xref: /freebsd/sys/netinet/ip_input.c (revision 4af86fb4f9a04f31e557deab6436606d3ffc6f55)
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 #ifdef VIMAGE
335 	if (!IS_DEFAULT_VNET(curvnet)) {
336 		netisr_register_vnet(&ip_nh);
337 #ifdef	RSS
338 		netisr_register_vnet(&ip_direct_nh);
339 #endif
340 		return;
341 	}
342 #endif
343 
344 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
345 	if (pr == NULL)
346 		panic("ip_init: PF_INET not found");
347 
348 	/* Initialize the entire ip_protox[] array to IPPROTO_RAW. */
349 	for (i = 0; i < IPPROTO_MAX; i++)
350 		ip_protox[i] = pr - inetsw;
351 	/*
352 	 * Cycle through IP protocols and put them into the appropriate place
353 	 * in ip_protox[].
354 	 */
355 	for (pr = inetdomain.dom_protosw;
356 	    pr < inetdomain.dom_protoswNPROTOSW; pr++)
357 		if (pr->pr_domain->dom_family == PF_INET &&
358 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
359 			/* Be careful to only index valid IP protocols. */
360 			if (pr->pr_protocol < IPPROTO_MAX)
361 				ip_protox[pr->pr_protocol] = pr - inetsw;
362 		}
363 
364 	netisr_register(&ip_nh);
365 #ifdef	RSS
366 	netisr_register(&ip_direct_nh);
367 #endif
368 }
369 
370 #ifdef VIMAGE
371 static void
372 ip_destroy(void *unused __unused)
373 {
374 	int error;
375 
376 #ifdef	RSS
377 	netisr_unregister_vnet(&ip_direct_nh);
378 #endif
379 	netisr_unregister_vnet(&ip_nh);
380 
381 	if ((error = pfil_head_unregister(&V_inet_pfil_hook)) != 0)
382 		printf("%s: WARNING: unable to unregister pfil hook, "
383 		    "error %d\n", __func__, error);
384 
385 	error = hhook_head_deregister(V_ipsec_hhh_in[HHOOK_IPSEC_INET]);
386 	if (error != 0) {
387 		printf("%s: WARNING: unable to deregister input helper hook "
388 		    "type HHOOK_TYPE_IPSEC_IN, id HHOOK_IPSEC_INET: "
389 		    "error %d returned\n", __func__, error);
390 	}
391 	error = hhook_head_deregister(V_ipsec_hhh_out[HHOOK_IPSEC_INET]);
392 	if (error != 0) {
393 		printf("%s: WARNING: unable to deregister output helper hook "
394 		    "type HHOOK_TYPE_IPSEC_OUT, id HHOOK_IPSEC_INET: "
395 		    "error %d returned\n", __func__, error);
396 	}
397 	/* Cleanup in_ifaddr hash table; should be empty. */
398 	hashdestroy(V_in_ifaddrhashtbl, M_IFADDR, V_in_ifaddrhmask);
399 
400 	/* Destroy IP reassembly queue. */
401 	ipreass_destroy();
402 }
403 
404 VNET_SYSUNINIT(ip, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip_destroy, NULL);
405 #endif
406 
407 #ifdef	RSS
408 /*
409  * IP direct input routine.
410  *
411  * This is called when reinjecting completed fragments where
412  * all of the previous checking and book-keeping has been done.
413  */
414 void
415 ip_direct_input(struct mbuf *m)
416 {
417 	struct ip *ip;
418 	int hlen;
419 
420 	ip = mtod(m, struct ip *);
421 	hlen = ip->ip_hl << 2;
422 
423 	IPSTAT_INC(ips_delivered);
424 	(*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
425 	return;
426 }
427 #endif
428 
429 /*
430  * Ip input routine.  Checksum and byte swap header.  If fragmented
431  * try to reassemble.  Process options.  Pass to next level.
