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. Neither the name of the University nor the names of its contributors
47 * may be used to endorse or promote products derived from this software
48 * without specific prior written permission.
50 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
51 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
52 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
53 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
54 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
55 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
56 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
57 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
58 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
59 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
62 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94
63 * $FreeBSD: src/sys/netinet/ip_input.c,v 1.130.2.52 2003/03/07 07:01:28 silby Exp $
68 #include "opt_bootp.h"
70 #include "opt_ipdivert.h"
71 #include "opt_ipstealth.h"
72 #include "opt_ipsec.h"
75 #include <sys/param.h>
76 #include <sys/systm.h>
78 #include <sys/malloc.h>
79 #include <sys/mpipe.h>
80 #include <sys/domain.h>
81 #include <sys/protosw.h>
82 #include <sys/socket.h>
84 #include <sys/globaldata.h>
85 #include <sys/thread.h>
86 #include <sys/kernel.h>
87 #include <sys/syslog.h>
88 #include <sys/sysctl.h>
89 #include <sys/in_cksum.h>
92 #include <sys/mplock2.h>
94 #include <machine/stdarg.h>
97 #include <net/if_types.h>
98 #include <net/if_var.h>
99 #include <net/if_dl.h>
100 #include <net/pfil.h>
101 #include <net/route.h>
102 #include <net/netisr2.h>
104 #include <netinet/in.h>
105 #include <netinet/in_systm.h>
106 #include <netinet/in_var.h>
107 #include <netinet/ip.h>
108 #include <netinet/in_pcb.h>
109 #include <netinet/ip_var.h>
110 #include <netinet/ip_icmp.h>
111 #include <netinet/ip_divert.h>
112 #include <netinet/ip_flow.h>
114 #include <sys/thread2.h>
115 #include <sys/msgport2.h>
116 #include <net/netmsg2.h>
118 #include <sys/socketvar.h>
120 #include <net/ipfw/ip_fw.h>
121 #include <net/dummynet/ip_dummynet.h>
124 #include <netinet6/ipsec.h>
125 #include <netproto/key/key.h>
129 #include <netproto/ipsec/ipsec.h>
130 #include <netproto/ipsec/key.h>
134 static int ip_rsvp_on;
135 struct socket *ip_rsvpd;
137 int ipforwarding = 0;
138 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW,
139 &ipforwarding, 0, "Enable IP forwarding between interfaces");
141 static int ipsendredirects = 1; /* XXX */
142 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW,
143 &ipsendredirects, 0, "Enable sending IP redirects");
145 int ip_defttl = IPDEFTTL;
146 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW,
147 &ip_defttl, 0, "Maximum TTL on IP packets");
149 static int ip_dosourceroute = 0;
150 SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW,
151 &ip_dosourceroute, 0, "Enable forwarding source routed IP packets");
153 static int ip_acceptsourceroute = 0;
154 SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute,
155 CTLFLAG_RW, &ip_acceptsourceroute, 0,
156 "Enable accepting source routed IP packets");
158 static int ip_keepfaith = 0;
159 SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW,
161 "Enable packet capture for FAITH IPv4->IPv6 translator daemon");
164 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragpackets, CTLFLAG_RW,
166 "Maximum number of IPv4 fragment reassembly queue entries");
168 static int maxfragsperpacket;
169 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW,
170 &maxfragsperpacket, 0,
171 "Maximum number of IPv4 fragments allowed per packet");
173 static int ip_sendsourcequench = 0;
174 SYSCTL_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW,
175 &ip_sendsourcequench, 0,
176 "Enable the transmission of source quench packets");
178 int ip_do_randomid = 1;
179 SYSCTL_INT(_net_inet_ip, OID_AUTO, random_id, CTLFLAG_RW,
181 "Assign random ip_id values");
183 * XXX - Setting ip_checkinterface mostly implements the receive side of
184 * the Strong ES model described in RFC 1122, but since the routing table
185 * and transmit implementation do not implement the Strong ES model,
186 * setting this to 1 results in an odd hybrid.
188 * XXX - ip_checkinterface currently must be disabled if you use ipnat
189 * to translate the destination address to another local interface.
191 * XXX - ip_checkinterface must be disabled if you add IP aliases
192 * to the loopback interface instead of the interface where the
193 * packets for those addresses are received.
195 static int ip_checkinterface = 0;
196 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW,
197 &ip_checkinterface, 0, "Verify packet arrives on correct interface");
199 static u_long ip_hash_count = 0;
200 SYSCTL_ULONG(_net_inet_ip, OID_AUTO, hash_count, CTLFLAG_RD,
201 &ip_hash_count, 0, "Number of packets hashed by IP");
204 static u_long ip_rehash_count = 0;
205 SYSCTL_ULONG(_net_inet_ip, OID_AUTO, rehash_count, CTLFLAG_RD,
206 &ip_rehash_count, 0, "Number of packets rehashed by IP");
208 static u_long ip_dispatch_fast = 0;
209 SYSCTL_ULONG(_net_inet_ip, OID_AUTO, dispatch_fast_count, CTLFLAG_RD,
210 &ip_dispatch_fast, 0, "Number of packets handled on current CPU");
212 static u_long ip_dispatch_slow = 0;
213 SYSCTL_ULONG(_net_inet_ip, OID_AUTO, dispatch_slow_count, CTLFLAG_RD,
214 &ip_dispatch_slow, 0, "Number of packets messaged to another CPU");
218 static int ipprintfs = 0;
221 extern struct domain inetdomain;
222 extern struct protosw inetsw[];
223 u_char ip_protox[IPPROTO_MAX];
224 struct in_ifaddrhead in_ifaddrheads[MAXCPU]; /* first inet address */
225 struct in_ifaddrhashhead *in_ifaddrhashtbls[MAXCPU];
226 /* inet addr hash table */
227 u_long in_ifaddrhmask; /* mask for hash table */
229 static struct mbuf *ipforward_mtemp[MAXCPU];
231 struct ip_stats ipstats_percpu[MAXCPU] __cachealign;
234 sysctl_ipstats(SYSCTL_HANDLER_ARGS)
238 for (cpu = 0; cpu < ncpus; ++cpu) {
239 if ((error = SYSCTL_OUT(req, &ipstats_percpu[cpu],
240 sizeof(struct ip_stats))))
242 if ((error = SYSCTL_IN(req, &ipstats_percpu[cpu],
243 sizeof(struct ip_stats))))
249 SYSCTL_PROC(_net_inet_ip, IPCTL_STATS, stats, (CTLTYPE_OPAQUE | CTLFLAG_RW),
250 0, 0, sysctl_ipstats, "S,ip_stats", "IP statistics");
252 /* Packet reassembly stuff */
253 #define IPREASS_NHASH_LOG2 6
254 #define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2)
255 #define IPREASS_HMASK (IPREASS_NHASH - 1)
256 #define IPREASS_HASH(x,y) \
257 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK)
259 TAILQ_HEAD(ipqhead, ipq);
260 struct ipfrag_queue {
262 struct netmsg_base timeo_netmsg;
263 struct netmsg_base drain_netmsg;
264 struct ipqhead ipq[IPREASS_NHASH];
267 static struct ipfrag_queue ipfrag_queue_pcpu[MAXCPU];
270 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
271 &ip_mtu, 0, "Default MTU");
275 static int ipstealth = 0;
276 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW, &ipstealth, 0, "");
278 static const int ipstealth = 0;
281 struct mbuf *(*ip_divert_p)(struct mbuf *, int, int);
283 struct pfil_head inet_pfil_hook;
286 * struct ip_srcrt_opt is used to store packet state while it travels
289 * XXX Note that the code even makes assumptions on the size and
290 * alignment of fields inside struct ip_srcrt so e.g. adding some
291 * fields will break the code. This needs to be fixed.
