2 * SPDX-License-Identifier: BSD-3-Clause
4 * Copyright (c) 1982, 1986, 1988, 1993
5 * The Regents of the University of California. All rights reserved.
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
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.
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
31 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94
34 #include <sys/cdefs.h>
35 __FBSDID("$FreeBSD$");
37 #include "opt_bootp.h"
39 #include "opt_ipstealth.h"
40 #include "opt_ipsec.h"
41 #include "opt_route.h"
45 #include <sys/param.h>
46 #include <sys/systm.h>
47 #include <sys/hhook.h>
49 #include <sys/malloc.h>
50 #include <sys/domain.h>
51 #include <sys/protosw.h>
52 #include <sys/socket.h>
54 #include <sys/kernel.h>
56 #include <sys/rmlock.h>
57 #include <sys/rwlock.h>
59 #include <sys/syslog.h>
60 #include <sys/sysctl.h>
63 #include <net/if_types.h>
64 #include <net/if_var.h>
65 #include <net/if_dl.h>
66 #include <net/if_private.h>
68 #include <net/route.h>
69 #include <net/route/nhop.h>
70 #include <net/netisr.h>
71 #include <net/rss_config.h>
74 #include <netinet/in.h>
75 #include <netinet/in_kdtrace.h>
76 #include <netinet/in_systm.h>
77 #include <netinet/in_var.h>
78 #include <netinet/ip.h>
79 #include <netinet/in_fib.h>
80 #include <netinet/in_pcb.h>
81 #include <netinet/ip_var.h>
82 #include <netinet/ip_encap.h>
83 #include <netinet/ip_fw.h>
84 #include <netinet/ip_icmp.h>
85 #include <netinet/igmp_var.h>
86 #include <netinet/ip_options.h>
87 #include <machine/in_cksum.h>
88 #include <netinet/ip_carp.h>
89 #include <netinet/in_rss.h>
91 #include <netinet/sctp_var.h>
94 #include <netipsec/ipsec_support.h>
96 #include <sys/socketvar.h>
98 #include <security/mac/mac_framework.h>
101 CTASSERT(sizeof(struct ip) == 20);
104 /* IP reassembly functions are defined in ip_reass.c. */
105 extern void ipreass_init(void);
106 extern void ipreass_vnet_init(void);
108 extern void ipreass_destroy(void);
111 VNET_DEFINE(int, rsvp_on);
113 VNET_DEFINE(int, ipforwarding);
114 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_VNET | CTLFLAG_RW,
115 &VNET_NAME(ipforwarding), 0,
116 "Enable IP forwarding between interfaces");
119 * Respond with an ICMP host redirect when we forward a packet out of
120 * the same interface on which it was received. See RFC 792.
122 VNET_DEFINE(int, ipsendredirects) = 1;
123 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_VNET | CTLFLAG_RW,
124 &VNET_NAME(ipsendredirects), 0,
125 "Enable sending IP redirects");
127 VNET_DEFINE_STATIC(bool, ip_strong_es) = false;
128 #define V_ip_strong_es VNET(ip_strong_es)
129 SYSCTL_BOOL(_net_inet_ip, OID_AUTO, rfc1122_strong_es,
130 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip_strong_es), false,
131 "Packet's IP destination address must match address on arrival interface");
133 VNET_DEFINE_STATIC(bool, ip_sav) = true;
134 #define V_ip_sav VNET(ip_sav)
135 SYSCTL_BOOL(_net_inet_ip, OID_AUTO, source_address_validation,
136 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip_sav), true,
137 "Drop incoming packets with source address that is a local address");
139 /* Packet filter hooks */
140 VNET_DEFINE(pfil_head_t, inet_pfil_head);
141 VNET_DEFINE(pfil_head_t, inet_local_pfil_head);
143 static struct netisr_handler ip_nh = {
145 .nh_handler = ip_input,
146 .nh_proto = NETISR_IP,
148 .nh_m2cpuid = rss_soft_m2cpuid_v4,
149 .nh_policy = NETISR_POLICY_CPU,
150 .nh_dispatch = NETISR_DISPATCH_HYBRID,
152 .nh_policy = NETISR_POLICY_FLOW,
158 * Directly dispatched frames are currently assumed
159 * to have a flowid already calculated.
161 * It should likely have something that assert it
162 * actually has valid flow details.