432  */
433 void
434 ip_input(struct mbuf *m)
435 {
436 	struct ip *ip = NULL;
437 	struct in_ifaddr *ia = NULL;
438 	struct ifaddr *ifa;
439 	struct ifnet *ifp;
440 	int    checkif, hlen = 0;
441 	uint16_t sum, ip_len;
442 	int dchg = 0;				/* dest changed after fw */
443 	struct in_addr odst;			/* original dst address */
444 
445 	M_ASSERTPKTHDR(m);
446 
447 	if (m->m_flags & M_FASTFWD_OURS) {
448 		m->m_flags &= ~M_FASTFWD_OURS;
449 		/* Set up some basics that will be used later. */
450 		ip = mtod(m, struct ip *);
451 		hlen = ip->ip_hl << 2;
452 		ip_len = ntohs(ip->ip_len);
453 		goto ours;
454 	}
455 
456 	IPSTAT_INC(ips_total);
457 
458 	if (m->m_pkthdr.len < sizeof(struct ip))
459 		goto tooshort;
460 
461 	if (m->m_len < sizeof (struct ip) &&
462 	    (m = m_pullup(m, sizeof (struct ip))) == NULL) {
463 		IPSTAT_INC(ips_toosmall);
464 		return;
465 	}
466 	ip = mtod(m, struct ip *);
467 
468 	if (ip->ip_v != IPVERSION) {
469 		IPSTAT_INC(ips_badvers);
470 		goto bad;
471 	}
472 
473 	hlen = ip->ip_hl << 2;
474 	if (hlen < sizeof(struct ip)) {	/* minimum header length */
475 		IPSTAT_INC(ips_badhlen);
476 		goto bad;
477 	}
478 	if (hlen > m->m_len) {
479 		if ((m = m_pullup(m, hlen)) == NULL) {
480 			IPSTAT_INC(ips_badhlen);
481 			return;
482 		}
483 		ip = mtod(m, struct ip *);
484 	}
485 
486 	IP_PROBE(receive, NULL, NULL, ip, m->m_pkthdr.rcvif, ip, NULL);
487 
488 	/* 127/8 must not appear on wire - RFC1122 */
489 	ifp = m->m_pkthdr.rcvif;
490 	if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
491 	    (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
492 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
493 			IPSTAT_INC(ips_badaddr);
494 			goto bad;
495 		}
496 	}
497 
498 	if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
499 		sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
500 	} else {
501 		if (hlen == sizeof(struct ip)) {
502 			sum = in_cksum_hdr(ip);
503 		} else {
504 			sum = in_cksum(m, hlen);
505 		}
506 	}
507 	if (sum) {
508 		IPSTAT_INC(ips_badsum);
509 		goto bad;
510 	}
511 
512 #ifdef ALTQ
513 	if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
514 		/* packet is dropped by traffic conditioner */
515 		return;
516 #endif
517 
518 	ip_len = ntohs(ip->ip_len);
519 	if (ip_len < hlen) {
520 		IPSTAT_INC(ips_badlen);
521 		goto bad;
522 	}
523 
524 	/*
525 	 * Check that the amount of data in the buffers
526 	 * is as at least much as the IP header would have us expect.
527 	 * Trim mbufs if longer than we expect.
528 	 * Drop packet if shorter than we expect.
529 	 */
530 	if (m->m_pkthdr.len < ip_len) {
531 tooshort:
532 		IPSTAT_INC(ips_tooshort);
533 		goto bad;
534 	}
535 	if (m->m_pkthdr.len > ip_len) {
536 		if (m->m_len == m->m_pkthdr.len) {
537 			m->m_len = ip_len;
538 			m->m_pkthdr.len = ip_len;
539 		} else
540 			m_adj(m, ip_len - m->m_pkthdr.len);
541 	}
542 
543 	/* Try to forward the packet, but if we fail continue */
544 #ifdef IPSEC
545 	/* For now we do not handle IPSEC in tryforward. */
546 	if (!key_havesp(IPSEC_DIR_INBOUND) && !key_havesp(IPSEC_DIR_OUTBOUND) &&
547 	    (V_ipforwarding == 1))
548 		if (ip_tryforward(m) == NULL)
549 			return;
550 	/*
551 	 * Bypass packet filtering for packets previously handled by IPsec.
552 	 */
553 	if (ip_ipsec_filtertunnel(m))
554 		goto passin;
555 #else
556 	if (V_ipforwarding == 1)
557 		if (ip_tryforward(m) == NULL)
558 			return;
559 #endif /* IPSEC */
560 
561 	/*
562 	 * Run through list of hooks for input packets.
563 	 *
564 	 * NB: Beware of the destination address changing (e.g.
565 	 *     by NAT rewriting).  When this happens, tell
566 	 *     ip_forward to do the right thing.
567 	 */
568 
569 	/* Jump over all PFIL processing if hooks are not active. */
570 	if (!PFIL_HOOKED(&V_inet_pfil_hook))
571 		goto passin;
572 
573 	odst = ip->ip_dst;
574 	if (pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_IN, NULL) != 0)
575 		return;
576 	if (m == NULL)			/* consumed by filter */
577 		return;
578 
579 	ip = mtod(m, struct ip *);
580 	dchg = (odst.s_addr != ip->ip_dst.s_addr);
581 	ifp = m->m_pkthdr.rcvif;
582 
583 	if (m->m_flags & M_FASTFWD_OURS) {
584 		m->m_flags &= ~M_FASTFWD_OURS;
585 		goto ours;
586 	}
587 	if (m->m_flags & M_IP_NEXTHOP) {
588 		if (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) {
589 			/*
590 			 * Directly ship the packet on.  This allows
591 			 * forwarding packets originally destined to us
592 			 * to some other directly connected host.
