xref: /freebsd/sys/netinet/ip_input.c (revision 2830819497fb2deae3dd71574592ace55f2fbdba)
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 #include <netipsec/ipsec.h>
83 #include <netipsec/key.h>
84 #endif /* IPSEC */
85 #include <netinet/in_rss.h>
86 
87 #include <sys/socketvar.h>
88 
89 #include <security/mac/mac_framework.h>
90 
91 #ifdef CTASSERT
92 CTASSERT(sizeof(struct ip) == 20);
93 #endif
94 
95 /* IP reassembly functions are defined in ip_reass.c. */
96 extern void ipreass_init(void);
97 extern void ipreass_drain(void);
98 extern void ipreass_slowtimo(void);
99 #ifdef VIMAGE
100 extern void ipreass_destroy(void);
101 #endif
102 
103 struct rmlock in_ifaddr_lock;
104 RM_SYSINIT(in_ifaddr_lock, &in_ifaddr_lock, "in_ifaddr_lock");
105 
106 VNET_DEFINE(int, rsvp_on);
107 
108 VNET_DEFINE(int, ipforwarding);
109 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_VNET | CTLFLAG_RW,
110     &VNET_NAME(ipforwarding), 0,
111     "Enable IP forwarding between interfaces");
112 
113 static VNET_DEFINE(int, ipsendredirects) = 1;	/* XXX */
114 #define	V_ipsendredirects	VNET(ipsendredirects)
115 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_VNET | CTLFLAG_RW,
116     &VNET_NAME(ipsendredirects), 0,
117     "Enable sending IP redirects");
118 
119 /*
120  * XXX - Setting ip_checkinterface mostly implements the receive side of
121  * the Strong ES model described in RFC 1122, but since the routing table
122  * and transmit implementation do not implement the Strong ES model,
123  * setting this to 1 results in an odd hybrid.
124  *
125  * XXX - ip_checkinterface currently must be disabled if you use ipnat
126  * to translate the destination address to another local interface.
127  *
128  * XXX - ip_checkinterface must be disabled if you add IP aliases
129  * to the loopback interface instead of the interface where the
130  * packets for those addresses are received.
131  */
132 static VNET_DEFINE(int, ip_checkinterface);
133 #define	V_ip_checkinterface	VNET(ip_checkinterface)
134 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_VNET | CTLFLAG_RW,
135     &VNET_NAME(ip_checkinterface), 0,
136     "Verify packet arrives on correct interface");
137 
138 VNET_DEFINE(struct pfil_head, inet_pfil_hook);	/* Packet filter hooks */
139 
140 static struct netisr_handler ip_nh = {
141 	.nh_name = "ip",
142 	.nh_handler = ip_input,
143 	.nh_proto = NETISR_IP,
144 #ifdef	RSS
145 	.nh_m2cpuid = rss_soft_m2cpuid_v4,
146 	.nh_policy = NETISR_POLICY_CPU,
147 	.nh_dispatch = NETISR_DISPATCH_HYBRID,
148 #else
149 	.nh_policy = NETISR_POLICY_FLOW,
150 #endif
151 };
152 
153 #ifdef	RSS
154 /*
155  * Directly dispatched frames are currently assumed
156  * to have a flowid already calculated.
157  *
158  * It should likely have something that assert it
159  * actually has valid flow details.
160  */
161 static struct netisr_handler ip_direct_nh = {
162 	.nh_name = "ip_direct",
163 	.nh_handler = ip_direct_input,
164 	.nh_proto = NETISR_IP_DIRECT,
165 	.nh_m2cpuid = rss_soft_m2cpuid_v4,
166 	.nh_policy = NETISR_POLICY_CPU,
167 	.nh_dispatch = NETISR_DISPATCH_HYBRID,
168 };
169 #endif
170 
171 extern	struct domain inetdomain;
172 extern	struct protosw inetsw[];
173 u_char	ip_protox[IPPROTO_MAX];
174 VNET_DEFINE(struct in_ifaddrhead, in_ifaddrhead);  /* first inet address */
175 VNET_DEFINE(struct in_ifaddrhashhead *, in_ifaddrhashtbl); /* inet addr hash table  */
176 VNET_DEFINE(u_long, in_ifaddrhmask);		/* mask for hash table */
177 
178 #ifdef IPCTL_DEFMTU
179 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
180     &ip_mtu, 0, "Default MTU");
181 #endif
182 
183 #ifdef IPSTEALTH
184 VNET_DEFINE(int, ipstealth);
185 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_VNET | CTLFLAG_RW,
186     &VNET_NAME(ipstealth), 0,
187     "IP stealth mode, no TTL decrementation on forwarding");
188 #endif
189 
190 /*
191  * IP statistics are stored in the "array" of counter(9)s.
192  */
193 VNET_PCPUSTAT_DEFINE(struct ipstat, ipstat);
194 VNET_PCPUSTAT_SYSINIT(ipstat);
195 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, IPCTL_STATS, stats, struct ipstat, ipstat,
196     "IP statistics (struct ipstat, netinet/ip_var.h)");
197 
198 #ifdef VIMAGE
199 VNET_PCPUSTAT_SYSUNINIT(ipstat);
200 #endif /* VIMAGE */
201 
202 /*
203  * Kernel module interface for updating ipstat.  The argument is an index
204  * into ipstat treated as an array.
