2 * Copyright (c) 1982, 1986, 1988, 1993
3 * The Regents of the University of California. All rights reserved.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * 3. All advertising materials mentioning features or use of this software
14 * must display the following acknowledgement:
15 * This product includes software developed by the University of
16 * California, Berkeley and its contributors.
17 * 4. Neither the name of the University nor the names of its contributors
18 * may be used to endorse or promote products derived from this software
19 * without specific prior written permission.
21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94
34 * $FreeBSD: src/sys/netinet/ip_input.c,v 1.130.2.52 2003/03/07 07:01:28 silby Exp $
35 * $DragonFly: src/sys/netinet/ip_input.c,v 1.12 2004/03/06 07:30:43 hsu Exp $
40 #include "opt_bootp.h"
43 #include "opt_ipdivert.h"
44 #include "opt_ipfilter.h"
45 #include "opt_ipstealth.h"
46 #include "opt_ipsec.h"
47 #include "opt_pfil_hooks.h"
48 #include "opt_random_ip_id.h"
50 #include <sys/param.h>
51 #include <sys/systm.h>
53 #include <sys/malloc.h>
54 #include <sys/domain.h>
55 #include <sys/protosw.h>
56 #include <sys/socket.h>
58 #include <sys/kernel.h>
59 #include <sys/syslog.h>
60 #include <sys/sysctl.h>
61 #include <sys/in_cksum.h>
64 #include <net/if_types.h>
65 #include <net/if_var.h>
66 #include <net/if_dl.h>
70 #include <net/route.h>
71 #include <net/netisr.h>
72 #include <net/intrq.h>
74 #include <netinet/in.h>
75 #include <netinet/in_systm.h>
76 #include <netinet/in_var.h>
77 #include <netinet/ip.h>
78 #include <netinet/in_pcb.h>
79 #include <netinet/ip_var.h>
80 #include <netinet/ip_icmp.h>
82 #include <netinet/ipprotosw.h>
84 #include <sys/socketvar.h>
86 #include <net/ipfw/ip_fw.h>
87 #include <net/dummynet/ip_dummynet.h>
90 #include <netinet6/ipsec.h>
91 #include <netproto/key/key.h>
95 #include <netipsec/ipsec.h>
96 #include <netipsec/key.h>
100 static int ip_rsvp_on;
101 struct socket *ip_rsvpd;
103 int ipforwarding = 0;
104 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW,
105 &ipforwarding, 0, "Enable IP forwarding between interfaces");
107 static int ipsendredirects = 1; /* XXX */
108 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW,
109 &ipsendredirects, 0, "Enable sending IP redirects");
111 int ip_defttl = IPDEFTTL;
112 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW,
113 &ip_defttl, 0, "Maximum TTL on IP packets");
115 static int ip_dosourceroute = 0;
116 SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW,
117 &ip_dosourceroute, 0, "Enable forwarding source routed IP packets");
119 static int ip_acceptsourceroute = 0;
120 SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute,
121 CTLFLAG_RW, &ip_acceptsourceroute, 0,
122 "Enable accepting source routed IP packets");
124 static int ip_keepfaith = 0;
125 SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW,
127 "Enable packet capture for FAITH IPv4->IPv6 translater daemon");
129 static int nipq = 0; /* total # of reass queues */
131 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragpackets, CTLFLAG_RW,
133 "Maximum number of IPv4 fragment reassembly queue entries");
135 static int maxfragsperpacket;
136 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW,
137 &maxfragsperpacket, 0,
138 "Maximum number of IPv4 fragments allowed per packet");
140 static int ip_sendsourcequench = 0;
141 SYSCTL_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW,
142 &ip_sendsourcequench, 0,
143 "Enable the transmission of source quench packets");
146 * XXX - Setting ip_checkinterface mostly implements the receive side of
147 * the Strong ES model described in RFC 1122, but since the routing table
148 * and transmit implementation do not implement the Strong ES model,
149 * setting this to 1 results in an odd hybrid.
151 * XXX - ip_checkinterface currently must be disabled if you use ipnat
152 * to translate the destination address to another local interface.
154 * XXX - ip_checkinterface must be disabled if you add IP aliases
155 * to the loopback interface instead of the interface where the
156 * packets for those addresses are received.
158 static int ip_checkinterface = 0;
159 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW,
160 &ip_checkinterface, 0, "Verify packet arrives on correct interface");
163 static int ipprintfs = 0;
166 static struct ifqueue ipintrq;
167 static int ipqmaxlen = IFQ_MAXLEN;
169 extern struct domain inetdomain;
170 extern struct ipprotosw inetsw[];
171 u_char ip_protox[IPPROTO_MAX];
172 struct in_ifaddrhead in_ifaddrhead; /* first inet address */
173 struct in_ifaddrhashhead *in_ifaddrhashtbl; /* inet addr hash table */
174 u_long in_ifaddrhmask; /* mask for hash table */
175 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW,
176 &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue");
177 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD,
178 &ipintrq.ifq_drops, 0, "Number of packets dropped from the IP input queue");
180 struct ipstat ipstat;
181 SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW,
182 &ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)");
184 /* Packet reassembly stuff */
185 #define IPREASS_NHASH_LOG2 6
186 #define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2)
187 #define IPREASS_HMASK (IPREASS_NHASH - 1)
188 #define IPREASS_HASH(x,y) \
189 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK)
191 static struct ipq ipq[IPREASS_NHASH];
192 const int ipintrq_present = 1;
195 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
196 &ip_mtu, 0, "Default MTU");
200 static int ipstealth = 0;
201 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW,
207 ip_fw_chk_t *ip_fw_chk_ptr;
212 ip_dn_io_t *ip_dn_io_ptr;
215 struct pfil_head inet_pfil_hook;
216 #endif /* PFIL_HOOKS */
219 * XXX this is ugly -- the following two global variables are
220 * used to store packet state while it travels through the stack.
221 * Note that the code even makes assumptions on the size and
222 * alignment of fields inside struct ip_srcrt so e.g. adding some
223 * fields will break the code. This needs to be fixed.
