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"
42 #include <sys/param.h>
43 #include <sys/systm.h>
44 #include <sys/hhook.h>
46 #include <sys/malloc.h>
47 #include <sys/domain.h>
48 #include <sys/protosw.h>
49 #include <sys/socket.h>
51 #include <sys/kernel.h>
53 #include <sys/rmlock.h>
54 #include <sys/rwlock.h>
56 #include <sys/syslog.h>
57 #include <sys/sysctl.h>
60 #include <net/if_types.h>
61 #include <net/if_var.h>
62 #include <net/if_dl.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_fw.h>
79 #include <netinet/ip_icmp.h>
80 #include <netinet/ip_options.h>
81 #include <machine/in_cksum.h>
82 #include <netinet/ip_carp.h>
83 #include <netinet/in_rss.h>
85 #include <netipsec/ipsec_support.h>
87 #include <sys/socketvar.h>
89 #include <security/mac/mac_framework.h>
92 CTASSERT(sizeof(struct ip) == 20);
95 /* IP reassembly functions are defined in ip_reass.c. */
96 extern void ipreass_init(void);
97 extern void ipreass_drain(void);
98 extern void ipreass_slowtimo(void);
100 extern void ipreass_destroy(void);
103 struct rmlock in_ifaddr_lock;
104 RM_SYSINIT(in_ifaddr_lock, &in_ifaddr_lock, "in_ifaddr_lock");
106 VNET_DEFINE(int, rsvp_on);
108 VNET_DEFINE(int, ipforwarding);
109 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_VNET | CTLFLAG_RW,
110 &VNET_NAME(ipforwarding), 0,
111 "Enable IP forwarding between interfaces");
114 * Respond with an ICMP host redirect when we forward a packet out of
115 * the same interface on which it was received. See RFC 792.
117 VNET_DEFINE(int, ipsendredirects) = 1;
118 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_VNET | CTLFLAG_RW,
119 &VNET_NAME(ipsendredirects), 0,
120 "Enable sending IP redirects");
123 * XXX - Setting ip_checkinterface mostly implements the receive side of
124 * the Strong ES model described in RFC 1122, but since the routing table
125 * and transmit implementation do not implement the Strong ES model,
126 * setting this to 1 results in an odd hybrid.
128 * XXX - ip_checkinterface currently must be disabled if you use ipnat
129 * to translate the destination address to another local interface.
131 * XXX - ip_checkinterface must be disabled if you add IP aliases
132 * to the loopback interface instead of the interface where the
133 * packets for those addresses are received.
135 VNET_DEFINE_STATIC(int, ip_checkinterface);
136 #define V_ip_checkinterface VNET(ip_checkinterface)
137 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_VNET | CTLFLAG_RW,
138 &VNET_NAME(ip_checkinterface), 0,
139 "Verify packet arrives on correct interface");
141 /* Packet filter hooks */
142 VNET_DEFINE(pfil_head_t, inet_pfil_head);
143 VNET_DEFINE(pfil_head_t, inet_local_pfil_head);
145 static struct netisr_handler ip_nh = {
147 .nh_handler = ip_input,
148 .nh_proto = NETISR_IP,
150 .nh_m2cpuid = rss_soft_m2cpuid_v4,
151 .nh_policy = NETISR_POLICY_CPU,
152 .nh_dispatch = NETISR_DISPATCH_HYBRID,
154 .nh_policy = NETISR_POLICY_FLOW,
160 * Directly dispatched frames are currently assumed
161 * to have a flowid already calculated.
163 * It should likely have something that assert it
164 * actually has valid flow details.
166 static struct netisr_handler ip_direct_nh = {
167 .nh_name = "ip_direct",
168 .nh_handler = ip_direct_input,
169 .nh_proto = NETISR_IP_DIRECT,
170 .nh_m2cpuid = rss_soft_m2cpuid_v4,
171 .nh_policy = NETISR_POLICY_CPU,
172 .nh_dispatch = NETISR_DISPATCH_HYBRID,
176 extern struct domain inetdomain;
177 extern struct protosw inetsw[];
178 u_char ip_protox[IPPROTO_MAX];
179 VNET_DEFINE(struct in_ifaddrhead, in_ifaddrhead); /* first inet address */
180 VNET_DEFINE(struct in_ifaddrhashhead *, in_ifaddrhashtbl); /* inet addr hash table */
181 VNET_DEFINE(u_long, in_ifaddrhmask); /* mask for hash table */
184 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
185 &ip_mtu, 0, "Default MTU");
189 VNET_DEFINE(int, ipstealth);
190 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_VNET | CTLFLAG_RW,
191 &VNET_NAME(ipstealth), 0,
192 "IP stealth mode, no TTL decrementation on forwarding");
196 * IP statistics are stored in the "array" of counter(9)s.
