1 //==========================================================================
3 // src/sys/netinet/ip_input.c
5 //==========================================================================
6 //####BSDCOPYRIGHTBEGIN####
8 // -------------------------------------------
10 // Portions of this software may have been derived from OpenBSD,
11 // FreeBSD or other sources, and are covered by the appropriate
12 // copyright disclaimers included herein.
14 // Portions created by Red Hat are
15 // Copyright (C) 2002 Red Hat, Inc. All Rights Reserved.
17 // -------------------------------------------
19 //####BSDCOPYRIGHTEND####
20 //==========================================================================
23 * Copyright (c) 1982, 1986, 1988, 1993
24 * The Regents of the University of California. All rights reserved.
26 * Redistribution and use in source and binary forms, with or without
27 * modification, are permitted provided that the following conditions
29 * 1. Redistributions of source code must retain the above copyright
30 * notice, this list of conditions and the following disclaimer.
31 * 2. Redistributions in binary form must reproduce the above copyright
32 * notice, this list of conditions and the following disclaimer in the
33 * documentation and/or other materials provided with the distribution.
34 * 3. All advertising materials mentioning features or use of this software
35 * must display the following acknowledgement:
36 * This product includes software developed by the University of
37 * California, Berkeley and its contributors.
38 * 4. Neither the name of the University nor the names of its contributors
39 * may be used to endorse or promote products derived from this software
40 * without specific prior written permission.
42 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
43 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
44 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
45 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
46 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
47 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
48 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
49 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
50 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
51 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
54 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94
55 * $FreeBSD: src/sys/netinet/ip_input.c,v 1.130.2.25 2001/08/29 21:41:37 jesper Exp $
60 #include <sys/param.h>
62 #include <sys/malloc.h>
63 #include <sys/domain.h>
64 #include <sys/protosw.h>
65 #include <sys/socket.h>
66 #include <sys/sysctl.h>
69 #include <net/if_var.h>
70 #include <net/if_dl.h>
71 #include <net/route.h>
72 #include <net/netisr.h>
73 #include <net/intrq.h>
75 #include <netinet/in.h>
76 #include <netinet/in_systm.h>
77 #include <netinet/in_var.h>
78 #include <netinet/ip.h>
79 #include <netinet/in_pcb.h>
80 #include <netinet/ip_var.h>
81 #include <netinet/ip_icmp.h>
83 #include <sys/socketvar.h>
85 #include <netinet/ip_fw.h>
88 #include <netinet6/ipsec.h>
89 #include <netkey/key.h>
93 #include <netinet/ip_dummynet.h>
97 static int ip_rsvp_on;
98 struct socket *ip_rsvpd;
100 int ipforwarding = 0;
101 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW,
102 &ipforwarding, 0, "Enable IP forwarding between interfaces");
104 static int ipsendredirects = 1; /* XXX */
105 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW,
106 &ipsendredirects, 0, "Enable sending IP redirects");
108 int ip_defttl = IPDEFTTL;
109 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW,
110 &ip_defttl, 0, "Maximum TTL on IP packets");
112 static int ip_dosourceroute = 0;
113 SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW,
114 &ip_dosourceroute, 0, "Enable forwarding source routed IP packets");
116 static int ip_acceptsourceroute = 0;
117 SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute,
118 CTLFLAG_RW, &ip_acceptsourceroute, 0,
119 "Enable accepting source routed IP packets");
121 #if defined(NFAITH) && 0 < NFAITH || defined(CYGPKG_NET_FREEBSD_SYSCTL)
122 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 #ifdef CYGPKG_NET_FREEBSD_SYSCTL
130 static int ip_maxfragpackets; /* initialized in ip_init() */
132 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragpackets, CTLFLAG_RW,
133 &ip_maxfragpackets, 0,
134 "Maximum number of IPv4 fragment reassembly queue entries");
136 static int nipq = 0; /* total # of reass queues */
140 * XXX - Setting ip_checkinterface mostly implements the receive side of
141 * the Strong ES model described in RFC 1122, but since the routing table
142 * and transmit implementation do not implement the Strong ES model,
143 * setting this to 1 results in an odd hybrid.
145 * XXX - ip_checkinterface currently must be disabled if you use ipnat
146 * to translate the destination address to another local interface.
148 * XXX - ip_checkinterface must be disabled if you add IP aliases
149 * to the loopback interface instead of the interface where the
150 * packets for those addresses are received.
