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
32 #include <sys/cdefs.h>
33 #include "opt_bootp.h"
35 #include "opt_ipstealth.h"
36 #include "opt_ipsec.h"
37 #include "opt_route.h"
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/hhook.h>
45 #include <sys/malloc.h>
46 #include <sys/domain.h>
47 #include <sys/protosw.h>
48 #include <sys/socket.h>
50 #include <sys/kernel.h>
52 #include <sys/rmlock.h>
53 #include <sys/rwlock.h>
55 #include <sys/syslog.h>
56 #include <sys/sysctl.h>
59 #include <net/if_types.h>
60 #include <net/if_var.h>
61 #include <net/if_dl.h>
62 #include <net/if_private.h>
64 #include <net/route.h>
65 #include <net/route/nhop.h>
66 #include <net/netisr.h>
67 #include <net/rss_config.h>
70 #include <netinet/in.h>
71 #include <netinet/in_kdtrace.h>
72 #include <netinet/in_systm.h>
73 #include <netinet/in_var.h>
74 #include <netinet/ip.h>
75 #include <netinet/in_fib.h>
76 #include <netinet/in_pcb.h>
77 #include <netinet/ip_var.h>
78 #include <netinet/ip_encap.h>
79 #include <netinet/ip_fw.h>
80 #include <netinet/ip_icmp.h>
81 #include <netinet/igmp_var.h>
82 #include <netinet/ip_options.h>
83 #include <machine/in_cksum.h>
84 #include <netinet/ip_carp.h>
85 #include <netinet/in_rss.h>
87 #include <netinet/sctp_var.h>
90 #include <netipsec/ipsec_support.h>
92 #include <sys/socketvar.h>
94 #include <security/mac/mac_framework.h>
97 CTASSERT(sizeof(struct ip) == 20);
100 /* IP reassembly functions are defined in ip_reass.c. */
101 extern void ipreass_init(void);
102 extern void ipreass_vnet_init(void);
104 extern void ipreass_destroy(void);
107 VNET_DEFINE(int, rsvp_on);
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");
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.
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");
123 VNET_DEFINE_STATIC(bool, ip_strong_es) = false;
124 #define V_ip_strong_es VNET(ip_strong_es)
125 SYSCTL_BOOL(_net_inet_ip, OID_AUTO, rfc1122_strong_es,
126 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip_strong_es), false,
127 "Packet's IP destination address must match address on arrival interface");
129 VNET_DEFINE_STATIC(bool, ip_sav) = true;
130 #define V_ip_sav VNET(ip_sav)
131 SYSCTL_BOOL(_net_inet_ip, OID_AUTO, source_address_validation,
132 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip_sav), true,
133 "Drop incoming packets with source address that is a local address");
135 /* Packet filter hooks */
136 VNET_DEFINE(pfil_head_t, inet_pfil_head);
137 VNET_DEFINE(pfil_head_t, inet_local_pfil_head);
139 static struct netisr_handler ip_nh = {
141 .nh_handler = ip_input,
142 .nh_proto = NETISR_IP,
144 .nh_m2cpuid = rss_soft_m2cpuid_v4,
145 .nh_policy = NETISR_POLICY_CPU,
146 .nh_dispatch = NETISR_DISPATCH_HYBRID,
148 .nh_policy = NETISR_POLICY_FLOW,
154 * Directly dispatched frames are currently assumed
155 * to have a flowid already calculated.
157 * It should likely have something that assert it
158 * actually has valid flow details.
160 static struct netisr_handler ip_direct_nh = {
161 .nh_name = "ip_direct",
162 .nh_handler = ip_direct_input,
163 .nh_proto = NETISR_IP_DIRECT,
164 .nh_m2cpuid = rss_soft_m2cpuid_v4,
165 .nh_policy = NETISR_POLICY_CPU,
166 .nh_dispatch = NETISR_DISPATCH_HYBRID,
170 ipproto_input_t *ip_protox[IPPROTO_MAX] = {
171 [0 ... IPPROTO_MAX - 1] = rip_input };
172 ipproto_ctlinput_t *ip_ctlprotox[IPPROTO_MAX] = {
173 [0 ... IPPROTO_MAX - 1] = rip_ctlinput };
175 VNET_DEFINE(struct in_ifaddrhead, in_ifaddrhead); /* first inet address */
176 VNET_DEFINE(struct in_ifaddrhashhead *, in_ifaddrhashtbl); /* inet addr hash table */
177 VNET_DEFINE(u_long, in_ifaddrhmask); /* mask for hash table */
179 /* Make sure it is safe to use hashinit(9) on CK_LIST. */
180 CTASSERT(sizeof(struct in_ifaddrhashhead) == sizeof(LIST_HEAD(, in_addr)));
183 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
184 &ip_mtu, 0, "Default MTU");
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");
195 * IP statistics are stored in the "array" of counter(9)s.
