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/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_pcb.h>
76 #include <netinet/ip_var.h>
77 #include <netinet/ip_fw.h>
78 #include <netinet/ip_icmp.h>
79 #include <netinet/ip_options.h>
80 #include <machine/in_cksum.h>
81 #include <netinet/ip_carp.h>
82 #include <netinet/in_rss.h>
84 #include <netipsec/ipsec_support.h>
86 #include <sys/socketvar.h>
88 #include <security/mac/mac_framework.h>
91 CTASSERT(sizeof(struct ip) == 20);
94 /* IP reassembly functions are defined in ip_reass.c. */
95 extern void ipreass_init(void);
96 extern void ipreass_drain(void);
97 extern void ipreass_slowtimo(void);
99 extern void ipreass_destroy(void);
102 struct rmlock in_ifaddr_lock;
103 RM_SYSINIT(in_ifaddr_lock, &in_ifaddr_lock, "in_ifaddr_lock");
105 VNET_DEFINE(int, rsvp_on);
107 VNET_DEFINE(int, ipforwarding);
108 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_VNET | CTLFLAG_RW,
109 &VNET_NAME(ipforwarding), 0,
110 "Enable IP forwarding between interfaces");
112 VNET_DEFINE_STATIC(int, ipsendredirects) = 1; /* XXX */
113 #define V_ipsendredirects VNET(ipsendredirects)
114 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_VNET | CTLFLAG_RW,
115 &VNET_NAME(ipsendredirects), 0,
116 "Enable sending IP redirects");
119 * XXX - Setting ip_checkinterface mostly implements the receive side of
120 * the Strong ES model described in RFC 1122, but since the routing table
121 * and transmit implementation do not implement the Strong ES model,
122 * setting this to 1 results in an odd hybrid.
124 * XXX - ip_checkinterface currently must be disabled if you use ipnat
125 * to translate the destination address to another local interface.
127 * XXX - ip_checkinterface must be disabled if you add IP aliases
128 * to the loopback interface instead of the interface where the
129 * packets for those addresses are received.
131 VNET_DEFINE_STATIC(int, ip_checkinterface);
132 #define V_ip_checkinterface VNET(ip_checkinterface)
133 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_VNET | CTLFLAG_RW,
134 &VNET_NAME(ip_checkinterface), 0,
135 "Verify packet arrives on correct interface");
137 VNET_DEFINE(pfil_head_t, inet_pfil_head); /* Packet filter hooks */
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 extern struct domain inetdomain;
171 extern struct protosw inetsw[];
172 u_char ip_protox[IPPROTO_MAX];
173 VNET_DEFINE(struct in_ifaddrhead, in_ifaddrhead); /* first inet address */
174 VNET_DEFINE(struct in_ifaddrhashhead *, in_ifaddrhashtbl); /* inet addr hash table */
175 VNET_DEFINE(u_long, in_ifaddrhmask); /* mask for hash table */
178 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
179 &ip_mtu, 0, "Default MTU");
183 VNET_DEFINE(int, ipstealth);
184 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_VNET | CTLFLAG_RW,
185 &VNET_NAME(ipstealth), 0,
186 "IP stealth mode, no TTL decrementation on forwarding");
190 * IP statistics are stored in the "array" of counter(9)s.
192 VNET_PCPUSTAT_DEFINE(struct ipstat, ipstat);
193 VNET_PCPUSTAT_SYSINIT(ipstat);
194 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, IPCTL_STATS, stats, struct ipstat, ipstat,
195 "IP statistics (struct ipstat, netinet/ip_var.h)");
198 VNET_PCPUSTAT_SYSUNINIT(ipstat);
202 * Kernel module interface for updating ipstat. The argument is an index
203 * into ipstat treated as an array.
