1/* $NetBSD: ip_input.c,v 1.344 2016/10/18 07:30:31 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 1.344 2016/10/18 07:30:31 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#include "opt_net_mpsafe.h"
105#endif
106
107#include "arp.h"
108
109#include <sys/param.h>
110#include <sys/systm.h>
111#include <sys/cpu.h>
112#include <sys/mbuf.h>
113#include <sys/domain.h>
114#include <sys/protosw.h>
115#include <sys/socket.h>
116#include <sys/socketvar.h>
117#include <sys/errno.h>
118#include <sys/time.h>
119#include <sys/kernel.h>
120#include <sys/pool.h>
121#include <sys/sysctl.h>
122#include <sys/kauth.h>
123
124#include <net/if.h>
125#include <net/if_dl.h>
126#include <net/route.h>
127#include <net/pktqueue.h>
128#include <net/pfil.h>
129
130#include <netinet/in.h>
131#include <netinet/in_systm.h>
132#include <netinet/ip.h>
133#include <netinet/in_pcb.h>
134#include <netinet/in_proto.h>
135#include <netinet/in_var.h>
136#include <netinet/ip_var.h>
137#include <netinet/ip_private.h>
138#include <netinet/ip_icmp.h>
139/* just for gif_ttl */
140#include <netinet/in_gif.h>
141#include "gif.h"
142#include <net/if_gre.h>
143#include "gre.h"
144
145#ifdef MROUTING
146#include <netinet/ip_mroute.h>
147#endif
148#include <netinet/portalgo.h>
149
150#ifdef IPSEC
151#include <netipsec/ipsec.h>
152#endif
153
154#ifndef IPFORWARDING
155#ifdef GATEWAY
156#define IPFORWARDING 1 /* forward IP packets not for us */
157#else /* GATEWAY */
158#define IPFORWARDING 0 /* don't forward IP packets not for us */
159#endif /* GATEWAY */
160#endif /* IPFORWARDING */
161#ifndef IPSENDREDIRECTS
162#define IPSENDREDIRECTS 1
163#endif
164#ifndef IPFORWSRCRT
165#define IPFORWSRCRT 1 /* forward source-routed packets */
166#endif
167#ifndef IPALLOWSRCRT
168#define IPALLOWSRCRT 1 /* allow source-routed packets */
169#endif
170#ifndef IPMTUDISC
171#define IPMTUDISC 1
172#endif
173#ifndef IPMTUDISCTIMEOUT
174#define IPMTUDISCTIMEOUT (10 * 60) /* as per RFC 1191 */
175#endif
176
177#ifdef COMPAT_50
178#include <compat/sys/time.h>
179#include <compat/sys/socket.h>
180#endif
181
182/*
183 * Note: DIRECTED_BROADCAST is handled this way so that previous
184 * configuration using this option will Just Work.
185 */
186#ifndef IPDIRECTEDBCAST
187#ifdef DIRECTED_BROADCAST
188#define IPDIRECTEDBCAST 1
189#else
190#define IPDIRECTEDBCAST 0
191#endif /* DIRECTED_BROADCAST */
192#endif /* IPDIRECTEDBCAST */
193int ipforwarding = IPFORWARDING;
194int ipsendredirects = IPSENDREDIRECTS;
195int ip_defttl = IPDEFTTL;
196int ip_forwsrcrt = IPFORWSRCRT;
197int ip_directedbcast = IPDIRECTEDBCAST;
198int ip_allowsrcrt = IPALLOWSRCRT;
199int ip_mtudisc = IPMTUDISC;
200int ip_mtudisc_timeout = IPMTUDISCTIMEOUT;
201#ifdef DIAGNOSTIC
202int ipprintfs = 0;
203#endif
204
205int ip_do_randomid = 0;
206
207/*
208 * XXX - Setting ip_checkinterface mostly implements the receive side of
209 * the Strong ES model described in RFC 1122, but since the routing table
210 * and transmit implementation do not implement the Strong ES model,
211 * setting this to 1 results in an odd hybrid.
212 *
213 * XXX - ip_checkinterface currently must be disabled if you use ipnat
214 * to translate the destination address to another local interface.
215 *
216 * XXX - ip_checkinterface must be disabled if you add IP aliases
217 * to the loopback interface instead of the interface where the
218 * packets for those addresses are received.
219 */
220static int ip_checkinterface __read_mostly = 0;
221
222struct rttimer_queue *ip_mtudisc_timeout_q = NULL;
223
224pktqueue_t * ip_pktq __read_mostly;
225pfil_head_t * inet_pfil_hook __read_mostly;
226ipid_state_t * ip_ids __read_mostly;
227percpu_t * ipstat_percpu __read_mostly;
228
229static percpu_t *ipforward_rt_percpu __cacheline_aligned;
230
231uint16_t ip_id;
232
233#ifdef INET_CSUM_COUNTERS
234#include <sys/device.h>
235
236struct evcnt ip_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
237 NULL, "inet", "hwcsum bad");
238struct evcnt ip_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
239 NULL, "inet", "hwcsum ok");
240struct evcnt ip_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
241 NULL, "inet", "swcsum");
242
243#define INET_CSUM_COUNTER_INCR(ev) (ev)->ev_count++
244
245EVCNT_ATTACH_STATIC(ip_hwcsum_bad);
246EVCNT_ATTACH_STATIC(ip_hwcsum_ok);
247EVCNT_ATTACH_STATIC(ip_swcsum);
248
249#else
250
251#define INET_CSUM_COUNTER_INCR(ev) /* nothing */
252
253#endif /* INET_CSUM_COUNTERS */
254
255/*
256 * We need to save the IP options in case a protocol wants to respond
257 * to an incoming packet over the same route if the packet got here
258 * using IP source routing. This allows connection establishment and
259 * maintenance when the remote end is on a network that is not known
260 * to us.
261 */
262
263static int ip_nhops = 0;
264
265static struct ip_srcrt {
266 struct in_addr dst; /* final destination */
267 char nop; /* one NOP to align */
268 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */
269 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)];
270} ip_srcrt;
271
272static int ip_drainwanted;
273
274struct sockaddr_in ipaddr = {
275 .sin_len = sizeof(ipaddr),
276 .sin_family = AF_INET,
277};
278
279static void save_rte(u_char *, struct in_addr);
280
281#ifdef MBUFTRACE
282struct mowner ip_rx_mowner = MOWNER_INIT("internet", "rx");
283struct mowner ip_tx_mowner = MOWNER_INIT("internet", "tx");
284#endif
285
286static void ipintr(void *);
287static void ip_input(struct mbuf *);
288static void ip_forward(struct mbuf *, int, struct ifnet *);
289static bool ip_dooptions(struct mbuf *);
290static struct in_ifaddr *ip_rtaddr(struct in_addr);
291static void sysctl_net_inet_ip_setup(struct sysctllog **);
292
293static struct in_ifaddr *ip_match_our_address(struct ifnet *, struct ip *,
294 int *);
295static struct in_ifaddr *ip_match_our_address_broadcast(struct ifnet *,
296 struct ip *);
297
298#ifdef NET_MPSAFE
299#define SOFTNET_LOCK() mutex_enter(softnet_lock)
300#define SOFTNET_UNLOCK() mutex_exit(softnet_lock)
301#else
302#define SOFTNET_LOCK() KASSERT(mutex_owned(softnet_lock))
303#define SOFTNET_UNLOCK() KASSERT(mutex_owned(softnet_lock))
304#endif
305
306/*
307 * IP initialization: fill in IP protocol switch table.
308 * All protocols not implemented in kernel go to raw IP protocol handler.
