xref: /freebsd/sys/netinet/ip_input.c (revision 08aba0aec7b7f676ccc3f7886f59f277d668d5b4)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1982, 1986, 1988, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the University nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	@(#)ip_input.c	8.2 (Berkeley) 1/4/94
32  */
33 
34 #include <sys/cdefs.h>
35 __FBSDID("$FreeBSD$");
36 
37 #include "opt_bootp.h"
38 #include "opt_ipstealth.h"
39 #include "opt_ipsec.h"
40 #include "opt_route.h"
41 #include "opt_rss.h"
42 #include "opt_sctp.h"
43 
44 #include <sys/param.h>
45 #include <sys/systm.h>
46 #include <sys/hhook.h>
47 #include <sys/mbuf.h>
48 #include <sys/malloc.h>
49 #include <sys/domain.h>
50 #include <sys/protosw.h>
51 #include <sys/socket.h>
52 #include <sys/time.h>
53 #include <sys/kernel.h>
54 #include <sys/lock.h>
55 #include <sys/rmlock.h>
56 #include <sys/rwlock.h>
57 #include <sys/sdt.h>
58 #include <sys/syslog.h>
59 #include <sys/sysctl.h>
60 
61 #include <net/if.h>
62 #include <net/if_types.h>
63 #include <net/if_var.h>
64 #include <net/if_dl.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 VNET_DEFINE(pfil_head_t, inet_pfil_head);	/* Packet filter hooks */
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 	if (hhook_head_register(HHOOK_TYPE_IPSEC_IN, AF_INET,
329 	    &V_ipsec_hhh_in[HHOOK_IPSEC_INET],
330 	    HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
331 		printf("%s: WARNING: unable to register input helper hook\n",
332 		    __func__);
333 	if (hhook_head_register(HHOOK_TYPE_IPSEC_OUT, AF_INET,
334 	    &V_ipsec_hhh_out[HHOOK_IPSEC_INET],
335 	    HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
336 		printf("%s: WARNING: unable to register output helper hook\n",
337 		    __func__);
338 
339 #ifdef VIMAGE
340 	netisr_register_vnet(&ip_nh);
341 #ifdef	RSS
342 	netisr_register_vnet(&ip_direct_nh);
343 #endif
344 #endif
345 }
346 VNET_SYSINIT(ip_vnet_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_FOURTH,
347     ip_vnet_init, NULL);
348 
349 static void
350 ip_init(const void *unused __unused)
351 {
352 
353 	ipreass_init();
354 
355 	/*
356 	 * Register statically compiled protocols, that are unlikely to
357 	 * ever become dynamic.
358 	 */
359 	IPPROTO_REGISTER(IPPROTO_ICMP, icmp_input, NULL);
360 	IPPROTO_REGISTER(IPPROTO_IGMP, igmp_input, NULL);
361 	IPPROTO_REGISTER(IPPROTO_RSVP, rsvp_input, NULL);
362 	IPPROTO_REGISTER(IPPROTO_IPV4, encap4_input, NULL);
363 	IPPROTO_REGISTER(IPPROTO_MOBILE, encap4_input, NULL);
364 	IPPROTO_REGISTER(IPPROTO_ETHERIP, encap4_input, NULL);
365 	IPPROTO_REGISTER(IPPROTO_GRE, encap4_input, NULL);
366 	IPPROTO_REGISTER(IPPROTO_IPV6, encap4_input, NULL);
367 	IPPROTO_REGISTER(IPPROTO_PIM, encap4_input, NULL);
368 #ifdef SCTP	/* XXX: has a loadable & static version */
369 	IPPROTO_REGISTER(IPPROTO_SCTP, sctp_input, sctp_ctlinput);
370 #endif
371 
372 	netisr_register(&ip_nh);
373 #ifdef	RSS
374 	netisr_register(&ip_direct_nh);
375 #endif
376 }
377 SYSINIT(ip_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip_init, NULL);
378 
379 #ifdef VIMAGE
380 static void
381 ip_destroy(void *unused __unused)
382 {
383 	int error;
384 
385 #ifdef	RSS
386 	netisr_unregister_vnet(&ip_direct_nh);
387 #endif
388 	netisr_unregister_vnet(&ip_nh);
389 
390 	pfil_head_unregister(V_inet_pfil_head);
391 	error = hhook_head_deregister(V_ipsec_hhh_in[HHOOK_IPSEC_INET]);
392 	if (error != 0) {
393 		printf("%s: WARNING: unable to deregister input helper hook "
394 		    "type HHOOK_TYPE_IPSEC_IN, id HHOOK_IPSEC_INET: "
395 		    "error %d returned\n", __func__, error);
396 	}
397 	error = hhook_head_deregister(V_ipsec_hhh_out[HHOOK_IPSEC_INET]);
398 	if (error != 0) {
399 		printf("%s: WARNING: unable to deregister output helper hook "
400 		    "type HHOOK_TYPE_IPSEC_OUT, id HHOOK_IPSEC_INET: "
401 		    "error %d returned\n", __func__, error);
402 	}
403 
404 	/* Remove the IPv4 addresses from all interfaces. */
405 	in_ifscrub_all();
406 
407 	/* Make sure the IPv4 routes are gone as well. */
408 	rib_flush_routes_family(AF_INET);
409 
410 	/* Destroy IP reassembly queue. */
411 	ipreass_destroy();
412 
413 	/* Cleanup in_ifaddr hash table; should be empty. */
414 	hashdestroy(V_in_ifaddrhashtbl, M_IFADDR, V_in_ifaddrhmask);
415 }
416 
417 VNET_SYSUNINIT(ip, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip_destroy, NULL);
418 #endif
419 
420 #ifdef	RSS
421 /*
422  * IP direct input routine.