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)");
203 VNET_PCPUSTAT_SYSUNINIT(ipstat);
207 * Kernel module interface for updating ipstat. The argument is an index
208 * into ipstat treated as an array.
211 kmod_ipstat_inc(int statnum)
214 counter_u64_add(VNET(ipstat)[statnum], 1);
218 kmod_ipstat_dec(int statnum)
221 counter_u64_add(VNET(ipstat)[statnum], -1);
225 sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS)
229 netisr_getqlimit(&ip_nh, &qlimit);
230 error = sysctl_handle_int(oidp, &qlimit, 0, req);
231 if (error || !req->newptr)
235 return (netisr_setqlimit(&ip_nh, qlimit));
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");
243 sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS)
245 u_int64_t qdrops_long;
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)
255 netisr_clearqdrops(&ip_nh);
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");
266 sysctl_netinet_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS)
270 netisr_getqlimit(&ip_direct_nh, &qlimit);
271 error = sysctl_handle_int(oidp, &qlimit, 0, req);
272 if (error || !req->newptr)
276 return (netisr_setqlimit(&ip_direct_nh, qlimit));
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");
284 sysctl_netinet_intr_direct_queue_drops(SYSCTL_HANDLER_ARGS)
286 u_int64_t qdrops_long;
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)
296 netisr_clearqdrops(&ip_direct_nh);
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");
307 * IP initialization: fill in IP protocol switch table.
308 * All protocols not implemented in kernel go to raw IP protocol handler.
311 ip_vnet_init(void *arg __unused)
313 struct pfil_head_args args;
315 CK_STAILQ_INIT(&V_in_ifaddrhead);
316 V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask);
318 /* Initialize IP reassembly queue. */
321 /* Initialize packet filter hooks. */
322 args.pa_version = PFIL_VERSION;
323 args.pa_flags = PFIL_IN | PFIL_OUT;
324 args.pa_type = PFIL_TYPE_IP4;
325 args.pa_headname = PFIL_INET_NAME;
326 V_inet_pfil_head = pfil_head_register(&args);
328 args.pa_flags = PFIL_OUT;
329 args.pa_headname = PFIL_INET_LOCAL_NAME;
330 V_inet_local_pfil_head = pfil_head_register(&args);
332 if (hhook_head_register(HHOOK_TYPE_IPSEC_IN, AF_INET,
333 &V_ipsec_hhh_in[HHOOK_IPSEC_INET],
334 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
335 printf("%s: WARNING: unable to register input helper hook\n",
337 if (hhook_head_register(HHOOK_TYPE_IPSEC_OUT, AF_INET,
338 &V_ipsec_hhh_out[HHOOK_IPSEC_INET],
339 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
340 printf("%s: WARNING: unable to register output helper hook\n",
344 netisr_register_vnet(&ip_nh);
346 netisr_register_vnet(&ip_direct_nh);
350 VNET_SYSINIT(ip_vnet_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_FOURTH,
354 ip_init(const void *unused __unused)
360 * Register statically compiled protocols, that are unlikely to
361 * ever become dynamic.