309 */
310void
311ip_init(void)
312{
313 const struct protosw *pr;
314
315 in_init();
316 sysctl_net_inet_ip_setup(NULL);
317
318 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
319 KASSERT(pr != NULL);
320
321 ip_pktq = pktq_create(IFQ_MAXLEN, ipintr, NULL);
322 KASSERT(ip_pktq != NULL);
323
324 for (u_int i = 0; i < IPPROTO_MAX; i++) {
325 ip_protox[i] = pr - inetsw;
326 }
327 for (pr = inetdomain.dom_protosw;
328 pr < inetdomain.dom_protoswNPROTOSW; pr++)
329 if (pr->pr_domain->dom_family == PF_INET &&
330 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
331 ip_protox[pr->pr_protocol] = pr - inetsw;
332
333 ip_reass_init();
334
335 ip_ids = ip_id_init();
336 ip_id = time_uptime & 0xfffff;
337
338 ip_mtudisc_timeout_q = rt_timer_queue_create(ip_mtudisc_timeout);
339#ifdef GATEWAY
340 ipflow_init();
341#endif
342
343 /* Register our Packet Filter hook. */
344 inet_pfil_hook = pfil_head_create(PFIL_TYPE_AF, (void *)AF_INET);
345 KASSERT(inet_pfil_hook != NULL);
346
347#ifdef MBUFTRACE
348 MOWNER_ATTACH(&ip_tx_mowner);
349 MOWNER_ATTACH(&ip_rx_mowner);
350#endif /* MBUFTRACE */
351
352 ipstat_percpu = percpu_alloc(sizeof(uint64_t) * IP_NSTATS);
353
354 ipforward_rt_percpu = percpu_alloc(sizeof(struct route));
355 if (ipforward_rt_percpu == NULL)
356 panic("failed to allocate ipforward_rt_percpu");
357}
358
359static struct in_ifaddr *
360ip_match_our_address(struct ifnet *ifp, struct ip *ip, int *downmatch)
361{
362 struct in_ifaddr *ia = NULL;
363 int checkif;
364
365 /*
366 * Enable a consistency check between the destination address
367 * and the arrival interface for a unicast packet (the RFC 1122
368 * strong ES model) if IP forwarding is disabled and the packet
369 * is not locally generated.
370 *
371 * XXX - Checking also should be disabled if the destination
372 * address is ipnat'ed to a different interface.
373 *
374 * XXX - Checking is incompatible with IP aliases added
375 * to the loopback interface instead of the interface where
376 * the packets are received.
377 *
378 * XXX - We need to add a per ifaddr flag for this so that
379 * we get finer grain control.
380 */
381 checkif = ip_checkinterface && (ipforwarding == 0) &&
382 (ifp->if_flags & IFF_LOOPBACK) == 0;
383
384 IN_ADDRHASH_READER_FOREACH(ia, ip->ip_dst.s_addr) {
385 if (in_hosteq(ia->ia_addr.sin_addr, ip->ip_dst)) {
386 if (ia->ia4_flags & IN_IFF_NOTREADY)
387 continue;
388 if (checkif && ia->ia_ifp != ifp)
389 continue;
390 if ((ia->ia_ifp->if_flags & IFF_UP) != 0 &&
391 (ia->ia4_flags & IN_IFF_DETACHED) == 0)
392 break;
393 else
394 (*downmatch)++;
395 }
396 }
397
398 return ia;
399}
400
401static struct in_ifaddr *
402ip_match_our_address_broadcast(struct ifnet *ifp, struct ip *ip)
403{
404 struct in_ifaddr *ia = NULL;
405 struct ifaddr *ifa;
406
407 IFADDR_READER_FOREACH(ifa, ifp) {
408 if (ifa->ifa_addr->sa_family != AF_INET)
409 continue;
410 ia = ifatoia(ifa);
411 if (ia->ia4_flags & (IN_IFF_NOTREADY | IN_IFF_DETACHED))
412 continue;
413 if (in_hosteq(ip->ip_dst, ia->ia_broadaddr.sin_addr) ||
414 in_hosteq(ip->ip_dst, ia->ia_netbroadcast) ||
415 /*
416 * Look for all-0's host part (old broadcast addr),
417 * either for subnet or net.
418 */
419 ip->ip_dst.s_addr == ia->ia_subnet ||
420 ip->ip_dst.s_addr == ia->ia_net)
421 goto matched;
422 /*
423 * An interface with IP address zero accepts
424 * all packets that arrive on that interface.
425 */
426 if (in_nullhost(ia->ia_addr.sin_addr))
427 goto matched;
428 }
429 ia = NULL;
430
431matched:
432 return ia;
433}
434
435/*
436 * IP software interrupt routine.
437 */
438static void
439ipintr(void *arg __unused)
440{
441 struct mbuf *m;
442
443 KASSERT(cpu_softintr_p());
444
445#ifndef NET_MPSAFE
446 mutex_enter(softnet_lock);
447#endif
448 while ((m = pktq_dequeue(ip_pktq)) != NULL) {
449 ip_input(m);
450 }
451#ifndef NET_MPSAFE
452 mutex_exit(softnet_lock);
453#endif
454}
455
456/*
457 * IP input routine. Checksum and byte swap header. If fragmented
458 * try to reassemble. Process options. Pass to next level.
459 */
460static void
461ip_input(struct mbuf *m)
462{
463 struct ip *ip = NULL;
464 struct in_ifaddr *ia = NULL;
465 int hlen = 0, len;
466 int downmatch;
467 int srcrt = 0;
468 ifnet_t *ifp;
469 struct psref psref;
470 int s;
471
472 KASSERTMSG(cpu_softintr_p(), "ip_input: not in the software "
473 "interrupt handler; synchronization assumptions violated");
474
475 MCLAIM(m, &ip_rx_mowner);
476 KASSERT((m->m_flags & M_PKTHDR) != 0);
477
478 ifp = m_get_rcvif_psref(m, &psref);
479 if (__predict_false(ifp == NULL))
480 goto out;
481
482 /*
483 * If no IP addresses have been set yet but the interfaces
484 * are receiving, can't do anything with incoming packets yet.
485 * Note: we pre-check without locks held.
486 */
487 if (IN_ADDRLIST_READER_EMPTY())
488 goto out;
489 IP_STATINC(IP_STAT_TOTAL);
490
491 /*
492 * If the IP header is not aligned, slurp it up into a new
493 * mbuf with space for link headers, in the event we forward
494 * it. Otherwise, if it is aligned, make sure the entire
495 * base IP header is in the first mbuf of the chain.
496 */
497 if (IP_HDR_ALIGNED_P(mtod(m, void *)) == 0) {
498 if ((m = m_copyup(m, sizeof(struct ip),
499 (max_linkhdr + 3) & ~3)) == NULL) {
500 /* XXXJRT new stat, please */
501 IP_STATINC(IP_STAT_TOOSMALL);
502 goto out;
503 }
504 } else if (__predict_false(m->m_len < sizeof (struct ip))) {
505 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
506 IP_STATINC(IP_STAT_TOOSMALL);
507 goto out;
508 }
509 }
510 ip = mtod(m, struct ip *);
511 if (ip->ip_v != IPVERSION) {
512 IP_STATINC(IP_STAT_BADVERS);
513 goto out;
514 }
515 hlen = ip->ip_hl << 2;
516 if (hlen < sizeof(struct ip)) { /* minimum header length */
517 IP_STATINC(IP_STAT_BADHLEN);
518 goto out;
519 }
520 if (hlen > m->m_len) {
521 if ((m = m_pullup(m, hlen)) == NULL) {
522 IP_STATINC(IP_STAT_BADHLEN);
523 goto out;
524 }
525 ip = mtod(m, struct ip *);
526 }
527
528 /*
529 * RFC1122: packets with a multicast source address are
530 * not allowed.
