ip_input.c revision 1.367 1 /* $NetBSD: ip_input.c,v 1.367 2018/02/05 13:34:20 maxv 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 *
49 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
50 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
51 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
52 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
53 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
54 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
55 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
56 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
57 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
58 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
59 * POSSIBILITY OF SUCH DAMAGE.
60 */
61
62 /*
63 * Copyright (c) 1982, 1986, 1988, 1993
64 * The Regents of the University of California. All rights reserved.
65 *
66 * Redistribution and use in source and binary forms, with or without
67 * modification, are permitted provided that the following conditions
68 * are met:
69 * 1. Redistributions of source code must retain the above copyright
70 * notice, this list of conditions and the following disclaimer.
71 * 2. Redistributions in binary form must reproduce the above copyright
72 * notice, this list of conditions and the following disclaimer in the
73 * documentation and/or other materials provided with the distribution.
74 * 3. Neither the name of the University nor the names of its contributors
75 * may be used to endorse or promote products derived from this software
76 * without specific prior written permission.
77 *
78 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
79 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
80 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
81 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
82 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
83 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
84 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
85 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
86 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
87 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
88 * SUCH DAMAGE.
89 *
90 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94
91 */
92
93 #include <sys/cdefs.h>
94 __KERNEL_RCSID(0, "$NetBSD: ip_input.c,v 1.367 2018/02/05 13:34:20 maxv Exp $");
95
96 #ifdef _KERNEL_OPT
97 #include "opt_inet.h"
98 #include "opt_gateway.h"
99 #include "opt_ipsec.h"
100 #include "opt_mrouting.h"
101 #include "opt_mbuftrace.h"
102 #include "opt_inet_csum.h"
103 #include "opt_net_mpsafe.h"
104 #endif
105
106 #include "arp.h"
107
108 #include <sys/param.h>
109 #include <sys/systm.h>
110 #include <sys/cpu.h>
111 #include <sys/mbuf.h>
112 #include <sys/domain.h>
113 #include <sys/protosw.h>
114 #include <sys/socket.h>
115 #include <sys/socketvar.h>
116 #include <sys/errno.h>
117 #include <sys/time.h>
118 #include <sys/kernel.h>
119 #include <sys/pool.h>
120 #include <sys/sysctl.h>
121 #include <sys/kauth.h>
122
123 #include <net/if.h>
124 #include <net/if_dl.h>
125 #include <net/route.h>
126 #include <net/pktqueue.h>
127 #include <net/pfil.h>
128
129 #include <netinet/in.h>
130 #include <netinet/in_systm.h>
131 #include <netinet/ip.h>
132 #include <netinet/in_pcb.h>
133 #include <netinet/in_proto.h>
134 #include <netinet/in_var.h>
135 #include <netinet/ip_var.h>
136 #include <netinet/ip_private.h>
137 #include <netinet/ip_icmp.h>
138 /* just for gif_ttl */
139 #include <netinet/in_gif.h>
140 #include "gif.h"
141 #include <net/if_gre.h>
142 #include "gre.h"
143
144 #ifdef MROUTING
145 #include <netinet/ip_mroute.h>
146 #endif
147 #include <netinet/portalgo.h>
148
149 #ifdef IPSEC
150 #include <netipsec/ipsec.h>
151 #endif
152
153 #ifndef IPFORWARDING
154 #ifdef GATEWAY
155 #define IPFORWARDING 1 /* forward IP packets not for us */
156 #else /* GATEWAY */
157 #define IPFORWARDING 0 /* don't forward IP packets not for us */
158 #endif /* GATEWAY */
159 #endif /* IPFORWARDING */
160 #ifndef IPSENDREDIRECTS
161 #define IPSENDREDIRECTS 1
162 #endif
163 #ifndef IPFORWSRCRT
164 #define IPFORWSRCRT 0 /* forward source-routed packets */
165 #endif
166 #ifndef IPALLOWSRCRT
167 #define IPALLOWSRCRT 0 /* allow source-routed packets */
168 #endif
169 #ifndef IPMTUDISC
170 #define IPMTUDISC 1
171 #endif
172 #ifndef IPMTUDISCTIMEOUT
173 #define IPMTUDISCTIMEOUT (10 * 60) /* as per RFC 1191 */
174 #endif
175
176 /*
177 * Note: DIRECTED_BROADCAST is handled this way so that previous
178 * configuration using this option will Just Work.
179 */
180 #ifndef IPDIRECTEDBCAST
181 #ifdef DIRECTED_BROADCAST
182 #define IPDIRECTEDBCAST 1
183 #else
184 #define IPDIRECTEDBCAST 0
185 #endif /* DIRECTED_BROADCAST */
186 #endif /* IPDIRECTEDBCAST */
187 int ipforwarding = IPFORWARDING;
188 int ipsendredirects = IPSENDREDIRECTS;
189 int ip_defttl = IPDEFTTL;
190 int ip_forwsrcrt = IPFORWSRCRT;
191 int ip_directedbcast = IPDIRECTEDBCAST;
192 int ip_allowsrcrt = IPALLOWSRCRT;
193 int ip_mtudisc = IPMTUDISC;
194 int ip_mtudisc_timeout = IPMTUDISCTIMEOUT;
195
196 int ip_do_randomid = 0;
197
198 /*
199 * XXX - Setting ip_checkinterface mostly implements the receive side of
200 * the Strong ES model described in RFC 1122, but since the routing table
201 * and transmit implementation do not implement the Strong ES model,
202 * setting this to 1 results in an odd hybrid.
203 *
204 * XXX - ip_checkinterface currently must be disabled if you use ipnat
205 * to translate the destination address to another local interface.
206 *
207 * XXX - ip_checkinterface must be disabled if you add IP aliases
208 * to the loopback interface instead of the interface where the
209 * packets for those addresses are received.
210 */
211 static int ip_checkinterface __read_mostly = 0;
212
213 struct rttimer_queue *ip_mtudisc_timeout_q = NULL;
214
215 pktqueue_t * ip_pktq __read_mostly;
216 pfil_head_t * inet_pfil_hook __read_mostly;
217 ipid_state_t * ip_ids __read_mostly;
218 percpu_t * ipstat_percpu __read_mostly;
219
220 static percpu_t *ipforward_rt_percpu __cacheline_aligned;
221
222 uint16_t ip_id;
223
224 #ifdef INET_CSUM_COUNTERS
225 #include <sys/device.h>
226
227 struct evcnt ip_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
228 NULL, "inet", "hwcsum bad");
229 struct evcnt ip_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
230 NULL, "inet", "hwcsum ok");
231 struct evcnt ip_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
232 NULL, "inet", "swcsum");
233
234 #define INET_CSUM_COUNTER_INCR(ev) (ev)->ev_count++
235
236 EVCNT_ATTACH_STATIC(ip_hwcsum_bad);
237 EVCNT_ATTACH_STATIC(ip_hwcsum_ok);
238 EVCNT_ATTACH_STATIC(ip_swcsum);
239
240 #else
241
242 #define INET_CSUM_COUNTER_INCR(ev) /* nothing */
243
244 #endif /* INET_CSUM_COUNTERS */
245
246 /*
247 * Used to save the IP options in case a protocol wants to respond
248 * to an incoming packet over the same route if the packet got here
249 * using IP source routing. This allows connection establishment and
250 * maintenance when the remote end is on a network that is not known
251 * to us.
