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 | CTLFLAG_MPSAFE, 0, 0,
234 sysctl_netinet_intr_queue_maxlen, "I",
235 "Maximum size of the IP input queue");
238 sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS)
240 u_int64_t qdrops_long;
243 netisr_getqdrops(&ip_nh, &qdrops_long);
244 qdrops = qdrops_long;
245 error = sysctl_handle_int(oidp, &qdrops, 0, req);
246 if (error || !req->newptr)
250 netisr_clearqdrops(&ip_nh);
254 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops,
255 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE,
256 0, 0, sysctl_netinet_intr_queue_drops, "I",
257 "Number of packets dropped from the IP input queue");
261 sysctl_netinet_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS)
265 netisr_getqlimit(&ip_direct_nh, &qlimit);
266 error = sysctl_handle_int(oidp, &qlimit, 0, req);
267 if (error || !req->newptr)
271 return (netisr_setqlimit(&ip_direct_nh, qlimit));
273 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRDQMAXLEN, intr_direct_queue_maxlen,
274 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
275 0, 0, sysctl_netinet_intr_direct_queue_maxlen,
276 "I", "Maximum size of the IP direct input queue");
279 sysctl_netinet_intr_direct_queue_drops(SYSCTL_HANDLER_ARGS)
281 u_int64_t qdrops_long;
284 netisr_getqdrops(&ip_direct_nh, &qdrops_long);
285 qdrops = qdrops_long;
286 error = sysctl_handle_int(oidp, &qdrops, 0, req);
287 if (error || !req->newptr)
291 netisr_clearqdrops(&ip_direct_nh);
295 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRDQDROPS, intr_direct_queue_drops,
296 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, 0, 0,
297 sysctl_netinet_intr_direct_queue_drops, "I",
298 "Number of packets dropped from the IP direct input queue");
302 * IP initialization: fill in IP protocol switch table.
303 * All protocols not implemented in kernel go to raw IP protocol handler.
308 struct pfil_head_args args;
312 CK_STAILQ_INIT(&V_in_ifaddrhead);
313 V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask);
315 /* Initialize IP reassembly queue. */
318 /* Initialize packet filter hooks. */
319 args.pa_version = PFIL_VERSION;
320 args.pa_flags = PFIL_IN | PFIL_OUT;
321 args.pa_type = PFIL_TYPE_IP4;
322 args.pa_headname = PFIL_INET_NAME;
323 V_inet_pfil_head = pfil_head_register(&args);
325 if (hhook_head_register(HHOOK_TYPE_IPSEC_IN, AF_INET,
326 &V_ipsec_hhh_in[HHOOK_IPSEC_INET],
327 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
328 printf("%s: WARNING: unable to register input helper hook\n",
330 if (hhook_head_register(HHOOK_TYPE_IPSEC_OUT, AF_INET,
331 &V_ipsec_hhh_out[HHOOK_IPSEC_INET],
332 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
333 printf("%s: WARNING: unable to register output helper hook\n",
336 /* Skip initialization of globals for non-default instances. */
338 if (!IS_DEFAULT_VNET(curvnet)) {
339 netisr_register_vnet(&ip_nh);
341 netisr_register_vnet(&ip_direct_nh);
347 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
349 panic("ip_init: PF_INET not found");
351 /* Initialize the entire ip_protox[] array to IPPROTO_RAW. */
352 for (i = 0; i < IPPROTO_MAX; i++)
353 ip_protox[i] = pr - inetsw;
355 * Cycle through IP protocols and put them into the appropriate place
358 for (pr = inetdomain.dom_protosw;
359 pr < inetdomain.dom_protoswNPROTOSW; pr++)
360 if (pr->pr_domain->dom_family == PF_INET &&
361 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
362 /* Be careful to only index valid IP protocols. */
363 if (pr->pr_protocol < IPPROTO_MAX)
364 ip_protox[pr->pr_protocol] = pr - inetsw;
367 netisr_register(&ip_nh);
369 netisr_register(&ip_direct_nh);
