xref: /freebsd/sys/netinet/ip_input.c (revision 3cbb4cc200f8a0ad7ed08233425ea54524a21f1c)
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 struct rmlock in_ifaddr_lock;
105 RM_SYSINIT(in_ifaddr_lock, &in_ifaddr_lock, "in_ifaddr_lock");
106 
107 VNET_DEFINE(int, rsvp_on);
108 
109 VNET_DEFINE(int, ipforwarding);
110 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_VNET | CTLFLAG_RW,
111     &VNET_NAME(ipforwarding), 0,
112     "Enable IP forwarding between interfaces");
113 
114 /*
115  * Respond with an ICMP host redirect when we forward a packet out of
116  * the same interface on which it was received.  See RFC 792.
117  */
118 VNET_DEFINE(int, ipsendredirects) = 1;
119 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_VNET | CTLFLAG_RW,
120     &VNET_NAME(ipsendredirects), 0,
121     "Enable sending IP redirects");
122 
123 /*
124  * XXX - Setting ip_checkinterface mostly implements the receive side of
125  * the Strong ES model described in RFC 1122, but since the routing table
126  * and transmit implementation do not implement the Strong ES model,
127  * setting this to 1 results in an odd hybrid.
128  *
129  * XXX - ip_checkinterface currently must be disabled if you use ipnat
130  * to translate the destination address to another local interface.
131  *
132  * XXX - ip_checkinterface must be disabled if you add IP aliases
133  * to the loopback interface instead of the interface where the
134  * packets for those addresses are received.
135  */
136 VNET_DEFINE_STATIC(int, ip_checkinterface);
137 #define	V_ip_checkinterface	VNET(ip_checkinterface)
138 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_VNET | CTLFLAG_RW,
139     &VNET_NAME(ip_checkinterface), 0,
140     "Verify packet arrives on correct interface");
141 
142 VNET_DEFINE(pfil_head_t, inet_pfil_head);	/* Packet filter hooks */
143 
144 static struct netisr_handler ip_nh = {
145 	.nh_name = "ip",
146 	.nh_handler = ip_input,
147 	.nh_proto = NETISR_IP,
148 #ifdef	RSS
149 	.nh_m2cpuid = rss_soft_m2cpuid_v4,
150 	.nh_policy = NETISR_POLICY_CPU,
151 	.nh_dispatch = NETISR_DISPATCH_HYBRID,
152 #else
153 	.nh_policy = NETISR_POLICY_FLOW,
154 #endif
155 };
156 
157 #ifdef	RSS
158 /*
159  * Directly dispatched frames are currently assumed
160  * to have a flowid already calculated.
161  *
162  * It should likely have something that assert it
163  * actually has valid flow details.
164  */
165 static struct netisr_handler ip_direct_nh = {
166 	.nh_name = "ip_direct",
167 	.nh_handler = ip_direct_input,
168 	.nh_proto = NETISR_IP_DIRECT,
169 	.nh_m2cpuid = rss_soft_m2cpuid_v4,
170 	.nh_policy = NETISR_POLICY_CPU,
171 	.nh_dispatch = NETISR_DISPATCH_HYBRID,
172 };
173 #endif
174 
175 extern	struct domain inetdomain;
176 extern	struct protosw inetsw[];
177 u_char	ip_protox[IPPROTO_MAX];
178 VNET_DEFINE(struct in_ifaddrhead, in_ifaddrhead);  /* first inet address */
179 VNET_DEFINE(struct in_ifaddrhashhead *, in_ifaddrhashtbl); /* inet addr hash table  */
180 VNET_DEFINE(u_long, in_ifaddrhmask);		/* mask for hash table */
181 
182 #ifdef IPCTL_DEFMTU
183 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
184     &ip_mtu, 0, "Default MTU");
185 #endif
186 
187 #ifdef IPSTEALTH
188 VNET_DEFINE(int, ipstealth);
189 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_VNET | CTLFLAG_RW,
190     &VNET_NAME(ipstealth), 0,
191     "IP stealth mode, no TTL decrementation on forwarding");
192 #endif
193 
194 /*
195  * IP statistics are stored in the "array" of counter(9)s.
196  */
197 VNET_PCPUSTAT_DEFINE(struct ipstat, ipstat);
198 VNET_PCPUSTAT_SYSINIT(ipstat);
199 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, IPCTL_STATS, stats, struct ipstat, ipstat,
200     "IP statistics (struct ipstat, netinet/ip_var.h)");
201 
202 #ifdef VIMAGE
203 VNET_PCPUSTAT_SYSUNINIT(ipstat);
204 #endif /* VIMAGE */
205 
206 /*
207  * Kernel module interface for updating ipstat.  The argument is an index
208  * into ipstat treated as an array.