293 * We need to save the IP options in case a protocol wants to respond
294 * to an incoming packet over the same route if the packet got here
295 * using IP source routing. This allows connection establishment and
296 * maintenance when the remote end is on a network that is not known
300 struct in_addr dst; /* final destination */
301 char nop; /* one NOP to align */
302 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */
303 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)];
306 struct ip_srcrt_opt {
308 struct ip_srcrt ip_srcrt;
311 #define IPFRAG_MPIPE_MAX 4096
312 #define MAXIPFRAG_MIN ((IPFRAG_MPIPE_MAX * 2) / 256)
314 static MALLOC_DEFINE(M_IPQ, "ipq", "IP Fragment Management");
315 static struct malloc_pipe ipq_mpipe;
317 static void save_rte(struct mbuf *, u_char *, struct in_addr);
318 static int ip_dooptions(struct mbuf *m, int, struct sockaddr_in *);
319 static void ip_freef(struct ipfrag_queue *, struct ipqhead *,
321 static void ip_input_handler(netmsg_t);
323 static void ipfrag_timeo_dispatch(netmsg_t);
324 static void ipfrag_drain_dispatch(netmsg_t);
327 * IP initialization: fill in IP protocol switch table.
328 * All protocols not implemented in kernel go to raw IP protocol handler.
337 * Make sure we can handle a reasonable number of fragments but
338 * cap it at IPFRAG_MPIPE_MAX.
340 mpipe_init(&ipq_mpipe, M_IPQ, sizeof(struct ipq),
341 IFQ_MAXLEN, IPFRAG_MPIPE_MAX, 0, NULL, NULL, NULL);
342 for (cpu = 0; cpu < ncpus; ++cpu) {
343 TAILQ_INIT(&in_ifaddrheads[cpu]);
344 in_ifaddrhashtbls[cpu] =
345 hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask);
347 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
350 for (i = 0; i < IPPROTO_MAX; i++)
351 ip_protox[i] = pr - inetsw;
352 for (pr = inetdomain.dom_protosw;
353 pr < inetdomain.dom_protoswNPROTOSW; pr++) {
354 if (pr->pr_domain->dom_family == PF_INET && pr->pr_protocol) {
355 if (pr->pr_protocol != IPPROTO_RAW)
356 ip_protox[pr->pr_protocol] = pr - inetsw;
360 inet_pfil_hook.ph_type = PFIL_TYPE_AF;
361 inet_pfil_hook.ph_af = AF_INET;
362 if ((i = pfil_head_register(&inet_pfil_hook)) != 0) {
363 kprintf("%s: WARNING: unable to register pfil hook, "
364 "error %d\n", __func__, i);
367 maxnipq = (nmbclusters / 32) / ncpus2;
368 if (maxnipq < MAXIPFRAG_MIN)
369 maxnipq = MAXIPFRAG_MIN;
370 maxfragsperpacket = 16;
372 ip_id = time_second & 0xffff; /* time_second survives reboots */
374 for (cpu = 0; cpu < ncpus; ++cpu) {
376 * Initialize IP statistics counters for each CPU.
378 bzero(&ipstats_percpu[cpu], sizeof(struct ip_stats));
381 * Preallocate mbuf template for forwarding
383 MGETHDR(ipforward_mtemp[cpu], MB_WAIT, MT_DATA);
386 * Initialize per-cpu ip fragments queues
388 for (i = 0; i < IPREASS_NHASH; i++) {
389 struct ipfrag_queue *fragq = &ipfrag_queue_pcpu[cpu];
391 TAILQ_INIT(&fragq->ipq[i]);
392 netmsg_init(&fragq->timeo_netmsg, NULL,
393 &netisr_adone_rport, 0, ipfrag_timeo_dispatch);
394 netmsg_init(&fragq->drain_netmsg, NULL,
395 &netisr_adone_rport, 0, ipfrag_drain_dispatch);
399 netisr_register(NETISR_IP, ip_input_handler, ip_hashfn_in);
400 netisr_register_hashcheck(NETISR_IP, ip_hashcheck);
403 /* Do transport protocol processing. */
405 transport_processing_oncpu(struct mbuf *m, int hlen, struct ip *ip)
407 const struct protosw *pr = &inetsw[ip_protox[ip->ip_p]];
410 * Switch out to protocol's input routine.
413 pr->pr_input(&m, &hlen, ip->ip_p);
418 transport_processing_handler(netmsg_t msg)
420 struct netmsg_packet *pmsg = &msg->packet;
424 ip = mtod(pmsg->nm_packet, struct ip *);
425 hlen = pmsg->base.lmsg.u.ms_result;
427 transport_processing_oncpu(pmsg->nm_packet, hlen, ip);
428 /* msg was embedded in the mbuf, do not reply! */
432 ip_input_handler(netmsg_t msg)
434 ip_input(msg->packet.nm_packet);
435 /* msg was embedded in the mbuf, do not reply! */
439 * IP input routine. Checksum and byte swap header. If fragmented
440 * try to reassemble. Process options. Pass to next level.
443 ip_input(struct mbuf *m)
446 struct in_ifaddr *ia = NULL;
447 struct in_ifaddr_container *iac;
450 struct in_addr pkt_dst;
451 boolean_t using_srcrt = FALSE; /* forward (by PFIL_HOOKS) */
452 struct in_addr odst; /* original dst address(NAT) */
454 struct sockaddr_in *next_hop = NULL;
457 struct tdb_ident *tdbi;
458 struct secpolicy *sp;
465 * This routine is called from numerous places which may not have
466 * characterized the packet.
468 ip = mtod(m, struct ip *);
469 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) ||
470 ntohs(ip->ip_off) & (IP_MF | IP_OFFMASK)) {
472 * Force hash recalculation for fragments and multicast
473 * packets; hardware may not do it correctly.
474 * XXX add flag to indicate the hash is from hardware
476 m->m_flags &= ~M_HASH;
478 if ((m->m_flags & M_HASH) == 0) {
479 ip_hashfn(&m, 0, IP_MPORT_IN);
482 KKASSERT(m->m_flags & M_HASH);
484 if (&curthread->td_msgport !=
485 netisr_hashport(m->m_pkthdr.hash)) {
486 netisr_queue(NETISR_IP, m);
487 /* Requeued to other netisr msgport; done */
491 /* mbuf could have been changed */
492 ip = mtod(m, struct ip *);
496 * Pull out certain tags
498 if (m->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) {
500 mtag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL);
501 KKASSERT(mtag != NULL);
502 next_hop = m_tag_data(mtag);
505 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
506 /* dummynet already filtered us */
507 ip = mtod(m, struct ip *);
508 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
514 /* length checks already done in ip_hashfn() */
515 KASSERT(m->m_len >= sizeof(struct ip), ("IP header not in one mbuf"));
517 if (IP_VHL_V(ip->ip_vhl) != IPVERSION) {
518 ipstat.ips_badvers++;
522 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
523 /* length checks already done in ip_hashfn() */
524 KASSERT(hlen >= sizeof(struct ip), ("IP header len too small"));
525 KASSERT(m->m_len >= hlen, ("complete IP header not in one mbuf"));
527 /* 127/8 must not appear on wire - RFC1122 */
528 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
529 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
530 if (!(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK)) {
531 ipstat.ips_badaddr++;
536 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
537 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
539 if (hlen == sizeof(struct ip))
540 sum = in_cksum_hdr(ip);
542 sum = in_cksum(m, hlen);
550 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) {
551 /* packet is dropped by traffic conditioner */
556 * Convert fields to host representation.
