ip_input.c revision 1.258 1 /* $NetBSD: ip_input.c,v 1.258 2007/12/21 18:58:55 matt Exp $ */
2
3 /*
4 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5 * All rights reserved.
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 * 3. Neither the name of the project nor the names of its contributors
16 * may be used to endorse or promote products derived from this software
17 * without specific prior written permission.
18 *
19 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 * SUCH DAMAGE.
30 */
31
32 /*-
33 * Copyright (c) 1998 The NetBSD Foundation, Inc.
34 * All rights reserved.
35 *
36 * This code is derived from software contributed to The NetBSD Foundation
37 * by Public Access Networks Corporation ("Panix"). It was developed under
38 * contract to Panix by Eric Haszlakiewicz and Thor Lancelot Simon.
39 *
40 * Redistribution and use in source and binary forms, with or without
41 * modification, are permitted provided that the following conditions
42 * are met:
43 * 1. Redistributions of source code must retain the above copyright
44 * notice, this list of conditions and the following disclaimer.
45 * 2. Redistributions in binary form must reproduce the above copyright
46 * notice, this list of conditions and the following disclaimer in the
47 * documentation and/or other materials provided with the distribution.
48 * 3. All advertising materials mentioning features or use of this software
49 * must display the following acknowledgement:
50 * This product includes software developed by the NetBSD
51 * Foundation, Inc. and its contributors.
52 * 4. Neither the name of The NetBSD Foundation nor the names of its
53 * contributors may be used to endorse or promote products derived
54 * from this software without specific prior written permission.
55 *
56 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
57 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
58 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
59 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
60 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
61 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
62 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
63 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
64 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
65 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
66 * POSSIBILITY OF SUCH DAMAGE.
67 */
68
69 /*
70 * Copyright (c) 1982, 1986, 1988, 1993
71 * The Regents of the University of California. All rights reserved.
72 *
73 * Redistribution and use in source and binary forms, with or without
74 * modification, are permitted provided that the following conditions
75 * are met:
76 * 1. Redistributions of source code must retain the above copyright
77 * notice, this list of conditions and the following disclaimer.
78 * 2. Redistributions in binary form must reproduce the above copyright
79 * notice, this list of conditions and the following disclaimer in the
80 * documentation and/or other materials provided with the distribution.
81 * 3. Neither the name of the University nor the names of its contributors
82 * may be used to endorse or promote products derived from this software
83 * without specific prior written permission.
84 *
85 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
86 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
87 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
88 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
89 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
90 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
91 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
92 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
93 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
94 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
95 * SUCH DAMAGE.
96 *
97 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94
98 */
99
100 #include <sys/cdefs.h>
101 __KERNEL_RCSID(0, "$NetBSD: ip_input.c,v 1.258 2007/12/21 18:58:55 matt Exp $");
102
103 #include "opt_inet.h"
104 #include "opt_gateway.h"
105 #include "opt_pfil_hooks.h"
106 #include "opt_ipsec.h"
107 #include "opt_mrouting.h"
108 #include "opt_mbuftrace.h"
109 #include "opt_inet_csum.h"
110
111 #include <sys/param.h>
112 #include <sys/systm.h>
113 #include <sys/malloc.h>
114 #include <sys/mbuf.h>
115 #include <sys/domain.h>
116 #include <sys/protosw.h>
117 #include <sys/socket.h>
118 #include <sys/socketvar.h>
119 #include <sys/errno.h>
120 #include <sys/time.h>
121 #include <sys/kernel.h>
122 #include <sys/pool.h>
123 #include <sys/sysctl.h>
124 #include <sys/kauth.h>
125
126 #include <net/if.h>
127 #include <net/if_dl.h>
128 #include <net/route.h>
129 #include <net/pfil.h>
130
131 #include <netinet/in.h>
132 #include <netinet/in_systm.h>
133 #include <netinet/ip.h>
134 #include <netinet/in_pcb.h>
135 #include <netinet/in_proto.h>
136 #include <netinet/in_var.h>
137 #include <netinet/ip_var.h>
138 #include <netinet/ip_icmp.h>
139 /* just for gif_ttl */
140 #include <netinet/in_gif.h>
141 #include "gif.h"
142 #include <net/if_gre.h>
143 #include "gre.h"
144
145 #ifdef MROUTING
146 #include <netinet/ip_mroute.h>
147 #endif
148
149 #ifdef IPSEC
150 #include <netinet6/ipsec.h>
151 #include <netkey/key.h>
152 #endif
153 #ifdef FAST_IPSEC
154 #include <netipsec/ipsec.h>
155 #include <netipsec/key.h>
156 #endif /* FAST_IPSEC*/
157
158 #ifndef IPFORWARDING
159 #ifdef GATEWAY
160 #define IPFORWARDING 1 /* forward IP packets not for us */
161 #else /* GATEWAY */
162 #define IPFORWARDING 0 /* don't forward IP packets not for us */
163 #endif /* GATEWAY */
164 #endif /* IPFORWARDING */
165 #ifndef IPSENDREDIRECTS
166 #define IPSENDREDIRECTS 1
167 #endif
168 #ifndef IPFORWSRCRT
169 #define IPFORWSRCRT 1 /* forward source-routed packets */
170 #endif
171 #ifndef IPALLOWSRCRT
172 #define IPALLOWSRCRT 1 /* allow source-routed packets */
173 #endif
174 #ifndef IPMTUDISC
175 #define IPMTUDISC 1
176 #endif
177 #ifndef IPMTUDISCTIMEOUT
178 #define IPMTUDISCTIMEOUT (10 * 60) /* as per RFC 1191 */
179 #endif
180
181 /*
182 * Note: DIRECTED_BROADCAST is handled this way so that previous
183 * configuration using this option will Just Work.
184 */
185 #ifndef IPDIRECTEDBCAST
186 #ifdef DIRECTED_BROADCAST
187 #define IPDIRECTEDBCAST 1
188 #else
189 #define IPDIRECTEDBCAST 0
190 #endif /* DIRECTED_BROADCAST */
191 #endif /* IPDIRECTEDBCAST */
192 int ipforwarding = IPFORWARDING;
193 int ipsendredirects = IPSENDREDIRECTS;
194 int ip_defttl = IPDEFTTL;
195 int ip_forwsrcrt = IPFORWSRCRT;
196 int ip_directedbcast = IPDIRECTEDBCAST;
197 int ip_allowsrcrt = IPALLOWSRCRT;
198 int ip_mtudisc = IPMTUDISC;
199 int ip_mtudisc_timeout = IPMTUDISCTIMEOUT;
200 #ifdef DIAGNOSTIC
201 int ipprintfs = 0;
202 #endif
203
204 int ip_do_randomid = 0;
205
206 /*
207 * XXX - Setting ip_checkinterface mostly implements the receive side of
208 * the Strong ES model described in RFC 1122, but since the routing table
209 * and transmit implementation do not implement the Strong ES model,
210 * setting this to 1 results in an odd hybrid.
211 *
212 * XXX - ip_checkinterface currently must be disabled if you use ipnat
213 * to translate the destination address to another local interface.
214 *
215 * XXX - ip_checkinterface must be disabled if you add IP aliases
216 * to the loopback interface instead of the interface where the
217 * packets for those addresses are received.
218 */
219 int ip_checkinterface = 0;
220
221
222 struct rttimer_queue *ip_mtudisc_timeout_q = NULL;
223
224 int ipqmaxlen = IFQ_MAXLEN;
225 u_long in_ifaddrhash; /* size of hash table - 1 */
226 int in_ifaddrentries; /* total number of addrs */
227 struct in_ifaddrhead in_ifaddrhead;
228 struct in_ifaddrhashhead *in_ifaddrhashtbl;
229 u_long in_multihash; /* size of hash table - 1 */
230 int in_multientries; /* total number of addrs */
231 struct in_multihashhead *in_multihashtbl;
232 struct ifqueue ipintrq;
233 struct ipstat ipstat;
234 uint16_t ip_id;
235
236 #ifdef PFIL_HOOKS
237 struct pfil_head inet_pfil_hook;
238 #endif
239
240 /*
241 * Cached copy of nmbclusters. If nbclusters is different,
242 * recalculate IP parameters derived from nmbclusters.
243 */
244 static int ip_nmbclusters; /* copy of nmbclusters */
245 static void ip_nmbclusters_changed(void); /* recalc limits */
246
247 #define CHECK_NMBCLUSTER_PARAMS() \
248 do { \
249 if (__predict_false(ip_nmbclusters != nmbclusters)) \
250 ip_nmbclusters_changed(); \
251 } while (/*CONSTCOND*/0)
252
253 /* IP datagram reassembly queues (hashed) */
254 #define IPREASS_NHASH_LOG2 6
255 #define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2)
256 #define IPREASS_HMASK (IPREASS_NHASH - 1)
257 #define IPREASS_HASH(x,y) \
258 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK)
259 struct ipqhead ipq[IPREASS_NHASH];
260 int ipq_locked;
261 static int ip_nfragpackets; /* packets in reass queue */
262 static int ip_nfrags; /* total fragments in reass queues */
263
264 int ip_maxfragpackets = 200; /* limit on packets. XXX sysctl */
265 int ip_maxfrags; /* limit on fragments. XXX sysctl */
266
267
268 /*
269 * Additive-Increase/Multiplicative-Decrease (AIMD) strategy for
270 * IP reassembly queue buffer managment.
271 *
272 * We keep a count of total IP fragments (NB: not fragmented packets!)
273 * awaiting reassembly (ip_nfrags) and a limit (ip_maxfrags) on fragments.
274 * If ip_nfrags exceeds ip_maxfrags the limit, we drop half the
275 * total fragments in reassembly queues.This AIMD policy avoids
276 * repeatedly deleting single packets under heavy fragmentation load
277 * (e.g., from lossy NFS peers).
278 */
279 static u_int ip_reass_ttl_decr(u_int ticks);
280 static void ip_reass_drophalf(void);
281
282
283 static inline int ipq_lock_try(void);
284 static inline void ipq_unlock(void);
285
286 static inline int
287 ipq_lock_try(void)
288 {
289 int s;
290
291 /*
292 * Use splvm() -- we're blocking things that would cause
293 * mbuf allocation.
