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 1.69 2007/08/11 18:57:34 dillon Exp $
73 #include "opt_bootp.h"
76 #include "opt_ipdivert.h"
77 #include "opt_ipfilter.h"
78 #include "opt_ipstealth.h"
79 #include "opt_ipsec.h"
81 #include <sys/param.h>
82 #include <sys/systm.h>
84 #include <sys/malloc.h>
85 #include <sys/mpipe.h>
86 #include <sys/domain.h>
87 #include <sys/protosw.h>
88 #include <sys/socket.h>
90 #include <sys/globaldata.h>
91 #include <sys/thread.h>
92 #include <sys/kernel.h>
93 #include <sys/syslog.h>
94 #include <sys/sysctl.h>
95 #include <sys/in_cksum.h>
97 #include <machine/stdarg.h>
100 #include <net/if_types.h>
101 #include <net/if_var.h>
102 #include <net/if_dl.h>
103 #include <net/pfil.h>
104 #include <net/route.h>
105 #include <net/netisr.h>
106 #include <net/intrq.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>
116 #include <sys/thread2.h>
117 #include <sys/msgport2.h>
118 #include <net/netmsg2.h>
120 #include <sys/socketvar.h>
122 #include <net/ipfw/ip_fw.h>
123 #include <net/dummynet/ip_dummynet.h>
126 #include <netinet6/ipsec.h>
127 #include <netproto/key/key.h>
131 #include <netproto/ipsec/ipsec.h>
132 #include <netproto/ipsec/key.h>
136 static int ip_rsvp_on;
137 struct socket *ip_rsvpd;
139 int ipforwarding = 0;
140 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW,
141 &ipforwarding, 0, "Enable IP forwarding between interfaces");
143 static int ipsendredirects = 1; /* XXX */
144 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW,
145 &ipsendredirects, 0, "Enable sending IP redirects");
147 int ip_defttl = IPDEFTTL;
148 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW,
149 &ip_defttl, 0, "Maximum TTL on IP packets");
151 static int ip_dosourceroute = 0;
152 SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW,
153 &ip_dosourceroute, 0, "Enable forwarding source routed IP packets");
155 static int ip_acceptsourceroute = 0;
156 SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute,
157 CTLFLAG_RW, &ip_acceptsourceroute, 0,
158 "Enable accepting source routed IP packets");
160 static int ip_keepfaith = 0;
161 SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW,
163 "Enable packet capture for FAITH IPv4->IPv6 translater daemon");
165 static int nipq = 0; /* total # of reass queues */
167 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragpackets, CTLFLAG_RW,
169 "Maximum number of IPv4 fragment reassembly queue entries");
171 static int maxfragsperpacket;
172 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW,
173 &maxfragsperpacket, 0,
174 "Maximum number of IPv4 fragments allowed per packet");
176 static int ip_sendsourcequench = 0;
177 SYSCTL_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW,
178 &ip_sendsourcequench, 0,
179 "Enable the transmission of source quench packets");
181 int ip_do_randomid = 0;
182 SYSCTL_INT(_net_inet_ip, OID_AUTO, random_id, CTLFLAG_RW,
184 "Assign random ip_id values");
186 * XXX - Setting ip_checkinterface mostly implements the receive side of
187 * the Strong ES model described in RFC 1122, but since the routing table
188 * and transmit implementation do not implement the Strong ES model,
189 * setting this to 1 results in an odd hybrid.
191 * XXX - ip_checkinterface currently must be disabled if you use ipnat
192 * to translate the destination address to another local interface.
194 * XXX - ip_checkinterface must be disabled if you add IP aliases
195 * to the loopback interface instead of the interface where the
196 * packets for those addresses are received.
198 static int ip_checkinterface = 0;
199 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW,
200 &ip_checkinterface, 0, "Verify packet arrives on correct interface");
203 static int ipprintfs = 0;
206 static struct ifqueue ipintrq;
207 static int ipqmaxlen = IFQ_MAXLEN;
209 extern struct domain inetdomain;
210 extern struct protosw inetsw[];
211 u_char ip_protox[IPPROTO_MAX];
212 struct in_ifaddrhead in_ifaddrhead; /* first inet address */
213 struct in_ifaddrhashhead *in_ifaddrhashtbl; /* inet addr hash table */
214 u_long in_ifaddrhmask; /* mask for hash table */
216 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW,
217 &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue");
218 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD,
219 &ipintrq.ifq_drops, 0, "Number of packets dropped from the IP input queue");
221 struct ip_stats ipstats_percpu[MAXCPU];
224 sysctl_ipstats(SYSCTL_HANDLER_ARGS)
228 for (cpu = 0; cpu < ncpus; ++cpu) {
229 if ((error = SYSCTL_OUT(req, &ipstats_percpu[cpu],
230 sizeof(struct ip_stats))))
232 if ((error = SYSCTL_IN(req, &ipstats_percpu[cpu],
233 sizeof(struct ip_stats))))
239 SYSCTL_PROC(_net_inet_ip, IPCTL_STATS, stats, (CTLTYPE_OPAQUE | CTLFLAG_RW),
240 0, 0, sysctl_ipstats, "S,ip_stats", "IP statistics");
242 SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW,
243 &ipstat, ip_stats, "IP statistics");
246 /* Packet reassembly stuff */
247 #define IPREASS_NHASH_LOG2 6
248 #define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2)
249 #define IPREASS_HMASK (IPREASS_NHASH - 1)
250 #define IPREASS_HASH(x,y) \
251 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK)
253 static struct ipq ipq[IPREASS_NHASH];
254 const int ipintrq_present = 1;
257 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
258 &ip_mtu, 0, "Default MTU");
262 static int ipstealth = 0;
263 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW, &ipstealth, 0, "");
265 static const int ipstealth = 0;
270 ip_fw_chk_t *ip_fw_chk_ptr;
275 ip_dn_io_t *ip_dn_io_ptr;
277 struct pfil_head inet_pfil_hook;
280 * XXX this is ugly -- the following two global variables are
281 * used to store packet state while it travels through the stack.
282 * Note that the code even makes assumptions on the size and
283 * alignment of fields inside struct ip_srcrt so e.g. adding some
284 * fields will break the code. This needs to be fixed.
286 * We need to save the IP options in case a protocol wants to respond
287 * to an incoming packet over the same route if the packet got here
288 * using IP source routing. This allows connection establishment and
289 * maintenance when the remote end is on a network that is not known
292 static int ip_nhops = 0;
294 static struct ip_srcrt {
295 struct in_addr dst; /* final destination */
296 char nop; /* one NOP to align */
297 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */
298 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)];
301 static MALLOC_DEFINE(M_IPQ, "ipq", "IP Fragment Management");
302 static struct malloc_pipe ipq_mpipe;
304 static void save_rte (u_char *, struct in_addr);
305 static int ip_dooptions (struct mbuf *m, int,
306 struct sockaddr_in *next_hop);
307 static void ip_forward (struct mbuf *m, boolean_t using_srcrt,
308 struct sockaddr_in *next_hop);
309 static void ip_freef (struct ipq *);
310 static void ip_input_handler (struct netmsg *);
311 static struct mbuf *ip_reass (struct mbuf *, struct ipq *,
312 struct ipq *, u_int32_t *);
315 * IP initialization: fill in IP protocol switch table.
