ip_input.c revision 1.319.4.7 1 /* $NetBSD: ip_input.c,v 1.319.4.7 2016/10/05 20:56:09 skrll 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 1.319.4.7 2016/10/05 20:56:09 skrll 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 goto out;
622 }
623 ip = mtod(m, struct ip *);
624 hlen = ip->ip_hl << 2;
625
626 /*
627 * XXX The setting of "srcrt" here is to prevent ip_forward()
628 * from generating ICMP redirects for packets that have
629 * been redirected by a hook back out on to the same LAN that
630 * they came from and is not an indication that the packet
631 * is being inffluenced by source routing options. This
632 * allows things like
633 * "rdr tlp0 0/0 port 80 -> 1.1.1.200 3128 tcp"
634 * where tlp0 is both on the 1.1.1.0/24 network and is the
635 * default route for hosts on 1.1.1.0/24. Of course this
636 * also requires a "map tlp0 ..." to complete the story.
637 * One might argue whether or not this kind of network config.
638 * should be supported in this manner...
639 */
640 srcrt = (odst.s_addr != ip->ip_dst.s_addr);
641 }
642
643 #ifdef ALTQ
644 /* XXX Temporary until ALTQ is changed to use a pfil hook */
645 if (altq_input) {
646 SOFTNET_LOCK();
647 if ((*altq_input)(m, AF_INET) == 0) {
648 /* Packet dropped by traffic conditioner. */
649 SOFTNET_UNLOCK();
650 goto out;
651 }
652 SOFTNET_UNLOCK();
653 }
654 #endif
655
656 /*
657 * Process options and, if not destined for us,
658 * ship it on. ip_dooptions returns 1 when an
659 * error was detected (causing an icmp message
660 * to be sent and the original packet to be freed).
661 */
662 ip_nhops = 0; /* for source routed packets */
663 if (hlen > sizeof (struct ip) && ip_dooptions(m))
664 goto out;
665
666 /*
667 * Check our list of addresses, to see if the packet is for us.
668 *
669 * Traditional 4.4BSD did not consult IFF_UP at all.
670 * The behavior here is to treat addresses on !IFF_UP interface
671 * or IN_IFF_NOTREADY addresses as not mine.
672 */
673 downmatch = 0;
674 s = pserialize_read_enter();
675 ia = ip_match_our_address(ifp, ip, &downmatch);
676 if (ia != NULL) {
677 pserialize_read_exit(s);
678 goto ours;
679 }
680
681 if (ifp->if_flags & IFF_BROADCAST) {
682 ia = ip_match_our_address_broadcast(ifp, ip);
683 if (ia != NULL) {
684 pserialize_read_exit(s);
685 goto ours;
686 }
687 }
688 pserialize_read_exit(s);
689
690 if (IN_MULTICAST(ip->ip_dst.s_addr)) {
691 #ifdef MROUTING
692 extern struct socket *ip_mrouter;
693
694 if (ip_mrouter) {
695 /*
696 * If we are acting as a multicast router, all
697 * incoming multicast packets are passed to the
698 * kernel-level multicast forwarding function.
699 * The packet is returned (relatively) intact; if
700 * ip_mforward() returns a non-zero value, the packet
701 * must be discarded, else it may be accepted below.
702 *
703 * (The IP ident field is put in the same byte order
704 * as expected when ip_mforward() is called from
705 * ip_output().)
706 */
707 SOFTNET_LOCK();
708 if (ip_mforward(m, ifp) != 0) {
709 SOFTNET_UNLOCK();
710 IP_STATINC(IP_STAT_CANTFORWARD);
711 goto out;
712 }
713 SOFTNET_UNLOCK();
714
715 /*
716 * The process-level routing demon needs to receive
717 * all multicast IGMP packets, whether or not this
718 * host belongs to their destination groups.
719 */
720 if (ip->ip_p == IPPROTO_IGMP) {
721 goto ours;
722 }
723 IP_STATINC(IP_STAT_CANTFORWARD);
724 }
725 #endif
726 /*
727 * See if we belong to the destination multicast group on the
728 * arrival interface.
729 */
730 if (!in_multi_group(ip->ip_dst, ifp, 0)) {
731 IP_STATINC(IP_STAT_CANTFORWARD);
732 goto out;
733 }
734 goto ours;
735 }
736 if (ip->ip_dst.s_addr == INADDR_BROADCAST ||
737 in_nullhost(ip->ip_dst))
738 goto ours;
739
740 /*
741 * Not for us; forward if possible and desirable.
742 */
743 if (ipforwarding == 0) {
744 m_put_rcvif_psref(ifp, &psref);
745 IP_STATINC(IP_STAT_CANTFORWARD);
746 m_freem(m);
747 } else {
748 /*
749 * If ip_dst matched any of my address on !IFF_UP interface,
750 * and there's no IFF_UP interface that matches ip_dst,
751 * send icmp unreach. Forwarding it will result in in-kernel
752 * forwarding loop till TTL goes to 0.
753 */
754 if (downmatch) {
755 m_put_rcvif_psref(ifp, &psref);
756 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
757 IP_STATINC(IP_STAT_CANTFORWARD);
758 return;
759 }
760 #ifdef IPSEC
761 /* Perform IPsec, if any. */
762 if (ipsec_used) {
763 SOFTNET_LOCK();
764 if (ipsec4_input(m, IP_FORWARDING |
765 (ip_directedbcast ? IP_ALLOWBROADCAST : 0)) != 0) {
766 SOFTNET_UNLOCK();
767 goto out;
768 }
769 SOFTNET_UNLOCK();
770 }
771 #endif
772 ip_forward(m, srcrt, ifp);
773 m_put_rcvif_psref(ifp, &psref);
774 }
775 return;
776
777 ours:
778 m_put_rcvif_psref(ifp, &psref);
779 ifp = NULL;
780
781 /*
782 * If offset or IP_MF are set, must reassemble.