164 static struct netisr_handler ip_direct_nh = {
165 .nh_name = "ip_direct",
166 .nh_handler = ip_direct_input,
167 .nh_proto = NETISR_IP_DIRECT,
168 .nh_m2cpuid = rss_soft_m2cpuid_v4,
169 .nh_policy = NETISR_POLICY_CPU,
170 .nh_dispatch = NETISR_DISPATCH_HYBRID,
174 ipproto_input_t *ip_protox[IPPROTO_MAX] = {
175 [0 ... IPPROTO_MAX - 1] = rip_input };
176 ipproto_ctlinput_t *ip_ctlprotox[IPPROTO_MAX] = {
177 [0 ... IPPROTO_MAX - 1] = rip_ctlinput };
179 VNET_DEFINE(struct in_ifaddrhead, in_ifaddrhead); /* first inet address */
180 VNET_DEFINE(struct in_ifaddrhashhead *, in_ifaddrhashtbl); /* inet addr hash table */
181 VNET_DEFINE(u_long, in_ifaddrhmask); /* mask for hash table */
183 /* Make sure it is safe to use hashinit(9) on CK_LIST. */
184 CTASSERT(sizeof(struct in_ifaddrhashhead) == sizeof(LIST_HEAD(, in_addr)));
187 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
188 &ip_mtu, 0, "Default MTU");
192 VNET_DEFINE(int, ipstealth);
193 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_VNET | CTLFLAG_RW,
194 &VNET_NAME(ipstealth), 0,
195 "IP stealth mode, no TTL decrementation on forwarding");
199 * IP statistics are stored in the "array" of counter(9)s.
201 VNET_PCPUSTAT_DEFINE(struct ipstat, ipstat);
202 VNET_PCPUSTAT_SYSINIT(ipstat);
203 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, IPCTL_STATS, stats, struct ipstat, ipstat,
204 "IP statistics (struct ipstat, netinet/ip_var.h)");
207 VNET_PCPUSTAT_SYSUNINIT(ipstat);
211 * Kernel module interface for updating ipstat. The argument is an index
212 * into ipstat treated as an array.
215 kmod_ipstat_inc(int statnum)
218 counter_u64_add(VNET(ipstat)[statnum], 1);
222 kmod_ipstat_dec(int statnum)
225 counter_u64_add(VNET(ipstat)[statnum], -1);
229 sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS)
233 netisr_getqlimit(&ip_nh, &qlimit);
234 error = sysctl_handle_int(oidp, &qlimit, 0, req);
235 if (error || !req->newptr)
239 return (netisr_setqlimit(&ip_nh, qlimit));
241 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen,
242 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 0,
243 sysctl_netinet_intr_queue_maxlen, "I",
244 "Maximum size of the IP input queue");
247 sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS)
249 u_int64_t qdrops_long;
252 netisr_getqdrops(&ip_nh, &qdrops_long);
253 qdrops = qdrops_long;
254 error = sysctl_handle_int(oidp, &qdrops, 0, req);
255 if (error || !req->newptr)
259 netisr_clearqdrops(&ip_nh);
263 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops,
264 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE,
265 0, 0, sysctl_netinet_intr_queue_drops, "I",
266 "Number of packets dropped from the IP input queue");
270 sysctl_netinet_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS)
274 netisr_getqlimit(&ip_direct_nh, &qlimit);
275 error = sysctl_handle_int(oidp, &qlimit, 0, req);
276 if (error || !req->newptr)
280 return (netisr_setqlimit(&ip_direct_nh, qlimit));
282 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRDQMAXLEN, intr_direct_queue_maxlen,
283 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
284 0, 0, sysctl_netinet_intr_direct_queue_maxlen,
285 "I", "Maximum size of the IP direct input queue");
288 sysctl_netinet_intr_direct_queue_drops(SYSCTL_HANDLER_ARGS)
290 u_int64_t qdrops_long;
293 netisr_getqdrops(&ip_direct_nh, &qdrops_long);
294 qdrops = qdrops_long;
295 error = sysctl_handle_int(oidp, &qdrops, 0, req);
296 if (error || !req->newptr)
300 netisr_clearqdrops(&ip_direct_nh);
304 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRDQDROPS, intr_direct_queue_drops,
305 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, 0, 0,
306 sysctl_netinet_intr_direct_queue_drops, "I",
307 "Number of packets dropped from the IP direct input queue");
311 * IP initialization: fill in IP protocol switch table.
312 * All protocols not implemented in kernel go to raw IP protocol handler.
315 ip_vnet_init(void *arg __unused)
317 struct pfil_head_args args;
319 CK_STAILQ_INIT(&V_in_ifaddrhead);
320 V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask);
322 /* Initialize IP reassembly queue. */
325 /* Initialize packet filter hooks. */
326 args.pa_version = PFIL_VERSION;
327 args.pa_flags = PFIL_IN | PFIL_OUT;
328 args.pa_type = PFIL_TYPE_IP4;
329 args.pa_headname = PFIL_INET_NAME;
330 V_inet_pfil_head = pfil_head_register(&args);
332 args.pa_flags = PFIL_OUT;
333 args.pa_headname = PFIL_INET_LOCAL_NAME;
334 V_inet_local_pfil_head = pfil_head_register(&args);
336 if (hhook_head_register(HHOOK_TYPE_IPSEC_IN, AF_INET,
337 &V_ipsec_hhh_in[HHOOK_IPSEC_INET],
338 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
339 printf("%s: WARNING: unable to register input helper hook\n",
341 if (hhook_head_register(HHOOK_TYPE_IPSEC_OUT, AF_INET,
342 &V_ipsec_hhh_out[HHOOK_IPSEC_INET],
343 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
344 printf("%s: WARNING: unable to register output helper hook\n",
348 netisr_register_vnet(&ip_nh);
350 netisr_register_vnet(&ip_direct_nh);
354 VNET_SYSINIT(ip_vnet_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_FOURTH,
358 ip_init(const void *unused __unused)
364 * Register statically compiled protocols, that are unlikely to
365 * ever become dynamic.
