2 * Copyright (c) 2003, 2004 Jeffrey M. Hsu. All rights reserved.
3 * Copyright (c) 2003, 2004 The DragonFly Project. All rights reserved.
5 * This code is derived from software contributed to The DragonFly Project
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. Neither the name of The DragonFly Project nor the names of its
17 * contributors may be used to endorse or promote products derived
18 * from this software without specific, prior written permission.
20 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
21 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
22 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
23 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
24 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
25 * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
26 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
27 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
28 * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
29 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
30 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
35 * Copyright (c) 1982, 1986, 1988, 1993
36 * The Regents of the University of California. All rights reserved.
38 * Redistribution and use in source and binary forms, with or without
39 * modification, are permitted provided that the following conditions
41 * 1. Redistributions of source code must retain the above copyright
42 * notice, this list of conditions and the following disclaimer.
43 * 2. Redistributions in binary form must reproduce the above copyright
44 * notice, this list of conditions and the following disclaimer in the
45 * documentation and/or other materials provided with the distribution.
46 * 3. Neither the name of the University nor the names of its contributors
47 * may be used to endorse or promote products derived from this software
48 * without specific prior written permission.
50 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
51 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
52 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
53 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
54 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
55 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
56 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
57 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
58 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
59 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
62 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94
63 * $FreeBSD: src/sys/netinet/ip_input.c,v 1.130.2.52 2003/03/07 07:01:28 silby Exp $
68 #include "opt_bootp.h"
70 #include "opt_ipdivert.h"
71 #include "opt_ipstealth.h"
72 #include "opt_ipsec.h"
75 #include <sys/param.h>
76 #include <sys/systm.h>
78 #include <sys/malloc.h>
79 #include <sys/mpipe.h>
80 #include <sys/domain.h>
81 #include <sys/protosw.h>
82 #include <sys/socket.h>
84 #include <sys/globaldata.h>
85 #include <sys/thread.h>
86 #include <sys/kernel.h>
87 #include <sys/syslog.h>
88 #include <sys/sysctl.h>
89 #include <sys/in_cksum.h>
92 #include <sys/mplock2.h>
94 #include <machine/stdarg.h>
97 #include <net/if_types.h>
98 #include <net/if_var.h>
99 #include <net/if_dl.h>
100 #include <net/pfil.h>
101 #include <net/route.h>
102 #include <net/netisr2.h>
104 #include <netinet/in.h>
105 #include <netinet/in_systm.h>
106 #include <netinet/in_var.h>
107 #include <netinet/ip.h>
108 #include <netinet/in_pcb.h>
109 #include <netinet/ip_var.h>
110 #include <netinet/ip_icmp.h>
111 #include <netinet/ip_divert.h>
112 #include <netinet/ip_flow.h>
114 #include <sys/thread2.h>
115 #include <sys/msgport2.h>
116 #include <net/netmsg2.h>
118 #include <sys/socketvar.h>
120 #include <net/ipfw/ip_fw.h>
121 #include <net/dummynet/ip_dummynet.h>
124 #include <netinet6/ipsec.h>
125 #include <netproto/key/key.h>
129 #include <netproto/ipsec/ipsec.h>
130 #include <netproto/ipsec/key.h>
134 static int ip_rsvp_on;
135 struct socket *ip_rsvpd;
137 int ipforwarding = 0;
138 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW,
139 &ipforwarding, 0, "Enable IP forwarding between interfaces");
141 static int ipsendredirects = 1; /* XXX */
142 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW,
143 &ipsendredirects, 0, "Enable sending IP redirects");
145 int ip_defttl = IPDEFTTL;
146 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW,
147 &ip_defttl, 0, "Maximum TTL on IP packets");
149 static int ip_dosourceroute = 0;
150 SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW,
151 &ip_dosourceroute, 0, "Enable forwarding source routed IP packets");
153 static int ip_acceptsourceroute = 0;
154 SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute,
155 CTLFLAG_RW, &ip_acceptsourceroute, 0,
156 "Enable accepting source routed IP packets");
158 static int ip_keepfaith = 0;
159 SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW,
161 "Enable packet capture for FAITH IPv4->IPv6 translator daemon");
163 static int nipq = 0; /* total # of reass queues */
165 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragpackets, CTLFLAG_RW,
167 "Maximum number of IPv4 fragment reassembly queue entries");
169 static int maxfragsperpacket;
170 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW,
171 &maxfragsperpacket, 0,
172 "Maximum number of IPv4 fragments allowed per packet");
174 static int ip_sendsourcequench = 0;
175 SYSCTL_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW,
176 &ip_sendsourcequench, 0,
177 "Enable the transmission of source quench packets");
179 int ip_do_randomid = 1;
180 SYSCTL_INT(_net_inet_ip, OID_AUTO, random_id, CTLFLAG_RW,
182 "Assign random ip_id values");
184 * XXX - Setting ip_checkinterface mostly implements the receive side of
185 * the Strong ES model described in RFC 1122, but since the routing table
186 * and transmit implementation do not implement the Strong ES model,
187 * setting this to 1 results in an odd hybrid.
189 * XXX - ip_checkinterface currently must be disabled if you use ipnat
190 * to translate the destination address to another local interface.
192 * XXX - ip_checkinterface must be disabled if you add IP aliases
193 * to the loopback interface instead of the interface where the
194 * packets for those addresses are received.
196 static int ip_checkinterface = 0;
197 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW,
198 &ip_checkinterface, 0, "Verify packet arrives on correct interface");
200 static u_long ip_hash_count = 0;
201 SYSCTL_ULONG(_net_inet_ip, OID_AUTO, hash_count, CTLFLAG_RD,
202 &ip_hash_count, 0, "Number of packets hashed by IP");
205 static u_long ip_rehash_count = 0;
206 SYSCTL_ULONG(_net_inet_ip, OID_AUTO, rehash_count, CTLFLAG_RD,
207 &ip_rehash_count, 0, "Number of packets rehashed by IP");
209 static u_long ip_dispatch_fast = 0;
210 SYSCTL_ULONG(_net_inet_ip, OID_AUTO, dispatch_fast_count, CTLFLAG_RD,
211 &ip_dispatch_fast, 0, "Number of packets handled on current CPU");
213 static u_long ip_dispatch_slow = 0;
214 SYSCTL_ULONG(_net_inet_ip, OID_AUTO, dispatch_slow_count, CTLFLAG_RD,
215 &ip_dispatch_slow, 0, "Number of packets messaged to another CPU");
218 static struct lwkt_token ipq_token = LWKT_TOKEN_INITIALIZER(ipq_token);
221 static int ipprintfs = 0;
224 extern struct domain inetdomain;
225 extern struct protosw inetsw[];
226 u_char ip_protox[IPPROTO_MAX];
227 struct in_ifaddrhead in_ifaddrheads[MAXCPU]; /* first inet address */
228 struct in_ifaddrhashhead *in_ifaddrhashtbls[MAXCPU];
229 /* inet addr hash table */
230 u_long in_ifaddrhmask; /* mask for hash table */
232 static struct mbuf *ipforward_mtemp[MAXCPU];
234 struct ip_stats ipstats_percpu[MAXCPU] __cachealign;
237 sysctl_ipstats(SYSCTL_HANDLER_ARGS)
241 for (cpu = 0; cpu < ncpus; ++cpu) {
242 if ((error = SYSCTL_OUT(req, &ipstats_percpu[cpu],
243 sizeof(struct ip_stats))))
245 if ((error = SYSCTL_IN(req, &ipstats_percpu[cpu],
246 sizeof(struct ip_stats))))
252 SYSCTL_PROC(_net_inet_ip, IPCTL_STATS, stats, (CTLTYPE_OPAQUE | CTLFLAG_RW),
253 0, 0, sysctl_ipstats, "S,ip_stats", "IP statistics");
255 /* Packet reassembly stuff */
256 #define IPREASS_NHASH_LOG2 6
257 #define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2)
258 #define IPREASS_HMASK (IPREASS_NHASH - 1)
259 #define IPREASS_HASH(x,y) \
260 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK)
262 static TAILQ_HEAD(ipqhead, ipq) ipq[IPREASS_NHASH];
265 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
266 &ip_mtu, 0, "Default MTU");
270 static int ipstealth = 0;
271 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW, &ipstealth, 0, "");
273 static const int ipstealth = 0;
276 struct mbuf *(*ip_divert_p)(struct mbuf *, int, int);
278 struct pfil_head inet_pfil_hook;
281 * struct ip_srcrt_opt is used to store packet state while it travels
284 * XXX Note that the code even makes assumptions on the size and
285 * alignment of fields inside struct ip_srcrt so e.g. adding some
286 * fields will break the code. This needs to be fixed.
