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) 2003, 2004 Jeffrey M. Hsu. All rights reserved.
37 * License terms: all terms for the DragonFly license above plus the following:
39 * 4. All advertising materials mentioning features or use of this software
40 * must display the following acknowledgement:
42 * This product includes software developed by Jeffrey M. Hsu
43 * for the DragonFly Project.
45 * This requirement may be waived with permission from Jeffrey Hsu.
46 * This requirement will sunset and may be removed on July 8 2005,
47 * after which the standard DragonFly license (as shown above) will
52 * Copyright (c) 1982, 1986, 1988, 1993
53 * The Regents of the University of California. All rights reserved.
55 * Redistribution and use in source and binary forms, with or without
56 * modification, are permitted provided that the following conditions
58 * 1. Redistributions of source code must retain the above copyright
59 * notice, this list of conditions and the following disclaimer.
60 * 2. Redistributions in binary form must reproduce the above copyright
61 * notice, this list of conditions and the following disclaimer in the
62 * documentation and/or other materials provided with the distribution.
63 * 3. All advertising materials mentioning features or use of this software
64 * must display the following acknowledgement:
65 * This product includes software developed by the University of
66 * California, Berkeley and its contributors.
67 * 4. Neither the name of the University nor the names of its contributors
68 * may be used to endorse or promote products derived from this software
69 * without specific prior written permission.
71 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
72 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
73 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
74 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
75 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
76 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
77 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
78 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
79 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
80 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
83 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94
84 * $FreeBSD: src/sys/netinet/ip_input.c,v 1.130.2.52 2003/03/07 07:01:28 silby Exp $
85 * $DragonFly: src/sys/netinet/ip_input.c,v 1.64 2006/12/22 23:57:52 swildner Exp $
90 #include "opt_bootp.h"
93 #include "opt_ipdivert.h"
94 #include "opt_ipfilter.h"
95 #include "opt_ipstealth.h"
96 #include "opt_ipsec.h"
98 #include <sys/param.h>
99 #include <sys/systm.h>
100 #include <sys/mbuf.h>
101 #include <sys/malloc.h>
102 #include <sys/mpipe.h>
103 #include <sys/domain.h>
104 #include <sys/protosw.h>
105 #include <sys/socket.h>
106 #include <sys/time.h>
107 #include <sys/globaldata.h>
108 #include <sys/thread.h>
109 #include <sys/kernel.h>
110 #include <sys/syslog.h>
111 #include <sys/sysctl.h>
112 #include <sys/in_cksum.h>
114 #include <sys/thread2.h>
115 #include <sys/msgport2.h>
117 #include <machine/stdarg.h>
120 #include <net/if_types.h>
121 #include <net/if_var.h>
122 #include <net/if_dl.h>
123 #include <net/pfil.h>
124 #include <net/route.h>
125 #include <net/netisr.h>
126 #include <net/intrq.h>
128 #include <netinet/in.h>
129 #include <netinet/in_systm.h>
130 #include <netinet/in_var.h>
131 #include <netinet/ip.h>
132 #include <netinet/in_pcb.h>
133 #include <netinet/ip_var.h>
134 #include <netinet/ip_icmp.h>
137 #include <sys/socketvar.h>
139 #include <net/ipfw/ip_fw.h>
140 #include <net/dummynet/ip_dummynet.h>
143 #include <netinet6/ipsec.h>
144 #include <netproto/key/key.h>
148 #include <netproto/ipsec/ipsec.h>
149 #include <netproto/ipsec/key.h>
153 static int ip_rsvp_on;
154 struct socket *ip_rsvpd;
156 int ipforwarding = 0;
157 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW,
158 &ipforwarding, 0, "Enable IP forwarding between interfaces");
160 static int ipsendredirects = 1; /* XXX */
161 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW,
162 &ipsendredirects, 0, "Enable sending IP redirects");
164 int ip_defttl = IPDEFTTL;
165 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW,
166 &ip_defttl, 0, "Maximum TTL on IP packets");
168 static int ip_dosourceroute = 0;
169 SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW,
170 &ip_dosourceroute, 0, "Enable forwarding source routed IP packets");
172 static int ip_acceptsourceroute = 0;
173 SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute,
174 CTLFLAG_RW, &ip_acceptsourceroute, 0,
175 "Enable accepting source routed IP packets");
177 static int ip_keepfaith = 0;
178 SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW,
180 "Enable packet capture for FAITH IPv4->IPv6 translater daemon");
182 static int nipq = 0; /* total # of reass queues */
184 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragpackets, CTLFLAG_RW,
186 "Maximum number of IPv4 fragment reassembly queue entries");
188 static int maxfragsperpacket;
189 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW,
190 &maxfragsperpacket, 0,
191 "Maximum number of IPv4 fragments allowed per packet");
193 static int ip_sendsourcequench = 0;
194 SYSCTL_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW,
195 &ip_sendsourcequench, 0,
196 "Enable the transmission of source quench packets");
198 int ip_do_randomid = 0;
199 SYSCTL_INT(_net_inet_ip, OID_AUTO, random_id, CTLFLAG_RW,
201 "Assign random ip_id values");
203 * XXX - Setting ip_checkinterface mostly implements the receive side of
204 * the Strong ES model described in RFC 1122, but since the routing table
205 * and transmit implementation do not implement the Strong ES model,
206 * setting this to 1 results in an odd hybrid.
208 * XXX - ip_checkinterface currently must be disabled if you use ipnat
209 * to translate the destination address to another local interface.
211 * XXX - ip_checkinterface must be disabled if you add IP aliases
212 * to the loopback interface instead of the interface where the
213 * packets for those addresses are received.
215 static int ip_checkinterface = 0;
216 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW,
217 &ip_checkinterface, 0, "Verify packet arrives on correct interface");
220 static int ipprintfs = 0;
223 static struct ifqueue ipintrq;
224 static int ipqmaxlen = IFQ_MAXLEN;
226 extern struct domain inetdomain;
227 extern struct protosw inetsw[];
228 u_char ip_protox[IPPROTO_MAX];
229 struct in_ifaddrhead in_ifaddrhead; /* first inet address */
230 struct in_ifaddrhashhead *in_ifaddrhashtbl; /* inet addr hash table */
231 u_long in_ifaddrhmask; /* mask for hash table */
233 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW,
234 &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue");
235 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD,
236 &ipintrq.ifq_drops, 0, "Number of packets dropped from the IP input queue");
238 struct ip_stats ipstats_percpu[MAXCPU];
241 sysctl_ipstats(SYSCTL_HANDLER_ARGS)
245 for (cpu = 0; cpu < ncpus; ++cpu) {
246 if ((error = SYSCTL_OUT(req, &ipstats_percpu[cpu],
247 sizeof(struct ip_stats))))
249 if ((error = SYSCTL_IN(req, &ipstats_percpu[cpu],
250 sizeof(struct ip_stats))))
256 SYSCTL_PROC(_net_inet_ip, IPCTL_STATS, stats, (CTLTYPE_OPAQUE | CTLFLAG_RW),
257 0, 0, sysctl_ipstats, "S,ip_stats", "IP statistics");
259 SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW,
260 &ipstat, ip_stats, "IP statistics");
263 /* Packet reassembly stuff */
264 #define IPREASS_NHASH_LOG2 6
265 #define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2)
266 #define IPREASS_HMASK (IPREASS_NHASH - 1)
267 #define IPREASS_HASH(x,y) \
268 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK)
270 static struct ipq ipq[IPREASS_NHASH];
271 const int ipintrq_present = 1;
274 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
275 &ip_mtu, 0, "Default MTU");
279 static int ipstealth = 0;
280 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW, &ipstealth, 0, "");
282 static const int ipstealth = 0;
287 ip_fw_chk_t *ip_fw_chk_ptr;
292 ip_dn_io_t *ip_dn_io_ptr;
294 struct pfil_head inet_pfil_hook;
297 * XXX this is ugly -- the following two global variables are
298 * used to store packet state while it travels through the stack.
