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 struct rmlock in_ifaddr_lock;
106 RM_SYSINIT(in_ifaddr_lock, &in_ifaddr_lock, "in_ifaddr_lock");
108 VNET_DEFINE(int, rsvp_on);
110 VNET_DEFINE(int, ipforwarding);
111 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_VNET | CTLFLAG_RW,
112 &VNET_NAME(ipforwarding), 0,
113 "Enable IP forwarding between interfaces");
116 * Respond with an ICMP host redirect when we forward a packet out of
117 * the same interface on which it was received. See RFC 792.
119 VNET_DEFINE(int, ipsendredirects) = 1;
120 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_VNET | CTLFLAG_RW,
121 &VNET_NAME(ipsendredirects), 0,
122 "Enable sending IP redirects");
125 * XXX - Setting ip_checkinterface mostly implements the receive side of
126 * the Strong ES model described in RFC 1122, but since the routing table
127 * and transmit implementation do not implement the Strong ES model,
128 * setting this to 1 results in an odd hybrid.
130 * XXX - ip_checkinterface currently must be disabled if you use ipnat
131 * to translate the destination address to another local interface.
133 * XXX - ip_checkinterface must be disabled if you add IP aliases
134 * to the loopback interface instead of the interface where the
135 * packets for those addresses are received.
137 VNET_DEFINE_STATIC(int, ip_checkinterface);
138 #define V_ip_checkinterface VNET(ip_checkinterface)
139 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_VNET | CTLFLAG_RW,
140 &VNET_NAME(ip_checkinterface), 0,
141 "Verify packet arrives on correct interface");
143 VNET_DEFINE(pfil_head_t, inet_pfil_head); /* Packet filter hooks */
145 static struct netisr_handler ip_nh = {
147 .nh_handler = ip_input,
148 .nh_proto = NETISR_IP,
150 .nh_m2cpuid = rss_soft_m2cpuid_v4,
151 .nh_policy = NETISR_POLICY_CPU,
152 .nh_dispatch = NETISR_DISPATCH_HYBRID,
154 .nh_policy = NETISR_POLICY_FLOW,
160 * Directly dispatched frames are currently assumed
161 * to have a flowid already calculated.
163 * It should likely have something that assert it
164 * actually has valid flow details.
166 static struct netisr_handler ip_direct_nh = {
167 .nh_name = "ip_direct",
168 .nh_handler = ip_direct_input,
169 .nh_proto = NETISR_IP_DIRECT,
170 .nh_m2cpuid = rss_soft_m2cpuid_v4,
171 .nh_policy = NETISR_POLICY_CPU,
172 .nh_dispatch = NETISR_DISPATCH_HYBRID,
176 extern struct domain inetdomain;
177 extern struct protosw inetsw[];
178 u_char ip_protox[IPPROTO_MAX];
179 VNET_DEFINE(struct in_ifaddrhead, in_ifaddrhead); /* first inet address */
180 VNET_DEFINE(struct in_ifaddrhashhead *, in_ifaddrhashtbl); /* inet addr hash table */
181 VNET_DEFINE(u_long, in_ifaddrhmask); /* mask for hash table */
184 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
185 &ip_mtu, 0, "Default MTU");
189 VNET_DEFINE(int, ipstealth);
190 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_VNET | CTLFLAG_RW,
191 &VNET_NAME(ipstealth), 0,
192 "IP stealth mode, no TTL decrementation on forwarding");
196 * IP statistics are stored in the "array" of counter(9)s.
198 VNET_PCPUSTAT_DEFINE(struct ipstat, ipstat);
199 VNET_PCPUSTAT_SYSINIT(ipstat);
200 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, IPCTL_STATS, stats, struct ipstat, ipstat,
201 "IP statistics (struct ipstat, netinet/ip_var.h)");
204 VNET_PCPUSTAT_SYSUNINIT(ipstat);
208 * Kernel module interface for updating ipstat. The argument is an index
209 * into ipstat treated as an array.
