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 1.9 2003/12/02 08:00:22 asmodai 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>
63 #include <net/if_types.h>
64 #include <net/if_var.h>
65 #include <net/if_dl.h>
69 #include <net/route.h>
70 #include <net/netisr.h>
71 #include <net/intrq.h>
73 #include <netinet/in.h>
74 #include <netinet/in_systm.h>
75 #include <netinet/in_var.h>
76 #include <netinet/ip.h>
77 #include <netinet/in_pcb.h>
78 #include <netinet/ip_var.h>
79 #include <netinet/ip_icmp.h>
80 #include <machine/in_cksum.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 mbuf *m)
310 struct in_ifaddr *ia = NULL;
312 int i, hlen, checkif;
314 struct in_addr pkt_dst;
315 u_int32_t divert_info = 0; /* packet divert/tee info */
316 struct ip_fw_args args;
317 int srcrt = 0; /* forward (by PFIL_HOOKS) */
319 struct in_addr odst; /* original dst address(NAT) */
323 struct tdb_ident *tdbi;
324 struct secpolicy *sp;
326 #endif /* FAST_IPSEC */
331 args.divert_rule = 0; /* divert cookie */
332 args.next_hop = NULL;
334 /* Grab info from MT_TAG mbufs prepended to the chain. */
335 for (; m && m->m_type == MT_TAG; m = m->m_next) {
336 switch(m->_m_tag_id) {
338 printf("ip_input: unrecognised MT_TAG tag %d\n",
342 case PACKET_TAG_DUMMYNET:
343 args.rule = ((struct dn_pkt *)m)->rule;
346 case PACKET_TAG_DIVERT:
347 args.divert_rule = (int)m->m_hdr.mh_data & 0xffff;
350 case PACKET_TAG_IPFORWARD:
351 args.next_hop = (struct sockaddr_in *)m->m_hdr.mh_data;
356 KASSERT(m != NULL && (m->m_flags & M_PKTHDR) != 0,
357 ("ip_input: no HDR"));
359 if (args.rule) { /* dummynet already filtered us */
360 ip = mtod(m, struct ip *);
361 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
367 if (m->m_pkthdr.len < sizeof(struct ip))
370 if (m->m_len < sizeof (struct ip) &&
371 (m = m_pullup(m, sizeof (struct ip))) == 0) {
372 ipstat.ips_toosmall++;
375 ip = mtod(m, struct ip *);
377 if (IP_VHL_V(ip->ip_vhl) != IPVERSION) {
378 ipstat.ips_badvers++;
382 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
383 if (hlen < sizeof(struct ip)) { /* minimum header length */
384 ipstat.ips_badhlen++;
387 if (hlen > m->m_len) {
388 if ((m = m_pullup(m, hlen)) == 0) {
389 ipstat.ips_badhlen++;
392 ip = mtod(m, struct ip *);
395 /* 127/8 must not appear on wire - RFC1122 */
396 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
397 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
398 if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
399 ipstat.ips_badaddr++;
404 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
405 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
407 if (hlen == sizeof(struct ip)) {
408 sum = in_cksum_hdr(ip);
410 sum = in_cksum(m, hlen);
419 * Convert fields to host representation.
421 ip->ip_len = ntohs(ip->ip_len);
422 if (ip->ip_len < hlen) {
426 ip->ip_off = ntohs(ip->ip_off);
429 * Check that the amount of data in the buffers
430 * is as at least much as the IP header would have us expect.
431 * Trim mbufs if longer than we expect.
432 * Drop packet if shorter than we expect.
434 if (m->m_pkthdr.len < ip->ip_len) {
436 ipstat.ips_tooshort++;
439 if (m->m_pkthdr.len > ip->ip_len) {
440 if (m->m_len == m->m_pkthdr.len) {
441 m->m_len = ip->ip_len;
442 m->m_pkthdr.len = ip->ip_len;
444 m_adj(m, ip->ip_len - m->m_pkthdr.len);
446 #if defined(IPSEC) && !defined(IPSEC_FILTERGIF)
448 * Bypass packet filtering for packets from a tunnel (gif).
450 if (ipsec_gethist(m, NULL))
456 * Right now when no processing on packet has done
457 * and it is still fresh out of network we do our black
459 * - Firewall: deny/allow/divert
460 * - Xlate: translate packet's addr/port (NAT).
461 * - Pipe: pass pkt through dummynet.
462 * - Wrap: fake packet's addr/port <unimpl.>
463 * - Encapsulate: put it in another IP and send out. <unimp.>
469 * Run through list of hooks for input packets.
471 * NB: Beware of the destination address changing (e.g.
472 * by NAT rewriting). When this happens, tell
473 * ip_forward to do the right thing.
476 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif,
479 if (m == NULL) /* consumed by filter */
481 ip = mtod(m, struct ip *);
482 srcrt = (odst.s_addr != ip->ip_dst.s_addr);
483 #endif /* PFIL_HOOKS */
484 if (fw_enable && IPFW_LOADED) {
486 * If we've been forwarded from the output side, then
487 * skip the firewall a second time
493 i = ip_fw_chk_ptr(&args);
496 if ( (i & IP_FW_PORT_DENY_FLAG) || m == NULL) { /* drop */
501 ip = mtod(m, struct ip *); /* just in case m changed */
502 if (i == 0 && args.next_hop == NULL) /* common case */
504 if (DUMMYNET_LOADED && (i & IP_FW_PORT_DYNT_FLAG) != 0) {
505 /* Send packet to the appropriate pipe */
506 ip_dn_io_ptr(m, i&0xffff, DN_TO_IP_IN, &args);
510 if (i != 0 && (i & IP_FW_PORT_DYNT_FLAG) == 0) {
511 /* Divert or tee packet */
516 if (i == 0 && args.next_hop != NULL)
519 * if we get here, the packet must be dropped
527 * Process options and, if not destined for us,
528 * ship it on. ip_dooptions returns 1 when an
529 * error was detected (causing an icmp message
530 * to be sent and the original packet to be freed).
