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>
62 #include <net/if_types.h>
63 #include <net/if_var.h>
64 #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 static VNET_DEFINE(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 static VNET_DEFINE(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(struct pfil_head, inet_pfil_hook); /* 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.
307 TAILQ_INIT(&V_in_ifaddrhead);
308 V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask);
310 /* Initialize IP reassembly queue. */
313 /* Initialize packet filter hooks. */
314 V_inet_pfil_hook.ph_type = PFIL_TYPE_AF;
315 V_inet_pfil_hook.ph_af = AF_INET;
316 if ((i = pfil_head_register(&V_inet_pfil_hook)) != 0)
317 printf("%s: WARNING: unable to register pfil hook, "
318 "error %d\n", __func__, i);
320 if (hhook_head_register(HHOOK_TYPE_IPSEC_IN, AF_INET,
321 &V_ipsec_hhh_in[HHOOK_IPSEC_INET],
322 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
323 printf("%s: WARNING: unable to register input helper hook\n",
325 if (hhook_head_register(HHOOK_TYPE_IPSEC_OUT, AF_INET,
326 &V_ipsec_hhh_out[HHOOK_IPSEC_INET],
327 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
328 printf("%s: WARNING: unable to register output helper hook\n",
331 /* Skip initialization of globals for non-default instances. */
333 if (!IS_DEFAULT_VNET(curvnet)) {
334 netisr_register_vnet(&ip_nh);
336 netisr_register_vnet(&ip_direct_nh);
342 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
344 panic("ip_init: PF_INET not found");
346 /* Initialize the entire ip_protox[] array to IPPROTO_RAW. */
347 for (i = 0; i < IPPROTO_MAX; i++)
348 ip_protox[i] = pr - inetsw;
350 * Cycle through IP protocols and put them into the appropriate place
353 for (pr = inetdomain.dom_protosw;
354 pr < inetdomain.dom_protoswNPROTOSW; pr++)
355 if (pr->pr_domain->dom_family == PF_INET &&
356 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
357 /* Be careful to only index valid IP protocols. */
358 if (pr->pr_protocol < IPPROTO_MAX)
359 ip_protox[pr->pr_protocol] = pr - inetsw;
362 netisr_register(&ip_nh);
364 netisr_register(&ip_direct_nh);
370 ip_destroy(void *unused __unused)
376 netisr_unregister_vnet(&ip_direct_nh);
378 netisr_unregister_vnet(&ip_nh);
380 if ((error = pfil_head_unregister(&V_inet_pfil_hook)) != 0)
381 printf("%s: WARNING: unable to unregister pfil hook, "
382 "error %d\n", __func__, error);
384 error = hhook_head_deregister(V_ipsec_hhh_in[HHOOK_IPSEC_INET]);
386 printf("%s: WARNING: unable to deregister input helper hook "
387 "type HHOOK_TYPE_IPSEC_IN, id HHOOK_IPSEC_INET: "
388 "error %d returned\n", __func__, error);
390 error = hhook_head_deregister(V_ipsec_hhh_out[HHOOK_IPSEC_INET]);
392 printf("%s: WARNING: unable to deregister output helper hook "
393 "type HHOOK_TYPE_IPSEC_OUT, id HHOOK_IPSEC_INET: "
394 "error %d returned\n", __func__, error);
397 /* Remove the IPv4 addresses from all interfaces. */
400 /* Make sure the IPv4 routes are gone as well. */
402 TAILQ_FOREACH(ifp, &V_ifnet, if_link)
403 rt_flushifroutes_af(ifp, AF_INET);
406 /* Destroy IP reassembly queue. */
409 /* Cleanup in_ifaddr hash table; should be empty. */
410 hashdestroy(V_in_ifaddrhashtbl, M_IFADDR, V_in_ifaddrhmask);
413 VNET_SYSUNINIT(ip, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip_destroy, NULL);
418 * IP direct input routine.
420 * This is called when reinjecting completed fragments where
421 * all of the previous checking and book-keeping has been done.
