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"
38 #include "opt_ipstealth.h"
39 #include "opt_ipsec.h"
40 #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>
65 #include <net/route.h>
66 #include <net/route/nhop.h>
67 #include <net/netisr.h>
68 #include <net/rss_config.h>
71 #include <netinet/in.h>
72 #include <netinet/in_kdtrace.h>
73 #include <netinet/in_systm.h>
74 #include <netinet/in_var.h>
75 #include <netinet/ip.h>
76 #include <netinet/in_fib.h>
77 #include <netinet/in_pcb.h>
78 #include <netinet/ip_var.h>
79 #include <netinet/ip_fw.h>
80 #include <netinet/ip_icmp.h>
81 #include <netinet/ip_options.h>
82 #include <machine/in_cksum.h>
83 #include <netinet/ip_carp.h>
84 #include <netinet/in_rss.h>
86 #include <netipsec/ipsec_support.h>
88 #include <sys/socketvar.h>
90 #include <security/mac/mac_framework.h>
93 CTASSERT(sizeof(struct ip) == 20);
96 /* IP reassembly functions are defined in ip_reass.c. */
97 extern void ipreass_init(void);
98 extern void ipreass_drain(void);
99 extern void ipreass_slowtimo(void);
101 extern void ipreass_destroy(void);
104 VNET_DEFINE(int, rsvp_on);
106 VNET_DEFINE(int, ipforwarding);
107 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_VNET | CTLFLAG_RW,
108 &VNET_NAME(ipforwarding), 0,
109 "Enable IP forwarding between interfaces");
112 * Respond with an ICMP host redirect when we forward a packet out of
113 * the same interface on which it was received. See RFC 792.
115 VNET_DEFINE(int, ipsendredirects) = 1;
116 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_VNET | CTLFLAG_RW,
117 &VNET_NAME(ipsendredirects), 0,
118 "Enable sending IP redirects");
120 VNET_DEFINE_STATIC(bool, ip_strong_es) = false;
121 #define V_ip_strong_es VNET(ip_strong_es)
122 SYSCTL_BOOL(_net_inet_ip, OID_AUTO, rfc1122_strong_es,
123 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip_strong_es), false,
124 "Packet's IP destination address must match address on arrival interface");
126 VNET_DEFINE_STATIC(bool, ip_sav) = true;
127 #define V_ip_sav VNET(ip_sav)
128 SYSCTL_BOOL(_net_inet_ip, OID_AUTO, source_address_validation,
129 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip_sav), true,
130 "Drop incoming packets with source address that is a local address");
132 VNET_DEFINE(pfil_head_t, inet_pfil_head); /* Packet filter hooks */
134 static struct netisr_handler ip_nh = {
136 .nh_handler = ip_input,
137 .nh_proto = NETISR_IP,
139 .nh_m2cpuid = rss_soft_m2cpuid_v4,
140 .nh_policy = NETISR_POLICY_CPU,
141 .nh_dispatch = NETISR_DISPATCH_HYBRID,
143 .nh_policy = NETISR_POLICY_FLOW,
149 * Directly dispatched frames are currently assumed
150 * to have a flowid already calculated.
152 * It should likely have something that assert it
153 * actually has valid flow details.
155 static struct netisr_handler ip_direct_nh = {
156 .nh_name = "ip_direct",
157 .nh_handler = ip_direct_input,
158 .nh_proto = NETISR_IP_DIRECT,
159 .nh_m2cpuid = rss_soft_m2cpuid_v4,
160 .nh_policy = NETISR_POLICY_CPU,
161 .nh_dispatch = NETISR_DISPATCH_HYBRID,
165 extern struct domain inetdomain;
166 extern struct protosw inetsw[];
167 u_char ip_protox[IPPROTO_MAX];
168 VNET_DEFINE(struct in_ifaddrhead, in_ifaddrhead); /* first inet address */
169 VNET_DEFINE(struct in_ifaddrhashhead *, in_ifaddrhashtbl); /* inet addr hash table */
170 VNET_DEFINE(u_long, in_ifaddrhmask); /* mask for hash table */
172 /* Make sure it is safe to use hashinit(9) on CK_LIST. */
173 CTASSERT(sizeof(struct in_ifaddrhashhead) == sizeof(LIST_HEAD(, in_addr)));
176 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
177 &ip_mtu, 0, "Default MTU");
181 VNET_DEFINE(int, ipstealth);
182 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_VNET | CTLFLAG_RW,
183 &VNET_NAME(ipstealth), 0,
184 "IP stealth mode, no TTL decrementation on forwarding");
188 * IP statistics are stored in the "array" of counter(9)s.
