1 /* $NetBSD: ip_input.c,v 1.212.2.3 2007/09/17 20:14:21 bouyer 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 1.212.2.3 2007/09/17 20:14:21 bouyer 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
125 #include <net/if.h>
126 #include <net/if_dl.h>
127 #include <net/route.h>
128 #include <net/pfil.h>
129
130 #include <netinet/in.h>
131 #include <netinet/in_systm.h>
132 #include <netinet/ip.h>
133 #include <netinet/in_pcb.h>
134 #include <netinet/in_var.h>
135 #include <netinet/ip_var.h>
136 #include <netinet/ip_icmp.h>
137 /* just for gif_ttl */
138 #include <netinet/in_gif.h>
139 #include "gif.h"
140 #include <net/if_gre.h>
141 #include "gre.h"
142
143 #ifdef MROUTING
144 #include <netinet/ip_mroute.h>
145 #endif
146
147 #ifdef IPSEC
148 #include <netinet6/ipsec.h>
149 #include <netkey/key.h>
150 #endif
151 #ifdef FAST_IPSEC
152 #include <netipsec/ipsec.h>
153 #include <netipsec/key.h>
154 #endif /* FAST_IPSEC*/
155
156 #ifndef IPFORWARDING
157 #ifdef GATEWAY
158 #define IPFORWARDING 1 /* forward IP packets not for us */
159 #else /* GATEWAY */
160 #define IPFORWARDING 0 /* don't forward IP packets not for us */
161 #endif /* GATEWAY */
162 #endif /* IPFORWARDING */
163 #ifndef IPSENDREDIRECTS
164 #define IPSENDREDIRECTS 1
165 #endif
166 #ifndef IPFORWSRCRT
167 #define IPFORWSRCRT 1 /* forward source-routed packets */
168 #endif
169 #ifndef IPALLOWSRCRT
170 #define IPALLOWSRCRT 1 /* allow source-routed packets */
171 #endif
172 #ifndef IPMTUDISC
173 #define IPMTUDISC 1
174 #endif
175 #ifndef IPMTUDISCTIMEOUT
176 #define IPMTUDISCTIMEOUT (10 * 60) /* as per RFC 1191 */
177 #endif
178
179 /*
180 * Note: DIRECTED_BROADCAST is handled this way so that previous
181 * configuration using this option will Just Work.
182 */
183 #ifndef IPDIRECTEDBCAST
184 #ifdef DIRECTED_BROADCAST
185 #define IPDIRECTEDBCAST 1
186 #else
187 #define IPDIRECTEDBCAST 0
188 #endif /* DIRECTED_BROADCAST */
189 #endif /* IPDIRECTEDBCAST */
190 int ipforwarding = IPFORWARDING;
191 int ipsendredirects = IPSENDREDIRECTS;
192 int ip_defttl = IPDEFTTL;
193 int ip_forwsrcrt = IPFORWSRCRT;
194 int ip_directedbcast = IPDIRECTEDBCAST;
195 int ip_allowsrcrt = IPALLOWSRCRT;
196 int ip_mtudisc = IPMTUDISC;
197 int ip_mtudisc_timeout = IPMTUDISCTIMEOUT;
198 #ifdef DIAGNOSTIC
199 int ipprintfs = 0;
200 #endif
201
202 int ip_do_randomid = 0;
203
204 /*
205 * XXX - Setting ip_checkinterface mostly implements the receive side of
206 * the Strong ES model described in RFC 1122, but since the routing table
207 * and transmit implementation do not implement the Strong ES model,
208 * setting this to 1 results in an odd hybrid.
209 *
210 * XXX - ip_checkinterface currently must be disabled if you use ipnat
211 * to translate the destination address to another local interface.
212 *
213 * XXX - ip_checkinterface must be disabled if you add IP aliases
214 * to the loopback interface instead of the interface where the
215 * packets for those addresses are received.
216 */
217 int ip_checkinterface = 0;
218
219
220 struct rttimer_queue *ip_mtudisc_timeout_q = NULL;
221
222 int ipqmaxlen = IFQ_MAXLEN;
223 u_long in_ifaddrhash; /* size of hash table - 1 */
224 int in_ifaddrentries; /* total number of addrs */
225 struct in_ifaddrhead in_ifaddrhead;
226 struct in_ifaddrhashhead *in_ifaddrhashtbl;
227 u_long in_multihash; /* size of hash table - 1 */
228 int in_multientries; /* total number of addrs */
229 struct in_multihashhead *in_multihashtbl;
230 struct ifqueue ipintrq;
231 struct ipstat ipstat;
232 uint16_t ip_id;
233
234 #ifdef PFIL_HOOKS
235 struct pfil_head inet_pfil_hook;
236 #endif
237
238 /*
239 * Cached copy of nmbclusters. If nbclusters is different,
240 * recalculate IP parameters derived from nmbclusters.
241 */
242 static int ip_nmbclusters; /* copy of nmbclusters */
243 static void ip_nmbclusters_changed(void); /* recalc limits */
244
245 #define CHECK_NMBCLUSTER_PARAMS() \
246 do { \
247 if (__predict_false(ip_nmbclusters != nmbclusters)) \
248 ip_nmbclusters_changed(); \
249 } while (/*CONSTCOND*/0)
250
251 /* IP datagram reassembly queues (hashed) */
252 #define IPREASS_NHASH_LOG2 6
253 #define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2)
254 #define IPREASS_HMASK (IPREASS_NHASH - 1)
255 #define IPREASS_HASH(x,y) \
256 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK)
257 struct ipqhead ipq[IPREASS_NHASH];
258 int ipq_locked;
259 static int ip_nfragpackets; /* packets in reass queue */
260 static int ip_nfrags; /* total fragments in reass queues */
261
262 int ip_maxfragpackets = 200; /* limit on packets. XXX sysctl */
263 int ip_maxfrags; /* limit on fragments. XXX sysctl */
264
265
266 /*
267 * Additive-Increase/Multiplicative-Decrease (AIMD) strategy for
268 * IP reassembly queue buffer managment.
269 *
270 * We keep a count of total IP fragments (NB: not fragmented packets!)
271 * awaiting reassembly (ip_nfrags) and a limit (ip_maxfrags) on fragments.
272 * If ip_nfrags exceeds ip_maxfrags the limit, we drop half the
273 * total fragments in reassembly queues.This AIMD policy avoids
274 * repeatedly deleting single packets under heavy fragmentation load
275 * (e.g., from lossy NFS peers).
276 */
277 static u_int ip_reass_ttl_decr(u_int ticks);
278 static void ip_reass_drophalf(void);
279
280
281 static __inline int ipq_lock_try(void);
282 static __inline void ipq_unlock(void);
283
284 static __inline int
285 ipq_lock_try(void)
286 {
287 int s;
288
289 /*
290 * Use splvm() -- we're blocking things that would cause
291 * mbuf allocation.
292 */
293 s = splvm();
294 if (ipq_locked) {
295 splx(s);
296 return (0);
297 }
298 ipq_locked = 1;
299 splx(s);
300 return (1);
301 }
302
303 static __inline void
304 ipq_unlock(void)
305 {
306 int s;
307
308 s = splvm();
309 ipq_locked = 0;
310 splx(s);
311 }
312
313 #ifdef DIAGNOSTIC
314 #define IPQ_LOCK() \
315 do { \
316 if (ipq_lock_try() == 0) { \
317 printf("%s:%d: ipq already locked\n", __FILE__, __LINE__); \
318 panic("ipq_lock"); \
319 } \
320 } while (/*CONSTCOND*/ 0)
321 #define IPQ_LOCK_CHECK() \
322 do { \
323 if (ipq_locked == 0) { \
324 printf("%s:%d: ipq lock not held\n", __FILE__, __LINE__); \
325 panic("ipq lock check"); \
326 } \
327 } while (/*CONSTCOND*/ 0)
328 #else
329 #define IPQ_LOCK() (void) ipq_lock_try()
330 #define IPQ_LOCK_CHECK() /* nothing */
331 #endif
332
333 #define IPQ_UNLOCK() ipq_unlock()
334
335 POOL_INIT(inmulti_pool, sizeof(struct in_multi), 0, 0, 0, "inmltpl", NULL);
336 POOL_INIT(ipqent_pool, sizeof(struct ipqent), 0, 0, 0, "ipqepl", NULL);
337
338 #ifdef INET_CSUM_COUNTERS
339 #include <sys/device.h>
340
341 struct evcnt ip_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
342 NULL, "inet", "hwcsum bad");
343 struct evcnt ip_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
344 NULL, "inet", "hwcsum ok");
345 struct evcnt ip_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
346 NULL, "inet", "swcsum");
347
348 #define INET_CSUM_COUNTER_INCR(ev) (ev)->ev_count++
349
350 EVCNT_ATTACH_STATIC(ip_hwcsum_bad);
351 EVCNT_ATTACH_STATIC(ip_hwcsum_ok);
352 EVCNT_ATTACH_STATIC(ip_swcsum);
353
354 #else
355
356 #define INET_CSUM_COUNTER_INCR(ev) /* nothing */
357
358 #endif /* INET_CSUM_COUNTERS */
359
360 /*
361 * We need to save the IP options in case a protocol wants to respond
362 * to an incoming packet over the same route if the packet got here
363 * using IP source routing. This allows connection establishment and
364 * maintenance when the remote end is on a network that is not known
365 * to us.
366 */
367 int ip_nhops = 0;
368 static struct ip_srcrt {
369 struct in_addr dst; /* final destination */
370 char nop; /* one NOP to align */
371 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */
372 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)];
373 } ip_srcrt;
374
375 static void save_rte(u_char *, struct in_addr);
376
377 #ifdef MBUFTRACE
378 struct mowner ip_rx_mowner = { "internet", "rx" };
379 struct mowner ip_tx_mowner = { "internet", "tx" };
380 #endif
381
382 /*
383 * Compute IP limits derived from the value of nmbclusters.
384 */
385 static void
386 ip_nmbclusters_changed(void)
387 {
388 ip_maxfrags = nmbclusters / 4;
389 ip_nmbclusters = nmbclusters;
390 }
391
392 /*
393 * IP initialization: fill in IP protocol switch table.
