2 * SPDX-License-Identifier: BSD-3-Clause
4 * Copyright (c) 1982, 1986, 1988, 1993
5 * The Regents of the University of California. All rights reserved.
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 * 3. Neither the name of the University nor the names of its contributors
16 * may be used to endorse or promote products derived from this software
17 * without specific prior written permission.
19 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94
34 #include <sys/cdefs.h>
35 __FBSDID("$FreeBSD$");
37 #include "opt_bootp.h"
38 #include "opt_ipstealth.h"
39 #include "opt_ipsec.h"
40 #include "opt_route.h"
43 #include <sys/param.h>
44 #include <sys/systm.h>
45 #include <sys/hhook.h>
47 #include <sys/malloc.h>
48 #include <sys/domain.h>
49 #include <sys/protosw.h>
50 #include <sys/socket.h>
52 #include <sys/kernel.h>
54 #include <sys/rmlock.h>
55 #include <sys/rwlock.h>
57 #include <sys/syslog.h>
58 #include <sys/sysctl.h>
61 #include <net/if_types.h>
62 #include <net/if_var.h>
63 #include <net/if_dl.h>
65 #include <net/route.h>
66 #include <net/route/nhop.h>
67 #include <net/netisr.h>
68 #include <net/rss_config.h>
71 #include <netinet/in.h>
72 #include <netinet/in_kdtrace.h>
73 #include <netinet/in_systm.h>
74 #include <netinet/in_var.h>
75 #include <netinet/ip.h>
76 #include <netinet/in_fib.h>
77 #include <netinet/in_pcb.h>
78 #include <netinet/ip_var.h>
79 #include <netinet/ip_fw.h>
80 #include <netinet/ip_icmp.h>
81 #include <netinet/ip_options.h>
82 #include <machine/in_cksum.h>
83 #include <netinet/ip_carp.h>
84 #include <netinet/in_rss.h>
86 #include <netipsec/ipsec_support.h>
88 #include <sys/socketvar.h>
90 #include <security/mac/mac_framework.h>
93 CTASSERT(sizeof(struct ip) == 20);
96 /* IP reassembly functions are defined in ip_reass.c. */
97 extern void ipreass_init(void);
98 extern void ipreass_drain(void);
99 extern void ipreass_slowtimo(void);
101 extern void ipreass_destroy(void);
104 struct rmlock in_ifaddr_lock;
105 RM_SYSINIT(in_ifaddr_lock, &in_ifaddr_lock, "in_ifaddr_lock");
107 VNET_DEFINE(int, rsvp_on);
109 VNET_DEFINE(int, ipforwarding);
110 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_VNET | CTLFLAG_RW,
111 &VNET_NAME(ipforwarding), 0,
112 "Enable IP forwarding between interfaces");
115 * Respond with an ICMP host redirect when we forward a packet out of
116 * the same interface on which it was received. See RFC 792.
118 VNET_DEFINE(int, ipsendredirects) = 1;
119 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_VNET | CTLFLAG_RW,
120 &VNET_NAME(ipsendredirects), 0,
121 "Enable sending IP redirects");
124 * XXX - Setting ip_checkinterface mostly implements the receive side of
125 * the Strong ES model described in RFC 1122, but since the routing table
126 * and transmit implementation do not implement the Strong ES model,
127 * setting this to 1 results in an odd hybrid.
129 * XXX - ip_checkinterface currently must be disabled if you use ipnat
130 * to translate the destination address to another local interface.
132 * XXX - ip_checkinterface must be disabled if you add IP aliases
133 * to the loopback interface instead of the interface where the
134 * packets for those addresses are received.
136 VNET_DEFINE_STATIC(int, ip_checkinterface);
137 #define V_ip_checkinterface VNET(ip_checkinterface)
138 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_VNET | CTLFLAG_RW,
139 &VNET_NAME(ip_checkinterface), 0,
140 "Verify packet arrives on correct interface");
142 VNET_DEFINE(pfil_head_t, inet_pfil_head); /* Packet filter hooks */
144 static struct netisr_handler ip_nh = {
146 .nh_handler = ip_input,
147 .nh_proto = NETISR_IP,
149 .nh_m2cpuid = rss_soft_m2cpuid_v4,
150 .nh_policy = NETISR_POLICY_CPU,
151 .nh_dispatch = NETISR_DISPATCH_HYBRID,
153 .nh_policy = NETISR_POLICY_FLOW,
159 * Directly dispatched frames are currently assumed
160 * to have a flowid already calculated.
162 * It should likely have something that assert it
163 * actually has valid flow details.
