xref: /freebsd/sys/netinet/ip_input.c (revision 4db78cacdee1b6f3b7880eb8c5560e8edaf91698)
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 VNET_DEFINE_STATIC(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 VNET_DEFINE_STATIC(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 	CK_STAILQ_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 	CK_STAILQ_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 rm_priotracker in_ifa_tracker;
452 	struct ip *ip = NULL;
453 	struct in_ifaddr *ia = NULL;
454 	struct ifaddr *ifa;
455 	struct ifnet *ifp;
456 	int    checkif, hlen = 0;
457 	uint16_t sum, ip_len;
458 	int dchg = 0;				/* dest changed after fw */
459 	struct in_addr odst;			/* original dst address */
460 
461 	M_ASSERTPKTHDR(m);
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 (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))) == NULL) {
479 		IPSTAT_INC(ips_toosmall);
480 		return;
481 	}
482 	ip = mtod(m, struct ip *);
483 
484 	if (ip->ip_v != IPVERSION) {
485 		IPSTAT_INC(ips_badvers);
486 		goto bad;
487 	}
488 
489 	hlen = ip->ip_hl << 2;
490 	if (hlen < sizeof(struct ip)) {	/* minimum header length */
491 		IPSTAT_INC(ips_badhlen);
492 		goto bad;
493 	}
494 	if (hlen > m->m_len) {
495 		if ((m = m_pullup(m, hlen)) == NULL) {
496 			IPSTAT_INC(ips_badhlen);
497 			return;
498 		}
499 		ip = mtod(m, struct ip *);
500 	}
501 
502 	IP_PROBE(receive, NULL, NULL, ip, m->m_pkthdr.rcvif, ip, NULL);
503 
504 	/* 127/8 must not appear on wire - RFC1122 */
505 	ifp = m->m_pkthdr.rcvif;
506 	if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
507 	    (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
508 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
509 			IPSTAT_INC(ips_badaddr);
510 			goto bad;
511 		}
512 	}
513 
514 	if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
515 		sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
516 	} else {
517 		if (hlen == sizeof(struct ip)) {
518 			sum = in_cksum_hdr(ip);
519 		} else {
520 			sum = in_cksum(m, hlen);
521 		}
522 	}
523 	if (sum) {
524 		IPSTAT_INC(ips_badsum);
525 		goto bad;
526 	}
527 
528 #ifdef ALTQ
529 	if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
530 		/* packet is dropped by traffic conditioner */
531 		return;
532 #endif
533 
534 	ip_len = ntohs(ip->ip_len);
535 	if (ip_len < hlen) {
536 		IPSTAT_INC(ips_badlen);
537 		goto bad;
538 	}
539 
540 	/*
541 	 * Check that the amount of data in the buffers
542 	 * is as at least much as the IP header would have us expect.
543 	 * Trim mbufs if longer than we expect.
544 	 * Drop packet if shorter than we expect.
545 	 */
546 	if (m->m_pkthdr.len < ip_len) {
547 tooshort:
548 		IPSTAT_INC(ips_tooshort);
549 		goto bad;
550 	}
551 	if (m->m_pkthdr.len > ip_len) {
552 		if (m->m_len == m->m_pkthdr.len) {
553 			m->m_len = ip_len;
554 			m->m_pkthdr.len = ip_len;
555 		} else
556 			m_adj(m, ip_len - m->m_pkthdr.len);
557 	}
558 
559 	/*
560 	 * Try to forward the packet, but if we fail continue.
561 	 * ip_tryforward() does not generate redirects, so fall
562 	 * through to normal processing if redirects are required.
563 	 * ip_tryforward() does inbound and outbound packet firewall
564 	 * processing. If firewall has decided that destination becomes
565 	 * our local address, it sets M_FASTFWD_OURS flag. In this
566 	 * case skip another inbound firewall processing and update
567 	 * ip pointer.
568 	 */
569 	if (V_ipforwarding != 0 && V_ipsendredirects == 0
570 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
571 	    && (!IPSEC_ENABLED(ipv4) ||
572 	    IPSEC_CAPS(ipv4, m, IPSEC_CAP_OPERABLE) == 0)
573 #endif
574 	    ) {
575 		if ((m = ip_tryforward(m)) == NULL)
576 			return;
577 		if (m->m_flags & M_FASTFWD_OURS) {
578 			m->m_flags &= ~M_FASTFWD_OURS;
579 			ip = mtod(m, struct ip *);
580 			goto ours;
581 		}
582 	}
583 
584 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
585 	/*
586 	 * Bypass packet filtering for packets previously handled by IPsec.