316 * All protocols not implemented in kernel go to raw IP protocol handler.
328 * Make sure we can handle a reasonable number of fragments but
329 * cap it at 4000 (XXX).
331 mpipe_init(&ipq_mpipe, M_IPQ, sizeof(struct ipq),
332 IFQ_MAXLEN, 4000, 0, NULL);
333 TAILQ_INIT(&in_ifaddrhead);
334 in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask);
335 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
338 for (i = 0; i < IPPROTO_MAX; i++)
339 ip_protox[i] = pr - inetsw;
340 for (pr = inetdomain.dom_protosw;
341 pr < inetdomain.dom_protoswNPROTOSW; pr++)
342 if (pr->pr_domain->dom_family == PF_INET &&
343 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
344 ip_protox[pr->pr_protocol] = pr - inetsw;
346 inet_pfil_hook.ph_type = PFIL_TYPE_AF;
347 inet_pfil_hook.ph_af = AF_INET;
348 if ((i = pfil_head_register(&inet_pfil_hook)) != 0) {
349 kprintf("%s: WARNING: unable to register pfil hook, "
350 "error %d\n", __func__, i);
353 for (i = 0; i < IPREASS_NHASH; i++)
354 ipq[i].next = ipq[i].prev = &ipq[i];
356 maxnipq = nmbclusters / 32;
357 maxfragsperpacket = 16;
359 ip_id = time_second & 0xffff;
360 ipintrq.ifq_maxlen = ipqmaxlen;
363 * Initialize IP statistics counters for each CPU.
367 for (cpu = 0; cpu < ncpus; ++cpu) {
368 bzero(&ipstats_percpu[cpu], sizeof(struct ip_stats));
371 bzero(&ipstat, sizeof(struct ip_stats));
374 netisr_register(NETISR_IP, ip_mport, ip_input_handler);
378 * XXX watch out this one. It is perhaps used as a cache for
379 * the most recently used route ? it is cleared in in_addroute()
380 * when a new route is successfully created.
382 struct route ipforward_rt[MAXCPU];
384 /* Do transport protocol processing. */
386 transport_processing_oncpu(struct mbuf *m, int hlen, struct ip *ip,
387 struct sockaddr_in *nexthop)
390 * Switch out to protocol's input routine.
392 if (nexthop && ip->ip_p == IPPROTO_TCP) {
393 /* TCP needs IPFORWARD info if available */
396 tag.mh_type = MT_TAG;
397 tag.mh_flags = PACKET_TAG_IPFORWARD;
398 tag.mh_data = (caddr_t)nexthop;
401 (*inetsw[ip_protox[ip->ip_p]].pr_input)
402 ((struct mbuf *)&tag, hlen, ip->ip_p);
404 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen, ip->ip_p);
408 struct netmsg_transport_packet {
409 struct netmsg nm_netmsg;
410 struct mbuf *nm_mbuf;
412 boolean_t nm_hasnexthop;
413 struct sockaddr_in nm_nexthop;
417 transport_processing_handler(netmsg_t netmsg)
419 struct netmsg_transport_packet *msg = (void *)netmsg;
420 struct sockaddr_in *nexthop;
423 ip = mtod(msg->nm_mbuf, struct ip *);
424 nexthop = msg->nm_hasnexthop ? &msg->nm_nexthop : NULL;
425 transport_processing_oncpu(msg->nm_mbuf, msg->nm_hlen, ip, nexthop);
426 lwkt_replymsg(&msg->nm_netmsg.nm_lmsg, 0);
430 ip_input_handler(struct netmsg *msg0)
432 struct mbuf *m = ((struct netmsg_packet *)msg0)->nm_packet;
435 /* msg0 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)
447 struct in_ifaddr *ia = NULL;
449 int i, hlen, checkif;
451 struct in_addr pkt_dst;
452 u_int32_t divert_info = 0; /* packet divert/tee info */
453 struct ip_fw_args args;
454 boolean_t using_srcrt = FALSE; /* forward (by PFIL_HOOKS) */
455 boolean_t needredispatch = FALSE;
456 struct in_addr odst; /* original dst address(NAT) */
457 #if defined(FAST_IPSEC) || defined(IPDIVERT)
461 struct tdb_ident *tdbi;
462 struct secpolicy *sp;
469 args.next_hop = NULL;
471 /* Grab info from MT_TAG mbufs prepended to the chain. */
472 while (m != NULL && m->m_type == MT_TAG) {
473 switch(m->_m_tag_id) {
474 case PACKET_TAG_DUMMYNET:
475 args.rule = ((struct dn_pkt *)m)->rule;
477 case PACKET_TAG_IPFORWARD:
478 args.next_hop = (struct sockaddr_in *)m->m_hdr.mh_data;
481 kprintf("ip_input: unrecognised MT_TAG tag %d\n",
487 KASSERT(m != NULL && (m->m_flags & M_PKTHDR), ("ip_input: no HDR"));
489 if (args.rule != NULL) { /* dummynet already filtered us */
490 ip = mtod(m, struct ip *);
491 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
497 /* length checks already done in ip_demux() */
498 KASSERT(m->m_len >= sizeof(ip), ("IP header not in one mbuf"));
500 ip = mtod(m, struct ip *);
502 if (IP_VHL_V(ip->ip_vhl) != IPVERSION) {
503 ipstat.ips_badvers++;
507 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
508 /* length checks already done in ip_demux() */
509 KASSERT(hlen >= sizeof(struct ip), ("IP header len too small"));
510 KASSERT(m->m_len >= hlen, ("packet shorter than IP header length"));
512 /* 127/8 must not appear on wire - RFC1122 */
513 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
514 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
515 if (!(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK)) {
516 ipstat.ips_badaddr++;
521 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
522 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
524 if (hlen == sizeof(struct ip)) {
525 sum = in_cksum_hdr(ip);
527 sum = in_cksum(m, hlen);
536 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) {
537 /* packet is dropped by traffic conditioner */
542 * Convert fields to host representation.
544 ip->ip_len = ntohs(ip->ip_len);
545 if (ip->ip_len < hlen) {
549 ip->ip_off = ntohs(ip->ip_off);
552 * Check that the amount of data in the buffers
553 * is as at least much as the IP header would have us expect.
554 * Trim mbufs if longer than we expect.
555 * Drop packet if shorter than we expect.
557 if (m->m_pkthdr.len < ip->ip_len) {
558 ipstat.ips_tooshort++;
561 if (m->m_pkthdr.len > ip->ip_len) {
562 if (m->m_len == m->m_pkthdr.len) {
563 m->m_len = ip->ip_len;
564 m->m_pkthdr.len = ip->ip_len;
566 m_adj(m, ip->ip_len - m->m_pkthdr.len);
568 #if defined(IPSEC) && !defined(IPSEC_FILTERGIF)
570 * Bypass packet filtering for packets from a tunnel (gif).