190 VNET_PCPUSTAT_DEFINE(struct ipstat, ipstat);
191 VNET_PCPUSTAT_SYSINIT(ipstat);
192 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, IPCTL_STATS, stats, struct ipstat, ipstat,
193 "IP statistics (struct ipstat, netinet/ip_var.h)");
196 VNET_PCPUSTAT_SYSUNINIT(ipstat);
200 * Kernel module interface for updating ipstat. The argument is an index
201 * into ipstat treated as an array.
204 kmod_ipstat_inc(int statnum)
207 counter_u64_add(VNET(ipstat)[statnum], 1);
211 kmod_ipstat_dec(int statnum)
214 counter_u64_add(VNET(ipstat)[statnum], -1);
218 sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS)
222 netisr_getqlimit(&ip_nh, &qlimit);
223 error = sysctl_handle_int(oidp, &qlimit, 0, req);
224 if (error || !req->newptr)
228 return (netisr_setqlimit(&ip_nh, qlimit));
230 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen,
231 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 0,
232 sysctl_netinet_intr_queue_maxlen, "I",
233 "Maximum size of the IP input queue");
236 sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS)
238 u_int64_t qdrops_long;
241 netisr_getqdrops(&ip_nh, &qdrops_long);
242 qdrops = qdrops_long;
243 error = sysctl_handle_int(oidp, &qdrops, 0, req);
244 if (error || !req->newptr)
248 netisr_clearqdrops(&ip_nh);
252 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops,
253 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE,
254 0, 0, sysctl_netinet_intr_queue_drops, "I",
255 "Number of packets dropped from the IP input queue");
259 sysctl_netinet_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS)
263 netisr_getqlimit(&ip_direct_nh, &qlimit);
264 error = sysctl_handle_int(oidp, &qlimit, 0, req);
265 if (error || !req->newptr)
269 return (netisr_setqlimit(&ip_direct_nh, qlimit));
271 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRDQMAXLEN, intr_direct_queue_maxlen,
272 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
273 0, 0, sysctl_netinet_intr_direct_queue_maxlen,
274 "I", "Maximum size of the IP direct input queue");
277 sysctl_netinet_intr_direct_queue_drops(SYSCTL_HANDLER_ARGS)
279 u_int64_t qdrops_long;
282 netisr_getqdrops(&ip_direct_nh, &qdrops_long);
283 qdrops = qdrops_long;
284 error = sysctl_handle_int(oidp, &qdrops, 0, req);
285 if (error || !req->newptr)
289 netisr_clearqdrops(&ip_direct_nh);
293 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRDQDROPS, intr_direct_queue_drops,
294 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, 0, 0,
295 sysctl_netinet_intr_direct_queue_drops, "I",
296 "Number of packets dropped from the IP direct input queue");
300 * IP initialization: fill in IP protocol switch table.
301 * All protocols not implemented in kernel go to raw IP protocol handler.
304 ip_vnet_init(void *arg __unused)
306 struct pfil_head_args args;
308 CK_STAILQ_INIT(&V_in_ifaddrhead);
309 V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask);
311 /* Initialize IP reassembly queue. */
314 /* Initialize packet filter hooks. */
315 args.pa_version = PFIL_VERSION;
316 args.pa_flags = PFIL_IN | PFIL_OUT;
317 args.pa_type = PFIL_TYPE_IP4;
318 args.pa_headname = PFIL_INET_NAME;
319 V_inet_pfil_head = pfil_head_register(&args);
321 if (hhook_head_register(HHOOK_TYPE_IPSEC_IN, AF_INET,
322 &V_ipsec_hhh_in[HHOOK_IPSEC_INET],
323 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
324 printf("%s: WARNING: unable to register input helper hook\n",
326 if (hhook_head_register(HHOOK_TYPE_IPSEC_OUT, AF_INET,
327 &V_ipsec_hhh_out[HHOOK_IPSEC_INET],
328 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
329 printf("%s: WARNING: unable to register output helper hook\n",
333 netisr_register_vnet(&ip_nh);
335 netisr_register_vnet(&ip_direct_nh);
339 VNET_SYSINIT(ip_vnet_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_FOURTH,
344 ip_init(const void *unused __unused)
348 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
349 KASSERT(pr, ("%s: PF_INET not found", __func__));
351 /* Initialize the entire ip_protox[] array to IPPROTO_RAW. */
352 for (int i = 0; i < IPPROTO_MAX; i++)
353 ip_protox[i] = pr - inetsw;
355 * Cycle through IP protocols and put them into the appropriate place
