xref: /freebsd/sys/netinet/ip_input.c (revision ef36b3f75658d201edb495068db5e1be49593de5)
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  * 3. 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 #include <netinet/in_rss.h>
81 
82 #include <netipsec/ipsec_support.h>
83 
84 #include <sys/socketvar.h>
85 
86 #include <security/mac/mac_framework.h>
87 
88 #ifdef CTASSERT
89 CTASSERT(sizeof(struct ip) == 20);
90 #endif
91 
92 /* IP reassembly functions are defined in ip_reass.c. */
93 extern void ipreass_init(void);
94 extern void ipreass_drain(void);
95 extern void ipreass_slowtimo(void);
96 #ifdef VIMAGE
97 extern void ipreass_destroy(void);
98 #endif
99 
100 struct rmlock in_ifaddr_lock;
101 RM_SYSINIT(in_ifaddr_lock, &in_ifaddr_lock, "in_ifaddr_lock");
102 
103 VNET_DEFINE(int, rsvp_on);
104 
105 VNET_DEFINE(int, ipforwarding);
106 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_VNET | CTLFLAG_RW,
107     &VNET_NAME(ipforwarding), 0,
108     "Enable IP forwarding between interfaces");
109 
110 static VNET_DEFINE(int, ipsendredirects) = 1;	/* XXX */
111 #define	V_ipsendredirects	VNET(ipsendredirects)
112 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_VNET | CTLFLAG_RW,
113     &VNET_NAME(ipsendredirects), 0,
114     "Enable sending IP redirects");
115 
116 /*
117  * XXX - Setting ip_checkinterface mostly implements the receive side of
118  * the Strong ES model described in RFC 1122, but since the routing table
119  * and transmit implementation do not implement the Strong ES model,
120  * setting this to 1 results in an odd hybrid.
121  *
122  * XXX - ip_checkinterface currently must be disabled if you use ipnat
123  * to translate the destination address to another local interface.
124  *
125  * XXX - ip_checkinterface must be disabled if you add IP aliases
126  * to the loopback interface instead of the interface where the
127  * packets for those addresses are received.
128  */
129 static VNET_DEFINE(int, ip_checkinterface);
130 #define	V_ip_checkinterface	VNET(ip_checkinterface)
131 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_VNET | CTLFLAG_RW,
132     &VNET_NAME(ip_checkinterface), 0,
133     "Verify packet arrives on correct interface");
134 
135 VNET_DEFINE(struct pfil_head, inet_pfil_hook);	/* Packet filter hooks */
136 
137 static struct netisr_handler ip_nh = {
138 	.nh_name = "ip",
139 	.nh_handler = ip_input,
140 	.nh_proto = NETISR_IP,
141 #ifdef	RSS
142 	.nh_m2cpuid = rss_soft_m2cpuid_v4,
143 	.nh_policy = NETISR_POLICY_CPU,
144 	.nh_dispatch = NETISR_DISPATCH_HYBRID,
145 #else
146 	.nh_policy = NETISR_POLICY_FLOW,
147 #endif
148 };
149 
150 #ifdef	RSS
151 /*
152  * Directly dispatched frames are currently assumed
153  * to have a flowid already calculated.
154  *
155  * It should likely have something that assert it
156  * actually has valid flow details.
157  */
158 static struct netisr_handler ip_direct_nh = {
159 	.nh_name = "ip_direct",
160 	.nh_handler = ip_direct_input,
161 	.nh_proto = NETISR_IP_DIRECT,
162 	.nh_m2cpuid = rss_soft_m2cpuid_v4,
163 	.nh_policy = NETISR_POLICY_CPU,
164 	.nh_dispatch = NETISR_DISPATCH_HYBRID,
165 };
166 #endif
167 
168 extern	struct domain inetdomain;
169 extern	struct protosw inetsw[];
170 u_char	ip_protox[IPPROTO_MAX];
171 VNET_DEFINE(struct in_ifaddrhead, in_ifaddrhead);  /* first inet address */
172 VNET_DEFINE(struct in_ifaddrhashhead *, in_ifaddrhashtbl); /* inet addr hash table  */
173 VNET_DEFINE(u_long, in_ifaddrhmask);		/* mask for hash table */
174 
175 #ifdef IPCTL_DEFMTU
176 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
177     &ip_mtu, 0, "Default MTU");
178 #endif
179 
180 #ifdef IPSTEALTH
181 VNET_DEFINE(int, ipstealth);
182 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_VNET | CTLFLAG_RW,
183     &VNET_NAME(ipstealth), 0,
184     "IP stealth mode, no TTL decrementation on forwarding");
185 #endif
186 
187 /*
188  * IP statistics are stored in the "array" of counter(9)s.
189  */
190 VNET_PCPUSTAT_DEFINE(struct ipstat, ipstat);
191 VNET_PCPUSTAT_SYSINIT(ipstat);
192 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, IPCTL_STATS, stats, struct ipstat, ipstat,
193     "IP statistics (struct ipstat, netinet/ip_var.h)");
194 
195 #ifdef VIMAGE
196 VNET_PCPUSTAT_SYSUNINIT(ipstat);
197 #endif /* VIMAGE */
198 
199 /*
200  * Kernel module interface for updating ipstat.  The argument is an index
201  * into ipstat treated as an array.