572 if (ipsec_gethist(m, NULL))
578 * Right now when no processing on packet has done
579 * and it is still fresh out of network we do our black
581 * - Firewall: deny/allow/divert
582 * - Xlate: translate packet's addr/port (NAT).
583 * - Pipe: pass pkt through dummynet.
584 * - Wrap: fake packet's addr/port <unimpl.>
585 * - Encapsulate: put it in another IP and send out. <unimp.>
591 * Run through list of hooks for input packets.
593 * NB: Beware of the destination address changing (e.g.
594 * by NAT rewriting). When this happens, tell
595 * ip_forward to do the right thing.
597 if (pfil_has_hooks(&inet_pfil_hook)) {
599 if (pfil_run_hooks(&inet_pfil_hook, &m,
600 m->m_pkthdr.rcvif, PFIL_IN)) {
603 if (m == NULL) /* consumed by filter */
605 ip = mtod(m, struct ip *);
606 using_srcrt = (odst.s_addr != ip->ip_dst.s_addr);
609 if (fw_enable && IPFW_LOADED) {
611 * If we've been forwarded from the output side, then
612 * skip the firewall a second time
614 if (args.next_hop != NULL)
618 i = ip_fw_chk_ptr(&args);
621 if ((i & IP_FW_PORT_DENY_FLAG) || m == NULL) { /* drop */
626 ip = mtod(m, struct ip *); /* just in case m changed */
627 if (i == 0 && args.next_hop == NULL) /* common case */
629 if (DUMMYNET_LOADED && (i & IP_FW_PORT_DYNT_FLAG)) {
630 /* Send packet to the appropriate pipe */
631 ip_dn_io_ptr(m, i&0xffff, DN_TO_IP_IN, &args);
635 if (i != 0 && !(i & IP_FW_PORT_DYNT_FLAG)) {
636 /* Divert or tee packet */
641 if (i == 0 && args.next_hop != NULL)
644 * if we get here, the packet must be dropped
652 * Process options and, if not destined for us,
653 * ship it on. ip_dooptions returns 1 when an
654 * error was detected (causing an icmp message
655 * to be sent and the original packet to be freed).
657 ip_nhops = 0; /* for source routed packets */
658 if (hlen > sizeof(struct ip) && ip_dooptions(m, 0, args.next_hop))
661 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
662 * matter if it is destined to another node, or whether it is
663 * a multicast one, RSVP wants it! and prevents it from being forwarded
664 * anywhere else. Also checks if the rsvp daemon is running before
665 * grabbing the packet.
667 if (rsvp_on && ip->ip_p == IPPROTO_RSVP)
671 * Check our list of addresses, to see if the packet is for us.
672 * If we don't have any addresses, assume any unicast packet
673 * we receive might be for us (and let the upper layers deal
676 if (TAILQ_EMPTY(&in_ifaddrhead) && !(m->m_flags & (M_MCAST | M_BCAST)))
680 * Cache the destination address of the packet; this may be
681 * changed by use of 'ipfw fwd'.
683 pkt_dst = args.next_hop ? args.next_hop->sin_addr : ip->ip_dst;
686 * Enable a consistency check between the destination address
687 * and the arrival interface for a unicast packet (the RFC 1122
688 * strong ES model) if IP forwarding is disabled and the packet
689 * is not locally generated and the packet is not subject to
692 * XXX - Checking also should be disabled if the destination
693 * address is ipnat'ed to a different interface.
695 * XXX - Checking is incompatible with IP aliases added
696 * to the loopback interface instead of the interface where
697 * the packets are received.
699 checkif = ip_checkinterface &&
701 m->m_pkthdr.rcvif != NULL &&
702 !(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) &&
703 (args.next_hop == NULL);
706 * Check for exact addresses in the hash bucket.
708 LIST_FOREACH(ia, INADDR_HASH(pkt_dst.s_addr), ia_hash) {
710 * If the address matches, verify that the packet
711 * arrived via the correct interface if checking is
714 if (IA_SIN(ia)->sin_addr.s_addr == pkt_dst.s_addr &&
715 (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif))
719 * Check for broadcast addresses.
721 * Only accept broadcast packets that arrive via the matching
722 * interface. Reception of forwarded directed broadcasts would
723 * be handled via ip_forward() and ether_output() with the loopback
724 * into the stack for SIMPLEX interfaces handled by ether_output().
726 if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
727 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
728 if (ifa->ifa_addr == NULL) /* shutdown/startup race */
730 if (ifa->ifa_addr->sa_family != AF_INET)
733 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
736 if (ia->ia_netbroadcast.s_addr == pkt_dst.s_addr)
739 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY)
744 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
745 struct in_multi *inm;
747 if (ip_mrouter != NULL) {
749 * If we are acting as a multicast router, all
750 * incoming multicast packets are passed to the
751 * kernel-level multicast forwarding function.
752 * The packet is returned (relatively) intact; if
753 * ip_mforward() returns a non-zero value, the packet
754 * must be discarded, else it may be accepted below.
756 if (ip_mforward != NULL &&
757 ip_mforward(ip, m->m_pkthdr.rcvif, m, NULL) != 0) {
758 ipstat.ips_cantforward++;
764 * The process-level routing daemon needs to receive
765 * all multicast IGMP packets, whether or not this
766 * host belongs to their destination groups.
768 if (ip->ip_p == IPPROTO_IGMP)
770 ipstat.ips_forward++;
773 * See if we belong to the destination multicast group on the
776 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
778 ipstat.ips_notmember++;
784 if (ip->ip_dst.s_addr == INADDR_BROADCAST)
786 if (ip->ip_dst.s_addr == INADDR_ANY)
790 * FAITH(Firewall Aided Internet Translator)
792 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
794 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP)
802 * Not for us; forward if possible and desirable.
805 ipstat.ips_cantforward++;
810 * Enforce inbound IPsec SPD.
812 if (ipsec4_in_reject(m, NULL)) {
813 ipsecstat.in_polvio++;
818 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
821 tdbi = (struct tdb_ident *)m_tag_data(mtag);
822 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
824 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
825 IP_FORWARDING, &error);
827 if (sp == NULL) { /* NB: can happen if error */
829 /*XXX error stat???*/
830 DPRINTF(("ip_input: no SP for forwarding\n")); /*XXX*/
835 * Check security policy against packet attributes.
837 error = ipsec_in_reject(sp, m);
841 ipstat.ips_cantforward++;
845 ip_forward(m, using_srcrt, args.next_hop);
852 * IPSTEALTH: Process non-routing options only
853 * if the packet is destined for us.
856 hlen > sizeof(struct ip) &&
857 ip_dooptions(m, 1, args.next_hop))
860 /* Count the packet in the ip address stats */
862 ia->ia_ifa.if_ipackets++;
863 ia->ia_ifa.if_ibytes += m->m_pkthdr.len;
867 * If offset or IP_MF are set, must reassemble.
