xref: /freebsd/sys/netinet/ip_input.c (revision 23648b7d730bcb5cd0ed1bc061ac18387545f814)
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_ipstealth.h"
37 #include "opt_ipsec.h"
38 #include "opt_route.h"
39 #include "opt_rss.h"
40 
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/hhook.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 	if (hhook_head_register(HHOOK_TYPE_IPSEC_IN, AF_INET,
322 	    &V_ipsec_hhh_in[HHOOK_IPSEC_INET],
323 	    HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
324 		printf("%s: WARNING: unable to register input helper hook\n",
325 		    __func__);
326 	if (hhook_head_register(HHOOK_TYPE_IPSEC_OUT, AF_INET,
327 	    &V_ipsec_hhh_out[HHOOK_IPSEC_INET],
328 	    HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
329 		printf("%s: WARNING: unable to register output helper hook\n",
330 		    __func__);
331 
332 	/* Skip initialization of globals for non-default instances. */
333 #ifdef VIMAGE
334 	if (!IS_DEFAULT_VNET(curvnet)) {
335 		netisr_register_vnet(&ip_nh);
336 #ifdef	RSS
337 		netisr_register_vnet(&ip_direct_nh);
338 #endif
339 		return;
340 	}
341 #endif
342 
343 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
344 	if (pr == NULL)
345 		panic("ip_init: PF_INET not found");
346 
347 	/* Initialize the entire ip_protox[] array to IPPROTO_RAW. */
348 	for (i = 0; i < IPPROTO_MAX; i++)
349 		ip_protox[i] = pr - inetsw;
350 	/*
351 	 * Cycle through IP protocols and put them into the appropriate place
352 	 * in ip_protox[].
353 	 */
354 	for (pr = inetdomain.dom_protosw;
355 	    pr < inetdomain.dom_protoswNPROTOSW; pr++)
356 		if (pr->pr_domain->dom_family == PF_INET &&
357 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
358 			/* Be careful to only index valid IP protocols. */
359 			if (pr->pr_protocol < IPPROTO_MAX)
360 				ip_protox[pr->pr_protocol] = pr - inetsw;
361 		}
362 
363 	netisr_register(&ip_nh);
364 #ifdef	RSS
365 	netisr_register(&ip_direct_nh);
366 #endif
367 }
368 
369 #ifdef VIMAGE
370 static void
371 ip_destroy(void *unused __unused)
372 {
373 	struct ifnet *ifp;
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 
398 	/* Remove the IPv4 addresses from all interfaces. */
399 	in_ifscrub_all();
400 
401 	/* Make sure the IPv4 routes are gone as well. */
402 	IFNET_RLOCK();
403 	TAILQ_FOREACH(ifp, &V_ifnet, if_link)
404 		rt_flushifroutes_af(ifp, AF_INET);
405 	IFNET_RUNLOCK();
406 
407 	/* Destroy IP reassembly queue. */
408 	ipreass_destroy();
409 
410 	/* Cleanup in_ifaddr hash table; should be empty. */
411 	hashdestroy(V_in_ifaddrhashtbl, M_IFADDR, V_in_ifaddrhmask);
412 }
413 
414 VNET_SYSUNINIT(ip, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip_destroy, NULL);
415 #endif
416 
417 #ifdef	RSS
418 /*
419  * IP direct input routine.
420  *
421  * This is called when reinjecting completed fragments where
422  * all of the previous checking and book-keeping has been done.
423  */
424 void
425 ip_direct_input(struct mbuf *m)
426 {
427 	struct ip *ip;
428 	int hlen;
429 
430 	ip = mtod(m, struct ip *);
431 	hlen = ip->ip_hl << 2;
432 
433 	IPSTAT_INC(ips_delivered);
434 	(*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
435 	return;
436 }
437 #endif
438 
439 /*
440  * Ip input routine.  Checksum and byte swap header.  If fragmented
441  * try to reassemble.  Process options.  Pass to next level.