288 * We need to save the IP options in case a protocol wants to respond
289 * to an incoming packet over the same route if the packet got here
290 * using IP source routing. This allows connection establishment and
291 * maintenance when the remote end is on a network that is not known
295 struct in_addr dst; /* final destination */
296 char nop; /* one NOP to align */
297 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */
298 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)];
301 struct ip_srcrt_opt {
303 struct ip_srcrt ip_srcrt;
306 static MALLOC_DEFINE(M_IPQ, "ipq", "IP Fragment Management");
307 static struct malloc_pipe ipq_mpipe;
309 static void save_rte(struct mbuf *, u_char *, struct in_addr);
310 static int ip_dooptions(struct mbuf *m, int, struct sockaddr_in *);
311 static void ip_freef(struct ipqhead *, struct ipq *);
312 static void ip_input_handler(netmsg_t);
315 * IP initialization: fill in IP protocol switch table.
316 * All protocols not implemented in kernel go to raw IP protocol handler.
326 * Make sure we can handle a reasonable number of fragments but
327 * cap it at 4000 (XXX).
329 mpipe_init(&ipq_mpipe, M_IPQ, sizeof(struct ipq),
330 IFQ_MAXLEN, 4000, 0, NULL, NULL, NULL);
331 for (i = 0; i < ncpus; ++i) {
332 TAILQ_INIT(&in_ifaddrheads[i]);
333 in_ifaddrhashtbls[i] =
334 hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask);
336 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
339 for (i = 0; i < IPPROTO_MAX; i++)
340 ip_protox[i] = pr - inetsw;
341 for (pr = inetdomain.dom_protosw;
342 pr < inetdomain.dom_protoswNPROTOSW; pr++) {
343 if (pr->pr_domain->dom_family == PF_INET && pr->pr_protocol) {
344 if (pr->pr_protocol != IPPROTO_RAW)
345 ip_protox[pr->pr_protocol] = pr - inetsw;
349 inet_pfil_hook.ph_type = PFIL_TYPE_AF;
350 inet_pfil_hook.ph_af = AF_INET;
351 if ((i = pfil_head_register(&inet_pfil_hook)) != 0) {
352 kprintf("%s: WARNING: unable to register pfil hook, "
353 "error %d\n", __func__, i);
356 for (i = 0; i < IPREASS_NHASH; i++)
359 maxnipq = nmbclusters / 32;
360 maxfragsperpacket = 16;
362 ip_id = time_second & 0xffff; /* time_second survives reboots */
364 for (cpu = 0; cpu < ncpus; ++cpu) {
366 * Initialize IP statistics counters for each CPU.
368 bzero(&ipstats_percpu[cpu], sizeof(struct ip_stats));
371 * Preallocate mbuf template for forwarding
373 MGETHDR(ipforward_mtemp[cpu], MB_WAIT, MT_DATA);
376 netisr_register(NETISR_IP, ip_input_handler, ip_hashfn_in);
377 netisr_register_hashcheck(NETISR_IP, ip_hashcheck);
380 /* Do transport protocol processing. */
382 transport_processing_oncpu(struct mbuf *m, int hlen, struct ip *ip)
384 const struct protosw *pr = &inetsw[ip_protox[ip->ip_p]];
387 * Switch out to protocol's input routine.
390 pr->pr_input(&m, &hlen, ip->ip_p);
395 transport_processing_handler(netmsg_t msg)
397 struct netmsg_packet *pmsg = &msg->packet;
401 ip = mtod(pmsg->nm_packet, struct ip *);
402 hlen = pmsg->base.lmsg.u.ms_result;
404 transport_processing_oncpu(pmsg->nm_packet, hlen, ip);
405 /* msg was embedded in the mbuf, do not reply! */
409 ip_input_handler(netmsg_t msg)
411 ip_input(msg->packet.nm_packet);
412 /* msg was embedded in the mbuf, do not reply! */
416 * IP input routine. Checksum and byte swap header. If fragmented
417 * try to reassemble. Process options. Pass to next level.
420 ip_input(struct mbuf *m)
423 struct in_ifaddr *ia = NULL;
424 struct in_ifaddr_container *iac;
427 struct in_addr pkt_dst;
428 boolean_t check_msgport = FALSE;
429 boolean_t using_srcrt = FALSE; /* forward (by PFIL_HOOKS) */
430 struct in_addr odst; /* original dst address(NAT) */
432 struct sockaddr_in *next_hop = NULL;
435 struct tdb_ident *tdbi;
436 struct secpolicy *sp;
443 * This routine is called from numerous places which may not have
444 * characterized the packet.
446 if ((m->m_flags & M_HASH) == 0) {
447 atomic_add_long(&ip_hash_count, 1);
448 ip_hashfn(&m, 0, IP_MPORT_IN);
451 KKASSERT(m->m_flags & M_HASH);
452 check_msgport = TRUE;
454 ip = mtod(m, struct ip *);
456 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
458 * XXX handle multicast on CPU0 for now.
460 * This could happen for packets hashed by hardware
461 * using RSS, which does not differentiate multicast
462 * packets from unicast packets.