209  */
210 void
211 kmod_ipstat_inc(int statnum)
212 {
213 
214 	counter_u64_add(VNET(ipstat)[statnum], 1);
215 }
216 
217 void
218 kmod_ipstat_dec(int statnum)
219 {
220 
221 	counter_u64_add(VNET(ipstat)[statnum], -1);
222 }
223 
224 static int
225 sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS)
226 {
227 	int error, qlimit;
228 
229 	netisr_getqlimit(&ip_nh, &qlimit);
230 	error = sysctl_handle_int(oidp, &qlimit, 0, req);
231 	if (error || !req->newptr)
232 		return (error);
233 	if (qlimit < 1)
234 		return (EINVAL);
235 	return (netisr_setqlimit(&ip_nh, qlimit));
236 }
237 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen,
238     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 0,
239     sysctl_netinet_intr_queue_maxlen, "I",
240     "Maximum size of the IP input queue");
241 
242 static int
243 sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS)
244 {
245 	u_int64_t qdrops_long;
246 	int error, qdrops;
247 
248 	netisr_getqdrops(&ip_nh, &qdrops_long);
249 	qdrops = qdrops_long;
250 	error = sysctl_handle_int(oidp, &qdrops, 0, req);
251 	if (error || !req->newptr)
252 		return (error);
253 	if (qdrops != 0)
254 		return (EINVAL);
255 	netisr_clearqdrops(&ip_nh);
256 	return (0);
257 }
258 
259 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops,
260     CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE,
261     0, 0, sysctl_netinet_intr_queue_drops, "I",
262     "Number of packets dropped from the IP input queue");
263 
264 #ifdef	RSS
265 static int
266 sysctl_netinet_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS)
267 {
268 	int error, qlimit;
269 
270 	netisr_getqlimit(&ip_direct_nh, &qlimit);
271 	error = sysctl_handle_int(oidp, &qlimit, 0, req);
272 	if (error || !req->newptr)
273 		return (error);
274 	if (qlimit < 1)
275 		return (EINVAL);
276 	return (netisr_setqlimit(&ip_direct_nh, qlimit));
277 }
278 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRDQMAXLEN, intr_direct_queue_maxlen,
279     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
280     0, 0, sysctl_netinet_intr_direct_queue_maxlen,
281     "I", "Maximum size of the IP direct input queue");
282 
283 static int
284 sysctl_netinet_intr_direct_queue_drops(SYSCTL_HANDLER_ARGS)
285 {
286 	u_int64_t qdrops_long;
287 	int error, qdrops;
288 
289 	netisr_getqdrops(&ip_direct_nh, &qdrops_long);
290 	qdrops = qdrops_long;
291 	error = sysctl_handle_int(oidp, &qdrops, 0, req);
292 	if (error || !req->newptr)
293 		return (error);
294 	if (qdrops != 0)
295 		return (EINVAL);
296 	netisr_clearqdrops(&ip_direct_nh);
297 	return (0);
298 }
299 
300 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRDQDROPS, intr_direct_queue_drops,
301     CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, 0, 0,
302     sysctl_netinet_intr_direct_queue_drops, "I",
303     "Number of packets dropped from the IP direct input queue");
304 #endif	/* RSS */
305 
306 /*
307  * IP initialization: fill in IP protocol switch table.
308  * All protocols not implemented in kernel go to raw IP protocol handler.
309  */
310 void
311 ip_init(void)
312 {
313 	struct pfil_head_args args;
314 	struct protosw *pr;
315 	int i;
316 
317 	CK_STAILQ_INIT(&V_in_ifaddrhead);
318 	V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask);
319 
320 	/* Initialize IP reassembly queue. */
321 	ipreass_init();
322 
323 	/* Initialize packet filter hooks. */
324 	args.pa_version = PFIL_VERSION;
325 	args.pa_flags = PFIL_IN | PFIL_OUT;
326 	args.pa_type = PFIL_TYPE_IP4;
327 	args.pa_headname = PFIL_INET_NAME;
328 	V_inet_pfil_head = pfil_head_register(&args);
329 
330 	if (hhook_head_register(HHOOK_TYPE_IPSEC_IN, AF_INET,
331 	    &V_ipsec_hhh_in[HHOOK_IPSEC_INET],
332 	    HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
333 		printf("%s: WARNING: unable to register input helper hook\n",
334 		    __func__);
335 	if (hhook_head_register(HHOOK_TYPE_IPSEC_OUT, AF_INET,
336 	    &V_ipsec_hhh_out[HHOOK_IPSEC_INET],
337 	    HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
338 		printf("%s: WARNING: unable to register output helper hook\n",
339 		    __func__);
340 
341 	/* Skip initialization of globals for non-default instances. */
342 #ifdef VIMAGE
343 	if (!IS_DEFAULT_VNET(curvnet)) {
344 		netisr_register_vnet(&ip_nh);
345 #ifdef	RSS
346 		netisr_register_vnet(&ip_direct_nh);
347 #endif
348 		return;
349 	}
350 #endif
351 
352 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
353 	if (pr == NULL)
354 		panic("ip_init: PF_INET not found");
355 
356 	/* Initialize the entire ip_protox[] array to IPPROTO_RAW. */
357 	for (i = 0; i < IPPROTO_MAX; i++)
358 		ip_protox[i] = pr - inetsw;
359 	/*
360 	 * Cycle through IP protocols and put them into the appropriate place
361 	 * in ip_protox[].
362 	 */
363 	for (pr = inetdomain.dom_protosw;
364 	    pr < inetdomain.dom_protoswNPROTOSW; pr++)
365 		if (pr->pr_domain->dom_family == PF_INET &&
366 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
367 			/* Be careful to only index valid IP protocols. */
368 			if (pr->pr_protocol < IPPROTO_MAX)
369 				ip_protox[pr->pr_protocol] = pr - inetsw;
370 		}
371 
372 	netisr_register(&ip_nh);
373 #ifdef	RSS
374 	netisr_register(&ip_direct_nh);
375 #endif
376 }
377 
378 #ifdef VIMAGE
379 static void
380 ip_destroy(void *unused __unused)
381 {
382 	struct ifnet *ifp;
383 	int error;
384 
385 #ifdef	RSS
386 	netisr_unregister_vnet(&ip_direct_nh);
387 #endif
388 	netisr_unregister_vnet(&ip_nh);
389 
390 	pfil_head_unregister(V_inet_pfil_head);
391 	error = hhook_head_deregister(V_ipsec_hhh_in[HHOOK_IPSEC_INET]);
392 	if (error != 0) {
393 		printf("%s: WARNING: unable to deregister input helper hook "
394 		    "type HHOOK_TYPE_IPSEC_IN, id HHOOK_IPSEC_INET: "
395 		    "error %d returned\n", __func__, error);
396 	}
397 	error = hhook_head_deregister(V_ipsec_hhh_out[HHOOK_IPSEC_INET]);
398 	if (error != 0) {
399 		printf("%s: WARNING: unable to deregister output helper hook "
400 		    "type HHOOK_TYPE_IPSEC_OUT, id HHOOK_IPSEC_INET: "
401 		    "error %d returned\n", __func__, error);
402 	}
403 
404 	/* Remove the IPv4 addresses from all interfaces. */
405 	in_ifscrub_all();
406 
407 	/* Make sure the IPv4 routes are gone as well. */
408 	IFNET_RLOCK();
409 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link)
410 		rt_flushifroutes_af(ifp, AF_INET);
411 	IFNET_RUNLOCK();
412 
413 	/* Destroy IP reassembly queue. */
414 	ipreass_destroy();
415 
416 	/* Cleanup in_ifaddr hash table; should be empty. */
417 	hashdestroy(V_in_ifaddrhashtbl, M_IFADDR, V_in_ifaddrhmask);
418 }
419 
420 VNET_SYSUNINIT(ip, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip_destroy, NULL);
421 #endif
422 
423 #ifdef	RSS
424 /*
425  * IP direct input routine.
