xref: /freebsd/sys/netinet/ip_input.c (revision c07d6445eb89d9dd3950361b065b7bd110e3a043)
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_inet.h"
39 #include "opt_ipstealth.h"
40 #include "opt_ipsec.h"
41 #include "opt_route.h"
42 #include "opt_rss.h"
43 #include "opt_sctp.h"
44 
45 #include <sys/param.h>
46 #include <sys/systm.h>
47 #include <sys/hhook.h>
48 #include <sys/mbuf.h>
49 #include <sys/malloc.h>
50 #include <sys/domain.h>
51 #include <sys/protosw.h>
52 #include <sys/socket.h>
53 #include <sys/time.h>
54 #include <sys/kernel.h>
55 #include <sys/lock.h>
56 #include <sys/rmlock.h>
57 #include <sys/rwlock.h>
58 #include <sys/sdt.h>
59 #include <sys/syslog.h>
60 #include <sys/sysctl.h>
61 
62 #include <net/if.h>
63 #include <net/if_types.h>
64 #include <net/if_var.h>
65 #include <net/if_dl.h>
66 #include <net/pfil.h>
67 #include <net/route.h>
68 #include <net/route/nhop.h>
69 #include <net/netisr.h>
70 #include <net/rss_config.h>
71 #include <net/vnet.h>
72 
73 #include <netinet/in.h>
74 #include <netinet/in_kdtrace.h>
75 #include <netinet/in_systm.h>
76 #include <netinet/in_var.h>
77 #include <netinet/ip.h>
78 #include <netinet/in_fib.h>
79 #include <netinet/in_pcb.h>
80 #include <netinet/ip_var.h>
81 #include <netinet/ip_encap.h>
82 #include <netinet/ip_fw.h>
83 #include <netinet/ip_icmp.h>
84 #include <netinet/igmp_var.h>
85 #include <netinet/ip_options.h>
86 #include <machine/in_cksum.h>
87 #include <netinet/ip_carp.h>
88 #include <netinet/in_rss.h>
89 #ifdef SCTP
90 #include <netinet/sctp_var.h>
91 #endif
92 
93 #include <netipsec/ipsec_support.h>
94 
95 #include <sys/socketvar.h>
96 
97 #include <security/mac/mac_framework.h>
98 
99 #ifdef CTASSERT
100 CTASSERT(sizeof(struct ip) == 20);
101 #endif
102 
103 /* IP reassembly functions are defined in ip_reass.c. */
104 extern void ipreass_init(void);
105 extern void ipreass_vnet_init(void);
106 #ifdef VIMAGE
107 extern void ipreass_destroy(void);
108 #endif
109 
110 VNET_DEFINE(int, rsvp_on);
111 
112 VNET_DEFINE(int, ipforwarding);
113 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_VNET | CTLFLAG_RW,
114     &VNET_NAME(ipforwarding), 0,
115     "Enable IP forwarding between interfaces");
116 
117 /*
118  * Respond with an ICMP host redirect when we forward a packet out of
119  * the same interface on which it was received.  See RFC 792.
120  */
121 VNET_DEFINE(int, ipsendredirects) = 1;
122 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_VNET | CTLFLAG_RW,
123     &VNET_NAME(ipsendredirects), 0,
124     "Enable sending IP redirects");
125 
126 VNET_DEFINE_STATIC(bool, ip_strong_es) = false;
127 #define	V_ip_strong_es	VNET(ip_strong_es)
128 SYSCTL_BOOL(_net_inet_ip, OID_AUTO, rfc1122_strong_es,
129     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip_strong_es), false,
130     "Packet's IP destination address must match address on arrival interface");
131 
132 VNET_DEFINE_STATIC(bool, ip_sav) = true;
133 #define	V_ip_sav	VNET(ip_sav)
134 SYSCTL_BOOL(_net_inet_ip, OID_AUTO, source_address_validation,
135     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip_sav), true,
136     "Drop incoming packets with source address that is a local address");
137 
138 VNET_DEFINE(pfil_head_t, inet_pfil_head);	/* Packet filter hooks */
139 
140 static struct netisr_handler ip_nh = {
141 	.nh_name = "ip",
142 	.nh_handler = ip_input,
143 	.nh_proto = NETISR_IP,
144 #ifdef	RSS
145 	.nh_m2cpuid = rss_soft_m2cpuid_v4,
146 	.nh_policy = NETISR_POLICY_CPU,
147 	.nh_dispatch = NETISR_DISPATCH_HYBRID,
148 #else
149 	.nh_policy = NETISR_POLICY_FLOW,
150 #endif
151 };
152 
153 #ifdef	RSS
154 /*
155  * Directly dispatched frames are currently assumed
156  * to have a flowid already calculated.
157  *
158  * It should likely have something that assert it
159  * actually has valid flow details.
160  */
161 static struct netisr_handler ip_direct_nh = {
162 	.nh_name = "ip_direct",
163 	.nh_handler = ip_direct_input,
164 	.nh_proto = NETISR_IP_DIRECT,
165 	.nh_m2cpuid = rss_soft_m2cpuid_v4,
166 	.nh_policy = NETISR_POLICY_CPU,
167 	.nh_dispatch = NETISR_DISPATCH_HYBRID,
168 };
169 #endif
170 
171 ipproto_input_t		*ip_protox[IPPROTO_MAX] = {
172 			    [0 ... IPPROTO_MAX - 1] = rip_input };
173 ipproto_ctlinput_t	*ip_ctlprotox[IPPROTO_MAX] = {
174 			    [0 ... IPPROTO_MAX - 1] = rip_ctlinput };
175 
176 VNET_DEFINE(struct in_ifaddrhead, in_ifaddrhead);  /* first inet address */
177 VNET_DEFINE(struct in_ifaddrhashhead *, in_ifaddrhashtbl); /* inet addr hash table  */
178 VNET_DEFINE(u_long, in_ifaddrhmask);		/* mask for hash table */
179 
180 /* Make sure it is safe to use hashinit(9) on CK_LIST. */
181 CTASSERT(sizeof(struct in_ifaddrhashhead) == sizeof(LIST_HEAD(, in_addr)));
182 
183 #ifdef IPCTL_DEFMTU
184 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
185     &ip_mtu, 0, "Default MTU");
186 #endif
187 
188 #ifdef IPSTEALTH
189 VNET_DEFINE(int, ipstealth);
190 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_VNET | CTLFLAG_RW,
191     &VNET_NAME(ipstealth), 0,
192     "IP stealth mode, no TTL decrementation on forwarding");
193 #endif
194 
195 /*
196  * IP statistics are stored in the "array" of counter(9)s.