294 */
295 s = splvm();
296 if (ipq_locked) {
297 splx(s);
298 return (0);
299 }
300 ipq_locked = 1;
301 splx(s);
302 return (1);
303 }
304
305 static inline void
306 ipq_unlock(void)
307 {
308 int s;
309
310 s = splvm();
311 ipq_locked = 0;
312 splx(s);
313 }
314
315 #ifdef DIAGNOSTIC
316 #define IPQ_LOCK() \
317 do { \
318 if (ipq_lock_try() == 0) { \
319 printf("%s:%d: ipq already locked\n", __FILE__, __LINE__); \
320 panic("ipq_lock"); \
321 } \
322 } while (/*CONSTCOND*/ 0)
323 #define IPQ_LOCK_CHECK() \
324 do { \
325 if (ipq_locked == 0) { \
326 printf("%s:%d: ipq lock not held\n", __FILE__, __LINE__); \
327 panic("ipq lock check"); \
328 } \
329 } while (/*CONSTCOND*/ 0)
330 #else
331 #define IPQ_LOCK() (void) ipq_lock_try()
332 #define IPQ_LOCK_CHECK() /* nothing */
333 #endif
334
335 #define IPQ_UNLOCK() ipq_unlock()
336
337 POOL_INIT(inmulti_pool, sizeof(struct in_multi), 0, 0, 0, "inmltpl", NULL,
338 IPL_SOFTNET);
339 POOL_INIT(ipqent_pool, sizeof(struct ipqent), 0, 0, 0, "ipqepl", NULL,
340 IPL_VM);
341
342 #ifdef INET_CSUM_COUNTERS
343 #include <sys/device.h>
344
345 struct evcnt ip_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
346 NULL, "inet", "hwcsum bad");
347 struct evcnt ip_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
348 NULL, "inet", "hwcsum ok");
349 struct evcnt ip_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
350 NULL, "inet", "swcsum");
351
352 #define INET_CSUM_COUNTER_INCR(ev) (ev)->ev_count++
353
354 EVCNT_ATTACH_STATIC(ip_hwcsum_bad);
355 EVCNT_ATTACH_STATIC(ip_hwcsum_ok);
356 EVCNT_ATTACH_STATIC(ip_swcsum);
357
358 #else
359
360 #define INET_CSUM_COUNTER_INCR(ev) /* nothing */
361
362 #endif /* INET_CSUM_COUNTERS */
363
364 /*
365 * We need to save the IP options in case a protocol wants to respond
366 * to an incoming packet over the same route if the packet got here
367 * using IP source routing. This allows connection establishment and
368 * maintenance when the remote end is on a network that is not known
369 * to us.
370 */
371 int ip_nhops = 0;
372 static struct ip_srcrt {
373 struct in_addr dst; /* final destination */
374 char nop; /* one NOP to align */
375 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */
376 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)];
377 } ip_srcrt;
378
379 static void save_rte(u_char *, struct in_addr);
380
381 #ifdef MBUFTRACE
382 struct mowner ip_rx_mowner = MOWNER_INIT("internet", "rx");
383 struct mowner ip_tx_mowner = MOWNER_INIT("internet", "tx");
384 #endif
385
386 /*
387 * Compute IP limits derived from the value of nmbclusters.
388 */
389 static void
390 ip_nmbclusters_changed(void)
391 {
392 ip_maxfrags = nmbclusters / 4;
393 ip_nmbclusters = nmbclusters;
394 }
395
396 /*
397 * IP initialization: fill in IP protocol switch table.
398 * All protocols not implemented in kernel go to raw IP protocol handler.
399 */
400 void
401 ip_init(void)
402 {
403 const struct protosw *pr;
404 int i;
405
406 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
407 if (pr == 0)
408 panic("ip_init");
409 for (i = 0; i < IPPROTO_MAX; i++)
410 ip_protox[i] = pr - inetsw;
411 for (pr = inetdomain.dom_protosw;
412 pr < inetdomain.dom_protoswNPROTOSW; pr++)
413 if (pr->pr_domain->dom_family == PF_INET &&
414 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
415 ip_protox[pr->pr_protocol] = pr - inetsw;
416
417 for (i = 0; i < IPREASS_NHASH; i++)
418 LIST_INIT(&ipq[i]);
419
420 ip_id = time_second & 0xfffff;
421
422 ipintrq.ifq_maxlen = ipqmaxlen;
423 ip_nmbclusters_changed();
424
425 TAILQ_INIT(&in_ifaddrhead);
426 in_ifaddrhashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, M_IFADDR,
427 M_WAITOK, &in_ifaddrhash);
428 in_multihashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, M_IPMADDR,
429 M_WAITOK, &in_multihash);
430 ip_mtudisc_timeout_q = rt_timer_queue_create(ip_mtudisc_timeout);
431 #ifdef GATEWAY
432 ipflow_init(ip_hashsize);
433 #endif
434
435 #ifdef PFIL_HOOKS
436 /* Register our Packet Filter hook. */
437 inet_pfil_hook.ph_type = PFIL_TYPE_AF;
438 inet_pfil_hook.ph_af = AF_INET;
439 i = pfil_head_register(&inet_pfil_hook);
440 if (i != 0)
441 printf("ip_init: WARNING: unable to register pfil hook, "
442 "error %d\n", i);
443 #endif /* PFIL_HOOKS */
444
445 #ifdef MBUFTRACE
446 MOWNER_ATTACH(&ip_tx_mowner);
447 MOWNER_ATTACH(&ip_rx_mowner);
448 #endif /* MBUFTRACE */
449 }
450
451 struct sockaddr_in ipaddr = {
452 .sin_len = sizeof(ipaddr),
453 .sin_family = AF_INET,
454 };
455 struct route ipforward_rt;
456
457 /*
458 * IP software interrupt routine
459 */
460 void
461 ipintr(void)
462 {
463 int s;
464 struct mbuf *m;
465
466 while (!IF_IS_EMPTY(&ipintrq)) {
467 s = splnet();
468 IF_DEQUEUE(&ipintrq, m);
469 splx(s);
470 if (m == 0)
471 return;
472 ip_input(m);
473 }
474 }
475
476 /*
477 * Ip input routine. Checksum and byte swap header. If fragmented
478 * try to reassemble. Process options. Pass to next level.
479 */
480 void
481 ip_input(struct mbuf *m)
482 {
483 struct ip *ip = NULL;
484 struct ipq *fp;
485 struct in_ifaddr *ia;
486 struct ifaddr *ifa;
487 struct ipqent *ipqe;
488 int hlen = 0, mff, len;
489 int downmatch;
490 int checkif;
491 int srcrt = 0;
492 int s;
493 u_int hash;
494 #ifdef FAST_IPSEC
495 struct m_tag *mtag;
496 struct tdb_ident *tdbi;
497 struct secpolicy *sp;
498 int error;
499 #endif /* FAST_IPSEC */
500
501 MCLAIM(m, &ip_rx_mowner);
502 #ifdef DIAGNOSTIC
503 if ((m->m_flags & M_PKTHDR) == 0)
504 panic("ipintr no HDR");
505 #endif
506
507 /*
508 * If no IP addresses have been set yet but the interfaces
509 * are receiving, can't do anything with incoming packets yet.
510 */
511 if (TAILQ_FIRST(&in_ifaddrhead) == 0)
512 goto bad;
513 ipstat.ips_total++;
514 /*
515 * If the IP header is not aligned, slurp it up into a new
516 * mbuf with space for link headers, in the event we forward
517 * it. Otherwise, if it is aligned, make sure the entire
518 * base IP header is in the first mbuf of the chain.
519 */
520 if (IP_HDR_ALIGNED_P(mtod(m, void *)) == 0) {
521 if ((m = m_copyup(m, sizeof(struct ip),
522 (max_linkhdr + 3) & ~3)) == NULL) {
523 /* XXXJRT new stat, please */
524 ipstat.ips_toosmall++;
525 return;
526 }
527 } else if (__predict_false(m->m_len < sizeof (struct ip))) {
528 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
529 ipstat.ips_toosmall++;
530 return;
531 }
532 }
533 ip = mtod(m, struct ip *);
534 if (ip->ip_v != IPVERSION) {
535 ipstat.ips_badvers++;
536 goto bad;
537 }
538 hlen = ip->ip_hl << 2;
539 if (hlen < sizeof(struct ip)) { /* minimum header length */
540 ipstat.ips_badhlen++;
541 goto bad;
542 }
543 if (hlen > m->m_len) {
544 if ((m = m_pullup(m, hlen)) == 0) {
545 ipstat.ips_badhlen++;
546 return;
547 }
548 ip = mtod(m, struct ip *);
549 }
550
551 /*
552 * RFC1122: packets with a multicast source address are
553 * not allowed.
554 */
555 if (IN_MULTICAST(ip->ip_src.s_addr)) {
556 ipstat.ips_badaddr++;
557 goto bad;
558 }
559
560 /* 127/8 must not appear on wire - RFC1122 */
561 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
562 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
563 if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
564 ipstat.ips_badaddr++;
565 goto bad;
566 }
567 }
568
569 switch (m->m_pkthdr.csum_flags &
570 ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_IPv4) |
571 M_CSUM_IPv4_BAD)) {
572 case M_CSUM_IPv4|M_CSUM_IPv4_BAD:
573 INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad);
574 goto badcsum;
575
576 case M_CSUM_IPv4:
577 /* Checksum was okay. */
578 INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok);
579 break;
580
581 default:
582 /*
583 * Must compute it ourselves. Maybe skip checksum on
584 * loopback interfaces.
585 */
586 if (__predict_true(!(m->m_pkthdr.rcvif->if_flags &
587 IFF_LOOPBACK) || ip_do_loopback_cksum)) {
588 INET_CSUM_COUNTER_INCR(&ip_swcsum);
589 if (in_cksum(m, hlen) != 0)
590 goto badcsum;
591 }
592 break;
593 }
594
595 /* Retrieve the packet length. */
596 len = ntohs(ip->ip_len);
597
598 /*
599 * Check for additional length bogosity
600 */
601 if (len < hlen) {
602 ipstat.ips_badlen++;
603 goto bad;
604 }
605
606 /*
607 * Check that the amount of data in the buffers
608 * is as at least much as the IP header would have us expect.
609 * Trim mbufs if longer than we expect.
610 * Drop packet if shorter than we expect.
611 */
612 if (m->m_pkthdr.len < len) {
613 ipstat.ips_tooshort++;
614 goto bad;
615 }
616 if (m->m_pkthdr.len > len) {
617 if (m->m_len == m->m_pkthdr.len) {
618 m->m_len = len;
619 m->m_pkthdr.len = len;
620 } else
621 m_adj(m, len - m->m_pkthdr.len);
622 }
623
624 #if defined(IPSEC)
625 /* ipflow (IP fast forwarding) is not compatible with IPsec. */
626 m->m_flags &= ~M_CANFASTFWD;
627 #else
628 /*
629 * Assume that we can create a fast-forward IP flow entry
630 * based on this packet.
631 */
632 m->m_flags |= M_CANFASTFWD;
633 #endif
634
635 #ifdef PFIL_HOOKS
636 /*
637 * Run through list of hooks for input packets. If there are any
638 * filters which require that additional packets in the flow are
639 * not fast-forwarded, they must clear the M_CANFASTFWD flag.
640 * Note that filters must _never_ set this flag, as another filter
641 * in the list may have previously cleared it.
642 */
643 /*
644 * let ipfilter look at packet on the wire,
645 * not the decapsulated packet.