375 ip_destroy(void *unused __unused)
381 netisr_unregister_vnet(&ip_direct_nh);
383 netisr_unregister_vnet(&ip_nh);
385 pfil_head_unregister(V_inet_pfil_head);
386 error = hhook_head_deregister(V_ipsec_hhh_in[HHOOK_IPSEC_INET]);
388 printf("%s: WARNING: unable to deregister input helper hook "
389 "type HHOOK_TYPE_IPSEC_IN, id HHOOK_IPSEC_INET: "
390 "error %d returned\n", __func__, error);
392 error = hhook_head_deregister(V_ipsec_hhh_out[HHOOK_IPSEC_INET]);
394 printf("%s: WARNING: unable to deregister output helper hook "
395 "type HHOOK_TYPE_IPSEC_OUT, id HHOOK_IPSEC_INET: "
396 "error %d returned\n", __func__, error);
399 /* Remove the IPv4 addresses from all interfaces. */
402 /* Make sure the IPv4 routes are gone as well. */
404 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link)
405 rt_flushifroutes_af(ifp, AF_INET);
408 /* Destroy IP reassembly queue. */
411 /* Cleanup in_ifaddr hash table; should be empty. */
412 hashdestroy(V_in_ifaddrhashtbl, M_IFADDR, V_in_ifaddrhmask);
415 VNET_SYSUNINIT(ip, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip_destroy, NULL);
420 * IP direct input routine.
422 * This is called when reinjecting completed fragments where
423 * all of the previous checking and book-keeping has been done.
426 ip_direct_input(struct mbuf *m)
431 ip = mtod(m, struct ip *);
432 hlen = ip->ip_hl << 2;
434 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
435 if (IPSEC_ENABLED(ipv4)) {
436 if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
440 IPSTAT_INC(ips_delivered);
441 (*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
447 * Ip input routine. Checksum and byte swap header. If fragmented
448 * try to reassemble. Process options. Pass to next level.
451 ip_input(struct mbuf *m)
453 struct rm_priotracker in_ifa_tracker;
454 struct ip *ip = NULL;
455 struct in_ifaddr *ia = NULL;
458 int checkif, hlen = 0;
459 uint16_t sum, ip_len;
460 int dchg = 0; /* dest changed after fw */
461 struct in_addr odst; /* original dst address */
466 if (m->m_flags & M_FASTFWD_OURS) {
467 m->m_flags &= ~M_FASTFWD_OURS;
468 /* Set up some basics that will be used later. */
469 ip = mtod(m, struct ip *);
470 hlen = ip->ip_hl << 2;
471 ip_len = ntohs(ip->ip_len);
475 IPSTAT_INC(ips_total);
477 if (m->m_pkthdr.len < sizeof(struct ip))
480 if (m->m_len < sizeof (struct ip) &&
481 (m = m_pullup(m, sizeof (struct ip))) == NULL) {
482 IPSTAT_INC(ips_toosmall);
485 ip = mtod(m, struct ip *);
487 if (ip->ip_v != IPVERSION) {
488 IPSTAT_INC(ips_badvers);
492 hlen = ip->ip_hl << 2;
493 if (hlen < sizeof(struct ip)) { /* minimum header length */
494 IPSTAT_INC(ips_badhlen);
497 if (hlen > m->m_len) {
498 if ((m = m_pullup(m, hlen)) == NULL) {
499 IPSTAT_INC(ips_badhlen);
502 ip = mtod(m, struct ip *);
505 IP_PROBE(receive, NULL, NULL, ip, m->m_pkthdr.rcvif, ip, NULL);
507 /* IN_LOOPBACK must not appear on the wire - RFC1122 */
508 ifp = m->m_pkthdr.rcvif;
509 if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) ||
510 IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) {
511 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
512 IPSTAT_INC(ips_badaddr);
517 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
518 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
520 if (hlen == sizeof(struct ip)) {
521 sum = in_cksum_hdr(ip);
523 sum = in_cksum(m, hlen);
527 IPSTAT_INC(ips_badsum);
532 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
533 /* packet is dropped by traffic conditioner */
537 ip_len = ntohs(ip->ip_len);
539 IPSTAT_INC(ips_badlen);
544 * Check that the amount of data in the buffers
545 * is as at least much as the IP header would have us expect.
