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 1.110 2008/09/24 14:26:39 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 <machine/stdarg.h>
101 #include <net/if_types.h>
102 #include <net/if_var.h>
103 #include <net/if_dl.h>
104 #include <net/pfil.h>
105 #include <net/route.h>
106 #include <net/netisr.h>
108 #include <netinet/in.h>
109 #include <netinet/in_systm.h>
110 #include <netinet/in_var.h>
111 #include <netinet/ip.h>
112 #include <netinet/in_pcb.h>
113 #include <netinet/ip_var.h>
114 #include <netinet/ip_icmp.h>
115 #include <netinet/ip_divert.h>
117 #include <sys/thread2.h>
118 #include <sys/msgport2.h>
119 #include <net/netmsg2.h>
121 #include <sys/socketvar.h>
123 #include <net/ipfw/ip_fw.h>
124 #include <net/dummynet/ip_dummynet.h>
127 #include <netinet6/ipsec.h>
128 #include <netproto/key/key.h>
132 #include <netproto/ipsec/ipsec.h>
133 #include <netproto/ipsec/key.h>
137 static int ip_rsvp_on;
138 struct socket *ip_rsvpd;
140 int ipforwarding = 0;
141 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW,
142 &ipforwarding, 0, "Enable IP forwarding between interfaces");
144 static int ipsendredirects = 1; /* XXX */
145 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW,
146 &ipsendredirects, 0, "Enable sending IP redirects");
148 int ip_defttl = IPDEFTTL;
149 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW,
150 &ip_defttl, 0, "Maximum TTL on IP packets");
152 static int ip_dosourceroute = 0;
153 SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW,
154 &ip_dosourceroute, 0, "Enable forwarding source routed IP packets");
156 static int ip_acceptsourceroute = 0;
157 SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute,
158 CTLFLAG_RW, &ip_acceptsourceroute, 0,
159 "Enable accepting source routed IP packets");
161 static int ip_keepfaith = 0;
162 SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW,
164 "Enable packet capture for FAITH IPv4->IPv6 translator daemon");
166 static int nipq = 0; /* total # of reass queues */
168 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragpackets, CTLFLAG_RW,
170 "Maximum number of IPv4 fragment reassembly queue entries");
172 static int maxfragsperpacket;
173 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW,
174 &maxfragsperpacket, 0,
175 "Maximum number of IPv4 fragments allowed per packet");
177 static int ip_sendsourcequench = 0;
178 SYSCTL_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW,
179 &ip_sendsourcequench, 0,
180 "Enable the transmission of source quench packets");
182 int ip_do_randomid = 1;
183 SYSCTL_INT(_net_inet_ip, OID_AUTO, random_id, CTLFLAG_RW,
185 "Assign random ip_id values");
187 * XXX - Setting ip_checkinterface mostly implements the receive side of
188 * the Strong ES model described in RFC 1122, but since the routing table
189 * and transmit implementation do not implement the Strong ES model,
190 * setting this to 1 results in an odd hybrid.
192 * XXX - ip_checkinterface currently must be disabled if you use ipnat
193 * to translate the destination address to another local interface.
195 * XXX - ip_checkinterface must be disabled if you add IP aliases
196 * to the loopback interface instead of the interface where the
197 * packets for those addresses are received.
199 static int ip_checkinterface = 0;
200 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW,
201 &ip_checkinterface, 0, "Verify packet arrives on correct interface");
204 static int ipprintfs = 0;
207 extern int udp_mpsafe_proto;
208 extern int tcp_mpsafe_proto;
210 extern struct domain inetdomain;
211 extern struct protosw inetsw[];
212 u_char ip_protox[IPPROTO_MAX];
213 struct in_ifaddrhead in_ifaddrheads[MAXCPU]; /* first inet address */
214 struct in_ifaddrhashhead *in_ifaddrhashtbls[MAXCPU];
215 /* inet addr hash table */
216 u_long in_ifaddrhmask; /* mask for hash table */
218 struct ip_stats ipstats_percpu[MAXCPU];
221 sysctl_ipstats(SYSCTL_HANDLER_ARGS)
225 for (cpu = 0; cpu < ncpus; ++cpu) {
226 if ((error = SYSCTL_OUT(req, &ipstats_percpu[cpu],
227 sizeof(struct ip_stats))))
229 if ((error = SYSCTL_IN(req, &ipstats_percpu[cpu],
230 sizeof(struct ip_stats))))
236 SYSCTL_PROC(_net_inet_ip, IPCTL_STATS, stats, (CTLTYPE_OPAQUE | CTLFLAG_RW),
237 0, 0, sysctl_ipstats, "S,ip_stats", "IP statistics");
239 SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW,
240 &ipstat, ip_stats, "IP statistics");
243 /* Packet reassembly stuff */
244 #define IPREASS_NHASH_LOG2 6
245 #define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2)
246 #define IPREASS_HMASK (IPREASS_NHASH - 1)
247 #define IPREASS_HASH(x,y) \
248 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK)
250 static struct ipq ipq[IPREASS_NHASH];
253 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
254 &ip_mtu, 0, "Default MTU");
258 static int ipstealth = 0;
259 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW, &ipstealth, 0, "");
261 static const int ipstealth = 0;
264 struct mbuf *(*ip_divert_p)(struct mbuf *, int, int);
266 struct pfil_head inet_pfil_hook;
269 * struct ip_srcrt_opt is used to store packet state while it travels
272 * XXX Note that the code even makes assumptions on the size and
273 * alignment of fields inside struct ip_srcrt so e.g. adding some
274 * fields will break the code. This needs to be fixed.
276 * We need to save the IP options in case a protocol wants to respond
277 * to an incoming packet over the same route if the packet got here
278 * using IP source routing. This allows connection establishment and
279 * maintenance when the remote end is on a network that is not known
283 struct in_addr dst; /* final destination */
284 char nop; /* one NOP to align */
285 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */
286 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)];
289 struct ip_srcrt_opt {
291 struct ip_srcrt ip_srcrt;
294 static MALLOC_DEFINE(M_IPQ, "ipq", "IP Fragment Management");
295 static struct malloc_pipe ipq_mpipe;
297 static void save_rte(struct mbuf *, u_char *, struct in_addr);
298 static int ip_dooptions(struct mbuf *m, int, struct sockaddr_in *);
299 static void ip_freef(struct ipq *);
300 static void ip_input_handler(struct netmsg *);
303 * IP initialization: fill in IP protocol switch table.
304 * All protocols not implemented in kernel go to raw IP protocol handler.
