xref: /freebsd/sys/netinet/ip_input.c (revision 05248206f720394d95c2a7475429311df670a2e9)
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 #include "opt_bootp.h"
36 #include "opt_inet.h"
37 #include "opt_ipstealth.h"
38 #include "opt_ipsec.h"
39 #include "opt_route.h"
40 #include "opt_rss.h"
41 #include "opt_sctp.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/if_private.h>
65 #include <net/pfil.h>
66 #include <net/route.h>
67 #include <net/route/nhop.h>
68 #include <net/netisr.h>
69 #include <net/rss_config.h>
70 #include <net/vnet.h>
71 
72 #include <netinet/in.h>
73 #include <netinet/in_kdtrace.h>
74 #include <netinet/in_systm.h>
75 #include <netinet/in_var.h>
76 #include <netinet/ip.h>
77 #include <netinet/in_fib.h>
78 #include <netinet/in_pcb.h>
79 #include <netinet/ip_var.h>
80 #include <netinet/ip_encap.h>
81 #include <netinet/ip_fw.h>
82 #include <netinet/ip_icmp.h>
83 #include <netinet/igmp_var.h>
84 #include <netinet/ip_options.h>
85 #include <machine/in_cksum.h>
86 #include <netinet/ip_carp.h>
87 #include <netinet/in_rss.h>
88 #ifdef SCTP
89 #include <netinet/sctp_var.h>
90 #endif
91 
92 #include <netipsec/ipsec_support.h>
93 
94 #include <sys/socketvar.h>
95 
96 #include <security/mac/mac_framework.h>
97 
98 #ifdef CTASSERT
99 CTASSERT(sizeof(struct ip) == 20);
100 #endif
101 
102 /* IP reassembly functions are defined in ip_reass.c. */
103 extern void ipreass_init(void);
104 extern void ipreass_vnet_init(void);
105 #ifdef VIMAGE
106 extern void ipreass_destroy(void);
107 #endif
108 
109 VNET_DEFINE(int, rsvp_on);
110 
111 VNET_DEFINE(int, ipforwarding);
112 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_VNET | CTLFLAG_RW,
113     &VNET_NAME(ipforwarding), 0,
114     "Enable IP forwarding between interfaces");
115 
116 /*
117  * Respond with an ICMP host redirect when we forward a packet out of
118  * the same interface on which it was received.  See RFC 792.
119  */
120 VNET_DEFINE(int, ipsendredirects) = 1;
121 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_VNET | CTLFLAG_RW,
122     &VNET_NAME(ipsendredirects), 0,
123     "Enable sending IP redirects");
124 
125 VNET_DEFINE_STATIC(bool, ip_strong_es) = false;
126 #define	V_ip_strong_es	VNET(ip_strong_es)
127 SYSCTL_BOOL(_net_inet_ip, OID_AUTO, rfc1122_strong_es,
128     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip_strong_es), false,
129     "Packet's IP destination address must match address on arrival interface");
130 
131 VNET_DEFINE_STATIC(bool, ip_sav) = true;
132 #define	V_ip_sav	VNET(ip_sav)
133 SYSCTL_BOOL(_net_inet_ip, OID_AUTO, source_address_validation,
134     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip_sav), true,
135     "Drop incoming packets with source address that is a local address");
136 
137 /* Packet filter hooks */
138 VNET_DEFINE(pfil_head_t, inet_pfil_head);
139 VNET_DEFINE(pfil_head_t, inet_local_pfil_head);
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 ipproto_input_t		*ip_protox[IPPROTO_MAX] = {
173 			    [0 ... IPPROTO_MAX - 1] = rip_input };
174 ipproto_ctlinput_t	*ip_ctlprotox[IPPROTO_MAX] = {
175 			    [0 ... IPPROTO_MAX - 1] = rip_ctlinput };
176 
177 VNET_DEFINE(struct in_ifaddrhead, in_ifaddrhead);  /* first inet address */
178 VNET_DEFINE(struct in_ifaddrhashhead *, in_ifaddrhashtbl); /* inet addr hash table  */
179 VNET_DEFINE(u_long, in_ifaddrhmask);		/* mask for hash table */
180 
181 /* Make sure it is safe to use hashinit(9) on CK_LIST. */
182 CTASSERT(sizeof(struct in_ifaddrhashhead) == sizeof(LIST_HEAD(, in_addr)));
183 
184 #ifdef IPCTL_DEFMTU
185 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
186     &ip_mtu, 0, "Default MTU");
187 #endif
188 
189 #ifdef IPSTEALTH
190 VNET_DEFINE(int, ipstealth);
191 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_VNET | CTLFLAG_RW,
192     &VNET_NAME(ipstealth), 0,
193     "IP stealth mode, no TTL decrementation on forwarding");
194 #endif
195 
196 /*
197  * IP statistics are stored in the "array" of counter(9)s.
198  */
199 VNET_PCPUSTAT_DEFINE(struct ipstat, ipstat);
200 VNET_PCPUSTAT_SYSINIT(ipstat);
201 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, IPCTL_STATS, stats, struct ipstat, ipstat,
202     "IP statistics (struct ipstat, netinet/ip_var.h)");
203 
204 #ifdef VIMAGE
205 VNET_PCPUSTAT_SYSUNINIT(ipstat);
206 #endif /* VIMAGE */
207 
208 /*
209  * Kernel module interface for updating ipstat.  The argument is an index
210  * into ipstat treated as an array.
