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 1.107 2008/09/18 11:19:42 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 struct domain inetdomain;
207 extern struct protosw inetsw[];
208 u_char ip_protox[IPPROTO_MAX];
209 struct in_ifaddrhead in_ifaddrheads[MAXCPU]; /* first inet address */
210 struct in_ifaddrhashhead *in_ifaddrhashtbls[MAXCPU];
211 /* inet addr hash table */
212 u_long in_ifaddrhmask; /* mask for hash table */
214 struct ip_stats ipstats_percpu[MAXCPU];
217 sysctl_ipstats(SYSCTL_HANDLER_ARGS)
221 for (cpu = 0; cpu < ncpus; ++cpu) {
222 if ((error = SYSCTL_OUT(req, &ipstats_percpu[cpu],
223 sizeof(struct ip_stats))))
225 if ((error = SYSCTL_IN(req, &ipstats_percpu[cpu],
226 sizeof(struct ip_stats))))
232 SYSCTL_PROC(_net_inet_ip, IPCTL_STATS, stats, (CTLTYPE_OPAQUE | CTLFLAG_RW),
233 0, 0, sysctl_ipstats, "S,ip_stats", "IP statistics");
235 SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW,
236 &ipstat, ip_stats, "IP statistics");
239 /* Packet reassembly stuff */
240 #define IPREASS_NHASH_LOG2 6
241 #define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2)
242 #define IPREASS_HMASK (IPREASS_NHASH - 1)
243 #define IPREASS_HASH(x,y) \
244 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK)
246 static struct ipq ipq[IPREASS_NHASH];
249 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
250 &ip_mtu, 0, "Default MTU");
254 static int ipstealth = 0;
255 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW, &ipstealth, 0, "");
257 static const int ipstealth = 0;
260 struct mbuf *(*ip_divert_p)(struct mbuf *, int, int);
262 struct pfil_head inet_pfil_hook;
265 * struct ip_srcrt_opt is used to store packet state while it travels
268 * XXX Note that the code even makes assumptions on the size and
269 * alignment of fields inside struct ip_srcrt so e.g. adding some
270 * fields will break the code. This needs to be fixed.
272 * We need to save the IP options in case a protocol wants to respond
273 * to an incoming packet over the same route if the packet got here
274 * using IP source routing. This allows connection establishment and
275 * maintenance when the remote end is on a network that is not known
279 struct in_addr dst; /* final destination */
280 char nop; /* one NOP to align */
281 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */
282 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)];
285 struct ip_srcrt_opt {
287 struct ip_srcrt ip_srcrt;
290 static MALLOC_DEFINE(M_IPQ, "ipq", "IP Fragment Management");
291 static struct malloc_pipe ipq_mpipe;
293 static void save_rte(struct mbuf *, u_char *, struct in_addr);
294 static int ip_dooptions(struct mbuf *m, int, struct sockaddr_in *);
295 static void ip_freef(struct ipq *);
296 static void ip_input_handler(struct netmsg *);
299 * IP initialization: fill in IP protocol switch table.
300 * All protocols not implemented in kernel go to raw IP protocol handler.
312 * Make sure we can handle a reasonable number of fragments but
313 * cap it at 4000 (XXX).
315 mpipe_init(&ipq_mpipe, M_IPQ, sizeof(struct ipq),
316 IFQ_MAXLEN, 4000, 0, NULL);
317 for (i = 0; i < ncpus; ++i) {
318 TAILQ_INIT(&in_ifaddrheads[i]);
319 in_ifaddrhashtbls[i] =
320 hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask);
322 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
325 for (i = 0; i < IPPROTO_MAX; i++)
326 ip_protox[i] = pr - inetsw;
327 for (pr = inetdomain.dom_protosw;
328 pr < inetdomain.dom_protoswNPROTOSW; pr++)
329 if (pr->pr_domain->dom_family == PF_INET &&
330 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
331 ip_protox[pr->pr_protocol] = pr - inetsw;
333 inet_pfil_hook.ph_type = PFIL_TYPE_AF;
334 inet_pfil_hook.ph_af = AF_INET;
335 if ((i = pfil_head_register(&inet_pfil_hook)) != 0) {
336 kprintf("%s: WARNING: unable to register pfil hook, "
337 "error %d\n", __func__, i);
340 for (i = 0; i < IPREASS_NHASH; i++)
341 ipq[i].next = ipq[i].prev = &ipq[i];
343 maxnipq = nmbclusters / 32;
344 maxfragsperpacket = 16;
346 ip_id = time_second & 0xffff;
349 * Initialize IP statistics counters for each CPU.
353 for (cpu = 0; cpu < ncpus; ++cpu) {
354 bzero(&ipstats_percpu[cpu], sizeof(struct ip_stats));
357 bzero(&ipstat, sizeof(struct ip_stats));
360 netisr_register(NETISR_IP, ip_mport_in, ip_input_handler);
364 * XXX watch out this one. It is perhaps used as a cache for
365 * the most recently used route ? it is cleared in in_addroute()
366 * when a new route is successfully created.
368 struct route ipforward_rt[MAXCPU];
370 /* Do transport protocol processing. */
372 transport_processing_oncpu(struct mbuf *m, int hlen, struct ip *ip)
375 * Switch out to protocol's input routine.
377 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen, ip->ip_p);
381 transport_processing_handler(netmsg_t netmsg)
383 struct netmsg_packet *pmsg = (struct netmsg_packet *)netmsg;
387 ip = mtod(pmsg->nm_packet, struct ip *);
388 hlen = pmsg->nm_netmsg.nm_lmsg.u.ms_result;
390 transport_processing_oncpu(pmsg->nm_packet, hlen, ip);
391 /* netmsg was embedded in the mbuf, do not reply! */
395 ip_input_handler(struct netmsg *msg0)
397 struct mbuf *m = ((struct netmsg_packet *)msg0)->nm_packet;
400 /* msg0 was embedded in the mbuf, do not reply! */
404 * IP input routine. Checksum and byte swap header. If fragmented
405 * try to reassemble. Process options. Pass to next level.