394 * All protocols not implemented in kernel go to raw IP protocol handler.
395 */
396 void
397 ip_init(void)
398 {
399 const struct protosw *pr;
400 int i;
401
402 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
403 if (pr == 0)
404 panic("ip_init");
405 for (i = 0; i < IPPROTO_MAX; i++)
406 ip_protox[i] = pr - inetsw;
407 for (pr = inetdomain.dom_protosw;
408 pr < inetdomain.dom_protoswNPROTOSW; pr++)
409 if (pr->pr_domain->dom_family == PF_INET &&
410 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
411 ip_protox[pr->pr_protocol] = pr - inetsw;
412
413 for (i = 0; i < IPREASS_NHASH; i++)
414 LIST_INIT(&ipq[i]);
415
416 ip_id = time.tv_sec & 0xfffff;
417
418 ipintrq.ifq_maxlen = ipqmaxlen;
419 ip_nmbclusters_changed();
420
421 TAILQ_INIT(&in_ifaddrhead);
422 in_ifaddrhashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, M_IFADDR,
423 M_WAITOK, &in_ifaddrhash);
424 in_multihashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, M_IPMADDR,
425 M_WAITOK, &in_multihash);
426 ip_mtudisc_timeout_q = rt_timer_queue_create(ip_mtudisc_timeout);
427 #ifdef GATEWAY
428 ipflow_init();
429 #endif
430
431 #ifdef PFIL_HOOKS
432 /* Register our Packet Filter hook. */
433 inet_pfil_hook.ph_type = PFIL_TYPE_AF;
434 inet_pfil_hook.ph_af = AF_INET;
435 i = pfil_head_register(&inet_pfil_hook);
436 if (i != 0)
437 printf("ip_init: WARNING: unable to register pfil hook, "
438 "error %d\n", i);
439 #endif /* PFIL_HOOKS */
440
441 #ifdef MBUFTRACE
442 MOWNER_ATTACH(&ip_tx_mowner);
443 MOWNER_ATTACH(&ip_rx_mowner);
444 #endif /* MBUFTRACE */
445 }
446
447 struct sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET };
448 struct route ipforward_rt;
449
450 /*
451 * IP software interrupt routine
452 */
453 void
454 ipintr(void)
455 {
456 int s;
457 struct mbuf *m;
458
459 while (1) {
460 s = splnet();
461 IF_DEQUEUE(&ipintrq, m);
462 splx(s);
463 if (m == 0)
464 return;
465 MCLAIM(m, &ip_rx_mowner);
466 ip_input(m);
467 }
468 }
469
470 /*
471 * Ip input routine. Checksum and byte swap header. If fragmented
472 * try to reassemble. Process options. Pass to next level.
473 */
474 void
475 ip_input(struct mbuf *m)
476 {
477 struct ip *ip = NULL;
478 struct ipq *fp;
479 struct in_ifaddr *ia;
480 struct ifaddr *ifa;
481 struct ipqent *ipqe;
482 int hlen = 0, mff, len;
483 int downmatch;
484 int checkif;
485 int srcrt = 0;
486 u_int hash;
487 #ifdef FAST_IPSEC
488 struct m_tag *mtag;
489 struct tdb_ident *tdbi;
490 struct secpolicy *sp;
491 int s, error;
492 #endif /* FAST_IPSEC */
493
494 MCLAIM(m, &ip_rx_mowner);
495 #ifdef DIAGNOSTIC
496 if ((m->m_flags & M_PKTHDR) == 0)
497 panic("ipintr no HDR");
498 #endif
499
500 /*
501 * If no IP addresses have been set yet but the interfaces
502 * are receiving, can't do anything with incoming packets yet.
503 */
504 if (TAILQ_FIRST(&in_ifaddrhead) == 0)
505 goto bad;
506 ipstat.ips_total++;
507 /*
508 * If the IP header is not aligned, slurp it up into a new
509 * mbuf with space for link headers, in the event we forward
510 * it. Otherwise, if it is aligned, make sure the entire
511 * base IP header is in the first mbuf of the chain.
512 */
513 if (IP_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
514 if ((m = m_copyup(m, sizeof(struct ip),
515 (max_linkhdr + 3) & ~3)) == NULL) {
516 /* XXXJRT new stat, please */
517 ipstat.ips_toosmall++;
518 return;
519 }
520 } else if (__predict_false(m->m_len < sizeof (struct ip))) {
521 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
522 ipstat.ips_toosmall++;
523 return;
524 }
525 }
526 ip = mtod(m, struct ip *);
527 if (ip->ip_v != IPVERSION) {
528 ipstat.ips_badvers++;
529 goto bad;
530 }
531 hlen = ip->ip_hl << 2;
532 if (hlen < sizeof(struct ip)) { /* minimum header length */
533 ipstat.ips_badhlen++;
534 goto bad;
535 }
536 if (hlen > m->m_len) {
537 if ((m = m_pullup(m, hlen)) == 0) {
538 ipstat.ips_badhlen++;
539 return;
540 }
541 ip = mtod(m, struct ip *);
542 }
543
544 /*
545 * RFC1122: packets with a multicast source address are
546 * not allowed.
547 */
548 if (IN_MULTICAST(ip->ip_src.s_addr)) {
549 ipstat.ips_badaddr++;
550 goto bad;
551 }
552
553 /* 127/8 must not appear on wire - RFC1122 */
554 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
555 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
556 if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
557 ipstat.ips_badaddr++;
558 goto bad;
559 }
560 }
561
562 switch (m->m_pkthdr.csum_flags &
563 ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_IPv4) |
564 M_CSUM_IPv4_BAD)) {
565 case M_CSUM_IPv4|M_CSUM_IPv4_BAD:
566 INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad);
567 goto badcsum;
568
569 case M_CSUM_IPv4:
570 /* Checksum was okay. */
571 INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok);
572 break;
573
574 default:
575 /*
576 * Must compute it ourselves. Maybe skip checksum on
577 * loopback interfaces.
578 */
579 if (__predict_true(!(m->m_pkthdr.rcvif->if_flags &
580 IFF_LOOPBACK) || ip_do_loopback_cksum)) {
581 INET_CSUM_COUNTER_INCR(&ip_swcsum);
582 if (in_cksum(m, hlen) != 0)
583 goto badcsum;
584 }
585 break;
586 }
587
588 /* Retrieve the packet length. */
589 len = ntohs(ip->ip_len);
590
591 /*
592 * Check for additional length bogosity
593 */
594 if (len < hlen) {
595 ipstat.ips_badlen++;
596 goto bad;
597 }
598
599 /*
600 * Check that the amount of data in the buffers
601 * is as at least much as the IP header would have us expect.
602 * Trim mbufs if longer than we expect.
603 * Drop packet if shorter than we expect.
604 */
605 if (m->m_pkthdr.len < len) {
606 ipstat.ips_tooshort++;
607 goto bad;
608 }
609 if (m->m_pkthdr.len > len) {
610 if (m->m_len == m->m_pkthdr.len) {
611 m->m_len = len;
612 m->m_pkthdr.len = len;
613 } else
614 m_adj(m, len - m->m_pkthdr.len);
615 }
616
617 #if defined(IPSEC)
618 /* ipflow (IP fast forwarding) is not compatible with IPsec. */
619 m->m_flags &= ~M_CANFASTFWD;
620 #else
621 /*
622 * Assume that we can create a fast-forward IP flow entry
623 * based on this packet.
624 */
625 m->m_flags |= M_CANFASTFWD;
626 #endif
627
628 #ifdef PFIL_HOOKS
629 /*
630 * Run through list of hooks for input packets. If there are any
631 * filters which require that additional packets in the flow are
632 * not fast-forwarded, they must clear the M_CANFASTFWD flag.
633 * Note that filters must _never_ set this flag, as another filter
634 * in the list may have previously cleared it.
635 */
636 /*
637 * let ipfilter look at packet on the wire,
638 * not the decapsulated packet.
639 */
640 #ifdef IPSEC
641 if (!ipsec_getnhist(m))
642 #elif defined(FAST_IPSEC)
643 if (!ipsec_indone(m))
644 #else
645 if (1)
646 #endif
647 {
648 struct in_addr odst;
649
650 odst = ip->ip_dst;
651 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif,
652 PFIL_IN) != 0)
653 return;
654 if (m == NULL)
655 return;
656 ip = mtod(m, struct ip *);
657 hlen = ip->ip_hl << 2;
658 /*
659 * XXX The setting of "srcrt" here is to prevent ip_forward()
660 * from generating ICMP redirects for packets that have
661 * been redirected by a hook back out on to the same LAN that
662 * they came from and is not an indication that the packet
663 * is being inffluenced by source routing options. This
664 * allows things like
665 * "rdr tlp0 0/0 port 80 -> 1.1.1.200 3128 tcp"
666 * where tlp0 is both on the 1.1.1.0/24 network and is the
667 * default route for hosts on 1.1.1.0/24. Of course this
668 * also requires a "map tlp0 ..." to complete the story.
669 * One might argue whether or not this kind of network config.
670 * should be supported in this manner...
671 */
672 srcrt = (odst.s_addr != ip->ip_dst.s_addr);
673 }
674 #endif /* PFIL_HOOKS */
675
676 #ifdef ALTQ
677 /* XXX Temporary until ALTQ is changed to use a pfil hook */
678 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) {
679 /* packet dropped by traffic conditioner */
680 return;
681 }
682 #endif
683
684 /*
685 * Process options and, if not destined for us,
686 * ship it on. ip_dooptions returns 1 when an
687 * error was detected (causing an icmp message
688 * to be sent and the original packet to be freed).
689 */
690 ip_nhops = 0; /* for source routed packets */
691 if (hlen > sizeof (struct ip) && ip_dooptions(m))
692 return;
693
694 /*
695 * Enable a consistency check between the destination address
696 * and the arrival interface for a unicast packet (the RFC 1122
697 * strong ES model) if IP forwarding is disabled and the packet
698 * is not locally generated.
699 *
700 * XXX - Checking also should be disabled if the destination
701 * address is ipnat'ed to a different interface.
702 *
703 * XXX - Checking is incompatible with IP aliases added
704 * to the loopback interface instead of the interface where
705 * the packets are received.
