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");
121 VNET_DEFINE_STATIC(bool, ip_strong_es) = false;
122 #define V_ip_strong_es VNET(ip_strong_es)
123 SYSCTL_BOOL(_net_inet_ip, OID_AUTO, rfc1122_strong_es,
124 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip_strong_es), false,
125 "Packet's IP destination address must match address on arrival interface");
127 VNET_DEFINE_STATIC(bool, ip_sav) = true;
128 #define V_ip_sav VNET(ip_sav)
129 SYSCTL_BOOL(_net_inet_ip, OID_AUTO, source_address_validation,
130 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip_sav), true,
131 "Drop incoming packets with source address that is a local address");
133 VNET_DEFINE(pfil_head_t, inet_pfil_head); /* Packet filter hooks */
135 static struct netisr_handler ip_nh = {
137 .nh_handler = ip_input,
138 .nh_proto = NETISR_IP,
140 .nh_m2cpuid = rss_soft_m2cpuid_v4,
141 .nh_policy = NETISR_POLICY_CPU,
142 .nh_dispatch = NETISR_DISPATCH_HYBRID,
144 .nh_policy = NETISR_POLICY_FLOW,
150 * Directly dispatched frames are currently assumed
151 * to have a flowid already calculated.
153 * It should likely have something that assert it
154 * actually has valid flow details.
156 static struct netisr_handler ip_direct_nh = {
157 .nh_name = "ip_direct",
158 .nh_handler = ip_direct_input,
159 .nh_proto = NETISR_IP_DIRECT,
160 .nh_m2cpuid = rss_soft_m2cpuid_v4,
161 .nh_policy = NETISR_POLICY_CPU,
162 .nh_dispatch = NETISR_DISPATCH_HYBRID,
166 extern struct domain inetdomain;
167 extern struct protosw inetsw[];
168 u_char ip_protox[IPPROTO_MAX];
169 VNET_DEFINE(struct in_ifaddrhead, in_ifaddrhead); /* first inet address */
170 VNET_DEFINE(struct in_ifaddrhashhead *, in_ifaddrhashtbl); /* inet addr hash table */
171 VNET_DEFINE(u_long, in_ifaddrhmask); /* mask for hash table */
173 /* Make sure it is safe to use hashinit(9) on CK_LIST. */
174 CTASSERT(sizeof(struct in_ifaddrhashhead) == sizeof(LIST_HEAD(, in_addr)));
177 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
178 &ip_mtu, 0, "Default MTU");
182 VNET_DEFINE(int, ipstealth);
183 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_VNET | CTLFLAG_RW,
184 &VNET_NAME(ipstealth), 0,
185 "IP stealth mode, no TTL decrementation on forwarding");
189 * IP statistics are stored in the "array" of counter(9)s.
191 VNET_PCPUSTAT_DEFINE(struct ipstat, ipstat);
192 VNET_PCPUSTAT_SYSINIT(ipstat);
193 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, IPCTL_STATS, stats, struct ipstat, ipstat,
194 "IP statistics (struct ipstat, netinet/ip_var.h)");
197 VNET_PCPUSTAT_SYSUNINIT(ipstat);
201 * Kernel module interface for updating ipstat. The argument is an index
202 * into ipstat treated as an array.
