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 1.115 2008/10/28 07:09:26 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>
98 #include <sys/mplock2.h>
100 #include <machine/stdarg.h>
103 #include <net/if_types.h>
104 #include <net/if_var.h>
105 #include <net/if_dl.h>
106 #include <net/pfil.h>
107 #include <net/route.h>
108 #include <net/netisr.h>
110 #include <netinet/in.h>
111 #include <netinet/in_systm.h>
112 #include <netinet/in_var.h>
113 #include <netinet/ip.h>
114 #include <netinet/in_pcb.h>
115 #include <netinet/ip_var.h>
116 #include <netinet/ip_icmp.h>
117 #include <netinet/ip_divert.h>
118 #include <netinet/ip_flow.h>
120 #include <sys/thread2.h>
121 #include <sys/msgport2.h>
122 #include <net/netmsg2.h>
124 #include <sys/socketvar.h>
126 #include <net/ipfw/ip_fw.h>
127 #include <net/dummynet/ip_dummynet.h>
130 #include <netinet6/ipsec.h>
131 #include <netproto/key/key.h>
135 #include <netproto/ipsec/ipsec.h>
136 #include <netproto/ipsec/key.h>
140 static int ip_rsvp_on;
141 struct socket *ip_rsvpd;
143 int ipforwarding = 0;
144 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW,
145 &ipforwarding, 0, "Enable IP forwarding between interfaces");
147 static int ipsendredirects = 1; /* XXX */
148 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW,
149 &ipsendredirects, 0, "Enable sending IP redirects");
151 int ip_defttl = IPDEFTTL;
152 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW,
153 &ip_defttl, 0, "Maximum TTL on IP packets");
155 static int ip_dosourceroute = 0;
156 SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW,
157 &ip_dosourceroute, 0, "Enable forwarding source routed IP packets");
159 static int ip_acceptsourceroute = 0;
160 SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute,
161 CTLFLAG_RW, &ip_acceptsourceroute, 0,
162 "Enable accepting source routed IP packets");
164 static int ip_keepfaith = 0;
165 SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW,
167 "Enable packet capture for FAITH IPv4->IPv6 translator daemon");
169 static int nipq = 0; /* total # of reass queues */
171 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragpackets, CTLFLAG_RW,
173 "Maximum number of IPv4 fragment reassembly queue entries");
175 static int maxfragsperpacket;
176 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW,
177 &maxfragsperpacket, 0,
178 "Maximum number of IPv4 fragments allowed per packet");
180 static int ip_sendsourcequench = 0;
181 SYSCTL_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW,
182 &ip_sendsourcequench, 0,
183 "Enable the transmission of source quench packets");
185 int ip_do_randomid = 1;
186 SYSCTL_INT(_net_inet_ip, OID_AUTO, random_id, CTLFLAG_RW,
188 "Assign random ip_id values");
190 * XXX - Setting ip_checkinterface mostly implements the receive side of
191 * the Strong ES model described in RFC 1122, but since the routing table
192 * and transmit implementation do not implement the Strong ES model,
193 * setting this to 1 results in an odd hybrid.
195 * XXX - ip_checkinterface currently must be disabled if you use ipnat
196 * to translate the destination address to another local interface.
198 * XXX - ip_checkinterface must be disabled if you add IP aliases
199 * to the loopback interface instead of the interface where the
200 * packets for those addresses are received.
202 static int ip_checkinterface = 0;
203 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW,
204 &ip_checkinterface, 0, "Verify packet arrives on correct interface");
206 static int ip_dispatch_fast = 0;
207 static int ip_dispatch_slow = 0;
208 static int ip_dispatch_recheck = 0;
209 static int ip_dispatch_software = 0;
210 SYSCTL_INT(_net_inet_ip, OID_AUTO, dispatch_fast_count, CTLFLAG_RW,
211 &ip_dispatch_fast, 0, "");
212 SYSCTL_INT(_net_inet_ip, OID_AUTO, dispatch_slow_count, CTLFLAG_RW,
213 &ip_dispatch_slow, 0, "");
214 SYSCTL_INT(_net_inet_ip, OID_AUTO, dispatch_software_count, CTLFLAG_RW,
215 &ip_dispatch_software, 0, "");
216 SYSCTL_INT(_net_inet_ip, OID_AUTO, dispatch_recheck_count, CTLFLAG_RW,
217 &ip_dispatch_recheck, 0, "");
219 static struct lwkt_token ipq_token = LWKT_TOKEN_MP_INITIALIZER(ipq_token);
222 static int ipprintfs = 0;
225 extern struct domain inetdomain;
226 extern struct protosw inetsw[];
227 u_char ip_protox[IPPROTO_MAX];
228 struct in_ifaddrhead in_ifaddrheads[MAXCPU]; /* first inet address */
229 struct in_ifaddrhashhead *in_ifaddrhashtbls[MAXCPU];
230 /* inet addr hash table */
231 u_long in_ifaddrhmask; /* mask for hash table */
233 struct ip_stats ipstats_percpu[MAXCPU];
236 sysctl_ipstats(SYSCTL_HANDLER_ARGS)
240 for (cpu = 0; cpu < ncpus; ++cpu) {
241 if ((error = SYSCTL_OUT(req, &ipstats_percpu[cpu],
242 sizeof(struct ip_stats))))
244 if ((error = SYSCTL_IN(req, &ipstats_percpu[cpu],
245 sizeof(struct ip_stats))))
251 SYSCTL_PROC(_net_inet_ip, IPCTL_STATS, stats, (CTLTYPE_OPAQUE | CTLFLAG_RW),
252 0, 0, sysctl_ipstats, "S,ip_stats", "IP statistics");
254 SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW,
255 &ipstat, ip_stats, "IP statistics");
258 /* Packet reassembly stuff */
259 #define IPREASS_NHASH_LOG2 6
260 #define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2)
261 #define IPREASS_HMASK (IPREASS_NHASH - 1)
262 #define IPREASS_HASH(x,y) \
263 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK)
265 static struct ipq ipq[IPREASS_NHASH];
268 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
269 &ip_mtu, 0, "Default MTU");
273 static int ipstealth = 0;
274 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW, &ipstealth, 0, "");
276 static const int ipstealth = 0;
279 struct mbuf *(*ip_divert_p)(struct mbuf *, int, int);
281 struct pfil_head inet_pfil_hook;
284 * struct ip_srcrt_opt is used to store packet state while it travels
287 * XXX Note that the code even makes assumptions on the size and
288 * alignment of fields inside struct ip_srcrt so e.g. adding some
289 * fields will break the code. This needs to be fixed.