646 */
647 #ifdef IPSEC
648 if (!ipsec_getnhist(m))
649 #elif defined(FAST_IPSEC)
650 if (!ipsec_indone(m))
651 #else
652 if (1)
653 #endif
654 {
655 struct in_addr odst;
656
657 odst = ip->ip_dst;
658 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif,
659 PFIL_IN) != 0)
660 return;
661 if (m == NULL)
662 return;
663 ip = mtod(m, struct ip *);
664 hlen = ip->ip_hl << 2;
665 /*
666 * XXX The setting of "srcrt" here is to prevent ip_forward()
667 * from generating ICMP redirects for packets that have
668 * been redirected by a hook back out on to the same LAN that
669 * they came from and is not an indication that the packet
670 * is being inffluenced by source routing options. This
671 * allows things like
672 * "rdr tlp0 0/0 port 80 -> 1.1.1.200 3128 tcp"
673 * where tlp0 is both on the 1.1.1.0/24 network and is the
674 * default route for hosts on 1.1.1.0/24. Of course this
675 * also requires a "map tlp0 ..." to complete the story.
676 * One might argue whether or not this kind of network config.
677 * should be supported in this manner...
678 */
679 srcrt = (odst.s_addr != ip->ip_dst.s_addr);
680 }
681 #endif /* PFIL_HOOKS */
682
683 #ifdef ALTQ
684 /* XXX Temporary until ALTQ is changed to use a pfil hook */
685 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) {
686 /* packet dropped by traffic conditioner */
687 return;
688 }
689 #endif
690
691 /*
692 * Process options and, if not destined for us,
693 * ship it on. ip_dooptions returns 1 when an
694 * error was detected (causing an icmp message
695 * to be sent and the original packet to be freed).
696 */
697 ip_nhops = 0; /* for source routed packets */
698 if (hlen > sizeof (struct ip) && ip_dooptions(m))
699 return;
700
701 /*
702 * Enable a consistency check between the destination address
703 * and the arrival interface for a unicast packet (the RFC 1122
704 * strong ES model) if IP forwarding is disabled and the packet
705 * is not locally generated.
706 *
707 * XXX - Checking also should be disabled if the destination
708 * address is ipnat'ed to a different interface.
709 *
710 * XXX - Checking is incompatible with IP aliases added
711 * to the loopback interface instead of the interface where
712 * the packets are received.
713 *
714 * XXX - We need to add a per ifaddr flag for this so that
715 * we get finer grain control.
716 */
717 checkif = ip_checkinterface && (ipforwarding == 0) &&
718 (m->m_pkthdr.rcvif != NULL) &&
719 ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0);
720
721 /*
722 * Check our list of addresses, to see if the packet is for us.
723 *
724 * Traditional 4.4BSD did not consult IFF_UP at all.
725 * The behavior here is to treat addresses on !IFF_UP interface
726 * as not mine.
727 */
728 downmatch = 0;
729 LIST_FOREACH(ia, &IN_IFADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
730 if (in_hosteq(ia->ia_addr.sin_addr, ip->ip_dst)) {
731 if (checkif && ia->ia_ifp != m->m_pkthdr.rcvif)
732 continue;
733 if ((ia->ia_ifp->if_flags & IFF_UP) != 0)
734 break;
735 else
736 downmatch++;
737 }
738 }
739 if (ia != NULL)
740 goto ours;
741 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
742 IFADDR_FOREACH(ifa, m->m_pkthdr.rcvif) {
743 if (ifa->ifa_addr->sa_family != AF_INET)
744 continue;
745 ia = ifatoia(ifa);
746 if (in_hosteq(ip->ip_dst, ia->ia_broadaddr.sin_addr) ||
747 in_hosteq(ip->ip_dst, ia->ia_netbroadcast) ||
748 /*
749 * Look for all-0's host part (old broadcast addr),
750 * either for subnet or net.
751 */
752 ip->ip_dst.s_addr == ia->ia_subnet ||
753 ip->ip_dst.s_addr == ia->ia_net)
754 goto ours;
755 /*
756 * An interface with IP address zero accepts
757 * all packets that arrive on that interface.
758 */
759 if (in_nullhost(ia->ia_addr.sin_addr))
760 goto ours;
761 }
762 }
763 if (IN_MULTICAST(ip->ip_dst.s_addr)) {
764 struct in_multi *inm;
765 #ifdef MROUTING
766 extern struct socket *ip_mrouter;
767
768 if (ip_mrouter) {
769 /*
770 * If we are acting as a multicast router, all
771 * incoming multicast packets are passed to the
772 * kernel-level multicast forwarding function.
773 * The packet is returned (relatively) intact; if
774 * ip_mforward() returns a non-zero value, the packet
775 * must be discarded, else it may be accepted below.
776 *
777 * (The IP ident field is put in the same byte order
778 * as expected when ip_mforward() is called from
779 * ip_output().)
780 */
781 if (ip_mforward(m, m->m_pkthdr.rcvif) != 0) {
782 ipstat.ips_cantforward++;
783 m_freem(m);
784 return;
785 }
786
787 /*
788 * The process-level routing demon needs to receive
789 * all multicast IGMP packets, whether or not this
790 * host belongs to their destination groups.
791 */
792 if (ip->ip_p == IPPROTO_IGMP)
793 goto ours;
794 ipstat.ips_forward++;
795 }
796 #endif
797 /*
798 * See if we belong to the destination multicast group on the
799 * arrival interface.
800 */
801 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
802 if (inm == NULL) {
803 ipstat.ips_cantforward++;
804 m_freem(m);
805 return;
806 }
807 goto ours;
808 }
809 if (ip->ip_dst.s_addr == INADDR_BROADCAST ||
810 in_nullhost(ip->ip_dst))
811 goto ours;
812
813 /*
814 * Not for us; forward if possible and desirable.
815 */
816 if (ipforwarding == 0) {
817 ipstat.ips_cantforward++;
818 m_freem(m);
819 } else {
820 /*
821 * If ip_dst matched any of my address on !IFF_UP interface,
822 * and there's no IFF_UP interface that matches ip_dst,
823 * send icmp unreach. Forwarding it will result in in-kernel
824 * forwarding loop till TTL goes to 0.
825 */
826 if (downmatch) {
827 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
828 ipstat.ips_cantforward++;
829 return;
830 }
831 #ifdef IPSEC
832 if (ipsec4_in_reject(m, NULL)) {
833 ipsecstat.in_polvio++;
834 goto bad;
835 }
836 #endif
837 #ifdef FAST_IPSEC
838 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
839 s = splsoftnet();
840 if (mtag != NULL) {
841 tdbi = (struct tdb_ident *)(mtag + 1);
842 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
843 } else {
844 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
845 IP_FORWARDING, &error);
846 }
847 if (sp == NULL) { /* NB: can happen if error */
848 splx(s);
849 /*XXX error stat???*/
850 DPRINTF(("ip_input: no SP for forwarding\n")); /*XXX*/
851 goto bad;
852 }
853
854 /*
855 * Check security policy against packet attributes.
856 */
857 error = ipsec_in_reject(sp, m);
858 KEY_FREESP(&sp);
859 splx(s);
860 if (error) {
861 ipstat.ips_cantforward++;
862 goto bad;
863 }
864
865 /*
866 * Peek at the outbound SP for this packet to determine if
867 * it's a Fast Forward candidate.
868 */
869 mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL);
870 if (mtag != NULL)
871 m->m_flags &= ~M_CANFASTFWD;
872 else {
873 s = splsoftnet();
874 sp = ipsec4_checkpolicy(m, IPSEC_DIR_OUTBOUND,
875 (IP_FORWARDING |
876 (ip_directedbcast ? IP_ALLOWBROADCAST : 0)),
877 &error, NULL);
878 if (sp != NULL) {
879 m->m_flags &= ~M_CANFASTFWD;
880 KEY_FREESP(&sp);
881 }
882 splx(s);
883 }
884 #endif /* FAST_IPSEC */
885
886 ip_forward(m, srcrt);
887 }
888 return;
889
890 ours:
891 /*
892 * If offset or IP_MF are set, must reassemble.
893 * Otherwise, nothing need be done.
894 * (We could look in the reassembly queue to see
895 * if the packet was previously fragmented,
896 * but it's not worth the time; just let them time out.)
897 */
898 if (ip->ip_off & ~htons(IP_DF|IP_RF)) {
899 uint16_t off;
900 /*
901 * Prevent TCP blind data attacks by not allowing non-initial
902 * fragments to start at less than 68 bytes (minimal fragment
903 * size).
904 */
905 off = htons(ip->ip_off) & ~(IP_DF|IP_EF|IP_MF);
906 if (off > 0 && off + hlen < IP_MINFRAGSIZE - 1) {
907 ipstat.ips_badfrags++;
908 goto bad;
909 }
910 /*
911 * Look for queue of fragments
912 * of this datagram.
913 */
914 IPQ_LOCK();
915 hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
916 LIST_FOREACH(fp, &ipq[hash], ipq_q) {
917 if (ip->ip_id == fp->ipq_id &&
918 in_hosteq(ip->ip_src, fp->ipq_src) &&
919 in_hosteq(ip->ip_dst, fp->ipq_dst) &&
920 ip->ip_p == fp->ipq_p)
921 goto found;
922
923 }
924 fp = 0;
925 found:
926
927 /*
928 * Adjust ip_len to not reflect header,
929 * set ipqe_mff if more fragments are expected,
930 * convert offset of this to bytes.
931 */
932 ip->ip_len = htons(ntohs(ip->ip_len) - hlen);
933 mff = (ip->ip_off & htons(IP_MF)) != 0;
934 if (mff) {
935 /*
936 * Make sure that fragments have a data length
937 * that's a non-zero multiple of 8 bytes.
938 */
939 if (ntohs(ip->ip_len) == 0 ||
940 (ntohs(ip->ip_len) & 0x7) != 0) {
941 ipstat.ips_badfrags++;
942 IPQ_UNLOCK();
943 goto bad;
944 }
945 }
946 ip->ip_off = htons((ntohs(ip->ip_off) & IP_OFFMASK) << 3);
947
948 /*
949 * If datagram marked as having more fragments
950 * or if this is not the first fragment,
951 * attempt reassembly; if it succeeds, proceed.
952 */
953 if (mff || ip->ip_off != htons(0)) {
954 ipstat.ips_fragments++;
955 s = splvm();
956 ipqe = pool_get(&ipqent_pool, PR_NOWAIT);
957 splx(s);
958 if (ipqe == NULL) {
959 ipstat.ips_rcvmemdrop++;
960 IPQ_UNLOCK();
961 goto bad;
962 }
963 ipqe->ipqe_mff = mff;
964 ipqe->ipqe_m = m;
965 ipqe->ipqe_ip = ip;
966 m = ip_reass(ipqe, fp, &ipq[hash]);
967 if (m == 0) {
968 IPQ_UNLOCK();
969 return;
970 }
971 ipstat.ips_reassembled++;
972 ip = mtod(m, struct ip *);
973 hlen = ip->ip_hl << 2;
974 ip->ip_len = htons(ntohs(ip->ip_len) + hlen);
975 } else
976 if (fp)
977 ip_freef(fp);
978 IPQ_UNLOCK();
979 }
980
981 #if defined(IPSEC)
982 /*
983 * enforce IPsec policy checking if we are seeing last header.
