2 * Copyright (c) 1982, 1986, 1988, 1993
3 * The Regents of the University of California. All rights reserved.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * 4. Neither the name of the University nor the names of its contributors
14 * may be used to endorse or promote products derived from this software
15 * without specific prior written permission.
17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94
32 #include <sys/cdefs.h>
33 __FBSDID("$FreeBSD$");
35 #include "opt_bootp.h"
36 #include "opt_ipstealth.h"
37 #include "opt_ipsec.h"
38 #include "opt_route.h"
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/hhook.h>
45 #include <sys/malloc.h>
46 #include <sys/domain.h>
47 #include <sys/protosw.h>
48 #include <sys/socket.h>
50 #include <sys/kernel.h>
52 #include <sys/rmlock.h>
53 #include <sys/rwlock.h>
55 #include <sys/syslog.h>
56 #include <sys/sysctl.h>
60 #include <net/if_types.h>
61 #include <net/if_var.h>
62 #include <net/if_dl.h>
63 #include <net/route.h>
64 #include <net/netisr.h>
65 #include <net/rss_config.h>
68 #include <netinet/in.h>
69 #include <netinet/in_kdtrace.h>
70 #include <netinet/in_systm.h>
71 #include <netinet/in_var.h>
72 #include <netinet/ip.h>
73 #include <netinet/in_pcb.h>
74 #include <netinet/ip_var.h>
75 #include <netinet/ip_fw.h>
76 #include <netinet/ip_icmp.h>
77 #include <netinet/ip_options.h>
78 #include <machine/in_cksum.h>
79 #include <netinet/ip_carp.h>
80 #include <netinet/in_rss.h>
82 #include <netipsec/ipsec_support.h>
84 #include <sys/socketvar.h>
86 #include <security/mac/mac_framework.h>
89 CTASSERT(sizeof(struct ip) == 20);
92 /* IP reassembly functions are defined in ip_reass.c. */
93 extern void ipreass_init(void);
94 extern void ipreass_drain(void);
95 extern void ipreass_slowtimo(void);
97 extern void ipreass_destroy(void);
100 struct rmlock in_ifaddr_lock;
101 RM_SYSINIT(in_ifaddr_lock, &in_ifaddr_lock, "in_ifaddr_lock");
103 VNET_DEFINE(int, rsvp_on);
105 VNET_DEFINE(int, ipforwarding);
106 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_VNET | CTLFLAG_RW,
107 &VNET_NAME(ipforwarding), 0,
108 "Enable IP forwarding between interfaces");
110 static VNET_DEFINE(int, ipsendredirects) = 1; /* XXX */
111 #define V_ipsendredirects VNET(ipsendredirects)
112 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_VNET | CTLFLAG_RW,
113 &VNET_NAME(ipsendredirects), 0,
114 "Enable sending IP redirects");
117 * XXX - Setting ip_checkinterface mostly implements the receive side of
118 * the Strong ES model described in RFC 1122, but since the routing table
119 * and transmit implementation do not implement the Strong ES model,
120 * setting this to 1 results in an odd hybrid.
122 * XXX - ip_checkinterface currently must be disabled if you use ipnat
123 * to translate the destination address to another local interface.
125 * XXX - ip_checkinterface must be disabled if you add IP aliases
126 * to the loopback interface instead of the interface where the
127 * packets for those addresses are received.
129 static VNET_DEFINE(int, ip_checkinterface);
130 #define V_ip_checkinterface VNET(ip_checkinterface)
131 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_VNET | CTLFLAG_RW,
132 &VNET_NAME(ip_checkinterface), 0,
133 "Verify packet arrives on correct interface");
135 VNET_DEFINE(struct pfil_head, inet_pfil_hook); /* Packet filter hooks */
137 static struct netisr_handler ip_nh = {
139 .nh_handler = ip_input,
140 .nh_proto = NETISR_IP,
142 .nh_m2cpuid = rss_soft_m2cpuid_v4,
143 .nh_policy = NETISR_POLICY_CPU,
144 .nh_dispatch = NETISR_DISPATCH_HYBRID,
146 .nh_policy = NETISR_POLICY_FLOW,
152 * Directly dispatched frames are currently assumed
153 * to have a flowid already calculated.
155 * It should likely have something that assert it
156 * actually has valid flow details.