558 ip->ip_len = ntohs(ip->ip_len);
559 ip->ip_off = ntohs(ip->ip_off);
561 /* length checks already done in ip_hashfn() */
562 KASSERT(ip->ip_len >= hlen, ("total length less then header length"));
563 KASSERT(m->m_pkthdr.len >= ip->ip_len, ("mbuf too short"));
566 * Trim mbufs if longer than the IP header would have us expect.
568 if (m->m_pkthdr.len > ip->ip_len) {
569 if (m->m_len == m->m_pkthdr.len) {
570 m->m_len = ip->ip_len;
571 m->m_pkthdr.len = ip->ip_len;
573 m_adj(m, ip->ip_len - m->m_pkthdr.len);
576 #if defined(IPSEC) && !defined(IPSEC_FILTERGIF)
578 * Bypass packet filtering for packets from a tunnel (gif).
580 if (ipsec_gethist(m, NULL))
586 * Right now when no processing on packet has done
587 * and it is still fresh out of network we do our black
589 * - Firewall: deny/allow/divert
590 * - Xlate: translate packet's addr/port (NAT).
591 * - Pipe: pass pkt through dummynet.
592 * - Wrap: fake packet's addr/port <unimpl.>
593 * - Encapsulate: put it in another IP and send out. <unimp.>
598 * If we've been forwarded from the output side, then
599 * skip the firewall a second time
601 if (next_hop != NULL)
605 if (!pfil_has_hooks(&inet_pfil_hook)) {
606 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
608 * Strip dummynet tags from stranded packets
610 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
611 KKASSERT(mtag != NULL);
612 m_tag_delete(m, mtag);
613 m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED;
619 * Run through list of hooks for input packets.
621 * NOTE! If the packet is rewritten pf/ipfw/whoever must
625 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif, PFIL_IN))
627 if (m == NULL) /* consumed by filter */
629 ip = mtod(m, struct ip *);
630 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
631 using_srcrt = (odst.s_addr != ip->ip_dst.s_addr);
633 if (m->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) {
634 mtag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL);
635 KKASSERT(mtag != NULL);
636 next_hop = m_tag_data(mtag);
638 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
642 if (m->m_pkthdr.fw_flags & FW_MBUF_REDISPATCH) {
643 m->m_pkthdr.fw_flags &= ~FW_MBUF_REDISPATCH;
647 * Process options and, if not destined for us,
648 * ship it on. ip_dooptions returns 1 when an
649 * error was detected (causing an icmp message
650 * to be sent and the original packet to be freed).
652 if (hlen > sizeof(struct ip) && ip_dooptions(m, 0, next_hop))
655 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
656 * matter if it is destined to another node, or whether it is
657 * a multicast one, RSVP wants it! and prevents it from being forwarded
658 * anywhere else. Also checks if the rsvp daemon is running before
659 * grabbing the packet.
661 if (rsvp_on && ip->ip_p == IPPROTO_RSVP)
665 * Check our list of addresses, to see if the packet is for us.
666 * If we don't have any addresses, assume any unicast packet
667 * we receive might be for us (and let the upper layers deal
670 if (TAILQ_EMPTY(&in_ifaddrheads[mycpuid]) &&
671 !(m->m_flags & (M_MCAST | M_BCAST)))
675 * Cache the destination address of the packet; this may be
676 * changed by use of 'ipfw fwd'.
678 pkt_dst = next_hop ? next_hop->sin_addr : ip->ip_dst;
681 * Enable a consistency check between the destination address
682 * and the arrival interface for a unicast packet (the RFC 1122
683 * strong ES model) if IP forwarding is disabled and the packet
684 * is not locally generated and the packet is not subject to
687 * XXX - Checking also should be disabled if the destination
688 * address is ipnat'ed to a different interface.
690 * XXX - Checking is incompatible with IP aliases added
691 * to the loopback interface instead of the interface where
692 * the packets are received.
694 checkif = ip_checkinterface &&
696 m->m_pkthdr.rcvif != NULL &&
697 !(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) &&
701 * Check for exact addresses in the hash bucket.
703 LIST_FOREACH(iac, INADDR_HASH(pkt_dst.s_addr), ia_hash) {
707 * If the address matches, verify that the packet
708 * arrived via the correct interface if checking is
711 if (IA_SIN(ia)->sin_addr.s_addr == pkt_dst.s_addr &&
712 (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif))
718 * Check for broadcast addresses.
720 * Only accept broadcast packets that arrive via the matching
721 * interface. Reception of forwarded directed broadcasts would
722 * be handled via ip_forward() and ether_output() with the loopback
723 * into the stack for SIMPLEX interfaces handled by ether_output().
725 if (m->m_pkthdr.rcvif != NULL &&
726 m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
727 struct ifaddr_container *ifac;
729 TAILQ_FOREACH(ifac, &m->m_pkthdr.rcvif->if_addrheads[mycpuid],
731 struct ifaddr *ifa = ifac->ifa;
733 if (ifa->ifa_addr == NULL) /* shutdown/startup race */
735 if (ifa->ifa_addr->sa_family != AF_INET)
738 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
741 if (ia->ia_netbroadcast.s_addr == pkt_dst.s_addr)
744 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY)
749 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
750 struct in_multi *inm;
752 if (ip_mrouter != NULL) {
753 /* XXX Multicast routing is not MPSAFE yet */
757 * If we are acting as a multicast router, all
758 * incoming multicast packets are passed to the
759 * kernel-level multicast forwarding function.
760 * The packet is returned (relatively) intact; if
761 * ip_mforward() returns a non-zero value, the packet
762 * must be discarded, else it may be accepted below.
764 if (ip_mforward != NULL &&
765 ip_mforward(ip, m->m_pkthdr.rcvif, m, NULL) != 0) {
767 ipstat.ips_cantforward++;
775 * The process-level routing daemon needs to receive
776 * all multicast IGMP packets, whether or not this
777 * host belongs to their destination groups.
779 if (ip->ip_p == IPPROTO_IGMP)
781 ipstat.ips_forward++;
784 * See if we belong to the destination multicast group on the
787 inm = IN_LOOKUP_MULTI(&ip->ip_dst, m->m_pkthdr.rcvif);
789 ipstat.ips_notmember++;
795 if (ip->ip_dst.s_addr == INADDR_BROADCAST)
797 if (ip->ip_dst.s_addr == INADDR_ANY)
801 * FAITH(Firewall Aided Internet Translator)
803 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
805 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP)
813 * Not for us; forward if possible and desirable.
816 ipstat.ips_cantforward++;
821 * Enforce inbound IPsec SPD.
823 if (ipsec4_in_reject(m, NULL)) {
824 ipsecstat.in_polvio++;
829 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
832 tdbi = (struct tdb_ident *)m_tag_data(mtag);
833 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
835 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
836 IP_FORWARDING, &error);
838 if (sp == NULL) { /* NB: can happen if error */
840 /*XXX error stat???*/
841 DPRINTF(("ip_input: no SP for forwarding\n")); /*XXX*/
846 * Check security policy against packet attributes.
848 error = ipsec_in_reject(sp, m);
852 ipstat.ips_cantforward++;
856 ip_forward(m, using_srcrt, next_hop);
863 * IPSTEALTH: Process non-routing options only
864 * if the packet is destined for us.
