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/rmlock.h>
53 #include <sys/rwlock.h>
55 #include <sys/syslog.h>
56 #include <sys/sysctl.h>
60 #include <net/if_types.h>
61 #include <net/if_var.h>
62 #include <net/if_dl.h>
63 #include <net/route.h>
64 #include <net/netisr.h>
65 #include <net/rss_config.h>
68 #include <netinet/in.h>
69 #include <netinet/in_kdtrace.h>
70 #include <netinet/in_systm.h>
71 #include <netinet/in_var.h>
72 #include <netinet/ip.h>
73 #include <netinet/in_pcb.h>
74 #include <netinet/ip_var.h>
75 #include <netinet/ip_fw.h>
76 #include <netinet/ip_icmp.h>
77 #include <netinet/ip_options.h>
78 #include <machine/in_cksum.h>
79 #include <netinet/ip_carp.h>
81 #include <netinet/ip_ipsec.h>
83 #include <netinet/in_rss.h>
85 #include <sys/socketvar.h>
87 #include <security/mac/mac_framework.h>
90 CTASSERT(sizeof(struct ip) == 20);
93 /* IP reassembly functions are defined in ip_reass.c. */
94 extern void ipreass_init(void);
95 extern void ipreass_drain(void);
96 extern void ipreass_slowtimo(void);
98 extern void ipreass_destroy(void);
101 struct rmlock in_ifaddr_lock;
102 RM_SYSINIT(in_ifaddr_lock, &in_ifaddr_lock, "in_ifaddr_lock");
104 VNET_DEFINE(int, rsvp_on);
106 VNET_DEFINE(int, ipforwarding);
107 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_VNET | CTLFLAG_RW,
108 &VNET_NAME(ipforwarding), 0,
109 "Enable IP forwarding between interfaces");
111 static VNET_DEFINE(int, ipsendredirects) = 1; /* XXX */
112 #define V_ipsendredirects VNET(ipsendredirects)
113 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_VNET | CTLFLAG_RW,
114 &VNET_NAME(ipsendredirects), 0,
115 "Enable sending IP redirects");
118 * XXX - Setting ip_checkinterface mostly implements the receive side of
119 * the Strong ES model described in RFC 1122, but since the routing table
120 * and transmit implementation do not implement the Strong ES model,
121 * setting this to 1 results in an odd hybrid.
123 * XXX - ip_checkinterface currently must be disabled if you use ipnat
124 * to translate the destination address to another local interface.
126 * XXX - ip_checkinterface must be disabled if you add IP aliases
127 * to the loopback interface instead of the interface where the
128 * packets for those addresses are received.
130 static VNET_DEFINE(int, ip_checkinterface);
131 #define V_ip_checkinterface VNET(ip_checkinterface)
132 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_VNET | CTLFLAG_RW,
133 &VNET_NAME(ip_checkinterface), 0,
134 "Verify packet arrives on correct interface");
136 VNET_DEFINE(struct pfil_head, inet_pfil_hook); /* Packet filter hooks */
138 static struct netisr_handler ip_nh = {
140 .nh_handler = ip_input,
141 .nh_proto = NETISR_IP,
143 .nh_m2cpuid = rss_soft_m2cpuid,
144 .nh_policy = NETISR_POLICY_CPU,
145 .nh_dispatch = NETISR_DISPATCH_HYBRID,
147 .nh_policy = NETISR_POLICY_FLOW,
153 * Directly dispatched frames are currently assumed
154 * to have a flowid already calculated.
156 * It should likely have something that assert it
157 * actually has valid flow details.
