xref: /freebsd/sys/netinet/ip_input.c (revision 65b71c48d226f3ef3eb1c4452dccc678a7d2ac96)
1 /*-
2  * Copyright (c) 1982, 1986, 1988, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 4. Neither the name of the University nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	@(#)ip_input.c	8.2 (Berkeley) 1/4/94
30  */
31 
32 #include <sys/cdefs.h>
33 __FBSDID("$FreeBSD$");
34 
35 #include "opt_bootp.h"
36 #include "opt_ipfw.h"
37 #include "opt_ipstealth.h"
38 #include "opt_ipsec.h"
39 #include "opt_route.h"
40 #include "opt_rss.h"
41 
42 #include <sys/param.h>
43 #include <sys/systm.h>
44 #include <sys/mbuf.h>
45 #include <sys/malloc.h>
46 #include <sys/domain.h>
47 #include <sys/protosw.h>
48 #include <sys/socket.h>
49 #include <sys/time.h>
50 #include <sys/kernel.h>
51 #include <sys/lock.h>
52 #include <sys/rmlock.h>
53 #include <sys/rwlock.h>
54 #include <sys/sdt.h>
55 #include <sys/syslog.h>
56 #include <sys/sysctl.h>
57 
58 #include <net/pfil.h>
59 #include <net/if.h>
60 #include <net/if_types.h>
61 #include <net/if_var.h>
62 #include <net/if_dl.h>
63 #include <net/route.h>
64 #include <net/netisr.h>
65 #include <net/rss_config.h>
66 #include <net/vnet.h>
67 
68 #include <netinet/in.h>
69 #include <netinet/in_kdtrace.h>
70 #include <netinet/in_systm.h>
71 #include <netinet/in_var.h>
72 #include <netinet/ip.h>
73 #include <netinet/in_pcb.h>
74 #include <netinet/ip_var.h>
75 #include <netinet/ip_fw.h>
76 #include <netinet/ip_icmp.h>
77 #include <netinet/ip_options.h>
78 #include <machine/in_cksum.h>
79 #include <netinet/ip_carp.h>
80 #ifdef IPSEC
81 #include <netinet/ip_ipsec.h>
82 #endif /* IPSEC */
83 #include <netinet/in_rss.h>
84 
85 #include <sys/socketvar.h>
86 
87 #include <security/mac/mac_framework.h>
88 
89 #ifdef CTASSERT
90 CTASSERT(sizeof(struct ip) == 20);
91 #endif
92 
93 /* IP reassembly functions are defined in ip_reass.c. */
94 extern void ipreass_init(void);
95 extern void ipreass_drain(void);
96 extern void ipreass_slowtimo(void);
97 #ifdef VIMAGE
98 extern void ipreass_destroy(void);
99 #endif
100 
101 struct rmlock in_ifaddr_lock;
102 RM_SYSINIT(in_ifaddr_lock, &in_ifaddr_lock, "in_ifaddr_lock");
103 
104 VNET_DEFINE(int, rsvp_on);
105 
106 VNET_DEFINE(int, ipforwarding);
107 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_VNET | CTLFLAG_RW,
108     &VNET_NAME(ipforwarding), 0,
109     "Enable IP forwarding between interfaces");
110 
111 static VNET_DEFINE(int, ipsendredirects) = 1;	/* XXX */
112 #define	V_ipsendredirects	VNET(ipsendredirects)
113 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_VNET | CTLFLAG_RW,
114     &VNET_NAME(ipsendredirects), 0,
115     "Enable sending IP redirects");
116 
117 /*
118  * XXX - Setting ip_checkinterface mostly implements the receive side of
119  * the Strong ES model described in RFC 1122, but since the routing table
120  * and transmit implementation do not implement the Strong ES model,
121  * setting this to 1 results in an odd hybrid.
122  *
123  * XXX - ip_checkinterface currently must be disabled if you use ipnat
124  * to translate the destination address to another local interface.
125  *
126  * XXX - ip_checkinterface must be disabled if you add IP aliases
127  * to the loopback interface instead of the interface where the
128  * packets for those addresses are received.
129  */
130 static VNET_DEFINE(int, ip_checkinterface);
131 #define	V_ip_checkinterface	VNET(ip_checkinterface)
132 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_VNET | CTLFLAG_RW,
133     &VNET_NAME(ip_checkinterface), 0,
134     "Verify packet arrives on correct interface");
135 
136 VNET_DEFINE(struct pfil_head, inet_pfil_hook);	/* Packet filter hooks */
137 
138 static struct netisr_handler ip_nh = {
139 	.nh_name = "ip",
140 	.nh_handler = ip_input,
141 	.nh_proto = NETISR_IP,
142 #ifdef	RSS
143 	.nh_m2cpuid = rss_soft_m2cpuid_v4,
144 	.nh_policy = NETISR_POLICY_CPU,
145 	.nh_dispatch = NETISR_DISPATCH_HYBRID,
146 #else
147 	.nh_policy = NETISR_POLICY_FLOW,
148 #endif
149 };
150 
151 #ifdef	RSS
152 /*
153  * Directly dispatched frames are currently assumed
154  * to have a flowid already calculated.
155  *
156  * It should likely have something that assert it
157  * actually has valid flow details.