158 static struct netisr_handler ip_direct_nh = {
159 .nh_name = "ip_direct",
160 .nh_handler = ip_direct_input,
161 .nh_proto = NETISR_IP_DIRECT,
162 .nh_m2cpuid = rss_soft_m2cpuid_v4,
163 .nh_policy = NETISR_POLICY_CPU,
164 .nh_dispatch = NETISR_DISPATCH_HYBRID,
168 extern struct domain inetdomain;
169 extern struct protosw inetsw[];
170 u_char ip_protox[IPPROTO_MAX];
171 VNET_DEFINE(struct in_ifaddrhead, in_ifaddrhead); /* first inet address */
172 VNET_DEFINE(struct in_ifaddrhashhead *, in_ifaddrhashtbl); /* inet addr hash table */
173 VNET_DEFINE(u_long, in_ifaddrhmask); /* mask for hash table */
176 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
177 &ip_mtu, 0, "Default MTU");
181 VNET_DEFINE(int, ipstealth);
182 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_VNET | CTLFLAG_RW,
183 &VNET_NAME(ipstealth), 0,
184 "IP stealth mode, no TTL decrementation on forwarding");
188 * IP statistics are stored in the "array" of counter(9)s.
190 VNET_PCPUSTAT_DEFINE(struct ipstat, ipstat);
191 VNET_PCPUSTAT_SYSINIT(ipstat);
192 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, IPCTL_STATS, stats, struct ipstat, ipstat,
193 "IP statistics (struct ipstat, netinet/ip_var.h)");
196 VNET_PCPUSTAT_SYSUNINIT(ipstat);
200 * Kernel module interface for updating ipstat. The argument is an index
201 * into ipstat treated as an array.
204 kmod_ipstat_inc(int statnum)
207 counter_u64_add(VNET(ipstat)[statnum], 1);
211 kmod_ipstat_dec(int statnum)
214 counter_u64_add(VNET(ipstat)[statnum], -1);
218 sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS)
222 netisr_getqlimit(&ip_nh, &qlimit);
223 error = sysctl_handle_int(oidp, &qlimit, 0, req);
224 if (error || !req->newptr)
228 return (netisr_setqlimit(&ip_nh, qlimit));
230 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen,
231 CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet_intr_queue_maxlen, "I",
232 "Maximum size of the IP input queue");
235 sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS)
237 u_int64_t qdrops_long;
240 netisr_getqdrops(&ip_nh, &qdrops_long);
241 qdrops = qdrops_long;
242 error = sysctl_handle_int(oidp, &qdrops, 0, req);
243 if (error || !req->newptr)
247 netisr_clearqdrops(&ip_nh);
251 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops,
252 CTLTYPE_INT|CTLFLAG_RD, 0, 0, sysctl_netinet_intr_queue_drops, "I",
253 "Number of packets dropped from the IP input queue");
257 sysctl_netinet_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS)
261 netisr_getqlimit(&ip_direct_nh, &qlimit);
262 error = sysctl_handle_int(oidp, &qlimit, 0, req);
263 if (error || !req->newptr)
267 return (netisr_setqlimit(&ip_direct_nh, qlimit));
269 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRDQMAXLEN, intr_direct_queue_maxlen,
270 CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet_intr_direct_queue_maxlen,
271 "I", "Maximum size of the IP direct input queue");
274 sysctl_netinet_intr_direct_queue_drops(SYSCTL_HANDLER_ARGS)
276 u_int64_t qdrops_long;
279 netisr_getqdrops(&ip_direct_nh, &qdrops_long);
280 qdrops = qdrops_long;
281 error = sysctl_handle_int(oidp, &qdrops, 0, req);
282 if (error || !req->newptr)
286 netisr_clearqdrops(&ip_direct_nh);
290 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRDQDROPS, intr_direct_queue_drops,
291 CTLTYPE_INT|CTLFLAG_RD, 0, 0, sysctl_netinet_intr_direct_queue_drops, "I",
292 "Number of packets dropped from the IP direct input queue");
296 * IP initialization: fill in IP protocol switch table.
297 * All protocols not implemented in kernel go to raw IP protocol handler.
