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 #include "opt_bootp.h"
37 #include "opt_ipstealth.h"
38 #include "opt_ipsec.h"
39 #include "opt_route.h"
43 #include <sys/param.h>
44 #include <sys/systm.h>
45 #include <sys/hhook.h>
47 #include <sys/malloc.h>
48 #include <sys/domain.h>
49 #include <sys/protosw.h>
50 #include <sys/socket.h>
52 #include <sys/kernel.h>
54 #include <sys/rmlock.h>
55 #include <sys/rwlock.h>
57 #include <sys/syslog.h>
58 #include <sys/sysctl.h>
61 #include <net/if_types.h>
62 #include <net/if_var.h>
63 #include <net/if_dl.h>
64 #include <net/if_private.h>
66 #include <net/route.h>
67 #include <net/route/nhop.h>
68 #include <net/netisr.h>
69 #include <net/rss_config.h>
72 #include <netinet/in.h>
73 #include <netinet/in_kdtrace.h>
74 #include <netinet/in_systm.h>
75 #include <netinet/in_var.h>
76 #include <netinet/ip.h>
77 #include <netinet/in_fib.h>
78 #include <netinet/in_pcb.h>
79 #include <netinet/ip_var.h>
80 #include <netinet/ip_encap.h>
81 #include <netinet/ip_fw.h>
82 #include <netinet/ip_icmp.h>
83 #include <netinet/igmp_var.h>
84 #include <netinet/ip_options.h>
85 #include <machine/in_cksum.h>
86 #include <netinet/ip_carp.h>
87 #include <netinet/in_rss.h>
89 #include <netinet/sctp_var.h>
92 #include <netipsec/ipsec_support.h>
94 #include <sys/socketvar.h>
96 #include <security/mac/mac_framework.h>
99 CTASSERT(sizeof(struct ip) == 20);
102 /* IP reassembly functions are defined in ip_reass.c. */
103 extern void ipreass_init(void);
104 extern void ipreass_vnet_init(void);
106 extern void ipreass_destroy(void);
109 VNET_DEFINE(int, rsvp_on);
111 VNET_DEFINE(int, ipforwarding);
112 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_VNET | CTLFLAG_RW,
113 &VNET_NAME(ipforwarding), 0,
114 "Enable IP forwarding between interfaces");
117 * Respond with an ICMP host redirect when we forward a packet out of
118 * the same interface on which it was received. See RFC 792.
120 VNET_DEFINE(int, ipsendredirects) = 1;
121 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_VNET | CTLFLAG_RW,
122 &VNET_NAME(ipsendredirects), 0,
123 "Enable sending IP redirects");
125 VNET_DEFINE_STATIC(bool, ip_strong_es) = false;
126 #define V_ip_strong_es VNET(ip_strong_es)
127 SYSCTL_BOOL(_net_inet_ip, OID_AUTO, rfc1122_strong_es,
128 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip_strong_es), false,
129 "Packet's IP destination address must match address on arrival interface");
131 VNET_DEFINE_STATIC(bool, ip_sav) = true;
132 #define V_ip_sav VNET(ip_sav)
133 SYSCTL_BOOL(_net_inet_ip, OID_AUTO, source_address_validation,
134 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip_sav), true,
135 "Drop incoming packets with source address that is a local address");
137 /* Packet filter hooks */
138 VNET_DEFINE(pfil_head_t, inet_pfil_head);
139 VNET_DEFINE(pfil_head_t, inet_local_pfil_head);
141 static struct netisr_handler ip_nh = {
143 .nh_handler = ip_input,
144 .nh_proto = NETISR_IP,
146 .nh_m2cpuid = rss_soft_m2cpuid_v4,
147 .nh_policy = NETISR_POLICY_CPU,
148 .nh_dispatch = NETISR_DISPATCH_HYBRID,
150 .nh_policy = NETISR_POLICY_FLOW,
156 * Directly dispatched frames are currently assumed
157 * to have a flowid already calculated.
159 * It should likely have something that assert it
160 * actually has valid flow details.
