2 * Copyright (c) 2003, 2004 Jeffrey M. Hsu. All rights reserved.
3 * Copyright (c) 2003, 2004 The DragonFly Project. All rights reserved.
5 * This code is derived from software contributed to The DragonFly Project
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. Neither the name of The DragonFly Project nor the names of its
17 * contributors may be used to endorse or promote products derived
18 * from this software without specific, prior written permission.
20 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
21 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
22 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
23 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
24 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
25 * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
26 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
27 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
28 * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
29 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
30 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
35 * Copyright (c) 1982, 1986, 1988, 1993
36 * The Regents of the University of California. All rights reserved.
38 * Redistribution and use in source and binary forms, with or without
39 * modification, are permitted provided that the following conditions
41 * 1. Redistributions of source code must retain the above copyright
42 * notice, this list of conditions and the following disclaimer.
43 * 2. Redistributions in binary form must reproduce the above copyright
44 * notice, this list of conditions and the following disclaimer in the
45 * documentation and/or other materials provided with the distribution.
46 * 3. All advertising materials mentioning features or use of this software
47 * must display the following acknowledgement:
48 * This product includes software developed by the University of
49 * California, Berkeley and its contributors.
50 * 4. Neither the name of the University nor the names of its contributors
51 * may be used to endorse or promote products derived from this software
52 * without specific prior written permission.
54 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
55 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
56 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
57 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
58 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
59 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
60 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
61 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
62 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
63 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
66 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94
67 * $FreeBSD: src/sys/netinet/ip_input.c,v 1.130.2.52 2003/03/07 07:01:28 silby Exp $
68 * $DragonFly: src/sys/netinet/ip_input.c,v 1.108 2008/09/23 11:28:49 sephe Exp $
73 #include "opt_bootp.h"
76 #include "opt_ipdivert.h"
77 #include "opt_ipfilter.h"
78 #include "opt_ipstealth.h"
79 #include "opt_ipsec.h"
81 #include <sys/param.h>
82 #include <sys/systm.h>
84 #include <sys/malloc.h>
85 #include <sys/mpipe.h>
86 #include <sys/domain.h>
87 #include <sys/protosw.h>
88 #include <sys/socket.h>
90 #include <sys/globaldata.h>
91 #include <sys/thread.h>
92 #include <sys/kernel.h>
93 #include <sys/syslog.h>
94 #include <sys/sysctl.h>
95 #include <sys/in_cksum.h>
97 #include <machine/stdarg.h>
100 #include <net/if_types.h>
101 #include <net/if_var.h>
102 #include <net/if_dl.h>
103 #include <net/pfil.h>
104 #include <net/route.h>
105 #include <net/netisr.h>
107 #include <netinet/in.h>
108 #include <netinet/in_systm.h>
109 #include <netinet/in_var.h>
110 #include <netinet/ip.h>
111 #include <netinet/in_pcb.h>
112 #include <netinet/ip_var.h>
113 #include <netinet/ip_icmp.h>
114 #include <netinet/ip_divert.h>
116 #include <sys/thread2.h>
117 #include <sys/msgport2.h>
118 #include <net/netmsg2.h>
120 #include <sys/socketvar.h>
122 #include <net/ipfw/ip_fw.h>
123 #include <net/dummynet/ip_dummynet.h>
126 #include <netinet6/ipsec.h>
127 #include <netproto/key/key.h>
131 #include <netproto/ipsec/ipsec.h>
132 #include <netproto/ipsec/key.h>
136 static int ip_rsvp_on;
137 struct socket *ip_rsvpd;
139 int ipforwarding = 0;
140 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW,
141 &ipforwarding, 0, "Enable IP forwarding between interfaces");
143 static int ipsendredirects = 1; /* XXX */
144 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW,
145 &ipsendredirects, 0, "Enable sending IP redirects");
147 int ip_defttl = IPDEFTTL;
148 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW,
149 &ip_defttl, 0, "Maximum TTL on IP packets");
151 static int ip_dosourceroute = 0;
152 SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW,
153 &ip_dosourceroute, 0, "Enable forwarding source routed IP packets");
155 static int ip_acceptsourceroute = 0;
156 SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute,
157 CTLFLAG_RW, &ip_acceptsourceroute, 0,
158 "Enable accepting source routed IP packets");
160 static int ip_keepfaith = 0;
161 SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW,
163 "Enable packet capture for FAITH IPv4->IPv6 translator daemon");
165 static int nipq = 0; /* total # of reass queues */
167 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragpackets, CTLFLAG_RW,
169 "Maximum number of IPv4 fragment reassembly queue entries");
171 static int maxfragsperpacket;
172 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW,
173 &maxfragsperpacket, 0,
174 "Maximum number of IPv4 fragments allowed per packet");
176 static int ip_sendsourcequench = 0;
177 SYSCTL_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW,
178 &ip_sendsourcequench, 0,
179 "Enable the transmission of source quench packets");
181 int ip_do_randomid = 1;
182 SYSCTL_INT(_net_inet_ip, OID_AUTO, random_id, CTLFLAG_RW,
184 "Assign random ip_id values");
186 * XXX - Setting ip_checkinterface mostly implements the receive side of
187 * the Strong ES model described in RFC 1122, but since the routing table
188 * and transmit implementation do not implement the Strong ES model,
189 * setting this to 1 results in an odd hybrid.
191 * XXX - ip_checkinterface currently must be disabled if you use ipnat
192 * to translate the destination address to another local interface.
194 * XXX - ip_checkinterface must be disabled if you add IP aliases
195 * to the loopback interface instead of the interface where the
196 * packets for those addresses are received.
198 static int ip_checkinterface = 0;
199 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW,
200 &ip_checkinterface, 0, "Verify packet arrives on correct interface");
203 static int ipprintfs = 0;
206 extern int udp_mpsafe_proto;
207 extern int tcp_mpsafe_proto;
209 extern struct domain inetdomain;
210 extern struct protosw inetsw[];
211 u_char ip_protox[IPPROTO_MAX];
212 struct in_ifaddrhead in_ifaddrheads[MAXCPU]; /* first inet address */
213 struct in_ifaddrhashhead *in_ifaddrhashtbls[MAXCPU];
214 /* inet addr hash table */
215 u_long in_ifaddrhmask; /* mask for hash table */
217 struct ip_stats ipstats_percpu[MAXCPU];
220 sysctl_ipstats(SYSCTL_HANDLER_ARGS)
224 for (cpu = 0; cpu < ncpus; ++cpu) {
225 if ((error = SYSCTL_OUT(req, &ipstats_percpu[cpu],
226 sizeof(struct ip_stats))))
228 if ((error = SYSCTL_IN(req, &ipstats_percpu[cpu],
229 sizeof(struct ip_stats))))
235 SYSCTL_PROC(_net_inet_ip, IPCTL_STATS, stats, (CTLTYPE_OPAQUE | CTLFLAG_RW),
236 0, 0, sysctl_ipstats, "S,ip_stats", "IP statistics");
238 SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW,
239 &ipstat, ip_stats, "IP statistics");
242 /* Packet reassembly stuff */
243 #define IPREASS_NHASH_LOG2 6
244 #define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2)
245 #define IPREASS_HMASK (IPREASS_NHASH - 1)
246 #define IPREASS_HASH(x,y) \
247 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK)
249 static struct ipq ipq[IPREASS_NHASH];
252 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
253 &ip_mtu, 0, "Default MTU");
257 static int ipstealth = 0;
258 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW, &ipstealth, 0, "");
260 static const int ipstealth = 0;
263 struct mbuf *(*ip_divert_p)(struct mbuf *, int, int);
265 struct pfil_head inet_pfil_hook;
268 * struct ip_srcrt_opt is used to store packet state while it travels
271 * XXX Note that the code even makes assumptions on the size and
272 * alignment of fields inside struct ip_srcrt so e.g. adding some
273 * fields will break the code. This needs to be fixed.