316 * Make sure we can handle a reasonable number of fragments but
317 * cap it at 4000 (XXX).
319 mpipe_init(&ipq_mpipe, M_IPQ, sizeof(struct ipq),
320 IFQ_MAXLEN, 4000, 0, NULL);
321 for (i = 0; i < ncpus; ++i) {
322 TAILQ_INIT(&in_ifaddrheads[i]);
323 in_ifaddrhashtbls[i] =
324 hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask);
326 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
329 for (i = 0; i < IPPROTO_MAX; i++)
330 ip_protox[i] = pr - inetsw;
331 for (pr = inetdomain.dom_protosw;
332 pr < inetdomain.dom_protoswNPROTOSW; pr++) {
333 if (pr->pr_domain->dom_family == PF_INET && pr->pr_protocol) {
334 if (pr->pr_protocol != IPPROTO_RAW)
335 ip_protox[pr->pr_protocol] = pr - inetsw;
338 switch (pr->pr_protocol) {
340 if (tcp_mpsafe_proto)
341 pr->pr_flags |= PR_MPSAFE;
345 if (udp_mpsafe_proto)
346 pr->pr_flags |= PR_MPSAFE;
352 inet_pfil_hook.ph_type = PFIL_TYPE_AF;
353 inet_pfil_hook.ph_af = AF_INET;
354 if ((i = pfil_head_register(&inet_pfil_hook)) != 0) {
355 kprintf("%s: WARNING: unable to register pfil hook, "
356 "error %d\n", __func__, i);
359 for (i = 0; i < IPREASS_NHASH; i++)
360 ipq[i].next = ipq[i].prev = &ipq[i];
362 maxnipq = nmbclusters / 32;
363 maxfragsperpacket = 16;
365 ip_id = time_second & 0xffff;
368 * Initialize IP statistics counters for each CPU.
372 for (cpu = 0; cpu < ncpus; ++cpu) {
373 bzero(&ipstats_percpu[cpu], sizeof(struct ip_stats));
376 bzero(&ipstat, sizeof(struct ip_stats));
379 netisr_register(NETISR_IP, ip_mport_in, ip_input_handler,
380 NETISR_FLAG_NOTMPSAFE);
384 * XXX watch out this one. It is perhaps used as a cache for
385 * the most recently used route ? it is cleared in in_addroute()
386 * when a new route is successfully created.
388 struct route ipforward_rt[MAXCPU];
390 /* Do transport protocol processing. */
392 transport_processing_oncpu(struct mbuf *m, int hlen, struct ip *ip)
394 const struct protosw *pr = &inetsw[ip_protox[ip->ip_p]];
397 * Switch out to protocol's input routine.
400 pr->pr_input(m, hlen, ip->ip_p);
405 transport_processing_handler(netmsg_t netmsg)
407 struct netmsg_packet *pmsg = (struct netmsg_packet *)netmsg;
411 ip = mtod(pmsg->nm_packet, struct ip *);
412 hlen = pmsg->nm_netmsg.nm_lmsg.u.ms_result;
414 transport_processing_oncpu(pmsg->nm_packet, hlen, ip);
415 /* netmsg was embedded in the mbuf, do not reply! */
419 ip_input_handler(struct netmsg *msg0)
421 struct mbuf *m = ((struct netmsg_packet *)msg0)->nm_packet;
424 /* msg0 was embedded in the mbuf, do not reply! */
428 * IP input routine. Checksum and byte swap header. If fragmented
429 * try to reassemble. Process options. Pass to next level.
432 ip_input(struct mbuf *m)
435 struct in_ifaddr *ia = NULL;
436 struct in_ifaddr_container *iac;
439 struct in_addr pkt_dst;
440 boolean_t using_srcrt = FALSE; /* forward (by PFIL_HOOKS) */
441 boolean_t needredispatch = FALSE;
442 struct in_addr odst; /* original dst address(NAT) */
444 struct sockaddr_in *next_hop = NULL;
446 struct tdb_ident *tdbi;
447 struct secpolicy *sp;
453 if (m->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) {
455 mtag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL);
456 KKASSERT(mtag != NULL);
457 next_hop = m_tag_data(mtag);
460 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
461 /* dummynet already filtered us */
462 ip = mtod(m, struct ip *);
463 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
469 /* length checks already done in ip_demux() */
470 KASSERT(m->m_len >= sizeof(ip), ("IP header not in one mbuf"));
472 ip = mtod(m, struct ip *);
474 if (IP_VHL_V(ip->ip_vhl) != IPVERSION) {
475 ipstat.ips_badvers++;
479 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
480 /* length checks already done in ip_demux() */
481 KASSERT(hlen >= sizeof(struct ip), ("IP header len too small"));
482 KASSERT(m->m_len >= hlen, ("packet shorter than IP header length"));
484 /* 127/8 must not appear on wire - RFC1122 */
485 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
486 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
487 if (!(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK)) {
488 ipstat.ips_badaddr++;
493 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
494 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
496 if (hlen == sizeof(struct ip)) {
497 sum = in_cksum_hdr(ip);
499 sum = in_cksum(m, hlen);
508 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) {
509 /* packet is dropped by traffic conditioner */
514 * Convert fields to host representation.
516 ip->ip_len = ntohs(ip->ip_len);
517 if (ip->ip_len < hlen) {
521 ip->ip_off = ntohs(ip->ip_off);
524 * Check that the amount of data in the buffers
525 * is as at least much as the IP header would have us expect.
526 * Trim mbufs if longer than we expect.
527 * Drop packet if shorter than we expect.
529 if (m->m_pkthdr.len < ip->ip_len) {
530 ipstat.ips_tooshort++;
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);
540 #if defined(IPSEC) && !defined(IPSEC_FILTERGIF)
542 * Bypass packet filtering for packets from a tunnel (gif).
544 if (ipsec_gethist(m, NULL))
550 * Right now when no processing on packet has done
551 * and it is still fresh out of network we do our black
553 * - Firewall: deny/allow/divert
554 * - Xlate: translate packet's addr/port (NAT).
555 * - Pipe: pass pkt through dummynet.
556 * - Wrap: fake packet's addr/port <unimpl.>
557 * - Encapsulate: put it in another IP and send out. <unimp.>
562 * If we've been forwarded from the output side, then
563 * skip the firewall a second time
565 if (next_hop != NULL)
569 if (!pfil_has_hooks(&inet_pfil_hook)) {
570 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
572 * Strip dummynet tags from stranded packets
574 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
575 KKASSERT(mtag != NULL);
576 m_tag_delete(m, mtag);
577 m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED;
583 * Run through list of hooks for input packets.