299 * Note that the code even makes assumptions on the size and
300 * alignment of fields inside struct ip_srcrt so e.g. adding some
301 * fields will break the code. This needs to be fixed.
303 * We need to save the IP options in case a protocol wants to respond
304 * to an incoming packet over the same route if the packet got here
305 * using IP source routing. This allows connection establishment and
306 * maintenance when the remote end is on a network that is not known
309 static int ip_nhops = 0;
311 static struct ip_srcrt {
312 struct in_addr dst; /* final destination */
313 char nop; /* one NOP to align */
314 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */
315 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)];
318 static MALLOC_DEFINE(M_IPQ, "ipq", "IP Fragment Management");
319 static struct malloc_pipe ipq_mpipe;
321 static void save_rte (u_char *, struct in_addr);
322 static int ip_dooptions (struct mbuf *m, int,
323 struct sockaddr_in *next_hop);
324 static void ip_forward (struct mbuf *m, boolean_t using_srcrt,
325 struct sockaddr_in *next_hop);
326 static void ip_freef (struct ipq *);
327 static int ip_input_handler (struct netmsg *);
328 static struct mbuf *ip_reass (struct mbuf *, struct ipq *,
329 struct ipq *, u_int32_t *);
332 * IP initialization: fill in IP protocol switch table.
333 * All protocols not implemented in kernel go to raw IP protocol handler.
345 * Make sure we can handle a reasonable number of fragments but
346 * cap it at 4000 (XXX).
348 mpipe_init(&ipq_mpipe, M_IPQ, sizeof(struct ipq),
349 IFQ_MAXLEN, 4000, 0, NULL);
350 TAILQ_INIT(&in_ifaddrhead);
351 in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask);
352 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
355 for (i = 0; i < IPPROTO_MAX; i++)
356 ip_protox[i] = pr - inetsw;
357 for (pr = inetdomain.dom_protosw;
358 pr < inetdomain.dom_protoswNPROTOSW; pr++)
359 if (pr->pr_domain->dom_family == PF_INET &&
360 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
361 ip_protox[pr->pr_protocol] = pr - inetsw;
363 inet_pfil_hook.ph_type = PFIL_TYPE_AF;
364 inet_pfil_hook.ph_af = AF_INET;
365 if ((i = pfil_head_register(&inet_pfil_hook)) != 0) {
366 kprintf("%s: WARNING: unable to register pfil hook, "
367 "error %d\n", __func__, i);
370 for (i = 0; i < IPREASS_NHASH; i++)
371 ipq[i].next = ipq[i].prev = &ipq[i];
373 maxnipq = nmbclusters / 32;
374 maxfragsperpacket = 16;
376 ip_id = time_second & 0xffff;
377 ipintrq.ifq_maxlen = ipqmaxlen;
380 * Initialize IP statistics counters for each CPU.
384 for (cpu = 0; cpu < ncpus; ++cpu) {
385 bzero(&ipstats_percpu[cpu], sizeof(struct ip_stats));
388 bzero(&ipstat, sizeof(struct ip_stats));
391 netisr_register(NETISR_IP, ip_mport, ip_input_handler);
395 * XXX watch out this one. It is perhaps used as a cache for
396 * the most recently used route ? it is cleared in in_addroute()
397 * when a new route is successfully created.
399 struct route ipforward_rt[MAXCPU];
401 /* Do transport protocol processing. */
403 transport_processing_oncpu(struct mbuf *m, int hlen, struct ip *ip,
404 struct sockaddr_in *nexthop)
407 * Switch out to protocol's input routine.
409 if (nexthop && ip->ip_p == IPPROTO_TCP) {
410 /* TCP needs IPFORWARD info if available */
413 tag.mh_type = MT_TAG;
414 tag.mh_flags = PACKET_TAG_IPFORWARD;
415 tag.mh_data = (caddr_t)nexthop;
418 (*inetsw[ip_protox[ip->ip_p]].pr_input)
419 ((struct mbuf *)&tag, hlen, ip->ip_p);
421 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen, ip->ip_p);
425 struct netmsg_transport_packet {
426 struct lwkt_msg nm_lmsg;
427 struct mbuf *nm_mbuf;
429 boolean_t nm_hasnexthop;
430 struct sockaddr_in nm_nexthop;
434 transport_processing_handler(lwkt_msg_t lmsg)
436 struct netmsg_transport_packet *msg = (void *)lmsg;
437 struct sockaddr_in *nexthop;
440 ip = mtod(msg->nm_mbuf, struct ip *);
441 nexthop = msg->nm_hasnexthop ? &msg->nm_nexthop : NULL;
442 transport_processing_oncpu(msg->nm_mbuf, msg->nm_hlen, ip, nexthop);
443 lwkt_replymsg(lmsg, 0);
448 ip_input_handler(struct netmsg *msg0)
450 struct mbuf *m = ((struct netmsg_packet *)msg0)->nm_packet;
453 /* msg0 was embedded in the mbuf, do not reply! */
458 * IP input routine. Checksum and byte swap header. If fragmented
459 * try to reassemble. Process options. Pass to next level.
462 ip_input(struct mbuf *m)
466 struct in_ifaddr *ia = NULL;
468 int i, hlen, checkif;
470 struct in_addr pkt_dst;
471 u_int32_t divert_info = 0; /* packet divert/tee info */
472 struct ip_fw_args args;
473 boolean_t using_srcrt = FALSE; /* forward (by PFIL_HOOKS) */
474 boolean_t needredispatch = FALSE;
475 struct in_addr odst; /* original dst address(NAT) */
476 #if defined(FAST_IPSEC) || defined(IPDIVERT)
480 struct tdb_ident *tdbi;
481 struct secpolicy *sp;
488 args.next_hop = NULL;
490 /* Grab info from MT_TAG mbufs prepended to the chain. */
491 while (m != NULL && m->m_type == MT_TAG) {
492 switch(m->_m_tag_id) {
493 case PACKET_TAG_DUMMYNET:
494 args.rule = ((struct dn_pkt *)m)->rule;
496 case PACKET_TAG_IPFORWARD:
497 args.next_hop = (struct sockaddr_in *)m->m_hdr.mh_data;
500 kprintf("ip_input: unrecognised MT_TAG tag %d\n",
506 KASSERT(m != NULL && (m->m_flags & M_PKTHDR), ("ip_input: no HDR"));
508 if (args.rule != NULL) { /* dummynet already filtered us */
509 ip = mtod(m, struct ip *);
510 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
516 /* length checks already done in ip_demux() */
517 KASSERT(m->m_len >= sizeof(ip), ("IP header not in one mbuf"));
519 ip = mtod(m, struct ip *);
521 if (IP_VHL_V(ip->ip_vhl) != IPVERSION) {
522 ipstat.ips_badvers++;
526 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
527 /* length checks already done in ip_demux() */
528 KASSERT(hlen >= sizeof(struct ip), ("IP header len too small"));
529 KASSERT(m->m_len >= hlen, ("packet shorter than IP header length"));
531 /* 127/8 must not appear on wire - RFC1122 */
532 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
533 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
534 if (!(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK)) {
535 ipstat.ips_badaddr++;
540 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
541 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
543 if (hlen == sizeof(struct ip)) {
544 sum = in_cksum_hdr(ip);
546 sum = in_cksum(m, hlen);
555 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) {
556 /* packet is dropped by traffic conditioner */
561 * Convert fields to host representation.