442  */
443 void
444 ip_input(struct mbuf *m)
445 {
446 	struct ip *ip = NULL;
447 	struct in_ifaddr *ia = NULL;
448 	struct ifaddr *ifa;
449 	struct ifnet *ifp;
450 	int    checkif, hlen = 0;
451 	uint16_t sum, ip_len;
452 	int dchg = 0;				/* dest changed after fw */
453 	struct in_addr odst;			/* original dst address */
454 
455 	M_ASSERTPKTHDR(m);
456 
457 	if (m->m_flags & M_FASTFWD_OURS) {
458 		m->m_flags &= ~M_FASTFWD_OURS;
459 		/* Set up some basics that will be used later. */
460 		ip = mtod(m, struct ip *);
461 		hlen = ip->ip_hl << 2;
462 		ip_len = ntohs(ip->ip_len);
463 		goto ours;
464 	}
465 
466 	IPSTAT_INC(ips_total);
467 
468 	if (m->m_pkthdr.len < sizeof(struct ip))
469 		goto tooshort;
470 
471 	if (m->m_len < sizeof (struct ip) &&
472 	    (m = m_pullup(m, sizeof (struct ip))) == NULL) {
473 		IPSTAT_INC(ips_toosmall);
474 		return;
475 	}
476 	ip = mtod(m, struct ip *);
477 
478 	if (ip->ip_v != IPVERSION) {
479 		IPSTAT_INC(ips_badvers);
480 		goto bad;
481 	}
482 
483 	hlen = ip->ip_hl << 2;
484 	if (hlen < sizeof(struct ip)) {	/* minimum header length */
485 		IPSTAT_INC(ips_badhlen);
486 		goto bad;
487 	}
488 	if (hlen > m->m_len) {
489 		if ((m = m_pullup(m, hlen)) == NULL) {
490 			IPSTAT_INC(ips_badhlen);
491 			return;
492 		}
493 		ip = mtod(m, struct ip *);
494 	}
495 
496 	IP_PROBE(receive, NULL, NULL, ip, m->m_pkthdr.rcvif, ip, NULL);
497 
498 	/* 127/8 must not appear on wire - RFC1122 */
499 	ifp = m->m_pkthdr.rcvif;
500 	if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
501 	    (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
502 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
503 			IPSTAT_INC(ips_badaddr);
504 			goto bad;
505 		}
506 	}
507 
508 	if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
509 		sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
510 	} else {
511 		if (hlen == sizeof(struct ip)) {
512 			sum = in_cksum_hdr(ip);
513 		} else {
514 			sum = in_cksum(m, hlen);
515 		}
516 	}
517 	if (sum) {
518 		IPSTAT_INC(ips_badsum);
519 		goto bad;
520 	}
521 
522 #ifdef ALTQ
523 	if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
524 		/* packet is dropped by traffic conditioner */
525 		return;
526 #endif
527 
528 	ip_len = ntohs(ip->ip_len);
529 	if (ip_len < hlen) {
530 		IPSTAT_INC(ips_badlen);
531 		goto bad;
532 	}
533 
534 	/*
535 	 * Check that the amount of data in the buffers
536 	 * is as at least much as the IP header would have us expect.
537 	 * Trim mbufs if longer than we expect.
538 	 * Drop packet if shorter than we expect.
539 	 */
540 	if (m->m_pkthdr.len < ip_len) {
541 tooshort:
542 		IPSTAT_INC(ips_tooshort);
543 		goto bad;
544 	}
545 	if (m->m_pkthdr.len > ip_len) {
546 		if (m->m_len == m->m_pkthdr.len) {
547 			m->m_len = ip_len;
548 			m->m_pkthdr.len = ip_len;
549 		} else
550 			m_adj(m, ip_len - m->m_pkthdr.len);
551 	}
552 
553 	/* Try to forward the packet, but if we fail continue */
554 #ifdef IPSEC
555 	/* For now we do not handle IPSEC in tryforward. */
556 	if (!key_havesp(IPSEC_DIR_INBOUND) && !key_havesp(IPSEC_DIR_OUTBOUND) &&
557 	    (V_ipforwarding == 1))
558 		if (ip_tryforward(m) == NULL)
559 			return;
560 	/*
561 	 * Bypass packet filtering for packets previously handled by IPsec.
562 	 */
563 	if (ip_ipsec_filtertunnel(m))
564 		goto passin;
565 #else
566 	if (V_ipforwarding == 1)
567 		if (ip_tryforward(m) == NULL)
568 			return;
569 #endif /* IPSEC */
570 
571 	/*
572 	 * Run through list of hooks for input packets.
573 	 *
574 	 * NB: Beware of the destination address changing (e.g.
575 	 *     by NAT rewriting).  When this happens, tell
576 	 *     ip_forward to do the right thing.
577 	 */
578 
579 	/* Jump over all PFIL processing if hooks are not active. */
580 	if (!PFIL_HOOKED(&V_inet_pfil_hook))
581 		goto passin;
582 
583 	odst = ip->ip_dst;
584 	if (pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_IN, NULL) != 0)
585 		return;
586 	if (m == NULL)			/* consumed by filter */
587 		return;
588 
589 	ip = mtod(m, struct ip *);
590 	dchg = (odst.s_addr != ip->ip_dst.s_addr);
591 	ifp = m->m_pkthdr.rcvif;
592 
593 	if (m->m_flags & M_FASTFWD_OURS) {
594 		m->m_flags &= ~M_FASTFWD_OURS;
595 		goto ours;
596 	}
597 	if (m->m_flags & M_IP_NEXTHOP) {
598 		if (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) {
599 			/*
600 			 * Directly ship the packet on.  This allows
601 			 * forwarding packets originally destined to us
602 			 * to some other directly connected host.
