ip_input.c revision 1.340 1 /* $NetBSD: ip_input.c,v 1.340 2016/08/31 09:14:47 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 1.340 2016/08/31 09:14:47 ozaki-r Exp $");
95
96 #ifdef _KERNEL_OPT
97 #include "opt_inet.h"
98 #include "opt_compat_netbsd.h"
99 #include "opt_gateway.h"
100 #include "opt_ipsec.h"
101 #include "opt_mrouting.h"
102 #include "opt_mbuftrace.h"
103 #include "opt_inet_csum.h"
104 #endif
105
106 #include "arp.h"
107
108 #include <sys/param.h>
109 #include <sys/systm.h>
110 #include <sys/cpu.h>
111 #include <sys/mbuf.h>
112 #include <sys/domain.h>
113 #include <sys/protosw.h>
114 #include <sys/socket.h>
115 #include <sys/socketvar.h>
116 #include <sys/errno.h>
117 #include <sys/time.h>
118 #include <sys/kernel.h>
119 #include <sys/pool.h>
120 #include <sys/sysctl.h>
121 #include <sys/kauth.h>
122
123 #include <net/if.h>
124 #include <net/if_dl.h>
125 #include <net/route.h>
126 #include <net/pktqueue.h>
127 #include <net/pfil.h>
128
129 #include <netinet/in.h>
130 #include <netinet/in_systm.h>
131 #include <netinet/ip.h>
132 #include <netinet/in_pcb.h>
133 #include <netinet/in_proto.h>
134 #include <netinet/in_var.h>
135 #include <netinet/ip_var.h>
136 #include <netinet/ip_private.h>
137 #include <netinet/ip_icmp.h>
138 /* just for gif_ttl */
139 #include <netinet/in_gif.h>
140 #include "gif.h"
141 #include <net/if_gre.h>
142 #include "gre.h"
143
144 #ifdef MROUTING
145 #include <netinet/ip_mroute.h>
146 #endif
147 #include <netinet/portalgo.h>
148
149 #ifdef IPSEC
150 #include <netipsec/ipsec.h>
151 #endif
152
153 #ifndef IPFORWARDING
154 #ifdef GATEWAY
155 #define IPFORWARDING 1 /* forward IP packets not for us */
156 #else /* GATEWAY */
157 #define IPFORWARDING 0 /* don't forward IP packets not for us */
158 #endif /* GATEWAY */
159 #endif /* IPFORWARDING */
160 #ifndef IPSENDREDIRECTS
161 #define IPSENDREDIRECTS 1
162 #endif
163 #ifndef IPFORWSRCRT
164 #define IPFORWSRCRT 1 /* forward source-routed packets */
165 #endif
166 #ifndef IPALLOWSRCRT
167 #define IPALLOWSRCRT 1 /* allow source-routed packets */
168 #endif
169 #ifndef IPMTUDISC
170 #define IPMTUDISC 1
171 #endif
172 #ifndef IPMTUDISCTIMEOUT
173 #define IPMTUDISCTIMEOUT (10 * 60) /* as per RFC 1191 */
174 #endif
175
176 #ifdef COMPAT_50
177 #include <compat/sys/time.h>
178 #include <compat/sys/socket.h>
179 #endif
180
181 /*
182 * Note: DIRECTED_BROADCAST is handled this way so that previous
183 * configuration using this option will Just Work.
184 */
185 #ifndef IPDIRECTEDBCAST
186 #ifdef DIRECTED_BROADCAST
187 #define IPDIRECTEDBCAST 1
188 #else
189 #define IPDIRECTEDBCAST 0
190 #endif /* DIRECTED_BROADCAST */
191 #endif /* IPDIRECTEDBCAST */
192 int ipforwarding = IPFORWARDING;
193 int ipsendredirects = IPSENDREDIRECTS;
194 int ip_defttl = IPDEFTTL;
195 int ip_forwsrcrt = IPFORWSRCRT;
196 int ip_directedbcast = IPDIRECTEDBCAST;
197 int ip_allowsrcrt = IPALLOWSRCRT;
198 int ip_mtudisc = IPMTUDISC;
199 int ip_mtudisc_timeout = IPMTUDISCTIMEOUT;
200 #ifdef DIAGNOSTIC
201 int ipprintfs = 0;
202 #endif
203
204 int ip_do_randomid = 0;
205
206 /*
207 * XXX - Setting ip_checkinterface mostly implements the receive side of
208 * the Strong ES model described in RFC 1122, but since the routing table
209 * and transmit implementation do not implement the Strong ES model,
210 * setting this to 1 results in an odd hybrid.
211 *
212 * XXX - ip_checkinterface currently must be disabled if you use ipnat
213 * to translate the destination address to another local interface.
214 *
215 * XXX - ip_checkinterface must be disabled if you add IP aliases
216 * to the loopback interface instead of the interface where the
217 * packets for those addresses are received.
218 */
219 static int ip_checkinterface __read_mostly = 0;
220
221 struct rttimer_queue *ip_mtudisc_timeout_q = NULL;
222
223 pktqueue_t * ip_pktq __read_mostly;
224 pfil_head_t * inet_pfil_hook __read_mostly;
225 ipid_state_t * ip_ids __read_mostly;
226 percpu_t * ipstat_percpu __read_mostly;
227
228 static percpu_t *ipforward_rt_percpu __cacheline_aligned;
229
230 uint16_t ip_id;
231
232 #ifdef INET_CSUM_COUNTERS
233 #include <sys/device.h>
234
235 struct evcnt ip_hwcsum_bad = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
236 NULL, "inet", "hwcsum bad");
237 struct evcnt ip_hwcsum_ok = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
238 NULL, "inet", "hwcsum ok");
239 struct evcnt ip_swcsum = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
240 NULL, "inet", "swcsum");
241
242 #define INET_CSUM_COUNTER_INCR(ev) (ev)->ev_count++
243
244 EVCNT_ATTACH_STATIC(ip_hwcsum_bad);
245 EVCNT_ATTACH_STATIC(ip_hwcsum_ok);
246 EVCNT_ATTACH_STATIC(ip_swcsum);
247
248 #else
249
250 #define INET_CSUM_COUNTER_INCR(ev) /* nothing */
251
252 #endif /* INET_CSUM_COUNTERS */
253
254 /*
255 * We need to save the IP options in case a protocol wants to respond
256 * to an incoming packet over the same route if the packet got here
257 * using IP source routing. This allows connection establishment and
258 * maintenance when the remote end is on a network that is not known
259 * to us.
260 */
261
262 static int ip_nhops = 0;
263
264 static struct ip_srcrt {
265 struct in_addr dst; /* final destination */
266 char nop; /* one NOP to align */
267 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */
268 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)];
269 } ip_srcrt;
270
271 static int ip_drainwanted;
272
273 struct sockaddr_in ipaddr = {
274 .sin_len = sizeof(ipaddr),
275 .sin_family = AF_INET,
276 };
277
278 static void save_rte(u_char *, struct in_addr);
279
280 #ifdef MBUFTRACE
281 struct mowner ip_rx_mowner = MOWNER_INIT("internet", "rx");
282 struct mowner ip_tx_mowner = MOWNER_INIT("internet", "tx");
283 #endif
284
285 static void ipintr(void *);
286 static void ip_input(struct mbuf *);
287 static void ip_forward(struct mbuf *, int, struct ifnet *);
288 static bool ip_dooptions(struct mbuf *);
289 static struct in_ifaddr *ip_rtaddr(struct in_addr);
290 static void sysctl_net_inet_ip_setup(struct sysctllog **);
291
292 static struct in_ifaddr *ip_match_our_address(struct ifnet *, struct ip *,
293 int *);
294 static struct in_ifaddr *ip_match_our_address_broadcast(struct ifnet *,
295 struct ip *);
296
297 /* XXX: Not yet enabled. */
298 #define SOFTNET_LOCK() KASSERT(mutex_owned(softnet_lock))
299 #define SOFTNET_UNLOCK() KASSERT(mutex_owned(softnet_lock))
300
301 /*
302 * IP initialization: fill in IP protocol switch table.