585 * NB: Beware of the destination address changing (e.g.
586 * by NAT rewriting). When this happens, tell
587 * ip_forward to do the right thing.
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 needredispatch = TRUE;
609 m->m_pkthdr.fw_flags &= ~FW_MBUF_REDISPATCH;
613 * Process options and, if not destined for us,
614 * ship it on. ip_dooptions returns 1 when an
615 * error was detected (causing an icmp message
616 * to be sent and the original packet to be freed).
618 if (hlen > sizeof(struct ip) && ip_dooptions(m, 0, next_hop))
621 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
622 * matter if it is destined to another node, or whether it is
623 * a multicast one, RSVP wants it! and prevents it from being forwarded
624 * anywhere else. Also checks if the rsvp daemon is running before
625 * grabbing the packet.
627 if (rsvp_on && ip->ip_p == IPPROTO_RSVP)
631 * Check our list of addresses, to see if the packet is for us.
632 * If we don't have any addresses, assume any unicast packet
633 * we receive might be for us (and let the upper layers deal
636 if (TAILQ_EMPTY(&in_ifaddrheads[mycpuid]) &&
637 !(m->m_flags & (M_MCAST | M_BCAST)))
641 * Cache the destination address of the packet; this may be
642 * changed by use of 'ipfw fwd'.
644 pkt_dst = next_hop ? next_hop->sin_addr : ip->ip_dst;
647 * Enable a consistency check between the destination address
648 * and the arrival interface for a unicast packet (the RFC 1122
649 * strong ES model) if IP forwarding is disabled and the packet
650 * is not locally generated and the packet is not subject to
653 * XXX - Checking also should be disabled if the destination
654 * address is ipnat'ed to a different interface.
656 * XXX - Checking is incompatible with IP aliases added
657 * to the loopback interface instead of the interface where
658 * the packets are received.
660 checkif = ip_checkinterface &&
662 m->m_pkthdr.rcvif != NULL &&
663 !(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) &&
667 * Check for exact addresses in the hash bucket.
669 LIST_FOREACH(iac, INADDR_HASH(pkt_dst.s_addr), ia_hash) {
673 * If the address matches, verify that the packet
674 * arrived via the correct interface if checking is
677 if (IA_SIN(ia)->sin_addr.s_addr == pkt_dst.s_addr &&
678 (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif))
684 * Check for broadcast addresses.
686 * Only accept broadcast packets that arrive via the matching
687 * interface. Reception of forwarded directed broadcasts would
688 * be handled via ip_forward() and ether_output() with the loopback
689 * into the stack for SIMPLEX interfaces handled by ether_output().
691 if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
692 struct ifaddr_container *ifac;
694 TAILQ_FOREACH(ifac, &m->m_pkthdr.rcvif->if_addrheads[mycpuid],
696 struct ifaddr *ifa = ifac->ifa;
698 if (ifa->ifa_addr == NULL) /* shutdown/startup race */
700 if (ifa->ifa_addr->sa_family != AF_INET)
703 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
706 if (ia->ia_netbroadcast.s_addr == pkt_dst.s_addr)
709 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY)
714 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
715 struct in_multi *inm;
717 if (ip_mrouter != NULL) {
719 * If we are acting as a multicast router, all
720 * incoming multicast packets are passed to the
721 * kernel-level multicast forwarding function.
722 * The packet is returned (relatively) intact; if
723 * ip_mforward() returns a non-zero value, the packet
724 * must be discarded, else it may be accepted below.
726 if (ip_mforward != NULL &&
727 ip_mforward(ip, m->m_pkthdr.rcvif, m, NULL) != 0) {
728 ipstat.ips_cantforward++;
734 * The process-level routing daemon needs to receive
735 * all multicast IGMP packets, whether or not this
736 * host belongs to their destination groups.
738 if (ip->ip_p == IPPROTO_IGMP)
740 ipstat.ips_forward++;
743 * See if we belong to the destination multicast group on the
746 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
748 ipstat.ips_notmember++;
754 if (ip->ip_dst.s_addr == INADDR_BROADCAST)
756 if (ip->ip_dst.s_addr == INADDR_ANY)
760 * FAITH(Firewall Aided Internet Translator)
762 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
764 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP)
772 * Not for us; forward if possible and desirable.
775 ipstat.ips_cantforward++;
780 * Enforce inbound IPsec SPD.
782 if (ipsec4_in_reject(m, NULL)) {
783 ipsecstat.in_polvio++;
788 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
791 tdbi = (struct tdb_ident *)m_tag_data(mtag);
792 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
794 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
795 IP_FORWARDING, &error);
797 if (sp == NULL) { /* NB: can happen if error */
799 /*XXX error stat???*/
800 DPRINTF(("ip_input: no SP for forwarding\n")); /*XXX*/
805 * Check security policy against packet attributes.
807 error = ipsec_in_reject(sp, m);
811 ipstat.ips_cantforward++;
815 ip_forward(m, using_srcrt, next_hop);
822 * IPSTEALTH: Process non-routing options only
823 * if the packet is destined for us.
826 hlen > sizeof(struct ip) &&
827 ip_dooptions(m, 1, next_hop))
830 /* Count the packet in the ip address stats */
832 ia->ia_ifa.if_ipackets++;
833 ia->ia_ifa.if_ibytes += m->m_pkthdr.len;
837 * If offset or IP_MF are set, must reassemble.
838 * Otherwise, nothing need be done.
839 * (We could look in the reassembly queue to see
840 * if the packet was previously fragmented,
841 * but it's not worth the time; just let them time out.)
843 if (ip->ip_off & (IP_MF | IP_OFFMASK)) {
845 * Attempt reassembly; if it succeeds, proceed.
846 * ip_reass() will return a different mbuf.
851 ip = mtod(m, struct ip *);
853 /* Get the header length of the reassembled packet */
854 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
856 needredispatch = TRUE;
863 * enforce IPsec policy checking if we are seeing last header.
864 * note that we do not visit this with protocols with pcb layer
865 * code - like udp/tcp/raw ip.
867 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) &&
868 ipsec4_in_reject(m, NULL)) {
869 ipsecstat.in_polvio++;
875 * enforce IPsec policy checking if we are seeing last header.
876 * note that we do not visit this with protocols with pcb layer
877 * code - like udp/tcp/raw ip.
879 if (inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) {
881 * Check if the packet has already had IPsec processing
882 * done. If so, then just pass it along. This tag gets
883 * set during AH, ESP, etc. input handling, before the
884 * packet is returned to the ip input queue for delivery.
