2 * SPDX-License-Identifier: BSD-3-Clause
4 * Copyright (c) 1982, 1986, 1988, 1993
5 * The Regents of the University of California. All rights reserved.
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 * 3. Neither the name of the University nor the names of its contributors
16 * may be used to endorse or promote products derived from this software
17 * without specific prior written permission.
19 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94
34 #include <sys/cdefs.h>
35 __FBSDID("$FreeBSD$");
37 #include "opt_bootp.h"
38 #include "opt_ipstealth.h"
39 #include "opt_ipsec.h"
40 #include "opt_route.h"
43 #include <sys/param.h>
44 #include <sys/systm.h>
45 #include <sys/hhook.h>
47 #include <sys/malloc.h>
48 #include <sys/domain.h>
49 #include <sys/protosw.h>
50 #include <sys/socket.h>
52 #include <sys/kernel.h>
54 #include <sys/rmlock.h>
55 #include <sys/rwlock.h>
57 #include <sys/syslog.h>
58 #include <sys/sysctl.h>
61 #include <net/if_types.h>
62 #include <net/if_var.h>
63 #include <net/if_dl.h>
65 #include <net/route.h>
66 #include <net/route/nhop.h>
67 #include <net/netisr.h>
68 #include <net/rss_config.h>
71 #include <netinet/in.h>
72 #include <netinet/in_kdtrace.h>
73 #include <netinet/in_systm.h>
74 #include <netinet/in_var.h>
75 #include <netinet/ip.h>
76 #include <netinet/in_fib.h>
77 #include <netinet/in_pcb.h>
78 #include <netinet/ip_var.h>
79 #include <netinet/ip_fw.h>
80 #include <netinet/ip_icmp.h>
81 #include <netinet/ip_options.h>
82 #include <machine/in_cksum.h>
83 #include <netinet/ip_carp.h>
84 #include <netinet/in_rss.h>
85 #include <netinet/ip_mroute.h>
87 #include <netipsec/ipsec_support.h>
89 #include <sys/socketvar.h>
91 #include <security/mac/mac_framework.h>
94 CTASSERT(sizeof(struct ip) == 20);
97 /* IP reassembly functions are defined in ip_reass.c. */
98 extern void ipreass_init(void);
99 extern void ipreass_drain(void);
100 extern void ipreass_slowtimo(void);
102 extern void ipreass_destroy(void);
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");
113 * Respond with an ICMP host redirect when we forward a packet out of
114 * the same interface on which it was received. See RFC 792.
116 VNET_DEFINE(int, ipsendredirects) = 1;
117 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_VNET | CTLFLAG_RW,
118 &VNET_NAME(ipsendredirects), 0,
119 "Enable sending IP redirects");
122 * XXX - Setting ip_checkinterface mostly implements the receive side of
123 * the Strong ES model described in RFC 1122, but since the routing table
124 * and transmit implementation do not implement the Strong ES model,
125 * setting this to 1 results in an odd hybrid.
127 * XXX - ip_checkinterface currently must be disabled if you use ipnat
128 * to translate the destination address to another local interface.
130 * XXX - ip_checkinterface must be disabled if you add IP aliases
131 * to the loopback interface instead of the interface where the
132 * packets for those addresses are received.
134 VNET_DEFINE_STATIC(int, ip_checkinterface);
135 #define V_ip_checkinterface VNET(ip_checkinterface)
136 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_VNET | CTLFLAG_RW,
137 &VNET_NAME(ip_checkinterface), 0,
138 "Verify packet arrives on correct interface");
140 VNET_DEFINE(pfil_head_t, inet_pfil_head); /* Packet filter hooks */
142 static struct netisr_handler ip_nh = {
144 .nh_handler = ip_input,
145 .nh_proto = NETISR_IP,
147 .nh_m2cpuid = rss_soft_m2cpuid_v4,
148 .nh_policy = NETISR_POLICY_CPU,
149 .nh_dispatch = NETISR_DISPATCH_HYBRID,
151 .nh_policy = NETISR_POLICY_FLOW,
157 * Directly dispatched frames are currently assumed
158 * to have a flowid already calculated.
160 * It should likely have something that assert it
161 * actually has valid flow details.