706 *
707 * XXX - We need to add a per ifaddr flag for this so that
708 * we get finer grain control.
709 */
710 checkif = ip_checkinterface && (ipforwarding == 0) &&
711 (m->m_pkthdr.rcvif != NULL) &&
712 ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0);
713
714 /*
715 * Check our list of addresses, to see if the packet is for us.
716 *
717 * Traditional 4.4BSD did not consult IFF_UP at all.
718 * The behavior here is to treat addresses on !IFF_UP interface
719 * as not mine.
720 */
721 downmatch = 0;
722 LIST_FOREACH(ia, &IN_IFADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
723 if (in_hosteq(ia->ia_addr.sin_addr, ip->ip_dst)) {
724 if (checkif && ia->ia_ifp != m->m_pkthdr.rcvif)
725 continue;
726 if ((ia->ia_ifp->if_flags & IFF_UP) != 0)
727 break;
728 else
729 downmatch++;
730 }
731 }
732 if (ia != NULL)
733 goto ours;
734 if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
735 IFADDR_FOREACH(ifa, m->m_pkthdr.rcvif) {
736 if (ifa->ifa_addr->sa_family != AF_INET)
737 continue;
738 ia = ifatoia(ifa);
739 if (in_hosteq(ip->ip_dst, ia->ia_broadaddr.sin_addr) ||
740 in_hosteq(ip->ip_dst, ia->ia_netbroadcast) ||
741 /*
742 * Look for all-0's host part (old broadcast addr),
743 * either for subnet or net.
744 */
745 ip->ip_dst.s_addr == ia->ia_subnet ||
746 ip->ip_dst.s_addr == ia->ia_net)
747 goto ours;
748 /*
749 * An interface with IP address zero accepts
750 * all packets that arrive on that interface.
751 */
752 if (in_nullhost(ia->ia_addr.sin_addr))
753 goto ours;
754 }
755 }
756 if (IN_MULTICAST(ip->ip_dst.s_addr)) {
757 struct in_multi *inm;
758 #ifdef MROUTING
759 extern struct socket *ip_mrouter;
760
761 if (ip_mrouter) {
762 /*
763 * If we are acting as a multicast router, all
764 * incoming multicast packets are passed to the
765 * kernel-level multicast forwarding function.
766 * The packet is returned (relatively) intact; if
767 * ip_mforward() returns a non-zero value, the packet
768 * must be discarded, else it may be accepted below.
769 *
770 * (The IP ident field is put in the same byte order
771 * as expected when ip_mforward() is called from
772 * ip_output().)
773 */
774 if (ip_mforward(m, m->m_pkthdr.rcvif) != 0) {
775 ipstat.ips_cantforward++;
776 m_freem(m);
777 return;
778 }
779
780 /*
781 * The process-level routing demon needs to receive
782 * all multicast IGMP packets, whether or not this
783 * host belongs to their destination groups.
784 */
785 if (ip->ip_p == IPPROTO_IGMP)
786 goto ours;
787 ipstat.ips_forward++;
788 }
789 #endif
790 /*
791 * See if we belong to the destination multicast group on the
792 * arrival interface.
793 */
794 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
795 if (inm == NULL) {
796 ipstat.ips_cantforward++;
797 m_freem(m);
798 return;
799 }
800 goto ours;
801 }
802 if (ip->ip_dst.s_addr == INADDR_BROADCAST ||
803 in_nullhost(ip->ip_dst))
804 goto ours;
805
806 /*
807 * Not for us; forward if possible and desirable.
808 */
809 if (ipforwarding == 0) {
810 ipstat.ips_cantforward++;
811 m_freem(m);
812 } else {
813 /*
814 * If ip_dst matched any of my address on !IFF_UP interface,
815 * and there's no IFF_UP interface that matches ip_dst,
816 * send icmp unreach. Forwarding it will result in in-kernel
817 * forwarding loop till TTL goes to 0.
818 */
819 if (downmatch) {
820 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
821 ipstat.ips_cantforward++;
822 return;
823 }
824 #ifdef IPSEC
825 if (ipsec4_in_reject(m, NULL)) {
826 ipsecstat.in_polvio++;
827 goto bad;
828 }
829 #endif
830 #ifdef FAST_IPSEC
831 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
832 s = splsoftnet();
833 if (mtag != NULL) {
834 tdbi = (struct tdb_ident *)(mtag + 1);
835 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
836 } else {
837 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
838 IP_FORWARDING, &error);
839 }
840 if (sp == NULL) { /* NB: can happen if error */
841 splx(s);
842 /*XXX error stat???*/
843 DPRINTF(("ip_input: no SP for forwarding\n")); /*XXX*/
844 goto bad;
845 }
846
847 /*
848 * Check security policy against packet attributes.
849 */
850 error = ipsec_in_reject(sp, m);
851 KEY_FREESP(&sp);
852 splx(s);
853 if (error) {
854 ipstat.ips_cantforward++;
855 goto bad;
856 }
857
858 /*
859 * Peek at the outbound SP for this packet to determine if
860 * it's a Fast Forward candidate.
861 */
862 mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL);
863 if (mtag != NULL)
864 m->m_flags &= ~M_CANFASTFWD;
865 else {
866 s = splsoftnet();
867 sp = ipsec4_checkpolicy(m, IPSEC_DIR_OUTBOUND,
868 (IP_FORWARDING |
869 (ip_directedbcast ? IP_ALLOWBROADCAST : 0)),
870 &error, NULL);
871 if (sp != NULL) {
872 m->m_flags &= ~M_CANFASTFWD;
873 KEY_FREESP(&sp);
874 }
875 splx(s);
876 }
877 #endif /* FAST_IPSEC */
878
879 ip_forward(m, srcrt);
880 }
881 return;
882
883 ours:
884 /*
885 * If offset or IP_MF are set, must reassemble.
886 * Otherwise, nothing need be done.
887 * (We could look in the reassembly queue to see
888 * if the packet was previously fragmented,
889 * but it's not worth the time; just let them time out.)
890 */
891 if (ip->ip_off & ~htons(IP_DF|IP_RF)) {
892
893 /*
894 * Look for queue of fragments
895 * of this datagram.
896 */
897 IPQ_LOCK();
898 hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
899 /* XXX LIST_FOREACH(fp, &ipq[hash], ipq_q) */
900 for (fp = LIST_FIRST(&ipq[hash]); fp != NULL;
901 fp = LIST_NEXT(fp, ipq_q)) {
902 if (ip->ip_id == fp->ipq_id &&
903 in_hosteq(ip->ip_src, fp->ipq_src) &&
904 in_hosteq(ip->ip_dst, fp->ipq_dst) &&
905 ip->ip_p == fp->ipq_p)
906 goto found;
907
908 }
909 fp = 0;
910 found:
911
912 /*
913 * Adjust ip_len to not reflect header,
914 * set ipqe_mff if more fragments are expected,
915 * convert offset of this to bytes.
916 */
917 ip->ip_len = htons(ntohs(ip->ip_len) - hlen);
918 mff = (ip->ip_off & htons(IP_MF)) != 0;
919 if (mff) {
920 /*
921 * Make sure that fragments have a data length
922 * that's a non-zero multiple of 8 bytes.
923 */
924 if (ntohs(ip->ip_len) == 0 ||
925 (ntohs(ip->ip_len) & 0x7) != 0) {
926 ipstat.ips_badfrags++;
927 IPQ_UNLOCK();
928 goto bad;
929 }
930 }
931 ip->ip_off = htons((ntohs(ip->ip_off) & IP_OFFMASK) << 3);
932
933 /*
934 * If datagram marked as having more fragments
935 * or if this is not the first fragment,
936 * attempt reassembly; if it succeeds, proceed.
937 */
938 if (mff || ip->ip_off != htons(0)) {
939 ipstat.ips_fragments++;
940 ipqe = pool_get(&ipqent_pool, PR_NOWAIT);
941 if (ipqe == NULL) {
942 ipstat.ips_rcvmemdrop++;
943 IPQ_UNLOCK();
944 goto bad;
945 }
946 ipqe->ipqe_mff = mff;
947 ipqe->ipqe_m = m;
948 ipqe->ipqe_ip = ip;
949 m = ip_reass(ipqe, fp, &ipq[hash]);
950 if (m == 0) {
951 IPQ_UNLOCK();
952 return;
953 }
954 ipstat.ips_reassembled++;
955 ip = mtod(m, struct ip *);
956 hlen = ip->ip_hl << 2;
957 ip->ip_len = htons(ntohs(ip->ip_len) + hlen);
958 } else
959 if (fp)
960 ip_freef(fp);
961 IPQ_UNLOCK();
962 }
963
964 #if defined(IPSEC)
965 /*
966 * enforce IPsec policy checking if we are seeing last header.
967 * note that we do not visit this with protocols with pcb layer
968 * code - like udp/tcp/raw ip.
969 */
970 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0 &&
971 ipsec4_in_reject(m, NULL)) {
972 ipsecstat.in_polvio++;
973 goto bad;
974 }
975 #endif
976 #if FAST_IPSEC
977 /*
978 * enforce IPsec policy checking if we are seeing last header.
979 * note that we do not visit this with protocols with pcb layer
980 * code - like udp/tcp/raw ip.
981 */
982 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) {
983 /*
984 * Check if the packet has already had IPsec processing
985 * done. If so, then just pass it along. This tag gets
986 * set during AH, ESP, etc. input handling, before the
987 * packet is returned to the ip input queue for delivery.
988 */
989 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
990 s = splsoftnet();
991 if (mtag != NULL) {
992 tdbi = (struct tdb_ident *)(mtag + 1);
993 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
994 } else {
995 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
996 IP_FORWARDING, &error);
997 }
998 if (sp != NULL) {
999 /*
1000 * Check security policy against packet attributes.
1001 */
1002 error = ipsec_in_reject(sp, m);
1003 KEY_FREESP(&sp);
1004 } else {
1005 /* XXX error stat??? */
1006 error = EINVAL;
1007 DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/
1008 }
1009 splx(s);
1010 if (error)
1011 goto bad;
1012 }
1013 #endif /* FAST_IPSEC */
1014
1015 /*
1016 * Switch out to protocol's input routine.