363 IPPROTO_REGISTER(IPPROTO_ICMP, icmp_input, NULL);
364 IPPROTO_REGISTER(IPPROTO_IGMP, igmp_input, NULL);
365 IPPROTO_REGISTER(IPPROTO_RSVP, rsvp_input, NULL);
366 IPPROTO_REGISTER(IPPROTO_IPV4, encap4_input, NULL);
367 IPPROTO_REGISTER(IPPROTO_MOBILE, encap4_input, NULL);
368 IPPROTO_REGISTER(IPPROTO_ETHERIP, encap4_input, NULL);
369 IPPROTO_REGISTER(IPPROTO_GRE, encap4_input, NULL);
370 IPPROTO_REGISTER(IPPROTO_IPV6, encap4_input, NULL);
371 IPPROTO_REGISTER(IPPROTO_PIM, encap4_input, NULL);
372 #ifdef SCTP /* XXX: has a loadable & static version */
373 IPPROTO_REGISTER(IPPROTO_SCTP, sctp_input, sctp_ctlinput);
376 netisr_register(&ip_nh);
378 netisr_register(&ip_direct_nh);
381 SYSINIT(ip_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip_init, NULL);
385 ip_destroy(void *unused __unused)
390 netisr_unregister_vnet(&ip_direct_nh);
392 netisr_unregister_vnet(&ip_nh);
394 pfil_head_unregister(V_inet_pfil_head);
395 error = hhook_head_deregister(V_ipsec_hhh_in[HHOOK_IPSEC_INET]);
397 printf("%s: WARNING: unable to deregister input helper hook "
398 "type HHOOK_TYPE_IPSEC_IN, id HHOOK_IPSEC_INET: "
399 "error %d returned\n", __func__, error);
401 error = hhook_head_deregister(V_ipsec_hhh_out[HHOOK_IPSEC_INET]);
403 printf("%s: WARNING: unable to deregister output helper hook "
404 "type HHOOK_TYPE_IPSEC_OUT, id HHOOK_IPSEC_INET: "
405 "error %d returned\n", __func__, error);
408 /* Remove the IPv4 addresses from all interfaces. */
411 /* Make sure the IPv4 routes are gone as well. */
412 rib_flush_routes_family(AF_INET);
414 /* Destroy IP reassembly queue. */
417 /* Cleanup in_ifaddr hash table; should be empty. */
418 hashdestroy(V_in_ifaddrhashtbl, M_IFADDR, V_in_ifaddrhmask);
421 VNET_SYSUNINIT(ip, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip_destroy, NULL);
426 * IP direct input routine.
428 * This is called when reinjecting completed fragments where
429 * all of the previous checking and book-keeping has been done.
432 ip_direct_input(struct mbuf *m)
437 ip = mtod(m, struct ip *);
438 hlen = ip->ip_hl << 2;
440 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
441 if (IPSEC_ENABLED(ipv4)) {
442 if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
446 IPSTAT_INC(ips_delivered);
447 ip_protox[ip->ip_p](&m, &hlen, ip->ip_p);
452 * Ip input routine. Checksum and byte swap header. If fragmented
453 * try to reassemble. Process options. Pass to next level.
456 ip_input(struct mbuf *m)
458 struct ip *ip = NULL;
459 struct in_ifaddr *ia = NULL;
463 uint16_t sum, ip_len;
464 int dchg = 0; /* dest changed after fw */
465 struct in_addr odst; /* original dst address */
471 if (m->m_flags & M_FASTFWD_OURS) {
472 m->m_flags &= ~M_FASTFWD_OURS;
473 /* Set up some basics that will be used later. */
474 ip = mtod(m, struct ip *);
475 hlen = ip->ip_hl << 2;
476 ip_len = ntohs(ip->ip_len);
480 IPSTAT_INC(ips_total);
482 if (__predict_false(m->m_pkthdr.len < sizeof(struct ip)))
485 if (m->m_len < sizeof(struct ip)) {
486 m = m_pullup(m, sizeof(struct ip));
487 if (__predict_false(m == NULL)) {
488 IPSTAT_INC(ips_toosmall);
492 ip = mtod(m, struct ip *);
494 if (__predict_false(ip->ip_v != IPVERSION)) {
495 IPSTAT_INC(ips_badvers);
499 hlen = ip->ip_hl << 2;
500 if (__predict_false(hlen < sizeof(struct ip))) { /* minimum header length */
501 IPSTAT_INC(ips_badhlen);
504 if (hlen > m->m_len) {
505 m = m_pullup(m, hlen);
506 if (__predict_false(m == NULL)) {
507 IPSTAT_INC(ips_badhlen);
510 ip = mtod(m, struct ip *);
513 IP_PROBE(receive, NULL, NULL, ip, m->m_pkthdr.rcvif, ip, NULL);
515 /* IN_LOOPBACK must not appear on the wire - RFC1122 */
516 ifp = m->m_pkthdr.rcvif;
517 if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) ||
518 IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) {
519 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
520 IPSTAT_INC(ips_badaddr);
524 /* The unspecified address can appear only as a src address - RFC1122 */
525 if (__predict_false(ntohl(ip->ip_dst.s_addr) == INADDR_ANY)) {
526 IPSTAT_INC(ips_badaddr);
530 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
531 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
533 if (hlen == sizeof(struct ip)) {
534 sum = in_cksum_hdr(ip);
536 sum = in_cksum(m, hlen);
539 if (__predict_false(sum)) {
540 IPSTAT_INC(ips_badsum);
544 ip_len = ntohs(ip->ip_len);
545 if (__predict_false(ip_len < hlen)) {
546 IPSTAT_INC(ips_badlen);
551 * Check that the amount of data in the buffers
552 * is as at least much as the IP header would have us expect.