886 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
889 tdbi = (struct tdb_ident *)m_tag_data(mtag);
890 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
892 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
893 IP_FORWARDING, &error);
897 * Check security policy against packet attributes.
899 error = ipsec_in_reject(sp, m);
902 /* XXX error stat??? */
904 DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/
911 #endif /* FAST_IPSEC */
913 ipstat.ips_delivered++;
914 if (needredispatch) {
915 struct netmsg_packet *pmsg;
918 ip->ip_off = htons(ip->ip_off);
919 ip->ip_len = htons(ip->ip_len);
920 port = ip_mport_in(&m);
924 pmsg = &m->m_hdr.mh_netmsg;
925 netmsg_init(&pmsg->nm_netmsg, &netisr_apanic_rport, MSGF_MPSAFE,
926 transport_processing_handler);
928 pmsg->nm_netmsg.nm_lmsg.u.ms_result = hlen;
930 ip = mtod(m, struct ip *);
931 ip->ip_len = ntohs(ip->ip_len);
932 ip->ip_off = ntohs(ip->ip_off);
933 lwkt_sendmsg(port, &pmsg->nm_netmsg.nm_lmsg);
935 transport_processing_oncpu(m, hlen, ip);
944 * Take incoming datagram fragment and try to reassemble it into
945 * whole datagram. If a chain for reassembly of this datagram already
946 * exists, then it is given as fp; otherwise have to make a chain.
949 ip_reass(struct mbuf *m)
951 struct ip *ip = mtod(m, struct ip *);
952 struct mbuf *p = NULL, *q, *nq;
954 struct ipq *fp = NULL;
955 int hlen = IP_VHL_HL(ip->ip_vhl) << 2;
959 /* If maxnipq is 0, never accept fragments. */
961 ipstat.ips_fragments++;
962 ipstat.ips_fragdropped++;
967 sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
969 * Look for queue of fragments of this datagram.
971 for (fp = ipq[sum].next; fp != &ipq[sum]; fp = fp->next)
972 if (ip->ip_id == fp->ipq_id &&
973 ip->ip_src.s_addr == fp->ipq_src.s_addr &&
974 ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
975 ip->ip_p == fp->ipq_p)
981 * Enforce upper bound on number of fragmented packets
982 * for which we attempt reassembly;
983 * If maxnipq is -1, accept all fragments without limitation.
985 if (nipq > maxnipq && maxnipq > 0) {
987 * drop something from the tail of the current queue
988 * before proceeding further
990 if (ipq[sum].prev == &ipq[sum]) { /* gak */
991 for (i = 0; i < IPREASS_NHASH; i++) {
992 if (ipq[i].prev != &ipq[i]) {
993 ipstat.ips_fragtimeout +=
994 ipq[i].prev->ipq_nfrags;
995 ip_freef(ipq[i].prev);
1000 ipstat.ips_fragtimeout +=
1001 ipq[sum].prev->ipq_nfrags;
1002 ip_freef(ipq[sum].prev);
1007 * Adjust ip_len to not reflect header,
1008 * convert offset of this to bytes.
1011 if (ip->ip_off & IP_MF) {
1013 * Make sure that fragments have a data length
1014 * that's a non-zero multiple of 8 bytes.
1016 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
1017 ipstat.ips_toosmall++; /* XXX */
1021 m->m_flags |= M_FRAG;
1023 m->m_flags &= ~M_FRAG;
1026 ipstat.ips_fragments++;
1027 m->m_pkthdr.header = ip;
1030 * If the hardware has not done csum over this fragment
1031 * then csum_data is not valid at all.
1033 if ((m->m_pkthdr.csum_flags & (CSUM_FRAG_NOT_CHECKED | CSUM_DATA_VALID))
1034 == (CSUM_FRAG_NOT_CHECKED | CSUM_DATA_VALID)) {
1035 m->m_pkthdr.csum_data = 0;
1036 m->m_pkthdr.csum_flags &= ~(CSUM_DATA_VALID | CSUM_PSEUDO_HDR);
1040 * Presence of header sizes in mbufs
1041 * would confuse code below.
1047 * If first fragment to arrive, create a reassembly queue.
1050 if ((fp = mpipe_alloc_nowait(&ipq_mpipe)) == NULL)
1052 insque(fp, &ipq[sum]);
1055 fp->ipq_ttl = IPFRAGTTL;
1056 fp->ipq_p = ip->ip_p;
1057 fp->ipq_id = ip->ip_id;
1058 fp->ipq_src = ip->ip_src;
1059 fp->ipq_dst = ip->ip_dst;
1061 m->m_nextpkt = NULL;
1067 #define GETIP(m) ((struct ip*)((m)->m_pkthdr.header))
1070 * Find a segment which begins after this one does.
1072 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt)
1073 if (GETIP(q)->ip_off > ip->ip_off)
1077 * If there is a preceding segment, it may provide some of
1078 * our data already. If so, drop the data from the incoming
1079 * segment. If it provides all of our data, drop us, otherwise
1080 * stick new segment in the proper place.
1082 * If some of the data is dropped from the the preceding
1083 * segment, then it's checksum is invalidated.
1086 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off;
1088 if (i >= ip->ip_len)
1091 m->m_pkthdr.csum_flags = 0;
1095 m->m_nextpkt = p->m_nextpkt;
1098 m->m_nextpkt = fp->ipq_frags;
1103 * While we overlap succeeding segments trim them or,
1104 * if they are completely covered, dequeue them.
1106 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off;
1108 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off;
1109 if (i < GETIP(q)->ip_len) {
1110 GETIP(q)->ip_len -= i;
1111 GETIP(q)->ip_off += i;
1113 q->m_pkthdr.csum_flags = 0;
1118 ipstat.ips_fragdropped++;
1120 q->m_nextpkt = NULL;
1126 * Check for complete reassembly and perform frag per packet
1129 * Frag limiting is performed here so that the nth frag has
1130 * a chance to complete the packet before we drop the packet.
1131 * As a result, n+1 frags are actually allowed per packet, but
1132 * only n will ever be stored. (n = maxfragsperpacket.)
1136 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
1137 if (GETIP(q)->ip_off != next) {
1138 if (fp->ipq_nfrags > maxfragsperpacket) {
1139 ipstat.ips_fragdropped += fp->ipq_nfrags;
1144 next += GETIP(q)->ip_len;
1146 /* Make sure the last packet didn't have the IP_MF flag */
1147 if (p->m_flags & M_FRAG) {
1148 if (fp->ipq_nfrags > maxfragsperpacket) {
1149 ipstat.ips_fragdropped += fp->ipq_nfrags;
1156 * Reassembly is complete. Make sure the packet is a sane size.
