xref: /freebsd/sys/netinet/ip_input.c (revision 81ad626541db97eb356e2c1d4a20eb2a26a766ab)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1982, 1986, 1988, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the University nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	@(#)ip_input.c	8.2 (Berkeley) 1/4/94
32  */
33 
34 #include <sys/cdefs.h>
35 __FBSDID("$FreeBSD$");
36 
37 #include "opt_bootp.h"
38 #include "opt_inet.h"
39 #include "opt_ipstealth.h"
40 #include "opt_ipsec.h"
41 #include "opt_route.h"
42 #include "opt_rss.h"
43 #include "opt_sctp.h"
44 
45 #include <sys/param.h>
46 #include <sys/systm.h>
47 #include <sys/hhook.h>
48 #include <sys/mbuf.h>
49 #include <sys/malloc.h>
50 #include <sys/domain.h>
51 #include <sys/protosw.h>
52 #include <sys/socket.h>
53 #include <sys/time.h>
54 #include <sys/kernel.h>
55 #include <sys/lock.h>
56 #include <sys/rmlock.h>
57 #include <sys/rwlock.h>
58 #include <sys/sdt.h>
59 #include <sys/syslog.h>
60 #include <sys/sysctl.h>
61 
62 #include <net/if.h>
63 #include <net/if_types.h>
64 #include <net/if_var.h>
65 #include <net/if_dl.h>
66 #include <net/if_private.h>
67 #include <net/pfil.h>
68 #include <net/route.h>
69 #include <net/route/nhop.h>
70 #include <net/netisr.h>
71 #include <net/rss_config.h>
72 #include <net/vnet.h>
73 
74 #include <netinet/in.h>
75 #include <netinet/in_kdtrace.h>
76 #include <netinet/in_systm.h>
77 #include <netinet/in_var.h>
78 #include <netinet/ip.h>
79 #include <netinet/in_fib.h>
80 #include <netinet/in_pcb.h>
81 #include <netinet/ip_var.h>
82 #include <netinet/ip_encap.h>
83 #include <netinet/ip_fw.h>
84 #include <netinet/ip_icmp.h>
85 #include <netinet/igmp_var.h>
86 #include <netinet/ip_options.h>
87 #include <machine/in_cksum.h>
88 #include <netinet/ip_carp.h>
89 #include <netinet/in_rss.h>
90 #ifdef SCTP
91 #include <netinet/sctp_var.h>
92 #endif
93 
94 #include <netipsec/ipsec_support.h>
95 
96 #include <sys/socketvar.h>
97 
98 #include <security/mac/mac_framework.h>
99 
100 #ifdef CTASSERT
101 CTASSERT(sizeof(struct ip) == 20);
102 #endif
103 
104 /* IP reassembly functions are defined in ip_reass.c. */
105 extern void ipreass_init(void);
106 extern void ipreass_vnet_init(void);
107 #ifdef VIMAGE
108 extern void ipreass_destroy(void);
109 #endif
110 
111 VNET_DEFINE(int, rsvp_on);
112 
113 VNET_DEFINE(int, ipforwarding);
114 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_VNET | CTLFLAG_RW,
115     &VNET_NAME(ipforwarding), 0,
116     "Enable IP forwarding between interfaces");
117 
118 /*
119  * Respond with an ICMP host redirect when we forward a packet out of
120  * the same interface on which it was received.  See RFC 792.
121  */
122 VNET_DEFINE(int, ipsendredirects) = 1;
123 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_VNET | CTLFLAG_RW,
124     &VNET_NAME(ipsendredirects), 0,
125     "Enable sending IP redirects");
126 
127 VNET_DEFINE_STATIC(bool, ip_strong_es) = false;
128 #define	V_ip_strong_es	VNET(ip_strong_es)
129 SYSCTL_BOOL(_net_inet_ip, OID_AUTO, rfc1122_strong_es,
130     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip_strong_es), false,
131     "Packet's IP destination address must match address on arrival interface");
132 
133 VNET_DEFINE_STATIC(bool, ip_sav) = true;
134 #define	V_ip_sav	VNET(ip_sav)
135 SYSCTL_BOOL(_net_inet_ip, OID_AUTO, source_address_validation,
136     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip_sav), true,
137     "Drop incoming packets with source address that is a local address");
138 
139 VNET_DEFINE(pfil_head_t, inet_pfil_head);	/* Packet filter hooks */
140 
141 static struct netisr_handler ip_nh = {
142 	.nh_name = "ip",
143 	.nh_handler = ip_input,
144 	.nh_proto = NETISR_IP,
145 #ifdef	RSS
146 	.nh_m2cpuid = rss_soft_m2cpuid_v4,
147 	.nh_policy = NETISR_POLICY_CPU,
148 	.nh_dispatch = NETISR_DISPATCH_HYBRID,
149 #else
150 	.nh_policy = NETISR_POLICY_FLOW,
151 #endif
152 };
153 
154 #ifdef	RSS
155 /*
156  * Directly dispatched frames are currently assumed
157  * to have a flowid already calculated.
158  *
159  * It should likely have something that assert it
160  * actually has valid flow details.