367 IPPROTO_REGISTER(IPPROTO_ICMP, icmp_input, NULL);
368 IPPROTO_REGISTER(IPPROTO_IGMP, igmp_input, NULL);
369 IPPROTO_REGISTER(IPPROTO_RSVP, rsvp_input, NULL);
370 IPPROTO_REGISTER(IPPROTO_IPV4, encap4_input, NULL);
371 IPPROTO_REGISTER(IPPROTO_MOBILE, encap4_input, NULL);
372 IPPROTO_REGISTER(IPPROTO_ETHERIP, encap4_input, NULL);
373 IPPROTO_REGISTER(IPPROTO_GRE, encap4_input, NULL);
374 IPPROTO_REGISTER(IPPROTO_IPV6, encap4_input, NULL);
375 IPPROTO_REGISTER(IPPROTO_PIM, encap4_input, NULL);
376 #ifdef SCTP /* XXX: has a loadable & static version */
377 IPPROTO_REGISTER(IPPROTO_SCTP, sctp_input, sctp_ctlinput);
380 netisr_register(&ip_nh);
382 netisr_register(&ip_direct_nh);
385 SYSINIT(ip_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip_init, NULL);
389 ip_destroy(void *unused __unused)
394 netisr_unregister_vnet(&ip_direct_nh);
396 netisr_unregister_vnet(&ip_nh);
398 pfil_head_unregister(V_inet_pfil_head);
399 error = hhook_head_deregister(V_ipsec_hhh_in[HHOOK_IPSEC_INET]);
401 printf("%s: WARNING: unable to deregister input helper hook "
402 "type HHOOK_TYPE_IPSEC_IN, id HHOOK_IPSEC_INET: "
403 "error %d returned\n", __func__, error);
405 error = hhook_head_deregister(V_ipsec_hhh_out[HHOOK_IPSEC_INET]);
407 printf("%s: WARNING: unable to deregister output helper hook "
408 "type HHOOK_TYPE_IPSEC_OUT, id HHOOK_IPSEC_INET: "
409 "error %d returned\n", __func__, error);
412 /* Remove the IPv4 addresses from all interfaces. */
415 /* Make sure the IPv4 routes are gone as well. */
416 rib_flush_routes_family(AF_INET);
418 /* Destroy IP reassembly queue. */
421 /* Cleanup in_ifaddr hash table; should be empty. */
422 hashdestroy(V_in_ifaddrhashtbl, M_IFADDR, V_in_ifaddrhmask);
425 VNET_SYSUNINIT(ip, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip_destroy, NULL);
430 * IP direct input routine.
432 * This is called when reinjecting completed fragments where
433 * all of the previous checking and book-keeping has been done.
436 ip_direct_input(struct mbuf *m)
441 ip = mtod(m, struct ip *);
442 hlen = ip->ip_hl << 2;
444 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
445 if (IPSEC_ENABLED(ipv4)) {
446 if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
450 IPSTAT_INC(ips_delivered);
451 ip_protox[ip->ip_p](&m, &hlen, ip->ip_p);
456 * Ip input routine. Checksum and byte swap header. If fragmented
457 * try to reassemble. Process options. Pass to next level.
