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 1.72 2007/11/16 02:45:45 sephe Exp $
73 #include "opt_bootp.h"
76 #include "opt_ipdivert.h"
77 #include "opt_ipfilter.h"
78 #include "opt_ipstealth.h"
79 #include "opt_ipsec.h"
81 #include <sys/param.h>
82 #include <sys/systm.h>
84 #include <sys/malloc.h>
85 #include <sys/mpipe.h>
86 #include <sys/domain.h>
87 #include <sys/protosw.h>
88 #include <sys/socket.h>
90 #include <sys/globaldata.h>
91 #include <sys/thread.h>
92 #include <sys/kernel.h>
93 #include <sys/syslog.h>
94 #include <sys/sysctl.h>
95 #include <sys/in_cksum.h>
97 #include <machine/stdarg.h>
100 #include <net/if_types.h>
101 #include <net/if_var.h>
102 #include <net/if_dl.h>
103 #include <net/pfil.h>
104 #include <net/route.h>
105 #include <net/netisr.h>
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;
271 ip_fw_dn_io_t *ip_fw_dn_io_ptr;
275 struct pfil_head inet_pfil_hook;
278 * XXX this is ugly -- the following two global variables are
279 * used to store packet state while it travels through the stack.
280 * Note that the code even makes assumptions on the size and
281 * alignment of fields inside struct ip_srcrt so e.g. adding some
282 * fields will break the code. This needs to be fixed.
284 * We need to save the IP options in case a protocol wants to respond
285 * to an incoming packet over the same route if the packet got here
286 * using IP source routing. This allows connection establishment and
287 * maintenance when the remote end is on a network that is not known
290 static int ip_nhops = 0;
292 static struct ip_srcrt {
293 struct in_addr dst; /* final destination */
294 char nop; /* one NOP to align */
295 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */
296 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)];
299 static MALLOC_DEFINE(M_IPQ, "ipq", "IP Fragment Management");
300 static struct malloc_pipe ipq_mpipe;
302 static void save_rte (u_char *, struct in_addr);
303 static int ip_dooptions (struct mbuf *m, int,
304 struct sockaddr_in *next_hop);
305 static void ip_forward (struct mbuf *m, boolean_t using_srcrt,
306 struct sockaddr_in *next_hop);
307 static void ip_freef (struct ipq *);
308 static void ip_input_handler (struct netmsg *);
309 static struct mbuf *ip_reass (struct mbuf *, struct ipq *,
310 struct ipq *, u_int32_t *);
313 * IP initialization: fill in IP protocol switch table.
314 * All protocols not implemented in kernel go to raw IP protocol handler.
326 * Make sure we can handle a reasonable number of fragments but
327 * cap it at 4000 (XXX).
329 mpipe_init(&ipq_mpipe, M_IPQ, sizeof(struct ipq),
330 IFQ_MAXLEN, 4000, 0, NULL);
331 TAILQ_INIT(&in_ifaddrhead);
332 in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask);
333 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
336 for (i = 0; i < IPPROTO_MAX; i++)
337 ip_protox[i] = pr - inetsw;
338 for (pr = inetdomain.dom_protosw;
339 pr < inetdomain.dom_protoswNPROTOSW; pr++)
340 if (pr->pr_domain->dom_family == PF_INET &&
341 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
342 ip_protox[pr->pr_protocol] = pr - inetsw;
344 inet_pfil_hook.ph_type = PFIL_TYPE_AF;
345 inet_pfil_hook.ph_af = AF_INET;
346 if ((i = pfil_head_register(&inet_pfil_hook)) != 0) {
347 kprintf("%s: WARNING: unable to register pfil hook, "
348 "error %d\n", __func__, i);
351 for (i = 0; i < IPREASS_NHASH; i++)
352 ipq[i].next = ipq[i].prev = &ipq[i];
354 maxnipq = nmbclusters / 32;
355 maxfragsperpacket = 16;
357 ip_id = time_second & 0xffff;
358 ipintrq.ifq_maxlen = ipqmaxlen;
361 * Initialize IP statistics counters for each CPU.
365 for (cpu = 0; cpu < ncpus; ++cpu) {
366 bzero(&ipstats_percpu[cpu], sizeof(struct ip_stats));
369 bzero(&ipstat, sizeof(struct ip_stats));
372 netisr_register(NETISR_IP, ip_mport, ip_input_handler);
376 * XXX watch out this one. It is perhaps used as a cache for
377 * the most recently used route ? it is cleared in in_addroute()
378 * when a new route is successfully created.
380 struct route ipforward_rt[MAXCPU];
382 /* Do transport protocol processing. */
384 transport_processing_oncpu(struct mbuf *m, int hlen, struct ip *ip,
385 struct sockaddr_in *nexthop)
388 * Switch out to protocol's input routine.
390 if (nexthop && ip->ip_p == IPPROTO_TCP) {
391 /* TCP needs IPFORWARD info if available */
394 tag.mh_type = MT_TAG;
395 tag.mh_flags = PACKET_TAG_IPFORWARD;
396 tag.mh_data = (caddr_t)nexthop;
399 (*inetsw[ip_protox[ip->ip_p]].pr_input)
400 ((struct mbuf *)&tag, hlen, ip->ip_p);
402 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen, ip->ip_p);
406 struct netmsg_transport_packet {
407 struct netmsg nm_netmsg;
408 struct mbuf *nm_mbuf;
410 boolean_t nm_hasnexthop;
411 struct sockaddr_in nm_nexthop;
415 transport_processing_handler(netmsg_t netmsg)
417 struct netmsg_transport_packet *msg = (void *)netmsg;
418 struct sockaddr_in *nexthop;
421 ip = mtod(msg->nm_mbuf, struct ip *);
422 nexthop = msg->nm_hasnexthop ? &msg->nm_nexthop : NULL;
423 transport_processing_oncpu(msg->nm_mbuf, msg->nm_hlen, ip, nexthop);
424 lwkt_replymsg(&msg->nm_netmsg.nm_lmsg, 0);
428 ip_input_handler(struct netmsg *msg0)
430 struct mbuf *m = ((struct netmsg_packet *)msg0)->nm_packet;
433 /* msg0 was embedded in the mbuf, do not reply! */
437 * IP input routine. Checksum and byte swap header. If fragmented
438 * try to reassemble. Process options. Pass to next level.