532 ip_nhops = 0; /* for source routed packets */
533 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0, args.next_hop))
536 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
537 * matter if it is destined to another node, or whether it is
538 * a multicast one, RSVP wants it! and prevents it from being forwarded
539 * anywhere else. Also checks if the rsvp daemon is running before
540 * grabbing the packet.
542 if (rsvp_on && ip->ip_p==IPPROTO_RSVP)
546 * Check our list of addresses, to see if the packet is for us.
547 * If we don't have any addresses, assume any unicast packet
548 * we receive might be for us (and let the upper layers deal
551 if (TAILQ_EMPTY(&in_ifaddrhead) &&
552 (m->m_flags & (M_MCAST|M_BCAST)) == 0)
556 * Cache the destination address of the packet; this may be
557 * changed by use of 'ipfw fwd'.
559 pkt_dst = args.next_hop ? args.next_hop->sin_addr : ip->ip_dst;
562 * Enable a consistency check between the destination address
563 * and the arrival interface for a unicast packet (the RFC 1122
564 * strong ES model) if IP forwarding is disabled and the packet
565 * is not locally generated and the packet is not subject to
568 * XXX - Checking also should be disabled if the destination
569 * address is ipnat'ed to a different interface.
571 * XXX - Checking is incompatible with IP aliases added
572 * to the loopback interface instead of the interface where
573 * the packets are received.
575 checkif = ip_checkinterface && (ipforwarding == 0) &&
576 m->m_pkthdr.rcvif != NULL &&
577 ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) &&
578 (args.next_hop == NULL);
581 * Check for exact addresses in the hash bucket.
583 LIST_FOREACH(ia, INADDR_HASH(pkt_dst.s_addr), ia_hash) {
585 * If the address matches, verify that the packet
586 * arrived via the correct interface if checking is
589 if (IA_SIN(ia)->sin_addr.s_addr == pkt_dst.s_addr &&
590 (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif))
594 * Check for broadcast addresses.
596 * Only accept broadcast packets that arrive via the matching
597 * interface. Reception of forwarded directed broadcasts would
598 * be handled via ip_forward() and ether_output() with the loopback
599 * into the stack for SIMPLEX interfaces handled by ether_output().
601 if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
602 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
603 if (ifa->ifa_addr->sa_family != AF_INET)
606 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
609 if (ia->ia_netbroadcast.s_addr == pkt_dst.s_addr)
612 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY)
617 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
618 struct in_multi *inm;
621 * If we are acting as a multicast router, all
622 * incoming multicast packets are passed to the
623 * kernel-level multicast forwarding function.
624 * The packet is returned (relatively) intact; if
625 * ip_mforward() returns a non-zero value, the packet
626 * must be discarded, else it may be accepted below.
629 ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) {
630 ipstat.ips_cantforward++;
636 * The process-level routing daemon needs to receive
637 * all multicast IGMP packets, whether or not this
638 * host belongs to their destination groups.
640 if (ip->ip_p == IPPROTO_IGMP)
642 ipstat.ips_forward++;
645 * See if we belong to the destination multicast group on the
648 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
650 ipstat.ips_notmember++;
656 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
658 if (ip->ip_dst.s_addr == INADDR_ANY)
662 * FAITH(Firewall Aided Internet Translator)
664 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
666 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP)
674 * Not for us; forward if possible and desirable.
676 if (ipforwarding == 0) {
677 ipstat.ips_cantforward++;
682 * Enforce inbound IPsec SPD.
684 if (ipsec4_in_reject(m, NULL)) {
685 ipsecstat.in_polvio++;
690 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
693 tdbi = (struct tdb_ident *)(mtag + 1);
694 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
696 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
697 IP_FORWARDING, &error);
699 if (sp == NULL) { /* NB: can happen if error */
701 /*XXX error stat???*/
702 DPRINTF(("ip_input: no SP for forwarding\n")); /*XXX*/
707 * Check security policy against packet attributes.
709 error = ipsec_in_reject(sp, m);
713 ipstat.ips_cantforward++;
716 #endif /* FAST_IPSEC */
717 ip_forward(m, srcrt, args.next_hop);
724 * IPSTEALTH: Process non-routing options only
725 * if the packet is destined for us.
727 if (ipstealth && hlen > sizeof (struct ip) &&
728 ip_dooptions(m, 1, args.next_hop))
730 #endif /* IPSTEALTH */
732 /* Count the packet in the ip address stats */
734 ia->ia_ifa.if_ipackets++;
735 ia->ia_ifa.if_ibytes += m->m_pkthdr.len;
739 * If offset or IP_MF are set, must reassemble.
740 * Otherwise, nothing need be done.
741 * (We could look in the reassembly queue to see
742 * if the packet was previously fragmented,
743 * but it's not worth the time; just let them time out.)