152 static int ip_checkinterface = 0;
153 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW,
154 &ip_checkinterface, 0, "Verify packet arrives on correct interface");
157 static int ipprintfs = 0;
160 extern struct domain inetdomain;
161 extern struct protosw inetsw[];
162 u_char ip_protox[IPPROTO_MAX];
163 static int ipqmaxlen = IFQ_MAXLEN;
164 struct in_ifaddrhead in_ifaddrhead; /* first inet address */
165 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW,
166 &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue");
167 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD,
168 &ipintrq.ifq_drops, 0, "Number of packets dropped from the IP input queue");
170 struct ipstat ipstat;
171 SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RD,
172 &ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)");
174 /* Packet reassembly stuff */
175 #define IPREASS_NHASH_LOG2 6
176 #define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2)
177 #define IPREASS_HMASK (IPREASS_NHASH - 1)
178 #define IPREASS_HASH(x,y) \
179 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK)
181 static struct ipq ipq[IPREASS_NHASH];
182 const int ipintrq_present = 1;
185 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
186 &ip_mtu, 0, "Default MTU");
190 static int ipstealth = 0;
191 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW,
197 ip_fw_chk_t *ip_fw_chk_ptr;
198 ip_fw_ctl_t *ip_fw_ctl_ptr;
202 ip_dn_ctl_t *ip_dn_ctl_ptr;
205 int (*fr_checkp) __P((struct ip *, int, struct ifnet *, int, struct mbuf **)) = NULL;
209 * We need to save the IP options in case a protocol wants to respond
210 * to an incoming packet over the same route if the packet got here
211 * using IP source routing. This allows connection establishment and
212 * maintenance when the remote end is on a network that is not known
215 static int ip_nhops = 0;
216 static struct ip_srcrt {
217 struct in_addr dst; /* final destination */
218 char nop; /* one NOP to align */
219 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */
220 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)];
223 struct sockaddr_in *ip_fw_fwd_addr;
225 static void save_rte __P((u_char *, struct in_addr));
226 static int ip_dooptions __P((struct mbuf *));
228 void ip_forward __P((struct mbuf *, int));
230 static void ip_forward __P((struct mbuf *, int));
232 static void ip_freef __P((struct ipq *));
234 static struct mbuf *ip_reass __P((struct mbuf *,
235 struct ipq *, struct ipq *, u_int32_t *, u_int16_t *));
237 static struct mbuf *ip_reass __P((struct mbuf *, struct ipq *, struct ipq *));
239 static struct in_ifaddr *ip_rtaddr __P((struct in_addr));
240 static void ipintr __P((void));
243 extern int ip6_protocol_tr;
244 int natpt_in4 __P((struct mbuf *, struct mbuf **));
245 extern void ip6_forward __P((struct mbuf *, int));
249 * IP initialization: fill in IP protocol switch table.
250 * All protocols not implemented in kernel go to raw IP protocol handler.
255 register struct protosw *pr;
258 TAILQ_INIT(&in_ifaddrhead);
259 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
262 for (i = 0; i < IPPROTO_MAX; i++)
263 ip_protox[i] = pr - inetsw;
264 for (pr = inetdomain.dom_protosw;
265 pr < inetdomain.dom_protoswNPROTOSW; pr++)
266 if (pr->pr_domain->dom_family == PF_INET &&
267 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
268 ip_protox[pr->pr_protocol] = pr - inetsw;
270 for (i = 0; i < IPREASS_NHASH; i++)
271 ipq[i].next = ipq[i].prev = &ipq[i];
273 maxnipq = nmbclusters / 4;
276 ip_id = time_second & 0xffff;
278 ipintrq.ifq_maxlen = ipqmaxlen;
280 register_netisr(NETISR_IP, ipintr);
283 static struct sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET };
284 static struct route ipforward_rt;
287 * Ip input routine. Checksum and byte swap header. If fragmented
288 * try to reassemble. Process options. Pass to next level.
291 ip_input(struct mbuf *m)
295 struct in_ifaddr *ia = NULL;
296 int i, hlen, mff, checkif;
298 u_int16_t divert_cookie; /* firewall cookie */
299 struct in_addr pkt_dst;
301 u_int32_t divert_info = 0; /* packet divert/tee info */
303 struct ip_fw_chain *rule = NULL;
306 /* Get and reset firewall cookie */
307 divert_cookie = ip_divert_cookie;
308 ip_divert_cookie = 0;
313 #if defined(IPFIREWALL) && defined(DUMMYNET)
315 * dummynet packet are prepended a vestigial mbuf with
316 * m_type = MT_DUMMYNET and m_data pointing to the matching
319 if (m->m_type == MT_DUMMYNET) {
320 rule = (struct ip_fw_chain *)(m->m_data) ;
322 ip = mtod(m, struct ip *);
323 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
330 if (m == NULL || (m->m_flags & M_PKTHDR) == 0)
331 panic("ip_input no HDR");
335 if (m->m_pkthdr.len < sizeof(struct ip))
338 if (m->m_len < sizeof (struct ip) &&
339 (m = m_pullup(m, sizeof (struct ip))) == 0) {
340 ipstat.ips_toosmall++;
343 ip = mtod(m, struct ip *);
345 if (IP_VHL_V(ip->ip_vhl) != IPVERSION) {
346 ipstat.ips_badvers++;
350 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
351 if (hlen < sizeof(struct ip)) { /* minimum header length */
352 ipstat.ips_badhlen++;
355 if (hlen > m->m_len) {
356 if ((m = m_pullup(m, hlen)) == 0) {
357 ipstat.ips_badhlen++;
360 ip = mtod(m, struct ip *);
363 /* 127/8 must not appear on wire - RFC1122 */
364 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
365 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
366 if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
367 ipstat.ips_badaddr++;
372 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
373 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
375 if (hlen == sizeof(struct ip)) {
376 sum = in_cksum_hdr(ip);
378 sum = in_cksum(m, hlen);
387 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
388 /* packet is dropped by traffic conditioner */
392 * Convert fields to host representation.
395 if (ip->ip_len < hlen) {
402 * Check that the amount of data in the buffers
403 * is as at least much as the IP header would have us expect.
404 * Trim mbufs if longer than we expect.
405 * Drop packet if shorter than we expect.
407 if (m->m_pkthdr.len < ip->ip_len) {
409 ipstat.ips_tooshort++;
412 if (m->m_pkthdr.len > ip->ip_len) {
413 if (m->m_len == m->m_pkthdr.len) {
414 m->m_len = ip->ip_len;
415 m->m_pkthdr.len = ip->ip_len;
417 m_adj(m, ip->ip_len - m->m_pkthdr.len);
421 if (ipsec_getnhist(m))
427 * Right now when no processing on packet has done
428 * and it is still fresh out of network we do our black
430 * - Firewall: deny/allow/divert
431 * - Xlate: translate packet's addr/port (NAT).
432 * - Pipe: pass pkt through dummynet.
433 * - Wrap: fake packet's addr/port <unimpl.>
434 * - Encapsulate: put it in another IP and send out. <unimp.>
437 #if defined(IPFIREWALL) && defined(DUMMYNET)
441 * Check if we want to allow this packet to be processed.