275 * We need to save the IP options in case a protocol wants to respond
276 * to an incoming packet over the same route if the packet got here
277 * using IP source routing. This allows connection establishment and
278 * maintenance when the remote end is on a network that is not known
282 struct in_addr dst; /* final destination */
283 char nop; /* one NOP to align */
284 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */
285 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)];
288 struct ip_srcrt_opt {
290 struct ip_srcrt ip_srcrt;
293 static MALLOC_DEFINE(M_IPQ, "ipq", "IP Fragment Management");
294 static struct malloc_pipe ipq_mpipe;
296 static void save_rte(struct mbuf *, u_char *, struct in_addr);
297 static int ip_dooptions(struct mbuf *m, int, struct sockaddr_in *);
298 static void ip_freef(struct ipq *);
299 static void ip_input_handler(struct netmsg *);
302 * IP initialization: fill in IP protocol switch table.
303 * All protocols not implemented in kernel go to raw IP protocol handler.
315 * Make sure we can handle a reasonable number of fragments but
316 * cap it at 4000 (XXX).
318 mpipe_init(&ipq_mpipe, M_IPQ, sizeof(struct ipq),
319 IFQ_MAXLEN, 4000, 0, NULL);
320 for (i = 0; i < ncpus; ++i) {
321 TAILQ_INIT(&in_ifaddrheads[i]);
322 in_ifaddrhashtbls[i] =
323 hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask);
325 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
328 for (i = 0; i < IPPROTO_MAX; i++)
329 ip_protox[i] = pr - inetsw;
330 for (pr = inetdomain.dom_protosw;
331 pr < inetdomain.dom_protoswNPROTOSW; pr++) {
332 if (pr->pr_domain->dom_family == PF_INET && pr->pr_protocol) {
333 if (pr->pr_protocol != IPPROTO_RAW)
334 ip_protox[pr->pr_protocol] = pr - inetsw;
337 switch (pr->pr_protocol) {
339 if (tcp_mpsafe_proto)
340 pr->pr_flags |= PR_MPSAFE;
344 if (udp_mpsafe_proto)
345 pr->pr_flags |= PR_MPSAFE;
351 inet_pfil_hook.ph_type = PFIL_TYPE_AF;
352 inet_pfil_hook.ph_af = AF_INET;
353 if ((i = pfil_head_register(&inet_pfil_hook)) != 0) {
354 kprintf("%s: WARNING: unable to register pfil hook, "
355 "error %d\n", __func__, i);
358 for (i = 0; i < IPREASS_NHASH; i++)
359 ipq[i].next = ipq[i].prev = &ipq[i];
361 maxnipq = nmbclusters / 32;
362 maxfragsperpacket = 16;
364 ip_id = time_second & 0xffff;
367 * Initialize IP statistics counters for each CPU.
371 for (cpu = 0; cpu < ncpus; ++cpu) {
372 bzero(&ipstats_percpu[cpu], sizeof(struct ip_stats));
375 bzero(&ipstat, sizeof(struct ip_stats));
378 netisr_register(NETISR_IP, ip_mport_in, ip_input_handler, 0);
382 * XXX watch out this one. It is perhaps used as a cache for
383 * the most recently used route ? it is cleared in in_addroute()
384 * when a new route is successfully created.
386 struct route ipforward_rt[MAXCPU];
388 /* Do transport protocol processing. */
390 transport_processing_oncpu(struct mbuf *m, int hlen, struct ip *ip)
392 const struct protosw *pr = &inetsw[ip_protox[ip->ip_p]];
395 * Switch out to protocol's input routine.
398 pr->pr_input(m, hlen, ip->ip_p);
403 transport_processing_handler(netmsg_t netmsg)
405 struct netmsg_packet *pmsg = (struct netmsg_packet *)netmsg;
409 ip = mtod(pmsg->nm_packet, struct ip *);
410 hlen = pmsg->nm_netmsg.nm_lmsg.u.ms_result;
412 transport_processing_oncpu(pmsg->nm_packet, hlen, ip);
413 /* netmsg was embedded in the mbuf, do not reply! */
417 ip_input_handler(struct netmsg *msg0)
419 struct mbuf *m = ((struct netmsg_packet *)msg0)->nm_packet;
422 /* msg0 was embedded in the mbuf, do not reply! */
426 * IP input routine. Checksum and byte swap header. If fragmented
427 * try to reassemble. Process options. Pass to next level.
430 ip_input(struct mbuf *m)
433 struct in_ifaddr *ia = NULL;
434 struct in_ifaddr_container *iac;
437 struct in_addr pkt_dst;
438 boolean_t using_srcrt = FALSE; /* forward (by PFIL_HOOKS) */
439 boolean_t needredispatch = FALSE;
440 struct in_addr odst; /* original dst address(NAT) */
442 struct sockaddr_in *next_hop = NULL;
444 struct tdb_ident *tdbi;
445 struct secpolicy *sp;
451 if (m->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) {
453 mtag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL);
454 KKASSERT(mtag != NULL);
455 next_hop = m_tag_data(mtag);
458 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
459 /* dummynet already filtered us */
460 ip = mtod(m, struct ip *);
461 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
467 /* length checks already done in ip_demux() */
468 KASSERT(m->m_len >= sizeof(ip), ("IP header not in one mbuf"));
470 ip = mtod(m, struct ip *);
472 if (IP_VHL_V(ip->ip_vhl) != IPVERSION) {
473 ipstat.ips_badvers++;
477 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
478 /* length checks already done in ip_demux() */
479 KASSERT(hlen >= sizeof(struct ip), ("IP header len too small"));
480 KASSERT(m->m_len >= hlen, ("packet shorter than IP header length"));
482 /* 127/8 must not appear on wire - RFC1122 */
483 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
484 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
485 if (!(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK)) {
486 ipstat.ips_badaddr++;
491 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
492 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
494 if (hlen == sizeof(struct ip)) {
495 sum = in_cksum_hdr(ip);
497 sum = in_cksum(m, hlen);
506 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) {
507 /* packet is dropped by traffic conditioner */
512 * Convert fields to host representation.
