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_INTRQMAXLEN, intr_direct_queue_maxlen,
270 CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet_intr_direct_queue_maxlen, "I",
271 "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_INTRQDROPS, 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 ip *ip = NULL;
450 struct in_ifaddr *ia = NULL;
453 int checkif, hlen = 0;
454 uint16_t sum, ip_len;
455 int dchg = 0; /* dest changed after fw */
456 struct in_addr odst; /* original dst address */
460 if (m->m_flags & M_FASTFWD_OURS) {
461 m->m_flags &= ~M_FASTFWD_OURS;
462 /* Set up some basics that will be used later. */
463 ip = mtod(m, struct ip *);
464 hlen = ip->ip_hl << 2;
465 ip_len = ntohs(ip->ip_len);
469 IPSTAT_INC(ips_total);
471 if (m->m_pkthdr.len < sizeof(struct ip))
474 if (m->m_len < sizeof (struct ip) &&
475 (m = m_pullup(m, sizeof (struct ip))) == NULL) {
476 IPSTAT_INC(ips_toosmall);
479 ip = mtod(m, struct ip *);
481 if (ip->ip_v != IPVERSION) {
482 IPSTAT_INC(ips_badvers);
486 hlen = ip->ip_hl << 2;
487 if (hlen < sizeof(struct ip)) { /* minimum header length */
488 IPSTAT_INC(ips_badhlen);
491 if (hlen > m->m_len) {
492 if ((m = m_pullup(m, hlen)) == NULL) {
493 IPSTAT_INC(ips_badhlen);
496 ip = mtod(m, struct ip *);
499 IP_PROBE(receive, NULL, NULL, ip, m->m_pkthdr.rcvif, ip, NULL);
501 /* 127/8 must not appear on wire - RFC1122 */
502 ifp = m->m_pkthdr.rcvif;
503 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
504 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
505 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
506 IPSTAT_INC(ips_badaddr);
511 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
512 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
514 if (hlen == sizeof(struct ip)) {
515 sum = in_cksum_hdr(ip);
517 sum = in_cksum(m, hlen);
521 IPSTAT_INC(ips_badsum);
526 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
527 /* packet is dropped by traffic conditioner */
531 ip_len = ntohs(ip->ip_len);
533 IPSTAT_INC(ips_badlen);
538 * Check that the amount of data in the buffers
539 * is as at least much as the IP header would have us expect.
540 * Trim mbufs if longer than we expect.
541 * Drop packet if shorter than we expect.
543 if (m->m_pkthdr.len < ip_len) {
545 IPSTAT_INC(ips_tooshort);
548 if (m->m_pkthdr.len > ip_len) {
549 if (m->m_len == m->m_pkthdr.len) {
551 m->m_pkthdr.len = ip_len;
553 m_adj(m, ip_len - m->m_pkthdr.len);
557 * Try to forward the packet, but if we fail continue.
558 * ip_tryforward() does inbound and outbound packet firewall
559 * processing. If firewall has decided that destination becomes
560 * our local address, it sets M_FASTFWD_OURS flag. In this
561 * case skip another inbound firewall processing and update
564 if (V_ipforwarding != 0
565 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
566 && (!IPSEC_ENABLED(ipv4) ||
567 IPSEC_CAPS(ipv4, m, IPSEC_CAP_OPERABLE) == 0)
570 if ((m = ip_tryforward(m)) == NULL)
572 if (m->m_flags & M_FASTFWD_OURS) {
573 m->m_flags &= ~M_FASTFWD_OURS;
574 ip = mtod(m, struct ip *);
579 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
581 * Bypass packet filtering for packets previously handled by IPsec.
583 if (IPSEC_ENABLED(ipv4) &&
584 IPSEC_CAPS(ipv4, m, IPSEC_CAP_BYPASS_FILTER) != 0)
589 * Run through list of hooks for input packets.
591 * NB: Beware of the destination address changing (e.g.
592 * by NAT rewriting). When this happens, tell
593 * ip_forward to do the right thing.
596 /* Jump over all PFIL processing if hooks are not active. */
597 if (!PFIL_HOOKED(&V_inet_pfil_hook))
601 if (pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_IN, NULL) != 0)
603 if (m == NULL) /* consumed by filter */
606 ip = mtod(m, struct ip *);
607 dchg = (odst.s_addr != ip->ip_dst.s_addr);
608 ifp = m->m_pkthdr.rcvif;
610 if (m->m_flags & M_FASTFWD_OURS) {
611 m->m_flags &= ~M_FASTFWD_OURS;
614 if (m->m_flags & M_IP_NEXTHOP) {
615 if (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) {
617 * Directly ship the packet on. This allows
618 * forwarding packets originally destined to us
619 * to some other directly connected host.