206 kmod_ipstat_inc(int statnum)
209 counter_u64_add(VNET(ipstat)[statnum], 1);
213 kmod_ipstat_dec(int statnum)
216 counter_u64_add(VNET(ipstat)[statnum], -1);
220 sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS)
224 netisr_getqlimit(&ip_nh, &qlimit);
225 error = sysctl_handle_int(oidp, &qlimit, 0, req);
226 if (error || !req->newptr)
230 return (netisr_setqlimit(&ip_nh, qlimit));
232 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen,
233 CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet_intr_queue_maxlen, "I",
234 "Maximum size of the IP input queue");
237 sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS)
239 u_int64_t qdrops_long;
242 netisr_getqdrops(&ip_nh, &qdrops_long);
243 qdrops = qdrops_long;
244 error = sysctl_handle_int(oidp, &qdrops, 0, req);
245 if (error || !req->newptr)
249 netisr_clearqdrops(&ip_nh);
253 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops,
254 CTLTYPE_INT|CTLFLAG_RD, 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, 0, 0, sysctl_netinet_intr_direct_queue_maxlen,
273 "I", "Maximum size of the IP direct input queue");
276 sysctl_netinet_intr_direct_queue_drops(SYSCTL_HANDLER_ARGS)
278 u_int64_t qdrops_long;
281 netisr_getqdrops(&ip_direct_nh, &qdrops_long);
282 qdrops = qdrops_long;
283 error = sysctl_handle_int(oidp, &qdrops, 0, req);
284 if (error || !req->newptr)
288 netisr_clearqdrops(&ip_direct_nh);
292 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRDQDROPS, intr_direct_queue_drops,
293 CTLTYPE_INT|CTLFLAG_RD, 0, 0, sysctl_netinet_intr_direct_queue_drops, "I",
294 "Number of packets dropped from the IP direct input queue");
298 * IP initialization: fill in IP protocol switch table.
299 * All protocols not implemented in kernel go to raw IP protocol handler.
304 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",
332 /* Skip initialization of globals for non-default instances. */
334 if (!IS_DEFAULT_VNET(curvnet)) {
335 netisr_register_vnet(&ip_nh);
337 netisr_register_vnet(&ip_direct_nh);
343 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
345 panic("ip_init: PF_INET not found");
347 /* Initialize the entire ip_protox[] array to IPPROTO_RAW. */
348 for (i = 0; i < IPPROTO_MAX; i++)
349 ip_protox[i] = pr - inetsw;
351 * Cycle through IP protocols and put them into the appropriate place
354 for (pr = inetdomain.dom_protosw;
355 pr < inetdomain.dom_protoswNPROTOSW; pr++)
356 if (pr->pr_domain->dom_family == PF_INET &&
357 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
358 /* Be careful to only index valid IP protocols. */
359 if (pr->pr_protocol < IPPROTO_MAX)
360 ip_protox[pr->pr_protocol] = pr - inetsw;
363 netisr_register(&ip_nh);
365 netisr_register(&ip_direct_nh);
371 ip_destroy(void *unused __unused)
377 netisr_unregister_vnet(&ip_direct_nh);
379 netisr_unregister_vnet(&ip_nh);
381 pfil_head_unregister(V_inet_pfil_head);
382 error = hhook_head_deregister(V_ipsec_hhh_in[HHOOK_IPSEC_INET]);
384 printf("%s: WARNING: unable to deregister input helper hook "
385 "type HHOOK_TYPE_IPSEC_IN, id HHOOK_IPSEC_INET: "
386 "error %d returned\n", __func__, error);
388 error = hhook_head_deregister(V_ipsec_hhh_out[HHOOK_IPSEC_INET]);
390 printf("%s: WARNING: unable to deregister output helper hook "
391 "type HHOOK_TYPE_IPSEC_OUT, id HHOOK_IPSEC_INET: "
392 "error %d returned\n", __func__, error);
395 /* Remove the IPv4 addresses from all interfaces. */
398 /* Make sure the IPv4 routes are gone as well. */
400 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link)
401 rt_flushifroutes_af(ifp, AF_INET);
404 /* Destroy IP reassembly queue. */
407 /* Cleanup in_ifaddr hash table; should be empty. */
408 hashdestroy(V_in_ifaddrhashtbl, M_IFADDR, V_in_ifaddrhmask);
411 VNET_SYSUNINIT(ip, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip_destroy, NULL);
416 * IP direct input routine.
418 * This is called when reinjecting completed fragments where
419 * all of the previous checking and book-keeping has been done.
