xref: /freebsd/sys/netinet/ip_input.c (revision 713d00e7b61b6d9c627a31cd3de92309fbe23c96)
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 VNET_DEFINE_STATIC(int, ipsendredirects) = 1;	/* XXX */
115 #define	V_ipsendredirects	VNET(ipsendredirects)
116 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_VNET | CTLFLAG_RW,
117     &VNET_NAME(ipsendredirects), 0,
118     "Enable sending IP redirects");
119 
120 /*
121  * XXX - Setting ip_checkinterface mostly implements the receive side of
122  * the Strong ES model described in RFC 1122, but since the routing table
123  * and transmit implementation do not implement the Strong ES model,
124  * setting this to 1 results in an odd hybrid.
125  *
126  * XXX - ip_checkinterface currently must be disabled if you use ipnat
127  * to translate the destination address to another local interface.
128  *
129  * XXX - ip_checkinterface must be disabled if you add IP aliases
130  * to the loopback interface instead of the interface where the
131  * packets for those addresses are received.
132  */
133 VNET_DEFINE_STATIC(int, ip_checkinterface);
134 #define	V_ip_checkinterface	VNET(ip_checkinterface)
135 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_VNET | CTLFLAG_RW,
136     &VNET_NAME(ip_checkinterface), 0,
137     "Verify packet arrives on correct interface");
138 
139 VNET_DEFINE(pfil_head_t, inet_pfil_head);	/* Packet filter hooks */
140 
141 static struct netisr_handler ip_nh = {
142 	.nh_name = "ip",
143 	.nh_handler = ip_input,
144 	.nh_proto = NETISR_IP,
145 #ifdef	RSS
146 	.nh_m2cpuid = rss_soft_m2cpuid_v4,
147 	.nh_policy = NETISR_POLICY_CPU,
148 	.nh_dispatch = NETISR_DISPATCH_HYBRID,
149 #else
150 	.nh_policy = NETISR_POLICY_FLOW,
151 #endif
152 };
153 
154 #ifdef	RSS
155 /*
156  * Directly dispatched frames are currently assumed
157  * to have a flowid already calculated.
158  *
159  * It should likely have something that assert it
160  * actually has valid flow details.
161  */
162 static struct netisr_handler ip_direct_nh = {
163 	.nh_name = "ip_direct",
164 	.nh_handler = ip_direct_input,
165 	.nh_proto = NETISR_IP_DIRECT,
166 	.nh_m2cpuid = rss_soft_m2cpuid_v4,
167 	.nh_policy = NETISR_POLICY_CPU,
168 	.nh_dispatch = NETISR_DISPATCH_HYBRID,
169 };
170 #endif
171 
172 extern	struct domain inetdomain;
173 extern	struct protosw inetsw[];
174 u_char	ip_protox[IPPROTO_MAX];
175 VNET_DEFINE(struct in_ifaddrhead, in_ifaddrhead);  /* first inet address */
176 VNET_DEFINE(struct in_ifaddrhashhead *, in_ifaddrhashtbl); /* inet addr hash table  */
177 VNET_DEFINE(u_long, in_ifaddrhmask);		/* mask for hash table */
178 
179 #ifdef IPCTL_DEFMTU
180 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
181     &ip_mtu, 0, "Default MTU");
182 #endif
183 
184 #ifdef IPSTEALTH
185 VNET_DEFINE(int, ipstealth);
186 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_VNET | CTLFLAG_RW,
187     &VNET_NAME(ipstealth), 0,
188     "IP stealth mode, no TTL decrementation on forwarding");
189 #endif
190 
191 /*
192  * IP statistics are stored in the "array" of counter(9)s.
193  */
194 VNET_PCPUSTAT_DEFINE(struct ipstat, ipstat);
195 VNET_PCPUSTAT_SYSINIT(ipstat);
196 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, IPCTL_STATS, stats, struct ipstat, ipstat,
197     "IP statistics (struct ipstat, netinet/ip_var.h)");
198 
199 #ifdef VIMAGE
200 VNET_PCPUSTAT_SYSUNINIT(ipstat);
201 #endif /* VIMAGE */
202 
203 /*
204  * Kernel module interface for updating ipstat.  The argument is an index
205  * into ipstat treated as an array.