205 kmod_ipstat_inc(int statnum)
208 counter_u64_add(VNET(ipstat)[statnum], 1);
212 kmod_ipstat_dec(int statnum)
215 counter_u64_add(VNET(ipstat)[statnum], -1);
219 sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS)
223 netisr_getqlimit(&ip_nh, &qlimit);
224 error = sysctl_handle_int(oidp, &qlimit, 0, req);
225 if (error || !req->newptr)
229 return (netisr_setqlimit(&ip_nh, qlimit));
231 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen,
232 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 0,
233 sysctl_netinet_intr_queue_maxlen, "I",
234 "Maximum size of the IP input queue");
237 sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS)
239 u_int64_t qdrops_long;
242 netisr_getqdrops(&ip_nh, &qdrops_long);
243 qdrops = qdrops_long;
244 error = sysctl_handle_int(oidp, &qdrops, 0, req);
245 if (error || !req->newptr)
249 netisr_clearqdrops(&ip_nh);
253 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops,
254 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE,
255 0, 0, sysctl_netinet_intr_queue_drops, "I",
256 "Number of packets dropped from the IP input queue");
260 sysctl_netinet_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS)
264 netisr_getqlimit(&ip_direct_nh, &qlimit);
265 error = sysctl_handle_int(oidp, &qlimit, 0, req);
266 if (error || !req->newptr)
270 return (netisr_setqlimit(&ip_direct_nh, qlimit));
272 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRDQMAXLEN, intr_direct_queue_maxlen,
273 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
274 0, 0, sysctl_netinet_intr_direct_queue_maxlen,
275 "I", "Maximum size of the IP direct input queue");
278 sysctl_netinet_intr_direct_queue_drops(SYSCTL_HANDLER_ARGS)
280 u_int64_t qdrops_long;
283 netisr_getqdrops(&ip_direct_nh, &qdrops_long);
284 qdrops = qdrops_long;
285 error = sysctl_handle_int(oidp, &qdrops, 0, req);
286 if (error || !req->newptr)
290 netisr_clearqdrops(&ip_direct_nh);
294 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRDQDROPS, intr_direct_queue_drops,
295 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, 0, 0,
296 sysctl_netinet_intr_direct_queue_drops, "I",
297 "Number of packets dropped from the IP direct input queue");
301 * IP initialization: fill in IP protocol switch table.
302 * All protocols not implemented in kernel go to raw IP protocol handler.
307 struct pfil_head_args args;
311 CK_STAILQ_INIT(&V_in_ifaddrhead);
312 V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask);
314 /* Initialize IP reassembly queue. */
317 /* Initialize packet filter hooks. */
318 args.pa_version = PFIL_VERSION;
319 args.pa_flags = PFIL_IN | PFIL_OUT;
320 args.pa_type = PFIL_TYPE_IP4;
321 args.pa_headname = PFIL_INET_NAME;
322 V_inet_pfil_head = pfil_head_register(&args);
324 if (hhook_head_register(HHOOK_TYPE_IPSEC_IN, AF_INET,
325 &V_ipsec_hhh_in[HHOOK_IPSEC_INET],
326 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
327 printf("%s: WARNING: unable to register input helper hook\n",
329 if (hhook_head_register(HHOOK_TYPE_IPSEC_OUT, AF_INET,
330 &V_ipsec_hhh_out[HHOOK_IPSEC_INET],
331 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
332 printf("%s: WARNING: unable to register output helper hook\n",
335 /* Skip initialization of globals for non-default instances. */
337 if (!IS_DEFAULT_VNET(curvnet)) {
338 netisr_register_vnet(&ip_nh);
340 netisr_register_vnet(&ip_direct_nh);
346 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
348 panic("ip_init: PF_INET not found");
350 /* Initialize the entire ip_protox[] array to IPPROTO_RAW. */
351 for (i = 0; i < IPPROTO_MAX; i++)
352 ip_protox[i] = pr - inetsw;
354 * Cycle through IP protocols and put them into the appropriate place
357 for (pr = inetdomain.dom_protosw;
358 pr < inetdomain.dom_protoswNPROTOSW; pr++)
359 if (pr->pr_domain->dom_family == PF_INET &&
360 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
361 /* Be careful to only index valid IP protocols. */
362 if (pr->pr_protocol < IPPROTO_MAX)
363 ip_protox[pr->pr_protocol] = pr - inetsw;
366 netisr_register(&ip_nh);
368 netisr_register(&ip_direct_nh);
374 ip_destroy(void *unused __unused)
379 netisr_unregister_vnet(&ip_direct_nh);
381 netisr_unregister_vnet(&ip_nh);
383 pfil_head_unregister(V_inet_pfil_head);
384 error = hhook_head_deregister(V_ipsec_hhh_in[HHOOK_IPSEC_INET]);
386 printf("%s: WARNING: unable to deregister input helper hook "
387 "type HHOOK_TYPE_IPSEC_IN, id HHOOK_IPSEC_INET: "
388 "error %d returned\n", __func__, error);
390 error = hhook_head_deregister(V_ipsec_hhh_out[HHOOK_IPSEC_INET]);
392 printf("%s: WARNING: unable to deregister output helper hook "
393 "type HHOOK_TYPE_IPSEC_OUT, id HHOOK_IPSEC_INET: "
394 "error %d returned\n", __func__, error);
397 /* Remove the IPv4 addresses from all interfaces. */
400 /* Make sure the IPv4 routes are gone as well. */
401 rib_flush_routes_family(AF_INET);
403 /* Destroy IP reassembly queue. */
406 /* Cleanup in_ifaddr hash table; should be empty. */
407 hashdestroy(V_in_ifaddrhashtbl, M_IFADDR, V_in_ifaddrhmask);
410 VNET_SYSUNINIT(ip, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip_destroy, NULL);
415 * IP direct input routine.