587 	 */
588 	if (IPSEC_ENABLED(ipv4) &&
589 	    IPSEC_CAPS(ipv4, m, IPSEC_CAP_BYPASS_FILTER) != 0)
590 			goto passin;
591 #endif
592 
593 	/*
594 	 * Run through list of hooks for input packets.
595 	 *
596 	 * NB: Beware of the destination address changing (e.g.
597 	 *     by NAT rewriting).  When this happens, tell
598 	 *     ip_forward to do the right thing.
599 	 */
600 
601 	/* Jump over all PFIL processing if hooks are not active. */
602 	if (!PFIL_HOOKED(&V_inet_pfil_hook))
603 		goto passin;
604 
605 	odst = ip->ip_dst;
606 	if (pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_IN, 0, NULL) != 0)
607 		return;
608 	if (m == NULL)			/* consumed by filter */
609 		return;
610 
611 	ip = mtod(m, struct ip *);
612 	dchg = (odst.s_addr != ip->ip_dst.s_addr);
613 	ifp = m->m_pkthdr.rcvif;
614 
615 	if (m->m_flags & M_FASTFWD_OURS) {
616 		m->m_flags &= ~M_FASTFWD_OURS;
617 		goto ours;
618 	}
619 	if (m->m_flags & M_IP_NEXTHOP) {
620 		if (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) {
621 			/*
622 			 * Directly ship the packet on.  This allows
623 			 * forwarding packets originally destined to us
624 			 * to some other directly connected host.
625 			 */
626 			ip_forward(m, 1);
627 			return;
628 		}
629 	}
630 passin:
631 
632 	/*
633 	 * Process options and, if not destined for us,
634 	 * ship it on.  ip_dooptions returns 1 when an
635 	 * error was detected (causing an icmp message
636 	 * to be sent and the original packet to be freed).
637 	 */
638 	if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
639 		return;
640 
641         /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
642          * matter if it is destined to another node, or whether it is
643          * a multicast one, RSVP wants it! and prevents it from being forwarded
644          * anywhere else. Also checks if the rsvp daemon is running before
645 	 * grabbing the packet.
646          */
647 	if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP)
648 		goto ours;
649 
650 	/*
651 	 * Check our list of addresses, to see if the packet is for us.
652 	 * If we don't have any addresses, assume any unicast packet
653 	 * we receive might be for us (and let the upper layers deal
654 	 * with it).
655 	 */
656 	if (CK_STAILQ_EMPTY(&V_in_ifaddrhead) &&
657 	    (m->m_flags & (M_MCAST|M_BCAST)) == 0)
658 		goto ours;
659 
660 	/*
661 	 * Enable a consistency check between the destination address
662 	 * and the arrival interface for a unicast packet (the RFC 1122
663 	 * strong ES model) if IP forwarding is disabled and the packet
664 	 * is not locally generated and the packet is not subject to
665 	 * 'ipfw fwd'.
666 	 *
667 	 * XXX - Checking also should be disabled if the destination
668 	 * address is ipnat'ed to a different interface.
669 	 *
670 	 * XXX - Checking is incompatible with IP aliases added
671 	 * to the loopback interface instead of the interface where
672 	 * the packets are received.
673 	 *
674 	 * XXX - This is the case for carp vhost IPs as well so we
675 	 * insert a workaround. If the packet got here, we already
676 	 * checked with carp_iamatch() and carp_forus().
677 	 */
678 	checkif = V_ip_checkinterface && (V_ipforwarding == 0) &&
679 	    ifp != NULL && ((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 	IN_IFADDR_RLOCK(&in_ifa_tracker);
686 	LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
687 		/*
688 		 * If the address matches, verify that the packet
689 		 * arrived via the correct interface if checking is
690 		 * enabled.
691 		 */
692 		if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr &&
693 		    (!checkif || ia->ia_ifp == ifp)) {
694 			counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
695 			counter_u64_add(ia->ia_ifa.ifa_ibytes,
696 			    m->m_pkthdr.len);
697 			IN_IFADDR_RUNLOCK(&in_ifa_tracker);
698 			goto ours;
699 		}
700 	}
701 	IN_IFADDR_RUNLOCK(&in_ifa_tracker);
702 
703 	/*
704 	 * Check for broadcast addresses.