464 m->m_pkthdr.hash = 0;
465 check_msgport = TRUE;
469 &curthread->td_msgport != netisr_hashport(m->m_pkthdr.hash)) {
470 netisr_queue(NETISR_IP, m);
471 /* Requeued to other netisr msgport; done */
476 * Pull out certain tags
478 if (m->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) {
480 mtag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL);
481 KKASSERT(mtag != NULL);
482 next_hop = m_tag_data(mtag);
485 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
486 /* dummynet already filtered us */
487 ip = mtod(m, struct ip *);
488 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
494 /* length checks already done in ip_hashfn() */
495 KASSERT(m->m_len >= sizeof(struct ip), ("IP header not in one mbuf"));
497 if (IP_VHL_V(ip->ip_vhl) != IPVERSION) {
498 ipstat.ips_badvers++;
502 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
503 /* length checks already done in ip_hashfn() */
504 KASSERT(hlen >= sizeof(struct ip), ("IP header len too small"));
505 KASSERT(m->m_len >= hlen, ("complete IP header not in one mbuf"));
507 /* 127/8 must not appear on wire - RFC1122 */
508 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
509 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
510 if (!(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK)) {
511 ipstat.ips_badaddr++;
516 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
517 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
519 if (hlen == sizeof(struct ip))
520 sum = in_cksum_hdr(ip);
522 sum = in_cksum(m, hlen);
530 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) {
531 /* packet is dropped by traffic conditioner */
536 * Convert fields to host representation.
538 ip->ip_len = ntohs(ip->ip_len);
539 ip->ip_off = ntohs(ip->ip_off);
541 /* length checks already done in ip_hashfn() */
542 KASSERT(ip->ip_len >= hlen, ("total length less then header length"));
543 KASSERT(m->m_pkthdr.len >= ip->ip_len, ("mbuf too short"));
546 * Trim mbufs if longer than the IP header would have us expect.
548 if (m->m_pkthdr.len > ip->ip_len) {
549 if (m->m_len == m->m_pkthdr.len) {
550 m->m_len = ip->ip_len;
551 m->m_pkthdr.len = ip->ip_len;
553 m_adj(m, ip->ip_len - m->m_pkthdr.len);
556 #if defined(IPSEC) && !defined(IPSEC_FILTERGIF)
558 * Bypass packet filtering for packets from a tunnel (gif).
560 if (ipsec_gethist(m, NULL))
566 * Right now when no processing on packet has done
567 * and it is still fresh out of network we do our black
569 * - Firewall: deny/allow/divert
570 * - Xlate: translate packet's addr/port (NAT).
571 * - Pipe: pass pkt through dummynet.
572 * - Wrap: fake packet's addr/port <unimpl.>
573 * - Encapsulate: put it in another IP and send out. <unimp.>
578 * If we've been forwarded from the output side, then
579 * skip the firewall a second time
581 if (next_hop != NULL)
585 if (!pfil_has_hooks(&inet_pfil_hook)) {
586 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
588 * Strip dummynet tags from stranded packets
590 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
591 KKASSERT(mtag != NULL);
592 m_tag_delete(m, mtag);
593 m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED;
599 * Run through list of hooks for input packets.
601 * NOTE! If the packet is rewritten pf/ipfw/whoever must
605 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif, PFIL_IN))
607 if (m == NULL) /* consumed by filter */
609 ip = mtod(m, struct ip *);
610 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
611 using_srcrt = (odst.s_addr != ip->ip_dst.s_addr);
613 if (m->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) {
614 mtag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL);
615 KKASSERT(mtag != NULL);
616 next_hop = m_tag_data(mtag);
618 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
622 if (m->m_pkthdr.fw_flags & FW_MBUF_REDISPATCH) {
623 m->m_pkthdr.fw_flags &= ~FW_MBUF_REDISPATCH;
627 * Process options and, if not destined for us,
628 * ship it on. ip_dooptions returns 1 when an
629 * error was detected (causing an icmp message
630 * to be sent and the original packet to be freed).
632 if (hlen > sizeof(struct ip) && ip_dooptions(m, 0, next_hop))
635 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
636 * matter if it is destined to another node, or whether it is
637 * a multicast one, RSVP wants it! and prevents it from being forwarded
638 * anywhere else. Also checks if the rsvp daemon is running before
639 * grabbing the packet.
641 if (rsvp_on && ip->ip_p == IPPROTO_RSVP)
645 * Check our list of addresses, to see if the packet is for us.
646 * If we don't have any addresses, assume any unicast packet
647 * we receive might be for us (and let the upper layers deal
650 if (TAILQ_EMPTY(&in_ifaddrheads[mycpuid]) &&
651 !(m->m_flags & (M_MCAST | M_BCAST)))
655 * Cache the destination address of the packet; this may be
656 * changed by use of 'ipfw fwd'.
658 pkt_dst = next_hop ? next_hop->sin_addr : ip->ip_dst;
661 * Enable a consistency check between the destination address
662 * and the arrival interface for a unicast packet (the RFC 1122
663 * strong ES model) if IP forwarding is disabled and the packet
664 * is not locally generated and the packet is not subject to
667 * XXX - Checking also should be disabled if the destination
668 * address is ipnat'ed to a different interface.
670 * XXX - Checking is incompatible with IP aliases added
671 * to the loopback interface instead of the interface where
672 * the packets are received.
674 checkif = ip_checkinterface &&
676 m->m_pkthdr.rcvif != NULL &&
677 !(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) &&
681 * Check for exact addresses in the hash bucket.
683 LIST_FOREACH(iac, INADDR_HASH(pkt_dst.s_addr), ia_hash) {
687 * If the address matches, verify that the packet
688 * arrived via the correct interface if checking is
691 if (IA_SIN(ia)->sin_addr.s_addr == pkt_dst.s_addr &&
692 (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif))
698 * Check for broadcast addresses.
700 * Only accept broadcast packets that arrive via the matching
701 * interface. Reception of forwarded directed broadcasts would
702 * be handled via ip_forward() and ether_output() with the loopback
703 * into the stack for SIMPLEX interfaces handled by ether_output().
705 if (m->m_pkthdr.rcvif != NULL &&
706 m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
707 struct ifaddr_container *ifac;
709 TAILQ_FOREACH(ifac, &m->m_pkthdr.rcvif->if_addrheads[mycpuid],
711 struct ifaddr *ifa = ifac->ifa;
713 if (ifa->ifa_addr == NULL) /* shutdown/startup race */
715 if (ifa->ifa_addr->sa_family != AF_INET)
718 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
721 if (ia->ia_netbroadcast.s_addr == pkt_dst.s_addr)
724 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY)
729 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
730 struct in_multi *inm;
732 /* XXX Multicast is not MPSAFE yet */
735 if (ip_mrouter != NULL) {
737 * If we are acting as a multicast router, all
738 * incoming multicast packets are passed to the
739 * kernel-level multicast forwarding function.
740 * The packet is returned (relatively) intact; if
741 * ip_mforward() returns a non-zero value, the packet
742 * must be discarded, else it may be accepted below.
744 if (ip_mforward != NULL &&
745 ip_mforward(ip, m->m_pkthdr.rcvif, m, NULL) != 0) {
747 ipstat.ips_cantforward++;
753 * The process-level routing daemon needs to receive
754 * all multicast IGMP packets, whether or not this
755 * host belongs to their destination groups.
757 if (ip->ip_p == IPPROTO_IGMP) {
761 ipstat.ips_forward++;
764 * See if we belong to the destination multicast group on the
767 inm = IN_LOOKUP_MULTI(&ip->ip_dst, m->m_pkthdr.rcvif);
770 ipstat.ips_notmember++;
778 if (ip->ip_dst.s_addr == INADDR_BROADCAST)
780 if (ip->ip_dst.s_addr == INADDR_ANY)
784 * FAITH(Firewall Aided Internet Translator)
786 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
788 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP)
796 * Not for us; forward if possible and desirable.