303 * All protocols not implemented in kernel go to raw IP protocol handler.
304 */
305 void
306 ip_init(void)
307 {
308 const struct protosw *pr;
309
310 in_init();
311 sysctl_net_inet_ip_setup(NULL);
312
313 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
314 KASSERT(pr != NULL);
315
316 ip_pktq = pktq_create(IFQ_MAXLEN, ipintr, NULL);
317 KASSERT(ip_pktq != NULL);
318
319 for (u_int i = 0; i < IPPROTO_MAX; i++) {
320 ip_protox[i] = pr - inetsw;
321 }
322 for (pr = inetdomain.dom_protosw;
323 pr < inetdomain.dom_protoswNPROTOSW; pr++)
324 if (pr->pr_domain->dom_family == PF_INET &&
325 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
326 ip_protox[pr->pr_protocol] = pr - inetsw;
327
328 ip_reass_init();
329
330 ip_ids = ip_id_init();
331 ip_id = time_uptime & 0xfffff;
332
333 ip_mtudisc_timeout_q = rt_timer_queue_create(ip_mtudisc_timeout);
334 #ifdef GATEWAY
335 ipflow_init();
336 #endif
337
338 /* Register our Packet Filter hook. */
339 inet_pfil_hook = pfil_head_create(PFIL_TYPE_AF, (void *)AF_INET);
340 KASSERT(inet_pfil_hook != NULL);
341
342 #ifdef MBUFTRACE
343 MOWNER_ATTACH(&ip_tx_mowner);
344 MOWNER_ATTACH(&ip_rx_mowner);
345 #endif /* MBUFTRACE */
346
347 ipstat_percpu = percpu_alloc(sizeof(uint64_t) * IP_NSTATS);
348
349 ipforward_rt_percpu = percpu_alloc(sizeof(struct route));
350 if (ipforward_rt_percpu == NULL)
351 panic("failed to allocate ipforward_rt_percpu");
352 }
353
354 static struct in_ifaddr *
355 ip_match_our_address(struct ifnet *ifp, struct ip *ip, int *downmatch)
356 {
357 struct in_ifaddr *ia = NULL;
358 int checkif;
359
360 /*
361 * Enable a consistency check between the destination address
362 * and the arrival interface for a unicast packet (the RFC 1122
363 * strong ES model) if IP forwarding is disabled and the packet
364 * is not locally generated.
365 *
366 * XXX - Checking also should be disabled if the destination
367 * address is ipnat'ed to a different interface.
368 *
369 * XXX - Checking is incompatible with IP aliases added
370 * to the loopback interface instead of the interface where
371 * the packets are received.
372 *
373 * XXX - We need to add a per ifaddr flag for this so that
374 * we get finer grain control.
375 */
376 checkif = ip_checkinterface && (ipforwarding == 0) &&
377 (ifp->if_flags & IFF_LOOPBACK) == 0;
378
379 IN_ADDRHASH_READER_FOREACH(ia, ip->ip_dst.s_addr) {
380 if (in_hosteq(ia->ia_addr.sin_addr, ip->ip_dst)) {
381 if (ia->ia4_flags & IN_IFF_NOTREADY)
382 continue;
383 if (checkif && ia->ia_ifp != ifp)
384 continue;
385 if ((ia->ia_ifp->if_flags & IFF_UP) != 0)
386 break;
387 else
388 (*downmatch)++;
389 }
390 }
391
392 return ia;
393 }
394
395 static struct in_ifaddr *
396 ip_match_our_address_broadcast(struct ifnet *ifp, struct ip *ip)
397 {
398 struct in_ifaddr *ia = NULL;
399 struct ifaddr *ifa;
400
401 IFADDR_READER_FOREACH(ifa, ifp) {
402 if (ifa->ifa_addr->sa_family != AF_INET)
403 continue;
404 ia = ifatoia(ifa);
405 if (ia->ia4_flags & IN_IFF_NOTREADY)
406 continue;
407 if (in_hosteq(ip->ip_dst, ia->ia_broadaddr.sin_addr) ||
408 in_hosteq(ip->ip_dst, ia->ia_netbroadcast) ||
409 /*
410 * Look for all-0's host part (old broadcast addr),
411 * either for subnet or net.
412 */
413 ip->ip_dst.s_addr == ia->ia_subnet ||
414 ip->ip_dst.s_addr == ia->ia_net)
415 goto matched;
416 /*
417 * An interface with IP address zero accepts
418 * all packets that arrive on that interface.
419 */
420 if (in_nullhost(ia->ia_addr.sin_addr))
421 goto matched;
422 }
423 ia = NULL;
424
425 matched:
426 return ia;
427 }
428
429 /*
430 * IP software interrupt routine.
431 */
432 static void
433 ipintr(void *arg __unused)
434 {
435 struct mbuf *m;
436
437 KASSERT(cpu_softintr_p());
438
439 mutex_enter(softnet_lock);
440 while ((m = pktq_dequeue(ip_pktq)) != NULL) {
441 ip_input(m);
442 }
443 mutex_exit(softnet_lock);
444 }
445
446 /*
447 * IP input routine. Checksum and byte swap header. If fragmented
448 * try to reassemble. Process options. Pass to next level.
449 */
450 static void
451 ip_input(struct mbuf *m)
452 {
453 struct ip *ip = NULL;
454 struct in_ifaddr *ia = NULL;
455 int hlen = 0, len;
456 int downmatch;
457 int srcrt = 0;
458 ifnet_t *ifp;
459 struct psref psref;
460 int s;
461
462 KASSERTMSG(cpu_softintr_p(), "ip_input: not in the software "
463 "interrupt handler; synchronization assumptions violated");
464
465 MCLAIM(m, &ip_rx_mowner);
466 KASSERT((m->m_flags & M_PKTHDR) != 0);
467
468 ifp = m_get_rcvif_psref(m, &psref);
469 if (__predict_false(ifp == NULL))
470 goto out;
471
472 /*
473 * If no IP addresses have been set yet but the interfaces
474 * are receiving, can't do anything with incoming packets yet.
475 * Note: we pre-check without locks held.
476 */
477 if (IN_ADDRLIST_READER_EMPTY())
478 goto out;
479 IP_STATINC(IP_STAT_TOTAL);
480
481 /*
482 * If the IP header is not aligned, slurp it up into a new
483 * mbuf with space for link headers, in the event we forward
484 * it. Otherwise, if it is aligned, make sure the entire
485 * base IP header is in the first mbuf of the chain.
486 */
487 if (IP_HDR_ALIGNED_P(mtod(m, void *)) == 0) {
488 if ((m = m_copyup(m, sizeof(struct ip),
489 (max_linkhdr + 3) & ~3)) == NULL) {
490 /* XXXJRT new stat, please */
491 IP_STATINC(IP_STAT_TOOSMALL);
492 goto out;
493 }
494 } else if (__predict_false(m->m_len < sizeof (struct ip))) {
495 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
496 IP_STATINC(IP_STAT_TOOSMALL);
497 goto out;
498 }
499 }
500 ip = mtod(m, struct ip *);
501 if (ip->ip_v != IPVERSION) {
502 IP_STATINC(IP_STAT_BADVERS);
503 goto out;
504 }
505 hlen = ip->ip_hl << 2;
506 if (hlen < sizeof(struct ip)) { /* minimum header length */
507 IP_STATINC(IP_STAT_BADHLEN);
508 goto out;
509 }
510 if (hlen > m->m_len) {
511 if ((m = m_pullup(m, hlen)) == NULL) {
512 IP_STATINC(IP_STAT_BADHLEN);
513 goto out;
514 }
515 ip = mtod(m, struct ip *);
516 }
517
518 /*
519 * RFC1122: packets with a multicast source address are
520 * not allowed.