984 * note that we do not visit this with protocols with pcb layer
985 * code - like udp/tcp/raw ip.
986 */
987 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0 &&
988 ipsec4_in_reject(m, NULL)) {
989 ipsecstat.in_polvio++;
990 goto bad;
991 }
992 #endif
993 #ifdef FAST_IPSEC
994 /*
995 * enforce IPsec policy checking if we are seeing last header.
996 * note that we do not visit this with protocols with pcb layer
997 * code - like udp/tcp/raw ip.
998 */
999 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) {
1000 /*
1001 * Check if the packet has already had IPsec processing
1002 * done. If so, then just pass it along. This tag gets
1003 * set during AH, ESP, etc. input handling, before the
1004 * packet is returned to the ip input queue for delivery.
1005 */
1006 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
1007 s = splsoftnet();
1008 if (mtag != NULL) {
1009 tdbi = (struct tdb_ident *)(mtag + 1);
1010 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
1011 } else {
1012 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
1013 IP_FORWARDING, &error);
1014 }
1015 if (sp != NULL) {
1016 /*
1017 * Check security policy against packet attributes.
1018 */
1019 error = ipsec_in_reject(sp, m);
1020 KEY_FREESP(&sp);
1021 } else {
1022 /* XXX error stat??? */
1023 error = EINVAL;
1024 DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/
1025 }
1026 splx(s);
1027 if (error)
1028 goto bad;
1029 }
1030 #endif /* FAST_IPSEC */
1031
1032 /*
1033 * Switch out to protocol's input routine.
1034 */
1035 #if IFA_STATS
1036 if (ia && ip)
1037 ia->ia_ifa.ifa_data.ifad_inbytes += ntohs(ip->ip_len);
1038 #endif
1039 ipstat.ips_delivered++;
1040 {
1041 int off = hlen, nh = ip->ip_p;
1042
1043 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh);
1044 return;
1045 }
1046 bad:
1047 m_freem(m);
1048 return;
1049
1050 badcsum:
1051 ipstat.ips_badsum++;
1052 m_freem(m);
1053 }
1054
1055 /*
1056 * Take incoming datagram fragment and try to
1057 * reassemble it into whole datagram. If a chain for
1058 * reassembly of this datagram already exists, then it
1059 * is given as fp; otherwise have to make a chain.
1060 */
1061 struct mbuf *
1062 ip_reass(struct ipqent *ipqe, struct ipq *fp, struct ipqhead *ipqhead)
1063 {
1064 struct mbuf *m = ipqe->ipqe_m;
1065 struct ipqent *nq, *p, *q;
1066 struct ip *ip;
1067 struct mbuf *t;
1068 int hlen = ipqe->ipqe_ip->ip_hl << 2;
1069 int i, next, s;
1070
1071 IPQ_LOCK_CHECK();
1072
1073 /*
1074 * Presence of header sizes in mbufs
1075 * would confuse code below.
1076 */
1077 m->m_data += hlen;
1078 m->m_len -= hlen;
1079
1080 #ifdef notyet
1081 /* make sure fragment limit is up-to-date */
1082 CHECK_NMBCLUSTER_PARAMS();
1083
1084 /* If we have too many fragments, drop the older half. */
1085 if (ip_nfrags >= ip_maxfrags)
1086 ip_reass_drophalf(void);
1087 #endif
1088
1089 /*
1090 * We are about to add a fragment; increment frag count.
1091 */
1092 ip_nfrags++;
1093
1094 /*
1095 * If first fragment to arrive, create a reassembly queue.
1096 */
1097 if (fp == 0) {
1098 /*
1099 * Enforce upper bound on number of fragmented packets
1100 * for which we attempt reassembly;
1101 * If maxfrag is 0, never accept fragments.
1102 * If maxfrag is -1, accept all fragments without limitation.
1103 */
1104 if (ip_maxfragpackets < 0)
1105 ;
1106 else if (ip_nfragpackets >= ip_maxfragpackets)
1107 goto dropfrag;
1108 ip_nfragpackets++;
1109 MALLOC(fp, struct ipq *, sizeof (struct ipq),
1110 M_FTABLE, M_NOWAIT);
1111 if (fp == NULL)
1112 goto dropfrag;
1113 LIST_INSERT_HEAD(ipqhead, fp, ipq_q);
1114 fp->ipq_nfrags = 1;
1115 fp->ipq_ttl = IPFRAGTTL;
1116 fp->ipq_p = ipqe->ipqe_ip->ip_p;
1117 fp->ipq_id = ipqe->ipqe_ip->ip_id;
1118 TAILQ_INIT(&fp->ipq_fragq);
1119 fp->ipq_src = ipqe->ipqe_ip->ip_src;
1120 fp->ipq_dst = ipqe->ipqe_ip->ip_dst;
1121 p = NULL;
1122 goto insert;
1123 } else {
1124 fp->ipq_nfrags++;
1125 }
1126
1127 /*
1128 * Find a segment which begins after this one does.
1129 */
1130 for (p = NULL, q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL;
1131 p = q, q = TAILQ_NEXT(q, ipqe_q))
1132 if (ntohs(q->ipqe_ip->ip_off) > ntohs(ipqe->ipqe_ip->ip_off))
1133 break;
1134
1135 /*
1136 * If there is a preceding segment, it may provide some of
1137 * our data already. If so, drop the data from the incoming
1138 * segment. If it provides all of our data, drop us.
1139 */
1140 if (p != NULL) {
1141 i = ntohs(p->ipqe_ip->ip_off) + ntohs(p->ipqe_ip->ip_len) -
1142 ntohs(ipqe->ipqe_ip->ip_off);
1143 if (i > 0) {
1144 if (i >= ntohs(ipqe->ipqe_ip->ip_len))
1145 goto dropfrag;
1146 m_adj(ipqe->ipqe_m, i);
1147 ipqe->ipqe_ip->ip_off =
1148 htons(ntohs(ipqe->ipqe_ip->ip_off) + i);
1149 ipqe->ipqe_ip->ip_len =
1150 htons(ntohs(ipqe->ipqe_ip->ip_len) - i);
1151 }
1152 }
1153
1154 /*
1155 * While we overlap succeeding segments trim them or,
1156 * if they are completely covered, dequeue them.
1157 */
1158 for (; q != NULL &&
1159 ntohs(ipqe->ipqe_ip->ip_off) + ntohs(ipqe->ipqe_ip->ip_len) >
1160 ntohs(q->ipqe_ip->ip_off); q = nq) {
1161 i = (ntohs(ipqe->ipqe_ip->ip_off) +
1162 ntohs(ipqe->ipqe_ip->ip_len)) - ntohs(q->ipqe_ip->ip_off);
1163 if (i < ntohs(q->ipqe_ip->ip_len)) {
1164 q->ipqe_ip->ip_len =
1165 htons(ntohs(q->ipqe_ip->ip_len) - i);
1166 q->ipqe_ip->ip_off =
1167 htons(ntohs(q->ipqe_ip->ip_off) + i);
1168 m_adj(q->ipqe_m, i);
1169 break;
1170 }
1171 nq = TAILQ_NEXT(q, ipqe_q);
1172 m_freem(q->ipqe_m);
1173 TAILQ_REMOVE(&fp->ipq_fragq, q, ipqe_q);
1174 s = splvm();
1175 pool_put(&ipqent_pool, q);
1176 splx(s);
1177 fp->ipq_nfrags--;
1178 ip_nfrags--;
1179 }
1180
1181 insert:
1182 /*
1183 * Stick new segment in its place;
1184 * check for complete reassembly.
1185 */
1186 if (p == NULL) {
1187 TAILQ_INSERT_HEAD(&fp->ipq_fragq, ipqe, ipqe_q);
1188 } else {
1189 TAILQ_INSERT_AFTER(&fp->ipq_fragq, p, ipqe, ipqe_q);
1190 }
1191 next = 0;
1192 for (p = NULL, q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL;
1193 p = q, q = TAILQ_NEXT(q, ipqe_q)) {
1194 if (ntohs(q->ipqe_ip->ip_off) != next)
1195 return (0);
1196 next += ntohs(q->ipqe_ip->ip_len);
1197 }
1198 if (p->ipqe_mff)
1199 return (0);
1200
1201 /*
1202 * Reassembly is complete. Check for a bogus message size and
1203 * concatenate fragments.
1204 */
1205 q = TAILQ_FIRST(&fp->ipq_fragq);
1206 ip = q->ipqe_ip;
1207 if ((next + (ip->ip_hl << 2)) > IP_MAXPACKET) {
1208 ipstat.ips_toolong++;
1209 ip_freef(fp);
1210 return (0);
1211 }
1212 m = q->ipqe_m;
1213 t = m->m_next;
1214 m->m_next = 0;
1215 m_cat(m, t);
1216 nq = TAILQ_NEXT(q, ipqe_q);
1217 s = splvm();
1218 pool_put(&ipqent_pool, q);
1219 splx(s);
1220 for (q = nq; q != NULL; q = nq) {
1221 t = q->ipqe_m;
1222 nq = TAILQ_NEXT(q, ipqe_q);
1223 s = splvm();
1224 pool_put(&ipqent_pool, q);
1225 splx(s);
1226 m_cat(m, t);
1227 }
1228 ip_nfrags -= fp->ipq_nfrags;
1229
1230 /*
1231 * Create header for new ip packet by
1232 * modifying header of first packet;
1233 * dequeue and discard fragment reassembly header.
1234 * Make header visible.
1235 */
1236 ip->ip_len = htons(next);
1237 ip->ip_src = fp->ipq_src;
1238 ip->ip_dst = fp->ipq_dst;
1239 LIST_REMOVE(fp, ipq_q);
1240 FREE(fp, M_FTABLE);
1241 ip_nfragpackets--;
1242 m->m_len += (ip->ip_hl << 2);
1243 m->m_data -= (ip->ip_hl << 2);
1244 /* some debugging cruft by sklower, below, will go away soon */
1245 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */
1246 int plen = 0;
1247 for (t = m; t; t = t->m_next)
1248 plen += t->m_len;
1249 m->m_pkthdr.len = plen;
1250 m->m_pkthdr.csum_flags = 0;
1251 }
1252 return (m);
1253
1254 dropfrag:
1255 if (fp != 0)
1256 fp->ipq_nfrags--;
1257 ip_nfrags--;
1258 ipstat.ips_fragdropped++;
1259 m_freem(m);
1260 s = splvm();
1261 pool_put(&ipqent_pool, ipqe);
1262 splx(s);
1263 return (0);
1264 }
1265
1266 /*
1267 * Free a fragment reassembly header and all
1268 * associated datagrams.