163 static struct netisr_handler ip_direct_nh = {
164 .nh_name = "ip_direct",
165 .nh_handler = ip_direct_input,
166 .nh_proto = NETISR_IP_DIRECT,
167 .nh_m2cpuid = rss_soft_m2cpuid_v4,
168 .nh_policy = NETISR_POLICY_CPU,
169 .nh_dispatch = NETISR_DISPATCH_HYBRID,
173 extern struct domain inetdomain;
174 extern struct protosw inetsw[];
175 u_char ip_protox[IPPROTO_MAX];
176 VNET_DEFINE(struct in_ifaddrhead, in_ifaddrhead); /* first inet address */
177 VNET_DEFINE(struct in_ifaddrhashhead *, in_ifaddrhashtbl); /* inet addr hash table */
178 VNET_DEFINE(u_long, in_ifaddrhmask); /* mask for hash table */
180 /* Make sure it is safe to use hashinit(9) on CK_LIST. */
181 CTASSERT(sizeof(struct in_ifaddrhashhead) == sizeof(LIST_HEAD(, in_addr)));
184 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
185 &ip_mtu, 0, "Default MTU");
189 VNET_DEFINE(int, ipstealth);
190 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_VNET | CTLFLAG_RW,
191 &VNET_NAME(ipstealth), 0,
192 "IP stealth mode, no TTL decrementation on forwarding");
196 * IP statistics are stored in the "array" of counter(9)s.
198 VNET_PCPUSTAT_DEFINE(struct ipstat, ipstat);
199 VNET_PCPUSTAT_SYSINIT(ipstat);
200 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, IPCTL_STATS, stats, struct ipstat, ipstat,
201 "IP statistics (struct ipstat, netinet/ip_var.h)");
204 VNET_PCPUSTAT_SYSUNINIT(ipstat);
208 * Kernel module interface for updating ipstat. The argument is an index
209 * into ipstat treated as an array.
212 kmod_ipstat_inc(int statnum)
215 counter_u64_add(VNET(ipstat)[statnum], 1);
219 kmod_ipstat_dec(int statnum)
222 counter_u64_add(VNET(ipstat)[statnum], -1);
226 sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS)
230 netisr_getqlimit(&ip_nh, &qlimit);
231 error = sysctl_handle_int(oidp, &qlimit, 0, req);
232 if (error || !req->newptr)
236 return (netisr_setqlimit(&ip_nh, qlimit));
238 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen,
239 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 0,
240 sysctl_netinet_intr_queue_maxlen, "I",
241 "Maximum size of the IP input queue");
244 sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS)
246 u_int64_t qdrops_long;
249 netisr_getqdrops(&ip_nh, &qdrops_long);
250 qdrops = qdrops_long;
251 error = sysctl_handle_int(oidp, &qdrops, 0, req);
252 if (error || !req->newptr)
256 netisr_clearqdrops(&ip_nh);
260 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops,
261 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE,
262 0, 0, sysctl_netinet_intr_queue_drops, "I",
263 "Number of packets dropped from the IP input queue");
267 sysctl_netinet_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS)
271 netisr_getqlimit(&ip_direct_nh, &qlimit);
272 error = sysctl_handle_int(oidp, &qlimit, 0, req);
273 if (error || !req->newptr)
277 return (netisr_setqlimit(&ip_direct_nh, qlimit));
279 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRDQMAXLEN, intr_direct_queue_maxlen,
280 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
281 0, 0, sysctl_netinet_intr_direct_queue_maxlen,
282 "I", "Maximum size of the IP direct input queue");
285 sysctl_netinet_intr_direct_queue_drops(SYSCTL_HANDLER_ARGS)
287 u_int64_t qdrops_long;
290 netisr_getqdrops(&ip_direct_nh, &qdrops_long);
291 qdrops = qdrops_long;
292 error = sysctl_handle_int(oidp, &qdrops, 0, req);
293 if (error || !req->newptr)
297 netisr_clearqdrops(&ip_direct_nh);
301 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRDQDROPS, intr_direct_queue_drops,
302 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, 0, 0,
303 sysctl_netinet_intr_direct_queue_drops, "I",
304 "Number of packets dropped from the IP direct input queue");
308 * IP initialization: fill in IP protocol switch table.
309 * All protocols not implemented in kernel go to raw IP protocol handler.
