xref: /freebsd-13.1/sys/netinet/ip_input.c (revision a7e7700f)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1982, 1986, 1988, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the University nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	@(#)ip_input.c	8.2 (Berkeley) 1/4/94
32  */
33 
34 #include <sys/cdefs.h>
35 __FBSDID("$FreeBSD$");
36 
37 #include "opt_bootp.h"
38 #include "opt_ipstealth.h"
39 #include "opt_ipsec.h"
40 #include "opt_route.h"
41 #include "opt_rss.h"
42 
43 #include <sys/param.h>
44 #include <sys/systm.h>
45 #include <sys/hhook.h>
46 #include <sys/mbuf.h>
47 #include <sys/malloc.h>
48 #include <sys/domain.h>
49 #include <sys/protosw.h>
50 #include <sys/socket.h>
51 #include <sys/time.h>
52 #include <sys/kernel.h>
53 #include <sys/lock.h>
54 #include <sys/rmlock.h>
55 #include <sys/rwlock.h>
56 #include <sys/sdt.h>
57 #include <sys/syslog.h>
58 #include <sys/sysctl.h>
59 
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/pfil.h>
65 #include <net/route.h>
66 #include <net/route/nhop.h>
67 #include <net/netisr.h>
68 #include <net/rss_config.h>
69 #include <net/vnet.h>
70 
71 #include <netinet/in.h>
72 #include <netinet/in_kdtrace.h>
73 #include <netinet/in_systm.h>
74 #include <netinet/in_var.h>
75 #include <netinet/ip.h>
76 #include <netinet/in_fib.h>
77 #include <netinet/in_pcb.h>
78 #include <netinet/ip_var.h>
79 #include <netinet/ip_fw.h>
80 #include <netinet/ip_icmp.h>
81 #include <netinet/ip_options.h>
82 #include <machine/in_cksum.h>
83 #include <netinet/ip_carp.h>
84 #include <netinet/in_rss.h>
85 
86 #include <netipsec/ipsec_support.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 /*
115  * Respond with an ICMP host redirect when we forward a packet out of
116  * the same interface on which it was received.  See RFC 792.
117  */
118 VNET_DEFINE(int, ipsendredirects) = 1;
119 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_VNET | CTLFLAG_RW,
120     &VNET_NAME(ipsendredirects), 0,
121     "Enable sending IP redirects");
122 
123 /*
124  * XXX - Setting ip_checkinterface mostly implements the receive side of
125  * the Strong ES model described in RFC 1122, but since the routing table
126  * and transmit implementation do not implement the Strong ES model,
127  * setting this to 1 results in an odd hybrid.
128  *
129  * XXX - ip_checkinterface currently must be disabled if you use ipnat
130  * to translate the destination address to another local interface.
131  *
132  * XXX - ip_checkinterface must be disabled if you add IP aliases
133  * to the loopback interface instead of the interface where the
134  * packets for those addresses are received.
135  */
136 VNET_DEFINE_STATIC(int, ip_checkinterface);
137 #define	V_ip_checkinterface	VNET(ip_checkinterface)
138 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_VNET | CTLFLAG_RW,
139     &VNET_NAME(ip_checkinterface), 0,
140     "Verify packet arrives on correct interface");
141 
142 VNET_DEFINE(pfil_head_t, inet_pfil_head);	/* Packet filter hooks */
143 
144 static struct netisr_handler ip_nh = {
145 	.nh_name = "ip",
146 	.nh_handler = ip_input,
147 	.nh_proto = NETISR_IP,
148 #ifdef	RSS
149 	.nh_m2cpuid = rss_soft_m2cpuid_v4,
150 	.nh_policy = NETISR_POLICY_CPU,
151 	.nh_dispatch = NETISR_DISPATCH_HYBRID,
152 #else
153 	.nh_policy = NETISR_POLICY_FLOW,
154 #endif
155 };
156 
157 #ifdef	RSS
158 /*
159  * Directly dispatched frames are currently assumed
160  * to have a flowid already calculated.
161  *
162  * It should likely have something that assert it
163  * actually has valid flow details.
164  */
165 static struct netisr_handler ip_direct_nh = {
166 	.nh_name = "ip_direct",
167 	.nh_handler = ip_direct_input,
168 	.nh_proto = NETISR_IP_DIRECT,
169 	.nh_m2cpuid = rss_soft_m2cpuid_v4,
170 	.nh_policy = NETISR_POLICY_CPU,
171 	.nh_dispatch = NETISR_DISPATCH_HYBRID,
172 };
173 #endif
174 
175 extern	struct domain inetdomain;
176 extern	struct protosw inetsw[];
177 u_char	ip_protox[IPPROTO_MAX];
178 VNET_DEFINE(struct in_ifaddrhead, in_ifaddrhead);  /* first inet address */
179 VNET_DEFINE(struct in_ifaddrhashhead *, in_ifaddrhashtbl); /* inet addr hash table  */
180 VNET_DEFINE(u_long, in_ifaddrhmask);		/* mask for hash table */
181 
182 #ifdef IPCTL_DEFMTU
183 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
184     &ip_mtu, 0, "Default MTU");
185 #endif
186 
187 #ifdef IPSTEALTH
188 VNET_DEFINE(int, ipstealth);
189 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_VNET | CTLFLAG_RW,
190     &VNET_NAME(ipstealth), 0,
191     "IP stealth mode, no TTL decrementation on forwarding");
192 #endif
193 
194 /*
195  * IP statistics are stored in the "array" of counter(9)s.
196  */
197 VNET_PCPUSTAT_DEFINE(struct ipstat, ipstat);
198 VNET_PCPUSTAT_SYSINIT(ipstat);
199 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, IPCTL_STATS, stats, struct ipstat, ipstat,
200     "IP statistics (struct ipstat, netinet/ip_var.h)");
201 
202 #ifdef VIMAGE
203 VNET_PCPUSTAT_SYSUNINIT(ipstat);
204 #endif /* VIMAGE */
205 
206 /*
207  * Kernel module interface for updating ipstat.  The argument is an index
208  * into ipstat treated as an array.