422 ip_direct_input(struct mbuf *m)
427 ip = mtod(m, struct ip *);
428 hlen = ip->ip_hl << 2;
430 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
431 if (IPSEC_ENABLED(ipv4)) {
432 if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
436 IPSTAT_INC(ips_delivered);
437 (*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
443 * Ip input routine. Checksum and byte swap header. If fragmented
444 * try to reassemble. Process options. Pass to next level.
447 ip_input(struct mbuf *m)
449 struct rm_priotracker in_ifa_tracker;
450 struct ip *ip = NULL;
451 struct in_ifaddr *ia = NULL;
454 int checkif, hlen = 0;
455 uint16_t sum, ip_len;
456 int dchg = 0; /* dest changed after fw */
457 struct in_addr odst; /* original dst address */
461 if (m->m_flags & M_FASTFWD_OURS) {
462 m->m_flags &= ~M_FASTFWD_OURS;
463 /* Set up some basics that will be used later. */
464 ip = mtod(m, struct ip *);
465 hlen = ip->ip_hl << 2;
466 ip_len = ntohs(ip->ip_len);
470 IPSTAT_INC(ips_total);
472 if (m->m_pkthdr.len < sizeof(struct ip))
475 if (m->m_len < sizeof (struct ip) &&
476 (m = m_pullup(m, sizeof (struct ip))) == NULL) {
477 IPSTAT_INC(ips_toosmall);
480 ip = mtod(m, struct ip *);
482 if (ip->ip_v != IPVERSION) {
483 IPSTAT_INC(ips_badvers);
487 hlen = ip->ip_hl << 2;
488 if (hlen < sizeof(struct ip)) { /* minimum header length */
489 IPSTAT_INC(ips_badhlen);
492 if (hlen > m->m_len) {
493 if ((m = m_pullup(m, hlen)) == NULL) {
494 IPSTAT_INC(ips_badhlen);
497 ip = mtod(m, struct ip *);
500 IP_PROBE(receive, NULL, NULL, ip, m->m_pkthdr.rcvif, ip, NULL);
502 /* 127/8 must not appear on wire - RFC1122 */
503 ifp = m->m_pkthdr.rcvif;
504 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
505 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
506 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
507 IPSTAT_INC(ips_badaddr);
512 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
513 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
515 if (hlen == sizeof(struct ip)) {
516 sum = in_cksum_hdr(ip);
518 sum = in_cksum(m, hlen);
522 IPSTAT_INC(ips_badsum);
527 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
528 /* packet is dropped by traffic conditioner */
532 ip_len = ntohs(ip->ip_len);
534 IPSTAT_INC(ips_badlen);
539 * Check that the amount of data in the buffers
540 * is as at least much as the IP header would have us expect.
541 * Trim mbufs if longer than we expect.
542 * Drop packet if shorter than we expect.
544 if (m->m_pkthdr.len < ip_len) {
546 IPSTAT_INC(ips_tooshort);
549 if (m->m_pkthdr.len > ip_len) {
550 if (m->m_len == m->m_pkthdr.len) {
552 m->m_pkthdr.len = ip_len;
554 m_adj(m, ip_len - m->m_pkthdr.len);
558 * Try to forward the packet, but if we fail continue.
559 * ip_tryforward() does not generate redirects, so fall
560 * through to normal processing if redirects are required.
561 * ip_tryforward() does inbound and outbound packet firewall
562 * processing. If firewall has decided that destination becomes
563 * our local address, it sets M_FASTFWD_OURS flag. In this
564 * case skip another inbound firewall processing and update
567 if (V_ipforwarding != 0 && V_ipsendredirects == 0
568 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
569 && (!IPSEC_ENABLED(ipv4) ||
570 IPSEC_CAPS(ipv4, m, IPSEC_CAP_OPERABLE) == 0)
573 if ((m = ip_tryforward(m)) == NULL)
575 if (m->m_flags & M_FASTFWD_OURS) {
576 m->m_flags &= ~M_FASTFWD_OURS;
577 ip = mtod(m, struct ip *);
582 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
584 * Bypass packet filtering for packets previously handled by IPsec.