161  */
162 static struct netisr_handler ip_direct_nh = {
163 	.nh_name = "ip_direct",
164 	.nh_handler = ip_direct_input,
165 	.nh_proto = NETISR_IP_DIRECT,
166 	.nh_m2cpuid = rss_soft_m2cpuid_v4,
167 	.nh_policy = NETISR_POLICY_CPU,
168 	.nh_dispatch = NETISR_DISPATCH_HYBRID,
169 };
170 #endif
171 
172 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 	if (hhook_head_register(HHOOK_TYPE_IPSEC_IN, AF_INET,
331 	    &V_ipsec_hhh_in[HHOOK_IPSEC_INET],
332 	    HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
333 		printf("%s: WARNING: unable to register input helper hook\n",
334 		    __func__);
335 	if (hhook_head_register(HHOOK_TYPE_IPSEC_OUT, AF_INET,
336 	    &V_ipsec_hhh_out[HHOOK_IPSEC_INET],
337 	    HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
338 		printf("%s: WARNING: unable to register output helper hook\n",
339 		    __func__);
340 
341 #ifdef VIMAGE
342 	netisr_register_vnet(&ip_nh);
343 #ifdef	RSS
344 	netisr_register_vnet(&ip_direct_nh);
345 #endif
346 #endif
347 }
348 VNET_SYSINIT(ip_vnet_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_FOURTH,
349     ip_vnet_init, NULL);
350 
351 static void
352 ip_init(const void *unused __unused)
353 {
354 
355 	ipreass_init();
356 
357 	/*
358 	 * Register statically compiled protocols, that are unlikely to
359 	 * ever become dynamic.
360 	 */
361 	IPPROTO_REGISTER(IPPROTO_ICMP, icmp_input, NULL);
362 	IPPROTO_REGISTER(IPPROTO_IGMP, igmp_input, NULL);
363 	IPPROTO_REGISTER(IPPROTO_RSVP, rsvp_input, NULL);
364 	IPPROTO_REGISTER(IPPROTO_IPV4, encap4_input, NULL);
365 	IPPROTO_REGISTER(IPPROTO_MOBILE, encap4_input, NULL);
366 	IPPROTO_REGISTER(IPPROTO_ETHERIP, encap4_input, NULL);
367 	IPPROTO_REGISTER(IPPROTO_GRE, encap4_input, NULL);
368 	IPPROTO_REGISTER(IPPROTO_IPV6, encap4_input, NULL);
369 	IPPROTO_REGISTER(IPPROTO_PIM, encap4_input, NULL);
370 #ifdef SCTP	/* XXX: has a loadable & static version */
371 	IPPROTO_REGISTER(IPPROTO_SCTP, sctp_input, sctp_ctlinput);
372 #endif
373 
374 	netisr_register(&ip_nh);
375 #ifdef	RSS
376 	netisr_register(&ip_direct_nh);
377 #endif
378 }
379 SYSINIT(ip_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip_init, NULL);
380 
381 #ifdef VIMAGE
382 static void
383 ip_destroy(void *unused __unused)
384 {
385 	int error;
386 
387 #ifdef	RSS
388 	netisr_unregister_vnet(&ip_direct_nh);
389 #endif
390 	netisr_unregister_vnet(&ip_nh);
391 
392 	pfil_head_unregister(V_inet_pfil_head);
393 	error = hhook_head_deregister(V_ipsec_hhh_in[HHOOK_IPSEC_INET]);
394 	if (error != 0) {
395 		printf("%s: WARNING: unable to deregister input helper hook "
396 		    "type HHOOK_TYPE_IPSEC_IN, id HHOOK_IPSEC_INET: "
397 		    "error %d returned\n", __func__, error);
398 	}
399 	error = hhook_head_deregister(V_ipsec_hhh_out[HHOOK_IPSEC_INET]);
400 	if (error != 0) {
401 		printf("%s: WARNING: unable to deregister output helper hook "
402 		    "type HHOOK_TYPE_IPSEC_OUT, id HHOOK_IPSEC_INET: "
403 		    "error %d returned\n", __func__, error);
404 	}
405 
406 	/* Remove the IPv4 addresses from all interfaces. */
407 	in_ifscrub_all();
408 
409 	/* Make sure the IPv4 routes are gone as well. */
410 	rib_flush_routes_family(AF_INET);
411 
412 	/* Destroy IP reassembly queue. */
413 	ipreass_destroy();
414 
415 	/* Cleanup in_ifaddr hash table; should be empty. */
416 	hashdestroy(V_in_ifaddrhashtbl, M_IFADDR, V_in_ifaddrhmask);
417 }
418 
419 VNET_SYSUNINIT(ip, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip_destroy, NULL);
420 #endif
421 
422 #ifdef	RSS
423 /*
424  * IP direct input routine.
425  *
426  * This is called when reinjecting completed fragments where
427  * all of the previous checking and book-keeping has been done.
428  */
429 void
430 ip_direct_input(struct mbuf *m)
431 {
432 	struct ip *ip;
433 	int hlen;
434 
435 	ip = mtod(m, struct ip *);
436 	hlen = ip->ip_hl << 2;
437 
438 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
439 	if (IPSEC_ENABLED(ipv4)) {
440 		if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
441 			return;
442 	}
443 #endif /* IPSEC */
444 	IPSTAT_INC(ips_delivered);
445 	ip_protox[ip->ip_p](&m, &hlen, ip->ip_p);
446 }
447 #endif
448 
449 /*
450  * Ip input routine.  Checksum and byte swap header.  If fragmented
451  * try to reassemble.  Process options.  Pass to next level.
