1 /* $NetBSD: ip_input.c,v 1.236.2.2 2008/03/30 15:27:49 jdc 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 1.236.2.2 2008/03/30 15:27:49 jdc 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 POOL_INIT(ipqent_pool, sizeof(struct ipqent), 0, 0, 0, "ipqepl", NULL);
339
340 #ifdef INET_CSUM_COUNTERS
341 #include <sys/device.h>
342
343 struct evcnt ip_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
344 NULL, "inet", "hwcsum bad");
345 struct evcnt ip_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
346 NULL, "inet", "hwcsum ok");
347 struct evcnt ip_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
348 NULL, "inet", "swcsum");
349
350 #define INET_CSUM_COUNTER_INCR(ev) (ev)->ev_count++
351
352 EVCNT_ATTACH_STATIC(ip_hwcsum_bad);
353 EVCNT_ATTACH_STATIC(ip_hwcsum_ok);
354 EVCNT_ATTACH_STATIC(ip_swcsum);
355
356 #else
357
358 #define INET_CSUM_COUNTER_INCR(ev) /* nothing */
359
360 #endif /* INET_CSUM_COUNTERS */
361
362 /*
363 * We need to save the IP options in case a protocol wants to respond
364 * to an incoming packet over the same route if the packet got here
365 * using IP source routing. This allows connection establishment and
366 * maintenance when the remote end is on a network that is not known
367 * to us.
368 */
369 int ip_nhops = 0;
370 static struct ip_srcrt {
371 struct in_addr dst; /* final destination */
372 char nop; /* one NOP to align */
373 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */
374 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)];
375 } ip_srcrt;
376
377 static void save_rte(u_char *, struct in_addr);
378
379 #ifdef MBUFTRACE
380 struct mowner ip_rx_mowner = MOWNER_INIT("internet", "rx");
381 struct mowner ip_tx_mowner = MOWNER_INIT("internet", "tx");
382 #endif
383
384 /*
385 * Compute IP limits derived from the value of nmbclusters.
386 */
387 static void
388 ip_nmbclusters_changed(void)
389 {
390 ip_maxfrags = nmbclusters / 4;
391 ip_nmbclusters = nmbclusters;
392 }
393
394 /*
395 * IP initialization: fill in IP protocol switch table.
396 * All protocols not implemented in kernel go to raw IP protocol handler.
397 */
398 void
399 ip_init(void)
400 {
401 const struct protosw *pr;
402 int i;
403
404 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
405 if (pr == 0)
406 panic("ip_init");
407 for (i = 0; i < IPPROTO_MAX; i++)
408 ip_protox[i] = pr - inetsw;
409 for (pr = inetdomain.dom_protosw;
410 pr < inetdomain.dom_protoswNPROTOSW; pr++)
411 if (pr->pr_domain->dom_family == PF_INET &&
412 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
413 ip_protox[pr->pr_protocol] = pr - inetsw;
414
415 for (i = 0; i < IPREASS_NHASH; i++)
416 LIST_INIT(&ipq[i]);
417
418 ip_id = time_second & 0xfffff;
419
420 ipintrq.ifq_maxlen = ipqmaxlen;
421 ip_nmbclusters_changed();
422
423 TAILQ_INIT(&in_ifaddrhead);
424 in_ifaddrhashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, M_IFADDR,
425 M_WAITOK, &in_ifaddrhash);
426 in_multihashtbl = hashinit(IN_IFADDR_HASH_SIZE, HASH_LIST, M_IPMADDR,
427 M_WAITOK, &in_multihash);
428 ip_mtudisc_timeout_q = rt_timer_queue_create(ip_mtudisc_timeout);
429 #ifdef GATEWAY
430 ipflow_init();
431 #endif
432
433 #ifdef PFIL_HOOKS
434 /* Register our Packet Filter hook. */
435 inet_pfil_hook.ph_type = PFIL_TYPE_AF;
436 inet_pfil_hook.ph_af = AF_INET;
437 i = pfil_head_register(&inet_pfil_hook);
438 if (i != 0)
439 printf("ip_init: WARNING: unable to register pfil hook, "
440 "error %d\n", i);
441 #endif /* PFIL_HOOKS */
442
443 #ifdef MBUFTRACE
444 MOWNER_ATTACH(&ip_tx_mowner);
445 MOWNER_ATTACH(&ip_rx_mowner);
446 #endif /* MBUFTRACE */
447 }
448
449 struct sockaddr_in ipaddr = {
450 .sin_len = sizeof(ipaddr),
451 .sin_family = AF_INET,
452 };
453 struct route ipforward_rt;
454
455 /*
456 * IP software interrupt routine
457 */
458 void
459 ipintr(void)
460 {
461 int s;
462 struct mbuf *m;
463
464 while (1) {
465 s = splnet();
466 IF_DEQUEUE(&ipintrq, m);
467 splx(s);
468 if (m == 0)
469 return;
470 MCLAIM(m, &ip_rx_mowner);
471 ip_input(m);
472 }
473 }
474
475 /*
476 * Ip input routine. Checksum and byte swap header. If fragmented
477 * try to reassemble. Process options. Pass to next level.
478 */
479 void
480 ip_input(struct mbuf *m)
481 {
482 struct ip *ip = NULL;
483 struct ipq *fp;
484 struct in_ifaddr *ia;
485 struct ifaddr *ifa;
486 struct ipqent *ipqe;
487 int hlen = 0, mff, len;
488 int downmatch;
489 int checkif;
490 int srcrt = 0;
491 int s;
492 u_int hash;
493 #ifdef FAST_IPSEC
494 struct m_tag *mtag;
495 struct tdb_ident *tdbi;
496 struct secpolicy *sp;
497 int error;
498 #endif /* FAST_IPSEC */
499
500 MCLAIM(m, &ip_rx_mowner);
501 #ifdef DIAGNOSTIC
502 if ((m->m_flags & M_PKTHDR) == 0)
503 panic("ipintr no HDR");
504 #endif
505
506 /*
507 * If no IP addresses have been set yet but the interfaces
508 * are receiving, can't do anything with incoming packets yet.
509 */
510 if (TAILQ_FIRST(&in_ifaddrhead) == 0)
511 goto bad;
512 ipstat.ips_total++;
513 /*
514 * If the IP header is not aligned, slurp it up into a new
515 * mbuf with space for link headers, in the event we forward
516 * it. Otherwise, if it is aligned, make sure the entire
517 * base IP header is in the first mbuf of the chain.
518 */
519 if (IP_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
520 if ((m = m_copyup(m, sizeof(struct ip),
521 (max_linkhdr + 3) & ~3)) == NULL) {
522 /* XXXJRT new stat, please */
523 ipstat.ips_toosmall++;
524 return;
525 }
526 } else if (__predict_false(m->m_len < sizeof (struct ip))) {
527 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
528 ipstat.ips_toosmall++;
529 return;
530 }
531 }
532 ip = mtod(m, struct ip *);
533 if (ip->ip_v != IPVERSION) {
534 ipstat.ips_badvers++;
535 goto bad;
536 }
537 hlen = ip->ip_hl << 2;
538 if (hlen < sizeof(struct ip)) { /* minimum header length */
539 ipstat.ips_badhlen++;
540 goto bad;
541 }
542 if (hlen > m->m_len) {
543 if ((m = m_pullup(m, hlen)) == 0) {
544 ipstat.ips_badhlen++;
545 return;
546 }
547 ip = mtod(m, struct ip *);
548 }
549
550 /*
551 * RFC1122: packets with a multicast source address are
552 * not allowed.
553 */
554 if (IN_MULTICAST(ip->ip_src.s_addr)) {
555 ipstat.ips_badaddr++;
556 goto bad;
557 }
558
559 /* 127/8 must not appear on wire - RFC1122 */
560 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
561 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
562 if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
563 ipstat.ips_badaddr++;
564 goto bad;
565 }
566 }
567
568 switch (m->m_pkthdr.csum_flags &
569 ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_IPv4) |
570 M_CSUM_IPv4_BAD)) {
571 case M_CSUM_IPv4|M_CSUM_IPv4_BAD:
572 INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad);
573 goto badcsum;
574
575 case M_CSUM_IPv4:
576 /* Checksum was okay. */
577 INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok);
578 break;
579
580 default:
581 /*
582 * Must compute it ourselves. Maybe skip checksum on
583 * loopback interfaces.