358 for (pr = inetdomain.dom_protosw;
359 pr < inetdomain.dom_protoswNPROTOSW; pr++)
360 if (pr->pr_domain->dom_family == PF_INET &&
361 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
362 /* Be careful to only index valid IP protocols. */
363 if (pr->pr_protocol < IPPROTO_MAX)
364 ip_protox[pr->pr_protocol] = pr - inetsw;
367 netisr_register(&ip_nh);
369 netisr_register(&ip_direct_nh);
372 SYSINIT(ip_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip_init, NULL);
376 ip_destroy(void *unused __unused)
381 netisr_unregister_vnet(&ip_direct_nh);
383 netisr_unregister_vnet(&ip_nh);
385 pfil_head_unregister(V_inet_pfil_head);
386 error = hhook_head_deregister(V_ipsec_hhh_in[HHOOK_IPSEC_INET]);
388 printf("%s: WARNING: unable to deregister input helper hook "
389 "type HHOOK_TYPE_IPSEC_IN, id HHOOK_IPSEC_INET: "
390 "error %d returned\n", __func__, error);
392 error = hhook_head_deregister(V_ipsec_hhh_out[HHOOK_IPSEC_INET]);
394 printf("%s: WARNING: unable to deregister output helper hook "
395 "type HHOOK_TYPE_IPSEC_OUT, id HHOOK_IPSEC_INET: "
396 "error %d returned\n", __func__, error);
399 /* Remove the IPv4 addresses from all interfaces. */
402 /* Make sure the IPv4 routes are gone as well. */
403 rib_flush_routes_family(AF_INET);
405 /* Destroy IP reassembly queue. */
408 /* Cleanup in_ifaddr hash table; should be empty. */
409 hashdestroy(V_in_ifaddrhashtbl, M_IFADDR, V_in_ifaddrhmask);
412 VNET_SYSUNINIT(ip, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip_destroy, NULL);
417 * IP direct input routine.
419 * This is called when reinjecting completed fragments where
420 * all of the previous checking and book-keeping has been done.
423 ip_direct_input(struct mbuf *m)
428 ip = mtod(m, struct ip *);
429 hlen = ip->ip_hl << 2;
431 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
432 if (IPSEC_ENABLED(ipv4)) {
433 if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
437 IPSTAT_INC(ips_delivered);
438 (*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
444 * Ip input routine. Checksum and byte swap header. If fragmented
445 * try to reassemble. Process options. Pass to next level.
448 ip_input(struct mbuf *m)
450 struct ip *ip = NULL;
451 struct in_ifaddr *ia = NULL;
455 uint16_t sum, ip_len;
456 int dchg = 0; /* dest changed after fw */
457 struct in_addr odst; /* original dst address */
463 if (m->m_flags & M_FASTFWD_OURS) {
464 m->m_flags &= ~M_FASTFWD_OURS;
465 /* Set up some basics that will be used later. */
466 ip = mtod(m, struct ip *);
467 hlen = ip->ip_hl << 2;
468 ip_len = ntohs(ip->ip_len);
472 IPSTAT_INC(ips_total);
474 if (__predict_false(m->m_pkthdr.len < sizeof(struct ip)))
477 if (m->m_len < sizeof(struct ip)) {
478 m = m_pullup(m, sizeof(struct ip));
479 if (__predict_false(m == NULL)) {
480 IPSTAT_INC(ips_toosmall);
484 ip = mtod(m, struct ip *);
486 if (__predict_false(ip->ip_v != IPVERSION)) {
487 IPSTAT_INC(ips_badvers);
491 hlen = ip->ip_hl << 2;
492 if (__predict_false(hlen < sizeof(struct ip))) { /* minimum header length */
493 IPSTAT_INC(ips_badhlen);
496 if (hlen > m->m_len) {
497 m = m_pullup(m, hlen);
498 if (__predict_false(m == NULL)) {
499 IPSTAT_INC(ips_badhlen);
502 ip = mtod(m, struct ip *);
505 IP_PROBE(receive, NULL, NULL, ip, m->m_pkthdr.rcvif, ip, NULL);
507 /* IN_LOOPBACK must not appear on the wire - RFC1122 */
508 ifp = m->m_pkthdr.rcvif;
509 if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) ||
510 IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) {
511 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
512 IPSTAT_INC(ips_badaddr);
517 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
518 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
520 if (hlen == sizeof(struct ip)) {
521 sum = in_cksum_hdr(ip);
523 sum = in_cksum(m, hlen);
526 if (__predict_false(sum)) {
527 IPSTAT_INC(ips_badsum);
532 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
533 /* packet is dropped by traffic conditioner */
537 ip_len = ntohs(ip->ip_len);
538 if (__predict_false(ip_len < hlen)) {
539 IPSTAT_INC(ips_badlen);
544 * Check that the amount of data in the buffers
545 * is as at least much as the IP header would have us expect.
