2 * SPDX-License-Identifier: BSD-3-Clause
4 * Copyright (c) 1982, 1986, 1988, 1993
5 * The Regents of the University of California. All rights reserved.
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 * 3. Neither the name of the University nor the names of its contributors
16 * may be used to endorse or promote products derived from this software
17 * without specific prior written permission.
19 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94
34 #include <sys/cdefs.h>
35 __FBSDID("$FreeBSD$");
37 #include "opt_bootp.h"
38 #include "opt_ipstealth.h"
39 #include "opt_ipsec.h"
40 #include "opt_route.h"
43 #include <sys/param.h>
44 #include <sys/systm.h>
45 #include <sys/hhook.h>
47 #include <sys/malloc.h>
48 #include <sys/domain.h>
49 #include <sys/protosw.h>
50 #include <sys/socket.h>
52 #include <sys/kernel.h>
54 #include <sys/rmlock.h>
55 #include <sys/rwlock.h>
57 #include <sys/syslog.h>
58 #include <sys/sysctl.h>
62 #include <net/if_types.h>
63 #include <net/if_var.h>
64 #include <net/if_dl.h>
65 #include <net/route.h>
66 #include <net/netisr.h>
67 #include <net/rss_config.h>
70 #include <netinet/in.h>
71 #include <netinet/in_kdtrace.h>
72 #include <netinet/in_systm.h>
73 #include <netinet/in_var.h>
74 #include <netinet/ip.h>
75 #include <netinet/in_pcb.h>
76 #include <netinet/ip_var.h>
77 #include <netinet/ip_fw.h>
78 #include <netinet/ip_icmp.h>
79 #include <netinet/ip_options.h>
80 #include <machine/in_cksum.h>
81 #include <netinet/ip_carp.h>
82 #include <netinet/in_rss.h>
84 #include <netipsec/ipsec_support.h>
86 #include <sys/socketvar.h>
88 #include <security/mac/mac_framework.h>
91 CTASSERT(sizeof(struct ip) == 20);
94 /* IP reassembly functions are defined in ip_reass.c. */
95 extern void ipreass_init(void);
96 extern void ipreass_drain(void);
97 extern void ipreass_slowtimo(void);
99 extern void ipreass_destroy(void);
102 struct rmlock in_ifaddr_lock;
103 RM_SYSINIT(in_ifaddr_lock, &in_ifaddr_lock, "in_ifaddr_lock");
105 VNET_DEFINE(int, rsvp_on);
107 VNET_DEFINE(int, ipforwarding);
108 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_VNET | CTLFLAG_RW,
109 &VNET_NAME(ipforwarding), 0,
110 "Enable IP forwarding between interfaces");
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(struct pfil_head, inet_pfil_hook); /* 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 */
181 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
182 &ip_mtu, 0, "Default MTU");
186 VNET_DEFINE(int, ipstealth);
187 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_VNET | CTLFLAG_RW,
188 &VNET_NAME(ipstealth), 0,
189 "IP stealth mode, no TTL decrementation on forwarding");
193 * IP statistics are stored in the "array" of counter(9)s.
195 VNET_PCPUSTAT_DEFINE(struct ipstat, ipstat);
196 VNET_PCPUSTAT_SYSINIT(ipstat);
197 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, IPCTL_STATS, stats, struct ipstat, ipstat,
198 "IP statistics (struct ipstat, netinet/ip_var.h)");
201 VNET_PCPUSTAT_SYSUNINIT(ipstat);
205 * Kernel module interface for updating ipstat. The argument is an index
206 * into ipstat treated as an array.