423  *
424  * This is called when reinjecting completed fragments where
425  * all of the previous checking and book-keeping has been done.
426  */
427 void
428 ip_direct_input(struct mbuf *m)
429 {
430 	struct ip *ip;
431 	int hlen;
432 
433 	ip = mtod(m, struct ip *);
434 	hlen = ip->ip_hl << 2;
435 
436 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
437 	if (IPSEC_ENABLED(ipv4)) {
438 		if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
439 			return;
440 	}
441 #endif /* IPSEC */
442 	IPSTAT_INC(ips_delivered);
443 	ip_protox[ip->ip_p](&m, &hlen, ip->ip_p);
444 }
445 #endif
446 
447 /*
448  * Ip input routine.  Checksum and byte swap header.  If fragmented
449  * try to reassemble.  Process options.  Pass to next level.
450  */
451 void
452 ip_input(struct mbuf *m)
453 {
454 	struct ip *ip = NULL;
455 	struct in_ifaddr *ia = NULL;
456 	struct ifaddr *ifa;
457 	struct ifnet *ifp;
458 	int hlen = 0;
459 	uint16_t sum, ip_len;
460 	int dchg = 0;				/* dest changed after fw */
461 	struct in_addr odst;			/* original dst address */
462 	bool strong_es;
463 
464 	M_ASSERTPKTHDR(m);
465 	NET_EPOCH_ASSERT();
466 
467 	if (m->m_flags & M_FASTFWD_OURS) {
468 		m->m_flags &= ~M_FASTFWD_OURS;
469 		/* Set up some basics that will be used later. */
470 		ip = mtod(m, struct ip *);
471 		hlen = ip->ip_hl << 2;
472 		ip_len = ntohs(ip->ip_len);
473 		goto ours;
474 	}
475 
476 	IPSTAT_INC(ips_total);
477 
478 	if (__predict_false(m->m_pkthdr.len < sizeof(struct ip)))
479 		goto tooshort;
480 
481 	if (m->m_len < sizeof(struct ip)) {
482 		m = m_pullup(m, sizeof(struct ip));
483 		if (__predict_false(m == NULL)) {
484 			IPSTAT_INC(ips_toosmall);
485 			return;
486 		}
487 	}
488 	ip = mtod(m, struct ip *);
489 
490 	if (__predict_false(ip->ip_v != IPVERSION)) {
491 		IPSTAT_INC(ips_badvers);
492 		goto bad;
493 	}
494 
495 	hlen = ip->ip_hl << 2;
496 	if (__predict_false(hlen < sizeof(struct ip))) {	/* minimum header length */
497 		IPSTAT_INC(ips_badhlen);
498 		goto bad;
499 	}
500 	if (hlen > m->m_len) {
501 		m = m_pullup(m, hlen);
502 		if (__predict_false(m == NULL)) {
503 			IPSTAT_INC(ips_badhlen);
504 			return;
505 		}
506 		ip = mtod(m, struct ip *);
507 	}
508 
509 	IP_PROBE(receive, NULL, NULL, ip, m->m_pkthdr.rcvif, ip, NULL);
510 
511 	/* IN_LOOPBACK must not appear on the wire - RFC1122 */
512 	ifp = m->m_pkthdr.rcvif;
513 	if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) ||
514 	    IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) {
515 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
516 			IPSTAT_INC(ips_badaddr);
517 			goto bad;
518 		}
519 	}
520 
521 	if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
522 		sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
523 	} else {
524 		if (hlen == sizeof(struct ip)) {
525 			sum = in_cksum_hdr(ip);
526 		} else {
527 			sum = in_cksum(m, hlen);
528 		}
529 	}
530 	if (__predict_false(sum)) {
531 		IPSTAT_INC(ips_badsum);
532 		goto bad;
533 	}
534 
535 #ifdef ALTQ
536 	if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
537 		/* packet is dropped by traffic conditioner */
538 		return;
539 #endif
540 
541 	ip_len = ntohs(ip->ip_len);
542 	if (__predict_false(ip_len < hlen)) {
543 		IPSTAT_INC(ips_badlen);
544 		goto bad;
545 	}
546 
547 	/*
548 	 * Check that the amount of data in the buffers
549 	 * is as at least much as the IP header would have us expect.
550 	 * Trim mbufs if longer than we expect.
551 	 * Drop packet if shorter than we expect.
552 	 */
553 	if (__predict_false(m->m_pkthdr.len < ip_len)) {
554 tooshort:
555 		IPSTAT_INC(ips_tooshort);
556 		goto bad;
557 	}
558 	if (m->m_pkthdr.len > ip_len) {
559 		if (m->m_len == m->m_pkthdr.len) {
560 			m->m_len = ip_len;
561 			m->m_pkthdr.len = ip_len;
562 		} else
563 			m_adj(m, ip_len - m->m_pkthdr.len);
564 	}
565 
566 	/*
567 	 * Try to forward the packet, but if we fail continue.