408 ip_input(struct mbuf *m)
411 struct in_ifaddr *ia = NULL;
412 struct in_ifaddr_container *iac;
415 struct in_addr pkt_dst;
416 boolean_t using_srcrt = FALSE; /* forward (by PFIL_HOOKS) */
417 boolean_t needredispatch = FALSE;
418 struct in_addr odst; /* original dst address(NAT) */
420 struct sockaddr_in *next_hop = NULL;
422 struct tdb_ident *tdbi;
423 struct secpolicy *sp;
429 if (m->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) {
431 mtag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL);
432 KKASSERT(mtag != NULL);
433 next_hop = m_tag_data(mtag);
436 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
437 /* dummynet already filtered us */
438 ip = mtod(m, struct ip *);
439 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
445 /* length checks already done in ip_demux() */
446 KASSERT(m->m_len >= sizeof(ip), ("IP header not in one mbuf"));
448 ip = mtod(m, struct ip *);
450 if (IP_VHL_V(ip->ip_vhl) != IPVERSION) {
451 ipstat.ips_badvers++;
455 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
456 /* length checks already done in ip_demux() */
457 KASSERT(hlen >= sizeof(struct ip), ("IP header len too small"));
458 KASSERT(m->m_len >= hlen, ("packet shorter than IP header length"));
460 /* 127/8 must not appear on wire - RFC1122 */
461 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
462 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
463 if (!(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK)) {
464 ipstat.ips_badaddr++;
469 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
470 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
472 if (hlen == sizeof(struct ip)) {
473 sum = in_cksum_hdr(ip);
475 sum = in_cksum(m, hlen);
484 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) {
485 /* packet is dropped by traffic conditioner */
490 * Convert fields to host representation.
492 ip->ip_len = ntohs(ip->ip_len);
493 if (ip->ip_len < hlen) {
497 ip->ip_off = ntohs(ip->ip_off);
500 * Check that the amount of data in the buffers
501 * is as at least much as the IP header would have us expect.
502 * Trim mbufs if longer than we expect.
503 * Drop packet if shorter than we expect.
505 if (m->m_pkthdr.len < ip->ip_len) {
506 ipstat.ips_tooshort++;
509 if (m->m_pkthdr.len > ip->ip_len) {
510 if (m->m_len == m->m_pkthdr.len) {
511 m->m_len = ip->ip_len;
512 m->m_pkthdr.len = ip->ip_len;
514 m_adj(m, ip->ip_len - m->m_pkthdr.len);
516 #if defined(IPSEC) && !defined(IPSEC_FILTERGIF)
518 * Bypass packet filtering for packets from a tunnel (gif).
520 if (ipsec_gethist(m, NULL))
526 * Right now when no processing on packet has done
527 * and it is still fresh out of network we do our black
529 * - Firewall: deny/allow/divert
530 * - Xlate: translate packet's addr/port (NAT).
531 * - Pipe: pass pkt through dummynet.
532 * - Wrap: fake packet's addr/port <unimpl.>
533 * - Encapsulate: put it in another IP and send out. <unimp.>
538 * If we've been forwarded from the output side, then
539 * skip the firewall a second time
541 if (next_hop != NULL)
545 if (!pfil_has_hooks(&inet_pfil_hook)) {
546 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
548 * Strip dummynet tags from stranded packets
550 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
551 KKASSERT(mtag != NULL);
552 m_tag_delete(m, mtag);
553 m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED;
559 * Run through list of hooks for input packets.
561 * NB: Beware of the destination address changing (e.g.
562 * by NAT rewriting). When this happens, tell
563 * ip_forward to do the right thing.
566 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif, PFIL_IN))
568 if (m == NULL) /* consumed by filter */
570 ip = mtod(m, struct ip *);
571 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
572 using_srcrt = (odst.s_addr != ip->ip_dst.s_addr);
574 if (m->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) {
575 mtag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL);
576 KKASSERT(mtag != NULL);
577 next_hop = m_tag_data(mtag);
579 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
583 if (m->m_pkthdr.fw_flags & FW_MBUF_REDISPATCH) {
584 needredispatch = TRUE;
585 m->m_pkthdr.fw_flags &= ~FW_MBUF_REDISPATCH;
589 * Process options and, if not destined for us,
590 * ship it on. ip_dooptions returns 1 when an
591 * error was detected (causing an icmp message
592 * to be sent and the original packet to be freed).
594 if (hlen > sizeof(struct ip) && ip_dooptions(m, 0, next_hop))
597 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
598 * matter if it is destined to another node, or whether it is
599 * a multicast one, RSVP wants it! and prevents it from being forwarded
600 * anywhere else. Also checks if the rsvp daemon is running before
601 * grabbing the packet.
603 if (rsvp_on && ip->ip_p == IPPROTO_RSVP)
607 * Check our list of addresses, to see if the packet is for us.
608 * If we don't have any addresses, assume any unicast packet
609 * we receive might be for us (and let the upper layers deal
612 if (TAILQ_EMPTY(&in_ifaddrheads[mycpuid]) &&
613 !(m->m_flags & (M_MCAST | M_BCAST)))
617 * Cache the destination address of the packet; this may be
618 * changed by use of 'ipfw fwd'.
620 pkt_dst = next_hop ? next_hop->sin_addr : ip->ip_dst;
623 * Enable a consistency check between the destination address
624 * and the arrival interface for a unicast packet (the RFC 1122
625 * strong ES model) if IP forwarding is disabled and the packet
626 * is not locally generated and the packet is not subject to
629 * XXX - Checking also should be disabled if the destination
630 * address is ipnat'ed to a different interface.
632 * XXX - Checking is incompatible with IP aliases added
633 * to the loopback interface instead of the interface where
634 * the packets are received.
636 checkif = ip_checkinterface &&
638 m->m_pkthdr.rcvif != NULL &&
639 !(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) &&
643 * Check for exact addresses in the hash bucket.
645 LIST_FOREACH(iac, INADDR_HASH(pkt_dst.s_addr), ia_hash) {
649 * If the address matches, verify that the packet
650 * arrived via the correct interface if checking is
653 if (IA_SIN(ia)->sin_addr.s_addr == pkt_dst.s_addr &&
654 (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif))
660 * Check for broadcast addresses.