1269 */
1270 void
1271 ip_freef(struct ipq *fp)
1272 {
1273 struct ipqent *q, *p;
1274 u_int nfrags = 0;
1275 int s;
1276
1277 IPQ_LOCK_CHECK();
1278
1279 for (q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL; q = p) {
1280 p = TAILQ_NEXT(q, ipqe_q);
1281 m_freem(q->ipqe_m);
1282 nfrags++;
1283 TAILQ_REMOVE(&fp->ipq_fragq, q, ipqe_q);
1284 s = splvm();
1285 pool_put(&ipqent_pool, q);
1286 splx(s);
1287 }
1288
1289 if (nfrags != fp->ipq_nfrags)
1290 printf("ip_freef: nfrags %d != %d\n", fp->ipq_nfrags, nfrags);
1291 ip_nfrags -= nfrags;
1292 LIST_REMOVE(fp, ipq_q);
1293 FREE(fp, M_FTABLE);
1294 ip_nfragpackets--;
1295 }
1296
1297 /*
1298 * IP reassembly TTL machinery for multiplicative drop.
1299 */
1300 static u_int fragttl_histo[(IPFRAGTTL+1)];
1301
1302
1303 /*
1304 * Decrement TTL of all reasembly queue entries by `ticks'.
1305 * Count number of distinct fragments (as opposed to partial, fragmented
1306 * datagrams) in the reassembly queue. While we traverse the entire
1307 * reassembly queue, compute and return the median TTL over all fragments.
1308 */
1309 static u_int
1310 ip_reass_ttl_decr(u_int ticks)
1311 {
1312 u_int nfrags, median, dropfraction, keepfraction;
1313 struct ipq *fp, *nfp;
1314 int i;
1315
1316 nfrags = 0;
1317 memset(fragttl_histo, 0, sizeof fragttl_histo);
1318
1319 for (i = 0; i < IPREASS_NHASH; i++) {
1320 for (fp = LIST_FIRST(&ipq[i]); fp != NULL; fp = nfp) {
1321 fp->ipq_ttl = ((fp->ipq_ttl <= ticks) ?
1322 0 : fp->ipq_ttl - ticks);
1323 nfp = LIST_NEXT(fp, ipq_q);
1324 if (fp->ipq_ttl == 0) {
1325 ipstat.ips_fragtimeout++;
1326 ip_freef(fp);
1327 } else {
1328 nfrags += fp->ipq_nfrags;
1329 fragttl_histo[fp->ipq_ttl] += fp->ipq_nfrags;
1330 }
1331 }
1332 }
1333
1334 KASSERT(ip_nfrags == nfrags);
1335
1336 /* Find median (or other drop fraction) in histogram. */
1337 dropfraction = (ip_nfrags / 2);
1338 keepfraction = ip_nfrags - dropfraction;
1339 for (i = IPFRAGTTL, median = 0; i >= 0; i--) {
1340 median += fragttl_histo[i];
1341 if (median >= keepfraction)
1342 break;
1343 }
1344
1345 /* Return TTL of median (or other fraction). */
1346 return (u_int)i;
1347 }
1348
1349 void
1350 ip_reass_drophalf(void)
1351 {
1352
1353 u_int median_ticks;
1354 /*
1355 * Compute median TTL of all fragments, and count frags
1356 * with that TTL or lower (roughly half of all fragments).
1357 */
1358 median_ticks = ip_reass_ttl_decr(0);
1359
1360 /* Drop half. */
1361 median_ticks = ip_reass_ttl_decr(median_ticks);
1362
1363 }
1364
1365 /*
1366 * IP timer processing;
1367 * if a timer expires on a reassembly
1368 * queue, discard it.
1369 */
1370 void
1371 ip_slowtimo(void)
1372 {
1373 static u_int dropscanidx = 0;
1374 u_int i;
1375 u_int median_ttl;
1376 int s = splsoftnet();
1377
1378 IPQ_LOCK();
1379
1380 /* Age TTL of all fragments by 1 tick .*/
1381 median_ttl = ip_reass_ttl_decr(1);
1382
1383 /* make sure fragment limit is up-to-date */
1384 CHECK_NMBCLUSTER_PARAMS();
1385
1386 /* If we have too many fragments, drop the older half. */
1387 if (ip_nfrags > ip_maxfrags)
1388 ip_reass_ttl_decr(median_ttl);
1389
1390 /*
1391 * If we are over the maximum number of fragmented packets
1392 * (due to the limit being lowered), drain off
1393 * enough to get down to the new limit. Start draining
1394 * from the reassembly hashqueue most recently drained.
1395 */
1396 if (ip_maxfragpackets < 0)
1397 ;
1398 else {
1399 int wrapped = 0;
1400
1401 i = dropscanidx;
1402 while (ip_nfragpackets > ip_maxfragpackets && wrapped == 0) {
1403 while (LIST_FIRST(&ipq[i]) != NULL)
1404 ip_freef(LIST_FIRST(&ipq[i]));
1405 if (++i >= IPREASS_NHASH) {
1406 i = 0;
1407 }
1408 /*
1409 * Dont scan forever even if fragment counters are
1410 * wrong: stop after scanning entire reassembly queue.
1411 */
1412 if (i == dropscanidx)
1413 wrapped = 1;
1414 }
1415 dropscanidx = i;
1416 }
1417 IPQ_UNLOCK();
1418 splx(s);
1419 }
1420
1421 /*
1422 * Drain off all datagram fragments.
1423 */
1424 void
1425 ip_drain(void)
1426 {
1427
1428 /*
1429 * We may be called from a device's interrupt context. If
1430 * the ipq is already busy, just bail out now.
1431 */
1432 if (ipq_lock_try() == 0)
1433 return;
1434
1435 /*
1436 * Drop half the total fragments now. If more mbufs are needed,
1437 * we will be called again soon.
1438 */
1439 ip_reass_drophalf();
1440
1441 IPQ_UNLOCK();
1442 }
1443
1444 /*
1445 * Do option processing on a datagram,
1446 * possibly discarding it if bad options are encountered,
1447 * or forwarding it if source-routed.
1448 * Returns 1 if packet has been forwarded/freed,
1449 * 0 if the packet should be processed further.
1450 */
1451 int
1452 ip_dooptions(struct mbuf *m)
1453 {
1454 struct ip *ip = mtod(m, struct ip *);
1455 u_char *cp, *cp0;
1456 struct ip_timestamp *ipt;
1457 struct in_ifaddr *ia;
1458 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0;
1459 struct in_addr dst;
1460 n_time ntime;
1461
1462 dst = ip->ip_dst;
1463 cp = (u_char *)(ip + 1);
1464 cnt = (ip->ip_hl << 2) - sizeof (struct ip);
1465 for (; cnt > 0; cnt -= optlen, cp += optlen) {
1466 opt = cp[IPOPT_OPTVAL];
1467 if (opt == IPOPT_EOL)
1468 break;
1469 if (opt == IPOPT_NOP)
1470 optlen = 1;
1471 else {
1472 if (cnt < IPOPT_OLEN + sizeof(*cp)) {
1473 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1474 goto bad;
1475 }
1476 optlen = cp[IPOPT_OLEN];
1477 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
1478 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1479 goto bad;
1480 }
1481 }
1482 switch (opt) {
1483
1484 default:
1485 break;
1486
1487 /*
1488 * Source routing with record.
1489 * Find interface with current destination address.
1490 * If none on this machine then drop if strictly routed,
1491 * or do nothing if loosely routed.
1492 * Record interface address and bring up next address
1493 * component. If strictly routed make sure next
1494 * address is on directly accessible net.
1495 */
1496 case IPOPT_LSRR:
1497 case IPOPT_SSRR:
1498 if (ip_allowsrcrt == 0) {
1499 type = ICMP_UNREACH;
1500 code = ICMP_UNREACH_NET_PROHIB;
1501 goto bad;
1502 }
1503 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1504 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1505 goto bad;
1506 }
1507 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1508 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1509 goto bad;
1510 }
1511 ipaddr.sin_addr = ip->ip_dst;
1512 ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr)));
1513 if (ia == 0) {
1514 if (opt == IPOPT_SSRR) {
1515 type = ICMP_UNREACH;
1516 code = ICMP_UNREACH_SRCFAIL;
1517 goto bad;
1518 }
1519 /*
1520 * Loose routing, and not at next destination
1521 * yet; nothing to do except forward.
1522 */
1523 break;
1524 }
1525 off--; /* 0 origin */
1526 if ((off + sizeof(struct in_addr)) > optlen) {
1527 /*
1528 * End of source route. Should be for us.
1529 */
1530 save_rte(cp, ip->ip_src);
1531 break;
1532 }
1533 /*
1534 * locate outgoing interface
1535 */
1536 bcopy((void *)(cp + off), (void *)&ipaddr.sin_addr,
1537 sizeof(ipaddr.sin_addr));
1538 if (opt == IPOPT_SSRR)
1539 ia = ifatoia(ifa_ifwithladdr(sintosa(&ipaddr)));
1540 else
1541 ia = ip_rtaddr(ipaddr.sin_addr);
1542 if (ia == 0) {
1543 type = ICMP_UNREACH;
1544 code = ICMP_UNREACH_SRCFAIL;
1545 goto bad;
1546 }
1547 ip->ip_dst = ipaddr.sin_addr;
1548 bcopy((void *)&ia->ia_addr.sin_addr,
1549 (void *)(cp + off), sizeof(struct in_addr));
1550 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1551 /*
1552 * Let ip_intr's mcast routing check handle mcast pkts
1553 */
1554 forward = !IN_MULTICAST(ip->ip_dst.s_addr);
1555 break;
1556
1557 case IPOPT_RR:
1558 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1559 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1560 goto bad;
1561 }
1562 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1563 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1564 goto bad;
1565 }
1566 /*
1567 * If no space remains, ignore.
1568 */
1569 off--; /* 0 origin */
1570 if ((off + sizeof(struct in_addr)) > optlen)
1571 break;
1572 bcopy((void *)(&ip->ip_dst), (void *)&ipaddr.sin_addr,
1573 sizeof(ipaddr.sin_addr));
1574 /*
1575 * locate outgoing interface; if we're the destination,
1576 * use the incoming interface (should be same).