593 			 */
594 			ip_forward(m, 1);
595 			return;
596 		}
597 	}
598 passin:
599 
600 	/*
601 	 * Process options and, if not destined for us,
602 	 * ship it on.  ip_dooptions returns 1 when an
603 	 * error was detected (causing an icmp message
604 	 * to be sent and the original packet to be freed).
605 	 */
606 	if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
607 		return;
608 
609         /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
610          * matter if it is destined to another node, or whether it is
611          * a multicast one, RSVP wants it! and prevents it from being forwarded
612          * anywhere else. Also checks if the rsvp daemon is running before
613 	 * grabbing the packet.
614          */
615 	if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP)
616 		goto ours;
617 
618 	/*
619 	 * Check our list of addresses, to see if the packet is for us.
620 	 * If we don't have any addresses, assume any unicast packet
621 	 * we receive might be for us (and let the upper layers deal
622 	 * with it).
623 	 */
624 	if (TAILQ_EMPTY(&V_in_ifaddrhead) &&
625 	    (m->m_flags & (M_MCAST|M_BCAST)) == 0)
626 		goto ours;
627 
628 	/*
629 	 * Enable a consistency check between the destination address
630 	 * and the arrival interface for a unicast packet (the RFC 1122
631 	 * strong ES model) if IP forwarding is disabled and the packet
632 	 * is not locally generated and the packet is not subject to
633 	 * 'ipfw fwd'.
634 	 *
635 	 * XXX - Checking also should be disabled if the destination
636 	 * address is ipnat'ed to a different interface.
637 	 *
638 	 * XXX - Checking is incompatible with IP aliases added
639 	 * to the loopback interface instead of the interface where
640 	 * the packets are received.
641 	 *
642 	 * XXX - This is the case for carp vhost IPs as well so we
643 	 * insert a workaround. If the packet got here, we already
644 	 * checked with carp_iamatch() and carp_forus().
645 	 */
646 	checkif = V_ip_checkinterface && (V_ipforwarding == 0) &&
647 	    ifp != NULL && ((ifp->if_flags & IFF_LOOPBACK) == 0) &&
648 	    ifp->if_carp == NULL && (dchg == 0);
649 
650 	/*
651 	 * Check for exact addresses in the hash bucket.
652 	 */
653 	/* IN_IFADDR_RLOCK(); */
654 	LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
655 		/*
656 		 * If the address matches, verify that the packet
657 		 * arrived via the correct interface if checking is
658 		 * enabled.
659 		 */
660 		if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr &&
661 		    (!checkif || ia->ia_ifp == ifp)) {
662 			counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
663 			counter_u64_add(ia->ia_ifa.ifa_ibytes,
664 			    m->m_pkthdr.len);
665 			/* IN_IFADDR_RUNLOCK(); */
666 			goto ours;
667 		}
668 	}
669 	/* IN_IFADDR_RUNLOCK(); */
670 
671 	/*
672 	 * Check for broadcast addresses.
673 	 *
674 	 * Only accept broadcast packets that arrive via the matching
675 	 * interface.  Reception of forwarded directed broadcasts would
676 	 * be handled via ip_forward() and ether_output() with the loopback
677 	 * into the stack for SIMPLEX interfaces handled by ether_output().
678 	 */
679 	if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
680 		IF_ADDR_RLOCK(ifp);
681 	        TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
682 			if (ifa->ifa_addr->sa_family != AF_INET)
683 				continue;
684 			ia = ifatoia(ifa);
685 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
686 			    ip->ip_dst.s_addr) {
687 				counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
688 				counter_u64_add(ia->ia_ifa.ifa_ibytes,
689 				    m->m_pkthdr.len);
690 				IF_ADDR_RUNLOCK(ifp);
691 				goto ours;
692 			}
693 #ifdef BOOTP_COMPAT
694 			if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) {
695 				counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
696 				counter_u64_add(ia->ia_ifa.ifa_ibytes,
697 				    m->m_pkthdr.len);
698 				IF_ADDR_RUNLOCK(ifp);
699 				goto ours;
700 			}
701 #endif
702 		}
703 		IF_ADDR_RUNLOCK(ifp);
704 		ia = NULL;
705 	}
706 	/* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */
707 	if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) {
708 		IPSTAT_INC(ips_cantforward);
709 		m_freem(m);
710 		return;
711 	}
712 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
713 		if (V_ip_mrouter) {
714 			/*
715 			 * If we are acting as a multicast router, all
716 			 * incoming multicast packets are passed to the
717 			 * kernel-level multicast forwarding function.