460 ip_input(struct mbuf *m)
462 struct ip *ip = NULL;
463 struct in_ifaddr *ia = NULL;
467 uint16_t sum, ip_len;
468 int dchg = 0; /* dest changed after fw */
469 struct in_addr odst; /* original dst address */
475 if (m->m_flags & M_FASTFWD_OURS) {
476 m->m_flags &= ~M_FASTFWD_OURS;
477 /* Set up some basics that will be used later. */
478 ip = mtod(m, struct ip *);
479 hlen = ip->ip_hl << 2;
480 ip_len = ntohs(ip->ip_len);
484 IPSTAT_INC(ips_total);
486 if (__predict_false(m->m_pkthdr.len < sizeof(struct ip)))
489 if (m->m_len < sizeof(struct ip)) {
490 m = m_pullup(m, sizeof(struct ip));
491 if (__predict_false(m == NULL)) {
492 IPSTAT_INC(ips_toosmall);
496 ip = mtod(m, struct ip *);
498 if (__predict_false(ip->ip_v != IPVERSION)) {
499 IPSTAT_INC(ips_badvers);
503 hlen = ip->ip_hl << 2;
504 if (__predict_false(hlen < sizeof(struct ip))) { /* minimum header length */
505 IPSTAT_INC(ips_badhlen);
508 if (hlen > m->m_len) {
509 m = m_pullup(m, hlen);
510 if (__predict_false(m == NULL)) {
511 IPSTAT_INC(ips_badhlen);
514 ip = mtod(m, struct ip *);
517 IP_PROBE(receive, NULL, NULL, ip, m->m_pkthdr.rcvif, ip, NULL);
519 /* IN_LOOPBACK must not appear on the wire - RFC1122 */
520 ifp = m->m_pkthdr.rcvif;
521 if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) ||
522 IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) {
523 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
524 IPSTAT_INC(ips_badaddr);
528 /* The unspecified address can appear only as a src address - RFC1122 */
529 if (__predict_false(ntohl(ip->ip_dst.s_addr) == INADDR_ANY)) {
530 IPSTAT_INC(ips_badaddr);
534 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
535 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
537 if (hlen == sizeof(struct ip)) {
538 sum = in_cksum_hdr(ip);
540 sum = in_cksum(m, hlen);
543 if (__predict_false(sum)) {
544 IPSTAT_INC(ips_badsum);
548 ip_len = ntohs(ip->ip_len);
549 if (__predict_false(ip_len < hlen)) {
550 IPSTAT_INC(ips_badlen);
555 * Check that the amount of data in the buffers
556 * is as at least much as the IP header would have us expect.
557 * Trim mbufs if longer than we expect.
558 * Drop packet if shorter than we expect.
560 if (__predict_false(m->m_pkthdr.len < ip_len)) {
562 IPSTAT_INC(ips_tooshort);
565 if (m->m_pkthdr.len > ip_len) {
566 if (m->m_len == m->m_pkthdr.len) {
568 m->m_pkthdr.len = ip_len;
570 m_adj(m, ip_len - m->m_pkthdr.len);
574 * Try to forward the packet, but if we fail continue.
575 * ip_tryforward() may generate redirects these days.
576 * XXX the logic below falling through to normal processing
577 * if redirects are required should be revisited as well.
578 * ip_tryforward() does inbound and outbound packet firewall
579 * processing. If firewall has decided that destination becomes
580 * our local address, it sets M_FASTFWD_OURS flag. In this
581 * case skip another inbound firewall processing and update
584 if (V_ipforwarding != 0
585 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
586 && (!IPSEC_ENABLED(ipv4) ||
587 IPSEC_CAPS(ipv4, m, IPSEC_CAP_OPERABLE) == 0)
591 * ip_dooptions() was run so we can ignore the source route (or
592 * any IP options case) case for redirects in ip_tryforward().
594 if ((m = ip_tryforward(m)) == NULL)
596 if (m->m_flags & M_FASTFWD_OURS) {
597 m->m_flags &= ~M_FASTFWD_OURS;
598 ip = mtod(m, struct ip *);
603 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
605 * Bypass packet filtering for packets previously handled by IPsec.
607 if (IPSEC_ENABLED(ipv4) &&
608 IPSEC_CAPS(ipv4, m, IPSEC_CAP_BYPASS_FILTER) != 0)
613 * Run through list of hooks for input packets.
615 * NB: Beware of the destination address changing (e.g.
616 * by NAT rewriting). When this happens, tell
617 * ip_forward to do the right thing.
620 /* Jump over all PFIL processing if hooks are not active. */
621 if (!PFIL_HOOKED_IN(V_inet_pfil_head))
625 if (pfil_mbuf_in(V_inet_pfil_head, &m, ifp, NULL) !=
628 if (m == NULL) /* consumed by filter */
631 ip = mtod(m, struct ip *);
632 dchg = (odst.s_addr != ip->ip_dst.s_addr);
634 if (m->m_flags & M_FASTFWD_OURS) {
635 m->m_flags &= ~M_FASTFWD_OURS;
638 if (m->m_flags & M_IP_NEXTHOP) {
639 if (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) {
641 * Directly ship the packet on. This allows
642 * forwarding packets originally destined to us
643 * to some other directly connected host.
652 * Process options and, if not destined for us,
653 * ship it on. ip_dooptions returns 1 when an
654 * error was detected (causing an icmp message
655 * to be sent and the original packet to be freed).
657 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
660 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
661 * matter if it is destined to another node, or whether it is
662 * a multicast one, RSVP wants it! and prevents it from being forwarded
663 * anywhere else. Also checks if the rsvp daemon is running before
664 * grabbing the packet.