211  */
212 void
213 kmod_ipstat_inc(int statnum)
214 {
215 
216 	counter_u64_add(VNET(ipstat)[statnum], 1);
217 }
218 
219 void
220 kmod_ipstat_dec(int statnum)
221 {
222 
223 	counter_u64_add(VNET(ipstat)[statnum], -1);
224 }
225 
226 static int
227 sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS)
228 {
229 	int error, qlimit;
230 
231 	netisr_getqlimit(&ip_nh, &qlimit);
232 	error = sysctl_handle_int(oidp, &qlimit, 0, req);
233 	if (error || !req->newptr)
234 		return (error);
235 	if (qlimit < 1)
236 		return (EINVAL);
237 	return (netisr_setqlimit(&ip_nh, qlimit));
238 }
239 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen,
240     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 0,
241     sysctl_netinet_intr_queue_maxlen, "I",
242     "Maximum size of the IP input queue");
243 
244 static int
245 sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS)
246 {
247 	u_int64_t qdrops_long;
248 	int error, qdrops;
249 
250 	netisr_getqdrops(&ip_nh, &qdrops_long);
251 	qdrops = qdrops_long;
252 	error = sysctl_handle_int(oidp, &qdrops, 0, req);
253 	if (error || !req->newptr)
254 		return (error);
255 	if (qdrops != 0)
256 		return (EINVAL);
257 	netisr_clearqdrops(&ip_nh);
258 	return (0);
259 }
260 
261 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops,
262     CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE,
263     0, 0, sysctl_netinet_intr_queue_drops, "I",
264     "Number of packets dropped from the IP input queue");
265 
266 #ifdef	RSS
267 static int
268 sysctl_netinet_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS)
269 {
270 	int error, qlimit;
271 
272 	netisr_getqlimit(&ip_direct_nh, &qlimit);
273 	error = sysctl_handle_int(oidp, &qlimit, 0, req);
274 	if (error || !req->newptr)
275 		return (error);
276 	if (qlimit < 1)
277 		return (EINVAL);
278 	return (netisr_setqlimit(&ip_direct_nh, qlimit));
279 }
280 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRDQMAXLEN, intr_direct_queue_maxlen,
281     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
282     0, 0, sysctl_netinet_intr_direct_queue_maxlen,
283     "I", "Maximum size of the IP direct input queue");
284 
285 static int
286 sysctl_netinet_intr_direct_queue_drops(SYSCTL_HANDLER_ARGS)
287 {
288 	u_int64_t qdrops_long;
289 	int error, qdrops;
290 
291 	netisr_getqdrops(&ip_direct_nh, &qdrops_long);
292 	qdrops = qdrops_long;
293 	error = sysctl_handle_int(oidp, &qdrops, 0, req);
294 	if (error || !req->newptr)
295 		return (error);
296 	if (qdrops != 0)
297 		return (EINVAL);
298 	netisr_clearqdrops(&ip_direct_nh);
299 	return (0);
300 }
301 
302 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRDQDROPS, intr_direct_queue_drops,
303     CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, 0, 0,
304     sysctl_netinet_intr_direct_queue_drops, "I",
305     "Number of packets dropped from the IP direct input queue");
306 #endif	/* RSS */
307 
308 /*
309  * IP initialization: fill in IP protocol switch table.
310  * All protocols not implemented in kernel go to raw IP protocol handler.
311  */
312 static void
313 ip_vnet_init(void *arg __unused)
314 {
315 	struct pfil_head_args args;
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_vnet_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 	args.pa_flags = PFIL_OUT;
331 	args.pa_headname = PFIL_INET_LOCAL_NAME;
332 	V_inet_local_pfil_head = pfil_head_register(&args);
333 
334 	if (hhook_head_register(HHOOK_TYPE_IPSEC_IN, AF_INET,
335 	    &V_ipsec_hhh_in[HHOOK_IPSEC_INET],
336 	    HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
337 		printf("%s: WARNING: unable to register input helper hook\n",
338 		    __func__);
339 	if (hhook_head_register(HHOOK_TYPE_IPSEC_OUT, AF_INET,
340 	    &V_ipsec_hhh_out[HHOOK_IPSEC_INET],
341 	    HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
342 		printf("%s: WARNING: unable to register output helper hook\n",
343 		    __func__);
344 
345 #ifdef VIMAGE
346 	netisr_register_vnet(&ip_nh);
347 #ifdef	RSS
348 	netisr_register_vnet(&ip_direct_nh);
349 #endif
350 #endif
351 }
352 VNET_SYSINIT(ip_vnet_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_FOURTH,
353     ip_vnet_init, NULL);
354 
355 static void
356 ip_init(const void *unused __unused)
357 {
358 
359 	ipreass_init();
360 
361 	/*
362 	 * Register statically compiled protocols, that are unlikely to
363 	 * ever become dynamic.
364 	 */
365 	IPPROTO_REGISTER(IPPROTO_ICMP, icmp_input, NULL);
366 	IPPROTO_REGISTER(IPPROTO_IGMP, igmp_input, NULL);
367 	IPPROTO_REGISTER(IPPROTO_RSVP, rsvp_input, NULL);
368 	IPPROTO_REGISTER(IPPROTO_IPV4, encap4_input, NULL);
369 	IPPROTO_REGISTER(IPPROTO_MOBILE, encap4_input, NULL);
370 	IPPROTO_REGISTER(IPPROTO_ETHERIP, encap4_input, NULL);
371 	IPPROTO_REGISTER(IPPROTO_GRE, encap4_input, NULL);
372 	IPPROTO_REGISTER(IPPROTO_IPV6, encap4_input, NULL);
373 	IPPROTO_REGISTER(IPPROTO_PIM, encap4_input, NULL);
374 #ifdef SCTP	/* XXX: has a loadable & static version */
375 	IPPROTO_REGISTER(IPPROTO_SCTP, sctp_input, sctp_ctlinput);
376 #endif
377 
378 	netisr_register(&ip_nh);
379 #ifdef	RSS
380 	netisr_register(&ip_direct_nh);
381 #endif
382 }
383 SYSINIT(ip_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip_init, NULL);
384 
385 #ifdef VIMAGE
386 static void
387 ip_destroy(void *unused __unused)
388 {
389 	int error;
390 
391 #ifdef	RSS
392 	netisr_unregister_vnet(&ip_direct_nh);
393 #endif
394 	netisr_unregister_vnet(&ip_nh);
395 
396 	pfil_head_unregister(V_inet_pfil_head);
397 	error = hhook_head_deregister(V_ipsec_hhh_in[HHOOK_IPSEC_INET]);
398 	if (error != 0) {
399 		printf("%s: WARNING: unable to deregister input helper hook "
400 		    "type HHOOK_TYPE_IPSEC_IN, id HHOOK_IPSEC_INET: "
401 		    "error %d returned\n", __func__, error);
402 	}
403 	error = hhook_head_deregister(V_ipsec_hhh_out[HHOOK_IPSEC_INET]);
404 	if (error != 0) {
405 		printf("%s: WARNING: unable to deregister output helper hook "
406 		    "type HHOOK_TYPE_IPSEC_OUT, id HHOOK_IPSEC_INET: "
407 		    "error %d returned\n", __func__, error);
408 	}
409 
410 	/* Remove the IPv4 addresses from all interfaces. */
411 	in_ifscrub_all();
412 
413 	/* Make sure the IPv4 routes are gone as well. */
414 	rib_flush_routes_family(AF_INET);
415 
416 	/* Destroy IP reassembly queue. */
417 	ipreass_destroy();
418 
419 	/* Cleanup in_ifaddr hash table; should be empty. */
420 	hashdestroy(V_in_ifaddrhashtbl, M_IFADDR, V_in_ifaddrhmask);
421 }
422 
423 VNET_SYSUNINIT(ip, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip_destroy, NULL);
424 #endif
425 
426 #ifdef	RSS
427 /*
428  * IP direct input routine.