441 ip_input(struct mbuf *m)
445 struct in_ifaddr *ia = NULL;
447 int i, hlen, checkif;
449 struct in_addr pkt_dst;
450 u_int32_t divert_info = 0; /* packet divert/tee info */
451 struct ip_fw_args args;
452 boolean_t using_srcrt = FALSE; /* forward (by PFIL_HOOKS) */
453 boolean_t needredispatch = FALSE;
454 struct in_addr odst; /* original dst address(NAT) */
457 struct tdb_ident *tdbi;
458 struct secpolicy *sp;
465 args.next_hop = NULL;
467 /* Grab info from MT_TAG mbufs prepended to the chain. */
468 while (m != NULL && m->m_type == MT_TAG) {
469 switch(m->_m_tag_id) {
470 case PACKET_TAG_IPFORWARD:
471 args.next_hop = (struct sockaddr_in *)m->m_hdr.mh_data;
474 kprintf("ip_input: unrecognised MT_TAG tag %d\n",
480 KASSERT(m != NULL && (m->m_flags & M_PKTHDR), ("ip_input: no HDR"));
482 /* Extract info from dummynet tag */
483 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
485 args.rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv;
487 m_tag_delete(m, mtag);
491 if (args.rule != NULL) { /* dummynet already filtered us */
492 ip = mtod(m, struct ip *);
493 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
499 /* length checks already done in ip_demux() */
500 KASSERT(m->m_len >= sizeof(ip), ("IP header not in one mbuf"));
502 ip = mtod(m, struct ip *);
504 if (IP_VHL_V(ip->ip_vhl) != IPVERSION) {
505 ipstat.ips_badvers++;
509 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
510 /* length checks already done in ip_demux() */
511 KASSERT(hlen >= sizeof(struct ip), ("IP header len too small"));
512 KASSERT(m->m_len >= hlen, ("packet shorter than IP header length"));
514 /* 127/8 must not appear on wire - RFC1122 */
515 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
516 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
517 if (!(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK)) {
518 ipstat.ips_badaddr++;
523 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
524 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
526 if (hlen == sizeof(struct ip)) {
527 sum = in_cksum_hdr(ip);
529 sum = in_cksum(m, hlen);
538 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) {
539 /* packet is dropped by traffic conditioner */
544 * Convert fields to host representation.
546 ip->ip_len = ntohs(ip->ip_len);
547 if (ip->ip_len < hlen) {
551 ip->ip_off = ntohs(ip->ip_off);
554 * Check that the amount of data in the buffers
555 * is as at least much as the IP header would have us expect.
556 * Trim mbufs if longer than we expect.
557 * Drop packet if shorter than we expect.
559 if (m->m_pkthdr.len < ip->ip_len) {
560 ipstat.ips_tooshort++;
563 if (m->m_pkthdr.len > ip->ip_len) {
564 if (m->m_len == m->m_pkthdr.len) {
565 m->m_len = ip->ip_len;
566 m->m_pkthdr.len = ip->ip_len;
568 m_adj(m, ip->ip_len - m->m_pkthdr.len);
570 #if defined(IPSEC) && !defined(IPSEC_FILTERGIF)
572 * Bypass packet filtering for packets from a tunnel (gif).
574 if (ipsec_gethist(m, NULL))
580 * Right now when no processing on packet has done
581 * and it is still fresh out of network we do our black
583 * - Firewall: deny/allow/divert
584 * - Xlate: translate packet's addr/port (NAT).
585 * - Pipe: pass pkt through dummynet.
586 * - Wrap: fake packet's addr/port <unimpl.>
587 * - Encapsulate: put it in another IP and send out. <unimp.>
593 * Run through list of hooks for input packets.
595 * NB: Beware of the destination address changing (e.g.
596 * by NAT rewriting). When this happens, tell
597 * ip_forward to do the right thing.
599 if (pfil_has_hooks(&inet_pfil_hook)) {
601 if (pfil_run_hooks(&inet_pfil_hook, &m,
602 m->m_pkthdr.rcvif, PFIL_IN)) {
605 if (m == NULL) /* consumed by filter */
607 ip = mtod(m, struct ip *);
608 using_srcrt = (odst.s_addr != ip->ip_dst.s_addr);
611 if (fw_enable && IPFW_LOADED) {
613 * If we've been forwarded from the output side, then
614 * skip the firewall a second time
616 if (args.next_hop != NULL)
620 i = ip_fw_chk_ptr(&args);
623 if ((i & IP_FW_PORT_DENY_FLAG) || m == NULL) { /* drop */
628 ip = mtod(m, struct ip *); /* just in case m changed */
629 if (i == 0 && args.next_hop == NULL) /* common case */
631 if (i & IP_FW_PORT_DYNT_FLAG) {
632 /* Send packet to the appropriate pipe */
633 ip_fw_dn_io_ptr(m, i&0xffff, DN_TO_IP_IN, &args);
637 if (i != 0 && !(i & IP_FW_PORT_DYNT_FLAG)) {
638 /* Divert or tee packet */
643 if (i == 0 && args.next_hop != NULL)
646 * if we get here, the packet must be dropped
654 * Process options and, if not destined for us,
655 * ship it on. ip_dooptions returns 1 when an
656 * error was detected (causing an icmp message
657 * to be sent and the original packet to be freed).
659 ip_nhops = 0; /* for source routed packets */
660 if (hlen > sizeof(struct ip) && ip_dooptions(m, 0, args.next_hop))
663 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
664 * matter if it is destined to another node, or whether it is
665 * a multicast one, RSVP wants it! and prevents it from being forwarded
666 * anywhere else. Also checks if the rsvp daemon is running before
667 * grabbing the packet.
669 if (rsvp_on && ip->ip_p == IPPROTO_RSVP)
673 * Check our list of addresses, to see if the packet is for us.
674 * If we don't have any addresses, assume any unicast packet
675 * we receive might be for us (and let the upper layers deal
678 if (TAILQ_EMPTY(&in_ifaddrhead) && !(m->m_flags & (M_MCAST | M_BCAST)))
682 * Cache the destination address of the packet; this may be
683 * changed by use of 'ipfw fwd'.
685 pkt_dst = args.next_hop ? args.next_hop->sin_addr : ip->ip_dst;
688 * Enable a consistency check between the destination address
689 * and the arrival interface for a unicast packet (the RFC 1122
690 * strong ES model) if IP forwarding is disabled and the packet
691 * is not locally generated and the packet is not subject to
694 * XXX - Checking also should be disabled if the destination
695 * address is ipnat'ed to a different interface.
697 * XXX - Checking is incompatible with IP aliases added
698 * to the loopback interface instead of the interface where
699 * the packets are received.
