xref: /freebsd/sys/netinet/ip_input.c (revision f55bd0e5798dd17251302f73904da76bdfd257b9)
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 
43 #include <sys/param.h>
44 #include <sys/systm.h>
45 #include <sys/hhook.h>
46 #include <sys/mbuf.h>
47 #include <sys/malloc.h>
48 #include <sys/domain.h>
49 #include <sys/protosw.h>
50 #include <sys/socket.h>
51 #include <sys/time.h>
52 #include <sys/kernel.h>
53 #include <sys/lock.h>
54 #include <sys/rmlock.h>
55 #include <sys/rwlock.h>
56 #include <sys/sdt.h>
57 #include <sys/syslog.h>
58 #include <sys/sysctl.h>
59 
60 #include <net/if.h>
61 #include <net/if_types.h>
62 #include <net/if_var.h>
63 #include <net/if_dl.h>
64 #include <net/pfil.h>
65 #include <net/route.h>
66 #include <net/route/nhop.h>
67 #include <net/netisr.h>
68 #include <net/rss_config.h>
69 #include <net/vnet.h>
70 
71 #include <netinet/in.h>
72 #include <netinet/in_kdtrace.h>
73 #include <netinet/in_systm.h>
74 #include <netinet/in_var.h>
75 #include <netinet/ip.h>
76 #include <netinet/in_fib.h>
77 #include <netinet/in_pcb.h>
78 #include <netinet/ip_var.h>
79 #include <netinet/ip_fw.h>
80 #include <netinet/ip_icmp.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 
87 #include <netipsec/ipsec_support.h>
88 
89 #include <sys/socketvar.h>
90 
91 #include <security/mac/mac_framework.h>
92 
93 #ifdef CTASSERT
94 CTASSERT(sizeof(struct ip) == 20);
95 #endif
96 
97 /* IP reassembly functions are defined in ip_reass.c. */
98 extern void ipreass_init(void);
99 extern void ipreass_drain(void);
100 extern void ipreass_slowtimo(void);
101 #ifdef VIMAGE
102 extern void ipreass_destroy(void);
103 #endif
104 
105 struct rmlock in_ifaddr_lock;
106 RM_SYSINIT(in_ifaddr_lock, &in_ifaddr_lock, "in_ifaddr_lock");
107 
108 VNET_DEFINE(int, rsvp_on);
109 
110 VNET_DEFINE(int, ipforwarding);
111 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_VNET | CTLFLAG_RW,
112     &VNET_NAME(ipforwarding), 0,
113     "Enable IP forwarding between interfaces");
114 
115 /*
116  * Respond with an ICMP host redirect when we forward a packet out of
117  * the same interface on which it was received.  See RFC 792.
118  */
119 VNET_DEFINE(int, ipsendredirects) = 1;
120 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_VNET | CTLFLAG_RW,
121     &VNET_NAME(ipsendredirects), 0,
122     "Enable sending IP redirects");
123 
124 /*
125  * XXX - Setting ip_checkinterface mostly implements the receive side of
126  * the Strong ES model described in RFC 1122, but since the routing table
127  * and transmit implementation do not implement the Strong ES model,
128  * setting this to 1 results in an odd hybrid.
129  *
130  * XXX - ip_checkinterface currently must be disabled if you use ipnat
131  * to translate the destination address to another local interface.
132  *
133  * XXX - ip_checkinterface must be disabled if you add IP aliases
134  * to the loopback interface instead of the interface where the
135  * packets for those addresses are received.
136  */
137 VNET_DEFINE_STATIC(int, ip_checkinterface);
138 #define	V_ip_checkinterface	VNET(ip_checkinterface)
139 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_VNET | CTLFLAG_RW,
140     &VNET_NAME(ip_checkinterface), 0,
141     "Verify packet arrives on correct interface");
142 
143 VNET_DEFINE(pfil_head_t, inet_pfil_head);	/* Packet filter hooks */
144 
145 static struct netisr_handler ip_nh = {
146 	.nh_name = "ip",
147 	.nh_handler = ip_input,
148 	.nh_proto = NETISR_IP,
149 #ifdef	RSS
150 	.nh_m2cpuid = rss_soft_m2cpuid_v4,
151 	.nh_policy = NETISR_POLICY_CPU,
152 	.nh_dispatch = NETISR_DISPATCH_HYBRID,
153 #else
154 	.nh_policy = NETISR_POLICY_FLOW,
155 #endif
156 };
157 
158 #ifdef	RSS
159 /*
160  * Directly dispatched frames are currently assumed
161  * to have a flowid already calculated.
162  *
163  * It should likely have something that assert it
164  * actually has valid flow details.
165  */
166 static struct netisr_handler ip_direct_nh = {
167 	.nh_name = "ip_direct",
168 	.nh_handler = ip_direct_input,
169 	.nh_proto = NETISR_IP_DIRECT,
170 	.nh_m2cpuid = rss_soft_m2cpuid_v4,
171 	.nh_policy = NETISR_POLICY_CPU,
172 	.nh_dispatch = NETISR_DISPATCH_HYBRID,
173 };
174 #endif
175 
176 extern	struct domain inetdomain;
177 extern	struct protosw inetsw[];
178 u_char	ip_protox[IPPROTO_MAX];
179 VNET_DEFINE(struct in_ifaddrhead, in_ifaddrhead);  /* first inet address */
180 VNET_DEFINE(struct in_ifaddrhashhead *, in_ifaddrhashtbl); /* inet addr hash table  */
181 VNET_DEFINE(u_long, in_ifaddrhmask);		/* mask for hash table */
182 
183 #ifdef IPCTL_DEFMTU
184 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
185     &ip_mtu, 0, "Default MTU");
186 #endif
187 
188 #ifdef IPSTEALTH
189 VNET_DEFINE(int, ipstealth);
190 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_VNET | CTLFLAG_RW,
191     &VNET_NAME(ipstealth), 0,
192     "IP stealth mode, no TTL decrementation on forwarding");
193 #endif
194 
195 /*
196  * IP statistics are stored in the "array" of counter(9)s.
