xref: /freebsd/sys/netinet/ip_input.c (revision f5147e312f43a9050468de539aeafa072caa1a60)
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/pfil.h>
61 #include <net/if.h>
62 #include <net/if_types.h>
63 #include <net/if_var.h>
64 #include <net/if_dl.h>
65 #include <net/route.h>
66 #include <net/netisr.h>
67 #include <net/rss_config.h>
68 #include <net/vnet.h>
69 
70 #include <netinet/in.h>
71 #include <netinet/in_kdtrace.h>
72 #include <netinet/in_systm.h>
73 #include <netinet/in_var.h>
74 #include <netinet/ip.h>
75 #include <netinet/in_pcb.h>
76 #include <netinet/ip_var.h>
77 #include <netinet/ip_fw.h>
78 #include <netinet/ip_icmp.h>
79 #include <netinet/ip_options.h>
80 #include <machine/in_cksum.h>
81 #include <netinet/ip_carp.h>
82 #include <netinet/in_rss.h>
83 
84 #include <netipsec/ipsec_support.h>
85 
86 #include <sys/socketvar.h>
87 
88 #include <security/mac/mac_framework.h>
89 
90 #ifdef CTASSERT
91 CTASSERT(sizeof(struct ip) == 20);
92 #endif
93 
94 /* IP reassembly functions are defined in ip_reass.c. */
95 extern void ipreass_init(void);
96 extern void ipreass_drain(void);
97 extern void ipreass_slowtimo(void);
98 #ifdef VIMAGE
99 extern void ipreass_destroy(void);
100 #endif
101 
102 struct rmlock in_ifaddr_lock;
103 RM_SYSINIT(in_ifaddr_lock, &in_ifaddr_lock, "in_ifaddr_lock");
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 static VNET_DEFINE(int, ipsendredirects) = 1;	/* XXX */
113 #define	V_ipsendredirects	VNET(ipsendredirects)
114 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_VNET | CTLFLAG_RW,
115     &VNET_NAME(ipsendredirects), 0,
116     "Enable sending IP redirects");
117 
118 /*
119  * XXX - Setting ip_checkinterface mostly implements the receive side of
120  * the Strong ES model described in RFC 1122, but since the routing table
121  * and transmit implementation do not implement the Strong ES model,
122  * setting this to 1 results in an odd hybrid.
123  *
124  * XXX - ip_checkinterface currently must be disabled if you use ipnat
125  * to translate the destination address to another local interface.
126  *
127  * XXX - ip_checkinterface must be disabled if you add IP aliases
128  * to the loopback interface instead of the interface where the
129  * packets for those addresses are received.
130  */
131 static VNET_DEFINE(int, ip_checkinterface);
132 #define	V_ip_checkinterface	VNET(ip_checkinterface)
133 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_VNET | CTLFLAG_RW,
134     &VNET_NAME(ip_checkinterface), 0,
135     "Verify packet arrives on correct interface");
136 
137 VNET_DEFINE(struct pfil_head, inet_pfil_hook);	/* Packet filter hooks */
138 
139 static struct netisr_handler ip_nh = {
140 	.nh_name = "ip",
141 	.nh_handler = ip_input,
142 	.nh_proto = NETISR_IP,
143 #ifdef	RSS
144 	.nh_m2cpuid = rss_soft_m2cpuid_v4,
145 	.nh_policy = NETISR_POLICY_CPU,
146 	.nh_dispatch = NETISR_DISPATCH_HYBRID,
147 #else
148 	.nh_policy = NETISR_POLICY_FLOW,
149 #endif
150 };
151 
152 #ifdef	RSS
153 /*
154  * Directly dispatched frames are currently assumed
155  * to have a flowid already calculated.
156  *
157  * It should likely have something that assert it
158  * actually has valid flow details.
159  */
160 static struct netisr_handler ip_direct_nh = {
161 	.nh_name = "ip_direct",
162 	.nh_handler = ip_direct_input,
163 	.nh_proto = NETISR_IP_DIRECT,
164 	.nh_m2cpuid = rss_soft_m2cpuid_v4,
165 	.nh_policy = NETISR_POLICY_CPU,
166 	.nh_dispatch = NETISR_DISPATCH_HYBRID,
167 };
168 #endif
169 
170 extern	struct domain inetdomain;
171 extern	struct protosw inetsw[];
172 u_char	ip_protox[IPPROTO_MAX];
173 VNET_DEFINE(struct in_ifaddrhead, in_ifaddrhead);  /* first inet address */
174 VNET_DEFINE(struct in_ifaddrhashhead *, in_ifaddrhashtbl); /* inet addr hash table  */
175 VNET_DEFINE(u_long, in_ifaddrhmask);		/* mask for hash table */
176 
177 #ifdef IPCTL_DEFMTU
178 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
179     &ip_mtu, 0, "Default MTU");
180 #endif
181 
182 #ifdef IPSTEALTH
183 VNET_DEFINE(int, ipstealth);
184 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_VNET | CTLFLAG_RW,
185     &VNET_NAME(ipstealth), 0,
186     "IP stealth mode, no TTL decrementation on forwarding");
187 #endif
188 
189 /*
190  * IP statistics are stored in the "array" of counter(9)s.
191  */
192 VNET_PCPUSTAT_DEFINE(struct ipstat, ipstat);
193 VNET_PCPUSTAT_SYSINIT(ipstat);
194 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, IPCTL_STATS, stats, struct ipstat, ipstat,
195     "IP statistics (struct ipstat, netinet/ip_var.h)");
196 
197 #ifdef VIMAGE
198 VNET_PCPUSTAT_SYSUNINIT(ipstat);
199 #endif /* VIMAGE */
200 
201 /*
202  * Kernel module interface for updating ipstat.  The argument is an index
203  * into ipstat treated as an array.