666 if (ip->ip_p == IPPROTO_RSVP && V_rsvp_on)
670 * Check our list of addresses, to see if the packet is for us.
671 * If we don't have any addresses, assume any unicast packet
672 * we receive might be for us (and let the upper layers deal
675 if (CK_STAILQ_EMPTY(&V_in_ifaddrhead) &&
676 (m->m_flags & (M_MCAST|M_BCAST)) == 0)
680 * Enable a consistency check between the destination address
681 * and the arrival interface for a unicast packet (the RFC 1122
682 * strong ES model) with a list of additional predicates:
683 * - if IP forwarding is disabled
684 * - the packet is not locally generated
685 * - the packet is not subject to 'ipfw fwd'
686 * - Interface is not running CARP. If the packet got here, we already
687 * checked it with carp_iamatch() and carp_forus().
689 strong_es = V_ip_strong_es && (V_ipforwarding == 0) &&
690 ((ifp->if_flags & IFF_LOOPBACK) == 0) &&
691 ifp->if_carp == NULL && (dchg == 0);
694 * Check for exact addresses in the hash bucket.
696 CK_LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
697 if (IA_SIN(ia)->sin_addr.s_addr != ip->ip_dst.s_addr)
701 * net.inet.ip.rfc1122_strong_es: the address matches, verify
702 * that the packet arrived via the correct interface.
704 if (__predict_false(strong_es && ia->ia_ifp != ifp)) {
705 IPSTAT_INC(ips_badaddr);
710 * net.inet.ip.source_address_validation: drop incoming
711 * packets that pretend to be ours.
713 if (V_ip_sav && !(ifp->if_flags & IFF_LOOPBACK) &&
714 __predict_false(in_localip_fib(ip->ip_src, ifp->if_fib))) {
715 IPSTAT_INC(ips_badaddr);
719 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
720 counter_u64_add(ia->ia_ifa.ifa_ibytes, m->m_pkthdr.len);
725 * Check for broadcast addresses.
727 * Only accept broadcast packets that arrive via the matching
728 * interface. Reception of forwarded directed broadcasts would
729 * be handled via ip_forward() and ether_output() with the loopback
730 * into the stack for SIMPLEX interfaces handled by ether_output().
732 if (ifp->if_flags & IFF_BROADCAST) {
733 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
734 if (ifa->ifa_addr->sa_family != AF_INET)
737 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
739 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
740 counter_u64_add(ia->ia_ifa.ifa_ibytes,
745 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) {
746 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
747 counter_u64_add(ia->ia_ifa.ifa_ibytes,
755 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
757 * RFC 3927 2.7: Do not forward multicast packets from
760 if (V_ip_mrouter && !IN_LINKLOCAL(ntohl(ip->ip_src.s_addr))) {
762 * If we are acting as a multicast router, all
763 * incoming multicast packets are passed to the
764 * kernel-level multicast forwarding function.
765 * The packet is returned (relatively) intact; if
766 * ip_mforward() returns a non-zero value, the packet
767 * must be discarded, else it may be accepted below.
769 if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) {
770 IPSTAT_INC(ips_cantforward);
776 * The process-level routing daemon needs to receive
777 * all multicast IGMP packets, whether or not this
778 * host belongs to their destination groups.
780 if (ip->ip_p == IPPROTO_IGMP) {
783 IPSTAT_INC(ips_forward);
786 * Assume the packet is for us, to avoid prematurely taking
787 * a lock on the in_multi hash. Protocols must perform
788 * their own filtering and update statistics accordingly.
792 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
794 if (ip->ip_dst.s_addr == INADDR_ANY)
796 /* RFC 3927 2.7: Do not forward packets to or from IN_LINKLOCAL. */
797 if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr)) ||
798 IN_LINKLOCAL(ntohl(ip->ip_src.s_addr))) {
799 IPSTAT_INC(ips_cantforward);
805 * Not for us; forward if possible and desirable.
807 if (V_ipforwarding == 0) {
808 IPSTAT_INC(ips_cantforward);
818 * IPSTEALTH: Process non-routing options only
819 * if the packet is destined for us.
821 if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1))
823 #endif /* IPSTEALTH */
826 * We are going to ship the packet to the local protocol stack. Call the
827 * filter again for this 'output' action, allowing redirect-like rules
828 * to adjust the source address.
830 if (PFIL_HOOKED_OUT(V_inet_local_pfil_head)) {
831 if (pfil_mbuf_out(V_inet_local_pfil_head, &m, V_loif, NULL) !=
834 if (m == NULL) /* consumed by filter */
836 ip = mtod(m, struct ip *);
840 * Attempt reassembly; if it succeeds, proceed.
841 * ip_reass() will return a different mbuf.
843 if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) {
844 /* XXXGL: shouldn't we save & set m_flags? */
848 ip = mtod(m, struct ip *);
849 /* Get the header length of the reassembled packet */
850 hlen = ip->ip_hl << 2;
853 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
854 if (IPSEC_ENABLED(ipv4)) {
855 if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
861 * Switch out to protocol's input routine.