158  */
159 static struct netisr_handler ip_direct_nh = {
160 	.nh_name = "ip_direct",
161 	.nh_handler = ip_direct_input,
162 	.nh_proto = NETISR_IP_DIRECT,
163 	.nh_m2cpuid = rss_soft_m2cpuid_v4,
164 	.nh_policy = NETISR_POLICY_CPU,
165 	.nh_dispatch = NETISR_DISPATCH_HYBRID,
166 };
167 #endif
168 
169 extern	struct domain inetdomain;
170 extern	struct protosw inetsw[];
171 u_char	ip_protox[IPPROTO_MAX];
172 VNET_DEFINE(struct in_ifaddrhead, in_ifaddrhead);  /* first inet address */
173 VNET_DEFINE(struct in_ifaddrhashhead *, in_ifaddrhashtbl); /* inet addr hash table  */
174 VNET_DEFINE(u_long, in_ifaddrhmask);		/* mask for hash table */
175 
176 #ifdef IPCTL_DEFMTU
177 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
178     &ip_mtu, 0, "Default MTU");
179 #endif
180 
181 #ifdef IPSTEALTH
182 VNET_DEFINE(int, ipstealth);
183 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_VNET | CTLFLAG_RW,
184     &VNET_NAME(ipstealth), 0,
185     "IP stealth mode, no TTL decrementation on forwarding");
186 #endif
187 
188 /*
189  * IP statistics are stored in the "array" of counter(9)s.
190  */
191 VNET_PCPUSTAT_DEFINE(struct ipstat, ipstat);
192 VNET_PCPUSTAT_SYSINIT(ipstat);
193 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, IPCTL_STATS, stats, struct ipstat, ipstat,
194     "IP statistics (struct ipstat, netinet/ip_var.h)");
195 
196 #ifdef VIMAGE
197 VNET_PCPUSTAT_SYSUNINIT(ipstat);
198 #endif /* VIMAGE */
199 
200 /*
201  * Kernel module interface for updating ipstat.  The argument is an index
202  * into ipstat treated as an array.
203  */
204 void
205 kmod_ipstat_inc(int statnum)
206 {
207 
208 	counter_u64_add(VNET(ipstat)[statnum], 1);
209 }
210 
211 void
212 kmod_ipstat_dec(int statnum)
213 {
214 
215 	counter_u64_add(VNET(ipstat)[statnum], -1);
216 }
217 
218 static int
219 sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS)
220 {
221 	int error, qlimit;
222 
223 	netisr_getqlimit(&ip_nh, &qlimit);
224 	error = sysctl_handle_int(oidp, &qlimit, 0, req);
225 	if (error || !req->newptr)
226 		return (error);
227 	if (qlimit < 1)
228 		return (EINVAL);
229 	return (netisr_setqlimit(&ip_nh, qlimit));
230 }
231 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen,
232     CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet_intr_queue_maxlen, "I",
233     "Maximum size of the IP input queue");
234 
235 static int
236 sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS)
237 {
238 	u_int64_t qdrops_long;
239 	int error, qdrops;
240 
241 	netisr_getqdrops(&ip_nh, &qdrops_long);
242 	qdrops = qdrops_long;
243 	error = sysctl_handle_int(oidp, &qdrops, 0, req);
244 	if (error || !req->newptr)
245 		return (error);
246 	if (qdrops != 0)
247 		return (EINVAL);
248 	netisr_clearqdrops(&ip_nh);
249 	return (0);
250 }
251 
252 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops,
253     CTLTYPE_INT|CTLFLAG_RD, 0, 0, sysctl_netinet_intr_queue_drops, "I",
254     "Number of packets dropped from the IP input queue");
255 
256 #ifdef	RSS
257 static int
258 sysctl_netinet_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS)
259 {
260 	int error, qlimit;
261 
262 	netisr_getqlimit(&ip_direct_nh, &qlimit);
263 	error = sysctl_handle_int(oidp, &qlimit, 0, req);
264 	if (error || !req->newptr)
265 		return (error);
266 	if (qlimit < 1)
267 		return (EINVAL);
268 	return (netisr_setqlimit(&ip_direct_nh, qlimit));
269 }
270 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_direct_queue_maxlen,
271     CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet_intr_direct_queue_maxlen, "I",
272     "Maximum size of the IP direct input queue");
273 
274 static int
275 sysctl_netinet_intr_direct_queue_drops(SYSCTL_HANDLER_ARGS)
276 {
277 	u_int64_t qdrops_long;
278 	int error, qdrops;
279 
280 	netisr_getqdrops(&ip_direct_nh, &qdrops_long);
281 	qdrops = qdrops_long;
282 	error = sysctl_handle_int(oidp, &qdrops, 0, req);
283 	if (error || !req->newptr)
284 		return (error);
285 	if (qdrops != 0)
286 		return (EINVAL);
287 	netisr_clearqdrops(&ip_direct_nh);
288 	return (0);
289 }
290 
291 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_direct_queue_drops,
292     CTLTYPE_INT|CTLFLAG_RD, 0, 0, sysctl_netinet_intr_direct_queue_drops, "I",
293     "Number of packets dropped from the IP direct input queue");
294 #endif	/* RSS */
295 
296 /*
297  * IP initialization: fill in IP protocol switch table.
298  * All protocols not implemented in kernel go to raw IP protocol handler.
299  */
300 void
301 ip_init(void)
302 {
303 	struct protosw *pr;
304 	int i;
305 
306 	TAILQ_INIT(&V_in_ifaddrhead);
307 	V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask);
308 
309 	/* Initialize IP reassembly queue. */
310 	ipreass_init();
311 
312 	/* Initialize packet filter hooks. */
313 	V_inet_pfil_hook.ph_type = PFIL_TYPE_AF;
314 	V_inet_pfil_hook.ph_af = AF_INET;
315 	if ((i = pfil_head_register(&V_inet_pfil_hook)) != 0)
316 		printf("%s: WARNING: unable to register pfil hook, "
317 			"error %d\n", __func__, i);
318 
319 	/* Skip initialization of globals for non-default instances. */
320 	if (!IS_DEFAULT_VNET(curvnet))
321 		return;
322 
323 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
324 	if (pr == NULL)
325 		panic("ip_init: PF_INET not found");
326 
327 	/* Initialize the entire ip_protox[] array to IPPROTO_RAW. */
328 	for (i = 0; i < IPPROTO_MAX; i++)
329 		ip_protox[i] = pr - inetsw;
330 	/*
331 	 * Cycle through IP protocols and put them into the appropriate place
332 	 * in ip_protox[].