531 */
532 if (IN_MULTICAST(ip->ip_src.s_addr)) {
533 IP_STATINC(IP_STAT_BADADDR);
534 goto out;
535 }
536
537 /* 127/8 must not appear on wire - RFC1122 */
538 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
539 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
540 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
541 IP_STATINC(IP_STAT_BADADDR);
542 goto out;
543 }
544 }
545
546 switch (m->m_pkthdr.csum_flags &
547 ((ifp->if_csum_flags_rx & M_CSUM_IPv4) |
548 M_CSUM_IPv4_BAD)) {
549 case M_CSUM_IPv4|M_CSUM_IPv4_BAD:
550 INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad);
551 IP_STATINC(IP_STAT_BADSUM);
552 goto out;
553
554 case M_CSUM_IPv4:
555 /* Checksum was okay. */
556 INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok);
557 break;
558
559 default:
560 /*
561 * Must compute it ourselves. Maybe skip checksum on
562 * loopback interfaces.
563 */
564 if (__predict_true(!(ifp->if_flags & IFF_LOOPBACK) ||
565 ip_do_loopback_cksum)) {
566 INET_CSUM_COUNTER_INCR(&ip_swcsum);
567 if (in_cksum(m, hlen) != 0) {
568 IP_STATINC(IP_STAT_BADSUM);
569 goto out;
570 }
571 }
572 break;
573 }
574
575 /* Retrieve the packet length. */
576 len = ntohs(ip->ip_len);
577
578 /*
579 * Check for additional length bogosity
580 */
581 if (len < hlen) {
582 IP_STATINC(IP_STAT_BADLEN);
583 goto out;
584 }
585
586 /*
587 * Check that the amount of data in the buffers
588 * is as at least much as the IP header would have us expect.
589 * Trim mbufs if longer than we expect.
590 * Drop packet if shorter than we expect.
591 */
592 if (m->m_pkthdr.len < len) {
593 IP_STATINC(IP_STAT_TOOSHORT);
594 goto out;
595 }
596 if (m->m_pkthdr.len > len) {
597 if (m->m_len == m->m_pkthdr.len) {
598 m->m_len = len;
599 m->m_pkthdr.len = len;
600 } else
601 m_adj(m, len - m->m_pkthdr.len);
602 }
603
604 /*
605 * Assume that we can create a fast-forward IP flow entry
606 * based on this packet.
607 */
608 m->m_flags |= M_CANFASTFWD;
609
610 /*
611 * Run through list of hooks for input packets. If there are any
612 * filters which require that additional packets in the flow are
613 * not fast-forwarded, they must clear the M_CANFASTFWD flag.
614 * Note that filters must _never_ set this flag, as another filter
615 * in the list may have previously cleared it.
616 */
617#if defined(IPSEC)
618 if (!ipsec_used || !ipsec_indone(m))
619#else
620 if (1)
621#endif
622 {
623 struct in_addr odst = ip->ip_dst;
624 bool freed;
625
626 freed = pfil_run_hooks(inet_pfil_hook, &m, ifp, PFIL_IN) != 0;
627 if (freed || m == NULL) {
628 m = NULL;
629 goto out;
630 }
631 ip = mtod(m, struct ip *);
632 hlen = ip->ip_hl << 2;
633
634 /*
635 * XXX The setting of "srcrt" here is to prevent ip_forward()
636 * from generating ICMP redirects for packets that have
637 * been redirected by a hook back out on to the same LAN that
638 * they came from and is not an indication that the packet
639 * is being inffluenced by source routing options. This
640 * allows things like
641 * "rdr tlp0 0/0 port 80 -> 1.1.1.200 3128 tcp"
642 * where tlp0 is both on the 1.1.1.0/24 network and is the
643 * default route for hosts on 1.1.1.0/24. Of course this
644 * also requires a "map tlp0 ..." to complete the story.
645 * One might argue whether or not this kind of network config.
646 * should be supported in this manner...
647 */
648 srcrt = (odst.s_addr != ip->ip_dst.s_addr);
649 }
650
651#ifdef ALTQ
652 /* XXX Temporary until ALTQ is changed to use a pfil hook */
653 if (altq_input) {
654 SOFTNET_LOCK();
655 if ((*altq_input)(m, AF_INET) == 0) {
656 /* Packet dropped by traffic conditioner. */
657 SOFTNET_UNLOCK();
658 m = NULL;
659 goto out;
660 }
661 SOFTNET_UNLOCK();
662 }
663#endif
664
665 /*
666 * Process options and, if not destined for us,
667 * ship it on. ip_dooptions returns 1 when an
668 * error was detected (causing an icmp message
669 * to be sent and the original packet to be freed).
670 */
671 ip_nhops = 0; /* for source routed packets */
672 if (hlen > sizeof (struct ip) && ip_dooptions(m)) {
673 m = NULL;
674 goto out;
675 }
676
677 /*
678 * Check our list of addresses, to see if the packet is for us.
679 *
680 * Traditional 4.4BSD did not consult IFF_UP at all.
681 * The behavior here is to treat addresses on !IFF_UP interface
682 * or IN_IFF_NOTREADY addresses as not mine.
683 */
684 downmatch = 0;
685 s = pserialize_read_enter();
686 ia = ip_match_our_address(ifp, ip, &downmatch);
687 if (ia != NULL) {
688 pserialize_read_exit(s);
689 goto ours;
690 }
691
692 if (ifp->if_flags & IFF_BROADCAST) {
693 ia = ip_match_our_address_broadcast(ifp, ip);
694 if (ia != NULL) {
695 pserialize_read_exit(s);
696 goto ours;
697 }
698 }
699 pserialize_read_exit(s);
700
701 if (IN_MULTICAST(ip->ip_dst.s_addr)) {
702#ifdef MROUTING
703 extern struct socket *ip_mrouter;
704
705 if (ip_mrouter) {
706 /*
707 * If we are acting as a multicast router, all
708 * incoming multicast packets are passed to the
709 * kernel-level multicast forwarding function.
710 * The packet is returned (relatively) intact; if
711 * ip_mforward() returns a non-zero value, the packet
712 * must be discarded, else it may be accepted below.
713 *
714 * (The IP ident field is put in the same byte order
715 * as expected when ip_mforward() is called from
716 * ip_output().)
717 */
718 SOFTNET_LOCK();
719 if (ip_mforward(m, ifp) != 0) {
720 SOFTNET_UNLOCK();
721 IP_STATINC(IP_STAT_CANTFORWARD);
722 goto out;
723 }
724 SOFTNET_UNLOCK();
725
726 /*
727 * The process-level routing demon needs to receive
728 * all multicast IGMP packets, whether or not this
729 * host belongs to their destination groups.
730 */
731 if (ip->ip_p == IPPROTO_IGMP) {
732 goto ours;
733 }
734 IP_STATINC(IP_STAT_CANTFORWARD);
735 }
736#endif
737 /*
738 * See if we belong to the destination multicast group on the
739 * arrival interface.
740 */
741 if (!in_multi_group(ip->ip_dst, ifp, 0)) {
742 IP_STATINC(IP_STAT_CANTFORWARD);
743 goto out;
744 }
745 goto ours;
746 }
747 if (ip->ip_dst.s_addr == INADDR_BROADCAST ||
748 in_nullhost(ip->ip_dst))
749 goto ours;
750
751 /*
752 * Not for us; forward if possible and desirable.
753 */
754 if (ipforwarding == 0) {
755 m_put_rcvif_psref(ifp, &psref);
756 IP_STATINC(IP_STAT_CANTFORWARD);
757 m_freem(m);
758 } else {
759 /*
760 * If ip_dst matched any of my address on !IFF_UP interface,
761 * and there's no IFF_UP interface that matches ip_dst,
762 * send icmp unreach. Forwarding it will result in in-kernel
763 * forwarding loop till TTL goes to 0.
764 */
765 if (downmatch) {
766 m_put_rcvif_psref(ifp, &psref);
767 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
768 IP_STATINC(IP_STAT_CANTFORWARD);
769 return;
770 }
771#ifdef IPSEC
772 /* Perform IPsec, if any. */
773 if (ipsec_used) {
774 SOFTNET_LOCK();
775 if (ipsec4_input(m, IP_FORWARDING |
776 (ip_directedbcast ? IP_ALLOWBROADCAST : 0)) != 0) {
777 SOFTNET_UNLOCK();
778 goto out;
779 }
780 SOFTNET_UNLOCK();
781 }
782#endif
783 ip_forward(m, srcrt, ifp);
784 m_put_rcvif_psref(ifp, &psref);
785 }
786 return;
787
788ours:
789 m_put_rcvif_psref(ifp, &psref);
790 ifp = NULL;
791
792 /*
793 * If offset or IP_MF are set, must reassemble.