546 * Trim mbufs if longer than we expect.
547 * Drop packet if shorter than we expect.
549 if (__predict_false(m->m_pkthdr.len < ip_len)) {
551 IPSTAT_INC(ips_tooshort);
554 if (m->m_pkthdr.len > ip_len) {
555 if (m->m_len == m->m_pkthdr.len) {
557 m->m_pkthdr.len = ip_len;
559 m_adj(m, ip_len - m->m_pkthdr.len);
563 * Try to forward the packet, but if we fail continue.
564 * ip_tryforward() may generate redirects these days.
565 * XXX the logic below falling through to normal processing
566 * if redirects are required should be revisited as well.
567 * ip_tryforward() does inbound and outbound packet firewall
568 * processing. If firewall has decided that destination becomes
569 * our local address, it sets M_FASTFWD_OURS flag. In this
570 * case skip another inbound firewall processing and update
573 if (V_ipforwarding != 0
574 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
575 && (!IPSEC_ENABLED(ipv4) ||
576 IPSEC_CAPS(ipv4, m, IPSEC_CAP_OPERABLE) == 0)
580 * ip_dooptions() was run so we can ignore the source route (or
581 * any IP options case) case for redirects in ip_tryforward().
583 if ((m = ip_tryforward(m)) == NULL)
585 if (m->m_flags & M_FASTFWD_OURS) {
586 m->m_flags &= ~M_FASTFWD_OURS;
587 ip = mtod(m, struct ip *);
592 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
594 * Bypass packet filtering for packets previously handled by IPsec.
596 if (IPSEC_ENABLED(ipv4) &&
597 IPSEC_CAPS(ipv4, m, IPSEC_CAP_BYPASS_FILTER) != 0)
602 * Run through list of hooks for input packets.
604 * NB: Beware of the destination address changing (e.g.
605 * by NAT rewriting). When this happens, tell
606 * ip_forward to do the right thing.
609 /* Jump over all PFIL processing if hooks are not active. */
610 if (!PFIL_HOOKED_IN(V_inet_pfil_head))
614 if (pfil_run_hooks(V_inet_pfil_head, &m, ifp, PFIL_IN, NULL) !=
617 if (m == NULL) /* consumed by filter */
620 ip = mtod(m, struct ip *);
621 dchg = (odst.s_addr != ip->ip_dst.s_addr);
623 if (m->m_flags & M_FASTFWD_OURS) {
624 m->m_flags &= ~M_FASTFWD_OURS;
627 if (m->m_flags & M_IP_NEXTHOP) {
628 if (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) {
630 * Directly ship the packet on. This allows
631 * forwarding packets originally destined to us
632 * to some other directly connected host.
641 * Process options and, if not destined for us,
642 * ship it on. ip_dooptions returns 1 when an
643 * error was detected (causing an icmp message
644 * to be sent and the original packet to be freed).
646 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
649 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
650 * matter if it is destined to another node, or whether it is
651 * a multicast one, RSVP wants it! and prevents it from being forwarded
652 * anywhere else. Also checks if the rsvp daemon is running before
653 * grabbing the packet.
655 if (ip->ip_p == IPPROTO_RSVP && V_rsvp_on)
659 * Check our list of addresses, to see if the packet is for us.
660 * If we don't have any addresses, assume any unicast packet
661 * we receive might be for us (and let the upper layers deal
664 if (CK_STAILQ_EMPTY(&V_in_ifaddrhead) &&
665 (m->m_flags & (M_MCAST|M_BCAST)) == 0)
669 * Enable a consistency check between the destination address
670 * and the arrival interface for a unicast packet (the RFC 1122
671 * strong ES model) with a list of additional predicates:
672 * - if IP forwarding is disabled
673 * - the packet is not locally generated
674 * - the packet is not subject to 'ipfw fwd'
675 * - Interface is not running CARP. If the packet got here, we already
676 * checked it with carp_iamatch() and carp_forus().
678 strong_es = V_ip_strong_es && (V_ipforwarding == 0) &&
679 ((ifp->if_flags & IFF_LOOPBACK) == 0) &&
680 ifp->if_carp == NULL && (dchg == 0);
683 * Check for exact addresses in the hash bucket.
685 CK_LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
686 if (IA_SIN(ia)->sin_addr.s_addr != ip->ip_dst.s_addr)
690 * net.inet.ip.rfc1122_strong_es: the address matches, verify
691 * that the packet arrived via the correct interface.
693 if (__predict_false(strong_es && ia->ia_ifp != ifp)) {
694 IPSTAT_INC(ips_badaddr);
699 * net.inet.ip.source_address_validation: drop incoming
700 * packets that pretend to be ours.