314 struct pfil_head_args args;
318 CK_STAILQ_INIT(&V_in_ifaddrhead);
319 V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask);
321 /* Initialize IP reassembly queue. */
324 /* Initialize packet filter hooks. */
325 args.pa_version = PFIL_VERSION;
326 args.pa_flags = PFIL_IN | PFIL_OUT;
327 args.pa_type = PFIL_TYPE_IP4;
328 args.pa_headname = PFIL_INET_NAME;
329 V_inet_pfil_head = pfil_head_register(&args);
331 if (hhook_head_register(HHOOK_TYPE_IPSEC_IN, AF_INET,
332 &V_ipsec_hhh_in[HHOOK_IPSEC_INET],
333 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
334 printf("%s: WARNING: unable to register input helper hook\n",
336 if (hhook_head_register(HHOOK_TYPE_IPSEC_OUT, AF_INET,
337 &V_ipsec_hhh_out[HHOOK_IPSEC_INET],
338 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
339 printf("%s: WARNING: unable to register output helper hook\n",
342 /* Skip initialization of globals for non-default instances. */
344 if (!IS_DEFAULT_VNET(curvnet)) {
345 netisr_register_vnet(&ip_nh);
347 netisr_register_vnet(&ip_direct_nh);
353 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
355 panic("ip_init: PF_INET not found");
357 /* Initialize the entire ip_protox[] array to IPPROTO_RAW. */
358 for (i = 0; i < IPPROTO_MAX; i++)
359 ip_protox[i] = pr - inetsw;
361 * Cycle through IP protocols and put them into the appropriate place
364 for (pr = inetdomain.dom_protosw;
365 pr < inetdomain.dom_protoswNPROTOSW; pr++)
366 if (pr->pr_domain->dom_family == PF_INET &&
367 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
368 /* Be careful to only index valid IP protocols. */
369 if (pr->pr_protocol < IPPROTO_MAX)
370 ip_protox[pr->pr_protocol] = pr - inetsw;
373 netisr_register(&ip_nh);
375 netisr_register(&ip_direct_nh);
381 ip_destroy(void *unused __unused)
386 netisr_unregister_vnet(&ip_direct_nh);
388 netisr_unregister_vnet(&ip_nh);
390 pfil_head_unregister(V_inet_pfil_head);
391 error = hhook_head_deregister(V_ipsec_hhh_in[HHOOK_IPSEC_INET]);
393 printf("%s: WARNING: unable to deregister input helper hook "
394 "type HHOOK_TYPE_IPSEC_IN, id HHOOK_IPSEC_INET: "
395 "error %d returned\n", __func__, error);
397 error = hhook_head_deregister(V_ipsec_hhh_out[HHOOK_IPSEC_INET]);
399 printf("%s: WARNING: unable to deregister output helper hook "
400 "type HHOOK_TYPE_IPSEC_OUT, id HHOOK_IPSEC_INET: "
401 "error %d returned\n", __func__, error);
404 /* Remove the IPv4 addresses from all interfaces. */
407 /* Make sure the IPv4 routes are gone as well. */
408 rib_flush_routes_family(AF_INET);
410 /* Destroy IP reassembly queue. */
413 /* Cleanup in_ifaddr hash table; should be empty. */
414 hashdestroy(V_in_ifaddrhashtbl, M_IFADDR, V_in_ifaddrhmask);
417 VNET_SYSUNINIT(ip, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip_destroy, NULL);
422 * IP direct input routine.
424 * This is called when reinjecting completed fragments where
425 * all of the previous checking and book-keeping has been done.
428 ip_direct_input(struct mbuf *m)
433 ip = mtod(m, struct ip *);
434 hlen = ip->ip_hl << 2;
436 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
437 if (IPSEC_ENABLED(ipv4)) {
438 if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
442 IPSTAT_INC(ips_delivered);
443 (*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
449 * Ip input routine. Checksum and byte swap header. If fragmented
450 * try to reassemble. Process options. Pass to next level.
453 ip_input(struct mbuf *m)
455 MROUTER_RLOCK_TRACKER;
456 struct ip *ip = NULL;
457 struct in_ifaddr *ia = NULL;
460 int checkif, hlen = 0;
461 uint16_t sum, ip_len;
462 int dchg = 0; /* dest changed after fw */
463 struct in_addr odst; /* original dst address */
468 if (m->m_flags & M_FASTFWD_OURS) {
469 m->m_flags &= ~M_FASTFWD_OURS;
470 /* Set up some basics that will be used later. */
471 ip = mtod(m, struct ip *);
472 hlen = ip->ip_hl << 2;
473 ip_len = ntohs(ip->ip_len);
477 IPSTAT_INC(ips_total);
479 if (m->m_pkthdr.len < sizeof(struct ip))
482 if (m->m_len < sizeof (struct ip) &&
483 (m = m_pullup(m, sizeof (struct ip))) == NULL) {
484 IPSTAT_INC(ips_toosmall);
487 ip = mtod(m, struct ip *);
489 if (ip->ip_v != IPVERSION) {
490 IPSTAT_INC(ips_badvers);
494 hlen = ip->ip_hl << 2;
495 if (hlen < sizeof(struct ip)) { /* minimum header length */
496 IPSTAT_INC(ips_badhlen);
499 if (hlen > m->m_len) {
500 if ((m = m_pullup(m, hlen)) == NULL) {
501 IPSTAT_INC(ips_badhlen);
504 ip = mtod(m, struct ip *);
507 IP_PROBE(receive, NULL, NULL, ip, m->m_pkthdr.rcvif, ip, NULL);
509 /* IN_LOOPBACK must not appear on the wire - RFC1122 */
510 ifp = m->m_pkthdr.rcvif;
511 if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) ||
512 IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) {
513 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
514 IPSTAT_INC(ips_badaddr);
519 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
520 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
522 if (hlen == sizeof(struct ip)) {
523 sum = in_cksum_hdr(ip);
525 sum = in_cksum(m, hlen);
529 IPSTAT_INC(ips_badsum);
534 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
535 /* packet is dropped by traffic conditioner */
539 ip_len = ntohs(ip->ip_len);
541 IPSTAT_INC(ips_badlen);
546 * Check that the amount of data in the buffers
547 * is as at least much as the IP header would have us expect.