162 static struct netisr_handler ip_direct_nh = {
163 .nh_name = "ip_direct",
164 .nh_handler = ip_direct_input,
165 .nh_proto = NETISR_IP_DIRECT,
166 .nh_m2cpuid = rss_soft_m2cpuid_v4,
167 .nh_policy = NETISR_POLICY_CPU,
168 .nh_dispatch = NETISR_DISPATCH_HYBRID,
172 ipproto_input_t *ip_protox[IPPROTO_MAX] = {
173 [0 ... IPPROTO_MAX - 1] = rip_input };
174 ipproto_ctlinput_t *ip_ctlprotox[IPPROTO_MAX] = {
175 [0 ... IPPROTO_MAX - 1] = rip_ctlinput };
177 VNET_DEFINE(struct in_ifaddrhead, in_ifaddrhead); /* first inet address */
178 VNET_DEFINE(struct in_ifaddrhashhead *, in_ifaddrhashtbl); /* inet addr hash table */
179 VNET_DEFINE(u_long, in_ifaddrhmask); /* mask for hash table */
181 /* Make sure it is safe to use hashinit(9) on CK_LIST. */
182 CTASSERT(sizeof(struct in_ifaddrhashhead) == sizeof(LIST_HEAD(, in_addr)));
185 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
186 &ip_mtu, 0, "Default MTU");
190 VNET_DEFINE(int, ipstealth);
191 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_VNET | CTLFLAG_RW,
192 &VNET_NAME(ipstealth), 0,
193 "IP stealth mode, no TTL decrementation on forwarding");
197 * IP statistics are stored in the "array" of counter(9)s.
199 VNET_PCPUSTAT_DEFINE(struct ipstat, ipstat);
200 VNET_PCPUSTAT_SYSINIT(ipstat);
201 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, IPCTL_STATS, stats, struct ipstat, ipstat,
202 "IP statistics (struct ipstat, netinet/ip_var.h)");
205 VNET_PCPUSTAT_SYSUNINIT(ipstat);
209 * Kernel module interface for updating ipstat. The argument is an index
210 * into ipstat treated as an array.
213 kmod_ipstat_inc(int statnum)
216 counter_u64_add(VNET(ipstat)[statnum], 1);
220 kmod_ipstat_dec(int statnum)
223 counter_u64_add(VNET(ipstat)[statnum], -1);
227 sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS)
231 netisr_getqlimit(&ip_nh, &qlimit);
232 error = sysctl_handle_int(oidp, &qlimit, 0, req);
233 if (error || !req->newptr)
237 return (netisr_setqlimit(&ip_nh, qlimit));
239 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen,
240 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 0,
241 sysctl_netinet_intr_queue_maxlen, "I",
242 "Maximum size of the IP input queue");
245 sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS)
247 u_int64_t qdrops_long;
250 netisr_getqdrops(&ip_nh, &qdrops_long);
251 qdrops = qdrops_long;
252 error = sysctl_handle_int(oidp, &qdrops, 0, req);
253 if (error || !req->newptr)
257 netisr_clearqdrops(&ip_nh);
261 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops,
262 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE,
263 0, 0, sysctl_netinet_intr_queue_drops, "I",
264 "Number of packets dropped from the IP input queue");
268 sysctl_netinet_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS)
272 netisr_getqlimit(&ip_direct_nh, &qlimit);
273 error = sysctl_handle_int(oidp, &qlimit, 0, req);
274 if (error || !req->newptr)
278 return (netisr_setqlimit(&ip_direct_nh, qlimit));
280 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRDQMAXLEN, intr_direct_queue_maxlen,
281 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
282 0, 0, sysctl_netinet_intr_direct_queue_maxlen,
283 "I", "Maximum size of the IP direct input queue");
286 sysctl_netinet_intr_direct_queue_drops(SYSCTL_HANDLER_ARGS)
288 u_int64_t qdrops_long;
291 netisr_getqdrops(&ip_direct_nh, &qdrops_long);
292 qdrops = qdrops_long;
293 error = sysctl_handle_int(oidp, &qdrops, 0, req);
294 if (error || !req->newptr)
298 netisr_clearqdrops(&ip_direct_nh);
302 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRDQDROPS, intr_direct_queue_drops,
303 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, 0, 0,
304 sysctl_netinet_intr_direct_queue_drops, "I",
305 "Number of packets dropped from the IP direct input queue");
309 * IP initialization: fill in IP protocol switch table.
310 * All protocols not implemented in kernel go to raw IP protocol handler.
313 ip_vnet_init(void *arg __unused)
315 struct pfil_head_args args;
317 CK_STAILQ_INIT(&V_in_ifaddrhead);
318 V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask);
320 /* Initialize IP reassembly queue. */
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);
330 args.pa_flags = PFIL_OUT;
331 args.pa_headname = PFIL_INET_LOCAL_NAME;
332 V_inet_local_pfil_head = pfil_head_register(&args);
334 if (hhook_head_register(HHOOK_TYPE_IPSEC_IN, AF_INET,
335 &V_ipsec_hhh_in[HHOOK_IPSEC_INET],
336 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
337 printf("%s: WARNING: unable to register input helper hook\n",
339 if (hhook_head_register(HHOOK_TYPE_IPSEC_OUT, AF_INET,
340 &V_ipsec_hhh_out[HHOOK_IPSEC_INET],
341 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
342 printf("%s: WARNING: unable to register output helper hook\n",
346 netisr_register_vnet(&ip_nh);
348 netisr_register_vnet(&ip_direct_nh);
352 VNET_SYSINIT(ip_vnet_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_FOURTH,
356 ip_init(const void *unused __unused)
362 * Register statically compiled protocols, that are unlikely to
363 * ever become dynamic.
