xref: /freebsd/sys/netinet/ip_input.c (revision 32cd3ee5901ea33d41ff550e5f40ce743c8d4165)
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 
32 #include "opt_bootp.h"
33 #include "opt_inet.h"
34 #include "opt_ipstealth.h"
35 #include "opt_ipsec.h"
36 #include "opt_route.h"
37 #include "opt_rss.h"
38 #include "opt_sctp.h"
39 
40 #include <sys/param.h>
41 #include <sys/systm.h>
42 #include <sys/hhook.h>
43 #include <sys/mbuf.h>
44 #include <sys/malloc.h>
45 #include <sys/domain.h>
46 #include <sys/protosw.h>
47 #include <sys/socket.h>
48 #include <sys/time.h>
49 #include <sys/kernel.h>
50 #include <sys/lock.h>
51 #include <sys/rmlock.h>
52 #include <sys/rwlock.h>
53 #include <sys/sdt.h>
54 #include <sys/syslog.h>
55 #include <sys/sysctl.h>
56 
57 #include <net/if.h>
58 #include <net/if_types.h>
59 #include <net/if_var.h>
60 #include <net/if_dl.h>
61 #include <net/if_private.h>
62 #include <net/pfil.h>
63 #include <net/route.h>
64 #include <net/route/nhop.h>
65 #include <net/netisr.h>
66 #include <net/rss_config.h>
67 #include <net/vnet.h>
68 
69 #include <netinet/in.h>
70 #include <netinet/in_kdtrace.h>
71 #include <netinet/in_systm.h>
72 #include <netinet/in_var.h>
73 #include <netinet/ip.h>
74 #include <netinet/in_fib.h>
75 #include <netinet/in_pcb.h>
76 #include <netinet/ip_var.h>
77 #include <netinet/ip_encap.h>
78 #include <netinet/ip_fw.h>
79 #include <netinet/ip_icmp.h>
80 #include <netinet/igmp_var.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 #include <netinet/ip_mroute.h>
86 #ifdef SCTP
87 #include <netinet/sctp_var.h>
88 #endif
89 
90 #include <netipsec/ipsec_support.h>
91 
92 #include <sys/socketvar.h>
93 
94 #include <security/mac/mac_framework.h>
95 
96 #ifdef CTASSERT
97 CTASSERT(sizeof(struct ip) == 20);
98 #endif
99 
100 /* IP reassembly functions are defined in ip_reass.c. */
101 extern void ipreass_init(void);
102 extern void ipreass_vnet_init(void);
103 #ifdef VIMAGE
104 extern void ipreass_destroy(void);
105 #endif
106 
107 VNET_DEFINE(int, rsvp_on);
108 
109 VNET_DEFINE(int, ipforwarding);
110 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_VNET | CTLFLAG_RW,
111     &VNET_NAME(ipforwarding), 0,
112     "Enable IP forwarding between interfaces");
113 
114 /*
115  * Respond with an ICMP host redirect when we forward a packet out of
116  * the same interface on which it was received.  See RFC 792.
117  */
118 VNET_DEFINE(int, ipsendredirects) = 1;
119 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_VNET | CTLFLAG_RW,
120     &VNET_NAME(ipsendredirects), 0,
121     "Enable sending IP redirects");
122 
123 VNET_DEFINE_STATIC(bool, ip_strong_es) = false;
124 #define	V_ip_strong_es	VNET(ip_strong_es)
125 SYSCTL_BOOL(_net_inet_ip, OID_AUTO, rfc1122_strong_es,
126     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip_strong_es), false,
127     "Packet's IP destination address must match address on arrival interface");
128 
129 VNET_DEFINE_STATIC(bool, ip_sav) = true;
130 #define	V_ip_sav	VNET(ip_sav)
131 SYSCTL_BOOL(_net_inet_ip, OID_AUTO, source_address_validation,
132     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip_sav), true,
133     "Drop incoming packets with source address that is a local address");
134 
135 /* Packet filter hooks */
136 VNET_DEFINE(pfil_head_t, inet_pfil_head);
137 VNET_DEFINE(pfil_head_t, inet_local_pfil_head);
138 
139 static struct netisr_handler ip_nh = {
140 	.nh_name = "ip",
141 	.nh_handler = ip_input,
142 	.nh_proto = NETISR_IP,
143 #ifdef	RSS
144 	.nh_m2cpuid = rss_soft_m2cpuid_v4,
145 	.nh_policy = NETISR_POLICY_CPU,
146 	.nh_dispatch = NETISR_DISPATCH_HYBRID,
147 #else
148 	.nh_policy = NETISR_POLICY_FLOW,
149 #endif
150 };
151 
152 #ifdef	RSS
153 /*
154  * Directly dispatched frames are currently assumed
155  * to have a flowid already calculated.
156  *
157  * It should likely have something that assert it
158  * actually has valid flow details.
159  */
160 static struct netisr_handler ip_direct_nh = {
161 	.nh_name = "ip_direct",
162 	.nh_handler = ip_direct_input,
163 	.nh_proto = NETISR_IP_DIRECT,
164 	.nh_m2cpuid = rss_soft_m2cpuid_v4,
165 	.nh_policy = NETISR_POLICY_CPU,
166 	.nh_dispatch = NETISR_DISPATCH_HYBRID,
167 };
168 #endif
169 
170 ipproto_input_t		*ip_protox[IPPROTO_MAX] = {
171 			    [0 ... IPPROTO_MAX - 1] = rip_input };
172 ipproto_ctlinput_t	*ip_ctlprotox[IPPROTO_MAX] = {
173 			    [0 ... IPPROTO_MAX - 1] = rip_ctlinput };
174 
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 /* Make sure it is safe to use hashinit(9) on CK_LIST. */
180 CTASSERT(sizeof(struct in_ifaddrhashhead) == sizeof(LIST_HEAD(, in_addr)));
181 
182 #ifdef IPCTL_DEFMTU
183 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
184     &ip_mtu, 0, "Default MTU");
185 #endif
186 
187 #ifdef IPSTEALTH
188 VNET_DEFINE(int, ipstealth);
189 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_VNET | CTLFLAG_RW,
190     &VNET_NAME(ipstealth), 0,
191     "IP stealth mode, no TTL decrementation on forwarding");
192 #endif
193 
194 /*
195  * IP statistics are stored in the "array" of counter(9)s.
196  */
197 VNET_PCPUSTAT_DEFINE(struct ipstat, ipstat);
198 VNET_PCPUSTAT_SYSINIT(ipstat);
199 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, IPCTL_STATS, stats, struct ipstat, ipstat,
200     "IP statistics (struct ipstat, netinet/ip_var.h)");
201 
202 #ifdef VIMAGE
203 VNET_PCPUSTAT_SYSUNINIT(ipstat);
204 #endif /* VIMAGE */
205 
206 /*
207  * Kernel module interface for updating ipstat.  The argument is an index
208  * into ipstat treated as an array.