745 if (ip->ip_off & (IP_MF | IP_OFFMASK)) {
747 /* If maxnipq is 0, never accept fragments. */
749 ipstat.ips_fragments++;
750 ipstat.ips_fragdropped++;
754 sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
756 * Look for queue of fragments
759 for (fp = ipq[sum].next; fp != &ipq[sum]; fp = fp->next)
760 if (ip->ip_id == fp->ipq_id &&
761 ip->ip_src.s_addr == fp->ipq_src.s_addr &&
762 ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
763 ip->ip_p == fp->ipq_p)
769 * Enforce upper bound on number of fragmented packets
770 * for which we attempt reassembly;
771 * If maxnipq is -1, accept all fragments without limitation.
773 if ((nipq > maxnipq) && (maxnipq > 0)) {
775 * drop something from the tail of the current queue
776 * before proceeding further
778 if (ipq[sum].prev == &ipq[sum]) { /* gak */
779 for (i = 0; i < IPREASS_NHASH; i++) {
780 if (ipq[i].prev != &ipq[i]) {
781 ipstat.ips_fragtimeout +=
782 ipq[i].prev->ipq_nfrags;
783 ip_freef(ipq[i].prev);
788 ipstat.ips_fragtimeout += ipq[sum].prev->ipq_nfrags;
789 ip_freef(ipq[sum].prev);
794 * Adjust ip_len to not reflect header,
795 * convert offset of this to bytes.
798 if (ip->ip_off & IP_MF) {
800 * Make sure that fragments have a data length
801 * that's a non-zero multiple of 8 bytes.
803 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
804 ipstat.ips_toosmall++; /* XXX */
807 m->m_flags |= M_FRAG;
809 m->m_flags &= ~M_FRAG;
813 * Attempt reassembly; if it succeeds, proceed.
814 * ip_reass() will return a different mbuf, and update
815 * the divert info in divert_info and args.divert_rule.
817 ipstat.ips_fragments++;
818 m->m_pkthdr.header = ip;
820 fp, &ipq[sum], &divert_info, &args.divert_rule);
823 ipstat.ips_reassembled++;
824 ip = mtod(m, struct ip *);
825 /* Get the header length of the reassembled packet */
826 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
828 /* Restore original checksum before diverting packet */
829 if (divert_info != 0) {
831 ip->ip_len = htons(ip->ip_len);
832 ip->ip_off = htons(ip->ip_off);
834 if (hlen == sizeof(struct ip))
835 ip->ip_sum = in_cksum_hdr(ip);
837 ip->ip_sum = in_cksum(m, hlen);
838 ip->ip_off = ntohs(ip->ip_off);
839 ip->ip_len = ntohs(ip->ip_len);
849 * Divert or tee packet to the divert protocol if required.
851 if (divert_info != 0) {
852 struct mbuf *clone = NULL;
854 /* Clone packet if we're doing a 'tee' */
855 if ((divert_info & IP_FW_PORT_TEE_FLAG) != 0)
856 clone = m_dup(m, M_DONTWAIT);
858 /* Restore packet header fields to original values */
860 ip->ip_len = htons(ip->ip_len);
861 ip->ip_off = htons(ip->ip_off);
863 /* Deliver packet to divert input routine */
864 divert_packet(m, 1, divert_info & 0xffff, args.divert_rule);
865 ipstat.ips_delivered++;
867 /* If 'tee', continue with original packet */
871 ip = mtod(m, struct ip *);
874 * Jump backwards to complete processing of the
875 * packet. But first clear divert_info to avoid
876 * entering this block again.
877 * We do not need to clear args.divert_rule
878 * or args.next_hop as they will not be used.
887 * enforce IPsec policy checking if we are seeing last header.
888 * note that we do not visit this with protocols with pcb layer
889 * code - like udp/tcp/raw ip.
891 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0 &&
892 ipsec4_in_reject(m, NULL)) {
893 ipsecstat.in_polvio++;
899 * enforce IPsec policy checking if we are seeing last header.
900 * note that we do not visit this with protocols with pcb layer
901 * code - like udp/tcp/raw ip.
903 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) {
905 * Check if the packet has already had IPsec processing
906 * done. If so, then just pass it along. This tag gets
907 * set during AH, ESP, etc. input handling, before the
908 * packet is returned to the ip input queue for delivery.
910 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
913 tdbi = (struct tdb_ident *)(mtag + 1);
914 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
916 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
917 IP_FORWARDING, &error);
921 * Check security policy against packet attributes.
923 error = ipsec_in_reject(sp, m);
926 /* XXX error stat??? */
928 DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/
935 #endif /* FAST_IPSEC */
938 * Switch out to protocol's input routine.
940 * XXX queue packet to protocol's message port.
942 ipstat.ips_delivered++;
943 if (args.next_hop && ip->ip_p == IPPROTO_TCP) {
944 /* TCP needs IPFORWARD info if available */
947 tag.mh_type = MT_TAG;
948 tag.mh_flags = PACKET_TAG_IPFORWARD;
949 tag.mh_data = (caddr_t)args.next_hop;
952 (*inetsw[ip_protox[ip->ip_p]].pr_input)(
953 (struct mbuf *)&tag, hlen, ip->ip_p);
955 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen, ip->ip_p);
963 * Take incoming datagram fragment and try to reassemble it into
964 * whole datagram. If a chain for reassembly of this datagram already
965 * exists, then it is given as fp; otherwise have to make a chain.
967 * When IPDIVERT enabled, keep additional state with each packet that
968 * tells us if we need to divert or tee the packet we're building.
969 * In particular, *divinfo includes the port and TEE flag,
970 * *divert_rule is the number of the matching rule.