202  */
203 void
204 kmod_ipstat_inc(int statnum)
205 {
206 
207 	counter_u64_add(VNET(ipstat)[statnum], 1);
208 }
209 
210 void
211 kmod_ipstat_dec(int statnum)
212 {
213 
214 	counter_u64_add(VNET(ipstat)[statnum], -1);
215 }
216 
217 static int
218 sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS)
219 {
220 	int error, qlimit;
221 
222 	netisr_getqlimit(&ip_nh, &qlimit);
223 	error = sysctl_handle_int(oidp, &qlimit, 0, req);
224 	if (error || !req->newptr)
225 		return (error);
226 	if (qlimit < 1)
227 		return (EINVAL);
228 	return (netisr_setqlimit(&ip_nh, qlimit));
229 }
230 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen,
231     CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet_intr_queue_maxlen, "I",
232     "Maximum size of the IP input queue");
233 
234 static int
235 sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS)
236 {
237 	u_int64_t qdrops_long;
238 	int error, qdrops;
239 
240 	netisr_getqdrops(&ip_nh, &qdrops_long);
241 	qdrops = qdrops_long;
242 	error = sysctl_handle_int(oidp, &qdrops, 0, req);
243 	if (error || !req->newptr)
244 		return (error);
245 	if (qdrops != 0)
246 		return (EINVAL);
247 	netisr_clearqdrops(&ip_nh);
248 	return (0);
249 }
250 
251 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops,
252     CTLTYPE_INT|CTLFLAG_RD, 0, 0, sysctl_netinet_intr_queue_drops, "I",
253     "Number of packets dropped from the IP input queue");
254 
255 #ifdef	RSS
256 static int
257 sysctl_netinet_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS)
258 {
259 	int error, qlimit;
260 
261 	netisr_getqlimit(&ip_direct_nh, &qlimit);
262 	error = sysctl_handle_int(oidp, &qlimit, 0, req);
263 	if (error || !req->newptr)
264 		return (error);
265 	if (qlimit < 1)
266 		return (EINVAL);
267 	return (netisr_setqlimit(&ip_direct_nh, qlimit));
268 }
269 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRDQMAXLEN, intr_direct_queue_maxlen,
270     CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet_intr_direct_queue_maxlen,
271     "I", "Maximum size of the IP direct input queue");
272 
273 static int
274 sysctl_netinet_intr_direct_queue_drops(SYSCTL_HANDLER_ARGS)
275 {
276 	u_int64_t qdrops_long;
277 	int error, qdrops;
278 
279 	netisr_getqdrops(&ip_direct_nh, &qdrops_long);
280 	qdrops = qdrops_long;
281 	error = sysctl_handle_int(oidp, &qdrops, 0, req);
282 	if (error || !req->newptr)
283 		return (error);
284 	if (qdrops != 0)
285 		return (EINVAL);
286 	netisr_clearqdrops(&ip_direct_nh);
287 	return (0);
288 }
289 
290 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRDQDROPS, intr_direct_queue_drops,
291     CTLTYPE_INT|CTLFLAG_RD, 0, 0, sysctl_netinet_intr_direct_queue_drops, "I",
292     "Number of packets dropped from the IP direct input queue");
293 #endif	/* RSS */
294 
295 /*
296  * IP initialization: fill in IP protocol switch table.
297  * All protocols not implemented in kernel go to raw IP protocol handler.
298  */
299 void
300 ip_init(void)
301 {
302 	struct protosw *pr;
303 	int i;
304 
305 	TAILQ_INIT(&V_in_ifaddrhead);
306 	V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask);
307 
308 	/* Initialize IP reassembly queue. */
309 	ipreass_init();
310 
311 	/* Initialize packet filter hooks. */
312 	V_inet_pfil_hook.ph_type = PFIL_TYPE_AF;
313 	V_inet_pfil_hook.ph_af = AF_INET;
314 	if ((i = pfil_head_register(&V_inet_pfil_hook)) != 0)
315 		printf("%s: WARNING: unable to register pfil hook, "
316 			"error %d\n", __func__, i);
317 
318 	if (hhook_head_register(HHOOK_TYPE_IPSEC_IN, AF_INET,
319 	    &V_ipsec_hhh_in[HHOOK_IPSEC_INET],
320 	    HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
321 		printf("%s: WARNING: unable to register input helper hook\n",
322 		    __func__);
323 	if (hhook_head_register(HHOOK_TYPE_IPSEC_OUT, AF_INET,
324 	    &V_ipsec_hhh_out[HHOOK_IPSEC_INET],
325 	    HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
326 		printf("%s: WARNING: unable to register output helper hook\n",
327 		    __func__);
328 
329 	/* Skip initialization of globals for non-default instances. */
330 #ifdef VIMAGE
331 	if (!IS_DEFAULT_VNET(curvnet)) {
332 		netisr_register_vnet(&ip_nh);
333 #ifdef	RSS
334 		netisr_register_vnet(&ip_direct_nh);
335 #endif
336 		return;
337 	}
338 #endif
339 
340 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
341 	if (pr == NULL)
342 		panic("ip_init: PF_INET not found");
343 
344 	/* Initialize the entire ip_protox[] array to IPPROTO_RAW. */
345 	for (i = 0; i < IPPROTO_MAX; i++)
346 		ip_protox[i] = pr - inetsw;
347 	/*
348 	 * Cycle through IP protocols and put them into the appropriate place
349 	 * in ip_protox[].