209  */
210 void
211 kmod_ipstat_inc(int statnum)
212 {
213 
214 	counter_u64_add(VNET(ipstat)[statnum], 1);
215 }
216 
217 void
218 kmod_ipstat_dec(int statnum)
219 {
220 
221 	counter_u64_add(VNET(ipstat)[statnum], -1);
222 }
223 
224 static int
225 sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS)
226 {
227 	int error, qlimit;
228 
229 	netisr_getqlimit(&ip_nh, &qlimit);
230 	error = sysctl_handle_int(oidp, &qlimit, 0, req);
231 	if (error || !req->newptr)
232 		return (error);
233 	if (qlimit < 1)
234 		return (EINVAL);
235 	return (netisr_setqlimit(&ip_nh, qlimit));
236 }
237 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen,
238     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 0,
239     sysctl_netinet_intr_queue_maxlen, "I",
240     "Maximum size of the IP input queue");
241 
242 static int
243 sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS)
244 {
245 	u_int64_t qdrops_long;
246 	int error, qdrops;
247 
248 	netisr_getqdrops(&ip_nh, &qdrops_long);
249 	qdrops = qdrops_long;
250 	error = sysctl_handle_int(oidp, &qdrops, 0, req);
251 	if (error || !req->newptr)
252 		return (error);
253 	if (qdrops != 0)
254 		return (EINVAL);
255 	netisr_clearqdrops(&ip_nh);
256 	return (0);
257 }
258 
259 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops,
260     CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE,
261     0, 0, sysctl_netinet_intr_queue_drops, "I",
262     "Number of packets dropped from the IP input queue");
263 
264 #ifdef	RSS
265 static int
266 sysctl_netinet_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS)
267 {
268 	int error, qlimit;
269 
270 	netisr_getqlimit(&ip_direct_nh, &qlimit);
271 	error = sysctl_handle_int(oidp, &qlimit, 0, req);
272 	if (error || !req->newptr)
273 		return (error);
274 	if (qlimit < 1)
275 		return (EINVAL);
276 	return (netisr_setqlimit(&ip_direct_nh, qlimit));
277 }
278 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRDQMAXLEN, intr_direct_queue_maxlen,
279     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
280     0, 0, sysctl_netinet_intr_direct_queue_maxlen,
281     "I", "Maximum size of the IP direct input queue");
282 
283 static int
284 sysctl_netinet_intr_direct_queue_drops(SYSCTL_HANDLER_ARGS)
285 {
286 	u_int64_t qdrops_long;
287 	int error, qdrops;
288 
289 	netisr_getqdrops(&ip_direct_nh, &qdrops_long);
290 	qdrops = qdrops_long;
291 	error = sysctl_handle_int(oidp, &qdrops, 0, req);
292 	if (error || !req->newptr)
293 		return (error);
294 	if (qdrops != 0)
295 		return (EINVAL);
296 	netisr_clearqdrops(&ip_direct_nh);
297 	return (0);
298 }
299 
300 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRDQDROPS, intr_direct_queue_drops,
301     CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, 0, 0,
302     sysctl_netinet_intr_direct_queue_drops, "I",
303     "Number of packets dropped from the IP direct input queue");
304 #endif	/* RSS */
305 
306 /*
307  * IP initialization: fill in IP protocol switch table.
308  * All protocols not implemented in kernel go to raw IP protocol handler.
309  */
310 static void
311 ip_vnet_init(void *arg __unused)
312 {
313 	struct pfil_head_args args;
314 
315 	CK_STAILQ_INIT(&V_in_ifaddrhead);
316 	V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask);
317 
318 	/* Initialize IP reassembly queue. */
319 	ipreass_vnet_init();
320 
321 	/* Initialize packet filter hooks. */
322 	args.pa_version = PFIL_VERSION;
323 	args.pa_flags = PFIL_IN | PFIL_OUT;
324 	args.pa_type = PFIL_TYPE_IP4;
325 	args.pa_headname = PFIL_INET_NAME;
326 	V_inet_pfil_head = pfil_head_register(&args);
327 
328 	args.pa_flags = PFIL_OUT;
329 	args.pa_headname = PFIL_INET_LOCAL_NAME;
330 	V_inet_local_pfil_head = pfil_head_register(&args);
331 
332 	if (hhook_head_register(HHOOK_TYPE_IPSEC_IN, AF_INET,
333 	    &V_ipsec_hhh_in[HHOOK_IPSEC_INET],
334 	    HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
335 		printf("%s: WARNING: unable to register input helper hook\n",
336 		    __func__);
337 	if (hhook_head_register(HHOOK_TYPE_IPSEC_OUT, AF_INET,
338 	    &V_ipsec_hhh_out[HHOOK_IPSEC_INET],
339 	    HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
340 		printf("%s: WARNING: unable to register output helper hook\n",
341 		    __func__);
342 
343 #ifdef VIMAGE
344 	netisr_register_vnet(&ip_nh);
345 #ifdef	RSS
346 	netisr_register_vnet(&ip_direct_nh);
347 #endif
348 #endif
349 }
350 VNET_SYSINIT(ip_vnet_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_FOURTH,
351     ip_vnet_init, NULL);
352 
353 static void
354 ip_init(const void *unused __unused)
355 {
356 	struct ifnet *ifp;
357 
358 	ipreass_init();
359 
360 	/*
361 	 * Register statically compiled protocols, that are unlikely to
362 	 * ever become dynamic.
363 	 */
364 	IPPROTO_REGISTER(IPPROTO_ICMP, icmp_input, NULL);
365 	IPPROTO_REGISTER(IPPROTO_IGMP, igmp_input, NULL);
366 	IPPROTO_REGISTER(IPPROTO_RSVP, rsvp_input, NULL);
367 	IPPROTO_REGISTER(IPPROTO_IPV4, encap4_input, NULL);
368 	IPPROTO_REGISTER(IPPROTO_MOBILE, encap4_input, NULL);
369 	IPPROTO_REGISTER(IPPROTO_ETHERIP, encap4_input, NULL);
370 	IPPROTO_REGISTER(IPPROTO_GRE, encap4_input, NULL);
371 	IPPROTO_REGISTER(IPPROTO_IPV6, encap4_input, NULL);
372 	IPPROTO_REGISTER(IPPROTO_PIM, encap4_input, NULL);
373 #ifdef SCTP	/* XXX: has a loadable & static version */
374 	IPPROTO_REGISTER(IPPROTO_SCTP, sctp_input, sctp_ctlinput);
375 #endif
376 
377 	netisr_register(&ip_nh);
378 #ifdef	RSS
379 	netisr_register(&ip_direct_nh);
380 #endif
381 	/*
382 	 * XXXGL: we use SYSINIT() here, but go over V_ifnet.  It was the same
383 	 * way before dom_ifattach removal.  This worked because when any
384 	 * non-default vnet is created, there are no interfaces inside.