718 			 * The packet is returned (relatively) intact; if
719 			 * ip_mforward() returns a non-zero value, the packet
720 			 * must be discarded, else it may be accepted below.
721 			 */
722 			if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) {
723 				IPSTAT_INC(ips_cantforward);
724 				m_freem(m);
725 				return;
726 			}
727 
728 			/*
729 			 * The process-level routing daemon needs to receive
730 			 * all multicast IGMP packets, whether or not this
731 			 * host belongs to their destination groups.
732 			 */
733 			if (ip->ip_p == IPPROTO_IGMP)
734 				goto ours;
735 			IPSTAT_INC(ips_forward);
736 		}
737 		/*
738 		 * Assume the packet is for us, to avoid prematurely taking
739 		 * a lock on the in_multi hash. Protocols must perform
740 		 * their own filtering and update statistics accordingly.
741 		 */
742 		goto ours;
743 	}
744 	if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
745 		goto ours;
746 	if (ip->ip_dst.s_addr == INADDR_ANY)
747 		goto ours;
748 
749 	/*
750 	 * Not for us; forward if possible and desirable.
751 	 */
752 	if (V_ipforwarding == 0) {
753 		IPSTAT_INC(ips_cantforward);
754 		m_freem(m);
755 	} else {
756 		ip_forward(m, dchg);
757 	}
758 	return;
759 
760 ours:
761 #ifdef IPSTEALTH
762 	/*
763 	 * IPSTEALTH: Process non-routing options only
764 	 * if the packet is destined for us.
765 	 */
766 	if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1))
767 		return;
768 #endif /* IPSTEALTH */
769 
770 	/*
771 	 * Attempt reassembly; if it succeeds, proceed.
772 	 * ip_reass() will return a different mbuf.
773 	 */
774 	if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) {
775 		/* XXXGL: shouldn't we save & set m_flags? */
776 		m = ip_reass(m);
777 		if (m == NULL)
778 			return;
779 		ip = mtod(m, struct ip *);
780 		/* Get the header length of the reassembled packet */
781 		hlen = ip->ip_hl << 2;
782 	}
783 
784 #ifdef IPSEC
785 	/*
786 	 * enforce IPsec policy checking if we are seeing last header.
787 	 * note that we do not visit this with protocols with pcb layer
788 	 * code - like udp/tcp/raw ip.
789 	 */
790 	if (ip_ipsec_input(m, ip->ip_p) != 0)
791 		goto bad;
792 #endif /* IPSEC */
793 
794 	/*
795 	 * Switch out to protocol's input routine.
796 	 */
797 	IPSTAT_INC(ips_delivered);
798 
799 	(*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
800 	return;
801 bad:
802 	m_freem(m);
803 }
804 
805 /*
806  * IP timer processing;
807  * if a timer expires on a reassembly
808  * queue, discard it.
809  */
810 void
811 ip_slowtimo(void)
812 {
813 	VNET_ITERATOR_DECL(vnet_iter);
814 
815 	VNET_LIST_RLOCK_NOSLEEP();
816 	VNET_FOREACH(vnet_iter) {
817 		CURVNET_SET(vnet_iter);
818 		ipreass_slowtimo();
819 		CURVNET_RESTORE();
820 	}
821 	VNET_LIST_RUNLOCK_NOSLEEP();
822 }
823 
824 void
825 ip_drain(void)
826 {
827 	VNET_ITERATOR_DECL(vnet_iter);
828 
829 	VNET_LIST_RLOCK_NOSLEEP();
830 	VNET_FOREACH(vnet_iter) {
831 		CURVNET_SET(vnet_iter);
832 		ipreass_drain();
833 		CURVNET_RESTORE();
834 	}
835 	VNET_LIST_RUNLOCK_NOSLEEP();
836 }
837 
838 /*
839  * The protocol to be inserted into ip_protox[] must be already registered
840  * in inetsw[], either statically or through pf_proto_register().
841  */
842 int
843 ipproto_register(short ipproto)
844 {
845 	struct protosw *pr;
846 
847 	/* Sanity checks. */
848 	if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
849 		return (EPROTONOSUPPORT);
850 
851 	/*
852 	 * The protocol slot must not be occupied by another protocol
853 	 * already.  An index pointing to IPPROTO_RAW is unused.