197  */
198 VNET_PCPUSTAT_DEFINE(struct ipstat, ipstat);
199 VNET_PCPUSTAT_SYSINIT(ipstat);
200 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, IPCTL_STATS, stats, struct ipstat, ipstat,
201     "IP statistics (struct ipstat, netinet/ip_var.h)");
202 
203 #ifdef VIMAGE
204 VNET_PCPUSTAT_SYSUNINIT(ipstat);
205 #endif /* VIMAGE */
206 
207 /*
208  * Kernel module interface for updating ipstat.  The argument is an index
209  * into ipstat treated as an array.
210  */
211 void
212 kmod_ipstat_inc(int statnum)
213 {
214 
215 	counter_u64_add(VNET(ipstat)[statnum], 1);
216 }
217 
218 void
219 kmod_ipstat_dec(int statnum)
220 {
221 
222 	counter_u64_add(VNET(ipstat)[statnum], -1);
223 }
224 
225 static int
226 sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS)
227 {
228 	int error, qlimit;
229 
230 	netisr_getqlimit(&ip_nh, &qlimit);
231 	error = sysctl_handle_int(oidp, &qlimit, 0, req);
232 	if (error || !req->newptr)
233 		return (error);
234 	if (qlimit < 1)
235 		return (EINVAL);
236 	return (netisr_setqlimit(&ip_nh, qlimit));
237 }
238 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen,
239     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 0,
240     sysctl_netinet_intr_queue_maxlen, "I",
241     "Maximum size of the IP input queue");
242 
243 static int
244 sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS)
245 {
246 	u_int64_t qdrops_long;
247 	int error, qdrops;
248 
249 	netisr_getqdrops(&ip_nh, &qdrops_long);
250 	qdrops = qdrops_long;
251 	error = sysctl_handle_int(oidp, &qdrops, 0, req);
252 	if (error || !req->newptr)
253 		return (error);
254 	if (qdrops != 0)
255 		return (EINVAL);
256 	netisr_clearqdrops(&ip_nh);
257 	return (0);
258 }
259 
260 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops,
261     CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE,
262     0, 0, sysctl_netinet_intr_queue_drops, "I",
263     "Number of packets dropped from the IP input queue");
264 
265 #ifdef	RSS
266 static int
267 sysctl_netinet_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS)
268 {
269 	int error, qlimit;
270 
271 	netisr_getqlimit(&ip_direct_nh, &qlimit);
272 	error = sysctl_handle_int(oidp, &qlimit, 0, req);
273 	if (error || !req->newptr)
274 		return (error);
275 	if (qlimit < 1)
276 		return (EINVAL);
277 	return (netisr_setqlimit(&ip_direct_nh, qlimit));
278 }
279 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRDQMAXLEN, intr_direct_queue_maxlen,
280     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
281     0, 0, sysctl_netinet_intr_direct_queue_maxlen,
282     "I", "Maximum size of the IP direct input queue");
283 
284 static int
285 sysctl_netinet_intr_direct_queue_drops(SYSCTL_HANDLER_ARGS)
286 {
287 	u_int64_t qdrops_long;
288 	int error, qdrops;
289 
290 	netisr_getqdrops(&ip_direct_nh, &qdrops_long);
291 	qdrops = qdrops_long;
292 	error = sysctl_handle_int(oidp, &qdrops, 0, req);
293 	if (error || !req->newptr)
294 		return (error);
295 	if (qdrops != 0)
296 		return (EINVAL);
297 	netisr_clearqdrops(&ip_direct_nh);
298 	return (0);
299 }
300 
301 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRDQDROPS, intr_direct_queue_drops,
302     CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, 0, 0,
303     sysctl_netinet_intr_direct_queue_drops, "I",
304     "Number of packets dropped from the IP direct input queue");
305 #endif	/* RSS */
306 
307 /*
308  * IP initialization: fill in IP protocol switch table.
309  * All protocols not implemented in kernel go to raw IP protocol handler.
310  */
311 void
312 ip_init(void)
313 {
314 	struct pfil_head_args args;
315 	struct protosw *pr;
316 	int i;
317 
318 	CK_STAILQ_INIT(&V_in_ifaddrhead);
319 	V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask);
320 
321 	/* Initialize IP reassembly queue. */
322 	ipreass_init();
323 
324 	/* Initialize packet filter hooks. */
325 	args.pa_version = PFIL_VERSION;
326 	args.pa_flags = PFIL_IN | PFIL_OUT;
327 	args.pa_type = PFIL_TYPE_IP4;
328 	args.pa_headname = PFIL_INET_NAME;
329 	V_inet_pfil_head = pfil_head_register(&args);
330 
331 	if (hhook_head_register(HHOOK_TYPE_IPSEC_IN, AF_INET,
332 	    &V_ipsec_hhh_in[HHOOK_IPSEC_INET],
333 	    HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
334 		printf("%s: WARNING: unable to register input helper hook\n",
335 		    __func__);
336 	if (hhook_head_register(HHOOK_TYPE_IPSEC_OUT, AF_INET,
337 	    &V_ipsec_hhh_out[HHOOK_IPSEC_INET],
338 	    HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
339 		printf("%s: WARNING: unable to register output helper hook\n",
340 		    __func__);
341 
342 	/* Skip initialization of globals for non-default instances. */
343 #ifdef VIMAGE
344 	if (!IS_DEFAULT_VNET(curvnet)) {
345 		netisr_register_vnet(&ip_nh);
346 #ifdef	RSS
347 		netisr_register_vnet(&ip_direct_nh);
348 #endif
349 		return;
350 	}
351 #endif
352 
353 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
354 	if (pr == NULL)
355 		panic("ip_init: PF_INET not found");
356 
357 	/* Initialize the entire ip_protox[] array to IPPROTO_RAW. */
358 	for (i = 0; i < IPPROTO_MAX; i++)
359 		ip_protox[i] = pr - inetsw;
360 	/*
361 	 * Cycle through IP protocols and put them into the appropriate place
362 	 * in ip_protox[].