206  */
207 void
208 kmod_ipstat_inc(int statnum)
209 {
210 
211 	counter_u64_add(VNET(ipstat)[statnum], 1);
212 }
213 
214 void
215 kmod_ipstat_dec(int statnum)
216 {
217 
218 	counter_u64_add(VNET(ipstat)[statnum], -1);
219 }
220 
221 static int
222 sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS)
223 {
224 	int error, qlimit;
225 
226 	netisr_getqlimit(&ip_nh, &qlimit);
227 	error = sysctl_handle_int(oidp, &qlimit, 0, req);
228 	if (error || !req->newptr)
229 		return (error);
230 	if (qlimit < 1)
231 		return (EINVAL);
232 	return (netisr_setqlimit(&ip_nh, qlimit));
233 }
234 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen,
235     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 0,
236     sysctl_netinet_intr_queue_maxlen, "I",
237     "Maximum size of the IP input queue");
238 
239 static int
240 sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS)
241 {
242 	u_int64_t qdrops_long;
243 	int error, qdrops;
244 
245 	netisr_getqdrops(&ip_nh, &qdrops_long);
246 	qdrops = qdrops_long;
247 	error = sysctl_handle_int(oidp, &qdrops, 0, req);
248 	if (error || !req->newptr)
249 		return (error);
250 	if (qdrops != 0)
251 		return (EINVAL);
252 	netisr_clearqdrops(&ip_nh);
253 	return (0);
254 }
255 
256 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops,
257     CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE,
258     0, 0, sysctl_netinet_intr_queue_drops, "I",
259     "Number of packets dropped from the IP input queue");
260 
261 #ifdef	RSS
262 static int
263 sysctl_netinet_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS)
264 {
265 	int error, qlimit;
266 
267 	netisr_getqlimit(&ip_direct_nh, &qlimit);
268 	error = sysctl_handle_int(oidp, &qlimit, 0, req);
269 	if (error || !req->newptr)
270 		return (error);
271 	if (qlimit < 1)
272 		return (EINVAL);
273 	return (netisr_setqlimit(&ip_direct_nh, qlimit));
274 }
275 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRDQMAXLEN, intr_direct_queue_maxlen,
276     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
277     0, 0, sysctl_netinet_intr_direct_queue_maxlen,
278     "I", "Maximum size of the IP direct input queue");
279 
280 static int
281 sysctl_netinet_intr_direct_queue_drops(SYSCTL_HANDLER_ARGS)
282 {
283 	u_int64_t qdrops_long;
284 	int error, qdrops;
285 
286 	netisr_getqdrops(&ip_direct_nh, &qdrops_long);
287 	qdrops = qdrops_long;
288 	error = sysctl_handle_int(oidp, &qdrops, 0, req);
289 	if (error || !req->newptr)
290 		return (error);
291 	if (qdrops != 0)
292 		return (EINVAL);
293 	netisr_clearqdrops(&ip_direct_nh);
294 	return (0);
295 }
296 
297 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRDQDROPS, intr_direct_queue_drops,
298     CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, 0, 0,
299     sysctl_netinet_intr_direct_queue_drops, "I",
300     "Number of packets dropped from the IP direct input queue");
301 #endif	/* RSS */
302 
303 /*
304  * IP initialization: fill in IP protocol switch table.
305  * All protocols not implemented in kernel go to raw IP protocol handler.
306  */
307 void
308 ip_init(void)
309 {
310 	struct pfil_head_args args;
311 	struct protosw *pr;
312 	int i;
313 
314 	CK_STAILQ_INIT(&V_in_ifaddrhead);
315 	V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask);
316 
317 	/* Initialize IP reassembly queue. */
318 	ipreass_init();
319 
320 	/* Initialize packet filter hooks. */
321 	args.pa_version = PFIL_VERSION;
322 	args.pa_flags = PFIL_IN | PFIL_OUT;
323 	args.pa_type = PFIL_TYPE_IP4;
324 	args.pa_headname = PFIL_INET_NAME;
325 	V_inet_pfil_head = pfil_head_register(&args);
326 
327 	if (hhook_head_register(HHOOK_TYPE_IPSEC_IN, AF_INET,
328 	    &V_ipsec_hhh_in[HHOOK_IPSEC_INET],
329 	    HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
330 		printf("%s: WARNING: unable to register input helper hook\n",
331 		    __func__);
332 	if (hhook_head_register(HHOOK_TYPE_IPSEC_OUT, AF_INET,
333 	    &V_ipsec_hhh_out[HHOOK_IPSEC_INET],
334 	    HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
335 		printf("%s: WARNING: unable to register output helper hook\n",
336 		    __func__);
337 
338 	/* Skip initialization of globals for non-default instances. */
339 #ifdef VIMAGE
340 	if (!IS_DEFAULT_VNET(curvnet)) {
341 		netisr_register_vnet(&ip_nh);
342 #ifdef	RSS
343 		netisr_register_vnet(&ip_direct_nh);
344 #endif
345 		return;
346 	}
347 #endif
348 
349 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
350 	if (pr == NULL)
351 		panic("ip_init: PF_INET not found");
352 
353 	/* Initialize the entire ip_protox[] array to IPPROTO_RAW. */
354 	for (i = 0; i < IPPROTO_MAX; i++)
355 		ip_protox[i] = pr - inetsw;
356 	/*
357 	 * Cycle through IP protocols and put them into the appropriate place
358 	 * in ip_protox[].
