xref: /freebsd/sys/netinet/ip_input.c (revision 277fbb92d5e4cd0938c67f77b08d9ba4ac9d54a6)
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/hhook.h>
45 #include <sys/mbuf.h>
46 #include <sys/malloc.h>
47 #include <sys/domain.h>
48 #include <sys/protosw.h>
49 #include <sys/socket.h>
50 #include <sys/time.h>
51 #include <sys/kernel.h>
52 #include <sys/lock.h>
53 #include <sys/rmlock.h>
54 #include <sys/rwlock.h>
55 #include <sys/sdt.h>
56 #include <sys/syslog.h>
57 #include <sys/sysctl.h>
58 
59 #include <net/pfil.h>
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/route.h>
65 #include <net/netisr.h>
66 #include <net/rss_config.h>
67 #include <net/vnet.h>
68 
69 #include <netinet/in.h>
70 #include <netinet/in_kdtrace.h>
71 #include <netinet/in_systm.h>
72 #include <netinet/in_var.h>
73 #include <netinet/ip.h>
74 #include <netinet/in_pcb.h>
75 #include <netinet/ip_var.h>
76 #include <netinet/ip_fw.h>
77 #include <netinet/ip_icmp.h>
78 #include <netinet/ip_options.h>
79 #include <machine/in_cksum.h>
80 #include <netinet/ip_carp.h>
81 #ifdef IPSEC
82 #include <netinet/ip_ipsec.h>
83 #include <netipsec/ipsec.h>
84 #include <netipsec/key.h>
85 #endif /* IPSEC */
86 #include <netinet/in_rss.h>
87 
88 #include <sys/socketvar.h>
89 
90 #include <security/mac/mac_framework.h>
91 
92 #ifdef CTASSERT
93 CTASSERT(sizeof(struct ip) == 20);
94 #endif
95 
96 /* IP reassembly functions are defined in ip_reass.c. */
97 extern void ipreass_init(void);
98 extern void ipreass_drain(void);
99 extern void ipreass_slowtimo(void);
100 #ifdef VIMAGE
101 extern void ipreass_destroy(void);
102 #endif
103 
104 struct rmlock in_ifaddr_lock;
105 RM_SYSINIT(in_ifaddr_lock, &in_ifaddr_lock, "in_ifaddr_lock");
106 
107 VNET_DEFINE(int, rsvp_on);
108 
109 VNET_DEFINE(int, ipforwarding);
110 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_VNET | CTLFLAG_RW,
111     &VNET_NAME(ipforwarding), 0,
112     "Enable IP forwarding between interfaces");
113 
114 static VNET_DEFINE(int, ipsendredirects) = 1;	/* XXX */
115 #define	V_ipsendredirects	VNET(ipsendredirects)
116 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_VNET | CTLFLAG_RW,
117     &VNET_NAME(ipsendredirects), 0,
118     "Enable sending IP redirects");
119 
120 /*
121  * XXX - Setting ip_checkinterface mostly implements the receive side of
122  * the Strong ES model described in RFC 1122, but since the routing table
123  * and transmit implementation do not implement the Strong ES model,
124  * setting this to 1 results in an odd hybrid.
125  *
126  * XXX - ip_checkinterface currently must be disabled if you use ipnat
127  * to translate the destination address to another local interface.
128  *
129  * XXX - ip_checkinterface must be disabled if you add IP aliases
130  * to the loopback interface instead of the interface where the
131  * packets for those addresses are received.
132  */
133 static VNET_DEFINE(int, ip_checkinterface);
134 #define	V_ip_checkinterface	VNET(ip_checkinterface)
135 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_VNET | CTLFLAG_RW,
136     &VNET_NAME(ip_checkinterface), 0,
137     "Verify packet arrives on correct interface");
138 
139 VNET_DEFINE(struct pfil_head, inet_pfil_hook);	/* Packet filter hooks */
140 
141 static struct netisr_handler ip_nh = {
142 	.nh_name = "ip",
143 	.nh_handler = ip_input,
144 	.nh_proto = NETISR_IP,
145 #ifdef	RSS
146 	.nh_m2cpuid = rss_soft_m2cpuid_v4,
147 	.nh_policy = NETISR_POLICY_CPU,
148 	.nh_dispatch = NETISR_DISPATCH_HYBRID,
149 #else
150 	.nh_policy = NETISR_POLICY_FLOW,
151 #endif
152 };
153 
154 #ifdef	RSS
155 /*
156  * Directly dispatched frames are currently assumed
157  * to have a flowid already calculated.
158  *
159  * It should likely have something that assert it
160  * actually has valid flow details.
161  */
162 static struct netisr_handler ip_direct_nh = {
163 	.nh_name = "ip_direct",
164 	.nh_handler = ip_direct_input,
165 	.nh_proto = NETISR_IP_DIRECT,
166 	.nh_m2cpuid = rss_soft_m2cpuid_v4,
167 	.nh_policy = NETISR_POLICY_CPU,
168 	.nh_dispatch = NETISR_DISPATCH_HYBRID,
169 };
170 #endif
171 
172 extern	struct domain inetdomain;
173 extern	struct protosw inetsw[];
174 u_char	ip_protox[IPPROTO_MAX];
175 VNET_DEFINE(struct in_ifaddrhead, in_ifaddrhead);  /* first inet address */
176 VNET_DEFINE(struct in_ifaddrhashhead *, in_ifaddrhashtbl); /* inet addr hash table  */
177 VNET_DEFINE(u_long, in_ifaddrhmask);		/* mask for hash table */
178 
179 #ifdef IPCTL_DEFMTU
180 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
181     &ip_mtu, 0, "Default MTU");
182 #endif
183 
184 #ifdef IPSTEALTH
185 VNET_DEFINE(int, ipstealth);
186 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_VNET | CTLFLAG_RW,
187     &VNET_NAME(ipstealth), 0,
188     "IP stealth mode, no TTL decrementation on forwarding");
189 #endif
190 
191 /*
192  * IP statistics are stored in the "array" of counter(9)s.