563 ip->ip_len = ntohs(ip->ip_len);
564 if (ip->ip_len < hlen) {
568 ip->ip_off = ntohs(ip->ip_off);
571 * Check that the amount of data in the buffers
572 * is as at least much as the IP header would have us expect.
573 * Trim mbufs if longer than we expect.
574 * Drop packet if shorter than we expect.
576 if (m->m_pkthdr.len < ip->ip_len) {
577 ipstat.ips_tooshort++;
580 if (m->m_pkthdr.len > ip->ip_len) {
581 if (m->m_len == m->m_pkthdr.len) {
582 m->m_len = ip->ip_len;
583 m->m_pkthdr.len = ip->ip_len;
585 m_adj(m, ip->ip_len - m->m_pkthdr.len);
587 #if defined(IPSEC) && !defined(IPSEC_FILTERGIF)
589 * Bypass packet filtering for packets from a tunnel (gif).
591 if (ipsec_gethist(m, NULL))
597 * Right now when no processing on packet has done
598 * and it is still fresh out of network we do our black
600 * - Firewall: deny/allow/divert
601 * - Xlate: translate packet's addr/port (NAT).
602 * - Pipe: pass pkt through dummynet.
603 * - Wrap: fake packet's addr/port <unimpl.>
604 * - Encapsulate: put it in another IP and send out. <unimp.>
610 * Run through list of hooks for input packets.
612 * NB: Beware of the destination address changing (e.g.
613 * by NAT rewriting). When this happens, tell
614 * ip_forward to do the right thing.
616 if (pfil_has_hooks(&inet_pfil_hook)) {
618 if (pfil_run_hooks(&inet_pfil_hook, &m,
619 m->m_pkthdr.rcvif, PFIL_IN)) {
622 if (m == NULL) /* consumed by filter */
624 ip = mtod(m, struct ip *);
625 using_srcrt = (odst.s_addr != ip->ip_dst.s_addr);
628 if (fw_enable && IPFW_LOADED) {
630 * If we've been forwarded from the output side, then
631 * skip the firewall a second time
633 if (args.next_hop != NULL)
637 i = ip_fw_chk_ptr(&args);
640 if ((i & IP_FW_PORT_DENY_FLAG) || m == NULL) { /* drop */
645 ip = mtod(m, struct ip *); /* just in case m changed */
646 if (i == 0 && args.next_hop == NULL) /* common case */
648 if (DUMMYNET_LOADED && (i & IP_FW_PORT_DYNT_FLAG)) {
649 /* Send packet to the appropriate pipe */
650 ip_dn_io_ptr(m, i&0xffff, DN_TO_IP_IN, &args);
654 if (i != 0 && !(i & IP_FW_PORT_DYNT_FLAG)) {
655 /* Divert or tee packet */
660 if (i == 0 && args.next_hop != NULL)
663 * if we get here, the packet must be dropped
671 * Process options and, if not destined for us,
672 * ship it on. ip_dooptions returns 1 when an
673 * error was detected (causing an icmp message
674 * to be sent and the original packet to be freed).
676 ip_nhops = 0; /* for source routed packets */
677 if (hlen > sizeof(struct ip) && ip_dooptions(m, 0, args.next_hop))
680 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
681 * matter if it is destined to another node, or whether it is
682 * a multicast one, RSVP wants it! and prevents it from being forwarded
683 * anywhere else. Also checks if the rsvp daemon is running before
684 * grabbing the packet.
686 if (rsvp_on && ip->ip_p == IPPROTO_RSVP)
690 * Check our list of addresses, to see if the packet is for us.
691 * If we don't have any addresses, assume any unicast packet
692 * we receive might be for us (and let the upper layers deal
695 if (TAILQ_EMPTY(&in_ifaddrhead) && !(m->m_flags & (M_MCAST | M_BCAST)))
699 * Cache the destination address of the packet; this may be
700 * changed by use of 'ipfw fwd'.
702 pkt_dst = args.next_hop ? args.next_hop->sin_addr : ip->ip_dst;
705 * Enable a consistency check between the destination address
706 * and the arrival interface for a unicast packet (the RFC 1122
707 * strong ES model) if IP forwarding is disabled and the packet
708 * is not locally generated and the packet is not subject to
711 * XXX - Checking also should be disabled if the destination
712 * address is ipnat'ed to a different interface.
714 * XXX - Checking is incompatible with IP aliases added
715 * to the loopback interface instead of the interface where
716 * the packets are received.
718 checkif = ip_checkinterface &&
720 m->m_pkthdr.rcvif != NULL &&
721 !(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) &&
722 (args.next_hop == NULL);
725 * Check for exact addresses in the hash bucket.
727 LIST_FOREACH(ia, INADDR_HASH(pkt_dst.s_addr), ia_hash) {
729 * If the address matches, verify that the packet
730 * arrived via the correct interface if checking is
733 if (IA_SIN(ia)->sin_addr.s_addr == pkt_dst.s_addr &&
734 (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif))
738 * Check for broadcast addresses.
740 * Only accept broadcast packets that arrive via the matching
741 * interface. Reception of forwarded directed broadcasts would
742 * be handled via ip_forward() and ether_output() with the loopback
743 * into the stack for SIMPLEX interfaces handled by ether_output().
745 if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
746 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
747 if (ifa->ifa_addr == NULL) /* shutdown/startup race */
749 if (ifa->ifa_addr->sa_family != AF_INET)
752 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
755 if (ia->ia_netbroadcast.s_addr == pkt_dst.s_addr)
758 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY)
763 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
764 struct in_multi *inm;
766 if (ip_mrouter != NULL) {
768 * If we are acting as a multicast router, all
769 * incoming multicast packets are passed to the
770 * kernel-level multicast forwarding function.
771 * The packet is returned (relatively) intact; if
772 * ip_mforward() returns a non-zero value, the packet
773 * must be discarded, else it may be accepted below.
775 if (ip_mforward != NULL &&
776 ip_mforward(ip, m->m_pkthdr.rcvif, m, NULL) != 0) {
777 ipstat.ips_cantforward++;
783 * The process-level routing daemon needs to receive
784 * all multicast IGMP packets, whether or not this
785 * host belongs to their destination groups.