365 IPPROTO_REGISTER(IPPROTO_ICMP, icmp_input, NULL);
366 IPPROTO_REGISTER(IPPROTO_IGMP, igmp_input, NULL);
367 IPPROTO_REGISTER(IPPROTO_RSVP, rsvp_input, NULL);
368 IPPROTO_REGISTER(IPPROTO_IPV4, encap4_input, NULL);
369 IPPROTO_REGISTER(IPPROTO_MOBILE, encap4_input, NULL);
370 IPPROTO_REGISTER(IPPROTO_ETHERIP, encap4_input, NULL);
371 IPPROTO_REGISTER(IPPROTO_GRE, encap4_input, NULL);
372 IPPROTO_REGISTER(IPPROTO_IPV6, encap4_input, NULL);
373 IPPROTO_REGISTER(IPPROTO_PIM, encap4_input, NULL);
374 #ifdef SCTP /* XXX: has a loadable & static version */
375 IPPROTO_REGISTER(IPPROTO_SCTP, sctp_input, sctp_ctlinput);
378 netisr_register(&ip_nh);
380 netisr_register(&ip_direct_nh);
383 SYSINIT(ip_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip_init, NULL);
387 ip_destroy(void *unused __unused)
392 netisr_unregister_vnet(&ip_direct_nh);
394 netisr_unregister_vnet(&ip_nh);
396 pfil_head_unregister(V_inet_pfil_head);
397 error = hhook_head_deregister(V_ipsec_hhh_in[HHOOK_IPSEC_INET]);
399 printf("%s: WARNING: unable to deregister input helper hook "
400 "type HHOOK_TYPE_IPSEC_IN, id HHOOK_IPSEC_INET: "
401 "error %d returned\n", __func__, error);
403 error = hhook_head_deregister(V_ipsec_hhh_out[HHOOK_IPSEC_INET]);
405 printf("%s: WARNING: unable to deregister output helper hook "
406 "type HHOOK_TYPE_IPSEC_OUT, id HHOOK_IPSEC_INET: "
407 "error %d returned\n", __func__, error);
410 /* Remove the IPv4 addresses from all interfaces. */
413 /* Make sure the IPv4 routes are gone as well. */
414 rib_flush_routes_family(AF_INET);
416 /* Destroy IP reassembly queue. */
419 /* Cleanup in_ifaddr hash table; should be empty. */
420 hashdestroy(V_in_ifaddrhashtbl, M_IFADDR, V_in_ifaddrhmask);
423 VNET_SYSUNINIT(ip, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip_destroy, NULL);
428 * IP direct input routine.
430 * This is called when reinjecting completed fragments where
431 * all of the previous checking and book-keeping has been done.
434 ip_direct_input(struct mbuf *m)
439 ip = mtod(m, struct ip *);
440 hlen = ip->ip_hl << 2;
442 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
443 if (IPSEC_ENABLED(ipv4)) {
444 if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
448 IPSTAT_INC(ips_delivered);
449 ip_protox[ip->ip_p](&m, &hlen, ip->ip_p);
454 * Ip input routine. Checksum and byte swap header. If fragmented
455 * try to reassemble. Process options. Pass to next level.
458 ip_input(struct mbuf *m)
460 struct ip *ip = NULL;
461 struct in_ifaddr *ia = NULL;
465 uint16_t sum, ip_len;
466 int dchg = 0; /* dest changed after fw */
467 struct in_addr odst; /* original dst address */
473 if (m->m_flags & M_FASTFWD_OURS) {
474 m->m_flags &= ~M_FASTFWD_OURS;
475 /* Set up some basics that will be used later. */
476 ip = mtod(m, struct ip *);
477 hlen = ip->ip_hl << 2;
478 ip_len = ntohs(ip->ip_len);
482 IPSTAT_INC(ips_total);
484 if (__predict_false(m->m_pkthdr.len < sizeof(struct ip)))
487 if (m->m_len < sizeof(struct ip)) {
488 m = m_pullup(m, sizeof(struct ip));
489 if (__predict_false(m == NULL)) {
490 IPSTAT_INC(ips_toosmall);
494 ip = mtod(m, struct ip *);
496 if (__predict_false(ip->ip_v != IPVERSION)) {
497 IPSTAT_INC(ips_badvers);
501 hlen = ip->ip_hl << 2;
502 if (__predict_false(hlen < sizeof(struct ip))) { /* minimum header length */
503 IPSTAT_INC(ips_badhlen);
506 if (hlen > m->m_len) {
507 m = m_pullup(m, hlen);
508 if (__predict_false(m == NULL)) {
509 IPSTAT_INC(ips_badhlen);
512 ip = mtod(m, struct ip *);
515 IP_PROBE(receive, NULL, NULL, ip, m->m_pkthdr.rcvif, ip, NULL);
517 /* IN_LOOPBACK must not appear on the wire - RFC1122 */
518 ifp = m->m_pkthdr.rcvif;
519 if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) ||
520 IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) {
521 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
522 IPSTAT_INC(ips_badaddr);
526 /* The unspecified address can appear only as a src address - RFC1122 */
527 if (__predict_false(ntohl(ip->ip_dst.s_addr) == INADDR_ANY)) {
528 IPSTAT_INC(ips_badaddr);
532 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
533 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
535 if (hlen == sizeof(struct ip)) {
536 sum = in_cksum_hdr(ip);
538 sum = in_cksum(m, hlen);
541 if (__predict_false(sum)) {
542 IPSTAT_INC(ips_badsum);
546 ip_len = ntohs(ip->ip_len);
547 if (__predict_false(ip_len < hlen)) {
548 IPSTAT_INC(ips_badlen);
553 * Check that the amount of data in the buffers
554 * is as at least much as the IP header would have us expect.
