ip_input.c revision 1.379 1 /* $NetBSD: ip_input.c,v 1.379 2018/04/11 08:11:20 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 1.379 2018/04/11 08:11:20 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
157 #define IPFORWARDING 0 /* don't forward IP packets not for us */
158 #endif
159 #endif
160
161 #define IPMTUDISCTIMEOUT (10 * 60) /* as per RFC 1191 */
162
163 int ipforwarding = IPFORWARDING;
164 int ipsendredirects = 1;
165 int ip_defttl = IPDEFTTL;
166 int ip_forwsrcrt = 0;
167 int ip_directedbcast = 0;
168 int ip_allowsrcrt = 0;
169 int ip_mtudisc = 1;
170 int ip_mtudisc_timeout = IPMTUDISCTIMEOUT;
171 int ip_do_randomid = 0;
172
173 /*
174 * XXX - Setting ip_checkinterface mostly implements the receive side of
175 * the Strong ES model described in RFC 1122, but since the routing table
176 * and transmit implementation do not implement the Strong ES model,
177 * setting this to 1 results in an odd hybrid.
178 *
179 * XXX - ip_checkinterface currently must be disabled if you use ipnat
180 * to translate the destination address to another local interface.
181 *
182 * XXX - ip_checkinterface must be disabled if you add IP aliases
183 * to the loopback interface instead of the interface where the
184 * packets for those addresses are received.
185 */
186 static int ip_checkinterface __read_mostly = 0;
187
188 struct rttimer_queue *ip_mtudisc_timeout_q = NULL;
189
190 pktqueue_t * ip_pktq __read_mostly;
191 pfil_head_t * inet_pfil_hook __read_mostly;
192 ipid_state_t * ip_ids __read_mostly;
193 percpu_t * ipstat_percpu __read_mostly;
194
195 static percpu_t *ipforward_rt_percpu __cacheline_aligned;
196
197 uint16_t ip_id;
198
199 #ifdef INET_CSUM_COUNTERS
200 #include <sys/device.h>
201
202 struct evcnt ip_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
203 NULL, "inet", "hwcsum bad");
204 struct evcnt ip_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
205 NULL, "inet", "hwcsum ok");
206 struct evcnt ip_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
207 NULL, "inet", "swcsum");
208
209 #define INET_CSUM_COUNTER_INCR(ev) (ev)->ev_count++
210
211 EVCNT_ATTACH_STATIC(ip_hwcsum_bad);
212 EVCNT_ATTACH_STATIC(ip_hwcsum_ok);
213 EVCNT_ATTACH_STATIC(ip_swcsum);
214
215 #else
216
217 #define INET_CSUM_COUNTER_INCR(ev) /* nothing */
218
219 #endif /* INET_CSUM_COUNTERS */
220
221 /*
222 * Used to save the IP options in case a protocol wants to respond
223 * to an incoming packet over the same route if the packet got here
224 * using IP source routing. This allows connection establishment and
225 * maintenance when the remote end is on a network that is not known
226 * to us.
227 */
228 struct ip_srcrt {
229 int isr_nhops; /* number of hops */
230 struct in_addr isr_dst; /* final destination */
231 char isr_nop; /* one NOP to align */
232 char isr_hdr[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN & OFFSET */
233 struct in_addr isr_routes[MAX_IPOPTLEN/sizeof(struct in_addr)];
234 };
235
236 static int ip_drainwanted;
237
238 static void save_rte(struct mbuf *, u_char *, struct in_addr);
239
240 #ifdef MBUFTRACE
241 struct mowner ip_rx_mowner = MOWNER_INIT("internet", "rx");
242 struct mowner ip_tx_mowner = MOWNER_INIT("internet", "tx");
243 #endif
244
245 static void ipintr(void *);
246 static void ip_input(struct mbuf *);
247 static void ip_forward(struct mbuf *, int, struct ifnet *);
248 static bool ip_dooptions(struct mbuf *);
249 static struct in_ifaddr *ip_rtaddr(struct in_addr, struct psref *);
250 static void sysctl_net_inet_ip_setup(struct sysctllog **);
251
252 static struct in_ifaddr *ip_match_our_address(struct ifnet *, struct ip *,
253 int *);
254 static struct in_ifaddr *ip_match_our_address_broadcast(struct ifnet *,
255 struct ip *);
256
257 #ifdef NET_MPSAFE
258 #define SOFTNET_LOCK() mutex_enter(softnet_lock)
259 #define SOFTNET_UNLOCK() mutex_exit(softnet_lock)
260 #else
261 #define SOFTNET_LOCK() KASSERT(mutex_owned(softnet_lock))
262 #define SOFTNET_UNLOCK() KASSERT(mutex_owned(softnet_lock))
263 #endif
264
265 /*
266 * IP initialization: fill in IP protocol switch table.
267 * All protocols not implemented in kernel go to raw IP protocol handler.
268 */
269 void
270 ip_init(void)
271 {
272 const struct protosw *pr;
273
274 in_init();
275 sysctl_net_inet_ip_setup(NULL);
276
277 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
278 KASSERT(pr != NULL);
279
280 ip_pktq = pktq_create(IFQ_MAXLEN, ipintr, NULL);
281 KASSERT(ip_pktq != NULL);
282
283 for (u_int i = 0; i < IPPROTO_MAX; i++) {
284 ip_protox[i] = pr - inetsw;
285 }
286 for (pr = inetdomain.dom_protosw;
287 pr < inetdomain.dom_protoswNPROTOSW; pr++)
288 if (pr->pr_domain->dom_family == PF_INET &&
289 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
290 ip_protox[pr->pr_protocol] = pr - inetsw;
291
292 ip_reass_init();
293
294 ip_ids = ip_id_init();
295 ip_id = time_uptime & 0xfffff;
296
297 #ifdef GATEWAY
298 ipflow_init();
299 #endif
300
301 /* Register our Packet Filter hook. */
302 inet_pfil_hook = pfil_head_create(PFIL_TYPE_AF, (void *)AF_INET);
303 KASSERT(inet_pfil_hook != NULL);
304
305 #ifdef MBUFTRACE
306 MOWNER_ATTACH(&ip_tx_mowner);
307 MOWNER_ATTACH(&ip_rx_mowner);
308 #endif
309
310 ipstat_percpu = percpu_alloc(sizeof(uint64_t) * IP_NSTATS);
311 ipforward_rt_percpu = percpu_alloc(sizeof(struct route));
312 ip_mtudisc_timeout_q = rt_timer_queue_create(ip_mtudisc_timeout);
313 }
314
315 static struct in_ifaddr *
316 ip_match_our_address(struct ifnet *ifp, struct ip *ip, int *downmatch)
317 {
318 struct in_ifaddr *ia = NULL;
319 int checkif;
320
321 /*
322 * Enable a consistency check between the destination address
323 * and the arrival interface for a unicast packet (the RFC 1122
324 * strong ES model) if IP forwarding is disabled and the packet
325 * is not locally generated.
326 *
327 * XXX - Checking also should be disabled if the destination
328 * address is ipnat'ed to a different interface.
329 *
330 * XXX - Checking is incompatible with IP aliases added
331 * to the loopback interface instead of the interface where
332 * the packets are received.
333 *
334 * XXX - We need to add a per ifaddr flag for this so that
335 * we get finer grain control.