584 */
585 if (__predict_true(!(m->m_pkthdr.rcvif->if_flags &
586 IFF_LOOPBACK) || ip_do_loopback_cksum)) {
587 INET_CSUM_COUNTER_INCR(&ip_swcsum);
588 if (in_cksum(m, hlen) != 0)
589 goto badcsum;
590 }
591 break;
592 }
593
594 /* Retrieve the packet length. */
595 len = ntohs(ip->ip_len);
596
597 /*
598 * Check for additional length bogosity
599 */
600 if (len < hlen) {
601 ipstat.ips_badlen++;
602 goto bad;
603 }
604
605 /*
606 * Check that the amount of data in the buffers
607 * is as at least much as the IP header would have us expect.
608 * Trim mbufs if longer than we expect.
609 * Drop packet if shorter than we expect.
610 */
611 if (m->m_pkthdr.len < len) {
612 ipstat.ips_tooshort++;
613 goto bad;
614 }
615 if (m->m_pkthdr.len > len) {
616 if (m->m_len == m->m_pkthdr.len) {
617 m->m_len = len;
618 m->m_pkthdr.len = len;
619 } else
620 m_adj(m, len - m->m_pkthdr.len);
621 }
622
623 #if defined(IPSEC)
624 /* ipflow (IP fast forwarding) is not compatible with IPsec. */
625 m->m_flags &= ~M_CANFASTFWD;
626 #else
627 /*
628 * Assume that we can create a fast-forward IP flow entry
629 * based on this packet.
630 */
631 m->m_flags |= M_CANFASTFWD;
632 #endif
633
634 #ifdef PFIL_HOOKS
635 /*
636 * Run through list of hooks for input packets. If there are any
637 * filters which require that additional packets in the flow are
638 * not fast-forwarded, they must clear the M_CANFASTFWD flag.
639 * Note that filters must _never_ set this flag, as another filter
640 * in the list may have previously cleared it.
641 */
642 /*
643 * let ipfilter look at packet on the wire,
644 * not the decapsulated packet.
645 */
646 #ifdef IPSEC
647 if (!ipsec_getnhist(m))
648 #elif defined(FAST_IPSEC)
649 if (!ipsec_indone(m))
650 #else
651 if (1)
652 #endif
653 {
654 struct in_addr odst;
655
656 odst = ip->ip_dst;
657 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif,
658 PFIL_IN) != 0)
659 return;
660 if (m == NULL)
661 return;
662 ip = mtod(m, struct ip *);
663 hlen = ip->ip_hl << 2;
664 /*
665 * XXX The setting of "srcrt" here is to prevent ip_forward()
666 * from generating ICMP redirects for packets that have
667 * been redirected by a hook back out on to the same LAN that
668 * they came from and is not an indication that the packet
669 * is being inffluenced by source routing options. This
670 * allows things like
671 * "rdr tlp0 0/0 port 80 -> 1.1.1.200 3128 tcp"
672 * where tlp0 is both on the 1.1.1.0/24 network and is the
673 * default route for hosts on 1.1.1.0/24. Of course this
674 * also requires a "map tlp0 ..." to complete the story.
675 * One might argue whether or not this kind of network config.
676 * should be supported in this manner...
677 */
678 srcrt = (odst.s_addr != ip->ip_dst.s_addr);
679 }
680 #endif /* PFIL_HOOKS */
681
682 #ifdef ALTQ
683 /* XXX Temporary until ALTQ is changed to use a pfil hook */
684 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) {
685 /* packet dropped by traffic conditioner */
686 return;
687 }
688 #endif
689
690 /*
691 * Process options and, if not destined for us,
692 * ship it on. ip_dooptions returns 1 when an
693 * error was detected (causing an icmp message
694 * to be sent and the original packet to be freed).
695 */
696 ip_nhops = 0; /* for source routed packets */
697 if (hlen > sizeof (struct ip) && ip_dooptions(m))
698 return;
699
700 /*
701 * Enable a consistency check between the destination address
702 * and the arrival interface for a unicast packet (the RFC 1122
703 * strong ES model) if IP forwarding is disabled and the packet
704 * is not locally generated.
705 *
706 * XXX - Checking also should be disabled if the destination
707 * address is ipnat'ed to a different interface.
708 *
709 * XXX - Checking is incompatible with IP aliases added
710 * to the loopback interface instead of the interface where
711 * the packets are received.
712 *
713 * XXX - We need to add a per ifaddr flag for this so that
714 * we get finer grain control.
715 */
716 checkif = ip_checkinterface && (ipforwarding == 0) &&
717 (m->m_pkthdr.rcvif != NULL) &&
718 ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0);
719
720 /*
721 * Check our list of addresses, to see if the packet is for us.
722 *
723 * Traditional 4.4BSD did not consult IFF_UP at all.
724 * The behavior here is to treat addresses on !IFF_UP interface
725 * as not mine.
726 */
727 downmatch = 0;
728 LIST_FOREACH(ia, &IN_IFADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
729 if (in_hosteq(ia->ia_addr.sin_addr, ip->ip_dst)) {
730 if (checkif && ia->ia_ifp != m->m_pkthdr.rcvif)
731 continue;
732 if ((ia->ia_ifp->if_flags & IFF_UP) != 0)
733 break;
734 else
735 downmatch++;
736 }
737 }
738 if (ia != NULL)
739 goto ours;
740 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
741 IFADDR_FOREACH(ifa, m->m_pkthdr.rcvif) {
742 if (ifa->ifa_addr->sa_family != AF_INET)
743 continue;
744 ia = ifatoia(ifa);
745 if (in_hosteq(ip->ip_dst, ia->ia_broadaddr.sin_addr) ||
746 in_hosteq(ip->ip_dst, ia->ia_netbroadcast) ||
747 /*
748 * Look for all-0's host part (old broadcast addr),
749 * either for subnet or net.
750 */
751 ip->ip_dst.s_addr == ia->ia_subnet ||
752 ip->ip_dst.s_addr == ia->ia_net)
753 goto ours;
754 /*
755 * An interface with IP address zero accepts
756 * all packets that arrive on that interface.
757 */
758 if (in_nullhost(ia->ia_addr.sin_addr))
759 goto ours;
760 }
761 }
762 if (IN_MULTICAST(ip->ip_dst.s_addr)) {
763 struct in_multi *inm;
764 #ifdef MROUTING
765 extern struct socket *ip_mrouter;
766
767 if (ip_mrouter) {
768 /*
769 * If we are acting as a multicast router, all
770 * incoming multicast packets are passed to the
771 * kernel-level multicast forwarding function.
772 * The packet is returned (relatively) intact; if
773 * ip_mforward() returns a non-zero value, the packet
774 * must be discarded, else it may be accepted below.
775 *
776 * (The IP ident field is put in the same byte order
777 * as expected when ip_mforward() is called from
778 * ip_output().)
779 */
780 if (ip_mforward(m, m->m_pkthdr.rcvif) != 0) {
781 ipstat.ips_cantforward++;
782 m_freem(m);
783 return;
784 }
785
786 /*
787 * The process-level routing demon needs to receive
788 * all multicast IGMP packets, whether or not this
789 * host belongs to their destination groups.
790 */
791 if (ip->ip_p == IPPROTO_IGMP)
792 goto ours;
793 ipstat.ips_forward++;
794 }
795 #endif
796 /*
797 * See if we belong to the destination multicast group on the
798 * arrival interface.
799 */
800 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
801 if (inm == NULL) {
802 ipstat.ips_cantforward++;
803 m_freem(m);
804 return;
805 }
806 goto ours;
807 }
808 if (ip->ip_dst.s_addr == INADDR_BROADCAST ||
809 in_nullhost(ip->ip_dst))
810 goto ours;
811
812 /*
813 * Not for us; forward if possible and desirable.
814 */
815 if (ipforwarding == 0) {
816 ipstat.ips_cantforward++;
817 m_freem(m);
818 } else {
819 /*
820 * If ip_dst matched any of my address on !IFF_UP interface,
821 * and there's no IFF_UP interface that matches ip_dst,
822 * send icmp unreach. Forwarding it will result in in-kernel
823 * forwarding loop till TTL goes to 0.
824 */
825 if (downmatch) {
826 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
827 ipstat.ips_cantforward++;
828 return;
829 }
830 #ifdef IPSEC
831 if (ipsec4_in_reject(m, NULL)) {
832 ipsecstat.in_polvio++;
833 goto bad;
834 }
835 #endif
836 #ifdef FAST_IPSEC
837 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
838 s = splsoftnet();
839 if (mtag != NULL) {
840 tdbi = (struct tdb_ident *)(mtag + 1);
841 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
842 } else {
843 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
844 IP_FORWARDING, &error);
845 }
846 if (sp == NULL) { /* NB: can happen if error */
847 splx(s);
848 /*XXX error stat???*/
849 DPRINTF(("ip_input: no SP for forwarding\n")); /*XXX*/
850 goto bad;
851 }
852
853 /*
854 * Check security policy against packet attributes.