452  */
453 void
454 ip_input(struct mbuf *m)
455 {
456 	struct ip *ip = NULL;
457 	struct in_ifaddr *ia = NULL;
458 	struct ifaddr *ifa;
459 	struct ifnet *ifp;
460 	int hlen = 0;
461 	uint16_t sum, ip_len;
462 	int dchg = 0;				/* dest changed after fw */
463 	struct in_addr odst;			/* original dst address */
464 	bool strong_es;
465 
466 	M_ASSERTPKTHDR(m);
467 	NET_EPOCH_ASSERT();
468 
469 	if (m->m_flags & M_FASTFWD_OURS) {
470 		m->m_flags &= ~M_FASTFWD_OURS;
471 		/* Set up some basics that will be used later. */
472 		ip = mtod(m, struct ip *);
473 		hlen = ip->ip_hl << 2;
474 		ip_len = ntohs(ip->ip_len);
475 		goto ours;
476 	}
477 
478 	IPSTAT_INC(ips_total);
479 
480 	if (__predict_false(m->m_pkthdr.len < sizeof(struct ip)))
481 		goto tooshort;
482 
483 	if (m->m_len < sizeof(struct ip)) {
484 		m = m_pullup(m, sizeof(struct ip));
485 		if (__predict_false(m == NULL)) {
486 			IPSTAT_INC(ips_toosmall);
487 			return;
488 		}
489 	}
490 	ip = mtod(m, struct ip *);
491 
492 	if (__predict_false(ip->ip_v != IPVERSION)) {
493 		IPSTAT_INC(ips_badvers);
494 		goto bad;
495 	}
496 
497 	hlen = ip->ip_hl << 2;
498 	if (__predict_false(hlen < sizeof(struct ip))) {	/* minimum header length */
499 		IPSTAT_INC(ips_badhlen);
500 		goto bad;
501 	}
502 	if (hlen > m->m_len) {
503 		m = m_pullup(m, hlen);
504 		if (__predict_false(m == NULL)) {
505 			IPSTAT_INC(ips_badhlen);
506 			return;
507 		}
508 		ip = mtod(m, struct ip *);
509 	}
510 
511 	IP_PROBE(receive, NULL, NULL, ip, m->m_pkthdr.rcvif, ip, NULL);
512 
513 	/* IN_LOOPBACK must not appear on the wire - RFC1122 */
514 	ifp = m->m_pkthdr.rcvif;
515 	if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) ||
516 	    IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) {
517 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
518 			IPSTAT_INC(ips_badaddr);
519 			goto bad;
520 		}
521 	}
522 
523 	if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
524 		sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
525 	} else {
526 		if (hlen == sizeof(struct ip)) {
527 			sum = in_cksum_hdr(ip);
528 		} else {
529 			sum = in_cksum(m, hlen);
530 		}
531 	}
532 	if (__predict_false(sum)) {
533 		IPSTAT_INC(ips_badsum);
534 		goto bad;
535 	}
536 
537 	ip_len = ntohs(ip->ip_len);
538 	if (__predict_false(ip_len < hlen)) {
539 		IPSTAT_INC(ips_badlen);
540 		goto bad;
541 	}
542 
543 	/*
544 	 * Check that the amount of data in the buffers
545 	 * is as at least much as the IP header would have us expect.
546 	 * Trim mbufs if longer than we expect.
547 	 * Drop packet if shorter than we expect.
548 	 */
549 	if (__predict_false(m->m_pkthdr.len < ip_len)) {
550 tooshort:
551 		IPSTAT_INC(ips_tooshort);
552 		goto bad;
553 	}
554 	if (m->m_pkthdr.len > ip_len) {
555 		if (m->m_len == m->m_pkthdr.len) {
556 			m->m_len = ip_len;
557 			m->m_pkthdr.len = ip_len;
558 		} else
559 			m_adj(m, ip_len - m->m_pkthdr.len);
560 	}
561 
562 	/*
563 	 * Try to forward the packet, but if we fail continue.
564 	 * ip_tryforward() may generate redirects these days.
565 	 * XXX the logic below falling through to normal processing
566 	 * if redirects are required should be revisited as well.
567 	 * ip_tryforward() does inbound and outbound packet firewall
568 	 * processing. If firewall has decided that destination becomes
569 	 * our local address, it sets M_FASTFWD_OURS flag. In this
570 	 * case skip another inbound firewall processing and update
571 	 * ip pointer.
572 	 */
573 	if (V_ipforwarding != 0
574 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
575 	    && (!IPSEC_ENABLED(ipv4) ||
576 	    IPSEC_CAPS(ipv4, m, IPSEC_CAP_OPERABLE) == 0)
577 #endif
578 	    ) {
579 		/*
580 		 * ip_dooptions() was run so we can ignore the source route (or
581 		 * any IP options case) case for redirects in ip_tryforward().
582 		 */
583 		if ((m = ip_tryforward(m)) == NULL)
584 			return;
585 		if (m->m_flags & M_FASTFWD_OURS) {
586 			m->m_flags &= ~M_FASTFWD_OURS;
587 			ip = mtod(m, struct ip *);
588 			goto ours;
589 		}
590 	}
591 
592 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
593 	/*
594 	 * Bypass packet filtering for packets previously handled by IPsec.
595 	 */
596 	if (IPSEC_ENABLED(ipv4) &&
597 	    IPSEC_CAPS(ipv4, m, IPSEC_CAP_BYPASS_FILTER) != 0)
598 			goto passin;
599 #endif
600 
601 	/*
602 	 * Run through list of hooks for input packets.
603 	 *
604 	 * NB: Beware of the destination address changing (e.g.
605 	 *     by NAT rewriting).  When this happens, tell
606 	 *     ip_forward to do the right thing.
607 	 */
608 
609 	/* Jump over all PFIL processing if hooks are not active. */
610 	if (!PFIL_HOOKED_IN(V_inet_pfil_head))
611 		goto passin;
612 
613 	odst = ip->ip_dst;
614 	if (pfil_mbuf_in(V_inet_pfil_head, &m, ifp, NULL) !=
615 	    PFIL_PASS)
616 		return;
617 	if (m == NULL)			/* consumed by filter */
618 		return;
619 
620 	ip = mtod(m, struct ip *);
621 	dchg = (odst.s_addr != ip->ip_dst.s_addr);
622 
623 	if (m->m_flags & M_FASTFWD_OURS) {
624 		m->m_flags &= ~M_FASTFWD_OURS;
625 		goto ours;
626 	}
627 	if (m->m_flags & M_IP_NEXTHOP) {
628 		if (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) {
629 			/*
630 			 * Directly ship the packet on.  This allows
631 			 * forwarding packets originally destined to us
632 			 * to some other directly connected host.