424 ip_direct_input(struct mbuf *m)
429 ip = mtod(m, struct ip *);
430 hlen = ip->ip_hl << 2;
432 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
433 if (IPSEC_ENABLED(ipv4)) {
434 if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
438 IPSTAT_INC(ips_delivered);
439 (*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
445 * Ip input routine. Checksum and byte swap header. If fragmented
446 * try to reassemble. Process options. Pass to next level.
449 ip_input(struct mbuf *m)
451 struct ip *ip = NULL;
452 struct in_ifaddr *ia = NULL;
455 int checkif, hlen = 0;
456 uint16_t sum, ip_len;
457 int dchg = 0; /* dest changed after fw */
458 struct in_addr odst; /* original dst address */
462 if (m->m_flags & M_FASTFWD_OURS) {
463 m->m_flags &= ~M_FASTFWD_OURS;
464 /* Set up some basics that will be used later. */
465 ip = mtod(m, struct ip *);
466 hlen = ip->ip_hl << 2;
467 ip_len = ntohs(ip->ip_len);
471 IPSTAT_INC(ips_total);
473 if (m->m_pkthdr.len < sizeof(struct ip))
476 if (m->m_len < sizeof (struct ip) &&
477 (m = m_pullup(m, sizeof (struct ip))) == NULL) {
478 IPSTAT_INC(ips_toosmall);
481 ip = mtod(m, struct ip *);
483 if (ip->ip_v != IPVERSION) {
484 IPSTAT_INC(ips_badvers);
488 hlen = ip->ip_hl << 2;
489 if (hlen < sizeof(struct ip)) { /* minimum header length */
490 IPSTAT_INC(ips_badhlen);
493 if (hlen > m->m_len) {
494 if ((m = m_pullup(m, hlen)) == NULL) {
495 IPSTAT_INC(ips_badhlen);
498 ip = mtod(m, struct ip *);
501 IP_PROBE(receive, NULL, NULL, ip, m->m_pkthdr.rcvif, ip, NULL);
503 /* 127/8 must not appear on wire - RFC1122 */
504 ifp = m->m_pkthdr.rcvif;
505 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
506 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
507 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
508 IPSTAT_INC(ips_badaddr);
513 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
514 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
516 if (hlen == sizeof(struct ip)) {
517 sum = in_cksum_hdr(ip);
519 sum = in_cksum(m, hlen);
523 IPSTAT_INC(ips_badsum);
528 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
529 /* packet is dropped by traffic conditioner */
533 ip_len = ntohs(ip->ip_len);
535 IPSTAT_INC(ips_badlen);
540 * Check that the amount of data in the buffers
541 * is as at least much as the IP header would have us expect.
542 * Trim mbufs if longer than we expect.
543 * Drop packet if shorter than we expect.
545 if (m->m_pkthdr.len < ip_len) {
547 IPSTAT_INC(ips_tooshort);
550 if (m->m_pkthdr.len > ip_len) {
551 if (m->m_len == m->m_pkthdr.len) {
553 m->m_pkthdr.len = ip_len;
555 m_adj(m, ip_len - m->m_pkthdr.len);
559 * Try to forward the packet, but if we fail continue.
560 * ip_tryforward() does inbound and outbound packet firewall
561 * processing. If firewall has decided that destination becomes
562 * our local address, it sets M_FASTFWD_OURS flag. In this
563 * case skip another inbound firewall processing and update
566 if (V_ipforwarding != 0
567 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
568 && (!IPSEC_ENABLED(ipv4) ||
569 IPSEC_CAPS(ipv4, m, IPSEC_CAP_OPERABLE) == 0)
572 if ((m = ip_tryforward(m)) == NULL)
574 if (m->m_flags & M_FASTFWD_OURS) {
575 m->m_flags &= ~M_FASTFWD_OURS;
576 ip = mtod(m, struct ip *);
581 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
583 * Bypass packet filtering for packets previously handled by IPsec.
585 if (IPSEC_ENABLED(ipv4) &&
586 IPSEC_CAPS(ipv4, m, IPSEC_CAP_BYPASS_FILTER) != 0)
591 * Run through list of hooks for input packets.
593 * NB: Beware of the destination address changing (e.g.
594 * by NAT rewriting). When this happens, tell
595 * ip_forward to do the right thing.