553 * Trim mbufs if longer than we expect.
554 * Drop packet if shorter than we expect.
556 if (__predict_false(m->m_pkthdr.len < ip_len)) {
558 IPSTAT_INC(ips_tooshort);
561 if (m->m_pkthdr.len > ip_len) {
562 if (m->m_len == m->m_pkthdr.len) {
564 m->m_pkthdr.len = ip_len;
566 m_adj(m, ip_len - m->m_pkthdr.len);
570 * Try to forward the packet, but if we fail continue.
571 * ip_tryforward() may generate redirects these days.
572 * XXX the logic below falling through to normal processing
573 * if redirects are required should be revisited as well.
574 * ip_tryforward() does inbound and outbound packet firewall
575 * processing. If firewall has decided that destination becomes
576 * our local address, it sets M_FASTFWD_OURS flag. In this
577 * case skip another inbound firewall processing and update
580 if (V_ipforwarding != 0
581 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
582 && (!IPSEC_ENABLED(ipv4) ||
583 IPSEC_CAPS(ipv4, m, IPSEC_CAP_OPERABLE) == 0)
587 * ip_dooptions() was run so we can ignore the source route (or
588 * any IP options case) case for redirects in ip_tryforward().
590 if ((m = ip_tryforward(m)) == NULL)
592 if (m->m_flags & M_FASTFWD_OURS) {
593 m->m_flags &= ~M_FASTFWD_OURS;
594 ip = mtod(m, struct ip *);
599 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
601 * Bypass packet filtering for packets previously handled by IPsec.
603 if (IPSEC_ENABLED(ipv4) &&
604 IPSEC_CAPS(ipv4, m, IPSEC_CAP_BYPASS_FILTER) != 0)
609 * Run through list of hooks for input packets.
611 * NB: Beware of the destination address changing (e.g.
612 * by NAT rewriting). When this happens, tell
613 * ip_forward to do the right thing.
616 /* Jump over all PFIL processing if hooks are not active. */
617 if (!PFIL_HOOKED_IN(V_inet_pfil_head))
621 if (pfil_mbuf_in(V_inet_pfil_head, &m, ifp, NULL) !=
625 ip = mtod(m, struct ip *);
626 dchg = (odst.s_addr != ip->ip_dst.s_addr);
628 if (m->m_flags & M_FASTFWD_OURS) {
629 m->m_flags &= ~M_FASTFWD_OURS;
632 if (m->m_flags & M_IP_NEXTHOP) {
633 if (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) {
635 * Directly ship the packet on. This allows
636 * forwarding packets originally destined to us
637 * to some other directly connected host.
646 * Process options and, if not destined for us,
647 * ship it on. ip_dooptions returns 1 when an
648 * error was detected (causing an icmp message
649 * to be sent and the original packet to be freed).
651 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
654 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
655 * matter if it is destined to another node, or whether it is
656 * a multicast one, RSVP wants it! and prevents it from being forwarded
657 * anywhere else. Also checks if the rsvp daemon is running before
658 * grabbing the packet.
660 if (ip->ip_p == IPPROTO_RSVP && V_rsvp_on)
664 * Check our list of addresses, to see if the packet is for us.
665 * If we don't have any addresses, assume any unicast packet
666 * we receive might be for us (and let the upper layers deal
669 if (CK_STAILQ_EMPTY(&V_in_ifaddrhead) &&
670 (m->m_flags & (M_MCAST|M_BCAST)) == 0)
674 * Enable a consistency check between the destination address
675 * and the arrival interface for a unicast packet (the RFC 1122
676 * strong ES model) with a list of additional predicates:
677 * - if IP forwarding is disabled
678 * - the packet is not locally generated
679 * - the packet is not subject to 'ipfw fwd'
680 * - Interface is not running CARP. If the packet got here, we already
681 * checked it with carp_iamatch() and carp_forus().
683 strong_es = V_ip_strong_es && (V_ipforwarding == 0) &&
684 ((ifp->if_flags & IFF_LOOPBACK) == 0) &&
685 ifp->if_carp == NULL && (dchg == 0);
688 * Check for exact addresses in the hash bucket.
690 CK_LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
691 if (IA_SIN(ia)->sin_addr.s_addr != ip->ip_dst.s_addr)
695 * net.inet.ip.rfc1122_strong_es: the address matches, verify
696 * that the packet arrived via the correct interface.