603 			 */
604 			ip_forward(m, 1);
605 			return;
606 		}
607 	}
608 passin:
609 
610 	/*
611 	 * Process options and, if not destined for us,
612 	 * ship it on.  ip_dooptions returns 1 when an
613 	 * error was detected (causing an icmp message
614 	 * to be sent and the original packet to be freed).
615 	 */
616 	if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
617 		return;
618 
619         /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
620          * matter if it is destined to another node, or whether it is
621          * a multicast one, RSVP wants it! and prevents it from being forwarded
622          * anywhere else. Also checks if the rsvp daemon is running before
623 	 * grabbing the packet.
624          */
625 	if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP)
626 		goto ours;
627 
628 	/*
629 	 * Check our list of addresses, to see if the packet is for us.
630 	 * If we don't have any addresses, assume any unicast packet
631 	 * we receive might be for us (and let the upper layers deal
632 	 * with it).
633 	 */
634 	if (TAILQ_EMPTY(&V_in_ifaddrhead) &&
635 	    (m->m_flags & (M_MCAST|M_BCAST)) == 0)
636 		goto ours;
637 
638 	/*
639 	 * Enable a consistency check between the destination address
640 	 * and the arrival interface for a unicast packet (the RFC 1122
641 	 * strong ES model) if IP forwarding is disabled and the packet
642 	 * is not locally generated and the packet is not subject to
643 	 * 'ipfw fwd'.
644 	 *
645 	 * XXX - Checking also should be disabled if the destination
646 	 * address is ipnat'ed to a different interface.
647 	 *
648 	 * XXX - Checking is incompatible with IP aliases added
649 	 * to the loopback interface instead of the interface where
650 	 * the packets are received.
651 	 *
652 	 * XXX - This is the case for carp vhost IPs as well so we
653 	 * insert a workaround. If the packet got here, we already
654 	 * checked with carp_iamatch() and carp_forus().
655 	 */
656 	checkif = V_ip_checkinterface && (V_ipforwarding == 0) &&
657 	    ifp != NULL && ((ifp->if_flags & IFF_LOOPBACK) == 0) &&
658 	    ifp->if_carp == NULL && (dchg == 0);
659 
660 	/*
661 	 * Check for exact addresses in the hash bucket.
662 	 */
663 	/* IN_IFADDR_RLOCK(); */
664 	LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
665 		/*
666 		 * If the address matches, verify that the packet
667 		 * arrived via the correct interface if checking is
668 		 * enabled.
669 		 */
670 		if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr &&
671 		    (!checkif || ia->ia_ifp == ifp)) {
672 			counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
673 			counter_u64_add(ia->ia_ifa.ifa_ibytes,
674 			    m->m_pkthdr.len);
675 			/* IN_IFADDR_RUNLOCK(); */
676 			goto ours;
677 		}
678 	}
679 	/* IN_IFADDR_RUNLOCK(); */
680 
681 	/*
682 	 * Check for broadcast addresses.
683 	 *
684 	 * Only accept broadcast packets that arrive via the matching
685 	 * interface.  Reception of forwarded directed broadcasts would
686 	 * be handled via ip_forward() and ether_output() with the loopback
687 	 * into the stack for SIMPLEX interfaces handled by ether_output().
688 	 */
689 	if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
690 		IF_ADDR_RLOCK(ifp);
691 	        TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
692 			if (ifa->ifa_addr->sa_family != AF_INET)
693 				continue;
694 			ia = ifatoia(ifa);
695 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
696 			    ip->ip_dst.s_addr) {
697 				counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
698 				counter_u64_add(ia->ia_ifa.ifa_ibytes,
699 				    m->m_pkthdr.len);
700 				IF_ADDR_RUNLOCK(ifp);
701 				goto ours;
702 			}
703 #ifdef BOOTP_COMPAT
704 			if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) {
705 				counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
706 				counter_u64_add(ia->ia_ifa.ifa_ibytes,
707 				    m->m_pkthdr.len);
708 				IF_ADDR_RUNLOCK(ifp);
709 				goto ours;
710 			}
711 #endif
712 		}
713 		IF_ADDR_RUNLOCK(ifp);
714 		ia = NULL;
715 	}
716 	/* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */
717 	if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) {
718 		IPSTAT_INC(ips_cantforward);
719 		m_freem(m);
720 		return;
721 	}
722 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
723 		if (V_ip_mrouter) {
724 			/*
725 			 * If we are acting as a multicast router, all
726 			 * incoming multicast packets are passed to the
727 			 * kernel-level multicast forwarding function.