1160 if (next + (IP_VHL_HL(ip->ip_vhl) << 2) > IP_MAXPACKET) {
1161 ipstat.ips_toolong++;
1162 ipstat.ips_fragdropped += fp->ipq_nfrags;
1168 * Concatenate fragments.
1175 q->m_nextpkt = NULL;
1176 for (q = nq; q != NULL; q = nq) {
1178 q->m_nextpkt = NULL;
1179 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
1180 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
1185 * Clean up the 1's complement checksum. Carry over 16 bits must
1186 * be added back. This assumes no more then 65535 packet fragments
1187 * were reassembled. A second carry can also occur (but not a third).
1189 m->m_pkthdr.csum_data = (m->m_pkthdr.csum_data & 0xffff) +
1190 (m->m_pkthdr.csum_data >> 16);
1191 if (m->m_pkthdr.csum_data > 0xFFFF)
1192 m->m_pkthdr.csum_data -= 0xFFFF;
1195 * Create header for new ip packet by
1196 * modifying header of first packet;
1197 * dequeue and discard fragment reassembly header.
1198 * Make header visible.
1201 ip->ip_src = fp->ipq_src;
1202 ip->ip_dst = fp->ipq_dst;
1205 mpipe_free(&ipq_mpipe, fp);
1206 m->m_len += (IP_VHL_HL(ip->ip_vhl) << 2);
1207 m->m_data -= (IP_VHL_HL(ip->ip_vhl) << 2);
1208 /* some debugging cruft by sklower, below, will go away soon */
1209 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */
1212 for (n = m; n; n = n->m_next)
1214 m->m_pkthdr.len = plen;
1217 ipstat.ips_reassembled++;
1221 ipstat.ips_fragdropped++;
1231 * Free a fragment reassembly header and all
1232 * associated datagrams.
1235 ip_freef(struct ipq *fp)
1239 while (fp->ipq_frags) {
1241 fp->ipq_frags = q->m_nextpkt;
1242 q->m_nextpkt = NULL;
1246 mpipe_free(&ipq_mpipe, fp);
1251 * IP timer processing;
1252 * if a timer expires on a reassembly
1253 * queue, discard it.
1262 for (i = 0; i < IPREASS_NHASH; i++) {
1266 while (fp != &ipq[i]) {
1269 if (fp->prev->ipq_ttl == 0) {
1270 ipstat.ips_fragtimeout += fp->prev->ipq_nfrags;
1276 * If we are over the maximum number of fragments
1277 * (due to the limit being lowered), drain off
1278 * enough to get down to the new limit.
1280 if (maxnipq >= 0 && nipq > maxnipq) {
1281 for (i = 0; i < IPREASS_NHASH; i++) {
1282 while (nipq > maxnipq &&
1283 (ipq[i].next != &ipq[i])) {
1284 ipstat.ips_fragdropped +=
1285 ipq[i].next->ipq_nfrags;
1286 ip_freef(ipq[i].next);
1295 * Drain off all datagram fragments.
1302 for (i = 0; i < IPREASS_NHASH; i++) {
1303 while (ipq[i].next != &ipq[i]) {
1304 ipstat.ips_fragdropped += ipq[i].next->ipq_nfrags;
1305 ip_freef(ipq[i].next);
1312 * Do option processing on a datagram,
1313 * possibly discarding it if bad options are encountered,
1314 * or forwarding it if source-routed.
1315 * The pass argument is used when operating in the IPSTEALTH
1316 * mode to tell what options to process:
1317 * [LS]SRR (pass 0) or the others (pass 1).
1318 * The reason for as many as two passes is that when doing IPSTEALTH,
1319 * non-routing options should be processed only if the packet is for us.
1320 * Returns 1 if packet has been forwarded/freed,
1321 * 0 if the packet should be processed further.
1324 ip_dooptions(struct mbuf *m, int pass, struct sockaddr_in *next_hop)
1326 struct sockaddr_in ipaddr = { sizeof ipaddr, AF_INET };
1327 struct ip *ip = mtod(m, struct ip *);
1329 struct in_ifaddr *ia;
1330 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB;
1331 boolean_t forward = FALSE;
1332 struct in_addr *sin, dst;
1336 cp = (u_char *)(ip + 1);
1337 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
1338 for (; cnt > 0; cnt -= optlen, cp += optlen) {
1339 opt = cp[IPOPT_OPTVAL];
1340 if (opt == IPOPT_EOL)
1342 if (opt == IPOPT_NOP)
1345 if (cnt < IPOPT_OLEN + sizeof(*cp)) {
1346 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1349 optlen = cp[IPOPT_OLEN];
1350 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
1351 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1361 * Source routing with record.
1362 * Find interface with current destination address.
1363 * If none on this machine then drop if strictly routed,
1364 * or do nothing if loosely routed.
1365 * Record interface address and bring up next address
1366 * component. If strictly routed make sure next
1367 * address is on directly accessible net.
1371 if (ipstealth && pass > 0)
1373 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1374 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1377 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1378 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1381 ipaddr.sin_addr = ip->ip_dst;
1382 ia = (struct in_ifaddr *)
1383 ifa_ifwithaddr((struct sockaddr *)&ipaddr);
1385 if (opt == IPOPT_SSRR) {
1386 type = ICMP_UNREACH;
1387 code = ICMP_UNREACH_SRCFAIL;
1390 if (!ip_dosourceroute)
1391 goto nosourcerouting;
1393 * Loose routing, and not at next destination
1394 * yet; nothing to do except forward.
1398 off--; /* 0 origin */
1399 if (off > optlen - (int)sizeof(struct in_addr)) {
1401 * End of source route. Should be for us.