205  */
206 void
207 kmod_ipstat_inc(int statnum)
208 {
209 
210 	counter_u64_add(VNET(ipstat)[statnum], 1);
211 }
212 
213 void
214 kmod_ipstat_dec(int statnum)
215 {
216 
217 	counter_u64_add(VNET(ipstat)[statnum], -1);
218 }
219 
220 static int
221 sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS)
222 {
223 	int error, qlimit;
224 
225 	netisr_getqlimit(&ip_nh, &qlimit);
226 	error = sysctl_handle_int(oidp, &qlimit, 0, req);
227 	if (error || !req->newptr)
228 		return (error);
229 	if (qlimit < 1)
230 		return (EINVAL);
231 	return (netisr_setqlimit(&ip_nh, qlimit));
232 }
233 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen,
234     CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet_intr_queue_maxlen, "I",
235     "Maximum size of the IP input queue");
236 
237 static int
238 sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS)
239 {
240 	u_int64_t qdrops_long;
241 	int error, qdrops;
242 
243 	netisr_getqdrops(&ip_nh, &qdrops_long);
244 	qdrops = qdrops_long;
245 	error = sysctl_handle_int(oidp, &qdrops, 0, req);
246 	if (error || !req->newptr)
247 		return (error);
248 	if (qdrops != 0)
249 		return (EINVAL);
250 	netisr_clearqdrops(&ip_nh);
251 	return (0);
252 }
253 
254 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops,
255     CTLTYPE_INT|CTLFLAG_RD, 0, 0, sysctl_netinet_intr_queue_drops, "I",
256     "Number of packets dropped from the IP input queue");
257 
258 #ifdef	RSS
259 static int
260 sysctl_netinet_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS)
261 {
262 	int error, qlimit;
263 
264 	netisr_getqlimit(&ip_direct_nh, &qlimit);
265 	error = sysctl_handle_int(oidp, &qlimit, 0, req);
266 	if (error || !req->newptr)
267 		return (error);
268 	if (qlimit < 1)
269 		return (EINVAL);
270 	return (netisr_setqlimit(&ip_direct_nh, qlimit));
271 }
272 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_direct_queue_maxlen,
273     CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet_intr_direct_queue_maxlen, "I",
274     "Maximum size of the IP direct input queue");
275 
276 static int
277 sysctl_netinet_intr_direct_queue_drops(SYSCTL_HANDLER_ARGS)
278 {
279 	u_int64_t qdrops_long;
280 	int error, qdrops;
281 
282 	netisr_getqdrops(&ip_direct_nh, &qdrops_long);
283 	qdrops = qdrops_long;
284 	error = sysctl_handle_int(oidp, &qdrops, 0, req);
285 	if (error || !req->newptr)
286 		return (error);
287 	if (qdrops != 0)
288 		return (EINVAL);
289 	netisr_clearqdrops(&ip_direct_nh);
290 	return (0);
291 }
292 
293 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_direct_queue_drops,
294     CTLTYPE_INT|CTLFLAG_RD, 0, 0, sysctl_netinet_intr_direct_queue_drops, "I",
295     "Number of packets dropped from the IP direct input queue");
296 #endif	/* RSS */
297 
298 /*
299  * IP initialization: fill in IP protocol switch table.
300  * All protocols not implemented in kernel go to raw IP protocol handler.
301  */
302 void
303 ip_init(void)
304 {
305 	struct protosw *pr;
306 	int i;
307 
308 	TAILQ_INIT(&V_in_ifaddrhead);
309 	V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask);
310 
311 	/* Initialize IP reassembly queue. */
312 	ipreass_init();
313 
314 	/* Initialize packet filter hooks. */
315 	V_inet_pfil_hook.ph_type = PFIL_TYPE_AF;
316 	V_inet_pfil_hook.ph_af = AF_INET;
317 	if ((i = pfil_head_register(&V_inet_pfil_hook)) != 0)
318 		printf("%s: WARNING: unable to register pfil hook, "
319 			"error %d\n", __func__, i);
320 
321 	/* Skip initialization of globals for non-default instances. */
322 	if (!IS_DEFAULT_VNET(curvnet))
323 		return;
324 
325 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
326 	if (pr == NULL)
327 		panic("ip_init: PF_INET not found");
328 
329 	/* Initialize the entire ip_protox[] array to IPPROTO_RAW. */
330 	for (i = 0; i < IPPROTO_MAX; i++)
331 		ip_protox[i] = pr - inetsw;
332 	/*
333 	 * Cycle through IP protocols and put them into the appropriate place
334 	 * in ip_protox[].
335 	 */
336 	for (pr = inetdomain.dom_protosw;
337 	    pr < inetdomain.dom_protoswNPROTOSW; pr++)
338 		if (pr->pr_domain->dom_family == PF_INET &&
339 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
340 			/* Be careful to only index valid IP protocols. */
341 			if (pr->pr_protocol < IPPROTO_MAX)
342 				ip_protox[pr->pr_protocol] = pr - inetsw;
343 		}
344 
345 	netisr_register(&ip_nh);
346 #ifdef	RSS
347 	netisr_register(&ip_direct_nh);
348 #endif
349 }
350 
351 #ifdef VIMAGE
352 void
353 ip_destroy(void)
354 {
355 	int i;
356 
357 	if ((i = pfil_head_unregister(&V_inet_pfil_hook)) != 0)
358 		printf("%s: WARNING: unable to unregister pfil hook, "
359 		    "error %d\n", __func__, i);
360 
361 	/* Cleanup in_ifaddr hash table; should be empty. */
362 	hashdestroy(V_in_ifaddrhashtbl, M_IFADDR, V_in_ifaddrhmask);
363 
364 	/* Destroy IP reassembly queue. */
365 	ipreass_destroy();
366 }
367 #endif
368 
369 #ifdef	RSS
370 /*
371  * IP direct input routine.
372  *
373  * This is called when reinjecting completed fragments where
374  * all of the previous checking and book-keeping has been done.