193  */
194 VNET_PCPUSTAT_DEFINE(struct ipstat, ipstat);
195 VNET_PCPUSTAT_SYSINIT(ipstat);
196 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, IPCTL_STATS, stats, struct ipstat, ipstat,
197     "IP statistics (struct ipstat, netinet/ip_var.h)");
198 
199 #ifdef VIMAGE
200 VNET_PCPUSTAT_SYSUNINIT(ipstat);
201 #endif /* VIMAGE */
202 
203 /*
204  * Kernel module interface for updating ipstat.  The argument is an index
205  * into ipstat treated as an array.
206  */
207 void
208 kmod_ipstat_inc(int statnum)
209 {
210 
211 	counter_u64_add(VNET(ipstat)[statnum], 1);
212 }
213 
214 void
215 kmod_ipstat_dec(int statnum)
216 {
217 
218 	counter_u64_add(VNET(ipstat)[statnum], -1);
219 }
220 
221 static int
222 sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS)
223 {
224 	int error, qlimit;
225 
226 	netisr_getqlimit(&ip_nh, &qlimit);
227 	error = sysctl_handle_int(oidp, &qlimit, 0, req);
228 	if (error || !req->newptr)
229 		return (error);
230 	if (qlimit < 1)
231 		return (EINVAL);
232 	return (netisr_setqlimit(&ip_nh, qlimit));
233 }
234 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen,
235     CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet_intr_queue_maxlen, "I",
236     "Maximum size of the IP input queue");
237 
238 static int
239 sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS)
240 {
241 	u_int64_t qdrops_long;
242 	int error, qdrops;
243 
244 	netisr_getqdrops(&ip_nh, &qdrops_long);
245 	qdrops = qdrops_long;
246 	error = sysctl_handle_int(oidp, &qdrops, 0, req);
247 	if (error || !req->newptr)
248 		return (error);
249 	if (qdrops != 0)
250 		return (EINVAL);
251 	netisr_clearqdrops(&ip_nh);
252 	return (0);
253 }
254 
255 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops,
256     CTLTYPE_INT|CTLFLAG_RD, 0, 0, sysctl_netinet_intr_queue_drops, "I",
257     "Number of packets dropped from the IP input queue");
258 
259 #ifdef	RSS
260 static int
261 sysctl_netinet_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS)
262 {
263 	int error, qlimit;
264 
265 	netisr_getqlimit(&ip_direct_nh, &qlimit);
266 	error = sysctl_handle_int(oidp, &qlimit, 0, req);
267 	if (error || !req->newptr)
268 		return (error);
269 	if (qlimit < 1)
270 		return (EINVAL);
271 	return (netisr_setqlimit(&ip_direct_nh, qlimit));
272 }
273 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_direct_queue_maxlen,
274     CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet_intr_direct_queue_maxlen, "I",
275     "Maximum size of the IP direct input queue");
276 
277 static int
278 sysctl_netinet_intr_direct_queue_drops(SYSCTL_HANDLER_ARGS)
279 {
280 	u_int64_t qdrops_long;
281 	int error, qdrops;
282 
283 	netisr_getqdrops(&ip_direct_nh, &qdrops_long);
284 	qdrops = qdrops_long;
285 	error = sysctl_handle_int(oidp, &qdrops, 0, req);
286 	if (error || !req->newptr)
287 		return (error);
288 	if (qdrops != 0)
289 		return (EINVAL);
290 	netisr_clearqdrops(&ip_direct_nh);
291 	return (0);
292 }
293 
294 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_direct_queue_drops,
295     CTLTYPE_INT|CTLFLAG_RD, 0, 0, sysctl_netinet_intr_direct_queue_drops, "I",
296     "Number of packets dropped from the IP direct input queue");
297 #endif	/* RSS */
298 
299 /*
300  * IP initialization: fill in IP protocol switch table.
301  * All protocols not implemented in kernel go to raw IP protocol handler.
302  */
303 void
304 ip_init(void)
305 {
306 	struct protosw *pr;
307 	int i;
308 
309 	TAILQ_INIT(&V_in_ifaddrhead);
310 	V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask);
311 
312 	/* Initialize IP reassembly queue. */
313 	ipreass_init();
314 
315 	/* Initialize packet filter hooks. */
316 	V_inet_pfil_hook.ph_type = PFIL_TYPE_AF;
317 	V_inet_pfil_hook.ph_af = AF_INET;
318 	if ((i = pfil_head_register(&V_inet_pfil_hook)) != 0)
319 		printf("%s: WARNING: unable to register pfil hook, "
320 			"error %d\n", __func__, i);
321 
322 	if (hhook_head_register(HHOOK_TYPE_IPSEC_IN, AF_INET,
323 	    &V_ipsec_hhh_in[HHOOK_IPSEC_INET],
324 	    HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
325 		printf("%s: WARNING: unable to register input helper hook\n",
326 		    __func__);
327 	if (hhook_head_register(HHOOK_TYPE_IPSEC_OUT, AF_INET,
328 	    &V_ipsec_hhh_out[HHOOK_IPSEC_INET],
329 	    HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
330 		printf("%s: WARNING: unable to register output helper hook\n",
331 		    __func__);
332 
333 	/* Skip initialization of globals for non-default instances. */
334 	if (!IS_DEFAULT_VNET(curvnet))
335 		return;
336 
337 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
338 	if (pr == NULL)
339 		panic("ip_init: PF_INET not found");
340 
341 	/* Initialize the entire ip_protox[] array to IPPROTO_RAW. */
342 	for (i = 0; i < IPPROTO_MAX; i++)
343 		ip_protox[i] = pr - inetsw;
344 	/*
345 	 * Cycle through IP protocols and put them into the appropriate place
346 	 * in ip_protox[].