783 */
784 if (ip->ip_off & ~htons(IP_DF|IP_RF)) {
785 /*
786 * Pass to IP reassembly mechanism.
787 */
788 if (ip_reass_packet(&m, ip) != 0) {
789 /* Failed; invalid fragment(s) or packet. */
790 goto out;
791 }
792 if (m == NULL) {
793 /* More fragments should come; silently return. */
794 goto out;
795 }
796 /*
797 * Reassembly is done, we have the final packet.
798 * Updated cached data in local variable(s).
799 */
800 ip = mtod(m, struct ip *);
801 hlen = ip->ip_hl << 2;
802 }
803
804 #ifdef IPSEC
805 /*
806 * Enforce IPsec policy checking if we are seeing last header.
807 * Note that we do not visit this with protocols with PCB layer
808 * code - like UDP/TCP/raw IP.
809 */
810 if (ipsec_used &&
811 (inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) {
812 SOFTNET_LOCK();
813 if (ipsec4_input(m, 0) != 0) {
814 SOFTNET_UNLOCK();
815 goto out;
816 }
817 SOFTNET_UNLOCK();
818 }
819 #endif
820
821 /*
822 * Switch out to protocol's input routine.
823 */
824 #if IFA_STATS
825 if (ia && ip) {
826 struct in_ifaddr *_ia;
827 /*
828 * Keep a reference from ip_match_our_address with psref
829 * is expensive, so explore ia here again.
830 */
831 s = pserialize_read_enter();
832 _ia = in_get_ia(ip->ip_dst.s_addr);
833 _ia->ia_ifa.ifa_data.ifad_inbytes += ntohs(ip->ip_len);
834 pserialize_read_exit(s);
835 }
836 #endif
837 IP_STATINC(IP_STAT_DELIVERED);
838
839 const int off = hlen, nh = ip->ip_p;
840
841 SOFTNET_LOCK();
842 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh);
843 SOFTNET_UNLOCK();
844 return;
845
846 out:
847 m_put_rcvif_psref(ifp, &psref);
848 if (m != NULL)
849 m_freem(m);
850 }
851
852 /*
853 * IP timer processing.
854 */
855 void
856 ip_slowtimo(void)
857 {
858
859 mutex_enter(softnet_lock);
860 KERNEL_LOCK(1, NULL);
861
862 ip_reass_slowtimo();
863
864 KERNEL_UNLOCK_ONE(NULL);
865 mutex_exit(softnet_lock);
866 }
867
868 /*
869 * IP drain processing.
870 */
871 void
872 ip_drain(void)
873 {
874
875 KERNEL_LOCK(1, NULL);
876 ip_reass_drain();
877 KERNEL_UNLOCK_ONE(NULL);
878 }
879
880 /*
881 * ip_dooptions: perform option processing on a datagram, possibly discarding
882 * it if bad options are encountered, or forwarding it if source-routed.
883 *
884 * => Returns true if packet has been forwarded/freed.
885 * => Returns false if the packet should be processed further.
886 */
887 static bool
888 ip_dooptions(struct mbuf *m)
889 {
890 struct ip *ip = mtod(m, struct ip *);
891 u_char *cp, *cp0;
892 struct ip_timestamp *ipt;
893 struct in_ifaddr *ia;
894 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0;
895 struct in_addr dst;
896 n_time ntime;
897 struct ifaddr *ifa;
898 int s;
899
900 dst = ip->ip_dst;
901 cp = (u_char *)(ip + 1);
902 cnt = (ip->ip_hl << 2) - sizeof (struct ip);
903 for (; cnt > 0; cnt -= optlen, cp += optlen) {
904 opt = cp[IPOPT_OPTVAL];
905 if (opt == IPOPT_EOL)
906 break;
907 if (opt == IPOPT_NOP)
908 optlen = 1;
909 else {
910 if (cnt < IPOPT_OLEN + sizeof(*cp)) {
911 code = &cp[IPOPT_OLEN] - (u_char *)ip;
912 goto bad;
913 }
914 optlen = cp[IPOPT_OLEN];
915 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
916 code = &cp[IPOPT_OLEN] - (u_char *)ip;
917 goto bad;
918 }
919 }
920 switch (opt) {
921
922 default:
923 break;
924
925 /*
926 * Source routing with record.
927 * Find interface with current destination address.
928 * If none on this machine then drop if strictly routed,
929 * or do nothing if loosely routed.
930 * Record interface address and bring up next address
931 * component. If strictly routed make sure next
932 * address is on directly accessible net.
933 */
934 case IPOPT_LSRR:
935 case IPOPT_SSRR:
936 if (ip_allowsrcrt == 0) {
937 type = ICMP_UNREACH;
938 code = ICMP_UNREACH_NET_PROHIB;
939 goto bad;
940 }
941 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
942 code = &cp[IPOPT_OLEN] - (u_char *)ip;
943 goto bad;
944 }
945 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
946 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
947 goto bad;
948 }
949 ipaddr.sin_addr = ip->ip_dst;
950
951 s = pserialize_read_enter();
952 ifa = ifa_ifwithaddr(sintosa(&ipaddr));
953 if (ifa == NULL) {
954 pserialize_read_exit(s);
955 if (opt == IPOPT_SSRR) {
956 type = ICMP_UNREACH;
957 code = ICMP_UNREACH_SRCFAIL;
958 goto bad;
959 }
960 /*
961 * Loose routing, and not at next destination
962 * yet; nothing to do except forward.