375  */
376 void
377 ip_direct_input(struct mbuf *m)
378 {
379 	struct ip *ip;
380 	int hlen;
381 
382 	ip = mtod(m, struct ip *);
383 	hlen = ip->ip_hl << 2;
384 
385 	IPSTAT_INC(ips_delivered);
386 	(*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
387 	return;
388 }
389 #endif
390 
391 /*
392  * Ip input routine.  Checksum and byte swap header.  If fragmented
393  * try to reassemble.  Process options.  Pass to next level.
394  */
395 void
396 ip_input(struct mbuf *m)
397 {
398 	struct ip *ip = NULL;
399 	struct in_ifaddr *ia = NULL;
400 	struct ifaddr *ifa;
401 	struct ifnet *ifp;
402 	int    checkif, hlen = 0;
403 	uint16_t sum, ip_len;
404 	int dchg = 0;				/* dest changed after fw */
405 	struct in_addr odst;			/* original dst address */
406 
407 	M_ASSERTPKTHDR(m);
408 
409 	if (m->m_flags & M_FASTFWD_OURS) {
410 		m->m_flags &= ~M_FASTFWD_OURS;
411 		/* Set up some basics that will be used later. */
412 		ip = mtod(m, struct ip *);
413 		hlen = ip->ip_hl << 2;
414 		ip_len = ntohs(ip->ip_len);
415 		goto ours;
416 	}
417 
418 	IPSTAT_INC(ips_total);
419 
420 	if (m->m_pkthdr.len < sizeof(struct ip))
421 		goto tooshort;
422 
423 	if (m->m_len < sizeof (struct ip) &&
424 	    (m = m_pullup(m, sizeof (struct ip))) == NULL) {
425 		IPSTAT_INC(ips_toosmall);
426 		return;
427 	}
428 	ip = mtod(m, struct ip *);
429 
430 	if (ip->ip_v != IPVERSION) {
431 		IPSTAT_INC(ips_badvers);
432 		goto bad;
433 	}
434 
435 	hlen = ip->ip_hl << 2;
436 	if (hlen < sizeof(struct ip)) {	/* minimum header length */
437 		IPSTAT_INC(ips_badhlen);
438 		goto bad;
439 	}
440 	if (hlen > m->m_len) {
441 		if ((m = m_pullup(m, hlen)) == NULL) {
442 			IPSTAT_INC(ips_badhlen);
443 			return;
444 		}
445 		ip = mtod(m, struct ip *);
446 	}
447 
448 	IP_PROBE(receive, NULL, NULL, ip, m->m_pkthdr.rcvif, ip, NULL);
449 
450 	/* 127/8 must not appear on wire - RFC1122 */
451 	ifp = m->m_pkthdr.rcvif;
452 	if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
453 	    (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
454 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
455 			IPSTAT_INC(ips_badaddr);
456 			goto bad;
457 		}
458 	}
459 
460 	if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
461 		sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
462 	} else {
463 		if (hlen == sizeof(struct ip)) {
464 			sum = in_cksum_hdr(ip);
465 		} else {
466 			sum = in_cksum(m, hlen);
467 		}
468 	}
469 	if (sum) {
470 		IPSTAT_INC(ips_badsum);
471 		goto bad;
472 	}
473 
474 #ifdef ALTQ
475 	if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
476 		/* packet is dropped by traffic conditioner */
477 		return;
478 #endif
479 
480 	ip_len = ntohs(ip->ip_len);
481 	if (ip_len < hlen) {
482 		IPSTAT_INC(ips_badlen);
483 		goto bad;
484 	}
485 
486 	/*
487 	 * Check that the amount of data in the buffers
488 	 * is as at least much as the IP header would have us expect.
489 	 * Trim mbufs if longer than we expect.
490 	 * Drop packet if shorter than we expect.
491 	 */
492 	if (m->m_pkthdr.len < ip_len) {
493 tooshort:
494 		IPSTAT_INC(ips_tooshort);
495 		goto bad;
496 	}
497 	if (m->m_pkthdr.len > ip_len) {
498 		if (m->m_len == m->m_pkthdr.len) {
499 			m->m_len = ip_len;
500 			m->m_pkthdr.len = ip_len;
501 		} else
502 			m_adj(m, ip_len - m->m_pkthdr.len);
503 	}
504 
505 	/* Try to forward the packet, but if we fail continue */
506 #ifdef IPSEC
507 	/* For now we do not handle IPSEC in tryforward. */
508 	if (!key_havesp(IPSEC_DIR_INBOUND) && !key_havesp(IPSEC_DIR_OUTBOUND) &&
509 	    (V_ipforwarding == 1))
510 		if (ip_tryforward(m) == NULL)
511 			return;
512 	/*
513 	 * Bypass packet filtering for packets previously handled by IPsec.
514 	 */
515 	if (ip_ipsec_filtertunnel(m))
516 		goto passin;
517 #else
518 	if (V_ipforwarding == 1)
519 		if (ip_tryforward(m) == NULL)
520 			return;
521 #endif /* IPSEC */
522 
523 	/*
524 	 * Run through list of hooks for input packets.
525 	 *
526 	 * NB: Beware of the destination address changing (e.g.
527 	 *     by NAT rewriting).  When this happens, tell
528 	 *     ip_forward to do the right thing.
529 	 */
530 
531 	/* Jump over all PFIL processing if hooks are not active. */
532 	if (!PFIL_HOOKED(&V_inet_pfil_hook))
533 		goto passin;
534 
535 	odst = ip->ip_dst;
536 	if (pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_IN, NULL) != 0)
537 		return;
538 	if (m == NULL)			/* consumed by filter */
539 		return;
540 
541 	ip = mtod(m, struct ip *);
542 	dchg = (odst.s_addr != ip->ip_dst.s_addr);
543 	ifp = m->m_pkthdr.rcvif;
544 
545 	if (m->m_flags & M_FASTFWD_OURS) {
546 		m->m_flags &= ~M_FASTFWD_OURS;
547 		goto ours;
548 	}
549 	if (m->m_flags & M_IP_NEXTHOP) {
550 		dchg = (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL);
551 		if (dchg != 0) {
552 			/*
553 			 * Directly ship the packet on.  This allows
554 			 * forwarding packets originally destined to us
555 			 * to some other directly connected host.