586 if (IPSEC_ENABLED(ipv4) &&
587 IPSEC_CAPS(ipv4, m, IPSEC_CAP_BYPASS_FILTER) != 0)
592 * Run through list of hooks for input packets.
594 * NB: Beware of the destination address changing (e.g.
595 * by NAT rewriting). When this happens, tell
596 * ip_forward to do the right thing.
599 /* Jump over all PFIL processing if hooks are not active. */
600 if (!PFIL_HOOKED_IN(V_inet_pfil_head))
604 if (pfil_run_hooks(V_inet_pfil_head, &m, ifp, PFIL_IN, NULL) !=
607 if (m == NULL) /* consumed by filter */
610 ip = mtod(m, struct ip *);
611 dchg = (odst.s_addr != ip->ip_dst.s_addr);
612 ifp = m->m_pkthdr.rcvif;
614 if (m->m_flags & M_FASTFWD_OURS) {
615 m->m_flags &= ~M_FASTFWD_OURS;
618 if (m->m_flags & M_IP_NEXTHOP) {
619 if (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) {
621 * Directly ship the packet on. This allows
622 * forwarding packets originally destined to us
623 * to some other directly connected host.
632 * Process options and, if not destined for us,
633 * ship it on. ip_dooptions returns 1 when an
634 * error was detected (causing an icmp message
635 * to be sent and the original packet to be freed).
637 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
640 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
641 * matter if it is destined to another node, or whether it is
642 * a multicast one, RSVP wants it! and prevents it from being forwarded
643 * anywhere else. Also checks if the rsvp daemon is running before
644 * grabbing the packet.
646 if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP)
650 * Check our list of addresses, to see if the packet is for us.
651 * If we don't have any addresses, assume any unicast packet
652 * we receive might be for us (and let the upper layers deal
655 if (CK_STAILQ_EMPTY(&V_in_ifaddrhead) &&
656 (m->m_flags & (M_MCAST|M_BCAST)) == 0)
660 * Enable a consistency check between the destination address
661 * and the arrival interface for a unicast packet (the RFC 1122
662 * strong ES model) if IP forwarding is disabled and the packet
663 * is not locally generated and the packet is not subject to
666 * XXX - Checking also should be disabled if the destination
667 * address is ipnat'ed to a different interface.
669 * XXX - Checking is incompatible with IP aliases added
670 * to the loopback interface instead of the interface where
671 * the packets are received.
673 * XXX - This is the case for carp vhost IPs as well so we
674 * insert a workaround. If the packet got here, we already
675 * checked with carp_iamatch() and carp_forus().
677 checkif = V_ip_checkinterface && (V_ipforwarding == 0) &&
678 ifp != NULL && ((ifp->if_flags & IFF_LOOPBACK) == 0) &&
679 ifp->if_carp == NULL && (dchg == 0);
682 * Check for exact addresses in the hash bucket.
684 IN_IFADDR_RLOCK(&in_ifa_tracker);
685 LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
687 * If the address matches, verify that the packet
688 * arrived via the correct interface if checking is
691 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr &&
692 (!checkif || ia->ia_ifp == ifp)) {
693 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
694 counter_u64_add(ia->ia_ifa.ifa_ibytes,
696 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
700 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
703 * Check for broadcast addresses.
705 * Only accept broadcast packets that arrive via the matching
706 * interface. Reception of forwarded directed broadcasts would
707 * be handled via ip_forward() and ether_output() with the loopback
708 * into the stack for SIMPLEX interfaces handled by ether_output().
710 if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
711 struct epoch_tracker et;
714 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
715 if (ifa->ifa_addr->sa_family != AF_INET)
718 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
720 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
721 counter_u64_add(ia->ia_ifa.ifa_ibytes,
727 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) {
728 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
729 counter_u64_add(ia->ia_ifa.ifa_ibytes,
739 /* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */
740 if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) {
741 IPSTAT_INC(ips_cantforward);
745 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
748 * If we are acting as a multicast router, all
749 * incoming multicast packets are passed to the
750 * kernel-level multicast forwarding function.
751 * The packet is returned (relatively) intact; if
752 * ip_mforward() returns a non-zero value, the packet
753 * must be discarded, else it may be accepted below.