705 	 *
706 	 * Only accept broadcast packets that arrive via the matching
707 	 * interface.  Reception of forwarded directed broadcasts would
708 	 * be handled via ip_forward() and ether_output() with the loopback
709 	 * into the stack for SIMPLEX interfaces handled by ether_output().
710 	 */
711 	if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
712 		IF_ADDR_RLOCK(ifp);
713 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
714 			if (ifa->ifa_addr->sa_family != AF_INET)
715 				continue;
716 			ia = ifatoia(ifa);
717 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
718 			    ip->ip_dst.s_addr) {
719 				counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
720 				counter_u64_add(ia->ia_ifa.ifa_ibytes,
721 				    m->m_pkthdr.len);
722 				IF_ADDR_RUNLOCK(ifp);
723 				goto ours;
724 			}
725 #ifdef BOOTP_COMPAT
726 			if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) {
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 				IF_ADDR_RUNLOCK(ifp);
731 				goto ours;
732 			}
733 #endif
734 		}
735 		IF_ADDR_RUNLOCK(ifp);
736 		ia = NULL;
737 	}
738 	/* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */
739 	if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) {
740 		IPSTAT_INC(ips_cantforward);
741 		m_freem(m);
742 		return;
743 	}
744 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
745 		if (V_ip_mrouter) {
746 			/*
747 			 * If we are acting as a multicast router, all
748 			 * incoming multicast packets are passed to the
749 			 * kernel-level multicast forwarding function.
750 			 * The packet is returned (relatively) intact; if
751 			 * ip_mforward() returns a non-zero value, the packet
752 			 * must be discarded, else it may be accepted below.
753 			 */
754 			if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) {
755 				IPSTAT_INC(ips_cantforward);
756 				m_freem(m);
757 				return;
758 			}
759 
760 			/*
761 			 * The process-level routing daemon needs to receive
762 			 * all multicast IGMP packets, whether or not this
763 			 * host belongs to their destination groups.
764 			 */
765 			if (ip->ip_p == IPPROTO_IGMP)
766 				goto ours;
767 			IPSTAT_INC(ips_forward);
768 		}
769 		/*
770 		 * Assume the packet is for us, to avoid prematurely taking
771 		 * a lock on the in_multi hash. Protocols must perform
772 		 * their own filtering and update statistics accordingly.
773 		 */
774 		goto ours;
775 	}
776 	if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
777 		goto ours;
778 	if (ip->ip_dst.s_addr == INADDR_ANY)
779 		goto ours;
780 
781 	/*
782 	 * Not for us; forward if possible and desirable.
783 	 */
784 	if (V_ipforwarding == 0) {
785 		IPSTAT_INC(ips_cantforward);
786 		m_freem(m);
787 	} else {
788 		ip_forward(m, dchg);
789 	}
790 	return;
791 
792 ours:
793 #ifdef IPSTEALTH
794 	/*
795 	 * IPSTEALTH: Process non-routing options only
796 	 * if the packet is destined for us.
797 	 */
798 	if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1))
799 		return;
800 #endif /* IPSTEALTH */
801 
802 	/*
803 	 * Attempt reassembly; if it succeeds, proceed.
804 	 * ip_reass() will return a different mbuf.
805 	 */
806 	if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) {
807 		/* XXXGL: shouldn't we save & set m_flags? */
808 		m = ip_reass(m);
809 		if (m == NULL)
810 			return;
811 		ip = mtod(m, struct ip *);
812 		/* Get the header length of the reassembled packet */
813 		hlen = ip->ip_hl << 2;
814 	}
815 
816 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
817 	if (IPSEC_ENABLED(ipv4)) {
818 		if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
819 			return;
820 	}
821 #endif /* IPSEC */
822 
823 	/*
824 	 * Switch out to protocol's input routine.
825 	 */
826 	IPSTAT_INC(ips_delivered);
827 
828 	(*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
829 	return;
830 bad:
831 	m_freem(m);
832 }
833 
834 /*
835  * IP timer processing;
836  * if a timer expires on a reassembly
837  * queue, discard it.