698 if (__predict_false(strong_es && ia->ia_ifp != ifp)) {
699 IPSTAT_INC(ips_badaddr);
704 * net.inet.ip.source_address_validation: drop incoming
705 * packets that pretend to be ours.
707 if (V_ip_sav && !(ifp->if_flags & IFF_LOOPBACK) &&
708 __predict_false(in_localip_fib(ip->ip_src, ifp->if_fib))) {
709 IPSTAT_INC(ips_badaddr);
713 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
714 counter_u64_add(ia->ia_ifa.ifa_ibytes, m->m_pkthdr.len);
719 * Check for broadcast addresses.
721 * Only accept broadcast packets that arrive via the matching
722 * interface. Reception of forwarded directed broadcasts would
723 * be handled via ip_forward() and ether_output() with the loopback
724 * into the stack for SIMPLEX interfaces handled by ether_output().
726 if (ifp->if_flags & IFF_BROADCAST) {
727 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
728 if (ifa->ifa_addr->sa_family != AF_INET)
731 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
733 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
734 counter_u64_add(ia->ia_ifa.ifa_ibytes,
739 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) {
740 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
741 counter_u64_add(ia->ia_ifa.ifa_ibytes,
749 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
751 * RFC 3927 2.7: Do not forward multicast packets from
754 if (V_ip_mrouter && !IN_LINKLOCAL(ntohl(ip->ip_src.s_addr))) {
756 * If we are acting as a multicast router, all
757 * incoming multicast packets are passed to the
758 * kernel-level multicast forwarding function.
759 * The packet is returned (relatively) intact; if
760 * ip_mforward() returns a non-zero value, the packet
761 * must be discarded, else it may be accepted below.
763 if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) {
764 IPSTAT_INC(ips_cantforward);
770 * The process-level routing daemon needs to receive
771 * all multicast IGMP packets, whether or not this
772 * host belongs to their destination groups.
774 if (ip->ip_p == IPPROTO_IGMP) {
777 IPSTAT_INC(ips_forward);
780 * Assume the packet is for us, to avoid prematurely taking
781 * a lock on the in_multi hash. Protocols must perform
782 * their own filtering and update statistics accordingly.
786 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
788 if (ip->ip_dst.s_addr == INADDR_ANY)
790 /* RFC 3927 2.7: Do not forward packets to or from IN_LINKLOCAL. */
791 if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr)) ||
792 IN_LINKLOCAL(ntohl(ip->ip_src.s_addr))) {
793 IPSTAT_INC(ips_cantforward);
799 * Not for us; forward if possible and desirable.
801 if (V_ipforwarding == 0) {
802 IPSTAT_INC(ips_cantforward);
812 * IPSTEALTH: Process non-routing options only
813 * if the packet is destined for us.
815 if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1))
817 #endif /* IPSTEALTH */
820 * We are going to ship the packet to the local protocol stack. Call the
821 * filter again for this 'output' action, allowing redirect-like rules
822 * to adjust the source address.
824 if (PFIL_HOOKED_OUT(V_inet_local_pfil_head)) {
825 if (pfil_mbuf_out(V_inet_local_pfil_head, &m, V_loif, NULL) !=
828 ip = mtod(m, struct ip *);
832 * Attempt reassembly; if it succeeds, proceed.
833 * ip_reass() will return a different mbuf.
835 if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) {
836 /* XXXGL: shouldn't we save & set m_flags? */
840 ip = mtod(m, struct ip *);
841 /* Get the header length of the reassembled packet */
842 hlen = ip->ip_hl << 2;
845 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
846 if (IPSEC_ENABLED(ipv4)) {
847 if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
853 * Switch out to protocol's input routine.
855 IPSTAT_INC(ips_delivered);
857 ip_protox[ip->ip_p](&m, &hlen, ip->ip_p);
864 ipproto_register(uint8_t proto, ipproto_input_t input, ipproto_ctlinput_t ctl)
870 * The protocol slot must not be occupied by another protocol
871 * already. An index pointing to rip_input() is unused.
873 if (ip_protox[proto] == rip_input) {
874 ip_protox[proto] = input;
875 ip_ctlprotox[proto] = ctl;
882 ipproto_unregister(uint8_t proto)
887 if (ip_protox[proto] != rip_input) {
888 ip_protox[proto] = rip_input;
889 ip_ctlprotox[proto] = rip_ctlinput;
896 * Forward a packet. If some error occurs return the sender
897 * an icmp packet. Note we can't always generate a meaningful
898 * icmp message because icmp doesn't have a large enough repertoire
899 * of codes and types.