514 ip->ip_len = ntohs(ip->ip_len);
515 if (ip->ip_len < hlen) {
519 ip->ip_off = ntohs(ip->ip_off);
522 * Check that the amount of data in the buffers
523 * is as at least much as the IP header would have us expect.
524 * Trim mbufs if longer than we expect.
525 * Drop packet if shorter than we expect.
527 if (m->m_pkthdr.len < ip->ip_len) {
528 ipstat.ips_tooshort++;
531 if (m->m_pkthdr.len > ip->ip_len) {
532 if (m->m_len == m->m_pkthdr.len) {
533 m->m_len = ip->ip_len;
534 m->m_pkthdr.len = ip->ip_len;
536 m_adj(m, ip->ip_len - m->m_pkthdr.len);
538 #if defined(IPSEC) && !defined(IPSEC_FILTERGIF)
540 * Bypass packet filtering for packets from a tunnel (gif).
542 if (ipsec_gethist(m, NULL))
548 * Right now when no processing on packet has done
549 * and it is still fresh out of network we do our black
551 * - Firewall: deny/allow/divert
552 * - Xlate: translate packet's addr/port (NAT).
553 * - Pipe: pass pkt through dummynet.
554 * - Wrap: fake packet's addr/port <unimpl.>
555 * - Encapsulate: put it in another IP and send out. <unimp.>
560 * If we've been forwarded from the output side, then
561 * skip the firewall a second time
563 if (next_hop != NULL)
567 if (!pfil_has_hooks(&inet_pfil_hook)) {
568 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
570 * Strip dummynet tags from stranded packets
572 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
573 KKASSERT(mtag != NULL);
574 m_tag_delete(m, mtag);
575 m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED;
581 * Run through list of hooks for input packets.
583 * NB: Beware of the destination address changing (e.g.
584 * by NAT rewriting). When this happens, tell
585 * ip_forward to do the right thing.
588 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif, PFIL_IN))
590 if (m == NULL) /* consumed by filter */
592 ip = mtod(m, struct ip *);
593 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
594 using_srcrt = (odst.s_addr != ip->ip_dst.s_addr);
596 if (m->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) {
597 mtag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL);
598 KKASSERT(mtag != NULL);
599 next_hop = m_tag_data(mtag);
601 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
605 if (m->m_pkthdr.fw_flags & FW_MBUF_REDISPATCH) {
606 needredispatch = TRUE;
607 m->m_pkthdr.fw_flags &= ~FW_MBUF_REDISPATCH;
611 * Process options and, if not destined for us,
612 * ship it on. ip_dooptions returns 1 when an
613 * error was detected (causing an icmp message
614 * to be sent and the original packet to be freed).
616 if (hlen > sizeof(struct ip) && ip_dooptions(m, 0, next_hop))
619 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
620 * matter if it is destined to another node, or whether it is
621 * a multicast one, RSVP wants it! and prevents it from being forwarded
622 * anywhere else. Also checks if the rsvp daemon is running before
623 * grabbing the packet.
625 if (rsvp_on && ip->ip_p == IPPROTO_RSVP)
629 * Check our list of addresses, to see if the packet is for us.
630 * If we don't have any addresses, assume any unicast packet
631 * we receive might be for us (and let the upper layers deal
634 if (TAILQ_EMPTY(&in_ifaddrheads[mycpuid]) &&
635 !(m->m_flags & (M_MCAST | M_BCAST)))
639 * Cache the destination address of the packet; this may be
640 * changed by use of 'ipfw fwd'.
642 pkt_dst = next_hop ? next_hop->sin_addr : ip->ip_dst;
645 * Enable a consistency check between the destination address
646 * and the arrival interface for a unicast packet (the RFC 1122
647 * strong ES model) if IP forwarding is disabled and the packet
648 * is not locally generated and the packet is not subject to
651 * XXX - Checking also should be disabled if the destination
652 * address is ipnat'ed to a different interface.
654 * XXX - Checking is incompatible with IP aliases added
655 * to the loopback interface instead of the interface where
656 * the packets are received.
658 checkif = ip_checkinterface &&
660 m->m_pkthdr.rcvif != NULL &&
661 !(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) &&
665 * Check for exact addresses in the hash bucket.
667 LIST_FOREACH(iac, INADDR_HASH(pkt_dst.s_addr), ia_hash) {
671 * If the address matches, verify that the packet
672 * arrived via the correct interface if checking is
675 if (IA_SIN(ia)->sin_addr.s_addr == pkt_dst.s_addr &&
676 (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif))
682 * Check for broadcast addresses.
684 * Only accept broadcast packets that arrive via the matching
685 * interface. Reception of forwarded directed broadcasts would
686 * be handled via ip_forward() and ether_output() with the loopback
687 * into the stack for SIMPLEX interfaces handled by ether_output().
689 if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
690 struct ifaddr_container *ifac;
692 TAILQ_FOREACH(ifac, &m->m_pkthdr.rcvif->if_addrheads[mycpuid],
694 struct ifaddr *ifa = ifac->ifa;
696 if (ifa->ifa_addr == NULL) /* shutdown/startup race */
698 if (ifa->ifa_addr->sa_family != AF_INET)
701 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
704 if (ia->ia_netbroadcast.s_addr == pkt_dst.s_addr)
707 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY)
712 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
713 struct in_multi *inm;
715 if (ip_mrouter != NULL) {
717 * If we are acting as a multicast router, all
718 * incoming multicast packets are passed to the
719 * kernel-level multicast forwarding function.
720 * The packet is returned (relatively) intact; if
721 * ip_mforward() returns a non-zero value, the packet
722 * must be discarded, else it may be accepted below.
724 if (ip_mforward != NULL &&
725 ip_mforward(ip, m->m_pkthdr.rcvif, m, NULL) != 0) {
726 ipstat.ips_cantforward++;
732 * The process-level routing daemon needs to receive
733 * all multicast IGMP packets, whether or not this
734 * host belongs to their destination groups.
736 if (ip->ip_p == IPPROTO_IGMP)
738 ipstat.ips_forward++;
741 * See if we belong to the destination multicast group on the
744 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
746 ipstat.ips_notmember++;
752 if (ip->ip_dst.s_addr == INADDR_BROADCAST)
754 if (ip->ip_dst.s_addr == INADDR_ANY)
758 * FAITH(Firewall Aided Internet Translator)
760 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
762 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP)
770 * Not for us; forward if possible and desirable.
773 ipstat.ips_cantforward++;
778 * Enforce inbound IPsec SPD.
780 if (ipsec4_in_reject(m, NULL)) {
781 ipsecstat.in_polvio++;
786 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
789 tdbi = (struct tdb_ident *)m_tag_data(mtag);
790 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
792 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
793 IP_FORWARDING, &error);
795 if (sp == NULL) { /* NB: can happen if error */
797 /*XXX error stat???*/
798 DPRINTF(("ip_input: no SP for forwarding\n")); /*XXX*/
803 * Check security policy against packet attributes.