212 kmod_ipstat_inc(int statnum)
215 counter_u64_add(VNET(ipstat)[statnum], 1);
219 kmod_ipstat_dec(int statnum)
222 counter_u64_add(VNET(ipstat)[statnum], -1);
226 sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS)
230 netisr_getqlimit(&ip_nh, &qlimit);
231 error = sysctl_handle_int(oidp, &qlimit, 0, req);
232 if (error || !req->newptr)
236 return (netisr_setqlimit(&ip_nh, qlimit));
238 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen,
239 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 0,
240 sysctl_netinet_intr_queue_maxlen, "I",
241 "Maximum size of the IP input queue");
244 sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS)
246 u_int64_t qdrops_long;
249 netisr_getqdrops(&ip_nh, &qdrops_long);
250 qdrops = qdrops_long;
251 error = sysctl_handle_int(oidp, &qdrops, 0, req);
252 if (error || !req->newptr)
256 netisr_clearqdrops(&ip_nh);
260 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops,
261 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE,
262 0, 0, sysctl_netinet_intr_queue_drops, "I",
263 "Number of packets dropped from the IP input queue");
267 sysctl_netinet_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS)
271 netisr_getqlimit(&ip_direct_nh, &qlimit);
272 error = sysctl_handle_int(oidp, &qlimit, 0, req);
273 if (error || !req->newptr)
277 return (netisr_setqlimit(&ip_direct_nh, qlimit));
279 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRDQMAXLEN, intr_direct_queue_maxlen,
280 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
281 0, 0, sysctl_netinet_intr_direct_queue_maxlen,
282 "I", "Maximum size of the IP direct input queue");
285 sysctl_netinet_intr_direct_queue_drops(SYSCTL_HANDLER_ARGS)
287 u_int64_t qdrops_long;
290 netisr_getqdrops(&ip_direct_nh, &qdrops_long);
291 qdrops = qdrops_long;
292 error = sysctl_handle_int(oidp, &qdrops, 0, req);
293 if (error || !req->newptr)
297 netisr_clearqdrops(&ip_direct_nh);
301 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRDQDROPS, intr_direct_queue_drops,
302 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, 0, 0,
303 sysctl_netinet_intr_direct_queue_drops, "I",
304 "Number of packets dropped from the IP direct input queue");
308 * IP initialization: fill in IP protocol switch table.
309 * All protocols not implemented in kernel go to raw IP protocol handler.
314 struct pfil_head_args args;
318 CK_STAILQ_INIT(&V_in_ifaddrhead);
319 V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask);
321 /* Initialize IP reassembly queue. */
324 /* Initialize packet filter hooks. */
325 args.pa_version = PFIL_VERSION;
326 args.pa_flags = PFIL_IN | PFIL_OUT;
327 args.pa_type = PFIL_TYPE_IP4;
328 args.pa_headname = PFIL_INET_NAME;
329 V_inet_pfil_head = pfil_head_register(&args);
331 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 rm_priotracker in_ifa_tracker;
457 struct ip *ip = NULL;
458 struct in_ifaddr *ia = NULL;
461 int checkif, hlen = 0;
462 uint16_t sum, ip_len;
463 int dchg = 0; /* dest changed after fw */
464 struct in_addr odst; /* original dst address */
469 if (m->m_flags & M_FASTFWD_OURS) {
470 m->m_flags &= ~M_FASTFWD_OURS;
471 /* Set up some basics that will be used later. */
472 ip = mtod(m, struct ip *);
473 hlen = ip->ip_hl << 2;
474 ip_len = ntohs(ip->ip_len);
478 IPSTAT_INC(ips_total);
480 if (m->m_pkthdr.len < sizeof(struct ip))
483 if (m->m_len < sizeof (struct ip) &&
484 (m = m_pullup(m, sizeof (struct ip))) == NULL) {
485 IPSTAT_INC(ips_toosmall);
488 ip = mtod(m, struct ip *);
490 if (ip->ip_v != IPVERSION) {
491 IPSTAT_INC(ips_badvers);
495 hlen = ip->ip_hl << 2;
496 if (hlen < sizeof(struct ip)) { /* minimum header length */
497 IPSTAT_INC(ips_badhlen);
500 if (hlen > m->m_len) {
501 if ((m = m_pullup(m, hlen)) == NULL) {
502 IPSTAT_INC(ips_badhlen);
505 ip = mtod(m, struct ip *);
508 IP_PROBE(receive, NULL, NULL, ip, m->m_pkthdr.rcvif, ip, NULL);
510 /* IN_LOOPBACK must not appear on the wire - RFC1122 */
511 ifp = m->m_pkthdr.rcvif;
512 if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) ||
513 IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) {
514 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
515 IPSTAT_INC(ips_badaddr);
520 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
521 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
523 if (hlen == sizeof(struct ip)) {
524 sum = in_cksum_hdr(ip);
526 sum = in_cksum(m, hlen);
530 IPSTAT_INC(ips_badsum);
535 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
536 /* packet is dropped by traffic conditioner */
540 ip_len = ntohs(ip->ip_len);
542 IPSTAT_INC(ips_badlen);
547 * Check that the amount of data in the buffers
548 * is as at least much as the IP header would have us expect.