336 */
337 checkif = ip_checkinterface && (ipforwarding == 0) &&
338 (ifp->if_flags & IFF_LOOPBACK) == 0;
339
340 IN_ADDRHASH_READER_FOREACH(ia, ip->ip_dst.s_addr) {
341 if (in_hosteq(ia->ia_addr.sin_addr, ip->ip_dst)) {
342 if (ia->ia4_flags & IN_IFF_NOTREADY)
343 continue;
344 if (checkif && ia->ia_ifp != ifp)
345 continue;
346 if ((ia->ia_ifp->if_flags & IFF_UP) == 0) {
347 (*downmatch)++;
348 continue;
349 }
350 if (ia->ia4_flags & IN_IFF_DETACHED &&
351 (ifp->if_flags & IFF_LOOPBACK) == 0)
352 continue;
353 break;
354 }
355 }
356
357 return ia;
358 }
359
360 static struct in_ifaddr *
361 ip_match_our_address_broadcast(struct ifnet *ifp, struct ip *ip)
362 {
363 struct in_ifaddr *ia = NULL;
364 struct ifaddr *ifa;
365
366 IFADDR_READER_FOREACH(ifa, ifp) {
367 if (ifa->ifa_addr->sa_family != AF_INET)
368 continue;
369 ia = ifatoia(ifa);
370 if (ia->ia4_flags & IN_IFF_NOTREADY)
371 continue;
372 if (ia->ia4_flags & IN_IFF_DETACHED &&
373 (ifp->if_flags & IFF_LOOPBACK) == 0)
374 continue;
375 if (in_hosteq(ip->ip_dst, ia->ia_broadaddr.sin_addr) ||
376 in_hosteq(ip->ip_dst, ia->ia_netbroadcast) ||
377 /*
378 * Look for all-0's host part (old broadcast addr),
379 * either for subnet or net.
380 */
381 ip->ip_dst.s_addr == ia->ia_subnet ||
382 ip->ip_dst.s_addr == ia->ia_net)
383 goto matched;
384 /*
385 * An interface with IP address zero accepts
386 * all packets that arrive on that interface.
387 */
388 if (in_nullhost(ia->ia_addr.sin_addr))
389 goto matched;
390 }
391 ia = NULL;
392
393 matched:
394 return ia;
395 }
396
397 /*
398 * IP software interrupt routine.
399 */
400 static void
401 ipintr(void *arg __unused)
402 {
403 struct mbuf *m;
404
405 KASSERT(cpu_softintr_p());
406
407 SOFTNET_KERNEL_LOCK_UNLESS_NET_MPSAFE();
408 while ((m = pktq_dequeue(ip_pktq)) != NULL) {
409 ip_input(m);
410 }
411 SOFTNET_KERNEL_UNLOCK_UNLESS_NET_MPSAFE();
412 }
413
414 /*
415 * IP input routine. Checksum and byte swap header. If fragmented
416 * try to reassemble. Process options. Pass to next level.
417 */
418 static void
419 ip_input(struct mbuf *m)
420 {
421 struct ip *ip = NULL;
422 struct in_ifaddr *ia = NULL;
423 int hlen = 0, len;
424 int downmatch;
425 int srcrt = 0;
426 ifnet_t *ifp;
427 struct psref psref;
428 int s;
429
430 KASSERTMSG(cpu_softintr_p(), "ip_input: not in the software "
431 "interrupt handler; synchronization assumptions violated");
432
433 MCLAIM(m, &ip_rx_mowner);
434 KASSERT((m->m_flags & M_PKTHDR) != 0);
435
436 ifp = m_get_rcvif_psref(m, &psref);
437 if (__predict_false(ifp == NULL))
438 goto out;
439
440 /*
441 * If no IP addresses have been set yet but the interfaces
442 * are receiving, can't do anything with incoming packets yet.
443 * Note: we pre-check without locks held.
444 */
445 if (IN_ADDRLIST_READER_EMPTY())
446 goto out;
447 IP_STATINC(IP_STAT_TOTAL);
448
449 /*
450 * If the IP header is not aligned, slurp it up into a new
451 * mbuf with space for link headers, in the event we forward
452 * it. Otherwise, if it is aligned, make sure the entire
453 * base IP header is in the first mbuf of the chain.
454 */
455 if (IP_HDR_ALIGNED_P(mtod(m, void *)) == 0) {
456 if ((m = m_copyup(m, sizeof(struct ip),
457 (max_linkhdr + 3) & ~3)) == NULL) {
458 /* XXXJRT new stat, please */
459 IP_STATINC(IP_STAT_TOOSMALL);
460 goto out;
461 }
462 } else if (__predict_false(m->m_len < sizeof(struct ip))) {
463 if ((m = m_pullup(m, sizeof(struct ip))) == NULL) {
464 IP_STATINC(IP_STAT_TOOSMALL);
465 goto out;
466 }
467 }
468 ip = mtod(m, struct ip *);
469 if (ip->ip_v != IPVERSION) {
470 IP_STATINC(IP_STAT_BADVERS);
471 goto out;
472 }
473 hlen = ip->ip_hl << 2;
474 if (hlen < sizeof(struct ip)) { /* minimum header length */
475 IP_STATINC(IP_STAT_BADHLEN);
476 goto out;
477 }
478 if (hlen > m->m_len) {
479 if ((m = m_pullup(m, hlen)) == NULL) {
480 IP_STATINC(IP_STAT_BADHLEN);
481 goto out;
482 }
483 ip = mtod(m, struct ip *);
484 }
485
486 /*
487 * RFC1122: packets with a multicast source address are
488 * not allowed.
489 */
490 if (IN_MULTICAST(ip->ip_src.s_addr)) {
491 IP_STATINC(IP_STAT_BADADDR);
492 goto out;
493 }
494
495 /* 127/8 must not appear on wire - RFC1122 */
496 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
497 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
498 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
499 IP_STATINC(IP_STAT_BADADDR);
500 goto out;
501 }
502 }
503
504 switch (m->m_pkthdr.csum_flags &
505 ((ifp->if_csum_flags_rx & M_CSUM_IPv4) | M_CSUM_IPv4_BAD)) {
506 case M_CSUM_IPv4|M_CSUM_IPv4_BAD:
507 INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad);
508 IP_STATINC(IP_STAT_BADSUM);
509 goto out;
510
511 case M_CSUM_IPv4:
512 /* Checksum was okay. */
513 INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok);
514 break;
515
516 default:
517 /*
518 * Must compute it ourselves. Maybe skip checksum on
519 * loopback interfaces.
520 */
521 if (__predict_true(!(ifp->if_flags & IFF_LOOPBACK) ||
522 ip_do_loopback_cksum)) {
523 INET_CSUM_COUNTER_INCR(&ip_swcsum);
524 if (in_cksum(m, hlen) != 0) {
525 IP_STATINC(IP_STAT_BADSUM);
526 goto out;
527 }
528 }
529 break;
530 }
531
532 /* Retrieve the packet length. */
533 len = ntohs(ip->ip_len);
534
535 /*
536 * Check for additional length bogosity
537 */
538 if (len < hlen) {
539 IP_STATINC(IP_STAT_BADLEN);
540 goto out;
541 }
542
543 /*
544 * Check that the amount of data in the buffers is at least as much
545 * as the IP header would have us expect. Trim mbufs if longer than
546 * we expect. Drop packet if shorter than we expect.
547 */
548 if (m->m_pkthdr.len < len) {
549 IP_STATINC(IP_STAT_TOOSHORT);
550 goto out;
551 }
552 if (m->m_pkthdr.len > len) {
553 if (m->m_len == m->m_pkthdr.len) {
554 m->m_len = len;
555 m->m_pkthdr.len = len;
556 } else
557 m_adj(m, len - m->m_pkthdr.len);
558 }
559
560 /*
561 * Assume that we can create a fast-forward IP flow entry
562 * based on this packet.
563 */
564 m->m_flags |= M_CANFASTFWD;
565
566 /*
567 * Run through list of hooks for input packets. If there are any
568 * filters which require that additional packets in the flow are
569 * not fast-forwarded, they must clear the M_CANFASTFWD flag.
