ip_input.c revision 1.343 1 /* $NetBSD: ip_input.c,v 1.343 2016/10/18 01:15:20 ozaki-r 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 1.343 2016/10/18 01:15:20 ozaki-r Exp $");
95
96 #ifdef _KERNEL_OPT
97 #include "opt_inet.h"
98 #include "opt_compat_netbsd.h"
99 #include "opt_gateway.h"
100 #include "opt_ipsec.h"
101 #include "opt_mrouting.h"
102 #include "opt_mbuftrace.h"
103 #include "opt_inet_csum.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 1 /* forward source-routed packets */
165 #endif
166 #ifndef IPALLOWSRCRT
167 #define IPALLOWSRCRT 1 /* 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 #ifdef COMPAT_50
177 #include <compat/sys/time.h>
178 #include <compat/sys/socket.h>
179 #endif
180
181 /*
182 * Note: DIRECTED_BROADCAST is handled this way so that previous
183 * configuration using this option will Just Work.
184 */
185 #ifndef IPDIRECTEDBCAST
186 #ifdef DIRECTED_BROADCAST
187 #define IPDIRECTEDBCAST 1
188 #else
189 #define IPDIRECTEDBCAST 0
190 #endif /* DIRECTED_BROADCAST */
191 #endif /* IPDIRECTEDBCAST */
192 int ipforwarding = IPFORWARDING;
193 int ipsendredirects = IPSENDREDIRECTS;
194 int ip_defttl = IPDEFTTL;
195 int ip_forwsrcrt = IPFORWSRCRT;
196 int ip_directedbcast = IPDIRECTEDBCAST;
197 int ip_allowsrcrt = IPALLOWSRCRT;
198 int ip_mtudisc = IPMTUDISC;
199 int ip_mtudisc_timeout = IPMTUDISCTIMEOUT;
200 #ifdef DIAGNOSTIC
201 int ipprintfs = 0;
202 #endif
203
204 int ip_do_randomid = 0;
205
206 /*
207 * XXX - Setting ip_checkinterface mostly implements the receive side of
208 * the Strong ES model described in RFC 1122, but since the routing table
209 * and transmit implementation do not implement the Strong ES model,
210 * setting this to 1 results in an odd hybrid.
211 *
212 * XXX - ip_checkinterface currently must be disabled if you use ipnat
213 * to translate the destination address to another local interface.
214 *
215 * XXX - ip_checkinterface must be disabled if you add IP aliases
216 * to the loopback interface instead of the interface where the
217 * packets for those addresses are received.
218 */
219 static int ip_checkinterface __read_mostly = 0;
220
221 struct rttimer_queue *ip_mtudisc_timeout_q = NULL;
222
223 pktqueue_t * ip_pktq __read_mostly;
224 pfil_head_t * inet_pfil_hook __read_mostly;
225 ipid_state_t * ip_ids __read_mostly;
226 percpu_t * ipstat_percpu __read_mostly;
227
228 static percpu_t *ipforward_rt_percpu __cacheline_aligned;
229
230 uint16_t ip_id;
231
232 #ifdef INET_CSUM_COUNTERS
233 #include <sys/device.h>
234
235 struct evcnt ip_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
236 NULL, "inet", "hwcsum bad");
237 struct evcnt ip_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
238 NULL, "inet", "hwcsum ok");
239 struct evcnt ip_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
240 NULL, "inet", "swcsum");
241
242 #define INET_CSUM_COUNTER_INCR(ev) (ev)->ev_count++
243
244 EVCNT_ATTACH_STATIC(ip_hwcsum_bad);
245 EVCNT_ATTACH_STATIC(ip_hwcsum_ok);
246 EVCNT_ATTACH_STATIC(ip_swcsum);
247
248 #else
249
250 #define INET_CSUM_COUNTER_INCR(ev) /* nothing */
251
252 #endif /* INET_CSUM_COUNTERS */
253
254 /*
255 * We need to save the IP options in case a protocol wants to respond
256 * to an incoming packet over the same route if the packet got here
257 * using IP source routing. This allows connection establishment and
258 * maintenance when the remote end is on a network that is not known
259 * to us.
260 */
261
262 static int ip_nhops = 0;
263
264 static struct ip_srcrt {
265 struct in_addr dst; /* final destination */
266 char nop; /* one NOP to align */
267 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */
268 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)];
269 } ip_srcrt;
270
271 static int ip_drainwanted;
272
273 struct sockaddr_in ipaddr = {
274 .sin_len = sizeof(ipaddr),
275 .sin_family = AF_INET,
276 };
277
278 static void save_rte(u_char *, struct in_addr);
279
280 #ifdef MBUFTRACE
281 struct mowner ip_rx_mowner = MOWNER_INIT("internet", "rx");
282 struct mowner ip_tx_mowner = MOWNER_INIT("internet", "tx");
283 #endif
284
285 static void ipintr(void *);
286 static void ip_input(struct mbuf *);
287 static void ip_forward(struct mbuf *, int, struct ifnet *);
288 static bool ip_dooptions(struct mbuf *);
289 static struct in_ifaddr *ip_rtaddr(struct in_addr);
290 static void sysctl_net_inet_ip_setup(struct sysctllog **);
291
292 static struct in_ifaddr *ip_match_our_address(struct ifnet *, struct ip *,
293 int *);
294 static struct in_ifaddr *ip_match_our_address_broadcast(struct ifnet *,
295 struct ip *);
296
297 /* XXX: Not yet enabled. */
298 #define SOFTNET_LOCK() KASSERT(mutex_owned(softnet_lock))
299 #define SOFTNET_UNLOCK() KASSERT(mutex_owned(softnet_lock))
300
301 /*
302 * IP initialization: fill in IP protocol switch table.
303 * All protocols not implemented in kernel go to raw IP protocol handler.
304 */
305 void
306 ip_init(void)
307 {
308 const struct protosw *pr;
309
310 in_init();
311 sysctl_net_inet_ip_setup(NULL);
312
313 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
314 KASSERT(pr != NULL);
315
316 ip_pktq = pktq_create(IFQ_MAXLEN, ipintr, NULL);
317 KASSERT(ip_pktq != NULL);
318
319 for (u_int i = 0; i < IPPROTO_MAX; i++) {
320 ip_protox[i] = pr - inetsw;
321 }
322 for (pr = inetdomain.dom_protosw;
323 pr < inetdomain.dom_protoswNPROTOSW; pr++)
324 if (pr->pr_domain->dom_family == PF_INET &&
325 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
326 ip_protox[pr->pr_protocol] = pr - inetsw;
327
328 ip_reass_init();
329
330 ip_ids = ip_id_init();
331 ip_id = time_uptime & 0xfffff;
332
333 ip_mtudisc_timeout_q = rt_timer_queue_create(ip_mtudisc_timeout);
334 #ifdef GATEWAY
335 ipflow_init();
336 #endif
337
338 /* Register our Packet Filter hook. */
339 inet_pfil_hook = pfil_head_create(PFIL_TYPE_AF, (void *)AF_INET);
340 KASSERT(inet_pfil_hook != NULL);
341
342 #ifdef MBUFTRACE
343 MOWNER_ATTACH(&ip_tx_mowner);
344 MOWNER_ATTACH(&ip_rx_mowner);
345 #endif /* MBUFTRACE */
346
347 ipstat_percpu = percpu_alloc(sizeof(uint64_t) * IP_NSTATS);
348
349 ipforward_rt_percpu = percpu_alloc(sizeof(struct route));
350 if (ipforward_rt_percpu == NULL)
351 panic("failed to allocate ipforward_rt_percpu");
352 }
353
354 static struct in_ifaddr *
355 ip_match_our_address(struct ifnet *ifp, struct ip *ip, int *downmatch)
356 {
357 struct in_ifaddr *ia = NULL;
358 int checkif;
359
360 /*
361 * Enable a consistency check between the destination address
362 * and the arrival interface for a unicast packet (the RFC 1122
363 * strong ES model) if IP forwarding is disabled and the packet
364 * is not locally generated.