555 * Trim mbufs if longer than we expect.
556 * Drop packet if shorter than we expect.
558 if (__predict_false(m->m_pkthdr.len < ip_len)) {
560 IPSTAT_INC(ips_tooshort);
563 if (m->m_pkthdr.len > ip_len) {
564 if (m->m_len == m->m_pkthdr.len) {
566 m->m_pkthdr.len = ip_len;
568 m_adj(m, ip_len - m->m_pkthdr.len);
572 * Try to forward the packet, but if we fail continue.
573 * ip_tryforward() may generate redirects these days.
574 * XXX the logic below falling through to normal processing
575 * if redirects are required should be revisited as well.
576 * ip_tryforward() does inbound and outbound packet firewall
577 * processing. If firewall has decided that destination becomes
578 * our local address, it sets M_FASTFWD_OURS flag. In this
579 * case skip another inbound firewall processing and update
582 if (V_ipforwarding != 0
583 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
584 && (!IPSEC_ENABLED(ipv4) ||
585 IPSEC_CAPS(ipv4, m, IPSEC_CAP_OPERABLE) == 0)
589 * ip_dooptions() was run so we can ignore the source route (or
590 * any IP options case) case for redirects in ip_tryforward().
592 if ((m = ip_tryforward(m)) == NULL)
594 if (m->m_flags & M_FASTFWD_OURS) {
595 m->m_flags &= ~M_FASTFWD_OURS;
596 ip = mtod(m, struct ip *);
601 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
603 * Bypass packet filtering for packets previously handled by IPsec.
605 if (IPSEC_ENABLED(ipv4) &&
606 IPSEC_CAPS(ipv4, m, IPSEC_CAP_BYPASS_FILTER) != 0)
611 * Run through list of hooks for input packets.
613 * NB: Beware of the destination address changing (e.g.
614 * by NAT rewriting). When this happens, tell
615 * ip_forward to do the right thing.
618 /* Jump over all PFIL processing if hooks are not active. */
619 if (!PFIL_HOOKED_IN(V_inet_pfil_head))
623 if (pfil_mbuf_in(V_inet_pfil_head, &m, ifp, NULL) !=
626 if (m == NULL) /* consumed by filter */
629 ip = mtod(m, struct ip *);
630 dchg = (odst.s_addr != ip->ip_dst.s_addr);
632 if (m->m_flags & M_FASTFWD_OURS) {
633 m->m_flags &= ~M_FASTFWD_OURS;
636 if (m->m_flags & M_IP_NEXTHOP) {
637 if (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) {
639 * Directly ship the packet on. This allows
640 * forwarding packets originally destined to us
641 * to some other directly connected host.
650 * Process options and, if not destined for us,
651 * ship it on. ip_dooptions returns 1 when an
652 * error was detected (causing an icmp message
653 * to be sent and the original packet to be freed).
655 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
658 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
659 * matter if it is destined to another node, or whether it is
660 * a multicast one, RSVP wants it! and prevents it from being forwarded
661 * anywhere else. Also checks if the rsvp daemon is running before
662 * grabbing the packet.
664 if (ip->ip_p == IPPROTO_RSVP && V_rsvp_on)
668 * Check our list of addresses, to see if the packet is for us.