974 ip_reass(struct mbuf *m, struct ipq *fp, struct ipq *where,
975 u_int32_t *divinfo, u_int16_t *divert_rule)
977 struct ip *ip = mtod(m, struct ip *);
978 struct mbuf *p = 0, *q, *nq;
980 int hlen = IP_VHL_HL(ip->ip_vhl) << 2;
984 * Presence of header sizes in mbufs
985 * would confuse code below.
991 * If first fragment to arrive, create a reassembly queue.
994 if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL)
996 fp = mtod(t, struct ipq *);
1000 fp->ipq_ttl = IPFRAGTTL;
1001 fp->ipq_p = ip->ip_p;
1002 fp->ipq_id = ip->ip_id;
1003 fp->ipq_src = ip->ip_src;
1004 fp->ipq_dst = ip->ip_dst;
1006 m->m_nextpkt = NULL;
1008 fp->ipq_div_info = 0;
1009 fp->ipq_div_cookie = 0;
1016 #define GETIP(m) ((struct ip*)((m)->m_pkthdr.header))
1019 * Find a segment which begins after this one does.
1021 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt)
1022 if (GETIP(q)->ip_off > ip->ip_off)
1026 * If there is a preceding segment, it may provide some of
1027 * our data already. If so, drop the data from the incoming
1028 * segment. If it provides all of our data, drop us, otherwise
1029 * stick new segment in the proper place.
1031 * If some of the data is dropped from the the preceding
1032 * segment, then it's checksum is invalidated.
1035 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off;
1037 if (i >= ip->ip_len)
1040 m->m_pkthdr.csum_flags = 0;
1044 m->m_nextpkt = p->m_nextpkt;
1047 m->m_nextpkt = fp->ipq_frags;
1052 * While we overlap succeeding segments trim them or,
1053 * if they are completely covered, dequeue them.
1055 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off;
1057 i = (ip->ip_off + ip->ip_len) -
1059 if (i < GETIP(q)->ip_len) {
1060 GETIP(q)->ip_len -= i;
1061 GETIP(q)->ip_off += i;
1063 q->m_pkthdr.csum_flags = 0;
1068 ipstat.ips_fragdropped++;
1077 * Transfer firewall instructions to the fragment structure.
1078 * Only trust info in the fragment at offset 0.
1080 if (ip->ip_off == 0) {
1081 fp->ipq_div_info = *divinfo;
1082 fp->ipq_div_cookie = *divert_rule;
1089 * Check for complete reassembly and perform frag per packet
1092 * Frag limiting is performed here so that the nth frag has
1093 * a chance to complete the packet before we drop the packet.
1094 * As a result, n+1 frags are actually allowed per packet, but
1095 * only n will ever be stored. (n = maxfragsperpacket.)
1099 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
1100 if (GETIP(q)->ip_off != next) {
1101 if (fp->ipq_nfrags > maxfragsperpacket) {
1102 ipstat.ips_fragdropped += fp->ipq_nfrags;
1107 next += GETIP(q)->ip_len;
1109 /* Make sure the last packet didn't have the IP_MF flag */
1110 if (p->m_flags & M_FRAG) {
1111 if (fp->ipq_nfrags > maxfragsperpacket) {
1112 ipstat.ips_fragdropped += fp->ipq_nfrags;
1119 * Reassembly is complete. Make sure the packet is a sane size.
1123 if (next + (IP_VHL_HL(ip->ip_vhl) << 2) > IP_MAXPACKET) {
1124 ipstat.ips_toolong++;
1125 ipstat.ips_fragdropped += fp->ipq_nfrags;
1131 * Concatenate fragments.
1139 for (q = nq; q != NULL; q = nq) {
1141 q->m_nextpkt = NULL;
1142 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
1143 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
1149 * Extract firewall instructions from the fragment structure.
1151 *divinfo = fp->ipq_div_info;
1152 *divert_rule = fp->ipq_div_cookie;
1156 * Create header for new ip packet by
1157 * modifying header of first packet;
1158 * dequeue and discard fragment reassembly header.
1159 * Make header visible.
1162 ip->ip_src = fp->ipq_src;
1163 ip->ip_dst = fp->ipq_dst;
1166 (void) m_free(dtom(fp));
1167 m->m_len += (IP_VHL_HL(ip->ip_vhl) << 2);
1168 m->m_data -= (IP_VHL_HL(ip->ip_vhl) << 2);
1169 /* some debugging cruft by sklower, below, will go away soon */
1170 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */
1172 for (t = m; t; t = t->m_next)
1174 m->m_pkthdr.len = plen;
1183 ipstat.ips_fragdropped++;
1193 * Free a fragment reassembly header and all
1194 * associated datagrams.
1202 while (fp->ipq_frags) {
1204 fp->ipq_frags = q->m_nextpkt;
1208 (void) m_free(dtom(fp));
1213 * IP timer processing;
1214 * if a timer expires on a reassembly
1215 * queue, discard it.
1224 for (i = 0; i < IPREASS_NHASH; i++) {
1228 while (fp != &ipq[i]) {
1231 if (fp->prev->ipq_ttl == 0) {
1232 ipstat.ips_fragtimeout += fp->prev->ipq_nfrags;
1238 * If we are over the maximum number of fragments
1239 * (due to the limit being lowered), drain off
1240 * enough to get down to the new limit.