417 * This is called when reinjecting completed fragments where
418 * all of the previous checking and book-keeping has been done.
421 ip_direct_input(struct mbuf *m)
426 ip = mtod(m, struct ip *);
427 hlen = ip->ip_hl << 2;
429 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
430 if (IPSEC_ENABLED(ipv4)) {
431 if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
435 IPSTAT_INC(ips_delivered);
436 (*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
442 * Ip input routine. Checksum and byte swap header. If fragmented
443 * try to reassemble. Process options. Pass to next level.
446 ip_input(struct mbuf *m)
448 MROUTER_RLOCK_TRACKER;
449 struct ip *ip = NULL;
450 struct in_ifaddr *ia = NULL;
454 uint16_t sum, ip_len;
455 int dchg = 0; /* dest changed after fw */
456 struct in_addr odst; /* original dst address */
462 if (m->m_flags & M_FASTFWD_OURS) {
463 m->m_flags &= ~M_FASTFWD_OURS;
464 /* Set up some basics that will be used later. */
465 ip = mtod(m, struct ip *);
466 hlen = ip->ip_hl << 2;
467 ip_len = ntohs(ip->ip_len);
471 IPSTAT_INC(ips_total);
473 if (__predict_false(m->m_pkthdr.len < sizeof(struct ip)))
476 if (m->m_len < sizeof(struct ip)) {
477 m = m_pullup(m, sizeof(struct ip));
478 if (__predict_false(m == NULL)) {
479 IPSTAT_INC(ips_toosmall);
483 ip = mtod(m, struct ip *);
485 if (__predict_false(ip->ip_v != IPVERSION)) {
486 IPSTAT_INC(ips_badvers);
490 hlen = ip->ip_hl << 2;
491 if (__predict_false(hlen < sizeof(struct ip))) { /* minimum header length */
492 IPSTAT_INC(ips_badhlen);
495 if (hlen > m->m_len) {
496 m = m_pullup(m, hlen);
497 if (__predict_false(m == NULL)) {
498 IPSTAT_INC(ips_badhlen);
501 ip = mtod(m, struct ip *);
504 IP_PROBE(receive, NULL, NULL, ip, m->m_pkthdr.rcvif, ip, NULL);
506 /* IN_LOOPBACK must not appear on the wire - RFC1122 */
507 ifp = m->m_pkthdr.rcvif;
508 if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) ||
509 IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) {
510 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
511 IPSTAT_INC(ips_badaddr);
516 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
517 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
519 if (hlen == sizeof(struct ip)) {
520 sum = in_cksum_hdr(ip);
522 sum = in_cksum(m, hlen);
525 if (__predict_false(sum)) {
526 IPSTAT_INC(ips_badsum);
531 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
532 /* packet is dropped by traffic conditioner */
536 ip_len = ntohs(ip->ip_len);
537 if (__predict_false(ip_len < hlen)) {
538 IPSTAT_INC(ips_badlen);
543 * Check that the amount of data in the buffers
544 * is as at least much as the IP header would have us expect.
545 * Trim mbufs if longer than we expect.
546 * Drop packet if shorter than we expect.
548 if (__predict_false(m->m_pkthdr.len < ip_len)) {
550 IPSTAT_INC(ips_tooshort);
553 if (m->m_pkthdr.len > ip_len) {
554 if (m->m_len == m->m_pkthdr.len) {
556 m->m_pkthdr.len = ip_len;
558 m_adj(m, ip_len - m->m_pkthdr.len);
562 * Try to forward the packet, but if we fail continue.
563 * ip_tryforward() does not generate redirects, so fall
564 * through to normal processing if redirects are required.