854 	 */
855 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
856 	if (pr == NULL)
857 		return (EPFNOSUPPORT);
858 	if (ip_protox[ipproto] != pr - inetsw)	/* IPPROTO_RAW */
859 		return (EEXIST);
860 
861 	/* Find the protocol position in inetsw[] and set the index. */
862 	for (pr = inetdomain.dom_protosw;
863 	     pr < inetdomain.dom_protoswNPROTOSW; pr++) {
864 		if (pr->pr_domain->dom_family == PF_INET &&
865 		    pr->pr_protocol && pr->pr_protocol == ipproto) {
866 			ip_protox[pr->pr_protocol] = pr - inetsw;
867 			return (0);
868 		}
869 	}
870 	return (EPROTONOSUPPORT);
871 }
872 
873 int
874 ipproto_unregister(short ipproto)
875 {
876 	struct protosw *pr;
877 
878 	/* Sanity checks. */
879 	if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
880 		return (EPROTONOSUPPORT);
881 
882 	/* Check if the protocol was indeed registered. */
883 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
884 	if (pr == NULL)
885 		return (EPFNOSUPPORT);
886 	if (ip_protox[ipproto] == pr - inetsw)  /* IPPROTO_RAW */
887 		return (ENOENT);
888 
889 	/* Reset the protocol slot to IPPROTO_RAW. */
890 	ip_protox[ipproto] = pr - inetsw;
891 	return (0);
892 }
893 
894 u_char inetctlerrmap[PRC_NCMDS] = {
895 	0,		0,		0,		0,
896 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
897 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
898 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
899 	0,		0,		EHOSTUNREACH,	0,
900 	ENOPROTOOPT,	ECONNREFUSED
901 };
902 
903 /*
904  * Forward a packet.  If some error occurs return the sender
905  * an icmp packet.  Note we can't always generate a meaningful
906  * icmp message because icmp doesn't have a large enough repertoire
907  * of codes and types.
908  *
909  * If not forwarding, just drop the packet.  This could be confusing
910  * if ipforwarding was zero but some routing protocol was advancing
911  * us as a gateway to somewhere.  However, we must let the routing
912  * protocol deal with that.
913  *
914  * The srcrt parameter indicates whether the packet is being forwarded
915  * via a source route.
916  */
917 void
918 ip_forward(struct mbuf *m, int srcrt)
919 {
920 	struct ip *ip = mtod(m, struct ip *);
921 	struct in_ifaddr *ia;
922 	struct mbuf *mcopy;
923 	struct sockaddr_in *sin;
924 	struct in_addr dest;
925 	struct route ro;
926 	int error, type = 0, code = 0, mtu = 0;
927 
928 	if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
929 		IPSTAT_INC(ips_cantforward);
930 		m_freem(m);
931 		return;
932 	}
933 #ifdef IPSEC
934 	if (ip_ipsec_fwd(m) != 0) {
935 		IPSTAT_INC(ips_cantforward);
936 		m_freem(m);
937 		return;
938 	}
939 #endif /* IPSEC */
940 #ifdef IPSTEALTH
941 	if (!V_ipstealth) {
942 #endif
943 		if (ip->ip_ttl <= IPTTLDEC) {
944 			icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS,
945 			    0, 0);
946 			return;
947 		}
948 #ifdef IPSTEALTH
949 	}
950 #endif
951 
952 	bzero(&ro, sizeof(ro));
953 	sin = (struct sockaddr_in *)&ro.ro_dst;
954 	sin->sin_family = AF_INET;
955 	sin->sin_len = sizeof(*sin);
956 	sin->sin_addr = ip->ip_dst;
957 #ifdef RADIX_MPATH
958 	rtalloc_mpath_fib(&ro,
959 	    ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
960 	    M_GETFIB(m));
961 #else
962 	in_rtalloc_ign(&ro, 0, M_GETFIB(m));
963 #endif
964 	if (ro.ro_rt != NULL) {
965 		ia = ifatoia(ro.ro_rt->rt_ifa);
966 		ifa_ref(&ia->ia_ifa);
967 	} else
968 		ia = NULL;
969 #ifndef IPSEC
970 	/*
971 	 * 'ia' may be NULL if there is no route for this destination.
972 	 * In case of IPsec, Don't discard it just yet, but pass it to
973 	 * ip_output in case of outgoing IPsec policy.
974 	 */
975 	if (!srcrt && ia == NULL) {
976 		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
977 		RO_RTFREE(&ro);
978 		return;
979 	}
980 #endif
981 
982 	/*
983 	 * Save the IP header and at most 8 bytes of the payload,
984 	 * in case we need to generate an ICMP message to the src.
985 	 *
986 	 * XXX this can be optimized a lot by saving the data in a local
987 	 * buffer on the stack (72 bytes at most), and only allocating the
988 	 * mbuf if really necessary. The vast majority of the packets
989 	 * are forwarded without having to send an ICMP back (either
990 	 * because unnecessary, or because rate limited), so we are
991 	 * really we are wasting a lot of work here.
