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>
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/rwlock.h>
54 #include <sys/syslog.h>
55 #include <sys/sysctl.h>
59 #include <net/if_types.h>
60 #include <net/if_var.h>
61 #include <net/if_dl.h>
62 #include <net/route.h>
63 #include <net/netisr.h>
64 #include <net/rss_config.h>
67 #include <netinet/in.h>
68 #include <netinet/in_kdtrace.h>
69 #include <netinet/in_systm.h>
70 #include <netinet/in_var.h>
71 #include <netinet/ip.h>
72 #include <netinet/in_pcb.h>
73 #include <netinet/ip_var.h>
74 #include <netinet/ip_fw.h>
75 #include <netinet/ip_icmp.h>
76 #include <netinet/ip_options.h>
77 #include <machine/in_cksum.h>
78 #include <netinet/ip_carp.h>
80 #include <netinet/ip_ipsec.h>
82 #include <netinet/in_rss.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 rwlock in_ifaddr_lock;
101 RW_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,
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_m2cpuid,
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 /* Skip initialization of globals for non-default instances. */
319 if (!IS_DEFAULT_VNET(curvnet))
322 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
324 panic("ip_init: PF_INET not found");
326 /* Initialize the entire ip_protox[] array to IPPROTO_RAW. */
327 for (i = 0; i < IPPROTO_MAX; i++)
328 ip_protox[i] = pr - inetsw;
330 * Cycle through IP protocols and put them into the appropriate place
333 for (pr = inetdomain.dom_protosw;
334 pr < inetdomain.dom_protoswNPROTOSW; pr++)
335 if (pr->pr_domain->dom_family == PF_INET &&
336 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
337 /* Be careful to only index valid IP protocols. */
338 if (pr->pr_protocol < IPPROTO_MAX)
339 ip_protox[pr->pr_protocol] = pr - inetsw;
342 netisr_register(&ip_nh);
344 netisr_register(&ip_direct_nh);
354 if ((i = pfil_head_unregister(&V_inet_pfil_hook)) != 0)
355 printf("%s: WARNING: unable to unregister pfil hook, "
356 "error %d\n", __func__, i);
358 /* Cleanup in_ifaddr hash table; should be empty. */
359 hashdestroy(V_in_ifaddrhashtbl, M_IFADDR, V_in_ifaddrhmask);
361 /* Destroy IP reassembly queue. */
368 * IP direct input routine.
370 * This is called when reinjecting completed fragments where
371 * all of the previous checking and book-keeping has been done.
374 ip_direct_input(struct mbuf *m)
379 ip = mtod(m, struct ip *);
380 hlen = ip->ip_hl << 2;
382 IPSTAT_INC(ips_delivered);
383 (*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
389 * Ip input routine. Checksum and byte swap header. If fragmented
390 * try to reassemble. Process options. Pass to next level.
393 ip_input(struct mbuf *m)
395 struct ip *ip = NULL;
396 struct in_ifaddr *ia = NULL;
399 int checkif, hlen = 0;
400 uint16_t sum, ip_len;
401 int dchg = 0; /* dest changed after fw */
402 struct in_addr odst; /* original dst address */
406 if (m->m_flags & M_FASTFWD_OURS) {
407 m->m_flags &= ~M_FASTFWD_OURS;
408 /* Set up some basics that will be used later. */
409 ip = mtod(m, struct ip *);
410 hlen = ip->ip_hl << 2;
411 ip_len = ntohs(ip->ip_len);
415 IPSTAT_INC(ips_total);
417 if (m->m_pkthdr.len < sizeof(struct ip))
420 if (m->m_len < sizeof (struct ip) &&
421 (m = m_pullup(m, sizeof (struct ip))) == NULL) {
422 IPSTAT_INC(ips_toosmall);
425 ip = mtod(m, struct ip *);
427 if (ip->ip_v != IPVERSION) {
428 IPSTAT_INC(ips_badvers);
432 hlen = ip->ip_hl << 2;
433 if (hlen < sizeof(struct ip)) { /* minimum header length */
434 IPSTAT_INC(ips_badhlen);
437 if (hlen > m->m_len) {
438 if ((m = m_pullup(m, hlen)) == NULL) {
439 IPSTAT_INC(ips_badhlen);
442 ip = mtod(m, struct ip *);
445 IP_PROBE(receive, NULL, NULL, ip, m->m_pkthdr.rcvif, ip, NULL);
447 /* 127/8 must not appear on wire - RFC1122 */
448 ifp = m->m_pkthdr.rcvif;
449 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
450 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
451 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
452 IPSTAT_INC(ips_badaddr);
457 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
458 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
460 if (hlen == sizeof(struct ip)) {
461 sum = in_cksum_hdr(ip);
463 sum = in_cksum(m, hlen);
467 IPSTAT_INC(ips_badsum);
472 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
473 /* packet is dropped by traffic conditioner */
477 ip_len = ntohs(ip->ip_len);
479 IPSTAT_INC(ips_badlen);
484 * Check that the amount of data in the buffers
485 * is as at least much as the IP header would have us expect.