565 * ip_tryforward() does inbound and outbound packet firewall
566 * processing. If firewall has decided that destination becomes
567 * our local address, it sets M_FASTFWD_OURS flag. In this
568 * case skip another inbound firewall processing and update
571 if (V_ipforwarding != 0
572 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
573 && (!IPSEC_ENABLED(ipv4) ||
574 IPSEC_CAPS(ipv4, m, IPSEC_CAP_OPERABLE) == 0)
577 if ((m = ip_tryforward(m)) == NULL)
579 if (m->m_flags & M_FASTFWD_OURS) {
580 m->m_flags &= ~M_FASTFWD_OURS;
581 ip = mtod(m, struct ip *);
586 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
588 * Bypass packet filtering for packets previously handled by IPsec.
590 if (IPSEC_ENABLED(ipv4) &&
591 IPSEC_CAPS(ipv4, m, IPSEC_CAP_BYPASS_FILTER) != 0)
596 * Run through list of hooks for input packets.
598 * NB: Beware of the destination address changing (e.g.
599 * by NAT rewriting). When this happens, tell
600 * ip_forward to do the right thing.
603 /* Jump over all PFIL processing if hooks are not active. */
604 if (!PFIL_HOOKED_IN(V_inet_pfil_head))
608 if (pfil_run_hooks(V_inet_pfil_head, &m, ifp, PFIL_IN, NULL) !=
611 if (m == NULL) /* consumed by filter */
614 ip = mtod(m, struct ip *);
615 dchg = (odst.s_addr != ip->ip_dst.s_addr);
617 if (m->m_flags & M_FASTFWD_OURS) {
618 m->m_flags &= ~M_FASTFWD_OURS;
621 if (m->m_flags & M_IP_NEXTHOP) {
622 if (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) {
624 * Directly ship the packet on. This allows
625 * forwarding packets originally destined to us
626 * to some other directly connected host.
635 * Process options and, if not destined for us,
636 * ship it on. ip_dooptions returns 1 when an
637 * error was detected (causing an icmp message
638 * to be sent and the original packet to be freed).
640 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
643 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
644 * matter if it is destined to another node, or whether it is
645 * a multicast one, RSVP wants it! and prevents it from being forwarded
646 * anywhere else. Also checks if the rsvp daemon is running before
647 * grabbing the packet.
649 if (ip->ip_p == IPPROTO_RSVP && V_rsvp_on)
653 * Check our list of addresses, to see if the packet is for us.
654 * If we don't have any addresses, assume any unicast packet
655 * we receive might be for us (and let the upper layers deal
658 if (CK_STAILQ_EMPTY(&V_in_ifaddrhead) &&
659 (m->m_flags & (M_MCAST|M_BCAST)) == 0)
663 * Enable a consistency check between the destination address
664 * and the arrival interface for a unicast packet (the RFC 1122
665 * strong ES model) with a list of additional predicates:
666 * - if IP forwarding is disabled
667 * - the packet is not locally generated
668 * - the packet is not subject to 'ipfw fwd'
669 * - Interface is not running CARP. If the packet got here, we already
670 * checked it with carp_iamatch() and carp_forus().
672 strong_es = V_ip_strong_es && (V_ipforwarding == 0) &&
673 ((ifp->if_flags & IFF_LOOPBACK) == 0) &&
674 ifp->if_carp == NULL && (dchg == 0);
677 * Check for exact addresses in the hash bucket.
679 CK_LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
680 if (IA_SIN(ia)->sin_addr.s_addr != ip->ip_dst.s_addr)
684 * net.inet.ip.rfc1122_strong_es: the address matches, verify
685 * that the packet arrived via the correct interface.
687 if (__predict_false(strong_es && ia->ia_ifp != ifp)) {
688 IPSTAT_INC(ips_badaddr);
693 * net.inet.ip.source_address_validation: drop incoming
694 * packets that pretend to be ours.
696 if (V_ip_sav && !(ifp->if_flags & IFF_LOOPBACK) &&
697 __predict_false(in_localip_fib(ip->ip_src, ifp->if_fib))) {
698 IPSTAT_INC(ips_badaddr);
702 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
703 counter_u64_add(ia->ia_ifa.ifa_ibytes, m->m_pkthdr.len);
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->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 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
741 * RFC 3927 2.7: Do not forward multicast packets from
744 if (V_ip_mrouter && !IN_LINKLOCAL(ntohl(ip->ip_src.s_addr))) {
746 * If we are acting as a multicast router, all
747 * incoming multicast packets are passed to the
748 * kernel-level multicast forwarding function.