992 	 *
993 	 * We don't use m_copy() because it might return a reference
994 	 * to a shared cluster. Both this function and ip_output()
995 	 * assume exclusive access to the IP header in `m', so any
996 	 * data in a cluster may change before we reach icmp_error().
997 	 */
998 	mcopy = m_gethdr(M_NOWAIT, m->m_type);
999 	if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_NOWAIT)) {
1000 		/*
1001 		 * It's probably ok if the pkthdr dup fails (because
1002 		 * the deep copy of the tag chain failed), but for now
1003 		 * be conservative and just discard the copy since
1004 		 * code below may some day want the tags.
1005 		 */
1006 		m_free(mcopy);
1007 		mcopy = NULL;
1008 	}
1009 	if (mcopy != NULL) {
1010 		mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy));
1011 		mcopy->m_pkthdr.len = mcopy->m_len;
1012 		m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1013 	}
1014 
1015 #ifdef IPSTEALTH
1016 	if (!V_ipstealth) {
1017 #endif
1018 		ip->ip_ttl -= IPTTLDEC;
1019 #ifdef IPSTEALTH
1020 	}
1021 #endif
1022 
1023 	/*
1024 	 * If forwarding packet using same interface that it came in on,
1025 	 * perhaps should send a redirect to sender to shortcut a hop.
1026 	 * Only send redirect if source is sending directly to us,
1027 	 * and if packet was not source routed (or has any options).
1028 	 * Also, don't send redirect if forwarding using a default route
1029 	 * or a route modified by a redirect.
1030 	 */
1031 	dest.s_addr = 0;
1032 	if (!srcrt && V_ipsendredirects &&
1033 	    ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) {
1034 		struct rtentry *rt;
1035 
1036 		rt = ro.ro_rt;
1037 
1038 		if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1039 		    satosin(rt_key(rt))->sin_addr.s_addr != 0) {
1040 #define	RTA(rt)	((struct in_ifaddr *)(rt->rt_ifa))
1041 			u_long src = ntohl(ip->ip_src.s_addr);
1042 
1043 			if (RTA(rt) &&
1044 			    (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1045 				if (rt->rt_flags & RTF_GATEWAY)
1046 					dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr;
1047 				else
1048 					dest.s_addr = ip->ip_dst.s_addr;
1049 				/* Router requirements says to only send host redirects */
1050 				type = ICMP_REDIRECT;
1051 				code = ICMP_REDIRECT_HOST;
1052 			}
1053 		}
1054 	}
1055 
1056 	error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
1057 
1058 	if (error == EMSGSIZE && ro.ro_rt)
1059 		mtu = ro.ro_rt->rt_mtu;
1060 	RO_RTFREE(&ro);
1061 
1062 	if (error)
1063 		IPSTAT_INC(ips_cantforward);
1064 	else {
1065 		IPSTAT_INC(ips_forward);
1066 		if (type)
1067 			IPSTAT_INC(ips_redirectsent);
1068 		else {
1069 			if (mcopy)
1070 				m_freem(mcopy);
1071 			if (ia != NULL)
1072 				ifa_free(&ia->ia_ifa);
1073 			return;
1074 		}
1075 	}
1076 	if (mcopy == NULL) {
1077 		if (ia != NULL)
1078 			ifa_free(&ia->ia_ifa);
1079 		return;
1080 	}
1081 
1082 	switch (error) {
1083 
1084 	case 0:				/* forwarded, but need redirect */
1085 		/* type, code set above */
1086 		break;
1087 
1088 	case ENETUNREACH:
1089 	case EHOSTUNREACH:
1090 	case ENETDOWN:
1091 	case EHOSTDOWN:
1092 	default:
1093 		type = ICMP_UNREACH;
1094 		code = ICMP_UNREACH_HOST;
1095 		break;
1096 
1097 	case EMSGSIZE:
1098 		type = ICMP_UNREACH;
1099 		code = ICMP_UNREACH_NEEDFRAG;
1100 
1101 #ifdef IPSEC
1102 		/*
1103 		 * If IPsec is configured for this path,
1104 		 * override any possibly mtu value set by ip_output.
1105 		 */
1106 		mtu = ip_ipsec_mtu(mcopy, mtu);
1107 #endif /* IPSEC */
1108 		/*
1109 		 * If the MTU was set before make sure we are below the
1110 		 * interface MTU.
1111 		 * If the MTU wasn't set before use the interface mtu or
1112 		 * fall back to the next smaller mtu step compared to the
1113 		 * current packet size.