225 * We need to save the IP options in case a protocol wants to respond
226 * to an incoming packet over the same route if the packet got here
227 * using IP source routing. This allows connection establishment and
228 * maintenance when the remote end is on a network that is not known
231 static int ip_nhops = 0;
232 static struct ip_srcrt {
233 struct in_addr dst; /* final destination */
234 char nop; /* one NOP to align */
235 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */
236 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)];
239 static void save_rte(u_char *, struct in_addr);
240 static int ip_dooptions(struct mbuf *m, int,
241 struct sockaddr_in *next_hop);
242 static void ip_forward(struct mbuf *m, int srcrt,
243 struct sockaddr_in *next_hop);
244 static void ip_freef(struct ipq *);
245 static struct mbuf *ip_reass(struct mbuf *, struct ipq *,
246 struct ipq *, u_int32_t *, u_int16_t *);
249 * IP initialization: fill in IP protocol switch table.
250 * All protocols not implemented in kernel go to raw IP protocol handler.
255 struct ipprotosw *pr;
258 TAILQ_INIT(&in_ifaddrhead);
259 in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask);
260 pr = (struct ipprotosw *)pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
263 for (i = 0; i < IPPROTO_MAX; i++)
264 ip_protox[i] = pr - inetsw;
265 for (pr = (struct ipprotosw *)inetdomain.dom_protosw;
266 pr < (struct ipprotosw *)inetdomain.dom_protoswNPROTOSW; pr++)
267 if (pr->pr_domain->dom_family == PF_INET &&
268 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
269 ip_protox[pr->pr_protocol] = pr - inetsw;
272 inet_pfil_hook.ph_type = PFIL_TYPE_AF;
273 inet_pfil_hook.ph_af = AF_INET;
274 if ((i = pfil_head_register(&inet_pfil_hook)) != 0)
275 printf("%s: WARNING: unable to register pfil hook, "
276 "error %d\n", __func__, i);
277 #endif /* PFIL_HOOKS */
279 for (i = 0; i < IPREASS_NHASH; i++)
280 ipq[i].next = ipq[i].prev = &ipq[i];
282 maxnipq = nmbclusters / 32;
283 maxfragsperpacket = 16;
286 ip_id = time_second & 0xffff;
288 ipintrq.ifq_maxlen = ipqmaxlen;
290 netisr_register(NETISR_IP, ip_mport, ip_input);
294 * XXX watch out this one. It is perhaps used as a cache for
295 * the most recently used route ? it is cleared in in_addroute()
296 * when a new route is successfully created.
298 struct route ipforward_rt;
299 static struct sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET };
302 * Ip input routine. Checksum and byte swap header. If fragmented
303 * try to reassemble. Process options. Pass to next level.
306 ip_input(struct netmsg *msg)
308 struct mbuf *m = ((struct netmsg_packet *)msg)->nm_packet;
311 struct in_ifaddr *ia = NULL;
313 int i, hlen, checkif;
315 struct in_addr pkt_dst;
316 u_int32_t divert_info = 0; /* packet divert/tee info */
317 struct ip_fw_args args;
318 int srcrt = 0; /* forward (by PFIL_HOOKS) */
320 struct in_addr odst; /* original dst address(NAT) */
324 struct tdb_ident *tdbi;
325 struct secpolicy *sp;
327 #endif /* FAST_IPSEC */
332 args.divert_rule = 0; /* divert cookie */
333 args.next_hop = NULL;
335 /* Grab info from MT_TAG mbufs prepended to the chain. */
336 for (; m && m->m_type == MT_TAG; m = m->m_next) {
337 switch(m->_m_tag_id) {
339 printf("ip_input: unrecognised MT_TAG tag %d\n",
343 case PACKET_TAG_DUMMYNET:
344 args.rule = ((struct dn_pkt *)m)->rule;
347 case PACKET_TAG_DIVERT:
348 args.divert_rule = (int)m->m_hdr.mh_data & 0xffff;
351 case PACKET_TAG_IPFORWARD:
352 args.next_hop = (struct sockaddr_in *)m->m_hdr.mh_data;
357 KASSERT(m != NULL && (m->m_flags & M_PKTHDR) != 0,
358 ("ip_input: no HDR"));
360 if (args.rule) { /* dummynet already filtered us */
361 ip = mtod(m, struct ip *);
362 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
368 if (m->m_pkthdr.len < sizeof(struct ip))
371 if (m->m_len < sizeof (struct ip) &&
372 (m = m_pullup(m, sizeof (struct ip))) == 0) {
373 ipstat.ips_toosmall++;
376 ip = mtod(m, struct ip *);
378 if (IP_VHL_V(ip->ip_vhl) != IPVERSION) {
379 ipstat.ips_badvers++;
383 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
384 if (hlen < sizeof(struct ip)) { /* minimum header length */
385 ipstat.ips_badhlen++;
388 if (hlen > m->m_len) {
389 if ((m = m_pullup(m, hlen)) == 0) {
390 ipstat.ips_badhlen++;
393 ip = mtod(m, struct ip *);
396 /* 127/8 must not appear on wire - RFC1122 */
397 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
398 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
399 if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
400 ipstat.ips_badaddr++;
405 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
406 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
408 if (hlen == sizeof(struct ip)) {
409 sum = in_cksum_hdr(ip);
411 sum = in_cksum(m, hlen);
420 * Convert fields to host representation.
422 ip->ip_len = ntohs(ip->ip_len);
423 if (ip->ip_len < hlen) {
427 ip->ip_off = ntohs(ip->ip_off);
430 * Check that the amount of data in the buffers
431 * is as at least much as the IP header would have us expect.
432 * Trim mbufs if longer than we expect.
433 * Drop packet if shorter than we expect.
435 if (m->m_pkthdr.len < ip->ip_len) {
437 ipstat.ips_tooshort++;
440 if (m->m_pkthdr.len > ip->ip_len) {
441 if (m->m_len == m->m_pkthdr.len) {
442 m->m_len = ip->ip_len;
443 m->m_pkthdr.len = ip->ip_len;
445 m_adj(m, ip->ip_len - m->m_pkthdr.len);
447 #if defined(IPSEC) && !defined(IPSEC_FILTERGIF)
449 * Bypass packet filtering for packets from a tunnel (gif).
451 if (ipsec_gethist(m, NULL))
457 * Right now when no processing on packet has done
458 * and it is still fresh out of network we do our black
460 * - Firewall: deny/allow/divert
461 * - Xlate: translate packet's addr/port (NAT).
462 * - Pipe: pass pkt through dummynet.
463 * - Wrap: fake packet's addr/port <unimpl.>
464 * - Encapsulate: put it in another IP and send out. <unimp.>
470 * Run through list of hooks for input packets.
472 * NB: Beware of the destination address changing (e.g.
473 * by NAT rewriting). When this happens, tell
474 * ip_forward to do the right thing.