197  */
198 VNET_PCPUSTAT_DEFINE(struct ipstat, ipstat);
199 VNET_PCPUSTAT_SYSINIT(ipstat);
200 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, IPCTL_STATS, stats, struct ipstat, ipstat,
201     "IP statistics (struct ipstat, netinet/ip_var.h)");
202 
203 #ifdef VIMAGE
204 VNET_PCPUSTAT_SYSUNINIT(ipstat);
205 #endif /* VIMAGE */
206 
207 /*
208  * Kernel module interface for updating ipstat.  The argument is an index
209  * into ipstat treated as an array.
210  */
211 void
212 kmod_ipstat_inc(int statnum)
213 {
214 
215 	counter_u64_add(VNET(ipstat)[statnum], 1);
216 }
217 
218 void
219 kmod_ipstat_dec(int statnum)
220 {
221 
222 	counter_u64_add(VNET(ipstat)[statnum], -1);
223 }
224 
225 static int
226 sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS)
227 {
228 	int error, qlimit;
229 
230 	netisr_getqlimit(&ip_nh, &qlimit);
231 	error = sysctl_handle_int(oidp, &qlimit, 0, req);
232 	if (error || !req->newptr)
233 		return (error);
234 	if (qlimit < 1)
235 		return (EINVAL);
236 	return (netisr_setqlimit(&ip_nh, qlimit));
237 }
238 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen,
239     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 0,
240     sysctl_netinet_intr_queue_maxlen, "I",
241     "Maximum size of the IP input queue");
242 
243 static int
244 sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS)
245 {
246 	u_int64_t qdrops_long;
247 	int error, qdrops;
248 
249 	netisr_getqdrops(&ip_nh, &qdrops_long);
250 	qdrops = qdrops_long;
251 	error = sysctl_handle_int(oidp, &qdrops, 0, req);
252 	if (error || !req->newptr)
253 		return (error);
254 	if (qdrops != 0)
255 		return (EINVAL);
256 	netisr_clearqdrops(&ip_nh);
257 	return (0);
258 }
259 
260 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops,
261     CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE,
262     0, 0, sysctl_netinet_intr_queue_drops, "I",
263     "Number of packets dropped from the IP input queue");
264 
265 #ifdef	RSS
266 static int
267 sysctl_netinet_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS)
268 {
269 	int error, qlimit;
270 
271 	netisr_getqlimit(&ip_direct_nh, &qlimit);
272 	error = sysctl_handle_int(oidp, &qlimit, 0, req);
273 	if (error || !req->newptr)
274 		return (error);
275 	if (qlimit < 1)
276 		return (EINVAL);
277 	return (netisr_setqlimit(&ip_direct_nh, qlimit));
278 }
279 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRDQMAXLEN, intr_direct_queue_maxlen,
280     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
281     0, 0, sysctl_netinet_intr_direct_queue_maxlen,
282     "I", "Maximum size of the IP direct input queue");
283 
284 static int
285 sysctl_netinet_intr_direct_queue_drops(SYSCTL_HANDLER_ARGS)
286 {
287 	u_int64_t qdrops_long;
288 	int error, qdrops;
289 
290 	netisr_getqdrops(&ip_direct_nh, &qdrops_long);
291 	qdrops = qdrops_long;
292 	error = sysctl_handle_int(oidp, &qdrops, 0, req);
293 	if (error || !req->newptr)
294 		return (error);
295 	if (qdrops != 0)
296 		return (EINVAL);
297 	netisr_clearqdrops(&ip_direct_nh);
298 	return (0);
299 }
300 
301 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRDQDROPS, intr_direct_queue_drops,
302     CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, 0, 0,
303     sysctl_netinet_intr_direct_queue_drops, "I",
304     "Number of packets dropped from the IP direct input queue");
305 #endif	/* RSS */
306 
307 /*
308  * IP initialization: fill in IP protocol switch table.
309  * All protocols not implemented in kernel go to raw IP protocol handler.
310  */
311 static void
312 ip_vnet_init(void *arg __unused)
313 {
314 	struct pfil_head_args args;
315 
316 	CK_STAILQ_INIT(&V_in_ifaddrhead);
317 	V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask);
318 
319 	/* Initialize IP reassembly queue. */
320 	ipreass_vnet_init();
321 
322 	/* Initialize packet filter hooks. */
323 	args.pa_version = PFIL_VERSION;
324 	args.pa_flags = PFIL_IN | PFIL_OUT;
325 	args.pa_type = PFIL_TYPE_IP4;
326 	args.pa_headname = PFIL_INET_NAME;
327 	V_inet_pfil_head = pfil_head_register(&args);
328 
329 	if (hhook_head_register(HHOOK_TYPE_IPSEC_IN, AF_INET,
330 	    &V_ipsec_hhh_in[HHOOK_IPSEC_INET],
331 	    HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
332 		printf("%s: WARNING: unable to register input helper hook\n",
333 		    __func__);
334 	if (hhook_head_register(HHOOK_TYPE_IPSEC_OUT, AF_INET,
335 	    &V_ipsec_hhh_out[HHOOK_IPSEC_INET],
336 	    HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
337 		printf("%s: WARNING: unable to register output helper hook\n",
338 		    __func__);
339 
340 #ifdef VIMAGE
341 	netisr_register_vnet(&ip_nh);
342 #ifdef	RSS
343 	netisr_register_vnet(&ip_direct_nh);
344 #endif
345 #endif
346 }
347 VNET_SYSINIT(ip_vnet_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_FOURTH,
348     ip_vnet_init, NULL);
349 
350 static void
351 ip_init(const void *unused __unused)
352 {
353 
354 	ipreass_init();
355 
356 	/*
357 	 * Register statically compiled protocols, that are unlikely to
358 	 * ever become dynamic.