549 * Trim mbufs if longer than we expect.
550 * Drop packet if shorter than we expect.
552 if (m->m_pkthdr.len < ip_len) {
554 IPSTAT_INC(ips_tooshort);
557 if (m->m_pkthdr.len > ip_len) {
558 if (m->m_len == m->m_pkthdr.len) {
560 m->m_pkthdr.len = ip_len;
562 m_adj(m, ip_len - m->m_pkthdr.len);
566 * Try to forward the packet, but if we fail continue.
567 * ip_tryforward() does not generate redirects, so fall
568 * through to normal processing if redirects are required.
569 * ip_tryforward() does inbound and outbound packet firewall
570 * processing. If firewall has decided that destination becomes
571 * our local address, it sets M_FASTFWD_OURS flag. In this
572 * case skip another inbound firewall processing and update
575 if (V_ipforwarding != 0
576 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
577 && (!IPSEC_ENABLED(ipv4) ||
578 IPSEC_CAPS(ipv4, m, IPSEC_CAP_OPERABLE) == 0)
581 if ((m = ip_tryforward(m)) == NULL)
583 if (m->m_flags & M_FASTFWD_OURS) {
584 m->m_flags &= ~M_FASTFWD_OURS;
585 ip = mtod(m, struct ip *);
590 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
592 * Bypass packet filtering for packets previously handled by IPsec.
594 if (IPSEC_ENABLED(ipv4) &&
595 IPSEC_CAPS(ipv4, m, IPSEC_CAP_BYPASS_FILTER) != 0)
600 * Run through list of hooks for input packets.
602 * NB: Beware of the destination address changing (e.g.
603 * by NAT rewriting). When this happens, tell
604 * ip_forward to do the right thing.
607 /* Jump over all PFIL processing if hooks are not active. */
608 if (!PFIL_HOOKED_IN(V_inet_pfil_head))
612 if (pfil_run_hooks(V_inet_pfil_head, &m, ifp, PFIL_IN, NULL) !=
615 if (m == NULL) /* consumed by filter */
618 ip = mtod(m, struct ip *);
619 dchg = (odst.s_addr != ip->ip_dst.s_addr);
620 ifp = m->m_pkthdr.rcvif;
622 if (m->m_flags & M_FASTFWD_OURS) {
623 m->m_flags &= ~M_FASTFWD_OURS;
626 if (m->m_flags & M_IP_NEXTHOP) {
627 if (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) {
629 * Directly ship the packet on. This allows
630 * forwarding packets originally destined to us
631 * to some other directly connected host.
640 * Process options and, if not destined for us,
641 * ship it on. ip_dooptions returns 1 when an
642 * error was detected (causing an icmp message
643 * to be sent and the original packet to be freed).
645 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
648 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
649 * matter if it is destined to another node, or whether it is
650 * a multicast one, RSVP wants it! and prevents it from being forwarded
651 * anywhere else. Also checks if the rsvp daemon is running before
652 * grabbing the packet.