794 */
795 if (ip->ip_off & ~htons(IP_DF|IP_RF)) {
796 /*
797 * Pass to IP reassembly mechanism.
798 */
799 if (ip_reass_packet(&m, ip) != 0) {
800 /* Failed; invalid fragment(s) or packet. */
801 goto out;
802 }
803 if (m == NULL) {
804 /* More fragments should come; silently return. */
805 goto out;
806 }
807 /*
808 * Reassembly is done, we have the final packet.
809 * Updated cached data in local variable(s).
810 */
811 ip = mtod(m, struct ip *);
812 hlen = ip->ip_hl << 2;
813 }
814
815#ifdef IPSEC
816 /*
817 * Enforce IPsec policy checking if we are seeing last header.
818 * Note that we do not visit this with protocols with PCB layer
819 * code - like UDP/TCP/raw IP.
820 */
821 if (ipsec_used &&
822 (inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) {
823 SOFTNET_LOCK();
824 if (ipsec4_input(m, 0) != 0) {
825 SOFTNET_UNLOCK();
826 goto out;
827 }
828 SOFTNET_UNLOCK();
829 }
830#endif
831
832 /*
833 * Switch out to protocol's input routine.
834 */
835#if IFA_STATS
836 if (ia && ip) {
837 struct in_ifaddr *_ia;
838 /*
839 * Keep a reference from ip_match_our_address with psref
840 * is expensive, so explore ia here again.
841 */
842 s = pserialize_read_enter();
843 _ia = in_get_ia(ip->ip_dst);
844 _ia->ia_ifa.ifa_data.ifad_inbytes += ntohs(ip->ip_len);
845 pserialize_read_exit(s);
846 }
847#endif
848 IP_STATINC(IP_STAT_DELIVERED);
849
850 const int off = hlen, nh = ip->ip_p;
851
852 SOFTNET_LOCK();
853 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh);
854 SOFTNET_UNLOCK();
855 return;
856
857out:
858 m_put_rcvif_psref(ifp, &psref);
859 if (m != NULL)
860 m_freem(m);
861}
862
863/*
864 * IP timer processing.
865 */
866void
867ip_slowtimo(void)
868{
869
870 mutex_enter(softnet_lock);
871 KERNEL_LOCK(1, NULL);
872
873 ip_reass_slowtimo();
874
875 KERNEL_UNLOCK_ONE(NULL);
876 mutex_exit(softnet_lock);
877}
878
879/*
880 * IP drain processing.
881 */
882void
883ip_drain(void)
884{
885
886 KERNEL_LOCK(1, NULL);
887 ip_reass_drain();
888 KERNEL_UNLOCK_ONE(NULL);
889}
890
891/*
892 * ip_dooptions: perform option processing on a datagram, possibly discarding
893 * it if bad options are encountered, or forwarding it if source-routed.
894 *
895 * => Returns true if packet has been forwarded/freed.
896 * => Returns false if the packet should be processed further.
897 */
898static bool
899ip_dooptions(struct mbuf *m)
900{
901 struct ip *ip = mtod(m, struct ip *);
902 u_char *cp, *cp0;
903 struct ip_timestamp *ipt;
904 struct in_ifaddr *ia;
905 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0;
906 struct in_addr dst;
907 n_time ntime;
908 struct ifaddr *ifa;
909 int s;
910
911 dst = ip->ip_dst;
912 cp = (u_char *)(ip + 1);
913 cnt = (ip->ip_hl << 2) - sizeof (struct ip);
914 for (; cnt > 0; cnt -= optlen, cp += optlen) {
915 opt = cp[IPOPT_OPTVAL];
916 if (opt == IPOPT_EOL)
917 break;
918 if (opt == IPOPT_NOP)
919 optlen = 1;
920 else {
921 if (cnt < IPOPT_OLEN + sizeof(*cp)) {
922 code = &cp[IPOPT_OLEN] - (u_char *)ip;
923 goto bad;
924 }
925 optlen = cp[IPOPT_OLEN];
926 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
927 code = &cp[IPOPT_OLEN] - (u_char *)ip;
928 goto bad;
929 }
930 }
931 switch (opt) {
932
933 default:
934 break;
935
936 /*
937 * Source routing with record.
938 * Find interface with current destination address.
939 * If none on this machine then drop if strictly routed,
940 * or do nothing if loosely routed.
941 * Record interface address and bring up next address
942 * component. If strictly routed make sure next
943 * address is on directly accessible net.
944 */
945 case IPOPT_LSRR:
946 case IPOPT_SSRR:
947 if (ip_allowsrcrt == 0) {
948 type = ICMP_UNREACH;
949 code = ICMP_UNREACH_NET_PROHIB;
950 goto bad;
951 }
952 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
953 code = &cp[IPOPT_OLEN] - (u_char *)ip;
954 goto bad;
955 }
956 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
957 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
958 goto bad;
959 }
960 ipaddr.sin_addr = ip->ip_dst;
961
962 s = pserialize_read_enter();
963 ifa = ifa_ifwithaddr(sintosa(&ipaddr));
964 if (ifa == NULL) {
965 pserialize_read_exit(s);
966 if (opt == IPOPT_SSRR) {
967 type = ICMP_UNREACH;
968 code = ICMP_UNREACH_SRCFAIL;
969 goto bad;
970 }
971 /*
972 * Loose routing, and not at next destination
973 * yet; nothing to do except forward.
974 */
975 break;
976 }
977 pserialize_read_exit(s);
978
979 off--; /* 0 origin */
980 if ((off + sizeof(struct in_addr)) > optlen) {
981 /*
982 * End of source route. Should be for us.
983 */
984 save_rte(cp, ip->ip_src);
985 break;
986 }
987 /*
988 * locate outgoing interface
989 */
990 memcpy((void *)&ipaddr.sin_addr, (void *)(cp + off),
991 sizeof(ipaddr.sin_addr));
992 s = pserialize_read_enter();
993 if (opt == IPOPT_SSRR) {
994 ifa = ifa_ifwithladdr(sintosa(&ipaddr));
995 if (ifa != NULL)
996 ia = ifatoia(ifa);
997 else
998 ia = NULL;
999 } else {
1000 ia = ip_rtaddr(ipaddr.sin_addr);
1001 }
1002 if (ia == NULL) {
1003 type = ICMP_UNREACH;
1004 code = ICMP_UNREACH_SRCFAIL;
1005 pserialize_read_exit(s);
1006 goto bad;
1007 }
1008 ip->ip_dst = ipaddr.sin_addr;
1009 bcopy((void *)&ia->ia_addr.sin_addr,
1010 (void *)(cp + off), sizeof(struct in_addr));
1011 pserialize_read_exit(s);
1012 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1013 /*
1014 * Let ip_intr's mcast routing check handle mcast pkts
1015 */
1016 forward = !IN_MULTICAST(ip->ip_dst.s_addr);
1017 break;
1018
1019 case IPOPT_RR:
1020 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1021 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1022 goto bad;
1023 }
1024 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1025 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1026 goto bad;
1027 }
1028 /*
1029 * If no space remains, ignore.
1030 */
1031 off--; /* 0 origin */
1032 if ((off + sizeof(struct in_addr)) > optlen)
1033 break;
1034 memcpy((void *)&ipaddr.sin_addr, (void *)(&ip->ip_dst),
1035 sizeof(ipaddr.sin_addr));
1036 /*
1037 * locate outgoing interface; if we're the destination,
1038 * use the incoming interface (should be same).