305 TAILQ_INIT(&V_in_ifaddrhead);
306 V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask);
308 /* Initialize IP reassembly queue. */
311 /* Initialize packet filter hooks. */
312 V_inet_pfil_hook.ph_type = PFIL_TYPE_AF;
313 V_inet_pfil_hook.ph_af = AF_INET;
314 if ((i = pfil_head_register(&V_inet_pfil_hook)) != 0)
315 printf("%s: WARNING: unable to register pfil hook, "
316 "error %d\n", __func__, i);
318 if (hhook_head_register(HHOOK_TYPE_IPSEC_IN, AF_INET,
319 &V_ipsec_hhh_in[HHOOK_IPSEC_INET],
320 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
321 printf("%s: WARNING: unable to register input helper hook\n",
323 if (hhook_head_register(HHOOK_TYPE_IPSEC_OUT, AF_INET,
324 &V_ipsec_hhh_out[HHOOK_IPSEC_INET],
325 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
326 printf("%s: WARNING: unable to register output helper hook\n",
329 /* Skip initialization of globals for non-default instances. */
331 if (!IS_DEFAULT_VNET(curvnet)) {
332 netisr_register_vnet(&ip_nh);
334 netisr_register_vnet(&ip_direct_nh);
340 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
342 panic("ip_init: PF_INET not found");
344 /* Initialize the entire ip_protox[] array to IPPROTO_RAW. */
345 for (i = 0; i < IPPROTO_MAX; i++)
346 ip_protox[i] = pr - inetsw;
348 * Cycle through IP protocols and put them into the appropriate place
351 for (pr = inetdomain.dom_protosw;
352 pr < inetdomain.dom_protoswNPROTOSW; pr++)
353 if (pr->pr_domain->dom_family == PF_INET &&
354 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
355 /* Be careful to only index valid IP protocols. */
356 if (pr->pr_protocol < IPPROTO_MAX)
357 ip_protox[pr->pr_protocol] = pr - inetsw;
360 netisr_register(&ip_nh);
362 netisr_register(&ip_direct_nh);
368 ip_destroy(void *unused __unused)
374 netisr_unregister_vnet(&ip_direct_nh);
376 netisr_unregister_vnet(&ip_nh);
378 if ((error = pfil_head_unregister(&V_inet_pfil_hook)) != 0)
379 printf("%s: WARNING: unable to unregister pfil hook, "
380 "error %d\n", __func__, error);
382 error = hhook_head_deregister(V_ipsec_hhh_in[HHOOK_IPSEC_INET]);
384 printf("%s: WARNING: unable to deregister input helper hook "
385 "type HHOOK_TYPE_IPSEC_IN, id HHOOK_IPSEC_INET: "
386 "error %d returned\n", __func__, error);
388 error = hhook_head_deregister(V_ipsec_hhh_out[HHOOK_IPSEC_INET]);
390 printf("%s: WARNING: unable to deregister output helper hook "
391 "type HHOOK_TYPE_IPSEC_OUT, id HHOOK_IPSEC_INET: "
392 "error %d returned\n", __func__, error);
395 /* Remove the IPv4 addresses from all interfaces. */
398 /* Make sure the IPv4 routes are gone as well. */
400 TAILQ_FOREACH(ifp, &V_ifnet, if_link)
401 rt_flushifroutes_af(ifp, AF_INET);
404 /* Destroy IP reassembly queue. */
407 /* Cleanup in_ifaddr hash table; should be empty. */
408 hashdestroy(V_in_ifaddrhashtbl, M_IFADDR, V_in_ifaddrhmask);
411 VNET_SYSUNINIT(ip, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip_destroy, NULL);
416 * IP direct input routine.
418 * This is called when reinjecting completed fragments where
419 * all of the previous checking and book-keeping has been done.
422 ip_direct_input(struct mbuf *m)
427 ip = mtod(m, struct ip *);
428 hlen = ip->ip_hl << 2;
430 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
431 if (IPSEC_ENABLED(ipv4)) {
432 if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
436 IPSTAT_INC(ips_delivered);
437 (*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
443 * Ip input routine. Checksum and byte swap header. If fragmented
444 * try to reassemble. Process options. Pass to next level.