165 static struct netisr_handler ip_direct_nh = {
166 .nh_name = "ip_direct",
167 .nh_handler = ip_direct_input,
168 .nh_proto = NETISR_IP_DIRECT,
169 .nh_m2cpuid = rss_soft_m2cpuid_v4,
170 .nh_policy = NETISR_POLICY_CPU,
171 .nh_dispatch = NETISR_DISPATCH_HYBRID,
175 extern struct domain inetdomain;
176 extern struct protosw inetsw[];
177 u_char ip_protox[IPPROTO_MAX];
178 VNET_DEFINE(struct in_ifaddrhead, in_ifaddrhead); /* first inet address */
179 VNET_DEFINE(struct in_ifaddrhashhead *, in_ifaddrhashtbl); /* inet addr hash table */
180 VNET_DEFINE(u_long, in_ifaddrhmask); /* mask for hash table */
183 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
184 &ip_mtu, 0, "Default MTU");
188 VNET_DEFINE(int, ipstealth);
189 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_VNET | CTLFLAG_RW,
190 &VNET_NAME(ipstealth), 0,
191 "IP stealth mode, no TTL decrementation on forwarding");
195 * IP statistics are stored in the "array" of counter(9)s.
197 VNET_PCPUSTAT_DEFINE(struct ipstat, ipstat);
198 VNET_PCPUSTAT_SYSINIT(ipstat);
199 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, IPCTL_STATS, stats, struct ipstat, ipstat,
200 "IP statistics (struct ipstat, netinet/ip_var.h)");
203 VNET_PCPUSTAT_SYSUNINIT(ipstat);
207 * Kernel module interface for updating ipstat. The argument is an index
208 * into ipstat treated as an array.
211 kmod_ipstat_inc(int statnum)
214 counter_u64_add(VNET(ipstat)[statnum], 1);
218 kmod_ipstat_dec(int statnum)
221 counter_u64_add(VNET(ipstat)[statnum], -1);
225 sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS)
229 netisr_getqlimit(&ip_nh, &qlimit);
230 error = sysctl_handle_int(oidp, &qlimit, 0, req);
231 if (error || !req->newptr)
235 return (netisr_setqlimit(&ip_nh, qlimit));
237 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen,
238 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 0,
239 sysctl_netinet_intr_queue_maxlen, "I",
240 "Maximum size of the IP input queue");
243 sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS)
245 u_int64_t qdrops_long;
248 netisr_getqdrops(&ip_nh, &qdrops_long);
249 qdrops = qdrops_long;
250 error = sysctl_handle_int(oidp, &qdrops, 0, req);
251 if (error || !req->newptr)
255 netisr_clearqdrops(&ip_nh);
259 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops,
260 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE,
261 0, 0, sysctl_netinet_intr_queue_drops, "I",
262 "Number of packets dropped from the IP input queue");
266 sysctl_netinet_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS)
270 netisr_getqlimit(&ip_direct_nh, &qlimit);
271 error = sysctl_handle_int(oidp, &qlimit, 0, req);
272 if (error || !req->newptr)
276 return (netisr_setqlimit(&ip_direct_nh, qlimit));
278 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRDQMAXLEN, intr_direct_queue_maxlen,
279 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
280 0, 0, sysctl_netinet_intr_direct_queue_maxlen,
281 "I", "Maximum size of the IP direct input queue");
284 sysctl_netinet_intr_direct_queue_drops(SYSCTL_HANDLER_ARGS)
286 u_int64_t qdrops_long;
289 netisr_getqdrops(&ip_direct_nh, &qdrops_long);
290 qdrops = qdrops_long;
291 error = sysctl_handle_int(oidp, &qdrops, 0, req);
292 if (error || !req->newptr)
296 netisr_clearqdrops(&ip_direct_nh);
300 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRDQDROPS, intr_direct_queue_drops,
301 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, 0, 0,
302 sysctl_netinet_intr_direct_queue_drops, "I",
303 "Number of packets dropped from the IP direct input queue");
307 * IP initialization: fill in IP protocol switch table.
308 * All protocols not implemented in kernel go to raw IP protocol handler.
