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"
37 #include "opt_ipstealth.h"
38 #include "opt_ipsec.h"
39 #include "opt_route.h"
42 #include <sys/param.h>
43 #include <sys/systm.h>
44 #include <sys/hhook.h>
46 #include <sys/malloc.h>
47 #include <sys/domain.h>
48 #include <sys/protosw.h>
49 #include <sys/socket.h>
51 #include <sys/kernel.h>
53 #include <sys/rmlock.h>
54 #include <sys/rwlock.h>
56 #include <sys/syslog.h>
57 #include <sys/sysctl.h>
61 #include <net/if_types.h>
62 #include <net/if_var.h>
63 #include <net/if_dl.h>
64 #include <net/route.h>
65 #include <net/netisr.h>
66 #include <net/rss_config.h>
69 #include <netinet/in.h>
70 #include <netinet/in_kdtrace.h>
71 #include <netinet/in_systm.h>
72 #include <netinet/in_var.h>
73 #include <netinet/ip.h>
74 #include <netinet/in_pcb.h>
75 #include <netinet/ip_var.h>
76 #include <netinet/ip_fw.h>
77 #include <netinet/ip_icmp.h>
78 #include <netinet/ip_options.h>
79 #include <machine/in_cksum.h>
80 #include <netinet/ip_carp.h>
82 #include <netinet/ip_ipsec.h>
83 #include <netipsec/ipsec.h>
84 #include <netipsec/key.h>
86 #include <netinet/in_rss.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");
114 static VNET_DEFINE(int, ipsendredirects) = 1; /* XXX */
115 #define V_ipsendredirects VNET(ipsendredirects)
116 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_VNET | CTLFLAG_RW,
117 &VNET_NAME(ipsendredirects), 0,
118 "Enable sending IP redirects");
121 * XXX - Setting ip_checkinterface mostly implements the receive side of
122 * the Strong ES model described in RFC 1122, but since the routing table
123 * and transmit implementation do not implement the Strong ES model,
124 * setting this to 1 results in an odd hybrid.
126 * XXX - ip_checkinterface currently must be disabled if you use ipnat
127 * to translate the destination address to another local interface.
129 * XXX - ip_checkinterface must be disabled if you add IP aliases
130 * to the loopback interface instead of the interface where the
131 * packets for those addresses are received.
133 static VNET_DEFINE(int, ip_checkinterface);
134 #define V_ip_checkinterface VNET(ip_checkinterface)
135 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_VNET | CTLFLAG_RW,
136 &VNET_NAME(ip_checkinterface), 0,
137 "Verify packet arrives on correct interface");
139 VNET_DEFINE(struct pfil_head, inet_pfil_hook); /* Packet filter hooks */
141 static struct netisr_handler ip_nh = {
143 .nh_handler = ip_input,
144 .nh_proto = NETISR_IP,
146 .nh_m2cpuid = rss_soft_m2cpuid_v4,
147 .nh_policy = NETISR_POLICY_CPU,
148 .nh_dispatch = NETISR_DISPATCH_HYBRID,
150 .nh_policy = NETISR_POLICY_FLOW,
156 * Directly dispatched frames are currently assumed
157 * to have a flowid already calculated.
159 * It should likely have something that assert it
160 * actually has valid flow details.
162 static struct netisr_handler ip_direct_nh = {
163 .nh_name = "ip_direct",
164 .nh_handler = ip_direct_input,
165 .nh_proto = NETISR_IP_DIRECT,
166 .nh_m2cpuid = rss_soft_m2cpuid_v4,
167 .nh_policy = NETISR_POLICY_CPU,
168 .nh_dispatch = NETISR_DISPATCH_HYBRID,
172 extern struct domain inetdomain;
173 extern struct protosw inetsw[];
174 u_char ip_protox[IPPROTO_MAX];
175 VNET_DEFINE(struct in_ifaddrhead, in_ifaddrhead); /* first inet address */
176 VNET_DEFINE(struct in_ifaddrhashhead *, in_ifaddrhashtbl); /* inet addr hash table */
177 VNET_DEFINE(u_long, in_ifaddrhmask); /* mask for hash table */
180 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
181 &ip_mtu, 0, "Default MTU");
185 VNET_DEFINE(int, ipstealth);
186 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_VNET | CTLFLAG_RW,
187 &VNET_NAME(ipstealth), 0,
188 "IP stealth mode, no TTL decrementation on forwarding");
192 * IP statistics are stored in the "array" of counter(9)s.