568 	 * ip_tryforward() may generate redirects these days.
569 	 * XXX the logic below falling through to normal processing
570 	 * if redirects are required should be revisited as well.
571 	 * ip_tryforward() does inbound and outbound packet firewall
572 	 * processing. If firewall has decided that destination becomes
573 	 * our local address, it sets M_FASTFWD_OURS flag. In this
574 	 * case skip another inbound firewall processing and update
575 	 * ip pointer.
576 	 */
577 	if (V_ipforwarding != 0
578 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
579 	    && (!IPSEC_ENABLED(ipv4) ||
580 	    IPSEC_CAPS(ipv4, m, IPSEC_CAP_OPERABLE) == 0)
581 #endif
582 	    ) {
583 		/*
584 		 * ip_dooptions() was run so we can ignore the source route (or
585 		 * any IP options case) case for redirects in ip_tryforward().
586 		 */
587 		if ((m = ip_tryforward(m)) == NULL)
588 			return;
589 		if (m->m_flags & M_FASTFWD_OURS) {
590 			m->m_flags &= ~M_FASTFWD_OURS;
591 			ip = mtod(m, struct ip *);
592 			goto ours;
593 		}
594 	}
595 
596 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
597 	/*
598 	 * Bypass packet filtering for packets previously handled by IPsec.
599 	 */
600 	if (IPSEC_ENABLED(ipv4) &&
601 	    IPSEC_CAPS(ipv4, m, IPSEC_CAP_BYPASS_FILTER) != 0)
602 			goto passin;
603 #endif
604 
605 	/*
606 	 * Run through list of hooks for input packets.
607 	 *
608 	 * NB: Beware of the destination address changing (e.g.
609 	 *     by NAT rewriting).  When this happens, tell
610 	 *     ip_forward to do the right thing.
611 	 */
612 
613 	/* Jump over all PFIL processing if hooks are not active. */
614 	if (!PFIL_HOOKED_IN(V_inet_pfil_head))
615 		goto passin;
616 
617 	odst = ip->ip_dst;
618 	if (pfil_run_hooks(V_inet_pfil_head, &m, ifp, PFIL_IN, NULL) !=
619 	    PFIL_PASS)
620 		return;
621 	if (m == NULL)			/* consumed by filter */
622 		return;
623 
624 	ip = mtod(m, struct ip *);
625 	dchg = (odst.s_addr != ip->ip_dst.s_addr);
626 
627 	if (m->m_flags & M_FASTFWD_OURS) {
628 		m->m_flags &= ~M_FASTFWD_OURS;
629 		goto ours;
630 	}
631 	if (m->m_flags & M_IP_NEXTHOP) {
632 		if (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) {
633 			/*
634 			 * Directly ship the packet on.  This allows
635 			 * forwarding packets originally destined to us
636 			 * to some other directly connected host.
637 			 */
638 			ip_forward(m, 1);
639 			return;
640 		}
641 	}
642 passin:
643 
644 	/*
645 	 * Process options and, if not destined for us,
646 	 * ship it on.  ip_dooptions returns 1 when an
647 	 * error was detected (causing an icmp message
648 	 * to be sent and the original packet to be freed).
649 	 */
650 	if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
651 		return;
652 
653         /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
654          * matter if it is destined to another node, or whether it is
655          * a multicast one, RSVP wants it! and prevents it from being forwarded
656          * anywhere else. Also checks if the rsvp daemon is running before
657 	 * grabbing the packet.
658          */
659 	if (ip->ip_p == IPPROTO_RSVP && V_rsvp_on)
660 		goto ours;
661 
662 	/*
663 	 * Check our list of addresses, to see if the packet is for us.
664 	 * If we don't have any addresses, assume any unicast packet
665 	 * we receive might be for us (and let the upper layers deal
666 	 * with it).
667 	 */
668 	if (CK_STAILQ_EMPTY(&V_in_ifaddrhead) &&
669 	    (m->m_flags & (M_MCAST|M_BCAST)) == 0)
670 		goto ours;
671 
672 	/*
673 	 * Enable a consistency check between the destination address
674 	 * and the arrival interface for a unicast packet (the RFC 1122
675 	 * strong ES model) with a list of additional predicates:
676 	 * - if IP forwarding is disabled
677 	 * - the packet is not locally generated
678 	 * - the packet is not subject to 'ipfw fwd'
679 	 * - Interface is not running CARP. If the packet got here, we already
680 	 *   checked it with carp_iamatch() and carp_forus().
681 	 */
682 	strong_es = V_ip_strong_es && (V_ipforwarding == 0) &&
683 	    ((ifp->if_flags & IFF_LOOPBACK) == 0) &&
684 	    ifp->if_carp == NULL && (dchg == 0);
685 
686 	/*
687 	 * Check for exact addresses in the hash bucket.
688 	 */
689 	CK_LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
690 		if (IA_SIN(ia)->sin_addr.s_addr != ip->ip_dst.s_addr)
691 			continue;
692 
693 		/*
694 		 * net.inet.ip.rfc1122_strong_es: the address matches, verify
695 		 * that the packet arrived via the correct interface.