654 if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP)
658 * Check our list of addresses, to see if the packet is for us.
659 * If we don't have any addresses, assume any unicast packet
660 * we receive might be for us (and let the upper layers deal
663 if (CK_STAILQ_EMPTY(&V_in_ifaddrhead) &&
664 (m->m_flags & (M_MCAST|M_BCAST)) == 0)
668 * Enable a consistency check between the destination address
669 * and the arrival interface for a unicast packet (the RFC 1122
670 * strong ES model) if IP forwarding is disabled and the packet
671 * is not locally generated and the packet is not subject to
674 * XXX - Checking also should be disabled if the destination
675 * address is ipnat'ed to a different interface.
677 * XXX - Checking is incompatible with IP aliases added
678 * to the loopback interface instead of the interface where
679 * the packets are received.
681 * XXX - This is the case for carp vhost IPs as well so we
682 * insert a workaround. If the packet got here, we already
683 * checked with carp_iamatch() and carp_forus().
685 checkif = V_ip_checkinterface && (V_ipforwarding == 0) &&
686 ifp != NULL && ((ifp->if_flags & IFF_LOOPBACK) == 0) &&
687 ifp->if_carp == NULL && (dchg == 0);
690 * Check for exact addresses in the hash bucket.
692 IN_IFADDR_RLOCK(&in_ifa_tracker);
693 LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
695 * If the address matches, verify that the packet
696 * arrived via the correct interface if checking is
699 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr &&
700 (!checkif || ia->ia_ifp == ifp)) {
701 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
702 counter_u64_add(ia->ia_ifa.ifa_ibytes,
704 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
708 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
711 * Check for broadcast addresses.
713 * Only accept broadcast packets that arrive via the matching
714 * interface. Reception of forwarded directed broadcasts would
715 * be handled via ip_forward() and ether_output() with the loopback
716 * into the stack for SIMPLEX interfaces handled by ether_output().
718 if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
719 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
720 if (ifa->ifa_addr->sa_family != AF_INET)
723 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
725 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
726 counter_u64_add(ia->ia_ifa.ifa_ibytes,
731 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) {
732 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
733 counter_u64_add(ia->ia_ifa.ifa_ibytes,
741 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
744 * RFC 3927 2.7: Do not forward multicast packets from
747 if (V_ip_mrouter && !IN_LINKLOCAL(ntohl(ip->ip_src.s_addr))) {
749 * If we are acting as a multicast router, all
750 * incoming multicast packets are passed to the
751 * kernel-level multicast forwarding function.
752 * The packet is returned (relatively) intact; if
753 * ip_mforward() returns a non-zero value, the packet
754 * must be discarded, else it may be accepted below.
756 if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) {
758 IPSTAT_INC(ips_cantforward);
764 * The process-level routing daemon needs to receive
765 * all multicast IGMP packets, whether or not this
766 * host belongs to their destination groups.
768 if (ip->ip_p == IPPROTO_IGMP) {
772 IPSTAT_INC(ips_forward);
776 * Assume the packet is for us, to avoid prematurely taking
777 * a lock on the in_multi hash. Protocols must perform
778 * their own filtering and update statistics accordingly.
782 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
784 if (ip->ip_dst.s_addr == INADDR_ANY)
786 /* RFC 3927 2.7: Do not forward packets to or from IN_LINKLOCAL. */
787 if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr)) ||
788 IN_LINKLOCAL(ntohl(ip->ip_src.s_addr))) {
789 IPSTAT_INC(ips_cantforward);
795 * Not for us; forward if possible and desirable.
797 if (V_ipforwarding == 0) {
798 IPSTAT_INC(ips_cantforward);
808 * IPSTEALTH: Process non-routing options only
809 * if the packet is destined for us.
811 if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1))
813 #endif /* IPSTEALTH */
816 * Attempt reassembly; if it succeeds, proceed.
817 * ip_reass() will return a different mbuf.