204  */
205 void
206 kmod_ipstat_inc(int statnum)
207 {
208 
209 	counter_u64_add(VNET(ipstat)[statnum], 1);
210 }
211 
212 void
213 kmod_ipstat_dec(int statnum)
214 {
215 
216 	counter_u64_add(VNET(ipstat)[statnum], -1);
217 }
218 
219 static int
220 sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS)
221 {
222 	int error, qlimit;
223 
224 	netisr_getqlimit(&ip_nh, &qlimit);
225 	error = sysctl_handle_int(oidp, &qlimit, 0, req);
226 	if (error || !req->newptr)
227 		return (error);
228 	if (qlimit < 1)
229 		return (EINVAL);
230 	return (netisr_setqlimit(&ip_nh, qlimit));
231 }
232 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen,
233     CTLTYPE_INT|CTLFLAG_RW, 0, 0, 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, 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, 0, 0, sysctl_netinet_intr_direct_queue_maxlen,
273     "I", "Maximum size of the IP direct input queue");
274 
275 static int
276 sysctl_netinet_intr_direct_queue_drops(SYSCTL_HANDLER_ARGS)
277 {
278 	u_int64_t qdrops_long;
279 	int error, qdrops;
280 
281 	netisr_getqdrops(&ip_direct_nh, &qdrops_long);
282 	qdrops = qdrops_long;
283 	error = sysctl_handle_int(oidp, &qdrops, 0, req);
284 	if (error || !req->newptr)
285 		return (error);
286 	if (qdrops != 0)
287 		return (EINVAL);
288 	netisr_clearqdrops(&ip_direct_nh);
289 	return (0);
290 }
291 
292 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRDQDROPS, intr_direct_queue_drops,
293     CTLTYPE_INT|CTLFLAG_RD, 0, 0, sysctl_netinet_intr_direct_queue_drops, "I",
294     "Number of packets dropped from the IP direct input queue");
295 #endif	/* RSS */
296 
297 /*
298  * IP initialization: fill in IP protocol switch table.
299  * All protocols not implemented in kernel go to raw IP protocol handler.
300  */
301 void
302 ip_init(void)
303 {
304 	struct protosw *pr;
305 	int i;
306 
307 	TAILQ_INIT(&V_in_ifaddrhead);
308 	V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask);
309 
310 	/* Initialize IP reassembly queue. */
311 	ipreass_init();
312 
313 	/* Initialize packet filter hooks. */
314 	V_inet_pfil_hook.ph_type = PFIL_TYPE_AF;
315 	V_inet_pfil_hook.ph_af = AF_INET;
316 	if ((i = pfil_head_register(&V_inet_pfil_hook)) != 0)
317 		printf("%s: WARNING: unable to register pfil hook, "
318 			"error %d\n", __func__, i);
319 
320 	if (hhook_head_register(HHOOK_TYPE_IPSEC_IN, AF_INET,
321 	    &V_ipsec_hhh_in[HHOOK_IPSEC_INET],
322 	    HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
323 		printf("%s: WARNING: unable to register input helper hook\n",
324 		    __func__);
325 	if (hhook_head_register(HHOOK_TYPE_IPSEC_OUT, AF_INET,
326 	    &V_ipsec_hhh_out[HHOOK_IPSEC_INET],
327 	    HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
328 		printf("%s: WARNING: unable to register output helper hook\n",
329 		    __func__);
330 
331 	/* Skip initialization of globals for non-default instances. */
332 #ifdef VIMAGE
333 	if (!IS_DEFAULT_VNET(curvnet)) {
334 		netisr_register_vnet(&ip_nh);
335 #ifdef	RSS
336 		netisr_register_vnet(&ip_direct_nh);
337 #endif
338 		return;
339 	}
340 #endif
341 
342 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
343 	if (pr == NULL)
344 		panic("ip_init: PF_INET not found");
345 
346 	/* Initialize the entire ip_protox[] array to IPPROTO_RAW. */
347 	for (i = 0; i < IPPROTO_MAX; i++)
348 		ip_protox[i] = pr - inetsw;
349 	/*
350 	 * Cycle through IP protocols and put them into the appropriate place
351 	 * in ip_protox[].
352 	 */
353 	for (pr = inetdomain.dom_protosw;
354 	    pr < inetdomain.dom_protoswNPROTOSW; pr++)
355 		if (pr->pr_domain->dom_family == PF_INET &&
356 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
357 			/* Be careful to only index valid IP protocols. */
358 			if (pr->pr_protocol < IPPROTO_MAX)
359 				ip_protox[pr->pr_protocol] = pr - inetsw;
360 		}
361 
362 	netisr_register(&ip_nh);
363 #ifdef	RSS
364 	netisr_register(&ip_direct_nh);
365 #endif
366 }
367 
368 #ifdef VIMAGE
369 static void
370 ip_destroy(void *unused __unused)
371 {
372 	struct ifnet *ifp;
373 	int error;
374 
375 #ifdef	RSS
376 	netisr_unregister_vnet(&ip_direct_nh);
377 #endif
378 	netisr_unregister_vnet(&ip_nh);
379 
380 	if ((error = pfil_head_unregister(&V_inet_pfil_hook)) != 0)
381 		printf("%s: WARNING: unable to unregister pfil hook, "
382 		    "error %d\n", __func__, error);
383 
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 	IFNET_RLOCK();
402 	TAILQ_FOREACH(ifp, &V_ifnet, if_link)
403 		rt_flushifroutes_af(ifp, AF_INET);
404 	IFNET_RUNLOCK();
405 
406 	/* Destroy IP reassembly queue. */
407 	ipreass_destroy();
408 
409 	/* Cleanup in_ifaddr hash table; should be empty. */
410 	hashdestroy(V_in_ifaddrhashtbl, M_IFADDR, V_in_ifaddrhmask);
411 }
412 
413 VNET_SYSUNINIT(ip, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip_destroy, NULL);
414 #endif
415 
416 #ifdef	RSS
417 /*
418  * IP direct input routine.
419  *
420  * This is called when reinjecting completed fragments where
421  * all of the previous checking and book-keeping has been done.