556 			 */
557 			ip_forward(m, 1);
558 			return;
559 		}
560 	}
561 passin:
562 
563 	/*
564 	 * Process options and, if not destined for us,
565 	 * ship it on.  ip_dooptions returns 1 when an
566 	 * error was detected (causing an icmp message
567 	 * to be sent and the original packet to be freed).
568 	 */
569 	if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
570 		return;
571 
572         /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
573          * matter if it is destined to another node, or whether it is
574          * a multicast one, RSVP wants it! and prevents it from being forwarded
575          * anywhere else. Also checks if the rsvp daemon is running before
576 	 * grabbing the packet.
577          */
578 	if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP)
579 		goto ours;
580 
581 	/*
582 	 * Check our list of addresses, to see if the packet is for us.
583 	 * If we don't have any addresses, assume any unicast packet
584 	 * we receive might be for us (and let the upper layers deal
585 	 * with it).
586 	 */
587 	if (TAILQ_EMPTY(&V_in_ifaddrhead) &&
588 	    (m->m_flags & (M_MCAST|M_BCAST)) == 0)
589 		goto ours;
590 
591 	/*
592 	 * Enable a consistency check between the destination address
593 	 * and the arrival interface for a unicast packet (the RFC 1122
594 	 * strong ES model) if IP forwarding is disabled and the packet
595 	 * is not locally generated and the packet is not subject to
596 	 * 'ipfw fwd'.
597 	 *
598 	 * XXX - Checking also should be disabled if the destination
599 	 * address is ipnat'ed to a different interface.
600 	 *
601 	 * XXX - Checking is incompatible with IP aliases added
602 	 * to the loopback interface instead of the interface where
603 	 * the packets are received.
604 	 *
605 	 * XXX - This is the case for carp vhost IPs as well so we
606 	 * insert a workaround. If the packet got here, we already
607 	 * checked with carp_iamatch() and carp_forus().
608 	 */
609 	checkif = V_ip_checkinterface && (V_ipforwarding == 0) &&
610 	    ifp != NULL && ((ifp->if_flags & IFF_LOOPBACK) == 0) &&
611 	    ifp->if_carp == NULL && (dchg == 0);
612 
613 	/*
614 	 * Check for exact addresses in the hash bucket.
615 	 */
616 	/* IN_IFADDR_RLOCK(); */
617 	LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
618 		/*
619 		 * If the address matches, verify that the packet
620 		 * arrived via the correct interface if checking is
621 		 * enabled.
622 		 */
623 		if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr &&
624 		    (!checkif || ia->ia_ifp == ifp)) {
625 			counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
626 			counter_u64_add(ia->ia_ifa.ifa_ibytes,
627 			    m->m_pkthdr.len);
628 			/* IN_IFADDR_RUNLOCK(); */
629 			goto ours;
630 		}
631 	}
632 	/* IN_IFADDR_RUNLOCK(); */
633 
634 	/*
635 	 * Check for broadcast addresses.
636 	 *
637 	 * Only accept broadcast packets that arrive via the matching
638 	 * interface.  Reception of forwarded directed broadcasts would
639 	 * be handled via ip_forward() and ether_output() with the loopback
640 	 * into the stack for SIMPLEX interfaces handled by ether_output().
641 	 */
642 	if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
643 		IF_ADDR_RLOCK(ifp);
644 	        TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
645 			if (ifa->ifa_addr->sa_family != AF_INET)
646 				continue;
647 			ia = ifatoia(ifa);
648 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
649 			    ip->ip_dst.s_addr) {
650 				counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
651 				counter_u64_add(ia->ia_ifa.ifa_ibytes,
652 				    m->m_pkthdr.len);
653 				IF_ADDR_RUNLOCK(ifp);
654 				goto ours;
655 			}
656 #ifdef BOOTP_COMPAT
657 			if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) {
658 				counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
659 				counter_u64_add(ia->ia_ifa.ifa_ibytes,
660 				    m->m_pkthdr.len);
661 				IF_ADDR_RUNLOCK(ifp);
662 				goto ours;
663 			}
664 #endif
665 		}
666 		IF_ADDR_RUNLOCK(ifp);
667 		ia = NULL;
668 	}
669 	/* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */
670 	if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) {
671 		IPSTAT_INC(ips_cantforward);
672 		m_freem(m);
673 		return;
674 	}
675 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
676 		if (V_ip_mrouter) {
677 			/*
678 			 * If we are acting as a multicast router, all
679 			 * incoming multicast packets are passed to the
680 			 * kernel-level multicast forwarding function.
681 			 * The packet is returned (relatively) intact; if
682 			 * ip_mforward() returns a non-zero value, the packet
683 			 * must be discarded, else it may be accepted below.
684 			 */
685 			if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) {
686 				IPSTAT_INC(ips_cantforward);
687 				m_freem(m);
688 				return;
689 			}
690 
691 			/*
692 			 * The process-level routing daemon needs to receive
693 			 * all multicast IGMP packets, whether or not this
694 			 * host belongs to their destination groups.
695 			 */
696 			if (ip->ip_p == IPPROTO_IGMP)
697 				goto ours;
698 			IPSTAT_INC(ips_forward);
699 		}
700 		/*
701 		 * Assume the packet is for us, to avoid prematurely taking
702 		 * a lock on the in_multi hash. Protocols must perform
703 		 * their own filtering and update statistics accordingly.