863 IPSTAT_INC(ips_delivered);
865 ip_protox[ip->ip_p](&m, &hlen, ip->ip_p);
872 ipproto_register(uint8_t proto, ipproto_input_t input, ipproto_ctlinput_t ctl)
878 * The protocol slot must not be occupied by another protocol
879 * already. An index pointing to rip_input() is unused.
881 if (ip_protox[proto] == rip_input) {
882 ip_protox[proto] = input;
883 ip_ctlprotox[proto] = ctl;
890 ipproto_unregister(uint8_t proto)
895 if (ip_protox[proto] != rip_input) {
896 ip_protox[proto] = rip_input;
897 ip_ctlprotox[proto] = rip_ctlinput;
904 * Forward a packet. If some error occurs return the sender
905 * an icmp packet. Note we can't always generate a meaningful
906 * icmp message because icmp doesn't have a large enough repertoire
907 * of codes and types.
909 * If not forwarding, just drop the packet. This could be confusing
910 * if ipforwarding was zero but some routing protocol was advancing
911 * us as a gateway to somewhere. However, we must let the routing
912 * protocol deal with that.
914 * The srcrt parameter indicates whether the packet is being forwarded
915 * via a source route.
918 ip_forward(struct mbuf *m, int srcrt)
920 struct ip *ip = mtod(m, struct ip *);
921 struct in_ifaddr *ia;
923 struct sockaddr_in *sin;
927 int error, type = 0, code = 0, mtu = 0;
931 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
932 IPSTAT_INC(ips_cantforward);
940 ip->ip_ttl <= IPTTLDEC) {
941 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 0, 0);
945 bzero(&ro, sizeof(ro));
946 sin = (struct sockaddr_in *)&ro.ro_dst;
947 sin->sin_family = AF_INET;
948 sin->sin_len = sizeof(*sin);
949 sin->sin_addr = ip->ip_dst;
950 flowid = m->m_pkthdr.flowid;
951 ro.ro_nh = fib4_lookup(M_GETFIB(m), ip->ip_dst, 0, NHR_REF, flowid);
952 if (ro.ro_nh != NULL) {
953 ia = ifatoia(ro.ro_nh->nh_ifa);
957 * Save the IP header and at most 8 bytes of the payload,
958 * in case we need to generate an ICMP message to the src.
960 * XXX this can be optimized a lot by saving the data in a local
961 * buffer on the stack (72 bytes at most), and only allocating the
962 * mbuf if really necessary. The vast majority of the packets
963 * are forwarded without having to send an ICMP back (either
964 * because unnecessary, or because rate limited), so we are
965 * really we are wasting a lot of work here.
967 * We don't use m_copym() because it might return a reference
968 * to a shared cluster. Both this function and ip_output()
969 * assume exclusive access to the IP header in `m', so any
970 * data in a cluster may change before we reach icmp_error().
972 mcopy = m_gethdr(M_NOWAIT, m->m_type);
973 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_NOWAIT)) {
975 * It's probably ok if the pkthdr dup fails (because
976 * the deep copy of the tag chain failed), but for now
977 * be conservative and just discard the copy since
978 * code below may some day want the tags.
984 mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy));
985 mcopy->m_pkthdr.len = mcopy->m_len;
986 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
989 if (V_ipstealth == 0)
991 ip->ip_ttl -= IPTTLDEC;
992 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
993 if (IPSEC_ENABLED(ipv4)) {
994 if ((error = IPSEC_FORWARD(ipv4, m)) != 0) {
995 /* mbuf consumed by IPsec */
998 if (error != EINPROGRESS)
999 IPSTAT_INC(ips_cantforward);
1002 /* No IPsec processing required */
1006 * If forwarding packet using same interface that it came in on,
1007 * perhaps should send a redirect to sender to shortcut a hop.
1008 * Only send redirect if source is sending directly to us,
1009 * and if packet was not source routed (or has any options).
1010 * Also, don't send redirect if forwarding using a default route
1011 * or a route modified by a redirect.