365 *
366 * XXX - Checking also should be disabled if the destination
367 * address is ipnat'ed to a different interface.
368 *
369 * XXX - Checking is incompatible with IP aliases added
370 * to the loopback interface instead of the interface where
371 * the packets are received.
372 *
373 * XXX - We need to add a per ifaddr flag for this so that
374 * we get finer grain control.
375 */
376 checkif = ip_checkinterface && (ipforwarding == 0) &&
377 (ifp->if_flags & IFF_LOOPBACK) == 0;
378
379 IN_ADDRHASH_READER_FOREACH(ia, ip->ip_dst.s_addr) {
380 if (in_hosteq(ia->ia_addr.sin_addr, ip->ip_dst)) {
381 if (ia->ia4_flags & IN_IFF_NOTREADY)
382 continue;
383 if (checkif && ia->ia_ifp != ifp)
384 continue;
385 if ((ia->ia_ifp->if_flags & IFF_UP) != 0 &&
386 (ia->ia4_flags & IN_IFF_DETACHED) == 0)
387 break;
388 else
389 (*downmatch)++;
390 }
391 }
392
393 return ia;
394 }
395
396 static struct in_ifaddr *
397 ip_match_our_address_broadcast(struct ifnet *ifp, struct ip *ip)
398 {
399 struct in_ifaddr *ia = NULL;
400 struct ifaddr *ifa;
401
402 IFADDR_READER_FOREACH(ifa, ifp) {
403 if (ifa->ifa_addr->sa_family != AF_INET)
404 continue;
405 ia = ifatoia(ifa);
406 if (ia->ia4_flags & (IN_IFF_NOTREADY | IN_IFF_DETACHED))
407 continue;
408 if (in_hosteq(ip->ip_dst, ia->ia_broadaddr.sin_addr) ||
409 in_hosteq(ip->ip_dst, ia->ia_netbroadcast) ||
410 /*
411 * Look for all-0's host part (old broadcast addr),
412 * either for subnet or net.
413 */
414 ip->ip_dst.s_addr == ia->ia_subnet ||
415 ip->ip_dst.s_addr == ia->ia_net)
416 goto matched;
417 /*
418 * An interface with IP address zero accepts
419 * all packets that arrive on that interface.
420 */
421 if (in_nullhost(ia->ia_addr.sin_addr))
422 goto matched;
423 }
424 ia = NULL;
425
426 matched:
427 return ia;
428 }
429
430 /*
431 * IP software interrupt routine.
432 */
433 static void
434 ipintr(void *arg __unused)
435 {
436 struct mbuf *m;
437
438 KASSERT(cpu_softintr_p());
439
440 mutex_enter(softnet_lock);
441 while ((m = pktq_dequeue(ip_pktq)) != NULL) {
442 ip_input(m);
443 }
444 mutex_exit(softnet_lock);
445 }
446
447 /*
448 * IP input routine. Checksum and byte swap header. If fragmented
449 * try to reassemble. Process options. Pass to next level.
450 */
451 static void
452 ip_input(struct mbuf *m)
453 {
454 struct ip *ip = NULL;
455 struct in_ifaddr *ia = NULL;
456 int hlen = 0, len;
457 int downmatch;
458 int srcrt = 0;
459 ifnet_t *ifp;
460 struct psref psref;
461 int s;
462
463 KASSERTMSG(cpu_softintr_p(), "ip_input: not in the software "
464 "interrupt handler; synchronization assumptions violated");
465
466 MCLAIM(m, &ip_rx_mowner);
467 KASSERT((m->m_flags & M_PKTHDR) != 0);
468
469 ifp = m_get_rcvif_psref(m, &psref);
470 if (__predict_false(ifp == NULL))
471 goto out;
472
473 /*
474 * If no IP addresses have been set yet but the interfaces
475 * are receiving, can't do anything with incoming packets yet.
476 * Note: we pre-check without locks held.
477 */
478 if (IN_ADDRLIST_READER_EMPTY())
479 goto out;
480 IP_STATINC(IP_STAT_TOTAL);
481
482 /*
483 * If the IP header is not aligned, slurp it up into a new
484 * mbuf with space for link headers, in the event we forward
485 * it. Otherwise, if it is aligned, make sure the entire
486 * base IP header is in the first mbuf of the chain.
487 */
488 if (IP_HDR_ALIGNED_P(mtod(m, void *)) == 0) {
489 if ((m = m_copyup(m, sizeof(struct ip),
490 (max_linkhdr + 3) & ~3)) == NULL) {
491 /* XXXJRT new stat, please */
492 IP_STATINC(IP_STAT_TOOSMALL);
493 goto out;
494 }
495 } else if (__predict_false(m->m_len < sizeof (struct ip))) {
496 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
497 IP_STATINC(IP_STAT_TOOSMALL);
498 goto out;
499 }
500 }
501 ip = mtod(m, struct ip *);
502 if (ip->ip_v != IPVERSION) {
503 IP_STATINC(IP_STAT_BADVERS);
504 goto out;
505 }
506 hlen = ip->ip_hl << 2;
507 if (hlen < sizeof(struct ip)) { /* minimum header length */
508 IP_STATINC(IP_STAT_BADHLEN);
509 goto out;
510 }
511 if (hlen > m->m_len) {
512 if ((m = m_pullup(m, hlen)) == NULL) {
513 IP_STATINC(IP_STAT_BADHLEN);
514 goto out;
515 }
516 ip = mtod(m, struct ip *);
517 }
518
519 /*
520 * RFC1122: packets with a multicast source address are
521 * not allowed.
522 */
523 if (IN_MULTICAST(ip->ip_src.s_addr)) {
524 IP_STATINC(IP_STAT_BADADDR);
525 goto out;
526 }
527
528 /* 127/8 must not appear on wire - RFC1122 */
529 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
530 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
531 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
532 IP_STATINC(IP_STAT_BADADDR);
533 goto out;
534 }
535 }
536
537 switch (m->m_pkthdr.csum_flags &
538 ((ifp->if_csum_flags_rx & M_CSUM_IPv4) |
539 M_CSUM_IPv4_BAD)) {
540 case M_CSUM_IPv4|M_CSUM_IPv4_BAD:
541 INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad);
542 IP_STATINC(IP_STAT_BADSUM);
543 goto out;
544
545 case M_CSUM_IPv4:
546 /* Checksum was okay. */
547 INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok);
548 break;
549
550 default:
551 /*
552 * Must compute it ourselves. Maybe skip checksum on
553 * loopback interfaces.
554 */
555 if (__predict_true(!(ifp->if_flags & IFF_LOOPBACK) ||
556 ip_do_loopback_cksum)) {
557 INET_CSUM_COUNTER_INCR(&ip_swcsum);
558 if (in_cksum(m, hlen) != 0) {
559 IP_STATINC(IP_STAT_BADSUM);
560 goto out;
561 }
562 }
563 break;
564 }
565
566 /* Retrieve the packet length. */
567 len = ntohs(ip->ip_len);
568
569 /*
570 * Check for additional length bogosity
571 */
572 if (len < hlen) {
573 IP_STATINC(IP_STAT_BADLEN);
574 goto out;
575 }
576
577 /*
578 * Check that the amount of data in the buffers
579 * is as at least much as the IP header would have us expect.
580 * Trim mbufs if longer than we expect.