546 * Trim mbufs if longer than we expect.
547 * Drop packet if shorter than we expect.
549 if (m->m_pkthdr.len < ip_len) {
551 IPSTAT_INC(ips_tooshort);
554 if (m->m_pkthdr.len > ip_len) {
555 if (m->m_len == m->m_pkthdr.len) {
557 m->m_pkthdr.len = ip_len;
559 m_adj(m, ip_len - m->m_pkthdr.len);
563 * Try to forward the packet, but if we fail continue.
564 * ip_tryforward() does not generate redirects, so fall
565 * through to normal processing if redirects are required.
566 * ip_tryforward() does inbound and outbound packet firewall
567 * processing. If firewall has decided that destination becomes
568 * our local address, it sets M_FASTFWD_OURS flag. In this
569 * case skip another inbound firewall processing and update
572 if (V_ipforwarding != 0 && V_ipsendredirects == 0
573 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
574 && (!IPSEC_ENABLED(ipv4) ||
575 IPSEC_CAPS(ipv4, m, IPSEC_CAP_OPERABLE) == 0)
578 if ((m = ip_tryforward(m)) == NULL)
580 if (m->m_flags & M_FASTFWD_OURS) {
581 m->m_flags &= ~M_FASTFWD_OURS;
582 ip = mtod(m, struct ip *);
587 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
589 * Bypass packet filtering for packets previously handled by IPsec.
591 if (IPSEC_ENABLED(ipv4) &&
592 IPSEC_CAPS(ipv4, m, IPSEC_CAP_BYPASS_FILTER) != 0)
597 * Run through list of hooks for input packets.
599 * NB: Beware of the destination address changing (e.g.
600 * by NAT rewriting). When this happens, tell
601 * ip_forward to do the right thing.
604 /* Jump over all PFIL processing if hooks are not active. */
605 if (!PFIL_HOOKED_IN(V_inet_pfil_head))
609 if (pfil_run_hooks(V_inet_pfil_head, &m, ifp, PFIL_IN, NULL) !=
612 if (m == NULL) /* consumed by filter */
615 ip = mtod(m, struct ip *);
616 dchg = (odst.s_addr != ip->ip_dst.s_addr);
617 ifp = m->m_pkthdr.rcvif;
619 if (m->m_flags & M_FASTFWD_OURS) {
620 m->m_flags &= ~M_FASTFWD_OURS;
623 if (m->m_flags & M_IP_NEXTHOP) {
624 if (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) {
626 * Directly ship the packet on. This allows
627 * forwarding packets originally destined to us
628 * to some other directly connected host.
637 * Process options and, if not destined for us,
638 * ship it on. ip_dooptions returns 1 when an
639 * error was detected (causing an icmp message
640 * to be sent and the original packet to be freed).
642 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
645 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
646 * matter if it is destined to another node, or whether it is
647 * a multicast one, RSVP wants it! and prevents it from being forwarded
648 * anywhere else. Also checks if the rsvp daemon is running before
649 * grabbing the packet.
651 if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP)
655 * Check our list of addresses, to see if the packet is for us.
656 * If we don't have any addresses, assume any unicast packet
657 * we receive might be for us (and let the upper layers deal
660 if (CK_STAILQ_EMPTY(&V_in_ifaddrhead) &&
661 (m->m_flags & (M_MCAST|M_BCAST)) == 0)
665 * Enable a consistency check between the destination address
666 * and the arrival interface for a unicast packet (the RFC 1122
667 * strong ES model) if IP forwarding is disabled and the packet
668 * is not locally generated and the packet is not subject to
671 * XXX - Checking also should be disabled if the destination
672 * address is ipnat'ed to a different interface.
674 * XXX - Checking is incompatible with IP aliases added
675 * to the loopback interface instead of the interface where
676 * the packets are received.
678 * XXX - This is the case for carp vhost IPs as well so we
679 * insert a workaround. If the packet got here, we already
680 * checked with carp_iamatch() and carp_forus().
682 checkif = V_ip_checkinterface && (V_ipforwarding == 0) &&
683 ifp != NULL && ((ifp->if_flags & IFF_LOOPBACK) == 0) &&
684 ifp->if_carp == NULL && (dchg == 0);
687 * Check for exact addresses in the hash bucket.