209 kmod_ipstat_inc(int statnum)
212 counter_u64_add(VNET(ipstat)[statnum], 1);
216 kmod_ipstat_dec(int statnum)
219 counter_u64_add(VNET(ipstat)[statnum], -1);
223 sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS)
227 netisr_getqlimit(&ip_nh, &qlimit);
228 error = sysctl_handle_int(oidp, &qlimit, 0, req);
229 if (error || !req->newptr)
233 return (netisr_setqlimit(&ip_nh, qlimit));
235 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen,
236 CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet_intr_queue_maxlen, "I",
237 "Maximum size of the IP input queue");
240 sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS)
242 u_int64_t qdrops_long;
245 netisr_getqdrops(&ip_nh, &qdrops_long);
246 qdrops = qdrops_long;
247 error = sysctl_handle_int(oidp, &qdrops, 0, req);
248 if (error || !req->newptr)
252 netisr_clearqdrops(&ip_nh);
256 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops,
257 CTLTYPE_INT|CTLFLAG_RD, 0, 0, sysctl_netinet_intr_queue_drops, "I",
258 "Number of packets dropped from the IP input queue");
262 sysctl_netinet_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS)
266 netisr_getqlimit(&ip_direct_nh, &qlimit);
267 error = sysctl_handle_int(oidp, &qlimit, 0, req);
268 if (error || !req->newptr)
272 return (netisr_setqlimit(&ip_direct_nh, qlimit));
274 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRDQMAXLEN, intr_direct_queue_maxlen,
275 CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet_intr_direct_queue_maxlen,
276 "I", "Maximum size of the IP direct input queue");
279 sysctl_netinet_intr_direct_queue_drops(SYSCTL_HANDLER_ARGS)
281 u_int64_t qdrops_long;
284 netisr_getqdrops(&ip_direct_nh, &qdrops_long);
285 qdrops = qdrops_long;
286 error = sysctl_handle_int(oidp, &qdrops, 0, req);
287 if (error || !req->newptr)
291 netisr_clearqdrops(&ip_direct_nh);
295 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRDQDROPS, intr_direct_queue_drops,
296 CTLTYPE_INT|CTLFLAG_RD, 0, 0, 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.
310 CK_STAILQ_INIT(&V_in_ifaddrhead);
311 V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask);
313 /* Initialize IP reassembly queue. */
316 /* Initialize packet filter hooks. */
317 V_inet_pfil_hook.ph_type = PFIL_TYPE_AF;
318 V_inet_pfil_hook.ph_af = AF_INET;
319 if ((i = pfil_head_register(&V_inet_pfil_hook)) != 0)
320 printf("%s: WARNING: unable to register pfil hook, "
321 "error %d\n", __func__, i);
323 if (hhook_head_register(HHOOK_TYPE_IPSEC_IN, AF_INET,
324 &V_ipsec_hhh_in[HHOOK_IPSEC_INET],
325 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
326 printf("%s: WARNING: unable to register input helper hook\n",
328 if (hhook_head_register(HHOOK_TYPE_IPSEC_OUT, AF_INET,
329 &V_ipsec_hhh_out[HHOOK_IPSEC_INET],
330 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
331 printf("%s: WARNING: unable to register output helper hook\n",
334 /* Skip initialization of globals for non-default instances. */
336 if (!IS_DEFAULT_VNET(curvnet)) {
337 netisr_register_vnet(&ip_nh);
339 netisr_register_vnet(&ip_direct_nh);
345 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
347 panic("ip_init: PF_INET not found");
349 /* Initialize the entire ip_protox[] array to IPPROTO_RAW. */
350 for (i = 0; i < IPPROTO_MAX; i++)
351 ip_protox[i] = pr - inetsw;
353 * Cycle through IP protocols and put them into the appropriate place
356 for (pr = inetdomain.dom_protosw;
357 pr < inetdomain.dom_protoswNPROTOSW; pr++)
358 if (pr->pr_domain->dom_family == PF_INET &&
359 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
360 /* Be careful to only index valid IP protocols. */
361 if (pr->pr_protocol < IPPROTO_MAX)
362 ip_protox[pr->pr_protocol] = pr - inetsw;
365 netisr_register(&ip_nh);
367 netisr_register(&ip_direct_nh);
373 ip_destroy(void *unused __unused)
379 netisr_unregister_vnet(&ip_direct_nh);
381 netisr_unregister_vnet(&ip_nh);
383 if ((error = pfil_head_unregister(&V_inet_pfil_hook)) != 0)
384 printf("%s: WARNING: unable to unregister pfil hook, "
385 "error %d\n", __func__, error);
387 error = hhook_head_deregister(V_ipsec_hhh_in[HHOOK_IPSEC_INET]);
389 printf("%s: WARNING: unable to deregister input helper hook "
390 "type HHOOK_TYPE_IPSEC_IN, id HHOOK_IPSEC_INET: "
391 "error %d returned\n", __func__, error);
393 error = hhook_head_deregister(V_ipsec_hhh_out[HHOOK_IPSEC_INET]);
395 printf("%s: WARNING: unable to deregister output helper hook "
396 "type HHOOK_TYPE_IPSEC_OUT, id HHOOK_IPSEC_INET: "
397 "error %d returned\n", __func__, error);
400 /* Remove the IPv4 addresses from all interfaces. */
403 /* Make sure the IPv4 routes are gone as well. */
405 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link)
406 rt_flushifroutes_af(ifp, AF_INET);
409 /* Destroy IP reassembly queue. */
412 /* Cleanup in_ifaddr hash table; should be empty. */
413 hashdestroy(V_in_ifaddrhashtbl, M_IFADDR, V_in_ifaddrhmask);
416 VNET_SYSUNINIT(ip, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip_destroy, NULL);
421 * IP direct input routine.