252 */
253 struct ip_srcrt {
254 int isr_nhops; /* number of hops */
255 struct in_addr isr_dst; /* final destination */
256 char isr_nop; /* one NOP to align */
257 char isr_hdr[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN & OFFSET */
258 struct in_addr isr_routes[MAX_IPOPTLEN/sizeof(struct in_addr)];
259 };
260
261 static int ip_drainwanted;
262
263 static void save_rte(struct mbuf *, u_char *, struct in_addr);
264
265 #ifdef MBUFTRACE
266 struct mowner ip_rx_mowner = MOWNER_INIT("internet", "rx");
267 struct mowner ip_tx_mowner = MOWNER_INIT("internet", "tx");
268 #endif
269
270 static void ipintr(void *);
271 static void ip_input(struct mbuf *);
272 static void ip_forward(struct mbuf *, int, struct ifnet *);
273 static bool ip_dooptions(struct mbuf *);
274 static struct in_ifaddr *ip_rtaddr(struct in_addr, struct psref *);
275 static void sysctl_net_inet_ip_setup(struct sysctllog **);
276
277 static struct in_ifaddr *ip_match_our_address(struct ifnet *, struct ip *,
278 int *);
279 static struct in_ifaddr *ip_match_our_address_broadcast(struct ifnet *,
280 struct ip *);
281
282 #ifdef NET_MPSAFE
283 #define SOFTNET_LOCK() mutex_enter(softnet_lock)
284 #define SOFTNET_UNLOCK() mutex_exit(softnet_lock)
285 #else
286 #define SOFTNET_LOCK() KASSERT(mutex_owned(softnet_lock))
287 #define SOFTNET_UNLOCK() KASSERT(mutex_owned(softnet_lock))
288 #endif
289
290 /*
291 * IP initialization: fill in IP protocol switch table.
292 * All protocols not implemented in kernel go to raw IP protocol handler.
293 */
294 void
295 ip_init(void)
296 {
297 const struct protosw *pr;
298
299 in_init();
300 sysctl_net_inet_ip_setup(NULL);
301
302 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
303 KASSERT(pr != NULL);
304
305 ip_pktq = pktq_create(IFQ_MAXLEN, ipintr, NULL);
306 KASSERT(ip_pktq != NULL);
307
308 for (u_int i = 0; i < IPPROTO_MAX; i++) {
309 ip_protox[i] = pr - inetsw;
310 }
311 for (pr = inetdomain.dom_protosw;
312 pr < inetdomain.dom_protoswNPROTOSW; pr++)
313 if (pr->pr_domain->dom_family == PF_INET &&
314 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
315 ip_protox[pr->pr_protocol] = pr - inetsw;
316
317 ip_reass_init();
318
319 ip_ids = ip_id_init();
320 ip_id = time_uptime & 0xfffff;
321
322 #ifdef GATEWAY
323 ipflow_init();
324 #endif
325
326 /* Register our Packet Filter hook. */
327 inet_pfil_hook = pfil_head_create(PFIL_TYPE_AF, (void *)AF_INET);
328 KASSERT(inet_pfil_hook != NULL);
329
330 #ifdef MBUFTRACE
331 MOWNER_ATTACH(&ip_tx_mowner);
332 MOWNER_ATTACH(&ip_rx_mowner);
333 #endif /* MBUFTRACE */
334
335 ipstat_percpu = percpu_alloc(sizeof(uint64_t) * IP_NSTATS);
336 ipforward_rt_percpu = percpu_alloc(sizeof(struct route));
337 ip_mtudisc_timeout_q = rt_timer_queue_create(ip_mtudisc_timeout);
338 }
339
340 static struct in_ifaddr *
341 ip_match_our_address(struct ifnet *ifp, struct ip *ip, int *downmatch)
342 {
343 struct in_ifaddr *ia = NULL;
344 int checkif;
345
346 /*
347 * Enable a consistency check between the destination address
348 * and the arrival interface for a unicast packet (the RFC 1122
349 * strong ES model) if IP forwarding is disabled and the packet
350 * is not locally generated.
351 *
352 * XXX - Checking also should be disabled if the destination
353 * address is ipnat'ed to a different interface.
354 *
355 * XXX - Checking is incompatible with IP aliases added
356 * to the loopback interface instead of the interface where
357 * the packets are received.
358 *
359 * XXX - We need to add a per ifaddr flag for this so that
360 * we get finer grain control.
361 */
362 checkif = ip_checkinterface && (ipforwarding == 0) &&
363 (ifp->if_flags & IFF_LOOPBACK) == 0;
364
365 IN_ADDRHASH_READER_FOREACH(ia, ip->ip_dst.s_addr) {
366 if (in_hosteq(ia->ia_addr.sin_addr, ip->ip_dst)) {
367 if (ia->ia4_flags & IN_IFF_NOTREADY)
368 continue;
369 if (checkif && ia->ia_ifp != ifp)
370 continue;
371 if ((ia->ia_ifp->if_flags & IFF_UP) == 0) {
372 (*downmatch)++;
373 continue;
374 }
375 if (ia->ia4_flags & IN_IFF_DETACHED &&
376 (ifp->if_flags & IFF_LOOPBACK) == 0)
377 continue;
378 break;
379 }
380 }
381
382 return ia;
383 }
384
385 static struct in_ifaddr *
386 ip_match_our_address_broadcast(struct ifnet *ifp, struct ip *ip)
387 {
388 struct in_ifaddr *ia = NULL;
389 struct ifaddr *ifa;
390
391 IFADDR_READER_FOREACH(ifa, ifp) {
392 if (ifa->ifa_addr->sa_family != AF_INET)
393 continue;
394 ia = ifatoia(ifa);
395 if (ia->ia4_flags & IN_IFF_NOTREADY)
396 continue;
397 if (ia->ia4_flags & IN_IFF_DETACHED &&
398 (ifp->if_flags & IFF_LOOPBACK) == 0)
399 continue;
400 if (in_hosteq(ip->ip_dst, ia->ia_broadaddr.sin_addr) ||
401 in_hosteq(ip->ip_dst, ia->ia_netbroadcast) ||
402 /*
403 * Look for all-0's host part (old broadcast addr),
404 * either for subnet or net.
405 */
406 ip->ip_dst.s_addr == ia->ia_subnet ||
407 ip->ip_dst.s_addr == ia->ia_net)
408 goto matched;
409 /*
410 * An interface with IP address zero accepts
411 * all packets that arrive on that interface.
412 */
413 if (in_nullhost(ia->ia_addr.sin_addr))
414 goto matched;
415 }
416 ia = NULL;
417
418 matched:
419 return ia;
420 }
421
422 /*
423 * IP software interrupt routine.
424 */
425 static void
426 ipintr(void *arg __unused)
427 {
428 struct mbuf *m;
429
430 KASSERT(cpu_softintr_p());
431
432 SOFTNET_LOCK_UNLESS_NET_MPSAFE();
433 while ((m = pktq_dequeue(ip_pktq)) != NULL) {
434 ip_input(m);
435 }
436 SOFTNET_UNLOCK_UNLESS_NET_MPSAFE();
437 }
438
439 /*
440 * IP input routine. Checksum and byte swap header. If fragmented
441 * try to reassemble. Process options. Pass to next level.
442 */
443 static void
444 ip_input(struct mbuf *m)
445 {
446 struct ip *ip = NULL;
447 struct in_ifaddr *ia = NULL;
448 int hlen = 0, len;
449 int downmatch;
450 int srcrt = 0;
451 ifnet_t *ifp;
452 struct psref psref;
453 int s;
454
455 KASSERTMSG(cpu_softintr_p(), "ip_input: not in the software "
456 "interrupt handler; synchronization assumptions violated");
457
458 MCLAIM(m, &ip_rx_mowner);
459 KASSERT((m->m_flags & M_PKTHDR) != 0);
460
461 ifp = m_get_rcvif_psref(m, &psref);
462 if (__predict_false(ifp == NULL))
463 goto out;
464
465 /*
466 * If no IP addresses have been set yet but the interfaces
467 * are receiving, can't do anything with incoming packets yet.
468 * Note: we pre-check without locks held.
469 */
470 if (IN_ADDRLIST_READER_EMPTY())
471 goto out;
472 IP_STATINC(IP_STAT_TOTAL);
473
474 /*
475 * If the IP header is not aligned, slurp it up into a new
476 * mbuf with space for link headers, in the event we forward
477 * it. Otherwise, if it is aligned, make sure the entire
478 * base IP header is in the first mbuf of the chain.