477 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif,
480 if (m == NULL) /* consumed by filter */
482 ip = mtod(m, struct ip *);
483 srcrt = (odst.s_addr != ip->ip_dst.s_addr);
484 #endif /* PFIL_HOOKS */
485 if (fw_enable && IPFW_LOADED) {
487 * If we've been forwarded from the output side, then
488 * skip the firewall a second time
494 i = ip_fw_chk_ptr(&args);
497 if ( (i & IP_FW_PORT_DENY_FLAG) || m == NULL) { /* drop */
502 ip = mtod(m, struct ip *); /* just in case m changed */
503 if (i == 0 && args.next_hop == NULL) /* common case */
505 if (DUMMYNET_LOADED && (i & IP_FW_PORT_DYNT_FLAG) != 0) {
506 /* Send packet to the appropriate pipe */
507 ip_dn_io_ptr(m, i&0xffff, DN_TO_IP_IN, &args);
511 if (i != 0 && (i & IP_FW_PORT_DYNT_FLAG) == 0) {
512 /* Divert or tee packet */
517 if (i == 0 && args.next_hop != NULL)
520 * if we get here, the packet must be dropped
528 * Process options and, if not destined for us,
529 * ship it on. ip_dooptions returns 1 when an
530 * error was detected (causing an icmp message
531 * to be sent and the original packet to be freed).
533 ip_nhops = 0; /* for source routed packets */
534 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0, args.next_hop))
537 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
538 * matter if it is destined to another node, or whether it is
539 * a multicast one, RSVP wants it! and prevents it from being forwarded
540 * anywhere else. Also checks if the rsvp daemon is running before
541 * grabbing the packet.
543 if (rsvp_on && ip->ip_p==IPPROTO_RSVP)
547 * Check our list of addresses, to see if the packet is for us.
548 * If we don't have any addresses, assume any unicast packet
549 * we receive might be for us (and let the upper layers deal
552 if (TAILQ_EMPTY(&in_ifaddrhead) &&
553 (m->m_flags & (M_MCAST|M_BCAST)) == 0)
557 * Cache the destination address of the packet; this may be
558 * changed by use of 'ipfw fwd'.
560 pkt_dst = args.next_hop ? args.next_hop->sin_addr : ip->ip_dst;
563 * Enable a consistency check between the destination address
564 * and the arrival interface for a unicast packet (the RFC 1122
565 * strong ES model) if IP forwarding is disabled and the packet
566 * is not locally generated and the packet is not subject to
569 * XXX - Checking also should be disabled if the destination
570 * address is ipnat'ed to a different interface.
572 * XXX - Checking is incompatible with IP aliases added
573 * to the loopback interface instead of the interface where
574 * the packets are received.
576 checkif = ip_checkinterface && (ipforwarding == 0) &&
577 m->m_pkthdr.rcvif != NULL &&
578 ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) &&
579 (args.next_hop == NULL);
582 * Check for exact addresses in the hash bucket.
584 LIST_FOREACH(ia, INADDR_HASH(pkt_dst.s_addr), ia_hash) {
586 * If the address matches, verify that the packet
587 * arrived via the correct interface if checking is
590 if (IA_SIN(ia)->sin_addr.s_addr == pkt_dst.s_addr &&
591 (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif))
595 * Check for broadcast addresses.
597 * Only accept broadcast packets that arrive via the matching
598 * interface. Reception of forwarded directed broadcasts would
599 * be handled via ip_forward() and ether_output() with the loopback
600 * into the stack for SIMPLEX interfaces handled by ether_output().
602 if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
603 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
604 if (ifa->ifa_addr->sa_family != AF_INET)
607 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
610 if (ia->ia_netbroadcast.s_addr == pkt_dst.s_addr)
613 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY)
618 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
619 struct in_multi *inm;
622 * If we are acting as a multicast router, all
623 * incoming multicast packets are passed to the
624 * kernel-level multicast forwarding function.
625 * The packet is returned (relatively) intact; if
626 * ip_mforward() returns a non-zero value, the packet
627 * must be discarded, else it may be accepted below.
630 ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) {
631 ipstat.ips_cantforward++;
637 * The process-level routing daemon needs to receive
638 * all multicast IGMP packets, whether or not this
639 * host belongs to their destination groups.
641 if (ip->ip_p == IPPROTO_IGMP)
643 ipstat.ips_forward++;
646 * See if we belong to the destination multicast group on the
649 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
651 ipstat.ips_notmember++;
657 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
659 if (ip->ip_dst.s_addr == INADDR_ANY)
663 * FAITH(Firewall Aided Internet Translator)
665 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
667 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP)
675 * Not for us; forward if possible and desirable.
677 if (ipforwarding == 0) {
678 ipstat.ips_cantforward++;
683 * Enforce inbound IPsec SPD.
685 if (ipsec4_in_reject(m, NULL)) {
686 ipsecstat.in_polvio++;
691 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
694 tdbi = (struct tdb_ident *)(mtag + 1);
695 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
697 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
698 IP_FORWARDING, &error);
700 if (sp == NULL) { /* NB: can happen if error */
702 /*XXX error stat???*/
703 DPRINTF(("ip_input: no SP for forwarding\n")); /*XXX*/
708 * Check security policy against packet attributes.
710 error = ipsec_in_reject(sp, m);
714 ipstat.ips_cantforward++;
717 #endif /* FAST_IPSEC */
718 ip_forward(m, srcrt, args.next_hop);
725 * IPSTEALTH: Process non-routing options only
726 * if the packet is destined for us.
728 if (ipstealth && hlen > sizeof (struct ip) &&
729 ip_dooptions(m, 1, args.next_hop))
731 #endif /* IPSTEALTH */
733 /* Count the packet in the ip address stats */
735 ia->ia_ifa.if_ipackets++;
736 ia->ia_ifa.if_ibytes += m->m_pkthdr.len;
740 * If offset or IP_MF are set, must reassemble.
741 * Otherwise, nothing need be done.
742 * (We could look in the reassembly queue to see
743 * if the packet was previously fragmented,
744 * but it's not worth the time; just let them time out.)
746 if (ip->ip_off & (IP_MF | IP_OFFMASK)) {
748 /* If maxnipq is 0, never accept fragments. */
750 ipstat.ips_fragments++;
751 ipstat.ips_fragdropped++;
755 sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
757 * Look for queue of fragments
760 for (fp = ipq[sum].next; fp != &ipq[sum]; fp = fp->next)
761 if (ip->ip_id == fp->ipq_id &&
762 ip->ip_src.s_addr == fp->ipq_src.s_addr &&
763 ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
764 ip->ip_p == fp->ipq_p)
770 * Enforce upper bound on number of fragmented packets
771 * for which we attempt reassembly;
772 * If maxnipq is -1, accept all fragments without limitation.