662 * Only accept broadcast packets that arrive via the matching
663 * interface. Reception of forwarded directed broadcasts would
664 * be handled via ip_forward() and ether_output() with the loopback
665 * into the stack for SIMPLEX interfaces handled by ether_output().
667 if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
668 struct ifaddr_container *ifac;
670 TAILQ_FOREACH(ifac, &m->m_pkthdr.rcvif->if_addrheads[mycpuid],
672 struct ifaddr *ifa = ifac->ifa;
674 if (ifa->ifa_addr == NULL) /* shutdown/startup race */
676 if (ifa->ifa_addr->sa_family != AF_INET)
679 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
682 if (ia->ia_netbroadcast.s_addr == pkt_dst.s_addr)
685 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY)
690 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
691 struct in_multi *inm;
693 if (ip_mrouter != NULL) {
695 * If we are acting as a multicast router, all
696 * incoming multicast packets are passed to the
697 * kernel-level multicast forwarding function.
698 * The packet is returned (relatively) intact; if
699 * ip_mforward() returns a non-zero value, the packet
700 * must be discarded, else it may be accepted below.
702 if (ip_mforward != NULL &&
703 ip_mforward(ip, m->m_pkthdr.rcvif, m, NULL) != 0) {
704 ipstat.ips_cantforward++;
710 * The process-level routing daemon needs to receive
711 * all multicast IGMP packets, whether or not this
712 * host belongs to their destination groups.
714 if (ip->ip_p == IPPROTO_IGMP)
716 ipstat.ips_forward++;
719 * See if we belong to the destination multicast group on the
722 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
724 ipstat.ips_notmember++;
730 if (ip->ip_dst.s_addr == INADDR_BROADCAST)
732 if (ip->ip_dst.s_addr == INADDR_ANY)
736 * FAITH(Firewall Aided Internet Translator)
738 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
740 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP)
748 * Not for us; forward if possible and desirable.
751 ipstat.ips_cantforward++;
756 * Enforce inbound IPsec SPD.
758 if (ipsec4_in_reject(m, NULL)) {
759 ipsecstat.in_polvio++;
764 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
767 tdbi = (struct tdb_ident *)m_tag_data(mtag);
768 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
770 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
771 IP_FORWARDING, &error);
773 if (sp == NULL) { /* NB: can happen if error */
775 /*XXX error stat???*/
776 DPRINTF(("ip_input: no SP for forwarding\n")); /*XXX*/
781 * Check security policy against packet attributes.
783 error = ipsec_in_reject(sp, m);
787 ipstat.ips_cantforward++;
791 ip_forward(m, using_srcrt, next_hop);
798 * IPSTEALTH: Process non-routing options only
799 * if the packet is destined for us.
802 hlen > sizeof(struct ip) &&
803 ip_dooptions(m, 1, next_hop))
806 /* Count the packet in the ip address stats */
808 ia->ia_ifa.if_ipackets++;
809 ia->ia_ifa.if_ibytes += m->m_pkthdr.len;
813 * If offset or IP_MF are set, must reassemble.
814 * Otherwise, nothing need be done.
815 * (We could look in the reassembly queue to see
816 * if the packet was previously fragmented,
817 * but it's not worth the time; just let them time out.)
819 if (ip->ip_off & (IP_MF | IP_OFFMASK)) {
821 * Attempt reassembly; if it succeeds, proceed.
822 * ip_reass() will return a different mbuf.
827 ip = mtod(m, struct ip *);
829 /* Get the header length of the reassembled packet */
830 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
832 needredispatch = TRUE;
839 * enforce IPsec policy checking if we are seeing last header.
840 * note that we do not visit this with protocols with pcb layer
841 * code - like udp/tcp/raw ip.
843 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) &&
844 ipsec4_in_reject(m, NULL)) {
845 ipsecstat.in_polvio++;
851 * enforce IPsec policy checking if we are seeing last header.
852 * note that we do not visit this with protocols with pcb layer
853 * code - like udp/tcp/raw ip.
855 if (inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) {
857 * Check if the packet has already had IPsec processing
858 * done. If so, then just pass it along. This tag gets
859 * set during AH, ESP, etc. input handling, before the
860 * packet is returned to the ip input queue for delivery.
862 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
865 tdbi = (struct tdb_ident *)m_tag_data(mtag);
866 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
868 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
869 IP_FORWARDING, &error);
873 * Check security policy against packet attributes.
875 error = ipsec_in_reject(sp, m);
878 /* XXX error stat??? */
880 DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/
887 #endif /* FAST_IPSEC */
889 ipstat.ips_delivered++;
890 if (needredispatch) {
891 struct netmsg_packet *pmsg;
894 ip->ip_off = htons(ip->ip_off);
895 ip->ip_len = htons(ip->ip_len);
896 port = ip_mport_in(&m);
900 pmsg = &m->m_hdr.mh_netmsg;
901 netmsg_init(&pmsg->nm_netmsg, &netisr_apanic_rport, 0,
902 transport_processing_handler);
904 pmsg->nm_netmsg.nm_lmsg.u.ms_result = hlen;
906 ip = mtod(m, struct ip *);
907 ip->ip_len = ntohs(ip->ip_len);
908 ip->ip_off = ntohs(ip->ip_off);
909 lwkt_sendmsg(port, &pmsg->nm_netmsg.nm_lmsg);
911 transport_processing_oncpu(m, hlen, ip);
920 * Take incoming datagram fragment and try to reassemble it into
921 * whole datagram. If a chain for reassembly of this datagram already
922 * exists, then it is given as fp; otherwise have to make a chain.
925 ip_reass(struct mbuf *m)
927 struct ip *ip = mtod(m, struct ip *);
928 struct mbuf *p = NULL, *q, *nq;
930 struct ipq *fp = NULL;
931 int hlen = IP_VHL_HL(ip->ip_vhl) << 2;
935 /* If maxnipq is 0, never accept fragments. */
937 ipstat.ips_fragments++;
938 ipstat.ips_fragdropped++;
943 sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
945 * Look for queue of fragments of this datagram.