702 if (V_ip_sav && !(ifp->if_flags & IFF_LOOPBACK) &&
703 __predict_false(in_localip_fib(ip->ip_src, ifp->if_fib))) {
704 IPSTAT_INC(ips_badaddr);
708 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
709 counter_u64_add(ia->ia_ifa.ifa_ibytes, m->m_pkthdr.len);
714 * Check for broadcast addresses.
716 * Only accept broadcast packets that arrive via the matching
717 * interface. Reception of forwarded directed broadcasts would
718 * be handled via ip_forward() and ether_output() with the loopback
719 * into the stack for SIMPLEX interfaces handled by ether_output().
721 if (ifp->if_flags & IFF_BROADCAST) {
722 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
723 if (ifa->ifa_addr->sa_family != AF_INET)
726 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
728 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
729 counter_u64_add(ia->ia_ifa.ifa_ibytes,
734 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) {
735 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
736 counter_u64_add(ia->ia_ifa.ifa_ibytes,
744 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
746 * RFC 3927 2.7: Do not forward multicast packets from
749 if (V_ip_mrouter && !IN_LINKLOCAL(ntohl(ip->ip_src.s_addr))) {
751 * If we are acting as a multicast router, all
752 * incoming multicast packets are passed to the
753 * kernel-level multicast forwarding function.
754 * The packet is returned (relatively) intact; if
755 * ip_mforward() returns a non-zero value, the packet
756 * must be discarded, else it may be accepted below.
758 if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) {
759 IPSTAT_INC(ips_cantforward);
765 * The process-level routing daemon needs to receive
766 * all multicast IGMP packets, whether or not this
767 * host belongs to their destination groups.
769 if (ip->ip_p == IPPROTO_IGMP) {
772 IPSTAT_INC(ips_forward);
775 * Assume the packet is for us, to avoid prematurely taking
776 * a lock on the in_multi hash. Protocols must perform
777 * their own filtering and update statistics accordingly.
781 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
783 if (ip->ip_dst.s_addr == INADDR_ANY)
785 /* RFC 3927 2.7: Do not forward packets to or from IN_LINKLOCAL. */
786 if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr)) ||
787 IN_LINKLOCAL(ntohl(ip->ip_src.s_addr))) {
788 IPSTAT_INC(ips_cantforward);
794 * Not for us; forward if possible and desirable.
796 if (V_ipforwarding == 0) {
797 IPSTAT_INC(ips_cantforward);
807 * IPSTEALTH: Process non-routing options only
808 * if the packet is destined for us.
810 if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1))
812 #endif /* IPSTEALTH */
815 * Attempt reassembly; if it succeeds, proceed.
816 * ip_reass() will return a different mbuf.
818 if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) {
819 /* XXXGL: shouldn't we save & set m_flags? */
823 ip = mtod(m, struct ip *);
824 /* Get the header length of the reassembled packet */
825 hlen = ip->ip_hl << 2;
828 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
829 if (IPSEC_ENABLED(ipv4)) {
830 if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
836 * Switch out to protocol's input routine.
838 IPSTAT_INC(ips_delivered);
840 (*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
847 * IP timer processing;
848 * if a timer expires on a reassembly
854 VNET_ITERATOR_DECL(vnet_iter);
856 VNET_LIST_RLOCK_NOSLEEP();
857 VNET_FOREACH(vnet_iter) {
858 CURVNET_SET(vnet_iter);
862 VNET_LIST_RUNLOCK_NOSLEEP();
868 VNET_ITERATOR_DECL(vnet_iter);
870 VNET_LIST_RLOCK_NOSLEEP();
871 VNET_FOREACH(vnet_iter) {
872 CURVNET_SET(vnet_iter);
876 VNET_LIST_RUNLOCK_NOSLEEP();
880 * The protocol to be inserted into ip_protox[] must be already registered
881 * in inetsw[], either statically or through pf_proto_register().
884 ipproto_register(short ipproto)
889 if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
890 return (EPROTONOSUPPORT);
893 * The protocol slot must not be occupied by another protocol
894 * already. An index pointing to IPPROTO_RAW is unused.