209  */
210 void
kmod_ipstat_inc(int statnum)211 kmod_ipstat_inc(int statnum)
212 {
213 
214 	counter_u64_add(VNET(ipstat)[statnum], 1);
215 }
216 
217 void
kmod_ipstat_dec(int statnum)218 kmod_ipstat_dec(int statnum)
219 {
220 
221 	counter_u64_add(VNET(ipstat)[statnum], -1);
222 }
223 
224 static int
sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS)225 sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS)
226 {
227 	int error, qlimit;
228 
229 	netisr_getqlimit(&ip_nh, &qlimit);
230 	error = sysctl_handle_int(oidp, &qlimit, 0, req);
231 	if (error || !req->newptr)
232 		return (error);
233 	if (qlimit < 1)
234 		return (EINVAL);
235 	return (netisr_setqlimit(&ip_nh, qlimit));
236 }
237 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen,
238     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 0,
239     sysctl_netinet_intr_queue_maxlen, "I",
240     "Maximum size of the IP input queue");
241 
242 static int
sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS)243 sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS)
244 {
245 	u_int64_t qdrops_long;
246 	int error, qdrops;
247 
248 	netisr_getqdrops(&ip_nh, &qdrops_long);
249 	qdrops = qdrops_long;
250 	error = sysctl_handle_int(oidp, &qdrops, 0, req);
251 	if (error || !req->newptr)
252 		return (error);
253 	if (qdrops != 0)
254 		return (EINVAL);
255 	netisr_clearqdrops(&ip_nh);
256 	return (0);
257 }
258 
259 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops,
260     CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE,
261     0, 0, sysctl_netinet_intr_queue_drops, "I",
262     "Number of packets dropped from the IP input queue");
263 
264 #ifdef	RSS
265 static int
sysctl_netinet_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS)266 sysctl_netinet_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS)
267 {
268 	int error, qlimit;
269 
270 	netisr_getqlimit(&ip_direct_nh, &qlimit);
271 	error = sysctl_handle_int(oidp, &qlimit, 0, req);
272 	if (error || !req->newptr)
273 		return (error);
274 	if (qlimit < 1)
275 		return (EINVAL);
276 	return (netisr_setqlimit(&ip_direct_nh, qlimit));
277 }
278 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRDQMAXLEN, intr_direct_queue_maxlen,
279     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
280     0, 0, sysctl_netinet_intr_direct_queue_maxlen,
281     "I", "Maximum size of the IP direct input queue");
282 
283 static int
sysctl_netinet_intr_direct_queue_drops(SYSCTL_HANDLER_ARGS)284 sysctl_netinet_intr_direct_queue_drops(SYSCTL_HANDLER_ARGS)
285 {
286 	u_int64_t qdrops_long;
287 	int error, qdrops;
288 
289 	netisr_getqdrops(&ip_direct_nh, &qdrops_long);
290 	qdrops = qdrops_long;
291 	error = sysctl_handle_int(oidp, &qdrops, 0, req);
292 	if (error || !req->newptr)
293 		return (error);
294 	if (qdrops != 0)
295 		return (EINVAL);
296 	netisr_clearqdrops(&ip_direct_nh);
297 	return (0);
298 }
299 
300 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRDQDROPS, intr_direct_queue_drops,
301     CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, 0, 0,
302     sysctl_netinet_intr_direct_queue_drops, "I",
303     "Number of packets dropped from the IP direct input queue");
304 #endif	/* RSS */
305 
306 /*
307  * IP initialization: fill in IP protocol switch table.
308  * All protocols not implemented in kernel go to raw IP protocol handler.
309  */
310 void
ip_init(void)311 ip_init(void)
312 {
313 	struct pfil_head_args args;
314 	struct protosw *pr;
315 	int i;
316 
317 	CK_STAILQ_INIT(&V_in_ifaddrhead);
318 	V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask);
319 
320 	/* Initialize IP reassembly queue. */
321 	ipreass_init();
322 
323 	/* Initialize packet filter hooks. */
324 	args.pa_version = PFIL_VERSION;
325 	args.pa_flags = PFIL_IN | PFIL_OUT;
326 	args.pa_type = PFIL_TYPE_IP4;
327 	args.pa_headname = PFIL_INET_NAME;
328 	V_inet_pfil_head = pfil_head_register(&args);
329 
330 	if (hhook_head_register(HHOOK_TYPE_IPSEC_IN, AF_INET,
331 	    &V_ipsec_hhh_in[HHOOK_IPSEC_INET],
332 	    HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
333 		printf("%s: WARNING: unable to register input helper hook\n",
334 		    __func__);
335 	if (hhook_head_register(HHOOK_TYPE_IPSEC_OUT, AF_INET,
336 	    &V_ipsec_hhh_out[HHOOK_IPSEC_INET],
337 	    HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
338 		printf("%s: WARNING: unable to register output helper hook\n",
339 		    __func__);
340 
341 	/* Skip initialization of globals for non-default instances. */
342 #ifdef VIMAGE
343 	if (!IS_DEFAULT_VNET(curvnet)) {
344 		netisr_register_vnet(&ip_nh);
345 #ifdef	RSS
346 		netisr_register_vnet(&ip_direct_nh);
347 #endif
348 		return;
349 	}
350 #endif
351 
352 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
353 	if (pr == NULL)
354 		panic("ip_init: PF_INET not found");
355 
356 	/* Initialize the entire ip_protox[] array to IPPROTO_RAW. */
357 	for (i = 0; i < IPPROTO_MAX; i++)
358 		ip_protox[i] = pr - inetsw;
359 	/*
360 	 * Cycle through IP protocols and put them into the appropriate place
361 	 * in ip_protox[].