669 * If we don't have any addresses, assume any unicast packet
670 * we receive might be for us (and let the upper layers deal
673 if (CK_STAILQ_EMPTY(&V_in_ifaddrhead) &&
674 (m->m_flags & (M_MCAST|M_BCAST)) == 0)
678 * Enable a consistency check between the destination address
679 * and the arrival interface for a unicast packet (the RFC 1122
680 * strong ES model) with a list of additional predicates:
681 * - if IP forwarding is disabled
682 * - the packet is not locally generated
683 * - the packet is not subject to 'ipfw fwd'
684 * - Interface is not running CARP. If the packet got here, we already
685 * checked it with carp_iamatch() and carp_forus().
687 strong_es = V_ip_strong_es && (V_ipforwarding == 0) &&
688 ((ifp->if_flags & IFF_LOOPBACK) == 0) &&
689 ifp->if_carp == NULL && (dchg == 0);
692 * Check for exact addresses in the hash bucket.
694 CK_LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
695 if (IA_SIN(ia)->sin_addr.s_addr != ip->ip_dst.s_addr)
699 * net.inet.ip.rfc1122_strong_es: the address matches, verify
700 * that the packet arrived via the correct interface.
702 if (__predict_false(strong_es && ia->ia_ifp != ifp)) {
703 IPSTAT_INC(ips_badaddr);
708 * net.inet.ip.source_address_validation: drop incoming
709 * packets that pretend to be ours.
711 if (V_ip_sav && !(ifp->if_flags & IFF_LOOPBACK) &&
712 __predict_false(in_localip_fib(ip->ip_src, ifp->if_fib))) {
713 IPSTAT_INC(ips_badaddr);
717 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
718 counter_u64_add(ia->ia_ifa.ifa_ibytes, m->m_pkthdr.len);
723 * Check for broadcast addresses.
725 * Only accept broadcast packets that arrive via the matching
726 * interface. Reception of forwarded directed broadcasts would
727 * be handled via ip_forward() and ether_output() with the loopback
728 * into the stack for SIMPLEX interfaces handled by ether_output().
730 if (ifp->if_flags & IFF_BROADCAST) {
731 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
732 if (ifa->ifa_addr->sa_family != AF_INET)
735 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
737 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
738 counter_u64_add(ia->ia_ifa.ifa_ibytes,
743 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) {
744 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
745 counter_u64_add(ia->ia_ifa.ifa_ibytes,
753 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
755 * RFC 3927 2.7: Do not forward multicast packets from
758 if (V_ip_mrouter && !IN_LINKLOCAL(ntohl(ip->ip_src.s_addr))) {
760 * If we are acting as a multicast router, all
761 * incoming multicast packets are passed to the
762 * kernel-level multicast forwarding function.
763 * The packet is returned (relatively) intact; if
764 * ip_mforward() returns a non-zero value, the packet
765 * must be discarded, else it may be accepted below.
767 if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) {
768 IPSTAT_INC(ips_cantforward);
774 * The process-level routing daemon needs to receive
775 * all multicast IGMP packets, whether or not this
776 * host belongs to their destination groups.
778 if (ip->ip_p == IPPROTO_IGMP) {
781 IPSTAT_INC(ips_forward);
784 * Assume the packet is for us, to avoid prematurely taking
785 * a lock on the in_multi hash. Protocols must perform
786 * their own filtering and update statistics accordingly.
790 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
792 if (ip->ip_dst.s_addr == INADDR_ANY)
794 /* RFC 3927 2.7: Do not forward packets to or from IN_LINKLOCAL. */
795 if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr)) ||
796 IN_LINKLOCAL(ntohl(ip->ip_src.s_addr))) {
797 IPSTAT_INC(ips_cantforward);
803 * Not for us; forward if possible and desirable.
805 if (V_ipforwarding == 0) {
806 IPSTAT_INC(ips_cantforward);
816 * IPSTEALTH: Process non-routing options only
817 * if the packet is destined for us.
819 if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1))
821 #endif /* IPSTEALTH */
824 * We are going to ship the packet to the local protocol stack. Call the
825 * filter again for this 'output' action, allowing redirect-like rules
826 * to adjust the source address.
828 if (PFIL_HOOKED_OUT(V_inet_local_pfil_head)) {
829 if (pfil_mbuf_out(V_inet_local_pfil_head, &m, V_loif, NULL) !=
832 if (m == NULL) /* consumed by filter */
834 ip = mtod(m, struct ip *);
838 * Attempt reassembly; if it succeeds, proceed.
839 * ip_reass() will return a different mbuf.
841 if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) {
842 /* XXXGL: shouldn't we save & set m_flags? */
846 ip = mtod(m, struct ip *);
847 /* Get the header length of the reassembled packet */
848 hlen = ip->ip_hl << 2;
851 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
852 if (IPSEC_ENABLED(ipv4)) {
853 if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
859 * Switch out to protocol's input routine.