486 * Trim mbufs if longer than we expect.
487 * Drop packet if shorter than we expect.
489 if (m->m_pkthdr.len < ip_len) {
491 IPSTAT_INC(ips_tooshort);
494 if (m->m_pkthdr.len > ip_len) {
495 if (m->m_len == m->m_pkthdr.len) {
497 m->m_pkthdr.len = ip_len;
499 m_adj(m, ip_len - m->m_pkthdr.len);
504 * Bypass packet filtering for packets previously handled by IPsec.
506 if (ip_ipsec_filtertunnel(m))
511 * Run through list of hooks for input packets.
513 * NB: Beware of the destination address changing (e.g.
514 * by NAT rewriting). When this happens, tell
515 * ip_forward to do the right thing.
518 /* Jump over all PFIL processing if hooks are not active. */
519 if (!PFIL_HOOKED(&V_inet_pfil_hook))
523 if (pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_IN, NULL) != 0)
525 if (m == NULL) /* consumed by filter */
528 ip = mtod(m, struct ip *);
529 dchg = (odst.s_addr != ip->ip_dst.s_addr);
530 ifp = m->m_pkthdr.rcvif;
532 if (m->m_flags & M_FASTFWD_OURS) {
533 m->m_flags &= ~M_FASTFWD_OURS;
536 if (m->m_flags & M_IP_NEXTHOP) {
537 dchg = (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL);
540 * Directly ship the packet on. This allows
541 * forwarding packets originally destined to us
542 * to some other directly connected host.
551 * Process options and, if not destined for us,
552 * ship it on. ip_dooptions returns 1 when an
553 * error was detected (causing an icmp message
554 * to be sent and the original packet to be freed).
556 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
559 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
560 * matter if it is destined to another node, or whether it is
561 * a multicast one, RSVP wants it! and prevents it from being forwarded
562 * anywhere else. Also checks if the rsvp daemon is running before
563 * grabbing the packet.
565 if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP)
569 * Check our list of addresses, to see if the packet is for us.
570 * If we don't have any addresses, assume any unicast packet
571 * we receive might be for us (and let the upper layers deal
574 if (TAILQ_EMPTY(&V_in_ifaddrhead) &&
575 (m->m_flags & (M_MCAST|M_BCAST)) == 0)
579 * Enable a consistency check between the destination address
580 * and the arrival interface for a unicast packet (the RFC 1122
581 * strong ES model) if IP forwarding is disabled and the packet
582 * is not locally generated and the packet is not subject to
585 * XXX - Checking also should be disabled if the destination
586 * address is ipnat'ed to a different interface.
588 * XXX - Checking is incompatible with IP aliases added
589 * to the loopback interface instead of the interface where
590 * the packets are received.
592 * XXX - This is the case for carp vhost IPs as well so we
593 * insert a workaround. If the packet got here, we already
594 * checked with carp_iamatch() and carp_forus().
596 checkif = V_ip_checkinterface && (V_ipforwarding == 0) &&
597 ifp != NULL && ((ifp->if_flags & IFF_LOOPBACK) == 0) &&
598 ifp->if_carp == NULL && (dchg == 0);
601 * Check for exact addresses in the hash bucket.