838  */
839 void
840 ip_slowtimo(void)
841 {
842 	VNET_ITERATOR_DECL(vnet_iter);
843 
844 	VNET_LIST_RLOCK_NOSLEEP();
845 	VNET_FOREACH(vnet_iter) {
846 		CURVNET_SET(vnet_iter);
847 		ipreass_slowtimo();
848 		CURVNET_RESTORE();
849 	}
850 	VNET_LIST_RUNLOCK_NOSLEEP();
851 }
852 
853 void
854 ip_drain(void)
855 {
856 	VNET_ITERATOR_DECL(vnet_iter);
857 
858 	VNET_LIST_RLOCK_NOSLEEP();
859 	VNET_FOREACH(vnet_iter) {
860 		CURVNET_SET(vnet_iter);
861 		ipreass_drain();
862 		CURVNET_RESTORE();
863 	}
864 	VNET_LIST_RUNLOCK_NOSLEEP();
865 }
866 
867 /*
868  * The protocol to be inserted into ip_protox[] must be already registered
869  * in inetsw[], either statically or through pf_proto_register().
870  */
871 int
872 ipproto_register(short ipproto)
873 {
874 	struct protosw *pr;
875 
876 	/* Sanity checks. */
877 	if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
878 		return (EPROTONOSUPPORT);
879 
880 	/*
881 	 * The protocol slot must not be occupied by another protocol
882 	 * already.  An index pointing to IPPROTO_RAW is unused.
883 	 */
884 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
885 	if (pr == NULL)
886 		return (EPFNOSUPPORT);
887 	if (ip_protox[ipproto] != pr - inetsw)	/* IPPROTO_RAW */
888 		return (EEXIST);
889 
890 	/* Find the protocol position in inetsw[] and set the index. */
891 	for (pr = inetdomain.dom_protosw;
892 	     pr < inetdomain.dom_protoswNPROTOSW; pr++) {
893 		if (pr->pr_domain->dom_family == PF_INET &&
894 		    pr->pr_protocol && pr->pr_protocol == ipproto) {
895 			ip_protox[pr->pr_protocol] = pr - inetsw;
896 			return (0);
897 		}
898 	}
899 	return (EPROTONOSUPPORT);
900 }
901 
902 int
903 ipproto_unregister(short ipproto)
904 {
905 	struct protosw *pr;
906 
907 	/* Sanity checks. */
908 	if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
909 		return (EPROTONOSUPPORT);
910 
911 	/* Check if the protocol was indeed registered. */
912 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
913 	if (pr == NULL)
914 		return (EPFNOSUPPORT);
915 	if (ip_protox[ipproto] == pr - inetsw)  /* IPPROTO_RAW */
916 		return (ENOENT);
917 
918 	/* Reset the protocol slot to IPPROTO_RAW. */
919 	ip_protox[ipproto] = pr - inetsw;
920 	return (0);
921 }
922 
923 u_char inetctlerrmap[PRC_NCMDS] = {
924 	0,		0,		0,		0,
925 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
926 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
927 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
928 	0,		0,		EHOSTUNREACH,	0,
929 	ENOPROTOOPT,	ECONNREFUSED
930 };
931 
932 /*
933  * Forward a packet.  If some error occurs return the sender
934  * an icmp packet.  Note we can't always generate a meaningful
935  * icmp message because icmp doesn't have a large enough repertoire
936  * of codes and types.
937  *
938  * If not forwarding, just drop the packet.  This could be confusing
939  * if ipforwarding was zero but some routing protocol was advancing
940  * us as a gateway to somewhere.  However, we must let the routing
941  * protocol deal with that.
942  *
943  * The srcrt parameter indicates whether the packet is being forwarded
944  * via a source route.
945  */
946 void
947 ip_forward(struct mbuf *m, int srcrt)
948 {
949 	struct ip *ip = mtod(m, struct ip *);
950 	struct in_ifaddr *ia;
951 	struct mbuf *mcopy;
952 	struct sockaddr_in *sin;
953 	struct in_addr dest;
954 	struct route ro;
955 	int error, type = 0, code = 0, mtu = 0;
956 
957 	if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
958 		IPSTAT_INC(ips_cantforward);
959 		m_freem(m);
960 		return;
961 	}
962 	if (
963 #ifdef IPSTEALTH
964 	    V_ipstealth == 0 &&
965 #endif
966 	    ip->ip_ttl <= IPTTLDEC) {
967 		icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 0, 0);
968 		return;
969 	}
970 
971 	bzero(&ro, sizeof(ro));
972 	sin = (struct sockaddr_in *)&ro.ro_dst;
973 	sin->sin_family = AF_INET;
974 	sin->sin_len = sizeof(*sin);
975 	sin->sin_addr = ip->ip_dst;
976 #ifdef RADIX_MPATH
977 	rtalloc_mpath_fib(&ro,
978 	    ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
979 	    M_GETFIB(m));
980 #else
981 	in_rtalloc_ign(&ro, 0, M_GETFIB(m));
982 #endif
983 	NET_EPOCH_ENTER();
984 	if (ro.ro_rt != NULL) {
985 		ia = ifatoia(ro.ro_rt->rt_ifa);
986 	} else
987 		ia = NULL;
988 	/*
989 	 * Save the IP header and at most 8 bytes of the payload,
990 	 * in case we need to generate an ICMP message to the src.