633 			 */
634 			ip_forward(m, 1);
635 			return;
636 		}
637 	}
638 passin:
639 
640 	/*
641 	 * Process options and, if not destined for us,
642 	 * ship it on.  ip_dooptions returns 1 when an
643 	 * error was detected (causing an icmp message
644 	 * to be sent and the original packet to be freed).
645 	 */
646 	if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
647 		return;
648 
649         /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
650          * matter if it is destined to another node, or whether it is
651          * a multicast one, RSVP wants it! and prevents it from being forwarded
652          * anywhere else. Also checks if the rsvp daemon is running before
653 	 * grabbing the packet.
654          */
655 	if (ip->ip_p == IPPROTO_RSVP && V_rsvp_on)
656 		goto ours;
657 
658 	/*
659 	 * Check our list of addresses, to see if the packet is for us.
660 	 * If we don't have any addresses, assume any unicast packet
661 	 * we receive might be for us (and let the upper layers deal
662 	 * with it).
663 	 */
664 	if (CK_STAILQ_EMPTY(&V_in_ifaddrhead) &&
665 	    (m->m_flags & (M_MCAST|M_BCAST)) == 0)
666 		goto ours;
667 
668 	/*
669 	 * Enable a consistency check between the destination address
670 	 * and the arrival interface for a unicast packet (the RFC 1122
671 	 * strong ES model) with a list of additional predicates:
672 	 * - if IP forwarding is disabled
673 	 * - the packet is not locally generated
674 	 * - the packet is not subject to 'ipfw fwd'
675 	 * - Interface is not running CARP. If the packet got here, we already
676 	 *   checked it with carp_iamatch() and carp_forus().
677 	 */
678 	strong_es = V_ip_strong_es && (V_ipforwarding == 0) &&
679 	    ((ifp->if_flags & IFF_LOOPBACK) == 0) &&
680 	    ifp->if_carp == NULL && (dchg == 0);
681 
682 	/*
683 	 * Check for exact addresses in the hash bucket.
684 	 */
685 	CK_LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
686 		if (IA_SIN(ia)->sin_addr.s_addr != ip->ip_dst.s_addr)
687 			continue;
688 
689 		/*
690 		 * net.inet.ip.rfc1122_strong_es: the address matches, verify
691 		 * that the packet arrived via the correct interface.
692 		 */
693 		if (__predict_false(strong_es && ia->ia_ifp != ifp)) {
694 			IPSTAT_INC(ips_badaddr);
695 			goto bad;
696 		}
697 
698 		/*
699 		 * net.inet.ip.source_address_validation: drop incoming
700 		 * packets that pretend to be ours.
701 		 */
702 		if (V_ip_sav && !(ifp->if_flags & IFF_LOOPBACK) &&
703 		    __predict_false(in_localip_fib(ip->ip_src, ifp->if_fib))) {
704 			IPSTAT_INC(ips_badaddr);
705 			goto bad;
706 		}
707 
708 		counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
709 		counter_u64_add(ia->ia_ifa.ifa_ibytes, m->m_pkthdr.len);
710 		goto ours;
711 	}
712 
713 	/*
714 	 * Check for broadcast addresses.
715 	 *
716 	 * Only accept broadcast packets that arrive via the matching
717 	 * interface.  Reception of forwarded directed broadcasts would
718 	 * be handled via ip_forward() and ether_output() with the loopback
719 	 * into the stack for SIMPLEX interfaces handled by ether_output().
720 	 */
721 	if (ifp->if_flags & IFF_BROADCAST) {
722 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
723 			if (ifa->ifa_addr->sa_family != AF_INET)
724 				continue;
725 			ia = ifatoia(ifa);
726 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
727 			    ip->ip_dst.s_addr) {
728 				counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
729 				counter_u64_add(ia->ia_ifa.ifa_ibytes,
730 				    m->m_pkthdr.len);
731 				goto ours;
732 			}
733 #ifdef BOOTP_COMPAT
734 			if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) {
735 				counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
736 				counter_u64_add(ia->ia_ifa.ifa_ibytes,
737 				    m->m_pkthdr.len);
738 				goto ours;
739 			}
740 #endif
741 		}
742 		ia = NULL;
743 	}
744 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
745 		/*
746 		 * RFC 3927 2.7: Do not forward multicast packets from
747 		 * IN_LINKLOCAL.
748 		 */
749 		if (V_ip_mrouter && !IN_LINKLOCAL(ntohl(ip->ip_src.s_addr))) {
750 			/*
751 			 * If we are acting as a multicast router, all
752 			 * incoming multicast packets are passed to the
753 			 * kernel-level multicast forwarding function.
754 			 * The packet is returned (relatively) intact; if
755 			 * ip_mforward() returns a non-zero value, the packet
756 			 * must be discarded, else it may be accepted below.
757 			 */
758 			if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) {
759 				IPSTAT_INC(ips_cantforward);
760 				m_freem(m);
761 				return;
762 			}
763 
764 			/*
765 			 * The process-level routing daemon needs to receive
766 			 * all multicast IGMP packets, whether or not this
767 			 * host belongs to their destination groups.
768 			 */
769 			if (ip->ip_p == IPPROTO_IGMP) {
770 				goto ours;
771 			}
772 			IPSTAT_INC(ips_forward);
773 		}
774 		/*
775 		 * Assume the packet is for us, to avoid prematurely taking
776 		 * a lock on the in_multi hash. Protocols must perform
777 		 * their own filtering and update statistics accordingly.