333 	 */
334 	for (pr = inetdomain.dom_protosw;
335 	    pr < inetdomain.dom_protoswNPROTOSW; pr++)
336 		if (pr->pr_domain->dom_family == PF_INET &&
337 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
338 			/* Be careful to only index valid IP protocols. */
339 			if (pr->pr_protocol < IPPROTO_MAX)
340 				ip_protox[pr->pr_protocol] = pr - inetsw;
341 		}
342 
343 	netisr_register(&ip_nh);
344 #ifdef	RSS
345 	netisr_register(&ip_direct_nh);
346 #endif
347 }
348 
349 #ifdef VIMAGE
350 void
351 ip_destroy(void)
352 {
353 	int i;
354 
355 	if ((i = pfil_head_unregister(&V_inet_pfil_hook)) != 0)
356 		printf("%s: WARNING: unable to unregister pfil hook, "
357 		    "error %d\n", __func__, i);
358 
359 	/* Cleanup in_ifaddr hash table; should be empty. */
360 	hashdestroy(V_in_ifaddrhashtbl, M_IFADDR, V_in_ifaddrhmask);
361 
362 	/* Destroy IP reassembly queue. */
363 	ipreass_destroy();
364 }
365 #endif
366 
367 #ifdef	RSS
368 /*
369  * IP direct input routine.
370  *
371  * This is called when reinjecting completed fragments where
372  * all of the previous checking and book-keeping has been done.
373  */
374 void
375 ip_direct_input(struct mbuf *m)
376 {
377 	struct ip *ip;
378 	int hlen;
379 
380 	ip = mtod(m, struct ip *);
381 	hlen = ip->ip_hl << 2;
382 
383 	IPSTAT_INC(ips_delivered);
384 	(*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
385 	return;
386 }
387 #endif
388 
389 /*
390  * Ip input routine.  Checksum and byte swap header.  If fragmented
391  * try to reassemble.  Process options.  Pass to next level.
392  */
393 void
394 ip_input(struct mbuf *m)
395 {
396 	struct ip *ip = NULL;
397 	struct in_ifaddr *ia = NULL;
398 	struct ifaddr *ifa;
399 	struct ifnet *ifp;
400 	int    checkif, hlen = 0;
401 	uint16_t sum, ip_len;
402 	int dchg = 0;				/* dest changed after fw */
403 	struct in_addr odst;			/* original dst address */
404 
405 	M_ASSERTPKTHDR(m);
406 
407 	if (m->m_flags & M_FASTFWD_OURS) {
408 		m->m_flags &= ~M_FASTFWD_OURS;
409 		/* Set up some basics that will be used later. */
410 		ip = mtod(m, struct ip *);
411 		hlen = ip->ip_hl << 2;
412 		ip_len = ntohs(ip->ip_len);
413 		goto ours;
414 	}
415 
416 	IPSTAT_INC(ips_total);
417 
418 	if (m->m_pkthdr.len < sizeof(struct ip))
419 		goto tooshort;
420 
421 	if (m->m_len < sizeof (struct ip) &&
422 	    (m = m_pullup(m, sizeof (struct ip))) == NULL) {
423 		IPSTAT_INC(ips_toosmall);
424 		return;
425 	}
426 	ip = mtod(m, struct ip *);
427 
428 	if (ip->ip_v != IPVERSION) {
429 		IPSTAT_INC(ips_badvers);
430 		goto bad;
431 	}
432 
433 	hlen = ip->ip_hl << 2;
434 	if (hlen < sizeof(struct ip)) {	/* minimum header length */
435 		IPSTAT_INC(ips_badhlen);
436 		goto bad;
437 	}
438 	if (hlen > m->m_len) {
439 		if ((m = m_pullup(m, hlen)) == NULL) {
440 			IPSTAT_INC(ips_badhlen);
441 			return;
442 		}
443 		ip = mtod(m, struct ip *);
444 	}
445 
446 	IP_PROBE(receive, NULL, NULL, ip, m->m_pkthdr.rcvif, ip, NULL);
447 
448 	/* 127/8 must not appear on wire - RFC1122 */
449 	ifp = m->m_pkthdr.rcvif;
450 	if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
451 	    (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
452 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
453 			IPSTAT_INC(ips_badaddr);
454 			goto bad;
455 		}
456 	}
457 
458 	if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
459 		sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
460 	} else {
461 		if (hlen == sizeof(struct ip)) {
462 			sum = in_cksum_hdr(ip);
463 		} else {
464 			sum = in_cksum(m, hlen);
465 		}
466 	}
467 	if (sum) {
468 		IPSTAT_INC(ips_badsum);
469 		goto bad;
470 	}
471 
472 #ifdef ALTQ
473 	if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
474 		/* packet is dropped by traffic conditioner */
475 		return;
476 #endif
477 
478 	ip_len = ntohs(ip->ip_len);
479 	if (ip_len < hlen) {
480 		IPSTAT_INC(ips_badlen);
481 		goto bad;
482 	}
483 
484 	/*
485 	 * Check that the amount of data in the buffers
486 	 * is as at least much as the IP header would have us expect.
487 	 * Trim mbufs if longer than we expect.
488 	 * Drop packet if shorter than we expect.
489 	 */
490 	if (m->m_pkthdr.len < ip_len) {
491 tooshort:
492 		IPSTAT_INC(ips_tooshort);
493 		goto bad;
494 	}
495 	if (m->m_pkthdr.len > ip_len) {
496 		if (m->m_len == m->m_pkthdr.len) {
497 			m->m_len = ip_len;
498 			m->m_pkthdr.len = ip_len;
499 		} else
500 			m_adj(m, ip_len - m->m_pkthdr.len);
501 	}
502 
503 #ifdef IPSEC
504 	/*
505 	 * Bypass packet filtering for packets previously handled by IPsec.
506 	 */
507 	if (ip_ipsec_filtertunnel(m))
508 		goto passin;
509 #endif /* IPSEC */
510 
511 	/*
512 	 * Run through list of hooks for input packets.