359 	 */
360 	for (pr = inetdomain.dom_protosw;
361 	    pr < inetdomain.dom_protoswNPROTOSW; pr++)
362 		if (pr->pr_domain->dom_family == PF_INET &&
363 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
364 			/* Be careful to only index valid IP protocols. */
365 			if (pr->pr_protocol < IPPROTO_MAX)
366 				ip_protox[pr->pr_protocol] = pr - inetsw;
367 		}
368 
369 	netisr_register(&ip_nh);
370 #ifdef	RSS
371 	netisr_register(&ip_direct_nh);
372 #endif
373 }
374 
375 #ifdef VIMAGE
376 static void
377 ip_destroy(void *unused __unused)
378 {
379 	struct ifnet *ifp;
380 	int error;
381 
382 #ifdef	RSS
383 	netisr_unregister_vnet(&ip_direct_nh);
384 #endif
385 	netisr_unregister_vnet(&ip_nh);
386 
387 	pfil_head_unregister(V_inet_pfil_head);
388 	error = hhook_head_deregister(V_ipsec_hhh_in[HHOOK_IPSEC_INET]);
389 	if (error != 0) {
390 		printf("%s: WARNING: unable to deregister input helper hook "
391 		    "type HHOOK_TYPE_IPSEC_IN, id HHOOK_IPSEC_INET: "
392 		    "error %d returned\n", __func__, error);
393 	}
394 	error = hhook_head_deregister(V_ipsec_hhh_out[HHOOK_IPSEC_INET]);
395 	if (error != 0) {
396 		printf("%s: WARNING: unable to deregister output helper hook "
397 		    "type HHOOK_TYPE_IPSEC_OUT, id HHOOK_IPSEC_INET: "
398 		    "error %d returned\n", __func__, error);
399 	}
400 
401 	/* Remove the IPv4 addresses from all interfaces. */
402 	in_ifscrub_all();
403 
404 	/* Make sure the IPv4 routes are gone as well. */
405 	IFNET_RLOCK();
406 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link)
407 		rt_flushifroutes_af(ifp, AF_INET);
408 	IFNET_RUNLOCK();
409 
410 	/* Destroy IP reassembly queue. */
411 	ipreass_destroy();
412 
413 	/* Cleanup in_ifaddr hash table; should be empty. */
414 	hashdestroy(V_in_ifaddrhashtbl, M_IFADDR, V_in_ifaddrhmask);
415 }
416 
417 VNET_SYSUNINIT(ip, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip_destroy, NULL);
418 #endif
419 
420 #ifdef	RSS
421 /*
422  * IP direct input routine.
423  *
424  * This is called when reinjecting completed fragments where
425  * all of the previous checking and book-keeping has been done.
426  */
427 void
428 ip_direct_input(struct mbuf *m)
429 {
430 	struct ip *ip;
431 	int hlen;
432 
433 	ip = mtod(m, struct ip *);
434 	hlen = ip->ip_hl << 2;
435 
436 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
437 	if (IPSEC_ENABLED(ipv4)) {
438 		if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
439 			return;
440 	}
441 #endif /* IPSEC */
442 	IPSTAT_INC(ips_delivered);
443 	(*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
444 	return;
445 }
446 #endif
447 
448 /*
449  * Ip input routine.  Checksum and byte swap header.  If fragmented
450  * try to reassemble.  Process options.  Pass to next level.