704 		 */
705 		goto ours;
706 	}
707 	if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
708 		goto ours;
709 	if (ip->ip_dst.s_addr == INADDR_ANY)
710 		goto ours;
711 
712 	/*
713 	 * Not for us; forward if possible and desirable.
714 	 */
715 	if (V_ipforwarding == 0) {
716 		IPSTAT_INC(ips_cantforward);
717 		m_freem(m);
718 	} else {
719 		ip_forward(m, dchg);
720 	}
721 	return;
722 
723 ours:
724 #ifdef IPSTEALTH
725 	/*
726 	 * IPSTEALTH: Process non-routing options only
727 	 * if the packet is destined for us.
728 	 */
729 	if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1))
730 		return;
731 #endif /* IPSTEALTH */
732 
733 	/*
734 	 * Attempt reassembly; if it succeeds, proceed.
735 	 * ip_reass() will return a different mbuf.
736 	 */
737 	if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) {
738 		/* XXXGL: shouldn't we save & set m_flags? */
739 		m = ip_reass(m);
740 		if (m == NULL)
741 			return;
742 		ip = mtod(m, struct ip *);
743 		/* Get the header length of the reassembled packet */
744 		hlen = ip->ip_hl << 2;
745 	}
746 
747 #ifdef IPSEC
748 	/*
749 	 * enforce IPsec policy checking if we are seeing last header.
750 	 * note that we do not visit this with protocols with pcb layer
751 	 * code - like udp/tcp/raw ip.
752 	 */
753 	if (ip_ipsec_input(m, ip->ip_p) != 0)
754 		goto bad;
755 #endif /* IPSEC */
756 
757 	/*
758 	 * Switch out to protocol's input routine.
759 	 */
760 	IPSTAT_INC(ips_delivered);
761 
762 	(*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
763 	return;
764 bad:
765 	m_freem(m);
766 }
767 
768 /*
769  * IP timer processing;
770  * if a timer expires on a reassembly
771  * queue, discard it.
772  */
773 void
774 ip_slowtimo(void)
775 {
776 	VNET_ITERATOR_DECL(vnet_iter);
777 
778 	VNET_LIST_RLOCK_NOSLEEP();
779 	VNET_FOREACH(vnet_iter) {
780 		CURVNET_SET(vnet_iter);
781 		ipreass_slowtimo();
782 		CURVNET_RESTORE();
783 	}
784 	VNET_LIST_RUNLOCK_NOSLEEP();
785 }
786 
787 void
788 ip_drain(void)
789 {
790 	VNET_ITERATOR_DECL(vnet_iter);
791 
792 	VNET_LIST_RLOCK_NOSLEEP();
793 	VNET_FOREACH(vnet_iter) {
794 		CURVNET_SET(vnet_iter);
795 		ipreass_drain();
796 		CURVNET_RESTORE();
797 	}
798 	VNET_LIST_RUNLOCK_NOSLEEP();
799 }
800 
801 /*
802  * The protocol to be inserted into ip_protox[] must be already registered
803  * in inetsw[], either statically or through pf_proto_register().
804  */
805 int
806 ipproto_register(short ipproto)
807 {
808 	struct protosw *pr;
809 
810 	/* Sanity checks. */
811 	if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
812 		return (EPROTONOSUPPORT);
813 
814 	/*
815 	 * The protocol slot must not be occupied by another protocol
816 	 * already.  An index pointing to IPPROTO_RAW is unused.
817 	 */
818 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
819 	if (pr == NULL)
820 		return (EPFNOSUPPORT);
821 	if (ip_protox[ipproto] != pr - inetsw)	/* IPPROTO_RAW */
822 		return (EEXIST);
823 
824 	/* Find the protocol position in inetsw[] and set the index. */
825 	for (pr = inetdomain.dom_protosw;
826 	     pr < inetdomain.dom_protoswNPROTOSW; pr++) {
827 		if (pr->pr_domain->dom_family == PF_INET &&
828 		    pr->pr_protocol && pr->pr_protocol == ipproto) {
829 			ip_protox[pr->pr_protocol] = pr - inetsw;
830 			return (0);
831 		}
832 	}
833 	return (EPROTONOSUPPORT);
834 }
835 
836 int
837 ipproto_unregister(short ipproto)
838 {
839 	struct protosw *pr;
840 
841 	/* Sanity checks. */
842 	if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
843 		return (EPROTONOSUPPORT);
844 
845 	/* Check if the protocol was indeed registered. */
846 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
847 	if (pr == NULL)
848 		return (EPFNOSUPPORT);
849 	if (ip_protox[ipproto] == pr - inetsw)  /* IPPROTO_RAW */
850 		return (ENOENT);
851 
852 	/* Reset the protocol slot to IPPROTO_RAW. */
853 	ip_protox[ipproto] = pr - inetsw;
854 	return (0);
855 }
856 
857 /*
858  * Given address of next destination (final or next hop), return (referenced)
859  * internet address info of interface to be used to get there.
860  */
861 struct in_ifaddr *
862 ip_rtaddr(struct in_addr dst, u_int fibnum)
863 {
864 	struct route sro;
865 	struct sockaddr_in *sin;
866 	struct in_ifaddr *ia;
867 
868 	bzero(&sro, sizeof(sro));
869 	sin = (struct sockaddr_in *)&sro.ro_dst;
870 	sin->sin_family = AF_INET;
871 	sin->sin_len = sizeof(*sin);
872 	sin->sin_addr = dst;
873 	in_rtalloc_ign(&sro, 0, fibnum);
874 
875 	if (sro.ro_rt == NULL)
876 		return (NULL);
877 
878 	ia = ifatoia(sro.ro_rt->rt_ifa);
879 	ifa_ref(&ia->ia_ifa);
880 	RTFREE(sro.ro_rt);
881 	return (ia);
882 }
883 
884 u_char inetctlerrmap[PRC_NCMDS] = {
885 	0,		0,		0,		0,
886 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
887 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
888 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
889 	0,		0,		EHOSTUNREACH,	0,
890 	ENOPROTOOPT,	ECONNREFUSED
891 };
892 
893 /*
894  * Forward a packet.  If some error occurs return the sender
895  * an icmp packet.  Note we can't always generate a meaningful
896  * icmp message because icmp doesn't have a large enough repertoire
897  * of codes and types.