194 VNET_PCPUSTAT_DEFINE(struct ipstat, ipstat);
195 VNET_PCPUSTAT_SYSINIT(ipstat);
196 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, IPCTL_STATS, stats, struct ipstat, ipstat,
197 "IP statistics (struct ipstat, netinet/ip_var.h)");
200 VNET_PCPUSTAT_SYSUNINIT(ipstat);
204 * Kernel module interface for updating ipstat. The argument is an index
205 * into ipstat treated as an array.
208 kmod_ipstat_inc(int statnum)
211 counter_u64_add(VNET(ipstat)[statnum], 1);
215 kmod_ipstat_dec(int statnum)
218 counter_u64_add(VNET(ipstat)[statnum], -1);
222 sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS)
226 netisr_getqlimit(&ip_nh, &qlimit);
227 error = sysctl_handle_int(oidp, &qlimit, 0, req);
228 if (error || !req->newptr)
232 return (netisr_setqlimit(&ip_nh, qlimit));
234 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen,
235 CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet_intr_queue_maxlen, "I",
236 "Maximum size of the IP input queue");
239 sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS)
241 u_int64_t qdrops_long;
244 netisr_getqdrops(&ip_nh, &qdrops_long);
245 qdrops = qdrops_long;
246 error = sysctl_handle_int(oidp, &qdrops, 0, req);
247 if (error || !req->newptr)
251 netisr_clearqdrops(&ip_nh);
255 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops,
256 CTLTYPE_INT|CTLFLAG_RD, 0, 0, sysctl_netinet_intr_queue_drops, "I",
257 "Number of packets dropped from the IP input queue");
261 sysctl_netinet_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS)
265 netisr_getqlimit(&ip_direct_nh, &qlimit);
266 error = sysctl_handle_int(oidp, &qlimit, 0, req);
267 if (error || !req->newptr)
271 return (netisr_setqlimit(&ip_direct_nh, qlimit));
273 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_direct_queue_maxlen,
274 CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet_intr_direct_queue_maxlen, "I",
275 "Maximum size of the IP direct input queue");
278 sysctl_netinet_intr_direct_queue_drops(SYSCTL_HANDLER_ARGS)
280 u_int64_t qdrops_long;
283 netisr_getqdrops(&ip_direct_nh, &qdrops_long);
284 qdrops = qdrops_long;
285 error = sysctl_handle_int(oidp, &qdrops, 0, req);
286 if (error || !req->newptr)
290 netisr_clearqdrops(&ip_direct_nh);
294 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_direct_queue_drops,
295 CTLTYPE_INT|CTLFLAG_RD, 0, 0, sysctl_netinet_intr_direct_queue_drops, "I",
296 "Number of packets dropped from the IP direct input queue");
300 * IP initialization: fill in IP protocol switch table.
301 * All protocols not implemented in kernel go to raw IP protocol handler.