755 if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) {
756 IPSTAT_INC(ips_cantforward);
762 * The process-level routing daemon needs to receive
763 * all multicast IGMP packets, whether or not this
764 * host belongs to their destination groups.
766 if (ip->ip_p == IPPROTO_IGMP)
768 IPSTAT_INC(ips_forward);
771 * Assume the packet is for us, to avoid prematurely taking
772 * a lock on the in_multi hash. Protocols must perform
773 * their own filtering and update statistics accordingly.
777 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
779 if (ip->ip_dst.s_addr == INADDR_ANY)
783 * Not for us; forward if possible and desirable.
785 if (V_ipforwarding == 0) {
786 IPSTAT_INC(ips_cantforward);
796 * IPSTEALTH: Process non-routing options only
797 * if the packet is destined for us.
799 if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1))
801 #endif /* IPSTEALTH */
804 * Attempt reassembly; if it succeeds, proceed.
805 * ip_reass() will return a different mbuf.
807 if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) {
808 /* XXXGL: shouldn't we save & set m_flags? */
812 ip = mtod(m, struct ip *);
813 /* Get the header length of the reassembled packet */
814 hlen = ip->ip_hl << 2;
817 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
818 if (IPSEC_ENABLED(ipv4)) {
819 if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
825 * Switch out to protocol's input routine.
827 IPSTAT_INC(ips_delivered);
829 (*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
836 * IP timer processing;
837 * if a timer expires on a reassembly
843 VNET_ITERATOR_DECL(vnet_iter);
845 VNET_LIST_RLOCK_NOSLEEP();
846 VNET_FOREACH(vnet_iter) {
847 CURVNET_SET(vnet_iter);
851 VNET_LIST_RUNLOCK_NOSLEEP();
857 VNET_ITERATOR_DECL(vnet_iter);
859 VNET_LIST_RLOCK_NOSLEEP();
860 VNET_FOREACH(vnet_iter) {
861 CURVNET_SET(vnet_iter);
865 VNET_LIST_RUNLOCK_NOSLEEP();
869 * The protocol to be inserted into ip_protox[] must be already registered
870 * in inetsw[], either statically or through pf_proto_register().
873 ipproto_register(short ipproto)
878 if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
879 return (EPROTONOSUPPORT);
882 * The protocol slot must not be occupied by another protocol
883 * already. An index pointing to IPPROTO_RAW is unused.
885 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
887 return (EPFNOSUPPORT);
888 if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */
891 /* Find the protocol position in inetsw[] and set the index. */
892 for (pr = inetdomain.dom_protosw;
893 pr < inetdomain.dom_protoswNPROTOSW; pr++) {
894 if (pr->pr_domain->dom_family == PF_INET &&
895 pr->pr_protocol && pr->pr_protocol == ipproto) {
896 ip_protox[pr->pr_protocol] = pr - inetsw;
900 return (EPROTONOSUPPORT);
904 ipproto_unregister(short ipproto)
909 if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
910 return (EPROTONOSUPPORT);
912 /* Check if the protocol was indeed registered. */
913 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
915 return (EPFNOSUPPORT);
916 if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */
919 /* Reset the protocol slot to IPPROTO_RAW. */
920 ip_protox[ipproto] = pr - inetsw;
924 u_char inetctlerrmap[PRC_NCMDS] = {
926 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
927 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
928 EMSGSIZE, EHOSTUNREACH, 0, 0,
929 0, 0, EHOSTUNREACH, 0,
930 ENOPROTOOPT, ECONNREFUSED
934 * Forward a packet. If some error occurs return the sender
935 * an icmp packet. Note we can't always generate a meaningful
936 * icmp message because icmp doesn't have a large enough repertoire
937 * of codes and types.
939 * If not forwarding, just drop the packet. This could be confusing
940 * if ipforwarding was zero but some routing protocol was advancing
941 * us as a gateway to somewhere. However, we must let the routing
942 * protocol deal with that.
944 * The srcrt parameter indicates whether the packet is being forwarded
945 * via a source route.