350 	 */
351 	for (pr = inetdomain.dom_protosw;
352 	    pr < inetdomain.dom_protoswNPROTOSW; pr++)
353 		if (pr->pr_domain->dom_family == PF_INET &&
354 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
355 			/* Be careful to only index valid IP protocols. */
356 			if (pr->pr_protocol < IPPROTO_MAX)
357 				ip_protox[pr->pr_protocol] = pr - inetsw;
358 		}
359 
360 	netisr_register(&ip_nh);
361 #ifdef	RSS
362 	netisr_register(&ip_direct_nh);
363 #endif
364 }
365 
366 #ifdef VIMAGE
367 static void
368 ip_destroy(void *unused __unused)
369 {
370 	struct ifnet *ifp;
371 	int error;
372 
373 #ifdef	RSS
374 	netisr_unregister_vnet(&ip_direct_nh);
375 #endif
376 	netisr_unregister_vnet(&ip_nh);
377 
378 	if ((error = pfil_head_unregister(&V_inet_pfil_hook)) != 0)
379 		printf("%s: WARNING: unable to unregister pfil hook, "
380 		    "error %d\n", __func__, error);
381 
382 	error = hhook_head_deregister(V_ipsec_hhh_in[HHOOK_IPSEC_INET]);
383 	if (error != 0) {
384 		printf("%s: WARNING: unable to deregister input helper hook "
385 		    "type HHOOK_TYPE_IPSEC_IN, id HHOOK_IPSEC_INET: "
386 		    "error %d returned\n", __func__, error);
387 	}
388 	error = hhook_head_deregister(V_ipsec_hhh_out[HHOOK_IPSEC_INET]);
389 	if (error != 0) {
390 		printf("%s: WARNING: unable to deregister output helper hook "
391 		    "type HHOOK_TYPE_IPSEC_OUT, id HHOOK_IPSEC_INET: "
392 		    "error %d returned\n", __func__, error);
393 	}
394 
395 	/* Remove the IPv4 addresses from all interfaces. */
396 	in_ifscrub_all();
397 
398 	/* Make sure the IPv4 routes are gone as well. */
399 	IFNET_RLOCK();
400 	TAILQ_FOREACH(ifp, &V_ifnet, if_link)
401 		rt_flushifroutes_af(ifp, AF_INET);
402 	IFNET_RUNLOCK();
403 
404 	/* Destroy IP reassembly queue. */
405 	ipreass_destroy();
406 
407 	/* Cleanup in_ifaddr hash table; should be empty. */
408 	hashdestroy(V_in_ifaddrhashtbl, M_IFADDR, V_in_ifaddrhmask);
409 }
410 
411 VNET_SYSUNINIT(ip, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip_destroy, NULL);
412 #endif
413 
414 #ifdef	RSS
415 /*
416  * IP direct input routine.
417  *
418  * This is called when reinjecting completed fragments where
419  * all of the previous checking and book-keeping has been done.
420  */
421 void
422 ip_direct_input(struct mbuf *m)
423 {
424 	struct ip *ip;
425 	int hlen;
426 
427 	ip = mtod(m, struct ip *);
428 	hlen = ip->ip_hl << 2;
429 
430 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
431 	if (IPSEC_ENABLED(ipv4)) {
432 		if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
433 			return;
434 	}
435 #endif /* IPSEC */
436 	IPSTAT_INC(ips_delivered);
437 	(*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
438 	return;
439 }
440 #endif
441 
442 /*
443  * Ip input routine.  Checksum and byte swap header.  If fragmented
444  * try to reassemble.  Process options.  Pass to next level.
445  */
446 void
447 ip_input(struct mbuf *m)
448 {
449 	struct ip *ip = NULL;
450 	struct in_ifaddr *ia = NULL;
451 	struct ifaddr *ifa;
452 	struct ifnet *ifp;
453 	int    checkif, hlen = 0;
454 	uint16_t sum, ip_len;
455 	int dchg = 0;				/* dest changed after fw */
456 	struct in_addr odst;			/* original dst address */
457 
458 	M_ASSERTPKTHDR(m);
459 
460 	if (m->m_flags & M_FASTFWD_OURS) {
461 		m->m_flags &= ~M_FASTFWD_OURS;
462 		/* Set up some basics that will be used later. */
463 		ip = mtod(m, struct ip *);
464 		hlen = ip->ip_hl << 2;
465 		ip_len = ntohs(ip->ip_len);
466 		goto ours;
467 	}
468 
469 	IPSTAT_INC(ips_total);
470 
471 	if (m->m_pkthdr.len < sizeof(struct ip))
472 		goto tooshort;
473 
474 	if (m->m_len < sizeof (struct ip) &&
475 	    (m = m_pullup(m, sizeof (struct ip))) == NULL) {
476 		IPSTAT_INC(ips_toosmall);
477 		return;
478 	}
479 	ip = mtod(m, struct ip *);
480 
481 	if (ip->ip_v != IPVERSION) {
482 		IPSTAT_INC(ips_badvers);
483 		goto bad;
484 	}
485 
486 	hlen = ip->ip_hl << 2;
487 	if (hlen < sizeof(struct ip)) {	/* minimum header length */
488 		IPSTAT_INC(ips_badhlen);
489 		goto bad;
490 	}
491 	if (hlen > m->m_len) {
492 		if ((m = m_pullup(m, hlen)) == NULL) {
493 			IPSTAT_INC(ips_badhlen);
494 			return;
495 		}
496 		ip = mtod(m, struct ip *);
497 	}
498 
499 	IP_PROBE(receive, NULL, NULL, ip, m->m_pkthdr.rcvif, ip, NULL);
500 
501 	/* 127/8 must not appear on wire - RFC1122 */
502 	ifp = m->m_pkthdr.rcvif;
503 	if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
504 	    (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
505 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
506 			IPSTAT_INC(ips_badaddr);
507 			goto bad;
508 		}
509 	}
510 
511 	if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
512 		sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
513 	} else {
514 		if (hlen == sizeof(struct ip)) {
515 			sum = in_cksum_hdr(ip);
516 		} else {
517 			sum = in_cksum(m, hlen);
518 		}
519 	}
520 	if (sum) {
521 		IPSTAT_INC(ips_badsum);
522 		goto bad;
523 	}
524 
525 #ifdef ALTQ
526 	if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
527 		/* packet is dropped by traffic conditioner */
528 		return;
529 #endif
530 
531 	ip_len = ntohs(ip->ip_len);
532 	if (ip_len < hlen) {
533 		IPSTAT_INC(ips_badlen);
534 		goto bad;
535 	}
536 
537 	/*
538 	 * Check that the amount of data in the buffers
539 	 * is as at least much as the IP header would have us expect.