896 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
898 return (EPFNOSUPPORT);
899 if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */
902 /* Find the protocol position in inetsw[] and set the index. */
903 for (pr = inetdomain.dom_protosw;
904 pr < inetdomain.dom_protoswNPROTOSW; pr++) {
905 if (pr->pr_domain->dom_family == PF_INET &&
906 pr->pr_protocol && pr->pr_protocol == ipproto) {
907 ip_protox[pr->pr_protocol] = pr - inetsw;
911 return (EPROTONOSUPPORT);
915 ipproto_unregister(short ipproto)
920 if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
921 return (EPROTONOSUPPORT);
923 /* Check if the protocol was indeed registered. */
924 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
926 return (EPFNOSUPPORT);
927 if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */
930 /* Reset the protocol slot to IPPROTO_RAW. */
931 ip_protox[ipproto] = pr - inetsw;
935 u_char inetctlerrmap[PRC_NCMDS] = {
937 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
938 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
939 EMSGSIZE, EHOSTUNREACH, 0, 0,
940 0, 0, EHOSTUNREACH, 0,
941 ENOPROTOOPT, ECONNREFUSED
945 * Forward a packet. If some error occurs return the sender
946 * an icmp packet. Note we can't always generate a meaningful
947 * icmp message because icmp doesn't have a large enough repertoire
948 * of codes and types.
950 * If not forwarding, just drop the packet. This could be confusing
951 * if ipforwarding was zero but some routing protocol was advancing
952 * us as a gateway to somewhere. However, we must let the routing
953 * protocol deal with that.
955 * The srcrt parameter indicates whether the packet is being forwarded
956 * via a source route.
959 ip_forward(struct mbuf *m, int srcrt)
961 struct ip *ip = mtod(m, struct ip *);
962 struct in_ifaddr *ia;
964 struct sockaddr_in *sin;
968 int error, type = 0, code = 0, mtu = 0;
972 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
973 IPSTAT_INC(ips_cantforward);
981 ip->ip_ttl <= IPTTLDEC) {
982 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 0, 0);
986 bzero(&ro, sizeof(ro));
987 sin = (struct sockaddr_in *)&ro.ro_dst;
988 sin->sin_family = AF_INET;
989 sin->sin_len = sizeof(*sin);
990 sin->sin_addr = ip->ip_dst;
991 flowid = m->m_pkthdr.flowid;
992 ro.ro_nh = fib4_lookup(M_GETFIB(m), ip->ip_dst, 0, NHR_REF, flowid);
993 if (ro.ro_nh != NULL) {
994 ia = ifatoia(ro.ro_nh->nh_ifa);
998 * Save the IP header and at most 8 bytes of the payload,
999 * in case we need to generate an ICMP message to the src.
1001 * XXX this can be optimized a lot by saving the data in a local
1002 * buffer on the stack (72 bytes at most), and only allocating the
1003 * mbuf if really necessary. The vast majority of the packets
1004 * are forwarded without having to send an ICMP back (either
1005 * because unnecessary, or because rate limited), so we are
1006 * really we are wasting a lot of work here.
1008 * We don't use m_copym() because it might return a reference
1009 * to a shared cluster. Both this function and ip_output()
1010 * assume exclusive access to the IP header in `m', so any
1011 * data in a cluster may change before we reach icmp_error().
1013 mcopy = m_gethdr(M_NOWAIT, m->m_type);
1014 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_NOWAIT)) {
1016 * It's probably ok if the pkthdr dup fails (because
1017 * the deep copy of the tag chain failed), but for now
1018 * be conservative and just discard the copy since
1019 * code below may some day want the tags.
1024 if (mcopy != NULL) {
1025 mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy));
1026 mcopy->m_pkthdr.len = mcopy->m_len;
1027 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1030 if (V_ipstealth == 0)
1032 ip->ip_ttl -= IPTTLDEC;
1033 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1034 if (IPSEC_ENABLED(ipv4)) {
1035 if ((error = IPSEC_FORWARD(ipv4, m)) != 0) {
1036 /* mbuf consumed by IPsec */
1039 if (error != EINPROGRESS)
1040 IPSTAT_INC(ips_cantforward);
1043 /* No IPsec processing required */
1047 * If forwarding packet using same interface that it came in on,
1048 * perhaps should send a redirect to sender to shortcut a hop.
1049 * Only send redirect if source is sending directly to us,
1050 * and if packet was not source routed (or has any options).
1051 * Also, don't send redirect if forwarding using a default route
1052 * or a route modified by a redirect.