774 if ((nipq > maxnipq) && (maxnipq > 0)) {
776 * drop something from the tail of the current queue
777 * before proceeding further
779 if (ipq[sum].prev == &ipq[sum]) { /* gak */
780 for (i = 0; i < IPREASS_NHASH; i++) {
781 if (ipq[i].prev != &ipq[i]) {
782 ipstat.ips_fragtimeout +=
783 ipq[i].prev->ipq_nfrags;
784 ip_freef(ipq[i].prev);
789 ipstat.ips_fragtimeout += ipq[sum].prev->ipq_nfrags;
790 ip_freef(ipq[sum].prev);
795 * Adjust ip_len to not reflect header,
796 * convert offset of this to bytes.
799 if (ip->ip_off & IP_MF) {
801 * Make sure that fragments have a data length
802 * that's a non-zero multiple of 8 bytes.
804 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
805 ipstat.ips_toosmall++; /* XXX */
808 m->m_flags |= M_FRAG;
810 m->m_flags &= ~M_FRAG;
814 * Attempt reassembly; if it succeeds, proceed.
815 * ip_reass() will return a different mbuf, and update
816 * the divert info in divert_info and args.divert_rule.
818 ipstat.ips_fragments++;
819 m->m_pkthdr.header = ip;
821 fp, &ipq[sum], &divert_info, &args.divert_rule);
824 ipstat.ips_reassembled++;
825 ip = mtod(m, struct ip *);
826 /* Get the header length of the reassembled packet */
827 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
829 /* Restore original checksum before diverting packet */
830 if (divert_info != 0) {
832 ip->ip_len = htons(ip->ip_len);
833 ip->ip_off = htons(ip->ip_off);
835 if (hlen == sizeof(struct ip))
836 ip->ip_sum = in_cksum_hdr(ip);
838 ip->ip_sum = in_cksum(m, hlen);
839 ip->ip_off = ntohs(ip->ip_off);
840 ip->ip_len = ntohs(ip->ip_len);
850 * Divert or tee packet to the divert protocol if required.
852 if (divert_info != 0) {
853 struct mbuf *clone = NULL;
855 /* Clone packet if we're doing a 'tee' */
856 if ((divert_info & IP_FW_PORT_TEE_FLAG) != 0)
857 clone = m_dup(m, M_DONTWAIT);
859 /* Restore packet header fields to original values */
861 ip->ip_len = htons(ip->ip_len);
862 ip->ip_off = htons(ip->ip_off);
864 /* Deliver packet to divert input routine */
865 divert_packet(m, 1, divert_info & 0xffff, args.divert_rule);
866 ipstat.ips_delivered++;
868 /* If 'tee', continue with original packet */
872 ip = mtod(m, struct ip *);
875 * Jump backwards to complete processing of the
876 * packet. But first clear divert_info to avoid
877 * entering this block again.
878 * We do not need to clear args.divert_rule
879 * or args.next_hop as they will not be used.
888 * enforce IPsec policy checking if we are seeing last header.
889 * note that we do not visit this with protocols with pcb layer
890 * code - like udp/tcp/raw ip.
892 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0 &&
893 ipsec4_in_reject(m, NULL)) {
894 ipsecstat.in_polvio++;
900 * enforce IPsec policy checking if we are seeing last header.
901 * note that we do not visit this with protocols with pcb layer
902 * code - like udp/tcp/raw ip.
904 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) {
906 * Check if the packet has already had IPsec processing
907 * done. If so, then just pass it along. This tag gets
908 * set during AH, ESP, etc. input handling, before the
909 * packet is returned to the ip input queue for delivery.
911 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
914 tdbi = (struct tdb_ident *)(mtag + 1);
915 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
917 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
918 IP_FORWARDING, &error);
922 * Check security policy against packet attributes.
924 error = ipsec_in_reject(sp, m);
927 /* XXX error stat??? */
929 DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/
936 #endif /* FAST_IPSEC */
939 * Switch out to protocol's input routine.
941 * XXX queue packet to protocol's message port.
943 ipstat.ips_delivered++;
944 if (args.next_hop && ip->ip_p == IPPROTO_TCP) {
945 /* TCP needs IPFORWARD info if available */
948 tag.mh_type = MT_TAG;
949 tag.mh_flags = PACKET_TAG_IPFORWARD;
950 tag.mh_data = (caddr_t)args.next_hop;
953 (*inetsw[ip_protox[ip->ip_p]].pr_input)(
954 (struct mbuf *)&tag, hlen, ip->ip_p);
956 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen, ip->ip_p);
964 * Take incoming datagram fragment and try to reassemble it into
965 * whole datagram. If a chain for reassembly of this datagram already
966 * exists, then it is given as fp; otherwise have to make a chain.
968 * When IPDIVERT enabled, keep additional state with each packet that
969 * tells us if we need to divert or tee the packet we're building.
970 * In particular, *divinfo includes the port and TEE flag,
971 * *divert_rule is the number of the matching rule.
975 ip_reass(struct mbuf *m, struct ipq *fp, struct ipq *where,
976 u_int32_t *divinfo, u_int16_t *divert_rule)
978 struct ip *ip = mtod(m, struct ip *);
979 struct mbuf *p = 0, *q, *nq;
981 int hlen = IP_VHL_HL(ip->ip_vhl) << 2;
985 * Presence of header sizes in mbufs
986 * would confuse code below.
992 * If first fragment to arrive, create a reassembly queue.
995 if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL)
997 fp = mtod(t, struct ipq *);
1001 fp->ipq_ttl = IPFRAGTTL;
1002 fp->ipq_p = ip->ip_p;
1003 fp->ipq_id = ip->ip_id;
1004 fp->ipq_src = ip->ip_src;
1005 fp->ipq_dst = ip->ip_dst;
1007 m->m_nextpkt = NULL;
1009 fp->ipq_div_info = 0;
1010 fp->ipq_div_cookie = 0;
1017 #define GETIP(m) ((struct ip*)((m)->m_pkthdr.header))
1020 * Find a segment which begins after this one does.