442 * Consider it to be bad if not.
447 if ((*fr_checkp)(ip, hlen, m->m_pkthdr.rcvif, 0, &m1) || !m1)
449 ip = mtod(m = m1, struct ip *);
451 if (fw_enable && ip_fw_chk_ptr) {
452 #ifdef IPFIREWALL_FORWARD
454 * If we've been forwarded from the output side, then
455 * skip the firewall a second time
459 #endif /* IPFIREWALL_FORWARD */
461 * See the comment in ip_output for the return values
462 * produced by the firewall.
464 i = (*ip_fw_chk_ptr)(&ip,
465 hlen, NULL, &divert_cookie, &m, &rule, &ip_fw_fwd_addr);
466 if ( (i & IP_FW_PORT_DENY_FLAG) || m == NULL) { /* drop */
471 ip = mtod(m, struct ip *); /* just in case m changed */
472 if (i == 0 && ip_fw_fwd_addr == NULL) /* common case */
475 if ((i & IP_FW_PORT_DYNT_FLAG) != 0) {
476 /* Send packet to the appropriate pipe */
477 dummynet_io(i&0xffff,DN_TO_IP_IN,m,NULL,NULL,0, rule,
483 if (i != 0 && (i & IP_FW_PORT_DYNT_FLAG) == 0) {
484 /* Divert or tee packet */
489 #ifdef IPFIREWALL_FORWARD
490 if (i == 0 && ip_fw_fwd_addr != NULL)
494 * if we get here, the packet must be dropped
502 * Process options and, if not destined for us,
503 * ship it on. ip_dooptions returns 1 when an
504 * error was detected (causing an icmp message
505 * to be sent and the original packet to be freed).
507 ip_nhops = 0; /* for source routed packets */
508 if (hlen > sizeof (struct ip) && ip_dooptions(m)) {
509 #ifdef IPFIREWALL_FORWARD
510 ip_fw_fwd_addr = NULL;
515 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
516 * matter if it is destined to another node, or whether it is
517 * a multicast one, RSVP wants it! and prevents it from being forwarded
518 * anywhere else. Also checks if the rsvp daemon is running before
519 * grabbing the packet.
521 if (rsvp_on && ip->ip_p==IPPROTO_RSVP)
526 * NATPT (Network Address Translation - Protocol Translation)
528 if (ip6_protocol_tr) {
529 struct mbuf *m1 = NULL;
531 switch (natpt_in4(m, &m1)) {
532 case IPPROTO_IP: /* this packet is not changed */
543 case IPPROTO_DONE: /* discard without free */
546 case IPPROTO_MAX: /* discard this packet */
560 * Check our list of addresses, to see if the packet is for us.
561 * If we don't have any addresses, assume any unicast packet
562 * we receive might be for us (and let the upper layers deal
565 if (TAILQ_EMPTY(&in_ifaddrhead) &&
566 (m->m_flags & (M_MCAST|M_BCAST)) == 0)
570 * Cache the destination address of the packet; this may be
571 * changed by use of 'ipfw fwd'.
573 pkt_dst = ip_fw_fwd_addr == NULL ?
574 ip->ip_dst : ip_fw_fwd_addr->sin_addr;
577 * Enable a consistency check between the destination address
578 * and the arrival interface for a unicast packet (the RFC 1122
579 * strong ES model) if IP forwarding is disabled and the packet
580 * is not locally generated and the packet is not subject to
583 * XXX - Checking also should be disabled if the destination
584 * address is ipnat'ed to a different interface.
586 * XXX - Checking is incompatible with IP aliases added
587 * to the loopback interface instead of the interface where
588 * the packets are received.
590 checkif = ip_checkinterface && (ipforwarding == 0) &&
591 ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) &&
592 (ip_fw_fwd_addr == NULL);
594 TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link) {
595 #define satosin(sa) ((struct sockaddr_in *)(sa))
598 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY)
602 * If the address matches, verify that the packet
603 * arrived via the correct interface if checking is
606 if (IA_SIN(ia)->sin_addr.s_addr == pkt_dst.s_addr &&
607 (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif))
610 * Only accept broadcast packets that arrive via the
611 * matching interface. Reception of forwarded directed
612 * broadcasts would be handled via ip_forward() and
613 * ether_output() with the loopback into the stack for
614 * SIMPLEX interfaces handled by ether_output().
616 if (ia->ia_ifp == m->m_pkthdr.rcvif &&
617 ia->ia_ifp && ia->ia_ifp->if_flags & IFF_BROADCAST) {
618 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
621 if (ia->ia_netbroadcast.s_addr == pkt_dst.s_addr)
625 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
626 struct in_multi *inm;
629 * If we are acting as a multicast router, all
630 * incoming multicast packets are passed to the
631 * kernel-level multicast forwarding function.
632 * The packet is returned (relatively) intact; if
633 * ip_mforward() returns a non-zero value, the packet
634 * must be discarded, else it may be accepted below.
636 if (ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) {
637 ipstat.ips_cantforward++;
643 * The process-level routing demon needs to receive
644 * all multicast IGMP packets, whether or not this
645 * host belongs to their destination groups.
647 if (ip->ip_p == IPPROTO_IGMP)
649 ipstat.ips_forward++;
652 * See if we belong to the destination multicast group on the
655 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
657 ipstat.ips_notmember++;
663 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
665 if (ip->ip_dst.s_addr == INADDR_ANY)
668 #if defined(NFAITH) && 0 < NFAITH
670 * FAITH(Firewall Aided Internet Translator)
672 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
674 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP)
682 * Not for us; forward if possible and desirable.