696 		 */
697 		if (__predict_false(strong_es && ia->ia_ifp != ifp)) {
698 			IPSTAT_INC(ips_badaddr);
699 			goto bad;
700 		}
701 
702 		/*
703 		 * net.inet.ip.source_address_validation: drop incoming
704 		 * packets that pretend to be ours.
705 		 */
706 		if (V_ip_sav && !(ifp->if_flags & IFF_LOOPBACK) &&
707 		    __predict_false(in_localip_fib(ip->ip_src, ifp->if_fib))) {
708 			IPSTAT_INC(ips_badaddr);
709 			goto bad;
710 		}
711 
712 		counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
713 		counter_u64_add(ia->ia_ifa.ifa_ibytes, m->m_pkthdr.len);
714 		goto ours;
715 	}
716 
717 	/*
718 	 * Check for broadcast addresses.
719 	 *
720 	 * Only accept broadcast packets that arrive via the matching
721 	 * interface.  Reception of forwarded directed broadcasts would
722 	 * be handled via ip_forward() and ether_output() with the loopback
723 	 * into the stack for SIMPLEX interfaces handled by ether_output().
724 	 */
725 	if (ifp->if_flags & IFF_BROADCAST) {
726 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
727 			if (ifa->ifa_addr->sa_family != AF_INET)
728 				continue;
729 			ia = ifatoia(ifa);
730 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
731 			    ip->ip_dst.s_addr) {
732 				counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
733 				counter_u64_add(ia->ia_ifa.ifa_ibytes,
734 				    m->m_pkthdr.len);
735 				goto ours;
736 			}
737 #ifdef BOOTP_COMPAT
738 			if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) {
739 				counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
740 				counter_u64_add(ia->ia_ifa.ifa_ibytes,
741 				    m->m_pkthdr.len);
742 				goto ours;
743 			}
744 #endif
745 		}
746 		ia = NULL;
747 	}
748 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
749 		/*
750 		 * RFC 3927 2.7: Do not forward multicast packets from
751 		 * IN_LINKLOCAL.
752 		 */
753 		if (V_ip_mrouter && !IN_LINKLOCAL(ntohl(ip->ip_src.s_addr))) {
754 			/*
755 			 * If we are acting as a multicast router, all
756 			 * incoming multicast packets are passed to the
757 			 * kernel-level multicast forwarding function.
758 			 * The packet is returned (relatively) intact; if
759 			 * ip_mforward() returns a non-zero value, the packet
760 			 * must be discarded, else it may be accepted below.
761 			 */
762 			if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) {
763 				IPSTAT_INC(ips_cantforward);
764 				m_freem(m);
765 				return;
766 			}
767 
768 			/*
769 			 * The process-level routing daemon needs to receive
770 			 * all multicast IGMP packets, whether or not this
771 			 * host belongs to their destination groups.
772 			 */
773 			if (ip->ip_p == IPPROTO_IGMP) {
774 				goto ours;
775 			}
776 			IPSTAT_INC(ips_forward);
777 		}
778 		/*
779 		 * Assume the packet is for us, to avoid prematurely taking
780 		 * a lock on the in_multi hash. Protocols must perform
781 		 * their own filtering and update statistics accordingly.
782 		 */
783 		goto ours;
784 	}
785 	if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
786 		goto ours;
787 	if (ip->ip_dst.s_addr == INADDR_ANY)
788 		goto ours;
789 	/* RFC 3927 2.7: Do not forward packets to or from IN_LINKLOCAL. */
790 	if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr)) ||
791 	    IN_LINKLOCAL(ntohl(ip->ip_src.s_addr))) {
792 		IPSTAT_INC(ips_cantforward);
793 		m_freem(m);
794 		return;
795 	}
796 
797 	/*
798 	 * Not for us; forward if possible and desirable.
799 	 */
800 	if (V_ipforwarding == 0) {
801 		IPSTAT_INC(ips_cantforward);
802 		m_freem(m);
803 	} else {
804 		ip_forward(m, dchg);
805 	}
806 	return;
807 
808 ours:
809 #ifdef IPSTEALTH
810 	/*
811 	 * IPSTEALTH: Process non-routing options only
812 	 * if the packet is destined for us.
813 	 */
814 	if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1))
815 		return;
816 #endif /* IPSTEALTH */
817 
818 	/*
819 	 * Attempt reassembly; if it succeeds, proceed.
820 	 * ip_reass() will return a different mbuf.
821 	 */
822 	if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) {
823 		/* XXXGL: shouldn't we save & set m_flags? */
824 		m = ip_reass(m);
825 		if (m == NULL)
826 			return;
827 		ip = mtod(m, struct ip *);
828 		/* Get the header length of the reassembled packet */
829 		hlen = ip->ip_hl << 2;
830 	}
831 
832 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
833 	if (IPSEC_ENABLED(ipv4)) {
834 		if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
835 			return;
836 	}
837 #endif /* IPSEC */
838 
839 	/*
840 	 * Switch out to protocol's input routine.