570 * Note that filters must _never_ set this flag, as another filter
571 * in the list may have previously cleared it.
572 *
573 * Don't call hooks if the packet has already been processed by
574 * IPsec (encapsulated, tunnel mode).
575 */
576 #if defined(IPSEC)
577 if (!ipsec_used || !ipsec_indone(m))
578 #else
579 if (1)
580 #endif
581 {
582 struct in_addr odst = ip->ip_dst;
583 bool freed;
584
585 freed = pfil_run_hooks(inet_pfil_hook, &m, ifp, PFIL_IN) != 0;
586 if (freed || m == NULL) {
587 m = NULL;
588 goto out;
589 }
590 ip = mtod(m, struct ip *);
591 hlen = ip->ip_hl << 2;
592
593 /*
594 * XXX The setting of "srcrt" here is to prevent ip_forward()
595 * from generating ICMP redirects for packets that have
596 * been redirected by a hook back out on to the same LAN that
597 * they came from and is not an indication that the packet
598 * is being influenced by source routing options. This
599 * allows things like
600 * "rdr tlp0 0/0 port 80 -> 1.1.1.200 3128 tcp"
601 * where tlp0 is both on the 1.1.1.0/24 network and is the
602 * default route for hosts on 1.1.1.0/24. Of course this
603 * also requires a "map tlp0 ..." to complete the story.
604 * One might argue whether or not this kind of network config.
605 * should be supported in this manner...
606 */
607 srcrt = (odst.s_addr != ip->ip_dst.s_addr);
608 }
609
610 #ifdef ALTQ
611 /* XXX Temporary until ALTQ is changed to use a pfil hook */
612 if (altq_input) {
613 SOFTNET_LOCK();
614 if ((*altq_input)(m, AF_INET) == 0) {
615 /* Packet dropped by traffic conditioner. */
616 SOFTNET_UNLOCK();
617 m = NULL;
618 goto out;
619 }
620 SOFTNET_UNLOCK();
621 }
622 #endif
623
624 /*
625 * Process options and, if not destined for us,
626 * ship it on. ip_dooptions returns 1 when an
627 * error was detected (causing an icmp message
628 * to be sent and the original packet to be freed).
629 */
630 if (hlen > sizeof(struct ip) && ip_dooptions(m)) {
631 m = NULL;
632 goto out;
633 }
634
635 /*
636 * Check our list of addresses, to see if the packet is for us.
637 *
638 * Traditional 4.4BSD did not consult IFF_UP at all.
639 * The behavior here is to treat addresses on !IFF_UP interface
640 * or IN_IFF_NOTREADY addresses as not mine.
641 */
642 downmatch = 0;
643 s = pserialize_read_enter();
644 ia = ip_match_our_address(ifp, ip, &downmatch);
645 if (ia != NULL) {
646 pserialize_read_exit(s);
647 goto ours;
648 }
649
650 if (ifp->if_flags & IFF_BROADCAST) {
651 ia = ip_match_our_address_broadcast(ifp, ip);
652 if (ia != NULL) {
653 pserialize_read_exit(s);
654 goto ours;
655 }
656 }
657 pserialize_read_exit(s);
658
659 if (IN_MULTICAST(ip->ip_dst.s_addr)) {
660 #ifdef MROUTING
661 extern struct socket *ip_mrouter;
662
663 if (ip_mrouter) {
664 /*
665 * If we are acting as a multicast router, all
666 * incoming multicast packets are passed to the
667 * kernel-level multicast forwarding function.
668 * The packet is returned (relatively) intact; if
669 * ip_mforward() returns a non-zero value, the packet
670 * must be discarded, else it may be accepted below.
671 *
672 * (The IP ident field is put in the same byte order
673 * as expected when ip_mforward() is called from
674 * ip_output().)
675 */
676 SOFTNET_LOCK();
677 if (ip_mforward(m, ifp) != 0) {
678 SOFTNET_UNLOCK();
679 IP_STATINC(IP_STAT_CANTFORWARD);
680 goto out;
681 }
682 SOFTNET_UNLOCK();
683
684 /*
685 * The process-level routing demon needs to receive
686 * all multicast IGMP packets, whether or not this
687 * host belongs to their destination groups.
688 */
689 if (ip->ip_p == IPPROTO_IGMP) {
690 goto ours;
691 }
692 IP_STATINC(IP_STAT_CANTFORWARD);
693 }
694 #endif
695 /*
696 * See if we belong to the destination multicast group on the
697 * arrival interface.
698 */
699 if (!in_multi_group(ip->ip_dst, ifp, 0)) {
700 IP_STATINC(IP_STAT_CANTFORWARD);
701 goto out;
702 }
703 goto ours;
704 }
705 if (ip->ip_dst.s_addr == INADDR_BROADCAST ||
706 in_nullhost(ip->ip_dst))
707 goto ours;
708
709 /*
710 * Not for us; forward if possible and desirable.
711 */
712 if (ipforwarding == 0) {
713 m_put_rcvif_psref(ifp, &psref);
714 IP_STATINC(IP_STAT_CANTFORWARD);
715 m_freem(m);
716 } else {
717 /*
718 * If ip_dst matched any of my address on !IFF_UP interface,
719 * and there's no IFF_UP interface that matches ip_dst,
720 * send icmp unreach. Forwarding it will result in in-kernel
721 * forwarding loop till TTL goes to 0.
722 */
723 if (downmatch) {
724 m_put_rcvif_psref(ifp, &psref);
725 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
726 IP_STATINC(IP_STAT_CANTFORWARD);
727 return;
728 }
729 #ifdef IPSEC
730 /* Check the security policy (SP) for the packet */
731 if (ipsec_used) {
732 if (ipsec4_input(m, IP_FORWARDING) != 0) {
733 goto out;
734 }
735 }
736 #endif
737 ip_forward(m, srcrt, ifp);
738 m_put_rcvif_psref(ifp, &psref);
739 }
740 return;
741
742 ours:
743 m_put_rcvif_psref(ifp, &psref);
744 ifp = NULL;
745
746 /*
747 * If offset or IP_MF are set, must reassemble.
748 */
749 if (ip->ip_off & ~htons(IP_DF|IP_RF)) {
750 /*
751 * Pass to IP reassembly mechanism.
752 */
753 if (ip_reass_packet(&m, ip) != 0) {
754 /* Failed; invalid fragment(s) or packet. */
755 goto out;
756 }
757 if (m == NULL) {
758 /* More fragments should come; silently return. */
759 goto out;
760 }
761 /*
762 * Reassembly is done, we have the final packet.
763 * Update cached data in local variable(s).
764 */
765 ip = mtod(m, struct ip *);
766 hlen = ip->ip_hl << 2;
767 }
768
769 #ifdef IPSEC
770 /*
771 * Enforce IPsec policy checking if we are seeing last header.
772 * Note that we do not visit this with protocols with PCB layer
773 * code - like UDP/TCP/raw IP.
774 */
775 if (ipsec_used &&
776 (inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) {
777 if (ipsec4_input(m, 0) != 0) {
778 goto out;
779 }
780 }
781 #endif
782
783 /*
784 * Switch out to protocol's input routine.
785 */
786 #if IFA_STATS
787 if (ia) {
788 struct in_ifaddr *_ia;
789 /*
790 * Keep a reference from ip_match_our_address with psref
791 * is expensive, so explore ia here again.
792 */
793 s = pserialize_read_enter();
794 _ia = in_get_ia(ip->ip_dst);
795 _ia->ia_ifa.ifa_data.ifad_inbytes += ntohs(ip->ip_len);
796 pserialize_read_exit(s);
797 }
798 #endif
799 IP_STATINC(IP_STAT_DELIVERED);
800
801 const int off = hlen, nh = ip->ip_p;
802
803 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh);
804 return;
805
806 out:
807 m_put_rcvif_psref(ifp, &psref);
808 if (m != NULL)
809 m_freem(m);
810 }
811
812 /*
813 * IP timer processing.