861 IPSTAT_INC(ips_delivered);
863 ip_protox[ip->ip_p](&m, &hlen, ip->ip_p);
870 ipproto_register(uint8_t proto, ipproto_input_t input, ipproto_ctlinput_t ctl)
876 * The protocol slot must not be occupied by another protocol
877 * already. An index pointing to rip_input() is unused.
879 if (ip_protox[proto] == rip_input) {
880 ip_protox[proto] = input;
881 ip_ctlprotox[proto] = ctl;
888 ipproto_unregister(uint8_t proto)
893 if (ip_protox[proto] != rip_input) {
894 ip_protox[proto] = rip_input;
895 ip_ctlprotox[proto] = rip_ctlinput;
902 * Forward a packet. If some error occurs return the sender
903 * an icmp packet. Note we can't always generate a meaningful
904 * icmp message because icmp doesn't have a large enough repertoire
905 * of codes and types.
907 * If not forwarding, just drop the packet. This could be confusing
908 * if ipforwarding was zero but some routing protocol was advancing
909 * us as a gateway to somewhere. However, we must let the routing
910 * protocol deal with that.
912 * The srcrt parameter indicates whether the packet is being forwarded
913 * via a source route.
916 ip_forward(struct mbuf *m, int srcrt)
918 struct ip *ip = mtod(m, struct ip *);
919 struct in_ifaddr *ia;
921 struct sockaddr_in *sin;
925 int error, type = 0, code = 0, mtu = 0;
929 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
930 IPSTAT_INC(ips_cantforward);
938 ip->ip_ttl <= IPTTLDEC) {
939 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 0, 0);
943 bzero(&ro, sizeof(ro));
944 sin = (struct sockaddr_in *)&ro.ro_dst;
945 sin->sin_family = AF_INET;
946 sin->sin_len = sizeof(*sin);
947 sin->sin_addr = ip->ip_dst;
948 flowid = m->m_pkthdr.flowid;
949 ro.ro_nh = fib4_lookup(M_GETFIB(m), ip->ip_dst, 0, NHR_REF, flowid);
950 if (ro.ro_nh != NULL) {
951 ia = ifatoia(ro.ro_nh->nh_ifa);
955 * Save the IP header and at most 8 bytes of the payload,
956 * in case we need to generate an ICMP message to the src.
958 * XXX this can be optimized a lot by saving the data in a local
959 * buffer on the stack (72 bytes at most), and only allocating the
960 * mbuf if really necessary. The vast majority of the packets
961 * are forwarded without having to send an ICMP back (either
962 * because unnecessary, or because rate limited), so we are
963 * really we are wasting a lot of work here.
965 * We don't use m_copym() because it might return a reference
966 * to a shared cluster. Both this function and ip_output()
967 * assume exclusive access to the IP header in `m', so any
968 * data in a cluster may change before we reach icmp_error().
970 mcopy = m_gethdr(M_NOWAIT, m->m_type);
971 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_NOWAIT)) {
973 * It's probably ok if the pkthdr dup fails (because
974 * the deep copy of the tag chain failed), but for now
975 * be conservative and just discard the copy since
976 * code below may some day want the tags.
982 mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy));
983 mcopy->m_pkthdr.len = mcopy->m_len;
984 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
987 if (V_ipstealth == 0)
989 ip->ip_ttl -= IPTTLDEC;
990 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
991 if (IPSEC_ENABLED(ipv4)) {
992 if ((error = IPSEC_FORWARD(ipv4, m)) != 0) {
993 /* mbuf consumed by IPsec */
996 if (error != EINPROGRESS)
997 IPSTAT_INC(ips_cantforward);
1000 /* No IPsec processing required */
1004 * If forwarding packet using same interface that it came in on,
1005 * perhaps should send a redirect to sender to shortcut a hop.
1006 * Only send redirect if source is sending directly to us,
1007 * and if packet was not source routed (or has any options).
1008 * Also, don't send redirect if forwarding using a default route
1009 * or a route modified by a redirect.