385 	 * Eventually this needs to be fixed.
386 	 */
387 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link)
388 		in_ifattach(NULL, ifp);
389 }
390 SYSINIT(ip_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip_init, NULL);
391 
392 #ifdef VIMAGE
393 static void
394 ip_destroy(void *unused __unused)
395 {
396 	int error;
397 
398 #ifdef	RSS
399 	netisr_unregister_vnet(&ip_direct_nh);
400 #endif
401 	netisr_unregister_vnet(&ip_nh);
402 
403 	pfil_head_unregister(V_inet_pfil_head);
404 	error = hhook_head_deregister(V_ipsec_hhh_in[HHOOK_IPSEC_INET]);
405 	if (error != 0) {
406 		printf("%s: WARNING: unable to deregister input helper hook "
407 		    "type HHOOK_TYPE_IPSEC_IN, id HHOOK_IPSEC_INET: "
408 		    "error %d returned\n", __func__, error);
409 	}
410 	error = hhook_head_deregister(V_ipsec_hhh_out[HHOOK_IPSEC_INET]);
411 	if (error != 0) {
412 		printf("%s: WARNING: unable to deregister output helper hook "
413 		    "type HHOOK_TYPE_IPSEC_OUT, id HHOOK_IPSEC_INET: "
414 		    "error %d returned\n", __func__, error);
415 	}
416 
417 	/* Remove the IPv4 addresses from all interfaces. */
418 	in_ifscrub_all();
419 
420 	/* Make sure the IPv4 routes are gone as well. */
421 	rib_flush_routes_family(AF_INET);
422 
423 	/* Destroy IP reassembly queue. */
424 	ipreass_destroy();
425 
426 	/* Cleanup in_ifaddr hash table; should be empty. */
427 	hashdestroy(V_in_ifaddrhashtbl, M_IFADDR, V_in_ifaddrhmask);
428 }
429 
430 VNET_SYSUNINIT(ip, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip_destroy, NULL);
431 #endif
432 
433 #ifdef	RSS
434 /*
435  * IP direct input routine.
436  *
437  * This is called when reinjecting completed fragments where
438  * all of the previous checking and book-keeping has been done.
439  */
440 void
441 ip_direct_input(struct mbuf *m)
442 {
443 	struct ip *ip;
444 	int hlen;
445 
446 	ip = mtod(m, struct ip *);
447 	hlen = ip->ip_hl << 2;
448 
449 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
450 	if (IPSEC_ENABLED(ipv4)) {
451 		if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
452 			return;
453 	}
454 #endif /* IPSEC */
455 	IPSTAT_INC(ips_delivered);
456 	ip_protox[ip->ip_p](&m, &hlen, ip->ip_p);
457 }
458 #endif
459 
460 /*
461  * Ip input routine.  Checksum and byte swap header.  If fragmented
462  * try to reassemble.  Process options.  Pass to next level.
463  */
464 void
465 ip_input(struct mbuf *m)
466 {
467 	struct ip *ip = NULL;
468 	struct in_ifaddr *ia = NULL;
469 	struct ifaddr *ifa;
470 	struct ifnet *ifp;
471 	int hlen = 0;
472 	uint16_t sum, ip_len;
473 	int dchg = 0;				/* dest changed after fw */
474 	struct in_addr odst;			/* original dst address */
475 	bool strong_es;
476 
477 	M_ASSERTPKTHDR(m);
478 	NET_EPOCH_ASSERT();
479 
480 	if (m->m_flags & M_FASTFWD_OURS) {
481 		m->m_flags &= ~M_FASTFWD_OURS;
482 		/* Set up some basics that will be used later. */
483 		ip = mtod(m, struct ip *);
484 		hlen = ip->ip_hl << 2;
485 		ip_len = ntohs(ip->ip_len);
486 		goto ours;
487 	}
488 
489 	IPSTAT_INC(ips_total);
490 
491 	if (__predict_false(m->m_pkthdr.len < sizeof(struct ip)))
492 		goto tooshort;
493 
494 	if (m->m_len < sizeof(struct ip)) {
495 		m = m_pullup(m, sizeof(struct ip));
496 		if (__predict_false(m == NULL)) {
497 			IPSTAT_INC(ips_toosmall);
498 			return;
499 		}
500 	}
501 	ip = mtod(m, struct ip *);
502 
503 	if (__predict_false(ip->ip_v != IPVERSION)) {
504 		IPSTAT_INC(ips_badvers);
505 		goto bad;
506 	}
507 
508 	hlen = ip->ip_hl << 2;
509 	if (__predict_false(hlen < sizeof(struct ip))) {	/* minimum header length */
510 		IPSTAT_INC(ips_badhlen);
511 		goto bad;
512 	}
513 	if (hlen > m->m_len) {
514 		m = m_pullup(m, hlen);
515 		if (__predict_false(m == NULL)) {
516 			IPSTAT_INC(ips_badhlen);
517 			return;
518 		}
519 		ip = mtod(m, struct ip *);
520 	}
521 
522 	IP_PROBE(receive, NULL, NULL, ip, m->m_pkthdr.rcvif, ip, NULL);
523 
524 	/* IN_LOOPBACK must not appear on the wire - RFC1122 */
525 	ifp = m->m_pkthdr.rcvif;
526 	if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) ||
527 	    IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) {
528 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
529 			IPSTAT_INC(ips_badaddr);
530 			goto bad;
531 		}
532 	}
533 
534 	if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
535 		sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
536 	} else if (m->m_pkthdr.csum_flags & CSUM_IP) {
537 		/*
538 		 * Packet from local host that offloaded checksum computation.
539 		 * Checksum not required since the packet wasn't on the wire.
540 		 */
541 		sum = 0;
542 	} else {
543 		if (hlen == sizeof(struct ip)) {
544 			sum = in_cksum_hdr(ip);
545 		} else {
546 			sum = in_cksum(m, hlen);
547 		}
548 	}
549 	if (__predict_false(sum)) {
550 		IPSTAT_INC(ips_badsum);
551 		goto bad;
552 	}
553 
554 	ip_len = ntohs(ip->ip_len);
555 	if (__predict_false(ip_len < hlen)) {
556 		IPSTAT_INC(ips_badlen);
557 		goto bad;
558 	}
559 
560 	/*
561 	 * Check that the amount of data in the buffers
562 	 * is as at least much as the IP header would have us expect.