701 checkif = ip_checkinterface &&
703 m->m_pkthdr.rcvif != NULL &&
704 !(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) &&
705 (args.next_hop == NULL);
708 * Check for exact addresses in the hash bucket.
710 LIST_FOREACH(ia, INADDR_HASH(pkt_dst.s_addr), ia_hash) {
712 * If the address matches, verify that the packet
713 * arrived via the correct interface if checking is
716 if (IA_SIN(ia)->sin_addr.s_addr == pkt_dst.s_addr &&
717 (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif))
721 * Check for broadcast addresses.
723 * Only accept broadcast packets that arrive via the matching
724 * interface. Reception of forwarded directed broadcasts would
725 * be handled via ip_forward() and ether_output() with the loopback
726 * into the stack for SIMPLEX interfaces handled by ether_output().
728 if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
729 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
730 if (ifa->ifa_addr == NULL) /* shutdown/startup race */
732 if (ifa->ifa_addr->sa_family != AF_INET)
735 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
738 if (ia->ia_netbroadcast.s_addr == pkt_dst.s_addr)
741 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY)
746 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
747 struct in_multi *inm;
749 if (ip_mrouter != NULL) {
751 * If we are acting as a multicast router, all
752 * incoming multicast packets are passed to the
753 * kernel-level multicast forwarding function.
754 * The packet is returned (relatively) intact; if
755 * ip_mforward() returns a non-zero value, the packet
756 * must be discarded, else it may be accepted below.
758 if (ip_mforward != NULL &&
759 ip_mforward(ip, m->m_pkthdr.rcvif, m, NULL) != 0) {
760 ipstat.ips_cantforward++;
766 * The process-level routing daemon needs to receive
767 * all multicast IGMP packets, whether or not this
768 * host belongs to their destination groups.
770 if (ip->ip_p == IPPROTO_IGMP)
772 ipstat.ips_forward++;
775 * See if we belong to the destination multicast group on the
778 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
780 ipstat.ips_notmember++;
786 if (ip->ip_dst.s_addr == INADDR_BROADCAST)
788 if (ip->ip_dst.s_addr == INADDR_ANY)
792 * FAITH(Firewall Aided Internet Translator)
794 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
796 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP)
804 * Not for us; forward if possible and desirable.
807 ipstat.ips_cantforward++;
812 * Enforce inbound IPsec SPD.
814 if (ipsec4_in_reject(m, NULL)) {
815 ipsecstat.in_polvio++;
820 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
823 tdbi = (struct tdb_ident *)m_tag_data(mtag);
824 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
826 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
827 IP_FORWARDING, &error);
829 if (sp == NULL) { /* NB: can happen if error */
831 /*XXX error stat???*/
832 DPRINTF(("ip_input: no SP for forwarding\n")); /*XXX*/
837 * Check security policy against packet attributes.
839 error = ipsec_in_reject(sp, m);
843 ipstat.ips_cantforward++;
847 ip_forward(m, using_srcrt, args.next_hop);
854 * IPSTEALTH: Process non-routing options only
855 * if the packet is destined for us.
858 hlen > sizeof(struct ip) &&
859 ip_dooptions(m, 1, args.next_hop))
862 /* Count the packet in the ip address stats */
864 ia->ia_ifa.if_ipackets++;
865 ia->ia_ifa.if_ibytes += m->m_pkthdr.len;
869 * If offset or IP_MF are set, must reassemble.
870 * Otherwise, nothing need be done.
871 * (We could look in the reassembly queue to see
872 * if the packet was previously fragmented,
873 * but it's not worth the time; just let them time out.)
875 if (ip->ip_off & (IP_MF | IP_OFFMASK)) {
877 /* If maxnipq is 0, never accept fragments. */
879 ipstat.ips_fragments++;
880 ipstat.ips_fragdropped++;
884 sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
886 * Look for queue of fragments
889 for (fp = ipq[sum].next; fp != &ipq[sum]; fp = fp->next)
890 if (ip->ip_id == fp->ipq_id &&
891 ip->ip_src.s_addr == fp->ipq_src.s_addr &&
892 ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
893 ip->ip_p == fp->ipq_p)
899 * Enforce upper bound on number of fragmented packets
900 * for which we attempt reassembly;
901 * If maxnipq is -1, accept all fragments without limitation.
903 if ((nipq > maxnipq) && (maxnipq > 0)) {
905 * drop something from the tail of the current queue
906 * before proceeding further
908 if (ipq[sum].prev == &ipq[sum]) { /* gak */
909 for (i = 0; i < IPREASS_NHASH; i++) {
910 if (ipq[i].prev != &ipq[i]) {
911 ipstat.ips_fragtimeout +=
912 ipq[i].prev->ipq_nfrags;
913 ip_freef(ipq[i].prev);
918 ipstat.ips_fragtimeout +=
919 ipq[sum].prev->ipq_nfrags;
920 ip_freef(ipq[sum].prev);
925 * Adjust ip_len to not reflect header,
926 * convert offset of this to bytes.
929 if (ip->ip_off & IP_MF) {
931 * Make sure that fragments have a data length
932 * that's a non-zero multiple of 8 bytes.
934 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
935 ipstat.ips_toosmall++; /* XXX */
938 m->m_flags |= M_FRAG;
940 m->m_flags &= ~M_FRAG;
944 * Attempt reassembly; if it succeeds, proceed.
945 * ip_reass() will return a different mbuf, and update
946 * the divert info in divert_info.
948 ipstat.ips_fragments++;
949 m->m_pkthdr.header = ip;
950 m = ip_reass(m, fp, &ipq[sum], &divert_info);
953 ipstat.ips_reassembled++;
954 needredispatch = TRUE;
955 ip = mtod(m, struct ip *);
956 /* Get the header length of the reassembled packet */
957 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
959 /* Restore original checksum before diverting packet */
960 if (divert_info != 0) {
962 ip->ip_len = htons(ip->ip_len);
963 ip->ip_off = htons(ip->ip_off);
965 if (hlen == sizeof(struct ip))
966 ip->ip_sum = in_cksum_hdr(ip);
968 ip->ip_sum = in_cksum(m, hlen);
969 ip->ip_off = ntohs(ip->ip_off);
970 ip->ip_len = ntohs(ip->ip_len);
980 * Divert or tee packet to the divert protocol if required.