728 			 * The packet is returned (relatively) intact; if
729 			 * ip_mforward() returns a non-zero value, the packet
730 			 * must be discarded, else it may be accepted below.
731 			 */
732 			if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) {
733 				IPSTAT_INC(ips_cantforward);
734 				m_freem(m);
735 				return;
736 			}
737 
738 			/*
739 			 * The process-level routing daemon needs to receive
740 			 * all multicast IGMP packets, whether or not this
741 			 * host belongs to their destination groups.
742 			 */
743 			if (ip->ip_p == IPPROTO_IGMP)
744 				goto ours;
745 			IPSTAT_INC(ips_forward);
746 		}
747 		/*
748 		 * Assume the packet is for us, to avoid prematurely taking
749 		 * a lock on the in_multi hash. Protocols must perform
750 		 * their own filtering and update statistics accordingly.
751 		 */
752 		goto ours;
753 	}
754 	if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
755 		goto ours;
756 	if (ip->ip_dst.s_addr == INADDR_ANY)
757 		goto ours;
758 
759 	/*
760 	 * Not for us; forward if possible and desirable.
761 	 */
762 	if (V_ipforwarding == 0) {
763 		IPSTAT_INC(ips_cantforward);
764 		m_freem(m);
765 	} else {
766 		ip_forward(m, dchg);
767 	}
768 	return;
769 
770 ours:
771 #ifdef IPSTEALTH
772 	/*
773 	 * IPSTEALTH: Process non-routing options only
774 	 * if the packet is destined for us.
775 	 */
776 	if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1))
777 		return;
778 #endif /* IPSTEALTH */
779 
780 	/*
781 	 * Attempt reassembly; if it succeeds, proceed.
782 	 * ip_reass() will return a different mbuf.
783 	 */
784 	if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) {
785 		/* XXXGL: shouldn't we save & set m_flags? */
786 		m = ip_reass(m);
787 		if (m == NULL)
788 			return;
789 		ip = mtod(m, struct ip *);
790 		/* Get the header length of the reassembled packet */
791 		hlen = ip->ip_hl << 2;
792 	}
793 
794 #ifdef IPSEC
795 	/*
796 	 * enforce IPsec policy checking if we are seeing last header.
797 	 * note that we do not visit this with protocols with pcb layer
798 	 * code - like udp/tcp/raw ip.
799 	 */
800 	if (ip_ipsec_input(m, ip->ip_p) != 0)
801 		goto bad;
802 #endif /* IPSEC */
803 
804 	/*
805 	 * Switch out to protocol's input routine.
806 	 */
807 	IPSTAT_INC(ips_delivered);
808 
809 	(*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
810 	return;
811 bad:
812 	m_freem(m);
813 }
814 
815 /*
816  * IP timer processing;
817  * if a timer expires on a reassembly
818  * queue, discard it.
819  */
820 void
821 ip_slowtimo(void)
822 {
823 	VNET_ITERATOR_DECL(vnet_iter);
824 
825 	VNET_LIST_RLOCK_NOSLEEP();
826 	VNET_FOREACH(vnet_iter) {
827 		CURVNET_SET(vnet_iter);
828 		ipreass_slowtimo();
829 		CURVNET_RESTORE();
830 	}
831 	VNET_LIST_RUNLOCK_NOSLEEP();
832 }
833 
834 void
835 ip_drain(void)
836 {
837 	VNET_ITERATOR_DECL(vnet_iter);
838 
839 	VNET_LIST_RLOCK_NOSLEEP();
840 	VNET_FOREACH(vnet_iter) {
841 		CURVNET_SET(vnet_iter);
842 		ipreass_drain();
843 		CURVNET_RESTORE();
844 	}
845 	VNET_LIST_RUNLOCK_NOSLEEP();
846 }
847 
848 /*
849  * The protocol to be inserted into ip_protox[] must be already registered
850  * in inetsw[], either statically or through pf_proto_register().
851  */
852 int
853 ipproto_register(short ipproto)
854 {
855 	struct protosw *pr;
856 
857 	/* Sanity checks. */
858 	if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
859 		return (EPROTONOSUPPORT);
860 
861 	/*
862 	 * The protocol slot must not be occupied by another protocol
863 	 * already.  An index pointing to IPPROTO_RAW is unused.