749 * The packet is returned (relatively) intact; if
750 * ip_mforward() returns a non-zero value, the packet
751 * must be discarded, else it may be accepted below.
753 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) {
769 IPSTAT_INC(ips_forward);
773 * Assume the packet is for us, to avoid prematurely taking
774 * a lock on the in_multi hash. Protocols must perform
775 * their own filtering and update statistics accordingly.
779 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
781 if (ip->ip_dst.s_addr == INADDR_ANY)
783 /* RFC 3927 2.7: Do not forward packets to or from IN_LINKLOCAL. */
784 if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr)) ||
785 IN_LINKLOCAL(ntohl(ip->ip_src.s_addr))) {
786 IPSTAT_INC(ips_cantforward);
792 * Not for us; forward if possible and desirable.
794 if (V_ipforwarding == 0) {
795 IPSTAT_INC(ips_cantforward);
805 * IPSTEALTH: Process non-routing options only
806 * if the packet is destined for us.
808 if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1))
810 #endif /* IPSTEALTH */
813 * Attempt reassembly; if it succeeds, proceed.
814 * ip_reass() will return a different mbuf.
816 if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) {
817 /* XXXGL: shouldn't we save & set m_flags? */
821 ip = mtod(m, struct ip *);
822 /* Get the header length of the reassembled packet */
823 hlen = ip->ip_hl << 2;
826 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
827 if (IPSEC_ENABLED(ipv4)) {
828 if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
834 * Switch out to protocol's input routine.
836 IPSTAT_INC(ips_delivered);
838 (*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
845 * IP timer processing;
846 * if a timer expires on a reassembly
852 VNET_ITERATOR_DECL(vnet_iter);
854 VNET_LIST_RLOCK_NOSLEEP();
855 VNET_FOREACH(vnet_iter) {
856 CURVNET_SET(vnet_iter);
860 VNET_LIST_RUNLOCK_NOSLEEP();
866 VNET_ITERATOR_DECL(vnet_iter);
868 VNET_LIST_RLOCK_NOSLEEP();
869 VNET_FOREACH(vnet_iter) {
870 CURVNET_SET(vnet_iter);
874 VNET_LIST_RUNLOCK_NOSLEEP();
878 * The protocol to be inserted into ip_protox[] must be already registered
879 * in inetsw[], either statically or through pf_proto_register().
882 ipproto_register(short ipproto)
887 if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
888 return (EPROTONOSUPPORT);
891 * The protocol slot must not be occupied by another protocol
892 * already. An index pointing to IPPROTO_RAW is unused.
894 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
896 return (EPFNOSUPPORT);
897 if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */
900 /* Find the protocol position in inetsw[] and set the index. */
901 for (pr = inetdomain.dom_protosw;
902 pr < inetdomain.dom_protoswNPROTOSW; pr++) {
903 if (pr->pr_domain->dom_family == PF_INET &&
904 pr->pr_protocol && pr->pr_protocol == ipproto) {
905 ip_protox[pr->pr_protocol] = pr - inetsw;
909 return (EPROTONOSUPPORT);
913 ipproto_unregister(short ipproto)
918 if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
919 return (EPROTONOSUPPORT);
921 /* Check if the protocol was indeed registered. */
922 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
924 return (EPFNOSUPPORT);
925 if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */
928 /* Reset the protocol slot to IPPROTO_RAW. */
929 ip_protox[ipproto] = pr - inetsw;
933 u_char inetctlerrmap[PRC_NCMDS] = {
935 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
936 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
937 EMSGSIZE, EHOSTUNREACH, 0, 0,
938 0, 0, EHOSTUNREACH, 0,
939 ENOPROTOOPT, ECONNREFUSED
943 * Forward a packet. If some error occurs return the sender
944 * an icmp packet. Note we can't always generate a meaningful
945 * icmp message because icmp doesn't have a large enough repertoire
946 * of codes and types.
948 * If not forwarding, just drop the packet. This could be confusing
949 * if ipforwarding was zero but some routing protocol was advancing
950 * us as a gateway to somewhere. However, we must let the routing
951 * protocol deal with that.
953 * The srcrt parameter indicates whether the packet is being forwarded
954 * via a source route.