868 * Otherwise, nothing need be done.
869 * (We could look in the reassembly queue to see
870 * if the packet was previously fragmented,
871 * but it's not worth the time; just let them time out.)
873 if (ip->ip_off & (IP_MF | IP_OFFMASK)) {
875 /* If maxnipq is 0, never accept fragments. */
877 ipstat.ips_fragments++;
878 ipstat.ips_fragdropped++;
882 sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
884 * Look for queue of fragments
887 for (fp = ipq[sum].next; fp != &ipq[sum]; fp = fp->next)
888 if (ip->ip_id == fp->ipq_id &&
889 ip->ip_src.s_addr == fp->ipq_src.s_addr &&
890 ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
891 ip->ip_p == fp->ipq_p)
897 * Enforce upper bound on number of fragmented packets
898 * for which we attempt reassembly;
899 * If maxnipq is -1, accept all fragments without limitation.
901 if ((nipq > maxnipq) && (maxnipq > 0)) {
903 * drop something from the tail of the current queue
904 * before proceeding further
906 if (ipq[sum].prev == &ipq[sum]) { /* gak */
907 for (i = 0; i < IPREASS_NHASH; i++) {
908 if (ipq[i].prev != &ipq[i]) {
909 ipstat.ips_fragtimeout +=
910 ipq[i].prev->ipq_nfrags;
911 ip_freef(ipq[i].prev);
916 ipstat.ips_fragtimeout +=
917 ipq[sum].prev->ipq_nfrags;
918 ip_freef(ipq[sum].prev);
923 * Adjust ip_len to not reflect header,
924 * convert offset of this to bytes.
927 if (ip->ip_off & IP_MF) {
929 * Make sure that fragments have a data length
930 * that's a non-zero multiple of 8 bytes.
932 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
933 ipstat.ips_toosmall++; /* XXX */
936 m->m_flags |= M_FRAG;
938 m->m_flags &= ~M_FRAG;
942 * Attempt reassembly; if it succeeds, proceed.
943 * ip_reass() will return a different mbuf, and update
944 * the divert info in divert_info.
946 ipstat.ips_fragments++;
947 m->m_pkthdr.header = ip;
948 m = ip_reass(m, fp, &ipq[sum], &divert_info);
951 ipstat.ips_reassembled++;
952 needredispatch = TRUE;
953 ip = mtod(m, struct ip *);
954 /* Get the header length of the reassembled packet */
955 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
957 /* Restore original checksum before diverting packet */
958 if (divert_info != 0) {
960 ip->ip_len = htons(ip->ip_len);
961 ip->ip_off = htons(ip->ip_off);
963 if (hlen == sizeof(struct ip))
964 ip->ip_sum = in_cksum_hdr(ip);
966 ip->ip_sum = in_cksum(m, hlen);
967 ip->ip_off = ntohs(ip->ip_off);
968 ip->ip_len = ntohs(ip->ip_len);
978 * Divert or tee packet to the divert protocol if required.
980 if (divert_info != 0) {
981 struct mbuf *clone = NULL;
983 /* Clone packet if we're doing a 'tee' */
984 if ((divert_info & IP_FW_PORT_TEE_FLAG) != 0)
985 clone = m_dup(m, MB_DONTWAIT);
987 /* Restore packet header fields to original values */
989 ip->ip_len = htons(ip->ip_len);
990 ip->ip_off = htons(ip->ip_off);
992 /* Deliver packet to divert input routine */
993 divert_packet(m, 1, divert_info & 0xffff);
994 ipstat.ips_delivered++;
996 /* If 'tee', continue with original packet */
1000 ip = mtod(m, struct ip *);
1003 * Jump backwards to complete processing of the
1004 * packet. But first clear divert_info to avoid
1005 * entering this block again.
1006 * We do not need to clear args.divert_rule
1007 * or args.next_hop as they will not be used.
1009 * XXX Better safe than sorry, remove the DIVERT tag.
1011 mtag = m_tag_find(m, PACKET_TAG_IPFW_DIVERT, NULL);
1013 m_tag_delete(m, mtag);
1022 * enforce IPsec policy checking if we are seeing last header.
1023 * note that we do not visit this with protocols with pcb layer
1024 * code - like udp/tcp/raw ip.
1026 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) &&
1027 ipsec4_in_reject(m, NULL)) {
1028 ipsecstat.in_polvio++;
1034 * enforce IPsec policy checking if we are seeing last header.
1035 * note that we do not visit this with protocols with pcb layer
1036 * code - like udp/tcp/raw ip.
1038 if (inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) {
1040 * Check if the packet has already had IPsec processing
1041 * done. If so, then just pass it along. This tag gets
1042 * set during AH, ESP, etc. input handling, before the
1043 * packet is returned to the ip input queue for delivery.
1045 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
1048 tdbi = (struct tdb_ident *)m_tag_data(mtag);
1049 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
1051 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
1052 IP_FORWARDING, &error);
1056 * Check security policy against packet attributes.
1058 error = ipsec_in_reject(sp, m);
1061 /* XXX error stat??? */
1063 DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/
1070 #endif /* FAST_IPSEC */
1072 ipstat.ips_delivered++;
1073 if (needredispatch) {
1074 struct netmsg_transport_packet *msg;
1077 ip->ip_off = htons(ip->ip_off);
1078 ip->ip_len = htons(ip->ip_len);
1079 port = ip_mport(&m);
1083 msg = kmalloc(sizeof(struct netmsg_transport_packet), M_LWKTMSG,
1084 M_INTWAIT | M_NULLOK);
1088 netmsg_init(&msg->nm_netmsg, &netisr_afree_rport, 0,
1089 transport_processing_handler);
1090 msg->nm_hlen = hlen;
1091 msg->nm_hasnexthop = (args.next_hop != NULL);
1092 if (msg->nm_hasnexthop)
1093 msg->nm_nexthop = *args.next_hop; /* structure copy */
1096 ip = mtod(m, struct ip *);
1097 ip->ip_len = ntohs(ip->ip_len);
1098 ip->ip_off = ntohs(ip->ip_off);
1099 lwkt_sendmsg(port, &msg->nm_netmsg.nm_lmsg);
1101 transport_processing_oncpu(m, hlen, ip, args.next_hop);
1110 * Take incoming datagram fragment and try to reassemble it into
1111 * whole datagram. If a chain for reassembly of this datagram already
1112 * exists, then it is given as fp; otherwise have to make a chain.
1114 * When IPDIVERT enabled, keep additional state with each packet that
1115 * tells us if we need to divert or tee the packet we're building.
1116 * In particular, *divinfo includes the port and TEE flag.
1119 static struct mbuf *
1120 ip_reass(struct mbuf *m, struct ipq *fp, struct ipq *where,
1123 struct ip *ip = mtod(m, struct ip *);
1124 struct mbuf *p = NULL, *q, *nq;
1126 int hlen = IP_VHL_HL(ip->ip_vhl) << 2;
1133 * Presence of header sizes in mbufs
1134 * would confuse code below.
1140 * If first fragment to arrive, create a reassembly queue.