362 	 */
363 	for (pr = inetdomain.dom_protosw;
364 	    pr < inetdomain.dom_protoswNPROTOSW; pr++)
365 		if (pr->pr_domain->dom_family == PF_INET &&
366 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
367 			/* Be careful to only index valid IP protocols. */
368 			if (pr->pr_protocol < IPPROTO_MAX)
369 				ip_protox[pr->pr_protocol] = pr - inetsw;
370 		}
371 
372 	netisr_register(&ip_nh);
373 #ifdef	RSS
374 	netisr_register(&ip_direct_nh);
375 #endif
376 }
377 
378 #ifdef VIMAGE
379 static void
ip_destroy(void * unused __unused)380 ip_destroy(void *unused __unused)
381 {
382 	int error;
383 
384 #ifdef	RSS
385 	netisr_unregister_vnet(&ip_direct_nh);
386 #endif
387 	netisr_unregister_vnet(&ip_nh);
388 
389 	pfil_head_unregister(V_inet_pfil_head);
390 	error = hhook_head_deregister(V_ipsec_hhh_in[HHOOK_IPSEC_INET]);
391 	if (error != 0) {
392 		printf("%s: WARNING: unable to deregister input helper hook "
393 		    "type HHOOK_TYPE_IPSEC_IN, id HHOOK_IPSEC_INET: "
394 		    "error %d returned\n", __func__, error);
395 	}
396 	error = hhook_head_deregister(V_ipsec_hhh_out[HHOOK_IPSEC_INET]);
397 	if (error != 0) {
398 		printf("%s: WARNING: unable to deregister output helper hook "
399 		    "type HHOOK_TYPE_IPSEC_OUT, id HHOOK_IPSEC_INET: "
400 		    "error %d returned\n", __func__, error);
401 	}
402 
403 	/* Remove the IPv4 addresses from all interfaces. */
404 	in_ifscrub_all();
405 
406 	/* Make sure the IPv4 routes are gone as well. */
407 	rib_flush_routes_family(AF_INET);
408 
409 	/* Destroy IP reassembly queue. */
410 	ipreass_destroy();
411 
412 	/* Cleanup in_ifaddr hash table; should be empty. */
413 	hashdestroy(V_in_ifaddrhashtbl, M_IFADDR, V_in_ifaddrhmask);
414 }
415 
416 VNET_SYSUNINIT(ip, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip_destroy, NULL);
417 #endif
418 
419 #ifdef	RSS
420 /*
421  * IP direct input routine.
422  *
423  * This is called when reinjecting completed fragments where
424  * all of the previous checking and book-keeping has been done.
425  */
426 void
ip_direct_input(struct mbuf * m)427 ip_direct_input(struct mbuf *m)
428 {
429 	struct ip *ip;
430 	int hlen;
431 
432 	ip = mtod(m, struct ip *);
433 	hlen = ip->ip_hl << 2;
434 
435 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
436 	if (IPSEC_ENABLED(ipv4)) {
437 		if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
438 			return;
439 	}
440 #endif /* IPSEC */
441 	IPSTAT_INC(ips_delivered);
442 	(*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
443 	return;
444 }
445 #endif
446 
447 /*
448  * Ip input routine.  Checksum and byte swap header.  If fragmented
449  * try to reassemble.  Process options.  Pass to next level.
450  */
451 void
ip_input(struct mbuf * m)452 ip_input(struct mbuf *m)
453 {
454 	struct rm_priotracker in_ifa_tracker;
455 	struct ip *ip = NULL;
456 	struct in_ifaddr *ia = NULL;
457 	struct ifaddr *ifa;
458 	struct ifnet *ifp;
459 	int    checkif, hlen = 0;
460 	uint16_t sum, ip_len;
461 	int dchg = 0;				/* dest changed after fw */
462 	struct in_addr odst;			/* original dst address */
463 
464 	M_ASSERTPKTHDR(m);
465 	NET_EPOCH_ASSERT();
466 
467 	if (m->m_flags & M_FASTFWD_OURS) {
468 		m->m_flags &= ~M_FASTFWD_OURS;
469 		/* Set up some basics that will be used later. */
470 		ip = mtod(m, struct ip *);
471 		hlen = ip->ip_hl << 2;
472 		ip_len = ntohs(ip->ip_len);
473 		goto ours;
474 	}
475 
476 	IPSTAT_INC(ips_total);
477 
478 	if (m->m_pkthdr.len < sizeof(struct ip))
479 		goto tooshort;
480 
481 	if (m->m_len < sizeof (struct ip) &&
482 	    (m = m_pullup(m, sizeof (struct ip))) == NULL) {
483 		IPSTAT_INC(ips_toosmall);
484 		return;
485 	}
486 	ip = mtod(m, struct ip *);
487 
488 	if (ip->ip_v != IPVERSION) {
489 		IPSTAT_INC(ips_badvers);
490 		goto bad;
491 	}
492 
493 	hlen = ip->ip_hl << 2;
494 	if (hlen < sizeof(struct ip)) {	/* minimum header length */
495 		IPSTAT_INC(ips_badhlen);
496 		goto bad;
497 	}
498 	if (hlen > m->m_len) {
499 		if ((m = m_pullup(m, hlen)) == NULL) {
500 			IPSTAT_INC(ips_badhlen);
501 			return;
502 		}
503 		ip = mtod(m, struct ip *);
504 	}
505 
506 	IP_PROBE(receive, NULL, NULL, ip, m->m_pkthdr.rcvif, ip, NULL);
507 
508 	/* IN_LOOPBACK must not appear on the wire - RFC1122 */
509 	ifp = m->m_pkthdr.rcvif;
510 	if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) ||
511 	    IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) {
512 		if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
513 			IPSTAT_INC(ips_badaddr);
514 			goto bad;
515 		}
516 	}
517 
518 	if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
519 		sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
520 	} else {
521 		if (hlen == sizeof(struct ip)) {
522 			sum = in_cksum_hdr(ip);
523 		} else {
524 			sum = in_cksum(m, hlen);
525 		}
526 	}
527 	if (sum) {
528 		IPSTAT_INC(ips_badsum);
529 		goto bad;
530 	}
531 
532 #ifdef ALTQ
533 	if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
534 		/* packet is dropped by traffic conditioner */
535 		return;
536 #endif
537 
538 	ip_len = ntohs(ip->ip_len);
539 	if (ip_len < hlen) {
540 		IPSTAT_INC(ips_badlen);
541 		goto bad;
542 	}
543 
544 	/*
545 	 * Check that the amount of data in the buffers
546 	 * is as at least much as the IP header would have us expect.
547 	 * Trim mbufs if longer than we expect.
548 	 * Drop packet if shorter than we expect.
549 	 */
550 	if (m->m_pkthdr.len < ip_len) {
551 tooshort:
552 		IPSTAT_INC(ips_tooshort);
553 		goto bad;
554 	}
555 	if (m->m_pkthdr.len > ip_len) {
556 		if (m->m_len == m->m_pkthdr.len) {
557 			m->m_len = ip_len;
558 			m->m_pkthdr.len = ip_len;
559 		} else
560 			m_adj(m, ip_len - m->m_pkthdr.len);
561 	}
562 
563 	/*
564 	 * Try to forward the packet, but if we fail continue.
565 	 * ip_tryforward() may generate redirects these days.
566 	 * XXX the logic below falling through to normal processing
567 	 * if redirects are required should be revisited as well.
568 	 * ip_tryforward() does inbound and outbound packet firewall
569 	 * processing. If firewall has decided that destination becomes
570 	 * our local address, it sets M_FASTFWD_OURS flag. In this
571 	 * case skip another inbound firewall processing and update
572 	 * ip pointer.