957 ip_forward(struct mbuf *m, int srcrt)
959 struct ip *ip = mtod(m, struct ip *);
960 struct in_ifaddr *ia;
962 struct sockaddr_in *sin;
966 int error, type = 0, code = 0, mtu = 0;
970 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
971 IPSTAT_INC(ips_cantforward);
979 ip->ip_ttl <= IPTTLDEC) {
980 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 0, 0);
984 bzero(&ro, sizeof(ro));
985 sin = (struct sockaddr_in *)&ro.ro_dst;
986 sin->sin_family = AF_INET;
987 sin->sin_len = sizeof(*sin);
988 sin->sin_addr = ip->ip_dst;
989 flowid = m->m_pkthdr.flowid;
990 ro.ro_nh = fib4_lookup(M_GETFIB(m), ip->ip_dst, 0, NHR_REF, flowid);
991 if (ro.ro_nh != NULL) {
992 ia = ifatoia(ro.ro_nh->nh_ifa);
996 * Save the IP header and at most 8 bytes of the payload,
997 * in case we need to generate an ICMP message to the src.
999 * XXX this can be optimized a lot by saving the data in a local
1000 * buffer on the stack (72 bytes at most), and only allocating the
1001 * mbuf if really necessary. The vast majority of the packets
1002 * are forwarded without having to send an ICMP back (either
1003 * because unnecessary, or because rate limited), so we are
1004 * really we are wasting a lot of work here.
1006 * We don't use m_copym() because it might return a reference
1007 * to a shared cluster. Both this function and ip_output()
1008 * assume exclusive access to the IP header in `m', so any
1009 * data in a cluster may change before we reach icmp_error().
1011 mcopy = m_gethdr(M_NOWAIT, m->m_type);
1012 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_NOWAIT)) {
1014 * It's probably ok if the pkthdr dup fails (because
1015 * the deep copy of the tag chain failed), but for now
1016 * be conservative and just discard the copy since
1017 * code below may some day want the tags.
1022 if (mcopy != NULL) {
1023 mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy));
1024 mcopy->m_pkthdr.len = mcopy->m_len;
1025 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1028 if (V_ipstealth == 0)
1030 ip->ip_ttl -= IPTTLDEC;
1031 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1032 if (IPSEC_ENABLED(ipv4)) {
1033 if ((error = IPSEC_FORWARD(ipv4, m)) != 0) {
1034 /* mbuf consumed by IPsec */
1037 if (error != EINPROGRESS)
1038 IPSTAT_INC(ips_cantforward);
1041 /* No IPsec processing required */
1045 * If forwarding packet using same interface that it came in on,
1046 * perhaps should send a redirect to sender to shortcut a hop.
1047 * Only send redirect if source is sending directly to us,
1048 * and if packet was not source routed (or has any options).
1049 * Also, don't send redirect if forwarding using a default route
1050 * or a route modified by a redirect.
1053 if (!srcrt && V_ipsendredirects &&
1054 ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) {
1055 struct nhop_object *nh;
1059 if (nh != NULL && ((nh->nh_flags & (NHF_REDIRECT|NHF_DEFAULT)) == 0)) {
1060 struct in_ifaddr *nh_ia = (struct in_ifaddr *)(nh->nh_ifa);
1061 u_long src = ntohl(ip->ip_src.s_addr);
1063 if (nh_ia != NULL &&
1064 (src & nh_ia->ia_subnetmask) == nh_ia->ia_subnet) {
1065 /* Router requirements says to only send host redirects */
1066 type = ICMP_REDIRECT;
1067 code = ICMP_REDIRECT_HOST;
1068 if (nh->nh_flags & NHF_GATEWAY) {
1069 if (nh->gw_sa.sa_family == AF_INET)
1070 dest.s_addr = nh->gw4_sa.sin_addr.s_addr;
1071 else /* Do not redirect in case gw is AF_INET6 */
1074 dest.s_addr = ip->ip_dst.s_addr;
1079 error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
1081 if (error == EMSGSIZE && ro.ro_nh)
1082 mtu = ro.ro_nh->nh_mtu;
1086 IPSTAT_INC(ips_cantforward);
1088 IPSTAT_INC(ips_forward);
1090 IPSTAT_INC(ips_redirectsent);
1101 case 0: /* forwarded, but need redirect */
1102 /* type, code set above */
1110 type = ICMP_UNREACH;
1111 code = ICMP_UNREACH_HOST;
1115 type = ICMP_UNREACH;
1116 code = ICMP_UNREACH_NEEDFRAG;
1118 * If the MTU was set before make sure we are below the
1120 * If the MTU wasn't set before use the interface mtu or
1121 * fall back to the next smaller mtu step compared to the
1122 * current packet size.