1403 if (!ip_acceptsourceroute)
1404 goto nosourcerouting;
1405 save_rte(m, cp, ip->ip_src);
1410 if (!ip_dosourceroute) {
1412 char buf[sizeof "aaa.bbb.ccc.ddd"];
1415 * Acting as a router, so generate ICMP
1418 strcpy(buf, inet_ntoa(ip->ip_dst));
1420 "attempted source route from %s to %s\n",
1421 inet_ntoa(ip->ip_src), buf);
1422 type = ICMP_UNREACH;
1423 code = ICMP_UNREACH_SRCFAIL;
1427 * Not acting as a router,
1431 ipstat.ips_cantforward++;
1438 * locate outgoing interface
1440 memcpy(&ipaddr.sin_addr, cp + off,
1441 sizeof ipaddr.sin_addr);
1443 if (opt == IPOPT_SSRR) {
1444 #define INA struct in_ifaddr *
1445 #define SA struct sockaddr *
1446 if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr))
1448 ia = (INA)ifa_ifwithnet((SA)&ipaddr);
1450 ia = ip_rtaddr(ipaddr.sin_addr,
1451 &ipforward_rt[mycpuid]);
1453 type = ICMP_UNREACH;
1454 code = ICMP_UNREACH_SRCFAIL;
1457 ip->ip_dst = ipaddr.sin_addr;
1458 memcpy(cp + off, &IA_SIN(ia)->sin_addr,
1459 sizeof(struct in_addr));
1460 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1462 * Let ip_intr's mcast routing check handle mcast pkts
1464 forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr));
1468 if (ipstealth && pass == 0)
1470 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1471 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1474 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1475 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1479 * If no space remains, ignore.
1481 off--; /* 0 origin */
1482 if (off > optlen - (int)sizeof(struct in_addr))
1484 memcpy(&ipaddr.sin_addr, &ip->ip_dst,
1485 sizeof ipaddr.sin_addr);
1487 * locate outgoing interface; if we're the destination,
1488 * use the incoming interface (should be same).
1490 if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == NULL &&
1491 (ia = ip_rtaddr(ipaddr.sin_addr,
1492 &ipforward_rt[mycpuid]))
1494 type = ICMP_UNREACH;
1495 code = ICMP_UNREACH_HOST;
1498 memcpy(cp + off, &IA_SIN(ia)->sin_addr,
1499 sizeof(struct in_addr));
1500 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1504 if (ipstealth && pass == 0)
1506 code = cp - (u_char *)ip;
1507 if (optlen < 4 || optlen > 40) {
1508 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1511 if ((off = cp[IPOPT_OFFSET]) < 5) {
1512 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1515 if (off > optlen - (int)sizeof(int32_t)) {
1516 cp[IPOPT_OFFSET + 1] += (1 << 4);
1517 if ((cp[IPOPT_OFFSET + 1] & 0xf0) == 0) {
1518 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1523 off--; /* 0 origin */
1524 sin = (struct in_addr *)(cp + off);
1525 switch (cp[IPOPT_OFFSET + 1] & 0x0f) {
1527 case IPOPT_TS_TSONLY:
1530 case IPOPT_TS_TSANDADDR:
1531 if (off + sizeof(n_time) +
1532 sizeof(struct in_addr) > optlen) {
1533 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1536 ipaddr.sin_addr = dst;
1537 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr,
1541 memcpy(sin, &IA_SIN(ia)->sin_addr,
1542 sizeof(struct in_addr));
1543 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1544 off += sizeof(struct in_addr);
1547 case IPOPT_TS_PRESPEC:
1548 if (off + sizeof(n_time) +
1549 sizeof(struct in_addr) > optlen) {
1550 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1553 memcpy(&ipaddr.sin_addr, sin,
1554 sizeof(struct in_addr));
1555 if (ifa_ifwithaddr((SA)&ipaddr) == NULL)
1557 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1558 off += sizeof(struct in_addr);
1562 code = &cp[IPOPT_OFFSET + 1] - (u_char *)ip;
1566 memcpy(cp + off, &ntime, sizeof(n_time));
1567 cp[IPOPT_OFFSET] += sizeof(n_time);
1570 if (forward && ipforwarding) {
1571 ip_forward(m, TRUE, next_hop);
1576 icmp_error(m, type, code, 0, 0);
1577 ipstat.ips_badoptions++;
1582 * Given address of next destination (final or next hop),
1583 * return internet address info of interface to be used to get there.
1586 ip_rtaddr(struct in_addr dst, struct route *ro)
1588 struct sockaddr_in *sin;
1590 sin = (struct sockaddr_in *)&ro->ro_dst;
1592 if (ro->ro_rt == NULL || dst.s_addr != sin->sin_addr.s_addr) {
1593 if (ro->ro_rt != NULL) {
1597 sin->sin_family = AF_INET;
1598 sin->sin_len = sizeof *sin;
1599 sin->sin_addr = dst;
1600 rtalloc_ign(ro, RTF_PRCLONING);
1603 if (ro->ro_rt == NULL)
1606 return (ifatoia(ro->ro_rt->rt_ifa));
1610 * Save incoming source route for use in replies,
1611 * to be picked up later by ip_srcroute if the receiver is interested.
1614 save_rte(struct mbuf *m, u_char *option, struct in_addr dst)
1617 struct ip_srcrt_opt *opt;
1620 mtag = m_tag_get(PACKET_TAG_IPSRCRT, sizeof(*opt), MB_DONTWAIT);
1623 opt = m_tag_data(mtag);
1625 olen = option[IPOPT_OLEN];
1628 kprintf("save_rte: olen %d\n", olen);
1630 if (olen > sizeof(opt->ip_srcrt) - (1 + sizeof(dst))) {
1634 bcopy(option, opt->ip_srcrt.srcopt, olen);
1635 opt->ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1636 opt->ip_srcrt.dst = dst;
1637 m_tag_prepend(m, mtag);
1641 * Retrieve incoming source route for use in replies,
1642 * in the same form used by setsockopt.
1643 * The first hop is placed before the options, will be removed later.
1646 ip_srcroute(struct mbuf *m0)
1648 struct in_addr *p, *q;
1651 struct ip_srcrt_opt *opt;
1656 mtag = m_tag_find(m0, PACKET_TAG_IPSRCRT, NULL);
1659 opt = m_tag_data(mtag);
1661 if (opt->ip_nhops == 0)
1663 m = m_get(MB_DONTWAIT, MT_HEADER);
1667 #define OPTSIZ (sizeof(opt->ip_srcrt.nop) + sizeof(opt->ip_srcrt.srcopt))
1669 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
1670 m->m_len = opt->ip_nhops * sizeof(struct in_addr) +
1671 sizeof(struct in_addr) + OPTSIZ;
1674 kprintf("ip_srcroute: nhops %d mlen %d",
1675 opt->ip_nhops, m->m_len);
1680 * First save first hop for return route
1682 p = &opt->ip_srcrt.route[opt->ip_nhops - 1];
1683 *(mtod(m, struct in_addr *)) = *p--;
1686 kprintf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr));
1690 * Copy option fields and padding (nop) to mbuf.