426  *
427  * This is called when reinjecting completed fragments where
428  * all of the previous checking and book-keeping has been done.
429  */
430 void
431 ip_direct_input(struct mbuf *m)
432 {
433 	struct ip *ip;
434 	int hlen;
435 
436 	ip = mtod(m, struct ip *);
437 	hlen = ip->ip_hl << 2;
438 
439 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
440 	if (IPSEC_ENABLED(ipv4)) {
441 		if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
442 			return;
443 	}
444 #endif /* IPSEC */
445 	IPSTAT_INC(ips_delivered);
446 	(*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
447 	return;
448 }
449 #endif
450 
451 /*
452  * Ip input routine.  Checksum and byte swap header.  If fragmented
453  * try to reassemble.  Process options.  Pass to next level.
454  */
455 void
456 ip_input(struct mbuf *m)
457 {
458 	struct rm_priotracker in_ifa_tracker;
459 	struct ip *ip = NULL;
460 	struct in_ifaddr *ia = NULL;
461 	struct ifaddr *ifa;
462 	struct ifnet *ifp;
463 	int    checkif, hlen = 0;
464 	uint16_t sum, ip_len;
465 	int dchg = 0;				/* dest changed after fw */
466 	struct in_addr odst;			/* original dst address */
467 
468 	M_ASSERTPKTHDR(m);
469 	NET_EPOCH_ASSERT();
470 
471 	if (m->m_flags & M_FASTFWD_OURS) {
472 		m->m_flags &= ~M_FASTFWD_OURS;
473 		/* Set up some basics that will be used later. */
474 		ip = mtod(m, struct ip *);
475 		hlen = ip->ip_hl << 2;
476 		ip_len = ntohs(ip->ip_len);
477 		goto ours;
478 	}
479 
480 	IPSTAT_INC(ips_total);
481 
482 	if (m->m_pkthdr.len < sizeof(struct ip))
483 		goto tooshort;
484 
485 	if (m->m_len < sizeof (struct ip) &&
486 	    (m = m_pullup(m, sizeof (struct ip))) == NULL) {
487 		IPSTAT_INC(ips_toosmall);
488 		return;
489 	}
490 	ip = mtod(m, struct ip *);
491 
492 	if (ip->ip_v != IPVERSION) {
493 		IPSTAT_INC(ips_badvers);
494 		goto bad;
495 	}
496 
497 	hlen = ip->ip_hl << 2;
498 	if (hlen < sizeof(struct ip)) {	/* minimum header length */
499 		IPSTAT_INC(ips_badhlen);
500 		goto bad;
501 	}
502 	if (hlen > m->m_len) {
503 		if ((m = m_pullup(m, hlen)) == NULL) {
504 			IPSTAT_INC(ips_badhlen);
505 			return;
506 		}
507 		ip = mtod(m, struct ip *);
508 	}
509 
510 	IP_PROBE(receive, NULL, NULL, ip, m->m_pkthdr.rcvif, ip, NULL);
511 
512 	/* IN_LOOPBACK must not appear on the wire - RFC1122 */
513 	ifp = m->m_pkthdr.rcvif;
514 	if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) ||
515 	    IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) {
516 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
517 			IPSTAT_INC(ips_badaddr);
518 			goto bad;
519 		}
520 	}
521 
522 	if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
523 		sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
524 	} else {
525 		if (hlen == sizeof(struct ip)) {
526 			sum = in_cksum_hdr(ip);
527 		} else {
528 			sum = in_cksum(m, hlen);
529 		}
530 	}
531 	if (sum) {
532 		IPSTAT_INC(ips_badsum);
533 		goto bad;
534 	}
535 
536 #ifdef ALTQ
537 	if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
538 		/* packet is dropped by traffic conditioner */
539 		return;
540 #endif
541 
542 	ip_len = ntohs(ip->ip_len);
543 	if (ip_len < hlen) {
544 		IPSTAT_INC(ips_badlen);
545 		goto bad;
546 	}
547 
548 	/*
549 	 * Check that the amount of data in the buffers
550 	 * is as at least much as the IP header would have us expect.
551 	 * Trim mbufs if longer than we expect.
552 	 * Drop packet if shorter than we expect.
553 	 */
554 	if (m->m_pkthdr.len < ip_len) {
555 tooshort:
556 		IPSTAT_INC(ips_tooshort);
557 		goto bad;
558 	}
559 	if (m->m_pkthdr.len > ip_len) {
560 		if (m->m_len == m->m_pkthdr.len) {
561 			m->m_len = ip_len;
562 			m->m_pkthdr.len = ip_len;
563 		} else
564 			m_adj(m, ip_len - m->m_pkthdr.len);
565 	}
566 
567 	/*
568 	 * Try to forward the packet, but if we fail continue.
569 	 * ip_tryforward() does not generate redirects, so fall
570 	 * through to normal processing if redirects are required.
571 	 * ip_tryforward() does inbound and outbound packet firewall
572 	 * processing. If firewall has decided that destination becomes
573 	 * our local address, it sets M_FASTFWD_OURS flag. In this
574 	 * case skip another inbound firewall processing and update
575 	 * ip pointer.