573 	 */
574 	if (V_ipforwarding != 0
575 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
576 	    && (!IPSEC_ENABLED(ipv4) ||
577 	    IPSEC_CAPS(ipv4, m, IPSEC_CAP_OPERABLE) == 0)
578 #endif
579 	    ) {
580 		/*
581 		 * ip_dooptions() was run so we can ignore the source route (or
582 		 * any IP options case) case for redirects in ip_tryforward().
583 		 */
584 		if ((m = ip_tryforward(m)) == NULL)
585 			return;
586 		if (m->m_flags & M_FASTFWD_OURS) {
587 			m->m_flags &= ~M_FASTFWD_OURS;
588 			ip = mtod(m, struct ip *);
589 			goto ours;
590 		}
591 	}
592 
593 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
594 	/*
595 	 * Bypass packet filtering for packets previously handled by IPsec.
596 	 */
597 	if (IPSEC_ENABLED(ipv4) &&
598 	    IPSEC_CAPS(ipv4, m, IPSEC_CAP_BYPASS_FILTER) != 0)
599 			goto passin;
600 #endif
601 
602 	/*
603 	 * Run through list of hooks for input packets.
604 	 *
605 	 * NB: Beware of the destination address changing (e.g.
606 	 *     by NAT rewriting).  When this happens, tell
607 	 *     ip_forward to do the right thing.
608 	 */
609 
610 	/* Jump over all PFIL processing if hooks are not active. */
611 	if (!PFIL_HOOKED_IN(V_inet_pfil_head))
612 		goto passin;
613 
614 	odst = ip->ip_dst;
615 	if (pfil_run_hooks(V_inet_pfil_head, &m, ifp, PFIL_IN, NULL) !=
616 	    PFIL_PASS)
617 		return;
618 	if (m == NULL)			/* consumed by filter */
619 		return;
620 
621 	ip = mtod(m, struct ip *);
622 	dchg = (odst.s_addr != ip->ip_dst.s_addr);
623 	ifp = m->m_pkthdr.rcvif;
624 
625 	if (m->m_flags & M_FASTFWD_OURS) {
626 		m->m_flags &= ~M_FASTFWD_OURS;
627 		goto ours;
628 	}
629 	if (m->m_flags & M_IP_NEXTHOP) {
630 		if (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) {
631 			/*
632 			 * Directly ship the packet on.  This allows
633 			 * forwarding packets originally destined to us
634 			 * to some other directly connected host.
635 			 */
636 			ip_forward(m, 1);
637 			return;
638 		}
639 	}
640 passin:
641 
642 	/*
643 	 * Process options and, if not destined for us,
644 	 * ship it on.  ip_dooptions returns 1 when an
645 	 * error was detected (causing an icmp message
646 	 * to be sent and the original packet to be freed).
647 	 */
648 	if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
649 		return;
650 
651         /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
652          * matter if it is destined to another node, or whether it is
653          * a multicast one, RSVP wants it! and prevents it from being forwarded
654          * anywhere else. Also checks if the rsvp daemon is running before
655 	 * grabbing the packet.
656          */
657 	if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP)
658 		goto ours;
659 
660 	/*
661 	 * Check our list of addresses, to see if the packet is for us.
662 	 * If we don't have any addresses, assume any unicast packet
663 	 * we receive might be for us (and let the upper layers deal
664 	 * with it).
665 	 */
666 	if (CK_STAILQ_EMPTY(&V_in_ifaddrhead) &&
667 	    (m->m_flags & (M_MCAST|M_BCAST)) == 0)
668 		goto ours;
669 
670 	/*
671 	 * Enable a consistency check between the destination address
672 	 * and the arrival interface for a unicast packet (the RFC 1122
673 	 * strong ES model) if IP forwarding is disabled and the packet
674 	 * is not locally generated and the packet is not subject to
675 	 * 'ipfw fwd'.
676 	 *
677 	 * XXX - Checking also should be disabled if the destination
678 	 * address is ipnat'ed to a different interface.
679 	 *
680 	 * XXX - Checking is incompatible with IP aliases added
681 	 * to the loopback interface instead of the interface where
682 	 * the packets are received.
683 	 *
684 	 * XXX - This is the case for carp vhost IPs as well so we
685 	 * insert a workaround. If the packet got here, we already
686 	 * checked with carp_iamatch() and carp_forus().
687 	 */
688 	checkif = V_ip_checkinterface && (V_ipforwarding == 0) &&
689 	    ifp != NULL && ((ifp->if_flags & IFF_LOOPBACK) == 0) &&
690 	    ifp->if_carp == NULL && (dchg == 0);
691 
692 	/*
693 	 * Check for exact addresses in the hash bucket.
694 	 */
695 	IN_IFADDR_RLOCK(&in_ifa_tracker);
696 	LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
697 		/*
698 		 * If the address matches, verify that the packet
699 		 * arrived via the correct interface if checking is
700 		 * enabled.
701 		 */
702 		if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr &&
703 		    (!checkif || ia->ia_ifp == ifp)) {
704 			counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
705 			counter_u64_add(ia->ia_ifa.ifa_ibytes,
706 			    m->m_pkthdr.len);
707 			IN_IFADDR_RUNLOCK(&in_ifa_tracker);
708 			goto ours;
709 		}
710 	}
711 	IN_IFADDR_RUNLOCK(&in_ifa_tracker);
712 
713 	/*
714 	 * Check for broadcast addresses.
715 	 *
716 	 * Only accept broadcast packets that arrive via the matching
717 	 * interface.  Reception of forwarded directed broadcasts would
718 	 * be handled via ip_forward() and ether_output() with the loopback
719 	 * into the stack for SIMPLEX interfaces handled by ether_output().
720 	 */
721 	if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
722 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
723 			if (ifa->ifa_addr->sa_family != AF_INET)
724 				continue;
725 			ia = ifatoia(ifa);
726 			if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
727 			    ip->ip_dst.s_addr) {
728 				counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
729 				counter_u64_add(ia->ia_ifa.ifa_ibytes,
730 				    m->m_pkthdr.len);
731 				goto ours;
732 			}
733 #ifdef BOOTP_COMPAT
734 			if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) {
735 				counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
736 				counter_u64_add(ia->ia_ifa.ifa_ibytes,
737 				    m->m_pkthdr.len);
738 				goto ours;
739 			}
740 #endif
741 		}
742 		ia = NULL;
743 	}
744 	if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
745 		/*
746 		 * RFC 3927 2.7: Do not forward multicast packets from
747 		 * IN_LINKLOCAL.