563 	 * Trim mbufs if longer than we expect.
564 	 * Drop packet if shorter than we expect.
565 	 */
566 	if (__predict_false(m->m_pkthdr.len < ip_len)) {
567 tooshort:
568 		IPSTAT_INC(ips_tooshort);
569 		goto bad;
570 	}
571 	if (m->m_pkthdr.len > ip_len) {
572 		if (m->m_len == m->m_pkthdr.len) {
573 			m->m_len = ip_len;
574 			m->m_pkthdr.len = ip_len;
575 		} else
576 			m_adj(m, ip_len - m->m_pkthdr.len);
577 	}
578 
579 	/*
580 	 * Try to forward the packet, but if we fail continue.
581 	 * ip_tryforward() may generate redirects these days.
582 	 * XXX the logic below falling through to normal processing
583 	 * if redirects are required should be revisited as well.
584 	 * ip_tryforward() does inbound and outbound packet firewall
585 	 * processing. If firewall has decided that destination becomes
586 	 * our local address, it sets M_FASTFWD_OURS flag. In this
587 	 * case skip another inbound firewall processing and update
588 	 * ip pointer.
589 	 */
590 	if (V_ipforwarding != 0
591 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
592 	    && (!IPSEC_ENABLED(ipv4) ||
593 	    IPSEC_CAPS(ipv4, m, IPSEC_CAP_OPERABLE) == 0)
594 #endif
595 	    ) {
596 		/*
597 		 * ip_dooptions() was run so we can ignore the source route (or
598 		 * any IP options case) case for redirects in ip_tryforward().
599 		 */
600 		if ((m = ip_tryforward(m)) == NULL)
601 			return;
602 		if (m->m_flags & M_FASTFWD_OURS) {
603 			m->m_flags &= ~M_FASTFWD_OURS;
604 			ip = mtod(m, struct ip *);
605 			goto ours;
606 		}
607 	}
608 
609 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
610 	/*
611 	 * Bypass packet filtering for packets previously handled by IPsec.
612 	 */
613 	if (IPSEC_ENABLED(ipv4) &&
614 	    IPSEC_CAPS(ipv4, m, IPSEC_CAP_BYPASS_FILTER) != 0)
615 			goto passin;
616 #endif
617 
618 	/*
619 	 * Run through list of hooks for input packets.
620 	 *
621 	 * NB: Beware of the destination address changing (e.g.
622 	 *     by NAT rewriting).  When this happens, tell
623 	 *     ip_forward to do the right thing.
624 	 */
625 
626 	/* Jump over all PFIL processing if hooks are not active. */
627 	if (!PFIL_HOOKED_IN(V_inet_pfil_head))
628 		goto passin;
629 
630 	odst = ip->ip_dst;
631 	if (pfil_mbuf_in(V_inet_pfil_head, &m, ifp, NULL) !=
632 	    PFIL_PASS)
633 		return;
634 
635 	ip = mtod(m, struct ip *);
636 	dchg = (odst.s_addr != ip->ip_dst.s_addr);
637 
638 	if (m->m_flags & M_FASTFWD_OURS) {
639 		m->m_flags &= ~M_FASTFWD_OURS;
640 		goto ours;
641 	}
642 	if (m->m_flags & M_IP_NEXTHOP) {
643 		if (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) {
644 			/*
645 			 * Directly ship the packet on.  This allows
646 			 * forwarding packets originally destined to us
647 			 * to some other directly connected host.
648 			 */
649 			ip_forward(m, 1);
650 			return;
651 		}
652 	}
653 passin:
654 	/*
655 	 * The unspecified address can appear only as a src address - RFC1122.
656 	 *
657 	 * The check is deferred to here to give firewalls a chance to block
658 	 * (and log) such packets.  ip_tryforward() will not process such
659 	 * packets.
660 	 */
661 	if (__predict_false(ntohl(ip->ip_dst.s_addr) == INADDR_ANY)) {
662 		IPSTAT_INC(ips_badaddr);
663 		goto bad;
664 	}
665 
666 	/*
667 	 * Process options and, if not destined for us,
668 	 * ship it on.  ip_dooptions returns 1 when an
669 	 * error was detected (causing an icmp message
670 	 * to be sent and the original packet to be freed).
671 	 */
672 	if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
673 		return;
674 
675         /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
676          * matter if it is destined to another node, or whether it is
677          * a multicast one, RSVP wants it! and prevents it from being forwarded
678          * anywhere else. Also checks if the rsvp daemon is running before
679 	 * grabbing the packet.
680          */
681 	if (ip->ip_p == IPPROTO_RSVP && V_rsvp_on)
682 		goto ours;
683 
684 	/*
685 	 * Check our list of addresses, to see if the packet is for us.
686 	 * If we don't have any addresses, assume any unicast packet
687 	 * we receive might be for us (and let the upper layers deal
688 	 * with it).
689 	 */
690 	if (CK_STAILQ_EMPTY(&V_in_ifaddrhead) &&
691 	    (m->m_flags & (M_MCAST|M_BCAST)) == 0)
692 		goto ours;
693 
694 	/*
695 	 * Enable a consistency check between the destination address
696 	 * and the arrival interface for a unicast packet (the RFC 1122
697 	 * strong ES model) with a list of additional predicates:
698 	 * - if IP forwarding is disabled
699 	 * - the packet is not locally generated
700 	 * - the packet is not subject to 'ipfw fwd'
701 	 * - Interface is not running CARP. If the packet got here, we already
702 	 *   checked it with carp_iamatch() and carp_forus().
703 	 */
704 	strong_es = V_ip_strong_es && (V_ipforwarding == 0) &&
705 	    ((ifp->if_flags & IFF_LOOPBACK) == 0) &&
706 	    ifp->if_carp == NULL && (dchg == 0);
707 
708 	/*
709 	 * Check for exact addresses in the hash bucket.
710 	 */
711 	CK_LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
712 		if (IA_SIN(ia)->sin_addr.s_addr != ip->ip_dst.s_addr)
713 			continue;
714 
715 		/*
716 		 * net.inet.ip.rfc1122_strong_es: the address matches, verify
717 		 * that the packet arrived via the correct interface.
718 		 */
719 		if (__predict_false(strong_es && ia->ia_ifp != ifp)) {
720 			IPSTAT_INC(ips_badaddr);
721 			goto bad;
722 		}
723 
724 		/*
725 		 * net.inet.ip.source_address_validation: drop incoming
726 		 * packets that pretend to be ours.