1017 */
1018 #if IFA_STATS
1019 if (ia && ip)
1020 ia->ia_ifa.ifa_data.ifad_inbytes += ntohs(ip->ip_len);
1021 #endif
1022 ipstat.ips_delivered++;
1023 {
1024 int off = hlen, nh = ip->ip_p;
1025
1026 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh);
1027 return;
1028 }
1029 bad:
1030 m_freem(m);
1031 return;
1032
1033 badcsum:
1034 ipstat.ips_badsum++;
1035 m_freem(m);
1036 }
1037
1038 /*
1039 * Take incoming datagram fragment and try to
1040 * reassemble it into whole datagram. If a chain for
1041 * reassembly of this datagram already exists, then it
1042 * is given as fp; otherwise have to make a chain.
1043 */
1044 struct mbuf *
1045 ip_reass(struct ipqent *ipqe, struct ipq *fp, struct ipqhead *ipqhead)
1046 {
1047 struct mbuf *m = ipqe->ipqe_m;
1048 struct ipqent *nq, *p, *q;
1049 struct ip *ip;
1050 struct mbuf *t;
1051 int hlen = ipqe->ipqe_ip->ip_hl << 2;
1052 int i, next;
1053
1054 IPQ_LOCK_CHECK();
1055
1056 /*
1057 * Presence of header sizes in mbufs
1058 * would confuse code below.
1059 */
1060 m->m_data += hlen;
1061 m->m_len -= hlen;
1062
1063 #ifdef notyet
1064 /* make sure fragment limit is up-to-date */
1065 CHECK_NMBCLUSTER_PARAMS();
1066
1067 /* If we have too many fragments, drop the older half. */
1068 if (ip_nfrags >= ip_maxfrags)
1069 ip_reass_drophalf(void);
1070 #endif
1071
1072 /*
1073 * We are about to add a fragment; increment frag count.
1074 */
1075 ip_nfrags++;
1076
1077 /*
1078 * If first fragment to arrive, create a reassembly queue.
1079 */
1080 if (fp == 0) {
1081 /*
1082 * Enforce upper bound on number of fragmented packets
1083 * for which we attempt reassembly;
1084 * If maxfrag is 0, never accept fragments.
1085 * If maxfrag is -1, accept all fragments without limitation.
1086 */
1087 if (ip_maxfragpackets < 0)
1088 ;
1089 else if (ip_nfragpackets >= ip_maxfragpackets)
1090 goto dropfrag;
1091 ip_nfragpackets++;
1092 MALLOC(fp, struct ipq *, sizeof (struct ipq),
1093 M_FTABLE, M_NOWAIT);
1094 if (fp == NULL)
1095 goto dropfrag;
1096 LIST_INSERT_HEAD(ipqhead, fp, ipq_q);
1097 fp->ipq_nfrags = 1;
1098 fp->ipq_ttl = IPFRAGTTL;
1099 fp->ipq_p = ipqe->ipqe_ip->ip_p;
1100 fp->ipq_id = ipqe->ipqe_ip->ip_id;
1101 TAILQ_INIT(&fp->ipq_fragq);
1102 fp->ipq_src = ipqe->ipqe_ip->ip_src;
1103 fp->ipq_dst = ipqe->ipqe_ip->ip_dst;
1104 p = NULL;
1105 goto insert;
1106 } else {
1107 fp->ipq_nfrags++;
1108 }
1109
1110 /*
1111 * Find a segment which begins after this one does.
1112 */
1113 for (p = NULL, q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL;
1114 p = q, q = TAILQ_NEXT(q, ipqe_q))
1115 if (ntohs(q->ipqe_ip->ip_off) > ntohs(ipqe->ipqe_ip->ip_off))
1116 break;
1117
1118 /*
1119 * If there is a preceding segment, it may provide some of
1120 * our data already. If so, drop the data from the incoming
1121 * segment. If it provides all of our data, drop us.
1122 */
1123 if (p != NULL) {
1124 i = ntohs(p->ipqe_ip->ip_off) + ntohs(p->ipqe_ip->ip_len) -
1125 ntohs(ipqe->ipqe_ip->ip_off);
1126 if (i > 0) {
1127 if (i >= ntohs(ipqe->ipqe_ip->ip_len))
1128 goto dropfrag;
1129 m_adj(ipqe->ipqe_m, i);
1130 ipqe->ipqe_ip->ip_off =
1131 htons(ntohs(ipqe->ipqe_ip->ip_off) + i);
1132 ipqe->ipqe_ip->ip_len =
1133 htons(ntohs(ipqe->ipqe_ip->ip_len) - i);
1134 }
1135 }
1136
1137 /*
1138 * While we overlap succeeding segments trim them or,
1139 * if they are completely covered, dequeue them.
1140 */
1141 for (; q != NULL &&
1142 ntohs(ipqe->ipqe_ip->ip_off) + ntohs(ipqe->ipqe_ip->ip_len) >
1143 ntohs(q->ipqe_ip->ip_off); q = nq) {
1144 i = (ntohs(ipqe->ipqe_ip->ip_off) +
1145 ntohs(ipqe->ipqe_ip->ip_len)) - ntohs(q->ipqe_ip->ip_off);
1146 if (i < ntohs(q->ipqe_ip->ip_len)) {
1147 q->ipqe_ip->ip_len =
1148 htons(ntohs(q->ipqe_ip->ip_len) - i);
1149 q->ipqe_ip->ip_off =
1150 htons(ntohs(q->ipqe_ip->ip_off) + i);
1151 m_adj(q->ipqe_m, i);
1152 break;
1153 }
1154 nq = TAILQ_NEXT(q, ipqe_q);
1155 m_freem(q->ipqe_m);
1156 TAILQ_REMOVE(&fp->ipq_fragq, q, ipqe_q);
1157 pool_put(&ipqent_pool, q);
1158 fp->ipq_nfrags--;
1159 ip_nfrags--;
1160 }
1161
1162 insert:
1163 /*
1164 * Stick new segment in its place;
1165 * check for complete reassembly.
1166 */
1167 if (p == NULL) {
1168 TAILQ_INSERT_HEAD(&fp->ipq_fragq, ipqe, ipqe_q);
1169 } else {
1170 TAILQ_INSERT_AFTER(&fp->ipq_fragq, p, ipqe, ipqe_q);
1171 }
1172 next = 0;
1173 for (p = NULL, q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL;
1174 p = q, q = TAILQ_NEXT(q, ipqe_q)) {
1175 if (ntohs(q->ipqe_ip->ip_off) != next)
1176 return (0);
1177 next += ntohs(q->ipqe_ip->ip_len);
1178 }
1179 if (p->ipqe_mff)
1180 return (0);
1181
1182 /*
1183 * Reassembly is complete. Check for a bogus message size and
1184 * concatenate fragments.
1185 */
1186 q = TAILQ_FIRST(&fp->ipq_fragq);
1187 ip = q->ipqe_ip;
1188 if ((next + (ip->ip_hl << 2)) > IP_MAXPACKET) {
1189 ipstat.ips_toolong++;
1190 ip_freef(fp);
1191 return (0);
1192 }
1193 m = q->ipqe_m;
1194 t = m->m_next;
1195 m->m_next = 0;
1196 m_cat(m, t);
1197 nq = TAILQ_NEXT(q, ipqe_q);
1198 pool_put(&ipqent_pool, q);
1199 for (q = nq; q != NULL; q = nq) {
1200 t = q->ipqe_m;
1201 nq = TAILQ_NEXT(q, ipqe_q);
1202 pool_put(&ipqent_pool, q);
1203 m_cat(m, t);
1204 }
1205 ip_nfrags -= fp->ipq_nfrags;
1206
1207 /*
1208 * Create header for new ip packet by
1209 * modifying header of first packet;
1210 * dequeue and discard fragment reassembly header.
1211 * Make header visible.
1212 */
1213 ip->ip_len = htons(next);
1214 ip->ip_src = fp->ipq_src;
1215 ip->ip_dst = fp->ipq_dst;
1216 LIST_REMOVE(fp, ipq_q);
1217 FREE(fp, M_FTABLE);
1218 ip_nfragpackets--;
1219 m->m_len += (ip->ip_hl << 2);
1220 m->m_data -= (ip->ip_hl << 2);
1221 /* some debugging cruft by sklower, below, will go away soon */
1222 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */
1223 int plen = 0;
1224 for (t = m; t; t = t->m_next)
1225 plen += t->m_len;
1226 m->m_pkthdr.len = plen;
1227 m->m_pkthdr.csum_flags = 0;
1228 }
1229 return (m);
1230
1231 dropfrag:
1232 if (fp != 0)
1233 fp->ipq_nfrags--;
1234 ip_nfrags--;
1235 ipstat.ips_fragdropped++;
1236 m_freem(m);
1237 pool_put(&ipqent_pool, ipqe);
1238 return (0);
1239 }
1240
1241 /*
1242 * Free a fragment reassembly header and all
1243 * associated datagrams.
1244 */
1245 void
1246 ip_freef(struct ipq *fp)
1247 {
1248 struct ipqent *q, *p;
1249 u_int nfrags = 0;
1250
1251 IPQ_LOCK_CHECK();
1252
1253 for (q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL; q = p) {
1254 p = TAILQ_NEXT(q, ipqe_q);
1255 m_freem(q->ipqe_m);
1256 nfrags++;
1257 TAILQ_REMOVE(&fp->ipq_fragq, q, ipqe_q);
1258 pool_put(&ipqent_pool, q);
1259 }
1260
1261 if (nfrags != fp->ipq_nfrags)
1262 printf("ip_freef: nfrags %d != %d\n", fp->ipq_nfrags, nfrags);
1263 ip_nfrags -= nfrags;
1264 LIST_REMOVE(fp, ipq_q);
1265 FREE(fp, M_FTABLE);
1266 ip_nfragpackets--;
1267 }
1268
1269 /*
1270 * IP reassembly TTL machinery for multiplicative drop.
1271 */
1272 static u_int fragttl_histo[(IPFRAGTTL+1)];
1273
1274
1275 /*
1276 * Decrement TTL of all reasembly queue entries by `ticks'.