855 */
856 error = ipsec_in_reject(sp, m);
857 KEY_FREESP(&sp);
858 splx(s);
859 if (error) {
860 ipstat.ips_cantforward++;
861 goto bad;
862 }
863
864 /*
865 * Peek at the outbound SP for this packet to determine if
866 * it's a Fast Forward candidate.
867 */
868 mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL);
869 if (mtag != NULL)
870 m->m_flags &= ~M_CANFASTFWD;
871 else {
872 s = splsoftnet();
873 sp = ipsec4_checkpolicy(m, IPSEC_DIR_OUTBOUND,
874 (IP_FORWARDING |
875 (ip_directedbcast ? IP_ALLOWBROADCAST : 0)),
876 &error, NULL);
877 if (sp != NULL) {
878 m->m_flags &= ~M_CANFASTFWD;
879 KEY_FREESP(&sp);
880 }
881 splx(s);
882 }
883 #endif /* FAST_IPSEC */
884
885 ip_forward(m, srcrt);
886 }
887 return;
888
889 ours:
890 /*
891 * If offset or IP_MF are set, must reassemble.
892 * Otherwise, nothing need be done.
893 * (We could look in the reassembly queue to see
894 * if the packet was previously fragmented,
895 * but it's not worth the time; just let them time out.)
896 */
897 if (ip->ip_off & ~htons(IP_DF|IP_RF)) {
898
899 /*
900 * Look for queue of fragments
901 * of this datagram.
902 */
903 IPQ_LOCK();
904 hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
905 /* XXX LIST_FOREACH(fp, &ipq[hash], ipq_q) */
906 for (fp = LIST_FIRST(&ipq[hash]); fp != NULL;
907 fp = LIST_NEXT(fp, ipq_q)) {
908 if (ip->ip_id == fp->ipq_id &&
909 in_hosteq(ip->ip_src, fp->ipq_src) &&
910 in_hosteq(ip->ip_dst, fp->ipq_dst) &&
911 ip->ip_p == fp->ipq_p)
912 goto found;
913
914 }
915 fp = 0;
916 found:
917
918 /*
919 * Adjust ip_len to not reflect header,
920 * set ipqe_mff if more fragments are expected,
921 * convert offset of this to bytes.
922 */
923 ip->ip_len = htons(ntohs(ip->ip_len) - hlen);
924 mff = (ip->ip_off & htons(IP_MF)) != 0;
925 if (mff) {
926 /*
927 * Make sure that fragments have a data length
928 * that's a non-zero multiple of 8 bytes.
929 */
930 if (ntohs(ip->ip_len) == 0 ||
931 (ntohs(ip->ip_len) & 0x7) != 0) {
932 ipstat.ips_badfrags++;
933 IPQ_UNLOCK();
934 goto bad;
935 }
936 }
937 ip->ip_off = htons((ntohs(ip->ip_off) & IP_OFFMASK) << 3);
938
939 /*
940 * If datagram marked as having more fragments
941 * or if this is not the first fragment,
942 * attempt reassembly; if it succeeds, proceed.
943 */
944 if (mff || ip->ip_off != htons(0)) {
945 ipstat.ips_fragments++;
946 s = splvm();
947 ipqe = pool_get(&ipqent_pool, PR_NOWAIT);
948 splx(s);
949 if (ipqe == NULL) {
950 ipstat.ips_rcvmemdrop++;
951 IPQ_UNLOCK();
952 goto bad;
953 }
954 ipqe->ipqe_mff = mff;
955 ipqe->ipqe_m = m;
956 ipqe->ipqe_ip = ip;
957 m = ip_reass(ipqe, fp, &ipq[hash]);
958 if (m == 0) {
959 IPQ_UNLOCK();
960 return;
961 }
962 ipstat.ips_reassembled++;
963 ip = mtod(m, struct ip *);
964 hlen = ip->ip_hl << 2;
965 ip->ip_len = htons(ntohs(ip->ip_len) + hlen);
966 } else
967 if (fp)
968 ip_freef(fp);
969 IPQ_UNLOCK();
970 }
971
972 #if defined(IPSEC)
973 /*
974 * enforce IPsec policy checking if we are seeing last header.
975 * note that we do not visit this with protocols with pcb layer
976 * code - like udp/tcp/raw ip.
977 */
978 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0 &&
979 ipsec4_in_reject(m, NULL)) {
980 ipsecstat.in_polvio++;
981 goto bad;
982 }
983 #endif
984 #ifdef FAST_IPSEC
985 /*
986 * enforce IPsec policy checking if we are seeing last header.
987 * note that we do not visit this with protocols with pcb layer
988 * code - like udp/tcp/raw ip.
989 */
990 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) {
991 /*
992 * Check if the packet has already had IPsec processing
993 * done. If so, then just pass it along. This tag gets
994 * set during AH, ESP, etc. input handling, before the
995 * packet is returned to the ip input queue for delivery.
996 */
997 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
998 s = splsoftnet();
999 if (mtag != NULL) {
1000 tdbi = (struct tdb_ident *)(mtag + 1);
1001 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
1002 } else {
1003 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
1004 IP_FORWARDING, &error);
1005 }
1006 if (sp != NULL) {
1007 /*
1008 * Check security policy against packet attributes.
1009 */
1010 error = ipsec_in_reject(sp, m);
1011 KEY_FREESP(&sp);
1012 } else {
1013 /* XXX error stat??? */
1014 error = EINVAL;
1015 DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/
1016 }
1017 splx(s);
1018 if (error)
1019 goto bad;
1020 }
1021 #endif /* FAST_IPSEC */
1022
1023 /*
1024 * Switch out to protocol's input routine.
1025 */
1026 #if IFA_STATS
1027 if (ia && ip)
1028 ia->ia_ifa.ifa_data.ifad_inbytes += ntohs(ip->ip_len);
1029 #endif
1030 ipstat.ips_delivered++;
1031 {
1032 int off = hlen, nh = ip->ip_p;
1033
1034 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh);
1035 return;
1036 }
1037 bad:
1038 m_freem(m);
1039 return;
1040
1041 badcsum:
1042 ipstat.ips_badsum++;
1043 m_freem(m);
1044 }
1045
1046 /*
1047 * Take incoming datagram fragment and try to
1048 * reassemble it into whole datagram. If a chain for
1049 * reassembly of this datagram already exists, then it
1050 * is given as fp; otherwise have to make a chain.
1051 */
1052 struct mbuf *
1053 ip_reass(struct ipqent *ipqe, struct ipq *fp, struct ipqhead *ipqhead)
1054 {
1055 struct mbuf *m = ipqe->ipqe_m;
1056 struct ipqent *nq, *p, *q;
1057 struct ip *ip;
1058 struct mbuf *t;
1059 int hlen = ipqe->ipqe_ip->ip_hl << 2;
1060 int i, next, s;
1061
1062 IPQ_LOCK_CHECK();
1063
1064 /*
1065 * Presence of header sizes in mbufs
1066 * would confuse code below.
1067 */
1068 m->m_data += hlen;
1069 m->m_len -= hlen;
1070
1071 #ifdef notyet
1072 /* make sure fragment limit is up-to-date */
1073 CHECK_NMBCLUSTER_PARAMS();
1074
1075 /* If we have too many fragments, drop the older half. */
1076 if (ip_nfrags >= ip_maxfrags)
1077 ip_reass_drophalf(void);
1078 #endif
1079
1080 /*
1081 * We are about to add a fragment; increment frag count.
1082 */
1083 ip_nfrags++;
1084
1085 /*
1086 * If first fragment to arrive, create a reassembly queue.
1087 */
1088 if (fp == 0) {
1089 /*
1090 * Enforce upper bound on number of fragmented packets
1091 * for which we attempt reassembly;
1092 * If maxfrag is 0, never accept fragments.
1093 * If maxfrag is -1, accept all fragments without limitation.
1094 */
1095 if (ip_maxfragpackets < 0)
1096 ;
1097 else if (ip_nfragpackets >= ip_maxfragpackets)
1098 goto dropfrag;
1099 ip_nfragpackets++;
1100 MALLOC(fp, struct ipq *, sizeof (struct ipq),
1101 M_FTABLE, M_NOWAIT);
1102 if (fp == NULL)
1103 goto dropfrag;
1104 LIST_INSERT_HEAD(ipqhead, fp, ipq_q);
1105 fp->ipq_nfrags = 1;
1106 fp->ipq_ttl = IPFRAGTTL;
1107 fp->ipq_p = ipqe->ipqe_ip->ip_p;
1108 fp->ipq_id = ipqe->ipqe_ip->ip_id;
1109 TAILQ_INIT(&fp->ipq_fragq);
1110 fp->ipq_src = ipqe->ipqe_ip->ip_src;
1111 fp->ipq_dst = ipqe->ipqe_ip->ip_dst;
1112 p = NULL;
1113 goto insert;
1114 } else {
1115 fp->ipq_nfrags++;
1116 }
1117
1118 /*
1119 * Find a segment which begins after this one does.