581 * Drop packet if shorter than we expect.
582 */
583 if (m->m_pkthdr.len < len) {
584 IP_STATINC(IP_STAT_TOOSHORT);
585 goto out;
586 }
587 if (m->m_pkthdr.len > len) {
588 if (m->m_len == m->m_pkthdr.len) {
589 m->m_len = len;
590 m->m_pkthdr.len = len;
591 } else
592 m_adj(m, len - m->m_pkthdr.len);
593 }
594
595 /*
596 * Assume that we can create a fast-forward IP flow entry
597 * based on this packet.
598 */
599 m->m_flags |= M_CANFASTFWD;
600
601 /*
602 * Run through list of hooks for input packets. If there are any
603 * filters which require that additional packets in the flow are
604 * not fast-forwarded, they must clear the M_CANFASTFWD flag.
605 * Note that filters must _never_ set this flag, as another filter
606 * in the list may have previously cleared it.
607 */
608 #if defined(IPSEC)
609 if (!ipsec_used || !ipsec_indone(m))
610 #else
611 if (1)
612 #endif
613 {
614 struct in_addr odst = ip->ip_dst;
615 bool freed;
616
617 SOFTNET_LOCK();
618 freed = pfil_run_hooks(inet_pfil_hook, &m, ifp, PFIL_IN) != 0;
619 SOFTNET_UNLOCK();
620 if (freed || m == NULL) {
621 m = NULL;
622 goto out;
623 }
624 ip = mtod(m, struct ip *);
625 hlen = ip->ip_hl << 2;
626
627 /*
628 * XXX The setting of "srcrt" here is to prevent ip_forward()
629 * from generating ICMP redirects for packets that have
630 * been redirected by a hook back out on to the same LAN that
631 * they came from and is not an indication that the packet
632 * is being inffluenced by source routing options. This
633 * allows things like
634 * "rdr tlp0 0/0 port 80 -> 1.1.1.200 3128 tcp"
635 * where tlp0 is both on the 1.1.1.0/24 network and is the
636 * default route for hosts on 1.1.1.0/24. Of course this
637 * also requires a "map tlp0 ..." to complete the story.
638 * One might argue whether or not this kind of network config.
639 * should be supported in this manner...
640 */
641 srcrt = (odst.s_addr != ip->ip_dst.s_addr);
642 }
643
644 #ifdef ALTQ
645 /* XXX Temporary until ALTQ is changed to use a pfil hook */
646 if (altq_input) {
647 SOFTNET_LOCK();
648 if ((*altq_input)(m, AF_INET) == 0) {
649 /* Packet dropped by traffic conditioner. */
650 SOFTNET_UNLOCK();
651 m = NULL;
652 goto out;
653 }
654 SOFTNET_UNLOCK();
655 }
656 #endif
657
658 /*
659 * Process options and, if not destined for us,
660 * ship it on. ip_dooptions returns 1 when an
661 * error was detected (causing an icmp message
662 * to be sent and the original packet to be freed).
663 */
664 ip_nhops = 0; /* for source routed packets */
665 if (hlen > sizeof (struct ip) && ip_dooptions(m)) {
666 m = NULL;
667 goto out;
668 }
669
670 /*
671 * Check our list of addresses, to see if the packet is for us.
672 *
673 * Traditional 4.4BSD did not consult IFF_UP at all.
674 * The behavior here is to treat addresses on !IFF_UP interface
675 * or IN_IFF_NOTREADY addresses as not mine.
676 */
677 downmatch = 0;
678 s = pserialize_read_enter();
679 ia = ip_match_our_address(ifp, ip, &downmatch);
680 if (ia != NULL) {
681 pserialize_read_exit(s);
682 goto ours;
683 }
684
685 if (ifp->if_flags & IFF_BROADCAST) {
686 ia = ip_match_our_address_broadcast(ifp, ip);
687 if (ia != NULL) {
688 pserialize_read_exit(s);
689 goto ours;
690 }
691 }
692 pserialize_read_exit(s);
693
694 if (IN_MULTICAST(ip->ip_dst.s_addr)) {
695 #ifdef MROUTING
696 extern struct socket *ip_mrouter;
697
698 if (ip_mrouter) {
699 /*
700 * If we are acting as a multicast router, all
701 * incoming multicast packets are passed to the
702 * kernel-level multicast forwarding function.
703 * The packet is returned (relatively) intact; if
704 * ip_mforward() returns a non-zero value, the packet
705 * must be discarded, else it may be accepted below.
706 *
707 * (The IP ident field is put in the same byte order
708 * as expected when ip_mforward() is called from
709 * ip_output().)
710 */
711 SOFTNET_LOCK();
712 if (ip_mforward(m, ifp) != 0) {
713 SOFTNET_UNLOCK();
714 IP_STATINC(IP_STAT_CANTFORWARD);
715 goto out;
716 }
717 SOFTNET_UNLOCK();
718
719 /*
720 * The process-level routing demon needs to receive
721 * all multicast IGMP packets, whether or not this
722 * host belongs to their destination groups.
723 */
724 if (ip->ip_p == IPPROTO_IGMP) {
725 goto ours;
726 }
727 IP_STATINC(IP_STAT_CANTFORWARD);
728 }
729 #endif
730 /*
731 * See if we belong to the destination multicast group on the
732 * arrival interface.
733 */
734 if (!in_multi_group(ip->ip_dst, ifp, 0)) {
735 IP_STATINC(IP_STAT_CANTFORWARD);
736 goto out;
737 }
738 goto ours;
739 }
740 if (ip->ip_dst.s_addr == INADDR_BROADCAST ||
741 in_nullhost(ip->ip_dst))
742 goto ours;
743
744 /*
745 * Not for us; forward if possible and desirable.
746 */
747 if (ipforwarding == 0) {
748 m_put_rcvif_psref(ifp, &psref);
749 IP_STATINC(IP_STAT_CANTFORWARD);
750 m_freem(m);
751 } else {
752 /*
753 * If ip_dst matched any of my address on !IFF_UP interface,
754 * and there's no IFF_UP interface that matches ip_dst,
755 * send icmp unreach. Forwarding it will result in in-kernel
756 * forwarding loop till TTL goes to 0.
757 */
758 if (downmatch) {
759 m_put_rcvif_psref(ifp, &psref);
760 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
761 IP_STATINC(IP_STAT_CANTFORWARD);
762 return;
763 }
764 #ifdef IPSEC
765 /* Perform IPsec, if any. */
766 if (ipsec_used) {
767 SOFTNET_LOCK();
768 if (ipsec4_input(m, IP_FORWARDING |
769 (ip_directedbcast ? IP_ALLOWBROADCAST : 0)) != 0) {
770 SOFTNET_UNLOCK();
771 goto out;
772 }
773 SOFTNET_UNLOCK();
774 }
775 #endif
776 ip_forward(m, srcrt, ifp);
777 m_put_rcvif_psref(ifp, &psref);
778 }
779 return;
780
781 ours:
782 m_put_rcvif_psref(ifp, &psref);
783 ifp = NULL;
784
785 /*
786 * If offset or IP_MF are set, must reassemble.
787 */
788 if (ip->ip_off & ~htons(IP_DF|IP_RF)) {
789 /*
790 * Pass to IP reassembly mechanism.
791 */
792 if (ip_reass_packet(&m, ip) != 0) {
793 /* Failed; invalid fragment(s) or packet. */
794 goto out;
795 }
796 if (m == NULL) {
797 /* More fragments should come; silently return. */
798 goto out;
799 }
800 /*
801 * Reassembly is done, we have the final packet.
802 * Updated cached data in local variable(s).