991 	 *
992 	 * XXX this can be optimized a lot by saving the data in a local
993 	 * buffer on the stack (72 bytes at most), and only allocating the
994 	 * mbuf if really necessary. The vast majority of the packets
995 	 * are forwarded without having to send an ICMP back (either
996 	 * because unnecessary, or because rate limited), so we are
997 	 * really we are wasting a lot of work here.
998 	 *
999 	 * We don't use m_copym() because it might return a reference
1000 	 * to a shared cluster. Both this function and ip_output()
1001 	 * assume exclusive access to the IP header in `m', so any
1002 	 * data in a cluster may change before we reach icmp_error().
1003 	 */
1004 	mcopy = m_gethdr(M_NOWAIT, m->m_type);
1005 	if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_NOWAIT)) {
1006 		/*
1007 		 * It's probably ok if the pkthdr dup fails (because
1008 		 * the deep copy of the tag chain failed), but for now
1009 		 * be conservative and just discard the copy since
1010 		 * code below may some day want the tags.
1011 		 */
1012 		m_free(mcopy);
1013 		mcopy = NULL;
1014 	}
1015 	if (mcopy != NULL) {
1016 		mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy));
1017 		mcopy->m_pkthdr.len = mcopy->m_len;
1018 		m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1019 	}
1020 #ifdef IPSTEALTH
1021 	if (V_ipstealth == 0)
1022 #endif
1023 		ip->ip_ttl -= IPTTLDEC;
1024 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1025 	if (IPSEC_ENABLED(ipv4)) {
1026 		if ((error = IPSEC_FORWARD(ipv4, m)) != 0) {
1027 			/* mbuf consumed by IPsec */
1028 			m_freem(mcopy);
1029 			if (error != EINPROGRESS)
1030 				IPSTAT_INC(ips_cantforward);
1031 			goto out;
1032 		}
1033 		/* No IPsec processing required */
1034 	}
1035 #endif /* IPSEC */
1036 	/*
1037 	 * If forwarding packet using same interface that it came in on,
1038 	 * perhaps should send a redirect to sender to shortcut a hop.
1039 	 * Only send redirect if source is sending directly to us,
1040 	 * and if packet was not source routed (or has any options).
1041 	 * Also, don't send redirect if forwarding using a default route
1042 	 * or a route modified by a redirect.
1043 	 */
1044 	dest.s_addr = 0;
1045 	if (!srcrt && V_ipsendredirects &&
1046 	    ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) {
1047 		struct rtentry *rt;
1048 
1049 		rt = ro.ro_rt;
1050 
1051 		if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1052 		    satosin(rt_key(rt))->sin_addr.s_addr != 0) {
1053 #define	RTA(rt)	((struct in_ifaddr *)(rt->rt_ifa))
1054 			u_long src = ntohl(ip->ip_src.s_addr);
1055 
1056 			if (RTA(rt) &&
1057 			    (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1058 				if (rt->rt_flags & RTF_GATEWAY)
1059 					dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr;
1060 				else
1061 					dest.s_addr = ip->ip_dst.s_addr;
1062 				/* Router requirements says to only send host redirects */
1063 				type = ICMP_REDIRECT;
1064 				code = ICMP_REDIRECT_HOST;
1065 			}
1066 		}
1067 	}
1068 
1069 	error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
1070 
1071 	if (error == EMSGSIZE && ro.ro_rt)
1072 		mtu = ro.ro_rt->rt_mtu;
1073 	RO_RTFREE(&ro);
1074 
1075 	if (error)
1076 		IPSTAT_INC(ips_cantforward);
1077 	else {
1078 		IPSTAT_INC(ips_forward);
1079 		if (type)
1080 			IPSTAT_INC(ips_redirectsent);
1081 		else {
1082 			if (mcopy)
1083 				m_freem(mcopy);
1084 			goto out;
1085 		}
1086 	}
1087 	if (mcopy == NULL)
1088 		goto out;
1089 
1090 
1091 	switch (error) {
1092 
1093 	case 0:				/* forwarded, but need redirect */
1094 		/* type, code set above */
1095 		break;
1096 
1097 	case ENETUNREACH:
1098 	case EHOSTUNREACH:
1099 	case ENETDOWN:
1100 	case EHOSTDOWN:
1101 	default:
1102 		type = ICMP_UNREACH;
1103 		code = ICMP_UNREACH_HOST;
1104 		break;
1105 
1106 	case EMSGSIZE:
1107 		type = ICMP_UNREACH;
1108 		code = ICMP_UNREACH_NEEDFRAG;
1109 		/*
1110 		 * If the MTU was set before make sure we are below the
1111 		 * interface MTU.