363 	 */
364 	for (pr = inetdomain.dom_protosw;
365 	    pr < inetdomain.dom_protoswNPROTOSW; pr++)
366 		if (pr->pr_domain->dom_family == PF_INET &&
367 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
368 			/* Be careful to only index valid IP protocols. */
369 			if (pr->pr_protocol < IPPROTO_MAX)
370 				ip_protox[pr->pr_protocol] = pr - inetsw;
371 		}
372 
373 	netisr_register(&ip_nh);
374 #ifdef	RSS
375 	netisr_register(&ip_direct_nh);
376 #endif
377 }
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 	(*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
444 	return;
445 }
446 #endif
447 
448 /*
449  * Ip input routine.  Checksum and byte swap header.  If fragmented
450  * try to reassemble.  Process options.  Pass to next level.
451  */
452 void
453 ip_input(struct mbuf *m)
454 {
455 	MROUTER_RLOCK_TRACKER;
456 	struct rm_priotracker in_ifa_tracker;
457 	struct ip *ip = NULL;
458 	struct in_ifaddr *ia = NULL;
459 	struct ifaddr *ifa;
460 	struct ifnet *ifp;
461 	int    checkif, hlen = 0;
462 	uint16_t sum, ip_len;
463 	int dchg = 0;				/* dest changed after fw */
464 	struct in_addr odst;			/* original dst address */
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 (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))) == NULL) {
485 		IPSTAT_INC(ips_toosmall);
486 		return;
487 	}
488 	ip = mtod(m, struct ip *);
489 
490 	if (ip->ip_v != IPVERSION) {
491 		IPSTAT_INC(ips_badvers);
492 		goto bad;
493 	}
494 
495 	hlen = ip->ip_hl << 2;
496 	if (hlen < sizeof(struct ip)) {	/* minimum header length */
497 		IPSTAT_INC(ips_badhlen);
498 		goto bad;
499 	}
500 	if (hlen > m->m_len) {
501 		if ((m = m_pullup(m, hlen)) == NULL) {
502 			IPSTAT_INC(ips_badhlen);
503 			return;
504 		}
505 		ip = mtod(m, struct ip *);
506 	}
507 
508 	IP_PROBE(receive, NULL, NULL, ip, m->m_pkthdr.rcvif, ip, NULL);
509 
510 	/* IN_LOOPBACK must not appear on the wire - RFC1122 */
511 	ifp = m->m_pkthdr.rcvif;
512 	if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) ||
513 	    IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) {
514 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
515 			IPSTAT_INC(ips_badaddr);
516 			goto bad;
517 		}
518 	}
519 
520 	if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
521 		sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
522 	} else {
523 		if (hlen == sizeof(struct ip)) {
524 			sum = in_cksum_hdr(ip);
525 		} else {
526 			sum = in_cksum(m, hlen);
527 		}
528 	}
529 	if (sum) {
530 		IPSTAT_INC(ips_badsum);
531 		goto bad;
532 	}
533 
534 #ifdef ALTQ
535 	if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
536 		/* packet is dropped by traffic conditioner */
537 		return;
538 #endif
539 
540 	ip_len = ntohs(ip->ip_len);
541 	if (ip_len < hlen) {
542 		IPSTAT_INC(ips_badlen);
543 		goto bad;
544 	}
545 
546 	/*
547 	 * Check that the amount of data in the buffers
548 	 * is as at least much as the IP header would have us expect.
549 	 * Trim mbufs if longer than we expect.
550 	 * Drop packet if shorter than we expect.
551 	 */
552 	if (m->m_pkthdr.len < ip_len) {
553 tooshort:
554 		IPSTAT_INC(ips_tooshort);
555 		goto bad;
556 	}
557 	if (m->m_pkthdr.len > ip_len) {
558 		if (m->m_len == m->m_pkthdr.len) {
559 			m->m_len = ip_len;
560 			m->m_pkthdr.len = ip_len;
561 		} else
562 			m_adj(m, ip_len - m->m_pkthdr.len);
563 	}
564 
565 	/*
566 	 * Try to forward the packet, but if we fail continue.
567 	 * ip_tryforward() does not generate redirects, so fall
568 	 * through to normal processing if redirects are required.
569 	 * ip_tryforward() does inbound and outbound packet firewall
570 	 * processing. If firewall has decided that destination becomes
571 	 * our local address, it sets M_FASTFWD_OURS flag. In this
572 	 * case skip another inbound firewall processing and update
573 	 * ip pointer.
574 	 */
575 	if (V_ipforwarding != 0
576 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
577 	    && (!IPSEC_ENABLED(ipv4) ||
578 	    IPSEC_CAPS(ipv4, m, IPSEC_CAP_OPERABLE) == 0)
579 #endif
580 	    ) {
581 		if ((m = ip_tryforward(m)) == NULL)
582 			return;
583 		if (m->m_flags & M_FASTFWD_OURS) {
584 			m->m_flags &= ~M_FASTFWD_OURS;
585 			ip = mtod(m, struct ip *);
586 			goto ours;
587 		}
588 	}
589 
590 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
591 	/*
592 	 * Bypass packet filtering for packets previously handled by IPsec.
593 	 */
594 	if (IPSEC_ENABLED(ipv4) &&
595 	    IPSEC_CAPS(ipv4, m, IPSEC_CAP_BYPASS_FILTER) != 0)
596 			goto passin;
597 #endif
598 
599 	/*
600 	 * Run through list of hooks for input packets.
601 	 *
602 	 * NB: Beware of the destination address changing (e.g.
603 	 *     by NAT rewriting).  When this happens, tell
604 	 *     ip_forward to do the right thing.