803 */
804 ip = mtod(m, struct ip *);
805 hlen = ip->ip_hl << 2;
806 }
807
808 #ifdef IPSEC
809 /*
810 * Enforce IPsec policy checking if we are seeing last header.
811 * Note that we do not visit this with protocols with PCB layer
812 * code - like UDP/TCP/raw IP.
813 */
814 if (ipsec_used &&
815 (inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) {
816 SOFTNET_LOCK();
817 if (ipsec4_input(m, 0) != 0) {
818 SOFTNET_UNLOCK();
819 goto out;
820 }
821 SOFTNET_UNLOCK();
822 }
823 #endif
824
825 /*
826 * Switch out to protocol's input routine.
827 */
828 #if IFA_STATS
829 if (ia && ip) {
830 struct in_ifaddr *_ia;
831 /*
832 * Keep a reference from ip_match_our_address with psref
833 * is expensive, so explore ia here again.
834 */
835 s = pserialize_read_enter();
836 _ia = in_get_ia(ip->ip_dst);
837 _ia->ia_ifa.ifa_data.ifad_inbytes += ntohs(ip->ip_len);
838 pserialize_read_exit(s);
839 }
840 #endif
841 IP_STATINC(IP_STAT_DELIVERED);
842
843 const int off = hlen, nh = ip->ip_p;
844
845 SOFTNET_LOCK();
846 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh);
847 SOFTNET_UNLOCK();
848 return;
849
850 out:
851 m_put_rcvif_psref(ifp, &psref);
852 if (m != NULL)
853 m_freem(m);
854 }
855
856 /*
857 * IP timer processing.
858 */
859 void
860 ip_slowtimo(void)
861 {
862
863 mutex_enter(softnet_lock);
864 KERNEL_LOCK(1, NULL);
865
866 ip_reass_slowtimo();
867
868 KERNEL_UNLOCK_ONE(NULL);
869 mutex_exit(softnet_lock);
870 }
871
872 /*
873 * IP drain processing.
874 */
875 void
876 ip_drain(void)
877 {
878
879 KERNEL_LOCK(1, NULL);
880 ip_reass_drain();
881 KERNEL_UNLOCK_ONE(NULL);
882 }
883
884 /*
885 * ip_dooptions: perform option processing on a datagram, possibly discarding
886 * it if bad options are encountered, or forwarding it if source-routed.
887 *
888 * => Returns true if packet has been forwarded/freed.
889 * => Returns false if the packet should be processed further.
890 */
891 static bool
892 ip_dooptions(struct mbuf *m)
893 {
894 struct ip *ip = mtod(m, struct ip *);
895 u_char *cp, *cp0;
896 struct ip_timestamp *ipt;
897 struct in_ifaddr *ia;
898 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0;
899 struct in_addr dst;
900 n_time ntime;
901 struct ifaddr *ifa;
902 int s;
903
904 dst = ip->ip_dst;
905 cp = (u_char *)(ip + 1);
906 cnt = (ip->ip_hl << 2) - sizeof (struct ip);
907 for (; cnt > 0; cnt -= optlen, cp += optlen) {
908 opt = cp[IPOPT_OPTVAL];
909 if (opt == IPOPT_EOL)
910 break;
911 if (opt == IPOPT_NOP)
912 optlen = 1;
913 else {
914 if (cnt < IPOPT_OLEN + sizeof(*cp)) {
915 code = &cp[IPOPT_OLEN] - (u_char *)ip;
916 goto bad;
917 }
918 optlen = cp[IPOPT_OLEN];
919 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
920 code = &cp[IPOPT_OLEN] - (u_char *)ip;
921 goto bad;
922 }
923 }
924 switch (opt) {
925
926 default:
927 break;
928
929 /*
930 * Source routing with record.
931 * Find interface with current destination address.
932 * If none on this machine then drop if strictly routed,
933 * or do nothing if loosely routed.
934 * Record interface address and bring up next address
935 * component. If strictly routed make sure next
936 * address is on directly accessible net.
937 */
938 case IPOPT_LSRR:
939 case IPOPT_SSRR:
940 if (ip_allowsrcrt == 0) {
941 type = ICMP_UNREACH;
942 code = ICMP_UNREACH_NET_PROHIB;
943 goto bad;
944 }
945 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
946 code = &cp[IPOPT_OLEN] - (u_char *)ip;
947 goto bad;
948 }
949 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
950 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
951 goto bad;
952 }
953 ipaddr.sin_addr = ip->ip_dst;
954
955 s = pserialize_read_enter();
956 ifa = ifa_ifwithaddr(sintosa(&ipaddr));
957 if (ifa == NULL) {
958 pserialize_read_exit(s);
959 if (opt == IPOPT_SSRR) {
960 type = ICMP_UNREACH;
961 code = ICMP_UNREACH_SRCFAIL;
962 goto bad;
963 }
964 /*
965 * Loose routing, and not at next destination
966 * yet; nothing to do except forward.
967 */
968 break;
969 }
970 pserialize_read_exit(s);
971
972 off--; /* 0 origin */
973 if ((off + sizeof(struct in_addr)) > optlen) {
974 /*
975 * End of source route. Should be for us.
976 */
977 save_rte(cp, ip->ip_src);
978 break;
979 }
980 /*
981 * locate outgoing interface
982 */
983 memcpy((void *)&ipaddr.sin_addr, (void *)(cp + off),
984 sizeof(ipaddr.sin_addr));
985 s = pserialize_read_enter();
986 if (opt == IPOPT_SSRR) {
987 ifa = ifa_ifwithladdr(sintosa(&ipaddr));
988 if (ifa != NULL)
989 ia = ifatoia(ifa);
990 else
991 ia = NULL;
992 } else {
993 ia = ip_rtaddr(ipaddr.sin_addr);
994 }
995 if (ia == NULL) {
996 type = ICMP_UNREACH;
997 code = ICMP_UNREACH_SRCFAIL;
998 pserialize_read_exit(s);
999 goto bad;
1000 }
1001 ip->ip_dst = ipaddr.sin_addr;
1002 bcopy((void *)&ia->ia_addr.sin_addr,
1003 (void *)(cp + off), sizeof(struct in_addr));
1004 pserialize_read_exit(s);
1005 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1006 /*
1007 * Let ip_intr's mcast routing check handle mcast pkts
1008 */
1009 forward = !IN_MULTICAST(ip->ip_dst.s_addr);
1010 break;
1011
1012 case IPOPT_RR:
1013 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1014 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1015 goto bad;
1016 }
1017 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1018 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1019 goto bad;
1020 }
1021 /*
1022 * If no space remains, ignore.
1023 */
1024 off--; /* 0 origin */
1025 if ((off + sizeof(struct in_addr)) > optlen)
1026 break;
1027 memcpy((void *)&ipaddr.sin_addr, (void *)(&ip->ip_dst),
1028 sizeof(ipaddr.sin_addr));
1029 /*
1030 * locate outgoing interface; if we're the destination,
1031 * use the incoming interface (should be same).