1577 */
1578 if ((ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))))
1579 == NULL &&
1580 (ia = ip_rtaddr(ipaddr.sin_addr)) == NULL) {
1581 type = ICMP_UNREACH;
1582 code = ICMP_UNREACH_HOST;
1583 goto bad;
1584 }
1585 bcopy((void *)&ia->ia_addr.sin_addr,
1586 (void *)(cp + off), sizeof(struct in_addr));
1587 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1588 break;
1589
1590 case IPOPT_TS:
1591 code = cp - (u_char *)ip;
1592 ipt = (struct ip_timestamp *)cp;
1593 if (ipt->ipt_len < 4 || ipt->ipt_len > 40) {
1594 code = (u_char *)&ipt->ipt_len - (u_char *)ip;
1595 goto bad;
1596 }
1597 if (ipt->ipt_ptr < 5) {
1598 code = (u_char *)&ipt->ipt_ptr - (u_char *)ip;
1599 goto bad;
1600 }
1601 if (ipt->ipt_ptr > ipt->ipt_len - sizeof (int32_t)) {
1602 if (++ipt->ipt_oflw == 0) {
1603 code = (u_char *)&ipt->ipt_ptr -
1604 (u_char *)ip;
1605 goto bad;
1606 }
1607 break;
1608 }
1609 cp0 = (cp + ipt->ipt_ptr - 1);
1610 switch (ipt->ipt_flg) {
1611
1612 case IPOPT_TS_TSONLY:
1613 break;
1614
1615 case IPOPT_TS_TSANDADDR:
1616 if (ipt->ipt_ptr - 1 + sizeof(n_time) +
1617 sizeof(struct in_addr) > ipt->ipt_len) {
1618 code = (u_char *)&ipt->ipt_ptr -
1619 (u_char *)ip;
1620 goto bad;
1621 }
1622 ipaddr.sin_addr = dst;
1623 ia = ifatoia(ifaof_ifpforaddr(sintosa(&ipaddr),
1624 m->m_pkthdr.rcvif));
1625 if (ia == 0)
1626 continue;
1627 bcopy(&ia->ia_addr.sin_addr,
1628 cp0, sizeof(struct in_addr));
1629 ipt->ipt_ptr += sizeof(struct in_addr);
1630 break;
1631
1632 case IPOPT_TS_PRESPEC:
1633 if (ipt->ipt_ptr - 1 + sizeof(n_time) +
1634 sizeof(struct in_addr) > ipt->ipt_len) {
1635 code = (u_char *)&ipt->ipt_ptr -
1636 (u_char *)ip;
1637 goto bad;
1638 }
1639 bcopy(cp0, &ipaddr.sin_addr,
1640 sizeof(struct in_addr));
1641 if (ifatoia(ifa_ifwithaddr(sintosa(&ipaddr)))
1642 == NULL)
1643 continue;
1644 ipt->ipt_ptr += sizeof(struct in_addr);
1645 break;
1646
1647 default:
1648 /* XXX can't take &ipt->ipt_flg */
1649 code = (u_char *)&ipt->ipt_ptr -
1650 (u_char *)ip + 1;
1651 goto bad;
1652 }
1653 ntime = iptime();
1654 cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */
1655 memmove((char *)cp + ipt->ipt_ptr - 1, cp0,
1656 sizeof(n_time));
1657 ipt->ipt_ptr += sizeof(n_time);
1658 }
1659 }
1660 if (forward) {
1661 if (ip_forwsrcrt == 0) {
1662 type = ICMP_UNREACH;
1663 code = ICMP_UNREACH_SRCFAIL;
1664 goto bad;
1665 }
1666 ip_forward(m, 1);
1667 return (1);
1668 }
1669 return (0);
1670 bad:
1671 icmp_error(m, type, code, 0, 0);
1672 ipstat.ips_badoptions++;
1673 return (1);
1674 }
1675
1676 /*
1677 * Given address of next destination (final or next hop),
1678 * return internet address info of interface to be used to get there.
1679 */
1680 struct in_ifaddr *
1681 ip_rtaddr(struct in_addr dst)
1682 {
1683 struct rtentry *rt;
1684 union {
1685 struct sockaddr dst;
1686 struct sockaddr_in dst4;
1687 } u;
1688
1689 sockaddr_in_init(&u.dst4, &dst, 0);
1690
1691 if ((rt = rtcache_lookup(&ipforward_rt, &u.dst)) == NULL)
1692 return NULL;
1693
1694 return ifatoia(rt->rt_ifa);
1695 }
1696
1697 /*
1698 * Save incoming source route for use in replies,
1699 * to be picked up later by ip_srcroute if the receiver is interested.
1700 */
1701 void
1702 save_rte(u_char *option, struct in_addr dst)
1703 {
1704 unsigned olen;
1705
1706 olen = option[IPOPT_OLEN];
1707 #ifdef DIAGNOSTIC
1708 if (ipprintfs)
1709 printf("save_rte: olen %d\n", olen);
1710 #endif /* 0 */
1711 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst)))
1712 return;
1713 bcopy((void *)option, (void *)ip_srcrt.srcopt, olen);
1714 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1715 ip_srcrt.dst = dst;
1716 }
1717
1718 /*
1719 * Retrieve incoming source route for use in replies,
1720 * in the same form used by setsockopt.
1721 * The first hop is placed before the options, will be removed later.
1722 */
1723 struct mbuf *
1724 ip_srcroute(void)
1725 {
1726 struct in_addr *p, *q;
1727 struct mbuf *m;
1728
1729 if (ip_nhops == 0)
1730 return NULL;
1731 m = m_get(M_DONTWAIT, MT_SOOPTS);
1732 if (m == 0)
1733 return NULL;
1734
1735 MCLAIM(m, &inetdomain.dom_mowner);
1736 #define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt))
1737
1738 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
1739 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) +
1740 OPTSIZ;
1741 #ifdef DIAGNOSTIC
1742 if (ipprintfs)
1743 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len);
1744 #endif
1745
1746 /*
1747 * First save first hop for return route
1748 */
1749 p = &ip_srcrt.route[ip_nhops - 1];
1750 *(mtod(m, struct in_addr *)) = *p--;
1751 #ifdef DIAGNOSTIC
1752 if (ipprintfs)
1753 printf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr));
1754 #endif
1755
1756 /*
1757 * Copy option fields and padding (nop) to mbuf.
1758 */
1759 ip_srcrt.nop = IPOPT_NOP;
1760 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
1761 memmove(mtod(m, char *) + sizeof(struct in_addr), &ip_srcrt.nop,
1762 OPTSIZ);
1763 q = (struct in_addr *)(mtod(m, char *) +
1764 sizeof(struct in_addr) + OPTSIZ);
1765 #undef OPTSIZ
1766 /*
1767 * Record return path as an IP source route,
1768 * reversing the path (pointers are now aligned).
1769 */
1770 while (p >= ip_srcrt.route) {
1771 #ifdef DIAGNOSTIC
1772 if (ipprintfs)
1773 printf(" %x", ntohl(q->s_addr));
1774 #endif
1775 *q++ = *p--;
1776 }
1777 /*
1778 * Last hop goes to final destination.
1779 */
1780 *q = ip_srcrt.dst;
1781 #ifdef DIAGNOSTIC
1782 if (ipprintfs)
1783 printf(" %x\n", ntohl(q->s_addr));
1784 #endif
1785 return (m);
1786 }
1787
1788 const int inetctlerrmap[PRC_NCMDS] = {
1789 [PRC_MSGSIZE] = EMSGSIZE,
1790 [PRC_HOSTDEAD] = EHOSTDOWN,
1791 [PRC_HOSTUNREACH] = EHOSTUNREACH,
1792 [PRC_UNREACH_NET] = EHOSTUNREACH,
1793 [PRC_UNREACH_HOST] = EHOSTUNREACH,
1794 [PRC_UNREACH_PROTOCOL] = ECONNREFUSED,
1795 [PRC_UNREACH_PORT] = ECONNREFUSED,
1796 [PRC_UNREACH_SRCFAIL] = EHOSTUNREACH,
1797 [PRC_PARAMPROB] = ENOPROTOOPT,
1798 };
1799
1800 /*
1801 * Forward a packet. If some error occurs return the sender
1802 * an icmp packet. Note we can't always generate a meaningful
1803 * icmp message because icmp doesn't have a large enough repertoire
1804 * of codes and types.
1805 *
1806 * If not forwarding, just drop the packet. This could be confusing
1807 * if ipforwarding was zero but some routing protocol was advancing
1808 * us as a gateway to somewhere. However, we must let the routing
1809 * protocol deal with that.
1810 *
1811 * The srcrt parameter indicates whether the packet is being forwarded
1812 * via a source route.
1813 */
1814 void
1815 ip_forward(struct mbuf *m, int srcrt)
1816 {
1817 struct ip *ip = mtod(m, struct ip *);
1818 struct rtentry *rt;
1819 int error, type = 0, code = 0, destmtu = 0;
1820 struct mbuf *mcopy;
1821 n_long dest;
1822 union {
1823 struct sockaddr dst;
1824 struct sockaddr_in dst4;
1825 } u;
1826
1827 /*
1828 * We are now in the output path.
1829 */
1830 MCLAIM(m, &ip_tx_mowner);
1831
1832 /*
1833 * Clear any in-bound checksum flags for this packet.
1834 */
1835 m->m_pkthdr.csum_flags = 0;
1836
1837 dest = 0;
1838 #ifdef DIAGNOSTIC
1839 if (ipprintfs) {
1840 printf("forward: src %s ", inet_ntoa(ip->ip_src));
1841 printf("dst %s ttl %x\n", inet_ntoa(ip->ip_dst), ip->ip_ttl);
1842 }
1843 #endif
1844 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
1845 ipstat.ips_cantforward++;
1846 m_freem(m);
1847 return;
1848 }
1849 if (ip->ip_ttl <= IPTTLDEC) {
1850 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0);
1851 return;
1852 }
1853
1854 sockaddr_in_init(&u.dst4, &ip->ip_dst, 0);
1855 if ((rt = rtcache_lookup(&ipforward_rt, &u.dst)) == NULL) {
1856 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_NET, dest, 0);
1857 return;
1858 }
1859
1860 /*
1861 * Save at most 68 bytes of the packet in case
1862 * we need to generate an ICMP message to the src.
1863 * Pullup to avoid sharing mbuf cluster between m and mcopy.
1864 */
1865 mcopy = m_copym(m, 0, imin(ntohs(ip->ip_len), 68), M_DONTWAIT);
1866 if (mcopy)
1867 mcopy = m_pullup(mcopy, ip->ip_hl << 2);
1868
1869 ip->ip_ttl -= IPTTLDEC;
1870
1871 /*
1872 * If forwarding packet using same interface that it came in on,
1873 * perhaps should send a redirect to sender to shortcut a hop.
1874 * Only send redirect if source is sending directly to us,
1875 * and if packet was not source routed (or has any options).
1876 * Also, don't send redirect if forwarding using a default route
1877 * or a route modified by a redirect.