429  *
430  * This is called when reinjecting completed fragments where
431  * all of the previous checking and book-keeping has been done.
432  */
433 void
434 ip_direct_input(struct mbuf *m)
435 {
436 	struct ip *ip;
437 	int hlen;
438 
439 	ip = mtod(m, struct ip *);
440 	hlen = ip->ip_hl << 2;
441 
442 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
443 	if (IPSEC_ENABLED(ipv4)) {
444 		if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
445 			return;
446 	}
447 #endif /* IPSEC */
448 	IPSTAT_INC(ips_delivered);
449 	ip_protox[ip->ip_p](&m, &hlen, ip->ip_p);
450 }
451 #endif
452 
453 /*
454  * Ip input routine.  Checksum and byte swap header.  If fragmented
455  * try to reassemble.  Process options.  Pass to next level.
456  */
457 void
458 ip_input(struct mbuf *m)
459 {
460 	struct ip *ip = NULL;
461 	struct in_ifaddr *ia = NULL;
462 	struct ifaddr *ifa;
463 	struct ifnet *ifp;
464 	int hlen = 0;
465 	uint16_t sum, ip_len;
466 	int dchg = 0;				/* dest changed after fw */
467 	struct in_addr odst;			/* original dst address */
468 	bool strong_es;
469 
470 	M_ASSERTPKTHDR(m);
471 	NET_EPOCH_ASSERT();
472 
473 	if (m->m_flags & M_FASTFWD_OURS) {
474 		m->m_flags &= ~M_FASTFWD_OURS;
475 		/* Set up some basics that will be used later. */
476 		ip = mtod(m, struct ip *);
477 		hlen = ip->ip_hl << 2;
478 		ip_len = ntohs(ip->ip_len);
479 		goto ours;
480 	}
481 
482 	IPSTAT_INC(ips_total);
483 
484 	if (__predict_false(m->m_pkthdr.len < sizeof(struct ip)))
485 		goto tooshort;
486 
487 	if (m->m_len < sizeof(struct ip)) {
488 		m = m_pullup(m, sizeof(struct ip));
489 		if (__predict_false(m == NULL)) {
490 			IPSTAT_INC(ips_toosmall);
491 			return;
492 		}
493 	}
494 	ip = mtod(m, struct ip *);
495 
496 	if (__predict_false(ip->ip_v != IPVERSION)) {
497 		IPSTAT_INC(ips_badvers);
498 		goto bad;
499 	}
500 
501 	hlen = ip->ip_hl << 2;
502 	if (__predict_false(hlen < sizeof(struct ip))) {	/* minimum header length */
503 		IPSTAT_INC(ips_badhlen);
504 		goto bad;
505 	}
506 	if (hlen > m->m_len) {
507 		m = m_pullup(m, hlen);
508 		if (__predict_false(m == NULL)) {
509 			IPSTAT_INC(ips_badhlen);
510 			return;
511 		}
512 		ip = mtod(m, struct ip *);
513 	}
514 
515 	IP_PROBE(receive, NULL, NULL, ip, m->m_pkthdr.rcvif, ip, NULL);
516 
517 	/* IN_LOOPBACK must not appear on the wire - RFC1122 */
518 	ifp = m->m_pkthdr.rcvif;
519 	if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) ||
520 	    IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) {
521 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
522 			IPSTAT_INC(ips_badaddr);
523 			goto bad;
524 		}
525 	}
526 	/* The unspecified address can appear only as a src address - RFC1122 */
527 	if (__predict_false(ntohl(ip->ip_dst.s_addr) == INADDR_ANY)) {
528 		IPSTAT_INC(ips_badaddr);
529 		goto bad;
530 	}
531 
532 	if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
533 		sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
534 	} else {
535 		if (hlen == sizeof(struct ip)) {
536 			sum = in_cksum_hdr(ip);
537 		} else {
538 			sum = in_cksum(m, hlen);
539 		}
540 	}
541 	if (__predict_false(sum)) {
542 		IPSTAT_INC(ips_badsum);
543 		goto bad;
544 	}
545 
546 	ip_len = ntohs(ip->ip_len);
547 	if (__predict_false(ip_len < hlen)) {
548 		IPSTAT_INC(ips_badlen);
549 		goto bad;
550 	}
551 
552 	/*
553 	 * Check that the amount of data in the buffers
554 	 * is as at least much as the IP header would have us expect.
555 	 * Trim mbufs if longer than we expect.
556 	 * Drop packet if shorter than we expect.
557 	 */
558 	if (__predict_false(m->m_pkthdr.len < ip_len)) {
559 tooshort:
560 		IPSTAT_INC(ips_tooshort);
561 		goto bad;
562 	}
563 	if (m->m_pkthdr.len > ip_len) {
564 		if (m->m_len == m->m_pkthdr.len) {
565 			m->m_len = ip_len;
566 			m->m_pkthdr.len = ip_len;
567 		} else
568 			m_adj(m, ip_len - m->m_pkthdr.len);
569 	}
570 
571 	/*
572 	 * Try to forward the packet, but if we fail continue.
573 	 * ip_tryforward() may generate redirects these days.
574 	 * XXX the logic below falling through to normal processing
575 	 * if redirects are required should be revisited as well.