598 /* Jump over all PFIL processing if hooks are not active. */
599 if (!PFIL_HOOKED(&V_inet_pfil_hook))
603 if (pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_IN, NULL) != 0)
605 if (m == NULL) /* consumed by filter */
608 ip = mtod(m, struct ip *);
609 dchg = (odst.s_addr != ip->ip_dst.s_addr);
610 ifp = m->m_pkthdr.rcvif;
612 if (m->m_flags & M_FASTFWD_OURS) {
613 m->m_flags &= ~M_FASTFWD_OURS;
616 if (m->m_flags & M_IP_NEXTHOP) {
617 if (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) {
619 * Directly ship the packet on. This allows
620 * forwarding packets originally destined to us
621 * to some other directly connected host.
630 * Process options and, if not destined for us,
631 * ship it on. ip_dooptions returns 1 when an
632 * error was detected (causing an icmp message
633 * to be sent and the original packet to be freed).
635 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
638 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
639 * matter if it is destined to another node, or whether it is
640 * a multicast one, RSVP wants it! and prevents it from being forwarded
641 * anywhere else. Also checks if the rsvp daemon is running before
642 * grabbing the packet.
644 if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP)
648 * Check our list of addresses, to see if the packet is for us.
649 * If we don't have any addresses, assume any unicast packet
650 * we receive might be for us (and let the upper layers deal
653 if (TAILQ_EMPTY(&V_in_ifaddrhead) &&
654 (m->m_flags & (M_MCAST|M_BCAST)) == 0)
658 * Enable a consistency check between the destination address
659 * and the arrival interface for a unicast packet (the RFC 1122
660 * strong ES model) if IP forwarding is disabled and the packet
661 * is not locally generated and the packet is not subject to
664 * XXX - Checking also should be disabled if the destination
665 * address is ipnat'ed to a different interface.
667 * XXX - Checking is incompatible with IP aliases added
668 * to the loopback interface instead of the interface where
669 * the packets are received.
671 * XXX - This is the case for carp vhost IPs as well so we
672 * insert a workaround. If the packet got here, we already
673 * checked with carp_iamatch() and carp_forus().
675 checkif = V_ip_checkinterface && (V_ipforwarding == 0) &&
676 ifp != NULL && ((ifp->if_flags & IFF_LOOPBACK) == 0) &&
677 ifp->if_carp == NULL && (dchg == 0);
680 * Check for exact addresses in the hash bucket.
682 /* IN_IFADDR_RLOCK(); */
683 LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
685 * If the address matches, verify that the packet
686 * arrived via the correct interface if checking is
689 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr &&
690 (!checkif || ia->ia_ifp == ifp)) {
691 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
692 counter_u64_add(ia->ia_ifa.ifa_ibytes,
694 /* IN_IFADDR_RUNLOCK(); */
698 /* IN_IFADDR_RUNLOCK(); */
701 * Check for broadcast addresses.
703 * Only accept broadcast packets that arrive via the matching
704 * interface. Reception of forwarded directed broadcasts would
705 * be handled via ip_forward() and ether_output() with the loopback
706 * into the stack for SIMPLEX interfaces handled by ether_output().
708 if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
710 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
711 if (ifa->ifa_addr->sa_family != AF_INET)
714 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
716 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
717 counter_u64_add(ia->ia_ifa.ifa_ibytes,
719 IF_ADDR_RUNLOCK(ifp);
723 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) {
724 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
725 counter_u64_add(ia->ia_ifa.ifa_ibytes,
727 IF_ADDR_RUNLOCK(ifp);
732 IF_ADDR_RUNLOCK(ifp);
735 /* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */
736 if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) {
737 IPSTAT_INC(ips_cantforward);
741 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
744 * If we are acting as a multicast router, all
745 * incoming multicast packets are passed to the
746 * kernel-level multicast forwarding function.
747 * The packet is returned (relatively) intact; if
748 * ip_mforward() returns a non-zero value, the packet
749 * must be discarded, else it may be accepted below.
751 if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) {
752 IPSTAT_INC(ips_cantforward);
758 * The process-level routing daemon needs to receive
759 * all multicast IGMP packets, whether or not this
760 * host belongs to their destination groups.