1039 */
1040 s = pserialize_read_enter();
1041 ifa = ifa_ifwithaddr(sintosa(&ipaddr));
1042 if (ifa == NULL) {
1043 ia = ip_rtaddr(ipaddr.sin_addr);
1044 if (ia == NULL) {
1045 pserialize_read_exit(s);
1046 type = ICMP_UNREACH;
1047 code = ICMP_UNREACH_HOST;
1048 goto bad;
1049 }
1050 } else {
1051 ia = ifatoia(ifa);
1052 }
1053 bcopy((void *)&ia->ia_addr.sin_addr,
1054 (void *)(cp + off), sizeof(struct in_addr));
1055 pserialize_read_exit(s);
1056 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1057 break;
1058
1059 case IPOPT_TS:
1060 code = cp - (u_char *)ip;
1061 ipt = (struct ip_timestamp *)cp;
1062 if (ipt->ipt_len < 4 || ipt->ipt_len > 40) {
1063 code = (u_char *)&ipt->ipt_len - (u_char *)ip;
1064 goto bad;
1065 }
1066 if (ipt->ipt_ptr < 5) {
1067 code = (u_char *)&ipt->ipt_ptr - (u_char *)ip;
1068 goto bad;
1069 }
1070 if (ipt->ipt_ptr > ipt->ipt_len - sizeof (int32_t)) {
1071 if (++ipt->ipt_oflw == 0) {
1072 code = (u_char *)&ipt->ipt_ptr -
1073 (u_char *)ip;
1074 goto bad;
1075 }
1076 break;
1077 }
1078 cp0 = (cp + ipt->ipt_ptr - 1);
1079 switch (ipt->ipt_flg) {
1080
1081 case IPOPT_TS_TSONLY:
1082 break;
1083
1084 case IPOPT_TS_TSANDADDR: {
1085 struct ifnet *rcvif;
1086 int _s, _ss;
1087
1088 if (ipt->ipt_ptr - 1 + sizeof(n_time) +
1089 sizeof(struct in_addr) > ipt->ipt_len) {
1090 code = (u_char *)&ipt->ipt_ptr -
1091 (u_char *)ip;
1092 goto bad;
1093 }
1094 ipaddr.sin_addr = dst;
1095 _ss = pserialize_read_enter();
1096 rcvif = m_get_rcvif(m, &_s);
1097 ifa = ifaof_ifpforaddr(sintosa(&ipaddr), rcvif);
1098 m_put_rcvif(rcvif, &_s);
1099 if (ifa == NULL) {
1100 pserialize_read_exit(_ss);
1101 break;
1102 }
1103 ia = ifatoia(ifa);
1104 bcopy(&ia->ia_addr.sin_addr,
1105 cp0, sizeof(struct in_addr));
1106 pserialize_read_exit(_ss);
1107 ipt->ipt_ptr += sizeof(struct in_addr);
1108 break;
1109 }
1110
1111 case IPOPT_TS_PRESPEC:
1112 if (ipt->ipt_ptr - 1 + sizeof(n_time) +
1113 sizeof(struct in_addr) > ipt->ipt_len) {
1114 code = (u_char *)&ipt->ipt_ptr -
1115 (u_char *)ip;
1116 goto bad;
1117 }
1118 memcpy(&ipaddr.sin_addr, cp0,
1119 sizeof(struct in_addr));
1120 s = pserialize_read_enter();
1121 ifa = ifa_ifwithaddr(sintosa(&ipaddr));
1122 if (ifa == NULL) {
1123 pserialize_read_exit(s);
1124 continue;
1125 }
1126 pserialize_read_exit(s);
1127 ipt->ipt_ptr += sizeof(struct in_addr);
1128 break;
1129
1130 default:
1131 /* XXX can't take &ipt->ipt_flg */
1132 code = (u_char *)&ipt->ipt_ptr -
1133 (u_char *)ip + 1;
1134 goto bad;
1135 }
1136 ntime = iptime();
1137 cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */
1138 memmove((char *)cp + ipt->ipt_ptr - 1, cp0,
1139 sizeof(n_time));
1140 ipt->ipt_ptr += sizeof(n_time);
1141 }
1142 }
1143 if (forward) {
1144 struct ifnet *rcvif;
1145 struct psref _psref;
1146
1147 if (ip_forwsrcrt == 0) {
1148 type = ICMP_UNREACH;
1149 code = ICMP_UNREACH_SRCFAIL;
1150 goto bad;
1151 }
1152
1153 rcvif = m_get_rcvif_psref(m, &_psref);
1154 if (__predict_false(rcvif == NULL)) {
1155 type = ICMP_UNREACH;
1156 code = ICMP_UNREACH_HOST;
1157 goto bad;
1158 }
1159 ip_forward(m, 1, rcvif);
1160 m_put_rcvif_psref(rcvif, &_psref);
1161 return true;
1162 }
1163 return false;
1164bad:
1165 icmp_error(m, type, code, 0, 0);
1166 IP_STATINC(IP_STAT_BADOPTIONS);
1167 return true;
1168}
1169
1170/*
1171 * ip_rtaddr: given address of next destination (final or next hop),
1172 * return internet address info of interface to be used to get there.
1173 */
1174static struct in_ifaddr *
1175ip_rtaddr(struct in_addr dst)
1176{
1177 struct rtentry *rt;
1178 union {
1179 struct sockaddr dst;
1180 struct sockaddr_in dst4;
1181 } u;
1182 struct route *ro;
1183
1184 sockaddr_in_init(&u.dst4, &dst, 0);
1185
1186 SOFTNET_LOCK();
1187 ro = percpu_getref(ipforward_rt_percpu);
1188 rt = rtcache_lookup(ro, &u.dst);
1189 percpu_putref(ipforward_rt_percpu);
1190 SOFTNET_UNLOCK();
1191 if (rt == NULL)
1192 return NULL;
1193
1194 return ifatoia(rt->rt_ifa);
1195}
1196
1197/*
1198 * save_rte: save incoming source route for use in replies, to be picked
1199 * up later by ip_srcroute if the receiver is interested.
1200 */
1201static void
1202save_rte(u_char *option, struct in_addr dst)
1203{
1204 unsigned olen;
1205
1206 olen = option[IPOPT_OLEN];
1207 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst)))
1208 return;
1209 memcpy((void *)ip_srcrt.srcopt, (void *)option, olen);
1210 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1211 ip_srcrt.dst = dst;
1212}
1213
1214/*
1215 * Retrieve incoming source route for use in replies,
1216 * in the same form used by setsockopt.
1217 * The first hop is placed before the options, will be removed later.
1218 */
1219struct mbuf *
1220ip_srcroute(void)
1221{
1222 struct in_addr *p, *q;
1223 struct mbuf *m;
1224
1225 if (ip_nhops == 0)
1226 return NULL;
1227 m = m_get(M_DONTWAIT, MT_SOOPTS);
1228 if (m == 0)
1229 return NULL;
1230
1231 MCLAIM(m, &inetdomain.dom_mowner);
1232#define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt))
1233
1234 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
1235 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) +
1236 OPTSIZ;
1237
1238 /*
1239 * First save first hop for return route
1240 */
1241 p = &ip_srcrt.route[ip_nhops - 1];
1242 *(mtod(m, struct in_addr *)) = *p--;
1243
1244 /*
1245 * Copy option fields and padding (nop) to mbuf.
1246 */
1247 ip_srcrt.nop = IPOPT_NOP;
1248 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
1249 memmove(mtod(m, char *) + sizeof(struct in_addr), &ip_srcrt.nop,
1250 OPTSIZ);
1251 q = (struct in_addr *)(mtod(m, char *) +
1252 sizeof(struct in_addr) + OPTSIZ);
1253#undef OPTSIZ
1254 /*
1255 * Record return path as an IP source route,
1256 * reversing the path (pointers are now aligned).
1257 */
1258 while (p >= ip_srcrt.route) {
1259 *q++ = *p--;
1260 }
1261 /*
1262 * Last hop goes to final destination.