799 ipstat.ips_cantforward++;
804 * Enforce inbound IPsec SPD.
806 if (ipsec4_in_reject(m, NULL)) {
807 ipsecstat.in_polvio++;
812 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
815 tdbi = (struct tdb_ident *)m_tag_data(mtag);
816 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
818 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
819 IP_FORWARDING, &error);
821 if (sp == NULL) { /* NB: can happen if error */
823 /*XXX error stat???*/
824 DPRINTF(("ip_input: no SP for forwarding\n")); /*XXX*/
829 * Check security policy against packet attributes.
831 error = ipsec_in_reject(sp, m);
835 ipstat.ips_cantforward++;
839 ip_forward(m, using_srcrt, next_hop);
846 * IPSTEALTH: Process non-routing options only
847 * if the packet is destined for us.
850 hlen > sizeof(struct ip) &&
851 ip_dooptions(m, 1, next_hop))
854 /* Count the packet in the ip address stats */
856 IFA_STAT_INC(&ia->ia_ifa, ipackets, 1);
857 IFA_STAT_INC(&ia->ia_ifa, ibytes, m->m_pkthdr.len);
861 * If offset or IP_MF are set, must reassemble.
862 * Otherwise, nothing need be done.
863 * (We could look in the reassembly queue to see
864 * if the packet was previously fragmented,
865 * but it's not worth the time; just let them time out.)
867 if (ip->ip_off & (IP_MF | IP_OFFMASK)) {
869 * Attempt reassembly; if it succeeds, proceed. ip_reass()
870 * will return a different mbuf.
872 * NOTE: ip_reass() returns m with M_HASH cleared to force
873 * us to recharacterize the packet.
878 ip = mtod(m, struct ip *);
880 /* Get the header length of the reassembled packet */
881 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
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) &&
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) {
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 *)m_tag_data(mtag);
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 * We must forward the packet to the correct protocol thread if
940 * we are not already in it.
942 * NOTE: ip_len is now in host form. ip_len is not adjusted
943 * further for protocol processing, instead we pass hlen
944 * to the protosw and let it deal with it.
946 ipstat.ips_delivered++;
948 if ((m->m_flags & M_HASH) == 0) {
950 atomic_add_long(&ip_rehash_count, 1);
952 ip->ip_len = htons(ip->ip_len + hlen);
953 ip->ip_off = htons(ip->ip_off);
955 ip_hashfn(&m, 0, IP_MPORT_IN);
959 ip = mtod(m, struct ip *);
960 ip->ip_len = ntohs(ip->ip_len) - hlen;
961 ip->ip_off = ntohs(ip->ip_off);
962 KKASSERT(m->m_flags & M_HASH);
964 port = netisr_hashport(m->m_pkthdr.hash);
966 if (port != &curthread->td_msgport) {
967 struct netmsg_packet *pmsg;
970 atomic_add_long(&ip_dispatch_slow, 1);
973 pmsg = &m->m_hdr.mh_netmsg;
974 netmsg_init(&pmsg->base, NULL, &netisr_apanic_rport,
975 0, transport_processing_handler);
977 pmsg->base.lmsg.u.ms_result = hlen;
978 lwkt_sendmsg(port, &pmsg->base.lmsg);
981 atomic_add_long(&ip_dispatch_fast, 1);
983 transport_processing_oncpu(m, hlen, ip);
992 * Take incoming datagram fragment and try to reassemble it into
993 * whole datagram. If a chain for reassembly of this datagram already
994 * exists, then it is given as fp; otherwise have to make a chain.
997 ip_reass(struct mbuf *m)
999 struct ip *ip = mtod(m, struct ip *);
1000 struct mbuf *p = NULL, *q, *nq;
1002 struct ipq *fp = NULL;
1003 struct ipqhead *head;
1004 int hlen = IP_VHL_HL(ip->ip_vhl) << 2;
1008 /* If maxnipq is 0, never accept fragments. */
1010 ipstat.ips_fragments++;
1011 ipstat.ips_fragdropped++;
1016 sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
1018 * Look for queue of fragments of this datagram.
1020 lwkt_gettoken(&ipq_token);
1022 TAILQ_FOREACH(fp, head, ipq_list) {
1023 if (ip->ip_id == fp->ipq_id &&
1024 ip->ip_src.s_addr == fp->ipq_src.s_addr &&
1025 ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
1026 ip->ip_p == fp->ipq_p)
1033 * Enforce upper bound on number of fragmented packets
1034 * for which we attempt reassembly;
1035 * If maxnipq is -1, accept all fragments without limitation.
1037 if (nipq > maxnipq && maxnipq > 0) {
1039 * drop something from the tail of the current queue
1040 * before proceeding further
1042 struct ipq *q = TAILQ_LAST(head, ipqhead);
1045 * The current queue is empty,
1046 * so drop from one of the others.
1048 for (i = 0; i < IPREASS_NHASH; i++) {
1049 struct ipq *r = TAILQ_LAST(&ipq[i], ipqhead);
1051 ipstat.ips_fragtimeout += r->ipq_nfrags;
1052 ip_freef(&ipq[i], r);
1057 ipstat.ips_fragtimeout += q->ipq_nfrags;
1063 * Adjust ip_len to not reflect header,
1064 * convert offset of this to bytes.
1067 if (ip->ip_off & IP_MF) {
1069 * Make sure that fragments have a data length
1070 * that's a non-zero multiple of 8 bytes.
1072 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
1073 ipstat.ips_toosmall++; /* XXX */
1077 m->m_flags |= M_FRAG;
1079 m->m_flags &= ~M_FRAG;
1083 ipstat.ips_fragments++;
1084 m->m_pkthdr.header = ip;
1087 * If the hardware has not done csum over this fragment
1088 * then csum_data is not valid at all.
1090 if ((m->m_pkthdr.csum_flags & (CSUM_FRAG_NOT_CHECKED | CSUM_DATA_VALID))
1091 == (CSUM_FRAG_NOT_CHECKED | CSUM_DATA_VALID)) {
1092 m->m_pkthdr.csum_data = 0;
1093 m->m_pkthdr.csum_flags &= ~(CSUM_DATA_VALID | CSUM_PSEUDO_HDR);
1097 * Presence of header sizes in mbufs
1098 * would confuse code below.
1104 * If first fragment to arrive, create a reassembly queue.
1107 if ((fp = mpipe_alloc_nowait(&ipq_mpipe)) == NULL)
1109 TAILQ_INSERT_HEAD(head, fp, ipq_list);
1112 fp->ipq_ttl = IPFRAGTTL;
1113 fp->ipq_p = ip->ip_p;
1114 fp->ipq_id = ip->ip_id;
1115 fp->ipq_src = ip->ip_src;
1116 fp->ipq_dst = ip->ip_dst;
1118 m->m_nextpkt = NULL;
1124 #define GETIP(m) ((struct ip*)((m)->m_pkthdr.header))
1127 * Find a segment which begins after this one does.
1129 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
1130 if (GETIP(q)->ip_off > ip->ip_off)
1135 * If there is a preceding segment, it may provide some of
1136 * our data already. If so, drop the data from the incoming
1137 * segment. If it provides all of our data, drop us, otherwise
1138 * stick new segment in the proper place.