814 */
815 void
816 ip_slowtimo(void)
817 {
818
819 SOFTNET_KERNEL_LOCK_UNLESS_NET_MPSAFE();
820
821 ip_reass_slowtimo();
822
823 SOFTNET_KERNEL_UNLOCK_UNLESS_NET_MPSAFE();
824 }
825
826 /*
827 * IP drain processing.
828 */
829 void
830 ip_drain(void)
831 {
832
833 KERNEL_LOCK(1, NULL);
834 ip_reass_drain();
835 KERNEL_UNLOCK_ONE(NULL);
836 }
837
838 /*
839 * ip_dooptions: perform option processing on a datagram, possibly discarding
840 * it if bad options are encountered, or forwarding it if source-routed.
841 *
842 * => Returns true if packet has been forwarded/freed.
843 * => Returns false if the packet should be processed further.
844 */
845 static bool
846 ip_dooptions(struct mbuf *m)
847 {
848 struct ip *ip = mtod(m, struct ip *);
849 u_char *cp, *cp0;
850 struct ip_timestamp *ipt;
851 struct in_ifaddr *ia;
852 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0;
853 int srr_present, rr_present, ts_present;
854 struct in_addr dst;
855 n_time ntime;
856 struct ifaddr *ifa = NULL;
857 int s;
858
859 srr_present = 0;
860 rr_present = 0;
861 ts_present = 0;
862
863 dst = ip->ip_dst;
864 cp = (u_char *)(ip + 1);
865 cnt = (ip->ip_hl << 2) - sizeof(struct ip);
866 for (; cnt > 0; cnt -= optlen, cp += optlen) {
867 opt = cp[IPOPT_OPTVAL];
868 if (opt == IPOPT_EOL)
869 break;
870 if (opt == IPOPT_NOP)
871 optlen = 1;
872 else {
873 if (cnt < IPOPT_OLEN + sizeof(*cp)) {
874 code = &cp[IPOPT_OLEN] - (u_char *)ip;
875 goto bad;
876 }
877 optlen = cp[IPOPT_OLEN];
878 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
879 code = &cp[IPOPT_OLEN] - (u_char *)ip;
880 goto bad;
881 }
882 }
883 switch (opt) {
884
885 default:
886 break;
887
888 /*
889 * Source routing with record.
890 * Find interface with current destination address.
891 * If none on this machine then drop if strictly routed,
892 * or do nothing if loosely routed.
893 * Record interface address and bring up next address
894 * component. If strictly routed make sure next
895 * address is on directly accessible net.
896 */
897 case IPOPT_LSRR:
898 case IPOPT_SSRR: {
899 struct psref psref;
900 struct sockaddr_in ipaddr = {
901 .sin_len = sizeof(ipaddr),
902 .sin_family = AF_INET,
903 };
904
905 if (ip_allowsrcrt == 0) {
906 type = ICMP_UNREACH;
907 code = ICMP_UNREACH_NET_PROHIB;
908 goto bad;
909 }
910 if (srr_present++) {
911 code = &cp[IPOPT_OPTVAL] - (u_char *)ip;
912 goto bad;
913 }
914 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
915 code = &cp[IPOPT_OLEN] - (u_char *)ip;
916 goto bad;
917 }
918 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
919 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
920 goto bad;
921 }
922 ipaddr.sin_addr = ip->ip_dst;
923
924 s = pserialize_read_enter();
925 ifa = ifa_ifwithaddr(sintosa(&ipaddr));
926 if (ifa == NULL) {
927 pserialize_read_exit(s);
928 if (opt == IPOPT_SSRR) {
929 type = ICMP_UNREACH;
930 code = ICMP_UNREACH_SRCFAIL;
931 goto bad;
932 }
933 /*
934 * Loose routing, and not at next destination
935 * yet; nothing to do except forward.
936 */
937 break;
938 }
939 pserialize_read_exit(s);
940
941 off--; /* 0 origin */
942 if ((off + sizeof(struct in_addr)) > optlen) {
943 /*
944 * End of source route. Should be for us.
945 */
946 save_rte(m, cp, ip->ip_src);
947 break;
948 }
949 /*
950 * locate outgoing interface
951 */
952 memcpy((void *)&ipaddr.sin_addr, (void *)(cp + off),
953 sizeof(ipaddr.sin_addr));
954 if (opt == IPOPT_SSRR) {
955 ifa = ifa_ifwithladdr_psref(sintosa(&ipaddr),
956 &psref);
957 if (ifa != NULL)
958 ia = ifatoia(ifa);
959 else
960 ia = NULL;
961 } else {
962 ia = ip_rtaddr(ipaddr.sin_addr, &psref);
963 }
964 if (ia == NULL) {
965 type = ICMP_UNREACH;
966 code = ICMP_UNREACH_SRCFAIL;
967 goto bad;
968 }
969 ip->ip_dst = ipaddr.sin_addr;
970 memcpy(cp + off, &ia->ia_addr.sin_addr,
971 sizeof(struct in_addr));
972 ia4_release(ia, &psref);
973 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
974 /*
975 * Let ip_intr's mcast routing check handle mcast pkts
976 */
977 forward = !IN_MULTICAST(ip->ip_dst.s_addr);
978 break;
979 }
980
981 case IPOPT_RR: {
982 struct psref psref;
983 struct sockaddr_in ipaddr = {
984 .sin_len = sizeof(ipaddr),
985 .sin_family = AF_INET,
986 };
987
988 if (rr_present++) {
989 code = &cp[IPOPT_OPTVAL] - (u_char *)ip;
990 goto bad;
991 }
992 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
993 code = &cp[IPOPT_OLEN] - (u_char *)ip;
994 goto bad;
995 }
996 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
997 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
998 goto bad;
999 }
1000 /*
1001 * If no space remains, ignore.
1002 */
1003 off--; /* 0 origin */
1004 if ((off + sizeof(struct in_addr)) > optlen)
1005 break;
1006 memcpy((void *)&ipaddr.sin_addr, (void *)&ip->ip_dst,
1007 sizeof(ipaddr.sin_addr));
1008 /*
1009 * locate outgoing interface; if we're the destination,
1010 * use the incoming interface (should be same).