819 if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) {
820 /* XXXGL: shouldn't we save & set m_flags? */
824 ip = mtod(m, struct ip *);
825 /* Get the header length of the reassembled packet */
826 hlen = ip->ip_hl << 2;
829 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
830 if (IPSEC_ENABLED(ipv4)) {
831 if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
837 * Switch out to protocol's input routine.
839 IPSTAT_INC(ips_delivered);
841 (*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
848 * IP timer processing;
849 * if a timer expires on a reassembly
855 VNET_ITERATOR_DECL(vnet_iter);
857 VNET_LIST_RLOCK_NOSLEEP();
858 VNET_FOREACH(vnet_iter) {
859 CURVNET_SET(vnet_iter);
863 VNET_LIST_RUNLOCK_NOSLEEP();
869 VNET_ITERATOR_DECL(vnet_iter);
871 VNET_LIST_RLOCK_NOSLEEP();
872 VNET_FOREACH(vnet_iter) {
873 CURVNET_SET(vnet_iter);
877 VNET_LIST_RUNLOCK_NOSLEEP();
881 * The protocol to be inserted into ip_protox[] must be already registered
882 * in inetsw[], either statically or through pf_proto_register().
885 ipproto_register(short ipproto)
890 if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
891 return (EPROTONOSUPPORT);
894 * The protocol slot must not be occupied by another protocol
895 * already. An index pointing to IPPROTO_RAW is unused.
897 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
899 return (EPFNOSUPPORT);
900 if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */
903 /* Find the protocol position in inetsw[] and set the index. */
904 for (pr = inetdomain.dom_protosw;
905 pr < inetdomain.dom_protoswNPROTOSW; pr++) {
906 if (pr->pr_domain->dom_family == PF_INET &&
907 pr->pr_protocol && pr->pr_protocol == ipproto) {
908 ip_protox[pr->pr_protocol] = pr - inetsw;
912 return (EPROTONOSUPPORT);
916 ipproto_unregister(short ipproto)
921 if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
922 return (EPROTONOSUPPORT);
924 /* Check if the protocol was indeed registered. */
925 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
927 return (EPFNOSUPPORT);
928 if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */
931 /* Reset the protocol slot to IPPROTO_RAW. */
932 ip_protox[ipproto] = pr - inetsw;
936 u_char inetctlerrmap[PRC_NCMDS] = {
938 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
939 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
940 EMSGSIZE, EHOSTUNREACH, 0, 0,
941 0, 0, EHOSTUNREACH, 0,
942 ENOPROTOOPT, ECONNREFUSED
946 * Forward a packet. If some error occurs return the sender
947 * an icmp packet. Note we can't always generate a meaningful
948 * icmp message because icmp doesn't have a large enough repertoire
949 * of codes and types.
951 * If not forwarding, just drop the packet. This could be confusing
952 * if ipforwarding was zero but some routing protocol was advancing
953 * us as a gateway to somewhere. However, we must let the routing
954 * protocol deal with that.
956 * The srcrt parameter indicates whether the packet is being forwarded
957 * via a source route.
960 ip_forward(struct mbuf *m, int srcrt)
962 struct ip *ip = mtod(m, struct ip *);
963 struct in_ifaddr *ia;
965 struct sockaddr_in *sin;
969 int error, type = 0, code = 0, mtu = 0;
973 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
974 IPSTAT_INC(ips_cantforward);
982 ip->ip_ttl <= IPTTLDEC) {
983 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 0, 0);
987 bzero(&ro, sizeof(ro));
988 sin = (struct sockaddr_in *)&ro.ro_dst;
989 sin->sin_family = AF_INET;
990 sin->sin_len = sizeof(*sin);
991 sin->sin_addr = ip->ip_dst;
992 flowid = m->m_pkthdr.flowid;
993 ro.ro_nh = fib4_lookup(M_GETFIB(m), ip->ip_dst, 0, NHR_REF, flowid);
994 if (ro.ro_nh != NULL) {
995 ia = ifatoia(ro.ro_nh->nh_ifa);
999 * Save the IP header and at most 8 bytes of the payload,
1000 * in case we need to generate an ICMP message to the src.