576 	 * ip_tryforward() does inbound and outbound packet firewall
577 	 * processing. If firewall has decided that destination becomes
578 	 * our local address, it sets M_FASTFWD_OURS flag. In this
579 	 * case skip another inbound firewall processing and update
580 	 * ip pointer.
581 	 */
582 	if (V_ipforwarding != 0
583 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
584 	    && (!IPSEC_ENABLED(ipv4) ||
585 	    IPSEC_CAPS(ipv4, m, IPSEC_CAP_OPERABLE) == 0)
586 #endif
587 	    ) {
588 		/*
589 		 * ip_dooptions() was run so we can ignore the source route (or
590 		 * any IP options case) case for redirects in ip_tryforward().
591 		 */
592 		if ((m = ip_tryforward(m)) == NULL)
593 			return;
594 		if (m->m_flags & M_FASTFWD_OURS) {
595 			m->m_flags &= ~M_FASTFWD_OURS;
596 			ip = mtod(m, struct ip *);
597 			goto ours;
598 		}
599 	}
600 
601 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
602 	/*
603 	 * Bypass packet filtering for packets previously handled by IPsec.
604 	 */
605 	if (IPSEC_ENABLED(ipv4) &&
606 	    IPSEC_CAPS(ipv4, m, IPSEC_CAP_BYPASS_FILTER) != 0)
607 			goto passin;
608 #endif
609 
610 	/*
611 	 * Run through list of hooks for input packets.
612 	 *
613 	 * NB: Beware of the destination address changing (e.g.
614 	 *     by NAT rewriting).  When this happens, tell
615 	 *     ip_forward to do the right thing.
616 	 */
617 
618 	/* Jump over all PFIL processing if hooks are not active. */
619 	if (!PFIL_HOOKED_IN(V_inet_pfil_head))
620 		goto passin;
621 
622 	odst = ip->ip_dst;
623 	if (pfil_mbuf_in(V_inet_pfil_head, &m, ifp, NULL) !=
624 	    PFIL_PASS)
625 		return;
626 	if (m == NULL)			/* consumed by filter */
627 		return;
628 
629 	ip = mtod(m, struct ip *);
630 	dchg = (odst.s_addr != ip->ip_dst.s_addr);
631 
632 	if (m->m_flags & M_FASTFWD_OURS) {
633 		m->m_flags &= ~M_FASTFWD_OURS;
634 		goto ours;
635 	}
636 	if (m->m_flags & M_IP_NEXTHOP) {
637 		if (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) {
638 			/*
639 			 * Directly ship the packet on.  This allows
640 			 * forwarding packets originally destined to us
641 			 * to some other directly connected host.
642 			 */
643 			ip_forward(m, 1);
644 			return;
645 		}
646 	}
647 passin:
648 
649 	/*
650 	 * Process options and, if not destined for us,
651 	 * ship it on.  ip_dooptions returns 1 when an
652 	 * error was detected (causing an icmp message
653 	 * to be sent and the original packet to be freed).
654 	 */
655 	if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
656 		return;
657 
658         /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
659          * matter if it is destined to another node, or whether it is
660          * a multicast one, RSVP wants it! and prevents it from being forwarded
661          * anywhere else. Also checks if the rsvp daemon is running before
662 	 * grabbing the packet.
663          */
664 	if (ip->ip_p == IPPROTO_RSVP && V_rsvp_on)
665 		goto ours;
666 
667 	/*
668 	 * Check our list of addresses, to see if the packet is for us.
669 	 * If we don't have any addresses, assume any unicast packet
670 	 * we receive might be for us (and let the upper layers deal
671 	 * with it).
672 	 */
673 	if (CK_STAILQ_EMPTY(&V_in_ifaddrhead) &&
674 	    (m->m_flags & (M_MCAST|M_BCAST)) == 0)
675 		goto ours;
676 
677 	/*
678 	 * Enable a consistency check between the destination address
679 	 * and the arrival interface for a unicast packet (the RFC 1122
680 	 * strong ES model) with a list of additional predicates:
681 	 * - if IP forwarding is disabled
682 	 * - the packet is not locally generated
683 	 * - the packet is not subject to 'ipfw fwd'
684 	 * - Interface is not running CARP. If the packet got here, we already
685 	 *   checked it with carp_iamatch() and carp_forus().
686 	 */
687 	strong_es = V_ip_strong_es && (V_ipforwarding == 0) &&
688 	    ((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 	CK_LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
695 		if (IA_SIN(ia)->sin_addr.s_addr != ip->ip_dst.s_addr)
696 			continue;
697 
698 		/*
699 		 * net.inet.ip.rfc1122_strong_es: the address matches, verify
700 		 * that the packet arrived via the correct interface.
701 		 */
702 		if (__predict_false(strong_es && ia->ia_ifp != ifp)) {
703 			IPSTAT_INC(ips_badaddr);
704 			goto bad;
705 		}
706 
707 		/*
708 		 * net.inet.ip.source_address_validation: drop incoming
709 		 * packets that pretend to be ours.
710 		 */
711 		if (V_ip_sav && !(ifp->if_flags & IFF_LOOPBACK) &&
712 		    __predict_false(in_localip_fib(ip->ip_src, ifp->if_fib))) {
713 			IPSTAT_INC(ips_badaddr);
714 			goto bad;
715 		}
716 
717 		counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
718 		counter_u64_add(ia->ia_ifa.ifa_ibytes, m->m_pkthdr.len);
719 		goto ours;
720 	}
721 
722 	/*
723 	 * Check for broadcast addresses.
724 	 *
725 	 * Only accept broadcast packets that arrive via the matching
726 	 * interface.  Reception of forwarded directed broadcasts would
727 	 * be handled via ip_forward() and ether_output() with the loopback
728 	 * into the stack for SIMPLEX interfaces handled by ether_output().