778 		 */
779 		goto ours;
780 	}
781 	if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
782 		goto ours;
783 	if (ip->ip_dst.s_addr == INADDR_ANY)
784 		goto ours;
785 	/* RFC 3927 2.7: Do not forward packets to or from IN_LINKLOCAL. */
786 	if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr)) ||
787 	    IN_LINKLOCAL(ntohl(ip->ip_src.s_addr))) {
788 		IPSTAT_INC(ips_cantforward);
789 		m_freem(m);
790 		return;
791 	}
792 
793 	/*
794 	 * Not for us; forward if possible and desirable.
795 	 */
796 	if (V_ipforwarding == 0) {
797 		IPSTAT_INC(ips_cantforward);
798 		m_freem(m);
799 	} else {
800 		ip_forward(m, dchg);
801 	}
802 	return;
803 
804 ours:
805 #ifdef IPSTEALTH
806 	/*
807 	 * IPSTEALTH: Process non-routing options only
808 	 * if the packet is destined for us.
809 	 */
810 	if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1))
811 		return;
812 #endif /* IPSTEALTH */
813 
814 	/*
815 	 * Attempt reassembly; if it succeeds, proceed.
816 	 * ip_reass() will return a different mbuf.
817 	 */
818 	if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) {
819 		/* XXXGL: shouldn't we save & set m_flags? */
820 		m = ip_reass(m);
821 		if (m == NULL)
822 			return;
823 		ip = mtod(m, struct ip *);
824 		/* Get the header length of the reassembled packet */
825 		hlen = ip->ip_hl << 2;
826 	}
827 
828 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
829 	if (IPSEC_ENABLED(ipv4)) {
830 		if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
831 			return;
832 	}
833 #endif /* IPSEC */
834 
835 	/*
836 	 * Switch out to protocol's input routine.
837 	 */
838 	IPSTAT_INC(ips_delivered);
839 
840 	ip_protox[ip->ip_p](&m, &hlen, ip->ip_p);
841 	return;
842 bad:
843 	m_freem(m);
844 }
845 
846 int
847 ipproto_register(uint8_t proto, ipproto_input_t input, ipproto_ctlinput_t ctl)
848 {
849 
850 	MPASS(proto > 0);
851 
852 	/*
853 	 * The protocol slot must not be occupied by another protocol
854 	 * already.  An index pointing to rip_input() is unused.
855 	 */
856 	if (ip_protox[proto] == rip_input) {
857 		ip_protox[proto] = input;
858 		ip_ctlprotox[proto] = ctl;
859 		return (0);
860 	} else
861 		return (EEXIST);
862 }
863 
864 int
865 ipproto_unregister(uint8_t proto)
866 {
867 
868 	MPASS(proto > 0);
869 
870 	if (ip_protox[proto] != rip_input) {
871 		ip_protox[proto] = rip_input;
872 		ip_ctlprotox[proto] = rip_ctlinput;
873 		return (0);
874 	} else
875 		return (ENOENT);
876 }
877 
878 /*
879  * Forward a packet.  If some error occurs return the sender
880  * an icmp packet.  Note we can't always generate a meaningful
881  * icmp message because icmp doesn't have a large enough repertoire
882  * of codes and types.
883  *
884  * If not forwarding, just drop the packet.  This could be confusing
885  * if ipforwarding was zero but some routing protocol was advancing
886  * us as a gateway to somewhere.  However, we must let the routing
887  * protocol deal with that.
888  *
889  * The srcrt parameter indicates whether the packet is being forwarded
890  * via a source route.
891  */
892 void
893 ip_forward(struct mbuf *m, int srcrt)
894 {
895 	struct ip *ip = mtod(m, struct ip *);
896 	struct in_ifaddr *ia;
897 	struct mbuf *mcopy;
898 	struct sockaddr_in *sin;
899 	struct in_addr dest;
900 	struct route ro;
901 	uint32_t flowid;
902 	int error, type = 0, code = 0, mtu = 0;
903 
904 	NET_EPOCH_ASSERT();
905 
906 	if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
907 		IPSTAT_INC(ips_cantforward);
908 		m_freem(m);
909 		return;
910 	}
911 	if (
912 #ifdef IPSTEALTH
913 	    V_ipstealth == 0 &&
914 #endif
915 	    ip->ip_ttl <= IPTTLDEC) {
916 		icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 0, 0);
917 		return;
918 	}
919 
920 	bzero(&ro, sizeof(ro));
921 	sin = (struct sockaddr_in *)&ro.ro_dst;
922 	sin->sin_family = AF_INET;
923 	sin->sin_len = sizeof(*sin);
924 	sin->sin_addr = ip->ip_dst;
925 	flowid = m->m_pkthdr.flowid;
926 	ro.ro_nh = fib4_lookup(M_GETFIB(m), ip->ip_dst, 0, NHR_REF, flowid);
927 	if (ro.ro_nh != NULL) {
928 		ia = ifatoia(ro.ro_nh->nh_ifa);
929 	} else
930 		ia = NULL;
931 	/*
932 	 * Save the IP header and at most 8 bytes of the payload,
933 	 * in case we need to generate an ICMP message to the src.
934 	 *
935 	 * XXX this can be optimized a lot by saving the data in a local
936 	 * buffer on the stack (72 bytes at most), and only allocating the
937 	 * mbuf if really necessary. The vast majority of the packets
938 	 * are forwarded without having to send an ICMP back (either
939 	 * because unnecessary, or because rate limited), so we are
940 	 * really we are wasting a lot of work here.