1055 if (!srcrt && V_ipsendredirects &&
1056 ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) {
1057 struct nhop_object *nh;
1061 if (nh != NULL && ((nh->nh_flags & (NHF_REDIRECT|NHF_DEFAULT)) == 0)) {
1062 struct in_ifaddr *nh_ia = (struct in_ifaddr *)(nh->nh_ifa);
1063 u_long src = ntohl(ip->ip_src.s_addr);
1065 if (nh_ia != NULL &&
1066 (src & nh_ia->ia_subnetmask) == nh_ia->ia_subnet) {
1067 /* Router requirements says to only send host redirects */
1068 type = ICMP_REDIRECT;
1069 code = ICMP_REDIRECT_HOST;
1070 if (nh->nh_flags & NHF_GATEWAY) {
1071 if (nh->gw_sa.sa_family == AF_INET)
1072 dest.s_addr = nh->gw4_sa.sin_addr.s_addr;
1073 else /* Do not redirect in case gw is AF_INET6 */
1076 dest.s_addr = ip->ip_dst.s_addr;
1081 error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
1083 if (error == EMSGSIZE && ro.ro_nh)
1084 mtu = ro.ro_nh->nh_mtu;
1088 IPSTAT_INC(ips_cantforward);
1090 IPSTAT_INC(ips_forward);
1092 IPSTAT_INC(ips_redirectsent);
1103 case 0: /* forwarded, but need redirect */
1104 /* type, code set above */
1112 type = ICMP_UNREACH;
1113 code = ICMP_UNREACH_HOST;
1117 type = ICMP_UNREACH;
1118 code = ICMP_UNREACH_NEEDFRAG;
1120 * If the MTU was set before make sure we are below the
1122 * If the MTU wasn't set before use the interface mtu or
1123 * fall back to the next smaller mtu step compared to the
1124 * current packet size.
1128 mtu = min(mtu, ia->ia_ifp->if_mtu);
1131 mtu = ia->ia_ifp->if_mtu;
1133 mtu = ip_next_mtu(ntohs(ip->ip_len), 0);
1135 IPSTAT_INC(ips_cantfrag);
1139 case EACCES: /* ipfw denied packet */
1143 icmp_error(mcopy, type, code, dest.s_addr, mtu);
1146 #define CHECK_SO_CT(sp, ct) \
1147 (((sp->so_options & SO_TIMESTAMP) && (sp->so_ts_clock == ct)) ? 1 : 0)
1150 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1156 if ((inp->inp_socket->so_options & SO_BINTIME) ||
1157 CHECK_SO_CT(inp->inp_socket, SO_TS_BINTIME)) {
1158 struct bintime boottimebin, bt;
1159 struct timespec ts1;
1161 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1163 mbuf_tstmp2timespec(m, &ts1);
1164 timespec2bintime(&ts1, &bt);
1165 getboottimebin(&boottimebin);
1166 bintime_add(&bt, &boottimebin);
1170 *mp = sbcreatecontrol((caddr_t)&bt, sizeof(bt),
1171 SCM_BINTIME, SOL_SOCKET);
1173 mp = &(*mp)->m_next;
1177 if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME_MICRO)) {
1178 struct bintime boottimebin, bt1;
1179 struct timespec ts1;
1182 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1184 mbuf_tstmp2timespec(m, &ts1);
1185 timespec2bintime(&ts1, &bt1);
1186 getboottimebin(&boottimebin);
1187 bintime_add(&bt1, &boottimebin);
1188 bintime2timeval(&bt1, &tv);
1192 *mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv),
1193 SCM_TIMESTAMP, SOL_SOCKET);
1195 mp = &(*mp)->m_next;
1198 } else if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME)) {
1199 struct bintime boottimebin;
1200 struct timespec ts, ts1;
1202 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1204 mbuf_tstmp2timespec(m, &ts);
1205 getboottimebin(&boottimebin);
1206 bintime2timespec(&boottimebin, &ts1);
1207 timespecadd(&ts, &ts1, &ts);
1211 *mp = sbcreatecontrol((caddr_t)&ts, sizeof(ts),
1212 SCM_REALTIME, SOL_SOCKET);
1214 mp = &(*mp)->m_next;
1217 } else if (CHECK_SO_CT(inp->inp_socket, SO_TS_MONOTONIC)) {
1220 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1222 mbuf_tstmp2timespec(m, &ts);
1225 *mp = sbcreatecontrol((caddr_t)&ts, sizeof(ts),
1226 SCM_MONOTONIC, SOL_SOCKET);
1228 mp = &(*mp)->m_next;
1232 if (stamped && (m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1234 struct sock_timestamp_info sti;
1236 bzero(&sti, sizeof(sti));
1237 sti.st_info_flags = ST_INFO_HW;
1238 if ((m->m_flags & M_TSTMP_HPREC) != 0)
1239 sti.st_info_flags |= ST_INFO_HW_HPREC;
1240 *mp = sbcreatecontrol((caddr_t)&sti, sizeof(sti), SCM_TIME_INFO,
1243 mp = &(*mp)->m_next;
1245 if (inp->inp_flags & INP_RECVDSTADDR) {
1246 *mp = sbcreatecontrol((caddr_t)&ip->ip_dst,
1247 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1249 mp = &(*mp)->m_next;
1251 if (inp->inp_flags & INP_RECVTTL) {
1252 *mp = sbcreatecontrol((caddr_t)&ip->ip_ttl,
1253 sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
1255 mp = &(*mp)->m_next;
1259 * Moving these out of udp_input() made them even more broken
1260 * than they already were.