901 * If not forwarding, just drop the packet. This could be confusing
902 * if ipforwarding was zero but some routing protocol was advancing
903 * us as a gateway to somewhere. However, we must let the routing
904 * protocol deal with that.
906 * The srcrt parameter indicates whether the packet is being forwarded
907 * via a source route.
910 ip_forward(struct mbuf *m, int srcrt)
912 struct ip *ip = mtod(m, struct ip *);
913 struct in_ifaddr *ia;
915 struct sockaddr_in *sin;
919 int error, type = 0, code = 0, mtu = 0;
923 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
924 IPSTAT_INC(ips_cantforward);
932 ip->ip_ttl <= IPTTLDEC) {
933 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 0, 0);
937 bzero(&ro, sizeof(ro));
938 sin = (struct sockaddr_in *)&ro.ro_dst;
939 sin->sin_family = AF_INET;
940 sin->sin_len = sizeof(*sin);
941 sin->sin_addr = ip->ip_dst;
942 flowid = m->m_pkthdr.flowid;
943 ro.ro_nh = fib4_lookup(M_GETFIB(m), ip->ip_dst, 0, NHR_REF, flowid);
944 if (ro.ro_nh != NULL) {
945 ia = ifatoia(ro.ro_nh->nh_ifa);
949 * Save the IP header and at most 8 bytes of the payload,
950 * in case we need to generate an ICMP message to the src.
952 * XXX this can be optimized a lot by saving the data in a local
953 * buffer on the stack (72 bytes at most), and only allocating the
954 * mbuf if really necessary. The vast majority of the packets
955 * are forwarded without having to send an ICMP back (either
956 * because unnecessary, or because rate limited), so we are
957 * really we are wasting a lot of work here.
959 * We don't use m_copym() because it might return a reference
960 * to a shared cluster. Both this function and ip_output()
961 * assume exclusive access to the IP header in `m', so any
962 * data in a cluster may change before we reach icmp_error().
964 mcopy = m_gethdr(M_NOWAIT, m->m_type);
965 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_NOWAIT)) {
967 * It's probably ok if the pkthdr dup fails (because
968 * the deep copy of the tag chain failed), but for now
969 * be conservative and just discard the copy since
970 * code below may some day want the tags.
976 mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy));
977 mcopy->m_pkthdr.len = mcopy->m_len;
978 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
981 if (V_ipstealth == 0)
983 ip->ip_ttl -= IPTTLDEC;
984 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
985 if (IPSEC_ENABLED(ipv4)) {
986 if ((error = IPSEC_FORWARD(ipv4, m)) != 0) {
987 /* mbuf consumed by IPsec */
990 if (error != EINPROGRESS)
991 IPSTAT_INC(ips_cantforward);
994 /* No IPsec processing required */
998 * If forwarding packet using same interface that it came in on,
999 * perhaps should send a redirect to sender to shortcut a hop.
1000 * Only send redirect if source is sending directly to us,
1001 * and if packet was not source routed (or has any options).
1002 * Also, don't send redirect if forwarding using a default route
1003 * or a route modified by a redirect.
1006 if (!srcrt && V_ipsendredirects &&
1007 ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) {
1008 struct nhop_object *nh;
1012 if (nh != NULL && ((nh->nh_flags & (NHF_REDIRECT|NHF_DEFAULT)) == 0)) {
1013 struct in_ifaddr *nh_ia = (struct in_ifaddr *)(nh->nh_ifa);
1014 u_long src = ntohl(ip->ip_src.s_addr);
1016 if (nh_ia != NULL &&
1017 (src & nh_ia->ia_subnetmask) == nh_ia->ia_subnet) {
1018 /* Router requirements says to only send host redirects */
1019 type = ICMP_REDIRECT;
1020 code = ICMP_REDIRECT_HOST;
1021 if (nh->nh_flags & NHF_GATEWAY) {
1022 if (nh->gw_sa.sa_family == AF_INET)
1023 dest.s_addr = nh->gw4_sa.sin_addr.s_addr;
1024 else /* Do not redirect in case gw is AF_INET6 */
1027 dest.s_addr = ip->ip_dst.s_addr;
1032 error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
1034 if (error == EMSGSIZE && ro.ro_nh)
1035 mtu = ro.ro_nh->nh_mtu;
1039 IPSTAT_INC(ips_cantforward);
1041 IPSTAT_INC(ips_forward);
1043 IPSTAT_INC(ips_redirectsent);
1054 case 0: /* forwarded, but need redirect */
1055 /* type, code set above */
1063 type = ICMP_UNREACH;
1064 code = ICMP_UNREACH_HOST;
1068 type = ICMP_UNREACH;
1069 code = ICMP_UNREACH_NEEDFRAG;
1071 * If the MTU was set before make sure we are below the
1073 * If the MTU wasn't set before use the interface mtu or
1074 * fall back to the next smaller mtu step compared to the
1075 * current packet size.