941 	 *
942 	 * We don't use m_copym() because it might return a reference
943 	 * to a shared cluster. Both this function and ip_output()
944 	 * assume exclusive access to the IP header in `m', so any
945 	 * data in a cluster may change before we reach icmp_error().
946 	 */
947 	mcopy = m_gethdr(M_NOWAIT, m->m_type);
948 	if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_NOWAIT)) {
949 		/*
950 		 * It's probably ok if the pkthdr dup fails (because
951 		 * the deep copy of the tag chain failed), but for now
952 		 * be conservative and just discard the copy since
953 		 * code below may some day want the tags.
954 		 */
955 		m_free(mcopy);
956 		mcopy = NULL;
957 	}
958 	if (mcopy != NULL) {
959 		mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy));
960 		mcopy->m_pkthdr.len = mcopy->m_len;
961 		m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
962 	}
963 #ifdef IPSTEALTH
964 	if (V_ipstealth == 0)
965 #endif
966 		ip->ip_ttl -= IPTTLDEC;
967 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
968 	if (IPSEC_ENABLED(ipv4)) {
969 		if ((error = IPSEC_FORWARD(ipv4, m)) != 0) {
970 			/* mbuf consumed by IPsec */
971 			RO_NHFREE(&ro);
972 			m_freem(mcopy);
973 			if (error != EINPROGRESS)
974 				IPSTAT_INC(ips_cantforward);
975 			return;
976 		}
977 		/* No IPsec processing required */
978 	}
979 #endif /* IPSEC */
980 	/*
981 	 * If forwarding packet using same interface that it came in on,
982 	 * perhaps should send a redirect to sender to shortcut a hop.
983 	 * Only send redirect if source is sending directly to us,
984 	 * and if packet was not source routed (or has any options).
985 	 * Also, don't send redirect if forwarding using a default route
986 	 * or a route modified by a redirect.
987 	 */
988 	dest.s_addr = 0;
989 	if (!srcrt && V_ipsendredirects &&
990 	    ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) {
991 		struct nhop_object *nh;
992 
993 		nh = ro.ro_nh;
994 
995 		if (nh != NULL && ((nh->nh_flags & (NHF_REDIRECT|NHF_DEFAULT)) == 0)) {
996 			struct in_ifaddr *nh_ia = (struct in_ifaddr *)(nh->nh_ifa);
997 			u_long src = ntohl(ip->ip_src.s_addr);
998 
999 			if (nh_ia != NULL &&
1000 			    (src & nh_ia->ia_subnetmask) == nh_ia->ia_subnet) {
1001 				/* Router requirements says to only send host redirects */
1002 				type = ICMP_REDIRECT;
1003 				code = ICMP_REDIRECT_HOST;
1004 				if (nh->nh_flags & NHF_GATEWAY) {
1005 				    if (nh->gw_sa.sa_family == AF_INET)
1006 					dest.s_addr = nh->gw4_sa.sin_addr.s_addr;
1007 				    else /* Do not redirect in case gw is AF_INET6 */
1008 					type = 0;
1009 				} else
1010 					dest.s_addr = ip->ip_dst.s_addr;
1011 			}
1012 		}
1013 	}
1014 
1015 	error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
1016 
1017 	if (error == EMSGSIZE && ro.ro_nh)
1018 		mtu = ro.ro_nh->nh_mtu;
1019 	RO_NHFREE(&ro);
1020 
1021 	if (error)
1022 		IPSTAT_INC(ips_cantforward);
1023 	else {
1024 		IPSTAT_INC(ips_forward);
1025 		if (type)
1026 			IPSTAT_INC(ips_redirectsent);
1027 		else {
1028 			if (mcopy)
1029 				m_freem(mcopy);
1030 			return;
1031 		}
1032 	}
1033 	if (mcopy == NULL)
1034 		return;
1035 
1036 	switch (error) {
1037 	case 0:				/* forwarded, but need redirect */
1038 		/* type, code set above */
1039 		break;
1040 
1041 	case ENETUNREACH:
1042 	case EHOSTUNREACH:
1043 	case ENETDOWN:
1044 	case EHOSTDOWN:
1045 	default:
1046 		type = ICMP_UNREACH;
1047 		code = ICMP_UNREACH_HOST;
1048 		break;
1049 
1050 	case EMSGSIZE:
1051 		type = ICMP_UNREACH;
1052 		code = ICMP_UNREACH_NEEDFRAG;
1053 		/*
1054 		 * If the MTU was set before make sure we are below the
1055 		 * interface MTU.
1056 		 * If the MTU wasn't set before use the interface mtu or
1057 		 * fall back to the next smaller mtu step compared to the
1058 		 * current packet size.