423 * This is called when reinjecting completed fragments where
424 * all of the previous checking and book-keeping has been done.
427 ip_direct_input(struct mbuf *m)
432 ip = mtod(m, struct ip *);
433 hlen = ip->ip_hl << 2;
435 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
436 if (IPSEC_ENABLED(ipv4)) {
437 if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
441 IPSTAT_INC(ips_delivered);
442 (*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
448 * Ip input routine. Checksum and byte swap header. If fragmented
449 * try to reassemble. Process options. Pass to next level.
452 ip_input(struct mbuf *m)
454 struct rm_priotracker in_ifa_tracker;
455 struct ip *ip = NULL;
456 struct in_ifaddr *ia = NULL;
459 int checkif, hlen = 0;
460 uint16_t sum, ip_len;
461 int dchg = 0; /* dest changed after fw */
462 struct in_addr odst; /* original dst address */
466 if (m->m_flags & M_FASTFWD_OURS) {
467 m->m_flags &= ~M_FASTFWD_OURS;
468 /* Set up some basics that will be used later. */
469 ip = mtod(m, struct ip *);
470 hlen = ip->ip_hl << 2;
471 ip_len = ntohs(ip->ip_len);
475 IPSTAT_INC(ips_total);
477 if (m->m_pkthdr.len < sizeof(struct ip))
480 if (m->m_len < sizeof (struct ip) &&
481 (m = m_pullup(m, sizeof (struct ip))) == NULL) {
482 IPSTAT_INC(ips_toosmall);
485 ip = mtod(m, struct ip *);
487 if (ip->ip_v != IPVERSION) {
488 IPSTAT_INC(ips_badvers);
492 hlen = ip->ip_hl << 2;
493 if (hlen < sizeof(struct ip)) { /* minimum header length */
494 IPSTAT_INC(ips_badhlen);
497 if (hlen > m->m_len) {
498 if ((m = m_pullup(m, hlen)) == NULL) {
499 IPSTAT_INC(ips_badhlen);
502 ip = mtod(m, struct ip *);
505 IP_PROBE(receive, NULL, NULL, ip, m->m_pkthdr.rcvif, ip, NULL);
507 /* IN_LOOPBACK must not appear on the wire - RFC1122 */
508 ifp = m->m_pkthdr.rcvif;
509 if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) ||
510 IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) {
511 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
512 IPSTAT_INC(ips_badaddr);
517 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
518 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
520 if (hlen == sizeof(struct ip)) {
521 sum = in_cksum_hdr(ip);
523 sum = in_cksum(m, hlen);
527 IPSTAT_INC(ips_badsum);
532 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
533 /* packet is dropped by traffic conditioner */
537 ip_len = ntohs(ip->ip_len);
539 IPSTAT_INC(ips_badlen);
544 * Check that the amount of data in the buffers
545 * is as at least much as the IP header would have us expect.
546 * Trim mbufs if longer than we expect.