1120 */
1121 for (p = NULL, q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL;
1122 p = q, q = TAILQ_NEXT(q, ipqe_q))
1123 if (ntohs(q->ipqe_ip->ip_off) > ntohs(ipqe->ipqe_ip->ip_off))
1124 break;
1125
1126 /*
1127 * If there is a preceding segment, it may provide some of
1128 * our data already. If so, drop the data from the incoming
1129 * segment. If it provides all of our data, drop us.
1130 */
1131 if (p != NULL) {
1132 i = ntohs(p->ipqe_ip->ip_off) + ntohs(p->ipqe_ip->ip_len) -
1133 ntohs(ipqe->ipqe_ip->ip_off);
1134 if (i > 0) {
1135 if (i >= ntohs(ipqe->ipqe_ip->ip_len))
1136 goto dropfrag;
1137 m_adj(ipqe->ipqe_m, i);
1138 ipqe->ipqe_ip->ip_off =
1139 htons(ntohs(ipqe->ipqe_ip->ip_off) + i);
1140 ipqe->ipqe_ip->ip_len =
1141 htons(ntohs(ipqe->ipqe_ip->ip_len) - i);
1142 }
1143 }
1144
1145 /*
1146 * While we overlap succeeding segments trim them or,
1147 * if they are completely covered, dequeue them.
1148 */
1149 for (; q != NULL &&
1150 ntohs(ipqe->ipqe_ip->ip_off) + ntohs(ipqe->ipqe_ip->ip_len) >
1151 ntohs(q->ipqe_ip->ip_off); q = nq) {
1152 i = (ntohs(ipqe->ipqe_ip->ip_off) +
1153 ntohs(ipqe->ipqe_ip->ip_len)) - ntohs(q->ipqe_ip->ip_off);
1154 if (i < ntohs(q->ipqe_ip->ip_len)) {
1155 q->ipqe_ip->ip_len =
1156 htons(ntohs(q->ipqe_ip->ip_len) - i);
1157 q->ipqe_ip->ip_off =
1158 htons(ntohs(q->ipqe_ip->ip_off) + i);
1159 m_adj(q->ipqe_m, i);
1160 break;
1161 }
1162 nq = TAILQ_NEXT(q, ipqe_q);
1163 m_freem(q->ipqe_m);
1164 TAILQ_REMOVE(&fp->ipq_fragq, q, ipqe_q);
1165 s = splvm();
1166 pool_put(&ipqent_pool, q);
1167 splx(s);
1168 fp->ipq_nfrags--;
1169 ip_nfrags--;
1170 }
1171
1172 insert:
1173 /*
1174 * Stick new segment in its place;
1175 * check for complete reassembly.
1176 */
1177 if (p == NULL) {
1178 TAILQ_INSERT_HEAD(&fp->ipq_fragq, ipqe, ipqe_q);
1179 } else {
1180 TAILQ_INSERT_AFTER(&fp->ipq_fragq, p, ipqe, ipqe_q);
1181 }
1182 next = 0;
1183 for (p = NULL, q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL;
1184 p = q, q = TAILQ_NEXT(q, ipqe_q)) {
1185 if (ntohs(q->ipqe_ip->ip_off) != next)
1186 return (0);
1187 next += ntohs(q->ipqe_ip->ip_len);
1188 }
1189 if (p->ipqe_mff)
1190 return (0);
1191
1192 /*
1193 * Reassembly is complete. Check for a bogus message size and
1194 * concatenate fragments.
1195 */
1196 q = TAILQ_FIRST(&fp->ipq_fragq);
1197 ip = q->ipqe_ip;
1198 if ((next + (ip->ip_hl << 2)) > IP_MAXPACKET) {
1199 ipstat.ips_toolong++;
1200 ip_freef(fp);
1201 return (0);
1202 }
1203 m = q->ipqe_m;
1204 t = m->m_next;
1205 m->m_next = 0;
1206 m_cat(m, t);
1207 nq = TAILQ_NEXT(q, ipqe_q);
1208 s = splvm();
1209 pool_put(&ipqent_pool, q);
1210 splx(s);
1211 for (q = nq; q != NULL; q = nq) {
1212 t = q->ipqe_m;
1213 nq = TAILQ_NEXT(q, ipqe_q);
1214 s = splvm();
1215 pool_put(&ipqent_pool, q);
1216 splx(s);
1217 m_cat(m, t);
1218 }
1219 ip_nfrags -= fp->ipq_nfrags;
1220
1221 /*
1222 * Create header for new ip packet by
1223 * modifying header of first packet;
1224 * dequeue and discard fragment reassembly header.
1225 * Make header visible.
1226 */
1227 ip->ip_len = htons(next);
1228 ip->ip_src = fp->ipq_src;
1229 ip->ip_dst = fp->ipq_dst;
1230 LIST_REMOVE(fp, ipq_q);
1231 FREE(fp, M_FTABLE);
1232 ip_nfragpackets--;
1233 m->m_len += (ip->ip_hl << 2);
1234 m->m_data -= (ip->ip_hl << 2);
1235 /* some debugging cruft by sklower, below, will go away soon */
1236 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */
1237 int plen = 0;
1238 for (t = m; t; t = t->m_next)
1239 plen += t->m_len;
1240 m->m_pkthdr.len = plen;
1241 m->m_pkthdr.csum_flags = 0;
1242 }
1243 return (m);
1244
1245 dropfrag:
1246 if (fp != 0)
1247 fp->ipq_nfrags--;
1248 ip_nfrags--;
1249 ipstat.ips_fragdropped++;
1250 m_freem(m);
1251 s = splvm();
1252 pool_put(&ipqent_pool, ipqe);
1253 splx(s);
1254 return (0);
1255 }
1256
1257 /*
1258 * Free a fragment reassembly header and all
1259 * associated datagrams.
1260 */
1261 void
1262 ip_freef(struct ipq *fp)
1263 {
1264 struct ipqent *q, *p;
1265 u_int nfrags = 0;
1266 int s;
1267
1268 IPQ_LOCK_CHECK();
1269
1270 for (q = TAILQ_FIRST(&fp->ipq_fragq); q != NULL; q = p) {
1271 p = TAILQ_NEXT(q, ipqe_q);
1272 m_freem(q->ipqe_m);
1273 nfrags++;
1274 TAILQ_REMOVE(&fp->ipq_fragq, q, ipqe_q);
1275 s = splvm();
1276 pool_put(&ipqent_pool, q);
1277 splx(s);
1278 }
1279
1280 if (nfrags != fp->ipq_nfrags)
1281 printf("ip_freef: nfrags %d != %d\n", fp->ipq_nfrags, nfrags);
1282 ip_nfrags -= nfrags;
1283 LIST_REMOVE(fp, ipq_q);
1284 FREE(fp, M_FTABLE);
1285 ip_nfragpackets--;
1286 }
1287
1288 /*
1289 * IP reassembly TTL machinery for multiplicative drop.
1290 */
1291 static u_int fragttl_histo[(IPFRAGTTL+1)];
1292
1293
1294 /*
1295 * Decrement TTL of all reasembly queue entries by `ticks'.
1296 * Count number of distinct fragments (as opposed to partial, fragmented
1297 * datagrams) in the reassembly queue. While we traverse the entire
1298 * reassembly queue, compute and return the median TTL over all fragments.
1299 */
1300 static u_int
1301 ip_reass_ttl_decr(u_int ticks)
1302 {
1303 u_int nfrags, median, dropfraction, keepfraction;
1304 struct ipq *fp, *nfp;
1305 int i;
1306
1307 nfrags = 0;
1308 memset(fragttl_histo, 0, sizeof fragttl_histo);
1309
1310 for (i = 0; i < IPREASS_NHASH; i++) {
1311 for (fp = LIST_FIRST(&ipq[i]); fp != NULL; fp = nfp) {
1312 fp->ipq_ttl = ((fp->ipq_ttl <= ticks) ?