684 if (ipforwarding == 0) {
685 ipstat.ips_cantforward++;
689 #ifdef IPFIREWALL_FORWARD
690 ip_fw_fwd_addr = NULL;
695 /* Count the packet in the ip address stats */
697 ia->ia_ifa.if_ipackets++;
698 ia->ia_ifa.if_ibytes += m->m_pkthdr.len;
702 * If offset or IP_MF are set, must reassemble.
703 * Otherwise, nothing need be done.
704 * (We could look in the reassembly queue to see
705 * if the packet was previously fragmented,
706 * but it's not worth the time; just let them time out.)
708 if (ip->ip_off & (IP_MF | IP_OFFMASK | IP_RF)) {
711 * Reassembly should be able to treat a mbuf cluster, for later
712 * operation of contiguous protocol headers on the cluster. (KAME)
714 if (m->m_flags & M_EXT) { /* XXX */
715 if ((m = m_pullup(m, hlen)) == 0) {
716 ipstat.ips_toosmall++;
717 #ifdef IPFIREWALL_FORWARD
718 ip_fw_fwd_addr = NULL;
722 ip = mtod(m, struct ip *);
726 /* If maxnipq is 0, never accept fragments. */
728 ipstat.ips_fragments++;
729 ipstat.ips_fragdropped++;
733 sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
735 * Look for queue of fragments
738 for (fp = ipq[sum].next; fp != &ipq[sum]; fp = fp->next)
739 if (ip->ip_id == fp->ipq_id &&
740 ip->ip_src.s_addr == fp->ipq_src.s_addr &&
741 ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
742 ip->ip_p == fp->ipq_p)
748 * Enforce upper bound on number of fragmented packets
749 * for which we attempt reassembly;
750 * If maxnipq is -1, accept all fragments without limitation.
752 if ((nipq > maxnipq) && (maxnipq > 0)) {
754 * drop something from the tail of the current queue
755 * before proceeding further
757 if (ipq[sum].prev == &ipq[sum]) { /* gak */
758 for (i = 0; i < IPREASS_NHASH; i++) {
759 if (ipq[i].prev != &ipq[i]) {
760 ip_freef(ipq[i].prev);
761 ipstat.ips_fragtimeout++;
766 ip_freef(ipq[sum].prev);
767 ipstat.ips_fragtimeout++;
772 * Adjust ip_len to not reflect header,
773 * set ip_mff if more fragments are expected,
774 * convert offset of this to bytes.
777 mff = (ip->ip_off & IP_MF) != 0;
780 * Make sure that fragments have a data length
781 * that's a non-zero multiple of 8 bytes.
783 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
784 ipstat.ips_toosmall++; /* XXX */
787 m->m_flags |= M_FRAG;
792 * If datagram marked as having more fragments
793 * or if this is not the first fragment,
794 * attempt reassembly; if it succeeds, proceed.
796 if (mff || ip->ip_off) {
797 ipstat.ips_fragments++;
798 m->m_pkthdr.header = ip;
801 fp, &ipq[sum], &divert_info, &divert_cookie);
803 m = ip_reass(m, fp, &ipq[sum]);
806 #ifdef IPFIREWALL_FORWARD
807 ip_fw_fwd_addr = NULL;
811 ipstat.ips_reassembled++;
812 ip = mtod(m, struct ip *);
813 /* Get the header length of the reassembled packet */
814 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
816 /* Restore original checksum before diverting packet */
817 if (divert_info != 0) {
822 if (hlen == sizeof(struct ip))
823 ip->ip_sum = in_cksum_hdr(ip);
825 ip->ip_sum = in_cksum(m, hlen);
839 * Divert or tee packet to the divert protocol if required.
841 * If divert_info is zero then cookie should be too, so we shouldn't
842 * need to clear them here. Assume divert_packet() does so also.
844 if (divert_info != 0) {
845 struct mbuf *clone = NULL;
847 /* Clone packet if we're doing a 'tee' */
848 if ((divert_info & IP_FW_PORT_TEE_FLAG) != 0)
849 clone = m_dup(m, M_DONTWAIT);
851 /* Restore packet header fields to original values */
856 /* Deliver packet to divert input routine */
857 ip_divert_cookie = divert_cookie;
858 divert_packet(m, 1, divert_info & 0xffff);
859 ipstat.ips_delivered++;
861 /* If 'tee', continue with original packet */
865 ip = mtod(m, struct ip *);
871 * enforce IPsec policy checking if we are seeing last header.
872 * note that we do not visit this with protocols with pcb layer
873 * code - like udp/tcp/raw ip.
875 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0 &&
876 ipsec4_in_reject(m, NULL)) {
877 ipsecstat.in_polvio++;
883 * Switch out to protocol's input routine.
885 ipstat.ips_delivered++;
889 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, off);
890 #ifdef IPFIREWALL_FORWARD
891 ip_fw_fwd_addr = NULL; /* tcp needed it */
896 #ifdef IPFIREWALL_FORWARD
897 ip_fw_fwd_addr = NULL;
903 * IP software interrupt routine - to go away sometime soon
913 IF_DEQUEUE(&ipintrq, m);
922 * Take incoming datagram fragment and try to reassemble it into
923 * whole datagram. If a chain for reassembly of this datagram already
924 * exists, then it is given as fp; otherwise have to make a chain.
926 * When IPDIVERT enabled, keep additional state with each packet that
927 * tells us if we need to divert or tee the packet we're building.
932 ip_reass(m, fp, where, divinfo, divcookie)
934 ip_reass(m, fp, where)
936 register struct mbuf *m;
937 register struct ipq *fp;
941 u_int16_t *divcookie;
944 struct ip *ip = mtod(m, struct ip *);
945 register struct mbuf *p = 0, *q, *nq;
947 int hlen = IP_VHL_HL(ip->ip_vhl) << 2;
951 * Presence of header sizes in mbufs
952 * would confuse code below.
958 * If first fragment to arrive, create a reassembly queue.