1002 * XXX this can be optimized a lot by saving the data in a local
1003 * buffer on the stack (72 bytes at most), and only allocating the
1004 * mbuf if really necessary. The vast majority of the packets
1005 * are forwarded without having to send an ICMP back (either
1006 * because unnecessary, or because rate limited), so we are
1007 * really we are wasting a lot of work here.
1009 * We don't use m_copym() because it might return a reference
1010 * to a shared cluster. Both this function and ip_output()
1011 * assume exclusive access to the IP header in `m', so any
1012 * data in a cluster may change before we reach icmp_error().
1014 mcopy = m_gethdr(M_NOWAIT, m->m_type);
1015 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_NOWAIT)) {
1017 * It's probably ok if the pkthdr dup fails (because
1018 * the deep copy of the tag chain failed), but for now
1019 * be conservative and just discard the copy since
1020 * code below may some day want the tags.
1025 if (mcopy != NULL) {
1026 mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy));
1027 mcopy->m_pkthdr.len = mcopy->m_len;
1028 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1031 if (V_ipstealth == 0)
1033 ip->ip_ttl -= IPTTLDEC;
1034 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1035 if (IPSEC_ENABLED(ipv4)) {
1036 if ((error = IPSEC_FORWARD(ipv4, m)) != 0) {
1037 /* mbuf consumed by IPsec */
1040 if (error != EINPROGRESS)
1041 IPSTAT_INC(ips_cantforward);
1044 /* No IPsec processing required */
1048 * If forwarding packet using same interface that it came in on,
1049 * perhaps should send a redirect to sender to shortcut a hop.
1050 * Only send redirect if source is sending directly to us,
1051 * and if packet was not source routed (or has any options).
1052 * Also, don't send redirect if forwarding using a default route
1053 * or a route modified by a redirect.
1056 if (!srcrt && V_ipsendredirects &&
1057 ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) {
1058 struct nhop_object *nh;
1062 if (nh != NULL && ((nh->nh_flags & (NHF_REDIRECT|NHF_DEFAULT)) == 0)) {
1063 struct in_ifaddr *nh_ia = (struct in_ifaddr *)(nh->nh_ifa);
1064 u_long src = ntohl(ip->ip_src.s_addr);
1066 if (nh_ia != NULL &&
1067 (src & nh_ia->ia_subnetmask) == nh_ia->ia_subnet) {
1068 /* Router requirements says to only send host redirects */
1069 type = ICMP_REDIRECT;
1070 code = ICMP_REDIRECT_HOST;
1071 if (nh->nh_flags & NHF_GATEWAY) {
1072 if (nh->gw_sa.sa_family == AF_INET)
1073 dest.s_addr = nh->gw4_sa.sin_addr.s_addr;
1074 else /* Do not redirect in case gw is AF_INET6 */
1077 dest.s_addr = ip->ip_dst.s_addr;
1082 error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
1084 if (error == EMSGSIZE && ro.ro_nh)
1085 mtu = ro.ro_nh->nh_mtu;
1089 IPSTAT_INC(ips_cantforward);
1091 IPSTAT_INC(ips_forward);
1093 IPSTAT_INC(ips_redirectsent);
1104 case 0: /* forwarded, but need redirect */
1105 /* type, code set above */
1113 type = ICMP_UNREACH;
1114 code = ICMP_UNREACH_HOST;
1118 type = ICMP_UNREACH;
1119 code = ICMP_UNREACH_NEEDFRAG;
1121 * If the MTU was set before make sure we are below the
1123 * If the MTU wasn't set before use the interface mtu or
1124 * fall back to the next smaller mtu step compared to the
1125 * current packet size.