982 if (divert_info != 0) {
983 struct mbuf *clone = NULL;
985 /* Clone packet if we're doing a 'tee' */
986 if ((divert_info & IP_FW_PORT_TEE_FLAG) != 0)
987 clone = m_dup(m, MB_DONTWAIT);
989 /* Restore packet header fields to original values */
991 ip->ip_len = htons(ip->ip_len);
992 ip->ip_off = htons(ip->ip_off);
994 /* Deliver packet to divert input routine */
995 divert_packet(m, 1, divert_info & 0xffff);
996 ipstat.ips_delivered++;
998 /* If 'tee', continue with original packet */
1002 ip = mtod(m, struct ip *);
1005 * Jump backwards to complete processing of the
1006 * packet. But first clear divert_info to avoid
1007 * entering this block again.
1008 * We do not need to clear args.divert_rule
1009 * or args.next_hop as they will not be used.
1011 * XXX Better safe than sorry, remove the DIVERT tag.
1013 mtag = m_tag_find(m, PACKET_TAG_IPFW_DIVERT, NULL);
1015 m_tag_delete(m, mtag);
1024 * enforce IPsec policy checking if we are seeing last header.
1025 * note that we do not visit this with protocols with pcb layer
1026 * code - like udp/tcp/raw ip.
1028 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) &&
1029 ipsec4_in_reject(m, NULL)) {
1030 ipsecstat.in_polvio++;
1036 * enforce IPsec policy checking if we are seeing last header.
1037 * note that we do not visit this with protocols with pcb layer
1038 * code - like udp/tcp/raw ip.
1040 if (inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) {
1042 * Check if the packet has already had IPsec processing
1043 * done. If so, then just pass it along. This tag gets
1044 * set during AH, ESP, etc. input handling, before the
1045 * packet is returned to the ip input queue for delivery.
1047 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
1050 tdbi = (struct tdb_ident *)m_tag_data(mtag);
1051 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
1053 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
1054 IP_FORWARDING, &error);
1058 * Check security policy against packet attributes.
1060 error = ipsec_in_reject(sp, m);
1063 /* XXX error stat??? */
1065 DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/
1072 #endif /* FAST_IPSEC */
1074 ipstat.ips_delivered++;
1075 if (needredispatch) {
1076 struct netmsg_transport_packet *msg;
1079 ip->ip_off = htons(ip->ip_off);
1080 ip->ip_len = htons(ip->ip_len);
1081 port = ip_mport(&m);
1085 msg = kmalloc(sizeof(struct netmsg_transport_packet), M_LWKTMSG,
1086 M_INTWAIT | M_NULLOK);
1090 netmsg_init(&msg->nm_netmsg, &netisr_afree_rport, 0,
1091 transport_processing_handler);
1092 msg->nm_hlen = hlen;
1093 msg->nm_hasnexthop = (args.next_hop != NULL);
1094 if (msg->nm_hasnexthop)
1095 msg->nm_nexthop = *args.next_hop; /* structure copy */
1098 ip = mtod(m, struct ip *);
1099 ip->ip_len = ntohs(ip->ip_len);
1100 ip->ip_off = ntohs(ip->ip_off);
1101 lwkt_sendmsg(port, &msg->nm_netmsg.nm_lmsg);
1103 transport_processing_oncpu(m, hlen, ip, args.next_hop);
1112 * Take incoming datagram fragment and try to reassemble it into
1113 * whole datagram. If a chain for reassembly of this datagram already
1114 * exists, then it is given as fp; otherwise have to make a chain.
1116 * When IPDIVERT enabled, keep additional state with each packet that
1117 * tells us if we need to divert or tee the packet we're building.
1118 * In particular, *divinfo includes the port and TEE flag.
1121 static struct mbuf *
1122 ip_reass(struct mbuf *m, struct ipq *fp, struct ipq *where,
1125 struct ip *ip = mtod(m, struct ip *);
1126 struct mbuf *p = NULL, *q, *nq;
1128 int hlen = IP_VHL_HL(ip->ip_vhl) << 2;
1135 * If the hardware has not done csum over this fragment
1136 * then csum_data is not valid at all.
1138 if ((m->m_pkthdr.csum_flags & (CSUM_FRAG_NOT_CHECKED | CSUM_DATA_VALID))
1139 == (CSUM_FRAG_NOT_CHECKED | CSUM_DATA_VALID)) {
1140 m->m_pkthdr.csum_data = 0;
1141 m->m_pkthdr.csum_flags &= ~(CSUM_DATA_VALID | CSUM_PSEUDO_HDR);
1145 * Presence of header sizes in mbufs
1146 * would confuse code below.
1152 * If first fragment to arrive, create a reassembly queue.
1155 if ((fp = mpipe_alloc_nowait(&ipq_mpipe)) == NULL)
1160 fp->ipq_ttl = IPFRAGTTL;
1161 fp->ipq_p = ip->ip_p;
1162 fp->ipq_id = ip->ip_id;
1163 fp->ipq_src = ip->ip_src;
1164 fp->ipq_dst = ip->ip_dst;
1166 m->m_nextpkt = NULL;
1168 fp->ipq_div_info = 0;
1175 #define GETIP(m) ((struct ip*)((m)->m_pkthdr.header))
1178 * Find a segment which begins after this one does.
1180 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt)
1181 if (GETIP(q)->ip_off > ip->ip_off)
1185 * If there is a preceding segment, it may provide some of
1186 * our data already. If so, drop the data from the incoming
1187 * segment. If it provides all of our data, drop us, otherwise
1188 * stick new segment in the proper place.
1190 * If some of the data is dropped from the the preceding
1191 * segment, then it's checksum is invalidated.
1194 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off;
1196 if (i >= ip->ip_len)
1199 m->m_pkthdr.csum_flags = 0;
1203 m->m_nextpkt = p->m_nextpkt;
1206 m->m_nextpkt = fp->ipq_frags;
1211 * While we overlap succeeding segments trim them or,
1212 * if they are completely covered, dequeue them.
1214 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off;
1216 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off;
1217 if (i < GETIP(q)->ip_len) {
1218 GETIP(q)->ip_len -= i;
1219 GETIP(q)->ip_off += i;
1221 q->m_pkthdr.csum_flags = 0;
1226 ipstat.ips_fragdropped++;
1228 q->m_nextpkt = NULL;
1236 * Transfer firewall instructions to the fragment structure.
1237 * Only trust info in the fragment at offset 0.
1239 if (ip->ip_off == 0) {
1240 fp->ipq_div_info = *divinfo;
1242 mtag = m_tag_find(m, PACKET_TAG_IPFW_DIVERT, NULL);
1244 m_tag_delete(m, mtag);
1250 * Check for complete reassembly and perform frag per packet
1253 * Frag limiting is performed here so that the nth frag has
1254 * a chance to complete the packet before we drop the packet.