841 	 */
842 	IPSTAT_INC(ips_delivered);
843 
844 	ip_protox[ip->ip_p](&m, &hlen, ip->ip_p);
845 	return;
846 bad:
847 	m_freem(m);
848 }
849 
850 int
851 ipproto_register(uint8_t proto, ipproto_input_t input, ipproto_ctlinput_t ctl)
852 {
853 
854 	MPASS(proto > 0);
855 
856 	/*
857 	 * The protocol slot must not be occupied by another protocol
858 	 * already.  An index pointing to rip_input() is unused.
859 	 */
860 	if (ip_protox[proto] == rip_input) {
861 		ip_protox[proto] = input;
862 		ip_ctlprotox[proto] = ctl;
863 		return (0);
864 	} else
865 		return (EEXIST);
866 }
867 
868 int
869 ipproto_unregister(uint8_t proto)
870 {
871 
872 	MPASS(proto > 0);
873 
874 	if (ip_protox[proto] != rip_input) {
875 		ip_protox[proto] = rip_input;
876 		ip_ctlprotox[proto] = rip_ctlinput;
877 		return (0);
878 	} else
879 		return (ENOENT);
880 }
881 
882 u_char inetctlerrmap[PRC_NCMDS] = {
883 	0,		0,		0,		0,
884 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
885 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
886 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
887 	0,		0,		EHOSTUNREACH,	0,
888 	ENOPROTOOPT,	ECONNREFUSED
889 };
890 
891 /*
892  * Forward a packet.  If some error occurs return the sender
893  * an icmp packet.  Note we can't always generate a meaningful
894  * icmp message because icmp doesn't have a large enough repertoire
895  * of codes and types.
896  *
897  * If not forwarding, just drop the packet.  This could be confusing
898  * if ipforwarding was zero but some routing protocol was advancing
899  * us as a gateway to somewhere.  However, we must let the routing
900  * protocol deal with that.
901  *
902  * The srcrt parameter indicates whether the packet is being forwarded
903  * via a source route.
904  */
905 void
906 ip_forward(struct mbuf *m, int srcrt)
907 {
908 	struct ip *ip = mtod(m, struct ip *);
909 	struct in_ifaddr *ia;
910 	struct mbuf *mcopy;
911 	struct sockaddr_in *sin;
912 	struct in_addr dest;
913 	struct route ro;
914 	uint32_t flowid;
915 	int error, type = 0, code = 0, mtu = 0;
916 
917 	NET_EPOCH_ASSERT();
918 
919 	if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
920 		IPSTAT_INC(ips_cantforward);
921 		m_freem(m);
922 		return;
923 	}
924 	if (
925 #ifdef IPSTEALTH
926 	    V_ipstealth == 0 &&
927 #endif
928 	    ip->ip_ttl <= IPTTLDEC) {
929 		icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 0, 0);
930 		return;
931 	}
932 
933 	bzero(&ro, sizeof(ro));
934 	sin = (struct sockaddr_in *)&ro.ro_dst;
935 	sin->sin_family = AF_INET;
936 	sin->sin_len = sizeof(*sin);
937 	sin->sin_addr = ip->ip_dst;
938 	flowid = m->m_pkthdr.flowid;
939 	ro.ro_nh = fib4_lookup(M_GETFIB(m), ip->ip_dst, 0, NHR_REF, flowid);
940 	if (ro.ro_nh != NULL) {
941 		ia = ifatoia(ro.ro_nh->nh_ifa);
942 	} else
943 		ia = NULL;
944 	/*
945 	 * Save the IP header and at most 8 bytes of the payload,
946 	 * in case we need to generate an ICMP message to the src.
947 	 *
948 	 * XXX this can be optimized a lot by saving the data in a local
949 	 * buffer on the stack (72 bytes at most), and only allocating the
950 	 * mbuf if really necessary. The vast majority of the packets
951 	 * are forwarded without having to send an ICMP back (either
952 	 * because unnecessary, or because rate limited), so we are
953 	 * really we are wasting a lot of work here.
954 	 *
955 	 * We don't use m_copym() because it might return a reference
956 	 * to a shared cluster. Both this function and ip_output()
957 	 * assume exclusive access to the IP header in `m', so any
958 	 * data in a cluster may change before we reach icmp_error().
959 	 */
960 	mcopy = m_gethdr(M_NOWAIT, m->m_type);
961 	if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_NOWAIT)) {
962 		/*
963 		 * It's probably ok if the pkthdr dup fails (because
964 		 * the deep copy of the tag chain failed), but for now
965 		 * be conservative and just discard the copy since
966 		 * code below may some day want the tags.
967 		 */
968 		m_free(mcopy);
969 		mcopy = NULL;
970 	}
971 	if (mcopy != NULL) {
972 		mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy));
973 		mcopy->m_pkthdr.len = mcopy->m_len;
974 		m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
975 	}
976 #ifdef IPSTEALTH
977 	if (V_ipstealth == 0)
978 #endif
979 		ip->ip_ttl -= IPTTLDEC;
980 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
981 	if (IPSEC_ENABLED(ipv4)) {
982 		if ((error = IPSEC_FORWARD(ipv4, m)) != 0) {
983 			/* mbuf consumed by IPsec */
984 			RO_NHFREE(&ro);
985 			m_freem(mcopy);
986 			if (error != EINPROGRESS)
987 				IPSTAT_INC(ips_cantforward);
988 			return;
989 		}
990 		/* No IPsec processing required */
991 	}
992 #endif /* IPSEC */
993 	/*
994 	 * If forwarding packet using same interface that it came in on,
995 	 * perhaps should send a redirect to sender to shortcut a hop.