1263 */
1264 *q = ip_srcrt.dst;
1265 return (m);
1266}
1267
1268const int inetctlerrmap[PRC_NCMDS] = {
1269 [PRC_MSGSIZE] = EMSGSIZE,
1270 [PRC_HOSTDEAD] = EHOSTDOWN,
1271 [PRC_HOSTUNREACH] = EHOSTUNREACH,
1272 [PRC_UNREACH_NET] = EHOSTUNREACH,
1273 [PRC_UNREACH_HOST] = EHOSTUNREACH,
1274 [PRC_UNREACH_PROTOCOL] = ECONNREFUSED,
1275 [PRC_UNREACH_PORT] = ECONNREFUSED,
1276 [PRC_UNREACH_SRCFAIL] = EHOSTUNREACH,
1277 [PRC_PARAMPROB] = ENOPROTOOPT,
1278};
1279
1280void
1281ip_fasttimo(void)
1282{
1283 if (ip_drainwanted) {
1284 ip_drain();
1285 ip_drainwanted = 0;
1286 }
1287}
1288
1289void
1290ip_drainstub(void)
1291{
1292 ip_drainwanted = 1;
1293}
1294
1295/*
1296 * Forward a packet. If some error occurs return the sender
1297 * an icmp packet. Note we can't always generate a meaningful
1298 * icmp message because icmp doesn't have a large enough repertoire
1299 * of codes and types.
1300 *
1301 * If not forwarding, just drop the packet. This could be confusing
1302 * if ipforwarding was zero but some routing protocol was advancing
1303 * us as a gateway to somewhere. However, we must let the routing
1304 * protocol deal with that.
1305 *
1306 * The srcrt parameter indicates whether the packet is being forwarded
1307 * via a source route.
1308 */
1309static void
1310ip_forward(struct mbuf *m, int srcrt, struct ifnet *rcvif)
1311{
1312 struct ip *ip = mtod(m, struct ip *);
1313 struct rtentry *rt;
1314 int error, type = 0, code = 0, destmtu = 0;
1315 struct mbuf *mcopy;
1316 n_long dest;
1317 union {
1318 struct sockaddr dst;
1319 struct sockaddr_in dst4;
1320 } u;
1321 uint64_t *ips;
1322 struct route *ro;
1323
1324 KASSERTMSG(cpu_softintr_p(), "ip_forward: not in the software "
1325 "interrupt handler; synchronization assumptions violated");
1326
1327 /*
1328 * We are now in the output path.
1329 */
1330 MCLAIM(m, &ip_tx_mowner);
1331
1332 /*
1333 * Clear any in-bound checksum flags for this packet.
1334 */
1335 m->m_pkthdr.csum_flags = 0;
1336
1337 dest = 0;
1338 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
1339 IP_STATINC(IP_STAT_CANTFORWARD);
1340 m_freem(m);
1341 return;
1342 }
1343
1344 if (ip->ip_ttl <= IPTTLDEC) {
1345 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0);
1346 return;
1347 }
1348
1349 sockaddr_in_init(&u.dst4, &ip->ip_dst, 0);
1350
1351 ro = percpu_getref(ipforward_rt_percpu);
1352 if ((rt = rtcache_lookup(ro, &u.dst)) == NULL) {
1353 percpu_putref(ipforward_rt_percpu);
1354 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_NET, dest, 0);
1355 return;
1356 }
1357
1358 /*
1359 * Save at most 68 bytes of the packet in case
1360 * we need to generate an ICMP message to the src.
1361 * Pullup to avoid sharing mbuf cluster between m and mcopy.
1362 */
1363 mcopy = m_copym(m, 0, imin(ntohs(ip->ip_len), 68), M_DONTWAIT);
1364 if (mcopy)
1365 mcopy = m_pullup(mcopy, ip->ip_hl << 2);
1366
1367 ip->ip_ttl -= IPTTLDEC;
1368
1369 /*
1370 * If forwarding packet using same interface that it came in on,
1371 * perhaps should send a redirect to sender to shortcut a hop.
1372 * Only send redirect if source is sending directly to us,
1373 * and if packet was not source routed (or has any options).
1374 * Also, don't send redirect if forwarding using a default route
1375 * or a route modified by a redirect.
1376 */
1377 if (rt->rt_ifp == rcvif &&
1378 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1379 !in_nullhost(satocsin(rt_getkey(rt))->sin_addr) &&
1380 ipsendredirects && !srcrt) {
1381 if (rt->rt_ifa &&
1382 (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) ==
1383 ifatoia(rt->rt_ifa)->ia_subnet) {
1384 if (rt->rt_flags & RTF_GATEWAY)
1385 dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
1386 else
1387 dest = ip->ip_dst.s_addr;
1388 /*
1389 * Router requirements says to only send host
1390 * redirects.
1391 */
1392 type = ICMP_REDIRECT;
1393 code = ICMP_REDIRECT_HOST;
1394 }
1395 }
1396
1397 error = ip_output(m, NULL, ro,
1398 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)),
1399 NULL, NULL);
1400
1401 if (error) {
1402 IP_STATINC(IP_STAT_CANTFORWARD);
1403 goto error;
1404 }
1405
1406 ips = IP_STAT_GETREF();
1407 ips[IP_STAT_FORWARD]++;
1408
1409 if (type) {
1410 ips[IP_STAT_REDIRECTSENT]++;
1411 IP_STAT_PUTREF();
1412 goto redirect;
1413 }
1414
1415 IP_STAT_PUTREF();
1416 if (mcopy) {
1417#ifdef GATEWAY
1418 if (mcopy->m_flags & M_CANFASTFWD)
1419 ipflow_create(ro, mcopy);
1420#endif
1421 m_freem(mcopy);
1422 }
1423
1424 percpu_putref(ipforward_rt_percpu);
1425 return;
1426
1427redirect:
1428error:
1429 if (mcopy == NULL) {
1430 percpu_putref(ipforward_rt_percpu);
1431 return;
1432 }
1433
1434 switch (error) {
1435
1436 case 0: /* forwarded, but need redirect */
1437 /* type, code set above */
1438 break;
1439
1440 case ENETUNREACH: /* shouldn't happen, checked above */
1441 case EHOSTUNREACH:
1442 case ENETDOWN:
1443 case EHOSTDOWN:
1444 default:
1445 type = ICMP_UNREACH;
1446 code = ICMP_UNREACH_HOST;
1447 break;
1448
1449 case EMSGSIZE:
1450 type = ICMP_UNREACH;
1451 code = ICMP_UNREACH_NEEDFRAG;
1452
1453 if ((rt = rtcache_validate(ro)) != NULL)
1454 destmtu = rt->rt_ifp->if_mtu;
1455#ifdef IPSEC
1456 if (ipsec_used)
1457 (void)ipsec4_forward(mcopy, &destmtu);
1458#endif
1459 IP_STATINC(IP_STAT_CANTFRAG);
1460 break;
1461
1462 case ENOBUFS:
1463 /*
1464 * Do not generate ICMP_SOURCEQUENCH as required in RFC 1812,
1465 * Requirements for IP Version 4 Routers. Source quench can
1466 * big problem under DoS attacks or if the underlying
1467 * interface is rate-limited.
