249 			mp = &(*mp)->m_next;
1250 	}
1251 	if (inp->inp_flags & INP_RECVTTL) {
1252 		*mp = sbcreatecontrol(&ip->ip_ttl, sizeof(u_char), IP_RECVTTL,
1253 		    IPPROTO_IP, M_NOWAIT);
1254 		if (*mp)
1255 			mp = &(*mp)->m_next;
1256 	}
1257 #ifdef notyet
1258 	/* XXX
1259 	 * Moving these out of udp_input() made them even more broken
1260 	 * than they already were.
1261 	 */
1262 	/* options were tossed already */
1263 	if (inp->inp_flags & INP_RECVOPTS) {
1264 		*mp = sbcreatecontrol(opts_deleted_above,
1265 		    sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP, M_NOWAIT);
1266 		if (*mp)
1267 			mp = &(*mp)->m_next;
1268 	}
1269 	/* ip_srcroute doesn't do what we want here, need to fix */
1270 	if (inp->inp_flags & INP_RECVRETOPTS) {
1271 		*mp = sbcreatecontrol(ip_srcroute(m), sizeof(struct in_addr),
1272 		    IP_RECVRETOPTS, IPPROTO_IP, M_NOWAIT);
1273 		if (*mp)
1274 			mp = &(*mp)->m_next;
1275 	}
1276 #endif
1277 	if (inp->inp_flags & INP_RECVIF) {
1278 		struct ifnet *ifp;
1279 		struct sdlbuf {
1280 			struct sockaddr_dl sdl;
1281 			u_char	pad[32];
1282 		} sdlbuf;
1283 		struct sockaddr_dl *sdp;
1284 		struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1285 
1286 		if ((ifp = m->m_pkthdr.rcvif)) {
1287 			sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1288 			/*
1289 			 * Change our mind and don't try copy.
1290 			 */
1291 			if (sdp->sdl_family != AF_LINK ||
1292 			    sdp->sdl_len > sizeof(sdlbuf)) {
1293 				goto makedummy;
1294 			}
1295 			bcopy(sdp, sdl2, sdp->sdl_len);
1296 		} else {
1297 makedummy:
1298 			sdl2->sdl_len =
1299 			    offsetof(struct sockaddr_dl, sdl_data[0]);
1300 			sdl2->sdl_family = AF_LINK;
1301 			sdl2->sdl_index = 0;
1302 			sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1303 		}
1304 		*mp = sbcreatecontrol(sdl2, sdl2->sdl_len, IP_RECVIF,
1305 		    IPPROTO_IP, M_NOWAIT);
1306 		if (*mp)
1307 			mp = &(*mp)->m_next;
1308 	}
1309 	if (inp->inp_flags & INP_RECVTOS) {
1310 		*mp = sbcreatecontrol(&ip->ip_tos, sizeof(u_char), IP_RECVTOS,
1311 		    IPPROTO_IP, M_NOWAIT);
1312 		if (*mp)
1313 			mp = &(*mp)->m_next;
1314 	}
1315 
1316 	if (inp->inp_flags2 & INP_RECVFLOWID) {
1317 		uint32_t flowid, flow_type;
1318 
1319 		flowid = m->m_pkthdr.flowid;
1320 		flow_type = M_HASHTYPE_GET(m);
1321 
1322 		/*
1323 		 * XXX should handle the failure of one or the
1324 		 * other - don't populate both?
1325 		 */
1326 		*mp = sbcreatecontrol(&flowid, sizeof(uint32_t), IP_FLOWID,
1327 		    IPPROTO_IP, M_NOWAIT);
1328 		if (*mp)
1329 			mp = &(*mp)->m_next;
1330 		*mp = sbcreatecontrol(&flow_type, sizeof(uint32_t),
1331 		    IP_FLOWTYPE, IPPROTO_IP, M_NOWAIT);
1332 		if (*mp)
1333 			mp = &(*mp)->m_next;
1334 	}
1335 
1336 #ifdef	RSS
1337 	if (inp->inp_flags2 & INP_RECVRSSBUCKETID) {
1338 		uint32_t flowid, flow_type;
1339 		uint32_t rss_bucketid;
1340 
1341 		flowid = m->m_pkthdr.flowid;
1342 		flow_type = M_HASHTYPE_GET(m);
1343 
1344 		if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) {
1345 			*mp = sbcreatecontrol(&rss_bucketid, sizeof(uint32_t),
1346 			    IP_RSSBUCKETID, IPPROTO_IP, M_NOWAIT);
1347 			if (*mp)
1348 				mp = &(*mp)->m_next;
1349 		}
1350 	}
1351 #endif
1352 }
1353 
1354 /*
1355  * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
1356  * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
1357  * locking.  This code remains in ip_input.c as ip_mroute.c is optionally
1358  * compiled.
1359  */
1360 VNET_DEFINE_STATIC(int, ip_rsvp_on);
1361 VNET_DEFINE(struct socket *, ip_rsvpd);
1362 
1363 #define	V_ip_rsvp_on		VNET(ip_rsvp_on)
1364 
1365 int
1366 ip_rsvp_init(struct socket *so)
1367 {
1368 
1369 	if (so->so_type != SOCK_RAW ||
1370 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
1371 		return EOPNOTSUPP;
1372 
1373 	if (V_ip_rsvpd != NULL)
1374 		return EADDRINUSE;
1375 
1376 	V_ip_rsvpd = so;
1377 	/*
1378 	 * This may seem silly, but we need to be sure we don't over-increment
1379 	 * the RSVP counter, in case something slips up.
1380 	 */
1381 	if (!V_ip_rsvp_on) {
1382 		V_ip_rsvp_on = 1;
1383 		V_rsvp_on++;
1384 	}
1385 
1386 	return 0;
1387 }
1388 
1389 int
1390 ip_rsvp_done(void)
1391 {
1392 
1393 	V_ip_rsvpd = NULL;
1394 	/*
1395 	 * This may seem silly, but we need to be sure we don't over-decrement
1396 	 * the RSVP counter, in case something slips up.
1397 	 */
1398 	if (V_ip_rsvp_on) {
1399 		V_ip_rsvp_on = 0;
1400 		V_rsvp_on--;
1401 	}
1402 	return 0;
1403 }
1404 
1405 int
1406 rsvp_input(struct mbuf **mp, int *offp, int proto)
1407 {
1408 	struct mbuf *m;
1409 
1410 	m = *mp;
1411 	*mp = NULL;
1412 
1413 	if (rsvp_input_p) { /* call the real one if loaded */
1414 		*mp = m;
1415 		rsvp_input_p(mp, offp, proto);
1416 		return (IPPROTO_DONE);
1417 	}
1418 
1419 	/* Can still get packets with rsvp_on = 0 if there is a local member
1420 	 * of the group to which the RSVP packet is addressed.  But in this
1421 	 * case we want to throw the packet away.
1422 	 */
1423 
1424 	if (!V_rsvp_on) {
1425 		m_freem(m);
1426 		return (IPPROTO_DONE);
1427 	}
1428 
1429 	if (V_ip_rsvpd != NULL) {
1430 		*mp = m;
1431 		rip_input(mp, offp, proto);
1432 		return (IPPROTO_DONE);
1433 	}
1434 	/* Drop the packet */
1435 	m_freem(m);
1436 	return (IPPROTO_DONE);
1437 }
1438