513 	 *
514 	 * NB: Beware of the destination address changing (e.g.
515 	 *     by NAT rewriting).  When this happens, tell
516 	 *     ip_forward to do the right thing.
517 	 */
518 
519 	/* Jump over all PFIL processing if hooks are not active. */
520 	if (!PFIL_HOOKED(&V_inet_pfil_hook))
521 		goto passin;
522 
523 	odst = ip->ip_dst;
524 	if (pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_IN, NULL) != 0)
525 		return;
526 	if (m == NULL)			/* consumed by filter */
527 		return;
528 
529 	ip = mtod(m, struct ip *);
530 	dchg = (odst.s_addr != ip->ip_dst.s_addr);
531 	ifp = m->m_pkthdr.rcvif;
532 
533 	if (m->m_flags & M_FASTFWD_OURS) {
534 		m->m_flags &= ~M_FASTFWD_OURS;
535 		goto ours;
536 	}
537 	if (m->m_flags & M_IP_NEXTHOP) {
538 		dchg = (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL);
539 		if (dchg != 0) {
540 			/*
541 			 * Directly ship the packet on.  This allows
542 			 * forwarding packets originally destined to us
543 			 * to some other directly connected host.
544 			 */
545 			ip_forward(m, 1);
546 			return;
547 		}
548 	}
549 passin:
550 
551 	/*
552 	 * Process options and, if not destined for us,
553 	 * ship it on.  ip_dooptions returns 1 when an
554 	 * error was detected (causing an icmp message
555 	 * to be sent and the original packet to be freed).
556 	 */
557 	if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
558 		return;
559 
560         /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
561          * matter if it is destined to another node, or whether it is
562          * a multicast one, RSVP wants it! and prevents it from being forwarded
563          * anywhere else. Also checks if the rsvp daemon is running before
564 	 * grabbing the packet.
565          */
566 	if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP)
567 		goto ours;
568 
569 	/*
570 	 * Check our list of addresses, to see if the packet is for us.
571 	 * If we don't have any addresses, assume any unicast packet
572 	 * we receive might be for us (and let the upper layers deal
573 	 * with it).
574 	 */
575 	if (TAILQ_EMPTY(&V_in_ifaddrhead) &&
576 	    (m->m_flags & (M_MCAST|M_BCAST)) == 0)
577 		goto ours;
578 
579 	/*
580 	 * Enable a consistency check between the destination address
581 	 * and the arrival interface for a unicast packet (the RFC 1122
582 	 * strong ES model) if IP forwarding is disabled and the packet
583 	 * is not locally generated and the packet is not subject to
584 	 * 'ipfw fwd'.
585 	 *
586 	 * XXX - Checking also should be disabled if the destination
587 	 * address is ipnat'ed to a different interface.
588 	 *
589 	 * XXX - Checking is incompatible with IP aliases added
590 	 * to the loopback interface instead of the interface where
591 	 * the packets are received.
592 	 *
593 	 * XXX - This is the case for carp vhost IPs as well so we
594 	 * insert a workaround. If the packet got here, we already
595 	 * checked with carp_iamatch() and carp_forus().
596 	 */
597 	checkif = V_ip_checkinterface && (V_ipforwarding == 0) &&
598 	    ifp != NULL && ((ifp->if_flags & IFF_LOOPBACK) == 0) &&
599 	    ifp->if_carp == NULL && (dchg == 0);
600 
601 	/*
602 	 * Check for exact addresses in the hash bucket.
603 	 */
604 	/* IN_IFADDR_RLOCK(); */
605 	LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
606 		/*
607 		 * If the address matches, verify that the packet
608 		 * arrived via the correct interface if checking is
609 		 * enabled.
610 		 */
611 		if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr &&
612 		    (!checkif || ia->ia_ifp == ifp)) {
613 			counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
614 			counter_u64_add(ia->ia_ifa.ifa_ibytes,
615 			    m->m_pkthdr.len);
616 			/* IN_IFADDR_RUNLOCK(); */
617 			goto ours;
618 		}
619 	}
620 	/* IN_IFADDR_RUNLOCK(); */
621 
622 	/*
623 	 * Check for broadcast addresses.
624 	 *
625 	 * Only accept broadcast packets that arrive via the matching
626 	 * interface.  Reception of forwarded directed broadcasts would
627 	 * be handled via ip_forward() and ether_output() with the loopback
628 	 * into the stack for SIMPLEX interfaces handled by ether_output().
629 	 */
630 	if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
631 		IF_ADDR_RLOCK(ifp);
632 	        TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
633 			if (ifa->ifa_addr->sa_family != AF_INET)
634 				continue;
635 			ia = ifatoia(ifa);
636 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
637 			    ip->ip_dst.s_addr) {
638 				counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
639 				counter_u64_add(ia->ia_ifa.ifa_ibytes,
640 				    m->m_pkthdr.len);
641 				IF_ADDR_RUNLOCK(ifp);
642 				goto ours;
643 			}
644 #ifdef BOOTP_COMPAT
645 			if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) {
646 				counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
647 				counter_u64_add(ia->ia_ifa.ifa_ibytes,
648 				    m->m_pkthdr.len);
649 				IF_ADDR_RUNLOCK(ifp);
650 				goto ours;
651 			}
652 #endif
653 		}
654 		IF_ADDR_RUNLOCK(ifp);
655 		ia = NULL;
656 	}
657 	/* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */
658 	if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) {
659 		IPSTAT_INC(ips_cantforward);
660 		m_freem(m);
661 		return;
662 	}
663 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
664 		if (V_ip_mrouter) {
665 			/*
666 			 * If we are acting as a multicast router, all
667 			 * incoming multicast packets are passed to the
668 			 * kernel-level multicast forwarding function.
669 			 * The packet is returned (relatively) intact; if
670 			 * ip_mforward() returns a non-zero value, the packet
671 			 * must be discarded, else it may be accepted below.