1014 if (!srcrt && V_ipsendredirects &&
1015 ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) {
1016 struct nhop_object *nh;
1020 if (nh != NULL && ((nh->nh_flags & (NHF_REDIRECT|NHF_DEFAULT)) == 0)) {
1021 struct in_ifaddr *nh_ia = (struct in_ifaddr *)(nh->nh_ifa);
1022 u_long src = ntohl(ip->ip_src.s_addr);
1024 if (nh_ia != NULL &&
1025 (src & nh_ia->ia_subnetmask) == nh_ia->ia_subnet) {
1026 /* Router requirements says to only send host redirects */
1027 type = ICMP_REDIRECT;
1028 code = ICMP_REDIRECT_HOST;
1029 if (nh->nh_flags & NHF_GATEWAY) {
1030 if (nh->gw_sa.sa_family == AF_INET)
1031 dest.s_addr = nh->gw4_sa.sin_addr.s_addr;
1032 else /* Do not redirect in case gw is AF_INET6 */
1035 dest.s_addr = ip->ip_dst.s_addr;
1040 error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
1042 if (error == EMSGSIZE && ro.ro_nh)
1043 mtu = ro.ro_nh->nh_mtu;
1047 IPSTAT_INC(ips_cantforward);
1049 IPSTAT_INC(ips_forward);
1051 IPSTAT_INC(ips_redirectsent);
1062 case 0: /* forwarded, but need redirect */
1063 /* type, code set above */
1071 type = ICMP_UNREACH;
1072 code = ICMP_UNREACH_HOST;
1076 type = ICMP_UNREACH;
1077 code = ICMP_UNREACH_NEEDFRAG;
1079 * If the MTU was set before make sure we are below the
1081 * If the MTU wasn't set before use the interface mtu or
1082 * fall back to the next smaller mtu step compared to the
1083 * current packet size.
1087 mtu = min(mtu, ia->ia_ifp->if_mtu);
1090 mtu = ia->ia_ifp->if_mtu;
1092 mtu = ip_next_mtu(ntohs(ip->ip_len), 0);
1094 IPSTAT_INC(ips_cantfrag);
1098 case EACCES: /* ipfw denied packet */
1102 icmp_error(mcopy, type, code, dest.s_addr, mtu);
1105 #define CHECK_SO_CT(sp, ct) \
1106 (((sp->so_options & SO_TIMESTAMP) && (sp->so_ts_clock == ct)) ? 1 : 0)
1109 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1115 if ((inp->inp_socket->so_options & SO_BINTIME) ||
1116 CHECK_SO_CT(inp->inp_socket, SO_TS_BINTIME)) {
1117 struct bintime boottimebin, bt;
1118 struct timespec ts1;
1120 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1122 mbuf_tstmp2timespec(m, &ts1);
1123 timespec2bintime(&ts1, &bt);
1124 getboottimebin(&boottimebin);
1125 bintime_add(&bt, &boottimebin);
1129 *mp = sbcreatecontrol(&bt, sizeof(bt), SCM_BINTIME,
1130 SOL_SOCKET, M_NOWAIT);
1132 mp = &(*mp)->m_next;
1136 if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME_MICRO)) {
1137 struct bintime boottimebin, bt1;
1138 struct timespec ts1;
1141 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1143 mbuf_tstmp2timespec(m, &ts1);
1144 timespec2bintime(&ts1, &bt1);
1145 getboottimebin(&boottimebin);
1146 bintime_add(&bt1, &boottimebin);
1147 bintime2timeval(&bt1, &tv);
1151 *mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv), SCM_TIMESTAMP,
1152 SOL_SOCKET, M_NOWAIT);
1154 mp = &(*mp)->m_next;
1157 } else if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME)) {
1158 struct bintime boottimebin;
1159 struct timespec ts, ts1;
1161 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1163 mbuf_tstmp2timespec(m, &ts);
1164 getboottimebin(&boottimebin);
1165 bintime2timespec(&boottimebin, &ts1);
1166 timespecadd(&ts, &ts1, &ts);
1170 *mp = sbcreatecontrol(&ts, sizeof(ts), SCM_REALTIME,
1171 SOL_SOCKET, M_NOWAIT);
1173 mp = &(*mp)->m_next;
1176 } else if (CHECK_SO_CT(inp->inp_socket, SO_TS_MONOTONIC)) {
1179 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1181 mbuf_tstmp2timespec(m, &ts);
1184 *mp = sbcreatecontrol(&ts, sizeof(ts), SCM_MONOTONIC,
1185 SOL_SOCKET, M_NOWAIT);
1187 mp = &(*mp)->m_next;
1191 if (stamped && (m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1193 struct sock_timestamp_info sti;
1195 bzero(&sti, sizeof(sti));
1196 sti.st_info_flags = ST_INFO_HW;
1197 if ((m->m_flags & M_TSTMP_HPREC) != 0)
1198 sti.st_info_flags |= ST_INFO_HW_HPREC;
1199 *mp = sbcreatecontrol(&sti, sizeof(sti), SCM_TIME_INFO,
1200 SOL_SOCKET, M_NOWAIT);
1202 mp = &(*mp)->m_next;
1204 if (inp->inp_flags & INP_RECVDSTADDR) {
1205 *mp = sbcreatecontrol(&ip->ip_dst, sizeof(struct in_addr),
1206 IP_RECVDSTADDR, IPPROTO_IP, M_NOWAIT);
1208 mp = &(*mp)->m_next;
1210 if (inp->inp_flags & INP_RECVTTL) {
1211 *mp = sbcreatecontrol(&ip->ip_ttl, sizeof(u_char), IP_RECVTTL,
1212 IPPROTO_IP, M_NOWAIT);
1214 mp = &(*mp)->m_next;
1218 * Moving these out of udp_input() made them even more broken
1219 * than they already were.