628 * Process options and, if not destined for us,
629 * ship it on. ip_dooptions returns 1 when an
630 * error was detected (causing an icmp message
631 * to be sent and the original packet to be freed).
633 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
636 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
637 * matter if it is destined to another node, or whether it is
638 * a multicast one, RSVP wants it! and prevents it from being forwarded
639 * anywhere else. Also checks if the rsvp daemon is running before
640 * grabbing the packet.
642 if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP)
646 * Check our list of addresses, to see if the packet is for us.
647 * If we don't have any addresses, assume any unicast packet
648 * we receive might be for us (and let the upper layers deal
651 if (TAILQ_EMPTY(&V_in_ifaddrhead) &&
652 (m->m_flags & (M_MCAST|M_BCAST)) == 0)
656 * Enable a consistency check between the destination address
657 * and the arrival interface for a unicast packet (the RFC 1122
658 * strong ES model) if IP forwarding is disabled and the packet
659 * is not locally generated and the packet is not subject to
662 * XXX - Checking also should be disabled if the destination
663 * address is ipnat'ed to a different interface.
665 * XXX - Checking is incompatible with IP aliases added
666 * to the loopback interface instead of the interface where
667 * the packets are received.
669 * XXX - This is the case for carp vhost IPs as well so we
670 * insert a workaround. If the packet got here, we already
671 * checked with carp_iamatch() and carp_forus().
673 checkif = V_ip_checkinterface && (V_ipforwarding == 0) &&
674 ifp != NULL && ((ifp->if_flags & IFF_LOOPBACK) == 0) &&
675 ifp->if_carp == NULL && (dchg == 0);
678 * Check for exact addresses in the hash bucket.
680 /* IN_IFADDR_RLOCK(); */
681 LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
683 * If the address matches, verify that the packet
684 * arrived via the correct interface if checking is
687 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr &&
688 (!checkif || ia->ia_ifp == ifp)) {
689 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
690 counter_u64_add(ia->ia_ifa.ifa_ibytes,
692 /* IN_IFADDR_RUNLOCK(); */
696 /* IN_IFADDR_RUNLOCK(); */
699 * Check for broadcast addresses.
701 * Only accept broadcast packets that arrive via the matching
702 * interface. Reception of forwarded directed broadcasts would
703 * be handled via ip_forward() and ether_output() with the loopback
704 * into the stack for SIMPLEX interfaces handled by ether_output().
706 if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
708 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
709 if (ifa->ifa_addr->sa_family != AF_INET)
712 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
714 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
715 counter_u64_add(ia->ia_ifa.ifa_ibytes,
717 IF_ADDR_RUNLOCK(ifp);
721 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) {
722 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
723 counter_u64_add(ia->ia_ifa.ifa_ibytes,
725 IF_ADDR_RUNLOCK(ifp);
730 IF_ADDR_RUNLOCK(ifp);
733 /* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */
734 if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) {
735 IPSTAT_INC(ips_cantforward);
739 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
742 * If we are acting as a multicast router, all
743 * incoming multicast packets are passed to the
744 * kernel-level multicast forwarding function.
745 * The packet is returned (relatively) intact; if
746 * ip_mforward() returns a non-zero value, the packet
747 * must be discarded, else it may be accepted below.
749 if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) {
750 IPSTAT_INC(ips_cantforward);
756 * The process-level routing daemon needs to receive
757 * all multicast IGMP packets, whether or not this
758 * host belongs to their destination groups.
760 if (ip->ip_p == IPPROTO_IGMP)
762 IPSTAT_INC(ips_forward);
765 * Assume the packet is for us, to avoid prematurely taking
766 * a lock on the in_multi hash. Protocols must perform
767 * their own filtering and update statistics accordingly.
771 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
773 if (ip->ip_dst.s_addr == INADDR_ANY)
777 * Not for us; forward if possible and desirable.
779 if (V_ipforwarding == 0) {
780 IPSTAT_INC(ips_cantforward);
790 * IPSTEALTH: Process non-routing options only
791 * if the packet is destined for us.
793 if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1))
795 #endif /* IPSTEALTH */
798 * Attempt reassembly; if it succeeds, proceed.
799 * ip_reass() will return a different mbuf.