359 	 */
360 	IPPROTO_REGISTER(IPPROTO_ICMP, icmp_input, NULL);
361 	IPPROTO_REGISTER(IPPROTO_IGMP, igmp_input, NULL);
362 	IPPROTO_REGISTER(IPPROTO_RSVP, rsvp_input, NULL);
363 	IPPROTO_REGISTER(IPPROTO_IPV4, encap4_input, NULL);
364 	IPPROTO_REGISTER(IPPROTO_MOBILE, encap4_input, NULL);
365 	IPPROTO_REGISTER(IPPROTO_ETHERIP, encap4_input, NULL);
366 	IPPROTO_REGISTER(IPPROTO_GRE, encap4_input, NULL);
367 	IPPROTO_REGISTER(IPPROTO_IPV6, encap4_input, NULL);
368 	IPPROTO_REGISTER(IPPROTO_PIM, encap4_input, NULL);
369 #ifdef SCTP	/* XXX: has a loadable & static version */
370 	IPPROTO_REGISTER(IPPROTO_SCTP, sctp_input, sctp_ctlinput);
371 #endif
372 
373 	netisr_register(&ip_nh);
374 #ifdef	RSS
375 	netisr_register(&ip_direct_nh);
376 #endif
377 }
378 SYSINIT(ip_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip_init, NULL);
379 
380 #ifdef VIMAGE
381 static void
382 ip_destroy(void *unused __unused)
383 {
384 	int error;
385 
386 #ifdef	RSS
387 	netisr_unregister_vnet(&ip_direct_nh);
388 #endif
389 	netisr_unregister_vnet(&ip_nh);
390 
391 	pfil_head_unregister(V_inet_pfil_head);
392 	error = hhook_head_deregister(V_ipsec_hhh_in[HHOOK_IPSEC_INET]);
393 	if (error != 0) {
394 		printf("%s: WARNING: unable to deregister input helper hook "
395 		    "type HHOOK_TYPE_IPSEC_IN, id HHOOK_IPSEC_INET: "
396 		    "error %d returned\n", __func__, error);
397 	}
398 	error = hhook_head_deregister(V_ipsec_hhh_out[HHOOK_IPSEC_INET]);
399 	if (error != 0) {
400 		printf("%s: WARNING: unable to deregister output helper hook "
401 		    "type HHOOK_TYPE_IPSEC_OUT, id HHOOK_IPSEC_INET: "
402 		    "error %d returned\n", __func__, error);
403 	}
404 
405 	/* Remove the IPv4 addresses from all interfaces. */
406 	in_ifscrub_all();
407 
408 	/* Make sure the IPv4 routes are gone as well. */
409 	rib_flush_routes_family(AF_INET);
410 
411 	/* Destroy IP reassembly queue. */
412 	ipreass_destroy();
413 
414 	/* Cleanup in_ifaddr hash table; should be empty. */
415 	hashdestroy(V_in_ifaddrhashtbl, M_IFADDR, V_in_ifaddrhmask);
416 }
417 
418 VNET_SYSUNINIT(ip, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip_destroy, NULL);
419 #endif
420 
421 #ifdef	RSS
422 /*
423  * IP direct input routine.
424  *
425  * This is called when reinjecting completed fragments where
426  * all of the previous checking and book-keeping has been done.
427  */
428 void
429 ip_direct_input(struct mbuf *m)
430 {
431 	struct ip *ip;
432 	int hlen;
433 
434 	ip = mtod(m, struct ip *);
435 	hlen = ip->ip_hl << 2;
436 
437 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
438 	if (IPSEC_ENABLED(ipv4)) {
439 		if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
440 			return;
441 	}
442 #endif /* IPSEC */
443 	IPSTAT_INC(ips_delivered);
444 	ip_protox[ip->ip_p](&m, &hlen, ip->ip_p);
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 ip *ip = NULL;
456 	struct in_ifaddr *ia = NULL;
457 	struct ifaddr *ifa;
458 	struct ifnet *ifp;
459 	int hlen = 0;
460 	uint16_t sum, ip_len;
461 	int dchg = 0;				/* dest changed after fw */
462 	struct in_addr odst;			/* original dst address */
463 	bool strong_es;
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 (__predict_false(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));
484 		if (__predict_false(m == NULL)) {
485 			IPSTAT_INC(ips_toosmall);
486 			return;
487 		}
488 	}
489 	ip = mtod(m, struct ip *);
490 
491 	if (__predict_false(ip->ip_v != IPVERSION)) {
492 		IPSTAT_INC(ips_badvers);
493 		goto bad;
494 	}
495 
496 	hlen = ip->ip_hl << 2;
497 	if (__predict_false(hlen < sizeof(struct ip))) {	/* minimum header length */
498 		IPSTAT_INC(ips_badhlen);
499 		goto bad;
500 	}
501 	if (hlen > m->m_len) {
502 		m = m_pullup(m, hlen);
503 		if (__predict_false(m == 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 (__predict_false(sum)) {
532 		IPSTAT_INC(ips_badsum);
533 		goto bad;
534 	}
535 
536 	ip_len = ntohs(ip->ip_len);
537 	if (__predict_false(ip_len < hlen)) {
538 		IPSTAT_INC(ips_badlen);
539 		goto bad;
540 	}
541 
542 	/*
543 	 * Check that the amount of data in the buffers
544 	 * is as at least much as the IP header would have us expect.
545 	 * Trim mbufs if longer than we expect.
546 	 * Drop packet if shorter than we expect.