313 struct pfil_head_args args;
317 CK_STAILQ_INIT(&V_in_ifaddrhead);
318 V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask);
320 /* Initialize IP reassembly queue. */
323 /* Initialize packet filter hooks. */
324 args.pa_version = PFIL_VERSION;
325 args.pa_flags = PFIL_IN | PFIL_OUT;
326 args.pa_type = PFIL_TYPE_IP4;
327 args.pa_headname = PFIL_INET_NAME;
328 V_inet_pfil_head = pfil_head_register(&args);
330 if (hhook_head_register(HHOOK_TYPE_IPSEC_IN, AF_INET,
331 &V_ipsec_hhh_in[HHOOK_IPSEC_INET],
332 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
333 printf("%s: WARNING: unable to register input helper hook\n",
335 if (hhook_head_register(HHOOK_TYPE_IPSEC_OUT, AF_INET,
336 &V_ipsec_hhh_out[HHOOK_IPSEC_INET],
337 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
338 printf("%s: WARNING: unable to register output helper hook\n",
341 /* Skip initialization of globals for non-default instances. */
343 if (!IS_DEFAULT_VNET(curvnet)) {
344 netisr_register_vnet(&ip_nh);
346 netisr_register_vnet(&ip_direct_nh);
352 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
354 panic("ip_init: PF_INET not found");
356 /* Initialize the entire ip_protox[] array to IPPROTO_RAW. */
357 for (i = 0; i < IPPROTO_MAX; i++)
358 ip_protox[i] = pr - inetsw;
360 * Cycle through IP protocols and put them into the appropriate place
363 for (pr = inetdomain.dom_protosw;
364 pr < inetdomain.dom_protoswNPROTOSW; pr++)
365 if (pr->pr_domain->dom_family == PF_INET &&
366 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
367 /* Be careful to only index valid IP protocols. */
368 if (pr->pr_protocol < IPPROTO_MAX)
369 ip_protox[pr->pr_protocol] = pr - inetsw;
372 netisr_register(&ip_nh);
374 netisr_register(&ip_direct_nh);
380 ip_destroy(void *unused __unused)
386 netisr_unregister_vnet(&ip_direct_nh);
388 netisr_unregister_vnet(&ip_nh);
390 pfil_head_unregister(V_inet_pfil_head);
391 error = hhook_head_deregister(V_ipsec_hhh_in[HHOOK_IPSEC_INET]);
393 printf("%s: WARNING: unable to deregister input helper hook "
394 "type HHOOK_TYPE_IPSEC_IN, id HHOOK_IPSEC_INET: "
395 "error %d returned\n", __func__, error);
397 error = hhook_head_deregister(V_ipsec_hhh_out[HHOOK_IPSEC_INET]);
399 printf("%s: WARNING: unable to deregister output helper hook "
400 "type HHOOK_TYPE_IPSEC_OUT, id HHOOK_IPSEC_INET: "
401 "error %d returned\n", __func__, error);
404 /* Remove the IPv4 addresses from all interfaces. */
407 /* Make sure the IPv4 routes are gone as well. */
409 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link)
410 rt_flushifroutes_af(ifp, AF_INET);
413 /* Destroy IP reassembly queue. */
416 /* Cleanup in_ifaddr hash table; should be empty. */
417 hashdestroy(V_in_ifaddrhashtbl, M_IFADDR, V_in_ifaddrhmask);
420 VNET_SYSUNINIT(ip, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip_destroy, NULL);
425 * IP direct input routine.
427 * This is called when reinjecting completed fragments where
428 * all of the previous checking and book-keeping has been done.
431 ip_direct_input(struct mbuf *m)
436 ip = mtod(m, struct ip *);
437 hlen = ip->ip_hl << 2;
439 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
440 if (IPSEC_ENABLED(ipv4)) {
441 if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
445 IPSTAT_INC(ips_delivered);
446 (*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
452 * Ip input routine. Checksum and byte swap header. If fragmented
453 * try to reassemble. Process options. Pass to next level.
456 ip_input(struct mbuf *m)
458 struct rm_priotracker in_ifa_tracker;
459 struct ip *ip = NULL;
460 struct in_ifaddr *ia = NULL;
463 int checkif, hlen = 0;
464 uint16_t sum, ip_len;
465 int dchg = 0; /* dest changed after fw */
466 struct in_addr odst; /* original dst address */
471 if (m->m_flags & M_FASTFWD_OURS) {
472 m->m_flags &= ~M_FASTFWD_OURS;
473 /* Set up some basics that will be used later. */
474 ip = mtod(m, struct ip *);
475 hlen = ip->ip_hl << 2;
476 ip_len = ntohs(ip->ip_len);
480 IPSTAT_INC(ips_total);
482 if (m->m_pkthdr.len < sizeof(struct ip))
485 if (m->m_len < sizeof (struct ip) &&
486 (m = m_pullup(m, sizeof (struct ip))) == NULL) {
487 IPSTAT_INC(ips_toosmall);
490 ip = mtod(m, struct ip *);
492 if (ip->ip_v != IPVERSION) {
493 IPSTAT_INC(ips_badvers);
497 hlen = ip->ip_hl << 2;
498 if (hlen < sizeof(struct ip)) { /* minimum header length */
499 IPSTAT_INC(ips_badhlen);
502 if (hlen > m->m_len) {
503 if ((m = m_pullup(m, hlen)) == NULL) {
504 IPSTAT_INC(ips_badhlen);
507 ip = mtod(m, struct ip *);
510 IP_PROBE(receive, NULL, NULL, ip, m->m_pkthdr.rcvif, ip, NULL);
512 /* IN_LOOPBACK must not appear on the wire - RFC1122 */
513 ifp = m->m_pkthdr.rcvif;
514 if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) ||
515 IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) {
516 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
517 IPSTAT_INC(ips_badaddr);
522 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
523 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
525 if (hlen == sizeof(struct ip)) {
526 sum = in_cksum_hdr(ip);
528 sum = in_cksum(m, hlen);
532 IPSTAT_INC(ips_badsum);
537 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
538 /* packet is dropped by traffic conditioner */
542 ip_len = ntohs(ip->ip_len);
544 IPSTAT_INC(ips_badlen);
549 * Check that the amount of data in the buffers
550 * is as at least much as the IP header would have us expect.