603 /* IN_IFADDR_RLOCK(); */
604 LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
606 * If the address matches, verify that the packet
607 * arrived via the correct interface if checking is
610 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr &&
611 (!checkif || ia->ia_ifp == ifp)) {
612 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
613 counter_u64_add(ia->ia_ifa.ifa_ibytes,
615 /* IN_IFADDR_RUNLOCK(); */
619 /* IN_IFADDR_RUNLOCK(); */
622 * Check for broadcast addresses.
624 * Only accept broadcast packets that arrive via the matching
625 * interface. Reception of forwarded directed broadcasts would
626 * be handled via ip_forward() and ether_output() with the loopback
627 * into the stack for SIMPLEX interfaces handled by ether_output().
629 if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
631 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
632 if (ifa->ifa_addr->sa_family != AF_INET)
635 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
637 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
638 counter_u64_add(ia->ia_ifa.ifa_ibytes,
640 IF_ADDR_RUNLOCK(ifp);
644 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) {
645 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
646 counter_u64_add(ia->ia_ifa.ifa_ibytes,
648 IF_ADDR_RUNLOCK(ifp);
653 IF_ADDR_RUNLOCK(ifp);
656 /* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */
657 if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) {
658 IPSTAT_INC(ips_cantforward);
662 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
665 * If we are acting as a multicast router, all
666 * incoming multicast packets are passed to the
667 * kernel-level multicast forwarding function.
668 * The packet is returned (relatively) intact; if
669 * ip_mforward() returns a non-zero value, the packet
670 * must be discarded, else it may be accepted below.
672 if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) {
673 IPSTAT_INC(ips_cantforward);
679 * The process-level routing daemon needs to receive
680 * all multicast IGMP packets, whether or not this
681 * host belongs to their destination groups.
683 if (ip->ip_p == IPPROTO_IGMP)
685 IPSTAT_INC(ips_forward);
688 * Assume the packet is for us, to avoid prematurely taking
689 * a lock on the in_multi hash. Protocols must perform
690 * their own filtering and update statistics accordingly.
694 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
696 if (ip->ip_dst.s_addr == INADDR_ANY)
700 * Not for us; forward if possible and desirable.
702 if (V_ipforwarding == 0) {
703 IPSTAT_INC(ips_cantforward);
713 * IPSTEALTH: Process non-routing options only
714 * if the packet is destined for us.
716 if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1))
718 #endif /* IPSTEALTH */
721 * Attempt reassembly; if it succeeds, proceed.
722 * ip_reass() will return a different mbuf.
724 if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) {
725 /* XXXGL: shouldn't we save & set m_flags? */
729 ip = mtod(m, struct ip *);
730 /* Get the header length of the reassembled packet */
731 hlen = ip->ip_hl << 2;
736 * enforce IPsec policy checking if we are seeing last header.
737 * note that we do not visit this with protocols with pcb layer
738 * code - like udp/tcp/raw ip.
740 if (ip_ipsec_input(m, ip->ip_p) != 0)
745 * Switch out to protocol's input routine.
747 IPSTAT_INC(ips_delivered);
749 (*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
756 * IP timer processing;
757 * if a timer expires on a reassembly
763 VNET_ITERATOR_DECL(vnet_iter);
765 VNET_LIST_RLOCK_NOSLEEP();
766 VNET_FOREACH(vnet_iter) {
767 CURVNET_SET(vnet_iter);
771 VNET_LIST_RUNLOCK_NOSLEEP();
777 VNET_ITERATOR_DECL(vnet_iter);
779 VNET_LIST_RLOCK_NOSLEEP();
780 VNET_FOREACH(vnet_iter) {
781 CURVNET_SET(vnet_iter);
785 VNET_LIST_RUNLOCK_NOSLEEP();
789 * The protocol to be inserted into ip_protox[] must be already registered
790 * in inetsw[], either statically or through pf_proto_register().
793 ipproto_register(short ipproto)
798 if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
799 return (EPROTONOSUPPORT);
802 * The protocol slot must not be occupied by another protocol
803 * already. An index pointing to IPPROTO_RAW is unused.