347 	 */
348 	for (pr = inetdomain.dom_protosw;
349 	    pr < inetdomain.dom_protoswNPROTOSW; pr++)
350 		if (pr->pr_domain->dom_family == PF_INET &&
351 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
352 			/* Be careful to only index valid IP protocols. */
353 			if (pr->pr_protocol < IPPROTO_MAX)
354 				ip_protox[pr->pr_protocol] = pr - inetsw;
355 		}
356 
357 	netisr_register(&ip_nh);
358 #ifdef	RSS
359 	netisr_register(&ip_direct_nh);
360 #endif
361 }
362 
363 #ifdef VIMAGE
364 void
365 ip_destroy(void)
366 {
367 	int error;
368 
369 	if ((error = pfil_head_unregister(&V_inet_pfil_hook)) != 0)
370 		printf("%s: WARNING: unable to unregister pfil hook, "
371 		    "error %d\n", __func__, error);
372 
373 	error = hhook_head_deregister(V_ipsec_hhh_in[HHOOK_IPSEC_INET]);
374 	if (error != 0) {
375 		printf("%s: WARNING: unable to deregister input helper hook "
376 		    "type HHOOK_TYPE_IPSEC_IN, id HHOOK_IPSEC_INET: "
377 		    "error %d returned\n", __func__, error);
378 	}
379 	error = hhook_head_deregister(V_ipsec_hhh_out[HHOOK_IPSEC_INET]);
380 	if (error != 0) {
381 		printf("%s: WARNING: unable to deregister output helper hook "
382 		    "type HHOOK_TYPE_IPSEC_OUT, id HHOOK_IPSEC_INET: "
383 		    "error %d returned\n", __func__, error);
384 	}
385 	/* Cleanup in_ifaddr hash table; should be empty. */
386 	hashdestroy(V_in_ifaddrhashtbl, M_IFADDR, V_in_ifaddrhmask);
387 
388 	/* Destroy IP reassembly queue. */
389 	ipreass_destroy();
390 }
391 #endif
392 
393 #ifdef	RSS
394 /*
395  * IP direct input routine.
396  *
397  * This is called when reinjecting completed fragments where
398  * all of the previous checking and book-keeping has been done.
399  */
400 void
401 ip_direct_input(struct mbuf *m)
402 {
403 	struct ip *ip;
404 	int hlen;
405 
406 	ip = mtod(m, struct ip *);
407 	hlen = ip->ip_hl << 2;
408 
409 	IPSTAT_INC(ips_delivered);
410 	(*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
411 	return;
412 }
413 #endif
414 
415 /*
416  * Ip input routine.  Checksum and byte swap header.  If fragmented
417  * try to reassemble.  Process options.  Pass to next level.
418  */
419 void
420 ip_input(struct mbuf *m)
421 {
422 	struct ip *ip = NULL;
423 	struct in_ifaddr *ia = NULL;
424 	struct ifaddr *ifa;
425 	struct ifnet *ifp;
426 	int    checkif, hlen = 0;
427 	uint16_t sum, ip_len;
428 	int dchg = 0;				/* dest changed after fw */
429 	struct in_addr odst;			/* original dst address */
430 
431 	M_ASSERTPKTHDR(m);
432 
433 	if (m->m_flags & M_FASTFWD_OURS) {
434 		m->m_flags &= ~M_FASTFWD_OURS;
435 		/* Set up some basics that will be used later. */
436 		ip = mtod(m, struct ip *);
437 		hlen = ip->ip_hl << 2;
438 		ip_len = ntohs(ip->ip_len);
439 		goto ours;
440 	}
441 
442 	IPSTAT_INC(ips_total);
443 
444 	if (m->m_pkthdr.len < sizeof(struct ip))
445 		goto tooshort;
446 
447 	if (m->m_len < sizeof (struct ip) &&
448 	    (m = m_pullup(m, sizeof (struct ip))) == NULL) {
449 		IPSTAT_INC(ips_toosmall);
450 		return;
451 	}
452 	ip = mtod(m, struct ip *);
453 
454 	if (ip->ip_v != IPVERSION) {
455 		IPSTAT_INC(ips_badvers);
456 		goto bad;
457 	}
458 
459 	hlen = ip->ip_hl << 2;
460 	if (hlen < sizeof(struct ip)) {	/* minimum header length */
461 		IPSTAT_INC(ips_badhlen);
462 		goto bad;
463 	}
464 	if (hlen > m->m_len) {
465 		if ((m = m_pullup(m, hlen)) == NULL) {
466 			IPSTAT_INC(ips_badhlen);
467 			return;
468 		}
469 		ip = mtod(m, struct ip *);
470 	}
471 
472 	IP_PROBE(receive, NULL, NULL, ip, m->m_pkthdr.rcvif, ip, NULL);
473 
474 	/* 127/8 must not appear on wire - RFC1122 */
475 	ifp = m->m_pkthdr.rcvif;
476 	if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
477 	    (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
478 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
479 			IPSTAT_INC(ips_badaddr);
480 			goto bad;
481 		}
482 	}
483 
484 	if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
485 		sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
486 	} else {
487 		if (hlen == sizeof(struct ip)) {
488 			sum = in_cksum_hdr(ip);
489 		} else {
490 			sum = in_cksum(m, hlen);
491 		}
492 	}
493 	if (sum) {
494 		IPSTAT_INC(ips_badsum);
495 		goto bad;
496 	}
497 
498 #ifdef ALTQ
499 	if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
500 		/* packet is dropped by traffic conditioner */
501 		return;
502 #endif
503 
504 	ip_len = ntohs(ip->ip_len);
505 	if (ip_len < hlen) {
506 		IPSTAT_INC(ips_badlen);
507 		goto bad;
508 	}
509 
510 	/*
511 	 * Check that the amount of data in the buffers
512 	 * is as at least much as the IP header would have us expect.
513 	 * Trim mbufs if longer than we expect.
514 	 * Drop packet if shorter than we expect.
515 	 */
516 	if (m->m_pkthdr.len < ip_len) {
517 tooshort:
518 		IPSTAT_INC(ips_tooshort);
519 		goto bad;
520 	}
521 	if (m->m_pkthdr.len > ip_len) {
522 		if (m->m_len == m->m_pkthdr.len) {
523 			m->m_len = ip_len;
524 			m->m_pkthdr.len = ip_len;
525 		} else
526 			m_adj(m, ip_len - m->m_pkthdr.len);
527 	}
528 
529 	/* Try to forward the packet, but if we fail continue */
530 #ifdef IPSEC
531 	/* For now we do not handle IPSEC in tryforward. */
532 	if (!key_havesp(IPSEC_DIR_INBOUND) && !key_havesp(IPSEC_DIR_OUTBOUND) &&
533 	    (V_ipforwarding == 1))
534 		if (ip_tryforward(m) == NULL)
535 			return;
536 	/*
537 	 * Bypass packet filtering for packets previously handled by IPsec.