291 * We need to save the IP options in case a protocol wants to respond
292 * to an incoming packet over the same route if the packet got here
293 * using IP source routing. This allows connection establishment and
294 * maintenance when the remote end is on a network that is not known
298 struct in_addr dst; /* final destination */
299 char nop; /* one NOP to align */
300 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */
301 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)];
304 struct ip_srcrt_opt {
306 struct ip_srcrt ip_srcrt;
309 static MALLOC_DEFINE(M_IPQ, "ipq", "IP Fragment Management");
310 static struct malloc_pipe ipq_mpipe;
312 static void save_rte(struct mbuf *, u_char *, struct in_addr);
313 static int ip_dooptions(struct mbuf *m, int, struct sockaddr_in *);
314 static void ip_freef(struct ipq *);
315 static void ip_input_handler(struct netmsg *);
318 * IP initialization: fill in IP protocol switch table.
319 * All protocols not implemented in kernel go to raw IP protocol handler.
331 * Make sure we can handle a reasonable number of fragments but
332 * cap it at 4000 (XXX).
334 mpipe_init(&ipq_mpipe, M_IPQ, sizeof(struct ipq),
335 IFQ_MAXLEN, 4000, 0, NULL);
336 for (i = 0; i < ncpus; ++i) {
337 TAILQ_INIT(&in_ifaddrheads[i]);
338 in_ifaddrhashtbls[i] =
339 hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask);
341 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
344 for (i = 0; i < IPPROTO_MAX; i++)
345 ip_protox[i] = pr - inetsw;
346 for (pr = inetdomain.dom_protosw;
347 pr < inetdomain.dom_protoswNPROTOSW; pr++) {
348 if (pr->pr_domain->dom_family == PF_INET && pr->pr_protocol) {
349 if (pr->pr_protocol != IPPROTO_RAW)
350 ip_protox[pr->pr_protocol] = pr - inetsw;
354 inet_pfil_hook.ph_type = PFIL_TYPE_AF;
355 inet_pfil_hook.ph_af = AF_INET;
356 if ((i = pfil_head_register(&inet_pfil_hook)) != 0) {
357 kprintf("%s: WARNING: unable to register pfil hook, "
358 "error %d\n", __func__, i);
361 for (i = 0; i < IPREASS_NHASH; i++)
362 ipq[i].next = ipq[i].prev = &ipq[i];
364 maxnipq = nmbclusters / 32;
365 maxfragsperpacket = 16;
367 ip_id = time_second & 0xffff;
370 * Initialize IP statistics counters for each CPU.
374 for (cpu = 0; cpu < ncpus; ++cpu) {
375 bzero(&ipstats_percpu[cpu], sizeof(struct ip_stats));
378 bzero(&ipstat, sizeof(struct ip_stats));
381 netisr_register(NETISR_IP, ip_input_handler, ip_cpufn_in);
384 /* Do transport protocol processing. */
386 transport_processing_oncpu(struct mbuf *m, int hlen, struct ip *ip)
388 const struct protosw *pr = &inetsw[ip_protox[ip->ip_p]];
391 * Switch out to protocol's input routine.
394 pr->pr_input(m, hlen, ip->ip_p);
399 transport_processing_handler(netmsg_t netmsg)
401 struct netmsg_packet *pmsg = (struct netmsg_packet *)netmsg;
405 ip = mtod(pmsg->nm_packet, struct ip *);
406 hlen = pmsg->nm_netmsg.nm_lmsg.u.ms_result;
408 transport_processing_oncpu(pmsg->nm_packet, hlen, ip);
409 /* netmsg was embedded in the mbuf, do not reply! */
413 ip_input_handler(struct netmsg *msg0)
415 struct mbuf *m = ((struct netmsg_packet *)msg0)->nm_packet;
418 /* msg0 was embedded in the mbuf, do not reply! */
422 * IP input routine. Checksum and byte swap header. If fragmented
423 * try to reassemble. Process options. Pass to next level.
426 ip_input(struct mbuf *m)
429 struct in_ifaddr *ia = NULL;
430 struct in_ifaddr_container *iac;
433 struct in_addr pkt_dst;
434 boolean_t using_srcrt = FALSE; /* forward (by PFIL_HOOKS) */
435 struct in_addr odst; /* original dst address(NAT) */
437 struct sockaddr_in *next_hop = NULL;
440 struct tdb_ident *tdbi;
441 struct secpolicy *sp;
448 * This routine is called from numerous places which may not have
449 * characterized the packet.
451 if ((m->m_flags & M_HASH) == 0) {
452 ++ip_dispatch_software;
453 ip_cpufn(&m, 0, IP_MPORT_IN);
456 KKASSERT(m->m_flags & M_HASH);
458 ip = mtod(m, struct ip *);
461 * Pull out certain tags
463 if (m->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) {
465 mtag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL);
466 KKASSERT(mtag != NULL);
467 next_hop = m_tag_data(mtag);
470 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
471 /* dummynet already filtered us */
472 ip = mtod(m, struct ip *);
473 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
479 /* length checks already done in ip_cpufn() */
480 KASSERT(m->m_len >= sizeof(struct ip), ("IP header not in one mbuf"));
482 if (IP_VHL_V(ip->ip_vhl) != IPVERSION) {
483 ipstat.ips_badvers++;
487 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
488 /* length checks already done in ip_cpufn() */
489 KASSERT(hlen >= sizeof(struct ip), ("IP header len too small"));
490 KASSERT(m->m_len >= hlen, ("complete IP header not in one mbuf"));
492 /* 127/8 must not appear on wire - RFC1122 */
493 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
494 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
495 if (!(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK)) {
496 ipstat.ips_badaddr++;
501 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
502 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
504 if (hlen == sizeof(struct ip))
505 sum = in_cksum_hdr(ip);
507 sum = in_cksum(m, hlen);
515 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) {
516 /* packet is dropped by traffic conditioner */
521 * Convert fields to host representation.
523 ip->ip_len = ntohs(ip->ip_len);
524 ip->ip_off = ntohs(ip->ip_off);
526 /* length checks already done in ip_cpufn() */
527 KASSERT(ip->ip_len >= hlen, ("total length less then header length"));
528 KASSERT(m->m_pkthdr.len >= ip->ip_len, ("mbuf too short"));
531 * Trim mbufs if longer than the IP header would have us expect.
533 if (m->m_pkthdr.len > ip->ip_len) {
534 if (m->m_len == m->m_pkthdr.len) {
535 m->m_len = ip->ip_len;
536 m->m_pkthdr.len = ip->ip_len;
538 m_adj(m, ip->ip_len - m->m_pkthdr.len);
541 #if defined(IPSEC) && !defined(IPSEC_FILTERGIF)
543 * Bypass packet filtering for packets from a tunnel (gif).
545 if (ipsec_gethist(m, NULL))
551 * Right now when no processing on packet has done
552 * and it is still fresh out of network we do our black
554 * - Firewall: deny/allow/divert
555 * - Xlate: translate packet's addr/port (NAT).
556 * - Pipe: pass pkt through dummynet.