727 		 */
728 		if (V_ip_sav && !(ifp->if_flags & IFF_LOOPBACK) &&
729 		    __predict_false(in_localip_fib(ip->ip_src, ifp->if_fib))) {
730 			IPSTAT_INC(ips_badaddr);
731 			goto bad;
732 		}
733 
734 		counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
735 		counter_u64_add(ia->ia_ifa.ifa_ibytes, m->m_pkthdr.len);
736 		goto ours;
737 	}
738 
739 	/*
740 	 * Check for broadcast addresses.
741 	 *
742 	 * Only accept broadcast packets that arrive via the matching
743 	 * interface.  Reception of forwarded directed broadcasts would
744 	 * be handled via ip_forward() and ether_output() with the loopback
745 	 * into the stack for SIMPLEX interfaces handled by ether_output().
746 	 */
747 	if (ifp->if_flags & IFF_BROADCAST) {
748 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
749 			if (ifa->ifa_addr->sa_family != AF_INET)
750 				continue;
751 			ia = ifatoia(ifa);
752 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
753 			    ip->ip_dst.s_addr) {
754 				counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
755 				counter_u64_add(ia->ia_ifa.ifa_ibytes,
756 				    m->m_pkthdr.len);
757 				goto ours;
758 			}
759 #ifdef BOOTP_COMPAT
760 			if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) {
761 				counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
762 				counter_u64_add(ia->ia_ifa.ifa_ibytes,
763 				    m->m_pkthdr.len);
764 				goto ours;
765 			}
766 #endif
767 		}
768 		ia = NULL;
769 	}
770 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
771 		/*
772 		 * RFC 3927 2.7: Do not forward multicast packets from
773 		 * IN_LINKLOCAL.
774 		 */
775 		if (V_ip_mrouting_enabled &&
776 		    !IN_LINKLOCAL(ntohl(ip->ip_src.s_addr))) {
777 			/*
778 			 * If we are acting as a multicast router, all
779 			 * incoming multicast packets are passed to the
780 			 * kernel-level multicast forwarding function.
781 			 * The packet is returned (relatively) intact; if
782 			 * ip_mforward() returns a non-zero value, the packet
783 			 * must be discarded, else it may be accepted below.
784 			 */
785 			if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) {
786 				IPSTAT_INC(ips_cantforward);
787 				m_freem(m);
788 				return;
789 			}
790 
791 			/*
792 			 * The process-level routing daemon needs to receive
793 			 * all multicast IGMP packets, whether or not this
794 			 * host belongs to their destination groups.
795 			 */
796 			if (ip->ip_p == IPPROTO_IGMP) {
797 				goto ours;
798 			}
799 			IPSTAT_INC(ips_forward);
800 		}
801 		/*
802 		 * Assume the packet is for us, to avoid prematurely taking
803 		 * a lock on the in_multi hash. Protocols must perform
804 		 * their own filtering and update statistics accordingly.
805 		 */
806 		goto ours;
807 	}
808 	if (in_broadcast(ip->ip_dst))
809 		goto ours;
810 	/* RFC 3927 2.7: Do not forward packets to or from IN_LINKLOCAL. */
811 	if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr)) ||
812 	    IN_LINKLOCAL(ntohl(ip->ip_src.s_addr))) {
813 		IPSTAT_INC(ips_cantforward);
814 		m_freem(m);
815 		return;
816 	}
817 
818 	/*
819 	 * Not for us; forward if possible and desirable.
820 	 */
821 	if (V_ipforwarding == 0) {
822 		IPSTAT_INC(ips_cantforward);
823 		m_freem(m);
824 	} else {
825 		ip_forward(m, dchg);
826 	}
827 	return;
828 
829 ours:
830 #ifdef IPSTEALTH
831 	/*
832 	 * IPSTEALTH: Process non-routing options only
833 	 * if the packet is destined for us.
834 	 */
835 	if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1))
836 		return;
837 #endif /* IPSTEALTH */
838 
839 	/*
840 	 * We are going to ship the packet to the local protocol stack. Call the
841 	 * filter again for this 'output' action, allowing redirect-like rules
842 	 * to adjust the source address.
843 	 */
844 	if (PFIL_HOOKED_OUT(V_inet_local_pfil_head)) {
845 		if (pfil_mbuf_out(V_inet_local_pfil_head, &m, V_loif, NULL) !=
846 		    PFIL_PASS)
847 			return;
848 		ip = mtod(m, struct ip *);
849 	}
850 
851 	/*
852 	 * Attempt reassembly; if it succeeds, proceed.
853 	 * ip_reass() will return a different mbuf.
854 	 */
855 	if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) {
856 		/* XXXGL: shouldn't we save & set m_flags? */
857 		m = ip_reass(m);
858 		if (m == NULL)
859 			return;
860 		ip = mtod(m, struct ip *);
861 		/* Get the header length of the reassembled packet */
862 		hlen = ip->ip_hl << 2;
863 	}
864 
865 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
866 	if (IPSEC_ENABLED(ipv4)) {
867 		if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
868 			return;
869 	}
870 #endif /* IPSEC */
871 
872 	/*
873 	 * Switch out to protocol's input routine.
874 	 */
875 	IPSTAT_INC(ips_delivered);
876 
877 	ip_protox[ip->ip_p](&m, &hlen, ip->ip_p);
878 	return;
879 bad:
880 	m_freem(m);
881 }
882 
883 int
884 ipproto_register(uint8_t proto, ipproto_input_t input, ipproto_ctlinput_t ctl)
885 {
886 
887 	MPASS(proto > 0);
888 
889 	/*
890 	 * The protocol slot must not be occupied by another protocol
891 	 * already.  An index pointing to rip_input() is unused.
892 	 */
893 	if (ip_protox[proto] == rip_input) {
894 		ip_protox[proto] = input;
895 		ip_ctlprotox[proto] = ctl;
896 		return (0);
897 	} else
898 		return (EEXIST);
899 }
900 
901 int
902 ipproto_unregister(uint8_t proto)
903 {
904 
905 	MPASS(proto > 0);
906 
907 	if (ip_protox[proto] != rip_input) {
908 		ip_protox[proto] = rip_input;
909 		ip_ctlprotox[proto] = rip_ctlinput;
910 		return (0);
911 	} else
912 		return (ENOENT);
913 }
914 
915 /*
916  * Forward a packet.  If some error occurs return the sender
917  * an icmp packet.  Note we can't always generate a meaningful
918  * icmp message because icmp doesn't have a large enough repertoire
919  * of codes and types.
920  *
921  * If not forwarding, just drop the packet.  This could be confusing
922  * if ipforwarding was zero but some routing protocol was advancing
923  * us as a gateway to somewhere.  However, we must let the routing
924  * protocol deal with that.