447 ip_input(struct mbuf *m)
449 struct rm_priotracker in_ifa_tracker;
450 struct ip *ip = NULL;
451 struct in_ifaddr *ia = NULL;
454 int checkif, hlen = 0;
455 uint16_t sum, ip_len;
456 int dchg = 0; /* dest changed after fw */
457 struct in_addr odst; /* original dst address */
461 if (m->m_flags & M_FASTFWD_OURS) {
462 m->m_flags &= ~M_FASTFWD_OURS;
463 /* Set up some basics that will be used later. */
464 ip = mtod(m, struct ip *);
465 hlen = ip->ip_hl << 2;
466 ip_len = ntohs(ip->ip_len);
470 IPSTAT_INC(ips_total);
472 if (m->m_pkthdr.len < sizeof(struct ip))
475 if (m->m_len < sizeof (struct ip) &&
476 (m = m_pullup(m, sizeof (struct ip))) == NULL) {
477 IPSTAT_INC(ips_toosmall);
480 ip = mtod(m, struct ip *);
482 if (ip->ip_v != IPVERSION) {
483 IPSTAT_INC(ips_badvers);
487 hlen = ip->ip_hl << 2;
488 if (hlen < sizeof(struct ip)) { /* minimum header length */
489 IPSTAT_INC(ips_badhlen);
492 if (hlen > m->m_len) {
493 if ((m = m_pullup(m, hlen)) == NULL) {
494 IPSTAT_INC(ips_badhlen);
497 ip = mtod(m, struct ip *);
500 IP_PROBE(receive, NULL, NULL, ip, m->m_pkthdr.rcvif, ip, NULL);
502 /* 127/8 must not appear on wire - RFC1122 */
503 ifp = m->m_pkthdr.rcvif;
504 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
505 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
506 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
507 IPSTAT_INC(ips_badaddr);
512 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
513 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
515 if (hlen == sizeof(struct ip)) {
516 sum = in_cksum_hdr(ip);
518 sum = in_cksum(m, hlen);
522 IPSTAT_INC(ips_badsum);
527 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
528 /* packet is dropped by traffic conditioner */
532 ip_len = ntohs(ip->ip_len);
534 IPSTAT_INC(ips_badlen);
539 * Check that the amount of data in the buffers
540 * is as at least much as the IP header would have us expect.
541 * Trim mbufs if longer than we expect.
542 * Drop packet if shorter than we expect.
544 if (m->m_pkthdr.len < ip_len) {
546 IPSTAT_INC(ips_tooshort);
549 if (m->m_pkthdr.len > ip_len) {
550 if (m->m_len == m->m_pkthdr.len) {
552 m->m_pkthdr.len = ip_len;
554 m_adj(m, ip_len - m->m_pkthdr.len);
558 * Try to forward the packet, but if we fail continue.
559 * ip_tryforward() does not generate redirects, so fall
560 * through to normal processing if redirects are required.
561 * ip_tryforward() does inbound and outbound packet firewall
562 * processing. If firewall has decided that destination becomes
563 * our local address, it sets M_FASTFWD_OURS flag. In this
564 * case skip another inbound firewall processing and update
567 if (V_ipforwarding != 0 && V_ipsendredirects == 0
568 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
569 && (!IPSEC_ENABLED(ipv4) ||
570 IPSEC_CAPS(ipv4, m, IPSEC_CAP_OPERABLE) == 0)
573 if ((m = ip_tryforward(m)) == NULL)
575 if (m->m_flags & M_FASTFWD_OURS) {
576 m->m_flags &= ~M_FASTFWD_OURS;
577 ip = mtod(m, struct ip *);
582 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
584 * Bypass packet filtering for packets previously handled by IPsec.
586 if (IPSEC_ENABLED(ipv4) &&
587 IPSEC_CAPS(ipv4, m, IPSEC_CAP_BYPASS_FILTER) != 0)
592 * Run through list of hooks for input packets.
594 * NB: Beware of the destination address changing (e.g.
595 * by NAT rewriting). When this happens, tell
596 * ip_forward to do the right thing.
599 /* Jump over all PFIL processing if hooks are not active. */
600 if (!PFIL_HOOKED(&V_inet_pfil_hook))
604 if (pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_IN, 0, NULL) != 0)
606 if (m == NULL) /* consumed by filter */
609 ip = mtod(m, struct ip *);
610 dchg = (odst.s_addr != ip->ip_dst.s_addr);
611 ifp = m->m_pkthdr.rcvif;
613 if (m->m_flags & M_FASTFWD_OURS) {
614 m->m_flags &= ~M_FASTFWD_OURS;
617 if (m->m_flags & M_IP_NEXTHOP) {
618 if (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) {
620 * Directly ship the packet on. This allows
621 * forwarding packets originally destined to us
622 * to some other directly connected host.
631 * Process options and, if not destined for us,
632 * ship it on. ip_dooptions returns 1 when an
633 * error was detected (causing an icmp message
634 * to be sent and the original packet to be freed).
636 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
639 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
640 * matter if it is destined to another node, or whether it is
641 * a multicast one, RSVP wants it! and prevents it from being forwarded
642 * anywhere else. Also checks if the rsvp daemon is running before
643 * grabbing the packet.
645 if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP)
649 * Check our list of addresses, to see if the packet is for us.