309 TAILQ_INIT(&V_in_ifaddrhead);
310 V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask);
312 /* Initialize IP reassembly queue. */
315 /* Initialize packet filter hooks. */
316 V_inet_pfil_hook.ph_type = PFIL_TYPE_AF;
317 V_inet_pfil_hook.ph_af = AF_INET;
318 if ((i = pfil_head_register(&V_inet_pfil_hook)) != 0)
319 printf("%s: WARNING: unable to register pfil hook, "
320 "error %d\n", __func__, i);
322 if (hhook_head_register(HHOOK_TYPE_IPSEC_IN, AF_INET,
323 &V_ipsec_hhh_in[HHOOK_IPSEC_INET],
324 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
325 printf("%s: WARNING: unable to register input helper hook\n",
327 if (hhook_head_register(HHOOK_TYPE_IPSEC_OUT, AF_INET,
328 &V_ipsec_hhh_out[HHOOK_IPSEC_INET],
329 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
330 printf("%s: WARNING: unable to register output helper hook\n",
333 /* Skip initialization of globals for non-default instances. */
334 if (!IS_DEFAULT_VNET(curvnet))
337 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
339 panic("ip_init: PF_INET not found");
341 /* Initialize the entire ip_protox[] array to IPPROTO_RAW. */
342 for (i = 0; i < IPPROTO_MAX; i++)
343 ip_protox[i] = pr - inetsw;
345 * Cycle through IP protocols and put them into the appropriate place
348 for (pr = inetdomain.dom_protosw;
349 pr < inetdomain.dom_protoswNPROTOSW; pr++)
350 if (pr->pr_domain->dom_family == PF_INET &&
351 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
352 /* Be careful to only index valid IP protocols. */
353 if (pr->pr_protocol < IPPROTO_MAX)
354 ip_protox[pr->pr_protocol] = pr - inetsw;
357 netisr_register(&ip_nh);
359 netisr_register(&ip_direct_nh);
369 if ((error = pfil_head_unregister(&V_inet_pfil_hook)) != 0)
370 printf("%s: WARNING: unable to unregister pfil hook, "
371 "error %d\n", __func__, error);
373 error = hhook_head_deregister(V_ipsec_hhh_in[HHOOK_IPSEC_INET]);
375 printf("%s: WARNING: unable to deregister input helper hook "
376 "type HHOOK_TYPE_IPSEC_IN, id HHOOK_IPSEC_INET: "
377 "error %d returned\n", __func__, error);
379 error = hhook_head_deregister(V_ipsec_hhh_out[HHOOK_IPSEC_INET]);
381 printf("%s: WARNING: unable to deregister output helper hook "
382 "type HHOOK_TYPE_IPSEC_OUT, id HHOOK_IPSEC_INET: "
383 "error %d returned\n", __func__, error);
385 /* Cleanup in_ifaddr hash table; should be empty. */
386 hashdestroy(V_in_ifaddrhashtbl, M_IFADDR, V_in_ifaddrhmask);
388 /* Destroy IP reassembly queue. */
395 * IP direct input routine.
397 * This is called when reinjecting completed fragments where
398 * all of the previous checking and book-keeping has been done.
401 ip_direct_input(struct mbuf *m)
406 ip = mtod(m, struct ip *);
407 hlen = ip->ip_hl << 2;
409 IPSTAT_INC(ips_delivered);
410 (*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
416 * Ip input routine. Checksum and byte swap header. If fragmented
417 * try to reassemble. Process options. Pass to next level.
420 ip_input(struct mbuf *m)
422 struct ip *ip = NULL;
423 struct in_ifaddr *ia = NULL;
426 int checkif, hlen = 0;
427 uint16_t sum, ip_len;
428 int dchg = 0; /* dest changed after fw */
429 struct in_addr odst; /* original dst address */
433 if (m->m_flags & M_FASTFWD_OURS) {
434 m->m_flags &= ~M_FASTFWD_OURS;
435 /* Set up some basics that will be used later. */
436 ip = mtod(m, struct ip *);
437 hlen = ip->ip_hl << 2;
438 ip_len = ntohs(ip->ip_len);
442 IPSTAT_INC(ips_total);
444 if (m->m_pkthdr.len < sizeof(struct ip))
447 if (m->m_len < sizeof (struct ip) &&
448 (m = m_pullup(m, sizeof (struct ip))) == NULL) {
449 IPSTAT_INC(ips_toosmall);
452 ip = mtod(m, struct ip *);
454 if (ip->ip_v != IPVERSION) {
455 IPSTAT_INC(ips_badvers);
459 hlen = ip->ip_hl << 2;
460 if (hlen < sizeof(struct ip)) { /* minimum header length */
461 IPSTAT_INC(ips_badhlen);
464 if (hlen > m->m_len) {
465 if ((m = m_pullup(m, hlen)) == NULL) {
466 IPSTAT_INC(ips_badhlen);
469 ip = mtod(m, struct ip *);
472 IP_PROBE(receive, NULL, NULL, ip, m->m_pkthdr.rcvif, ip, NULL);
474 /* 127/8 must not appear on wire - RFC1122 */
475 ifp = m->m_pkthdr.rcvif;
476 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
477 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
478 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
479 IPSTAT_INC(ips_badaddr);
484 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
485 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
487 if (hlen == sizeof(struct ip)) {
488 sum = in_cksum_hdr(ip);
490 sum = in_cksum(m, hlen);
494 IPSTAT_INC(ips_badsum);
499 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
500 /* packet is dropped by traffic conditioner */
504 ip_len = ntohs(ip->ip_len);
506 IPSTAT_INC(ips_badlen);
511 * Check that the amount of data in the buffers
512 * is as at least much as the IP header would have us expect.