451  */
452 void
453 ip_input(struct mbuf *m)
454 {
455 	struct rm_priotracker in_ifa_tracker;
456 	struct ip *ip = NULL;
457 	struct in_ifaddr *ia = NULL;
458 	struct ifaddr *ifa;
459 	struct ifnet *ifp;
460 	int    checkif, hlen = 0;
461 	uint16_t sum, ip_len;
462 	int dchg = 0;				/* dest changed after fw */
463 	struct in_addr odst;			/* original dst address */
464 
465 	M_ASSERTPKTHDR(m);
466 	NET_EPOCH_ASSERT();
467 
468 	if (m->m_flags & M_FASTFWD_OURS) {
469 		m->m_flags &= ~M_FASTFWD_OURS;
470 		/* Set up some basics that will be used later. */
471 		ip = mtod(m, struct ip *);
472 		hlen = ip->ip_hl << 2;
473 		ip_len = ntohs(ip->ip_len);
474 		goto ours;
475 	}
476 
477 	IPSTAT_INC(ips_total);
478 
479 	if (m->m_pkthdr.len < sizeof(struct ip))
480 		goto tooshort;
481 
482 	if (m->m_len < sizeof (struct ip) &&
483 	    (m = m_pullup(m, sizeof (struct ip))) == NULL) {
484 		IPSTAT_INC(ips_toosmall);
485 		return;
486 	}
487 	ip = mtod(m, struct ip *);
488 
489 	if (ip->ip_v != IPVERSION) {
490 		IPSTAT_INC(ips_badvers);
491 		goto bad;
492 	}
493 
494 	hlen = ip->ip_hl << 2;
495 	if (hlen < sizeof(struct ip)) {	/* minimum header length */
496 		IPSTAT_INC(ips_badhlen);
497 		goto bad;
498 	}
499 	if (hlen > m->m_len) {
500 		if ((m = m_pullup(m, hlen)) == NULL) {
501 			IPSTAT_INC(ips_badhlen);
502 			return;
503 		}
504 		ip = mtod(m, struct ip *);
505 	}
506 
507 	IP_PROBE(receive, NULL, NULL, ip, m->m_pkthdr.rcvif, ip, NULL);
508 
509 	/* IN_LOOPBACK must not appear on the wire - RFC1122 */
510 	ifp = m->m_pkthdr.rcvif;
511 	if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) ||
512 	    IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) {
513 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
514 			IPSTAT_INC(ips_badaddr);
515 			goto bad;
516 		}
517 	}
518 
519 	if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
520 		sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
521 	} else {
522 		if (hlen == sizeof(struct ip)) {
523 			sum = in_cksum_hdr(ip);
524 		} else {
525 			sum = in_cksum(m, hlen);
526 		}
527 	}
528 	if (sum) {
529 		IPSTAT_INC(ips_badsum);
530 		goto bad;
531 	}
532 
533 #ifdef ALTQ
534 	if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
535 		/* packet is dropped by traffic conditioner */
536 		return;
537 #endif
538 
539 	ip_len = ntohs(ip->ip_len);
540 	if (ip_len < hlen) {
541 		IPSTAT_INC(ips_badlen);
542 		goto bad;
543 	}
544 
545 	/*
546 	 * Check that the amount of data in the buffers
547 	 * is as at least much as the IP header would have us expect.
548 	 * Trim mbufs if longer than we expect.
549 	 * Drop packet if shorter than we expect.
550 	 */
551 	if (m->m_pkthdr.len < ip_len) {
552 tooshort:
553 		IPSTAT_INC(ips_tooshort);
554 		goto bad;
555 	}
556 	if (m->m_pkthdr.len > ip_len) {
557 		if (m->m_len == m->m_pkthdr.len) {
558 			m->m_len = ip_len;
559 			m->m_pkthdr.len = ip_len;
560 		} else
561 			m_adj(m, ip_len - m->m_pkthdr.len);
562 	}
563 
564 	/*
565 	 * Try to forward the packet, but if we fail continue.
566 	 * ip_tryforward() does not generate redirects, so fall
567 	 * through to normal processing if redirects are required.
568 	 * ip_tryforward() does inbound and outbound packet firewall
569 	 * processing. If firewall has decided that destination becomes
570 	 * our local address, it sets M_FASTFWD_OURS flag. In this
571 	 * case skip another inbound firewall processing and update
572 	 * ip pointer.
573 	 */
574 	if (V_ipforwarding != 0 && V_ipsendredirects == 0
575 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
576 	    && (!IPSEC_ENABLED(ipv4) ||
577 	    IPSEC_CAPS(ipv4, m, IPSEC_CAP_OPERABLE) == 0)
578 #endif
579 	    ) {
580 		if ((m = ip_tryforward(m)) == NULL)
581 			return;
582 		if (m->m_flags & M_FASTFWD_OURS) {
583 			m->m_flags &= ~M_FASTFWD_OURS;
584 			ip = mtod(m, struct ip *);
585 			goto ours;
586 		}
587 	}
588 
589 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
590 	/*
591 	 * Bypass packet filtering for packets previously handled by IPsec.
592 	 */
593 	if (IPSEC_ENABLED(ipv4) &&
594 	    IPSEC_CAPS(ipv4, m, IPSEC_CAP_BYPASS_FILTER) != 0)
595 			goto passin;
596 #endif
597 
598 	/*
599 	 * Run through list of hooks for input packets.
600 	 *
601 	 * NB: Beware of the destination address changing (e.g.
602 	 *     by NAT rewriting).  When this happens, tell
603 	 *     ip_forward to do the right thing.