1112 		 * If the MTU wasn't set before use the interface mtu or
1113 		 * fall back to the next smaller mtu step compared to the
1114 		 * current packet size.
1115 		 */
1116 		if (mtu != 0) {
1117 			if (ia != NULL)
1118 				mtu = min(mtu, ia->ia_ifp->if_mtu);
1119 		} else {
1120 			if (ia != NULL)
1121 				mtu = ia->ia_ifp->if_mtu;
1122 			else
1123 				mtu = ip_next_mtu(ntohs(ip->ip_len), 0);
1124 		}
1125 		IPSTAT_INC(ips_cantfrag);
1126 		break;
1127 
1128 	case ENOBUFS:
1129 	case EACCES:			/* ipfw denied packet */
1130 		m_freem(mcopy);
1131 		goto out;
1132 	}
1133 	icmp_error(mcopy, type, code, dest.s_addr, mtu);
1134  out:
1135 	NET_EPOCH_EXIT();
1136 }
1137 
1138 #define	CHECK_SO_CT(sp, ct) \
1139     (((sp->so_options & SO_TIMESTAMP) && (sp->so_ts_clock == ct)) ? 1 : 0)
1140 
1141 void
1142 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1143     struct mbuf *m)
1144 {
1145 	bool stamped;
1146 
1147 	stamped = false;
1148 	if ((inp->inp_socket->so_options & SO_BINTIME) ||
1149 	    CHECK_SO_CT(inp->inp_socket, SO_TS_BINTIME)) {
1150 		struct bintime boottimebin, bt;
1151 		struct timespec ts1;
1152 
1153 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1154 		    M_TSTMP)) {
1155 			mbuf_tstmp2timespec(m, &ts1);
1156 			timespec2bintime(&ts1, &bt);
1157 			getboottimebin(&boottimebin);
1158 			bintime_add(&bt, &boottimebin);
1159 		} else {
1160 			bintime(&bt);
1161 		}
1162 		*mp = sbcreatecontrol((caddr_t)&bt, sizeof(bt),
1163 		    SCM_BINTIME, SOL_SOCKET);
1164 		if (*mp != NULL) {
1165 			mp = &(*mp)->m_next;
1166 			stamped = true;
1167 		}
1168 	}
1169 	if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME_MICRO)) {
1170 		struct bintime boottimebin, bt1;
1171 		struct timespec ts1;;
1172 		struct timeval tv;
1173 
1174 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1175 		    M_TSTMP)) {
1176 			mbuf_tstmp2timespec(m, &ts1);
1177 			timespec2bintime(&ts1, &bt1);
1178 			getboottimebin(&boottimebin);
1179 			bintime_add(&bt1, &boottimebin);
1180 			bintime2timeval(&bt1, &tv);
1181 		} else {
1182 			microtime(&tv);
1183 		}
1184 		*mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv),
1185 		    SCM_TIMESTAMP, SOL_SOCKET);
1186 		if (*mp != NULL) {
1187 			mp = &(*mp)->m_next;
1188 			stamped = true;
1189 		}
1190 	} else if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME)) {
1191 		struct bintime boottimebin;
1192 		struct timespec ts, ts1;
1193 
1194 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1195 		    M_TSTMP)) {
1196 			mbuf_tstmp2timespec(m, &ts);
1197 			getboottimebin(&boottimebin);
1198 			bintime2timespec(&boottimebin, &ts1);
1199 			timespecadd(&ts, &ts1, &ts);
1200 		} else {
1201 			nanotime(&ts);
1202 		}
1203 		*mp = sbcreatecontrol((caddr_t)&ts, sizeof(ts),
1204 		    SCM_REALTIME, SOL_SOCKET);
1205 		if (*mp != NULL) {
1206 			mp = &(*mp)->m_next;
1207 			stamped = true;
1208 		}
1209 	} else if (CHECK_SO_CT(inp->inp_socket, SO_TS_MONOTONIC)) {
1210 		struct timespec ts;
1211 
1212 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1213 		    M_TSTMP))
1214 			mbuf_tstmp2timespec(m, &ts);
1215 		else
1216 			nanouptime(&ts);
1217 		*mp = sbcreatecontrol((caddr_t)&ts, sizeof(ts),