513 * Trim mbufs if longer than we expect.
514 * Drop packet if shorter than we expect.
516 if (m->m_pkthdr.len < ip_len) {
518 IPSTAT_INC(ips_tooshort);
521 if (m->m_pkthdr.len > ip_len) {
522 if (m->m_len == m->m_pkthdr.len) {
524 m->m_pkthdr.len = ip_len;
526 m_adj(m, ip_len - m->m_pkthdr.len);
529 /* Try to forward the packet, but if we fail continue */
531 /* For now we do not handle IPSEC in tryforward. */
532 if (!key_havesp(IPSEC_DIR_INBOUND) && !key_havesp(IPSEC_DIR_OUTBOUND) &&
533 (V_ipforwarding == 1))
534 if (ip_tryforward(m) == NULL)
537 * Bypass packet filtering for packets previously handled by IPsec.
539 if (ip_ipsec_filtertunnel(m))
542 if (V_ipforwarding == 1)
543 if (ip_tryforward(m) == NULL)
548 * Run through list of hooks for input packets.
550 * NB: Beware of the destination address changing (e.g.
551 * by NAT rewriting). When this happens, tell
552 * ip_forward to do the right thing.
555 /* Jump over all PFIL processing if hooks are not active. */
556 if (!PFIL_HOOKED(&V_inet_pfil_hook))
560 if (pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_IN, NULL) != 0)
562 if (m == NULL) /* consumed by filter */
565 ip = mtod(m, struct ip *);
566 dchg = (odst.s_addr != ip->ip_dst.s_addr);
567 ifp = m->m_pkthdr.rcvif;
569 if (m->m_flags & M_FASTFWD_OURS) {
570 m->m_flags &= ~M_FASTFWD_OURS;
573 if (m->m_flags & M_IP_NEXTHOP) {
574 dchg = (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL);
577 * Directly ship the packet on. This allows
578 * forwarding packets originally destined to us
579 * to some other directly connected host.
588 * Process options and, if not destined for us,
589 * ship it on. ip_dooptions returns 1 when an
590 * error was detected (causing an icmp message
591 * to be sent and the original packet to be freed).
593 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
596 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
597 * matter if it is destined to another node, or whether it is
598 * a multicast one, RSVP wants it! and prevents it from being forwarded
599 * anywhere else. Also checks if the rsvp daemon is running before
600 * grabbing the packet.
602 if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP)
606 * Check our list of addresses, to see if the packet is for us.
607 * If we don't have any addresses, assume any unicast packet
608 * we receive might be for us (and let the upper layers deal
611 if (TAILQ_EMPTY(&V_in_ifaddrhead) &&
612 (m->m_flags & (M_MCAST|M_BCAST)) == 0)
616 * Enable a consistency check between the destination address
617 * and the arrival interface for a unicast packet (the RFC 1122
618 * strong ES model) if IP forwarding is disabled and the packet
619 * is not locally generated and the packet is not subject to
622 * XXX - Checking also should be disabled if the destination
623 * address is ipnat'ed to a different interface.
625 * XXX - Checking is incompatible with IP aliases added
626 * to the loopback interface instead of the interface where
627 * the packets are received.
629 * XXX - This is the case for carp vhost IPs as well so we
630 * insert a workaround. If the packet got here, we already
631 * checked with carp_iamatch() and carp_forus().