650 * If we don't have any addresses, assume any unicast packet
651 * we receive might be for us (and let the upper layers deal
654 if (TAILQ_EMPTY(&V_in_ifaddrhead) &&
655 (m->m_flags & (M_MCAST|M_BCAST)) == 0)
659 * Enable a consistency check between the destination address
660 * and the arrival interface for a unicast packet (the RFC 1122
661 * strong ES model) if IP forwarding is disabled and the packet
662 * is not locally generated and the packet is not subject to
665 * XXX - Checking also should be disabled if the destination
666 * address is ipnat'ed to a different interface.
668 * XXX - Checking is incompatible with IP aliases added
669 * to the loopback interface instead of the interface where
670 * the packets are received.
672 * XXX - This is the case for carp vhost IPs as well so we
673 * insert a workaround. If the packet got here, we already
674 * checked with carp_iamatch() and carp_forus().
676 checkif = V_ip_checkinterface && (V_ipforwarding == 0) &&
677 ifp != NULL && ((ifp->if_flags & IFF_LOOPBACK) == 0) &&
678 ifp->if_carp == NULL && (dchg == 0);
681 * Check for exact addresses in the hash bucket.
683 IN_IFADDR_RLOCK(&in_ifa_tracker);
684 LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
686 * If the address matches, verify that the packet
687 * arrived via the correct interface if checking is
690 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr &&
691 (!checkif || ia->ia_ifp == ifp)) {
692 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
693 counter_u64_add(ia->ia_ifa.ifa_ibytes,
695 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
699 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
702 * Check for broadcast addresses.
704 * Only accept broadcast packets that arrive via the matching
705 * interface. Reception of forwarded directed broadcasts would
706 * be handled via ip_forward() and ether_output() with the loopback
707 * into the stack for SIMPLEX interfaces handled by ether_output().
709 if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
711 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
712 if (ifa->ifa_addr->sa_family != AF_INET)
715 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
717 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
718 counter_u64_add(ia->ia_ifa.ifa_ibytes,
720 IF_ADDR_RUNLOCK(ifp);
724 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) {
725 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
726 counter_u64_add(ia->ia_ifa.ifa_ibytes,
728 IF_ADDR_RUNLOCK(ifp);
733 IF_ADDR_RUNLOCK(ifp);
736 /* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */
737 if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) {
738 IPSTAT_INC(ips_cantforward);
742 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
745 * If we are acting as a multicast router, all
746 * incoming multicast packets are passed to the
747 * kernel-level multicast forwarding function.
748 * The packet is returned (relatively) intact; if
749 * ip_mforward() returns a non-zero value, the packet
750 * must be discarded, else it may be accepted below.
752 if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) {
753 IPSTAT_INC(ips_cantforward);
759 * The process-level routing daemon needs to receive
760 * all multicast IGMP packets, whether or not this
761 * host belongs to their destination groups.
763 if (ip->ip_p == IPPROTO_IGMP)
765 IPSTAT_INC(ips_forward);
768 * Assume the packet is for us, to avoid prematurely taking
769 * a lock on the in_multi hash. Protocols must perform
770 * their own filtering and update statistics accordingly.
774 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
776 if (ip->ip_dst.s_addr == INADDR_ANY)
780 * Not for us; forward if possible and desirable.
782 if (V_ipforwarding == 0) {
783 IPSTAT_INC(ips_cantforward);
793 * IPSTEALTH: Process non-routing options only
794 * if the packet is destined for us.
796 if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1))
798 #endif /* IPSTEALTH */
801 * Attempt reassembly; if it succeeds, proceed.
802 * ip_reass() will return a different mbuf.
804 if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) {
805 /* XXXGL: shouldn't we save & set m_flags? */
809 ip = mtod(m, struct ip *);
810 /* Get the header length of the reassembled packet */
811 hlen = ip->ip_hl << 2;
814 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
815 if (IPSEC_ENABLED(ipv4)) {
816 if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
822 * Switch out to protocol's input routine.
824 IPSTAT_INC(ips_delivered);
826 (*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
833 * IP timer processing;
834 * if a timer expires on a reassembly
840 VNET_ITERATOR_DECL(vnet_iter);
842 VNET_LIST_RLOCK_NOSLEEP();
843 VNET_FOREACH(vnet_iter) {
844 CURVNET_SET(vnet_iter);
848 VNET_LIST_RUNLOCK_NOSLEEP();
854 VNET_ITERATOR_DECL(vnet_iter);
856 VNET_LIST_RLOCK_NOSLEEP();
857 VNET_FOREACH(vnet_iter) {
858 CURVNET_SET(vnet_iter);
862 VNET_LIST_RUNLOCK_NOSLEEP();
866 * The protocol to be inserted into ip_protox[] must be already registered
867 * in inetsw[], either statically or through pf_proto_register().