805 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
807 return (EPFNOSUPPORT);
808 if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */
811 /* Find the protocol position in inetsw[] and set the index. */
812 for (pr = inetdomain.dom_protosw;
813 pr < inetdomain.dom_protoswNPROTOSW; pr++) {
814 if (pr->pr_domain->dom_family == PF_INET &&
815 pr->pr_protocol && pr->pr_protocol == ipproto) {
816 ip_protox[pr->pr_protocol] = pr - inetsw;
820 return (EPROTONOSUPPORT);
824 ipproto_unregister(short ipproto)
829 if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
830 return (EPROTONOSUPPORT);
832 /* Check if the protocol was indeed registered. */
833 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
835 return (EPFNOSUPPORT);
836 if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */
839 /* Reset the protocol slot to IPPROTO_RAW. */
840 ip_protox[ipproto] = pr - inetsw;
845 * Given address of next destination (final or next hop), return (referenced)
846 * internet address info of interface to be used to get there.
849 ip_rtaddr(struct in_addr dst, u_int fibnum)
852 struct sockaddr_in *sin;
853 struct in_ifaddr *ia;
855 bzero(&sro, sizeof(sro));
856 sin = (struct sockaddr_in *)&sro.ro_dst;
857 sin->sin_family = AF_INET;
858 sin->sin_len = sizeof(*sin);
860 in_rtalloc_ign(&sro, 0, fibnum);
862 if (sro.ro_rt == NULL)
865 ia = ifatoia(sro.ro_rt->rt_ifa);
866 ifa_ref(&ia->ia_ifa);
871 u_char inetctlerrmap[PRC_NCMDS] = {
873 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
874 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
875 EMSGSIZE, EHOSTUNREACH, 0, 0,
876 0, 0, EHOSTUNREACH, 0,
877 ENOPROTOOPT, ECONNREFUSED
881 * Forward a packet. If some error occurs return the sender
882 * an icmp packet. Note we can't always generate a meaningful
883 * icmp message because icmp doesn't have a large enough repertoire
884 * of codes and types.
886 * If not forwarding, just drop the packet. This could be confusing
887 * if ipforwarding was zero but some routing protocol was advancing
888 * us as a gateway to somewhere. However, we must let the routing
889 * protocol deal with that.
891 * The srcrt parameter indicates whether the packet is being forwarded
892 * via a source route.
895 ip_forward(struct mbuf *m, int srcrt)
897 struct ip *ip = mtod(m, struct ip *);
898 struct in_ifaddr *ia;
900 struct sockaddr_in *sin;
903 int error, type = 0, code = 0, mtu = 0;
905 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
906 IPSTAT_INC(ips_cantforward);
911 if (ip_ipsec_fwd(m) != 0) {
912 IPSTAT_INC(ips_cantforward);
920 if (ip->ip_ttl <= IPTTLDEC) {
921 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS,
929 bzero(&ro, sizeof(ro));
930 sin = (struct sockaddr_in *)&ro.ro_dst;
931 sin->sin_family = AF_INET;
932 sin->sin_len = sizeof(*sin);
933 sin->sin_addr = ip->ip_dst;
935 rtalloc_mpath_fib(&ro,
936 ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
939 in_rtalloc_ign(&ro, 0, M_GETFIB(m));
941 if (ro.ro_rt != NULL) {
942 ia = ifatoia(ro.ro_rt->rt_ifa);
943 ifa_ref(&ia->ia_ifa);
948 * 'ia' may be NULL if there is no route for this destination.
949 * In case of IPsec, Don't discard it just yet, but pass it to
950 * ip_output in case of outgoing IPsec policy.
952 if (!srcrt && ia == NULL) {
953 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
960 * Save the IP header and at most 8 bytes of the payload,
961 * in case we need to generate an ICMP message to the src.
963 * XXX this can be optimized a lot by saving the data in a local
964 * buffer on the stack (72 bytes at most), and only allocating the
965 * mbuf if really necessary. The vast majority of the packets
966 * are forwarded without having to send an ICMP back (either
967 * because unnecessary, or because rate limited), so we are
968 * really we are wasting a lot of work here.
970 * We don't use m_copy() because it might return a reference
971 * to a shared cluster. Both this function and ip_output()
972 * assume exclusive access to the IP header in `m', so any
973 * data in a cluster may change before we reach icmp_error().
975 mcopy = m_gethdr(M_NOWAIT, m->m_type);
976 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_NOWAIT)) {
978 * It's probably ok if the pkthdr dup fails (because
979 * the deep copy of the tag chain failed), but for now
980 * be conservative and just discard the copy since
981 * code below may some day want the tags.