538 	 */
539 	if (ip_ipsec_filtertunnel(m))
540 		goto passin;
541 #else
542 	if (V_ipforwarding == 1)
543 		if (ip_tryforward(m) == NULL)
544 			return;
545 #endif /* IPSEC */
546 
547 	/*
548 	 * Run through list of hooks for input packets.
549 	 *
550 	 * NB: Beware of the destination address changing (e.g.
551 	 *     by NAT rewriting).  When this happens, tell
552 	 *     ip_forward to do the right thing.
553 	 */
554 
555 	/* Jump over all PFIL processing if hooks are not active. */
556 	if (!PFIL_HOOKED(&V_inet_pfil_hook))
557 		goto passin;
558 
559 	odst = ip->ip_dst;
560 	if (pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_IN, NULL) != 0)
561 		return;
562 	if (m == NULL)			/* consumed by filter */
563 		return;
564 
565 	ip = mtod(m, struct ip *);
566 	dchg = (odst.s_addr != ip->ip_dst.s_addr);
567 	ifp = m->m_pkthdr.rcvif;
568 
569 	if (m->m_flags & M_FASTFWD_OURS) {
570 		m->m_flags &= ~M_FASTFWD_OURS;
571 		goto ours;
572 	}
573 	if (m->m_flags & M_IP_NEXTHOP) {
574 		dchg = (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL);
575 		if (dchg != 0) {
576 			/*
577 			 * Directly ship the packet on.  This allows
578 			 * forwarding packets originally destined to us
579 			 * to some other directly connected host.
580 			 */
581 			ip_forward(m, 1);
582 			return;
583 		}
584 	}
585 passin:
586 
587 	/*
588 	 * Process options and, if not destined for us,
589 	 * ship it on.  ip_dooptions returns 1 when an
590 	 * error was detected (causing an icmp message
591 	 * to be sent and the original packet to be freed).
592 	 */
593 	if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
594 		return;
595 
596         /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
597          * matter if it is destined to another node, or whether it is
598          * a multicast one, RSVP wants it! and prevents it from being forwarded
599          * anywhere else. Also checks if the rsvp daemon is running before
600 	 * grabbing the packet.
601          */
602 	if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP)
603 		goto ours;
604 
605 	/*
606 	 * Check our list of addresses, to see if the packet is for us.
607 	 * If we don't have any addresses, assume any unicast packet
608 	 * we receive might be for us (and let the upper layers deal
609 	 * with it).
610 	 */
611 	if (TAILQ_EMPTY(&V_in_ifaddrhead) &&
612 	    (m->m_flags & (M_MCAST|M_BCAST)) == 0)
613 		goto ours;
614 
615 	/*
616 	 * Enable a consistency check between the destination address
617 	 * and the arrival interface for a unicast packet (the RFC 1122
618 	 * strong ES model) if IP forwarding is disabled and the packet
619 	 * is not locally generated and the packet is not subject to
620 	 * 'ipfw fwd'.
621 	 *
622 	 * XXX - Checking also should be disabled if the destination
623 	 * address is ipnat'ed to a different interface.
624 	 *
625 	 * XXX - Checking is incompatible with IP aliases added
626 	 * to the loopback interface instead of the interface where
627 	 * the packets are received.
628 	 *
629 	 * XXX - This is the case for carp vhost IPs as well so we
630 	 * insert a workaround. If the packet got here, we already
631 	 * checked with carp_iamatch() and carp_forus().
632 	 */
633 	checkif = V_ip_checkinterface && (V_ipforwarding == 0) &&
634 	    ifp != NULL && ((ifp->if_flags & IFF_LOOPBACK) == 0) &&
635 	    ifp->if_carp == NULL && (dchg == 0);
636 
637 	/*
638 	 * Check for exact addresses in the hash bucket.
639 	 */
640 	/* IN_IFADDR_RLOCK(); */
641 	LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
642 		/*
643 		 * If the address matches, verify that the packet
644 		 * arrived via the correct interface if checking is
645 		 * enabled.
646 		 */
647 		if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr &&
648 		    (!checkif || ia->ia_ifp == ifp)) {
649 			counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
650 			counter_u64_add(ia->ia_ifa.ifa_ibytes,
651 			    m->m_pkthdr.len);
652 			/* IN_IFADDR_RUNLOCK(); */
653 			goto ours;
654 		}
655 	}
656 	/* IN_IFADDR_RUNLOCK(); */
657 
658 	/*
659 	 * Check for broadcast addresses.
660 	 *
661 	 * Only accept broadcast packets that arrive via the matching
662 	 * interface.  Reception of forwarded directed broadcasts would
663 	 * be handled via ip_forward() and ether_output() with the loopback
664 	 * into the stack for SIMPLEX interfaces handled by ether_output().
665 	 */
666 	if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
667 		IF_ADDR_RLOCK(ifp);
668 	        TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
669 			if (ifa->ifa_addr->sa_family != AF_INET)
670 				continue;
671 			ia = ifatoia(ifa);
672 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
673 			    ip->ip_dst.s_addr) {
674 				counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
675 				counter_u64_add(ia->ia_ifa.ifa_ibytes,
676 				    m->m_pkthdr.len);
677 				IF_ADDR_RUNLOCK(ifp);
678 				goto ours;
679 			}
680 #ifdef BOOTP_COMPAT
681 			if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) {
682 				counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
683 				counter_u64_add(ia->ia_ifa.ifa_ibytes,
684 				    m->m_pkthdr.len);
685 				IF_ADDR_RUNLOCK(ifp);
686 				goto ours;
687 			}
688 #endif
689 		}
690 		IF_ADDR_RUNLOCK(ifp);
691 		ia = NULL;
692 	}
693 	/* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */
694 	if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) {
695 		IPSTAT_INC(ips_cantforward);
696 		m_freem(m);
697 		return;
698 	}
699 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
700 		if (V_ip_mrouter) {
701 			/*
702 			 * If we are acting as a multicast router, all
703 			 * incoming multicast packets are passed to the
704 			 * kernel-level multicast forwarding function.