867 hlen > sizeof(struct ip) &&
868 ip_dooptions(m, 1, next_hop))
871 /* Count the packet in the ip address stats */
873 IFA_STAT_INC(&ia->ia_ifa, ipackets, 1);
874 IFA_STAT_INC(&ia->ia_ifa, ibytes, m->m_pkthdr.len);
878 * If offset or IP_MF are set, must reassemble.
879 * Otherwise, nothing need be done.
880 * (We could look in the reassembly queue to see
881 * if the packet was previously fragmented,
882 * but it's not worth the time; just let them time out.)
884 if (ip->ip_off & (IP_MF | IP_OFFMASK)) {
886 * Attempt reassembly; if it succeeds, proceed. ip_reass()
887 * will return a different mbuf.
889 * NOTE: ip_reass() returns m with M_HASH cleared to force
890 * us to recharacterize the packet.
895 ip = mtod(m, struct ip *);
897 /* Get the header length of the reassembled packet */
898 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
905 * enforce IPsec policy checking if we are seeing last header.
906 * note that we do not visit this with protocols with pcb layer
907 * code - like udp/tcp/raw ip.
909 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) &&
910 ipsec4_in_reject(m, NULL)) {
911 ipsecstat.in_polvio++;
917 * enforce IPsec policy checking if we are seeing last header.
918 * note that we do not visit this with protocols with pcb layer
919 * code - like udp/tcp/raw ip.
921 if (inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) {
923 * Check if the packet has already had IPsec processing
924 * done. If so, then just pass it along. This tag gets
925 * set during AH, ESP, etc. input handling, before the
926 * packet is returned to the ip input queue for delivery.
928 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
931 tdbi = (struct tdb_ident *)m_tag_data(mtag);
932 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
934 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
935 IP_FORWARDING, &error);
939 * Check security policy against packet attributes.
941 error = ipsec_in_reject(sp, m);
944 /* XXX error stat??? */
946 DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/
954 #endif /* FAST_IPSEC */
957 * We must forward the packet to the correct protocol thread if
958 * we are not already in it.
960 * NOTE: ip_len is now in host form. ip_len is not adjusted
961 * further for protocol processing, instead we pass hlen
962 * to the protosw and let it deal with it.
964 ipstat.ips_delivered++;
966 if ((m->m_flags & M_HASH) == 0) {
968 atomic_add_long(&ip_rehash_count, 1);
970 ip->ip_len = htons(ip->ip_len + hlen);
971 ip->ip_off = htons(ip->ip_off);
973 ip_hashfn(&m, 0, IP_MPORT_IN);
977 ip = mtod(m, struct ip *);
978 ip->ip_len = ntohs(ip->ip_len) - hlen;
979 ip->ip_off = ntohs(ip->ip_off);
980 KKASSERT(m->m_flags & M_HASH);
982 port = netisr_hashport(m->m_pkthdr.hash);
984 if (port != &curthread->td_msgport) {
985 struct netmsg_packet *pmsg;
988 atomic_add_long(&ip_dispatch_slow, 1);
991 pmsg = &m->m_hdr.mh_netmsg;
992 netmsg_init(&pmsg->base, NULL, &netisr_apanic_rport,
993 0, transport_processing_handler);
995 pmsg->base.lmsg.u.ms_result = hlen;
996 lwkt_sendmsg(port, &pmsg->base.lmsg);
999 atomic_add_long(&ip_dispatch_fast, 1);
1001 transport_processing_oncpu(m, hlen, ip);
1010 * Take incoming datagram fragment and try to reassemble it into
1011 * whole datagram. If a chain for reassembly of this datagram already
1012 * exists, then it is given as fp; otherwise have to make a chain.
1015 ip_reass(struct mbuf *m)
1017 struct ipfrag_queue *fragq = &ipfrag_queue_pcpu[mycpuid];
1018 struct ip *ip = mtod(m, struct ip *);
1019 struct mbuf *p = NULL, *q, *nq;
1021 struct ipq *fp = NULL;
1022 struct ipqhead *head;
1023 int hlen = IP_VHL_HL(ip->ip_vhl) << 2;
1027 /* If maxnipq is 0, never accept fragments. */
1029 ipstat.ips_fragments++;
1030 ipstat.ips_fragdropped++;
1035 sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
1037 * Look for queue of fragments of this datagram.
1039 head = &fragq->ipq[sum];
1040 TAILQ_FOREACH(fp, head, ipq_list) {
1041 if (ip->ip_id == fp->ipq_id &&
1042 ip->ip_src.s_addr == fp->ipq_src.s_addr &&
1043 ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
1044 ip->ip_p == fp->ipq_p)
1051 * Enforce upper bound on number of fragmented packets
1052 * for which we attempt reassembly;
1053 * If maxnipq is -1, accept all fragments without limitation.
1055 if (fragq->nipq > maxnipq && maxnipq > 0) {
1057 * drop something from the tail of the current queue
1058 * before proceeding further
1060 struct ipq *q = TAILQ_LAST(head, ipqhead);
1063 * The current queue is empty,
1064 * so drop from one of the others.
1066 for (i = 0; i < IPREASS_NHASH; i++) {
1067 struct ipq *r = TAILQ_LAST(&fragq->ipq[i],
1070 ipstat.ips_fragtimeout += r->ipq_nfrags;
1071 ip_freef(fragq, &fragq->ipq[i], r);
1076 ipstat.ips_fragtimeout += q->ipq_nfrags;
1077 ip_freef(fragq, head, q);
1082 * Adjust ip_len to not reflect header,
1083 * convert offset of this to bytes.
1086 if (ip->ip_off & IP_MF) {
1088 * Make sure that fragments have a data length
1089 * that's a non-zero multiple of 8 bytes.
1091 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
1092 ipstat.ips_toosmall++; /* XXX */
1096 m->m_flags |= M_FRAG;
1098 m->m_flags &= ~M_FRAG;
1102 ipstat.ips_fragments++;
1103 m->m_pkthdr.header = ip;
1106 * If the hardware has not done csum over this fragment
1107 * then csum_data is not valid at all.
1109 if ((m->m_pkthdr.csum_flags & (CSUM_FRAG_NOT_CHECKED | CSUM_DATA_VALID))
1110 == (CSUM_FRAG_NOT_CHECKED | CSUM_DATA_VALID)) {
1111 m->m_pkthdr.csum_data = 0;
1112 m->m_pkthdr.csum_flags &= ~(CSUM_DATA_VALID | CSUM_PSEUDO_HDR);
1116 * Presence of header sizes in mbufs
1117 * would confuse code below.
1123 * If first fragment to arrive, create a reassembly queue.
1126 if ((fp = mpipe_alloc_nowait(&ipq_mpipe)) == NULL)
1128 TAILQ_INSERT_HEAD(head, fp, ipq_list);
1131 fp->ipq_ttl = IPFRAGTTL;
1132 fp->ipq_p = ip->ip_p;
1133 fp->ipq_id = ip->ip_id;
1134 fp->ipq_src = ip->ip_src;
1135 fp->ipq_dst = ip->ip_dst;
1137 m->m_nextpkt = NULL;
1143 #define GETIP(m) ((struct ip*)((m)->m_pkthdr.header))
1146 * Find a segment which begins after this one does.
1148 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
1149 if (GETIP(q)->ip_off > ip->ip_off)
1154 * If there is a preceding segment, it may provide some of
1155 * our data already. If so, drop the data from the incoming
1156 * segment. If it provides all of our data, drop us, otherwise
1157 * stick new segment in the proper place.
1159 * If some of the data is dropped from the the preceding
1160 * segment, then it's checksum is invalidated.