633 checkif = V_ip_checkinterface && (V_ipforwarding == 0) &&
634 ifp != NULL && ((ifp->if_flags & IFF_LOOPBACK) == 0) &&
635 ifp->if_carp == NULL && (dchg == 0);
638 * Check for exact addresses in the hash bucket.
640 /* IN_IFADDR_RLOCK(); */
641 LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
643 * If the address matches, verify that the packet
644 * arrived via the correct interface if checking is
647 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr &&
648 (!checkif || ia->ia_ifp == ifp)) {
649 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
650 counter_u64_add(ia->ia_ifa.ifa_ibytes,
652 /* IN_IFADDR_RUNLOCK(); */
656 /* IN_IFADDR_RUNLOCK(); */
659 * Check for broadcast addresses.
661 * Only accept broadcast packets that arrive via the matching
662 * interface. Reception of forwarded directed broadcasts would
663 * be handled via ip_forward() and ether_output() with the loopback
664 * into the stack for SIMPLEX interfaces handled by ether_output().
666 if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
668 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
669 if (ifa->ifa_addr->sa_family != AF_INET)
672 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
674 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
675 counter_u64_add(ia->ia_ifa.ifa_ibytes,
677 IF_ADDR_RUNLOCK(ifp);
681 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) {
682 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
683 counter_u64_add(ia->ia_ifa.ifa_ibytes,
685 IF_ADDR_RUNLOCK(ifp);
690 IF_ADDR_RUNLOCK(ifp);
693 /* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */
694 if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) {
695 IPSTAT_INC(ips_cantforward);
699 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
702 * If we are acting as a multicast router, all
703 * incoming multicast packets are passed to the
704 * kernel-level multicast forwarding function.
705 * The packet is returned (relatively) intact; if
706 * ip_mforward() returns a non-zero value, the packet
707 * must be discarded, else it may be accepted below.
709 if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) {
710 IPSTAT_INC(ips_cantforward);
716 * The process-level routing daemon needs to receive
717 * all multicast IGMP packets, whether or not this
718 * host belongs to their destination groups.
720 if (ip->ip_p == IPPROTO_IGMP)
722 IPSTAT_INC(ips_forward);
725 * Assume the packet is for us, to avoid prematurely taking
726 * a lock on the in_multi hash. Protocols must perform
727 * their own filtering and update statistics accordingly.
731 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
733 if (ip->ip_dst.s_addr == INADDR_ANY)
737 * Not for us; forward if possible and desirable.
739 if (V_ipforwarding == 0) {
740 IPSTAT_INC(ips_cantforward);
750 * IPSTEALTH: Process non-routing options only
751 * if the packet is destined for us.
753 if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1))
755 #endif /* IPSTEALTH */
758 * Attempt reassembly; if it succeeds, proceed.
759 * ip_reass() will return a different mbuf.
761 if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) {
762 /* XXXGL: shouldn't we save & set m_flags? */
766 ip = mtod(m, struct ip *);
767 /* Get the header length of the reassembled packet */
768 hlen = ip->ip_hl << 2;
773 * enforce IPsec policy checking if we are seeing last header.
774 * note that we do not visit this with protocols with pcb layer
775 * code - like udp/tcp/raw ip.
777 if (ip_ipsec_input(m, ip->ip_p) != 0)
782 * Switch out to protocol's input routine.
784 IPSTAT_INC(ips_delivered);
786 (*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
793 * IP timer processing;
794 * if a timer expires on a reassembly
800 VNET_ITERATOR_DECL(vnet_iter);
802 VNET_LIST_RLOCK_NOSLEEP();
803 VNET_FOREACH(vnet_iter) {
804 CURVNET_SET(vnet_iter);
808 VNET_LIST_RUNLOCK_NOSLEEP();
814 VNET_ITERATOR_DECL(vnet_iter);
816 VNET_LIST_RLOCK_NOSLEEP();
817 VNET_FOREACH(vnet_iter) {
818 CURVNET_SET(vnet_iter);
822 VNET_LIST_RUNLOCK_NOSLEEP();
826 * The protocol to be inserted into ip_protox[] must be already registered
827 * in inetsw[], either statically or through pf_proto_register().