422  */
423 void
424 ip_direct_input(struct mbuf *m)
425 {
426 	struct ip *ip;
427 	int hlen;
428 
429 	ip = mtod(m, struct ip *);
430 	hlen = ip->ip_hl << 2;
431 
432 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
433 	if (IPSEC_ENABLED(ipv4)) {
434 		if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
435 			return;
436 	}
437 #endif /* IPSEC */
438 	IPSTAT_INC(ips_delivered);
439 	(*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
440 	return;
441 }
442 #endif
443 
444 /*
445  * Ip input routine.  Checksum and byte swap header.  If fragmented
446  * try to reassemble.  Process options.  Pass to next level.
447  */
448 void
449 ip_input(struct mbuf *m)
450 {
451 	struct ip *ip = NULL;
452 	struct in_ifaddr *ia = NULL;
453 	struct ifaddr *ifa;
454 	struct ifnet *ifp;
455 	int    checkif, hlen = 0;
456 	uint16_t sum, ip_len;
457 	int dchg = 0;				/* dest changed after fw */
458 	struct in_addr odst;			/* original dst address */
459 
460 	M_ASSERTPKTHDR(m);
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 (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))) == NULL) {
478 		IPSTAT_INC(ips_toosmall);
479 		return;
480 	}
481 	ip = mtod(m, struct ip *);
482 
483 	if (ip->ip_v != IPVERSION) {
484 		IPSTAT_INC(ips_badvers);
485 		goto bad;
486 	}
487 
488 	hlen = ip->ip_hl << 2;
489 	if (hlen < sizeof(struct ip)) {	/* minimum header length */
490 		IPSTAT_INC(ips_badhlen);
491 		goto bad;
492 	}
493 	if (hlen > m->m_len) {
494 		if ((m = m_pullup(m, hlen)) == NULL) {
495 			IPSTAT_INC(ips_badhlen);
496 			return;
497 		}
498 		ip = mtod(m, struct ip *);
499 	}
500 
501 	IP_PROBE(receive, NULL, NULL, ip, m->m_pkthdr.rcvif, ip, NULL);
502 
503 	/* 127/8 must not appear on wire - RFC1122 */
504 	ifp = m->m_pkthdr.rcvif;
505 	if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
506 	    (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
507 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
508 			IPSTAT_INC(ips_badaddr);
509 			goto bad;
510 		}
511 	}
512 
513 	if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
514 		sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
515 	} else {
516 		if (hlen == sizeof(struct ip)) {
517 			sum = in_cksum_hdr(ip);
518 		} else {
519 			sum = in_cksum(m, hlen);
520 		}
521 	}
522 	if (sum) {
523 		IPSTAT_INC(ips_badsum);
524 		goto bad;
525 	}
526 
527 #ifdef ALTQ
528 	if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
529 		/* packet is dropped by traffic conditioner */
530 		return;
531 #endif
532 
533 	ip_len = ntohs(ip->ip_len);
534 	if (ip_len < hlen) {
535 		IPSTAT_INC(ips_badlen);
536 		goto bad;
537 	}
538 
539 	/*
540 	 * Check that the amount of data in the buffers
541 	 * is as at least much as the IP header would have us expect.
542 	 * Trim mbufs if longer than we expect.
543 	 * Drop packet if shorter than we expect.
544 	 */
545 	if (m->m_pkthdr.len < ip_len) {
546 tooshort:
547 		IPSTAT_INC(ips_tooshort);
548 		goto bad;
549 	}
550 	if (m->m_pkthdr.len > ip_len) {
551 		if (m->m_len == m->m_pkthdr.len) {
552 			m->m_len = ip_len;
553 			m->m_pkthdr.len = ip_len;
554 		} else
555 			m_adj(m, ip_len - m->m_pkthdr.len);
556 	}
557 
558 	/*
559 	 * Try to forward the packet, but if we fail continue.
560 	 * ip_tryforward() does inbound and outbound packet firewall
561 	 * processing. If firewall has decided that destination becomes
562 	 * our local address, it sets M_FASTFWD_OURS flag. In this
563 	 * case skip another inbound firewall processing and update
564 	 * ip pointer.
565 	 */
566 	if (V_ipforwarding != 0
567 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
568 	    && (!IPSEC_ENABLED(ipv4) ||
569 	    IPSEC_CAPS(ipv4, m, IPSEC_CAP_OPERABLE) == 0)
570 #endif
571 	    ) {
572 		if ((m = ip_tryforward(m)) == NULL)
573 			return;
574 		if (m->m_flags & M_FASTFWD_OURS) {
575 			m->m_flags &= ~M_FASTFWD_OURS;
576 			ip = mtod(m, struct ip *);
577 			goto ours;
578 		}
579 	}
580 
581 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
582 	/*
583 	 * Bypass packet filtering for packets previously handled by IPsec.
584 	 */
585 	if (IPSEC_ENABLED(ipv4) &&
586 	    IPSEC_CAPS(ipv4, m, IPSEC_CAP_BYPASS_FILTER) != 0)
587 			goto passin;
588 #endif
589 
590 	/*
591 	 * Run through list of hooks for input packets.
592 	 *
593 	 * NB: Beware of the destination address changing (e.g.
594 	 *     by NAT rewriting).  When this happens, tell
595 	 *     ip_forward to do the right thing.
596 	 */
597 
598 	/* Jump over all PFIL processing if hooks are not active. */
599 	if (!PFIL_HOOKED(&V_inet_pfil_hook))
600 		goto passin;
601 
602 	odst = ip->ip_dst;
603 	if (pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_IN, 0, NULL) != 0)
604 		return;
605 	if (m == NULL)			/* consumed by filter */
606 		return;
607 
608 	ip = mtod(m, struct ip *);
609 	dchg = (odst.s_addr != ip->ip_dst.s_addr);
610 	ifp = m->m_pkthdr.rcvif;
611 
612 	if (m->m_flags & M_FASTFWD_OURS) {
613 		m->m_flags &= ~M_FASTFWD_OURS;
614 		goto ours;
615 	}
616 	if (m->m_flags & M_IP_NEXTHOP) {
617 		if (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) {
618 			/*
619 			 * Directly ship the packet on.  This allows
620 			 * forwarding packets originally destined to us
621 			 * to some other directly connected host.