1313 0 : fp->ipq_ttl - ticks);
1314 nfp = LIST_NEXT(fp, ipq_q);
1315 if (fp->ipq_ttl == 0) {
1316 ipstat.ips_fragtimeout++;
1317 ip_freef(fp);
1318 } else {
1319 nfrags += fp->ipq_nfrags;
1320 fragttl_histo[fp->ipq_ttl] += fp->ipq_nfrags;
1321 }
1322 }
1323 }
1324
1325 KASSERT(ip_nfrags == nfrags);
1326
1327 /* Find median (or other drop fraction) in histogram. */
1328 dropfraction = (ip_nfrags / 2);
1329 keepfraction = ip_nfrags - dropfraction;
1330 for (i = IPFRAGTTL, median = 0; i >= 0; i--) {
1331 median += fragttl_histo[i];
1332 if (median >= keepfraction)
1333 break;
1334 }
1335
1336 /* Return TTL of median (or other fraction). */
1337 return (u_int)i;
1338 }
1339
1340 void
1341 ip_reass_drophalf(void)
1342 {
1343
1344 u_int median_ticks;
1345 /*
1346 * Compute median TTL of all fragments, and count frags
1347 * with that TTL or lower (roughly half of all fragments).
1348 */
1349 median_ticks = ip_reass_ttl_decr(0);
1350
1351 /* Drop half. */
1352 median_ticks = ip_reass_ttl_decr(median_ticks);
1353
1354 }
1355
1356 /*
1357 * IP timer processing;
1358 * if a timer expires on a reassembly
1359 * queue, discard it.
1360 */
1361 void
1362 ip_slowtimo(void)
1363 {
1364 static u_int dropscanidx = 0;
1365 u_int i;
1366 u_int median_ttl;
1367 int s = splsoftnet();
1368
1369 IPQ_LOCK();
1370
1371 /* Age TTL of all fragments by 1 tick .*/
1372 median_ttl = ip_reass_ttl_decr(1);
1373
1374 /* make sure fragment limit is up-to-date */
1375 CHECK_NMBCLUSTER_PARAMS();
1376
1377 /* If we have too many fragments, drop the older half. */
1378 if (ip_nfrags > ip_maxfrags)
1379 ip_reass_ttl_decr(median_ttl);
1380
1381 /*
1382 * If we are over the maximum number of fragmented packets
1383 * (due to the limit being lowered), drain off
1384 * enough to get down to the new limit. Start draining
1385 * from the reassembly hashqueue most recently drained.
1386 */
1387 if (ip_maxfragpackets < 0)
1388 ;
1389 else {
1390 int wrapped = 0;
1391
1392 i = dropscanidx;
1393 while (ip_nfragpackets > ip_maxfragpackets && wrapped == 0) {
1394 while (LIST_FIRST(&ipq[i]) != NULL)
1395 ip_freef(LIST_FIRST(&ipq[i]));
1396 if (++i >= IPREASS_NHASH) {
1397 i = 0;
1398 }
1399 /*
1400 * Dont scan forever even if fragment counters are
1401 * wrong: stop after scanning entire reassembly queue.
1402 */
1403 if (i == dropscanidx)
1404 wrapped = 1;
1405 }
1406 dropscanidx = i;
1407 }
1408 IPQ_UNLOCK();
1409 #ifdef GATEWAY
1410 ipflow_slowtimo();
1411 #endif
1412 splx(s);
1413 }
1414
1415 /*
1416 * Drain off all datagram fragments.
1417 */
1418 void
1419 ip_drain(void)
1420 {
1421
1422 /*
1423 * We may be called from a device's interrupt context. If
1424 * the ipq is already busy, just bail out now.
1425 */
1426 if (ipq_lock_try() == 0)
1427 return;
1428
1429 /*
1430 * Drop half the total fragments now. If more mbufs are needed,
1431 * we will be called again soon.
1432 */
1433 ip_reass_drophalf();
1434
1435 IPQ_UNLOCK();
1436 }
1437
1438 /*
1439 * Do option processing on a datagram,
1440 * possibly discarding it if bad options are encountered,
1441 * or forwarding it if source-routed.
1442 * Returns 1 if packet has been forwarded/freed,
1443 * 0 if the packet should be processed further.
1444 */
1445 int
1446 ip_dooptions(struct mbuf *m)
1447 {
1448 struct ip *ip = mtod(m, struct ip *);
1449 u_char *cp, *cp0;
1450 struct ip_timestamp *ipt;
1451 struct in_ifaddr *ia;
1452 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0;
1453 struct in_addr dst;
1454 n_time ntime;
1455
1456 dst = ip->ip_dst;
1457 cp = (u_char *)(ip + 1);
1458 cnt = (ip->ip_hl << 2) - sizeof (struct ip);
1459 for (; cnt > 0; cnt -= optlen, cp += optlen) {
1460 opt = cp[IPOPT_OPTVAL];
1461 if (opt == IPOPT_EOL)
1462 break;
1463 if (opt == IPOPT_NOP)
1464 optlen = 1;
1465 else {
1466 if (cnt < IPOPT_OLEN + sizeof(*cp)) {
1467 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1468 goto bad;
1469 }
1470 optlen = cp[IPOPT_OLEN];
1471 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
1472 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1473 goto bad;
1474 }
1475 }
1476 switch (opt) {
1477
1478 default:
1479 break;
1480
1481 /*
1482 * Source routing with record.
1483 * Find interface with current destination address.
1484 * If none on this machine then drop if strictly routed,
1485 * or do nothing if loosely routed.
1486 * Record interface address and bring up next address
1487 * component. If strictly routed make sure next
1488 * address is on directly accessible net.
1489 */
1490 case IPOPT_LSRR:
1491 case IPOPT_SSRR:
1492 if (ip_allowsrcrt == 0) {
1493 type = ICMP_UNREACH;
1494 code = ICMP_UNREACH_NET_PROHIB;
1495 goto bad;
1496 }
1497 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1498 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1499 goto bad;
1500 }
1501 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1502 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1503 goto bad;
1504 }
1505 ipaddr.sin_addr = ip->ip_dst;
1506 ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr)));
1507 if (ia == 0) {
1508 if (opt == IPOPT_SSRR) {
1509 type = ICMP_UNREACH;
1510 code = ICMP_UNREACH_SRCFAIL;
1511 goto bad;
1512 }
1513 /*
1514 * Loose routing, and not at next destination
1515 * yet; nothing to do except forward.
1516 */
1517 break;
1518 }
1519 off--; /* 0 origin */
1520 if ((off + sizeof(struct in_addr)) > optlen) {
1521 /*
1522 * End of source route. Should be for us.
1523 */
1524 save_rte(cp, ip->ip_src);
1525 break;
1526 }
1527 /*
1528 * locate outgoing interface
1529 */
1530 bcopy((caddr_t)(cp + off), (caddr_t)&ipaddr.sin_addr,
1531 sizeof(ipaddr.sin_addr));
1532 if (opt == IPOPT_SSRR)
1533 ia = ifatoia(ifa_ifwithladdr(sintosa(&ipaddr)));
1534 else
1535 ia = ip_rtaddr(ipaddr.sin_addr);
1536 if (ia == 0) {
1537 type = ICMP_UNREACH;
1538 code = ICMP_UNREACH_SRCFAIL;
1539 goto bad;
1540 }
1541 ip->ip_dst = ipaddr.sin_addr;
1542 bcopy((caddr_t)&ia->ia_addr.sin_addr,
1543 (caddr_t)(cp + off), sizeof(struct in_addr));
1544 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1545 /*
1546 * Let ip_intr's mcast routing check handle mcast pkts
1547 */
1548 forward = !IN_MULTICAST(ip->ip_dst.s_addr);
1549 break;
1550
1551 case IPOPT_RR:
1552 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1553 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1554 goto bad;
1555 }
1556 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1557 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1558 goto bad;
1559 }
1560 /*
1561 * If no space remains, ignore.
1562 */
1563 off--; /* 0 origin */
1564 if ((off + sizeof(struct in_addr)) > optlen)
1565 break;
1566 bcopy((caddr_t)(&ip->ip_dst), (caddr_t)&ipaddr.sin_addr,
1567 sizeof(ipaddr.sin_addr));
1568 /*
1569 * locate outgoing interface; if we're the destination,
1570 * use the incoming interface (should be same).