547 	 */
548 	if (__predict_false(m->m_pkthdr.len < ip_len)) {
549 tooshort:
550 		IPSTAT_INC(ips_tooshort);
551 		goto bad;
552 	}
553 	if (m->m_pkthdr.len > ip_len) {
554 		if (m->m_len == m->m_pkthdr.len) {
555 			m->m_len = ip_len;
556 			m->m_pkthdr.len = ip_len;
557 		} else
558 			m_adj(m, ip_len - m->m_pkthdr.len);
559 	}
560 
561 	/*
562 	 * Try to forward the packet, but if we fail continue.
563 	 * ip_tryforward() may generate redirects these days.
564 	 * XXX the logic below falling through to normal processing
565 	 * if redirects are required should be revisited as well.
566 	 * ip_tryforward() does inbound and outbound packet firewall
567 	 * processing. If firewall has decided that destination becomes
568 	 * our local address, it sets M_FASTFWD_OURS flag. In this
569 	 * case skip another inbound firewall processing and update
570 	 * ip pointer.
571 	 */
572 	if (V_ipforwarding != 0
573 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
574 	    && (!IPSEC_ENABLED(ipv4) ||
575 	    IPSEC_CAPS(ipv4, m, IPSEC_CAP_OPERABLE) == 0)
576 #endif
577 	    ) {
578 		/*
579 		 * ip_dooptions() was run so we can ignore the source route (or
580 		 * any IP options case) case for redirects in ip_tryforward().
581 		 */
582 		if ((m = ip_tryforward(m)) == NULL)
583 			return;
584 		if (m->m_flags & M_FASTFWD_OURS) {
585 			m->m_flags &= ~M_FASTFWD_OURS;
586 			ip = mtod(m, struct ip *);
587 			goto ours;
588 		}
589 	}
590 
591 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
592 	/*
593 	 * Bypass packet filtering for packets previously handled by IPsec.
594 	 */
595 	if (IPSEC_ENABLED(ipv4) &&
596 	    IPSEC_CAPS(ipv4, m, IPSEC_CAP_BYPASS_FILTER) != 0)
597 			goto passin;
598 #endif
599 
600 	/*
601 	 * Run through list of hooks for input packets.
602 	 *
603 	 * NB: Beware of the destination address changing (e.g.
604 	 *     by NAT rewriting).  When this happens, tell
605 	 *     ip_forward to do the right thing.
606 	 */
607 
608 	/* Jump over all PFIL processing if hooks are not active. */
609 	if (!PFIL_HOOKED_IN(V_inet_pfil_head))
610 		goto passin;
611 
612 	odst = ip->ip_dst;
613 	if (pfil_mbuf_in(V_inet_pfil_head, &m, ifp, NULL) !=
614 	    PFIL_PASS)
615 		return;
616 	if (m == NULL)			/* consumed by filter */
617 		return;
618 
619 	ip = mtod(m, struct ip *);
620 	dchg = (odst.s_addr != ip->ip_dst.s_addr);
621 
622 	if (m->m_flags & M_FASTFWD_OURS) {
623 		m->m_flags &= ~M_FASTFWD_OURS;
624 		goto ours;
625 	}
626 	if (m->m_flags & M_IP_NEXTHOP) {
627 		if (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) {
628 			/*
629 			 * Directly ship the packet on.  This allows
630 			 * forwarding packets originally destined to us
631 			 * to some other directly connected host.
632 			 */
633 			ip_forward(m, 1);
634 			return;
635 		}
636 	}
637 passin:
638 
639 	/*
640 	 * Process options and, if not destined for us,
641 	 * ship it on.  ip_dooptions returns 1 when an
642 	 * error was detected (causing an icmp message
643 	 * to be sent and the original packet to be freed).
644 	 */
645 	if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
646 		return;
647 
648         /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
649          * matter if it is destined to another node, or whether it is
650          * a multicast one, RSVP wants it! and prevents it from being forwarded
651          * anywhere else. Also checks if the rsvp daemon is running before
652 	 * grabbing the packet.
653          */
654 	if (ip->ip_p == IPPROTO_RSVP && V_rsvp_on)
655 		goto ours;
656 
657 	/*
658 	 * Check our list of addresses, to see if the packet is for us.
659 	 * If we don't have any addresses, assume any unicast packet
660 	 * we receive might be for us (and let the upper layers deal
661 	 * with it).
662 	 */
663 	if (CK_STAILQ_EMPTY(&V_in_ifaddrhead) &&
664 	    (m->m_flags & (M_MCAST|M_BCAST)) == 0)
665 		goto ours;
666 
667 	/*
668 	 * Enable a consistency check between the destination address
669 	 * and the arrival interface for a unicast packet (the RFC 1122
670 	 * strong ES model) with a list of additional predicates:
671 	 * - if IP forwarding is disabled
672 	 * - the packet is not locally generated
673 	 * - the packet is not subject to 'ipfw fwd'
674 	 * - Interface is not running CARP. If the packet got here, we already
675 	 *   checked it with carp_iamatch() and carp_forus().
676 	 */
677 	strong_es = V_ip_strong_es && (V_ipforwarding == 0) &&
678 	    ((ifp->if_flags & IFF_LOOPBACK) == 0) &&
679 	    ifp->if_carp == NULL && (dchg == 0);
680 
681 	/*
682 	 * Check for exact addresses in the hash bucket.
683 	 */
684 	CK_LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
685 		if (IA_SIN(ia)->sin_addr.s_addr != ip->ip_dst.s_addr)
686 			continue;
687 
688 		/*
689 		 * net.inet.ip.rfc1122_strong_es: the address matches, verify
690 		 * that the packet arrived via the correct interface.
691 		 */
692 		if (__predict_false(strong_es && ia->ia_ifp != ifp)) {
693 			IPSTAT_INC(ips_badaddr);
694 			goto bad;
695 		}
696 
697 		/*
698 		 * net.inet.ip.source_address_validation: drop incoming
699 		 * packets that pretend to be ours.