1277 * Count number of distinct fragments (as opposed to partial, fragmented
1278 * datagrams) in the reassembly queue. While we traverse the entire
1279 * reassembly queue, compute and return the median TTL over all fragments.
1280 */
1281 static u_int
1282 ip_reass_ttl_decr(u_int ticks)
1283 {
1284 u_int nfrags, median, dropfraction, keepfraction;
1285 struct ipq *fp, *nfp;
1286 int i;
1287
1288 nfrags = 0;
1289 memset(fragttl_histo, 0, sizeof fragttl_histo);
1290
1291 for (i = 0; i < IPREASS_NHASH; i++) {
1292 for (fp = LIST_FIRST(&ipq[i]); fp != NULL; fp = nfp) {
1293 fp->ipq_ttl = ((fp->ipq_ttl <= ticks) ?
1294 0 : fp->ipq_ttl - ticks);
1295 nfp = LIST_NEXT(fp, ipq_q);
1296 if (fp->ipq_ttl == 0) {
1297 ipstat.ips_fragtimeout++;
1298 ip_freef(fp);
1299 } else {
1300 nfrags += fp->ipq_nfrags;
1301 fragttl_histo[fp->ipq_ttl] += fp->ipq_nfrags;
1302 }
1303 }
1304 }
1305
1306 KASSERT(ip_nfrags == nfrags);
1307
1308 /* Find median (or other drop fraction) in histogram. */
1309 dropfraction = (ip_nfrags / 2);
1310 keepfraction = ip_nfrags - dropfraction;
1311 for (i = IPFRAGTTL, median = 0; i >= 0; i--) {
1312 median += fragttl_histo[i];
1313 if (median >= keepfraction)
1314 break;
1315 }
1316
1317 /* Return TTL of median (or other fraction). */
1318 return (u_int)i;
1319 }
1320
1321 void
1322 ip_reass_drophalf(void)
1323 {
1324
1325 u_int median_ticks;
1326 /*
1327 * Compute median TTL of all fragments, and count frags
1328 * with that TTL or lower (roughly half of all fragments).
1329 */
1330 median_ticks = ip_reass_ttl_decr(0);
1331
1332 /* Drop half. */
1333 median_ticks = ip_reass_ttl_decr(median_ticks);
1334
1335 }
1336
1337 /*
1338 * IP timer processing;
1339 * if a timer expires on a reassembly
1340 * queue, discard it.
1341 */
1342 void
1343 ip_slowtimo(void)
1344 {
1345 static u_int dropscanidx = 0;
1346 u_int i;
1347 u_int median_ttl;
1348 int s = splsoftnet();
1349
1350 IPQ_LOCK();
1351
1352 /* Age TTL of all fragments by 1 tick .*/
1353 median_ttl = ip_reass_ttl_decr(1);
1354
1355 /* make sure fragment limit is up-to-date */
1356 CHECK_NMBCLUSTER_PARAMS();
1357
1358 /* If we have too many fragments, drop the older half. */
1359 if (ip_nfrags > ip_maxfrags)
1360 ip_reass_ttl_decr(median_ttl);
1361
1362 /*
1363 * If we are over the maximum number of fragmented packets
1364 * (due to the limit being lowered), drain off
1365 * enough to get down to the new limit. Start draining
1366 * from the reassembly hashqueue most recently drained.
1367 */
1368 if (ip_maxfragpackets < 0)
1369 ;
1370 else {
1371 int wrapped = 0;
1372
1373 i = dropscanidx;
1374 while (ip_nfragpackets > ip_maxfragpackets && wrapped == 0) {
1375 while (LIST_FIRST(&ipq[i]) != NULL)
1376 ip_freef(LIST_FIRST(&ipq[i]));
1377 if (++i >= IPREASS_NHASH) {
1378 i = 0;
1379 }
1380 /*
1381 * Dont scan forever even if fragment counters are
1382 * wrong: stop after scanning entire reassembly queue.
1383 */
1384 if (i == dropscanidx)
1385 wrapped = 1;
1386 }
1387 dropscanidx = i;
1388 }
1389 IPQ_UNLOCK();
1390 #ifdef GATEWAY
1391 ipflow_slowtimo();
1392 #endif
1393 splx(s);
1394 }
1395
1396 /*
1397 * Drain off all datagram fragments.
1398 */
1399 void
1400 ip_drain(void)
1401 {
1402
1403 /*
1404 * We may be called from a device's interrupt context. If
1405 * the ipq is already busy, just bail out now.
1406 */
1407 if (ipq_lock_try() == 0)
1408 return;
1409
1410 /*
1411 * Drop half the total fragments now. If more mbufs are needed,
1412 * we will be called again soon.
1413 */
1414 ip_reass_drophalf();
1415
1416 IPQ_UNLOCK();
1417 }
1418
1419 /*
1420 * Do option processing on a datagram,
1421 * possibly discarding it if bad options are encountered,
1422 * or forwarding it if source-routed.
1423 * Returns 1 if packet has been forwarded/freed,
1424 * 0 if the packet should be processed further.
1425 */
1426 int
1427 ip_dooptions(struct mbuf *m)
1428 {
1429 struct ip *ip = mtod(m, struct ip *);
1430 u_char *cp, *cp0;
1431 struct ip_timestamp *ipt;
1432 struct in_ifaddr *ia;
1433 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0;
1434 struct in_addr dst;
1435 n_time ntime;
1436
1437 dst = ip->ip_dst;
1438 cp = (u_char *)(ip + 1);
1439 cnt = (ip->ip_hl << 2) - sizeof (struct ip);
1440 for (; cnt > 0; cnt -= optlen, cp += optlen) {
1441 opt = cp[IPOPT_OPTVAL];
1442 if (opt == IPOPT_EOL)
1443 break;
1444 if (opt == IPOPT_NOP)
1445 optlen = 1;
1446 else {
1447 if (cnt < IPOPT_OLEN + sizeof(*cp)) {
1448 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1449 goto bad;
1450 }
1451 optlen = cp[IPOPT_OLEN];
1452 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
1453 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1454 goto bad;
1455 }
1456 }
1457 switch (opt) {
1458
1459 default:
1460 break;
1461
1462 /*
1463 * Source routing with record.
1464 * Find interface with current destination address.
1465 * If none on this machine then drop if strictly routed,
1466 * or do nothing if loosely routed.
1467 * Record interface address and bring up next address
1468 * component. If strictly routed make sure next
1469 * address is on directly accessible net.
1470 */
1471 case IPOPT_LSRR:
1472 case IPOPT_SSRR:
1473 if (ip_allowsrcrt == 0) {
1474 type = ICMP_UNREACH;
1475 code = ICMP_UNREACH_NET_PROHIB;
1476 goto bad;
1477 }
1478 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1479 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1480 goto bad;
1481 }
1482 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1483 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1484 goto bad;
1485 }
1486 ipaddr.sin_addr = ip->ip_dst;
1487 ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr)));
1488 if (ia == 0) {
1489 if (opt == IPOPT_SSRR) {
1490 type = ICMP_UNREACH;
1491 code = ICMP_UNREACH_SRCFAIL;
1492 goto bad;
1493 }
1494 /*
1495 * Loose routing, and not at next destination
1496 * yet; nothing to do except forward.
1497 */
1498 break;
1499 }
1500 off--; /* 0 origin */
1501 if ((off + sizeof(struct in_addr)) > optlen) {
1502 /*
1503 * End of source route. Should be for us.
1504 */
1505 save_rte(cp, ip->ip_src);
1506 break;
1507 }
1508 /*
1509 * locate outgoing interface
1510 */
1511 bcopy((caddr_t)(cp + off), (caddr_t)&ipaddr.sin_addr,
1512 sizeof(ipaddr.sin_addr));
1513 if (opt == IPOPT_SSRR)
1514 ia = ifatoia(ifa_ifwithladdr(sintosa(&ipaddr)));
1515 else
1516 ia = ip_rtaddr(ipaddr.sin_addr);
1517 if (ia == 0) {
1518 type = ICMP_UNREACH;
1519 code = ICMP_UNREACH_SRCFAIL;
1520 goto bad;
1521 }
1522 ip->ip_dst = ipaddr.sin_addr;
1523 bcopy((caddr_t)&ia->ia_addr.sin_addr,
1524 (caddr_t)(cp + off), sizeof(struct in_addr));
1525 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1526 /*
1527 * Let ip_intr's mcast routing check handle mcast pkts
1528 */
1529 forward = !IN_MULTICAST(ip->ip_dst.s_addr);
1530 break;
1531
1532 case IPOPT_RR:
1533 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1534 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1535 goto bad;
1536 }
1537 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1538 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1539 goto bad;
1540 }
1541 /*
1542 * If no space remains, ignore.
1543 */
1544 off--; /* 0 origin */
1545 if ((off + sizeof(struct in_addr)) > optlen)
1546 break;
1547 bcopy((caddr_t)(&ip->ip_dst), (caddr_t)&ipaddr.sin_addr,
1548 sizeof(ipaddr.sin_addr));
1549 /*
1550 * locate outgoing interface; if we're the destination,
1551 * use the incoming interface (should be same).