521 */
522 if (IN_MULTICAST(ip->ip_src.s_addr)) {
523 IP_STATINC(IP_STAT_BADADDR);
524 goto out;
525 }
526
527 /* 127/8 must not appear on wire - RFC1122 */
528 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
529 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
530 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
531 IP_STATINC(IP_STAT_BADADDR);
532 goto out;
533 }
534 }
535
536 switch (m->m_pkthdr.csum_flags &
537 ((ifp->if_csum_flags_rx & M_CSUM_IPv4) |
538 M_CSUM_IPv4_BAD)) {
539 case M_CSUM_IPv4|M_CSUM_IPv4_BAD:
540 INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad);
541 IP_STATINC(IP_STAT_BADSUM);
542 goto out;
543
544 case M_CSUM_IPv4:
545 /* Checksum was okay. */
546 INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok);
547 break;
548
549 default:
550 /*
551 * Must compute it ourselves. Maybe skip checksum on
552 * loopback interfaces.
553 */
554 if (__predict_true(!(ifp->if_flags & IFF_LOOPBACK) ||
555 ip_do_loopback_cksum)) {
556 INET_CSUM_COUNTER_INCR(&ip_swcsum);
557 if (in_cksum(m, hlen) != 0) {
558 IP_STATINC(IP_STAT_BADSUM);
559 goto out;
560 }
561 }
562 break;
563 }
564
565 /* Retrieve the packet length. */
566 len = ntohs(ip->ip_len);
567
568 /*
569 * Check for additional length bogosity
570 */
571 if (len < hlen) {
572 IP_STATINC(IP_STAT_BADLEN);
573 goto out;
574 }
575
576 /*
577 * Check that the amount of data in the buffers
578 * is as at least much as the IP header would have us expect.
579 * Trim mbufs if longer than we expect.
580 * Drop packet if shorter than we expect.
581 */
582 if (m->m_pkthdr.len < len) {
583 IP_STATINC(IP_STAT_TOOSHORT);
584 goto out;
585 }
586 if (m->m_pkthdr.len > len) {
587 if (m->m_len == m->m_pkthdr.len) {
588 m->m_len = len;
589 m->m_pkthdr.len = len;
590 } else
591 m_adj(m, len - m->m_pkthdr.len);
592 }
593
594 /*
595 * Assume that we can create a fast-forward IP flow entry
596 * based on this packet.
597 */
598 m->m_flags |= M_CANFASTFWD;
599
600 /*
601 * Run through list of hooks for input packets. If there are any
602 * filters which require that additional packets in the flow are
603 * not fast-forwarded, they must clear the M_CANFASTFWD flag.
604 * Note that filters must _never_ set this flag, as another filter
605 * in the list may have previously cleared it.
606 */
607 #if defined(IPSEC)
608 if (!ipsec_used || !ipsec_indone(m))
609 #else
610 if (1)
611 #endif
612 {
613 struct in_addr odst = ip->ip_dst;
614 bool freed;
615
616 SOFTNET_LOCK();
617 freed = pfil_run_hooks(inet_pfil_hook, &m, ifp, PFIL_IN) != 0;
618 SOFTNET_UNLOCK();
619 if (freed || m == NULL) {
620 goto out;
621 }
622 ip = mtod(m, struct ip *);
623 hlen = ip->ip_hl << 2;
624
625 /*
626 * XXX The setting of "srcrt" here is to prevent ip_forward()
627 * from generating ICMP redirects for packets that have
628 * been redirected by a hook back out on to the same LAN that
629 * they came from and is not an indication that the packet
630 * is being inffluenced by source routing options. This
631 * allows things like
632 * "rdr tlp0 0/0 port 80 -> 1.1.1.200 3128 tcp"
633 * where tlp0 is both on the 1.1.1.0/24 network and is the
634 * default route for hosts on 1.1.1.0/24. Of course this
635 * also requires a "map tlp0 ..." to complete the story.
636 * One might argue whether or not this kind of network config.
637 * should be supported in this manner...
638 */
639 srcrt = (odst.s_addr != ip->ip_dst.s_addr);
640 }
641
642 #ifdef ALTQ
643 /* XXX Temporary until ALTQ is changed to use a pfil hook */
644 if (altq_input) {
645 SOFTNET_LOCK();
646 if ((*altq_input)(m, AF_INET) == 0) {
647 /* Packet dropped by traffic conditioner. */
648 SOFTNET_UNLOCK();
649 goto out;
650 }
651 SOFTNET_UNLOCK();
652 }
653 #endif
654
655 /*
656 * Process options and, if not destined for us,
657 * ship it on. ip_dooptions returns 1 when an
658 * error was detected (causing an icmp message
659 * to be sent and the original packet to be freed).
660 */
661 ip_nhops = 0; /* for source routed packets */
662 if (hlen > sizeof (struct ip) && ip_dooptions(m))
663 goto out;
664
665 /*
666 * Check our list of addresses, to see if the packet is for us.
667 *
668 * Traditional 4.4BSD did not consult IFF_UP at all.
669 * The behavior here is to treat addresses on !IFF_UP interface
670 * or IN_IFF_NOTREADY addresses as not mine.
671 */
672 downmatch = 0;
673 s = pserialize_read_enter();
674 ia = ip_match_our_address(ifp, ip, &downmatch);
675 if (ia != NULL) {
676 pserialize_read_exit(s);
677 goto ours;
678 }
679
680 if (ifp->if_flags & IFF_BROADCAST) {
681 ia = ip_match_our_address_broadcast(ifp, ip);
682 if (ia != NULL) {
683 pserialize_read_exit(s);
684 goto ours;
685 }
686 }
687 pserialize_read_exit(s);
688
689 if (IN_MULTICAST(ip->ip_dst.s_addr)) {
690 #ifdef MROUTING
691 extern struct socket *ip_mrouter;
692
693 if (ip_mrouter) {
694 /*
695 * If we are acting as a multicast router, all
696 * incoming multicast packets are passed to the
697 * kernel-level multicast forwarding function.
698 * The packet is returned (relatively) intact; if
699 * ip_mforward() returns a non-zero value, the packet
700 * must be discarded, else it may be accepted below.
701 *
702 * (The IP ident field is put in the same byte order
703 * as expected when ip_mforward() is called from
704 * ip_output().)
705 */
706 SOFTNET_LOCK();
707 if (ip_mforward(m, ifp) != 0) {
708 SOFTNET_UNLOCK();
709 IP_STATINC(IP_STAT_CANTFORWARD);
710 goto out;
711 }
712 SOFTNET_UNLOCK();
713
714 /*
715 * The process-level routing demon needs to receive
716 * all multicast IGMP packets, whether or not this
717 * host belongs to their destination groups.
718 */
719 if (ip->ip_p == IPPROTO_IGMP) {
720 goto ours;
721 }
722 IP_STATINC(IP_STAT_CANTFORWARD);
723 }
724 #endif
725 /*
726 * See if we belong to the destination multicast group on the
727 * arrival interface.
728 */
729 if (!in_multi_group(ip->ip_dst, ifp, 0)) {
730 IP_STATINC(IP_STAT_CANTFORWARD);
731 goto out;
732 }
733 goto ours;
734 }
735 if (ip->ip_dst.s_addr == INADDR_BROADCAST ||
736 in_nullhost(ip->ip_dst))
737 goto ours;
738
739 /*
740 * Not for us; forward if possible and desirable.
741 */
742 if (ipforwarding == 0) {
743 m_put_rcvif_psref(ifp, &psref);
744 IP_STATINC(IP_STAT_CANTFORWARD);
745 m_freem(m);
746 } else {
747 /*
748 * If ip_dst matched any of my address on !IFF_UP interface,
749 * and there's no IFF_UP interface that matches ip_dst,
750 * send icmp unreach. Forwarding it will result in in-kernel
751 * forwarding loop till TTL goes to 0.
752 */
753 if (downmatch) {
754 m_put_rcvif_psref(ifp, &psref);
755 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
756 IP_STATINC(IP_STAT_CANTFORWARD);
757 return;
758 }
759 #ifdef IPSEC
760 /* Perform IPsec, if any. */
761 if (ipsec_used) {
762 SOFTNET_LOCK();
763 if (ipsec4_input(m, IP_FORWARDING |
764 (ip_directedbcast ? IP_ALLOWBROADCAST : 0)) != 0) {
765 SOFTNET_UNLOCK();
766 goto out;
767 }
768 SOFTNET_UNLOCK();
769 }
770 #endif
771 ip_forward(m, srcrt, ifp);
772 m_put_rcvif_psref(ifp, &psref);
773 }
774 return;
775
776 ours:
777 m_put_rcvif_psref(ifp, &psref);
778 ifp = NULL;
779
780 /*
781 * If offset or IP_MF are set, must reassemble.