898  *
899  * If not forwarding, just drop the packet.  This could be confusing
900  * if ipforwarding was zero but some routing protocol was advancing
901  * us as a gateway to somewhere.  However, we must let the routing
902  * protocol deal with that.
903  *
904  * The srcrt parameter indicates whether the packet is being forwarded
905  * via a source route.
906  */
907 void
908 ip_forward(struct mbuf *m, int srcrt)
909 {
910 	struct ip *ip = mtod(m, struct ip *);
911 	struct in_ifaddr *ia;
912 	struct mbuf *mcopy;
913 	struct sockaddr_in *sin;
914 	struct in_addr dest;
915 	struct route ro;
916 	int error, type = 0, code = 0, mtu = 0;
917 
918 	if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
919 		IPSTAT_INC(ips_cantforward);
920 		m_freem(m);
921 		return;
922 	}
923 #ifdef IPSEC
924 	if (ip_ipsec_fwd(m) != 0) {
925 		IPSTAT_INC(ips_cantforward);
926 		m_freem(m);
927 		return;
928 	}
929 #endif /* IPSEC */
930 #ifdef IPSTEALTH
931 	if (!V_ipstealth) {
932 #endif
933 		if (ip->ip_ttl <= IPTTLDEC) {
934 			icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS,
935 			    0, 0);
936 			return;
937 		}
938 #ifdef IPSTEALTH
939 	}
940 #endif
941 
942 	bzero(&ro, sizeof(ro));
943 	sin = (struct sockaddr_in *)&ro.ro_dst;
944 	sin->sin_family = AF_INET;
945 	sin->sin_len = sizeof(*sin);
946 	sin->sin_addr = ip->ip_dst;
947 #ifdef RADIX_MPATH
948 	rtalloc_mpath_fib(&ro,
949 	    ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
950 	    M_GETFIB(m));
951 #else
952 	in_rtalloc_ign(&ro, 0, M_GETFIB(m));
953 #endif
954 	if (ro.ro_rt != NULL) {
955 		ia = ifatoia(ro.ro_rt->rt_ifa);
956 		ifa_ref(&ia->ia_ifa);
957 	} else
958 		ia = NULL;
959 #ifndef IPSEC
960 	/*
961 	 * 'ia' may be NULL if there is no route for this destination.
962 	 * In case of IPsec, Don't discard it just yet, but pass it to
963 	 * ip_output in case of outgoing IPsec policy.
964 	 */
965 	if (!srcrt && ia == NULL) {
966 		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
967 		RO_RTFREE(&ro);
968 		return;
969 	}
970 #endif
971 
972 	/*
973 	 * Save the IP header and at most 8 bytes of the payload,
974 	 * in case we need to generate an ICMP message to the src.
975 	 *
976 	 * XXX this can be optimized a lot by saving the data in a local
977 	 * buffer on the stack (72 bytes at most), and only allocating the
978 	 * mbuf if really necessary. The vast majority of the packets
979 	 * are forwarded without having to send an ICMP back (either
980 	 * because unnecessary, or because rate limited), so we are
981 	 * really we are wasting a lot of work here.
982 	 *
983 	 * We don't use m_copy() because it might return a reference
984 	 * to a shared cluster. Both this function and ip_output()
985 	 * assume exclusive access to the IP header in `m', so any
986 	 * data in a cluster may change before we reach icmp_error().
987 	 */
988 	mcopy = m_gethdr(M_NOWAIT, m->m_type);
989 	if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_NOWAIT)) {
990 		/*
991 		 * It's probably ok if the pkthdr dup fails (because
992 		 * the deep copy of the tag chain failed), but for now
993 		 * be conservative and just discard the copy since
994 		 * code below may some day want the tags.
995 		 */
996 		m_free(mcopy);
997 		mcopy = NULL;
998 	}
999 	if (mcopy != NULL) {
1000 		mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy));
1001 		mcopy->m_pkthdr.len = mcopy->m_len;
1002 		m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1003 	}
1004 
1005 #ifdef IPSTEALTH
1006 	if (!V_ipstealth) {
1007 #endif
1008 		ip->ip_ttl -= IPTTLDEC;
1009 #ifdef IPSTEALTH
1010 	}
1011 #endif
1012 
1013 	/*
1014 	 * If forwarding packet using same interface that it came in on,
1015 	 * perhaps should send a redirect to sender to shortcut a hop.
1016 	 * Only send redirect if source is sending directly to us,
1017 	 * and if packet was not source routed (or has any options).
1018 	 * Also, don't send redirect if forwarding using a default route
1019 	 * or a route modified by a redirect.