963 */
964 break;
965 }
966 pserialize_read_exit(s);
967
968 off--; /* 0 origin */
969 if ((off + sizeof(struct in_addr)) > optlen) {
970 /*
971 * End of source route. Should be for us.
972 */
973 save_rte(cp, ip->ip_src);
974 break;
975 }
976 /*
977 * locate outgoing interface
978 */
979 memcpy((void *)&ipaddr.sin_addr, (void *)(cp + off),
980 sizeof(ipaddr.sin_addr));
981 s = pserialize_read_enter();
982 if (opt == IPOPT_SSRR) {
983 ifa = ifa_ifwithladdr(sintosa(&ipaddr));
984 if (ifa != NULL)
985 ia = ifatoia(ifa);
986 else
987 ia = NULL;
988 } else {
989 ia = ip_rtaddr(ipaddr.sin_addr);
990 }
991 if (ia == NULL) {
992 type = ICMP_UNREACH;
993 code = ICMP_UNREACH_SRCFAIL;
994 pserialize_read_exit(s);
995 goto bad;
996 }
997 ip->ip_dst = ipaddr.sin_addr;
998 bcopy((void *)&ia->ia_addr.sin_addr,
999 (void *)(cp + off), sizeof(struct in_addr));
1000 pserialize_read_exit(s);
1001 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1002 /*
1003 * Let ip_intr's mcast routing check handle mcast pkts
1004 */
1005 forward = !IN_MULTICAST(ip->ip_dst.s_addr);
1006 break;
1007
1008 case IPOPT_RR:
1009 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1010 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1011 goto bad;
1012 }
1013 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1014 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1015 goto bad;
1016 }
1017 /*
1018 * If no space remains, ignore.
1019 */
1020 off--; /* 0 origin */
1021 if ((off + sizeof(struct in_addr)) > optlen)
1022 break;
1023 memcpy((void *)&ipaddr.sin_addr, (void *)(&ip->ip_dst),
1024 sizeof(ipaddr.sin_addr));
1025 /*
1026 * locate outgoing interface; if we're the destination,
1027 * use the incoming interface (should be same).
1028 */
1029 s = pserialize_read_enter();
1030 ifa = ifa_ifwithaddr(sintosa(&ipaddr));
1031 if (ifa == NULL) {
1032 ia = ip_rtaddr(ipaddr.sin_addr);
1033 if (ia == NULL) {
1034 pserialize_read_exit(s);
1035 type = ICMP_UNREACH;
1036 code = ICMP_UNREACH_HOST;
1037 goto bad;
1038 }
1039 } else {
1040 ia = ifatoia(ifa);
1041 }
1042 bcopy((void *)&ia->ia_addr.sin_addr,
1043 (void *)(cp + off), sizeof(struct in_addr));
1044 pserialize_read_exit(s);
1045 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1046 break;
1047
1048 case IPOPT_TS:
1049 code = cp - (u_char *)ip;
1050 ipt = (struct ip_timestamp *)cp;
1051 if (ipt->ipt_len < 4 || ipt->ipt_len > 40) {
1052 code = (u_char *)&ipt->ipt_len - (u_char *)ip;
1053 goto bad;
1054 }
1055 if (ipt->ipt_ptr < 5) {
1056 code = (u_char *)&ipt->ipt_ptr - (u_char *)ip;
1057 goto bad;
1058 }
1059 if (ipt->ipt_ptr > ipt->ipt_len - sizeof (int32_t)) {
1060 if (++ipt->ipt_oflw == 0) {
1061 code = (u_char *)&ipt->ipt_ptr -
1062 (u_char *)ip;
1063 goto bad;
1064 }
1065 break;
1066 }
1067 cp0 = (cp + ipt->ipt_ptr - 1);
1068 switch (ipt->ipt_flg) {
1069
1070 case IPOPT_TS_TSONLY:
1071 break;
1072
1073 case IPOPT_TS_TSANDADDR: {
1074 struct ifnet *rcvif;
1075 int _s, _ss;
1076
1077 if (ipt->ipt_ptr - 1 + sizeof(n_time) +
1078 sizeof(struct in_addr) > ipt->ipt_len) {
1079 code = (u_char *)&ipt->ipt_ptr -
1080 (u_char *)ip;
1081 goto bad;
1082 }
1083 ipaddr.sin_addr = dst;
1084 _ss = pserialize_read_enter();
1085 rcvif = m_get_rcvif(m, &_s);
1086 ifa = ifaof_ifpforaddr(sintosa(&ipaddr), rcvif);
1087 m_put_rcvif(rcvif, &_s);
1088 if (ifa == NULL) {
1089 pserialize_read_exit(_ss);
1090 break;
1091 }
1092 ia = ifatoia(ifa);
1093 bcopy(&ia->ia_addr.sin_addr,
1094 cp0, sizeof(struct in_addr));
1095 pserialize_read_exit(_ss);
1096 ipt->ipt_ptr += sizeof(struct in_addr);
1097 break;
1098 }
1099
1100 case IPOPT_TS_PRESPEC:
1101 if (ipt->ipt_ptr - 1 + sizeof(n_time) +
1102 sizeof(struct in_addr) > ipt->ipt_len) {
1103 code = (u_char *)&ipt->ipt_ptr -
1104 (u_char *)ip;
1105 goto bad;
1106 }
1107 memcpy(&ipaddr.sin_addr, cp0,
1108 sizeof(struct in_addr));
1109 s = pserialize_read_enter();
1110 ifa = ifa_ifwithaddr(sintosa(&ipaddr));
1111 if (ifa == NULL) {
1112 pserialize_read_exit(s);
1113 continue;
1114 }
1115 pserialize_read_exit(s);
1116 ipt->ipt_ptr += sizeof(struct in_addr);
1117 break;
1118
1119 default:
1120 /* XXX can't take &ipt->ipt_flg */
1121 code = (u_char *)&ipt->ipt_ptr -
1122 (u_char *)ip + 1;
1123 goto bad;
1124 }
1125 ntime = iptime();
1126 cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */
1127 memmove((char *)cp + ipt->ipt_ptr - 1, cp0,
1128 sizeof(n_time));
1129 ipt->ipt_ptr += sizeof(n_time);
1130 }
1131 }
1132 if (forward) {
1133 struct ifnet *rcvif;
1134 struct psref _psref;
1135
1136 if (ip_forwsrcrt == 0) {
1137 type = ICMP_UNREACH;
1138 code = ICMP_UNREACH_SRCFAIL;
1139 goto bad;
1140 }
1141
1142 rcvif = m_get_rcvif_psref(m, &_psref);
1143 if (__predict_false(rcvif == NULL)) {
1144 type = ICMP_UNREACH;
1145 code = ICMP_UNREACH_HOST;
1146 goto bad;
1147 }
1148 ip_forward(m, 1, rcvif);
1149 m_put_rcvif_psref(rcvif, &_psref);
1150 return true;
1151 }
1152 return false;
1153 bad:
1154 icmp_error(m, type, code, 0, 0);
1155 IP_STATINC(IP_STAT_BADOPTIONS);
1156 return true;
1157 }
1158
1159 /*
1160 * ip_rtaddr: given address of next destination (final or next hop),
1161 * return internet address info of interface to be used to get there.