1022 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt)
1023 if (GETIP(q)->ip_off > ip->ip_off)
1027 * If there is a preceding segment, it may provide some of
1028 * our data already. If so, drop the data from the incoming
1029 * segment. If it provides all of our data, drop us, otherwise
1030 * stick new segment in the proper place.
1032 * If some of the data is dropped from the the preceding
1033 * segment, then it's checksum is invalidated.
1036 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off;
1038 if (i >= ip->ip_len)
1041 m->m_pkthdr.csum_flags = 0;
1045 m->m_nextpkt = p->m_nextpkt;
1048 m->m_nextpkt = fp->ipq_frags;
1053 * While we overlap succeeding segments trim them or,
1054 * if they are completely covered, dequeue them.
1056 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off;
1058 i = (ip->ip_off + ip->ip_len) -
1060 if (i < GETIP(q)->ip_len) {
1061 GETIP(q)->ip_len -= i;
1062 GETIP(q)->ip_off += i;
1064 q->m_pkthdr.csum_flags = 0;
1069 ipstat.ips_fragdropped++;
1078 * Transfer firewall instructions to the fragment structure.
1079 * Only trust info in the fragment at offset 0.
1081 if (ip->ip_off == 0) {
1082 fp->ipq_div_info = *divinfo;
1083 fp->ipq_div_cookie = *divert_rule;
1090 * Check for complete reassembly and perform frag per packet
1093 * Frag limiting is performed here so that the nth frag has
1094 * a chance to complete the packet before we drop the packet.
1095 * As a result, n+1 frags are actually allowed per packet, but
1096 * only n will ever be stored. (n = maxfragsperpacket.)
1100 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
1101 if (GETIP(q)->ip_off != next) {
1102 if (fp->ipq_nfrags > maxfragsperpacket) {
1103 ipstat.ips_fragdropped += fp->ipq_nfrags;
1108 next += GETIP(q)->ip_len;
1110 /* Make sure the last packet didn't have the IP_MF flag */
1111 if (p->m_flags & M_FRAG) {
1112 if (fp->ipq_nfrags > maxfragsperpacket) {
1113 ipstat.ips_fragdropped += fp->ipq_nfrags;
1120 * Reassembly is complete. Make sure the packet is a sane size.
1124 if (next + (IP_VHL_HL(ip->ip_vhl) << 2) > IP_MAXPACKET) {
1125 ipstat.ips_toolong++;
1126 ipstat.ips_fragdropped += fp->ipq_nfrags;
1132 * Concatenate fragments.
1140 for (q = nq; q != NULL; q = nq) {
1142 q->m_nextpkt = NULL;
1143 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
1144 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
1150 * Extract firewall instructions from the fragment structure.
1152 *divinfo = fp->ipq_div_info;
1153 *divert_rule = fp->ipq_div_cookie;
1157 * Create header for new ip packet by
1158 * modifying header of first packet;
1159 * dequeue and discard fragment reassembly header.
1160 * Make header visible.
1163 ip->ip_src = fp->ipq_src;
1164 ip->ip_dst = fp->ipq_dst;
1167 (void) m_free(dtom(fp));
1168 m->m_len += (IP_VHL_HL(ip->ip_vhl) << 2);
1169 m->m_data -= (IP_VHL_HL(ip->ip_vhl) << 2);
1170 /* some debugging cruft by sklower, below, will go away soon */
1171 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */
1173 for (t = m; t; t = t->m_next)
1175 m->m_pkthdr.len = plen;
1184 ipstat.ips_fragdropped++;
1194 * Free a fragment reassembly header and all
1195 * associated datagrams.
1203 while (fp->ipq_frags) {
1205 fp->ipq_frags = q->m_nextpkt;
1209 (void) m_free(dtom(fp));
1214 * IP timer processing;
1215 * if a timer expires on a reassembly
1216 * queue, discard it.
1225 for (i = 0; i < IPREASS_NHASH; i++) {
1229 while (fp != &ipq[i]) {
1232 if (fp->prev->ipq_ttl == 0) {
1233 ipstat.ips_fragtimeout += fp->prev->ipq_nfrags;
1239 * If we are over the maximum number of fragments
1240 * (due to the limit being lowered), drain off
1241 * enough to get down to the new limit.
1243 if (maxnipq >= 0 && nipq > maxnipq) {
1244 for (i = 0; i < IPREASS_NHASH; i++) {
1245 while (nipq > maxnipq &&
1246 (ipq[i].next != &ipq[i])) {
1247 ipstat.ips_fragdropped +=
1248 ipq[i].next->ipq_nfrags;
1249 ip_freef(ipq[i].next);
1258 * Drain off all datagram fragments.
1265 for (i = 0; i < IPREASS_NHASH; i++) {
1266 while (ipq[i].next != &ipq[i]) {
1267 ipstat.ips_fragdropped += ipq[i].next->ipq_nfrags;
1268 ip_freef(ipq[i].next);
1275 * Do option processing on a datagram,
1276 * possibly discarding it if bad options are encountered,
1277 * or forwarding it if source-routed.
1278 * The pass argument is used when operating in the IPSTEALTH
1279 * mode to tell what options to process:
1280 * [LS]SRR (pass 0) or the others (pass 1).
1281 * The reason for as many as two passes is that when doing IPSTEALTH,
1282 * non-routing options should be processed only if the packet is for us.
1283 * Returns 1 if packet has been forwarded/freed,
1284 * 0 if the packet should be processed further.
1287 ip_dooptions(struct mbuf *m, int pass, struct sockaddr_in *next_hop)
1289 struct ip *ip = mtod(m, struct ip *);
1291 struct in_ifaddr *ia;
1292 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0;
1293 struct in_addr *sin, dst;
1297 cp = (u_char *)(ip + 1);
1298 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip);
1299 for (; cnt > 0; cnt -= optlen, cp += optlen) {
1300 opt = cp[IPOPT_OPTVAL];
1301 if (opt == IPOPT_EOL)
1303 if (opt == IPOPT_NOP)
1306 if (cnt < IPOPT_OLEN + sizeof(*cp)) {
1307 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1310 optlen = cp[IPOPT_OLEN];
1311 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
1312 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1322 * Source routing with record.
1323 * Find interface with current destination address.
1324 * If none on this machine then drop if strictly routed,
1325 * or do nothing if loosely routed.
1326 * Record interface address and bring up next address
1327 * component. If strictly routed make sure next
1328 * address is on directly accessible net.