1571 */
1572 if ((ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr))))
1573 == NULL &&
1574 (ia = ip_rtaddr(ipaddr.sin_addr)) == NULL) {
1575 type = ICMP_UNREACH;
1576 code = ICMP_UNREACH_HOST;
1577 goto bad;
1578 }
1579 bcopy((caddr_t)&ia->ia_addr.sin_addr,
1580 (caddr_t)(cp + off), sizeof(struct in_addr));
1581 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1582 break;
1583
1584 case IPOPT_TS:
1585 code = cp - (u_char *)ip;
1586 ipt = (struct ip_timestamp *)cp;
1587 if (ipt->ipt_len < 4 || ipt->ipt_len > 40) {
1588 code = (u_char *)&ipt->ipt_len - (u_char *)ip;
1589 goto bad;
1590 }
1591 if (ipt->ipt_ptr < 5) {
1592 code = (u_char *)&ipt->ipt_ptr - (u_char *)ip;
1593 goto bad;
1594 }
1595 if (ipt->ipt_ptr > ipt->ipt_len - sizeof (int32_t)) {
1596 if (++ipt->ipt_oflw == 0) {
1597 code = (u_char *)&ipt->ipt_ptr -
1598 (u_char *)ip;
1599 goto bad;
1600 }
1601 break;
1602 }
1603 cp0 = (cp + ipt->ipt_ptr - 1);
1604 switch (ipt->ipt_flg) {
1605
1606 case IPOPT_TS_TSONLY:
1607 break;
1608
1609 case IPOPT_TS_TSANDADDR:
1610 if (ipt->ipt_ptr - 1 + sizeof(n_time) +
1611 sizeof(struct in_addr) > ipt->ipt_len) {
1612 code = (u_char *)&ipt->ipt_ptr -
1613 (u_char *)ip;
1614 goto bad;
1615 }
1616 ipaddr.sin_addr = dst;
1617 ia = ifatoia(ifaof_ifpforaddr(sintosa(&ipaddr),
1618 m->m_pkthdr.rcvif));
1619 if (ia == 0)
1620 continue;
1621 bcopy(&ia->ia_addr.sin_addr,
1622 cp0, sizeof(struct in_addr));
1623 ipt->ipt_ptr += sizeof(struct in_addr);
1624 break;
1625
1626 case IPOPT_TS_PRESPEC:
1627 if (ipt->ipt_ptr - 1 + sizeof(n_time) +
1628 sizeof(struct in_addr) > ipt->ipt_len) {
1629 code = (u_char *)&ipt->ipt_ptr -
1630 (u_char *)ip;
1631 goto bad;
1632 }
1633 bcopy(cp0, &ipaddr.sin_addr,
1634 sizeof(struct in_addr));
1635 if (ifatoia(ifa_ifwithaddr(sintosa(&ipaddr)))
1636 == NULL)
1637 continue;
1638 ipt->ipt_ptr += sizeof(struct in_addr);
1639 break;
1640
1641 default:
1642 /* XXX can't take &ipt->ipt_flg */
1643 code = (u_char *)&ipt->ipt_ptr -
1644 (u_char *)ip + 1;
1645 goto bad;
1646 }
1647 ntime = iptime();
1648 cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */
1649 bcopy(cp0, (caddr_t)cp + ipt->ipt_ptr - 1,
1650 sizeof(n_time));
1651 ipt->ipt_ptr += sizeof(n_time);
1652 }
1653 }
1654 if (forward) {
1655 if (ip_forwsrcrt == 0) {
1656 type = ICMP_UNREACH;
1657 code = ICMP_UNREACH_SRCFAIL;
1658 goto bad;
1659 }
1660 ip_forward(m, 1);
1661 return (1);
1662 }
1663 return (0);
1664 bad:
1665 icmp_error(m, type, code, 0, 0);
1666 ipstat.ips_badoptions++;
1667 return (1);
1668 }
1669
1670 /*
1671 * Given address of next destination (final or next hop),
1672 * return internet address info of interface to be used to get there.
1673 */
1674 struct in_ifaddr *
1675 ip_rtaddr(struct in_addr dst)
1676 {
1677 struct sockaddr_in *sin;
1678
1679 sin = satosin(&ipforward_rt.ro_dst);
1680
1681 if (ipforward_rt.ro_rt == 0 || !in_hosteq(dst, sin->sin_addr)) {
1682 if (ipforward_rt.ro_rt) {
1683 RTFREE(ipforward_rt.ro_rt);
1684 ipforward_rt.ro_rt = 0;
1685 }
1686 sin->sin_family = AF_INET;
1687 sin->sin_len = sizeof(*sin);
1688 sin->sin_addr = dst;
1689
1690 rtalloc(&ipforward_rt);
1691 }
1692 if (ipforward_rt.ro_rt == 0)
1693 return ((struct in_ifaddr *)0);
1694 return (ifatoia(ipforward_rt.ro_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((caddr_t)option, (caddr_t)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 ((struct mbuf *)0);
1731 m = m_get(M_DONTWAIT, MT_SOOPTS);
1732 if (m == 0)
1733 return ((struct mbuf *)0);
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 bcopy((caddr_t)&ip_srcrt.nop,
1762 mtod(m, caddr_t) + sizeof(struct in_addr), OPTSIZ);
1763 q = (struct in_addr *)(mtod(m, caddr_t) +
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 /*
1789 * Strip out IP options, at higher
1790 * level protocol in the kernel.
1791 * Second argument is buffer to which options
1792 * will be moved, and return value is their length.
1793 * XXX should be deleted; last arg currently ignored.
1794 */
1795 void
1796 ip_stripoptions(struct mbuf *m, struct mbuf *mopt)
1797 {
1798 int i;
1799 struct ip *ip = mtod(m, struct ip *);
1800 caddr_t opts;
1801 int olen;
1802
1803 olen = (ip->ip_hl << 2) - sizeof (struct ip);
1804 opts = (caddr_t)(ip + 1);
1805 i = m->m_len - (sizeof (struct ip) + olen);
1806 bcopy(opts + olen, opts, (unsigned)i);
1807 m->m_len -= olen;
1808 if (m->m_flags & M_PKTHDR)
1809 m->m_pkthdr.len -= olen;
1810 ip->ip_len = htons(ntohs(ip->ip_len) - olen);
1811 ip->ip_hl = sizeof (struct ip) >> 2;
1812 }
1813
1814 const int inetctlerrmap[PRC_NCMDS] = {
1815 0, 0, 0, 0,
1816 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH,
1817 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED,
1818 EMSGSIZE, EHOSTUNREACH, 0, 0,
1819 0, 0, 0, 0,
1820 ENOPROTOOPT
1821 };
1822
1823 /*
1824 * Forward a packet. If some error occurs return the sender
1825 * an icmp packet. Note we can't always generate a meaningful
1826 * icmp message because icmp doesn't have a large enough repertoire
1827 * of codes and types.
1828 *
1829 * If not forwarding, just drop the packet. This could be confusing
1830 * if ipforwarding was zero but some routing protocol was advancing
1831 * us as a gateway to somewhere. However, we must let the routing
1832 * protocol deal with that.
1833 *
1834 * The srcrt parameter indicates whether the packet is being forwarded
1835 * via a source route.
1836 */
1837 void
1838 ip_forward(struct mbuf *m, int srcrt)
1839 {
1840 struct ip *ip = mtod(m, struct ip *);
1841 struct sockaddr_in *sin;
1842 struct rtentry *rt;
1843 int error, type = 0, code = 0, destmtu = 0;
1844 struct mbuf *mcopy;
1845 n_long dest;
1846
1847 /*
1848 * We are now in the output path.
1849 */
1850 MCLAIM(m, &ip_tx_mowner);
1851
1852 /*
1853 * Clear any in-bound checksum flags for this packet.
1854 */
1855 m->m_pkthdr.csum_flags = 0;
1856
1857 dest = 0;
1858 #ifdef DIAGNOSTIC
1859 if (ipprintfs) {
1860 printf("forward: src %s ", inet_ntoa(ip->ip_src));
1861 printf("dst %s ttl %x\n", inet_ntoa(ip->ip_dst), ip->ip_ttl);
1862 }
1863 #endif
1864 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
1865 ipstat.ips_cantforward++;
1866 m_freem(m);
1867 return;
1868 }
1869 if (ip->ip_ttl <= IPTTLDEC) {
1870 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0);
1871 return;
1872 }
1873
1874 sin = satosin(&ipforward_rt.ro_dst);
1875 if ((rt = ipforward_rt.ro_rt) == 0 ||
1876 !in_hosteq(ip->ip_dst, sin->sin_addr)) {
1877 if (ipforward_rt.ro_rt) {
1878 RTFREE(ipforward_rt.ro_rt);
1879 ipforward_rt.ro_rt = 0;
1880 }
1881 sin->sin_family = AF_INET;
1882 sin->sin_len = sizeof(struct sockaddr_in);
1883 sin->sin_addr = ip->ip_dst;
1884
1885 rtalloc(&ipforward_rt);
1886 if (ipforward_rt.ro_rt == 0) {
1887 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_NET, dest, 0);
1888 return;
1889 }
1890 rt = ipforward_rt.ro_rt;
1891 }
1892
1893 /*
1894 * Save at most 68 bytes of the packet in case
1895 * we need to generate an ICMP message to the src.