870 ipproto_register(short ipproto)
875 if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
876 return (EPROTONOSUPPORT);
879 * The protocol slot must not be occupied by another protocol
880 * already. An index pointing to IPPROTO_RAW is unused.
882 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
884 return (EPFNOSUPPORT);
885 if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */
888 /* Find the protocol position in inetsw[] and set the index. */
889 for (pr = inetdomain.dom_protosw;
890 pr < inetdomain.dom_protoswNPROTOSW; pr++) {
891 if (pr->pr_domain->dom_family == PF_INET &&
892 pr->pr_protocol && pr->pr_protocol == ipproto) {
893 ip_protox[pr->pr_protocol] = pr - inetsw;
897 return (EPROTONOSUPPORT);
901 ipproto_unregister(short ipproto)
906 if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
907 return (EPROTONOSUPPORT);
909 /* Check if the protocol was indeed registered. */
910 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
912 return (EPFNOSUPPORT);
913 if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */
916 /* Reset the protocol slot to IPPROTO_RAW. */
917 ip_protox[ipproto] = pr - inetsw;
921 u_char inetctlerrmap[PRC_NCMDS] = {
923 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
924 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
925 EMSGSIZE, EHOSTUNREACH, 0, 0,
926 0, 0, EHOSTUNREACH, 0,
927 ENOPROTOOPT, ECONNREFUSED
931 * Forward a packet. If some error occurs return the sender
932 * an icmp packet. Note we can't always generate a meaningful
933 * icmp message because icmp doesn't have a large enough repertoire
934 * of codes and types.
936 * If not forwarding, just drop the packet. This could be confusing
937 * if ipforwarding was zero but some routing protocol was advancing
938 * us as a gateway to somewhere. However, we must let the routing
939 * protocol deal with that.
941 * The srcrt parameter indicates whether the packet is being forwarded
942 * via a source route.
945 ip_forward(struct mbuf *m, int srcrt)
947 struct ip *ip = mtod(m, struct ip *);
948 struct in_ifaddr *ia;
950 struct sockaddr_in *sin;
953 int error, type = 0, code = 0, mtu = 0;
955 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
956 IPSTAT_INC(ips_cantforward);
964 ip->ip_ttl <= IPTTLDEC) {
965 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 0, 0);
969 bzero(&ro, sizeof(ro));
970 sin = (struct sockaddr_in *)&ro.ro_dst;
971 sin->sin_family = AF_INET;
972 sin->sin_len = sizeof(*sin);
973 sin->sin_addr = ip->ip_dst;
975 rtalloc_mpath_fib(&ro,
976 ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
979 in_rtalloc_ign(&ro, 0, M_GETFIB(m));
981 if (ro.ro_rt != NULL) {
982 ia = ifatoia(ro.ro_rt->rt_ifa);
983 ifa_ref(&ia->ia_ifa);
987 * Save the IP header and at most 8 bytes of the payload,
988 * in case we need to generate an ICMP message to the src.
990 * XXX this can be optimized a lot by saving the data in a local
991 * buffer on the stack (72 bytes at most), and only allocating the
992 * mbuf if really necessary. The vast majority of the packets
993 * are forwarded without having to send an ICMP back (either
994 * because unnecessary, or because rate limited), so we are
995 * really we are wasting a lot of work here.
997 * We don't use m_copy() because it might return a reference
998 * to a shared cluster. Both this function and ip_output()
999 * assume exclusive access to the IP header in `m', so any
1000 * data in a cluster may change before we reach icmp_error().
1002 mcopy = m_gethdr(M_NOWAIT, m->m_type);
1003 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_NOWAIT)) {
1005 * It's probably ok if the pkthdr dup fails (because
1006 * the deep copy of the tag chain failed), but for now
1007 * be conservative and just discard the copy since
1008 * code below may some day want the tags.