1692 opt->ip_srcrt.nop = IPOPT_NOP;
1693 opt->ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
1694 memcpy(mtod(m, caddr_t) + sizeof(struct in_addr), &opt->ip_srcrt.nop,
1696 q = (struct in_addr *)(mtod(m, caddr_t) +
1697 sizeof(struct in_addr) + OPTSIZ);
1700 * Record return path as an IP source route,
1701 * reversing the path (pointers are now aligned).
1703 while (p >= opt->ip_srcrt.route) {
1706 kprintf(" %x", ntohl(q->s_addr));
1711 * Last hop goes to final destination.
1713 *q = opt->ip_srcrt.dst;
1714 m_tag_delete(m0, mtag);
1717 kprintf(" %x\n", ntohl(q->s_addr));
1723 * Strip out IP options.
1726 ip_stripoptions(struct mbuf *m)
1729 struct ip *ip = mtod(m, struct ip *);
1733 optlen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
1734 opts = (caddr_t)(ip + 1);
1735 datalen = m->m_len - (sizeof(struct ip) + optlen);
1736 bcopy(opts + optlen, opts, datalen);
1738 if (m->m_flags & M_PKTHDR)
1739 m->m_pkthdr.len -= optlen;
1740 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, sizeof(struct ip) >> 2);
1743 u_char inetctlerrmap[PRC_NCMDS] = {
1745 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
1746 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
1747 EMSGSIZE, EHOSTUNREACH, 0, 0,
1749 ENOPROTOOPT, ECONNREFUSED
1753 * Forward a packet. If some error occurs return the sender
1754 * an icmp packet. Note we can't always generate a meaningful
1755 * icmp message because icmp doesn't have a large enough repertoire
1756 * of codes and types.
1758 * If not forwarding, just drop the packet. This could be confusing
1759 * if ipforwarding was zero but some routing protocol was advancing
1760 * us as a gateway to somewhere. However, we must let the routing
1761 * protocol deal with that.
1763 * The using_srcrt parameter indicates whether the packet is being forwarded
1764 * via a source route.
1767 ip_forward(struct mbuf *m, boolean_t using_srcrt, struct sockaddr_in *next_hop)
1769 struct ip *ip = mtod(m, struct ip *);
1770 struct sockaddr_in *ipforward_rtaddr;
1772 int error, type = 0, code = 0, destmtu = 0;
1775 struct in_addr pkt_dst;
1776 struct route *cache_rt = &ipforward_rt[mycpuid];
1780 * Cache the destination address of the packet; this may be
1781 * changed by use of 'ipfw fwd'.
1783 pkt_dst = (next_hop != NULL) ? next_hop->sin_addr : ip->ip_dst;
1787 kprintf("forward: src %x dst %x ttl %x\n",
1788 ip->ip_src.s_addr, pkt_dst.s_addr, ip->ip_ttl);
1791 if (m->m_flags & (M_BCAST | M_MCAST) || !in_canforward(pkt_dst)) {
1792 ipstat.ips_cantforward++;
1796 if (!ipstealth && ip->ip_ttl <= IPTTLDEC) {
1797 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0);
1801 ipforward_rtaddr = (struct sockaddr_in *) &cache_rt->ro_dst;
1802 if (cache_rt->ro_rt == NULL ||
1803 ipforward_rtaddr->sin_addr.s_addr != pkt_dst.s_addr) {
1804 if (cache_rt->ro_rt != NULL) {
1805 RTFREE(cache_rt->ro_rt);
1806 cache_rt->ro_rt = NULL;
1808 ipforward_rtaddr->sin_family = AF_INET;
1809 ipforward_rtaddr->sin_len = sizeof(struct sockaddr_in);
1810 ipforward_rtaddr->sin_addr = pkt_dst;
1811 rtalloc_ign(cache_rt, RTF_PRCLONING);
1812 if (cache_rt->ro_rt == NULL) {
1813 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0);
1817 rt = cache_rt->ro_rt;
1820 * Save the IP header and at most 8 bytes of the payload,
1821 * in case we need to generate an ICMP message to the src.
1823 * XXX this can be optimized a lot by saving the data in a local
1824 * buffer on the stack (72 bytes at most), and only allocating the
1825 * mbuf if really necessary. The vast majority of the packets
1826 * are forwarded without having to send an ICMP back (either
1827 * because unnecessary, or because rate limited), so we are
1828 * really we are wasting a lot of work here.
1830 * We don't use m_copy() because it might return a reference
1831 * to a shared cluster. Both this function and ip_output()
1832 * assume exclusive access to the IP header in `m', so any
1833 * data in a cluster may change before we reach icmp_error().
1835 MGETHDR(mcopy, MB_DONTWAIT, m->m_type);
1836 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, MB_DONTWAIT)) {
1838 * It's probably ok if the pkthdr dup fails (because
1839 * the deep copy of the tag chain failed), but for now
1840 * be conservative and just discard the copy since
1841 * code below may some day want the tags.
1846 if (mcopy != NULL) {
1847 mcopy->m_len = imin((IP_VHL_HL(ip->ip_vhl) << 2) + 8,
1849 mcopy->m_pkthdr.len = mcopy->m_len;
1850 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1854 ip->ip_ttl -= IPTTLDEC;
1857 * If forwarding packet using same interface that it came in on,
1858 * perhaps should send a redirect to sender to shortcut a hop.
1859 * Only send redirect if source is sending directly to us,
1860 * and if packet was not source routed (or has any options).
1861 * Also, don't send redirect if forwarding using a default route
1862 * or a route modified by a redirect.
1864 if (rt->rt_ifp == m->m_pkthdr.rcvif &&
1865 !(rt->rt_flags & (RTF_DYNAMIC | RTF_MODIFIED)) &&
1866 satosin(rt_key(rt))->sin_addr.s_addr != INADDR_ANY &&
1867 ipsendredirects && !using_srcrt && next_hop == NULL) {
1868 u_long src = ntohl(ip->ip_src.s_addr);
1869 struct in_ifaddr *rt_ifa = (struct in_ifaddr *)rt->rt_ifa;
1871 if (rt_ifa != NULL &&
1872 (src & rt_ifa->ia_subnetmask) == rt_ifa->ia_subnet) {
1873 if (rt->rt_flags & RTF_GATEWAY)
1874 dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
1876 dest = pkt_dst.s_addr;
1878 * Router requirements says to only send
1881 type = ICMP_REDIRECT;
1882 code = ICMP_REDIRECT_HOST;
1885 kprintf("redirect (%d) to %x\n", code, dest);
1890 error = ip_output(m, NULL, cache_rt, IP_FORWARDING, NULL, NULL);
1892 ipstat.ips_forward++;
1895 ipflow_create(cache_rt, mcopy);
1898 return; /* most common case */
1900 ipstat.ips_redirectsent++;
1903 ipstat.ips_cantforward++;
1910 * Send ICMP message.