987 mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy));
988 mcopy->m_pkthdr.len = mcopy->m_len;
989 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
995 ip->ip_ttl -= IPTTLDEC;
1001 * If forwarding packet using same interface that it came in on,
1002 * perhaps should send a redirect to sender to shortcut a hop.
1003 * Only send redirect if source is sending directly to us,
1004 * and if packet was not source routed (or has any options).
1005 * Also, don't send redirect if forwarding using a default route
1006 * or a route modified by a redirect.
1009 if (!srcrt && V_ipsendredirects &&
1010 ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) {
1015 if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1016 satosin(rt_key(rt))->sin_addr.s_addr != 0) {
1017 #define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa))
1018 u_long src = ntohl(ip->ip_src.s_addr);
1021 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1022 if (rt->rt_flags & RTF_GATEWAY)
1023 dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr;
1025 dest.s_addr = ip->ip_dst.s_addr;
1026 /* Router requirements says to only send host redirects */
1027 type = ICMP_REDIRECT;
1028 code = ICMP_REDIRECT_HOST;
1033 error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
1035 if (error == EMSGSIZE && ro.ro_rt)
1036 mtu = ro.ro_rt->rt_mtu;
1040 IPSTAT_INC(ips_cantforward);
1042 IPSTAT_INC(ips_forward);
1044 IPSTAT_INC(ips_redirectsent);
1049 ifa_free(&ia->ia_ifa);
1053 if (mcopy == NULL) {
1055 ifa_free(&ia->ia_ifa);
1061 case 0: /* forwarded, but need redirect */
1062 /* type, code set above */
1070 type = ICMP_UNREACH;
1071 code = ICMP_UNREACH_HOST;
1075 type = ICMP_UNREACH;
1076 code = ICMP_UNREACH_NEEDFRAG;
1080 * If IPsec is configured for this path,
1081 * override any possibly mtu value set by ip_output.
1083 mtu = ip_ipsec_mtu(mcopy, mtu);
1086 * If the MTU was set before make sure we are below the
1088 * If the MTU wasn't set before use the interface mtu or
1089 * fall back to the next smaller mtu step compared to the
1090 * current packet size.
1094 mtu = min(mtu, ia->ia_ifp->if_mtu);
1097 mtu = ia->ia_ifp->if_mtu;
1099 mtu = ip_next_mtu(ntohs(ip->ip_len), 0);
1101 IPSTAT_INC(ips_cantfrag);
1105 case EACCES: /* ipfw denied packet */
1108 ifa_free(&ia->ia_ifa);
1112 ifa_free(&ia->ia_ifa);
1113 icmp_error(mcopy, type, code, dest.s_addr, mtu);
1117 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1121 if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) {
1125 if (inp->inp_socket->so_options & SO_BINTIME) {
1126 *mp = sbcreatecontrol((caddr_t)&bt, sizeof(bt),
1127 SCM_BINTIME, SOL_SOCKET);
1129 mp = &(*mp)->m_next;
1131 if (inp->inp_socket->so_options & SO_TIMESTAMP) {
1134 bintime2timeval(&bt, &tv);
1135 *mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv),
1136 SCM_TIMESTAMP, SOL_SOCKET);
1138 mp = &(*mp)->m_next;
1141 if (inp->inp_flags & INP_RECVDSTADDR) {
1142 *mp = sbcreatecontrol((caddr_t)&ip->ip_dst,
1143 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1145 mp = &(*mp)->m_next;
1147 if (inp->inp_flags & INP_RECVTTL) {
1148 *mp = sbcreatecontrol((caddr_t)&ip->ip_ttl,
1149 sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
1151 mp = &(*mp)->m_next;
1155 * Moving these out of udp_input() made them even more broken
1156 * than they already were.