540 	 * Trim mbufs if longer than we expect.
541 	 * Drop packet if shorter than we expect.
542 	 */
543 	if (m->m_pkthdr.len < ip_len) {
544 tooshort:
545 		IPSTAT_INC(ips_tooshort);
546 		goto bad;
547 	}
548 	if (m->m_pkthdr.len > ip_len) {
549 		if (m->m_len == m->m_pkthdr.len) {
550 			m->m_len = ip_len;
551 			m->m_pkthdr.len = ip_len;
552 		} else
553 			m_adj(m, ip_len - m->m_pkthdr.len);
554 	}
555 
556 	/*
557 	 * Try to forward the packet, but if we fail continue.
558 	 * ip_tryforward() does inbound and outbound packet firewall
559 	 * processing. If firewall has decided that destination becomes
560 	 * our local address, it sets M_FASTFWD_OURS flag. In this
561 	 * case skip another inbound firewall processing and update
562 	 * ip pointer.
563 	 */
564 	if (V_ipforwarding != 0
565 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
566 	    && (!IPSEC_ENABLED(ipv4) ||
567 	    IPSEC_CAPS(ipv4, m, IPSEC_CAP_OPERABLE) == 0)
568 #endif
569 	    ) {
570 		if ((m = ip_tryforward(m)) == NULL)
571 			return;
572 		if (m->m_flags & M_FASTFWD_OURS) {
573 			m->m_flags &= ~M_FASTFWD_OURS;
574 			ip = mtod(m, struct ip *);
575 			goto ours;
576 		}
577 	}
578 
579 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
580 	/*
581 	 * Bypass packet filtering for packets previously handled by IPsec.
582 	 */
583 	if (IPSEC_ENABLED(ipv4) &&
584 	    IPSEC_CAPS(ipv4, m, IPSEC_CAP_BYPASS_FILTER) != 0)
585 			goto passin;
586 #endif
587 
588 	/*
589 	 * Run through list of hooks for input packets.
590 	 *
591 	 * NB: Beware of the destination address changing (e.g.
592 	 *     by NAT rewriting).  When this happens, tell
593 	 *     ip_forward to do the right thing.
594 	 */
595 
596 	/* Jump over all PFIL processing if hooks are not active. */
597 	if (!PFIL_HOOKED(&V_inet_pfil_hook))
598 		goto passin;
599 
600 	odst = ip->ip_dst;
601 	if (pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_IN, NULL) != 0)
602 		return;
603 	if (m == NULL)			/* consumed by filter */
604 		return;
605 
606 	ip = mtod(m, struct ip *);
607 	dchg = (odst.s_addr != ip->ip_dst.s_addr);
608 	ifp = m->m_pkthdr.rcvif;
609 
610 	if (m->m_flags & M_FASTFWD_OURS) {
611 		m->m_flags &= ~M_FASTFWD_OURS;
612 		goto ours;
613 	}
614 	if (m->m_flags & M_IP_NEXTHOP) {
615 		if (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) {
616 			/*
617 			 * Directly ship the packet on.  This allows
618 			 * forwarding packets originally destined to us
619 			 * to some other directly connected host.
620 			 */
621 			ip_forward(m, 1);
622 			return;
623 		}
624 	}
625 passin:
626 
627 	/*
628 	 * Process options and, if not destined for us,
629 	 * ship it on.  ip_dooptions returns 1 when an
630 	 * error was detected (causing an icmp message
631 	 * to be sent and the original packet to be freed).
632 	 */
633 	if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
634 		return;
635 
636         /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
637          * matter if it is destined to another node, or whether it is
638          * a multicast one, RSVP wants it! and prevents it from being forwarded
639          * anywhere else. Also checks if the rsvp daemon is running before
640 	 * grabbing the packet.
641          */
642 	if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP)
643 		goto ours;
644 
645 	/*
646 	 * Check our list of addresses, to see if the packet is for us.
647 	 * If we don't have any addresses, assume any unicast packet
648 	 * we receive might be for us (and let the upper layers deal
649 	 * with it).
650 	 */
651 	if (TAILQ_EMPTY(&V_in_ifaddrhead) &&
652 	    (m->m_flags & (M_MCAST|M_BCAST)) == 0)
653 		goto ours;
654 
655 	/*
656 	 * Enable a consistency check between the destination address
657 	 * and the arrival interface for a unicast packet (the RFC 1122
658 	 * strong ES model) if IP forwarding is disabled and the packet
659 	 * is not locally generated and the packet is not subject to
660 	 * 'ipfw fwd'.
661 	 *
662 	 * XXX - Checking also should be disabled if the destination
663 	 * address is ipnat'ed to a different interface.
664 	 *
665 	 * XXX - Checking is incompatible with IP aliases added
666 	 * to the loopback interface instead of the interface where
667 	 * the packets are received.
668 	 *
669 	 * XXX - This is the case for carp vhost IPs as well so we
670 	 * insert a workaround. If the packet got here, we already
671 	 * checked with carp_iamatch() and carp_forus().
672 	 */
673 	checkif = V_ip_checkinterface && (V_ipforwarding == 0) &&
674 	    ifp != NULL && ((ifp->if_flags & IFF_LOOPBACK) == 0) &&
675 	    ifp->if_carp == NULL && (dchg == 0);
676 
677 	/*
678 	 * Check for exact addresses in the hash bucket.
679 	 */
680 	/* IN_IFADDR_RLOCK(); */
681 	LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
682 		/*
683 		 * If the address matches, verify that the packet
684 		 * arrived via the correct interface if checking is
685 		 * enabled.