1011 */
1012 ifa = ifa_ifwithaddr_psref(sintosa(&ipaddr), &psref);
1013 if (ifa == NULL) {
1014 ia = ip_rtaddr(ipaddr.sin_addr, &psref);
1015 if (ia == NULL) {
1016 type = ICMP_UNREACH;
1017 code = ICMP_UNREACH_HOST;
1018 goto bad;
1019 }
1020 } else {
1021 ia = ifatoia(ifa);
1022 }
1023 memcpy(cp + off, &ia->ia_addr.sin_addr,
1024 sizeof(struct in_addr));
1025 ia4_release(ia, &psref);
1026 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1027 break;
1028 }
1029
1030 case IPOPT_TS:
1031 code = cp - (u_char *)ip;
1032 ipt = (struct ip_timestamp *)cp;
1033 if (ts_present++) {
1034 code = &cp[IPOPT_OPTVAL] - (u_char *)ip;
1035 goto bad;
1036 }
1037 if (ipt->ipt_len < 4 || ipt->ipt_len > 40) {
1038 code = (u_char *)&ipt->ipt_len - (u_char *)ip;
1039 goto bad;
1040 }
1041 if (ipt->ipt_ptr < 5) {
1042 code = (u_char *)&ipt->ipt_ptr - (u_char *)ip;
1043 goto bad;
1044 }
1045 if (ipt->ipt_ptr > ipt->ipt_len - sizeof(int32_t)) {
1046 if (++ipt->ipt_oflw == 0) {
1047 code = (u_char *)&ipt->ipt_ptr -
1048 (u_char *)ip;
1049 goto bad;
1050 }
1051 break;
1052 }
1053 cp0 = (cp + ipt->ipt_ptr - 1);
1054 switch (ipt->ipt_flg) {
1055
1056 case IPOPT_TS_TSONLY:
1057 break;
1058
1059 case IPOPT_TS_TSANDADDR: {
1060 struct ifnet *rcvif;
1061 int _s, _ss;
1062 struct sockaddr_in ipaddr = {
1063 .sin_len = sizeof(ipaddr),
1064 .sin_family = AF_INET,
1065 };
1066
1067 if (ipt->ipt_ptr - 1 + sizeof(n_time) +
1068 sizeof(struct in_addr) > ipt->ipt_len) {
1069 code = (u_char *)&ipt->ipt_ptr -
1070 (u_char *)ip;
1071 goto bad;
1072 }
1073 ipaddr.sin_addr = dst;
1074 _ss = pserialize_read_enter();
1075 rcvif = m_get_rcvif(m, &_s);
1076 if (__predict_true(rcvif != NULL)) {
1077 ifa = ifaof_ifpforaddr(sintosa(&ipaddr),
1078 rcvif);
1079 }
1080 m_put_rcvif(rcvif, &_s);
1081 if (ifa == NULL) {
1082 pserialize_read_exit(_ss);
1083 break;
1084 }
1085 ia = ifatoia(ifa);
1086 memcpy(cp0, &ia->ia_addr.sin_addr,
1087 sizeof(struct in_addr));
1088 pserialize_read_exit(_ss);
1089 ipt->ipt_ptr += sizeof(struct in_addr);
1090 break;
1091 }
1092
1093 case IPOPT_TS_PRESPEC: {
1094 struct sockaddr_in ipaddr = {
1095 .sin_len = sizeof(ipaddr),
1096 .sin_family = AF_INET,
1097 };
1098
1099 if (ipt->ipt_ptr - 1 + sizeof(n_time) +
1100 sizeof(struct in_addr) > ipt->ipt_len) {
1101 code = (u_char *)&ipt->ipt_ptr -
1102 (u_char *)ip;
1103 goto bad;
1104 }
1105 memcpy(&ipaddr.sin_addr, cp0,
1106 sizeof(struct in_addr));
1107 s = pserialize_read_enter();
1108 ifa = ifa_ifwithaddr(sintosa(&ipaddr));
1109 if (ifa == NULL) {
1110 pserialize_read_exit(s);
1111 continue;
1112 }
1113 pserialize_read_exit(s);
1114 ipt->ipt_ptr += sizeof(struct in_addr);
1115 break;
1116 }
1117
1118 default:
1119 /* XXX can't take &ipt->ipt_flg */
1120 code = (u_char *)&ipt->ipt_ptr -
1121 (u_char *)ip + 1;
1122 goto bad;
1123 }
1124 ntime = iptime();
1125 cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */
1126 memmove((char *)cp + ipt->ipt_ptr - 1, cp0,
1127 sizeof(n_time));
1128 ipt->ipt_ptr += sizeof(n_time);
1129 }
1130 }
1131 if (forward) {
1132 struct ifnet *rcvif;
1133 struct psref _psref;
1134
1135 if (ip_forwsrcrt == 0) {
1136 type = ICMP_UNREACH;
1137 code = ICMP_UNREACH_SRCFAIL;
1138 goto bad;
1139 }
1140
1141 rcvif = m_get_rcvif_psref(m, &_psref);
1142 if (__predict_false(rcvif == NULL)) {
1143 type = ICMP_UNREACH;
1144 code = ICMP_UNREACH_HOST;
1145 goto bad;
1146 }
1147 ip_forward(m, 1, rcvif);
1148 m_put_rcvif_psref(rcvif, &_psref);
1149 return true;
1150 }
1151 return false;
1152 bad:
1153 icmp_error(m, type, code, 0, 0);
1154 IP_STATINC(IP_STAT_BADOPTIONS);
1155 return true;
1156 }
1157
1158 /*
1159 * ip_rtaddr: given address of next destination (final or next hop),
1160 * return internet address info of interface to be used to get there.
1161 */
1162 static struct in_ifaddr *
1163 ip_rtaddr(struct in_addr dst, struct psref *psref)
1164 {
1165 struct rtentry *rt;
1166 union {
1167 struct sockaddr dst;
1168 struct sockaddr_in dst4;
1169 } u;
1170 struct route *ro;
1171
1172 sockaddr_in_init(&u.dst4, &dst, 0);
1173
1174 ro = percpu_getref(ipforward_rt_percpu);
1175 rt = rtcache_lookup(ro, &u.dst);
1176 if (rt == NULL) {
1177 percpu_putref(ipforward_rt_percpu);
1178 return NULL;
1179 }
1180
1181 ia4_acquire(ifatoia(rt->rt_ifa), psref);
1182 rtcache_unref(rt, ro);
1183 percpu_putref(ipforward_rt_percpu);
1184
1185 return ifatoia(rt->rt_ifa);
1186 }
1187
1188 /*
1189 * save_rte: save incoming source route for use in replies, to be picked
1190 * up later by ip_srcroute if the receiver is interested.
1191 */
1192 static void
1193 save_rte(struct mbuf *m, u_char *option, struct in_addr dst)
1194 {
1195 struct ip_srcrt *isr;
1196 struct m_tag *mtag;
1197 unsigned olen;
1198
1199 olen = option[IPOPT_OLEN];
1200 if (olen > sizeof(isr->isr_hdr) + sizeof(isr->isr_routes))
1201 return;
1202
1203 mtag = m_tag_get(PACKET_TAG_SRCROUTE, sizeof(*isr), M_NOWAIT);
1204 if (mtag == NULL)
1205 return;
1206 isr = (struct ip_srcrt *)(mtag + 1);
1207
1208 memcpy(isr->isr_hdr, option, olen);
1209 isr->isr_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1210 isr->isr_dst = dst;
1211 m_tag_prepend(m, mtag);
1212 }
1213
1214 /*
1215 * Retrieve incoming source route for use in replies,
1216 * in the same form used by setsockopt.
1217 * The first hop is placed before the options, will be removed later.
1218 */
1219 struct mbuf *
1220 ip_srcroute(struct mbuf *m0)
1221 {
1222 struct in_addr *p, *q;
1223 struct mbuf *m;
1224 struct ip_srcrt *isr;
1225 struct m_tag *mtag;
1226
1227 mtag = m_tag_find(m0, PACKET_TAG_SRCROUTE, NULL);
1228 if (mtag == NULL)
1229 return NULL;
1230 isr = (struct ip_srcrt *)(mtag + 1);
1231
1232 if (isr->isr_nhops == 0)
1233 return NULL;
1234
1235 m = m_get(M_DONTWAIT, MT_SOOPTS);
1236 if (m == NULL)
1237 return NULL;
1238
1239 MCLAIM(m, &inetdomain.dom_mowner);
1240 #define OPTSIZ (sizeof(isr->isr_nop) + sizeof(isr->isr_hdr))
1241
1242 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + header) */
1243 m->m_len = (isr->isr_nhops + 1) * sizeof(struct in_addr) + OPTSIZ;
1244
1245 /*
1246 * First save first hop for return route
1247 */
1248 p = &(isr->isr_routes[isr->isr_nhops - 1]);
1249 *(mtod(m, struct in_addr *)) = *p--;
1250
1251 /*
1252 * Copy option fields and padding (nop) to mbuf.