729 	 */
730 	if (ifp->if_flags & IFF_BROADCAST) {
731 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
732 			if (ifa->ifa_addr->sa_family != AF_INET)
733 				continue;
734 			ia = ifatoia(ifa);
735 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
736 			    ip->ip_dst.s_addr) {
737 				counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
738 				counter_u64_add(ia->ia_ifa.ifa_ibytes,
739 				    m->m_pkthdr.len);
740 				goto ours;
741 			}
742 #ifdef BOOTP_COMPAT
743 			if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) {
744 				counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
745 				counter_u64_add(ia->ia_ifa.ifa_ibytes,
746 				    m->m_pkthdr.len);
747 				goto ours;
748 			}
749 #endif
750 		}
751 		ia = NULL;
752 	}
753 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
754 		/*
755 		 * RFC 3927 2.7: Do not forward multicast packets from
756 		 * IN_LINKLOCAL.
757 		 */
758 		if (V_ip_mrouter && !IN_LINKLOCAL(ntohl(ip->ip_src.s_addr))) {
759 			/*
760 			 * If we are acting as a multicast router, all
761 			 * incoming multicast packets are passed to the
762 			 * kernel-level multicast forwarding function.
763 			 * The packet is returned (relatively) intact; if
764 			 * ip_mforward() returns a non-zero value, the packet
765 			 * must be discarded, else it may be accepted below.
766 			 */
767 			if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) {
768 				IPSTAT_INC(ips_cantforward);
769 				m_freem(m);
770 				return;
771 			}
772 
773 			/*
774 			 * The process-level routing daemon needs to receive
775 			 * all multicast IGMP packets, whether or not this
776 			 * host belongs to their destination groups.
777 			 */
778 			if (ip->ip_p == IPPROTO_IGMP) {
779 				goto ours;
780 			}
781 			IPSTAT_INC(ips_forward);
782 		}
783 		/*
784 		 * Assume the packet is for us, to avoid prematurely taking
785 		 * a lock on the in_multi hash. Protocols must perform
786 		 * their own filtering and update statistics accordingly.
787 		 */
788 		goto ours;
789 	}
790 	if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
791 		goto ours;
792 	if (ip->ip_dst.s_addr == INADDR_ANY)
793 		goto ours;
794 	/* RFC 3927 2.7: Do not forward packets to or from IN_LINKLOCAL. */
795 	if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr)) ||
796 	    IN_LINKLOCAL(ntohl(ip->ip_src.s_addr))) {
797 		IPSTAT_INC(ips_cantforward);
798 		m_freem(m);
799 		return;
800 	}
801 
802 	/*
803 	 * Not for us; forward if possible and desirable.
804 	 */
805 	if (V_ipforwarding == 0) {
806 		IPSTAT_INC(ips_cantforward);
807 		m_freem(m);
808 	} else {
809 		ip_forward(m, dchg);
810 	}
811 	return;
812 
813 ours:
814 #ifdef IPSTEALTH
815 	/*
816 	 * IPSTEALTH: Process non-routing options only
817 	 * if the packet is destined for us.
818 	 */
819 	if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1))
820 		return;
821 #endif /* IPSTEALTH */
822 
823 	/*
824 	 * We are going to ship the packet to the local protocol stack. Call the
825 	 * filter again for this 'output' action, allowing redirect-like rules
826 	 * to adjust the source address.
827 	 */
828 	if (PFIL_HOOKED_OUT(V_inet_local_pfil_head)) {
829 		if (pfil_mbuf_out(V_inet_local_pfil_head, &m, V_loif, NULL) !=
830 		    PFIL_PASS)
831 			return;
832 		if (m == NULL)			/* consumed by filter */
833 			return;
834 		ip = mtod(m, struct ip *);
835 	}
836 
837 	/*
838 	 * Attempt reassembly; if it succeeds, proceed.
839 	 * ip_reass() will return a different mbuf.
840 	 */
841 	if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) {
842 		/* XXXGL: shouldn't we save & set m_flags? */
843 		m = ip_reass(m);
844 		if (m == NULL)
845 			return;
846 		ip = mtod(m, struct ip *);
847 		/* Get the header length of the reassembled packet */
848 		hlen = ip->ip_hl << 2;
849 	}
850 
851 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
852 	if (IPSEC_ENABLED(ipv4)) {
853 		if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
854 			return;
855 	}
856 #endif /* IPSEC */
857 
858 	/*
859 	 * Switch out to protocol's input routine.
860 	 */
861 	IPSTAT_INC(ips_delivered);
862 
863 	ip_protox[ip->ip_p](&m, &hlen, ip->ip_p);
864 	return;
865 bad:
866 	m_freem(m);
867 }
868 
869 int
870 ipproto_register(uint8_t proto, ipproto_input_t input, ipproto_ctlinput_t ctl)
871 {
872 
873 	MPASS(proto > 0);
874 
875 	/*
876 	 * The protocol slot must not be occupied by another protocol
877 	 * already.  An index pointing to rip_input() is unused.
878 	 */
879 	if (ip_protox[proto] == rip_input) {
880 		ip_protox[proto] = input;
881 		ip_ctlprotox[proto] = ctl;
882 		return (0);
883 	} else
884 		return (EEXIST);
885 }
886 
887 int
888 ipproto_unregister(uint8_t proto)
889 {
890 
891 	MPASS(proto > 0);
892 
893 	if (ip_protox[proto] != rip_input) {
894 		ip_protox[proto] = rip_input;
895 		ip_ctlprotox[proto] = rip_ctlinput;
896 		return (0);
897 	} else
898 		return (ENOENT);
899 }
900 
901 /*
902  * Forward a packet.  If some error occurs return the sender
903  * an icmp packet.  Note we can't always generate a meaningful
904  * icmp message because icmp doesn't have a large enough repertoire
905  * of codes and types.
906  *
907  * If not forwarding, just drop the packet.  This could be confusing
908  * if ipforwarding was zero but some routing protocol was advancing
909  * us as a gateway to somewhere.  However, we must let the routing
910  * protocol deal with that.
911  *
912  * The srcrt parameter indicates whether the packet is being forwarded
913  * via a source route.