622 			 */
623 			ip_forward(m, 1);
624 			return;
625 		}
626 	}
627 passin:
628 
629 	/*
630 	 * Process options and, if not destined for us,
631 	 * ship it on.  ip_dooptions returns 1 when an
632 	 * error was detected (causing an icmp message
633 	 * to be sent and the original packet to be freed).
634 	 */
635 	if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
636 		return;
637 
638         /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
639          * matter if it is destined to another node, or whether it is
640          * a multicast one, RSVP wants it! and prevents it from being forwarded
641          * anywhere else. Also checks if the rsvp daemon is running before
642 	 * grabbing the packet.
643          */
644 	if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP)
645 		goto ours;
646 
647 	/*
648 	 * Check our list of addresses, to see if the packet is for us.
649 	 * If we don't have any addresses, assume any unicast packet
650 	 * we receive might be for us (and let the upper layers deal
651 	 * with it).
652 	 */
653 	if (TAILQ_EMPTY(&V_in_ifaddrhead) &&
654 	    (m->m_flags & (M_MCAST|M_BCAST)) == 0)
655 		goto ours;
656 
657 	/*
658 	 * Enable a consistency check between the destination address
659 	 * and the arrival interface for a unicast packet (the RFC 1122
660 	 * strong ES model) if IP forwarding is disabled and the packet
661 	 * is not locally generated and the packet is not subject to
662 	 * 'ipfw fwd'.
663 	 *
664 	 * XXX - Checking also should be disabled if the destination
665 	 * address is ipnat'ed to a different interface.
666 	 *
667 	 * XXX - Checking is incompatible with IP aliases added
668 	 * to the loopback interface instead of the interface where
669 	 * the packets are received.
670 	 *
671 	 * XXX - This is the case for carp vhost IPs as well so we
672 	 * insert a workaround. If the packet got here, we already
673 	 * checked with carp_iamatch() and carp_forus().
674 	 */
675 	checkif = V_ip_checkinterface && (V_ipforwarding == 0) &&
676 	    ifp != NULL && ((ifp->if_flags & IFF_LOOPBACK) == 0) &&
677 	    ifp->if_carp == NULL && (dchg == 0);
678 
679 	/*
680 	 * Check for exact addresses in the hash bucket.
681 	 */
682 	/* IN_IFADDR_RLOCK(); */
683 	LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
684 		/*
685 		 * If the address matches, verify that the packet
686 		 * arrived via the correct interface if checking is
687 		 * enabled.
688 		 */
689 		if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr &&
690 		    (!checkif || ia->ia_ifp == ifp)) {
691 			counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
692 			counter_u64_add(ia->ia_ifa.ifa_ibytes,
693 			    m->m_pkthdr.len);
694 			/* IN_IFADDR_RUNLOCK(); */
695 			goto ours;
696 		}
697 	}
698 	/* IN_IFADDR_RUNLOCK(); */
699 
700 	/*
701 	 * Check for broadcast addresses.
702 	 *
703 	 * Only accept broadcast packets that arrive via the matching
704 	 * interface.  Reception of forwarded directed broadcasts would
705 	 * be handled via ip_forward() and ether_output() with the loopback
706 	 * into the stack for SIMPLEX interfaces handled by ether_output().
707 	 */
708 	if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
709 		IF_ADDR_RLOCK(ifp);
710 	        TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
711 			if (ifa->ifa_addr->sa_family != AF_INET)
712 				continue;
713 			ia = ifatoia(ifa);
714 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
715 			    ip->ip_dst.s_addr) {
716 				counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
717 				counter_u64_add(ia->ia_ifa.ifa_ibytes,
718 				    m->m_pkthdr.len);
719 				IF_ADDR_RUNLOCK(ifp);
720 				goto ours;
721 			}
722 #ifdef BOOTP_COMPAT
723 			if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) {
724 				counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
725 				counter_u64_add(ia->ia_ifa.ifa_ibytes,
726 				    m->m_pkthdr.len);
727 				IF_ADDR_RUNLOCK(ifp);
728 				goto ours;
729 			}
730 #endif
731 		}
732 		IF_ADDR_RUNLOCK(ifp);
733 		ia = NULL;
734 	}
735 	/* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */
736 	if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) {
737 		IPSTAT_INC(ips_cantforward);
738 		m_freem(m);
739 		return;
740 	}
741 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
742 		if (V_ip_mrouter) {
743 			/*
744 			 * If we are acting as a multicast router, all
745 			 * incoming multicast packets are passed to the
746 			 * kernel-level multicast forwarding function.
747 			 * The packet is returned (relatively) intact; if
748 			 * ip_mforward() returns a non-zero value, the packet
749 			 * must be discarded, else it may be accepted below.
750 			 */
751 			if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) {
752 				IPSTAT_INC(ips_cantforward);
753 				m_freem(m);
754 				return;
755 			}
756 
757 			/*
758 			 * The process-level routing daemon needs to receive
759 			 * all multicast IGMP packets, whether or not this
760 			 * host belongs to their destination groups.
761 			 */
762 			if (ip->ip_p == IPPROTO_IGMP)
763 				goto ours;
764 			IPSTAT_INC(ips_forward);
765 		}
766 		/*
767 		 * Assume the packet is for us, to avoid prematurely taking
768 		 * a lock on the in_multi hash. Protocols must perform
769 		 * their own filtering and update statistics accordingly.
770 		 */
771 		goto ours;
772 	}
773 	if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
774 		goto ours;
775 	if (ip->ip_dst.s_addr == INADDR_ANY)
776 		goto ours;
777 
778 	/*
779 	 * Not for us; forward if possible and desirable.
780 	 */
781 	if (V_ipforwarding == 0) {
782 		IPSTAT_INC(ips_cantforward);
783 		m_freem(m);
784 	} else {
785 		ip_forward(m, dchg);
786 	}
787 	return;
788 
789 ours:
790 #ifdef IPSTEALTH
791 	/*
792 	 * IPSTEALTH: Process non-routing options only
793 	 * if the packet is destined for us.