576 	 */
577 	if (V_ipforwarding != 0
578 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
579 	    && (!IPSEC_ENABLED(ipv4) ||
580 	    IPSEC_CAPS(ipv4, m, IPSEC_CAP_OPERABLE) == 0)
581 #endif
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 	ifp = m->m_pkthdr.rcvif;
623 
624 	if (m->m_flags & M_FASTFWD_OURS) {
625 		m->m_flags &= ~M_FASTFWD_OURS;
626 		goto ours;
627 	}
628 	if (m->m_flags & M_IP_NEXTHOP) {
629 		if (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) {
630 			/*
631 			 * Directly ship the packet on.  This allows
632 			 * forwarding packets originally destined to us
633 			 * to some other directly connected host.
634 			 */
635 			ip_forward(m, 1);
636 			return;
637 		}
638 	}
639 passin:
640 
641 	/*
642 	 * Process options and, if not destined for us,
643 	 * ship it on.  ip_dooptions returns 1 when an
644 	 * error was detected (causing an icmp message
645 	 * to be sent and the original packet to be freed).
646 	 */
647 	if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
648 		return;
649 
650         /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
651          * matter if it is destined to another node, or whether it is
652          * a multicast one, RSVP wants it! and prevents it from being forwarded
653          * anywhere else. Also checks if the rsvp daemon is running before
654 	 * grabbing the packet.
655          */
656 	if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP)
657 		goto ours;
658 
659 	/*
660 	 * Check our list of addresses, to see if the packet is for us.
661 	 * If we don't have any addresses, assume any unicast packet
662 	 * we receive might be for us (and let the upper layers deal
663 	 * with it).
664 	 */
665 	if (CK_STAILQ_EMPTY(&V_in_ifaddrhead) &&
666 	    (m->m_flags & (M_MCAST|M_BCAST)) == 0)
667 		goto ours;
668 
669 	/*
670 	 * Enable a consistency check between the destination address
671 	 * and the arrival interface for a unicast packet (the RFC 1122
672 	 * strong ES model) if IP forwarding is disabled and the packet
673 	 * is not locally generated and the packet is not subject to
674 	 * 'ipfw fwd'.
675 	 *
676 	 * XXX - Checking also should be disabled if the destination
677 	 * address is ipnat'ed to a different interface.
678 	 *
679 	 * XXX - Checking is incompatible with IP aliases added
680 	 * to the loopback interface instead of the interface where
681 	 * the packets are received.
682 	 *
683 	 * XXX - This is the case for carp vhost IPs as well so we
684 	 * insert a workaround. If the packet got here, we already
685 	 * checked with carp_iamatch() and carp_forus().
686 	 */
687 	checkif = V_ip_checkinterface && (V_ipforwarding == 0) &&
688 	    ifp != NULL && ((ifp->if_flags & IFF_LOOPBACK) == 0) &&
689 	    ifp->if_carp == NULL && (dchg == 0);
690 
691 	/*
692 	 * Check for exact addresses in the hash bucket.
693 	 */
694 	IN_IFADDR_RLOCK(&in_ifa_tracker);
695 	LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
696 		/*
697 		 * If the address matches, verify that the packet
698 		 * arrived via the correct interface if checking is
699 		 * enabled.
700 		 */
701 		if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr &&
702 		    (!checkif || ia->ia_ifp == ifp)) {
703 			counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
704 			counter_u64_add(ia->ia_ifa.ifa_ibytes,
705 			    m->m_pkthdr.len);
706 			IN_IFADDR_RUNLOCK(&in_ifa_tracker);
707 			goto ours;
708 		}
709 	}
710 	IN_IFADDR_RUNLOCK(&in_ifa_tracker);
711 
712 	/*
713 	 * Check for broadcast addresses.
714 	 *
715 	 * Only accept broadcast packets that arrive via the matching
716 	 * interface.  Reception of forwarded directed broadcasts would
717 	 * be handled via ip_forward() and ether_output() with the loopback
718 	 * into the stack for SIMPLEX interfaces handled by ether_output().
719 	 */
720 	if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
721 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
722 			if (ifa->ifa_addr->sa_family != AF_INET)
723 				continue;
724 			ia = ifatoia(ifa);
725 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
726 			    ip->ip_dst.s_addr) {
727 				counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
728 				counter_u64_add(ia->ia_ifa.ifa_ibytes,
729 				    m->m_pkthdr.len);
730 				goto ours;
731 			}
732 #ifdef BOOTP_COMPAT
733 			if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) {
734 				counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
735 				counter_u64_add(ia->ia_ifa.ifa_ibytes,
736 				    m->m_pkthdr.len);
737 				goto ours;
738 			}
739 #endif
740 		}
741 		ia = NULL;
742 	}
743 	/* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */
744 	if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) {
745 		IPSTAT_INC(ips_cantforward);
746 		m_freem(m);
747 		return;
748 	}
749 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
750 		if (V_ip_mrouter) {
751 			/*
752 			 * If we are acting as a multicast router, all
753 			 * incoming multicast packets are passed to the
754 			 * kernel-level multicast forwarding function.
755 			 * The packet is returned (relatively) intact; if
756 			 * ip_mforward() returns a non-zero value, the packet
757 			 * must be discarded, else it may be accepted below.
758 			 */
759 			if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) {
760 				IPSTAT_INC(ips_cantforward);
761 				m_freem(m);
762 				return;
763 			}
764 
765 			/*
766 			 * The process-level routing daemon needs to receive
767 			 * all multicast IGMP packets, whether or not this
768 			 * host belongs to their destination groups.
769 			 */
770 			if (ip->ip_p == IPPROTO_IGMP)
771 				goto ours;
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 
786 	/*
787 	 * Not for us; forward if possible and desirable.
788 	 */
789 	if (V_ipforwarding == 0) {
790 		IPSTAT_INC(ips_cantforward);
791 		m_freem(m);
792 	} else {
793 		ip_forward(m, dchg);
794 	}
795 	return;
796 
797 ours:
798 #ifdef IPSTEALTH
799 	/*
800 	 * IPSTEALTH: Process non-routing options only
801 	 * if the packet is destined for us.