1896 * Pullup to avoid sharing mbuf cluster between m and mcopy.
1897 */
1898 mcopy = m_copym(m, 0, imin(ntohs(ip->ip_len), 68), M_DONTWAIT);
1899 if (mcopy)
1900 mcopy = m_pullup(mcopy, ip->ip_hl << 2);
1901
1902 ip->ip_ttl -= IPTTLDEC;
1903
1904 /*
1905 * If forwarding packet using same interface that it came in on,
1906 * perhaps should send a redirect to sender to shortcut a hop.
1907 * Only send redirect if source is sending directly to us,
1908 * and if packet was not source routed (or has any options).
1909 * Also, don't send redirect if forwarding using a default route
1910 * or a route modified by a redirect.
1911 */
1912 if (rt->rt_ifp == m->m_pkthdr.rcvif &&
1913 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1914 !in_nullhost(satosin(rt_key(rt))->sin_addr) &&
1915 ipsendredirects && !srcrt) {
1916 if (rt->rt_ifa &&
1917 (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) ==
1918 ifatoia(rt->rt_ifa)->ia_subnet) {
1919 if (rt->rt_flags & RTF_GATEWAY)
1920 dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
1921 else
1922 dest = ip->ip_dst.s_addr;
1923 /*
1924 * Router requirements says to only send host
1925 * redirects.
1926 */
1927 type = ICMP_REDIRECT;
1928 code = ICMP_REDIRECT_HOST;
1929 #ifdef DIAGNOSTIC
1930 if (ipprintfs)
1931 printf("redirect (%d) to %x\n", code,
1932 (u_int32_t)dest);
1933 #endif
1934 }
1935 }
1936
1937 error = ip_output(m, (struct mbuf *)0, &ipforward_rt,
1938 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)),
1939 (struct ip_moptions *)NULL, (struct socket *)NULL);
1940
1941 if (error)
1942 ipstat.ips_cantforward++;
1943 else {
1944 ipstat.ips_forward++;
1945 if (type)
1946 ipstat.ips_redirectsent++;
1947 else {
1948 if (mcopy) {
1949 #ifdef GATEWAY
1950 if (mcopy->m_flags & M_CANFASTFWD)
1951 ipflow_create(&ipforward_rt, mcopy);
1952 #endif
1953 m_freem(mcopy);
1954 }
1955 return;
1956 }
1957 }
1958 if (mcopy == NULL)
1959 return;
1960
1961 switch (error) {
1962
1963 case 0: /* forwarded, but need redirect */
1964 /* type, code set above */
1965 break;
1966
1967 case ENETUNREACH: /* shouldn't happen, checked above */
1968 case EHOSTUNREACH:
1969 case ENETDOWN:
1970 case EHOSTDOWN:
1971 default:
1972 type = ICMP_UNREACH;
1973 code = ICMP_UNREACH_HOST;
1974 break;
1975
1976 case EMSGSIZE:
1977 type = ICMP_UNREACH;
1978 code = ICMP_UNREACH_NEEDFRAG;
1979
1980 if (ipforward_rt.ro_rt) {
1981
1982 #if defined(IPSEC) || defined(FAST_IPSEC)
1983 /*
1984 * If the packet is routed over IPsec tunnel, tell the
1985 * originator the tunnel MTU.
1986 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
1987 * XXX quickhack!!!
1988 */
1989
1990 struct secpolicy *sp;
1991 int ipsecerror;
1992 size_t ipsechdr;
1993 struct route *ro;
1994
1995 sp = ipsec4_getpolicybyaddr(mcopy,
1996 IPSEC_DIR_OUTBOUND, IP_FORWARDING,
1997 &ipsecerror);
1998 #endif
1999
2000 destmtu = ipforward_rt.ro_rt->rt_ifp->if_mtu;
2001 #if defined(IPSEC) || defined(FAST_IPSEC)
2002 if (sp != NULL) {
2003 /* count IPsec header size */
2004 ipsechdr = ipsec4_hdrsiz(mcopy,
2005 IPSEC_DIR_OUTBOUND, NULL);
2006
2007 /*
2008 * find the correct route for outer IPv4
2009 * header, compute tunnel MTU.
2010 */
2011
2012 if (sp->req != NULL
2013 && sp->req->sav != NULL
2014 && sp->req->sav->sah != NULL) {
2015 ro = &sp->req->sav->sah->sa_route;
2016 if (ro->ro_rt && ro->ro_rt->rt_ifp) {
2017 destmtu =
2018 ro->ro_rt->rt_rmx.rmx_mtu ?
2019 ro->ro_rt->rt_rmx.rmx_mtu :
2020 ro->ro_rt->rt_ifp->if_mtu;
2021 destmtu -= ipsechdr;
2022 }
2023 }
2024
2025 #ifdef IPSEC
2026 key_freesp(sp);
2027 #else
2028 KEY_FREESP(&sp);
2029 #endif
2030 }
2031 #endif /*defined(IPSEC) || defined(FAST_IPSEC)*/
2032 }
2033 ipstat.ips_cantfrag++;
2034 break;
2035
2036 case ENOBUFS:
2037 #if 1
2038 /*
2039 * a router should not generate ICMP_SOURCEQUENCH as
2040 * required in RFC1812 Requirements for IP Version 4 Routers.
2041 * source quench could be a big problem under DoS attacks,
2042 * or if the underlying interface is rate-limited.
2043 */
2044 if (mcopy)
2045 m_freem(mcopy);
2046 return;
2047 #else
2048 type = ICMP_SOURCEQUENCH;
2049 code = 0;
2050 break;
2051 #endif
2052 }
2053 icmp_error(mcopy, type, code, dest, destmtu);
2054 }
2055
2056 void
2057 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
2058 struct mbuf *m)
2059 {
2060
2061 if (inp->inp_socket->so_options & SO_TIMESTAMP) {
2062 struct timeval tv;
2063
2064 microtime(&tv);
2065 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
2066 SCM_TIMESTAMP, SOL_SOCKET);
2067 if (*mp)
2068 mp = &(*mp)->m_next;
2069 }
2070 if (inp->inp_flags & INP_RECVDSTADDR) {
2071 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
2072 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
2073 if (*mp)
2074 mp = &(*mp)->m_next;
2075 }
2076 #ifdef notyet
2077 /*
2078 * XXX
2079 * Moving these out of udp_input() made them even more broken
2080 * than they already were.
2081 * - fenner@parc.xerox.com
2082 */
2083 /* options were tossed already */
2084 if (inp->inp_flags & INP_RECVOPTS) {
2085 *mp = sbcreatecontrol((caddr_t) opts_deleted_above,
2086 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
2087 if (*mp)
2088 mp = &(*mp)->m_next;
2089 }
2090 /* ip_srcroute doesn't do what we want here, need to fix */
2091 if (inp->inp_flags & INP_RECVRETOPTS) {
2092 *mp = sbcreatecontrol((caddr_t) ip_srcroute(),
2093 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
2094 if (*mp)
2095 mp = &(*mp)->m_next;
2096 }
2097 #endif
2098 if (inp->inp_flags & INP_RECVIF) {
2099 struct sockaddr_dl sdl;
2100
2101 sdl.sdl_len = offsetof(struct sockaddr_dl, sdl_data[0]);
2102 sdl.sdl_family = AF_LINK;
2103 sdl.sdl_index = m->m_pkthdr.rcvif ?
2104 m->m_pkthdr.rcvif->if_index : 0;
2105 sdl.sdl_nlen = sdl.sdl_alen = sdl.sdl_slen = 0;
2106 *mp = sbcreatecontrol((caddr_t) &sdl, sdl.sdl_len,
2107 IP_RECVIF, IPPROTO_IP);
2108 if (*mp)
2109 mp = &(*mp)->m_next;
2110 }
2111 }
2112
2113 /*
2114 * sysctl helper routine for net.inet.ip.forwsrcrt.
2115 */
2116 static int
2117 sysctl_net_inet_ip_forwsrcrt(SYSCTLFN_ARGS)
2118 {
2119 int error, tmp;
2120 struct sysctlnode node;
2121
2122 node = *rnode;
2123 tmp = ip_forwsrcrt;
2124 node.sysctl_data = &tmp;
2125 error = sysctl_lookup(SYSCTLFN_CALL(&node));
2126 if (error || newp == NULL)
2127 return (error);
2128
2129 if (kauth_authorize_network(l->l_cred, KAUTH_NETWORK_FORWSRCRT,
2130 0, NULL, NULL, NULL))
2131 return (EPERM);
2132
2133 ip_forwsrcrt = tmp;
2134
2135 return (0);
2136 }
2137
2138 /*
2139 * sysctl helper routine for net.inet.ip.mtudisctimeout. checks the
2140 * range of the new value and tweaks timers if it changes.