1468 */
1469 if (mcopy)
1470 m_freem(mcopy);
1471 percpu_putref(ipforward_rt_percpu);
1472 return;
1473 }
1474 icmp_error(mcopy, type, code, dest, destmtu);
1475 percpu_putref(ipforward_rt_percpu);
1476}
1477
1478void
1479ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1480 struct mbuf *m)
1481{
1482 struct socket *so = inp->inp_socket;
1483 ifnet_t *ifp;
1484 int inpflags = inp->inp_flags;
1485 struct psref psref;
1486
1487 ifp = m_get_rcvif_psref(m, &psref);
1488 if (__predict_false(ifp == NULL))
1489 return; /* XXX should report error? */
1490
1491 if (so->so_options & SO_TIMESTAMP
1492#ifdef SO_OTIMESTAMP
1493 || so->so_options & SO_OTIMESTAMP
1494#endif
1495 ) {
1496 struct timeval tv;
1497
1498 microtime(&tv);
1499#ifdef SO_OTIMESTAMP
1500 if (so->so_options & SO_OTIMESTAMP) {
1501 struct timeval50 tv50;
1502 timeval_to_timeval50(&tv, &tv50);
1503 *mp = sbcreatecontrol((void *) &tv50, sizeof(tv50),
1504 SCM_OTIMESTAMP, SOL_SOCKET);
1505 } else
1506#endif
1507 *mp = sbcreatecontrol((void *) &tv, sizeof(tv),
1508 SCM_TIMESTAMP, SOL_SOCKET);
1509 if (*mp)
1510 mp = &(*mp)->m_next;
1511 }
1512 if (inpflags & INP_RECVDSTADDR) {
1513 *mp = sbcreatecontrol((void *) &ip->ip_dst,
1514 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1515 if (*mp)
1516 mp = &(*mp)->m_next;
1517 }
1518 if (inpflags & INP_RECVPKTINFO) {
1519 struct in_pktinfo ipi;
1520 ipi.ipi_addr = ip->ip_src;
1521 ipi.ipi_ifindex = ifp->if_index;
1522 *mp = sbcreatecontrol((void *) &ipi,
1523 sizeof(ipi), IP_RECVPKTINFO, IPPROTO_IP);
1524 if (*mp)
1525 mp = &(*mp)->m_next;
1526 }
1527 if (inpflags & INP_PKTINFO) {
1528 struct in_pktinfo ipi;
1529 ipi.ipi_addr = ip->ip_dst;
1530 ipi.ipi_ifindex = ifp->if_index;
1531 *mp = sbcreatecontrol((void *) &ipi,
1532 sizeof(ipi), IP_PKTINFO, IPPROTO_IP);
1533 if (*mp)
1534 mp = &(*mp)->m_next;
1535 }
1536 if (inpflags & INP_RECVIF) {
1537 struct sockaddr_dl sdl;
1538
1539 sockaddr_dl_init(&sdl, sizeof(sdl), ifp->if_index, 0, NULL, 0,
1540 NULL, 0);
1541 *mp = sbcreatecontrol(&sdl, sdl.sdl_len, IP_RECVIF, IPPROTO_IP);
1542 if (*mp)
1543 mp = &(*mp)->m_next;
1544 }
1545 if (inpflags & INP_RECVTTL) {
1546 *mp = sbcreatecontrol((void *) &ip->ip_ttl,
1547 sizeof(uint8_t), IP_RECVTTL, IPPROTO_IP);
1548 if (*mp)
1549 mp = &(*mp)->m_next;
1550 }
1551 m_put_rcvif_psref(ifp, &psref);
1552}
1553
1554/*
1555 * sysctl helper routine for net.inet.ip.forwsrcrt.
1556 */
1557static int
1558sysctl_net_inet_ip_forwsrcrt(SYSCTLFN_ARGS)
1559{
1560 int error, tmp;
1561 struct sysctlnode node;
1562
1563 node = *rnode;
1564 tmp = ip_forwsrcrt;
1565 node.sysctl_data = &tmp;
1566 error = sysctl_lookup(SYSCTLFN_CALL(&node));
1567 if (error || newp == NULL)
1568 return (error);
1569
1570 error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_FORWSRCRT,
1571 0, NULL, NULL, NULL);
1572 if (error)
1573 return (error);
1574
1575 ip_forwsrcrt = tmp;
1576
1577 return (0);
1578}
1579
1580/*
1581 * sysctl helper routine for net.inet.ip.mtudisctimeout. checks the
1582 * range of the new value and tweaks timers if it changes.
1583 */
1584static int
1585sysctl_net_inet_ip_pmtudto(SYSCTLFN_ARGS)
1586{
1587 int error, tmp;
1588 struct sysctlnode node;
1589
1590 node = *rnode;
1591 tmp = ip_mtudisc_timeout;
1592 node.sysctl_data = &tmp;
1593 error = sysctl_lookup(SYSCTLFN_CALL(&node));
1594 if (error || newp == NULL)
1595 return (error);
1596 if (tmp < 0)
1597 return (EINVAL);
1598
1599 mutex_enter(softnet_lock);
1600
1601 ip_mtudisc_timeout = tmp;
1602 rt_timer_queue_change(ip_mtudisc_timeout_q, ip_mtudisc_timeout);
1603
1604 mutex_exit(softnet_lock);
1605
1606 return (0);
1607}
1608
1609static int
1610sysctl_net_inet_ip_stats(SYSCTLFN_ARGS)
1611{
1612
1613 return (NETSTAT_SYSCTL(ipstat_percpu, IP_NSTATS));
1614}
1615
1616static void
1617sysctl_net_inet_ip_setup(struct sysctllog **clog)
1618{
1619 sysctl_createv(clog, 0, NULL, NULL,
1620 CTLFLAG_PERMANENT,
1621 CTLTYPE_NODE, "inet",
1622 SYSCTL_DESCR("PF_INET related settings"),
1623 NULL, 0, NULL, 0,
1624 CTL_NET, PF_INET, CTL_EOL);
1625 sysctl_createv(clog, 0, NULL, NULL,
1626 CTLFLAG_PERMANENT,
1627 CTLTYPE_NODE, "ip",
1628 SYSCTL_DESCR("IPv4 related settings"),
1629 NULL, 0, NULL, 0,
1630 CTL_NET, PF_INET, IPPROTO_IP, CTL_EOL);
1631
1632 sysctl_createv(clog, 0, NULL, NULL,
1633 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1634 CTLTYPE_INT, "forwarding",
1635 SYSCTL_DESCR("Enable forwarding of INET datagrams"),
1636 NULL, 0, &ipforwarding, 0,
1637 CTL_NET, PF_INET, IPPROTO_IP,
1638 IPCTL_FORWARDING, CTL_EOL);
1639 sysctl_createv(clog, 0, NULL, NULL,
1640 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1641 CTLTYPE_INT, "redirect",
1642 SYSCTL_DESCR("Enable sending of ICMP redirect messages"),
1643 NULL, 0, &ipsendredirects, 0,
1644 CTL_NET, PF_INET, IPPROTO_IP,
1645 IPCTL_SENDREDIRECTS, CTL_EOL);
1646 sysctl_createv(clog, 0, NULL, NULL,
1647 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1648 CTLTYPE_INT, "ttl",
1649 SYSCTL_DESCR("Default TTL for an INET datagram"),
1650 NULL, 0, &ip_defttl, 0,
1651 CTL_NET, PF_INET, IPPROTO_IP,
1652 IPCTL_DEFTTL, CTL_EOL);
1653#ifdef IPCTL_DEFMTU
1654 sysctl_createv(clog, 0, NULL, NULL,
1655 CTLFLAG_PERMANENT /* |CTLFLAG_READWRITE? */,
1656 CTLTYPE_INT, "mtu",
1657 SYSCTL_DESCR("Default MTA for an INET route"),
1658 NULL, 0, &ip_mtu, 0,
1659 CTL_NET, PF_INET, IPPROTO_IP,
1660 IPCTL_DEFMTU, CTL_EOL);
1661#endif /* IPCTL_DEFMTU */
1662 sysctl_createv(clog, 0, NULL, NULL,
1663 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1664 CTLTYPE_INT, "forwsrcrt",
1665 SYSCTL_DESCR("Enable forwarding of source-routed "
1666 "datagrams"),
1667 sysctl_net_inet_ip_forwsrcrt, 0, &ip_forwsrcrt, 0,