1333 if (ipstealth && pass > 0)
1336 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1337 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1340 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1341 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1344 ipaddr.sin_addr = ip->ip_dst;
1345 ia = (struct in_ifaddr *)
1346 ifa_ifwithaddr((struct sockaddr *)&ipaddr);
1348 if (opt == IPOPT_SSRR) {
1349 type = ICMP_UNREACH;
1350 code = ICMP_UNREACH_SRCFAIL;
1353 if (!ip_dosourceroute)
1354 goto nosourcerouting;
1356 * Loose routing, and not at next destination
1357 * yet; nothing to do except forward.
1361 off--; /* 0 origin */
1362 if (off > optlen - (int)sizeof(struct in_addr)) {
1364 * End of source route. Should be for us.
1366 if (!ip_acceptsourceroute)
1367 goto nosourcerouting;
1368 save_rte(cp, ip->ip_src);
1375 if (!ip_dosourceroute) {
1377 char buf[16]; /* aaa.bbb.ccc.ddd\0 */
1379 * Acting as a router, so generate ICMP
1382 strcpy(buf, inet_ntoa(ip->ip_dst));
1384 "attempted source route from %s to %s\n",
1385 inet_ntoa(ip->ip_src), buf);
1386 type = ICMP_UNREACH;
1387 code = ICMP_UNREACH_SRCFAIL;
1391 * Not acting as a router, so silently drop.
1396 ipstat.ips_cantforward++;
1403 * locate outgoing interface
1405 (void)memcpy(&ipaddr.sin_addr, cp + off,
1406 sizeof(ipaddr.sin_addr));
1408 if (opt == IPOPT_SSRR) {
1409 #define INA struct in_ifaddr *
1410 #define SA struct sockaddr *
1411 if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == 0)
1412 ia = (INA)ifa_ifwithnet((SA)&ipaddr);
1414 ia = ip_rtaddr(ipaddr.sin_addr, &ipforward_rt);
1416 type = ICMP_UNREACH;
1417 code = ICMP_UNREACH_SRCFAIL;
1420 ip->ip_dst = ipaddr.sin_addr;
1421 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
1422 sizeof(struct in_addr));
1423 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1425 * Let ip_intr's mcast routing check handle mcast pkts
1427 forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr));
1432 if (ipstealth && pass == 0)
1435 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1436 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1439 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1440 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1444 * If no space remains, ignore.
1446 off--; /* 0 origin */
1447 if (off > optlen - (int)sizeof(struct in_addr))
1449 (void)memcpy(&ipaddr.sin_addr, &ip->ip_dst,
1450 sizeof(ipaddr.sin_addr));
1452 * locate outgoing interface; if we're the destination,
1453 * use the incoming interface (should be same).
1455 if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == 0 &&
1456 (ia = ip_rtaddr(ipaddr.sin_addr,
1457 &ipforward_rt)) == 0) {
1458 type = ICMP_UNREACH;
1459 code = ICMP_UNREACH_HOST;
1462 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
1463 sizeof(struct in_addr));
1464 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1469 if (ipstealth && pass == 0)
1472 code = cp - (u_char *)ip;
1473 if (optlen < 4 || optlen > 40) {
1474 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1477 if ((off = cp[IPOPT_OFFSET]) < 5) {
1478 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1481 if (off > optlen - (int)sizeof(int32_t)) {
1482 cp[IPOPT_OFFSET + 1] += (1 << 4);
1483 if ((cp[IPOPT_OFFSET + 1] & 0xf0) == 0) {
1484 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1489 off--; /* 0 origin */
1490 sin = (struct in_addr *)(cp + off);
1491 switch (cp[IPOPT_OFFSET + 1] & 0x0f) {
1493 case IPOPT_TS_TSONLY:
1496 case IPOPT_TS_TSANDADDR:
1497 if (off + sizeof(n_time) +
1498 sizeof(struct in_addr) > optlen) {
1499 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1502 ipaddr.sin_addr = dst;
1503 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr,
1507 (void)memcpy(sin, &IA_SIN(ia)->sin_addr,
1508 sizeof(struct in_addr));
1509 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1510 off += sizeof(struct in_addr);
1513 case IPOPT_TS_PRESPEC:
1514 if (off + sizeof(n_time) +
1515 sizeof(struct in_addr) > optlen) {
1516 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1519 (void)memcpy(&ipaddr.sin_addr, sin,
1520 sizeof(struct in_addr));
1521 if (ifa_ifwithaddr((SA)&ipaddr) == 0)
1523 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1524 off += sizeof(struct in_addr);
1528 code = &cp[IPOPT_OFFSET + 1] - (u_char *)ip;
1532 (void)memcpy(cp + off, &ntime, sizeof(n_time));
1533 cp[IPOPT_OFFSET] += sizeof(n_time);
1536 if (forward && ipforwarding) {
1537 ip_forward(m, 1, next_hop);
1542 icmp_error(m, type, code, 0, 0);
1543 ipstat.ips_badoptions++;
1548 * Given address of next destination (final or next hop),
1549 * return internet address info of interface to be used to get there.
1556 struct sockaddr_in *sin;
1558 sin = (struct sockaddr_in *)&rt->ro_dst;
1560 if (rt->ro_rt == 0 ||
1561 dst.s_addr != sin->sin_addr.s_addr) {
1566 sin->sin_family = AF_INET;
1567 sin->sin_len = sizeof(*sin);
1568 sin->sin_addr = dst;
1570 rtalloc_ign(rt, RTF_PRCLONING);
1573 return ((struct in_ifaddr *)0);
1574 return (ifatoia(rt->ro_rt->rt_ifa));
1578 * Save incoming source route for use in replies,
1579 * to be picked up later by ip_srcroute if the receiver is interested.
1582 save_rte(option, dst)
1588 olen = option[IPOPT_OLEN];
1591 printf("save_rte: olen %d\n", olen);
1593 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst)))
1595 bcopy(option, ip_srcrt.srcopt, olen);
1596 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1601 * Retrieve incoming source route for use in replies,
1602 * in the same form used by setsockopt.
1603 * The first hop is placed before the options, will be removed later.
1608 struct in_addr *p, *q;
1612 return ((struct mbuf *)0);
1613 m = m_get(M_DONTWAIT, MT_HEADER);
1615 return ((struct mbuf *)0);
1617 #define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt))
1619 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
1620 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) +
1624 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len);
1628 * First save first hop for return route
1630 p = &ip_srcrt.route[ip_nhops - 1];
1631 *(mtod(m, struct in_addr *)) = *p--;
1634 printf(" hops %lx", (u_long)ntohl(mtod(m, struct in_addr *)->s_addr));
1638 * Copy option fields and padding (nop) to mbuf.