1143 if ((fp = mpipe_alloc_nowait(&ipq_mpipe)) == NULL)
1148 fp->ipq_ttl = IPFRAGTTL;
1149 fp->ipq_p = ip->ip_p;
1150 fp->ipq_id = ip->ip_id;
1151 fp->ipq_src = ip->ip_src;
1152 fp->ipq_dst = ip->ip_dst;
1154 m->m_nextpkt = NULL;
1156 fp->ipq_div_info = 0;
1163 #define GETIP(m) ((struct ip*)((m)->m_pkthdr.header))
1166 * Find a segment which begins after this one does.
1168 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt)
1169 if (GETIP(q)->ip_off > ip->ip_off)
1173 * If there is a preceding segment, it may provide some of
1174 * our data already. If so, drop the data from the incoming
1175 * segment. If it provides all of our data, drop us, otherwise
1176 * stick new segment in the proper place.
1178 * If some of the data is dropped from the the preceding
1179 * segment, then it's checksum is invalidated.
1182 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off;
1184 if (i >= ip->ip_len)
1187 m->m_pkthdr.csum_flags = 0;
1191 m->m_nextpkt = p->m_nextpkt;
1194 m->m_nextpkt = fp->ipq_frags;
1199 * While we overlap succeeding segments trim them or,
1200 * if they are completely covered, dequeue them.
1202 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off;
1204 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off;
1205 if (i < GETIP(q)->ip_len) {
1206 GETIP(q)->ip_len -= i;
1207 GETIP(q)->ip_off += i;
1209 q->m_pkthdr.csum_flags = 0;
1214 ipstat.ips_fragdropped++;
1216 q->m_nextpkt = NULL;
1224 * Transfer firewall instructions to the fragment structure.
1225 * Only trust info in the fragment at offset 0.
1227 if (ip->ip_off == 0) {
1228 fp->ipq_div_info = *divinfo;
1230 mtag = m_tag_find(m, PACKET_TAG_IPFW_DIVERT, NULL);
1232 m_tag_delete(m, mtag);
1238 * Check for complete reassembly and perform frag per packet
1241 * Frag limiting is performed here so that the nth frag has
1242 * a chance to complete the packet before we drop the packet.
1243 * As a result, n+1 frags are actually allowed per packet, but
1244 * only n will ever be stored. (n = maxfragsperpacket.)
1248 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
1249 if (GETIP(q)->ip_off != next) {
1250 if (fp->ipq_nfrags > maxfragsperpacket) {
1251 ipstat.ips_fragdropped += fp->ipq_nfrags;
1256 next += GETIP(q)->ip_len;
1258 /* Make sure the last packet didn't have the IP_MF flag */
1259 if (p->m_flags & M_FRAG) {
1260 if (fp->ipq_nfrags > maxfragsperpacket) {
1261 ipstat.ips_fragdropped += fp->ipq_nfrags;
1268 * Reassembly is complete. Make sure the packet is a sane size.
1272 if (next + (IP_VHL_HL(ip->ip_vhl) << 2) > IP_MAXPACKET) {
1273 ipstat.ips_toolong++;
1274 ipstat.ips_fragdropped += fp->ipq_nfrags;
1280 * Concatenate fragments.
1287 q->m_nextpkt = NULL;
1288 for (q = nq; q != NULL; q = nq) {
1290 q->m_nextpkt = NULL;
1291 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
1292 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
1297 * Clean up the 1's complement checksum. Carry over 16 bits must
1298 * be added back. This assumes no more then 65535 packet fragments
1299 * were reassembled. A second carry can also occur (but not a third).
1301 m->m_pkthdr.csum_data = (m->m_pkthdr.csum_data & 0xffff) +
1302 (m->m_pkthdr.csum_data >> 16);
1303 if (m->m_pkthdr.csum_data > 0xFFFF)
1304 m->m_pkthdr.csum_data -= 0xFFFF;
1309 * Extract firewall instructions from the fragment structure.
1311 *divinfo = fp->ipq_div_info;
1315 * Create header for new ip packet by
1316 * modifying header of first packet;
1317 * dequeue and discard fragment reassembly header.
1318 * Make header visible.
1321 ip->ip_src = fp->ipq_src;
1322 ip->ip_dst = fp->ipq_dst;
1325 mpipe_free(&ipq_mpipe, fp);
1326 m->m_len += (IP_VHL_HL(ip->ip_vhl) << 2);
1327 m->m_data -= (IP_VHL_HL(ip->ip_vhl) << 2);
1328 /* some debugging cruft by sklower, below, will go away soon */
1329 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */
1332 for (n = m; n; n = n->m_next)
1334 m->m_pkthdr.len = plen;
1342 ipstat.ips_fragdropped++;
1352 * Free a fragment reassembly header and all
1353 * associated datagrams.
1356 ip_freef(struct ipq *fp)
1360 while (fp->ipq_frags) {
1362 fp->ipq_frags = q->m_nextpkt;
1363 q->m_nextpkt = NULL;
1367 mpipe_free(&ipq_mpipe, fp);
1372 * IP timer processing;
1373 * if a timer expires on a reassembly
1374 * queue, discard it.
1383 for (i = 0; i < IPREASS_NHASH; i++) {
1387 while (fp != &ipq[i]) {
1390 if (fp->prev->ipq_ttl == 0) {
1391 ipstat.ips_fragtimeout += fp->prev->ipq_nfrags;
1397 * If we are over the maximum number of fragments
1398 * (due to the limit being lowered), drain off
1399 * enough to get down to the new limit.
1401 if (maxnipq >= 0 && nipq > maxnipq) {
1402 for (i = 0; i < IPREASS_NHASH; i++) {
1403 while (nipq > maxnipq &&
1404 (ipq[i].next != &ipq[i])) {
1405 ipstat.ips_fragdropped +=
1406 ipq[i].next->ipq_nfrags;
1407 ip_freef(ipq[i].next);
1416 * Drain off all datagram fragments.
1423 for (i = 0; i < IPREASS_NHASH; i++) {
1424 while (ipq[i].next != &ipq[i]) {
1425 ipstat.ips_fragdropped += ipq[i].next->ipq_nfrags;
1426 ip_freef(ipq[i].next);
1433 * Do option processing on a datagram,
1434 * possibly discarding it if bad options are encountered,
1435 * or forwarding it if source-routed.
1436 * The pass argument is used when operating in the IPSTEALTH
1437 * mode to tell what options to process:
1438 * [LS]SRR (pass 0) or the others (pass 1).
1439 * The reason for as many as two passes is that when doing IPSTEALTH,
1440 * non-routing options should be processed only if the packet is for us.
1441 * Returns 1 if packet has been forwarded/freed,
1442 * 0 if the packet should be processed further.