830 ipproto_register(short ipproto)
835 if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
836 return (EPROTONOSUPPORT);
839 * The protocol slot must not be occupied by another protocol
840 * already. An index pointing to IPPROTO_RAW is unused.
842 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
844 return (EPFNOSUPPORT);
845 if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */
848 /* Find the protocol position in inetsw[] and set the index. */
849 for (pr = inetdomain.dom_protosw;
850 pr < inetdomain.dom_protoswNPROTOSW; pr++) {
851 if (pr->pr_domain->dom_family == PF_INET &&
852 pr->pr_protocol && pr->pr_protocol == ipproto) {
853 ip_protox[pr->pr_protocol] = pr - inetsw;
857 return (EPROTONOSUPPORT);
861 ipproto_unregister(short ipproto)
866 if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
867 return (EPROTONOSUPPORT);
869 /* Check if the protocol was indeed registered. */
870 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
872 return (EPFNOSUPPORT);
873 if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */
876 /* Reset the protocol slot to IPPROTO_RAW. */
877 ip_protox[ipproto] = pr - inetsw;
881 u_char inetctlerrmap[PRC_NCMDS] = {
883 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
884 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
885 EMSGSIZE, EHOSTUNREACH, 0, 0,
886 0, 0, EHOSTUNREACH, 0,
887 ENOPROTOOPT, ECONNREFUSED
891 * Forward a packet. If some error occurs return the sender
892 * an icmp packet. Note we can't always generate a meaningful
893 * icmp message because icmp doesn't have a large enough repertoire
894 * of codes and types.
896 * If not forwarding, just drop the packet. This could be confusing
897 * if ipforwarding was zero but some routing protocol was advancing
898 * us as a gateway to somewhere. However, we must let the routing
899 * protocol deal with that.
901 * The srcrt parameter indicates whether the packet is being forwarded
902 * via a source route.
905 ip_forward(struct mbuf *m, int srcrt)
907 struct ip *ip = mtod(m, struct ip *);
908 struct in_ifaddr *ia;
910 struct sockaddr_in *sin;
913 int error, type = 0, code = 0, mtu = 0;
915 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
916 IPSTAT_INC(ips_cantforward);
921 if (ip_ipsec_fwd(m) != 0) {
922 IPSTAT_INC(ips_cantforward);
930 if (ip->ip_ttl <= IPTTLDEC) {
931 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS,
939 bzero(&ro, sizeof(ro));
940 sin = (struct sockaddr_in *)&ro.ro_dst;
941 sin->sin_family = AF_INET;
942 sin->sin_len = sizeof(*sin);
943 sin->sin_addr = ip->ip_dst;
945 rtalloc_mpath_fib(&ro,
946 ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
949 in_rtalloc_ign(&ro, 0, M_GETFIB(m));
951 if (ro.ro_rt != NULL) {
952 ia = ifatoia(ro.ro_rt->rt_ifa);
953 ifa_ref(&ia->ia_ifa);
958 * 'ia' may be NULL if there is no route for this destination.
959 * In case of IPsec, Don't discard it just yet, but pass it to
960 * ip_output in case of outgoing IPsec policy.
962 if (!srcrt && ia == NULL) {
963 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
970 * Save the IP header and at most 8 bytes of the payload,
971 * in case we need to generate an ICMP message to the src.
973 * XXX this can be optimized a lot by saving the data in a local
974 * buffer on the stack (72 bytes at most), and only allocating the
975 * mbuf if really necessary. The vast majority of the packets
976 * are forwarded without having to send an ICMP back (either
977 * because unnecessary, or because rate limited), so we are
978 * really we are wasting a lot of work here.
980 * We don't use m_copy() because it might return a reference
981 * to a shared cluster. Both this function and ip_output()
982 * assume exclusive access to the IP header in `m', so any
983 * data in a cluster may change before we reach icmp_error().
985 mcopy = m_gethdr(M_NOWAIT, m->m_type);
986 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_NOWAIT)) {
988 * It's probably ok if the pkthdr dup fails (because
989 * the deep copy of the tag chain failed), but for now
990 * be conservative and just discard the copy since
991 * code below may some day want the tags.