1163 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off;
1165 if (i >= ip->ip_len)
1168 m->m_pkthdr.csum_flags = 0;
1172 m->m_nextpkt = p->m_nextpkt;
1175 m->m_nextpkt = fp->ipq_frags;
1180 * While we overlap succeeding segments trim them or,
1181 * if they are completely covered, dequeue them.
1183 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off;
1185 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off;
1186 if (i < GETIP(q)->ip_len) {
1187 GETIP(q)->ip_len -= i;
1188 GETIP(q)->ip_off += i;
1190 q->m_pkthdr.csum_flags = 0;
1195 ipstat.ips_fragdropped++;
1197 q->m_nextpkt = NULL;
1203 * Check for complete reassembly and perform frag per packet
1206 * Frag limiting is performed here so that the nth frag has
1207 * a chance to complete the packet before we drop the packet.
1208 * As a result, n+1 frags are actually allowed per packet, but
1209 * only n will ever be stored. (n = maxfragsperpacket.)
1213 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
1214 if (GETIP(q)->ip_off != next) {
1215 if (fp->ipq_nfrags > maxfragsperpacket) {
1216 ipstat.ips_fragdropped += fp->ipq_nfrags;
1217 ip_freef(fragq, head, fp);
1221 next += GETIP(q)->ip_len;
1223 /* Make sure the last packet didn't have the IP_MF flag */
1224 if (p->m_flags & M_FRAG) {
1225 if (fp->ipq_nfrags > maxfragsperpacket) {
1226 ipstat.ips_fragdropped += fp->ipq_nfrags;
1227 ip_freef(fragq, head, fp);
1233 * Reassembly is complete. Make sure the packet is a sane size.
1237 if (next + (IP_VHL_HL(ip->ip_vhl) << 2) > IP_MAXPACKET) {
1238 ipstat.ips_toolong++;
1239 ipstat.ips_fragdropped += fp->ipq_nfrags;
1240 ip_freef(fragq, head, fp);
1245 * Concatenate fragments.
1252 q->m_nextpkt = NULL;
1253 for (q = nq; q != NULL; q = nq) {
1255 q->m_nextpkt = NULL;
1256 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
1257 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
1262 * Clean up the 1's complement checksum. Carry over 16 bits must
1263 * be added back. This assumes no more then 65535 packet fragments
1264 * were reassembled. A second carry can also occur (but not a third).
1266 m->m_pkthdr.csum_data = (m->m_pkthdr.csum_data & 0xffff) +
1267 (m->m_pkthdr.csum_data >> 16);
1268 if (m->m_pkthdr.csum_data > 0xFFFF)
1269 m->m_pkthdr.csum_data -= 0xFFFF;
1272 * Create header for new ip packet by
1273 * modifying header of first packet;
1274 * dequeue and discard fragment reassembly header.
1275 * Make header visible.
1278 ip->ip_src = fp->ipq_src;
1279 ip->ip_dst = fp->ipq_dst;
1280 TAILQ_REMOVE(head, fp, ipq_list);
1282 mpipe_free(&ipq_mpipe, fp);
1283 m->m_len += (IP_VHL_HL(ip->ip_vhl) << 2);
1284 m->m_data -= (IP_VHL_HL(ip->ip_vhl) << 2);
1285 /* some debugging cruft by sklower, below, will go away soon */
1286 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */
1289 for (n = m; n; n = n->m_next)
1291 m->m_pkthdr.len = plen;
1295 * Reassembly complete, return the next protocol.
1297 * Be sure to clear M_HASH to force the packet
1298 * to be re-characterized.
1300 * Clear M_FRAG, we are no longer a fragment.
1302 m->m_flags &= ~(M_HASH | M_FRAG);
1304 ipstat.ips_reassembled++;
1308 ipstat.ips_fragdropped++;
1319 * Free a fragment reassembly header and all
1320 * associated datagrams.
1323 ip_freef(struct ipfrag_queue *fragq, struct ipqhead *fhp, struct ipq *fp)
1328 * Remove first to protect against blocking
1330 TAILQ_REMOVE(fhp, fp, ipq_list);
1333 * Clean out at our leisure
1335 while (fp->ipq_frags) {
1337 fp->ipq_frags = q->m_nextpkt;
1338 q->m_nextpkt = NULL;
1341 mpipe_free(&ipq_mpipe, fp);
1346 * If a timer expires on a reassembly queue, discard it.
1349 ipfrag_timeo_dispatch(netmsg_t nmsg)
1351 struct ipfrag_queue *fragq = &ipfrag_queue_pcpu[mycpuid];
1352 struct ipq *fp, *fp_temp;
1353 struct ipqhead *head;
1357 lwkt_replymsg(&nmsg->lmsg, 0); /* reply ASAP */
1360 for (i = 0; i < IPREASS_NHASH; i++) {
1361 head = &fragq->ipq[i];
1362 TAILQ_FOREACH_MUTABLE(fp, head, ipq_list, fp_temp) {
1363 if (--fp->ipq_ttl == 0) {
1364 ipstat.ips_fragtimeout += fp->ipq_nfrags;
1365 ip_freef(fragq, head, fp);
1370 * If we are over the maximum number of fragments
1371 * (due to the limit being lowered), drain off
1372 * enough to get down to the new limit.
1374 if (maxnipq >= 0 && fragq->nipq > maxnipq) {
1375 for (i = 0; i < IPREASS_NHASH; i++) {
1376 head = &fragq->ipq[i];
1377 while (fragq->nipq > maxnipq && !TAILQ_EMPTY(head)) {
1378 ipstat.ips_fragdropped +=
1379 TAILQ_FIRST(head)->ipq_nfrags;
1380 ip_freef(fragq, head, TAILQ_FIRST(head));
1387 ipfrag_timeo_ipi(void *arg __unused)
1390 struct lwkt_msg *msg = &ipfrag_queue_pcpu[cpu].timeo_netmsg.lmsg;
1393 if (msg->ms_flags & MSGF_DONE)
1394 lwkt_sendmsg_oncpu(netisr_cpuport(cpu), msg);
1399 ipfrag_slowtimo(void)
1403 CPUMASK_ASSBMASK(mask, ncpus);
1404 CPUMASK_ANDMASK(mask, smp_active_mask);
1405 if (CPUMASK_TESTNZERO(mask))
1406 lwkt_send_ipiq_mask(mask, ipfrag_timeo_ipi, NULL);
1410 * IP timer processing
1420 * Drain off all datagram fragments.
1423 ipfrag_drain_dispatch(netmsg_t nmsg)
1425 struct ipfrag_queue *fragq = &ipfrag_queue_pcpu[mycpuid];
1426 struct ipqhead *head;
1430 lwkt_replymsg(&nmsg->lmsg, 0); /* reply ASAP */
1433 for (i = 0; i < IPREASS_NHASH; i++) {
1434 head = &fragq->ipq[i];
1435 while (!TAILQ_EMPTY(head)) {
1436 ipstat.ips_fragdropped += TAILQ_FIRST(head)->ipq_nfrags;
1437 ip_freef(fragq, head, TAILQ_FIRST(head));
1443 ipfrag_drain_ipi(void *arg __unused)
1446 struct lwkt_msg *msg = &ipfrag_queue_pcpu[cpu].drain_netmsg.lmsg;
1449 if (msg->ms_flags & MSGF_DONE)
1450 lwkt_sendmsg_oncpu(netisr_cpuport(cpu), msg);
1459 CPUMASK_ASSBMASK(mask, ncpus);
1460 CPUMASK_ANDMASK(mask, smp_active_mask);
1461 if (CPUMASK_TESTNZERO(mask))
1462 lwkt_send_ipiq_mask(mask, ipfrag_drain_ipi, NULL);
1473 * Do option processing on a datagram,
1474 * possibly discarding it if bad options are encountered,
1475 * or forwarding it if source-routed.