479 */
480 if (IP_HDR_ALIGNED_P(mtod(m, void *)) == 0) {
481 if ((m = m_copyup(m, sizeof(struct ip),
482 (max_linkhdr + 3) & ~3)) == NULL) {
483 /* XXXJRT new stat, please */
484 IP_STATINC(IP_STAT_TOOSMALL);
485 goto out;
486 }
487 } else if (__predict_false(m->m_len < sizeof(struct ip))) {
488 if ((m = m_pullup(m, sizeof(struct ip))) == NULL) {
489 IP_STATINC(IP_STAT_TOOSMALL);
490 goto out;
491 }
492 }
493 ip = mtod(m, struct ip *);
494 if (ip->ip_v != IPVERSION) {
495 IP_STATINC(IP_STAT_BADVERS);
496 goto out;
497 }
498 hlen = ip->ip_hl << 2;
499 if (hlen < sizeof(struct ip)) { /* minimum header length */
500 IP_STATINC(IP_STAT_BADHLEN);
501 goto out;
502 }
503 if (hlen > m->m_len) {
504 if ((m = m_pullup(m, hlen)) == NULL) {
505 IP_STATINC(IP_STAT_BADHLEN);
506 goto out;
507 }
508 ip = mtod(m, struct ip *);
509 }
510
511 /*
512 * RFC1122: packets with a multicast source address are
513 * not allowed.
514 */
515 if (IN_MULTICAST(ip->ip_src.s_addr)) {
516 IP_STATINC(IP_STAT_BADADDR);
517 goto out;
518 }
519
520 /* 127/8 must not appear on wire - RFC1122 */
521 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
522 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
523 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
524 IP_STATINC(IP_STAT_BADADDR);
525 goto out;
526 }
527 }
528
529 switch (m->m_pkthdr.csum_flags &
530 ((ifp->if_csum_flags_rx & M_CSUM_IPv4) | M_CSUM_IPv4_BAD)) {
531 case M_CSUM_IPv4|M_CSUM_IPv4_BAD:
532 INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad);
533 IP_STATINC(IP_STAT_BADSUM);
534 goto out;
535
536 case M_CSUM_IPv4:
537 /* Checksum was okay. */
538 INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok);
539 break;
540
541 default:
542 /*
543 * Must compute it ourselves. Maybe skip checksum on
544 * loopback interfaces.
545 */
546 if (__predict_true(!(ifp->if_flags & IFF_LOOPBACK) ||
547 ip_do_loopback_cksum)) {
548 INET_CSUM_COUNTER_INCR(&ip_swcsum);
549 if (in_cksum(m, hlen) != 0) {
550 IP_STATINC(IP_STAT_BADSUM);
551 goto out;
552 }
553 }
554 break;
555 }
556
557 /* Retrieve the packet length. */
558 len = ntohs(ip->ip_len);
559
560 /*
561 * Check for additional length bogosity
562 */
563 if (len < hlen) {
564 IP_STATINC(IP_STAT_BADLEN);
565 goto out;
566 }
567
568 /*
569 * Check that the amount of data in the buffers is at least as much
570 * as the IP header would have us expect. Trim mbufs if longer than
571 * we expect. Drop packet if shorter than we expect.
572 */
573 if (m->m_pkthdr.len < len) {
574 IP_STATINC(IP_STAT_TOOSHORT);
575 goto out;
576 }
577 if (m->m_pkthdr.len > len) {
578 if (m->m_len == m->m_pkthdr.len) {
579 m->m_len = len;
580 m->m_pkthdr.len = len;
581 } else
582 m_adj(m, len - m->m_pkthdr.len);
583 }
584
585 /*
586 * Assume that we can create a fast-forward IP flow entry
587 * based on this packet.
588 */
589 m->m_flags |= M_CANFASTFWD;
590
591 /*
592 * Run through list of hooks for input packets. If there are any
593 * filters which require that additional packets in the flow are
594 * not fast-forwarded, they must clear the M_CANFASTFWD flag.
595 * Note that filters must _never_ set this flag, as another filter
596 * in the list may have previously cleared it.
597 */
598 #if defined(IPSEC)
599 if (!ipsec_used || !ipsec_indone(m))
600 #else
601 if (1)
602 #endif
603 {
604 struct in_addr odst = ip->ip_dst;
605 bool freed;
606
607 freed = pfil_run_hooks(inet_pfil_hook, &m, ifp, PFIL_IN) != 0;
608 if (freed || m == NULL) {
609 m = NULL;
610 goto out;
611 }
612 ip = mtod(m, struct ip *);
613 hlen = ip->ip_hl << 2;
614
615 /*
616 * XXX The setting of "srcrt" here is to prevent ip_forward()
617 * from generating ICMP redirects for packets that have
618 * been redirected by a hook back out on to the same LAN that
619 * they came from and is not an indication that the packet
620 * is being inffluenced by source routing options. This
621 * allows things like
622 * "rdr tlp0 0/0 port 80 -> 1.1.1.200 3128 tcp"
623 * where tlp0 is both on the 1.1.1.0/24 network and is the
624 * default route for hosts on 1.1.1.0/24. Of course this
625 * also requires a "map tlp0 ..." to complete the story.
626 * One might argue whether or not this kind of network config.
627 * should be supported in this manner...
628 */
629 srcrt = (odst.s_addr != ip->ip_dst.s_addr);
630 }
631
632 #ifdef ALTQ
633 /* XXX Temporary until ALTQ is changed to use a pfil hook */
634 if (altq_input) {
635 SOFTNET_LOCK();
636 if ((*altq_input)(m, AF_INET) == 0) {
637 /* Packet dropped by traffic conditioner. */
638 SOFTNET_UNLOCK();
639 m = NULL;
640 goto out;
641 }
642 SOFTNET_UNLOCK();
643 }
644 #endif
645
646 /*
647 * Process options and, if not destined for us,
648 * ship it on. ip_dooptions returns 1 when an
649 * error was detected (causing an icmp message
650 * to be sent and the original packet to be freed).
651 */
652 if (hlen > sizeof(struct ip) && ip_dooptions(m)) {
653 m = NULL;
654 goto out;
655 }
656
657 /*
658 * Check our list of addresses, to see if the packet is for us.
659 *
660 * Traditional 4.4BSD did not consult IFF_UP at all.
661 * The behavior here is to treat addresses on !IFF_UP interface
662 * or IN_IFF_NOTREADY addresses as not mine.
663 */
664 downmatch = 0;
665 s = pserialize_read_enter();
666 ia = ip_match_our_address(ifp, ip, &downmatch);
667 if (ia != NULL) {
668 pserialize_read_exit(s);
669 goto ours;
670 }
671
672 if (ifp->if_flags & IFF_BROADCAST) {
673 ia = ip_match_our_address_broadcast(ifp, ip);
674 if (ia != NULL) {
675 pserialize_read_exit(s);
676 goto ours;
677 }
678 }
679 pserialize_read_exit(s);
680
681 if (IN_MULTICAST(ip->ip_dst.s_addr)) {
682 #ifdef MROUTING
683 extern struct socket *ip_mrouter;
684
685 if (ip_mrouter) {
686 /*
687 * If we are acting as a multicast router, all
688 * incoming multicast packets are passed to the
689 * kernel-level multicast forwarding function.
690 * The packet is returned (relatively) intact; if
691 * ip_mforward() returns a non-zero value, the packet
692 * must be discarded, else it may be accepted below.
693 *
694 * (The IP ident field is put in the same byte order
695 * as expected when ip_mforward() is called from
696 * ip_output().)
697 */
698 SOFTNET_LOCK();
699 if (ip_mforward(m, ifp) != 0) {
700 SOFTNET_UNLOCK();
701 IP_STATINC(IP_STAT_CANTFORWARD);
702 goto out;
703 }
704 SOFTNET_UNLOCK();
705
706 /*
707 * The process-level routing demon needs to receive
708 * all multicast IGMP packets, whether or not this
709 * host belongs to their destination groups.
710 */
711 if (ip->ip_p == IPPROTO_IGMP) {
712 goto ours;
713 }
714 IP_STATINC(IP_STAT_CANTFORWARD);
715 }
716 #endif
717 /*
718 * See if we belong to the destination multicast group on the
719 * arrival interface.
720 */
721 if (!in_multi_group(ip->ip_dst, ifp, 0)) {
722 IP_STATINC(IP_STAT_CANTFORWARD);
723 goto out;
724 }
725 goto ours;
726 }
727 if (ip->ip_dst.s_addr == INADDR_BROADCAST ||
728 in_nullhost(ip->ip_dst))
729 goto ours;
730
731 /*
732 * Not for us; forward if possible and desirable.