762 if (ip->ip_p == IPPROTO_IGMP)
764 IPSTAT_INC(ips_forward);
767 * Assume the packet is for us, to avoid prematurely taking
768 * a lock on the in_multi hash. Protocols must perform
769 * their own filtering and update statistics accordingly.
773 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
775 if (ip->ip_dst.s_addr == INADDR_ANY)
779 * Not for us; forward if possible and desirable.
781 if (V_ipforwarding == 0) {
782 IPSTAT_INC(ips_cantforward);
792 * IPSTEALTH: Process non-routing options only
793 * if the packet is destined for us.
795 if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1))
797 #endif /* IPSTEALTH */
800 * Attempt reassembly; if it succeeds, proceed.
801 * ip_reass() will return a different mbuf.
803 if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) {
804 /* XXXGL: shouldn't we save & set m_flags? */
808 ip = mtod(m, struct ip *);
809 /* Get the header length of the reassembled packet */
810 hlen = ip->ip_hl << 2;
813 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
814 if (IPSEC_ENABLED(ipv4)) {
815 if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
821 * Switch out to protocol's input routine.
823 IPSTAT_INC(ips_delivered);
825 (*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
832 * IP timer processing;
833 * if a timer expires on a reassembly
839 VNET_ITERATOR_DECL(vnet_iter);
841 VNET_LIST_RLOCK_NOSLEEP();
842 VNET_FOREACH(vnet_iter) {
843 CURVNET_SET(vnet_iter);
847 VNET_LIST_RUNLOCK_NOSLEEP();
853 VNET_ITERATOR_DECL(vnet_iter);
855 VNET_LIST_RLOCK_NOSLEEP();
856 VNET_FOREACH(vnet_iter) {
857 CURVNET_SET(vnet_iter);
861 VNET_LIST_RUNLOCK_NOSLEEP();
865 * The protocol to be inserted into ip_protox[] must be already registered
866 * in inetsw[], either statically or through pf_proto_register().
869 ipproto_register(short ipproto)
874 if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
875 return (EPROTONOSUPPORT);
878 * The protocol slot must not be occupied by another protocol
879 * already. An index pointing to IPPROTO_RAW is unused.
881 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
883 return (EPFNOSUPPORT);
884 if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */
887 /* Find the protocol position in inetsw[] and set the index. */
888 for (pr = inetdomain.dom_protosw;
889 pr < inetdomain.dom_protoswNPROTOSW; pr++) {
890 if (pr->pr_domain->dom_family == PF_INET &&
891 pr->pr_protocol && pr->pr_protocol == ipproto) {
892 ip_protox[pr->pr_protocol] = pr - inetsw;
896 return (EPROTONOSUPPORT);
900 ipproto_unregister(short ipproto)
905 if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
906 return (EPROTONOSUPPORT);
908 /* Check if the protocol was indeed registered. */
909 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
911 return (EPFNOSUPPORT);
912 if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */
915 /* Reset the protocol slot to IPPROTO_RAW. */
916 ip_protox[ipproto] = pr - inetsw;
920 u_char inetctlerrmap[PRC_NCMDS] = {
922 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
923 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
924 EMSGSIZE, EHOSTUNREACH, 0, 0,
925 0, 0, EHOSTUNREACH, 0,
926 ENOPROTOOPT, ECONNREFUSED
930 * Forward a packet. If some error occurs return the sender
931 * an icmp packet. Note we can't always generate a meaningful
932 * icmp message because icmp doesn't have a large enough repertoire
933 * of codes and types.
935 * If not forwarding, just drop the packet. This could be confusing
936 * if ipforwarding was zero but some routing protocol was advancing
937 * us as a gateway to somewhere. However, we must let the routing
938 * protocol deal with that.
940 * The srcrt parameter indicates whether the packet is being forwarded
941 * via a source route.