1114 		 */
1115 		if (mtu != 0) {
1116 			if (ia != NULL)
1117 				mtu = min(mtu, ia->ia_ifp->if_mtu);
1118 		} else {
1119 			if (ia != NULL)
1120 				mtu = ia->ia_ifp->if_mtu;
1121 			else
1122 				mtu = ip_next_mtu(ntohs(ip->ip_len), 0);
1123 		}
1124 		IPSTAT_INC(ips_cantfrag);
1125 		break;
1126 
1127 	case ENOBUFS:
1128 	case EACCES:			/* ipfw denied packet */
1129 		m_freem(mcopy);
1130 		if (ia != NULL)
1131 			ifa_free(&ia->ia_ifa);
1132 		return;
1133 	}
1134 	if (ia != NULL)
1135 		ifa_free(&ia->ia_ifa);
1136 	icmp_error(mcopy, type, code, dest.s_addr, mtu);
1137 }
1138 
1139 void
1140 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1141     struct mbuf *m)
1142 {
1143 
1144 	if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) {
1145 		struct bintime bt;
1146 
1147 		bintime(&bt);
1148 		if (inp->inp_socket->so_options & SO_BINTIME) {
1149 			*mp = sbcreatecontrol((caddr_t)&bt, sizeof(bt),
1150 			    SCM_BINTIME, SOL_SOCKET);
1151 			if (*mp)
1152 				mp = &(*mp)->m_next;
1153 		}
1154 		if (inp->inp_socket->so_options & SO_TIMESTAMP) {
1155 			struct timeval tv;
1156 
1157 			bintime2timeval(&bt, &tv);
1158 			*mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv),
1159 			    SCM_TIMESTAMP, SOL_SOCKET);
1160 			if (*mp)
1161 				mp = &(*mp)->m_next;
1162 		}
1163 	}
1164 	if (inp->inp_flags & INP_RECVDSTADDR) {
1165 		*mp = sbcreatecontrol((caddr_t)&ip->ip_dst,
1166 		    sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1167 		if (*mp)
1168 			mp = &(*mp)->m_next;
1169 	}
1170 	if (inp->inp_flags & INP_RECVTTL) {
1171 		*mp = sbcreatecontrol((caddr_t)&ip->ip_ttl,
1172 		    sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
1173 		if (*mp)
1174 			mp = &(*mp)->m_next;
1175 	}
1176 #ifdef notyet
1177 	/* XXX
1178 	 * Moving these out of udp_input() made them even more broken
1179 	 * than they already were.
1180 	 */
1181 	/* options were tossed already */
1182 	if (inp->inp_flags & INP_RECVOPTS) {
1183 		*mp = sbcreatecontrol((caddr_t)opts_deleted_above,
1184 		    sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1185 		if (*mp)
1186 			mp = &(*mp)->m_next;
1187 	}
1188 	/* ip_srcroute doesn't do what we want here, need to fix */
1189 	if (inp->inp_flags & INP_RECVRETOPTS) {
1190 		*mp = sbcreatecontrol((caddr_t)ip_srcroute(m),
1191 		    sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1192 		if (*mp)
1193 			mp = &(*mp)->m_next;
1194 	}
1195 #endif
1196 	if (inp->inp_flags & INP_RECVIF) {
1197 		struct ifnet *ifp;
1198 		struct sdlbuf {
1199 			struct sockaddr_dl sdl;
1200 			u_char	pad[32];
1201 		} sdlbuf;
1202 		struct sockaddr_dl *sdp;
1203 		struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1204 
1205 		if ((ifp = m->m_pkthdr.rcvif) &&
1206 		    ifp->if_index && ifp->if_index <= V_if_index) {
1207 			sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1208 			/*
1209 			 * Change our mind and don't try copy.
1210 			 */
1211 			if (sdp->sdl_family != AF_LINK ||
1212 			    sdp->sdl_len > sizeof(sdlbuf)) {
1213 				goto makedummy;
1214 			}
1215 			bcopy(sdp, sdl2, sdp->sdl_len);
1216 		} else {
1217 makedummy:
1218 			sdl2->sdl_len =
1219 			    offsetof(struct sockaddr_dl, sdl_data[0]);
1220 			sdl2->sdl_family = AF_LINK;
1221 			sdl2->sdl_index = 0;
1222 			sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1223 		}
1224 		*mp = sbcreatecontrol((caddr_t)sdl2, sdl2->sdl_len,
1225 		    IP_RECVIF, IPPROTO_IP);
1226 		if (*mp)
1227 			mp = &(*mp)->m_next;
1228 	}
1229 	if (inp->inp_flags & INP_RECVTOS) {
1230 		*mp = sbcreatecontrol((caddr_t)&ip->ip_tos,
1231 		    sizeof(u_char), IP_RECVTOS, IPPROTO_IP);
1232 		if (*mp)
1233 			mp = &(*mp)->m_next;
1234 	}
1235 
1236 	if (inp->inp_flags2 & INP_RECVFLOWID) {
1237 		uint32_t flowid, flow_type;
1238 
1239 		flowid = m->m_pkthdr.flowid;
1240 		flow_type = M_HASHTYPE_GET(m);
1241 
1242 		/*
1243 		 * XXX should handle the failure of one or the
1244 		 * other - don't populate both?