2141 */
2142 static int
2143 sysctl_net_inet_ip_pmtudto(SYSCTLFN_ARGS)
2144 {
2145 int error, tmp;
2146 struct sysctlnode node;
2147
2148 node = *rnode;
2149 tmp = ip_mtudisc_timeout;
2150 node.sysctl_data = &tmp;
2151 error = sysctl_lookup(SYSCTLFN_CALL(&node));
2152 if (error || newp == NULL)
2153 return (error);
2154 if (tmp < 0)
2155 return (EINVAL);
2156
2157 ip_mtudisc_timeout = tmp;
2158 rt_timer_queue_change(ip_mtudisc_timeout_q, ip_mtudisc_timeout);
2159
2160 return (0);
2161 }
2162
2163 #ifdef GATEWAY
2164 /*
2165 * sysctl helper routine for net.inet.ip.maxflows. apparently if
2166 * maxflows is even looked up, we "reap flows".
2167 */
2168 static int
2169 sysctl_net_inet_ip_maxflows(SYSCTLFN_ARGS)
2170 {
2171 int s;
2172
2173 s = sysctl_lookup(SYSCTLFN_CALL(rnode));
2174 if (s)
2175 return (s);
2176
2177 s = splsoftnet();
2178 ipflow_reap(0);
2179 splx(s);
2180
2181 return (0);
2182 }
2183 #endif /* GATEWAY */
2184
2185
2186 SYSCTL_SETUP(sysctl_net_inet_ip_setup, "sysctl net.inet.ip subtree setup")
2187 {
2188 extern int subnetsarelocal, hostzeroisbroadcast;
2189
2190 sysctl_createv(clog, 0, NULL, NULL,
2191 CTLFLAG_PERMANENT,
2192 CTLTYPE_NODE, "net", NULL,
2193 NULL, 0, NULL, 0,
2194 CTL_NET, CTL_EOL);
2195 sysctl_createv(clog, 0, NULL, NULL,
2196 CTLFLAG_PERMANENT,
2197 CTLTYPE_NODE, "inet",
2198 SYSCTL_DESCR("PF_INET related settings"),
2199 NULL, 0, NULL, 0,
2200 CTL_NET, PF_INET, CTL_EOL);
2201 sysctl_createv(clog, 0, NULL, NULL,
2202 CTLFLAG_PERMANENT,
2203 CTLTYPE_NODE, "ip",
2204 SYSCTL_DESCR("IPv4 related settings"),
2205 NULL, 0, NULL, 0,
2206 CTL_NET, PF_INET, IPPROTO_IP, CTL_EOL);
2207
2208 sysctl_createv(clog, 0, NULL, NULL,
2209 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2210 CTLTYPE_INT, "forwarding",
2211 SYSCTL_DESCR("Enable forwarding of INET datagrams"),
2212 NULL, 0, &ipforwarding, 0,
2213 CTL_NET, PF_INET, IPPROTO_IP,
2214 IPCTL_FORWARDING, CTL_EOL);
2215 sysctl_createv(clog, 0, NULL, NULL,
2216 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2217 CTLTYPE_INT, "redirect",
2218 SYSCTL_DESCR("Enable sending of ICMP redirect messages"),
2219 NULL, 0, &ipsendredirects, 0,
2220 CTL_NET, PF_INET, IPPROTO_IP,
2221 IPCTL_SENDREDIRECTS, CTL_EOL);
2222 sysctl_createv(clog, 0, NULL, NULL,
2223 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2224 CTLTYPE_INT, "ttl",
2225 SYSCTL_DESCR("Default TTL for an INET datagram"),
2226 NULL, 0, &ip_defttl, 0,
2227 CTL_NET, PF_INET, IPPROTO_IP,
2228 IPCTL_DEFTTL, CTL_EOL);
2229 #ifdef IPCTL_DEFMTU
2230 sysctl_createv(clog, 0, NULL, NULL,
2231 CTLFLAG_PERMANENT /* |CTLFLAG_READWRITE? */,
2232 CTLTYPE_INT, "mtu",
2233 SYSCTL_DESCR("Default MTA for an INET route"),
2234 NULL, 0, &ip_mtu, 0,
2235 CTL_NET, PF_INET, IPPROTO_IP,
2236 IPCTL_DEFMTU, CTL_EOL);
2237 #endif /* IPCTL_DEFMTU */
2238 sysctl_createv(clog, 0, NULL, NULL,
2239 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2240 CTLTYPE_INT, "forwsrcrt",
2241 SYSCTL_DESCR("Enable forwarding of source-routed "
2242 "datagrams"),
2243 sysctl_net_inet_ip_forwsrcrt, 0, &ip_forwsrcrt, 0,
2244 CTL_NET, PF_INET, IPPROTO_IP,
2245 IPCTL_FORWSRCRT, CTL_EOL);
2246 sysctl_createv(clog, 0, NULL, NULL,
2247 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2248 CTLTYPE_INT, "directed-broadcast",
2249 SYSCTL_DESCR("Enable forwarding of broadcast datagrams"),
2250 NULL, 0, &ip_directedbcast, 0,
2251 CTL_NET, PF_INET, IPPROTO_IP,
2252 IPCTL_DIRECTEDBCAST, CTL_EOL);
2253 sysctl_createv(clog, 0, NULL, NULL,
2254 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2255 CTLTYPE_INT, "allowsrcrt",
2256 SYSCTL_DESCR("Accept source-routed datagrams"),
2257 NULL, 0, &ip_allowsrcrt, 0,
2258 CTL_NET, PF_INET, IPPROTO_IP,
2259 IPCTL_ALLOWSRCRT, CTL_EOL);
2260 sysctl_createv(clog, 0, NULL, NULL,
2261 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2262 CTLTYPE_INT, "subnetsarelocal",
2263 SYSCTL_DESCR("Whether logical subnets are considered "
2264 "local"),
2265 NULL, 0, &subnetsarelocal, 0,
2266 CTL_NET, PF_INET, IPPROTO_IP,
2267 IPCTL_SUBNETSARELOCAL, CTL_EOL);
2268 sysctl_createv(clog, 0, NULL, NULL,
2269 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2270 CTLTYPE_INT, "mtudisc",
2271 SYSCTL_DESCR("Use RFC1191 Path MTU Discovery"),
2272 NULL, 0, &ip_mtudisc, 0,
2273 CTL_NET, PF_INET, IPPROTO_IP,
2274 IPCTL_MTUDISC, CTL_EOL);
2275 sysctl_createv(clog, 0, NULL, NULL,
2276 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2277 CTLTYPE_INT, "anonportmin",
2278 SYSCTL_DESCR("Lowest ephemeral port number to assign"),
2279 sysctl_net_inet_ip_ports, 0, &anonportmin, 0,
2280 CTL_NET, PF_INET, IPPROTO_IP,
2281 IPCTL_ANONPORTMIN, CTL_EOL);
2282 sysctl_createv(clog, 0, NULL, NULL,
2283 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2284 CTLTYPE_INT, "anonportmax",
2285 SYSCTL_DESCR("Highest ephemeral port number to assign"),
2286 sysctl_net_inet_ip_ports, 0, &anonportmax, 0,
2287 CTL_NET, PF_INET, IPPROTO_IP,
2288 IPCTL_ANONPORTMAX, CTL_EOL);
2289 sysctl_createv(clog, 0, NULL, NULL,
2290 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2291 CTLTYPE_INT, "mtudisctimeout",
2292 SYSCTL_DESCR("Lifetime of a Path MTU Discovered route"),
2293 sysctl_net_inet_ip_pmtudto, 0, &ip_mtudisc_timeout, 0,
2294 CTL_NET, PF_INET, IPPROTO_IP,
2295 IPCTL_MTUDISCTIMEOUT, CTL_EOL);
2296 #ifdef GATEWAY
2297 sysctl_createv(clog, 0, NULL, NULL,
2298 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2299 CTLTYPE_INT, "maxflows",
2300 SYSCTL_DESCR("Number of flows for fast forwarding"),
2301 sysctl_net_inet_ip_maxflows, 0, &ip_maxflows, 0,
2302 CTL_NET, PF_INET, IPPROTO_IP,
2303 IPCTL_MAXFLOWS, 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 }
Cache object: 6723685a05bdc37ba8a25585dfded10b
|