159 static struct netisr_handler ip_direct_nh = {
160 .nh_name = "ip_direct",
161 .nh_handler = ip_direct_input,
162 .nh_proto = NETISR_IP_DIRECT,
163 .nh_m2cpuid = rss_m2cpuid,
164 .nh_policy = NETISR_POLICY_CPU,
165 .nh_dispatch = NETISR_DISPATCH_HYBRID,
169 extern struct domain inetdomain;
170 extern struct protosw inetsw[];
171 u_char ip_protox[IPPROTO_MAX];
172 VNET_DEFINE(struct in_ifaddrhead, in_ifaddrhead); /* first inet address */
173 VNET_DEFINE(struct in_ifaddrhashhead *, in_ifaddrhashtbl); /* inet addr hash table */
174 VNET_DEFINE(u_long, in_ifaddrhmask); /* mask for hash table */
177 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
178 &ip_mtu, 0, "Default MTU");
182 VNET_DEFINE(int, ipstealth);
183 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_VNET | CTLFLAG_RW,
184 &VNET_NAME(ipstealth), 0,
185 "IP stealth mode, no TTL decrementation on forwarding");
189 * IP statistics are stored in the "array" of counter(9)s.
191 VNET_PCPUSTAT_DEFINE(struct ipstat, ipstat);
192 VNET_PCPUSTAT_SYSINIT(ipstat);
193 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, IPCTL_STATS, stats, struct ipstat, ipstat,
194 "IP statistics (struct ipstat, netinet/ip_var.h)");
197 VNET_PCPUSTAT_SYSUNINIT(ipstat);
201 * Kernel module interface for updating ipstat. The argument is an index
202 * into ipstat treated as an array.
205 kmod_ipstat_inc(int statnum)
208 counter_u64_add(VNET(ipstat)[statnum], 1);
212 kmod_ipstat_dec(int statnum)
215 counter_u64_add(VNET(ipstat)[statnum], -1);
219 sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS)
223 netisr_getqlimit(&ip_nh, &qlimit);
224 error = sysctl_handle_int(oidp, &qlimit, 0, req);
225 if (error || !req->newptr)
229 return (netisr_setqlimit(&ip_nh, qlimit));
231 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen,
232 CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet_intr_queue_maxlen, "I",
233 "Maximum size of the IP input queue");
236 sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS)
238 u_int64_t qdrops_long;
241 netisr_getqdrops(&ip_nh, &qdrops_long);
242 qdrops = qdrops_long;
243 error = sysctl_handle_int(oidp, &qdrops, 0, req);
244 if (error || !req->newptr)
248 netisr_clearqdrops(&ip_nh);
252 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops,
253 CTLTYPE_INT|CTLFLAG_RD, 0, 0, sysctl_netinet_intr_queue_drops, "I",
254 "Number of packets dropped from the IP input queue");
258 sysctl_netinet_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS)
262 netisr_getqlimit(&ip_direct_nh, &qlimit);
263 error = sysctl_handle_int(oidp, &qlimit, 0, req);
264 if (error || !req->newptr)
268 return (netisr_setqlimit(&ip_direct_nh, qlimit));
270 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_direct_queue_maxlen,
271 CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet_intr_direct_queue_maxlen, "I",
272 "Maximum size of the IP direct input queue");
275 sysctl_netinet_intr_direct_queue_drops(SYSCTL_HANDLER_ARGS)
277 u_int64_t qdrops_long;
280 netisr_getqdrops(&ip_direct_nh, &qdrops_long);
281 qdrops = qdrops_long;
282 error = sysctl_handle_int(oidp, &qdrops, 0, req);
283 if (error || !req->newptr)
287 netisr_clearqdrops(&ip_direct_nh);
291 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_direct_queue_drops,
292 CTLTYPE_INT|CTLFLAG_RD, 0, 0, sysctl_netinet_intr_direct_queue_drops, "I",
293 "Number of packets dropped from the IP direct input queue");
297 * IP initialization: fill in IP protocol switch table.
298 * All protocols not implemented in kernel go to raw IP protocol handler.