1915 case 0: /* forwarded, but need redirect */
1916 /* type, code set above */
1919 case ENETUNREACH: /* shouldn't happen, checked above */
1924 type = ICMP_UNREACH;
1925 code = ICMP_UNREACH_HOST;
1929 type = ICMP_UNREACH;
1930 code = ICMP_UNREACH_NEEDFRAG;
1933 * If the packet is routed over IPsec tunnel, tell the
1934 * originator the tunnel MTU.
1935 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
1938 if (cache_rt->ro_rt != NULL) {
1939 struct secpolicy *sp = NULL;
1944 sp = ipsec4_getpolicybyaddr(mcopy,
1950 destmtu = cache_rt->ro_rt->rt_ifp->if_mtu;
1952 /* count IPsec header size */
1953 ipsechdr = ipsec4_hdrsiz(mcopy,
1958 * find the correct route for outer IPv4
1959 * header, compute tunnel MTU.
1962 if (sp->req != NULL && sp->req->sav != NULL &&
1963 sp->req->sav->sah != NULL) {
1964 ro = &sp->req->sav->sah->sa_route;
1965 if (ro->ro_rt != NULL &&
1966 ro->ro_rt->rt_ifp != NULL) {
1968 ro->ro_rt->rt_ifp->if_mtu;
1969 destmtu -= ipsechdr;
1978 * If the packet is routed over IPsec tunnel, tell the
1979 * originator the tunnel MTU.
1980 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
1983 if (cache_rt->ro_rt != NULL) {
1984 struct secpolicy *sp = NULL;
1989 sp = ipsec_getpolicybyaddr(mcopy,
1995 destmtu = cache_rt->ro_rt->rt_ifp->if_mtu;
1997 /* count IPsec header size */
1998 ipsechdr = ipsec4_hdrsiz(mcopy,
2003 * find the correct route for outer IPv4
2004 * header, compute tunnel MTU.
2007 if (sp->req != NULL &&
2008 sp->req->sav != NULL &&
2009 sp->req->sav->sah != NULL) {
2010 ro = &sp->req->sav->sah->sa_route;
2011 if (ro->ro_rt != NULL &&
2012 ro->ro_rt->rt_ifp != NULL) {
2014 ro->ro_rt->rt_ifp->if_mtu;
2015 destmtu -= ipsechdr;
2022 #else /* !IPSEC && !FAST_IPSEC */
2023 if (cache_rt->ro_rt != NULL)
2024 destmtu = cache_rt->ro_rt->rt_ifp->if_mtu;
2026 ipstat.ips_cantfrag++;
2031 * A router should not generate ICMP_SOURCEQUENCH as
2032 * required in RFC1812 Requirements for IP Version 4 Routers.
2033 * Source quench could be a big problem under DoS attacks,
2034 * or if the underlying interface is rate-limited.
2035 * Those who need source quench packets may re-enable them
2036 * via the net.inet.ip.sendsourcequench sysctl.
2038 if (!ip_sendsourcequench) {
2042 type = ICMP_SOURCEQUENCH;
2047 case EACCES: /* ipfw denied packet */
2051 icmp_error(mcopy, type, code, dest, destmtu);
2055 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
2058 if (inp->inp_socket->so_options & SO_TIMESTAMP) {
2062 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
2063 SCM_TIMESTAMP, SOL_SOCKET);
2065 mp = &(*mp)->m_next;
2067 if (inp->inp_flags & INP_RECVDSTADDR) {
2068 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
2069 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
2071 mp = &(*mp)->m_next;
2073 if (inp->inp_flags & INP_RECVTTL) {
2074 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl,
2075 sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
2077 mp = &(*mp)->m_next;
2081 * Moving these out of udp_input() made them even more broken
2082 * than they already were.
2084 /* options were tossed already */
2085 if (inp->inp_flags & INP_RECVOPTS) {
2086 *mp = sbcreatecontrol((caddr_t) opts_deleted_above,
2087 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
2089 mp = &(*mp)->m_next;
2091 /* ip_srcroute doesn't do what we want here, need to fix */
2092 if (inp->inp_flags & INP_RECVRETOPTS) {
2093 *mp = sbcreatecontrol((caddr_t) ip_srcroute(m),
2094 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
2096 mp = &(*mp)->m_next;
2099 if (inp->inp_flags & INP_RECVIF) {
2102 struct sockaddr_dl sdl;
2105 struct sockaddr_dl *sdp;
2106 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
2108 if (((ifp = m->m_pkthdr.rcvif)) &&
2109 ((ifp->if_index != 0) && (ifp->if_index <= if_index))) {
2110 sdp = IF_LLSOCKADDR(ifp);
2112 * Change our mind and don't try copy.
2114 if ((sdp->sdl_family != AF_LINK) ||
2115 (sdp->sdl_len > sizeof(sdlbuf))) {
2118 bcopy(sdp, sdl2, sdp->sdl_len);
2122 offsetof(struct sockaddr_dl, sdl_data[0]);
2123 sdl2->sdl_family = AF_LINK;
2124 sdl2->sdl_index = 0;
2125 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
2127 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len,
2128 IP_RECVIF, IPPROTO_IP);
2130 mp = &(*mp)->m_next;
2135 * XXX these routines are called from the upper part of the kernel.
2137 * They could also be moved to ip_mroute.c, since all the RSVP
2138 * handling is done there already.
2141 ip_rsvp_init(struct socket *so)
2143 if (so->so_type != SOCK_RAW ||
2144 so->so_proto->pr_protocol != IPPROTO_RSVP)
2147 if (ip_rsvpd != NULL)
2152 * This may seem silly, but we need to be sure we don't over-increment
2153 * the RSVP counter, in case something slips up.
2168 * This may seem silly, but we need to be sure we don't over-decrement
2169 * the RSVP counter, in case something slips up.
2179 rsvp_input(struct mbuf *m, ...) /* XXX must fixup manually */
2185 off = __va_arg(ap, int);
2186 proto = __va_arg(ap, int);
2189 if (rsvp_input_p) { /* call the real one if loaded */
2190 rsvp_input_p(m, off, proto);
2194 /* Can still get packets with rsvp_on = 0 if there is a local member
2195 * of the group to which the RSVP packet is addressed. But in this
2196 * case we want to throw the packet away.
2204 if (ip_rsvpd != NULL) {
2205 rip_input(m, off, proto);
2208 /* Drop the packet */