672 			 */
673 			if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) {
674 				IPSTAT_INC(ips_cantforward);
675 				m_freem(m);
676 				return;
677 			}
678 
679 			/*
680 			 * The process-level routing daemon needs to receive
681 			 * all multicast IGMP packets, whether or not this
682 			 * host belongs to their destination groups.
683 			 */
684 			if (ip->ip_p == IPPROTO_IGMP)
685 				goto ours;
686 			IPSTAT_INC(ips_forward);
687 		}
688 		/*
689 		 * Assume the packet is for us, to avoid prematurely taking
690 		 * a lock on the in_multi hash. Protocols must perform
691 		 * their own filtering and update statistics accordingly.
692 		 */
693 		goto ours;
694 	}
695 	if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
696 		goto ours;
697 	if (ip->ip_dst.s_addr == INADDR_ANY)
698 		goto ours;
699 
700 	/*
701 	 * Not for us; forward if possible and desirable.
702 	 */
703 	if (V_ipforwarding == 0) {
704 		IPSTAT_INC(ips_cantforward);
705 		m_freem(m);
706 	} else {
707 		ip_forward(m, dchg);
708 	}
709 	return;
710 
711 ours:
712 #ifdef IPSTEALTH
713 	/*
714 	 * IPSTEALTH: Process non-routing options only
715 	 * if the packet is destined for us.
716 	 */
717 	if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1))
718 		return;
719 #endif /* IPSTEALTH */
720 
721 	/*
722 	 * Attempt reassembly; if it succeeds, proceed.
723 	 * ip_reass() will return a different mbuf.
724 	 */
725 	if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) {
726 		/* XXXGL: shouldn't we save & set m_flags? */
727 		m = ip_reass(m);
728 		if (m == NULL)
729 			return;
730 		ip = mtod(m, struct ip *);
731 		/* Get the header length of the reassembled packet */
732 		hlen = ip->ip_hl << 2;
733 	}
734 
735 #ifdef IPSEC
736 	/*
737 	 * enforce IPsec policy checking if we are seeing last header.
738 	 * note that we do not visit this with protocols with pcb layer
739 	 * code - like udp/tcp/raw ip.
740 	 */
741 	if (ip_ipsec_input(m, ip->ip_p) != 0)
742 		goto bad;
743 #endif /* IPSEC */
744 
745 	/*
746 	 * Switch out to protocol's input routine.
747 	 */
748 	IPSTAT_INC(ips_delivered);
749 
750 	(*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
751 	return;
752 bad:
753 	m_freem(m);
754 }
755 
756 /*
757  * IP timer processing;
758  * if a timer expires on a reassembly
759  * queue, discard it.
760  */
761 void
762 ip_slowtimo(void)
763 {
764 	VNET_ITERATOR_DECL(vnet_iter);
765 
766 	VNET_LIST_RLOCK_NOSLEEP();
767 	VNET_FOREACH(vnet_iter) {
768 		CURVNET_SET(vnet_iter);
769 		ipreass_slowtimo();
770 		CURVNET_RESTORE();
771 	}
772 	VNET_LIST_RUNLOCK_NOSLEEP();
773 }
774 
775 void
776 ip_drain(void)
777 {
778 	VNET_ITERATOR_DECL(vnet_iter);
779 
780 	VNET_LIST_RLOCK_NOSLEEP();
781 	VNET_FOREACH(vnet_iter) {
782 		CURVNET_SET(vnet_iter);
783 		ipreass_drain();
784 		CURVNET_RESTORE();
785 	}
786 	VNET_LIST_RUNLOCK_NOSLEEP();
787 }
788 
789 /*
790  * The protocol to be inserted into ip_protox[] must be already registered
791  * in inetsw[], either statically or through pf_proto_register().
792  */
793 int
794 ipproto_register(short ipproto)
795 {
796 	struct protosw *pr;
797 
798 	/* Sanity checks. */
799 	if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
800 		return (EPROTONOSUPPORT);
801 
802 	/*
803 	 * The protocol slot must not be occupied by another protocol
804 	 * already.  An index pointing to IPPROTO_RAW is unused.
805 	 */
806 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
807 	if (pr == NULL)
808 		return (EPFNOSUPPORT);
809 	if (ip_protox[ipproto] != pr - inetsw)	/* IPPROTO_RAW */
810 		return (EEXIST);
811 
812 	/* Find the protocol position in inetsw[] and set the index. */
813 	for (pr = inetdomain.dom_protosw;
814 	     pr < inetdomain.dom_protoswNPROTOSW; pr++) {
815 		if (pr->pr_domain->dom_family == PF_INET &&
816 		    pr->pr_protocol && pr->pr_protocol == ipproto) {
817 			ip_protox[pr->pr_protocol] = pr - inetsw;
818 			return (0);
819 		}
820 	}
821 	return (EPROTONOSUPPORT);
822 }
823 
824 int
825 ipproto_unregister(short ipproto)
826 {
827 	struct protosw *pr;
828 
829 	/* Sanity checks. */
830 	if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
831 		return (EPROTONOSUPPORT);
832 
833 	/* Check if the protocol was indeed registered. */
834 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
835 	if (pr == NULL)
836 		return (EPFNOSUPPORT);
837 	if (ip_protox[ipproto] == pr - inetsw)  /* IPPROTO_RAW */
838 		return (ENOENT);
839 
840 	/* Reset the protocol slot to IPPROTO_RAW. */
841 	ip_protox[ipproto] = pr - inetsw;
842 	return (0);
843 }
844 
845 /*
846  * Given address of next destination (final or next hop), return (referenced)
847  * internet address info of interface to be used to get there.