547 * Drop packet if shorter than we expect.
549 if (m->m_pkthdr.len < ip_len) {
551 IPSTAT_INC(ips_tooshort);
554 if (m->m_pkthdr.len > ip_len) {
555 if (m->m_len == m->m_pkthdr.len) {
557 m->m_pkthdr.len = ip_len;
559 m_adj(m, ip_len - m->m_pkthdr.len);
563 * Try to forward the packet, but if we fail continue.
564 * ip_tryforward() does not generate redirects, so fall
565 * through to normal processing if redirects are required.
566 * ip_tryforward() does inbound and outbound packet firewall
567 * processing. If firewall has decided that destination becomes
568 * our local address, it sets M_FASTFWD_OURS flag. In this
569 * case skip another inbound firewall processing and update
572 if (V_ipforwarding != 0
573 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
574 && (!IPSEC_ENABLED(ipv4) ||
575 IPSEC_CAPS(ipv4, m, IPSEC_CAP_OPERABLE) == 0)
578 if ((m = ip_tryforward(m)) == NULL)
580 if (m->m_flags & M_FASTFWD_OURS) {
581 m->m_flags &= ~M_FASTFWD_OURS;
582 ip = mtod(m, struct ip *);
587 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
589 * Bypass packet filtering for packets previously handled by IPsec.
591 if (IPSEC_ENABLED(ipv4) &&
592 IPSEC_CAPS(ipv4, m, IPSEC_CAP_BYPASS_FILTER) != 0)
597 * Run through list of hooks for input packets.
599 * NB: Beware of the destination address changing (e.g.
600 * by NAT rewriting). When this happens, tell
601 * ip_forward to do the right thing.
604 /* Jump over all PFIL processing if hooks are not active. */
605 if (!PFIL_HOOKED(&V_inet_pfil_hook))
609 if (pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_IN, 0, NULL) != 0)
611 if (m == NULL) /* consumed by filter */
614 ip = mtod(m, struct ip *);
615 dchg = (odst.s_addr != ip->ip_dst.s_addr);
616 ifp = m->m_pkthdr.rcvif;
618 if (m->m_flags & M_FASTFWD_OURS) {
619 m->m_flags &= ~M_FASTFWD_OURS;
622 if (m->m_flags & M_IP_NEXTHOP) {
623 if (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) {
625 * Directly ship the packet on. This allows
626 * forwarding packets originally destined to us
627 * to some other directly connected host.
636 * Process options and, if not destined for us,
637 * ship it on. ip_dooptions returns 1 when an
638 * error was detected (causing an icmp message
639 * to be sent and the original packet to be freed).
641 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
644 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
645 * matter if it is destined to another node, or whether it is
646 * a multicast one, RSVP wants it! and prevents it from being forwarded
647 * anywhere else. Also checks if the rsvp daemon is running before
648 * grabbing the packet.
650 if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP)
654 * Check our list of addresses, to see if the packet is for us.
655 * If we don't have any addresses, assume any unicast packet
656 * we receive might be for us (and let the upper layers deal
659 if (CK_STAILQ_EMPTY(&V_in_ifaddrhead) &&
660 (m->m_flags & (M_MCAST|M_BCAST)) == 0)
664 * Enable a consistency check between the destination address
665 * and the arrival interface for a unicast packet (the RFC 1122
666 * strong ES model) if IP forwarding is disabled and the packet
667 * is not locally generated and the packet is not subject to
670 * XXX - Checking also should be disabled if the destination
671 * address is ipnat'ed to a different interface.
673 * XXX - Checking is incompatible with IP aliases added
674 * to the loopback interface instead of the interface where
675 * the packets are received.
677 * XXX - This is the case for carp vhost IPs as well so we
678 * insert a workaround. If the packet got here, we already
679 * checked with carp_iamatch() and carp_forus().
681 checkif = V_ip_checkinterface && (V_ipforwarding == 0) &&
682 ifp != NULL && ((ifp->if_flags & IFF_LOOPBACK) == 0) &&
683 ifp->if_carp == NULL && (dchg == 0);
686 * Check for exact addresses in the hash bucket.