914  */
915 void
916 ip_forward(struct mbuf *m, int srcrt)
917 {
918 	struct ip *ip = mtod(m, struct ip *);
919 	struct in_ifaddr *ia;
920 	struct mbuf *mcopy;
921 	struct sockaddr_in *sin;
922 	struct in_addr dest;
923 	struct route ro;
924 	uint32_t flowid;
925 	int error, type = 0, code = 0, mtu = 0;
926 
927 	NET_EPOCH_ASSERT();
928 
929 	if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
930 		IPSTAT_INC(ips_cantforward);
931 		m_freem(m);
932 		return;
933 	}
934 	if (
935 #ifdef IPSTEALTH
936 	    V_ipstealth == 0 &&
937 #endif
938 	    ip->ip_ttl <= IPTTLDEC) {
939 		icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 0, 0);
940 		return;
941 	}
942 
943 	bzero(&ro, sizeof(ro));
944 	sin = (struct sockaddr_in *)&ro.ro_dst;
945 	sin->sin_family = AF_INET;
946 	sin->sin_len = sizeof(*sin);
947 	sin->sin_addr = ip->ip_dst;
948 	flowid = m->m_pkthdr.flowid;
949 	ro.ro_nh = fib4_lookup(M_GETFIB(m), ip->ip_dst, 0, NHR_REF, flowid);
950 	if (ro.ro_nh != NULL) {
951 		ia = ifatoia(ro.ro_nh->nh_ifa);
952 	} else
953 		ia = NULL;
954 	/*
955 	 * Save the IP header and at most 8 bytes of the payload,
956 	 * in case we need to generate an ICMP message to the src.
957 	 *
958 	 * XXX this can be optimized a lot by saving the data in a local
959 	 * buffer on the stack (72 bytes at most), and only allocating the
960 	 * mbuf if really necessary. The vast majority of the packets
961 	 * are forwarded without having to send an ICMP back (either
962 	 * because unnecessary, or because rate limited), so we are
963 	 * really we are wasting a lot of work here.
964 	 *
965 	 * We don't use m_copym() because it might return a reference
966 	 * to a shared cluster. Both this function and ip_output()
967 	 * assume exclusive access to the IP header in `m', so any
968 	 * data in a cluster may change before we reach icmp_error().
969 	 */
970 	mcopy = m_gethdr(M_NOWAIT, m->m_type);
971 	if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_NOWAIT)) {
972 		/*
973 		 * It's probably ok if the pkthdr dup fails (because
974 		 * the deep copy of the tag chain failed), but for now
975 		 * be conservative and just discard the copy since
976 		 * code below may some day want the tags.
977 		 */
978 		m_free(mcopy);
979 		mcopy = NULL;
980 	}
981 	if (mcopy != NULL) {
982 		mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy));
983 		mcopy->m_pkthdr.len = mcopy->m_len;
984 		m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
985 	}
986 #ifdef IPSTEALTH
987 	if (V_ipstealth == 0)
988 #endif
989 		ip->ip_ttl -= IPTTLDEC;
990 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
991 	if (IPSEC_ENABLED(ipv4)) {
992 		if ((error = IPSEC_FORWARD(ipv4, m)) != 0) {
993 			/* mbuf consumed by IPsec */
994 			RO_NHFREE(&ro);
995 			m_freem(mcopy);
996 			if (error != EINPROGRESS)
997 				IPSTAT_INC(ips_cantforward);
998 			return;
999 		}
1000 		/* No IPsec processing required */
1001 	}
1002 #endif /* IPSEC */
1003 	/*
1004 	 * If forwarding packet using same interface that it came in on,
1005 	 * perhaps should send a redirect to sender to shortcut a hop.
1006 	 * Only send redirect if source is sending directly to us,
1007 	 * and if packet was not source routed (or has any options).
1008 	 * Also, don't send redirect if forwarding using a default route
1009 	 * or a route modified by a redirect.
1010 	 */
1011 	dest.s_addr = 0;
1012 	if (!srcrt && V_ipsendredirects &&
1013 	    ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) {
1014 		struct nhop_object *nh;
1015 
1016 		nh = ro.ro_nh;
1017 
1018 		if (nh != NULL && ((nh->nh_flags & (NHF_REDIRECT|NHF_DEFAULT)) == 0)) {
1019 			struct in_ifaddr *nh_ia = (struct in_ifaddr *)(nh->nh_ifa);
1020 			u_long src = ntohl(ip->ip_src.s_addr);
1021 
1022 			if (nh_ia != NULL &&
1023 			    (src & nh_ia->ia_subnetmask) == nh_ia->ia_subnet) {
1024 				/* Router requirements says to only send host redirects */
1025 				type = ICMP_REDIRECT;
1026 				code = ICMP_REDIRECT_HOST;
1027 				if (nh->nh_flags & NHF_GATEWAY) {
1028 				    if (nh->gw_sa.sa_family == AF_INET)
1029 					dest.s_addr = nh->gw4_sa.sin_addr.s_addr;
1030 				    else /* Do not redirect in case gw is AF_INET6 */
1031 					type = 0;
1032 				} else
1033 					dest.s_addr = ip->ip_dst.s_addr;
1034 			}
1035 		}
1036 	}
1037 
1038 	error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
1039 
1040 	if (error == EMSGSIZE && ro.ro_nh)
1041 		mtu = ro.ro_nh->nh_mtu;
1042 	RO_NHFREE(&ro);
1043 
1044 	if (error)
1045 		IPSTAT_INC(ips_cantforward);
1046 	else {
1047 		IPSTAT_INC(ips_forward);
1048 		if (type)
1049 			IPSTAT_INC(ips_redirectsent);
1050 		else {
1051 			if (mcopy)
1052 				m_freem(mcopy);
1053 			return;
1054 		}
1055 	}
1056 	if (mcopy == NULL)
1057 		return;
1058 
1059 	switch (error) {
1060 	case 0:				/* forwarded, but need redirect */
1061 		/* type, code set above */
1062 		break;
1063 
1064 	case ENETUNREACH:
1065 	case EHOSTUNREACH:
1066 	case ENETDOWN:
1067 	case EHOSTDOWN:
1068 	default:
1069 		type = ICMP_UNREACH;
1070 		code = ICMP_UNREACH_HOST;
1071 		break;
1072 
1073 	case EMSGSIZE:
1074 		type = ICMP_UNREACH;
1075 		code = ICMP_UNREACH_NEEDFRAG;
1076 		/*
1077 		 * If the MTU was set before make sure we are below the
1078 		 * interface MTU.
1079 		 * If the MTU wasn't set before use the interface mtu or
1080 		 * fall back to the next smaller mtu step compared to the
1081 		 * current packet size.