787 if (ip->ip_p == IPPROTO_IGMP)
789 ipstat.ips_forward++;
792 * See if we belong to the destination multicast group on the
795 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
797 ipstat.ips_notmember++;
803 if (ip->ip_dst.s_addr == INADDR_BROADCAST)
805 if (ip->ip_dst.s_addr == INADDR_ANY)
809 * FAITH(Firewall Aided Internet Translator)
811 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
813 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP)
821 * Not for us; forward if possible and desirable.
824 ipstat.ips_cantforward++;
829 * Enforce inbound IPsec SPD.
831 if (ipsec4_in_reject(m, NULL)) {
832 ipsecstat.in_polvio++;
837 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
840 tdbi = (struct tdb_ident *)m_tag_data(mtag);
841 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
843 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
844 IP_FORWARDING, &error);
846 if (sp == NULL) { /* NB: can happen if error */
848 /*XXX error stat???*/
849 DPRINTF(("ip_input: no SP for forwarding\n")); /*XXX*/
854 * Check security policy against packet attributes.
856 error = ipsec_in_reject(sp, m);
860 ipstat.ips_cantforward++;
864 ip_forward(m, using_srcrt, args.next_hop);
871 * IPSTEALTH: Process non-routing options only
872 * if the packet is destined for us.
875 hlen > sizeof(struct ip) &&
876 ip_dooptions(m, 1, args.next_hop))
879 /* Count the packet in the ip address stats */
881 ia->ia_ifa.if_ipackets++;
882 ia->ia_ifa.if_ibytes += m->m_pkthdr.len;
886 * If offset or IP_MF are set, must reassemble.
887 * Otherwise, nothing need be done.
888 * (We could look in the reassembly queue to see
889 * if the packet was previously fragmented,
890 * but it's not worth the time; just let them time out.)
892 if (ip->ip_off & (IP_MF | IP_OFFMASK)) {
894 /* If maxnipq is 0, never accept fragments. */
896 ipstat.ips_fragments++;
897 ipstat.ips_fragdropped++;
901 sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
903 * Look for queue of fragments
906 for (fp = ipq[sum].next; fp != &ipq[sum]; fp = fp->next)
907 if (ip->ip_id == fp->ipq_id &&
908 ip->ip_src.s_addr == fp->ipq_src.s_addr &&
909 ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
910 ip->ip_p == fp->ipq_p)
916 * Enforce upper bound on number of fragmented packets
917 * for which we attempt reassembly;
918 * If maxnipq is -1, accept all fragments without limitation.
920 if ((nipq > maxnipq) && (maxnipq > 0)) {
922 * drop something from the tail of the current queue
923 * before proceeding further
925 if (ipq[sum].prev == &ipq[sum]) { /* gak */
926 for (i = 0; i < IPREASS_NHASH; i++) {
927 if (ipq[i].prev != &ipq[i]) {
928 ipstat.ips_fragtimeout +=
929 ipq[i].prev->ipq_nfrags;
930 ip_freef(ipq[i].prev);
935 ipstat.ips_fragtimeout +=
936 ipq[sum].prev->ipq_nfrags;
937 ip_freef(ipq[sum].prev);
942 * Adjust ip_len to not reflect header,
943 * convert offset of this to bytes.
946 if (ip->ip_off & IP_MF) {
948 * Make sure that fragments have a data length
949 * that's a non-zero multiple of 8 bytes.
951 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
952 ipstat.ips_toosmall++; /* XXX */
955 m->m_flags |= M_FRAG;
957 m->m_flags &= ~M_FRAG;
961 * Attempt reassembly; if it succeeds, proceed.
962 * ip_reass() will return a different mbuf, and update
963 * the divert info in divert_info.
965 ipstat.ips_fragments++;
966 m->m_pkthdr.header = ip;
967 m = ip_reass(m, fp, &ipq[sum], &divert_info);
970 ipstat.ips_reassembled++;
971 needredispatch = TRUE;
972 ip = mtod(m, struct ip *);
973 /* Get the header length of the reassembled packet */
974 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
976 /* Restore original checksum before diverting packet */
977 if (divert_info != 0) {
979 ip->ip_len = htons(ip->ip_len);
980 ip->ip_off = htons(ip->ip_off);
982 if (hlen == sizeof(struct ip))
983 ip->ip_sum = in_cksum_hdr(ip);
985 ip->ip_sum = in_cksum(m, hlen);
986 ip->ip_off = ntohs(ip->ip_off);
987 ip->ip_len = ntohs(ip->ip_len);
997 * Divert or tee packet to the divert protocol if required.
999 if (divert_info != 0) {
1000 struct mbuf *clone = NULL;
1002 /* Clone packet if we're doing a 'tee' */
1003 if ((divert_info & IP_FW_PORT_TEE_FLAG) != 0)
1004 clone = m_dup(m, MB_DONTWAIT);
1006 /* Restore packet header fields to original values */
1008 ip->ip_len = htons(ip->ip_len);
1009 ip->ip_off = htons(ip->ip_off);
1011 /* Deliver packet to divert input routine */
1012 divert_packet(m, 1, divert_info & 0xffff);
1013 ipstat.ips_delivered++;
1015 /* If 'tee', continue with original packet */
1019 ip = mtod(m, struct ip *);
1022 * Jump backwards to complete processing of the
1023 * packet. But first clear divert_info to avoid
1024 * entering this block again.
1025 * We do not need to clear args.divert_rule
1026 * or args.next_hop as they will not be used.
1028 * XXX Better safe than sorry, remove the DIVERT tag.
1030 mtag = m_tag_find(m, PACKET_TAG_IPFW_DIVERT, NULL);
1032 m_tag_delete(m, mtag);
1041 * enforce IPsec policy checking if we are seeing last header.
1042 * note that we do not visit this with protocols with pcb layer
1043 * code - like udp/tcp/raw ip.
1045 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) &&
1046 ipsec4_in_reject(m, NULL)) {
1047 ipsecstat.in_polvio++;
1053 * enforce IPsec policy checking if we are seeing last header.
1054 * note that we do not visit this with protocols with pcb layer
1055 * code - like udp/tcp/raw ip.
1057 if (inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) {
1059 * Check if the packet has already had IPsec processing
1060 * done. If so, then just pass it along. This tag gets
1061 * set during AH, ESP, etc. input handling, before the
1062 * packet is returned to the ip input queue for delivery.
1064 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
1067 tdbi = (struct tdb_ident *)m_tag_data(mtag);
1068 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
1070 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
1071 IP_FORWARDING, &error);
1075 * Check security policy against packet attributes.