306 TAILQ_INIT(&V_in_ifaddrhead);
307 V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask);
309 /* Initialize IP reassembly queue. */
312 /* Initialize packet filter hooks. */
313 V_inet_pfil_hook.ph_type = PFIL_TYPE_AF;
314 V_inet_pfil_hook.ph_af = AF_INET;
315 if ((i = pfil_head_register(&V_inet_pfil_hook)) != 0)
316 printf("%s: WARNING: unable to register pfil hook, "
317 "error %d\n", __func__, i);
319 /* Skip initialization of globals for non-default instances. */
320 if (!IS_DEFAULT_VNET(curvnet))
323 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
325 panic("ip_init: PF_INET not found");
327 /* Initialize the entire ip_protox[] array to IPPROTO_RAW. */
328 for (i = 0; i < IPPROTO_MAX; i++)
329 ip_protox[i] = pr - inetsw;
331 * Cycle through IP protocols and put them into the appropriate place
334 for (pr = inetdomain.dom_protosw;
335 pr < inetdomain.dom_protoswNPROTOSW; pr++)
336 if (pr->pr_domain->dom_family == PF_INET &&
337 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
338 /* Be careful to only index valid IP protocols. */
339 if (pr->pr_protocol < IPPROTO_MAX)
340 ip_protox[pr->pr_protocol] = pr - inetsw;
343 netisr_register(&ip_nh);
345 netisr_register(&ip_direct_nh);
355 if ((i = pfil_head_unregister(&V_inet_pfil_hook)) != 0)
356 printf("%s: WARNING: unable to unregister pfil hook, "
357 "error %d\n", __func__, i);
359 /* Cleanup in_ifaddr hash table; should be empty. */
360 hashdestroy(V_in_ifaddrhashtbl, M_IFADDR, V_in_ifaddrhmask);
362 /* Destroy IP reassembly queue. */
369 * IP direct input routine.
371 * This is called when reinjecting completed fragments where
372 * all of the previous checking and book-keeping has been done.
375 ip_direct_input(struct mbuf *m)
380 ip = mtod(m, struct ip *);
381 hlen = ip->ip_hl << 2;
383 IPSTAT_INC(ips_delivered);
384 (*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
390 * Ip input routine. Checksum and byte swap header. If fragmented
391 * try to reassemble. Process options. Pass to next level.
394 ip_input(struct mbuf *m)
396 struct ip *ip = NULL;
397 struct in_ifaddr *ia = NULL;
400 int checkif, hlen = 0;
401 uint16_t sum, ip_len;
402 int dchg = 0; /* dest changed after fw */
403 struct in_addr odst; /* original dst address */
407 if (m->m_flags & M_FASTFWD_OURS) {
408 m->m_flags &= ~M_FASTFWD_OURS;
409 /* Set up some basics that will be used later. */
410 ip = mtod(m, struct ip *);
411 hlen = ip->ip_hl << 2;
412 ip_len = ntohs(ip->ip_len);
416 IPSTAT_INC(ips_total);
418 if (m->m_pkthdr.len < sizeof(struct ip))
421 if (m->m_len < sizeof (struct ip) &&
422 (m = m_pullup(m, sizeof (struct ip))) == NULL) {
423 IPSTAT_INC(ips_toosmall);
426 ip = mtod(m, struct ip *);
428 if (ip->ip_v != IPVERSION) {
429 IPSTAT_INC(ips_badvers);
433 hlen = ip->ip_hl << 2;
434 if (hlen < sizeof(struct ip)) { /* minimum header length */
435 IPSTAT_INC(ips_badhlen);
438 if (hlen > m->m_len) {
439 if ((m = m_pullup(m, hlen)) == NULL) {
440 IPSTAT_INC(ips_badhlen);
443 ip = mtod(m, struct ip *);
446 IP_PROBE(receive, NULL, NULL, ip, m->m_pkthdr.rcvif, ip, NULL);
448 /* 127/8 must not appear on wire - RFC1122 */
449 ifp = m->m_pkthdr.rcvif;
450 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
451 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
452 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
453 IPSTAT_INC(ips_badaddr);
458 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
459 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
461 if (hlen == sizeof(struct ip)) {
462 sum = in_cksum_hdr(ip);
464 sum = in_cksum(m, hlen);
468 IPSTAT_INC(ips_badsum);
473 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
474 /* packet is dropped by traffic conditioner */
478 ip_len = ntohs(ip->ip_len);
480 IPSTAT_INC(ips_badlen);
485 * Check that the amount of data in the buffers
486 * is as at least much as the IP header would have us expect.