1079 mtu = min(mtu, ia->ia_ifp->if_mtu);
1082 mtu = ia->ia_ifp->if_mtu;
1084 mtu = ip_next_mtu(ntohs(ip->ip_len), 0);
1086 IPSTAT_INC(ips_cantfrag);
1090 case EACCES: /* ipfw denied packet */
1094 icmp_error(mcopy, type, code, dest.s_addr, mtu);
1097 #define CHECK_SO_CT(sp, ct) \
1098 (((sp->so_options & SO_TIMESTAMP) && (sp->so_ts_clock == ct)) ? 1 : 0)
1101 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1107 if ((inp->inp_socket->so_options & SO_BINTIME) ||
1108 CHECK_SO_CT(inp->inp_socket, SO_TS_BINTIME)) {
1109 struct bintime boottimebin, bt;
1110 struct timespec ts1;
1112 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1114 mbuf_tstmp2timespec(m, &ts1);
1115 timespec2bintime(&ts1, &bt);
1116 getboottimebin(&boottimebin);
1117 bintime_add(&bt, &boottimebin);
1121 *mp = sbcreatecontrol(&bt, sizeof(bt), SCM_BINTIME,
1122 SOL_SOCKET, M_NOWAIT);
1124 mp = &(*mp)->m_next;
1128 if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME_MICRO)) {
1129 struct bintime boottimebin, bt1;
1130 struct timespec ts1;
1133 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1135 mbuf_tstmp2timespec(m, &ts1);
1136 timespec2bintime(&ts1, &bt1);
1137 getboottimebin(&boottimebin);
1138 bintime_add(&bt1, &boottimebin);
1139 bintime2timeval(&bt1, &tv);
1143 *mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv), SCM_TIMESTAMP,
1144 SOL_SOCKET, M_NOWAIT);
1146 mp = &(*mp)->m_next;
1149 } else if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME)) {
1150 struct bintime boottimebin;
1151 struct timespec ts, ts1;
1153 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1155 mbuf_tstmp2timespec(m, &ts);
1156 getboottimebin(&boottimebin);
1157 bintime2timespec(&boottimebin, &ts1);
1158 timespecadd(&ts, &ts1, &ts);
1162 *mp = sbcreatecontrol(&ts, sizeof(ts), SCM_REALTIME,
1163 SOL_SOCKET, M_NOWAIT);
1165 mp = &(*mp)->m_next;
1168 } else if (CHECK_SO_CT(inp->inp_socket, SO_TS_MONOTONIC)) {
1171 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1173 mbuf_tstmp2timespec(m, &ts);
1176 *mp = sbcreatecontrol(&ts, sizeof(ts), SCM_MONOTONIC,
1177 SOL_SOCKET, M_NOWAIT);
1179 mp = &(*mp)->m_next;
1183 if (stamped && (m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1185 struct sock_timestamp_info sti;
1187 bzero(&sti, sizeof(sti));
1188 sti.st_info_flags = ST_INFO_HW;
1189 if ((m->m_flags & M_TSTMP_HPREC) != 0)
1190 sti.st_info_flags |= ST_INFO_HW_HPREC;
1191 *mp = sbcreatecontrol(&sti, sizeof(sti), SCM_TIME_INFO,
1192 SOL_SOCKET, M_NOWAIT);
1194 mp = &(*mp)->m_next;
1196 if (inp->inp_flags & INP_RECVDSTADDR) {
1197 *mp = sbcreatecontrol(&ip->ip_dst, sizeof(struct in_addr),
1198 IP_RECVDSTADDR, IPPROTO_IP, M_NOWAIT);
1200 mp = &(*mp)->m_next;
1202 if (inp->inp_flags & INP_RECVTTL) {
1203 *mp = sbcreatecontrol(&ip->ip_ttl, sizeof(u_char), IP_RECVTTL,
1204 IPPROTO_IP, M_NOWAIT);
1206 mp = &(*mp)->m_next;
1210 * Moving these out of udp_input() made them even more broken
1211 * than they already were.