748 		 */
749 		if (V_ip_mrouter && !IN_LINKLOCAL(ntohl(ip->ip_src.s_addr))) {
750 			/*
751 			 * If we are acting as a multicast router, all
752 			 * incoming multicast packets are passed to the
753 			 * kernel-level multicast forwarding function.
754 			 * The packet is returned (relatively) intact; if
755 			 * ip_mforward() returns a non-zero value, the packet
756 			 * must be discarded, else it may be accepted below.
757 			 */
758 			if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) {
759 				IPSTAT_INC(ips_cantforward);
760 				m_freem(m);
761 				return;
762 			}
763 
764 			/*
765 			 * The process-level routing daemon needs to receive
766 			 * all multicast IGMP packets, whether or not this
767 			 * host belongs to their destination groups.
768 			 */
769 			if (ip->ip_p == IPPROTO_IGMP)
770 				goto ours;
771 			IPSTAT_INC(ips_forward);
772 		}
773 		/*
774 		 * Assume the packet is for us, to avoid prematurely taking
775 		 * a lock on the in_multi hash. Protocols must perform
776 		 * their own filtering and update statistics accordingly.
777 		 */
778 		goto ours;
779 	}
780 	if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
781 		goto ours;
782 	if (ip->ip_dst.s_addr == INADDR_ANY)
783 		goto ours;
784 	/* RFC 3927 2.7: Do not forward packets to or from IN_LINKLOCAL. */
785 	if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr)) ||
786 	    IN_LINKLOCAL(ntohl(ip->ip_src.s_addr))) {
787 		IPSTAT_INC(ips_cantforward);
788 		m_freem(m);
789 		return;
790 	}
791 
792 	/*
793 	 * Not for us; forward if possible and desirable.
794 	 */
795 	if (V_ipforwarding == 0) {
796 		IPSTAT_INC(ips_cantforward);
797 		m_freem(m);
798 	} else {
799 		ip_forward(m, dchg);
800 	}
801 	return;
802 
803 ours:
804 #ifdef IPSTEALTH
805 	/*
806 	 * IPSTEALTH: Process non-routing options only
807 	 * if the packet is destined for us.
808 	 */
809 	if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1))
810 		return;
811 #endif /* IPSTEALTH */
812 
813 	/*
814 	 * Attempt reassembly; if it succeeds, proceed.
815 	 * ip_reass() will return a different mbuf.
816 	 */
817 	if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) {
818 		/* XXXGL: shouldn't we save & set m_flags? */
819 		m = ip_reass(m);
820 		if (m == NULL)
821 			return;
822 		ip = mtod(m, struct ip *);
823 		/* Get the header length of the reassembled packet */
824 		hlen = ip->ip_hl << 2;
825 	}
826 
827 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
828 	if (IPSEC_ENABLED(ipv4)) {
829 		if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
830 			return;
831 	}
832 #endif /* IPSEC */
833 
834 	/*
835 	 * Switch out to protocol's input routine.
836 	 */
837 	IPSTAT_INC(ips_delivered);
838 
839 	(*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
840 	return;
841 bad:
842 	m_freem(m);
843 }
844 
845 /*
846  * IP timer processing;
847  * if a timer expires on a reassembly
848  * queue, discard it.
849  */
850 void
ip_slowtimo(void)851 ip_slowtimo(void)
852 {
853 	VNET_ITERATOR_DECL(vnet_iter);
854 
855 	VNET_LIST_RLOCK_NOSLEEP();
856 	VNET_FOREACH(vnet_iter) {
857 		CURVNET_SET(vnet_iter);
858 		ipreass_slowtimo();
859 		CURVNET_RESTORE();
860 	}
861 	VNET_LIST_RUNLOCK_NOSLEEP();
862 }
863 
864 void
ip_drain(void)865 ip_drain(void)
866 {
867 	VNET_ITERATOR_DECL(vnet_iter);
868 
869 	VNET_LIST_RLOCK_NOSLEEP();
870 	VNET_FOREACH(vnet_iter) {
871 		CURVNET_SET(vnet_iter);
872 		ipreass_drain();
873 		CURVNET_RESTORE();
874 	}
875 	VNET_LIST_RUNLOCK_NOSLEEP();
876 }
877 
878 /*
879  * The protocol to be inserted into ip_protox[] must be already registered
880  * in inetsw[], either statically or through pf_proto_register().
881  */
882 int
ipproto_register(short ipproto)883 ipproto_register(short ipproto)
884 {
885 	struct protosw *pr;
886 
887 	/* Sanity checks. */
888 	if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
889 		return (EPROTONOSUPPORT);
890 
891 	/*
892 	 * The protocol slot must not be occupied by another protocol
893 	 * already.  An index pointing to IPPROTO_RAW is unused.
894 	 */
895 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
896 	if (pr == NULL)
897 		return (EPFNOSUPPORT);
898 	if (ip_protox[ipproto] != pr - inetsw)	/* IPPROTO_RAW */
899 		return (EEXIST);
900 
901 	/* Find the protocol position in inetsw[] and set the index. */
902 	for (pr = inetdomain.dom_protosw;
903 	     pr < inetdomain.dom_protoswNPROTOSW; pr++) {
904 		if (pr->pr_domain->dom_family == PF_INET &&
905 		    pr->pr_protocol && pr->pr_protocol == ipproto) {
906 			ip_protox[pr->pr_protocol] = pr - inetsw;
907 			return (0);
908 		}
909 	}
910 	return (EPROTONOSUPPORT);
911 }
912 
913 int
ipproto_unregister(short ipproto)914 ipproto_unregister(short ipproto)
915 {
916 	struct protosw *pr;
917 
918 	/* Sanity checks. */
919 	if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
920 		return (EPROTONOSUPPORT);
921 
922 	/* Check if the protocol was indeed registered. */
923 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
924 	if (pr == NULL)
925 		return (EPFNOSUPPORT);
926 	if (ip_protox[ipproto] == pr - inetsw)  /* IPPROTO_RAW */
927 		return (ENOENT);
928 
929 	/* Reset the protocol slot to IPPROTO_RAW. */
930 	ip_protox[ipproto] = pr - inetsw;
931 	return (0);
932 }
933 
934 u_char inetctlerrmap[PRC_NCMDS] = {
935 	0,		0,		0,		0,
936 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
937 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
938 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
939 	0,		0,		EHOSTUNREACH,	0,
940 	ENOPROTOOPT,	ECONNREFUSED
941 };
942 
943 /*
944  * Forward a packet.  If some error occurs return the sender
945  * an icmp packet.  Note we can't always generate a meaningful
946  * icmp message because icmp doesn't have a large enough repertoire
947  * of codes and types.