1245 		 */
1246 		*mp = sbcreatecontrol((caddr_t) &flowid,
1247 		    sizeof(uint32_t), IP_FLOWID, IPPROTO_IP);
1248 		if (*mp)
1249 			mp = &(*mp)->m_next;
1250 		*mp = sbcreatecontrol((caddr_t) &flow_type,
1251 		    sizeof(uint32_t), IP_FLOWTYPE, IPPROTO_IP);
1252 		if (*mp)
1253 			mp = &(*mp)->m_next;
1254 	}
1255 
1256 #ifdef	RSS
1257 	if (inp->inp_flags2 & INP_RECVRSSBUCKETID) {
1258 		uint32_t flowid, flow_type;
1259 		uint32_t rss_bucketid;
1260 
1261 		flowid = m->m_pkthdr.flowid;
1262 		flow_type = M_HASHTYPE_GET(m);
1263 
1264 		if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) {
1265 			*mp = sbcreatecontrol((caddr_t) &rss_bucketid,
1266 			   sizeof(uint32_t), IP_RSSBUCKETID, IPPROTO_IP);
1267 			if (*mp)
1268 				mp = &(*mp)->m_next;
1269 		}
1270 	}
1271 #endif
1272 }
1273 
1274 /*
1275  * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
1276  * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
1277  * locking.  This code remains in ip_input.c as ip_mroute.c is optionally
1278  * compiled.
1279  */
1280 static VNET_DEFINE(int, ip_rsvp_on);
1281 VNET_DEFINE(struct socket *, ip_rsvpd);
1282 
1283 #define	V_ip_rsvp_on		VNET(ip_rsvp_on)
1284 
1285 int
1286 ip_rsvp_init(struct socket *so)
1287 {
1288 
1289 	if (so->so_type != SOCK_RAW ||
1290 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
1291 		return EOPNOTSUPP;
1292 
1293 	if (V_ip_rsvpd != NULL)
1294 		return EADDRINUSE;
1295 
1296 	V_ip_rsvpd = so;
1297 	/*
1298 	 * This may seem silly, but we need to be sure we don't over-increment
1299 	 * the RSVP counter, in case something slips up.
1300 	 */
1301 	if (!V_ip_rsvp_on) {
1302 		V_ip_rsvp_on = 1;
1303 		V_rsvp_on++;
1304 	}
1305 
1306 	return 0;
1307 }
1308 
1309 int
1310 ip_rsvp_done(void)
1311 {
1312 
1313 	V_ip_rsvpd = NULL;
1314 	/*
1315 	 * This may seem silly, but we need to be sure we don't over-decrement
1316 	 * the RSVP counter, in case something slips up.
1317 	 */
1318 	if (V_ip_rsvp_on) {
1319 		V_ip_rsvp_on = 0;
1320 		V_rsvp_on--;
1321 	}
1322 	return 0;
1323 }
1324 
1325 int
1326 rsvp_input(struct mbuf **mp, int *offp, int proto)
1327 {
1328 	struct mbuf *m;
1329 
1330 	m = *mp;
1331 	*mp = NULL;
1332 
1333 	if (rsvp_input_p) { /* call the real one if loaded */
1334 		*mp = m;
1335 		rsvp_input_p(mp, offp, proto);
1336 		return (IPPROTO_DONE);
1337 	}
1338 
1339 	/* Can still get packets with rsvp_on = 0 if there is a local member
1340 	 * of the group to which the RSVP packet is addressed.  But in this
1341 	 * case we want to throw the packet away.
1342 	 */
1343 
1344 	if (!V_rsvp_on) {
1345 		m_freem(m);
1346 		return (IPPROTO_DONE);
1347 	}
1348 
1349 	if (V_ip_rsvpd != NULL) {
1350 		*mp = m;
1351 		rip_input(mp, offp, proto);
1352 		return (IPPROTO_DONE);
1353 	}
1354 	/* Drop the packet */
1355 	m_freem(m);
1356 	return (IPPROTO_DONE);
1357 }
1358