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. All advertising materials mentioning features or use of this software
47 * must display the following acknowledgement:
48 * This product includes software developed by the University of
49 * California, Berkeley and its contributors.
50 * 4. Neither the name of the University nor the names of its contributors
51 * may be used to endorse or promote products derived from this software
52 * without specific prior written permission.
54 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
55 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
56 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
57 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
58 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
59 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
60 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
61 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
62 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
63 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
66 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94
67 * $FreeBSD: src/sys/netinet/ip_input.c,v 1.130.2.52 2003/03/07 07:01:28 silby Exp $
68 * $DragonFly: src/sys/netinet/ip_input.c,v 1.85 2008/08/22 09:14:16 sephe Exp $
73 #include "opt_bootp.h"
76 #include "opt_ipdivert.h"
77 #include "opt_ipfilter.h"
78 #include "opt_ipstealth.h"
79 #include "opt_ipsec.h"
81 #include <sys/param.h>
82 #include <sys/systm.h>
84 #include <sys/malloc.h>
85 #include <sys/mpipe.h>
86 #include <sys/domain.h>
87 #include <sys/protosw.h>
88 #include <sys/socket.h>
90 #include <sys/globaldata.h>
91 #include <sys/thread.h>
92 #include <sys/kernel.h>
93 #include <sys/syslog.h>
94 #include <sys/sysctl.h>
95 #include <sys/in_cksum.h>
97 #include <machine/stdarg.h>
100 #include <net/if_types.h>
101 #include <net/if_var.h>
102 #include <net/if_dl.h>
103 #include <net/pfil.h>
104 #include <net/route.h>
105 #include <net/netisr.h>
107 #include <netinet/in.h>
108 #include <netinet/in_systm.h>
109 #include <netinet/in_var.h>
110 #include <netinet/ip.h>
111 #include <netinet/in_pcb.h>
112 #include <netinet/ip_var.h>
113 #include <netinet/ip_icmp.h>
115 #include <sys/thread2.h>
116 #include <sys/msgport2.h>
117 #include <net/netmsg2.h>
119 #include <sys/socketvar.h>
121 #include <net/ipfw/ip_fw.h>
122 #include <net/dummynet/ip_dummynet.h>
125 #include <netinet6/ipsec.h>
126 #include <netproto/key/key.h>
130 #include <netproto/ipsec/ipsec.h>
131 #include <netproto/ipsec/key.h>
135 static int ip_rsvp_on
;
136 struct socket
*ip_rsvpd
;
138 int ipforwarding
= 0;
139 SYSCTL_INT(_net_inet_ip
, IPCTL_FORWARDING
, forwarding
, CTLFLAG_RW
,
140 &ipforwarding
, 0, "Enable IP forwarding between interfaces");
142 static int ipsendredirects
= 1; /* XXX */
143 SYSCTL_INT(_net_inet_ip
, IPCTL_SENDREDIRECTS
, redirect
, CTLFLAG_RW
,
144 &ipsendredirects
, 0, "Enable sending IP redirects");
146 int ip_defttl
= IPDEFTTL
;
147 SYSCTL_INT(_net_inet_ip
, IPCTL_DEFTTL
, ttl
, CTLFLAG_RW
,
148 &ip_defttl
, 0, "Maximum TTL on IP packets");
150 static int ip_dosourceroute
= 0;
151 SYSCTL_INT(_net_inet_ip
, IPCTL_SOURCEROUTE
, sourceroute
, CTLFLAG_RW
,
152 &ip_dosourceroute
, 0, "Enable forwarding source routed IP packets");
154 static int ip_acceptsourceroute
= 0;
155 SYSCTL_INT(_net_inet_ip
, IPCTL_ACCEPTSOURCEROUTE
, accept_sourceroute
,
156 CTLFLAG_RW
, &ip_acceptsourceroute
, 0,
157 "Enable accepting source routed IP packets");
159 static int ip_keepfaith
= 0;
160 SYSCTL_INT(_net_inet_ip
, IPCTL_KEEPFAITH
, keepfaith
, CTLFLAG_RW
,
162 "Enable packet capture for FAITH IPv4->IPv6 translator daemon");
164 static int nipq
= 0; /* total # of reass queues */
166 SYSCTL_INT(_net_inet_ip
, OID_AUTO
, maxfragpackets
, CTLFLAG_RW
,
168 "Maximum number of IPv4 fragment reassembly queue entries");
170 static int maxfragsperpacket
;
171 SYSCTL_INT(_net_inet_ip
, OID_AUTO
, maxfragsperpacket
, CTLFLAG_RW
,
172 &maxfragsperpacket
, 0,
173 "Maximum number of IPv4 fragments allowed per packet");
175 static int ip_sendsourcequench
= 0;
176 SYSCTL_INT(_net_inet_ip
, OID_AUTO
, sendsourcequench
, CTLFLAG_RW
,
177 &ip_sendsourcequench
, 0,
178 "Enable the transmission of source quench packets");
180 int ip_do_randomid
= 1;
181 SYSCTL_INT(_net_inet_ip
, OID_AUTO
, random_id
, CTLFLAG_RW
,
183 "Assign random ip_id values");
185 * XXX - Setting ip_checkinterface mostly implements the receive side of
186 * the Strong ES model described in RFC 1122, but since the routing table
187 * and transmit implementation do not implement the Strong ES model,
188 * setting this to 1 results in an odd hybrid.
190 * XXX - ip_checkinterface currently must be disabled if you use ipnat
191 * to translate the destination address to another local interface.
193 * XXX - ip_checkinterface must be disabled if you add IP aliases
194 * to the loopback interface instead of the interface where the
195 * packets for those addresses are received.
197 static int ip_checkinterface
= 0;
198 SYSCTL_INT(_net_inet_ip
, OID_AUTO
, check_interface
, CTLFLAG_RW
,
199 &ip_checkinterface
, 0, "Verify packet arrives on correct interface");
202 static int ipprintfs
= 0;
205 extern struct domain inetdomain
;
206 extern struct protosw inetsw
[];
207 u_char ip_protox
[IPPROTO_MAX
];
208 struct in_ifaddrhead in_ifaddrheads
[MAXCPU
]; /* first inet address */
209 struct in_ifaddrhashhead
*in_ifaddrhashtbls
[MAXCPU
];
210 /* inet addr hash table */
211 u_long in_ifaddrhmask
; /* mask for hash table */
213 struct ip_stats ipstats_percpu
[MAXCPU
];
216 sysctl_ipstats(SYSCTL_HANDLER_ARGS
)
220 for (cpu
= 0; cpu
< ncpus
; ++cpu
) {
221 if ((error
= SYSCTL_OUT(req
, &ipstats_percpu
[cpu
],
222 sizeof(struct ip_stats
))))
224 if ((error
= SYSCTL_IN(req
, &ipstats_percpu
[cpu
],
225 sizeof(struct ip_stats
))))
231 SYSCTL_PROC(_net_inet_ip
, IPCTL_STATS
, stats
, (CTLTYPE_OPAQUE
| CTLFLAG_RW
),
232 0, 0, sysctl_ipstats
, "S,ip_stats", "IP statistics");
234 SYSCTL_STRUCT(_net_inet_ip
, IPCTL_STATS
, stats
, CTLFLAG_RW
,
235 &ipstat
, ip_stats
, "IP statistics");
238 /* Packet reassembly stuff */
239 #define IPREASS_NHASH_LOG2 6
240 #define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2)
241 #define IPREASS_HMASK (IPREASS_NHASH - 1)
242 #define IPREASS_HASH(x,y) \
243 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK)
245 static struct ipq ipq
[IPREASS_NHASH
];
248 SYSCTL_INT(_net_inet_ip
, IPCTL_DEFMTU
, mtu
, CTLFLAG_RW
,
249 &ip_mtu
, 0, "Default MTU");
253 static int ipstealth
= 0;
254 SYSCTL_INT(_net_inet_ip
, OID_AUTO
, stealth
, CTLFLAG_RW
, &ipstealth
, 0, "");
256 static const int ipstealth
= 0;
261 ip_fw_chk_t
*ip_fw_chk_ptr
;
262 ip_fw_dn_io_t
*ip_fw_dn_io_ptr
;
267 struct pfil_head inet_pfil_hook
;
270 * XXX this is ugly -- the following two global variables are
271 * used to store packet state while it travels through the stack.