948  *
949  * If not forwarding, just drop the packet.  This could be confusing
950  * if ipforwarding was zero but some routing protocol was advancing
951  * us as a gateway to somewhere.  However, we must let the routing
952  * protocol deal with that.
953  *
954  * The srcrt parameter indicates whether the packet is being forwarded
955  * via a source route.
956  */
957 void
ip_forward(struct mbuf * m,int srcrt)958 ip_forward(struct mbuf *m, int srcrt)
959 {
960 	struct ip *ip = mtod(m, struct ip *);
961 	struct in_ifaddr *ia;
962 	struct mbuf *mcopy;
963 	struct sockaddr_in *sin;
964 	struct in_addr dest;
965 	struct route ro;
966 	uint32_t flowid;
967 	int error, type = 0, code = 0, mtu = 0;
968 
969 	NET_EPOCH_ASSERT();
970 
971 	if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
972 		IPSTAT_INC(ips_cantforward);
973 		m_freem(m);
974 		return;
975 	}
976 	if (
977 #ifdef IPSTEALTH
978 	    V_ipstealth == 0 &&
979 #endif
980 	    ip->ip_ttl <= IPTTLDEC) {
981 		icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 0, 0);
982 		return;
983 	}
984 
985 	bzero(&ro, sizeof(ro));
986 	sin = (struct sockaddr_in *)&ro.ro_dst;
987 	sin->sin_family = AF_INET;
988 	sin->sin_len = sizeof(*sin);
989 	sin->sin_addr = ip->ip_dst;
990 	flowid = m->m_pkthdr.flowid;
991 	ro.ro_nh = fib4_lookup(M_GETFIB(m), ip->ip_dst, 0, NHR_REF, flowid);
992 	if (ro.ro_nh != NULL) {
993 		ia = ifatoia(ro.ro_nh->nh_ifa);
994 	} else
995 		ia = NULL;
996 	/*
997 	 * Save the IP header and at most 8 bytes of the payload,
998 	 * in case we need to generate an ICMP message to the src.
999 	 *
1000 	 * XXX this can be optimized a lot by saving the data in a local
1001 	 * buffer on the stack (72 bytes at most), and only allocating the
1002 	 * mbuf if really necessary. The vast majority of the packets
1003 	 * are forwarded without having to send an ICMP back (either
1004 	 * because unnecessary, or because rate limited), so we are
1005 	 * really we are wasting a lot of work here.
1006 	 *
1007 	 * We don't use m_copym() because it might return a reference
1008 	 * to a shared cluster. Both this function and ip_output()
1009 	 * assume exclusive access to the IP header in `m', so any
1010 	 * data in a cluster may change before we reach icmp_error().
1011 	 */
1012 	mcopy = m_gethdr(M_NOWAIT, m->m_type);
1013 	if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_NOWAIT)) {
1014 		/*
1015 		 * It's probably ok if the pkthdr dup fails (because
1016 		 * the deep copy of the tag chain failed), but for now
1017 		 * be conservative and just discard the copy since
1018 		 * code below may some day want the tags.
1019 		 */
1020 		m_free(mcopy);
1021 		mcopy = NULL;
1022 	}
1023 	if (mcopy != NULL) {
1024 		mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy));
1025 		mcopy->m_pkthdr.len = mcopy->m_len;
1026 		m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1027 	}
1028 #ifdef IPSTEALTH
1029 	if (V_ipstealth == 0)
1030 #endif
1031 		ip->ip_ttl -= IPTTLDEC;
1032 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1033 	if (IPSEC_ENABLED(ipv4)) {
1034 		if ((error = IPSEC_FORWARD(ipv4, m)) != 0) {
1035 			/* mbuf consumed by IPsec */
1036 			RO_NHFREE(&ro);
1037 			m_freem(mcopy);
1038 			if (error != EINPROGRESS)
1039 				IPSTAT_INC(ips_cantforward);
1040 			return;
1041 		}
1042 		/* No IPsec processing required */
1043 	}
1044 #endif /* IPSEC */
1045 	/*
1046 	 * If forwarding packet using same interface that it came in on,
1047 	 * perhaps should send a redirect to sender to shortcut a hop.
1048 	 * Only send redirect if source is sending directly to us,
1049 	 * and if packet was not source routed (or has any options).
1050 	 * Also, don't send redirect if forwarding using a default route
1051 	 * or a route modified by a redirect.