925  *
926  * The srcrt parameter indicates whether the packet is being forwarded
927  * via a source route.
928  */
929 void
930 ip_forward(struct mbuf *m, int srcrt)
931 {
932 	struct ip *ip = mtod(m, struct ip *);
933 	struct in_ifaddr *ia;
934 	struct mbuf *mcopy;
935 	struct sockaddr_in *sin;
936 	struct in_addr dest;
937 	struct route ro;
938 	uint32_t flowid;
939 	int error, type = 0, code = 0, mtu = 0;
940 
941 	NET_EPOCH_ASSERT();
942 
943 	if (m->m_flags & (M_BCAST|M_MCAST) || !in_canforward(ip->ip_dst)) {
944 		IPSTAT_INC(ips_cantforward);
945 		m_freem(m);
946 		return;
947 	}
948 	if (
949 #ifdef IPSTEALTH
950 	    V_ipstealth == 0 &&
951 #endif
952 	    ip->ip_ttl <= IPTTLDEC) {
953 		icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 0, 0);
954 		return;
955 	}
956 
957 	bzero(&ro, sizeof(ro));
958 	sin = (struct sockaddr_in *)&ro.ro_dst;
959 	sin->sin_family = AF_INET;
960 	sin->sin_len = sizeof(*sin);
961 	sin->sin_addr = ip->ip_dst;
962 	flowid = m->m_pkthdr.flowid;
963 	ro.ro_nh = fib4_lookup(M_GETFIB(m), ip->ip_dst, 0, NHR_REF, flowid);
964 	if (ro.ro_nh != NULL) {
965 		if (ro.ro_nh->nh_flags & (NHF_BLACKHOLE | NHF_BROADCAST)) {
966 			IPSTAT_INC(ips_cantforward);
967 			m_freem(m);
968 			NH_FREE(ro.ro_nh);
969 			return;
970 		}
971 		if (ro.ro_nh->nh_flags & NHF_REJECT) {
972 			IPSTAT_INC(ips_cantforward);
973 			NH_FREE(ro.ro_nh);
974 			icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
975 			return;
976 		}
977 		ia = ifatoia(ro.ro_nh->nh_ifa);
978 	} else
979 		ia = NULL;
980 	/*
981 	 * Save the IP header and at most 8 bytes of the payload,
982 	 * in case we need to generate an ICMP message to the src.
983 	 *
984 	 * XXX this can be optimized a lot by saving the data in a local
985 	 * buffer on the stack (72 bytes at most), and only allocating the
986 	 * mbuf if really necessary. The vast majority of the packets
987 	 * are forwarded without having to send an ICMP back (either
988 	 * because unnecessary, or because rate limited), so we are
989 	 * really we are wasting a lot of work here.
990 	 *
991 	 * We don't use m_copym() because it might return a reference
992 	 * to a shared cluster. Both this function and ip_output()
993 	 * assume exclusive access to the IP header in `m', so any
994 	 * data in a cluster may change before we reach icmp_error().
995 	 */
996 	mcopy = m_gethdr(M_NOWAIT, m->m_type);
997 	if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_NOWAIT)) {
998 		/*
999 		 * It's probably ok if the pkthdr dup fails (because
1000 		 * the deep copy of the tag chain failed), but for now
1001 		 * be conservative and just discard the copy since
1002 		 * code below may some day want the tags.
1003 		 */
1004 		m_free(mcopy);
1005 		mcopy = NULL;
1006 	}
1007 	if (mcopy != NULL) {
1008 		mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy));
1009 		mcopy->m_pkthdr.len = mcopy->m_len;
1010 		m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1011 	}
1012 #ifdef IPSTEALTH
1013 	if (V_ipstealth == 0)
1014 #endif
1015 		ip->ip_ttl -= IPTTLDEC;
1016 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1017 	if (IPSEC_ENABLED(ipv4)) {
1018 		if ((error = IPSEC_FORWARD(ipv4, m)) != 0) {
1019 			/* mbuf consumed by IPsec */
1020 			RO_NHFREE(&ro);
1021 			m_freem(mcopy);
1022 			if (error != EINPROGRESS)
1023 				IPSTAT_INC(ips_cantforward);
1024 			return;
1025 		}
1026 		/* No IPsec processing required */
1027 	}
1028 #endif /* IPSEC */
1029 	/*
1030 	 * If forwarding packet using same interface that it came in on,
1031 	 * perhaps should send a redirect to sender to shortcut a hop.
1032 	 * Only send redirect if source is sending directly to us,
1033 	 * and if packet was not source routed (or has any options).
1034 	 * Also, don't send redirect if forwarding using a default route
1035 	 * or a route modified by a redirect.
1036 	 */
1037 	dest.s_addr = 0;
1038 	if (!srcrt && V_ipsendredirects &&
1039 	    ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) {
1040 		struct nhop_object *nh;
1041 
1042 		nh = ro.ro_nh;
1043 
1044 		if (nh != NULL && ((nh->nh_flags & (NHF_REDIRECT|NHF_DEFAULT)) == 0)) {
1045 			struct in_ifaddr *nh_ia = (struct in_ifaddr *)(nh->nh_ifa);
1046 			u_long src = ntohl(ip->ip_src.s_addr);
1047 
1048 			if (nh_ia != NULL &&
1049 			    (src & nh_ia->ia_subnetmask) == nh_ia->ia_subnet) {
1050 				/* Router requirements says to only send host redirects */
1051 				type = ICMP_REDIRECT;
1052 				code = ICMP_REDIRECT_HOST;
1053 				if (nh->nh_flags & NHF_GATEWAY) {
1054 				    if (nh->gw_sa.sa_family == AF_INET)
1055 					dest.s_addr = nh->gw4_sa.sin_addr.s_addr;
1056 				    else /* Do not redirect in case gw is AF_INET6 */
1057 					type = 0;
1058 				} else
1059 					dest.s_addr = ip->ip_dst.s_addr;
1060 			}
1061 		}
1062 	}
1063 
1064 	error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
1065 
1066 	if (error == EMSGSIZE && ro.ro_nh)
1067 		mtu = ro.ro_nh->nh_mtu;
1068 	RO_NHFREE(&ro);
1069 
1070 	if (error)
1071 		IPSTAT_INC(ips_cantforward);
1072 	else {
1073 		IPSTAT_INC(ips_forward);
1074 		if (type)
1075 			IPSTAT_INC(ips_redirectsent);
1076 		else {
1077 			if (mcopy)
1078 				m_freem(mcopy);
1079 			return;
1080 		}
1081 	}
1082 	if (mcopy == NULL)
1083 		return;
1084 
1085 	switch (error) {
1086 	case 0:				/* forwarded, but need redirect */
1087 		/* type, code set above */
1088 		break;
1089 
1090 	case ENETUNREACH:
1091 	case EHOSTUNREACH:
1092 	case ENETDOWN:
1093 	case EHOSTDOWN:
1094 	default:
1095 		type = ICMP_UNREACH;
1096 		code = ICMP_UNREACH_HOST;
1097 		break;
1098 
1099 	case EMSGSIZE:
1100 		type = ICMP_UNREACH;
1101 		code = ICMP_UNREACH_NEEDFRAG;
1102 		/*
1103 		 * If the MTU was set before make sure we are below the
1104 		 * interface MTU.