1476 * The pass argument is used when operating in the IPSTEALTH
1477 * mode to tell what options to process:
1478 * [LS]SRR (pass 0) or the others (pass 1).
1479 * The reason for as many as two passes is that when doing IPSTEALTH,
1480 * non-routing options should be processed only if the packet is for us.
1481 * Returns 1 if packet has been forwarded/freed,
1482 * 0 if the packet should be processed further.
1485 ip_dooptions(struct mbuf *m, int pass, struct sockaddr_in *next_hop)
1487 struct sockaddr_in ipaddr = { sizeof ipaddr, AF_INET };
1488 struct ip *ip = mtod(m, struct ip *);
1490 struct in_ifaddr *ia;
1491 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB;
1492 boolean_t forward = FALSE;
1493 struct in_addr *sin, dst;
1497 cp = (u_char *)(ip + 1);
1498 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
1499 for (; cnt > 0; cnt -= optlen, cp += optlen) {
1500 opt = cp[IPOPT_OPTVAL];
1501 if (opt == IPOPT_EOL)
1503 if (opt == IPOPT_NOP)
1506 if (cnt < IPOPT_OLEN + sizeof(*cp)) {
1507 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1510 optlen = cp[IPOPT_OLEN];
1511 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
1512 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1522 * Source routing with record.
1523 * Find interface with current destination address.
1524 * If none on this machine then drop if strictly routed,
1525 * or do nothing if loosely routed.
1526 * Record interface address and bring up next address
1527 * component. If strictly routed make sure next
1528 * address is on directly accessible net.
1532 if (ipstealth && pass > 0)
1534 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1535 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1538 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1539 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1542 ipaddr.sin_addr = ip->ip_dst;
1543 ia = (struct in_ifaddr *)
1544 ifa_ifwithaddr((struct sockaddr *)&ipaddr);
1546 if (opt == IPOPT_SSRR) {
1547 type = ICMP_UNREACH;
1548 code = ICMP_UNREACH_SRCFAIL;
1551 if (!ip_dosourceroute)
1552 goto nosourcerouting;
1554 * Loose routing, and not at next destination
1555 * yet; nothing to do except forward.
1559 off--; /* 0 origin */
1560 if (off > optlen - (int)sizeof(struct in_addr)) {
1562 * End of source route. Should be for us.
1564 if (!ip_acceptsourceroute)
1565 goto nosourcerouting;
1566 save_rte(m, cp, ip->ip_src);
1571 if (!ip_dosourceroute) {
1573 char buf[sizeof "aaa.bbb.ccc.ddd"];
1576 * Acting as a router, so generate ICMP
1579 strcpy(buf, inet_ntoa(ip->ip_dst));
1581 "attempted source route from %s to %s\n",
1582 inet_ntoa(ip->ip_src), buf);
1583 type = ICMP_UNREACH;
1584 code = ICMP_UNREACH_SRCFAIL;
1588 * Not acting as a router,
1592 ipstat.ips_cantforward++;
1599 * locate outgoing interface
1601 memcpy(&ipaddr.sin_addr, cp + off,
1602 sizeof ipaddr.sin_addr);
1604 if (opt == IPOPT_SSRR) {
1605 #define INA struct in_ifaddr *
1606 #define SA struct sockaddr *
1607 if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr))
1609 ia = (INA)ifa_ifwithnet((SA)&ipaddr);
1611 ia = ip_rtaddr(ipaddr.sin_addr, NULL);
1614 type = ICMP_UNREACH;
1615 code = ICMP_UNREACH_SRCFAIL;
1618 ip->ip_dst = ipaddr.sin_addr;
1619 memcpy(cp + off, &IA_SIN(ia)->sin_addr,
1620 sizeof(struct in_addr));
1621 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1623 * Let ip_intr's mcast routing check handle mcast pkts
1625 forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr));
1629 if (ipstealth && pass == 0)
1631 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1632 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1635 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1636 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1640 * If no space remains, ignore.
1642 off--; /* 0 origin */
1643 if (off > optlen - (int)sizeof(struct in_addr))
1645 memcpy(&ipaddr.sin_addr, &ip->ip_dst,
1646 sizeof ipaddr.sin_addr);
1648 * locate outgoing interface; if we're the destination,
1649 * use the incoming interface (should be same).
1651 if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == NULL &&
1652 (ia = ip_rtaddr(ipaddr.sin_addr, NULL)) == NULL) {
1653 type = ICMP_UNREACH;
1654 code = ICMP_UNREACH_HOST;
1657 memcpy(cp + off, &IA_SIN(ia)->sin_addr,
1658 sizeof(struct in_addr));
1659 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1663 if (ipstealth && pass == 0)
1665 code = cp - (u_char *)ip;
1666 if (optlen < 4 || optlen > 40) {
1667 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1670 if ((off = cp[IPOPT_OFFSET]) < 5) {
1671 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1674 if (off > optlen - (int)sizeof(int32_t)) {
1675 cp[IPOPT_OFFSET + 1] += (1 << 4);
1676 if ((cp[IPOPT_OFFSET + 1] & 0xf0) == 0) {
1677 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1682 off--; /* 0 origin */
1683 sin = (struct in_addr *)(cp + off);
1684 switch (cp[IPOPT_OFFSET + 1] & 0x0f) {
1686 case IPOPT_TS_TSONLY:
1689 case IPOPT_TS_TSANDADDR:
1690 if (off + sizeof(n_time) +
1691 sizeof(struct in_addr) > optlen) {
1692 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1695 ipaddr.sin_addr = dst;
1696 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr,
1700 memcpy(sin, &IA_SIN(ia)->sin_addr,
1701 sizeof(struct in_addr));
1702 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1703 off += sizeof(struct in_addr);
1706 case IPOPT_TS_PRESPEC:
1707 if (off + sizeof(n_time) +
1708 sizeof(struct in_addr) > optlen) {
1709 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1712 memcpy(&ipaddr.sin_addr, sin,
1713 sizeof(struct in_addr));
1714 if (ifa_ifwithaddr((SA)&ipaddr) == NULL)
1716 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1717 off += sizeof(struct in_addr);
1721 code = &cp[IPOPT_OFFSET + 1] - (u_char *)ip;
1725 memcpy(cp + off, &ntime, sizeof(n_time));
1726 cp[IPOPT_OFFSET] += sizeof(n_time);
1729 if (forward && ipforwarding) {
1730 ip_forward(m, TRUE, next_hop);
1735 icmp_error(m, type, code, 0, 0);
1736 ipstat.ips_badoptions++;
1741 * Given address of next destination (final or next hop),
1742 * return internet address info of interface to be used to get there.
1745 ip_rtaddr(struct in_addr dst, struct route *ro0)
1747 struct route sro, *ro;
1748 struct sockaddr_in *sin;
1749 struct in_ifaddr *ia;
1754 bzero(&sro, sizeof(sro));
1758 sin = (struct sockaddr_in *)&ro->ro_dst;
1760 if (ro->ro_rt == NULL || dst.s_addr != sin->sin_addr.s_addr) {
1761 if (ro->ro_rt != NULL) {
1765 sin->sin_family = AF_INET;
1766 sin->sin_len = sizeof *sin;
1767 sin->sin_addr = dst;
1768 rtalloc_ign(ro, RTF_PRCLONING);
1771 if (ro->ro_rt == NULL)
1774 ia = ifatoia(ro->ro_rt->rt_ifa);
1782 * Save incoming source route for use in replies,
1783 * to be picked up later by ip_srcroute if the receiver is interested.