733 */
734 if (ipforwarding == 0) {
735 m_put_rcvif_psref(ifp, &psref);
736 IP_STATINC(IP_STAT_CANTFORWARD);
737 m_freem(m);
738 } else {
739 /*
740 * If ip_dst matched any of my address on !IFF_UP interface,
741 * and there's no IFF_UP interface that matches ip_dst,
742 * send icmp unreach. Forwarding it will result in in-kernel
743 * forwarding loop till TTL goes to 0.
744 */
745 if (downmatch) {
746 m_put_rcvif_psref(ifp, &psref);
747 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
748 IP_STATINC(IP_STAT_CANTFORWARD);
749 return;
750 }
751 #ifdef IPSEC
752 /* Check the security policy (SP) for the packet */
753 if (ipsec_used) {
754 if (ipsec4_input(m, IP_FORWARDING |
755 (ip_directedbcast ? IP_ALLOWBROADCAST : 0)) != 0) {
756 goto out;
757 }
758 }
759 #endif
760 ip_forward(m, srcrt, ifp);
761 m_put_rcvif_psref(ifp, &psref);
762 }
763 return;
764
765 ours:
766 m_put_rcvif_psref(ifp, &psref);
767 ifp = NULL;
768
769 /*
770 * If offset or IP_MF are set, must reassemble.
771 */
772 if (ip->ip_off & ~htons(IP_DF|IP_RF)) {
773 /*
774 * Pass to IP reassembly mechanism.
775 */
776 if (ip_reass_packet(&m, ip) != 0) {
777 /* Failed; invalid fragment(s) or packet. */
778 goto out;
779 }
780 if (m == NULL) {
781 /* More fragments should come; silently return. */
782 goto out;
783 }
784 /*
785 * Reassembly is done, we have the final packet.
786 * Update cached data in local variable(s).
787 */
788 ip = mtod(m, struct ip *);
789 hlen = ip->ip_hl << 2;
790 }
791
792 #ifdef IPSEC
793 /*
794 * Enforce IPsec policy checking if we are seeing last header.
795 * Note that we do not visit this with protocols with PCB layer
796 * code - like UDP/TCP/raw IP.
797 */
798 if (ipsec_used &&
799 (inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) {
800 if (ipsec4_input(m, 0) != 0) {
801 goto out;
802 }
803 }
804 #endif
805
806 /*
807 * Switch out to protocol's input routine.
808 */
809 #if IFA_STATS
810 if (ia && ip) {
811 struct in_ifaddr *_ia;
812 /*
813 * Keep a reference from ip_match_our_address with psref
814 * is expensive, so explore ia here again.
815 */
816 s = pserialize_read_enter();
817 _ia = in_get_ia(ip->ip_dst);
818 _ia->ia_ifa.ifa_data.ifad_inbytes += ntohs(ip->ip_len);
819 pserialize_read_exit(s);
820 }
821 #endif
822 IP_STATINC(IP_STAT_DELIVERED);
823
824 const int off = hlen, nh = ip->ip_p;
825
826 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh);
827 return;
828
829 out:
830 m_put_rcvif_psref(ifp, &psref);
831 if (m != NULL)
832 m_freem(m);
833 }
834
835 /*
836 * IP timer processing.
837 */
838 void
839 ip_slowtimo(void)
840 {
841
842 SOFTNET_KERNEL_LOCK_UNLESS_NET_MPSAFE();
843
844 ip_reass_slowtimo();
845
846 SOFTNET_KERNEL_UNLOCK_UNLESS_NET_MPSAFE();
847 }
848
849 /*
850 * IP drain processing.
851 */
852 void
853 ip_drain(void)
854 {
855
856 KERNEL_LOCK(1, NULL);
857 ip_reass_drain();
858 KERNEL_UNLOCK_ONE(NULL);
859 }
860
861 /*
862 * ip_dooptions: perform option processing on a datagram, possibly discarding
863 * it if bad options are encountered, or forwarding it if source-routed.
864 *
865 * => Returns true if packet has been forwarded/freed.
866 * => Returns false if the packet should be processed further.
867 */
868 static bool
869 ip_dooptions(struct mbuf *m)
870 {
871 struct ip *ip = mtod(m, struct ip *);
872 u_char *cp, *cp0;
873 struct ip_timestamp *ipt;
874 struct in_ifaddr *ia;
875 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0;
876 struct in_addr dst;
877 n_time ntime;
878 struct ifaddr *ifa = NULL;
879 int s;
880
881 dst = ip->ip_dst;
882 cp = (u_char *)(ip + 1);
883 cnt = (ip->ip_hl << 2) - sizeof(struct ip);
884 for (; cnt > 0; cnt -= optlen, cp += optlen) {
885 opt = cp[IPOPT_OPTVAL];
886 if (opt == IPOPT_EOL)
887 break;
888 if (opt == IPOPT_NOP)
889 optlen = 1;
890 else {
891 if (cnt < IPOPT_OLEN + sizeof(*cp)) {
892 code = &cp[IPOPT_OLEN] - (u_char *)ip;
893 goto bad;
894 }
895 optlen = cp[IPOPT_OLEN];
896 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
897 code = &cp[IPOPT_OLEN] - (u_char *)ip;
898 goto bad;
899 }
900 }
901 switch (opt) {
902
903 default:
904 break;
905
906 /*
907 * Source routing with record.
908 * Find interface with current destination address.
909 * If none on this machine then drop if strictly routed,
910 * or do nothing if loosely routed.
911 * Record interface address and bring up next address
912 * component. If strictly routed make sure next
913 * address is on directly accessible net.
914 */
915 case IPOPT_LSRR:
916 case IPOPT_SSRR: {
917 struct psref psref;
918 struct sockaddr_in ipaddr = {
919 .sin_len = sizeof(ipaddr),
920 .sin_family = AF_INET,
921 };
922
923 if (ip_allowsrcrt == 0) {
924 type = ICMP_UNREACH;
925 code = ICMP_UNREACH_NET_PROHIB;
926 goto bad;
927 }
928 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
929 code = &cp[IPOPT_OLEN] - (u_char *)ip;
930 goto bad;
931 }
932 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
933 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
934 goto bad;
935 }
936 ipaddr.sin_addr = ip->ip_dst;
937
938 s = pserialize_read_enter();
939 ifa = ifa_ifwithaddr(sintosa(&ipaddr));
940 if (ifa == NULL) {
941 pserialize_read_exit(s);
942 if (opt == IPOPT_SSRR) {
943 type = ICMP_UNREACH;
944 code = ICMP_UNREACH_SRCFAIL;
945 goto bad;
946 }
947 /*
948 * Loose routing, and not at next destination
949 * yet; nothing to do except forward.
950 */
951 break;
952 }
953 pserialize_read_exit(s);
954
955 off--; /* 0 origin */
956 if ((off + sizeof(struct in_addr)) > optlen) {
957 /*
958 * End of source route. Should be for us.
959 */
960 save_rte(m, cp, ip->ip_src);
961 break;
962 }
963 /*
964 * locate outgoing interface
965 */
966 memcpy((void *)&ipaddr.sin_addr, (void *)(cp + off),
967 sizeof(ipaddr.sin_addr));
968 if (opt == IPOPT_SSRR) {
969 ifa = ifa_ifwithladdr_psref(sintosa(&ipaddr),
970 &psref);
971 if (ifa != NULL)
972 ia = ifatoia(ifa);
973 else
974 ia = NULL;
975 } else {
976 ia = ip_rtaddr(ipaddr.sin_addr, &psref);
977 }
978 if (ia == NULL) {
979 type = ICMP_UNREACH;
980 code = ICMP_UNREACH_SRCFAIL;
981 goto bad;
982 }
983 ip->ip_dst = ipaddr.sin_addr;
984 bcopy((void *)&ia->ia_addr.sin_addr,
985 (void *)(cp + off), sizeof(struct in_addr));
986 ia4_release(ia, &psref);
987 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
988 /*
989 * Let ip_intr's mcast routing check handle mcast pkts
990 */
991 forward = !IN_MULTICAST(ip->ip_dst.s_addr);
992 break;
993 }
994
995 case IPOPT_RR: {
996 struct psref psref;
997 struct sockaddr_in ipaddr = {
998 .sin_len = sizeof(ipaddr),
999 .sin_family = AF_INET,
1000 };
1001
1002 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1003 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1004 goto bad;
1005 }
1006 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1007 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1008 goto bad;
1009 }
1010 /*
1011 * If no space remains, ignore.