686 		 */
687 		if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr &&
688 		    (!checkif || ia->ia_ifp == ifp)) {
689 			counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
690 			counter_u64_add(ia->ia_ifa.ifa_ibytes,
691 			    m->m_pkthdr.len);
692 			/* IN_IFADDR_RUNLOCK(); */
693 			goto ours;
694 		}
695 	}
696 	/* IN_IFADDR_RUNLOCK(); */
697 
698 	/*
699 	 * Check for broadcast addresses.
700 	 *
701 	 * Only accept broadcast packets that arrive via the matching
702 	 * interface.  Reception of forwarded directed broadcasts would
703 	 * be handled via ip_forward() and ether_output() with the loopback
704 	 * into the stack for SIMPLEX interfaces handled by ether_output().
705 	 */
706 	if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
707 		IF_ADDR_RLOCK(ifp);
708 	        TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
709 			if (ifa->ifa_addr->sa_family != AF_INET)
710 				continue;
711 			ia = ifatoia(ifa);
712 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
713 			    ip->ip_dst.s_addr) {
714 				counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
715 				counter_u64_add(ia->ia_ifa.ifa_ibytes,
716 				    m->m_pkthdr.len);
717 				IF_ADDR_RUNLOCK(ifp);
718 				goto ours;
719 			}
720 #ifdef BOOTP_COMPAT
721 			if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) {
722 				counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
723 				counter_u64_add(ia->ia_ifa.ifa_ibytes,
724 				    m->m_pkthdr.len);
725 				IF_ADDR_RUNLOCK(ifp);
726 				goto ours;
727 			}
728 #endif
729 		}
730 		IF_ADDR_RUNLOCK(ifp);
731 		ia = NULL;
732 	}
733 	/* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */
734 	if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) {
735 		IPSTAT_INC(ips_cantforward);
736 		m_freem(m);
737 		return;
738 	}
739 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
740 		if (V_ip_mrouter) {
741 			/*
742 			 * If we are acting as a multicast router, all
743 			 * incoming multicast packets are passed to the
744 			 * kernel-level multicast forwarding function.
745 			 * The packet is returned (relatively) intact; if
746 			 * ip_mforward() returns a non-zero value, the packet
747 			 * must be discarded, else it may be accepted below.
748 			 */
749 			if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) {
750 				IPSTAT_INC(ips_cantforward);
751 				m_freem(m);
752 				return;
753 			}
754 
755 			/*
756 			 * The process-level routing daemon needs to receive
757 			 * all multicast IGMP packets, whether or not this
758 			 * host belongs to their destination groups.
759 			 */
760 			if (ip->ip_p == IPPROTO_IGMP)
761 				goto ours;
762 			IPSTAT_INC(ips_forward);
763 		}
764 		/*
765 		 * Assume the packet is for us, to avoid prematurely taking
766 		 * a lock on the in_multi hash. Protocols must perform
767 		 * their own filtering and update statistics accordingly.
768 		 */
769 		goto ours;
770 	}
771 	if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
772 		goto ours;
773 	if (ip->ip_dst.s_addr == INADDR_ANY)
774 		goto ours;
775 
776 	/*
777 	 * Not for us; forward if possible and desirable.
778 	 */
779 	if (V_ipforwarding == 0) {
780 		IPSTAT_INC(ips_cantforward);
781 		m_freem(m);
782 	} else {
783 		ip_forward(m, dchg);
784 	}
785 	return;
786 
787 ours:
788 #ifdef IPSTEALTH
789 	/*
790 	 * IPSTEALTH: Process non-routing options only
791 	 * if the packet is destined for us.
792 	 */
793 	if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1))
794 		return;
795 #endif /* IPSTEALTH */
796 
797 	/*
798 	 * Attempt reassembly; if it succeeds, proceed.
799 	 * ip_reass() will return a different mbuf.
800 	 */
801 	if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) {
802 		/* XXXGL: shouldn't we save & set m_flags? */
803 		m = ip_reass(m);
804 		if (m == NULL)
805 			return;
806 		ip = mtod(m, struct ip *);
807 		/* Get the header length of the reassembled packet */
808 		hlen = ip->ip_hl << 2;
809 	}
810 
811 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
812 	if (IPSEC_ENABLED(ipv4)) {
813 		if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
814 			return;
815 	}
816 #endif /* IPSEC */
817 
818 	/*
819 	 * Switch out to protocol's input routine.
820 	 */
821 	IPSTAT_INC(ips_delivered);
822 
823 	(*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
824 	return;
825 bad:
826 	m_freem(m);
827 }
828 
829 /*
830  * IP timer processing;
831  * if a timer expires on a reassembly
832  * queue, discard it.
833  */
834 void
835 ip_slowtimo(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_slowtimo();
843 		CURVNET_RESTORE();
844 	}
845 	VNET_LIST_RUNLOCK_NOSLEEP();
846 }
847 
848 void
849 ip_drain(void)
850 {
851 	VNET_ITERATOR_DECL(vnet_iter);
852 
853 	VNET_LIST_RLOCK_NOSLEEP();
854 	VNET_FOREACH(vnet_iter) {
855 		CURVNET_SET(vnet_iter);
856 		ipreass_drain();
857 		CURVNET_RESTORE();
858 	}
859 	VNET_LIST_RUNLOCK_NOSLEEP();
860 }
861 
862 /*
863  * The protocol to be inserted into ip_protox[] must be already registered
864  * in inetsw[], either statically or through pf_proto_register().
865  */
866 int
867 ipproto_register(short ipproto)
868 {
869 	struct protosw *pr;
870 
871 	/* Sanity checks. */
872 	if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
873 		return (EPROTONOSUPPORT);
874 
875 	/*
876 	 * The protocol slot must not be occupied by another protocol
877 	 * already.  An index pointing to IPPROTO_RAW is unused.