802 	 */
803 	if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1))
804 		return;
805 #endif /* IPSTEALTH */
806 
807 	/*
808 	 * Attempt reassembly; if it succeeds, proceed.
809 	 * ip_reass() will return a different mbuf.
810 	 */
811 	if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) {
812 		/* XXXGL: shouldn't we save & set m_flags? */
813 		m = ip_reass(m);
814 		if (m == NULL)
815 			return;
816 		ip = mtod(m, struct ip *);
817 		/* Get the header length of the reassembled packet */
818 		hlen = ip->ip_hl << 2;
819 	}
820 
821 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
822 	if (IPSEC_ENABLED(ipv4)) {
823 		if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
824 			return;
825 	}
826 #endif /* IPSEC */
827 
828 	/*
829 	 * Switch out to protocol's input routine.
830 	 */
831 	IPSTAT_INC(ips_delivered);
832 
833 	(*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
834 	return;
835 bad:
836 	m_freem(m);
837 }
838 
839 /*
840  * IP timer processing;
841  * if a timer expires on a reassembly
842  * queue, discard it.
843  */
844 void
845 ip_slowtimo(void)
846 {
847 	VNET_ITERATOR_DECL(vnet_iter);
848 
849 	VNET_LIST_RLOCK_NOSLEEP();
850 	VNET_FOREACH(vnet_iter) {
851 		CURVNET_SET(vnet_iter);
852 		ipreass_slowtimo();
853 		CURVNET_RESTORE();
854 	}
855 	VNET_LIST_RUNLOCK_NOSLEEP();
856 }
857 
858 void
859 ip_drain(void)
860 {
861 	VNET_ITERATOR_DECL(vnet_iter);
862 
863 	VNET_LIST_RLOCK_NOSLEEP();
864 	VNET_FOREACH(vnet_iter) {
865 		CURVNET_SET(vnet_iter);
866 		ipreass_drain();
867 		CURVNET_RESTORE();
868 	}
869 	VNET_LIST_RUNLOCK_NOSLEEP();
870 }
871 
872 /*
873  * The protocol to be inserted into ip_protox[] must be already registered
874  * in inetsw[], either statically or through pf_proto_register().
875  */
876 int
877 ipproto_register(short ipproto)
878 {
879 	struct protosw *pr;
880 
881 	/* Sanity checks. */
882 	if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
883 		return (EPROTONOSUPPORT);
884 
885 	/*
886 	 * The protocol slot must not be occupied by another protocol
887 	 * already.  An index pointing to IPPROTO_RAW is unused.
888 	 */
889 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
890 	if (pr == NULL)
891 		return (EPFNOSUPPORT);
892 	if (ip_protox[ipproto] != pr - inetsw)	/* IPPROTO_RAW */
893 		return (EEXIST);
894 
895 	/* Find the protocol position in inetsw[] and set the index. */
896 	for (pr = inetdomain.dom_protosw;
897 	     pr < inetdomain.dom_protoswNPROTOSW; pr++) {
898 		if (pr->pr_domain->dom_family == PF_INET &&
899 		    pr->pr_protocol && pr->pr_protocol == ipproto) {
900 			ip_protox[pr->pr_protocol] = pr - inetsw;
901 			return (0);
902 		}
903 	}
904 	return (EPROTONOSUPPORT);
905 }
906 
907 int
908 ipproto_unregister(short ipproto)
909 {
910 	struct protosw *pr;
911 
912 	/* Sanity checks. */
913 	if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
914 		return (EPROTONOSUPPORT);
915 
916 	/* Check if the protocol was indeed registered. */
917 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
918 	if (pr == NULL)
919 		return (EPFNOSUPPORT);
920 	if (ip_protox[ipproto] == pr - inetsw)  /* IPPROTO_RAW */
921 		return (ENOENT);
922 
923 	/* Reset the protocol slot to IPPROTO_RAW. */
924 	ip_protox[ipproto] = pr - inetsw;
925 	return (0);
926 }
927 
928 u_char inetctlerrmap[PRC_NCMDS] = {
929 	0,		0,		0,		0,
930 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
931 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
932 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
933 	0,		0,		EHOSTUNREACH,	0,
934 	ENOPROTOOPT,	ECONNREFUSED
935 };
936 
937 /*
938  * Forward a packet.  If some error occurs return the sender
939  * an icmp packet.  Note we can't always generate a meaningful
940  * icmp message because icmp doesn't have a large enough repertoire
941  * of codes and types.
942  *
943  * If not forwarding, just drop the packet.  This could be confusing
944  * if ipforwarding was zero but some routing protocol was advancing
945  * us as a gateway to somewhere.  However, we must let the routing
946  * protocol deal with that.
947  *
948  * The srcrt parameter indicates whether the packet is being forwarded
949  * via a source route.
950  */
951 void
952 ip_forward(struct mbuf *m, int srcrt)
953 {
954 	struct ip *ip = mtod(m, struct ip *);
955 	struct in_ifaddr *ia;
956 	struct mbuf *mcopy;
957 	struct sockaddr_in *sin;
958 	struct in_addr dest;
959 	struct route ro;
960 	uint32_t flowid;
961 	int error, type = 0, code = 0, mtu = 0;
962 
963 	NET_EPOCH_ASSERT();
964 
965 	if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
966 		IPSTAT_INC(ips_cantforward);
967 		m_freem(m);
968 		return;
969 	}
970 	if (
971 #ifdef IPSTEALTH
972 	    V_ipstealth == 0 &&
973 #endif
974 	    ip->ip_ttl <= IPTTLDEC) {
975 		icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 0, 0);
976 		return;
977 	}
978 
979 	bzero(&ro, sizeof(ro));
980 	sin = (struct sockaddr_in *)&ro.ro_dst;
981 	sin->sin_family = AF_INET;
982 	sin->sin_len = sizeof(*sin);
983 	sin->sin_addr = ip->ip_dst;
984 #ifdef RADIX_MPATH
985 	flowid = ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr);
986 #else
987 	flowid = m->m_pkthdr.flowid;
988 #endif
989 	ro.ro_nh = fib4_lookup(M_GETFIB(m), ip->ip_dst, 0, NHR_REF, flowid);
990 	if (ro.ro_nh != NULL) {
991 		ia = ifatoia(ro.ro_nh->nh_ifa);
992 	} else
993 		ia = NULL;
994 	/*
995 	 * Save the IP header and at most 8 bytes of the payload,
996 	 * in case we need to generate an ICMP message to the src.