864 	 */
865 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
866 	if (pr == NULL)
867 		return (EPFNOSUPPORT);
868 	if (ip_protox[ipproto] != pr - inetsw)	/* IPPROTO_RAW */
869 		return (EEXIST);
870 
871 	/* Find the protocol position in inetsw[] and set the index. */
872 	for (pr = inetdomain.dom_protosw;
873 	     pr < inetdomain.dom_protoswNPROTOSW; pr++) {
874 		if (pr->pr_domain->dom_family == PF_INET &&
875 		    pr->pr_protocol && pr->pr_protocol == ipproto) {
876 			ip_protox[pr->pr_protocol] = pr - inetsw;
877 			return (0);
878 		}
879 	}
880 	return (EPROTONOSUPPORT);
881 }
882 
883 int
884 ipproto_unregister(short ipproto)
885 {
886 	struct protosw *pr;
887 
888 	/* Sanity checks. */
889 	if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
890 		return (EPROTONOSUPPORT);
891 
892 	/* Check if the protocol was indeed registered. */
893 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
894 	if (pr == NULL)
895 		return (EPFNOSUPPORT);
896 	if (ip_protox[ipproto] == pr - inetsw)  /* IPPROTO_RAW */
897 		return (ENOENT);
898 
899 	/* Reset the protocol slot to IPPROTO_RAW. */
900 	ip_protox[ipproto] = pr - inetsw;
901 	return (0);
902 }
903 
904 u_char inetctlerrmap[PRC_NCMDS] = {
905 	0,		0,		0,		0,
906 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
907 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
908 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
909 	0,		0,		EHOSTUNREACH,	0,
910 	ENOPROTOOPT,	ECONNREFUSED
911 };
912 
913 /*
914  * Forward a packet.  If some error occurs return the sender
915  * an icmp packet.  Note we can't always generate a meaningful
916  * icmp message because icmp doesn't have a large enough repertoire
917  * of codes and types.
918  *
919  * If not forwarding, just drop the packet.  This could be confusing
920  * if ipforwarding was zero but some routing protocol was advancing
921  * us as a gateway to somewhere.  However, we must let the routing
922  * protocol deal with that.
923  *
924  * The srcrt parameter indicates whether the packet is being forwarded
925  * via a source route.
926  */
927 void
928 ip_forward(struct mbuf *m, int srcrt)
929 {
930 	struct ip *ip = mtod(m, struct ip *);
931 	struct in_ifaddr *ia;
932 	struct mbuf *mcopy;
933 	struct sockaddr_in *sin;
934 	struct in_addr dest;
935 	struct route ro;
936 	int error, type = 0, code = 0, mtu = 0;
937 
938 	if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
939 		IPSTAT_INC(ips_cantforward);
940 		m_freem(m);
941 		return;
942 	}
943 #ifdef IPSEC
944 	if (ip_ipsec_fwd(m) != 0) {
945 		IPSTAT_INC(ips_cantforward);
946 		m_freem(m);
947 		return;
948 	}
949 #endif /* IPSEC */
950 #ifdef IPSTEALTH
951 	if (!V_ipstealth) {
952 #endif
953 		if (ip->ip_ttl <= IPTTLDEC) {
954 			icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS,
955 			    0, 0);
956 			return;
957 		}
958 #ifdef IPSTEALTH
959 	}
960 #endif
961 
962 	bzero(&ro, sizeof(ro));
963 	sin = (struct sockaddr_in *)&ro.ro_dst;
964 	sin->sin_family = AF_INET;
965 	sin->sin_len = sizeof(*sin);
966 	sin->sin_addr = ip->ip_dst;
967 #ifdef RADIX_MPATH
968 	rtalloc_mpath_fib(&ro,
969 	    ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
970 	    M_GETFIB(m));
971 #else
972 	in_rtalloc_ign(&ro, 0, M_GETFIB(m));
973 #endif
974 	if (ro.ro_rt != NULL) {
975 		ia = ifatoia(ro.ro_rt->rt_ifa);
976 		ifa_ref(&ia->ia_ifa);
977 	} else
978 		ia = NULL;
979 #ifndef IPSEC
980 	/*
981 	 * 'ia' may be NULL if there is no route for this destination.
982 	 * In case of IPsec, Don't discard it just yet, but pass it to
983 	 * ip_output in case of outgoing IPsec policy.
984 	 */
985 	if (!srcrt && ia == NULL) {
986 		icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
987 		RO_RTFREE(&ro);
988 		return;
989 	}
990 #endif
991 
992 	/*
993 	 * Save the IP header and at most 8 bytes of the payload,
994 	 * in case we need to generate an ICMP message to the src.