700 		 */
701 		if (V_ip_sav && !(ifp->if_flags & IFF_LOOPBACK) &&
702 		    __predict_false(in_localip_fib(ip->ip_src, ifp->if_fib))) {
703 			IPSTAT_INC(ips_badaddr);
704 			goto bad;
705 		}
706 
707 		counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
708 		counter_u64_add(ia->ia_ifa.ifa_ibytes, m->m_pkthdr.len);
709 		goto ours;
710 	}
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->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 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
744 		/*
745 		 * RFC 3927 2.7: Do not forward multicast packets from
746 		 * IN_LINKLOCAL.
747 		 */
748 		if (V_ip_mrouter && !IN_LINKLOCAL(ntohl(ip->ip_src.s_addr))) {
749 			/*
750 			 * If we are acting as a multicast router, all
751 			 * incoming multicast packets are passed to the
752 			 * kernel-level multicast forwarding function.
753 			 * The packet is returned (relatively) intact; if
754 			 * ip_mforward() returns a non-zero value, the packet
755 			 * must be discarded, else it may be accepted below.
756 			 */
757 			if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) {
758 				IPSTAT_INC(ips_cantforward);
759 				m_freem(m);
760 				return;
761 			}
762 
763 			/*
764 			 * The process-level routing daemon needs to receive
765 			 * all multicast IGMP packets, whether or not this
766 			 * host belongs to their destination groups.
767 			 */
768 			if (ip->ip_p == IPPROTO_IGMP) {
769 				goto ours;
770 			}
771 			IPSTAT_INC(ips_forward);
772 		}
773 		/*
774 		 * Assume the packet is for us, to avoid prematurely taking
775 		 * a lock on the in_multi hash. Protocols must perform
776 		 * their own filtering and update statistics accordingly.
777 		 */
778 		goto ours;
779 	}
780 	if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
781 		goto ours;
782 	if (ip->ip_dst.s_addr == INADDR_ANY)
783 		goto ours;
784 	/* RFC 3927 2.7: Do not forward packets to or from IN_LINKLOCAL. */
785 	if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr)) ||
786 	    IN_LINKLOCAL(ntohl(ip->ip_src.s_addr))) {
787 		IPSTAT_INC(ips_cantforward);
788 		m_freem(m);
789 		return;
790 	}
791 
792 	/*
793 	 * Not for us; forward if possible and desirable.
794 	 */
795 	if (V_ipforwarding == 0) {
796 		IPSTAT_INC(ips_cantforward);
797 		m_freem(m);
798 	} else {
799 		ip_forward(m, dchg);
800 	}
801 	return;
802 
803 ours:
804 #ifdef IPSTEALTH
805 	/*
806 	 * IPSTEALTH: Process non-routing options only
807 	 * if the packet is destined for us.
808 	 */
809 	if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1))
810 		return;
811 #endif /* IPSTEALTH */
812 
813 	/*
814 	 * Attempt reassembly; if it succeeds, proceed.
815 	 * ip_reass() will return a different mbuf.
816 	 */
817 	if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) {
818 		/* XXXGL: shouldn't we save & set m_flags? */
819 		m = ip_reass(m);
820 		if (m == NULL)
821 			return;
822 		ip = mtod(m, struct ip *);
823 		/* Get the header length of the reassembled packet */
824 		hlen = ip->ip_hl << 2;
825 	}
826 
827 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
828 	if (IPSEC_ENABLED(ipv4)) {
829 		if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
830 			return;
831 	}
832 #endif /* IPSEC */
833 
834 	/*
835 	 * Switch out to protocol's input routine.
836 	 */
837 	IPSTAT_INC(ips_delivered);
838 
839 	ip_protox[ip->ip_p](&m, &hlen, ip->ip_p);
840 	return;
841 bad:
842 	m_freem(m);
843 }
844 
845 int
846 ipproto_register(uint8_t proto, ipproto_input_t input, ipproto_ctlinput_t ctl)
847 {
848 
849 	MPASS(proto > 0);
850 
851 	/*
852 	 * The protocol slot must not be occupied by another protocol
853 	 * already.  An index pointing to rip_input() is unused.
854 	 */
855 	if (ip_protox[proto] == rip_input) {
856 		ip_protox[proto] = input;
857 		ip_ctlprotox[proto] = ctl;
858 		return (0);
859 	} else
860 		return (EEXIST);
861 }
862 
863 int
864 ipproto_unregister(uint8_t proto)
865 {
866 
867 	MPASS(proto > 0);
868 
869 	if (ip_protox[proto] != rip_input) {
870 		ip_protox[proto] = rip_input;
871 		ip_ctlprotox[proto] = rip_ctlinput;
872 		return (0);
873 	} else
874 		return (ENOENT);
875 }
876 
877 /*
878  * Forward a packet.  If some error occurs return the sender
879  * an icmp packet.  Note we can't always generate a meaningful
880  * icmp message because icmp doesn't have a large enough repertoire
881  * of codes and types.
882  *
883  * If not forwarding, just drop the packet.  This could be confusing
884  * if ipforwarding was zero but some routing protocol was advancing
885  * us as a gateway to somewhere.  However, we must let the routing
886  * protocol deal with that.
887  *
888  * The srcrt parameter indicates whether the packet is being forwarded
889  * via a source route.
890  */
891 void
892 ip_forward(struct mbuf *m, int srcrt)
893 {
894 	struct ip *ip = mtod(m, struct ip *);
895 	struct in_ifaddr *ia;
896 	struct mbuf *mcopy;
897 	struct sockaddr_in *sin;
898 	struct in_addr dest;
899 	struct route ro;
900 	uint32_t flowid;
901 	int error, type = 0, code = 0, mtu = 0;
902 
903 	NET_EPOCH_ASSERT();
904 
905 	if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
906 		IPSTAT_INC(ips_cantforward);
907 		m_freem(m);
908 		return;
909 	}
910 	if (
911 #ifdef IPSTEALTH
912 	    V_ipstealth == 0 &&
913 #endif
914 	    ip->ip_ttl <= IPTTLDEC) {
915 		icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 0, 0);
916 		return;
917 	}
918 
919 	bzero(&ro, sizeof(ro));
920 	sin = (struct sockaddr_in *)&ro.ro_dst;
921 	sin->sin_family = AF_INET;
922 	sin->sin_len = sizeof(*sin);
923 	sin->sin_addr = ip->ip_dst;
924 	flowid = m->m_pkthdr.flowid;
925 	ro.ro_nh = fib4_lookup(M_GETFIB(m), ip->ip_dst, 0, NHR_REF, flowid);
926 	if (ro.ro_nh != NULL) {
927 		ia = ifatoia(ro.ro_nh->nh_ifa);
928 	} else
929 		ia = NULL;
930 	/*
931 	 * Save the IP header and at most 8 bytes of the payload,
932 	 * in case we need to generate an ICMP message to the src.