605 	 */
606 
607 	/* Jump over all PFIL processing if hooks are not active. */
608 	if (!PFIL_HOOKED_IN(V_inet_pfil_head))
609 		goto passin;
610 
611 	odst = ip->ip_dst;
612 	if (pfil_run_hooks(V_inet_pfil_head, &m, ifp, PFIL_IN, NULL) !=
613 	    PFIL_PASS)
614 		return;
615 	if (m == NULL)			/* consumed by filter */
616 		return;
617 
618 	ip = mtod(m, struct ip *);
619 	dchg = (odst.s_addr != ip->ip_dst.s_addr);
620 	ifp = m->m_pkthdr.rcvif;
621 
622 	if (m->m_flags & M_FASTFWD_OURS) {
623 		m->m_flags &= ~M_FASTFWD_OURS;
624 		goto ours;
625 	}
626 	if (m->m_flags & M_IP_NEXTHOP) {
627 		if (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) {
628 			/*
629 			 * Directly ship the packet on.  This allows
630 			 * forwarding packets originally destined to us
631 			 * to some other directly connected host.
632 			 */
633 			ip_forward(m, 1);
634 			return;
635 		}
636 	}
637 passin:
638 
639 	/*
640 	 * Process options and, if not destined for us,
641 	 * ship it on.  ip_dooptions returns 1 when an
642 	 * error was detected (causing an icmp message
643 	 * to be sent and the original packet to be freed).
644 	 */
645 	if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
646 		return;
647 
648         /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
649          * matter if it is destined to another node, or whether it is
650          * a multicast one, RSVP wants it! and prevents it from being forwarded
651          * anywhere else. Also checks if the rsvp daemon is running before
652 	 * grabbing the packet.
653          */
654 	if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP)
655 		goto ours;
656 
657 	/*
658 	 * Check our list of addresses, to see if the packet is for us.
659 	 * If we don't have any addresses, assume any unicast packet
660 	 * we receive might be for us (and let the upper layers deal
661 	 * with it).
662 	 */
663 	if (CK_STAILQ_EMPTY(&V_in_ifaddrhead) &&
664 	    (m->m_flags & (M_MCAST|M_BCAST)) == 0)
665 		goto ours;
666 
667 	/*
668 	 * Enable a consistency check between the destination address
669 	 * and the arrival interface for a unicast packet (the RFC 1122
670 	 * strong ES model) if IP forwarding is disabled and the packet
671 	 * is not locally generated and the packet is not subject to
672 	 * 'ipfw fwd'.
673 	 *
674 	 * XXX - Checking also should be disabled if the destination
675 	 * address is ipnat'ed to a different interface.
676 	 *
677 	 * XXX - Checking is incompatible with IP aliases added
678 	 * to the loopback interface instead of the interface where
679 	 * the packets are received.
680 	 *
681 	 * XXX - This is the case for carp vhost IPs as well so we
682 	 * insert a workaround. If the packet got here, we already
683 	 * checked with carp_iamatch() and carp_forus().
684 	 */
685 	checkif = V_ip_checkinterface && (V_ipforwarding == 0) &&
686 	    ifp != NULL && ((ifp->if_flags & IFF_LOOPBACK) == 0) &&
687 	    ifp->if_carp == NULL && (dchg == 0);
688 
689 	/*
690 	 * Check for exact addresses in the hash bucket.
691 	 */
692 	IN_IFADDR_RLOCK(&in_ifa_tracker);
693 	LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
694 		/*
695 		 * If the address matches, verify that the packet
696 		 * arrived via the correct interface if checking is
697 		 * enabled.
698 		 */
699 		if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr &&
700 		    (!checkif || ia->ia_ifp == ifp)) {
701 			counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
702 			counter_u64_add(ia->ia_ifa.ifa_ibytes,
703 			    m->m_pkthdr.len);
704 			IN_IFADDR_RUNLOCK(&in_ifa_tracker);
705 			goto ours;
706 		}
707 	}
708 	IN_IFADDR_RUNLOCK(&in_ifa_tracker);
709 
710 	/*
711 	 * Check for broadcast addresses.
712 	 *
713 	 * Only accept broadcast packets that arrive via the matching
714 	 * interface.  Reception of forwarded directed broadcasts would
715 	 * be handled via ip_forward() and ether_output() with the loopback
716 	 * into the stack for SIMPLEX interfaces handled by ether_output().
717 	 */
718 	if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
719 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
720 			if (ifa->ifa_addr->sa_family != AF_INET)
721 				continue;
722 			ia = ifatoia(ifa);
723 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
724 			    ip->ip_dst.s_addr) {
725 				counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
726 				counter_u64_add(ia->ia_ifa.ifa_ibytes,
727 				    m->m_pkthdr.len);
728 				goto ours;
729 			}
730 #ifdef BOOTP_COMPAT
731 			if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) {
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 #endif
738 		}
739 		ia = NULL;
740 	}
741 	/* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */
742 	if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) {
743 		IPSTAT_INC(ips_cantforward);
744 		m_freem(m);
745 		return;
746 	}
747 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
748 		MROUTER_RLOCK();
749 		if (V_ip_mrouter) {
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 				MROUTER_RUNLOCK();
760 				IPSTAT_INC(ips_cantforward);
761 				m_freem(m);
762 				return;
763 			}
764 
765 			/*
766 			 * The process-level routing daemon needs to receive
767 			 * all multicast IGMP packets, whether or not this
768 			 * host belongs to their destination groups.
769 			 */
770 			if (ip->ip_p == IPPROTO_IGMP) {
771 				MROUTER_RUNLOCK();
772 				goto ours;
773 			}
774 			IPSTAT_INC(ips_forward);
775 		}
776 		MROUTER_RUNLOCK();
777 		/*
778 		 * Assume the packet is for us, to avoid prematurely taking
779 		 * a lock on the in_multi hash. Protocols must perform
780 		 * their own filtering and update statistics accordingly.
781 		 */
782 		goto ours;
783 	}
784 	if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
785 		goto ours;
786 	if (ip->ip_dst.s_addr == INADDR_ANY)
787 		goto ours;
788 
789 	/*
790 	 * Not for us; forward if possible and desirable.