948 ip_forward(struct mbuf *m, int srcrt)
950 struct ip *ip = mtod(m, struct ip *);
951 struct in_ifaddr *ia;
953 struct sockaddr_in *sin;
956 struct epoch_tracker et;
957 int error, type = 0, code = 0, mtu = 0;
959 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
960 IPSTAT_INC(ips_cantforward);
968 ip->ip_ttl <= IPTTLDEC) {
969 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 0, 0);
973 bzero(&ro, sizeof(ro));
974 sin = (struct sockaddr_in *)&ro.ro_dst;
975 sin->sin_family = AF_INET;
976 sin->sin_len = sizeof(*sin);
977 sin->sin_addr = ip->ip_dst;
979 rtalloc_mpath_fib(&ro,
980 ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
983 in_rtalloc_ign(&ro, 0, M_GETFIB(m));
986 if (ro.ro_rt != NULL) {
987 ia = ifatoia(ro.ro_rt->rt_ifa);
991 * Save the IP header and at most 8 bytes of the payload,
992 * in case we need to generate an ICMP message to the src.
994 * XXX this can be optimized a lot by saving the data in a local
995 * buffer on the stack (72 bytes at most), and only allocating the
996 * mbuf if really necessary. The vast majority of the packets
997 * are forwarded without having to send an ICMP back (either
998 * because unnecessary, or because rate limited), so we are
999 * really we are wasting a lot of work here.
1001 * We don't use m_copym() because it might return a reference
1002 * to a shared cluster. Both this function and ip_output()
1003 * assume exclusive access to the IP header in `m', so any
1004 * data in a cluster may change before we reach icmp_error().
1006 mcopy = m_gethdr(M_NOWAIT, m->m_type);
1007 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_NOWAIT)) {
1009 * It's probably ok if the pkthdr dup fails (because
1010 * the deep copy of the tag chain failed), but for now
1011 * be conservative and just discard the copy since
1012 * code below may some day want the tags.
1017 if (mcopy != NULL) {
1018 mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy));
1019 mcopy->m_pkthdr.len = mcopy->m_len;
1020 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1023 if (V_ipstealth == 0)
1025 ip->ip_ttl -= IPTTLDEC;
1026 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1027 if (IPSEC_ENABLED(ipv4)) {
1028 if ((error = IPSEC_FORWARD(ipv4, m)) != 0) {
1029 /* mbuf consumed by IPsec */
1031 if (error != EINPROGRESS)
1032 IPSTAT_INC(ips_cantforward);
1035 /* No IPsec processing required */
1039 * If forwarding packet using same interface that it came in on,
1040 * perhaps should send a redirect to sender to shortcut a hop.
1041 * Only send redirect if source is sending directly to us,
1042 * and if packet was not source routed (or has any options).
1043 * Also, don't send redirect if forwarding using a default route
1044 * or a route modified by a redirect.
1047 if (!srcrt && V_ipsendredirects &&
1048 ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) {
1053 if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1054 satosin(rt_key(rt))->sin_addr.s_addr != 0) {
1055 #define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa))
1056 u_long src = ntohl(ip->ip_src.s_addr);
1059 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1060 if (rt->rt_flags & RTF_GATEWAY)
1061 dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr;
1063 dest.s_addr = ip->ip_dst.s_addr;
1064 /* Router requirements says to only send host redirects */
1065 type = ICMP_REDIRECT;
1066 code = ICMP_REDIRECT_HOST;
1071 error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
1073 if (error == EMSGSIZE && ro.ro_rt)
1074 mtu = ro.ro_rt->rt_mtu;
1078 IPSTAT_INC(ips_cantforward);
1080 IPSTAT_INC(ips_forward);
1082 IPSTAT_INC(ips_redirectsent);
1095 case 0: /* forwarded, but need redirect */
1096 /* type, code set above */
1104 type = ICMP_UNREACH;
1105 code = ICMP_UNREACH_HOST;
1109 type = ICMP_UNREACH;
1110 code = ICMP_UNREACH_NEEDFRAG;
1112 * If the MTU was set before make sure we are below the
1114 * If the MTU wasn't set before use the interface mtu or
1115 * fall back to the next smaller mtu step compared to the
1116 * current packet size.