996 	 * Only send redirect if source is sending directly to us,
997 	 * and if packet was not source routed (or has any options).
998 	 * Also, don't send redirect if forwarding using a default route
999 	 * or a route modified by a redirect.
1000 	 */
1001 	dest.s_addr = 0;
1002 	if (!srcrt && V_ipsendredirects &&
1003 	    ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) {
1004 		struct nhop_object *nh;
1005 
1006 		nh = ro.ro_nh;
1007 
1008 		if (nh != NULL && ((nh->nh_flags & (NHF_REDIRECT|NHF_DEFAULT)) == 0)) {
1009 			struct in_ifaddr *nh_ia = (struct in_ifaddr *)(nh->nh_ifa);
1010 			u_long src = ntohl(ip->ip_src.s_addr);
1011 
1012 			if (nh_ia != NULL &&
1013 			    (src & nh_ia->ia_subnetmask) == nh_ia->ia_subnet) {
1014 				/* Router requirements says to only send host redirects */
1015 				type = ICMP_REDIRECT;
1016 				code = ICMP_REDIRECT_HOST;
1017 				if (nh->nh_flags & NHF_GATEWAY) {
1018 				    if (nh->gw_sa.sa_family == AF_INET)
1019 					dest.s_addr = nh->gw4_sa.sin_addr.s_addr;
1020 				    else /* Do not redirect in case gw is AF_INET6 */
1021 					type = 0;
1022 				} else
1023 					dest.s_addr = ip->ip_dst.s_addr;
1024 			}
1025 		}
1026 	}
1027 
1028 	error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
1029 
1030 	if (error == EMSGSIZE && ro.ro_nh)
1031 		mtu = ro.ro_nh->nh_mtu;
1032 	RO_NHFREE(&ro);
1033 
1034 	if (error)
1035 		IPSTAT_INC(ips_cantforward);
1036 	else {
1037 		IPSTAT_INC(ips_forward);
1038 		if (type)
1039 			IPSTAT_INC(ips_redirectsent);
1040 		else {
1041 			if (mcopy)
1042 				m_freem(mcopy);
1043 			return;
1044 		}
1045 	}
1046 	if (mcopy == NULL)
1047 		return;
1048 
1049 	switch (error) {
1050 	case 0:				/* forwarded, but need redirect */
1051 		/* type, code set above */
1052 		break;
1053 
1054 	case ENETUNREACH:
1055 	case EHOSTUNREACH:
1056 	case ENETDOWN:
1057 	case EHOSTDOWN:
1058 	default:
1059 		type = ICMP_UNREACH;
1060 		code = ICMP_UNREACH_HOST;
1061 		break;
1062 
1063 	case EMSGSIZE:
1064 		type = ICMP_UNREACH;
1065 		code = ICMP_UNREACH_NEEDFRAG;
1066 		/*
1067 		 * If the MTU was set before make sure we are below the
1068 		 * interface MTU.
1069 		 * If the MTU wasn't set before use the interface mtu or
1070 		 * fall back to the next smaller mtu step compared to the
1071 		 * current packet size.
1072 		 */
1073 		if (mtu != 0) {
1074 			if (ia != NULL)
1075 				mtu = min(mtu, ia->ia_ifp->if_mtu);
1076 		} else {
1077 			if (ia != NULL)
1078 				mtu = ia->ia_ifp->if_mtu;
1079 			else
1080 				mtu = ip_next_mtu(ntohs(ip->ip_len), 0);
1081 		}
1082 		IPSTAT_INC(ips_cantfrag);
1083 		break;
1084 
1085 	case ENOBUFS:
1086 	case EACCES:			/* ipfw denied packet */
1087 		m_freem(mcopy);
1088 		return;
1089 	}
1090 	icmp_error(mcopy, type, code, dest.s_addr, mtu);
1091 }
1092 
1093 #define	CHECK_SO_CT(sp, ct) \
1094     (((sp->so_options & SO_TIMESTAMP) && (sp->so_ts_clock == ct)) ? 1 : 0)
1095 
1096 void
1097 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1098     struct mbuf *m)
1099 {
1100 	bool stamped;
1101 
1102 	stamped = false;
1103 	if ((inp->inp_socket->so_options & SO_BINTIME) ||
1104 	    CHECK_SO_CT(inp->inp_socket, SO_TS_BINTIME)) {
1105 		struct bintime boottimebin, bt;
1106 		struct timespec ts1;
1107 
1108 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1109 		    M_TSTMP)) {
1110 			mbuf_tstmp2timespec(m, &ts1);
1111 			timespec2bintime(&ts1, &bt);
1112 			getboottimebin(&boottimebin);
1113 			bintime_add(&bt, &boottimebin);
1114 		} else {
1115 			bintime(&bt);
1116 		}
1117 		*mp = sbcreatecontrol(&bt, sizeof(bt), SCM_BINTIME,
1118 		    SOL_SOCKET, M_NOWAIT);
1119 		if (*mp != NULL) {
1120 			mp = &(*mp)->m_next;
1121 			stamped = true;
1122 		}
1123 	}