689 IN_IFADDR_RLOCK(&in_ifa_tracker);
690 LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
692 * If the address matches, verify that the packet
693 * arrived via the correct interface if checking is
696 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr &&
697 (!checkif || ia->ia_ifp == ifp)) {
698 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
699 counter_u64_add(ia->ia_ifa.ifa_ibytes,
701 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
705 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
708 * Check for broadcast addresses.
710 * Only accept broadcast packets that arrive via the matching
711 * interface. Reception of forwarded directed broadcasts would
712 * be handled via ip_forward() and ether_output() with the loopback
713 * into the stack for SIMPLEX interfaces handled by ether_output().
715 if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
716 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
717 if (ifa->ifa_addr->sa_family != AF_INET)
720 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
722 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
723 counter_u64_add(ia->ia_ifa.ifa_ibytes,
728 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) {
729 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
730 counter_u64_add(ia->ia_ifa.ifa_ibytes,
738 /* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */
739 if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) {
740 IPSTAT_INC(ips_cantforward);
744 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
747 * If we are acting as a multicast router, all
748 * incoming multicast packets are passed to the
749 * kernel-level multicast forwarding function.
750 * The packet is returned (relatively) intact; if
751 * ip_mforward() returns a non-zero value, the packet
752 * must be discarded, else it may be accepted below.
754 if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) {
755 IPSTAT_INC(ips_cantforward);
761 * The process-level routing daemon needs to receive
762 * all multicast IGMP packets, whether or not this
763 * host belongs to their destination groups.
765 if (ip->ip_p == IPPROTO_IGMP)
767 IPSTAT_INC(ips_forward);
770 * Assume the packet is for us, to avoid prematurely taking
771 * a lock on the in_multi hash. Protocols must perform
772 * their own filtering and update statistics accordingly.
776 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
778 if (ip->ip_dst.s_addr == INADDR_ANY)
782 * Not for us; forward if possible and desirable.
784 if (V_ipforwarding == 0) {
785 IPSTAT_INC(ips_cantforward);
795 * IPSTEALTH: Process non-routing options only
796 * if the packet is destined for us.
798 if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1))
800 #endif /* IPSTEALTH */
803 * Attempt reassembly; if it succeeds, proceed.
804 * ip_reass() will return a different mbuf.
806 if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) {
807 /* XXXGL: shouldn't we save & set m_flags? */
811 ip = mtod(m, struct ip *);
812 /* Get the header length of the reassembled packet */
813 hlen = ip->ip_hl << 2;
816 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
817 if (IPSEC_ENABLED(ipv4)) {
818 if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
824 * Switch out to protocol's input routine.
826 IPSTAT_INC(ips_delivered);
828 (*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
835 * IP timer processing;
836 * if a timer expires on a reassembly
842 VNET_ITERATOR_DECL(vnet_iter);
844 VNET_LIST_RLOCK_NOSLEEP();
845 VNET_FOREACH(vnet_iter) {
846 CURVNET_SET(vnet_iter);
850 VNET_LIST_RUNLOCK_NOSLEEP();
856 VNET_ITERATOR_DECL(vnet_iter);
858 VNET_LIST_RLOCK_NOSLEEP();
859 VNET_FOREACH(vnet_iter) {
860 CURVNET_SET(vnet_iter);
864 VNET_LIST_RUNLOCK_NOSLEEP();
868 * The protocol to be inserted into ip_protox[] must be already registered
869 * in inetsw[], either statically or through pf_proto_register().
872 ipproto_register(short ipproto)
877 if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
878 return (EPROTONOSUPPORT);
881 * The protocol slot must not be occupied by another protocol
882 * already. An index pointing to IPPROTO_RAW is unused.