1445 ip_dooptions(struct mbuf *m, int pass, struct sockaddr_in *next_hop)
1447 struct sockaddr_in ipaddr = { sizeof ipaddr, AF_INET };
1448 struct ip *ip = mtod(m, struct ip *);
1450 struct in_ifaddr *ia;
1451 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB;
1452 boolean_t forward = FALSE;
1453 struct in_addr *sin, dst;
1457 cp = (u_char *)(ip + 1);
1458 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
1459 for (; cnt > 0; cnt -= optlen, cp += optlen) {
1460 opt = cp[IPOPT_OPTVAL];
1461 if (opt == IPOPT_EOL)
1463 if (opt == IPOPT_NOP)
1466 if (cnt < IPOPT_OLEN + sizeof(*cp)) {
1467 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1470 optlen = cp[IPOPT_OLEN];
1471 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
1472 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1482 * Source routing with record.
1483 * Find interface with current destination address.
1484 * If none on this machine then drop if strictly routed,
1485 * or do nothing if loosely routed.
1486 * Record interface address and bring up next address
1487 * component. If strictly routed make sure next
1488 * address is on directly accessible net.
1492 if (ipstealth && pass > 0)
1494 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1495 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1498 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1499 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1502 ipaddr.sin_addr = ip->ip_dst;
1503 ia = (struct in_ifaddr *)
1504 ifa_ifwithaddr((struct sockaddr *)&ipaddr);
1506 if (opt == IPOPT_SSRR) {
1507 type = ICMP_UNREACH;
1508 code = ICMP_UNREACH_SRCFAIL;
1511 if (!ip_dosourceroute)
1512 goto nosourcerouting;
1514 * Loose routing, and not at next destination
1515 * yet; nothing to do except forward.
1519 off--; /* 0 origin */
1520 if (off > optlen - (int)sizeof(struct in_addr)) {
1522 * End of source route. Should be for us.
1524 if (!ip_acceptsourceroute)
1525 goto nosourcerouting;
1526 save_rte(cp, ip->ip_src);
1531 if (!ip_dosourceroute) {
1533 char buf[sizeof "aaa.bbb.ccc.ddd"];
1536 * Acting as a router, so generate ICMP
1539 strcpy(buf, inet_ntoa(ip->ip_dst));
1541 "attempted source route from %s to %s\n",
1542 inet_ntoa(ip->ip_src), buf);
1543 type = ICMP_UNREACH;
1544 code = ICMP_UNREACH_SRCFAIL;
1548 * Not acting as a router,
1552 ipstat.ips_cantforward++;
1559 * locate outgoing interface
1561 memcpy(&ipaddr.sin_addr, cp + off,
1562 sizeof ipaddr.sin_addr);
1564 if (opt == IPOPT_SSRR) {
1565 #define INA struct in_ifaddr *
1566 #define SA struct sockaddr *
1567 if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr))
1569 ia = (INA)ifa_ifwithnet((SA)&ipaddr);
1571 ia = ip_rtaddr(ipaddr.sin_addr,
1572 &ipforward_rt[mycpuid]);
1574 type = ICMP_UNREACH;
1575 code = ICMP_UNREACH_SRCFAIL;
1578 ip->ip_dst = ipaddr.sin_addr;
1579 memcpy(cp + off, &IA_SIN(ia)->sin_addr,
1580 sizeof(struct in_addr));
1581 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1583 * Let ip_intr's mcast routing check handle mcast pkts
1585 forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr));
1589 if (ipstealth && pass == 0)
1591 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1592 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1595 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1596 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1600 * If no space remains, ignore.
1602 off--; /* 0 origin */
1603 if (off > optlen - (int)sizeof(struct in_addr))
1605 memcpy(&ipaddr.sin_addr, &ip->ip_dst,
1606 sizeof ipaddr.sin_addr);
1608 * locate outgoing interface; if we're the destination,
1609 * use the incoming interface (should be same).
1611 if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == NULL &&
1612 (ia = ip_rtaddr(ipaddr.sin_addr,
1613 &ipforward_rt[mycpuid]))
1615 type = ICMP_UNREACH;
1616 code = ICMP_UNREACH_HOST;
1619 memcpy(cp + off, &IA_SIN(ia)->sin_addr,
1620 sizeof(struct in_addr));
1621 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1625 if (ipstealth && pass == 0)
1627 code = cp - (u_char *)ip;
1628 if (optlen < 4 || optlen > 40) {
1629 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1632 if ((off = cp[IPOPT_OFFSET]) < 5) {
1633 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1636 if (off > optlen - (int)sizeof(int32_t)) {
1637 cp[IPOPT_OFFSET + 1] += (1 << 4);
1638 if ((cp[IPOPT_OFFSET + 1] & 0xf0) == 0) {
1639 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1644 off--; /* 0 origin */
1645 sin = (struct in_addr *)(cp + off);
1646 switch (cp[IPOPT_OFFSET + 1] & 0x0f) {
1648 case IPOPT_TS_TSONLY:
1651 case IPOPT_TS_TSANDADDR:
1652 if (off + sizeof(n_time) +
1653 sizeof(struct in_addr) > optlen) {
1654 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1657 ipaddr.sin_addr = dst;
1658 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr,
1662 memcpy(sin, &IA_SIN(ia)->sin_addr,
1663 sizeof(struct in_addr));
1664 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1665 off += sizeof(struct in_addr);
1668 case IPOPT_TS_PRESPEC:
1669 if (off + sizeof(n_time) +
1670 sizeof(struct in_addr) > optlen) {
1671 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1674 memcpy(&ipaddr.sin_addr, sin,
1675 sizeof(struct in_addr));
1676 if (ifa_ifwithaddr((SA)&ipaddr) == NULL)
1678 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1679 off += sizeof(struct in_addr);
1683 code = &cp[IPOPT_OFFSET + 1] - (u_char *)ip;
1687 memcpy(cp + off, &ntime, sizeof(n_time));
1688 cp[IPOPT_OFFSET] += sizeof(n_time);
1691 if (forward && ipforwarding) {
1692 ip_forward(m, TRUE, next_hop);
1697 icmp_error(m, type, code, 0, 0);
1698 ipstat.ips_badoptions++;
1703 * Given address of next destination (final or next hop),
1704 * return internet address info of interface to be used to get there.
1707 ip_rtaddr(struct in_addr dst, struct route *ro)
1709 struct sockaddr_in *sin;
1711 sin = (struct sockaddr_in *)&ro->ro_dst;
1713 if (ro->ro_rt == NULL || dst.s_addr != sin->sin_addr.s_addr) {
1714 if (ro->ro_rt != NULL) {
1718 sin->sin_family = AF_INET;
1719 sin->sin_len = sizeof *sin;
1720 sin->sin_addr = dst;
1721 rtalloc_ign(ro, RTF_PRCLONING);
1724 if (ro->ro_rt == NULL)
1727 return (ifatoia(ro->ro_rt->rt_ifa));
1731 * Save incoming source route for use in replies,
1732 * to be picked up later by ip_srcroute if the receiver is interested.
1735 save_rte(u_char *option, struct in_addr dst)
1739 olen = option[IPOPT_OLEN];
1742 kprintf("save_rte: olen %d\n", olen);
1744 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst)))
1746 bcopy(option, ip_srcrt.srcopt, olen);
1747 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1752 * Retrieve incoming source route for use in replies,
1753 * in the same form used by setsockopt.