947 for (fp = ipq[sum].next; fp != &ipq[sum]; fp = fp->next)
948 if (ip->ip_id == fp->ipq_id &&
949 ip->ip_src.s_addr == fp->ipq_src.s_addr &&
950 ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
951 ip->ip_p == fp->ipq_p)
957 * Enforce upper bound on number of fragmented packets
958 * for which we attempt reassembly;
959 * If maxnipq is -1, accept all fragments without limitation.
961 if (nipq > maxnipq && maxnipq > 0) {
963 * drop something from the tail of the current queue
964 * before proceeding further
966 if (ipq[sum].prev == &ipq[sum]) { /* gak */
967 for (i = 0; i < IPREASS_NHASH; i++) {
968 if (ipq[i].prev != &ipq[i]) {
969 ipstat.ips_fragtimeout +=
970 ipq[i].prev->ipq_nfrags;
971 ip_freef(ipq[i].prev);
976 ipstat.ips_fragtimeout +=
977 ipq[sum].prev->ipq_nfrags;
978 ip_freef(ipq[sum].prev);
983 * Adjust ip_len to not reflect header,
984 * convert offset of this to bytes.
987 if (ip->ip_off & IP_MF) {
989 * Make sure that fragments have a data length
990 * that's a non-zero multiple of 8 bytes.
992 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
993 ipstat.ips_toosmall++; /* XXX */
997 m->m_flags |= M_FRAG;
999 m->m_flags &= ~M_FRAG;
1002 ipstat.ips_fragments++;
1003 m->m_pkthdr.header = ip;
1006 * If the hardware has not done csum over this fragment
1007 * then csum_data is not valid at all.
1009 if ((m->m_pkthdr.csum_flags & (CSUM_FRAG_NOT_CHECKED | CSUM_DATA_VALID))
1010 == (CSUM_FRAG_NOT_CHECKED | CSUM_DATA_VALID)) {
1011 m->m_pkthdr.csum_data = 0;
1012 m->m_pkthdr.csum_flags &= ~(CSUM_DATA_VALID | CSUM_PSEUDO_HDR);
1016 * Presence of header sizes in mbufs
1017 * would confuse code below.
1023 * If first fragment to arrive, create a reassembly queue.
1026 if ((fp = mpipe_alloc_nowait(&ipq_mpipe)) == NULL)
1028 insque(fp, &ipq[sum]);
1031 fp->ipq_ttl = IPFRAGTTL;
1032 fp->ipq_p = ip->ip_p;
1033 fp->ipq_id = ip->ip_id;
1034 fp->ipq_src = ip->ip_src;
1035 fp->ipq_dst = ip->ip_dst;
1037 m->m_nextpkt = NULL;
1043 #define GETIP(m) ((struct ip*)((m)->m_pkthdr.header))
1046 * Find a segment which begins after this one does.
1048 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt)
1049 if (GETIP(q)->ip_off > ip->ip_off)
1053 * If there is a preceding segment, it may provide some of
1054 * our data already. If so, drop the data from the incoming
1055 * segment. If it provides all of our data, drop us, otherwise
1056 * stick new segment in the proper place.
1058 * If some of the data is dropped from the the preceding
1059 * segment, then it's checksum is invalidated.
1062 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off;
1064 if (i >= ip->ip_len)
1067 m->m_pkthdr.csum_flags = 0;
1071 m->m_nextpkt = p->m_nextpkt;
1074 m->m_nextpkt = fp->ipq_frags;
1079 * While we overlap succeeding segments trim them or,
1080 * if they are completely covered, dequeue them.
1082 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off;
1084 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off;
1085 if (i < GETIP(q)->ip_len) {
1086 GETIP(q)->ip_len -= i;
1087 GETIP(q)->ip_off += i;
1089 q->m_pkthdr.csum_flags = 0;
1094 ipstat.ips_fragdropped++;
1096 q->m_nextpkt = NULL;
1102 * Check for complete reassembly and perform frag per packet
1105 * Frag limiting is performed here so that the nth frag has
1106 * a chance to complete the packet before we drop the packet.
1107 * As a result, n+1 frags are actually allowed per packet, but
1108 * only n will ever be stored. (n = maxfragsperpacket.)
1112 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
1113 if (GETIP(q)->ip_off != next) {
1114 if (fp->ipq_nfrags > maxfragsperpacket) {
1115 ipstat.ips_fragdropped += fp->ipq_nfrags;
1120 next += GETIP(q)->ip_len;
1122 /* Make sure the last packet didn't have the IP_MF flag */
1123 if (p->m_flags & M_FRAG) {
1124 if (fp->ipq_nfrags > maxfragsperpacket) {
1125 ipstat.ips_fragdropped += fp->ipq_nfrags;
1132 * Reassembly is complete. Make sure the packet is a sane size.
1136 if (next + (IP_VHL_HL(ip->ip_vhl) << 2) > IP_MAXPACKET) {
1137 ipstat.ips_toolong++;
1138 ipstat.ips_fragdropped += fp->ipq_nfrags;
1144 * Concatenate fragments.
1151 q->m_nextpkt = NULL;
1152 for (q = nq; q != NULL; q = nq) {
1154 q->m_nextpkt = NULL;
1155 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
1156 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
1161 * Clean up the 1's complement checksum. Carry over 16 bits must
1162 * be added back. This assumes no more then 65535 packet fragments
1163 * were reassembled. A second carry can also occur (but not a third).
1165 m->m_pkthdr.csum_data = (m->m_pkthdr.csum_data & 0xffff) +
1166 (m->m_pkthdr.csum_data >> 16);
1167 if (m->m_pkthdr.csum_data > 0xFFFF)
1168 m->m_pkthdr.csum_data -= 0xFFFF;
1171 * Create header for new ip packet by
1172 * modifying header of first packet;
1173 * dequeue and discard fragment reassembly header.
1174 * Make header visible.
1177 ip->ip_src = fp->ipq_src;
1178 ip->ip_dst = fp->ipq_dst;
1181 mpipe_free(&ipq_mpipe, fp);
1182 m->m_len += (IP_VHL_HL(ip->ip_vhl) << 2);
1183 m->m_data -= (IP_VHL_HL(ip->ip_vhl) << 2);
1184 /* some debugging cruft by sklower, below, will go away soon */
1185 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */
1188 for (n = m; n; n = n->m_next)
1190 m->m_pkthdr.len = plen;
1193 ipstat.ips_reassembled++;
1197 ipstat.ips_fragdropped++;
1207 * Free a fragment reassembly header and all
1208 * associated datagrams.