1020 	 */
1021 	dest.s_addr = 0;
1022 	if (!srcrt && V_ipsendredirects &&
1023 	    ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) {
1024 		struct rtentry *rt;
1025 
1026 		rt = ro.ro_rt;
1027 
1028 		if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1029 		    satosin(rt_key(rt))->sin_addr.s_addr != 0) {
1030 #define	RTA(rt)	((struct in_ifaddr *)(rt->rt_ifa))
1031 			u_long src = ntohl(ip->ip_src.s_addr);
1032 
1033 			if (RTA(rt) &&
1034 			    (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1035 				if (rt->rt_flags & RTF_GATEWAY)
1036 					dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr;
1037 				else
1038 					dest.s_addr = ip->ip_dst.s_addr;
1039 				/* Router requirements says to only send host redirects */
1040 				type = ICMP_REDIRECT;
1041 				code = ICMP_REDIRECT_HOST;
1042 			}
1043 		}
1044 	}
1045 
1046 	error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
1047 
1048 	if (error == EMSGSIZE && ro.ro_rt)
1049 		mtu = ro.ro_rt->rt_mtu;
1050 	RO_RTFREE(&ro);
1051 
1052 	if (error)
1053 		IPSTAT_INC(ips_cantforward);
1054 	else {
1055 		IPSTAT_INC(ips_forward);
1056 		if (type)
1057 			IPSTAT_INC(ips_redirectsent);
1058 		else {
1059 			if (mcopy)
1060 				m_freem(mcopy);
1061 			if (ia != NULL)
1062 				ifa_free(&ia->ia_ifa);
1063 			return;
1064 		}
1065 	}
1066 	if (mcopy == NULL) {
1067 		if (ia != NULL)
1068 			ifa_free(&ia->ia_ifa);
1069 		return;
1070 	}
1071 
1072 	switch (error) {
1073 
1074 	case 0:				/* forwarded, but need redirect */
1075 		/* type, code set above */
1076 		break;
1077 
1078 	case ENETUNREACH:
1079 	case EHOSTUNREACH:
1080 	case ENETDOWN:
1081 	case EHOSTDOWN:
1082 	default:
1083 		type = ICMP_UNREACH;
1084 		code = ICMP_UNREACH_HOST;
1085 		break;
1086 
1087 	case EMSGSIZE:
1088 		type = ICMP_UNREACH;
1089 		code = ICMP_UNREACH_NEEDFRAG;
1090 
1091 #ifdef IPSEC
1092 		/*
1093 		 * If IPsec is configured for this path,
1094 		 * override any possibly mtu value set by ip_output.
1095 		 */
1096 		mtu = ip_ipsec_mtu(mcopy, mtu);
1097 #endif /* IPSEC */
1098 		/*
1099 		 * If the MTU was set before make sure we are below the
1100 		 * interface MTU.
1101 		 * If the MTU wasn't set before use the interface mtu or
1102 		 * fall back to the next smaller mtu step compared to the
1103 		 * current packet size.
1104 		 */
1105 		if (mtu != 0) {
1106 			if (ia != NULL)
1107 				mtu = min(mtu, ia->ia_ifp->if_mtu);
1108 		} else {
1109 			if (ia != NULL)
1110 				mtu = ia->ia_ifp->if_mtu;
1111 			else
1112 				mtu = ip_next_mtu(ntohs(ip->ip_len), 0);
1113 		}
1114 		IPSTAT_INC(ips_cantfrag);
1115 		break;
1116 
1117 	case ENOBUFS:
1118 	case EACCES:			/* ipfw denied packet */
1119 		m_freem(mcopy);
1120 		if (ia != NULL)
1121 			ifa_free(&ia->ia_ifa);
1122 		return;
1123 	}
1124 	if (ia != NULL)
1125 		ifa_free(&ia->ia_ifa);
1126 	icmp_error(mcopy, type, code, dest.s_addr, mtu);
1127 }
1128 
1129 void
1130 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1131     struct mbuf *m)
1132 {
1133 
1134 	if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) {
1135 		struct bintime bt;
1136 
1137 		bintime(&bt);
1138 		if (inp->inp_socket->so_options & SO_BINTIME) {
1139 			*mp = sbcreatecontrol((caddr_t)&bt, sizeof(bt),
1140 			    SCM_BINTIME, SOL_SOCKET);
1141 			if (*mp)
1142 				mp = &(*mp)->m_next;
1143 		}
1144 		if (inp->inp_socket->so_options & SO_TIMESTAMP) {
1145 			struct timeval tv;
1146 
1147 			bintime2timeval(&bt, &tv);
1148 			*mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv),
1149 			    SCM_TIMESTAMP, SOL_SOCKET);
1150 			if (*mp)
1151 				mp = &(*mp)->m_next;
1152 		}
1153 	}
1154 	if (inp->inp_flags & INP_RECVDSTADDR) {
1155 		*mp = sbcreatecontrol((caddr_t)&ip->ip_dst,
1156 		    sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1157 		if (*mp)
1158 			mp = &(*mp)->m_next;
1159 	}
1160 	if (inp->inp_flags & INP_RECVTTL) {
1161 		*mp = sbcreatecontrol((caddr_t)&ip->ip_ttl,
1162 		    sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
1163 		if (*mp)
1164 			mp = &(*mp)->m_next;
1165 	}
1166 #ifdef notyet
1167 	/* XXX
1168 	 * Moving these out of udp_input() made them even more broken
1169 	 * than they already were.
1170 	 */
1171 	/* options were tossed already */
1172 	if (inp->inp_flags & INP_RECVOPTS) {
1173 		*mp = sbcreatecontrol((caddr_t)opts_deleted_above,
1174 		    sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1175 		if (*mp)
1176 			mp = &(*mp)->m_next;
1177 	}
1178 	/* ip_srcroute doesn't do what we want here, need to fix */
1179 	if (inp->inp_flags & INP_RECVRETOPTS) {
1180 		*mp = sbcreatecontrol((caddr_t)ip_srcroute(m),
1181 		    sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1182 		if (*mp)
1183 			mp = &(*mp)->m_next;
1184 	}
1185 #endif
1186 	if (inp->inp_flags & INP_RECVIF) {
1187 		struct ifnet *ifp;
1188 		struct sdlbuf {
1189 			struct sockaddr_dl sdl;
1190 			u_char	pad[32];
1191 		} sdlbuf;
1192 		struct sockaddr_dl *sdp;
1193 		struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1194 
1195 		if ((ifp = m->m_pkthdr.rcvif) &&
1196 		    ifp->if_index && ifp->if_index <= V_if_index) {
1197 			sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1198 			/*
1199 			 * Change our mind and don't try copy.