1242 if (maxnipq >= 0 && nipq > maxnipq) {
1243 for (i = 0; i < IPREASS_NHASH; i++) {
1244 while (nipq > maxnipq &&
1245 (ipq[i].next != &ipq[i])) {
1246 ipstat.ips_fragdropped +=
1247 ipq[i].next->ipq_nfrags;
1248 ip_freef(ipq[i].next);
1257 * Drain off all datagram fragments.
1264 for (i = 0; i < IPREASS_NHASH; i++) {
1265 while (ipq[i].next != &ipq[i]) {
1266 ipstat.ips_fragdropped += ipq[i].next->ipq_nfrags;
1267 ip_freef(ipq[i].next);
1274 * Do option processing on a datagram,
1275 * possibly discarding it if bad options are encountered,
1276 * or forwarding it if source-routed.
1277 * The pass argument is used when operating in the IPSTEALTH
1278 * mode to tell what options to process:
1279 * [LS]SRR (pass 0) or the others (pass 1).
1280 * The reason for as many as two passes is that when doing IPSTEALTH,
1281 * non-routing options should be processed only if the packet is for us.
1282 * Returns 1 if packet has been forwarded/freed,
1283 * 0 if the packet should be processed further.
1286 ip_dooptions(struct mbuf *m, int pass, struct sockaddr_in *next_hop)
1288 struct ip *ip = mtod(m, struct ip *);
1290 struct in_ifaddr *ia;
1291 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0;
1292 struct in_addr *sin, dst;
1296 cp = (u_char *)(ip + 1);
1297 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip);
1298 for (; cnt > 0; cnt -= optlen, cp += optlen) {
1299 opt = cp[IPOPT_OPTVAL];
1300 if (opt == IPOPT_EOL)
1302 if (opt == IPOPT_NOP)
1305 if (cnt < IPOPT_OLEN + sizeof(*cp)) {
1306 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1309 optlen = cp[IPOPT_OLEN];
1310 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
1311 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1321 * Source routing with record.
1322 * Find interface with current destination address.
1323 * If none on this machine then drop if strictly routed,
1324 * or do nothing if loosely routed.
1325 * Record interface address and bring up next address
1326 * component. If strictly routed make sure next
1327 * address is on directly accessible net.
1332 if (ipstealth && pass > 0)
1335 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1336 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1339 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1340 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1343 ipaddr.sin_addr = ip->ip_dst;
1344 ia = (struct in_ifaddr *)
1345 ifa_ifwithaddr((struct sockaddr *)&ipaddr);
1347 if (opt == IPOPT_SSRR) {
1348 type = ICMP_UNREACH;
1349 code = ICMP_UNREACH_SRCFAIL;
1352 if (!ip_dosourceroute)
1353 goto nosourcerouting;
1355 * Loose routing, and not at next destination
1356 * yet; nothing to do except forward.
1360 off--; /* 0 origin */
1361 if (off > optlen - (int)sizeof(struct in_addr)) {
1363 * End of source route. Should be for us.
1365 if (!ip_acceptsourceroute)
1366 goto nosourcerouting;
1367 save_rte(cp, ip->ip_src);
1374 if (!ip_dosourceroute) {
1376 char buf[16]; /* aaa.bbb.ccc.ddd\0 */
1378 * Acting as a router, so generate ICMP
1381 strcpy(buf, inet_ntoa(ip->ip_dst));
1383 "attempted source route from %s to %s\n",
1384 inet_ntoa(ip->ip_src), buf);
1385 type = ICMP_UNREACH;
1386 code = ICMP_UNREACH_SRCFAIL;
1390 * Not acting as a router, so silently drop.
1395 ipstat.ips_cantforward++;
1402 * locate outgoing interface
1404 (void)memcpy(&ipaddr.sin_addr, cp + off,
1405 sizeof(ipaddr.sin_addr));
1407 if (opt == IPOPT_SSRR) {
1408 #define INA struct in_ifaddr *
1409 #define SA struct sockaddr *
1410 if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == 0)
1411 ia = (INA)ifa_ifwithnet((SA)&ipaddr);
1413 ia = ip_rtaddr(ipaddr.sin_addr, &ipforward_rt);
1415 type = ICMP_UNREACH;
1416 code = ICMP_UNREACH_SRCFAIL;
1419 ip->ip_dst = ipaddr.sin_addr;
1420 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
1421 sizeof(struct in_addr));
1422 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1424 * Let ip_intr's mcast routing check handle mcast pkts
1426 forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr));
1431 if (ipstealth && pass == 0)
1434 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1435 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1438 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1439 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1443 * If no space remains, ignore.
1445 off--; /* 0 origin */
1446 if (off > optlen - (int)sizeof(struct in_addr))
1448 (void)memcpy(&ipaddr.sin_addr, &ip->ip_dst,
1449 sizeof(ipaddr.sin_addr));
1451 * locate outgoing interface; if we're the destination,
1452 * use the incoming interface (should be same).