1255 * As a result, n+1 frags are actually allowed per packet, but
1256 * only n will ever be stored. (n = maxfragsperpacket.)
1260 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
1261 if (GETIP(q)->ip_off != next) {
1262 if (fp->ipq_nfrags > maxfragsperpacket) {
1263 ipstat.ips_fragdropped += fp->ipq_nfrags;
1268 next += GETIP(q)->ip_len;
1270 /* Make sure the last packet didn't have the IP_MF flag */
1271 if (p->m_flags & M_FRAG) {
1272 if (fp->ipq_nfrags > maxfragsperpacket) {
1273 ipstat.ips_fragdropped += fp->ipq_nfrags;
1280 * Reassembly is complete. Make sure the packet is a sane size.
1284 if (next + (IP_VHL_HL(ip->ip_vhl) << 2) > IP_MAXPACKET) {
1285 ipstat.ips_toolong++;
1286 ipstat.ips_fragdropped += fp->ipq_nfrags;
1292 * Concatenate fragments.
1299 q->m_nextpkt = NULL;
1300 for (q = nq; q != NULL; q = nq) {
1302 q->m_nextpkt = NULL;
1303 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
1304 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
1309 * Clean up the 1's complement checksum. Carry over 16 bits must
1310 * be added back. This assumes no more then 65535 packet fragments
1311 * were reassembled. A second carry can also occur (but not a third).
1313 m->m_pkthdr.csum_data = (m->m_pkthdr.csum_data & 0xffff) +
1314 (m->m_pkthdr.csum_data >> 16);
1315 if (m->m_pkthdr.csum_data > 0xFFFF)
1316 m->m_pkthdr.csum_data -= 0xFFFF;
1321 * Extract firewall instructions from the fragment structure.
1323 *divinfo = fp->ipq_div_info;
1327 * Create header for new ip packet by
1328 * modifying header of first packet;
1329 * dequeue and discard fragment reassembly header.
1330 * Make header visible.
1333 ip->ip_src = fp->ipq_src;
1334 ip->ip_dst = fp->ipq_dst;
1337 mpipe_free(&ipq_mpipe, fp);
1338 m->m_len += (IP_VHL_HL(ip->ip_vhl) << 2);
1339 m->m_data -= (IP_VHL_HL(ip->ip_vhl) << 2);
1340 /* some debugging cruft by sklower, below, will go away soon */
1341 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */
1344 for (n = m; n; n = n->m_next)
1346 m->m_pkthdr.len = plen;
1354 ipstat.ips_fragdropped++;
1364 * Free a fragment reassembly header and all
1365 * associated datagrams.
1368 ip_freef(struct ipq *fp)
1372 while (fp->ipq_frags) {
1374 fp->ipq_frags = q->m_nextpkt;
1375 q->m_nextpkt = NULL;
1379 mpipe_free(&ipq_mpipe, fp);
1384 * IP timer processing;
1385 * if a timer expires on a reassembly
1386 * queue, discard it.
1395 for (i = 0; i < IPREASS_NHASH; i++) {
1399 while (fp != &ipq[i]) {
1402 if (fp->prev->ipq_ttl == 0) {
1403 ipstat.ips_fragtimeout += fp->prev->ipq_nfrags;
1409 * If we are over the maximum number of fragments
1410 * (due to the limit being lowered), drain off
1411 * enough to get down to the new limit.
1413 if (maxnipq >= 0 && nipq > maxnipq) {
1414 for (i = 0; i < IPREASS_NHASH; i++) {
1415 while (nipq > maxnipq &&
1416 (ipq[i].next != &ipq[i])) {
1417 ipstat.ips_fragdropped +=
1418 ipq[i].next->ipq_nfrags;
1419 ip_freef(ipq[i].next);
1428 * Drain off all datagram fragments.
1435 for (i = 0; i < IPREASS_NHASH; i++) {
1436 while (ipq[i].next != &ipq[i]) {
1437 ipstat.ips_fragdropped += ipq[i].next->ipq_nfrags;
1438 ip_freef(ipq[i].next);
1445 * Do option processing on a datagram,
1446 * possibly discarding it if bad options are encountered,
1447 * or forwarding it if source-routed.
1448 * The pass argument is used when operating in the IPSTEALTH
1449 * mode to tell what options to process:
1450 * [LS]SRR (pass 0) or the others (pass 1).
1451 * The reason for as many as two passes is that when doing IPSTEALTH,
1452 * non-routing options should be processed only if the packet is for us.
1453 * Returns 1 if packet has been forwarded/freed,
1454 * 0 if the packet should be processed further.
1457 ip_dooptions(struct mbuf *m, int pass, struct sockaddr_in *next_hop)
1459 struct sockaddr_in ipaddr = { sizeof ipaddr, AF_INET };
1460 struct ip *ip = mtod(m, struct ip *);
1462 struct in_ifaddr *ia;
1463 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB;
1464 boolean_t forward = FALSE;
1465 struct in_addr *sin, dst;
1469 cp = (u_char *)(ip + 1);
1470 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
1471 for (; cnt > 0; cnt -= optlen, cp += optlen) {
1472 opt = cp[IPOPT_OPTVAL];
1473 if (opt == IPOPT_EOL)
1475 if (opt == IPOPT_NOP)
1478 if (cnt < IPOPT_OLEN + sizeof(*cp)) {
1479 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1482 optlen = cp[IPOPT_OLEN];
1483 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
1484 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1494 * Source routing with record.
1495 * Find interface with current destination address.
1496 * If none on this machine then drop if strictly routed,
1497 * or do nothing if loosely routed.
1498 * Record interface address and bring up next address
1499 * component. If strictly routed make sure next
1500 * address is on directly accessible net.
1504 if (ipstealth && pass > 0)
1506 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1507 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1510 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1511 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1514 ipaddr.sin_addr = ip->ip_dst;
1515 ia = (struct in_ifaddr *)
1516 ifa_ifwithaddr((struct sockaddr *)&ipaddr);
1518 if (opt == IPOPT_SSRR) {
1519 type = ICMP_UNREACH;
1520 code = ICMP_UNREACH_SRCFAIL;
1523 if (!ip_dosourceroute)
1524 goto nosourcerouting;
1526 * Loose routing, and not at next destination
1527 * yet; nothing to do except forward.
1531 off--; /* 0 origin */
1532 if (off > optlen - (int)sizeof(struct in_addr)) {
1534 * End of source route. Should be for us.