551 * Trim mbufs if longer than we expect.
552 * Drop packet if shorter than we expect.
554 if (m->m_pkthdr.len < ip_len) {
556 IPSTAT_INC(ips_tooshort);
559 if (m->m_pkthdr.len > ip_len) {
560 if (m->m_len == m->m_pkthdr.len) {
562 m->m_pkthdr.len = ip_len;
564 m_adj(m, ip_len - m->m_pkthdr.len);
568 * Try to forward the packet, but if we fail continue.
569 * ip_tryforward() does not generate redirects, so fall
570 * through to normal processing if redirects are required.
571 * ip_tryforward() does inbound and outbound packet firewall
572 * processing. If firewall has decided that destination becomes
573 * our local address, it sets M_FASTFWD_OURS flag. In this
574 * case skip another inbound firewall processing and update
577 if (V_ipforwarding != 0
578 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
579 && (!IPSEC_ENABLED(ipv4) ||
580 IPSEC_CAPS(ipv4, m, IPSEC_CAP_OPERABLE) == 0)
583 if ((m = ip_tryforward(m)) == NULL)
585 if (m->m_flags & M_FASTFWD_OURS) {
586 m->m_flags &= ~M_FASTFWD_OURS;
587 ip = mtod(m, struct ip *);
592 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
594 * Bypass packet filtering for packets previously handled by IPsec.
596 if (IPSEC_ENABLED(ipv4) &&
597 IPSEC_CAPS(ipv4, m, IPSEC_CAP_BYPASS_FILTER) != 0)
602 * Run through list of hooks for input packets.
604 * NB: Beware of the destination address changing (e.g.
605 * by NAT rewriting). When this happens, tell
606 * ip_forward to do the right thing.
609 /* Jump over all PFIL processing if hooks are not active. */
610 if (!PFIL_HOOKED_IN(V_inet_pfil_head))
614 if (pfil_run_hooks(V_inet_pfil_head, &m, ifp, PFIL_IN, NULL) !=
617 if (m == NULL) /* consumed by filter */
620 ip = mtod(m, struct ip *);
621 dchg = (odst.s_addr != ip->ip_dst.s_addr);
622 ifp = m->m_pkthdr.rcvif;
624 if (m->m_flags & M_FASTFWD_OURS) {
625 m->m_flags &= ~M_FASTFWD_OURS;
628 if (m->m_flags & M_IP_NEXTHOP) {
629 if (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) {
631 * Directly ship the packet on. This allows
632 * forwarding packets originally destined to us
633 * to some other directly connected host.
642 * Process options and, if not destined for us,
643 * ship it on. ip_dooptions returns 1 when an
644 * error was detected (causing an icmp message
645 * to be sent and the original packet to be freed).
647 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
650 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
651 * matter if it is destined to another node, or whether it is
652 * a multicast one, RSVP wants it! and prevents it from being forwarded
653 * anywhere else. Also checks if the rsvp daemon is running before
654 * grabbing the packet.
656 if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP)
660 * Check our list of addresses, to see if the packet is for us.
661 * If we don't have any addresses, assume any unicast packet
662 * we receive might be for us (and let the upper layers deal
665 if (CK_STAILQ_EMPTY(&V_in_ifaddrhead) &&
666 (m->m_flags & (M_MCAST|M_BCAST)) == 0)
670 * Enable a consistency check between the destination address
671 * and the arrival interface for a unicast packet (the RFC 1122
672 * strong ES model) if IP forwarding is disabled and the packet
673 * is not locally generated and the packet is not subject to
676 * XXX - Checking also should be disabled if the destination
677 * address is ipnat'ed to a different interface.