794 	 */
795 	if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1))
796 		return;
797 #endif /* IPSTEALTH */
798 
799 	/*
800 	 * Attempt reassembly; if it succeeds, proceed.
801 	 * ip_reass() will return a different mbuf.
802 	 */
803 	if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) {
804 		/* XXXGL: shouldn't we save & set m_flags? */
805 		m = ip_reass(m);
806 		if (m == NULL)
807 			return;
808 		ip = mtod(m, struct ip *);
809 		/* Get the header length of the reassembled packet */
810 		hlen = ip->ip_hl << 2;
811 	}
812 
813 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
814 	if (IPSEC_ENABLED(ipv4)) {
815 		if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
816 			return;
817 	}
818 #endif /* IPSEC */
819 
820 	/*
821 	 * Switch out to protocol's input routine.
822 	 */
823 	IPSTAT_INC(ips_delivered);
824 
825 	(*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
826 	return;
827 bad:
828 	m_freem(m);
829 }
830 
831 /*
832  * IP timer processing;
833  * if a timer expires on a reassembly
834  * queue, discard it.
835  */
836 void
837 ip_slowtimo(void)
838 {
839 	VNET_ITERATOR_DECL(vnet_iter);
840 
841 	VNET_LIST_RLOCK_NOSLEEP();
842 	VNET_FOREACH(vnet_iter) {
843 		CURVNET_SET(vnet_iter);
844 		ipreass_slowtimo();
845 		CURVNET_RESTORE();
846 	}
847 	VNET_LIST_RUNLOCK_NOSLEEP();
848 }
849 
850 void
851 ip_drain(void)
852 {
853 	VNET_ITERATOR_DECL(vnet_iter);
854 
855 	VNET_LIST_RLOCK_NOSLEEP();
856 	VNET_FOREACH(vnet_iter) {
857 		CURVNET_SET(vnet_iter);
858 		ipreass_drain();
859 		CURVNET_RESTORE();
860 	}
861 	VNET_LIST_RUNLOCK_NOSLEEP();
862 }
863 
864 /*
865  * The protocol to be inserted into ip_protox[] must be already registered
866  * in inetsw[], either statically or through pf_proto_register().
867  */
868 int
869 ipproto_register(short ipproto)
870 {
871 	struct protosw *pr;
872 
873 	/* Sanity checks. */
874 	if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
875 		return (EPROTONOSUPPORT);
876 
877 	/*
878 	 * The protocol slot must not be occupied by another protocol
879 	 * already.  An index pointing to IPPROTO_RAW is unused.
880 	 */
881 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
882 	if (pr == NULL)
883 		return (EPFNOSUPPORT);
884 	if (ip_protox[ipproto] != pr - inetsw)	/* IPPROTO_RAW */
885 		return (EEXIST);
886 
887 	/* Find the protocol position in inetsw[] and set the index. */
888 	for (pr = inetdomain.dom_protosw;
889 	     pr < inetdomain.dom_protoswNPROTOSW; pr++) {
890 		if (pr->pr_domain->dom_family == PF_INET &&
891 		    pr->pr_protocol && pr->pr_protocol == ipproto) {
892 			ip_protox[pr->pr_protocol] = pr - inetsw;
893 			return (0);
894 		}
895 	}
896 	return (EPROTONOSUPPORT);
897 }
898 
899 int
900 ipproto_unregister(short ipproto)
901 {
902 	struct protosw *pr;
903 
904 	/* Sanity checks. */
905 	if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
906 		return (EPROTONOSUPPORT);
907 
908 	/* Check if the protocol was indeed registered. */
909 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
910 	if (pr == NULL)
911 		return (EPFNOSUPPORT);
912 	if (ip_protox[ipproto] == pr - inetsw)  /* IPPROTO_RAW */
913 		return (ENOENT);
914 
915 	/* Reset the protocol slot to IPPROTO_RAW. */
916 	ip_protox[ipproto] = pr - inetsw;
917 	return (0);
918 }
919 
920 u_char inetctlerrmap[PRC_NCMDS] = {
921 	0,		0,		0,		0,
922 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
923 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
924 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
925 	0,		0,		EHOSTUNREACH,	0,
926 	ENOPROTOOPT,	ECONNREFUSED
927 };
928 
929 /*
930  * Forward a packet.  If some error occurs return the sender
931  * an icmp packet.  Note we can't always generate a meaningful
932  * icmp message because icmp doesn't have a large enough repertoire
933  * of codes and types.
934  *
935  * If not forwarding, just drop the packet.  This could be confusing
936  * if ipforwarding was zero but some routing protocol was advancing
937  * us as a gateway to somewhere.  However, we must let the routing
938  * protocol deal with that.
939  *
940  * The srcrt parameter indicates whether the packet is being forwarded
941  * via a source route.
942  */
943 void
944 ip_forward(struct mbuf *m, int srcrt)
945 {
946 	struct ip *ip = mtod(m, struct ip *);
947 	struct in_ifaddr *ia;
948 	struct mbuf *mcopy;
949 	struct sockaddr_in *sin;
950 	struct in_addr dest;
951 	struct route ro;
952 	int error, type = 0, code = 0, mtu = 0;
953 
954 	if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
955 		IPSTAT_INC(ips_cantforward);
956 		m_freem(m);
957 		return;
958 	}
959 	if (
960 #ifdef IPSTEALTH
961 	    V_ipstealth == 0 &&
962 #endif
963 	    ip->ip_ttl <= IPTTLDEC) {
964 		icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 0, 0);
965 		return;
966 	}
967 
968 	bzero(&ro, sizeof(ro));
969 	sin = (struct sockaddr_in *)&ro.ro_dst;
970 	sin->sin_family = AF_INET;
971 	sin->sin_len = sizeof(*sin);
972 	sin->sin_addr = ip->ip_dst;
973 #ifdef RADIX_MPATH
974 	rtalloc_mpath_fib(&ro,
975 	    ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
976 	    M_GETFIB(m));
977 #else
978 	in_rtalloc_ign(&ro, 0, M_GETFIB(m));
979 #endif
980 	if (ro.ro_rt != NULL) {
981 		ia = ifatoia(ro.ro_rt->rt_ifa);
982 		ifa_ref(&ia->ia_ifa);
983 	} else
984 		ia = NULL;
985 	/*
986 	 * Save the IP header and at most 8 bytes of the payload,
987 	 * in case we need to generate an ICMP message to the src.