548 * Trim mbufs if longer than we expect.
549 * Drop packet if shorter than we expect.
551 if (m->m_pkthdr.len < ip_len) {
553 IPSTAT_INC(ips_tooshort);
556 if (m->m_pkthdr.len > ip_len) {
557 if (m->m_len == m->m_pkthdr.len) {
559 m->m_pkthdr.len = ip_len;
561 m_adj(m, ip_len - m->m_pkthdr.len);
565 * Try to forward the packet, but if we fail continue.
566 * ip_tryforward() does not generate redirects, so fall
567 * through to normal processing if redirects are required.
568 * ip_tryforward() does inbound and outbound packet firewall
569 * processing. If firewall has decided that destination becomes
570 * our local address, it sets M_FASTFWD_OURS flag. In this
571 * case skip another inbound firewall processing and update
574 if (V_ipforwarding != 0
575 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
576 && (!IPSEC_ENABLED(ipv4) ||
577 IPSEC_CAPS(ipv4, m, IPSEC_CAP_OPERABLE) == 0)
580 if ((m = ip_tryforward(m)) == NULL)
582 if (m->m_flags & M_FASTFWD_OURS) {
583 m->m_flags &= ~M_FASTFWD_OURS;
584 ip = mtod(m, struct ip *);
589 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
591 * Bypass packet filtering for packets previously handled by IPsec.
593 if (IPSEC_ENABLED(ipv4) &&
594 IPSEC_CAPS(ipv4, m, IPSEC_CAP_BYPASS_FILTER) != 0)
599 * Run through list of hooks for input packets.
601 * NB: Beware of the destination address changing (e.g.
602 * by NAT rewriting). When this happens, tell
603 * ip_forward to do the right thing.
606 /* Jump over all PFIL processing if hooks are not active. */
607 if (!PFIL_HOOKED_IN(V_inet_pfil_head))
611 if (pfil_run_hooks(V_inet_pfil_head, &m, ifp, PFIL_IN, NULL) !=
614 if (m == NULL) /* consumed by filter */
617 ip = mtod(m, struct ip *);
618 dchg = (odst.s_addr != ip->ip_dst.s_addr);
619 ifp = m->m_pkthdr.rcvif;
621 if (m->m_flags & M_FASTFWD_OURS) {
622 m->m_flags &= ~M_FASTFWD_OURS;
625 if (m->m_flags & M_IP_NEXTHOP) {
626 if (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) {
628 * Directly ship the packet on. This allows
629 * forwarding packets originally destined to us
630 * to some other directly connected host.
639 * Process options and, if not destined for us,
640 * ship it on. ip_dooptions returns 1 when an
641 * error was detected (causing an icmp message
642 * to be sent and the original packet to be freed).
644 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
647 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
648 * matter if it is destined to another node, or whether it is
649 * a multicast one, RSVP wants it! and prevents it from being forwarded
650 * anywhere else. Also checks if the rsvp daemon is running before
651 * grabbing the packet.
653 if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP)
657 * Check our list of addresses, to see if the packet is for us.
658 * If we don't have any addresses, assume any unicast packet
659 * we receive might be for us (and let the upper layers deal
662 if (CK_STAILQ_EMPTY(&V_in_ifaddrhead) &&
663 (m->m_flags & (M_MCAST|M_BCAST)) == 0)
667 * Enable a consistency check between the destination address
668 * and the arrival interface for a unicast packet (the RFC 1122
669 * strong ES model) if IP forwarding is disabled and the packet
670 * is not locally generated and the packet is not subject to
673 * XXX - Checking also should be disabled if the destination
674 * address is ipnat'ed to a different interface.
676 * XXX - Checking is incompatible with IP aliases added
677 * to the loopback interface instead of the interface where
678 * the packets are received.
680 * XXX - This is the case for carp vhost IPs as well so we
681 * insert a workaround. If the packet got here, we already
682 * checked with carp_iamatch() and carp_forus().