933 	 *
934 	 * XXX this can be optimized a lot by saving the data in a local
935 	 * buffer on the stack (72 bytes at most), and only allocating the
936 	 * mbuf if really necessary. The vast majority of the packets
937 	 * are forwarded without having to send an ICMP back (either
938 	 * because unnecessary, or because rate limited), so we are
939 	 * really we are wasting a lot of work here.
940 	 *
941 	 * We don't use m_copym() because it might return a reference
942 	 * to a shared cluster. Both this function and ip_output()
943 	 * assume exclusive access to the IP header in `m', so any
944 	 * data in a cluster may change before we reach icmp_error().
945 	 */
946 	mcopy = m_gethdr(M_NOWAIT, m->m_type);
947 	if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_NOWAIT)) {
948 		/*
949 		 * It's probably ok if the pkthdr dup fails (because
950 		 * the deep copy of the tag chain failed), but for now
951 		 * be conservative and just discard the copy since
952 		 * code below may some day want the tags.
953 		 */
954 		m_free(mcopy);
955 		mcopy = NULL;
956 	}
957 	if (mcopy != NULL) {
958 		mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy));
959 		mcopy->m_pkthdr.len = mcopy->m_len;
960 		m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
961 	}
962 #ifdef IPSTEALTH
963 	if (V_ipstealth == 0)
964 #endif
965 		ip->ip_ttl -= IPTTLDEC;
966 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
967 	if (IPSEC_ENABLED(ipv4)) {
968 		if ((error = IPSEC_FORWARD(ipv4, m)) != 0) {
969 			/* mbuf consumed by IPsec */
970 			RO_NHFREE(&ro);
971 			m_freem(mcopy);
972 			if (error != EINPROGRESS)
973 				IPSTAT_INC(ips_cantforward);
974 			return;
975 		}
976 		/* No IPsec processing required */
977 	}
978 #endif /* IPSEC */
979 	/*
980 	 * If forwarding packet using same interface that it came in on,
981 	 * perhaps should send a redirect to sender to shortcut a hop.
982 	 * Only send redirect if source is sending directly to us,
983 	 * and if packet was not source routed (or has any options).
984 	 * Also, don't send redirect if forwarding using a default route
985 	 * or a route modified by a redirect.
986 	 */
987 	dest.s_addr = 0;
988 	if (!srcrt && V_ipsendredirects &&
989 	    ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) {
990 		struct nhop_object *nh;
991 
992 		nh = ro.ro_nh;
993 
994 		if (nh != NULL && ((nh->nh_flags & (NHF_REDIRECT|NHF_DEFAULT)) == 0)) {
995 			struct in_ifaddr *nh_ia = (struct in_ifaddr *)(nh->nh_ifa);
996 			u_long src = ntohl(ip->ip_src.s_addr);
997 
998 			if (nh_ia != NULL &&
999 			    (src & nh_ia->ia_subnetmask) == nh_ia->ia_subnet) {
1000 				/* Router requirements says to only send host redirects */
1001 				type = ICMP_REDIRECT;
1002 				code = ICMP_REDIRECT_HOST;
1003 				if (nh->nh_flags & NHF_GATEWAY) {
1004 				    if (nh->gw_sa.sa_family == AF_INET)
1005 					dest.s_addr = nh->gw4_sa.sin_addr.s_addr;
1006 				    else /* Do not redirect in case gw is AF_INET6 */
1007 					type = 0;
1008 				} else
1009 					dest.s_addr = ip->ip_dst.s_addr;
1010 			}
1011 		}
1012 	}
1013 
1014 	error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
1015 
1016 	if (error == EMSGSIZE && ro.ro_nh)
1017 		mtu = ro.ro_nh->nh_mtu;
1018 	RO_NHFREE(&ro);
1019 
1020 	if (error)
1021 		IPSTAT_INC(ips_cantforward);
1022 	else {
1023 		IPSTAT_INC(ips_forward);
1024 		if (type)
1025 			IPSTAT_INC(ips_redirectsent);
1026 		else {
1027 			if (mcopy)
1028 				m_freem(mcopy);
1029 			return;
1030 		}
1031 	}
1032 	if (mcopy == NULL)
1033 		return;
1034 
1035 	switch (error) {
1036 	case 0:				/* forwarded, but need redirect */
1037 		/* type, code set above */
1038 		break;
1039 
1040 	case ENETUNREACH:
1041 	case EHOSTUNREACH:
1042 	case ENETDOWN:
1043 	case EHOSTDOWN:
1044 	default:
1045 		type = ICMP_UNREACH;
1046 		code = ICMP_UNREACH_HOST;
1047 		break;
1048 
1049 	case EMSGSIZE:
1050 		type = ICMP_UNREACH;
1051 		code = ICMP_UNREACH_NEEDFRAG;
1052 		/*
1053 		 * If the MTU was set before make sure we are below the
1054 		 * interface MTU.
1055 		 * If the MTU wasn't set before use the interface mtu or
1056 		 * fall back to the next smaller mtu step compared to the
1057 		 * current packet size.