1162 */
1163 static struct in_ifaddr *
1164 ip_rtaddr(struct in_addr dst)
1165 {
1166 struct rtentry *rt;
1167 union {
1168 struct sockaddr dst;
1169 struct sockaddr_in dst4;
1170 } u;
1171 struct route *ro;
1172
1173 sockaddr_in_init(&u.dst4, &dst, 0);
1174
1175 SOFTNET_LOCK();
1176 ro = percpu_getref(ipforward_rt_percpu);
1177 rt = rtcache_lookup(ro, &u.dst);
1178 percpu_putref(ipforward_rt_percpu);
1179 SOFTNET_UNLOCK();
1180 if (rt == NULL)
1181 return NULL;
1182
1183 return ifatoia(rt->rt_ifa);
1184 }
1185
1186 /*
1187 * save_rte: save incoming source route for use in replies, to be picked
1188 * up later by ip_srcroute if the receiver is interested.
1189 */
1190 static void
1191 save_rte(u_char *option, struct in_addr dst)
1192 {
1193 unsigned olen;
1194
1195 olen = option[IPOPT_OLEN];
1196 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst)))
1197 return;
1198 memcpy((void *)ip_srcrt.srcopt, (void *)option, olen);
1199 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1200 ip_srcrt.dst = dst;
1201 }
1202
1203 /*
1204 * Retrieve incoming source route for use in replies,
1205 * in the same form used by setsockopt.
1206 * The first hop is placed before the options, will be removed later.
1207 */
1208 struct mbuf *
1209 ip_srcroute(void)
1210 {
1211 struct in_addr *p, *q;
1212 struct mbuf *m;
1213
1214 if (ip_nhops == 0)
1215 return NULL;
1216 m = m_get(M_DONTWAIT, MT_SOOPTS);
1217 if (m == 0)
1218 return NULL;
1219
1220 MCLAIM(m, &inetdomain.dom_mowner);
1221 #define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt))
1222
1223 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
1224 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) +
1225 OPTSIZ;
1226
1227 /*
1228 * First save first hop for return route
1229 */
1230 p = &ip_srcrt.route[ip_nhops - 1];
1231 *(mtod(m, struct in_addr *)) = *p--;
1232
1233 /*
1234 * Copy option fields and padding (nop) to mbuf.
1235 */
1236 ip_srcrt.nop = IPOPT_NOP;
1237 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
1238 memmove(mtod(m, char *) + sizeof(struct in_addr), &ip_srcrt.nop,
1239 OPTSIZ);
1240 q = (struct in_addr *)(mtod(m, char *) +
1241 sizeof(struct in_addr) + OPTSIZ);
1242 #undef OPTSIZ
1243 /*
1244 * Record return path as an IP source route,
1245 * reversing the path (pointers are now aligned).
1246 */
1247 while (p >= ip_srcrt.route) {
1248 *q++ = *p--;
1249 }
1250 /*
1251 * Last hop goes to final destination.
1252 */
1253 *q = ip_srcrt.dst;
1254 return (m);
1255 }
1256
1257 const int inetctlerrmap[PRC_NCMDS] = {
1258 [PRC_MSGSIZE] = EMSGSIZE,
1259 [PRC_HOSTDEAD] = EHOSTDOWN,
1260 [PRC_HOSTUNREACH] = EHOSTUNREACH,
1261 [PRC_UNREACH_NET] = EHOSTUNREACH,
1262 [PRC_UNREACH_HOST] = EHOSTUNREACH,
1263 [PRC_UNREACH_PROTOCOL] = ECONNREFUSED,
1264 [PRC_UNREACH_PORT] = ECONNREFUSED,
1265 [PRC_UNREACH_SRCFAIL] = EHOSTUNREACH,
1266 [PRC_PARAMPROB] = ENOPROTOOPT,
1267 };
1268
1269 void
1270 ip_fasttimo(void)
1271 {
1272 if (ip_drainwanted) {
1273 ip_drain();
1274 ip_drainwanted = 0;
1275 }
1276 }
1277
1278 void
1279 ip_drainstub(void)
1280 {
1281 ip_drainwanted = 1;
1282 }
1283
1284 /*
1285 * Forward a packet. If some error occurs return the sender
1286 * an icmp packet. Note we can't always generate a meaningful
1287 * icmp message because icmp doesn't have a large enough repertoire
1288 * of codes and types.