272 * Note that the code even makes assumptions on the size and
273 * alignment of fields inside struct ip_srcrt so e.g. adding some
274 * fields will break the code. This needs to be fixed.
276 * We need to save the IP options in case a protocol wants to respond
277 * to an incoming packet over the same route if the packet got here
278 * using IP source routing. This allows connection establishment and
279 * maintenance when the remote end is on a network that is not known
282 static int ip_nhops
= 0;
284 static struct ip_srcrt
{
285 struct in_addr dst
; /* final destination */
286 char nop
; /* one NOP to align */
287 char srcopt
[IPOPT_OFFSET
+ 1]; /* OPTVAL, OLEN and OFFSET */
288 struct in_addr route
[MAX_IPOPTLEN
/sizeof(struct in_addr
)];
291 static MALLOC_DEFINE(M_IPQ
, "ipq", "IP Fragment Management");
292 static struct malloc_pipe ipq_mpipe
;
294 static void save_rte (u_char
*, struct in_addr
);
295 static int ip_dooptions (struct mbuf
*m
, int,
296 struct sockaddr_in
*next_hop
);
297 static void ip_freef (struct ipq
*);
298 static void ip_input_handler (struct netmsg
*);
299 static struct mbuf
*ip_reass (struct mbuf
*, struct ipq
*,
300 struct ipq
*, u_int32_t
*);
303 * IP initialization: fill in IP protocol switch table.
304 * All protocols not implemented in kernel go to raw IP protocol handler.
316 * Make sure we can handle a reasonable number of fragments but
317 * cap it at 4000 (XXX).
319 mpipe_init(&ipq_mpipe
, M_IPQ
, sizeof(struct ipq
),
320 IFQ_MAXLEN
, 4000, 0, NULL
);
321 for (i
= 0; i
< ncpus
; ++i
) {
322 TAILQ_INIT(&in_ifaddrheads
[i
]);
323 in_ifaddrhashtbls
[i
] =
324 hashinit(INADDR_NHASH
, M_IFADDR
, &in_ifaddrhmask
);
326 pr
= pffindproto(PF_INET
, IPPROTO_RAW
, SOCK_RAW
);
329 for (i
= 0; i
< IPPROTO_MAX
; i
++)
330 ip_protox
[i
] = pr
- inetsw
;
331 for (pr
= inetdomain
.dom_protosw
;
332 pr
< inetdomain
.dom_protoswNPROTOSW
; pr
++)
333 if (pr
->pr_domain
->dom_family
== PF_INET
&&
334 pr
->pr_protocol
&& pr
->pr_protocol
!= IPPROTO_RAW
)
335 ip_protox
[pr
->pr_protocol
] = pr
- inetsw
;
337 inet_pfil_hook
.ph_type
= PFIL_TYPE_AF
;
338 inet_pfil_hook
.ph_af
= AF_INET
;
339 if ((i
= pfil_head_register(&inet_pfil_hook
)) != 0) {
340 kprintf("%s: WARNING: unable to register pfil hook, "
341 "error %d\n", __func__
, i
);
344 for (i
= 0; i
< IPREASS_NHASH
; i
++)
345 ipq
[i
].next
= ipq
[i
].prev
= &ipq
[i
];
347 maxnipq
= nmbclusters
/ 32;
348 maxfragsperpacket
= 16;
350 ip_id
= time_second
& 0xffff;
353 * Initialize IP statistics counters for each CPU.
357 for (cpu
= 0; cpu
< ncpus
; ++cpu
) {
358 bzero(&ipstats_percpu
[cpu
], sizeof(struct ip_stats
));
361 bzero(&ipstat
, sizeof(struct ip_stats
));
364 netisr_register(NETISR_IP
, ip_mport_in
, ip_input_handler
);
368 * XXX watch out this one. It is perhaps used as a cache for
369 * the most recently used route ? it is cleared in in_addroute()
370 * when a new route is successfully created.
372 struct route ipforward_rt
[MAXCPU
];
374 /* Do transport protocol processing. */
376 transport_processing_oncpu(struct mbuf
*m
, int hlen
, struct ip
*ip
)
379 * Switch out to protocol's input routine.
381 (*inetsw
[ip_protox
[ip
->ip_p
]].pr_input
)(m
, hlen
, ip
->ip_p
);
384 struct netmsg_transport_packet
{
385 struct netmsg nm_netmsg
;
386 struct mbuf
*nm_mbuf
;
391 transport_processing_handler(netmsg_t netmsg
)
393 struct netmsg_transport_packet
*msg
= (void *)netmsg
;
396 ip
= mtod(msg
->nm_mbuf
, struct ip
*);
397 transport_processing_oncpu(msg
->nm_mbuf
, msg
->nm_hlen
, ip
);
398 lwkt_replymsg(&msg
->nm_netmsg
.nm_lmsg
, 0);
402 ip_input_handler(struct netmsg
*msg0
)
404 struct mbuf
*m
= ((struct netmsg_packet
*)msg0
)->nm_packet
;
407 /* msg0 was embedded in the mbuf, do not reply! */
411 * IP input routine. Checksum and byte swap header. If fragmented
412 * try to reassemble. Process options. Pass to next level.
415 ip_input(struct mbuf
*m
)
419 struct in_ifaddr
*ia
= NULL
;
420 struct in_ifaddr_container
*iac
;
421 int i
, hlen
, checkif
;
423 struct in_addr pkt_dst
;
424 u_int32_t divert_info
= 0; /* packet divert/tee info */
425 struct ip_fw_args args
;
426 boolean_t using_srcrt
= FALSE
; /* forward (by PFIL_HOOKS) */
427 boolean_t needredispatch
= FALSE
;
428 struct in_addr odst
; /* original dst address(NAT) */
430 struct sockaddr_in
*next_hop
= NULL
;
432 struct tdb_ident
*tdbi
;
433 struct secpolicy
*sp
;
443 if (m
->m_pkthdr
.fw_flags
& IPFORWARD_MBUF_TAGGED
) {
445 mtag
= m_tag_find(m
, PACKET_TAG_IPFORWARD
, NULL
);
446 KKASSERT(mtag
!= NULL
);
447 next_hop
= m_tag_data(mtag
);
450 /* Extract info from dummynet tag */
451 mtag
= m_tag_find(m
, PACKET_TAG_DUMMYNET
, NULL
);
453 args
.rule
= ((struct dn_pkt
*)m_tag_data(mtag
))->dn_priv
;
454 m_tag_delete(m
, mtag
);
457 if (args
.rule
!= NULL
) { /* dummynet already filtered us */
458 ip
= mtod(m
, struct ip
*);
459 hlen
= IP_VHL_HL(ip
->ip_vhl
) << 2;
465 /* length checks already done in ip_demux() */
466 KASSERT(m
->m_len
>= sizeof(ip
), ("IP header not in one mbuf"));
468 ip
= mtod(m
, struct ip
*);
470 if (IP_VHL_V(ip
->ip_vhl
) != IPVERSION
) {
471 ipstat
.ips_badvers
++;
475 hlen
= IP_VHL_HL(ip
->ip_vhl
) << 2;
476 /* length checks already done in ip_demux() */
477 KASSERT(hlen
>= sizeof(struct ip
), ("IP header len too small"));
478 KASSERT(m
->m_len
>= hlen
, ("packet shorter than IP header length"));
480 /* 127/8 must not appear on wire - RFC1122 */
481 if ((ntohl(ip
->ip_dst
.s_addr
) >> IN_CLASSA_NSHIFT
) == IN_LOOPBACKNET
||
482 (ntohl(ip
->ip_src
.s_addr
) >> IN_CLASSA_NSHIFT
) == IN_LOOPBACKNET
) {
483 if (!(m
->m_pkthdr
.rcvif
->if_flags
& IFF_LOOPBACK
)) {
484 ipstat
.ips_badaddr
++;
489 if (m
->m_pkthdr
.csum_flags
& CSUM_IP_CHECKED
) {
490 sum
= !(m
->m_pkthdr
.csum_flags
& CSUM_IP_VALID
);
492 if (hlen
== sizeof(struct ip
)) {
493 sum
= in_cksum_hdr(ip
);
495 sum
= in_cksum(m
, hlen
);
504 if (altq_input
!= NULL
&& (*altq_input
)(m
, AF_INET
) == 0) {
505 /* packet is dropped by traffic conditioner */
510 * Convert fields to host representation.