604 	 */
605 
606 	/* Jump over all PFIL processing if hooks are not active. */
607 	if (!PFIL_HOOKED_IN(V_inet_pfil_head))
608 		goto passin;
609 
610 	odst = ip->ip_dst;
611 	if (pfil_run_hooks(V_inet_pfil_head, &m, ifp, PFIL_IN, NULL) !=
612 	    PFIL_PASS)
613 		return;
614 	if (m == NULL)			/* consumed by filter */
615 		return;
616 
617 	ip = mtod(m, struct ip *);
618 	dchg = (odst.s_addr != ip->ip_dst.s_addr);
619 	ifp = m->m_pkthdr.rcvif;
620 
621 	if (m->m_flags & M_FASTFWD_OURS) {
622 		m->m_flags &= ~M_FASTFWD_OURS;
623 		goto ours;
624 	}
625 	if (m->m_flags & M_IP_NEXTHOP) {
626 		if (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) {
627 			/*
628 			 * Directly ship the packet on.  This allows
629 			 * forwarding packets originally destined to us
630 			 * to some other directly connected host.
631 			 */
632 			ip_forward(m, 1);
633 			return;
634 		}
635 	}
636 passin:
637 
638 	/*
639 	 * Process options and, if not destined for us,
640 	 * ship it on.  ip_dooptions returns 1 when an
641 	 * error was detected (causing an icmp message
642 	 * to be sent and the original packet to be freed).
643 	 */
644 	if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
645 		return;
646 
647         /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
648          * matter if it is destined to another node, or whether it is
649          * a multicast one, RSVP wants it! and prevents it from being forwarded
650          * anywhere else. Also checks if the rsvp daemon is running before
651 	 * grabbing the packet.
652          */
653 	if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP)
654 		goto ours;
655 
656 	/*
657 	 * Check our list of addresses, to see if the packet is for us.
658 	 * If we don't have any addresses, assume any unicast packet
659 	 * we receive might be for us (and let the upper layers deal
660 	 * with it).
661 	 */
662 	if (CK_STAILQ_EMPTY(&V_in_ifaddrhead) &&
663 	    (m->m_flags & (M_MCAST|M_BCAST)) == 0)
664 		goto ours;
665 
666 	/*
667 	 * Enable a consistency check between the destination address
668 	 * and the arrival interface for a unicast packet (the RFC 1122
669 	 * strong ES model) if IP forwarding is disabled and the packet
670 	 * is not locally generated and the packet is not subject to
671 	 * 'ipfw fwd'.
672 	 *
673 	 * XXX - Checking also should be disabled if the destination
674 	 * address is ipnat'ed to a different interface.
675 	 *
676 	 * XXX - Checking is incompatible with IP aliases added
677 	 * to the loopback interface instead of the interface where
678 	 * the packets are received.
679 	 *
680 	 * XXX - This is the case for carp vhost IPs as well so we
681 	 * insert a workaround. If the packet got here, we already
682 	 * checked with carp_iamatch() and carp_forus().
683 	 */
684 	checkif = V_ip_checkinterface && (V_ipforwarding == 0) &&
685 	    ifp != NULL && ((ifp->if_flags & IFF_LOOPBACK) == 0) &&
686 	    ifp->if_carp == NULL && (dchg == 0);
687 
688 	/*
689 	 * Check for exact addresses in the hash bucket.
690 	 */
691 	IN_IFADDR_RLOCK(&in_ifa_tracker);
692 	LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
693 		/*
694 		 * If the address matches, verify that the packet
695 		 * arrived via the correct interface if checking is
696 		 * enabled.
697 		 */
698 		if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr &&
699 		    (!checkif || ia->ia_ifp == ifp)) {
700 			counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
701 			counter_u64_add(ia->ia_ifa.ifa_ibytes,
702 			    m->m_pkthdr.len);
703 			IN_IFADDR_RUNLOCK(&in_ifa_tracker);
704 			goto ours;
705 		}
706 	}
707 	IN_IFADDR_RUNLOCK(&in_ifa_tracker);
708 
709 	/*
710 	 * Check for broadcast addresses.
711 	 *
712 	 * Only accept broadcast packets that arrive via the matching
713 	 * interface.  Reception of forwarded directed broadcasts would
714 	 * be handled via ip_forward() and ether_output() with the loopback
715 	 * into the stack for SIMPLEX interfaces handled by ether_output().
716 	 */
717 	if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
718 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
719 			if (ifa->ifa_addr->sa_family != AF_INET)
720 				continue;
721 			ia = ifatoia(ifa);
722 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
723 			    ip->ip_dst.s_addr) {
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 				goto ours;
728 			}
729 #ifdef BOOTP_COMPAT
730 			if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) {
731 				counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
732 				counter_u64_add(ia->ia_ifa.ifa_ibytes,
733 				    m->m_pkthdr.len);
734 				goto ours;
735 			}
736 #endif
737 		}
738 		ia = NULL;
739 	}
740 	/* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */
741 	if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) {
742 		IPSTAT_INC(ips_cantforward);
743 		m_freem(m);
744 		return;
745 	}
746 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
747 		if (V_ip_mrouter) {
748 			/*
749 			 * If we are acting as a multicast router, all
750 			 * incoming multicast packets are passed to the
751 			 * kernel-level multicast forwarding function.