1012 */
1013 off--; /* 0 origin */
1014 if ((off + sizeof(struct in_addr)) > optlen)
1015 break;
1016 memcpy((void *)&ipaddr.sin_addr, (void *)(&ip->ip_dst),
1017 sizeof(ipaddr.sin_addr));
1018 /*
1019 * locate outgoing interface; if we're the destination,
1020 * use the incoming interface (should be same).
1021 */
1022 ifa = ifa_ifwithaddr_psref(sintosa(&ipaddr), &psref);
1023 if (ifa == NULL) {
1024 ia = ip_rtaddr(ipaddr.sin_addr, &psref);
1025 if (ia == NULL) {
1026 type = ICMP_UNREACH;
1027 code = ICMP_UNREACH_HOST;
1028 goto bad;
1029 }
1030 } else {
1031 ia = ifatoia(ifa);
1032 }
1033 bcopy((void *)&ia->ia_addr.sin_addr,
1034 (void *)(cp + off), sizeof(struct in_addr));
1035 ia4_release(ia, &psref);
1036 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1037 break;
1038 }
1039
1040 case IPOPT_TS:
1041 code = cp - (u_char *)ip;
1042 ipt = (struct ip_timestamp *)cp;
1043 if (ipt->ipt_len < 4 || ipt->ipt_len > 40) {
1044 code = (u_char *)&ipt->ipt_len - (u_char *)ip;
1045 goto bad;
1046 }
1047 if (ipt->ipt_ptr < 5) {
1048 code = (u_char *)&ipt->ipt_ptr - (u_char *)ip;
1049 goto bad;
1050 }
1051 if (ipt->ipt_ptr > ipt->ipt_len - sizeof(int32_t)) {
1052 if (++ipt->ipt_oflw == 0) {
1053 code = (u_char *)&ipt->ipt_ptr -
1054 (u_char *)ip;
1055 goto bad;
1056 }
1057 break;
1058 }
1059 cp0 = (cp + ipt->ipt_ptr - 1);
1060 switch (ipt->ipt_flg) {
1061
1062 case IPOPT_TS_TSONLY:
1063 break;
1064
1065 case IPOPT_TS_TSANDADDR: {
1066 struct ifnet *rcvif;
1067 int _s, _ss;
1068 struct sockaddr_in ipaddr = {
1069 .sin_len = sizeof(ipaddr),
1070 .sin_family = AF_INET,
1071 };
1072
1073 if (ipt->ipt_ptr - 1 + sizeof(n_time) +
1074 sizeof(struct in_addr) > ipt->ipt_len) {
1075 code = (u_char *)&ipt->ipt_ptr -
1076 (u_char *)ip;
1077 goto bad;
1078 }
1079 ipaddr.sin_addr = dst;
1080 _ss = pserialize_read_enter();
1081 rcvif = m_get_rcvif(m, &_s);
1082 if (__predict_true(rcvif != NULL)) {
1083 ifa = ifaof_ifpforaddr(sintosa(&ipaddr),
1084 rcvif);
1085 }
1086 m_put_rcvif(rcvif, &_s);
1087 if (ifa == NULL) {
1088 pserialize_read_exit(_ss);
1089 break;
1090 }
1091 ia = ifatoia(ifa);
1092 bcopy(&ia->ia_addr.sin_addr,
1093 cp0, sizeof(struct in_addr));
1094 pserialize_read_exit(_ss);
1095 ipt->ipt_ptr += sizeof(struct in_addr);
1096 break;
1097 }
1098
1099 case IPOPT_TS_PRESPEC: {
1100 struct sockaddr_in ipaddr = {
1101 .sin_len = sizeof(ipaddr),
1102 .sin_family = AF_INET,
1103 };
1104
1105 if (ipt->ipt_ptr - 1 + sizeof(n_time) +
1106 sizeof(struct in_addr) > ipt->ipt_len) {
1107 code = (u_char *)&ipt->ipt_ptr -
1108 (u_char *)ip;
1109 goto bad;
1110 }
1111 memcpy(&ipaddr.sin_addr, cp0,
1112 sizeof(struct in_addr));
1113 s = pserialize_read_enter();
1114 ifa = ifa_ifwithaddr(sintosa(&ipaddr));
1115 if (ifa == NULL) {
1116 pserialize_read_exit(s);
1117 continue;
1118 }
1119 pserialize_read_exit(s);
1120 ipt->ipt_ptr += sizeof(struct in_addr);
1121 break;
1122 }
1123
1124 default:
1125 /* XXX can't take &ipt->ipt_flg */
1126 code = (u_char *)&ipt->ipt_ptr -
1127 (u_char *)ip + 1;
1128 goto bad;
1129 }
1130 ntime = iptime();
1131 cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */
1132 memmove((char *)cp + ipt->ipt_ptr - 1, cp0,
1133 sizeof(n_time));
1134 ipt->ipt_ptr += sizeof(n_time);
1135 }
1136 }
1137 if (forward) {
1138 struct ifnet *rcvif;
1139 struct psref _psref;
1140
1141 if (ip_forwsrcrt == 0) {
1142 type = ICMP_UNREACH;
1143 code = ICMP_UNREACH_SRCFAIL;
1144 goto bad;
1145 }
1146
1147 rcvif = m_get_rcvif_psref(m, &_psref);
1148 if (__predict_false(rcvif == NULL)) {
1149 type = ICMP_UNREACH;
1150 code = ICMP_UNREACH_HOST;
1151 goto bad;
1152 }
1153 ip_forward(m, 1, rcvif);
1154 m_put_rcvif_psref(rcvif, &_psref);
1155 return true;
1156 }
1157 return false;
1158 bad:
1159 icmp_error(m, type, code, 0, 0);
1160 IP_STATINC(IP_STAT_BADOPTIONS);
1161 return true;
1162 }
1163
1164 /*
1165 * ip_rtaddr: given address of next destination (final or next hop),
1166 * return internet address info of interface to be used to get there.
1167 */
1168 static struct in_ifaddr *
1169 ip_rtaddr(struct in_addr dst, struct psref *psref)
1170 {
1171 struct rtentry *rt;
1172 union {
1173 struct sockaddr dst;
1174 struct sockaddr_in dst4;
1175 } u;
1176 struct route *ro;
1177
1178 sockaddr_in_init(&u.dst4, &dst, 0);
1179
1180 ro = percpu_getref(ipforward_rt_percpu);
1181 rt = rtcache_lookup(ro, &u.dst);
1182 if (rt == NULL) {
1183 percpu_putref(ipforward_rt_percpu);
1184 return NULL;
1185 }
1186
1187 ia4_acquire(ifatoia(rt->rt_ifa), psref);
1188 rtcache_unref(rt, ro);
1189 percpu_putref(ipforward_rt_percpu);
1190
1191 return ifatoia(rt->rt_ifa);
1192 }
1193
1194 /*
1195 * save_rte: save incoming source route for use in replies, to be picked
1196 * up later by ip_srcroute if the receiver is interested.
1197 */
1198 static void
1199 save_rte(struct mbuf *m, u_char *option, struct in_addr dst)
1200 {
1201 struct ip_srcrt *isr;
1202 struct m_tag *mtag;
1203 unsigned olen;
1204
1205 olen = option[IPOPT_OLEN];
1206 if (olen > sizeof(isr->isr_hdr) + sizeof(isr->isr_routes))
1207 return;
1208
1209 mtag = m_tag_get(PACKET_TAG_SRCROUTE, sizeof(*isr), M_NOWAIT);
1210 if (mtag == NULL)
1211 return;
1212 isr = (struct ip_srcrt *)(mtag + 1);
1213
1214 memcpy(isr->isr_hdr, option, olen);
1215 isr->isr_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1216 isr->isr_dst = dst;
1217 m_tag_prepend(m, mtag);
1218 }
1219
1220 /*
1221 * Retrieve incoming source route for use in replies,
1222 * in the same form used by setsockopt.
1223 * The first hop is placed before the options, will be removed later.