988 	 *
989 	 * XXX this can be optimized a lot by saving the data in a local
990 	 * buffer on the stack (72 bytes at most), and only allocating the
991 	 * mbuf if really necessary. The vast majority of the packets
992 	 * are forwarded without having to send an ICMP back (either
993 	 * because unnecessary, or because rate limited), so we are
994 	 * really we are wasting a lot of work here.
995 	 *
996 	 * We don't use m_copym() because it might return a reference
997 	 * to a shared cluster. Both this function and ip_output()
998 	 * assume exclusive access to the IP header in `m', so any
999 	 * data in a cluster may change before we reach icmp_error().
1000 	 */
1001 	mcopy = m_gethdr(M_NOWAIT, m->m_type);
1002 	if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_NOWAIT)) {
1003 		/*
1004 		 * It's probably ok if the pkthdr dup fails (because
1005 		 * the deep copy of the tag chain failed), but for now
1006 		 * be conservative and just discard the copy since
1007 		 * code below may some day want the tags.
1008 		 */
1009 		m_free(mcopy);
1010 		mcopy = NULL;
1011 	}
1012 	if (mcopy != NULL) {
1013 		mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy));
1014 		mcopy->m_pkthdr.len = mcopy->m_len;
1015 		m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1016 	}
1017 #ifdef IPSTEALTH
1018 	if (V_ipstealth == 0)
1019 #endif
1020 		ip->ip_ttl -= IPTTLDEC;
1021 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1022 	if (IPSEC_ENABLED(ipv4)) {
1023 		if ((error = IPSEC_FORWARD(ipv4, m)) != 0) {
1024 			/* mbuf consumed by IPsec */
1025 			m_freem(mcopy);
1026 			if (error != EINPROGRESS)
1027 				IPSTAT_INC(ips_cantforward);
1028 			return;
1029 		}
1030 		/* No IPsec processing required */
1031 	}
1032 #endif /* IPSEC */
1033 	/*
1034 	 * If forwarding packet using same interface that it came in on,
1035 	 * perhaps should send a redirect to sender to shortcut a hop.
1036 	 * Only send redirect if source is sending directly to us,
1037 	 * and if packet was not source routed (or has any options).
1038 	 * Also, don't send redirect if forwarding using a default route
1039 	 * or a route modified by a redirect.
1040 	 */
1041 	dest.s_addr = 0;
1042 	if (!srcrt && V_ipsendredirects &&
1043 	    ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) {
1044 		struct rtentry *rt;
1045 
1046 		rt = ro.ro_rt;
1047 
1048 		if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1049 		    satosin(rt_key(rt))->sin_addr.s_addr != 0) {
1050 #define	RTA(rt)	((struct in_ifaddr *)(rt->rt_ifa))
1051 			u_long src = ntohl(ip->ip_src.s_addr);
1052 
1053 			if (RTA(rt) &&
1054 			    (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1055 				if (rt->rt_flags & RTF_GATEWAY)
1056 					dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr;
1057 				else
1058 					dest.s_addr = ip->ip_dst.s_addr;
1059 				/* Router requirements says to only send host redirects */
1060 				type = ICMP_REDIRECT;
1061 				code = ICMP_REDIRECT_HOST;
1062 			}
1063 		}
1064 	}
1065 
1066 	error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
1067 
1068 	if (error == EMSGSIZE && ro.ro_rt)
1069 		mtu = ro.ro_rt->rt_mtu;
1070 	RO_RTFREE(&ro);
1071 
1072 	if (error)
1073 		IPSTAT_INC(ips_cantforward);
1074 	else {
1075 		IPSTAT_INC(ips_forward);
1076 		if (type)
1077 			IPSTAT_INC(ips_redirectsent);
1078 		else {
1079 			if (mcopy)
1080 				m_freem(mcopy);
1081 			if (ia != NULL)
1082 				ifa_free(&ia->ia_ifa);
1083 			return;
1084 		}
1085 	}
1086 	if (mcopy == NULL) {
1087 		if (ia != NULL)
1088 			ifa_free(&ia->ia_ifa);
1089 		return;
1090 	}
1091 
1092 	switch (error) {
1093 
1094 	case 0:				/* forwarded, but need redirect */
1095 		/* type, code set above */
1096 		break;
1097 
1098 	case ENETUNREACH:
1099 	case EHOSTUNREACH:
1100 	case ENETDOWN:
1101 	case EHOSTDOWN:
1102 	default:
1103 		type = ICMP_UNREACH;
1104 		code = ICMP_UNREACH_HOST;
1105 		break;
1106 
1107 	case EMSGSIZE:
1108 		type = ICMP_UNREACH;
1109 		code = ICMP_UNREACH_NEEDFRAG;
1110 		/*
1111 		 * If the MTU was set before make sure we are below the
1112 		 * interface MTU.
1113 		 * If the MTU wasn't set before use the interface mtu or
1114 		 * fall back to the next smaller mtu step compared to the
1115 		 * current packet size.