1262 /* options were tossed already */
1263 if (inp->inp_flags & INP_RECVOPTS) {
1264 *mp = sbcreatecontrol((caddr_t)opts_deleted_above,
1265 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1267 mp = &(*mp)->m_next;
1269 /* ip_srcroute doesn't do what we want here, need to fix */
1270 if (inp->inp_flags & INP_RECVRETOPTS) {
1271 *mp = sbcreatecontrol((caddr_t)ip_srcroute(m),
1272 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1274 mp = &(*mp)->m_next;
1277 if (inp->inp_flags & INP_RECVIF) {
1280 struct sockaddr_dl sdl;
1283 struct sockaddr_dl *sdp;
1284 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1286 if ((ifp = m->m_pkthdr.rcvif)) {
1287 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1289 * Change our mind and don't try copy.
1291 if (sdp->sdl_family != AF_LINK ||
1292 sdp->sdl_len > sizeof(sdlbuf)) {
1295 bcopy(sdp, sdl2, sdp->sdl_len);
1299 offsetof(struct sockaddr_dl, sdl_data[0]);
1300 sdl2->sdl_family = AF_LINK;
1301 sdl2->sdl_index = 0;
1302 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1304 *mp = sbcreatecontrol((caddr_t)sdl2, sdl2->sdl_len,
1305 IP_RECVIF, IPPROTO_IP);
1307 mp = &(*mp)->m_next;
1309 if (inp->inp_flags & INP_RECVTOS) {
1310 *mp = sbcreatecontrol((caddr_t)&ip->ip_tos,
1311 sizeof(u_char), IP_RECVTOS, IPPROTO_IP);
1313 mp = &(*mp)->m_next;
1316 if (inp->inp_flags2 & INP_RECVFLOWID) {
1317 uint32_t flowid, flow_type;
1319 flowid = m->m_pkthdr.flowid;
1320 flow_type = M_HASHTYPE_GET(m);
1323 * XXX should handle the failure of one or the
1324 * other - don't populate both?
1326 *mp = sbcreatecontrol((caddr_t) &flowid,
1327 sizeof(uint32_t), IP_FLOWID, IPPROTO_IP);
1329 mp = &(*mp)->m_next;
1330 *mp = sbcreatecontrol((caddr_t) &flow_type,
1331 sizeof(uint32_t), IP_FLOWTYPE, IPPROTO_IP);
1333 mp = &(*mp)->m_next;
1337 if (inp->inp_flags2 & INP_RECVRSSBUCKETID) {
1338 uint32_t flowid, flow_type;
1339 uint32_t rss_bucketid;
1341 flowid = m->m_pkthdr.flowid;
1342 flow_type = M_HASHTYPE_GET(m);
1344 if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) {
1345 *mp = sbcreatecontrol((caddr_t) &rss_bucketid,
1346 sizeof(uint32_t), IP_RSSBUCKETID, IPPROTO_IP);
1348 mp = &(*mp)->m_next;
1355 * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
1356 * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
1357 * locking. This code remains in ip_input.c as ip_mroute.c is optionally
1360 VNET_DEFINE_STATIC(int, ip_rsvp_on);
1361 VNET_DEFINE(struct socket *, ip_rsvpd);
1363 #define V_ip_rsvp_on VNET(ip_rsvp_on)
1366 ip_rsvp_init(struct socket *so)
1369 if (so->so_type != SOCK_RAW ||
1370 so->so_proto->pr_protocol != IPPROTO_RSVP)
1373 if (V_ip_rsvpd != NULL)
1378 * This may seem silly, but we need to be sure we don't over-increment
1379 * the RSVP counter, in case something slips up.
1381 if (!V_ip_rsvp_on) {
1395 * This may seem silly, but we need to be sure we don't over-decrement
1396 * the RSVP counter, in case something slips up.
1406 rsvp_input(struct mbuf **mp, int *offp, int proto)
1413 if (rsvp_input_p) { /* call the real one if loaded */
1415 rsvp_input_p(mp, offp, proto);
1416 return (IPPROTO_DONE);
1419 /* Can still get packets with rsvp_on = 0 if there is a local member
1420 * of the group to which the RSVP packet is addressed. But in this
1421 * case we want to throw the packet away.
1426 return (IPPROTO_DONE);
1429 if (V_ip_rsvpd != NULL) {
1431 rip_input(mp, offp, proto);
1432 return (IPPROTO_DONE);
1434 /* Drop the packet */
1436 return (IPPROTO_DONE);