684 checkif = V_ip_checkinterface && (V_ipforwarding == 0) &&
685 ifp != NULL && ((ifp->if_flags & IFF_LOOPBACK) == 0) &&
686 ifp->if_carp == NULL && (dchg == 0);
689 * Check for exact addresses in the hash bucket.
691 CK_LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
693 * If the address matches, verify that the packet
694 * arrived via the correct interface if checking is
697 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr &&
698 (!checkif || ia->ia_ifp == ifp)) {
699 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
700 counter_u64_add(ia->ia_ifa.ifa_ibytes,
707 * Check for broadcast addresses.
709 * Only accept broadcast packets that arrive via the matching
710 * interface. Reception of forwarded directed broadcasts would
711 * be handled via ip_forward() and ether_output() with the loopback
712 * into the stack for SIMPLEX interfaces handled by ether_output().
714 if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
715 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
716 if (ifa->ifa_addr->sa_family != AF_INET)
719 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
721 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
722 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,
737 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
740 * RFC 3927 2.7: Do not forward multicast packets from
743 if (V_ip_mrouter && !IN_LINKLOCAL(ntohl(ip->ip_src.s_addr))) {
745 * If we are acting as a multicast router, all
746 * incoming multicast packets are passed to the
747 * kernel-level multicast forwarding function.
748 * The packet is returned (relatively) intact; if
749 * ip_mforward() returns a non-zero value, the packet
750 * must be discarded, else it may be accepted below.
752 if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) {
754 IPSTAT_INC(ips_cantforward);
760 * The process-level routing daemon needs to receive
761 * all multicast IGMP packets, whether or not this
762 * host belongs to their destination groups.
764 if (ip->ip_p == IPPROTO_IGMP) {
768 IPSTAT_INC(ips_forward);
772 * Assume the packet is for us, to avoid prematurely taking
773 * a lock on the in_multi hash. Protocols must perform
774 * their own filtering and update statistics accordingly.
778 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
780 if (ip->ip_dst.s_addr == INADDR_ANY)
782 /* RFC 3927 2.7: Do not forward packets to or from IN_LINKLOCAL. */
783 if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr)) ||
784 IN_LINKLOCAL(ntohl(ip->ip_src.s_addr))) {
785 IPSTAT_INC(ips_cantforward);
791 * Not for us; forward if possible and desirable.
793 if (V_ipforwarding == 0) {
794 IPSTAT_INC(ips_cantforward);
804 * IPSTEALTH: Process non-routing options only
805 * if the packet is destined for us.
807 if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1))
809 #endif /* IPSTEALTH */
812 * Attempt reassembly; if it succeeds, proceed.
813 * ip_reass() will return a different mbuf.
815 if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) {
816 /* XXXGL: shouldn't we save & set m_flags? */
820 ip = mtod(m, struct ip *);
821 /* Get the header length of the reassembled packet */
822 hlen = ip->ip_hl << 2;
825 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
826 if (IPSEC_ENABLED(ipv4)) {
827 if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
833 * Switch out to protocol's input routine.
835 IPSTAT_INC(ips_delivered);
837 (*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
844 * IP timer processing;
845 * if a timer expires on a reassembly
851 VNET_ITERATOR_DECL(vnet_iter);
853 VNET_LIST_RLOCK_NOSLEEP();
854 VNET_FOREACH(vnet_iter) {
855 CURVNET_SET(vnet_iter);
859 VNET_LIST_RUNLOCK_NOSLEEP();
865 VNET_ITERATOR_DECL(vnet_iter);
867 VNET_LIST_RLOCK_NOSLEEP();
868 VNET_FOREACH(vnet_iter) {
869 CURVNET_SET(vnet_iter);
873 VNET_LIST_RUNLOCK_NOSLEEP();
877 * The protocol to be inserted into ip_protox[] must be already registered
878 * in inetsw[], either statically or through pf_proto_register().
881 ipproto_register(short ipproto)
886 if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
887 return (EPROTONOSUPPORT);
890 * The protocol slot must not be occupied by another protocol
891 * already. An index pointing to IPPROTO_RAW is unused.