848  */
849 struct in_ifaddr *
850 ip_rtaddr(struct in_addr dst, u_int fibnum)
851 {
852 	struct route sro;
853 	struct sockaddr_in *sin;
854 	struct in_ifaddr *ia;
855 
856 	bzero(&sro, sizeof(sro));
857 	sin = (struct sockaddr_in *)&sro.ro_dst;
858 	sin->sin_family = AF_INET;
859 	sin->sin_len = sizeof(*sin);
860 	sin->sin_addr = dst;
861 	in_rtalloc_ign(&sro, 0, fibnum);
862 
863 	if (sro.ro_rt == NULL)
864 		return (NULL);
865 
866 	ia = ifatoia(sro.ro_rt->rt_ifa);
867 	ifa_ref(&ia->ia_ifa);
868 	RTFREE(sro.ro_rt);
869 	return (ia);
870 }
871 
872 u_char inetctlerrmap[PRC_NCMDS] = {
873 	0,		0,		0,		0,
874 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
875 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
876 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
877 	0,		0,		EHOSTUNREACH,	0,
878 	ENOPROTOOPT,	ECONNREFUSED
879 };
880 
881 /*
882  * Forward a packet.  If some error occurs return the sender
883  * an icmp packet.  Note we can't always generate a meaningful
884  * icmp message because icmp doesn't have a large enough repertoire
885  * of codes and types.
886  *
887  * If not forwarding, just drop the packet.  This could be confusing
888  * if ipforwarding was zero but some routing protocol was advancing
889  * us as a gateway to somewhere.  However, we must let the routing
890  * protocol deal with that.
891  *
892  * The srcrt parameter indicates whether the packet is being forwarded
893  * via a source route.
894  */
895 void
896 ip_forward(struct mbuf *m, int srcrt)
897 {
898 	struct ip *ip = mtod(m, struct ip *);
899 	struct in_ifaddr *ia;
900 	struct mbuf *mcopy;
901 	struct sockaddr_in *sin;
902 	struct in_addr dest;
903 	struct route ro;
904 	int error, type = 0, code = 0, mtu = 0;
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 #ifdef IPSEC
912 	if (ip_ipsec_fwd(m) != 0) {
913 		IPSTAT_INC(ips_cantforward);
914 		m_freem(m);
915 		return;
916 	}
917 #endif /* IPSEC */
918 #ifdef IPSTEALTH
919 	if (!V_ipstealth) {
920 #endif
921 		if (ip->ip_ttl <= IPTTLDEC) {
922 			icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS,
923 			    0, 0);
924 			return;
925 		}
926 #ifdef IPSTEALTH
927 	}
928 #endif
929 
930 	bzero(&ro, sizeof(ro));
931 	sin = (struct sockaddr_in *)&ro.ro_dst;
932 	sin->sin_family = AF_INET;
933 	sin->sin_len = sizeof(*sin);
934 	sin->sin_addr = ip->ip_dst;
935 #ifdef RADIX_MPATH
936 	rtalloc_mpath_fib(&ro,
937 	    ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
938 	    M_GETFIB(m));
939 #else
940 	in_rtalloc_ign(&ro, 0, M_GETFIB(m));
941 #endif
942 	if (ro.ro_rt != NULL) {
943 		ia = ifatoia(ro.ro_rt->rt_ifa);
944 		ifa_ref(&ia->ia_ifa);
945 	} else
946 		ia = NULL;
947 #ifndef IPSEC
948 	/*
949 	 * 'ia' may be NULL if there is no route for this destination.
950 	 * In case of IPsec, Don't discard it just yet, but pass it to
951 	 * ip_output in case of outgoing IPsec policy.
952 	 */
953 	if (!srcrt && ia == NULL) {
954 		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
955 		RO_RTFREE(&ro);
956 		return;
957 	}
958 #endif
959 
960 	/*
961 	 * Save the IP header and at most 8 bytes of the payload,
962 	 * in case we need to generate an ICMP message to the src.
963 	 *
964 	 * XXX this can be optimized a lot by saving the data in a local
965 	 * buffer on the stack (72 bytes at most), and only allocating the
966 	 * mbuf if really necessary. The vast majority of the packets
967 	 * are forwarded without having to send an ICMP back (either
968 	 * because unnecessary, or because rate limited), so we are
969 	 * really we are wasting a lot of work here.
970 	 *
971 	 * We don't use m_copy() because it might return a reference
972 	 * to a shared cluster. Both this function and ip_output()
973 	 * assume exclusive access to the IP header in `m', so any
974 	 * data in a cluster may change before we reach icmp_error().
975 	 */
976 	mcopy = m_gethdr(M_NOWAIT, m->m_type);
977 	if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_NOWAIT)) {
978 		/*
979 		 * It's probably ok if the pkthdr dup fails (because
980 		 * the deep copy of the tag chain failed), but for now
981 		 * be conservative and just discard the copy since
982 		 * code below may some day want the tags.
983 		 */
984 		m_free(mcopy);
985 		mcopy = NULL;
986 	}
987 	if (mcopy != NULL) {
988 		mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy));
989 		mcopy->m_pkthdr.len = mcopy->m_len;
990 		m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
991 	}
992 
993 #ifdef IPSTEALTH
994 	if (!V_ipstealth) {
995 #endif
996 		ip->ip_ttl -= IPTTLDEC;
997 #ifdef IPSTEALTH
998 	}
999 #endif
1000 
1001 	/*
1002 	 * If forwarding packet using same interface that it came in on,
1003 	 * perhaps should send a redirect to sender to shortcut a hop.
1004 	 * Only send redirect if source is sending directly to us,
1005 	 * and if packet was not source routed (or has any options).
1006 	 * Also, don't send redirect if forwarding using a default route
1007 	 * or a route modified by a redirect.