1213 /* options were tossed already */
1214 if (inp->inp_flags & INP_RECVOPTS) {
1215 *mp = sbcreatecontrol(opts_deleted_above,
1216 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP, M_NOWAIT);
1218 mp = &(*mp)->m_next;
1220 /* ip_srcroute doesn't do what we want here, need to fix */
1221 if (inp->inp_flags & INP_RECVRETOPTS) {
1222 *mp = sbcreatecontrol(ip_srcroute(m), sizeof(struct in_addr),
1223 IP_RECVRETOPTS, IPPROTO_IP, M_NOWAIT);
1225 mp = &(*mp)->m_next;
1228 if (inp->inp_flags & INP_RECVIF) {
1231 struct sockaddr_dl sdl;
1234 struct sockaddr_dl *sdp;
1235 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1237 if ((ifp = m->m_pkthdr.rcvif)) {
1238 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1240 * Change our mind and don't try copy.
1242 if (sdp->sdl_family != AF_LINK ||
1243 sdp->sdl_len > sizeof(sdlbuf)) {
1246 bcopy(sdp, sdl2, sdp->sdl_len);
1250 offsetof(struct sockaddr_dl, sdl_data[0]);
1251 sdl2->sdl_family = AF_LINK;
1252 sdl2->sdl_index = 0;
1253 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1255 *mp = sbcreatecontrol(sdl2, sdl2->sdl_len, IP_RECVIF,
1256 IPPROTO_IP, M_NOWAIT);
1258 mp = &(*mp)->m_next;
1260 if (inp->inp_flags & INP_RECVTOS) {
1261 *mp = sbcreatecontrol(&ip->ip_tos, sizeof(u_char), IP_RECVTOS,
1262 IPPROTO_IP, M_NOWAIT);
1264 mp = &(*mp)->m_next;
1267 if (inp->inp_flags2 & INP_RECVFLOWID) {
1268 uint32_t flowid, flow_type;
1270 flowid = m->m_pkthdr.flowid;
1271 flow_type = M_HASHTYPE_GET(m);
1274 * XXX should handle the failure of one or the
1275 * other - don't populate both?
1277 *mp = sbcreatecontrol(&flowid, sizeof(uint32_t), IP_FLOWID,
1278 IPPROTO_IP, M_NOWAIT);
1280 mp = &(*mp)->m_next;
1281 *mp = sbcreatecontrol(&flow_type, sizeof(uint32_t),
1282 IP_FLOWTYPE, IPPROTO_IP, M_NOWAIT);
1284 mp = &(*mp)->m_next;
1288 if (inp->inp_flags2 & INP_RECVRSSBUCKETID) {
1289 uint32_t flowid, flow_type;
1290 uint32_t rss_bucketid;
1292 flowid = m->m_pkthdr.flowid;
1293 flow_type = M_HASHTYPE_GET(m);
1295 if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) {
1296 *mp = sbcreatecontrol(&rss_bucketid, sizeof(uint32_t),
1297 IP_RSSBUCKETID, IPPROTO_IP, M_NOWAIT);
1299 mp = &(*mp)->m_next;
1306 * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
1307 * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
1308 * locking. This code remains in ip_input.c as ip_mroute.c is optionally
1311 VNET_DEFINE_STATIC(int, ip_rsvp_on);
1312 VNET_DEFINE(struct socket *, ip_rsvpd);
1314 #define V_ip_rsvp_on VNET(ip_rsvp_on)
1317 ip_rsvp_init(struct socket *so)
1320 if (V_ip_rsvpd != NULL)
1325 * This may seem silly, but we need to be sure we don't over-increment
1326 * the RSVP counter, in case something slips up.
1328 if (!V_ip_rsvp_on) {
1342 * This may seem silly, but we need to be sure we don't over-decrement
1343 * the RSVP counter, in case something slips up.
1353 rsvp_input(struct mbuf **mp, int *offp, int proto)
1360 if (rsvp_input_p) { /* call the real one if loaded */
1362 rsvp_input_p(mp, offp, proto);
1363 return (IPPROTO_DONE);
1366 /* Can still get packets with rsvp_on = 0 if there is a local member
1367 * of the group to which the RSVP packet is addressed. But in this
1368 * case we want to throw the packet away.
1373 return (IPPROTO_DONE);
1376 if (V_ip_rsvpd != NULL) {
1378 rip_input(mp, offp, proto);
1379 return (IPPROTO_DONE);
1381 /* Drop the packet */
1383 return (IPPROTO_DONE);