1454 if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == 0 &&
1455 (ia = ip_rtaddr(ipaddr.sin_addr,
1456 &ipforward_rt)) == 0) {
1457 type = ICMP_UNREACH;
1458 code = ICMP_UNREACH_HOST;
1461 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
1462 sizeof(struct in_addr));
1463 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1468 if (ipstealth && pass == 0)
1471 code = cp - (u_char *)ip;
1472 if (optlen < 4 || optlen > 40) {
1473 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1476 if ((off = cp[IPOPT_OFFSET]) < 5) {
1477 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1480 if (off > optlen - (int)sizeof(int32_t)) {
1481 cp[IPOPT_OFFSET + 1] += (1 << 4);
1482 if ((cp[IPOPT_OFFSET + 1] & 0xf0) == 0) {
1483 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1488 off--; /* 0 origin */
1489 sin = (struct in_addr *)(cp + off);
1490 switch (cp[IPOPT_OFFSET + 1] & 0x0f) {
1492 case IPOPT_TS_TSONLY:
1495 case IPOPT_TS_TSANDADDR:
1496 if (off + sizeof(n_time) +
1497 sizeof(struct in_addr) > optlen) {
1498 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1501 ipaddr.sin_addr = dst;
1502 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr,
1506 (void)memcpy(sin, &IA_SIN(ia)->sin_addr,
1507 sizeof(struct in_addr));
1508 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1509 off += sizeof(struct in_addr);
1512 case IPOPT_TS_PRESPEC:
1513 if (off + sizeof(n_time) +
1514 sizeof(struct in_addr) > optlen) {
1515 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1518 (void)memcpy(&ipaddr.sin_addr, sin,
1519 sizeof(struct in_addr));
1520 if (ifa_ifwithaddr((SA)&ipaddr) == 0)
1522 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1523 off += sizeof(struct in_addr);
1527 code = &cp[IPOPT_OFFSET + 1] - (u_char *)ip;
1531 (void)memcpy(cp + off, &ntime, sizeof(n_time));
1532 cp[IPOPT_OFFSET] += sizeof(n_time);
1535 if (forward && ipforwarding) {
1536 ip_forward(m, 1, next_hop);
1541 icmp_error(m, type, code, 0, 0);
1542 ipstat.ips_badoptions++;
1547 * Given address of next destination (final or next hop),
1548 * return internet address info of interface to be used to get there.
1555 struct sockaddr_in *sin;
1557 sin = (struct sockaddr_in *)&rt->ro_dst;
1559 if (rt->ro_rt == 0 ||
1560 dst.s_addr != sin->sin_addr.s_addr) {
1565 sin->sin_family = AF_INET;
1566 sin->sin_len = sizeof(*sin);
1567 sin->sin_addr = dst;
1569 rtalloc_ign(rt, RTF_PRCLONING);
1572 return ((struct in_ifaddr *)0);
1573 return (ifatoia(rt->ro_rt->rt_ifa));
1577 * Save incoming source route for use in replies,
1578 * to be picked up later by ip_srcroute if the receiver is interested.
1581 save_rte(option, dst)
1587 olen = option[IPOPT_OLEN];
1590 printf("save_rte: olen %d\n", olen);
1592 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst)))
1594 bcopy(option, ip_srcrt.srcopt, olen);
1595 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1600 * Retrieve incoming source route for use in replies,
1601 * in the same form used by setsockopt.
1602 * The first hop is placed before the options, will be removed later.
1607 struct in_addr *p, *q;
1611 return ((struct mbuf *)0);
1612 m = m_get(M_DONTWAIT, MT_HEADER);
1614 return ((struct mbuf *)0);
1616 #define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt))
1618 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
1619 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) +
1623 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len);
1627 * First save first hop for return route
1629 p = &ip_srcrt.route[ip_nhops - 1];
1630 *(mtod(m, struct in_addr *)) = *p--;
1633 printf(" hops %lx", (u_long)ntohl(mtod(m, struct in_addr *)->s_addr));
1637 * Copy option fields and padding (nop) to mbuf.
1639 ip_srcrt.nop = IPOPT_NOP;
1640 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
1641 (void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr),
1642 &ip_srcrt.nop, OPTSIZ);
1643 q = (struct in_addr *)(mtod(m, caddr_t) +
1644 sizeof(struct in_addr) + OPTSIZ);
1647 * Record return path as an IP source route,
1648 * reversing the path (pointers are now aligned).
1650 while (p >= ip_srcrt.route) {
1653 printf(" %lx", (u_long)ntohl(q->s_addr));
1658 * Last hop goes to final destination.
1663 printf(" %lx\n", (u_long)ntohl(q->s_addr));
1669 * Strip out IP options, at higher
1670 * level protocol in the kernel.
1671 * Second argument is buffer to which options
1672 * will be moved, and return value is their length.
1673 * XXX should be deleted; last arg currently ignored.
1676 ip_stripoptions(m, mopt)
1681 struct ip *ip = mtod(m, struct ip *);
1685 olen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip);
1686 opts = (caddr_t)(ip + 1);
1687 i = m->m_len - (sizeof (struct ip) + olen);
1688 bcopy(opts + olen, opts, (unsigned)i);
1690 if (m->m_flags & M_PKTHDR)
1691 m->m_pkthdr.len -= olen;
1692 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, sizeof(struct ip) >> 2);
1695 u_char inetctlerrmap[PRC_NCMDS] = {
1697 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
1698 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
1699 EMSGSIZE, EHOSTUNREACH, 0, 0,
1701 ENOPROTOOPT, ECONNREFUSED
1705 * Forward a packet. If some error occurs return the sender
1706 * an icmp packet. Note we can't always generate a meaningful
1707 * icmp message because icmp doesn't have a large enough repertoire
1708 * of codes and types.
1710 * If not forwarding, just drop the packet. This could be confusing
1711 * if ipforwarding was zero but some routing protocol was advancing
1712 * us as a gateway to somewhere. However, we must let the routing
1713 * protocol deal with that.
1715 * The srcrt parameter indicates whether the packet is being forwarded
1716 * via a source route.