791 	 */
792 	if (V_ipforwarding == 0) {
793 		IPSTAT_INC(ips_cantforward);
794 		m_freem(m);
795 	} else {
796 		ip_forward(m, dchg);
797 	}
798 	return;
799 
800 ours:
801 #ifdef IPSTEALTH
802 	/*
803 	 * IPSTEALTH: Process non-routing options only
804 	 * if the packet is destined for us.
805 	 */
806 	if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1))
807 		return;
808 #endif /* IPSTEALTH */
809 
810 	/*
811 	 * Attempt reassembly; if it succeeds, proceed.
812 	 * ip_reass() will return a different mbuf.
813 	 */
814 	if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) {
815 		/* XXXGL: shouldn't we save & set m_flags? */
816 		m = ip_reass(m);
817 		if (m == NULL)
818 			return;
819 		ip = mtod(m, struct ip *);
820 		/* Get the header length of the reassembled packet */
821 		hlen = ip->ip_hl << 2;
822 	}
823 
824 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
825 	if (IPSEC_ENABLED(ipv4)) {
826 		if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
827 			return;
828 	}
829 #endif /* IPSEC */
830 
831 	/*
832 	 * Switch out to protocol's input routine.
833 	 */
834 	IPSTAT_INC(ips_delivered);
835 
836 	(*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
837 	return;
838 bad:
839 	m_freem(m);
840 }
841 
842 /*
843  * IP timer processing;
844  * if a timer expires on a reassembly
845  * queue, discard it.
846  */
847 void
848 ip_slowtimo(void)
849 {
850 	VNET_ITERATOR_DECL(vnet_iter);
851 
852 	VNET_LIST_RLOCK_NOSLEEP();
853 	VNET_FOREACH(vnet_iter) {
854 		CURVNET_SET(vnet_iter);
855 		ipreass_slowtimo();
856 		CURVNET_RESTORE();
857 	}
858 	VNET_LIST_RUNLOCK_NOSLEEP();
859 }
860 
861 void
862 ip_drain(void)
863 {
864 	VNET_ITERATOR_DECL(vnet_iter);
865 
866 	VNET_LIST_RLOCK_NOSLEEP();
867 	VNET_FOREACH(vnet_iter) {
868 		CURVNET_SET(vnet_iter);
869 		ipreass_drain();
870 		CURVNET_RESTORE();
871 	}
872 	VNET_LIST_RUNLOCK_NOSLEEP();
873 }
874 
875 /*
876  * The protocol to be inserted into ip_protox[] must be already registered
877  * in inetsw[], either statically or through pf_proto_register().
878  */
879 int
880 ipproto_register(short ipproto)
881 {
882 	struct protosw *pr;
883 
884 	/* Sanity checks. */
885 	if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
886 		return (EPROTONOSUPPORT);
887 
888 	/*
889 	 * The protocol slot must not be occupied by another protocol
890 	 * already.  An index pointing to IPPROTO_RAW is unused.
891 	 */
892 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
893 	if (pr == NULL)
894 		return (EPFNOSUPPORT);
895 	if (ip_protox[ipproto] != pr - inetsw)	/* IPPROTO_RAW */
896 		return (EEXIST);
897 
898 	/* Find the protocol position in inetsw[] and set the index. */
899 	for (pr = inetdomain.dom_protosw;
900 	     pr < inetdomain.dom_protoswNPROTOSW; pr++) {
901 		if (pr->pr_domain->dom_family == PF_INET &&
902 		    pr->pr_protocol && pr->pr_protocol == ipproto) {
903 			ip_protox[pr->pr_protocol] = pr - inetsw;
904 			return (0);
905 		}
906 	}
907 	return (EPROTONOSUPPORT);
908 }
909 
910 int
911 ipproto_unregister(short ipproto)
912 {
913 	struct protosw *pr;
914 
915 	/* Sanity checks. */
916 	if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
917 		return (EPROTONOSUPPORT);
918 
919 	/* Check if the protocol was indeed registered. */
920 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
921 	if (pr == NULL)
922 		return (EPFNOSUPPORT);
923 	if (ip_protox[ipproto] == pr - inetsw)  /* IPPROTO_RAW */
924 		return (ENOENT);
925 
926 	/* Reset the protocol slot to IPPROTO_RAW. */
927 	ip_protox[ipproto] = pr - inetsw;
928 	return (0);
929 }
930 
931 u_char inetctlerrmap[PRC_NCMDS] = {
932 	0,		0,		0,		0,
933 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
934 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
935 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
936 	0,		0,		EHOSTUNREACH,	0,
937 	ENOPROTOOPT,	ECONNREFUSED
938 };
939 
940 /*
941  * Forward a packet.  If some error occurs return the sender
942  * an icmp packet.  Note we can't always generate a meaningful
943  * icmp message because icmp doesn't have a large enough repertoire
944  * of codes and types.
945  *
946  * If not forwarding, just drop the packet.  This could be confusing
947  * if ipforwarding was zero but some routing protocol was advancing
948  * us as a gateway to somewhere.  However, we must let the routing
949  * protocol deal with that.
950  *
951  * The srcrt parameter indicates whether the packet is being forwarded
952  * via a source route.
953  */
954 void
955 ip_forward(struct mbuf *m, int srcrt)
956 {
957 	struct ip *ip = mtod(m, struct ip *);
958 	struct in_ifaddr *ia;
959 	struct mbuf *mcopy;
960 	struct sockaddr_in *sin;
961 	struct in_addr dest;
962 	struct route ro;
963 	uint32_t flowid;
964 	int error, type = 0, code = 0, mtu = 0;
965 
966 	NET_EPOCH_ASSERT();
967 
968 	if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
969 		IPSTAT_INC(ips_cantforward);
970 		m_freem(m);
971 		return;
972 	}
973 	if (
974 #ifdef IPSTEALTH
975 	    V_ipstealth == 0 &&
976 #endif
977 	    ip->ip_ttl <= IPTTLDEC) {
978 		icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 0, 0);
979 		return;
980 	}
981 
982 	bzero(&ro, sizeof(ro));
983 	sin = (struct sockaddr_in *)&ro.ro_dst;
984 	sin->sin_family = AF_INET;
985 	sin->sin_len = sizeof(*sin);
986 	sin->sin_addr = ip->ip_dst;
987 	flowid = m->m_pkthdr.flowid;
988 	ro.ro_nh = fib4_lookup(M_GETFIB(m), ip->ip_dst, 0, NHR_REF, flowid);
989 	if (ro.ro_nh != NULL) {
990 		ia = ifatoia(ro.ro_nh->nh_ifa);
991 	} else
992 		ia = NULL;
993 	/*
994 	 * Save the IP header and at most 8 bytes of the payload,
995 	 * in case we need to generate an ICMP message to the src.