893 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
895 return (EPFNOSUPPORT);
896 if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */
899 /* Find the protocol position in inetsw[] and set the index. */
900 for (pr = inetdomain.dom_protosw;
901 pr < inetdomain.dom_protoswNPROTOSW; pr++) {
902 if (pr->pr_domain->dom_family == PF_INET &&
903 pr->pr_protocol && pr->pr_protocol == ipproto) {
904 ip_protox[pr->pr_protocol] = pr - inetsw;
908 return (EPROTONOSUPPORT);
912 ipproto_unregister(short ipproto)
917 if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
918 return (EPROTONOSUPPORT);
920 /* Check if the protocol was indeed registered. */
921 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
923 return (EPFNOSUPPORT);
924 if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */
927 /* Reset the protocol slot to IPPROTO_RAW. */
928 ip_protox[ipproto] = pr - inetsw;
932 u_char inetctlerrmap[PRC_NCMDS] = {
934 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
935 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
936 EMSGSIZE, EHOSTUNREACH, 0, 0,
937 0, 0, EHOSTUNREACH, 0,
938 ENOPROTOOPT, ECONNREFUSED
942 * Forward a packet. If some error occurs return the sender
943 * an icmp packet. Note we can't always generate a meaningful
944 * icmp message because icmp doesn't have a large enough repertoire
945 * of codes and types.
947 * If not forwarding, just drop the packet. This could be confusing
948 * if ipforwarding was zero but some routing protocol was advancing
949 * us as a gateway to somewhere. However, we must let the routing
950 * protocol deal with that.
952 * The srcrt parameter indicates whether the packet is being forwarded
953 * via a source route.
956 ip_forward(struct mbuf *m, int srcrt)
958 struct ip *ip = mtod(m, struct ip *);
959 struct in_ifaddr *ia;
961 struct sockaddr_in *sin;
965 int error, type = 0, code = 0, mtu = 0;
969 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
970 IPSTAT_INC(ips_cantforward);
978 ip->ip_ttl <= IPTTLDEC) {
979 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 0, 0);
983 bzero(&ro, sizeof(ro));
984 sin = (struct sockaddr_in *)&ro.ro_dst;
985 sin->sin_family = AF_INET;
986 sin->sin_len = sizeof(*sin);
987 sin->sin_addr = ip->ip_dst;
988 flowid = m->m_pkthdr.flowid;
989 ro.ro_nh = fib4_lookup(M_GETFIB(m), ip->ip_dst, 0, NHR_REF, flowid);
990 if (ro.ro_nh != NULL) {
991 ia = ifatoia(ro.ro_nh->nh_ifa);
995 * Save the IP header and at most 8 bytes of the payload,
996 * in case we need to generate an ICMP message to the src.
998 * XXX this can be optimized a lot by saving the data in a local
999 * buffer on the stack (72 bytes at most), and only allocating the
1000 * mbuf if really necessary. The vast majority of the packets
1001 * are forwarded without having to send an ICMP back (either
1002 * because unnecessary, or because rate limited), so we are
1003 * really we are wasting a lot of work here.
1005 * We don't use m_copym() because it might return a reference
1006 * to a shared cluster. Both this function and ip_output()
1007 * assume exclusive access to the IP header in `m', so any
1008 * data in a cluster may change before we reach icmp_error().
1010 mcopy = m_gethdr(M_NOWAIT, m->m_type);
1011 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_NOWAIT)) {
1013 * It's probably ok if the pkthdr dup fails (because
1014 * the deep copy of the tag chain failed), but for now
1015 * be conservative and just discard the copy since
1016 * code below may some day want the tags.
1021 if (mcopy != NULL) {
1022 mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy));
1023 mcopy->m_pkthdr.len = mcopy->m_len;
1024 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1027 if (V_ipstealth == 0)
1029 ip->ip_ttl -= IPTTLDEC;
1030 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1031 if (IPSEC_ENABLED(ipv4)) {
1032 if ((error = IPSEC_FORWARD(ipv4, m)) != 0) {
1033 /* mbuf consumed by IPsec */
1036 if (error != EINPROGRESS)
1037 IPSTAT_INC(ips_cantforward);
1040 /* No IPsec processing required */
1044 * If forwarding packet using same interface that it came in on,
1045 * perhaps should send a redirect to sender to shortcut a hop.
1046 * Only send redirect if source is sending directly to us,
1047 * and if packet was not source routed (or has any options).
1048 * Also, don't send redirect if forwarding using a default route
1049 * or a route modified by a redirect.