884 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
886 return (EPFNOSUPPORT);
887 if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */
890 /* Find the protocol position in inetsw[] and set the index. */
891 for (pr = inetdomain.dom_protosw;
892 pr < inetdomain.dom_protoswNPROTOSW; pr++) {
893 if (pr->pr_domain->dom_family == PF_INET &&
894 pr->pr_protocol && pr->pr_protocol == ipproto) {
895 ip_protox[pr->pr_protocol] = pr - inetsw;
899 return (EPROTONOSUPPORT);
903 ipproto_unregister(short ipproto)
908 if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
909 return (EPROTONOSUPPORT);
911 /* Check if the protocol was indeed registered. */
912 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
914 return (EPFNOSUPPORT);
915 if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */
918 /* Reset the protocol slot to IPPROTO_RAW. */
919 ip_protox[ipproto] = pr - inetsw;
923 u_char inetctlerrmap[PRC_NCMDS] = {
925 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
926 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
927 EMSGSIZE, EHOSTUNREACH, 0, 0,
928 0, 0, EHOSTUNREACH, 0,
929 ENOPROTOOPT, ECONNREFUSED
933 * Forward a packet. If some error occurs return the sender
934 * an icmp packet. Note we can't always generate a meaningful
935 * icmp message because icmp doesn't have a large enough repertoire
936 * of codes and types.
938 * If not forwarding, just drop the packet. This could be confusing
939 * if ipforwarding was zero but some routing protocol was advancing
940 * us as a gateway to somewhere. However, we must let the routing
941 * protocol deal with that.
943 * The srcrt parameter indicates whether the packet is being forwarded
944 * via a source route.
947 ip_forward(struct mbuf *m, int srcrt)
949 struct ip *ip = mtod(m, struct ip *);
950 struct in_ifaddr *ia;
952 struct sockaddr_in *sin;
955 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));
985 if (ro.ro_rt != NULL) {
986 ia = ifatoia(ro.ro_rt->rt_ifa);
990 * Save the IP header and at most 8 bytes of the payload,
991 * in case we need to generate an ICMP message to the src.
993 * XXX this can be optimized a lot by saving the data in a local
994 * buffer on the stack (72 bytes at most), and only allocating the
995 * mbuf if really necessary. The vast majority of the packets
996 * are forwarded without having to send an ICMP back (either
997 * because unnecessary, or because rate limited), so we are
998 * really we are wasting a lot of work here.
1000 * We don't use m_copym() because it might return a reference
1001 * to a shared cluster. Both this function and ip_output()
1002 * assume exclusive access to the IP header in `m', so any
1003 * data in a cluster may change before we reach icmp_error().
1005 mcopy = m_gethdr(M_NOWAIT, m->m_type);
1006 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_NOWAIT)) {
1008 * It's probably ok if the pkthdr dup fails (because
1009 * the deep copy of the tag chain failed), but for now
1010 * be conservative and just discard the copy since
1011 * code below may some day want the tags.
1016 if (mcopy != NULL) {
1017 mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy));
1018 mcopy->m_pkthdr.len = mcopy->m_len;
1019 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1022 if (V_ipstealth == 0)
1024 ip->ip_ttl -= IPTTLDEC;
1025 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1026 if (IPSEC_ENABLED(ipv4)) {
1027 if ((error = IPSEC_FORWARD(ipv4, m)) != 0) {
1028 /* mbuf consumed by IPsec */
1030 if (error != EINPROGRESS)
1031 IPSTAT_INC(ips_cantforward);
1034 /* No IPsec processing required */
1038 * If forwarding packet using same interface that it came in on,
1039 * perhaps should send a redirect to sender to shortcut a hop.
1040 * Only send redirect if source is sending directly to us,
1041 * and if packet was not source routed (or has any options).
1042 * Also, don't send redirect if forwarding using a default route
1043 * or a route modified by a redirect.