198 VNET_PCPUSTAT_DEFINE(struct ipstat, ipstat);
199 VNET_PCPUSTAT_SYSINIT(ipstat);
200 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, IPCTL_STATS, stats, struct ipstat, ipstat,
201 "IP statistics (struct ipstat, netinet/ip_var.h)");
204 VNET_PCPUSTAT_SYSUNINIT(ipstat);
208 * Kernel module interface for updating ipstat. The argument is an index
209 * into ipstat treated as an array.
212 kmod_ipstat_inc(int statnum)
215 counter_u64_add(VNET(ipstat)[statnum], 1);
219 kmod_ipstat_dec(int statnum)
222 counter_u64_add(VNET(ipstat)[statnum], -1);
226 sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS)
230 netisr_getqlimit(&ip_nh, &qlimit);
231 error = sysctl_handle_int(oidp, &qlimit, 0, req);
232 if (error || !req->newptr)
236 return (netisr_setqlimit(&ip_nh, qlimit));
238 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen,
239 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 0,
240 sysctl_netinet_intr_queue_maxlen, "I",
241 "Maximum size of the IP input queue");
244 sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS)
246 u_int64_t qdrops_long;
249 netisr_getqdrops(&ip_nh, &qdrops_long);
250 qdrops = qdrops_long;
251 error = sysctl_handle_int(oidp, &qdrops, 0, req);
252 if (error || !req->newptr)
256 netisr_clearqdrops(&ip_nh);
260 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops,
261 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE,
262 0, 0, sysctl_netinet_intr_queue_drops, "I",
263 "Number of packets dropped from the IP input queue");
267 sysctl_netinet_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS)
271 netisr_getqlimit(&ip_direct_nh, &qlimit);
272 error = sysctl_handle_int(oidp, &qlimit, 0, req);
273 if (error || !req->newptr)
277 return (netisr_setqlimit(&ip_direct_nh, qlimit));
279 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRDQMAXLEN, intr_direct_queue_maxlen,
280 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
281 0, 0, sysctl_netinet_intr_direct_queue_maxlen,
282 "I", "Maximum size of the IP direct input queue");
285 sysctl_netinet_intr_direct_queue_drops(SYSCTL_HANDLER_ARGS)
287 u_int64_t qdrops_long;
290 netisr_getqdrops(&ip_direct_nh, &qdrops_long);
291 qdrops = qdrops_long;
292 error = sysctl_handle_int(oidp, &qdrops, 0, req);
293 if (error || !req->newptr)
297 netisr_clearqdrops(&ip_direct_nh);
301 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRDQDROPS, intr_direct_queue_drops,
302 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, 0, 0,
303 sysctl_netinet_intr_direct_queue_drops, "I",
304 "Number of packets dropped from the IP direct input queue");
308 * IP initialization: fill in IP protocol switch table.
309 * All protocols not implemented in kernel go to raw IP protocol handler.