997 mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy));
998 mcopy->m_pkthdr.len = mcopy->m_len;
999 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1005 ip->ip_ttl -= IPTTLDEC;
1011 * If forwarding packet using same interface that it came in on,
1012 * perhaps should send a redirect to sender to shortcut a hop.
1013 * Only send redirect if source is sending directly to us,
1014 * and if packet was not source routed (or has any options).
1015 * Also, don't send redirect if forwarding using a default route
1016 * or a route modified by a redirect.
1019 if (!srcrt && V_ipsendredirects &&
1020 ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) {
1025 if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1026 satosin(rt_key(rt))->sin_addr.s_addr != 0) {
1027 #define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa))
1028 u_long src = ntohl(ip->ip_src.s_addr);
1031 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1032 if (rt->rt_flags & RTF_GATEWAY)
1033 dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr;
1035 dest.s_addr = ip->ip_dst.s_addr;
1036 /* Router requirements says to only send host redirects */
1037 type = ICMP_REDIRECT;
1038 code = ICMP_REDIRECT_HOST;
1043 error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
1045 if (error == EMSGSIZE && ro.ro_rt)
1046 mtu = ro.ro_rt->rt_mtu;
1050 IPSTAT_INC(ips_cantforward);
1052 IPSTAT_INC(ips_forward);
1054 IPSTAT_INC(ips_redirectsent);
1059 ifa_free(&ia->ia_ifa);
1063 if (mcopy == NULL) {
1065 ifa_free(&ia->ia_ifa);
1071 case 0: /* forwarded, but need redirect */
1072 /* type, code set above */
1080 type = ICMP_UNREACH;
1081 code = ICMP_UNREACH_HOST;
1085 type = ICMP_UNREACH;
1086 code = ICMP_UNREACH_NEEDFRAG;
1090 * If IPsec is configured for this path,
1091 * override any possibly mtu value set by ip_output.
1093 mtu = ip_ipsec_mtu(mcopy, mtu);
1096 * If the MTU was set before make sure we are below the
1098 * If the MTU wasn't set before use the interface mtu or
1099 * fall back to the next smaller mtu step compared to the
1100 * current packet size.
1104 mtu = min(mtu, ia->ia_ifp->if_mtu);
1107 mtu = ia->ia_ifp->if_mtu;
1109 mtu = ip_next_mtu(ntohs(ip->ip_len), 0);
1111 IPSTAT_INC(ips_cantfrag);
1115 case EACCES: /* ipfw denied packet */
1118 ifa_free(&ia->ia_ifa);
1122 ifa_free(&ia->ia_ifa);
1123 icmp_error(mcopy, type, code, dest.s_addr, mtu);
1127 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1131 if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) {
1135 if (inp->inp_socket->so_options & SO_BINTIME) {
1136 *mp = sbcreatecontrol((caddr_t)&bt, sizeof(bt),
1137 SCM_BINTIME, SOL_SOCKET);
1139 mp = &(*mp)->m_next;
1141 if (inp->inp_socket->so_options & SO_TIMESTAMP) {
1144 bintime2timeval(&bt, &tv);
1145 *mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv),
1146 SCM_TIMESTAMP, SOL_SOCKET);
1148 mp = &(*mp)->m_next;
1151 if (inp->inp_flags & INP_RECVDSTADDR) {
1152 *mp = sbcreatecontrol((caddr_t)&ip->ip_dst,
1153 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1155 mp = &(*mp)->m_next;
1157 if (inp->inp_flags & INP_RECVTTL) {
1158 *mp = sbcreatecontrol((caddr_t)&ip->ip_ttl,
1159 sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
1161 mp = &(*mp)->m_next;
1165 * Moving these out of udp_input() made them even more broken
1166 * than they already were.