1158 /* options were tossed already */
1159 if (inp->inp_flags & INP_RECVOPTS) {
1160 *mp = sbcreatecontrol((caddr_t)opts_deleted_above,
1161 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1163 mp = &(*mp)->m_next;
1165 /* ip_srcroute doesn't do what we want here, need to fix */
1166 if (inp->inp_flags & INP_RECVRETOPTS) {
1167 *mp = sbcreatecontrol((caddr_t)ip_srcroute(m),
1168 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1170 mp = &(*mp)->m_next;
1173 if (inp->inp_flags & INP_RECVIF) {
1176 struct sockaddr_dl sdl;
1179 struct sockaddr_dl *sdp;
1180 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1182 if ((ifp = m->m_pkthdr.rcvif) &&
1183 ifp->if_index && ifp->if_index <= V_if_index) {
1184 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1186 * Change our mind and don't try copy.
1188 if (sdp->sdl_family != AF_LINK ||
1189 sdp->sdl_len > sizeof(sdlbuf)) {
1192 bcopy(sdp, sdl2, sdp->sdl_len);
1196 offsetof(struct sockaddr_dl, sdl_data[0]);
1197 sdl2->sdl_family = AF_LINK;
1198 sdl2->sdl_index = 0;
1199 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1201 *mp = sbcreatecontrol((caddr_t)sdl2, sdl2->sdl_len,
1202 IP_RECVIF, IPPROTO_IP);
1204 mp = &(*mp)->m_next;
1206 if (inp->inp_flags & INP_RECVTOS) {
1207 *mp = sbcreatecontrol((caddr_t)&ip->ip_tos,
1208 sizeof(u_char), IP_RECVTOS, IPPROTO_IP);
1210 mp = &(*mp)->m_next;
1213 if (inp->inp_flags2 & INP_RECVFLOWID) {
1214 uint32_t flowid, flow_type;
1216 flowid = m->m_pkthdr.flowid;
1217 flow_type = M_HASHTYPE_GET(m);
1220 * XXX should handle the failure of one or the
1221 * other - don't populate both?
1223 *mp = sbcreatecontrol((caddr_t) &flowid,
1224 sizeof(uint32_t), IP_FLOWID, IPPROTO_IP);
1226 mp = &(*mp)->m_next;
1227 *mp = sbcreatecontrol((caddr_t) &flow_type,
1228 sizeof(uint32_t), IP_FLOWTYPE, IPPROTO_IP);
1230 mp = &(*mp)->m_next;
1234 if (inp->inp_flags2 & INP_RECVRSSBUCKETID) {
1235 uint32_t flowid, flow_type;
1236 uint32_t rss_bucketid;
1238 flowid = m->m_pkthdr.flowid;
1239 flow_type = M_HASHTYPE_GET(m);
1241 if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) {
1242 *mp = sbcreatecontrol((caddr_t) &rss_bucketid,
1243 sizeof(uint32_t), IP_RSSBUCKETID, IPPROTO_IP);
1245 mp = &(*mp)->m_next;
1252 * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
1253 * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
1254 * locking. This code remains in ip_input.c as ip_mroute.c is optionally
1257 static VNET_DEFINE(int, ip_rsvp_on);
1258 VNET_DEFINE(struct socket *, ip_rsvpd);
1260 #define V_ip_rsvp_on VNET(ip_rsvp_on)
1263 ip_rsvp_init(struct socket *so)
1266 if (so->so_type != SOCK_RAW ||
1267 so->so_proto->pr_protocol != IPPROTO_RSVP)
1270 if (V_ip_rsvpd != NULL)
1275 * This may seem silly, but we need to be sure we don't over-increment
1276 * the RSVP counter, in case something slips up.
1278 if (!V_ip_rsvp_on) {
1292 * This may seem silly, but we need to be sure we don't over-decrement
1293 * the RSVP counter, in case something slips up.
1303 rsvp_input(struct mbuf **mp, int *offp, int proto)
1310 if (rsvp_input_p) { /* call the real one if loaded */
1312 rsvp_input_p(mp, offp, proto);
1313 return (IPPROTO_DONE);
1316 /* Can still get packets with rsvp_on = 0 if there is a local member
1317 * of the group to which the RSVP packet is addressed. But in this
1318 * case we want to throw the packet away.
1323 return (IPPROTO_DONE);
1326 if (V_ip_rsvpd != NULL) {
1328 rip_input(mp, offp, proto);
1329 return (IPPROTO_DONE);
1331 /* Drop the packet */
1333 return (IPPROTO_DONE);