1719 ip_forward(struct mbuf *m, int srcrt, struct sockaddr_in *next_hop)
1721 struct ip *ip = mtod(m, struct ip *);
1722 struct sockaddr_in *sin;
1724 int error, type = 0, code = 0;
1727 struct in_addr pkt_dst;
1728 struct ifnet *destifp;
1729 #if defined(IPSEC) || defined(FAST_IPSEC)
1730 struct ifnet dummyifp;
1735 * Cache the destination address of the packet; this may be
1736 * changed by use of 'ipfw fwd'.
1738 pkt_dst = next_hop ? next_hop->sin_addr : ip->ip_dst;
1742 printf("forward: src %lx dst %lx ttl %x\n",
1743 (u_long)ip->ip_src.s_addr, (u_long)pkt_dst.s_addr,
1748 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(pkt_dst) == 0) {
1749 ipstat.ips_cantforward++;
1756 if (ip->ip_ttl <= IPTTLDEC) {
1757 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS,
1765 sin = (struct sockaddr_in *)&ipforward_rt.ro_dst;
1766 if ((rt = ipforward_rt.ro_rt) == 0 ||
1767 pkt_dst.s_addr != sin->sin_addr.s_addr) {
1768 if (ipforward_rt.ro_rt) {
1769 RTFREE(ipforward_rt.ro_rt);
1770 ipforward_rt.ro_rt = 0;
1772 sin->sin_family = AF_INET;
1773 sin->sin_len = sizeof(*sin);
1774 sin->sin_addr = pkt_dst;
1776 rtalloc_ign(&ipforward_rt, RTF_PRCLONING);
1777 if (ipforward_rt.ro_rt == 0) {
1778 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0);
1781 rt = ipforward_rt.ro_rt;
1785 * Save the IP header and at most 8 bytes of the payload,
1786 * in case we need to generate an ICMP message to the src.
1788 * XXX this can be optimized a lot by saving the data in a local
1789 * buffer on the stack (72 bytes at most), and only allocating the
1790 * mbuf if really necessary. The vast majority of the packets
1791 * are forwarded without having to send an ICMP back (either
1792 * because unnecessary, or because rate limited), so we are
1793 * really we are wasting a lot of work here.
1795 * We don't use m_copy() because it might return a reference
1796 * to a shared cluster. Both this function and ip_output()
1797 * assume exclusive access to the IP header in `m', so any
1798 * data in a cluster may change before we reach icmp_error().
1800 MGET(mcopy, M_DONTWAIT, m->m_type);
1801 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_DONTWAIT)) {
1803 * It's probably ok if the pkthdr dup fails (because
1804 * the deep copy of the tag chain failed), but for now
1805 * be conservative and just discard the copy since
1806 * code below may some day want the tags.
1811 if (mcopy != NULL) {
1812 mcopy->m_len = imin((IP_VHL_HL(ip->ip_vhl) << 2) + 8,
1814 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1820 ip->ip_ttl -= IPTTLDEC;
1826 * If forwarding packet using same interface that it came in on,
1827 * perhaps should send a redirect to sender to shortcut a hop.
1828 * Only send redirect if source is sending directly to us,
1829 * and if packet was not source routed (or has any options).
1830 * Also, don't send redirect if forwarding using a default route
1831 * or a route modified by a redirect.
1833 if (rt->rt_ifp == m->m_pkthdr.rcvif &&
1834 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1835 satosin(rt_key(rt))->sin_addr.s_addr != 0 &&
1836 ipsendredirects && !srcrt && !next_hop) {
1837 #define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa))
1838 u_long src = ntohl(ip->ip_src.s_addr);
1841 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1842 if (rt->rt_flags & RTF_GATEWAY)
1843 dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
1845 dest = pkt_dst.s_addr;
1846 /* Router requirements says to only send host redirects */
1847 type = ICMP_REDIRECT;
1848 code = ICMP_REDIRECT_HOST;
1851 printf("redirect (%d) to %lx\n", code, (u_long)dest);
1860 /* Pass IPFORWARD info if available */
1862 tag.mh_type = MT_TAG;
1863 tag.mh_flags = PACKET_TAG_IPFORWARD;
1864 tag.mh_data = (caddr_t)next_hop;
1866 m = (struct mbuf *)&tag;
1868 error = ip_output(m, (struct mbuf *)0, &ipforward_rt,
1869 IP_FORWARDING, 0, NULL);
1872 ipstat.ips_cantforward++;
1874 ipstat.ips_forward++;
1876 ipstat.ips_redirectsent++;
1879 ipflow_create(&ipforward_rt, mcopy);
1891 case 0: /* forwarded, but need redirect */
1892 /* type, code set above */
1895 case ENETUNREACH: /* shouldn't happen, checked above */
1900 type = ICMP_UNREACH;
1901 code = ICMP_UNREACH_HOST;
1905 type = ICMP_UNREACH;
1906 code = ICMP_UNREACH_NEEDFRAG;
1909 * If the packet is routed over IPsec tunnel, tell the
1910 * originator the tunnel MTU.
1911 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
1914 if (ipforward_rt.ro_rt) {
1915 struct secpolicy *sp = NULL;
1920 sp = ipsec4_getpolicybyaddr(mcopy,
1926 destifp = ipforward_rt.ro_rt->rt_ifp;
1928 /* count IPsec header size */
1929 ipsechdr = ipsec4_hdrsiz(mcopy,
1934 * find the correct route for outer IPv4
1935 * header, compute tunnel MTU.
1938 * The "dummyifp" code relies upon the fact
1939 * that icmp_error() touches only ifp->if_mtu.