1289 *
1290 * If not forwarding, just drop the packet. This could be confusing
1291 * if ipforwarding was zero but some routing protocol was advancing
1292 * us as a gateway to somewhere. However, we must let the routing
1293 * protocol deal with that.
1294 *
1295 * The srcrt parameter indicates whether the packet is being forwarded
1296 * via a source route.
1297 */
1298 static void
1299 ip_forward(struct mbuf *m, int srcrt, struct ifnet *rcvif)
1300 {
1301 struct ip *ip = mtod(m, struct ip *);
1302 struct rtentry *rt;
1303 int error, type = 0, code = 0, destmtu = 0;
1304 struct mbuf *mcopy;
1305 n_long dest;
1306 union {
1307 struct sockaddr dst;
1308 struct sockaddr_in dst4;
1309 } u;
1310 uint64_t *ips;
1311 struct route *ro;
1312
1313 KASSERTMSG(cpu_softintr_p(), "ip_forward: not in the software "
1314 "interrupt handler; synchronization assumptions violated");
1315
1316 /*
1317 * We are now in the output path.
1318 */
1319 MCLAIM(m, &ip_tx_mowner);
1320
1321 /*
1322 * Clear any in-bound checksum flags for this packet.
1323 */
1324 m->m_pkthdr.csum_flags = 0;
1325
1326 dest = 0;
1327 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
1328 IP_STATINC(IP_STAT_CANTFORWARD);
1329 m_freem(m);
1330 return;
1331 }
1332
1333 SOFTNET_LOCK();
1334
1335 if (ip->ip_ttl <= IPTTLDEC) {
1336 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0);
1337 SOFTNET_UNLOCK();
1338 return;
1339 }
1340
1341 sockaddr_in_init(&u.dst4, &ip->ip_dst, 0);
1342
1343 ro = percpu_getref(ipforward_rt_percpu);
1344 if ((rt = rtcache_lookup(ro, &u.dst)) == NULL) {
1345 percpu_putref(ipforward_rt_percpu);
1346 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_NET, dest, 0);
1347 SOFTNET_UNLOCK();
1348 return;
1349 }
1350
1351 /*
1352 * Save at most 68 bytes of the packet in case
1353 * we need to generate an ICMP message to the src.
1354 * Pullup to avoid sharing mbuf cluster between m and mcopy.
1355 */
1356 mcopy = m_copym(m, 0, imin(ntohs(ip->ip_len), 68), M_DONTWAIT);
1357 if (mcopy)
1358 mcopy = m_pullup(mcopy, ip->ip_hl << 2);
1359
1360 ip->ip_ttl -= IPTTLDEC;
1361
1362 /*
1363 * If forwarding packet using same interface that it came in on,
1364 * perhaps should send a redirect to sender to shortcut a hop.
1365 * Only send redirect if source is sending directly to us,
1366 * and if packet was not source routed (or has any options).
1367 * Also, don't send redirect if forwarding using a default route
1368 * or a route modified by a redirect.
1369 */
1370 if (rt->rt_ifp == rcvif &&
1371 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1372 !in_nullhost(satocsin(rt_getkey(rt))->sin_addr) &&
1373 ipsendredirects && !srcrt) {
1374 if (rt->rt_ifa &&
1375 (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) ==
1376 ifatoia(rt->rt_ifa)->ia_subnet) {
1377 if (rt->rt_flags & RTF_GATEWAY)
1378 dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
1379 else
1380 dest = ip->ip_dst.s_addr;
1381 /*
1382 * Router requirements says to only send host
1383 * redirects.
1384 */
1385 type = ICMP_REDIRECT;
1386 code = ICMP_REDIRECT_HOST;
1387 }
1388 }
1389
1390 error = ip_output(m, NULL, ro,
1391 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)),
1392 NULL, NULL);
1393
1394 if (error) {
1395 IP_STATINC(IP_STAT_CANTFORWARD);
1396 goto error;
1397 }
1398
1399 ips = IP_STAT_GETREF();
1400 ips[IP_STAT_FORWARD]++;
1401
1402 if (type) {
1403 ips[IP_STAT_REDIRECTSENT]++;
1404 IP_STAT_PUTREF();
1405 goto redirect;
1406 }
1407
1408 IP_STAT_PUTREF();
1409 if (mcopy) {
1410 #ifdef GATEWAY
1411 if (mcopy->m_flags & M_CANFASTFWD)
1412 ipflow_create(ro, mcopy);
1413 #endif
1414 m_freem(mcopy);
1415 }
1416
1417 percpu_putref(ipforward_rt_percpu);
1418 SOFTNET_UNLOCK();
1419 return;
1420
1421 redirect:
1422 error:
1423 if (mcopy == NULL) {
1424 percpu_putref(ipforward_rt_percpu);
1425 SOFTNET_UNLOCK();
1426 return;
1427 }
1428
1429 switch (error) {
1430
1431 case 0: /* forwarded, but need redirect */
1432 /* type, code set above */
1433 break;
1434
1435 case ENETUNREACH: /* shouldn't happen, checked above */
1436 case EHOSTUNREACH:
1437 case ENETDOWN:
1438 case EHOSTDOWN:
1439 default:
1440 type = ICMP_UNREACH;
1441 code = ICMP_UNREACH_HOST;
1442 break;
1443
1444 case EMSGSIZE:
1445 type = ICMP_UNREACH;
1446 code = ICMP_UNREACH_NEEDFRAG;
1447
1448 if ((rt = rtcache_validate(ro)) != NULL)
1449 destmtu = rt->rt_ifp->if_mtu;
1450 #ifdef IPSEC
1451 if (ipsec_used)
1452 (void)ipsec4_forward(mcopy, &destmtu);
1453 #endif
1454 IP_STATINC(IP_STAT_CANTFRAG);
1455 break;
1456
1457 case ENOBUFS:
1458 /*
1459 * Do not generate ICMP_SOURCEQUENCH as required in RFC 1812,
1460 * Requirements for IP Version 4 Routers. Source quench can
1461 * big problem under DoS attacks or if the underlying
1462 * interface is rate-limited.