878 	 */
879 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
880 	if (pr == NULL)
881 		return (EPFNOSUPPORT);
882 	if (ip_protox[ipproto] != pr - inetsw)	/* IPPROTO_RAW */
883 		return (EEXIST);
884 
885 	/* Find the protocol position in inetsw[] and set the index. */
886 	for (pr = inetdomain.dom_protosw;
887 	     pr < inetdomain.dom_protoswNPROTOSW; pr++) {
888 		if (pr->pr_domain->dom_family == PF_INET &&
889 		    pr->pr_protocol && pr->pr_protocol == ipproto) {
890 			ip_protox[pr->pr_protocol] = pr - inetsw;
891 			return (0);
892 		}
893 	}
894 	return (EPROTONOSUPPORT);
895 }
896 
897 int
898 ipproto_unregister(short ipproto)
899 {
900 	struct protosw *pr;
901 
902 	/* Sanity checks. */
903 	if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
904 		return (EPROTONOSUPPORT);
905 
906 	/* Check if the protocol was indeed registered. */
907 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
908 	if (pr == NULL)
909 		return (EPFNOSUPPORT);
910 	if (ip_protox[ipproto] == pr - inetsw)  /* IPPROTO_RAW */
911 		return (ENOENT);
912 
913 	/* Reset the protocol slot to IPPROTO_RAW. */
914 	ip_protox[ipproto] = pr - inetsw;
915 	return (0);
916 }
917 
918 u_char inetctlerrmap[PRC_NCMDS] = {
919 	0,		0,		0,		0,
920 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
921 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
922 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
923 	0,		0,		EHOSTUNREACH,	0,
924 	ENOPROTOOPT,	ECONNREFUSED
925 };
926 
927 /*
928  * Forward a packet.  If some error occurs return the sender
929  * an icmp packet.  Note we can't always generate a meaningful
930  * icmp message because icmp doesn't have a large enough repertoire
931  * of codes and types.
932  *
933  * If not forwarding, just drop the packet.  This could be confusing
934  * if ipforwarding was zero but some routing protocol was advancing
935  * us as a gateway to somewhere.  However, we must let the routing
936  * protocol deal with that.
937  *
938  * The srcrt parameter indicates whether the packet is being forwarded
939  * via a source route.
940  */
941 void
942 ip_forward(struct mbuf *m, int srcrt)
943 {
944 	struct ip *ip = mtod(m, struct ip *);
945 	struct in_ifaddr *ia;
946 	struct mbuf *mcopy;
947 	struct sockaddr_in *sin;
948 	struct in_addr dest;
949 	struct route ro;
950 	int error, type = 0, code = 0, mtu = 0;
951 
952 	if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
953 		IPSTAT_INC(ips_cantforward);
954 		m_freem(m);
955 		return;
956 	}
957 	if (
958 #ifdef IPSTEALTH
959 	    V_ipstealth == 0 &&
960 #endif
961 	    ip->ip_ttl <= IPTTLDEC) {
962 		icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 0, 0);
963 		return;
964 	}
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 	/*
984 	 * Save the IP header and at most 8 bytes of the payload,
985 	 * in case we need to generate an ICMP message to the src.
986 	 *
987 	 * XXX this can be optimized a lot by saving the data in a local
988 	 * buffer on the stack (72 bytes at most), and only allocating the
989 	 * mbuf if really necessary. The vast majority of the packets
990 	 * are forwarded without having to send an ICMP back (either
991 	 * because unnecessary, or because rate limited), so we are
992 	 * really we are wasting a lot of work here.
993 	 *
994 	 * We don't use m_copym() because it might return a reference
995 	 * to a shared cluster. Both this function and ip_output()
996 	 * assume exclusive access to the IP header in `m', so any
997 	 * data in a cluster may change before we reach icmp_error().
998 	 */
999 	mcopy = m_gethdr(M_NOWAIT, m->m_type);
1000 	if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_NOWAIT)) {
1001 		/*
1002 		 * It's probably ok if the pkthdr dup fails (because
1003 		 * the deep copy of the tag chain failed), but for now
1004 		 * be conservative and just discard the copy since
1005 		 * code below may some day want the tags.
1006 		 */
1007 		m_free(mcopy);
1008 		mcopy = NULL;
1009 	}
1010 	if (mcopy != NULL) {
1011 		mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy));
1012 		mcopy->m_pkthdr.len = mcopy->m_len;
1013 		m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1014 	}
1015 #ifdef IPSTEALTH
1016 	if (V_ipstealth == 0)
1017 #endif
1018 		ip->ip_ttl -= IPTTLDEC;
1019 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1020 	if (IPSEC_ENABLED(ipv4)) {
1021 		if ((error = IPSEC_FORWARD(ipv4, m)) != 0) {
1022 			/* mbuf consumed by IPsec */
1023 			m_freem(mcopy);
1024 			if (error != EINPROGRESS)
1025 				IPSTAT_INC(ips_cantforward);
1026 			return;
1027 		}
1028 		/* No IPsec processing required */
1029 	}
1030 #endif /* IPSEC */
1031 	/*
1032 	 * If forwarding packet using same interface that it came in on,
1033 	 * perhaps should send a redirect to sender to shortcut a hop.
1034 	 * Only send redirect if source is sending directly to us,
1035 	 * and if packet was not source routed (or has any options).
1036 	 * Also, don't send redirect if forwarding using a default route
1037 	 * or a route modified by a redirect.