1536 if (!ip_acceptsourceroute)
1537 goto nosourcerouting;
1538 save_rte(cp, ip->ip_src);
1543 if (!ip_dosourceroute) {
1545 char buf[sizeof "aaa.bbb.ccc.ddd"];
1548 * Acting as a router, so generate ICMP
1551 strcpy(buf, inet_ntoa(ip->ip_dst));
1553 "attempted source route from %s to %s\n",
1554 inet_ntoa(ip->ip_src), buf);
1555 type = ICMP_UNREACH;
1556 code = ICMP_UNREACH_SRCFAIL;
1560 * Not acting as a router,
1564 ipstat.ips_cantforward++;
1571 * locate outgoing interface
1573 memcpy(&ipaddr.sin_addr, cp + off,
1574 sizeof ipaddr.sin_addr);
1576 if (opt == IPOPT_SSRR) {
1577 #define INA struct in_ifaddr *
1578 #define SA struct sockaddr *
1579 if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr))
1581 ia = (INA)ifa_ifwithnet((SA)&ipaddr);
1583 ia = ip_rtaddr(ipaddr.sin_addr,
1584 &ipforward_rt[mycpuid]);
1586 type = ICMP_UNREACH;
1587 code = ICMP_UNREACH_SRCFAIL;
1590 ip->ip_dst = ipaddr.sin_addr;
1591 memcpy(cp + off, &IA_SIN(ia)->sin_addr,
1592 sizeof(struct in_addr));
1593 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1595 * Let ip_intr's mcast routing check handle mcast pkts
1597 forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr));
1601 if (ipstealth && pass == 0)
1603 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1604 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1607 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1608 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1612 * If no space remains, ignore.
1614 off--; /* 0 origin */
1615 if (off > optlen - (int)sizeof(struct in_addr))
1617 memcpy(&ipaddr.sin_addr, &ip->ip_dst,
1618 sizeof ipaddr.sin_addr);
1620 * locate outgoing interface; if we're the destination,
1621 * use the incoming interface (should be same).
1623 if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == NULL &&
1624 (ia = ip_rtaddr(ipaddr.sin_addr,
1625 &ipforward_rt[mycpuid]))
1627 type = ICMP_UNREACH;
1628 code = ICMP_UNREACH_HOST;
1631 memcpy(cp + off, &IA_SIN(ia)->sin_addr,
1632 sizeof(struct in_addr));
1633 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1637 if (ipstealth && pass == 0)
1639 code = cp - (u_char *)ip;
1640 if (optlen < 4 || optlen > 40) {
1641 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1644 if ((off = cp[IPOPT_OFFSET]) < 5) {
1645 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1648 if (off > optlen - (int)sizeof(int32_t)) {
1649 cp[IPOPT_OFFSET + 1] += (1 << 4);
1650 if ((cp[IPOPT_OFFSET + 1] & 0xf0) == 0) {
1651 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1656 off--; /* 0 origin */
1657 sin = (struct in_addr *)(cp + off);
1658 switch (cp[IPOPT_OFFSET + 1] & 0x0f) {
1660 case IPOPT_TS_TSONLY:
1663 case IPOPT_TS_TSANDADDR:
1664 if (off + sizeof(n_time) +
1665 sizeof(struct in_addr) > optlen) {
1666 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1669 ipaddr.sin_addr = dst;
1670 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr,
1674 memcpy(sin, &IA_SIN(ia)->sin_addr,
1675 sizeof(struct in_addr));
1676 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1677 off += sizeof(struct in_addr);
1680 case IPOPT_TS_PRESPEC:
1681 if (off + sizeof(n_time) +
1682 sizeof(struct in_addr) > optlen) {
1683 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1686 memcpy(&ipaddr.sin_addr, sin,
1687 sizeof(struct in_addr));
1688 if (ifa_ifwithaddr((SA)&ipaddr) == NULL)
1690 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1691 off += sizeof(struct in_addr);
1695 code = &cp[IPOPT_OFFSET + 1] - (u_char *)ip;
1699 memcpy(cp + off, &ntime, sizeof(n_time));
1700 cp[IPOPT_OFFSET] += sizeof(n_time);
1703 if (forward && ipforwarding) {
1704 ip_forward(m, TRUE, next_hop);
1709 icmp_error(m, type, code, 0, 0);
1710 ipstat.ips_badoptions++;
1715 * Given address of next destination (final or next hop),
1716 * return internet address info of interface to be used to get there.
1719 ip_rtaddr(struct in_addr dst, struct route *ro)
1721 struct sockaddr_in *sin;
1723 sin = (struct sockaddr_in *)&ro->ro_dst;
1725 if (ro->ro_rt == NULL || dst.s_addr != sin->sin_addr.s_addr) {
1726 if (ro->ro_rt != NULL) {
1730 sin->sin_family = AF_INET;
1731 sin->sin_len = sizeof *sin;
1732 sin->sin_addr = dst;
1733 rtalloc_ign(ro, RTF_PRCLONING);
1736 if (ro->ro_rt == NULL)
1739 return (ifatoia(ro->ro_rt->rt_ifa));
1743 * Save incoming source route for use in replies,
1744 * to be picked up later by ip_srcroute if the receiver is interested.
1747 save_rte(u_char *option, struct in_addr dst)
1751 olen = option[IPOPT_OLEN];
1754 kprintf("save_rte: olen %d\n", olen);
1756 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst)))
1758 bcopy(option, ip_srcrt.srcopt, olen);
1759 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1764 * Retrieve incoming source route for use in replies,
1765 * in the same form used by setsockopt.
1766 * The first hop is placed before the options, will be removed later.
1771 struct in_addr *p, *q;
1776 m = m_get(MB_DONTWAIT, MT_HEADER);
1780 #define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt))
1782 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
1783 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) +
1787 kprintf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len);
1791 * First save first hop for return route
1793 p = &ip_srcrt.route[ip_nhops - 1];
1794 *(mtod(m, struct in_addr *)) = *p--;
1797 kprintf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr));
1801 * Copy option fields and padding (nop) to mbuf.
1803 ip_srcrt.nop = IPOPT_NOP;
1804 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
1805 memcpy(mtod(m, caddr_t) + sizeof(struct in_addr), &ip_srcrt.nop,
1807 q = (struct in_addr *)(mtod(m, caddr_t) +
1808 sizeof(struct in_addr) + OPTSIZ);
1811 * Record return path as an IP source route,
1812 * reversing the path (pointers are now aligned).