782 */
783 if (ip->ip_off & ~htons(IP_DF|IP_RF)) {
784 /*
785 * Pass to IP reassembly mechanism.
786 */
787 if (ip_reass_packet(&m, ip) != 0) {
788 /* Failed; invalid fragment(s) or packet. */
789 goto out;
790 }
791 if (m == NULL) {
792 /* More fragments should come; silently return. */
793 goto out;
794 }
795 /*
796 * Reassembly is done, we have the final packet.
797 * Updated cached data in local variable(s).
798 */
799 ip = mtod(m, struct ip *);
800 hlen = ip->ip_hl << 2;
801 }
802
803 #ifdef IPSEC
804 /*
805 * Enforce IPsec policy checking if we are seeing last header.
806 * Note that we do not visit this with protocols with PCB layer
807 * code - like UDP/TCP/raw IP.
808 */
809 if (ipsec_used &&
810 (inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) {
811 SOFTNET_LOCK();
812 if (ipsec4_input(m, 0) != 0) {
813 SOFTNET_UNLOCK();
814 goto out;
815 }
816 SOFTNET_UNLOCK();
817 }
818 #endif
819
820 /*
821 * Switch out to protocol's input routine.
822 */
823 #if IFA_STATS
824 if (ia && ip) {
825 struct in_ifaddr *_ia;
826 /*
827 * Keep a reference from ip_match_our_address with psref
828 * is expensive, so explore ia here again.
829 */
830 s = pserialize_read_enter();
831 _ia = in_get_ia(ip->ip_dst.s_addr);
832 _ia->ia_ifa.ifa_data.ifad_inbytes += ntohs(ip->ip_len);
833 pserialize_read_exit(s);
834 }
835 #endif
836 IP_STATINC(IP_STAT_DELIVERED);
837
838 const int off = hlen, nh = ip->ip_p;
839
840 SOFTNET_LOCK();
841 (*inetsw[ip_protox[nh]].pr_input)(m, off, nh);
842 SOFTNET_UNLOCK();
843 return;
844
845 out:
846 m_put_rcvif_psref(ifp, &psref);
847 if (m != NULL)
848 m_freem(m);
849 }
850
851 /*
852 * IP timer processing.
853 */
854 void
855 ip_slowtimo(void)
856 {
857
858 mutex_enter(softnet_lock);
859 KERNEL_LOCK(1, NULL);
860
861 ip_reass_slowtimo();
862
863 KERNEL_UNLOCK_ONE(NULL);
864 mutex_exit(softnet_lock);
865 }
866
867 /*
868 * IP drain processing.
869 */
870 void
871 ip_drain(void)
872 {
873
874 KERNEL_LOCK(1, NULL);
875 ip_reass_drain();
876 KERNEL_UNLOCK_ONE(NULL);
877 }
878
879 /*
880 * ip_dooptions: perform option processing on a datagram, possibly discarding
881 * it if bad options are encountered, or forwarding it if source-routed.
882 *
883 * => Returns true if packet has been forwarded/freed.
884 * => Returns false if the packet should be processed further.
885 */
886 static bool
887 ip_dooptions(struct mbuf *m)
888 {
889 struct ip *ip = mtod(m, struct ip *);
890 u_char *cp, *cp0;
891 struct ip_timestamp *ipt;
892 struct in_ifaddr *ia;
893 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0;
894 struct in_addr dst;
895 n_time ntime;
896 struct ifaddr *ifa;
897 int s;
898
899 dst = ip->ip_dst;
900 cp = (u_char *)(ip + 1);
901 cnt = (ip->ip_hl << 2) - sizeof (struct ip);
902 for (; cnt > 0; cnt -= optlen, cp += optlen) {
903 opt = cp[IPOPT_OPTVAL];
904 if (opt == IPOPT_EOL)
905 break;
906 if (opt == IPOPT_NOP)
907 optlen = 1;
908 else {
909 if (cnt < IPOPT_OLEN + sizeof(*cp)) {
910 code = &cp[IPOPT_OLEN] - (u_char *)ip;
911 goto bad;
912 }
913 optlen = cp[IPOPT_OLEN];
914 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
915 code = &cp[IPOPT_OLEN] - (u_char *)ip;
916 goto bad;
917 }
918 }
919 switch (opt) {
920
921 default:
922 break;
923
924 /*
925 * Source routing with record.
926 * Find interface with current destination address.
927 * If none on this machine then drop if strictly routed,
928 * or do nothing if loosely routed.
929 * Record interface address and bring up next address
930 * component. If strictly routed make sure next
931 * address is on directly accessible net.
932 */
933 case IPOPT_LSRR:
934 case IPOPT_SSRR:
935 if (ip_allowsrcrt == 0) {
936 type = ICMP_UNREACH;
937 code = ICMP_UNREACH_NET_PROHIB;
938 goto bad;
939 }
940 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
941 code = &cp[IPOPT_OLEN] - (u_char *)ip;
942 goto bad;
943 }
944 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
945 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
946 goto bad;
947 }
948 ipaddr.sin_addr = ip->ip_dst;
949
950 s = pserialize_read_enter();
951 ifa = ifa_ifwithaddr(sintosa(&ipaddr));
952 if (ifa == NULL) {
953 pserialize_read_exit(s);
954 if (opt == IPOPT_SSRR) {
955 type = ICMP_UNREACH;
956 code = ICMP_UNREACH_SRCFAIL;
957 goto bad;
958 }
959 /*
960 * Loose routing, and not at next destination
961 * yet; nothing to do except forward.
962 */
963 break;
964 }
965 pserialize_read_exit(s);
966
967 off--; /* 0 origin */
968 if ((off + sizeof(struct in_addr)) > optlen) {
969 /*
970 * End of source route. Should be for us.
971 */
972 save_rte(cp, ip->ip_src);
973 break;
974 }
975 /*
976 * locate outgoing interface
977 */
978 memcpy((void *)&ipaddr.sin_addr, (void *)(cp + off),
979 sizeof(ipaddr.sin_addr));
980 s = pserialize_read_enter();
981 if (opt == IPOPT_SSRR) {
982 ifa = ifa_ifwithladdr(sintosa(&ipaddr));
983 if (ifa != NULL)
984 ia = ifatoia(ifa);
985 else
986 ia = NULL;
987 } else {
988 ia = ip_rtaddr(ipaddr.sin_addr);
989 }
990 if (ia == NULL) {
991 type = ICMP_UNREACH;
992 code = ICMP_UNREACH_SRCFAIL;
993 pserialize_read_exit(s);
994 goto bad;
995 }
996 ip->ip_dst = ipaddr.sin_addr;
997 bcopy((void *)&ia->ia_addr.sin_addr,
998 (void *)(cp + off), sizeof(struct in_addr));
999 pserialize_read_exit(s);
1000 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1001 /*
1002 * Let ip_intr's mcast routing check handle mcast pkts
1003 */
1004 forward = !IN_MULTICAST(ip->ip_dst.s_addr);
1005 break;
1006
1007 case IPOPT_RR:
1008 if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1009 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1010 goto bad;
1011 }
1012 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1013 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1014 goto bad;
1015 }
1016 /*
1017 * If no space remains, ignore.
1018 */
1019 off--; /* 0 origin */
1020 if ((off + sizeof(struct in_addr)) > optlen)
1021 break;
1022 memcpy((void *)&ipaddr.sin_addr, (void *)(&ip->ip_dst),
1023 sizeof(ipaddr.sin_addr));
1024 /*
1025 * locate outgoing interface; if we're the destination,
1026 * use the incoming interface (should be same).