1552 */
1553 if ((ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))))
1554 == NULL &&
1555 (ia = ip_rtaddr(ipaddr.sin_addr)) == NULL) {
1556 type = ICMP_UNREACH;
1557 code = ICMP_UNREACH_HOST;
1558 goto bad;
1559 }
1560 bcopy((caddr_t)&ia->ia_addr.sin_addr,
1561 (caddr_t)(cp + off), sizeof(struct in_addr));
1562 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1563 break;
1564
1565 case IPOPT_TS:
1566 code = cp - (u_char *)ip;
1567 ipt = (struct ip_timestamp *)cp;
1568 if (ipt->ipt_len < 4 || ipt->ipt_len > 40) {
1569 code = (u_char *)&ipt->ipt_len - (u_char *)ip;
1570 goto bad;
1571 }
1572 if (ipt->ipt_ptr < 5) {
1573 code = (u_char *)&ipt->ipt_ptr - (u_char *)ip;
1574 goto bad;
1575 }
1576 if (ipt->ipt_ptr > ipt->ipt_len - sizeof (int32_t)) {
1577 if (++ipt->ipt_oflw == 0) {
1578 code = (u_char *)&ipt->ipt_ptr -
1579 (u_char *)ip;
1580 goto bad;
1581 }
1582 break;
1583 }
1584 cp0 = (cp + ipt->ipt_ptr - 1);
1585 switch (ipt->ipt_flg) {
1586
1587 case IPOPT_TS_TSONLY:
1588 break;
1589
1590 case IPOPT_TS_TSANDADDR:
1591 if (ipt->ipt_ptr - 1 + sizeof(n_time) +
1592 sizeof(struct in_addr) > ipt->ipt_len) {
1593 code = (u_char *)&ipt->ipt_ptr -
1594 (u_char *)ip;
1595 goto bad;
1596 }
1597 ipaddr.sin_addr = dst;
1598 ia = ifatoia(ifaof_ifpforaddr(sintosa(&ipaddr),
1599 m->m_pkthdr.rcvif));
1600 if (ia == 0)
1601 continue;
1602 bcopy(&ia->ia_addr.sin_addr,
1603 cp0, sizeof(struct in_addr));
1604 ipt->ipt_ptr += sizeof(struct in_addr);
1605 break;
1606
1607 case IPOPT_TS_PRESPEC:
1608 if (ipt->ipt_ptr - 1 + sizeof(n_time) +
1609 sizeof(struct in_addr) > ipt->ipt_len) {
1610 code = (u_char *)&ipt->ipt_ptr -
1611 (u_char *)ip;
1612 goto bad;
1613 }
1614 bcopy(cp0, &ipaddr.sin_addr,
1615 sizeof(struct in_addr));
1616 if (ifatoia(ifa_ifwithaddr(sintosa(&ipaddr)))
1617 == NULL)
1618 continue;
1619 ipt->ipt_ptr += sizeof(struct in_addr);
1620 break;
1621
1622 default:
1623 /* XXX can't take &ipt->ipt_flg */
1624 code = (u_char *)&ipt->ipt_ptr -
1625 (u_char *)ip + 1;
1626 goto bad;
1627 }
1628 ntime = iptime();
1629 cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */
1630 bcopy(cp0, (caddr_t)cp + ipt->ipt_ptr - 1,
1631 sizeof(n_time));
1632 ipt->ipt_ptr += sizeof(n_time);
1633 }
1634 }
1635 if (forward) {
1636 if (ip_forwsrcrt == 0) {
1637 type = ICMP_UNREACH;
1638 code = ICMP_UNREACH_SRCFAIL;
1639 goto bad;
1640 }
1641 ip_forward(m, 1);
1642 return (1);
1643 }
1644 return (0);
1645 bad:
1646 icmp_error(m, type, code, 0, 0);
1647 ipstat.ips_badoptions++;
1648 return (1);
1649 }
1650
1651 /*
1652 * Given address of next destination (final or next hop),
1653 * return internet address info of interface to be used to get there.
1654 */
1655 struct in_ifaddr *
1656 ip_rtaddr(struct in_addr dst)
1657 {
1658 struct sockaddr_in *sin;
1659
1660 sin = satosin(&ipforward_rt.ro_dst);
1661
1662 if (ipforward_rt.ro_rt == 0 || !in_hosteq(dst, sin->sin_addr)) {
1663 if (ipforward_rt.ro_rt) {
1664 RTFREE(ipforward_rt.ro_rt);
1665 ipforward_rt.ro_rt = 0;
1666 }
1667 sin->sin_family = AF_INET;
1668 sin->sin_len = sizeof(*sin);
1669 sin->sin_addr = dst;
1670
1671 rtalloc(&ipforward_rt);
1672 }
1673 if (ipforward_rt.ro_rt == 0)
1674 return ((struct in_ifaddr *)0);
1675 return (ifatoia(ipforward_rt.ro_rt->rt_ifa));
1676 }
1677
1678 /*
1679 * Save incoming source route for use in replies,
1680 * to be picked up later by ip_srcroute if the receiver is interested.
1681 */
1682 void
1683 save_rte(u_char *option, struct in_addr dst)
1684 {
1685 unsigned olen;
1686
1687 olen = option[IPOPT_OLEN];
1688 #ifdef DIAGNOSTIC
1689 if (ipprintfs)
1690 printf("save_rte: olen %d\n", olen);
1691 #endif /* 0 */
1692 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst)))
1693 return;
1694 bcopy((caddr_t)option, (caddr_t)ip_srcrt.srcopt, olen);
1695 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1696 ip_srcrt.dst = dst;
1697 }
1698
1699 /*
1700 * Retrieve incoming source route for use in replies,
1701 * in the same form used by setsockopt.
1702 * The first hop is placed before the options, will be removed later.
1703 */
1704 struct mbuf *
1705 ip_srcroute(void)
1706 {
1707 struct in_addr *p, *q;
1708 struct mbuf *m;
1709
1710 if (ip_nhops == 0)
1711 return ((struct mbuf *)0);
1712 m = m_get(M_DONTWAIT, MT_SOOPTS);
1713 if (m == 0)
1714 return ((struct mbuf *)0);
1715
1716 MCLAIM(m, &inetdomain.dom_mowner);
1717 #define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt))
1718
1719 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
1720 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) +
1721 OPTSIZ;
1722 #ifdef DIAGNOSTIC
1723 if (ipprintfs)
1724 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len);
1725 #endif
1726
1727 /*
1728 * First save first hop for return route
1729 */
1730 p = &ip_srcrt.route[ip_nhops - 1];
1731 *(mtod(m, struct in_addr *)) = *p--;
1732 #ifdef DIAGNOSTIC
1733 if (ipprintfs)
1734 printf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr));
1735 #endif
1736
1737 /*
1738 * Copy option fields and padding (nop) to mbuf.
1739 */
1740 ip_srcrt.nop = IPOPT_NOP;
1741 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
1742 bcopy((caddr_t)&ip_srcrt.nop,
1743 mtod(m, caddr_t) + sizeof(struct in_addr), OPTSIZ);
1744 q = (struct in_addr *)(mtod(m, caddr_t) +
1745 sizeof(struct in_addr) + OPTSIZ);
1746 #undef OPTSIZ
1747 /*
1748 * Record return path as an IP source route,
1749 * reversing the path (pointers are now aligned).
1750 */
1751 while (p >= ip_srcrt.route) {
1752 #ifdef DIAGNOSTIC
1753 if (ipprintfs)
1754 printf(" %x", ntohl(q->s_addr));
1755 #endif
1756 *q++ = *p--;
1757 }
1758 /*
1759 * Last hop goes to final destination.
1760 */
1761 *q = ip_srcrt.dst;
1762 #ifdef DIAGNOSTIC
1763 if (ipprintfs)
1764 printf(" %x\n", ntohl(q->s_addr));
1765 #endif
1766 return (m);
1767 }
1768
1769 /*
1770 * Strip out IP options, at higher
1771 * level protocol in the kernel.
1772 * Second argument is buffer to which options
1773 * will be moved, and return value is their length.
1774 * XXX should be deleted; last arg currently ignored.
1775 */
1776 void
1777 ip_stripoptions(struct mbuf *m, struct mbuf *mopt)
1778 {
1779 int i;
1780 struct ip *ip = mtod(m, struct ip *);
1781 caddr_t opts;
1782 int olen;
1783
1784 olen = (ip->ip_hl << 2) - sizeof (struct ip);
1785 opts = (caddr_t)(ip + 1);
1786 i = m->m_len - (sizeof (struct ip) + olen);
1787 bcopy(opts + olen, opts, (unsigned)i);
1788 m->m_len -= olen;
1789 if (m->m_flags & M_PKTHDR)
1790 m->m_pkthdr.len -= olen;
1791 ip->ip_len = htons(ntohs(ip->ip_len) - olen);
1792 ip->ip_hl = sizeof (struct ip) >> 2;
1793 }
1794
1795 const int inetctlerrmap[PRC_NCMDS] = {
1796 0, 0, 0, 0,
1797 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
1798 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
1799 EMSGSIZE, EHOSTUNREACH, 0, 0,
1800 0, 0, 0, 0,
1801 ENOPROTOOPT
1802 };
1803
1804 /*
1805 * Forward a packet. If some error occurs return the sender
1806 * an icmp packet. Note we can't always generate a meaningful
1807 * icmp message because icmp doesn't have a large enough repertoire
1808 * of codes and types.
1809 *
1810 * If not forwarding, just drop the packet. This could be confusing
1811 * if ipforwarding was zero but some routing protocol was advancing
1812 * us as a gateway to somewhere. However, we must let the routing
1813 * protocol deal with that.
1814 *
1815 * The srcrt parameter indicates whether the packet is being forwarded
1816 * via a source route.
1817 */
1818 void
1819 ip_forward(struct mbuf *m, int srcrt)
1820 {
1821 struct ip *ip = mtod(m, struct ip *);
1822 struct sockaddr_in *sin;
1823 struct rtentry *rt;
1824 int error, type = 0, code = 0;
1825 struct mbuf *mcopy;
1826 n_long dest;
1827 struct ifnet *destifp;
1828 #if defined(IPSEC) || defined(FAST_IPSEC)
1829 struct ifnet dummyifp;
1830 #endif
1831
1832 /*
1833 * We are now in the output path.
1834 */
1835 MCLAIM(m, &ip_tx_mowner);
1836
1837 /*
1838 * Clear any in-bound checksum flags for this packet.