801 if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) {
802 /* XXXGL: shouldn't we save & set m_flags? */
806 ip = mtod(m, struct ip *);
807 /* Get the header length of the reassembled packet */
808 hlen = ip->ip_hl << 2;
811 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
812 if (IPSEC_ENABLED(ipv4)) {
813 if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
819 * Switch out to protocol's input routine.
821 IPSTAT_INC(ips_delivered);
823 (*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
830 * IP timer processing;
831 * if a timer expires on a reassembly
837 VNET_ITERATOR_DECL(vnet_iter);
839 VNET_LIST_RLOCK_NOSLEEP();
840 VNET_FOREACH(vnet_iter) {
841 CURVNET_SET(vnet_iter);
845 VNET_LIST_RUNLOCK_NOSLEEP();
851 VNET_ITERATOR_DECL(vnet_iter);
853 VNET_LIST_RLOCK_NOSLEEP();
854 VNET_FOREACH(vnet_iter) {
855 CURVNET_SET(vnet_iter);
859 VNET_LIST_RUNLOCK_NOSLEEP();
863 * The protocol to be inserted into ip_protox[] must be already registered
864 * in inetsw[], either statically or through pf_proto_register().
867 ipproto_register(short ipproto)
872 if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
873 return (EPROTONOSUPPORT);
876 * The protocol slot must not be occupied by another protocol
877 * already. An index pointing to IPPROTO_RAW is unused.
879 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
881 return (EPFNOSUPPORT);
882 if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */
885 /* Find the protocol position in inetsw[] and set the index. */
886 for (pr = inetdomain.dom_protosw;
887 pr < inetdomain.dom_protoswNPROTOSW; pr++) {
888 if (pr->pr_domain->dom_family == PF_INET &&
889 pr->pr_protocol && pr->pr_protocol == ipproto) {
890 ip_protox[pr->pr_protocol] = pr - inetsw;
894 return (EPROTONOSUPPORT);
898 ipproto_unregister(short ipproto)
903 if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
904 return (EPROTONOSUPPORT);
906 /* Check if the protocol was indeed registered. */
907 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
909 return (EPFNOSUPPORT);
910 if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */
913 /* Reset the protocol slot to IPPROTO_RAW. */
914 ip_protox[ipproto] = pr - inetsw;
918 u_char inetctlerrmap[PRC_NCMDS] = {
920 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
921 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
922 EMSGSIZE, EHOSTUNREACH, 0, 0,
923 0, 0, EHOSTUNREACH, 0,
924 ENOPROTOOPT, ECONNREFUSED
928 * Forward a packet. If some error occurs return the sender
929 * an icmp packet. Note we can't always generate a meaningful
930 * icmp message because icmp doesn't have a large enough repertoire
931 * of codes and types.
933 * If not forwarding, just drop the packet. This could be confusing
934 * if ipforwarding was zero but some routing protocol was advancing
935 * us as a gateway to somewhere. However, we must let the routing
936 * protocol deal with that.
938 * The srcrt parameter indicates whether the packet is being forwarded
939 * via a source route.
942 ip_forward(struct mbuf *m, int srcrt)
944 struct ip *ip = mtod(m, struct ip *);
945 struct in_ifaddr *ia;
947 struct sockaddr_in *sin;
950 int error, type = 0, code = 0, mtu = 0;
952 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
953 IPSTAT_INC(ips_cantforward);
961 ip->ip_ttl <= IPTTLDEC) {
962 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 0, 0);
966 bzero(&ro, sizeof(ro));
967 sin = (struct sockaddr_in *)&ro.ro_dst;
968 sin->sin_family = AF_INET;
969 sin->sin_len = sizeof(*sin);
970 sin->sin_addr = ip->ip_dst;
972 rtalloc_mpath_fib(&ro,
973 ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
976 in_rtalloc_ign(&ro, 0, M_GETFIB(m));
978 if (ro.ro_rt != NULL) {
979 ia = ifatoia(ro.ro_rt->rt_ifa);
980 ifa_ref(&ia->ia_ifa);
984 * Save the IP header and at most 8 bytes of the payload,
985 * in case we need to generate an ICMP message to the src.
987 * XXX this can be optimized a lot by saving the data in a local
988 * buffer on the stack (72 bytes at most), and only allocating the
989 * mbuf if really necessary. The vast majority of the packets
990 * are forwarded without having to send an ICMP back (either
991 * because unnecessary, or because rate limited), so we are
992 * really we are wasting a lot of work here.