705 			 * The packet is returned (relatively) intact; if
706 			 * ip_mforward() returns a non-zero value, the packet
707 			 * must be discarded, else it may be accepted below.
708 			 */
709 			if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) {
710 				IPSTAT_INC(ips_cantforward);
711 				m_freem(m);
712 				return;
713 			}
714 
715 			/*
716 			 * The process-level routing daemon needs to receive
717 			 * all multicast IGMP packets, whether or not this
718 			 * host belongs to their destination groups.
719 			 */
720 			if (ip->ip_p == IPPROTO_IGMP)
721 				goto ours;
722 			IPSTAT_INC(ips_forward);
723 		}
724 		/*
725 		 * Assume the packet is for us, to avoid prematurely taking
726 		 * a lock on the in_multi hash. Protocols must perform
727 		 * their own filtering and update statistics accordingly.
728 		 */
729 		goto ours;
730 	}
731 	if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
732 		goto ours;
733 	if (ip->ip_dst.s_addr == INADDR_ANY)
734 		goto ours;
735 
736 	/*
737 	 * Not for us; forward if possible and desirable.
738 	 */
739 	if (V_ipforwarding == 0) {
740 		IPSTAT_INC(ips_cantforward);
741 		m_freem(m);
742 	} else {
743 		ip_forward(m, dchg);
744 	}
745 	return;
746 
747 ours:
748 #ifdef IPSTEALTH
749 	/*
750 	 * IPSTEALTH: Process non-routing options only
751 	 * if the packet is destined for us.
752 	 */
753 	if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1))
754 		return;
755 #endif /* IPSTEALTH */
756 
757 	/*
758 	 * Attempt reassembly; if it succeeds, proceed.
759 	 * ip_reass() will return a different mbuf.
760 	 */
761 	if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) {
762 		/* XXXGL: shouldn't we save & set m_flags? */
763 		m = ip_reass(m);
764 		if (m == NULL)
765 			return;
766 		ip = mtod(m, struct ip *);
767 		/* Get the header length of the reassembled packet */
768 		hlen = ip->ip_hl << 2;
769 	}
770 
771 #ifdef IPSEC
772 	/*
773 	 * enforce IPsec policy checking if we are seeing last header.
774 	 * note that we do not visit this with protocols with pcb layer
775 	 * code - like udp/tcp/raw ip.
776 	 */
777 	if (ip_ipsec_input(m, ip->ip_p) != 0)
778 		goto bad;
779 #endif /* IPSEC */
780 
781 	/*
782 	 * Switch out to protocol's input routine.
783 	 */
784 	IPSTAT_INC(ips_delivered);
785 
786 	(*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
787 	return;
788 bad:
789 	m_freem(m);
790 }
791 
792 /*
793  * IP timer processing;
794  * if a timer expires on a reassembly
795  * queue, discard it.
796  */
797 void
798 ip_slowtimo(void)
799 {
800 	VNET_ITERATOR_DECL(vnet_iter);
801 
802 	VNET_LIST_RLOCK_NOSLEEP();
803 	VNET_FOREACH(vnet_iter) {
804 		CURVNET_SET(vnet_iter);
805 		ipreass_slowtimo();
806 		CURVNET_RESTORE();
807 	}
808 	VNET_LIST_RUNLOCK_NOSLEEP();
809 }
810 
811 void
812 ip_drain(void)
813 {
814 	VNET_ITERATOR_DECL(vnet_iter);
815 
816 	VNET_LIST_RLOCK_NOSLEEP();
817 	VNET_FOREACH(vnet_iter) {
818 		CURVNET_SET(vnet_iter);
819 		ipreass_drain();
820 		CURVNET_RESTORE();
821 	}
822 	VNET_LIST_RUNLOCK_NOSLEEP();
823 }
824 
825 /*
826  * The protocol to be inserted into ip_protox[] must be already registered
827  * in inetsw[], either statically or through pf_proto_register().
828  */
829 int
830 ipproto_register(short ipproto)
831 {
832 	struct protosw *pr;
833 
834 	/* Sanity checks. */
835 	if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
836 		return (EPROTONOSUPPORT);
837 
838 	/*
839 	 * The protocol slot must not be occupied by another protocol
840 	 * already.  An index pointing to IPPROTO_RAW is unused.
841 	 */
842 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
843 	if (pr == NULL)
844 		return (EPFNOSUPPORT);
845 	if (ip_protox[ipproto] != pr - inetsw)	/* IPPROTO_RAW */
846 		return (EEXIST);
847 
848 	/* Find the protocol position in inetsw[] and set the index. */
849 	for (pr = inetdomain.dom_protosw;
850 	     pr < inetdomain.dom_protoswNPROTOSW; pr++) {
851 		if (pr->pr_domain->dom_family == PF_INET &&
852 		    pr->pr_protocol && pr->pr_protocol == ipproto) {
853 			ip_protox[pr->pr_protocol] = pr - inetsw;
854 			return (0);
855 		}
856 	}
857 	return (EPROTONOSUPPORT);
858 }
859 
860 int
861 ipproto_unregister(short ipproto)
862 {
863 	struct protosw *pr;
864 
865 	/* Sanity checks. */
866 	if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
867 		return (EPROTONOSUPPORT);
868 
869 	/* Check if the protocol was indeed registered. */
870 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
871 	if (pr == NULL)
872 		return (EPFNOSUPPORT);
873 	if (ip_protox[ipproto] == pr - inetsw)  /* IPPROTO_RAW */
874 		return (ENOENT);
875 
876 	/* Reset the protocol slot to IPPROTO_RAW. */
877 	ip_protox[ipproto] = pr - inetsw;
878 	return (0);
879 }
880 
881 /*
882  * Given address of next destination (final or next hop), return (referenced)
883  * internet address info of interface to be used to get there.