688 IN_IFADDR_RLOCK(&in_ifa_tracker);
689 LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
691 * If the address matches, verify that the packet
692 * arrived via the correct interface if checking is
695 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr &&
696 (!checkif || ia->ia_ifp == ifp)) {
697 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
698 counter_u64_add(ia->ia_ifa.ifa_ibytes,
700 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
704 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
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) {
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,
725 IF_ADDR_RUNLOCK(ifp);
729 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) {
730 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
731 counter_u64_add(ia->ia_ifa.ifa_ibytes,
733 IF_ADDR_RUNLOCK(ifp);
738 IF_ADDR_RUNLOCK(ifp);
741 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
743 * RFC 3927 2.7: Do not forward multicast packets from
746 if (V_ip_mrouter && !IN_LINKLOCAL(ntohl(ip->ip_src.s_addr))) {
748 * If we are acting as a multicast router, all
749 * incoming multicast packets are passed to the
750 * kernel-level multicast forwarding function.
751 * The packet is returned (relatively) intact; if
752 * ip_mforward() returns a non-zero value, the packet
753 * must be discarded, else it may be accepted below.
755 if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) {
756 IPSTAT_INC(ips_cantforward);
762 * The process-level routing daemon needs to receive
763 * all multicast IGMP packets, whether or not this
764 * host belongs to their destination groups.
766 if (ip->ip_p == IPPROTO_IGMP)
768 IPSTAT_INC(ips_forward);
771 * Assume the packet is for us, to avoid prematurely taking
772 * a lock on the in_multi hash. Protocols must perform
773 * their own filtering and update statistics accordingly.
777 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
779 if (ip->ip_dst.s_addr == INADDR_ANY)
781 /* RFC 3927 2.7: Do not forward packets to or from IN_LINKLOCAL. */
782 if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr)) ||
783 IN_LINKLOCAL(ntohl(ip->ip_src.s_addr))) {
784 IPSTAT_INC(ips_cantforward);
790 * Not for us; forward if possible and desirable.
792 if (V_ipforwarding == 0) {
793 IPSTAT_INC(ips_cantforward);
803 * IPSTEALTH: Process non-routing options only
804 * if the packet is destined for us.
806 if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1))
808 #endif /* IPSTEALTH */
811 * Attempt reassembly; if it succeeds, proceed.
812 * ip_reass() will return a different mbuf.
814 if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) {
815 /* XXXGL: shouldn't we save & set m_flags? */
819 ip = mtod(m, struct ip *);
820 /* Get the header length of the reassembled packet */
821 hlen = ip->ip_hl << 2;
824 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
825 if (IPSEC_ENABLED(ipv4)) {
826 if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
832 * Switch out to protocol's input routine.
834 IPSTAT_INC(ips_delivered);
836 (*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
843 * IP timer processing;
844 * if a timer expires on a reassembly
850 VNET_ITERATOR_DECL(vnet_iter);
852 VNET_LIST_RLOCK_NOSLEEP();
853 VNET_FOREACH(vnet_iter) {
854 CURVNET_SET(vnet_iter);
858 VNET_LIST_RUNLOCK_NOSLEEP();
864 VNET_ITERATOR_DECL(vnet_iter);
866 VNET_LIST_RLOCK_NOSLEEP();
867 VNET_FOREACH(vnet_iter) {
868 CURVNET_SET(vnet_iter);
872 VNET_LIST_RUNLOCK_NOSLEEP();
876 * The protocol to be inserted into ip_protox[] must be already registered
877 * in inetsw[], either statically or through pf_proto_register().
880 ipproto_register(short ipproto)
885 if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
886 return (EPROTONOSUPPORT);
889 * The protocol slot must not be occupied by another protocol
890 * already. An index pointing to IPPROTO_RAW is unused.