314 struct pfil_head_args args;
318 CK_STAILQ_INIT(&V_in_ifaddrhead);
319 V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask);
321 /* Initialize IP reassembly queue. */
324 /* Initialize packet filter hooks. */
325 args.pa_version = PFIL_VERSION;
326 args.pa_flags = PFIL_IN | PFIL_OUT;
327 args.pa_type = PFIL_TYPE_IP4;
328 args.pa_headname = PFIL_INET_NAME;
329 V_inet_pfil_head = pfil_head_register(&args);
331 args.pa_flags = PFIL_OUT;
332 args.pa_headname = PFIL_INET_LOCAL_NAME;
333 V_inet_local_pfil_head = pfil_head_register(&args);
335 if (hhook_head_register(HHOOK_TYPE_IPSEC_IN, AF_INET,
336 &V_ipsec_hhh_in[HHOOK_IPSEC_INET],
337 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
338 printf("%s: WARNING: unable to register input helper hook\n",
340 if (hhook_head_register(HHOOK_TYPE_IPSEC_OUT, AF_INET,
341 &V_ipsec_hhh_out[HHOOK_IPSEC_INET],
342 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
343 printf("%s: WARNING: unable to register output helper hook\n",
346 /* Skip initialization of globals for non-default instances. */
348 if (!IS_DEFAULT_VNET(curvnet)) {
349 netisr_register_vnet(&ip_nh);
351 netisr_register_vnet(&ip_direct_nh);
357 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
359 panic("ip_init: PF_INET not found");
361 /* Initialize the entire ip_protox[] array to IPPROTO_RAW. */
362 for (i = 0; i < IPPROTO_MAX; i++)
363 ip_protox[i] = pr - inetsw;
365 * Cycle through IP protocols and put them into the appropriate place
368 for (pr = inetdomain.dom_protosw;
369 pr < inetdomain.dom_protoswNPROTOSW; pr++)
370 if (pr->pr_domain->dom_family == PF_INET &&
371 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
372 /* Be careful to only index valid IP protocols. */
373 if (pr->pr_protocol < IPPROTO_MAX)
374 ip_protox[pr->pr_protocol] = pr - inetsw;
377 netisr_register(&ip_nh);
379 netisr_register(&ip_direct_nh);
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 (*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
453 * Ip input routine. Checksum and byte swap header. If fragmented
454 * try to reassemble. Process options. Pass to next level.
457 ip_input(struct mbuf *m)
459 struct rm_priotracker in_ifa_tracker;
460 struct ip *ip = NULL;
461 struct in_ifaddr *ia = NULL;
464 int checkif, hlen = 0;
465 uint16_t sum, ip_len;
466 int dchg = 0; /* dest changed after fw */
467 struct in_addr odst; /* original dst address */
472 if (m->m_flags & M_FASTFWD_OURS) {
473 m->m_flags &= ~M_FASTFWD_OURS;
474 /* Set up some basics that will be used later. */
475 ip = mtod(m, struct ip *);
476 hlen = ip->ip_hl << 2;
477 ip_len = ntohs(ip->ip_len);
481 IPSTAT_INC(ips_total);
483 if (m->m_pkthdr.len < sizeof(struct ip))
486 if (m->m_len < sizeof (struct ip) &&
487 (m = m_pullup(m, sizeof (struct ip))) == NULL) {
488 IPSTAT_INC(ips_toosmall);
491 ip = mtod(m, struct ip *);
493 if (ip->ip_v != IPVERSION) {
494 IPSTAT_INC(ips_badvers);
498 hlen = ip->ip_hl << 2;
499 if (hlen < sizeof(struct ip)) { /* minimum header length */
500 IPSTAT_INC(ips_badhlen);
503 if (hlen > m->m_len) {
504 if ((m = m_pullup(m, hlen)) == NULL) {
505 IPSTAT_INC(ips_badhlen);
508 ip = mtod(m, struct ip *);
511 IP_PROBE(receive, NULL, NULL, ip, m->m_pkthdr.rcvif, ip, NULL);
513 /* IN_LOOPBACK must not appear on the wire - RFC1122 */
514 ifp = m->m_pkthdr.rcvif;
515 if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) ||
516 IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) {
517 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
518 IPSTAT_INC(ips_badaddr);
522 /* The unspecified address can appear only as a src address - RFC1122 */
523 if (__predict_false(ntohl(ip->ip_dst.s_addr) == INADDR_ANY)) {
524 IPSTAT_INC(ips_badaddr);
528 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
529 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
531 if (hlen == sizeof(struct ip)) {
532 sum = in_cksum_hdr(ip);
534 sum = in_cksum(m, hlen);
538 IPSTAT_INC(ips_badsum);
543 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
544 /* packet is dropped by traffic conditioner */
548 ip_len = ntohs(ip->ip_len);
550 IPSTAT_INC(ips_badlen);
555 * Check that the amount of data in the buffers
556 * is as at least much as the IP header would have us expect.
557 * Trim mbufs if longer than we expect.
558 * Drop packet if shorter than we expect.