1077 error = ipsec_in_reject(sp, m);
1080 /* XXX error stat??? */
1082 DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/
1089 #endif /* FAST_IPSEC */
1091 ipstat.ips_delivered++;
1092 if (needredispatch) {
1093 struct netmsg_transport_packet *msg;
1096 ip->ip_off = htons(ip->ip_off);
1097 ip->ip_len = htons(ip->ip_len);
1098 port = ip_mport(&m);
1102 msg = kmalloc(sizeof(struct netmsg_transport_packet), M_LWKTMSG,
1103 M_INTWAIT | M_NULLOK);
1107 lwkt_initmsg(&msg->nm_lmsg, &netisr_afree_rport, 0,
1108 lwkt_cmd_func(transport_processing_handler),
1110 msg->nm_hlen = hlen;
1111 msg->nm_hasnexthop = (args.next_hop != NULL);
1112 if (msg->nm_hasnexthop)
1113 msg->nm_nexthop = *args.next_hop; /* structure copy */
1116 ip = mtod(m, struct ip *);
1117 ip->ip_len = ntohs(ip->ip_len);
1118 ip->ip_off = ntohs(ip->ip_off);
1119 lwkt_sendmsg(port, &msg->nm_lmsg);
1121 transport_processing_oncpu(m, hlen, ip, args.next_hop);
1130 * Take incoming datagram fragment and try to reassemble it into
1131 * whole datagram. If a chain for reassembly of this datagram already
1132 * exists, then it is given as fp; otherwise have to make a chain.
1134 * When IPDIVERT enabled, keep additional state with each packet that
1135 * tells us if we need to divert or tee the packet we're building.
1136 * In particular, *divinfo includes the port and TEE flag.
1139 static struct mbuf *
1140 ip_reass(struct mbuf *m, struct ipq *fp, struct ipq *where,
1143 struct ip *ip = mtod(m, struct ip *);
1144 struct mbuf *p = NULL, *q, *nq;
1146 int hlen = IP_VHL_HL(ip->ip_vhl) << 2;
1153 * Presence of header sizes in mbufs
1154 * would confuse code below.
1160 * If first fragment to arrive, create a reassembly queue.
1163 if ((fp = mpipe_alloc_nowait(&ipq_mpipe)) == NULL)
1168 fp->ipq_ttl = IPFRAGTTL;
1169 fp->ipq_p = ip->ip_p;
1170 fp->ipq_id = ip->ip_id;
1171 fp->ipq_src = ip->ip_src;
1172 fp->ipq_dst = ip->ip_dst;
1174 m->m_nextpkt = NULL;
1176 fp->ipq_div_info = 0;
1183 #define GETIP(m) ((struct ip*)((m)->m_pkthdr.header))
1186 * Find a segment which begins after this one does.
1188 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt)
1189 if (GETIP(q)->ip_off > ip->ip_off)
1193 * If there is a preceding segment, it may provide some of
1194 * our data already. If so, drop the data from the incoming
1195 * segment. If it provides all of our data, drop us, otherwise
1196 * stick new segment in the proper place.
1198 * If some of the data is dropped from the the preceding
1199 * segment, then it's checksum is invalidated.
1202 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off;
1204 if (i >= ip->ip_len)
1207 m->m_pkthdr.csum_flags = 0;
1211 m->m_nextpkt = p->m_nextpkt;
1214 m->m_nextpkt = fp->ipq_frags;
1219 * While we overlap succeeding segments trim them or,
1220 * if they are completely covered, dequeue them.
1222 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off;
1224 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off;
1225 if (i < GETIP(q)->ip_len) {
1226 GETIP(q)->ip_len -= i;
1227 GETIP(q)->ip_off += i;
1229 q->m_pkthdr.csum_flags = 0;
1234 ipstat.ips_fragdropped++;
1236 q->m_nextpkt = NULL;
1244 * Transfer firewall instructions to the fragment structure.
1245 * Only trust info in the fragment at offset 0.
1247 if (ip->ip_off == 0) {
1248 fp->ipq_div_info = *divinfo;
1250 mtag = m_tag_find(m, PACKET_TAG_IPFW_DIVERT, NULL);
1252 m_tag_delete(m, mtag);
1258 * Check for complete reassembly and perform frag per packet
1261 * Frag limiting is performed here so that the nth frag has
1262 * a chance to complete the packet before we drop the packet.
1263 * As a result, n+1 frags are actually allowed per packet, but
1264 * only n will ever be stored. (n = maxfragsperpacket.)
1268 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
1269 if (GETIP(q)->ip_off != next) {
1270 if (fp->ipq_nfrags > maxfragsperpacket) {
1271 ipstat.ips_fragdropped += fp->ipq_nfrags;
1276 next += GETIP(q)->ip_len;
1278 /* Make sure the last packet didn't have the IP_MF flag */
1279 if (p->m_flags & M_FRAG) {
1280 if (fp->ipq_nfrags > maxfragsperpacket) {
1281 ipstat.ips_fragdropped += fp->ipq_nfrags;
1288 * Reassembly is complete. Make sure the packet is a sane size.
1292 if (next + (IP_VHL_HL(ip->ip_vhl) << 2) > IP_MAXPACKET) {
1293 ipstat.ips_toolong++;
1294 ipstat.ips_fragdropped += fp->ipq_nfrags;
1300 * Concatenate fragments.
1307 q->m_nextpkt = NULL;
1308 for (q = nq; q != NULL; q = nq) {
1310 q->m_nextpkt = NULL;
1311 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
1312 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
1318 * Extract firewall instructions from the fragment structure.
1320 *divinfo = fp->ipq_div_info;
1324 * Create header for new ip packet by
1325 * modifying header of first packet;
1326 * dequeue and discard fragment reassembly header.
1327 * Make header visible.
1330 ip->ip_src = fp->ipq_src;
1331 ip->ip_dst = fp->ipq_dst;
1334 mpipe_free(&ipq_mpipe, fp);
1335 m->m_len += (IP_VHL_HL(ip->ip_vhl) << 2);
1336 m->m_data -= (IP_VHL_HL(ip->ip_vhl) << 2);
1337 /* some debugging cruft by sklower, below, will go away soon */
1338 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */
1341 for (n = m; n; n = n->m_next)
1343 m->m_pkthdr.len = plen;
1351 ipstat.ips_fragdropped++;
1361 * Free a fragment reassembly header and all
1362 * associated datagrams.
1365 ip_freef(struct ipq *fp)
1369 while (fp->ipq_frags) {
1371 fp->ipq_frags = q->m_nextpkt;
1372 q->m_nextpkt = NULL;
1376 mpipe_free(&ipq_mpipe, fp);
1381 * IP timer processing;
1382 * if a timer expires on a reassembly
1383 * queue, discard it.
1392 for (i = 0; i < IPREASS_NHASH; i++) {
1396 while (fp != &ipq[i]) {
1399 if (fp->prev->ipq_ttl == 0) {
1400 ipstat.ips_fragtimeout += fp->prev->ipq_nfrags;
1406 * If we are over the maximum number of fragments
1407 * (due to the limit being lowered), drain off
1408 * enough to get down to the new limit.
1410 if (maxnipq >= 0 && nipq > maxnipq) {
1411 for (i = 0; i < IPREASS_NHASH; i++) {
1412 while (nipq > maxnipq &&
1413 (ipq[i].next != &ipq[i])) {
1414 ipstat.ips_fragdropped +=
1415 ipq[i].next->ipq_nfrags;
1416 ip_freef(ipq[i].next);
1425 * Drain off all datagram fragments.
1432 for (i = 0; i < IPREASS_NHASH; i++) {
1433 while (ipq[i].next != &ipq[i]) {
1434 ipstat.ips_fragdropped += ipq[i].next->ipq_nfrags;
1435 ip_freef(ipq[i].next);
1442 * Do option processing on a datagram,
1443 * possibly discarding it if bad options are encountered,
1444 * or forwarding it if source-routed.