961 if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL)
963 fp = mtod(t, struct ipq *);
966 fp->ipq_ttl = IPFRAGTTL;
967 fp->ipq_p = ip->ip_p;
968 fp->ipq_id = ip->ip_id;
969 fp->ipq_src = ip->ip_src;
970 fp->ipq_dst = ip->ip_dst;
974 fp->ipq_div_info = 0;
975 fp->ipq_div_cookie = 0;
980 #define GETIP(m) ((struct ip*)((m)->m_pkthdr.header))
983 * Find a segment which begins after this one does.
985 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt)
986 if (GETIP(q)->ip_off > ip->ip_off)
990 * If there is a preceding segment, it may provide some of
991 * our data already. If so, drop the data from the incoming
992 * segment. If it provides all of our data, drop us, otherwise
993 * stick new segment in the proper place.
995 * If some of the data is dropped from the the preceding
996 * segment, then it's checksum is invalidated.
999 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off;
1001 if (i >= ip->ip_len)
1004 m->m_pkthdr.csum_flags = 0;
1008 m->m_nextpkt = p->m_nextpkt;
1011 m->m_nextpkt = fp->ipq_frags;
1016 * While we overlap succeeding segments trim them or,
1017 * if they are completely covered, dequeue them.
1019 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off;
1021 i = (ip->ip_off + ip->ip_len) -
1023 if (i < GETIP(q)->ip_len) {
1024 GETIP(q)->ip_len -= i;
1025 GETIP(q)->ip_off += i;
1027 q->m_pkthdr.csum_flags = 0;
1039 * Transfer firewall instructions to the fragment structure.
1040 * Any fragment diverting causes the whole packet to divert.
1042 fp->ipq_div_info = *divinfo;
1043 fp->ipq_div_cookie = *divcookie;
1049 * Check for complete reassembly.
1052 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
1053 if (GETIP(q)->ip_off != next)
1055 next += GETIP(q)->ip_len;
1057 /* Make sure the last packet didn't have the IP_MF flag */
1058 if (p->m_flags & M_FRAG)
1062 * Reassembly is complete. Make sure the packet is a sane size.
1066 if (next + (IP_VHL_HL(ip->ip_vhl) << 2) > IP_MAXPACKET) {
1067 ipstat.ips_toolong++;
1073 * Concatenate fragments.
1081 for (q = nq; q != NULL; q = nq) {
1083 q->m_nextpkt = NULL;
1084 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
1085 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
1091 * Extract firewall instructions from the fragment structure.
1093 *divinfo = fp->ipq_div_info;
1094 *divcookie = fp->ipq_div_cookie;
1098 * Create header for new ip packet by
1099 * modifying header of first packet;
1100 * dequeue and discard fragment reassembly header.
1101 * Make header visible.
1104 ip->ip_src = fp->ipq_src;
1105 ip->ip_dst = fp->ipq_dst;
1108 (void) m_free(dtom(fp));
1109 m->m_len += (IP_VHL_HL(ip->ip_vhl) << 2);
1110 m->m_data -= (IP_VHL_HL(ip->ip_vhl) << 2);
1111 /* some debugging cruft by sklower, below, will go away soon */
1112 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */
1113 register int plen = 0;
1114 for (t = m; t; t = t->m_next)
1116 m->m_pkthdr.len = plen;
1125 ipstat.ips_fragdropped++;
1133 * Free a fragment reassembly header and all
1134 * associated datagrams.
1140 register struct mbuf *q;
1142 while (fp->ipq_frags) {
1144 fp->ipq_frags = q->m_nextpkt;
1148 (void) m_free(dtom(fp));
1153 * IP timer processing;
1154 * if a timer expires on a reassembly
1155 * queue, discard it.
1160 register struct ipq *fp;
1164 for (i = 0; i < IPREASS_NHASH; i++) {
1168 while (fp != &ipq[i]) {
1171 if (fp->prev->ipq_ttl == 0) {
1172 ipstat.ips_fragtimeout++;
1178 * If we are over the maximum number of fragments
1179 * (due to the limit being lowered), drain off
1180 * enough to get down to the new limit.
1182 if (maxnipq >= 0 && nipq > maxnipq) {
1183 for (i = 0; i < IPREASS_NHASH; i++) {
1184 while ((nipq > maxnipq) &&
1185 (ipq[i].next != &ipq[i])) {
1186 ipstat.ips_fragdropped++;
1187 ip_freef(ipq[i].next);
1196 * Drain off all datagram fragments.
1203 for (i = 0; i < IPREASS_NHASH; i++) {
1204 while (ipq[i].next != &ipq[i]) {
1205 ipstat.ips_fragdropped++;
1206 ip_freef(ipq[i].next);
1213 * Do option processing on a datagram,
1214 * possibly discarding it if bad options are encountered,
1215 * or forwarding it if source-routed.
1216 * Returns 1 if packet has been forwarded/freed,
1217 * 0 if the packet should be processed further.
1223 register struct ip *ip = mtod(m, struct ip *);
1224 register u_char *cp;
1225 register struct ip_timestamp *ipt;
1226 register struct in_ifaddr *ia;
1227 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0;
1228 struct in_addr *sin, dst;
1232 cp = (u_char *)(ip + 1);
1233 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip);
1234 for (; cnt > 0; cnt -= optlen, cp += optlen) {
1235 opt = cp[IPOPT_OPTVAL];
1236 if (opt == IPOPT_EOL)
1238 if (opt == IPOPT_NOP)
1241 if (cnt < IPOPT_OLEN + sizeof(*cp)) {
1242 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1245 optlen = cp[IPOPT_OLEN];
1246 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
1247 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1257 * Source routing with record.
1258 * Find interface with current destination address.
1259 * If none on this machine then drop if strictly routed,
1260 * or do nothing if loosely routed.