1032 */
1033 s = pserialize_read_enter();
1034 ifa = ifa_ifwithaddr(sintosa(&ipaddr));
1035 if (ifa == NULL) {
1036 ia = ip_rtaddr(ipaddr.sin_addr);
1037 if (ia == NULL) {
1038 pserialize_read_exit(s);
1039 type = ICMP_UNREACH;
1040 code = ICMP_UNREACH_HOST;
1041 goto bad;
1042 }
1043 } else {
1044 ia = ifatoia(ifa);
1045 }
1046 bcopy((void *)&ia->ia_addr.sin_addr,
1047 (void *)(cp + off), sizeof(struct in_addr));
1048 pserialize_read_exit(s);
1049 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1050 break;
1051
1052 case IPOPT_TS:
1053 code = cp - (u_char *)ip;
1054 ipt = (struct ip_timestamp *)cp;
1055 if (ipt->ipt_len < 4 || ipt->ipt_len > 40) {
1056 code = (u_char *)&ipt->ipt_len - (u_char *)ip;
1057 goto bad;
1058 }
1059 if (ipt->ipt_ptr < 5) {
1060 code = (u_char *)&ipt->ipt_ptr - (u_char *)ip;
1061 goto bad;
1062 }
1063 if (ipt->ipt_ptr > ipt->ipt_len - sizeof (int32_t)) {
1064 if (++ipt->ipt_oflw == 0) {
1065 code = (u_char *)&ipt->ipt_ptr -
1066 (u_char *)ip;
1067 goto bad;
1068 }
1069 break;
1070 }
1071 cp0 = (cp + ipt->ipt_ptr - 1);
1072 switch (ipt->ipt_flg) {
1073
1074 case IPOPT_TS_TSONLY:
1075 break;
1076
1077 case IPOPT_TS_TSANDADDR: {
1078 struct ifnet *rcvif;
1079 int _s, _ss;
1080
1081 if (ipt->ipt_ptr - 1 + sizeof(n_time) +
1082 sizeof(struct in_addr) > ipt->ipt_len) {
1083 code = (u_char *)&ipt->ipt_ptr -
1084 (u_char *)ip;
1085 goto bad;
1086 }
1087 ipaddr.sin_addr = dst;
1088 _ss = pserialize_read_enter();
1089 rcvif = m_get_rcvif(m, &_s);
1090 ifa = ifaof_ifpforaddr(sintosa(&ipaddr), rcvif);
1091 m_put_rcvif(rcvif, &_s);
1092 if (ifa == NULL) {
1093 pserialize_read_exit(_ss);
1094 break;
1095 }
1096 ia = ifatoia(ifa);
1097 bcopy(&ia->ia_addr.sin_addr,
1098 cp0, sizeof(struct in_addr));
1099 pserialize_read_exit(_ss);
1100 ipt->ipt_ptr += sizeof(struct in_addr);
1101 break;
1102 }
1103
1104 case IPOPT_TS_PRESPEC:
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 default:
1124 /* XXX can't take &ipt->ipt_flg */
1125 code = (u_char *)&ipt->ipt_ptr -
1126 (u_char *)ip + 1;
1127 goto bad;
1128 }
1129 ntime = iptime();
1130 cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */
1131 memmove((char *)cp + ipt->ipt_ptr - 1, cp0,
1132 sizeof(n_time));
1133 ipt->ipt_ptr += sizeof(n_time);
1134 }
1135 }
1136 if (forward) {
1137 struct ifnet *rcvif;
1138 struct psref _psref;
1139
1140 if (ip_forwsrcrt == 0) {
1141 type = ICMP_UNREACH;
1142 code = ICMP_UNREACH_SRCFAIL;
1143 goto bad;
1144 }
1145
1146 rcvif = m_get_rcvif_psref(m, &_psref);
1147 if (__predict_false(rcvif == NULL)) {
1148 type = ICMP_UNREACH;
1149 code = ICMP_UNREACH_HOST;
1150 goto bad;
1151 }
1152 ip_forward(m, 1, rcvif);
1153 m_put_rcvif_psref(rcvif, &_psref);
1154 return true;
1155 }
1156 return false;
1157 bad:
1158 icmp_error(m, type, code, 0, 0);
1159 IP_STATINC(IP_STAT_BADOPTIONS);
1160 return true;
1161 }
1162
1163 /*
1164 * ip_rtaddr: given address of next destination (final or next hop),
1165 * return internet address info of interface to be used to get there.
1166 */
1167 static struct in_ifaddr *
1168 ip_rtaddr(struct in_addr dst)
1169 {
1170 struct rtentry *rt;
1171 union {
1172 struct sockaddr dst;
1173 struct sockaddr_in dst4;
1174 } u;
1175 struct route *ro;
1176
1177 sockaddr_in_init(&u.dst4, &dst, 0);
1178
1179 SOFTNET_LOCK();
1180 ro = percpu_getref(ipforward_rt_percpu);
1181 rt = rtcache_lookup(ro, &u.dst);
1182 percpu_putref(ipforward_rt_percpu);
1183 SOFTNET_UNLOCK();
1184 if (rt == NULL)
1185 return NULL;
1186
1187 return ifatoia(rt->rt_ifa);
1188 }
1189
1190 /*
1191 * save_rte: save incoming source route for use in replies, to be picked
1192 * up later by ip_srcroute if the receiver is interested.
1193 */
1194 static void
1195 save_rte(u_char *option, struct in_addr dst)
1196 {
1197 unsigned olen;
1198
1199 olen = option[IPOPT_OLEN];
1200 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst)))
1201 return;
1202 memcpy((void *)ip_srcrt.srcopt, (void *)option, olen);
1203 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1204 ip_srcrt.dst = dst;
1205 }
1206
1207 /*
1208 * Retrieve incoming source route for use in replies,
1209 * in the same form used by setsockopt.
1210 * The first hop is placed before the options, will be removed later.
1211 */
1212 struct mbuf *
1213 ip_srcroute(void)
1214 {
1215 struct in_addr *p, *q;
1216 struct mbuf *m;
1217
1218 if (ip_nhops == 0)
1219 return NULL;
1220 m = m_get(M_DONTWAIT, MT_SOOPTS);
1221 if (m == 0)
1222 return NULL;
1223
1224 MCLAIM(m, &inetdomain.dom_mowner);
1225 #define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt))
1226
1227 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
1228 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) +
1229 OPTSIZ;
1230
1231 /*
1232 * First save first hop for return route
1233 */
1234 p = &ip_srcrt.route[ip_nhops - 1];
1235 *(mtod(m, struct in_addr *)) = *p--;
1236
1237 /*
1238 * Copy option fields and padding (nop) to mbuf.
1239 */
1240 ip_srcrt.nop = IPOPT_NOP;
1241 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
1242 memmove(mtod(m, char *) + sizeof(struct in_addr), &ip_srcrt.nop,
1243 OPTSIZ);
1244 q = (struct in_addr *)(mtod(m, char *) +
1245 sizeof(struct in_addr) + OPTSIZ);
1246 #undef OPTSIZ
1247 /*
1248 * Record return path as an IP source route,
1249 * reversing the path (pointers are now aligned).
1250 */
1251 while (p >= ip_srcrt.route) {
1252 *q++ = *p--;
1253 }
1254 /*
1255 * Last hop goes to final destination.
1256 */
1257 *q = ip_srcrt.dst;
1258 return (m);
1259 }
1260
1261 const int inetctlerrmap[PRC_NCMDS] = {
1262 [PRC_MSGSIZE] = EMSGSIZE,
1263 [PRC_HOSTDEAD] = EHOSTDOWN,
1264 [PRC_HOSTUNREACH] = EHOSTUNREACH,
1265 [PRC_UNREACH_NET] = EHOSTUNREACH,
1266 [PRC_UNREACH_HOST] = EHOSTUNREACH,
1267 [PRC_UNREACH_PROTOCOL] = ECONNREFUSED,
1268 [PRC_UNREACH_PORT] = ECONNREFUSED,
1269 [PRC_UNREACH_SRCFAIL] = EHOSTUNREACH,
1270 [PRC_PARAMPROB] = ENOPROTOOPT,
1271 };
1272
1273 void
1274 ip_fasttimo(void)
1275 {
1276 if (ip_drainwanted) {
1277 ip_drain();
1278 ip_drainwanted = 0;
1279 }
1280 }
1281
1282 void
1283 ip_drainstub(void)
1284 {
1285 ip_drainwanted = 1;
1286 }
1287
1288 /*
1289 * Forward a packet. If some error occurs return the sender
1290 * an icmp packet. Note we can't always generate a meaningful
1291 * icmp message because icmp doesn't have a large enough repertoire
1292 * of codes and types.