892 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
894 return (EPFNOSUPPORT);
895 if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */
898 /* Find the protocol position in inetsw[] and set the index. */
899 for (pr = inetdomain.dom_protosw;
900 pr < inetdomain.dom_protoswNPROTOSW; pr++) {
901 if (pr->pr_domain->dom_family == PF_INET &&
902 pr->pr_protocol && pr->pr_protocol == ipproto) {
903 ip_protox[pr->pr_protocol] = pr - inetsw;
907 return (EPROTONOSUPPORT);
911 ipproto_unregister(short ipproto)
916 if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
917 return (EPROTONOSUPPORT);
919 /* Check if the protocol was indeed registered. */
920 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
922 return (EPFNOSUPPORT);
923 if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */
926 /* Reset the protocol slot to IPPROTO_RAW. */
927 ip_protox[ipproto] = pr - inetsw;
931 u_char inetctlerrmap[PRC_NCMDS] = {
933 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
934 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
935 EMSGSIZE, EHOSTUNREACH, 0, 0,
936 0, 0, EHOSTUNREACH, 0,
937 ENOPROTOOPT, ECONNREFUSED
941 * Forward a packet. If some error occurs return the sender
942 * an icmp packet. Note we can't always generate a meaningful
943 * icmp message because icmp doesn't have a large enough repertoire
944 * of codes and types.
946 * If not forwarding, just drop the packet. This could be confusing
947 * if ipforwarding was zero but some routing protocol was advancing
948 * us as a gateway to somewhere. However, we must let the routing
949 * protocol deal with that.
951 * The srcrt parameter indicates whether the packet is being forwarded
952 * via a source route.
955 ip_forward(struct mbuf *m, int srcrt)
957 struct ip *ip = mtod(m, struct ip *);
958 struct in_ifaddr *ia;
960 struct sockaddr_in *sin;
963 int error, type = 0, code = 0, mtu = 0;
965 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
966 IPSTAT_INC(ips_cantforward);
974 ip->ip_ttl <= IPTTLDEC) {
975 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 0, 0);
979 bzero(&ro, sizeof(ro));
980 sin = (struct sockaddr_in *)&ro.ro_dst;
981 sin->sin_family = AF_INET;
982 sin->sin_len = sizeof(*sin);
983 sin->sin_addr = ip->ip_dst;
985 rtalloc_mpath_fib(&ro,
986 ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
989 in_rtalloc_ign(&ro, 0, M_GETFIB(m));
992 if (ro.ro_rt != NULL) {
993 ia = ifatoia(ro.ro_rt->rt_ifa);
997 * Save the IP header and at most 8 bytes of the payload,
998 * in case we need to generate an ICMP message to the src.
1000 * XXX this can be optimized a lot by saving the data in a local
1001 * buffer on the stack (72 bytes at most), and only allocating the
1002 * mbuf if really necessary. The vast majority of the packets
1003 * are forwarded without having to send an ICMP back (either
1004 * because unnecessary, or because rate limited), so we are
1005 * really we are wasting a lot of work here.
1007 * We don't use m_copym() because it might return a reference
1008 * to a shared cluster. Both this function and ip_output()
1009 * assume exclusive access to the IP header in `m', so any
1010 * data in a cluster may change before we reach icmp_error().
1012 mcopy = m_gethdr(M_NOWAIT, m->m_type);
1013 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_NOWAIT)) {
1015 * It's probably ok if the pkthdr dup fails (because
1016 * the deep copy of the tag chain failed), but for now
1017 * be conservative and just discard the copy since
1018 * code below may some day want the tags.