679 * XXX - Checking is incompatible with IP aliases added
680 * to the loopback interface instead of the interface where
681 * the packets are received.
683 * XXX - This is the case for carp vhost IPs as well so we
684 * insert a workaround. If the packet got here, we already
685 * checked with carp_iamatch() and carp_forus().
687 checkif = V_ip_checkinterface && (V_ipforwarding == 0) &&
688 ifp != NULL && ((ifp->if_flags & IFF_LOOPBACK) == 0) &&
689 ifp->if_carp == NULL && (dchg == 0);
692 * Check for exact addresses in the hash bucket.
694 IN_IFADDR_RLOCK(&in_ifa_tracker);
695 LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
697 * If the address matches, verify that the packet
698 * arrived via the correct interface if checking is
701 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr &&
702 (!checkif || ia->ia_ifp == ifp)) {
703 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
704 counter_u64_add(ia->ia_ifa.ifa_ibytes,
706 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
710 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
713 * Check for broadcast addresses.
715 * Only accept broadcast packets that arrive via the matching
716 * interface. Reception of forwarded directed broadcasts would
717 * be handled via ip_forward() and ether_output() with the loopback
718 * into the stack for SIMPLEX interfaces handled by ether_output().
720 if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
721 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
722 if (ifa->ifa_addr->sa_family != AF_INET)
725 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
727 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
728 counter_u64_add(ia->ia_ifa.ifa_ibytes,
733 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) {
734 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
735 counter_u64_add(ia->ia_ifa.ifa_ibytes,
743 /* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */
744 if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) {
745 IPSTAT_INC(ips_cantforward);
749 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
752 * If we are acting as a multicast router, all
753 * incoming multicast packets are passed to the
754 * kernel-level multicast forwarding function.
755 * The packet is returned (relatively) intact; if
756 * ip_mforward() returns a non-zero value, the packet
757 * must be discarded, else it may be accepted below.
759 if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) {
760 IPSTAT_INC(ips_cantforward);
766 * The process-level routing daemon needs to receive
767 * all multicast IGMP packets, whether or not this
768 * host belongs to their destination groups.
770 if (ip->ip_p == IPPROTO_IGMP)
772 IPSTAT_INC(ips_forward);
775 * Assume the packet is for us, to avoid prematurely taking
776 * a lock on the in_multi hash. Protocols must perform
777 * their own filtering and update statistics accordingly.
781 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
783 if (ip->ip_dst.s_addr == INADDR_ANY)
787 * Not for us; forward if possible and desirable.
789 if (V_ipforwarding == 0) {
790 IPSTAT_INC(ips_cantforward);
800 * IPSTEALTH: Process non-routing options only
801 * if the packet is destined for us.
803 if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1))
805 #endif /* IPSTEALTH */
808 * Attempt reassembly; if it succeeds, proceed.
809 * ip_reass() will return a different mbuf.
811 if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) {
812 /* XXXGL: shouldn't we save & set m_flags? */
816 ip = mtod(m, struct ip *);
817 /* Get the header length of the reassembled packet */
818 hlen = ip->ip_hl << 2;
821 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
822 if (IPSEC_ENABLED(ipv4)) {
823 if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
829 * Switch out to protocol's input routine.
831 IPSTAT_INC(ips_delivered);
833 (*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
840 * IP timer processing;
841 * if a timer expires on a reassembly
847 VNET_ITERATOR_DECL(vnet_iter);
849 VNET_LIST_RLOCK_NOSLEEP();
850 VNET_FOREACH(vnet_iter) {
851 CURVNET_SET(vnet_iter);
855 VNET_LIST_RUNLOCK_NOSLEEP();
861 VNET_ITERATOR_DECL(vnet_iter);
863 VNET_LIST_RLOCK_NOSLEEP();
864 VNET_FOREACH(vnet_iter) {
865 CURVNET_SET(vnet_iter);
869 VNET_LIST_RUNLOCK_NOSLEEP();
873 * The protocol to be inserted into ip_protox[] must be already registered
874 * in inetsw[], either statically or through pf_proto_register().
877 ipproto_register(short ipproto)
882 if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
883 return (EPROTONOSUPPORT);
886 * The protocol slot must not be occupied by another protocol
887 * already. An index pointing to IPPROTO_RAW is unused.