994 * We don't use m_copy() because it might return a reference
995 * to a shared cluster. Both this function and ip_output()
996 * assume exclusive access to the IP header in `m', so any
997 * data in a cluster may change before we reach icmp_error().
999 mcopy = m_gethdr(M_NOWAIT, m->m_type);
1000 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_NOWAIT)) {
1002 * It's probably ok if the pkthdr dup fails (because
1003 * the deep copy of the tag chain failed), but for now
1004 * be conservative and just discard the copy since
1005 * code below may some day want the tags.
1010 if (mcopy != NULL) {
1011 mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy));
1012 mcopy->m_pkthdr.len = mcopy->m_len;
1013 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1016 if (V_ipstealth == 0)
1018 ip->ip_ttl -= IPTTLDEC;
1019 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1020 if (IPSEC_ENABLED(ipv4)) {
1021 if ((error = IPSEC_FORWARD(ipv4, m)) != 0) {
1022 /* mbuf consumed by IPsec */
1024 if (error != EINPROGRESS)
1025 IPSTAT_INC(ips_cantforward);
1028 /* No IPsec processing required */
1032 * If forwarding packet using same interface that it came in on,
1033 * perhaps should send a redirect to sender to shortcut a hop.
1034 * Only send redirect if source is sending directly to us,
1035 * and if packet was not source routed (or has any options).
1036 * Also, don't send redirect if forwarding using a default route
1037 * or a route modified by a redirect.
1040 if (!srcrt && V_ipsendredirects &&
1041 ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) {
1046 if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1047 satosin(rt_key(rt))->sin_addr.s_addr != 0) {
1048 #define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa))
1049 u_long src = ntohl(ip->ip_src.s_addr);
1052 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1053 if (rt->rt_flags & RTF_GATEWAY)
1054 dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr;
1056 dest.s_addr = ip->ip_dst.s_addr;
1057 /* Router requirements says to only send host redirects */
1058 type = ICMP_REDIRECT;
1059 code = ICMP_REDIRECT_HOST;
1064 error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
1066 if (error == EMSGSIZE && ro.ro_rt)
1067 mtu = ro.ro_rt->rt_mtu;
1071 IPSTAT_INC(ips_cantforward);
1073 IPSTAT_INC(ips_forward);
1075 IPSTAT_INC(ips_redirectsent);
1080 ifa_free(&ia->ia_ifa);
1084 if (mcopy == NULL) {
1086 ifa_free(&ia->ia_ifa);
1092 case 0: /* forwarded, but need redirect */
1093 /* type, code set above */
1101 type = ICMP_UNREACH;
1102 code = ICMP_UNREACH_HOST;
1106 type = ICMP_UNREACH;
1107 code = ICMP_UNREACH_NEEDFRAG;
1109 * If the MTU was set before make sure we are below the
1111 * If the MTU wasn't set before use the interface mtu or
1112 * fall back to the next smaller mtu step compared to the
1113 * current packet size.
1117 mtu = min(mtu, ia->ia_ifp->if_mtu);
1120 mtu = ia->ia_ifp->if_mtu;
1122 mtu = ip_next_mtu(ntohs(ip->ip_len), 0);
1124 IPSTAT_INC(ips_cantfrag);
1128 case EACCES: /* ipfw denied packet */
1131 ifa_free(&ia->ia_ifa);
1135 ifa_free(&ia->ia_ifa);
1136 icmp_error(mcopy, type, code, dest.s_addr, mtu);
1140 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1144 if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) {
1148 if (inp->inp_socket->so_options & SO_BINTIME) {
1149 *mp = sbcreatecontrol((caddr_t)&bt, sizeof(bt),
1150 SCM_BINTIME, SOL_SOCKET);
1152 mp = &(*mp)->m_next;
1154 if (inp->inp_socket->so_options & SO_TIMESTAMP) {
1157 bintime2timeval(&bt, &tv);
1158 *mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv),
1159 SCM_TIMESTAMP, SOL_SOCKET);
1161 mp = &(*mp)->m_next;
1164 if (inp->inp_flags & INP_RECVDSTADDR) {
1165 *mp = sbcreatecontrol((caddr_t)&ip->ip_dst,
1166 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1168 mp = &(*mp)->m_next;
1170 if (inp->inp_flags & INP_RECVTTL) {
1171 *mp = sbcreatecontrol((caddr_t)&ip->ip_ttl,
1172 sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
1174 mp = &(*mp)->m_next;
1178 * Moving these out of udp_input() made them even more broken
1179 * than they already were.