1058 		 */
1059 		if (mtu != 0) {
1060 			if (ia != NULL)
1061 				mtu = min(mtu, ia->ia_ifp->if_mtu);
1062 		} else {
1063 			if (ia != NULL)
1064 				mtu = ia->ia_ifp->if_mtu;
1065 			else
1066 				mtu = ip_next_mtu(ntohs(ip->ip_len), 0);
1067 		}
1068 		IPSTAT_INC(ips_cantfrag);
1069 		break;
1070 
1071 	case ENOBUFS:
1072 	case EACCES:			/* ipfw denied packet */
1073 		m_freem(mcopy);
1074 		return;
1075 	}
1076 	icmp_error(mcopy, type, code, dest.s_addr, mtu);
1077 }
1078 
1079 #define	CHECK_SO_CT(sp, ct) \
1080     (((sp->so_options & SO_TIMESTAMP) && (sp->so_ts_clock == ct)) ? 1 : 0)
1081 
1082 void
1083 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1084     struct mbuf *m)
1085 {
1086 	bool stamped;
1087 
1088 	stamped = false;
1089 	if ((inp->inp_socket->so_options & SO_BINTIME) ||
1090 	    CHECK_SO_CT(inp->inp_socket, SO_TS_BINTIME)) {
1091 		struct bintime boottimebin, bt;
1092 		struct timespec ts1;
1093 
1094 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1095 		    M_TSTMP)) {
1096 			mbuf_tstmp2timespec(m, &ts1);
1097 			timespec2bintime(&ts1, &bt);
1098 			getboottimebin(&boottimebin);
1099 			bintime_add(&bt, &boottimebin);
1100 		} else {
1101 			bintime(&bt);
1102 		}
1103 		*mp = sbcreatecontrol(&bt, sizeof(bt), SCM_BINTIME,
1104 		    SOL_SOCKET, M_NOWAIT);
1105 		if (*mp != NULL) {
1106 			mp = &(*mp)->m_next;
1107 			stamped = true;
1108 		}
1109 	}
1110 	if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME_MICRO)) {
1111 		struct bintime boottimebin, bt1;
1112 		struct timespec ts1;
1113 		struct timeval tv;
1114 
1115 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1116 		    M_TSTMP)) {
1117 			mbuf_tstmp2timespec(m, &ts1);
1118 			timespec2bintime(&ts1, &bt1);
1119 			getboottimebin(&boottimebin);
1120 			bintime_add(&bt1, &boottimebin);
1121 			bintime2timeval(&bt1, &tv);
1122 		} else {
1123 			microtime(&tv);
1124 		}
1125 		*mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv), SCM_TIMESTAMP,
1126 		    SOL_SOCKET, M_NOWAIT);
1127 		if (*mp != NULL) {
1128 			mp = &(*mp)->m_next;
1129 			stamped = true;
1130 		}
1131 	} else if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME)) {
1132 		struct bintime boottimebin;
1133 		struct timespec ts, ts1;
1134 
1135 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1136 		    M_TSTMP)) {
1137 			mbuf_tstmp2timespec(m, &ts);
1138 			getboottimebin(&boottimebin);
1139 			bintime2timespec(&boottimebin, &ts1);
1140 			timespecadd(&ts, &ts1, &ts);
1141 		} else {
1142 			nanotime(&ts);
1143 		}
1144 		*mp = sbcreatecontrol(&ts, sizeof(ts), SCM_REALTIME,
1145 		    SOL_SOCKET, M_NOWAIT);
1146 		if (*mp != NULL) {
1147 			mp = &(*mp)->m_next;
1148 			stamped = true;
1149 		}
1150 	} else if (CHECK_SO_CT(inp->inp_socket, SO_TS_MONOTONIC)) {
1151 		struct timespec ts;
1152 
1153 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1154 		    M_TSTMP))
1155 			mbuf_tstmp2timespec(m, &ts);
1156 		else
1157 			nanouptime(&ts);
1158 		*mp = sbcreatecontrol(&ts, sizeof(ts), SCM_MONOTONIC,
1159 		    SOL_SOCKET, M_NOWAIT);
1160 		if (*mp != NULL) {
1161 			mp = &(*mp)->m_next;
1162 			stamped = true;
1163 		}
1164 	}
1165 	if (stamped && (m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1166 	    M_TSTMP)) {
1167 		struct sock_timestamp_info sti;
1168 
1169 		bzero(&sti, sizeof(sti));
1170 		sti.st_info_flags = ST_INFO_HW;
1171 		if ((m->m_flags & M_TSTMP_HPREC) != 0)
1172 			sti.st_info_flags |= ST_INFO_HW_HPREC;
1173 		*mp = sbcreatecontrol(&sti, sizeof(sti), SCM_TIME_INFO,
1174 		    SOL_SOCKET, M_NOWAIT);
1175 		if (*mp != NULL)
1176 			mp = &(*mp)->m_next;
1177 	}
1178 	if (inp->inp_flags & INP_RECVDSTADDR) {
1179 		*mp = sbcreatecontrol(&ip->ip_dst, sizeof(struct in_addr),
1180 		    IP_RECVDSTADDR, IPPROTO_IP, M_NOWAIT);
1181 		if (*mp)
1182 			mp = &(*mp)->m_next;
1183 	}
1184 	if (inp->inp_flags & INP_RECVTTL) {
1185 		*mp = sbcreatecontrol(&ip->ip_ttl, sizeof(u_char), IP_RECVTTL,
1186 		    IPPROTO_IP, M_NOWAIT);
1187 		if (*mp)
1188 			mp = &(*mp)->m_next;
1189 	}
1190 #ifdef notyet
1191 	/* XXX
1192 	 * Moving these out of udp_input() made them even more broken
1193 	 * than they already were.
1194 	 */
1195 	/* options were tossed already */
1196 	if (inp->inp_flags & INP_RECVOPTS) {
1197 		*mp = sbcreatecontrol(opts_deleted_above,
1198 		    sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP, M_NOWAIT);
1199 		if (*mp)
1200 			mp = &(*mp)->m_next;
1201 	}
1202 	/* ip_srcroute doesn't do what we want here, need to fix */
1203 	if (inp->inp_flags & INP_RECVRETOPTS) {
1204 		*mp = sbcreatecontrol(ip_srcroute(m), sizeof(struct in_addr),
1205 		    IP_RECVRETOPTS, IPPROTO_IP, M_NOWAIT);
1206 		if (*mp)
1207 			mp = &(*mp)->m_next;
1208 	}
1209 #endif
1210 	if (inp->inp_flags & INP_RECVIF) {
1211 		struct ifnet *ifp;
1212 		struct sdlbuf {
1213 			struct sockaddr_dl sdl;
1214 			u_char	pad[32];
1215 		} sdlbuf;
1216 		struct sockaddr_dl *sdp;
1217 		struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1218 
1219 		if ((ifp = m->m_pkthdr.rcvif)) {
1220 			sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1221 			/*
1222 			 * Change our mind and don't try copy.