1013 if (mcopy != NULL) {
1014 mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy));
1015 mcopy->m_pkthdr.len = mcopy->m_len;
1016 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1019 if (V_ipstealth == 0)
1021 ip->ip_ttl -= IPTTLDEC;
1022 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1023 if (IPSEC_ENABLED(ipv4)) {
1024 if ((error = IPSEC_FORWARD(ipv4, m)) != 0) {
1025 /* mbuf consumed by IPsec */
1027 if (error != EINPROGRESS)
1028 IPSTAT_INC(ips_cantforward);
1031 /* No IPsec processing required */
1035 * If forwarding packet using same interface that it came in on,
1036 * perhaps should send a redirect to sender to shortcut a hop.
1037 * Only send redirect if source is sending directly to us,
1038 * and if packet was not source routed (or has any options).
1039 * Also, don't send redirect if forwarding using a default route
1040 * or a route modified by a redirect.
1043 if (!srcrt && V_ipsendredirects &&
1044 ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) {
1049 if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1050 satosin(rt_key(rt))->sin_addr.s_addr != 0) {
1051 #define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa))
1052 u_long src = ntohl(ip->ip_src.s_addr);
1055 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1056 if (rt->rt_flags & RTF_GATEWAY)
1057 dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr;
1059 dest.s_addr = ip->ip_dst.s_addr;
1060 /* Router requirements says to only send host redirects */
1061 type = ICMP_REDIRECT;
1062 code = ICMP_REDIRECT_HOST;
1067 error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
1069 if (error == EMSGSIZE && ro.ro_rt)
1070 mtu = ro.ro_rt->rt_mtu;
1074 IPSTAT_INC(ips_cantforward);
1076 IPSTAT_INC(ips_forward);
1078 IPSTAT_INC(ips_redirectsent);
1083 ifa_free(&ia->ia_ifa);
1087 if (mcopy == NULL) {
1089 ifa_free(&ia->ia_ifa);
1095 case 0: /* forwarded, but need redirect */
1096 /* type, code set above */
1104 type = ICMP_UNREACH;
1105 code = ICMP_UNREACH_HOST;
1109 type = ICMP_UNREACH;
1110 code = ICMP_UNREACH_NEEDFRAG;
1112 * If the MTU was set before make sure we are below the
1114 * If the MTU wasn't set before use the interface mtu or
1115 * fall back to the next smaller mtu step compared to the
1116 * current packet size.
1120 mtu = min(mtu, ia->ia_ifp->if_mtu);
1123 mtu = ia->ia_ifp->if_mtu;
1125 mtu = ip_next_mtu(ntohs(ip->ip_len), 0);
1127 IPSTAT_INC(ips_cantfrag);
1131 case EACCES: /* ipfw denied packet */
1134 ifa_free(&ia->ia_ifa);
1138 ifa_free(&ia->ia_ifa);
1139 icmp_error(mcopy, type, code, dest.s_addr, mtu);
1142 #define CHECK_SO_CT(sp, ct) \
1143 (((sp->so_options & SO_TIMESTAMP) && (sp->so_ts_clock == ct)) ? 1 : 0)
1146 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1150 if ((inp->inp_socket->so_options & SO_BINTIME) ||
1151 CHECK_SO_CT(inp->inp_socket, SO_TS_BINTIME)) {
1155 *mp = sbcreatecontrol((caddr_t)&bt, sizeof(bt),
1156 SCM_BINTIME, SOL_SOCKET);
1158 mp = &(*mp)->m_next;
1160 if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME_MICRO)) {
1164 *mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv),
1165 SCM_TIMESTAMP, SOL_SOCKET);
1167 mp = &(*mp)->m_next;
1168 } else if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME)) {
1172 *mp = sbcreatecontrol((caddr_t)&ts, sizeof(ts),
1173 SCM_REALTIME, SOL_SOCKET);
1175 mp = &(*mp)->m_next;
1176 } else if (CHECK_SO_CT(inp->inp_socket, SO_TS_MONOTONIC)) {
1180 *mp = sbcreatecontrol((caddr_t)&ts, sizeof(ts),
1181 SCM_MONOTONIC, SOL_SOCKET);
1183 mp = &(*mp)->m_next;
1185 if (inp->inp_flags & INP_RECVDSTADDR) {
1186 *mp = sbcreatecontrol((caddr_t)&ip->ip_dst,
1187 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1189 mp = &(*mp)->m_next;
1191 if (inp->inp_flags & INP_RECVTTL) {
1192 *mp = sbcreatecontrol((caddr_t)&ip->ip_ttl,
1193 sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
1195 mp = &(*mp)->m_next;
1199 * Moving these out of udp_input() made them even more broken
1200 * than they already were.