884  */
885 struct in_ifaddr *
886 ip_rtaddr(struct in_addr dst, u_int fibnum)
887 {
888 	struct route sro;
889 	struct sockaddr_in *sin;
890 	struct in_ifaddr *ia;
891 
892 	bzero(&sro, sizeof(sro));
893 	sin = (struct sockaddr_in *)&sro.ro_dst;
894 	sin->sin_family = AF_INET;
895 	sin->sin_len = sizeof(*sin);
896 	sin->sin_addr = dst;
897 	in_rtalloc_ign(&sro, 0, fibnum);
898 
899 	if (sro.ro_rt == NULL)
900 		return (NULL);
901 
902 	ia = ifatoia(sro.ro_rt->rt_ifa);
903 	ifa_ref(&ia->ia_ifa);
904 	RTFREE(sro.ro_rt);
905 	return (ia);
906 }
907 
908 u_char inetctlerrmap[PRC_NCMDS] = {
909 	0,		0,		0,		0,
910 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
911 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
912 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
913 	0,		0,		EHOSTUNREACH,	0,
914 	ENOPROTOOPT,	ECONNREFUSED
915 };
916 
917 /*
918  * Forward a packet.  If some error occurs return the sender
919  * an icmp packet.  Note we can't always generate a meaningful
920  * icmp message because icmp doesn't have a large enough repertoire
921  * of codes and types.
922  *
923  * If not forwarding, just drop the packet.  This could be confusing
924  * if ipforwarding was zero but some routing protocol was advancing
925  * us as a gateway to somewhere.  However, we must let the routing
926  * protocol deal with that.
927  *
928  * The srcrt parameter indicates whether the packet is being forwarded
929  * via a source route.
930  */
931 void
932 ip_forward(struct mbuf *m, int srcrt)
933 {
934 	struct ip *ip = mtod(m, struct ip *);
935 	struct in_ifaddr *ia;
936 	struct mbuf *mcopy;
937 	struct sockaddr_in *sin;
938 	struct in_addr dest;
939 	struct route ro;
940 	int error, type = 0, code = 0, mtu = 0;
941 
942 	if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
943 		IPSTAT_INC(ips_cantforward);
944 		m_freem(m);
945 		return;
946 	}
947 #ifdef IPSEC
948 	if (ip_ipsec_fwd(m) != 0) {
949 		IPSTAT_INC(ips_cantforward);
950 		m_freem(m);
951 		return;
952 	}
953 #endif /* IPSEC */
954 #ifdef IPSTEALTH
955 	if (!V_ipstealth) {
956 #endif
957 		if (ip->ip_ttl <= IPTTLDEC) {
958 			icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS,
959 			    0, 0);
960 			return;
961 		}
962 #ifdef IPSTEALTH
963 	}
964 #endif
965 
966 	bzero(&ro, sizeof(ro));
967 	sin = (struct sockaddr_in *)&ro.ro_dst;
968 	sin->sin_family = AF_INET;
969 	sin->sin_len = sizeof(*sin);
970 	sin->sin_addr = ip->ip_dst;
971 #ifdef RADIX_MPATH
972 	rtalloc_mpath_fib(&ro,
973 	    ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
974 	    M_GETFIB(m));
975 #else
976 	in_rtalloc_ign(&ro, 0, M_GETFIB(m));
977 #endif
978 	if (ro.ro_rt != NULL) {
979 		ia = ifatoia(ro.ro_rt->rt_ifa);
980 		ifa_ref(&ia->ia_ifa);
981 	} else
982 		ia = NULL;
983 #ifndef IPSEC
984 	/*
985 	 * 'ia' may be NULL if there is no route for this destination.
986 	 * In case of IPsec, Don't discard it just yet, but pass it to
987 	 * ip_output in case of outgoing IPsec policy.
988 	 */
989 	if (!srcrt && ia == NULL) {
990 		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
991 		RO_RTFREE(&ro);
992 		return;
993 	}
994 #endif
995 
996 	/*
997 	 * Save the IP header and at most 8 bytes of the payload,
998 	 * in case we need to generate an ICMP message to the src.
999 	 *
1000 	 * XXX this can be optimized a lot by saving the data in a local
1001 	 * buffer on the stack (72 bytes at most), and only allocating the
1002 	 * mbuf if really necessary. The vast majority of the packets
1003 	 * are forwarded without having to send an ICMP back (either
1004 	 * because unnecessary, or because rate limited), so we are
1005 	 * really we are wasting a lot of work here.
1006 	 *
1007 	 * We don't use m_copy() because it might return a reference
1008 	 * to a shared cluster. Both this function and ip_output()
1009 	 * assume exclusive access to the IP header in `m', so any
1010 	 * data in a cluster may change before we reach icmp_error().
1011 	 */
1012 	mcopy = m_gethdr(M_NOWAIT, m->m_type);
1013 	if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_NOWAIT)) {
1014 		/*
1015 		 * It's probably ok if the pkthdr dup fails (because
1016 		 * the deep copy of the tag chain failed), but for now
1017 		 * be conservative and just discard the copy since
1018 		 * code below may some day want the tags.
1019 		 */
1020 		m_free(mcopy);
1021 		mcopy = NULL;
1022 	}
1023 	if (mcopy != NULL) {
1024 		mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy));
1025 		mcopy->m_pkthdr.len = mcopy->m_len;
1026 		m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1027 	}
1028 
1029 #ifdef IPSTEALTH
1030 	if (!V_ipstealth) {
1031 #endif
1032 		ip->ip_ttl -= IPTTLDEC;
1033 #ifdef IPSTEALTH
1034 	}
1035 #endif
1036 
1037 	/*
1038 	 * If forwarding packet using same interface that it came in on,
1039 	 * perhaps should send a redirect to sender to shortcut a hop.
1040 	 * Only send redirect if source is sending directly to us,
1041 	 * and if packet was not source routed (or has any options).
1042 	 * Also, don't send redirect if forwarding using a default route
1043 	 * or a route modified by a redirect.