557 * - Wrap: fake packet's addr/port <unimpl.>
558 * - Encapsulate: put it in another IP and send out. <unimp.>
563 * If we've been forwarded from the output side, then
564 * skip the firewall a second time
566 if (next_hop != NULL)
570 if (!pfil_has_hooks(&inet_pfil_hook)) {
571 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
573 * Strip dummynet tags from stranded packets
575 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
576 KKASSERT(mtag != NULL);
577 m_tag_delete(m, mtag);
578 m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED;
584 * Run through list of hooks for input packets.
586 * NOTE! If the packet is rewritten pf/ipfw/whoever must
590 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif, PFIL_IN))
592 if (m == NULL) /* consumed by filter */
594 ip = mtod(m, struct ip *);
595 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
596 using_srcrt = (odst.s_addr != ip->ip_dst.s_addr);
598 if (m->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) {
599 mtag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL);
600 KKASSERT(mtag != NULL);
601 next_hop = m_tag_data(mtag);
603 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
607 if (m->m_pkthdr.fw_flags & FW_MBUF_REDISPATCH) {
608 m->m_pkthdr.fw_flags &= ~FW_MBUF_REDISPATCH;
612 * Process options and, if not destined for us,
613 * ship it on. ip_dooptions returns 1 when an
614 * error was detected (causing an icmp message
615 * to be sent and the original packet to be freed).
617 if (hlen > sizeof(struct ip) && ip_dooptions(m, 0, next_hop))
620 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
621 * matter if it is destined to another node, or whether it is
622 * a multicast one, RSVP wants it! and prevents it from being forwarded
623 * anywhere else. Also checks if the rsvp daemon is running before
624 * grabbing the packet.
626 if (rsvp_on && ip->ip_p == IPPROTO_RSVP)
630 * Check our list of addresses, to see if the packet is for us.
631 * If we don't have any addresses, assume any unicast packet
632 * we receive might be for us (and let the upper layers deal
635 if (TAILQ_EMPTY(&in_ifaddrheads[mycpuid]) &&
636 !(m->m_flags & (M_MCAST | M_BCAST)))
640 * Cache the destination address of the packet; this may be
641 * changed by use of 'ipfw fwd'.
643 pkt_dst = next_hop ? next_hop->sin_addr : ip->ip_dst;
646 * Enable a consistency check between the destination address
647 * and the arrival interface for a unicast packet (the RFC 1122
648 * strong ES model) if IP forwarding is disabled and the packet
649 * is not locally generated and the packet is not subject to
652 * XXX - Checking also should be disabled if the destination
653 * address is ipnat'ed to a different interface.
655 * XXX - Checking is incompatible with IP aliases added
656 * to the loopback interface instead of the interface where
657 * the packets are received.
659 checkif = ip_checkinterface &&
661 m->m_pkthdr.rcvif != NULL &&
662 !(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) &&
666 * Check for exact addresses in the hash bucket.
668 LIST_FOREACH(iac, INADDR_HASH(pkt_dst.s_addr), ia_hash) {
672 * If the address matches, verify that the packet
673 * arrived via the correct interface if checking is
676 if (IA_SIN(ia)->sin_addr.s_addr == pkt_dst.s_addr &&
677 (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif))
683 * Check for broadcast addresses.
685 * Only accept broadcast packets that arrive via the matching
686 * interface. Reception of forwarded directed broadcasts would
687 * be handled via ip_forward() and ether_output() with the loopback
688 * into the stack for SIMPLEX interfaces handled by ether_output().
690 if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
691 struct ifaddr_container *ifac;
693 TAILQ_FOREACH(ifac, &m->m_pkthdr.rcvif->if_addrheads[mycpuid],
695 struct ifaddr *ifa = ifac->ifa;
697 if (ifa->ifa_addr == NULL) /* shutdown/startup race */
699 if (ifa->ifa_addr->sa_family != AF_INET)
702 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
705 if (ia->ia_netbroadcast.s_addr == pkt_dst.s_addr)
708 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY)
713 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
714 struct in_multi *inm;
716 /* XXX Multicast is not MPSAFE yet */
719 if (ip_mrouter != NULL) {
721 * If we are acting as a multicast router, all
722 * incoming multicast packets are passed to the
723 * kernel-level multicast forwarding function.
724 * The packet is returned (relatively) intact; if
725 * ip_mforward() returns a non-zero value, the packet
726 * must be discarded, else it may be accepted below.
728 if (ip_mforward != NULL &&
729 ip_mforward(ip, m->m_pkthdr.rcvif, m, NULL) != 0) {
731 ipstat.ips_cantforward++;
737 * The process-level routing daemon needs to receive
738 * all multicast IGMP packets, whether or not this
739 * host belongs to their destination groups.
741 if (ip->ip_p == IPPROTO_IGMP) {
745 ipstat.ips_forward++;
748 * See if we belong to the destination multicast group on the
751 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
754 ipstat.ips_notmember++;
762 if (ip->ip_dst.s_addr == INADDR_BROADCAST)
764 if (ip->ip_dst.s_addr == INADDR_ANY)
768 * FAITH(Firewall Aided Internet Translator)
770 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
772 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP)
780 * Not for us; forward if possible and desirable.
783 ipstat.ips_cantforward++;
788 * Enforce inbound IPsec SPD.
790 if (ipsec4_in_reject(m, NULL)) {
791 ipsecstat.in_polvio++;
796 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
799 tdbi = (struct tdb_ident *)m_tag_data(mtag);
800 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
802 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
803 IP_FORWARDING, &error);
805 if (sp == NULL) { /* NB: can happen if error */
807 /*XXX error stat???*/
808 DPRINTF(("ip_input: no SP for forwarding\n")); /*XXX*/
813 * Check security policy against packet attributes.
815 error = ipsec_in_reject(sp, m);
819 ipstat.ips_cantforward++;
823 ip_forward(m, using_srcrt, next_hop);
830 * IPSTEALTH: Process non-routing options only
831 * if the packet is destined for us.
834 hlen > sizeof(struct ip) &&
835 ip_dooptions(m, 1, next_hop))
838 /* Count the packet in the ip address stats */
840 ia->ia_ifa.if_ipackets++;
841 ia->ia_ifa.if_ibytes += m->m_pkthdr.len;
845 * If offset or IP_MF are set, must reassemble.
846 * Otherwise, nothing need be done.
847 * (We could look in the reassembly queue to see
848 * if the packet was previously fragmented,
849 * but it's not worth the time; just let them time out.)
851 if (ip->ip_off & (IP_MF | IP_OFFMASK)) {
853 * Attempt reassembly; if it succeeds, proceed. ip_reass()
854 * will return a different mbuf.
856 * NOTE: ip_reass() returns m with M_HASH cleared to force
857 * us to recharacterize the packet.