1202 /* options were tossed already */
1203 if (inp->inp_flags & INP_RECVOPTS) {
1204 *mp = sbcreatecontrol((caddr_t)opts_deleted_above,
1205 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1207 mp = &(*mp)->m_next;
1209 /* ip_srcroute doesn't do what we want here, need to fix */
1210 if (inp->inp_flags & INP_RECVRETOPTS) {
1211 *mp = sbcreatecontrol((caddr_t)ip_srcroute(m),
1212 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1214 mp = &(*mp)->m_next;
1217 if (inp->inp_flags & INP_RECVIF) {
1220 struct sockaddr_dl sdl;
1223 struct sockaddr_dl *sdp;
1224 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1226 if ((ifp = m->m_pkthdr.rcvif) &&
1227 ifp->if_index && ifp->if_index <= V_if_index) {
1228 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1230 * Change our mind and don't try copy.
1232 if (sdp->sdl_family != AF_LINK ||
1233 sdp->sdl_len > sizeof(sdlbuf)) {
1236 bcopy(sdp, sdl2, sdp->sdl_len);
1240 offsetof(struct sockaddr_dl, sdl_data[0]);
1241 sdl2->sdl_family = AF_LINK;
1242 sdl2->sdl_index = 0;
1243 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1245 *mp = sbcreatecontrol((caddr_t)sdl2, sdl2->sdl_len,
1246 IP_RECVIF, IPPROTO_IP);
1248 mp = &(*mp)->m_next;
1250 if (inp->inp_flags & INP_RECVTOS) {
1251 *mp = sbcreatecontrol((caddr_t)&ip->ip_tos,
1252 sizeof(u_char), IP_RECVTOS, IPPROTO_IP);
1254 mp = &(*mp)->m_next;
1257 if (inp->inp_flags2 & INP_RECVFLOWID) {
1258 uint32_t flowid, flow_type;
1260 flowid = m->m_pkthdr.flowid;
1261 flow_type = M_HASHTYPE_GET(m);
1264 * XXX should handle the failure of one or the
1265 * other - don't populate both?
1267 *mp = sbcreatecontrol((caddr_t) &flowid,
1268 sizeof(uint32_t), IP_FLOWID, IPPROTO_IP);
1270 mp = &(*mp)->m_next;
1271 *mp = sbcreatecontrol((caddr_t) &flow_type,
1272 sizeof(uint32_t), IP_FLOWTYPE, IPPROTO_IP);
1274 mp = &(*mp)->m_next;
1278 if (inp->inp_flags2 & INP_RECVRSSBUCKETID) {
1279 uint32_t flowid, flow_type;
1280 uint32_t rss_bucketid;
1282 flowid = m->m_pkthdr.flowid;
1283 flow_type = M_HASHTYPE_GET(m);
1285 if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) {
1286 *mp = sbcreatecontrol((caddr_t) &rss_bucketid,
1287 sizeof(uint32_t), IP_RSSBUCKETID, IPPROTO_IP);
1289 mp = &(*mp)->m_next;
1296 * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
1297 * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
1298 * locking. This code remains in ip_input.c as ip_mroute.c is optionally
1301 static VNET_DEFINE(int, ip_rsvp_on);
1302 VNET_DEFINE(struct socket *, ip_rsvpd);
1304 #define V_ip_rsvp_on VNET(ip_rsvp_on)
1307 ip_rsvp_init(struct socket *so)
1310 if (so->so_type != SOCK_RAW ||
1311 so->so_proto->pr_protocol != IPPROTO_RSVP)
1314 if (V_ip_rsvpd != NULL)
1319 * This may seem silly, but we need to be sure we don't over-increment
1320 * the RSVP counter, in case something slips up.
1322 if (!V_ip_rsvp_on) {
1336 * This may seem silly, but we need to be sure we don't over-decrement
1337 * the RSVP counter, in case something slips up.
1347 rsvp_input(struct mbuf **mp, int *offp, int proto)
1354 if (rsvp_input_p) { /* call the real one if loaded */
1356 rsvp_input_p(mp, offp, proto);
1357 return (IPPROTO_DONE);
1360 /* Can still get packets with rsvp_on = 0 if there is a local member
1361 * of the group to which the RSVP packet is addressed. But in this
1362 * case we want to throw the packet away.
1367 return (IPPROTO_DONE);
1370 if (V_ip_rsvpd != NULL) {
1372 rip_input(mp, offp, proto);
1373 return (IPPROTO_DONE);
1375 /* Drop the packet */
1377 return (IPPROTO_DONE);