1786 save_rte(struct mbuf *m, u_char *option, struct in_addr dst)
1789 struct ip_srcrt_opt *opt;
1792 mtag = m_tag_get(PACKET_TAG_IPSRCRT, sizeof(*opt), MB_DONTWAIT);
1795 opt = m_tag_data(mtag);
1797 olen = option[IPOPT_OLEN];
1800 kprintf("save_rte: olen %d\n", olen);
1802 if (olen > sizeof(opt->ip_srcrt) - (1 + sizeof(dst))) {
1806 bcopy(option, opt->ip_srcrt.srcopt, olen);
1807 opt->ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1808 opt->ip_srcrt.dst = dst;
1809 m_tag_prepend(m, mtag);
1813 * Retrieve incoming source route for use in replies,
1814 * in the same form used by setsockopt.
1815 * The first hop is placed before the options, will be removed later.
1818 ip_srcroute(struct mbuf *m0)
1820 struct in_addr *p, *q;
1823 struct ip_srcrt_opt *opt;
1828 mtag = m_tag_find(m0, PACKET_TAG_IPSRCRT, NULL);
1831 opt = m_tag_data(mtag);
1833 if (opt->ip_nhops == 0)
1835 m = m_get(MB_DONTWAIT, MT_HEADER);
1839 #define OPTSIZ (sizeof(opt->ip_srcrt.nop) + sizeof(opt->ip_srcrt.srcopt))
1841 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
1842 m->m_len = opt->ip_nhops * sizeof(struct in_addr) +
1843 sizeof(struct in_addr) + OPTSIZ;
1846 kprintf("ip_srcroute: nhops %d mlen %d",
1847 opt->ip_nhops, m->m_len);
1852 * First save first hop for return route
1854 p = &opt->ip_srcrt.route[opt->ip_nhops - 1];
1855 *(mtod(m, struct in_addr *)) = *p--;
1858 kprintf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr));
1862 * Copy option fields and padding (nop) to mbuf.
1864 opt->ip_srcrt.nop = IPOPT_NOP;
1865 opt->ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
1866 memcpy(mtod(m, caddr_t) + sizeof(struct in_addr), &opt->ip_srcrt.nop,
1868 q = (struct in_addr *)(mtod(m, caddr_t) +
1869 sizeof(struct in_addr) + OPTSIZ);
1872 * Record return path as an IP source route,
1873 * reversing the path (pointers are now aligned).
1875 while (p >= opt->ip_srcrt.route) {
1878 kprintf(" %x", ntohl(q->s_addr));
1883 * Last hop goes to final destination.
1885 *q = opt->ip_srcrt.dst;
1886 m_tag_delete(m0, mtag);
1889 kprintf(" %x\n", ntohl(q->s_addr));
1895 * Strip out IP options.
1898 ip_stripoptions(struct mbuf *m)
1901 struct ip *ip = mtod(m, struct ip *);
1905 optlen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
1906 opts = (caddr_t)(ip + 1);
1907 datalen = m->m_len - (sizeof(struct ip) + optlen);
1908 bcopy(opts + optlen, opts, datalen);
1910 if (m->m_flags & M_PKTHDR)
1911 m->m_pkthdr.len -= optlen;
1912 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, sizeof(struct ip) >> 2);
1915 u_char inetctlerrmap[PRC_NCMDS] = {
1917 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
1918 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
1919 EMSGSIZE, EHOSTUNREACH, 0, 0,
1921 ENOPROTOOPT, ECONNREFUSED
1925 * Forward a packet. If some error occurs return the sender
1926 * an icmp packet. Note we can't always generate a meaningful
1927 * icmp message because icmp doesn't have a large enough repertoire
1928 * of codes and types.
1930 * If not forwarding, just drop the packet. This could be confusing
1931 * if ipforwarding was zero but some routing protocol was advancing
1932 * us as a gateway to somewhere. However, we must let the routing
1933 * protocol deal with that.
1935 * The using_srcrt parameter indicates whether the packet is being forwarded
1936 * via a source route.
1939 ip_forward(struct mbuf *m, boolean_t using_srcrt, struct sockaddr_in *next_hop)
1941 struct ip *ip = mtod(m, struct ip *);
1943 struct route fwd_ro;
1944 int error, type = 0, code = 0, destmtu = 0;
1945 struct mbuf *mcopy, *mtemp = NULL;
1947 struct in_addr pkt_dst;
1951 * Cache the destination address of the packet; this may be
1952 * changed by use of 'ipfw fwd'.
1954 pkt_dst = (next_hop != NULL) ? next_hop->sin_addr : ip->ip_dst;
1958 kprintf("forward: src %x dst %x ttl %x\n",
1959 ip->ip_src.s_addr, pkt_dst.s_addr, ip->ip_ttl);
1962 if (m->m_flags & (M_BCAST | M_MCAST) || !in_canforward(pkt_dst)) {
1963 ipstat.ips_cantforward++;
1967 if (!ipstealth && ip->ip_ttl <= IPTTLDEC) {
1968 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0);
1972 bzero(&fwd_ro, sizeof(fwd_ro));
1973 ip_rtaddr(pkt_dst, &fwd_ro);
1974 if (fwd_ro.ro_rt == NULL) {
1975 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0);
1980 if (curthread->td_type == TD_TYPE_NETISR) {
1982 * Save the IP header and at most 8 bytes of the payload,
1983 * in case we need to generate an ICMP message to the src.
1985 mtemp = ipforward_mtemp[mycpuid];
1986 KASSERT((mtemp->m_flags & M_EXT) == 0 &&
1987 mtemp->m_data == mtemp->m_pktdat &&
1988 m_tag_first(mtemp) == NULL,
1989 ("ip_forward invalid mtemp1"));
1991 if (!m_dup_pkthdr(mtemp, m, MB_DONTWAIT)) {
1993 * It's probably ok if the pkthdr dup fails (because
1994 * the deep copy of the tag chain failed), but for now
1995 * be conservative and just discard the copy since
1996 * code below may some day want the tags.
2000 mtemp->m_type = m->m_type;
2001 mtemp->m_len = imin((IP_VHL_HL(ip->ip_vhl) << 2) + 8,
2003 mtemp->m_pkthdr.len = mtemp->m_len;
2004 m_copydata(m, 0, mtemp->m_len, mtod(mtemp, caddr_t));
2009 ip->ip_ttl -= IPTTLDEC;
2012 * If forwarding packet using same interface that it came in on,
2013 * perhaps should send a redirect to sender to shortcut a hop.
2014 * Only send redirect if source is sending directly to us,
2015 * and if packet was not source routed (or has any options).
2016 * Also, don't send redirect if forwarding using a default route
2017 * or a route modified by a redirect.
2019 if (rt->rt_ifp == m->m_pkthdr.rcvif &&
2020 !(rt->rt_flags & (RTF_DYNAMIC | RTF_MODIFIED)) &&
2021 satosin(rt_key(rt))->sin_addr.s_addr != INADDR_ANY &&
2022 ipsendredirects && !using_srcrt && next_hop == NULL) {
2023 u_long src = ntohl(ip->ip_src.s_addr);
2024 struct in_ifaddr *rt_ifa = (struct in_ifaddr *)rt->rt_ifa;
2026 if (rt_ifa != NULL &&
2027 (src & rt_ifa->ia_subnetmask) == rt_ifa->ia_subnet) {
2028 if (rt->rt_flags & RTF_GATEWAY)
2029 dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
2031 dest = pkt_dst.s_addr;
2033 * Router requirements says to only send
2036 type = ICMP_REDIRECT;
2037 code = ICMP_REDIRECT_HOST;
2040 kprintf("redirect (%d) to %x\n", code, dest);
2045 error = ip_output(m, NULL, &fwd_ro, IP_FORWARDING, NULL, NULL);
2047 ipstat.ips_forward++;
2050 ipflow_create(&fwd_ro, mtemp);
2053 ipstat.ips_redirectsent++;
2056 ipstat.ips_cantforward++;
2063 * Errors that do not require generating ICMP message
2068 * A router should not generate ICMP_SOURCEQUENCH as
2069 * required in RFC1812 Requirements for IP Version 4 Routers.