1211 ip_freef(struct ipq *fp)
1215 while (fp->ipq_frags) {
1217 fp->ipq_frags = q->m_nextpkt;
1218 q->m_nextpkt = NULL;
1222 mpipe_free(&ipq_mpipe, fp);
1227 * IP timer processing;
1228 * if a timer expires on a reassembly
1229 * queue, discard it.
1238 for (i = 0; i < IPREASS_NHASH; i++) {
1242 while (fp != &ipq[i]) {
1245 if (fp->prev->ipq_ttl == 0) {
1246 ipstat.ips_fragtimeout += fp->prev->ipq_nfrags;
1252 * If we are over the maximum number of fragments
1253 * (due to the limit being lowered), drain off
1254 * enough to get down to the new limit.
1256 if (maxnipq >= 0 && nipq > maxnipq) {
1257 for (i = 0; i < IPREASS_NHASH; i++) {
1258 while (nipq > maxnipq &&
1259 (ipq[i].next != &ipq[i])) {
1260 ipstat.ips_fragdropped +=
1261 ipq[i].next->ipq_nfrags;
1262 ip_freef(ipq[i].next);
1271 * Drain off all datagram fragments.
1278 for (i = 0; i < IPREASS_NHASH; i++) {
1279 while (ipq[i].next != &ipq[i]) {
1280 ipstat.ips_fragdropped += ipq[i].next->ipq_nfrags;
1281 ip_freef(ipq[i].next);
1288 * Do option processing on a datagram,
1289 * possibly discarding it if bad options are encountered,
1290 * or forwarding it if source-routed.
1291 * The pass argument is used when operating in the IPSTEALTH
1292 * mode to tell what options to process:
1293 * [LS]SRR (pass 0) or the others (pass 1).
1294 * The reason for as many as two passes is that when doing IPSTEALTH,
1295 * non-routing options should be processed only if the packet is for us.
1296 * Returns 1 if packet has been forwarded/freed,
1297 * 0 if the packet should be processed further.
1300 ip_dooptions(struct mbuf *m, int pass, struct sockaddr_in *next_hop)
1302 struct sockaddr_in ipaddr = { sizeof ipaddr, AF_INET };
1303 struct ip *ip = mtod(m, struct ip *);
1305 struct in_ifaddr *ia;
1306 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB;
1307 boolean_t forward = FALSE;
1308 struct in_addr *sin, dst;
1312 cp = (u_char *)(ip + 1);
1313 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
1314 for (; cnt > 0; cnt -= optlen, cp += optlen) {
1315 opt = cp[IPOPT_OPTVAL];
1316 if (opt == IPOPT_EOL)
1318 if (opt == IPOPT_NOP)
1321 if (cnt < IPOPT_OLEN + sizeof(*cp)) {
1322 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1325 optlen = cp[IPOPT_OLEN];
1326 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
1327 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1337 * Source routing with record.
1338 * Find interface with current destination address.
1339 * If none on this machine then drop if strictly routed,
1340 * or do nothing if loosely routed.
1341 * Record interface address and bring up next address
1342 * component. If strictly routed make sure next
1343 * address is on directly accessible net.
1347 if (ipstealth && pass > 0)
1349 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1350 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1353 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1354 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1357 ipaddr.sin_addr = ip->ip_dst;
1358 ia = (struct in_ifaddr *)
1359 ifa_ifwithaddr((struct sockaddr *)&ipaddr);
1361 if (opt == IPOPT_SSRR) {
1362 type = ICMP_UNREACH;
1363 code = ICMP_UNREACH_SRCFAIL;
1366 if (!ip_dosourceroute)
1367 goto nosourcerouting;
1369 * Loose routing, and not at next destination
1370 * yet; nothing to do except forward.
1374 off--; /* 0 origin */
1375 if (off > optlen - (int)sizeof(struct in_addr)) {
1377 * End of source route. Should be for us.
1379 if (!ip_acceptsourceroute)
1380 goto nosourcerouting;
1381 save_rte(m, cp, ip->ip_src);
1386 if (!ip_dosourceroute) {
1388 char buf[sizeof "aaa.bbb.ccc.ddd"];
1391 * Acting as a router, so generate ICMP
1394 strcpy(buf, inet_ntoa(ip->ip_dst));
1396 "attempted source route from %s to %s\n",
1397 inet_ntoa(ip->ip_src), buf);
1398 type = ICMP_UNREACH;
1399 code = ICMP_UNREACH_SRCFAIL;
1403 * Not acting as a router,
1407 ipstat.ips_cantforward++;
1414 * locate outgoing interface
1416 memcpy(&ipaddr.sin_addr, cp + off,
1417 sizeof ipaddr.sin_addr);
1419 if (opt == IPOPT_SSRR) {
1420 #define INA struct in_ifaddr *
1421 #define SA struct sockaddr *
1422 if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr))
1424 ia = (INA)ifa_ifwithnet((SA)&ipaddr);
1426 ia = ip_rtaddr(ipaddr.sin_addr,
1427 &ipforward_rt[mycpuid]);
1429 type = ICMP_UNREACH;
1430 code = ICMP_UNREACH_SRCFAIL;
1433 ip->ip_dst = ipaddr.sin_addr;
1434 memcpy(cp + off, &IA_SIN(ia)->sin_addr,
1435 sizeof(struct in_addr));
1436 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1438 * Let ip_intr's mcast routing check handle mcast pkts
1440 forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr));
1444 if (ipstealth && pass == 0)
1446 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1447 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1450 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1451 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1455 * If no space remains, ignore.
1457 off--; /* 0 origin */
1458 if (off > optlen - (int)sizeof(struct in_addr))
1460 memcpy(&ipaddr.sin_addr, &ip->ip_dst,
1461 sizeof ipaddr.sin_addr);
1463 * locate outgoing interface; if we're the destination,
1464 * use the incoming interface (should be same).