1027 */
1028 s = pserialize_read_enter();
1029 ifa = ifa_ifwithaddr(sintosa(&ipaddr));
1030 if (ifa == NULL) {
1031 ia = ip_rtaddr(ipaddr.sin_addr);
1032 if (ia == NULL) {
1033 pserialize_read_exit(s);
1034 type = ICMP_UNREACH;
1035 code = ICMP_UNREACH_HOST;
1036 goto bad;
1037 }
1038 } else {
1039 ia = ifatoia(ifa);
1040 }
1041 bcopy((void *)&ia->ia_addr.sin_addr,
1042 (void *)(cp + off), sizeof(struct in_addr));
1043 pserialize_read_exit(s);
1044 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1045 break;
1046
1047 case IPOPT_TS:
1048 code = cp - (u_char *)ip;
1049 ipt = (struct ip_timestamp *)cp;
1050 if (ipt->ipt_len < 4 || ipt->ipt_len > 40) {
1051 code = (u_char *)&ipt->ipt_len - (u_char *)ip;
1052 goto bad;
1053 }
1054 if (ipt->ipt_ptr < 5) {
1055 code = (u_char *)&ipt->ipt_ptr - (u_char *)ip;
1056 goto bad;
1057 }
1058 if (ipt->ipt_ptr > ipt->ipt_len - sizeof (int32_t)) {
1059 if (++ipt->ipt_oflw == 0) {
1060 code = (u_char *)&ipt->ipt_ptr -
1061 (u_char *)ip;
1062 goto bad;
1063 }
1064 break;
1065 }
1066 cp0 = (cp + ipt->ipt_ptr - 1);
1067 switch (ipt->ipt_flg) {
1068
1069 case IPOPT_TS_TSONLY:
1070 break;
1071
1072 case IPOPT_TS_TSANDADDR: {
1073 struct ifnet *rcvif;
1074 int _s, _ss;
1075
1076 if (ipt->ipt_ptr - 1 + sizeof(n_time) +
1077 sizeof(struct in_addr) > ipt->ipt_len) {
1078 code = (u_char *)&ipt->ipt_ptr -
1079 (u_char *)ip;
1080 goto bad;
1081 }
1082 ipaddr.sin_addr = dst;
1083 _ss = pserialize_read_enter();
1084 rcvif = m_get_rcvif(m, &_s);
1085 ifa = ifaof_ifpforaddr(sintosa(&ipaddr), rcvif);
1086 m_put_rcvif(rcvif, &_s);
1087 if (ifa == NULL) {
1088 pserialize_read_exit(_ss);
1089 break;
1090 }
1091 ia = ifatoia(ifa);
1092 bcopy(&ia->ia_addr.sin_addr,
1093 cp0, sizeof(struct in_addr));
1094 pserialize_read_exit(_ss);
1095 ipt->ipt_ptr += sizeof(struct in_addr);
1096 break;
1097 }
1098
1099 case IPOPT_TS_PRESPEC:
1100 if (ipt->ipt_ptr - 1 + sizeof(n_time) +
1101 sizeof(struct in_addr) > ipt->ipt_len) {
1102 code = (u_char *)&ipt->ipt_ptr -
1103 (u_char *)ip;
1104 goto bad;
1105 }
1106 memcpy(&ipaddr.sin_addr, cp0,
1107 sizeof(struct in_addr));
1108 s = pserialize_read_enter();
1109 ifa = ifa_ifwithaddr(sintosa(&ipaddr));
1110 if (ifa == NULL) {
1111 pserialize_read_exit(s);
1112 continue;
1113 }
1114 pserialize_read_exit(s);
1115 ipt->ipt_ptr += sizeof(struct in_addr);
1116 break;
1117
1118 default:
1119 /* XXX can't take &ipt->ipt_flg */
1120 code = (u_char *)&ipt->ipt_ptr -
1121 (u_char *)ip + 1;
1122 goto bad;
1123 }
1124 ntime = iptime();
1125 cp0 = (u_char *) &ntime; /* XXX grumble, GCC... */
1126 memmove((char *)cp + ipt->ipt_ptr - 1, cp0,
1127 sizeof(n_time));
1128 ipt->ipt_ptr += sizeof(n_time);
1129 }
1130 }
1131 if (forward) {
1132 struct ifnet *rcvif;
1133 struct psref _psref;
1134
1135 if (ip_forwsrcrt == 0) {
1136 type = ICMP_UNREACH;
1137 code = ICMP_UNREACH_SRCFAIL;
1138 goto bad;
1139 }
1140
1141 rcvif = m_get_rcvif_psref(m, &_psref);
1142 if (__predict_false(rcvif == NULL)) {
1143 type = ICMP_UNREACH;
1144 code = ICMP_UNREACH_HOST;
1145 goto bad;
1146 }
1147 ip_forward(m, 1, rcvif);
1148 m_put_rcvif_psref(rcvif, &_psref);
1149 return true;
1150 }
1151 return false;
1152 bad:
1153 icmp_error(m, type, code, 0, 0);
1154 IP_STATINC(IP_STAT_BADOPTIONS);
1155 return true;
1156 }
1157
1158 /*
1159 * ip_rtaddr: given address of next destination (final or next hop),
1160 * return internet address info of interface to be used to get there.
1161 */
1162 static struct in_ifaddr *
1163 ip_rtaddr(struct in_addr dst)
1164 {
1165 struct rtentry *rt;
1166 union {
1167 struct sockaddr dst;
1168 struct sockaddr_in dst4;
1169 } u;
1170 struct route *ro;
1171
1172 sockaddr_in_init(&u.dst4, &dst, 0);
1173
1174 SOFTNET_LOCK();
1175 ro = percpu_getref(ipforward_rt_percpu);
1176 rt = rtcache_lookup(ro, &u.dst);
1177 percpu_putref(ipforward_rt_percpu);
1178 SOFTNET_UNLOCK();
1179 if (rt == NULL)
1180 return NULL;
1181
1182 return ifatoia(rt->rt_ifa);
1183 }
1184
1185 /*
1186 * save_rte: save incoming source route for use in replies, to be picked
1187 * up later by ip_srcroute if the receiver is interested.
1188 */
1189 static void
1190 save_rte(u_char *option, struct in_addr dst)
1191 {
1192 unsigned olen;
1193
1194 olen = option[IPOPT_OLEN];
1195 if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst)))
1196 return;
1197 memcpy((void *)ip_srcrt.srcopt, (void *)option, olen);
1198 ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1199 ip_srcrt.dst = dst;
1200 }
1201
1202 /*
1203 * Retrieve incoming source route for use in replies,
1204 * in the same form used by setsockopt.
1205 * The first hop is placed before the options, will be removed later.
1206 */
1207 struct mbuf *
1208 ip_srcroute(void)
1209 {
1210 struct in_addr *p, *q;
1211 struct mbuf *m;
1212
1213 if (ip_nhops == 0)
1214 return NULL;
1215 m = m_get(M_DONTWAIT, MT_SOOPTS);
1216 if (m == 0)
1217 return NULL;
1218
1219 MCLAIM(m, &inetdomain.dom_mowner);
1220 #define OPTSIZ (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt))
1221
1222 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
1223 m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) +
1224 OPTSIZ;
1225
1226 /*
1227 * First save first hop for return route
1228 */
1229 p = &ip_srcrt.route[ip_nhops - 1];
1230 *(mtod(m, struct in_addr *)) = *p--;
1231
1232 /*
1233 * Copy option fields and padding (nop) to mbuf.
1234 */
1235 ip_srcrt.nop = IPOPT_NOP;
1236 ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
1237 memmove(mtod(m, char *) + sizeof(struct in_addr), &ip_srcrt.nop,
1238 OPTSIZ);
1239 q = (struct in_addr *)(mtod(m, char *) +
1240 sizeof(struct in_addr) + OPTSIZ);
1241 #undef OPTSIZ
1242 /*
1243 * Record return path as an IP source route,
1244 * reversing the path (pointers are now aligned).
1245 */
1246 while (p >= ip_srcrt.route) {
1247 *q++ = *p--;
1248 }
1249 /*
1250 * Last hop goes to final destination.