1754 * The first hop is placed before the options, will be removed later.
1759 struct in_addr *p, *q;
1764 m = m_get(MB_DONTWAIT, MT_HEADER);
1768 #define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt))
1770 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
1771 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) +
1775 kprintf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len);
1779 * First save first hop for return route
1781 p = &ip_srcrt.route[ip_nhops - 1];
1782 *(mtod(m, struct in_addr *)) = *p--;
1785 kprintf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr));
1789 * Copy option fields and padding (nop) to mbuf.
1791 ip_srcrt.nop = IPOPT_NOP;
1792 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
1793 memcpy(mtod(m, caddr_t) + sizeof(struct in_addr), &ip_srcrt.nop,
1795 q = (struct in_addr *)(mtod(m, caddr_t) +
1796 sizeof(struct in_addr) + OPTSIZ);
1799 * Record return path as an IP source route,
1800 * reversing the path (pointers are now aligned).
1802 while (p >= ip_srcrt.route) {
1805 kprintf(" %x", ntohl(q->s_addr));
1810 * Last hop goes to final destination.
1815 kprintf(" %x\n", ntohl(q->s_addr));
1821 * Strip out IP options.
1824 ip_stripoptions(struct mbuf *m)
1827 struct ip *ip = mtod(m, struct ip *);
1831 optlen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
1832 opts = (caddr_t)(ip + 1);
1833 datalen = m->m_len - (sizeof(struct ip) + optlen);
1834 bcopy(opts + optlen, opts, datalen);
1836 if (m->m_flags & M_PKTHDR)
1837 m->m_pkthdr.len -= optlen;
1838 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, sizeof(struct ip) >> 2);
1841 u_char inetctlerrmap[PRC_NCMDS] = {
1843 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
1844 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
1845 EMSGSIZE, EHOSTUNREACH, 0, 0,
1847 ENOPROTOOPT, ECONNREFUSED
1851 * Forward a packet. If some error occurs return the sender
1852 * an icmp packet. Note we can't always generate a meaningful
1853 * icmp message because icmp doesn't have a large enough repertoire
1854 * of codes and types.
1856 * If not forwarding, just drop the packet. This could be confusing
1857 * if ipforwarding was zero but some routing protocol was advancing
1858 * us as a gateway to somewhere. However, we must let the routing
1859 * protocol deal with that.
1861 * The using_srcrt parameter indicates whether the packet is being forwarded
1862 * via a source route.
1865 ip_forward(struct mbuf *m, boolean_t using_srcrt, struct sockaddr_in *next_hop)
1867 struct ip *ip = mtod(m, struct ip *);
1868 struct sockaddr_in *ipforward_rtaddr;
1870 int error, type = 0, code = 0, destmtu = 0;
1873 struct in_addr pkt_dst;
1875 struct route *cache_rt = &ipforward_rt[mycpuid];
1879 * Cache the destination address of the packet; this may be
1880 * changed by use of 'ipfw fwd'.
1882 pkt_dst = (next_hop != NULL) ? next_hop->sin_addr : ip->ip_dst;
1886 kprintf("forward: src %x dst %x ttl %x\n",
1887 ip->ip_src.s_addr, pkt_dst.s_addr, ip->ip_ttl);
1890 if (m->m_flags & (M_BCAST | M_MCAST) || !in_canforward(pkt_dst)) {
1891 ipstat.ips_cantforward++;
1895 if (!ipstealth && ip->ip_ttl <= IPTTLDEC) {
1896 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0);
1900 ipforward_rtaddr = (struct sockaddr_in *) &cache_rt->ro_dst;
1901 if (cache_rt->ro_rt == NULL ||
1902 ipforward_rtaddr->sin_addr.s_addr != pkt_dst.s_addr) {
1903 if (cache_rt->ro_rt != NULL) {
1904 RTFREE(cache_rt->ro_rt);
1905 cache_rt->ro_rt = NULL;
1907 ipforward_rtaddr->sin_family = AF_INET;
1908 ipforward_rtaddr->sin_len = sizeof(struct sockaddr_in);
1909 ipforward_rtaddr->sin_addr = pkt_dst;
1910 rtalloc_ign(cache_rt, RTF_PRCLONING);
1911 if (cache_rt->ro_rt == NULL) {
1912 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0);
1916 rt = cache_rt->ro_rt;
1919 * Save the IP header and at most 8 bytes of the payload,
1920 * in case we need to generate an ICMP message to the src.
1922 * XXX this can be optimized a lot by saving the data in a local
1923 * buffer on the stack (72 bytes at most), and only allocating the
1924 * mbuf if really necessary. The vast majority of the packets
1925 * are forwarded without having to send an ICMP back (either
1926 * because unnecessary, or because rate limited), so we are
1927 * really we are wasting a lot of work here.
1929 * We don't use m_copy() because it might return a reference
1930 * to a shared cluster. Both this function and ip_output()
1931 * assume exclusive access to the IP header in `m', so any
1932 * data in a cluster may change before we reach icmp_error().
1934 MGETHDR(mcopy, MB_DONTWAIT, m->m_type);
1935 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, MB_DONTWAIT)) {
1937 * It's probably ok if the pkthdr dup fails (because
1938 * the deep copy of the tag chain failed), but for now
1939 * be conservative and just discard the copy since
1940 * code below may some day want the tags.
1945 if (mcopy != NULL) {
1946 mcopy->m_len = imin((IP_VHL_HL(ip->ip_vhl) << 2) + 8,
1948 mcopy->m_pkthdr.len = mcopy->m_len;
1949 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1953 ip->ip_ttl -= IPTTLDEC;
1956 * If forwarding packet using same interface that it came in on,
1957 * perhaps should send a redirect to sender to shortcut a hop.
1958 * Only send redirect if source is sending directly to us,
1959 * and if packet was not source routed (or has any options).
1960 * Also, don't send redirect if forwarding using a default route
1961 * or a route modified by a redirect.
1963 if (rt->rt_ifp == m->m_pkthdr.rcvif &&
1964 !(rt->rt_flags & (RTF_DYNAMIC | RTF_MODIFIED)) &&
1965 satosin(rt_key(rt))->sin_addr.s_addr != INADDR_ANY &&
1966 ipsendredirects && !using_srcrt && next_hop == NULL) {
1967 u_long src = ntohl(ip->ip_src.s_addr);
1968 struct in_ifaddr *rt_ifa = (struct in_ifaddr *)rt->rt_ifa;
1970 if (rt_ifa != NULL &&
1971 (src & rt_ifa->ia_subnetmask) == rt_ifa->ia_subnet) {
1972 if (rt->rt_flags & RTF_GATEWAY)
1973 dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
1975 dest = pkt_dst.s_addr;
1977 * Router requirements says to only send
1980 type = ICMP_REDIRECT;
1981 code = ICMP_REDIRECT_HOST;
1984 kprintf("redirect (%d) to %x\n", code, dest);
1989 if (next_hop != NULL) {
1990 /* Pass IPFORWARD info if available */
1991 tag.mh_type = MT_TAG;
1992 tag.mh_flags = PACKET_TAG_IPFORWARD;
1993 tag.mh_data = (caddr_t)next_hop;
1995 m = (struct mbuf *)&tag;
1998 error = ip_output(m, NULL, cache_rt, IP_FORWARDING, NULL,
2001 ipstat.ips_forward++;
2004 ipflow_create(cache_rt, mcopy);
2007 return; /* most common case */
2009 ipstat.ips_redirectsent++;
2012 ipstat.ips_cantforward++;
2019 * Send ICMP message.