1023 if (mcopy != NULL) {
1024 mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy));
1025 mcopy->m_pkthdr.len = mcopy->m_len;
1026 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1029 if (V_ipstealth == 0)
1031 ip->ip_ttl -= IPTTLDEC;
1032 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1033 if (IPSEC_ENABLED(ipv4)) {
1034 if ((error = IPSEC_FORWARD(ipv4, m)) != 0) {
1035 /* 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) {
1059 if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1060 satosin(rt_key(rt))->sin_addr.s_addr != 0) {
1061 #define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa))
1062 u_long src = ntohl(ip->ip_src.s_addr);
1065 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1066 if (rt->rt_flags & RTF_GATEWAY)
1067 dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr;
1069 dest.s_addr = ip->ip_dst.s_addr;
1070 /* Router requirements says to only send host redirects */
1071 type = ICMP_REDIRECT;
1072 code = ICMP_REDIRECT_HOST;
1077 error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
1079 if (error == EMSGSIZE && ro.ro_rt)
1080 mtu = ro.ro_rt->rt_mtu;
1084 IPSTAT_INC(ips_cantforward);
1086 IPSTAT_INC(ips_forward);
1088 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);
1146 #define CHECK_SO_CT(sp, ct) \
1147 (((sp->so_options & SO_TIMESTAMP) && (sp->so_ts_clock == ct)) ? 1 : 0)
1150 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1156 if ((inp->inp_socket->so_options & SO_BINTIME) ||
1157 CHECK_SO_CT(inp->inp_socket, SO_TS_BINTIME)) {
1158 struct bintime boottimebin, bt;
1159 struct timespec ts1;
1161 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1163 mbuf_tstmp2timespec(m, &ts1);
1164 timespec2bintime(&ts1, &bt);
1165 getboottimebin(&boottimebin);
1166 bintime_add(&bt, &boottimebin);
1170 *mp = sbcreatecontrol((caddr_t)&bt, sizeof(bt),
1171 SCM_BINTIME, SOL_SOCKET);
1173 mp = &(*mp)->m_next;
1177 if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME_MICRO)) {
1178 struct bintime boottimebin, bt1;
1179 struct timespec ts1;;
1182 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1184 mbuf_tstmp2timespec(m, &ts1);
1185 timespec2bintime(&ts1, &bt1);
1186 getboottimebin(&boottimebin);
1187 bintime_add(&bt1, &boottimebin);
1188 bintime2timeval(&bt1, &tv);
1192 *mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv),
1193 SCM_TIMESTAMP, SOL_SOCKET);
1195 mp = &(*mp)->m_next;
1198 } else if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME)) {
1199 struct bintime boottimebin;
1200 struct timespec ts, ts1;
1202 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1204 mbuf_tstmp2timespec(m, &ts);
1205 getboottimebin(&boottimebin);
1206 bintime2timespec(&boottimebin, &ts1);
1207 timespecadd(&ts, &ts1, &ts);
1211 *mp = sbcreatecontrol((caddr_t)&ts, sizeof(ts),
1212 SCM_REALTIME, SOL_SOCKET);
1214 mp = &(*mp)->m_next;
1217 } else if (CHECK_SO_CT(inp->inp_socket, SO_TS_MONOTONIC)) {
1220 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1222 mbuf_tstmp2timespec(m, &ts);
1225 *mp = sbcreatecontrol((caddr_t)&ts, sizeof(ts),
1226 SCM_MONOTONIC, SOL_SOCKET);
1228 mp = &(*mp)->m_next;
1232 if (stamped && (m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1234 struct sock_timestamp_info sti;
1236 bzero(&sti, sizeof(sti));
1237 sti.st_info_flags = ST_INFO_HW;
1238 if ((m->m_flags & M_TSTMP_HPREC) != 0)
1239 sti.st_info_flags |= ST_INFO_HW_HPREC;
1240 *mp = sbcreatecontrol((caddr_t)&sti, sizeof(sti), SCM_TIME_INFO,
1243 mp = &(*mp)->m_next;
1245 if (inp->inp_flags & INP_RECVDSTADDR) {
1246 *mp = sbcreatecontrol((caddr_t)&ip->ip_dst,
1247 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1249 mp = &(*mp)->m_next;
1251 if (inp->inp_flags & INP_RECVTTL) {
1252 *mp = sbcreatecontrol((caddr_t)&ip->ip_ttl,
1253 sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
1255 mp = &(*mp)->m_next;
1259 * Moving these out of udp_input() made them even more broken
1260 * than they already were.
1262 /* options were tossed already */
1263 if (inp->inp_flags & INP_RECVOPTS) {
1264 *mp = sbcreatecontrol((caddr_t)opts_deleted_above,
1265 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1267 mp = &(*mp)->m_next;
1269 /* ip_srcroute doesn't do what we want here, need to fix */
1270 if (inp->inp_flags & INP_RECVRETOPTS) {
1271 *mp = sbcreatecontrol((caddr_t)ip_srcroute(m),
1272 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1274 mp = &(*mp)->m_next;
1277 if (inp->inp_flags & INP_RECVIF) {
1280 struct sockaddr_dl sdl;
1283 struct sockaddr_dl *sdp;
1284 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1286 if ((ifp = m->m_pkthdr.rcvif) &&
1287 ifp->if_index && ifp->if_index <= V_if_index) {
1288 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1290 * Change our mind and don't try copy.