1640 ip_srcrt.nop = IPOPT_NOP;
1641 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
1642 (void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr),
1643 &ip_srcrt.nop, OPTSIZ);
1644 q = (struct in_addr *)(mtod(m, caddr_t) +
1645 sizeof(struct in_addr) + OPTSIZ);
1648 * Record return path as an IP source route,
1649 * reversing the path (pointers are now aligned).
1651 while (p >= ip_srcrt.route) {
1654 printf(" %lx", (u_long)ntohl(q->s_addr));
1659 * Last hop goes to final destination.
1664 printf(" %lx\n", (u_long)ntohl(q->s_addr));
1670 * Strip out IP options.
1673 ip_stripoptions(struct mbuf *m)
1676 struct ip *ip = mtod(m, struct ip *);
1680 optlen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
1681 opts = (caddr_t)(ip + 1);
1682 datalen = m->m_len - (sizeof(struct ip) + optlen);
1683 bcopy(opts + optlen, opts, datalen);
1685 if (m->m_flags & M_PKTHDR)
1686 m->m_pkthdr.len -= optlen;
1687 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, sizeof(struct ip) >> 2);
1690 u_char inetctlerrmap[PRC_NCMDS] = {
1692 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
1693 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
1694 EMSGSIZE, EHOSTUNREACH, 0, 0,
1696 ENOPROTOOPT, ECONNREFUSED
1700 * Forward a packet. If some error occurs return the sender
1701 * an icmp packet. Note we can't always generate a meaningful
1702 * icmp message because icmp doesn't have a large enough repertoire
1703 * of codes and types.
1705 * If not forwarding, just drop the packet. This could be confusing
1706 * if ipforwarding was zero but some routing protocol was advancing
1707 * us as a gateway to somewhere. However, we must let the routing
1708 * protocol deal with that.
1710 * The srcrt parameter indicates whether the packet is being forwarded
1711 * via a source route.
1714 ip_forward(struct mbuf *m, int srcrt, struct sockaddr_in *next_hop)
1716 struct ip *ip = mtod(m, struct ip *);
1717 struct sockaddr_in *sin;
1719 int error, type = 0, code = 0;
1722 struct in_addr pkt_dst;
1723 struct ifnet *destifp;
1724 #if defined(IPSEC) || defined(FAST_IPSEC)
1725 struct ifnet dummyifp;
1730 * Cache the destination address of the packet; this may be
1731 * changed by use of 'ipfw fwd'.
1733 pkt_dst = next_hop ? next_hop->sin_addr : ip->ip_dst;
1737 printf("forward: src %lx dst %lx ttl %x\n",
1738 (u_long)ip->ip_src.s_addr, (u_long)pkt_dst.s_addr,
1743 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(pkt_dst) == 0) {
1744 ipstat.ips_cantforward++;
1751 if (ip->ip_ttl <= IPTTLDEC) {
1752 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS,
1760 sin = (struct sockaddr_in *)&ipforward_rt.ro_dst;
1761 if ((rt = ipforward_rt.ro_rt) == 0 ||
1762 pkt_dst.s_addr != sin->sin_addr.s_addr) {
1763 if (ipforward_rt.ro_rt) {
1764 RTFREE(ipforward_rt.ro_rt);
1765 ipforward_rt.ro_rt = 0;
1767 sin->sin_family = AF_INET;
1768 sin->sin_len = sizeof(*sin);
1769 sin->sin_addr = pkt_dst;
1771 rtalloc_ign(&ipforward_rt, RTF_PRCLONING);
1772 if (ipforward_rt.ro_rt == 0) {
1773 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0);
1776 rt = ipforward_rt.ro_rt;
1780 * Save the IP header and at most 8 bytes of the payload,
1781 * in case we need to generate an ICMP message to the src.
1783 * XXX this can be optimized a lot by saving the data in a local
1784 * buffer on the stack (72 bytes at most), and only allocating the
1785 * mbuf if really necessary. The vast majority of the packets
1786 * are forwarded without having to send an ICMP back (either
1787 * because unnecessary, or because rate limited), so we are
1788 * really we are wasting a lot of work here.
1790 * We don't use m_copy() because it might return a reference
1791 * to a shared cluster. Both this function and ip_output()
1792 * assume exclusive access to the IP header in `m', so any
1793 * data in a cluster may change before we reach icmp_error().
1795 MGET(mcopy, M_DONTWAIT, m->m_type);
1796 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_DONTWAIT)) {
1798 * It's probably ok if the pkthdr dup fails (because
1799 * the deep copy of the tag chain failed), but for now
1800 * be conservative and just discard the copy since
1801 * code below may some day want the tags.
1806 if (mcopy != NULL) {
1807 mcopy->m_len = imin((IP_VHL_HL(ip->ip_vhl) << 2) + 8,
1809 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1815 ip->ip_ttl -= IPTTLDEC;
1821 * If forwarding packet using same interface that it came in on,
1822 * perhaps should send a redirect to sender to shortcut a hop.
1823 * Only send redirect if source is sending directly to us,
1824 * and if packet was not source routed (or has any options).
1825 * Also, don't send redirect if forwarding using a default route
1826 * or a route modified by a redirect.
1828 if (rt->rt_ifp == m->m_pkthdr.rcvif &&
1829 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1830 satosin(rt_key(rt))->sin_addr.s_addr != 0 &&
1831 ipsendredirects && !srcrt && !next_hop) {
1832 #define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa))
1833 u_long src = ntohl(ip->ip_src.s_addr);
1836 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1837 if (rt->rt_flags & RTF_GATEWAY)
1838 dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
1840 dest = pkt_dst.s_addr;
1841 /* Router requirements says to only send host redirects */
1842 type = ICMP_REDIRECT;
1843 code = ICMP_REDIRECT_HOST;
1846 printf("redirect (%d) to %lx\n", code, (u_long)dest);
1855 /* Pass IPFORWARD info if available */
1857 tag.mh_type = MT_TAG;
1858 tag.mh_flags = PACKET_TAG_IPFORWARD;
1859 tag.mh_data = (caddr_t)next_hop;
1861 m = (struct mbuf *)&tag;
1863 error = ip_output(m, (struct mbuf *)0, &ipforward_rt,
1864 IP_FORWARDING, 0, NULL);
1867 ipstat.ips_cantforward++;
1869 ipstat.ips_forward++;
1871 ipstat.ips_redirectsent++;
1874 ipflow_create(&ipforward_rt, mcopy);
1886 case 0: /* forwarded, but need redirect */
1887 /* type, code set above */
1890 case ENETUNREACH: /* shouldn't happen, checked above */
1895 type = ICMP_UNREACH;
1896 code = ICMP_UNREACH_HOST;
1900 type = ICMP_UNREACH;
1901 code = ICMP_UNREACH_NEEDFRAG;
1904 * If the packet is routed over IPsec tunnel, tell the
1905 * originator the tunnel MTU.