1463 */
1464 if (mcopy)
1465 m_freem(mcopy);
1466 percpu_putref(ipforward_rt_percpu);
1467 SOFTNET_UNLOCK();
1468 return;
1469 }
1470 icmp_error(mcopy, type, code, dest, destmtu);
1471 percpu_putref(ipforward_rt_percpu);
1472 SOFTNET_UNLOCK();
1473 }
1474
1475 void
1476 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1477 struct mbuf *m)
1478 {
1479 struct socket *so = inp->inp_socket;
1480 ifnet_t *ifp;
1481 int inpflags = inp->inp_flags;
1482 struct psref psref;
1483
1484 ifp = m_get_rcvif_psref(m, &psref);
1485 if (__predict_false(ifp == NULL))
1486 return; /* XXX should report error? */
1487
1488 if (so->so_options & SO_TIMESTAMP
1489 #ifdef SO_OTIMESTAMP
1490 || so->so_options & SO_OTIMESTAMP
1491 #endif
1492 ) {
1493 struct timeval tv;
1494
1495 microtime(&tv);
1496 #ifdef SO_OTIMESTAMP
1497 if (so->so_options & SO_OTIMESTAMP) {
1498 struct timeval50 tv50;
1499 timeval_to_timeval50(&tv, &tv50);
1500 *mp = sbcreatecontrol((void *) &tv50, sizeof(tv50),
1501 SCM_OTIMESTAMP, SOL_SOCKET);
1502 } else
1503 #endif
1504 *mp = sbcreatecontrol((void *) &tv, sizeof(tv),
1505 SCM_TIMESTAMP, SOL_SOCKET);
1506 if (*mp)
1507 mp = &(*mp)->m_next;
1508 }
1509 if (inpflags & INP_RECVDSTADDR) {
1510 *mp = sbcreatecontrol((void *) &ip->ip_dst,
1511 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1512 if (*mp)
1513 mp = &(*mp)->m_next;
1514 }
1515 if (inpflags & INP_RECVPKTINFO) {
1516 struct in_pktinfo ipi;
1517 ipi.ipi_addr = ip->ip_src;
1518 ipi.ipi_ifindex = ifp->if_index;
1519 *mp = sbcreatecontrol((void *) &ipi,
1520 sizeof(ipi), IP_RECVPKTINFO, IPPROTO_IP);
1521 if (*mp)
1522 mp = &(*mp)->m_next;
1523 }
1524 if (inpflags & INP_PKTINFO) {
1525 struct in_pktinfo ipi;
1526 ipi.ipi_addr = ip->ip_dst;
1527 ipi.ipi_ifindex = ifp->if_index;
1528 *mp = sbcreatecontrol((void *) &ipi,
1529 sizeof(ipi), IP_PKTINFO, IPPROTO_IP);
1530 if (*mp)
1531 mp = &(*mp)->m_next;
1532 }
1533 if (inpflags & INP_RECVIF) {
1534 struct sockaddr_dl sdl;
1535
1536 sockaddr_dl_init(&sdl, sizeof(sdl), ifp->if_index, 0, NULL, 0,
1537 NULL, 0);
1538 *mp = sbcreatecontrol(&sdl, sdl.sdl_len, IP_RECVIF, IPPROTO_IP);
1539 if (*mp)
1540 mp = &(*mp)->m_next;
1541 }
1542 if (inpflags & INP_RECVTTL) {
1543 *mp = sbcreatecontrol((void *) &ip->ip_ttl,
1544 sizeof(uint8_t), IP_RECVTTL, IPPROTO_IP);
1545 if (*mp)
1546 mp = &(*mp)->m_next;
1547 }
1548 m_put_rcvif_psref(ifp, &psref);
1549 }
1550
1551 /*
1552 * sysctl helper routine for net.inet.ip.forwsrcrt.
1553 */
1554 static int
1555 sysctl_net_inet_ip_forwsrcrt(SYSCTLFN_ARGS)
1556 {
1557 int error, tmp;
1558 struct sysctlnode node;
1559
1560 node = *rnode;
1561 tmp = ip_forwsrcrt;
1562 node.sysctl_data = &tmp;
1563 error = sysctl_lookup(SYSCTLFN_CALL(&node));
1564 if (error || newp == NULL)
1565 return (error);
1566
1567 error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_FORWSRCRT,
1568 0, NULL, NULL, NULL);
1569 if (error)
1570 return (error);
1571
1572 ip_forwsrcrt = tmp;
1573
1574 return (0);
1575 }
1576
1577 /*
1578 * sysctl helper routine for net.inet.ip.mtudisctimeout. checks the
1579 * range of the new value and tweaks timers if it changes.