1218 		    SCM_MONOTONIC, SOL_SOCKET);
1219 		if (*mp != NULL) {
1220 			mp = &(*mp)->m_next;
1221 			stamped = true;
1222 		}
1223 	}
1224 	if (stamped && (m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1225 	    M_TSTMP)) {
1226 		struct sock_timestamp_info sti;
1227 
1228 		bzero(&sti, sizeof(sti));
1229 		sti.st_info_flags = ST_INFO_HW;
1230 		if ((m->m_flags & M_TSTMP_HPREC) != 0)
1231 			sti.st_info_flags |= ST_INFO_HW_HPREC;
1232 		*mp = sbcreatecontrol((caddr_t)&sti, sizeof(sti), SCM_TIME_INFO,
1233 		    SOL_SOCKET);
1234 		if (*mp != NULL)
1235 			mp = &(*mp)->m_next;
1236 	}
1237 	if (inp->inp_flags & INP_RECVDSTADDR) {
1238 		*mp = sbcreatecontrol((caddr_t)&ip->ip_dst,
1239 		    sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1240 		if (*mp)
1241 			mp = &(*mp)->m_next;
1242 	}
1243 	if (inp->inp_flags & INP_RECVTTL) {
1244 		*mp = sbcreatecontrol((caddr_t)&ip->ip_ttl,
1245 		    sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
1246 		if (*mp)
1247 			mp = &(*mp)->m_next;
1248 	}
1249 #ifdef notyet
1250 	/* XXX
1251 	 * Moving these out of udp_input() made them even more broken
1252 	 * than they already were.
1253 	 */
1254 	/* options were tossed already */
1255 	if (inp->inp_flags & INP_RECVOPTS) {
1256 		*mp = sbcreatecontrol((caddr_t)opts_deleted_above,
1257 		    sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1258 		if (*mp)
1259 			mp = &(*mp)->m_next;
1260 	}
1261 	/* ip_srcroute doesn't do what we want here, need to fix */
1262 	if (inp->inp_flags & INP_RECVRETOPTS) {
1263 		*mp = sbcreatecontrol((caddr_t)ip_srcroute(m),
1264 		    sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1265 		if (*mp)
1266 			mp = &(*mp)->m_next;
1267 	}
1268 #endif
1269 	if (inp->inp_flags & INP_RECVIF) {
1270 		struct ifnet *ifp;
1271 		struct sdlbuf {
1272 			struct sockaddr_dl sdl;
1273 			u_char	pad[32];
1274 		} sdlbuf;
1275 		struct sockaddr_dl *sdp;
1276 		struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1277 
1278 		if ((ifp = m->m_pkthdr.rcvif) &&
1279 		    ifp->if_index && ifp->if_index <= V_if_index) {
1280 			sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1281 			/*
1282 			 * Change our mind and don't try copy.
1283 			 */
1284 			if (sdp->sdl_family != AF_LINK ||
1285 			    sdp->sdl_len > sizeof(sdlbuf)) {
1286 				goto makedummy;
1287 			}
1288 			bcopy(sdp, sdl2, sdp->sdl_len);
1289 		} else {
1290 makedummy:
1291 			sdl2->sdl_len =
1292 			    offsetof(struct sockaddr_dl, sdl_data[0]);
1293 			sdl2->sdl_family = AF_LINK;
1294 			sdl2->sdl_index = 0;
1295 			sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1296 		}
1297 		*mp = sbcreatecontrol((caddr_t)sdl2, sdl2->sdl_len,
1298 		    IP_RECVIF, IPPROTO_IP);
1299 		if (*mp)
1300 			mp = &(*mp)->m_next;
1301 	}
1302 	if (inp->inp_flags & INP_RECVTOS) {
1303 		*mp = sbcreatecontrol((caddr_t)&ip->ip_tos,
1304 		    sizeof(u_char), IP_RECVTOS, IPPROTO_IP);
1305 		if (*mp)
1306 			mp = &(*mp)->m_next;
1307 	}
1308 
1309 	if (inp->inp_flags2 & INP_RECVFLOWID) {
1310 		uint32_t flowid, flow_type;
1311 
1312 		flowid = m->m_pkthdr.flowid;
1313 		flow_type = M_HASHTYPE_GET(m);
1314 
1315 		/*
1316 		 * XXX should handle the failure of one or the
1317 		 * other - don't populate both?