487 * Trim mbufs if longer than we expect.
488 * Drop packet if shorter than we expect.
490 if (m->m_pkthdr.len < ip_len) {
492 IPSTAT_INC(ips_tooshort);
495 if (m->m_pkthdr.len > ip_len) {
496 if (m->m_len == m->m_pkthdr.len) {
498 m->m_pkthdr.len = ip_len;
500 m_adj(m, ip_len - m->m_pkthdr.len);
505 * Bypass packet filtering for packets previously handled by IPsec.
507 if (ip_ipsec_filtertunnel(m))
512 * Run through list of hooks for input packets.
514 * NB: Beware of the destination address changing (e.g.
515 * by NAT rewriting). When this happens, tell
516 * ip_forward to do the right thing.
519 /* Jump over all PFIL processing if hooks are not active. */
520 if (!PFIL_HOOKED(&V_inet_pfil_hook))
524 if (pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_IN, NULL) != 0)
526 if (m == NULL) /* consumed by filter */
529 ip = mtod(m, struct ip *);
530 dchg = (odst.s_addr != ip->ip_dst.s_addr);
531 ifp = m->m_pkthdr.rcvif;
533 if (m->m_flags & M_FASTFWD_OURS) {
534 m->m_flags &= ~M_FASTFWD_OURS;
537 if (m->m_flags & M_IP_NEXTHOP) {
538 dchg = (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL);
541 * Directly ship the packet on. This allows
542 * forwarding packets originally destined to us
543 * to some other directly connected host.
552 * Process options and, if not destined for us,
553 * ship it on. ip_dooptions returns 1 when an
554 * error was detected (causing an icmp message
555 * to be sent and the original packet to be freed).
557 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
560 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
561 * matter if it is destined to another node, or whether it is
562 * a multicast one, RSVP wants it! and prevents it from being forwarded
563 * anywhere else. Also checks if the rsvp daemon is running before
564 * grabbing the packet.
566 if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP)
570 * Check our list of addresses, to see if the packet is for us.
571 * If we don't have any addresses, assume any unicast packet
572 * we receive might be for us (and let the upper layers deal
575 if (TAILQ_EMPTY(&V_in_ifaddrhead) &&
576 (m->m_flags & (M_MCAST|M_BCAST)) == 0)
580 * Enable a consistency check between the destination address
581 * and the arrival interface for a unicast packet (the RFC 1122
582 * strong ES model) if IP forwarding is disabled and the packet
583 * is not locally generated and the packet is not subject to
586 * XXX - Checking also should be disabled if the destination
587 * address is ipnat'ed to a different interface.
589 * XXX - Checking is incompatible with IP aliases added
590 * to the loopback interface instead of the interface where
591 * the packets are received.
593 * XXX - This is the case for carp vhost IPs as well so we
594 * insert a workaround. If the packet got here, we already
595 * checked with carp_iamatch() and carp_forus().
597 checkif = V_ip_checkinterface && (V_ipforwarding == 0) &&
598 ifp != NULL && ((ifp->if_flags & IFF_LOOPBACK) == 0) &&
599 ifp->if_carp == NULL && (dchg == 0);
602 * Check for exact addresses in the hash bucket.
604 /* IN_IFADDR_RLOCK(); */
605 LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
607 * If the address matches, verify that the packet
608 * arrived via the correct interface if checking is
611 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr &&
612 (!checkif || ia->ia_ifp == ifp)) {
613 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
614 counter_u64_add(ia->ia_ifa.ifa_ibytes,
616 /* IN_IFADDR_RUNLOCK(); */
620 /* IN_IFADDR_RUNLOCK(); */
623 * Check for broadcast addresses.
625 * Only accept broadcast packets that arrive via the matching
626 * interface. Reception of forwarded directed broadcasts would
627 * be handled via ip_forward() and ether_output() with the loopback
628 * into the stack for SIMPLEX interfaces handled by ether_output().