1293 *
1294 * If not forwarding, just drop the packet. This could be confusing
1295 * if ipforwarding was zero but some routing protocol was advancing
1296 * us as a gateway to somewhere. However, we must let the routing
1297 * protocol deal with that.
1298 *
1299 * The srcrt parameter indicates whether the packet is being forwarded
1300 * via a source route.
1301 */
1302 static void
1303 ip_forward(struct mbuf *m, int srcrt, struct ifnet *rcvif)
1304 {
1305 struct ip *ip = mtod(m, struct ip *);
1306 struct rtentry *rt;
1307 int error, type = 0, code = 0, destmtu = 0;
1308 struct mbuf *mcopy;
1309 n_long dest;
1310 union {
1311 struct sockaddr dst;
1312 struct sockaddr_in dst4;
1313 } u;
1314 uint64_t *ips;
1315 struct route *ro;
1316
1317 KASSERTMSG(cpu_softintr_p(), "ip_forward: not in the software "
1318 "interrupt handler; synchronization assumptions violated");
1319
1320 /*
1321 * We are now in the output path.
1322 */
1323 MCLAIM(m, &ip_tx_mowner);
1324
1325 /*
1326 * Clear any in-bound checksum flags for this packet.
1327 */
1328 m->m_pkthdr.csum_flags = 0;
1329
1330 dest = 0;
1331 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
1332 IP_STATINC(IP_STAT_CANTFORWARD);
1333 m_freem(m);
1334 return;
1335 }
1336
1337 SOFTNET_LOCK();
1338
1339 if (ip->ip_ttl <= IPTTLDEC) {
1340 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0);
1341 SOFTNET_UNLOCK();
1342 return;
1343 }
1344
1345 sockaddr_in_init(&u.dst4, &ip->ip_dst, 0);
1346
1347 ro = percpu_getref(ipforward_rt_percpu);
1348 if ((rt = rtcache_lookup(ro, &u.dst)) == NULL) {
1349 percpu_putref(ipforward_rt_percpu);
1350 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_NET, dest, 0);
1351 SOFTNET_UNLOCK();
1352 return;
1353 }
1354
1355 /*
1356 * Save at most 68 bytes of the packet in case
1357 * we need to generate an ICMP message to the src.
1358 * Pullup to avoid sharing mbuf cluster between m and mcopy.
1359 */
1360 mcopy = m_copym(m, 0, imin(ntohs(ip->ip_len), 68), M_DONTWAIT);
1361 if (mcopy)
1362 mcopy = m_pullup(mcopy, ip->ip_hl << 2);
1363
1364 ip->ip_ttl -= IPTTLDEC;
1365
1366 /*
1367 * If forwarding packet using same interface that it came in on,
1368 * perhaps should send a redirect to sender to shortcut a hop.
1369 * Only send redirect if source is sending directly to us,
1370 * and if packet was not source routed (or has any options).
1371 * Also, don't send redirect if forwarding using a default route
1372 * or a route modified by a redirect.
1373 */
1374 if (rt->rt_ifp == rcvif &&
1375 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1376 !in_nullhost(satocsin(rt_getkey(rt))->sin_addr) &&
1377 ipsendredirects && !srcrt) {
1378 if (rt->rt_ifa &&
1379 (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) ==
1380 ifatoia(rt->rt_ifa)->ia_subnet) {
1381 if (rt->rt_flags & RTF_GATEWAY)
1382 dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
1383 else
1384 dest = ip->ip_dst.s_addr;
1385 /*
1386 * Router requirements says to only send host
1387 * redirects.
1388 */
1389 type = ICMP_REDIRECT;
1390 code = ICMP_REDIRECT_HOST;
1391 }
1392 }
1393
1394 error = ip_output(m, NULL, ro,
1395 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)),
1396 NULL, NULL);
1397
1398 if (error) {
1399 IP_STATINC(IP_STAT_CANTFORWARD);
1400 goto error;
1401 }
1402
1403 ips = IP_STAT_GETREF();
1404 ips[IP_STAT_FORWARD]++;
1405
1406 if (type) {
1407 ips[IP_STAT_REDIRECTSENT]++;
1408 IP_STAT_PUTREF();
1409 goto redirect;
1410 }
1411
1412 IP_STAT_PUTREF();
1413 if (mcopy) {
1414 #ifdef GATEWAY
1415 if (mcopy->m_flags & M_CANFASTFWD)
1416 ipflow_create(ro, mcopy);
1417 #endif
1418 m_freem(mcopy);
1419 }
1420
1421 percpu_putref(ipforward_rt_percpu);
1422 SOFTNET_UNLOCK();
1423 return;
1424
1425 redirect:
1426 error:
1427 if (mcopy == NULL) {
1428 percpu_putref(ipforward_rt_percpu);
1429 SOFTNET_UNLOCK();
1430 return;
1431 }
1432
1433 switch (error) {
1434
1435 case 0: /* forwarded, but need redirect */
1436 /* type, code set above */
1437 break;
1438
1439 case ENETUNREACH: /* shouldn't happen, checked above */
1440 case EHOSTUNREACH:
1441 case ENETDOWN:
1442 case EHOSTDOWN:
1443 default:
1444 type = ICMP_UNREACH;
1445 code = ICMP_UNREACH_HOST;
1446 break;
1447
1448 case EMSGSIZE:
1449 type = ICMP_UNREACH;
1450 code = ICMP_UNREACH_NEEDFRAG;
1451
1452 if ((rt = rtcache_validate(ro)) != NULL)
1453 destmtu = rt->rt_ifp->if_mtu;
1454 #ifdef IPSEC
1455 if (ipsec_used)
1456 (void)ipsec4_forward(mcopy, &destmtu);
1457 #endif
1458 IP_STATINC(IP_STAT_CANTFRAG);
1459 break;
1460
1461 case ENOBUFS:
1462 /*
1463 * Do not generate ICMP_SOURCEQUENCH as required in RFC 1812,
1464 * Requirements for IP Version 4 Routers. Source quench can
1465 * big problem under DoS attacks or if the underlying
1466 * interface is rate-limited.