1668 CTL_NET, PF_INET, IPPROTO_IP,
1669 IPCTL_FORWSRCRT, CTL_EOL);
1670 sysctl_createv(clog, 0, NULL, NULL,
1671 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1672 CTLTYPE_INT, "directed-broadcast",
1673 SYSCTL_DESCR("Enable forwarding of broadcast datagrams"),
1674 NULL, 0, &ip_directedbcast, 0,
1675 CTL_NET, PF_INET, IPPROTO_IP,
1676 IPCTL_DIRECTEDBCAST, CTL_EOL);
1677 sysctl_createv(clog, 0, NULL, NULL,
1678 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1679 CTLTYPE_INT, "allowsrcrt",
1680 SYSCTL_DESCR("Accept source-routed datagrams"),
1681 NULL, 0, &ip_allowsrcrt, 0,
1682 CTL_NET, PF_INET, IPPROTO_IP,
1683 IPCTL_ALLOWSRCRT, CTL_EOL);
1684
1685 sysctl_createv(clog, 0, NULL, NULL,
1686 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1687 CTLTYPE_INT, "mtudisc",
1688 SYSCTL_DESCR("Use RFC1191 Path MTU Discovery"),
1689 NULL, 0, &ip_mtudisc, 0,
1690 CTL_NET, PF_INET, IPPROTO_IP,
1691 IPCTL_MTUDISC, CTL_EOL);
1692 sysctl_createv(clog, 0, NULL, NULL,
1693 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1694 CTLTYPE_INT, "anonportmin",
1695 SYSCTL_DESCR("Lowest ephemeral port number to assign"),
1696 sysctl_net_inet_ip_ports, 0, &anonportmin, 0,
1697 CTL_NET, PF_INET, IPPROTO_IP,
1698 IPCTL_ANONPORTMIN, CTL_EOL);
1699 sysctl_createv(clog, 0, NULL, NULL,
1700 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1701 CTLTYPE_INT, "anonportmax",
1702 SYSCTL_DESCR("Highest ephemeral port number to assign"),
1703 sysctl_net_inet_ip_ports, 0, &anonportmax, 0,
1704 CTL_NET, PF_INET, IPPROTO_IP,
1705 IPCTL_ANONPORTMAX, CTL_EOL);
1706 sysctl_createv(clog, 0, NULL, NULL,
1707 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1708 CTLTYPE_INT, "mtudisctimeout",
1709 SYSCTL_DESCR("Lifetime of a Path MTU Discovered route"),
1710 sysctl_net_inet_ip_pmtudto, 0, (void *)&ip_mtudisc_timeout, 0,
1711 CTL_NET, PF_INET, IPPROTO_IP,
1712 IPCTL_MTUDISCTIMEOUT, CTL_EOL);
1713#ifndef IPNOPRIVPORTS
1714 sysctl_createv(clog, 0, NULL, NULL,
1715 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1716 CTLTYPE_INT, "lowportmin",
1717 SYSCTL_DESCR("Lowest privileged ephemeral port number "
1718 "to assign"),
1719 sysctl_net_inet_ip_ports, 0, &lowportmin, 0,
1720 CTL_NET, PF_INET, IPPROTO_IP,
1721 IPCTL_LOWPORTMIN, CTL_EOL);
1722 sysctl_createv(clog, 0, NULL, NULL,
1723 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1724 CTLTYPE_INT, "lowportmax",
1725 SYSCTL_DESCR("Highest privileged ephemeral port number "
1726 "to assign"),
1727 sysctl_net_inet_ip_ports, 0, &lowportmax, 0,
1728 CTL_NET, PF_INET, IPPROTO_IP,
1729 IPCTL_LOWPORTMAX, CTL_EOL);
1730#endif /* IPNOPRIVPORTS */
1731#if NGRE > 0
1732 sysctl_createv(clog, 0, NULL, NULL,
1733 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1734 CTLTYPE_INT, "grettl",
1735 SYSCTL_DESCR("Default TTL for a gre tunnel datagram"),
1736 NULL, 0, &ip_gre_ttl, 0,
1737 CTL_NET, PF_INET, IPPROTO_IP,
1738 IPCTL_GRE_TTL, CTL_EOL);
1739#endif /* NGRE */
1740 sysctl_createv(clog, 0, NULL, NULL,
1741 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1742 CTLTYPE_INT, "checkinterface",
1743 SYSCTL_DESCR("Enable receive side of Strong ES model "
1744 "from RFC1122"),
1745 NULL, 0, &ip_checkinterface, 0,
1746 CTL_NET, PF_INET, IPPROTO_IP,
1747 IPCTL_CHECKINTERFACE, CTL_EOL);
1748 sysctl_createv(clog, 0, NULL, NULL,
1749 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1750 CTLTYPE_INT, "random_id",
1751 SYSCTL_DESCR("Assign random ip_id values"),
1752 NULL, 0, &ip_do_randomid, 0,
1753 CTL_NET, PF_INET, IPPROTO_IP,
1754 IPCTL_RANDOMID, CTL_EOL);
1755 sysctl_createv(clog, 0, NULL, NULL,
1756 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1757 CTLTYPE_INT, "do_loopback_cksum",
1758 SYSCTL_DESCR("Perform IP checksum on loopback"),
1759 NULL, 0, &ip_do_loopback_cksum, 0,
1760 CTL_NET, PF_INET, IPPROTO_IP,
1761 IPCTL_LOOPBACKCKSUM, CTL_EOL);
1762 sysctl_createv(clog, 0, NULL, NULL,
1763 CTLFLAG_PERMANENT,
1764 CTLTYPE_STRUCT, "stats",
1765 SYSCTL_DESCR("IP statistics"),
1766 sysctl_net_inet_ip_stats, 0, NULL, 0,
1767 CTL_NET, PF_INET, IPPROTO_IP, IPCTL_STATS,
1768 CTL_EOL);
1769#if NARP
1770 sysctl_createv(clog, 0, NULL, NULL,
1771 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1772 CTLTYPE_INT, "dad_count",
1773 SYSCTL_DESCR("Number of Duplicate Address Detection "
1774 "probes to send"),
1775 NULL, 0, &ip_dad_count, 0,
1776 CTL_NET, PF_INET, IPPROTO_IP,
1777 IPCTL_DAD_COUNT, CTL_EOL);
1778#endif
1779
1780 /* anonportalgo RFC6056 subtree */
1781 const struct sysctlnode *portalgo_node;
1782 sysctl_createv(clog, 0, NULL, &portalgo_node,
1783 CTLFLAG_PERMANENT,
1784 CTLTYPE_NODE, "anonportalgo",
1785 SYSCTL_DESCR("Anonymous Port Algorithm Selection (RFC 6056)"),
1786 NULL, 0, NULL, 0,
1787 CTL_NET, PF_INET, IPPROTO_IP, CTL_CREATE, CTL_EOL);
1788 sysctl_createv(clog, 0, &portalgo_node, NULL,
1789 CTLFLAG_PERMANENT,
1790 CTLTYPE_STRING, "available",
1791 SYSCTL_DESCR("available algorithms"),
1792 sysctl_portalgo_available, 0, NULL, PORTALGO_MAXLEN,
1793 CTL_CREATE, CTL_EOL);
1794 sysctl_createv(clog, 0, &portalgo_node, NULL,
1795 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1796 CTLTYPE_STRING, "selected",
1797 SYSCTL_DESCR("selected algorithm"),
1798 sysctl_portalgo_selected4, 0, NULL, PORTALGO_MAXLEN,
1799 CTL_CREATE, CTL_EOL);
1800 sysctl_createv(clog, 0, &portalgo_node, NULL,
1801 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1802 CTLTYPE_STRUCT, "reserve",
1803 SYSCTL_DESCR("bitmap of reserved ports"),
1804 sysctl_portalgo_reserve4, 0, NULL, 0,
1805 CTL_CREATE, CTL_EOL);
1806}
1807
1808void
1809ip_statinc(u_int stat)
1810{
1811
1812 KASSERT(stat < IP_NSTATS);
1813 IP_STATINC(stat);
1814}
1815