1253 */
1254 isr->isr_nop = IPOPT_NOP;
1255 isr->isr_hdr[IPOPT_OFFSET] = IPOPT_MINOFF;
1256 memmove(mtod(m, char *) + sizeof(struct in_addr), &isr->isr_nop,
1257 OPTSIZ);
1258 q = (struct in_addr *)(mtod(m, char *) +
1259 sizeof(struct in_addr) + OPTSIZ);
1260 #undef OPTSIZ
1261 /*
1262 * Record return path as an IP source route,
1263 * reversing the path (pointers are now aligned).
1264 */
1265 while (p >= isr->isr_routes) {
1266 *q++ = *p--;
1267 }
1268 /*
1269 * Last hop goes to final destination.
1270 */
1271 *q = isr->isr_dst;
1272 m_tag_delete(m0, mtag);
1273 return m;
1274 }
1275
1276 const int inetctlerrmap[PRC_NCMDS] = {
1277 [PRC_MSGSIZE] = EMSGSIZE,
1278 [PRC_HOSTDEAD] = EHOSTDOWN,
1279 [PRC_HOSTUNREACH] = EHOSTUNREACH,
1280 [PRC_UNREACH_NET] = EHOSTUNREACH,
1281 [PRC_UNREACH_HOST] = EHOSTUNREACH,
1282 [PRC_UNREACH_PROTOCOL] = ECONNREFUSED,
1283 [PRC_UNREACH_PORT] = ECONNREFUSED,
1284 [PRC_UNREACH_SRCFAIL] = EHOSTUNREACH,
1285 [PRC_PARAMPROB] = ENOPROTOOPT,
1286 };
1287
1288 void
1289 ip_fasttimo(void)
1290 {
1291 if (ip_drainwanted) {
1292 ip_drain();
1293 ip_drainwanted = 0;
1294 }
1295 }
1296
1297 void
1298 ip_drainstub(void)
1299 {
1300 ip_drainwanted = 1;
1301 }
1302
1303 /*
1304 * Forward a packet. If some error occurs return the sender
1305 * an icmp packet. Note we can't always generate a meaningful
1306 * icmp message because icmp doesn't have a large enough repertoire
1307 * of codes and types.
1308 *
1309 * If not forwarding, just drop the packet. This could be confusing
1310 * if ipforwarding was zero but some routing protocol was advancing
1311 * us as a gateway to somewhere. However, we must let the routing
1312 * protocol deal with that.
1313 *
1314 * The srcrt parameter indicates whether the packet is being forwarded
1315 * via a source route.
1316 */
1317 static void
1318 ip_forward(struct mbuf *m, int srcrt, struct ifnet *rcvif)
1319 {
1320 struct ip *ip = mtod(m, struct ip *);
1321 struct rtentry *rt;
1322 int error, type = 0, code = 0, destmtu = 0;
1323 struct mbuf *mcopy;
1324 n_long dest;
1325 union {
1326 struct sockaddr dst;
1327 struct sockaddr_in dst4;
1328 } u;
1329 uint64_t *ips;
1330 struct route *ro;
1331
1332 KASSERTMSG(cpu_softintr_p(), "ip_forward: not in the software "
1333 "interrupt handler; synchronization assumptions violated");
1334
1335 /*
1336 * We are now in the output path.
1337 */
1338 MCLAIM(m, &ip_tx_mowner);
1339
1340 /*
1341 * Clear any in-bound checksum flags for this packet.
1342 */
1343 m->m_pkthdr.csum_flags = 0;
1344
1345 dest = 0;
1346 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
1347 IP_STATINC(IP_STAT_CANTFORWARD);
1348 m_freem(m);
1349 return;
1350 }
1351
1352 if (ip->ip_ttl <= IPTTLDEC) {
1353 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0);
1354 return;
1355 }
1356
1357 sockaddr_in_init(&u.dst4, &ip->ip_dst, 0);
1358
1359 ro = percpu_getref(ipforward_rt_percpu);
1360 rt = rtcache_lookup(ro, &u.dst);
1361 if (rt == NULL) {
1362 percpu_putref(ipforward_rt_percpu);
1363 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_NET, dest, 0);
1364 return;
1365 }
1366
1367 /*
1368 * Save at most 68 bytes of the packet in case
1369 * we need to generate an ICMP message to the src.
1370 * Pullup to avoid sharing mbuf cluster between m and mcopy.
1371 */
1372 mcopy = m_copym(m, 0, imin(ntohs(ip->ip_len), 68), M_DONTWAIT);
1373 if (mcopy)
1374 mcopy = m_pullup(mcopy, ip->ip_hl << 2);
1375
1376 ip->ip_ttl -= IPTTLDEC;
1377
1378 /*
1379 * If forwarding packet using same interface that it came in on,
1380 * perhaps should send a redirect to sender to shortcut a hop.
1381 * Only send redirect if source is sending directly to us,
1382 * and if packet was not source routed (or has any options).
1383 * Also, don't send redirect if forwarding using a default route
1384 * or a route modified by a redirect.
1385 */
1386 if (rt->rt_ifp == rcvif &&
1387 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1388 !in_nullhost(satocsin(rt_getkey(rt))->sin_addr) &&
1389 ipsendredirects && !srcrt) {
1390 if (rt->rt_ifa &&
1391 (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) ==
1392 ifatoia(rt->rt_ifa)->ia_subnet) {
1393 if (rt->rt_flags & RTF_GATEWAY)
1394 dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
1395 else
1396 dest = ip->ip_dst.s_addr;
1397 /*
1398 * Router requirements says to only send host
1399 * redirects.
1400 */
1401 type = ICMP_REDIRECT;
1402 code = ICMP_REDIRECT_HOST;
1403 }
1404 }
1405 rtcache_unref(rt, ro);
1406
1407 error = ip_output(m, NULL, ro,
1408 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)),
1409 NULL, NULL);
1410
1411 if (error) {
1412 IP_STATINC(IP_STAT_CANTFORWARD);
1413 goto error;
1414 }
1415
1416 ips = IP_STAT_GETREF();
1417 ips[IP_STAT_FORWARD]++;
1418
1419 if (type) {
1420 ips[IP_STAT_REDIRECTSENT]++;
1421 IP_STAT_PUTREF();
1422 goto redirect;
1423 }
1424
1425 IP_STAT_PUTREF();
1426 if (mcopy) {
1427 #ifdef GATEWAY
1428 if (mcopy->m_flags & M_CANFASTFWD)
1429 ipflow_create(ro, mcopy);
1430 #endif
1431 m_freem(mcopy);
1432 }
1433
1434 percpu_putref(ipforward_rt_percpu);
1435 return;
1436
1437 redirect:
1438 error:
1439 if (mcopy == NULL) {
1440 percpu_putref(ipforward_rt_percpu);
1441 return;
1442 }
1443
1444 switch (error) {
1445
1446 case 0: /* forwarded, but need redirect */
1447 /* type, code set above */
1448 break;
1449
1450 case ENETUNREACH: /* shouldn't happen, checked above */
1451 case EHOSTUNREACH:
1452 case ENETDOWN:
1453 case EHOSTDOWN:
1454 default:
1455 type = ICMP_UNREACH;
1456 code = ICMP_UNREACH_HOST;
1457 break;
1458
1459 case EMSGSIZE:
1460 type = ICMP_UNREACH;
1461 code = ICMP_UNREACH_NEEDFRAG;
1462
1463 if ((rt = rtcache_validate(ro)) != NULL) {
1464 destmtu = rt->rt_ifp->if_mtu;
1465 rtcache_unref(rt, ro);
1466 }
1467 #ifdef IPSEC
1468 if (ipsec_used)
1469 (void)ipsec4_forward(mcopy, &destmtu);
1470 #endif
1471 IP_STATINC(IP_STAT_CANTFRAG);
1472 break;
1473
1474 case ENOBUFS:
1475 /*
1476 * Do not generate ICMP_SOURCEQUENCH as required in RFC 1812,
1477 * Requirements for IP Version 4 Routers. Source quench can
1478 * be a big problem under DoS attacks or if the underlying
1479 * interface is rate-limited.