1052 	 */
1053 	dest.s_addr = 0;
1054 	if (!srcrt && V_ipsendredirects &&
1055 	    ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) {
1056 		struct nhop_object *nh;
1057 
1058 		nh = ro.ro_nh;
1059 
1060 		if (nh != NULL && ((nh->nh_flags & (NHF_REDIRECT|NHF_DEFAULT)) == 0)) {
1061 			struct in_ifaddr *nh_ia = (struct in_ifaddr *)(nh->nh_ifa);
1062 			u_long src = ntohl(ip->ip_src.s_addr);
1063 
1064 			if (nh_ia != NULL &&
1065 			    (src & nh_ia->ia_subnetmask) == nh_ia->ia_subnet) {
1066 				/* Router requirements says to only send host redirects */
1067 				type = ICMP_REDIRECT;
1068 				code = ICMP_REDIRECT_HOST;
1069 				if (nh->nh_flags & NHF_GATEWAY) {
1070 				    if (nh->gw_sa.sa_family == AF_INET)
1071 					dest.s_addr = nh->gw4_sa.sin_addr.s_addr;
1072 				    else /* Do not redirect in case gw is AF_INET6 */
1073 					type = 0;
1074 				} else
1075 					dest.s_addr = ip->ip_dst.s_addr;
1076 			}
1077 		}
1078 	}
1079 
1080 	error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
1081 
1082 	if (error == EMSGSIZE && ro.ro_nh)
1083 		mtu = ro.ro_nh->nh_mtu;
1084 	RO_NHFREE(&ro);
1085 
1086 	if (error)
1087 		IPSTAT_INC(ips_cantforward);
1088 	else {
1089 		IPSTAT_INC(ips_forward);
1090 		if (type)
1091 			IPSTAT_INC(ips_redirectsent);
1092 		else {
1093 			if (mcopy)
1094 				m_freem(mcopy);
1095 			return;
1096 		}
1097 	}
1098 	if (mcopy == NULL)
1099 		return;
1100 
1101 	switch (error) {
1102 	case 0:				/* forwarded, but need redirect */
1103 		/* type, code set above */
1104 		break;
1105 
1106 	case ENETUNREACH:
1107 	case EHOSTUNREACH:
1108 	case ENETDOWN:
1109 	case EHOSTDOWN:
1110 	default:
1111 		type = ICMP_UNREACH;
1112 		code = ICMP_UNREACH_HOST;
1113 		break;
1114 
1115 	case EMSGSIZE:
1116 		type = ICMP_UNREACH;
1117 		code = ICMP_UNREACH_NEEDFRAG;
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 		return;
1141 	}
1142 	icmp_error(mcopy, type, code, dest.s_addr, mtu);
1143 }
1144 
1145 #define	CHECK_SO_CT(sp, ct) \
1146     (((sp->so_options & SO_TIMESTAMP) && (sp->so_ts_clock == ct)) ? 1 : 0)
1147 
1148 void
ip_savecontrol(struct inpcb * inp,struct mbuf ** mp,struct ip * ip,struct mbuf * m)1149 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1150     struct mbuf *m)
1151 {
1152 	bool stamped;
1153 
1154 	stamped = false;
1155 	if ((inp->inp_socket->so_options & SO_BINTIME) ||
1156 	    CHECK_SO_CT(inp->inp_socket, SO_TS_BINTIME)) {
1157 		struct bintime boottimebin, bt;
1158 		struct timespec ts1;
1159 
1160 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1161 		    M_TSTMP)) {
1162 			mbuf_tstmp2timespec(m, &ts1);
1163 			timespec2bintime(&ts1, &bt);
1164 			getboottimebin(&boottimebin);
1165 			bintime_add(&bt, &boottimebin);
1166 		} else {
1167 			bintime(&bt);
1168 		}
1169 		*mp = sbcreatecontrol((caddr_t)&bt, sizeof(bt),
1170 		    SCM_BINTIME, SOL_SOCKET);
1171 		if (*mp != NULL) {
1172 			mp = &(*mp)->m_next;
1173 			stamped = true;
1174 		}
1175 	}
1176 	if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME_MICRO)) {
1177 		struct bintime boottimebin, bt1;
1178 		struct timespec ts1;
1179 		struct timeval tv;
1180 
1181 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1182 		    M_TSTMP)) {
1183 			mbuf_tstmp2timespec(m, &ts1);
1184 			timespec2bintime(&ts1, &bt1);
1185 			getboottimebin(&boottimebin);
1186 			bintime_add(&bt1, &boottimebin);
1187 			bintime2timeval(&bt1, &tv);
1188 		} else {
1189 			microtime(&tv);
1190 		}
1191 		*mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv),
1192 		    SCM_TIMESTAMP, SOL_SOCKET);
1193 		if (*mp != NULL) {
1194 			mp = &(*mp)->m_next;
1195 			stamped = true;
1196 		}
1197 	} else if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME)) {
1198 		struct bintime boottimebin;
1199 		struct timespec ts, ts1;
1200 
1201 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1202 		    M_TSTMP)) {
1203 			mbuf_tstmp2timespec(m, &ts);
1204 			getboottimebin(&boottimebin);
1205 			bintime2timespec(&boottimebin, &ts1);
1206 			timespecadd(&ts, &ts1, &ts);
1207 		} else {
1208 			nanotime(&ts);
1209 		}
1210 		*mp = sbcreatecontrol((caddr_t)&ts, sizeof(ts),
1211 		    SCM_REALTIME, SOL_SOCKET);
1212 		if (*mp != NULL) {
1213 			mp = &(*mp)->m_next;
1214 			stamped = true;
1215 		}
1216 	} else if (CHECK_SO_CT(inp->inp_socket, SO_TS_MONOTONIC)) {
1217 		struct timespec ts;
1218 
1219 		if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1220 		    M_TSTMP))
1221 			mbuf_tstmp2timespec(m, &ts);
1222 		else
1223 			nanouptime(&ts);
1224 		*mp = sbcreatecontrol((caddr_t)&ts, sizeof(ts),
1225 		    SCM_MONOTONIC, SOL_SOCKET);
1226 		if (*mp != NULL) {
1227 			mp = &(*mp)->m_next;
1228 			stamped = true;
1229 		}
1230 	}
1231 	if (stamped && (m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1232 	    M_TSTMP)) {
1233 		struct sock_timestamp_info sti;
1234 
1235 		bzero(&sti, sizeof(sti));
1236 		sti.st_info_flags = ST_INFO_HW;
1237 		if ((m->m_flags & M_TSTMP_HPREC) != 0)
1238 			sti.st_info_flags |= ST_INFO_HW_HPREC;
1239 		*mp = sbcreatecontrol((caddr_t)&sti, sizeof(sti), SCM_TIME_INFO,
1240 		    SOL_SOCKET);
1241 		if (*mp != NULL)
1242 			mp = &(*mp)->m_next;
1243 	}
1244 	if (inp->inp_flags & INP_RECVDSTADDR) {
1245 		*mp = sbcreatecontrol((caddr_t)&ip->ip_dst,
1246 		    sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1247 		if (*mp)
1248 			mp = &(*mp)->m_next;
1249 	}
1250 	if (inp->inp_flags & INP_RECVTTL) {
1251 		*mp = sbcreatecontrol((caddr_t)&ip->ip_ttl,
1252 		    sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
1253 		if (*mp)
1254 			mp = &(*mp)->m_next;
1255 	}
1256 #ifdef notyet
1257 	/* XXX
1258 	 * Moving these out of udp_input() made them even more broken
1259 	 * than they already were.