1224 */
1225 struct mbuf *
1226 ip_srcroute(struct mbuf *m0)
1227 {
1228 struct in_addr *p, *q;
1229 struct mbuf *m;
1230 struct ip_srcrt *isr;
1231 struct m_tag *mtag;
1232
1233 mtag = m_tag_find(m0, PACKET_TAG_SRCROUTE, NULL);
1234 if (mtag == NULL)
1235 return NULL;
1236 isr = (struct ip_srcrt *)(mtag + 1);
1237
1238 if (isr->isr_nhops == 0)
1239 return NULL;
1240
1241 m = m_get(M_DONTWAIT, MT_SOOPTS);
1242 if (m == NULL)
1243 return NULL;
1244
1245 MCLAIM(m, &inetdomain.dom_mowner);
1246 #define OPTSIZ (sizeof(isr->isr_nop) + sizeof(isr->isr_hdr))
1247
1248 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + header) */
1249 m->m_len = (isr->isr_nhops + 1) * sizeof(struct in_addr) + OPTSIZ;
1250
1251 /*
1252 * First save first hop for return route
1253 */
1254 p = &(isr->isr_routes[isr->isr_nhops - 1]);
1255 *(mtod(m, struct in_addr *)) = *p--;
1256
1257 /*
1258 * Copy option fields and padding (nop) to mbuf.
1259 */
1260 isr->isr_nop = IPOPT_NOP;
1261 isr->isr_hdr[IPOPT_OFFSET] = IPOPT_MINOFF;
1262 memmove(mtod(m, char *) + sizeof(struct in_addr), &isr->isr_nop,
1263 OPTSIZ);
1264 q = (struct in_addr *)(mtod(m, char *) +
1265 sizeof(struct in_addr) + OPTSIZ);
1266 #undef OPTSIZ
1267 /*
1268 * Record return path as an IP source route,
1269 * reversing the path (pointers are now aligned).
1270 */
1271 while (p >= isr->isr_routes) {
1272 *q++ = *p--;
1273 }
1274 /*
1275 * Last hop goes to final destination.
1276 */
1277 *q = isr->isr_dst;
1278 m_tag_delete(m0, mtag);
1279 return m;
1280 }
1281
1282 const int inetctlerrmap[PRC_NCMDS] = {
1283 [PRC_MSGSIZE] = EMSGSIZE,
1284 [PRC_HOSTDEAD] = EHOSTDOWN,
1285 [PRC_HOSTUNREACH] = EHOSTUNREACH,
1286 [PRC_UNREACH_NET] = EHOSTUNREACH,
1287 [PRC_UNREACH_HOST] = EHOSTUNREACH,
1288 [PRC_UNREACH_PROTOCOL] = ECONNREFUSED,
1289 [PRC_UNREACH_PORT] = ECONNREFUSED,
1290 [PRC_UNREACH_SRCFAIL] = EHOSTUNREACH,
1291 [PRC_PARAMPROB] = ENOPROTOOPT,
1292 };
1293
1294 void
1295 ip_fasttimo(void)
1296 {
1297 if (ip_drainwanted) {
1298 ip_drain();
1299 ip_drainwanted = 0;
1300 }
1301 }
1302
1303 void
1304 ip_drainstub(void)
1305 {
1306 ip_drainwanted = 1;
1307 }
1308
1309 /*
1310 * Forward a packet. If some error occurs return the sender
1311 * an icmp packet. Note we can't always generate a meaningful
1312 * icmp message because icmp doesn't have a large enough repertoire
1313 * of codes and types.
1314 *
1315 * If not forwarding, just drop the packet. This could be confusing
1316 * if ipforwarding was zero but some routing protocol was advancing
1317 * us as a gateway to somewhere. However, we must let the routing
1318 * protocol deal with that.
1319 *
1320 * The srcrt parameter indicates whether the packet is being forwarded
1321 * via a source route.
1322 */
1323 static void
1324 ip_forward(struct mbuf *m, int srcrt, struct ifnet *rcvif)
1325 {
1326 struct ip *ip = mtod(m, struct ip *);
1327 struct rtentry *rt;
1328 int error, type = 0, code = 0, destmtu = 0;
1329 struct mbuf *mcopy;
1330 n_long dest;
1331 union {
1332 struct sockaddr dst;
1333 struct sockaddr_in dst4;
1334 } u;
1335 uint64_t *ips;
1336 struct route *ro;
1337
1338 KASSERTMSG(cpu_softintr_p(), "ip_forward: not in the software "
1339 "interrupt handler; synchronization assumptions violated");
1340
1341 /*
1342 * We are now in the output path.
1343 */
1344 MCLAIM(m, &ip_tx_mowner);
1345
1346 /*
1347 * Clear any in-bound checksum flags for this packet.
1348 */
1349 m->m_pkthdr.csum_flags = 0;
1350
1351 dest = 0;
1352 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
1353 IP_STATINC(IP_STAT_CANTFORWARD);
1354 m_freem(m);
1355 return;
1356 }
1357
1358 if (ip->ip_ttl <= IPTTLDEC) {
1359 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0);
1360 return;
1361 }
1362
1363 sockaddr_in_init(&u.dst4, &ip->ip_dst, 0);
1364
1365 ro = percpu_getref(ipforward_rt_percpu);
1366 rt = rtcache_lookup(ro, &u.dst);
1367 if (rt == NULL) {
1368 percpu_putref(ipforward_rt_percpu);
1369 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_NET, dest, 0);
1370 return;
1371 }
1372
1373 /*
1374 * Save at most 68 bytes of the packet in case
1375 * we need to generate an ICMP message to the src.
1376 * Pullup to avoid sharing mbuf cluster between m and mcopy.
1377 */
1378 mcopy = m_copym(m, 0, imin(ntohs(ip->ip_len), 68), M_DONTWAIT);
1379 if (mcopy)
1380 mcopy = m_pullup(mcopy, ip->ip_hl << 2);
1381
1382 ip->ip_ttl -= IPTTLDEC;
1383
1384 /*
1385 * If forwarding packet using same interface that it came in on,
1386 * perhaps should send a redirect to sender to shortcut a hop.
1387 * Only send redirect if source is sending directly to us,
1388 * and if packet was not source routed (or has any options).
1389 * Also, don't send redirect if forwarding using a default route
1390 * or a route modified by a redirect.
1391 */
1392 if (rt->rt_ifp == rcvif &&
1393 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1394 !in_nullhost(satocsin(rt_getkey(rt))->sin_addr) &&
1395 ipsendredirects && !srcrt) {
1396 if (rt->rt_ifa &&
1397 (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) ==
1398 ifatoia(rt->rt_ifa)->ia_subnet) {
1399 if (rt->rt_flags & RTF_GATEWAY)
1400 dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
1401 else
1402 dest = ip->ip_dst.s_addr;
1403 /*
1404 * Router requirements says to only send host
1405 * redirects.
1406 */
1407 type = ICMP_REDIRECT;
1408 code = ICMP_REDIRECT_HOST;
1409 }
1410 }
1411 rtcache_unref(rt, ro);
1412
1413 error = ip_output(m, NULL, ro,
1414 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)),
1415 NULL, NULL);
1416
1417 if (error) {
1418 IP_STATINC(IP_STAT_CANTFORWARD);
1419 goto error;
1420 }
1421
1422 ips = IP_STAT_GETREF();
1423 ips[IP_STAT_FORWARD]++;
1424
1425 if (type) {
1426 ips[IP_STAT_REDIRECTSENT]++;
1427 IP_STAT_PUTREF();
1428 goto redirect;
1429 }
1430
1431 IP_STAT_PUTREF();
1432 if (mcopy) {
1433 #ifdef GATEWAY
1434 if (mcopy->m_flags & M_CANFASTFWD)
1435 ipflow_create(ro, mcopy);
1436 #endif
1437 m_freem(mcopy);
1438 }
1439
1440 percpu_putref(ipforward_rt_percpu);
1441 return;
1442
1443 redirect:
1444 error:
1445 if (mcopy == NULL) {
1446 percpu_putref(ipforward_rt_percpu);
1447 return;
1448 }
1449
1450 switch (error) {
1451
1452 case 0: /* forwarded, but need redirect */
1453 /* type, code set above */
1454 break;
1455
1456 case ENETUNREACH: /* shouldn't happen, checked above */
1457 case EHOSTUNREACH:
1458 case ENETDOWN:
1459 case EHOSTDOWN:
1460 default:
1461 type = ICMP_UNREACH;
1462 code = ICMP_UNREACH_HOST;
1463 break;
1464
1465 case EMSGSIZE:
1466 type = ICMP_UNREACH;
1467 code = ICMP_UNREACH_NEEDFRAG;
1468
1469 if ((rt = rtcache_validate(ro)) != NULL) {
1470 destmtu = rt->rt_ifp->if_mtu;
1471 rtcache_unref(rt, ro);
1472 }
1473 #ifdef IPSEC
1474 if (ipsec_used)
1475 (void)ipsec4_forward(mcopy, &destmtu);
1476 #endif
1477 IP_STATINC(IP_STAT_CANTFRAG);
1478 break;
1479
1480 case ENOBUFS:
1481 /*
1482 * Do not generate ICMP_SOURCEQUENCH as required in RFC 1812,
1483 * Requirements for IP Version 4 Routers. Source quench can
1484 * big problem under DoS attacks or if the underlying
1485 * interface is rate-limited.