862 ip = mtod(m, struct ip *);
864 /* Get the header length of the reassembled packet */
865 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
872 * enforce IPsec policy checking if we are seeing last header.
873 * note that we do not visit this with protocols with pcb layer
874 * code - like udp/tcp/raw ip.
876 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) &&
877 ipsec4_in_reject(m, NULL)) {
878 ipsecstat.in_polvio++;
884 * enforce IPsec policy checking if we are seeing last header.
885 * note that we do not visit this with protocols with pcb layer
886 * code - like udp/tcp/raw ip.
888 if (inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) {
890 * Check if the packet has already had IPsec processing
891 * done. If so, then just pass it along. This tag gets
892 * set during AH, ESP, etc. input handling, before the
893 * packet is returned to the ip input queue for delivery.
895 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
898 tdbi = (struct tdb_ident *)m_tag_data(mtag);
899 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
901 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
902 IP_FORWARDING, &error);
906 * Check security policy against packet attributes.
908 error = ipsec_in_reject(sp, m);
911 /* XXX error stat??? */
913 DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/
920 #endif /* FAST_IPSEC */
923 * We must forward the packet to the correct protocol thread if
924 * we are not already in it.
926 * NOTE: ip_len is now in host form. ip_len is not adjusted
927 * further for protocol processing, instead we pass hlen
928 * to the protosw and let it deal with it.
930 ipstat.ips_delivered++;
932 if ((m->m_flags & M_HASH) == 0) {
933 ++ip_dispatch_recheck;
934 ip->ip_len = htons(ip->ip_len);
935 ip->ip_off = htons(ip->ip_off);
937 ip_cpufn(&m, 0, IP_MPORT_IN);
941 ip = mtod(m, struct ip *);
942 ip->ip_len = ntohs(ip->ip_len);
943 ip->ip_off = ntohs(ip->ip_off);
944 KKASSERT(m->m_flags & M_HASH);
946 port = cpu_portfn(m->m_pkthdr.hash);
948 if (port != &curthread->td_msgport) {
949 struct netmsg_packet *pmsg;
953 pmsg = &m->m_hdr.mh_netmsg;
954 netmsg_init(&pmsg->nm_netmsg, NULL, &netisr_apanic_rport,
955 0, transport_processing_handler);
957 pmsg->nm_netmsg.nm_lmsg.u.ms_result = hlen;
959 lwkt_sendmsg(port, &pmsg->nm_netmsg.nm_lmsg);
962 transport_processing_oncpu(m, hlen, ip);
971 * Take incoming datagram fragment and try to reassemble it into
972 * whole datagram. If a chain for reassembly of this datagram already
973 * exists, then it is given as fp; otherwise have to make a chain.
976 ip_reass(struct mbuf *m)
978 struct ip *ip = mtod(m, struct ip *);
979 struct mbuf *p = NULL, *q, *nq;
981 struct ipq *fp = NULL;
982 int hlen = IP_VHL_HL(ip->ip_vhl) << 2;
986 /* If maxnipq is 0, never accept fragments. */
988 ipstat.ips_fragments++;
989 ipstat.ips_fragdropped++;
994 sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
996 * Look for queue of fragments of this datagram.
998 lwkt_gettoken(&ipq_token);
999 for (fp = ipq[sum].next; fp != &ipq[sum]; fp = fp->next) {
1000 if (ip->ip_id == fp->ipq_id &&
1001 ip->ip_src.s_addr == fp->ipq_src.s_addr &&
1002 ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
1003 ip->ip_p == fp->ipq_p)
1010 * Enforce upper bound on number of fragmented packets
1011 * for which we attempt reassembly;
1012 * If maxnipq is -1, accept all fragments without limitation.
1014 if (nipq > maxnipq && maxnipq > 0) {
1016 * drop something from the tail of the current queue
1017 * before proceeding further
1019 if (ipq[sum].prev == &ipq[sum]) { /* gak */
1020 for (i = 0; i < IPREASS_NHASH; i++) {
1021 if (ipq[i].prev != &ipq[i]) {
1022 ipstat.ips_fragtimeout +=
1023 ipq[i].prev->ipq_nfrags;
1024 ip_freef(ipq[i].prev);
1029 ipstat.ips_fragtimeout +=
1030 ipq[sum].prev->ipq_nfrags;
1031 ip_freef(ipq[sum].prev);
1036 * Adjust ip_len to not reflect header,
1037 * convert offset of this to bytes.
1040 if (ip->ip_off & IP_MF) {
1042 * Make sure that fragments have a data length
1043 * that's a non-zero multiple of 8 bytes.
1045 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
1046 ipstat.ips_toosmall++; /* XXX */
1050 m->m_flags |= M_FRAG;
1052 m->m_flags &= ~M_FRAG;
1056 ipstat.ips_fragments++;
1057 m->m_pkthdr.header = ip;
1060 * If the hardware has not done csum over this fragment
1061 * then csum_data is not valid at all.
1063 if ((m->m_pkthdr.csum_flags & (CSUM_FRAG_NOT_CHECKED | CSUM_DATA_VALID))
1064 == (CSUM_FRAG_NOT_CHECKED | CSUM_DATA_VALID)) {
1065 m->m_pkthdr.csum_data = 0;
1066 m->m_pkthdr.csum_flags &= ~(CSUM_DATA_VALID | CSUM_PSEUDO_HDR);
1070 * Presence of header sizes in mbufs
1071 * would confuse code below.
1077 * If first fragment to arrive, create a reassembly queue.
1080 if ((fp = mpipe_alloc_nowait(&ipq_mpipe)) == NULL)
1082 insque(fp, &ipq[sum]);
1085 fp->ipq_ttl = IPFRAGTTL;
1086 fp->ipq_p = ip->ip_p;
1087 fp->ipq_id = ip->ip_id;
1088 fp->ipq_src = ip->ip_src;
1089 fp->ipq_dst = ip->ip_dst;
1091 m->m_nextpkt = NULL;
1097 #define GETIP(m) ((struct ip*)((m)->m_pkthdr.header))
1100 * Find a segment which begins after this one does.
1102 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
1103 if (GETIP(q)->ip_off > ip->ip_off)
1108 * If there is a preceding segment, it may provide some of
1109 * our data already. If so, drop the data from the incoming
1110 * segment. If it provides all of our data, drop us, otherwise
1111 * stick new segment in the proper place.
1113 * If some of the data is dropped from the the preceding
1114 * segment, then it's checksum is invalidated.
1117 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off;
1119 if (i >= ip->ip_len)
1122 m->m_pkthdr.csum_flags = 0;
1126 m->m_nextpkt = p->m_nextpkt;
1129 m->m_nextpkt = fp->ipq_frags;
1134 * While we overlap succeeding segments trim them or,
1135 * if they are completely covered, dequeue them.