1105 		 * If the MTU wasn't set before use the interface mtu or
1106 		 * fall back to the next smaller mtu step compared to the
1107 		 * current packet size.
1108 		 */
1109 		if (mtu != 0) {
1110 			if (ia != NULL)
1111 				mtu = min(mtu, ia->ia_ifp->if_mtu);
1112 		} else {
1113 			if (ia != NULL)
1114 				mtu = ia->ia_ifp->if_mtu;
1115 			else
1116 				mtu = ip_next_mtu(ntohs(ip->ip_len), 0);
1117 		}
1118 		IPSTAT_INC(ips_cantfrag);
1119 		break;
1120 
1121 	case ENOBUFS:
1122 	case EACCES:			/* ipfw denied packet */
1123 		m_freem(mcopy);
1124 		return;
1125 	}
1126 	icmp_error(mcopy, type, code, dest.s_addr, mtu);
1127 }
1128 
1129 #define	CHECK_SO_CT(sp, ct) \
1130     (((sp->so_options & SO_TIMESTAMP) && (sp->so_ts_clock == ct)) ? 1 : 0)
1131 
1132 void
1133 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1134     struct mbuf *m)
1135 {
1136 	bool stamped;
1137 
1138 	stamped = false;
1139 	if ((inp->inp_socket->so_options & SO_BINTIME) ||
1140 	    CHECK_SO_CT(inp->inp_socket, SO_TS_BINTIME)) {
1141 		struct bintime boottimebin, bt;
1142 		struct timespec ts1;
1143 
1144 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1145 		    M_TSTMP)) {
1146 			mbuf_tstmp2timespec(m, &ts1);
1147 			timespec2bintime(&ts1, &bt);
1148 			getboottimebin(&boottimebin);
1149 			bintime_add(&bt, &boottimebin);
1150 		} else {
1151 			bintime(&bt);
1152 		}
1153 		*mp = sbcreatecontrol(&bt, sizeof(bt), SCM_BINTIME,
1154 		    SOL_SOCKET, M_NOWAIT);
1155 		if (*mp != NULL) {
1156 			mp = &(*mp)->m_next;
1157 			stamped = true;
1158 		}
1159 	}
1160 	if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME_MICRO)) {
1161 		struct bintime boottimebin, bt1;
1162 		struct timespec ts1;
1163 		struct timeval tv;
1164 
1165 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1166 		    M_TSTMP)) {
1167 			mbuf_tstmp2timespec(m, &ts1);
1168 			timespec2bintime(&ts1, &bt1);
1169 			getboottimebin(&boottimebin);
1170 			bintime_add(&bt1, &boottimebin);
1171 			bintime2timeval(&bt1, &tv);
1172 		} else {
1173 			microtime(&tv);
1174 		}
1175 		*mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv), SCM_TIMESTAMP,
1176 		    SOL_SOCKET, M_NOWAIT);
1177 		if (*mp != NULL) {
1178 			mp = &(*mp)->m_next;
1179 			stamped = true;
1180 		}
1181 	} else if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME)) {
1182 		struct bintime boottimebin;
1183 		struct timespec ts, ts1;
1184 
1185 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1186 		    M_TSTMP)) {
1187 			mbuf_tstmp2timespec(m, &ts);
1188 			getboottimebin(&boottimebin);
1189 			bintime2timespec(&boottimebin, &ts1);
1190 			timespecadd(&ts, &ts1, &ts);
1191 		} else {
1192 			nanotime(&ts);
1193 		}
1194 		*mp = sbcreatecontrol(&ts, sizeof(ts), SCM_REALTIME,
1195 		    SOL_SOCKET, M_NOWAIT);
1196 		if (*mp != NULL) {
1197 			mp = &(*mp)->m_next;
1198 			stamped = true;
1199 		}
1200 	} else if (CHECK_SO_CT(inp->inp_socket, SO_TS_MONOTONIC)) {
1201 		struct timespec ts;
1202 
1203 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1204 		    M_TSTMP))
1205 			mbuf_tstmp2timespec(m, &ts);
1206 		else
1207 			nanouptime(&ts);
1208 		*mp = sbcreatecontrol(&ts, sizeof(ts), SCM_MONOTONIC,
1209 		    SOL_SOCKET, M_NOWAIT);
1210 		if (*mp != NULL) {
1211 			mp = &(*mp)->m_next;
1212 			stamped = true;
1213 		}
1214 	}
1215 	if (stamped && (m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1216 	    M_TSTMP)) {
1217 		struct sock_timestamp_info sti;
1218 
1219 		bzero(&sti, sizeof(sti));
1220 		sti.st_info_flags = ST_INFO_HW;
1221 		if ((m->m_flags & M_TSTMP_HPREC) != 0)
1222 			sti.st_info_flags |= ST_INFO_HW_HPREC;
1223 		*mp = sbcreatecontrol(&sti, sizeof(sti), SCM_TIME_INFO,
1224 		    SOL_SOCKET, M_NOWAIT);
1225 		if (*mp != NULL)
1226 			mp = &(*mp)->m_next;
1227 	}
1228 	if (inp->inp_flags & INP_RECVDSTADDR) {
1229 		*mp = sbcreatecontrol(&ip->ip_dst, sizeof(struct in_addr),
1230 		    IP_RECVDSTADDR, IPPROTO_IP, M_NOWAIT);
1231 		if (*mp)
1232 			mp = &(*mp)->m_next;
1233 	}
1234 	if (inp->inp_flags & INP_RECVTTL) {
1235 		*mp = sbcreatecontrol(&ip->ip_ttl, sizeof(u_char), IP_RECVTTL,
1236 		    IPPROTO_IP, M_NOWAIT);
1237 		if (*mp)
1238 			mp = &(*mp)->m_next;
1239 	}
1240 #ifdef notyet
1241 	/* XXX
1242 	 * Moving these out of udp_input() made them even more broken
1243 	 * than they already were.