1814 while (p >= ip_srcrt.route) {
1817 kprintf(" %x", ntohl(q->s_addr));
1822 * Last hop goes to final destination.
1827 kprintf(" %x\n", ntohl(q->s_addr));
1833 * Strip out IP options.
1836 ip_stripoptions(struct mbuf *m)
1839 struct ip *ip = mtod(m, struct ip *);
1843 optlen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
1844 opts = (caddr_t)(ip + 1);
1845 datalen = m->m_len - (sizeof(struct ip) + optlen);
1846 bcopy(opts + optlen, opts, datalen);
1848 if (m->m_flags & M_PKTHDR)
1849 m->m_pkthdr.len -= optlen;
1850 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, sizeof(struct ip) >> 2);
1853 u_char inetctlerrmap[PRC_NCMDS] = {
1855 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
1856 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
1857 EMSGSIZE, EHOSTUNREACH, 0, 0,
1859 ENOPROTOOPT, ECONNREFUSED
1863 * Forward a packet. If some error occurs return the sender
1864 * an icmp packet. Note we can't always generate a meaningful
1865 * icmp message because icmp doesn't have a large enough repertoire
1866 * of codes and types.
1868 * If not forwarding, just drop the packet. This could be confusing
1869 * if ipforwarding was zero but some routing protocol was advancing
1870 * us as a gateway to somewhere. However, we must let the routing
1871 * protocol deal with that.
1873 * The using_srcrt parameter indicates whether the packet is being forwarded
1874 * via a source route.
1877 ip_forward(struct mbuf *m, boolean_t using_srcrt, struct sockaddr_in *next_hop)
1879 struct ip *ip = mtod(m, struct ip *);
1880 struct sockaddr_in *ipforward_rtaddr;
1882 int error, type = 0, code = 0, destmtu = 0;
1885 struct in_addr pkt_dst;
1887 struct route *cache_rt = &ipforward_rt[mycpuid];
1891 * Cache the destination address of the packet; this may be
1892 * changed by use of 'ipfw fwd'.
1894 pkt_dst = (next_hop != NULL) ? next_hop->sin_addr : ip->ip_dst;
1898 kprintf("forward: src %x dst %x ttl %x\n",
1899 ip->ip_src.s_addr, pkt_dst.s_addr, ip->ip_ttl);
1902 if (m->m_flags & (M_BCAST | M_MCAST) || !in_canforward(pkt_dst)) {
1903 ipstat.ips_cantforward++;
1907 if (!ipstealth && ip->ip_ttl <= IPTTLDEC) {
1908 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0);
1912 ipforward_rtaddr = (struct sockaddr_in *) &cache_rt->ro_dst;
1913 if (cache_rt->ro_rt == NULL ||
1914 ipforward_rtaddr->sin_addr.s_addr != pkt_dst.s_addr) {
1915 if (cache_rt->ro_rt != NULL) {
1916 RTFREE(cache_rt->ro_rt);
1917 cache_rt->ro_rt = NULL;
1919 ipforward_rtaddr->sin_family = AF_INET;
1920 ipforward_rtaddr->sin_len = sizeof(struct sockaddr_in);
1921 ipforward_rtaddr->sin_addr = pkt_dst;
1922 rtalloc_ign(cache_rt, RTF_PRCLONING);
1923 if (cache_rt->ro_rt == NULL) {
1924 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0);
1928 rt = cache_rt->ro_rt;
1931 * Save the IP header and at most 8 bytes of the payload,
1932 * in case we need to generate an ICMP message to the src.
1934 * XXX this can be optimized a lot by saving the data in a local
1935 * buffer on the stack (72 bytes at most), and only allocating the
1936 * mbuf if really necessary. The vast majority of the packets
1937 * are forwarded without having to send an ICMP back (either
1938 * because unnecessary, or because rate limited), so we are
1939 * really we are wasting a lot of work here.
1941 * We don't use m_copy() because it might return a reference
1942 * to a shared cluster. Both this function and ip_output()
1943 * assume exclusive access to the IP header in `m', so any
1944 * data in a cluster may change before we reach icmp_error().
1946 MGETHDR(mcopy, MB_DONTWAIT, m->m_type);
1947 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, MB_DONTWAIT)) {
1949 * It's probably ok if the pkthdr dup fails (because
1950 * the deep copy of the tag chain failed), but for now
1951 * be conservative and just discard the copy since
1952 * code below may some day want the tags.
1957 if (mcopy != NULL) {
1958 mcopy->m_len = imin((IP_VHL_HL(ip->ip_vhl) << 2) + 8,
1960 mcopy->m_pkthdr.len = mcopy->m_len;
1961 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1965 ip->ip_ttl -= IPTTLDEC;
1968 * If forwarding packet using same interface that it came in on,
1969 * perhaps should send a redirect to sender to shortcut a hop.
1970 * Only send redirect if source is sending directly to us,
1971 * and if packet was not source routed (or has any options).
1972 * Also, don't send redirect if forwarding using a default route
1973 * or a route modified by a redirect.
1975 if (rt->rt_ifp == m->m_pkthdr.rcvif &&
1976 !(rt->rt_flags & (RTF_DYNAMIC | RTF_MODIFIED)) &&
1977 satosin(rt_key(rt))->sin_addr.s_addr != INADDR_ANY &&
1978 ipsendredirects && !using_srcrt && next_hop == NULL) {
1979 u_long src = ntohl(ip->ip_src.s_addr);
1980 struct in_ifaddr *rt_ifa = (struct in_ifaddr *)rt->rt_ifa;
1982 if (rt_ifa != NULL &&
1983 (src & rt_ifa->ia_subnetmask) == rt_ifa->ia_subnet) {
1984 if (rt->rt_flags & RTF_GATEWAY)
1985 dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
1987 dest = pkt_dst.s_addr;
1989 * Router requirements says to only send
1992 type = ICMP_REDIRECT;
1993 code = ICMP_REDIRECT_HOST;
1996 kprintf("redirect (%d) to %x\n", code, dest);
2001 if (next_hop != NULL) {
2002 /* Pass IPFORWARD info if available */
2003 tag.mh_type = MT_TAG;
2004 tag.mh_flags = PACKET_TAG_IPFORWARD;
2005 tag.mh_data = (caddr_t)next_hop;
2007 m = (struct mbuf *)&tag;
2010 error = ip_output(m, NULL, cache_rt, IP_FORWARDING, NULL,
2013 ipstat.ips_forward++;
2016 ipflow_create(cache_rt, mcopy);
2019 return; /* most common case */
2021 ipstat.ips_redirectsent++;
2024 ipstat.ips_cantforward++;
2031 * Send ICMP message.