1052 if (!srcrt && V_ipsendredirects &&
1053 ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) {
1054 struct nhop_object *nh;
1058 if (nh != NULL && ((nh->nh_flags & (NHF_REDIRECT|NHF_DEFAULT)) == 0)) {
1059 struct in_ifaddr *nh_ia = (struct in_ifaddr *)(nh->nh_ifa);
1060 u_long src = ntohl(ip->ip_src.s_addr);
1062 if (nh_ia != NULL &&
1063 (src & nh_ia->ia_subnetmask) == nh_ia->ia_subnet) {
1064 /* Router requirements says to only send host redirects */
1065 type = ICMP_REDIRECT;
1066 code = ICMP_REDIRECT_HOST;
1067 if (nh->nh_flags & NHF_GATEWAY) {
1068 if (nh->gw_sa.sa_family == AF_INET)
1069 dest.s_addr = nh->gw4_sa.sin_addr.s_addr;
1070 else /* Do not redirect in case gw is AF_INET6 */
1073 dest.s_addr = ip->ip_dst.s_addr;
1078 error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
1080 if (error == EMSGSIZE && ro.ro_nh)
1081 mtu = ro.ro_nh->nh_mtu;
1085 IPSTAT_INC(ips_cantforward);
1087 IPSTAT_INC(ips_forward);
1089 IPSTAT_INC(ips_redirectsent);
1100 case 0: /* forwarded, but need redirect */
1101 /* type, code set above */
1109 type = ICMP_UNREACH;
1110 code = ICMP_UNREACH_HOST;
1114 type = ICMP_UNREACH;
1115 code = ICMP_UNREACH_NEEDFRAG;
1117 * If the MTU was set before make sure we are below the
1119 * If the MTU wasn't set before use the interface mtu or
1120 * fall back to the next smaller mtu step compared to the
1121 * current packet size.
1125 mtu = min(mtu, ia->ia_ifp->if_mtu);
1128 mtu = ia->ia_ifp->if_mtu;
1130 mtu = ip_next_mtu(ntohs(ip->ip_len), 0);
1132 IPSTAT_INC(ips_cantfrag);
1136 case EACCES: /* ipfw denied packet */
1140 icmp_error(mcopy, type, code, dest.s_addr, mtu);
1143 #define CHECK_SO_CT(sp, ct) \
1144 (((sp->so_options & SO_TIMESTAMP) && (sp->so_ts_clock == ct)) ? 1 : 0)
1147 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1153 if ((inp->inp_socket->so_options & SO_BINTIME) ||
1154 CHECK_SO_CT(inp->inp_socket, SO_TS_BINTIME)) {
1155 struct bintime boottimebin, bt;
1156 struct timespec ts1;
1158 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1160 mbuf_tstmp2timespec(m, &ts1);
1161 timespec2bintime(&ts1, &bt);
1162 getboottimebin(&boottimebin);
1163 bintime_add(&bt, &boottimebin);
1167 *mp = sbcreatecontrol((caddr_t)&bt, sizeof(bt),
1168 SCM_BINTIME, SOL_SOCKET);
1170 mp = &(*mp)->m_next;
1174 if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME_MICRO)) {
1175 struct bintime boottimebin, bt1;
1176 struct timespec ts1;
1179 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1181 mbuf_tstmp2timespec(m, &ts1);
1182 timespec2bintime(&ts1, &bt1);
1183 getboottimebin(&boottimebin);
1184 bintime_add(&bt1, &boottimebin);
1185 bintime2timeval(&bt1, &tv);
1189 *mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv),
1190 SCM_TIMESTAMP, SOL_SOCKET);
1192 mp = &(*mp)->m_next;
1195 } else if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME)) {
1196 struct bintime boottimebin;
1197 struct timespec ts, ts1;
1199 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1201 mbuf_tstmp2timespec(m, &ts);
1202 getboottimebin(&boottimebin);
1203 bintime2timespec(&boottimebin, &ts1);
1204 timespecadd(&ts, &ts1, &ts);
1208 *mp = sbcreatecontrol((caddr_t)&ts, sizeof(ts),
1209 SCM_REALTIME, SOL_SOCKET);
1211 mp = &(*mp)->m_next;
1214 } else if (CHECK_SO_CT(inp->inp_socket, SO_TS_MONOTONIC)) {
1217 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1219 mbuf_tstmp2timespec(m, &ts);
1222 *mp = sbcreatecontrol((caddr_t)&ts, sizeof(ts),
1223 SCM_MONOTONIC, SOL_SOCKET);
1225 mp = &(*mp)->m_next;
1229 if (stamped && (m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1231 struct sock_timestamp_info sti;
1233 bzero(&sti, sizeof(sti));
1234 sti.st_info_flags = ST_INFO_HW;
1235 if ((m->m_flags & M_TSTMP_HPREC) != 0)
1236 sti.st_info_flags |= ST_INFO_HW_HPREC;
1237 *mp = sbcreatecontrol((caddr_t)&sti, sizeof(sti), SCM_TIME_INFO,
1240 mp = &(*mp)->m_next;
1242 if (inp->inp_flags & INP_RECVDSTADDR) {
1243 *mp = sbcreatecontrol((caddr_t)&ip->ip_dst,
1244 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1246 mp = &(*mp)->m_next;
1248 if (inp->inp_flags & INP_RECVTTL) {
1249 *mp = sbcreatecontrol((caddr_t)&ip->ip_ttl,
1250 sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
1252 mp = &(*mp)->m_next;
1256 * Moving these out of udp_input() made them even more broken
1257 * than they already were.