1137 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off;
1139 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off;
1140 if (i < GETIP(q)->ip_len) {
1141 GETIP(q)->ip_len -= i;
1142 GETIP(q)->ip_off += i;
1144 q->m_pkthdr.csum_flags = 0;
1149 ipstat.ips_fragdropped++;
1151 q->m_nextpkt = NULL;
1157 * Check for complete reassembly and perform frag per packet
1160 * Frag limiting is performed here so that the nth frag has
1161 * a chance to complete the packet before we drop the packet.
1162 * As a result, n+1 frags are actually allowed per packet, but
1163 * only n will ever be stored. (n = maxfragsperpacket.)
1167 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
1168 if (GETIP(q)->ip_off != next) {
1169 if (fp->ipq_nfrags > maxfragsperpacket) {
1170 ipstat.ips_fragdropped += fp->ipq_nfrags;
1175 next += GETIP(q)->ip_len;
1177 /* Make sure the last packet didn't have the IP_MF flag */
1178 if (p->m_flags & M_FRAG) {
1179 if (fp->ipq_nfrags > maxfragsperpacket) {
1180 ipstat.ips_fragdropped += fp->ipq_nfrags;
1187 * Reassembly is complete. Make sure the packet is a sane size.
1191 if (next + (IP_VHL_HL(ip->ip_vhl) << 2) > IP_MAXPACKET) {
1192 ipstat.ips_toolong++;
1193 ipstat.ips_fragdropped += fp->ipq_nfrags;
1199 * Concatenate fragments.
1206 q->m_nextpkt = NULL;
1207 for (q = nq; q != NULL; q = nq) {
1209 q->m_nextpkt = NULL;
1210 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
1211 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
1216 * Clean up the 1's complement checksum. Carry over 16 bits must
1217 * be added back. This assumes no more then 65535 packet fragments
1218 * were reassembled. A second carry can also occur (but not a third).
1220 m->m_pkthdr.csum_data = (m->m_pkthdr.csum_data & 0xffff) +
1221 (m->m_pkthdr.csum_data >> 16);
1222 if (m->m_pkthdr.csum_data > 0xFFFF)
1223 m->m_pkthdr.csum_data -= 0xFFFF;
1226 * Create header for new ip packet by
1227 * modifying header of first packet;
1228 * dequeue and discard fragment reassembly header.
1229 * Make header visible.
1232 ip->ip_src = fp->ipq_src;
1233 ip->ip_dst = fp->ipq_dst;
1236 mpipe_free(&ipq_mpipe, fp);
1237 m->m_len += (IP_VHL_HL(ip->ip_vhl) << 2);
1238 m->m_data -= (IP_VHL_HL(ip->ip_vhl) << 2);
1239 /* some debugging cruft by sklower, below, will go away soon */
1240 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */
1243 for (n = m; n; n = n->m_next)
1245 m->m_pkthdr.len = plen;
1249 * Reassembly complete, return the next protocol.
1251 * Be sure to clear M_HASH to force the packet
1252 * to be re-characterized.
1254 * Clear M_FRAG, we are no longer a fragment.
1256 m->m_flags &= ~(M_HASH | M_FRAG);
1258 ipstat.ips_reassembled++;
1259 lwkt_reltoken(&ipq_token);
1263 ipstat.ips_fragdropped++;
1268 lwkt_reltoken(&ipq_token);
1275 * Free a fragment reassembly header and all
1276 * associated datagrams.
1278 * Called with ipq_token held.
1281 ip_freef(struct ipq *fp)
1286 * Remove first to protect against blocking
1291 * Clean out at our leisure
1293 while (fp->ipq_frags) {
1295 fp->ipq_frags = q->m_nextpkt;
1296 q->m_nextpkt = NULL;
1299 mpipe_free(&ipq_mpipe, fp);
1304 * IP timer processing;
1305 * if a timer expires on a reassembly
1306 * queue, discard it.
1314 lwkt_gettoken(&ipq_token);
1315 for (i = 0; i < IPREASS_NHASH; i++) {
1319 while (fp != &ipq[i]) {
1322 if (fp->prev->ipq_ttl == 0) {
1323 ipstat.ips_fragtimeout += fp->prev->ipq_nfrags;
1329 * If we are over the maximum number of fragments
1330 * (due to the limit being lowered), drain off
1331 * enough to get down to the new limit.
1333 if (maxnipq >= 0 && nipq > maxnipq) {
1334 for (i = 0; i < IPREASS_NHASH; i++) {
1335 while (nipq > maxnipq &&
1336 (ipq[i].next != &ipq[i])) {
1337 ipstat.ips_fragdropped +=
1338 ipq[i].next->ipq_nfrags;
1339 ip_freef(ipq[i].next);
1343 lwkt_reltoken(&ipq_token);
1348 * Drain off all datagram fragments.
1355 lwkt_gettoken(&ipq_token);
1356 for (i = 0; i < IPREASS_NHASH; i++) {
1357 while (ipq[i].next != &ipq[i]) {
1358 ipstat.ips_fragdropped += ipq[i].next->ipq_nfrags;
1359 ip_freef(ipq[i].next);
1362 lwkt_reltoken(&ipq_token);
1367 * Do option processing on a datagram,
1368 * possibly discarding it if bad options are encountered,
1369 * or forwarding it if source-routed.
1370 * The pass argument is used when operating in the IPSTEALTH
1371 * mode to tell what options to process:
1372 * [LS]SRR (pass 0) or the others (pass 1).
1373 * The reason for as many as two passes is that when doing IPSTEALTH,
1374 * non-routing options should be processed only if the packet is for us.
1375 * Returns 1 if packet has been forwarded/freed,
1376 * 0 if the packet should be processed further.
1379 ip_dooptions(struct mbuf *m, int pass, struct sockaddr_in *next_hop)
1381 struct sockaddr_in ipaddr = { sizeof ipaddr, AF_INET };
1382 struct ip *ip = mtod(m, struct ip *);
1384 struct in_ifaddr *ia;
1385 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB;
1386 boolean_t forward = FALSE;
1387 struct in_addr *sin, dst;
1391 cp = (u_char *)(ip + 1);
1392 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
1393 for (; cnt > 0; cnt -= optlen, cp += optlen) {
1394 opt = cp[IPOPT_OPTVAL];
1395 if (opt == IPOPT_EOL)
1397 if (opt == IPOPT_NOP)
1400 if (cnt < IPOPT_OLEN + sizeof(*cp)) {
1401 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1404 optlen = cp[IPOPT_OLEN];
1405 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
1406 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1416 * Source routing with record.
1417 * Find interface with current destination address.
1418 * If none on this machine then drop if strictly routed,
1419 * or do nothing if loosely routed.
1420 * Record interface address and bring up next address
1421 * component. If strictly routed make sure next
1422 * address is on directly accessible net.