752 			 * The packet is returned (relatively) intact; if
753 			 * ip_mforward() returns a non-zero value, the packet
754 			 * must be discarded, else it may be accepted below.
755 			 */
756 			if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) {
757 				IPSTAT_INC(ips_cantforward);
758 				m_freem(m);
759 				return;
760 			}
761 
762 			/*
763 			 * The process-level routing daemon needs to receive
764 			 * all multicast IGMP packets, whether or not this
765 			 * host belongs to their destination groups.
766 			 */
767 			if (ip->ip_p == IPPROTO_IGMP)
768 				goto ours;
769 			IPSTAT_INC(ips_forward);
770 		}
771 		/*
772 		 * Assume the packet is for us, to avoid prematurely taking
773 		 * a lock on the in_multi hash. Protocols must perform
774 		 * their own filtering and update statistics accordingly.
775 		 */
776 		goto ours;
777 	}
778 	if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
779 		goto ours;
780 	if (ip->ip_dst.s_addr == INADDR_ANY)
781 		goto ours;
782 
783 	/*
784 	 * Not for us; forward if possible and desirable.
785 	 */
786 	if (V_ipforwarding == 0) {
787 		IPSTAT_INC(ips_cantforward);
788 		m_freem(m);
789 	} else {
790 		ip_forward(m, dchg);
791 	}
792 	return;
793 
794 ours:
795 #ifdef IPSTEALTH
796 	/*
797 	 * IPSTEALTH: Process non-routing options only
798 	 * if the packet is destined for us.
799 	 */
800 	if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1))
801 		return;
802 #endif /* IPSTEALTH */
803 
804 	/*
805 	 * Attempt reassembly; if it succeeds, proceed.
806 	 * ip_reass() will return a different mbuf.
807 	 */
808 	if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) {
809 		/* XXXGL: shouldn't we save & set m_flags? */
810 		m = ip_reass(m);
811 		if (m == NULL)
812 			return;
813 		ip = mtod(m, struct ip *);
814 		/* Get the header length of the reassembled packet */
815 		hlen = ip->ip_hl << 2;
816 	}
817 
818 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
819 	if (IPSEC_ENABLED(ipv4)) {
820 		if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
821 			return;
822 	}
823 #endif /* IPSEC */
824 
825 	/*
826 	 * Switch out to protocol's input routine.
827 	 */
828 	IPSTAT_INC(ips_delivered);
829 
830 	(*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
831 	return;
832 bad:
833 	m_freem(m);
834 }
835 
836 /*
837  * IP timer processing;
838  * if a timer expires on a reassembly
839  * queue, discard it.
840  */
841 void
842 ip_slowtimo(void)
843 {
844 	VNET_ITERATOR_DECL(vnet_iter);
845 
846 	VNET_LIST_RLOCK_NOSLEEP();
847 	VNET_FOREACH(vnet_iter) {
848 		CURVNET_SET(vnet_iter);
849 		ipreass_slowtimo();
850 		CURVNET_RESTORE();
851 	}
852 	VNET_LIST_RUNLOCK_NOSLEEP();
853 }
854 
855 void
856 ip_drain(void)
857 {
858 	VNET_ITERATOR_DECL(vnet_iter);
859 
860 	VNET_LIST_RLOCK_NOSLEEP();
861 	VNET_FOREACH(vnet_iter) {
862 		CURVNET_SET(vnet_iter);
863 		ipreass_drain();
864 		CURVNET_RESTORE();
865 	}
866 	VNET_LIST_RUNLOCK_NOSLEEP();
867 }
868 
869 /*
870  * The protocol to be inserted into ip_protox[] must be already registered
871  * in inetsw[], either statically or through pf_proto_register().
872  */
873 int
874 ipproto_register(short ipproto)
875 {
876 	struct protosw *pr;
877 
878 	/* Sanity checks. */
879 	if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
880 		return (EPROTONOSUPPORT);
881 
882 	/*
883 	 * The protocol slot must not be occupied by another protocol
884 	 * already.  An index pointing to IPPROTO_RAW is unused.