630 if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
632 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
633 if (ifa->ifa_addr->sa_family != AF_INET)
636 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
638 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
639 counter_u64_add(ia->ia_ifa.ifa_ibytes,
641 IF_ADDR_RUNLOCK(ifp);
645 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) {
646 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
647 counter_u64_add(ia->ia_ifa.ifa_ibytes,
649 IF_ADDR_RUNLOCK(ifp);
654 IF_ADDR_RUNLOCK(ifp);
657 /* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */
658 if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) {
659 IPSTAT_INC(ips_cantforward);
663 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
666 * If we are acting as a multicast router, all
667 * incoming multicast packets are passed to the
668 * kernel-level multicast forwarding function.
669 * The packet is returned (relatively) intact; if
670 * ip_mforward() returns a non-zero value, the packet
671 * must be discarded, else it may be accepted below.
673 if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) {
674 IPSTAT_INC(ips_cantforward);
680 * The process-level routing daemon needs to receive
681 * all multicast IGMP packets, whether or not this
682 * host belongs to their destination groups.
684 if (ip->ip_p == IPPROTO_IGMP)
686 IPSTAT_INC(ips_forward);
689 * Assume the packet is for us, to avoid prematurely taking
690 * a lock on the in_multi hash. Protocols must perform
691 * their own filtering and update statistics accordingly.
695 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
697 if (ip->ip_dst.s_addr == INADDR_ANY)
701 * Not for us; forward if possible and desirable.
703 if (V_ipforwarding == 0) {
704 IPSTAT_INC(ips_cantforward);
714 * IPSTEALTH: Process non-routing options only
715 * if the packet is destined for us.
717 if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1))
719 #endif /* IPSTEALTH */
722 * Attempt reassembly; if it succeeds, proceed.
723 * ip_reass() will return a different mbuf.
725 if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) {
726 /* XXXGL: shouldn't we save & set m_flags? */
730 ip = mtod(m, struct ip *);
731 /* Get the header length of the reassembled packet */
732 hlen = ip->ip_hl << 2;
737 * enforce IPsec policy checking if we are seeing last header.
738 * note that we do not visit this with protocols with pcb layer
739 * code - like udp/tcp/raw ip.
741 if (ip_ipsec_input(m, ip->ip_p) != 0)
746 * Switch out to protocol's input routine.
748 IPSTAT_INC(ips_delivered);
750 (*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p);
757 * IP timer processing;
758 * if a timer expires on a reassembly
764 VNET_ITERATOR_DECL(vnet_iter);
766 VNET_LIST_RLOCK_NOSLEEP();
767 VNET_FOREACH(vnet_iter) {
768 CURVNET_SET(vnet_iter);
772 VNET_LIST_RUNLOCK_NOSLEEP();
778 VNET_ITERATOR_DECL(vnet_iter);
780 VNET_LIST_RLOCK_NOSLEEP();
781 VNET_FOREACH(vnet_iter) {
782 CURVNET_SET(vnet_iter);
786 VNET_LIST_RUNLOCK_NOSLEEP();
790 * The protocol to be inserted into ip_protox[] must be already registered
791 * in inetsw[], either statically or through pf_proto_register().
794 ipproto_register(short ipproto)
799 if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
800 return (EPROTONOSUPPORT);
803 * The protocol slot must not be occupied by another protocol
804 * already. An index pointing to IPPROTO_RAW is unused.
806 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
808 return (EPFNOSUPPORT);
809 if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */
812 /* Find the protocol position in inetsw[] and set the index. */
813 for (pr = inetdomain.dom_protosw;
814 pr < inetdomain.dom_protoswNPROTOSW; pr++) {
815 if (pr->pr_domain->dom_family == PF_INET &&
816 pr->pr_protocol && pr->pr_protocol == ipproto) {
817 ip_protox[pr->pr_protocol] = pr - inetsw;
821 return (EPROTONOSUPPORT);
825 ipproto_unregister(short ipproto)
830 if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
831 return (EPROTONOSUPPORT);
833 /* Check if the protocol was indeed registered. */
834 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
836 return (EPFNOSUPPORT);
837 if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */
840 /* Reset the protocol slot to IPPROTO_RAW. */
841 ip_protox[ipproto] = pr - inetsw;
846 * Given address of next destination (final or next hop), return (referenced)
847 * internet address info of interface to be used to get there.