1116 		 */
1117 		if (mtu != 0) {
1118 			if (ia != NULL)
1119 				mtu = min(mtu, ia->ia_ifp->if_mtu);
1120 		} else {
1121 			if (ia != NULL)
1122 				mtu = ia->ia_ifp->if_mtu;
1123 			else
1124 				mtu = ip_next_mtu(ntohs(ip->ip_len), 0);
1125 		}
1126 		IPSTAT_INC(ips_cantfrag);
1127 		break;
1128 
1129 	case ENOBUFS:
1130 	case EACCES:			/* ipfw denied packet */
1131 		m_freem(mcopy);
1132 		if (ia != NULL)
1133 			ifa_free(&ia->ia_ifa);
1134 		return;
1135 	}
1136 	if (ia != NULL)
1137 		ifa_free(&ia->ia_ifa);
1138 	icmp_error(mcopy, type, code, dest.s_addr, mtu);
1139 }
1140 
1141 #define	CHECK_SO_CT(sp, ct) \
1142     (((sp->so_options & SO_TIMESTAMP) && (sp->so_ts_clock == ct)) ? 1 : 0)
1143 
1144 void
1145 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1146     struct mbuf *m)
1147 {
1148 	bool stamped;
1149 
1150 	stamped = false;
1151 	if ((inp->inp_socket->so_options & SO_BINTIME) ||
1152 	    CHECK_SO_CT(inp->inp_socket, SO_TS_BINTIME)) {
1153 		struct bintime boottimebin, bt;
1154 		struct timespec ts1;
1155 
1156 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1157 		    M_TSTMP)) {
1158 			mbuf_tstmp2timespec(m, &ts1);
1159 			timespec2bintime(&ts1, &bt);
1160 			getboottimebin(&boottimebin);
1161 			bintime_add(&bt, &boottimebin);
1162 		} else {
1163 			bintime(&bt);
1164 		}
1165 		*mp = sbcreatecontrol((caddr_t)&bt, sizeof(bt),
1166 		    SCM_BINTIME, SOL_SOCKET);
1167 		if (*mp != NULL) {
1168 			mp = &(*mp)->m_next;
1169 			stamped = true;
1170 		}
1171 	}
1172 	if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME_MICRO)) {
1173 		struct bintime boottimebin, bt1;
1174 		struct timespec ts1;;
1175 		struct timeval tv;
1176 
1177 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1178 		    M_TSTMP)) {
1179 			mbuf_tstmp2timespec(m, &ts1);
1180 			timespec2bintime(&ts1, &bt1);
1181 			getboottimebin(&boottimebin);
1182 			bintime_add(&bt1, &boottimebin);
1183 			bintime2timeval(&bt1, &tv);
1184 		} else {
1185 			microtime(&tv);
1186 		}
1187 		*mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv),
1188 		    SCM_TIMESTAMP, SOL_SOCKET);
1189 		if (*mp != NULL) {
1190 			mp = &(*mp)->m_next;
1191 			stamped = true;
1192 		}
1193 	} else if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME)) {
1194 		struct bintime boottimebin;
1195 		struct timespec ts, ts1;
1196 
1197 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1198 		    M_TSTMP)) {
1199 			mbuf_tstmp2timespec(m, &ts);
1200 			getboottimebin(&boottimebin);
1201 			bintime2timespec(&boottimebin, &ts1);
1202 			timespecadd(&ts, &ts1);
1203 		} else {
1204 			nanotime(&ts);
1205 		}
1206 		*mp = sbcreatecontrol((caddr_t)&ts, sizeof(ts),
1207 		    SCM_REALTIME, SOL_SOCKET);
1208 		if (*mp != NULL) {
1209 			mp = &(*mp)->m_next;
1210 			stamped = true;
1211 		}
1212 	} else if (CHECK_SO_CT(inp->inp_socket, SO_TS_MONOTONIC)) {
1213 		struct timespec ts;
1214 
1215 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1216 		    M_TSTMP))
1217 			mbuf_tstmp2timespec(m, &ts);
1218 		else
1219 			nanouptime(&ts);
1220 		*mp = sbcreatecontrol((caddr_t)&ts, sizeof(ts),
1221 		    SCM_MONOTONIC, SOL_SOCKET);
1222 		if (*mp != NULL) {
1223 			mp = &(*mp)->m_next;
1224 			stamped = true;
1225 		}
1226 	}
1227 	if (stamped && (m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1228 	    M_TSTMP)) {
1229 		struct sock_timestamp_info sti;
1230 
1231 		bzero(&sti, sizeof(sti));
1232 		sti.st_info_flags = ST_INFO_HW;
1233 		if ((m->m_flags & M_TSTMP_HPREC) != 0)
1234 			sti.st_info_flags |= ST_INFO_HW_HPREC;
1235 		*mp = sbcreatecontrol((caddr_t)&sti, sizeof(sti), SCM_TIME_INFO,
1236 		    SOL_SOCKET);
1237 		if (*mp != NULL)
1238 			mp = &(*mp)->m_next;
1239 	}
1240 	if (inp->inp_flags & INP_RECVDSTADDR) {
1241 		*mp = sbcreatecontrol((caddr_t)&ip->ip_dst,
1242 		    sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1243 		if (*mp)
1244 			mp = &(*mp)->m_next;
1245 	}
1246 	if (inp->inp_flags & INP_RECVTTL) {
1247 		*mp = sbcreatecontrol((caddr_t)&ip->ip_ttl,
1248 		    sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
1249 		if (*mp)
1250 			mp = &(*mp)->m_next;
1251 	}
1252 #ifdef notyet
1253 	/* XXX
1254 	 * Moving these out of udp_input() made them even more broken
1255 	 * than they already were.
1256 	 */
1257 	/* options were tossed already */
1258 	if (inp->inp_flags & INP_RECVOPTS) {
1259 		*mp = sbcreatecontrol((caddr_t)opts_deleted_above,
1260 		    sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1261 		if (*mp)
1262 			mp = &(*mp)->m_next;
1263 	}
1264 	/* ip_srcroute doesn't do what we want here, need to fix */
1265 	if (inp->inp_flags & INP_RECVRETOPTS) {
1266 		*mp = sbcreatecontrol((caddr_t)ip_srcroute(m),
1267 		    sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1268 		if (*mp)
1269 			mp = &(*mp)->m_next;
1270 	}
1271 #endif
1272 	if (inp->inp_flags & INP_RECVIF) {
1273 		struct ifnet *ifp;
1274 		struct sdlbuf {
1275 			struct sockaddr_dl sdl;
1276 			u_char	pad[32];
1277 		} sdlbuf;
1278 		struct sockaddr_dl *sdp;
1279 		struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1280 
1281 		if ((ifp = m->m_pkthdr.rcvif) &&
1282 		    ifp->if_index && ifp->if_index <= V_if_index) {
1283 			sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1284 			/*
1285 			 * Change our mind and don't try copy.