1046 if (!srcrt && V_ipsendredirects &&
1047 ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) {
1052 if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1053 satosin(rt_key(rt))->sin_addr.s_addr != 0) {
1054 #define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa))
1055 u_long src = ntohl(ip->ip_src.s_addr);
1058 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1059 if (rt->rt_flags & RTF_GATEWAY)
1060 dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr;
1062 dest.s_addr = ip->ip_dst.s_addr;
1063 /* Router requirements says to only send host redirects */
1064 type = ICMP_REDIRECT;
1065 code = ICMP_REDIRECT_HOST;
1070 error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
1072 if (error == EMSGSIZE && ro.ro_rt)
1073 mtu = ro.ro_rt->rt_mtu;
1077 IPSTAT_INC(ips_cantforward);
1079 IPSTAT_INC(ips_forward);
1081 IPSTAT_INC(ips_redirectsent);
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 */
1134 icmp_error(mcopy, type, code, dest.s_addr, mtu);
1137 #define CHECK_SO_CT(sp, ct) \
1138 (((sp->so_options & SO_TIMESTAMP) && (sp->so_ts_clock == ct)) ? 1 : 0)
1141 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1147 if ((inp->inp_socket->so_options & SO_BINTIME) ||
1148 CHECK_SO_CT(inp->inp_socket, SO_TS_BINTIME)) {
1149 struct bintime boottimebin, bt;
1150 struct timespec ts1;
1152 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1154 mbuf_tstmp2timespec(m, &ts1);
1155 timespec2bintime(&ts1, &bt);
1156 getboottimebin(&boottimebin);
1157 bintime_add(&bt, &boottimebin);
1161 *mp = sbcreatecontrol((caddr_t)&bt, sizeof(bt),
1162 SCM_BINTIME, SOL_SOCKET);
1164 mp = &(*mp)->m_next;
1168 if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME_MICRO)) {
1169 struct bintime boottimebin, bt1;
1170 struct timespec ts1;
1173 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1175 mbuf_tstmp2timespec(m, &ts1);
1176 timespec2bintime(&ts1, &bt1);
1177 getboottimebin(&boottimebin);
1178 bintime_add(&bt1, &boottimebin);
1179 bintime2timeval(&bt1, &tv);
1183 *mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv),
1184 SCM_TIMESTAMP, SOL_SOCKET);
1186 mp = &(*mp)->m_next;
1189 } else if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME)) {
1190 struct bintime boottimebin;
1191 struct timespec ts, ts1;
1193 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1195 mbuf_tstmp2timespec(m, &ts);
1196 getboottimebin(&boottimebin);
1197 bintime2timespec(&boottimebin, &ts1);
1198 timespecadd(&ts, &ts1, &ts);
1202 *mp = sbcreatecontrol((caddr_t)&ts, sizeof(ts),
1203 SCM_REALTIME, SOL_SOCKET);
1205 mp = &(*mp)->m_next;
1208 } else if (CHECK_SO_CT(inp->inp_socket, SO_TS_MONOTONIC)) {
1211 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1213 mbuf_tstmp2timespec(m, &ts);
1216 *mp = sbcreatecontrol((caddr_t)&ts, sizeof(ts),
1217 SCM_MONOTONIC, SOL_SOCKET);
1219 mp = &(*mp)->m_next;
1223 if (stamped && (m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1225 struct sock_timestamp_info sti;
1227 bzero(&sti, sizeof(sti));
1228 sti.st_info_flags = ST_INFO_HW;
1229 if ((m->m_flags & M_TSTMP_HPREC) != 0)
1230 sti.st_info_flags |= ST_INFO_HW_HPREC;
1231 *mp = sbcreatecontrol((caddr_t)&sti, sizeof(sti), SCM_TIME_INFO,
1234 mp = &(*mp)->m_next;
1236 if (inp->inp_flags & INP_RECVDSTADDR) {
1237 *mp = sbcreatecontrol((caddr_t)&ip->ip_dst,
1238 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1240 mp = &(*mp)->m_next;
1242 if (inp->inp_flags & INP_RECVTTL) {
1243 *mp = sbcreatecontrol((caddr_t)&ip->ip_ttl,
1244 sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
1246 mp = &(*mp)->m_next;
1250 * Moving these out of udp_input() made them even more broken
1251 * than they already were.