1261 * Record interface address and bring up next address
1262 * component. If strictly routed make sure next
1263 * address is on directly accessible net.
1267 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1268 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1271 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1272 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1275 ipaddr.sin_addr = ip->ip_dst;
1276 ia = (struct in_ifaddr *)
1277 ifa_ifwithaddr((struct sockaddr *)&ipaddr);
1279 if (opt == IPOPT_SSRR) {
1280 type = ICMP_UNREACH;
1281 code = ICMP_UNREACH_SRCFAIL;
1284 if (!ip_dosourceroute)
1285 goto nosourcerouting;
1287 * Loose routing, and not at next destination
1288 * yet; nothing to do except forward.
1292 off--; /* 0 origin */
1293 if (off > optlen - (int)sizeof(struct in_addr)) {
1295 * End of source route. Should be for us.
1297 if (!ip_acceptsourceroute)
1298 goto nosourcerouting;
1299 save_rte(cp, ip->ip_src);
1303 if (!ip_dosourceroute) {
1305 char buf[16]; /* aaa.bbb.ccc.ddd\0 */
1307 * Acting as a router, so generate ICMP
1310 strcpy(buf, inet_ntoa(ip->ip_dst));
1312 "attempted source route from %s to %s\n",
1313 inet_ntoa(ip->ip_src), buf);
1314 type = ICMP_UNREACH;
1315 code = ICMP_UNREACH_SRCFAIL;
1319 * Not acting as a router, so silently drop.
1321 ipstat.ips_cantforward++;
1328 * locate outgoing interface
1330 (void)memcpy(&ipaddr.sin_addr, cp + off,
1331 sizeof(ipaddr.sin_addr));
1333 if (opt == IPOPT_SSRR) {
1334 #define INA struct in_ifaddr *
1335 #define SA struct sockaddr *
1336 if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == 0)
1337 ia = (INA)ifa_ifwithnet((SA)&ipaddr);
1339 ia = ip_rtaddr(ipaddr.sin_addr);
1341 type = ICMP_UNREACH;
1342 code = ICMP_UNREACH_SRCFAIL;
1345 ip->ip_dst = ipaddr.sin_addr;
1346 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
1347 sizeof(struct in_addr));
1348 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1350 * Let ip_intr's mcast routing check handle mcast pkts
1352 forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr));
1356 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1357 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1360 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1361 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1365 * If no space remains, ignore.
1367 off--; /* 0 origin */
1368 if (off > optlen - (int)sizeof(struct in_addr))
1370 (void)memcpy(&ipaddr.sin_addr, &ip->ip_dst,
1371 sizeof(ipaddr.sin_addr));
1373 * locate outgoing interface; if we're the destination,
1374 * use the incoming interface (should be same).
1376 if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == 0 &&
1377 (ia = ip_rtaddr(ipaddr.sin_addr)) == 0) {
1378 type = ICMP_UNREACH;
1379 code = ICMP_UNREACH_HOST;
1382 (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
1383 sizeof(struct in_addr));
1384 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1388 code = cp - (u_char *)ip;
1389 ipt = (struct ip_timestamp *)cp;
1390 if (ipt->ipt_len < 4 || ipt->ipt_len > 40) {
1391 code = (u_char *)&ipt->ipt_len - (u_char *)ip;
1394 if (ipt->ipt_ptr < 5) {
1395 code = (u_char *)&ipt->ipt_ptr - (u_char *)ip;
1399 ipt->ipt_len - (int)sizeof(int32_t)) {
1400 if (++ipt->ipt_oflw == 0) {
1401 code = (u_char *)&ipt->ipt_ptr -
1407 sin = (struct in_addr *)(cp + ipt->ipt_ptr - 1);
1408 switch (ipt->ipt_flg) {
1410 case IPOPT_TS_TSONLY:
1413 case IPOPT_TS_TSANDADDR:
1414 if (ipt->ipt_ptr - 1 + sizeof(n_time) +
1415 sizeof(struct in_addr) > ipt->ipt_len) {
1416 code = (u_char *)&ipt->ipt_ptr -
1420 ipaddr.sin_addr = dst;
1421 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr,
1425 (void)memcpy(sin, &IA_SIN(ia)->sin_addr,
1426 sizeof(struct in_addr));
1427 ipt->ipt_ptr += sizeof(struct in_addr);
1430 case IPOPT_TS_PRESPEC:
1431 if (ipt->ipt_ptr - 1 + sizeof(n_time) +
1432 sizeof(struct in_addr) > ipt->ipt_len) {
1433 code = (u_char *)&ipt->ipt_ptr -
1437 (void)memcpy(&ipaddr.sin_addr, sin,
1438 sizeof(struct in_addr));
1439 if (ifa_ifwithaddr((SA)&ipaddr) == 0)
1441 ipt->ipt_ptr += sizeof(struct in_addr);
1445 /* XXX can't take &ipt->ipt_flg */
1446 code = (u_char *)&ipt->ipt_ptr -
1451 (void)memcpy(cp + ipt->ipt_ptr - 1, &ntime,
1453 ipt->ipt_ptr += sizeof(n_time);
1456 if (forward && ipforwarding) {
1462 icmp_error(m, type, code, 0, 0);
1463 ipstat.ips_badoptions++;
1468 * Given address of next destination (final or next hop),
1469 * return internet address info of interface to be used to get there.
1471 static struct in_ifaddr *
1475 register struct sockaddr_in *sin;
1477 sin = (struct sockaddr_in *) &ipforward_rt.ro_dst;
1479 if (ipforward_rt.ro_rt == 0 || dst.s_addr != sin->sin_addr.s_addr) {
1480 if (ipforward_rt.ro_rt) {
1481 RTFREE(ipforward_rt.ro_rt);
1482 ipforward_rt.ro_rt = 0;
1484 sin->sin_family = AF_INET;
1485 sin->sin_len = sizeof(*sin);
1486 sin->sin_addr = dst;
1488 rtalloc_ign(&ipforward_rt, RTF_PRCLONING);
1490 if (ipforward_rt.ro_rt == 0)
1491 return ((struct in_ifaddr *)0);
1492 return ((struct in_ifaddr *) ipforward_rt.ro_rt->rt_ifa);
1496 * Save incoming source route for use in replies,
1497 * to be picked up later by ip_srcroute if the receiver is interested.