1251 */
1252 *q = ip_srcrt.dst;
1253 return (m);
1254 }
1255
1256 const int inetctlerrmap[PRC_NCMDS] = {
1257 [PRC_MSGSIZE] = EMSGSIZE,
1258 [PRC_HOSTDEAD] = EHOSTDOWN,
1259 [PRC_HOSTUNREACH] = EHOSTUNREACH,
1260 [PRC_UNREACH_NET] = EHOSTUNREACH,
1261 [PRC_UNREACH_HOST] = EHOSTUNREACH,
1262 [PRC_UNREACH_PROTOCOL] = ECONNREFUSED,
1263 [PRC_UNREACH_PORT] = ECONNREFUSED,
1264 [PRC_UNREACH_SRCFAIL] = EHOSTUNREACH,
1265 [PRC_PARAMPROB] = ENOPROTOOPT,
1266 };
1267
1268 void
1269 ip_fasttimo(void)
1270 {
1271 if (ip_drainwanted) {
1272 ip_drain();
1273 ip_drainwanted = 0;
1274 }
1275 }
1276
1277 void
1278 ip_drainstub(void)
1279 {
1280 ip_drainwanted = 1;
1281 }
1282
1283 /*
1284 * Forward a packet. If some error occurs return the sender
1285 * an icmp packet. Note we can't always generate a meaningful
1286 * icmp message because icmp doesn't have a large enough repertoire
1287 * of codes and types.
1288 *
1289 * If not forwarding, just drop the packet. This could be confusing
1290 * if ipforwarding was zero but some routing protocol was advancing
1291 * us as a gateway to somewhere. However, we must let the routing
1292 * protocol deal with that.
1293 *
1294 * The srcrt parameter indicates whether the packet is being forwarded
1295 * via a source route.
1296 */
1297 static void
1298 ip_forward(struct mbuf *m, int srcrt, struct ifnet *rcvif)
1299 {
1300 struct ip *ip = mtod(m, struct ip *);
1301 struct rtentry *rt;
1302 int error, type = 0, code = 0, destmtu = 0;
1303 struct mbuf *mcopy;
1304 n_long dest;
1305 union {
1306 struct sockaddr dst;
1307 struct sockaddr_in dst4;
1308 } u;
1309 uint64_t *ips;
1310 struct route *ro;
1311
1312 KASSERTMSG(cpu_softintr_p(), "ip_forward: not in the software "
1313 "interrupt handler; synchronization assumptions violated");
1314
1315 /*
1316 * We are now in the output path.
1317 */
1318 MCLAIM(m, &ip_tx_mowner);
1319
1320 /*
1321 * Clear any in-bound checksum flags for this packet.
1322 */
1323 m->m_pkthdr.csum_flags = 0;
1324
1325 dest = 0;
1326 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
1327 IP_STATINC(IP_STAT_CANTFORWARD);
1328 m_freem(m);
1329 return;
1330 }
1331
1332 SOFTNET_LOCK();
1333
1334 if (ip->ip_ttl <= IPTTLDEC) {
1335 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0);
1336 SOFTNET_UNLOCK();
1337 return;
1338 }
1339
1340 sockaddr_in_init(&u.dst4, &ip->ip_dst, 0);
1341
1342 ro = percpu_getref(ipforward_rt_percpu);
1343 if ((rt = rtcache_lookup(ro, &u.dst)) == NULL) {
1344 percpu_putref(ipforward_rt_percpu);
1345 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_NET, dest, 0);
1346 SOFTNET_UNLOCK();
1347 return;
1348 }
1349
1350 /*
1351 * Save at most 68 bytes of the packet in case
1352 * we need to generate an ICMP message to the src.
1353 * Pullup to avoid sharing mbuf cluster between m and mcopy.
1354 */
1355 mcopy = m_copym(m, 0, imin(ntohs(ip->ip_len), 68), M_DONTWAIT);
1356 if (mcopy)
1357 mcopy = m_pullup(mcopy, ip->ip_hl << 2);
1358
1359 ip->ip_ttl -= IPTTLDEC;
1360
1361 /*
1362 * If forwarding packet using same interface that it came in on,
1363 * perhaps should send a redirect to sender to shortcut a hop.
1364 * Only send redirect if source is sending directly to us,
1365 * and if packet was not source routed (or has any options).
1366 * Also, don't send redirect if forwarding using a default route
1367 * or a route modified by a redirect.
1368 */
1369 if (rt->rt_ifp == rcvif &&
1370 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1371 !in_nullhost(satocsin(rt_getkey(rt))->sin_addr) &&
1372 ipsendredirects && !srcrt) {
1373 if (rt->rt_ifa &&
1374 (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) ==
1375 ifatoia(rt->rt_ifa)->ia_subnet) {
1376 if (rt->rt_flags & RTF_GATEWAY)
1377 dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
1378 else
1379 dest = ip->ip_dst.s_addr;
1380 /*
1381 * Router requirements says to only send host
1382 * redirects.
1383 */
1384 type = ICMP_REDIRECT;
1385 code = ICMP_REDIRECT_HOST;
1386 }
1387 }
1388
1389 error = ip_output(m, NULL, ro,
1390 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)),
1391 NULL, NULL);
1392
1393 if (error) {
1394 IP_STATINC(IP_STAT_CANTFORWARD);
1395 goto error;
1396 }
1397
1398 ips = IP_STAT_GETREF();
1399 ips[IP_STAT_FORWARD]++;
1400
1401 if (type) {
1402 ips[IP_STAT_REDIRECTSENT]++;
1403 IP_STAT_PUTREF();
1404 goto redirect;
1405 }
1406
1407 IP_STAT_PUTREF();
1408 if (mcopy) {
1409 #ifdef GATEWAY
1410 if (mcopy->m_flags & M_CANFASTFWD)
1411 ipflow_create(ro, mcopy);
1412 #endif
1413 m_freem(mcopy);
1414 }
1415
1416 percpu_putref(ipforward_rt_percpu);
1417 SOFTNET_UNLOCK();
1418 return;
1419
1420 redirect:
1421 error:
1422 if (mcopy == NULL) {
1423 percpu_putref(ipforward_rt_percpu);
1424 SOFTNET_UNLOCK();
1425 return;
1426 }
1427
1428 switch (error) {
1429
1430 case 0: /* forwarded, but need redirect */
1431 /* type, code set above */
1432 break;
1433
1434 case ENETUNREACH: /* shouldn't happen, checked above */
1435 case EHOSTUNREACH:
1436 case ENETDOWN:
1437 case EHOSTDOWN:
1438 default:
1439 type = ICMP_UNREACH;
1440 code = ICMP_UNREACH_HOST;
1441 break;
1442
1443 case EMSGSIZE:
1444 type = ICMP_UNREACH;
1445 code = ICMP_UNREACH_NEEDFRAG;
1446
1447 if ((rt = rtcache_validate(ro)) != NULL)
1448 destmtu = rt->rt_ifp->if_mtu;
1449 #ifdef IPSEC
1450 if (ipsec_used)
1451 (void)ipsec4_forward(mcopy, &destmtu);
1452 #endif
1453 IP_STATINC(IP_STAT_CANTFRAG);
1454 break;
1455
1456 case ENOBUFS:
1457 /*
1458 * Do not generate ICMP_SOURCEQUENCH as required in RFC 1812,
1459 * Requirements for IP Version 4 Routers. Source quench can
1460 * big problem under DoS attacks or if the underlying
1461 * interface is rate-limited.