889 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
891 return (EPFNOSUPPORT);
892 if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */
895 /* Find the protocol position in inetsw[] and set the index. */
896 for (pr = inetdomain.dom_protosw;
897 pr < inetdomain.dom_protoswNPROTOSW; pr++) {
898 if (pr->pr_domain->dom_family == PF_INET &&
899 pr->pr_protocol && pr->pr_protocol == ipproto) {
900 ip_protox[pr->pr_protocol] = pr - inetsw;
904 return (EPROTONOSUPPORT);
908 ipproto_unregister(short ipproto)
913 if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
914 return (EPROTONOSUPPORT);
916 /* Check if the protocol was indeed registered. */
917 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
919 return (EPFNOSUPPORT);
920 if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */
923 /* Reset the protocol slot to IPPROTO_RAW. */
924 ip_protox[ipproto] = pr - inetsw;
928 u_char inetctlerrmap[PRC_NCMDS] = {
930 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
931 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
932 EMSGSIZE, EHOSTUNREACH, 0, 0,
933 0, 0, EHOSTUNREACH, 0,
934 ENOPROTOOPT, ECONNREFUSED
938 * Forward a packet. If some error occurs return the sender
939 * an icmp packet. Note we can't always generate a meaningful
940 * icmp message because icmp doesn't have a large enough repertoire
941 * of codes and types.
943 * If not forwarding, just drop the packet. This could be confusing
944 * if ipforwarding was zero but some routing protocol was advancing
945 * us as a gateway to somewhere. However, we must let the routing
946 * protocol deal with that.
948 * The srcrt parameter indicates whether the packet is being forwarded
949 * via a source route.
952 ip_forward(struct mbuf *m, int srcrt)
954 struct ip *ip = mtod(m, struct ip *);
955 struct in_ifaddr *ia;
957 struct sockaddr_in *sin;
961 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;
984 flowid = m->m_pkthdr.flowid;
985 ro.ro_nh = fib4_lookup(M_GETFIB(m), ip->ip_dst, 0, NHR_REF, flowid);
986 if (ro.ro_nh != NULL) {
987 ia = ifatoia(ro.ro_nh->nh_ifa);
991 * Save the IP header and at most 8 bytes of the payload,
992 * in case we need to generate an ICMP message to the src.
994 * XXX this can be optimized a lot by saving the data in a local
995 * buffer on the stack (72 bytes at most), and only allocating the
996 * mbuf if really necessary. The vast majority of the packets
997 * are forwarded without having to send an ICMP back (either
998 * because unnecessary, or because rate limited), so we are
999 * really we are wasting a lot of work here.
1001 * We don't use m_copym() because it might return a reference
1002 * to a shared cluster. Both this function and ip_output()
1003 * assume exclusive access to the IP header in `m', so any
1004 * data in a cluster may change before we reach icmp_error().
1006 mcopy = m_gethdr(M_NOWAIT, m->m_type);
1007 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_NOWAIT)) {
1009 * It's probably ok if the pkthdr dup fails (because
1010 * the deep copy of the tag chain failed), but for now
1011 * be conservative and just discard the copy since
1012 * code below may some day want the tags.
1017 if (mcopy != NULL) {
1018 mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy));
1019 mcopy->m_pkthdr.len = mcopy->m_len;
1020 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1023 if (V_ipstealth == 0)
1025 ip->ip_ttl -= IPTTLDEC;
1026 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1027 if (IPSEC_ENABLED(ipv4)) {
1028 if ((error = IPSEC_FORWARD(ipv4, m)) != 0) {
1029 /* mbuf consumed by IPsec */
1032 if (error != EINPROGRESS)
1033 IPSTAT_INC(ips_cantforward);
1036 /* No IPsec processing required */
1040 * If forwarding packet using same interface that it came in on,
1041 * perhaps should send a redirect to sender to shortcut a hop.
1042 * Only send redirect if source is sending directly to us,
1043 * and if packet was not source routed (or has any options).
1044 * Also, don't send redirect if forwarding using a default route
1045 * or a route modified by a redirect.