2070 * Source quench could be a big problem under DoS attacks,
2071 * or if the underlying interface is rate-limited.
2072 * Those who need source quench packets may re-enable them
2073 * via the net.inet.ip.sendsourcequench sysctl.
2075 if (!ip_sendsourcequench)
2079 case EACCES: /* ipfw denied packet */
2083 KASSERT((mtemp->m_flags & M_EXT) == 0 &&
2084 mtemp->m_data == mtemp->m_pktdat,
2085 ("ip_forward invalid mtemp2"));
2086 mcopy = m_copym(mtemp, 0, mtemp->m_len, MB_DONTWAIT);
2091 * Send ICMP message.
2094 case 0: /* forwarded, but need redirect */
2095 /* type, code set above */
2098 case ENETUNREACH: /* shouldn't happen, checked above */
2103 type = ICMP_UNREACH;
2104 code = ICMP_UNREACH_HOST;
2108 type = ICMP_UNREACH;
2109 code = ICMP_UNREACH_NEEDFRAG;
2112 * If the packet is routed over IPsec tunnel, tell the
2113 * originator the tunnel MTU.
2114 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
2117 if (fwd_ro.ro_rt != NULL) {
2118 struct secpolicy *sp = NULL;
2123 sp = ipsec4_getpolicybyaddr(mcopy,
2129 destmtu = fwd_ro.ro_rt->rt_ifp->if_mtu;
2131 /* count IPsec header size */
2132 ipsechdr = ipsec4_hdrsiz(mcopy,
2137 * find the correct route for outer IPv4
2138 * header, compute tunnel MTU.
2141 if (sp->req != NULL && sp->req->sav != NULL &&
2142 sp->req->sav->sah != NULL) {
2143 ro = &sp->req->sav->sah->sa_route;
2144 if (ro->ro_rt != NULL &&
2145 ro->ro_rt->rt_ifp != NULL) {
2147 ro->ro_rt->rt_ifp->if_mtu;
2148 destmtu -= ipsechdr;
2155 #elif defined(FAST_IPSEC)
2157 * If the packet is routed over IPsec tunnel, tell the
2158 * originator the tunnel MTU.
2159 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
2162 if (fwd_ro.ro_rt != NULL) {
2163 struct secpolicy *sp = NULL;
2168 sp = ipsec_getpolicybyaddr(mcopy,
2174 destmtu = fwd_ro.ro_rt->rt_ifp->if_mtu;
2176 /* count IPsec header size */
2177 ipsechdr = ipsec4_hdrsiz(mcopy,
2182 * find the correct route for outer IPv4
2183 * header, compute tunnel MTU.
2186 if (sp->req != NULL &&
2187 sp->req->sav != NULL &&
2188 sp->req->sav->sah != NULL) {
2189 ro = &sp->req->sav->sah->sa_route;
2190 if (ro->ro_rt != NULL &&
2191 ro->ro_rt->rt_ifp != NULL) {
2193 ro->ro_rt->rt_ifp->if_mtu;
2194 destmtu -= ipsechdr;
2201 #else /* !IPSEC && !FAST_IPSEC */
2202 if (fwd_ro.ro_rt != NULL)
2203 destmtu = fwd_ro.ro_rt->rt_ifp->if_mtu;
2205 ipstat.ips_cantfrag++;
2209 type = ICMP_SOURCEQUENCH;
2213 case EACCES: /* ipfw denied packet */
2214 panic("ip_forward EACCES should not reach");
2216 icmp_error(mcopy, type, code, dest, destmtu);
2219 m_tag_delete_chain(mtemp);
2220 if (fwd_ro.ro_rt != NULL)
2221 RTFREE(fwd_ro.ro_rt);
2225 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
2228 if (inp->inp_socket->so_options & SO_TIMESTAMP) {
2232 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
2233 SCM_TIMESTAMP, SOL_SOCKET);
2235 mp = &(*mp)->m_next;
2237 if (inp->inp_flags & INP_RECVDSTADDR) {
2238 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
2239 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
2241 mp = &(*mp)->m_next;
2243 if (inp->inp_flags & INP_RECVTTL) {
2244 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl,
2245 sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
2247 mp = &(*mp)->m_next;
2251 * Moving these out of udp_input() made them even more broken
2252 * than they already were.
2254 /* options were tossed already */
2255 if (inp->inp_flags & INP_RECVOPTS) {
2256 *mp = sbcreatecontrol((caddr_t) opts_deleted_above,
2257 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
2259 mp = &(*mp)->m_next;
2261 /* ip_srcroute doesn't do what we want here, need to fix */
2262 if (inp->inp_flags & INP_RECVRETOPTS) {
2263 *mp = sbcreatecontrol((caddr_t) ip_srcroute(m),
2264 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
2266 mp = &(*mp)->m_next;
2269 if (inp->inp_flags & INP_RECVIF) {
2272 struct sockaddr_dl sdl;
2275 struct sockaddr_dl *sdp;
2276 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
2278 if (((ifp = m->m_pkthdr.rcvif)) &&
2279 ((ifp->if_index != 0) && (ifp->if_index <= if_index))) {
2280 sdp = IF_LLSOCKADDR(ifp);
2282 * Change our mind and don't try copy.
2284 if ((sdp->sdl_family != AF_LINK) ||
2285 (sdp->sdl_len > sizeof(sdlbuf))) {
2288 bcopy(sdp, sdl2, sdp->sdl_len);
2292 offsetof(struct sockaddr_dl, sdl_data[0]);
2293 sdl2->sdl_family = AF_LINK;
2294 sdl2->sdl_index = 0;
2295 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
2297 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len,
2298 IP_RECVIF, IPPROTO_IP);
2300 mp = &(*mp)->m_next;
2305 * XXX these routines are called from the upper part of the kernel.
2307 * They could also be moved to ip_mroute.c, since all the RSVP
2308 * handling is done there already.
2311 ip_rsvp_init(struct socket *so)
2313 if (so->so_type != SOCK_RAW ||
2314 so->so_proto->pr_protocol != IPPROTO_RSVP)
2317 if (ip_rsvpd != NULL)
2322 * This may seem silly, but we need to be sure we don't over-increment
2323 * the RSVP counter, in case something slips up.
2338 * This may seem silly, but we need to be sure we don't over-decrement
2339 * the RSVP counter, in case something slips up.
2349 rsvp_input(struct mbuf **mp, int *offp, int proto)
2351 struct mbuf *m = *mp;
2355 if (rsvp_input_p) { /* call the real one if loaded */
2357 rsvp_input_p(mp, offp, proto);
2358 return(IPPROTO_DONE);
2361 /* Can still get packets with rsvp_on = 0 if there is a local member
2362 * of the group to which the RSVP packet is addressed. But in this
2363 * case we want to throw the packet away.
2368 return(IPPROTO_DONE);
2371 if (ip_rsvpd != NULL) {
2373 rip_input(mp, offp, proto);
2374 return(IPPROTO_DONE);
2376 /* Drop the packet */
2378 return(IPPROTO_DONE);