1259 /* options were tossed already */
1260 if (inp->inp_flags & INP_RECVOPTS) {
1261 *mp = sbcreatecontrol((caddr_t)opts_deleted_above,
1262 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1264 mp = &(*mp)->m_next;
1266 /* ip_srcroute doesn't do what we want here, need to fix */
1267 if (inp->inp_flags & INP_RECVRETOPTS) {
1268 *mp = sbcreatecontrol((caddr_t)ip_srcroute(m),
1269 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1271 mp = &(*mp)->m_next;
1274 if (inp->inp_flags & INP_RECVIF) {
1277 struct sockaddr_dl sdl;
1280 struct sockaddr_dl *sdp;
1281 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1283 if ((ifp = m->m_pkthdr.rcvif) &&
1284 ifp->if_index && ifp->if_index <= V_if_index) {
1285 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1287 * Change our mind and don't try copy.
1289 if (sdp->sdl_family != AF_LINK ||
1290 sdp->sdl_len > sizeof(sdlbuf)) {
1293 bcopy(sdp, sdl2, sdp->sdl_len);
1297 offsetof(struct sockaddr_dl, sdl_data[0]);
1298 sdl2->sdl_family = AF_LINK;
1299 sdl2->sdl_index = 0;
1300 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1302 *mp = sbcreatecontrol((caddr_t)sdl2, sdl2->sdl_len,
1303 IP_RECVIF, IPPROTO_IP);
1305 mp = &(*mp)->m_next;
1307 if (inp->inp_flags & INP_RECVTOS) {
1308 *mp = sbcreatecontrol((caddr_t)&ip->ip_tos,
1309 sizeof(u_char), IP_RECVTOS, IPPROTO_IP);
1311 mp = &(*mp)->m_next;
1314 if (inp->inp_flags2 & INP_RECVFLOWID) {
1315 uint32_t flowid, flow_type;
1317 flowid = m->m_pkthdr.flowid;
1318 flow_type = M_HASHTYPE_GET(m);
1321 * XXX should handle the failure of one or the
1322 * other - don't populate both?
1324 *mp = sbcreatecontrol((caddr_t) &flowid,
1325 sizeof(uint32_t), IP_FLOWID, IPPROTO_IP);
1327 mp = &(*mp)->m_next;
1328 *mp = sbcreatecontrol((caddr_t) &flow_type,
1329 sizeof(uint32_t), IP_FLOWTYPE, IPPROTO_IP);
1331 mp = &(*mp)->m_next;
1335 if (inp->inp_flags2 & INP_RECVRSSBUCKETID) {
1336 uint32_t flowid, flow_type;
1337 uint32_t rss_bucketid;
1339 flowid = m->m_pkthdr.flowid;
1340 flow_type = M_HASHTYPE_GET(m);
1342 if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) {
1343 *mp = sbcreatecontrol((caddr_t) &rss_bucketid,
1344 sizeof(uint32_t), IP_RSSBUCKETID, IPPROTO_IP);
1346 mp = &(*mp)->m_next;
1353 * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
1354 * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
1355 * locking. This code remains in ip_input.c as ip_mroute.c is optionally
1358 VNET_DEFINE_STATIC(int, ip_rsvp_on);
1359 VNET_DEFINE(struct socket *, ip_rsvpd);
1361 #define V_ip_rsvp_on VNET(ip_rsvp_on)
1364 ip_rsvp_init(struct socket *so)
1367 if (so->so_type != SOCK_RAW ||
1368 so->so_proto->pr_protocol != IPPROTO_RSVP)
1371 if (V_ip_rsvpd != NULL)
1376 * This may seem silly, but we need to be sure we don't over-increment
1377 * the RSVP counter, in case something slips up.
1379 if (!V_ip_rsvp_on) {
1393 * This may seem silly, but we need to be sure we don't over-decrement
1394 * the RSVP counter, in case something slips up.
1404 rsvp_input(struct mbuf **mp, int *offp, int proto)
1411 if (rsvp_input_p) { /* call the real one if loaded */
1413 rsvp_input_p(mp, offp, proto);
1414 return (IPPROTO_DONE);
1417 /* Can still get packets with rsvp_on = 0 if there is a local member
1418 * of the group to which the RSVP packet is addressed. But in this
1419 * case we want to throw the packet away.
1424 return (IPPROTO_DONE);
1427 if (V_ip_rsvpd != NULL) {
1429 rip_input(mp, offp, proto);
1430 return (IPPROTO_DONE);
1432 /* Drop the packet */
1434 return (IPPROTO_DONE);