1012 if (!srcrt && V_ipsendredirects &&
1013 ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) {
1014 struct nhop_object *nh;
1018 if (nh != NULL && ((nh->nh_flags & (NHF_REDIRECT|NHF_DEFAULT)) == 0)) {
1019 struct in_ifaddr *nh_ia = (struct in_ifaddr *)(nh->nh_ifa);
1020 u_long src = ntohl(ip->ip_src.s_addr);
1022 if (nh_ia != NULL &&
1023 (src & nh_ia->ia_subnetmask) == nh_ia->ia_subnet) {
1024 /* Router requirements says to only send host redirects */
1025 type = ICMP_REDIRECT;
1026 code = ICMP_REDIRECT_HOST;
1027 if (nh->nh_flags & NHF_GATEWAY) {
1028 if (nh->gw_sa.sa_family == AF_INET)
1029 dest.s_addr = nh->gw4_sa.sin_addr.s_addr;
1030 else /* Do not redirect in case gw is AF_INET6 */
1033 dest.s_addr = ip->ip_dst.s_addr;
1038 error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
1040 if (error == EMSGSIZE && ro.ro_nh)
1041 mtu = ro.ro_nh->nh_mtu;
1045 IPSTAT_INC(ips_cantforward);
1047 IPSTAT_INC(ips_forward);
1049 IPSTAT_INC(ips_redirectsent);
1060 case 0: /* forwarded, but need redirect */
1061 /* type, code set above */
1069 type = ICMP_UNREACH;
1070 code = ICMP_UNREACH_HOST;
1074 type = ICMP_UNREACH;
1075 code = ICMP_UNREACH_NEEDFRAG;
1077 * If the MTU was set before make sure we are below the
1079 * If the MTU wasn't set before use the interface mtu or
1080 * fall back to the next smaller mtu step compared to the
1081 * current packet size.
1085 mtu = min(mtu, ia->ia_ifp->if_mtu);
1088 mtu = ia->ia_ifp->if_mtu;
1090 mtu = ip_next_mtu(ntohs(ip->ip_len), 0);
1092 IPSTAT_INC(ips_cantfrag);
1096 case EACCES: /* ipfw denied packet */
1100 icmp_error(mcopy, type, code, dest.s_addr, mtu);
1103 #define CHECK_SO_CT(sp, ct) \
1104 (((sp->so_options & SO_TIMESTAMP) && (sp->so_ts_clock == ct)) ? 1 : 0)
1107 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1113 if ((inp->inp_socket->so_options & SO_BINTIME) ||
1114 CHECK_SO_CT(inp->inp_socket, SO_TS_BINTIME)) {
1115 struct bintime boottimebin, bt;
1116 struct timespec ts1;
1118 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1120 mbuf_tstmp2timespec(m, &ts1);
1121 timespec2bintime(&ts1, &bt);
1122 getboottimebin(&boottimebin);
1123 bintime_add(&bt, &boottimebin);
1127 *mp = sbcreatecontrol(&bt, sizeof(bt), SCM_BINTIME,
1128 SOL_SOCKET, M_NOWAIT);
1130 mp = &(*mp)->m_next;
1134 if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME_MICRO)) {
1135 struct bintime boottimebin, bt1;
1136 struct timespec ts1;
1139 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1141 mbuf_tstmp2timespec(m, &ts1);
1142 timespec2bintime(&ts1, &bt1);
1143 getboottimebin(&boottimebin);
1144 bintime_add(&bt1, &boottimebin);
1145 bintime2timeval(&bt1, &tv);
1149 *mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv), SCM_TIMESTAMP,
1150 SOL_SOCKET, M_NOWAIT);
1152 mp = &(*mp)->m_next;
1155 } else if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME)) {
1156 struct bintime boottimebin;
1157 struct timespec ts, ts1;
1159 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1161 mbuf_tstmp2timespec(m, &ts);
1162 getboottimebin(&boottimebin);
1163 bintime2timespec(&boottimebin, &ts1);
1164 timespecadd(&ts, &ts1, &ts);
1168 *mp = sbcreatecontrol(&ts, sizeof(ts), SCM_REALTIME,
1169 SOL_SOCKET, M_NOWAIT);
1171 mp = &(*mp)->m_next;
1174 } else if (CHECK_SO_CT(inp->inp_socket, SO_TS_MONOTONIC)) {
1177 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1179 mbuf_tstmp2timespec(m, &ts);
1182 *mp = sbcreatecontrol(&ts, sizeof(ts), SCM_MONOTONIC,
1183 SOL_SOCKET, M_NOWAIT);
1185 mp = &(*mp)->m_next;
1189 if (stamped && (m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1191 struct sock_timestamp_info sti;
1193 bzero(&sti, sizeof(sti));
1194 sti.st_info_flags = ST_INFO_HW;
1195 if ((m->m_flags & M_TSTMP_HPREC) != 0)
1196 sti.st_info_flags |= ST_INFO_HW_HPREC;
1197 *mp = sbcreatecontrol(&sti, sizeof(sti), SCM_TIME_INFO,
1198 SOL_SOCKET, M_NOWAIT);
1200 mp = &(*mp)->m_next;
1202 if (inp->inp_flags & INP_RECVDSTADDR) {
1203 *mp = sbcreatecontrol(&ip->ip_dst, sizeof(struct in_addr),
1204 IP_RECVDSTADDR, IPPROTO_IP, M_NOWAIT);
1206 mp = &(*mp)->m_next;
1208 if (inp->inp_flags & INP_RECVTTL) {
1209 *mp = sbcreatecontrol(&ip->ip_ttl, sizeof(u_char), IP_RECVTTL,
1210 IPPROTO_IP, M_NOWAIT);
1212 mp = &(*mp)->m_next;
1216 * Moving these out of udp_input() made them even more broken
1217 * than they already were.