1839 */
1840 m->m_pkthdr.csum_flags = 0;
1841
1842 dest = 0;
1843 #ifdef DIAGNOSTIC
1844 if (ipprintfs)
1845 printf("forward: src %2.2x dst %2.2x ttl %x\n",
1846 ntohl(ip->ip_src.s_addr),
1847 ntohl(ip->ip_dst.s_addr), ip->ip_ttl);
1848 #endif
1849 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
1850 ipstat.ips_cantforward++;
1851 m_freem(m);
1852 return;
1853 }
1854 if (ip->ip_ttl <= IPTTLDEC) {
1855 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0);
1856 return;
1857 }
1858 ip->ip_ttl -= IPTTLDEC;
1859
1860 sin = satosin(&ipforward_rt.ro_dst);
1861 if ((rt = ipforward_rt.ro_rt) == 0 ||
1862 !in_hosteq(ip->ip_dst, sin->sin_addr)) {
1863 if (ipforward_rt.ro_rt) {
1864 RTFREE(ipforward_rt.ro_rt);
1865 ipforward_rt.ro_rt = 0;
1866 }
1867 sin->sin_family = AF_INET;
1868 sin->sin_len = sizeof(struct sockaddr_in);
1869 sin->sin_addr = ip->ip_dst;
1870
1871 rtalloc(&ipforward_rt);
1872 if (ipforward_rt.ro_rt == 0) {
1873 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0);
1874 return;
1875 }
1876 rt = ipforward_rt.ro_rt;
1877 }
1878
1879 /*
1880 * Save at most 68 bytes of the packet in case
1881 * we need to generate an ICMP message to the src.
1882 * Pullup to avoid sharing mbuf cluster between m and mcopy.
1883 */
1884 mcopy = m_copym(m, 0, imin(ntohs(ip->ip_len), 68), M_DONTWAIT);
1885 if (mcopy)
1886 mcopy = m_pullup(mcopy, ip->ip_hl << 2);
1887
1888 /*
1889 * If forwarding packet using same interface that it came in on,
1890 * perhaps should send a redirect to sender to shortcut a hop.
1891 * Only send redirect if source is sending directly to us,
1892 * and if packet was not source routed (or has any options).
1893 * Also, don't send redirect if forwarding using a default route
1894 * or a route modified by a redirect.
1895 */
1896 if (rt->rt_ifp == m->m_pkthdr.rcvif &&
1897 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1898 !in_nullhost(satosin(rt_key(rt))->sin_addr) &&
1899 ipsendredirects && !srcrt) {
1900 if (rt->rt_ifa &&
1901 (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) ==
1902 ifatoia(rt->rt_ifa)->ia_subnet) {
1903 if (rt->rt_flags & RTF_GATEWAY)
1904 dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
1905 else
1906 dest = ip->ip_dst.s_addr;
1907 /*
1908 * Router requirements says to only send host
1909 * redirects.
1910 */
1911 type = ICMP_REDIRECT;
1912 code = ICMP_REDIRECT_HOST;
1913 #ifdef DIAGNOSTIC
1914 if (ipprintfs)
1915 printf("redirect (%d) to %x\n", code,
1916 (u_int32_t)dest);
1917 #endif
1918 }
1919 }
1920
1921 error = ip_output(m, (struct mbuf *)0, &ipforward_rt,
1922 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)),
1923 (struct ip_moptions *)NULL, (struct socket *)NULL);
1924
1925 if (error)
1926 ipstat.ips_cantforward++;
1927 else {
1928 ipstat.ips_forward++;
1929 if (type)
1930 ipstat.ips_redirectsent++;
1931 else {
1932 if (mcopy) {
1933 #ifdef GATEWAY
1934 if (mcopy->m_flags & M_CANFASTFWD)
1935 ipflow_create(&ipforward_rt, mcopy);
1936 #endif
1937 m_freem(mcopy);
1938 }
1939 return;
1940 }
1941 }
1942 if (mcopy == NULL)
1943 return;
1944 destifp = NULL;
1945
1946 switch (error) {
1947
1948 case 0: /* forwarded, but need redirect */
1949 /* type, code set above */
1950 break;
1951
1952 case ENETUNREACH: /* shouldn't happen, checked above */
1953 case EHOSTUNREACH:
1954 case ENETDOWN:
1955 case EHOSTDOWN:
1956 default:
1957 type = ICMP_UNREACH;
1958 code = ICMP_UNREACH_HOST;
1959 break;
1960
1961 case EMSGSIZE:
1962 type = ICMP_UNREACH;
1963 code = ICMP_UNREACH_NEEDFRAG;
1964 #if !defined(IPSEC) && !defined(FAST_IPSEC)
1965 if (ipforward_rt.ro_rt)
1966 destifp = ipforward_rt.ro_rt->rt_ifp;
1967 #else
1968 /*
1969 * If the packet is routed over IPsec tunnel, tell the
1970 * originator the tunnel MTU.
1971 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
1972 * XXX quickhack!!!
1973 */
1974 if (ipforward_rt.ro_rt) {
1975 struct secpolicy *sp;
1976 int ipsecerror;
1977 size_t ipsechdr;
1978 struct route *ro;
1979
1980 sp = ipsec4_getpolicybyaddr(mcopy,
1981 IPSEC_DIR_OUTBOUND, IP_FORWARDING,
1982 &ipsecerror);
1983
1984 if (sp == NULL)
1985 destifp = ipforward_rt.ro_rt->rt_ifp;
1986 else {
1987 /* count IPsec header size */
1988 ipsechdr = ipsec4_hdrsiz(mcopy,
1989 IPSEC_DIR_OUTBOUND, NULL);
1990
1991 /*
1992 * find the correct route for outer IPv4
1993 * header, compute tunnel MTU.
1994 *
1995 * XXX BUG ALERT
1996 * The "dummyifp" code relies upon the fact
1997 * that icmp_error() touches only ifp->if_mtu.
1998 */
1999 /*XXX*/
2000 destifp = NULL;
2001 if (sp->req != NULL
2002 && sp->req->sav != NULL
2003 && sp->req->sav->sah != NULL) {
2004 ro = &sp->req->sav->sah->sa_route;
2005 if (ro->ro_rt && ro->ro_rt->rt_ifp) {
2006 dummyifp.if_mtu =
2007 ro->ro_rt->rt_rmx.rmx_mtu ?
2008 ro->ro_rt->rt_rmx.rmx_mtu :
2009 ro->ro_rt->rt_ifp->if_mtu;
2010 dummyifp.if_mtu -= ipsechdr;
2011 destifp = &dummyifp;
2012 }
2013 }
2014
2015 #ifdef IPSEC
2016 key_freesp(sp);
2017 #else
2018 KEY_FREESP(&sp);
2019 #endif
2020 }
2021 }
2022 #endif /*IPSEC*/
2023 ipstat.ips_cantfrag++;
2024 break;
2025
2026 case ENOBUFS:
2027 #if 1
2028 /*
2029 * a router should not generate ICMP_SOURCEQUENCH as
2030 * required in RFC1812 Requirements for IP Version 4 Routers.
2031 * source quench could be a big problem under DoS attacks,
2032 * or if the underlying interface is rate-limited.
2033 */
2034 if (mcopy)
2035 m_freem(mcopy);
2036 return;
2037 #else
2038 type = ICMP_SOURCEQUENCH;
2039 code = 0;
2040 break;
2041 #endif
2042 }
2043 icmp_error(mcopy, type, code, dest, destifp);
2044 }
2045
2046 void
2047 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
2048 struct mbuf *m)
2049 {
2050
2051 if (inp->inp_socket->so_options & SO_TIMESTAMP) {
2052 struct timeval tv;
2053
2054 microtime(&tv);
2055 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
2056 SCM_TIMESTAMP, SOL_SOCKET);
2057 if (*mp)
2058 mp = &(*mp)->m_next;
2059 }
2060 if (inp->inp_flags & INP_RECVDSTADDR) {
2061 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
2062 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
2063 if (*mp)
2064 mp = &(*mp)->m_next;
2065 }
2066 #ifdef notyet
2067 /*
2068 * XXX
2069 * Moving these out of udp_input() made them even more broken
2070 * than they already were.
2071 * - fenner@parc.xerox.com
2072 */
2073 /* options were tossed already */
2074 if (inp->inp_flags & INP_RECVOPTS) {
2075 *mp = sbcreatecontrol((caddr_t) opts_deleted_above,
2076 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
2077 if (*mp)
2078 mp = &(*mp)->m_next;
2079 }
2080 /* ip_srcroute doesn't do what we want here, need to fix */
2081 if (inp->inp_flags & INP_RECVRETOPTS) {
2082 *mp = sbcreatecontrol((caddr_t) ip_srcroute(),
2083 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
2084 if (*mp)
2085 mp = &(*mp)->m_next;
2086 }
2087 #endif
2088 if (inp->inp_flags & INP_RECVIF) {
2089 struct sockaddr_dl sdl;
2090
2091 sdl.sdl_len = offsetof(struct sockaddr_dl, sdl_data[0]);
2092 sdl.sdl_family = AF_LINK;
2093 sdl.sdl_index = m->m_pkthdr.rcvif ?
2094 m->m_pkthdr.rcvif->if_index : 0;
2095 sdl.sdl_nlen = sdl.sdl_alen = sdl.sdl_slen = 0;
2096 *mp = sbcreatecontrol((caddr_t) &sdl, sdl.sdl_len,
2097 IP_RECVIF, IPPROTO_IP);
2098 if (*mp)
2099 mp = &(*mp)->m_next;
2100 }
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. apparently if
2131 * maxflows is even looked up, we "reap flows".