944 ip_forward(struct mbuf *m, int srcrt)
946 struct ip *ip = mtod(m, struct ip *);
947 struct in_ifaddr *ia;
949 struct sockaddr_in *sin;
952 int error, type = 0, code = 0, mtu = 0;
954 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
955 IPSTAT_INC(ips_cantforward);
963 ip->ip_ttl <= IPTTLDEC) {
964 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 0, 0);
968 bzero(&ro, sizeof(ro));
969 sin = (struct sockaddr_in *)&ro.ro_dst;
970 sin->sin_family = AF_INET;
971 sin->sin_len = sizeof(*sin);
972 sin->sin_addr = ip->ip_dst;
974 rtalloc_mpath_fib(&ro,
975 ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
978 in_rtalloc_ign(&ro, 0, M_GETFIB(m));
980 if (ro.ro_rt != NULL) {
981 ia = ifatoia(ro.ro_rt->rt_ifa);
982 ifa_ref(&ia->ia_ifa);
986 * Save the IP header and at most 8 bytes of the payload,
987 * in case we need to generate an ICMP message to the src.
989 * XXX this can be optimized a lot by saving the data in a local
990 * buffer on the stack (72 bytes at most), and only allocating the
991 * mbuf if really necessary. The vast majority of the packets
992 * are forwarded without having to send an ICMP back (either
993 * because unnecessary, or because rate limited), so we are
994 * really we are wasting a lot of work here.
996 * We don't use m_copym() because it might return a reference
997 * to a shared cluster. Both this function and ip_output()
998 * assume exclusive access to the IP header in `m', so any
999 * data in a cluster may change before we reach icmp_error().
1001 mcopy = m_gethdr(M_NOWAIT, m->m_type);
1002 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_NOWAIT)) {
1004 * It's probably ok if the pkthdr dup fails (because
1005 * the deep copy of the tag chain failed), but for now
1006 * be conservative and just discard the copy since
1007 * code below may some day want the tags.
1012 if (mcopy != NULL) {
1013 mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy));
1014 mcopy->m_pkthdr.len = mcopy->m_len;
1015 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1018 if (V_ipstealth == 0)
1020 ip->ip_ttl -= IPTTLDEC;
1021 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1022 if (IPSEC_ENABLED(ipv4)) {
1023 if ((error = IPSEC_FORWARD(ipv4, m)) != 0) {
1024 /* mbuf consumed by IPsec */
1026 if (error != EINPROGRESS)
1027 IPSTAT_INC(ips_cantforward);
1030 /* No IPsec processing required */
1034 * If forwarding packet using same interface that it came in on,
1035 * perhaps should send a redirect to sender to shortcut a hop.
1036 * Only send redirect if source is sending directly to us,
1037 * and if packet was not source routed (or has any options).
1038 * Also, don't send redirect if forwarding using a default route
1039 * or a route modified by a redirect.
1042 if (!srcrt && V_ipsendredirects &&
1043 ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) {
1048 if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1049 satosin(rt_key(rt))->sin_addr.s_addr != 0) {
1050 #define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa))
1051 u_long src = ntohl(ip->ip_src.s_addr);
1054 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1055 if (rt->rt_flags & RTF_GATEWAY)
1056 dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr;
1058 dest.s_addr = ip->ip_dst.s_addr;
1059 /* Router requirements says to only send host redirects */
1060 type = ICMP_REDIRECT;
1061 code = ICMP_REDIRECT_HOST;
1066 error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
1068 if (error == EMSGSIZE && ro.ro_rt)
1069 mtu = ro.ro_rt->rt_mtu;
1073 IPSTAT_INC(ips_cantforward);
1075 IPSTAT_INC(ips_forward);
1077 IPSTAT_INC(ips_redirectsent);
1082 ifa_free(&ia->ia_ifa);
1086 if (mcopy == NULL) {
1088 ifa_free(&ia->ia_ifa);
1094 case 0: /* forwarded, but need redirect */
1095 /* type, code set above */
1103 type = ICMP_UNREACH;
1104 code = ICMP_UNREACH_HOST;
1108 type = ICMP_UNREACH;
1109 code = ICMP_UNREACH_NEEDFRAG;
1111 * If the MTU was set before make sure we are below the
1113 * If the MTU wasn't set before use the interface mtu or
1114 * fall back to the next smaller mtu step compared to the
1115 * current packet size.