1140 * If some of the data is dropped from the the preceding
1141 * segment, then it's checksum is invalidated.
1144 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off;
1146 if (i >= ip->ip_len)
1149 m->m_pkthdr.csum_flags = 0;
1153 m->m_nextpkt = p->m_nextpkt;
1156 m->m_nextpkt = fp->ipq_frags;
1161 * While we overlap succeeding segments trim them or,
1162 * if they are completely covered, dequeue them.
1164 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off;
1166 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off;
1167 if (i < GETIP(q)->ip_len) {
1168 GETIP(q)->ip_len -= i;
1169 GETIP(q)->ip_off += i;
1171 q->m_pkthdr.csum_flags = 0;
1176 ipstat.ips_fragdropped++;
1178 q->m_nextpkt = NULL;
1184 * Check for complete reassembly and perform frag per packet
1187 * Frag limiting is performed here so that the nth frag has
1188 * a chance to complete the packet before we drop the packet.
1189 * As a result, n+1 frags are actually allowed per packet, but
1190 * only n will ever be stored. (n = maxfragsperpacket.)
1194 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
1195 if (GETIP(q)->ip_off != next) {
1196 if (fp->ipq_nfrags > maxfragsperpacket) {
1197 ipstat.ips_fragdropped += fp->ipq_nfrags;
1202 next += GETIP(q)->ip_len;
1204 /* Make sure the last packet didn't have the IP_MF flag */
1205 if (p->m_flags & M_FRAG) {
1206 if (fp->ipq_nfrags > maxfragsperpacket) {
1207 ipstat.ips_fragdropped += fp->ipq_nfrags;
1214 * Reassembly is complete. Make sure the packet is a sane size.
1218 if (next + (IP_VHL_HL(ip->ip_vhl) << 2) > IP_MAXPACKET) {
1219 ipstat.ips_toolong++;
1220 ipstat.ips_fragdropped += fp->ipq_nfrags;
1226 * Concatenate fragments.
1233 q->m_nextpkt = NULL;
1234 for (q = nq; q != NULL; q = nq) {
1236 q->m_nextpkt = NULL;
1237 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
1238 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
1243 * Clean up the 1's complement checksum. Carry over 16 bits must
1244 * be added back. This assumes no more then 65535 packet fragments
1245 * were reassembled. A second carry can also occur (but not a third).
1247 m->m_pkthdr.csum_data = (m->m_pkthdr.csum_data & 0xffff) +
1248 (m->m_pkthdr.csum_data >> 16);
1249 if (m->m_pkthdr.csum_data > 0xFFFF)
1250 m->m_pkthdr.csum_data -= 0xFFFF;
1253 * Create header for new ip packet by
1254 * modifying header of first packet;
1255 * dequeue and discard fragment reassembly header.
1256 * Make header visible.
1259 ip->ip_src = fp->ipq_src;
1260 ip->ip_dst = fp->ipq_dst;
1261 TAILQ_REMOVE(head, fp, ipq_list);
1263 mpipe_free(&ipq_mpipe, fp);
1264 m->m_len += (IP_VHL_HL(ip->ip_vhl) << 2);
1265 m->m_data -= (IP_VHL_HL(ip->ip_vhl) << 2);
1266 /* some debugging cruft by sklower, below, will go away soon */
1267 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */
1270 for (n = m; n; n = n->m_next)
1272 m->m_pkthdr.len = plen;
1276 * Reassembly complete, return the next protocol.
1278 * Be sure to clear M_HASH to force the packet
1279 * to be re-characterized.
1281 * Clear M_FRAG, we are no longer a fragment.
1283 m->m_flags &= ~(M_HASH | M_FRAG);
1285 ipstat.ips_reassembled++;
1286 lwkt_reltoken(&ipq_token);
1290 ipstat.ips_fragdropped++;
1295 lwkt_reltoken(&ipq_token);
1302 * Free a fragment reassembly header and all
1303 * associated datagrams.
1305 * Called with ipq_token held.
1308 ip_freef(struct ipqhead *fhp, struct ipq *fp)
1313 * Remove first to protect against blocking
1315 TAILQ_REMOVE(fhp, fp, ipq_list);
1318 * Clean out at our leisure
1320 while (fp->ipq_frags) {
1322 fp->ipq_frags = q->m_nextpkt;
1323 q->m_nextpkt = NULL;
1326 mpipe_free(&ipq_mpipe, fp);
1331 * IP timer processing;
1332 * if a timer expires on a reassembly
1333 * queue, discard it.
1338 struct ipq *fp, *fp_temp;
1339 struct ipqhead *head;
1342 lwkt_gettoken(&ipq_token);
1343 for (i = 0; i < IPREASS_NHASH; i++) {
1345 TAILQ_FOREACH_MUTABLE(fp, head, ipq_list, fp_temp) {
1346 if (--fp->ipq_ttl == 0) {
1347 ipstat.ips_fragtimeout += fp->ipq_nfrags;
1353 * If we are over the maximum number of fragments
1354 * (due to the limit being lowered), drain off
1355 * enough to get down to the new limit.
1357 if (maxnipq >= 0 && nipq > maxnipq) {
1358 for (i = 0; i < IPREASS_NHASH; i++) {
1360 while (nipq > maxnipq && !TAILQ_EMPTY(head)) {
1361 ipstat.ips_fragdropped +=
1362 TAILQ_FIRST(head)->ipq_nfrags;
1363 ip_freef(head, TAILQ_FIRST(head));
1367 lwkt_reltoken(&ipq_token);
1372 * Drain off all datagram fragments.
1377 struct ipqhead *head;
1380 lwkt_gettoken(&ipq_token);
1381 for (i = 0; i < IPREASS_NHASH; i++) {
1383 while (!TAILQ_EMPTY(head)) {
1384 ipstat.ips_fragdropped += TAILQ_FIRST(head)->ipq_nfrags;
1385 ip_freef(head, TAILQ_FIRST(head));
1388 lwkt_reltoken(&ipq_token);
1393 * Do option processing on a datagram,
1394 * possibly discarding it if bad options are encountered,
1395 * or forwarding it if source-routed.
1396 * The pass argument is used when operating in the IPSTEALTH
1397 * mode to tell what options to process:
1398 * [LS]SRR (pass 0) or the others (pass 1).
1399 * The reason for as many as two passes is that when doing IPSTEALTH,
1400 * non-routing options should be processed only if the packet is for us.
1401 * Returns 1 if packet has been forwarded/freed,
1402 * 0 if the packet should be processed further.
1405 ip_dooptions(struct mbuf *m, int pass, struct sockaddr_in *next_hop)
1407 struct sockaddr_in ipaddr = { sizeof ipaddr, AF_INET };
1408 struct ip *ip = mtod(m, struct ip *);
1410 struct in_ifaddr *ia;
1411 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB;
1412 boolean_t forward = FALSE;
1413 struct in_addr *sin, dst;
1417 cp = (u_char *)(ip + 1);
1418 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
1419 for (; cnt > 0; cnt -= optlen, cp += optlen) {
1420 opt = cp[IPOPT_OPTVAL];
1421 if (opt == IPOPT_EOL)
1423 if (opt == IPOPT_NOP)
1426 if (cnt < IPOPT_OLEN + sizeof(*cp)) {
1427 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1430 optlen = cp[IPOPT_OLEN];
1431 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
1432 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1442 * Source routing with record.