1008 	 */
1009 	dest.s_addr = 0;
1010 	if (!srcrt && V_ipsendredirects &&
1011 	    ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) {
1012 		struct rtentry *rt;
1013 
1014 		rt = ro.ro_rt;
1015 
1016 		if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1017 		    satosin(rt_key(rt))->sin_addr.s_addr != 0) {
1018 #define	RTA(rt)	((struct in_ifaddr *)(rt->rt_ifa))
1019 			u_long src = ntohl(ip->ip_src.s_addr);
1020 
1021 			if (RTA(rt) &&
1022 			    (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1023 				if (rt->rt_flags & RTF_GATEWAY)
1024 					dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr;
1025 				else
1026 					dest.s_addr = ip->ip_dst.s_addr;
1027 				/* Router requirements says to only send host redirects */
1028 				type = ICMP_REDIRECT;
1029 				code = ICMP_REDIRECT_HOST;
1030 			}
1031 		}
1032 	}
1033 
1034 	error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
1035 
1036 	if (error == EMSGSIZE && ro.ro_rt)
1037 		mtu = ro.ro_rt->rt_mtu;
1038 	RO_RTFREE(&ro);
1039 
1040 	if (error)
1041 		IPSTAT_INC(ips_cantforward);
1042 	else {
1043 		IPSTAT_INC(ips_forward);
1044 		if (type)
1045 			IPSTAT_INC(ips_redirectsent);
1046 		else {
1047 			if (mcopy)
1048 				m_freem(mcopy);
1049 			if (ia != NULL)
1050 				ifa_free(&ia->ia_ifa);
1051 			return;
1052 		}
1053 	}
1054 	if (mcopy == NULL) {
1055 		if (ia != NULL)
1056 			ifa_free(&ia->ia_ifa);
1057 		return;
1058 	}
1059 
1060 	switch (error) {
1061 
1062 	case 0:				/* forwarded, but need redirect */
1063 		/* type, code set above */
1064 		break;
1065 
1066 	case ENETUNREACH:
1067 	case EHOSTUNREACH:
1068 	case ENETDOWN:
1069 	case EHOSTDOWN:
1070 	default:
1071 		type = ICMP_UNREACH;
1072 		code = ICMP_UNREACH_HOST;
1073 		break;
1074 
1075 	case EMSGSIZE:
1076 		type = ICMP_UNREACH;
1077 		code = ICMP_UNREACH_NEEDFRAG;
1078 
1079 #ifdef IPSEC
1080 		/*
1081 		 * If IPsec is configured for this path,
1082 		 * override any possibly mtu value set by ip_output.
1083 		 */
1084 		mtu = ip_ipsec_mtu(mcopy, mtu);
1085 #endif /* IPSEC */
1086 		/*
1087 		 * If the MTU was set before make sure we are below the
1088 		 * interface MTU.
1089 		 * If the MTU wasn't set before use the interface mtu or
1090 		 * fall back to the next smaller mtu step compared to the
1091 		 * current packet size.
1092 		 */
1093 		if (mtu != 0) {
1094 			if (ia != NULL)
1095 				mtu = min(mtu, ia->ia_ifp->if_mtu);
1096 		} else {
1097 			if (ia != NULL)
1098 				mtu = ia->ia_ifp->if_mtu;
1099 			else
1100 				mtu = ip_next_mtu(ntohs(ip->ip_len), 0);
1101 		}
1102 		IPSTAT_INC(ips_cantfrag);
1103 		break;
1104 
1105 	case ENOBUFS:
1106 	case EACCES:			/* ipfw denied packet */
1107 		m_freem(mcopy);
1108 		if (ia != NULL)
1109 			ifa_free(&ia->ia_ifa);
1110 		return;
1111 	}
1112 	if (ia != NULL)
1113 		ifa_free(&ia->ia_ifa);
1114 	icmp_error(mcopy, type, code, dest.s_addr, mtu);
1115 }
1116 
1117 void
1118 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1119     struct mbuf *m)
1120 {
1121 
1122 	if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) {
1123 		struct bintime bt;
1124 
1125 		bintime(&bt);
1126 		if (inp->inp_socket->so_options & SO_BINTIME) {
1127 			*mp = sbcreatecontrol((caddr_t)&bt, sizeof(bt),
1128 			    SCM_BINTIME, SOL_SOCKET);
1129 			if (*mp)
1130 				mp = &(*mp)->m_next;
1131 		}
1132 		if (inp->inp_socket->so_options & SO_TIMESTAMP) {
1133 			struct timeval tv;
1134 
1135 			bintime2timeval(&bt, &tv);
1136 			*mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv),
1137 			    SCM_TIMESTAMP, SOL_SOCKET);
1138 			if (*mp)
1139 				mp = &(*mp)->m_next;
1140 		}
1141 	}
1142 	if (inp->inp_flags & INP_RECVDSTADDR) {
1143 		*mp = sbcreatecontrol((caddr_t)&ip->ip_dst,
1144 		    sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1145 		if (*mp)
1146 			mp = &(*mp)->m_next;
1147 	}
1148 	if (inp->inp_flags & INP_RECVTTL) {
1149 		*mp = sbcreatecontrol((caddr_t)&ip->ip_ttl,
1150 		    sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
1151 		if (*mp)
1152 			mp = &(*mp)->m_next;
1153 	}
1154 #ifdef notyet
1155 	/* XXX
1156 	 * Moving these out of udp_input() made them even more broken
1157 	 * than they already were.
1158 	 */
1159 	/* options were tossed already */
1160 	if (inp->inp_flags & INP_RECVOPTS) {
1161 		*mp = sbcreatecontrol((caddr_t)opts_deleted_above,
1162 		    sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1163 		if (*mp)
1164 			mp = &(*mp)->m_next;
1165 	}
1166 	/* ip_srcroute doesn't do what we want here, need to fix */
1167 	if (inp->inp_flags & INP_RECVRETOPTS) {
1168 		*mp = sbcreatecontrol((caddr_t)ip_srcroute(m),
1169 		    sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1170 		if (*mp)
1171 			mp = &(*mp)->m_next;
1172 	}
1173 #endif
1174 	if (inp->inp_flags & INP_RECVIF) {
1175 		struct ifnet *ifp;
1176 		struct sdlbuf {
1177 			struct sockaddr_dl sdl;
1178 			u_char	pad[32];
1179 		} sdlbuf;
1180 		struct sockaddr_dl *sdp;
1181 		struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1182 
1183 		if ((ifp = m->m_pkthdr.rcvif) &&
1184 		    ifp->if_index && ifp->if_index <= V_if_index) {
1185 			sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1186 			/*
1187 			 * Change our mind and don't try copy.