1048 if (!srcrt && V_ipsendredirects &&
1049 ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) {
1050 struct nhop_object *nh;
1054 if (nh != NULL && ((nh->nh_flags & (NHF_REDIRECT|NHF_DEFAULT)) == 0)) {
1055 struct in_ifaddr *nh_ia = (struct in_ifaddr *)(nh->nh_ifa);
1056 u_long src = ntohl(ip->ip_src.s_addr);
1058 if (nh_ia != NULL &&
1059 (src & nh_ia->ia_subnetmask) == nh_ia->ia_subnet) {
1060 if (nh->nh_flags & NHF_GATEWAY)
1061 dest.s_addr = nh->gw4_sa.sin_addr.s_addr;
1063 dest.s_addr = ip->ip_dst.s_addr;
1064 /* Router requirements says to only send host redirects */
1065 type = ICMP_REDIRECT;
1066 code = ICMP_REDIRECT_HOST;
1071 error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
1073 if (error == EMSGSIZE && ro.ro_nh)
1074 mtu = ro.ro_nh->nh_mtu;
1078 IPSTAT_INC(ips_cantforward);
1080 IPSTAT_INC(ips_forward);
1082 IPSTAT_INC(ips_redirectsent);
1093 case 0: /* forwarded, but need redirect */
1094 /* type, code set above */
1102 type = ICMP_UNREACH;
1103 code = ICMP_UNREACH_HOST;
1107 type = ICMP_UNREACH;
1108 code = ICMP_UNREACH_NEEDFRAG;
1110 * If the MTU was set before make sure we are below the
1112 * If the MTU wasn't set before use the interface mtu or
1113 * fall back to the next smaller mtu step compared to the
1114 * current packet size.
1118 mtu = min(mtu, ia->ia_ifp->if_mtu);
1121 mtu = ia->ia_ifp->if_mtu;
1123 mtu = ip_next_mtu(ntohs(ip->ip_len), 0);
1125 IPSTAT_INC(ips_cantfrag);
1129 case EACCES: /* ipfw denied packet */
1133 icmp_error(mcopy, type, code, dest.s_addr, mtu);
1136 #define CHECK_SO_CT(sp, ct) \
1137 (((sp->so_options & SO_TIMESTAMP) && (sp->so_ts_clock == ct)) ? 1 : 0)
1140 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1146 if ((inp->inp_socket->so_options & SO_BINTIME) ||
1147 CHECK_SO_CT(inp->inp_socket, SO_TS_BINTIME)) {
1148 struct bintime boottimebin, bt;
1149 struct timespec ts1;
1151 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1153 mbuf_tstmp2timespec(m, &ts1);
1154 timespec2bintime(&ts1, &bt);
1155 getboottimebin(&boottimebin);
1156 bintime_add(&bt, &boottimebin);
1160 *mp = sbcreatecontrol((caddr_t)&bt, sizeof(bt),
1161 SCM_BINTIME, SOL_SOCKET);
1163 mp = &(*mp)->m_next;
1167 if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME_MICRO)) {
1168 struct bintime boottimebin, bt1;
1169 struct timespec ts1;
1172 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1174 mbuf_tstmp2timespec(m, &ts1);
1175 timespec2bintime(&ts1, &bt1);
1176 getboottimebin(&boottimebin);
1177 bintime_add(&bt1, &boottimebin);
1178 bintime2timeval(&bt1, &tv);
1182 *mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv),
1183 SCM_TIMESTAMP, SOL_SOCKET);
1185 mp = &(*mp)->m_next;
1188 } else if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME)) {
1189 struct bintime boottimebin;
1190 struct timespec ts, ts1;
1192 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1194 mbuf_tstmp2timespec(m, &ts);
1195 getboottimebin(&boottimebin);
1196 bintime2timespec(&boottimebin, &ts1);
1197 timespecadd(&ts, &ts1, &ts);
1201 *mp = sbcreatecontrol((caddr_t)&ts, sizeof(ts),
1202 SCM_REALTIME, SOL_SOCKET);
1204 mp = &(*mp)->m_next;
1207 } else if (CHECK_SO_CT(inp->inp_socket, SO_TS_MONOTONIC)) {
1210 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1212 mbuf_tstmp2timespec(m, &ts);
1215 *mp = sbcreatecontrol((caddr_t)&ts, sizeof(ts),
1216 SCM_MONOTONIC, SOL_SOCKET);
1218 mp = &(*mp)->m_next;
1222 if (stamped && (m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1224 struct sock_timestamp_info sti;
1226 bzero(&sti, sizeof(sti));
1227 sti.st_info_flags = ST_INFO_HW;
1228 if ((m->m_flags & M_TSTMP_HPREC) != 0)
1229 sti.st_info_flags |= ST_INFO_HW_HPREC;
1230 *mp = sbcreatecontrol((caddr_t)&sti, sizeof(sti), SCM_TIME_INFO,
1233 mp = &(*mp)->m_next;
1235 if (inp->inp_flags & INP_RECVDSTADDR) {
1236 *mp = sbcreatecontrol((caddr_t)&ip->ip_dst,
1237 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1239 mp = &(*mp)->m_next;
1241 if (inp->inp_flags & INP_RECVTTL) {
1242 *mp = sbcreatecontrol((caddr_t)&ip->ip_ttl,
1243 sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
1245 mp = &(*mp)->m_next;
1249 * Moving these out of udp_input() made them even more broken
1250 * than they already were.