805 error = ipsec_in_reject(sp, m);
809 ipstat.ips_cantforward++;
813 ip_forward(m, using_srcrt, next_hop);
820 * IPSTEALTH: Process non-routing options only
821 * if the packet is destined for us.
824 hlen > sizeof(struct ip) &&
825 ip_dooptions(m, 1, next_hop))
828 /* Count the packet in the ip address stats */
830 ia->ia_ifa.if_ipackets++;
831 ia->ia_ifa.if_ibytes += m->m_pkthdr.len;
835 * If offset or IP_MF are set, must reassemble.
836 * Otherwise, nothing need be done.
837 * (We could look in the reassembly queue to see
838 * if the packet was previously fragmented,
839 * but it's not worth the time; just let them time out.)
841 if (ip->ip_off & (IP_MF | IP_OFFMASK)) {
843 * Attempt reassembly; if it succeeds, proceed.
844 * ip_reass() will return a different mbuf.
849 ip = mtod(m, struct ip *);
851 /* Get the header length of the reassembled packet */
852 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
854 needredispatch = TRUE;
861 * enforce IPsec policy checking if we are seeing last header.
862 * note that we do not visit this with protocols with pcb layer
863 * code - like udp/tcp/raw ip.
865 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) &&
866 ipsec4_in_reject(m, NULL)) {
867 ipsecstat.in_polvio++;
873 * enforce IPsec policy checking if we are seeing last header.
874 * note that we do not visit this with protocols with pcb layer
875 * code - like udp/tcp/raw ip.
877 if (inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) {
879 * Check if the packet has already had IPsec processing
880 * done. If so, then just pass it along. This tag gets
881 * set during AH, ESP, etc. input handling, before the
882 * packet is returned to the ip input queue for delivery.
884 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
887 tdbi = (struct tdb_ident *)m_tag_data(mtag);
888 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
890 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
891 IP_FORWARDING, &error);
895 * Check security policy against packet attributes.
897 error = ipsec_in_reject(sp, m);
900 /* XXX error stat??? */
902 DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/
909 #endif /* FAST_IPSEC */
911 ipstat.ips_delivered++;
912 if (needredispatch) {
913 struct netmsg_packet *pmsg;
916 ip->ip_off = htons(ip->ip_off);
917 ip->ip_len = htons(ip->ip_len);
918 port = ip_mport_in(&m);
922 pmsg = &m->m_hdr.mh_netmsg;
923 netmsg_init(&pmsg->nm_netmsg, &netisr_apanic_rport, MSGF_MPSAFE,
924 transport_processing_handler);
926 pmsg->nm_netmsg.nm_lmsg.u.ms_result = hlen;
928 ip = mtod(m, struct ip *);
929 ip->ip_len = ntohs(ip->ip_len);
930 ip->ip_off = ntohs(ip->ip_off);
931 lwkt_sendmsg(port, &pmsg->nm_netmsg.nm_lmsg);
933 transport_processing_oncpu(m, hlen, ip);
942 * Take incoming datagram fragment and try to reassemble it into
943 * whole datagram. If a chain for reassembly of this datagram already
944 * exists, then it is given as fp; otherwise have to make a chain.
947 ip_reass(struct mbuf *m)
949 struct ip *ip = mtod(m, struct ip *);
950 struct mbuf *p = NULL, *q, *nq;
952 struct ipq *fp = NULL;
953 int hlen = IP_VHL_HL(ip->ip_vhl) << 2;
957 /* If maxnipq is 0, never accept fragments. */
959 ipstat.ips_fragments++;
960 ipstat.ips_fragdropped++;
965 sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
967 * Look for queue of fragments of this datagram.
969 for (fp = ipq[sum].next; fp != &ipq[sum]; fp = fp->next)
970 if (ip->ip_id == fp->ipq_id &&
971 ip->ip_src.s_addr == fp->ipq_src.s_addr &&
972 ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
973 ip->ip_p == fp->ipq_p)
979 * Enforce upper bound on number of fragmented packets
980 * for which we attempt reassembly;
981 * If maxnipq is -1, accept all fragments without limitation.
983 if (nipq > maxnipq && maxnipq > 0) {
985 * drop something from the tail of the current queue
986 * before proceeding further
988 if (ipq[sum].prev == &ipq[sum]) { /* gak */
989 for (i = 0; i < IPREASS_NHASH; i++) {
990 if (ipq[i].prev != &ipq[i]) {
991 ipstat.ips_fragtimeout +=
992 ipq[i].prev->ipq_nfrags;
993 ip_freef(ipq[i].prev);
998 ipstat.ips_fragtimeout +=
999 ipq[sum].prev->ipq_nfrags;
1000 ip_freef(ipq[sum].prev);
1005 * Adjust ip_len to not reflect header,
1006 * convert offset of this to bytes.
1009 if (ip->ip_off & IP_MF) {
1011 * Make sure that fragments have a data length
1012 * that's a non-zero multiple of 8 bytes.
1014 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
1015 ipstat.ips_toosmall++; /* XXX */
1019 m->m_flags |= M_FRAG;
1021 m->m_flags &= ~M_FRAG;
1024 ipstat.ips_fragments++;
1025 m->m_pkthdr.header = ip;
1028 * If the hardware has not done csum over this fragment
1029 * then csum_data is not valid at all.
1031 if ((m->m_pkthdr.csum_flags & (CSUM_FRAG_NOT_CHECKED | CSUM_DATA_VALID))
1032 == (CSUM_FRAG_NOT_CHECKED | CSUM_DATA_VALID)) {
1033 m->m_pkthdr.csum_data = 0;
1034 m->m_pkthdr.csum_flags &= ~(CSUM_DATA_VALID | CSUM_PSEUDO_HDR);
1038 * Presence of header sizes in mbufs
1039 * would confuse code below.
1045 * If first fragment to arrive, create a reassembly queue.
1048 if ((fp = mpipe_alloc_nowait(&ipq_mpipe)) == NULL)
1050 insque(fp, &ipq[sum]);
1053 fp->ipq_ttl = IPFRAGTTL;
1054 fp->ipq_p = ip->ip_p;
1055 fp->ipq_id = ip->ip_id;
1056 fp->ipq_src = ip->ip_src;
1057 fp->ipq_dst = ip->ip_dst;
1059 m->m_nextpkt = NULL;
1065 #define GETIP(m) ((struct ip*)((m)->m_pkthdr.header))
1068 * Find a segment which begins after this one does.
1070 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt)
1071 if (GETIP(q)->ip_off > ip->ip_off)
1075 * If there is a preceding segment, it may provide some of
1076 * our data already. If so, drop the data from the incoming
1077 * segment. If it provides all of our data, drop us, otherwise
1078 * stick new segment in the proper place.