2024 case 0: /* forwarded, but need redirect */
2025 /* type, code set above */
2028 case ENETUNREACH: /* shouldn't happen, checked above */
2033 type = ICMP_UNREACH;
2034 code = ICMP_UNREACH_HOST;
2038 type = ICMP_UNREACH;
2039 code = ICMP_UNREACH_NEEDFRAG;
2042 * If the packet is routed over IPsec tunnel, tell the
2043 * originator the tunnel MTU.
2044 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
2047 if (cache_rt->ro_rt != NULL) {
2048 struct secpolicy *sp = NULL;
2053 sp = ipsec4_getpolicybyaddr(mcopy,
2059 destmtu = cache_rt->ro_rt->rt_ifp->if_mtu;
2061 /* count IPsec header size */
2062 ipsechdr = ipsec4_hdrsiz(mcopy,
2067 * find the correct route for outer IPv4
2068 * header, compute tunnel MTU.
2071 if (sp->req != NULL && sp->req->sav != NULL &&
2072 sp->req->sav->sah != NULL) {
2073 ro = &sp->req->sav->sah->sa_route;
2074 if (ro->ro_rt != NULL &&
2075 ro->ro_rt->rt_ifp != NULL) {
2077 ro->ro_rt->rt_ifp->if_mtu;
2078 destmtu -= ipsechdr;
2087 * If the packet is routed over IPsec tunnel, tell the
2088 * originator the tunnel MTU.
2089 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
2092 if (cache_rt->ro_rt != NULL) {
2093 struct secpolicy *sp = NULL;
2098 sp = ipsec_getpolicybyaddr(mcopy,
2104 destmtu = cache_rt->ro_rt->rt_ifp->if_mtu;
2106 /* count IPsec header size */
2107 ipsechdr = ipsec4_hdrsiz(mcopy,
2112 * find the correct route for outer IPv4
2113 * header, compute tunnel MTU.
2116 if (sp->req != NULL &&
2117 sp->req->sav != NULL &&
2118 sp->req->sav->sah != NULL) {
2119 ro = &sp->req->sav->sah->sa_route;
2120 if (ro->ro_rt != NULL &&
2121 ro->ro_rt->rt_ifp != NULL) {
2123 ro->ro_rt->rt_ifp->if_mtu;
2124 destmtu -= ipsechdr;
2131 #else /* !IPSEC && !FAST_IPSEC */
2132 if (cache_rt->ro_rt != NULL)
2133 destmtu = cache_rt->ro_rt->rt_ifp->if_mtu;
2135 ipstat.ips_cantfrag++;
2140 * A router should not generate ICMP_SOURCEQUENCH as
2141 * required in RFC1812 Requirements for IP Version 4 Routers.
2142 * Source quench could be a big problem under DoS attacks,
2143 * or if the underlying interface is rate-limited.
2144 * Those who need source quench packets may re-enable them
2145 * via the net.inet.ip.sendsourcequench sysctl.
2147 if (!ip_sendsourcequench) {
2151 type = ICMP_SOURCEQUENCH;
2156 case EACCES: /* ipfw denied packet */
2160 icmp_error(mcopy, type, code, dest, destmtu);
2164 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
2167 if (inp->inp_socket->so_options & SO_TIMESTAMP) {
2171 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
2172 SCM_TIMESTAMP, SOL_SOCKET);
2174 mp = &(*mp)->m_next;
2176 if (inp->inp_flags & INP_RECVDSTADDR) {
2177 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
2178 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
2180 mp = &(*mp)->m_next;
2182 if (inp->inp_flags & INP_RECVTTL) {
2183 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl,
2184 sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
2186 mp = &(*mp)->m_next;
2190 * Moving these out of udp_input() made them even more broken
2191 * than they already were.
2193 /* options were tossed already */
2194 if (inp->inp_flags & INP_RECVOPTS) {
2195 *mp = sbcreatecontrol((caddr_t) opts_deleted_above,
2196 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
2198 mp = &(*mp)->m_next;
2200 /* ip_srcroute doesn't do what we want here, need to fix */
2201 if (inp->inp_flags & INP_RECVRETOPTS) {
2202 *mp = sbcreatecontrol((caddr_t) ip_srcroute(),
2203 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
2205 mp = &(*mp)->m_next;
2208 if (inp->inp_flags & INP_RECVIF) {
2211 struct sockaddr_dl sdl;
2214 struct sockaddr_dl *sdp;
2215 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
2217 if (((ifp = m->m_pkthdr.rcvif)) &&
2218 ((ifp->if_index != 0) && (ifp->if_index <= if_index))) {
2219 sdp = IF_LLSOCKADDR(ifp);
2221 * Change our mind and don't try copy.
2223 if ((sdp->sdl_family != AF_LINK) ||
2224 (sdp->sdl_len > sizeof(sdlbuf))) {
2227 bcopy(sdp, sdl2, sdp->sdl_len);
2231 offsetof(struct sockaddr_dl, sdl_data[0]);
2232 sdl2->sdl_family = AF_LINK;
2233 sdl2->sdl_index = 0;
2234 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
2236 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len,
2237 IP_RECVIF, IPPROTO_IP);
2239 mp = &(*mp)->m_next;
2244 * XXX these routines are called from the upper part of the kernel.
2246 * They could also be moved to ip_mroute.c, since all the RSVP
2247 * handling is done there already.
2250 ip_rsvp_init(struct socket *so)
2252 if (so->so_type != SOCK_RAW ||
2253 so->so_proto->pr_protocol != IPPROTO_RSVP)
2256 if (ip_rsvpd != NULL)
2261 * This may seem silly, but we need to be sure we don't over-increment
2262 * the RSVP counter, in case something slips up.
2277 * This may seem silly, but we need to be sure we don't over-decrement
2278 * the RSVP counter, in case something slips up.
2288 rsvp_input(struct mbuf *m, ...) /* XXX must fixup manually */
2294 off = __va_arg(ap, int);
2295 proto = __va_arg(ap, int);
2298 if (rsvp_input_p) { /* call the real one if loaded */
2299 rsvp_input_p(m, off, proto);
2303 /* Can still get packets with rsvp_on = 0 if there is a local member
2304 * of the group to which the RSVP packet is addressed. But in this
2305 * case we want to throw the packet away.
2313 if (ip_rsvpd != NULL) {
2314 rip_input(m, off, proto);
2317 /* Drop the packet */