1082 		 */
1083 		if (mtu != 0) {
1084 			if (ia != NULL)
1085 				mtu = min(mtu, ia->ia_ifp->if_mtu);
1086 		} else {
1087 			if (ia != NULL)
1088 				mtu = ia->ia_ifp->if_mtu;
1089 			else
1090 				mtu = ip_next_mtu(ntohs(ip->ip_len), 0);
1091 		}
1092 		IPSTAT_INC(ips_cantfrag);
1093 		break;
1094 
1095 	case ENOBUFS:
1096 	case EACCES:			/* ipfw denied packet */
1097 		m_freem(mcopy);
1098 		return;
1099 	}
1100 	icmp_error(mcopy, type, code, dest.s_addr, mtu);
1101 }
1102 
1103 #define	CHECK_SO_CT(sp, ct) \
1104     (((sp->so_options & SO_TIMESTAMP) && (sp->so_ts_clock == ct)) ? 1 : 0)
1105 
1106 void
1107 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1108     struct mbuf *m)
1109 {
1110 	bool stamped;
1111 
1112 	stamped = false;
1113 	if ((inp->inp_socket->so_options & SO_BINTIME) ||
1114 	    CHECK_SO_CT(inp->inp_socket, SO_TS_BINTIME)) {
1115 		struct bintime boottimebin, bt;
1116 		struct timespec ts1;
1117 
1118 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1119 		    M_TSTMP)) {
1120 			mbuf_tstmp2timespec(m, &ts1);
1121 			timespec2bintime(&ts1, &bt);
1122 			getboottimebin(&boottimebin);
1123 			bintime_add(&bt, &boottimebin);
1124 		} else {
1125 			bintime(&bt);
1126 		}
1127 		*mp = sbcreatecontrol(&bt, sizeof(bt), SCM_BINTIME,
1128 		    SOL_SOCKET, M_NOWAIT);
1129 		if (*mp != NULL) {
1130 			mp = &(*mp)->m_next;
1131 			stamped = true;
1132 		}
1133 	}
1134 	if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME_MICRO)) {
1135 		struct bintime boottimebin, bt1;
1136 		struct timespec ts1;
1137 		struct timeval tv;
1138 
1139 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1140 		    M_TSTMP)) {
1141 			mbuf_tstmp2timespec(m, &ts1);
1142 			timespec2bintime(&ts1, &bt1);
1143 			getboottimebin(&boottimebin);
1144 			bintime_add(&bt1, &boottimebin);
1145 			bintime2timeval(&bt1, &tv);
1146 		} else {
1147 			microtime(&tv);
1148 		}
1149 		*mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv), SCM_TIMESTAMP,
1150 		    SOL_SOCKET, M_NOWAIT);
1151 		if (*mp != NULL) {
1152 			mp = &(*mp)->m_next;
1153 			stamped = true;
1154 		}
1155 	} else if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME)) {
1156 		struct bintime boottimebin;
1157 		struct timespec ts, ts1;
1158 
1159 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1160 		    M_TSTMP)) {
1161 			mbuf_tstmp2timespec(m, &ts);
1162 			getboottimebin(&boottimebin);
1163 			bintime2timespec(&boottimebin, &ts1);
1164 			timespecadd(&ts, &ts1, &ts);
1165 		} else {
1166 			nanotime(&ts);
1167 		}
1168 		*mp = sbcreatecontrol(&ts, sizeof(ts), SCM_REALTIME,
1169 		    SOL_SOCKET, M_NOWAIT);
1170 		if (*mp != NULL) {
1171 			mp = &(*mp)->m_next;
1172 			stamped = true;
1173 		}
1174 	} else if (CHECK_SO_CT(inp->inp_socket, SO_TS_MONOTONIC)) {
1175 		struct timespec ts;
1176 
1177 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1178 		    M_TSTMP))
1179 			mbuf_tstmp2timespec(m, &ts);
1180 		else
1181 			nanouptime(&ts);
1182 		*mp = sbcreatecontrol(&ts, sizeof(ts), SCM_MONOTONIC,
1183 		    SOL_SOCKET, M_NOWAIT);
1184 		if (*mp != NULL) {
1185 			mp = &(*mp)->m_next;
1186 			stamped = true;
1187 		}
1188 	}
1189 	if (stamped && (m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1190 	    M_TSTMP)) {
1191 		struct sock_timestamp_info sti;
1192 
1193 		bzero(&sti, sizeof(sti));
1194 		sti.st_info_flags = ST_INFO_HW;
1195 		if ((m->m_flags & M_TSTMP_HPREC) != 0)
1196 			sti.st_info_flags |= ST_INFO_HW_HPREC;
1197 		*mp = sbcreatecontrol(&sti, sizeof(sti), SCM_TIME_INFO,
1198 		    SOL_SOCKET, M_NOWAIT);
1199 		if (*mp != NULL)
1200 			mp = &(*mp)->m_next;
1201 	}
1202 	if (inp->inp_flags & INP_RECVDSTADDR) {
1203 		*mp = sbcreatecontrol(&ip->ip_dst, sizeof(struct in_addr),
1204 		    IP_RECVDSTADDR, IPPROTO_IP, M_NOWAIT);
1205 		if (*mp)
1206 			mp = &(*mp)->m_next;
1207 	}
1208 	if (inp->inp_flags & INP_RECVTTL) {
1209 		*mp = sbcreatecontrol(&ip->ip_ttl, sizeof(u_char), IP_RECVTTL,
1210 		    IPPROTO_IP, M_NOWAIT);
1211 		if (*mp)
1212 			mp = &(*mp)->m_next;
1213 	}
1214 #ifdef notyet
1215 	/* XXX
1216 	 * Moving these out of udp_input() made them even more broken
1217 	 * than they already were.