1059 		 */
1060 		if (mtu != 0) {
1061 			if (ia != NULL)
1062 				mtu = min(mtu, ia->ia_ifp->if_mtu);
1063 		} else {
1064 			if (ia != NULL)
1065 				mtu = ia->ia_ifp->if_mtu;
1066 			else
1067 				mtu = ip_next_mtu(ntohs(ip->ip_len), 0);
1068 		}
1069 		IPSTAT_INC(ips_cantfrag);
1070 		break;
1071 
1072 	case ENOBUFS:
1073 	case EACCES:			/* ipfw denied packet */
1074 		m_freem(mcopy);
1075 		return;
1076 	}
1077 	icmp_error(mcopy, type, code, dest.s_addr, mtu);
1078 }
1079 
1080 #define	CHECK_SO_CT(sp, ct) \
1081     (((sp->so_options & SO_TIMESTAMP) && (sp->so_ts_clock == ct)) ? 1 : 0)
1082 
1083 void
1084 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1085     struct mbuf *m)
1086 {
1087 	bool stamped;
1088 
1089 	stamped = false;
1090 	if ((inp->inp_socket->so_options & SO_BINTIME) ||
1091 	    CHECK_SO_CT(inp->inp_socket, SO_TS_BINTIME)) {
1092 		struct bintime boottimebin, bt;
1093 		struct timespec ts1;
1094 
1095 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1096 		    M_TSTMP)) {
1097 			mbuf_tstmp2timespec(m, &ts1);
1098 			timespec2bintime(&ts1, &bt);
1099 			getboottimebin(&boottimebin);
1100 			bintime_add(&bt, &boottimebin);
1101 		} else {
1102 			bintime(&bt);
1103 		}
1104 		*mp = sbcreatecontrol(&bt, sizeof(bt), SCM_BINTIME,
1105 		    SOL_SOCKET, M_NOWAIT);
1106 		if (*mp != NULL) {
1107 			mp = &(*mp)->m_next;
1108 			stamped = true;
1109 		}
1110 	}
1111 	if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME_MICRO)) {
1112 		struct bintime boottimebin, bt1;
1113 		struct timespec ts1;
1114 		struct timeval tv;
1115 
1116 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1117 		    M_TSTMP)) {
1118 			mbuf_tstmp2timespec(m, &ts1);
1119 			timespec2bintime(&ts1, &bt1);
1120 			getboottimebin(&boottimebin);
1121 			bintime_add(&bt1, &boottimebin);
1122 			bintime2timeval(&bt1, &tv);
1123 		} else {
1124 			microtime(&tv);
1125 		}
1126 		*mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv), SCM_TIMESTAMP,
1127 		    SOL_SOCKET, M_NOWAIT);
1128 		if (*mp != NULL) {
1129 			mp = &(*mp)->m_next;
1130 			stamped = true;
1131 		}
1132 	} else if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME)) {
1133 		struct bintime boottimebin;
1134 		struct timespec ts, ts1;
1135 
1136 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1137 		    M_TSTMP)) {
1138 			mbuf_tstmp2timespec(m, &ts);
1139 			getboottimebin(&boottimebin);
1140 			bintime2timespec(&boottimebin, &ts1);
1141 			timespecadd(&ts, &ts1, &ts);
1142 		} else {
1143 			nanotime(&ts);
1144 		}
1145 		*mp = sbcreatecontrol(&ts, sizeof(ts), SCM_REALTIME,
1146 		    SOL_SOCKET, M_NOWAIT);
1147 		if (*mp != NULL) {
1148 			mp = &(*mp)->m_next;
1149 			stamped = true;
1150 		}
1151 	} else if (CHECK_SO_CT(inp->inp_socket, SO_TS_MONOTONIC)) {
1152 		struct timespec ts;
1153 
1154 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1155 		    M_TSTMP))
1156 			mbuf_tstmp2timespec(m, &ts);
1157 		else
1158 			nanouptime(&ts);
1159 		*mp = sbcreatecontrol(&ts, sizeof(ts), SCM_MONOTONIC,
1160 		    SOL_SOCKET, M_NOWAIT);
1161 		if (*mp != NULL) {
1162 			mp = &(*mp)->m_next;
1163 			stamped = true;
1164 		}
1165 	}
1166 	if (stamped && (m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1167 	    M_TSTMP)) {
1168 		struct sock_timestamp_info sti;
1169 
1170 		bzero(&sti, sizeof(sti));
1171 		sti.st_info_flags = ST_INFO_HW;
1172 		if ((m->m_flags & M_TSTMP_HPREC) != 0)
1173 			sti.st_info_flags |= ST_INFO_HW_HPREC;
1174 		*mp = sbcreatecontrol(&sti, sizeof(sti), SCM_TIME_INFO,
1175 		    SOL_SOCKET, M_NOWAIT);
1176 		if (*mp != NULL)
1177 			mp = &(*mp)->m_next;
1178 	}
1179 	if (inp->inp_flags & INP_RECVDSTADDR) {
1180 		*mp = sbcreatecontrol(&ip->ip_dst, sizeof(struct in_addr),
1181 		    IP_RECVDSTADDR, IPPROTO_IP, M_NOWAIT);
1182 		if (*mp)
1183 			mp = &(*mp)->m_next;
1184 	}
1185 	if (inp->inp_flags & INP_RECVTTL) {
1186 		*mp = sbcreatecontrol(&ip->ip_ttl, sizeof(u_char), IP_RECVTTL,
1187 		    IPPROTO_IP, M_NOWAIT);
1188 		if (*mp)
1189 			mp = &(*mp)->m_next;
1190 	}
1191 #ifdef notyet
1192 	/* XXX
1193 	 * Moving these out of udp_input() made them even more broken
1194 	 * than they already were.
1195 	 */
1196 	/* options were tossed already */
1197 	if (inp->inp_flags & INP_RECVOPTS) {
1198 		*mp = sbcreatecontrol(opts_deleted_above,
1199 		    sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP, M_NOWAIT);
1200 		if (*mp)
1201 			mp = &(*mp)->m_next;
1202 	}
1203 	/* ip_srcroute doesn't do what we want here, need to fix */
1204 	if (inp->inp_flags & INP_RECVRETOPTS) {
1205 		*mp = sbcreatecontrol(ip_srcroute(m), sizeof(struct in_addr),
1206 		    IP_RECVRETOPTS, IPPROTO_IP, M_NOWAIT);
1207 		if (*mp)
1208 			mp = &(*mp)->m_next;
1209 	}
1210 #endif
1211 	if (inp->inp_flags & INP_RECVIF) {
1212 		struct ifnet *ifp;
1213 		struct sdlbuf {
1214 			struct sockaddr_dl sdl;
1215 			u_char	pad[32];
1216 		} sdlbuf;
1217 		struct sockaddr_dl *sdp;
1218 		struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1219 
1220 		if ((ifp = m->m_pkthdr.rcvif)) {
1221 			sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1222 			/*
1223 			 * Change our mind and don't try copy.