1878 */
1879 if (rt->rt_ifp == m->m_pkthdr.rcvif &&
1880 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1881 !in_nullhost(satocsin(rt_getkey(rt))->sin_addr) &&
1882 ipsendredirects && !srcrt) {
1883 if (rt->rt_ifa &&
1884 (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) ==
1885 ifatoia(rt->rt_ifa)->ia_subnet) {
1886 if (rt->rt_flags & RTF_GATEWAY)
1887 dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
1888 else
1889 dest = ip->ip_dst.s_addr;
1890 /*
1891 * Router requirements says to only send host
1892 * redirects.
1893 */
1894 type = ICMP_REDIRECT;
1895 code = ICMP_REDIRECT_HOST;
1896 #ifdef DIAGNOSTIC
1897 if (ipprintfs)
1898 printf("redirect (%d) to %x\n", code,
1899 (u_int32_t)dest);
1900 #endif
1901 }
1902 }
1903
1904 error = ip_output(m, NULL, &ipforward_rt,
1905 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)),
1906 (struct ip_moptions *)NULL, (struct socket *)NULL);
1907
1908 if (error)
1909 ipstat.ips_cantforward++;
1910 else {
1911 ipstat.ips_forward++;
1912 if (type)
1913 ipstat.ips_redirectsent++;
1914 else {
1915 if (mcopy) {
1916 #ifdef GATEWAY
1917 if (mcopy->m_flags & M_CANFASTFWD)
1918 ipflow_create(&ipforward_rt, mcopy);
1919 #endif
1920 m_freem(mcopy);
1921 }
1922 return;
1923 }
1924 }
1925 if (mcopy == NULL)
1926 return;
1927
1928 switch (error) {
1929
1930 case 0: /* forwarded, but need redirect */
1931 /* type, code set above */
1932 break;
1933
1934 case ENETUNREACH: /* shouldn't happen, checked above */
1935 case EHOSTUNREACH:
1936 case ENETDOWN:
1937 case EHOSTDOWN:
1938 default:
1939 type = ICMP_UNREACH;
1940 code = ICMP_UNREACH_HOST;
1941 break;
1942
1943 case EMSGSIZE:
1944 type = ICMP_UNREACH;
1945 code = ICMP_UNREACH_NEEDFRAG;
1946 #if !defined(IPSEC) && !defined(FAST_IPSEC)
1947 if ((rt = rtcache_getrt(&ipforward_rt)) != NULL)
1948 destmtu = rt->rt_ifp->if_mtu;
1949 #else
1950 /*
1951 * If the packet is routed over IPsec tunnel, tell the
1952 * originator the tunnel MTU.
1953 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
1954 * XXX quickhack!!!
1955 */
1956 if ((rt = rtcache_getrt(&ipforward_rt)) != NULL) {
1957 struct secpolicy *sp;
1958 int ipsecerror;
1959 size_t ipsechdr;
1960 struct route *ro;
1961
1962 sp = ipsec4_getpolicybyaddr(mcopy,
1963 IPSEC_DIR_OUTBOUND, IP_FORWARDING,
1964 &ipsecerror);
1965
1966 if (sp == NULL)
1967 destmtu = rt->rt_ifp->if_mtu;
1968 else {
1969 /* count IPsec header size */
1970 ipsechdr = ipsec4_hdrsiz(mcopy,
1971 IPSEC_DIR_OUTBOUND, NULL);
1972
1973 /*
1974 * find the correct route for outer IPv4
1975 * header, compute tunnel MTU.
1976 */
1977
1978 if (sp->req != NULL
1979 && sp->req->sav != NULL
1980 && sp->req->sav->sah != NULL) {
1981 ro = &sp->req->sav->sah->sa_route;
1982 if (rt && rt->rt_ifp) {
1983 destmtu =
1984 rt->rt_rmx.rmx_mtu ?
1985 rt->rt_rmx.rmx_mtu :
1986 rt->rt_ifp->if_mtu;
1987 destmtu -= ipsechdr;
1988 }
1989 }
1990
1991 #ifdef IPSEC
1992 key_freesp(sp);
1993 #else
1994 KEY_FREESP(&sp);
1995 #endif
1996 }
1997 }
1998 #endif /*IPSEC*/
1999 ipstat.ips_cantfrag++;
2000 break;
2001
2002 case ENOBUFS:
2003 #if 1
2004 /*
2005 * a router should not generate ICMP_SOURCEQUENCH as
2006 * required in RFC1812 Requirements for IP Version 4 Routers.
2007 * source quench could be a big problem under DoS attacks,
2008 * or if the underlying interface is rate-limited.
2009 */
2010 if (mcopy)
2011 m_freem(mcopy);
2012 return;
2013 #else
2014 type = ICMP_SOURCEQUENCH;
2015 code = 0;
2016 break;
2017 #endif
2018 }
2019 icmp_error(mcopy, type, code, dest, destmtu);
2020 }
2021
2022 void
2023 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
2024 struct mbuf *m)
2025 {
2026
2027 if (inp->inp_socket->so_options & SO_TIMESTAMP) {
2028 struct timeval tv;
2029
2030 microtime(&tv);
2031 *mp = sbcreatecontrol((void *) &tv, sizeof(tv),
2032 SCM_TIMESTAMP, SOL_SOCKET);
2033 if (*mp)
2034 mp = &(*mp)->m_next;
2035 }
2036 if (inp->inp_flags & INP_RECVDSTADDR) {
2037 *mp = sbcreatecontrol((void *) &ip->ip_dst,
2038 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
2039 if (*mp)
2040 mp = &(*mp)->m_next;
2041 }
2042 #ifdef notyet
2043 /*
2044 * XXX
2045 * Moving these out of udp_input() made them even more broken
2046 * than they already were.
2047 * - fenner (at) parc.xerox.com
2048 */
2049 /* options were tossed already */
2050 if (inp->inp_flags & INP_RECVOPTS) {
2051 *mp = sbcreatecontrol((void *) opts_deleted_above,
2052 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
2053 if (*mp)
2054 mp = &(*mp)->m_next;
2055 }
2056 /* ip_srcroute doesn't do what we want here, need to fix */
2057 if (inp->inp_flags & INP_RECVRETOPTS) {
2058 *mp = sbcreatecontrol((void *) ip_srcroute(),
2059 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
2060 if (*mp)
2061 mp = &(*mp)->m_next;
2062 }
2063 #endif
2064 if (inp->inp_flags & INP_RECVIF) {
2065 struct sockaddr_dl sdl;
2066
2067 sockaddr_dl_init(&sdl, sizeof(sdl),
2068 (m->m_pkthdr.rcvif != NULL)
2069 ? m->m_pkthdr.rcvif->if_index
2070 : 0,
2071 0, NULL, 0, NULL, 0);
2072 *mp = sbcreatecontrol(&sdl, sdl.sdl_len, IP_RECVIF, IPPROTO_IP);
2073 if (*mp)
2074 mp = &(*mp)->m_next;
2075 }
2076 }
2077
2078 /*
2079 * sysctl helper routine for net.inet.ip.forwsrcrt.
2080 */
2081 static int
2082 sysctl_net_inet_ip_forwsrcrt(SYSCTLFN_ARGS)
2083 {
2084 int error, tmp;
2085 struct sysctlnode node;
2086
2087 node = *rnode;
2088 tmp = ip_forwsrcrt;
2089 node.sysctl_data = &tmp;
2090 error = sysctl_lookup(SYSCTLFN_CALL(&node));
2091 if (error || newp == NULL)
2092 return (error);
2093
2094 if (kauth_authorize_network(l->l_cred, KAUTH_NETWORK_FORWSRCRT,
2095 0, NULL, NULL, NULL))
2096 return (EPERM);
2097
2098 ip_forwsrcrt = tmp;
2099
2100 return (0);
2101 }
2102
2103 /*
2104 * sysctl helper routine for net.inet.ip.mtudisctimeout. checks the
2105 * range of the new value and tweaks timers if it changes.
2106 */
2107 static int
2108 sysctl_net_inet_ip_pmtudto(SYSCTLFN_ARGS)
2109 {
2110 int error, tmp;
2111 struct sysctlnode node;
2112
2113 node = *rnode;
2114 tmp = ip_mtudisc_timeout;
2115 node.sysctl_data = &tmp;
2116 error = sysctl_lookup(SYSCTLFN_CALL(&node));
2117 if (error || newp == NULL)
2118 return (error);
2119 if (tmp < 0)
2120 return (EINVAL);
2121
2122 ip_mtudisc_timeout = tmp;
2123 rt_timer_queue_change(ip_mtudisc_timeout_q, ip_mtudisc_timeout);
2124
2125 return (0);
2126 }
2127
2128 #ifdef GATEWAY
2129 /*
2130 * sysctl helper routine for net.inet.ip.maxflows.