1200 			 */
1201 			if (sdp->sdl_family != AF_LINK ||
1202 			    sdp->sdl_len > sizeof(sdlbuf)) {
1203 				goto makedummy;
1204 			}
1205 			bcopy(sdp, sdl2, sdp->sdl_len);
1206 		} else {
1207 makedummy:
1208 			sdl2->sdl_len =
1209 			    offsetof(struct sockaddr_dl, sdl_data[0]);
1210 			sdl2->sdl_family = AF_LINK;
1211 			sdl2->sdl_index = 0;
1212 			sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1213 		}
1214 		*mp = sbcreatecontrol((caddr_t)sdl2, sdl2->sdl_len,
1215 		    IP_RECVIF, IPPROTO_IP);
1216 		if (*mp)
1217 			mp = &(*mp)->m_next;
1218 	}
1219 	if (inp->inp_flags & INP_RECVTOS) {
1220 		*mp = sbcreatecontrol((caddr_t)&ip->ip_tos,
1221 		    sizeof(u_char), IP_RECVTOS, IPPROTO_IP);
1222 		if (*mp)
1223 			mp = &(*mp)->m_next;
1224 	}
1225 
1226 	if (inp->inp_flags2 & INP_RECVFLOWID) {
1227 		uint32_t flowid, flow_type;
1228 
1229 		flowid = m->m_pkthdr.flowid;
1230 		flow_type = M_HASHTYPE_GET(m);
1231 
1232 		/*
1233 		 * XXX should handle the failure of one or the
1234 		 * other - don't populate both?
1235 		 */
1236 		*mp = sbcreatecontrol((caddr_t) &flowid,
1237 		    sizeof(uint32_t), IP_FLOWID, IPPROTO_IP);
1238 		if (*mp)
1239 			mp = &(*mp)->m_next;
1240 		*mp = sbcreatecontrol((caddr_t) &flow_type,
1241 		    sizeof(uint32_t), IP_FLOWTYPE, IPPROTO_IP);
1242 		if (*mp)
1243 			mp = &(*mp)->m_next;
1244 	}
1245 
1246 #ifdef	RSS
1247 	if (inp->inp_flags2 & INP_RECVRSSBUCKETID) {
1248 		uint32_t flowid, flow_type;
1249 		uint32_t rss_bucketid;
1250 
1251 		flowid = m->m_pkthdr.flowid;
1252 		flow_type = M_HASHTYPE_GET(m);
1253 
1254 		if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) {
1255 			*mp = sbcreatecontrol((caddr_t) &rss_bucketid,
1256 			   sizeof(uint32_t), IP_RSSBUCKETID, IPPROTO_IP);
1257 			if (*mp)
1258 				mp = &(*mp)->m_next;
1259 		}
1260 	}
1261 #endif
1262 }
1263 
1264 /*
1265  * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
1266  * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
1267  * locking.  This code remains in ip_input.c as ip_mroute.c is optionally
1268  * compiled.
1269  */
1270 static VNET_DEFINE(int, ip_rsvp_on);
1271 VNET_DEFINE(struct socket *, ip_rsvpd);
1272 
1273 #define	V_ip_rsvp_on		VNET(ip_rsvp_on)
1274 
1275 int
1276 ip_rsvp_init(struct socket *so)
1277 {
1278 
1279 	if (so->so_type != SOCK_RAW ||
1280 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
1281 		return EOPNOTSUPP;
1282 
1283 	if (V_ip_rsvpd != NULL)
1284 		return EADDRINUSE;
1285 
1286 	V_ip_rsvpd = so;
1287 	/*
1288 	 * This may seem silly, but we need to be sure we don't over-increment
1289 	 * the RSVP counter, in case something slips up.
1290 	 */
1291 	if (!V_ip_rsvp_on) {
1292 		V_ip_rsvp_on = 1;
1293 		V_rsvp_on++;
1294 	}
1295 
1296 	return 0;
1297 }
1298 
1299 int
1300 ip_rsvp_done(void)
1301 {
1302 
1303 	V_ip_rsvpd = NULL;
1304 	/*
1305 	 * This may seem silly, but we need to be sure we don't over-decrement
1306 	 * the RSVP counter, in case something slips up.
1307 	 */
1308 	if (V_ip_rsvp_on) {
1309 		V_ip_rsvp_on = 0;
1310 		V_rsvp_on--;
1311 	}
1312 	return 0;
1313 }
1314 
1315 int
1316 rsvp_input(struct mbuf **mp, int *offp, int proto)
1317 {
1318 	struct mbuf *m;
1319 
1320 	m = *mp;
1321 	*mp = NULL;
1322 
1323 	if (rsvp_input_p) { /* call the real one if loaded */
1324 		*mp = m;
1325 		rsvp_input_p(mp, offp, proto);
1326 		return (IPPROTO_DONE);
1327 	}
1328 
1329 	/* Can still get packets with rsvp_on = 0 if there is a local member
1330 	 * of the group to which the RSVP packet is addressed.  But in this
1331 	 * case we want to throw the packet away.
1332 	 */
1333 
1334 	if (!V_rsvp_on) {
1335 		m_freem(m);
1336 		return (IPPROTO_DONE);
1337 	}
1338 
1339 	if (V_ip_rsvpd != NULL) {
1340 		*mp = m;
1341 		rip_input(mp, offp, proto);
1342 		return (IPPROTO_DONE);
1343 	}
1344 	/* Drop the packet */
1345 	m_freem(m);
1346 	return (IPPROTO_DONE);
1347 }
1348