1462 */
1463 if (mcopy)
1464 m_freem(mcopy);
1465 percpu_putref(ipforward_rt_percpu);
1466 SOFTNET_UNLOCK();
1467 return;
1468 }
1469 icmp_error(mcopy, type, code, dest, destmtu);
1470 percpu_putref(ipforward_rt_percpu);
1471 SOFTNET_UNLOCK();
1472 }
1473
1474 void
1475 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1476 struct mbuf *m)
1477 {
1478 struct socket *so = inp->inp_socket;
1479 ifnet_t *ifp;
1480 int inpflags = inp->inp_flags;
1481 struct psref psref;
1482
1483 ifp = m_get_rcvif_psref(m, &psref);
1484 if (__predict_false(ifp == NULL))
1485 return; /* XXX should report error? */
1486
1487 if (so->so_options & SO_TIMESTAMP
1488 #ifdef SO_OTIMESTAMP
1489 || so->so_options & SO_OTIMESTAMP
1490 #endif
1491 ) {
1492 struct timeval tv;
1493
1494 microtime(&tv);
1495 #ifdef SO_OTIMESTAMP
1496 if (so->so_options & SO_OTIMESTAMP) {
1497 struct timeval50 tv50;
1498 timeval_to_timeval50(&tv, &tv50);
1499 *mp = sbcreatecontrol((void *) &tv50, sizeof(tv50),
1500 SCM_OTIMESTAMP, SOL_SOCKET);
1501 } else
1502 #endif
1503 *mp = sbcreatecontrol((void *) &tv, sizeof(tv),
1504 SCM_TIMESTAMP, SOL_SOCKET);
1505 if (*mp)
1506 mp = &(*mp)->m_next;
1507 }
1508 if (inpflags & INP_RECVDSTADDR) {
1509 *mp = sbcreatecontrol((void *) &ip->ip_dst,
1510 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
1511 if (*mp)
1512 mp = &(*mp)->m_next;
1513 }
1514 if (inpflags & INP_RECVPKTINFO) {
1515 struct in_pktinfo ipi;
1516 ipi.ipi_addr = ip->ip_src;
1517 ipi.ipi_ifindex = ifp->if_index;
1518 *mp = sbcreatecontrol((void *) &ipi,
1519 sizeof(ipi), IP_RECVPKTINFO, IPPROTO_IP);
1520 if (*mp)
1521 mp = &(*mp)->m_next;
1522 }
1523 if (inpflags & INP_PKTINFO) {
1524 struct in_pktinfo ipi;
1525 ipi.ipi_addr = ip->ip_dst;
1526 ipi.ipi_ifindex = ifp->if_index;
1527 *mp = sbcreatecontrol((void *) &ipi,
1528 sizeof(ipi), IP_PKTINFO, IPPROTO_IP);
1529 if (*mp)
1530 mp = &(*mp)->m_next;
1531 }
1532 if (inpflags & INP_RECVIF) {
1533 struct sockaddr_dl sdl;
1534
1535 sockaddr_dl_init(&sdl, sizeof(sdl), ifp->if_index, 0, NULL, 0,
1536 NULL, 0);
1537 *mp = sbcreatecontrol(&sdl, sdl.sdl_len, IP_RECVIF, IPPROTO_IP);
1538 if (*mp)
1539 mp = &(*mp)->m_next;
1540 }
1541 if (inpflags & INP_RECVTTL) {
1542 *mp = sbcreatecontrol((void *) &ip->ip_ttl,
1543 sizeof(uint8_t), IP_RECVTTL, IPPROTO_IP);
1544 if (*mp)
1545 mp = &(*mp)->m_next;
1546 }
1547 m_put_rcvif_psref(ifp, &psref);
1548 }
1549
1550 /*
1551 * sysctl helper routine for net.inet.ip.forwsrcrt.
1552 */
1553 static int
1554 sysctl_net_inet_ip_forwsrcrt(SYSCTLFN_ARGS)
1555 {
1556 int error, tmp;
1557 struct sysctlnode node;
1558
1559 node = *rnode;
1560 tmp = ip_forwsrcrt;
1561 node.sysctl_data = &tmp;
1562 error = sysctl_lookup(SYSCTLFN_CALL(&node));
1563 if (error || newp == NULL)
1564 return (error);
1565
1566 error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_FORWSRCRT,
1567 0, NULL, NULL, NULL);
1568 if (error)
1569 return (error);
1570
1571 ip_forwsrcrt = tmp;
1572
1573 return (0);
1574 }
1575
1576 /*
1577 * sysctl helper routine for net.inet.ip.mtudisctimeout. checks the
1578 * range of the new value and tweaks timers if it changes.
1579 */
1580 static int
1581 sysctl_net_inet_ip_pmtudto(SYSCTLFN_ARGS)
1582 {
1583 int error, tmp;
1584 struct sysctlnode node;
1585
1586 node = *rnode;
1587 tmp = ip_mtudisc_timeout;
1588 node.sysctl_data = &tmp;
1589 error = sysctl_lookup(SYSCTLFN_CALL(&node));
1590 if (error || newp == NULL)
1591 return (error);
1592 if (tmp < 0)
1593 return (EINVAL);
1594
1595 mutex_enter(softnet_lock);
1596
1597 ip_mtudisc_timeout = tmp;
1598 rt_timer_queue_change(ip_mtudisc_timeout_q, ip_mtudisc_timeout);
1599
1600 mutex_exit(softnet_lock);
1601
1602 return (0);
1603 }
1604
1605 static int
1606 sysctl_net_inet_ip_stats(SYSCTLFN_ARGS)
1607 {
1608
1609 return (NETSTAT_SYSCTL(ipstat_percpu, IP_NSTATS));
1610 }
1611
1612 static void
1613 sysctl_net_inet_ip_setup(struct sysctllog **clog)
1614 {
1615 sysctl_createv(clog, 0, NULL, NULL,
1616 CTLFLAG_PERMANENT,
1617 CTLTYPE_NODE, "inet",
1618 SYSCTL_DESCR("PF_INET related settings"),
1619 NULL, 0, NULL, 0,
1620 CTL_NET, PF_INET, CTL_EOL);
1621 sysctl_createv(clog, 0, NULL, NULL,
1622 CTLFLAG_PERMANENT,
1623 CTLTYPE_NODE, "ip",
1624 SYSCTL_DESCR("IPv4 related settings"),
1625 NULL, 0, NULL, 0,
1626 CTL_NET, PF_INET, IPPROTO_IP, CTL_EOL);
1627
1628 sysctl_createv(clog, 0, NULL, NULL,
1629 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1630 CTLTYPE_INT, "forwarding",
1631 SYSCTL_DESCR("Enable forwarding of INET datagrams"),
1632 NULL, 0, &ipforwarding, 0,
1633 CTL_NET, PF_INET, IPPROTO_IP,
1634 IPCTL_FORWARDING, CTL_EOL);
1635 sysctl_createv(clog, 0, NULL, NULL,
1636 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1637 CTLTYPE_INT, "redirect",
1638 SYSCTL_DESCR("Enable sending of ICMP redirect messages"),
1639 NULL, 0, &ipsendredirects, 0,
1640 CTL_NET, PF_INET, IPPROTO_IP,
1641 IPCTL_SENDREDIRECTS, CTL_EOL);
1642 sysctl_createv(clog, 0, NULL, NULL,
1643 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1644 CTLTYPE_INT, "ttl",
1645 SYSCTL_DESCR("Default TTL for an INET datagram"),
1646 NULL, 0, &ip_defttl, 0,
1647 CTL_NET, PF_INET, IPPROTO_IP,
1648 IPCTL_DEFTTL, CTL_EOL);
1649 #ifdef IPCTL_DEFMTU
1650 sysctl_createv(clog, 0, NULL, NULL,
1651 CTLFLAG_PERMANENT /* |CTLFLAG_READWRITE? */,
1652 CTLTYPE_INT, "mtu",
1653 SYSCTL_DESCR("Default MTA for an INET route"),
1654 NULL, 0, &ip_mtu, 0,
1655 CTL_NET, PF_INET, IPPROTO_IP,
1656 IPCTL_DEFMTU, CTL_EOL);
1657 #endif /* IPCTL_DEFMTU */
1658 sysctl_createv(clog, 0, NULL, NULL,
1659 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1660 CTLTYPE_INT, "forwsrcrt",
1661 SYSCTL_DESCR("Enable forwarding of source-routed "
1662 "datagrams"),