1223 			 */
1224 			if (sdp->sdl_family != AF_LINK ||
1225 			    sdp->sdl_len > sizeof(sdlbuf)) {
1226 				goto makedummy;
1227 			}
1228 			bcopy(sdp, sdl2, sdp->sdl_len);
1229 		} else {
1230 makedummy:
1231 			sdl2->sdl_len =
1232 			    offsetof(struct sockaddr_dl, sdl_data[0]);
1233 			sdl2->sdl_family = AF_LINK;
1234 			sdl2->sdl_index = 0;
1235 			sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1236 		}
1237 		*mp = sbcreatecontrol(sdl2, sdl2->sdl_len, IP_RECVIF,
1238 		    IPPROTO_IP, M_NOWAIT);
1239 		if (*mp)
1240 			mp = &(*mp)->m_next;
1241 	}
1242 	if (inp->inp_flags & INP_RECVTOS) {
1243 		*mp = sbcreatecontrol(&ip->ip_tos, sizeof(u_char), IP_RECVTOS,
1244 		    IPPROTO_IP, M_NOWAIT);
1245 		if (*mp)
1246 			mp = &(*mp)->m_next;
1247 	}
1248 
1249 	if (inp->inp_flags2 & INP_RECVFLOWID) {
1250 		uint32_t flowid, flow_type;
1251 
1252 		flowid = m->m_pkthdr.flowid;
1253 		flow_type = M_HASHTYPE_GET(m);
1254 
1255 		/*
1256 		 * XXX should handle the failure of one or the
1257 		 * other - don't populate both?
1258 		 */
1259 		*mp = sbcreatecontrol(&flowid, sizeof(uint32_t), IP_FLOWID,
1260 		    IPPROTO_IP, M_NOWAIT);
1261 		if (*mp)
1262 			mp = &(*mp)->m_next;
1263 		*mp = sbcreatecontrol(&flow_type, sizeof(uint32_t),
1264 		    IP_FLOWTYPE, IPPROTO_IP, M_NOWAIT);
1265 		if (*mp)
1266 			mp = &(*mp)->m_next;
1267 	}
1268 
1269 #ifdef	RSS
1270 	if (inp->inp_flags2 & INP_RECVRSSBUCKETID) {
1271 		uint32_t flowid, flow_type;
1272 		uint32_t rss_bucketid;
1273 
1274 		flowid = m->m_pkthdr.flowid;
1275 		flow_type = M_HASHTYPE_GET(m);
1276 
1277 		if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) {
1278 			*mp = sbcreatecontrol(&rss_bucketid, sizeof(uint32_t),
1279 			    IP_RSSBUCKETID, IPPROTO_IP, M_NOWAIT);
1280 			if (*mp)
1281 				mp = &(*mp)->m_next;
1282 		}
1283 	}
1284 #endif
1285 }
1286 
1287 /*
1288  * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
1289  * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
1290  * locking.  This code remains in ip_input.c as ip_mroute.c is optionally
1291  * compiled.
1292  */
1293 VNET_DEFINE_STATIC(int, ip_rsvp_on);
1294 VNET_DEFINE(struct socket *, ip_rsvpd);
1295 
1296 #define	V_ip_rsvp_on		VNET(ip_rsvp_on)
1297 
1298 int
1299 ip_rsvp_init(struct socket *so)
1300 {
1301 
1302 	if (V_ip_rsvpd != NULL)
1303 		return EADDRINUSE;
1304 
1305 	V_ip_rsvpd = so;
1306 	/*
1307 	 * This may seem silly, but we need to be sure we don't over-increment
1308 	 * the RSVP counter, in case something slips up.
1309 	 */
1310 	if (!V_ip_rsvp_on) {
1311 		V_ip_rsvp_on = 1;
1312 		V_rsvp_on++;
1313 	}
1314 
1315 	return 0;
1316 }
1317 
1318 int
1319 ip_rsvp_done(void)
1320 {
1321 
1322 	V_ip_rsvpd = NULL;
1323 	/*
1324 	 * This may seem silly, but we need to be sure we don't over-decrement
1325 	 * the RSVP counter, in case something slips up.
1326 	 */
1327 	if (V_ip_rsvp_on) {
1328 		V_ip_rsvp_on = 0;
1329 		V_rsvp_on--;
1330 	}
1331 	return 0;
1332 }
1333 
1334 int
1335 rsvp_input(struct mbuf **mp, int *offp, int proto)
1336 {
1337 	struct mbuf *m;
1338 
1339 	m = *mp;
1340 	*mp = NULL;
1341 
1342 	if (rsvp_input_p) { /* call the real one if loaded */
1343 		*mp = m;
1344 		rsvp_input_p(mp, offp, proto);
1345 		return (IPPROTO_DONE);
1346 	}
1347 
1348 	/* Can still get packets with rsvp_on = 0 if there is a local member
1349 	 * of the group to which the RSVP packet is addressed.  But in this
1350 	 * case we want to throw the packet away.
1351 	 */
1352 
1353 	if (!V_rsvp_on) {
1354 		m_freem(m);
1355 		return (IPPROTO_DONE);
1356 	}
1357 
1358 	if (V_ip_rsvpd != NULL) {
1359 		*mp = m;
1360 		rip_input(mp, offp, proto);
1361 		return (IPPROTO_DONE);
1362 	}
1363 	/* Drop the packet */
1364 	m_freem(m);
1365 	return (IPPROTO_DONE);
1366 }
1367