560 if (m->m_pkthdr.len < ip_len) {
562 IPSTAT_INC(ips_tooshort);
565 if (m->m_pkthdr.len > ip_len) {
566 if (m->m_len == m->m_pkthdr.len) {
568 m->m_pkthdr.len = ip_len;
570 m_adj(m, ip_len - m->m_pkthdr.len);
574 * Try to forward the packet, but if we fail continue.
575 * ip_tryforward() may generate redirects these days.
576 * XXX the logic below falling through to normal processing
577 * if redirects are required should be revisited as well.
578 * ip_tryforward() does inbound and outbound packet firewall
579 * processing. If firewall has decided that destination becomes
580 * our local address, it sets M_FASTFWD_OURS flag. In this
581 * case skip another inbound firewall processing and update
584 if (V_ipforwarding != 0
585 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
586 && (!IPSEC_ENABLED(ipv4) ||
587 IPSEC_CAPS(ipv4, m, IPSEC_CAP_OPERABLE) == 0)
591 * ip_dooptions() was run so we can ignore the source route (or
592 * any IP options case) case for redirects in ip_tryforward().
594 if ((m = ip_tryforward(m)) == NULL)
596 if (m->m_flags & M_FASTFWD_OURS) {
597 m->m_flags &= ~M_FASTFWD_OURS;
598 ip = mtod(m, struct ip *);
603 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
605 * Bypass packet filtering for packets previously handled by IPsec.
607 if (IPSEC_ENABLED(ipv4) &&
608 IPSEC_CAPS(ipv4, m, IPSEC_CAP_BYPASS_FILTER) != 0)
613 * Run through list of hooks for input packets.
615 * NB: Beware of the destination address changing (e.g.
616 * by NAT rewriting). When this happens, tell
617 * ip_forward to do the right thing.
620 /* Jump over all PFIL processing if hooks are not active. */
621 if (!PFIL_HOOKED_IN(V_inet_pfil_head))
625 if (pfil_run_hooks(V_inet_pfil_head, &m, ifp, PFIL_IN, NULL) !=
628 if (m == NULL) /* consumed by filter */
631 ip = mtod(m, struct ip *);
632 dchg = (odst.s_addr != ip->ip_dst.s_addr);
633 ifp = m->m_pkthdr.rcvif;
635 if (m->m_flags & M_FASTFWD_OURS) {
636 m->m_flags &= ~M_FASTFWD_OURS;
639 if (m->m_flags & M_IP_NEXTHOP) {
640 if (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) {
642 * Directly ship the packet on. This allows
643 * forwarding packets originally destined to us
644 * to some other directly connected host.
653 * Process options and, if not destined for us,
654 * ship it on. ip_dooptions returns 1 when an
655 * error was detected (causing an icmp message
656 * to be sent and the original packet to be freed).
658 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
661 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
662 * matter if it is destined to another node, or whether it is
663 * a multicast one, RSVP wants it! and prevents it from being forwarded
664 * anywhere else. Also checks if the rsvp daemon is running before
665 * grabbing the packet.
667 if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP)
671 * Check our list of addresses, to see if the packet is for us.
672 * If we don't have any addresses, assume any unicast packet
673 * we receive might be for us (and let the upper layers deal
676 if (CK_STAILQ_EMPTY(&V_in_ifaddrhead) &&
677 (m->m_flags & (M_MCAST|M_BCAST)) == 0)
681 * Enable a consistency check between the destination address
682 * and the arrival interface for a unicast packet (the RFC 1122
683 * strong ES model) if IP forwarding is disabled and the packet
684 * is not locally generated and the packet is not subject to
687 * XXX - Checking also should be disabled if the destination
688 * address is ipnat'ed to a different interface.
690 * XXX - Checking is incompatible with IP aliases added
691 * to the loopback interface instead of the interface where
692 * the packets are received.
694 * XXX - This is the case for carp vhost IPs as well so we
695 * insert a workaround. If the packet got here, we already
696 * checked with carp_iamatch() and carp_forus().
698 checkif = V_ip_checkinterface && (V_ipforwarding == 0) &&
699 ifp != NULL && ((ifp->if_flags & IFF_LOOPBACK) == 0) &&
700 ifp->if_carp == NULL && (dchg == 0);
703 * Check for exact addresses in the hash bucket.
705 IN_IFADDR_RLOCK(&in_ifa_tracker);
706 LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
708 * If the address matches, verify that the packet
709 * arrived via the correct interface if checking is
712 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr &&
713 (!checkif || ia->ia_ifp == ifp)) {
714 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
715 counter_u64_add(ia->ia_ifa.ifa_ibytes,
717 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
721 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
724 * Check for broadcast addresses.