512 ip
->ip_len
= ntohs(ip
->ip_len
);
513 if (ip
->ip_len
< hlen
) {
517 ip
->ip_off
= ntohs(ip
->ip_off
);
520 * Check that the amount of data in the buffers
521 * is as at least much as the IP header would have us expect.
522 * Trim mbufs if longer than we expect.
523 * Drop packet if shorter than we expect.
525 if (m
->m_pkthdr
.len
< ip
->ip_len
) {
526 ipstat
.ips_tooshort
++;
529 if (m
->m_pkthdr
.len
> ip
->ip_len
) {
530 if (m
->m_len
== m
->m_pkthdr
.len
) {
531 m
->m_len
= ip
->ip_len
;
532 m
->m_pkthdr
.len
= ip
->ip_len
;
534 m_adj(m
, ip
->ip_len
- m
->m_pkthdr
.len
);
536 #if defined(IPSEC) && !defined(IPSEC_FILTERGIF)
538 * Bypass packet filtering for packets from a tunnel (gif).
540 if (ipsec_gethist(m
, NULL
))
546 * Right now when no processing on packet has done
547 * and it is still fresh out of network we do our black
549 * - Firewall: deny/allow/divert
550 * - Xlate: translate packet's addr/port (NAT).
551 * - Pipe: pass pkt through dummynet.
552 * - Wrap: fake packet's addr/port <unimpl.>
553 * - Encapsulate: put it in another IP and send out. <unimp.>
559 * Run through list of hooks for input packets.
561 * NB: Beware of the destination address changing (e.g.
562 * by NAT rewriting). When this happens, tell
563 * ip_forward to do the right thing.
565 if (pfil_has_hooks(&inet_pfil_hook
)) {
567 if (pfil_run_hooks(&inet_pfil_hook
, &m
,
568 m
->m_pkthdr
.rcvif
, PFIL_IN
)) {
571 if (m
== NULL
) /* consumed by filter */
573 ip
= mtod(m
, struct ip
*);
574 using_srcrt
= (odst
.s_addr
!= ip
->ip_dst
.s_addr
);
577 if (fw_enable
&& IPFW_LOADED
) {
579 * If we've been forwarded from the output side, then
580 * skip the firewall a second time
582 if (next_hop
!= NULL
)
586 i
= ip_fw_chk_ptr(&args
);
589 if ((i
& IP_FW_PORT_DENY_FLAG
) || m
== NULL
) { /* drop */
594 ip
= mtod(m
, struct ip
*); /* just in case m changed */
596 if (m
->m_pkthdr
.fw_flags
& IPFORWARD_MBUF_TAGGED
) {
597 mtag
= m_tag_find(m
, PACKET_TAG_IPFORWARD
, NULL
);
598 KKASSERT(mtag
!= NULL
);
599 next_hop
= m_tag_data(mtag
);
602 if (i
== 0 && next_hop
== NULL
) /* common case */
604 if (i
& IP_FW_PORT_DYNT_FLAG
) {
605 /* Send packet to the appropriate pipe */
606 ip_fw_dn_io_ptr(m
, i
&0xffff, DN_TO_IP_IN
, &args
);
610 if (i
!= 0 && !(i
& IP_FW_PORT_DYNT_FLAG
)) {
611 /* Divert or tee packet */
616 if (i
== 0 && next_hop
!= NULL
)
619 * if we get here, the packet must be dropped
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 ip_nhops
= 0; /* for source routed packets */
633 if (hlen
> sizeof(struct ip
) && ip_dooptions(m
, 0, next_hop
))
636 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
637 * matter if it is destined to another node, or whether it is
638 * a multicast one, RSVP wants it! and prevents it from being forwarded
639 * anywhere else. Also checks if the rsvp daemon is running before
640 * grabbing the packet.
642 if (rsvp_on
&& ip
->ip_p
== IPPROTO_RSVP
)
646 * Check our list of addresses, to see if the packet is for us.
647 * If we don't have any addresses, assume any unicast packet
648 * we receive might be for us (and let the upper layers deal
651 if (TAILQ_EMPTY(&in_ifaddrheads
[mycpuid
]) &&
652 !(m
->m_flags
& (M_MCAST
| M_BCAST
)))
656 * Cache the destination address of the packet; this may be
657 * changed by use of 'ipfw fwd'.
659 pkt_dst
= next_hop
? next_hop
->sin_addr
: ip
->ip_dst
;
662 * Enable a consistency check between the destination address
663 * and the arrival interface for a unicast packet (the RFC 1122
664 * strong ES model) if IP forwarding is disabled and the packet
665 * is not locally generated and the packet is not subject to
668 * XXX - Checking also should be disabled if the destination
669 * address is ipnat'ed to a different interface.
671 * XXX - Checking is incompatible with IP aliases added
672 * to the loopback interface instead of the interface where
673 * the packets are received.
675 checkif
= ip_checkinterface
&&
677 m
->m_pkthdr
.rcvif
!= NULL
&&
678 !(m
->m_pkthdr
.rcvif
->if_flags
& IFF_LOOPBACK
) &&
682 * Check for exact addresses in the hash bucket.
684 LIST_FOREACH(iac
, INADDR_HASH(pkt_dst
.s_addr
), ia_hash
) {
688 * If the address matches, verify that the packet
689 * arrived via the correct interface if checking is
692 if (IA_SIN(ia
)->sin_addr
.s_addr
== pkt_dst
.s_addr
&&
693 (!checkif
|| ia
->ia_ifp
== m
->m_pkthdr
.rcvif
))
699 * Check for broadcast addresses.
701 * Only accept broadcast packets that arrive via the matching
702 * interface. Reception of forwarded directed broadcasts would
703 * be handled via ip_forward() and ether_output() with the loopback
704 * into the stack for SIMPLEX interfaces handled by ether_output().
706 if (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 if (ip_mrouter
!= NULL
) {
734 * If we are acting as a multicast router, all
735 * incoming multicast packets are passed to the
736 * kernel-level multicast forwarding function.
737 * The packet is returned (relatively) intact; if
738 * ip_mforward() returns a non-zero value, the packet
739 * must be discarded, else it may be accepted below.
741 if (ip_mforward
!= NULL
&&
742 ip_mforward(ip
, m
->m_pkthdr
.rcvif
, m
, NULL
) != 0) {
743 ipstat
.ips_cantforward
++;
749 * The process-level routing daemon needs to receive
750 * all multicast IGMP packets, whether or not this
751 * host belongs to their destination groups.
753 if (ip
->ip_p
== IPPROTO_IGMP
)
755 ipstat
.ips_forward
++;
758 * See if we belong to the destination multicast group on the
761 IN_LOOKUP_MULTI(ip
->ip_dst
, m
->m_pkthdr
.rcvif
, inm
);
763 ipstat
.ips_notmember
++;
769 if (ip
->ip_dst
.s_addr
== INADDR_BROADCAST
)
771 if (ip
->ip_dst
.s_addr
== INADDR_ANY
)
775 * FAITH(Firewall Aided Internet Translator)
777 if (m
->m_pkthdr
.rcvif
&& m
->m_pkthdr
.rcvif
->if_type
== IFT_FAITH
) {
779 if (ip
->ip_p
== IPPROTO_TCP
|| ip
->ip_p
== IPPROTO_ICMP
)
787 * Not for us; forward if possible and desirable.
790 ipstat
.ips_cantforward
++;
795 * Enforce inbound IPsec SPD.
797 if (ipsec4_in_reject(m
, NULL
)) {
798 ipsecstat
.in_polvio
++;
803 mtag
= m_tag_find(m
, PACKET_TAG_IPSEC_IN_DONE
, NULL
);
806 tdbi
= (struct tdb_ident
*)m_tag_data(mtag
);
807 sp
= ipsec_getpolicy(tdbi
, IPSEC_DIR_INBOUND
);
809 sp
= ipsec_getpolicybyaddr(m
, IPSEC_DIR_INBOUND
,
810 IP_FORWARDING
, &error
);
812 if (sp
== NULL
) { /* NB: can happen if error */
814 /*XXX error stat???*/
815 DPRINTF(("ip_input: no SP for forwarding\n")); /*XXX*/
820 * Check security policy against packet attributes.