1292 if (sdp->sdl_family != AF_LINK ||
1293 sdp->sdl_len > sizeof(sdlbuf)) {
1296 bcopy(sdp, sdl2, sdp->sdl_len);
1300 offsetof(struct sockaddr_dl, sdl_data[0]);
1301 sdl2->sdl_family = AF_LINK;
1302 sdl2->sdl_index = 0;
1303 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1305 *mp = sbcreatecontrol((caddr_t)sdl2, sdl2->sdl_len,
1306 IP_RECVIF, IPPROTO_IP);
1308 mp = &(*mp)->m_next;
1310 if (inp->inp_flags & INP_RECVTOS) {
1311 *mp = sbcreatecontrol((caddr_t)&ip->ip_tos,
1312 sizeof(u_char), IP_RECVTOS, IPPROTO_IP);
1314 mp = &(*mp)->m_next;
1317 if (inp->inp_flags2 & INP_RECVFLOWID) {
1318 uint32_t flowid, flow_type;
1320 flowid = m->m_pkthdr.flowid;
1321 flow_type = M_HASHTYPE_GET(m);
1324 * XXX should handle the failure of one or the
1325 * other - don't populate both?
1327 *mp = sbcreatecontrol((caddr_t) &flowid,
1328 sizeof(uint32_t), IP_FLOWID, IPPROTO_IP);
1330 mp = &(*mp)->m_next;
1331 *mp = sbcreatecontrol((caddr_t) &flow_type,
1332 sizeof(uint32_t), IP_FLOWTYPE, IPPROTO_IP);
1334 mp = &(*mp)->m_next;
1338 if (inp->inp_flags2 & INP_RECVRSSBUCKETID) {
1339 uint32_t flowid, flow_type;
1340 uint32_t rss_bucketid;
1342 flowid = m->m_pkthdr.flowid;
1343 flow_type = M_HASHTYPE_GET(m);
1345 if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) {
1346 *mp = sbcreatecontrol((caddr_t) &rss_bucketid,
1347 sizeof(uint32_t), IP_RSSBUCKETID, IPPROTO_IP);
1349 mp = &(*mp)->m_next;
1356 * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
1357 * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
1358 * locking. This code remains in ip_input.c as ip_mroute.c is optionally
1361 VNET_DEFINE_STATIC(int, ip_rsvp_on);
1362 VNET_DEFINE(struct socket *, ip_rsvpd);
1364 #define V_ip_rsvp_on VNET(ip_rsvp_on)
1367 ip_rsvp_init(struct socket *so)
1370 if (so->so_type != SOCK_RAW ||
1371 so->so_proto->pr_protocol != IPPROTO_RSVP)
1374 if (V_ip_rsvpd != NULL)
1379 * This may seem silly, but we need to be sure we don't over-increment
1380 * the RSVP counter, in case something slips up.
1382 if (!V_ip_rsvp_on) {
1396 * This may seem silly, but we need to be sure we don't over-decrement
1397 * the RSVP counter, in case something slips up.
1407 rsvp_input(struct mbuf **mp, int *offp, int proto)
1414 if (rsvp_input_p) { /* call the real one if loaded */
1416 rsvp_input_p(mp, offp, proto);
1417 return (IPPROTO_DONE);
1420 /* Can still get packets with rsvp_on = 0 if there is a local member
1421 * of the group to which the RSVP packet is addressed. But in this
1422 * case we want to throw the packet away.
1427 return (IPPROTO_DONE);
1430 if (V_ip_rsvpd != NULL) {
1432 rip_input(mp, offp, proto);
1433 return (IPPROTO_DONE);
1435 /* Drop the packet */
1437 return (IPPROTO_DONE);