997 	 *
998 	 * XXX this can be optimized a lot by saving the data in a local
999 	 * buffer on the stack (72 bytes at most), and only allocating the
1000 	 * mbuf if really necessary. The vast majority of the packets
1001 	 * are forwarded without having to send an ICMP back (either
1002 	 * because unnecessary, or because rate limited), so we are
1003 	 * really we are wasting a lot of work here.
1004 	 *
1005 	 * We don't use m_copym() because it might return a reference
1006 	 * to a shared cluster. Both this function and ip_output()
1007 	 * assume exclusive access to the IP header in `m', so any
1008 	 * data in a cluster may change before we reach icmp_error().
1009 	 */
1010 	mcopy = m_gethdr(M_NOWAIT, m->m_type);
1011 	if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_NOWAIT)) {
1012 		/*
1013 		 * It's probably ok if the pkthdr dup fails (because
1014 		 * the deep copy of the tag chain failed), but for now
1015 		 * be conservative and just discard the copy since
1016 		 * code below may some day want the tags.
1017 		 */
1018 		m_free(mcopy);
1019 		mcopy = NULL;
1020 	}
1021 	if (mcopy != NULL) {
1022 		mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy));
1023 		mcopy->m_pkthdr.len = mcopy->m_len;
1024 		m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1025 	}
1026 #ifdef IPSTEALTH
1027 	if (V_ipstealth == 0)
1028 #endif
1029 		ip->ip_ttl -= IPTTLDEC;
1030 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1031 	if (IPSEC_ENABLED(ipv4)) {
1032 		if ((error = IPSEC_FORWARD(ipv4, m)) != 0) {
1033 			/* mbuf consumed by IPsec */
1034 			RO_NHFREE(&ro);
1035 			m_freem(mcopy);
1036 			if (error != EINPROGRESS)
1037 				IPSTAT_INC(ips_cantforward);
1038 			return;
1039 		}
1040 		/* No IPsec processing required */
1041 	}
1042 #endif /* IPSEC */
1043 	/*
1044 	 * If forwarding packet using same interface that it came in on,
1045 	 * perhaps should send a redirect to sender to shortcut a hop.
1046 	 * Only send redirect if source is sending directly to us,
1047 	 * and if packet was not source routed (or has any options).
1048 	 * Also, don't send redirect if forwarding using a default route
1049 	 * or a route modified by a redirect.
1050 	 */
1051 	dest.s_addr = 0;
1052 	if (!srcrt && V_ipsendredirects &&
1053 	    ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) {
1054 		struct nhop_object *nh;
1055 
1056 		nh = ro.ro_nh;
1057 
1058 		if (nh != NULL && ((nh->nh_flags & (NHF_REDIRECT|NHF_DEFAULT)) == 0)) {
1059 			struct in_ifaddr *nh_ia = (struct in_ifaddr *)(nh->nh_ifa);
1060 			u_long src = ntohl(ip->ip_src.s_addr);
1061 
1062 			if (nh_ia != NULL &&
1063 			    (src & nh_ia->ia_subnetmask) == nh_ia->ia_subnet) {
1064 				if (nh->nh_flags & NHF_GATEWAY)
1065 					dest.s_addr = nh->gw4_sa.sin_addr.s_addr;
1066 				else
1067 					dest.s_addr = ip->ip_dst.s_addr;
1068 				/* Router requirements says to only send host redirects */
1069 				type = ICMP_REDIRECT;
1070 				code = ICMP_REDIRECT_HOST;
1071 			}
1072 		}
1073 	}
1074 
1075 	error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
1076 
1077 	if (error == EMSGSIZE && ro.ro_nh)
1078 		mtu = ro.ro_nh->nh_mtu;
1079 	RO_NHFREE(&ro);
1080 
1081 	if (error)
1082 		IPSTAT_INC(ips_cantforward);
1083 	else {
1084 		IPSTAT_INC(ips_forward);
1085 		if (type)
1086 			IPSTAT_INC(ips_redirectsent);
1087 		else {
1088 			if (mcopy)
1089 				m_freem(mcopy);
1090 			return;
1091 		}
1092 	}
1093 	if (mcopy == NULL)
1094 		return;
1095 
1096 	switch (error) {
1097 	case 0:				/* forwarded, but need redirect */
1098 		/* type, code set above */
1099 		break;
1100 
1101 	case ENETUNREACH:
1102 	case EHOSTUNREACH:
1103 	case ENETDOWN:
1104 	case EHOSTDOWN:
1105 	default:
1106 		type = ICMP_UNREACH;
1107 		code = ICMP_UNREACH_HOST;
1108 		break;
1109 
1110 	case EMSGSIZE:
1111 		type = ICMP_UNREACH;
1112 		code = ICMP_UNREACH_NEEDFRAG;
1113 		/*
1114 		 * If the MTU was set before make sure we are below the
1115 		 * interface MTU.
1116 		 * If the MTU wasn't set before use the interface mtu or
1117 		 * fall back to the next smaller mtu step compared to the
1118 		 * current packet size.