1080 * If some of the data is dropped from the the preceding
1081 * segment, then it's checksum is invalidated.
1084 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off;
1086 if (i >= ip->ip_len)
1089 m->m_pkthdr.csum_flags = 0;
1093 m->m_nextpkt = p->m_nextpkt;
1096 m->m_nextpkt = fp->ipq_frags;
1101 * While we overlap succeeding segments trim them or,
1102 * if they are completely covered, dequeue them.
1104 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off;
1106 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off;
1107 if (i < GETIP(q)->ip_len) {
1108 GETIP(q)->ip_len -= i;
1109 GETIP(q)->ip_off += i;
1111 q->m_pkthdr.csum_flags = 0;
1116 ipstat.ips_fragdropped++;
1118 q->m_nextpkt = NULL;
1124 * Check for complete reassembly and perform frag per packet
1127 * Frag limiting is performed here so that the nth frag has
1128 * a chance to complete the packet before we drop the packet.
1129 * As a result, n+1 frags are actually allowed per packet, but
1130 * only n will ever be stored. (n = maxfragsperpacket.)
1134 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
1135 if (GETIP(q)->ip_off != next) {
1136 if (fp->ipq_nfrags > maxfragsperpacket) {
1137 ipstat.ips_fragdropped += fp->ipq_nfrags;
1142 next += GETIP(q)->ip_len;
1144 /* Make sure the last packet didn't have the IP_MF flag */
1145 if (p->m_flags & M_FRAG) {
1146 if (fp->ipq_nfrags > maxfragsperpacket) {
1147 ipstat.ips_fragdropped += fp->ipq_nfrags;
1154 * Reassembly is complete. Make sure the packet is a sane size.
1158 if (next + (IP_VHL_HL(ip->ip_vhl) << 2) > IP_MAXPACKET) {
1159 ipstat.ips_toolong++;
1160 ipstat.ips_fragdropped += fp->ipq_nfrags;
1166 * Concatenate fragments.
1173 q->m_nextpkt = NULL;
1174 for (q = nq; q != NULL; q = nq) {
1176 q->m_nextpkt = NULL;
1177 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
1178 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
1183 * Clean up the 1's complement checksum. Carry over 16 bits must
1184 * be added back. This assumes no more then 65535 packet fragments
1185 * were reassembled. A second carry can also occur (but not a third).
1187 m->m_pkthdr.csum_data = (m->m_pkthdr.csum_data & 0xffff) +
1188 (m->m_pkthdr.csum_data >> 16);
1189 if (m->m_pkthdr.csum_data > 0xFFFF)
1190 m->m_pkthdr.csum_data -= 0xFFFF;
1193 * Create header for new ip packet by
1194 * modifying header of first packet;
1195 * dequeue and discard fragment reassembly header.
1196 * Make header visible.
1199 ip->ip_src = fp->ipq_src;
1200 ip->ip_dst = fp->ipq_dst;
1203 mpipe_free(&ipq_mpipe, fp);
1204 m->m_len += (IP_VHL_HL(ip->ip_vhl) << 2);
1205 m->m_data -= (IP_VHL_HL(ip->ip_vhl) << 2);
1206 /* some debugging cruft by sklower, below, will go away soon */
1207 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */
1210 for (n = m; n; n = n->m_next)
1212 m->m_pkthdr.len = plen;
1215 ipstat.ips_reassembled++;
1219 ipstat.ips_fragdropped++;
1229 * Free a fragment reassembly header and all
1230 * associated datagrams.
1233 ip_freef(struct ipq *fp)
1237 while (fp->ipq_frags) {
1239 fp->ipq_frags = q->m_nextpkt;
1240 q->m_nextpkt = NULL;
1244 mpipe_free(&ipq_mpipe, fp);
1249 * IP timer processing;
1250 * if a timer expires on a reassembly
1251 * queue, discard it.
1260 for (i = 0; i < IPREASS_NHASH; i++) {
1264 while (fp != &ipq[i]) {
1267 if (fp->prev->ipq_ttl == 0) {
1268 ipstat.ips_fragtimeout += fp->prev->ipq_nfrags;
1274 * If we are over the maximum number of fragments
1275 * (due to the limit being lowered), drain off
1276 * enough to get down to the new limit.
1278 if (maxnipq >= 0 && nipq > maxnipq) {
1279 for (i = 0; i < IPREASS_NHASH; i++) {
1280 while (nipq > maxnipq &&
1281 (ipq[i].next != &ipq[i])) {
1282 ipstat.ips_fragdropped +=
1283 ipq[i].next->ipq_nfrags;
1284 ip_freef(ipq[i].next);
1293 * Drain off all datagram fragments.
1300 for (i = 0; i < IPREASS_NHASH; i++) {
1301 while (ipq[i].next != &ipq[i]) {
1302 ipstat.ips_fragdropped += ipq[i].next->ipq_nfrags;
1303 ip_freef(ipq[i].next);
1310 * Do option processing on a datagram,
1311 * possibly discarding it if bad options are encountered,
1312 * or forwarding it if source-routed.
1313 * The pass argument is used when operating in the IPSTEALTH
1314 * mode to tell what options to process:
1315 * [LS]SRR (pass 0) or the others (pass 1).
1316 * The reason for as many as two passes is that when doing IPSTEALTH,
1317 * non-routing options should be processed only if the packet is for us.
1318 * Returns 1 if packet has been forwarded/freed,
1319 * 0 if the packet should be processed further.
1322 ip_dooptions(struct mbuf *m, int pass, struct sockaddr_in *next_hop)
1324 struct sockaddr_in ipaddr = { sizeof ipaddr, AF_INET };
1325 struct ip *ip = mtod(m, struct ip *);
1327 struct in_ifaddr *ia;
1328 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB;
1329 boolean_t forward = FALSE;
1330 struct in_addr *sin, dst;
1334 cp = (u_char *)(ip + 1);
1335 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
1336 for (; cnt > 0; cnt -= optlen, cp += optlen) {
1337 opt = cp[IPOPT_OPTVAL];
1338 if (opt == IPOPT_EOL)
1340 if (opt == IPOPT_NOP)
1343 if (cnt < IPOPT_OLEN + sizeof(*cp)) {
1344 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1347 optlen = cp[IPOPT_OLEN];
1348 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
1349 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1359 * Source routing with record.
1360 * Find interface with current destination address.
1361 * If none on this machine then drop if strictly routed,
1362 * or do nothing if loosely routed.
1363 * Record interface address and bring up next address
1364 * component. If strictly routed make sure next
1365 * address is on directly accessible net.
1369 if (ipstealth && pass > 0)
1371 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1372 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1375 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1376 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1379 ipaddr.sin_addr = ip->ip_dst;
1380 ia = (struct in_ifaddr *)
1381 ifa_ifwithaddr((struct sockaddr *)&ipaddr);
1383 if (opt == IPOPT_SSRR) {
1384 type = ICMP_UNREACH;
1385 code = ICMP_UNREACH_SRCFAIL;
1388 if (!ip_dosourceroute)
1389 goto nosourcerouting;
1391 * Loose routing, and not at next destination
1392 * yet; nothing to do except forward.