822 error
= ipsec_in_reject(sp
, m
);
826 ipstat
.ips_cantforward
++;
830 ip_forward(m
, using_srcrt
, next_hop
);
837 * IPSTEALTH: Process non-routing options only
838 * if the packet is destined for us.
841 hlen
> sizeof(struct ip
) &&
842 ip_dooptions(m
, 1, next_hop
))
845 /* Count the packet in the ip address stats */
847 ia
->ia_ifa
.if_ipackets
++;
848 ia
->ia_ifa
.if_ibytes
+= m
->m_pkthdr
.len
;
852 * If offset or IP_MF are set, must reassemble.
853 * Otherwise, nothing need be done.
854 * (We could look in the reassembly queue to see
855 * if the packet was previously fragmented,
856 * but it's not worth the time; just let them time out.)
858 if (ip
->ip_off
& (IP_MF
| IP_OFFMASK
)) {
860 /* If maxnipq is 0, never accept fragments. */
862 ipstat
.ips_fragments
++;
863 ipstat
.ips_fragdropped
++;
867 sum
= IPREASS_HASH(ip
->ip_src
.s_addr
, ip
->ip_id
);
869 * Look for queue of fragments
872 for (fp
= ipq
[sum
].next
; fp
!= &ipq
[sum
]; fp
= fp
->next
)
873 if (ip
->ip_id
== fp
->ipq_id
&&
874 ip
->ip_src
.s_addr
== fp
->ipq_src
.s_addr
&&
875 ip
->ip_dst
.s_addr
== fp
->ipq_dst
.s_addr
&&
876 ip
->ip_p
== fp
->ipq_p
)
882 * Enforce upper bound on number of fragmented packets
883 * for which we attempt reassembly;
884 * If maxnipq is -1, accept all fragments without limitation.
886 if ((nipq
> maxnipq
) && (maxnipq
> 0)) {
888 * drop something from the tail of the current queue
889 * before proceeding further
891 if (ipq
[sum
].prev
== &ipq
[sum
]) { /* gak */
892 for (i
= 0; i
< IPREASS_NHASH
; i
++) {
893 if (ipq
[i
].prev
!= &ipq
[i
]) {
894 ipstat
.ips_fragtimeout
+=
895 ipq
[i
].prev
->ipq_nfrags
;
896 ip_freef(ipq
[i
].prev
);
901 ipstat
.ips_fragtimeout
+=
902 ipq
[sum
].prev
->ipq_nfrags
;
903 ip_freef(ipq
[sum
].prev
);
908 * Adjust ip_len to not reflect header,
909 * convert offset of this to bytes.
912 if (ip
->ip_off
& IP_MF
) {
914 * Make sure that fragments have a data length
915 * that's a non-zero multiple of 8 bytes.
917 if (ip
->ip_len
== 0 || (ip
->ip_len
& 0x7) != 0) {
918 ipstat
.ips_toosmall
++; /* XXX */
921 m
->m_flags
|= M_FRAG
;
923 m
->m_flags
&= ~M_FRAG
;
927 * Attempt reassembly; if it succeeds, proceed.
928 * ip_reass() will return a different mbuf, and update
929 * the divert info in divert_info.
931 ipstat
.ips_fragments
++;
932 m
->m_pkthdr
.header
= ip
;
933 m
= ip_reass(m
, fp
, &ipq
[sum
], &divert_info
);
936 ipstat
.ips_reassembled
++;
937 needredispatch
= TRUE
;
938 ip
= mtod(m
, struct ip
*);
939 /* Get the header length of the reassembled packet */
940 hlen
= IP_VHL_HL(ip
->ip_vhl
) << 2;
942 /* Restore original checksum before diverting packet */
943 if (divert_info
!= 0) {
945 ip
->ip_len
= htons(ip
->ip_len
);
946 ip
->ip_off
= htons(ip
->ip_off
);
948 if (hlen
== sizeof(struct ip
))
949 ip
->ip_sum
= in_cksum_hdr(ip
);
951 ip
->ip_sum
= in_cksum(m
, hlen
);
952 ip
->ip_off
= ntohs(ip
->ip_off
);
953 ip
->ip_len
= ntohs(ip
->ip_len
);
963 * Divert or tee packet to the divert protocol if required.
965 if (divert_info
!= 0) {
966 struct mbuf
*clone
= NULL
;
968 /* Clone packet if we're doing a 'tee' */
969 if ((divert_info
& IP_FW_PORT_TEE_FLAG
) != 0)
970 clone
= m_dup(m
, MB_DONTWAIT
);
972 /* Restore packet header fields to original values */
974 ip
->ip_len
= htons(ip
->ip_len
);
975 ip
->ip_off
= htons(ip
->ip_off
);
977 /* Deliver packet to divert input routine */
978 divert_packet(m
, 1, divert_info
& 0xffff);
979 ipstat
.ips_delivered
++;
981 /* If 'tee', continue with original packet */
985 ip
= mtod(m
, struct ip
*);
988 * Jump backwards to complete processing of the
989 * packet. But first clear divert_info to avoid
990 * entering this block again.
991 * We do not need to clear args.divert_rule as
992 * it will not be used.
994 * XXX Better safe than sorry, remove the DIVERT tag.
996 mtag
= m_tag_find(m
, PACKET_TAG_IPFW_DIVERT
, NULL
);
998 m_tag_delete(m
, mtag
);
1007 * enforce IPsec policy checking if we are seeing last header.
1008 * note that we do not visit this with protocols with pcb layer
1009 * code - like udp/tcp/raw ip.
1011 if ((inetsw
[ip_protox
[ip
->ip_p
]].pr_flags
& PR_LASTHDR
) &&
1012 ipsec4_in_reject(m
, NULL
)) {
1013 ipsecstat
.in_polvio
++;
1019 * enforce IPsec policy checking if we are seeing last header.
1020 * note that we do not visit this with protocols with pcb layer
1021 * code - like udp/tcp/raw ip.
1023 if (inetsw
[ip_protox
[ip
->ip_p
]].pr_flags
& PR_LASTHDR
) {
1025 * Check if the packet has already had IPsec processing
1026 * done. If so, then just pass it along. This tag gets
1027 * set during AH, ESP, etc. input handling, before the
1028 * packet is returned to the ip input queue for delivery.
1030 mtag
= m_tag_find(m
, PACKET_TAG_IPSEC_IN_DONE
, NULL
);
1033 tdbi
= (struct tdb_ident
*)m_tag_data(mtag
);
1034 sp
= ipsec_getpolicy(tdbi
, IPSEC_DIR_INBOUND
);
1036 sp
= ipsec_getpolicybyaddr(m
, IPSEC_DIR_INBOUND
,
1037 IP_FORWARDING
, &error
);
1041 * Check security policy against packet attributes.
1043 error
= ipsec_in_reject(sp
, m
);
1046 /* XXX error stat??? */
1048 DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/
1055 #endif /* FAST_IPSEC */
1057 ipstat
.ips_delivered
++;
1058 if (needredispatch
) {
1059 struct netmsg_transport_packet
*msg
;
1062 ip
->ip_off
= htons(ip
->ip_off
);
1063 ip
->ip_len
= htons(ip
->ip_len
);
1064 port
= ip_mport_in(&m
);
1068 msg
= kmalloc(sizeof(struct netmsg_transport_packet
), M_LWKTMSG
,
1069 M_INTWAIT
| M_NULLOK
);
1073 netmsg_init(&msg
->nm_netmsg
, &netisr_afree_rport
, 0,
1074 transport_processing_handler
);
1075 msg
->nm_hlen
= hlen
;
1078 ip
= mtod(m
, struct ip
*);
1079 ip
->ip_len
= ntohs(ip
->ip_len
);
1080 ip
->ip_off
= ntohs(ip
->ip_off
);
1081 lwkt_sendmsg(port
, &msg
->nm_netmsg
.nm_lmsg
);
1083 transport_processing_oncpu(m
, hlen
, ip
);
1092 * Take incoming datagram fragment and try to reassemble it into
1093 * whole datagram. If a chain for reassembly of this datagram already
1094 * exists, then it is given as fp; otherwise have to make a chain.