1426 if (ipstealth && pass > 0)
1428 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1429 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1432 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1433 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1436 ipaddr.sin_addr = ip->ip_dst;
1437 ia = (struct in_ifaddr *)
1438 ifa_ifwithaddr((struct sockaddr *)&ipaddr);
1440 if (opt == IPOPT_SSRR) {
1441 type = ICMP_UNREACH;
1442 code = ICMP_UNREACH_SRCFAIL;
1445 if (!ip_dosourceroute)
1446 goto nosourcerouting;
1448 * Loose routing, and not at next destination
1449 * yet; nothing to do except forward.
1453 off--; /* 0 origin */
1454 if (off > optlen - (int)sizeof(struct in_addr)) {
1456 * End of source route. Should be for us.
1458 if (!ip_acceptsourceroute)
1459 goto nosourcerouting;
1460 save_rte(m, cp, ip->ip_src);
1465 if (!ip_dosourceroute) {
1467 char buf[sizeof "aaa.bbb.ccc.ddd"];
1470 * Acting as a router, so generate ICMP
1473 strcpy(buf, inet_ntoa(ip->ip_dst));
1475 "attempted source route from %s to %s\n",
1476 inet_ntoa(ip->ip_src), buf);
1477 type = ICMP_UNREACH;
1478 code = ICMP_UNREACH_SRCFAIL;
1482 * Not acting as a router,
1486 ipstat.ips_cantforward++;
1493 * locate outgoing interface
1495 memcpy(&ipaddr.sin_addr, cp + off,
1496 sizeof ipaddr.sin_addr);
1498 if (opt == IPOPT_SSRR) {
1499 #define INA struct in_ifaddr *
1500 #define SA struct sockaddr *
1501 if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr))
1503 ia = (INA)ifa_ifwithnet((SA)&ipaddr);
1505 ia = ip_rtaddr(ipaddr.sin_addr, NULL);
1508 type = ICMP_UNREACH;
1509 code = ICMP_UNREACH_SRCFAIL;
1512 ip->ip_dst = ipaddr.sin_addr;
1513 memcpy(cp + off, &IA_SIN(ia)->sin_addr,
1514 sizeof(struct in_addr));
1515 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1517 * Let ip_intr's mcast routing check handle mcast pkts
1519 forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr));
1523 if (ipstealth && pass == 0)
1525 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1526 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1529 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1530 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1534 * If no space remains, ignore.
1536 off--; /* 0 origin */
1537 if (off > optlen - (int)sizeof(struct in_addr))
1539 memcpy(&ipaddr.sin_addr, &ip->ip_dst,
1540 sizeof ipaddr.sin_addr);
1542 * locate outgoing interface; if we're the destination,
1543 * use the incoming interface (should be same).
1545 if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == NULL &&
1546 (ia = ip_rtaddr(ipaddr.sin_addr, NULL)) == NULL) {
1547 type = ICMP_UNREACH;
1548 code = ICMP_UNREACH_HOST;
1551 memcpy(cp + off, &IA_SIN(ia)->sin_addr,
1552 sizeof(struct in_addr));
1553 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1557 if (ipstealth && pass == 0)
1559 code = cp - (u_char *)ip;
1560 if (optlen < 4 || optlen > 40) {
1561 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1564 if ((off = cp[IPOPT_OFFSET]) < 5) {
1565 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1568 if (off > optlen - (int)sizeof(int32_t)) {
1569 cp[IPOPT_OFFSET + 1] += (1 << 4);
1570 if ((cp[IPOPT_OFFSET + 1] & 0xf0) == 0) {
1571 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1576 off--; /* 0 origin */
1577 sin = (struct in_addr *)(cp + off);
1578 switch (cp[IPOPT_OFFSET + 1] & 0x0f) {
1580 case IPOPT_TS_TSONLY:
1583 case IPOPT_TS_TSANDADDR:
1584 if (off + sizeof(n_time) +
1585 sizeof(struct in_addr) > optlen) {
1586 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1589 ipaddr.sin_addr = dst;
1590 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr,
1594 memcpy(sin, &IA_SIN(ia)->sin_addr,
1595 sizeof(struct in_addr));
1596 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1597 off += sizeof(struct in_addr);
1600 case IPOPT_TS_PRESPEC:
1601 if (off + sizeof(n_time) +
1602 sizeof(struct in_addr) > optlen) {
1603 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1606 memcpy(&ipaddr.sin_addr, sin,
1607 sizeof(struct in_addr));
1608 if (ifa_ifwithaddr((SA)&ipaddr) == NULL)
1610 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1611 off += sizeof(struct in_addr);
1615 code = &cp[IPOPT_OFFSET + 1] - (u_char *)ip;
1619 memcpy(cp + off, &ntime, sizeof(n_time));
1620 cp[IPOPT_OFFSET] += sizeof(n_time);
1623 if (forward && ipforwarding) {
1624 ip_forward(m, TRUE, next_hop);
1629 icmp_error(m, type, code, 0, 0);
1630 ipstat.ips_badoptions++;
1635 * Given address of next destination (final or next hop),
1636 * return internet address info of interface to be used to get there.
1639 ip_rtaddr(struct in_addr dst, struct route *ro0)
1641 struct route sro, *ro;
1642 struct sockaddr_in *sin;
1643 struct in_ifaddr *ia;
1648 bzero(&sro, sizeof(sro));
1652 sin = (struct sockaddr_in *)&ro->ro_dst;
1654 if (ro->ro_rt == NULL || dst.s_addr != sin->sin_addr.s_addr) {
1655 if (ro->ro_rt != NULL) {
1659 sin->sin_family = AF_INET;
1660 sin->sin_len = sizeof *sin;
1661 sin->sin_addr = dst;
1662 rtalloc_ign(ro, RTF_PRCLONING);
1665 if (ro->ro_rt == NULL)
1668 ia = ifatoia(ro->ro_rt->rt_ifa);
1676 * Save incoming source route for use in replies,
1677 * to be picked up later by ip_srcroute if the receiver is interested.
1680 save_rte(struct mbuf *m, u_char *option, struct in_addr dst)
1683 struct ip_srcrt_opt *opt;
1686 mtag = m_tag_get(PACKET_TAG_IPSRCRT, sizeof(*opt), MB_DONTWAIT);
1689 opt = m_tag_data(mtag);
1691 olen = option[IPOPT_OLEN];
1694 kprintf("save_rte: olen %d\n", olen);
1696 if (olen > sizeof(opt->ip_srcrt) - (1 + sizeof(dst))) {
1700 bcopy(option, opt->ip_srcrt.srcopt, olen);
1701 opt->ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1702 opt->ip_srcrt.dst = dst;
1703 m_tag_prepend(m, mtag);
1707 * Retrieve incoming source route for use in replies,
1708 * in the same form used by setsockopt.
1709 * The first hop is placed before the options, will be removed later.