1119 mtu = min(mtu, ia->ia_ifp->if_mtu);
1122 mtu = ia->ia_ifp->if_mtu;
1124 mtu = ip_next_mtu(ntohs(ip->ip_len), 0);
1126 IPSTAT_INC(ips_cantfrag);
1130 case EACCES: /* ipfw denied packet */
1133 ifa_free(&ia->ia_ifa);
1137 ifa_free(&ia->ia_ifa);
1138 icmp_error(mcopy, type, code, dest.s_addr, mtu);
1141 #define CHECK_SO_CT(sp, ct) \
1142 (((sp->so_options & SO_TIMESTAMP) && (sp->so_ts_clock == ct)) ? 1 : 0)
1145 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1151 if ((inp->inp_socket->so_options & SO_BINTIME) ||
1152 CHECK_SO_CT(inp->inp_socket, SO_TS_BINTIME)) {
1153 struct bintime boottimebin, bt;
1154 struct timespec ts1;
1156 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1158 mbuf_tstmp2timespec(m, &ts1);
1159 timespec2bintime(&ts1, &bt);
1160 getboottimebin(&boottimebin);
1161 bintime_add(&bt, &boottimebin);
1165 *mp = sbcreatecontrol((caddr_t)&bt, sizeof(bt),
1166 SCM_BINTIME, SOL_SOCKET);
1168 mp = &(*mp)->m_next;
1172 if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME_MICRO)) {
1173 struct bintime boottimebin, bt1;
1174 struct timespec ts1;;
1177 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1179 mbuf_tstmp2timespec(m, &ts1);
1180 timespec2bintime(&ts1, &bt1);
1181 getboottimebin(&boottimebin);
1182 bintime_add(&bt1, &boottimebin);
1183 bintime2timeval(&bt1, &tv);
1187 *mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv),
1188 SCM_TIMESTAMP, SOL_SOCKET);
1190 mp = &(*mp)->m_next;
1193 } else if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME)) {
1194 struct bintime boottimebin;
1195 struct timespec ts, ts1;
1197 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1199 mbuf_tstmp2timespec(m, &ts);
1200 getboottimebin(&boottimebin);
1201 bintime2timespec(&boottimebin, &ts1);
1202 timespecadd(&ts, &ts1);
1206 *mp = sbcreatecontrol((caddr_t)&ts, sizeof(ts),
1207 SCM_REALTIME, SOL_SOCKET);
1209 mp = &(*mp)->m_next;
1212 } else if (CHECK_SO_CT(inp->inp_socket, SO_TS_MONOTONIC)) {
1215 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1217 mbuf_tstmp2timespec(m, &ts);
1220 *mp = sbcreatecontrol((caddr_t)&ts, sizeof(ts),
1221 SCM_MONOTONIC, SOL_SOCKET);
1223 mp = &(*mp)->m_next;
1227 if (stamped && (m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1229 struct sock_timestamp_info sti;
1231 bzero(&sti, sizeof(sti));
1232 sti.st_info_flags = ST_INFO_HW;
1233 if ((m->m_flags & M_TSTMP_HPREC) != 0)
1234 sti.st_info_flags |= ST_INFO_HW_HPREC;
1235 *mp = sbcreatecontrol((caddr_t)&sti, sizeof(sti), SCM_TIME_INFO,
1238 mp = &(*mp)->m_next;
1240 if (inp->inp_flags & INP_RECVDSTADDR) {
1241 *mp = sbcreatecontrol((caddr_t)&ip->ip_dst,
1242 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1244 mp = &(*mp)->m_next;
1246 if (inp->inp_flags & INP_RECVTTL) {
1247 *mp = sbcreatecontrol((caddr_t)&ip->ip_ttl,
1248 sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
1250 mp = &(*mp)->m_next;
1254 * Moving these out of udp_input() made them even more broken
1255 * than they already were.