1096 * When IPDIVERT enabled, keep additional state with each packet that
1097 * tells us if we need to divert or tee the packet we're building.
1098 * In particular, *divinfo includes the port and TEE flag.
1101 static struct mbuf
*
1102 ip_reass(struct mbuf
*m
, struct ipq
*fp
, struct ipq
*where
,
1105 struct ip
*ip
= mtod(m
, struct ip
*);
1106 struct mbuf
*p
= NULL
, *q
, *nq
;
1108 int hlen
= IP_VHL_HL(ip
->ip_vhl
) << 2;
1115 * If the hardware has not done csum over this fragment
1116 * then csum_data is not valid at all.
1118 if ((m
->m_pkthdr
.csum_flags
& (CSUM_FRAG_NOT_CHECKED
| CSUM_DATA_VALID
))
1119 == (CSUM_FRAG_NOT_CHECKED
| CSUM_DATA_VALID
)) {
1120 m
->m_pkthdr
.csum_data
= 0;
1121 m
->m_pkthdr
.csum_flags
&= ~(CSUM_DATA_VALID
| CSUM_PSEUDO_HDR
);
1125 * Presence of header sizes in mbufs
1126 * would confuse code below.
1132 * If first fragment to arrive, create a reassembly queue.
1135 if ((fp
= mpipe_alloc_nowait(&ipq_mpipe
)) == NULL
)
1140 fp
->ipq_ttl
= IPFRAGTTL
;
1141 fp
->ipq_p
= ip
->ip_p
;
1142 fp
->ipq_id
= ip
->ip_id
;
1143 fp
->ipq_src
= ip
->ip_src
;
1144 fp
->ipq_dst
= ip
->ip_dst
;
1146 m
->m_nextpkt
= NULL
;
1148 fp
->ipq_div_info
= 0;
1155 #define GETIP(m) ((struct ip*)((m)->m_pkthdr.header))
1158 * Find a segment which begins after this one does.
1160 for (p
= NULL
, q
= fp
->ipq_frags
; q
; p
= q
, q
= q
->m_nextpkt
)
1161 if (GETIP(q
)->ip_off
> ip
->ip_off
)
1165 * If there is a preceding segment, it may provide some of
1166 * our data already. If so, drop the data from the incoming
1167 * segment. If it provides all of our data, drop us, otherwise
1168 * stick new segment in the proper place.
1170 * If some of the data is dropped from the the preceding
1171 * segment, then it's checksum is invalidated.
1174 i
= GETIP(p
)->ip_off
+ GETIP(p
)->ip_len
- ip
->ip_off
;
1176 if (i
>= ip
->ip_len
)
1179 m
->m_pkthdr
.csum_flags
= 0;
1183 m
->m_nextpkt
= p
->m_nextpkt
;
1186 m
->m_nextpkt
= fp
->ipq_frags
;
1191 * While we overlap succeeding segments trim them or,
1192 * if they are completely covered, dequeue them.
1194 for (; q
!= NULL
&& ip
->ip_off
+ ip
->ip_len
> GETIP(q
)->ip_off
;
1196 i
= (ip
->ip_off
+ ip
->ip_len
) - GETIP(q
)->ip_off
;
1197 if (i
< GETIP(q
)->ip_len
) {
1198 GETIP(q
)->ip_len
-= i
;
1199 GETIP(q
)->ip_off
+= i
;
1201 q
->m_pkthdr
.csum_flags
= 0;
1206 ipstat
.ips_fragdropped
++;
1208 q
->m_nextpkt
= NULL
;
1216 * Transfer firewall instructions to the fragment structure.
1217 * Only trust info in the fragment at offset 0.
1219 if (ip
->ip_off
== 0) {
1220 fp
->ipq_div_info
= *divinfo
;
1222 mtag
= m_tag_find(m
, PACKET_TAG_IPFW_DIVERT
, NULL
);
1224 m_tag_delete(m
, mtag
);
1230 * Check for complete reassembly and perform frag per packet
1233 * Frag limiting is performed here so that the nth frag has
1234 * a chance to complete the packet before we drop the packet.
1235 * As a result, n+1 frags are actually allowed per packet, but
1236 * only n will ever be stored. (n = maxfragsperpacket.)
1240 for (p
= NULL
, q
= fp
->ipq_frags
; q
; p
= q
, q
= q
->m_nextpkt
) {
1241 if (GETIP(q
)->ip_off
!= next
) {
1242 if (fp
->ipq_nfrags
> maxfragsperpacket
) {
1243 ipstat
.ips_fragdropped
+= fp
->ipq_nfrags
;
1248 next
+= GETIP(q
)->ip_len
;
1250 /* Make sure the last packet didn't have the IP_MF flag */
1251 if (p
->m_flags
& M_FRAG
) {
1252 if (fp
->ipq_nfrags
> maxfragsperpacket
) {
1253 ipstat
.ips_fragdropped
+= fp
->ipq_nfrags
;
1260 * Reassembly is complete. Make sure the packet is a sane size.
1264 if (next
+ (IP_VHL_HL(ip
->ip_vhl
) << 2) > IP_MAXPACKET
) {
1265 ipstat
.ips_toolong
++;
1266 ipstat
.ips_fragdropped
+= fp
->ipq_nfrags
;
1272 * Concatenate fragments.
1279 q
->m_nextpkt
= NULL
;
1280 for (q
= nq
; q
!= NULL
; q
= nq
) {
1282 q
->m_nextpkt
= NULL
;
1283 m
->m_pkthdr
.csum_flags
&= q
->m_pkthdr
.csum_flags
;
1284 m
->m_pkthdr
.csum_data
+= q
->m_pkthdr
.csum_data
;
1289 * Clean up the 1's complement checksum. Carry over 16 bits must
1290 * be added back. This assumes no more then 65535 packet fragments
1291 * were reassembled. A second carry can also occur (but not a third).
1293 m
->m_pkthdr
.csum_data
= (m
->m_pkthdr
.csum_data
& 0xffff) +
1294 (m
->m_pkthdr
.csum_data
>> 16);
1295 if (m
->m_pkthdr
.csum_data
> 0xFFFF)
1296 m
->m_pkthdr
.csum_data
-= 0xFFFF;
1301 * Extract firewall instructions from the fragment structure.
1303 *divinfo
= fp
->ipq_div_info
;
1307 * Create header for new ip packet by
1308 * modifying header of first packet;
1309 * dequeue and discard fragment reassembly header.
1310 * Make header visible.
1313 ip
->ip_src
= fp
->ipq_src
;
1314 ip
->ip_dst
= fp
->ipq_dst
;
1317 mpipe_free(&ipq_mpipe
, fp
);
1318 m
->m_len
+= (IP_VHL_HL(ip
->ip_vhl
) << 2);
1319 m
->m_data
-= (IP_VHL_HL(ip
->ip_vhl
) << 2);
1320 /* some debugging cruft by sklower, below, will go away soon */
1321 if (m
->m_flags
& M_PKTHDR
) { /* XXX this should be done elsewhere */
1324 for (n
= m
; n
; n
= n
->m_next
)
1326 m
->m_pkthdr
.len
= plen
;
1334 ipstat
.ips_fragdropped
++;
1344 * Free a fragment reassembly header and all
1345 * associated datagrams.
1348 ip_freef(struct ipq
*fp
)
1352 while (fp
->ipq_frags
) {
1354 fp
->ipq_frags
= q
->m_nextpkt
;
1355 q
->m_nextpkt
= NULL
;
1359 mpipe_free(&ipq_mpipe
, fp
);
1364 * IP timer processing;
1365 * if a timer expires on a reassembly
1366 * queue, discard it.
1375 for (i
= 0; i
< IPREASS_NHASH
; i
++) {
1379 while (fp
!= &ipq
[i
]) {
1382 if (fp
->prev
->ipq_ttl
== 0) {
1383 ipstat
.ips_fragtimeout
+= fp
->prev
->ipq_nfrags
;
1389 * If we are over the maximum number of fragments
1390 * (due to the limit being lowered), drain off
1391 * enough to get down to the new limit.