1126 mtu = min(mtu, ia->ia_ifp->if_mtu);
1129 mtu = ia->ia_ifp->if_mtu;
1131 mtu = ip_next_mtu(ntohs(ip->ip_len), 0);
1133 IPSTAT_INC(ips_cantfrag);
1137 case EACCES: /* ipfw denied packet */
1141 icmp_error(mcopy, type, code, dest.s_addr, mtu);
1144 #define CHECK_SO_CT(sp, ct) \
1145 (((sp->so_options & SO_TIMESTAMP) && (sp->so_ts_clock == ct)) ? 1 : 0)
1148 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1154 if ((inp->inp_socket->so_options & SO_BINTIME) ||
1155 CHECK_SO_CT(inp->inp_socket, SO_TS_BINTIME)) {
1156 struct bintime boottimebin, bt;
1157 struct timespec ts1;
1159 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1161 mbuf_tstmp2timespec(m, &ts1);
1162 timespec2bintime(&ts1, &bt);
1163 getboottimebin(&boottimebin);
1164 bintime_add(&bt, &boottimebin);
1168 *mp = sbcreatecontrol((caddr_t)&bt, sizeof(bt),
1169 SCM_BINTIME, SOL_SOCKET);
1171 mp = &(*mp)->m_next;
1175 if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME_MICRO)) {
1176 struct bintime boottimebin, bt1;
1177 struct timespec ts1;
1180 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1182 mbuf_tstmp2timespec(m, &ts1);
1183 timespec2bintime(&ts1, &bt1);
1184 getboottimebin(&boottimebin);
1185 bintime_add(&bt1, &boottimebin);
1186 bintime2timeval(&bt1, &tv);
1190 *mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv),
1191 SCM_TIMESTAMP, SOL_SOCKET);
1193 mp = &(*mp)->m_next;
1196 } else if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME)) {
1197 struct bintime boottimebin;
1198 struct timespec ts, ts1;
1200 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1202 mbuf_tstmp2timespec(m, &ts);
1203 getboottimebin(&boottimebin);
1204 bintime2timespec(&boottimebin, &ts1);
1205 timespecadd(&ts, &ts1, &ts);
1209 *mp = sbcreatecontrol((caddr_t)&ts, sizeof(ts),
1210 SCM_REALTIME, SOL_SOCKET);
1212 mp = &(*mp)->m_next;
1215 } else if (CHECK_SO_CT(inp->inp_socket, SO_TS_MONOTONIC)) {
1218 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1220 mbuf_tstmp2timespec(m, &ts);
1223 *mp = sbcreatecontrol((caddr_t)&ts, sizeof(ts),
1224 SCM_MONOTONIC, SOL_SOCKET);
1226 mp = &(*mp)->m_next;
1230 if (stamped && (m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1232 struct sock_timestamp_info sti;
1234 bzero(&sti, sizeof(sti));
1235 sti.st_info_flags = ST_INFO_HW;
1236 if ((m->m_flags & M_TSTMP_HPREC) != 0)
1237 sti.st_info_flags |= ST_INFO_HW_HPREC;
1238 *mp = sbcreatecontrol((caddr_t)&sti, sizeof(sti), SCM_TIME_INFO,
1241 mp = &(*mp)->m_next;
1243 if (inp->inp_flags & INP_RECVDSTADDR) {
1244 *mp = sbcreatecontrol((caddr_t)&ip->ip_dst,
1245 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1247 mp = &(*mp)->m_next;
1249 if (inp->inp_flags & INP_RECVTTL) {
1250 *mp = sbcreatecontrol((caddr_t)&ip->ip_ttl,
1251 sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
1253 mp = &(*mp)->m_next;
1257 * Moving these out of udp_input() made them even more broken
1258 * than they already were.