1129 mtu = min(mtu, ia->ia_ifp->if_mtu);
1132 mtu = ia->ia_ifp->if_mtu;
1134 mtu = ip_next_mtu(ntohs(ip->ip_len), 0);
1136 IPSTAT_INC(ips_cantfrag);
1140 case EACCES: /* ipfw denied packet */
1144 icmp_error(mcopy, type, code, dest.s_addr, mtu);
1147 #define CHECK_SO_CT(sp, ct) \
1148 (((sp->so_options & SO_TIMESTAMP) && (sp->so_ts_clock == ct)) ? 1 : 0)
1151 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1157 if ((inp->inp_socket->so_options & SO_BINTIME) ||
1158 CHECK_SO_CT(inp->inp_socket, SO_TS_BINTIME)) {
1159 struct bintime boottimebin, bt;
1160 struct timespec ts1;
1162 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1164 mbuf_tstmp2timespec(m, &ts1);
1165 timespec2bintime(&ts1, &bt);
1166 getboottimebin(&boottimebin);
1167 bintime_add(&bt, &boottimebin);
1171 *mp = sbcreatecontrol((caddr_t)&bt, sizeof(bt),
1172 SCM_BINTIME, SOL_SOCKET);
1174 mp = &(*mp)->m_next;
1178 if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME_MICRO)) {
1179 struct bintime boottimebin, bt1;
1180 struct timespec ts1;
1183 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1185 mbuf_tstmp2timespec(m, &ts1);
1186 timespec2bintime(&ts1, &bt1);
1187 getboottimebin(&boottimebin);
1188 bintime_add(&bt1, &boottimebin);
1189 bintime2timeval(&bt1, &tv);
1193 *mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv),
1194 SCM_TIMESTAMP, SOL_SOCKET);
1196 mp = &(*mp)->m_next;
1199 } else if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME)) {
1200 struct bintime boottimebin;
1201 struct timespec ts, ts1;
1203 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1205 mbuf_tstmp2timespec(m, &ts);
1206 getboottimebin(&boottimebin);
1207 bintime2timespec(&boottimebin, &ts1);
1208 timespecadd(&ts, &ts1, &ts);
1212 *mp = sbcreatecontrol((caddr_t)&ts, sizeof(ts),
1213 SCM_REALTIME, SOL_SOCKET);
1215 mp = &(*mp)->m_next;
1218 } else if (CHECK_SO_CT(inp->inp_socket, SO_TS_MONOTONIC)) {
1221 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1223 mbuf_tstmp2timespec(m, &ts);
1226 *mp = sbcreatecontrol((caddr_t)&ts, sizeof(ts),
1227 SCM_MONOTONIC, SOL_SOCKET);
1229 mp = &(*mp)->m_next;
1233 if (stamped && (m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1235 struct sock_timestamp_info sti;
1237 bzero(&sti, sizeof(sti));
1238 sti.st_info_flags = ST_INFO_HW;
1239 if ((m->m_flags & M_TSTMP_HPREC) != 0)
1240 sti.st_info_flags |= ST_INFO_HW_HPREC;
1241 *mp = sbcreatecontrol((caddr_t)&sti, sizeof(sti), SCM_TIME_INFO,
1244 mp = &(*mp)->m_next;
1246 if (inp->inp_flags & INP_RECVDSTADDR) {
1247 *mp = sbcreatecontrol((caddr_t)&ip->ip_dst,
1248 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1250 mp = &(*mp)->m_next;
1252 if (inp->inp_flags & INP_RECVTTL) {
1253 *mp = sbcreatecontrol((caddr_t)&ip->ip_ttl,
1254 sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
1256 mp = &(*mp)->m_next;
1260 * Moving these out of udp_input() made them even more broken
1261 * than they already were.
1263 /* options were tossed already */
1264 if (inp->inp_flags & INP_RECVOPTS) {
1265 *mp = sbcreatecontrol((caddr_t)opts_deleted_above,
1266 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1268 mp = &(*mp)->m_next;
1270 /* ip_srcroute doesn't do what we want here, need to fix */
1271 if (inp->inp_flags & INP_RECVRETOPTS) {
1272 *mp = sbcreatecontrol((caddr_t)ip_srcroute(m),
1273 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1275 mp = &(*mp)->m_next;
1278 if (inp->inp_flags & INP_RECVIF) {
1281 struct sockaddr_dl sdl;
1284 struct sockaddr_dl *sdp;
1285 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1287 if ((ifp = m->m_pkthdr.rcvif) &&
1288 ifp->if_index && ifp->if_index <= V_if_index) {
1289 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1291 * Change our mind and don't try copy.