1124 	if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME_MICRO)) {
1125 		struct bintime boottimebin, bt1;
1126 		struct timespec ts1;
1127 		struct timeval tv;
1128 
1129 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1130 		    M_TSTMP)) {
1131 			mbuf_tstmp2timespec(m, &ts1);
1132 			timespec2bintime(&ts1, &bt1);
1133 			getboottimebin(&boottimebin);
1134 			bintime_add(&bt1, &boottimebin);
1135 			bintime2timeval(&bt1, &tv);
1136 		} else {
1137 			microtime(&tv);
1138 		}
1139 		*mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv), SCM_TIMESTAMP,
1140 		    SOL_SOCKET, M_NOWAIT);
1141 		if (*mp != NULL) {
1142 			mp = &(*mp)->m_next;
1143 			stamped = true;
1144 		}
1145 	} else if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME)) {
1146 		struct bintime boottimebin;
1147 		struct timespec ts, ts1;
1148 
1149 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1150 		    M_TSTMP)) {
1151 			mbuf_tstmp2timespec(m, &ts);
1152 			getboottimebin(&boottimebin);
1153 			bintime2timespec(&boottimebin, &ts1);
1154 			timespecadd(&ts, &ts1, &ts);
1155 		} else {
1156 			nanotime(&ts);
1157 		}
1158 		*mp = sbcreatecontrol(&ts, sizeof(ts), SCM_REALTIME,
1159 		    SOL_SOCKET, M_NOWAIT);
1160 		if (*mp != NULL) {
1161 			mp = &(*mp)->m_next;
1162 			stamped = true;
1163 		}
1164 	} else if (CHECK_SO_CT(inp->inp_socket, SO_TS_MONOTONIC)) {
1165 		struct timespec ts;
1166 
1167 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1168 		    M_TSTMP))
1169 			mbuf_tstmp2timespec(m, &ts);
1170 		else
1171 			nanouptime(&ts);
1172 		*mp = sbcreatecontrol(&ts, sizeof(ts), SCM_MONOTONIC,
1173 		    SOL_SOCKET, M_NOWAIT);
1174 		if (*mp != NULL) {
1175 			mp = &(*mp)->m_next;
1176 			stamped = true;
1177 		}
1178 	}
1179 	if (stamped && (m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1180 	    M_TSTMP)) {
1181 		struct sock_timestamp_info sti;
1182 
1183 		bzero(&sti, sizeof(sti));
1184 		sti.st_info_flags = ST_INFO_HW;
1185 		if ((m->m_flags & M_TSTMP_HPREC) != 0)
1186 			sti.st_info_flags |= ST_INFO_HW_HPREC;
1187 		*mp = sbcreatecontrol(&sti, sizeof(sti), SCM_TIME_INFO,
1188 		    SOL_SOCKET, M_NOWAIT);
1189 		if (*mp != NULL)
1190 			mp = &(*mp)->m_next;
1191 	}
1192 	if (inp->inp_flags & INP_RECVDSTADDR) {
1193 		*mp = sbcreatecontrol(&ip->ip_dst, sizeof(struct in_addr),
1194 		    IP_RECVDSTADDR, IPPROTO_IP, M_NOWAIT);
1195 		if (*mp)
1196 			mp = &(*mp)->m_next;
1197 	}
1198 	if (inp->inp_flags & INP_RECVTTL) {
1199 		*mp = sbcreatecontrol(&ip->ip_ttl, sizeof(u_char), IP_RECVTTL,
1200 		    IPPROTO_IP, M_NOWAIT);
1201 		if (*mp)
1202 			mp = &(*mp)->m_next;
1203 	}
1204 #ifdef notyet
1205 	/* XXX
1206 	 * Moving these out of udp_input() made them even more broken
1207 	 * than they already were.
1208 	 */
1209 	/* options were tossed already */
1210 	if (inp->inp_flags & INP_RECVOPTS) {
1211 		*mp = sbcreatecontrol(opts_deleted_above,
1212 		    sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP, M_NOWAIT);
1213 		if (*mp)
1214 			mp = &(*mp)->m_next;
1215 	}
1216 	/* ip_srcroute doesn't do what we want here, need to fix */
1217 	if (inp->inp_flags & INP_RECVRETOPTS) {
1218 		*mp = sbcreatecontrol(ip_srcroute(m), sizeof(struct in_addr),
1219 		    IP_RECVRETOPTS, IPPROTO_IP, M_NOWAIT);
1220 		if (*mp)
1221 			mp = &(*mp)->m_next;
1222 	}
1223 #endif
1224 	if (inp->inp_flags & INP_RECVIF) {
1225 		struct ifnet *ifp;
1226 		struct sdlbuf {
1227 			struct sockaddr_dl sdl;
1228 			u_char	pad[32];
1229 		} sdlbuf;
1230 		struct sockaddr_dl *sdp;
1231 		struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1232 
1233 		if ((ifp = m->m_pkthdr.rcvif)) {
1234 			sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1235 			/*
1236 			 * Change our mind and don't try copy.