1286 			 */
1287 			if (sdp->sdl_family != AF_LINK ||
1288 			    sdp->sdl_len > sizeof(sdlbuf)) {
1289 				goto makedummy;
1290 			}
1291 			bcopy(sdp, sdl2, sdp->sdl_len);
1292 		} else {
1293 makedummy:
1294 			sdl2->sdl_len =
1295 			    offsetof(struct sockaddr_dl, sdl_data[0]);
1296 			sdl2->sdl_family = AF_LINK;
1297 			sdl2->sdl_index = 0;
1298 			sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1299 		}
1300 		*mp = sbcreatecontrol((caddr_t)sdl2, sdl2->sdl_len,
1301 		    IP_RECVIF, IPPROTO_IP);
1302 		if (*mp)
1303 			mp = &(*mp)->m_next;
1304 	}
1305 	if (inp->inp_flags & INP_RECVTOS) {
1306 		*mp = sbcreatecontrol((caddr_t)&ip->ip_tos,
1307 		    sizeof(u_char), IP_RECVTOS, IPPROTO_IP);
1308 		if (*mp)
1309 			mp = &(*mp)->m_next;
1310 	}
1311 
1312 	if (inp->inp_flags2 & INP_RECVFLOWID) {
1313 		uint32_t flowid, flow_type;
1314 
1315 		flowid = m->m_pkthdr.flowid;
1316 		flow_type = M_HASHTYPE_GET(m);
1317 
1318 		/*
1319 		 * XXX should handle the failure of one or the
1320 		 * other - don't populate both?
1321 		 */
1322 		*mp = sbcreatecontrol((caddr_t) &flowid,
1323 		    sizeof(uint32_t), IP_FLOWID, IPPROTO_IP);
1324 		if (*mp)
1325 			mp = &(*mp)->m_next;
1326 		*mp = sbcreatecontrol((caddr_t) &flow_type,
1327 		    sizeof(uint32_t), IP_FLOWTYPE, IPPROTO_IP);
1328 		if (*mp)
1329 			mp = &(*mp)->m_next;
1330 	}
1331 
1332 #ifdef	RSS
1333 	if (inp->inp_flags2 & INP_RECVRSSBUCKETID) {
1334 		uint32_t flowid, flow_type;
1335 		uint32_t rss_bucketid;
1336 
1337 		flowid = m->m_pkthdr.flowid;
1338 		flow_type = M_HASHTYPE_GET(m);
1339 
1340 		if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) {
1341 			*mp = sbcreatecontrol((caddr_t) &rss_bucketid,
1342 			   sizeof(uint32_t), IP_RSSBUCKETID, IPPROTO_IP);
1343 			if (*mp)
1344 				mp = &(*mp)->m_next;
1345 		}
1346 	}
1347 #endif
1348 }
1349 
1350 /*
1351  * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
1352  * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
1353  * locking.  This code remains in ip_input.c as ip_mroute.c is optionally
1354  * compiled.
1355  */
1356 static VNET_DEFINE(int, ip_rsvp_on);
1357 VNET_DEFINE(struct socket *, ip_rsvpd);
1358 
1359 #define	V_ip_rsvp_on		VNET(ip_rsvp_on)
1360 
1361 int
1362 ip_rsvp_init(struct socket *so)
1363 {
1364 
1365 	if (so->so_type != SOCK_RAW ||
1366 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
1367 		return EOPNOTSUPP;
1368 
1369 	if (V_ip_rsvpd != NULL)
1370 		return EADDRINUSE;
1371 
1372 	V_ip_rsvpd = so;
1373 	/*
1374 	 * This may seem silly, but we need to be sure we don't over-increment
1375 	 * the RSVP counter, in case something slips up.
1376 	 */
1377 	if (!V_ip_rsvp_on) {
1378 		V_ip_rsvp_on = 1;
1379 		V_rsvp_on++;
1380 	}
1381 
1382 	return 0;
1383 }
1384 
1385 int
1386 ip_rsvp_done(void)
1387 {
1388 
1389 	V_ip_rsvpd = NULL;
1390 	/*
1391 	 * This may seem silly, but we need to be sure we don't over-decrement
1392 	 * the RSVP counter, in case something slips up.
1393 	 */
1394 	if (V_ip_rsvp_on) {
1395 		V_ip_rsvp_on = 0;
1396 		V_rsvp_on--;
1397 	}
1398 	return 0;
1399 }
1400 
1401 int
1402 rsvp_input(struct mbuf **mp, int *offp, int proto)
1403 {
1404 	struct mbuf *m;
1405 
1406 	m = *mp;
1407 	*mp = NULL;
1408 
1409 	if (rsvp_input_p) { /* call the real one if loaded */
1410 		*mp = m;
1411 		rsvp_input_p(mp, offp, proto);
1412 		return (IPPROTO_DONE);
1413 	}
1414 
1415 	/* Can still get packets with rsvp_on = 0 if there is a local member
1416 	 * of the group to which the RSVP packet is addressed.  But in this
1417 	 * case we want to throw the packet away.
1418 	 */
1419 
1420 	if (!V_rsvp_on) {
1421 		m_freem(m);
1422 		return (IPPROTO_DONE);
1423 	}
1424 
1425 	if (V_ip_rsvpd != NULL) {
1426 		*mp = m;
1427 		rip_input(mp, offp, proto);
1428 		return (IPPROTO_DONE);
1429 	}
1430 	/* Drop the packet */
1431 	m_freem(m);
1432 	return (IPPROTO_DONE);
1433 }
1434