1219 /* options were tossed already */
1220 if (inp->inp_flags & INP_RECVOPTS) {
1221 *mp = sbcreatecontrol(opts_deleted_above,
1222 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP, M_NOWAIT);
1224 mp = &(*mp)->m_next;
1226 /* ip_srcroute doesn't do what we want here, need to fix */
1227 if (inp->inp_flags & INP_RECVRETOPTS) {
1228 *mp = sbcreatecontrol(ip_srcroute(m), sizeof(struct in_addr),
1229 IP_RECVRETOPTS, IPPROTO_IP, M_NOWAIT);
1231 mp = &(*mp)->m_next;
1234 if (inp->inp_flags & INP_RECVIF) {
1237 struct sockaddr_dl sdl;
1240 struct sockaddr_dl *sdp;
1241 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1243 if ((ifp = m->m_pkthdr.rcvif)) {
1244 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1246 * Change our mind and don't try copy.
1248 if (sdp->sdl_family != AF_LINK ||
1249 sdp->sdl_len > sizeof(sdlbuf)) {
1252 bcopy(sdp, sdl2, sdp->sdl_len);
1256 offsetof(struct sockaddr_dl, sdl_data[0]);
1257 sdl2->sdl_family = AF_LINK;
1258 sdl2->sdl_index = 0;
1259 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1261 *mp = sbcreatecontrol(sdl2, sdl2->sdl_len, IP_RECVIF,
1262 IPPROTO_IP, M_NOWAIT);
1264 mp = &(*mp)->m_next;
1266 if (inp->inp_flags & INP_RECVTOS) {
1267 *mp = sbcreatecontrol(&ip->ip_tos, sizeof(u_char), IP_RECVTOS,
1268 IPPROTO_IP, M_NOWAIT);
1270 mp = &(*mp)->m_next;
1273 if (inp->inp_flags2 & INP_RECVFLOWID) {
1274 uint32_t flowid, flow_type;
1276 flowid = m->m_pkthdr.flowid;
1277 flow_type = M_HASHTYPE_GET(m);
1280 * XXX should handle the failure of one or the
1281 * other - don't populate both?
1283 *mp = sbcreatecontrol(&flowid, sizeof(uint32_t), IP_FLOWID,
1284 IPPROTO_IP, M_NOWAIT);
1286 mp = &(*mp)->m_next;
1287 *mp = sbcreatecontrol(&flow_type, sizeof(uint32_t),
1288 IP_FLOWTYPE, IPPROTO_IP, M_NOWAIT);
1290 mp = &(*mp)->m_next;
1294 if (inp->inp_flags2 & INP_RECVRSSBUCKETID) {
1295 uint32_t flowid, flow_type;
1296 uint32_t rss_bucketid;
1298 flowid = m->m_pkthdr.flowid;
1299 flow_type = M_HASHTYPE_GET(m);
1301 if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) {
1302 *mp = sbcreatecontrol(&rss_bucketid, sizeof(uint32_t),
1303 IP_RSSBUCKETID, IPPROTO_IP, M_NOWAIT);
1305 mp = &(*mp)->m_next;
1312 * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
1313 * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
1314 * locking. This code remains in ip_input.c as ip_mroute.c is optionally
1317 VNET_DEFINE_STATIC(int, ip_rsvp_on);
1318 VNET_DEFINE(struct socket *, ip_rsvpd);
1320 #define V_ip_rsvp_on VNET(ip_rsvp_on)
1323 ip_rsvp_init(struct socket *so)
1326 if (V_ip_rsvpd != NULL)
1331 * This may seem silly, but we need to be sure we don't over-increment
1332 * the RSVP counter, in case something slips up.
1334 if (!V_ip_rsvp_on) {
1348 * This may seem silly, but we need to be sure we don't over-decrement
1349 * the RSVP counter, in case something slips up.
1359 rsvp_input(struct mbuf **mp, int *offp, int proto)
1366 if (rsvp_input_p) { /* call the real one if loaded */
1368 rsvp_input_p(mp, offp, proto);
1369 return (IPPROTO_DONE);
1372 /* Can still get packets with rsvp_on = 0 if there is a local member
1373 * of the group to which the RSVP packet is addressed. But in this
1374 * case we want to throw the packet away.
1379 return (IPPROTO_DONE);
1382 if (V_ip_rsvpd != NULL) {
1384 rip_input(mp, offp, proto);
1385 return (IPPROTO_DONE);
1387 /* Drop the packet */
1389 return (IPPROTO_DONE);