1467 */
1468 if (mcopy)
1469 m_freem(mcopy);
1470 percpu_putref(ipforward_rt_percpu);
1471 SOFTNET_UNLOCK();
1472 return;
1473 }
1474 icmp_error(mcopy, type, code, dest, destmtu);
1475 percpu_putref(ipforward_rt_percpu);
1476 SOFTNET_UNLOCK();
1477 }
1478
1479 void
1480 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1481 struct mbuf *m)
1482 {
1483 struct socket *so = inp->inp_socket;
1484 ifnet_t *ifp;
1485 int inpflags = inp->inp_flags;
1486 struct psref psref;
1487
1488 ifp = m_get_rcvif_psref(m, &psref);
1489 if (__predict_false(ifp == NULL))
1490 return; /* XXX should report error? */
1491
1492 if (so->so_options & SO_TIMESTAMP
1493 #ifdef SO_OTIMESTAMP
1494 || so->so_options & SO_OTIMESTAMP
1495 #endif
1496 ) {
1497 struct timeval tv;
1498
1499 microtime(&tv);
1500 #ifdef SO_OTIMESTAMP
1501 if (so->so_options & SO_OTIMESTAMP) {
1502 struct timeval50 tv50;
1503 timeval_to_timeval50(&tv, &tv50);
1504 *mp = sbcreatecontrol((void *) &tv50, sizeof(tv50),
1505 SCM_OTIMESTAMP, SOL_SOCKET);
1506 } else
1507 #endif
1508 *mp = sbcreatecontrol((void *) &tv, sizeof(tv),
1509 SCM_TIMESTAMP, SOL_SOCKET);
1510 if (*mp)
1511 mp = &(*mp)->m_next;
1512 }
1513 if (inpflags & INP_RECVDSTADDR) {
1514 *mp = sbcreatecontrol((void *) &ip->ip_dst,
1515 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1516 if (*mp)
1517 mp = &(*mp)->m_next;
1518 }
1519 if (inpflags & INP_RECVPKTINFO) {
1520 struct in_pktinfo ipi;
1521 ipi.ipi_addr = ip->ip_src;
1522 ipi.ipi_ifindex = ifp->if_index;
1523 *mp = sbcreatecontrol((void *) &ipi,
1524 sizeof(ipi), IP_RECVPKTINFO, IPPROTO_IP);
1525 if (*mp)
1526 mp = &(*mp)->m_next;
1527 }
1528 if (inpflags & INP_PKTINFO) {
1529 struct in_pktinfo ipi;
1530 ipi.ipi_addr = ip->ip_dst;
1531 ipi.ipi_ifindex = ifp->if_index;
1532 *mp = sbcreatecontrol((void *) &ipi,
1533 sizeof(ipi), IP_PKTINFO, IPPROTO_IP);
1534 if (*mp)
1535 mp = &(*mp)->m_next;
1536 }
1537 if (inpflags & INP_RECVIF) {
1538 struct sockaddr_dl sdl;
1539
1540 sockaddr_dl_init(&sdl, sizeof(sdl), ifp->if_index, 0, NULL, 0,
1541 NULL, 0);
1542 *mp = sbcreatecontrol(&sdl, sdl.sdl_len, IP_RECVIF, IPPROTO_IP);
1543 if (*mp)
1544 mp = &(*mp)->m_next;
1545 }
1546 if (inpflags & INP_RECVTTL) {
1547 *mp = sbcreatecontrol((void *) &ip->ip_ttl,
1548 sizeof(uint8_t), IP_RECVTTL, IPPROTO_IP);
1549 if (*mp)
1550 mp = &(*mp)->m_next;
1551 }
1552 m_put_rcvif_psref(ifp, &psref);
1553 }
1554
1555 /*
1556 * sysctl helper routine for net.inet.ip.forwsrcrt.
1557 */
1558 static int
1559 sysctl_net_inet_ip_forwsrcrt(SYSCTLFN_ARGS)
1560 {
1561 int error, tmp;
1562 struct sysctlnode node;
1563
1564 node = *rnode;
1565 tmp = ip_forwsrcrt;
1566 node.sysctl_data = &tmp;
1567 error = sysctl_lookup(SYSCTLFN_CALL(&node));
1568 if (error || newp == NULL)
1569 return (error);
1570
1571 error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_FORWSRCRT,
1572 0, NULL, NULL, NULL);
1573 if (error)
1574 return (error);
1575
1576 ip_forwsrcrt = tmp;
1577
1578 return (0);
1579 }
1580
1581 /*
1582 * sysctl helper routine for net.inet.ip.mtudisctimeout. checks the
1583 * range of the new value and tweaks timers if it changes.
1584 */
1585 static int
1586 sysctl_net_inet_ip_pmtudto(SYSCTLFN_ARGS)
1587 {
1588 int error, tmp;
1589 struct sysctlnode node;
1590
1591 node = *rnode;
1592 tmp = ip_mtudisc_timeout;
1593 node.sysctl_data = &tmp;
1594 error = sysctl_lookup(SYSCTLFN_CALL(&node));
1595 if (error || newp == NULL)
1596 return (error);
1597 if (tmp < 0)
1598 return (EINVAL);
1599
1600 mutex_enter(softnet_lock);
1601
1602 ip_mtudisc_timeout = tmp;
1603 rt_timer_queue_change(ip_mtudisc_timeout_q, ip_mtudisc_timeout);
1604
1605 mutex_exit(softnet_lock);
1606
1607 return (0);
1608 }
1609
1610 static int
1611 sysctl_net_inet_ip_stats(SYSCTLFN_ARGS)
1612 {
1613
1614 return (NETSTAT_SYSCTL(ipstat_percpu, IP_NSTATS));
1615 }
1616
1617 static void
1618 sysctl_net_inet_ip_setup(struct sysctllog **clog)
1619 {
1620 sysctl_createv(clog, 0, NULL, NULL,
1621 CTLFLAG_PERMANENT,
1622 CTLTYPE_NODE, "inet",
1623 SYSCTL_DESCR("PF_INET related settings"),
1624 NULL, 0, NULL, 0,
1625 CTL_NET, PF_INET, CTL_EOL);
1626 sysctl_createv(clog, 0, NULL, NULL,
1627 CTLFLAG_PERMANENT,
1628 CTLTYPE_NODE, "ip",
1629 SYSCTL_DESCR("IPv4 related settings"),
1630 NULL, 0, NULL, 0,
1631 CTL_NET, PF_INET, IPPROTO_IP, CTL_EOL);
1632
1633 sysctl_createv(clog, 0, NULL, NULL,
1634 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1635 CTLTYPE_INT, "forwarding",
1636 SYSCTL_DESCR("Enable forwarding of INET datagrams"),
1637 NULL, 0, &ipforwarding, 0,
1638 CTL_NET, PF_INET, IPPROTO_IP,
1639 IPCTL_FORWARDING, CTL_EOL);
1640 sysctl_createv(clog, 0, NULL, NULL,
1641 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1642 CTLTYPE_INT, "redirect",
1643 SYSCTL_DESCR("Enable sending of ICMP redirect messages"),
1644 NULL, 0, &ipsendredirects, 0,
1645 CTL_NET, PF_INET, IPPROTO_IP,
1646 IPCTL_SENDREDIRECTS, CTL_EOL);
1647 sysctl_createv(clog, 0, NULL, NULL,
1648 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1649 CTLTYPE_INT, "ttl",
1650 SYSCTL_DESCR("Default TTL for an INET datagram"),
1651 NULL, 0, &ip_defttl, 0,
1652 CTL_NET, PF_INET, IPPROTO_IP,
1653 IPCTL_DEFTTL, CTL_EOL);
1654 #ifdef IPCTL_DEFMTU
1655 sysctl_createv(clog, 0, NULL, NULL,
1656 CTLFLAG_PERMANENT /* |CTLFLAG_READWRITE? */,
1657 CTLTYPE_INT, "mtu",
1658 SYSCTL_DESCR("Default MTA for an INET route"),
1659 NULL, 0, &ip_mtu, 0,
1660 CTL_NET, PF_INET, IPPROTO_IP,
1661 IPCTL_DEFMTU, CTL_EOL);
1662 #endif /* IPCTL_DEFMTU */
1663 sysctl_createv(clog, 0, NULL, NULL,
1664 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1665 CTLTYPE_INT, "forwsrcrt",
1666 SYSCTL_DESCR("Enable forwarding of source-routed "
1667 "datagrams"),