1318 		 */
1319 		*mp = sbcreatecontrol((caddr_t) &flowid,
1320 		    sizeof(uint32_t), IP_FLOWID, IPPROTO_IP);
1321 		if (*mp)
1322 			mp = &(*mp)->m_next;
1323 		*mp = sbcreatecontrol((caddr_t) &flow_type,
1324 		    sizeof(uint32_t), IP_FLOWTYPE, IPPROTO_IP);
1325 		if (*mp)
1326 			mp = &(*mp)->m_next;
1327 	}
1328 
1329 #ifdef	RSS
1330 	if (inp->inp_flags2 & INP_RECVRSSBUCKETID) {
1331 		uint32_t flowid, flow_type;
1332 		uint32_t rss_bucketid;
1333 
1334 		flowid = m->m_pkthdr.flowid;
1335 		flow_type = M_HASHTYPE_GET(m);
1336 
1337 		if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) {
1338 			*mp = sbcreatecontrol((caddr_t) &rss_bucketid,
1339 			   sizeof(uint32_t), IP_RSSBUCKETID, IPPROTO_IP);
1340 			if (*mp)
1341 				mp = &(*mp)->m_next;
1342 		}
1343 	}
1344 #endif
1345 }
1346 
1347 /*
1348  * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
1349  * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
1350  * locking.  This code remains in ip_input.c as ip_mroute.c is optionally
1351  * compiled.
1352  */
1353 VNET_DEFINE_STATIC(int, ip_rsvp_on);
1354 VNET_DEFINE(struct socket *, ip_rsvpd);
1355 
1356 #define	V_ip_rsvp_on		VNET(ip_rsvp_on)
1357 
1358 int
1359 ip_rsvp_init(struct socket *so)
1360 {
1361 
1362 	if (so->so_type != SOCK_RAW ||
1363 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
1364 		return EOPNOTSUPP;
1365 
1366 	if (V_ip_rsvpd != NULL)
1367 		return EADDRINUSE;
1368 
1369 	V_ip_rsvpd = so;
1370 	/*
1371 	 * This may seem silly, but we need to be sure we don't over-increment
1372 	 * the RSVP counter, in case something slips up.
1373 	 */
1374 	if (!V_ip_rsvp_on) {
1375 		V_ip_rsvp_on = 1;
1376 		V_rsvp_on++;
1377 	}
1378 
1379 	return 0;
1380 }
1381 
1382 int
1383 ip_rsvp_done(void)
1384 {
1385 
1386 	V_ip_rsvpd = NULL;
1387 	/*
1388 	 * This may seem silly, but we need to be sure we don't over-decrement
1389 	 * the RSVP counter, in case something slips up.
1390 	 */
1391 	if (V_ip_rsvp_on) {
1392 		V_ip_rsvp_on = 0;
1393 		V_rsvp_on--;
1394 	}
1395 	return 0;
1396 }
1397 
1398 int
1399 rsvp_input(struct mbuf **mp, int *offp, int proto)
1400 {
1401 	struct mbuf *m;
1402 
1403 	m = *mp;
1404 	*mp = NULL;
1405 
1406 	if (rsvp_input_p) { /* call the real one if loaded */
1407 		*mp = m;
1408 		rsvp_input_p(mp, offp, proto);
1409 		return (IPPROTO_DONE);
1410 	}
1411 
1412 	/* Can still get packets with rsvp_on = 0 if there is a local member
1413 	 * of the group to which the RSVP packet is addressed.  But in this
1414 	 * case we want to throw the packet away.
1415 	 */
1416 
1417 	if (!V_rsvp_on) {
1418 		m_freem(m);
1419 		return (IPPROTO_DONE);
1420 	}
1421 
1422 	if (V_ip_rsvpd != NULL) {
1423 		*mp = m;
1424 		rip_input(mp, offp, proto);
1425 		return (IPPROTO_DONE);
1426 	}
1427 	/* Drop the packet */
1428 	m_freem(m);
1429 	return (IPPROTO_DONE);
1430 }
1431