726 * Only accept broadcast packets that arrive via the matching
727 * interface. Reception of forwarded directed broadcasts would
728 * be handled via ip_forward() and ether_output() with the loopback
729 * into the stack for SIMPLEX interfaces handled by ether_output().
731 if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
732 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
733 if (ifa->ifa_addr->sa_family != AF_INET)
736 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
738 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
739 counter_u64_add(ia->ia_ifa.ifa_ibytes,
744 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) {
745 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
746 counter_u64_add(ia->ia_ifa.ifa_ibytes,
754 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
756 * RFC 3927 2.7: Do not forward multicast packets from
759 if (V_ip_mrouter && !IN_LINKLOCAL(ntohl(ip->ip_src.s_addr))) {
761 * If we are acting as a multicast router, all
762 * incoming multicast packets are passed to the
763 * kernel-level multicast forwarding function.
764 * The packet is returned (relatively) intact; if
765 * ip_mforward() returns a non-zero value, the packet
766 * must be discarded, else it may be accepted below.
768 if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) {
769 IPSTAT_INC(ips_cantforward);
775 * The process-level routing daemon needs to receive
776 * all multicast IGMP packets, whether or not this
777 * host belongs to their destination groups.
779 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_run_hooks(V_inet_local_pfil_head, &m, V_loif, PFIL_OUT, 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 (*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
870 * IP timer processing;
871 * if a timer expires on a reassembly
877 VNET_ITERATOR_DECL(vnet_iter);
879 VNET_LIST_RLOCK_NOSLEEP();
880 VNET_FOREACH(vnet_iter) {
881 CURVNET_SET(vnet_iter);
885 VNET_LIST_RUNLOCK_NOSLEEP();
891 VNET_ITERATOR_DECL(vnet_iter);
893 VNET_LIST_RLOCK_NOSLEEP();
894 VNET_FOREACH(vnet_iter) {
895 CURVNET_SET(vnet_iter);
899 VNET_LIST_RUNLOCK_NOSLEEP();
903 * The protocol to be inserted into ip_protox[] must be already registered
904 * in inetsw[], either statically or through pf_proto_register().
907 ipproto_register(short ipproto)
912 if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
913 return (EPROTONOSUPPORT);
916 * The protocol slot must not be occupied by another protocol
917 * already. An index pointing to IPPROTO_RAW is unused.
919 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
921 return (EPFNOSUPPORT);
922 if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */
925 /* Find the protocol position in inetsw[] and set the index. */
926 for (pr = inetdomain.dom_protosw;
927 pr < inetdomain.dom_protoswNPROTOSW; pr++) {
928 if (pr->pr_domain->dom_family == PF_INET &&
929 pr->pr_protocol && pr->pr_protocol == ipproto) {
930 ip_protox[pr->pr_protocol] = pr - inetsw;
934 return (EPROTONOSUPPORT);
938 ipproto_unregister(short ipproto)
943 if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
944 return (EPROTONOSUPPORT);
946 /* Check if the protocol was indeed registered. */
947 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
949 return (EPFNOSUPPORT);
950 if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */
953 /* Reset the protocol slot to IPPROTO_RAW. */
954 ip_protox[ipproto] = pr - inetsw;
958 u_char inetctlerrmap[PRC_NCMDS] = {
960 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
961 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
962 EMSGSIZE, EHOSTUNREACH, 0, 0,
963 0, 0, EHOSTUNREACH, 0,
964 ENOPROTOOPT, ECONNREFUSED
968 * Forward a packet. If some error occurs return the sender
969 * an icmp packet. Note we can't always generate a meaningful
970 * icmp message because icmp doesn't have a large enough repertoire
971 * of codes and types.
973 * If not forwarding, just drop the packet. This could be confusing
974 * if ipforwarding was zero but some routing protocol was advancing
975 * us as a gateway to somewhere. However, we must let the routing
976 * protocol deal with that.
978 * The srcrt parameter indicates whether the packet is being forwarded
979 * via a source route.