1252 /* options were tossed already */
1253 if (inp->inp_flags & INP_RECVOPTS) {
1254 *mp = sbcreatecontrol((caddr_t)opts_deleted_above,
1255 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1257 mp = &(*mp)->m_next;
1259 /* ip_srcroute doesn't do what we want here, need to fix */
1260 if (inp->inp_flags & INP_RECVRETOPTS) {
1261 *mp = sbcreatecontrol((caddr_t)ip_srcroute(m),
1262 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1264 mp = &(*mp)->m_next;
1267 if (inp->inp_flags & INP_RECVIF) {
1270 struct sockaddr_dl sdl;
1273 struct sockaddr_dl *sdp;
1274 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1276 if ((ifp = m->m_pkthdr.rcvif) &&
1277 ifp->if_index && ifp->if_index <= V_if_index) {
1278 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1280 * Change our mind and don't try copy.
1282 if (sdp->sdl_family != AF_LINK ||
1283 sdp->sdl_len > sizeof(sdlbuf)) {
1286 bcopy(sdp, sdl2, sdp->sdl_len);
1290 offsetof(struct sockaddr_dl, sdl_data[0]);
1291 sdl2->sdl_family = AF_LINK;
1292 sdl2->sdl_index = 0;
1293 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1295 *mp = sbcreatecontrol((caddr_t)sdl2, sdl2->sdl_len,
1296 IP_RECVIF, IPPROTO_IP);
1298 mp = &(*mp)->m_next;
1300 if (inp->inp_flags & INP_RECVTOS) {
1301 *mp = sbcreatecontrol((caddr_t)&ip->ip_tos,
1302 sizeof(u_char), IP_RECVTOS, IPPROTO_IP);
1304 mp = &(*mp)->m_next;
1307 if (inp->inp_flags2 & INP_RECVFLOWID) {
1308 uint32_t flowid, flow_type;
1310 flowid = m->m_pkthdr.flowid;
1311 flow_type = M_HASHTYPE_GET(m);
1314 * XXX should handle the failure of one or the
1315 * other - don't populate both?
1317 *mp = sbcreatecontrol((caddr_t) &flowid,
1318 sizeof(uint32_t), IP_FLOWID, IPPROTO_IP);
1320 mp = &(*mp)->m_next;
1321 *mp = sbcreatecontrol((caddr_t) &flow_type,
1322 sizeof(uint32_t), IP_FLOWTYPE, IPPROTO_IP);
1324 mp = &(*mp)->m_next;
1328 if (inp->inp_flags2 & INP_RECVRSSBUCKETID) {
1329 uint32_t flowid, flow_type;
1330 uint32_t rss_bucketid;
1332 flowid = m->m_pkthdr.flowid;
1333 flow_type = M_HASHTYPE_GET(m);
1335 if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) {
1336 *mp = sbcreatecontrol((caddr_t) &rss_bucketid,
1337 sizeof(uint32_t), IP_RSSBUCKETID, IPPROTO_IP);
1339 mp = &(*mp)->m_next;
1346 * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
1347 * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
1348 * locking. This code remains in ip_input.c as ip_mroute.c is optionally
1351 VNET_DEFINE_STATIC(int, ip_rsvp_on);
1352 VNET_DEFINE(struct socket *, ip_rsvpd);
1354 #define V_ip_rsvp_on VNET(ip_rsvp_on)
1357 ip_rsvp_init(struct socket *so)
1360 if (so->so_type != SOCK_RAW ||
1361 so->so_proto->pr_protocol != IPPROTO_RSVP)
1364 if (V_ip_rsvpd != NULL)
1369 * This may seem silly, but we need to be sure we don't over-increment
1370 * the RSVP counter, in case something slips up.
1372 if (!V_ip_rsvp_on) {
1386 * This may seem silly, but we need to be sure we don't over-decrement
1387 * the RSVP counter, in case something slips up.
1397 rsvp_input(struct mbuf **mp, int *offp, int proto)
1404 if (rsvp_input_p) { /* call the real one if loaded */
1406 rsvp_input_p(mp, offp, proto);
1407 return (IPPROTO_DONE);
1410 /* Can still get packets with rsvp_on = 0 if there is a local member
1411 * of the group to which the RSVP packet is addressed. But in this
1412 * case we want to throw the packet away.
1417 return (IPPROTO_DONE);
1420 if (V_ip_rsvpd != NULL) {
1422 rip_input(mp, offp, proto);
1423 return (IPPROTO_DONE);
1425 /* Drop the packet */
1427 return (IPPROTO_DONE);