1712 ip_srcroute(struct mbuf *m0)
1714 struct in_addr *p, *q;
1717 struct ip_srcrt_opt *opt;
1722 mtag = m_tag_find(m0, PACKET_TAG_IPSRCRT, NULL);
1725 opt = m_tag_data(mtag);
1727 if (opt->ip_nhops == 0)
1729 m = m_get(MB_DONTWAIT, MT_HEADER);
1733 #define OPTSIZ (sizeof(opt->ip_srcrt.nop) + sizeof(opt->ip_srcrt.srcopt))
1735 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
1736 m->m_len = opt->ip_nhops * sizeof(struct in_addr) +
1737 sizeof(struct in_addr) + OPTSIZ;
1740 kprintf("ip_srcroute: nhops %d mlen %d",
1741 opt->ip_nhops, m->m_len);
1746 * First save first hop for return route
1748 p = &opt->ip_srcrt.route[opt->ip_nhops - 1];
1749 *(mtod(m, struct in_addr *)) = *p--;
1752 kprintf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr));
1756 * Copy option fields and padding (nop) to mbuf.
1758 opt->ip_srcrt.nop = IPOPT_NOP;
1759 opt->ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
1760 memcpy(mtod(m, caddr_t) + sizeof(struct in_addr), &opt->ip_srcrt.nop,
1762 q = (struct in_addr *)(mtod(m, caddr_t) +
1763 sizeof(struct in_addr) + OPTSIZ);
1766 * Record return path as an IP source route,
1767 * reversing the path (pointers are now aligned).
1769 while (p >= opt->ip_srcrt.route) {
1772 kprintf(" %x", ntohl(q->s_addr));
1777 * Last hop goes to final destination.
1779 *q = opt->ip_srcrt.dst;
1780 m_tag_delete(m0, mtag);
1783 kprintf(" %x\n", ntohl(q->s_addr));
1789 * Strip out IP options.
1792 ip_stripoptions(struct mbuf *m)
1795 struct ip *ip = mtod(m, struct ip *);
1799 optlen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
1800 opts = (caddr_t)(ip + 1);
1801 datalen = m->m_len - (sizeof(struct ip) + optlen);
1802 bcopy(opts + optlen, opts, datalen);
1804 if (m->m_flags & M_PKTHDR)
1805 m->m_pkthdr.len -= optlen;
1806 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, sizeof(struct ip) >> 2);
1809 u_char inetctlerrmap[PRC_NCMDS] = {
1811 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
1812 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
1813 EMSGSIZE, EHOSTUNREACH, 0, 0,
1815 ENOPROTOOPT, ECONNREFUSED
1819 * Forward a packet. If some error occurs return the sender
1820 * an icmp packet. Note we can't always generate a meaningful
1821 * icmp message because icmp doesn't have a large enough repertoire
1822 * of codes and types.
1824 * If not forwarding, just drop the packet. This could be confusing
1825 * if ipforwarding was zero but some routing protocol was advancing
1826 * us as a gateway to somewhere. However, we must let the routing
1827 * protocol deal with that.
1829 * The using_srcrt parameter indicates whether the packet is being forwarded
1830 * via a source route.
1833 ip_forward(struct mbuf *m, boolean_t using_srcrt, struct sockaddr_in *next_hop)
1835 struct ip *ip = mtod(m, struct ip *);
1837 struct route fwd_ro;
1838 int error, type = 0, code = 0, destmtu = 0;
1841 struct in_addr pkt_dst;
1845 * Cache the destination address of the packet; this may be
1846 * changed by use of 'ipfw fwd'.
1848 pkt_dst = (next_hop != NULL) ? next_hop->sin_addr : ip->ip_dst;
1852 kprintf("forward: src %x dst %x ttl %x\n",
1853 ip->ip_src.s_addr, pkt_dst.s_addr, ip->ip_ttl);
1856 if (m->m_flags & (M_BCAST | M_MCAST) || !in_canforward(pkt_dst)) {
1857 ipstat.ips_cantforward++;
1861 if (!ipstealth && ip->ip_ttl <= IPTTLDEC) {
1862 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0);
1866 bzero(&fwd_ro, sizeof(fwd_ro));
1867 ip_rtaddr(pkt_dst, &fwd_ro);
1868 if (fwd_ro.ro_rt == NULL) {
1869 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0);
1875 * Save the IP header and at most 8 bytes of the payload,
1876 * in case we need to generate an ICMP message to the src.
1878 * XXX this can be optimized a lot by saving the data in a local
1879 * buffer on the stack (72 bytes at most), and only allocating the
1880 * mbuf if really necessary. The vast majority of the packets
1881 * are forwarded without having to send an ICMP back (either
1882 * because unnecessary, or because rate limited), so we are
1883 * really we are wasting a lot of work here.
1885 * We don't use m_copy() because it might return a reference
1886 * to a shared cluster. Both this function and ip_output()
1887 * assume exclusive access to the IP header in `m', so any
1888 * data in a cluster may change before we reach icmp_error().
1890 MGETHDR(mcopy, MB_DONTWAIT, m->m_type);
1891 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, MB_DONTWAIT)) {
1893 * It's probably ok if the pkthdr dup fails (because
1894 * the deep copy of the tag chain failed), but for now
1895 * be conservative and just discard the copy since
1896 * code below may some day want the tags.
1901 if (mcopy != NULL) {
1902 mcopy->m_len = imin((IP_VHL_HL(ip->ip_vhl) << 2) + 8,
1904 mcopy->m_pkthdr.len = mcopy->m_len;
1905 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1909 ip->ip_ttl -= IPTTLDEC;
1912 * If forwarding packet using same interface that it came in on,
1913 * perhaps should send a redirect to sender to shortcut a hop.
1914 * Only send redirect if source is sending directly to us,
1915 * and if packet was not source routed (or has any options).
1916 * Also, don't send redirect if forwarding using a default route
1917 * or a route modified by a redirect.
1919 if (rt->rt_ifp == m->m_pkthdr.rcvif &&
1920 !(rt->rt_flags & (RTF_DYNAMIC | RTF_MODIFIED)) &&
1921 satosin(rt_key(rt))->sin_addr.s_addr != INADDR_ANY &&
1922 ipsendredirects && !using_srcrt && next_hop == NULL) {
1923 u_long src = ntohl(ip->ip_src.s_addr);
1924 struct in_ifaddr *rt_ifa = (struct in_ifaddr *)rt->rt_ifa;
1926 if (rt_ifa != NULL &&
1927 (src & rt_ifa->ia_subnetmask) == rt_ifa->ia_subnet) {
1928 if (rt->rt_flags & RTF_GATEWAY)
1929 dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
1931 dest = pkt_dst.s_addr;
1933 * Router requirements says to only send
1936 type = ICMP_REDIRECT;
1937 code = ICMP_REDIRECT_HOST;
1940 kprintf("redirect (%d) to %x\n", code, dest);
1945 error = ip_output(m, NULL, &fwd_ro, IP_FORWARDING, NULL, NULL);
1947 ipstat.ips_forward++;
1950 ipflow_create(&fwd_ro, mcopy);
1955 ipstat.ips_redirectsent++;
1958 ipstat.ips_cantforward++;
1965 * Send ICMP message.