1480 */
1481 if (mcopy)
1482 m_freem(mcopy);
1483 percpu_putref(ipforward_rt_percpu);
1484 return;
1485 }
1486 icmp_error(mcopy, type, code, dest, destmtu);
1487 percpu_putref(ipforward_rt_percpu);
1488 }
1489
1490 void
1491 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1492 struct mbuf *m)
1493 {
1494 struct socket *so = inp->inp_socket;
1495 int inpflags = inp->inp_flags;
1496
1497 if (SOOPT_TIMESTAMP(so->so_options))
1498 mp = sbsavetimestamp(so->so_options, m, mp);
1499
1500 if (inpflags & INP_RECVDSTADDR) {
1501 *mp = sbcreatecontrol(&ip->ip_dst,
1502 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1503 if (*mp)
1504 mp = &(*mp)->m_next;
1505 }
1506
1507 if (inpflags & INP_RECVTTL) {
1508 *mp = sbcreatecontrol(&ip->ip_ttl,
1509 sizeof(uint8_t), IP_RECVTTL, IPPROTO_IP);
1510 if (*mp)
1511 mp = &(*mp)->m_next;
1512 }
1513
1514 struct psref psref;
1515 ifnet_t *ifp = m_get_rcvif_psref(m, &psref);
1516 if (__predict_false(ifp == NULL)) {
1517 #ifdef DIAGNOSTIC
1518 printf("%s: missing receive interface\n", __func__);
1519 #endif
1520 return; /* XXX should report error? */
1521 }
1522
1523 if (inpflags & INP_RECVPKTINFO) {
1524 struct in_pktinfo ipi;
1525 ipi.ipi_addr = ip->ip_dst;
1526 ipi.ipi_ifindex = ifp->if_index;
1527 *mp = sbcreatecontrol(&ipi,
1528 sizeof(ipi), IP_PKTINFO, IPPROTO_IP);
1529 if (*mp)
1530 mp = &(*mp)->m_next;
1531 }
1532 if (inpflags & INP_RECVIF) {
1533 struct sockaddr_dl sdl;
1534
1535 sockaddr_dl_init(&sdl, sizeof(sdl), ifp->if_index, 0, NULL, 0,
1536 NULL, 0);
1537 *mp = sbcreatecontrol(&sdl, sdl.sdl_len, IP_RECVIF, IPPROTO_IP);
1538 if (*mp)
1539 mp = &(*mp)->m_next;
1540 }
1541 m_put_rcvif_psref(ifp, &psref);
1542 }
1543
1544 /*
1545 * sysctl helper routine for net.inet.ip.forwsrcrt.
1546 */
1547 static int
1548 sysctl_net_inet_ip_forwsrcrt(SYSCTLFN_ARGS)
1549 {
1550 int error, tmp;
1551 struct sysctlnode node;
1552
1553 node = *rnode;
1554 tmp = ip_forwsrcrt;
1555 node.sysctl_data = &tmp;
1556 error = sysctl_lookup(SYSCTLFN_CALL(&node));
1557 if (error || newp == NULL)
1558 return (error);
1559
1560 error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_FORWSRCRT,
1561 0, NULL, NULL, NULL);
1562 if (error)
1563 return (error);
1564
1565 ip_forwsrcrt = tmp;
1566
1567 return (0);
1568 }
1569
1570 /*
1571 * sysctl helper routine for net.inet.ip.mtudisctimeout. checks the
1572 * range of the new value and tweaks timers if it changes.
1573 */
1574 static int
1575 sysctl_net_inet_ip_pmtudto(SYSCTLFN_ARGS)
1576 {
1577 int error, tmp;
1578 struct sysctlnode node;
1579
1580 icmp_mtudisc_lock();
1581
1582 node = *rnode;
1583 tmp = ip_mtudisc_timeout;
1584 node.sysctl_data = &tmp;
1585 error = sysctl_lookup(SYSCTLFN_CALL(&node));
1586 if (error || newp == NULL)
1587 goto out;
1588 if (tmp < 0) {
1589 error = EINVAL;
1590 goto out;
1591 }
1592
1593 ip_mtudisc_timeout = tmp;
1594 rt_timer_queue_change(ip_mtudisc_timeout_q, ip_mtudisc_timeout);
1595 error = 0;
1596 out:
1597 icmp_mtudisc_unlock();
1598 return error;
1599 }
1600
1601 static int
1602 sysctl_net_inet_ip_stats(SYSCTLFN_ARGS)
1603 {
1604
1605 return (NETSTAT_SYSCTL(ipstat_percpu, IP_NSTATS));
1606 }
1607
1608 static void
1609 sysctl_net_inet_ip_setup(struct sysctllog **clog)
1610 {
1611 sysctl_createv(clog, 0, NULL, NULL,
1612 CTLFLAG_PERMANENT,
1613 CTLTYPE_NODE, "inet",
1614 SYSCTL_DESCR("PF_INET related settings"),
1615 NULL, 0, NULL, 0,
1616 CTL_NET, PF_INET, CTL_EOL);
1617 sysctl_createv(clog, 0, NULL, NULL,
1618 CTLFLAG_PERMANENT,
1619 CTLTYPE_NODE, "ip",
1620 SYSCTL_DESCR("IPv4 related settings"),
1621 NULL, 0, NULL, 0,
1622 CTL_NET, PF_INET, IPPROTO_IP, CTL_EOL);
1623
1624 sysctl_createv(clog, 0, NULL, NULL,
1625 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1626 CTLTYPE_INT, "forwarding",
1627 SYSCTL_DESCR("Enable forwarding of INET datagrams"),
1628 NULL, 0, &ipforwarding, 0,
1629 CTL_NET, PF_INET, IPPROTO_IP,
1630 IPCTL_FORWARDING, CTL_EOL);
1631 sysctl_createv(clog, 0, NULL, NULL,
1632 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1633 CTLTYPE_INT, "redirect",
1634 SYSCTL_DESCR("Enable sending of ICMP redirect messages"),
1635 NULL, 0, &ipsendredirects, 0,
1636 CTL_NET, PF_INET, IPPROTO_IP,
1637 IPCTL_SENDREDIRECTS, CTL_EOL);
1638 sysctl_createv(clog, 0, NULL, NULL,
1639 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1640 CTLTYPE_INT, "ttl",
1641 SYSCTL_DESCR("Default TTL for an INET datagram"),
1642 NULL, 0, &ip_defttl, 0,
1643 CTL_NET, PF_INET, IPPROTO_IP,
1644 IPCTL_DEFTTL, CTL_EOL);
1645 sysctl_createv(clog, 0, NULL, NULL,
1646 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1647 CTLTYPE_INT, "forwsrcrt",
1648 SYSCTL_DESCR("Enable forwarding of source-routed "
1649 "datagrams"),
1650 sysctl_net_inet_ip_forwsrcrt, 0, &ip_forwsrcrt, 0,
1651 CTL_NET, PF_INET, IPPROTO_IP,
1652 IPCTL_FORWSRCRT, CTL_EOL);
1653 sysctl_createv(clog, 0, NULL, NULL,
1654 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1655 CTLTYPE_INT, "directed-broadcast",
1656 SYSCTL_DESCR("Enable forwarding of broadcast datagrams"),
1657 NULL, 0, &ip_directedbcast, 0,
1658 CTL_NET, PF_INET, IPPROTO_IP,
1659 IPCTL_DIRECTEDBCAST, CTL_EOL);
1660 sysctl_createv(clog, 0, NULL, NULL,