1120 mtu = min(mtu, ia->ia_ifp->if_mtu);
1123 mtu = ia->ia_ifp->if_mtu;
1125 mtu = ip_next_mtu(ntohs(ip->ip_len), 0);
1127 IPSTAT_INC(ips_cantfrag);
1131 case EACCES: /* ipfw denied packet */
1135 icmp_error(mcopy, type, code, dest.s_addr, mtu);
1140 #define CHECK_SO_CT(sp, ct) \
1141 (((sp->so_options & SO_TIMESTAMP) && (sp->so_ts_clock == ct)) ? 1 : 0)
1144 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1150 if ((inp->inp_socket->so_options & SO_BINTIME) ||
1151 CHECK_SO_CT(inp->inp_socket, SO_TS_BINTIME)) {
1152 struct bintime boottimebin, bt;
1153 struct timespec ts1;
1155 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1157 mbuf_tstmp2timespec(m, &ts1);
1158 timespec2bintime(&ts1, &bt);
1159 getboottimebin(&boottimebin);
1160 bintime_add(&bt, &boottimebin);
1164 *mp = sbcreatecontrol((caddr_t)&bt, sizeof(bt),
1165 SCM_BINTIME, SOL_SOCKET);
1167 mp = &(*mp)->m_next;
1171 if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME_MICRO)) {
1172 struct bintime boottimebin, bt1;
1173 struct timespec ts1;;
1176 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1178 mbuf_tstmp2timespec(m, &ts1);
1179 timespec2bintime(&ts1, &bt1);
1180 getboottimebin(&boottimebin);
1181 bintime_add(&bt1, &boottimebin);
1182 bintime2timeval(&bt1, &tv);
1186 *mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv),
1187 SCM_TIMESTAMP, SOL_SOCKET);
1189 mp = &(*mp)->m_next;
1192 } else if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME)) {
1193 struct bintime boottimebin;
1194 struct timespec ts, ts1;
1196 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1198 mbuf_tstmp2timespec(m, &ts);
1199 getboottimebin(&boottimebin);
1200 bintime2timespec(&boottimebin, &ts1);
1201 timespecadd(&ts, &ts1, &ts);
1205 *mp = sbcreatecontrol((caddr_t)&ts, sizeof(ts),
1206 SCM_REALTIME, SOL_SOCKET);
1208 mp = &(*mp)->m_next;
1211 } else if (CHECK_SO_CT(inp->inp_socket, SO_TS_MONOTONIC)) {
1214 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1216 mbuf_tstmp2timespec(m, &ts);
1219 *mp = sbcreatecontrol((caddr_t)&ts, sizeof(ts),
1220 SCM_MONOTONIC, SOL_SOCKET);
1222 mp = &(*mp)->m_next;
1226 if (stamped && (m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1228 struct sock_timestamp_info sti;
1230 bzero(&sti, sizeof(sti));
1231 sti.st_info_flags = ST_INFO_HW;
1232 if ((m->m_flags & M_TSTMP_HPREC) != 0)
1233 sti.st_info_flags |= ST_INFO_HW_HPREC;
1234 *mp = sbcreatecontrol((caddr_t)&sti, sizeof(sti), SCM_TIME_INFO,
1237 mp = &(*mp)->m_next;
1239 if (inp->inp_flags & INP_RECVDSTADDR) {
1240 *mp = sbcreatecontrol((caddr_t)&ip->ip_dst,
1241 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1243 mp = &(*mp)->m_next;
1245 if (inp->inp_flags & INP_RECVTTL) {
1246 *mp = sbcreatecontrol((caddr_t)&ip->ip_ttl,
1247 sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
1249 mp = &(*mp)->m_next;
1253 * Moving these out of udp_input() made them even more broken
1254 * than they already were.