1486 */
1487 if (mcopy)
1488 m_freem(mcopy);
1489 percpu_putref(ipforward_rt_percpu);
1490 return;
1491 }
1492 icmp_error(mcopy, type, code, dest, destmtu);
1493 percpu_putref(ipforward_rt_percpu);
1494 }
1495
1496 void
1497 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1498 struct mbuf *m)
1499 {
1500 struct socket *so = inp->inp_socket;
1501 int inpflags = inp->inp_flags;
1502
1503 if (SOOPT_TIMESTAMP(so->so_options))
1504 mp = sbsavetimestamp(so->so_options, m, mp);
1505
1506 if (inpflags & INP_RECVDSTADDR) {
1507 *mp = sbcreatecontrol(&ip->ip_dst,
1508 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1509 if (*mp)
1510 mp = &(*mp)->m_next;
1511 }
1512
1513 if (inpflags & INP_RECVTTL) {
1514 *mp = sbcreatecontrol(&ip->ip_ttl,
1515 sizeof(uint8_t), IP_RECVTTL, IPPROTO_IP);
1516 if (*mp)
1517 mp = &(*mp)->m_next;
1518 }
1519
1520 struct psref psref;
1521 ifnet_t *ifp = m_get_rcvif_psref(m, &psref);
1522 if (__predict_false(ifp == NULL)) {
1523 #ifdef DIAGNOSTIC
1524 printf("%s: missing receive interface\n", __func__);
1525 #endif
1526 return; /* XXX should report error? */
1527 }
1528
1529 if (inpflags & INP_RECVPKTINFO) {
1530 struct in_pktinfo ipi;
1531 ipi.ipi_addr = ip->ip_dst;
1532 ipi.ipi_ifindex = ifp->if_index;
1533 *mp = sbcreatecontrol(&ipi,
1534 sizeof(ipi), IP_PKTINFO, IPPROTO_IP);
1535 if (*mp)
1536 mp = &(*mp)->m_next;
1537 }
1538 if (inpflags & INP_RECVIF) {
1539 struct sockaddr_dl sdl;
1540
1541 sockaddr_dl_init(&sdl, sizeof(sdl), ifp->if_index, 0, NULL, 0,
1542 NULL, 0);
1543 *mp = sbcreatecontrol(&sdl, sdl.sdl_len, IP_RECVIF, IPPROTO_IP);
1544 if (*mp)
1545 mp = &(*mp)->m_next;
1546 }
1547 m_put_rcvif_psref(ifp, &psref);
1548 }
1549
1550 /*
1551 * sysctl helper routine for net.inet.ip.forwsrcrt.
1552 */
1553 static int
1554 sysctl_net_inet_ip_forwsrcrt(SYSCTLFN_ARGS)
1555 {
1556 int error, tmp;
1557 struct sysctlnode node;
1558
1559 node = *rnode;
1560 tmp = ip_forwsrcrt;
1561 node.sysctl_data = &tmp;
1562 error = sysctl_lookup(SYSCTLFN_CALL(&node));
1563 if (error || newp == NULL)
1564 return (error);
1565
1566 error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_FORWSRCRT,
1567 0, NULL, NULL, NULL);
1568 if (error)
1569 return (error);
1570
1571 ip_forwsrcrt = tmp;
1572
1573 return (0);
1574 }
1575
1576 /*
1577 * sysctl helper routine for net.inet.ip.mtudisctimeout. checks the
1578 * range of the new value and tweaks timers if it changes.
1579 */
1580 static int
1581 sysctl_net_inet_ip_pmtudto(SYSCTLFN_ARGS)
1582 {
1583 int error, tmp;
1584 struct sysctlnode node;
1585
1586 icmp_mtudisc_lock();
1587
1588 node = *rnode;
1589 tmp = ip_mtudisc_timeout;
1590 node.sysctl_data = &tmp;
1591 error = sysctl_lookup(SYSCTLFN_CALL(&node));
1592 if (error || newp == NULL)
1593 goto out;
1594 if (tmp < 0) {
1595 error = EINVAL;
1596 goto out;
1597 }
1598
1599 ip_mtudisc_timeout = tmp;
1600 rt_timer_queue_change(ip_mtudisc_timeout_q, ip_mtudisc_timeout);
1601 error = 0;
1602 out:
1603 icmp_mtudisc_unlock();
1604 return error;
1605 }
1606
1607 static int
1608 sysctl_net_inet_ip_stats(SYSCTLFN_ARGS)
1609 {
1610
1611 return (NETSTAT_SYSCTL(ipstat_percpu, IP_NSTATS));
1612 }
1613
1614 static void
1615 sysctl_net_inet_ip_setup(struct sysctllog **clog)
1616 {
1617 sysctl_createv(clog, 0, NULL, NULL,
1618 CTLFLAG_PERMANENT,
1619 CTLTYPE_NODE, "inet",
1620 SYSCTL_DESCR("PF_INET related settings"),
1621 NULL, 0, NULL, 0,
1622 CTL_NET, PF_INET, CTL_EOL);
1623 sysctl_createv(clog, 0, NULL, NULL,
1624 CTLFLAG_PERMANENT,
1625 CTLTYPE_NODE, "ip",
1626 SYSCTL_DESCR("IPv4 related settings"),
1627 NULL, 0, NULL, 0,
1628 CTL_NET, PF_INET, IPPROTO_IP, CTL_EOL);
1629
1630 sysctl_createv(clog, 0, NULL, NULL,
1631 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1632 CTLTYPE_INT, "forwarding",
1633 SYSCTL_DESCR("Enable forwarding of INET datagrams"),
1634 NULL, 0, &ipforwarding, 0,
1635 CTL_NET, PF_INET, IPPROTO_IP,
1636 IPCTL_FORWARDING, CTL_EOL);
1637 sysctl_createv(clog, 0, NULL, NULL,
1638 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1639 CTLTYPE_INT, "redirect",
1640 SYSCTL_DESCR("Enable sending of ICMP redirect messages"),
1641 NULL, 0, &ipsendredirects, 0,
1642 CTL_NET, PF_INET, IPPROTO_IP,
1643 IPCTL_SENDREDIRECTS, CTL_EOL);
1644 sysctl_createv(clog, 0, NULL, NULL,
1645 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1646 CTLTYPE_INT, "ttl",
1647 SYSCTL_DESCR("Default TTL for an INET datagram"),
1648 NULL, 0, &ip_defttl, 0,
1649 CTL_NET, PF_INET, IPPROTO_IP,
1650 IPCTL_DEFTTL, CTL_EOL);
1651 #ifdef IPCTL_DEFMTU
1652 sysctl_createv(clog, 0, NULL, NULL,
1653 CTLFLAG_PERMANENT /* |CTLFLAG_READWRITE? */,
1654 CTLTYPE_INT, "mtu",
1655 SYSCTL_DESCR("Default MTA for an INET route"),
1656 NULL, 0, &ip_mtu, 0,
1657 CTL_NET, PF_INET, IPPROTO_IP,
1658 IPCTL_DEFMTU, CTL_EOL);
1659 #endif /* IPCTL_DEFMTU */
1660 sysctl_createv(clog, 0, NULL, NULL,
1661 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1662 CTLTYPE_INT, "forwsrcrt",
1663 SYSCTL_DESCR("Enable forwarding of source-routed "
1664 "datagrams"),
1665 sysctl_net_inet_ip_forwsrcrt, 0, &ip_forwsrcrt, 0,
1666 CTL_NET, PF_INET, IPPROTO_IP,
1667 IPCTL_FORWSRCRT, CTL_EOL);
1668 sysctl_createv(clog, 0, NULL, NULL,
1669 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1670 CTLTYPE_INT, "directed-broadcast",
1671 SYSCTL_DESCR("Enable forwarding of broadcast datagrams"),
1672 NULL, 0, &ip_directedbcast, 0,