1253 /* options were tossed already */
1254 if (inp->inp_flags & INP_RECVOPTS) {
1255 *mp = sbcreatecontrol((caddr_t)opts_deleted_above,
1256 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1258 mp = &(*mp)->m_next;
1260 /* ip_srcroute doesn't do what we want here, need to fix */
1261 if (inp->inp_flags & INP_RECVRETOPTS) {
1262 *mp = sbcreatecontrol((caddr_t)ip_srcroute(m),
1263 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1265 mp = &(*mp)->m_next;
1268 if (inp->inp_flags & INP_RECVIF) {
1271 struct sockaddr_dl sdl;
1274 struct sockaddr_dl *sdp;
1275 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1277 if ((ifp = m->m_pkthdr.rcvif) &&
1278 ifp->if_index && ifp->if_index <= V_if_index) {
1279 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1281 * Change our mind and don't try copy.
1283 if (sdp->sdl_family != AF_LINK ||
1284 sdp->sdl_len > sizeof(sdlbuf)) {
1287 bcopy(sdp, sdl2, sdp->sdl_len);
1291 offsetof(struct sockaddr_dl, sdl_data[0]);
1292 sdl2->sdl_family = AF_LINK;
1293 sdl2->sdl_index = 0;
1294 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1296 *mp = sbcreatecontrol((caddr_t)sdl2, sdl2->sdl_len,
1297 IP_RECVIF, IPPROTO_IP);
1299 mp = &(*mp)->m_next;
1301 if (inp->inp_flags & INP_RECVTOS) {
1302 *mp = sbcreatecontrol((caddr_t)&ip->ip_tos,
1303 sizeof(u_char), IP_RECVTOS, IPPROTO_IP);
1305 mp = &(*mp)->m_next;
1308 if (inp->inp_flags2 & INP_RECVFLOWID) {
1309 uint32_t flowid, flow_type;
1311 flowid = m->m_pkthdr.flowid;
1312 flow_type = M_HASHTYPE_GET(m);
1315 * XXX should handle the failure of one or the
1316 * other - don't populate both?
1318 *mp = sbcreatecontrol((caddr_t) &flowid,
1319 sizeof(uint32_t), IP_FLOWID, IPPROTO_IP);
1321 mp = &(*mp)->m_next;
1322 *mp = sbcreatecontrol((caddr_t) &flow_type,
1323 sizeof(uint32_t), IP_FLOWTYPE, IPPROTO_IP);
1325 mp = &(*mp)->m_next;
1329 if (inp->inp_flags2 & INP_RECVRSSBUCKETID) {
1330 uint32_t flowid, flow_type;
1331 uint32_t rss_bucketid;
1333 flowid = m->m_pkthdr.flowid;
1334 flow_type = M_HASHTYPE_GET(m);
1336 if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) {
1337 *mp = sbcreatecontrol((caddr_t) &rss_bucketid,
1338 sizeof(uint32_t), IP_RSSBUCKETID, IPPROTO_IP);
1340 mp = &(*mp)->m_next;
1347 * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
1348 * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
1349 * locking. This code remains in ip_input.c as ip_mroute.c is optionally
1352 VNET_DEFINE_STATIC(int, ip_rsvp_on);
1353 VNET_DEFINE(struct socket *, ip_rsvpd);
1355 #define V_ip_rsvp_on VNET(ip_rsvp_on)
1358 ip_rsvp_init(struct socket *so)
1361 if (so->so_type != SOCK_RAW ||
1362 so->so_proto->pr_protocol != IPPROTO_RSVP)
1365 if (V_ip_rsvpd != NULL)
1370 * This may seem silly, but we need to be sure we don't over-increment
1371 * the RSVP counter, in case something slips up.
1373 if (!V_ip_rsvp_on) {
1387 * This may seem silly, but we need to be sure we don't over-decrement
1388 * the RSVP counter, in case something slips up.
1398 rsvp_input(struct mbuf **mp, int *offp, int proto)
1405 if (rsvp_input_p) { /* call the real one if loaded */
1407 rsvp_input_p(mp, offp, proto);
1408 return (IPPROTO_DONE);
1411 /* Can still get packets with rsvp_on = 0 if there is a local member
1412 * of the group to which the RSVP packet is addressed. But in this
1413 * case we want to throw the packet away.
1418 return (IPPROTO_DONE);
1421 if (V_ip_rsvpd != NULL) {
1423 rip_input(mp, offp, proto);
1424 return (IPPROTO_DONE);
1426 /* Drop the packet */
1428 return (IPPROTO_DONE);