2036 case 0: /* forwarded, but need redirect */
2037 /* type, code set above */
2040 case ENETUNREACH: /* shouldn't happen, checked above */
2045 type = ICMP_UNREACH;
2046 code = ICMP_UNREACH_HOST;
2050 type = ICMP_UNREACH;
2051 code = ICMP_UNREACH_NEEDFRAG;
2054 * If the packet is routed over IPsec tunnel, tell the
2055 * originator the tunnel MTU.
2056 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
2059 if (cache_rt->ro_rt != NULL) {
2060 struct secpolicy *sp = NULL;
2065 sp = ipsec4_getpolicybyaddr(mcopy,
2071 destmtu = cache_rt->ro_rt->rt_ifp->if_mtu;
2073 /* count IPsec header size */
2074 ipsechdr = ipsec4_hdrsiz(mcopy,
2079 * find the correct route for outer IPv4
2080 * header, compute tunnel MTU.
2083 if (sp->req != NULL && sp->req->sav != NULL &&
2084 sp->req->sav->sah != NULL) {
2085 ro = &sp->req->sav->sah->sa_route;
2086 if (ro->ro_rt != NULL &&
2087 ro->ro_rt->rt_ifp != NULL) {
2089 ro->ro_rt->rt_ifp->if_mtu;
2090 destmtu -= ipsechdr;
2099 * If the packet is routed over IPsec tunnel, tell the
2100 * originator the tunnel MTU.
2101 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
2104 if (cache_rt->ro_rt != NULL) {
2105 struct secpolicy *sp = NULL;
2110 sp = ipsec_getpolicybyaddr(mcopy,
2116 destmtu = cache_rt->ro_rt->rt_ifp->if_mtu;
2118 /* count IPsec header size */
2119 ipsechdr = ipsec4_hdrsiz(mcopy,
2124 * find the correct route for outer IPv4
2125 * header, compute tunnel MTU.
2128 if (sp->req != NULL &&
2129 sp->req->sav != NULL &&
2130 sp->req->sav->sah != NULL) {
2131 ro = &sp->req->sav->sah->sa_route;
2132 if (ro->ro_rt != NULL &&
2133 ro->ro_rt->rt_ifp != NULL) {
2135 ro->ro_rt->rt_ifp->if_mtu;
2136 destmtu -= ipsechdr;
2143 #else /* !IPSEC && !FAST_IPSEC */
2144 if (cache_rt->ro_rt != NULL)
2145 destmtu = cache_rt->ro_rt->rt_ifp->if_mtu;
2147 ipstat.ips_cantfrag++;
2152 * A router should not generate ICMP_SOURCEQUENCH as
2153 * required in RFC1812 Requirements for IP Version 4 Routers.
2154 * Source quench could be a big problem under DoS attacks,
2155 * or if the underlying interface is rate-limited.
2156 * Those who need source quench packets may re-enable them
2157 * via the net.inet.ip.sendsourcequench sysctl.
2159 if (!ip_sendsourcequench) {
2163 type = ICMP_SOURCEQUENCH;
2168 case EACCES: /* ipfw denied packet */
2172 icmp_error(mcopy, type, code, dest, destmtu);
2176 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
2179 if (inp->inp_socket->so_options & SO_TIMESTAMP) {
2183 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
2184 SCM_TIMESTAMP, SOL_SOCKET);
2186 mp = &(*mp)->m_next;
2188 if (inp->inp_flags & INP_RECVDSTADDR) {
2189 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
2190 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
2192 mp = &(*mp)->m_next;
2194 if (inp->inp_flags & INP_RECVTTL) {
2195 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl,
2196 sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
2198 mp = &(*mp)->m_next;
2202 * Moving these out of udp_input() made them even more broken
2203 * than they already were.
2205 /* options were tossed already */
2206 if (inp->inp_flags & INP_RECVOPTS) {
2207 *mp = sbcreatecontrol((caddr_t) opts_deleted_above,
2208 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
2210 mp = &(*mp)->m_next;
2212 /* ip_srcroute doesn't do what we want here, need to fix */
2213 if (inp->inp_flags & INP_RECVRETOPTS) {
2214 *mp = sbcreatecontrol((caddr_t) ip_srcroute(),
2215 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
2217 mp = &(*mp)->m_next;
2220 if (inp->inp_flags & INP_RECVIF) {
2223 struct sockaddr_dl sdl;
2226 struct sockaddr_dl *sdp;
2227 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
2229 if (((ifp = m->m_pkthdr.rcvif)) &&
2230 ((ifp->if_index != 0) && (ifp->if_index <= if_index))) {
2231 sdp = IF_LLSOCKADDR(ifp);
2233 * Change our mind and don't try copy.
2235 if ((sdp->sdl_family != AF_LINK) ||
2236 (sdp->sdl_len > sizeof(sdlbuf))) {
2239 bcopy(sdp, sdl2, sdp->sdl_len);
2243 offsetof(struct sockaddr_dl, sdl_data[0]);
2244 sdl2->sdl_family = AF_LINK;
2245 sdl2->sdl_index = 0;
2246 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
2248 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len,
2249 IP_RECVIF, IPPROTO_IP);
2251 mp = &(*mp)->m_next;
2256 * XXX these routines are called from the upper part of the kernel.
2258 * They could also be moved to ip_mroute.c, since all the RSVP
2259 * handling is done there already.
2262 ip_rsvp_init(struct socket *so)
2264 if (so->so_type != SOCK_RAW ||
2265 so->so_proto->pr_protocol != IPPROTO_RSVP)
2268 if (ip_rsvpd != NULL)
2273 * This may seem silly, but we need to be sure we don't over-increment
2274 * the RSVP counter, in case something slips up.
2289 * This may seem silly, but we need to be sure we don't over-decrement
2290 * the RSVP counter, in case something slips up.
2300 rsvp_input(struct mbuf *m, ...) /* XXX must fixup manually */
2306 off = __va_arg(ap, int);
2307 proto = __va_arg(ap, int);
2310 if (rsvp_input_p) { /* call the real one if loaded */
2311 rsvp_input_p(m, off, proto);
2315 /* Can still get packets with rsvp_on = 0 if there is a local member
2316 * of the group to which the RSVP packet is addressed. But in this
2317 * case we want to throw the packet away.
2325 if (ip_rsvpd != NULL) {
2326 rip_input(m, off, proto);
2329 /* Drop the packet */