850 ip_rtaddr(struct in_addr dst, u_int fibnum)
853 struct sockaddr_in *sin;
854 struct in_ifaddr *ia;
856 bzero(&sro, sizeof(sro));
857 sin = (struct sockaddr_in *)&sro.ro_dst;
858 sin->sin_family = AF_INET;
859 sin->sin_len = sizeof(*sin);
861 in_rtalloc_ign(&sro, 0, fibnum);
863 if (sro.ro_rt == NULL)
866 ia = ifatoia(sro.ro_rt->rt_ifa);
867 ifa_ref(&ia->ia_ifa);
872 u_char inetctlerrmap[PRC_NCMDS] = {
874 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
875 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
876 EMSGSIZE, EHOSTUNREACH, 0, 0,
877 0, 0, EHOSTUNREACH, 0,
878 ENOPROTOOPT, ECONNREFUSED
882 * Forward a packet. If some error occurs return the sender
883 * an icmp packet. Note we can't always generate a meaningful
884 * icmp message because icmp doesn't have a large enough repertoire
885 * of codes and types.
887 * If not forwarding, just drop the packet. This could be confusing
888 * if ipforwarding was zero but some routing protocol was advancing
889 * us as a gateway to somewhere. However, we must let the routing
890 * protocol deal with that.
892 * The srcrt parameter indicates whether the packet is being forwarded
893 * via a source route.
896 ip_forward(struct mbuf *m, int srcrt)
898 struct ip *ip = mtod(m, struct ip *);
899 struct in_ifaddr *ia;
901 struct sockaddr_in *sin;
904 int error, type = 0, code = 0, mtu = 0;
906 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
907 IPSTAT_INC(ips_cantforward);
912 if (ip_ipsec_fwd(m) != 0) {
913 IPSTAT_INC(ips_cantforward);
921 if (ip->ip_ttl <= IPTTLDEC) {
922 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS,
930 bzero(&ro, sizeof(ro));
931 sin = (struct sockaddr_in *)&ro.ro_dst;
932 sin->sin_family = AF_INET;
933 sin->sin_len = sizeof(*sin);
934 sin->sin_addr = ip->ip_dst;
936 rtalloc_mpath_fib(&ro,
937 ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr),
940 in_rtalloc_ign(&ro, 0, M_GETFIB(m));
942 if (ro.ro_rt != NULL) {
943 ia = ifatoia(ro.ro_rt->rt_ifa);
944 ifa_ref(&ia->ia_ifa);
949 * 'ia' may be NULL if there is no route for this destination.
950 * In case of IPsec, Don't discard it just yet, but pass it to
951 * ip_output in case of outgoing IPsec policy.
953 if (!srcrt && ia == NULL) {
954 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
961 * Save the IP header and at most 8 bytes of the payload,
962 * in case we need to generate an ICMP message to the src.
964 * XXX this can be optimized a lot by saving the data in a local
965 * buffer on the stack (72 bytes at most), and only allocating the
966 * mbuf if really necessary. The vast majority of the packets
967 * are forwarded without having to send an ICMP back (either
968 * because unnecessary, or because rate limited), so we are
969 * really we are wasting a lot of work here.
971 * We don't use m_copy() because it might return a reference
972 * to a shared cluster. Both this function and ip_output()
973 * assume exclusive access to the IP header in `m', so any
974 * data in a cluster may change before we reach icmp_error().
976 mcopy = m_gethdr(M_NOWAIT, m->m_type);
977 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_NOWAIT)) {
979 * It's probably ok if the pkthdr dup fails (because
980 * the deep copy of the tag chain failed), but for now
981 * be conservative and just discard the copy since
982 * code below may some day want the tags.
988 mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy));
989 mcopy->m_pkthdr.len = mcopy->m_len;
990 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
996 ip->ip_ttl -= IPTTLDEC;
1002 * If forwarding packet using same interface that it came in on,
1003 * perhaps should send a redirect to sender to shortcut a hop.