982 ip_forward(struct mbuf *m, int srcrt)
984 struct ip *ip = mtod(m, struct ip *);
985 struct in_ifaddr *ia;
987 struct sockaddr_in *sin;
991 int error, type = 0, code = 0, mtu = 0;
995 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
996 IPSTAT_INC(ips_cantforward);
1004 ip->ip_ttl <= IPTTLDEC) {
1005 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 0, 0);
1009 bzero(&ro, sizeof(ro));
1010 sin = (struct sockaddr_in *)&ro.ro_dst;
1011 sin->sin_family = AF_INET;
1012 sin->sin_len = sizeof(*sin);
1013 sin->sin_addr = ip->ip_dst;
1014 flowid = m->m_pkthdr.flowid;
1015 ro.ro_nh = fib4_lookup(M_GETFIB(m), ip->ip_dst, 0, NHR_REF, flowid);
1016 if (ro.ro_nh != NULL) {
1017 ia = ifatoia(ro.ro_nh->nh_ifa);
1021 * Save the IP header and at most 8 bytes of the payload,
1022 * in case we need to generate an ICMP message to the src.
1024 * XXX this can be optimized a lot by saving the data in a local
1025 * buffer on the stack (72 bytes at most), and only allocating the
1026 * mbuf if really necessary. The vast majority of the packets
1027 * are forwarded without having to send an ICMP back (either
1028 * because unnecessary, or because rate limited), so we are
1029 * really we are wasting a lot of work here.
1031 * We don't use m_copym() because it might return a reference
1032 * to a shared cluster. Both this function and ip_output()
1033 * assume exclusive access to the IP header in `m', so any
1034 * data in a cluster may change before we reach icmp_error().
1036 mcopy = m_gethdr(M_NOWAIT, m->m_type);
1037 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_NOWAIT)) {
1039 * It's probably ok if the pkthdr dup fails (because
1040 * the deep copy of the tag chain failed), but for now
1041 * be conservative and just discard the copy since
1042 * code below may some day want the tags.
1047 if (mcopy != NULL) {
1048 mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy));
1049 mcopy->m_pkthdr.len = mcopy->m_len;
1050 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1053 if (V_ipstealth == 0)
1055 ip->ip_ttl -= IPTTLDEC;
1056 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1057 if (IPSEC_ENABLED(ipv4)) {
1058 if ((error = IPSEC_FORWARD(ipv4, m)) != 0) {
1059 /* mbuf consumed by IPsec */
1062 if (error != EINPROGRESS)
1063 IPSTAT_INC(ips_cantforward);
1066 /* No IPsec processing required */
1070 * If forwarding packet using same interface that it came in on,
1071 * perhaps should send a redirect to sender to shortcut a hop.
1072 * Only send redirect if source is sending directly to us,
1073 * and if packet was not source routed (or has any options).
1074 * Also, don't send redirect if forwarding using a default route
1075 * or a route modified by a redirect.
1078 if (!srcrt && V_ipsendredirects &&
1079 ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) {
1080 struct nhop_object *nh;
1084 if (nh != NULL && ((nh->nh_flags & (NHF_REDIRECT|NHF_DEFAULT)) == 0)) {
1085 struct in_ifaddr *nh_ia = (struct in_ifaddr *)(nh->nh_ifa);
1086 u_long src = ntohl(ip->ip_src.s_addr);
1088 if (nh_ia != NULL &&
1089 (src & nh_ia->ia_subnetmask) == nh_ia->ia_subnet) {
1090 /* Router requirements says to only send host redirects */
1091 type = ICMP_REDIRECT;
1092 code = ICMP_REDIRECT_HOST;
1093 if (nh->nh_flags & NHF_GATEWAY) {
1094 if (nh->gw_sa.sa_family == AF_INET)
1095 dest.s_addr = nh->gw4_sa.sin_addr.s_addr;
1096 else /* Do not redirect in case gw is AF_INET6 */
1099 dest.s_addr = ip->ip_dst.s_addr;
1104 error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
1106 if (error == EMSGSIZE && ro.ro_nh)
1107 mtu = ro.ro_nh->nh_mtu;
1111 IPSTAT_INC(ips_cantforward);
1113 IPSTAT_INC(ips_forward);
1115 IPSTAT_INC(ips_redirectsent);
1126 case 0: /* forwarded, but need redirect */
1127 /* type, code set above */
1135 type = ICMP_UNREACH;
1136 code = ICMP_UNREACH_HOST;
1140 type = ICMP_UNREACH;
1141 code = ICMP_UNREACH_NEEDFRAG;
1143 * If the MTU was set before make sure we are below the
1145 * If the MTU wasn't set before use the interface mtu or
1146 * fall back to the next smaller mtu step compared to the
1147 * current packet size.