1260 	 */
1261 	/* options were tossed already */
1262 	if (inp->inp_flags & INP_RECVOPTS) {
1263 		*mp = sbcreatecontrol((caddr_t)opts_deleted_above,
1264 		    sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1265 		if (*mp)
1266 			mp = &(*mp)->m_next;
1267 	}
1268 	/* ip_srcroute doesn't do what we want here, need to fix */
1269 	if (inp->inp_flags & INP_RECVRETOPTS) {
1270 		*mp = sbcreatecontrol((caddr_t)ip_srcroute(m),
1271 		    sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1272 		if (*mp)
1273 			mp = &(*mp)->m_next;
1274 	}
1275 #endif
1276 	if (inp->inp_flags & INP_RECVIF) {
1277 		struct ifnet *ifp;
1278 		struct sdlbuf {
1279 			struct sockaddr_dl sdl;
1280 			u_char	pad[32];
1281 		} sdlbuf;
1282 		struct sockaddr_dl *sdp;
1283 		struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1284 
1285 		if ((ifp = m->m_pkthdr.rcvif) &&
1286 		    ifp->if_index && ifp->if_index <= V_if_index) {
1287 			sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1288 			/*
1289 			 * Change our mind and don't try copy.
1290 			 */
1291 			if (sdp->sdl_family != AF_LINK ||
1292 			    sdp->sdl_len > sizeof(sdlbuf)) {
1293 				goto makedummy;
1294 			}
1295 			bcopy(sdp, sdl2, sdp->sdl_len);
1296 		} else {
1297 makedummy:
1298 			sdl2->sdl_len =
1299 			    offsetof(struct sockaddr_dl, sdl_data[0]);
1300 			sdl2->sdl_family = AF_LINK;
1301 			sdl2->sdl_index = 0;
1302 			sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1303 		}
1304 		*mp = sbcreatecontrol((caddr_t)sdl2, sdl2->sdl_len,
1305 		    IP_RECVIF, IPPROTO_IP);
1306 		if (*mp)
1307 			mp = &(*mp)->m_next;
1308 	}
1309 	if (inp->inp_flags & INP_RECVTOS) {
1310 		*mp = sbcreatecontrol((caddr_t)&ip->ip_tos,
1311 		    sizeof(u_char), IP_RECVTOS, IPPROTO_IP);
1312 		if (*mp)
1313 			mp = &(*mp)->m_next;
1314 	}
1315 
1316 	if (inp->inp_flags2 & INP_RECVFLOWID) {
1317 		uint32_t flowid, flow_type;
1318 
1319 		flowid = m->m_pkthdr.flowid;
1320 		flow_type = M_HASHTYPE_GET(m);
1321 
1322 		/*
1323 		 * XXX should handle the failure of one or the
1324 		 * other - don't populate both?
1325 		 */
1326 		*mp = sbcreatecontrol((caddr_t) &flowid,
1327 		    sizeof(uint32_t), IP_FLOWID, IPPROTO_IP);
1328 		if (*mp)
1329 			mp = &(*mp)->m_next;
1330 		*mp = sbcreatecontrol((caddr_t) &flow_type,
1331 		    sizeof(uint32_t), IP_FLOWTYPE, IPPROTO_IP);
1332 		if (*mp)
1333 			mp = &(*mp)->m_next;
1334 	}
1335 
1336 #ifdef	RSS
1337 	if (inp->inp_flags2 & INP_RECVRSSBUCKETID) {
1338 		uint32_t flowid, flow_type;
1339 		uint32_t rss_bucketid;
1340 
1341 		flowid = m->m_pkthdr.flowid;
1342 		flow_type = M_HASHTYPE_GET(m);
1343 
1344 		if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) {
1345 			*mp = sbcreatecontrol((caddr_t) &rss_bucketid,
1346 			   sizeof(uint32_t), IP_RSSBUCKETID, IPPROTO_IP);
1347 			if (*mp)
1348 				mp = &(*mp)->m_next;
1349 		}
1350 	}
1351 #endif
1352 }
1353 
1354 /*
1355  * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
1356  * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
1357  * locking.  This code remains in ip_input.c as ip_mroute.c is optionally
1358  * compiled.
1359  */
1360 VNET_DEFINE_STATIC(int, ip_rsvp_on);
1361 VNET_DEFINE(struct socket *, ip_rsvpd);
1362 
1363 #define	V_ip_rsvp_on		VNET(ip_rsvp_on)
1364 
1365 int
ip_rsvp_init(struct socket * so)1366 ip_rsvp_init(struct socket *so)
1367 {
1368 
1369 	if (so->so_type != SOCK_RAW ||
1370 	    so->so_proto->pr_protocol != IPPROTO_RSVP)
1371 		return EOPNOTSUPP;
1372 
1373 	if (V_ip_rsvpd != NULL)
1374 		return EADDRINUSE;
1375 
1376 	V_ip_rsvpd = so;
1377 	/*
1378 	 * This may seem silly, but we need to be sure we don't over-increment
1379 	 * the RSVP counter, in case something slips up.
1380 	 */
1381 	if (!V_ip_rsvp_on) {
1382 		V_ip_rsvp_on = 1;
1383 		V_rsvp_on++;
1384 	}
1385 
1386 	return 0;
1387 }
1388 
1389 int
ip_rsvp_done(void)1390 ip_rsvp_done(void)
1391 {
1392 
1393 	V_ip_rsvpd = NULL;
1394 	/*
1395 	 * This may seem silly, but we need to be sure we don't over-decrement
1396 	 * the RSVP counter, in case something slips up.
1397 	 */
1398 	if (V_ip_rsvp_on) {
1399 		V_ip_rsvp_on = 0;
1400 		V_rsvp_on--;
1401 	}
1402 	return 0;
1403 }
1404 
1405 int
rsvp_input(struct mbuf ** mp,int * offp,int proto)1406 rsvp_input(struct mbuf **mp, int *offp, int proto)
1407 {
1408 	struct mbuf *m;
1409 
1410 	m = *mp;
1411 	*mp = NULL;
1412 
1413 	if (rsvp_input_p) { /* call the real one if loaded */
1414 		*mp = m;
1415 		rsvp_input_p(mp, offp, proto);
1416 		return (IPPROTO_DONE);
1417 	}
1418 
1419 	/* Can still get packets with rsvp_on = 0 if there is a local member
1420 	 * of the group to which the RSVP packet is addressed.  But in this
1421 	 * case we want to throw the packet away.
1422 	 */
1423 
1424 	if (!V_rsvp_on) {
1425 		m_freem(m);
1426 		return (IPPROTO_DONE);
1427 	}
1428 
1429 	if (V_ip_rsvpd != NULL) {
1430 		*mp = m;
1431 		rip_input(mp, offp, proto);
1432 		return (IPPROTO_DONE);
1433 	}
1434 	/* Drop the packet */
1435 	m_freem(m);
1436 	return (IPPROTO_DONE);
1437 }
1438