2132 */
2133 static int
2134 sysctl_net_inet_ip_maxflows(SYSCTLFN_ARGS)
2135 {
2136 int s;
2137
2138 s = sysctl_lookup(SYSCTLFN_CALL(rnode));
2139 if (s)
2140 return (s);
2141
2142 s = splsoftnet();
2143 ipflow_reap(0);
2144 splx(s);
2145
2146 return (0);
2147 }
2148 #endif /* GATEWAY */
2149
2150
2151 SYSCTL_SETUP(sysctl_net_inet_ip_setup, "sysctl net.inet.ip subtree setup")
2152 {
2153 extern int subnetsarelocal, hostzeroisbroadcast;
2154
2155 sysctl_createv(clog, 0, NULL, NULL,
2156 CTLFLAG_PERMANENT,
2157 CTLTYPE_NODE, "net", NULL,
2158 NULL, 0, NULL, 0,
2159 CTL_NET, CTL_EOL);
2160 sysctl_createv(clog, 0, NULL, NULL,
2161 CTLFLAG_PERMANENT,
2162 CTLTYPE_NODE, "inet",
2163 SYSCTL_DESCR("PF_INET related settings"),
2164 NULL, 0, NULL, 0,
2165 CTL_NET, PF_INET, CTL_EOL);
2166 sysctl_createv(clog, 0, NULL, NULL,
2167 CTLFLAG_PERMANENT,
2168 CTLTYPE_NODE, "ip",
2169 SYSCTL_DESCR("IPv4 related settings"),
2170 NULL, 0, NULL, 0,
2171 CTL_NET, PF_INET, IPPROTO_IP, CTL_EOL);
2172
2173 sysctl_createv(clog, 0, NULL, NULL,
2174 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2175 CTLTYPE_INT, "forwarding",
2176 SYSCTL_DESCR("Enable forwarding of INET datagrams"),
2177 NULL, 0, &ipforwarding, 0,
2178 CTL_NET, PF_INET, IPPROTO_IP,
2179 IPCTL_FORWARDING, CTL_EOL);
2180 sysctl_createv(clog, 0, NULL, NULL,
2181 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2182 CTLTYPE_INT, "redirect",
2183 SYSCTL_DESCR("Enable sending of ICMP redirect messages"),
2184 NULL, 0, &ipsendredirects, 0,
2185 CTL_NET, PF_INET, IPPROTO_IP,
2186 IPCTL_SENDREDIRECTS, CTL_EOL);
2187 sysctl_createv(clog, 0, NULL, NULL,
2188 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2189 CTLTYPE_INT, "ttl",
2190 SYSCTL_DESCR("Default TTL for an INET datagram"),
2191 NULL, 0, &ip_defttl, 0,
2192 CTL_NET, PF_INET, IPPROTO_IP,
2193 IPCTL_DEFTTL, CTL_EOL);
2194 #ifdef IPCTL_DEFMTU
2195 sysctl_createv(clog, 0, NULL, NULL,
2196 CTLFLAG_PERMANENT /* |CTLFLAG_READWRITE? */,
2197 CTLTYPE_INT, "mtu",
2198 SYSCTL_DESCR("Default MTA for an INET route"),
2199 NULL, 0, &ip_mtu, 0,
2200 CTL_NET, PF_INET, IPPROTO_IP,
2201 IPCTL_DEFMTU, CTL_EOL);
2202 #endif /* IPCTL_DEFMTU */
2203 sysctl_createv(clog, 0, NULL, NULL,
2204 CTLFLAG_PERMANENT|CTLFLAG_READONLY1,
2205 CTLTYPE_INT, "forwsrcrt",
2206 SYSCTL_DESCR("Enable forwarding of source-routed "
2207 "datagrams"),
2208 NULL, 0, &ip_forwsrcrt, 0,
2209 CTL_NET, PF_INET, IPPROTO_IP,
2210 IPCTL_FORWSRCRT, CTL_EOL);
2211 sysctl_createv(clog, 0, NULL, NULL,
2212 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2213 CTLTYPE_INT, "directed-broadcast",
2214 SYSCTL_DESCR("Enable forwarding of broadcast datagrams"),
2215 NULL, 0, &ip_directedbcast, 0,
2216 CTL_NET, PF_INET, IPPROTO_IP,
2217 IPCTL_DIRECTEDBCAST, CTL_EOL);
2218 sysctl_createv(clog, 0, NULL, NULL,
2219 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2220 CTLTYPE_INT, "allowsrcrt",
2221 SYSCTL_DESCR("Accept source-routed datagrams"),
2222 NULL, 0, &ip_allowsrcrt, 0,
2223 CTL_NET, PF_INET, IPPROTO_IP,
2224 IPCTL_ALLOWSRCRT, CTL_EOL);
2225 sysctl_createv(clog, 0, NULL, NULL,
2226 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2227 CTLTYPE_INT, "subnetsarelocal",
2228 SYSCTL_DESCR("Whether logical subnets are considered "
2229 "local"),
2230 NULL, 0, &subnetsarelocal, 0,
2231 CTL_NET, PF_INET, IPPROTO_IP,
2232 IPCTL_SUBNETSARELOCAL, CTL_EOL);
2233 sysctl_createv(clog, 0, NULL, NULL,
2234 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2235 CTLTYPE_INT, "mtudisc",
2236 SYSCTL_DESCR("Use RFC1191 Path MTU Discovery"),
2237 NULL, 0, &ip_mtudisc, 0,
2238 CTL_NET, PF_INET, IPPROTO_IP,
2239 IPCTL_MTUDISC, CTL_EOL);
2240 sysctl_createv(clog, 0, NULL, NULL,
2241 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2242 CTLTYPE_INT, "anonportmin",
2243 SYSCTL_DESCR("Lowest ephemeral port number to assign"),
2244 sysctl_net_inet_ip_ports, 0, &anonportmin, 0,
2245 CTL_NET, PF_INET, IPPROTO_IP,
2246 IPCTL_ANONPORTMIN, CTL_EOL);
2247 sysctl_createv(clog, 0, NULL, NULL,
2248 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2249 CTLTYPE_INT, "anonportmax",
2250 SYSCTL_DESCR("Highest ephemeral port number to assign"),
2251 sysctl_net_inet_ip_ports, 0, &anonportmax, 0,
2252 CTL_NET, PF_INET, IPPROTO_IP,
2253 IPCTL_ANONPORTMAX, CTL_EOL);
2254 sysctl_createv(clog, 0, NULL, NULL,
2255 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2256 CTLTYPE_INT, "mtudisctimeout",
2257 SYSCTL_DESCR("Lifetime of a Path MTU Discovered route"),
2258 sysctl_net_inet_ip_pmtudto, 0, &ip_mtudisc_timeout, 0,
2259 CTL_NET, PF_INET, IPPROTO_IP,
2260 IPCTL_MTUDISCTIMEOUT, CTL_EOL);
2261 #ifdef GATEWAY
2262 sysctl_createv(clog, 0, NULL, NULL,
2263 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2264 CTLTYPE_INT, "maxflows",
2265 SYSCTL_DESCR("Number of flows for fast forwarding"),
2266 sysctl_net_inet_ip_maxflows, 0, &ip_maxflows, 0,
2267 CTL_NET, PF_INET, IPPROTO_IP,
2268 IPCTL_MAXFLOWS, CTL_EOL);
2269 #endif /* GATEWAY */
2270 sysctl_createv(clog, 0, NULL, NULL,
2271 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2272 CTLTYPE_INT, "hostzerobroadcast",
2273 SYSCTL_DESCR("All zeroes address is broadcast address"),
2274 NULL, 0, &hostzeroisbroadcast, 0,
2275 CTL_NET, PF_INET, IPPROTO_IP,
2276 IPCTL_HOSTZEROBROADCAST, CTL_EOL);
2277 #if NGIF > 0
2278 sysctl_createv(clog, 0, NULL, NULL,
2279 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2280 CTLTYPE_INT, "gifttl",
2281 SYSCTL_DESCR("Default TTL for a gif tunnel datagram"),
2282 NULL, 0, &ip_gif_ttl, 0,
2283 CTL_NET, PF_INET, IPPROTO_IP,
2284 IPCTL_GIF_TTL, CTL_EOL);
2285 #endif /* NGIF */
2286 #ifndef IPNOPRIVPORTS
2287 sysctl_createv(clog, 0, NULL, NULL,
2288 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2289 CTLTYPE_INT, "lowportmin",
2290 SYSCTL_DESCR("Lowest privileged ephemeral port number "
2291 "to assign"),
2292 sysctl_net_inet_ip_ports, 0, &lowportmin, 0,
2293 CTL_NET, PF_INET, IPPROTO_IP,
2294 IPCTL_LOWPORTMIN, CTL_EOL);
2295 sysctl_createv(clog, 0, NULL, NULL,
2296 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2297 CTLTYPE_INT, "lowportmax",
2298 SYSCTL_DESCR("Highest privileged ephemeral port number "
2299 "to assign"),
2300 sysctl_net_inet_ip_ports, 0, &lowportmax, 0,
2301 CTL_NET, PF_INET, IPPROTO_IP,
2302 IPCTL_LOWPORTMAX, CTL_EOL);
2303 #endif /* IPNOPRIVPORTS */
2304 sysctl_createv(clog, 0, NULL, NULL,
2305 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2306 CTLTYPE_INT, "maxfragpackets",
2307 SYSCTL_DESCR("Maximum number of fragments to retain for "
2308 "possible reassembly"),
2309 NULL, 0, &ip_maxfragpackets, 0,
2310 CTL_NET, PF_INET, IPPROTO_IP,
2311 IPCTL_MAXFRAGPACKETS, CTL_EOL);
2312 #if NGRE > 0
2313 sysctl_createv(clog, 0, NULL, NULL,
2314 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2315 CTLTYPE_INT, "grettl",
2316 SYSCTL_DESCR("Default TTL for a gre tunnel datagram"),
2317 NULL, 0, &ip_gre_ttl, 0,
2318 CTL_NET, PF_INET, IPPROTO_IP,
2319 IPCTL_GRE_TTL, CTL_EOL);
2320 #endif /* NGRE */
2321 sysctl_createv(clog, 0, NULL, NULL,
2322 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2323 CTLTYPE_INT, "checkinterface",
2324 SYSCTL_DESCR("Enable receive side of Strong ES model "
2325 "from RFC1122"),
2326 NULL, 0, &ip_checkinterface, 0,
2327 CTL_NET, PF_INET, IPPROTO_IP,
2328 IPCTL_CHECKINTERFACE, CTL_EOL);
2329 sysctl_createv(clog, 0, NULL, NULL,
2330 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2331 CTLTYPE_INT, "random_id",
2332 SYSCTL_DESCR("Assign random ip_id values"),
2333 NULL, 0, &ip_do_randomid, 0,
2334 CTL_NET, PF_INET, IPPROTO_IP,
2335 IPCTL_RANDOMID, CTL_EOL);
2336 sysctl_createv(clog, 0, NULL, NULL,
2337 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2338 CTLTYPE_INT, "do_loopback_cksum",
2339 SYSCTL_DESCR("Perform IP checksum on loopback"),
2340 NULL, 0, &ip_do_loopback_cksum, 0,
2341 CTL_NET, PF_INET, IPPROTO_IP,
2342 IPCTL_LOOPBACKCKSUM, CTL_EOL);
2343 }
Cache object: 84f111767a6bcc364d858eabec1e31a2
|