1119 		 */
1120 		if (mtu != 0) {
1121 			if (ia != NULL)
1122 				mtu = min(mtu, ia->ia_ifp->if_mtu);
1123 		} else {
1124 			if (ia != NULL)
1125 				mtu = ia->ia_ifp->if_mtu;
1126 			else
1127 				mtu = ip_next_mtu(ntohs(ip->ip_len), 0);
1128 		}
1129 		IPSTAT_INC(ips_cantfrag);
1130 		break;
1131 
1132 	case ENOBUFS:
1133 	case EACCES:			/* ipfw denied packet */
1134 		m_freem(mcopy);
1135 		return;
1136 	}
1137 	icmp_error(mcopy, type, code, dest.s_addr, mtu);
1138 }
1139 
1140 #define	CHECK_SO_CT(sp, ct) \
1141     (((sp->so_options & SO_TIMESTAMP) && (sp->so_ts_clock == ct)) ? 1 : 0)
1142 
1143 void
1144 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1145     struct mbuf *m)
1146 {
1147 	bool stamped;
1148 
1149 	stamped = false;
1150 	if ((inp->inp_socket->so_options & SO_BINTIME) ||
1151 	    CHECK_SO_CT(inp->inp_socket, SO_TS_BINTIME)) {
1152 		struct bintime boottimebin, bt;
1153 		struct timespec ts1;
1154 
1155 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1156 		    M_TSTMP)) {
1157 			mbuf_tstmp2timespec(m, &ts1);
1158 			timespec2bintime(&ts1, &bt);
1159 			getboottimebin(&boottimebin);
1160 			bintime_add(&bt, &boottimebin);
1161 		} else {
1162 			bintime(&bt);
1163 		}
1164 		*mp = sbcreatecontrol((caddr_t)&bt, sizeof(bt),
1165 		    SCM_BINTIME, SOL_SOCKET);
1166 		if (*mp != NULL) {
1167 			mp = &(*mp)->m_next;
1168 			stamped = true;
1169 		}
1170 	}
1171 	if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME_MICRO)) {
1172 		struct bintime boottimebin, bt1;
1173 		struct timespec ts1;
1174 		struct timeval tv;
1175 
1176 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1177 		    M_TSTMP)) {
1178 			mbuf_tstmp2timespec(m, &ts1);
1179 			timespec2bintime(&ts1, &bt1);
1180 			getboottimebin(&boottimebin);
1181 			bintime_add(&bt1, &boottimebin);
1182 			bintime2timeval(&bt1, &tv);
1183 		} else {
1184 			microtime(&tv);
1185 		}
1186 		*mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv),
1187 		    SCM_TIMESTAMP, SOL_SOCKET);
1188 		if (*mp != NULL) {
1189 			mp = &(*mp)->m_next;
1190 			stamped = true;
1191 		}
1192 	} else if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME)) {
1193 		struct bintime boottimebin;
1194 		struct timespec ts, ts1;
1195 
1196 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1197 		    M_TSTMP)) {
1198 			mbuf_tstmp2timespec(m, &ts);
1199 			getboottimebin(&boottimebin);
1200 			bintime2timespec(&boottimebin, &ts1);
1201 			timespecadd(&ts, &ts1, &ts);
1202 		} else {
1203 			nanotime(&ts);
1204 		}
1205 		*mp = sbcreatecontrol((caddr_t)&ts, sizeof(ts),
1206 		    SCM_REALTIME, SOL_SOCKET);
1207 		if (*mp != NULL) {
1208 			mp = &(*mp)->m_next;
1209 			stamped = true;
1210 		}
1211 	} else if (CHECK_SO_CT(inp->inp_socket, SO_TS_MONOTONIC)) {
1212 		struct timespec ts;
1213 
1214 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1215 		    M_TSTMP))
1216 			mbuf_tstmp2timespec(m, &ts);
1217 		else
1218 			nanouptime(&ts);
1219 		*mp = sbcreatecontrol((caddr_t)&ts, sizeof(ts),
1220 		    SCM_MONOTONIC, SOL_SOCKET);
1221 		if (*mp != NULL) {
1222 			mp = &(*mp)->m_next;
1223 			stamped = true;
1224 		}
1225 	}
1226 	if (stamped && (m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1227 	    M_TSTMP)) {
1228 		struct sock_timestamp_info sti;
1229 
1230 		bzero(&sti, sizeof(sti));
1231 		sti.st_info_flags = ST_INFO_HW;
1232 		if ((m->m_flags & M_TSTMP_HPREC) != 0)
1233 			sti.st_info_flags |= ST_INFO_HW_HPREC;
1234 		*mp = sbcreatecontrol((caddr_t)&sti, sizeof(sti), SCM_TIME_INFO,
1235 		    SOL_SOCKET);
1236 		if (*mp != NULL)
1237 			mp = &(*mp)->m_next;
1238 	}
1239 	if (inp->inp_flags & INP_RECVDSTADDR) {
1240 		*mp = sbcreatecontrol((caddr_t)&ip->ip_dst,
1241 		    sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1242 		if (*mp)
1243 			mp = &(*mp)->m_next;
1244 	}
1245 	if (inp->inp_flags & INP_RECVTTL) {
1246 		*mp = sbcreatecontrol((caddr_t)&ip->ip_ttl,
1247 		    sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
1248 		if (*mp)
1249 			mp = &(*mp)->m_next;
1250 	}
1251 #ifdef notyet
1252 	/* XXX
1253 	 * Moving these out of udp_input() made them even more broken
1254 	 * than they already were.
1255 	 */
1256 	/* options were tossed already */
1257 	if (inp->inp_flags & INP_RECVOPTS) {
1258 		*mp = sbcreatecontrol((caddr_t)opts_deleted_above,
1259 		    sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1260 		if (*mp)
1261 			mp = &(*mp)->m_next;
1262 	}
1263 	/* ip_srcroute doesn't do what we want here, need to fix */
1264 	if (inp->inp_flags & INP_RECVRETOPTS) {
1265 		*mp = sbcreatecontrol((caddr_t)ip_srcroute(m),
1266 		    sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1267 		if (*mp)
1268 			mp = &(*mp)->m_next;
1269 	}
1270 #endif
1271 	if (inp->inp_flags & INP_RECVIF) {
1272 		struct ifnet *ifp;
1273 		struct sdlbuf {
1274 			struct sockaddr_dl sdl;
1275 			u_char	pad[32];
1276 		} sdlbuf;
1277 		struct sockaddr_dl *sdp;
1278 		struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1279 
1280 		if ((ifp = m->m_pkthdr.rcvif) &&
1281 		    ifp->if_index && ifp->if_index <= V_if_index) {
1282 			sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1283 			/*
1284 			 * Change our mind and don't try copy.