1151 mtu = min(mtu, ia->ia_ifp->if_mtu);
1154 mtu = ia->ia_ifp->if_mtu;
1156 mtu = ip_next_mtu(ntohs(ip->ip_len), 0);
1158 IPSTAT_INC(ips_cantfrag);
1162 case EACCES: /* ipfw denied packet */
1166 icmp_error(mcopy, type, code, dest.s_addr, mtu);
1169 #define CHECK_SO_CT(sp, ct) \
1170 (((sp->so_options & SO_TIMESTAMP) && (sp->so_ts_clock == ct)) ? 1 : 0)
1173 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1179 if ((inp->inp_socket->so_options & SO_BINTIME) ||
1180 CHECK_SO_CT(inp->inp_socket, SO_TS_BINTIME)) {
1181 struct bintime boottimebin, bt;
1182 struct timespec ts1;
1184 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1186 mbuf_tstmp2timespec(m, &ts1);
1187 timespec2bintime(&ts1, &bt);
1188 getboottimebin(&boottimebin);
1189 bintime_add(&bt, &boottimebin);
1193 *mp = sbcreatecontrol((caddr_t)&bt, sizeof(bt),
1194 SCM_BINTIME, SOL_SOCKET);
1196 mp = &(*mp)->m_next;
1200 if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME_MICRO)) {
1201 struct bintime boottimebin, bt1;
1202 struct timespec ts1;
1205 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1207 mbuf_tstmp2timespec(m, &ts1);
1208 timespec2bintime(&ts1, &bt1);
1209 getboottimebin(&boottimebin);
1210 bintime_add(&bt1, &boottimebin);
1211 bintime2timeval(&bt1, &tv);
1215 *mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv),
1216 SCM_TIMESTAMP, SOL_SOCKET);
1218 mp = &(*mp)->m_next;
1221 } else if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME)) {
1222 struct bintime boottimebin;
1223 struct timespec ts, ts1;
1225 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1227 mbuf_tstmp2timespec(m, &ts);
1228 getboottimebin(&boottimebin);
1229 bintime2timespec(&boottimebin, &ts1);
1230 timespecadd(&ts, &ts1, &ts);
1234 *mp = sbcreatecontrol((caddr_t)&ts, sizeof(ts),
1235 SCM_REALTIME, SOL_SOCKET);
1237 mp = &(*mp)->m_next;
1240 } else if (CHECK_SO_CT(inp->inp_socket, SO_TS_MONOTONIC)) {
1243 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1245 mbuf_tstmp2timespec(m, &ts);
1248 *mp = sbcreatecontrol((caddr_t)&ts, sizeof(ts),
1249 SCM_MONOTONIC, SOL_SOCKET);
1251 mp = &(*mp)->m_next;
1255 if (stamped && (m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1257 struct sock_timestamp_info sti;
1259 bzero(&sti, sizeof(sti));
1260 sti.st_info_flags = ST_INFO_HW;
1261 if ((m->m_flags & M_TSTMP_HPREC) != 0)
1262 sti.st_info_flags |= ST_INFO_HW_HPREC;
1263 *mp = sbcreatecontrol((caddr_t)&sti, sizeof(sti), SCM_TIME_INFO,
1266 mp = &(*mp)->m_next;
1268 if (inp->inp_flags & INP_RECVDSTADDR) {
1269 *mp = sbcreatecontrol((caddr_t)&ip->ip_dst,
1270 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1272 mp = &(*mp)->m_next;
1274 if (inp->inp_flags & INP_RECVTTL) {
1275 *mp = sbcreatecontrol((caddr_t)&ip->ip_ttl,
1276 sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
1278 mp = &(*mp)->m_next;
1282 * Moving these out of udp_input() made them even more broken
1283 * than they already were.