1396 off--; /* 0 origin */
1397 if (off > optlen - (int)sizeof(struct in_addr)) {
1399 * End of source route. Should be for us.
1401 if (!ip_acceptsourceroute)
1402 goto nosourcerouting;
1403 save_rte(m, cp, ip->ip_src);
1408 if (!ip_dosourceroute) {
1410 char buf[sizeof "aaa.bbb.ccc.ddd"];
1413 * Acting as a router, so generate ICMP
1416 strcpy(buf, inet_ntoa(ip->ip_dst));
1418 "attempted source route from %s to %s\n",
1419 inet_ntoa(ip->ip_src), buf);
1420 type = ICMP_UNREACH;
1421 code = ICMP_UNREACH_SRCFAIL;
1425 * Not acting as a router,
1429 ipstat.ips_cantforward++;
1436 * locate outgoing interface
1438 memcpy(&ipaddr.sin_addr, cp + off,
1439 sizeof ipaddr.sin_addr);
1441 if (opt == IPOPT_SSRR) {
1442 #define INA struct in_ifaddr *
1443 #define SA struct sockaddr *
1444 if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr))
1446 ia = (INA)ifa_ifwithnet((SA)&ipaddr);
1448 ia = ip_rtaddr(ipaddr.sin_addr,
1449 &ipforward_rt[mycpuid]);
1451 type = ICMP_UNREACH;
1452 code = ICMP_UNREACH_SRCFAIL;
1455 ip->ip_dst = ipaddr.sin_addr;
1456 memcpy(cp + off, &IA_SIN(ia)->sin_addr,
1457 sizeof(struct in_addr));
1458 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1460 * Let ip_intr's mcast routing check handle mcast pkts
1462 forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr));
1466 if (ipstealth && pass == 0)
1468 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1469 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1472 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1473 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1477 * If no space remains, ignore.
1479 off--; /* 0 origin */
1480 if (off > optlen - (int)sizeof(struct in_addr))
1482 memcpy(&ipaddr.sin_addr, &ip->ip_dst,
1483 sizeof ipaddr.sin_addr);
1485 * locate outgoing interface; if we're the destination,
1486 * use the incoming interface (should be same).
1488 if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == NULL &&
1489 (ia = ip_rtaddr(ipaddr.sin_addr,
1490 &ipforward_rt[mycpuid]))
1492 type = ICMP_UNREACH;
1493 code = ICMP_UNREACH_HOST;
1496 memcpy(cp + off, &IA_SIN(ia)->sin_addr,
1497 sizeof(struct in_addr));
1498 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1502 if (ipstealth && pass == 0)
1504 code = cp - (u_char *)ip;
1505 if (optlen < 4 || optlen > 40) {
1506 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1509 if ((off = cp[IPOPT_OFFSET]) < 5) {
1510 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1513 if (off > optlen - (int)sizeof(int32_t)) {
1514 cp[IPOPT_OFFSET + 1] += (1 << 4);
1515 if ((cp[IPOPT_OFFSET + 1] & 0xf0) == 0) {
1516 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1521 off--; /* 0 origin */
1522 sin = (struct in_addr *)(cp + off);
1523 switch (cp[IPOPT_OFFSET + 1] & 0x0f) {
1525 case IPOPT_TS_TSONLY:
1528 case IPOPT_TS_TSANDADDR:
1529 if (off + sizeof(n_time) +
1530 sizeof(struct in_addr) > optlen) {
1531 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1534 ipaddr.sin_addr = dst;
1535 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr,
1539 memcpy(sin, &IA_SIN(ia)->sin_addr,
1540 sizeof(struct in_addr));
1541 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1542 off += sizeof(struct in_addr);
1545 case IPOPT_TS_PRESPEC:
1546 if (off + sizeof(n_time) +
1547 sizeof(struct in_addr) > optlen) {
1548 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1551 memcpy(&ipaddr.sin_addr, sin,
1552 sizeof(struct in_addr));
1553 if (ifa_ifwithaddr((SA)&ipaddr) == NULL)
1555 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1556 off += sizeof(struct in_addr);
1560 code = &cp[IPOPT_OFFSET + 1] - (u_char *)ip;
1564 memcpy(cp + off, &ntime, sizeof(n_time));
1565 cp[IPOPT_OFFSET] += sizeof(n_time);
1568 if (forward && ipforwarding) {
1569 ip_forward(m, TRUE, next_hop);
1574 icmp_error(m, type, code, 0, 0);
1575 ipstat.ips_badoptions++;
1580 * Given address of next destination (final or next hop),
1581 * return internet address info of interface to be used to get there.
1584 ip_rtaddr(struct in_addr dst, struct route *ro)
1586 struct sockaddr_in *sin;
1588 sin = (struct sockaddr_in *)&ro->ro_dst;
1590 if (ro->ro_rt == NULL || dst.s_addr != sin->sin_addr.s_addr) {
1591 if (ro->ro_rt != NULL) {
1595 sin->sin_family = AF_INET;
1596 sin->sin_len = sizeof *sin;
1597 sin->sin_addr = dst;
1598 rtalloc_ign(ro, RTF_PRCLONING);
1601 if (ro->ro_rt == NULL)
1604 return (ifatoia(ro->ro_rt->rt_ifa));
1608 * Save incoming source route for use in replies,
1609 * to be picked up later by ip_srcroute if the receiver is interested.
1612 save_rte(struct mbuf *m, u_char *option, struct in_addr dst)
1615 struct ip_srcrt_opt *opt;
1618 mtag = m_tag_get(PACKET_TAG_IPSRCRT, sizeof(*opt), MB_DONTWAIT);
1621 opt = m_tag_data(mtag);
1623 olen = option[IPOPT_OLEN];
1626 kprintf("save_rte: olen %d\n", olen);
1628 if (olen > sizeof(opt->ip_srcrt) - (1 + sizeof(dst))) {
1632 bcopy(option, opt->ip_srcrt.srcopt, olen);
1633 opt->ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1634 opt->ip_srcrt.dst = dst;
1635 m_tag_prepend(m, mtag);
1639 * Retrieve incoming source route for use in replies,
1640 * in the same form used by setsockopt.
1641 * The first hop is placed before the options, will be removed later.