1218 	 */
1219 	/* options were tossed already */
1220 	if (inp->inp_flags & INP_RECVOPTS) {
1221 		*mp = sbcreatecontrol(opts_deleted_above,
1222 		    sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP, M_NOWAIT);
1223 		if (*mp)
1224 			mp = &(*mp)->m_next;
1225 	}
1226 	/* ip_srcroute doesn't do what we want here, need to fix */
1227 	if (inp->inp_flags & INP_RECVRETOPTS) {
1228 		*mp = sbcreatecontrol(ip_srcroute(m), sizeof(struct in_addr),
1229 		    IP_RECVRETOPTS, IPPROTO_IP, M_NOWAIT);
1230 		if (*mp)
1231 			mp = &(*mp)->m_next;
1232 	}
1233 #endif
1234 	if (inp->inp_flags & INP_RECVIF) {
1235 		struct ifnet *ifp;
1236 		struct sdlbuf {
1237 			struct sockaddr_dl sdl;
1238 			u_char	pad[32];
1239 		} sdlbuf;
1240 		struct sockaddr_dl *sdp;
1241 		struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1242 
1243 		if ((ifp = m->m_pkthdr.rcvif)) {
1244 			sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1245 			/*
1246 			 * Change our mind and don't try copy.
1247 			 */
1248 			if (sdp->sdl_family != AF_LINK ||
1249 			    sdp->sdl_len > sizeof(sdlbuf)) {
1250 				goto makedummy;
1251 			}
1252 			bcopy(sdp, sdl2, sdp->sdl_len);
1253 		} else {
1254 makedummy:
1255 			sdl2->sdl_len =
1256 			    offsetof(struct sockaddr_dl, sdl_data[0]);
1257 			sdl2->sdl_family = AF_LINK;
1258 			sdl2->sdl_index = 0;
1259 			sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1260 		}
1261 		*mp = sbcreatecontrol(sdl2, sdl2->sdl_len, IP_RECVIF,
1262 		    IPPROTO_IP, M_NOWAIT);
1263 		if (*mp)
1264 			mp = &(*mp)->m_next;
1265 	}
1266 	if (inp->inp_flags & INP_RECVTOS) {
1267 		*mp = sbcreatecontrol(&ip->ip_tos, sizeof(u_char), IP_RECVTOS,
1268 		    IPPROTO_IP, M_NOWAIT);
1269 		if (*mp)
1270 			mp = &(*mp)->m_next;
1271 	}
1272 
1273 	if (inp->inp_flags2 & INP_RECVFLOWID) {
1274 		uint32_t flowid, flow_type;
1275 
1276 		flowid = m->m_pkthdr.flowid;
1277 		flow_type = M_HASHTYPE_GET(m);
1278 
1279 		/*
1280 		 * XXX should handle the failure of one or the
1281 		 * other - don't populate both?
1282 		 */
1283 		*mp = sbcreatecontrol(&flowid, sizeof(uint32_t), IP_FLOWID,
1284 		    IPPROTO_IP, M_NOWAIT);
1285 		if (*mp)
1286 			mp = &(*mp)->m_next;
1287 		*mp = sbcreatecontrol(&flow_type, sizeof(uint32_t),
1288 		    IP_FLOWTYPE, IPPROTO_IP, M_NOWAIT);
1289 		if (*mp)
1290 			mp = &(*mp)->m_next;
1291 	}
1292 
1293 #ifdef	RSS
1294 	if (inp->inp_flags2 & INP_RECVRSSBUCKETID) {
1295 		uint32_t flowid, flow_type;
1296 		uint32_t rss_bucketid;
1297 
1298 		flowid = m->m_pkthdr.flowid;
1299 		flow_type = M_HASHTYPE_GET(m);
1300 
1301 		if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) {
1302 			*mp = sbcreatecontrol(&rss_bucketid, sizeof(uint32_t),
1303 			    IP_RSSBUCKETID, IPPROTO_IP, M_NOWAIT);
1304 			if (*mp)
1305 				mp = &(*mp)->m_next;
1306 		}
1307 	}
1308 #endif
1309 }
1310 
1311 /*
1312  * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
1313  * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
1314  * locking.  This code remains in ip_input.c as ip_mroute.c is optionally
1315  * compiled.
1316  */
1317 VNET_DEFINE_STATIC(int, ip_rsvp_on);
1318 VNET_DEFINE(struct socket *, ip_rsvpd);
1319 
1320 #define	V_ip_rsvp_on		VNET(ip_rsvp_on)
1321 
1322 int
1323 ip_rsvp_init(struct socket *so)
1324 {
1325 
1326 	if (V_ip_rsvpd != NULL)
1327 		return EADDRINUSE;
1328 
1329 	V_ip_rsvpd = so;
1330 	/*
1331 	 * This may seem silly, but we need to be sure we don't over-increment
1332 	 * the RSVP counter, in case something slips up.
1333 	 */
1334 	if (!V_ip_rsvp_on) {
1335 		V_ip_rsvp_on = 1;
1336 		V_rsvp_on++;
1337 	}
1338 
1339 	return 0;
1340 }
1341 
1342 int
1343 ip_rsvp_done(void)
1344 {
1345 
1346 	V_ip_rsvpd = NULL;
1347 	/*
1348 	 * This may seem silly, but we need to be sure we don't over-decrement
1349 	 * the RSVP counter, in case something slips up.
1350 	 */
1351 	if (V_ip_rsvp_on) {
1352 		V_ip_rsvp_on = 0;
1353 		V_rsvp_on--;
1354 	}
1355 	return 0;
1356 }
1357 
1358 int
1359 rsvp_input(struct mbuf **mp, int *offp, int proto)
1360 {
1361 	struct mbuf *m;
1362 
1363 	m = *mp;
1364 	*mp = NULL;
1365 
1366 	if (rsvp_input_p) { /* call the real one if loaded */
1367 		*mp = m;
1368 		rsvp_input_p(mp, offp, proto);
1369 		return (IPPROTO_DONE);
1370 	}
1371 
1372 	/* Can still get packets with rsvp_on = 0 if there is a local member
1373 	 * of the group to which the RSVP packet is addressed.  But in this
1374 	 * case we want to throw the packet away.
1375 	 */
1376 
1377 	if (!V_rsvp_on) {
1378 		m_freem(m);
1379 		return (IPPROTO_DONE);
1380 	}
1381 
1382 	if (V_ip_rsvpd != NULL) {
1383 		*mp = m;
1384 		rip_input(mp, offp, proto);
1385 		return (IPPROTO_DONE);
1386 	}
1387 	/* Drop the packet */
1388 	m_freem(m);
1389 	return (IPPROTO_DONE);
1390 }
1391