1443 * Find interface with current destination address.
1444 * If none on this machine then drop if strictly routed,
1445 * or do nothing if loosely routed.
1446 * Record interface address and bring up next address
1447 * component. If strictly routed make sure next
1448 * address is on directly accessible net.
1452 if (ipstealth && pass > 0)
1454 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1455 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1458 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1459 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1462 ipaddr.sin_addr = ip->ip_dst;
1463 ia = (struct in_ifaddr *)
1464 ifa_ifwithaddr((struct sockaddr *)&ipaddr);
1466 if (opt == IPOPT_SSRR) {
1467 type = ICMP_UNREACH;
1468 code = ICMP_UNREACH_SRCFAIL;
1471 if (!ip_dosourceroute)
1472 goto nosourcerouting;
1474 * Loose routing, and not at next destination
1475 * yet; nothing to do except forward.
1479 off--; /* 0 origin */
1480 if (off > optlen - (int)sizeof(struct in_addr)) {
1482 * End of source route. Should be for us.
1484 if (!ip_acceptsourceroute)
1485 goto nosourcerouting;
1486 save_rte(m, cp, ip->ip_src);
1491 if (!ip_dosourceroute) {
1493 char buf[sizeof "aaa.bbb.ccc.ddd"];
1496 * Acting as a router, so generate ICMP
1499 strcpy(buf, inet_ntoa(ip->ip_dst));
1501 "attempted source route from %s to %s\n",
1502 inet_ntoa(ip->ip_src), buf);
1503 type = ICMP_UNREACH;
1504 code = ICMP_UNREACH_SRCFAIL;
1508 * Not acting as a router,
1512 ipstat.ips_cantforward++;
1519 * locate outgoing interface
1521 memcpy(&ipaddr.sin_addr, cp + off,
1522 sizeof ipaddr.sin_addr);
1524 if (opt == IPOPT_SSRR) {
1525 #define INA struct in_ifaddr *
1526 #define SA struct sockaddr *
1527 if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr))
1529 ia = (INA)ifa_ifwithnet((SA)&ipaddr);
1531 ia = ip_rtaddr(ipaddr.sin_addr, NULL);
1534 type = ICMP_UNREACH;
1535 code = ICMP_UNREACH_SRCFAIL;
1538 ip->ip_dst = ipaddr.sin_addr;
1539 memcpy(cp + off, &IA_SIN(ia)->sin_addr,
1540 sizeof(struct in_addr));
1541 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1543 * Let ip_intr's mcast routing check handle mcast pkts
1545 forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr));
1549 if (ipstealth && pass == 0)
1551 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1552 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1555 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1556 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1560 * If no space remains, ignore.
1562 off--; /* 0 origin */
1563 if (off > optlen - (int)sizeof(struct in_addr))
1565 memcpy(&ipaddr.sin_addr, &ip->ip_dst,
1566 sizeof ipaddr.sin_addr);
1568 * locate outgoing interface; if we're the destination,
1569 * use the incoming interface (should be same).
1571 if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == NULL &&
1572 (ia = ip_rtaddr(ipaddr.sin_addr, NULL)) == NULL) {
1573 type = ICMP_UNREACH;
1574 code = ICMP_UNREACH_HOST;
1577 memcpy(cp + off, &IA_SIN(ia)->sin_addr,
1578 sizeof(struct in_addr));
1579 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1583 if (ipstealth && pass == 0)
1585 code = cp - (u_char *)ip;
1586 if (optlen < 4 || optlen > 40) {
1587 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1590 if ((off = cp[IPOPT_OFFSET]) < 5) {
1591 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1594 if (off > optlen - (int)sizeof(int32_t)) {
1595 cp[IPOPT_OFFSET + 1] += (1 << 4);
1596 if ((cp[IPOPT_OFFSET + 1] & 0xf0) == 0) {
1597 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1602 off--; /* 0 origin */
1603 sin = (struct in_addr *)(cp + off);
1604 switch (cp[IPOPT_OFFSET + 1] & 0x0f) {
1606 case IPOPT_TS_TSONLY:
1609 case IPOPT_TS_TSANDADDR:
1610 if (off + sizeof(n_time) +
1611 sizeof(struct in_addr) > optlen) {
1612 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1615 ipaddr.sin_addr = dst;
1616 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr,
1620 memcpy(sin, &IA_SIN(ia)->sin_addr,
1621 sizeof(struct in_addr));
1622 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1623 off += sizeof(struct in_addr);
1626 case IPOPT_TS_PRESPEC:
1627 if (off + sizeof(n_time) +
1628 sizeof(struct in_addr) > optlen) {
1629 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1632 memcpy(&ipaddr.sin_addr, sin,
1633 sizeof(struct in_addr));
1634 if (ifa_ifwithaddr((SA)&ipaddr) == NULL)
1636 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1637 off += sizeof(struct in_addr);
1641 code = &cp[IPOPT_OFFSET + 1] - (u_char *)ip;
1645 memcpy(cp + off, &ntime, sizeof(n_time));
1646 cp[IPOPT_OFFSET] += sizeof(n_time);
1649 if (forward && ipforwarding) {
1650 ip_forward(m, TRUE, next_hop);
1655 icmp_error(m, type, code, 0, 0);
1656 ipstat.ips_badoptions++;
1661 * Given address of next destination (final or next hop),
1662 * return internet address info of interface to be used to get there.
1665 ip_rtaddr(struct in_addr dst, struct route *ro0)
1667 struct route sro, *ro;
1668 struct sockaddr_in *sin;
1669 struct in_ifaddr *ia;
1674 bzero(&sro, sizeof(sro));
1678 sin = (struct sockaddr_in *)&ro->ro_dst;
1680 if (ro->ro_rt == NULL || dst.s_addr != sin->sin_addr.s_addr) {
1681 if (ro->ro_rt != NULL) {
1685 sin->sin_family = AF_INET;
1686 sin->sin_len = sizeof *sin;
1687 sin->sin_addr = dst;
1688 rtalloc_ign(ro, RTF_PRCLONING);
1691 if (ro->ro_rt == NULL)
1694 ia = ifatoia(ro->ro_rt->rt_ifa);
1702 * Save incoming source route for use in replies,
1703 * to be picked up later by ip_srcroute if the receiver is interested.
1706 save_rte(struct mbuf *m, u_char *option, struct in_addr dst)
1709 struct ip_srcrt_opt *opt;
1712 mtag = m_tag_get(PACKET_TAG_IPSRCRT, sizeof(*opt), MB_DONTWAIT);
1715 opt = m_tag_data(mtag);
1717 olen = option[IPOPT_OLEN];
1720 kprintf("save_rte: olen %d\n", olen);
1722 if (olen > sizeof(opt->ip_srcrt) - (1 + sizeof(dst))) {
1726 bcopy(option, opt->ip_srcrt.srcopt, olen);
1727 opt->ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1728 opt->ip_srcrt.dst = dst;
1729 m_tag_prepend(m, mtag);
1733 * Retrieve incoming source route for use in replies,
1734 * in the same form used by setsockopt.