1393 if (maxnipq
>= 0 && nipq
> maxnipq
) {
1394 for (i
= 0; i
< IPREASS_NHASH
; i
++) {
1395 while (nipq
> maxnipq
&&
1396 (ipq
[i
].next
!= &ipq
[i
])) {
1397 ipstat
.ips_fragdropped
+=
1398 ipq
[i
].next
->ipq_nfrags
;
1399 ip_freef(ipq
[i
].next
);
1408 * Drain off all datagram fragments.
1415 for (i
= 0; i
< IPREASS_NHASH
; i
++) {
1416 while (ipq
[i
].next
!= &ipq
[i
]) {
1417 ipstat
.ips_fragdropped
+= ipq
[i
].next
->ipq_nfrags
;
1418 ip_freef(ipq
[i
].next
);
1425 * Do option processing on a datagram,
1426 * possibly discarding it if bad options are encountered,
1427 * or forwarding it if source-routed.
1428 * The pass argument is used when operating in the IPSTEALTH
1429 * mode to tell what options to process:
1430 * [LS]SRR (pass 0) or the others (pass 1).
1431 * The reason for as many as two passes is that when doing IPSTEALTH,
1432 * non-routing options should be processed only if the packet is for us.
1433 * Returns 1 if packet has been forwarded/freed,
1434 * 0 if the packet should be processed further.
1437 ip_dooptions(struct mbuf
*m
, int pass
, struct sockaddr_in
*next_hop
)
1439 struct sockaddr_in ipaddr
= { sizeof ipaddr
, AF_INET
};
1440 struct ip
*ip
= mtod(m
, struct ip
*);
1442 struct in_ifaddr
*ia
;
1443 int opt
, optlen
, cnt
, off
, code
, type
= ICMP_PARAMPROB
;
1444 boolean_t forward
= FALSE
;
1445 struct in_addr
*sin
, dst
;
1449 cp
= (u_char
*)(ip
+ 1);
1450 cnt
= (IP_VHL_HL(ip
->ip_vhl
) << 2) - sizeof(struct ip
);
1451 for (; cnt
> 0; cnt
-= optlen
, cp
+= optlen
) {
1452 opt
= cp
[IPOPT_OPTVAL
];
1453 if (opt
== IPOPT_EOL
)
1455 if (opt
== IPOPT_NOP
)
1458 if (cnt
< IPOPT_OLEN
+ sizeof(*cp
)) {
1459 code
= &cp
[IPOPT_OLEN
] - (u_char
*)ip
;
1462 optlen
= cp
[IPOPT_OLEN
];
1463 if (optlen
< IPOPT_OLEN
+ sizeof(*cp
) || optlen
> cnt
) {
1464 code
= &cp
[IPOPT_OLEN
] - (u_char
*)ip
;
1474 * Source routing with record.
1475 * Find interface with current destination address.
1476 * If none on this machine then drop if strictly routed,
1477 * or do nothing if loosely routed.
1478 * Record interface address and bring up next address
1479 * component. If strictly routed make sure next
1480 * address is on directly accessible net.
1484 if (ipstealth
&& pass
> 0)
1486 if (optlen
< IPOPT_OFFSET
+ sizeof(*cp
)) {
1487 code
= &cp
[IPOPT_OLEN
] - (u_char
*)ip
;
1490 if ((off
= cp
[IPOPT_OFFSET
]) < IPOPT_MINOFF
) {
1491 code
= &cp
[IPOPT_OFFSET
] - (u_char
*)ip
;
1494 ipaddr
.sin_addr
= ip
->ip_dst
;
1495 ia
= (struct in_ifaddr
*)
1496 ifa_ifwithaddr((struct sockaddr
*)&ipaddr
);
1498 if (opt
== IPOPT_SSRR
) {
1499 type
= ICMP_UNREACH
;
1500 code
= ICMP_UNREACH_SRCFAIL
;
1503 if (!ip_dosourceroute
)
1504 goto nosourcerouting
;
1506 * Loose routing, and not at next destination
1507 * yet; nothing to do except forward.
1511 off
--; /* 0 origin */
1512 if (off
> optlen
- (int)sizeof(struct in_addr
)) {
1514 * End of source route. Should be for us.
1516 if (!ip_acceptsourceroute
)
1517 goto nosourcerouting
;
1518 save_rte(cp
, ip
->ip_src
);
1523 if (!ip_dosourceroute
) {
1525 char buf
[sizeof "aaa.bbb.ccc.ddd"];
1528 * Acting as a router, so generate ICMP
1531 strcpy(buf
, inet_ntoa(ip
->ip_dst
));
1533 "attempted source route from %s to %s\n",
1534 inet_ntoa(ip
->ip_src
), buf
);
1535 type
= ICMP_UNREACH
;
1536 code
= ICMP_UNREACH_SRCFAIL
;
1540 * Not acting as a router,
1544 ipstat
.ips_cantforward
++;
1551 * locate outgoing interface
1553 memcpy(&ipaddr
.sin_addr
, cp
+ off
,
1554 sizeof ipaddr
.sin_addr
);
1556 if (opt
== IPOPT_SSRR
) {
1557 #define INA struct in_ifaddr *
1558 #define SA struct sockaddr *
1559 if ((ia
= (INA
)ifa_ifwithdstaddr((SA
)&ipaddr
))
1561 ia
= (INA
)ifa_ifwithnet((SA
)&ipaddr
);
1563 ia
= ip_rtaddr(ipaddr
.sin_addr
,
1564 &ipforward_rt
[mycpuid
]);
1566 type
= ICMP_UNREACH
;
1567 code
= ICMP_UNREACH_SRCFAIL
;
1570 ip
->ip_dst
= ipaddr
.sin_addr
;
1571 memcpy(cp
+ off
, &IA_SIN(ia
)->sin_addr
,
1572 sizeof(struct in_addr
));
1573 cp
[IPOPT_OFFSET
] += sizeof(struct in_addr
);
1575 * Let ip_intr's mcast routing check handle mcast pkts
1577 forward
= !IN_MULTICAST(ntohl(ip
->ip_dst
.s_addr
));
1581 if (ipstealth
&& pass
== 0)
1583 if (optlen
< IPOPT_OFFSET
+ sizeof(*cp
)) {
1584 code
= &cp
[IPOPT_OFFSET
] - (u_char
*)ip
;
1587 if ((off
= cp
[IPOPT_OFFSET
]) < IPOPT_MINOFF
) {
1588 code
= &cp
[IPOPT_OFFSET
] - (u_char
*)ip
;
1592 * If no space remains, ignore.
1594 off
--; /* 0 origin */
1595 if (off
> optlen
- (int)sizeof(struct in_addr
))
1597 memcpy(&ipaddr
.sin_addr
, &ip
->ip_dst
,
1598 sizeof ipaddr
.sin_addr
);
1600 * locate outgoing interface; if we're the destination,
1601 * use the incoming interface (should be same).