1224 			 */
1225 			if (sdp->sdl_family != AF_LINK ||
1226 			    sdp->sdl_len > sizeof(sdlbuf)) {
1227 				goto makedummy;
1228 			}
1229 			bcopy(sdp, sdl2, sdp->sdl_len);
1230 		} else {
1231 makedummy:
1232 			sdl2->sdl_len =
1233 			    offsetof(struct sockaddr_dl, sdl_data[0]);
1234 			sdl2->sdl_family = AF_LINK;
1235 			sdl2->sdl_index = 0;
1236 			sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1237 		}
1238 		*mp = sbcreatecontrol(sdl2, sdl2->sdl_len, IP_RECVIF,
1239 		    IPPROTO_IP, M_NOWAIT);
1240 		if (*mp)
1241 			mp = &(*mp)->m_next;
1242 	}
1243 	if (inp->inp_flags & INP_RECVTOS) {
1244 		*mp = sbcreatecontrol(&ip->ip_tos, sizeof(u_char), IP_RECVTOS,
1245 		    IPPROTO_IP, M_NOWAIT);
1246 		if (*mp)
1247 			mp = &(*mp)->m_next;
1248 	}
1249 
1250 	if (inp->inp_flags2 & INP_RECVFLOWID) {
1251 		uint32_t flowid, flow_type;
1252 
1253 		flowid = m->m_pkthdr.flowid;
1254 		flow_type = M_HASHTYPE_GET(m);
1255 
1256 		/*
1257 		 * XXX should handle the failure of one or the
1258 		 * other - don't populate both?
1259 		 */
1260 		*mp = sbcreatecontrol(&flowid, sizeof(uint32_t), IP_FLOWID,
1261 		    IPPROTO_IP, M_NOWAIT);
1262 		if (*mp)
1263 			mp = &(*mp)->m_next;
1264 		*mp = sbcreatecontrol(&flow_type, sizeof(uint32_t),
1265 		    IP_FLOWTYPE, IPPROTO_IP, M_NOWAIT);
1266 		if (*mp)
1267 			mp = &(*mp)->m_next;
1268 	}
1269 
1270 #ifdef	RSS
1271 	if (inp->inp_flags2 & INP_RECVRSSBUCKETID) {
1272 		uint32_t flowid, flow_type;
1273 		uint32_t rss_bucketid;
1274 
1275 		flowid = m->m_pkthdr.flowid;
1276 		flow_type = M_HASHTYPE_GET(m);
1277 
1278 		if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) {
1279 			*mp = sbcreatecontrol(&rss_bucketid, sizeof(uint32_t),
1280 			    IP_RSSBUCKETID, IPPROTO_IP, M_NOWAIT);
1281 			if (*mp)
1282 				mp = &(*mp)->m_next;
1283 		}
1284 	}
1285 #endif
1286 }
1287 
1288 /*
1289  * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
1290  * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
1291  * locking.  This code remains in ip_input.c as ip_mroute.c is optionally
1292  * compiled.
1293  */
1294 VNET_DEFINE_STATIC(int, ip_rsvp_on);
1295 VNET_DEFINE(struct socket *, ip_rsvpd);
1296 
1297 #define	V_ip_rsvp_on		VNET(ip_rsvp_on)
1298 
1299 int
1300 ip_rsvp_init(struct socket *so)
1301 {
1302 
1303 	if (V_ip_rsvpd != NULL)
1304 		return EADDRINUSE;
1305 
1306 	V_ip_rsvpd = so;
1307 	/*
1308 	 * This may seem silly, but we need to be sure we don't over-increment
1309 	 * the RSVP counter, in case something slips up.
1310 	 */
1311 	if (!V_ip_rsvp_on) {
1312 		V_ip_rsvp_on = 1;
1313 		V_rsvp_on++;
1314 	}
1315 
1316 	return 0;
1317 }
1318 
1319 int
1320 ip_rsvp_done(void)
1321 {
1322 
1323 	V_ip_rsvpd = NULL;
1324 	/*
1325 	 * This may seem silly, but we need to be sure we don't over-decrement
1326 	 * the RSVP counter, in case something slips up.
1327 	 */
1328 	if (V_ip_rsvp_on) {
1329 		V_ip_rsvp_on = 0;
1330 		V_rsvp_on--;
1331 	}
1332 	return 0;
1333 }
1334 
1335 int
1336 rsvp_input(struct mbuf **mp, int *offp, int proto)
1337 {
1338 	struct mbuf *m;
1339 
1340 	m = *mp;
1341 	*mp = NULL;
1342 
1343 	if (rsvp_input_p) { /* call the real one if loaded */
1344 		*mp = m;
1345 		rsvp_input_p(mp, offp, proto);
1346 		return (IPPROTO_DONE);
1347 	}
1348 
1349 	/* Can still get packets with rsvp_on = 0 if there is a local member
1350 	 * of the group to which the RSVP packet is addressed.  But in this
1351 	 * case we want to throw the packet away.
1352 	 */
1353 
1354 	if (!V_rsvp_on) {
1355 		m_freem(m);
1356 		return (IPPROTO_DONE);
1357 	}
1358 
1359 	if (V_ip_rsvpd != NULL) {
1360 		*mp = m;
1361 		rip_input(mp, offp, proto);
1362 		return (IPPROTO_DONE);
1363 	}
1364 	/* Drop the packet */
1365 	m_freem(m);
1366 	return (IPPROTO_DONE);
1367 }
1368