995 	 *
996 	 * XXX this can be optimized a lot by saving the data in a local
997 	 * buffer on the stack (72 bytes at most), and only allocating the
998 	 * mbuf if really necessary. The vast majority of the packets
999 	 * are forwarded without having to send an ICMP back (either
1000 	 * because unnecessary, or because rate limited), so we are
1001 	 * really we are wasting a lot of work here.
1002 	 *
1003 	 * We don't use m_copym() because it might return a reference
1004 	 * to a shared cluster. Both this function and ip_output()
1005 	 * assume exclusive access to the IP header in `m', so any
1006 	 * data in a cluster may change before we reach icmp_error().
1007 	 */
1008 	mcopy = m_gethdr(M_NOWAIT, m->m_type);
1009 	if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_NOWAIT)) {
1010 		/*
1011 		 * It's probably ok if the pkthdr dup fails (because
1012 		 * the deep copy of the tag chain failed), but for now
1013 		 * be conservative and just discard the copy since
1014 		 * code below may some day want the tags.
1015 		 */
1016 		m_free(mcopy);
1017 		mcopy = NULL;
1018 	}
1019 	if (mcopy != NULL) {
1020 		mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy));
1021 		mcopy->m_pkthdr.len = mcopy->m_len;
1022 		m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1023 	}
1024 
1025 #ifdef IPSTEALTH
1026 	if (!V_ipstealth) {
1027 #endif
1028 		ip->ip_ttl -= IPTTLDEC;
1029 #ifdef IPSTEALTH
1030 	}
1031 #endif
1032 
1033 	/*
1034 	 * If forwarding packet using same interface that it came in on,
1035 	 * perhaps should send a redirect to sender to shortcut a hop.
1036 	 * Only send redirect if source is sending directly to us,
1037 	 * and if packet was not source routed (or has any options).
1038 	 * Also, don't send redirect if forwarding using a default route
1039 	 * or a route modified by a redirect.
1040 	 */
1041 	dest.s_addr = 0;
1042 	if (!srcrt && V_ipsendredirects &&
1043 	    ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) {
1044 		struct rtentry *rt;
1045 
1046 		rt = ro.ro_rt;
1047 
1048 		if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1049 		    satosin(rt_key(rt))->sin_addr.s_addr != 0) {
1050 #define	RTA(rt)	((struct in_ifaddr *)(rt->rt_ifa))
1051 			u_long src = ntohl(ip->ip_src.s_addr);
1052 
1053 			if (RTA(rt) &&
1054 			    (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1055 				if (rt->rt_flags & RTF_GATEWAY)
1056 					dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr;
1057 				else
1058 					dest.s_addr = ip->ip_dst.s_addr;
1059 				/* Router requirements says to only send host redirects */
1060 				type = ICMP_REDIRECT;
1061 				code = ICMP_REDIRECT_HOST;
1062 			}
1063 		}
1064 	}
1065 
1066 	error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
1067 
1068 	if (error == EMSGSIZE && ro.ro_rt)
1069 		mtu = ro.ro_rt->rt_mtu;
1070 	RO_RTFREE(&ro);
1071 
1072 	if (error)
1073 		IPSTAT_INC(ips_cantforward);
1074 	else {
1075 		IPSTAT_INC(ips_forward);
1076 		if (type)
1077 			IPSTAT_INC(ips_redirectsent);
1078 		else {
1079 			if (mcopy)
1080 				m_freem(mcopy);
1081 			if (ia != NULL)
1082 				ifa_free(&ia->ia_ifa);
1083 			return;
1084 		}
1085 	}
1086 	if (mcopy == NULL) {
1087 		if (ia != NULL)
1088 			ifa_free(&ia->ia_ifa);
1089 		return;
1090 	}
1091 
1092 	switch (error) {
1093 
1094 	case 0:				/* forwarded, but need redirect */
1095 		/* type, code set above */
1096 		break;
1097 
1098 	case ENETUNREACH:
1099 	case EHOSTUNREACH:
1100 	case ENETDOWN:
1101 	case EHOSTDOWN:
1102 	default:
1103 		type = ICMP_UNREACH;
1104 		code = ICMP_UNREACH_HOST;
1105 		break;
1106 
1107 	case EMSGSIZE:
1108 		type = ICMP_UNREACH;
1109 		code = ICMP_UNREACH_NEEDFRAG;
1110 
1111 #ifdef IPSEC
1112 		/*
1113 		 * If IPsec is configured for this path,
1114 		 * override any possibly mtu value set by ip_output.
1115 		 */
1116 		mtu = ip_ipsec_mtu(mcopy, mtu);
1117 #endif /* IPSEC */
1118 		/*
1119 		 * If the MTU was set before make sure we are below the
1120 		 * interface MTU.
1121 		 * If the MTU wasn't set before use the interface mtu or
1122 		 * fall back to the next smaller mtu step compared to the
1123 		 * current packet size.