1260 /* options were tossed already */
1261 if (inp->inp_flags & INP_RECVOPTS) {
1262 *mp = sbcreatecontrol((caddr_t)opts_deleted_above,
1263 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1265 mp = &(*mp)->m_next;
1267 /* ip_srcroute doesn't do what we want here, need to fix */
1268 if (inp->inp_flags & INP_RECVRETOPTS) {
1269 *mp = sbcreatecontrol((caddr_t)ip_srcroute(m),
1270 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1272 mp = &(*mp)->m_next;
1275 if (inp->inp_flags & INP_RECVIF) {
1278 struct sockaddr_dl sdl;
1281 struct sockaddr_dl *sdp;
1282 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1284 if ((ifp = m->m_pkthdr.rcvif) &&
1285 ifp->if_index && ifp->if_index <= V_if_index) {
1286 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1288 * Change our mind and don't try copy.
1290 if (sdp->sdl_family != AF_LINK ||
1291 sdp->sdl_len > sizeof(sdlbuf)) {
1294 bcopy(sdp, sdl2, sdp->sdl_len);
1298 offsetof(struct sockaddr_dl, sdl_data[0]);
1299 sdl2->sdl_family = AF_LINK;
1300 sdl2->sdl_index = 0;
1301 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1303 *mp = sbcreatecontrol((caddr_t)sdl2, sdl2->sdl_len,
1304 IP_RECVIF, IPPROTO_IP);
1306 mp = &(*mp)->m_next;
1308 if (inp->inp_flags & INP_RECVTOS) {
1309 *mp = sbcreatecontrol((caddr_t)&ip->ip_tos,
1310 sizeof(u_char), IP_RECVTOS, IPPROTO_IP);
1312 mp = &(*mp)->m_next;
1315 if (inp->inp_flags2 & INP_RECVFLOWID) {
1316 uint32_t flowid, flow_type;
1318 flowid = m->m_pkthdr.flowid;
1319 flow_type = M_HASHTYPE_GET(m);
1322 * XXX should handle the failure of one or the
1323 * other - don't populate both?
1325 *mp = sbcreatecontrol((caddr_t) &flowid,
1326 sizeof(uint32_t), IP_FLOWID, IPPROTO_IP);
1328 mp = &(*mp)->m_next;
1329 *mp = sbcreatecontrol((caddr_t) &flow_type,
1330 sizeof(uint32_t), IP_FLOWTYPE, IPPROTO_IP);
1332 mp = &(*mp)->m_next;
1336 if (inp->inp_flags2 & INP_RECVRSSBUCKETID) {
1337 uint32_t flowid, flow_type;
1338 uint32_t rss_bucketid;
1340 flowid = m->m_pkthdr.flowid;
1341 flow_type = M_HASHTYPE_GET(m);
1343 if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) {
1344 *mp = sbcreatecontrol((caddr_t) &rss_bucketid,
1345 sizeof(uint32_t), IP_RSSBUCKETID, IPPROTO_IP);
1347 mp = &(*mp)->m_next;
1354 * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
1355 * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
1356 * locking. This code remains in ip_input.c as ip_mroute.c is optionally
1359 VNET_DEFINE_STATIC(int, ip_rsvp_on);
1360 VNET_DEFINE(struct socket *, ip_rsvpd);
1362 #define V_ip_rsvp_on VNET(ip_rsvp_on)
1365 ip_rsvp_init(struct socket *so)
1368 if (so->so_type != SOCK_RAW ||
1369 so->so_proto->pr_protocol != IPPROTO_RSVP)
1372 if (V_ip_rsvpd != NULL)
1377 * This may seem silly, but we need to be sure we don't over-increment
1378 * the RSVP counter, in case something slips up.
1380 if (!V_ip_rsvp_on) {
1394 * This may seem silly, but we need to be sure we don't over-decrement
1395 * the RSVP counter, in case something slips up.
1405 rsvp_input(struct mbuf **mp, int *offp, int proto)
1412 if (rsvp_input_p) { /* call the real one if loaded */
1414 rsvp_input_p(mp, offp, proto);
1415 return (IPPROTO_DONE);
1418 /* Can still get packets with rsvp_on = 0 if there is a local member
1419 * of the group to which the RSVP packet is addressed. But in this
1420 * case we want to throw the packet away.
1425 return (IPPROTO_DONE);
1428 if (V_ip_rsvpd != NULL) {
1430 rip_input(mp, offp, proto);
1431 return (IPPROTO_DONE);
1433 /* Drop the packet */
1435 return (IPPROTO_DONE);