1285 			 */
1286 			if (sdp->sdl_family != AF_LINK ||
1287 			    sdp->sdl_len > sizeof(sdlbuf)) {
1288 				goto makedummy;
1289 			}
1290 			bcopy(sdp, sdl2, sdp->sdl_len);
1291 		} else {
1292 makedummy:
1293 			sdl2->sdl_len =
1294 			    offsetof(struct sockaddr_dl, sdl_data[0]);
1295 			sdl2->sdl_family = AF_LINK;
1296 			sdl2->sdl_index = 0;
1297 			sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1298 		}
1299 		*mp = sbcreatecontrol((caddr_t)sdl2, sdl2->sdl_len,
1300 		    IP_RECVIF, IPPROTO_IP);
1301 		if (*mp)
1302 			mp = &(*mp)->m_next;
1303 	}
1304 	if (inp->inp_flags & INP_RECVTOS) {
1305 		*mp = sbcreatecontrol((caddr_t)&ip->ip_tos,
1306 		    sizeof(u_char), IP_RECVTOS, IPPROTO_IP);
1307 		if (*mp)
1308 			mp = &(*mp)->m_next;
1309 	}
1310 
1311 	if (inp->inp_flags2 & INP_RECVFLOWID) {
1312 		uint32_t flowid, flow_type;
1313 
1314 		flowid = m->m_pkthdr.flowid;
1315 		flow_type = M_HASHTYPE_GET(m);
1316 
1317 		/*
1318 		 * XXX should handle the failure of one or the
1319 		 * other - don't populate both?
1320 		 */
1321 		*mp = sbcreatecontrol((caddr_t) &flowid,
1322 		    sizeof(uint32_t), IP_FLOWID, IPPROTO_IP);
1323 		if (*mp)
1324 			mp = &(*mp)->m_next;
1325 		*mp = sbcreatecontrol((caddr_t) &flow_type,
1326 		    sizeof(uint32_t), IP_FLOWTYPE, IPPROTO_IP);
1327 		if (*mp)
1328 			mp = &(*mp)->m_next;
1329 	}
1330 
1331 #ifdef	RSS
1332 	if (inp->inp_flags2 & INP_RECVRSSBUCKETID) {
1333 		uint32_t flowid, flow_type;
1334 		uint32_t rss_bucketid;
1335 
1336 		flowid = m->m_pkthdr.flowid;
1337 		flow_type = M_HASHTYPE_GET(m);
1338 
1339 		if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) {
1340 			*mp = sbcreatecontrol((caddr_t) &rss_bucketid,
1341 			   sizeof(uint32_t), IP_RSSBUCKETID, IPPROTO_IP);
1342 			if (*mp)
1343 				mp = &(*mp)->m_next;
1344 		}
1345 	}
1346 #endif
1347 }
1348 
1349 /*
1350  * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
1351  * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
1352  * locking.  This code remains in ip_input.c as ip_mroute.c is optionally
1353  * compiled.
1354  */
1355 VNET_DEFINE_STATIC(int, ip_rsvp_on);
1356 VNET_DEFINE(struct socket *, ip_rsvpd);
1357 
1358 #define	V_ip_rsvp_on		VNET(ip_rsvp_on)
1359 
1360 int
1361 ip_rsvp_init(struct socket *so)
1362 {
1363 
1364 	if (so->so_type != SOCK_RAW ||
1365 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
1366 		return EOPNOTSUPP;
1367 
1368 	if (V_ip_rsvpd != NULL)
1369 		return EADDRINUSE;
1370 
1371 	V_ip_rsvpd = so;
1372 	/*
1373 	 * This may seem silly, but we need to be sure we don't over-increment
1374 	 * the RSVP counter, in case something slips up.
1375 	 */
1376 	if (!V_ip_rsvp_on) {
1377 		V_ip_rsvp_on = 1;
1378 		V_rsvp_on++;
1379 	}
1380 
1381 	return 0;
1382 }
1383 
1384 int
1385 ip_rsvp_done(void)
1386 {
1387 
1388 	V_ip_rsvpd = NULL;
1389 	/*
1390 	 * This may seem silly, but we need to be sure we don't over-decrement
1391 	 * the RSVP counter, in case something slips up.
1392 	 */
1393 	if (V_ip_rsvp_on) {
1394 		V_ip_rsvp_on = 0;
1395 		V_rsvp_on--;
1396 	}
1397 	return 0;
1398 }
1399 
1400 int
1401 rsvp_input(struct mbuf **mp, int *offp, int proto)
1402 {
1403 	struct mbuf *m;
1404 
1405 	m = *mp;
1406 	*mp = NULL;
1407 
1408 	if (rsvp_input_p) { /* call the real one if loaded */
1409 		*mp = m;
1410 		rsvp_input_p(mp, offp, proto);
1411 		return (IPPROTO_DONE);
1412 	}
1413 
1414 	/* Can still get packets with rsvp_on = 0 if there is a local member
1415 	 * of the group to which the RSVP packet is addressed.  But in this
1416 	 * case we want to throw the packet away.
1417 	 */
1418 
1419 	if (!V_rsvp_on) {
1420 		m_freem(m);
1421 		return (IPPROTO_DONE);
1422 	}
1423 
1424 	if (V_ip_rsvpd != NULL) {
1425 		*mp = m;
1426 		rip_input(mp, offp, proto);
1427 		return (IPPROTO_DONE);
1428 	}
1429 	/* Drop the packet */
1430 	m_freem(m);
1431 	return (IPPROTO_DONE);
1432 }
1433