1038 	 */
1039 	dest.s_addr = 0;
1040 	if (!srcrt && V_ipsendredirects &&
1041 	    ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) {
1042 		struct rtentry *rt;
1043 
1044 		rt = ro.ro_rt;
1045 
1046 		if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1047 		    satosin(rt_key(rt))->sin_addr.s_addr != 0) {
1048 #define	RTA(rt)	((struct in_ifaddr *)(rt->rt_ifa))
1049 			u_long src = ntohl(ip->ip_src.s_addr);
1050 
1051 			if (RTA(rt) &&
1052 			    (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1053 				if (rt->rt_flags & RTF_GATEWAY)
1054 					dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr;
1055 				else
1056 					dest.s_addr = ip->ip_dst.s_addr;
1057 				/* Router requirements says to only send host redirects */
1058 				type = ICMP_REDIRECT;
1059 				code = ICMP_REDIRECT_HOST;
1060 			}
1061 		}
1062 	}
1063 
1064 	error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
1065 
1066 	if (error == EMSGSIZE && ro.ro_rt)
1067 		mtu = ro.ro_rt->rt_mtu;
1068 	RO_RTFREE(&ro);
1069 
1070 	if (error)
1071 		IPSTAT_INC(ips_cantforward);
1072 	else {
1073 		IPSTAT_INC(ips_forward);
1074 		if (type)
1075 			IPSTAT_INC(ips_redirectsent);
1076 		else {
1077 			if (mcopy)
1078 				m_freem(mcopy);
1079 			if (ia != NULL)
1080 				ifa_free(&ia->ia_ifa);
1081 			return;
1082 		}
1083 	}
1084 	if (mcopy == NULL) {
1085 		if (ia != NULL)
1086 			ifa_free(&ia->ia_ifa);
1087 		return;
1088 	}
1089 
1090 	switch (error) {
1091 
1092 	case 0:				/* forwarded, but need redirect */
1093 		/* type, code set above */
1094 		break;
1095 
1096 	case ENETUNREACH:
1097 	case EHOSTUNREACH:
1098 	case ENETDOWN:
1099 	case EHOSTDOWN:
1100 	default:
1101 		type = ICMP_UNREACH;
1102 		code = ICMP_UNREACH_HOST;
1103 		break;
1104 
1105 	case EMSGSIZE:
1106 		type = ICMP_UNREACH;
1107 		code = ICMP_UNREACH_NEEDFRAG;
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 #define	CHECK_SO_CT(sp, ct) \
1140     (((sp->so_options & SO_TIMESTAMP) && (sp->so_ts_clock == ct)) ? 1 : 0)
1141 
1142 void
1143 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1144     struct mbuf *m)
1145 {
1146 
1147 	if ((inp->inp_socket->so_options & SO_BINTIME) ||
1148 	    CHECK_SO_CT(inp->inp_socket, SO_TS_BINTIME)) {
1149 		struct bintime bt;
1150 
1151 		bintime(&bt);
1152 		*mp = sbcreatecontrol((caddr_t)&bt, sizeof(bt),
1153 		    SCM_BINTIME, SOL_SOCKET);
1154 		if (*mp)
1155 			mp = &(*mp)->m_next;
1156 	}
1157 	if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME_MICRO)) {
1158 		struct timeval tv;
1159 
1160 		microtime(&tv);
1161 		*mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv),
1162 		    SCM_TIMESTAMP, SOL_SOCKET);
1163 		if (*mp)
1164 			mp = &(*mp)->m_next;
1165 	} else if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME)) {
1166 		struct timespec ts;
1167 
1168 		nanotime(&ts);
1169 		*mp = sbcreatecontrol((caddr_t)&ts, sizeof(ts),
1170 		    SCM_REALTIME, SOL_SOCKET);
1171 		if (*mp)
1172 			mp = &(*mp)->m_next;
1173 	} else if (CHECK_SO_CT(inp->inp_socket, SO_TS_MONOTONIC)) {
1174 		struct timespec ts;
1175 
1176 		nanouptime(&ts);
1177 		*mp = sbcreatecontrol((caddr_t)&ts, sizeof(ts),
1178 		    SCM_MONOTONIC, SOL_SOCKET);
1179 		if (*mp)
1180 			mp = &(*mp)->m_next;
1181 	}
1182 	if (inp->inp_flags & INP_RECVDSTADDR) {
1183 		*mp = sbcreatecontrol((caddr_t)&ip->ip_dst,
1184 		    sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1185 		if (*mp)
1186 			mp = &(*mp)->m_next;
1187 	}
1188 	if (inp->inp_flags & INP_RECVTTL) {
1189 		*mp = sbcreatecontrol((caddr_t)&ip->ip_ttl,
1190 		    sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
1191 		if (*mp)
1192 			mp = &(*mp)->m_next;
1193 	}
1194 #ifdef notyet
1195 	/* XXX
1196 	 * Moving these out of udp_input() made them even more broken
1197 	 * than they already were.
1198 	 */
1199 	/* options were tossed already */
1200 	if (inp->inp_flags & INP_RECVOPTS) {
1201 		*mp = sbcreatecontrol((caddr_t)opts_deleted_above,
1202 		    sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1203 		if (*mp)
1204 			mp = &(*mp)->m_next;
1205 	}
1206 	/* ip_srcroute doesn't do what we want here, need to fix */
1207 	if (inp->inp_flags & INP_RECVRETOPTS) {
1208 		*mp = sbcreatecontrol((caddr_t)ip_srcroute(m),
1209 		    sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1210 		if (*mp)
1211 			mp = &(*mp)->m_next;
1212 	}
1213 #endif
1214 	if (inp->inp_flags & INP_RECVIF) {
1215 		struct ifnet *ifp;
1216 		struct sdlbuf {
1217 			struct sockaddr_dl sdl;
1218 			u_char	pad[32];
1219 		} sdlbuf;
1220 		struct sockaddr_dl *sdp;
1221 		struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1222 
1223 		if ((ifp = m->m_pkthdr.rcvif) &&
1224 		    ifp->if_index && ifp->if_index <= V_if_index) {
1225 			sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1226 			/*
1227 			 * Change our mind and don't try copy.