1181 /* options were tossed already */
1182 if (inp->inp_flags & INP_RECVOPTS) {
1183 *mp = sbcreatecontrol((caddr_t)opts_deleted_above,
1184 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1186 mp = &(*mp)->m_next;
1188 /* ip_srcroute doesn't do what we want here, need to fix */
1189 if (inp->inp_flags & INP_RECVRETOPTS) {
1190 *mp = sbcreatecontrol((caddr_t)ip_srcroute(m),
1191 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1193 mp = &(*mp)->m_next;
1196 if (inp->inp_flags & INP_RECVIF) {
1199 struct sockaddr_dl sdl;
1202 struct sockaddr_dl *sdp;
1203 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1205 if ((ifp = m->m_pkthdr.rcvif) &&
1206 ifp->if_index && ifp->if_index <= V_if_index) {
1207 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1209 * Change our mind and don't try copy.
1211 if (sdp->sdl_family != AF_LINK ||
1212 sdp->sdl_len > sizeof(sdlbuf)) {
1215 bcopy(sdp, sdl2, sdp->sdl_len);
1219 offsetof(struct sockaddr_dl, sdl_data[0]);
1220 sdl2->sdl_family = AF_LINK;
1221 sdl2->sdl_index = 0;
1222 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1224 *mp = sbcreatecontrol((caddr_t)sdl2, sdl2->sdl_len,
1225 IP_RECVIF, IPPROTO_IP);
1227 mp = &(*mp)->m_next;
1229 if (inp->inp_flags & INP_RECVTOS) {
1230 *mp = sbcreatecontrol((caddr_t)&ip->ip_tos,
1231 sizeof(u_char), IP_RECVTOS, IPPROTO_IP);
1233 mp = &(*mp)->m_next;
1236 if (inp->inp_flags2 & INP_RECVFLOWID) {
1237 uint32_t flowid, flow_type;
1239 flowid = m->m_pkthdr.flowid;
1240 flow_type = M_HASHTYPE_GET(m);
1243 * XXX should handle the failure of one or the
1244 * other - don't populate both?
1246 *mp = sbcreatecontrol((caddr_t) &flowid,
1247 sizeof(uint32_t), IP_FLOWID, IPPROTO_IP);
1249 mp = &(*mp)->m_next;
1250 *mp = sbcreatecontrol((caddr_t) &flow_type,
1251 sizeof(uint32_t), IP_FLOWTYPE, IPPROTO_IP);
1253 mp = &(*mp)->m_next;
1257 if (inp->inp_flags2 & INP_RECVRSSBUCKETID) {
1258 uint32_t flowid, flow_type;
1259 uint32_t rss_bucketid;
1261 flowid = m->m_pkthdr.flowid;
1262 flow_type = M_HASHTYPE_GET(m);
1264 if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) {
1265 *mp = sbcreatecontrol((caddr_t) &rss_bucketid,
1266 sizeof(uint32_t), IP_RSSBUCKETID, IPPROTO_IP);
1268 mp = &(*mp)->m_next;
1275 * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
1276 * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
1277 * locking. This code remains in ip_input.c as ip_mroute.c is optionally
1280 static VNET_DEFINE(int, ip_rsvp_on);
1281 VNET_DEFINE(struct socket *, ip_rsvpd);
1283 #define V_ip_rsvp_on VNET(ip_rsvp_on)
1286 ip_rsvp_init(struct socket *so)
1289 if (so->so_type != SOCK_RAW ||
1290 so->so_proto->pr_protocol != IPPROTO_RSVP)
1293 if (V_ip_rsvpd != NULL)
1298 * This may seem silly, but we need to be sure we don't over-increment
1299 * the RSVP counter, in case something slips up.
1301 if (!V_ip_rsvp_on) {
1315 * This may seem silly, but we need to be sure we don't over-decrement
1316 * the RSVP counter, in case something slips up.
1326 rsvp_input(struct mbuf **mp, int *offp, int proto)
1333 if (rsvp_input_p) { /* call the real one if loaded */
1335 rsvp_input_p(mp, offp, proto);
1336 return (IPPROTO_DONE);
1339 /* Can still get packets with rsvp_on = 0 if there is a local member
1340 * of the group to which the RSVP packet is addressed. But in this
1341 * case we want to throw the packet away.
1346 return (IPPROTO_DONE);
1349 if (V_ip_rsvpd != NULL) {
1351 rip_input(mp, offp, proto);
1352 return (IPPROTO_DONE);
1354 /* Drop the packet */
1356 return (IPPROTO_DONE);