1004 * Only send redirect if source is sending directly to us,
1005 * and if packet was not source routed (or has any options).
1006 * Also, don't send redirect if forwarding using a default route
1007 * or a route modified by a redirect.
1010 if (!srcrt && V_ipsendredirects &&
1011 ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) {
1016 if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1017 satosin(rt_key(rt))->sin_addr.s_addr != 0) {
1018 #define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa))
1019 u_long src = ntohl(ip->ip_src.s_addr);
1022 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1023 if (rt->rt_flags & RTF_GATEWAY)
1024 dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr;
1026 dest.s_addr = ip->ip_dst.s_addr;
1027 /* Router requirements says to only send host redirects */
1028 type = ICMP_REDIRECT;
1029 code = ICMP_REDIRECT_HOST;
1034 error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
1036 if (error == EMSGSIZE && ro.ro_rt)
1037 mtu = ro.ro_rt->rt_mtu;
1041 IPSTAT_INC(ips_cantforward);
1043 IPSTAT_INC(ips_forward);
1045 IPSTAT_INC(ips_redirectsent);
1050 ifa_free(&ia->ia_ifa);
1054 if (mcopy == NULL) {
1056 ifa_free(&ia->ia_ifa);
1062 case 0: /* forwarded, but need redirect */
1063 /* type, code set above */
1071 type = ICMP_UNREACH;
1072 code = ICMP_UNREACH_HOST;
1076 type = ICMP_UNREACH;
1077 code = ICMP_UNREACH_NEEDFRAG;
1081 * If IPsec is configured for this path,
1082 * override any possibly mtu value set by ip_output.
1084 mtu = ip_ipsec_mtu(mcopy, mtu);
1087 * If the MTU was set before make sure we are below the
1089 * If the MTU wasn't set before use the interface mtu or
1090 * fall back to the next smaller mtu step compared to the
1091 * current packet size.
1095 mtu = min(mtu, ia->ia_ifp->if_mtu);
1098 mtu = ia->ia_ifp->if_mtu;
1100 mtu = ip_next_mtu(ntohs(ip->ip_len), 0);
1102 IPSTAT_INC(ips_cantfrag);
1106 case EACCES: /* ipfw denied packet */
1109 ifa_free(&ia->ia_ifa);
1113 ifa_free(&ia->ia_ifa);
1114 icmp_error(mcopy, type, code, dest.s_addr, mtu);
1118 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1122 if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) {
1126 if (inp->inp_socket->so_options & SO_BINTIME) {
1127 *mp = sbcreatecontrol((caddr_t)&bt, sizeof(bt),
1128 SCM_BINTIME, SOL_SOCKET);
1130 mp = &(*mp)->m_next;
1132 if (inp->inp_socket->so_options & SO_TIMESTAMP) {
1135 bintime2timeval(&bt, &tv);
1136 *mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv),
1137 SCM_TIMESTAMP, SOL_SOCKET);
1139 mp = &(*mp)->m_next;
1142 if (inp->inp_flags & INP_RECVDSTADDR) {
1143 *mp = sbcreatecontrol((caddr_t)&ip->ip_dst,
1144 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1146 mp = &(*mp)->m_next;
1148 if (inp->inp_flags & INP_RECVTTL) {
1149 *mp = sbcreatecontrol((caddr_t)&ip->ip_ttl,
1150 sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
1152 mp = &(*mp)->m_next;
1156 * Moving these out of udp_input() made them even more broken
1157 * than they already were.
1159 /* options were tossed already */
1160 if (inp->inp_flags & INP_RECVOPTS) {
1161 *mp = sbcreatecontrol((caddr_t)opts_deleted_above,
1162 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1164 mp = &(*mp)->m_next;
1166 /* ip_srcroute doesn't do what we want here, need to fix */
1167 if (inp->inp_flags & INP_RECVRETOPTS) {
1168 *mp = sbcreatecontrol((caddr_t)ip_srcroute(m),
1169 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1171 mp = &(*mp)->m_next;
1174 if (inp->inp_flags & INP_RECVIF) {
1177 struct sockaddr_dl sdl;
1180 struct sockaddr_dl *sdp;
1181 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1183 if ((ifp = m->m_pkthdr.rcvif) &&
1184 ifp->if_index && ifp->if_index <= V_if_index) {
1185 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1187 * Change our mind and don't try copy.