1237 			 */
1238 			if (sdp->sdl_family != AF_LINK ||
1239 			    sdp->sdl_len > sizeof(sdlbuf)) {
1240 				goto makedummy;
1241 			}
1242 			bcopy(sdp, sdl2, sdp->sdl_len);
1243 		} else {
1244 makedummy:
1245 			sdl2->sdl_len =
1246 			    offsetof(struct sockaddr_dl, sdl_data[0]);
1247 			sdl2->sdl_family = AF_LINK;
1248 			sdl2->sdl_index = 0;
1249 			sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1250 		}
1251 		*mp = sbcreatecontrol(sdl2, sdl2->sdl_len, IP_RECVIF,
1252 		    IPPROTO_IP, M_NOWAIT);
1253 		if (*mp)
1254 			mp = &(*mp)->m_next;
1255 	}
1256 	if (inp->inp_flags & INP_RECVTOS) {
1257 		*mp = sbcreatecontrol(&ip->ip_tos, sizeof(u_char), IP_RECVTOS,
1258 		    IPPROTO_IP, M_NOWAIT);
1259 		if (*mp)
1260 			mp = &(*mp)->m_next;
1261 	}
1262 
1263 	if (inp->inp_flags2 & INP_RECVFLOWID) {
1264 		uint32_t flowid, flow_type;
1265 
1266 		flowid = m->m_pkthdr.flowid;
1267 		flow_type = M_HASHTYPE_GET(m);
1268 
1269 		/*
1270 		 * XXX should handle the failure of one or the
1271 		 * other - don't populate both?
1272 		 */
1273 		*mp = sbcreatecontrol(&flowid, sizeof(uint32_t), IP_FLOWID,
1274 		    IPPROTO_IP, M_NOWAIT);
1275 		if (*mp)
1276 			mp = &(*mp)->m_next;
1277 		*mp = sbcreatecontrol(&flow_type, sizeof(uint32_t),
1278 		    IP_FLOWTYPE, IPPROTO_IP, M_NOWAIT);
1279 		if (*mp)
1280 			mp = &(*mp)->m_next;
1281 	}
1282 
1283 #ifdef	RSS
1284 	if (inp->inp_flags2 & INP_RECVRSSBUCKETID) {
1285 		uint32_t flowid, flow_type;
1286 		uint32_t rss_bucketid;
1287 
1288 		flowid = m->m_pkthdr.flowid;
1289 		flow_type = M_HASHTYPE_GET(m);
1290 
1291 		if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) {
1292 			*mp = sbcreatecontrol(&rss_bucketid, sizeof(uint32_t),
1293 			    IP_RSSBUCKETID, IPPROTO_IP, M_NOWAIT);
1294 			if (*mp)
1295 				mp = &(*mp)->m_next;
1296 		}
1297 	}
1298 #endif
1299 }
1300 
1301 /*
1302  * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
1303  * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
1304  * locking.  This code remains in ip_input.c as ip_mroute.c is optionally
1305  * compiled.
1306  */
1307 VNET_DEFINE_STATIC(int, ip_rsvp_on);
1308 VNET_DEFINE(struct socket *, ip_rsvpd);
1309 
1310 #define	V_ip_rsvp_on		VNET(ip_rsvp_on)
1311 
1312 int
1313 ip_rsvp_init(struct socket *so)
1314 {
1315 
1316 	if (so->so_type != SOCK_RAW ||
1317 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
1318 		return EOPNOTSUPP;
1319 
1320 	if (V_ip_rsvpd != NULL)
1321 		return EADDRINUSE;
1322 
1323 	V_ip_rsvpd = so;
1324 	/*
1325 	 * This may seem silly, but we need to be sure we don't over-increment
1326 	 * the RSVP counter, in case something slips up.
1327 	 */
1328 	if (!V_ip_rsvp_on) {
1329 		V_ip_rsvp_on = 1;
1330 		V_rsvp_on++;
1331 	}
1332 
1333 	return 0;
1334 }
1335 
1336 int
1337 ip_rsvp_done(void)
1338 {
1339 
1340 	V_ip_rsvpd = NULL;
1341 	/*
1342 	 * This may seem silly, but we need to be sure we don't over-decrement
1343 	 * the RSVP counter, in case something slips up.
1344 	 */
1345 	if (V_ip_rsvp_on) {
1346 		V_ip_rsvp_on = 0;
1347 		V_rsvp_on--;
1348 	}
1349 	return 0;
1350 }
1351 
1352 int
1353 rsvp_input(struct mbuf **mp, int *offp, int proto)
1354 {
1355 	struct mbuf *m;
1356 
1357 	m = *mp;
1358 	*mp = NULL;
1359 
1360 	if (rsvp_input_p) { /* call the real one if loaded */
1361 		*mp = m;
1362 		rsvp_input_p(mp, offp, proto);
1363 		return (IPPROTO_DONE);
1364 	}
1365 
1366 	/* Can still get packets with rsvp_on = 0 if there is a local member
1367 	 * of the group to which the RSVP packet is addressed.  But in this
1368 	 * case we want to throw the packet away.
1369 	 */
1370 
1371 	if (!V_rsvp_on) {
1372 		m_freem(m);
1373 		return (IPPROTO_DONE);
1374 	}
1375 
1376 	if (V_ip_rsvpd != NULL) {
1377 		*mp = m;
1378 		rip_input(mp, offp, proto);
1379 		return (IPPROTO_DONE);
1380 	}
1381 	/* Drop the packet */
1382 	m_freem(m);
1383 	return (IPPROTO_DONE);
1384 }
1385