1244 	 */
1245 	/* options were tossed already */
1246 	if (inp->inp_flags & INP_RECVOPTS) {
1247 		*mp = sbcreatecontrol(opts_deleted_above,
1248 		    sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP, M_NOWAIT);
1249 		if (*mp)
1250 			mp = &(*mp)->m_next;
1251 	}
1252 	/* ip_srcroute doesn't do what we want here, need to fix */
1253 	if (inp->inp_flags & INP_RECVRETOPTS) {
1254 		*mp = sbcreatecontrol(ip_srcroute(m), sizeof(struct in_addr),
1255 		    IP_RECVRETOPTS, IPPROTO_IP, M_NOWAIT);
1256 		if (*mp)
1257 			mp = &(*mp)->m_next;
1258 	}
1259 #endif
1260 	if (inp->inp_flags & INP_RECVIF) {
1261 		struct ifnet *ifp;
1262 		struct sdlbuf {
1263 			struct sockaddr_dl sdl;
1264 			u_char	pad[32];
1265 		} sdlbuf;
1266 		struct sockaddr_dl *sdp;
1267 		struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1268 
1269 		if ((ifp = m->m_pkthdr.rcvif)) {
1270 			sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1271 			/*
1272 			 * Change our mind and don't try copy.
1273 			 */
1274 			if (sdp->sdl_family != AF_LINK ||
1275 			    sdp->sdl_len > sizeof(sdlbuf)) {
1276 				goto makedummy;
1277 			}
1278 			bcopy(sdp, sdl2, sdp->sdl_len);
1279 		} else {
1280 makedummy:
1281 			sdl2->sdl_len =
1282 			    offsetof(struct sockaddr_dl, sdl_data[0]);
1283 			sdl2->sdl_family = AF_LINK;
1284 			sdl2->sdl_index = 0;
1285 			sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1286 		}
1287 		*mp = sbcreatecontrol(sdl2, sdl2->sdl_len, IP_RECVIF,
1288 		    IPPROTO_IP, M_NOWAIT);
1289 		if (*mp)
1290 			mp = &(*mp)->m_next;
1291 	}
1292 	if (inp->inp_flags & INP_RECVTOS) {
1293 		*mp = sbcreatecontrol(&ip->ip_tos, sizeof(u_char), IP_RECVTOS,
1294 		    IPPROTO_IP, M_NOWAIT);
1295 		if (*mp)
1296 			mp = &(*mp)->m_next;
1297 	}
1298 
1299 	if (inp->inp_flags2 & INP_RECVFLOWID) {
1300 		uint32_t flowid, flow_type;
1301 
1302 		flowid = m->m_pkthdr.flowid;
1303 		flow_type = M_HASHTYPE_GET(m);
1304 
1305 		/*
1306 		 * XXX should handle the failure of one or the
1307 		 * other - don't populate both?
1308 		 */
1309 		*mp = sbcreatecontrol(&flowid, sizeof(uint32_t), IP_FLOWID,
1310 		    IPPROTO_IP, M_NOWAIT);
1311 		if (*mp)
1312 			mp = &(*mp)->m_next;
1313 		*mp = sbcreatecontrol(&flow_type, sizeof(uint32_t),
1314 		    IP_FLOWTYPE, IPPROTO_IP, M_NOWAIT);
1315 		if (*mp)
1316 			mp = &(*mp)->m_next;
1317 	}
1318 
1319 #ifdef	RSS
1320 	if (inp->inp_flags2 & INP_RECVRSSBUCKETID) {
1321 		uint32_t flowid, flow_type;
1322 		uint32_t rss_bucketid;
1323 
1324 		flowid = m->m_pkthdr.flowid;
1325 		flow_type = M_HASHTYPE_GET(m);
1326 
1327 		if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) {
1328 			*mp = sbcreatecontrol(&rss_bucketid, sizeof(uint32_t),
1329 			    IP_RSSBUCKETID, IPPROTO_IP, M_NOWAIT);
1330 			if (*mp)
1331 				mp = &(*mp)->m_next;
1332 		}
1333 	}
1334 #endif
1335 }
1336 
1337 /*
1338  * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
1339  * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
1340  * locking.  This code remains in ip_input.c as ip_mroute.c is optionally
1341  * compiled.
1342  */
1343 VNET_DEFINE_STATIC(int, ip_rsvp_on);
1344 VNET_DEFINE(struct socket *, ip_rsvpd);
1345 
1346 #define	V_ip_rsvp_on		VNET(ip_rsvp_on)
1347 
1348 int
1349 ip_rsvp_init(struct socket *so)
1350 {
1351 
1352 	if (V_ip_rsvpd != NULL)
1353 		return EADDRINUSE;
1354 
1355 	V_ip_rsvpd = so;
1356 	/*
1357 	 * This may seem silly, but we need to be sure we don't over-increment
1358 	 * the RSVP counter, in case something slips up.
1359 	 */
1360 	if (!V_ip_rsvp_on) {
1361 		V_ip_rsvp_on = 1;
1362 		V_rsvp_on++;
1363 	}
1364 
1365 	return 0;
1366 }
1367 
1368 int
1369 ip_rsvp_done(void)
1370 {
1371 
1372 	V_ip_rsvpd = NULL;
1373 	/*
1374 	 * This may seem silly, but we need to be sure we don't over-decrement
1375 	 * the RSVP counter, in case something slips up.
1376 	 */
1377 	if (V_ip_rsvp_on) {
1378 		V_ip_rsvp_on = 0;
1379 		V_rsvp_on--;
1380 	}
1381 	return 0;
1382 }
1383 
1384 int
1385 rsvp_input(struct mbuf **mp, int *offp, int proto)
1386 {
1387 	struct mbuf *m;
1388 
1389 	m = *mp;
1390 	*mp = NULL;
1391 
1392 	if (rsvp_input_p) { /* call the real one if loaded */
1393 		*mp = m;
1394 		rsvp_input_p(mp, offp, proto);
1395 		return (IPPROTO_DONE);
1396 	}
1397 
1398 	/* Can still get packets with rsvp_on = 0 if there is a local member
1399 	 * of the group to which the RSVP packet is addressed.  But in this
1400 	 * case we want to throw the packet away.
1401 	 */
1402 
1403 	if (!V_rsvp_on) {
1404 		m_freem(m);
1405 		return (IPPROTO_DONE);
1406 	}
1407 
1408 	if (V_ip_rsvpd != NULL) {
1409 		*mp = m;
1410 		rip_input(mp, offp, proto);
1411 		return (IPPROTO_DONE);
1412 	}
1413 	/* Drop the packet */
1414 	m_freem(m);
1415 	return (IPPROTO_DONE);
1416 }
1417