1044 	 */
1045 	dest.s_addr = 0;
1046 	if (!srcrt && V_ipsendredirects &&
1047 	    ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) {
1048 		struct rtentry *rt;
1049 
1050 		rt = ro.ro_rt;
1051 
1052 		if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1053 		    satosin(rt_key(rt))->sin_addr.s_addr != 0) {
1054 #define	RTA(rt)	((struct in_ifaddr *)(rt->rt_ifa))
1055 			u_long src = ntohl(ip->ip_src.s_addr);
1056 
1057 			if (RTA(rt) &&
1058 			    (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1059 				if (rt->rt_flags & RTF_GATEWAY)
1060 					dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr;
1061 				else
1062 					dest.s_addr = ip->ip_dst.s_addr;
1063 				/* Router requirements says to only send host redirects */
1064 				type = ICMP_REDIRECT;
1065 				code = ICMP_REDIRECT_HOST;
1066 			}
1067 		}
1068 	}
1069 
1070 	error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
1071 
1072 	if (error == EMSGSIZE && ro.ro_rt)
1073 		mtu = ro.ro_rt->rt_mtu;
1074 	RO_RTFREE(&ro);
1075 
1076 	if (error)
1077 		IPSTAT_INC(ips_cantforward);
1078 	else {
1079 		IPSTAT_INC(ips_forward);
1080 		if (type)
1081 			IPSTAT_INC(ips_redirectsent);
1082 		else {
1083 			if (mcopy)
1084 				m_freem(mcopy);
1085 			if (ia != NULL)
1086 				ifa_free(&ia->ia_ifa);
1087 			return;
1088 		}
1089 	}
1090 	if (mcopy == NULL) {
1091 		if (ia != NULL)
1092 			ifa_free(&ia->ia_ifa);
1093 		return;
1094 	}
1095 
1096 	switch (error) {
1097 
1098 	case 0:				/* forwarded, but need redirect */
1099 		/* type, code set above */
1100 		break;
1101 
1102 	case ENETUNREACH:
1103 	case EHOSTUNREACH:
1104 	case ENETDOWN:
1105 	case EHOSTDOWN:
1106 	default:
1107 		type = ICMP_UNREACH;
1108 		code = ICMP_UNREACH_HOST;
1109 		break;
1110 
1111 	case EMSGSIZE:
1112 		type = ICMP_UNREACH;
1113 		code = ICMP_UNREACH_NEEDFRAG;
1114 
1115 #ifdef IPSEC
1116 		/*
1117 		 * If IPsec is configured for this path,
1118 		 * override any possibly mtu value set by ip_output.
1119 		 */
1120 		mtu = ip_ipsec_mtu(mcopy, mtu);
1121 #endif /* IPSEC */
1122 		/*
1123 		 * If the MTU was set before make sure we are below the
1124 		 * interface MTU.
1125 		 * If the MTU wasn't set before use the interface mtu or
1126 		 * fall back to the next smaller mtu step compared to the
1127 		 * current packet size.
1128 		 */
1129 		if (mtu != 0) {
1130 			if (ia != NULL)
1131 				mtu = min(mtu, ia->ia_ifp->if_mtu);
1132 		} else {
1133 			if (ia != NULL)
1134 				mtu = ia->ia_ifp->if_mtu;
1135 			else
1136 				mtu = ip_next_mtu(ntohs(ip->ip_len), 0);
1137 		}
1138 		IPSTAT_INC(ips_cantfrag);
1139 		break;
1140 
1141 	case ENOBUFS:
1142 	case EACCES:			/* ipfw denied packet */
1143 		m_freem(mcopy);
1144 		if (ia != NULL)
1145 			ifa_free(&ia->ia_ifa);
1146 		return;
1147 	}
1148 	if (ia != NULL)
1149 		ifa_free(&ia->ia_ifa);
1150 	icmp_error(mcopy, type, code, dest.s_addr, mtu);
1151 }
1152 
1153 void
1154 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1155     struct mbuf *m)
1156 {
1157 
1158 	if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) {
1159 		struct bintime bt;
1160 
1161 		bintime(&bt);
1162 		if (inp->inp_socket->so_options & SO_BINTIME) {
1163 			*mp = sbcreatecontrol((caddr_t)&bt, sizeof(bt),
1164 			    SCM_BINTIME, SOL_SOCKET);
1165 			if (*mp)
1166 				mp = &(*mp)->m_next;
1167 		}
1168 		if (inp->inp_socket->so_options & SO_TIMESTAMP) {
1169 			struct timeval tv;
1170 
1171 			bintime2timeval(&bt, &tv);
1172 			*mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv),
1173 			    SCM_TIMESTAMP, SOL_SOCKET);
1174 			if (*mp)
1175 				mp = &(*mp)->m_next;
1176 		}
1177 	}
1178 	if (inp->inp_flags & INP_RECVDSTADDR) {
1179 		*mp = sbcreatecontrol((caddr_t)&ip->ip_dst,
1180 		    sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1181 		if (*mp)
1182 			mp = &(*mp)->m_next;
1183 	}
1184 	if (inp->inp_flags & INP_RECVTTL) {
1185 		*mp = sbcreatecontrol((caddr_t)&ip->ip_ttl,
1186 		    sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
1187 		if (*mp)
1188 			mp = &(*mp)->m_next;
1189 	}
1190 #ifdef notyet
1191 	/* XXX
1192 	 * Moving these out of udp_input() made them even more broken
1193 	 * than they already were.
1194 	 */
1195 	/* options were tossed already */
1196 	if (inp->inp_flags & INP_RECVOPTS) {
1197 		*mp = sbcreatecontrol((caddr_t)opts_deleted_above,
1198 		    sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1199 		if (*mp)
1200 			mp = &(*mp)->m_next;
1201 	}
1202 	/* ip_srcroute doesn't do what we want here, need to fix */
1203 	if (inp->inp_flags & INP_RECVRETOPTS) {
1204 		*mp = sbcreatecontrol((caddr_t)ip_srcroute(m),
1205 		    sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1206 		if (*mp)
1207 			mp = &(*mp)->m_next;
1208 	}
1209 #endif
1210 	if (inp->inp_flags & INP_RECVIF) {
1211 		struct ifnet *ifp;
1212 		struct sdlbuf {
1213 			struct sockaddr_dl sdl;
1214 			u_char	pad[32];
1215 		} sdlbuf;
1216 		struct sockaddr_dl *sdp;
1217 		struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1218 
1219 		if ((ifp = m->m_pkthdr.rcvif) &&
1220 		    ifp->if_index && ifp->if_index <= V_if_index) {
1221 			sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1222 			/*
1223 			 * Change our mind and don't try copy.