1466 if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == NULL &&
1467 (ia = ip_rtaddr(ipaddr.sin_addr,
1468 &ipforward_rt[mycpuid]))
1470 type = ICMP_UNREACH;
1471 code = ICMP_UNREACH_HOST;
1474 memcpy(cp + off, &IA_SIN(ia)->sin_addr,
1475 sizeof(struct in_addr));
1476 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1480 if (ipstealth && pass == 0)
1482 code = cp - (u_char *)ip;
1483 if (optlen < 4 || optlen > 40) {
1484 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1487 if ((off = cp[IPOPT_OFFSET]) < 5) {
1488 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1491 if (off > optlen - (int)sizeof(int32_t)) {
1492 cp[IPOPT_OFFSET + 1] += (1 << 4);
1493 if ((cp[IPOPT_OFFSET + 1] & 0xf0) == 0) {
1494 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1499 off--; /* 0 origin */
1500 sin = (struct in_addr *)(cp + off);
1501 switch (cp[IPOPT_OFFSET + 1] & 0x0f) {
1503 case IPOPT_TS_TSONLY:
1506 case IPOPT_TS_TSANDADDR:
1507 if (off + sizeof(n_time) +
1508 sizeof(struct in_addr) > optlen) {
1509 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1512 ipaddr.sin_addr = dst;
1513 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr,
1517 memcpy(sin, &IA_SIN(ia)->sin_addr,
1518 sizeof(struct in_addr));
1519 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1520 off += sizeof(struct in_addr);
1523 case IPOPT_TS_PRESPEC:
1524 if (off + sizeof(n_time) +
1525 sizeof(struct in_addr) > optlen) {
1526 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1529 memcpy(&ipaddr.sin_addr, sin,
1530 sizeof(struct in_addr));
1531 if (ifa_ifwithaddr((SA)&ipaddr) == NULL)
1533 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1534 off += sizeof(struct in_addr);
1538 code = &cp[IPOPT_OFFSET + 1] - (u_char *)ip;
1542 memcpy(cp + off, &ntime, sizeof(n_time));
1543 cp[IPOPT_OFFSET] += sizeof(n_time);
1546 if (forward && ipforwarding) {
1547 ip_forward(m, TRUE, next_hop);
1552 icmp_error(m, type, code, 0, 0);
1553 ipstat.ips_badoptions++;
1558 * Given address of next destination (final or next hop),
1559 * return internet address info of interface to be used to get there.
1562 ip_rtaddr(struct in_addr dst, struct route *ro)
1564 struct sockaddr_in *sin;
1566 sin = (struct sockaddr_in *)&ro->ro_dst;
1568 if (ro->ro_rt == NULL || dst.s_addr != sin->sin_addr.s_addr) {
1569 if (ro->ro_rt != NULL) {
1573 sin->sin_family = AF_INET;
1574 sin->sin_len = sizeof *sin;
1575 sin->sin_addr = dst;
1576 rtalloc_ign(ro, RTF_PRCLONING);
1579 if (ro->ro_rt == NULL)
1582 return (ifatoia(ro->ro_rt->rt_ifa));
1586 * Save incoming source route for use in replies,
1587 * to be picked up later by ip_srcroute if the receiver is interested.
1590 save_rte(struct mbuf *m, u_char *option, struct in_addr dst)
1593 struct ip_srcrt_opt *opt;
1596 mtag = m_tag_get(PACKET_TAG_IPSRCRT, sizeof(*opt), MB_DONTWAIT);
1599 opt = m_tag_data(mtag);
1601 olen = option[IPOPT_OLEN];
1604 kprintf("save_rte: olen %d\n", olen);
1606 if (olen > sizeof(opt->ip_srcrt) - (1 + sizeof(dst))) {
1610 bcopy(option, opt->ip_srcrt.srcopt, olen);
1611 opt->ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1612 opt->ip_srcrt.dst = dst;
1613 m_tag_prepend(m, mtag);
1617 * Retrieve incoming source route for use in replies,
1618 * in the same form used by setsockopt.
1619 * The first hop is placed before the options, will be removed later.
1622 ip_srcroute(struct mbuf *m0)
1624 struct in_addr *p, *q;
1627 struct ip_srcrt_opt *opt;
1632 mtag = m_tag_find(m0, PACKET_TAG_IPSRCRT, NULL);
1635 opt = m_tag_data(mtag);
1637 if (opt->ip_nhops == 0)
1639 m = m_get(MB_DONTWAIT, MT_HEADER);
1643 #define OPTSIZ (sizeof(opt->ip_srcrt.nop) + sizeof(opt->ip_srcrt.srcopt))
1645 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
1646 m->m_len = opt->ip_nhops * sizeof(struct in_addr) +
1647 sizeof(struct in_addr) + OPTSIZ;
1650 kprintf("ip_srcroute: nhops %d mlen %d",
1651 opt->ip_nhops, m->m_len);
1656 * First save first hop for return route
1658 p = &opt->ip_srcrt.route[opt->ip_nhops - 1];
1659 *(mtod(m, struct in_addr *)) = *p--;
1662 kprintf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr));
1666 * Copy option fields and padding (nop) to mbuf.
1668 opt->ip_srcrt.nop = IPOPT_NOP;
1669 opt->ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
1670 memcpy(mtod(m, caddr_t) + sizeof(struct in_addr), &opt->ip_srcrt.nop,
1672 q = (struct in_addr *)(mtod(m, caddr_t) +
1673 sizeof(struct in_addr) + OPTSIZ);
1676 * Record return path as an IP source route,
1677 * reversing the path (pointers are now aligned).
1679 while (p >= opt->ip_srcrt.route) {
1682 kprintf(" %x", ntohl(q->s_addr));
1687 * Last hop goes to final destination.
1689 *q = opt->ip_srcrt.dst;
1690 m_tag_delete(m0, mtag);
1693 kprintf(" %x\n", ntohl(q->s_addr));
1699 * Strip out IP options.