1257 /* options were tossed already */
1258 if (inp->inp_flags & INP_RECVOPTS) {
1259 *mp = sbcreatecontrol((caddr_t)opts_deleted_above,
1260 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1262 mp = &(*mp)->m_next;
1264 /* ip_srcroute doesn't do what we want here, need to fix */
1265 if (inp->inp_flags & INP_RECVRETOPTS) {
1266 *mp = sbcreatecontrol((caddr_t)ip_srcroute(m),
1267 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1269 mp = &(*mp)->m_next;
1272 if (inp->inp_flags & INP_RECVIF) {
1275 struct sockaddr_dl sdl;
1278 struct sockaddr_dl *sdp;
1279 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1281 if ((ifp = m->m_pkthdr.rcvif) &&
1282 ifp->if_index && ifp->if_index <= V_if_index) {
1283 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1285 * Change our mind and don't try copy.
1287 if (sdp->sdl_family != AF_LINK ||
1288 sdp->sdl_len > sizeof(sdlbuf)) {
1291 bcopy(sdp, sdl2, sdp->sdl_len);
1295 offsetof(struct sockaddr_dl, sdl_data[0]);
1296 sdl2->sdl_family = AF_LINK;
1297 sdl2->sdl_index = 0;
1298 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1300 *mp = sbcreatecontrol((caddr_t)sdl2, sdl2->sdl_len,
1301 IP_RECVIF, IPPROTO_IP);
1303 mp = &(*mp)->m_next;
1305 if (inp->inp_flags & INP_RECVTOS) {
1306 *mp = sbcreatecontrol((caddr_t)&ip->ip_tos,
1307 sizeof(u_char), IP_RECVTOS, IPPROTO_IP);
1309 mp = &(*mp)->m_next;
1312 if (inp->inp_flags2 & INP_RECVFLOWID) {
1313 uint32_t flowid, flow_type;
1315 flowid = m->m_pkthdr.flowid;
1316 flow_type = M_HASHTYPE_GET(m);
1319 * XXX should handle the failure of one or the
1320 * other - don't populate both?
1322 *mp = sbcreatecontrol((caddr_t) &flowid,
1323 sizeof(uint32_t), IP_FLOWID, IPPROTO_IP);
1325 mp = &(*mp)->m_next;
1326 *mp = sbcreatecontrol((caddr_t) &flow_type,
1327 sizeof(uint32_t), IP_FLOWTYPE, IPPROTO_IP);
1329 mp = &(*mp)->m_next;
1333 if (inp->inp_flags2 & INP_RECVRSSBUCKETID) {
1334 uint32_t flowid, flow_type;
1335 uint32_t rss_bucketid;
1337 flowid = m->m_pkthdr.flowid;
1338 flow_type = M_HASHTYPE_GET(m);
1340 if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) {
1341 *mp = sbcreatecontrol((caddr_t) &rss_bucketid,
1342 sizeof(uint32_t), IP_RSSBUCKETID, IPPROTO_IP);
1344 mp = &(*mp)->m_next;
1351 * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
1352 * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
1353 * locking. This code remains in ip_input.c as ip_mroute.c is optionally
1356 static VNET_DEFINE(int, ip_rsvp_on);
1357 VNET_DEFINE(struct socket *, ip_rsvpd);
1359 #define V_ip_rsvp_on VNET(ip_rsvp_on)
1362 ip_rsvp_init(struct socket *so)
1365 if (so->so_type != SOCK_RAW ||
1366 so->so_proto->pr_protocol != IPPROTO_RSVP)
1369 if (V_ip_rsvpd != NULL)
1374 * This may seem silly, but we need to be sure we don't over-increment
1375 * the RSVP counter, in case something slips up.
1377 if (!V_ip_rsvp_on) {
1391 * This may seem silly, but we need to be sure we don't over-decrement
1392 * the RSVP counter, in case something slips up.
1402 rsvp_input(struct mbuf **mp, int *offp, int proto)
1409 if (rsvp_input_p) { /* call the real one if loaded */
1411 rsvp_input_p(mp, offp, proto);
1412 return (IPPROTO_DONE);
1415 /* Can still get packets with rsvp_on = 0 if there is a local member
1416 * of the group to which the RSVP packet is addressed. But in this
1417 * case we want to throw the packet away.
1422 return (IPPROTO_DONE);
1425 if (V_ip_rsvpd != NULL) {
1427 rip_input(mp, offp, proto);
1428 return (IPPROTO_DONE);
1430 /* Drop the packet */
1432 return (IPPROTO_DONE);