1285 /* options were tossed already */
1286 if (inp->inp_flags & INP_RECVOPTS) {
1287 *mp = sbcreatecontrol((caddr_t)opts_deleted_above,
1288 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1290 mp = &(*mp)->m_next;
1292 /* ip_srcroute doesn't do what we want here, need to fix */
1293 if (inp->inp_flags & INP_RECVRETOPTS) {
1294 *mp = sbcreatecontrol((caddr_t)ip_srcroute(m),
1295 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1297 mp = &(*mp)->m_next;
1300 if (inp->inp_flags & INP_RECVIF) {
1303 struct sockaddr_dl sdl;
1306 struct sockaddr_dl *sdp;
1307 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1309 if ((ifp = m->m_pkthdr.rcvif) &&
1310 ifp->if_index && ifp->if_index <= V_if_index) {
1311 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1313 * Change our mind and don't try copy.
1315 if (sdp->sdl_family != AF_LINK ||
1316 sdp->sdl_len > sizeof(sdlbuf)) {
1319 bcopy(sdp, sdl2, sdp->sdl_len);
1323 offsetof(struct sockaddr_dl, sdl_data[0]);
1324 sdl2->sdl_family = AF_LINK;
1325 sdl2->sdl_index = 0;
1326 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1328 *mp = sbcreatecontrol((caddr_t)sdl2, sdl2->sdl_len,
1329 IP_RECVIF, IPPROTO_IP);
1331 mp = &(*mp)->m_next;
1333 if (inp->inp_flags & INP_RECVTOS) {
1334 *mp = sbcreatecontrol((caddr_t)&ip->ip_tos,
1335 sizeof(u_char), IP_RECVTOS, IPPROTO_IP);
1337 mp = &(*mp)->m_next;
1340 if (inp->inp_flags2 & INP_RECVFLOWID) {
1341 uint32_t flowid, flow_type;
1343 flowid = m->m_pkthdr.flowid;
1344 flow_type = M_HASHTYPE_GET(m);
1347 * XXX should handle the failure of one or the
1348 * other - don't populate both?
1350 *mp = sbcreatecontrol((caddr_t) &flowid,
1351 sizeof(uint32_t), IP_FLOWID, IPPROTO_IP);
1353 mp = &(*mp)->m_next;
1354 *mp = sbcreatecontrol((caddr_t) &flow_type,
1355 sizeof(uint32_t), IP_FLOWTYPE, IPPROTO_IP);
1357 mp = &(*mp)->m_next;
1361 if (inp->inp_flags2 & INP_RECVRSSBUCKETID) {
1362 uint32_t flowid, flow_type;
1363 uint32_t rss_bucketid;
1365 flowid = m->m_pkthdr.flowid;
1366 flow_type = M_HASHTYPE_GET(m);
1368 if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) {
1369 *mp = sbcreatecontrol((caddr_t) &rss_bucketid,
1370 sizeof(uint32_t), IP_RSSBUCKETID, IPPROTO_IP);
1372 mp = &(*mp)->m_next;
1379 * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
1380 * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
1381 * locking. This code remains in ip_input.c as ip_mroute.c is optionally
1384 VNET_DEFINE_STATIC(int, ip_rsvp_on);
1385 VNET_DEFINE(struct socket *, ip_rsvpd);
1387 #define V_ip_rsvp_on VNET(ip_rsvp_on)
1390 ip_rsvp_init(struct socket *so)
1393 if (so->so_type != SOCK_RAW ||
1394 so->so_proto->pr_protocol != IPPROTO_RSVP)
1397 if (V_ip_rsvpd != NULL)
1402 * This may seem silly, but we need to be sure we don't over-increment
1403 * the RSVP counter, in case something slips up.
1405 if (!V_ip_rsvp_on) {
1419 * This may seem silly, but we need to be sure we don't over-decrement
1420 * the RSVP counter, in case something slips up.
1430 rsvp_input(struct mbuf **mp, int *offp, int proto)
1437 if (rsvp_input_p) { /* call the real one if loaded */
1439 rsvp_input_p(mp, offp, proto);
1440 return (IPPROTO_DONE);
1443 /* Can still get packets with rsvp_on = 0 if there is a local member
1444 * of the group to which the RSVP packet is addressed. But in this
1445 * case we want to throw the packet away.
1450 return (IPPROTO_DONE);
1453 if (V_ip_rsvpd != NULL) {
1455 rip_input(mp, offp, proto);
1456 return (IPPROTO_DONE);
1458 /* Drop the packet */
1460 return (IPPROTO_DONE);