1224 			 */
1225 			if (sdp->sdl_family != AF_LINK ||
1226 			    sdp->sdl_len > sizeof(sdlbuf)) {
1227 				goto makedummy;
1228 			}
1229 			bcopy(sdp, sdl2, sdp->sdl_len);
1230 		} else {
1231 makedummy:
1232 			sdl2->sdl_len =
1233 			    offsetof(struct sockaddr_dl, sdl_data[0]);
1234 			sdl2->sdl_family = AF_LINK;
1235 			sdl2->sdl_index = 0;
1236 			sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1237 		}
1238 		*mp = sbcreatecontrol((caddr_t)sdl2, sdl2->sdl_len,
1239 		    IP_RECVIF, IPPROTO_IP);
1240 		if (*mp)
1241 			mp = &(*mp)->m_next;
1242 	}
1243 	if (inp->inp_flags & INP_RECVTOS) {
1244 		*mp = sbcreatecontrol((caddr_t)&ip->ip_tos,
1245 		    sizeof(u_char), IP_RECVTOS, IPPROTO_IP);
1246 		if (*mp)
1247 			mp = &(*mp)->m_next;
1248 	}
1249 
1250 	if (inp->inp_flags2 & INP_RECVFLOWID) {
1251 		uint32_t flowid, flow_type;
1252 
1253 		flowid = m->m_pkthdr.flowid;
1254 		flow_type = M_HASHTYPE_GET(m);
1255 
1256 		/*
1257 		 * XXX should handle the failure of one or the
1258 		 * other - don't populate both?
1259 		 */
1260 		*mp = sbcreatecontrol((caddr_t) &flowid,
1261 		    sizeof(uint32_t), IP_FLOWID, IPPROTO_IP);
1262 		if (*mp)
1263 			mp = &(*mp)->m_next;
1264 		*mp = sbcreatecontrol((caddr_t) &flow_type,
1265 		    sizeof(uint32_t), IP_FLOWTYPE, IPPROTO_IP);
1266 		if (*mp)
1267 			mp = &(*mp)->m_next;
1268 	}
1269 
1270 #ifdef	RSS
1271 	if (inp->inp_flags2 & INP_RECVRSSBUCKETID) {
1272 		uint32_t flowid, flow_type;
1273 		uint32_t rss_bucketid;
1274 
1275 		flowid = m->m_pkthdr.flowid;
1276 		flow_type = M_HASHTYPE_GET(m);
1277 
1278 		if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) {
1279 			*mp = sbcreatecontrol((caddr_t) &rss_bucketid,
1280 			   sizeof(uint32_t), IP_RSSBUCKETID, IPPROTO_IP);
1281 			if (*mp)
1282 				mp = &(*mp)->m_next;
1283 		}
1284 	}
1285 #endif
1286 }
1287 
1288 /*
1289  * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
1290  * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
1291  * locking.  This code remains in ip_input.c as ip_mroute.c is optionally
1292  * compiled.
1293  */
1294 static VNET_DEFINE(int, ip_rsvp_on);
1295 VNET_DEFINE(struct socket *, ip_rsvpd);
1296 
1297 #define	V_ip_rsvp_on		VNET(ip_rsvp_on)
1298 
1299 int
1300 ip_rsvp_init(struct socket *so)
1301 {
1302 
1303 	if (so->so_type != SOCK_RAW ||
1304 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
1305 		return EOPNOTSUPP;
1306 
1307 	if (V_ip_rsvpd != NULL)
1308 		return EADDRINUSE;
1309 
1310 	V_ip_rsvpd = so;
1311 	/*
1312 	 * This may seem silly, but we need to be sure we don't over-increment
1313 	 * the RSVP counter, in case something slips up.
1314 	 */
1315 	if (!V_ip_rsvp_on) {
1316 		V_ip_rsvp_on = 1;
1317 		V_rsvp_on++;
1318 	}
1319 
1320 	return 0;
1321 }
1322 
1323 int
1324 ip_rsvp_done(void)
1325 {
1326 
1327 	V_ip_rsvpd = NULL;
1328 	/*
1329 	 * This may seem silly, but we need to be sure we don't over-decrement
1330 	 * the RSVP counter, in case something slips up.
1331 	 */
1332 	if (V_ip_rsvp_on) {
1333 		V_ip_rsvp_on = 0;
1334 		V_rsvp_on--;
1335 	}
1336 	return 0;
1337 }
1338 
1339 int
1340 rsvp_input(struct mbuf **mp, int *offp, int proto)
1341 {
1342 	struct mbuf *m;
1343 
1344 	m = *mp;
1345 	*mp = NULL;
1346 
1347 	if (rsvp_input_p) { /* call the real one if loaded */
1348 		*mp = m;
1349 		rsvp_input_p(mp, offp, proto);
1350 		return (IPPROTO_DONE);
1351 	}
1352 
1353 	/* Can still get packets with rsvp_on = 0 if there is a local member
1354 	 * of the group to which the RSVP packet is addressed.  But in this
1355 	 * case we want to throw the packet away.
1356 	 */
1357 
1358 	if (!V_rsvp_on) {
1359 		m_freem(m);
1360 		return (IPPROTO_DONE);
1361 	}
1362 
1363 	if (V_ip_rsvpd != NULL) {
1364 		*mp = m;
1365 		rip_input(mp, offp, proto);
1366 		return (IPPROTO_DONE);
1367 	}
1368 	/* Drop the packet */
1369 	m_freem(m);
1370 	return (IPPROTO_DONE);
1371 }
1372