1580 */
1581 static int
1582 sysctl_net_inet_ip_pmtudto(SYSCTLFN_ARGS)
1583 {
1584 int error, tmp;
1585 struct sysctlnode node;
1586
1587 node = *rnode;
1588 tmp = ip_mtudisc_timeout;
1589 node.sysctl_data = &tmp;
1590 error = sysctl_lookup(SYSCTLFN_CALL(&node));
1591 if (error || newp == NULL)
1592 return (error);
1593 if (tmp < 0)
1594 return (EINVAL);
1595
1596 mutex_enter(softnet_lock);
1597
1598 ip_mtudisc_timeout = tmp;
1599 rt_timer_queue_change(ip_mtudisc_timeout_q, ip_mtudisc_timeout);
1600
1601 mutex_exit(softnet_lock);
1602
1603 return (0);
1604 }
1605
1606 static int
1607 sysctl_net_inet_ip_stats(SYSCTLFN_ARGS)
1608 {
1609
1610 return (NETSTAT_SYSCTL(ipstat_percpu, IP_NSTATS));
1611 }
1612
1613 static void
1614 sysctl_net_inet_ip_setup(struct sysctllog **clog)
1615 {
1616 sysctl_createv(clog, 0, NULL, NULL,
1617 CTLFLAG_PERMANENT,
1618 CTLTYPE_NODE, "inet",
1619 SYSCTL_DESCR("PF_INET related settings"),
1620 NULL, 0, NULL, 0,
1621 CTL_NET, PF_INET, CTL_EOL);
1622 sysctl_createv(clog, 0, NULL, NULL,
1623 CTLFLAG_PERMANENT,
1624 CTLTYPE_NODE, "ip",
1625 SYSCTL_DESCR("IPv4 related settings"),
1626 NULL, 0, NULL, 0,
1627 CTL_NET, PF_INET, IPPROTO_IP, CTL_EOL);
1628
1629 sysctl_createv(clog, 0, NULL, NULL,
1630 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1631 CTLTYPE_INT, "forwarding",
1632 SYSCTL_DESCR("Enable forwarding of INET datagrams"),
1633 NULL, 0, &ipforwarding, 0,
1634 CTL_NET, PF_INET, IPPROTO_IP,
1635 IPCTL_FORWARDING, CTL_EOL);
1636 sysctl_createv(clog, 0, NULL, NULL,
1637 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1638 CTLTYPE_INT, "redirect",
1639 SYSCTL_DESCR("Enable sending of ICMP redirect messages"),
1640 NULL, 0, &ipsendredirects, 0,
1641 CTL_NET, PF_INET, IPPROTO_IP,
1642 IPCTL_SENDREDIRECTS, CTL_EOL);
1643 sysctl_createv(clog, 0, NULL, NULL,
1644 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1645 CTLTYPE_INT, "ttl",
1646 SYSCTL_DESCR("Default TTL for an INET datagram"),
1647 NULL, 0, &ip_defttl, 0,
1648 CTL_NET, PF_INET, IPPROTO_IP,
1649 IPCTL_DEFTTL, CTL_EOL);
1650 #ifdef IPCTL_DEFMTU
1651 sysctl_createv(clog, 0, NULL, NULL,
1652 CTLFLAG_PERMANENT /* |CTLFLAG_READWRITE? */,
1653 CTLTYPE_INT, "mtu",
1654 SYSCTL_DESCR("Default MTA for an INET route"),
1655 NULL, 0, &ip_mtu, 0,
1656 CTL_NET, PF_INET, IPPROTO_IP,
1657 IPCTL_DEFMTU, CTL_EOL);
1658 #endif /* IPCTL_DEFMTU */
1659 sysctl_createv(clog, 0, NULL, NULL,
1660 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1661 CTLTYPE_INT, "forwsrcrt",
1662 SYSCTL_DESCR("Enable forwarding of source-routed "
1663 "datagrams"),
1664 sysctl_net_inet_ip_forwsrcrt, 0, &ip_forwsrcrt, 0,
1665 CTL_NET, PF_INET, IPPROTO_IP,
1666 IPCTL_FORWSRCRT, CTL_EOL);
1667 sysctl_createv(clog, 0, NULL, NULL,
1668 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1669 CTLTYPE_INT, "directed-broadcast",
1670 SYSCTL_DESCR("Enable forwarding of broadcast datagrams"),
1671 NULL, 0, &ip_directedbcast, 0,
1672 CTL_NET, PF_INET, IPPROTO_IP,
1673 IPCTL_DIRECTEDBCAST, CTL_EOL);
1674 sysctl_createv(clog, 0, NULL, NULL,
1675 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1676 CTLTYPE_INT, "allowsrcrt",
1677 SYSCTL_DESCR("Accept source-routed datagrams"),
1678 NULL, 0, &ip_allowsrcrt, 0,
1679 CTL_NET, PF_INET, IPPROTO_IP,
1680 IPCTL_ALLOWSRCRT, CTL_EOL);
1681
1682 sysctl_createv(clog, 0, NULL, NULL,
1683 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1684 CTLTYPE_INT, "mtudisc",
1685 SYSCTL_DESCR("Use RFC1191 Path MTU Discovery"),
1686 NULL, 0, &ip_mtudisc, 0,
1687 CTL_NET, PF_INET, IPPROTO_IP,
1688 IPCTL_MTUDISC, CTL_EOL);
1689 sysctl_createv(clog, 0, NULL, NULL,
1690 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1691 CTLTYPE_INT, "anonportmin",
1692 SYSCTL_DESCR("Lowest ephemeral port number to assign"),
1693 sysctl_net_inet_ip_ports, 0, &anonportmin, 0,
1694 CTL_NET, PF_INET, IPPROTO_IP,
1695 IPCTL_ANONPORTMIN, CTL_EOL);
1696 sysctl_createv(clog, 0, NULL, NULL,
1697 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1698 CTLTYPE_INT, "anonportmax",
1699 SYSCTL_DESCR("Highest ephemeral port number to assign"),
1700 sysctl_net_inet_ip_ports, 0, &anonportmax, 0,
1701 CTL_NET, PF_INET, IPPROTO_IP,