1970 case 0: /* forwarded, but need redirect */
1971 /* type, code set above */
1974 case ENETUNREACH: /* shouldn't happen, checked above */
1979 type = ICMP_UNREACH;
1980 code = ICMP_UNREACH_HOST;
1984 type = ICMP_UNREACH;
1985 code = ICMP_UNREACH_NEEDFRAG;
1988 * If the packet is routed over IPsec tunnel, tell the
1989 * originator the tunnel MTU.
1990 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
1993 if (fwd_ro.ro_rt != NULL) {
1994 struct secpolicy *sp = NULL;
1999 sp = ipsec4_getpolicybyaddr(mcopy,
2005 destmtu = fwd_ro.ro_rt->rt_ifp->if_mtu;
2007 /* count IPsec header size */
2008 ipsechdr = ipsec4_hdrsiz(mcopy,
2013 * find the correct route for outer IPv4
2014 * header, compute tunnel MTU.
2017 if (sp->req != NULL && sp->req->sav != NULL &&
2018 sp->req->sav->sah != NULL) {
2019 ro = &sp->req->sav->sah->sa_route;
2020 if (ro->ro_rt != NULL &&
2021 ro->ro_rt->rt_ifp != NULL) {
2023 ro->ro_rt->rt_ifp->if_mtu;
2024 destmtu -= ipsechdr;
2033 * If the packet is routed over IPsec tunnel, tell the
2034 * originator the tunnel MTU.
2035 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
2038 if (fwd_ro.ro_rt != NULL) {
2039 struct secpolicy *sp = NULL;
2044 sp = ipsec_getpolicybyaddr(mcopy,
2050 destmtu = fwd_ro.ro_rt->rt_ifp->if_mtu;
2052 /* count IPsec header size */
2053 ipsechdr = ipsec4_hdrsiz(mcopy,
2058 * find the correct route for outer IPv4
2059 * header, compute tunnel MTU.
2062 if (sp->req != NULL &&
2063 sp->req->sav != NULL &&
2064 sp->req->sav->sah != NULL) {
2065 ro = &sp->req->sav->sah->sa_route;
2066 if (ro->ro_rt != NULL &&
2067 ro->ro_rt->rt_ifp != NULL) {
2069 ro->ro_rt->rt_ifp->if_mtu;
2070 destmtu -= ipsechdr;
2077 #else /* !IPSEC && !FAST_IPSEC */
2078 if (fwd_ro.ro_rt != NULL)
2079 destmtu = fwd_ro.ro_rt->rt_ifp->if_mtu;
2081 ipstat.ips_cantfrag++;
2086 * A router should not generate ICMP_SOURCEQUENCH as
2087 * required in RFC1812 Requirements for IP Version 4 Routers.
2088 * Source quench could be a big problem under DoS attacks,
2089 * or if the underlying interface is rate-limited.
2090 * Those who need source quench packets may re-enable them
2091 * via the net.inet.ip.sendsourcequench sysctl.
2093 if (!ip_sendsourcequench) {
2097 type = ICMP_SOURCEQUENCH;
2102 case EACCES: /* ipfw denied packet */
2106 icmp_error(mcopy, type, code, dest, destmtu);
2108 if (fwd_ro.ro_rt != NULL)
2109 RTFREE(fwd_ro.ro_rt);
2113 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
2116 if (inp->inp_socket->so_options & SO_TIMESTAMP) {
2120 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
2121 SCM_TIMESTAMP, SOL_SOCKET);
2123 mp = &(*mp)->m_next;
2125 if (inp->inp_flags & INP_RECVDSTADDR) {
2126 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
2127 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
2129 mp = &(*mp)->m_next;
2131 if (inp->inp_flags & INP_RECVTTL) {
2132 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl,
2133 sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
2135 mp = &(*mp)->m_next;
2139 * Moving these out of udp_input() made them even more broken
2140 * than they already were.
2142 /* options were tossed already */
2143 if (inp->inp_flags & INP_RECVOPTS) {
2144 *mp = sbcreatecontrol((caddr_t) opts_deleted_above,
2145 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
2147 mp = &(*mp)->m_next;
2149 /* ip_srcroute doesn't do what we want here, need to fix */
2150 if (inp->inp_flags & INP_RECVRETOPTS) {
2151 *mp = sbcreatecontrol((caddr_t) ip_srcroute(m),
2152 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
2154 mp = &(*mp)->m_next;
2157 if (inp->inp_flags & INP_RECVIF) {
2160 struct sockaddr_dl sdl;
2163 struct sockaddr_dl *sdp;
2164 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
2166 if (((ifp = m->m_pkthdr.rcvif)) &&
2167 ((ifp->if_index != 0) && (ifp->if_index <= if_index))) {
2168 sdp = IF_LLSOCKADDR(ifp);
2170 * Change our mind and don't try copy.
2172 if ((sdp->sdl_family != AF_LINK) ||
2173 (sdp->sdl_len > sizeof(sdlbuf))) {
2176 bcopy(sdp, sdl2, sdp->sdl_len);
2180 offsetof(struct sockaddr_dl, sdl_data[0]);
2181 sdl2->sdl_family = AF_LINK;
2182 sdl2->sdl_index = 0;
2183 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
2185 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len,
2186 IP_RECVIF, IPPROTO_IP);
2188 mp = &(*mp)->m_next;
2193 * XXX these routines are called from the upper part of the kernel.
2195 * They could also be moved to ip_mroute.c, since all the RSVP
2196 * handling is done there already.
2199 ip_rsvp_init(struct socket *so)
2201 if (so->so_type != SOCK_RAW ||
2202 so->so_proto->pr_protocol != IPPROTO_RSVP)
2205 if (ip_rsvpd != NULL)
2210 * This may seem silly, but we need to be sure we don't over-increment
2211 * the RSVP counter, in case something slips up.
2226 * This may seem silly, but we need to be sure we don't over-decrement
2227 * the RSVP counter, in case something slips up.
2237 rsvp_input(struct mbuf *m, ...) /* XXX must fixup manually */
2243 off = __va_arg(ap, int);
2244 proto = __va_arg(ap, int);
2247 if (rsvp_input_p) { /* call the real one if loaded */
2248 rsvp_input_p(m, off, proto);
2252 /* Can still get packets with rsvp_on = 0 if there is a local member
2253 * of the group to which the RSVP packet is addressed. But in this
2254 * case we want to throw the packet away.
2262 if (ip_rsvpd != NULL) {
2263 rip_input(m, off, proto);
2266 /* Drop the packet */