1673 CTL_NET, PF_INET, IPPROTO_IP,
1674 IPCTL_DIRECTEDBCAST, CTL_EOL);
1675 sysctl_createv(clog, 0, NULL, NULL,
1676 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1677 CTLTYPE_INT, "allowsrcrt",
1678 SYSCTL_DESCR("Accept source-routed datagrams"),
1679 NULL, 0, &ip_allowsrcrt, 0,
1680 CTL_NET, PF_INET, IPPROTO_IP,
1681 IPCTL_ALLOWSRCRT, CTL_EOL);
1682
1683 sysctl_createv(clog, 0, NULL, NULL,
1684 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1685 CTLTYPE_INT, "mtudisc",
1686 SYSCTL_DESCR("Use RFC1191 Path MTU Discovery"),
1687 NULL, 0, &ip_mtudisc, 0,
1688 CTL_NET, PF_INET, IPPROTO_IP,
1689 IPCTL_MTUDISC, CTL_EOL);
1690 sysctl_createv(clog, 0, NULL, NULL,
1691 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1692 CTLTYPE_INT, "anonportmin",
1693 SYSCTL_DESCR("Lowest ephemeral port number to assign"),
1694 sysctl_net_inet_ip_ports, 0, &anonportmin, 0,
1695 CTL_NET, PF_INET, IPPROTO_IP,
1696 IPCTL_ANONPORTMIN, CTL_EOL);
1697 sysctl_createv(clog, 0, NULL, NULL,
1698 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1699 CTLTYPE_INT, "anonportmax",
1700 SYSCTL_DESCR("Highest ephemeral port number to assign"),
1701 sysctl_net_inet_ip_ports, 0, &anonportmax, 0,
1702 CTL_NET, PF_INET, IPPROTO_IP,
1703 IPCTL_ANONPORTMAX, CTL_EOL);
1704 sysctl_createv(clog, 0, NULL, NULL,
1705 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1706 CTLTYPE_INT, "mtudisctimeout",
1707 SYSCTL_DESCR("Lifetime of a Path MTU Discovered route"),
1708 sysctl_net_inet_ip_pmtudto, 0, (void *)&ip_mtudisc_timeout, 0,
1709 CTL_NET, PF_INET, IPPROTO_IP,
1710 IPCTL_MTUDISCTIMEOUT, CTL_EOL);
1711 #ifndef IPNOPRIVPORTS
1712 sysctl_createv(clog, 0, NULL, NULL,
1713 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1714 CTLTYPE_INT, "lowportmin",
1715 SYSCTL_DESCR("Lowest privileged ephemeral port number "
1716 "to assign"),
1717 sysctl_net_inet_ip_ports, 0, &lowportmin, 0,
1718 CTL_NET, PF_INET, IPPROTO_IP,
1719 IPCTL_LOWPORTMIN, CTL_EOL);
1720 sysctl_createv(clog, 0, NULL, NULL,
1721 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1722 CTLTYPE_INT, "lowportmax",
1723 SYSCTL_DESCR("Highest privileged ephemeral port number "
1724 "to assign"),
1725 sysctl_net_inet_ip_ports, 0, &lowportmax, 0,
1726 CTL_NET, PF_INET, IPPROTO_IP,
1727 IPCTL_LOWPORTMAX, CTL_EOL);
1728 #endif /* IPNOPRIVPORTS */
1729 #if NGRE > 0
1730 sysctl_createv(clog, 0, NULL, NULL,
1731 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1732 CTLTYPE_INT, "grettl",
1733 SYSCTL_DESCR("Default TTL for a gre tunnel datagram"),
1734 NULL, 0, &ip_gre_ttl, 0,
1735 CTL_NET, PF_INET, IPPROTO_IP,
1736 IPCTL_GRE_TTL, CTL_EOL);
1737 #endif /* NGRE */
1738 sysctl_createv(clog, 0, NULL, NULL,
1739 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1740 CTLTYPE_INT, "checkinterface",
1741 SYSCTL_DESCR("Enable receive side of Strong ES model "
1742 "from RFC1122"),
1743 NULL, 0, &ip_checkinterface, 0,
1744 CTL_NET, PF_INET, IPPROTO_IP,
1745 IPCTL_CHECKINTERFACE, CTL_EOL);
1746 sysctl_createv(clog, 0, NULL, NULL,
1747 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1748 CTLTYPE_INT, "random_id",
1749 SYSCTL_DESCR("Assign random ip_id values"),
1750 NULL, 0, &ip_do_randomid, 0,
1751 CTL_NET, PF_INET, IPPROTO_IP,
1752 IPCTL_RANDOMID, CTL_EOL);
1753 sysctl_createv(clog, 0, NULL, NULL,
1754 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1755 CTLTYPE_INT, "do_loopback_cksum",
1756 SYSCTL_DESCR("Perform IP checksum on loopback"),
1757 NULL, 0, &ip_do_loopback_cksum, 0,
1758 CTL_NET, PF_INET, IPPROTO_IP,
1759 IPCTL_LOOPBACKCKSUM, CTL_EOL);
1760 sysctl_createv(clog, 0, NULL, NULL,
1761 CTLFLAG_PERMANENT,
1762 CTLTYPE_STRUCT, "stats",
1763 SYSCTL_DESCR("IP statistics"),
1764 sysctl_net_inet_ip_stats, 0, NULL, 0,
1765 CTL_NET, PF_INET, IPPROTO_IP, IPCTL_STATS,
1766 CTL_EOL);
1767 #if NARP
1768 sysctl_createv(clog, 0, NULL, NULL,
1769 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1770 CTLTYPE_INT, "dad_count",
1771 SYSCTL_DESCR("Number of Duplicate Address Detection "
1772 "probes to send"),
1773 NULL, 0, &ip_dad_count, 0,
1774 CTL_NET, PF_INET, IPPROTO_IP,
1775 IPCTL_DAD_COUNT, CTL_EOL);
1776 #endif
1777
1778 /* anonportalgo RFC6056 subtree */
1779 const struct sysctlnode *portalgo_node;
1780 sysctl_createv(clog, 0, NULL, &portalgo_node,
1781 CTLFLAG_PERMANENT,
1782 CTLTYPE_NODE, "anonportalgo",
1783 SYSCTL_DESCR("Anonymous Port Algorithm Selection (RFC 6056)"),
1784 NULL, 0, NULL, 0,
1785 CTL_NET, PF_INET, IPPROTO_IP, CTL_CREATE, CTL_EOL);
1786 sysctl_createv(clog, 0, &portalgo_node, NULL,
1787 CTLFLAG_PERMANENT,
1788 CTLTYPE_STRING, "available",
1789 SYSCTL_DESCR("available algorithms"),
1790 sysctl_portalgo_available, 0, NULL, PORTALGO_MAXLEN,
1791 CTL_CREATE, CTL_EOL);
1792 sysctl_createv(clog, 0, &portalgo_node, NULL,
1793 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1794 CTLTYPE_STRING, "selected",
1795 SYSCTL_DESCR("selected algorithm"),
1796 sysctl_portalgo_selected4, 0, NULL, PORTALGO_MAXLEN,
1797 CTL_CREATE, CTL_EOL);
1798 sysctl_createv(clog, 0, &portalgo_node, NULL,
1799 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1800 CTLTYPE_STRUCT, "reserve",
1801 SYSCTL_DESCR("bitmap of reserved ports"),
1802 sysctl_portalgo_reserve4, 0, NULL, 0,
1803 CTL_CREATE, CTL_EOL);
1804 }
1805
1806 void
1807 ip_statinc(u_int stat)
1808 {
1809
1810 KASSERT(stat < IP_NSTATS);
1811 IP_STATINC(stat);
1812 }
1813