1188 			 */
1189 			if (sdp->sdl_family != AF_LINK ||
1190 			    sdp->sdl_len > sizeof(sdlbuf)) {
1191 				goto makedummy;
1192 			}
1193 			bcopy(sdp, sdl2, sdp->sdl_len);
1194 		} else {
1195 makedummy:
1196 			sdl2->sdl_len =
1197 			    offsetof(struct sockaddr_dl, sdl_data[0]);
1198 			sdl2->sdl_family = AF_LINK;
1199 			sdl2->sdl_index = 0;
1200 			sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1201 		}
1202 		*mp = sbcreatecontrol((caddr_t)sdl2, sdl2->sdl_len,
1203 		    IP_RECVIF, IPPROTO_IP);
1204 		if (*mp)
1205 			mp = &(*mp)->m_next;
1206 	}
1207 	if (inp->inp_flags & INP_RECVTOS) {
1208 		*mp = sbcreatecontrol((caddr_t)&ip->ip_tos,
1209 		    sizeof(u_char), IP_RECVTOS, IPPROTO_IP);
1210 		if (*mp)
1211 			mp = &(*mp)->m_next;
1212 	}
1213 
1214 	if (inp->inp_flags2 & INP_RECVFLOWID) {
1215 		uint32_t flowid, flow_type;
1216 
1217 		flowid = m->m_pkthdr.flowid;
1218 		flow_type = M_HASHTYPE_GET(m);
1219 
1220 		/*
1221 		 * XXX should handle the failure of one or the
1222 		 * other - don't populate both?
1223 		 */
1224 		*mp = sbcreatecontrol((caddr_t) &flowid,
1225 		    sizeof(uint32_t), IP_FLOWID, IPPROTO_IP);
1226 		if (*mp)
1227 			mp = &(*mp)->m_next;
1228 		*mp = sbcreatecontrol((caddr_t) &flow_type,
1229 		    sizeof(uint32_t), IP_FLOWTYPE, IPPROTO_IP);
1230 		if (*mp)
1231 			mp = &(*mp)->m_next;
1232 	}
1233 
1234 #ifdef	RSS
1235 	if (inp->inp_flags2 & INP_RECVRSSBUCKETID) {
1236 		uint32_t flowid, flow_type;
1237 		uint32_t rss_bucketid;
1238 
1239 		flowid = m->m_pkthdr.flowid;
1240 		flow_type = M_HASHTYPE_GET(m);
1241 
1242 		if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) {
1243 			*mp = sbcreatecontrol((caddr_t) &rss_bucketid,
1244 			   sizeof(uint32_t), IP_RSSBUCKETID, IPPROTO_IP);
1245 			if (*mp)
1246 				mp = &(*mp)->m_next;
1247 		}
1248 	}
1249 #endif
1250 }
1251 
1252 /*
1253  * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
1254  * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
1255  * locking.  This code remains in ip_input.c as ip_mroute.c is optionally
1256  * compiled.
1257  */
1258 static VNET_DEFINE(int, ip_rsvp_on);
1259 VNET_DEFINE(struct socket *, ip_rsvpd);
1260 
1261 #define	V_ip_rsvp_on		VNET(ip_rsvp_on)
1262 
1263 int
1264 ip_rsvp_init(struct socket *so)
1265 {
1266 
1267 	if (so->so_type != SOCK_RAW ||
1268 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
1269 		return EOPNOTSUPP;
1270 
1271 	if (V_ip_rsvpd != NULL)
1272 		return EADDRINUSE;
1273 
1274 	V_ip_rsvpd = so;
1275 	/*
1276 	 * This may seem silly, but we need to be sure we don't over-increment
1277 	 * the RSVP counter, in case something slips up.
1278 	 */
1279 	if (!V_ip_rsvp_on) {
1280 		V_ip_rsvp_on = 1;
1281 		V_rsvp_on++;
1282 	}
1283 
1284 	return 0;
1285 }
1286 
1287 int
1288 ip_rsvp_done(void)
1289 {
1290 
1291 	V_ip_rsvpd = NULL;
1292 	/*
1293 	 * This may seem silly, but we need to be sure we don't over-decrement
1294 	 * the RSVP counter, in case something slips up.
1295 	 */
1296 	if (V_ip_rsvp_on) {
1297 		V_ip_rsvp_on = 0;
1298 		V_rsvp_on--;
1299 	}
1300 	return 0;
1301 }
1302 
1303 int
1304 rsvp_input(struct mbuf **mp, int *offp, int proto)
1305 {
1306 	struct mbuf *m;
1307 
1308 	m = *mp;
1309 	*mp = NULL;
1310 
1311 	if (rsvp_input_p) { /* call the real one if loaded */
1312 		*mp = m;
1313 		rsvp_input_p(mp, offp, proto);
1314 		return (IPPROTO_DONE);
1315 	}
1316 
1317 	/* Can still get packets with rsvp_on = 0 if there is a local member
1318 	 * of the group to which the RSVP packet is addressed.  But in this
1319 	 * case we want to throw the packet away.
1320 	 */
1321 
1322 	if (!V_rsvp_on) {
1323 		m_freem(m);
1324 		return (IPPROTO_DONE);
1325 	}
1326 
1327 	if (V_ip_rsvpd != NULL) {
1328 		*mp = m;
1329 		rip_input(mp, offp, proto);
1330 		return (IPPROTO_DONE);
1331 	}
1332 	/* Drop the packet */
1333 	m_freem(m);
1334 	return (IPPROTO_DONE);
1335 }
1336