1906 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
1909 if (ipforward_rt.ro_rt) {
1910 struct secpolicy *sp = NULL;
1915 sp = ipsec4_getpolicybyaddr(mcopy,
1921 destifp = ipforward_rt.ro_rt->rt_ifp;
1923 /* count IPsec header size */
1924 ipsechdr = ipsec4_hdrsiz(mcopy,
1929 * find the correct route for outer IPv4
1930 * header, compute tunnel MTU.
1933 * The "dummyifp" code relies upon the fact
1934 * that icmp_error() touches only ifp->if_mtu.
1939 && sp->req->sav != NULL
1940 && sp->req->sav->sah != NULL) {
1941 ro = &sp->req->sav->sah->sa_route;
1942 if (ro->ro_rt && ro->ro_rt->rt_ifp) {
1944 ro->ro_rt->rt_ifp->if_mtu;
1945 dummyifp.if_mtu -= ipsechdr;
1946 destifp = &dummyifp;
1955 * If the packet is routed over IPsec tunnel, tell the
1956 * originator the tunnel MTU.
1957 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
1960 if (ipforward_rt.ro_rt) {
1961 struct secpolicy *sp = NULL;
1966 sp = ipsec_getpolicybyaddr(mcopy,
1972 destifp = ipforward_rt.ro_rt->rt_ifp;
1974 /* count IPsec header size */
1975 ipsechdr = ipsec4_hdrsiz(mcopy,
1980 * find the correct route for outer IPv4
1981 * header, compute tunnel MTU.
1984 * The "dummyifp" code relies upon the fact
1985 * that icmp_error() touches only ifp->if_mtu.
1990 && sp->req->sav != NULL
1991 && sp->req->sav->sah != NULL) {
1992 ro = &sp->req->sav->sah->sa_route;
1993 if (ro->ro_rt && ro->ro_rt->rt_ifp) {
1995 ro->ro_rt->rt_ifp->if_mtu;
1996 dummyifp.if_mtu -= ipsechdr;
1997 destifp = &dummyifp;
2004 #else /* !IPSEC && !FAST_IPSEC */
2005 if (ipforward_rt.ro_rt)
2006 destifp = ipforward_rt.ro_rt->rt_ifp;
2008 ipstat.ips_cantfrag++;
2013 * A router should not generate ICMP_SOURCEQUENCH as
2014 * required in RFC1812 Requirements for IP Version 4 Routers.
2015 * Source quench could be a big problem under DoS attacks,
2016 * or if the underlying interface is rate-limited.
2017 * Those who need source quench packets may re-enable them
2018 * via the net.inet.ip.sendsourcequench sysctl.
2020 if (ip_sendsourcequench == 0) {
2024 type = ICMP_SOURCEQUENCH;
2029 case EACCES: /* ipfw denied packet */
2033 icmp_error(mcopy, type, code, dest, destifp);
2037 ip_savecontrol(inp, mp, ip, m)
2043 if (inp->inp_socket->so_options & SO_TIMESTAMP) {
2047 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
2048 SCM_TIMESTAMP, SOL_SOCKET);
2050 mp = &(*mp)->m_next;
2052 if (inp->inp_flags & INP_RECVDSTADDR) {
2053 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
2054 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
2056 mp = &(*mp)->m_next;
2060 * Moving these out of udp_input() made them even more broken
2061 * than they already were.
2063 /* options were tossed already */
2064 if (inp->inp_flags & INP_RECVOPTS) {
2065 *mp = sbcreatecontrol((caddr_t) opts_deleted_above,
2066 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
2068 mp = &(*mp)->m_next;
2070 /* ip_srcroute doesn't do what we want here, need to fix */
2071 if (inp->inp_flags & INP_RECVRETOPTS) {
2072 *mp = sbcreatecontrol((caddr_t) ip_srcroute(),
2073 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
2075 mp = &(*mp)->m_next;
2078 if (inp->inp_flags & INP_RECVIF) {
2081 struct sockaddr_dl sdl;
2084 struct sockaddr_dl *sdp;
2085 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
2087 if (((ifp = m->m_pkthdr.rcvif))
2088 && ( ifp->if_index && (ifp->if_index <= if_index))) {
2089 sdp = (struct sockaddr_dl *)(ifnet_addrs
2090 [ifp->if_index - 1]->ifa_addr);
2092 * Change our mind and don't try copy.
2094 if ((sdp->sdl_family != AF_LINK)
2095 || (sdp->sdl_len > sizeof(sdlbuf))) {
2098 bcopy(sdp, sdl2, sdp->sdl_len);
2102 = offsetof(struct sockaddr_dl, sdl_data[0]);
2103 sdl2->sdl_family = AF_LINK;
2104 sdl2->sdl_index = 0;
2105 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
2107 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len,
2108 IP_RECVIF, IPPROTO_IP);
2110 mp = &(*mp)->m_next;
2115 * XXX these routines are called from the upper part of the kernel.
2117 * They could also be moved to ip_mroute.c, since all the RSVP
2118 * handling is done there already.
2121 ip_rsvp_init(struct socket *so)
2123 if (so->so_type != SOCK_RAW ||
2124 so->so_proto->pr_protocol != IPPROTO_RSVP)
2127 if (ip_rsvpd != NULL)
2132 * This may seem silly, but we need to be sure we don't over-increment
2133 * the RSVP counter, in case something slips up.
2148 * This may seem silly, but we need to be sure we don't over-decrement
2149 * the RSVP counter, in case something slips up.
2159 rsvp_input(struct mbuf *m, int off, int proto) /* XXX must fixup manually */
2161 if (rsvp_input_p) { /* call the real one if loaded */
2162 rsvp_input_p(m, off, proto);
2166 /* Can still get packets with rsvp_on = 0 if there is a local member
2167 * of the group to which the RSVP packet is addressed. But in this
2168 * case we want to throw the packet away.
2176 if (ip_rsvpd != NULL) {
2177 rip_input(m, off, proto);
2180 /* Drop the packet */