1702 IPCTL_ANONPORTMAX, CTL_EOL);
1703 sysctl_createv(clog, 0, NULL, NULL,
1704 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1705 CTLTYPE_INT, "mtudisctimeout",
1706 SYSCTL_DESCR("Lifetime of a Path MTU Discovered route"),
1707 sysctl_net_inet_ip_pmtudto, 0, (void *)&ip_mtudisc_timeout, 0,
1708 CTL_NET, PF_INET, IPPROTO_IP,
1709 IPCTL_MTUDISCTIMEOUT, CTL_EOL);
1710 #ifndef IPNOPRIVPORTS
1711 sysctl_createv(clog, 0, NULL, NULL,
1712 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1713 CTLTYPE_INT, "lowportmin",
1714 SYSCTL_DESCR("Lowest privileged ephemeral port number "
1715 "to assign"),
1716 sysctl_net_inet_ip_ports, 0, &lowportmin, 0,
1717 CTL_NET, PF_INET, IPPROTO_IP,
1718 IPCTL_LOWPORTMIN, CTL_EOL);
1719 sysctl_createv(clog, 0, NULL, NULL,
1720 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1721 CTLTYPE_INT, "lowportmax",
1722 SYSCTL_DESCR("Highest privileged ephemeral port number "
1723 "to assign"),
1724 sysctl_net_inet_ip_ports, 0, &lowportmax, 0,
1725 CTL_NET, PF_INET, IPPROTO_IP,
1726 IPCTL_LOWPORTMAX, CTL_EOL);
1727 #endif /* IPNOPRIVPORTS */
1728 #if NGRE > 0
1729 sysctl_createv(clog, 0, NULL, NULL,
1730 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1731 CTLTYPE_INT, "grettl",
1732 SYSCTL_DESCR("Default TTL for a gre tunnel datagram"),
1733 NULL, 0, &ip_gre_ttl, 0,
1734 CTL_NET, PF_INET, IPPROTO_IP,
1735 IPCTL_GRE_TTL, CTL_EOL);
1736 #endif /* NGRE */
1737 sysctl_createv(clog, 0, NULL, NULL,
1738 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1739 CTLTYPE_INT, "checkinterface",
1740 SYSCTL_DESCR("Enable receive side of Strong ES model "
1741 "from RFC1122"),
1742 NULL, 0, &ip_checkinterface, 0,
1743 CTL_NET, PF_INET, IPPROTO_IP,
1744 IPCTL_CHECKINTERFACE, CTL_EOL);
1745 sysctl_createv(clog, 0, NULL, NULL,
1746 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1747 CTLTYPE_INT, "random_id",
1748 SYSCTL_DESCR("Assign random ip_id values"),
1749 NULL, 0, &ip_do_randomid, 0,
1750 CTL_NET, PF_INET, IPPROTO_IP,
1751 IPCTL_RANDOMID, CTL_EOL);
1752 sysctl_createv(clog, 0, NULL, NULL,
1753 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1754 CTLTYPE_INT, "do_loopback_cksum",
1755 SYSCTL_DESCR("Perform IP checksum on loopback"),
1756 NULL, 0, &ip_do_loopback_cksum, 0,
1757 CTL_NET, PF_INET, IPPROTO_IP,
1758 IPCTL_LOOPBACKCKSUM, CTL_EOL);
1759 sysctl_createv(clog, 0, NULL, NULL,
1760 CTLFLAG_PERMANENT,
1761 CTLTYPE_STRUCT, "stats",
1762 SYSCTL_DESCR("IP statistics"),
1763 sysctl_net_inet_ip_stats, 0, NULL, 0,
1764 CTL_NET, PF_INET, IPPROTO_IP, IPCTL_STATS,
1765 CTL_EOL);
1766 #if NARP
1767 sysctl_createv(clog, 0, NULL, NULL,
1768 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1769 CTLTYPE_INT, "dad_count",
1770 SYSCTL_DESCR("Number of Duplicate Address Detection "
1771 "probes to send"),
1772 NULL, 0, &ip_dad_count, 0,
1773 CTL_NET, PF_INET, IPPROTO_IP,
1774 IPCTL_DAD_COUNT, CTL_EOL);
1775 #endif
1776
1777 /* anonportalgo RFC6056 subtree */
1778 const struct sysctlnode *portalgo_node;
1779 sysctl_createv(clog, 0, NULL, &portalgo_node,
1780 CTLFLAG_PERMANENT,
1781 CTLTYPE_NODE, "anonportalgo",
1782 SYSCTL_DESCR("Anonymous Port Algorithm Selection (RFC 6056)"),
1783 NULL, 0, NULL, 0,
1784 CTL_NET, PF_INET, IPPROTO_IP, CTL_CREATE, CTL_EOL);
1785 sysctl_createv(clog, 0, &portalgo_node, NULL,
1786 CTLFLAG_PERMANENT,
1787 CTLTYPE_STRING, "available",
1788 SYSCTL_DESCR("available algorithms"),
1789 sysctl_portalgo_available, 0, NULL, PORTALGO_MAXLEN,
1790 CTL_CREATE, CTL_EOL);
1791 sysctl_createv(clog, 0, &portalgo_node, NULL,
1792 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1793 CTLTYPE_STRING, "selected",
1794 SYSCTL_DESCR("selected algorithm"),
1795 sysctl_portalgo_selected4, 0, NULL, PORTALGO_MAXLEN,
1796 CTL_CREATE, CTL_EOL);
1797 sysctl_createv(clog, 0, &portalgo_node, NULL,
1798 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1799 CTLTYPE_STRUCT, "reserve",
1800 SYSCTL_DESCR("bitmap of reserved ports"),
1801 sysctl_portalgo_reserve4, 0, NULL, 0,
1802 CTL_CREATE, CTL_EOL);
1803 }
1804
1805 void
1806 ip_statinc(u_int stat)
1807 {
1808
1809 KASSERT(stat < IP_NSTATS);
1810 IP_STATINC(stat);
1811 }
1812