1735 * The first hop is placed before the options, will be removed later.
1738 ip_srcroute(struct mbuf *m0)
1740 struct in_addr *p, *q;
1743 struct ip_srcrt_opt *opt;
1748 mtag = m_tag_find(m0, PACKET_TAG_IPSRCRT, NULL);
1751 opt = m_tag_data(mtag);
1753 if (opt->ip_nhops == 0)
1755 m = m_get(MB_DONTWAIT, MT_HEADER);
1759 #define OPTSIZ (sizeof(opt->ip_srcrt.nop) + sizeof(opt->ip_srcrt.srcopt))
1761 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
1762 m->m_len = opt->ip_nhops * sizeof(struct in_addr) +
1763 sizeof(struct in_addr) + OPTSIZ;
1766 kprintf("ip_srcroute: nhops %d mlen %d",
1767 opt->ip_nhops, m->m_len);
1772 * First save first hop for return route
1774 p = &opt->ip_srcrt.route[opt->ip_nhops - 1];
1775 *(mtod(m, struct in_addr *)) = *p--;
1778 kprintf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr));
1782 * Copy option fields and padding (nop) to mbuf.
1784 opt->ip_srcrt.nop = IPOPT_NOP;
1785 opt->ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
1786 memcpy(mtod(m, caddr_t) + sizeof(struct in_addr), &opt->ip_srcrt.nop,
1788 q = (struct in_addr *)(mtod(m, caddr_t) +
1789 sizeof(struct in_addr) + OPTSIZ);
1792 * Record return path as an IP source route,
1793 * reversing the path (pointers are now aligned).
1795 while (p >= opt->ip_srcrt.route) {
1798 kprintf(" %x", ntohl(q->s_addr));
1803 * Last hop goes to final destination.
1805 *q = opt->ip_srcrt.dst;
1806 m_tag_delete(m0, mtag);
1809 kprintf(" %x\n", ntohl(q->s_addr));
1815 * Strip out IP options.
1818 ip_stripoptions(struct mbuf *m)
1821 struct ip *ip = mtod(m, struct ip *);
1825 optlen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
1826 opts = (caddr_t)(ip + 1);
1827 datalen = m->m_len - (sizeof(struct ip) + optlen);
1828 bcopy(opts + optlen, opts, datalen);
1830 if (m->m_flags & M_PKTHDR)
1831 m->m_pkthdr.len -= optlen;
1832 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, sizeof(struct ip) >> 2);
1835 u_char inetctlerrmap[PRC_NCMDS] = {
1837 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
1838 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
1839 EMSGSIZE, EHOSTUNREACH, 0, 0,
1841 ENOPROTOOPT, ECONNREFUSED
1845 * Forward a packet. If some error occurs return the sender
1846 * an icmp packet. Note we can't always generate a meaningful
1847 * icmp message because icmp doesn't have a large enough repertoire
1848 * of codes and types.
1850 * If not forwarding, just drop the packet. This could be confusing
1851 * if ipforwarding was zero but some routing protocol was advancing
1852 * us as a gateway to somewhere. However, we must let the routing
1853 * protocol deal with that.
1855 * The using_srcrt parameter indicates whether the packet is being forwarded
1856 * via a source route.
1859 ip_forward(struct mbuf *m, boolean_t using_srcrt, struct sockaddr_in *next_hop)
1861 struct ip *ip = mtod(m, struct ip *);
1863 struct route fwd_ro;
1864 int error, type = 0, code = 0, destmtu = 0;
1865 struct mbuf *mcopy, *mtemp = NULL;
1867 struct in_addr pkt_dst;
1871 * Cache the destination address of the packet; this may be
1872 * changed by use of 'ipfw fwd'.
1874 pkt_dst = (next_hop != NULL) ? next_hop->sin_addr : ip->ip_dst;
1878 kprintf("forward: src %x dst %x ttl %x\n",
1879 ip->ip_src.s_addr, pkt_dst.s_addr, ip->ip_ttl);
1882 if (m->m_flags & (M_BCAST | M_MCAST) || !in_canforward(pkt_dst)) {
1883 ipstat.ips_cantforward++;
1887 if (!ipstealth && ip->ip_ttl <= IPTTLDEC) {
1888 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0);
1892 bzero(&fwd_ro, sizeof(fwd_ro));
1893 ip_rtaddr(pkt_dst, &fwd_ro);
1894 if (fwd_ro.ro_rt == NULL) {
1895 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0);
1900 if (curthread->td_type == TD_TYPE_NETISR) {
1902 * Save the IP header and at most 8 bytes of the payload,
1903 * in case we need to generate an ICMP message to the src.
1905 mtemp = ipforward_mtemp[mycpuid];
1906 KASSERT((mtemp->m_flags & M_EXT) == 0 &&
1907 mtemp->m_data == mtemp->m_pktdat &&
1908 m_tag_first(mtemp) == NULL,
1909 ("ip_forward invalid mtemp1"));
1911 if (!m_dup_pkthdr(mtemp, m, MB_DONTWAIT)) {
1913 * It's probably ok if the pkthdr dup fails (because
1914 * the deep copy of the tag chain failed), but for now
1915 * be conservative and just discard the copy since
1916 * code below may some day want the tags.
1920 mtemp->m_type = m->m_type;
1921 mtemp->m_len = imin((IP_VHL_HL(ip->ip_vhl) << 2) + 8,
1923 mtemp->m_pkthdr.len = mtemp->m_len;
1924 m_copydata(m, 0, mtemp->m_len, mtod(mtemp, caddr_t));
1929 ip->ip_ttl -= IPTTLDEC;
1932 * If forwarding packet using same interface that it came in on,
1933 * perhaps should send a redirect to sender to shortcut a hop.
1934 * Only send redirect if source is sending directly to us,
1935 * and if packet was not source routed (or has any options).
1936 * Also, don't send redirect if forwarding using a default route
1937 * or a route modified by a redirect.
1939 if (rt->rt_ifp == m->m_pkthdr.rcvif &&
1940 !(rt->rt_flags & (RTF_DYNAMIC | RTF_MODIFIED)) &&
1941 satosin(rt_key(rt))->sin_addr.s_addr != INADDR_ANY &&
1942 ipsendredirects && !using_srcrt && next_hop == NULL) {
1943 u_long src = ntohl(ip->ip_src.s_addr);
1944 struct in_ifaddr *rt_ifa = (struct in_ifaddr *)rt->rt_ifa;
1946 if (rt_ifa != NULL &&
1947 (src & rt_ifa->ia_subnetmask) == rt_ifa->ia_subnet) {
1948 if (rt->rt_flags & RTF_GATEWAY)
1949 dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
1951 dest = pkt_dst.s_addr;
1953 * Router requirements says to only send
1956 type = ICMP_REDIRECT;
1957 code = ICMP_REDIRECT_HOST;
1960 kprintf("redirect (%d) to %x\n", code, dest);
1965 error = ip_output(m, NULL, &fwd_ro, IP_FORWARDING, NULL, NULL);
1967 ipstat.ips_forward++;
1970 ipflow_create(&fwd_ro, mtemp);
1973 ipstat.ips_redirectsent++;
1976 ipstat.ips_cantforward++;
1983 * Errors that do not require generating ICMP message
1988 * A router should not generate ICMP_SOURCEQUENCH as
1989 * required in RFC1812 Requirements for IP Version 4 Routers.