1661 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1662 CTLTYPE_INT, "allowsrcrt",
1663 SYSCTL_DESCR("Accept source-routed datagrams"),
1664 NULL, 0, &ip_allowsrcrt, 0,
1665 CTL_NET, PF_INET, IPPROTO_IP,
1666 IPCTL_ALLOWSRCRT, CTL_EOL);
1667
1668 sysctl_createv(clog, 0, NULL, NULL,
1669 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1670 CTLTYPE_INT, "mtudisc",
1671 SYSCTL_DESCR("Use RFC1191 Path MTU Discovery"),
1672 NULL, 0, &ip_mtudisc, 0,
1673 CTL_NET, PF_INET, IPPROTO_IP,
1674 IPCTL_MTUDISC, CTL_EOL);
1675 sysctl_createv(clog, 0, NULL, NULL,
1676 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1677 CTLTYPE_INT, "anonportmin",
1678 SYSCTL_DESCR("Lowest ephemeral port number to assign"),
1679 sysctl_net_inet_ip_ports, 0, &anonportmin, 0,
1680 CTL_NET, PF_INET, IPPROTO_IP,
1681 IPCTL_ANONPORTMIN, CTL_EOL);
1682 sysctl_createv(clog, 0, NULL, NULL,
1683 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1684 CTLTYPE_INT, "anonportmax",
1685 SYSCTL_DESCR("Highest ephemeral port number to assign"),
1686 sysctl_net_inet_ip_ports, 0, &anonportmax, 0,
1687 CTL_NET, PF_INET, IPPROTO_IP,
1688 IPCTL_ANONPORTMAX, CTL_EOL);
1689 sysctl_createv(clog, 0, NULL, NULL,
1690 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1691 CTLTYPE_INT, "mtudisctimeout",
1692 SYSCTL_DESCR("Lifetime of a Path MTU Discovered route"),
1693 sysctl_net_inet_ip_pmtudto, 0, (void *)&ip_mtudisc_timeout, 0,
1694 CTL_NET, PF_INET, IPPROTO_IP,
1695 IPCTL_MTUDISCTIMEOUT, CTL_EOL);
1696 #ifndef IPNOPRIVPORTS
1697 sysctl_createv(clog, 0, NULL, NULL,
1698 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1699 CTLTYPE_INT, "lowportmin",
1700 SYSCTL_DESCR("Lowest privileged ephemeral port number "
1701 "to assign"),
1702 sysctl_net_inet_ip_ports, 0, &lowportmin, 0,
1703 CTL_NET, PF_INET, IPPROTO_IP,
1704 IPCTL_LOWPORTMIN, CTL_EOL);
1705 sysctl_createv(clog, 0, NULL, NULL,
1706 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1707 CTLTYPE_INT, "lowportmax",
1708 SYSCTL_DESCR("Highest privileged ephemeral port number "
1709 "to assign"),
1710 sysctl_net_inet_ip_ports, 0, &lowportmax, 0,
1711 CTL_NET, PF_INET, IPPROTO_IP,
1712 IPCTL_LOWPORTMAX, CTL_EOL);
1713 #endif /* IPNOPRIVPORTS */
1714 #if NGRE > 0
1715 sysctl_createv(clog, 0, NULL, NULL,
1716 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1717 CTLTYPE_INT, "grettl",
1718 SYSCTL_DESCR("Default TTL for a gre tunnel datagram"),
1719 NULL, 0, &ip_gre_ttl, 0,
1720 CTL_NET, PF_INET, IPPROTO_IP,
1721 IPCTL_GRE_TTL, CTL_EOL);
1722 #endif /* NGRE */
1723 sysctl_createv(clog, 0, NULL, NULL,
1724 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1725 CTLTYPE_INT, "checkinterface",
1726 SYSCTL_DESCR("Enable receive side of Strong ES model "
1727 "from RFC1122"),
1728 NULL, 0, &ip_checkinterface, 0,
1729 CTL_NET, PF_INET, IPPROTO_IP,
1730 IPCTL_CHECKINTERFACE, CTL_EOL);
1731 sysctl_createv(clog, 0, NULL, NULL,
1732 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1733 CTLTYPE_INT, "random_id",
1734 SYSCTL_DESCR("Assign random ip_id values"),
1735 NULL, 0, &ip_do_randomid, 0,
1736 CTL_NET, PF_INET, IPPROTO_IP,
1737 IPCTL_RANDOMID, CTL_EOL);
1738 sysctl_createv(clog, 0, NULL, NULL,
1739 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1740 CTLTYPE_INT, "do_loopback_cksum",
1741 SYSCTL_DESCR("Perform IP checksum on loopback"),
1742 NULL, 0, &ip_do_loopback_cksum, 0,
1743 CTL_NET, PF_INET, IPPROTO_IP,
1744 IPCTL_LOOPBACKCKSUM, CTL_EOL);
1745 sysctl_createv(clog, 0, NULL, NULL,
1746 CTLFLAG_PERMANENT,
1747 CTLTYPE_STRUCT, "stats",
1748 SYSCTL_DESCR("IP statistics"),
1749 sysctl_net_inet_ip_stats, 0, NULL, 0,
1750 CTL_NET, PF_INET, IPPROTO_IP, IPCTL_STATS,
1751 CTL_EOL);
1752 #if NARP
1753 sysctl_createv(clog, 0, NULL, NULL,
1754 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1755 CTLTYPE_INT, "dad_count",
1756 SYSCTL_DESCR("Number of Duplicate Address Detection "
1757 "probes to send"),
1758 NULL, 0, &ip_dad_count, 0,
1759 CTL_NET, PF_INET, IPPROTO_IP,
1760 IPCTL_DAD_COUNT, CTL_EOL);
1761 #endif
1762
1763 /* anonportalgo RFC6056 subtree */
1764 const struct sysctlnode *portalgo_node;
1765 sysctl_createv(clog, 0, NULL, &portalgo_node,
1766 CTLFLAG_PERMANENT,
1767 CTLTYPE_NODE, "anonportalgo",
1768 SYSCTL_DESCR("Anonymous Port Algorithm Selection (RFC 6056)"),
1769 NULL, 0, NULL, 0,
1770 CTL_NET, PF_INET, IPPROTO_IP, CTL_CREATE, CTL_EOL);
1771 sysctl_createv(clog, 0, &portalgo_node, NULL,
1772 CTLFLAG_PERMANENT,
1773 CTLTYPE_STRING, "available",
1774 SYSCTL_DESCR("available algorithms"),
1775 sysctl_portalgo_available, 0, NULL, PORTALGO_MAXLEN,
1776 CTL_CREATE, CTL_EOL);
1777 sysctl_createv(clog, 0, &portalgo_node, NULL,
1778 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1779 CTLTYPE_STRING, "selected",
1780 SYSCTL_DESCR("selected algorithm"),
1781 sysctl_portalgo_selected4, 0, NULL, PORTALGO_MAXLEN,
1782 CTL_CREATE, CTL_EOL);
1783 sysctl_createv(clog, 0, &portalgo_node, NULL,
1784 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1785 CTLTYPE_STRUCT, "reserve",
1786 SYSCTL_DESCR("bitmap of reserved ports"),
1787 sysctl_portalgo_reserve4, 0, NULL, 0,
1788 CTL_CREATE, CTL_EOL);
1789 }
1790
1791 void
1792 ip_statinc(u_int stat)
1793 {
1794
1795 KASSERT(stat < IP_NSTATS);
1796 IP_STATINC(stat);
1797 }
1798