1124 		 */
1125 		if (mtu != 0) {
1126 			if (ia != NULL)
1127 				mtu = min(mtu, ia->ia_ifp->if_mtu);
1128 		} else {
1129 			if (ia != NULL)
1130 				mtu = ia->ia_ifp->if_mtu;
1131 			else
1132 				mtu = ip_next_mtu(ntohs(ip->ip_len), 0);
1133 		}
1134 		IPSTAT_INC(ips_cantfrag);
1135 		break;
1136 
1137 	case ENOBUFS:
1138 	case EACCES:			/* ipfw denied packet */
1139 		m_freem(mcopy);
1140 		if (ia != NULL)
1141 			ifa_free(&ia->ia_ifa);
1142 		return;
1143 	}
1144 	if (ia != NULL)
1145 		ifa_free(&ia->ia_ifa);
1146 	icmp_error(mcopy, type, code, dest.s_addr, mtu);
1147 }
1148 
1149 void
1150 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1151     struct mbuf *m)
1152 {
1153 
1154 	if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) {
1155 		struct bintime bt;
1156 
1157 		bintime(&bt);
1158 		if (inp->inp_socket->so_options & SO_BINTIME) {
1159 			*mp = sbcreatecontrol((caddr_t)&bt, sizeof(bt),
1160 			    SCM_BINTIME, SOL_SOCKET);
1161 			if (*mp)
1162 				mp = &(*mp)->m_next;
1163 		}
1164 		if (inp->inp_socket->so_options & SO_TIMESTAMP) {
1165 			struct timeval tv;
1166 
1167 			bintime2timeval(&bt, &tv);
1168 			*mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv),
1169 			    SCM_TIMESTAMP, SOL_SOCKET);
1170 			if (*mp)
1171 				mp = &(*mp)->m_next;
1172 		}
1173 	}
1174 	if (inp->inp_flags & INP_RECVDSTADDR) {
1175 		*mp = sbcreatecontrol((caddr_t)&ip->ip_dst,
1176 		    sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1177 		if (*mp)
1178 			mp = &(*mp)->m_next;
1179 	}
1180 	if (inp->inp_flags & INP_RECVTTL) {
1181 		*mp = sbcreatecontrol((caddr_t)&ip->ip_ttl,
1182 		    sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
1183 		if (*mp)
1184 			mp = &(*mp)->m_next;
1185 	}
1186 #ifdef notyet
1187 	/* XXX
1188 	 * Moving these out of udp_input() made them even more broken
1189 	 * than they already were.
1190 	 */
1191 	/* options were tossed already */
1192 	if (inp->inp_flags & INP_RECVOPTS) {
1193 		*mp = sbcreatecontrol((caddr_t)opts_deleted_above,
1194 		    sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1195 		if (*mp)
1196 			mp = &(*mp)->m_next;
1197 	}
1198 	/* ip_srcroute doesn't do what we want here, need to fix */
1199 	if (inp->inp_flags & INP_RECVRETOPTS) {
1200 		*mp = sbcreatecontrol((caddr_t)ip_srcroute(m),
1201 		    sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1202 		if (*mp)
1203 			mp = &(*mp)->m_next;
1204 	}
1205 #endif
1206 	if (inp->inp_flags & INP_RECVIF) {
1207 		struct ifnet *ifp;
1208 		struct sdlbuf {
1209 			struct sockaddr_dl sdl;
1210 			u_char	pad[32];
1211 		} sdlbuf;
1212 		struct sockaddr_dl *sdp;
1213 		struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1214 
1215 		if ((ifp = m->m_pkthdr.rcvif) &&
1216 		    ifp->if_index && ifp->if_index <= V_if_index) {
1217 			sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1218 			/*
1219 			 * Change our mind and don't try copy.
1220 			 */
1221 			if (sdp->sdl_family != AF_LINK ||
1222 			    sdp->sdl_len > sizeof(sdlbuf)) {
1223 				goto makedummy;
1224 			}
1225 			bcopy(sdp, sdl2, sdp->sdl_len);
1226 		} else {
1227 makedummy:
1228 			sdl2->sdl_len =
1229 			    offsetof(struct sockaddr_dl, sdl_data[0]);
1230 			sdl2->sdl_family = AF_LINK;
1231 			sdl2->sdl_index = 0;
1232 			sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1233 		}
1234 		*mp = sbcreatecontrol((caddr_t)sdl2, sdl2->sdl_len,
1235 		    IP_RECVIF, IPPROTO_IP);
1236 		if (*mp)
1237 			mp = &(*mp)->m_next;
1238 	}
1239 	if (inp->inp_flags & INP_RECVTOS) {
1240 		*mp = sbcreatecontrol((caddr_t)&ip->ip_tos,
1241 		    sizeof(u_char), IP_RECVTOS, IPPROTO_IP);
1242 		if (*mp)
1243 			mp = &(*mp)->m_next;
1244 	}
1245 
1246 	if (inp->inp_flags2 & INP_RECVFLOWID) {
1247 		uint32_t flowid, flow_type;
1248 
1249 		flowid = m->m_pkthdr.flowid;
1250 		flow_type = M_HASHTYPE_GET(m);
1251 
1252 		/*
1253 		 * XXX should handle the failure of one or the
1254 		 * other - don't populate both?