1944 && sp->req->sav != NULL
1945 && sp->req->sav->sah != NULL) {
1946 ro = &sp->req->sav->sah->sa_route;
1947 if (ro->ro_rt && ro->ro_rt->rt_ifp) {
1949 ro->ro_rt->rt_ifp->if_mtu;
1950 dummyifp.if_mtu -= ipsechdr;
1951 destifp = &dummyifp;
1960 * If the packet is routed over IPsec tunnel, tell the
1961 * originator the tunnel MTU.
1962 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
1965 if (ipforward_rt.ro_rt) {
1966 struct secpolicy *sp = NULL;
1971 sp = ipsec_getpolicybyaddr(mcopy,
1977 destifp = ipforward_rt.ro_rt->rt_ifp;
1979 /* count IPsec header size */
1980 ipsechdr = ipsec4_hdrsiz(mcopy,
1985 * find the correct route for outer IPv4
1986 * header, compute tunnel MTU.
1989 * The "dummyifp" code relies upon the fact
1990 * that icmp_error() touches only ifp->if_mtu.
1995 && sp->req->sav != NULL
1996 && sp->req->sav->sah != NULL) {
1997 ro = &sp->req->sav->sah->sa_route;
1998 if (ro->ro_rt && ro->ro_rt->rt_ifp) {
2000 ro->ro_rt->rt_ifp->if_mtu;
2001 dummyifp.if_mtu -= ipsechdr;
2002 destifp = &dummyifp;
2009 #else /* !IPSEC && !FAST_IPSEC */
2010 if (ipforward_rt.ro_rt)
2011 destifp = ipforward_rt.ro_rt->rt_ifp;
2013 ipstat.ips_cantfrag++;
2018 * A router should not generate ICMP_SOURCEQUENCH as
2019 * required in RFC1812 Requirements for IP Version 4 Routers.
2020 * Source quench could be a big problem under DoS attacks,
2021 * or if the underlying interface is rate-limited.
2022 * Those who need source quench packets may re-enable them
2023 * via the net.inet.ip.sendsourcequench sysctl.
2025 if (ip_sendsourcequench == 0) {
2029 type = ICMP_SOURCEQUENCH;
2034 case EACCES: /* ipfw denied packet */
2038 icmp_error(mcopy, type, code, dest, destifp);
2042 ip_savecontrol(inp, mp, ip, m)
2048 if (inp->inp_socket->so_options & SO_TIMESTAMP) {
2052 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
2053 SCM_TIMESTAMP, SOL_SOCKET);
2055 mp = &(*mp)->m_next;
2057 if (inp->inp_flags & INP_RECVDSTADDR) {
2058 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
2059 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
2061 mp = &(*mp)->m_next;
2065 * Moving these out of udp_input() made them even more broken
2066 * than they already were.
2068 /* options were tossed already */
2069 if (inp->inp_flags & INP_RECVOPTS) {
2070 *mp = sbcreatecontrol((caddr_t) opts_deleted_above,
2071 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
2073 mp = &(*mp)->m_next;
2075 /* ip_srcroute doesn't do what we want here, need to fix */
2076 if (inp->inp_flags & INP_RECVRETOPTS) {
2077 *mp = sbcreatecontrol((caddr_t) ip_srcroute(),
2078 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
2080 mp = &(*mp)->m_next;
2083 if (inp->inp_flags & INP_RECVIF) {
2086 struct sockaddr_dl sdl;
2089 struct sockaddr_dl *sdp;
2090 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
2092 if (((ifp = m->m_pkthdr.rcvif))
2093 && ( ifp->if_index && (ifp->if_index <= if_index))) {
2094 sdp = (struct sockaddr_dl *)(ifnet_addrs
2095 [ifp->if_index - 1]->ifa_addr);
2097 * Change our mind and don't try copy.
2099 if ((sdp->sdl_family != AF_LINK)
2100 || (sdp->sdl_len > sizeof(sdlbuf))) {
2103 bcopy(sdp, sdl2, sdp->sdl_len);
2107 = offsetof(struct sockaddr_dl, sdl_data[0]);
2108 sdl2->sdl_family = AF_LINK;
2109 sdl2->sdl_index = 0;
2110 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
2112 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len,
2113 IP_RECVIF, IPPROTO_IP);
2115 mp = &(*mp)->m_next;
2120 * XXX these routines are called from the upper part of the kernel.
2122 * They could also be moved to ip_mroute.c, since all the RSVP
2123 * handling is done there already.
2126 ip_rsvp_init(struct socket *so)
2128 if (so->so_type != SOCK_RAW ||
2129 so->so_proto->pr_protocol != IPPROTO_RSVP)
2132 if (ip_rsvpd != NULL)
2137 * This may seem silly, but we need to be sure we don't over-increment
2138 * the RSVP counter, in case something slips up.
2153 * This may seem silly, but we need to be sure we don't over-decrement
2154 * the RSVP counter, in case something slips up.
2164 rsvp_input(struct mbuf *m, int off, int proto) /* XXX must fixup manually */
2166 if (rsvp_input_p) { /* call the real one if loaded */
2167 rsvp_input_p(m, off, proto);
2171 /* Can still get packets with rsvp_on = 0 if there is a local member
2172 * of the group to which the RSVP packet is addressed. But in this
2173 * case we want to throw the packet away.
2181 if (ip_rsvpd != NULL) {
2182 rip_input(m, off, proto);
2185 /* Drop the packet */