1445 * The pass argument is used when operating in the IPSTEALTH
1446 * mode to tell what options to process:
1447 * [LS]SRR (pass 0) or the others (pass 1).
1448 * The reason for as many as two passes is that when doing IPSTEALTH,
1449 * non-routing options should be processed only if the packet is for us.
1450 * Returns 1 if packet has been forwarded/freed,
1451 * 0 if the packet should be processed further.
1454 ip_dooptions(struct mbuf *m, int pass, struct sockaddr_in *next_hop)
1456 struct sockaddr_in ipaddr = { sizeof ipaddr, AF_INET };
1457 struct ip *ip = mtod(m, struct ip *);
1459 struct in_ifaddr *ia;
1460 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB;
1461 boolean_t forward = FALSE;
1462 struct in_addr *sin, dst;
1466 cp = (u_char *)(ip + 1);
1467 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
1468 for (; cnt > 0; cnt -= optlen, cp += optlen) {
1469 opt = cp[IPOPT_OPTVAL];
1470 if (opt == IPOPT_EOL)
1472 if (opt == IPOPT_NOP)
1475 if (cnt < IPOPT_OLEN + sizeof(*cp)) {
1476 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1479 optlen = cp[IPOPT_OLEN];
1480 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
1481 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1491 * Source routing with record.
1492 * Find interface with current destination address.
1493 * If none on this machine then drop if strictly routed,
1494 * or do nothing if loosely routed.
1495 * Record interface address and bring up next address
1496 * component. If strictly routed make sure next
1497 * address is on directly accessible net.
1501 if (ipstealth && pass > 0)
1503 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1504 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1507 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1508 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1511 ipaddr.sin_addr = ip->ip_dst;
1512 ia = (struct in_ifaddr *)
1513 ifa_ifwithaddr((struct sockaddr *)&ipaddr);
1515 if (opt == IPOPT_SSRR) {
1516 type = ICMP_UNREACH;
1517 code = ICMP_UNREACH_SRCFAIL;
1520 if (!ip_dosourceroute)
1521 goto nosourcerouting;
1523 * Loose routing, and not at next destination
1524 * yet; nothing to do except forward.
1528 off--; /* 0 origin */
1529 if (off > optlen - (int)sizeof(struct in_addr)) {
1531 * End of source route. Should be for us.
1533 if (!ip_acceptsourceroute)
1534 goto nosourcerouting;
1535 save_rte(cp, ip->ip_src);
1540 if (!ip_dosourceroute) {
1542 char buf[sizeof "aaa.bbb.ccc.ddd"];
1545 * Acting as a router, so generate ICMP
1548 strcpy(buf, inet_ntoa(ip->ip_dst));
1550 "attempted source route from %s to %s\n",
1551 inet_ntoa(ip->ip_src), buf);
1552 type = ICMP_UNREACH;
1553 code = ICMP_UNREACH_SRCFAIL;
1557 * Not acting as a router,
1561 ipstat.ips_cantforward++;
1568 * locate outgoing interface
1570 memcpy(&ipaddr.sin_addr, cp + off,
1571 sizeof ipaddr.sin_addr);
1573 if (opt == IPOPT_SSRR) {
1574 #define INA struct in_ifaddr *
1575 #define SA struct sockaddr *
1576 if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr))
1578 ia = (INA)ifa_ifwithnet((SA)&ipaddr);
1580 ia = ip_rtaddr(ipaddr.sin_addr,
1581 &ipforward_rt[mycpuid]);
1583 type = ICMP_UNREACH;
1584 code = ICMP_UNREACH_SRCFAIL;
1587 ip->ip_dst = ipaddr.sin_addr;
1588 memcpy(cp + off, &IA_SIN(ia)->sin_addr,
1589 sizeof(struct in_addr));
1590 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1592 * Let ip_intr's mcast routing check handle mcast pkts
1594 forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr));
1598 if (ipstealth && pass == 0)
1600 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1601 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1604 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1605 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1609 * If no space remains, ignore.
1611 off--; /* 0 origin */
1612 if (off > optlen - (int)sizeof(struct in_addr))
1614 memcpy(&ipaddr.sin_addr, &ip->ip_dst,
1615 sizeof ipaddr.sin_addr);
1617 * locate outgoing interface; if we're the destination,
1618 * use the incoming interface (should be same).
1620 if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == NULL &&
1621 (ia = ip_rtaddr(ipaddr.sin_addr,
1622 &ipforward_rt[mycpuid]))
1624 type = ICMP_UNREACH;
1625 code = ICMP_UNREACH_HOST;
1628 memcpy(cp + off, &IA_SIN(ia)->sin_addr,
1629 sizeof(struct in_addr));
1630 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1634 if (ipstealth && pass == 0)
1636 code = cp - (u_char *)ip;
1637 if (optlen < 4 || optlen > 40) {
1638 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1641 if ((off = cp[IPOPT_OFFSET]) < 5) {
1642 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1645 if (off > optlen - (int)sizeof(int32_t)) {
1646 cp[IPOPT_OFFSET + 1] += (1 << 4);
1647 if ((cp[IPOPT_OFFSET + 1] & 0xf0) == 0) {
1648 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1653 off--; /* 0 origin */
1654 sin = (struct in_addr *)(cp + off);
1655 switch (cp[IPOPT_OFFSET + 1] & 0x0f) {
1657 case IPOPT_TS_TSONLY:
1660 case IPOPT_TS_TSANDADDR:
1661 if (off + sizeof(n_time) +
1662 sizeof(struct in_addr) > optlen) {
1663 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1666 ipaddr.sin_addr = dst;
1667 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr,
1671 memcpy(sin, &IA_SIN(ia)->sin_addr,
1672 sizeof(struct in_addr));
1673 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1674 off += sizeof(struct in_addr);
1677 case IPOPT_TS_PRESPEC:
1678 if (off + sizeof(n_time) +
1679 sizeof(struct in_addr) > optlen) {
1680 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1683 memcpy(&ipaddr.sin_addr, sin,
1684 sizeof(struct in_addr));
1685 if (ifa_ifwithaddr((SA)&ipaddr) == NULL)
1687 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1688 off += sizeof(struct in_addr);
1692 code = &cp[IPOPT_OFFSET + 1] - (u_char *)ip;
1696 memcpy(cp + off, &ntime, sizeof(n_time));
1697 cp[IPOPT_OFFSET] += sizeof(n_time);
1700 if (forward && ipforwarding) {
1701 ip_forward(m, TRUE, next_hop);
1706 icmp_error(m, type, code, 0, 0);
1707 ipstat.ips_badoptions++;
1712 * Given address of next destination (final or next hop),
1713 * return internet address info of interface to be used to get there.
1716 ip_rtaddr(struct in_addr dst, struct route *ro)
1718 struct sockaddr_in *sin;
1720 sin = (struct sockaddr_in *)&ro->ro_dst;
1722 if (ro->ro_rt == NULL || dst.s_addr != sin->sin_addr.s_addr) {
1723 if (ro->ro_rt != NULL) {
1727 sin->sin_family = AF_INET;
1728 sin->sin_len = sizeof *sin;
1729 sin->sin_addr = dst;
1730 rtalloc_ign(ro, RTF_PRCLONING);
1733 if (ro->ro_rt == NULL)
1736 return (ifatoia(ro->ro_rt->rt_ifa));
1740 * Save incoming source route for use in replies,
1741 * to be picked up later by ip_srcroute if the receiver is interested.