1255 		 */
1256 		*mp = sbcreatecontrol((caddr_t) &flowid,
1257 		    sizeof(uint32_t), IP_FLOWID, IPPROTO_IP);
1258 		if (*mp)
1259 			mp = &(*mp)->m_next;
1260 		*mp = sbcreatecontrol((caddr_t) &flow_type,
1261 		    sizeof(uint32_t), IP_FLOWTYPE, IPPROTO_IP);
1262 		if (*mp)
1263 			mp = &(*mp)->m_next;
1264 	}
1265 
1266 #ifdef	RSS
1267 	if (inp->inp_flags2 & INP_RECVRSSBUCKETID) {
1268 		uint32_t flowid, flow_type;
1269 		uint32_t rss_bucketid;
1270 
1271 		flowid = m->m_pkthdr.flowid;
1272 		flow_type = M_HASHTYPE_GET(m);
1273 
1274 		if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) {
1275 			*mp = sbcreatecontrol((caddr_t) &rss_bucketid,
1276 			   sizeof(uint32_t), IP_RSSBUCKETID, IPPROTO_IP);
1277 			if (*mp)
1278 				mp = &(*mp)->m_next;
1279 		}
1280 	}
1281 #endif
1282 }
1283 
1284 /*
1285  * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
1286  * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
1287  * locking.  This code remains in ip_input.c as ip_mroute.c is optionally
1288  * compiled.
1289  */
1290 static VNET_DEFINE(int, ip_rsvp_on);
1291 VNET_DEFINE(struct socket *, ip_rsvpd);
1292 
1293 #define	V_ip_rsvp_on		VNET(ip_rsvp_on)
1294 
1295 int
1296 ip_rsvp_init(struct socket *so)
1297 {
1298 
1299 	if (so->so_type != SOCK_RAW ||
1300 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
1301 		return EOPNOTSUPP;
1302 
1303 	if (V_ip_rsvpd != NULL)
1304 		return EADDRINUSE;
1305 
1306 	V_ip_rsvpd = so;
1307 	/*
1308 	 * This may seem silly, but we need to be sure we don't over-increment
1309 	 * the RSVP counter, in case something slips up.
1310 	 */
1311 	if (!V_ip_rsvp_on) {
1312 		V_ip_rsvp_on = 1;
1313 		V_rsvp_on++;
1314 	}
1315 
1316 	return 0;
1317 }
1318 
1319 int
1320 ip_rsvp_done(void)
1321 {
1322 
1323 	V_ip_rsvpd = NULL;
1324 	/*
1325 	 * This may seem silly, but we need to be sure we don't over-decrement
1326 	 * the RSVP counter, in case something slips up.
1327 	 */
1328 	if (V_ip_rsvp_on) {
1329 		V_ip_rsvp_on = 0;
1330 		V_rsvp_on--;
1331 	}
1332 	return 0;
1333 }
1334 
1335 int
1336 rsvp_input(struct mbuf **mp, int *offp, int proto)
1337 {
1338 	struct mbuf *m;
1339 
1340 	m = *mp;
1341 	*mp = NULL;
1342 
1343 	if (rsvp_input_p) { /* call the real one if loaded */
1344 		*mp = m;
1345 		rsvp_input_p(mp, offp, proto);
1346 		return (IPPROTO_DONE);
1347 	}
1348 
1349 	/* Can still get packets with rsvp_on = 0 if there is a local member
1350 	 * of the group to which the RSVP packet is addressed.  But in this
1351 	 * case we want to throw the packet away.
1352 	 */
1353 
1354 	if (!V_rsvp_on) {
1355 		m_freem(m);
1356 		return (IPPROTO_DONE);
1357 	}
1358 
1359 	if (V_ip_rsvpd != NULL) {
1360 		*mp = m;
1361 		rip_input(mp, offp, proto);
1362 		return (IPPROTO_DONE);
1363 	}
1364 	/* Drop the packet */
1365 	m_freem(m);
1366 	return (IPPROTO_DONE);
1367 }
1368