1702 ip_stripoptions(struct mbuf *m)
1705 struct ip *ip = mtod(m, struct ip *);
1709 optlen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
1710 opts = (caddr_t)(ip + 1);
1711 datalen = m->m_len - (sizeof(struct ip) + optlen);
1712 bcopy(opts + optlen, opts, datalen);
1714 if (m->m_flags & M_PKTHDR)
1715 m->m_pkthdr.len -= optlen;
1716 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, sizeof(struct ip) >> 2);
1719 u_char inetctlerrmap[PRC_NCMDS] = {
1721 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
1722 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
1723 EMSGSIZE, EHOSTUNREACH, 0, 0,
1725 ENOPROTOOPT, ECONNREFUSED
1729 * Forward a packet. If some error occurs return the sender
1730 * an icmp packet. Note we can't always generate a meaningful
1731 * icmp message because icmp doesn't have a large enough repertoire
1732 * of codes and types.
1734 * If not forwarding, just drop the packet. This could be confusing
1735 * if ipforwarding was zero but some routing protocol was advancing
1736 * us as a gateway to somewhere. However, we must let the routing
1737 * protocol deal with that.
1739 * The using_srcrt parameter indicates whether the packet is being forwarded
1740 * via a source route.
1743 ip_forward(struct mbuf *m, boolean_t using_srcrt, struct sockaddr_in *next_hop)
1745 struct ip *ip = mtod(m, struct ip *);
1746 struct sockaddr_in *ipforward_rtaddr;
1748 int error, type = 0, code = 0, destmtu = 0;
1751 struct in_addr pkt_dst;
1752 struct route *cache_rt = &ipforward_rt[mycpuid];
1756 * Cache the destination address of the packet; this may be
1757 * changed by use of 'ipfw fwd'.
1759 pkt_dst = (next_hop != NULL) ? next_hop->sin_addr : ip->ip_dst;
1763 kprintf("forward: src %x dst %x ttl %x\n",
1764 ip->ip_src.s_addr, pkt_dst.s_addr, ip->ip_ttl);
1767 if (m->m_flags & (M_BCAST | M_MCAST) || !in_canforward(pkt_dst)) {
1768 ipstat.ips_cantforward++;
1772 if (!ipstealth && ip->ip_ttl <= IPTTLDEC) {
1773 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0);
1777 ipforward_rtaddr = (struct sockaddr_in *) &cache_rt->ro_dst;
1778 if (cache_rt->ro_rt == NULL ||
1779 ipforward_rtaddr->sin_addr.s_addr != pkt_dst.s_addr) {
1780 if (cache_rt->ro_rt != NULL) {
1781 RTFREE(cache_rt->ro_rt);
1782 cache_rt->ro_rt = NULL;
1784 ipforward_rtaddr->sin_family = AF_INET;
1785 ipforward_rtaddr->sin_len = sizeof(struct sockaddr_in);
1786 ipforward_rtaddr->sin_addr = pkt_dst;
1787 rtalloc_ign(cache_rt, RTF_PRCLONING);
1788 if (cache_rt->ro_rt == NULL) {
1789 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0);
1793 rt = cache_rt->ro_rt;
1796 * Save the IP header and at most 8 bytes of the payload,
1797 * in case we need to generate an ICMP message to the src.
1799 * XXX this can be optimized a lot by saving the data in a local
1800 * buffer on the stack (72 bytes at most), and only allocating the
1801 * mbuf if really necessary. The vast majority of the packets
1802 * are forwarded without having to send an ICMP back (either
1803 * because unnecessary, or because rate limited), so we are
1804 * really we are wasting a lot of work here.
1806 * We don't use m_copy() because it might return a reference
1807 * to a shared cluster. Both this function and ip_output()
1808 * assume exclusive access to the IP header in `m', so any
1809 * data in a cluster may change before we reach icmp_error().
1811 MGETHDR(mcopy, MB_DONTWAIT, m->m_type);
1812 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, MB_DONTWAIT)) {
1814 * It's probably ok if the pkthdr dup fails (because
1815 * the deep copy of the tag chain failed), but for now
1816 * be conservative and just discard the copy since
1817 * code below may some day want the tags.
1822 if (mcopy != NULL) {
1823 mcopy->m_len = imin((IP_VHL_HL(ip->ip_vhl) << 2) + 8,
1825 mcopy->m_pkthdr.len = mcopy->m_len;
1826 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1830 ip->ip_ttl -= IPTTLDEC;
1833 * If forwarding packet using same interface that it came in on,
1834 * perhaps should send a redirect to sender to shortcut a hop.
1835 * Only send redirect if source is sending directly to us,
1836 * and if packet was not source routed (or has any options).
1837 * Also, don't send redirect if forwarding using a default route
1838 * or a route modified by a redirect.
1840 if (rt->rt_ifp == m->m_pkthdr.rcvif &&
1841 !(rt->rt_flags & (RTF_DYNAMIC | RTF_MODIFIED)) &&
1842 satosin(rt_key(rt))->sin_addr.s_addr != INADDR_ANY &&
1843 ipsendredirects && !using_srcrt && next_hop == NULL) {
1844 u_long src = ntohl(ip->ip_src.s_addr);
1845 struct in_ifaddr *rt_ifa = (struct in_ifaddr *)rt->rt_ifa;
1847 if (rt_ifa != NULL &&
1848 (src & rt_ifa->ia_subnetmask) == rt_ifa->ia_subnet) {
1849 if (rt->rt_flags & RTF_GATEWAY)
1850 dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
1852 dest = pkt_dst.s_addr;
1854 * Router requirements says to only send
1857 type = ICMP_REDIRECT;
1858 code = ICMP_REDIRECT_HOST;
1861 kprintf("redirect (%d) to %x\n", code, dest);
1866 error = ip_output(m, NULL, cache_rt, IP_FORWARDING, NULL, NULL);
1868 ipstat.ips_forward++;
1871 ipflow_create(cache_rt, mcopy);
1874 return; /* most common case */
1876 ipstat.ips_redirectsent++;
1879 ipstat.ips_cantforward++;
1886 * Send ICMP message.
1891 case 0: /* forwarded, but need redirect */
1892 /* type, code set above */
1895 case ENETUNREACH: /* shouldn't happen, checked above */
1900 type = ICMP_UNREACH;
1901 code = ICMP_UNREACH_HOST;
1905 type = ICMP_UNREACH;
1906 code = ICMP_UNREACH_NEEDFRAG;
1909 * If the packet is routed over IPsec tunnel, tell the
1910 * originator the tunnel MTU.