1228 			 */
1229 			if (sdp->sdl_family != AF_LINK ||
1230 			    sdp->sdl_len > sizeof(sdlbuf)) {
1231 				goto makedummy;
1232 			}
1233 			bcopy(sdp, sdl2, sdp->sdl_len);
1234 		} else {
1235 makedummy:
1236 			sdl2->sdl_len =
1237 			    offsetof(struct sockaddr_dl, sdl_data[0]);
1238 			sdl2->sdl_family = AF_LINK;
1239 			sdl2->sdl_index = 0;
1240 			sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1241 		}
1242 		*mp = sbcreatecontrol((caddr_t)sdl2, sdl2->sdl_len,
1243 		    IP_RECVIF, IPPROTO_IP);
1244 		if (*mp)
1245 			mp = &(*mp)->m_next;
1246 	}
1247 	if (inp->inp_flags & INP_RECVTOS) {
1248 		*mp = sbcreatecontrol((caddr_t)&ip->ip_tos,
1249 		    sizeof(u_char), IP_RECVTOS, IPPROTO_IP);
1250 		if (*mp)
1251 			mp = &(*mp)->m_next;
1252 	}
1253 
1254 	if (inp->inp_flags2 & INP_RECVFLOWID) {
1255 		uint32_t flowid, flow_type;
1256 
1257 		flowid = m->m_pkthdr.flowid;
1258 		flow_type = M_HASHTYPE_GET(m);
1259 
1260 		/*
1261 		 * XXX should handle the failure of one or the
1262 		 * other - don't populate both?
1263 		 */
1264 		*mp = sbcreatecontrol((caddr_t) &flowid,
1265 		    sizeof(uint32_t), IP_FLOWID, IPPROTO_IP);
1266 		if (*mp)
1267 			mp = &(*mp)->m_next;
1268 		*mp = sbcreatecontrol((caddr_t) &flow_type,
1269 		    sizeof(uint32_t), IP_FLOWTYPE, IPPROTO_IP);
1270 		if (*mp)
1271 			mp = &(*mp)->m_next;
1272 	}
1273 
1274 #ifdef	RSS
1275 	if (inp->inp_flags2 & INP_RECVRSSBUCKETID) {
1276 		uint32_t flowid, flow_type;
1277 		uint32_t rss_bucketid;
1278 
1279 		flowid = m->m_pkthdr.flowid;
1280 		flow_type = M_HASHTYPE_GET(m);
1281 
1282 		if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) {
1283 			*mp = sbcreatecontrol((caddr_t) &rss_bucketid,
1284 			   sizeof(uint32_t), IP_RSSBUCKETID, IPPROTO_IP);
1285 			if (*mp)
1286 				mp = &(*mp)->m_next;
1287 		}
1288 	}
1289 #endif
1290 }
1291 
1292 /*
1293  * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
1294  * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
1295  * locking.  This code remains in ip_input.c as ip_mroute.c is optionally
1296  * compiled.
1297  */
1298 static VNET_DEFINE(int, ip_rsvp_on);
1299 VNET_DEFINE(struct socket *, ip_rsvpd);
1300 
1301 #define	V_ip_rsvp_on		VNET(ip_rsvp_on)
1302 
1303 int
1304 ip_rsvp_init(struct socket *so)
1305 {
1306 
1307 	if (so->so_type != SOCK_RAW ||
1308 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
1309 		return EOPNOTSUPP;
1310 
1311 	if (V_ip_rsvpd != NULL)
1312 		return EADDRINUSE;
1313 
1314 	V_ip_rsvpd = so;
1315 	/*
1316 	 * This may seem silly, but we need to be sure we don't over-increment
1317 	 * the RSVP counter, in case something slips up.
1318 	 */
1319 	if (!V_ip_rsvp_on) {
1320 		V_ip_rsvp_on = 1;
1321 		V_rsvp_on++;
1322 	}
1323 
1324 	return 0;
1325 }
1326 
1327 int
1328 ip_rsvp_done(void)
1329 {
1330 
1331 	V_ip_rsvpd = NULL;
1332 	/*
1333 	 * This may seem silly, but we need to be sure we don't over-decrement
1334 	 * the RSVP counter, in case something slips up.
1335 	 */
1336 	if (V_ip_rsvp_on) {
1337 		V_ip_rsvp_on = 0;
1338 		V_rsvp_on--;
1339 	}
1340 	return 0;
1341 }
1342 
1343 int
1344 rsvp_input(struct mbuf **mp, int *offp, int proto)
1345 {
1346 	struct mbuf *m;
1347 
1348 	m = *mp;
1349 	*mp = NULL;
1350 
1351 	if (rsvp_input_p) { /* call the real one if loaded */
1352 		*mp = m;
1353 		rsvp_input_p(mp, offp, proto);
1354 		return (IPPROTO_DONE);
1355 	}
1356 
1357 	/* Can still get packets with rsvp_on = 0 if there is a local member
1358 	 * of the group to which the RSVP packet is addressed.  But in this
1359 	 * case we want to throw the packet away.
1360 	 */
1361 
1362 	if (!V_rsvp_on) {
1363 		m_freem(m);
1364 		return (IPPROTO_DONE);
1365 	}
1366 
1367 	if (V_ip_rsvpd != NULL) {
1368 		*mp = m;
1369 		rip_input(mp, offp, proto);
1370 		return (IPPROTO_DONE);
1371 	}
1372 	/* Drop the packet */
1373 	m_freem(m);
1374 	return (IPPROTO_DONE);
1375 }
1376