996 	 *
997 	 * XXX this can be optimized a lot by saving the data in a local
998 	 * buffer on the stack (72 bytes at most), and only allocating the
999 	 * mbuf if really necessary. The vast majority of the packets
1000 	 * are forwarded without having to send an ICMP back (either
1001 	 * because unnecessary, or because rate limited), so we are
1002 	 * really we are wasting a lot of work here.
1003 	 *
1004 	 * We don't use m_copym() because it might return a reference
1005 	 * to a shared cluster. Both this function and ip_output()
1006 	 * assume exclusive access to the IP header in `m', so any
1007 	 * data in a cluster may change before we reach icmp_error().
1008 	 */
1009 	mcopy = m_gethdr(M_NOWAIT, m->m_type);
1010 	if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_NOWAIT)) {
1011 		/*
1012 		 * It's probably ok if the pkthdr dup fails (because
1013 		 * the deep copy of the tag chain failed), but for now
1014 		 * be conservative and just discard the copy since
1015 		 * code below may some day want the tags.
1016 		 */
1017 		m_free(mcopy);
1018 		mcopy = NULL;
1019 	}
1020 	if (mcopy != NULL) {
1021 		mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy));
1022 		mcopy->m_pkthdr.len = mcopy->m_len;
1023 		m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1024 	}
1025 #ifdef IPSTEALTH
1026 	if (V_ipstealth == 0)
1027 #endif
1028 		ip->ip_ttl -= IPTTLDEC;
1029 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1030 	if (IPSEC_ENABLED(ipv4)) {
1031 		if ((error = IPSEC_FORWARD(ipv4, m)) != 0) {
1032 			/* mbuf consumed by IPsec */
1033 			RO_NHFREE(&ro);
1034 			m_freem(mcopy);
1035 			if (error != EINPROGRESS)
1036 				IPSTAT_INC(ips_cantforward);
1037 			return;
1038 		}
1039 		/* No IPsec processing required */
1040 	}
1041 #endif /* IPSEC */
1042 	/*
1043 	 * If forwarding packet using same interface that it came in on,
1044 	 * perhaps should send a redirect to sender to shortcut a hop.
1045 	 * Only send redirect if source is sending directly to us,
1046 	 * and if packet was not source routed (or has any options).
1047 	 * Also, don't send redirect if forwarding using a default route
1048 	 * or a route modified by a redirect.
1049 	 */
1050 	dest.s_addr = 0;
1051 	if (!srcrt && V_ipsendredirects &&
1052 	    ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) {
1053 		struct nhop_object *nh;
1054 
1055 		nh = ro.ro_nh;
1056 
1057 		if (nh != NULL && ((nh->nh_flags & (NHF_REDIRECT|NHF_DEFAULT)) == 0)) {
1058 			struct in_ifaddr *nh_ia = (struct in_ifaddr *)(nh->nh_ifa);
1059 			u_long src = ntohl(ip->ip_src.s_addr);
1060 
1061 			if (nh_ia != NULL &&
1062 			    (src & nh_ia->ia_subnetmask) == nh_ia->ia_subnet) {
1063 				if (nh->nh_flags & NHF_GATEWAY)
1064 					dest.s_addr = nh->gw4_sa.sin_addr.s_addr;
1065 				else
1066 					dest.s_addr = ip->ip_dst.s_addr;
1067 				/* Router requirements says to only send host redirects */
1068 				type = ICMP_REDIRECT;
1069 				code = ICMP_REDIRECT_HOST;
1070 			}
1071 		}
1072 	}
1073 
1074 	error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
1075 
1076 	if (error == EMSGSIZE && ro.ro_nh)
1077 		mtu = ro.ro_nh->nh_mtu;
1078 	RO_NHFREE(&ro);
1079 
1080 	if (error)
1081 		IPSTAT_INC(ips_cantforward);
1082 	else {
1083 		IPSTAT_INC(ips_forward);
1084 		if (type)
1085 			IPSTAT_INC(ips_redirectsent);
1086 		else {
1087 			if (mcopy)
1088 				m_freem(mcopy);
1089 			return;
1090 		}
1091 	}
1092 	if (mcopy == NULL)
1093 		return;
1094 
1095 	switch (error) {
1096 	case 0:				/* forwarded, but need redirect */
1097 		/* type, code set above */
1098 		break;
1099 
1100 	case ENETUNREACH:
1101 	case EHOSTUNREACH:
1102 	case ENETDOWN:
1103 	case EHOSTDOWN:
1104 	default:
1105 		type = ICMP_UNREACH;
1106 		code = ICMP_UNREACH_HOST;
1107 		break;
1108 
1109 	case EMSGSIZE:
1110 		type = ICMP_UNREACH;
1111 		code = ICMP_UNREACH_NEEDFRAG;
1112 		/*
1113 		 * If the MTU was set before make sure we are below the
1114 		 * interface MTU.
1115 		 * If the MTU wasn't set before use the interface mtu or
1116 		 * fall back to the next smaller mtu step compared to the
1117 		 * current packet size.