1911 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
1914 if (cache_rt->ro_rt != NULL) {
1915 struct secpolicy *sp = NULL;
1920 sp = ipsec4_getpolicybyaddr(mcopy,
1926 destmtu = cache_rt->ro_rt->rt_ifp->if_mtu;
1928 /* count IPsec header size */
1929 ipsechdr = ipsec4_hdrsiz(mcopy,
1934 * find the correct route for outer IPv4
1935 * header, compute tunnel MTU.
1938 if (sp->req != NULL && sp->req->sav != NULL &&
1939 sp->req->sav->sah != NULL) {
1940 ro = &sp->req->sav->sah->sa_route;
1941 if (ro->ro_rt != NULL &&
1942 ro->ro_rt->rt_ifp != NULL) {
1944 ro->ro_rt->rt_ifp->if_mtu;
1945 destmtu -= ipsechdr;
1954 * If the packet is routed over IPsec tunnel, tell the
1955 * originator the tunnel MTU.
1956 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
1959 if (cache_rt->ro_rt != NULL) {
1960 struct secpolicy *sp = NULL;
1965 sp = ipsec_getpolicybyaddr(mcopy,
1971 destmtu = cache_rt->ro_rt->rt_ifp->if_mtu;
1973 /* count IPsec header size */
1974 ipsechdr = ipsec4_hdrsiz(mcopy,
1979 * find the correct route for outer IPv4
1980 * header, compute tunnel MTU.
1983 if (sp->req != NULL &&
1984 sp->req->sav != NULL &&
1985 sp->req->sav->sah != NULL) {
1986 ro = &sp->req->sav->sah->sa_route;
1987 if (ro->ro_rt != NULL &&
1988 ro->ro_rt->rt_ifp != NULL) {
1990 ro->ro_rt->rt_ifp->if_mtu;
1991 destmtu -= ipsechdr;
1998 #else /* !IPSEC && !FAST_IPSEC */
1999 if (cache_rt->ro_rt != NULL)
2000 destmtu = cache_rt->ro_rt->rt_ifp->if_mtu;
2002 ipstat.ips_cantfrag++;
2007 * A router should not generate ICMP_SOURCEQUENCH as
2008 * required in RFC1812 Requirements for IP Version 4 Routers.
2009 * Source quench could be a big problem under DoS attacks,
2010 * or if the underlying interface is rate-limited.
2011 * Those who need source quench packets may re-enable them
2012 * via the net.inet.ip.sendsourcequench sysctl.
2014 if (!ip_sendsourcequench) {
2018 type = ICMP_SOURCEQUENCH;
2023 case EACCES: /* ipfw denied packet */
2027 icmp_error(mcopy, type, code, dest, destmtu);
2031 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
2034 if (inp->inp_socket->so_options & SO_TIMESTAMP) {
2038 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
2039 SCM_TIMESTAMP, SOL_SOCKET);
2041 mp = &(*mp)->m_next;
2043 if (inp->inp_flags & INP_RECVDSTADDR) {
2044 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
2045 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
2047 mp = &(*mp)->m_next;
2049 if (inp->inp_flags & INP_RECVTTL) {
2050 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl,
2051 sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
2053 mp = &(*mp)->m_next;
2057 * Moving these out of udp_input() made them even more broken
2058 * than they already were.
2060 /* options were tossed already */
2061 if (inp->inp_flags & INP_RECVOPTS) {
2062 *mp = sbcreatecontrol((caddr_t) opts_deleted_above,
2063 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
2065 mp = &(*mp)->m_next;
2067 /* ip_srcroute doesn't do what we want here, need to fix */
2068 if (inp->inp_flags & INP_RECVRETOPTS) {
2069 *mp = sbcreatecontrol((caddr_t) ip_srcroute(m),
2070 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
2072 mp = &(*mp)->m_next;
2075 if (inp->inp_flags & INP_RECVIF) {
2078 struct sockaddr_dl sdl;
2081 struct sockaddr_dl *sdp;
2082 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
2084 if (((ifp = m->m_pkthdr.rcvif)) &&
2085 ((ifp->if_index != 0) && (ifp->if_index <= if_index))) {
2086 sdp = IF_LLSOCKADDR(ifp);
2088 * Change our mind and don't try copy.
2090 if ((sdp->sdl_family != AF_LINK) ||
2091 (sdp->sdl_len > sizeof(sdlbuf))) {
2094 bcopy(sdp, sdl2, sdp->sdl_len);
2098 offsetof(struct sockaddr_dl, sdl_data[0]);
2099 sdl2->sdl_family = AF_LINK;
2100 sdl2->sdl_index = 0;
2101 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
2103 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len,
2104 IP_RECVIF, IPPROTO_IP);
2106 mp = &(*mp)->m_next;
2111 * XXX these routines are called from the upper part of the kernel.
2113 * They could also be moved to ip_mroute.c, since all the RSVP
2114 * handling is done there already.
2117 ip_rsvp_init(struct socket *so)
2119 if (so->so_type != SOCK_RAW ||
2120 so->so_proto->pr_protocol != IPPROTO_RSVP)
2123 if (ip_rsvpd != NULL)
2128 * This may seem silly, but we need to be sure we don't over-increment
2129 * the RSVP counter, in case something slips up.
2144 * This may seem silly, but we need to be sure we don't over-decrement
2145 * the RSVP counter, in case something slips up.
2155 rsvp_input(struct mbuf *m, ...) /* XXX must fixup manually */
2161 off = __va_arg(ap, int);
2162 proto = __va_arg(ap, int);
2165 if (rsvp_input_p) { /* call the real one if loaded */
2166 rsvp_input_p(m, off, proto);
2170 /* Can still get packets with rsvp_on = 0 if there is a local member
2171 * of the group to which the RSVP packet is addressed. But in this
2172 * case we want to throw the packet away.
2180 if (ip_rsvpd != NULL) {
2181 rip_input(m, off, proto);
2184 /* Drop the packet */