2131 */
2132 static int
2133 sysctl_net_inet_ip_maxflows(SYSCTLFN_ARGS)
2134 {
2135 int s;
2136
2137 s = sysctl_lookup(SYSCTLFN_CALL(rnode));
2138 if (s || newp == NULL)
2139 return (s);
2140
2141 s = splsoftnet();
2142 ipflow_reap(0);
2143 splx(s);
2144
2145 return (0);
2146 }
2147
2148 static int
2149 sysctl_net_inet_ip_hashsize(SYSCTLFN_ARGS)
2150 {
2151 int error, tmp;
2152 struct sysctlnode node;
2153
2154 node = *rnode;
2155 tmp = ip_hashsize;
2156 node.sysctl_data = &tmp;
2157 error = sysctl_lookup(SYSCTLFN_CALL(&node));
2158 if (error || newp == NULL)
2159 return (error);
2160
2161 if ((tmp & (tmp - 1)) == 0 && tmp != 0) {
2162 /*
2163 * Can only fail due to malloc()
2164 */
2165 if (ipflow_invalidate_all(tmp))
2166 return ENOMEM;
2167 } else {
2168 /*
2169 * EINVAL if not a power of 2
2170 */
2171 return EINVAL;
2172 }
2173
2174 return (0);
2175 }
2176 #endif /* GATEWAY */
2177
2178
2179 SYSCTL_SETUP(sysctl_net_inet_ip_setup, "sysctl net.inet.ip subtree setup")
2180 {
2181 extern int subnetsarelocal, hostzeroisbroadcast;
2182
2183 sysctl_createv(clog, 0, NULL, NULL,
2184 CTLFLAG_PERMANENT,
2185 CTLTYPE_NODE, "net", NULL,
2186 NULL, 0, NULL, 0,
2187 CTL_NET, CTL_EOL);
2188 sysctl_createv(clog, 0, NULL, NULL,
2189 CTLFLAG_PERMANENT,
2190 CTLTYPE_NODE, "inet",
2191 SYSCTL_DESCR("PF_INET related settings"),
2192 NULL, 0, NULL, 0,
2193 CTL_NET, PF_INET, CTL_EOL);
2194 sysctl_createv(clog, 0, NULL, NULL,
2195 CTLFLAG_PERMANENT,
2196 CTLTYPE_NODE, "ip",
2197 SYSCTL_DESCR("IPv4 related settings"),
2198 NULL, 0, NULL, 0,
2199 CTL_NET, PF_INET, IPPROTO_IP, CTL_EOL);
2200
2201 sysctl_createv(clog, 0, NULL, NULL,
2202 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2203 CTLTYPE_INT, "forwarding",
2204 SYSCTL_DESCR("Enable forwarding of INET datagrams"),
2205 NULL, 0, &ipforwarding, 0,
2206 CTL_NET, PF_INET, IPPROTO_IP,
2207 IPCTL_FORWARDING, CTL_EOL);
2208 sysctl_createv(clog, 0, NULL, NULL,
2209 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2210 CTLTYPE_INT, "redirect",
2211 SYSCTL_DESCR("Enable sending of ICMP redirect messages"),
2212 NULL, 0, &ipsendredirects, 0,
2213 CTL_NET, PF_INET, IPPROTO_IP,
2214 IPCTL_SENDREDIRECTS, CTL_EOL);
2215 sysctl_createv(clog, 0, NULL, NULL,
2216 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2217 CTLTYPE_INT, "ttl",
2218 SYSCTL_DESCR("Default TTL for an INET datagram"),
2219 NULL, 0, &ip_defttl, 0,
2220 CTL_NET, PF_INET, IPPROTO_IP,
2221 IPCTL_DEFTTL, CTL_EOL);
2222 #ifdef IPCTL_DEFMTU
2223 sysctl_createv(clog, 0, NULL, NULL,
2224 CTLFLAG_PERMANENT /* |CTLFLAG_READWRITE? */,
2225 CTLTYPE_INT, "mtu",
2226 SYSCTL_DESCR("Default MTA for an INET route"),
2227 NULL, 0, &ip_mtu, 0,
2228 CTL_NET, PF_INET, IPPROTO_IP,
2229 IPCTL_DEFMTU, CTL_EOL);
2230 #endif /* IPCTL_DEFMTU */
2231 sysctl_createv(clog, 0, NULL, NULL,
2232 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2233 CTLTYPE_INT, "forwsrcrt",
2234 SYSCTL_DESCR("Enable forwarding of source-routed "
2235 "datagrams"),
2236 sysctl_net_inet_ip_forwsrcrt, 0, &ip_forwsrcrt, 0,
2237 CTL_NET, PF_INET, IPPROTO_IP,
2238 IPCTL_FORWSRCRT, CTL_EOL);
2239 sysctl_createv(clog, 0, NULL, NULL,
2240 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2241 CTLTYPE_INT, "directed-broadcast",
2242 SYSCTL_DESCR("Enable forwarding of broadcast datagrams"),
2243 NULL, 0, &ip_directedbcast, 0,
2244 CTL_NET, PF_INET, IPPROTO_IP,
2245 IPCTL_DIRECTEDBCAST, CTL_EOL);
2246 sysctl_createv(clog, 0, NULL, NULL,
2247 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2248 CTLTYPE_INT, "allowsrcrt",
2249 SYSCTL_DESCR("Accept source-routed datagrams"),
2250 NULL, 0, &ip_allowsrcrt, 0,
2251 CTL_NET, PF_INET, IPPROTO_IP,
2252 IPCTL_ALLOWSRCRT, CTL_EOL);
2253 sysctl_createv(clog, 0, NULL, NULL,
2254 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2255 CTLTYPE_INT, "subnetsarelocal",
2256 SYSCTL_DESCR("Whether logical subnets are considered "
2257 "local"),
2258 NULL, 0, &subnetsarelocal, 0,
2259 CTL_NET, PF_INET, IPPROTO_IP,
2260 IPCTL_SUBNETSARELOCAL, CTL_EOL);
2261 sysctl_createv(clog, 0, NULL, NULL,
2262 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2263 CTLTYPE_INT, "mtudisc",
2264 SYSCTL_DESCR("Use RFC1191 Path MTU Discovery"),
2265 NULL, 0, &ip_mtudisc, 0,
2266 CTL_NET, PF_INET, IPPROTO_IP,
2267 IPCTL_MTUDISC, CTL_EOL);
2268 sysctl_createv(clog, 0, NULL, NULL,
2269 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2270 CTLTYPE_INT, "anonportmin",
2271 SYSCTL_DESCR("Lowest ephemeral port number to assign"),
2272 sysctl_net_inet_ip_ports, 0, &anonportmin, 0,
2273 CTL_NET, PF_INET, IPPROTO_IP,
2274 IPCTL_ANONPORTMIN, CTL_EOL);
2275 sysctl_createv(clog, 0, NULL, NULL,
2276 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2277 CTLTYPE_INT, "anonportmax",
2278 SYSCTL_DESCR("Highest ephemeral port number to assign"),
2279 sysctl_net_inet_ip_ports, 0, &anonportmax, 0,
2280 CTL_NET, PF_INET, IPPROTO_IP,
2281 IPCTL_ANONPORTMAX, CTL_EOL);
2282 sysctl_createv(clog, 0, NULL, NULL,
2283 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2284 CTLTYPE_INT, "mtudisctimeout",
2285 SYSCTL_DESCR("Lifetime of a Path MTU Discovered route"),
2286 sysctl_net_inet_ip_pmtudto, 0, &ip_mtudisc_timeout, 0,
2287 CTL_NET, PF_INET, IPPROTO_IP,
2288 IPCTL_MTUDISCTIMEOUT, CTL_EOL);
2289 #ifdef GATEWAY
2290 sysctl_createv(clog, 0, NULL, NULL,
2291 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2292 CTLTYPE_INT, "maxflows",
2293 SYSCTL_DESCR("Number of flows for fast forwarding"),
2294 sysctl_net_inet_ip_maxflows, 0, &ip_maxflows, 0,
2295 CTL_NET, PF_INET, IPPROTO_IP,
2296 IPCTL_MAXFLOWS, CTL_EOL);
2297 sysctl_createv(clog, 0, NULL, NULL,
2298 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2299 CTLTYPE_INT, "hashsize",
2300 SYSCTL_DESCR("Size of hash table for fast forwarding (IPv4)"),
2301 sysctl_net_inet_ip_hashsize, 0, &ip_hashsize, 0,
2302 CTL_NET, PF_INET, IPPROTO_IP,
2303 CTL_CREATE, CTL_EOL);
2304 #endif /* GATEWAY */
2305 sysctl_createv(clog, 0, NULL, NULL,
2306 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2307 CTLTYPE_INT, "hostzerobroadcast",
2308 SYSCTL_DESCR("All zeroes address is broadcast address"),
2309 NULL, 0, &hostzeroisbroadcast, 0,
2310 CTL_NET, PF_INET, IPPROTO_IP,
2311 IPCTL_HOSTZEROBROADCAST, CTL_EOL);
2312 #if NGIF > 0
2313 sysctl_createv(clog, 0, NULL, NULL,
2314 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2315 CTLTYPE_INT, "gifttl",
2316 SYSCTL_DESCR("Default TTL for a gif tunnel datagram"),
2317 NULL, 0, &ip_gif_ttl, 0,
2318 CTL_NET, PF_INET, IPPROTO_IP,
2319 IPCTL_GIF_TTL, CTL_EOL);
2320 #endif /* NGIF */
2321 #ifndef IPNOPRIVPORTS
2322 sysctl_createv(clog, 0, NULL, NULL,
2323 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2324 CTLTYPE_INT, "lowportmin",
2325 SYSCTL_DESCR("Lowest privileged ephemeral port number "
2326 "to assign"),
2327 sysctl_net_inet_ip_ports, 0, &lowportmin, 0,
2328 CTL_NET, PF_INET, IPPROTO_IP,
2329 IPCTL_LOWPORTMIN, CTL_EOL);
2330 sysctl_createv(clog, 0, NULL, NULL,
2331 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2332 CTLTYPE_INT, "lowportmax",
2333 SYSCTL_DESCR("Highest privileged ephemeral port number "
2334 "to assign"),
2335 sysctl_net_inet_ip_ports, 0, &lowportmax, 0,
2336 CTL_NET, PF_INET, IPPROTO_IP,
2337 IPCTL_LOWPORTMAX, CTL_EOL);
2338 #endif /* IPNOPRIVPORTS */
2339 sysctl_createv(clog, 0, NULL, NULL,
2340 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2341 CTLTYPE_INT, "maxfragpackets",
2342 SYSCTL_DESCR("Maximum number of fragments to retain for "
2343 "possible reassembly"),
2344 NULL, 0, &ip_maxfragpackets, 0,
2345 CTL_NET, PF_INET, IPPROTO_IP,
2346 IPCTL_MAXFRAGPACKETS, CTL_EOL);
2347 #if NGRE > 0
2348 sysctl_createv(clog, 0, NULL, NULL,
2349 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2350 CTLTYPE_INT, "grettl",
2351 SYSCTL_DESCR("Default TTL for a gre tunnel datagram"),
2352 NULL, 0, &ip_gre_ttl, 0,
2353 CTL_NET, PF_INET, IPPROTO_IP,
2354 IPCTL_GRE_TTL, CTL_EOL);
2355 #endif /* NGRE */
2356 sysctl_createv(clog, 0, NULL, NULL,
2357 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2358 CTLTYPE_INT, "checkinterface",
2359 SYSCTL_DESCR("Enable receive side of Strong ES model "
2360 "from RFC1122"),
2361 NULL, 0, &ip_checkinterface, 0,
2362 CTL_NET, PF_INET, IPPROTO_IP,
2363 IPCTL_CHECKINTERFACE, CTL_EOL);
2364 sysctl_createv(clog, 0, NULL, NULL,
2365 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2366 CTLTYPE_INT, "random_id",
2367 SYSCTL_DESCR("Assign random ip_id values"),
2368 NULL, 0, &ip_do_randomid, 0,
2369 CTL_NET, PF_INET, IPPROTO_IP,
2370 IPCTL_RANDOMID, CTL_EOL);
2371 sysctl_createv(clog, 0, NULL, NULL,
2372 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2373 CTLTYPE_INT, "do_loopback_cksum",
2374 SYSCTL_DESCR("Perform IP checksum on loopback"),
2375 NULL, 0, &ip_do_loopback_cksum, 0,
2376 CTL_NET, PF_INET, IPPROTO_IP,
2377 IPCTL_LOOPBACKCKSUM, CTL_EOL);
2378 sysctl_createv(clog, 0, NULL, NULL,
2379 CTLFLAG_PERMANENT,
2380 CTLTYPE_STRUCT, "stats",
2381 SYSCTL_DESCR("IP statistics"),
2382 NULL, 0, &ipstat, sizeof(ipstat),
2383 CTL_NET, PF_INET, IPPROTO_IP, IPCTL_STATS,
2384 CTL_EOL);
2385 }
2386