1118 		 */
1119 		if (mtu != 0) {
1120 			if (ia != NULL)
1121 				mtu = min(mtu, ia->ia_ifp->if_mtu);
1122 		} else {
1123 			if (ia != NULL)
1124 				mtu = ia->ia_ifp->if_mtu;
1125 			else
1126 				mtu = ip_next_mtu(ntohs(ip->ip_len), 0);
1127 		}
1128 		IPSTAT_INC(ips_cantfrag);
1129 		break;
1130 
1131 	case ENOBUFS:
1132 	case EACCES:			/* ipfw denied packet */
1133 		m_freem(mcopy);
1134 		return;
1135 	}
1136 	icmp_error(mcopy, type, code, dest.s_addr, mtu);
1137 }
1138 
1139 #define	CHECK_SO_CT(sp, ct) \
1140     (((sp->so_options & SO_TIMESTAMP) && (sp->so_ts_clock == ct)) ? 1 : 0)
1141 
1142 void
1143 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1144     struct mbuf *m)
1145 {
1146 	bool stamped;
1147 
1148 	stamped = false;
1149 	if ((inp->inp_socket->so_options & SO_BINTIME) ||
1150 	    CHECK_SO_CT(inp->inp_socket, SO_TS_BINTIME)) {
1151 		struct bintime boottimebin, bt;
1152 		struct timespec ts1;
1153 
1154 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1155 		    M_TSTMP)) {
1156 			mbuf_tstmp2timespec(m, &ts1);
1157 			timespec2bintime(&ts1, &bt);
1158 			getboottimebin(&boottimebin);
1159 			bintime_add(&bt, &boottimebin);
1160 		} else {
1161 			bintime(&bt);
1162 		}
1163 		*mp = sbcreatecontrol((caddr_t)&bt, sizeof(bt),
1164 		    SCM_BINTIME, SOL_SOCKET);
1165 		if (*mp != NULL) {
1166 			mp = &(*mp)->m_next;
1167 			stamped = true;
1168 		}
1169 	}
1170 	if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME_MICRO)) {
1171 		struct bintime boottimebin, bt1;
1172 		struct timespec ts1;
1173 		struct timeval tv;
1174 
1175 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1176 		    M_TSTMP)) {
1177 			mbuf_tstmp2timespec(m, &ts1);
1178 			timespec2bintime(&ts1, &bt1);
1179 			getboottimebin(&boottimebin);
1180 			bintime_add(&bt1, &boottimebin);
1181 			bintime2timeval(&bt1, &tv);
1182 		} else {
1183 			microtime(&tv);
1184 		}
1185 		*mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv),
1186 		    SCM_TIMESTAMP, SOL_SOCKET);
1187 		if (*mp != NULL) {
1188 			mp = &(*mp)->m_next;
1189 			stamped = true;
1190 		}
1191 	} else if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME)) {
1192 		struct bintime boottimebin;
1193 		struct timespec ts, ts1;
1194 
1195 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1196 		    M_TSTMP)) {
1197 			mbuf_tstmp2timespec(m, &ts);
1198 			getboottimebin(&boottimebin);
1199 			bintime2timespec(&boottimebin, &ts1);
1200 			timespecadd(&ts, &ts1, &ts);
1201 		} else {
1202 			nanotime(&ts);
1203 		}
1204 		*mp = sbcreatecontrol((caddr_t)&ts, sizeof(ts),
1205 		    SCM_REALTIME, SOL_SOCKET);
1206 		if (*mp != NULL) {
1207 			mp = &(*mp)->m_next;
1208 			stamped = true;
1209 		}
1210 	} else if (CHECK_SO_CT(inp->inp_socket, SO_TS_MONOTONIC)) {
1211 		struct timespec ts;
1212 
1213 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1214 		    M_TSTMP))
1215 			mbuf_tstmp2timespec(m, &ts);
1216 		else
1217 			nanouptime(&ts);
1218 		*mp = sbcreatecontrol((caddr_t)&ts, sizeof(ts),
1219 		    SCM_MONOTONIC, SOL_SOCKET);
1220 		if (*mp != NULL) {
1221 			mp = &(*mp)->m_next;
1222 			stamped = true;
1223 		}
1224 	}
1225 	if (stamped && (m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1226 	    M_TSTMP)) {
1227 		struct sock_timestamp_info sti;
1228 
1229 		bzero(&sti, sizeof(sti));
1230 		sti.st_info_flags = ST_INFO_HW;
1231 		if ((m->m_flags & M_TSTMP_HPREC) != 0)
1232 			sti.st_info_flags |= ST_INFO_HW_HPREC;
1233 		*mp = sbcreatecontrol((caddr_t)&sti, sizeof(sti), SCM_TIME_INFO,
1234 		    SOL_SOCKET);
1235 		if (*mp != NULL)
1236 			mp = &(*mp)->m_next;
1237 	}
1238 	if (inp->inp_flags & INP_RECVDSTADDR) {
1239 		*mp = sbcreatecontrol((caddr_t)&ip->ip_dst,
1240 		    sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1241 		if (*mp)
1242 			mp = &(*mp)->m_next;
1243 	}
1244 	if (inp->inp_flags & INP_RECVTTL) {
1245 		*mp = sbcreatecontrol((caddr_t)&ip->ip_ttl,
1246 		    sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
1247 		if (*mp)
1248 			mp = &(*mp)->m_next;
1249 	}
1250 #ifdef notyet
1251 	/* XXX
1252 	 * Moving these out of udp_input() made them even more broken
1253 	 * than they already were.
1254 	 */
1255 	/* options were tossed already */
1256 	if (inp->inp_flags & INP_RECVOPTS) {
1257 		*mp = sbcreatecontrol((caddr_t)opts_deleted_above,
1258 		    sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1259 		if (*mp)
1260 			mp = &(*mp)->m_next;
1261 	}
1262 	/* ip_srcroute doesn't do what we want here, need to fix */
1263 	if (inp->inp_flags & INP_RECVRETOPTS) {
1264 		*mp = sbcreatecontrol((caddr_t)ip_srcroute(m),
1265 		    sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1266 		if (*mp)
1267 			mp = &(*mp)->m_next;
1268 	}
1269 #endif
1270 	if (inp->inp_flags & INP_RECVIF) {
1271 		struct ifnet *ifp;
1272 		struct sdlbuf {
1273 			struct sockaddr_dl sdl;
1274 			u_char	pad[32];
1275 		} sdlbuf;
1276 		struct sockaddr_dl *sdp;
1277 		struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1278 
1279 		if ((ifp = m->m_pkthdr.rcvif) &&
1280 		    ifp->if_index && ifp->if_index <= V_if_index) {
1281 			sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1282 			/*
1283 			 * Change our mind and don't try copy.