1189 if (sdp->sdl_family != AF_LINK ||
1190 sdp->sdl_len > sizeof(sdlbuf)) {
1193 bcopy(sdp, sdl2, sdp->sdl_len);
1197 offsetof(struct sockaddr_dl, sdl_data[0]);
1198 sdl2->sdl_family = AF_LINK;
1199 sdl2->sdl_index = 0;
1200 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1202 *mp = sbcreatecontrol((caddr_t)sdl2, sdl2->sdl_len,
1203 IP_RECVIF, IPPROTO_IP);
1205 mp = &(*mp)->m_next;
1207 if (inp->inp_flags & INP_RECVTOS) {
1208 *mp = sbcreatecontrol((caddr_t)&ip->ip_tos,
1209 sizeof(u_char), IP_RECVTOS, IPPROTO_IP);
1211 mp = &(*mp)->m_next;
1214 if (inp->inp_flags2 & INP_RECVFLOWID) {
1215 uint32_t flowid, flow_type;
1217 flowid = m->m_pkthdr.flowid;
1218 flow_type = M_HASHTYPE_GET(m);
1221 * XXX should handle the failure of one or the
1222 * other - don't populate both?
1224 *mp = sbcreatecontrol((caddr_t) &flowid,
1225 sizeof(uint32_t), IP_FLOWID, IPPROTO_IP);
1227 mp = &(*mp)->m_next;
1228 *mp = sbcreatecontrol((caddr_t) &flow_type,
1229 sizeof(uint32_t), IP_FLOWTYPE, IPPROTO_IP);
1231 mp = &(*mp)->m_next;
1235 if (inp->inp_flags2 & INP_RECVRSSBUCKETID) {
1236 uint32_t flowid, flow_type;
1237 uint32_t rss_bucketid;
1239 flowid = m->m_pkthdr.flowid;
1240 flow_type = M_HASHTYPE_GET(m);
1242 if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) {
1243 *mp = sbcreatecontrol((caddr_t) &rss_bucketid,
1244 sizeof(uint32_t), IP_RSSBUCKETID, IPPROTO_IP);
1246 mp = &(*mp)->m_next;
1253 * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
1254 * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
1255 * locking. This code remains in ip_input.c as ip_mroute.c is optionally
1258 static VNET_DEFINE(int, ip_rsvp_on);
1259 VNET_DEFINE(struct socket *, ip_rsvpd);
1261 #define V_ip_rsvp_on VNET(ip_rsvp_on)
1264 ip_rsvp_init(struct socket *so)
1267 if (so->so_type != SOCK_RAW ||
1268 so->so_proto->pr_protocol != IPPROTO_RSVP)
1271 if (V_ip_rsvpd != NULL)
1276 * This may seem silly, but we need to be sure we don't over-increment
1277 * the RSVP counter, in case something slips up.
1279 if (!V_ip_rsvp_on) {
1293 * This may seem silly, but we need to be sure we don't over-decrement
1294 * the RSVP counter, in case something slips up.
1304 rsvp_input(struct mbuf **mp, int *offp, int proto)
1311 if (rsvp_input_p) { /* call the real one if loaded */
1313 rsvp_input_p(mp, offp, proto);
1314 return (IPPROTO_DONE);
1317 /* Can still get packets with rsvp_on = 0 if there is a local member
1318 * of the group to which the RSVP packet is addressed. But in this
1319 * case we want to throw the packet away.
1324 return (IPPROTO_DONE);
1327 if (V_ip_rsvpd != NULL) {
1329 rip_input(mp, offp, proto);
1330 return (IPPROTO_DONE);
1332 /* Drop the packet */
1334 return (IPPROTO_DONE);