1500 save_rte(option, dst)
1506 olen = option[IPOPT_OLEN];
1509 printf("save_rte: olen %d\n", olen);
1511 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst)))
1513 bcopy(option, ip_srcrt.srcopt, olen);
1514 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1519 * Retrieve incoming source route for use in replies,
1520 * in the same form used by setsockopt.
1521 * The first hop is placed before the options, will be removed later.
1526 register struct in_addr *p, *q;
1527 register struct mbuf *m;
1530 return ((struct mbuf *)0);
1531 m = m_get(M_DONTWAIT, MT_HEADER);
1533 return ((struct mbuf *)0);
1535 #define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt))
1537 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
1538 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) +
1542 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len);
1546 * First save first hop for return route
1548 p = &ip_srcrt.route[ip_nhops - 1];
1549 *(mtod(m, struct in_addr *)) = *p--;
1552 printf(" hops %lx", (u_long)ntohl(mtod(m, struct in_addr *)->s_addr));
1556 * Copy option fields and padding (nop) to mbuf.
1558 ip_srcrt.nop = IPOPT_NOP;
1559 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
1560 (void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr),
1561 &ip_srcrt.nop, OPTSIZ);
1562 q = (struct in_addr *)(mtod(m, caddr_t) +
1563 sizeof(struct in_addr) + OPTSIZ);
1566 * Record return path as an IP source route,
1567 * reversing the path (pointers are now aligned).
1569 while (p >= ip_srcrt.route) {
1572 printf(" %lx", (u_long)ntohl(q->s_addr));
1577 * Last hop goes to final destination.
1582 printf(" %lx\n", (u_long)ntohl(q->s_addr));
1588 * Strip out IP options, at higher
1589 * level protocol in the kernel.
1590 * Second argument is buffer to which options
1591 * will be moved, and return value is their length.
1592 * XXX should be deleted; last arg currently ignored.
1595 ip_stripoptions(m, mopt)
1596 register struct mbuf *m;
1600 struct ip *ip = mtod(m, struct ip *);
1601 register caddr_t opts;
1604 olen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip);
1605 opts = (caddr_t)(ip + 1);
1606 i = m->m_len - (sizeof (struct ip) + olen);
1607 bcopy(opts + olen, opts, (unsigned)i);
1609 if (m->m_flags & M_PKTHDR)
1610 m->m_pkthdr.len -= olen;
1611 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, sizeof(struct ip) >> 2);
1614 int inetctlerrmap[PRC_NCMDS] = {
1616 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
1617 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
1618 EMSGSIZE, EHOSTUNREACH, 0, 0,
1620 ENOPROTOOPT, ECONNREFUSED
1624 * Forward a packet. If some error occurs return the sender
1625 * an icmp packet. Note we can't always generate a meaningful
1626 * icmp message because icmp doesn't have a large enough repertoire
1627 * of codes and types.
1629 * If not forwarding, just drop the packet. This could be confusing
1630 * if ipforwarding was zero but some routing protocol was advancing
1631 * us as a gateway to somewhere. However, we must let the routing
1632 * protocol deal with that.
1634 * The srcrt parameter indicates whether the packet is being forwarded
1635 * via a source route.
1642 ip_forward(m, srcrt)
1646 register struct ip *ip = mtod(m, struct ip *);
1647 register struct sockaddr_in *sin;
1648 register struct rtentry *rt;
1649 int error, type = 0, code = 0;
1652 struct ifnet *destifp;
1654 struct ifnet dummyifp;
1660 printf("forward: src %lx dst %lx ttl %x\n",
1661 (u_long)ip->ip_src.s_addr, (u_long)ip->ip_dst.s_addr,
1666 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
1667 ipstat.ips_cantforward++;
1674 if (ip->ip_ttl <= IPTTLDEC) {
1675 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS,
1683 sin = (struct sockaddr_in *)&ipforward_rt.ro_dst;
1684 if ((rt = ipforward_rt.ro_rt) == 0 ||
1685 ip->ip_dst.s_addr != sin->sin_addr.s_addr) {
1686 if (ipforward_rt.ro_rt) {
1687 RTFREE(ipforward_rt.ro_rt);
1688 ipforward_rt.ro_rt = 0;
1690 sin->sin_family = AF_INET;
1691 sin->sin_len = sizeof(*sin);
1692 sin->sin_addr = ip->ip_dst;
1694 rtalloc_ign(&ipforward_rt, RTF_PRCLONING);
1695 if (ipforward_rt.ro_rt == 0) {
1696 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0);
1699 rt = ipforward_rt.ro_rt;
1703 * Save the IP header and at most 8 bytes of the payload,
1704 * in case we need to generate an ICMP message to the src.
1706 * We don't use m_copy() because it might return a reference
1707 * to a shared cluster. Both this function and ip_output()
1708 * assume exclusive access to the IP header in `m', so any
1709 * data in a cluster may change before we reach icmp_error().
1711 MGET(mcopy, M_DONTWAIT, m->m_type);
1712 if (mcopy != NULL) {
1713 M_COPY_PKTHDR(mcopy, m);
1714 mcopy->m_len = imin((IP_VHL_HL(ip->ip_vhl) << 2) + 8,
1716 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1722 ip->ip_ttl -= IPTTLDEC;
1728 * If forwarding packet using same interface that it came in on,
1729 * perhaps should send a redirect to sender to shortcut a hop.