1744 save_rte(u_char *option, struct in_addr dst)
1748 olen = option[IPOPT_OLEN];
1751 kprintf("save_rte: olen %d\n", olen);
1753 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst)))
1755 bcopy(option, ip_srcrt.srcopt, olen);
1756 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1761 * Retrieve incoming source route for use in replies,
1762 * in the same form used by setsockopt.
1763 * The first hop is placed before the options, will be removed later.
1768 struct in_addr *p, *q;
1773 m = m_get(MB_DONTWAIT, MT_HEADER);
1777 #define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt))
1779 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
1780 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) +
1784 kprintf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len);
1788 * First save first hop for return route
1790 p = &ip_srcrt.route[ip_nhops - 1];
1791 *(mtod(m, struct in_addr *)) = *p--;
1794 kprintf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr));
1798 * Copy option fields and padding (nop) to mbuf.
1800 ip_srcrt.nop = IPOPT_NOP;
1801 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
1802 memcpy(mtod(m, caddr_t) + sizeof(struct in_addr), &ip_srcrt.nop,
1804 q = (struct in_addr *)(mtod(m, caddr_t) +
1805 sizeof(struct in_addr) + OPTSIZ);
1808 * Record return path as an IP source route,
1809 * reversing the path (pointers are now aligned).
1811 while (p >= ip_srcrt.route) {
1814 kprintf(" %x", ntohl(q->s_addr));
1819 * Last hop goes to final destination.
1824 kprintf(" %x\n", ntohl(q->s_addr));
1830 * Strip out IP options.
1833 ip_stripoptions(struct mbuf *m)
1836 struct ip *ip = mtod(m, struct ip *);
1840 optlen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
1841 opts = (caddr_t)(ip + 1);
1842 datalen = m->m_len - (sizeof(struct ip) + optlen);
1843 bcopy(opts + optlen, opts, datalen);
1845 if (m->m_flags & M_PKTHDR)
1846 m->m_pkthdr.len -= optlen;
1847 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, sizeof(struct ip) >> 2);
1850 u_char inetctlerrmap[PRC_NCMDS] = {
1852 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
1853 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
1854 EMSGSIZE, EHOSTUNREACH, 0, 0,
1856 ENOPROTOOPT, ECONNREFUSED
1860 * Forward a packet. If some error occurs return the sender
1861 * an icmp packet. Note we can't always generate a meaningful
1862 * icmp message because icmp doesn't have a large enough repertoire
1863 * of codes and types.
1865 * If not forwarding, just drop the packet. This could be confusing
1866 * if ipforwarding was zero but some routing protocol was advancing
1867 * us as a gateway to somewhere. However, we must let the routing
1868 * protocol deal with that.
1870 * The using_srcrt parameter indicates whether the packet is being forwarded
1871 * via a source route.
1874 ip_forward(struct mbuf *m, boolean_t using_srcrt, struct sockaddr_in *next_hop)
1876 struct ip *ip = mtod(m, struct ip *);
1877 struct sockaddr_in *ipforward_rtaddr;
1879 int error, type = 0, code = 0, destmtu = 0;
1882 struct in_addr pkt_dst;
1884 struct route *cache_rt = &ipforward_rt[mycpuid];
1888 * Cache the destination address of the packet; this may be
1889 * changed by use of 'ipfw fwd'.
1891 pkt_dst = (next_hop != NULL) ? next_hop->sin_addr : ip->ip_dst;
1895 kprintf("forward: src %x dst %x ttl %x\n",
1896 ip->ip_src.s_addr, pkt_dst.s_addr, ip->ip_ttl);
1899 if (m->m_flags & (M_BCAST | M_MCAST) || !in_canforward(pkt_dst)) {
1900 ipstat.ips_cantforward++;
1904 if (!ipstealth && ip->ip_ttl <= IPTTLDEC) {
1905 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0);
1909 ipforward_rtaddr = (struct sockaddr_in *) &cache_rt->ro_dst;
1910 if (cache_rt->ro_rt == NULL ||
1911 ipforward_rtaddr->sin_addr.s_addr != pkt_dst.s_addr) {
1912 if (cache_rt->ro_rt != NULL) {
1913 RTFREE(cache_rt->ro_rt);
1914 cache_rt->ro_rt = NULL;
1916 ipforward_rtaddr->sin_family = AF_INET;
1917 ipforward_rtaddr->sin_len = sizeof(struct sockaddr_in);
1918 ipforward_rtaddr->sin_addr = pkt_dst;
1919 rtalloc_ign(cache_rt, RTF_PRCLONING);
1920 if (cache_rt->ro_rt == NULL) {
1921 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0);
1925 rt = cache_rt->ro_rt;
1928 * Save the IP header and at most 8 bytes of the payload,
1929 * in case we need to generate an ICMP message to the src.
1931 * XXX this can be optimized a lot by saving the data in a local
1932 * buffer on the stack (72 bytes at most), and only allocating the
1933 * mbuf if really necessary. The vast majority of the packets
1934 * are forwarded without having to send an ICMP back (either
1935 * because unnecessary, or because rate limited), so we are
1936 * really we are wasting a lot of work here.
1938 * We don't use m_copy() because it might return a reference
1939 * to a shared cluster. Both this function and ip_output()
1940 * assume exclusive access to the IP header in `m', so any
1941 * data in a cluster may change before we reach icmp_error().
1943 MGETHDR(mcopy, MB_DONTWAIT, m->m_type);
1944 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, MB_DONTWAIT)) {
1946 * It's probably ok if the pkthdr dup fails (because
1947 * the deep copy of the tag chain failed), but for now
1948 * be conservative and just discard the copy since
1949 * code below may some day want the tags.
1954 if (mcopy != NULL) {
1955 mcopy->m_len = imin((IP_VHL_HL(ip->ip_vhl) << 2) + 8,
1957 mcopy->m_pkthdr.len = mcopy->m_len;
1958 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1962 ip->ip_ttl -= IPTTLDEC;
1965 * If forwarding packet using same interface that it came in on,
1966 * perhaps should send a redirect to sender to shortcut a hop.
1967 * Only send redirect if source is sending directly to us,
1968 * and if packet was not source routed (or has any options).
1969 * Also, don't send redirect if forwarding using a default route
1970 * or a route modified by a redirect.