1990 * Source quench could be a big problem under DoS attacks,
1991 * or if the underlying interface is rate-limited.
1992 * Those who need source quench packets may re-enable them
1993 * via the net.inet.ip.sendsourcequench sysctl.
1995 if (!ip_sendsourcequench)
1999 case EACCES: /* ipfw denied packet */
2003 KASSERT((mtemp->m_flags & M_EXT) == 0 &&
2004 mtemp->m_data == mtemp->m_pktdat,
2005 ("ip_forward invalid mtemp2"));
2006 mcopy = m_copym(mtemp, 0, mtemp->m_len, MB_DONTWAIT);
2011 * Send ICMP message.
2014 case 0: /* forwarded, but need redirect */
2015 /* type, code set above */
2018 case ENETUNREACH: /* shouldn't happen, checked above */
2023 type = ICMP_UNREACH;
2024 code = ICMP_UNREACH_HOST;
2028 type = ICMP_UNREACH;
2029 code = ICMP_UNREACH_NEEDFRAG;
2032 * If the packet is routed over IPsec tunnel, tell the
2033 * originator the tunnel MTU.
2034 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
2037 if (fwd_ro.ro_rt != NULL) {
2038 struct secpolicy *sp = NULL;
2043 sp = ipsec4_getpolicybyaddr(mcopy,
2049 destmtu = fwd_ro.ro_rt->rt_ifp->if_mtu;
2051 /* count IPsec header size */
2052 ipsechdr = ipsec4_hdrsiz(mcopy,
2057 * find the correct route for outer IPv4
2058 * header, compute tunnel MTU.
2061 if (sp->req != NULL && sp->req->sav != NULL &&
2062 sp->req->sav->sah != NULL) {
2063 ro = &sp->req->sav->sah->sa_route;
2064 if (ro->ro_rt != NULL &&
2065 ro->ro_rt->rt_ifp != NULL) {
2067 ro->ro_rt->rt_ifp->if_mtu;
2068 destmtu -= ipsechdr;
2077 * If the packet is routed over IPsec tunnel, tell the
2078 * originator the tunnel MTU.
2079 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
2082 if (fwd_ro.ro_rt != NULL) {
2083 struct secpolicy *sp = NULL;
2088 sp = ipsec_getpolicybyaddr(mcopy,
2094 destmtu = fwd_ro.ro_rt->rt_ifp->if_mtu;
2096 /* count IPsec header size */
2097 ipsechdr = ipsec4_hdrsiz(mcopy,
2102 * find the correct route for outer IPv4
2103 * header, compute tunnel MTU.
2106 if (sp->req != NULL &&
2107 sp->req->sav != NULL &&
2108 sp->req->sav->sah != NULL) {
2109 ro = &sp->req->sav->sah->sa_route;
2110 if (ro->ro_rt != NULL &&
2111 ro->ro_rt->rt_ifp != NULL) {
2113 ro->ro_rt->rt_ifp->if_mtu;
2114 destmtu -= ipsechdr;
2121 #else /* !IPSEC && !FAST_IPSEC */
2122 if (fwd_ro.ro_rt != NULL)
2123 destmtu = fwd_ro.ro_rt->rt_ifp->if_mtu;
2125 ipstat.ips_cantfrag++;
2129 type = ICMP_SOURCEQUENCH;
2133 case EACCES: /* ipfw denied packet */
2134 panic("ip_forward EACCES should not reach");
2136 icmp_error(mcopy, type, code, dest, destmtu);
2139 m_tag_delete_chain(mtemp);
2140 if (fwd_ro.ro_rt != NULL)
2141 RTFREE(fwd_ro.ro_rt);
2145 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
2148 if (inp->inp_socket->so_options & SO_TIMESTAMP) {
2152 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
2153 SCM_TIMESTAMP, SOL_SOCKET);
2155 mp = &(*mp)->m_next;
2157 if (inp->inp_flags & INP_RECVDSTADDR) {
2158 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
2159 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
2161 mp = &(*mp)->m_next;
2163 if (inp->inp_flags & INP_RECVTTL) {
2164 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl,
2165 sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
2167 mp = &(*mp)->m_next;
2171 * Moving these out of udp_input() made them even more broken
2172 * than they already were.
2174 /* options were tossed already */
2175 if (inp->inp_flags & INP_RECVOPTS) {
2176 *mp = sbcreatecontrol((caddr_t) opts_deleted_above,
2177 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
2179 mp = &(*mp)->m_next;
2181 /* ip_srcroute doesn't do what we want here, need to fix */
2182 if (inp->inp_flags & INP_RECVRETOPTS) {
2183 *mp = sbcreatecontrol((caddr_t) ip_srcroute(m),
2184 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
2186 mp = &(*mp)->m_next;
2189 if (inp->inp_flags & INP_RECVIF) {
2192 struct sockaddr_dl sdl;
2195 struct sockaddr_dl *sdp;
2196 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
2198 if (((ifp = m->m_pkthdr.rcvif)) &&
2199 ((ifp->if_index != 0) && (ifp->if_index <= if_index))) {
2200 sdp = IF_LLSOCKADDR(ifp);
2202 * Change our mind and don't try copy.
2204 if ((sdp->sdl_family != AF_LINK) ||
2205 (sdp->sdl_len > sizeof(sdlbuf))) {
2208 bcopy(sdp, sdl2, sdp->sdl_len);
2212 offsetof(struct sockaddr_dl, sdl_data[0]);
2213 sdl2->sdl_family = AF_LINK;
2214 sdl2->sdl_index = 0;
2215 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
2217 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len,
2218 IP_RECVIF, IPPROTO_IP);
2220 mp = &(*mp)->m_next;
2225 * XXX these routines are called from the upper part of the kernel.
2227 * They could also be moved to ip_mroute.c, since all the RSVP
2228 * handling is done there already.
2231 ip_rsvp_init(struct socket *so)
2233 if (so->so_type != SOCK_RAW ||
2234 so->so_proto->pr_protocol != IPPROTO_RSVP)
2237 if (ip_rsvpd != NULL)
2242 * This may seem silly, but we need to be sure we don't over-increment
2243 * the RSVP counter, in case something slips up.
2258 * This may seem silly, but we need to be sure we don't over-decrement
2259 * the RSVP counter, in case something slips up.
2269 rsvp_input(struct mbuf **mp, int *offp, int proto)
2271 struct mbuf *m = *mp;
2275 if (rsvp_input_p) { /* call the real one if loaded */
2277 rsvp_input_p(mp, offp, proto);
2278 return(IPPROTO_DONE);
2281 /* Can still get packets with rsvp_on = 0 if there is a local member
2282 * of the group to which the RSVP packet is addressed. But in this
2283 * case we want to throw the packet away.
2288 return(IPPROTO_DONE);
2291 if (ip_rsvpd != NULL) {
2293 rip_input(mp, offp, proto);
2294 return(IPPROTO_DONE);
2296 /* Drop the packet */
2298 return(IPPROTO_DONE);