885 	 */
886 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
887 	if (pr == NULL)
888 		return (EPFNOSUPPORT);
889 	if (ip_protox[ipproto] != pr - inetsw)	/* IPPROTO_RAW */
890 		return (EEXIST);
891 
892 	/* Find the protocol position in inetsw[] and set the index. */
893 	for (pr = inetdomain.dom_protosw;
894 	     pr < inetdomain.dom_protoswNPROTOSW; pr++) {
895 		if (pr->pr_domain->dom_family == PF_INET &&
896 		    pr->pr_protocol && pr->pr_protocol == ipproto) {
897 			ip_protox[pr->pr_protocol] = pr - inetsw;
898 			return (0);
899 		}
900 	}
901 	return (EPROTONOSUPPORT);
902 }
903 
904 int
905 ipproto_unregister(short ipproto)
906 {
907 	struct protosw *pr;
908 
909 	/* Sanity checks. */
910 	if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
911 		return (EPROTONOSUPPORT);
912 
913 	/* Check if the protocol was indeed registered. */
914 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
915 	if (pr == NULL)
916 		return (EPFNOSUPPORT);
917 	if (ip_protox[ipproto] == pr - inetsw)  /* IPPROTO_RAW */
918 		return (ENOENT);
919 
920 	/* Reset the protocol slot to IPPROTO_RAW. */
921 	ip_protox[ipproto] = pr - inetsw;
922 	return (0);
923 }
924 
925 u_char inetctlerrmap[PRC_NCMDS] = {
926 	0,		0,		0,		0,
927 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
928 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
929 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
930 	0,		0,		EHOSTUNREACH,	0,
931 	ENOPROTOOPT,	ECONNREFUSED
932 };
933 
934 /*
935  * Forward a packet.  If some error occurs return the sender
936  * an icmp packet.  Note we can't always generate a meaningful
937  * icmp message because icmp doesn't have a large enough repertoire
938  * of codes and types.
939  *
940  * If not forwarding, just drop the packet.  This could be confusing
941  * if ipforwarding was zero but some routing protocol was advancing
942  * us as a gateway to somewhere.  However, we must let the routing
943  * protocol deal with that.
944  *
945  * The srcrt parameter indicates whether the packet is being forwarded
946  * via a source route.
947  */
948 void
949 ip_forward(struct mbuf *m, int srcrt)
950 {
951 	struct ip *ip = mtod(m, struct ip *);
952 	struct in_ifaddr *ia;
953 	struct mbuf *mcopy;
954 	struct sockaddr_in *sin;
955 	struct in_addr dest;
956 	struct route ro;
957 	uint32_t flowid;
958 	int error, type = 0, code = 0, mtu = 0;
959 
960 	NET_EPOCH_ASSERT();
961 
962 	if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
963 		IPSTAT_INC(ips_cantforward);
964 		m_freem(m);
965 		return;
966 	}
967 	if (
968 #ifdef IPSTEALTH
969 	    V_ipstealth == 0 &&
970 #endif
971 	    ip->ip_ttl <= IPTTLDEC) {
972 		icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 0, 0);
973 		return;
974 	}
975 
976 	bzero(&ro, sizeof(ro));
977 	sin = (struct sockaddr_in *)&ro.ro_dst;
978 	sin->sin_family = AF_INET;
979 	sin->sin_len = sizeof(*sin);
980 	sin->sin_addr = ip->ip_dst;
981 #ifdef RADIX_MPATH
982 	flowid = ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr);
983 #else
984 	flowid = m->m_pkthdr.flowid;
985 #endif
986 	ro.ro_nh = fib4_lookup(M_GETFIB(m), ip->ip_dst, 0, NHR_REF, flowid);
987 	if (ro.ro_nh != NULL) {
988 		ia = ifatoia(ro.ro_nh->nh_ifa);
989 	} else
990 		ia = NULL;
991 	/*
992 	 * Save the IP header and at most 8 bytes of the payload,
993 	 * in case we need to generate an ICMP message to the src.
994 	 *
995 	 * XXX this can be optimized a lot by saving the data in a local
996 	 * buffer on the stack (72 bytes at most), and only allocating the
997 	 * mbuf if really necessary. The vast majority of the packets
998 	 * are forwarded without having to send an ICMP back (either
999 	 * because unnecessary, or because rate limited), so we are
1000 	 * really we are wasting a lot of work here.
1001 	 *
1002 	 * We don't use m_copym() because it might return a reference
1003 	 * to a shared cluster. Both this function and ip_output()
1004 	 * assume exclusive access to the IP header in `m', so any
1005 	 * data in a cluster may change before we reach icmp_error().
1006 	 */
1007 	mcopy = m_gethdr(M_NOWAIT, m->m_type);
1008 	if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_NOWAIT)) {
1009 		/*
1010 		 * It's probably ok if the pkthdr dup fails (because
1011 		 * the deep copy of the tag chain failed), but for now
1012 		 * be conservative and just discard the copy since
1013 		 * code below may some day want the tags.