1221 /* options were tossed already */
1222 if (inp->inp_flags & INP_RECVOPTS) {
1223 *mp = sbcreatecontrol(opts_deleted_above,
1224 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP, M_NOWAIT);
1226 mp = &(*mp)->m_next;
1228 /* ip_srcroute doesn't do what we want here, need to fix */
1229 if (inp->inp_flags & INP_RECVRETOPTS) {
1230 *mp = sbcreatecontrol(ip_srcroute(m), sizeof(struct in_addr),
1231 IP_RECVRETOPTS, IPPROTO_IP, M_NOWAIT);
1233 mp = &(*mp)->m_next;
1236 if (inp->inp_flags & INP_RECVIF) {
1239 struct sockaddr_dl sdl;
1242 struct sockaddr_dl *sdp;
1243 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1245 if ((ifp = m->m_pkthdr.rcvif)) {
1246 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1248 * Change our mind and don't try copy.
1250 if (sdp->sdl_family != AF_LINK ||
1251 sdp->sdl_len > sizeof(sdlbuf)) {
1254 bcopy(sdp, sdl2, sdp->sdl_len);
1258 offsetof(struct sockaddr_dl, sdl_data[0]);
1259 sdl2->sdl_family = AF_LINK;
1260 sdl2->sdl_index = 0;
1261 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1263 *mp = sbcreatecontrol(sdl2, sdl2->sdl_len, IP_RECVIF,
1264 IPPROTO_IP, M_NOWAIT);
1266 mp = &(*mp)->m_next;
1268 if (inp->inp_flags & INP_RECVTOS) {
1269 *mp = sbcreatecontrol(&ip->ip_tos, sizeof(u_char), IP_RECVTOS,
1270 IPPROTO_IP, M_NOWAIT);
1272 mp = &(*mp)->m_next;
1275 if (inp->inp_flags2 & INP_RECVFLOWID) {
1276 uint32_t flowid, flow_type;
1278 flowid = m->m_pkthdr.flowid;
1279 flow_type = M_HASHTYPE_GET(m);
1282 * XXX should handle the failure of one or the
1283 * other - don't populate both?
1285 *mp = sbcreatecontrol(&flowid, sizeof(uint32_t), IP_FLOWID,
1286 IPPROTO_IP, M_NOWAIT);
1288 mp = &(*mp)->m_next;
1289 *mp = sbcreatecontrol(&flow_type, sizeof(uint32_t),
1290 IP_FLOWTYPE, IPPROTO_IP, M_NOWAIT);
1292 mp = &(*mp)->m_next;
1296 if (inp->inp_flags2 & INP_RECVRSSBUCKETID) {
1297 uint32_t flowid, flow_type;
1298 uint32_t rss_bucketid;
1300 flowid = m->m_pkthdr.flowid;
1301 flow_type = M_HASHTYPE_GET(m);
1303 if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) {
1304 *mp = sbcreatecontrol(&rss_bucketid, sizeof(uint32_t),
1305 IP_RSSBUCKETID, IPPROTO_IP, M_NOWAIT);
1307 mp = &(*mp)->m_next;
1314 * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
1315 * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
1316 * locking. This code remains in ip_input.c as ip_mroute.c is optionally
1319 VNET_DEFINE_STATIC(int, ip_rsvp_on);
1320 VNET_DEFINE(struct socket *, ip_rsvpd);
1322 #define V_ip_rsvp_on VNET(ip_rsvp_on)
1325 ip_rsvp_init(struct socket *so)
1328 if (V_ip_rsvpd != NULL)
1333 * This may seem silly, but we need to be sure we don't over-increment
1334 * the RSVP counter, in case something slips up.
1336 if (!V_ip_rsvp_on) {
1350 * This may seem silly, but we need to be sure we don't over-decrement
1351 * the RSVP counter, in case something slips up.
1361 rsvp_input(struct mbuf **mp, int *offp, int proto)
1368 if (rsvp_input_p) { /* call the real one if loaded */
1370 rsvp_input_p(mp, offp, proto);
1371 return (IPPROTO_DONE);
1374 /* Can still get packets with rsvp_on = 0 if there is a local member
1375 * of the group to which the RSVP packet is addressed. But in this
1376 * case we want to throw the packet away.
1381 return (IPPROTO_DONE);
1384 if (V_ip_rsvpd != NULL) {
1386 rip_input(mp, offp, proto);
1387 return (IPPROTO_DONE);
1389 /* Drop the packet */
1391 return (IPPROTO_DONE);