1663 sysctl_net_inet_ip_forwsrcrt, 0, &ip_forwsrcrt, 0,
1664 CTL_NET, PF_INET, IPPROTO_IP,
1665 IPCTL_FORWSRCRT, CTL_EOL);
1666 sysctl_createv(clog, 0, NULL, NULL,
1667 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1668 CTLTYPE_INT, "directed-broadcast",
1669 SYSCTL_DESCR("Enable forwarding of broadcast datagrams"),
1670 NULL, 0, &ip_directedbcast, 0,
1671 CTL_NET, PF_INET, IPPROTO_IP,
1672 IPCTL_DIRECTEDBCAST, CTL_EOL);
1673 sysctl_createv(clog, 0, NULL, NULL,
1674 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1675 CTLTYPE_INT, "allowsrcrt",
1676 SYSCTL_DESCR("Accept source-routed datagrams"),
1677 NULL, 0, &ip_allowsrcrt, 0,
1678 CTL_NET, PF_INET, IPPROTO_IP,
1679 IPCTL_ALLOWSRCRT, CTL_EOL);
1680
1681 sysctl_createv(clog, 0, NULL, NULL,
1682 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1683 CTLTYPE_INT, "mtudisc",
1684 SYSCTL_DESCR("Use RFC1191 Path MTU Discovery"),
1685 NULL, 0, &ip_mtudisc, 0,
1686 CTL_NET, PF_INET, IPPROTO_IP,
1687 IPCTL_MTUDISC, CTL_EOL);
1688 sysctl_createv(clog, 0, NULL, NULL,
1689 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1690 CTLTYPE_INT, "anonportmin",
1691 SYSCTL_DESCR("Lowest ephemeral port number to assign"),
1692 sysctl_net_inet_ip_ports, 0, &anonportmin, 0,
1693 CTL_NET, PF_INET, IPPROTO_IP,
1694 IPCTL_ANONPORTMIN, CTL_EOL);
1695 sysctl_createv(clog, 0, NULL, NULL,
1696 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1697 CTLTYPE_INT, "anonportmax",
1698 SYSCTL_DESCR("Highest ephemeral port number to assign"),
1699 sysctl_net_inet_ip_ports, 0, &anonportmax, 0,
1700 CTL_NET, PF_INET, IPPROTO_IP,
1701 IPCTL_ANONPORTMAX, CTL_EOL);
1702 sysctl_createv(clog, 0, NULL, NULL,
1703 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1704 CTLTYPE_INT, "mtudisctimeout",
1705 SYSCTL_DESCR("Lifetime of a Path MTU Discovered route"),
1706 sysctl_net_inet_ip_pmtudto, 0, (void *)&ip_mtudisc_timeout, 0,
1707 CTL_NET, PF_INET, IPPROTO_IP,
1708 IPCTL_MTUDISCTIMEOUT, CTL_EOL);
1709 #ifndef IPNOPRIVPORTS
1710 sysctl_createv(clog, 0, NULL, NULL,
1711 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1712 CTLTYPE_INT, "lowportmin",
1713 SYSCTL_DESCR("Lowest privileged ephemeral port number "
1714 "to assign"),
1715 sysctl_net_inet_ip_ports, 0, &lowportmin, 0,
1716 CTL_NET, PF_INET, IPPROTO_IP,
1717 IPCTL_LOWPORTMIN, CTL_EOL);
1718 sysctl_createv(clog, 0, NULL, NULL,
1719 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1720 CTLTYPE_INT, "lowportmax",
1721 SYSCTL_DESCR("Highest privileged ephemeral port number "
1722 "to assign"),
1723 sysctl_net_inet_ip_ports, 0, &lowportmax, 0,
1724 CTL_NET, PF_INET, IPPROTO_IP,
1725 IPCTL_LOWPORTMAX, CTL_EOL);
1726 #endif /* IPNOPRIVPORTS */
1727 #if NGRE > 0
1728 sysctl_createv(clog, 0, NULL, NULL,
1729 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1730 CTLTYPE_INT, "grettl",
1731 SYSCTL_DESCR("Default TTL for a gre tunnel datagram"),
1732 NULL, 0, &ip_gre_ttl, 0,
1733 CTL_NET, PF_INET, IPPROTO_IP,
1734 IPCTL_GRE_TTL, CTL_EOL);
1735 #endif /* NGRE */
1736 sysctl_createv(clog, 0, NULL, NULL,
1737 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1738 CTLTYPE_INT, "checkinterface",
1739 SYSCTL_DESCR("Enable receive side of Strong ES model "
1740 "from RFC1122"),
1741 NULL, 0, &ip_checkinterface, 0,
1742 CTL_NET, PF_INET, IPPROTO_IP,
1743 IPCTL_CHECKINTERFACE, CTL_EOL);
1744 sysctl_createv(clog, 0, NULL, NULL,
1745 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1746 CTLTYPE_INT, "random_id",
1747 SYSCTL_DESCR("Assign random ip_id values"),
1748 NULL, 0, &ip_do_randomid, 0,
1749 CTL_NET, PF_INET, IPPROTO_IP,
1750 IPCTL_RANDOMID, CTL_EOL);
1751 sysctl_createv(clog, 0, NULL, NULL,
1752 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1753 CTLTYPE_INT, "do_loopback_cksum",
1754 SYSCTL_DESCR("Perform IP checksum on loopback"),
1755 NULL, 0, &ip_do_loopback_cksum, 0,
1756 CTL_NET, PF_INET, IPPROTO_IP,
1757 IPCTL_LOOPBACKCKSUM, CTL_EOL);
1758 sysctl_createv(clog, 0, NULL, NULL,
1759 CTLFLAG_PERMANENT,
1760 CTLTYPE_STRUCT, "stats",
1761 SYSCTL_DESCR("IP statistics"),
1762 sysctl_net_inet_ip_stats, 0, NULL, 0,
1763 CTL_NET, PF_INET, IPPROTO_IP, IPCTL_STATS,
1764 CTL_EOL);
1765 #if NARP
1766 sysctl_createv(clog, 0, NULL, NULL,
1767 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1768 CTLTYPE_INT, "dad_count",
1769 SYSCTL_DESCR("Number of Duplicate Address Detection "
1770 "probes to send"),
1771 NULL, 0, &ip_dad_count, 0,
1772 CTL_NET, PF_INET, IPPROTO_IP,
1773 IPCTL_DAD_COUNT, CTL_EOL);
1774 #endif
1775
1776 /* anonportalgo RFC6056 subtree */
1777 const struct sysctlnode *portalgo_node;
1778 sysctl_createv(clog, 0, NULL, &portalgo_node,
1779 CTLFLAG_PERMANENT,
1780 CTLTYPE_NODE, "anonportalgo",
1781 SYSCTL_DESCR("Anonymous Port Algorithm Selection (RFC 6056)"),
1782 NULL, 0, NULL, 0,
1783 CTL_NET, PF_INET, IPPROTO_IP, CTL_CREATE, CTL_EOL);
1784 sysctl_createv(clog, 0, &portalgo_node, NULL,
1785 CTLFLAG_PERMANENT,
1786 CTLTYPE_STRING, "available",
1787 SYSCTL_DESCR("available algorithms"),
1788 sysctl_portalgo_available, 0, NULL, PORTALGO_MAXLEN,
1789 CTL_CREATE, CTL_EOL);
1790 sysctl_createv(clog, 0, &portalgo_node, NULL,
1791 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1792 CTLTYPE_STRING, "selected",
1793 SYSCTL_DESCR("selected algorithm"),
1794 sysctl_portalgo_selected4, 0, NULL, PORTALGO_MAXLEN,
1795 CTL_CREATE, CTL_EOL);
1796 sysctl_createv(clog, 0, &portalgo_node, NULL,
1797 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1798 CTLTYPE_STRUCT, "reserve",
1799 SYSCTL_DESCR("bitmap of reserved ports"),
1800 sysctl_portalgo_reserve4, 0, NULL, 0,
1801 CTL_CREATE, CTL_EOL);
1802 }
1803
1804 void
1805 ip_statinc(u_int stat)
1806 {
1807
1808 KASSERT(stat < IP_NSTATS);
1809 IP_STATINC(stat);
1810 }
1811