1603 if ((ia
= (INA
)ifa_ifwithaddr((SA
)&ipaddr
)) == NULL
&&
1604 (ia
= ip_rtaddr(ipaddr
.sin_addr
,
1605 &ipforward_rt
[mycpuid
]))
1607 type
= ICMP_UNREACH
;
1608 code
= ICMP_UNREACH_HOST
;
1611 memcpy(cp
+ off
, &IA_SIN(ia
)->sin_addr
,
1612 sizeof(struct in_addr
));
1613 cp
[IPOPT_OFFSET
] += sizeof(struct in_addr
);
1617 if (ipstealth
&& pass
== 0)
1619 code
= cp
- (u_char
*)ip
;
1620 if (optlen
< 4 || optlen
> 40) {
1621 code
= &cp
[IPOPT_OLEN
] - (u_char
*)ip
;
1624 if ((off
= cp
[IPOPT_OFFSET
]) < 5) {
1625 code
= &cp
[IPOPT_OLEN
] - (u_char
*)ip
;
1628 if (off
> optlen
- (int)sizeof(int32_t)) {
1629 cp
[IPOPT_OFFSET
+ 1] += (1 << 4);
1630 if ((cp
[IPOPT_OFFSET
+ 1] & 0xf0) == 0) {
1631 code
= &cp
[IPOPT_OFFSET
] - (u_char
*)ip
;
1636 off
--; /* 0 origin */
1637 sin
= (struct in_addr
*)(cp
+ off
);
1638 switch (cp
[IPOPT_OFFSET
+ 1] & 0x0f) {
1640 case IPOPT_TS_TSONLY
:
1643 case IPOPT_TS_TSANDADDR
:
1644 if (off
+ sizeof(n_time
) +
1645 sizeof(struct in_addr
) > optlen
) {
1646 code
= &cp
[IPOPT_OFFSET
] - (u_char
*)ip
;
1649 ipaddr
.sin_addr
= dst
;
1650 ia
= (INA
)ifaof_ifpforaddr((SA
)&ipaddr
,
1654 memcpy(sin
, &IA_SIN(ia
)->sin_addr
,
1655 sizeof(struct in_addr
));
1656 cp
[IPOPT_OFFSET
] += sizeof(struct in_addr
);
1657 off
+= sizeof(struct in_addr
);
1660 case IPOPT_TS_PRESPEC
:
1661 if (off
+ sizeof(n_time
) +
1662 sizeof(struct in_addr
) > optlen
) {
1663 code
= &cp
[IPOPT_OFFSET
] - (u_char
*)ip
;
1666 memcpy(&ipaddr
.sin_addr
, sin
,
1667 sizeof(struct in_addr
));
1668 if (ifa_ifwithaddr((SA
)&ipaddr
) == NULL
)
1670 cp
[IPOPT_OFFSET
] += sizeof(struct in_addr
);
1671 off
+= sizeof(struct in_addr
);
1675 code
= &cp
[IPOPT_OFFSET
+ 1] - (u_char
*)ip
;
1679 memcpy(cp
+ off
, &ntime
, sizeof(n_time
));
1680 cp
[IPOPT_OFFSET
] += sizeof(n_time
);
1683 if (forward
&& ipforwarding
) {
1684 ip_forward(m
, TRUE
, next_hop
);
1689 icmp_error(m
, type
, code
, 0, 0);
1690 ipstat
.ips_badoptions
++;
1695 * Given address of next destination (final or next hop),
1696 * return internet address info of interface to be used to get there.
1699 ip_rtaddr(struct in_addr dst
, struct route
*ro
)
1701 struct sockaddr_in
*sin
;
1703 sin
= (struct sockaddr_in
*)&ro
->ro_dst
;
1705 if (ro
->ro_rt
== NULL
|| dst
.s_addr
!= sin
->sin_addr
.s_addr
) {
1706 if (ro
->ro_rt
!= NULL
) {
1710 sin
->sin_family
= AF_INET
;
1711 sin
->sin_len
= sizeof *sin
;
1712 sin
->sin_addr
= dst
;
1713 rtalloc_ign(ro
, RTF_PRCLONING
);
1716 if (ro
->ro_rt
== NULL
)
1719 return (ifatoia(ro
->ro_rt
->rt_ifa
));
1723 * Save incoming source route for use in replies,
1724 * to be picked up later by ip_srcroute if the receiver is interested.
1727 save_rte(u_char
*option
, struct in_addr dst
)
1731 olen
= option
[IPOPT_OLEN
];
1734 kprintf("save_rte: olen %d\n", olen
);
1736 if (olen
> sizeof(ip_srcrt
) - (1 + sizeof(dst
)))
1738 bcopy(option
, ip_srcrt
.srcopt
, olen
);
1739 ip_nhops
= (olen
- IPOPT_OFFSET
- 1) / sizeof(struct in_addr
);
1744 * Retrieve incoming source route for use in replies,
1745 * in the same form used by setsockopt.
1746 * The first hop is placed before the options, will be removed later.
1751 struct in_addr
*p
, *q
;
1756 m
= m_get(MB_DONTWAIT
, MT_HEADER
);
1760 #define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt))
1762 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
1763 m
->m_len
= ip_nhops
* sizeof(struct in_addr
) + sizeof(struct in_addr
) +
1767 kprintf("ip_srcroute: nhops %d mlen %d", ip_nhops
, m
->m_len
);
1771 * First save first hop for return route
1773 p
= &ip_srcrt
.route
[ip_nhops
- 1];
1774 *(mtod(m
, struct in_addr
*)) = *p
--;
1777 kprintf(" hops %x", ntohl(mtod(m
, struct in_addr
*)->s_addr
));
1781 * Copy option fields and padding (nop) to mbuf.
1783 ip_srcrt
.nop
= IPOPT_NOP
;
1784 ip_srcrt
.srcopt
[IPOPT_OFFSET
] = IPOPT_MINOFF
;
1785 memcpy(mtod(m
, caddr_t
) + sizeof(struct in_addr
), &ip_srcrt
.nop
,
1787 q
= (struct in_addr
*)(mtod(m
, caddr_t
) +
1788 sizeof(struct in_addr
) + OPTSIZ
);
1791 * Record return path as an IP source route,
1792 * reversing the path (pointers are now aligned).
1794 while (p
>= ip_srcrt
.route
) {
1797 kprintf(" %x", ntohl(q
->s_addr
));
1802 * Last hop goes to final destination.
1807 kprintf(" %x\n", ntohl(q
->s_addr
));
1813 * Strip out IP options.
1816 ip_stripoptions(struct mbuf
*m
)
1819 struct ip
*ip
= mtod(m
, struct ip
*);
1823 optlen
= (IP_VHL_HL(ip
->ip_vhl
) << 2) - sizeof(struct ip
);
1824 opts
= (caddr_t
)(ip
+ 1);
1825 datalen
= m
->m_len
- (sizeof(struct ip
) + optlen
);
1826 bcopy(opts
+ optlen
, opts
, datalen
);
1828 if (m
->m_flags
& M_PKTHDR
)
1829 m
->m_pkthdr
.len
-= optlen
;
1830 ip
->ip_vhl
= IP_MAKE_VHL(IPVERSION
, sizeof(struct ip
) >> 2);
1833 u_char inetctlerrmap
[PRC_NCMDS
] = {
1835 0, EMSGSIZE
, EHOSTDOWN
, EHOSTUNREACH
,
1836 EHOSTUNREACH
, EHOSTUNREACH
, ECONNREFUSED
, ECONNREFUSED
,
1837 EMSGSIZE
, EHOSTUNREACH
, 0, 0,
1839 ENOPROTOOPT
, ECONNREFUSED
1843 * Forward a packet. If some error occurs return the sender
1844 * an icmp packet. Note we can't always generate a meaningful
1845 * icmp message because icmp doesn't have a large enough repertoire
1846 * of codes and types.
1848 * If not forwarding, just drop the packet. This could be confusing
1849 * if ipforwarding was zero but some routing protocol was advancing
1850 * us as a gateway to somewhere. However, we must let the routing
1851 * protocol deal with that.
1853 * The using_srcrt parameter indicates whether the packet is being forwarded
1854 * via a source route.
1857 ip_forward(struct mbuf
*m
, boolean_t using_srcrt
, struct sockaddr_in
*next_hop
)
1859 struct ip
*ip
= mtod(m
, struct ip
*);
1860 struct sockaddr_in
*ipforward_rtaddr
;
1862 int error
, type
= 0, code
= 0, destmtu
= 0;
1865 struct in_addr pkt_dst
;
1866 struct route
*cache_rt
= &ipforward_rt
[mycpuid
];
1870 * Cache the destination address of the packet; this may be
1871 * changed by use of 'ipfw fwd'.
1873 pkt_dst
= (next_hop
!= NULL
) ? next_hop
->sin_addr
: ip
->ip_dst
;
1877 kprintf("forward: src %x dst %x ttl %x\n",
1878 ip
->ip_src
.s_addr
, pkt_dst
.s_addr
, ip
->ip_ttl
);
1881 if (m
->m_flags
& (M_BCAST
| M_MCAST
) || !in_canforward(pkt_dst
)) {
1882 ipstat
.ips_cantforward
++;
1886 if (!ipstealth
&& ip
->ip_ttl
<= IPTTLDEC
) {
1887 icmp_error(m
, ICMP_TIMXCEED
, ICMP_TIMXCEED_INTRANS
, dest
, 0);
1891 ipforward_rtaddr
= (struct sockaddr_in
*) &cache_rt
->ro_dst
;
1892 if (cache_rt
->ro_rt
== NULL
||
1893 ipforward_rtaddr
->sin_addr
.s_addr
!= pkt_dst
.s_addr
) {
1894 if (cache_rt
->ro_rt
!= NULL
) {
1895 RTFREE(cache_rt
->ro_rt
);
1896 cache_rt
->ro_rt
= NULL
;
1898 ipforward_rtaddr
->sin_family
= AF_INET
;
1899 ipforward_rtaddr
->sin_len
= sizeof(struct sockaddr_in
);
1900 ipforward_rtaddr
->sin_addr
= pkt_dst
;
1901 rtalloc_ign(cache_rt
, RTF_PRCLONING
);
1902 if (cache_rt
->ro_rt
== NULL
) {
1903 icmp_error(m
, ICMP_UNREACH
, ICMP_UNREACH_HOST
, dest
, 0);
1907 rt
= cache_rt
->ro_rt
;
1910 * Save the IP header and at most 8 bytes of the payload,
1911 * in case we need to generate an ICMP message to the src.