1972 if (rt->rt_ifp == m->m_pkthdr.rcvif &&
1973 !(rt->rt_flags & (RTF_DYNAMIC | RTF_MODIFIED)) &&
1974 satosin(rt_key(rt))->sin_addr.s_addr != INADDR_ANY &&
1975 ipsendredirects && !using_srcrt && next_hop == NULL) {
1976 u_long src = ntohl(ip->ip_src.s_addr);
1977 struct in_ifaddr *rt_ifa = (struct in_ifaddr *)rt->rt_ifa;
1979 if (rt_ifa != NULL &&
1980 (src & rt_ifa->ia_subnetmask) == rt_ifa->ia_subnet) {
1981 if (rt->rt_flags & RTF_GATEWAY)
1982 dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
1984 dest = pkt_dst.s_addr;
1986 * Router requirements says to only send
1989 type = ICMP_REDIRECT;
1990 code = ICMP_REDIRECT_HOST;
1993 kprintf("redirect (%d) to %x\n", code, dest);
1998 if (next_hop != NULL) {
1999 /* Pass IPFORWARD info if available */
2000 tag.mh_type = MT_TAG;
2001 tag.mh_flags = PACKET_TAG_IPFORWARD;
2002 tag.mh_data = (caddr_t)next_hop;
2004 m = (struct mbuf *)&tag;
2007 error = ip_output(m, NULL, cache_rt, IP_FORWARDING, NULL,
2010 ipstat.ips_forward++;
2013 ipflow_create(cache_rt, mcopy);
2016 return; /* most common case */
2018 ipstat.ips_redirectsent++;
2021 ipstat.ips_cantforward++;
2028 * Send ICMP message.
2033 case 0: /* forwarded, but need redirect */
2034 /* type, code set above */
2037 case ENETUNREACH: /* shouldn't happen, checked above */
2042 type = ICMP_UNREACH;
2043 code = ICMP_UNREACH_HOST;
2047 type = ICMP_UNREACH;
2048 code = ICMP_UNREACH_NEEDFRAG;
2051 * If the packet is routed over IPsec tunnel, tell the
2052 * originator the tunnel MTU.
2053 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
2056 if (cache_rt->ro_rt != NULL) {
2057 struct secpolicy *sp = NULL;
2062 sp = ipsec4_getpolicybyaddr(mcopy,
2068 destmtu = cache_rt->ro_rt->rt_ifp->if_mtu;
2070 /* count IPsec header size */
2071 ipsechdr = ipsec4_hdrsiz(mcopy,
2076 * find the correct route for outer IPv4
2077 * header, compute tunnel MTU.
2080 if (sp->req != NULL && sp->req->sav != NULL &&
2081 sp->req->sav->sah != NULL) {
2082 ro = &sp->req->sav->sah->sa_route;
2083 if (ro->ro_rt != NULL &&
2084 ro->ro_rt->rt_ifp != NULL) {
2086 ro->ro_rt->rt_ifp->if_mtu;
2087 destmtu -= ipsechdr;
2096 * If the packet is routed over IPsec tunnel, tell the
2097 * originator the tunnel MTU.
2098 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
2101 if (cache_rt->ro_rt != NULL) {
2102 struct secpolicy *sp = NULL;
2107 sp = ipsec_getpolicybyaddr(mcopy,
2113 destmtu = cache_rt->ro_rt->rt_ifp->if_mtu;
2115 /* count IPsec header size */
2116 ipsechdr = ipsec4_hdrsiz(mcopy,
2121 * find the correct route for outer IPv4
2122 * header, compute tunnel MTU.
2125 if (sp->req != NULL &&
2126 sp->req->sav != NULL &&
2127 sp->req->sav->sah != NULL) {
2128 ro = &sp->req->sav->sah->sa_route;
2129 if (ro->ro_rt != NULL &&
2130 ro->ro_rt->rt_ifp != NULL) {
2132 ro->ro_rt->rt_ifp->if_mtu;
2133 destmtu -= ipsechdr;
2140 #else /* !IPSEC && !FAST_IPSEC */
2141 if (cache_rt->ro_rt != NULL)
2142 destmtu = cache_rt->ro_rt->rt_ifp->if_mtu;
2144 ipstat.ips_cantfrag++;
2149 * A router should not generate ICMP_SOURCEQUENCH as
2150 * required in RFC1812 Requirements for IP Version 4 Routers.
2151 * Source quench could be a big problem under DoS attacks,
2152 * or if the underlying interface is rate-limited.
2153 * Those who need source quench packets may re-enable them
2154 * via the net.inet.ip.sendsourcequench sysctl.
2156 if (!ip_sendsourcequench) {
2160 type = ICMP_SOURCEQUENCH;
2165 case EACCES: /* ipfw denied packet */
2169 icmp_error(mcopy, type, code, dest, destmtu);
2173 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
2176 if (inp->inp_socket->so_options & SO_TIMESTAMP) {
2180 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
2181 SCM_TIMESTAMP, SOL_SOCKET);
2183 mp = &(*mp)->m_next;
2185 if (inp->inp_flags & INP_RECVDSTADDR) {
2186 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
2187 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
2189 mp = &(*mp)->m_next;
2193 * Moving these out of udp_input() made them even more broken
2194 * than they already were.
2196 /* options were tossed already */
2197 if (inp->inp_flags & INP_RECVOPTS) {
2198 *mp = sbcreatecontrol((caddr_t) opts_deleted_above,
2199 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
2201 mp = &(*mp)->m_next;
2203 /* ip_srcroute doesn't do what we want here, need to fix */
2204 if (inp->inp_flags & INP_RECVRETOPTS) {
2205 *mp = sbcreatecontrol((caddr_t) ip_srcroute(),
2206 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
2208 mp = &(*mp)->m_next;
2211 if (inp->inp_flags & INP_RECVIF) {
2214 struct sockaddr_dl sdl;
2217 struct sockaddr_dl *sdp;
2218 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
2220 if (((ifp = m->m_pkthdr.rcvif)) &&
2221 ((ifp->if_index != 0) && (ifp->if_index <= if_index))) {
2222 sdp = IF_LLSOCKADDR(ifp);
2224 * Change our mind and don't try copy.
2226 if ((sdp->sdl_family != AF_LINK) ||
2227 (sdp->sdl_len > sizeof(sdlbuf))) {
2230 bcopy(sdp, sdl2, sdp->sdl_len);
2234 offsetof(struct sockaddr_dl, sdl_data[0]);
2235 sdl2->sdl_family = AF_LINK;
2236 sdl2->sdl_index = 0;
2237 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
2239 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len,
2240 IP_RECVIF, IPPROTO_IP);
2242 mp = &(*mp)->m_next;
2247 * XXX these routines are called from the upper part of the kernel.
2249 * They could also be moved to ip_mroute.c, since all the RSVP
2250 * handling is done there already.
2253 ip_rsvp_init(struct socket *so)
2255 if (so->so_type != SOCK_RAW ||
2256 so->so_proto->pr_protocol != IPPROTO_RSVP)
2259 if (ip_rsvpd != NULL)
2264 * This may seem silly, but we need to be sure we don't over-increment
2265 * the RSVP counter, in case something slips up.
2280 * This may seem silly, but we need to be sure we don't over-decrement
2281 * the RSVP counter, in case something slips up.
2291 rsvp_input(struct mbuf *m, ...) /* XXX must fixup manually */
2297 off = __va_arg(ap, int);
2298 proto = __va_arg(ap, int);
2301 if (rsvp_input_p) { /* call the real one if loaded */
2302 rsvp_input_p(m, off, proto);
2306 /* Can still get packets with rsvp_on = 0 if there is a local member
2307 * of the group to which the RSVP packet is addressed. But in this
2308 * case we want to throw the packet away.
2316 if (ip_rsvpd != NULL) {
2317 rip_input(m, off, proto);
2320 /* Drop the packet */