1644 ip_srcroute(struct mbuf *m0)
1646 struct in_addr *p, *q;
1649 struct ip_srcrt_opt *opt;
1654 mtag = m_tag_find(m0, PACKET_TAG_IPSRCRT, NULL);
1657 opt = m_tag_data(mtag);
1659 if (opt->ip_nhops == 0)
1661 m = m_get(MB_DONTWAIT, MT_HEADER);
1665 #define OPTSIZ (sizeof(opt->ip_srcrt.nop) + sizeof(opt->ip_srcrt.srcopt))
1667 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
1668 m->m_len = opt->ip_nhops * sizeof(struct in_addr) +
1669 sizeof(struct in_addr) + OPTSIZ;
1672 kprintf("ip_srcroute: nhops %d mlen %d",
1673 opt->ip_nhops, m->m_len);
1678 * First save first hop for return route
1680 p = &opt->ip_srcrt.route[opt->ip_nhops - 1];
1681 *(mtod(m, struct in_addr *)) = *p--;
1684 kprintf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr));
1688 * Copy option fields and padding (nop) to mbuf.
1690 opt->ip_srcrt.nop = IPOPT_NOP;
1691 opt->ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
1692 memcpy(mtod(m, caddr_t) + sizeof(struct in_addr), &opt->ip_srcrt.nop,
1694 q = (struct in_addr *)(mtod(m, caddr_t) +
1695 sizeof(struct in_addr) + OPTSIZ);
1698 * Record return path as an IP source route,
1699 * reversing the path (pointers are now aligned).
1701 while (p >= opt->ip_srcrt.route) {
1704 kprintf(" %x", ntohl(q->s_addr));
1709 * Last hop goes to final destination.
1711 *q = opt->ip_srcrt.dst;
1712 m_tag_delete(m0, mtag);
1715 kprintf(" %x\n", ntohl(q->s_addr));
1721 * Strip out IP options.
1724 ip_stripoptions(struct mbuf *m)
1727 struct ip *ip = mtod(m, struct ip *);
1731 optlen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
1732 opts = (caddr_t)(ip + 1);
1733 datalen = m->m_len - (sizeof(struct ip) + optlen);
1734 bcopy(opts + optlen, opts, datalen);
1736 if (m->m_flags & M_PKTHDR)
1737 m->m_pkthdr.len -= optlen;
1738 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, sizeof(struct ip) >> 2);
1741 u_char inetctlerrmap[PRC_NCMDS] = {
1743 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
1744 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
1745 EMSGSIZE, EHOSTUNREACH, 0, 0,
1747 ENOPROTOOPT, ECONNREFUSED
1751 * Forward a packet. If some error occurs return the sender
1752 * an icmp packet. Note we can't always generate a meaningful
1753 * icmp message because icmp doesn't have a large enough repertoire
1754 * of codes and types.
1756 * If not forwarding, just drop the packet. This could be confusing
1757 * if ipforwarding was zero but some routing protocol was advancing
1758 * us as a gateway to somewhere. However, we must let the routing
1759 * protocol deal with that.
1761 * The using_srcrt parameter indicates whether the packet is being forwarded
1762 * via a source route.
1765 ip_forward(struct mbuf *m, boolean_t using_srcrt, struct sockaddr_in *next_hop)
1767 struct ip *ip = mtod(m, struct ip *);
1768 struct sockaddr_in *ipforward_rtaddr;
1770 int error, type = 0, code = 0, destmtu = 0;
1773 struct in_addr pkt_dst;
1774 struct route *cache_rt = &ipforward_rt[mycpuid];
1778 * Cache the destination address of the packet; this may be
1779 * changed by use of 'ipfw fwd'.
1781 pkt_dst = (next_hop != NULL) ? next_hop->sin_addr : ip->ip_dst;
1785 kprintf("forward: src %x dst %x ttl %x\n",
1786 ip->ip_src.s_addr, pkt_dst.s_addr, ip->ip_ttl);
1789 if (m->m_flags & (M_BCAST | M_MCAST) || !in_canforward(pkt_dst)) {
1790 ipstat.ips_cantforward++;
1794 if (!ipstealth && ip->ip_ttl <= IPTTLDEC) {
1795 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0);
1799 ipforward_rtaddr = (struct sockaddr_in *) &cache_rt->ro_dst;
1800 if (cache_rt->ro_rt == NULL ||
1801 ipforward_rtaddr->sin_addr.s_addr != pkt_dst.s_addr) {
1802 if (cache_rt->ro_rt != NULL) {
1803 RTFREE(cache_rt->ro_rt);
1804 cache_rt->ro_rt = NULL;
1806 ipforward_rtaddr->sin_family = AF_INET;
1807 ipforward_rtaddr->sin_len = sizeof(struct sockaddr_in);
1808 ipforward_rtaddr->sin_addr = pkt_dst;
1809 rtalloc_ign(cache_rt, RTF_PRCLONING);
1810 if (cache_rt->ro_rt == NULL) {
1811 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0);
1815 rt = cache_rt->ro_rt;
1818 * Save the IP header and at most 8 bytes of the payload,
1819 * in case we need to generate an ICMP message to the src.
1821 * XXX this can be optimized a lot by saving the data in a local
1822 * buffer on the stack (72 bytes at most), and only allocating the
1823 * mbuf if really necessary. The vast majority of the packets
1824 * are forwarded without having to send an ICMP back (either
1825 * because unnecessary, or because rate limited), so we are
1826 * really we are wasting a lot of work here.
1828 * We don't use m_copy() because it might return a reference
1829 * to a shared cluster. Both this function and ip_output()
1830 * assume exclusive access to the IP header in `m', so any
1831 * data in a cluster may change before we reach icmp_error().
1833 MGETHDR(mcopy, MB_DONTWAIT, m->m_type);
1834 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, MB_DONTWAIT)) {
1836 * It's probably ok if the pkthdr dup fails (because
1837 * the deep copy of the tag chain failed), but for now
1838 * be conservative and just discard the copy since
1839 * code below may some day want the tags.
1844 if (mcopy != NULL) {
1845 mcopy->m_len = imin((IP_VHL_HL(ip->ip_vhl) << 2) + 8,
1847 mcopy->m_pkthdr.len = mcopy->m_len;
1848 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1852 ip->ip_ttl -= IPTTLDEC;
1855 * If forwarding packet using same interface that it came in on,
1856 * perhaps should send a redirect to sender to shortcut a hop.
1857 * Only send redirect if source is sending directly to us,
1858 * and if packet was not source routed (or has any options).
1859 * Also, don't send redirect if forwarding using a default route
1860 * or a route modified by a redirect.
1862 if (rt->rt_ifp == m->m_pkthdr.rcvif &&
1863 !(rt->rt_flags & (RTF_DYNAMIC | RTF_MODIFIED)) &&
1864 satosin(rt_key(rt))->sin_addr.s_addr != INADDR_ANY &&
1865 ipsendredirects && !using_srcrt && next_hop == NULL) {
1866 u_long src = ntohl(ip->ip_src.s_addr);
1867 struct in_ifaddr *rt_ifa = (struct in_ifaddr *)rt->rt_ifa;
1869 if (rt_ifa != NULL &&
1870 (src & rt_ifa->ia_subnetmask) == rt_ifa->ia_subnet) {
1871 if (rt->rt_flags & RTF_GATEWAY)
1872 dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
1874 dest = pkt_dst.s_addr;
1876 * Router requirements says to only send
1879 type = ICMP_REDIRECT;
1880 code = ICMP_REDIRECT_HOST;
1883 kprintf("redirect (%d) to %x\n", code, dest);
1888 error = ip_output(m, NULL, cache_rt, IP_FORWARDING, NULL, NULL);
1890 ipstat.ips_forward++;
1893 ipflow_create(cache_rt, mcopy);
1896 return; /* most common case */
1898 ipstat.ips_redirectsent++;
1901 ipstat.ips_cantforward++;
1908 * Send ICMP message.