1913 * XXX this can be optimized a lot by saving the data in a local
1914 * buffer on the stack (72 bytes at most), and only allocating the
1915 * mbuf if really necessary. The vast majority of the packets
1916 * are forwarded without having to send an ICMP back (either
1917 * because unnecessary, or because rate limited), so we are
1918 * really we are wasting a lot of work here.
1920 * We don't use m_copy() because it might return a reference
1921 * to a shared cluster. Both this function and ip_output()
1922 * assume exclusive access to the IP header in `m', so any
1923 * data in a cluster may change before we reach icmp_error().
1925 MGETHDR(mcopy
, MB_DONTWAIT
, m
->m_type
);
1926 if (mcopy
!= NULL
&& !m_dup_pkthdr(mcopy
, m
, MB_DONTWAIT
)) {
1928 * It's probably ok if the pkthdr dup fails (because
1929 * the deep copy of the tag chain failed), but for now
1930 * be conservative and just discard the copy since
1931 * code below may some day want the tags.
1936 if (mcopy
!= NULL
) {
1937 mcopy
->m_len
= imin((IP_VHL_HL(ip
->ip_vhl
) << 2) + 8,
1939 mcopy
->m_pkthdr
.len
= mcopy
->m_len
;
1940 m_copydata(m
, 0, mcopy
->m_len
, mtod(mcopy
, caddr_t
));
1944 ip
->ip_ttl
-= IPTTLDEC
;
1947 * If forwarding packet using same interface that it came in on,
1948 * perhaps should send a redirect to sender to shortcut a hop.
1949 * Only send redirect if source is sending directly to us,
1950 * and if packet was not source routed (or has any options).
1951 * Also, don't send redirect if forwarding using a default route
1952 * or a route modified by a redirect.
1954 if (rt
->rt_ifp
== m
->m_pkthdr
.rcvif
&&
1955 !(rt
->rt_flags
& (RTF_DYNAMIC
| RTF_MODIFIED
)) &&
1956 satosin(rt_key(rt
))->sin_addr
.s_addr
!= INADDR_ANY
&&
1957 ipsendredirects
&& !using_srcrt
&& next_hop
== NULL
) {
1958 u_long src
= ntohl(ip
->ip_src
.s_addr
);
1959 struct in_ifaddr
*rt_ifa
= (struct in_ifaddr
*)rt
->rt_ifa
;
1961 if (rt_ifa
!= NULL
&&
1962 (src
& rt_ifa
->ia_subnetmask
) == rt_ifa
->ia_subnet
) {
1963 if (rt
->rt_flags
& RTF_GATEWAY
)
1964 dest
= satosin(rt
->rt_gateway
)->sin_addr
.s_addr
;
1966 dest
= pkt_dst
.s_addr
;
1968 * Router requirements says to only send
1971 type
= ICMP_REDIRECT
;
1972 code
= ICMP_REDIRECT_HOST
;
1975 kprintf("redirect (%d) to %x\n", code
, dest
);
1980 error
= ip_output(m
, NULL
, cache_rt
, IP_FORWARDING
, NULL
, NULL
);
1982 ipstat
.ips_forward
++;
1985 ipflow_create(cache_rt
, mcopy
);
1988 return; /* most common case */
1990 ipstat
.ips_redirectsent
++;
1993 ipstat
.ips_cantforward
++;
2000 * Send ICMP message.
2005 case 0: /* forwarded, but need redirect */
2006 /* type, code set above */
2009 case ENETUNREACH
: /* shouldn't happen, checked above */
2014 type
= ICMP_UNREACH
;
2015 code
= ICMP_UNREACH_HOST
;
2019 type
= ICMP_UNREACH
;
2020 code
= ICMP_UNREACH_NEEDFRAG
;
2023 * If the packet is routed over IPsec tunnel, tell the
2024 * originator the tunnel MTU.
2025 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
2028 if (cache_rt
->ro_rt
!= NULL
) {
2029 struct secpolicy
*sp
= NULL
;
2034 sp
= ipsec4_getpolicybyaddr(mcopy
,
2040 destmtu
= cache_rt
->ro_rt
->rt_ifp
->if_mtu
;
2042 /* count IPsec header size */
2043 ipsechdr
= ipsec4_hdrsiz(mcopy
,
2048 * find the correct route for outer IPv4
2049 * header, compute tunnel MTU.
2052 if (sp
->req
!= NULL
&& sp
->req
->sav
!= NULL
&&
2053 sp
->req
->sav
->sah
!= NULL
) {
2054 ro
= &sp
->req
->sav
->sah
->sa_route
;
2055 if (ro
->ro_rt
!= NULL
&&
2056 ro
->ro_rt
->rt_ifp
!= NULL
) {
2058 ro
->ro_rt
->rt_ifp
->if_mtu
;
2059 destmtu
-= ipsechdr
;
2068 * If the packet is routed over IPsec tunnel, tell the
2069 * originator the tunnel MTU.
2070 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
2073 if (cache_rt
->ro_rt
!= NULL
) {
2074 struct secpolicy
*sp
= NULL
;
2079 sp
= ipsec_getpolicybyaddr(mcopy
,
2085 destmtu
= cache_rt
->ro_rt
->rt_ifp
->if_mtu
;
2087 /* count IPsec header size */
2088 ipsechdr
= ipsec4_hdrsiz(mcopy
,
2093 * find the correct route for outer IPv4
2094 * header, compute tunnel MTU.
2097 if (sp
->req
!= NULL
&&
2098 sp
->req
->sav
!= NULL
&&
2099 sp
->req
->sav
->sah
!= NULL
) {
2100 ro
= &sp
->req
->sav
->sah
->sa_route
;
2101 if (ro
->ro_rt
!= NULL
&&
2102 ro
->ro_rt
->rt_ifp
!= NULL
) {
2104 ro
->ro_rt
->rt_ifp
->if_mtu
;
2105 destmtu
-= ipsechdr
;
2112 #else /* !IPSEC && !FAST_IPSEC */
2113 if (cache_rt
->ro_rt
!= NULL
)
2114 destmtu
= cache_rt
->ro_rt
->rt_ifp
->if_mtu
;
2116 ipstat
.ips_cantfrag
++;
2121 * A router should not generate ICMP_SOURCEQUENCH as
2122 * required in RFC1812 Requirements for IP Version 4 Routers.
2123 * Source quench could be a big problem under DoS attacks,
2124 * or if the underlying interface is rate-limited.
2125 * Those who need source quench packets may re-enable them
2126 * via the net.inet.ip.sendsourcequench sysctl.
2128 if (!ip_sendsourcequench
) {
2132 type
= ICMP_SOURCEQUENCH
;
2137 case EACCES
: /* ipfw denied packet */
2141 icmp_error(mcopy
, type
, code
, dest
, destmtu
);
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(),
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
*m
, ...) /* XXX must fixup manually */
2275 off
= __va_arg(ap
, int);
2276 proto
= __va_arg(ap
, int);
2279 if (rsvp_input_p
) { /* call the real one if loaded */
2280 rsvp_input_p(m
, off
, proto
);
2284 /* Can still get packets with rsvp_on = 0 if there is a local member
2285 * of the group to which the RSVP packet is addressed. But in this
2286 * case we want to throw the packet away.
2294 if (ip_rsvpd
!= NULL
) {
2295 rip_input(m
, off
, proto
);
2298 /* Drop the packet */