1168 /* options were tossed already */
1169 if (inp->inp_flags & INP_RECVOPTS) {
1170 *mp = sbcreatecontrol((caddr_t)opts_deleted_above,
1171 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1173 mp = &(*mp)->m_next;
1175 /* ip_srcroute doesn't do what we want here, need to fix */
1176 if (inp->inp_flags & INP_RECVRETOPTS) {
1177 *mp = sbcreatecontrol((caddr_t)ip_srcroute(m),
1178 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1180 mp = &(*mp)->m_next;
1183 if (inp->inp_flags & INP_RECVIF) {
1186 struct sockaddr_dl sdl;
1189 struct sockaddr_dl *sdp;
1190 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1192 if ((ifp = m->m_pkthdr.rcvif) &&
1193 ifp->if_index && ifp->if_index <= V_if_index) {
1194 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1196 * Change our mind and don't try copy.
1198 if (sdp->sdl_family != AF_LINK ||
1199 sdp->sdl_len > sizeof(sdlbuf)) {
1202 bcopy(sdp, sdl2, sdp->sdl_len);
1206 offsetof(struct sockaddr_dl, sdl_data[0]);
1207 sdl2->sdl_family = AF_LINK;
1208 sdl2->sdl_index = 0;
1209 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1211 *mp = sbcreatecontrol((caddr_t)sdl2, sdl2->sdl_len,
1212 IP_RECVIF, IPPROTO_IP);
1214 mp = &(*mp)->m_next;
1216 if (inp->inp_flags & INP_RECVTOS) {
1217 *mp = sbcreatecontrol((caddr_t)&ip->ip_tos,
1218 sizeof(u_char), IP_RECVTOS, IPPROTO_IP);
1220 mp = &(*mp)->m_next;
1223 if (inp->inp_flags2 & INP_RECVFLOWID) {
1224 uint32_t flowid, flow_type;
1226 flowid = m->m_pkthdr.flowid;
1227 flow_type = M_HASHTYPE_GET(m);
1230 * XXX should handle the failure of one or the
1231 * other - don't populate both?
1233 *mp = sbcreatecontrol((caddr_t) &flowid,
1234 sizeof(uint32_t), IP_FLOWID, IPPROTO_IP);
1236 mp = &(*mp)->m_next;
1237 *mp = sbcreatecontrol((caddr_t) &flow_type,
1238 sizeof(uint32_t), IP_FLOWTYPE, IPPROTO_IP);
1240 mp = &(*mp)->m_next;
1244 if (inp->inp_flags2 & INP_RECVRSSBUCKETID) {
1245 uint32_t flowid, flow_type;
1246 uint32_t rss_bucketid;
1248 flowid = m->m_pkthdr.flowid;
1249 flow_type = M_HASHTYPE_GET(m);
1251 if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) {
1252 *mp = sbcreatecontrol((caddr_t) &rss_bucketid,
1253 sizeof(uint32_t), IP_RSSBUCKETID, IPPROTO_IP);
1255 mp = &(*mp)->m_next;
1262 * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
1263 * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
1264 * locking. This code remains in ip_input.c as ip_mroute.c is optionally
1267 static VNET_DEFINE(int, ip_rsvp_on);
1268 VNET_DEFINE(struct socket *, ip_rsvpd);
1270 #define V_ip_rsvp_on VNET(ip_rsvp_on)
1273 ip_rsvp_init(struct socket *so)
1276 if (so->so_type != SOCK_RAW ||
1277 so->so_proto->pr_protocol != IPPROTO_RSVP)
1280 if (V_ip_rsvpd != NULL)
1285 * This may seem silly, but we need to be sure we don't over-increment
1286 * the RSVP counter, in case something slips up.
1288 if (!V_ip_rsvp_on) {
1302 * This may seem silly, but we need to be sure we don't over-decrement
1303 * the RSVP counter, in case something slips up.
1313 rsvp_input(struct mbuf **mp, int *offp, int proto)
1320 if (rsvp_input_p) { /* call the real one if loaded */
1322 rsvp_input_p(mp, offp, proto);
1323 return (IPPROTO_DONE);
1326 /* Can still get packets with rsvp_on = 0 if there is a local member
1327 * of the group to which the RSVP packet is addressed. But in this
1328 * case we want to throw the packet away.
1333 return (IPPROTO_DONE);
1336 if (V_ip_rsvpd != NULL) {
1338 rip_input(mp, offp, proto);
1339 return (IPPROTO_DONE);
1341 /* Drop the packet */
1343 return (IPPROTO_DONE);