1284 			 */
1285 			if (sdp->sdl_family != AF_LINK ||
1286 			    sdp->sdl_len > sizeof(sdlbuf)) {
1287 				goto makedummy;
1288 			}
1289 			bcopy(sdp, sdl2, sdp->sdl_len);
1290 		} else {
1291 makedummy:
1292 			sdl2->sdl_len =
1293 			    offsetof(struct sockaddr_dl, sdl_data[0]);
1294 			sdl2->sdl_family = AF_LINK;
1295 			sdl2->sdl_index = 0;
1296 			sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1297 		}
1298 		*mp = sbcreatecontrol((caddr_t)sdl2, sdl2->sdl_len,
1299 		    IP_RECVIF, IPPROTO_IP);
1300 		if (*mp)
1301 			mp = &(*mp)->m_next;
1302 	}
1303 	if (inp->inp_flags & INP_RECVTOS) {
1304 		*mp = sbcreatecontrol((caddr_t)&ip->ip_tos,
1305 		    sizeof(u_char), IP_RECVTOS, IPPROTO_IP);
1306 		if (*mp)
1307 			mp = &(*mp)->m_next;
1308 	}
1309 
1310 	if (inp->inp_flags2 & INP_RECVFLOWID) {
1311 		uint32_t flowid, flow_type;
1312 
1313 		flowid = m->m_pkthdr.flowid;
1314 		flow_type = M_HASHTYPE_GET(m);
1315 
1316 		/*
1317 		 * XXX should handle the failure of one or the
1318 		 * other - don't populate both?
1319 		 */
1320 		*mp = sbcreatecontrol((caddr_t) &flowid,
1321 		    sizeof(uint32_t), IP_FLOWID, IPPROTO_IP);
1322 		if (*mp)
1323 			mp = &(*mp)->m_next;
1324 		*mp = sbcreatecontrol((caddr_t) &flow_type,
1325 		    sizeof(uint32_t), IP_FLOWTYPE, IPPROTO_IP);
1326 		if (*mp)
1327 			mp = &(*mp)->m_next;
1328 	}
1329 
1330 #ifdef	RSS
1331 	if (inp->inp_flags2 & INP_RECVRSSBUCKETID) {
1332 		uint32_t flowid, flow_type;
1333 		uint32_t rss_bucketid;
1334 
1335 		flowid = m->m_pkthdr.flowid;
1336 		flow_type = M_HASHTYPE_GET(m);
1337 
1338 		if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) {
1339 			*mp = sbcreatecontrol((caddr_t) &rss_bucketid,
1340 			   sizeof(uint32_t), IP_RSSBUCKETID, IPPROTO_IP);
1341 			if (*mp)
1342 				mp = &(*mp)->m_next;
1343 		}
1344 	}
1345 #endif
1346 }
1347 
1348 /*
1349  * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
1350  * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
1351  * locking.  This code remains in ip_input.c as ip_mroute.c is optionally
1352  * compiled.
1353  */
1354 VNET_DEFINE_STATIC(int, ip_rsvp_on);
1355 VNET_DEFINE(struct socket *, ip_rsvpd);
1356 
1357 #define	V_ip_rsvp_on		VNET(ip_rsvp_on)
1358 
1359 int
1360 ip_rsvp_init(struct socket *so)
1361 {
1362 
1363 	if (so->so_type != SOCK_RAW ||
1364 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
1365 		return EOPNOTSUPP;
1366 
1367 	if (V_ip_rsvpd != NULL)
1368 		return EADDRINUSE;
1369 
1370 	V_ip_rsvpd = so;
1371 	/*
1372 	 * This may seem silly, but we need to be sure we don't over-increment
1373 	 * the RSVP counter, in case something slips up.
1374 	 */
1375 	if (!V_ip_rsvp_on) {
1376 		V_ip_rsvp_on = 1;
1377 		V_rsvp_on++;
1378 	}
1379 
1380 	return 0;
1381 }
1382 
1383 int
1384 ip_rsvp_done(void)
1385 {
1386 
1387 	V_ip_rsvpd = NULL;
1388 	/*
1389 	 * This may seem silly, but we need to be sure we don't over-decrement
1390 	 * the RSVP counter, in case something slips up.
1391 	 */
1392 	if (V_ip_rsvp_on) {
1393 		V_ip_rsvp_on = 0;
1394 		V_rsvp_on--;
1395 	}
1396 	return 0;
1397 }
1398 
1399 int
1400 rsvp_input(struct mbuf **mp, int *offp, int proto)
1401 {
1402 	struct mbuf *m;
1403 
1404 	m = *mp;
1405 	*mp = NULL;
1406 
1407 	if (rsvp_input_p) { /* call the real one if loaded */
1408 		*mp = m;
1409 		rsvp_input_p(mp, offp, proto);
1410 		return (IPPROTO_DONE);
1411 	}
1412 
1413 	/* Can still get packets with rsvp_on = 0 if there is a local member
1414 	 * of the group to which the RSVP packet is addressed.  But in this
1415 	 * case we want to throw the packet away.
1416 	 */
1417 
1418 	if (!V_rsvp_on) {
1419 		m_freem(m);
1420 		return (IPPROTO_DONE);
1421 	}
1422 
1423 	if (V_ip_rsvpd != NULL) {
1424 		*mp = m;
1425 		rip_input(mp, offp, proto);
1426 		return (IPPROTO_DONE);
1427 	}
1428 	/* Drop the packet */
1429 	m_freem(m);
1430 	return (IPPROTO_DONE);
1431 }
1432