1293 if (sdp->sdl_family != AF_LINK ||
1294 sdp->sdl_len > sizeof(sdlbuf)) {
1297 bcopy(sdp, sdl2, sdp->sdl_len);
1301 offsetof(struct sockaddr_dl, sdl_data[0]);
1302 sdl2->sdl_family = AF_LINK;
1303 sdl2->sdl_index = 0;
1304 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1306 *mp = sbcreatecontrol((caddr_t)sdl2, sdl2->sdl_len,
1307 IP_RECVIF, IPPROTO_IP);
1309 mp = &(*mp)->m_next;
1311 if (inp->inp_flags & INP_RECVTOS) {
1312 *mp = sbcreatecontrol((caddr_t)&ip->ip_tos,
1313 sizeof(u_char), IP_RECVTOS, IPPROTO_IP);
1315 mp = &(*mp)->m_next;
1318 if (inp->inp_flags2 & INP_RECVFLOWID) {
1319 uint32_t flowid, flow_type;
1321 flowid = m->m_pkthdr.flowid;
1322 flow_type = M_HASHTYPE_GET(m);
1325 * XXX should handle the failure of one or the
1326 * other - don't populate both?
1328 *mp = sbcreatecontrol((caddr_t) &flowid,
1329 sizeof(uint32_t), IP_FLOWID, IPPROTO_IP);
1331 mp = &(*mp)->m_next;
1332 *mp = sbcreatecontrol((caddr_t) &flow_type,
1333 sizeof(uint32_t), IP_FLOWTYPE, IPPROTO_IP);
1335 mp = &(*mp)->m_next;
1339 if (inp->inp_flags2 & INP_RECVRSSBUCKETID) {
1340 uint32_t flowid, flow_type;
1341 uint32_t rss_bucketid;
1343 flowid = m->m_pkthdr.flowid;
1344 flow_type = M_HASHTYPE_GET(m);
1346 if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) {
1347 *mp = sbcreatecontrol((caddr_t) &rss_bucketid,
1348 sizeof(uint32_t), IP_RSSBUCKETID, IPPROTO_IP);
1350 mp = &(*mp)->m_next;
1357 * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
1358 * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
1359 * locking. This code remains in ip_input.c as ip_mroute.c is optionally
1362 VNET_DEFINE_STATIC(int, ip_rsvp_on);
1363 VNET_DEFINE(struct socket *, ip_rsvpd);
1365 #define V_ip_rsvp_on VNET(ip_rsvp_on)
1368 ip_rsvp_init(struct socket *so)
1371 if (so->so_type != SOCK_RAW ||
1372 so->so_proto->pr_protocol != IPPROTO_RSVP)
1375 if (V_ip_rsvpd != NULL)
1380 * This may seem silly, but we need to be sure we don't over-increment
1381 * the RSVP counter, in case something slips up.
1383 if (!V_ip_rsvp_on) {
1397 * This may seem silly, but we need to be sure we don't over-decrement
1398 * the RSVP counter, in case something slips up.
1408 rsvp_input(struct mbuf **mp, int *offp, int proto)
1415 if (rsvp_input_p) { /* call the real one if loaded */
1417 rsvp_input_p(mp, offp, proto);
1418 return (IPPROTO_DONE);
1421 /* Can still get packets with rsvp_on = 0 if there is a local member
1422 * of the group to which the RSVP packet is addressed. But in this
1423 * case we want to throw the packet away.
1428 return (IPPROTO_DONE);
1431 if (V_ip_rsvpd != NULL) {
1433 rip_input(mp, offp, proto);
1434 return (IPPROTO_DONE);
1436 /* Drop the packet */
1438 return (IPPROTO_DONE);