1668 sysctl_net_inet_ip_forwsrcrt, 0, &ip_forwsrcrt, 0,
1669 CTL_NET, PF_INET, IPPROTO_IP,
1670 IPCTL_FORWSRCRT, CTL_EOL);
1671 sysctl_createv(clog, 0, NULL, NULL,
1672 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1673 CTLTYPE_INT, "directed-broadcast",
1674 SYSCTL_DESCR("Enable forwarding of broadcast datagrams"),
1675 NULL, 0, &ip_directedbcast, 0,
1676 CTL_NET, PF_INET, IPPROTO_IP,
1677 IPCTL_DIRECTEDBCAST, CTL_EOL);
1678 sysctl_createv(clog, 0, NULL, NULL,
1679 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1680 CTLTYPE_INT, "allowsrcrt",
1681 SYSCTL_DESCR("Accept source-routed datagrams"),
1682 NULL, 0, &ip_allowsrcrt, 0,
1683 CTL_NET, PF_INET, IPPROTO_IP,
1684 IPCTL_ALLOWSRCRT, CTL_EOL);
1685
1686 sysctl_createv(clog, 0, NULL, NULL,
1687 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1688 CTLTYPE_INT, "mtudisc",
1689 SYSCTL_DESCR("Use RFC1191 Path MTU Discovery"),
1690 NULL, 0, &ip_mtudisc, 0,
1691 CTL_NET, PF_INET, IPPROTO_IP,
1692 IPCTL_MTUDISC, CTL_EOL);
1693 sysctl_createv(clog, 0, NULL, NULL,
1694 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1695 CTLTYPE_INT, "anonportmin",
1696 SYSCTL_DESCR("Lowest ephemeral port number to assign"),
1697 sysctl_net_inet_ip_ports, 0, &anonportmin, 0,
1698 CTL_NET, PF_INET, IPPROTO_IP,
1699 IPCTL_ANONPORTMIN, CTL_EOL);
1700 sysctl_createv(clog, 0, NULL, NULL,
1701 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1702 CTLTYPE_INT, "anonportmax",
1703 SYSCTL_DESCR("Highest ephemeral port number to assign"),
1704 sysctl_net_inet_ip_ports, 0, &anonportmax, 0,
1705 CTL_NET, PF_INET, IPPROTO_IP,
1706 IPCTL_ANONPORTMAX, CTL_EOL);
1707 sysctl_createv(clog, 0, NULL, NULL,
1708 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1709 CTLTYPE_INT, "mtudisctimeout",
1710 SYSCTL_DESCR("Lifetime of a Path MTU Discovered route"),
1711 sysctl_net_inet_ip_pmtudto, 0, (void *)&ip_mtudisc_timeout, 0,
1712 CTL_NET, PF_INET, IPPROTO_IP,
1713 IPCTL_MTUDISCTIMEOUT, CTL_EOL);
1714 #ifndef IPNOPRIVPORTS
1715 sysctl_createv(clog, 0, NULL, NULL,
1716 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1717 CTLTYPE_INT, "lowportmin",
1718 SYSCTL_DESCR("Lowest privileged ephemeral port number "
1719 "to assign"),
1720 sysctl_net_inet_ip_ports, 0, &lowportmin, 0,
1721 CTL_NET, PF_INET, IPPROTO_IP,
1722 IPCTL_LOWPORTMIN, CTL_EOL);
1723 sysctl_createv(clog, 0, NULL, NULL,
1724 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1725 CTLTYPE_INT, "lowportmax",
1726 SYSCTL_DESCR("Highest privileged ephemeral port number "
1727 "to assign"),
1728 sysctl_net_inet_ip_ports, 0, &lowportmax, 0,
1729 CTL_NET, PF_INET, IPPROTO_IP,
1730 IPCTL_LOWPORTMAX, CTL_EOL);
1731 #endif /* IPNOPRIVPORTS */
1732 #if NGRE > 0
1733 sysctl_createv(clog, 0, NULL, NULL,
1734 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1735 CTLTYPE_INT, "grettl",
1736 SYSCTL_DESCR("Default TTL for a gre tunnel datagram"),
1737 NULL, 0, &ip_gre_ttl, 0,
1738 CTL_NET, PF_INET, IPPROTO_IP,
1739 IPCTL_GRE_TTL, CTL_EOL);
1740 #endif /* NGRE */
1741 sysctl_createv(clog, 0, NULL, NULL,
1742 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1743 CTLTYPE_INT, "checkinterface",
1744 SYSCTL_DESCR("Enable receive side of Strong ES model "
1745 "from RFC1122"),
1746 NULL, 0, &ip_checkinterface, 0,
1747 CTL_NET, PF_INET, IPPROTO_IP,
1748 IPCTL_CHECKINTERFACE, CTL_EOL);
1749 sysctl_createv(clog, 0, NULL, NULL,
1750 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1751 CTLTYPE_INT, "random_id",
1752 SYSCTL_DESCR("Assign random ip_id values"),
1753 NULL, 0, &ip_do_randomid, 0,
1754 CTL_NET, PF_INET, IPPROTO_IP,
1755 IPCTL_RANDOMID, CTL_EOL);
1756 sysctl_createv(clog, 0, NULL, NULL,
1757 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1758 CTLTYPE_INT, "do_loopback_cksum",
1759 SYSCTL_DESCR("Perform IP checksum on loopback"),
1760 NULL, 0, &ip_do_loopback_cksum, 0,
1761 CTL_NET, PF_INET, IPPROTO_IP,
1762 IPCTL_LOOPBACKCKSUM, CTL_EOL);
1763 sysctl_createv(clog, 0, NULL, NULL,
1764 CTLFLAG_PERMANENT,
1765 CTLTYPE_STRUCT, "stats",
1766 SYSCTL_DESCR("IP statistics"),
1767 sysctl_net_inet_ip_stats, 0, NULL, 0,
1768 CTL_NET, PF_INET, IPPROTO_IP, IPCTL_STATS,
1769 CTL_EOL);
1770 #if NARP
1771 sysctl_createv(clog, 0, NULL, NULL,
1772 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1773 CTLTYPE_INT, "dad_count",
1774 SYSCTL_DESCR("Number of Duplicate Address Detection "
1775 "probes to send"),
1776 NULL, 0, &ip_dad_count, 0,
1777 CTL_NET, PF_INET, IPPROTO_IP,
1778 IPCTL_DAD_COUNT, CTL_EOL);
1779 #endif
1780
1781 /* anonportalgo RFC6056 subtree */
1782 const struct sysctlnode *portalgo_node;
1783 sysctl_createv(clog, 0, NULL, &portalgo_node,
1784 CTLFLAG_PERMANENT,
1785 CTLTYPE_NODE, "anonportalgo",
1786 SYSCTL_DESCR("Anonymous Port Algorithm Selection (RFC 6056)"),
1787 NULL, 0, NULL, 0,
1788 CTL_NET, PF_INET, IPPROTO_IP, CTL_CREATE, CTL_EOL);
1789 sysctl_createv(clog, 0, &portalgo_node, NULL,
1790 CTLFLAG_PERMANENT,
1791 CTLTYPE_STRING, "available",
1792 SYSCTL_DESCR("available algorithms"),
1793 sysctl_portalgo_available, 0, NULL, PORTALGO_MAXLEN,
1794 CTL_CREATE, CTL_EOL);
1795 sysctl_createv(clog, 0, &portalgo_node, NULL,
1796 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1797 CTLTYPE_STRING, "selected",
1798 SYSCTL_DESCR("selected algorithm"),
1799 sysctl_portalgo_selected4, 0, NULL, PORTALGO_MAXLEN,
1800 CTL_CREATE, CTL_EOL);
1801 sysctl_createv(clog, 0, &portalgo_node, NULL,
1802 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1803 CTLTYPE_STRUCT, "reserve",
1804 SYSCTL_DESCR("bitmap of reserved ports"),
1805 sysctl_portalgo_reserve4, 0, NULL, 0,
1806 CTL_CREATE, CTL_EOL);
1807 }
1808
1809 void
1810 ip_statinc(u_int stat)
1811 {
1812
1813 KASSERT(stat < IP_NSTATS);
1814 IP_STATINC(stat);
1815 }
1816