1014 		 */
1015 		m_free(mcopy);
1016 		mcopy = NULL;
1017 	}
1018 	if (mcopy != NULL) {
1019 		mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy));
1020 		mcopy->m_pkthdr.len = mcopy->m_len;
1021 		m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1022 	}
1023 #ifdef IPSTEALTH
1024 	if (V_ipstealth == 0)
1025 #endif
1026 		ip->ip_ttl -= IPTTLDEC;
1027 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1028 	if (IPSEC_ENABLED(ipv4)) {
1029 		if ((error = IPSEC_FORWARD(ipv4, m)) != 0) {
1030 			/* mbuf consumed by IPsec */
1031 			m_freem(mcopy);
1032 			if (error != EINPROGRESS)
1033 				IPSTAT_INC(ips_cantforward);
1034 			return;
1035 		}
1036 		/* No IPsec processing required */
1037 	}
1038 #endif /* IPSEC */
1039 	/*
1040 	 * If forwarding packet using same interface that it came in on,
1041 	 * perhaps should send a redirect to sender to shortcut a hop.
1042 	 * Only send redirect if source is sending directly to us,
1043 	 * and if packet was not source routed (or has any options).
1044 	 * Also, don't send redirect if forwarding using a default route
1045 	 * or a route modified by a redirect.
1046 	 */
1047 	dest.s_addr = 0;
1048 	if (!srcrt && V_ipsendredirects &&
1049 	    ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) {
1050 		struct nhop_object *nh;
1051 
1052 		nh = ro.ro_nh;
1053 
1054 		if (nh != NULL && ((nh->nh_flags & (NHF_REDIRECT|NHF_DEFAULT)) == 0)) {
1055 			struct in_ifaddr *nh_ia = (struct in_ifaddr *)(nh->nh_ifa);
1056 			u_long src = ntohl(ip->ip_src.s_addr);
1057 
1058 			if (nh_ia != NULL &&
1059 			    (src & nh_ia->ia_subnetmask) == nh_ia->ia_subnet) {
1060 				if (nh->nh_flags & NHF_GATEWAY)
1061 					dest.s_addr = nh->gw4_sa.sin_addr.s_addr;
1062 				else
1063 					dest.s_addr = ip->ip_dst.s_addr;
1064 				/* Router requirements says to only send host redirects */
1065 				type = ICMP_REDIRECT;
1066 				code = ICMP_REDIRECT_HOST;
1067 			}
1068 		}
1069 	}
1070 
1071 	error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
1072 
1073 	if (error == EMSGSIZE && ro.ro_nh)
1074 		mtu = ro.ro_nh->nh_mtu;
1075 	RO_NHFREE(&ro);
1076 
1077 	if (error)
1078 		IPSTAT_INC(ips_cantforward);
1079 	else {
1080 		IPSTAT_INC(ips_forward);
1081 		if (type)
1082 			IPSTAT_INC(ips_redirectsent);
1083 		else {
1084 			if (mcopy)
1085 				m_freem(mcopy);
1086 			return;
1087 		}
1088 	}
1089 	if (mcopy == NULL)
1090 		return;
1091 
1092 
1093 	switch (error) {
1094 
1095 	case 0:				/* forwarded, but need redirect */
1096 		/* type, code set above */
1097 		break;
1098 
1099 	case ENETUNREACH:
1100 	case EHOSTUNREACH:
1101 	case ENETDOWN:
1102 	case EHOSTDOWN:
1103 	default:
1104 		type = ICMP_UNREACH;
1105 		code = ICMP_UNREACH_HOST;
1106 		break;
1107 
1108 	case EMSGSIZE:
1109 		type = ICMP_UNREACH;
1110 		code = ICMP_UNREACH_NEEDFRAG;
1111 		/*
1112 		 * If the MTU was set before make sure we are below the
1113 		 * interface MTU.
1114 		 * If the MTU wasn't set before use the interface mtu or
1115 		 * fall back to the next smaller mtu step compared to the
1116 		 * current packet size.
1117 		 */
1118 		if (mtu != 0) {
1119 			if (ia != NULL)
1120 				mtu = min(mtu, ia->ia_ifp->if_mtu);
1121 		} else {
1122 			if (ia != NULL)
1123 				mtu = ia->ia_ifp->if_mtu;
1124 			else
1125 				mtu = ip_next_mtu(ntohs(ip->ip_len), 0);
1126 		}
1127 		IPSTAT_INC(ips_cantfrag);
1128 		break;
1129 
1130 	case ENOBUFS:
1131 	case EACCES:			/* ipfw denied packet */
1132 		m_freem(mcopy);
1133 		return;
1134 	}
1135 	icmp_error(mcopy, type, code, dest.s_addr, mtu);
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