1730 * Only send redirect if source is sending directly to us,
1731 * and if packet was not source routed (or has any options).
1732 * Also, don't send redirect if forwarding using a default route
1733 * or a route modified by a redirect.
1735 #define satosin(sa) ((struct sockaddr_in *)(sa))
1736 if (rt->rt_ifp == m->m_pkthdr.rcvif &&
1737 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1738 satosin(rt_key(rt))->sin_addr.s_addr != 0 &&
1739 ipsendredirects && !srcrt) {
1740 #define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa))
1741 u_long src = ntohl(ip->ip_src.s_addr);
1744 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1745 if (rt->rt_flags & RTF_GATEWAY)
1746 dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
1748 dest = ip->ip_dst.s_addr;
1749 /* Router requirements says to only send host redirects */
1750 type = ICMP_REDIRECT;
1751 code = ICMP_REDIRECT_HOST;
1754 printf("redirect (%d) to %lx\n", code, (u_long)dest);
1759 error = ip_output(m, (struct mbuf *)0, &ipforward_rt,
1762 ipstat.ips_cantforward++;
1764 ipstat.ips_forward++;
1766 ipstat.ips_redirectsent++;
1769 ipflow_create(&ipforward_rt, mcopy);
1781 case 0: /* forwarded, but need redirect */
1782 /* type, code set above */
1785 case ENETUNREACH: /* shouldn't happen, checked above */
1790 type = ICMP_UNREACH;
1791 code = ICMP_UNREACH_HOST;
1795 type = ICMP_UNREACH;
1796 code = ICMP_UNREACH_NEEDFRAG;
1798 if (ipforward_rt.ro_rt)
1799 destifp = ipforward_rt.ro_rt->rt_ifp;
1802 * If the packet is routed over IPsec tunnel, tell the
1803 * originator the tunnel MTU.
1804 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
1807 if (ipforward_rt.ro_rt) {
1808 struct secpolicy *sp = NULL;
1813 sp = ipsec4_getpolicybyaddr(mcopy,
1819 destifp = ipforward_rt.ro_rt->rt_ifp;
1821 /* count IPsec header size */
1822 ipsechdr = ipsec4_hdrsiz(mcopy,
1827 * find the correct route for outer IPv4
1828 * header, compute tunnel MTU.
1831 * The "dummyifp" code relies upon the fact
1832 * that icmp_error() touches only ifp->if_mtu.
1837 && sp->req->sav != NULL
1838 && sp->req->sav->sah != NULL) {
1839 ro = &sp->req->sav->sah->sa_route;
1840 if (ro->ro_rt && ro->ro_rt->rt_ifp) {
1842 ro->ro_rt->rt_ifp->if_mtu;
1843 dummyifp.if_mtu -= ipsechdr;
1844 destifp = &dummyifp;
1852 ipstat.ips_cantfrag++;
1858 * don't generate ICMP_SOURCEQUENCH
1859 * (RFC1812 Requirements for IP Version 4 Routers)
1865 type = ICMP_SOURCEQUENCH;
1870 case EACCES: /* ipfw denied packet */
1874 icmp_error(mcopy, type, code, dest, destifp);
1878 ip_savecontrol(inp, mp, ip, m)
1879 register struct inpcb *inp;
1880 register struct mbuf **mp;
1881 register struct ip *ip;
1882 register struct mbuf *m;
1884 if (inp->inp_socket->so_options & SO_TIMESTAMP) {
1888 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1889 SCM_TIMESTAMP, SOL_SOCKET);
1891 mp = &(*mp)->m_next;
1893 if (inp->inp_flags & INP_RECVDSTADDR) {
1894 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
1895 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1897 mp = &(*mp)->m_next;
1901 * Moving these out of udp_input() made them even more broken
1902 * than they already were.
1904 /* options were tossed already */
1905 if (inp->inp_flags & INP_RECVOPTS) {
1906 *mp = sbcreatecontrol((caddr_t) opts_deleted_above,
1907 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
1909 mp = &(*mp)->m_next;
1911 /* ip_srcroute doesn't do what we want here, need to fix */
1912 if (inp->inp_flags & INP_RECVRETOPTS) {
1913 *mp = sbcreatecontrol((caddr_t) ip_srcroute(),
1914 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
1916 mp = &(*mp)->m_next;
1919 if (inp->inp_flags & INP_RECVIF) {
1922 struct sockaddr_dl sdl;
1925 struct sockaddr_dl *sdp;
1926 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1928 if (((ifp = m->m_pkthdr.rcvif))
1929 && ( ifp->if_index && (ifp->if_index <= if_index))) {
1930 sdp = (struct sockaddr_dl *)(ifnet_addrs
1931 [ifp->if_index - 1]->ifa_addr);
1933 * Change our mind and don't try copy.
1935 if ((sdp->sdl_family != AF_LINK)
1936 || (sdp->sdl_len > sizeof(sdlbuf))) {
1939 bcopy(sdp, sdl2, sdp->sdl_len);
1943 = offsetof(struct sockaddr_dl, sdl_data[0]);
1944 sdl2->sdl_family = AF_LINK;
1945 sdl2->sdl_index = 0;
1946 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1948 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len,
1949 IP_RECVIF, IPPROTO_IP);
1951 mp = &(*mp)->m_next;
1956 ip_rsvp_init(struct socket *so)
1958 if (so->so_type != SOCK_RAW ||
1959 so->so_proto->pr_protocol != IPPROTO_RSVP)
1962 if (ip_rsvpd != NULL)
1967 * This may seem silly, but we need to be sure we don't over-increment
1968 * the RSVP counter, in case something slips up.
1983 * This may seem silly, but we need to be sure we don't over-decrement
1984 * the RSVP counter, in case something slips up.