1913 case 0: /* forwarded, but need redirect */
1914 /* type, code set above */
1917 case ENETUNREACH: /* shouldn't happen, checked above */
1922 type = ICMP_UNREACH;
1923 code = ICMP_UNREACH_HOST;
1927 type = ICMP_UNREACH;
1928 code = ICMP_UNREACH_NEEDFRAG;
1931 * If the packet is routed over IPsec tunnel, tell the
1932 * originator the tunnel MTU.
1933 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
1936 if (cache_rt->ro_rt != NULL) {
1937 struct secpolicy *sp = NULL;
1942 sp = ipsec4_getpolicybyaddr(mcopy,
1948 destmtu = cache_rt->ro_rt->rt_ifp->if_mtu;
1950 /* count IPsec header size */
1951 ipsechdr = ipsec4_hdrsiz(mcopy,
1956 * find the correct route for outer IPv4
1957 * header, compute tunnel MTU.
1960 if (sp->req != NULL && sp->req->sav != NULL &&
1961 sp->req->sav->sah != NULL) {
1962 ro = &sp->req->sav->sah->sa_route;
1963 if (ro->ro_rt != NULL &&
1964 ro->ro_rt->rt_ifp != NULL) {
1966 ro->ro_rt->rt_ifp->if_mtu;
1967 destmtu -= ipsechdr;
1976 * If the packet is routed over IPsec tunnel, tell the
1977 * originator the tunnel MTU.
1978 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
1981 if (cache_rt->ro_rt != NULL) {
1982 struct secpolicy *sp = NULL;
1987 sp = ipsec_getpolicybyaddr(mcopy,
1993 destmtu = cache_rt->ro_rt->rt_ifp->if_mtu;
1995 /* count IPsec header size */
1996 ipsechdr = ipsec4_hdrsiz(mcopy,
2001 * find the correct route for outer IPv4
2002 * header, compute tunnel MTU.
2005 if (sp->req != NULL &&
2006 sp->req->sav != NULL &&
2007 sp->req->sav->sah != NULL) {
2008 ro = &sp->req->sav->sah->sa_route;
2009 if (ro->ro_rt != NULL &&
2010 ro->ro_rt->rt_ifp != NULL) {
2012 ro->ro_rt->rt_ifp->if_mtu;
2013 destmtu -= ipsechdr;
2020 #else /* !IPSEC && !FAST_IPSEC */
2021 if (cache_rt->ro_rt != NULL)
2022 destmtu = cache_rt->ro_rt->rt_ifp->if_mtu;
2024 ipstat.ips_cantfrag++;
2029 * A router should not generate ICMP_SOURCEQUENCH as
2030 * required in RFC1812 Requirements for IP Version 4 Routers.
2031 * Source quench could be a big problem under DoS attacks,
2032 * or if the underlying interface is rate-limited.
2033 * Those who need source quench packets may re-enable them
2034 * via the net.inet.ip.sendsourcequench sysctl.
2036 if (!ip_sendsourcequench) {
2040 type = ICMP_SOURCEQUENCH;
2045 case EACCES: /* ipfw denied packet */
2049 icmp_error(mcopy, type, code, dest, destmtu);
2053 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
2056 if (inp->inp_socket->so_options & SO_TIMESTAMP) {
2060 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
2061 SCM_TIMESTAMP, SOL_SOCKET);
2063 mp = &(*mp)->m_next;
2065 if (inp->inp_flags & INP_RECVDSTADDR) {
2066 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
2067 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
2069 mp = &(*mp)->m_next;
2071 if (inp->inp_flags & INP_RECVTTL) {
2072 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl,
2073 sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
2075 mp = &(*mp)->m_next;
2079 * Moving these out of udp_input() made them even more broken
2080 * than they already were.
2082 /* options were tossed already */
2083 if (inp->inp_flags & INP_RECVOPTS) {
2084 *mp = sbcreatecontrol((caddr_t) opts_deleted_above,
2085 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
2087 mp = &(*mp)->m_next;
2089 /* ip_srcroute doesn't do what we want here, need to fix */
2090 if (inp->inp_flags & INP_RECVRETOPTS) {
2091 *mp = sbcreatecontrol((caddr_t) ip_srcroute(m),
2092 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
2094 mp = &(*mp)->m_next;
2097 if (inp->inp_flags & INP_RECVIF) {
2100 struct sockaddr_dl sdl;
2103 struct sockaddr_dl *sdp;
2104 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
2106 if (((ifp = m->m_pkthdr.rcvif)) &&
2107 ((ifp->if_index != 0) && (ifp->if_index <= if_index))) {
2108 sdp = IF_LLSOCKADDR(ifp);
2110 * Change our mind and don't try copy.
2112 if ((sdp->sdl_family != AF_LINK) ||
2113 (sdp->sdl_len > sizeof(sdlbuf))) {
2116 bcopy(sdp, sdl2, sdp->sdl_len);
2120 offsetof(struct sockaddr_dl, sdl_data[0]);
2121 sdl2->sdl_family = AF_LINK;
2122 sdl2->sdl_index = 0;
2123 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
2125 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len,
2126 IP_RECVIF, IPPROTO_IP);
2128 mp = &(*mp)->m_next;
2133 * XXX these routines are called from the upper part of the kernel.
2135 * They could also be moved to ip_mroute.c, since all the RSVP
2136 * handling is done there already.
2139 ip_rsvp_init(struct socket *so)
2141 if (so->so_type != SOCK_RAW ||
2142 so->so_proto->pr_protocol != IPPROTO_RSVP)
2145 if (ip_rsvpd != NULL)
2150 * This may seem silly, but we need to be sure we don't over-increment
2151 * the RSVP counter, in case something slips up.
2166 * This may seem silly, but we need to be sure we don't over-decrement
2167 * the RSVP counter, in case something slips up.
2177 rsvp_input(struct mbuf *m, ...) /* XXX must fixup manually */
2183 off = __va_arg(ap, int);
2184 proto = __va_arg(ap, int);
2187 if (rsvp_input_p) { /* call the real one if loaded */
2188 rsvp_input_p(m, off, proto);
2192 /* Can still get packets with rsvp_on = 0 if there is a local member
2193 * of the group to which the RSVP packet is addressed. But in this
2194 * case we want to throw the packet away.
2202 if (ip_rsvpd != NULL) {
2203 rip_input(m, off, proto);
2206 /* Drop the packet */