1256 /* options were tossed already */
1257 if (inp->inp_flags & INP_RECVOPTS) {
1258 *mp = sbcreatecontrol((caddr_t)opts_deleted_above,
1259 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1261 mp = &(*mp)->m_next;
1263 /* ip_srcroute doesn't do what we want here, need to fix */
1264 if (inp->inp_flags & INP_RECVRETOPTS) {
1265 *mp = sbcreatecontrol((caddr_t)ip_srcroute(m),
1266 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1268 mp = &(*mp)->m_next;
1271 if (inp->inp_flags & INP_RECVIF) {
1274 struct sockaddr_dl sdl;
1277 struct sockaddr_dl *sdp;
1278 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1280 if ((ifp = m->m_pkthdr.rcvif) &&
1281 ifp->if_index && ifp->if_index <= V_if_index) {
1282 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1284 * Change our mind and don't try copy.
1286 if (sdp->sdl_family != AF_LINK ||
1287 sdp->sdl_len > sizeof(sdlbuf)) {
1290 bcopy(sdp, sdl2, sdp->sdl_len);
1294 offsetof(struct sockaddr_dl, sdl_data[0]);
1295 sdl2->sdl_family = AF_LINK;
1296 sdl2->sdl_index = 0;
1297 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1299 *mp = sbcreatecontrol((caddr_t)sdl2, sdl2->sdl_len,
1300 IP_RECVIF, IPPROTO_IP);
1302 mp = &(*mp)->m_next;
1304 if (inp->inp_flags & INP_RECVTOS) {
1305 *mp = sbcreatecontrol((caddr_t)&ip->ip_tos,
1306 sizeof(u_char), IP_RECVTOS, IPPROTO_IP);
1308 mp = &(*mp)->m_next;
1311 if (inp->inp_flags2 & INP_RECVFLOWID) {
1312 uint32_t flowid, flow_type;
1314 flowid = m->m_pkthdr.flowid;
1315 flow_type = M_HASHTYPE_GET(m);
1318 * XXX should handle the failure of one or the
1319 * other - don't populate both?
1321 *mp = sbcreatecontrol((caddr_t) &flowid,
1322 sizeof(uint32_t), IP_FLOWID, IPPROTO_IP);
1324 mp = &(*mp)->m_next;
1325 *mp = sbcreatecontrol((caddr_t) &flow_type,
1326 sizeof(uint32_t), IP_FLOWTYPE, IPPROTO_IP);
1328 mp = &(*mp)->m_next;
1332 if (inp->inp_flags2 & INP_RECVRSSBUCKETID) {
1333 uint32_t flowid, flow_type;
1334 uint32_t rss_bucketid;
1336 flowid = m->m_pkthdr.flowid;
1337 flow_type = M_HASHTYPE_GET(m);
1339 if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) {
1340 *mp = sbcreatecontrol((caddr_t) &rss_bucketid,
1341 sizeof(uint32_t), IP_RSSBUCKETID, IPPROTO_IP);
1343 mp = &(*mp)->m_next;
1350 * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
1351 * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
1352 * locking. This code remains in ip_input.c as ip_mroute.c is optionally
1355 VNET_DEFINE_STATIC(int, ip_rsvp_on);
1356 VNET_DEFINE(struct socket *, ip_rsvpd);
1358 #define V_ip_rsvp_on VNET(ip_rsvp_on)
1361 ip_rsvp_init(struct socket *so)
1364 if (so->so_type != SOCK_RAW ||
1365 so->so_proto->pr_protocol != IPPROTO_RSVP)
1368 if (V_ip_rsvpd != NULL)
1373 * This may seem silly, but we need to be sure we don't over-increment
1374 * the RSVP counter, in case something slips up.
1376 if (!V_ip_rsvp_on) {
1390 * This may seem silly, but we need to be sure we don't over-decrement
1391 * the RSVP counter, in case something slips up.
1401 rsvp_input(struct mbuf **mp, int *offp, int proto)
1408 if (rsvp_input_p) { /* call the real one if loaded */
1410 rsvp_input_p(mp, offp, proto);
1411 return (IPPROTO_DONE);
1414 /* Can still get packets with rsvp_on = 0 if there is a local member
1415 * of the group to which the RSVP packet is addressed. But in this
1416 * case we want to throw the packet away.
1421 return (IPPROTO_DONE);
1424 if (V_ip_rsvpd != NULL) {
1426 rip_input(mp, offp, proto);
1427 return (IPPROTO_DONE);
1429 /* Drop the packet */
1431 return (IPPROTO_DONE);